Spectroscopy of Pharmaceutical Solids PDF

DK5783_half-series-title.
qxd
2/21/06
2:35 PM
Page A
Spectroscopy of
Pharmaceutical Solids
2006 by Taylor & Francis Group, LLC
DK5783_half-series-title.qxd
2/21/06
2:35 PM
Page B
DRUGS AND THE PHARMACEUTICAL SCIENCES

A Series of Textbooks and Monographs
Executive Editor
James Swarbrick
PharmaceuTech, Inc.
Pinehurst, North Carolina
Advisory Board
Larry L. Augsburger
Harry G. Brittain
University of Maryland
Baltimore, Maryland
Center for Pharmaceutical Physics

Milford, New Jersey
Jennifer B. Dressman
Johann Wolfgang Goethe University
Frankfurt, Germany
Anthony J. Hickey
University of North Carolina School of
Pharmacy
Chapel Hill, North Carolina
Jeffrey A. Hughes
University of Florida College of
Pharmacy
Gainesville, Florida
Trevor M. Jones
The Association of the
British Pharmaceutical Industry
London, United Kingdom
Vincent H. L. Lee
Ajaz Hussain
U.S. Food and Drug Administration
Frederick, Maryland
Hans E. Junginger
Leiden/Amsterdam Center
for Drug Research
Leiden, The Netherlands
Stephen G. Schulman
University of Southern California

Los Angeles, California
University of Florida
Gainesville, Florida
Jerome P. Skelly
Elizabeth M. Topp
Alexandria, Virginia
Geoffrey T. Tucker
University of Sheffield
Royal Hallamshire Hospital
Sheffield, United Kingdom
University of Kansas School of

Pharmacy
Lawrence, Kansas
Peter York
University of Bradford School of
Pharmacy
Bradford, United Kingdom
2/21/06
2:35 PM
Page C
1. Pharmacokinetics, Milo Gibaldi and Donald Perrier

2. Good Manufacturing Practices for Pharmaceuticals: A Plan for Total
Quality Control, Sidney H. Willig, Murray M. Tuckerman,
and William S. Hitchings IV
3. Microencapsulation, edited by J. R. Nixon
4. Drug Metabolism: Chemical and Biochemical Aspects, Bernard Testa
and Peter Jenner
5. New Drugs: Discovery and Development, edited by Alan A. Rubin
6. Sustained and Controlled Release Drug Delivery Systems, edited by
Joseph R. Robinson
7. Modern Pharmaceutics, edited by Gilbert S. Banker
and Christopher T. Rhodes
8. Prescription Drugs in Short Supply: Case Histories, Michael A. Schwartz
9. Activated Charcoal: Antidotal and Other Medical Uses, David O. Cooney
10. Concepts in Drug Metabolism (in two parts), edited by Peter Jenner
and Bernard Testa
11. Pharmaceutical Analysis: Modern Methods (in two parts), edited by
James W. Munson
12. Techniques of Solubilization of Drugs, edited by Samuel H. Yalkowsky
13. Orphan Drugs, edited by Fred E. Karch
14. Novel Drug Delivery Systems: Fundamentals, Developmental Concepts,
Biomedical Assessments, Yie W. Chien
15. Pharmacokinetics: Second Edition, Revised and Expanded, Milo Gibaldi
and Donald Perrier
Quality Control, Second Edition, Revised and Expanded, Sidney H. Willig,
Murray M. Tuckerman, and William S. Hitchings IV
17. Formulation of Veterinary Dosage Forms, edited by Jack Blodinger
18. Dermatological Formulations: Percutaneous Absorption, Brian W. Barry
19. The Clinical Research Process in the Pharmaceutical Industry, edited by
Gary M. Matoren
20. Microencapsulation and Related Drug Processes, Patrick B. Deasy
21. Drugs and Nutrients: The Interactive Effects, edited by Daphne A. Roe
and T. Colin Campbell
22. Biotechnology of Industrial Antibiotics, Erick J. Vandamme
23. Pharmaceutical Process Validation, edited by Bernard T. Loftus
and Robert A. Nash
24. Anticancer and Interferon Agents: Synthesis and Properties, edited by
Raphael M. Ottenbrite and George B. Butler
25. Pharmaceutical Statistics: Practical and Clinical Applications,
Sanford Bolton
2/21/06
2:35 PM
Page D
26. Drug Dynamics for Analytical, Clinical, and Biological Chemists,

Benjamin J. Gudzinowicz, Burrows T. Younkin, Jr.,
and Michael J. Gudzinowicz
27. Modern Analysis of Antibiotics, edited by Adjoran Aszalos
28. Solubility and Related Properties, Kenneth C. James
29. Controlled Drug Delivery: Fundamentals and Applications, Second Edition,
Revised and Expanded, edited by Joseph R. Robinson and Vincent H. Lee
30. New Drug Approval Process: Clinical and Regulatory Management,
edited by Richard A. Guarino
31. Transdermal Controlled Systemic Medications, edited by Yie W. Chien
32. Drug Delivery Devices: Fundamentals and Applications, edited by
Praveen Tyle
33. Pharmacokinetics: Regulatory Industrial Academic Perspectives,
edited by Peter G. Welling and Francis L. S. Tse
34. Clinical Drug Trials and Tribulations, edited by Allen E. Cato
35. Transdermal Drug Delivery: Developmental Issues and Research
Initiatives, edited by Jonathan Hadgraft and Richard H. Guy
36. Aqueous Polymeric Coatings for Pharmaceutical Dosage Forms,
edited by James W. McGinity
37. Pharmaceutical Pelletization Technology, edited by Isaac Ghebre-Sellassie
38. Good Laboratory Practice Regulations, edited by Allen F. Hirsch
39. Nasal Systemic Drug Delivery, Yie W. Chien, Kenneth S. E. Su,
and Shyi-Feu Chang
40. Modern Pharmaceutics: Second Edition, Revised and Expanded,
edited by Gilbert S. Banker and Christopher T. Rhodes
41. Specialized Drug Delivery Systems: Manufacturing and Production
Technology, edited by Praveen Tyle
42. Topical Drug Delivery Formulations, edited by David W. Osborne
and Anton H. Amann
43. Drug Stability: Principles and Practices, Jens T. Carstensen
Second Edition, Revised and Expanded, Sanford Bolton
45. Biodegradable Polymers as Drug Delivery Systems, edited by
Mark Chasin and Robert Langer
46. Preclinical Drug Disposition: A Laboratory Handbook, Francis L. S. Tse
and James J. Jaffe
47. HPLC in the Pharmaceutical Industry, edited by Godwin W. Fong
and Stanley K. Lam
48. Pharmaceutical Bioequivalence, edited by Peter G. Welling,
Francis L. S. Tse, and Shrikant V. Dinghe
49. Pharmaceutical Dissolution Testing, Umesh V. Banakar
2/21/06
2:35 PM
Page E
50. Novel Drug Delivery Systems: Second Edition, Revised and Expanded,
Yie W. Chien
51. Managing the Clinical Drug Development Process, David M. Cocchetto
and Ronald V. Nardi
Quality Control, Third Edition, edited by Sidney H. Willig
and James R. Stoker
53. Prodrugs: Topical and Ocular Drug Delivery, edited by Kenneth B. Sloan
54. Pharmaceutical Inhalation Aerosol Technology, edited by
Anthony J. Hickey
55. Radiopharmaceuticals: Chemistry and Pharmacology, edited by
Adrian D. Nunn
56. New Drug Approval Process: Second Edition, Revised and Expanded,
57. Pharmaceutical Process Validation: Second Edition, Revised
and Expanded, edited by Ira R. Berry and Robert A. Nash
58. Ophthalmic Drug Delivery Systems, edited by Ashim K. Mitra
59. Pharmaceutical Skin Penetration Enhancement, edited by
Kenneth A. Walters and Jonathan Hadgraft
60. Colonic Drug Absorption and Metabolism, edited by Peter R. Bieck
61. Pharmaceutical Particulate Carriers: Therapeutic Applications, edited by
Alain Rolland
62. Drug Permeation Enhancement: Theory and Applications, edited by
Dean S. Hsieh
63. Glycopeptide Antibiotics, edited by Ramakrishnan Nagarajan
64. Achieving Sterility in Medical and Pharmaceutical Products, Nigel A. Halls
65. Multiparticulate Oral Drug Delivery, edited by Isaac Ghebre-Sellassie
66. Colloidal Drug Delivery Systems, edited by Jrg Kreuter
67. Pharmacokinetics: Regulatory Industrial Academic Perspectives,
Second Edition, edited by Peter G. Welling and Francis L. S. Tse
68. Drug Stability: Principles and Practices, Second Edition, Revised
and Expanded, Jens T. Carstensen
69. Good Laboratory Practice Regulations: Second Edition, Revised
and Expanded, edited by Sandy Weinberg
70. Physical Characterization of Pharmaceutical Solids, edited by
Harry G. Brittain
71. Pharmaceutical Powder Compaction Technology, edited by
Gran Alderborn and Christer Nystrm
72. Modern Pharmaceutics: Third Edition, Revised and Expanded, edited by
Gilbert S. Banker and Christopher T. Rhodes
73. Microencapsulation: Methods and Industrial Applications, edited by
Simon Benita
2/21/06
2:35 PM
Page F
74. Oral Mucosal Drug Delivery, edited by Michael J. Rathbone

75. Clinical Research in Pharmaceutical Development, edited by Barry Bleidt
and Michael Montagne
76. The Drug Development Process: Increasing Efficiency and Cost
Effectiveness, edited by Peter G. Welling, Louis Lasagna,
and Umesh V. Banakar
77. Microparticulate Systems for the Delivery of Proteins and Vaccines,
edited by Smadar Cohen and Howard Bernstein
Quality Control, Fourth Edition, Revised and Expanded, Sidney H. Willig
and James R. Stoker
79. Aqueous Polymeric Coatings for Pharmaceutical Dosage Forms:
Second Edition, Revised and Expanded, edited by James W. McGinity
Third Edition, Sanford Bolton
81. Handbook of Pharmaceutical Granulation Technology, edited by
Dilip M. Parikh
82. Biotechnology of Antibiotics: Second Edition, Revised and Expanded,
edited by William R. Strohl
83. Mechanisms of Transdermal Drug Delivery, edited by Russell O. Potts
and Richard H. Guy
84. Pharmaceutical Enzymes, edited by Albert Lauwers and Simon Scharp
85. Development of Biopharmaceutical Parenteral Dosage Forms, edited by
John A. Bontempo
86. Pharmaceutical Project Management, edited by Tony Kennedy
87. Drug Products for Clinical Trials: An International Guide to Formulation
Production Quality Control, edited by Donald C. Monkhouse
and Christopher T. Rhodes
88. Development and Formulation of Veterinary Dosage Forms:
Second Edition, Revised and Expanded, edited by Gregory E. Hardee
and J. Desmond Baggot
89. Receptor-Based Drug Design, edited by Paul Leff
90. Automation and Validation of Information in Pharmaceutical Processing,
edited by Joseph F. deSpautz
91. Dermal Absorption and Toxicity Assessment, edited by Michael S. Roberts
and Kenneth A. Walters
92. Pharmaceutical Experimental Design, Gareth A. Lewis, Didier Mathieu,
and Roger Phan-Tan-Luu
93. Preparing for FDA Pre-Approval Inspections, edited by Martin D. Hynes III
94. Pharmaceutical Excipients: Characterization by IR, Raman, and NMR
Spectroscopy, David E. Bugay and W. Paul Findlay
95. Polymorphism in Pharmaceutical Solids, edited by Harry G. Brittain
2/21/06
2:35 PM
Page G
96. Freeze-Drying/Lyophilization of Pharmaceutical and Biological Products,

edited by Louis Rey and Joan C. May
97. Percutaneous Absorption: DrugsCosmeticsMechanismsMethodology,
Third Edition, Revised and Expanded, edited by Robert L. Bronaugh
and Howard I. Maibach
98. Bioadhesive Drug Delivery Systems: Fundamentals, Novel Approaches,
and Development, edited by Edith Mathiowitz, Donald E. Chickering III,
and Claus-Michael Lehr
99. Protein Formulation and Delivery, edited by Eugene J. McNally
100. New Drug Approval Process: Third Edition, The Global Challenge,
101. Peptide and Protein Drug Analysis, edited by Ronald E. Reid
102. Transport Processes in Pharmaceutical Systems, edited by
Gordon L. Amidon, Ping I. Lee, and Elizabeth M. Topp
103. Excipient Toxicity and Safety, edited by Myra L. Weiner
and Lois A. Kotkoskie
104. The Clinical Audit in Pharmaceutical Development, edited by
Michael R. Hamrell
105. Pharmaceutical Emulsions and Suspensions, edited by Francoise Nielloud
and Gilberte Marti-Mestres
106. Oral Drug Absorption: Prediction and Assessment, edited by
Jennifer B. Dressman and Hans Lennerns
107. Drug Stability: Principles and Practices, Third Edition, Revised
and Expanded, edited by Jens T. Carstensen and C. T. Rhodes
108. Containment in the Pharmaceutical Industry, edited by James P. Wood
Quality Control from Manufacturer to Consumer, Fifth Edition, Revised
and Expanded, Sidney H. Willig
110. Advanced Pharmaceutical Solids, Jens T. Carstensen
111. Endotoxins: Pyrogens, LAL Testing, and Depyrogenation, Second Edition,
Revised and Expanded, Kevin L. Williams
112. Pharmaceutical Process Engineering, Anthony J. Hickey
and David Ganderton
113. Pharmacogenomics, edited by Werner Kalow, Urs A. Meyer
and Rachel F. Tyndale
114. Handbook of Drug Screening, edited by Ramakrishna Seethala
and Prabhavathi B. Fernandes
115. Drug Targeting Technology: Physical Chemical Biological Methods,
edited by Hans Schreier
116. DrugDrug Interactions, edited by A. David Rodrigues
117. Handbook of Pharmaceutical Analysis, edited by Lena Ohannesian
and Anthony J. Streeter
118. Pharmaceutical Process Scale-Up, edited by Michael Levin
2/21/06
2:35 PM
Page H
119. Dermatological and Transdermal Formulations, edited by

Kenneth A. Walters
120. Clinical Drug Trials and Tribulations: Second Edition, Revised
and Expanded, edited by Allen Cato, Lynda Sutton, and Allen Cato III
121. Modern Pharmaceutics: Fourth Edition, Revised and Expanded, edited by
Gilbert S. Banker and Christopher T. Rhodes
122. Surfactants and Polymers in Drug Delivery, Martin Malmsten
123. Transdermal Drug Delivery: Second Edition, Revised and Expanded,
edited by Richard H. Guy and Jonathan Hadgraft
124. Good Laboratory Practice Regulations: Second Edition, Revised
and Expanded, edited by Sandy Weinberg
125. Parenteral Quality Control: Sterility, Pyrogen, Particulate, and Package
Integrity Testing: Third Edition, Revised and Expanded, Michael J. Akers,
Daniel S. Larrimore, and Dana Morton Guazzo
126. Modified-Release Drug Delivery Technology, edited by
Michael J. Rathbone, Jonathan Hadgraft, and Michael S. Roberts
127. Simulation for Designing Clinical Trials: A PharmacokineticPharmacodynamic Modeling Perspective, edited by Hui C. Kimko
and Stephen B. Duffull
128. Affinity Capillary Electrophoresis in Pharmaceutics and Biopharmaceutics,
edited by Reinhard H. H. Neubert and Hans-Hermann Rttinger
129. Pharmaceutical Process Validation: An International Third Edition,
Revised and Expanded, edited by Robert A. Nash and Alfred H. Wachter
130. Ophthalmic Drug Delivery Systems: Second Edition, Revised
and Expanded, edited by Ashim K. Mitra
131. Pharmaceutical Gene Delivery Systems, edited by Alain Rolland
and Sean M. Sullivan
132. Biomarkers in Clinical Drug Development, edited by John C. Bloom
and Robert A. Dean
133. Pharmaceutical Extrusion Technology, edited by Isaac Ghebre-Sellassie
and Charles Martin
134. Pharmaceutical Inhalation Aerosol Technology: Second Edition,
Revised and Expanded, edited by Anthony J. Hickey
Fourth Edition, Sanford Bolton and Charles Bon
136. Compliance Handbook for Pharmaceuticals, Medical Devices,
and Biologics, edited by Carmen Medina
137. Freeze-Drying/Lyophilization of Pharmaceutical and Biological Products:
Second Edition, Revised and Expanded, edited by Louis Rey
and Joan C. May
138. Supercritical Fluid Technology for Drug Product Development, edited by
Peter York, Uday B. Kompella, and Boris Y. Shekunov
139. New Drug Approval Process: Fourth Edition, Accelerating Global
Registrations, edited by Richard A. Guarino
2/21/06
2:35 PM
Page I
140. Microbial Contamination Control in Parenteral Manufacturing, edited by

Kevin L. Williams
141. New Drug Development: Regulatory Paradigms for Clinical Pharmacology
and Biopharmaceutics, edited by Chandrahas G. Sahajwalla
142. Microbial Contamination Control in the Pharmaceutical Industry, edited by
Luis Jimenez
143. Generic Drug Product Development: Solid Oral Dosage Forms, edited by
Leon Shargel and Izzy Kanfer
144. Introduction to the Pharmaceutical Regulatory Process, edited by
Ira R. Berry
145. Drug Delivery to the Oral Cavity: Molecules to Market, edited by
Tapash K. Ghosh and William R. Pfister
146. Good Design Practices for GMP Pharmaceutical Facilities, edited by
Andrew Signore and Terry Jacobs
147. Drug Products for Clinical Trials, Second Edition, edited by Donald
Monkhouse, Charles Carney, and Jim Clark
148. Polymeric Drug Delivery Systems, edited by Glen S. Kwon
149. Injectable Dispersed Systems: Formulation, Processing, and Performance,
edited by Diane J. Burgess
150. Laboratory Auditing for Quality and Regulatory Compliance,
Donald Singer, Raluca-Ioana Stefan, and Jacobus van Staden
151. Active Pharmaceutical Ingredients: Development, Manufacturing,
and Regulation, edited by Stanley Nusim
152. Preclinical Drug Development, edited by Mark C. Rogge and David R. Taft
153. Pharmaceutical Stress Testing: Predicting Drug Degradation, edited by
Steven W. Baertschi
154. Handbook of Pharmaceutical Granulation Technology: Second Edition,
edited by Dilip M. Parikh
155. Percutaneous Absorption: DrugsCosmeticsMechanismsMethodology,
Fourth Edition, edited by Robert L. Bronaugh and Howard I. Maibach
156. Pharmacogenomics: Second Edition, edited by Werner Kalow,
Urs A. Meyer and Rachel F. Tyndale
157. Pharmaceutical Process Scale-Up, Second Edition, edited by
Michael Levin
158. Microencapsulation: Methods and Industrial Applications, Second Edition,
edited by Simon Benita
159. Nanoparticle Technology for Drug Delivery, edited by Ram B. Gupta
and Uday B. Kompella
160. Spectroscopy of Pharmaceutical Solids, edited by Harry G. Brittain
2/21/06
2:35 PM
Page i
Spectroscopy of
Pharmaceutical Solids
edited by
Harry G. Brittain
Center for Pharmaceutical Physics
Milford, New Jersey, U.S.A.
New York London
DK5783_Discl.fm Page 1 Monday, November 14, 2005 11:32 AM
Published in 2006 by
Taylor & Francis Group
270 Madison Avenue
New York, NY 10016
No claim to original U.S. Government works
Printed in the United States of America on acid-free paper
10 9 8 7 6 5 4 3 2 1
International Standard Book Number-10: 1-57444-893-5 (Hardcover)
International Standard Book Number-13: 978-1-57444-893-1 (Hardcover)
This book contains information obtained from authentic and highly regarded sources. Reprinted material is
quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts
have been made to publish reliable data and information, but the author and the publisher cannot assume
responsibility for the validity of all materials or for the consequences of their use.
No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic,
mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and
recording, or in any information storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.com
(http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC) 222 Rosewood Drive,
Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration
for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate
system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only
for identification and explanation without intent to infringe.
Library of Congress Cataloging-in-Publication Data

Catalog record is available from the Library of Congress
Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com
Taylor & Francis Group
is the Academic Division of Informa plc.
Preface
Spectroscopic techniques have been widely used in the pharmaceutical sciences

to obtain both fundamental and applied information. During the development of
any given drug candidate, spectroscopy will be used to establish the structure of
the compound and understand its interaction with other constituents. It is also
often used as a means for evaluating the analytical characteristics of the bulk
drug substance and its formulations. It is no surprise, therefore, that solid-state
spectroscopic methods have become extremely important to successful modern
drug development.
Many scientists believe themselves to be familiar with the principles that
govern the interaction of electromagnetic radiation with matter, and yet their
knowledge is often based on partial truths. For instance, most would state that
for a molecule to absorb ultraviolet light, an electron must be promoted from
one energy level to another. While in some cases there is validity to this
belief, the true origin of the transition is a change in the orbital angular momentum of the molecule, and the absorption of a quantum of light causes the
transition from one molecular state to another. Genuine knowledge as to the
origin of spectroscopic phenomena might not change the routine use of a particular technique, but it would provide a basis that could lead to a more advanced
application for that technique.
The reasoning just stated has led to the need for the present volume.
However great the use of solid-state spectroscopy might be, a greater degree of
fundamental understanding is necessary to obtain maximal use out of each technique. In the present work, the underlying principles of each technique will be
sufficiently outlined to provide a thorough and proper understanding of
the physics involved, and then applications will be used to illustrate what can
be learned through the employment of the method under discussion. Whenever
possible, the examples will be drawn from the pharmaceutical literature, but
this rule will be violated whenever the author feels that an application from
another field might inspire analogous work by a pharmaceutical scientist.
iii
iv
Preface
In 1995, I edited a volume entitled Physical Characterization of Pharmaceutical Solids in which a fairly extensive overview of methods suitable for
work at the molecular, particulate, and bulk levels was provided. Since a substantial portion of this earlier book was concerned with the use of spectroscopy
for the characterization of solids having pharmaceutical, the present volume
may be viewed as being Volume 1 in the second edition of the older book. In
the present volume, the use of spectroscopy for the characterization of pharmaceutical solids has been taken much further, and the range of topics has been
greatly extended relative to the coverage of the earlier volume.
Harry G. Brittain
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iii
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ix
Part I.
1.
Introduction to Spectroscopy
Electromagnetic Radiation and Spectroscopy . . . . . . . . . . . . . . . . 1

Harry G. Brittain
Nature of Electromagnetic Radiation . . . . 1
Classical Description of Electromagnetic Radiation . . . . 3
The Quantum Theory of Electromagnetic Radiation . . . . 8
Ionization and Excitation of Atoms . . . . 12
Development of the Quantum Theory . . . . 17
Overview of Spectroscopy . . . . 28
Summary . . . . 30
References . . . . 31
Part II.
Core Electron Spectroscopy
2.
Core Electron States and X-Ray Absorption Spectroscopy . . . . . 33

Harry G. Brittain
Wave Mechanical Description of the Hydrogen Atom . . . . 33
X-Ray Absorption Spectroscopy . . . . 53
3.
X-Ray Photoelectron and X-Ray Fluorescence Spectroscopy

Harry G. Brittain
X-Ray Emission Spectroscopies . . . . 67
v
. . . 67
vi
Contents
X-Ray Photoelectron Spectroscopy . . . . 69

X-Ray Fluorescence Spectroscopy . . . . 77
Part III.
Valence Electron Spectroscopy
4.
Molecular Orbital Theory and the Electronic Structure

of Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Harry G. Brittain
Molecular Orbital (MO) Theory of Organic Molecules . . . . 91
LCAO-MO Theory . . . . 93
LCAO-MO Theory and the Spectroscopy of
Diatomic Molecules . . . . 97
MO Theory and the Spectroscopy of Polyatomic Molecules . . . . 104
General Features of the Absorption
Spectroscopy of Organic Molecules . . . . 112
5.
UV/VIS Reflectance Spectroscopy . . . . . . . . . . . . . . . . . . . . . . 121

Harry G. Brittain
Introduction . . . . 121
Categories of Light Reflectance . . . . 122
Theory Associated with Diffuse Reflectance Measurements . . . . 124
Instrumentation for the Measurement of Diffuse Reflectance . . . . 127
Quantitative Measurement of Color . . . . 129
Applications of UV/VIS Diffuse Reflectance Spectroscopy
to the Study of Solids Having Pharmaceutical Interest . . . . 134
Use of Adsorbed Indicators for the Evaluation
of Surface Acidity and Basicity . . . . 141
Summary . . . . 147
6.
Luminescence Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

Harry G. Brittain
Photophysical Processes in Isolated Organic Molecules . . . . 152
Photophysical Processes Associated with
Interacting Molecules . . . . 157
Instrumentation for the Measurement of Luminescence
Spectra . . . . 170
Photoluminescence and the Structure of
Organic Molecules . . . . 178
Studies of Luminescence in the Crystalline State . . . . 179
Photoluminescence from Molecules Adsorbed on Surfaces . . . . 184
Contents
Part IV.
vii
Vibrational Level Spectroscopy
7.
Molecular Motion and Vibrational Spectroscopy . . . . . . . . . . . 205

Harry G. Brittain
Motion of Nuclei in Molecules . . . . 206
Rotation and Vibration in Diatomic Molecules . . . . 210
Vibrational Energy Levels and Spectroscopy of
Diatomic Molecules . . . . 213
Vibrational Energy Levels and Spectroscopy of the
Fundamental Transitions of Polyatomic Molecules . . . . 219
Spectroscopy of Transitions Associated with Overtones and
Combinations of Fundamental Vibrational Modes . . . . 227
Summary . . . . 231
8.
Infrared Absorption Spectroscopy . . . . . . . . . . . . . . . . . . . . . . 235

David E. Bugay and Harry G. Brittain
Some Fundamental Principles of IR Absorption Spectroscopy . . . . 236
Instrumentation for the Measurement of
IR Spectra in the Solid State . . . . 239
Applications of IR Spectroscopy to Areas of
Pharmaceutical Interest . . . . 246
9.
Raman Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271

David E. Bugay and Harry G. Brittain
Some Fundamental Principles of the Raman Effect . . . . 272
Instrumentation for Measurement of Raman
Spectra in the Solid State . . . . 278
Group Frequency Correlations in Raman Spectroscopy
and Effects of the Solid State . . . . 283
Applications of Raman Spectroscopy to
Areas of Pharmaceutical Interest . . . . 286
10.
Near-Infrared Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 313

Robert P. Cogdill and James K. Drennen, III
Fundamental Principles . . . . 315
Practice of NIR Spectroscopy . . . . 344
Applications in Pharmaceutical Solids Analysis . . . . 389
viii
Contents
Part V.
11.
Nuclear Spin Level Spectroscopy
Solid-State Nuclear Magnetic Resonance Spectrometry . . . . . . 413

Ales Medek
Fundamental Principles of NMR . . . . 413
SSNMR of Powder Solids . . . . 420
Pharmaceutically Relevant Applications of SSNMR . . . . 427
Contributors
Harry G. Brittain
U.S.A.
Center for Pharmaceutical Physics, Milford, New Jersey,

SSCI, Inc., West Lafayette, Indiana, U.S.A.
David E. Bugay
Robert P. Cogdill
Pennsylvania, U.S.A.
School of Pharmacy, Duquesne University, Pittsburgh,
James K. Drennen, III

School of Pharmacy, Duquesne University,
Pittsburgh, Pennsylvania, U.S.A.
Ales Medek
Pfizer Global R&D, Groton, Connecticut, U.S.A.
ix
PART I.
INTRODUCTION TO SPECTROSCOPY
1
Electromagnetic Radiation and
Spectroscopy
Harry G. Brittain
Center for Pharmaceutical Physics, Milford, New Jersey, U.S.A.
NATURE OF ELECTROMAGNETIC RADIATION

It had been known since ancient times that a moist atmosphere would split light
from the sun into a rainbow of colors. In 1672, Newton used an apparatus similar
to that shown in Figure 1 to demonstrate that the same type of splitting could be
effected using a glass prism as the active element. In this work, he showed
that the refractability of the light increased on passing from red to violet
and postulated a corpuscular theory for the nature of light. Unfortunately,
Newtons great reputation discouraged others from challenging his theory, and
this situation persisted until a new wave theory was presented to the Royal
Society by Young in 1801. Needless to say, a fierce scientific and philosophical
debate ensued. In 1815, Fresnel developed a mathematical theory to interpret the
phenomenon of interference and also explained the polarization of light by
assuming that the light vibrations, which pass through a medium, are contained
in a plane transverse to the direction of propagation. The great controversy
over the wave versus corpuscular nature of light was resolved in 1850 by
Foucault, who designed a revolving-mirror apparatus that could measure the
velocity of light through different media. He conclusively demonstrated that
light travels more slowly in water than it does in air, a finding that was required
by the wave theory but incompatible with the corpuscular theory of Newton.
In 1873, Maxwell published his, Treatise on Electricity and Magnetism,
in which he presented the most definitive statement regarding the classical
1
Brittain
Figure 1 Spectroscopic apparatus of the type used by Newton to observe the splitting of
sunlight into visible colors.
theory of electromagnetic radiation. In this theory, electromagnetic radiation is

described entirely by a wave theory that associated oscillating electric and magnetic fields with the radiation. Ordinary radiation traveling along the z-axis can be
treated in terms of electric (E) and magnetic (H) fields that are mutually perpendicular to each other and to the direction of propagation. Polarized light is then
considered to be the electromagnetic radiation whose electric and magnetic
vectors are constrained to vibrate in a single plane parallel to the z-axis. These
relationships are illustrated in Figure 2.
An electric charge has an electric field associated with it, which radiates
outward in a uniform manner from its center. If this charge moves with constant
velocity or is at rest, then the spherical symmetry of the field is preserved. If the
charge accelerates, the field will tend to lag behind, because it will take time to
accommodate itself to the changing velocity of the charge. This change causes a
disturbance in the direction of the field, which radiates out as a spherical front
into the surrounding space. This disturbance, which is propagated with a definite
velocity, c, constitutes electromagnetic radiation. It has been found that this velocity is constant for all electromagnetic radiation in a vacuum, having a value of
c 2.998 108 m/sec.
By convention, electromagnetic radiation is taken to propagate along the
z-axis, and the oscillation of the electric and magnetic fields lies in the xz- and
yz-planes. The associated electric field strength vector varies sinusoidally in
phase with the oscillating dipole, and the energy spreads outward as a continuous
train of electromagnetic radiation, polarized along the direction of oscillation.
Figure 2
Illustration of the wave nature of electromagnetic radiation.
Electromagnetic Radiation and Spectroscopy
The value of the electric field strength vector, E, along the z-axis at a given time
and position on the z-axis is given by:
E Eo cos 2p n(t z=c)
(1)
where Eo is a vector that defines the amplitude of the electric field, and n is the
frequency of the electromagnetic radiation.
One often finds it convenient to characterize electromagnetic radiation by
the distance between equivalent points along the traveling wave, a parameter
known as the wavelength, l. In any homogenous medium of refractive index,
n, the velocity of propagation is now given by c/n. Consequently the frequency
and the wavelength are interrelated by:
c nnl
(2)
A unit that has found extensive use for the description of certain types of
electromagnetic radiation is the wavenumber, which is defined as the number
, is obtained
of waves contained in a path length of 1 cm. The wavenumber, w
as the reciprocal of the wavelength, so that:
w~ 1=l n=c
(3)
As mentioned before, there is a magnetic field lying in a plane perpendicular to the electric vector, associated with the oscillating electric field, and oscillating in phase with it. The amplitude of the two fields are not independent and
are related by:
Eo =Ho (m=1)1=2
(4)
where Ho is a vector that defines the amplitude of the magnetic field, m is the magnetic permeability, and 1 is the dielectric constant of the medium in which the
radiation is propagating.
The spectrum of electromagnetic radiation is divided into various regions
for convenience, and definitions for these are given in Table 1. Most spectroscopists delineate a particular region by its wavelength interval, and it is clear from
the table that the preferred SI unit of the meter is only convenient for radiowaves.
In the ultraviolet and visible regions, one finds it more convenient instead to
specify wavelengths in units of nanometers.
Micrometers are more appropriate for the infrared (IR) region, and centimeters best describe wavelengths in the microwave portion of the spectrum.
As will be developed in succeeding chapters, each spectral region is characterized
by electromagnetic radiation having a given energy, and the magnitude of this
energy determines the nature of its interaction with atoms and molecules.
CLASSICAL DESCRIPTION OF ELECTROMAGNETIC
RADIATION
In the classical world, a molecule had to contain either a permanent or transient
dipole moment in order to interact with electromagnetic radiation. A bond
Table 1
Spectral Regions of Electromagnetic Radiation
Electromagnetic
radiation type
g-ray
X-Ray
Ultraviolet
Far
Near
Visible
Infrared
Near
Mid
Far
Microwave
Radiowave
Wavelength (m)
Wavenumber (cm21)
Frequency (Hz)
Less than 1.0 10210

1.0 10210 to 1.0 1028
Greater than 100,000,000

1,000,000 to 100,000,000
Greater than 3.0 1018

3.0 1016 to 3.0 1018
1.0 1028 to 2.0 1027

(10 to 200 nm)
2.0 1027 to 4.0 1027
(200 to 400 nm)
4.0 1027 to 7.5 1027
(400 to 750 nm)
50,000 to 1,000,000
1.5 1015 to 3.0 1016
25,000 to 50,000
7.5 1014 to 1.5 1015
13,350 to 25,000
4.0 1014 to 7.5 1014
7.5 1027 to 2.5 1026

(0.75 to 2.5 mm)
2.5 1026 to 2.5 1025
(2.5 to 25 mm)
2.5 1025 to 4.0 1024
(25 to 400 mm)
4.0 1024 to 1.0 100
(0.04 to 100 cm)
Greater than 1
4,000 to 13,350
1.2 1014 to 4.0 1014
400 to 4000
1.2 1013 to 1.2 1014
25 to 400
7.5 1011 to 1.2 1013
0.01 to 25
3.0 108 to 7.5 1011
Less than 0.01
Less than 3.0 108

Brittain
formed between atoms having different electronegativity values results in a

charge separation and in the formation of a permanent electric dipole. The
total electric moment, P, of a molecule involves a summation of charge
separations over all the electronic and nuclear coordinates:
X
P
qi r i
(5)
i
where qi defines the magnitude of charges separated by a distance ri. To a first

approximation, one can assume that the nuclear coordinates are fixed and that
inner-shell electrons have a spherical charge distribution. Therefore, the major
contribution to the total electric moment arises from the electronic cloud of the
bonding and nonbonding valence electrons. The total electric dipole moment
of a polyatomic molecule is properly obtained as the resultant from the vectorial
addition of the products of the time-averaged electronic and nuclear coordinates,
and their corresponding charges, over the entire molecule. Molecules that possess
a permanent dipole moment are said to be polar.
Neutral atoms or molecules that have a symmetrical charge distribution
cannot yield a permanent dipole moment and are termed nonpolar. However,
when such species are placed in an electric field, both the electronic and
nuclear charge distributions become distorted, resulting in an induced dipole
moment. For symmetrical atoms or molecules, the magnitude of the induced
moment is proportional to the strength of the external field. The proportionality
constant in the relationship is the polarizability (a) and will consist of contributions from the atomic and electronic polarizabilities.
The classical description as to how atoms and molecules interact with electromagnetic radiation becomes a determination as to the circumstances whereby
electric dipole absorption or emission can occur. For example, during the rotation
of a polar molecule, the change in its dipole moment will be solely a consequence
of the change in orientation. If there is no molecular electric dipole moment, then
there can be no interaction with the radiation field. If radiation is to be absorbed
or emitted, the distortion associated with a molecular vibration must give rise to a
change in the magnitude of the dipole moment. Changes in magnetic dipole
moment can also give rise to absorption or emission through an interaction
with the magnetic vector in the radiation, but the observed intensity will be
many orders of magnitude lower than for the electric dipole case. Such interactions will be seen, however, to be of central importance for nuclear magnetic
resonance and electron paramagnetic resonance.
Unfortunately for the workers in the late 1800s, the Maxwell theory was
maddeningly vague as to the intimate details of the mechanisms whereby electromagnetic radiation interacted with atoms and molecules. However, the most
important property of electromagnetic radiation was that the waves conveyed
energy. This energy (w, in units of ergs/cm) was distributed in space at the rate of:
w
1E2 mH2
8p
(6)
Brittain
In this description, 1E 2 mH 2 (i.e., the amounts of electric and magnetic

energies are equal), so:
w 1E2 =4p
(7)
This energy density is carried along with the waves but is not really the parameter
of importance. What one really needs to know is the amount of energy that flows
per second across the unit area of a plane drawn perpendicular to the direction of
wave propagation. This quantity is the intensity, I. As mentioned earlier, the wave
moves through a medium with a velocity given by c/n, so integration of the motion
equation yields:
I (cw)=n
(8)
One may take advantage of the fact that w is defined by Equation (7), and
that 1 n 2, to obtain:
I cE2 =4p
(9)
Since E is defined by Equation (1), one may determine that the average intensity of
the waves is given by:
I cnE2o =8p
(10)
The classical laws of mechanical and electromagnetic theory permit a

discussion of the emission and absorption of electromagnetic radiation by a
system of electrically charged particles. According to the classical theory,
the rate of emission of radiant energy by an accelerated particle of electric
charge, q, is:
dE 2q2 a~ 2
dt
3c3
(11)
in which 2dE/dt is the rate at which the energy, E, of the particle is converted
into radiant energy, and a is the acceleration of the particle.
First, consider a special system in which a particle of charge q carries out
simple harmonic oscillation, with frequency, n, along the x-axis. The equation
describing this harmonic motion is given by:
x Xo cos (2pnt)
(12)
Differentiating Equation (12), and assuming that the amplitude factor Xo

is independent of time, we obtain an equation describing the acceleration of
this particle:
a~ 4p2 n2 X o cos (2pnt)
(13)
The average rate of emission of radiant energy by such a system is consequently:

dE 16p4 q2 n4 X2o
dt
3c3
(14)
As a result of the emission of energy, the amplitude factor Xo of the motion must
decrease with time, but if the fractional change in energy during a cycle of the
motion is small, Equation (14) retains its validity. The radiation emitted by
such a system has the frequency, n, of the emitting system. It is plane-polarized,
with the plane of the electric vector being the plane defined by the x-axis and the
direction of propagation of the light.
In case the particle carries out independent harmonic oscillations along all
three Cartesian axes, this motion would be characterized by respective frequencies of nX, nY, and nZ and amplitudes of Xo, Yo, and Zo. In that case, the total rate
of emission of radiant energy would be given as the sum of three terms similar to
the right side of Equation (14). If the motion of the particle cannot be described as
simple harmonic, it can be represented by Fourier series or as Fourier integral as a
sum or integral of harmonic terms similar to that of Equation (12). Light having a
frequency characteristic of each of these terms will then be emitted at a rate given
by Equation (14) but with the coefficient of the Fourier term being introduced in
place of Xo.
The emission of light by a system composed of several interacting electrically charged particles is handled in an analogous fashion. A Fourier analysis is
first made of the motion of the system in a given state to resolve it into harmonic
terms. For a given term, corresponding to a particular frequency of motion, n, the
coefficient resulting from the analysis (which is a function of the coordinates of
the particles) is expanded as a power series in the quantities xn/l, yn/l, and zn/l.
xn, yn, and zn are the coordinates of the particles relative to some origin, and l is
the wavelength of the electromagnetic radiation. The zeroth-degree term in this
expansion must equal zero, because the electric charge distribution of the system
does not change with time. The first-degree term of the expansion involves
(in addition to the harmonic function of time) only a function of the spatial
coordinates. The aggregate of these first-degree terms in the coordinates with
their associated time factors, summed over all frequency values occurring in the
original Fourier analysis, represents a dynamical quantity known as the electric
moment of the system. This quantity has already been defined in Equation (5).
To the degree that classical theory is applicable, the radiation emitted by a
system of several particles can be discussed by performing a Fourier analysis of
the electric moment. Corresponding to each frequency term in this representation
of P, radiation will be emitted having a frequency n. The rate of this emission is
given by an expression similar to that of Equation (14) but with the term (qXo)
being replaced by the Fourier coefficient in the electric moment expansion.
The emission of radiation by this mechanism is usually called dipole emission,
with the radiation itself sometimes being described as dipole radiation.
However, as shall be discussed in the next section, experimental developments and theoretical explanations arose around the beginning of the twentieth
century that amply demonstrated the absolute inadequacy of classical theories
to explain the interaction of electromagnetic radiation with atoms and molecules.
Hanna, in his treatise Quantum Mechanics in Chemistry, has succinctly
Brittain
summarized the three most important statements (and problems) associated with
the application of classical mechanics to molecular theory and spectroscopy.
First, there could be no limit to the accuracy with which one or more of the
dynamical variables of a classical system could be simultaneously measured,
except by the limits imposed by the precision of measuring instruments.
Second, there was no restriction on the number of variables that could be
measured simultaneously. Finally, because the expressions for velocity are continuously varying functions of time, the velocity (and hence the kinetic energy) of
a system could be continuously varied so that there were no restrictions on the
values permissible to a variable. Although the system of classical mechanics
worked well for particles as large as billiard balls, all three assumptions cannot
be applied to the mechanics that takes place at the subatomic or atomic scale.
For this reason, further discussion of the classical theory will not be pursued,
and our focus moves entirely to the use of quantum theory.
Nevertheless, the power of the electromagnetic theory of Maxwell
(especially as developed by Lorentz, Hertz, and other workers) proved to be
quite successful in explaining many phenomena and in the prediction of new
technologies. For instance, the development of radiotechnology at the beginning
of the twentieth century represented a direct application of the classical theory for
electromagnetic radiation. It also spawned some more dubious research, such as
the fruitless search for the ther, believed to be the medium by which electromagnetic radiation was propagated. This substance remained an integral part
of classical physics until its existence was finally disproved by the experiments
of Michelson and Morley.
THE QUANTUM THEORY OF ELECTROMAGNETIC
RADIATION
A number of experimental advances were made around the end of the nineteenth
century that demonstrated the inadequacy of the carefully constructed and
elegant field of physics existing at that moment. One of the most important of
these concerned the well-known emission of light by a heated solid. Physicists
had known for a long time that objects (such as horseshoes in a blacksmiths
fire) would glow when heated to high temperatures and that the color of the
emitted light differed with the temperature of the solid.
In 1884, Boltzmann derived a theoretical relationship describing the variation of the emissive power of an ideal heated solid (known as a blackbody) as a
function of the absolute temperature, T. This relationship arose from a consideration of a Carnot engine operated with radiation as the working substance, and
was finally stated as the Stefan-Boltzmann law:
W sT 4
(15)
In Equation (15), W is the total amount of energy radiated per square

centimeter of surface per second, and s is the Stefan-Boltzmann constant
(5.6694 1025erg/sec cm2). Being a consequence of the second law of thermodynamics, this relation was totally compatible with the tenets of the classical
theory. Unfortunately, although the Stefan-Boltzmann law provided a reasonably
accurate prediction of light emission at long wavelengths, it failed completely at
short wavelengths.
Rayleigh and Jeans attempted to address the problem of blackbody
radiation using classical statistical mechanics and the properties of wave
motion, and derived an alternate formula:
r(n, T ) dn
8 p n2 k T
dn
c3
(16)
where k is the Boltzmann constant, r(n,T) dn is the energy density of radiation

between frequencies n and (n dn) at absolute temperature T, and is proportional
to the intensity of light emitted by the body at this temperature. A plot of
Equation (16) for T 2000 K is shown in Figure 3, along with a spectrum of
the actual spectral energy output of a real blackbody at that same temperature.
Figure 3 Spectral energy distribution for a blackbody radiator operating at 2000 K (solid
trace) and the distribution predicted on the basis of the Rayleigh-Jeans equation for a
2000 K radiator (dashed trace).
10
Brittain
The problem with all of the classical theories advanced to explain the
wavelength dependence of blackbody radiation is that although they were
fairly successful in predicting the spectrum of energy density at wavelengths
longer than the observed maximum, they all predicted an essentially infinite
output of energy at shorter wavelengths. This clearly contradicted all experimental observations and was therefore termed the ultraviolet catastrophe.
Wien also derived an equation based on the second law of thermodynamics
and found that the monochromatic energy density in an isothermal enclosure
and the monochromatic emissive power of a blackbody when obtained at corresponding wavelengths were both directly proportional to the fifth power of the
absolute temperature. From this fact, Wien showed that the wavelength at
which the emissive power is a maximum was related to the temperature by the
simple relation:
lmax T 0:289
(17)
This equation came to be known as the Wien displacement law, and the units of
the constant on the right-hand side of Equation (17) are cm.deg. Unfortunately,
although the Wien law was successful in certain wavelength regions, it failed
completely in others.
Classical physics could not be made to yield a solution to the problem of
blackbody radiation, but a new theory advanced by Planck was able to explain
all of the experimental observations. He began with the assumption made by
Rayleigh and Jeans that the radiation on the inside of a heated body arises
from the emission of the vibrating constituents of the material. But rather than
assuming that these vibrations could emit energy of any value, he proposed
that the energy could instead only be emitted in discrete amounts called
quanta. Planck further hypothesized that these quanta were characterized by an
energy given by:
E hn
(18)
where h is the constant now known as Plancks constant (equal to 6.6256

10227 erg sec). The energy of any oscillator would therefore have to be a multiple
of hn. Planck then derived a new equation for the energy density of blackbody
radiation, finding:
r0 (n, T) dn
8phn3 hn=kT
(e
1)1 dn
c3
(19)
The validity of Plancks theory is proven in that Equation (19) was found to
yield predictions that were in complete agreement with the experiment. It can be
shown that in the limit of small wavelengths, Equation (19) has the form of Wiens
law as stated in Equation (17). In addition, for long wavelengths, Equation (19)
has the form of the Rayleigh-Jeans law of Equation (16). The most important
aspect of the new theory was that one could only obtain agreement with
11
observation if one assumed that the energy of the oscillators could not vary in a
continuous fashion. The new theory required that the oscillators could only
assume definite energies that were a function of the frequency of oscillation. In
addition, energy could not be absorbed or emitted in a continuous manner but
only in definite amounts that were equal to multiples of hn. This required the
further assumption that an oscillator had to remain in a state of constant energy
(termed a stationary state) until a quantum of energy, or its multiple, had been
absorbed or emitted.
As one could imagine, the new theory of Plank caused a storm of controversy. The wave theory of electromagnetic radiation had been thoroughly
worked out and was able to provide reasonably good explanations for many emission and absorption phenomena. An oscillating electric charge has oscillating
electric and magnetic fields, and at distant points these fields constitute the radiation field, or the light wave emitted by the charge. The field carries out energy at
a uniform and continuous rate, and the charge loses energy at the same rate, as
required by the law of conservation of energy. This situation requires that the
system gradually comes to rest, accounting for the emission process. To describe
absorption, one assumes a light wave with an alternating electric field that acts on
an electric charge that is capable of oscillation. The field exerts its force on the
charge, gradually setting it into motion with greater and greater amplitude so
that it gradually and continuously absorbs energy. Both the absorption and emission processes are therefore seen to be absolutely continuous according to the
classical wave theory. This is to be contrasted with the quantum theory of
Planck, which required that energy could only change by finite amounts equal
to hn.
Fortunately for Planck and the new theory, other scientists were conducting
investigations in various phenomena that could not be explained by the existing
classical theory. For instance, Einstein had been attempting to explain the photoelectric effect, discovered by Hertz in 1887. This effect occurs when light of a
suitable frequency illuminates the surface of an electropositive metal, which is
maintained at a negative voltage relative to a plate (i.e., acting as a cathode).
If the experiment is performed in an evacuated chamber, one observes a flow
of electric current from the cathode to the plate as long as the cathode is illuminated. In 1900, Lenard showed that the absorption of light by the metal was
accompanied by the emission of cathode rays (shown later to be a stream of electrons) from the metal. Over the course of time it became known that the magnitude of the photocurrent is directly proportional to the intensity of the incident
radiation, but paradoxically the maximum kinetic energy of the emitted electrons
was found to be independent of the incident radiant intensity. The fact that the
kinetic energy of the photoelectrons was rigorously independent of the radiant
intensity could not be explained by any application of the classical theory.
In 1905, Einstein published a theoretical explanation for the photoelectric
effect that required an assumption extremely similar to that made by Planck. He
assumed that the energy of a radiation field could not be considered as being
12
Brittain
continuously distributed through space (which would be required by classical

wave theory), but instead that it was carried by particles. These light quanta
(named photons by Lewis in 1926) were each of energy equal to hn. Einsteins
hypothesis required that the absorption or emission of light having a frequency
of n must consist of the absorption or emission of a photon having energy of
hn. The simple explanation for the photoelectric effect was that if the energy
of a photon was insufficient to remove an electron from the cathode, no
current would flow. As the frequency of the radiation was increased, the
quanta became characterized by higher and higher energies, and eventually a frequency is reached that is sufficient to eject electrons and cause the flow of current.
In addition, because the energy of the photons depends only on the frequency of
the irradiating light, its intensity cannot have any effect on the energies of the
ejected electrons. But only the absolute number of electrons emitted can be
affected by the light intensity.
The philosophical debate that accompanied these new theories revolved
around a number of conceptual difficulties. When Planck made his assertion
that the energies of an ensemble of oscillators were restricted to certain fixed
states, he imposed constraints on a hitherto continuous variable. The postulate
by Einstein required that light be transmitted by corpuscular photons, so how
did the concepts of frequency and wavelength fit into the picture? Finally, the
wave theory had completely explained a large number of phenomena, such as
interference fringes, so how could this be if radiation was corpuscular? The
only way out of this dilemma would ultimately be through the generation of a
quantum wave mechanics theory.
IONIZATION AND EXCITATION OF ATOMS

The final difficulty with wave theory proved to be its failure to explain the
structure of atoms, which led to the end of attempts to use classical mechanics
as a means to describe microscopic phenomena. As early as 1750, Melvill
noticed that when different gases were placed in a flame, light of various
colors was emitted. Most importantly, he found that when the emitted radiation
was passed through one of Newtons prisms, the light actually was found to
consist of bright spots of color separated by completely dark intervals. He
also noted that the exact colors of the emitted light were specific and characteristic of the identity of the substance excited (the term used by Melvill to
describe his observations) by the flame. As an example of typical experimental
results of this type, the flame-excited line emission spectrum of hydrogen is
shown in Figure 4.
It soon became possible to qualitatively identify elements by the presence
of characteristic lines in their emission spectra, and many new elements were
discovered for the first time through their spectral emissions. In 1885, Balmer
found that the wavelengths associated with the emission spectrum of atomic
13
Figure 4 Emission spectrum of the hydrogen atom, indicating the spectral lines associated with transitions within the Lyman, Balmer, and Paschen series. The intensity axis is
not drawn to scale.
hydrogen could be empirically fitted by a very simple equation:
n2 K
(n2 22 )
(20)
where n is one of the integers 3, 4, 5, or 6, and K is a constant (which equals

3.6456 1027 m). The Balmer equation was generalized by Rydberg in 1906
and Ritz in 1908 to yield the energy of atomic hydrogen emission lines in
units of wavenumbers:

1
1
1
w~ R 2 2
(21)
l
n1 n2
where n1 and n2 are integers. R is the factor now known as the Rydberg constant,
and has the value of 4/K, or 109,677.8 cm21. The success of the Balmer-RydbergRitz equation was in that not only did it accurately fit the energies of all known
hydrogen emission lines but that it also accurately predicted the positions of
many additional lines that were subsequently discovered. Rydberg also found
that the spectra of many other atoms could be predicted fairly well by the
general equation:
Tn
R
(n a)2
(22)
14
Brittain
where Tn represents the differences in energies between two energy states, and a
is an empirically determined constant.
The problem for physicists at this time was that although the empirical
formula were very successful in interpreting atomic emission spectra, no theoretical explanation could be developed from the existing classical theory that
satisfactorily explained the observed phenomena. The best classical theory
could only predict integral ratios of frequencies, and the results of computations
based on these theories were not even approximately correct. Furthermore, the
development of spectral equipment having high degrees of resolution indicated
that even the empirical relationships were adequate for the prediction of the
fine structure that seemed to be associated with all known spectral lines. All
theories advanced prior to 1913 were deemed to be complete failures.
Fortunately, new empirical observations were being made that led to the
development of new thinking. Around 1890, Thompson discovered the electron,
and shortly thereafter the charge and mass of this elementary particle had been
determined. Around 1910, Rutherford bombarded a piece of gold foil with an
a-particle beam and studied the angular distribution of the transmitted and
reflected particles relative to the incident trajectory. Rutherford used his observations to advance a theory in 1911 that the structure of an atom was basically
that of a small positively charged nucleus, about which negatively charged
electrons rotated in a type of planetary motion.
Rutherfords model was consistent with experimental evidence but was
also totally contradicted by the current electromagnetic theory. According to
classical theory, because an electron was subject to the attractive force of the
nucleus, it was required to undergo continuous acceleration to maintain a
stable orbit. An additional requirement of classical electrodynamics was that
this accelerated charge must constantly lose energy through radiation. As
energy was lost, the electron would have to spiral closer and closer to the
nucleus, eventually completely collapsing with a continuous emission of radiation during the process. The theory indicated that the lifetime of an atom could
not be longer than 1028 seconds, a conclusion that obviously disagreed with
reality.
In 1913, Bohr proposed a hypothesis to explain the discrepancies between
classical theory and experiment, which is now seen to be a logical extension of
the theories previously set forth by Planck and Einstein concerning radiation
and the energy of stationary states. Noting that the dimensions of Plancks
constant were those of angular momentum (energy times time), he proposed
that Maxwells laws were invalid at the subatomic level. Instead, the angular
momentum of an orbiting electron in a stable atom had to be quantized in
units of h/2p, a quantity that is ordinarily referred to as h . If the orbit of the electron was to be stable, it was further required that the centrifugal force of orbiting
electron had to be balanced by the coulombic attraction between the electron and
the nucleus. Bohr was then able to obtain a solution to the simultaneous equations
for the electron angular momentum, the balance of coulombic attraction and
15
centrifugal force, and Plancks equation, to derive a formula analogous to the

Balmer-Rydberg-Ritz equation.
From Bohrs postulates, it is possible to derive the energies of the possible
stationary states responsible for the radiation that is absorbed or emitted by an
atom consisting of a single electron and nucleus. The specification of these
states then permits one to compute the frequency of the associated electromagnetic radiation. To begin, one assumes that the charge on the nucleus to be
Z times the fundamental electronic charge, e, and that Coulombs law provides
the attractive force, F, between the nucleus and electron:
F Ze2 =r 2
(23)
where r is the radius of the electron orbit. This force must be exactly balanced by
the centrifugal force on the rotating electron:
F mr v2
(24)
where v is the angular velocity of the electron, and m is its mass. Because
Equations (23) and (24) must be equal, we find that:
mrv2 Ze2 =r 2
(25)
The angular momentum, L, of a circulating particle is defined as its linear

momentum multiplied by the radius of its circular motion. Bohrs second
postulate requires that the angular momentum of a rotating electron be an integral
multiple of h :
L nh mr2 v
(26)
where n is an integer having the values of 1, 2, 3, and so on. Solving Equations

(25) and (26) for the orbital radius yields:
r
n2 h 2
me2 Z
(27)
One now finds that the orbital radius restricted us to certain values, termed orbits,
having values equal to h 2/me 2Z, 4h 2/me 2Z, 9h 2/me 2Z, and so on. For the smallest allowed orbit of a hydrogen atom (defined by Z 1 and n 1), one finds
that:
ro ; ao h 2 =me2
(28)
).
and that ao 5.29 10211 m (0.529 A
The Bohr theory also permits a calculation of the total energy, E, of the
atom from Hamiltons equation:
E T V
(29)
16
Brittain
where the kinetic energy, T, exists by virtue of the rotation motion of the electron:
1
1
T mr 2 v2 Ze2 =r
2
2
(30)
and the potential energy, V, exists by virtue of the position of the electron relative
to the nucleus:
V Ze2 =r
(31)
Substituting Equations (30) and (31) into (29), one finds that the total energy is
given by:
E 12 Ze2 =r
(32)
and if the Bohr Equation for r (27) is substituted into Equation (32), one finds:
E
mZ 2 e4
2n2 h 2
(33)
It is therefore concluded that only certain stationary states are possible and that
the state having the lowest total energy is defined by n 1.
A transition between two states requires the absorption or emission of
energy equal to the energy difference between the states. By the postulate of
Bohr, this difference in energy must be quantized, so that:
DE E1 E2 hn

me4 1
1
2 2 2 2
2n h n1 n2
(34)
(35)
Remembering that n c/l, one obtains an equation having the form of the
Balmer-Rydberg-Ritz formula:

1 2p2 me4 1
1

w^
(36)
l
h3 c
n21 n22
with the Rydberg constant, R, being found to equal:
R
2p2 me4
h3 c
(37)
substitution of values for m, c, e, and h into Equation (37) yields the exact value
as that determined on the basis of the empirical observations (109,737 cm21).
The power of the Bohr theory was demonstrated in its ability to correlate
the series of lines observed in the emission spectrum of atomic hydrogen by
Lyman, Balmer, and Paschen, as has been illustrated in Figure 5. These correlations confirmed the postulates of Bohr that energy states were quantized and
that absorption or emission spectroscopy could only take place between these
17
Figure 5 Energy level diagram of the hydrogen atom, illustrating assignments for the
observed emission spectral lines. The family of transitions terminating in the n 1
level belong to the Lyman series, those terminating in the n 2 level belong to the
Balmer series, and transitions terminating in the n 3 level belong to the Paschen series.
quantized levels. In time, the integers defined as n became known as quantum

numbers.
Although Bohrs theory quickly became replaced by a wave mechanical
quantum theory, one of his enduring conclusions became known as the correspondence principle. This represented a recognition that the laws applicable to
the subatomic and atomic level must ultimately reduce to the classical laws
under the limiting conditions for which the latter were known to be valid.
DEVELOPMENT OF THE QUANTUM THEORY

Although the Bohr theory worked very well for one-electron atoms (H, He,
Li2, Be3, etc.), it could not handle the spectrum of any atom containing
18
Brittain
more than one electron. Furthermore, it could not provide an explanation for the
fine structure associated within a given series of emission lines. Refinements
were made to the Bohr theory by Sommerfeld to permit the existence of elliptical
orbits and to include relativistic effects, but neither of these modifications led to
the development of a satisfactory general theory. It was generally agreed among
the physicists of the time that a new type of mechanics was needed to treat atomic
and molecular behavior.
Prior to 1920, the corpuscular nature of electrons appeared to be well established on the basis of ample experimental evidence. This included the deflection
of cathode rays by electric and magnetic fields and the measurement of the
charge/mass ratio of an electron. Following this came the oil-drop experiment
of Millikan that enabled a direct determination of the charge on an electron.
Finally, it had also been observed that electrons would leave cloud tracks in
cloud chambers and that the velocities of these electrons could be affected by
the nature of the transversed medium.
The field of atomic mechanics was broken wide open by the hypothesis
made by de Broglie in 1924 that the dual wave/particulate nature of light was
extendable to matter. He reasoned that if light consisted of photons, then the
momentum of a photon would be given by its mass times its velocity, and because:
E mc2 hn
(38)
then both mc and hn/c must each equal h/l. de Broglie argued that a mass having a
momentum equal to p should also have associated with it a matter wave characterized by h/l. In other words,
l h=p h=mv
(39)
where if m and v are the respective mass and velocity of the moving particle, then l
would be the wavelength of its matter wave.
The experimental verification proving the existence of the wave character
of matter was provided shortly thereafter by Davisson and Germer, who
bombarded the surface of a nickel crystal with a narrow beam of low velocity
electrons. They evaluated the intensities of reflected electrons as a function of
their angles and found a pattern that was extremely similar to that produced by
the diffraction of X-rays off crystalline surfaces. Using a modified Bragg
equation, they obtained a value for the wavelength of the electrons that agreed
excellently with that calculated by Equation (39).
It was now evident that any valid explanation for atomic or molecular
structure would necessarily require a theory that incorporated both the wave
and corpuscular properties of matter. Erwin Schrodinger and Werner Heisenberg,
who developed theories along two different pathways, solved the problem almost
simultaneously. Schrodinger reasoned that electronic motion could be treated as
waves and used equations for wave motion and stationary waves available
from classical physics to develop a wave mechanical picture. In this approach,
19
techniques of partial differential equations were used to handle the equations for
wave motion. Heisenberg used the properties of matrices and developed a matrix
mechanics theory based on matrix algebra that was later proved to be equivalent
to the wave mechanics of Schrodinger. Dirac and von Neumann developed more
general quantum mechanical theories, and both the Schrodinger and Heisenberg
approaches were shown to be specific instances of the general theory.
It is beyond the scope of this introductory chapter to fully develop these
quantum mechanical theories in detail, and interested readers can consult any
of the references cited at the end of this chapter. However, one can look at
some of the simplest applications of the quantum theory to gain a degree of
understanding as to the defining properties of energy states in atoms and
molecules. Knowing that spectroscopy entails transitions among energy states
will provide the impetus for this analysis. Because the theory of Schrodinger is
more accessible, we shall adopt the usual practice of presenting elements of
the wave mechanical theory for our spectroscopic explanations.
In the same way that classical mechanics and thermodynamics were based
on a core of assumed postulates, quantum mechanics was formulated upon a
series of assumptions that were assumed to be correct as long as they did not
lead to the deduction of contradictory predictions. One may gain a considerable
degree of understanding into quantum mechanics by considering the fundamental
postulates that provides its basis. The first postulate defines the state of a system
consisting of N particles by a function, c (qi, t), which is a function of the coordinates of each particle and the time dependence of their motion. Furthermore,
this state function must be single-valued at all points, finite, and continuous.
The nature of this function is such that its product with its complex conjugate
c (i.e., the function obtained by replacing i by 2i, where i is the square root
of 1) is proportional to the probability of finding the particle in a region
between qi and (qi dqi) at time t.
Among the consequences of this first postulate is the understanding that the
wave function describes the system as completely as possible. However, it must
be recognized that one cannot specify the exact positions and momenta of all
particles contained within a given system and predict their future behavior in
the classical manner. Instead, as stated most explicitly by Heisenberg in his
uncertainty principle, it is possible to specify only the probability that certain
events will occur. Because c must contain both positional and time dependencies,
in some cases it is possible to separate out the time-dependent function and therefore obtain a time-invariant function termed a stationary state. Spectroscopists
are most interested in the determination and definition of these stationary
states, because spectroscopy consists of the characterization of the transitions
between such states.
The second postulate of the wave mechanical theory states that there must
exist a mathematical operator for every observable property of a system, and the
physical properties of the observable can be inferred from the mathematical
properties of its associated operator. When one operates on a state function
20
Brittain
with the operator, the exact experimental values of the observable property are
obtained as the eigenvalues of the following equation:
a ci ai ci
(40)
In other words, when the operator, a, acts on the particular wave function, ci, one
obtains a value for the observable property, ai, multiplied by the unchanged wave
function. Examples of operators of interest would be those used to generate
atomic coordinates or particle momenta.
Undoubtedly, the most important operator is the one used to obtain the total
energy of a system. Given earlier as Equation (29), Hamiltons equation states
that the total energy of a system is the sum of the potential and kinetic energies.
Therefore, the Hamiltonian operator is defined as:
T^ V^
H
(41)
where T and V are the respective operators for the kinetic and potential energy
terms. The definition of a Hamiltonian operator is the most important step
during an application of the Schrodinger quantum mechanics, so an example is
appropriate. In classical mechanics, the kinetic energy of a particle moving in
a system defined by Cartesian coordinates is
T

1 2
pX p2Y p2Z
2m
(42)
where m is the mass of the particle, and the pi quantities represent the linear
momenta along each Cartesian axis. To deduce a quantum mechanical operator
expression, one substitutes the quantum mechanical differential operator for
the pi momenta:
pi ! ih (@=@qi )
(43)
obtaining:

h 2 @2
@2
@2
T^
2m @X 2 @Y 2 @X 2
h 2 r2
(44)
2m
is only a function of the particle coordinates, its quantum expression
Because V
remains equal to its classical expression. In our example, the Hamiltonian operator is then given by:
2
h r V(x, y, z)
H
2m
(45)
The third postulate of the wave mechanical theory bridges the gap between
the mathematical formalism of quantum mechanical theory and real experimental
observation and is used to obtain values of experimental parameters not defined
21
by an eigenvalue equation. When a number of measurements are made on the

) will be given by:
dynamical variable, A, then the average value of A(A
~A c a c dt
(46)
Equation (46) assumes that the wave function, c, is normalized (i.e., c c dt 1)

and that the integration is carried out over all space permitted to the particle. The
average value for the observable, A, is also referred to as its expectation value, or
kAl. The fourth postulate provides another linkage between theory and experiment. When the operator, a, and the state function, c, are eigenfunctions of
each other, then the experimentally observed parameter, a, will be that given
by Equation (40).
The fifth, and final, postulate leads to the fundamental equation of the
Schrodinger statement of quantum mechanics. The evolution of a state function,
c(qi, t), is given by the equation:
c (qi , t) ih (@cqi , t=@t)
H
(47)
is the
where the factor ih (@/@t) is the energy operator for the system, and H
Hamiltonian operator. Equation (47) is often denoted as the time-dependent
nor c(qi, t) exhibits time dependence, or if
Schrodinger equation. If neither H
the time dependence in c(qi, t) can be factored out, then Equation (47) reduces
to the stationary state Schrodinger equation:
ci Ei ci
H
(48)
where Ei is the energy associated with the wave function ci.

We shall often refer to the state characterized by the lowest energy as the
ground state of the system, and therefore c0 is wave function defining the ground
state. It follows that any other state must necessarily be characterized by a higher
energy than the ground state, and such states are referred to as excited states.
Every excited state is defined by its characteristic excited state wave function
that contains a total specification of the state and its properties.
One may now define the two general types of spectroscopic transitions.
One of these is absorption spectroscopy, where a transition takes place from
the unique ground state to one of the excited states. The other is emission spectroscopy, where a transition takes place from a state of high energy to a state of
lower energy. The terminating state in an emissive transition may be (but is not
necessarily) the ground state.
The formalism just described can be brought to life through the exposition
of a simple, yet illustrative, example. We will consider the example of a single
particle moving in one dimension, known colloquially as the one-dimensional
box problem. In this case, the Hamiltonian operator of Equation (45) will
involve the second derivative in only one coordinate (such as the x-axis), so
22
Brittain
Equation (47) becomes:

h 2 @2 c (x, t)
@c (x, t)
V(x)c(x, t) E
@t
2m @X 2
(49)
The potential function, V(x), depends on the nature of the interactions that are
exerted upon the particle. If the particle is confined to a definite region in
space, satisfactory solutions for c (x, t) can be found only for certain values of
the energy, E, of the particle. The wave functions that satisfy Equation (49)
reveal the position of the particle in the system through the probability function,
c (x, t)2.
A number of assumptions make possible the solution of the Schrodinger
equation for this particle in a one-dimensional box. The solutions to this problem
will entail wave equations that are time-independent, so the [@c(x, t)/@t]
term must equal zero. Outside of the box that is defined by x 0 and x a,
the potential energy of the particle must be infinite, so there will be zero probability [i.e., c (x, t)2 0] of finding the particle outside of the box. This boundary condition therefore requires that c(x, t) 0 outside of the box as well. To be
well behaved, and avoid a discontinuity at x 0 and x a, the wave function
must also equal zero at these boundaries. Finally, because the motion of the particle is not constrained in any way when it is inside the box, the potential energy
term must equal zero. In that case, the Schrodinger equation becomes:
h 2 @2 c
Ec (x)
2m @X 2
(50)
The form of Equation (50) is well known, and its solution has the general form of:
c (x) A sin (npx=a)
(51)
where n is an integer, having the values 1, 2, 3, 4, and so on, and A is a constant

equal to (2/a)1/2. The existence of quantum numbers is required by the use of
quantum wave mechanics, even for this simple example.
Substitution of Equation (51) back into equation (50) and rearranging
yields an equation for the energy of the particle as:
En
n2 h2
8ma2
(52)
The conclusion is obtained that for each wave function, cn, defined by Equation
(51) there is a single energy, En, defined by Equation (52). In that case, each wave
function is defined by a unique quantum number, and the state of lowest energy
(the ground state) must be defined by n 1 (for which E1 h 2/8ma 2). The state
of the next highest energy (the lowest energy excited state) would be defined by
n 2 and would be characterized by E2 h 2/2ma 2.
One could then envision the simplest absorption spectroscopy possible
within an one-dimensional box as consisting of a transition from the ground
23
state defined by c1 to the excited state defined by c2. The energy of this lowest
energy transition is given by:
DE12 E2 E1 3h2 =8ma2
(53)
The energy of the excited state defined by c3 is calculated to be 9h 2/8ma 2, so the

energy pertaining to the c1 ! c3 transition is:
DE13 E3 E1 7h2 =8ma2
(54)
Because it is evident that DE13 . DE12, then the c1 ! c3 absorption transition

would be predicted to take place at a higher frequency (shorter wavelength)
than would the c1 ! c2 absorption. Were one able to empirically measure the
energies of these two hypothetical transitions, then a correlation of the experimental DE values with the theoretical DE values would permit a deduction to
be made regarding the quality of the wave functions used in the computation.
The correlation of energy states with experimentally observed spectral transitions
represents an assignment of the empirical data and is the ultimate goal of theoretical spectroscopy.
Additional insight into wave functions, energy states, and transition energies can be obtained by permitting the motion of the particle to take place in
three dimensions rather than one. The wave function pertinent to this situation
will have the form c(x, y, z), and one assumes that the boundaries of the box
run from x 0, y 0, and z 0 to x a, y b, and z c.
Applying all of the same assumptions as used previously to the threedimensional problem, the time-independent Schrodinger equation becomes:

h 2 @2 c (x; y; z) @2 c (x; y; z) @2 c (x; y; z)
E c (x; y; z)
@X 2
@Y 2
@Z 2
2m
(55)
But since c (x, y, z) is a simultaneous function of all three x, y, and z Cartesian

coordinates, a solution to Equation (55) can only be found if the variables in
c (x, y, z) are first separated. This can be accomplished by assuming that any
motion of the particle along any given direction is not affected by its motion
along any other direction. This is equivalent to an assumption that the overall
state wave function is the product of three independent wave functions, each
of which is in turn a function of only one variable:
c (x, y, z) c (x)c (y)c (z)
(56)
so that the total energy, E, of the c (x, y, z) function will be given by:
E EX EY E Z
(57)
24
Brittain
Upon substitution of Equation (56) into Equation (55) and collection of variables,
one obtains three equations:
h 2 @2 c (x)
Ec(x)
2m @X 2
2
2
h @ c (y)
E c(y)

2m @Y 2
h 2 @2 c (z)
E c(z)

2m @Z 2
(58)
(59)
(60)
for which the solutions are:
c (x) A sin (np x=a)

c (y) B sin (pp y=b)
c (z) C sin (qp z=c)
(61)
(62)
(63)
where n, p, and q are integers (separately permitted to have values of 1, 2, 3, 4,

etc.), and A (2/a)1/2, B (2/b)1/2, and C (2/c)1/2. The complete solution to
the particle in a three-dimensional box is obtained by substituting Equations (61
63) into Equation (56) to obtain:
c (x, y, z) D sin (np x=a) sin (pp y=b) sin (qp z=c)
(64)
where D ABC. One immediately notes from Equation (64) that the wave function of a particle able to move in three dimensions can only be completely specified by the use of three quantum numbers.
The energy of a given c(x, y, z) function is given by:

h2 n2 p2 q2
Enpq
(65)
8m a2 b2 c2
The unique ground state (c111) associated with this system is defined by
n p q 1, and will have an energy of:

h2 1
1
1
E111
(66)
8m a2 b2 c2
The energies of the lowest energy excited states will be given by:

h2 4
1
1
E211

8m a2 b2 c2

h2 1
4
1

E121
8m a2 b2 c2

h2 1
1
4

E112
8m a2 b2 c2
(67)
(68)
(69)
25
but without knowing values for a, b, and c, the relative orders of c211, c121, and
c112 cannot be deduced.
A very interesting situation arises for the case where the three-dimensional
box is cubic or where a b c. In that case, the energy of the c(x, y, z) functions
is given by:
Enpq
h2
n2 p2 q2
8ma2
(70)
The unique ground state (c111) associated with this cubic system is still defined
by n p q 1 and has an energy of:
E111
3h2
8ma2
(71)
The energies of the lowest energy excited states will then be given by:
3h2
4ma2
E121 E112
E211
(72)
(73)
One finds in this situation the existence of three states, each of which is defined by
a unique set of quantum numbers, which are characterized by the same energy.
These three states are said to be degenerate, and an energy level is termed to
have a degeneracy corresponding to the number of states that occur at that
energy. Figure 6 illustrates the pattern of excited states that is predicted by
Equation (70). The three absorption transitions corresponding to c111 ! c211,
c111 ! c121, and c111 ! c112 must each take place at identical energies and
could not therefore be distinguished in the absence of an additional perturbation
on the system.
One can obtain the hypothetical absorption spectrum that would be
observed for the particle in a three-dimensional box, by computing the energy
differences between the states in Figure 6. The energy difference between the
c111 state and any excited cnpq state must be equal to the energy required to
effect the absorption transition. By Plancks law, this energy difference is proportional to the frequency of radiation needed to effect the absorption and inversely proportional to the wavelength of the radiation. An absorption spectrum of
this type is shown in Figure 7, where the wavelength scale has been scaled so as
to appear in reasonable units.
Thus far, we have examined the nature of state functions, the energies of
these, and the energy associated with transitions between the levels described
by these state functions. The quantum theory also permits one to calculate the
intensity of light that is absorbed or emitted during a spectroscopic transition.
In quantum mechanics, the intensity associated with a transition is determined
by the probability of its occurrence. As discussed earlier, such transitions
require the presence of an electric dipole moment within the transitioning
26
Figure 6
Brittain
Energy level diagram for the states of a particle in a three-dimensional box.
Figure 7 Hypothetical absorption spectrum for a particle in a three-dimensional box.

The intensity axis has been plotted to be proportional to the degeneracy of the final
state, and the wavelength axis has been arbitrarily fitted to a pseudonanometer scale so
as to appear in reasonable units.
27
species, and wave mechanics indicates that the spectral transition probability, Pij,
from a lower state (ci) to an upper state (cj) is given by:
Pij ci mc
^ j dt
(74)
where m is the total electric dipole moment operator. The probability of the
ci ! cj transition is known as the transition moment, and its use replaces the
classical Fourier analysis of the electric moment to obtain the radiation intensity
associated with the transition. When the value of the transition moment of
Equation (74) does not equal zero, the spectral transition is said to be electric
dipole allowed.
One can also define the transition moment for a magnetic dipole
transition as:
c j dt
Mij ci m
(75)
is the total magnetic dipole moment operator. In general, pure magnetic
where m
dipole allowed transitions are significantly weaker than the corresponding pure
electric dipole allowed transitions. However, the simultaneous existence of
nonzero transition moments of both types will be required for the observation
of chiroptical absorption or emission.
For the example of a particle in a three-dimensional box, the transition
moment can be factored into three component integrals after a separation of
variables is conducted:
Pij (x) ci m^ X c j dt
(76)
(77)
Pij ( y) ci m^ Y c j dt
(78)
Pij (z) ci m^ Z c j dt
where the three mi operators represent the components of the electric dipole
moment operator along each of the three Cartesian axes. Depending on the
nature of ci and cj, any one of the Pij(x), Pij( y), or Pij(z) moments can equal
zero (or not). When a given moment has a nonzero value along a particular
axis, the transition is said to be polarized along that axis. For real molecules,
the exact computational task can be formidable, but we shall eventually see
that the use of certain symmetry arguments will permit one to make qualitative
deductions of great value.
To recapitulate, in classical mechanics all the necessary coordinates and
momenta of a system are totally specified, and the interactions among these
are totally accounted for. This allows one to calculate the configuration and
motion within the system at any future time with complete predictability.
28
Brittain
Quantum mechanics, on the other hand, concedes the exact determination of

some dynamical variables and provides their computational solutions in terms
of probability functions. The inability of a state function to provide exact
values for certain variables is often consistent with the inability of some experimental measurements to yield more exact determinations. One concludes that
quantum mechanics is consistent with observations for situations where the
application of classical mechanics leads to serious inconsistencies and problems.
OVERVIEW OF SPECTROSCOPY
Having developed the basic concepts needed to define the nature of spectroscopy,
it now remains to discuss how electromagnetic radiation quanta of various magnitudes may interact with the subatomic and atomic constituents of atoms and
molecules. One way to classify the range of spectroscopic possibilities would
be the brute force interpretation of each of the spectral energy regions defined
in Table 1, but this scheme lacks elegance and impedes clarity in the presentation.
Instances exist where the interaction of matter with quanta of different magnitudes (from different spectral regions) leads to the occurrence of the same type
of state transition. These processes may proceed through different mechanisms
and may take place within different spectral regions, so the requisite crossreferrals unnecessarily complicate matters.
The superior method of spectral classification is to consider the possible
types of energy level transitions that might take place in atoms or molecules
and to examine the magnitude of the light quanta and associated processes that
cause such transitions to take place. This latter approach has the advantage
that one might proceed directly from the precepts discussed in this chapter to
an area of importance for the practicing pharmaceutical spectroscopist. After
that, other types of spectroscopy can be introduced as refinements on this basic
model. This approach also permits the introduction of additional theory on an
as-needed basis, which will often be spliced into the elements of a preceding
section.
Several types of atomic and molecular states are characterized by wave
functions that are largely functions of electronic character. Electrons in these
states are ordinarily subdivided as to whether they are actively involved in
chemical bonding (known as valence electrons) or not (known as core electrons).
Other states are characterized by wave functions that are largely functions of
vibrational or rotational nuclear motion. Still other states are defined by the
paramagnetic properties and spin states of either the electrons or of the nuclei.
Altogether, it is useful to define six broad classes of wave function types and
the spectroscopy associated with each grouping.
The first category will concern spectroscopic transitions among states
defined as having predominantly core electron properties. The energies separating these states have magnitudes comparable with the energies of X-rays,
and so X-ray absorption spectroscopy is one of the techniques used to probe
29
coordinative environments of absorbing atoms. Also within this grouping are the
various techniques of atomic spectroscopy and photoelectron spectroscopy.
These techniques are characterized by the use of light quanta possessing
sufficient energy to cause an electron to be completely ejected (i.e., ionized)
from an atom or a molecule. One might study the excess kinetic energy associated
with the ejected electron when the energy of the excitation photon is larger than
the ionization energy. This approach encompasses the various techniques of
photoelectron spectroscopy, where one allows a beam of photons of known
energy to interact with a sample and measures the excess kinetic energy of the
ejected electrons. By subtracting the excess kinetic energy from the incident
photon energy, one obtains the binding energy of the electron. Electronic energies are then correlated with structural features of the molecule or compared
with theoretical calculations. Depending on the choice of experimental circumstances and observing mode of the analyst, the ionization process can be
accompanied by the absorption or emission of electromagnetic radiation.
This gives rise to the families of spectroscopy conveniently classed as atomic
absorption or as atomic emission spectroscopies.
The second category concerns spectroscopic transitions among states
defined as predominantly having valence electron properties. This category contains some of the most familiar types of spectroscopy, namely absorption in the
ultraviolet, visible, and near-IR regions of the spectrum. Techniques have been
developed that permit such work to be conducted on the basis of either light transmission or of reflected light. A variety of complementary techniques have also
been developed that depend on the interaction of polarized light with molecules
of certain types, and these have great importance in the study of chiral compounds of pharmaceutical interest. Emission spectroscopies, such as fluorescence
and phosphorescence, are also characterized by transitions among valence electron states, and they have found extensive use in pharmaceutically oriented
spectroscopy.
The third category concerns spectroscopic transitions among states whose
wave functions are independent of electronic character and instead are defined
as having predominantly nuclear motion properties. These wave functions
describe various states of molecular vibration, and transitions between vibrational
states in molecules are caused by the absorption of energy in the IR region of the
electromagnetic spectrum. IR absorption spectroscopy can be observed in either
transmission or reflectance modes. Interestingly, radiation in this region can be
produced by hot glowing metal and represents a practical use of blackbody radiation. Transitions among vibrational states can also be measured using Raman
spectroscopy, where one observes the frequency of scattered light rather than
that of absorbed light. Energy differences between weak scattering lines and the
main irradiation line correspond to vibrational and/or rotational transitions in
the system being studied.
The fourth category also concerns spectroscopic transitions among states
whose wave functions are characterized by nuclear motion properties, but
30
Brittain
where transitions among these states is associated with pure rotational motion of
the molecule. Consequently, transitions among these states constitute rotational
spectroscopy. Most of these transitions occur in the microwave region, except in
the case of very light molecules where the transitions take place in the farIR. Microwaves are characterized by the fact that they are generated by
klystrons and magnetrons and are conducted by waveguides (hollow pieces of
metal tubing with either a circular or rectangular cross-section).
The fifth category encompasses the states defined by wave functions
specifically related to the properties of unpaired electrons in a molecule. One
might think this to be a technique of limited utility, but examples of applicable
species include organic free radicals, radiation-induced radicals, paramagnetic
transition metal ions, defect centers in solids, and conduction electrons in
metals. When a material containing one or more unpaired electrons is placed
in an external magnetic field, transitions between different electron spin states
can be induced by radiation in the microwave region. Such spectroscopy is
properly termed electron paramagnetic resonance, but one often encounters
the inaccurate term electron spin resonance to describe the same phenomenon.
Experiments are usually performed at frequencies anywhere in the range of
2000 to 36,000 MHz. A frequency of 10,000 MHz corresponds to a wavelength
of 3 cm, and an energy of 0.3 cm21.
The sixth category encompasses the states defined by wave functions
specifically related to the properties of nuclear paramagnetism. Here, transitions
among nuclear spin states are induced by absorption of electromagnetic radiation
in the radio-frequency region of the spectrum after these have been split by the use
of an external magnetic field. Nuclear magnetic resonance spectroscopy detects
transitions between nuclear spin states in an applied magnetic field, whereas
nuclear quadrupole resonance spectroscopy detects the splitting in the nuclear
spin levels arising from the interaction of an unsymmetrical charge distribution
of certain nuclei with an electric field gradient. The size of quanta used in this
work is very small (i.e., much less than kT at room temperature), so that these
types of spectroscopy require the use of sophisticated techniques to achieve the
population differences in nuclear spin states required for transitions to take place.
SUMMARY
It is now firmly established that spectroscopy originates from transitions among
states that are properly described by quantum mechanical wave functions. The
irradiation of a sample by electromagnetic radiation will only result in its absorption if the energies of the quanta, which constitute that radiation correspond to a
difference in energy levels within the substance. Any electromagnetic radiation
emitted by a sample under a suitable stimulus condition will be of energy
equal to an energy level difference within the emitting substance.
Because the wave functions describing the properties of a substance can
ordinarily be factored according to a majority property of the atom or molecule,
31
the absorption or emission of electromagnetic radiation can be considered to

affect the electronic, vibrational, or rotational energies of a molecule. Quanta
of very low energy will affect the spin states of unpaired electrons and paramagnetic nuclei. Owing to the differing magnitudes of the energy level differences
associated with these different states, the type of electromagnetic radiation
used to effect the spectroscopic change will generally fall in a characteristic
region of the spectrum.
REFERENCES
Classical Mechanics
1. Born M, Wolf E. Principles of Optics. New York: Macmillan, 1964.
2. Corben HC, Stehle P. Classical Mechanics. New York: John Willey & Sons, 1960.
3. Feynman RP, Leighton RB, Sands M. The Feynman Lectures on Physics, Vol. 1.
Reading: Addison-Wesley, 1963.
4. Goldstein H. Classical Mechanics. Reading: Addison-Wesley, 1980.
5. Leech JW. Classical Mechanics. London: Methuen, 1958.
6. Kibble TWB. Classical Mechanics. New York: John Wiley & Sons, 1973.
7. Richtmeyer FK, Kennard EH. Chapters 1, 2. Introduction to Modern Physics.
New York: McGraw-Hill Book Co., 1942.
8. Rutherford DE. Classical Mechanics. New York: Wiley-Interscience, 1951.
Quantum Mechanics
9. Atkins PW. Molecular Quantum Mechanics. Oxford: Clarendon Press, 1970.
10. Daudel R, Lefebvre R, Moser C. Quantum Chemistry. New York: Wiley-Interscience,
1959.
11. Davis JC. Advanced Physical Chemistry. New York: Ronald Press, 1965.
12. Eyring H, Walter J, Kimball GE. Quantum Chemistry. New York: John Wiley & Sons,
1944.
13. Hanna MW. Quantum Mechanics in Chemistry. Menlo Park: Benjamin/Cummings,
1981.
14. Kauzmann W. Quantum Chemistry. New York: Academic Press, 1957.
15. Levine IN. Quantum Chemistry. Boston: Allyn and Bacon, 1970.
16. Messiah A. Quantum Mechanics. North-Holland, Amsterdam, 1961.
17. Pauling L, Wilson EB. Introduction to Quantum Mechanics. New York:
McGraw-Hill, 1935.
18. Pilar FL. Elementary Quantum Chemistry. New York: McGraw-Hill, 1968.
19. Rojansky V. Introductory Quantum Mechanics. Englewood Cliffs: Prentice-Hall, 1938.
20. Slater JC. Quantum Theory of Molecules and Solids. New York: McGraw-Hill, 1963.
General Spectroscopy
21. Barrow GM. Molecular Spectroscopy. New York: McGraw-Hill, 1962.
22. Brittain EFH, George WO, Wells CHJ. Introduction to Molecular Spectroscopy.
London: Academic Press, 1970.
32
Brittain
23. Guillory WA. Introduction to Molecular Structure and Spectroscopy. Boston: Allyn
and Bacon, 1977.
24. King GW. Spectroscopy and Molecular Structure. New York: Holt, Rinehart and
Winston, 1964.
25. Reid C. Excited States in Chemistry and Biology. New York: Academic Press, 1957.
26. Richards WG, Scott PR. Structure and Spectra of Molecules. Chichester: John Wiley
& Sons, 1985.
27. Sandorfy C. Electronic Spectra and Quantum Chemistry. Englewood Cliffs:
Prentice-Hall, 1964.
28. Simons JP. Photochemistry and Spectroscopy. London: Wiley-Interscience, 1971.
PART II.
CORE ELECTRON SPECTROSCOPY
2
Core Electron States and X-Ray
Absorption Spectroscopy
Harry G. Brittain
WAVE MECHANICAL DESCRIPTION OF THE HYDROGEN ATOM

Although the theory of Bohr sufficed to interpret the spectral lines associated with
elemental hydrogen, which had been excited by a flame, the extension of this
theory to entirely explain the line spectra of heavier atoms proved to be
elusive. In addition, an explanation for the doubling of lines in the hydrogen spectrum eluded the simple theory. As indicated in the last chapter, only the wave
mechanical theories of Schrodinger, Heisenberg, and Dirac were able to
provide the necessary mathematical description of atomic structure that permitted
physicists to adequately interpret the empirical data that had become widely
available. It became obvious to these workers that an understanding of the
spectroscopy of atomic and molecular states derived from core electrons in a molecule required the development of a comprehensive picture for the quantum
numbers that define such states. The solution to the wave mechanical equation
of the hydrogen atom was found to meet this need, because it could be solved
exactly. Furthermore, its solutions would then be used to develop approximate
treatments for heavier atoms.
As with all quantum mechanical derivations, one begins with the
Schrodinger equation:
^ ci Ei ci
H
(1)
33
34
Brittain
where Ei is the energy associated with the wave function ci. The heart of the
derivation is deducing the correct form for the Hamiltonian operator:
^ T^ V^
H
(2)
terms. The form for the kinetic energy remains as defined in chapter 1:

h 2 @2
@2
@2
r2
T^
(3)
h2
2
2
2
2m @X
@Y
@X
2m
The potential energy term, V, for the electron is given by the attractive
force between the negatively charged electron and the positively charged
nucleus. This expression is available as Coulombs Law:
V e2 =r
(4)
where r is the distance of the electron from the nucleus, and e is the electron
fundamental charge. The Schrodinger equation now takes the form:
Ec
h 2 2
e2
r c c
2m
r
(5)
The presence of a radial distance as one of the parameters in the potential function precludes one from working in the Cartesian coordinate system. Fortunately,
the situation of a single electron moving about a nucleus permits a ready application of the spherical polar coordinate system. In this system, space is defined by
a radial distance varying from 0 to 1, an angle u varying from 0 to p, and an
angle f varying from 0 to 2p. As illustrated in the polar coordinate system illustrated in Figure 1, the position of a point at location (x,y,z) becomes specified by:
x r sin u cos f
(6)
y r sin u sin f
z r cos u
(7)
(8)
By inserting the coordinate transformation relationships of Equations (6) through (8)

into Equation (5), one obtains:

1 @ 2 @c
1
@
@c
1
@2 c h 2
e2
r
E
sin u
2 2
0 2
2
r @r
r sin u @u
@r
@u
r
r sin u @f2 2m
(9)
The advantage of using spherical polar coordinates is that it permits a facile separation of the variables, which is essential to obtain a solution to the equation. One
assumes that the motion of the electron along one coordinate direction is independent of its motion along the other two directions (the same assumption that was
used to obtain a solution for the particle in the three-dimensional box), so then
one can write:
c(r, u, f) R(r)Q(u)F(f)
(10)
Core Electron States and XAS
35
Figure 1 The spherical polar coordinate system used to describe the position of an object
with respect to an origin. The magnitude of the radial distance (r) describes a sphere upon
whose surface the object is found, and its position on the surface of this sphere is defined
by the intersection of the two planes described by the u and f angles.
In Equation (10), each of the component wave functions is a function of one variable
only. This equation is substituted into Equation (9), whereupon one obtains:

1 d 2 dR
1
d
dQ
r
sin u
0 2
2
r R dr
dr
r ( sin u)Q du
du
(11)

2
2
2
1
d F h
e
E
2
r
r ( sin2 u)F df2 2m
In Equation (11), the partial differentiation symbols have been replaced by total
differentiation symbols, because each function is now a function of a single variable.
By multiplying through by r 2(sin2 u) and rearranging, one obtains:

1 d2 F sin2 u d 2 dR
sin u d
dQ
r
sin u
F df 2
dr
du
R dr
Q du
(12)

2
2
h
e
2
2
E
(r sin u)
2m
r
Because the left-hand side of Equation (12) contains only r and u as variables
and because the right-hand side is a function of f only, each side must therefore
36
Brittain
be equal to some common constant. For reasons, which will soon become apparent,
this constant is denoted as 2m 2. In that case, the left-hand side of Equation (12) can
be written in the form known as the F equation:
d2 F
m2 F
df2
(13)
The right-hand side of Equation (12) must also equal m 2, hence substituting in this
value and dividing through by sin2 u, one obtains:

2

h
1 d 2 dR
e2 2
r
E
r
R dr
dr
2m
r

m2
1
d
dQ
2
sin u
du
sin u Q sin u du
(14)
One has now achieved the separation of all functions in r to the left-hand side of
Equation (14) and all functions in u to the right-hand side. Because these two sets
of terms are independent of each other, but their sum is a constant (equal to zero),
each set must be equal to a constant. Setting the sum of the terms in u equal to
the constant 2 b, we find:

h 2
1 d 2 dR
b
e2
r
E

R0
r 2 dr
dr
r2
2m
r
(15)

1 d
dQ
m2 Q
sin u
2 bQ 0
sin u du
du
sin u
(16)
and
Equation (15) is known as the R equation, and Equation (16) is known as the Q
equation.
Through the use of an appropriate coordinate system, the Schrodinger
equation for the hydrogen atom has been reduced to the solution of three separate
equations, each of which is constructed in terms of a single variable. It is interesting to note that although the energy of the system depends only on r, a complete description of the wave functions requires that all three equations be solved.
To simplify the course of the present discussion, however, we will not
pursue the traditional method of solving each wave equation separately but
will instead follow an alternate approach. In our treatment, we will factor the
overall wave function into a radial function, R(r), and an angular function,
Y(u, f):
C(r, u, f) R(r)Y(u, f)
(17)
37
Upon substitution of Equation (17) into Equation (9), one obtains:

2

h
1 d 2 dR
e2 2
1
@
@Y
1
@2 Y
r
E
sin u
r
2
R dr
dr
Y( sin u) @u
@u
2m
r
Y( sin u) @f2
(18)
In Equation (18), the partial differentiation symbols for the R equation have
been replaced by total differentiation symbols, because that function is a function
of a single variable. As soon earlier, both sides of Equation (18) must each be
equal to a constant that shall be termed b. One then obtains:

h 2
1 d 2 dR
b
e2
r
E

R0
(19)
r 2 dr
dr
r2
2m
r
and

1 d
dY
1 @2 Y
sin u
bY 0

sin u du
du
( sin2 u) @f2
(20)
Equation (19) has the same form as the previously derived R equation, and
Equation (20) is known as the Angular equation. A detailed analysis of the
mathematics is available in the standard references (1 6), but it shall suffice to
note that the constant b is better defined by the term l(l 1).
Solution of the Angular Equation

Equation (20) is not a simple differential equation and consequently does not
have an obvious solution. However, through appropriate mathematical manipulation one can rearrange the angular equation into a form that is equivalent to
that of the associated Legendre equation. The solutions to this equation are
found by assuming polynomial functions that are single-valued and that exist
for only certain combinations of defining constants. The solutions to the
Legendre equation are often termed spherical harmonics, whose value is determined by two different types of integers (i.e., quantum numbers). The first of
these is m, which can only assume the values:
m 0, +1, +2, +3, +4, etc:
(21)
The constant m is called the magnetic quantum number, because it effectively

corresponds to the quantum number of the same number invoked in the Bohr
theory to explain the splitting of spectral lines in the presence of a magnetic
field. The second constant, l, is called the azimuthal quantum number and is
restricted to values such that:
jmj
(22)
38
Brittain
Table 1
Selected Angular Wave Functions for the Hydrogen Atom
Azimuthal quantum
number (l)
Magnetic quantum
number (m)
Angular
wave equation
0
1
1
0
0
+1
2
2
0
+1
+2
(1/4p)1/2
(3/4p)1/2 cos u
(3/4p)1/2 sin u cos f
and
(3/4p)1/2 sin u sin f
(5/16p)1/2 (3 cos2 u 2 1)
(15/4p)1/2 sin u cos u cos f
and
(15/4p)1/2 sin u cos u sin f
(15/16p)1/2 sin2 u cos 2f
and
(15/16p)1/2 sin2 u sin 2f
The solution to Equation (20) is written in the form:

imf
Y(u, f) Pjmj
( cos u)e
(23)
where m is permitted to assume the values 0, +1, +2, +3, . . . ,+l. An excellent
description of the properties of the associated Legendre polynomials, Pjmj
, is
available (7). An examination of associated Legendre functions reveals that the
functions equal zero for all cases where the value of jmj exceeds the value of
l. Alternatively, the solutions to the angular wave equation can be written in
the equivalent real forms:
Y(u, f) Pjmj
( cos u) sin (mf)
(24)
Pjmj
( cos u) cos (mf)
(25)
Y(u, f)
In Equations (24) and (25), m is a zero or positive integer whose value does not
exceed that of l.
A number of angular wave functions that are of particular utility to a discussion of atomic properties are given in Table 1. Considering the restrictions
existing for the magnetic and azimuthal quantum numbers, it is indicative that
for a given value of l, there are (2l 1) different spherical harmonic functions,
each of which is characterized by a different value of m (i.e., m 2l, 2l 1,
2l 2, . . . , 0, . . . , l 2, l 1, l).
Solution of the Radial Equation
Through suitable mathematical manipulation, the radial Equation (19) can be
reduced to the form of a standard differential equation whose solutions are in
the form of polynomials. This process is facilitated by the introduction of two
39
new constants:
a2 2mE=h 2
(26)
l me2 =h 2 a
(27)
and
The analysis is also aided by the definition of a new variable, r:
r 2ar
The R equation then has the simpler form:

1 d 2 dR
b l 1
r
2 R0
dr
r
r 4
r2 dr
(28)
(29)
where the radial function now has the form of R(r). Because the radial wave
function must go to zero as r ! 1, the general solution to Equation (29) has
the form:
R(r) er=2 F(r)
(30)
where the F(r) functions have the form of the associated Laguerre polynomials
(7). These have the general form:
X
F(r) r
ak rk
(31)
where the summation takes place from k 0 to k 1. If the new expression for
R(r) is substituted back into Equation (29), one obtains an infinite number of
terms in increasing powers of r. Because the coefficients of all terms in a
given power of r must equal zero, it is possible to deduce the recursion formula:

k1l
ak1 ak
(32)
(k 1)(k 2 2)
It is easily shown that as r ! 1, R(r) cannot be finite unless the series terminates. To terminate the series defined by the coefficients of Equation (32), it
must be that:
k1l0
(33)
which in turn requires that l be an integer. It is common practice to use the

symbol n for l, so that:
nk1
(34)
It is immediately apparent from Equation (34) that n can have any positive
integral value but that it cannot equal zero. This quantum number is known as the
principal quantum number because it is related to the energy states of the system.
40
Brittain
Table 2
Selected Radial Wave Functions for the Hydrogen Atom
Principal quantum
number (n)
Azimuthal quantum
number (l )
Radial wave
equation
1
2
2
3
0
0
1
0
3
3
1
2
2 (me 2/h2)3/2 e 2r
p
(1/2 2) (me 2/h2)3/2 (2 2 r) e2r/2
p
(1/2 6) (me 2/h2)3/2 r e2r/2
p
(2/81 3) (me 2/h2)3/2
(27 2 18r r2) e2r/3
p
(4/81 6) (me 2/h2)3/2 (6r 2 r2) e2r/3
p
(4/81 30) (me2/h2)3/2 r2 e 2r/3
Note: r rme 2/h2.
A number of radial wave functions, which are of particular utility to a discussion of atomic properties, are given in Table 2. For a given value of l, n is
restricted to the values l 1, l 2, and so on. But because the value of n
determines the energy of the system, it is more convenient to consider n as the
fundamental quantum number of the system. In that case, for a given value
n. one may have l 0, 1, 2, . . . , n 2 1. It is already shown that for a given
value of n, one may have m 2l, 2l 1, 2l 2, . . . , 0, l 2, l 1, l.
Wave Functions of the Hydrogen Atom and Hydrogen-Like Atoms

Equation (17) defined the wave function of the hydrogen atom, and this description is now extended to include subscripts pertaining to the associated quantum
numbers:
cnm (r, u, f) Rn (r)Ym (u, f)
(35)
Solutions for both the angular wave functions, Ym (u; f), have been provided in
Table 1, and solutions for the radial wave functions, Rn (r), are found in Table 2.
A particular state of a hydrogen atom is now seen to be defined by a definite set of
n, l, and m quantum numbers, and its wave function is obtained by multiplying
the appropriate individual wave functions.
We can readily extend the scope of these derivations to include all possible
one-electron atoms, which are ordinarily termed as being hydrogen-like atoms.
Atoms in this series include He, Li2, Be3, and so forth. The main difference
that arises for these hydrogen-like atoms is found in the statement of the potential
energy, which is still derived from Coulombs Law but is now given as:
V Ze2 =r
(36)
where Z is the atomic number of the atom in question. The inclusion of higher
values for the nuclear charge does not materially affect any of the wave function
41
derivations and is simply carried through during the solution of the radial
equation.
As we have stated earlier, for a given value of n, the azimuthal quantum
number can have values of l 0 to l n 2 1. Furthermore, a state characterized by a given value of l can, in turn, exist as any one of the substates characterized by different values of the magnetic quantum number, m. Because m can
possess all values from 2l, through 0, and up to l, it follows that the number of
such states is given by the expression (2l 1). Following the nomenclature of
Mulliken, we will refer to the cn m (r, u, f) wave function as an orbital.
States of the hydrogen atom characterized by l 0 are known as s orbitals,
those by l 1 are p orbitals, those by l 2 are d orbitals, and those by l 3 are
f orbitals. For example a state or orbital characterized by n 3 and l 1 is
termed a 3p state or a 3p orbital. There are actually three substates within a p
state, and these are characterized by m 21, m 0, and m 1. Because
these three p states differ only in their angular properties, we can plot their
angular properties using the pertinent aspects of their angular wave functions.
The advantage of having obtained the solutions to the angular equation in the
real forms that were summarized in Table 1 is that their three-dimensional probability density graphs have recognizable shapes that permit some physical
interpretation to be assigned to them. For instance, all p orbitals characterized
by the (sin u cos f) term have their major density aligned along the x-axis and
are therefore known as px orbitals. Similarly, all p orbitals characterized by
the (sin u sin f) term have their major density aligned along the y-axis and are
therefore known as py orbitals. Finally, the p orbitals characterized by the
(cos u) term have their major density aligned along the z-axis and are therefore
known as pz orbitals.
Taking advantage of this additional nomenclature, one may write the full
wave functions that describe the various orbitals for hydrogen-like atoms, and
a number of these are collected in Table 3.
Energy Levels and Spectroscopy of the Hydrogen Atom

Substitution of n for l back into the defining Equations (26) and (27) yields the
result:
En
Z 2 me4 1
2h 2 n2
(37)
Equation (37) represents exactly the same result that had been obtained from the
Bohr theory (Equation 32 of chap. 1). One immediately notes that the total energy
of a state is solely determined by the identity of the principal quantum number,
n. All of the substates associated with a given value of n (and defined by their l
and m quantum numbers) are all therefore of equal energy in the absence of an
additional perturbation. These states of the same energy are said to be degenerate.
42
Brittain
Table 3
Normalized Wave Functions of the Hydrogen Atom
Principal Azimuthal Magnetic

quantum quantum quantum
number
number
number
(m)
(l)
(n)
1
2
2
2
0
0
1
1
0
0
0
+1
3
3
3
0
1
1
0
0
+1
+1
+2
Normalized
wave equation
p
c1S (1/ p) (Zme 2/h 2)3/2 e2r
p
c2S (1/4 2p) (Zme 2/h 2)3/2 (2 2 r) e2r/2
p
c2Pz (1/4 2p) (Zme 2/h 2)3/2 r e 2r/2 cos u
p
c2Px (1/4 2p) (Zme 2/h 2)3/2 r e 2r/2 sin u cos f
p
c2Py (1/4 2p) (Zme 2/h 2)3/2 r e2r/2 sin u sin f
p
c3S (1/81 3p) (Zme 2/h 2)3/2 (27 2 18r r2) e 2r/3
p
c3Pz (2/81 p) (Zme 2/h 2)3/2 (6r 2 r2) e 2r/3 cos u
p
c3Px (2/81 p) (Zme 2/h 2)3/2 (6r 2 r2) e 2r/3
sin u cos f
p
c3Py (2/81 p) (Zme 2/h 2)3/2 (6r 2 r2) e 2r/3
sin u sin f
p
c3dz 2 (1/81 6p) (Zme 2/h 2)3/2 r2 e 2r/3
2
(3cos u 2 1)
p
p
c3dxz ( 2/81 p) (Zme 2/h 2)3/2 r2 e 2r/3
sin u cos u cos f
p
p
c3dyz ( 2/81 p) (Zme 2/h 2)3/2 r2 e 2r/3
sin u cos u sin f
p
c3dx2 y2 (1/81 p) (Zme 2/h 2)3/2 r2 e 2r/3
2
sin u cos 2f
p
c3dxy (1/81 p) (Zme 2/h 2)?? r2 e 2r/3
2
sin u sin 2f
2
Note: r rZme 2/h
.
The azimuthal quantum number l is the quantum number associated with

the total angular momentum of the electron in the given state (i.e., orbital). In
quantum mechanical terminology, the spherical harmonic functions (Pjmj
) are
eigenfunctions of the total orbital angular momentum operator (L), and have
an eigenvalue equal to l (l 1) h 2, or:
2 jmj
L^ 2 Pjmj
( 1)h P
(38)
As stated earlier, the quantum number l is restricted to integral values

between 0 and (n 2 1) and gives the number of nodes existing in the angular
wave function. For more complex atoms, the energies of the substates
defined by differing values of l will not be equal, and these states will not be
degenerate.
The magnetic quantum number m is associated with the component of
angular momentum along a specific axis in the atom that is usually the z-axis.
Because all atoms are spherically symmetric, there is no way to define a
43
particular axis, unless the atom is placed in an electric or magnetic field. Consequently, the energy of a state is not affected by its value of m unless an external
field is present. The value of m does determine the degeneracy of a state,
however, because there are 2l 1 values of m for a state defined by the
quantum number l. The splitting of states defined by different values for m is
called the Zeeman effect. The magnetic quantum is defined explicitly during
the solution to the F equation, and the solutions to this equation are found to
be eigenfunctions of the z-component of the angular momentum operator:
L^ z F(m) mh F(m)
(39)
It should be completely evident at this point that the properties of a given

electron in an atom are critically determined by its position relative to the nucleus
(through the radial wave function) and by its angular momentum state (through
the angular wave function).
Having evaluated some of the properties associated with the wave functions of the hydrogen-like atoms, it now becomes possible to discuss the spectroscopy associated with these systems. In the previous chapter, we established
that the intensity associated with a transition is determined by the probability
of its occurrence, for which the electric dipole spectral transition probability,
Pij, from a lower state (ci) to an upper state (cj) was given by:
Pij ci mc j dt
(40)
where m is the total electric dipole moment operator. For the hydrogen atom, the
instantaneous dipole moment is 2e r, where r is the vector from the nucleus to
the electron.
Substitution of the various wave functions provided in Table 3 into
Equation (38) and performance of the necessary integration operations leads to
the realization that definite restrictions exist on the number and type of spectral
transitions that may actually take place. Recognizing that every wave function is
defined by a unique set of n, l, and m quantum numbers, one defines the lower
state wave function by ci (n0 , l0 , m0 ) and the upper state wave function by cj
(n00 , l00 , m00 ). In that case, Equation (38) becomes:
Pij ci (n0 , 0 , m0 )er cj (n00 , 00 , m00 ) dt
(41)
To evaluate this integral, each component must be evaluated separately. This is

accomplished by using the definitions of the spherical polar coordinate system as
given in Equations (6) through (8) to define the components of the electric dipole
spectral transition probability along each of the Cartesian axes. One obtains these
44
Brittain
quantities through the solution of the triple integrals:
Pij (x) e
ci (n0 , 0 , m0 )(r sin u cos f)cj (n00 , 00 , m00 )r2 sin u dru df (42)
ci (n0 , 0 , m0 )cj (n00 , 00 , m00 )r sin u cos f r2 sin u dr df (43)

Pij (y) c
ci (n0 , 0 , m0 )cj (n00 , 00 , m00 )r cos u r2 sin u dr du df

(44)
Pij (z) e
where the integration over dr takes place from 0 to 1, the integration over du
takes place from 0 to p, and the integration over df takes place from 0 to 2p.
The integrals defined in Equations (42) through (44) can be evaluated
from general Legendre polynomials to obtain selection rules that govern the
spectroscopy of the hydrogen atom, but these operations are outside the course
of this narrative. Interested readers can look up the details of the solution (2),
but the outcome of these derivations can be summarized in three relations.
n0 ! n00 ,
no restriction
(45)
For a hydrogen-like, one-electron atom, Equation (45) signifies that there are
no rules imposed on spectral transitions that are related to the relative magnitudes
of the principal quantum numbers of either the initial or final states.
0 ! 00 ,
D +1
(46)
Equation (46) signifies that transitions may take place from s-orbital
wave functions to p-orbital wave functions or that transitions may take place
from p-orbital wave functions to d-orbital wave functions, but that the transition from s-orbital wave functions to d-orbital wave functions are strictly
forbidden.
m0 ! m00 ,
Dm 0, +1
(47)
The selection rule of Equation (47) is restrictive only in those situations where
significant degeneracy exists in the wave functions by virtue of their possessing
larger values of l.
It is worth noting that the symmetry of the wave functions describing the
states of a system has an important bearing on whether integrals such as these
will vanish or not. Often, the selection rules can be determined without
needing to undertake a detailed evaluation of the integrals. It is a requirement
of all atomic wave functions that they have either even or odd parity, or that
they must either be symmetric or antisymmetric with respect to inversion of
the coordinates of all the particles through the origin. The electric dipole
moment operator involved in the matrix element will also be even or odd in
parity, because it changes sign upon inversion of the coordinates.
Now if both
ci and cj are even, then the elements of the integral ci m cj dt will also be
odd, because the product of two even functions is even and that of an even
45
function with an odd function is odd. But if the integral is odd, it will contribute
equal positive and negative amounts to the integration over all values of the coordinates, and therefore the integral must equal zero. Similarly, two odd states and
the odd dipole moment will result in an odd integral, which also results in a vanishing integral. Only when one state is even and the other odd will the integral be
even and have a finite value.
The effect of wave function symmetry is summarized in the Laporte rule,
which states that dipole radiation transitions can take place only from even
states to odd states or from odd states to even states. In the hydrogen atom situation, states characterized by l 0, 2, 4, 6, . . . are even, and states characterized
by l 1, 3, 5, 7, . . . are odd. When one deals with complex atoms, the use of
the Laporte rule greatly simplifies the evaluation of possible spectroscopic
transitions, because it is relatively easy to assign a parity value to even the
most complicated wave functions.
In the earlier discussion regarding the Bohr theory and the spectra of the
hydrogen atom, it was concluded that spectral transitions could take place
between any two energy levels specified by the principal quantum number,
n. The quantum mechanical treatment just concluded for this system shows the
assumption made by Bohr to be correct, but a restriction was found for which
the azimuthal quantum number, l, could only change by +1. This means that
because the Lyman series terminates in the 1s ground state, its spectral lines
must originate from p-states of differing values of n. Similarly, transitions with
the Balmer series must originate only from s-states and d-states, because they
terminate in the 2p-state.
Electron Spin
Although the methods of quantum mechanics just outlined were found to yield
excellent agreement with the observed spectra of atoms, it was recognized that
the theory still failed to explain several features in observed atomic spectra.
These consisted of small shifts in the expected frequencies of spectral lines
and the resolution of many spectral lines into closely spaced multiplets of
several lines. This multiplicity of lines, implying the existence of additional
electronic states and transitions, could not be predicted from the theory.
As it happened, an explanation for the multiplicity of spectral lines had
been independently advanced by Goudsmit and Uhlenbeck (8) and by Bichowsky
and Urey (9) shortly before quantum mechanics had been fully developed. These
workers postulated that an electron had an intrinsic angular momentum that
they called spin angular momentum. Some examples of phenomena that had
been explained by the concept of spin were the Stern-Gerlach experiments, the
degeneracy of the excited states of atoms and molecules, the anomalous
Zeeman effect, and the fine structure splittings of atomic spectra. The proper
theory of these properties was not obtained until Dirac applied relativity theory
to the quantum mechanical formulation (10). The mathematical considerations
46
Brittain
necessary in Diracs theory are extremely complex, but there is some value in
briefly examining the consequences of his approach.
According to the theory of relativity, the energy of a body depends not only
on its kinetic and potential energies but also on its rest mass, or its mass when it is
stationary with respect to the frame of reference being used to make measurements. If this energy term, in addition to the momentum and potential energy
terms, is inserted into the Hamiltonian operator during the construction of
the Schrodinger equation, it is found that the equation is of such a form that it
cannot be solved without the introduction of several additional operators,
whose nature cannot be discussed here. In addition, the resulting equation
cannot be solved as an ordinary differential equation but requires the use of a
more complex matrix formulation.
Some understanding of the important findings that result from the relativistic equation that Dirac solved can be acquired through a qualitative discussion of
the results. If one considers the simple case of a single particle moving in one
dimension and confined in a potential well with infinite walls, with the
nonrelativistic Schrodinger equation, one finds that only certain positive
energy levels are possible. When the Dirac theory is used, the essential difference
between the relativistic and nonrelativistic solutions is that the former includes
the relativistic rest-mass energy equivalent mc 2. However, because transitions
between electronic states involve differences in energies, the constant mc 2 is
not observed experimentally.
In the nonrelativistic solution of the Schrodinger equation, one finds that
for each energy level there is a single wave function describing the state of the
system. The solution of the relativistic equation gives the surprising result
that there are two independent, orthogonal wave functions, each of which are
the solutions for each energy level. Unfortunately, it is difficult to describe the
significance of these two states with a simple physical picture. An investigation
of the angular momentum properties of the wave function solutions reveals
that the z-component of the orbital angular momentum of a state in a hydrogen
atom does not really remain constant with time, as predicted. One is forced to
construct another angular momentum operator for which the relativistic solutions
are eigenfunctions. One of the two solutions is an eigenfunction having a positive
sign on one of its terms, and the other is an eigenfunction with a negative sign on
the analogous term.
It is plausible, then, to advance a physical explanation for the selfcontained angular momentum, which an electron apparently possesses in addition
to angular momentum arising from its orbital motions. The intrinsic angular
momentum arises because of a spinning motion of the electron mass about an
axis through its center. Such a rotational motion of mass would obviously give
rise to angular momentum, even though the particle might not itself be moving
in some path about a central nucleus. Ordinarily, this intrinsic angular momentum
of the electron is termed spin angular momentum, or simply electron spin.
One should remember, however, that this interpretation is not required by
47
relativistic quantum mechanics and is, in fact, introduced only by a process of

analogy (i.e., comparing an operator, which is a constant of motion in the relativistic case with the orbital angular-momentum operator from nonrelativistic
quantum mechanics). Be that as it may, the concept of electron spin is useful
and will be used henceforth whenever necessary.
Because there is no classical analog of electron spin, one cannot use the
normal procedure of stating a classical expression and substituting in the appropriate quantum mechanical operators. The concept of electron spin is therefore
introduced into quantum mechanics through a series of three postulates. The
first of these is that the operators for spin angular momentum commute and
combine in the same way as those for ordinary angular momentum. This yields
a series of spin angular momentum operators (S^ 2 , S^ x , S^ y , and S^ z ), which
are exactly analogous to the orbital angular momentum operators (L^ 2 , L^ x , L^ y ,
and L^ z ) that are discussed earlier. The second postulate states that for a single
electron, there are only two simultaneous eigenfunctions of the operators S^ 2
and Sz. By convention, these are called ca and cb and obey the eigenvalue
equations:
S^ z ca 12 h ca
S^ z cb 12 h cb
(48)
(49)
A comparison of Equations (48) and (49) with Equation (39) indicates that spin
angular momentum, like orbital angular momentum, can be expressed in multiples of h. The eigenvalue of a is seen to be 12, and the eigenvalue of b is 2 12,
leading to the definition of a new quantum number, ms. The permissible values
of ms are seen to be +12.
The third postulate is that the spinning electron acts like a magnet, having a
magnetic dipole moment operator equal to:
M go bm S^
(48)
where go is the spectroscopic splitting factor (equal to 2.0023), and bm is the Bohr
magneton (equal to 9.2732 10221 erg/G). It is interesting to note that the value
of go was originally obtained empirically but was later derived a priori from the
Dirac theory.
Because the spin operators can only affect coordinates in spin space, they
must commute with all operators that are a function only of spatial coordinates.
With this in mind, one is able to partition the total wave function between orbital
and spin contributions:
ctotal corbital cspin
(49)
where corbital is the wave function previous denoted as c (r, u, f) in Equation (10),
and cspin is the spin wave function (ca or cb). Incorporation of the spin operators
in the Hamiltonian operator ultimately yields atomic wave functions, which
48
Brittain
are characterized by four quantum numbers, namely n, l, ml, and ms. The
ml quantum number is the one previously termed m in the spin-free
(nonrelativistic) Schrodinger system but which is termed ml when electron
spin is included in the relativistic wave functions. A single electron can therefore
be considered as moving in four-dimensional space, which consists of three space
coordinates and one spin coordinate.
The Pauli Exclusion and Aufbau Principles, and Multielectron Atoms
It has now been established that a single electron moving in a spherically symmetrical, but non-Coulomb force field can be characterized by two quantum
numbers (n and l). The energy differences between states of different l and
equal n are ordinarily less than that between states with different n. The possible
states of an electron can therefore be divided into principal groups (or levels)
that differ from one another in their n values, and into subgroups (or sublevels)
of the same n but different l values. In the presence of a magnetic field, states
of given n and l, the states characterized by different values of ml, will not be
degenerate.
In an atom containing several electrons, the motion of each individual electron can be approximated as motion in a centrally symmetric but non-Coulomb
field of force. This field results from the overlapping of the Coulomb field of
the nucleus and the mean field of the other electrons. In this approximation, a
definite value of n and l can also be attributed to each electron in a complicated
atom. The approximation will be particularly good when considering a single
electron with large n, as is usually the case for most of the higher terms of an
atom. The action of the remaining electrons may then be approximated as
being due to their mean field. This approximation permits the characterization
of electrons as being either core electrons or as being valence electrons.
But when considering electrons having equal n and l, these must necessarily be roughly equidistant from the nucleus, and thus the assumption of a
mean field is relatively poor. The action of the other electrons on a given electron
will be strongly dependent on their momentary positions, hence the field in which
the given electron moves can no longer be considered as being centrally symmetric. This causes the n and l quantum numbers to lose an exactly definable
and physically interpretable meaning. Nevertheless, we shall see that it will
still be possible to obtain the number and type of the electronic states, and this
will permit major deductions to be made about spectroscopic transitions.
In order to understand the assembly of the periodic system, and the periodicity in the properties of elements and their energy level diagrams, a new assumption must be introduced. This is termed the Pauli exclusion principle, and its use
prevents the filling of various electron shells with an arbitrary number of
electrons. To formulate the principle conveniently, it is easy to imagine an
atom brought into a very strong magnetic field, which is so strong that each
individual li and si are space-quantized independently of one another in the
49
direction of the field. In other words, for each single electron, the components of
l in the direction of the field will take one of the values ml l, l 2 1, l 2 2, . . . ,
2l, and the components of s can take one of the values ms +12. The total
number of possible states, of course, will not be altered by the assumption of a
strong magnetic field.
The Pauli principle states that in one and the same atom, no two electrons
can have the same set of values for the four quantum numbers, n, l, ml, and ms. It
follows that only a limited number of electrons can have the same set of values
for the n and l quantum numbers. The Pauli principle does not result from the
fundamentals of quantum mechanics but is an assumption that fits very well
into quantum mechanics. It cannot be properly theoretically justified without
going into the Dirac theory. Accepting the validity of the Pauli principle,
however, permits the ready classification of electrons in an atom according to
a pattern that explains a considerable body of spectroscopic information.
The utility of the Pauli principle has been illustrated in Table 4, which list
out all of the possible states of an electron in an atom, and the divisions into
groups and subgroups, for shells up to n 3. On the basis of the Pauli principle,
only two electrons can be present in each cell defined by a specific set of (n, l,
and ml) quantum numbers, and then only when these have antiparallel spin
directions (different values of ms).
The ordering of states in Table 4 is roughly that of the relative energies
of the various states, hence according to the Pauli principle electrons would fill
the states according to the following order:
1s 2s 2p 3s 3p 3d 4s 4p . . .
(50)
The electronic configuration of an atom is built up by the filling of electrons into

the lowest possible sequence of levels that is permitted by the Pauli principle.
This concept is commonly referred to as the Aufbau, or Building-Up, principle.
Thus, the single electron of the hydrogen atom is placed into the 1s orbital of
the K shell (the shell nomenclature being that of Bohr), and one writes its electronic configurations as 1s 1. Similarly, the Pauli principle permits both electrons
of the helium atom to occupy the 1s orbital of the K shell, yielding the electronic
configuration of 1s 2. Of course, the two electrons of helium must exist in different
spin states so as not to violate the Pauli principle.
Owing to the restrictions on quantum numbers imposed by the Pauli principle, it is not possible to place any more electrons into the K shell. When any
shell in a multielectron atom contains the maximum number of electrons permitted by the Pauli principle, it is termed a closed shell or a full shell, and electrons in a full shell are often denoted as core electrons. For example, the 1s 2
configuration of helium represents the maximal filling of the K shell, and these
1s electrons will be core electrons for any atom containing additional electrons.
The outer electrons of an atom are often termed the valence electrons, because it
will be only these electrons that become involved in chemical bonding.
50
Brittain
Table 4
Possible States of an Electron in a Multielectron Atom
Shell
Orbital
type
1s
2s
Principal
quantum
number (n)
Azimuthal
quantum
number (l )
Magnetic
quantum
number (ml)
Spin
quantum
number (ms)
1
1
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
0
0
0
0
1
1
1
1
1
1
0
0
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
0
0
0
0
1
1
0
0
21
21
0
0
1
1
0
0
21
21
2
2
1
1
0
0
21
21
22
22
1/2
21/2
1/2
21/2
1/2
21/2
1/2
21/2
1/2
21/2
1/2
21/2
1/2
21/2
1/2
21/2
1/2
21/2
1/2
21/2
1/2
21/2
1/2
21/2
1/2
21/2
1/2
21/2
2p
3s
3p
3d
The ground state of an atom is the one in which all electrons are in the
lowest possible energy states. On the basis of the Pauli principle, this is the
state in which all the lower shells are filled as far as the Pauli principle allows.
We shall return to this point later, but excited states of an atom result when
one or more of its electrons are raised to any of the higher orbitals.
The configuration of an atom having a nuclear charge of Z 1 is obtained
by taking the ground state electronic configuration of the atom characterized by
the nuclear charge of Z and adding an additional electron to one of the shells not
yet filled. The electron configurations of the atoms belonging to the first four rows
of the periodic table are dictated by the pattern of quantum numbers given in
51
Table 5 Configurations of Multielectron Atoms Making Up the First Four Rows of the
Periodic Table
Z
Element
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
H
He
Li
Be
B
C
N
O
F
Ne
Na
Mg
Al
Si
P
S
Cl
Ar
Electron configuration
Element
1s 1
1s 2 ; [He]
[He] 2s 1
[He] 2s 2
[He] 2s 2 2p 1
[He] 2s 2 2p 2
[He] 2s 2 2p 3
[He] 2s 2 2p 4
[He] 2s 2 2p 5
[He] 2s 2 2p 6 ; [Ne]
[Ne] 3s 1
[Ne] 3s 2
[Ne] 3s 2 3p 1
[Ne] 3s 2 3p 2
[Ne] 3s 2 3p 3
[Ne] 3s 2 3p 4
[Ne] 3s 2 3p 5
[Ne] 3s 2 3p 6 ; [Ar]
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
K
Ca
Sc
Ti
V
Cr
Mn
Fe
Co
Ni
Cu
Zn
Ga
Ge
As
Se
Br
Kr
Electron configuration
[Ar] 4s1
[Ar] 4s2
[Ar] 3d 1 4s 2
[Ar] 3d 2 4s 2
[Ar] 3d 3 4s 2
[Ar] 3d 5 4s 1
[Ar] 3d 5 4s 2
[Ar] 3d 6 4s 2
[Ar] 3d 7 4s 2
[Ar] 3d 8 4s 2
[Ar] 3d 10 4s 1
[Ar] 3d 10 4s 2
[Ar] 3d 10 4s 2
[Ar] 3d 10 4s 2
[Ar] 3d 10 4s 2
[Ar] 3d 10 4s 2
[Ar] 3d 10 4s 2
[Ar] 3d 10 4s 2
4p 1
4p 2
4p 3
4p 4
4p 5
4p 6 ; [Kr]
Table 4 and are summarized in Table 5. It must be noted that interelectronic

interactions can cause the relative energies of the orbitals to vary slightly from
atom to atom, and the order can differ from the ideal sequence when two
orbitals lie close together in energy. For example, the ground state configuration
of chromium is [Ar] 3d 5 4s 1 rather than [Ar] 3d 4 4s 2 as a result of the favorable
situation of an exactly half-filled shell of d-electrons.
One will note that the electron configurations of the ground states
show periodicity, because after a certain number of electrons have been added,
the outermost electron will be of a type previously encountered. For
instance, all the alkali metals of Group IA contain a single outer electron
and exhibit electron configurations that can be written as [core] ns 1.
Similarly, all the alkaline earth metals of Group IIA contain two electrons
outside of the core and exhibit electron configurations that can be written as
[core] ns 2.
Spectroscopy of Core Electron States

To discuss the possible spectroscopic transitions, which can originate out of core
electron states, a reexamination of the energy levels of the hydrogen atom is
appropriate. Earlier, Equation (21) of chapter 1 was cited as the relation that
52
Brittain
fitted the emission spectrum of atomic hydrogen:

~ 1R 1 1
W
l
n21 n22
(51)
where n1 and n2 are each integers. Having completed the quantum mechanical
treatment of this system, we now see that the integers n1 and n2 are in fact the
principal quantum numbers of the initial and final states, respectively. This
permits calculation of transition energies and are collected in Table 6 for transitions originating out of the n 1 (1s) ground state of the hydrogen atom.
An examination of the pattern of transition energies and wavelengths of
Table 6 yields three important conclusions. The first of these is that the energy
required to effect the promotion of an electron out of the 1s level is exclusively
associated with the electromagnetic radiation within the X-ray region of the
spectrum. Second, when the principal quantum number of the terminal state is
fairly low (i.e., less than 7), the spectral transitions are separated by a sufficient
energy difference so that they remain discrete and are easily resolved. However,
when n exceeds 10, the close spacing of succeeding transitions requires such high
Table 6
Absorption Spectra of the Hydrogen Atom Originating from the

n 1 State
n of terminal
(state)
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
50
100
Energy (cm21)
)
Wavelength (A
Wavelength (nm)
82,258.4
97,491.4
102,822.9
105,290.7
106,631.2
107,439.5
107,964.1
108,323.8
108,581.0
108,771.4
108,916.1
109,028.8
109,118.2
109,190.3
109,249.4
109,298.3
109,339.3
109,374.0
109,403.6
109,633.9
109,666.8
1215.68
1025.73
972.55
949.75
937.81
930.76
926.23
923.16
920.97
919.36
918.14
917.19
916.44
915.83
915.34
914.93
914.58
914.29
914.05
912.13
911.85
121.57
102.57
97.25
94.98
93.78
93.08
92.62
92.32
92.10
91.94
91.81
91.72
91.64
91.58
91.53
91.49
91.46
91.43
91.40
91.21
91.19
53
resolution to detect individual spectral transitions that these eventually degrade

into a continuum of unresolvable spectroscopy. Finally, no energy levels
exist when the transition energy exceeds the value of the Rydberg constant
(109,677.8 cm21, equivalent to a wavelength of 91.18 nm), hence any electron
excited at a higher energy must be ionized out of the atom.
These conclusions permit the classification of core electron spectroscopy
into two broad divisions. The first entails the excitation of core electrons into
states, which are defined by the quantum numbers previously discussed and
is termed X-ray absorption spectroscopy, or XAS. The excited state produced
by XAS can undergo a multitude of secondary transitions, leading to a variety
of phenomena, which can be grouped together under the broad category of
atomic spectroscopy.
The second type of spectroscopy concerns transitions where the electron is
excited above the continuum of energy levels and is completely ionized. This
type of spectroscopy is termed X-ray photoelectron spectroscopy, where one
follows these processes through a measurement of the kinetic energy of the
ionized electron. The excited state can be deactivated through the emission of
electromagnetic radiation, a process known as X-ray fluorescence. When the
excited state is deactivated through nonradiative processes, but which results
in the ejection of electrons nonetheless, the phenomena constitute Auger electron
spectroscopy.
X-RAY ABSORPTION SPECTROSCOPY
To understand the specificity of the X-ray absorption spectra, it is first necessary
to consider the mechanism of the interaction of X-rays with a material. When an
X-ray beam passes through a specimen, the intensity of the rays becomes weaker
because of the dual processes of absorption and scattering. The mechanism of
X-ray absorption differs from that of optical absorption, in that the absorption
of X-ray energy occurs as a result of a single processionization of the atom
at the expense of inner electrons. The energy of the absorbed radiation is thus
transformed into kinetic energy of these displaced electrons (known as photoelectrons), in addition to the potential energy of the excited atom that equals
the bonding energy of the displaced electron.
The X-ray radiation characterized by the least energy (i.e., greatest
wavelength) causes the displacement of electrons from the outer shells. With
increasing energy of the excitation quanta, an ever smaller quantity is required
to remove electrons from a given shell, which is characterized by reduced
absorption. A continuous lessening of absorption continues for as long as the
emission energy is sufficient to force an electron out of the next, deeper-lying,
inner shell. This phenomenon gives rise to a sharp increase in absorption that
manifests itself in the form of an absorption edge.
Thus, the X-ray absorption spectra of materials are characterized by the
existence of three elements. These are a continuously rising absorption
54
Brittain
coefficient with increasing wavelength, an absorption edge, and the fine structure
of the absorption edges.
Most experimental work is conducted using synchrotron radiation,
although in principle there is no reason why conventional high-intensity X-ray
sources (i.e., rotating anode generators) could not be used. The absorption
spectroscopy experiment is conducted by inserting the sample into the X-ray
beam and measuring the incident and transmitted beam flux.
Another phenomenon that produces a weakening of X-ray intensity during
the passage through the substance is scattering that is known to occur in two
types. Coherent (or Thompson) scattering can take place as a result of the
collision of the X-ray photon (having energy equal to hn) by the electrons
(which have energies of Eel) of the atom. Should energy of the X-ray photons
be less than that of the electron bonding (hn , Eel), the photons cannot displace
the electron from a given inner shell. Following an elastic collision with fixed
electrons, the photons are scattered, hence their energy (and wavelength)
remain unchanged. On the other hand, incoherent (or Compton) scattering
results if the energy of the X-ray photons is greater than that of the binding electrons (hn . Eel), so that the photons force the electron out of the corresponding
inner shell. When the photoelectrons collide with electrons, they partition some
of their energy to them. This results in the scattering photons displaying a lesser
energy and a greater wavelength. Because the ejection of an electron is the first
condition for the emergence of all X-ray and electron spectra, incoherent
scattering must accompany the effect. Furthermore, because the atom has at
the same time more and less strongly bound core electrons, one can observe
two lines in the spectrum of scattered emission, those with unchanged and
changed (increased) wavelength.
The intensity of scattering increases with the atomic number of the
material, so that the more electrons present in an atom, the greater will be their
scattering effect. The X-rays are scattered poorly by light atoms and strongly
by heavy ones. The correlation of the coherent and incoherent scattering
depends on two factors, namely, the wavelengths of the incident radiation and
the atomic number of the absorbing atom. The smaller the wavelength of
the incident radiation (i.e., the higher its energy) and the lesser the atomic
number (i.e., the smaller the energy of the electrons bonding in these atoms),
the greater will be the incoherent scattering of the emission.
The quantitative estimation of the reduced intensity of the X-rays following
their passage through the substance is effected by using the attenuation factor
m, which is the sum of the coefficient of pure (photoelectric) absorption (t)
and the scattering (dispersion) factor (s). The attenuation factor is often
referred to as the coefficient of absorption. With wavelengths exceeding 0.5 A

and for elements with atomic numbers exceeding 26, the attenuation is effectively
caused entirely by absorption. In that case, one concludes that m t.
The coefficient of absorption (in units of cm21) is deduced from the usual
statement of Beers law. This is employed in evaluating the transparency or
55
opacity of the specimen with its given thickness (at a specific wavelength), but
because it depends on the state of the substance it is not a constant characterizing
the absorption of a given element.
More frequent use is made of a mass absorption coefficient, equal to the
absorption coefficient divided by the density of the substance. The mass absorption coefficient is independent of the state of the substance and of the chemical
compound to which the given element belongs, and it has a definite significance
for the given element at the given wavelength. It is a common practice to indicate
the wavelength for which the value of the mass absorption coefficient is cited
or to indicate the characteristic X-ray line whose radiation is used in measuring
the absorption. Values for the most important absorption coefficients of the
elements, measured at commonly used X-ray lines, have been published (11).
When viewed under conditions of low resolution, the X-ray absorption
spectrum appears as a continuously rising absorption with increasing wavelength
but which is superimposed with abrupt discontinuities. When the spectrum is
examined at higher resolution, one finds the existence of a fine structure made
up of distinctive elements. One element is the position of the absorption edge
itself. Another is the Kossel structure of the initial region of absorption, consisting of additional weaker maxima near the absorption edge and having energies up
to 20 30 eV at higher energies. Finally, one observes the Kroning extended fine
structure, continuing from the absorption edge up to higher energies as large as
400 eV and being evident in the form of weak absorption fluctuations. Because
substances in different phases will exhibit differences in their electronic structure,
the nature of the fine structure transitions is a function of the phase under study.
In atoms, the absorption edge corresponds to the first allowed transition to
an empty level. In the case of argon, for example, the electron configuration
(Table 5) is 1s 2 2s 2 2p 6 3s 2 3p 6, and the lowest unoccupied levels are the 4s,
4p, and 3d orbitals. Recalling the selection rule Dl +1 from Equation (46),
one finds that the K-edge of absorption corresponds to the transition to the first
vacant level, or 1s ! 4p. The next allowed transitions from the 1s (K) level
will be 1s ! 5p, 1s ! 6p, 1s ! 7p, and so on right out to the limit of the
p-level series. Superposition of these lines forms a fine structure of the argon
K-edge absorption.
The X-ray absorption spectrum can be divided into two basic regions, each
of which provides different information. One of these is near the absorption edge,
and is termed the XANES (X-ray absorption near edge structure) region. The
XANES region is known to be sensitive to the coordination symmetry and the
oxidation state of the target atom. The other is the EXAFS (extended X-ray
absorption fine structure) region, and its spectroscopy can be used to obtain
information on particle size and composition, coordination numbers and distances, and interparticulate interactions.
Owing to the favorable absorption coefficients of heavy atoms, XANES
and EXAFS studies are usually conducted on systems containing at least one
atom having an atomic number exceeding 26. As an example, the K-edge
56
Brittain
absorption in metallic nickel (for which the binding energy of the Ni K-electron
is 8332.8 eV) is plotted against the energy of the ejected photoelectron in
Figure 2A. The EXAFS spectrum extracted from this spectrum is found in
Figure 2B, plotted in the usual manner of EXAFS absorption function as a
function of the photoelectron wave vector.
The X-ray absorption spectrophotometric techniques have found extensive
use in the characterization of all systems that contain heavy atoms, such as
homogeneous and heterogeneous catalysts (12). Not surprisingly, this methodology has proved to be extremely useful in the characterization of small-molecule
inorganic compounds and extraordinarily informative in the study of metalloproteins. The utility of this approach will be outlined through the description
of appropriate selected illustrative examples.
Figure 2 (A) X-ray absorption spectrum and (B) corresponding EXAFS spectrum for
metallic nickel. Abbreviation: EXAFS, extended X-ray absorption fine structure.
57
XAS Studies of Inorganic Compounds

Different EXAFS and XANES in the K-edge of Mn(III) bound to dodecaphenylporphyrin complexes were used to study the activation mechanism induced
upon reaction with ozone (13). It was deduced that the coordination environment
of the Mn(III) ion consisted of the four equatorial pyrrolic nitrogen atoms of the
porphyrin ligand plus two ozone molecules, these being arranged symmetrically
with respect to the porphyrin core. The two Mn oxygen bonds were found to be
and 2.49 + 0.05 A
.
nonequivalent, exhibiting bond lengths of 2.09 + 0.02 A
Interestingly, only one of these bonds was affected upon this formation of the
catalytically active species.
Both K-edge and L-edge absorption spectroscopies were used to study the
light-induced excited spin state trapping of bis(phenanthroline)-bis(thiocyanate)Fe(II) (14). The spectra obtained for low- and high-spin forms exhibited trend
discontinuities that paralleled trends in other structural work, indicating the
utility of the spectroscopic study. In another study of this type, XAS was used
in conjunction with other techniques to study the spin crossover phenomena
associated with other Fe(II) complexes (15). It was possible to obtain values
for changes in Fe N bond distances, which were caused by changes in the
spin state of the iron atom in the complexes.
A number of Fe(III) complexes with various saccharide ligands were
characterized using K-edge XAS techniques, together with a number of other
methodologies (16). A variety of mononuclear, dinuclear, and trinuclear complexes were obtained, and in some cases mixed ligand complexes were obtained.
The XANES studies of these complexes exhibited a pre-edge structure possessing
1s-3d transition structure, which is indicative of the existence of octahedral
Fe(III) centers. The EXAFS studies delineated the binding of the saccharide
ligands through the pattern of Fe22O bonds and also indicated the existence of
weaker interactions between the central Fe(III) ion and carbon atoms in the
ligands.
The K-edge XANES spectra were obtained for two polymorphs of
bis(1,2-dione-dioximato) Ni(II), and the compound dissolved in pyridine (17).
In the crystalline materials, the XANES features consisted mainly of the
1s ! 4p transition of Ni(II), with the features around 8335 8350 eV being
significantly affected by the interaction between the central Ni(II) ion and the
atoms of adjacent molecules in the stacking structures. In pyridine solution,
the molecules were found to stack above each other, with slight deviations
from the central axis being noted.
The XAS studies were performed to study the Cu(I) complexes of crown
ether appended bis([2-pyridyl]-ethyl)amines and their oxygen adducts (18).
Following a crystallographic study of a model compound, the contribution of
pyridine ring atoms to the EXAFS was simulated. Oxygenation of the complexes
resulted in a large increase of the intensity of the major peak in the phasecorrected Fourier transform, which was interpreted as evidence for a peroxo
58
Brittain
coordination mode. This in turn caused a valency change from Cu(I) to Cu(II), as
judged from the edge position.
One compound of pharmaceutical interest, which has been studied by XAS
is zinc stearate (19). It was concluded in this work that the Zn(II) ion contains
four oxygen atoms in its inner coordination sphere and that the Zn22O distances
. Furthermore, the carboxylate groups were found to bind in a
were 1.95 A
bridging bidentate mode. Zinc stearate was observed to melt at 1308C, but the
coordination structure of the Zn(II) ions was found to be maintained in the
molten state. Examples of the Zn K-edge EXAFS spectra, and the Fourier
transforms of these, at various temperatures are found in Figure 3. The Zn22O
, and
bond lengths at 258C, 22038C, and 1708C were all found to be 1.95 A
the coordination number of the central Zn(II) ion was deduced to be 4.1 + 0.1
at all temperatures.
Figure 3 (A) Zn K-edge EXAFS spectra of zinc stearate at 258C (thick curve), 22038C
(thin curve), and 1708C (dotted curve), and (B) the Fourier transforms of these. Abbreviation: EXAFS, extended X-ray absorption fine struction. Source: Adapted from Ref. 19.
59
The XAS Studies of Metallo-Proteins

Without a doubt, the most powerful method available to establish the threedimensional structure of a chemical species is that of single-crystal X-ray diffraction. However, this technique cannot be used when the substance in question
cannot be crystallized. This difficulty is often encountered during the study of
peptides and protein systems, and as a result a large number of investigations
have been conducted where structural information has been obtained from the
conduct of the XAS work. The ability of this technique to provide information
on the immediate coordinative environments of metal ions has enabled
workers to obtain important information about such binding sites in peptides
and proteins.
The solution structures of the binuclear Mn(II) centers in arginase, Mncatalase, and the Mn(II) substituted forms of selected iron enzymes (ribonucleotide reductase and hemerythrin) have been determined using XAS (20). The
spectra obtained on model Mn(II) compounds showed an inverse correlation
between the XANES peak maximum and the root-mean-square (RMS) deviation
in metal ligand bond lengths. Arginase and Mn-substituted ribonucleotide
reductase protein systems were found to have symmetric nearest neighbor
environments characterized by a low RMS deviation in bond length, whereas
Mn catalase and Mn-substituted hemerythrin appeared to have a larger bond
length deviation. The 1s ! 3d areas for arginase and Mn-substituted ribonucleotide reductase were consistent with six coordinate Mn, whereas the 1s ! 3d
areas for Mn catalase and Mn-substituted hemerythrin were larger, suggesting
that one or both of the Mn(II) ions were five-coordinate in these proteins. The
EXAFS spectra for Mn-substituted ribonucleotide reductase and arginase were
nearly identical, with symmetric Mn nearest neighbor environments and
outer-shell scattering being consistent with a lower limit of one histidine per
Mn(II) core. In contrast, the EXAFS data for Mn catalase and Mn-substituted
hemerythrin showed two distinct Mn nearest neighbor shells and outer-shell
carbon scattering consistent with a lower limit of approximately two to three histidine residues per Mn(II) core.
The XAS in the Mn(II) L-edge was used to study the metal ion binding sites
of Mn-catalase from Lactobacillus plantarum and the coordination chemistry of
mixedvalence Mn(II) complexes that served as model compounds (21). Both
reduced and superoxidized Mn-catalase were examined by fluorescence-detected
soft XAS, and it was learned that the corresponding Mn(II) L-edge spectra were
dramatically different. The spectrum of reduced Mn(II)-Mn(II)-catalase was
interpreted using ligand field atomic multiplet calculations and by comparison
to model compound spectra. For the interpretation of mixed valence Mn(II)
spectra, an empirical simulation procedure based on the addition of homovalent
model compound spectra was developed and tested on a variety of Mn(II)
complexes and superoxidized Mn-catalase.
60
Brittain
The Oxygen-Evolving Complex (OCE) of photosystem II catalyzes the

four-electron oxidation of two water molecules to dioxygen. The importance
of this system is that it is the source of most of the molecular oxygen in the atmosphere and provides the reducing equivalents needed for the reduction of carbon
dioxide to glucose. The XAS has proven to be of extreme value in the study of
this system, because the activity appears to center around the inorganic cofactors
required for oxygen evolution (four Mn(II) ions, Ca(II), and chloride). In one
study, EXAFS spectroscopy was used to characterize the local structural environment of Mn(II) in the resting state and in two different reduced derivatives of
the photosynthetic OEC complex (22). The same group has reported on the use
of EXAFS for the characterization of the structural consequences accompanying
Ca(II) replacement in the reaction center complex of the OEC complex (23).
Others have used XANES and EXAFS to assess the degree of similarity
between the OEC in photosystem II and a series of synthetic manganese complexes containing the distorted cubane fMn4O3Xg core (where X benzoate,
acetate, methoxide, hydroxide, azide, fluoride, chloride, or bromide) (24). Most
recently, edge spectroscopy and EXAFS was used to study the structural
change of the Mn(II) cluster during the onset of substrate/water oxidation by
the system (25). It was proposed in this work that the changes in Mn Mn distances observed during the transition were the result of ligand or water oxidation,
leading to the formation of an oxyl radical intermediate at a bridging or terminal
position. The reaction of the oxyl radical with hydroxide ion, water, or an oxo
group during the subsequent conversion was proposed to lead to the formation
of the O22O bond.
The iron L-edge X-ray absorption spectra of deoxymyoglobin (deoxy-Mb),
oxymyoglobin (Mb-O2), carbonmonoxymyoglobin (Mb-CO), and some photoproducts of these (MB -CO and Mb -O2) were obtained and compared to simulated spectra derived from a ligand field multiplet calculation (26). The analysis
indicated that both Mb-CO and Mb-O2 were low spin contradicting other work
that suggested Mb-O2 to be of intermediate spin. These workers summarized
the advantages of L-edge XAS (relative to the more energetic K-edge spectroscopy) as being a direct metal-selective probe for metalloproteins,
characterized by intense electric dipole allowed transitions, capable of yielding
higher resolution in the structural details, and having energy levels accessible
to d-orbital calculations.
The iron K-edge X-ray absorption spectrum of Rhodococcus sp. R312
nitrite hydratase (in frozen solutions at pH 7 and 9) has been analyzed to
determine details of the iron site coordination (27). The EXAFS analysis
permitted the deduction of two or three sulfur ligands per iron, and an overall
six-fold coordination sphere implies an N3S2O ligation sphere. The bond
lengths obtained from the EXAFS analysis support cis-coordination of two
cysteine ligands and conclusively rule out nitric oxide coordination to the iron.
Iron K-edge spectroscopy was used to study the binding of nitric oxide to iron
in inactive nitrile hydratase, confirming a 1:1 stoichiometry of nitric oxide
61
bound to low-spin Fe(III) (28). Studies were also performed on four iron complexes of various pentadentate ligands that served as model compounds for the
metal-ion binding site in the protein system, enabling a better modeling of the
metal-ion coordination environment.
The binding of nitric oxide to ferric nitric oxide reductase from Fusarium
oxysporum is studied using IR, resonance Raman, and XAS (29). Trends in
vibrational frequencies indicated that NO bound to Fe(III) more strongly in the
parent enzyme than it did with the d-camphor-bound form of Pseudomonas
putida camphor hydroxylase cytochrome (a typical model of the monooxygenase). Support for this conclusion was provided from an EXAFS study, which
for the Fusarium oxysporum
yielded a Fe22NO bond distance of 1.66 + 0.02 A
ferric nitric oxide reductase, and 1.76 + 0.02 A for the d-camphor-bound form of
Pseudomonas putida camphor hydroxylase cytochrome.
The EXAFS studies of the heme in human cystathionine b-synthase have
been performed in order to assign the axial ligands provided by the protein
(30). The ray absorption data revealed that iron in ferric cystathionine b-synthase
is six-coordinate, and the scattering intensities were judged consistent with the
presence of five nitrogen donors and one sulfur ligand. All of the data support
this assignment of the axial ligands as cysteine and imidazole.
A key step in the catalytic cycle of coenzyme B12-dependent enzymes is the
homolysis of the organometallic bond of the cofactor, leading to the formation of
a 50 -deoxyadenosyl radical. For the adenosylcobalamin-dependent enzyme
methylmalonyl CoA mutase (MCM), it has been suggested that this step is
mediated by a protein-induced lengthening of the cofactor axial cobaltnitrogen
bond. An EXAFS study indicated a normal axial cobaltnitrogen bond of the
cofactor in the related coenzyme B12-dependent enzymes glutamate mutase
from Clostridium cochlearium and 2-methyleneglutarate mutase from Clostridium barkeri. The XANES part of the various spectra were similar to each
other but deviated significantly from the corresponding spectra of aquocob(III)alamin and for cob(II)alamin. In addition, the spectra contained a pronounced
pre-edge peak, indicating the presence of a covalently attached sixth carbon
ligand to the cobalt center (31).
Structural information has been obtained from the analysis of Ni(II) K-edge
absorption spectroscopy of [Ni22Fe] hydrogenases from a number of sources
(32). It was reported that the nickel sites of all but the oxidized sample of Alaligenes eutrophus hydrogenase were quite similar, with the nickel K-edge energies
shifting 0.9 1.5 eV to lower energy upon reduction from oxidized to fully
reduced forms. Analysis of the XANES features assigned to 1s ! 3d electronic
transitions indicates that the shift in energy, which accompanies a reduction of
the enzymes, may be attributed at least in part to an increase in the coordination
number from five to six. With the exception of the oxidized sample of Alaligenes
eutrophus hydrogenase, the EXAFS data were dominated by the scattering from
. The data obtained from this hydrogenase
S-donor ligands at approximately 2.2 A
were unique in that they indicated a significant structural change to occur upon
62
Brittain
reduction of the enzyme. Data obtained from the oxidized enzyme indicate that
,
the EXAFS is dominated by scattering from 3-4 N,O donor atoms at 2.06(2) A
with contributions from 2-3 S-donor ligands at 2.35(2) A. This coordination

shell changes upon reduction to a more typical nickel site composed of four
. The results of the EXAFS
S-donor ligands at a Ni S distance of 2.19(2) A
analysis was shown to be in general agreement with published crystal structures.
Nickel L-edge absorption spectroscopy is used to study the electronic structure of Ni(II) in several Ni22Fe hydrogenases under a variety of conditions (33).
The L-edge spectra were interpreted by ligand field multiplet simulations and by
the comparison with data for Ni(II) model complexes, and evidence was obtained
for high-spin Ni(II) in the active enzyme. The heterogeneous Ni(II) sites in
carbon monoxide dehydrogenases derived from Clostridium thermoaceticum
and Rhodospirillum rubrum have been characterized using nickel L-edge spectroscopy (34). The data indicated that most of the Ni(II) in the isolated
enzymes is low-spin, but upon treatment with CO, a fraction of the Ni(II) was
converted either to high-spin NI(II) and/or to Ni(I).
The zinc site in tRNA-guanine transglycosylase derived from Escherichia
coli has been studied by zinc K-edge absorption spectroscopy (35). The EXAFS
data were most consistent with a tetracoordinate zinc atom being bound to one
nitrogen and three sulfur ligands. It was concluded in this work that careful analysis of EXAFS data could reveal subtle conformational changes in metal binding
sites, which were not observable using more conventional spectroscopic
methods. The zinc K-edge X-ray absorption spectra of the ferric uptake regulation protein isolated from Escherichia coli have been obtained in frozen
solution to determine details of the zinc coordination (36). The EXAFS spectra
of the apo-protein and of the Co(II)-substituted protein indicated the existence
of a tetrahedral environment for the zinc atom, with two sulfur donor ligands
and two N/O donor
(from cysteine residues) located at a distance of 2.3 A
. The XANES of the two forms of the protein were sufficiently
ligands at 2.0 A
different so as to suggest a change of the conformation of the zinc site upon
incorporation of cobalt.
Zinc is essential for thousands of proteins in organisms, but despite the
undisputed need for information regarding local zinc-site structure, the majority
of these environments remain uncharacterized. This is primarily due to the difficulty of studying the diamagnetic Zn(II) ion with conventional spectroscopic
probes. In many cases, the coordinative environment of Zn(II) in proteins has
been inferred on the basis of sequence similarity with one of the crystallographically characterized Zn proteins. Unfortunately, many zinc sites do not show
significant sequence similarity to crystallographically characterized Zn proteins,
and thus sequence alignment alone is often not sufficient to give the correct Zn
coordination sphere. To address this problem, X-ray absorption fine structure
spectra have been measured for a series of structurally characterized zinc
model complexes that mimic the zinc sites found in metalloproteins (37).
Because Zn22N and Zn22S EXAFS oscillations are nearly out of phase over
63
the accessible energy range, it is difficult to determine the relative number of scattering sulfurs and nitrogens. The authors described a protocol that can be used to
obtain an accurate quantitation of the number of low-Z (S) ligands. It was shown
that two of the spectroscopic variables (the scale factor and shift in the threshold
energy) ordinarily treated as freely adjustable parameters could lead to erroneous
results if not carefully controlled.
The molybdenum enzymes comprise a group having a wide range of functionalities, catalyzing a variety of two-electron redox reactions coupled to the
transfer of an oxygen atom. All of the structural information obtained on the
molybdenum site indicates that the enzymes contain a novel pterin-molybdenum
cofactor, in which the molybdenum appears bound by a dithiolene side chain
of the pterin ring. A number of molybdenum complexes, which could serve
as model compounds for the enzymes, have been characterized by XAS. In
one such study, molybdenum K-edge spectra and EXAFS analyses were used
to measure shifts of edge energies over the M(IV,V,VI) oxidation states, and
features in the second-derivative edge spectra were correlated with the number
of bound oxo ligands (38). In another work, the X-ray absorption spectra at the
molybdenum and selenium K-edges and the tungsten L2,3-edges were acquired
for a set of 14 Mo(IV) and W(IV,VI) bis(dithiolene) complexes related to the
active sites of molybdo- and tungstoenzymes (39). The edge shifts of these
complexes were correlated with ligand electronegativities, and terminal ligand
binding was clearly distinguished in the presence of four methyl-dithiolene
interactions.
Sulfite oxidase is an oxo-transferase, which is responsible for the physiologically vital oxidation of sulfite to sulfate. The XAS at the molybdenum and
sulfur K-edges was used to probe the active site of wild-type and cysteine
207serine mutant human sulfite oxidases (40). The active site structure in
, and three Mo22S
the wild-type enzyme consisted of Mo55O ligands at 1.71 A
. The mutant molybdenum site consisted of a trioxo site characligands at 2.41 A
, and two Mo22S ligands at 2.47 A
. The
terized by Mo55O bond lengths of 1.74 A
X-ray absorption spectra were measured at the sulfur K-edge, chlorine K-edge,
and Mo L3- and L2-edges for several model compounds of sulfite oxidase to
investigate ligand metal covalency and its effects on oxo transfer reactivity
(41). It was concluded that anisotropic covalency of the Mo cysteine bond in
sulfate oxidase could promote preferential transfer of one oxo group during the
performance of its catalytic function.
In another noteworthy study, XAS at the molybdenum K-edge is used to
probe the molybdenum coordination of dimethyl sulfoxide reductase obtained
from Rhodobacter sphaeroides (42). The molybdenum site of the oxidized
), with
protein possessed a novel Mo(VI) mono-oxo site (Mo55O at 1.68 A
and
additional coordination by approximately four thiolate ligands at 2.44 A
. The reduced Mo(IV) form of the
either an oxygen or nitrogen ligand at 1.92 A
enzyme is a des-oxomolybdenum, coordinated with 3-4 thiolates at 2.33 A

and 1.92 A
. The stable Mo(V)
and two different Mo22O/N ligands at 2.16 A
64
Brittain
glycerol-inhibited species was found to be a des-oxomolybdenum, with approxi and two similarly coordinated oxygen or
mately four thiolate ligands at 2.40 A
nitrogen ligands at 1.96 A.

The X-ray absorption spectroscopy at the molybdenum and selenium
K-edges is used to probe the active site structure of formate dehydrogenase H
obtained from Escherichia coli (43). The selenium K-edge EXAFS not only
agreed with the molybdenum K-edge data but also indicated the unexpected
. This finding led to the sugpresence of Se22S ligation at a bond length of 2.19 A
gestion that the active site of Escherichia coli formate dehydrogenase H contains
a novel selenosulfide ligand to molybdenum, where the selenium and sulfur
originated from selenocysteine and one of the pterin-cofactor dithiolenes.
Not all of the systems studied by XAS have involved edge studies of metal
ions, and some very interesting work have been performed using i-edge XAS of
the bound ligands (44). The sulfur K-edge EXAFS studies of cadmium-, zinc-,
copper-, and silver-rabbit liver metallothioneins are used to obtain the first
direct evidence for mixtures of bridging and terminal sulfur ligands (45). The
XAS at the sulfur K-edge is applied to a series of mononuclear iron-sulfur complexes to determine covalency, and its distribution over d-orbitals split by the
ligand field (46). It was found that the covalency decreased in the proteins relative
to the models, indicating that the thiolate22Fe(III) bond was highly covalent and
that a modulation of covalency in the proteins could contribute to the redox
properties of the active site.
Investigation of the electronic structure of 2Fe-2S model complexes and
the Rieske protein is performed using sulfur K-edge XAS (47). The degree of
sulfide-iron covalency in the [Fe2S2(SR)4]22 complexes was higher than the
degree of thiolate-iron covalency, indicating extensive charge donation of the
bridging sulfides. It was determined that the degree of thiolate covalency of
the Fe(III) center is the same in both the oxidized and reduced Rieske clusters
and similar to that of the [Fe2S2(SR)4]22 model complexes.
The sulfur K-edge XAS was also used to investigate the distribution of
sulfur types in whole blood cell samples, in selected subcellular blood fractions,
and in cell-free plasma from the tunicate Ascidia ceratodes (48). The sulfur
K-edge XAS spectrum of washed blood cell membranes revealed traces of
sulfate and low-valence sulfur but no sulfate ester or sulfonate. It was concluded
that the sulfur species contained within the blood cells was almost exclusively
cytosolic.
REFERENCES
1. Atkins PW. Molecular Quantum Mechanics. Oxford: Clarendon Press, 1970.
1944.
1981.
65

6. Pauling L, Wilson EB. Introduction to Quantum Mechanics. New York:
McGraw-Hill, 1935.
7. Davis JC. Advanced Physical Chemistry. Appendix D New York: Ronald Press,
1965:596 600.
8. Goudsmit GE, Uhlenbeck S. Naturwissenshaften. 1925; 13:953.
9. Bichowsky R, Urey HC. Proc Nat Acad Sci 1926; 12:80.
10. Dirac PAM. Proc Roy Soc 1928; A117: 610; Proc Roy Soc 1928; A118:351.
11. Birks LS X-Ray Spectrochemical Analysis. New York: Interscience, 1969.
12. Evans J. Chem Soc Rev 1997; 26:11.
13. Gotte V, Goulon J, Goulon-Ginet C, Rogalev A, Natoli CR, Perie K, Barge J-M,
Guilard R. J Phys Chem B 2000; 104:1927.
14. Lee JJ, Sheu H-S, Lee C-R, Chen J-M, Lee J-F, Wang C-C, Huang C-H, Wang Y.
J Am Chem Soc 2003; 122:5742.
15. Real JA, Castro I, Bousseksou A, Verdaguer M, Burriel R, Castro M, Linares J,
Varret F. Inorg Chem 1997; 36:455.
16. Rao PP, Geetha K, Raghavan MSS, Sreedhara A, Tokunaga K, Yamaguchi T,
Jadhav V, Ganesh KN, Krishnamoorthy T, Ramaiah KVA, Bhattacharyya RK.
Inorg Chim Acta 2000; 297:373.
17. Matsuo S, Yamaguchi T, Wakita H. J Phys Chem B 2000; 104:3471.
18. Feiters MC, Gebbink RJMK, Sole VA, Nolting H-F, Karlin KD, Nolte JM. Inorg
Chem 1999; 38:6171.
19. Ishioka T, Maeda K, Watanabe I, Kawauchi S, Harada M. Spectrochim Acta 2000;
A56:1731.
20. Stemmler TL, Sossong TM, Goldstein JI, Ash DE, Elgren TE, Kurtz DM,
Penner-Hahn, JE. Biochem 1997; 36:9847.
21. Grush MM, Chen J, Stemmier TL, George SJ, Ralston CY, Stibrany RT, Gelasco A,
Christou G, Gorun SM, Penner-Hahn JE, Cramer SP. J Am Chem Soc 1996; 118:65.
22. Riggs-Gelasco PJ, Mei R, Yocum CF, Penner-Hahn JE. J Am Chem Soc 1996;
118:2387.
23. Riggs-Gelasco PJ, Mei R, Ghanotakis DF, Yocum CF, Penner-Hahn JE. J Am Soc
1996; 118:2400.
24. Cinco, RM, Rompel A, Visser H, Aromi G, Christou G, Sauer K, Klein MP,
Yachandra VK. Inorg Chem 1999; 38:5988.
25. Liang W, Roelofs TA, Cinco RM, Rompel A, Latimer MJ, Yu WO, Sauer K, Klein
MP, Yachandra VK. J Am Chem Soc 2000; 122:3399.
26. Wang H, Peng G, Miller LM, Scheuring EM, George SJ, Chance MR, Cramer SP.
J Am Chem Soc 1997; 119:4921.
27. Scarrow RC, Brennan BA, Cummings JG, Jin H, Duong DJ, Kindt JT, Nelson MJ.
Biochem 1996; 35:10078.
28. Scarrow RC, Strickler BS, Ellison JJ, Shoner SC, Kovacs JA, Cummings JG,
Nelson MJ. J Am Chem Soc 1998; 120:9237.
29. Obayashi E, Tsukamoto K, Adachi S-I, Takahashi S, Nomura M, Iizuka T, Shoun H,
Shiro Y. J Am Chem Soc 1997; 119:7807.
30. Ojha S, Hwang J, Kabil O, Penner-Hahn JE, Banerjee R. Biochem 2000; 39:10542.
31. Champloy F, Jogl G, Reitzer R, Buckel W, Bothe H, Beatrix B, Broeker G,
Michalowicz A, Meyer-Klaucke W, Kratky C. J Am Chem Soc 1999; 121:11780.
66
Brittain
32. Gu Z, Dong J, Allan CB, Choudhury SB, Franco R, Moura JJG, Moura I, LeGall J,
Przybyla AE, Roseboom W, Albracht SPJ, Axley MJ, Scott RA, Maroney MJ.
J Am Chem Soc 1996; 118:11155.
33. Wang H, Raiston CY, Patil DS, Jones RM, Gu W, Verhagen M, Adams M, Ge P,
Riordan C, Marganian CA, Mascharak P, Kovacs J, Miller CG, Collins TJ, Brooker S,
Croucher PD, Wang K, Stiefel EI, Cramer SP. J Am Chem Soc 2000; 122:10544.
34. Ralston CY, Wang H, Ragsdale SW, Kumar M, Spangler NJ, Ludden PW, Gu W,
Jones RM, Patil DS, Cramer SP. J Am Chem Soc 2000; 122:10553.
35. Garcia GA, Tiemey DL, Chong S, Clark K, Penner-Hahn JE. Biochemistry 1996;
35:3133.
36. Jacquamet L, Aberdam D, Adrait A, Hazemann J-L, Latour J-M, Michaud-Soret I.
Biochem 1998; 37:2564.
37. Clark-Baldwin K, Tierney DL, Govindaswamy N, Gruff ES, Kim C, Berg J, Koch SA,
James E. Penner-Hahn. J Am Chem Soc 1998; 120:8401.
38. Musgrave KB, Donahue JP, Lorber C, Holm RH, Hedman B, Hodgson KO. J Am
Chem Soc 1999; 121:10297.
39. Musgrave KB, Lim BS, Sung K-M, Holm RH, Hedman B, Hodgson KO. Inorg Chem
2000; 39:5238.
40. George GN, Garrett RM, Prince RC, Rajagopalan KV. J Am Chem Soc 1996;
118:8588.
41. Izumi Y, Glaser T, Rose K, McMaster, J, Basu P, Enemark JH, Hedman B,
Hodgson KO, Solomon EI. J Am Chem Soc 1999; 121:10035.
42. George GN, Hilton J, Raiagopalan KV. J Am Chem Soc 1996; 118:1113.
43. George GN, Colangelo CM, Dong J, Scott RA, Khangulov SV, Gladyshev VN,
Stadtman TC. J Am Chem Soc 1998; 120:1267.
44. Glaser T, Hedman B, Hodgson KO, Solomon EI. Acc Chem Res 2000; 33:859.
45. Gui Z, Green AR, Kasrai M, Bancroft GM, Stillman MJ. Inorg Chem 1996; 35:6520.
46. Rose K, Shadle SE, Eidsness MK, Kurtz DM, Scott RA, Hedman B, Hodgson KO,
Solomon EI. J Am Chem Soc 1998; 120:10743.
47. Rose K, Shadle SE, Glaser T, de Vries, S, Cherepanov A, Canters GW, Hedman B,
Hodgson KO, Solomon EI. J Am Chem Soc 1999; 121:2353.
48. Frank P, Hedman B, Hodgson KO. Inorg Chem 1996; 38:260.
3
X-Ray Photoelectron and X-Ray
Fluorescence Spectroscopy
Harry G. Brittain
X-RAY EMISSION SPECTROSCOPIES

Up to this point, we have considered only the spectroscopy associated with transitions from states characterized by core electron wave functions to other states
that are equally well defined by different wave functions. This spectroscopy
description is appropriate as long as the quantum numbers of the terminal state
are relatively low in magnitude, but different phenomena become important
when the quantum numbers become sufficiently large. At high quantum
numbers, the energy spacings between higher excited states become successively
smaller until the states effectively merge into a continuum. In fact, when an
anharmonic potential energy term is used in the Hamiltonian operator, one
finds that at sufficiently high quantum numbers dissociation of an excited electron
may take place. This information is of great interest, because knowledge of the
energy required to dissociate an electron would be directly related to the
energy of the state from which it had been removed.
The ionization potential of an electron is defined as the energy required
for its removal from its orbital of occupancy. Originally, investigation of
the magnitude of ion currents obtained as a function of the energy of an
impacting species were used to deduce ionization potentials, but these studies generally yielded information only on a few electrons in the valence shells. However,
the ionization of electrons from atoms upon irradiation by X-rays is a phenomenon
known as the photoelectric effect, the process being expressed as:
A hn ! A e
(1)
67
68
Brittain
where hn is the energy of the impacting photon. One could study the characteristics
of the ionized atoms (as is done in the practice of mass spectrometry), or one may
study the characteristics of the ejected photoelectrons. For the latter, the Einstein
photoelectric law provides a relation that describes the partitioning of energy
during the photoionization process:
KE hn IP
(2)
where IP is the ionization potential of the state from which the electron is ejected,
and KE is the kinetic energy of the electron.
Ejected photoelectrons are separated according to their kinetic energies in
an analyzer system. The photoelectron spectrum is therefore a measure of the
number of electrons detected, and using Koopmans theorem one associates
the kinetic energy of the observed electrons (Eobs) with their characteristic
binding energies (EB):
Eobs hn EB
(3)
If the system undergoing photoionization consists of molecules, then the

observed energies will exhibit fine structure due to the additional possibilities
of vibrational or rotational excitation upon ionization. In that case:
Eobs hn EB EVIB EROT
(4)
If the resolution of the spectrometer is sufficient, then the individual

vibrational and/or rotational fine structure can be resolved and observed. If
not, then the observed photoelectron peak will consist of a broader band that
encompasses all of the unresolved fine structure.
From its introduction, electron spectroscopy has been attractive for the
study of atomic and molecular electronic structure of each band in the spectrum
corresponding to ionization from a single atomic or molecular orbital. In
addition, each occupied orbital having binding energy less than the excitation
energy will yield a single band in the spectrum.
Although there are a multitude of photoionization spectroscopies that
have been developed for the study of core electronic states, the present discourse
will center on two main techniques that have found significant use in the characterization of pharmaceutical solids. These have been termed x-ray photoelectron
spectroscopy (XPS) and x-ray fluorescence (XRF), and the relevant photophysics
associated with these is illustrated in Figure 1. In XPS, irradiation by X-rays
of energy hn is used to photoeject core electrons, whose kinetic energies are
directly determined. In XRF, one does not observe the photoejected electron
but instead monitors the fluorescence that is emitted by the system after one of
the valence electrons drops down into a core electron level to fill the vacancy
created by the initial photoionization. It is worth noting that the pathway of the
XRF process is independent of the primary photoionization step and can be
initiated by electron impact and also photon impact.
XPS and XRF Spectroscopy
69
Figure 1 Photophysics associated with X-ray photoelectron spectroscopy and X-ray fluorescence. As illustrated, in the XPS experiment one monitors the energy of the electron
ejected from the M-shell upon photoionization ( process 1). In the XRF experiment, one
monitors the fluorescence emitted either from the M-shell after photoionization ( process
2a) or from the L-shell after photoionization and radiationless decay ( process 2b).
X-RAY PHOTOELECTRON SPECTROSCOPY

Because the energy of the x-ray source used to effect the photoionization is
accurately known, the XPS experiment requires the determination of the
kinetic energies of ejected core photoelectrons to deduce their ionization potentials and hence the energy of their original binding energies (1). As is the case for
most spectroscopic methods, this process is easy to describe but not necessarily
easy to execute. Equation (3) is only appropriate to photoionization studies
performed in the gaseous state, and the equation applicable to the solid state is
somewhat more complicated:
Eobs hn EB W C
(5)
W is a quantity termed the work function, and is the energy required to

remove the already photoionized electron out of the solid-state matrix in which
it is located. One may envision this energy as the overcoming of all barriers
that are associated with interaction of the photoelectron with the material.
70
Brittain
For many substances, the work function is approximately 5 eV, and this is not
insignificant compared with the magnitude of typical binding energies. Unless
one chooses to measure the work function, XPS spectra are ordinarily presented
in terms of energy shifts with respect to a defined reference and not as absolute
binding energies. The surface charging effect, C, is another correction that
reflects the charge distribution in a solid, and the effect of atomic potentials on
core electron levels.
Siegbahn developed a model to refine the concept of binding energy,
attempting to account for the interaction of electron i with all nuclei in the
solid and all other electrons (1):
EB E0 k q Si=j q j =rij
(6)
where E0 is the binding energy of the electron associated with the isolated
atom, q is the valence charge, k is an empirical constant, and Si=j qj/rij is a
Madelung-type term reflecting the potential field of the environment. Sets of
charge-binding energy relations deduced from quantum mechanical calculations
are available (2).
Experimental Details
Only the briefest outline of the experimental methodology will be provided here,
because very excellent reviews of instrumentation (3,4) are available. In addition,
the ongoing annual reviews in Analytical Chemistry (5) provide another source of
information regarding the state of the art. The essential components of a typical
system consist of the X-ray source, an energy analyzer for the ejected photoelectrons, and a detector for the analyzed electrons, with the entire system
being enclosed in a high-vacuum chamber. A schematic diagram of the basic
XPS spectrometer system is given in Figure 2. The photoelectrons ejected
from the sample upon excitation by the source are drawn into the analyzer, separated on the basis of their kinetic energy, and their intensity as a function of
kinetic energy measured by a detector. The sample is mounted on a probe face
in such a way that the photoejected electrons can easily pass into the analyzer.
It is clear from equation (5) that the energy of the irradiation energy must
exceed the ionization potential of the electronic states to be studied, so any X-ray
source having sufficient energy can be used. The two most important sources are
the Ka lines of magnesium (1254 eV, line width 0.8 eV) and aluminum
(1487 eV, line width 0.9 eV), and for measurement of higher binding energies
one may use the Ka lines of titanium (4511 eV, line width 1.4 eV) or copper
(8048 eV, line width 2.5 eV). A discussion of the relative merits of different
sources is available (6). For very high-resolution work, the line width of the
X-ray source can be further reduced through the use of a monochromator.
The determination of the kinetic energy of ejected photoelectrons is the
most important aspect of the technique, and hence the analyzers used must be
simultaneously selective and sensitive. One analyzer class is the retarding
71
Figure 2 Basic components of the apparatus used for the measurement of X-ray photoelectron spectroscopy. X-rays from the source (A) are allowed to impinge on the sample
mounted on a probe (B), the energies of the ejected photoelectrons discriminated by the
electron analyzer (C), and finally measured by the electron detector (D). All of these components are contained in a high-vacuum chamber.
potential type, where photoelectrons are subjected to a variable retarding voltage

and the photoelectron spectrum results from differences in the relative transmission of electrons of different energies at a given retarding field. Examples
of retarding potential analyzers are the cylindrical grid and spherical grid analyzers. The other analyzer class is the dispersive type, where one obtains a spatial
resolution of photoelectrons on the basis of their differing kinetic energies.
Examples of dispersive analyzers are the cylindrical condenser and spherical condenser analyzers.
Once the ejected photoelectrons have been discriminated on the basis of
their kinetic energies, detection of these is usually effected using electron multipliers. These devices can vary significantly in their design, with the simplest
consisting of a plain charge collector (termed a Faraday cup). More typical
applications make use of electron multipliers having gains in the range of 106,
a factor that is necessary to obtain accurately measurable currents in the
microampere range.
Assignment of Observed XPS Bands and their Chemical Shifts
Because the binding energies associated with a given electron in a defined core
electron state is largely determined by the identity of the atom and the particular
set of n, l, and m quantum numbers defining the state, assignment of XPS peaks
can be readily achieved through the use of standard tables (1,7,8). Some of the
assignments made for core electron binding energies of elements in the free
atom state are found in Table 1, and Figure 3 illustrates the regular increase in
binding energies of various states with increasing atomic number. Thus,
72
Table 1
Element
H
He
Li
Be
B
C
N
O
F
Ne
Na
Mg
Al
Si
P
S
Cl
Ar
Brittain
Binding Energies (eV) of Core Electrons in Various Free Atom States
Electron
configuration
1s 1
1s 2 ; [He]
[He] 2s 1
[He] 2s 2
[He] 2s 2 2p 1
[He] 2s 2 2p 2
[He] 2s 2 2p 3
[He] 2s 2 2p 4
[He] 2s 2 2p 5
[He] 2s 2 2p 6 ; [Ne]
[Ne] 3s 1
[Ne] 3s 2
[Ne] 3s 2 3p 1
[Ne] 3s 2 3p 2
[Ne] 3s 2 3p 3
[Ne] 3s 2 3p 4
[Ne] 3s 2 3p 5
[Ne] 3s 2 3p 6 ; [Ar]
L2-shell
L3-shell
K-shell
L1-shell
(1s orbital) (2s orbital) (2p1/2 orbital) (2p3/2 orbital)
13.60
24.59
54.7
111.5
188
284.2
409.9
543.1
696.7
870.21
1070.8
1303.0
1559.0
1839
2145.5
2472
2822.0
3205.9
(5.4)
(9.3)
(12.9)
(16.6)
(37.3)
(41.6)
(37.9)
48.47
63.5
88.6
117.8
149.7
189
230.9
270
326.3
(8.3)
(11.3)
(14.5)
(13.6)
(17.4)
21.7
30.4
49.6
72.9
99.8
136
163.6
202
250.6
21.6
30.5
49.2
72.5
99.2
135
162.5
200
248.4
Note: Binding energies of some valence electron states have been listed for comparison purposes.
Source: Adapted From Refs. 4,5.
identification of a particular XPS peak with a state of known n, l, and m quantum

numbers is relatively straight-forward.
Once the peak system is identified, the real chemical information is derived
from the fine structure within each system. Because the valence electrons are intimately involved in chemical bonding and because they provide an electrostatic
shield for the core electrons, the interaction between the core electrons and the
valence electrons causes the energies of the core electrons to be affected by
details of their molecular environment. Any change in the valence electron
shield will be reflected in a change in the amount of energy required to
remove a core electron through that valence shell. For example, the beryllium
1S band in the oxide is shifted by 2.9 eV relative to that of the metallic
element, which is attributed to the difference in oxidation state between the
beryllium atom in its two states (9). In beryllium fluoride, the 1S band is
shifted further by 4.6 eV, which is relative to that of the metallic element,
reflecting the additional electronegativity of fluorine relative to that of oxygen.
For the organic compounds of pharmaceutical interest, the chemical shifts
of the 1S bands of carbon and oxygen would be of primary interest, and consideration of a few simple molecules will serve to illustrate the energy ranges.
Electronegativity is found to exert a strong influence on the carbon 1S binding
73
Figure 3 Binding energies of (A) K-shell electrons and (B) L-shell electrons as a
function of atomic number. Source: From Ref. 5.
energies, as illustrated by the series CH4 (290.7 eV), CH3OH (292.3 eV), CHF3
(298.8 eV), and CF4 (301.8 eV) (10). The two carbons in ethanol can be differentiated, with the carbon of the terminal methyl group exhibiting a 1S binding
energy of 290.9 eV and the methylene carbon directly bound to the hydroxyl
group exhibiting a binding energy of 292.3 eV. The chemical shift differentiation
in acetic acid is even more pronounced, where the carbon of the terminal methyl
group exhibits a 1S binding energy of 291.4 eV and the carboxylate carbon
exhibits a binding energy of 294.5 eV (10).
Applications of XPS to the Characterization of Solids
In its simplest application, XPS can be used as a complementary technique to
X-ray diffraction or solid-state nuclear magnetic resonance in the study of the
74
Brittain
properties of different compounds. For instance, the strong ionic [NH....N]

hydrogen bonds in a series of perisubstituted naphthalenes were found to strongly
influence the binding energies of the core electrons of the donor and acceptor
atoms (11). These authors were able to use the difference in binding energies
of the donor and acceptor as a means to evaluate the strength and asymmetry
of the strong hydrogen bonds. The properties of the [NH....N] hydrogen
bonds developed between 1,8-bis(dimethylamino)naphthalene with various
organic and inorganic acids were further studied in other publications (12,13).
XPS has been used to study the core electronic levels of triaminotrinitrobenzene and as a means to understand the processes associated with its
decomposition (14). The nitrogen and oxygen core spectra exhibited unusually
high-energy satellite structure, with the separation of the satellites from the
main core lines decreasing with the successive removal of donor amine
groups. The core level spectra of the component nitrogen and oxygen decreased
when the compound was subjected to isothermal and photolytic decomposition,
indicating the breakage of the C22NO2 bond.
The use of XPS as a means to yield quantitative elemental analysis through
the use of detail scans was used to characterize human immunoglobulins (15). In
this methodology, a small portion of the binding energy range is studies with a
higher degree of resolution to obtain better definition of the fine details in the
spectra, and thus minimizing the use of curve resolution techniques. This particular formulation consisted of IgG-dispersed monolithic matrices, prepared using a
nonbiodegradable polymer carrier. An initial burst of drug release was noted, and
the XPS studies showed that this effect might be associated with the surface concentration of IgG molecules. This and other findings suggested that the kinetics of
in vitro drug release were determined by the rate of IgG release through the
polymer matrix.
The incident X-ray radiation used as the source in XPS does not penetrate
far beyond the upper molecular layers of a solid sample, typically reaching to
beneath the surface. Consequently, it is not surprising
depths approximately 50A
that the technique has been used to study the surfaces of materials having
pharmaceutical interest. For instance, XPS was used to study the surface concentration of PEG 400 distearate contained in poly(D,L )-lactic acid microspheres
(16). In this study, the carbon 1S lines were found to contain more useful
information than did the oxygen 1S lines. As illustrated in Figure 4, as the
amount of PEG 400 distearate increased in the microspheres, the carbon lines
were found to exhibit a greater shift to higher energies.
XPS and time of flight secondary ion mass spectrometry were used to
show that the polypropylene oxide component of adsorbed poloxamers adhered
to the surface of polystyrene particles (17). Prior to the performance of this
work, it had been known that the adsorption of poloxamers altered the electrostatic
charge, adhesion behavior, and handling properties of such particles, and from the
XPS studies the authors were able to deduce that there was a polyethylene oxiderich outer surface that appeared to influence the charge alteration.
75
Figure 4 X-ray photoelectron spectra of poly(D,L )-lactic acid microspheres containing

PEG 400 distearate at levels of (A) 0%, (B) 1%, (C) 10%, and (D) 100%. Source:
From Ref. 16.
Poly(methyl methacrylate) is used to fill the gap between a prosthesis and

its surrounding bone in cemented arthroplasties; biocompatibility problems
related to bone cement application are known to limit the clinical success of
these devices. XPS was used to learn if there was a correlation between
surface changes and aging time (18). Evidence for the breaking of old bonds,
and the formation of new bonds, with the new bonding patterns are clearly
visible in oxygen 1S spectra of the hydroxyl groups. These effects were
interpreted as indicating the hydrolysis of PMMA ester groups, and the subsequent formation of hydrogen bonds between the hydroxyl and carbonyl
groups of adjacent polymer chains.
The changes in chemical composition after the surface modification and
coating of dextran-based ion-exchange microspheres were studied by XPS
(19), because with the incorporation of proper standardizations the technique
can be used to deduce elemental analyses of surface species. For instance,
after introduction of palmitoyl groups to sulfopropylated dextran microspheres,
the relative content of hydrocarbon carbon 1S response could be correlated
with concentration. Because the relative abundance of carbon associated with
carboxylic acid groups increased effectively from 0% to over 6%, this suggested
that the palmitoyl groups were linked to the microspheres through ester bonds.
76
Brittain
Given the utility of XPS as a technique for surface analysis, it is logical

that the method could be used to study the properties of coating films. For
instance, XPS has been used to study the surface chemical composition of a
range of poly(methacrylate) films that are commonly used as pharmaceutical
film coatings (20). Through the use of curve resolution methods, it was determined that the carbon 1S region consisted of four unresolved peaks (Fig. 5
representative example of a carbon 1S spectrum) derived from carbon atoms in
the polymer backbone and also in the ester side chains. Knowledge of chemical
shifts in model compounds enabled peak assignments, and these were used to
elucidate the differences in film character associated with Eudragits S, L, E,
and RS. In a similar application, XPS was used to develop a method that could
be used to quantitate the interfacial thickness of a film-tablet layer (21).
Another application of XPS that takes advantage of its responsiveness to
surface phenomena is in the study of molecular adsorption. In one study, XPD
was used to evaluate the absorbed state of glucose that had been sorbed onto
alumina (22). The alumina used in this work was a specific type (Mizusawa
Kagaku, type Neobead-P) containing a little percentage of sodium, which acts
as a basic site in the mutarotation of the adsorbed glucose. However, the XPS
data indicated that the aluminum-oxygen position of the surface served as the
Figure 5
Ref. 20.
High-resolution X-ray photoelectron spectra of Eudragit L. Source: From
77
adsorption sites for the glucose and that no preferential adsorption of glucose
onto the sodium sites took place.
XPS has been used to identify the functional states of carbon existing on the
surfaces of different activated charcoals (23). Because the adsorption on phenobarbital on such materials had been studied in the past, the system represented a
good model to evaluate the utility of XPS. A good correlation was found to exist
between the apparent areas occupied on the adsorbent surface per phenobarbital
molecule and the relative percentages of a C22O functional state, which was
determined to be the primary site involved in the binding.
Evaluating the ability of a pharmaceutical powder to be wetted is an
important criterion in understanding its dissolution properties, its ability to be
granulated, the adhesion of film coatings, and its tendency to aggregate when
placed in suspended form. In order to learn whether XPS could be used as a
method to study the wet ability of a powder, the binding energies of various
barbiturates in powdered and compressed form, and compared with wet ability
data obtained by other methods (24). There appeared to be no correlation,
however, between binding energies and surface wet ability. Although compaction
of the powders would change surface energetics, that process would not change
the chemical composition of the surface.
A capillary penetration technique was used to determine the wet ability of
morphine sulfate powders, and the surface chemistry of these powders studied
by XPS (25). Contact angles were shown to correlate with both the aspect ratio
of the component crystals and the atomic nitrogen-to-oxygen ratios at the
surface. It was concluded that the relative exposure of different crystal faces
(which could be evaluated by XPS) played an important role in controlling
the wet ability of the powdered morphine sulfate. In a similar study, XPS
was used to characterize the surface composition of poly(lactic acid) polymers
and poly(lactic/glycolic acid) copolymers, but the surface wet ability was
evaluated by other methods (26). The wet ability improvement of these surfaces
by pluronic surfactants was then studied and found to be a function of surfactant
composition.
X-RAY FLUORESCENCE SPECTROSCOPY

As illustrated in Figure 1, the electromagnetic radiation measured in an XRF
experiment is the result of one or more valence electrons filling the vacancy
created by an initial photoionization where a core electron was ejected upon
absorption of X-ray photons. The quantity of radiation from a certain level will
be dependent on the relative efficiency of the radiationless and radiative deactivation processes, with this relative efficiency being denoted at the fluorescent
yield. The fluorescent yield is defined as the number of X-ray photons emitted
within a given series divided by the number of vacancies formed in the associated
level within the same time period.
78
Brittain
As discussed in earlier sections, regardless of whether one uses the theory

of Bohr or the quantum theory, the quantized energy difference between the states
of a multielectron atom is given by:

Z 2 me4
1
1
DE 2 2 2 2
(7)
n1 n2
2n h
where Z is the atomic number of the element in question. In equation (7), one
identifies the levels having n 1 as being from the K shell, levels having
n 2 are from the L shell, levels having n 3 are from the M shell, and so
on. When an electron of the L shell is transferred into a vacancy in the K
shell, the energy released in the process equals the energy difference between
the states:

Z 2 me4
1
1
DEKa 2 2 2 2
(8)
1
2
2n h
This particular transition results in the emission of X-ray radiation known
as the Ka line. For transfer from the M shell into the K shell, the energy of the Kb
line is given by:

Z 2 me4
1
1
DEKb 2 2 2 2
(9)
1
3
2n h
Because both transitions described in equations (8) and (9) terminate in the
K shell, they are known to belong to the spectrum of the K series. Similarly, the
spectral lines that result when electrons fall into the L shell define the spectrum of
the L series, and the energies of these lines are given by:

me4
1
1
DEL 2 2 2 2
(10)
2
n
2n h
Equation (7) indicates that the energy difference between the two states
involved in a XRF transition is proportional to the atomic number of the
element in question, a fact realized some time ago by Mosley. The simple
equations given previously do not account for the different orbitals that exist
for each principal quantum number, and the inclusion of these levels into the
spectral considerations explain the fine structure observed within almost all
XRF bands. For instance, the L shell (defined by n 2) contains the 2S and
2P levels and the transfer of electrons from these levels down to a vacancy in
the K shell (defined by n 1) results in the emission of X-ray radiation
denoted as the Ka1 and the Ka2 lines.
A detailed summary of the characteristic XRF lines for the elements is
available (27), as are the more detailed discussions of the phenomena of X-ray
absorption and X-ray fluorescence (28 31).
79
Experimental Details
There are numerous types of instrumentation available for the measurement of
XRF, but most of these are based either on wavelength dispersive methodology
(typically referred to as WDX) or on the energy dispersive technique (typically
known as EDX). For a detailed comparison of the two approaches for XRF
measurement, the reader is referred to an excellent discussion by Jenkins (32).
The first XRF spectrometers employed the WDX methodology, which is
schematically illustrated in the upper half of Figure 6. The X-rays emanating from
the source are passed through a suitable filter or filters to remove any undesired
wavelengths and collimated into a beam that is used to irradiate the sample. For
instance, one typically uses a thin layer of elemental nickel to isolate the Ka lines
of a copper X-ray source from contamination by the Kb lines, because the K-edge
absorption of nickel will serve to pass the Ka radiation but not the Kb radiation.
The various atoms making up the sample emit their characteristic XRF at
an angle of u relative to the excitation beam, and the resulting X-rays are
Figure 6 Basic components of the apparatus used for the measurement of XRF by the
wavelength and energy dispersive methods. WDX: X-rays from the source (A) are
allowed to impinge on the sample (B); the resulting XRF is discriminated by the crystal
(C) and finally measured by the detector (D). EDX: X-rays from the source (A) are
allowed to impinge on the sample (B), and the resulting XRF is measured by the detector
(D). Abbreviations: WDX, wavelength dispersive XRF; EDX, energy dispersive XRF;
XRF, X-ray fluorescence.
80
Brittain
discriminated by a monochromator that uses a single crystal as a diffraction

grating. The diffraction off the crystal must obey Braggs law:
nl 2d sin (u )
(11)
where n is the diffraction order, l is the wavelength of X-rays undergoing diffraction, d is the distance between the planes of atoms acting as the diffraction
grating, and u is the angle of the excitation beam with the sample face. It is
clear from equation (11) that no single crystal can serve as an efficient diffraction
grating for all wavelengths of X-rays, and so WDX instruments typically contain
a variety of different crystals that the user can select to optimize the XRF region
of greatest interest.
The detector can be either a gas-filled tube detector or a scintillation
detector, which is systematically swept over the sample and which measures
the X-ray intensity as a function of the 2u scattering angle. Through suitable
calibration, each 2u angle is converted into a wavelength value for display.
The major drawback associated with WDX spectrometers is the reality
derived from the Bragg scattering law that one cannot measure all wavelengths in a XRF spectrum in a single scan. Hence, one might be required
to run multiple spectra if the range of elements to be studied is wide in
terms of atomic number.
The EDX methodology was developed as a means to permit an analyst
to acquire the entire XRF spectrum simultaneously, therefore eliminating
the requirement to acquire data in parcels. A simple schematic of an EDX spectrometer is illustrated in the lower half of Figure 6. As with the WDX instrumentation, X-rays emanating from the source are filtered and collimated into a beam
that is used to irradiate the sample. The XRF emitted by the sample is caused to
fall onto a semiconductor diode detector, and one obtains the spectrum through
the use of multichannel analysis of the detector output.
Although the instrumentation of the EDX method is simpler, it does not
provide the same degree of wavelength resolution associated with WDX
spectrometers. Jenkins (32) has provided an extremely useful set of criteria
that permit a judicious choice as to which particular technique would be most
appropriate for a given application:
1.
2.
3.
A WDX spectrometer capable of simultaneously recording a span of

wavelength information is most suitable for high-specimen throughput
quantitative analysis, where speed of the essence and the high initial
cost can be justified.
A sequential WDX spectrometer is most useful where time of analysis
is not critical and where a moderately high initial cost can be justified.
An EDX system is usually chosen when initial cost is a major consideration or where the analyses to be run will be qualitative or semiqualitative in nature.
81
One of the problems faced in XRF spectroscopy is the fact that the absolute
sensitivity of an element decreases with atomic number, and this decrease is most
considerable for light elements. For quite some time, XRF spectrometers could
not detect elements having atomic numbers less than 12, which did not permit
analysis of any second row elements, such as carbon, nitrogen, oxygen, and
fluorine. Fortunately, advances in detector design have been made, and now
through the use of ultrathin beryllium window analysts, one can now obtain
reliable results on the light elements. When EDX is combined with scanning
electron microscopy, one can even use windowless detectors to maximize the
response of the light elements (33).
Qualitative XRF Analysis
By virtue of their capacity to observe the entire range of X-ray fluorescence
and speed of data acquisition, EDX spectrometers are admirably suited for
qualitative analysis work. Equally well suited would be WDX spectrometers
capable of observing XRF over a wide range of scattering angles. As discussed
earlier and illustrated in Figure 3, the Ka and Kb XRF of a given element is
directly related to its atomic number, and extensive tables are available that
provide accurate energies for the fluorescence (27). Such information can be
stored in the analysis computer, and used in a peak-match mode to easily
identify the elemental origins of the fluorescing atoms within the sample. The
elemental identification is facilitated by the fact that XRF originates from core
electron levels that are only weakly affected by the details of chemical
bonding. It must be remembered, however, that the selectivity associated with
XRF analysis ends with elemental speciation and that the technique is generally
unable to distinguish between the same element contained in different chemical
compounds.
X-ray fluorescence spectra of most elements consist largely of bands
associated with the K, L, and M series. The XRF associated with the K-series
is dominated by the a1/a2 doublet, although some very weak b XRF can often
be observed at higher energies. The L-series XRF will consist of three main
groups of lines, associated with the a, b, and g structure. Ordinarily, the a1
line will be the strongest, and the b1 line will be the next most intense. The
peaks within the M-series are observed only for the heavier elements and
consist of an unresolved a1/a2 doublet as the strongest feature, followed by a
band of peaks associated with the b XRF.
For instance, the Ka1 and Ka2 lines of elemental copper are observed at
and 1.544A
, respectively, and Kb lines will be observed
wavelengths of 1.540A
and 1.381A
(33). The unresolved La1 and La2 lines are observed at a
at 1.392A
, and Lb lines are observed at 13.053A
and 12.094A
.
wavelength of 13.330A
The presence of copper in a sample would be indicated by the presence of
XRF peaks detected either at these wavelengths or at their corresponding
energies.
82
Brittain
XRF qualitative analysis entails the identification of each line in the

measured spectrum. The analysis begins with the assumption that the most
intense lines will be due to either Ka or La emission, and these are used to
match the observed lines with a given element. Once the most intense a-lines
are assigned, one then goes on to assign the b-lines, and these assignments serve to confirm those made from the analysis of the a-lines. Of course,
this task has been made quite easy through the use of computer systems
that store all the information and execute the peak matching analysis.
For example, the EDX analysis of a fumarate salt of a drug substance
was unexpectedly found to exhibit definite responses for chlorine, indicating
that the sample actually consisted of mixed fumarate and hydrochloride
salts (34).
In another qualitative study, EDX analysis was used to study the nature of
the precipitate occasionally formed in Zn insulin solutions (35). Identification of
the EDX peaks obtained for the crystalline precipitates enabled the deduction that
the solid consisted of a Zn insulin complex, and a rough analysis of the peak
intensities indicated that the composition of the precipitate was comparable
with that existing in the starting materials. The combination of the EDX technique with scanning electron microscopy enabled the analyses to be conducted
on relatively few numbers of extremely small particles.
Quantitative XRF Analysis
Owing to their superior fluorescent yield, heavy elements ordinarily yield considerably more intense XRF bands than the light elements. This feature can be
exploited to determine the concentration of inorganic species in a sample or
the concentration of a compound that contains a heavy element in some
matrix. Many potential XRF applications have never been developed owing to
the rise of atomic spectroscopic methods, particularly inductively coupled
plasma atomic emission spectrometry (36). Nevertheless, under the right set of
circumstances, XRF analysis can be profitably employed.
A number of experimental considerations must be addressed in order to
use XRF as a quantitative tool, and these have been discussed at length in a
number of texts (37 39). The effects on the usual analytical performance
parameters (accuracy, precision, linearity, limits of detection and quantitation,
and ruggedness) associated with the instrument are usually minimal. However,
the effects associated with sample handling, preparation, and presentation
cannot be ignored, as they have the potential to exert a major influence over
the quality of the analysis.
The background emission detected in a XRF spectrum is usually due to
scattering of the source radiation. Because the scattering intensity from a
sample is inversely proportional to the atomic number of the scattering atom,
it follows that background effects are more pronounced for samples consisting
largely of second row elements (i.e., organic molecules of pharmaceutical
83
interest). Hence, background correction routines play a major role in transforming raw XRF spectra into spectra suitable for quantitative analysis.
The performance of a quantitative XRF analysis requires that correction for
the interelement interactions that are associated with the matrix of the sample.
For homogeneous samples, matrix effects can serve to enhance or diminish the
response of a given element. The intensity retardation effects usually originate
from competition among scattering units in the sample for the incident X-ray
radiation or from the attenuation of emitted XRF by its passage through the
sample itself. Enhancement effects arise when emitted XRF is reabsorbed by
other atoms in the sample, causing those atoms to exhibit a higher XRF intensity
than would be associated with the incident X-ray radiation itself.
Because matrix and background effects are notoriously difficult to identify
and control, quantitative XRF analysis usually entails the use of standards. One
can write an expression relating the concentration of an analyte in a sample with
the XRF intensity of one of its emission lines:
Ci KIi MB
(12)
where Ci is the concentration of the analyte exhibiting the intensity response Ii, K
is a constant composed of instrumental factors, M is a constant composed of
matrix factors, and B is a constant composed of background factors. Although
it is conceivable that one could derive an equation describing how K would
be calculated, it is clear that calculation of the M or B constants would be
extraordinarily difficult.
However, when an internal standard containing an element of known
concentration CS is added to the sample, its concentration would be derived
from its observed intensity IS by:
CS KIS MB
(13)
Because the analyte and standard are equally distributed in the analyzed
sample, one can assume that the M and B constants would be the same for
both species. In that case, dividing equation (12) by equation (13) and rearranging yields a relation enabling the calculation of the amount of analyte in the
sample:
Ci (Ii =IS )CS
(14)
This method of using an internal standard to deal with matrix and background effects has, of course, been used for ages in analytical chemistry.
Equally useful would be another analytical method, namely that of standard
additions. In this latter approach, the sample is spiked with successively increasing amounts of the pure analyte, and the intercept in the response concentration
curve is used to calculate the amount of analyte in the original sample.
There has always been great interest in developing algorithms that enable
quantitative XRF spectroscopy to be performed without the use of standards,
84
Brittain
methodology that is often referred to as the fundamental parameters approach.

The method was first proposed by Criss, who published a program suitable for
its implementation (40). The advantage of the fundamental parameters approach
is that it represents a more time-efficient and less tedious method of analysis that,
under the right circumstances, does not sacrifice accuracy for the sake of
expediency.
For example, fundamental parameter calculations have been used for the
analysis of europium over concentration ranges of 0.1 30.0% w/w in alumina,
calcia, magnesia, lanthania, and thoria oxidic catalyst supports (41). Results comparable with the multiple regression method were obtained when the matrix stoichiometry was defined as Eu203 and the catalyst oxide equally defined by its
stoichiometry. It was found to be necessary to use two standards, which bracketed
the Eu(III) concentrations in the actual samples. However, when these conditions
were met, the results obtained were comparable with those obtained using a 10point multiple regression calibration curve. It was also found that the percent
relative difference between the fundamental parameters and multiple regression
results was approximately 2%, but the fundamental parameter results were
obtained with a considerable saving of standard preparation time.
Total reflection X-ray fluorescence (TXRF) has become very popular for
the conduct of microanalysis and trace elemental analysis (42 44). TXRF
relies on scatter properties near and below the Bragg angle to reduce background
interference and to improve limits of detection that can amount to an order of
magnitude (or more) over more traditional XRF measurements. As illustrated
in Figure 7, if X-rays are directed at a smooth surface at a very small angle,
virtually all of the radiation will be reflected at an equally small angle.
However, few X-rays will excite atoms immediately at the surface, and those
atoms will emit their characteristic radiation in all directions. One obtains very
clean analytical signals when using the TXRF mode because there is essentially
no backscatter radiation impinging on the detector.
TRXF was used to determine the trace elements in samples of lecithin,
insulin, procaine, and tryptophan in an attempt to develop elemental fingerprints
that could be used to determine the origin of the sample (45). It was reported that
through the use of matrix-independent sample preparation and an internal standard, one could use TXRF to facilitate characterization of the samples without
the need for extensive pretreatment. In another work, a study was made of the
capability of TXRF for the determination of trace elements in pharmaceutical
substances with and without preconcentration (46).
Trace amounts of bromine in sodium diclofenac, sodium {2-[(2,6-dichlorophenyl)amino]phenyl}acetate, have been determined using XRF (47), because
the drug substance should not contain more than 100 ppm of organic
bromine remaining after completion of the chemical synthesis. Pellets containing
the analyte were compressed over a boric acid support, which yielded stable
samples for analysis, and selected XRF spectra obtained in this study are
shown in Figure 8. It was found that samples from the Far East contained
85
Figure 7 Instrumental arrangement for the measurement of total reflection XRF. The
X-rays from the source (A) are allowed to impinge on the sample mounted on a reflector
plate (B). Most of the incident radiation bounces off the sample (C) but some results in the
production of XRF (D), which is measured by the detector (E). Abbreviation: XRF, X-ray
fluorescence.
more than 4000 ppm of organic bromine, various samples from Europe contained
about 500 ppm, and samples from an Italian source contained less than 10 ppm of
organic bromine.
No sample pretreatment was required for analysis of the Cs, Ba2, Cu2,
and La3 content in poly(vinyl alcoholvinyl sulfate) copolymers by TXRF
(48). These assays facilitate the estimation of the efficiency associated with the
ion-exchange preparation of polyelectrolytes and for determinations of the
stoichiometry of the end product. The results of the analyses were crossreferenced against data obtained using ICP-AES and titrimetric technology,
and good agreement was obtained between the various methods.
TXRF was used to characterize high-viscosity polymer dispersions (49),
with special attention being paid to the different drying techniques and their
effect on the uniformity of the deposited films. TXRF was also used as a
means to classify different polymers on the basis of their incoherently scattered
peaks (50). Dispersive XRF has been used to asses the level of aluminum in
antacid tablets (51).
Probably the most effective use of XRF and TXRF continues to be in the
analysis of samples of biological origin. For instance, TXRF has been used
86
Brittain
Figure 8 XRF spectra of pure diclofenac (lower trace), and diclofenac-containing

,
organic bromine (upper trace). The bromine Ka peak is the marked peak at 1.04A
. Source: From Ref. 47.
whereas the bromine Kb peak is the marked peak located at 9.93A
without a significant amount of sample preparation to determine the metal

cofactors in enzyme complexes (52). The protein content in a number of
enzymes has been deduced through a TXRF of the sulfur content of the components methionine and crysteine (53). It was found that for enzymes with
low molecular weights and minor amounts of buffer components that a reliable
determination of sulfur was possible. In other works, TXRF was used to
determine trace elements in serum and homogenized brain samples (54),
selenium and other trace elements in serum and urine (55), lead in whole human
blood (56), and the Zn/Cu ratio in serum as a means to aid in cancer diagnosis (57).
87
Thin sections of biological materials were prepared by means of a freezing

microtome, placed on quartz carriers and analyzed by XXRF (58). Quantitation
was effected by spiking the sections with an internal standard, and the time frame
for the analysis was suitable for use as screening and monitoring methodology.
The calcium and iodine content in gall bladder stones has been determined
using EDX techniques, where quantitation was enabled by the method of
standard additions (59). In other works, TXRF methods have been used to
study the changes in trace elements present in different human tissues when
these have be altered due to the presence of disease states (60 61).
REFERENCES
1. Siegbahn K, Nordling C, Fahlman A, Nordberg R, Hamrin K, Hedman J, Johansson G,
Bergmark T, Karlsson SE, Lindgren I, Lindberg B. ESCA Atomic, Molecular, and
Solid State Structure Studied by Means of Electron Spectroscopy. Uppsala, 1967.
2. Sleigh C, Pijpers AP, Jaspers A, Coussens B, Meier RJ. J Electron Spect Rel Phen
1996; 77:41.
3. Carlson TA. Photoelectron and Auger Spectroscopy. New York: Plenum Press,
1975:15 63.
4. Ghosh PK. Introduction to Photoelectron Spectroscopy. New York: John Wiley &
Sons, 1983:23 52.
5. See, for example, Turner NH, Schreifels JA. Anal Chem 2000; 72:99R; ibid., 1998;
70:229R.
6. Steinhardt RG,. Granados FAD, Post GL. Anal Chem 1955; 27:1046.
7. Jolly WL, Bomben KD, Eyermann CJ. At Nucl Data Tables 1984; 31:433.
8. In: CRC Handbook of Chemistry and Physics. Lide DR, ed. 74th ed. Boca Raton, FL:
CRC Press, 1993:10 272 through 10 277.
9. See Ref. 49, pp. 76 77.
10. Ghosh PK. Introduction to Photoelectron Spectroscopy. New York: John Wiley &
Sons, 1983:55.
11. Wozniak K, He H, Klinowski J, Barr TL, Hardcastle SE. J Phys Chem 1996;
100:11408.
12. Wozniak K, He H, Klinowski J, Jones W, Barr TL. J Phys Chem 1995; 99:14667.
13. Wozniak K, He H, Klinowski J, Barr TL, Milart P. J Phys Chem 1996; 100:11420.
14. Sharma J, Garrett WL, Owens FJ, Vogel VL. J Phys Chem 1982; 86:1657.
15. Wang C-H, Sengothi K, Wong HM, Lee T. J Pharm Sci 1999; 88:221.
16. Lacasse FX, Hildgen P, McMullen JN. Int J Pharm 1998; 174:101.
17. Cassidy OE, Rowley G, Fletcher IW, Davies SF, Briggs D. Int J Pharm 1999; 182:199.
18. Bettencourt A, Calado A, Amaral J, Alfaia A, Vale FM. Monteiro J, Montemor MF,
Ferreira MGS, Castro M. Int J Pharm 2004; 278:181.
19. Liu Z, Wu XY, Ballinger JR, Bendayan R. J Pharm Sci 2000; 89:807.
20. Davies MC, Wilding LR, Short RD, Khan MA, Watts JF, Melia CD. Int J Pharm 1989;
57:183.
21. Felton LA, Perry WL, Pharm Dev Tech 2002; 7:43.
22. Nakatani J, Ozawa S, Ogino Y. J Phys Chem 1989; 93:3255.
88
Brittain
23. Burke GM, Wurster DE, Berg MJ, Veng-Pedersen P, Schottelius DD. Pharm Res
1992; 9:126.
24. Buckton G, Bulpett R, Verma N. Int J Pharm 1991; 72:157
25. Prestidge CA, Tsatouhas G. Int J Pharm 2000; 198:201
26. Kiss E, Bertoti I, Vargha-Butler EL. J Colloid Interfac Sci 2002; 245:91.
27. Cauchois Y, Senemand C. Wavelengths of X-Ray Emission Lines and Absorption
Edges, Volume 18 of International Tables of Selected Constants, Oxford: Pergamon
Press, 1978.
28. Herglotz HK, Birks LS. X-Ray Spectrometry. New York: Marcel Dekker, 1978.
29. Fabian DJ. Soft X-Ray Band Spectra. London: Academic Press, 1968.
30. Jenkins R. An Introduction to X-Ray Spectrometry. London: Heyden, 1970.
31. Van Griken RE, Markowicz AA. Handbook of X-Ray Spectrometry. New York:
Marcel Dekker, 1992.
32. Jenkins R. Comparison of Wavelength and Energy Dispersive Spectrometers. Chapter
nd
7 in X-Ray Fluorescence Spectrometry, 2 edn. New York: Wiley-Interscience,
1999:111 121.
33. Newman AW, Brittain HG. Optical and Electron Microscopies. Chapter 5 in Physical
Characterization of Pharmaceutical Solids. New York: Marcel Dekker, 1995:127
156.
34. Berridge JC. J Pharm Biomed Anal 1995; 14:7
35. Salemink PJ, Elzerman HJW, Stenfert JTh, Gribnau TCJ. J Pharm Biomed Anal 1989;
1989; 7:1261.
36. Wang T, Jia X, Wu J. J Pharm Biomed Anal 2003; 33:639.
37. Muller RO. Spectrochemical Analysis by X-Ray Fluorescence. New York: Plenum,
1972.
nd
38. Betrin EP. Principles and Practice of X-Ray Spectrometric Analysis. 2 edn. New
York: Plenum, 1975.
39. Tertian R, Claisse F. Principles of Quantitative X-Ray Fluorescence Analysis.
London: Heyden, 1982.
40. Criss JW. Adv X-Ray Anal 1980; 23:93
41. Kenny EM, Brittain HG. Adv X-Ray Anal 1985; 28:209.
42. Prange A. Spectrochim Acta 1989; B44:437
43. Aiginger H. Spectrochim Acta 1991; B46:1313
44. Wobrauschek P. J Anal Atom Spectromet 1998; 13:333.
45. Wagner W, Rostam-Khani P, Wittershagen A, Rittmeyer C, Kolbesen BO, Hoffmann
H. Spectrochim Acta 1997; B52:961.
46. Kelki-Levai A, Varga I, Zih-Perenyi K, Laszity A. Spectrochim Acta 1999; B54:827.
47. Da Re MAP, Lucchini F, Parisi F, Salvi A. J Pharm Biomed Anal 1990; 8:975.
48. Varga I, Nagy M. Spectrochim Acta 2001; 2001; B56:2229.
49. Vasquez C. Spectrochim Acta 2004; B59:1215.
50. Boeykens S, Vasquez C. Spectrochim Acta 2004; B59:1189.
51. Georgiades CA. J Assoc Off Anal Chem 1990; 73:385.
52. Wittershagen A, Rostam-Khani P, Klimmek O, Gross R, Zickermann V, Zickermann
I, Gemeinhardt S, Kroger A, Ludwig B, Kolbesen BO. Spectrochim Acta 1997;
B52:1033.
53. Mertens M, Rittmeyer C, Kolbesen BO. Spectrochim Acta 2001; B56:2157.
54. Marco LM, Greaves ED, Alvardo J. Spectrochim Acta 1999; B54:1469.
89
55. Bellisola G, Pasti F, Valdes M, Torboli A. Spectrochim Acta 1999;B54:1481.

56. Ayalaa RE, Alvarez EM, Wobrauschek P. Spectrochim Acta 1991; B46:1429.
57. Marco LM, Jimenez E, Hernandez EA, Rojas A, Greaves ED. Spectrochim Acta
2001;. B56:2195.
58. Klockenkamper R, von Bohlen A, Wiecken B. Spectrochim Acta 1989; B44:511.
59. Ekinci N, Sahin Y. Spectrochim Acta 2002; B57:167.
60. Benninghoff L, von Czarnowski D, Denkhaus E, Lemke K. Spectrochim Acta 1997;
B52:1039.
61. von Czarnowski D, Denkhaus E, Lemke K. Spectrochim Acta 1997; B52:1047.
PART III.
VALENCE ELECTRON SPECTROSCOPY
4
Molecular Orbital Theory and the
Electronic Structure of Molecules
Harry G. Brittain
MOLECULAR ORBITAL (MO) THEORY OF ORGANIC MOLECULES

In the previous chapters, the electronic structure of atoms has been developed
by first considering the orbitals, energy levels, and states of the hydrogen atom.
Subsequently, these concepts were generalized and extended to deduce similar
descriptions of the properties of multielectron atoms. Because electrons occupying completely filled shells of a given principal quantum number are localized
around their respective nuclei, it was possible to discuss a variety of core electron
spectroscopies in terms of atomic orbitals and the states derived from these.
In considering the electronic structure of polyatomic molecules, it is
tempting to try a rigorous quantum mechanical approach, beginning with the
Schrodinger equation:
^ ci Ei ci
H
(1)
where Ei is the energy associated with the wave function ci. As discussed in the
previous chapter, the solution to the Schrodinger equation requires a specification
of the form of the Hamiltonian operator:
^ T^ V^
H
(2)
terms. Although one may write the equations that describe all of the kinetic
91
92
Brittain
energy terms in a polyatomic molecule, great difficulty is encountered once

equations for the potential energy terms are written. In the case of the hydrogen
atom, the potential term for the electron was given by the attractive force between
the negatively charged electron and the positively charged nucleus, that is,
Coulombs Law:
V e2 =r
(3)
where r is the distance of the electron from the nucleus, and e is the electron fundamental charge. For the hydrogen atom, the Hamiltonian operator took the form:
^ h r2 e
H
2me
r
(4)
where me is the mass of the electron. Through the use of appropriate mathematical techniques, one was able to substitute Equation (4) into Equation (1) and
solve for the wave functions that described the electronic structure of the atom.
Let us now consider the case of simplest molecule possible, namely the
hydrogen molecular ion (H
2 ). This species would consist of two nuclei and a
single electron, and the Hamiltonian operator becomes considerably more complicated owing to the nature of the potential energy term:
2

2
2
2
^ h r2 e e e
H
(5)
2me
rA
rB rAB
where rA is the distance of the electron from nucleus A, rB is the distance of the
electron from nucleus B, and rAB is the internuclear separation.
The Schrodinger equation obtained by the substitution of Equation (5) into
Equation (1) can be solved analytically by transforming the system into one
described by confocal elliptical coordinates. When this approach is taken, the
resulting equations have been solved with great accuracy, yielding an exact solution that calculates the correct internuclear separation and electron dissociation
energy (1). The calculated electron distribution for the ground state of H
2 indicates that the most probable region for the electron is to be located between
the two nuclei. The localization of electron density between the two nuclei is
described by a wave function known as a molecular orbital and represents the
source of stability that holds the nuclei together to form the molecule.
Unfortunately, the ability to derive an exact solution to the Schrodinger
equation for molecules containing more than one electron is not possible. Just
as the addition of more electrons to the hydrogen made an exact solution of
the atomic system impossible, the addition of more electrons and nuclei to the
hydrogen molecular ion tremendously complicates the situation. The basic
problem is that the Hamiltonian operator must contain terms that account for
the attraction of each electron for each proton and a potential energy term that
shows the effect of repulsion by each electron for one another:
V ee e1 e2 =r12
(6)
MO Theory and the Electronic Structure of Molecules
93
where r12 is the interelectron separation. The resulting Schrodinger equation

is an example of the classic multibody problem in physics, where the large
number of interparticle potential energy terms makes solution of the equation
impossible.
Because there can be no exact solutions of the Schrodinger equation for real
molecules, it follows that rigorous wave functions for molecules cannot be
derived. In order to develop theories of molecular bonding and electronic structure, approximation methods have been used to derive the best possible descriptions. One such theory has been termed as the valence-bond model, where
covalent molecules are assembled on the basis of shared pairs of electrons (2).
Although the valence-bond provides a qualitative understanding of molecular
properties, true quantitative results require the use of variational and perturbation
methods. There are better approaches, however, that yield useful qualitative and
quantitative descriptions of molecular structure and electronic properties; these
approaches will be discussed in the following sections.
LCAO-MO THEORY
The hydrogen molecule is formed by the combination of two hydrogen atoms,
where shared electron density between the nuclei represents the bonding force
that holds the two atoms together. However, it is evident that if one steadily
increases the separation between the two nuclei, then at some internuclear
separation the molecule must dissociate into two nonbonded hydrogen atoms.
If these hydrogen atoms are sufficiently separated so that there is no significant
interaction between them, then the result is a pair of simple hydrogen atoms
whose wave functions can be obtained using the Schrodinger equation in the
manner described in the previous sections. The following discussion is only a
brief exposition of the basic principles of linear combination of atomic orbitals
(LCAO)-MO theory, and interested readers should consult the standard references in the field (3 8).
The opposite concept of separation is the situation where two hydrogen
atoms are brought successively closer to each other until electron density
becomes shared between them. This suggests that one could reasonably
approximate the MO of the hydrogen molecule as some sort of combination
of the atomic orbitals of the constituent hydrogen atoms. The linear combination of atomic orbitals to form MO (commonly known as the LCAO-MO
theory) is one such approach (3 8). The LCAO MOs of the hydrogen molecule
are expressed as:
F N(ac1 bc2 )
(7)
where F is the wave function of the MO, c1 and c2 are the atomic orbitals of
the individual atoms, N is a normalization factor, and a and b are the linear
mixing coefficients.
94
Brittain
Without specifying the form of the Hamiltonian operator, we can substitute

the wave function of Equation (7) into the Schrodinger equation and solve for the
energy of that wave function by a calculation of its expectation value:
^ dt
E F HF
(8)
where the integration is to be performed over all space. Because E is a constant,
Equation (8) can be rewritten as:
^ E)F dt 0
F (H
(9)
Equation (9) can only be exactly satisfied for the Hamiltonian operator in question if the MO wave function is the exact wave function. But because we are
working within an approximation framework, then the solution of Equation (9)
will exceed zero by the residual, r:
^ E)F dt r
F (H
(10)
The smaller the value of r, the closer is the expectation value of the Hamiltonian
expressing the correct energy, and the closer is the LCAO wave function
expressing the true wave function.
If one substitutes the LCAO wave function of Equation (7) into Equation
(10) and assumes the existence of real functions for F, then:
^ E)(ac1 bc2 ) dt r
N2 (ac1 bc2 )(H
(11)
The calculus of variations states that a function will be minimal whenever the
partial derivatives with respect to each of its variables are simultaneously equal
to zero. Considering the linear mixing parameters, a and b, to be the variables,
then r will be at a minimum and the expectation value for the Hamiltonian over
the LCAO wave function will be as close as possible to the true energy when:
(@r=@a) 0
(12)
(@r=@b) 0
(13)
Multiplying Equation (11) and collecting the terms in the mixing

parameters, one obtains:

2
2
2
^
^
N a
c1 H c1 d t E c1 c 1 d t b
c2 Hc2 dt E c2 c2 dt

^ c2 d t E c 1 c2 d t r
(14)
2ab c1 H
95
If the individual atomic orbitals are normalized in the usual manner, then:
c 1 c1 d t 1
(15)
c 2 c2 d t 1
(16)
It is usual practice to define the Coulomb integral as:
^ ci d t
ai c i H
(17)
the reasonance integral as:
^ c j dt
bij ci H
(18)
and the overlap integral as:
Sij ci c j dt
(19)
Substituting Equations (17), (18), and (19) into Equation (14), and dividing
throughout by N2, yields:
a2 (a1 E) b2 (a2 E) 2ab(b12 ES12 ) r=N2
(20)
To apply the variational principle to Equation (20), one must minimize r/N2 with
respect to both a and b, yielding the two equations:
@(r=N2 )=@a 2a(a1 E) 2b(b12 ES12 ) 0
2
@(r=N )=@b 2a(b12 ES12 ) 2b(a2 E) 0
(21)
(22)
The result of the minimization process is two equations in the linear mixing
coefficients. The only situation whereby one can have this set of linear equations
simultaneously equalling zero is for the determinant, which multiplies the
coefficients to equal zero:

a1 E
b12 ES12

0
(23)
b ES12
a2 E
12
Because both atoms are equivalent in the case of the hydrogen molecule, the
subscripts to the individual atoms are unnecessary. In this case, the value of the
determinant becomes:
(a E)(a E) (b ES)(b ES) 0
(24)
E2 (1 S2 ) 2E(a Sb) a2 Sb2 0
(25)
or
96
Brittain
Equation (25) can be directly solved using the quadratic equation, and the two solutions are found to be:
E1 (a b)=(1 S)
E2 (a b)=(1 S)
(26)
(27)
Evaluation of the integrals reveals that b has a negative value, whereas S

has a relatively small positive value. One, therefore, finds that the two energy
levels will be nearly symmetrically arrayed about the value of a, which is the
energy of the state derived from the occupancy of one electron in the lowest
energy atomic orbital of the isolated hydrogen atom. The implication of Equations
(26) and (27) is that the LCAO method represents the bringing together of two
hydrogen atoms and the consequent formation of two MOs from their respective
atomic orbitals. As illustrated in Figure 1, one of these MOs will have a lower
energy than the isolated atoms, and the other will have a higher energy than the
isolated atoms.
Figure 1 Energy level diagram for the two LCAO-MO energy levels resulting from
the combination of the atomic orbitals of two atoms. Abbreviation: LCAO-MO, linear
combination of atomic orbitals molecular orbital.
97
It follows that if each hydrogen atom contributes one electron to the hydrogen molecule, then a more stable system is achieved if both of these electrons
occupy the lower energy MO than if they remained in their atomic orbitals.
For this reason, the lower energy MO is termed as a bonding orbital, and for
reasons to be discussed next, the higher energy MO is termed as an antibonding
orbital.
LCAO-MO THEORY AND THE SPECTROSCOPY OF
DIATOMIC MOLECULES
In the previous discussion of atomic orbitals, it was recognized that the rules
associated with quantum number possibilities placed restrictions on possible
atomic wave functions. As a result, the electrons in the polyelectronic atoms of
the first and second rows of the periodic table fill atomic orbitals in the order
1s, 2s, and then 2p, a trend commonly referred to as the Aufbau principle.
One, therefore, deduces the following electronic structures for these atoms:
H
1
He
2
Li
3
Be
4
B
5
C
6
N
7
O
8
F
9
Ne 10
electron )
electrons )
electrons )
electrons )
electrons )
electrons )
electrons )
electrons )
electrons )
electrons )
1s1
1s2
1s2 2s1
1s2 2s2
1s2 2s2 2p1
1s2 2s2 2p2
1s2 2s2 2p3
1s2 2s2 2p4
1s2 2s2 2p5
1s2 2s2 2p6
It follows that the LCAO MOs that can be formed from combination of these
elements will necessarily involve electrons derived from s-orbitals and p-orbitals.
The boundary surface of an s-orbital consists of a sphere, and so the mathematical
sign of the orbital is the same everywhere. On the other hand, the boundary
surfaces of p-orbitals are defined by a dumbbell shape, with one lobe having
one type of sign and the other having the opposite sign. A much more detailed
discussion of the electronic structure and states of diatomic molecules is available
in Ref. (9).
As has been established during the discussion of the hydrogen atom, electron density that is shared between two atoms constitutes the attractive force that
is commonly known as a chemical bond. Electron density can only be shared by
the overlap of orbitals having the same mathematical sign, and Figure 2 show the
types of bonds that can be formed by the combination of s- and p-orbitals. The
bond types are further differentiated by the presence or absence of a node of
electron density lying along the internuclear axis. Bonds formed by the overlap
of two s-orbitals or one s-orbital and one pX-orbital lack a node along this axis
98
Brittain
and are termed s-bonds, with the two-electron wave functions describing
these bond types are known as s-orbitals. Bonds formed by the overlap of two
pY-orbitals or two pZ-orbitals have a single node along this axis and are termed
p-bonds, and therefore the two-electron wave functions describing these bond
types are known as p-orbitals. Antibonding orbitals are defined by the same symmetry types and are marked with an asterisk to distinguish them from bonding
orbitals.
For homonuclear diatomic molecules, the MOs are further differentiated as
being gerade (symmetric) or ungerade (antisymmetric) with respect to inversion
through the center of gravity of the molecule. Hence, the bonding s-orbitals
and antibonding p-orbitals are symmetric in character (gerade), whereas the
bonding p-orbitals and antibonding s-orbitals are antisymmetric in character
(ungerade).
The difference in energy between the component atomic orbitals also
governs the types of MOs that can be formed between two atoms. For instance,
the energies of the 2s electrons are considerably greater than the energies of the
1s electrons, so MOs cannot be formed by the overlap of the 1s orbital of one
Figure 2 Types of bonds that can be formed by the combination of s- and p-orbitals. The
open orbital types signify a positive wave function, and the shaded orbital types signify a
negative wave function.
99
atom with the 2s orbital of the other. Likewise, the energy of the 2s orbital
is sufficiently different from that of the pX-orbital so that no MO is formed by
these.
Considering both the symmetry of the MOs and the energetics of the component atomic orbitals, MOs are listed in the following sequence of increasing
energy:
sg (1s) su (1s) sg (2s) su (2s) sg (2p) pu (2p) pg (2p) su (2p)
In this sequence, it has been recognized that the pu(2pY) and pu(2pZ) two-electron
orbitals have exactly the same energy, and are therefore identified as the pu(2p)
four-electron orbital. Similarly, the pg(2p) four-electron orbital is seen to consist
of the isoenergetic pair of pg(2pY) and pg(2pZ) two-electron orbitals. In addition,
the sg(2p) and su(2p) orbitals are understood to actually be the sg( pX) and su( pX)
orbitals.
The Aufbau principle can be applied to write the MO sequences of the
homonuclear diatomic molecules formed by the elements of the first row of
the periodic table:
H
2
H2
He
2
He2
1
2
3
4
electron
electrons
electrons
electrons
)
)
)
)
sg (1s)1
sg (1s)2
sg (1s)2 su (1s)1
sg (1s)2 su (1s)2
For the diatomic molecules of the second row, electrons in the [sg(1s)2 su(1s)2]
kernel are the core electrons, and the electrons in the other MO are the valence
electrons. In the absence of configurational interactions, the Aufbau principle
can be applied to write the MO sequences of the homonuclear diatomic
molecules formed by the elements of the first and second rows of the periodic
table:
Li2
Be2
B2
C2
6
8
10
12
electrons
electrons
electrons
electrons
)
)
)
)
sg (1s)2 su (1s)2 sg (2s)2

sg (1s)2 su (1s)2 sg (2s)2
sg (1s)2 su (1s)2 sg (2s)2
sg (1s)2 su (1s)2 sg (2s)2
pu (2p)2
sg (1s)2 su (1s)2 sg (2s)2
pu (2p)4
sg (1s)2 su (1s)2 sg (2s)2
pu (2p)4 pg (2p)2
su (2s)2
su (2s)2
su (2s)2 sg (2p)2
su (2s)2 sg (2p)2
su (2s)2 sg (2p)2
N2
14 electrons )
O2
16 electrons )
F2
18 electrons )
sg (1s)2 su (1s)2 sg (2s)2 su (2s)2 sg (2p)2

pu (2p)4 pg (2p)4
Ne2
20 electrons )

pu (2p)4 pg (2p)4 su (2p)2
su (2s)2 sg (2p)2
100
Brittain
The bond order (BO) of a molecule is calculated as:

BO (# bonding electrons # antibonding electrons)=2
(28)
The existence of a stable diatomic molecule is indicated by a bond order that

exceeds zero. Any molecule that contains an equal number of bonding and
antibonding electrons is characterized by a bond order of zero and should not
exist, because there would be no stabilization mechanism to hold the atoms
together. It is a fact that no evidence has been found implicating the existence
of He2, Be2, or Ne2 in their ground states.
The MO sequences of heteronuclear diatomic molecules follow a similar
manner. Hence, for instance, the sequences for the following molecules are
written as:
BeO
12 electrons )
sg (1s)2 su (1s)2 sg (2s)2 su (2s)2 sg (2p)2 pu (2p)2
CN
13 electrons )
CO
14 electrons )
NO
15 electrons )
pu (2p)4 pg (2p)1

NF
16 electrons )
pu (2p)4 pg (2p)2

OF
17 electrons )
pu (2p)4 pg (2p)3

It is tempting at this point to describe the electronic spectroscopy of one
of these diatomic molecules as the promotion from a lower MO to one of the
higher, unoccupied MOs, but this would be inappropriate. As was the case for
atomic spectroscopy, electronic transitions are governed by the selection rules
derived from their angular momenta properties, and the same situation holds
for the electronic transitions of diatomic molecules. Only after the nature of
the electronic states of the component atoms are established in the molecule,
the nature of the electronic states of the molecule itself are established, and
then only an understanding of the nature of the electronic transition can be
developed.
It was established in an earlier section that the wave functions of a hydrogen-like atom are specified by the values of three quantum numbers. For a given
value of the principal quantum number, n, the angular momentum quantum
number, l, allows all possible values between 0 and n 2 1: orbitals for which
l 0 are the s-orbitals, orbitals for which l 1 are the p-orbitals, orbitals for
101
which l 2 are the d-orbitals, and so on. Finally, for a given orbital type defined
by its angular momentum quantum number, the magnetic quantum number, m,
can take all values from l to 0 to l. For atoms characterized by the existence
of spherical symmetry, the energies must be the same for all orbitals having
the same value 8 of n and l. It is important to understand that the value of the
angular momentum quantum number implies an orbital angular momentum of
(h/2p)[l(l 1)]1/2, and the value of the magnetic quantum number implies a
component of this angular momentum equal to m(h/2p) along a specified direction. In a multielectronic atom, the energy of a given electron is found to depend
on its value of all three quantum numbers.
For multielectronic atoms, there is a simple relation between the values of
the angular momentum quantum numbers and the total angular momentum, L, of
the atom. For an atom having two electrons characterized by l1 and l2, the permissible L values will be (l1 l2), (l1 l2 2 1), (l1 l2 2 1), . . . , (l1 2 l2). In
the same way, the atomic orbitals that are labeled as s, p, or d, indicate l
values of 0, 1, or 2, and the electronic states for which L 0, 1, or 2 are
labeled as S, P, or D. The electron spin of an atomic orbital is described by
the spin quantum number, ms, for which the permissible values are +1/2. The
total spin angular momentum, S, for a two electron system will therefore be
S 0 (spins paired) or S 1 (spins unpaired). The electronic state of an atom
is defined by its L and S values, and the atomic spectroscopy discussed in the
previous chapters is concerned with the selection rules associated with changes
in L and S.
For diatomic molecules created by bonding of two separated atoms, the
angular and spin momenta of the individual electrons combine to form the resultant orbital (L) and spin (S) angular momenta that now define the electronic states
of the molecule. The following term symbols are used to identify electronic states
by their resultant total orbital angular momentum values:
L0
Symbol S
L1
Symbol P
L2
L3
Symbol F
Symbol D
Electronic states are further identified by their total spin angular momentum
values, by placing the value of the spin multiplicity, (2S 1), as a left-hand
superscript to the term symbol of the state. Any MO that contains its
maximum number of allowed electrons (i.e., completely full) would be characterized by L 0 and S 0.
The number and types of possible electronic states will depend on the
number and types of electrons occupying the various MOs. Equivalent electrons
are defined as those occupying the same MO and therefore have the same n and l
quantum numbers. The electronic terms that arise from the coupling of equivalent
102
Brittain
s- or p-electrons are:
s1
s2
p1
p2
p3
p4
S
S
2
P
1
S, 3 S, 1 D
2
P
1
S
1
Hunds rule states that the ground states of an atom or molecule is that
having the highest value for the spin multiplicity, and for states of equal spin
multiplicity the state of the lower orbital angular momentum will be more
stable. Diatomic molecules are described by the equivalent electron cases, and
their descriptions summarized in Table 1. It should be noted that the effects of
configuration interaction were neglected in deriving the ground state terms of
Table 1, and therefore some of the states indicated for the diatomic molecules
are not rigorously correct. However, they will suffice for the purposes of this
discussion.
Nonequivalent electrons are those that occupy different MO and therefore
have no restriction on their n and l quantum numbers. This case is of interest,
because the excited states of diatomic molecules are described by the nonequivalent electron cases. The electronic terms arising from the coupling of
nonequivalent s- or p-electrons are:
si sj
S, 3 S
si pj
P, 3 P
pi pj
(2)1 S, (2)3 S, 1 D, 3 D
Consider the ground electronic configuration of C2:

GS sg (1s)2 su (1s)2 sg (2s)2 su (2s)2 sg (2p)2 pu (2p)2
for which the term symbol of the ground state is 3S. This configuration would
also cause the excited 1S and 1D states. One excited state configuration for this
diatomic molecule is:
ES1 sg (1s)2 su (1s)2 sg (2s)2 su (2s)2 sg (2p)2 pu (2p)1 pg (2p)1
which would be characterized by two 1S, two 3S, one 1D, and one 3D excited electronic states. Another excited state configuration for this diatomic molecule is:
ES2 sg (1s)2 su (1s)2 sg (2s)2 su (2s)2 sg (2p)2 pu (2p)1 su (2p)1
which would be characterized by one 1P and one 3P excited electronic states. In
principle, electronic transitions take place from the 3S ground electronic state to
any of these excited electronic states.
103
Table 1 Summaries of the Properties of Some Diatomic Molecules in Their Ground

States, in the Absence of Configuration Interaction
Electronic configuration
sg(1s)1
sg(1s)2
sg(1s)2 su(1s)1
sg(1s)2 su(1s)2
[sg(1s)2 su(1s)2] sg(2s)2 su(2s)2
[sg(1s)2 su(1s)2] sg(2s)2 su(2s)2
[sg(1s)2 su(1s)2] sg(2s)2 su(2s)2 sg(2p)2
pu(2p)2
pu(2p)4
pu(2p)4 pg(2p)2
pu(2p)4 pg(2p)4
pu(2p)4 pg(2p)4 su(2p)2
[s(1s)2 s (1s)2] s(2s)2 s (2s)2 s(2p)2
p(2p)2
[s(1s)2 s (1s)2] s(2s)2 s (2s)2 s(2p)2
p(2p)3
[s(1s)2 s (1s)2] s(2s)2 s (2s)2 s(2p)2
p(2p)4
[s(1s)2 s (1s)2] s(2s)2 s (2s)2 s(2p)2
p(2p)4 p (2p)1
[s(1s)2 su(1s)2] s(2s)2 s (2s)2 s(2p)2
p(2p)4 p (2p)2
[s(1s)2 s (1s)2] s(2s)2 s (2s)2 s(2p)2
p(2p)4 p (2p)3
H
2
H2
He+2
He2
Li2
Be2
B2
C2
N2
O2
F2
Ne2
BeO
CN
CO
NO
NF
OF
Bond
order
1
2
1
1
2
0
1
0
1
2
Ground
state
2
S
S
2
S
1
S
1
S
3
S
1
112
112
1
2
However, the rules developed by Hund state that electronic transitions

characterized by DL 0 or +1, and DS 0 only are allowed. In that case, the
plausible absorption spectroscopy of the C2 molecule simplifies to the following:
3
S ! 3 SA (ES1 ),
SB (ES1 ),
P(ES2 )
There are other Laporte and parity selection rules that would apply to the C2
illustration (and which would further simplify the range possible for electronic
transitions) but have not been considered in this discussion for the sake of
simplicity.
104
Brittain
Spectroscopy among electronic states is determined by the expectation

value of the transition moment, Re:
(29)
Re c1 mc2 dt
where Re is the electronic dipole moment, summed over the electronic coordinates. For those instances where the value of the transition moment equals
zero, the transition is said to be forbidden. Hence, transitions among the electronic states of a molecule are referred to as electric-dipole transitions.
MO THEORY AND THE SPECTROSCOPY OF

POLYATOMIC MOLECULES
The electronic spectra of diatomic molecules and their associated vibrational and
rotation spectroscopies have been studied in far greater detail than most polyelectronic molecules. Because often, the energy of an electronically polyatomic molecule can be distributed so as to break weak chemical bonds, the spectroscopy of
high-energy states is usually only studied through the use of photoelectron spectroscopy. As a result, the only electronic transitions routinely studied for polyatomic organic molecules are those that involve transitions between the ground
state and its lowest-lying electronic states.
Nevertheless, the principles just developed to describe the electronic spectroscopy of diatomic molecules can be extended as a means to gain an understanding of the states of polyatomic molecules and their associated
spectroscopy. Each MO, Fi, can be constructed as a linear combination of
atomic orbitals, cp, but now with the orbitals of each nuclei appearing in expansion:
Fi Sci pcp
(30)
and weighted by their respective linear mixing coefficients, cip. Hence, the MOs
that can be constructed for polyatomic molecules are nonlocalized and extend
over all atoms in the molecule.
Symmetry of Polyatomic Molecules and its Effect
on their Electronic States
The problem with nonlinear polyatomic molecules is that there is no preferential
direction (like the internuclear axis of a diatomic molecule) onto which the
orbital angular momenta can be quantized. Although one can use the concept
of s- and p-orbitals to describe the character of individual electrons, it is not
possible to define meaningful angular momenta for the whole system.
However, by making use of the symmetry of the molecule and the associated
science of group theory, it becomes possible to understand the electronic
spectra of polyatomic molecules. The use of group theory in chemistry has
105
been well documented elsewhere (10 14), and therefore only the essential
aspects of the methodology will be presented here.
On considering the three-dimensional structures of molecules, it quickly
becomes apparent that such molecules (especially small molecules) possess
certain elements of recognizable symmetry. A symmetry operation is a movement of the atoms in a molecular species that, when performed, leaves the molecule in a configuration that is not distinguishable from the original configuration.
A symmetry element is a property of a molecule about which symmetry
operations can be carried out, such as a point, a line, or a plane. The action of
a symmetry operation takes an atom having coordinates of (x, y, z) and transforms
it to one having coordinates of (x0 , y0 , z0 ). For instance, if a molecule is rotated by
1808 about the z-axis, there is the effect of changing the (xi, yi, zi) coordinates of
every atom in the molecule to (2xi, 2yi, 2zi). The actions of symmetry operations can be visualized through the use of 3 3 matrices, although a considerable simplification will be achieved by using the characters of these matrices.
The properties of the five fundamental symmetry operations can be illustrated through consideration of how their performance affects the coordinates
of a hypothetical cyclic molecule consisting of four identical atoms. The first
element is the identity operation (E), whose role sounds trivial but plays an
essential part in the definition of a group:
There are many reflection planes (s) associated with this particular molecule, but the
one that executes a reflection along the direction of this page can be illustrated as:
The performance of a second reflection of the molecule brings it back to its

original position, an operation that is equivalent to the performance of the identity
operation. The molecule also possesses an inversion center (i), whose action is
illustrated as:
106
Brittain
The performance of a second inversion operation on the molecule brings it back to its
original position, an operation that is equivalent to the performance of the identity
operation.
A proper axis of rotation (CN) is defined as a rotation of the molecule by an
angle of 360/N degrees about a defined axis. Proper rotation axes are often
denoted by the number of times their performance brings the system back to
its original position. For example, consider the fourfold axis of rotation (i.e.,
rotation by 908) that is perpendicular to the molecular plane and passes
through the center of gravity of the molecule:
After the C4 operation is performed four times, the molecule is back to its
original position, which is equivalent to the performance of the identity operation. Alternatively, consider the twofold axis of rotation (i.e., rotation by
1808) that passes through the line connecting atoms A1 and A4:
Because the axis passes through atoms A1 and A4, their positions are not altered
upon performance of this symmetry operation. After the C2 operation is performed twice, the molecule reverts to its original position, which is equivalent
to the performance of the identity operation.
Finally, an improper axis of rotation (SN) is defined as a rotation of the molecule by an angle of 360/N degrees about a defined axis, followed by a reflection
operation through the plane. Improper rotation axes are also denoted by the
number of times their performance brings the system back to its original position.
We consider now the improper twofold axis (i.e., rotation by 1808, followed by a
reflection) that passes through the line connecting atoms A1 and A4. The operation can be envisioned as consisting first of rotation about the twofold axis:
107
To complete the S2 operation, the molecule is then subjected to a reflection operation along the direction of the twofold axis of rotation:
Interestingly, one finds that performance of the S2 operation is equivalent to the

performance of the identity operation.
In a more formal development of group theory, it can be shown that there
are only 47 possible combinations of symmetry operations that can mutually
coexist. These various symmetry combinations are known as point groups,
because the symmetry elements of a molecule leave a point of the molecule
fixed in space. This stands in contrast with the space groups found in crystals,
where additional symmetry operations result in a translation of the molecule
(or a unit cell) to a new location in the crystal. A more formal definition of a
point group is that it is a complete set of nonredundant symmetry operations possessing all the properties of a mathematical group.
Molecules can be systematically classified according to the number and
types of symmetry elements and operations they possess. Because all molecules
must be characterized by a finite set of symmetry operations, it follows that the
observable properties of molecules will be described and governed by the point
group that describes their symmetry properties. For instance, the application
of group theory to a particular molecule can be used to understand its patterns
of molecular vibration, its electronic properties, and the allowedness of the
electronic transitions associated with its MO.
The characteristics of every point group are encapsulated in its character
table. Readers who want to know how character tables are derived and assembled
should consult the leading Refs. (10 14), but in this development we will focus
exclusively on the use of character tables for our purposes. Character tables are
derived from the transformation matrices that describe the three-dimensional
effects of symmetry elements, with the character (the sum of the diagonal elements
of each matrix) of each symmetry element making up the entries. It is a consequence of group theory that there is a finite set of irreducible representations
that describe the symmetry properties of various combinations of symmetry
elements. Each property of the molecule will have symmetry properties that
cause it to be described by one of the irreducible representations of a point group.
All symmetry-related information is summed up in the character table of a
point group, which is ordinarily divided into the four main areas. To illustrate the
structure of a character table, the table for the C2V point group is shown in
Figure 3. Area I contains the names for the various irreducible representations,
while Area II contains sets of characters relating the symmetry properties of
108
Figure 3
areas.
Brittain
The character table of the C2V point group, illustrating its subdivision into four
each representation. Every wave function of a molecule will transform according

to one of the representations in the point group, and therefore its properties will
be defined by the characters of that group.
Area III indicates which representations appropriately describe motions
along the x, y, and z coordinates and rotations about these same axes. The significance of Area III becomes apparent once one understands that the elements of the
dipole moment operator transform according to the representations associated
with x, y, and z. Then, because the symmetry properties of the group representations are the symmetry properties of MO, it will be possible to derive selection
rules for allowing electric dipole transitions.
Finally, Area IV indicates the representations that describe squares and
binary products of coordinates according to their properties. Area IV assumes
significance because the elements of the polarizability tensor transform according
to the representations of the cross products as indicated in Area IV of the
character table. Because transitions in Raman spectroscopy entail changes in
molecular polarizability, selection rules for this method of observing molecular
motion can be derived as well.
Worked Example: The Electronic Spectroscopy of Formaldehyde
To illustrate the symmetry approach to MO theory and the electronic spectra of
polyatomic molecules, we will first consider formaldehyde, H2CO:
109
In chapter 2, the following atomic configurations were established:

H
1s1
1s2 2s2 2p2
1s2 2s2 2p4
We have also established in previous chapters that the oxygen 1s and carbon 1s
core electrons will not be involved in the chemical bonding or in the electronic
spectroscopy of formaldehyde. Thus, of the 16 electrons in formaldehyde, only
14 can be considered as being valence electrons.
In the conventional view, the central carbon is considered as being
sp 2-hybridized to achieve the trigonal geometry for the molecule. In other
words, its two 2s atomic orbitals are combined with one of its 2p atomic orbitals,
yielding a hybridized atomic orbital holding three electrons. The remaining, nonhybridized, 2p orbital is the one perpendicular to the molecular plane. As a result,
the new atomic configuration of the hybridized carbon can be written as:
C(sp2 )
1s2 2sp3 2p1
In this view, one of the three electrons of the sp 2-hybrid will have the proper symmetry to form a s-bond with the carbon atom.
In order to achieve the right type of orbital symmetry that enables formation of a double bond between the oxygen and carbon atoms, it is assumed
that the valence atomic orbitals of the oxygen atom also become sp 2-hybridized.
In such a case, the new atomic configuration of the hybridized oxygen can be
written as:
O(sp2 )
1s2 2sp3 2p3
In this view, the three electrons of the sp 2-hybrid will have the proper symmetry
to form s-bonds with the two hydrogen atoms and with the oxygen atom. It
should be noted that one of the nonhybridized 2p orbitals (the one perpendicular
to the molecular plane) will have the right symmetry to form a p-bond with the
carbon atom, thus, explaining the double bond as consisting of one s-bond and
one p-bond.
Assuming that the high-energy 2p electrons of the oxygen atom drop down
to fill the lower-energy orbitals of the sp 2 oxygen hybrids, the 16 electrons of formaldehyde can be placed into the following MO sequence describing the ground
state of the molecule:
H2 CO(G)
core (sCH1 )2 (sCH2 )2 (sCO )2 (pCO )2 (sp21 )2 (sp22 )2
The four 2-electron bonds of formaldehyde are represented by the sCH1, sCH2,
sCO, and pCO orbitals, and the four nonbonding electrons are represented by
the sp 21 and sp 22 orbitals. For the sake of simplicity, the (1sC)2 and (1sO)2 electrons
have been written simply as the core electrons.
110
Brittain
Of course, the formation of every bonding orbital must be accompanied by

the formation of its associated antibonding orbital. A large number of excitedstate configurations are written, each of which corresponds to an excited state
of the formaldehyde molecule. Although some of these are energetically unreasonable, the following three excited configurations and their states are of particular significance:
H2 CO(E1 )
core (sCH1 )2 (sCH2 )2 (sCO )2 (pCO )2 (sp21 )2 (sp22 )1 (pCO )1
H2 CO(E2 )
core (sCH1 )2 (sCH2 )2 (sCO )2 (pCO )1 (sp21 )2 (sp22 )2 (pCO )1
H2 CO(E3 )
core (sCH1 )2 (sCH2 )2 (sCO )2 (pCO )2 (sp21 )2 (sp22 )1 (sCO )1
Owing to the nature of the MO involved, the E1 transition is commonly identified

as the n ! p transition, the E2 transition is denoted as the p ! p transition,
and the E3 transition is classified as the n ! s transition.
In order to determine whether any of these spectroscopic transitions can
actually take
place, one must determine the expectation value of the transition
moment, c1 mc
^ 2 dt. Given the complexity of the involved MO wave functions,
an exact calculation of the transition is not practical. Fortunately, group theory
permits a ready determination of whether the expectation value will have a
finite value if the integration is successfully performed. Because every electronic
state of a molecule must possess symmetry properties that are described by one of
the representations of its molecular point group, the representations in its
character table can be used to determine whether the transition moment equals
zero or not. Because the integration step necessary to evaluate the transition
moment is performed over all space, any antisymmetric result must integrate
to zero. Therefore, if one takes the direct product of the representations for
c1 , m, and c2, unless the resulting representation contains the totally symmetric
representation (i.e., the one where all characters equal 1), the transition moment
of which must equal zero.
The task is greatly simplified when one recognizes that if the atomic orbitals of the ground state molecular wave function are all fully occupied, then that
electronic state will transform as the totally symmetric representation of the point
group. In the case of formaldehyde, whose point group is C2V and whose character table is given in Figure 3, the A1 representation summarizes the symmetry
properties of the ground state. This recognition greatly reduces the complexity
of the evaluation, because now one only needs to determine whether the direct
product of the representation of the excited state and the representation of the
components of the dipole moment operator contain the totally symmetric representation. But, it is a principle of group theory that only the direct product of
a representation with itself can yield a totally symmetric result.
Therefore, as long as the ground state transforms as the totally symmetric
representation of the point group (and it usually does for organic molecules), the
transition moment will be nonzero if the representation of the excited state is the
111
same as the representation of one of the dipole moment components. Consideration of the symmetry properties of the excited state orbitals enables one to
deduce that the E1 excited state transforms according to the A2 representation,
the E2 excited state transforms according to the A1 representation, and the E3
excited state transforms according to the B2 representation.
Examination of the C2V character table of Figure 3 reveals that the x component of the dipole moment operator transforms as the B1 representation, the y
component of the dipole moment operator transforms as the B2 representation,
and the z component of the dipole moment operator transforms as the A1 representation. One may, therefore, now deduce the following set of selection
rules for the spectroscopy of formaldehyde:
1. Because the E1 excited state transforms according to the A2 representation and because none of the components of the dipole moment
operator transform according to the A2 representation, the n ! p
transition is forbidden.
2. Because the E2 excited state transforms according to the A1 representation and because the z component of the dipole moment operator
transforms according to the A1 representation, the p ! p transition
is allowed.
3. Because the E3 excited state transforms according to the B2 representation and because the y component of the dipole moment operator
transforms according to the B2 representation, the n ! s transition
is allowed.
The power of group theory is that it permits one to determine whether a transition
will or will not be allowed, but it does not yield information regarding the
intensity of the transition.
We have not yet considered the effect of electron spin, and it is clear that
the MO configurations of the E1, E2, and E3 excited states could exist with either
the spins of the unpaired electrons being paired or unpaired. For paired electrons,
the multiplicity (2S 1) would equal one, and such states are known as singlet
states. For unpaired electrons the multiplicity equals three, and such states are
known as triplet states. Remembering that the multiplicity of a ground state
whose MO is composed of completely filled atomic orbitals must equal unity
and that the spin selection DS 0, the spin selection rule states that only
transitions to singlet excited states will be allowed.
The final stage in a spectroscopic analysis would be to use the results of the
theoretical study as a means to assign the nature of empirically observed electronic transitions. If assignments of the absorption bands to transitions between
given states can be made, the relative energies of the various electronic states
of the molecule can be deduced. It is a general rule that transitions allowed by
symmetry considerations will be extremely strong, whereas transitions forbidden
by symmetry will be extremely weak or not observed at all. For many carbonyl
compounds, a weak band is observed around 280 nm, and this is attributed to the
112
Brittain
forbidden n ! p transition. Clearly, some relaxation of the selection rule must

have taken place to permit this transition to take place at all. Very strong absorption bands are observed in the 170 180 nm region, which are attributed to the
overlapping p ! p and n ! s transitions.
GENERAL FEATURES OF THE ABSORPTION SPECTROSCOPY

OF ORGANIC MOLECULES
For the large molecules of pharmaceutical interest, the number of MOs becomes
so unwieldy that the type of detailed analysis just conducted for formaldehyde is
not feasible without the use of computers and semiempirical computer calculations. However, it has become established that although each MO extends
over the entire molecule, many of these orbitals have the majority of their electron density localized at one particular part of the molecule. For instance, the porbital forming half of the double bond between two atoms will be mostly localized in the vicinity of those atoms. By their nature, nonbonding electrons will be
localized on the atom, which furnished their atomic orbitals.
As a result of this electron localization, there are only a limited number of
electronic transition types that a molecule might experience (15 20). For organic
molecules, two of these involve transitions from bonding MOs (s ! s or
p ! p ) and are generally allowed by the symmetry-based selection rules.
The s ! s transitions are generally observed in the far-ultraviolet (UV)
region, whereas the p ! p transitions occur in the visible (VIS) or UV
regions of the spectrum. Two other transition types originate from nonbonding
MOs (n ! s or n ! p ). The fairly intense n ! s transition usually is
observed in the middle to far UV region, whereas the weak n ! p transition
is usually observed in the VIS or near-UV regions of the spectrum.
For organic molecules of pharmaceutical interest, the n ! p and p ! p
transitions are of the most interest and will therefore be discussed in more detail.
n ! p Transitions
Unsaturated molecules containing atoms that have localized nonbonding electrons (oxygen, nitrogen, sulfur, etc.) often exhibit a weak band in their absorption
spectra that can be assigned to a n ! p transition. As described earlier for formaldehyde, and extendable to carbonyl groups in general, this transition is
associated with an excitation from the lone-pair orbital localized on the
oxygen to the p -orbital of the carbonyl group. The observation that the bands
corresponding to the n ! p transition are weak suggests that the transition is
symmetry forbidden. The transition is completely forbidden in strict C2V symmetry, but the transition gains intensity owing to the symmetry breaking of the
carbonyl group caused by the presence of other atoms in the molecule. In the
113
language of Beers Law:

A abc
(31)
where A is the total absorbance, a is the molar absorptivity, b is the path length of
absorption, and c is the molar concentration, the molar absorptivity of a n ! p
transition will be in the range of 1 10 L/mol cm.
In saturated aldehydes and ketones, the band due to the n ! p transition is
generally observed around 270 300 nm, with a molar absorptivity of 1 2 L/
mol cm. In unsaturated carbonyl compounds, the absorption spectrum is
usually a composition of ethylenic and carbonyl absorptions. If the carbonyl
group forms part of a conjugated system, the n ! p transition will be observed
in the range of 300 350 nm, with a molar absorptivity in the order of 10 L/
mol cm.
The nature of the substituents in the immediate vicinity of the carbonyl
group affects the wavelength of the n ! p transition. Inductive (i.e., electrostatic) effects tend to alter the energies of the ground n-state and the excited
p -state by roughly an equivalent amount, and therefore the energy of the
n ! p transition tends not to be perturbed by inductive effects. Substituents
that themselves possess nonbonding electrons act as electron donors to the
p-system through a resonance effect, causing increase in the energies of all
p-orbitals (bonding and antibonding) without affecting the energy of the norbital. The effect of resonance effects is to increase the energy of the n ! p
transition, which is manifested in a shift toward shorter wavelengths (commonly
referred to as a blue shift). Replacement of the hydrogen atom in an aldehyde
group by an alkyl group will have the same effect, and hence the n ! p transition of a ketone is always blue-shifted relative to its aldehyde equivalent.
One of the simplest methods to assign a given band as being due to a
n ! p transition is to measure the spectrum of the corresponding molecule
lacking the functional group. For instance, in the absorption spectrum of pyrazine, one observes a weak absorption band around 300 nm, which is not
present in the spectrum of benzene. These findings enable one to deduce that
the band must arise from a transition from the nonbonding orbitals of the nitrogen
atom into the p -network of the aromatic system.
When the absorption spectra of compounds capable of exhibiting a n ! p
transition are measured in the solution phase, one finds that the solvent can exert a
strong influence on the energy of the transition. For instance, the energy of the
n ! p transition will shift to shorter wavelengths when the solvent is
changed from hydrocarbon to alcohol. This behavior has been explained that
in the alcoholic solvent, the carbonyl group is hydrogen-bonded with the
solvent as part of the overall solvation process, which stabilizes the ground
state of the compound. The hydrogen bonding has no effect on the excited
p -state, and hence the lowering of the energy of the ground n-state relative to
the excited p -state is responsible for the increase in energy of the n ! p
114
Brittain
transition in the interacting alcoholic solvent relative to the energy in the noninteracting hydrocarbon solvent.
Another characteristic of n ! p transitions is that they are often strongly
affected by the acidity of their solution. For instance, the n ! p transition of
pyridine disappears in acidic solutions owing to the reaction:
Once the pyridine nitrogen becomes protonated, the nonbonding electrons

become involved in bonding, which in turn changes the overall pattern of molecular orbitals and their associated electronic transitions.
p ! p Transitions
A p ! p transition entails excitation from a p-electron system and can therefore occur in any molecule containing such a system if symmetry allows. Generally, such organic molecules can be characterized as containing two general types
of p-systems, namely one where a localized p-bond exists and another where a
delocalized p-bonding system exists.
Ethylene contains a total of 12 valence electrons, with 10 of these being
contained within the s-bonding system and the remaining two in the localized
p-bonding orbital. The electronic configurations of the ground and excited
states of ethylene can therefore be written as:
C2 H4 (G)
core (pCC )2
C2 H4 (E1 )
core (pCC )1 (pCC )1
Ethylene belongs to the D2h point group, whose character table is provided in
Figure 4. A consideration of the symmetry of the molecular orbitals indicates
that the p-orbital transforms as the B3u representation and that the p -orbital
transforms as the B2g representation. Because all of the molecular orbitals in
the ground state of the molecule are completely full, it must transform as the
totally symmetric Ag representation. To derive the symmetry of the E1 excited
state, the direct product of the B3u and B2g representations is taken, which
shows that the E1 excited state transforms as the B1u representation.
Examination of the D2h character table of Figure 4 indicates that the
Ag ! B1u, p ! p , transition will be electric dipole allowed and polarized
along the internuclear z-axis. Because this band is allowed by symmetry, one
would anticipate that its degree of absorptivity would be large. In fact, ethylene
Figure 4
115
The character table of the D2h point group.
exhibits a strong absorbance band at 174 nm, characterized by a molar absorptivity of 16,000 L/mol cm.
Benzene contains a total of 30 valence electrons, with 24 of these being
contained within the s-bonding system and the remaining six the delocalized
p-bonding system. Owing to its hexagonal and planar structure, it is not really
possible to discuss the MO sequence of benzene without first considering its
point group symmetry. The compound belongs to the D6h point group, whose
character table is shown in Figure 5. One of the interesting aspects of this
point group is the existence of the E-type representations, because MOs
having E-type symmetry and transforming according to one of the E representations will be capable of degeneracy (i.e., two MOs characterized by the same
energy).
In benzene, the six 2pz orbitals (perpendicular to the molecular plane) of
the carbon atoms become six p-type molecular orbitals. The bonding p-orbital
of lowest energy has A2u symmetry, whereas the other two bonding p-orbitals
of higher energy are found to be degenerate and transform according to the
E1g representation. The lowest energy antibonding p-orbital is of the degenerate
E2u symmetry, and the highest energy antibonding p-orbital transforms according to the B1g representation. The electronic configurations of the ground and
lowest excited states of benzene can therefore be written as:
C6 H6 (G) core
(pL2u )2 (pE1g )4
C6 H6 (E1 ) core
(pL2u )2 (pE1g )3 (pE2u )1
It can be shown that the E1 configuration will lead to the existence of three
exhibited electronic states having E1u, B2u, and B1u symmetry. Because the electrons in the (pE1g )3 (pE2u )1 configuration can occupy those orbitals either with
116
Figure 5
Brittain
The character table of the D6h point group.
spins paired or unpaired, each of the electronic states may exist as either as a
singlet (spins paired, so 2S 1 1) or as a triplet state (spins unpaired, so
2S 1 3). Because all of the MOs in the ground state of the molecule are completely full, it must transform as the totally symmetric A1g representation.
The theory therefore predicts the existence of three possible spin-allowed
p ! p transitions for benzene, and the symmetry rules can be used to conclude
that only one of these transitions is actually allowed. Nevertheless, various symmetry breaking mechanisms enable all three electronic transitions of benzene
to be observed in its absorption spectrum. A weak band at approximately
255 nm (molar absorptivity of 200 L/mol cm) is assigned to the symmetryforbidden 1A1g ! 1B2u transition. The somewhat stronger band of the
symmetry-forbidden 1A1g ! 1B1u, transition is located at approximately
200 nm (molar absorptivity of 6300 L/mol cm), and is in fact observed as a
long-wavelength shoulder on the intense 1A1g ! 1E1u transition, which has its
maximum at 180 nm (molar absorptivity of 100,000 L/mol cm).
The effect of solvents on the energies of p ! p transitions is opposite to
that previously described for n ! p transitions. In the case of benzene, the distribution of p-electrons in the ground state is symmetric and thus is not perturbed
when the molecule is in a polar or nonpolar environment. On the other hand, the
distribution of p-electrons in the excited state is antisymmetric, and hence the
molecular orbital will be stabilized in a polar environment. Therefore, it is
117
generally found that the energy of a p ! p transition undergoes a shift to longer

wavelength (known as a red shift) when the polarity of the solvent dissolving
an aromatic molecule is increased. However, the generality of this rule is sometimes broken when the solvent is capable of undergoing substantial hydrogen
bonding with the solute. For example, the absorption spectrum of b-naphthol dissolved in dioxane exhibits additional absorption bands associated with hydrogenbonded species relative to those of the compound dissolved in hexane.
Chromophores
As discussed earlier, the molecular orbitals of many functional groups are indeed
localized on the grouping of atoms and are only secondarily perturbed by the
remainder of the molecule. In other words, the absorption bands of a related
series of molecules containing a particular chemical group are similar and
occur in approximately the same region of the spectrum. Such groups are
termed chromophores (literally color carriers). The transferability of a group
absorption band and intensity over a given wavelength range from one molecule
to another requires the existence of a lack of interaction between the chromophore and the rest of the molecule. The absorption bands of a number of chromophores were summarized by Guillory (19), and some of these have been
abstracted in Table 2.
Recognizing that most transitions observed in the UV or VIS region of the
spectrum are either n ! p or p ! p in character, Scott has provided an extensive summary of the properties of various chromophores (21). In addition, Scott
has also developed a number of rules for the prediction of the absorption bands in
substituted benzene derivatives. For instance, the absorption maximum of phenol
is observed at 230 nm. Knowing that the effect of additional substituents on the
phenol system would cause a red shift of the absorption, Scott developed tables
for the estimation of the band maximum. For instance, the contribution for a
amine group at the ortho, meta, and para positions would be 15, 15, and
58, so that the absorption maximum would shift to 245 nm for the ortho- and
meta-substituted aminophenols and 288 nm for para-aminophenol.
Charge Transfer Transitions
For molecules containing two chromophores, or a molecule formed by the complexation of a donor and an acceptor, one often can measure broad featureless
absorption bands that are not observed with either of the individual components.
Because these bands often entail the redistribution of electron density from one
chromophore to the other, they are frequently denoted as charge transfer
bands. For example, the absorption spectra of 1,2,4,5-tetramethyl-benzene, tetracyano-ethylene, and their 1:1 mixture are shown in Figure 6, with the strong band
at approximately 490 nm, which is due to the charge transfer absorption band.
Owing to the loose nature of the bonding between the donor and acceptor in
the ground state of the complex, charge transfer bands are generally very broad
118
Table 2
Brittain
Absorption Spectra of Selected Molecules Containing Chromophores
Group
Example
C5
5C
5CH2
H2C5
C5
5O
H2C5
5O
2
2OH
CH32
2OH
2
2NO2
CH32
2NO2
2
2NH2
CH32
2NH2
Penyl
C6H6
Naphthyl
C10H8
Anthryl
C14H10
Wavelength
(nm)
Molar absorptivity
(L/mol cm)
182.5
174.4
170.4
162.0
295.0
185.0
183.0
150.0
277.5
210.0
215.0
190.5
255.0
200.0
180.0
211.0
270.0
221.0
360.0
250.0
250
16,000
16,500
10,000
10
20,000
200
1,900
10
10,000
580
3,200
200
6,300
100,000
250
5,000
100,000
6,000
150,000
Source: Adapted from Ref. 19.
and show no fine structure due to an underlying vibrational structure. Such bands
will frequently change energy when the donor acceptor complexes are crystallized, and the existence of polymorphic structures will often yield differently
colored solids. For example, the charge-transfer complex formed by picric acid
and 2-iodoaniline can be obtained in two polymorphs, depending on the crystallization process used (22). Monoclinic Form-I contains four cations (2-iodoanilinium) and four anions (picrate) per unit cell, and its crystals are yellow in
color. The triclinic Form-II contains two iodoanilinium cations and two picrate
anions per unit cell, and exists as a green solid. In Form-I, the anion cation
pairs are stacked to form segregated columns, in which the same ion types are
stacked with each other. In Form-II, the anions and cations are alternately
stacked to form continuous columns. The structural arrangement of Form-I
does not allow for the generation of a charge-transfer absorption band, and
hence, the solid maintains the yellow color of picric acid. On the other hand,
the alternate stacking present in Form-II facilitates the overlap of suitable
molecular orbitals to yield a new charge-transfer absorption band, which is
equally as absorptive as the locally excited band of the picrate anion.
119
Figure 6 Absorption spectra of 1,2,4,5-tetramethylbenzene (long dashed trace), tetracyanoethylene (short dashed trace), and their 1:1 donor acceptor complex (solid trace).
The compounds of transition elements often exhibit absorption bands that

are due to intramolecular charge transfer transitions. The transitions can involve
transfer of electron density from the coordinated ligands to the empty d-orbitals
of the metal ions or transfer of electron density from metal ion d-orbitals to empty
p -orbitals of the ligands.
REFERENCES
1. Jaffe GZ Physik 1934; 87: 535.
2. Pauling L. The Nature of the Chemical Bond, 3rd edn. Ithaca: Cornell University
Press, 1960.
4. Slater JC. Quantum Theory of Molecules and Solids, volume 1. New York: Mc-Graw
Hill Co., 1963.
5. Ballhausen CJ, Gray HB. Molecular Orbital Theory. Reading, MA: Benjamin/
Cummings Pub., 1964.
6. Flurry RL. Molecular Orbital Theories of Bonding in Organic Molecules. New York:
7. McGlynn SP, Vanquickenborne LG, Kinoshita M, Carroll DG. Introduction to
Applied Quantum Chemistry. New York: Holt, Rinehart and Winston, 1972.
120
Brittain
8. Hanna MW. Molecular Quantum Mechanics in Chemistry. 3rd edn. Menlo Park, CA:
Benjamin/Cummings Pub., 1981.
9. Herzberg G. Spectra of Diatomic Molecules, 2nd edn. Cincinnati, OH: Van Nostrand
Reinhold Co., 1950.
10. Tinkham M. Group Theory and Quantum Mechanics. New York: McGraw-Hill Book
Co., 1964.
11. Hochstrasser RM. Molecular Aspects of Symmetry. New York: W.A. Benjamin Inc.,
1966.
12. Orchin M, Jaffe HH. Symmetry, Orbitals, and Spectra. New York: Wiley-Interscience,
1971.
13. Cotton FA. Chemical Applications of Group Theory, 2nd edn. New York:
Wiley-Interscience, 1971.
14. Harris DC, Bertolucci MD. Symmetry and Spectroscopy. New York: Oxford
University Press, 1978.
15. Herzberg G. Electronic Spectra and Electronic Structure of Polyatomic Molecules.
Princeton, NJ: Van Nostrand Reinhold Co., 1967.
16. Barrow GM. Introduction to Molecular Spectroscopy. New York: McGraw-Hill Book
Co., 1962.
Winston, 1964.
18. Brittain EFH, George WO, Wells CHJ. Introduction to Molecular Spectroscopy
Theory and Experiment. London: Academic Press, 1970.
19. Guillory WA. Introduction to Molecular Structure and Spectroscopy. Boston, MA:
Allyn and Bacon, Inc., 1977.
20. McHale JL. Molecular Spectroscopy. Upper Saddle River, NJ: Prentice Hall, 1999.
21. Scott AI. Interpretation of the Ultraviolet Spectra of Natural Products. New York:
Pergamon Press, 1964.
22. Tanaka M, Matsui H, Mizoguchi J-I, Kashino S. Bull Chem Soc Jap 1994; 67: 1572.
5
UV/VIS Reflectance Spectroscopy
Harry G. Brittain
INTRODUCTION
Everyone capable of sight is a practitioner of reflectance spectroscopy in the visible
region of the spectrum. We are constantly bathed by light, some which is characterized by wavelengths between 400 nm (i.e., blue light) and 650 nm (i.e., red light).
When our eyes perceive, for instance, a red piece of clothing, we are actually observing the reflected portion of the irradiation light. The clothing absorbs the blue light,
reflects the unabsorbed red light, our eyes perceive the reflected red light, and our
mind then interprets this measurement to conclude that the piece of clothing was
actually red. However, our perception can be altered by changing the conditions
of the experiment. If the same piece of clothing is now irradiated only with blue
light, that light will be completely absorbed, and no light will be reflected.
Because in that case our eye would not be able to detect any reflected color from
the clothing, we would then conclude that the same piece of clothing was black.
Everyday experience therefore defines the process of reflectance spectroscopy. An object is irradiated with electromagnetic radiation, then the
object absorbs whatever wavelengths it is capable of absorbing, the unabsorbed
light is reflected, and ultimately quantitated by some sort of optical detector.
From the discussions of the previous chapters, it is now completely clear that
the criteria for wavelength absorption will be that dictated by the molecular
orbitals of the molecules involved in the absorption process.
Transmission spectroscopy has been used for a long time to study the
absorption spectroscopy of molecules dissolved in a solution phase, but owing
to the opaque nature of most solids, such methodology cannot usually be used
to study the electronic spectroscopy of molecules in the solid state. Reflectance
121
122
Brittain
spectroscopy takes advantage of the scattering of incident energy off a solid

surface, so instead of measuring the light intensity before and after passing
through a specimen, one measures the amount of light reflected off the sample
surface with the aid of a suitably modified instrumental system.
The most important applications of reflectance spectroscopy in the ultraviolet
and visible regions of the spectrum (UV/VIS) have been in the fields of color
measurement and color matching, areas of great importance to the dye, paint,
paper, textile, and printing industries. The pharmaceutical industry has similar
interests in that the use of coloring agents in formulations requires specification.
It is logical to conclude that because most pharmaceutical agents are administered
as solid dosage forms, the most appropriate forms of UV/VIS spectroscopy would
make use of reflectance techniques. As will be seen, reflectance spectroscopy has
been used by a number of workers to study the kinetics and mechanisms associated
with a variety of reactions, which were found to take place in the solid state.
The scope of this chapter will be mostly concerned with investigations of
diffuse reflectance work performed in the UV/VIS region of the spectrum and
with colors that can be perceived by the human eye. An enormous amount of
very important work has been conducted using irradiation wavelengths in the
near-infrared region of the spectrum, but that aspect will be covered elsewhere
in this book. Several standard references on UV/VIS reflectance spectroscopy
are available (1 5).
CATEGORIES OF LIGHT REFLECTANCE
When light impinges on a nonsmooth surface, its spatial distribution is altered so
that it travels in numerous directions, a phenomenon known as scatter. One
defines the specular reflectance, RS, as the ratio of light reflected according to
the laws of regular reflection to that of the incident light energy. The diffuse
reflectance, RD, is then defined as the ratio of light diffusely reflected in all directions (apart from that of direct reflection) to that of the incident light energy. The
total reflectance is all of the light that is reflected by a solid and is defined as the
sum of the specular and diffuse reflectances.
Specular reflectance (also called direct or mirror-type reflectance) represents the ordinary reflection of the incident light by the surface(s) of the interacting object(s) and is characterized by a complete absence of light transmission
through the interiors of the object(s). The characteristics of reflected energy
depend on wavelength, degree of polarization, and angle with respect to the
surface. For instance, clear glass will reflect approximately 4% of the incident
light if it impinges on the glass surface at a high angle. As the angle of incidence
decreases, an increasingly higher percentage of the light becomes reflected until,
at the grazing angle, the entire incident light is reflected. The theory associated
with specular reflectance is fully outlined in chapter 2 of Ref. (2).
In contrast, diffuse reflectance is associated with the radiation, which penetrates into the particles to some extent and then emerges from the bulk solid. This
123
light will exhibit spectral characteristics, which are modified from those of the
incident beam by the wavelength-specific attenuation that took place as a
result of induced electronic transitions caused by light absorption at the boundaries of the component particles. The attenuation of the diffuse part of the reflectance caused by absorption can be expressed by the Bouguer-Lambert law:
I I0 e1d
(1)
where I is the intensity of diffusively reflected light, I0 is the incident light intensity, 1 is the molar extinction coefficient, and d is the mean thickness of
penetration.
Specifically referring to powdered solids, diffusion of light is caused by the
effects induced by multiple scattering of the individual particles within
the interior of the sample. Equation (1) indicates that the distribution of wavelengths leaving the solid will be modified from that of the incident light by
any wavelength-selective absorption within the sample. The electromagnetic
radiation, which is diffusely reflected, must penetrate the sample before returning
to the surface through multiple scattering. If one defines the depth of penetration
of light into a layer as the distance required for the intensity to decrease by the
factor, e, then the mean depth is defined by:
d 1=k
(2)
In Equation (2), k is the absorption coefficient of the solid in question and

represents the contribution to the index of refraction due to absorption.
The strength of absorption of a sample can be related to the wavelength of
the incident light, l, as long as the same units are used throughout:
Strong absorber ) d , l
Weak absorber ) d . l
(3)
(4)
From these relationships, one can see that for strong absorbers (i.e., d is
small), the degree of diffuse reflectance will be small so that the total reflectance
will be dominated by the specular reflectance contribution. On the other hand, for
weak absorbers (i.e., d is large), the degree of diffuse reflectance will be large
because such materials can absorb a considerable amount of the incident light
owing to the existence of numerous multiple reflections.
The size of the particles making up the interacting layer will also exert a
strong effect on the diffuse reflectance. Because samples having a fine particle
size distribution will scatter the incident radiation more efficiently than the
larger particles, one finds that the absorption of a solid will be more pronounced
for an ensemble of large particles than a similar ensemble of small particles. To
illustrate this point, the reflectance spectra obtained for the xanthene dye eosin-Y
absorbed onto fine particles of anhydrous lactose (i.e., material that passed
through a 100-mesh sieve) and onto coarse anhydrous lactose particles (i.e.,
material that was retained on a 60-mesh sieve) are shown in Figure 1.
124
Brittain
100
Percent Reflectance
80
60
40
20
0
450
500
550
600
650
700
Wavelength (nm)
750
800
Figure 1 Reflectance spectra of eosin-Y absorbed onto anhydrous lactose passed

through a 100-mesh sieve (solid trace) and onto anhydrous lactose retained on a 60mesh sieve (dashed trace). The data were obtained using a Varian DMS 200 UV/VIS
spectrophotometer fitted with a diffuse reflectance accessory. Source: Adapted from HG
Brittain, (unpublished results).
Owing to the complementary nature of specular and diffuse reflectances, it

is most essential to design experimental conditions for which the diffuse reflectance component of the total reflectance is maximized. High levels of specular
reflectance are undesirable in this work, and both the collection optics and sample
preparation must be optimized to minimize the effects of specular reflectance.
THEORY ASSOCIATED WITH DIFFUSE REFLECTANCE
MEASUREMENTS
The theory associated with diffuse reflectance has been fully developed, and
interested readers can obtain complete details in the texts by Wendlandt and
Hecht (2), and Kortum (3). The most useful theory concerning diffuse reflectance
was developed by Kubelka and Munk (6,7), the derivation of which was given
by Frei and MacNeil (4) in appendix 1 of their text. In this narrative, the
essential aspects of the Kubelka-Munk theory will be summarized.
Consider a layer of substance, having a thickness equal to d, that is irradiated by a monochromatic beam of incident light having intensity equal to I0.
125
The intensity of the light along the incident direction (taken to be the x-axis) is
represented by I, and the light intensity in the negative x-direction (caused by
scatter) is represented by J. If the maximum depth of penetration in the layer
is dx, then the average path of the radiation (df) for these two components is
given by:

1
dI
dfI dx
df
(5)
I cos f df
and

dfI dx

1
dJ
df
J cos f df
(6)
In Equations (5) and (6), the integration limits are 0 to p/2, and the
quantities (dI/df) and (dJ/df) represent the angular distribution of the light.
Assuming conditions for ideal diffuse radiation:
dI
I sin 2f
df
dJ
J sin 2f
df
(7)
(8)
If one defines the absorption coefficient (k) as the efficiency of absorbance,

and the scattering coefficient (s) as the efficiency of scattering, then the component absorbed in the layer will equal (kI dx) and the back-scatter component
will equal (sI dx). The amount of scattering in the forward direction will equal
(sJ dx). Therefore, the change in light intensity along the direction of light propagation within the layer will be given by the sum of these factors, or:
dI sJ dx (k s)I dx
(9)
and the change in light intensity in the negative direction is given by:
dJ (k s)J dx sI dx
(10)
Integration of Equations (9) and (10) yields:

I A(1 b)esx B(1 b)esx
J A(1 b)e
sx
B(1 b)e
s x
(11)
(12)
In Equations (11) and (12), the following definitions have been used:
s k(k 2s)1=2
b k=(k 2s)
1=2
(13)
(14)
The constants A and B of Equations (11) and (12) are determined by the
limiting conditions of the experiment. If the entire thickness of the layer is d,
126
Brittain
then it can be shown that:

(1 b)esd I0
(1 b)2 esd (1 b)2 esd
(1 b)esd I0
B
(1 b)2 esd (1 b)2 esd
(15)
(16)
The transmission (T ) through the layer would be given by I (when x d )

divided by I0, or:
4b
(1
(1 b)2 esd
2b
(1 b)2 sin(hsd) 2b cos(hsd)
b)2 esd
(17)
(18)
The diffuse reflectance (R) of the layer would be given by I (when x 0)

divided by I0, or:
(1 b)2 esd esd
(1 b)2 esd (1 b)2 esd
(1 b)2 sin(hsd)
2
(1 b) sin(hsd) 2b cos(hsd)
(19)
(20)
For the case of pure transmission, s 0 and b 1, Equation (17) reduces

to the Bouguer-Lambert law quoted previously as Equation (1), and R 0.
For a layer having infinite thickness, d approaches zero, and one obtains R1
(the absolute reflectance of an effectively infinitely thick layer) as:
R1
s k k(k 2s)1=2
s
(21)
These conditions can be achieved experimentally if four conditions are

met. The first of these is that the extent of the horizontal layer is so large relative
to its thickness, and so the light diffused horizontally out of the edges of the
sample can be neglected relative to the light moving perpendicular to the layer
front. The second condition is that the material in the layer is homogeneous in
its composition for the entire distance through which the light passes. The
third condition is that the light incident on the top of the layer is so perfectly
diffused that all points on the surface receive equal irradiation. The final condition is that the top of the sample has the same index of refraction as the
medium (typically air) in contact with the sample. Equation (21) is usually
127
rearranged to the more familiar Kubelka-Munk equation:

k (1 R1 )2
s
2R1
(22)
To study the UV or VIS absorption spectroscopy of a solid material, the

radiation reflected from the surface of the sample is detected and recorded as a
function of the incident wavelength. The choice of standard material is subject
to the discretion of the investigator but is usually either MgO or BaSO4. The
most appropriate standard is one whose matrix permits the most useful data to
be obtained on the system of interest. For such work, it is common practice
to measure reflectance values obtained at each wavelength relative to the
measured reflectance of a standard at that wavelength:
R01 R1 (sample)=R1 (standard)
(23)
0
R1
and to use
in the Kubelka-Munk equation. The term on the left side of
Equation (22) is often termed the remission function (or the Kubelka-Munk function) and is frequently identified by f (R1). Equation (22) indicates that a linear
relationship should exist between f (R1) and the sample absorption spectrum.
Taking the logarithm of the Kubelka-Munk function yields:
log f (R1 ) log k log s
(24)
It is concluded that when log[ f (R1)] is plotted against the analyzing wavelength, the resulting diffuse reflectance spectrum will be identical to the transmission spectrum of the compound. The only difference will be a displacement
in the ordinate scale by a quantity equal to 2(log s).
It needs to be reemphasized that the form of the Kubelka-Munk equation is
valid only for weak absorbing systems, does not allow for any significant contribution from specular reflectance, and requires that the particle size of the powdered sample be relatively small. However, the theory for more complicated
systems is available, and the equations pertaining to a wide range of possibilities
are available (7). Other approaches to the theory of reflectance spectroscopy have
been made using the statistical single particle theory (7 9).
INSTRUMENTATION FOR THE MEASUREMENT OF DIFFUSE

REFLECTANCE
Probably, the first instrument designed for reflectance studies was that described
by Nutting (10), which was improved by Taylor (11). Since that time, a large
number of instruments have been proposed, and numerous modifications were
suggested to continually improve signal-to-noise ratios (2 4).
The basic requirements for a diffuse reflectance spectrometer are a light
source and a monochromator (to obtain monochromatic radiation), an integrating
sphere, and a detection and recording system. The incident energy from the
128
Brittain
source need not be diffuse in itself, and good results are obtained using direct
irradiation of the sample. The light is rendered monochromatic by either a monochromator or filter arrangement, split, and allowed to fall on both the standard and
sample faces. The light diffusely reflected by the reference and the sample are
collected by the internal reflections of the sphere, and the intensity is ultimately
measured by the detector. The data are processed (usually according to the
Kubelka-Munk equation) and displayed on a suitable device.
The essential and unique aspect of a diffuse reflectance spectrometer is
the integrating sphere, the design of which is shown in Figure 2. The inner surfaces of the sphere are coated with highly reflective materials (typically either
MgO or BaSO4) to minimize absorption of the diffuse reflectance. The efficiency
of the sphere is defined as the energy lost at the detector port divided by the total
energy lost within the sphere. Detailed discussions are available regarding the
design of and theory associated with integrating spheres (2 4). For an ideal
sphere, the intensity due to reflected light at any point is independent of the
spatial distribution within the sphere. Detection of the light intensity at any
point is therefore an accurate determination of the diffuse reflectance.
A variety of sample-handling devices can be incorporated into the integrating sphere so as to permit the performance of necessary experiments. For
instance, the addition of a hot stage to the sphere allows the reflectance of
Figure 2 Schematic diagram of the integrating sphere portion of a diffuse reflectance

spectrometer, illustrating the key elements of the optical train. Although the detector has
been placed in the plane of the sample and reference materials, it would be mounted orthogonal to the plane created by the intersection of the optical beams in common practice.
129
materials to be obtained at elevated temperatures (12,13). This particular aspect

has been found to be especially important in the detection and monitoring of the
progress of solid-state chemical reactions. The technique of high-temperature
diffuse reflectance spectroscopy becomes particularly useful with the combination of a gas evolution detection technique (14).
To meet the requirements of the Kubelka-Munk theory, the average particle
size of the sample should be small and ideally between 0.1 and 1 mm. At particle
sizes less than 0.1 mm, the scattering coefficient exhibits an unacceptably strong
dependence on the frequency of the incident radiation. For particles larger than
1 mm, the specular reflectance becomes excessively large and depresses the magnitude of the diffuse reflectance spectrum. The effect of particle size on the quality
of the diffuse spectrum was illustrated previously in Figure 1. The thickness of the
sample layer should be between 1 and 5 mm, ensuring that the path length meets
the requirements of the Kubelka-Munk theory so as to be infinitely deep.
QUANTITATIVE MEASUREMENT OF COLOR

Color is basically a perception that is developed in the mind of a given individual,
and consequently different people can perceive a particular color in alternate
fashions (15 17). Such variability in interpretation causes great difficulty in
the evaluation of color-related phenomena and leads to subjective rather than
objective judgments. For obvious reasons, the development of a quantitative
method for color determination is highly desirable in that a good system would
lead to elimination of the subjectivity associated with a visual interpretative
measurement (18).
Although other systems have been developed for the quantitative
expression of color (19,20), only the CIE (Commission Internationale de lEclairage) system and its multiple incarnations have gained general acceptance (22).
The evolution of the CIE system and its alternates has been detailed in Refs.
(20,23). The CIE system assumes that color may be expressed as the summation
of selected spectral components in a three-dimensional manner, with the three
primary colors normally added for such purposes being blue, green, and red.
The CIE system is based on the fact that human sight is trichromatic in its
color perception, and that two stimuli will produce the same color if each of
the three tristimulus values (X, Y, and Z ) are equal for the two:
X k S(l)R(l)x(l) dl
Y k S(l)R(l)y(l) dl
Z k S(l)R(l)z(l) dl
(25)
(26)
(27)
130
Brittain
where:
k
100
S(l)y(l) dl
(28)
In Equations (25 28), S(l) is the spectral power distribution of the illuminant, and R(l) is the spectral reflectance factor of the object. x(l), y(l), and z(l)
are the color-matching functions of the observer. In the usual practice, k is
defined so that the tristimulus value, Y, for a perfect reflecting diffusor [serving
as the reference for R(l)] equals 100. Using the functions proposed by the CIE
in 1931, y(l) was made identical to the spectral photopic luminous efficiency
function, and consequently its tristimulus value, Y, is a measure of the brightness
of objects. The X and Z values describe aspects of color, which permit identification with various spectral regions.
The x(l), y(l), and z(l) terms were derived by the CIE from data obtained
in visual experiments where observers matched colors obtained by the mixing of
the blue, green, and red primary colors. The average result for human observers
were defined as the CIE 1931 28 standard observer, and the wavelength dependencies of these color matching functions are illustrated in Figure 3 (23,24).
The proper implementation of the CIE system requires the use of a standard
illumination source for calculation of the tristimulus values. Three standard sources were recommended in the 1931 CIE system and may be presented
in terms of color temperatures (the temperature at which the color of a
Figure 3 Color matching functions x(l), y(l), and z(l) of the CIE 1931 28 standard
observer. Source: Adapted from Ref. 23.
131
black-body radiator matches that of the illuminant). The simplest source is an

incandescent lamp, operating at a color temperature of 2856 K. The other two
sources are combinations of lamps and solution filters designed to provide the
equivalent of sunlight at noon or the daylight associated with an overcast sky.
The latter two sources are equivalent to color temperatures of 5000 and
6800 K, respectively.
Because color is a perceived quantity, a strict mathematical relation relating tristimulus values to a concept of color is not possible. However, an equation
has been proposed that relates the perception of color to these values. The trichromatic equation for tristimulus values is normally put into the general form:
color (C) x(X) y(Y) z(Z)
(29)
where x, y, and z represent the chromaticity coordinates of C, obtained via the

following relations:
x X=(X Y Z)
(30)
y Y=(X Y Z)
(31)
z Z=(X Y Z)
(32)
Only two of the three chromaticity coordinates need to be actually specified, since:
1xyz
(33)
A chromaticity diagram can therefore be drawn presenting colors in terms

of their x and y coordinates, which are often termed as hue and saturation,
respectively. A diagram of this type is presented in Figure 4 (23,24).
In terms of the chromaticity diagram, it can be said that two colors having
the same chromaticities (x, y) and the same luminous reflectance (Y ) are matched.
Technically, no information is provided about the appearance of two matched
colors, although subjective judgments about specific colors identified by a pair
of (x, y) coordinates are often made.
The central point (marked W in Fig. 4) of a chromaticity diagram is termed
the achromatic point and represents the combination of x and y values that yield
white light. A straight line passing through W would connect complementary
colors. Combination of the colors specified at the endpoints of this line in the relative amounts indicated by the distances from their respective points to W will
result in the generation of white light.
The actual determination of color is made with a photoelectric tristimulus
colorimeter fitted with a CIE illuminating source. Measurements are made relative to a standard (usually MgO or BaSO4) taken through each of the provided
filters. The experimental readings are converted to tristimulus values by means
of instrumental factors provided by the instrument manufacturer, and taking
the filter specifications into account. The tristimulus values are converted to
chromaticity coordinates using Equations (30 32). The accuracy of this
132
Brittain
Figure 4 Chromaticity diagram presented in terms of the x and y coordinates. Source:

Adapted from Ref. 22.
approach is determined by how well the combination of detector and tristimulus

filters approximates the characteristics of the CIE standard observer.
With the introduction of computers and microprocessor-controlled instrumentation, it has become possible to use spectrophotometry to obtain far more
accurate determinations of color and to obtain tristimulus values after integration
of the data according to Equations (25 27). This degree of sophistication permits
the use of more advanced methods of color quantitation, such as the 1976 CIE
L u v system (25) or other systems.
Raff has provided a discussion summarizing how the technology of the CIE
system could be applied to the characterization of materials having pharmaceutical interest (26). At the time when this article was written, the U.S. Food
and Drug Administration (FDA) was just becoming concerned about the stability
of coloring agents and dyes, and the quantitative CIE system held a definite
appeal for those seeking to eliminate subjectivity from such determinations.
In this work, Raff provided the reflectance spectra of several pigments and
133
attempted to study the changes in color that took place upon exposure to strong
illumination.
In a subsequent work, Raff (27) used the CIE system to quantify the colors
that could be obtained when using FD&C aluminum lakes as colorants in tablet
formulations. He reported on the concentration dependence of the tristimulus
values obtained when calcium sulfate dihydrate was compressed with various
amounts of FD&C Blue No. 2 aluminum lake, and one example of the reported
data is found in Table 1. The Y tristimulus value may be taken as the relative
lightness of the tablet surface, with a value of 1.0 being the maximum. It is
evident that increasing colorant concentrations decrease the apparent brightness
of the tablets. The chromaticity coordinates, x and y, indicate that the apparent
color of the tablets shifts as the colorant levels are raised, and that no limiting
color can be reached solely through the use of FD&C Blue No. 2 aluminum
lake. Because the perceived color cannot be saturated, careful control of the colorant level in this particular formulation is required to ensure that all batches
yield equivalent appearances.
Goodhart et al. (28) used the CIE parameters to evaluate the chromaticity
coordinates and brightness of nearly 50 pharmaceutical colorants compressed
with lactose. The colorants were categorized as belonging to blue, black,
brown, green, orange, red, violet, and yellow classes, and several iron oxide colorants were also studied. The aim of this work was to produce a data set that could
be used to match any given color through the combination of pharmaceutically
acceptable agents.
In an extension of their work, Goodhart et al. (29) developed a system
whereby the final desired color of a compressed table formulation was first
chosen from a standard color chart [such as the Munsell compilation (23)].
This color was then analyzed as to its CIE parameters, and these parameters
were in turn used to develop a colorant combination that would produce a
match of the desired color. The ultimate end of this work was to produce a database of sufficient depth that the empirical nature of color-matching could be
eliminated.
Table 1 Tristimulus Values Obtained on a Series of Tablets Colored with FD&C Blue
No. 2 Aluminum Lake
Percent dye (w/w)
0.065
0.29
0.90
2.60
6.20
10.00
0.744
0.568
0.439
0.279
0.175
0.142
0.762
0.582
0.447
0.280
0.172
0.138
1.012
0.860
0.718
0.518
0.362
0.304
0.296
0.282
0.274
0.259
0.247
0.243
0.303
0.289
0.279
0.260
0.243
0.236
134
Brittain
Bogdansky (30) had used a combination of the CIE and alternate color
systems to deduce color parameters associated with tablet colorants. In this
work, thin color dispersions on chromatography paper were made, and the tristimulus values determined through appropriate colorimetry. The materials
studied were ordered through their perceived color, and these judgments correlated with the quantitative color parameters. Most importantly, an evaluation procedure was established for the acceptability criteria to be used in judging the
range of chromaticity values, which signified equivalence in the color of solid
dosage forms produced as different lots. The compilation of tristimulus and
chromaticity information was shown to provide a permanent description of a
colored sample and to define any difference between an analyte and its standard.
The effect of particle size, and hence dispersion, on the coloring properties
of aluminum lake dyes has been studied through quantitative measurement of
color in compressed formulations (31). It was found that reduction in the particle
size for the input lake material resulted in an increase in color strength, and those
particles of submicron size contributed greatly to the observed effects. Analysis
of the formulations using the parameters of the 1931 CIE system could only lead
to a qualitative estimation of the effects, but the use of the 1976 CIE L u v
system provided a superior evaluation of the trends. With the latter system, the
effects of dispersion on hue, chroma, lightness, and total color differences
were quantitatively related to human visual perception.
The appearance testing of tablets through measurement of color changes
has been automated through the use of fiber optic probes and factor analysis of
the data (32). Good correlation between measured chromaticity parameters and
visual subjective judgment was demonstrated, with samples of differing
degrees of whiteness being used to develop the correlation. It was pointed out,
however, that surface defects on the analyzed materials could compromise the
quality of the correlation and that more sophisticated methods for data evaluation
would be useful.
APPLICATIONS OF UV/VIS DIFFUSE REFLECTANCE
SPECTROSCOPY TO THE STUDY OF SOLIDS HAVING
PHARMACEUTICAL INTEREST
Although diffuse reflectance spectroscopy in the UV/VIS region of the spectrum
has not played a major role in the characterization of pharmaceutical solids, it can
be used to obtain basic information on formulated products. For instance, the
color of formulations containing yellow nifumic acid with a variety of white
excipients has been used as a means to evaluate the organization and compactibility of binary powder mixtures (33).
UV/VIS diffuse reflectance spectroscopy has been used, however, with
great success to follow the course of reactions for which color changes were
directly related to stability-indicating parameters. The most important of these
have concerned the color stability of active pharmaceutical ingredients or
135
excipients used in solid dose formulations, where the reflectance spectra were
used to evaluate the relative rates of decomposition among the ingredients.
Another area of investigation concerns the study of either drug excipient or
excipient excipient reactions, and here the technique has been particularly
useful in those instances where color either develops or disappears as the reaction
proceeds.
Studies of Color Stability in Solid Dose Formulations
One of the most important uses for reflectance spectroscopy in the characterization of pharmaceutical solids concerned the stability of coloring agents in
tablet formulations. With the description of a device, which enabled the
surface of intact tablets to be studied (34), the photostability of various dyes
and lakes in tablets was followed (35,36). Exposure of formulations to both
normal and exaggerated light conditions was investigated, and the kinetics of
the photodegradation evaluated. In most cases, the photoreactions appeared to
follow first-order kinetics.
As an example of this type of work, the reflectance spectra of tablets formulated with FD&C Blue No. 1 and exposed to various light intensities are
shown in Figure 5. Under normal illumination (45 foot candles), the decomposition kinetics was found to be modest. When the tablets were exposed to
higher levels of illumination (550 foot candles), degradation proceeding by at
least three different reaction pathways was observed, accompanied by very complicated kinetics (35).
In general, the kinetics associated with the fading of the various dyes
studied was found to vary over wide ranges. Lakes prepared from several
dyes were found to exhibit less photostability than did the corresponding dyes
(36). In most instances, the initial rates of decomposition were significantly
faster than the rates observed at later time points. The technique has also been
applied to studies of the photostability of naturally sourced colorants in
powder beds rather than in compressed products (37).
The ability of various protection schemes to retard the photodecomposition
was investigated by means of diffuse reflectance spectroscopy. Glasses of different colors were evaluated as to their ability to retard the fading of tablets colored
with FD&C Blue No. 1 and D&C Yellow No. 10 (38). The main conclusion of
this work was that amber glass afforded the best protection against the effects
of photoillumination, although any glass containing an effective UV-absorber
could be equally effective. The incorporation of 2,4-dihydroxy-benzophenone
into tablet formulations as an inherent UV-absorber was studied, and its effect
on dye fading rates was evaluated (39). It was found that this compound could
effectively protect against the photodecomposition reaction, as long as its absorption spectrum matched reasonably well with the dye under question.
In a detailed study of the color stability of tablets containing FD&C Red
No. 3 dye, the ability of a variety of sunscreen agents to retard the color loss
136
Brittain
Figure 5 Reflectance spectra of tablets formulated with FD&C Blue No. 1 and exposed
to ordinary illumination for various time periods. Full spectra are shown in the upper box
for tablets (A) as initially prepared, and after (B) 42 days, and (C) 84 days of illumination.
The dependence of dye absorbance on elapsed time is shown in the lower trace. Source:
was investigated (40). The agents were incorporated in films coated upon the
model tablet formulation, and the tablets subsequently exposed to light of 1000
foot candles. The film coating method was not found to yield acceptable stability
when using either glyceryl p-aminobenzoate or salicylates as sunscreens. The
greatest protection against fading was observed when 2-ethoxyethyl p-methoxycinnamate was used as the protection agent.
The stability of a wide variety of certified dyes in tablets has been evaluated
upon exposure to various illumination conditions (41 43). In these works, more
137
detailed studies of dye concentration, exposure time, and irradiation light intensity were performed, and the data analyzed more rigorously. The effect of varying
tablet excipients was also investigated with respect to the decomposition kinetics.
This work finalized with the development of a systematic approach for the testing
of color stability, and the investigators developed a ranking of colorant stabilities
in tablets (43).
Studies of Solid-State Reactions in Formulated Materials
It was recognized very early that diffuse reflectance spectroscopy could be used
to study the interactions of various compounds in a formulation, and the technique has been particularly useful in the characterization of solid-state reactions
(44). Lach concluded that diffuse reflectance spectroscopy could also be used to
verify the potency of a drug in its formulation. In addition, studies conducted
under stress conditions would be useful in the study of drug excipient interactions, drug degradation pathways, and alterations in bioavailability owing to
chemisorption of the drug onto other components in the formulation (44).
In a series of works, Lach et al. (45 47) studied the solid-state interactions
of a variety of compounds with various adjuvants. Working predominantly with
compounds containing conjugated aromatic systems (oxytetracycline, anthracene, phenothiazine, salicylic acid, prednisone, and hydrochlorothiazide), it
was deduced that the complexes formed with the adjuvants were of the donor
acceptor type. Large bathochromic and hyperchromic spectral shifts in absorption maxima upon complexation were inferred to be the result of strong charge
transfer interactions. In the specific instance where the drug possessed a chelation
center and the adjuvant contained a metallic center, the site-specific complexation could be identified as the source of the change in the reflectance spectrum
(47,48).
The stability of ascorbic acid formulations has been successfully studied
using diffuse reflectance spectroscopy because the degradation is accompanied
by a color reaction, and this color change is directly related to the potency of
the active component (49,50). Tablets stored at ambient conditions will gradually
age from white to a yellowish-brown color, with the degree of stability being
greatly affected by the excipients in the formulation. It was determined that
materials containing metal ions (i.e., magnesium and calcium stearate, or talc)
accelerated the color reaction, whereas metal-free excipients (such as stearic
acid or hydrogenated vegetable oil) imparted maximal color stability (50). The
effect of temperature and relative humidity on the stability of the formulations
was also followed using diffuse reflectance spectroscopy, and it was found that
the effect of elevated humidity was more important than the effect of elevated
temperatures.
Diffuse reflectance spectroscopy was used to screen the possible interactions between a large number of adjuvants and several dyes (51). It was concluded that supposedly inert excipients (such as starch or lactose) were capable
138
Brittain
of undergoing significant reactions with the dyes investigated (Red No. 3, Blue
No. 1, and Yellow No. 5). For adjuvants containing metal ions (zinc oxide, or
calcium, magnesium, and aluminum hydroxides), the degree of interaction
could be considerable. It was concluded from these studies that dye excipient
interactions could also be responsible for the lack of color stability in certain
tablet formulations.
The reaction of a primary amine with lactose is accompanied by a browning
of the solids, and the path of such reactions is easily followed by means of diffuse
reflectance spectroscopy. For instance, the reaction of isonicotinic acid hydrazide
(isoniazid) with lactose could be followed through changes in the reflectance
spectrum (52). As may be seen in Figure 6, as the sample was heated for increasing amounts of time a steady decrease in reflectance was noted. The spectral data
were used to deduce the rate constants for the browning reaction at various
heating temperatures, and these rates could be correlated with those obtained
through chemical analysis (formation of isonicotinoyl hydrazone of lactose) of
the mixtures. In addition, the active was found to both chemisorb as well as physisorb onto magnesium oxide, with accompanying changes in the reflectance
spectrum (31).
Figure 6 Reflectance spectra of the reaction products formed when isonicotinic acid
hydrazide reacts with lactose. Reflectance spectra are shown for (A) the initially prepared
material, and for samples illuminated for (B) 4 hours, (C) 10 hours, (D) 21 hours, and (E)
44 hours. Source: Adapted from Ref. 52.
139
A similar study has been conducted in which the interaction of

d-amphetamine sulfate with spray-dried lactose was investigated (53). Upon
storage at elevated temperatures, discoloration of the powder blends was noted,
and the new absorption bands were characterized. A maximum was noted at
340 nm and was attributed to the chemisorption of the amine onto the lactose
particles. The other band appeared at 295 nm and was attributed to the new compound (d-amphetamine-hydroxymethylfurfural) formed as a result of the reaction
between the two. Through the use of Arrhenius plots, the browning rate anticipated for a temperature of 258C was predicted.
The interaction of tetracycline and its derivatives with excipients has
been studied by diffuse reflectance spectroscopy (54,55). Adjuvants containing
metal ions were found to yield the largest degree of change in the reflectance
spectra, with the effects of calcium being the most prominent. Complexation
of the metal ion with specific functional groups on the drug molecules was postulated as the origin of these effects. The complexation would be anticipated to
influence the bioavailability of the compound, and therefore the technique
is demonstrated to provide useful information regarding the degree of drug
excipient compatibility.
Other workers have used the tristimulus parameters to study the kinetics of
decomposition reactions. The fading of tablet colorants was shown to follow firstorder reaction kinetics, with the source of the illumination energy apparently not
affecting the kinetics (49). The effect of excipients on the discoloration of
ascorbic acid in tablet formulations has also been followed through determination
of color changes (57). In this latter work, it was established that lactose and
Emdexw influenced color changes less than sorbitol.
Compounds known to undergo changes in their absorption spectra upon
sorption onto a solid surface are termed adsorptiochromic, and such effects
would be ideally studied by means of diffuse reflectance spectroscopy. In one
such study, the absorption of various spiropyrans onto many different solids
was investigated (58,59). For the compounds studied, the reflectance spectra
were dominated by bands at 550 nm and in the range of 400 500 nm (most
often at 472-nm). As an example, the reflectance spectra obtained for
6-nitrobenxospiropyran are shown in Figure 7. When the difference spectrum
was taken between the spectrum of the pure compound and that obtained after
sorption onto silicic acid, the bands characteristic of the adsorbed species were
clearly evident.
The interaction between drug compounds and excipients, as these influence
drug dissolution, can be successfully studied by means of reflectance spectroscopy. In one study concerning probucol and indomethacin, it was deduced
that hydrogen bonding and van der Waals forces determined the physisorption
between the active and the excipients in several model formulations (60). Chemisorption forces were found to play only minor roles in these interactions. These
studies indicated that surface catalytic effects could be important during the
selection of formulation excipients.
140
Brittain
Figure 7 Diffuse reflectance spectra of (A) pure 6-nitrobenxospiropyran, and (B) after
its sorption onto silicic acid. Source: Adapted from Ref. 58.
The appearance of tablets and powders during accelerated stability testing

can be quantified using tristimulus colorimetry (61). In this work, various formulations were stored under stress conditions, and the tristimulus parameters of the
tablets were measured at various time points during the storage time. Because the
particular systems studied were characterized by changes in tablet color as
the decomposition reaction proceeded, the X, Y, and Z parameters were judged
useful in deducing the kinetics of the reactions (Fig. 8). Both first-order or
zero-order reactions were noted for the systems studied, and a formal Arrhenius
treatment of the data could be used to predict shelf lives for the formulations.
Instrumental evaluation of the color of solid dosage forms has been incorporated into the stability testing protocol of a number of pharmaceutical products,
including formulations of captopril, flucloxacillin sodium, cefoxitin sodium, and
theophylline (62). The rate of color formulation in the dosage forms obeyed the
Arrhenius equation is each case, and the CIE parameters were used as part of
potency predictions when a relationship between the color change and degree
of drug decomposition has been established.
Taking the application of quantitative color one step further, color
parameters were shown to be a quantifiable variable for the degree of whiteness
of uncoated tablets (63). Color intensity parameters related to yellowish or
brownish colors were defined that were used to describe the discoloration kinetics
with acceptable regression coefficients. The discoloration rates determined under
several storage temperatures were found to follow Arrhenius kinetics, and
141
Figure 8 pH dependence of the Qy chromaticity parameters obtained during the potentiometric titration of Abbreviations: NQS, 1,2-naphthoquinone-2-semicarbazone; NQT,
1,2-naphthoquinone-2-thiosemicarbazone. Source: Adapted from Ref. 61.
therefore the activation energy could be estimated for the products. It was determined that through the use of the color intensity parameters, it was statistically
possible to determine the period of time for which the uncoated tablets remained
white.
USE OF ADSORBED INDICATORS FOR THE EVALUATION
OF SURFACE ACIDITY AND BASICITY
Although the relative degree of acidity or basicity of a substance dissolved in an
aqueous medium can be adequately defined in terms of the solution pH, a comparable expression for the acidity or basicity of the surface of a solid is not as
easily expressed. However, such an evaluation would be of great importance to
a preformulation study, as it is well known that acidic or basic surfaces of
solids are eminently suitable as catalytic agents (64). In the absence of such
knowledge, formulators usually assume certain characteristics of the materials
they are working with and then base their compounding on such assumptions.
The concept of the acid strength of a surface was first discussed in a systematic manner by Walling (65), who sought to differentiate the acid strength
of a surface from the stoichiometry of acidic sites. Walling extended the indicator
theory developed by Hammett to explain the color changes associated with
certain substances as the acidity of their medium was changed (66 68) to
142
Brittain
solids. In particular, Walling defined the acidity of a solid surface as the ability of
the surface to convert an adsorbed neutral base into its conjugate acid. In order to
see how the Walling model can be applied to solids, it is first necessary to work
through some details of ionic equilibria.
Consider the hydrolysis reaction of a weak acid with water:
HA H2 O
! H3 O A
(34)
The position of equilibrium at a given temperature is defined by the magnitude of

the equilibrium constant expression:
aH O aa
K 3
(35)
aHA aH2 O
where the amount of the ith interacting species is specified in terms of its activity,
ai. Owing to the fact that the mole fraction of the solvent in most aqueous solutions is effectively unity, the activity of water (aH2O) in such solutions equals
one. In that case, the ionization constant expression becomes:
aH O aA
Ka 3
(36)
aHA
The activity of the ith interacting species in a solution is related to its molar
concentration, [i], by the relation:
ai i f i
(37)
where fi is a dimensionless quantity known as the activity coefficient. The activity

coefficient (and consequently the activity) of a species varies with the ionic
strength of the medium such that the use of activities instead of molar concentrations in expressions of constant equilibrium frees the numerical value of the
constant from its dependence on ionic strength. Therefore, the ionization constant
expression for a hypothetical weak acid can also be written as:

H3 O A f H3 O f A
Ka
(38)
HA
f HA
The activity coefficients in Equation (38) vary with the overall ionic strength to
keep the value of Ka numerically constant over a range of ionic strengths.
When the ionic strength of the medium is low (i.e., high dilution), all of the
activity coefficients in Equation (38) approach unity, and the ionization constant
expression becomes:
K0a
H3 O A
HA
(39)
It is important to recognize that when an ionization constant is expressed in terms

of molar concentrations the value for Ka0 will be a function of ionic strength. For
instance, Ka0 of acetic acid equals 1.75 1025 at an ionic strength of 0.0,
143
2.85 1025 at an ionic strength of 0.1, and 3.31 1025 at an ionic strength of
0.5.
Using the p scale of Srensen, where the factors in constant equilibrium
expressions are expressed as the negative of their base 10 logarithms, Equation
(39) can be arranged as:
logH3 O pK0a log {A =HA}
(40)
When the weak acid HA is an indicator, the concentration ratio [A2]/[HA] can
be determined using colorimetry. In aqueous solutions, the left-hand side of
Equation (40) is the pH, and Walling (65) defined that quantity as the acid
strength, HO, for solids. The HO function can be equally defined as:
HO aH3 O ( f A = f HA )
(41)
It is important to remember that decreasing values of HO signify increasing acid

strength. If the definition of acid strength is applicable to surfaces, then the
acid strength of a surface should be readily measurable by the observation of
the colors of absorbed indicators. In particular, the appearance of the color
characteristic of the acid form of the indicator would be taken as indicating a
value of the HO function for the surface lower than the pKa of the indicator used.
Walling (65) pointed out a number of requirements essential for the successful use of basic indicators for the determination of the acid strength of
surfaces:
1. The indicators used are restricted to those for which the basic form is
neutral and that are converted to their conjugate acids by the addition
of a single proton.
2. The indicators must be capable of adsorbing onto the surface under
investigation.
3. The solvent used to interact with the indicator with the solid surface
must not perturb the properties of the surface itself.
4. The method requires substrate surfaces whose inherent color does not
interfere with the color determination of the adsorbed indicator.
Walling then continued to define the indicator types that would eventually
be termed the Hammett indicators. Any substance that is neutral or slightly basic
in water, that has a different color in concentrated sulfuric acid, and that is an 1:1
electrolyte can act as a basic indicator at acidity strengths between those of water
and sulfuric acid. Furthermore, the substance must owe its color change to
protonation by a single proton and not by the addition of water or some other
molecule. In the original work, 15 such indicators were indicated as having
appropriate properties (65).
Benesi extended the methodology over a much wider range and used the
color changes of adsorbed indicators to evaluate the acid strength of a variety
of catalyst surfaces (69). Surprisingly, it was found that carefully dried
144
Brittain
samples of many solids yielded the acid response with all tested indicators,
demonstrating that the tested material surfaces (metal oxides, acids supported
on silica gel, clays, and cracking catalysts) were considerably more acidic than
had been anticipated. In fact, the surfaces of some of the clay and catalysts
were found to be so acidic that they could not be completely neutralized with
sodium hydroxide.
Around the same time, the use of Hammett indicators in the quantitation of
acid sites on the surfaces of acidic materials by amine titration became established (70,71). In this method, one first sorbs the indicator onto the surface of
the substance to be titrated, which places the indicator into its acidic form as indicated by its characteristic color. One then titrates the solids with a stronger base
that serves to displace the indicator and monitors the course of the reaction on the
basis of the color change associated with the released indicator (which is then in
its basic form, having a different color).
It follows that if the material surface under investigation is basic in character, then titration with a strong acid would be appropriate to evaluate the acid
strength and stoichiometry of binding sites. For example, a series of solids
have been suspended in benzene, and then titrated with trichloroacetic acid in
the presence of Hammett indicators (72). This approach enabled the determination of the basic strength of a surface in terms of the Hammett acidity function,
facilitating the evaluation of HO across the entire range. It was found that the
basicity of the studied solids increased across the series ZnO . TiO2 .
g-Al2O3 . BaO . B2O3 . ZrO2.
Although the Hammett indicators could be used to provide visual determinations of the acidity or basicity of surfaces, the use of reflectance spectroscopy
facilitated an extension of these studies. For instance, Pichat used UV reflectance
spectroscopy to study the character of the binding sites in the adsorption of three
diazine molecules on silica, alumina, and silica-alumina (73). In the case of silica
or alumina, the spectral bands of adsorbed diazines were comparable with those
obtained for the molecules dissolved in a hydrocarbon solvent, and it was concluded that the adsorption sites involved surface hydroxyl groups similar to
those existing in aqueous or alcoholic solutions. However, significant shifts in
the wavelength of the n ! p transitions of the diazines were observed when
the molecules were sorbed onto silica-alumina, confirming the strong protonic
acidity of this material.
In a series of studies, Leermakers et al. (74 77) used UV spectroscopy to
study the electronic spectra and structure of a number of molecules absorbed onto
silica gel. Because the n,p excited state is less polar than the ground state molecule, then after sorption onto the support, the n ! p transition would be
observed at higher energy and therefore shifted to shorter wavelengths. Conversely, because the p,p excited is more polar than the ground state, then the p !
p transition would be shifted to lower energies and hence longer wavelengths.
This trend was rationalized on the basis that activated silica gel contains highly
polar protonated sites capable of interacting with the oxygen atoms of the
145
carbonyl groups, and these groups would cause a greater perturbation on the
p excited state. This effect is illustrated in Figure 9 for the particular instance
of biacetyl.
Limitations in the indicator methods exist, however, and have been found
to be particularly important in the study of the surface acidity of zeolites (78,79).
In these materials, the acid strength of the surface can be readily established
through the use of Hammett indicators, but the quantitative titration of acid
sites turned out to be strongly affected by the ability of the indicator dye to
enter the zeolite cavities. A completely potassium-exchanged yttrium-zeolite
was not found to exhibit observable acidity, but lanthanum- and gadoliniumexchanged zeolites were found to be strongly acidic even at low degrees of
exchange (80).
Jozwiakowski and Connors (81) used a slightly different approach to study
the surface polarity of silica samples. The samples to be characterized were
suspended in glycerin and then allowed to interact with the 6-nitrobenzoindolinopyran as the adsorption indicator. Because the powders remained in suspension,
conventional UV spectroscopy was used to evaluate the color changes. It was
found that the absorption maximum of the indicator could be related to the
surface polarity of the silica samples by a relatively simple relationship.
Schmidt et al. have carried out a series of determinations of the surface
acidity of pharmaceutical excipients, using a variety of absorption indicators to
deduce the acid strengths. In the first study, the surface acidity of microcrystalline
cellulose and dicalcium phosphate anhydrate was evaluated using thymol blue,
Figure 9 UV absorption spectra of biacetyl dissolved in cyclohexane (solid trace) and

absorbed onto silica gel (dashed trace). Source: Adapted from Ref. 74.
146
Brittain
bromocresol green, bromocresol purple, and phenol red as the absorption indicators (82). As shown in Figure 10, bromocresol green exists primarily in its
acid form when adsorbed onto anhydrous dicalcium phosphate but is distributed
between its acidic and basic forms when adsorbed onto microcrystalline cellulose. Using the pH dependence of dye absorbance published in the paper, one
would conclude that HO would be approximately 1.0 for dicalcium phosphate
anhydrate and approximately 4.0 for microcrystalline cellulose. Granules were
also prepared by the buffering of anhydrous dicalcium phosphate with a
phosphate/citrate/borate system, and the HO of the buffered granules could be
estimated as approximately 7.0.
In a subsequent work, the effect of excipient surface acidity on the
decomposition rate of acetylsalicylic acid was studied (83). The surface of anhydrous dicalcium phosphate was buffered to varying degrees, and then the surface
acidity determined using the appropriate absorption indicators. A methanolic solution of acetylsalicylic acid was used to coat the granules, and then the dried
coated granules were stored under various conditions to induce decomposition
of the drug substance. It was found that the surface acidity decomposition-rate
profiles were similar to the pH-rate profiles of acetylsalicylic acid in solution,
indicating that the decomposition took place in water layers within the coated
granules.
1.0
F(R)
0.8
0.6
0.4
0.2
0.0
250
350
450
550
650
750
Wavelength (nm)
Figure 10 Diffuse reflectance spectra of bromocresol green adsorbed onto microcrystalline cellulose (solid trace) and onto dicalcium phosphate anhydrate (dashed trace).
147
The surface acidities of a wide variety of excipients for solid dosage forms
have been determined using absorption indicators (84). A surprisingly wide
number of fillers, binders, disintegrants, and lubricants were found to exhibit
HO values in the range of 3 4, and only a few excipients exhibited HO values
larger than this. In fact, potassium bicarbonate (HO 8.05), sodium bicarbonate
(HO 8.27), and sodium carbonate (HO 8.58) were found to be the only excipients studied whose surfaces exhibited basic properties.
Referring back to Figure 1, from the wavelength of the absorption maxima
(approximately 525 nm) it can be concluded that the sorbed eosin-Y is present on
the lactose samples in its acid form. Because the pKa of this dye is approximately
3.0, this finding implies that the HO value of the lactose is less than 3.0.
SUMMARY
Although UV/VIS diffuse reflectance spectroscopy has not been used as extensively in the study of pharmaceutical solids as have other spectroscopic techniques, its applications have been sufficiently numerous that the power of the
methodology is evident. The full reflectance spectra, or the derived colorimetry
parameters, can be very useful in the study of solids, which are characterized
by color detectable by the human eye. It is evident that questions pertaining to
the colorants used for identification purposes in tablet formulations can be
fully answered through the use of appropriately designed diffuse reflectance
spectral experiments.
REFERENCES
1. Hardy AC. Handbook of Colorimetry. Cambridge, MA: M.I.T. Press, 1936.
2. Wendlandt WW, Hecht HG. Reflectance Spectroscopy. New York: Interscience Pub.,
1966.
3. Kortum G. Reflectance Spectroscopy: Principles, Methods, Applications. New York:
Springer-Verlag, 1969.
4. Frei RW, MacNeil JD. Diffuse Reflectance Spectroscopy in Environmental Problem
Solving. Cleveland OH: CRC Press, 1973.
5. Frei RW, Frodyma MM, Lieu VT. Diffuse reflectance spectroscopy. In: Svehla G, ed.
Comprehensive Analytical Chemistry. Vol. IV, Chapter 3. Amsterdam: Elsevier,
1975: 263 354.
6. Kubelka P, Munk F. Z Tech Physik 1931; 12:593.
7. Kubelka P. J Opt Sci Am 1948; 38:448.
8. Melamed NT. J Appl Phys 1963; 34:560.
9. Simmons EL. Opt Acta 1971; 18:59.
10. Nutting PG. Trans Illum Eng Soc 1912; 7:412.
11. Taylor AH. J Opt Sci Am 1919; 4:9.
12. Wendlandt WW, Franke PH, Smith JP. Anal Chem 1963; 35:105.
13. Wendlandt WW. Pure Appl Chem 1971; 25:825.
14. Wendlandt WW, Bradley WS. Thermochim Acta 1970; 1:143.
148
Brittain
15. Evans RM. An Introduction to Color. New York: John Wiley & Sons, 1948.
16. Committee on Colorimetry, Optical Society of America, The Science of Color.
New York: Thomas Y. Crowell Co., 1953.
17. Wright WD. The Measurement of Color. New York: Macmillan Co., 1958.
18. Judd DB, Wyszecki G. Color in Business, Science, and Industry. 2nd edn. New York:
John Wiley & Sons, 1963.
19. Kelly KL, Judd DB. The ISCC-NBS Method of Designating Colors and a Dictionary
of Color Names. National Bureau of Standards Circular 553, Washington, D.C.: U.S.
Government Printing Office, 1955.
20. Wyszecki G, Stiles WS. Color Science: Concepts and Methods, Quantitative Data and
Formulae. 2nd edn. New York: John Wiley & Sons, 1982.
21. Hunter RS, Harold RW. The Measurement of Appearance. New York: WileyInterscience, 1987.
22. CIE, International Conference on Illumination, Proceedings of the Eighth Session,
Cambridge, UK, Bureau Centrale de la CIE, Paris, 1931.
23. Billmeyer FW Jr, Saltzman M. Principles of Color Technology. New York: WileyInterscience, 1966.
24. Berns RS. Billmeyer and Saltzmans Principles of Color Technology. 3rd edn.
New York: Wiley-Interscience, 2000.
25. CIE Publication 15.2, Colorimetry. 2nd edn. Central Bureau of CIE, Austria: Vienna,
1986.
26. Raff AM. J Pharm Sci 1963; 52:291.
27. Raff AM. J. Pharm Sci 1964; 53:380.
28. Everhard ME, Dickcius DA, Goodhart FW. J Pharm Sci 1964; 53:173.
29. Goodhart FW, Kelly MA, Lieferman HA. J Pharm Sci 1965; 54:1799.
30. Bogdansky FM. J Pharm Sci 1975; 64:323.
31. Wou LS, Mulley BA. J Pharm Sci 1988; 77:866.
32. Wirth M. J Pharm Sci 1991; 80:1177.
33. Barra J, Ullrich A, Falson-Rieg F, Doelker E. Pharm Dev Tech 2000; 5:87.
34. Urbanyi T, Swartz CJ, Lachman L. J Am Pharm Assoc Sci Ed 1960; 49:163.
35. Lachman L, Swartz CJ, Urbanyi T, Cooper J. J Am Pharm Assoc Sci Ed 1960; 49:165.
36. Lachman L, Weinstein S, Swartz CJ, Urbanyi T, Cooper J. J Pharm Sci 1961; 50:141.
37. Dehner EJ, Shiromani PK. Drug Dev Indust Pharm 1993; 19:1659.
38. Swartz CJ, Lachman L, Urbanyi T, Cooper J. J Pharm Sci 1961; 50:145.
39. Lachman L, Urbanyi T, Weinstein S, Cooper J, Swartz CJ. J Pharm Sci 1962; 51:321.
40. Hajratwala BR. J Pharm Sci 1974; 63:129.
41. Everhard ME, Goodhart FW. J Pharm Sci 1963; 52:281.
42. Goodhart FW, Everhard ME, Dickcius DA. J Pharm Sci 1964; 53:338.
43. Goodhart FW, Leiberman HA, Mody DS, Ninger FC. J Pharm Sci 1967; 56:63.
44. Pope DG, Lach JL. Pharm Acta Helv 1975; 50:165.
45. Lach JL, Bornstein M. J Pharm Sci 1965; 54:1730.
46. Bornstein M, Lach, JL. J Pharm Sci 1966; 55:1033.
47. Lach JL, Bornstein M. J Pharm Sci 1966; 55:1040.
48. Bornstein M, Lach JL, Munden BJ. J Pharm Sci 1968; 57:1653.
49. Carstensen JT, Johnson J, Valentine W, Vance J. J Pharm Sci 1964; 53:1050.
50. Wortz RB. J Pharm Sci 1967; 56:1169.
51. Bornstein M, Walsh JP, Munden BJ, Lach JL. J Pharm Sci 1967; 56:1410.

52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
149
Wu W-H, Chin T-F, Lach JL. J Pharm Sci 1970; 59:1286.

Blaug SM, Huang W-T. J Pharm Sci 1972; 61:1770.
Lach JL, Bighley LD. J Pharm Sci 1970; 59:1261.
McCallister JD, Chin T-F, Lach JL. J Pharm Sci 1970; 59:1286.
Turi P, Brusco D, Maulding HV, Tausendfreund RA, Michaelis AF. J Pharm Sci 1972;
61:1811.
Vemuri S, Taracatac C, Skluzacek R. Drug Dev Indust Pharm 1985; 11:207.
Connors KM, Jozwiakowski MJ. J Pharm Sci 1987; 76:892.
Jozwiakowski MJ, Connors KM. J Pharm Sci 1988; 77:241.
Monkhouse DC, Lach JL. J Pharm Sci 1972; 61:1435.
Bosch E, Casassas E, Izquierdo A, Roses M. Anal Chem 1984; 56:1422.
Stark G, Fawcett JP, Tucker IG, Weacherall IL. Int J Pharm 1996; 143:93.
Berberich J, Dee K-H, Hayauchi Y, Portner C. Int J Pharm 2002; 234:55.
Tanabe K. Solids Acids and Bases. New York: Academic Press, 1970.
Walling C. J Am Chem Soc 1950; 72:1164.
Hammett LP, Deyrup AJ. J Am Chem Soc 1932; 54:2721.
Hammett LP, Deyrup AJ. J Am Chem Soc 1932; 54:4239.
Hammett LP, Paul MA. J Am Chem Soc 1934; 56:827.
Benesi HA. J Am Chem Soc 1956; 78:5490.
Johnson O. J Phys Chem 1955; 59:827.
Benesi HA. J Phys Chem 1957; 61:970.
Yamanaka T, Tanabe K. J Phys Chem 1975; 79:2409.
Pichat P. J Phys Chem 1974; 78:2376.
Leermakers PA, Thomas HT. J Am Chem Soc 1965; 87:1620.
Leermakers PA, Thomas HT, Weis LD, James C. J Am Chem Soc 1966; 88:5075.
Evans TR, Toth AF, Leermakers PA. J Am Chem Soc 1967; 89:5060.
Weis LD, Evans TR, Leermakers PA. J Am Chem Soc 1968; 90:6109.
Kladnig WF. J Phys Chem 1979; 83:765.
Barthomeuf D. J Phys Chem 1979; 83:767.
Kladnig WF. J Phys Chem 1976; 80:262.
Jozwiakowski MJ, Connors KA. Pharm Res 1987; 4:398.
Glombitza BW, Oelkrug D, Schmidt PC. Eur J Pharm Biopharm 1994; 40:289.
Glombitza BW, Schmidt PC. Eur J Pharm Biopharm 1995; 41:114.
Scheef C-A, Oelkrug D, Schmidt PC. Eur J Pharm Biopharm 1998; 46:209.
6
Luminescence Spectroscopy
Harry G. Brittain
INTRODUCTION
Luminescence is a general term for the spontaneous emission of electromagnetic
radiation by a material. This results from the transition of molecules from
their electronically excited states down to their ground states. Different types
of luminescence were historically differentiated on the basis of their respective
timescales, but now their primary classification is derived from changes in
spin angular momenta. Luminescence is most often thought of as being initiated
by electromagnetic radiation having wavelengths in the X-ray, ultraviolet (UV),
or visible (VIS) regions of the spectrum, with such processes being grouped
under the category of photoluminescence. However, luminescence is initiated
by a variety of other sources, such as ionizing radiation (scintillation), electron
beams (cathodoluminescence), heat (thermoluminescence), chemical reaction
(chemiluminescence), electrical current (electroluminescence), or mechanical
impact (triboluminescence). Generally, as long as the resulting emission spectrum
originates from transitions among electronic states of the molecule, it will exhibit
the same wavelength dependence regardless of the mode of excitation.
The subject of molecular luminescence has been discussed at great length,
with special emphasis being placed on the study of photophysical processes in
fluid solution (1 25). Many of these general concepts are of great importance
to any investigation of the luminescence phenomena in solids, such as the excitation spectrum, the emission spectrum, the quantum yield of luminescence, and
the lifetime of emission. When dealing with spectroscopic investigations of
fluorophores in single crystals, the polarization of the luminescence is also
found to yield important information. A comprehensive summary of the older
151
152
Brittain
literature is available for those workers who are involved in the research of the
early history of this field (26).
PHOTOPHYSICAL PROCESSES IN ISOLATED
ORGANIC MOLECULES
In chapter 4, the LCAO-MO theory was used to develop an understanding of the
ground and excited states of a molecule, and the conditions were established
under which a given molecule might undergo an absorption transition from its
ground electronic state to one of its excited electronic states. In that discussion,
the consequences of the phrase what goes up must come down were not considered, but it is clear that if molecules only possessed the capability to absorb
energy, then sooner or later the absorption process would cease when all of the
molecules resided in their excited states. Although such population inversions
can be created under certain conditions (i.e., for laser-associated processes),
this situation is not ordinarily realized. Several mechanisms exist that facilitate
deactivation of the molecule from its excited state and back to its ground state.
In the absence of excitation, the ground state of a molecule can be
described as an electronic wave function onto which is superimposed a sequence
Figure 1 Jablonski-type energy level diagram of the ground and first excited state of a
fluorescent organic molecule. Excitation is drawn as the S0(v 0) ! S1(v 3) transition,
radiationless deactivation of the initial state is the S1(v 3) ! S1(v 0) transition, and
fluorescence is the S1(v 0) ! S0(v 2) transition. The vertical lines correspond to
transitions characterized by the most favorable Franck-Condon overlap.
153
of vibrational wave functions. This has been illustrated in Figure 1 by horizontal

levels proper to each curve (rotational states have not been included for the sake
of clarity in the figure). Due to the magnitude of the energetics involved, a molecule in thermal equilibrium at ambient temperature will exist largely in the
lowest vibrational state of its ground state (i.e., the v 0 level). For the vast
majority of organic molecules, this state will have all electron spins paired and
hence its spin multiplicity will be 2S 1 0, which is identified as a singlet
state. In this discussion, the singlet ground state will be identified as the state S0.
When a molecule is irradiated with light that has an energy matching a separation between states, it is absorbed in a time frame of approximately 10215
seconds. Because the strict spin selection rule is DS 0, excitation into the
excited singlet states (S1, S2, S3, etc.) is only possible, and only then the
orbital angular momentum selection rules are also satisfied. Of course, these
S0 ! Si transitions are the same as those transitions among electronic states discussed in a previous chapter. Because the absorption process is the action that
places molecules in excited electronic states, the wavelength dependence of
absorption is termed the excitation spectrum.
Because the time frame for electronic absorption is very short, the atoms in
the molecule do not have any opportunity to undergo any change in their positions
or momenta within this time period. Consequently, the nuclear configuration and
relative motions in the excited state immediately after absorption are the same as
those of the ground state molecule prior to the light absorption. This understanding
is termed as the Franck-Condon Principle, which states that the most probable
transition will be the one in which there is no change in nuclear position or momentum between the two electronic states. Referring to Figure 1 again, it is seen that
the most intense vibrational component of the transition will be the one that takes
place at the same configuration of the ground state (drawn as the v 3 level of
state S1). In general, the energy of the most probabe Franck-Condon peak will
be the energy of the band maximum in an electronic absorption spectrum. In
other words, for the S0 ! S1 transition of Figure 1, the position of maximal absorbance would correspond to the energy of the S0(v 0) ! S1(v 3) transition.
A molecule raised into an upper vibrational level of the excited state
rapidly loses its excess vibrational energy and decays into the lowest vibrational
level of the excited state (i.e., the v 0 level of state S1) through a process known
as internal conversion. For molecules dissolved in a solution phase, the
vibrational deactivation is most commonly achieved through collisions with
solvent molecules. In the solid state, the internal conversion process will
involve transfer of energy between molecules until the requisite energy is lost
to the system as heat. Many molecules existing in the v 0 level of state S1
return to the ground state through radiationless deactivation, and the excess electronic energy is liberated in the form of heat.
However, other molecules instead spontaneously emit a quantum of
light from the excited state, and the wavelength dependence of the emitted light
is known as the emission spectrum. When the emission takes place at the S1 state
and terminates in the S0 state, the process is known as fluorescence. Fluorescence
154
Brittain
takes place in a time frame of approximately 1029 seconds, and so the FranckCondon principle still holds. Consequently, the most intense fluorescence transition
of Figure 1 would be the one that takes place at the same geometrical configuration
of the excited state, or the S1(v 0) ! S0(v 2) transition.
Because room temperature absorbance takes place practically entirely from the
v 0 level of the ground S0 state and because room temperature fluorescence takes
place almost completely from the v 0 level of the excited S1 state, it follows that
only one vibrational component of the two processes can take place at the same
energy. That is, only the S0(v 0) $ S1(v 0) transition can be observed in both
the excitation and emission spectra. It is, therefore, evident that all excitationtransitions other than the S0(v 0) ! S1(v 0) absorption must take place at
higher energies, whereas all emission transitions other than the S1(v 0) !
S0(v 0) fluorescence must take place at lower energies. This fact is formalized
as Stokes Law, which states that the wavelength of fluorescence will always be
longer (i.e., take place at lower energies) than that of the excited light.
Frequently, an emission spectrum will be a mirror image of the lowest
energy absorption band because the pattern of vibrational energy levels in the
two states is approximately the same. As an example, the excitation and emission
spectra of sodium salicylate (27) dissolved in water are shown in Figure 2. The
emission spectrum exhibits its maximum at 411.5 nm (24,301 cm21), the
1000
EX = 300 nm
EM = 415 nm
Intensity (arbitrary units)
900
800
700
600
500
400
300
200
100
0
250
300
350
400
450
Wavelength (nm)
500
550
Figure 2 Excitation (solid trace) and emission (dashed trace) spectra of sodium salicylate dissolved in water. Abbreviations: EX, excitation; EM, emission. Source: Adapted
from Ref. 27.
155
excitation spectrum exhibits its maximum at 304.3 nm (32,862 cm21), and so

the Stokes shift equals 8561 cm21. The excitation and emission spectra interest at
349.9 nm (28,582 cm21), which becomes the best estimate of the S0(v
0) $ S1(v 0) transition. The mirror image relationship between the spectra
is indicated by the fact that during the excitation process the molecule is
excited into a vibrational energy level that is 4280 cm21 higher than the
S1(v 0) state and that during the emission process the molecule is in an
excited vibrational state that is 4281 cm21 higher than the S0(v 0) state.
In the preceding discussion, only those excited states where the resultant
spin angular momentum equalled zero were considered. However, through a
process known as intersystem crossing, the excited molecule might enter into
another excited state for which the resultant spin angular momentum equals
one (i.e., spins unpaired). Because the multiplicity of this system would be
2(1) 1 or three, the state is a triplet state, for which direct excitation from
the ground singlet state would be forbidden. When it is observed, luminescence
from the excited triplet state down to the ground state (also a forbidden process) is
known as phosphorescence.
Because the lowest-lying triplet state has less energy than the lowest-lying
singlet state, the phosphorescence of a molecule will invariably be observed at
100
90
Absorption
Fluorescence Phosphorescence
80
70
60
50
40
30
20
10
0
300
400
500
600
Wavelength (nm)
700
Figure 3 Room temperature absorption (solid trace) and fluorescence (broken trace)
spectra of sodium fluorescein dissolved in water and the phosphorescence (dashed trace)
spectrum obtained in an aqueous alcohol glass at 77K. Source: Adapted from Ref. 27.
156
Brittain
lower energies than its fluorescence. This behavior has been illustrated in
Figure 3, where the room temperature absorption and fluorescence spectra of
sodium fluorescein dissolved in water are shown. Also shown is the phosphorescence spectrum of the same compound obtained in an aqueous alcohol glass
at 77K (27).
The rate associated with the fluorescence process (Q) must equal the rate of
absorption of excited excitation light (IA, measured in quanta/sec) multiplied by
the quantum efficiency (wF) of the fluorescence:
Q IA wF
(1)
Q (I0 IT )wF
(2)
or
where I0 is the intensity of the incident light and IT is the intensity of light that is
transmitted through the sample. The quantum efficiency, or quantum yield, of the
fluorescence process is defined as:
wF NF =NA
(3)
where NF is the number of photons emitted as fluorescence, and NA is the total

number of photons absorbed.
The absorption of excitation energy by the sample is given by the familiar
Beer-Lambert law:
IT =I0 ekcb
(4)
where c is the concentration of solute (in units of molecules/mL), k is the effective cross-section of one molecule (in units of cm2), and b is the path length of the
sample (in units of cm).
Substitution of Equation (4) into Equation (2) yields:
Q I0 (1 ckcb )wF
The exponential of Equation (5) can be expanded as:

(kcb)2 (kcb)3

Q I0 wF kcb
2
6
(5)
(6)
For dilute solutions, the amount of absorbance is usually small, and hence
the term (kcb) will also be small. Under such circumstances, Equation (6) is
simplified as:
Q I0 wF kcb
(7)
and one concludes that the intensity of fluorescence will be proportional to both
the intensity of the excitation energy and the magnitude of the absorptivity of the
sample.
157
If a system initially consists of an ensemble of excited molecules, then the

probability that any one of the molecules will emit electromagnetic radiation and
return to the ground state is independent of the presence of other excited molecules. In this case, the change in the number of emissive molecules per unit
time (dn/dt) will be given by the simple rate equation:
dn=dt kF n
(8)
where n is the number of emissive molecules present at time t. Integration of

Equation (8) yields:
n n0 exp ( kF t)
(9)
where n0 is the initial number of excited molecules. By the initial definitions:

Q dn=dt
(10)
So Equation (9) becomes:

Q Q0 exp ( kF t)
(11)
It is therefore concluded that the decay of fluorescence will be exponential in

nature. One defines the mean radiative lifetime (tR) of the fluorescence as the
reciprocal of the fluorescence rate constant:
tR 1=kF
(12)
Bowen (28) has provided a simple equation for estimation of the radiative lifetime:
2 2
(13)
1=tR 2900 n n0 a dn
where n0 is the wavenumber of the maximum in the absorption band, n is the
refractive index of the solvent, and a dn is the area under the curve when the
molar absorptivity is plotted against the wavenumber. More complicated
equations for evaluation of radiative lifetimes have been developed (29).
Typical lifetimes of fluorescence are in the order of 1029 seconds, or on a
nanosecond scale. On the other hand, the typical lifetimes of phosphorescence are
in the order of 1023 seconds (or higher), or typically on a millisecond scale.
PHOTOPHYSICAL PROCESSES ASSOCIATED WITH

INTERACTING MOLECULES
The photophysics of well-behaved systems can usually be described using
the isolated molecule principles outlined in the previous section, but it should
be recognized that even isolated molecules can exhibit deviations from their
ideal nature. For example, if the molecule undergoes some type of change in
its configuration during the photophysical process, this will cause sufficient
158
Brittain
alteration in the dipole moment or polarizability so as to alter the energies of the

respective states. The most typical example of this behavior is encountered in
molecules for which the ground and excited states are differently solvated. The
situation is most important for polar solutes dissolved in polar solvents, and in
fact measurements of the 0 0 band shift in differing solvents have been used
to deduce the dipole moment of the molecule in its excited state.
However, the deviations from isolated molecule ideality become highly
important for organic molecules in the solid state. One key difference is that
once absorbed, excitation energy might undergo transfer between molecules if
such interaction is facilitated by the details of the solid-state environment.
Another possibility is that cooperative phenomena existing in the solid state
can create excited-state dimeric (or greater) species that will take up luminescence characteristics of their own.
A basis for understanding solid-state phenomena can be built on the model
of the isolated molecule to include different types of interactive processes that
might take place among the molecules subsequently after their excitation.
Intermolecular Energy Transfer
One of the processes competing with fluorescence is a nonradiative deactivation
of the excited state through transfer of the absorbed molecule to some other
species. The primary excitation step of a hypothetical molecule is described in
the usual kinetic scheme as:
D hn ! D
(14)
where D is the molecule in its ground state, and D represents that molecule in its
excited state. The fluorescence process is then expressed as:
D ! D hn0
(15)
and the rate constant for this process is denoted at ke. Alternatively, the excited
molecule can return to its ground state by a radiationless process:
D ! D D
(16)
where the process is characterized by the rate constant kn, and D is the amount of
heat liberated by the nonradiative deactivation of the excited state. Should the
molecule undergo internal conversion to its triplet state (T ):
D ! T D0
(17)
0
The corresponding rate constant would be kg, and D would be the amount of heat
liberated by the internal conversion process.
However, if during the lifetime of its excited state, D encounters another
species capable of accepting energy from one of the D excited states, then the
following reaction would occur:
D Q ! D Q
(18)
159
This process is known as quenching, and characterized by the rate constant kq.
The rate of the energy transfer obviously will also have a dependence on the concentration of the quenching species, Q. Once its excited state becomes populated,
Q will return to the ground state either by the radiative mechanism characterized
by the rate constant ke0
Q ! Q hn
(19)
or by the non-radiative mechanism characterized by the rate constant kn0 , and

release of an amount of heat equal to D0 :
Q ! Q D
(20)
Consider now a luminescent material undergoing irradiation with excitation energy for an illumination period, which is longer in comparison with its
emission lifetime. Under these conditions, a steady state is set up in which the
rate of production of excited molecules is exactly balanced by the rate of disappearance of these species. In the absence of any photochemical reaction or intersystem crossing from any of the upper excited states, one may express the rate of
production of excited molecules as being equal to the rate of light absorption, Ia:
Ia (ke kn kg ) D
(21)
where [D ] represents the concentration of molecules in the excited luminescent

state. When the energy transfer process is operative, an additional source of
excited state deactivation becomes operative, and Equation (21) becomes:
Ia (ke kn kg kq Q) D
(22)
where [Q] is the concentration of the quencher.

Earlier in Equation (3), the quantum efficiency was defined as the number
of photons emitted in fluorescence divided by the total number of photons
absorbed. This quantity is clearly related to the steady-state rate constants, and
in the absence of any quenching process the luminescence quantum efficiency
(w0F) is given as:
w0F ke =(ke kn kg )
(23)
When luminescence quenching is also present, Equation (23) becomes:
wF ke =(ke kn kg kq Q)
(24)
Taking the dividend of Equations (23) and (24) yields:
w0F =wF 1
kq Q
(ke kn kg )
(25)
If we define a new constant as:

kSV w kq =(ke kn kg )
(26)
160
Brittain
then Equation (25) becomes:
wF 0 =wF 1 kSV w Q
(27)
Equation (27) is of the form developed for luminescence quenching by Stern and
Volmer (30), and kwSV is termed the Stern-Volmer constant for luminescence
quenching.
In a similar manner, one may develop a similar Stern-Volmer expression to
describe the shortening of luminescence lifetime that would be caused by the
quenching agent:
t0 =t 1 kSV t Q
(28)
In Equation (28), t0 and t are the emission lifetimes in the absence and
presence, respectively, when the quencher concentration equals [Q].
It is important to note that Equations (27) and (28) are valid only as long as
the donor and quencher species encounter each other via collisional processes.
This mechanism is termed dynamic quenching and would be anticipated when
the donor and quencher are not physically bound. The diffusion of monomeric
donors and acceptors, which encounter each other solely through collisional processes, is seen to fulfill these requirements. Under these conditions, it must be
that:
kSV w kSV t
(29)
A comparison of energy transfer as studied simultaneously through intensity and

lifetime quenching can, therefore, serve to identify the conditions under which no
association of donor and quencher takes place. If the conditions of Equation (29)
are found to be met, then the presence of pure dynamic quenching may be taken
as indicating the lack of physical association between the donor and acceptor
compounds. This situation has been demonstrated as being characteristic of luminescence processes that take place in the vapor phase (31).
A very different situation would result when the donor and acceptor species
are bound in close proximity as part of an association-type structure. In this case,
the energy transfer process would not require the encounter of Q and D during the
excited state lifetime of the donor, as all geometrical considerations would be fulfilled as a result of formation of the dimer species. One would anticipate that if
the energy transfer was sufficiently efficient, then the DQ pairs would be nonluminescent as far as the donor is concerned. The only luminescence that
could be observed in the system would therefore have to originate either from
Q in either the DQ complex or from the intrinsic emission from uncomplexed
D or Q. When the quenching mechanism involves complex formation, the
process is termed static quenching. Static quenching would also result in the
observation of reduced emission intensities and in the observation of a linear
Stern-Volmer dependence. However, the significance of the quenching constant
161
would necessarily be very different when compared with the constant obtained
from pure collisional quenching.
Pure static quenching cannot affect the observed luminescence lifetime, as
this quantity can only be determined for the species responsible for the emission.
As long as Q and DQ are not luminescent, any experimentally measured lifetime
would only correspond to free D. Therefore, in the absence of collisional quenching, it would follow that t0 t, and therefore from Equation (28) it would be
necessary that ktSV[Q] 0. Thus, the presence of pure-static quenching in a
system can be demonstrated by the collection of lifetime data during the
course of intensity experiments. If the luminescence intensities decrease with
increasing concentrations of the quencher and yet no change in the lifetime of
luminescence is noted, then this indicates the existence of pure static quenching.
However, in most situations the energy transfer process would entail a
combination of simultaneous dynamic and static quenching. Whatever quantity
of donor/acceptor pairs that could be formed would do so prior to the absorption
of excitation energy by the donor:
D Q $ DQ
(30)
The association constant, KC, corresponding to formation of the DQ pair would

be given by the usual equilibrium constant expression:
KC
DQ
DQ
(31)
where [D], [Q], and [DQ] are the respective activities of the donor, the quencher,
and the dimer complex.
The intensity of light that would be emitted from a solution in which both
static and quenching processes were operative would be given by:
I AFD FLUM
(32)
where A is a proportionality constant, FD is the fraction of excitation energy that

actually excites the free D, and FLUM is the fraction of free D, which is not dynamically deactivated by the quencher Q. The assumptions behind Equation (32)
are that D is dynamically quenched and any DQ pair that might be formed is
nonluminescent but capable of absorbing excitation energy. In addition, free Q
is assumed not to absorb any of the incident radiation, and it is also assumed
that the solution is optically dense so that all excitation energy will be absorbed
by either D or DQ.
FD is defined from the Beers Law contributions of the D and DQ species:
FD
aD D
aD D aDQ DQ
(33)
where aD and aDQ are the molar absorptivities of the donor and dimer complex,
respectively. Unless the association of donor and acceptor leads to the formation
162
Brittain
of a charge-transfer complex, it usually follows that:

aD aDQ
(34)
and, therefore, Equation (33) is simplified to a relation based on concentration

only:
FD
D
D DQ
(35)
The fraction of free D that is not dynamically quenched by Q is simply the

reciprocal of Equation (28) and is therefore given by:
FLUM
1
1 kSV t Q
Substitution of Equations (34) and (35) into Equation (32) yields:

D
1
IA
D DQ 1 kSV t Q
(34)
(35)
From Equation (31), it is known that:

DQ KCDQ
Therefore, Equation (35) becomes:

1
1
IA
1 kC Q 1 kSV t Q
(36)
(37)
The proportionality constant, A, must equal the incident light intensity (I0),
because when [Q] 0, it follows that:
A I0
(38)
A general Stern-Volmer equation covering simultaneous dynamic and

static quenching may be obtained by dividing Equation (38) by Equation (37).
One then finds:
I0 =I 1 (kSV t kC )Q (kSV t kC )Q2
(39)
Several publications have appeared that treat nonlinear Stern-Volmer

quenching kinetics and the subject of complexation (32 35). In one interesting
study, the critical micelle concentration of a fluorescent detergent was measured
through determinations of Stern-Volmer quenching parameters as a function of
detergent concentration (36). Besides enabling the determination of the quenching mechanism, one of the practical and useful consequences of simultaneous
measurements of emission intensity and lifetime quenching is in the calculation
of association constants.
163
When the luminescence quenching takes place entirely through the static
mechanism, it follows that ktSV 0, and, therefore, Equation (39) reduces to:
I0 =I 1 kC Q
(40)
The form of the Stern-Volmer relation for pure static quenching is seen to be the
same as that obtained for pure dynamic quenching, except that now the SternVolmer quenching constant is identified as the formation constant of the
donor/quencher complex. It can, therefore, be noted that the observation of a
linear Stern-Volmer relationship does not indicate the existence of either pure
static or dynamic quenching, and that identification of mechanism can only be
deduced using simultaneous studies of luminescence intensity and lifetime
quenching. Of course, the observation of a non-linear Stern-Volmer relation
would necessarily imply the existence of both types of quenching in the system.
It is worth noting that the bimolecular rate constant corresponding to the
quenching process may be calculated by:
kq kSV =t0
(41)
In many studies of energy transfer among organic or donor/quencher

systems, when the quenching mechanism is fully dynamic (or nearly so), the
reactions are observed to proceed in the diffusion-controlled limit.
Photophysics of Associated Molecules
It will be shown in the subsequent section that the excitation and emission spectra
of organic molecules in the solid state is often dominated by the cooperative
effects associated with energy transfer. It is, therefore, informative to briefly consider the transient excited state of dimeric species that can exist in the solution
phase as a means to introduce the phenomena.
The primary photoexcitation step of a molecule was described earlier by
the following equation:
D hn ! D
(14)
and the excited D molecule was described as being capable of undergoing a

number of subsequent reactions involving energy and its transfer. One possible
reaction of the excited D molecule would be to form an excited-state complex
species with an unexcited molecule:
D D ! {DD}
(42)
The fDDg species is known as an excimer, and an extensive review of excimer

photophysics is available (37). In general, quenching by formation of excimers is
favored in nonpolar solvents, and quenching via electron transfer is the mechanism observed in polar solvent systems.
The polynuclear aromatic hydrocarbon pyrene represents one of the best
known examples of excimer fluorescence and was first studied by Forster and
164
Brittain
Kasper (38,39). These workers found that at low concentrations of pyrene

in benzene, only the structured fluorescence characteristic of monomeric molecules was observed. As the concentration of pyrene in the solution was
increased, the monomer fluorescence decreased in intensity and became replaced
by a featureless emission observed at much longer wavelengths. In their work,
Forster and Kasper proved that no ground state association existed between
pyrene molecules.
To illustrate the photophysical characteristics of excimers, Figure 4 shows
the emission spectrum that was obtained for pyrene dissolved in cyclohexane at
concentrations of 0.01 mg/mL (27). The structured emission is characteristic of
the well-known fluorescence of the monomeric form of the solute. Also shown in
Figure 4 is the emission spectrum of pyrene but dissolved at a concentration of
0.1 mg/mL. In this spectrum, the broad, red-shifted, fluorescence spectrum is
characteristic of the excimer species.
From a detailed kinetic analysis of the photophysics associated with the monomeric and excimer species of pyrene, several deductions were made that were
found to be experimentally verifiable (40). From a consideration of the halfvalue concentration (the concentration at which the fluorescence of the monomer
has been quenched by 50%, and the concentration at which the fluorescence of
the excimer has risen to 50% of its maximal value), it was deduced that the
association reaction to form the fDDg species would be diffusion-controlled. In
100
75
50
25
0
350
400
450
500
550
Wavelength (nm)
600
650
Figure 4 Fluorescence spectra of pyrene dissolved in cyclohexane. Spectra are shown in

arbitrary units for solute dissolved at a concentration of 0.01 mg/mL (solid trace) and
0.1 mg/mL (dashed trace). Source: Adapted from Ref. 27.
165
such a case, the reciprocal of the associated rate constant would be proportional to
the viscosity (41,42). This relationship was shown to exist in a number of different
solvent systems, which confirms the mechanism.
Doller and Forster studied the temperature dependence of the emission efficiencies for pyrene in hydrocarbon solutions and were able to demonstrate that
the excimer formation was a reversible reaction that could be described by thermodynamics (42). It was found that in dilute solution, the fluorescence intensity
decreased with increasing temperature, demonstrating the effect of increasing
amounts of radiationless deactivation. For mixtures of monomer and excimer,
a complicated relationship between temperature and the fluorescence efficiencies
of both species was observed, and the fitting of these results enabled a calculation
of the enthalpy of formation for the excimer species.
Since the original work, a large number of molecules have been shown to
be capable of forming excimer species at suitable concentrations in appropriate
solvent systems. One of the most studied systems are molecules containing
two identical fluorophores separated by an alkane chain. Excimer emission has
been detected in systems, such as the diphenylalkanes (43), dinaphthylalkanes
(44), and dipyridylalkanes (45), as long as the separating carbon chain consisted
of three methylene units.
When the pyrene molecule contains a dissymmetric center, the excimer can
exhibit enormous optical activity relative to the monomer molecule. The excitedstate optical activity (as measured by the circularly polarized luminescence) of
the two enantiomers of 1-(1-hydroxyhexyl)pyrene was found to be negligible
at low concentrations but became very intense as the solute concentration was
increased to the point where significant amounts of the excimer were formed
(46). These findings were interpreted as demonstrating a strong stereoselective
preference for the stacking of the pyrene aromatic rings in the excimer. The intensity of the circularly polarized luminescence of the enantiomeric excimers was
found to be equal in magnitude, but opposite in sign, at equivalent concentrations,
suggesting that the sense of chiral stacking was opposite for the excimer species.
Studies of excimer fluorescence have become important in the study of
certain polymeric systems (47). For example, intramolecular excimers have
been studied in aqueous solutions of poly(styrenesulfonic acid) and its salts
(48,49). For the Li(I), Na(I), Cs(I), and Mg(II) salts, the excimer-to-monomer
ratios were found to be independent of the salt concentrations, indicating that
condensation with these counter ions did not result in a conformational change
in the polyelectrolyte. On the other hand, for the K(I), Ca(II), Sr(II), and
La(III) salts, the excimer-to-monomer ratios increased as the salt concentration
was increased, indicating the existence of a conformational change that facilitated adjacent phenyl groups to become more interactive.
Another possible reaction of the excited D molecule would be to form an
excited-state complex species, but with an unexcited molecule that has a different
chemical identity:
D Q ! {DQ}
(43)
166
Brittain
When D and Q are different molecules, the term exciplex is used to describe the
luminescent fDQg species. Of course, if the fDQg species is not luminescent, it
follows that the previous discussion pertaining to luminescence quenching
applies. Studies of exciplexes have been thoroughly reviewed (50).
To illustrate the spectral characteristics of exciplex formation, the experiments of Knibbe et al. (51) have been reproduced. Figure 5 shows the fluorescence
spectra of anthracene dissolved at a concentration of 1.8 mg/mL in toluene and
the spectra of the same solution when increasing amounts of N, N-diethylaniline
were added (27). It is obvious from the figure that the addition of diethylaniline
quenches the fluorescence of anthracene and that the fluorescence of a new entity
increases in intensity with the amount of diethylaniline added. It was shown by
Knibbe et al. (51) that the broad fluorescence originated from the 1:1 exciplex
and that the degree of formation of ground-state complex species was very small.
The ability of amines to form exciplexes with a multitude of fluorescent
organic molecules has become an extremely well-studied phenomenon, and the
family of stilbene-like molecules represents one such research interest (52).
For example, trans-3-aminostilbene and its N-methyl derivatives are strongly fluorescent when dissolved in cyclohexane, and form exciplexes when alkylamines
are added (53). Modeling of the concentration dependence of the quenching of
100
80
60
40
B
20
0
350
400
450
500
550
600
Wavelength (nm)
Figure 5 Fluorescence spectra of anthracene dissolved at a concentration of 1.8 mg/mL

in toluene (solid trace). Spectra are also shown for this solution but also containing N,Ndiethylaniline at concentrations of 0.75 mg/mL (A), 3.75 mg/mL (B), and 15.0 mg/mL
(C). Source: Adapted from Ref. 27.
167
the monomer fluorescence, and the growth of exciplex fluorescence, with primary
amine concentration led to the conclusion that excited-state complexes having
1:1 and 1:2 stilbene-amine stoichiometries could be formed. Secondary amines
tended to yield only the 1:1 exciplexes. The formation of the exciplex was attributed to a Lewis acid/base interaction between the lone pair acceptor of the stilbene moiety with the lone pair donor of the alkylamine.
Stereoselectivity has been detected in the fluorescence quenching of 1,10 binaphthyl by chiral amines (54). It was reported that specific geometrical
requirements were associated with the fluorescence quenching of 1,10 -binaphthyl
by the resolved enantiomers of N, N-dimethyl-a-phenethylamine, although the
overall details of the quenching were similar to those of other quenching
studies of aromatic molecules by dialkylamines. In this study, the degree of
stereoselectivity was found to be strongly dependent on the solvent used, with
solvents of low dielectric constant yielding the highest enantiomeric discrimination values. One part of the solvent dependence was attributed to alterations
in the quenching mechanism and another part to stabilization of certain conformers of the 1,10 -binaphthyl donor.
Photophysics in the Solid State
It is evident from the discussion of excimers and exciplexes that the photophysics
of organic compounds can become complex when the molecules are brought into
close proximity and allowed to interact with each other. In fact, many compounds
that are highly fluorescent in solution become absolutely nonfluorescent in the
solid state. For example, the fluorescence quantum yield of the sodium salt of fluorescein is close to unity, but no fluorescence can be detected for solid uranine. In
other instances, a nonfluorescent material can become luminescent if the defect
sites in the crystal can serve as energy traps. A number of texts have appeared
that treat the solid-state photoluminescence of inorganic and organic systems,
and the coverage in these is of considerable historical interest (55 60).
The repetitive three-dimensional structure of the crystalline state of a luminescent molecule locks the component molecules in well-defined orientations
with respect to each other that are defined by the details of the unit cell and its
translational symmetry. If the degree of interaction between molecules is
weak, excitation into low-lying electronic states results in the population of
excited states that are similar to those for an individual molecule in the vapor
phase. However, this excited state cannot be a stationary state of the crystal,
and, over time, the probability of finding the excitation on the initial molecule
decreases, and the probability on the adjoining molecules increases.
As a result, energy transfer between molecules, via the common radiation
field, is often facilitated, and such radiationless transfer of energy can take place
over appreciable distances in the crystal. Although the ground electronic state is
localized on individualized molecules even in the crystalline state, the excited
electronic states of these can be so strongly interactive that excitation energy
168
Brittain
becomes delocalized among the coupled molecules. The excitation energy, therefore, spreads outward through the crystal like a wave, or alternatively, as the
motion of a quasiparticle through the crystal. Frenkel first termed such a
delocalized excited state as an exciton (61), and the theory of excitons has
been thoroughly applied to organic crystals by Davydov (62). Additional discussions of exciton theory can be found in the texts by McClure (63) and Knox (64),
and only a brief summary will be given here.
Craig and Walmsley (65) have summarized a number of characteristics pertaining to excitons. Because the transfer of electronic energy between the excited
molecules is rapid relative to the time frame of fluorescence, the delocalization of
excitation energy has the effect of extending the excited state molecular orbital
over the ensemble of molecules involved in the energy transfer. For a pair of molecules, the interaction leads to a splitting of the single-molecule energy level into
a pair of levels, where the magnitude of the splitting is determined by the strength
of the coupling. This type of splitting is commonly referred to as Davydov
splitting and becomes manifest in the appearance of new bands in the excitation
spectrum. In addition, the theory predicts that the mean frequency of the Davydov
components would be displaced to lower energies from that of the free molecule
value as a result of cooperative interactions in the crystalline state.
For an ensemble of N identical, decoupled, molecules, the wave functions
(fi) of the excited states where one molecule is excited and the others are in the
ground state can be written as:
f1 c (1) c(2) c(3) c(n)

f2 c(1) c (2) c(3) c(n)
f3 c(1) c(2) c (3) c(n)

fN c(1) c(2) c(3) c (n)
(44)
where c(i) is the wave function of the molecule it its ground state and c ( j ) is
that of a molecule in the excited state. If there is no interaction between the molecules, then all of the excited state wave functions will have the same energy. But,
if transfer of excitation energy between the excited states takes place, then those
states are not stationary states of the system and must be mixed. The mixing
process yields a new series of states (Cn) that will be formed as linear combinations of the fi functions and that will differ from each other by the values of
the coefficients, Cnm:
C1 C11 f1 C12 f2 C13 f3 C1m fN
C2 C21 f1 C22 f2 C23 f3 C2m fN
C3 C31 f1 C32 f2 C33 f3 C3m fN

CN Cn1 f1 Cn2 f2 Cn3 f3 Cnm fN
(45)
169
Figure 6 Schematic energy level diagram illustrating the effect of exciton coupling that
results in a splitting of the excited state into a manifold of states.
As illustrated in Figure 6, the result of exciton coupling is to produce

an excitation multiplet corresponding to a band of n levels, each differing
in energy by small amounts. The overall spread of the excitation band is determined by the strength of the coupling, and one can actually observe a separation
between the symmetric linear combinations and the antisymmetric linear combinations if the coupling is sufficiently strong.
In molecular crystals, the magnitude of the interaction energy and the
degree of exciton coupling would necessarily be dependent on the relative orientation of the molecules in the crystal and on their distance separation. It follows
that because polymorphic crystal forms are characterized by the existence of
differing structural properties, the nature of the exciton coupling in the various
forms would be dictated by the particular structural characteristics of each
form. As a result, one would anticipate that the magnitude of the Davydov
splitting, and the degree of shifting of the levels, would be dependent on the
exact structural details existing in the different polymorphic forms.
The effect of the crystalline state on the fluorescence of an organic compound can be illustrated using the example of 4-methylsalicylic acid, and
Figure 7 shows the normalized solution-phase and solid-state excitation and
emission spectra obtained for this compound (66). A considerable amount of
energy transfer quenching must exist in the solid state, as these spectra were
weaker on the order of 100 times than the corresponding solution phase
spectra. The emission spectrum obtained for the solution phase exhibited a
170
Brittain
100
Normalized Intensity
80
60
40
20
0
250
300
350
400
450
500
550
Wavelength (nm)
Figure 7 Fluorescence spectra of 4-methylsalicylic acid. Spectra are shown for the solution-phase excitation (A) and emission (C) obtained at a concentration of 15 mg/mL and
for the solid-state excitation (B) and emission (D) spectra. All of the spectra have been
normalized to a common intensity factor. Source: Adapted from Ref. 66.
maximum at a wavelength of 408 nm, whereas the emission maximum of the

solid-state spectrum was only slightly red-shifted to 413 nm, indicating that
the fluorescent state of 4-methylsalicylic acid was equivalent in both phases.
On the other hand, the maximum of the excitation spectra strongly differed,
being observed at 302 nm in the solution phase and at 342 nm in the solid
state. This strong red-shift of the excitation spectra is a direct consequence of
the intermolecular energy transfer that was also associated with the overall loss
in fluorescence intensity in the solid state.
INSTRUMENTATION FOR THE MEASUREMENT

OF LUMINESCENCE SPECTRA
Steady-State Photoluminescence Spectroscopy
At a minimum, a fluorescence system will contain the essential components of
a source of excitation energy, a holder for the sample, a device to analyze the
resulting emission, and a detector for its quantitative measurement. Figure 8
shows a block diagram that illustrates the most commonly employed right-angle
method of measurement. Very detailed descriptions of steady-state luminescence
Figure 8
171
Block diagram of a steady-state fluorescence spectrometer.
spectrometers are available, and those needing more detailed discussions of individual topics should consult these literature sources (7 12,15 19,22,67).
In the most simple fluorescence system, the wavelengths of excited
and emitted light are selected by broad-pass filters that allow measurements
to be made at any pair of fixed groups of wavelengths. Such devices are
termed filter fluorimeters and have the advantage that their use facilitates the
greatest degree of analytical sensitivity. For maximum sensitivity, the sample
can be excited by a broad range of wavelengths that would span its entire absorption band, and the intensity of the entire emission spectrum can be measured
at once.
More sophisticated fluorescence spectrometers are based upon the monochromator discrimination of either the excitation or emission optical trains. In
such systems, one may set the emission analyzer to a fixed wavelength and systematically vary the excitation wavelength to measure the excitation spectrum.
Alternatively, one could set the excitation monochromator at a fixed wavelength
and systematically vary the analyzing emission wavelength to measure the emission or fluorescence spectrum.
Commonly employed sources in fluorescence spectrometry are characterized either as having spectral outputs as a continuum of energy over a wide
range of wavelengths or as a series of discrete lines. An example of the latter
type of source that has received much use in the past as an excitation source is
172
Brittain
the mercury lamp, where the output from a low-pressure lamp would be concentrated in the UV region of the spectrum. Medium- and high-pressure lamps will
have an output that effectively covers the entire range of wavelengths in the UV
and VIS regions.
However, the use of a monochromator to select the proper excitation wavelength really requires the use of a continuum source, and the most useful source
of this type is undoubtedly xenon arc source. When used as light sources for
steady-state spectroscopy, these sources are operated on a continuous basis,
but they can also be operated stroboscopically for time-resolved investigations.
The output is essentially a continuum of electromagnetic energy, onto which is
superimposed a number of sharp lines (8,10). The xenon source effectively
allows the use of any wavelength throughout the UV and VIS regions of the
spectrum.
Arc lamps are inherently more unstable than other sources, and a method to
compensate for the drift is usually employed when long-term stability is required.
Typically, this is accomplished by splitting the excitation beam so that a small
portion is quantitated by a reference detector. This reference signal is used to
ratio the signal from the detector observing the sample, and the arrangement
serves to account for variations in source intensity. Arc lamps can produce
ozone from the photolysis of atmospheric oxygen, which ordinarily requires
their venting. However, ozone-free lamps are available that minimize this
problem, although at the expense of the lowest wavelengths of the source.
A filter fluorimeter system uses fixed filters to isolate both the excited and
emitted wavelengths, and often a pair of cut-off filters is used to isolate one particular wavelength from a source emitting a line spectrum. The filters may be
either commercial glass slides or can be solutions contained in cuvettes. The
emission filter must be chosen so that any Rayleigh or Tyndall scattered light
is not passed, but that the light emitted by the sample is transmitted. Such a
simple filter system can be used for routine quantitative work, but only after
the sufficient characterization of the photophysics of the system to know the
most appropriate combination of filters.
When selectivity in excitation or emission is required, monochromators
containing diffraction gratings are used to select the desired wavelengths.
These types of fluorescence spectrometers are then capable of recording both
excitation and emission spectra and are, therefore, the most versatile systems.
Although solution-phase fluorescence studies do not require high resolution
(and in fact are usually operated under low-resolution conditions to achieve sensitivity), the study of luminescent materials in the solid state can require the use
of high resolution and narrow monochromator slit widths.
In the past, commercial fluorescence systems made use of photomultiplier
tubes as their detectors (68). A wide variety of tube types are available, which
differ mainly in the substance from which the photocathode is made. Photomultiplier tubes having S5 response are responsive to light having wavelengths
ranging from the UV to wavelengths as long as 650 nm. For studies conducted
173
in the near-infrared region (NIR), a photomultiplier tube having S20 response is

ordinarily used. The sensitivity of a photomultiplier tube is limited by the level of
its dark current (the signal from the tube in the absence of light), but this can be
reduced by cooling the tube.
The spectral response of all photomultiplier tubes varies with wavelength,
but when it is necessary to determine the actual quantum intensity of the incident
radiation, or obtain a response-free spectrum, the detector response is corrected.
This is most conveniently achieved through the use of a quantum counter, which
is a substance for which the quantum yield of fluorescence is unity. One such
quantum counter is a concentrated solution of Rhodamine 101 in ethylene
glycol, which will emit the same number of quanta of light as it absorbs but
over a very wide wavelength range. Thus, by measuring the output of the
quantum counter at one wavelength, the number of incident quanta over a
wide wavelength range can be measured.
The versatility of luminescence spectroscopy has been greatly enhanced
through the use of multichannel analyzers, because such devices permit the
acquisition of complete spectroscopic data almost instantaneously over a range
of wavelengths (69,70). Among the most commonly used detectors are diode
arrays, charge-coupled and charge-injection detectors, and vidicons.
Sample configurations for solid-state work usually involve front-face excitation of the sample, with the luminescence being recorded at a grazing angle
with respect to the irradiation beam. For single crystals, films, and supporting
matrices, the sample can be mounted on a suitable holder and placed at the
proper geometry in the optical train. Powdered materials can be placed in a
thin-walled tube (preferably made of vitreous quartz, but glass can serve as
well) prior to its placement in the spectrometer. Whatever the arrangement,
care must be paid to the problem of spurious signals being detected that originated from inadvertent reflection of either the excitation or emission beams.
Time-Resolved Luminescence Spectroscopy
In many instances, the basic techniques of steady-state luminescence spectroscopy suffice to solve a problem of physical, analytical, or other scientific
interest. When these methodologies prove to be insufficient, the next stage in
sophistication would be to use synchronous scanning of excitation and emission
spectroscopies as a means to achieve additional selectivity. Synchronous fluorimetry entails the simultaneous scanning of both the excitation and emission
monochromator wavelengths, but where the wavelength difference between the
two settings is maintained at a constant value. This combination of parameters
results in a simplification of the observed spectra features and a reduction in
the observed peak widths, both factors that lead to enhanced selectivity (71 74).
Synchronous luminescence spectroscopy has proven to be very useful in the
characterization of mixtures of compounds for which overlapping of spectroscopic
bands is a problem, and for which an evaluation of the entire excitationemission
174
Brittain
matrix is not necessary. For example, the combination of synchronous fluorimetry

and derivative data analysis has been used to determine the levels of salicylic
acid and its metabolites in urine samples (75,76). Additional advances have included
the use of variable angle (77) and constant-energy (78) synchronous scanning fluorescence methods. However, all of the methods suffer from the same limitation,
namely the requirement for the optimal value of the scanning wavelength offset
that yields the greatest photophysical advantage to be determined, which can
sometimes be a lengthy process.
In the search for additional analytical selectivity and sensitivity in luminescence spectroscopy, a considerable amount of work has been conducted where
the time scale of the photophysical processes is incorporated into the measurement. In preceding sections of this chapter, a number of spectroscopic transitions
(both radiative and nonradiative) have been discussed, but until now the time
evolution of these has not been discussed. It is evident that every transition
must take place within a finite amount of time, and, therefore, the incorporation
of the time element into luminescence techniques will necessarily impart a
selectivity into the methodology that can be exploited (79,80).
Two methods have become widely used for the determination of timeresolved luminescence properties, with these operating in the time domain and
in the frequency domain. In time-domain luminescence spectroscopy, the
sample is excited using pulsed excitation energy, and the spectroscopy observable is measured at some finite time after the pulse is complete. In frequencydomain spectroscopy, the sample is excited with modulated excitation energy,
and the spectroscopy observable is found in the phase shift of the measured
signal with respect to the initial excitation. Each approach has its strengths and
limitations (81), and these will be briefly reviewed in the following sections.
Photophysical Measurements in the Time Domain
In time domain fluorimetry, the initial excitation of the sample is achieved using a
light pulse of sufficiently short duration so that the width of the pulse is short relative to the decay time of the sample. As discussed earlier, the change in the
number of luminescent molecules per unit time (dn/dt) is given by:
dn=dt kF n
(46)
where n is the number of emissive molecules present at time t. Defining Q as

2dn/dt, and integrating Equation (46) yields the familiar equation for the exponential decay of fluorescence:
Q Q0 exp (kF t)
(47)
The mean radiative lifetime (tR) of a fluorescence decay equals the reciprocal of
the fluorescence rate constant, so therefore:
Q Q0 exp (t=tR )
(48)
175
In the more general instance, an experimentally observed luminescence lifetime,

t, will be determined by both the nonradiative (kNR) and radiative (kF) decay
constants
t 1=(kF kNR )
(49)
Methods for determination of luminescence lifetimes are available in all of the

general references, but the most complete discussion is found in the monograph
by Demas (20).
The fluorescence decay curve of a hypothetical fluorescent substance
exhibiting a 25 nsec emission decay is illustrated in Figure 9, along with the
decay curves of a pulsed flashlamp characterized by a 2 nsec pulse. It can be concluded that if one sought to measure the spectra characteristics of the compound
free from contamination by energy from the source, then one would delay the
recording of data for a time period of at least 15 nsec after the initial pulse.
One approach for the detection of authentic luminescence is to use a gating
method, where the detector is only activated after a definite time period following
the excitation pulse. Alternatively, various investigators have made use of boxcar
integrators, transient signal recorders, multichannel analyzers, and video fluorimeters to monitor the time dependence of fluorescence.
The power of using time resolution to enhance the selectivity of a photophysical measurement is illustrated in Figure 10, where the observed decay
1.00
Intensity
0.75
0.50
0.25
0.00
0
25
50
75
100
125
150
Time (nsec)
Figure 9 Decay curves of a 2 nsec pulsed flashlamp (dashed trace) and of a hypothetical
fluorescent substance exhibiting a 25 nsec emission decay.
176
Brittain
1.00
Intensity
0.75
0.50
0.25
0.00
0
25
50
75
100
Time (nsec)
125
150
Figure 10 Decay curves of an equimolar mixture of hypothetical fluorescent substances,

one exhibiting a 25 nsec emission decay (short-dashed trace) and the other exhibiting a
75 nsec emission decay (long-dashed trace). The total observed decay curve is the solid
trace.
curve of an equimolar mixture of fluorescent substances is shown along with the

individual decay curves of the components. For the two hypothetical substances
exhibiting fluorescence lifetimes of 25 and 75 nsec, it is clear that one could
observe the fluorescence of only the 75 nsec component in the mixture by
employing a 100 nsec gate time. As long as this particular substance was the
analyte of interest, one could obtain the full range of information accessible
with luminescence techniques without the need to first conduct a separation step.
The technique of time-correlated single-photon counting is probably the
most widespread method used to acquire photophysical information in the time
domain (82,83). In this approach, the sample is excited using a pulse source,
and only those photons that are time-correlated to the excitation pulse are
counted to develop the fluorescence signal. By using a repetitive pulsed
source, the detection system builds up a histogram of detector counts in bins of
time channels. Taken together, this histogram has the form of a fluorescence
decay. The observed data are then deconvoluted into contributions from the
system and associated with optical elements to ultimately obtain the decay
curve of the sample itself (84).
Photophysical Measurements in the Frequency Domain
Another approach to the measurement of time-resolved luminescence spectroscopy involves modulated light sources and phase shift measurements and,
177
hence, is typically known as phase-resolved fluorescence spectroscopy. This technique uses a source whose intensity is sinusoidally modulated and yields the lifetime of the sample from the phase shift between the excitation waveform and that
of the sample emission. It has been the subject of numerous reviews (8587).
The time-dependent intensity of the excitation energy, IT, is described by:
IT I0 1 MEX sin(vt)
(50)
where I0 is the DC intensity component of the excitation beam, and v is the

angular modulation frequency related to the linear modulation frequency, f, by
the relation v 2pf. The degree of modulation, MEX, is the ratio of the amplitude of the AC intensity to the DC intensity component.
After excitation by the modulated exciting beam, the resulting fluorescence
will be characterized by its own time dependence, QT and will also be demodulated and phase shifted by an amount related to the fluorescence lifetime of the
emissive species:
QT Q0 1 MEX m sin(vt w)
(51)
In Equation (50), Q0 is the DC component of the fluorescence, w is the phase shift

of the emission relative to the excitation, and m is the demodulation factor
(m cos w). The demodulation factor can also be evaluated as the ratio of the
degree of AC modulation of the emissive species to the degree of AC modulation
of a pure scattering reference.
Modulation of the source can be effected using either a Pockels cell or a
photoelastic modulator, and the magnitude of the phase shift is measured using
phase-sensitive detection (typically with a lock-in amplifier). Additional precision in the measurements can be achieved through the use of cross-correlation
techniques.
From knowledge of the phase shift in the fluorescence, the luminescence
lifetime of the fluorescent species can be calculated using:
t (1=v) tan w
(52)
The lifetime can also be determined using the demodulation factor:

1 1

1
v m2
(53)
The considerable advances in instrumental sophistication, and methods for

the suppression of interfering signals, have made the use of phase-resolution
luminescence spectroscopy as the most versatile of the time-resolved methods.
Advances in data analysis and handling have facilitated the study of complex
mixtures, and the technique has become the investigative tool of choice for the
study of complicated biomolecular systems.
178
Brittain
PHOTOLUMINESCENCE AND THE STRUCTURE

OF ORGANIC MOLECULES
As stated earlier, not all organic molecules are capable of exhibiting fluorescence
in either the solid state or in the solution phase, and the subject of molecular structure and luminescence efficiency has been discussed by a number of authors in the
general texts cited earlier. Particularly useful compilations of fluorescence
spectra are those of Berlman (6), Schmillen and Legler (88), and Maeda (89).
Other general discussions include fluorescence and phosphorescence of pharmaceutical compounds (90) and the fluorescence of organic natural products (91).
When considering the molecules, or functional groupings within molecules
that lead to the observation of measurable fluorescence, one must keep in mind
the photophysical reactions described earlier in Equations (14) through (20).
Molecules in their lowest-energy excited state (typically the lowest-singlet
state of an organic molecule) can be deactivated through fluorescence, but that
state can also become depopulated through radiationless deactivation (internal
conversion or intersystem crossing) or by the conduct of a photochemical reaction. The relative rates of the reaction for the different processes determine the
outcome of a photoexcitation process, and the existence of fluorescence requires
that the rate constant for radiative transitions be large relative to those for nonradiative decay or photodecomposition.
Guibault has summarized the characteristics of molecules that should
exhibit strong fluorescence (92), and these are summarized as follows.
.
The spin-allowed electronic absorption transition of lowest energy is

very intense. Because the intensity of absorption is directly proportional
to the rate constant for the radiative transition and because fluorescence
is the reverse of absorption, it follows that the more probable the absorption transition, the more probable will be the fluorescence transition.
The energy of the lowest spin-allowed absorption transition should be
reasonably low. High energy excitation of a molecule is more likely to
result in its photodissociation rather than in the initiation of the fluorescence process.
The terminal state of the absorption transition (i.e., the excitation
process) should not be strongly involved in chemical bonding. If the
state is strongly bonded, its excitation could result in bond dissociation
rather than fluorescence.
The molecule should not contain structural features or functional
groups that enhance the rates of radiationless transitions. Certain structural features greatly increase the rates of radiationless processes and
therefore adversely affect fluorescence efficiencies.
Guibault goes on to use these simple considerations to understand why aromatic hydrocarbons are usually very intensely fluorescent. In molecules of this
type, it is well known that the states derived from orbitals in the p-bonding
179
system are higher in energy than are the states derived from orbitals in the
s-bonding system. Thus, aromatic hydrocarbons can be excited at relatively
low energies to undergo the strongly allowed p ! p transition without significant alteration in the chemical bonding patterns of the molecule. Excitation into
the high energy s ! p transitions does not usually result in observable fluorescence, owing to the large perturbations in chemical bonding that accompany
the excitation process.
The situation becomes more complicated in molecules containing functional groups that contain nonbonding electrons, such as the carbonyl groups
present in aldehydes, ketones, and carboxylic acids. In these instances, the
energy of the n ! p transition will usually be lower relative to the energy of
the p ! p transition. However, the intensity of the n ! p transition is invariably less than the intensity of the p ! p transition, and therefore one would
anticipate that aromatic carbonyl compounds exhibit lower intensities of fluorescence than the parent aromatic hydrocarbons.
Interested readers seeking more information regarding the effects of
molecular structure on the fluorescence of aromatic molecules should consult
the monograph by Birks (14).
STUDIES OF LUMINESCENCE IN THE CRYSTALLINE STATE
The first investigations into the luminescence phenomena of pure organic molecules in their crystalline state dealt with aromatic hydrocarbons, such as
naphthalene and anthracene. Many of the earliest studies were flawed by the
presence of impurities, which acted as energy sinks, leading to photoluminescence from the impurity but not from the host crystal. For example, Shpak
and Sheka showed that many previously recorded fluorescence spectra of
naphthalene in its crystalline state were actually the fluorescence spectra of the
b-methyl-naphthalene impurity (93).
The crystalline naphthalene system has been used in a number of chemical
physics studies to examine the flow of excitons in both pure and mixed crystals. For
example, the energy transfer phenomena have been formulated in terms of a static
percolation theory, where the dynamic motion of excitons has been described as
being a function of the excitation lifetime, jump time, coherence time, and trapping
time (94). In another model, a random-walk formulism was used to investigate the
anisotropic interactions associated with the energy transfer of excitons and to
better understand the first singlet excited state of naphthalene (95). The flow of
excitation energy, and the concept of exciton energy funnels, has been studied
in different mixed naphthalene crystals, where the luminescence of the impurity
species has been used as a means to probe the photophysics in the crystal (96,97).
Crystalline anthracene represents another system where the photophysical
phenomena in the crystalline state are very well understood, and the absorption
(98,99) and fluorescence (100,101) spectra in the crystalline state have been
well detailed. A considerable amount of work has been expended in an effort
180
Brittain
to understand the 25,000 cm21 (approximately 400 nm) band system and the
nearly 500 cm21 gap between the absorption and fluorescence maxima. From
low temperature studies conducted on zone-refined material, the origin of the fluorescence spectrum (i.e., the v 0 to v 0 transition) has been identified as
having an energy of 25,103 cm21, which correlates well with the b-polarized
absorption transition at 25,150 cm21. In one study, the energy of the Davydov
split state was found to be 24,705 cm21, indicating that the magnitude of the
Davydov splitting was approximately 400 cm21.
Photophysical phenomena have been studied in a number of other molecular
crystals, largely taking the chemical physics approach of understanding the structure of energy levels, monitoring the flow of excitation energy, and evaluating the
resulting photoluminescence. For example, investigations of the photophysics in
crystalline biphenyl (102,103), trans-stilbene (104), dimethylbenzophenone
(105,106), and dichlorobenzophenone (107) have been reported. Fluorescence
studies have been used to study the phase transition, which takes place at a temperature of 191 K (2828C) in crystals of p-terphenyl (108).
It is probably true, however, that pharmaceutical scientists are more interested in spectroscopic studies that provide insights into the structures of substances, rather than in studies that are conducted to probe photophysical
pathways. Such work would entail investigations of the relationship between fluorescence and molecular geometry, where X-ray structural analysis results would
be correlated with solid-state luminescence properties. For example, the redshifted, structureless emission associated with crystalline spiroanthronyl-substituted dihydrobenzanthracenes has been attributed to the presence of excimer
luminescence in the solid state (109). In other works, this same group studied
the structure-mediated intramolecular energy transfer in a series of anthronylanthracenes (110), and the effect of molecular stereochemistry on the excimer
phenomena of crystalline dihydroanthracenes (111).
The utility of solid-state photoluminescence as a means to study compounds of pharmaceutical interest can be illustrated through a recent study conducted on erythromycin-B derivatives (27). Figure 11 shows the excitation and
emission spectra of erythromycin-B dihydrate, where it may be observed that
the excitation spectrum consists of two bands having maxima at approximately
350 and 390 nm. The maxima of the fluorescence resulting in the excitation of
these bands are nearly equivalent, exhibiting maxima at 482 and 488 nm, respectively. Dehydration of the substance causes the fluorescence intensity to be
reduced by a factor of two, and a substantial red-shift in the maxima of the
fluorescence bands. Excitation of the dehydrated material at 350 nm now
yields fluorescence having a maximum at 459 nm, while excitation at 390 nm
yields fluorescence having a maximum at 472 nm.
Quite different behavior was observed for the hydrochloride salt of erythromycin-B, as shown in Figure 12. Once again, the excitation spectrum consists of two
bands having maxima at approximately 350 and 390 nm, and the maxima of the fluorescence obtained on excitation of these bands are found to exhibit maxima at 455
181
Erythromycin-B dihydrate
60
Intensity
45
30
15
0
300
350
400
450
500
550
600
650
550
600
650
Wavelength (nm)
Desolvated Erythromycin-B dihydrate
Intensity
30
20
10
0
300
350
400
450
500
Wavelength (nm)
Figure 11 Solid-state fluorescence spectra of erythromycin-B dihydrate (upper spectra)

and on its dehydrated product (lower spectra). The excitation spectra (open circle) were
obtained using an analyzing emission at wavelength of 465 nm, whereas the emission
spectra were obtained upon excitation at 350 nm (solid square) and at 390 nm (solid
circle). Although the intensity scales are arbitrary, the spectra are scaled to permit comparisons to be made. Source: Adapted from Ref. 27.
and 470 nm. The intensity of fluorescence obtained from the hydrated hydrochloride
salt was found to be more than twice that of the hydrated free base. Dehydration of
the hydrated hydrochloride salt was found to have little effect on the emission
maxima (455 nm for excitation at 350 nm, and 472 nm for excitation at 390 nm),
but the intensity of emission decreased by more than a factor of 15.
182
Brittain
Hydrated Erythromycin-B hydrochloride

140
120
Intensity
100
80
60
40
20
0
300
350
400
450
500
550
600
650
Wavelength (nm)
Dehydrated Erythromycin-B hydrochloride
Intensity
8
6
4
2
0
300
350
400
450
500
550
600
650
Wavelength (nm)
Figure 12 Solid-state fluorescence spectra of the hydrochloride salt of erythromycin-B,

obtained on the precipitated hydrate (upper spectra) and on its dehydrated product (lower
spectra). The excitation spectra (open circle) were obtained using an analyzing emission at
wavelength of 465 nm, while the emission spectra were obtained upon excitation at
350 nm (solid square) and at 390 nm (solid circle). Although the intensity scales are arbitrary,
the spectra are scaled to permit comparisons to be made. Source: Adapted from Ref. 27.
The erythromycin-B example clearly illustrated the sensitivity of photophysics to details of crystal structure. In another example, the different forms
of the disodium salt of a HMG-CoA reductase inhibitor (known to be capable
of existing in a number of hydrate species) were found to exhibit varying fluorescence properties in their respective solid states (112). Type I was known to be a
monohydrate species and was characterized by an excitation maximum at 345 nm
183
and a main fluorescence maximum at 371 nm. Type II was a dihydrate whose
photophysical parameters were equivalent to those of Type I but for which the
overall levels of fluorescence were greatly reduced. Incorporation of additional
waters of hydration into the substance produced Type III material, which exhibited an excitation maximum at 400 nm and a fluorescence maximum at 485 nm.
Finally, upon binding of nine water molecules, the structure relaxed into a liquid
crystalline form, which exhibited photophysical properties equivalent to those of
type III, but which were considerably reduced in intensity.
The trihydrate phases of ampicillin and amoxicillin are almost isostructural, and yet their photoluminescence spectra have been found to be quite sensitive to small differences in solid-state structure (113). The solid-state
excitation and emission spectra of both solids were dominated by energy transfer
and exciton effects, which were manifested as decreases in the energy of the excitation and emission bands of the solid-state systems relative to those of the free
molecule in solution. The photoluminescence data revealed that in spite of the
known structural similarity of ampicillin trihydrate and amoxicillin trihydrate,
the magnitude of the Davydov splitting and the degree of band energy shifting
differed between the two systems. This finding indicated that the small differences in crystal structure existing between the two compounds leads to measurable differences in the patterns of energy transfer.
When the crystal structures of different polymorphic forms of the same
compound are sufficiently different, these differences can be manifested in the
photophysics of the system. It was found that exciton effects dominated the excitation spectra of four polymorphic forms of diflunisal, leading to a decrease in the
energy of the excitation bands relative to that observed for the free molecule in
fluid solution, and in a splitting of the excitation peak into two Davydov components
(114). The trends in the excitation and emission spectra led to the grouping of diflunisal Forms I, II, and III into one category, and diflunisal Form-IV into a separate
category. This categorization is illustrated in Figure 13, where the excitation
spectra of Forms III and IV are plotted. Because it had been established that
Form-IV was characterized by the highest crystal density and degree of intermolecular interaction, the magnitude of the exciton coupling could be deduced as being
derived from the face-to-face overlap of the fluorescent salicylate-type fluorophores.
The anhydrous Form-III and dihydrate phases of carbamazepine were
found to exhibit identical maxima in their excitation spectra (347 nm), but the
intensities of the fluorescence associated with these solvatomorphs differed
widely (115). The peak maxima associated with the emission spectra were also
found to differ between the two forms, with 407 nm being observed for
Form-III and 416 nm for the dihydrate phase. Because the emissive intensity
of the dihydrate phase was approximately 3.15 times that of the anhydrate
phase, this difference was exploited to develop a method for the study of the kinetics of the aqueous solution-mediated phase transformation between these
forms. Studies were conducted at temperatures in the range of 18 408C, and
it was found that the phase transformation was adequately characterized by
184
Brittain
100
Intensity
75
50
25
0
250
300
350
400
Wavelength (nm)
450
Figure 13 Excitation spectra of diflunisal, Form-III (solid trace) and Form-IV (dashed
trace), normalized to a common intensity factor. Source: Adapted from Ref. 93.
first-order reaction kinetics. The temperature dependence in the calculated rate

constants was used to calculate an activation energy of 11.2 kCal/mol
(47.4 cal/g) for the anhydrate-to-dihydrate phase conversion.
PHOTOLUMINESCENCE FROM MOLECULES
ADSORBED ON SURFACES
As has been discussed earlier, the photophysics of molecules in their bulk crystalline state is often significantly changed by the effects of intermolecular energy
transfer, and the resulting excitation and emission spectra are often quite different
from those of the corresponding isolated molecule. It has been found that the
photophysics of luminescent molecules can also be altered in an entirely different
way by adsorbing these molecules onto a solid surface (116 118). In this latter
instance, the photophysics of the adsorbates will generally reflect details of their
interaction with the host surface and can even provide information regarding the
properties of the surface itself.
The types of research conducted on the photoluminescence from molecules
adsorbed onto a solid surface can be divided into two broad classes. One of these
is primarily chemical physics-oriented, where one typically uses the excitation
and emission spectroscopy of an adsorbed molecule to probe its intrinsic
185
photophysics as perturbed by the surface interaction. In this manner, one can

obtain information on either the probe molecule itself or on the surface contacting
the probe molecule. The other broad class of investigation is more analytical in
nature, where an analyst seeks to enhance some photophysical property of the
absorbed molecules in order to obtain qualitative or quantitative information
on their composition or concentration.
Photophysical Studies of FluorophoreSurface Interactions
Xanthene Dyes as Surface Probes
Owing to their favorable photophysical properties, and the sensitivity of these to
their immediate environment, xanthene dyes have been used to study a number of
solid surfaces. Lendvay published a comprehensive paper that illustrated how the
adsorption of fluorescein onto various basic surfaces perturbed the photophysics
of the system to differing degrees depending on the identity of the adsorbing substrate (119). In this study, fluorescein was adsorbed onto the surfaces of powdered
basic metal hydroxides, and the resulting changes in fluorescence spectra were
documented. Some of these experiments have been repeated, and the resulting
emission spectra are shown in Figure 14 (27). Even though fluorescein contains
only basic functional groups, its interactions with the surface hydroxyl groups of
the solids causes a red-shifting of the luminescence that is highly dependent on
the substrate. As evident in the figure, the shift of the emission band toward
longer wavelengths followed the order of Zn(OH)2 . Mg(OH)2 . Ca(OH)2,
undoubtedly reflecting differing degrees and types of hydrogen bonding
between the surface and the adsorbed dye.
When dye molecules are adsorbed onto crystal surfaces, it has been found
that the presence of distorted adsorption sites can lead to the observation of
reduced fluorescence lifetimes and altered photophysics. In one study, this
phenomenon was interpreted in terms of a quenching mechanism where rhodamine B probe molecules at the defect sites of glass and naphthalene crystals
acted as energy traps and as subsequent photoluminescent sites (120). In a
more comprehensive study, the interactions of rhodamine B, rhodamine 101,
and pyronine B with single crystals of naphthalene and phenanthrene were
studied (121). In this latter work, the additional mechanism of enhanced internal
conversion was also found to shorten the fluorescence lifetimes. In a detailed
analysis of the decay kinetics, it was learned that the shortest lifetime fluorescence was associated with the most heavily distorted adsorption site.
From studies conducted on rhodamine B in the solution phase, it was
learned that the compound could be induced to dimerize under the effect of
hydrostatic pressure (122). The monomer and dimer peaks of rhodamine B
were found to shift to lower energy levels with increasing pressure, as was anticipated for a p ! p excitation process. When the dye was adsorbed onto the
(0001) face of single crystals of zinc oxide, the presence of dimeric dye molecules was evident in the emission spectra as a second red-shifted peak, whose
186
Brittain
Relative Intensity
100
75
50
25
0
450
500
550
600
650
Wavelength (nm)
700
Figure 14 Emission spectra of fluorescein, normalized to a common intensity factor.

Spectra are shown for the solution phase emission (solid trace), and for the fluorophore
adsorbed onto Zn(OH)2 (C), Mg(OH)2 (A), and Ca(OH)2 (B). Source: Adapted from
Ref. 27.
intensity could exceed that of the monomer emission that was observed at low
applied pressures. Through a study of the changes in molecular volume upon
compression, it was deduced that the mechanism whereby rhodamine B interacted with the crystalline ZnO differed from the mode of interaction for cresyl
violet adsorbed onto the same substrate (123).
The effect on dye photophysics upon adsorption onto amorphous phases
has also been investigated, with picosecond photon counting fluorescence
profiles being used to evaluate the properties of cresyl violet adsorbed onto amorphous quartz (124). The fluorescence decays were found to be consistent with the
presence of monomeric cresyl violet molecules, and no conclusive evidence for
dimeric species could be obtained. In another study, time-resolved fluorescence
depolarization was measured by time-correlated photon counting for submonolayers of rhodamine 3B absorbed onto amorphous quartz (125). In this work, it
was found that when the surface coverage exceeded 0.4 times that of a monolayer, aggregates of rhodamine 3B that formed were perturbed the photophysics
in a less readily interpretable manner.
The phosphorescence of erythrosin B (tetraiodofluorescein) dispersed in
amorphous thin films of maltose and maltitol has been used to monitor the
187
molecular mobility of these matrices over the temperature range of 2258C to

over 1108C (126). Analysis of the energies of the emission peaks and their associated bandwidths, together with time-resolved intensity decay parameters, was
used to deduce information about the modes of thermally induced mobility,
because this mobility affected the rate of dipolar relaxation of the triplet state
and the rate of intersystem crossing to the ground state. The mobility measured
in mixed matrices was found to be less than that of either of the pure substance,
a phenomenon that was considered to be comparable with the antiplasticization
seen in mixtures of small molecule plasticizers with synthetic polymers and
starch.
Pyrene as a Surface Probe
Another probe molecule that has been used in many studies is the polynuclear
aromatic hydrocarbon, pyrene. This compound is known to form excimers in
both the solution phase and in its crystalline state (14), and the details of its
vibrational structure within its monomer emission are known to be dependent
on the polarity of its environment (127). In a particular work, it was found
from time-resolved spectral studies that both pyrene and naphthalene exhibited
excimer-like fluorescence after being adsorbed onto silica gel (128). Because
no ground-state complex was known to exist, this finding was viewed as being
indicative of substantial interaction between the adsorbed molecules in their
excited states.
A detailed study of the excitation and emission spectra, and the associated
fluorescence decay, of pyrene adsorbed on silica gel, porous Vycor glass,
alumina, and calcium fluoride have been reported (129). When silica was used
as the solid substrate, monomer emission and a longer-wavelength, broadband, emission that decayed in a nonexponential manner was observed. The distribution of pyrene on the silica surfaces was deduced not to be random with its
bimolecular association and its associated excimer-type fluorescence was stabilized at certain preferential sites. The data also indicated that the phenomenon was
probably related to peculiarities in the hydrogen-bonding interaction between the
surface and the p-electron system of the polynuclear aromatic hydrocarbon.
The interaction of surfactants with solid surfaces can be studied by first
forming the substrate, then adsorbing the luminescent probe, and finally using
the photophysical characteristics of the system to evaluate the surface characteristics. For example, the interaction of the surfactant Tritron X-100 [i.e., octylphenol poly(oxyethylene)] with solid silica surfaces has been investigated using
pyrene as the fluorescent probe molecule (130). The approach in this work was
to solubilize the probe molecule within the hydrophobic regions of the adsorption
layer and to use the decay kinetics to understand the spatial extent to which the
probe was free to move. It was concluded that the surfactant aggregates formed
on the surface greatly resembled those of regular solution-phase micelles,
although the regimes of structure type were found to be concentration dependent.
188
Brittain
Because 1,3-di-1-pyrenylpropane is capable of forming intramolecular

excimers, its use as a bifunctional pyrene probe molecule has been advanced
as a means to avoid experimental complications associated with ground-state
aggregation (131). The compound was adsorbed onto silica in the presence and
absence of coadsorbed octanol, and also onto octadecylsilica, and then the photophysics studied using steady-state and single-photon counting methods. Evidence
was obtained for dynamic intramolecular excimers on silica surfaces in the presence of octanol but not when the probe was adsorbed onto untreated silica. This
latter effect on untreated silica was attributed to complete immobilization of the
fluorescent probe, which prevented the occurrence of the excited-state association. The presence of a monolayer of adsorbed octanol on the silica surface
served to provide the necessary molecular mobility to form excimer species,
which paralleled those formed by pyrene in the same system.
The polarity of both polymeric and monomeric C18 chromatographic
stationary phases has been evaluated from the fluorescence vibronic fine structure
of pyrene adsorbed onto the surfaces of the packing materials (132). It was
learned that the intercalation of the organic modifier into the stationary phase produced an inverse relationship between the polarity of the particle surface and that
of the mobile phase, and that when the mobile phase contained the highest water
content to adsorbed probe became partially exposed to the solution phase. This
latter effect became manifested in perturbations of the photophysics of the
probe molecule.
The illustrations given here for studies involving the characterization of
xanthene dye and pyrene fluorescence probes adsorbed onto various solid surfaces represent only a few of the countless works published in this area. It is
safe to state that significant advances in colloid science have been derived
from the study of fluorescent probe molecules and their interactions with
various interfacial structures, and interested readers are referred to the literature
for discussions of these applications (133 135).
RTP of Adsorbed Luminescent Molecules
It is well established that techniques associated with luminescence spectroscopy
have great analytical potential owing to the high degrees of sensitivity that can be
achieved, and the selectivity that can be derived from a judicious choice of excitation and emission wavelengths. When luminescence studies are performed on
solid surfaces, a variety of mechanisms serve to enhance the photophysical processes, and it is not uncommon to achieve limits of detection in the nanogram or
subnanogram range. Because the instrumentation required for such work can be
relatively simple and inexpensive, and because the analysis time can be exceedingly rapid, solid-surface luminescence spectroscopy has been extensively investigated for its potential as analytical methodology.
One of the more interesting consequences of adsorbing a sample onto a
solid surface is that intersystem crossing from the manifold of singlet states
189
can be facilitated into the lowest energy triplet state. This process is illustrated in
the Jablonski-type energy level diagram of Figure 15, where the primary excitation into the lowest energy singlet state does not result in fluorescence but is
instead followed by intersystem crossing into the lowest energy triplet state.
The process of phosphorescence can then take place from the triplet state, in
the form of the spin-forbidden transition down to the ground state. Phosphorimetry can, of course, be practiced in the solution phase but usually requires
that the sample be cooled to cryogenic temperatures before the effect can be
detected. When the luminescent material is absorbed onto certain solid surfaces,
typically coabsorbed with an agent containing an atom of high atomic number
that can promote the intersystem crossing process, room temperature phosphorescence can often be observed.
Vo-Dinh has summarized a number of general trends regarding the room
temperature phosphorescence (RTP) of organic compounds and has provided
extensive tables of organic compounds and their RTP properties (117). Polynuclear aromatic compounds are often intrinsically weakly phosphorescent due
to the large singlet-triplet energy gap and the weak spin-orbit coupling between
the pp singlet and triplet states. Upon adsorption onto a solid surface in the
Figure 15 Jablonski-type energy level diagram of the ground and excited states of a fluorescent organic molecule, illustrating (A) the primary excitation into the lowest energy
singlet state, (B) intersystem crossing into the lowest energy triplet state, and finally (C)
phosphorescence taking place from the triplet state in the form of the spin-forbidden transition down to the ground state.
190
Brittain
presence of a heavy atom, RTP is often observed. Nitrogen heterocyclic compounds normally do not exhibit phosphorescence but can exhibit a highly
appreciable RTP upon adsorption on cellulose surfaces in the presence of
silver or thallium salts. Several other substantial reviews of the phenomenon
are available (116 118,136 138).
Photophysical and Mechanistic Studies
During a study of the decomposition of benzoyl peroxide in tetrachloroethylene
by thin-layer chromatography, strong blue-green phosphorescence was observed
from dried spots corresponding to the ammonium salts of biphenylcarboxylic
acids (139). In studying this effect further, it was found that RTP could be
obtained from the salts of a wide variety of the salts of polynuclear carboxylic
or sulfonic acids, phenols, and amines adsorbed onto paper, silica, alumina,
and other supports. A general trend reported in this paper was that the emission
and excitation spectra of the adsorbed species were very similar to the spectra
obtained from frozen solutions at cryogenic temperatures and that the emission
lifetimes of the adsorbed molecules on the millisecond time scale were known
to be typical for phosphorescence.
The effects of moisture, oxygen, substrate on the photophysical properties
of the sodium salts of 4-biphenylcarboxylate and 1-naphthoate adsorbed onto cellulose paper, sucrose, starch, glass fiber paper, silanized paper, and stainless steel
have been studied (140). It was learned that RTP could only be generated when
the surface contained large quantities of available hydroxyl groups, presumably
to appropriately bind and orient the adsorbed fluorophores through hydrogen
bonding. At the same time, it was found that moisture and oxygen quenched
the RTP, and hence an appropriate sample matrix would be one that was set
up to resist the penetration and subsequent quenching effects of these agents.
The effect of temperature on the quantum yield for solid-surface luminescence of p-aminobenzoic acid adsorbed onto sodium acetate has been reported
over the temperature range of 21808C to 238C, with these data being supplemented by measurements of the phosphorescence lifetime over a slightly
wider temperature range (141). It was found that although the fluorescence
quantum yields were almost temperature independent, the quantum yield for
phosphorescence exhibited a strong temperature dependence. The results supported the model of a rigidly held fluorophore adsorbed onto the substrate, but
where the adsorbed acid had been converted into its anion.
The effect of oxygen on the solid-matrix phosphorescence of perdeuterated
phenanthrene adsorbed on different grades and types of paper has been reported
(142). Effects associated with molecular diffusion on the surfaces were found to
be minimal, and diffusional quenching did not contribute greatly to the photophysics. Static quenching by oxygen was found to be more significant on partially
hydrophobic paper than on the usual Whatman filter papers, and this effect is
illustrated in Figure 16. The luminescence decay results were evaluated using
both linear and nonlinear fitting routines but yielded comparable results.
191
1.4
1.3
0 /
1.2
1.1
1.0
0.9
0
12
16
[Oxygen] (%)
20
Figure 16 Ratios of the solid-matrix phosphorescence lifetimes obtained in the absence

(t0) and presence (t) of oxygen for perdeuterated phenanthrene adsorbed onto partially
hydrophobic filter paper (solid trace) and on Whatman No. 1 filter paper (short dashed
trace). Source: Adapted from Ref. 142.
The effect of coadsorbed alkali halide salts on the RTP of 2-naphthalenesulfonate adsorbed onto filter paper has been examined to determine the contribution
made by the heavy atom effect (143). It was found that the RTP of this fluorophore
was enhanced in the anticipated order of NaI . NaBr . NaCl . NaF, indicating
that the external heavy atoms increase the radiative triplet decay constant more
than the competing nonradiative rate constant. The results were consistent with
the existence of an exchange interaction model between the heavy atom perturber
and the fluorophore, where singlet $ triplet transition intensity was borrowed
from electronic transitions of the perturbing halide ion.
The RTP characteristics of several aromatic ketones adsorbed onto filter
paper were established, and compared with the phosphorescence spectra obtained
at low temperatures in frozen matrices (144). It was found that ketones for which
the cryogenic luminescence originated from the 3(np ) state did not exhibit
appreciable RPT, but that when the glassy phosphorescence originated from
the 3(pp ) or triplet charge-transfer state, the room temperature phosphorescence
was fairly intense. In addition, strong hydrogen bonding of the fluorophore to the
hydroxyl groups of the filter paper was a prerequisite for the observation of RTP.
Both room temperature phosphorescence and delayed charge-transfer
fluorescence were observed from coronene absorbed onto thermally activated
192
Brittain
g-alumina (145). The fluorescence was only observed at low temperatures and was
deduced to originate from the charge-transfer singlet state of an electron donor
acceptor complex between the adsorbed coronene and Lewis acceptor centers on
the surface. On the other hand, the phosphorescence efficiency was found to be
almost constant over the entire range of temperatures studied (100 to 500 K) and
was attributed to a locally excited triplet state of the adsorbed coronene.
A method for the determination of quantum yield efficiencies of room temperature fluorescence or phosphorescence of organic molecules adsorbed onto a
variety of solid surfaces has been described (146). The method makes use of a conventional scanning luminescence spectrometer and entails careful measurement of
instrumental and sample blanks for proper referencing. The quantum yield of
sodium salicylate adsorbed onto solid sodium acetate was established as a reference standard, with a value of 0.52 + 0.03 being reported for a 1% dispersion
in the absence of a nitrogen flow. The quantum yields of unknowns could then
be easily determined as relative values through comparison with the standard,
and such values were reported for a variety of adsorbed fluorophores.
Hurtubise has contributed an extensive review on the trace analysis of
compounds using solid-matrix luminescence analysis (147). In this review, he
compared the quantum yields for fluorescence and phosphorescence of several
emissive compounds adsorbed onto a variety of solid matrices and compared
many of these with the analogous solution phase values. Also discussed were
the importance of rigidity in the substrate, the effects of water and gases on the
photophysics, the effects of heavy atoms, and the heat capacity of the various
systems.
Studies of Substrate Surfaces Appropriate for RTP Work
When adsorbed fluorophores are incapable of interacting in a site-specific manner
with functional groups on the substrate surface, RTP will not be observed. As a
result, a number of studies have been conducted where specific fluorophore types
were adsorbed onto a wide variety of solid surfaces, and the resulting photophysical properties of the adsorbates were used to characterize the nature of the
fluorophore substrate interaction. For example, three molecules having known
spectroscopic properties (benzo[a]phenazine, 1,4-diazatriphenylene, and Michlers ketone) were adsorbed onto Whatman 4 filter paper, and it was learned
that the paper provided a very polar environment with a high degree of hydrogen
bonding activity for the adsorbed molecules (148).
The RTP phenomenon has been investigated for eight indolecarboxylic
acids adsorbed onto several ion-exchange filter papers (149). It was found that
the largest RTP signals were obtained when using DE-81 anion-exchange
paper, and substantial heavy atom effects were noted when using sodium
iodide as a coadsorbate. In a related study, various Whatman filter and anionexchange papers were evaluated for their utility as RTP substrates (150). The
intensity of the RTP signals on these papers was found to be largely independent
of the method used for the sample preparation, and also generally were more
193
intense than the RTP intensities obtained for the same fluorophores when those
were adsorbed onto silica gel thin-layer chromatography plates.
In another study, the RTP characteristics of a large number of aromatic
ketones adsorbed onto a variety of solid substrates have been reported (151). In
this work, the ability of silica gel chromaplates, filter paper, filter paper containing
poly(acrylic) acid, and sodium chloride doped with 0.5% poly(acrylic) acid to
promote RTP from adsorbed fluorophores was studied. Although it was found
that no one surface yielded optimal RTP for all of the model compounds used,
good analytical data were obtained from the majority of the surfaces investigated.
However, the choice of the substrate still remained somewhat empirical, as no
strong predictive model emerged from the work. For example, 2-acetonaphthone
did not yield RTP after adsorption onto filter paper but did so from the other
substrates studied. Were one to restrict the choice of surfaces to the family of
filter papers (as has been recommended elsewhere), one would reach the erroneous
conclusion that this particular compound could not exhibit the RTP phenomenon.
The RTF and RTP quantum yields, intersystem crossing efficiencies, and
phosphorescence lifetime values have been obtained for p-amino-benzoic acid
adsorbed onto sodium acetate or onto mixtures of sodium acetate and sodium
chloride (152). It was deduced that the efficiency of RTF and RTP were determined primarily by how efficiently the support matrix became packed with the
fluorophore molecules, although the effect was more important for the fluorescence process than it was for the phosphorescence process. Key to the luminescence induction process appeared to be the formation of a salt upon interaction of
the p-aminobenzoic acid with the sodium acetate substrate, with the resulting
anion being the species adsorbed into the matrix.
The luminescence properties of the p-aminobenzoate anion adsorbed onto
sodium acetate and filter paper were compared to evaluate the interactions that
determined the relative yields for fluorescence and phosphorescence (153). A
relationship was found between the reciprocal of the phosphorescence lifetimes
and the thermal processes that led to deactivation of the triplet state. Analysis of
the results indicated that the p-aminobenzoate anion was incorporated into the
sodium acetate crystal structure, and that its triplet energy was lost primarily
into the skeletal vibrations of the support. For the anion adsorbed onto filter
paper, it was concluded that although some of the triplet energy was lost
into the vibrational modes of the filter paper, other factors were involved in
the deactivation of the triplet state.
Interesting adsorption effects were observed when generating analytical
calibration curves for benzo[ f ]quinoline adsorbed onto a silica thin-layer
chromatography place under different pH conditions (154). As the concentration
of analyte was increased, the calibration curves for the RTF were found to eventually level off at a constant value, whereas those for RTP increased to a
maximum value and then decreased later. These findings were interpreted to indicate that the fluorescence process could take place from molecules adsorbed onto
the surface and from multilayers of these molecules, but that the phosphorescence
194
Brittain
process could only originate from molecules adsorbed onto the surface itself and
not in the multilayers.
Mixtures of cyclodextrins with sodium chloride have been found to constitute good substrates for the generation of RTP. Fifty-five organic compounds
containing various functional groups were adsorbed onto an 80% a-cyclodextrin NaCl matrix, and a large number of these yielded photophysical properties
that were deemed to be appropriately selective and sensitive for development of
analytical methods (155). In another study, it was found that use of a matrix consisting of 30% b-cyclodextrin NaCl yielded appreciable RTP in most of the
molecules studied without the use of an external heavy atom (156). In this
latter work, it was shown that the luminescence spectra obtained using the cyclodextrin matrix possessed more fine structure than did the corresponding spectra
obtained using filter paper as the substrate and that the degree of fine structure
approach that was observed for the fluorophores was frozen in cryogenic media.
The enhancement of RTP from molecules adsorbed onto filter paper by the
co-adsorption of a-, b-, or g-cyclodextrins in the presence of heavy atoms has
been studied (157). For most compounds, the use of a-cyclodextrin did not
yield appreciable RTP, and it was also found that the RTP intensities could be
amplified if the filter paper was pretreated with the cyclodextrin. However, it
was clear that once the substrate surface sites were saturated with the cyclodextrin, the degree of RTP enhancement reached a limiting value.
The energy gap law showing an inverse correlation between the rates of
non-radiative transitions from the lowest excited triplet states to the ground
states and the differences in energies between the lowest vibrational energy
levels of the excited and ground states has been shown to be applicable for the
RTP of polynuclear aromatic molecules adsorbed onto cyclodextrin-salt matrices
(158). Changes in the phosphorescence lifetime of the adsorbed fluorophore were
correlated with the magnitude of the nonradiative rate constants, which were in
turn correlated with the energy gap between the excited triplet and the ground
states. Thus, upon measurement of the phosphorescence lifetime for a polynuclear aromatic hydrocarbon and the energy of its triplet state, one could
predict whether the magnitude of the observed RTP intensity would be strong
or weak. For example, given the phosphorescence lifetimes characteristic for
polynuclear aromatic hydrocarbons, those molecules whose triplet energy was
less than 12,000 cm21 should not exhibit RTP.
Glasses prepared by the controlled evaporation of glucose or trehalose solutions have been found to function as matrices for the induction of RTP and suitable for use in trace analysis (159). The advantage of immobilizing the analyte in
a glassy matrix is that the photophysics is essentially unaffected by moisture or
oxygen, thus yielding a more robust analytical procedure. The disadvantage of
the procedure is that the RTP signals tended to decrease over time, suggesting
that the glassy matrices were not sufficiently rigid to withstand UV-induced
damage. However, under the right set of circumstances, the easily prepared
glucose glasses would be a superior choice for solid-matrix luminescence studies.
195
Glucose glasses have also been obtained from the molten state, and the
solid-matrix luminescence of heterocyclic aromatic amines in these solids has
been reported (160). The melt technique that permitted the exclusion of water
from the system enhanced the luminescence quantum yields in most cases. In
another work, the effects of glass preparative method and role of heavy atoms
in the matrices were studied (161). It was deduced that the phosphorescence
intensity was a function of how efficiently the triplet state could be populated
and how quickly the triplet state became depopulated through phosphorescence.
When sodium iodide was included in the glass matrix as the heavy atom, the
triplet state formation became temperature independent, making the phosphorescence intensity dependent only on its lifetime.
The quenching of the solid-matrix phosphorescence associated with heterocyclic aromatic amines by water was studied and additional characterization performed by NIR spectroscopy and differential scanning calorimetry (DSC) (162).
The NIR method was used to determine the water content in the glasses, and a
modified Stern-Volmer equation was used to describe the changes in phosphorescence intensities and lifetimes. The DSC method was used to evaluate the
glass transition temperatures of the matrices, and these values correlated with
the photophysical results. The microenvironment of the glucose glasses was evaluated using a polarity probe, facilitating development of a model that described
the effects of the different parameters on the photophysical properties. For
example, water was found to be able to diffuse though channels in the glasses
to cause dynamic quenching and also served to disrupt the hydrogen bonding
network that structured the glasses.
Analysis of Mixtures
As discussed earlier, luminescence spectroscopy contains an extra element of
selectivity in that even for steady-state work, the analyst has control over both
the excitation and emission wavelengths. The selectivity possible with the technique becomes even more advantageous when one adds in additional refinements,
such as time resolution or mathematical processing of the data. For example,
derivative techniques were used for the identification of components in binary
and ternary mixtures of nitrogen heterocyclic compounds (160). Direct, first,
and second derivative RTP spectra were used for such identifications by matching
spectral wavelengths from mixtures with the spectral wavelengths of standards.
Different nitrogen heterocyclic compounds and a polycyclic aromatic
hydrocarbon were combined to form various binary and ternary mixtures, and
then the components were determined at the nanogram level by RTF or RTP
in conjunction with selective excitation or emission monitoring (161). Multicomponent mixtures of polynuclear aromatic hydrocarbons were absorbed onto
a-cyclodextrin/sodium chloride mixtures, and then characterized by their
solid-state photoluminescence (162). Using a protocol involving extraction,
selective excitation, and photoluminescent detection, mixtures containing as
many as nine components could be analyzed at the nanogram level.
196
Brittain
The RTP obtained from eight benzoquinone isomers can be resolved into the
linear or angular subgroups on the basis of the conventional luminescence spectra,
but the selectivity can be improved using second derivative and synchronous scanning techniques (163). The approach was applied to determining the levels of acridine, benzo(h)quinoline, and phenanthridine in a coal tar fraction. In this work,
substantial differences were noted between luminescence and HPLC measurements, suggesting that although photoluminescence can be used as a screening
tool, it should be combined with a separation step to obtain optimal results.
In another report, the use of combined solid-surface fluorescence and phosphorescence for the determination of several luminescence compounds in different
mixtures has been studied (164). Parameterssuch as the pH of the sample environment, the nature and preparation of the solid substrates used, and the choice of heavy
atom perturberswere used to selectively determine the components. It was found
that even more selectivity could be introduced into the RTF and RTP approaches
by working with the optimal substrate and heavy atom, effectively removing contributions made by substances other than the analytes.
An analysis method, based on solid-matrix luminescence, has been
reported for the determination of 1-hydroxypyrene, tetrol, and 9-hydroxyphenanthrene in urine fractions (165). The greatest analytical sensitivity was obtained in
the phosphorescence mode, with the compounds being absorbed onto Whatman
1PS paper with thallium nitrate as the heavy-atom agent. The 9-hydroxyphenanthrene metabolite was found to undergo photobleaching under the analysis conditions, so its limit of detection could only be estimated. However, by taking
suitable advantage of the differing photophysical parameters, it was possible to
identify all compounds in spiked urine fractions.
Compounds Having Pharmaceutical Interest
The utility of RTP spectroscopy to the determination of 28 pharmaceutical compounds has been reported, with deliberate changes in experimental conditions
being used to maximize the sensitivity (166). It was reported that the best
heavy atom perturber and promoter of RTP was iodide ion and that Ag(I)
appeared to have little effect. Generally, when RTP was observed, the emission
bands were shifted to lower energies relative to the phosphorescence bands
obtained in frozen cryogenic matrices. Interestingly, a connection between
molecular structure and phosphorescence characteristics could not be drawn,
with closely related compounds often exhibiting vast different RTP intensities.
However, the RTP methodology was shown to be simple and selective and,
therefore, suitable for use in clinical analysis.
The effects of anionic, nonionic, and cationic surfactants on the RTP of 11
phenothiazine derivatives have been described (167). For these compounds,
strong RTP could be observed once the compounds were adsorbed onto filter
paper, but in the presence of coadsorbed thallium(I) nitrate as a heavy atom perturbing agent, a tripling of the RTP signal could be obtained. Interestingly, use of
the Tl(I) heavy atom in the presence of the surfactants resulted in a quenching of
the RTP. As part of a much larger study of 30 compounds having pharmaceutical
197
interest, a study of eight phenothiazine compounds was conducted (168). A wide

variety of photophysical properties were noted for the various compounds, and
the intensity of RTP for a given substance was not readily predictable. In fact,
more compounds were found to phosphoresce at room temperature when
adsorbed onto filter paper that could phosphoresce in frozen ethanol solutions.
Room temperature phosphorimetry of a mixture of two methylxanthines
has been reported, but because the spectral characteristics of caffeine and
theophylline are so similar, the prior use of a separation procedure was required
(169). In another work, RTP was used as an alternative detection technique for
the chromatographic determination of caffeine, theophylline, and theobromine
in urine samples (170). Although the analytical performance parameters associated with the RTP and conventional UV absorbance methods were comparable,
superior selectivity was achieved with the RTP. One of the more interesting
aspects of the latter method was the idea that chromatograms could be stored
to document the assay results, but this proposal has become dated in this day
of automated data documentation.
The RTP properties of diazepam were investigated as a function of the type
of filter paper used as the solid substrate, the choice of heavy atom used to
enhance the triplet state yield, and the pH of the initial spotting solution (171).
It was concluded that the largest RTP signal for diazepam was obtained when
the substance was adsorbed onto Whatman No. 1 filter paper from an acidic solution, and using Hg(II) as the heavy atom. Tablets containing 10 mg of active
ingredient could be assayed by the RTP method with good analytical performance parameters, and the method could also be used to obtain good results for
the determination of diazepam in serum. Spiking of diazepam into serum
yielded low recoveries because the substance is known to be highly proteinbound, but through the use of an extraction procedure the analyte could be quantitatively recovered.
The mutagenic and carcinogenic properties of polycyclic aromatic hydrocarbons are well documented, and therefore there is a continuing interest in methodology to evaluate their toxic effects. Some of these compounds can be
metabolized into compounds that are able to form adducts with DNA, and
these adducts have been the focus in cancer research. One compound that has
received extensive interest is benzo[a]pyrene, and not surprisingly it has been
found that solid-matrix RTP a useful technique for the characterization of its
adducts with DNA.
Solid-matrix fluorescence and phosphorescence were obtained for DNA
adducts of benzo[a]pyrene when these complexes were adsorbed on Whatman
No. 1 and 1PS filter papers together with thallium nitrate as the heavy-atom
inducer (172). Using selective excitation, it was possible to excite different
forms of the adducts, and moisture quenching of the phosphorescence was
used to detect quasi-intercalated adducts. In other studies, solid-matrix phosphorescence was used study the adducts formed by DNA with benzo[a]pyrene-7,8dihydrodiol-9,10 epoxide (173) and racemic anti-dibenzo[a,1]pyrene11,12-diol-13,14-epoxide (174).
198
Brittain
It has been established that isomeric tetrols of benzo[a]pyrene are formed by

the acid hydrolysis of the benzo[a]pyrene-DNA adducts, and consequently study
of the spectroscopic properties of these tetrols has proven important. In one study,
four tetrols were separated by HPLC, and the chromatographic fractions characterized as to their solid-matrix phosphorescence properties (175). It was reported
that the compounds could be detected at the pictogram levels, and that the luminescence method could be used to identify levels of compounds that were
undetectable by HPLC. Additional luminescence studies have been reported for
tetrols formed from benzo[a]pyrene-DNA adducts and that have been supported
on a-cyclodextrin/NaCl (176) and b-cyclodextrin/salt mixtures (177).
The effects of a heavy-atom salt on the solid matrix luminescence properties of the racemic trans-7,8-dihydroxy-anti-9,10-epoxy-7,8,9,10-tetrahydrobenzo[a]pyrene-DNA adducts and tetrol I-1 were studied on several substrates (178).
Although thallium nitrate was effective in promoting RTP, the RTF was
quenched. It was futher reported that the most superior substrates for inducing
phosphorescence were 1% a-cyclodextrin/TlNO3, 10% trehalose/TlNO3, and
30% TlNO3/sodium acetate. Subsequently, the quenching of the solid matrix fluorescence by sodium iodide and thallium nitrate was used to investigate the
nature of the adducts (179). When using thallium nitrate as the quenching
agent, a two-site model with two independent quenching sites was applicable
to the data. However, for sodium iodide quenching, the data for tetrol I-1 were
fit to the sphere of action model, but the data for the DNA adducts was qualitatively related to a BET isotherm. Additional insight into the quenching mechanisms was derived in a subsequent study, where it was also learned that the DNA in
the adducts were held very rigidly to the matrix, thus facilitating their interaction
with the heavy atom (180).
The RTP approach has been applied to the study of other polynuclear aromatic
hydrocarbons. The solid matrix spectroscopy of benzo[ f ]quinoline and benzo[h]quinoline adsorbed on silica gel chromatography subsequently submerged in
mixed chloroform/hexane solvents (181). The phosphorescence data were interpreted to indicate that the emissive compounds were protected from collisional
deactivation by the matrix and that adsorbed chloroform minimally disrupted the
interactions between the adsorbate and the substrate. Fluorescence and phosphorescence quantum yields, polarization values, and emission lifetimes were obtained
for benzo[ f ]quinoline adsorbed onto b-cyclodextrin/NaCl matrices over a wide
range of temperatures (182). As is often noted in systems of this type, the fluorescence data exhibited only a minor temperature dependence, but temperature
was found to exert a strong influence on the phosphorescence parameters.
REFERENCES
1. Pringsheim P. Luminescence of Liquids and Solids. New York: Interscience Pub., 1943.
2. De Ment J. Fluorochemistry. New York: Chemical Pub. Co., 1945.
3. Pringsheim P. Fluorescence and Phosphorescence. New York: Interscience Pub., 1949.
199
4. Radley JA. Fluorescence Analysis in Ultraviolet Light. London: Chapman & Hall,
1954.
5. Forster T. Fluoreszenz Organischer Verbindungen. Gottingen: Vandenhoeck und
Ruprecht, 1951.
6. Berlman IB. Handbook of Fluorescence Spectra of Aromatic Molecules. New York:
Academic Press, 1965.
7. Hercules DM. Fluorescence and Phosphorescence Analysis. New York: Interscience
Pub., 1966.
8. Calvert JG, Pitts JN. Photochemistry. New York: John Wiley & Sons, 1966.
9. Guilbault GG. Fluorescence: Theory, Instrumentation, and Practice. New York:
10. Parker CA. Photoluminescence of Solutions. New York: Elsevier Pub., 1968.
11. Zander M. Phosphorimetry. New York: Academic Press, 1968.
12. Becker RS. Theory and Interpretation of Fluorescence and Phosphorescence.
New York: Wiley-Interscience, 1969.
13. Glynn SP, Azumi T, Kinoshita M. Molecular Spectroscopy of the Triplet State. Englewood Cliffs, NJ: Prentice-Hall, 1969.
14. Birks JB. Photophysics of Aromatic Molecules. New York: J. Wiley & Sons, Inc.,
1970.
15. White CE, Argauer RJ. Fluorescence Analysis: A Practical Approach. New York:
16. Winefordner JD, Schulman SG, OHaver, TC. Luminescence Spectrometry in
Analytical Chemistry. New York: John Wiley & Sons, 1972.
17. Guilbault GG. Practical Fluorescence: Theory, Methods, and Techniques.
New York: Marcel Dekker, 1973.
18. Schulman SG. Fluorescence and Phosphorescence Spectroscopy: Physicochemical
Principles and Practice. Oxford: Pergamon Press, 1977.
19. Lakowicz JR. Principles of Fluorescence Spectroscopy. New York: Plenum Press,
1983.
20. Demas JN. Excited State Lifetime Measurements. New York: Academic Press, 1983.
21. Baeyens WRG, De Keukeleire D, Korkidis K. Luminescence Techniques in Chemical and Biochemical Analysis. New York: Marcel Dekker, 1991.
22. Lakowicz JR. Principles of Fluorescence Spectroscopy. 2nd ed. New York: Kluwer
Academic/Plenum Pub., 1999.
23. Udenfriend S. Fluorescence Assay in Biology and Medicine. New York: Academic
Press, volume 1 (1962); volume 2 (1969).
24. Wehry EL. Modern Fluorescence Spectroscopy. New York: Plenum Press, volume 1
(1976); volume 2 (1976); volume 3 (1981); volume 4 (1981).
25. Schulman SG. Molecular Luminescence Spectroscopy. New York: John Wiley &
Sons, volume 1 (1985); volume 2 (1988); volume 3 (1993).
26. Passwater RA. Guide to the Fluorescence Literature. New York: Plenum Press,
volume 1 (1967); volume 2 (1970); volume 3 (1974).
27. Brittain HG. unpublished fluorescence results obtained using a Perkin-Elmer model
LS-5B luminescence spectrometer, whose sample compartment has been modified to
facilitate front-surface excitation and emission spectroscopies of solids contained in
thin-walled glass tubes.
28. Bowen EJ. Proc Royal Soc London 1936; A154:349.
29. Strickler SJ, Berg RA. J. Chem Phys 1962; 37:814.
200
Brittain
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
Stern O, Volmer M. Phys Z 1919; 20:183.

Bowen EJ. Trans Far Soc 1954; 50:97.
Boaz H, Rollefson GK. J Am Chem Soc 1950; 72:3435.
Bowen EJ, Metcalf WS. Proc Royal Soc London 1951; A206:437.
Keizer J. J Am Chem Soc 1983; 105:1494.
Blatt E, Mau AW-H, Sasse WHF, Sawyer WH. Aust J Chem 1988; 41:127.
Gabor G, Turro NJ. Photochem Photobiol 1985; 42:447.
See reference [14], pp. 420 491.
Forster T, Kasper K. Z Physik Chem (Frankfurt) 1954; 1:275.
Forster T, Kasper K. Z Elektrochem 1955; 59:977.
For a detailed discussion of excimer kinetics, see reference [10], pp. 347 356.
Kasper K. Z Physik Chem (Frankfurt) 1959; 12:52.
Doller E, Forster T. Z Physik Chem (Frankfurt) 1962; 34:132.
Hirayama F. J Chem Phys 1965; 42:3363.
Chandross EA, Dempster CJ. J Am Chem Soc 1970; 92:3586.
Handa T, Utena Y, Yajima H. J Phys Chem 1984; 88:5150.
Brittain HG, Ambrozich DL; Saburi M, Fendler JH. J Am Chem Soc 1980; 102:6372.
Anufrieva EV, Ghiggino KP, Gotlig YY, Phillips D, Roberts AJ. Luminescence,
Advances in Polymer Science. 40:1981.
Ander P, Mahmoudhagh MK. Macromolecules 1982; 15:213.
Turro NJ, Ukubo T. J Phys Chem 1982; 86:1485.
Froehlich P, Wehry EL. The Study of Excited-State Complexes (Exciplexes) by
Fluorescence Spectroscopy, chapter 8 in Modern Fluorescence Spectroscopy.
Wehry EL. ed., volume 2, New York: Plenum Press, 1976, pp. 319 438.
Knibbe H, Rehm D, Weller A. Ber Bunsenges Phys Chem 1968; 72:257.
Lewis FD. Acc Chem Res 1979; 12:152.
Lewis FD, Kalgutkar RS, Kurth TL. J Phys Chem 2004; 108:1425.
Irie M, Yorozu T, Hayashi K. J Am Chem Soc 1978; 100:2236.
Kroger FA. Some Aspects of the Luminescence of Solids. New York: Elsevier, 1948.
Garlick GFJ. Luminescent Materials. Oxford: Clarendon Press, 1949.
Leverenz HW. An Introduction to Luminescence of Solids. New York: John Wiley &
Sons, 1950.
Kallmann HP, Spruch GM. Luminescence of Organic and Inorganic Materials.
New York: John Wiley & Sons, 1962.
Curie D. Luminescence in Crystals. London: Methuen & Co., 1963. This book is a
translation of Luminescence Cristalline. Paris: Dunod Pub., 1960.
Goldberg P. Luminescence of Inorganic Solids. New York: Academic Press, 1966.
Frenkel JI. Phys Rev 1931; 37(17): 1276.
Davydov AS. Theory of Molecular Excitons. New York: McGraw-Hill, 1962.
McClure DS. Electronic Spectra of Molecules and Ions in Crystals. New York: Academic Press, 1959.
Knox RS. Theory of Excitons. New York: Academic Press, 1963.
Craig DP, Walmsley SH. Excitons in Molecular Crystals. New York: Benjamin WA,
1968.
Brittain HG, unpublished fluorescence results obtained using high-resolution
spectrometers constructed at the Center for Pharmaceutical Physics. The excitation
spectrometer used a 0.5 meter monochromator (Spex model 1870) to discriminate
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
201
the output of a 300 watt xenon arc lamp at a resolution of 0.6 nm. The fluorescence
was obtained using front-face excitation, allowed to pass through a long-pass filter
(1% w/v solution of sodium nitrite in a 1 cm cell) to remove scattering light, and
then detected by an end-on photomultiplier tube (Thorn EMI type 6256QB having
S-11 response). The emission spectrometer used a combination of glass and solution
filters to obtain the desired output of a 250 watt xenon arc lamp. The filtered light
impinged on the sample in the front-face excitation mode, and the fluorescence analyzed by a 0.5 meter monochromator (Spex model 1870) at a resolution of 0.8 nm.
The fluorescence was detected by an end-on photomultiplier tube (Thorn EMI
type 9558QB having S-20 response).
Mielenz KD. Measurement of Photoluminescence. New York: Academic Press, 1982.
Budde W. Physical Detectors of Optical Radiation. New York: Academic Press, 1983.
Talmi Y. Multichannel Image Detectors. Washington, DC: ACS Symposium Series
102, American Chemical Society, 1979.
Christian GD, James B. Callis, Davidson ER. Array Detectors and ExcitationEmission Matrices in Multicomponent Analysis, chapter 4 in Modern Fluorescence
Spectroscopy. volume 4, Wehry EL. ed., New York: Plenum Press, 1981, pp. 111165.
Vo-Dinh T. Anal Chem 1978; 50:396.
Lloyd JBF, Evett IW. Anal Chem 1978; 50:1710.
Green GL, OHaver TC. Anal Chem 1974; 46:2191.
John P, Soutar I. Anal Chem 1976; 48:520.
Munoz de la Pena A, Salinas F, Duran-Meras I. Anal Chem 1988; 60:2493.
Munoz de la Pena A, Duran-Meras I, Duran MS. Analyst 1990; 115: 1007.
Miller JM. Analyst 1984; 109, 191.
Inman EL, Winefordner JD. Anal Chim Acta 1982; 138, 245.
Warner IM, Patonay G, Thomas MP. Anal Chem 1985; 57, 463A.
Ndou TT, Warner IM. Chem Rev 1991; 91:493.
Demas JN. Time-Resolved and Phase-Resolved Emission Spectroscopy, chapter 2
in Molecular Luminescence Spectroscopy, volume 2, Schulman SG. ed., New York:
John Wiley & Sons, 1988, pp. 79 127.
Cundall RB, Dale RE. Time-Resolved Fluorescence Spectroscopy in Biochemistry
and Biology. New York: Plenum Press, 1983.
OConnor DV, Phillips D. Time-Correlated Single Photon Counting. London: Academic Press, 1984.
Reference 22, pp. 101 110.
McGown LB, Bright FV. Anal Chem 1984; 56:1400A.
Gratton E, Jameson DM, Hall RD. Ann Rev Biophys Bioeng 1984; 13:105.
McGown LB, Bright FV. Crit Rev Anal Chem 1987; 18:245.
Schmillen A, Legler R. Luminescence of Organic Substances. Berlin: SpringerVerlag, 1967.
Maeda M. Laser Dyes. Tokyo: Academic Press, 1984.
Baeyens WRG. Fluorescence and Phosphorescence of Pharmaceuticals, Molecular
Luminescence Spectroscopy. Volume 1, Schulman SG. ed., New York: J. Wiley &
Sons, Inc., 1985, pp. 29 166.
Wolfbeis OS. Fluorescence of Organic Natural Products, Molecular Luminescence
Spectroscopy. Volume 1, Schulman SG. ed., New York: J. Wiley & Sons, Inc., 1985,
pp. 167 370.
202
Brittain
92.
93.
94.
95.
96.
97.
98.
Reference [17], pp. 80 81.

Shpak MT, Sheka EF. Optics and Spectroscopy 1960; 9:29.
Kopelman R. J Phys Chem 1976; 80:2191.
Argyrakis P, Blumen A, Kopelman R, Zumofen G. J Phys Chem 1984; 88:1973.
Argyrakis P, Hooper D, Kopelman R. J Phys Chem 1983; 87:1467.
Gentry ST, Kopelman R. J Phys Chem 1984; 88:3170.
Craig DP, Hobbins PC. J Chem Soc 539 (1955); Craig DP. ibid., 2302 (1955); Craig
DP, Hobbins PC. ibid., 2309 (1955).
Bree A, Lyons LE. J Chem Soc 2658 (1956); Bree A, Lyons LE. ibid., 2662 (1956);
Lyons LE, Morris GC. ibid., 1551 (1959); Bree A, Lyons LE. ibid., 5206 (1960).
Lacey AR, Lyons LE. J Chem Soc 5393 (1964).
Akon CD, Craig DP. Trans Far Soc 1966; 62:1673.
Taliani C, Bree A. J Phys Chem 1984; 88:2351.
Taen S. J Phys Chem 1988; 92:107.
Ikeyama T, Azumi T. J Phys Chem 1985; 89:5332.
Bonno B, Laporte JL, Rousset Y. J Luminescence 1983; 28:443.
Grahm DJ. J Phys Chem 1985; 89:5330.
Singham SB, Pratt DW. J Phys Chem 1982; 86:507.
Wakayama NI. J Luminescence 1982; 27:299.
Becker H-D, Sandros K, Skelton BW, White AH. J Phys Chem 1981; 85:2923.
Brittain HG, Ranadive SA, Serajuddin ATM. Pharm Res 1995; 12:556.
Brittain HG. AAPS PharmSciTech 2005; 5. In press.
Brittain HG, Elder BJ, Isbester PK, Salerno AH. Pharm Res 2005; 22:999.
Brittain HG. J Pharm Sci 2004; 93:375.
Hurtubise RJ. Solid Surface Luminescence Analysis. New York: Marcel Dekker, 1981.
Vo-Dinh, T. Room Temperature Phosphorimetry for Chemical Analysis. New York:
Hurtubise RJ. Luminescence from Solid Surfaces, chapter 1 in Molecular Luminescence Spectroscopy. volume 2, Schulman SG. ed., New York: John Wiley &
Sons, 1988, pp. 1 77.
Lendvay E. J Phys Chem 1965; 69:738.
Klemnitz K, Murao T, Yamazaki I, Nakashima N, Yoshihara K. Chem Phys Lett
1983; 101:337.
Klemnitz K, Tamai N, Yamazaki I, Nakashima N, Yoshihara K. J Phys Chem 1987;
91:1423.
Roberts ER, Drickamer HG. J Phys Chem 1985; 89:3092.
Clark FT, Drickamer HG. J Chem Phys 1984; 81:1024.
Anfinrud P, Crackel RL, Struve WS. J Phys Chem 1984; 88:5873.
Anfinrud P, Hart DE, Hedstrom JF, Struve WS. J Phys Chem 1986; 90:3116.
Shirke S, Takhistov P, Ludescher RD. J Phys Chem B 2005; 109:16119.
Kalyanasundaram K, Thomas JK. J Am Chem Soc 1977; 99:2039.
Hara K, de Mayo P, Ware WR, Weedon AC, Wong GS, Wu KC. Chem Phys Lett
1980; 69:105.
Bauer RK, de Mayo P, Ware WR, Wu KC. J Phys Chem 1982; 86:3781.
Levitz P, van Damme H, Keravis D. J Phys Chem 1984; 88:2228.
99.
100.
101.
102.
103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
114.
115.
116.
117.
118.
119.
120.
121.
122.
123.
124.
125.
126.
127.
128.
129.
130.
203
131. Avnir D, Russe R, Ottolenghi M, Wellner E, Zachariasse KA. J Phys Chem 1985;
89:3521.
132. Carr JW, Harris JM. Anal Chem 1986; 58:626.
133. Thomas JK. Chemistry of Excitation at Interfaces. Washington, DC: ACS Monograph 191, American Chemical Society, 1983.
134. Fendler JH. Membrane Mimetic Chemistry. New York: John Wiley & Sons, 1982.
135. Thomas JK. J Phys Chem 1987; 91:267.
136. Vo-Dinh T, Winefordner JD, Appl Spect Rev 1977; 13:261.
137. Hurtubise RJ. Anal Chem 1989; 61:889A.
138. Gunshefski M, Santana JJ, Stephenson J, Winefordner JD. Appl Spect Rev 1992;
27:143.
139. Schulman EM, Walling C. J Phys Chem 1973; 77:902.
140. Schulman EM, Parker RT. J Phys Chem 1977; 81:1932.
141. Ramasamy SM, Hurtubise RJ. Anal Chem 1987; 59:432.
142. Hurtubise RJ, Ackerman AH, Smith BW. Appl Spectrosc. 2001; 55:490.
143. White W, Seybold PG. J Phys Chem 1977; 81:2035.
144. Scharf G, Winefordner JD. Talanta 1986; 33:17.
145. Honnen W, Krabicher G, Uhl S, Oelkrug D. J Phys Chem 1983; 87:4872.
146. Ramasamy SM, Senthilnathan VP, Hurtubise RJ. Anal Chem 1986; 58:612.
147. Hurtubise RJ. Anal Chim Acta 1997; 351, 1.
148. Suter GW, Kallir AJ, Wild UP, T Vo-Dinh. J Phys Chem 1986; 90:4941.
149. Andino M, Aaron JJ, Winefordner JD. Talanta 1986; 33:27.
150. Fidanza J, Aaron JJ. Talanta 1986; 33:215.
151. Citta LA, Hurtubise RJ. Microchem J 1986; 34:56.
152. Ramasamy SM, Hurtubise RJ. Anal Chem 1987; 59:2144.
153. Ramasamy SM, Hurtubise RJ. Anal Chem 1990: 62:1060.
154. Burrell GJ, Hurtubise RJ. Anal Chem 1987; 59:965.
155. Bello J, Hurtubise RJ. Appl Spect 1986; 40:790.
156. Richmond MD, Hurtubise RJ. Anal Chem 1989; 61:2643.
157. Alak AM, Contolini N, Vo-Dinh T. Anal Chim Acta 1989; 217:171.
158. Ramasamy SM, Hurtubise RJ. Appl Spect 1996; 50:115.
159. Wang J, Hurtubise RJ. Appl Spect 1996; 50:53.
160. Mendonsa SD, Hurtubise RJ. Appl Spect 2000; 54:456.
161. Mendonsa SD, Hurtubise RJ. J Lumin. 2002; 97:19.
162. Bello LM, Hurtubise RJ. Anal Chem 1988; 60:1285.
163. Vo-Dinh T, Miller GH, Abbott DW, Moody RL, Ma CY, Ho C-H. Anal Chim Acta
1985; 175:181.
164. Asafu-Ajaye EB, Su SY. Anal Chem 1986; 58:539.
165. Smith BW, Hurtubise RJ. Appl Spectrosc, 2000; 54:1357.
166. Bower ELY, Winefordner JD. Anal Chim Acta 1978; 101:319.
167. MCG Alvarez-Coque, Ramos GR, OReilly AM, Winefordner JD. Anal Chim Acta
1988; 204:247.
168. Khasawneh IM, Alvarez-Coque MCG, Ramos GR, Winefordner JD. J Pharm
Biomed Anal 1989; 7:29.
169. Perry LM, Shao EY, Winefordner JD, Talanta 1989; 36:1037.
170. Campiglia AD, Iaserna JJ, Berthod A, Winefordner JD. Anal Chim Acta 1991;
244:215.
204
171.
172.
173.
174.
175.
176.
177.
178.
179.
180.
181.
182.
Brittain
Andino MM, Winefordner JD. J Pharm Biomed Anal, 1986; 4:317.
Tjioe SW, Hurtubise RJ. Appl Spectrosc 1998; 52:414.
Li M, Hurtubise RJ, Weston A. Anal Chem 1999; 71:4679.
Thompson AJ, Hurtubise RJ. Appl Spectrosc 2005; 59:126.
Tjioe SW, Hurtubise RJ. Talanta 1995; 42, 59.
Corley J, Hurtubise RJ. Anal Chem 1993; 65:2601.
Richmond MD, Hurtubise RJ. Anal Chim Acta 1991; 255:335.
Chu Y, Hurtubise RJ. Appl Spectrosc 1996; 50:476.
Smith BW, Hurtubise RJ. Appl Spectrosc 2003; 57:943.
Smith BW, Hurtubise RJ. Anal Chim Acta 2004; 502:149.
Burrell GJ, Hurtubise RJ. Anal Chem 1988; 60:564.
Richmond MD, Hurtubise RJ. Anal Chem 1991; 63:169.
PART IV. VIBRATIONAL LEVEL SPECTROSCOPY
7
Molecular Motion and Vibrational
Spectroscopy
Harry G. Brittain
INTRODUCTION
The energies associated with the fundamental vibrational modes of a chemical
compound lie within the range of 4004000 cm21, a spectral region corresponding
to midinfrared (mid-IR) electromagnetic radiation. Transitions among vibrational
energy levels can be observed directly through their absorbance in the infrared
(IR) region of the spectrum or through inelastic scattering of incident energy via
the Raman effect. Overtones and combination bands of vibrational modes are
observed in the near-infrared (NIR) region of the spectrum (4000 to 13,350 cm21).
IR absorption spectroscopy, especially when measured by means of the
Fourier transform method (FTIR), is a powerful technique for the physical characterization of pharmaceutical solids. When the structural characteristics of a solid
perturb the pattern of vibrational motion for a given molecule, one can use these
alterations as a means to study the solid-state chemistry of the system. FTIR
spectra are often used to evaluate the type of polymorphism existing in a drug substance and can be very useful in studies of water contained within a hydrate species.
The vibrational energy levels of a molecule in its solid form can also be
studied using Raman spectroscopy. Because most compounds of pharmaceutical
interest are characterized by low molecular symmetry, the same bands observed
in the IR absorption spectrum will also be observed in the Raman spectrum.
However, the fundamentally different nature of the selection rules associated with
the Raman effect leads to the observation of significant differences in
peak intensity between the two methods. In general, symmetric vibrations and
205
206
Brittain
nonpolar groups yield the most intense Raman scattering bands, whereas antisymmetric vibrations and polar groups yield the most intense IR absorption bands.
Spectral transitions that are observed in the NIR region of the spectrum are
associated with overtone and combinations of the fundamental bands detected in
IR absorption or Raman spectroscopies. Some of the spectral features that have
found the greatest utility are those functional groups that contain unique hydrogen atoms. For example, studies of water in solids can be easily performed
through systematic characterization of the characteristic 22OH band, usually
observed around 5170 cm21. The determination of hydrate species in an anhydrous matrix can easily be performed using NIR analysis.
In this chapter, an overview of the theoretical foundations for all three types
of vibrational spectroscopy will be presented. It goes without saying that interested readers can considerably find more complexity and details in the literature,
with the two books by Gerhard Herzberg forming the best possible references to
consult (1,2). Support for the discussed quantum mechanical presentations are
available in references (3 8), and more specific details of molecular spectroscopy can be found in references (9 13). Finally, a listing of significant
books covering vibrational spectroscopy is located in references (14 24).
MOTION OF NUCLEI IN MOLECULES
Certain types of molecular spectroscopy have their origins in the patterns of
motion executed by the atoms making up the molecule, and it is intuitive that
the energies associated with these motions are the key to a theoretical interpretation of the observed spectra. It is therefore important to consider the nature of
these motions and the manner in which energy can be distributed among the
various types of motion. In order to simplify this discussion, the perturbing
effects of other molecules in a condensed state will not be considered, but
instead we will develop the state of a molecule that is essentially free to move
through space. The perturbing effects of the solid state on the properties of the
individual molecule will be developed in the succeeding chapters.
At the outset, it is essential to define a useful term in this connection,
namely the degree of freedom. Essentially, the number of degrees of freedom
that any body night possess is related to the number of independent coordinates
that are needed to specify its position in space. For instance, a sphere that is able
to move randomly in space would require the specification of all three of the
Cartesian coordinates (x, y, and z) to define its position. It can therefore be
stated that this free object has three degrees of freedom. However, if the object
was constrained to move only in the xy-plane, then the value of the z-coordinate
would be fixed and only the x and y coordinates would have to be specified during
the motion. In that case, the object has only two degrees of freedom. If it is further
required that the object moves along a line in this plane, then only the coordinate
along which the object moved (i.e., one degree of freedom) needs to be specified.
If one now considers a molecule consisting of N atoms, it is clear that one
must specify the x, y, and z coordinates for each atom, and therefore the total
Molecular Motion and Vibrational Spectroscopy
207
number of degrees of freedom possible for this molecule would be equal to 3N. At
the same time, the atoms in the molecule are bound, and this bonding will affect
how the 3N degrees of freedom would be distributed among the different types of
motion that the molecule can perform.
For a single particle, only one type of motion is possible, which corresponds
to its translation through space. For unrestrained motion, this situation is completely described when all three coordinates of the particle are specified and the
object has three degrees of translational freedom. In a molecule, however, each
atom is not free to move independently of the others, but instead the molecule
translates as a whole. Hence, the translational motion of the entire molecule
can be described in terms of the three coordinates of the center of mass,
namely the point in the molecule where all the mass can be considered to be concentrated. The translational motions for a nonlinear triatomic molecule are shown
in Figure 1. It can be concluded from this discussion that both monatomic and
polyatomic molecules possess three translational degrees of freedom.
For polyatomic molecules consisting of N atoms, there are still (3N 2 3)
degrees of freedom remaining to be specified. These are termed internal modes
of motion, which consist of rotational and vibrational motions. The rotational
motions may be most easily visualized by considering the atoms to be held
fixed by rigid bonds so that the molecule as a whole rotates about axes through
its center of mass. As shown in Figure 2 for a simple diatomic, a linear molecule
can rotate about two mutually perpendicular axes through the center of mass.
Motion about the molecular axis will not contribute to the rotational energy of
the molecule, so only two angles of rotation are required to describe the rotational motion of a linear molecule. Therefore, linear molecules possess two
rotational degrees of freedom. On the other hand, a nonlinear polyatomic
molecule can rotate about three mutually perpendicular axes through its center
of mass (Fig. 3). Consequently, three angles of rotation must be specified, and
non-linear molecules, therefore, possess three degrees of rotational freedom.
Now 3N is the total number of degrees of freedom for a molecule with N
atoms, of which three are translational and two and three for linear and nonlinear
molecules, respectively, are rotational. For a linear molecule the remaining
number of degrees of freedom must equal (3N 2 5), and for a nonlinear molecule
the remaining number of degrees of freedom must equal (3N 2 6). These remaining degrees of freedom must be due to another form of internal motion and is
known as vibrational motion.
For a simple diatomic molecule, the number of degrees of freedom is given
by [3(2) 2 5], or 1. It is easy to see that the only nontranslational or nonrotational type of motion accessible to a diatomic molecule is the simple stretching of the bond connecting the two atoms. This type of molecular motion is often
termed as stretching mode of vibration and consists of a periodic lengthening and
shortening of the distance between the two atoms about some equilibrium position. Because of this motion, the diatomic molecule is seen to possess both
vibrational kinetic and potential energies, and therefore the vibrational modes
of motion will have two energy terms associated with it.
208
Figure 1
Brittain
Translational motions of the center of mass for a nonlinear triatomic molecule.
As discussed earlier, a linear polyatomic molecule consisting of N atoms

will be characterized by (3N 2 5) nontranslational and nonrotational vibrational
motions. For a linear triatomic molecule, the number of authentic vibrations
will therefore be equal [3(3) 2 5], for four vibrational normal modes of
motion. These are depicted in Figure 4, where the atoms are shown in their equilibrium positions and the arrows indicate the directions of motion of the atoms
away from these positions for one phase of the vibration. The four vibrational
modes of a linear triatomic molecule are termed as the symmetric stretching
motion, the antisymmetric stretching motion, and two bending motions. The
symmetric stretching mode involves atomic movement where the bonds stretch
Figure 2
209
Rotational motion in a diatomic molecule about its center of mass.
symmetrically from the equilibrium position. One phase of the anti-symmetric

stretching vibration involves the left hand and center atoms moving together and
the right-hand atom moving away from the center atom, effecting a compression
of the bond on the left and a stretching of the bond on the right. Finally, one
bending vibration arises from a slight displacement of the end atoms above (or
below) the bond axis, while the central atom moves simultaneously below (or
above) this axis. The second bending vibration (the fourth vibration for this
molecule) is obtained from the first by rotating the molecule by 908 about the molecular axis. The result is that the end atoms move into (or out of) the plane of the
paper while the central atom moves out of (or into) this plane.
On the other hand, a nonlinear polyatomic molecule consisting of N
atoms will be characterized by (3N 2 6) nontranslational and nonrotational
vibrational motions. For a nonlinear triatomic molecule, the number of authentic
vibrations will therefore be equal [3(3) 6] for three vibrational normal modes of
motion, which are depicted in Figure 5. To preserve the pure vibrational nature of
the symmetric stretching motion, the central atom must move an equal amount in
the direction opposite to the motion of the end atoms, so that the net result of the
simultaneous motions of the end and central atoms is a symmetric stretching of
210
Figure 3
Brittain
Rotational motion in a non-linear triatomic molecule about its center of mass.
the bonds. The atomic motions for the asymmetric stretching vibration gives rise
to a compression of one bond and a stretching of the other bond, whereas the
atomic motions for the bending vibration yield an opening and closing of the
bond angle.
ROTATION AND VIBRATION IN DIATOMIC MOLECULES

In order to develop a more rigorous treatment of the wave functions associated with
rotational and vibrational motion, the application of wave mechanics to diatomic
molecules will be considered. In the approximation of Born and Oppenheimer,
it is recognized that the time frame for nuclear motion is very slow when
compared with that of electronic motion, and hence the two may be separated
Figure 4
Vibrational motion in a linear triatomic molecule.
Figure 5
Vibrational motion in a nonlinear triatomic molecule.
211
212
Brittain
from each other. In other words, the molecule will remain in the same electronic
state during the entire lifetime of nuclear motion. As a result, one can develop the
wave mechanical expressions for nuclear motion independent of the expressions
for electronic motion.
For a diatomic molecule, the Hamiltonian operator for the nuclear motion
is given by the sum of the kinetic and the potential energy operators. One can then
write the Schrodinger equation for nuclear motion as:
TN V(r)c Ec
(1)
where TN is the kinetic energy operator, V(r) is the potential energy function, and
it is understood that c is the wave function for nuclear motion. For the diatomic
molecule A-B, Equation (1) becomes:
1
1
rA 2 c
rB 2 c 2E V(r)c 0
mA
mB
(2)
where mA and mB are the masses of the two atoms comprising the molecule.
To eliminate terms associated with translation of the nuclei, the motion of
the atoms will be described in terms of a set of axes that move with the molecule.
In that case, the only forces to be considered are due to the effect of the potential
field that is directed along the internuclear axis. Mathematically, this situation is
equivalent to that of a single particle moving in the field defined by V(r), where
the mass of this equivalent particle is given by the reduced mass (m) of the two
atoms:
mA mB
mA mB
(3)
The wave equation now becomes:

2m
r2 c 2 E V(r)c 0
h
(4)
which is analogous to Equation (5) of chapter 2, except that the form of the potential energy function has not yet been specified. As previously, after the wave
function is transformed into spherical polar coordinates, it can be factored into
functions of a single variable:
c (r, u, f) R(r)Q(u)F(f)
(5)
Once the Schrodinger equation is solved in the manner described in the chapter,
the resulting wave function will have the form:
c (r, u, f) R(r)Y(u, f)
(6)
213
where the Y(u, f) aspect are the surface harmonics defined in Equation (23) of
chapter 2. The radial functions are solutions of the equation:

1 d 2 dR
2m
C
r
2 E V(r) 2 R 0
r 2 dr
dr
r
h
(7)
where C equals the constant of chapter 2, Equation (13). If we now substitute:

C J(J 1)
S(r) rR(r)
(8)
(9)
where S(r) is a radial distribution function. The meaning of J will be made clear
shortly. Equation (7) now becomes:

d2 S 2m
J(J 1)h 2
E V(r)
S0
dr 2 h 2
2mr 2
(10)
Equation (10) describes two types of nuclear motion. One of these is termed
vibrational motion, because the wave functions that are a function of r describe
the motion of the nuclei directed along the axis of the diatomic molecule. In other
words, Equation (10) is the wave equation for one-dimensional motion, where the
potential field is given by:
V(r)
J(J 1)h 2
2mr 2
(11)
The second term in Equation (10) accounts for the centrifugal potential due to
rotation of the molecule with respect to the space fixed coordinate axes.
The solution to Equation (10) yields a set of vibrational wave functions
and their energies, each of which is distinguished by a quantum number, v.
The vibrational wave functions are further defined by the permissible values of
J, the quantum number for the angular momentum of the system.
VIBRATIONAL ENERGY LEVELS AND SPECTROSCOPY OF
DIATOMIC MOLECULES
In order to understand the quantization of vibrational energy, it is useful
to consider the model where the molecule is not permitted to undergo rotational
motion. More simply stated, one merely sets J 0 in Equation (10) to obtain:
d2 S 2m
E V(r)S 0
dr 2 h 2
(12)
Equation (12) cannot be considered further until one chooses a form for the
potential function, V(r).
214
Brittain
The Harmonic Oscillator Model

In the simplest model, V(r) is assumed to have the form associated with simple
classical harmonic motion:
V(r) 12 kr 2
(13)
where k is the force constant associated with the magnitude of the restoring
force on the particle.
In the classical theory, the vibrational frequency (y OSC) of the harmonic
oscillator is:
yOSC (1=2p)(k=m)1=2
(14)
The solution to Equation (12) is simplified if three new variables are defined:
a 2mE=h 2
b (mk)
1=2
(15)
=h
g r=(b)1=2
After these substitutions are made, Equation (12) becomes:

d2 S
a
2

g
S0
dg2
b
(16)
(17)
(18)
One approach to a solution for Equation (18) is to see how the equation behaves
as very large values of g, namely g 1. In that case, g a/b, and the solution
becomes:
S(g) C exp (+g2 =2)
(19)
where C is a constant. In fact, only the equation having the negative exponent is
acceptable, because the positive exponent solution has the unacceptable result of
S ! 1 as g ! 1. Therefore, at large values of g, the wave equation has the
form:
S(g) C exp ( g2 =2)
(20)
To obtain wave functions that are acceptable at all values of g and that yield
Equation (20) at large values of g, it is useful to introduce another function U(g),
which is a correction function. In that case:
S(g) CU(g) exp ( g2 =2)
(21)
The most convenient form of U(g) is that of a power series in g that approaches
unity as g ! 1. If one substitutes Equation (21) into Equation (18), then:

d2 U
dU
a
2g
1 U0
(22)
dg2
dg
b
215
It happens that Equation (22) is in the form of a well-known standard equation,

namely that of Hermite. The solutions to the equation have the form of a complex
power series, but recognizing that x g and 2V [(a/b) 2 1], then the equation
takes the general form:
d2 Y
dY
2VY 0
2x
dx2
dx
(23)
The solutions to Equation (23) are known as the Hermite polynomials, HV (x),
and are given by the following equation as long as V 0, or if V is a positive
number:
V(V 1)(2x)V2
1!
V(V 1)(V 2)(V 3)(2x)V4

2!
HV (x) (2x)V
(24)
The relationship between Equations (22) and (23) is apparent if:

Y U(g)
Xg
(25)
(26)
2V (a=b) 1
(27)
The Hermite polynomials provide solutions to Equation (22):

U(g) C0 HV (g)
(28)
where C0 is another constant, and the HV(g) functions are defined by a vibrational
quantum number, namely:
v 0, 1, 2, 3, . . .
(29)
The vibrational eigenfunctions of the Schrodinger equation are in turn defined by

their characteristic v quantum number and have the general form:
SV NV HV (g) exp ( g2 =2)
(30)
where NV is the normalization factor.

The lowest energy vibrational wave functions have the form:
S0 (a=p)1=4 exp ( ag2 =2)
3
S1 (4a =p)
S2 (a=4p)
3
1=4
1=4
S3 (9a =p)
(31)
g exp ( ag =2)
2
(32)
2
(1 2ag ) exp ( ag =2)
1=4
2=3
(g
ag ) exp ( ag =2)
(33)
(34)
216
Brittain
Spectroscopy of a Harmonically Oscillating Diatomic Molecule

If the definitions of a from Equation (15) and b from Equation (16) are substituted into Equation (27), then the energies (EV) of the vibrational wave functions are functions of the reduced mass and the bond force constant:
EV h (k=m)1=2 (v 12)
(35)
Using the definition of Equation (14), the energies (in units of ergs) are also
equal to:
EV h nOSC (v 12)
(36)
Alternatively, one can express the energy of a given vibrational state in units of
reciprocal centimeters (cm21) by dividing Equation (36) by the speed of light:
GV
h nOSC
(v 12)
c
(37)
It is conventional to define the oscillator energy as v (in units of cm21), and in

that case:
GV v(v 12)
(38)
For most molecules, v will have values in the range of 200 4000 cm21.
Because a homonuclear diatomic molecule has no permanent dipole
moment, it cannot have an IR absorption spectrum. The reason for this is that
the electric dipole spectral transition probability, Pij, from a lower vibrational
state (Si) to an upper vibrational state (Sj) must equal zero:
Pij Si mS j dt
(39)
For heteronuclear diatomic molecules that have a permanent dipole
moment, it can be shown that Equation (39) will equal zero unless the following
selection rule is met:
Dv +1
(40)
The energetics of vibrational energy levels is such that the ground vibrational
state of a given molecule will be characterized by v 0. The implication of
the harmonic oscillator selection rule is that the only allowed vibrational
transition would be S0 ! S1, which is known as the fundamental vibrational
transition. In the harmonic oscillator model, transitions from S0 to higher
excited states SV are forbidden. It will be shown in a later section that the harmonic oscillator selection rule of Dv +1 can be relaxed by anharmonic effects.
217
In the harmonic oscillator model, therefore, the energy of the E(v 0) state
is given by:
E0
h
(k=m)1=2
2
(41)
and the energy of the E(v 1) state is given by:

E1
3h
(k=m)1=2
2
(42)
Therefore, energy of the (v 0) ! (v 1) transition is given by:

E1 E0 h (k=m)1=2
(43)
G1 G0 v
(44)
or
A considerable insight into vibrational spectroscopy can be gleaned

through a consideration of the equations derived previously. According to
Equation (43), the energy of a vibrational transition may be seen to be partially
defined by the magnitude of the force constant (which is interpreted as a measure
of the bond strength). In particular, it can be shown that one can calculate a value
for the force constant of a harmonic oscillator using the relation:
k m(4p2 v2 c2 )
(45)
where the reduced mass m is give in units of grams/molecule, v is in units of

cm21, and the speed of light c equals 2.9979246 1010 cm/sec. Wilson,
Decius, and Cross have compiled a listing of force constants for various bond
types (24).
The series of diatomic hydrogen halides constitutes a good illustration of
the effect of force constant on the fundamental stretching frequency of a diatomic
molecule. The reduced masses of these molecules are dominated by the hydrogen
atom, being and equal to 0.957159, 0.979726, 0.995384, and 0.999998 atomic
units for HF, HCl, HBr, and HI, respectively. From the extensive compilation
of measured vibrational frequencies of diatomic molecules published by
Herzberg (2), it is known that the energies of the fundamental vibrational transitions decrease in the order of 4138.52 cm21 (HF), 2989.74 cm21 (HCl),
2649.47 cm21 (HBr), and 2309.5 cm21 (HI). Using Equation (45), the calculated
harmonic oscillator force constants are found to be 9.659 105 dynes/cm for
HF, 5.160 105 dynes/cm for HCl, 4.117 105 dynes/cm for HBr, and
3.143 105 dynes/cm for HI. These calculated bond force constants correlate
well with the known bond energies (25) for these molecules, which is illustrated
in Figure 6.
As discussed previously, the energy of a vibrational transition is also critically dependent on the reduced mass of the atoms involved in the vibrational
218
Brittain
motion. This can be illustrated by various interhalogen diatomic molecules for

which the force constants would be fairly similar but which would differ significantly in the respective reduced masses. The reduced masses of ClBr, ClI, and BrI
are 24.3389, 27.4338, and 49.0317 atomic units, respectively, and the energies of
the respective vibrational absorption bands are 430.0, 384.2, and 268.4 cm21.
The harmonic oscillator force constants are roughly equivalent, being calculated
as 2.65 105 dynes/cm for ClBr, 2.39 105 dynes/cm for ClI, and
2.08 105 dynes/cm for BrI.
The diatomic molecule harmonic oscillator has one additional lesson to
teach regarding the effect of reduced mass and force constant on the energies
of vibrational transitions. Using the average bond stretching force constants of
Wilson, Decius, and Cross (24), and the relation:
(k=m)1=2
2pc
(46)
it becomes possible to calculate the transition energies for the fundamental

vibrational transitions of some chemically relevant bond types. For instance, the
Figure 6 Correlation of harmonic oscillator force constants with bond energies for the
series of hydrogen halide molecules.
219
energy of the C22H stretching mode is calculated to be 3444 cm21, and the energy of
the C22C stretching mode is calculated to be 1189 cm21. In addition, the energy of
the N22H stretching mode is calculated to be 3425 cm21, and the energy of the N22N
stretching mode is calculated to be 1044 cm21. Finally, the energy of the O22H
stretching mode is calculated to be 3737 cm21, and the energy of the C22C stretching
mode is calculated to be 950 cm21. From this analysis, the general trend emerges that
the stretching frequencies of molecules containing a hydrogen atom will be much
higher than the stretching frequencies of molecules that do not contain a hydrogen
atom. The spectroscopic consequence of this trend (which will be developed
further in forthcoming sections) is that vibrational transitions associated with
atomhydrogen stretching modes will be observed at much higher energies than
the transitions of other functional groups.
VIBRATIONAL ENERGY LEVELS AND SPECTROSCOPY OF THE
FUNDAMENTAL TRANSITIONS OF POLYATOMIC MOLECULES
Even for polyatomic molecules, the harmonic oscillator model can be used with a
fair degree of adequacy to describe their fundamental vibrational transitions. In
fact, one may gain considerable insight into the vibrational spectra of complicated models by considering them as summations of a number of harmonic oscillators. In addition, many vibrational wave functions turn out to be essentially
localized on the atoms of a functional group and its nearest neighbors, a fact
that will ultimately lead to the development of the concept of group frequencies.
Vibrational Modes and Wave Functions of Polyatomic Molecules
As discussed earlier, the number of remaining degrees of freedom is not limited
for a nonlinear polyatomic molecule, and therefore the number of possible
vibrational modes will equal 3N 2 6. For instance, a nonlinear triatomic molecule will exhibit [3(3) 2 6] genuine modes of vibrational motion, of three
vibrational modes. These motions are illustrated in Figure 5, but now, one can
identify the symmetric stretching mode as v1, the scissors bending mode as
v2, and the antisymmetric stretching mode as v3.
Because the individual vibrational modes are independent of each other,
the vibrational wave function describing the states of this nonlinear triatomic
molecule will be:
c (n1 , n2 , n3 ) c(n1 )c(n2 )c(n3 )
(47)
The ground state of this non-linear triatomic molecule will be characterized by

c(n1), c(n2), and c(n3) each being in their respective v 0 states, and its wave
function will be abbreviated as c000. The lowest energy excited states will be
characterized by two of the c(n1), c(n2), or c(n3) functions each being in
the v 0 state and the third function being in the v 1 state. These three possible
excited states would be identified by the functions c100, c010, and c001.
220
Brittain
The effects of force constant and reduced mass on the vibrational frequencies of the constituent groups of atoms in a molecule leads to certain generalizations about the vibrational spectroscopy of polyatomic molecules. Because
the force constants of most bonds are fairly constant from molecule to molecule
and because the reduced masses are often comparable as well, it follows that the
vibrational frequency of a given functional group is primarily determined by the
identity of the atoms involved. This concept has led to the concept of group frequencies, and this understanding has greatly facilitated the interpretation of
vibrational spectra.
The vibrations associated with atoms in a molecule linked by covalent
bonds are classified as being due to either stretching or deformation (i.e.,
bending) modes of molecular motion. The force constants associated with the
deformation modes are usually weaker than those of stretching modes by
approximately one order of magnitude and therefore consistently observed at
lower energies. The weaker nature of the deformation modes also makes these
more sensitive to details of chemical bonding, a fact that has proven to be very
useful to vibrational spectroscopists over the years. For complex molecules,
the same atom may be involved with both stretching and deformational modes
of motion.
The constancy of force constants (and therefore of group frequencies) is
valid only for similar chemical environments, and therefore one observes that
changes in molecular bonding can yield shifts in the vibrational frequency of a
given functional group. As a result, the published correlation tables usually
present a range of frequencies as being characteristic of a particular functional
group. For instance, the bending mode of C22H is observed around
1400 cm21, but nevertheless exhibits variability depending on the remainder of
the molecule (18). In the 22O22CO22CH3 fragment, the frequency of the
C22H bending mode will be in the range of 1380 1365 cm21. In the
22CO22CH3 fragment, the frequency of the C22H bending mode decreases
from 1360 1355 cm21, and in the 22COO22CH3 fragment, the frequency of
the C22H bending mode increases from 1440 to 1435 cm21.
Practically, every book covering vibrational spectroscopy features some
type of summary regarding the frequencies associated with various functional
groups, but the most complete compilation of characteristic group frequencies
may be found in monographs developed especially for this purpose (26 28).
Electric Dipole Absorption Spectroscopy of the Fundamental
Modes of Polyatomic Molecules
Although one can readily determine that the nonlinear triatomic molecule mentioned previously must exhibit three genuine modes of vibrational motions, it
remains to be determined whether transitions can take place within the states
associated with these vibrational modes. Because the harmonic oscillator selection rule is still Dv +1, it follows that the only possible fundamental
221
vibrational spectroscopic transitions would correspond to:
c000 ! c100
c000 ! c010
c000 ! c001
(48)
(49)
(50)
As previously discussed in the case of electric dipole transitions, the value

of the transition moment:
Pij ci mc j dt
(51)
for an allowed transition cannot equal zero. Fortunately, one can determine
whether or not a given transition equals zero or (or not) solely on the basis of
molecular symmetry-based group theory. This is because the integration step
necessary to evaluate Pij is to be performed over all space, and it follows that
any antisymmetric result must integrate to zero.
In practice, one uses the symmetry properties (i.e., its representation) of ci,
m, and cj to deduce whether Pij does or does not equal zero. As discussed in
chapter 4, the ground state of a molecule (denoted as c000 for the nonlinear
triatomic molecule of the illustration) must transform as the total symmetric representation in its point group. To achieve a symmetric end result, the representation of the excited state wave function must be the same as that of the
representation for one of the elements of the dipole moment operator (29 32).
For the nonlinear triatomic molecule, the n1 symmetric stretching mode
transforms according to the A1 representation, the n2 bending mode also as the
A1 representation, and the n3 antisymmetric stretching mode as the B1 representation. Examination of the character table for the C2V point group (Fig. 3 of chap.
4) reveals that the z-component of the dipole moment operator transforms as the
A1 representation and that the x-component of the dipole moment operator transforms as the B1 representation. Consequently, it can be concluded that all three
vibrational modes of a nonlinear triatomic molecule will exhibit absorbance in
the IR region of the spectrum.
For water (a typical nonlinear triatomic molecule), the energies of the
vibrational spectroscopic transitions in the gas phase are:
c000 ! c100
(n1 ) 3652 cm1
(52)
c000 ! c010
(n2 ) 3756 cm
1
(53)
(n3 ) 1545 cm
1
(54)
c000 ! c001
As per the spectrum of bulk water shown in Figure 7, these energies are substantially shifted owing to the extensive degree of hydrogen bonding.
Analysis of the vibrational spectrum of a polyatomic molecule becomes
progressively more complicated as the number of atoms increases. As another
example, we will consider ethylene, C2H4, a molecule having D2h symmetry and
222
Brittain
Figure 7 Infrared absorption spectrum of bulk water (the intensity scale is in KubelkaMunk units).
whose character table is found in Figure 4 of chapter 4. This molecule contains a

total of six atoms, and so the number of genuine molecular vibrational modes
will be given by [3(6) 2 6], or 12 modes of vibration. Each vibrational mode
must transform according to one of the irreducible representations of the D2h
point group as summarized in Table 1. Because the three elements of the dipole
moment operator transform according to the B1u, B2u, and B3u irreducible representations, it follows that only those vibrational modes that transform according to one
of these representations will exhibit allowed transitions in the IR spectrum. As a
result, only five out of 12 vibrational modes will exhibit an IR absorption band.
For very complicated organic molecules, there is usually a sufficient lack of
molecular symmetry that all possible vibrational modes will yield a fundamental
absorption band in the IR spectrum. However, the intensity of the resulting bands
will still be determined by the magnitude of the transition moment, and not all
will exhibit strong degrees of absorption. For this reason, one ordinarily includes
an intensity estimation (i.e., very strong, strong, moderate, weak, very weak) in
any listing of peak energies.
Raman Spectroscopy of the Fundamental Modes of
Polyatomic Molecules
Alternatively, the vibrational modes of a compound may be measured using
Raman spectroscopy, where one measures the inelastic scattering of radiation
by a nonabsorbing medium (19 23). In the Raman effect, when an intense
223
Table 1 Molecular Motion Descriptions and Symmetry of the Vibrational Modes of

Ethylene, a Molecule Having D2h Symmetry
n1
n2
n3
n4
n5
n6
n7
n8
n9
n10
n11
n12
Description of molecular motion
Corresponding irreducible
representation in the D2h
point group
Carbon carbon stretching mode

Carbon hydrogen stretching mode (1)
Hydrogen carbon hydrogen bending mode (1)
Hydrogen carbon hydrogen bending mode (2)
Wagging mode (1)
Wagging mode (2)
Out-of-plane bending mode (1)
Ag
Ag
B1u
B2g
B3u
Ag
B1u
B2g
B3u
Au
B2u
B3g
beam of light is passed through a medium, approximately one in every million

incident photons is scattered with a loss or gain of energy. The inelastic-scattered
radiation can occur at lower (Stokes lines) and higher (anti-Stokes lines) frequencies relative to that of the incident (or elastically scattered) light, corresponding to
the vibrational transition frequencies of the sample or of the medium. The actual
intensities of the Stokes and anti-Stokes lines are determined by the Boltzmann
factor characterizing the vibrational population. For high-frequency vibrations,
the Stokes lines are relatively intense relative to the anti-Stokes lines, so conventional Raman spectroscopy makes exclusive use of the Stokes component.
The Raman effect originates from the interaction of the oscillating-induced
polarization or dipole moment of the medium with the electric field vector of the
incident radiation, and a significant degree of understanding can be obtained by
considering the classical theory of scattering. The theory of scattering begins with
the fact that because light consists of electromagnetic radiation, the oscillating
electric field can induce a dipole moment in an irradiated molecule. This
molecule will in turn radiate light having the same frequency as the incident
radiation, but which is instead scattered in directions that are different from
that of the incident light.
The magnitude of the induced dipole moment, M, will depend both on the
intensity of the incident radiation and on the polarizability of the molecule. Thus,
one may consider the polarizability as a measure of how readily the electrons in
the molecule are displaced by the electric field of the light wave. In other words,
M aE
(55)
224
Brittain
where a is the tensor describing the polarizability molecule, and E is the intensity
of the electric field of the incident radiation.
When electromagnetic radiation interacts with a molecule, the induced
electric field is given by:
E E0 cos (2p n0 t)
(56)
where E0 is the amplitude of the electric vector of the incident light, and n0 is the
frequency of incident light interacting with the molecule. The electric moment
induced in the molecule will have the value:
M aE0 cos (2p n0 t)
(57)
Equation (57) describes Rayleigh scattering, or the situation where the frequency
of the scattered light is the same as that of the incident light.
The polarizability, a, is a scalar quantity only for isotropic systems, and the
full tensor description must be used for anisotropic molecules. For such molecules, the following equations are required to relate the nonequal polarizability
along the principal axes of the molecule:
MX aXX EX aXY EY aXZ EZ
MY aYX EX aYY EY aYZ EZ
(58)
(59)
MX aZX EX aZY EY aZZ EZ
(60)
where Mi is the component of the induced dipole moment induced along the ith
axis.
For small amplitudes of molecular vibration, the polarizability of the
molecule can be expressed in terms of vibrational coordinates by the equation:

@a
a a0
QV
(61)
@QV 0
where a0 is the molecular polarizability at the equilibrium configuration of the
molecule, and (@a/@QV)0 is the rate of change of the polarization with change
in bond length (evaluated at the equilibrium configuration). The vibrational
displacement function, QV, is a function of time because the molecule vibrates
with a fundamental frequency equal to nV:
QV Q0V cos (2p nV t)
where Q0V is the vibrational amplitude.
Substitution of Equations (61) and (62) into Equation (57) yields:

@a
0
M E0 a0 cos (2p n0 t) E0 QV
cos (2p n0 t) cos (2p nV t)
@QV 0
(62)
(63)
225
Through the use of the trigonometric identity:

1
cos (g) cos (b) cos (g b) cos (g b)
2
(64)
Equation (63) can be arranged to:

1
@a
cos2p (n0 nV )t
M E0 a0 cos (2p n0 t) E0 Q0V
2
@QV 0

1
@a
E0 Q0V
cos2p (n0 nV )t
2
@QV 0
(65)
The first term on the right-hand side of Equation (65) corresponds to Rayleigh
scattering, where the frequency of the scattered radiation is the same as that of
the incident light. The second term will give rise to weak scattering of light
whose frequency equals (n0 nV), whereas the third term gives rise to weak scattering of light whose frequency equals (n0 2 nV). The weakness of the scattering
associated with the second and third terms of Equation (65) originates in the fact
that the polarizability changes are relatively less during the process.
The scattering of electromagnetic radiation at frequencies shifted from the
incident radiation by amounts equal to vibrational frequencies is known as the
Raman effect. The family of scattering bands observed at frequencies of (n0 nV)
are known as the Stokes lines, whereas the family of scattering bands observed
at frequencies of (n0 nV) are known as the anti-Stokes lines. The classical
theory predicts that the Stokes and anti-Stokes lines should be observed with
equal probability (and hence with equal intensity), but in reality the antiStokes lines are found to be much weaker in intensity than the Stokes lines,
which are themselves much weaker than the Rayleigh scattering intensity. As
will be shown in the succeeding chapter, the quantum mechanical treatment of
the Raman effect correctly predicts the relative intensities of the Stokes and
anti-Stokes lines.
A vibrational mode will only appear in the Raman spectrum if, and only if,
the displacements in atomic position change the polarizability of the molecule.
Because this criterion is completely different from that of the electric dipole
absorption, it is possible that some vibrational modes will be seen in the
Raman spectrum and not in the IR absorption spectrum (or vice versa). When
a vibrational mode is allowed under both electric dipole and polarizability selection rules, the observed frequency will be identical. Owing to the different nature
of the selection rules, however, the intensities of corresponding bands may
exhibit considerable differences.
The intensity of a Raman transition is governed by the magnitude of its
transition moment:
Pij E ci ac j dt
(66)
226
The transition moment can be divided into its various components:
Pij (x) E ci aXX c j dt E ci aXY c j dt E ci aXZ c j dt
Pij ( y) E ci aYX c j dt E ci aYY c j dt E ci aYZ c j dt
Pij (z) E ci aZX c j dt E ci aZY c j dt E ci aZZ c j dt
Brittain
(67)
(68)
(69)
If any one of the transition moment components of Equations (67) through (69)
are nonzero, then the transition will be allowed in the Raman spectrum.
As discussed previously, one can use molecular symmetry-based group
theory to determine whether or not a given transition component will equal
zero or (or not). This process uses the symmetry properties (i.e., its representation) of ci, a, and cj to deduce whether any of the Pij components is nonzero. As before, the ground state of a molecule (which would be the state for
which all vibrational wave functions are in their v 0 level) must transform as
the totally symmetric representation in its point group. To achieve a symmetric
end result, the representation of the excited state wave function must be the
same as that of the representation for one of the elements of the polarizability
operator (29 32). As discussed in chapter 4, the elements of the polarizability
tensor will transform as the squares and binary products of coordinates listed
in area IV of the point group character table.
Returning to the example of the nonlinear triatomic molecule, it may be
recalled that the n1 symmetric stretching mode also transforms according to
the A1 representation, the n2 bending mode also transforms as the A1 representation, and the n3 antisymmetric stretching mode as the B1 representation. Examination of the character table for the C2V point group (Fig. 3 of chap. 4) reveals
that the x 2, y 2, and z 2 coordinates of the polarizability tensor transforms as the
A1 representation, and that the xy cross-product of the polarizability tensor
transforms as the B1 representation. Consequently, it can be concluded that all
three vibrational modes of a nonlinear triatomic molecule will exhibit Raman
scattering.
Turning to the previously considered D2h symmetry example of ethylene
(C2H4), examination of the character table of Figure 4 of chapter 4 indicates
that the x 2, y 2, and z 2 coordinates of the polarizability tensor transform as the
Ag representation and that the xy, xz, and yx cross-products of the polarizability
tensor transform as the B1g, B2g, and B3g representation. As shown in Table 1,
each of the genuine molecular vibrational modes transforms according to one
of the irreducible representations of the D2h point group. In order for a vibrational
mode to exhibit Raman scattering, the symmetry of the molecular motion must
transform according to one of the B1g, B2g, or B3g irreducible representations.
As a result, only six out of the 12 vibrational modes will exhibit Raman
scattering.
227
The D2h symmetry example reveals an interesting fact regarding

vibrational spectroscopy. When a molecule contains an inversion center as one
of its symmetry elements, vibrational modes will be active either in electric
dipole absorption spectroscopy or in Raman scattering but not in both types of
spectroscopy. In the case of ethylene, only the n1, n2, n4, n6, n8, and n12
vibrational modes of Table 1 will exhibit Raman scattering, and only the n3,
n5, n7, n9, and n11 vibrational modes will exhibit absorption in the IR spectrum.
Most interestingly, because the vibrational mode identified as n10 (one of the outof-plane bending modes) transforms as the Au representation, it is forbidden by
both the electric dipole and polarizability selection rules, and therefore its fundamental cannot be observed in any mode of vibrational spectroscopy.
It should be noted that the same summaries of characteristic group frequencies that are found in vibrational spectroscopy books and compilations for correlation of IR absorption bands apply equally well to the interpretation of Raman
spectra. Generally, symmetric vibrations and nonpolar groups yield the most
intense Raman scattering bands, whereas antisymmetric vibrations and polar
groups yield the most intense IR absorption bands. Thus, even though lowsymmetry organic molecules (typically those of pharmaceutical interest) will
feature all vibrational modes in both the IR absorption and Raman spectra, the
spectra often look quite different owing to the different origins of the transitions.
Nevertheless, the measured frequency of a given vibrational mode will be the
same whether it is detected through its IR absorption or through its Raman
scattering.
SPECTROSCOPY OF TRANSITIONS ASSOCIATED WITH
OVERTONES AND COMBINATIONS OF FUNDAMENTAL
VIBRATIONAL MODES
As discussed previously, the harmonic oscillator selection rule states that only
transitions for which Dv +1 will be allowed. Because at room temperature,
the majority of molecules will be in the v 0 state, this implies that only the fundamental transitions are observed in either the IR absorption or Raman spectra.
However, it should be remembered that the harmonic oscillator model pertains
only to one type of potential energy function substituted into Equation (1).
A more realistic description of the energy level sequence is achieved
through the use of potential energy functions that permit molecular distortion
in highly excited vibrational states. The use of more realistic potential energy
functions serve to weaken the Dv +1 selection rule as a result of anharmonic
effects. Three requirements are associated with more realistic potential energy
functions. The first of these is that as the magnitude of the vibration approaches
infinity, the magnitude of the potential function must approach a constant value.
This constant value is often termed the dissociation energy, De, because when
that much energy is absorbed by the system the bond must break. Second, as
the magnitude of the vibration approaches zero, the magnitude of the potential
228
Brittain
function must approach infinity. In other words, the nuclei are not allowed to
touch one another no matter what the degree of molecular motion. Finally, the
potential energy function must exhibit a minimal value at the equilibrium position of the nuclei.
The most widely used potential energy function is the Morse potential:
V(r) De 1 ea(rre) 2
(70)
The rigorous approach would then be to substitute this form for the potential
energy function back into Equation (1) and to then solve the resulting Schrodinger
equation in the usual manner. However, this solution leads to great mathematical
difficulty, and hence more empirical approaches are usually taken.
The Anharmonic Oscillator Model

A more practical approach toward vibrational spectroscopy is to expand the harmonic oscillator Equation (38) in an infinite power series based on (v 12) to
include anharmonic effects:
G0V ve (v 12 ) ve Xe (v 12)2 ve Ye (v 12)3 :
(71)
where ve is the harmonic oscillator constant, and where Xe and Ye are anharmonic constants. Above the energy of the dissociation limit, there exists a continuum
of levels that are infinitely close together in energy, and these describe the
breakup of the molecule. This dissociation energy, De, is given by the sum:
De SDG0 (v 12)
(72)
where the summation is performed over the range of v 0 to the highest

vibrational level for which the bond in the molecule is still intact.
Usually, the veXe term will be considerably larger than the veYe term, and
thus most experimentally determined vibrational spectra can be fitted to the simplified equation:
G0V ve (v 12) ve Xe (v 12)2
(73)
For most molecules, the veXe term will be positive in character, so the separation
between successively higher excited states will decrease as v increases.
0
In the simplified model, the separation between the GV0 state and the GV1
state will be given by:
DG0V ve ve Xe (2v 2)
(74)
In this simplified model of the anharmonic oscillator, the dissociation energy is

given by:
De ve =4Xe
75
229
Equation (73) indicates that progressions of experimentally observed

vibrational transitions that can be envisioned as a sum of a harmonic and an
anharmonic part. However, at ambient conditions the majority of molecules
are in the v 0 state because the thermal energy available at 258C is not sufficient to raise more than a few percent of molecules into the v 1 state. In other
words, for a transition from v 0 to v 1, the contribution from the
veXe(2v 2) term in Equation (73) is relatively minor, and consequently the
fundamental transition can be adequately understood in terms of the harmonic
oscillator model.
Overtone Vibrational Spectroscopy

Probably the most important consequence of including anharmonic effects in
vibrational energy states is that the strict harmonic oscillator selection rule of
Dv +1 becomes relaxed. This situation causes transitions from the ground
vibrational state to higher vibrationally excited states to gain some allowedness
in their character. However, the overtone transitions (i.e., where Dv 2, 3, 4,
etc.) will generally be less intense than the corresponding fundamental transition
because the overtone transitions can take place only through relaxation of the
primary selection rule. Nevertheless, overtone transitions are easily detected,
and are observed in the NIR region of the electromagnetic spectrum (33 36).
For example, in the vapor phase, HCl exhibits its fundamental
(v 0 ! v 1) absorption at an energy of 2885.9 cm21 (3465.1 nm). The
energy of the first overtone (v 0 ! v 2) transition is observed
at 5668.0 cm21 (1764.3 nm), and the energy of the second overtone
(v 0 ! v 3) transition is observed at 8347.0 cm21 (1198.0 nm).
Returning to the illustration of the nonlinear triatomic molecule, the
number of genuine vibrational modes was found to be [3(3) 2 6], or three. As discussed earlier, the individual vibrational modes are independent of each other,
and so the vibrational wave function describing the states of this nonlinear
triatomic molecule will be:
c(n1 , n2 , n3 ) c(n1 )c(n2 )c(n3 )
(76)
The ground state of this non-linear triatomic molecule will be characterized by

c(n1), c(n2), and c(n3) each being in their respective v 0 states (c000). The
lowest energy excited states will be characterized by two of the c(n1), c(n2),
or c(n3) functions each being in the v 0 states and the third function in the
v 1 state (i.e., c100, c010, or c001), and so the fundamental vibrational bands
would be c000 ! c100, c000 ! c010, and c000 ! c001.
Once the effects of anharmonicity are factored into the transition moment
calculations, almost any type of vibrational excitation becomes possible. One
230
Brittain
could observe sequences of simple overtone transitions, such as:
c000 ! c200 , c300 , c400 , etc:

c000 ! c020 , c030 , c040 , etc:
c000 ! c002 , c003 , c004 , etc:
(77)
(78)
(79)
or any one of a number of combination transitions, some of which can be

written as:
c000 ! c110 , c101 , c011 , etc:

c000 ! c120 , c201 , c021 , etc:
(80)
(81)
It should be remembered that the overtone and combination transitions are

still of the electric dipole type, and will therefore not be observed unless the value
of the transition moment
Pij ci mc j dt
(82)
does not equal zero. As before, one can use molecular symmetry-based group
theory to determine whether or not a given transition moment will equal zero.
This is because the integration step necessary to evaluate Pij is to be performed
over all space, and it follows that any antisymmetric result must integrate to zero.
It was established earlier that for a nonlinear triatomic molecule, the n1
symmetric stretching mode transforms according to the A1 representation, the n2
bending mode also transforms as the A1 representation, and the n3 antisymmetric
stretching mode as the B1 representation. The ground c000 state must transform as
the totally symmetric representation (A1). To determine the symmetries of the overtone and combination excited states, one takes the direct product of the
representations of the two excited modes and determines the representation of
the point group at which the resulting product transforms. The simplicity of the
C2V point group, and the symmetries of the fundamental vibrations, make this a relatively simple task. For instance, the c110 state will transform as the A1 representation,
the c101 state as the B1 representation, and the c011 state also as the B1 representation.
Because it has already been established that the z-component of the dipole moment
operator transforms as the A1 representation and that the x-component of the dipole
moment operator transforms as the B1 representation, it can be concluded that the
c000 ! c110, c000 ! c101, and c000 ! c011 transitions will be allowed by group
theory and would exhibit absorbance in the NIR region of the spectrum.
A summary of the overtone and combination bands of water in the
vapor phase is given in Table 2, and a portion of its liquid phase absorption
spectrum is shown in Figure 8. For bulk water, the moderate band observed at
1455 nm can be assigned to the c000 ! c101 transition, the weak band at
1806 nm can be assigned to the c000 ! c021 transition, and the strong bad
at 1936 nm can be assigned to the c000 ! c011 transition.
231
Table 2 Overtone and Combination Vibrational Transitions of Water in the

Vapor Phase
Upper state
n1
n2
n3
Energy (cm21)
Wavelength (nm)
Intensity
0
0
1
1
2
0
2
3
1
1
2
0
1
0
0
1
0
0
1
1
1
1
1
3
1
1
3
5,332.0
6,874.0
7,251.6
8,807.1
10,613.1
11,032.4
12,151.2
13,830.9
14,318.8
1875.5
1454.8
1379.0
1135.4
942.2
906.4
823.0
723.0
698.4
Medium
Weak
Medium
Strong
Strong
Medium
Medium
Weak
Weak
SUMMARY
Not only are the vibrational modes of a compound affected by the identity and
character of the rest of the atoms in the molecule, the molecular motion can
also be further perturbed by details of the solid-state structure. This feature of
vibrational spectroscopy makes the various techniques highly useful as adjunct
Figure 8 Near-infrared absorption spectrum of bulk water (the intensity scale is in

arbitrary units).
232
Brittain
methods to the standard procedures of crystallography. As will be seen in the

following chapters, through the use of appropriate sample manipulation, one
can obtain valuable information regarding molecules in their solid state.
REFERENCES
1. Herzberg G. Infrared and Raman Spectra of Polyatomic Molecules. New York, NY:
Van Nostrand Reinhold Co., 1945.
2. Herzberg G. Spectra of Diatomic Molecules, 2nd edn. New York, NY: Van Nostrand
Reinhold Co., 1950.
3. Pauling L, Wilson EB. Introduction to Quantum Mechanics. New York: McGrawHill, 1935.
1944.
6. Davis JC. Advanced Physical Chemistry. New York: Ronald Press, 1965.
1981.
9. Barrow GM. Molecular Spectroscopy. New York: McGraw-Hill, 1962.
Winston, 1964.
11. Brittain EFH, George WO, Wells CHJ. Introduction to Molecular Spectroscopy.
London: Academic Press, 1970.
12. Guillory WA. Introduction to Molecular Structure and Spectroscopy. Boston: Allyn
and Bacon, 1977.
13. McHale JL. Molecular Spectroscopy. Upper Saddle River. NJ: Prentice Hall, 1999.
14. Rao CNR. Chemical Applications of Infrared Spectroscopy. New York: Academic
Press, 1963.
15. Kendall DN. Applied Infrared Spectroscopy. New York: Reinhold Pub., 1966.
16. Conley RT. Infrared Spectroscopy. Boston, MA: Allyn and Bacon, 1966.
17. Colthup NB, Daly LH, Wiberley SE. Introduction to Infrared and Raman Spectroscopy, 2nd ed. London: Academic Press, 1975.
18. Nakanishi K, Solomon PH. Infrared Absorption Spectroscopy. San Francisco, CA:
Holden-Day, 1977.
19. Bhagavantam S. Scattering of Light and the Raman Effect. New York: Chemical
Publishing Co., 1942.
20. Grasselli JG, Snavely MK, Bulkin BJ. Chemical Applications of Raman Spectroscopy. New York: Wiley-Interscience, 1981.
21. Grasselli JG, Bulkin BJ. Analytical Raman Spectroscopy. New York: John Wiley &
Sons, 1991.
22. Ferraro JR, Nakamoto K. Introductory Raman Spectroscopy. New York: Academic
Press, 1994.
23. Lewis IR, Edwards HGM. Handbook of Raman Spectroscopy. New York: Marcel
Dekker, 2001.
233
24. Wilson EB, Decius JC, Cross PC. Molecular Vibrations: The Theory of Infrared and
Raman Vibrational Spectra. New York: McGraw-Hill Book Co., 1955.
25. Weast RC. Handbook of Chemistry and Physics, 50th edn. Cleveland OH: Chemical
Rubber Co., 1969:F-158 F-161.
26. Flett MC. Characteristic Frequencies of Groups in the Infra-Red. Amsterdam: Elsevier
Pub., 1963.
27. Bellamy LJ. Advances in Infrared Group Frequencies. London: Methuen & Co., 1968.
28. Socrates G. Infrared and Raman Characteristic Group FrequenciesTables and
Charts, 3rd edn. Chichester: John Wiley & Sons, 2001.
29. Hochstrasser RM. Molecular Aspects of Symmetry. New York: WA Benjamin Inc.,
1966.
30. Orchin M, Jaffe HH. Symmetry, Orbitals, and Spectra. New York: Wiley-Interscience,
1971.
31. Cotton FA. Chemical Applications of Group Theory, 2nd ed. New York: Wiley-Interscience, 1971.
32. Harris DC, Bertolucci MD. Symmetry and Spectroscopy. New York: Oxford University Press, 1978.
33. Fearn T, Osborne BG. Near Infrared Spectroscopy in Food Analysis. Dublin, Ireland:
Longman Scientific & Technical Pub., 1986.
34. Williams P, Norris K. Near-Infrared Technology in the Agricultural and Food Industries. Saint Paul, MN: American Association of Cereal Chemists, 2001.
35. Siesler HW, Ozaki Y, Kawata S, Heise HM. Near-Infrared Spectroscopy: Principles,
Instruments, Applications. Weinheim: Wiley-VCH Pub., 2002.
36. Burns DA, Ciurczak EW. Handbook of Near-Infrared Analysis, 2nd edn. New York:
37. Ciurczak EW, Drennen JK. Pharmaceutical and Medical Applications of NearInfrared Spectroscopy. New York: Marcel Dekker, 2002.
8
Infrared Absorption Spectroscopy
David E. Bugay
Harry G. Brittain
INTRODUCTION
The applications of infrared (IR) absorption spectroscopy are one of the most fundamental and useful spectroscopic techniques available in the arsenal of a
pharmaceutical scientist. Typically, when most investigator scientists think of
vibrational spectroscopy, studies in the IR region of the spectrum come to
mind, owing to the commercial availability of spectrophotometers, the
common understanding of the technique, the ease of spectral interpretation
with the aid of correlation tables, and the widespread reference to the technique
in pharmaceutically based regulatory guidance documents. Within this framework, the IR spectroscopy is routinely used as an identification assay method
for various intermediate compounds, active pharmaceutical ingredients, excipients, and formulated drug products, and the methodology can also be developed
as a quantitative technique for any of these.
As important as IR spectroscopy might be, however, it must be remembered that the full vibrational analysis of a molecule would also include
Raman spectroscopy. Although IR and Raman spectroscopies are complementary
techniques, the widespread use of the Raman technique in pharmaceutical investigations has been limited. Before the advent of Fourier transform (FT) techniques and lasers, experimental difficulties tended to limit the use of Raman
spectroscopy. However, over the last 25 years a renaissance of the Raman
235
236
Bugay and Brittain
technique has been seen, due mainly to instrumentation development. Applications of the Raman effect in solid-state pharmaceutics will be the subject of
the following chapter.
The use of vibrational spectroscopy, specifically IR spectroscopy, in
pharmaceutical applications has become very diverse. Bulk drug substance
characterization uses traditional techniques, including structure elucidation,
routine compound identification, and solid-state characterization (15). In addition,
hyphenated techniques, such as liquid chromatography-IR spectroscopy (LC-IR)
(6) and thermogravimetric (TG) analysis-IR spectroscopy (TG-IR) (7,8) have
entered the arsenals of analytical and pharmaceutical scientists. Drug excipient
interactions, drug and/or excipient interactions with storage vessels, particulate
identification, contaminant analysis, and mapping of pharmaceutical tablets are
just a few of the microscopy-related applications of IR spectroscopy. Infrared
absorption spectroscopy has also found extensive use in biopharmaceutics as a
technique for the evaluation of peptide secondary structure (9), in reaction monitoring systems for chemical process engineers (10) and in on-line or at-line
systems for monitoring chemical processes, manufacturing operations, or clinical
packaging procedures (11).
It follows that in order for an investigator to extract the highest level of
information from the conduct of a vibrational analysis, it is imperative that the
basic concepts of the effect, the limitations of the instrumentation, and sampling
techniques be fully understood. In some cases, Raman spectroscopy may be the
more appropriate technique but might be cost prohibitive. Infrared spectroscopy
might be applicable, but the scientist must first appreciate the limitations of the
technique.
Since the advent of FT techniques and advances in laser technology and
computer systems, a myriad of sampling techniques have become available for
the vibrational spectroscopist. Analysis can now be performed on virtually any
type of sample, such as single crystals, bulk particulate material, slurries,
creams, films, solutions (aqueous or organic), oils, gas-phase samples, and
on-process streams. Additional advantages of vibrational analysis include their
typically nondestructive nature (with the ability to recover the material for
further characterization), quantitative technique under proper sampling conditions, and complementary to other characterization techniques.
SOME FUNDAMENTAL PRINCIPLES OF IR ABSORPTION
SPECTROSCOPY
A brief description of the IR spectroscopic theory will be presented so that the
important basics associated with the spectroscopic effect are available to the
reader. A more complete description of the underlying theory to IR spectroscopy
is found in the previous chapter and in the literature (12 17).
Upon irradiation of a sample by a broad-band source of IR energy, the
absorption of light having appropriate frequencies results in transitions between
237
molecular vibrational and rotational energy levels. As discussed in the previous

chapter, a vibrational transition might be approximated by the consideration of
atoms bonded together within a molecule as a harmonic oscillator. Based on
Hookes law, the vibrational frequency between two atoms might be approximated as:
s
1 k
n
2p m
(1)
where m is the reduced mass of the two atoms, such that m (m1m2)/(m1 + m2),
and k is the force constant of the bond (dynes/cm). Quantum mechanical
analysis of the harmonic oscillator model reveals a series of equally spaced
vibrational energy levels (defined by the vibrational quantum number v, where
v 0, 1, 2, 3 . . .) that are expressed as:

1
Ev v hn0
2
(2)
where Ev is the energy of the vth level, h is Plancks constant, and n0 is the fundamental vibrational frequency.
It must be noted that the fundamental vibrational frequencies in a polyatomic molecule do not correspond to the vibrations of single pairs of atoms,
but rather originate from the motions of groups of linked atoms. The absorption
of IR energy by a molecule corresponds to approximately 2 10-kcal/mole,
which in turn equals the stretching and bending vibrational frequencies of most
bonds in covalently bonded molecules.
The number of fundamental vibrational modes of a molecule is equal to
the number of degrees of vibrational freedom. As discussed in the previous
chapter, for a nonlinear molecule consisting of N atoms, 3N 2 6 of vibrational
freedom exist, and so there would be 3N 2 6 fundamental vibrational modes.
For a linear molecule, the number of fundamental vibrational modes equals
3N 2 5, and so in a total vibrational analysis of a linear molecule by complementary IR and Raman techniques, a total of five vibrational transitions
would be observed. Additional, but weak, absorption bands can also be observed
due to transitions not allowed by the harmonic oscillator model but that gain
intensity from anharmonic effects.
For a fundamental vibrational mode to be able to absorb IR energy, a
change in the molecular dipole must take place during the molecular vibration.
Typically, atoms in a molecule that possess different electronegativities and
that are chemically bonded will undergo a change in the net dipole moment
during normal molecular motion. As a result, antisymmetric vibrational modes
and vibrations due to polar groups are more likely to exhibit prominent IR
absorption bands.
238
Bugay and Brittain
The percent transmission (%T ) of an IR absorption spectrum might be calculated using:

Percent Transmission
I
100
I0
(3)
where I equals the intensity of transmitted IR radiation, and I0 equals the intensity
of the incident IR beam. In this representation, the abscissa of a percent transmission IR spectrum would be in units of wavenumbers (cm21) and %T on the
ordinate (Fig. 1).
The IR absorption spectra is presented in absorbance units using:

I0
Absorbance log
abc
(4)
I
where a is the absorptivity, b is the sample cell thickness, and c is the concentration. When there is an intention to use IR spectroscopy for quantitative purposes, one uses the absorbance presentation.
As previously mentioned, one of the most important pharmaceutical applications of IR spectroscopy is for the identification of molecular entities. This arises
from the fact that the fundamental vibrational frequencies of a small group of
atoms might be approximated by the harmonic oscillator model, indicating that
the energy of the vibration is determined by the force constant of the bond and
the reduced mass of the atoms in motion. For example, if the force constant (k)
for a double bond is 10 105 dynes/cm, then by the harmonic oscillator equation
Figure 1 The infrared vibrational spectrum of a compound represented displayed as

percent transmittance (upper trace) and absorbance (lower trace).
one can approximate the vibrational frequency for C55C as:

s
s
k
10 105
4:12
n 4:12
1682 cm1
m
12 12=12 12
239
(5)
A fairly good agreement exists between the calculated value of 1682 cm21
and the empirically observed value of 1650 cm21. Based on the harmonic oscillator approximation, numerous correlation tables are generated that allow one to
estimate the characteristic absorption frequency of a specific functionality (16).
On this basis, one might recognize contributions by the various functional
groups in a molecule in the IR spectrum, ultimately enabling the use of IR spectroscopy to identify a molecular entity. This understanding can subsequently
enable the performance of quantitative analysis on the molecule.
The far-IR region might be considered an extension of the mid-IR region,
because this area of the spectrum is typically used to investigate the fundamental
vibrational modes of heavy atoms. The region extends from 400 to 10 cm21, and
the low frequency molecular vibrations arise from the vibrational motion of
heavy atoms (such as metal ligand bonds). In addition, this region of the spectrum is used to study the intermolecular vibrations of crystalline materials. The
motion of molecules relative to one another gives rise to lattice modes, which
are at times referred to as external modes. These low-frequency modes are
very sensitive to changes in the conformation or structure of a molecule and
are observed in the range 200 to 10 cm21. Although this spectral region is
important for the study of polymorphism, very few literature reports on such
phenomena have been published. From a practical standpoint, investigations in
the far-IR region require nontraditional optical windows, such as polyethylene
or polypropylene because alkali halide windows do not transmit in this region.
The spectral quality also is compromised due to the existence of relatively
poor sources and insensitive detectors.
INSTRUMENTATION FOR THE MEASUREMENT OF
IR SPECTRA IN THE SOLID STATE
The instrumentation for the measurement of IR spectra varies almost as widely
as do its applications. Historically, mid-IR instrumentation made use of the conventional dispersive optical layout, where a source was rendered monochromatic, allowed to impinge on a sample, and the amount of beam attenuation
measured as a function of incident wavelength (18). The dispersive methodology suffers from a lack of sensitivity and worked as well as it did owing
to the strong degree of molar absorptivity associated with the electric dipole
allowed vibrational transitions. Although dispersive mid-IR instruments
remain in use, only a few manufacturers still produce such instrumentation.
However, dispersive mid-IR instruments have found a niche in the processmonitoring field.
240
Bugay and Brittain
The acquisition of high-quality IR spectra on solid materials has been made

possible by the introduction of the FTIR method, because this approach
minimizes transmission and beam attenuation problems. Essentially all FTIR
spectrometers use a Michelson interferometer, where IR energy entering the
interferometer is split into two beams by means of a beam splitter. One beam
follows a path of fixed distance before it gets reflected back into the beam splitter,
while the other beam travels a variable distance before being recombined with the
first beam. The recombination of these two beams yields an interference pattern,
where the time-dependent constructive and destructive interferences have the
effect of forming a cosine signal. Each component wavelength of the source
will yield a unique cosine wave, having a maximum at the zero pathlength difference (ZPD) and which decays with increasing distance from the ZPD. The position of the detector is such that the radiation in the central image of the
interference pattern will be incident upon it, and therefore the intensity variations
in the recombined beam are manifest as phase differences. The observed signal at
the detector is a summation of all the cosine waves, having a maximum at the
ZPD and decaying rapidly with increasing distance from the ZPD. If the component cosine waves can be resolved, then the contribution from individual wavelengths are observed. The frequency domain spectrum is obtained from the
interferogram by performing the Fourier transformation mathematical operation.
The review in the classic book by Griffiths and DeHaseth (19) gives a full
description of the components of an IR spectrophotometer.
Nevertheless, the greatest challenge for the IR spectroscopist is that of
sample preparation. Since the development of the FTIR spectrophotometers, an
abundance of sampling techniques for analytical purposes has been developed,
and it is most appropriate to examine the most important of these that are
widely used for pharmaceutical problem solving and method development.
Alkali Halide Pellet
The classic and most time-honored IR spectroscopic sampling technique is the
preparation of alkali halide pellet (20). This technique involves mixing the
sample of interest with powdered and dry alkali halide (typically KBr or KCl)
at a 1 2% w/w ratio of sample to alkali halide. The mixture is pulverized
until it is finely ground and homogeneous, placed into a die (typically made of
stainless steel), and subjected to approximately 10,000-psi pressure for a sufficient period to produce a glassy pellet. The pellet (with the sample finely
dispersed throughout the glass) might then be placed into the IR spectrophotometer for spectral data acquisition.
This sampling technique is widely used for the preparation of samples for
chemical identity testing and is commonly incorporated into a regulatory submission as an identification test for bulk drug substances or excipients (21).
The advantage of the pellet technique is that only a small amount of sample is
required (usually 1 mg) and a high-quality spectrum is obtained in a matter of
241
minutes. However, the investigator should know that disadvantages, such as

solid-state transformation of the sample due to the pressure requirements to
form the glass pellet and possible halide exchange between KBr or KCl and
the sample of interest (22,23) may take place. The disadvantages become critical
on considering the IR spectroscopy as one of the investigative tools in the study
of solid-state properties, because the act of sample preparation can serve to alter
the characteristics of interest.
Quantitation of mixtures was attempted with fair success utilizing the alkali
halide pellet sampling technique (24). However, owing to the disadvantages just
discussed, it is appropriate that this sampling technique be used only for simple
compound identification assays.
Mineral-Oil Mull
Another classical sampling technique for solids is the suspension of the analyte in
mineral oil (20). In this technique, a small amount of the sample (approximately
1 mg) is placed into an agate mortar, a small amount of mineral oil is added, and
the sample and oil are mixed to achieve an even consistency. The mixture that is
placed onto an IR-transmissive optical window is sampled by the IR spectrophotometer. One advantage to this technique is that there is no likelihood of solidstate transformations due to mixing and/or grinding, and hence the use of
mineral-oil mulls represents a good technique for the qualitative determination
of the polymorphic identity of a compound. Unfortunately, the mineral oil
itself exhibits a number of intense absorption bands (observed at 2952, 2923,
2853, 1458, and 1376 cm21) that may overlap important absorption bands associated with the sample of iterest. Typically, this technique is used for qualitative
identification assays whenever the alkali halide pellet technique is inappropriate.
Diffuse Reflectance Technique
The diffuse reflectance (DR) technique is one of the most important solid-state
sampling techniques for pharmaceutical problem solving and method development (25). Sometimes referred to as DRIFTS (diffuse reflectance infrared
Fourier transform spectroscopy), this technique has found extensive use in the
mid- and near-IR spectral regions. The technique involves irradiation of the
powdered sample by an IR beam, where the incident radiation undergoes absorption, reflection, and diffraction by the particles of the sample and only the
incident radiation that undergoes DR contains absorptivity information about
the sample. Figure 2 shows a schematic diagram for a DR sampling accessory.
A number of significant advantages are associated with the DR method of
sampling. Some samples may be investigated in their as-received condition, or at
worst would be studied after dilution with a nonabsorbing matrix, such as KBr or
KCl (typically at a 1 5%-w/w active to nonabsorbing matrix material ratio).
Macro- and micro-sampling cups are usually provided with the DR accessory,
with approximately 400 and 10 mg of sample being required for each cup,
242
Bugay and Brittain
Figure 2 Schematic diagram of a diffuse reflectance sampling accessory. The incident

infrared radiation is shown as entering from an opening in the collection mirrors and
impacting with the surface of the sample. The diffuse reflectance from the sample is collected by the mirrors and ultimately passed on to the detector.
respectively. The sample also is recovered completely, enabling other solid-state

investigations to be conducted on the same material. By the use of an environmental chamber attachment, variable-temperature and -humidity DR experiments
can be performed (26). By varying the temperature of the sample, information
about temperature-mediated crystal form transformations and the nature of the
interaction of a solvate with the parent molecule can be determined.
The DR technique lends itself to studies of polymorphic composition
because it is noninvasive, its use causes no changes in polymorph character
due to its inherent limited sample handling, and the technique can be used for
quantitative purposes (27). One disadvantage of DR IR spectroscopy is that the
spectral results are sometimes affected by the particle size of the analyte
(25,28). The development of DRIFTS quantitative assays for polymorphic composition must take the particle size of each component into account. Variations in
absorption band intensity may be falsely interpreted as concentration changes in
the analyte when, in reality, the variation is actually due to particle size effects.
Attenuated Total Reflectance

Attenuated total reflectance (ATR) is a sampling technique suitable for the study
of an analyte in any physical form other than the vapor phase (29,30). In the ATR
technique, IR radiation is passed through a crystal at an angle less than the critical
angle, which causes the light to undergo total internal reflection. At each such
reflection, the radiation penetrates a small distance beyond the crystal surface,
and if a substance capable of absorbing IR energy is in physical contact with
243
the crystal, then the internally reflected energy will be attenuated at those frequencies corresponding to changes in molecular vibrational states. As illustrated
in Figure 3, the IR beam passes through the optically denser crystal and reflects at
the surface of the sample.
An advantage of this technique is that it requires very little sample preparation, because one simply needs to place the sample in contact with the
crystal. Typically, the sample is clamped onto the surface of the crystal with
moderate pressure to ensure a sufficient degree of optical contact with the irradiating IR beam. The simplicity of sample handling makes the IR-ATR
method most attractive for the study of polymorphic and solvatomorphic
solids, because there is no opportunity for the sample to encounter any outside
substances or sources of energy that might perturb its nature and composition.
Obviously, the analyst must be scrupulous in the cleaning of the crystal
surface between analyses in order to avoid cross contamination. Because the
internal reflectance process does not permit the IR beam to pass very deeply
into the sample, it is typical to determine the composition or an analyte up to a
sampling depth in the range of 5 10 mm.
Microspectroscopy
The first linkage between a microscope and an IR spectrophotometer was
reported in 1949 (31), but now every manufacturer of IR spectrophotometers
offers an optical/IR microscope sampling accessory. Because optical microscopy
can be used to obtain significant information about a sample, such as its crystalline or amorphous nature, particle morphology, and size, the inclusion of IR
microspectroscopy in a program of study is appropriate for any solid-state
investigation. Interfacing the microscope to an IR spectrophotometer enables
Figure 3 Schematic diagram of a sampling accessory suitable for multiple reflection

attenuated total reflectance. The incident infrared beam enters the crystal and undergoes
multiple reflections at the crystal surfaces before ultimately leaving the crystal on its
way to the detector. At each reflection from the sample mounted at the crystal upper
surface, the beam is sufficiently attenuated by sample absorbance so that the absorption
spectrum can be derived from the beam.
244
Bugay and Brittain
investigators to obtain molecular identification of individual particles or crystals

and to provide a valuable means for the identification of particulate contamination in bulk or formulated drug products. Other advantages for the technique
are nondestructive sampling, the ability to make reflectance and/or transmittance
measurements, and minimal sample requirements (32).
The IR microspectroscopy sampling technique can be considered to be the
ultimate sampling technique, because the entire IR spectrum can be obtained on
only a single particle. However, due to the limitation of diffraction, the particles
of interest must typically be larger than 10 10 mm. The sample of interest is
positioned upon an IR optical window, and the slide placed onto the microscope
stage for its visual inspection. Once the sample is in focus, the field of view is
apertured down to the sample, where, depending upon sample morphology, thickness, and transmittance properties, a reflectance and/or transmittance IR spectrum may be acquired by the IR microscope accessory.
By using the hot-stage microscopy, one may monitor the IR spectra of
products being generated during the course of solid-state transformations.
Alternatively, one can use an environmentally controlled chamber to change
the atmosphere about a small sample and simultaneously observe the sample
using both optical and spectroscopic means. This approach would be particularly
useful when observing temperature-mediated crystal form changes that are
accompanied by changes in crystal morphology.
TG/IR Analysis
In this sampling technique, a TG analyzer is interfaced to an IR spectrophotometer so that the evolved gas from the heated sample in the TG furnace
is directed to an IR gas-phase cell. This IR sampling technique lends itself
to the identification and quantitation of residual solvent content for a
pharmaceutical solid (7) and for the investigation of pharmaceutical solvates/
hydrates (8).
Analogous to standard TG analysis procedures, the sample of interest for
TG/IR analysis is placed into a TG sample cup and introduced into the TG
furnace. The TG balance then monitors the weight loss of the sample as a function of temperature (typical heating rate of 108C/min). In ordinary practice, one
deduces the residual solvent content of a sample from the TG weight loss data,
that is, the amount of weight loss over a specific temperature range. However,
by acquiring the IR absorption data in addition to the weight loss measurement,
one can make an unequivocal identification of the evolved gases. This enables the
analyst to identify the species responsible for the weight loss and facilitates the
determination of the residual solvent level.
Although the residual solvent content of a sample can be obtained by the
TG/IR analysis, gas chromatography techniques usually outperform TG/IR
experiments because of their lower detection limits and ease of quantitation.
One area where the TG/IR technique excels is in the investigation of
245
pharmaceutical solvates and hydrates. Conventional thermal analysis methods,

such as thermogravimetry and differential scanning calorimetry are used to determine the presence of solvatomorphs after correlating a TG weight loss with a
DSC endotherm over the same range of temperatures. However, the use of
these methods cannot yield the identity of the solvent of crystallization, but
such information is obtained by measuring the IR spectrum analysis of the gas
evolved within the studied temperature range (8).
Photoacoustic Detection
The photoacoustic effect was first discovered by Alexander Graham Bell in
the early 1880s, and the instrumentation for its detection was once known as
the spectrophone (33). It took nearly a century for the detection scheme to be
applied to FTIR spectroscopy (34,35), when it was recognized that photoacoustic
spectroscopy (PAS) could be used to acquire spectra on opaque materials (which
are commonly found in pharmaceutical formulations). It was also recognized that
minimal sample preparation is necessary for measurement of the effect and that
depth profiling of the sample is possible.
The PAS phenomenon involves the selective absorption of modulated IR
radiation by the sample and subsequent measurement of the heat released when
the excited molecules return to their ground states. Once the IR radiation is
absorbed (obviously at frequencies that correspond to the fundamental vibrational
frequencies of the sample) and reradiated as IR energy having shifted frequencies,
the evolved heat escapes from the solid sample and heats a boundary layer of gas.
Typically, this conversion from modulated IR radiation to heat involves only a
small degree of temperature increase at the surface of the sample (approximately
10268C). Because the sample is placed into a closed cavity cell, which is filled
with a coupling gas (usually helium), the increase in temperature produces
pressure changes in the surrounding gas that can be detected by a very sensitive
microphone as sound waves. Because the incident IR radiation is modulated,
the pressure changes in the coupling gas take place at the same modulation frequency, causing the acoustic wave to vibrate at the same frequency. After detection by the microphone, the resulting signal is amplified by phase-sensitive
detection and Fourier-processed to yield the PAS spectrum.
The depth profiling of a solid sample may be performed by varying the
interferometer moving mirror velocity. By increasing the mirror velocity, the
sampling depth varies and surface studies might be performed. Limitations do
exist, but the technique has proven to be quite effective for solid samples (36).
In addition, particle size has a minimal effect upon the photoacoustic measurement, thus representing an advantage over DR sampling techniques.
Fiber Optics
The development of fiber optics in IR spectroscopy grew out of need to analyze
hazardous materials from a remote location. Solids are analyzed with a DR
246
Bugay and Brittain
arrangement of optical fibers, whereas liquids can be analyzed with an ATR

arrangement of optical fibers on the end of the probe. Much of the early development of IR spectroscopy using fiber optic probes was carried out in the near-IR
spectral region due to the limitations of the materials available for use as optical
fibers. With the development of chalcogenide fibers, the use of fiber optics has
extended into the mid-IR spectral region. Transmission measurements with chalcogenide fibers are suitable over the range of 4000 to 1000 cm21, but absorptions
due to hydrogen-containing impurities in the fiber limit the spectral region of
2250 2100 cm21. The normal use of the chalcogenide fibers is limited to fiber
lengths of approximately two meters in length (37).
APPLICATIONS OF IR SPECTROSCOPY TO AREAS OF
Infrared absorption spectroscopy has a significant history in the chemical and
physical characterization of pharmaceutical solids. Tremendous hardware developments in IR spectrophotometers took place after World War II. IR analysis
is the most widely reported spectroscopic technique utilized in the study of
solid-state crystal form issues because of the hardware developments and
because the IR spectroscopy is widely understood and the spectrophotometers
are relatively inexpensive (as compared to, e.g., solid-state nuclear magnetic resonance spectrometers). Additionally, because solid-state IR spectroscopy can be
used to probe the nature of polymorphism on the molecular level, the method is
particularly useful in instances where full crystallographic characterization of
polymorphism is not possible. Additionally, recent generations of hardware
design allows for relatively high spatial resolution of IR data acquisition
through microspectroscopy. This technique allows for drug product mapping
applications and for contaminant analysis.
Qualitative Identification of Pharmaceutical Compounds
The identity testing of compounds of pharmaceutical interest represents one
very important role of a spectroscopy laboratory. This mode of testing can be
accomplished using methods associated with the mid-IR or near-IR spectral
regions, and FT-Raman spectroscopy has begun to play a similar role. As
stated in the official general test of the U.S. Pharmacopeia (21), the IR absorption spectrum of a substance provides the most conclusive evidence of identity
that can be obtained from a single test. For bulk drug substances, the identity test
most often used is that of the general test <197k>, where the sample is processed
as a KBr pellet.
The identity testing of solid dosage forms of drugs usually involves a
number of wet chemical extraction steps, followed the processing of the
extract into a KBr pellet by the general method <197k>. For example, in the identity testing of leucovorin calcium tablets, one begins by grinding tablets,
247
dissolving the powder in water, sonicating the dispersion, and filtering. The filtrate is mixed with ammonium oxalate, shaken, and centrifuged. Methanol is
added to the preparation until it is clear, and then the preparation is placed in a
08C freezer until a precipitate is formed. The supernatant is decanted, methanol
is added to dissolve the precipitate, and the solution is allowed to evaporate. After
the remaining precipitate is dried, the KBr pellet is prepared, and the IR spectrum
obtained is compared with the IR absorption spectrum obtained from a reference
standard prepared by the same procedure.
Studies of the physical properties of pharmaceutical solids, and especially
the active pharmaceutical ingredient, are a very important aspect of the drug
development process. To that end, IR spectroscopy has become a widely utilized
physical characterization technique for the qualitative identification of compounds. For example, the IR absorption spectra of benzoic acid and sodium
benzoate are shown in Figure 4, where the differentiation between the two is
self-evident (38). The identity of the salt is readily confirmed by the shift in
its carbonyl band frequency to 1595 cm21 from the value of 1678 cm21 that is
characteristic of the free acid.
Figure 4 The FTIR ATR spectra in the fingerprint region of benzoic acid (solid trace)
and sodium benzoate (dashed trace). Abbreviation: FTIR ATR, Fourier transform infrared attenuated total reflectance. Source: Adapted from Ref. 38.
248
Bugay and Brittain
IR Spectral Studies of Polymorphic Systems

When coupled with a nonperturbing sampling mode, such as DR or ATR, the IR
spectroscopy can also be used to establish the phase identity of a potentially polymorphic substance. For example, as shown in Figure 5, the polymorphic identity
of famotidine can be readily established using FTIR-ATR spectroscopy, because
the spectra are sufficiently different in the fingerprint region to enable such a
distinction to be made (38). It should be noted, however, that the magnitude of
peak shifting between two forms can often be as little as 10 cm21, hence
careful attention needs to be paid to sample handling and peak measurement.
In the case of the monotropic polymorphic system formed by DuP-747, it
was found that distinctly different DR, Raman, and solid-state 13C-NMR spectra
existed for each physical form (39). The complementary nature of the IR and
Raman spectroscopies provided evidence that the polymorphic pair was
roughly equivalent in conformation, and it was deduced that the crystal structure
differences between the polymorphs of DuP-747 resulted from different modes of
molecular packing even though more accurate crystallographic information could
not be obtained.
Figure 5 The FTIR ATR spectra in the fingerprint region of famotidine Form-A (solid
trace) and Form-B (dashed trace). Abbreviation: FTIR ATR, Fourier transform infrared
attenuated total reflectance. Source: Adapted from Ref. 38.
249
The structure of only one of the two polymorphs of fosinopril sodium was
solved, because it has not been found possible to grow useful crystals of the
other form (40). Consequently, the two nonsolvated crystal forms were studied
using a multidisciplinary approach consisting of XRPD, IR, NMR, and thermal
analysis. Through detailed studies of the IR absorption and solid-state
13
C22NMR spectra, it was possible to conclude that the environment of the
acetal side chain of the fosinopril moiety differed in the two forms. In addition,
possible cis-trans isomerization about the C622N peptide bond was also identified. In the absence of single-crystal structural results, these conformational
differences were postulated as the origin of the observed polymorphism in
fosinopril sodium.
A common theme throughout many of the reports on polymorphic characterization is the use of a number of different physical analytical techniques. In the
majority of cases, XRPD is the primary method of choice for determining the
number of solid-state forms, because the effect originates from crystallographic
differences. Thermal analysis and solid-state spectroscopy were used to support
the XPRD results, and further characterize the structural aspects of the solid-state
system. A number of examples where a multidisciplinary approach to solid-state
form characterization was used are provided in Table 1. Outside of pure crystallographic investigations, the majority of studies on polymorphic systems make
use of multiple techniques, very often including solid-state IR absorption
spectroscopy.
In one such multidisciplinary study, five differently colored, solid-state
forms of 5-methyl-2-[(2-nitrophenyl)amino]-3-thiophenecarbonitrile were discussed (41). It has been deduced that the color of the polymorphic crystal
forms has a molecular origin and is related to the angle between the planes of
the phenyl and thiophene rings, which changes the characters of the molecular
orbitals of the systems and affects the degree of electron delocalization. The
IR spectra for three of the solid-state forms illustrate shifts in frequencies that
originate from the structural differences. For example, the nitrile stretching frequency varies by approximately 10 cm21 each for the yellow, orange, and red
solid-state forms of the compound. In the solid-state 13C-NMR spectra, the C-3
carbon of the thiophene ring shows a chemical shift difference of about 5 ppm
for the same three forms.
In a different type of multidisciplinary study, a theoretical and experimental study have been carried out on the three polymorphs of octahydro-1,3,5,7tetranitro-1,3,5,7-tetrazocine (82). The density functional theory and scaled
force-field methods were used to calculate the gas phase vibrational spectroscopy
for the molecule in different conformations and molecular symmetries, and the
authors compared these calculated spectra with experimentally observed data
to interpret the various crystal spectra. It was concluded that the molecule
existed in different conformations in each of the three polymorphs, with the molecule in the b-phase exhibiting Ci symmetry and those in the a- and d-phases
being characterized by C2V symmetry.
250
Table 1
Bugay and Brittain

Polymorphic Systems Studied Using Infrared Spectroscopy Techniques
Compound
AG-337
Carbamazepine
anhydrate/dihydrate
Carbovir
Cefamandole nafate
Cefazolin
Cefepime.2HCl
Compound H
Cortisone acetate
Delavirdine mesylate
Diflunisal
1,2-Dihydro-6-neopentyl2-oxonicotinic
11-a-Dimethylamino-2b-ethoxy-3-a-hydroxy5-a-pregnan-20-one
Dirithromycin
DuP-747
b-Estradiol
Fluocinolone acetonide
Fosinopril sodium
Frusemide
Indomethacin
Lactose
Losartan
Mefloquine HCl
40 -Methyl-20 nitroacetanilide
5-methyl-2-[(2nitrophenyl)amino]-3thiophenecarbonitrile
Nedocromil magnesium
Nicardipine HCl
Nimodipine
Paroxetine HCl
Ranitidine HCl
RG-12525
SC-41930
Spironolactone
SQ-33600
Character of the study
References
Qualitative characterization
Qualitative and quantitative
characterization
characterization
Quantitative characterization
(42)
(43 45)
(27)
(50)
(51 53)
(54 56)
(57)
(58)
(59)
Qualitative
Qualitative
Qualitative
Qualitative
Qualitative
Qualitative
Qualitative
Qualitative
Qualitative
Qualitative
Qualitative
characterization
characterization
characterization
characterization
characterization
characterization
characterization
characterization
characterization
characterization
characterization
(60)
(39)
(61)
(62)
(40)
(63)
(64)
(65)
(66)
(67)
(68)
(41)
characterization
(69)
(70)
(71)
(72)
(73,74)
(46)
(47)
(48,49)
(75)
(76)
(77)
(78)
(Continued)
251
Table 1 Polymorphic Systems Studied Using Infrared Spectroscopy Techniques

(Continued)
Compound
Sulfamethoxazole
Sulfaproxiline
Testosterone
Character of the study

characterization
References
(79)
(80)
(81)
In a study of the amorphous salt formed by the coprecipitation of cimetidine and diflunisal, the solid-state IR absorption spectroscopy was used to
prove the existence of the salt species (83). The prominent carbonyl absorption
band observed at 1650 cm21 in crystalline diflunisal could not be observed in
the spectrum of the amorphous salt, but a new peak was noted at 1580 cm21
that was assigned to an asymmetric stretching mode of a carboxylate group.
These findings were taken to indicate that the amorphous character of precipitates
formed by cimetidine and other nonsteroidal antiinflammatory agents were due
to salt formation and not nonbonding intermolecular interactions.
IR Spectral Studies of Solvatomorphic Systems
When the solid-state properties of solvates and hydrates are catalogued by means of
the multidisciplinary approach, one often finds that IR absorption spectroscopy
yields particularly useful information. For example, as shown in Figure 6, the characteristic absorption bands in the fingerprint and high-frequency regions permit the
ready differentiation of lactose anhydrate from lactose monohydrate (38). For
instance, the presence of a strong absorption band at an energy of 1068 cm21
readily identifies a lactose sample as consisting of the monohydrate phase. In the
high-frequency region, the well-defined absorbance band located at 3522 cm21
demonstrates the presence of crystalline water in the monohydrate phase.
Fluconazole has been isolated in a number of solvated and nonsolvated
forms, and the IR spectroscopy has proven to be an important tool in the characterization of these. In one study involving two nonsolvated polymorphs (Forms I
and II) and several solvatomorphs (the 14-acetone solvate, a 17-benzene solvate, and
a monohydrate), the IR spectra of the different forms showed differentiation in
bands associated with the triazole and 2,4-difluorobenzyl groups, and in the
propane backbone (84). In another study, the diagnostic IR spectral characteristics of nonsolvated Form-III and two solvatomorphs (the 14-ethyl acetate
solvate, and a monohydrate) were used to demonstrate the novelty of the new
forms relative to those in the literature (85).
During a study of the physicochemical properties of niclosamide anhydrate
and two of its monohydrates, it was found that the carbonyl frequency was particularly sensitive to the crystal form of the drug substance (86). Previously, the drug
252
Bugay and Brittain
Figure 6 The FTIR ATR spectra of lactose monohydrate (solid trace) and lactose anhydrate (dashed trace). Abbreviation: FTIR ATR, Fourier transform infrared attenuated
total reflectance. Source: Adapted from Ref. 38.
substance was known to be able to exist in two conformations, and the IR spectra
were used to demonstrate the conformational state of the compound in the
solvatomorphic forms. Similarly, the IR spectra obtained in the fingerprint and
high-frequency regions for four nonsolvated polymorphs of tenoxicam, and of
its acetonitrile, dioxane, dimethyl formamide, ethyl acetate, acetone, and isopropanol solvates facilitated a differentiation between the various crystal forms (87).
On the other hand, the mid-IR spectra of three solvatomorphs of tetroxoprim were
found to be virtually identical, indicating that vibrational spectroscopy is not
always the method of choice for such work (88).
The SQ-33600 was investigated by a variety of characterization techniques,
including solid-state 13C-NMR and IR absorption spectroscopies (78). Humiditydependent changes were noted in the crystal properties of this HMG-CoA
reductase inhibitor, where different types of hydrate formation could be achieved
depending upon the relative humidity of the storage condition. In each hydrate
type, a distinct IR vibrational frequency could be detected in the high frequency
region, indicating that different hydrogen-binding effects perturbed the
253
vibrational frequency of the crystalline water in the hydrates. It was postulated

that the hydration continuously varied within lattice channels.
Norfloxacin is somewhat unusual in that the hydrate form is more soluble in
water than is the anhydrate form. To investigate its hydration behavior, FTIR
microspectroscopy was used to examine the structural changes associated with
the hydration/dehydration processes that took place at different relative humidities (89). It was found that when norfloxacin anhydrate was converted to its
hydrate, the intensities of the absorption bands at 1732 and 1253 cm21 (assigned
to the C55O and C55O moieties of the carboxylic acid group) decreased gradually
with increasing water content. On the other hand, the intensities of the absorption
bands at 1584 and 1339 cm21 (corresponding to asymmetric and symmetric carboxylate modes) increased with the water content. In addition, the peak at
21
2553 cm21 (assigned to the NH
as the water
2 group) shifted from 2558 cm
content increased. These spectral changes correlated with transformations of
COOH to COO2 and from NH to NH
2 , attributable to a proton transfer from carboxylic acid group and suggestive that hydration can induce proton transfer processes in the solid state.
IR Spectroscopy Studies of Phase Transformations
As long as the pattern of the vibrational modes is sufficiently perturbed by differences caused by crystallographic variations between polymorphs or solvatomorphs,
the IR absorption spectroscopy can be used to study the phase transformation
process. For example, it is well known that granulation of anhydrous lactose with
water will result in the conversion of the anhydrate phase to the monohydrate
phase, and a change from the b-anomer to the a-anomer. Figure 7 shows the
FTIRATR spectra obtained in the high-frequency region of lactose monohydrate
and at different stages in its solution-mediated conversion to the monohydrate phase
(38). It is amply evident that, if desired, one could use the absorbance band at
3522 cm21 to follow the kinetics of the formation of the monohydrate.
The solution-mediated phase transformation of the metastable b-phase of
glycine to its stable a-phase was followed by IR absorption spectroscopy (90).
A number of changes in the IR spectra were found to take place during the
phase change, most notably the disappearance of the weak peak at 1660 cm21
once the transformation was complete. Other bands that were observed to shift
in energy during the process were the antisymmetric stretch of the carboxyl
group (1580 1590 cm21), the deformation of the NH
3 group (1515
1500 cm21), and the O22C55O group bend (around 700 cm21).
The IR absorption spectroscopy played an important role in the characterization of the two polymorphs of racemic [2-[4-(3-ethoxy-2-hydroxypropyl)phenylcarbamoyl]ethyl]trimethyl-ammonium benzene sulfonate, and in the studies
of the phase conversion of the least stable g-phase into the most stable 1-phase
(91). Very large band frequencies were observed for the O22H stretching
modes (3340 3600 cm21) and the SH stretching modes (1160 1260 cm21),
254
Bugay and Brittain
Figure 7 The FTIR-ATR spectra in the high-frequency region of lactose monohydrate

(dotted trace), after approximately 50% conversion to the monohydrate phase (solid
trace), and full conversion to the monohydrate phase (dashed trace). Abbreviation:
FTIR ATR, Fourier transform infrared attenuated total reflectance. Source: Adapted
from Ref. 38.
indicating that the bonding modes of the hydroxyl and benzene sulfonate groups
were distinctly different in the two crystal forms. Because the kinetics of the
phase transition was exceedingly rapid, the process could only be followed
using in situ FTIR ATR techniques.
By the use of FTIR microspectroscopy, the phase transformation of acetaminophen polymorphs could be studied (92). The as-received substance (Form-I)
was found to melt around 1658C but to solidify into an amorphous state upon
cooling. If this glassy substance was heated again, it was found to crystallize
at 858C into Form-III. Continued heating of the newly formed phase led to observation of another phase transformation into Form-II around 1188C. Although the
changes could be followed visually through the microscopy, changes in the profiles and bandshapes of the IR absorption peaks proved to be a much more selective means to deduce the temperatures of the phase transitions.
The dehydration, rehydration, and solidification processes of trehalose
dihydrate have been studied using vibrational spectroscopy (93). Trehalose dihydrate undergoes a dehydration transition around 1008C, and the IR spectra obtained
during this process were found to change significantly over the 15001800 cm21
region. For example, the band intensities at 1640 and 1687 cm21 underwent sharp
decreases around 658C but remained relatively constant at higher temperatures.
255
These phenomena were taken to imply the existence of a phase transition of trehalose dihydrate over the temperature range of 64678C, reflecting a transformation
where the lattice water became liquid-like prior to the eventual dehydration.
During the rehydration process where trehalose anhydrate became trehalose dihydrate, the liquid-like water became solid-like water over the same temperature
range. It was postulated that this phase transition could be related to the protective
effect of trehalose in preserving protein stability.
IR Spectroscopic Studies of Manufacture and Processing
When the vibrational spectra of polymorphic or solvatomorphic systems are
suitable, IR absorption spectroscopy and ATR sampling can be used to follow
and study the crystallization of the respective forms. Fevotte (94) has published
an excellent summary of perspectives for the use of on-line monitoring of
pharmaceutical crystallization processes with the use of in situ IR absorption
spectroscopy. As long as the spectroscopy is appropriate, one can insert an
IR ATR probe into a crystallizing vessel, and determine the solubility profiles
of drug substances, as well as the limits of their metastable zones. In addition,
one can evaluate appropriate seeding policies and the parameters required to
obtain a real increase in the quality of the final product. Finally, the use of in
situ IR absorption spectroscopy should have potential in following the impurity
content in a vessel, and the results are used to minimize such species.
For example, the FTIR ATR methodology was used to monitor the supersaturation level during a batch cooling crystallization of succinic acid (95).
The system was set up to monitor the absorbance ratio of the peak area within
the 1806 1675 cm21 region (carboxylate carbonyl absorption) to the peak
area of the 1671 1496 cm21 region (water absorption) and could be used with
prior knowledge of kinetic data. Using a feedback control loop, the system
could maintain the supersaturation level close to the equilibrium solubility
during the cooling process and, hence, could be optimized to yield the appropriate
crystallite size.
In other studies of this type, the FTIR ATR methodology and suitable in
situ probes have been used to study the crystallization of mandelic acid (96),
paracetamol (97), and glycine (98) from water. These and other examples
amply demonstrate the utility of in situ IR absorption spectroscopy as a
general method applicable for process analytical control.
The effect of secondary processing on the solid-state properties of substances having pharmaceutical interest is well known, and a variety of methodologies have been used to study such transformations (99). For example, when
benzoic acid was compressed with methylated cyclodextrins, the carbonyl
band normally observed at 1688 cm21 was found to shift up to 1720 cm21
(100). It was determined that the compression process induced proton transfer
in the hydrogen-bonded eight-membered ring of dimeric benzoic acid pairs
and that this proton transfer was responsible for the frequency shift.
256
Bugay and Brittain
The effect of particle pulverization on the properties of glisentide was

studied using a multitude of techniques, including IR absorption spectroscopy
(101). The grinding of the drug substance, or cogrinding it with polyvinylpyrrolidone, was not found to induce a change in the polymorphic composition, but it
did result in the formation of an amorphous content and a consequent increase in
the dissolution rate. The various peaks in the IR spectrum reflected this trend in
that no shifting of peak maxima was observed, and the bands underwent a general
reduction of intensity and a loss of spectral resolution.
The effect of secondary processes, such as size reduction, wet granulation,
consolidation, and compression on celecoxib and its solvatomorphs, was studied
(102). Although the polymorphic state of the nonsolvated form and the dimethylacetamide solvate did not change with milling, shifts in the IR spectrum that
were caused by milling the dimethylformamide solvate indicated a loss of
bound solvent and transformation to the nonsolvated form. In the DMF solvate
form the NH stretching doublet is absent, but upon milling (and the accompanying desolvation), the band characteristic of the nonsolvate developed at
3235 cm21. Wet granulation of celecoxib did not lead to any change in polymorphic condition, but the solvates were converted into the nonsolvated form
by the action of the aqueous granulation fluid. The same type of desolvation
process could be induced by the compression of the solvates into a compact.
The effect of grinding on the photostability of two polymorphs of nicardipine hydrochloride was studied by a number of techniques, including IR reflectance spectroscopy (103). The nicardipine content at the surface of a solid
dosage form could be readily determined on the basis of its carbonyl absorbance
at 1700 cm21, and the formation of photodegradation products could be detected
by their absorption at 1734 cm21. It was found that the photodegradation rate constant of the b-form was greater than that of the a-form, although both phases were
found to decompose much more quickly when being ground to a fine powder.
Quantitative Analytical Applications
In the studies of celecoxib and nicardipine hydrochloride just discussed, qualitative IR spectroscopy was used to obtain information about the phase transformations involved. However, in both of these works the investigators took
the qualitative aspects discussed earlier and validated quantitative methods
for the determination of the species involved. In the case of celecoxib, the
ratio of the intensity of the doublet associated with the S55O symmetric stretching mode (1135 and 1165 cm21) to the intensity of the singlet associated with the
aromatic C55C stretching mode (1498 cm21) was found to be a linear function of
the free celecoxib concentration in its mixtures with the dimethylacetamide
solvate (102). For nicardipine hydrochloride, the intensity of carbonyl absorbance at 1700 cm21 was used as a quantitative indicator of the amount of residual
drug substance after photodecomposition (103).
When working with the DR technique, two critical factors must be kept in
mind when developing a quantitative assay, namely the production of
257
homogeneous calibration and validation samples and a consistent particle size for
all components, including subsequent samples for analysis. During the assay
development and method validation, one must be cognizant of mixing techniques
available to achieve homogeneous samples, because inhomogeneity in the
calibration and validation samples can lead to inaccurate band absorption
values and subsequent prediction errors. In addition, variation in the particle
size of the nonabsorbing matrix or of the sample can influence the DR IR spectrum, which can also lead to prediction errors.
These considerations were fully taken into account during the development
of a quantitative DR assay for cefepime dihydrochloride dihydrate in a matrix of
cefepime dihydrochloride monohydrate (27). The validation process was hampered by difficulties encountered in mixing reference materials having very
different particle morphologies, but this was overcome by slurrying the materials
in hexane, mixing thoroughly, and then evaporating off the inert solvent.
A working range of 1.0 8.0% was established, with a detection limit of 0.3%
and a quantitation limit of 1.0%. The effect of particle size on the analytical parameters was also evaluated, and it was established that the IR method was valid
for particles larger than 125 mm but smaller than 590 mm.
Sulfamethoxazole is known to exist in at least two polymorphic forms that
have been fully characterized (79). Because the different forms exhibit distinctly
different DR mid-IR spectra, it was possible to quantitate one form in the presence of the other. During the assay development, a number of mixing techniques
were investigated in an effort to achieve homogeneous samples. After mixing and
particle size factors were optimized, a quantitative DR IR spectroscopic assay
method was developed in which independent validation samples were predicted
within 4% of theoretical values.
In some pharmaceutical spectroscopy laboratories, TG/IR analysis is routinely used for solvate identification of pseudopolymorphic compounds. For
example, a bulk drug substance exhibited distinctly different X-ray powder diffraction and differential scanning calorimetry data when compared to the reference standard data (8). Figure 8 displays the TG analysis weight loss curve,
and the mid-IR spectra that were acquired at various time points and temperatures
for the sample in question. Based upon the existence of absorption bands at 2972
and 1066 cm21, an organic species containing aliphatic C22H and C22C22O
moieties, respectively, was deduced to be present in the evolved gas. Subsequent
analysis of the IR spectra collected at the completion of the weight loss (spectral
library matching) revealed that an ethanol solvate was present. Thus, TG/IR
provided an unequivocal characterization as to the origin of the solvatomorphic
character of this particular compound (8).
The IR spectroscopy and ATR have been used as a means to detect concentrations of principal species during the titration of malic acid (104). After subtracting the various types of water (pure, acidic, or basic water) from the
spectra, factor analysis was used to deduce the three malic species as being
nonionized malic acid, a monosodium malate, and a doubly ionized salt.
Although these studies were conducted in aqueous media, there is no reason
258
Bugay and Brittain
Figure 8 Thermogravimetric weight loss curve and subsequent infrared spectra

measured as a function of temperature. A slight lag time exists between the TG weight
loss and the infrared spectral acquisition due to the evolved gas having to be swept into
the gas sampling cell by the helium carrier gas Abbreviation: TG, thermogravimetric.
why an investigator could not use the results of a similar solid-state study to
evaluate the salt species existing in a bulk drug substance. In fact, IR absorption
spectroscopy is one of the methods whereby one can differentiate the patented
hemisodium salt of valproic acid from the nonpatented monosodium salt (105).
IR Spectroscopic Studies During the Preformulation Stage
of Drug Development
During preformulation, the possible interactions between a drug substance and its
proposed excipients are typically studied using either thermal analysis (106) or
chromatographic (107) techniques. Unfortunately, the use of chromatography
requires extraction and/or dissolution techniques that may destroy critical physical and chemical information, and the use of thermal analysis is often
accompanied by false results associated with the measurement technique itself.
There is no doubt that a methodology with the ability to study drug excipient
interactions in a noninvasive manner is of great value, and IR absorption spectroscopy certainly can meet the need.
The utility of IR absorption spectroscopy in support of drug interaction
studies can be illustrated through a study of intentionally selected incompatible
materials. It is well known that omeprazole is unstable under acidic conditions,
and that the substance undergoes rapid decomposition when allowed to interact
with an acid (108). A 50%-w/w blend of omeprazole and salicylic acid was
259
prepared, and as would be expected the IR spectrum of the blend consisted of the
amalgamation of the spectra of the individual components (38). After this blend
was exposed to an environment of 408C and 75% relative humidity for 14 days
(Fig. 9), the IR spectrum was now found to reflect almost entirely of the spectrum
of salicylic acid. This finding demonstrates the practically complete decomposition of the omeprazole during this accelerated stability study and the incompatibility of the two substances.
The reaction of isoniazid (isonicotinic acid hydrazide) with magnesium
oxide was studied by several techniques, including IR absorption spectroscopy
(109). Chemisorption and physisorption led to new ultraviolet absorption
bands, imparting color into stressed mixtures. Through a comparison of the IR
spectrum of a genuine coordination complex of isoniazid with those of the starting material and the MgO product, it was concluded that no formation of a
Figure 9 The FTIR ATR spectra in the fingerprint region of a 50%-w/w blend of omeprazole with salicylic acid (dashed trace), and the same sample after being exposed to
408C and 75% relative humidity storage condition for 14 days (solid trace). Also shown
for comparison purposes is the spectrum of salicylic acid (dotted trace). Abbreviation:
FTIR ATR, Fourier transform infrared attenuated total reflectance. Source: Adapted
from Ref. 38.
260
Bugay and Brittain
Mg(II)-isoniazid coordination complex took place. Rather, the incompatibility

between the drug and excipient was attributed to a donor acceptor mechanism,
with the pyridine ring of the drug substance acting as the donor.
The interaction of dextro-amphetamine sulfate with lactose has been
studied in detail, with the degree of reaction being more pronounced with
spray-dried lactose owing to the higher amounts of 5-hydroxymethylfurfual in
the excipient (110). The presence of a C55N vibrational band, which was not
present in the initial materials, in the IR absorption spectrum of the decomposition product was used to assist in an identification of the brown product as
being the amphetamine-hydroxymethylfurfual Schiff base reaction product commonly associated with the Maillard-type reaction.
Differential scanning calorimetry and DR IR spectroscopy have been used
to study the yellow or brown color that develops when aminophylline is mixed
with lactose (111). The DSC thermogram of the aminophylline/lactose
mixture is not a direct superposition of the individual components, and this observation was taken as an indication of an incompatibility. After the complete analysis of the IR spectra of the individual components, physical mixtures of these, and
various blended samples subjected to stress conditions (608C for three weeks), it
was concluded that ethylenediamine is liberated from the aminophylline complex
and reacts with lactose through the usual Schiff base intermediate. This reaction,
in turn, results in brown discoloration of the sample.
Combinatorial chemistry is rapidly becoming a reliable way for pharmaceutical companies to discover and identify drug candidates and to study their
interactions with potential excipient materials. For example, an IR spectroscopic
method for the analysis of solid-phase organic reactions that occur on resin beads
was described (112,113). This method represents an improvement over previous
IR spectroscopic methods that required preparing a KBr pellet from at least
10 mg of resin beads. In the described method, a drop of resin solution was
removed from the reaction vessel, washed in organic solvents, and dried under
vacuum. The dried resin beads are placed on an NaCl window, flattened to
10 15 mm thickness, and spectra acquired in the transmission mode of the IR
microscope. The authors demonstrated that the flattened beads provided superior
results to nonflattened beads.
Bioanalytical Applications
Pharmaceutical companies are showing great interest in developing products
based on proteins, enzymes, and peptides, and with the development of such
products come the need for methods to evaluate the purity and structural
nature of these biopharmaceuticals. However, in the solution phase these biopolymers are susceptible to a variety of chemical and physical degradation
processes (114), which has caused many formulators to developed lyophilized
dosage forms. Apart from a wide range of chemical instabilities (115), protein
pharmaceuticals are prone to physical changes in their secondary structure that
261
lead to the deleterious effects of denaturation, aggregation, surface adsorption,

and precipitation (116). Consequently, methods to evaluate the secondary
structure content of pharmaceutically active biopolymers are required, and IR
absorption spectroscopy has proven to be one of the important methods
(117 119).
The secondary structure of proteins is reflected in the amide-I (1600
1700 cm21), amide-II (15001600 cm21), and amide-III (12251300 cm21)
absorption bands, as these are primarily associated with the stretching vibrations
of peptide carbonyl groups. Of these, the amide-I and amide II bands were found
to be particularly useful for IR-based work, and the measurement of the various
band energies enable an identification of the peptide secondary structure type. For
example, poly(L -lysine) exists completely in the a-helix conformation at pH 11.2,
exhibiting an amide-I band at 1644 cm21 and a split amide-II band located at
1521 and 1548 cm21 (120). Heating this solution at 30408C causes a conformational change to the antiparallel b-sheet, whereupon the amide-I band splits to
absorptions at 1615 and 1691 cm21, and the split amide-II bands shift to 1535
and 1563 cm21. At pH 6.5, poly(D,L -alanine) is known to exist in the random coil
conformation, exhibiting its amide-I band at 1650 cm21 and its amide-II band at
1551 cm21.
The effect of lyophilization on the secondary structure of a-chymotrypsin
was reported, and a comparison of the solution phase spectrum within the amide-I
band with that of the dried solid is shown in Figure 9 (121). Quantitative analysis
of the solid-phase spectrum yielded values for b-sheet, a-helix, b-turn, and
random coil percentages that agreed closely with the values derived from an
X-ray crystallographic analysis. However, the lyophilization process was found
to lead to a reduction in the percentages of random coil and b-turn percentages
relative to the solution phase composition, and an increase in the content of bsheet and a-helix. It was concluded that even though these differences were
not large, the protein was not in its native form in the lyophilized solid.
As mentioned, maintaining a biopolymer as close as possible to its native
state during lyophilization is crucial for that substance to maintain its efficacy,
and for this reason formulators normally include a cryoprotectant in their formulations. It should be recognized, however, that the freezing and subsequent dehydration steps of lyophilization present different types of stress to the secondary
structure of a protein, and these should be treated separately for optimal
protein stabilization. The effect of dimethyl sulfoxide, glycerol, and ethylene
glycol on the secondary structures of cytochrome C and lysozyme were
studied using IR absorption spectroscopy (122). Although the secondary structure
of cytochrome C seemed to remain unchanged by the cosolvents, shifts in the
amide-I band of lysozyme indicated the formation of additional a-helical
content at the expense of the b-turn and random coil contents.
The secondary structures of two labile enzymes, lactate dehydrogenase and
phosphofructokinase, have been evaluated using IR absorption spectroscopy in
the aqueous and lyophilized states (123). The addition of 10-mM mannitol,
262
Bugay and Brittain
lactose, or trehalose or 1% polyethylene glycol (PEG) to the enzyme solutions

attenuated the protein unfolding, but the IR spectra still indicated the existence
of significant spectral differences for the enzymes between the dried state and
the aqueous conformation. However, when a combination of 1% PEG and
either10-mM mannitol, lactose, or trehalose was added, the native structure
was preserved during lyophilization, and essentially full enzymatic activity
was recovered upon reconstitution. The authors concluded that for labile proteins,
the preservation of the native structure during lyophilization is esssential for the
recovery of activity following rehydration.
The FTIR spectroscopy was used to characterize the interaction of stabilizing carbohydrates with dried proteins (124). The lyophilization of lysozyme with
trehalose, lactose, or myoinositol resulted in substantial alterations of the IR
spectra of the dried carbohydrates. It was also found that dehydration-induced
shifts in the positions of the lysozyme amide-I and amide-II bands could be partially and fully reversed, respectively, when the protein was lyophilized in the
presence of either trehalose or lactose. The results indicated that not only did
hydrogen bonding exist between the dried proteins and the carbohydrates but
also that carbohydrate binding to the biopolymers was required for the carbohydrates to induce protein stabilization during the processes of lyophilization
and rehydration.
The interactions between proteins and carbohydrates in the dried state have
been further studied using DR FTIR methods (125). Recombinant human deoxyribonuclease I and recombinant human insulin were lyophilized with mannitol,
sucrose, trehalose, and dextrans of different molecular weights, and the IR spectra
demonstrated that changes in the protein secondary structure accompanied
the lyophilization process. For both proteins, the presence of intermolecular
b-sheets due to aggregation was detected, and the a-helical content decreased
significantly. The use of carbohydrates as cryoprotectants was found to inhibit
rearrangements in the protein secondary structure, with sucrose and trehalose
being the most efficient for this purpose. However, whatever conformational changes were noted for these proteins appeared to be reversible after
rehydration.
The IR spectroscopic examination of the secondary structure in the driedstate conformation of interleukin-2 was used to determine the pH conditions and
stabilizers that provide optimal storage stability for the lyophilized product (126).
At pH 7, the biopolymer unfolds extensively upon lyophilization, whereas at pH
values less than 5 it remained essentially in its native form. A strong correlation
was observed between the retention of the native structure during freeze-drying
and enhanced stability, as interleukin-2 prepared at pH 5 is approximately an
order of magnitude more stable with regard to the formation of soluble and insoluble aggregate protein prepared at pH 7. A similar pH stability profile was
observed in the presence of excipients, although the excipients alter the overall
stability profile. It was concluded that excipients with the capacity to substitute
for water upon dehydration better preserve the native structure, and this,
in turn, results in enhanced stability.
263
Chemical Mapping and Imaging

Probably one of the most exciting new applications of IR spectroscopy is that of
chemical mapping and imaging, where investigators generate a chemical image
of a two-dimensional area of a sample (127). The spectroscopic chemical
imaging technique relies on the interface of an optical microscope, equipped
with a motorized stage, to an IR spectrophotometer. The operator is able to visually focus upon a sample of interest and make optical observations regarding the
sample, such as morphology, separation of layers, and the like. Subsequently, the
same two-dimensional visual area is defined for spectroscopic analysis. Utilizing a
raster pattern (i.e., the stepwise movement of the motorized stage along the x- and
y-axes), individual spectra are acquired for each spatial location within the twodimensional area. Spatial resolution is typically defined by the technique (approximately 5 10 mm for IR work). Once the individual spectra are obtained for the
two-dimensional area, the intensity of a specific spectral feature within each spectrum can be plotted as a function of the spatial position. In this manner, one is able
to obtain a contour plot showing the spatial position of a chemical entity.
One example of the use of IR mapping is the physical characterization of
multicomponent pharmaceutical formulations. In order to provide immediate
or controlled release rates of the active pharmaceutical ingredient, pharmaceutical formulations are becoming more complex. In addition, the mode of
action for a drug substance may require specific dissolution of the compound
within the intestinal tract (typically via an enteric coating) or within the
stomach. To facilitate these types of drug delivery, multiple-layered solid dose
formulations have become commonplace and are ideally suited to the IR
mapping methodology.
To facilitate the release of the drug substance within the intestinal tract, a
formulation was produced where the active pharmaceutical ingredient was spray
coated onto a 90 100-mm sugar sphere. This coated core was overlaid with an
intermediate excipient layer, and then a final enteric coating was applied to
provide protection to the microsphere from acidic dissolution within the
stomach. Chemical and physical characterization of the microsphere was
required to determine the thickness of the various layers, their integrity, and
their degree of uniformity. In order to obtain this information, a number of microspheres were embedded within an epoxy matrix such that the block could be
mounted into a microtome. The epoxy block was then sectioned in controlled
cuts so that a half-sphere microsphere was exposed on itssurface. The block
was then mounted on a leveling stage positioned on the stage of an IR microscope. It was required that the surface of analysis be kept orthogonal to the IR
beam so that a constant focus could be maintained for the entire two-dimensional
area sampled by the mapping experiment.
After obtaining an optical microscopy picture of the area to be sampled
(Fig. 10A), a two-dimensional area of analysis was defined by the IR mapping
software in conjunction with the operators view of the sample. For this particular
example, a line map was performed where a line of analysis was drawn by the
264
Bugay and Brittain
Figure 10A Comparison of the infrared absorption spectra of a-chymotrypsin in its

aqueous (dashed trace) and dried solid (solid trace) states. Source: Adapted from
Ref. 113 and the absorbance scale is in arbitrary units
mapping software, and the operator designated that IR spectra be collected along
this line in 20-mm increments. In actuality, a rectangular sampling area having
dimensions of 20 70 mm was designated. After the set-up of the sampling parameters, an IR spectrum was acquired at each spatial position. As one can
imagine, thousands of spectra can easily be accumulated for very large sampling
areas that were designated with small spatial increments. Fortunately, automated
processing software is available to subsequently produce the spectral images in
accord with a diagnostic spectral feature.
In this case, Figure 10B shows the entire IR spectrum for each spatial position sampled by the experiment. The x-axis represents traditional IR energy (in
units of wavenumbers), and the y-axis represents a spatial position along the line
of analysis. The z-axis in this case represents the intensity of each spectral
feature, and in the figure, blue signifies strong intensity and red signifies weak
intensity). By examining Figure 10B, one can clearly discern unique spatial
areas due to the envelope of common spectral features from one spatial position
to another. Starting at the bottom of Figure 10B, the first 20 mm represents one
chemical entity based upon its unique spectral features. After the spectral
interpretation based on consultation with IR libraries (128), this layer was identified as a particular enteric coating excipient. A specific polymorphic form of the
drug substance was determined to reside approximately 20 80 mm further in,
265
Figure 10B The infrared mapping results for sampling across a microtomed beadlet
containing a drug substance. An optical image of the half-sphere beadlet is displayed in
the upper right corner of the figure. The spectrum presented in Section A represents one
spatial position from the total sampling area. Section B of the figure presents all of the
spectra collected from the total sample area. Within Section B of the figure, the y-axis
represents distance along the sampling area (top of the arrow pointing toward the core of
the beadlet), the x-axis is the infrared absorption band (in cm21), and the black and gray
colors represent the intensity of spectral features (black being most intense, and gray
being least intense).
and a relatively small intermediate layer was then identified from approximately
80 110 mm. Finally, the sugar core was identified from approximately the
110-mm position and into the core of the microsphere. From this type of analysis,
one can easily identify different chemical species (i.e., drug substances or formulated excipients), different physical forms of the drug substance, and the spatial
locations of different chemical species. This technique has been widely applied to
drug products (tablets, microspheres, etc.) and drug delivery devices (transdermal
patches, stents, etc.).
REFERENCES
1. Markovich RJ, Pidgeon C. Pharm Res 1991; 8:663.
2. Bugay DE, Williams AC. Vibrational spectroscopy. In: Brittain HG, ed. Physical
Characterization of Pharmaceutical Solids. New York: Marcel Dekker, 1995:59 91.
3. Bugay D. Pharm Res 1993; 10:317; Threlfall TL. Analyst 1995; 120:2435.
4. Brittain HG. J Pharm Sci 1997; 86:405.
5. Brittain HG. Methods for the characterization of polymorphs and solvates. In:
Brittain HG, ed. Polymorphism in Pharmaceutical Solids. New York: Marcel
Dekker, 1999:227 278.
266
Bugay and Brittain
6.
7.
8.
9.
10.
Griffiths PR, Lange AJ. J Chromatogr Sci 1992; 30:93.

Johnson DJ. Compton DAC Spectroscopy. 1988; 3:47.
Rodriguez C, Bugay DE. J Pharm Sci 1997; 86:263.
Kim Y, Rose CA, Liu Y, Ozaki Y, Datta G, Tu AT. J Pharm Sci 1994; 83:1175.
Lynch J, Riseman S, Laswell W, Tschaen D, Volante R, Smith G, Shinkai I. J Org
Chem 1989; 54:3792.
Dempster MA, MacDonald BF, Gemperline PF, Boyer NR. Anal Chim Acta 1995;
310:43.
Herzberg G. Infrared and Raman Spectra of Polyatomic Molecules. 6th edn.
New York, NY: D. Van Nostrand Reinhold Co., 1954.
Rao CNR. Chemical Applications of Infrared Spectroscopy. New York: Academic
Press, 1963.
Kendall DN. Applied Infrared Spectroscopy. New York: Reinhold Pub., 1966.
Conley RT. Infrared Spectroscopy. Boston, MA: Allyn and Bacon, 1966.
Colthup NB, Daly LH, Wiberley SE. Introduction to Infrared and Raman Spectroscopy, 2nd edn. London: Academic Press, 1975.
Nakanishi K, Solomon PH. Infrared Absorption Spectroscopy. San Francisco, CA:
Holden-Day, 1977.
Anderson DH, Woodall NB, Infrared spectroscopy. In: Weissberger A, Rossiter BW,
eds. Physical Methods of Chemistry. Vol. I, New York: Wiley-Interscience, 1972:
1 84.
Griffiths PR, de Haseth JA. Fourier Transform Infrared Spectrometry. New York:
Stewart JE. Infrared Spectroscopy: Experimental Methods and Techniques.
New York: Marcel Dekker, 1970.
Spectrophotometric Identification Tests, Infrared Absorption, General Test <197>,
United States Pharmacopeia, 28th edn. United States Pharmacopeial Convention,
Rockville, MD, 2005:2295.
Bell VA, Citro VR, Hodge GD. Clays and Clay Minerals 1991; 39:290.
Mutha SC, Ludemann WB. J Pharm Sci 1976; 65:1400.
Hlavay J, Inczedy J. Spectrochim Acta 1985; 41A:783.
Fuller MP, Griffiths PR. Anal Chem 1978; 50:1906.
The Complete Guide to FT-IR, Spectra-Tech, Inc., Stamford CT, 1993.
Bugay DE, Newman AW, Findlay WP. J Pharm Biomed Anal 1996; 15:49.
Culler SR. Diffuse reflectance spectroscopy: sampling techniques for qualitative/
quantitative analysis of solids. In: Coleman PB, ed. Practical Sampling Techniques
for Infrared Analysis. Boca Raton, FL: CRC Press, 1993:93 105.
Compton SV, Compton DAC. Optimization of data by internal reflectance spectroscopy. In: Coleman PB, ed. Practical Sampling Techniques for Infrared Analysis.
Boca Raton, FL: CRC Press, 1993:55 92.
Urban MW. Attenuated Total Reflectance Spectroscopy of Polymers. Washington,
DC: American Chemical Society, 1996.
Barer R, Cole ARH. Nature 1949; 63:198.
Reffner JA, Coates JP, Messerschmidt RG. Amer Lab 1987; 19:5.
Bell AG. Amer Assoc Proc 1881; 29:115.
Vidrine DW. Appl Spectrosc 1980; 34:314.
Rockley MG. Appl Spectrosc 1980; 34:405.
McClelland JF, Luo S, Jones RW, Seaverson LM. In: Bicanic D, ed. Photoacoustic
and Photothermal Phenomena III. Berlin: Springer-Verlag, 1992:113 124.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
267
37. MacLaurin P, Crabb NC, Wells I, Worsfold PJ, Coombs D. Anal Chem 1996;
68:1116.
38. Brittain HG. Unpublished Results for infrared spectra obtained at a resolution of
2 cm 1, using a Shimadzu model 8400 Fourier-transform infrared spectrometer,
and sampled against the ZnSe crystal of a Pike MIRacleTM single reflection horizontal ATR sampling accessory.
39. Raghavan K, Dwivedi A, Campbell GC, Nemeth G, Hussain MA. J Pharm Biomed
Anal 1994; 12:777.
40. Brittain HG, Morris KR, Bugay DE, Thakur DE, Serajuddin ATM. J Pharm Biomed
Anal 1994; 11:1063.
41. Stephenson GA, Borchardt TB, Byrn SR, Bowyer J, Bunnell CA, Snorek SV, Yu L.
J Pharm Sci 1995; 84:1385.
42. Dash AK, Tyle P. J Pharm Sci 1996; 85:1123.
43. McMahon LE, Timmins P, Williams AC, York P. J Pharm Sci 1996; 85:1064.
44. Suryanarayanan R, Wiedmann TS. Pharm Res 1990; 7:184.
45. Matsuda Y, Akazawa R, Teraoka R, Otsuka M. J Pharm Pharmacol 1994; 46:162.
46. Nguyen NAT, Ghosh S, Gatlin LA, Grant DJW. J Pharm Sci 1994; 83:1116.
47. Kalinkova GN, Dimitrova L. Vib Spectros 1995; 10:41.
48. Byrn SR, Gray G, Pfeiffer RR, Frye J. J Pharm Sci 1985; 74:565.
49. Kamat MS, Osawa T, DeAngelis RJ, Koyama Y, DeLuca PP. Pharm Res 1988;
5:426.
50. Langkilde FW, Sjoblom J, Tekenbergs-Hjelte L, Mrak J. J Pharm Biomed Anal
1997; 15:687.
51. Harris RK, Kenwright AM, Say BJ, Yeung RR, Fletton RA, Lancaster RW,
Hardgrove GL. Spectrochim Acta 1990; 46A:927.
52. Deeley CM, Spragg RA, Threlfall TL. Spectrochim Acta 1991; 47A:1217.
53. Christopher EA, Harris RK, Fletton RA. Solid State Nucl Magn Reson 1992; 1:93.
54. Gao P. Pharm Res 1998; 15:1425.
55. Gao P. Pharm Res 1996; 13:1095.
56. Sarver RW, Meulman PA, Bowerman DK, Havens JL. Int J Pharm 1998; 167:105.
57. Martnez-Oharriz MC, Martn C, Goni MM, Rodrguez-Espinosa MC, Tros de
Ilarduya-Apaolaza, Sanchez M. J Pharm Sci 1994; 83:174.
58. Chao RS, Vail KC. Pharm Res 1987; 4:429.
59. Harris RK, Kenwright AM, Fletton RA, Lancaster RW. Spectrochim Acta 1998;
54A:1837.
60. Stephenson GA, Stowell JG, Toma PH, Dorman DE, Greene JR, Byrn SR. J Am
Chem Soc 1994; 116:5766.
61. Variankaval NE, Jacol KI, Dinh SM. J Biomed Mat Res 1999; 44:397.
62. Bartolomei M, Ramusino MC, Ghetti P. J Pharm Biomed Anal 1997; 15:1813.
63. Doherty C, York P. Int J Pharm 1988; 47:141.
64. Lin S-Y. J Pharm Sci 1992; 81:572.
65. Brittain HG, Bogdanowich SJ, Bugay DE, DeVincentis J, Lewen G, Newman AW.
Pharm Res 1991; 8:963.
66. Raghavan K, Dwivedi A, Campbell GC, Johnston E, Levorse D, McCauley J,
Hussain M. Pharm Res 1993; 10:900.
67. Kiss A, Repasi J, Salamon Z, Novak C, Pokol G, Tomor K. J Pharm Biomed Anal
1994; 12:889.
68. Fletton RA, Lancaster RW, Harris RK, Kenwright AM, Packer KJ, Waters
DN, Yeadon A. J Chem Soc Perkin Trans II 1986; 1705.
268
Bugay and Brittain
69. Zhu H, Khankari RK, Padden BE, Munson EJ, Gleason WB, Grant DJW. J Pharm
Sci 1996; 85:1026.
70. Yan J, Giunchedi P. Boll Chim Farmaceutico 1990; 129:276.
71. Grunenberg A, Keil B, Henck J-O. Int J Pharm 1995; 118:11.
72. Lynch IR, Buxton PC, Roe JM. Anal Proc 1988; 25:305.
73. Agatonovic-Kustrin S, Tucker IG, Schmierer D. Pharm Res 1999; 16:1477.
74. Madan T, Kakkar AP. Drug Dev Ind Pharm 1994; 20:1571.
75. Carlton RA, Difeo TJ, Powner TH, Santos I, Thompson MD. J Pharm Sci 1996;
85:461.
76. Roston DA, Walters MC, Rhinebarger RR, Ferro LJ. J Pharm Biomed Anal 1993;
11:293.
77. Neville GA, Beckstead HD, Shurvell HF. J Pharm Sci 1992; 81:1141.
78. Morris KR, Newman AW, Bugay DE, Ranadive SA, Singh AK, Szyper M, Varia SA,
Brittain HG, Serajuddin ATM. Int J Pharm 1994; 108:195.
79. Hartauer KJ, Miller ES, Guillory JK. Int J Pharm 1992; 85:163.
80. Pitre` D, Stradi R. Arch Pharm (Weinheim) 1991; 324:57.
81. Fletton RA, Harris RK, Kenwright AM, Lancaster RW, Packer KJ, Sheppard N.
Spectrochim Acta 1987; 43A:1111.
82. Brand HV, Rabie RL, Funk DJ, Diaz-Acousta I, Pulay P, Lippert TK. J Phys Chem B
2002; 106:10594.
83. Yamamura S, Gotoh H, Sakamoto Y, Momose Y. Int J Pharm 2002; 241:213.
84. Alkhamis KA, Obaidat AA, Nuseirat AF. Pharm Dev Tech 2002; 7:491.
85. Caira MR, Alkhamis KA, Obaidat RM. J Pharm Sci 2004; 93:601.
86. van Tonder EC, Maleka TSP, Liebenberg W, Song M, Wurster DE, de Villiers MM.
Int J Pharm 2004; 269:417.
87. Cantera RG, Leza MG, Bachiller CM. J Pharm Sci 2002; 91:2240.
88. Caira MR, Bettinetti G, Sorrenti M. J Pharm Sci 2002; 91:467.
89. Hu T-C, Wang S-L, Chen T-F, Lin S-Y. J Pharm Sci 2002; 91:1351.
90. Ferrari ES, Davey RJ, Cross WI, Gillon AL, Towler CS. Cryst Growth Design
2003; 3:53.
91. Fujimoto D, Tamura R, Lepp Z, Takahashi H, Ushio T. Cryst Growth Design
2003; 3:973.
92. Wang S-L, Lin S-Y, Wei Y-S. Chem Pharm Bull 2002; 50:153.
93. Lin S-Y, Chien J-L. Pharm Res 2003; 20:1926.
94. Fevotte G. Int J Pharm 2002; 241:263.
95. Feng L, Berglund KA. Cryst Growth Design 2002; 2:449.
96. Profir VM, Furusjo E, Danielsson L-G, Rasmuson AC. Cryst Growth Design 2002;
2:273.
97. Fujiwara M, Chow PS, Ma DL, Braatz RD. Cryst Growth Design 2002; 2:363.
98. Doki N, Seki H, Takano K, Asatani H, Yokota M, Kubota N. Cryst Growth Design
2004; 4:949.
100. Moribe K, Yonemochi E, Oguchi T, Nakai Y, Yamamoto K. Chem Pharm Bull 1995;
43:666.
101. Mura P, Cirri M, Faucci MT, Gines-Dorago JM, Bettinetti GP. J Pharm Biomed Anal
2002; 30:227.
102. Chawla G, Bansal AK. Pharm Dev Tech 2004; 9:419.
269
103. Teraoka R, Otsuka M, Matsuda Y. Int J Pharm 2004; 286:1.

104. Max J-J, Chapados C. J Phys Chem A 2002; 106:6452.
105. Meade EM. Sodium Hydrogen Divalproate Oligomer, United States Patent
4,988,731, issued January 29, 1991; United States Patent 5,212,326, issued May
18, 1993.
106. Guillory JK, Hwang SC, Lach JL. J Pharm Sci 1969; 58:301.
107. Serajuddin ATM, Thakur AB, Ghoshal RN, Fakes MG, Ranadive SA, Morris KR,
Varia SA. J Pharm Sci 1999; 88:696.
108. Yang R. Schulman, Zavala PJ. Anal Chim Acta 2003; 481:155.
109. Wu W-H, Chin T-F, Lach JL. J Pharm Sci 1970; 59:1234.
110. Blaug SM, Huang W-T. J Pharm Sci 1972; 61:1770.
111. Hartauer KJ, Guillory JK. Drug Dev Indust Pharm 1991; 17:617.
112. Yan B, Kumaravel G. Tetrahedron 1996; 52:843.
113. Yan B, Kumaravel G, Anjaria H, Wu A, Petter RC, Jewell CF, Wareing JR. J Org
Chem 1995; 60:5736.
114. Manning MC, Patel K, Borchardt RT. Pharm Res 1989; 6:903.
115. Lai MC, Topp EM. J Pharm Sci 1999; 88:489.
116. Costantino HR, Langer R, Klibanov AM. J Pharm Sci 1994; 83:1662.
117. Byler DM, Susi H. Biopolymers 1986; 25:469.
118. Sarver RW, Krueger WC. Anal Biochem 1991; 199:61.
119. Harris PI, Chapman D. Biopolymers (Peptide Science) 1995; 37:251.
120. Venyaminov S, Kalnin NN. Biopolymers 1990; 30:1259.
121. Dong A, Prestrelski SJ, Allison SD, Carpenter JF. J Pharm Sci 1995; 84:415.
122. Huang P, Dong A, Caughey WS. J Pharm Sci 1995; 84:387.
123. Prestrelski SJ, Arakawa T, Carpenter JF. Arch Biochem Biophys 1993; 303:465.
124. Carpenter JF, Crowe JH. Biochem 1989; 28:3916.
125. Souillac PO, Middaugh CR, Rytting JH. Int J Pharm 2002; 235:207.
126. Prestrelski SJ, Pikal KA, Arakawa T. Pharm Res 1995; 12:1250.
127. Krishnan K, Powell JR, Hill SL. Infrared microimaging. In: Humecki H, ed. Practical
Guide to Infrared Microspectroscopy. New York: Marcel Dekker, 1995: 85 110.
128. Bugay DE, Findlay WP. Pharmaceutical Excipients: Characterization by IR, Raman,
and NMR Spectroscopy. New York: Marcel Dekker, 1999.
9
Raman Spectroscopy
David E. Bugay
Harry G. Brittain
INTRODUCTION
Raman spectroscopy is a form of vibrational spectroscopy that is finding an
ever-increasing degree of use in pharmaceutical investigations. Applications
include chemical structure elucidation (1), routine chemical identification (2),
and characterization of materials in the solid (3 5). Raman spectroscopy is
also applicable to drug product characterization within drug delivery devices
(6), contaminant analysis, drug excipient interaction, and problem solving.
Analysis can be performed on virtually any type of sample, such as single crystals, bulk materials, slurries, creams, particulates, films, solutions (aqueous and
organic), oils, gas-phase samples, and process streams (the latter through the
use of fiber optic probes). Furthermore, Raman spectroscopy is usual typically
nondestructive in nature, enabling the analyst to recover the sample for further
characterization.
When a compound is irradiated with monochromatic radiation, a certain
amount of the incident light will be scattered by the molecule. Of the scattered
radiation, the majority of the photons are scattered at the same frequency as
the frequency of the incident radiation frequency. As described in the section,
Raman Spectroscopy of the Fundamental Modes of Polyatomic Molecules of
chapter 7, this form of scattering has been termed elastic, or Rayleigh, scattering.
271
272
Bugay and Brittain
Figure 1 Comparison of the diffuse reflectance infrared absorption spectrum (lower

trace) and the FT-Raman spectrum (upper trace) of an active pharmaceutical ingredient.
If the scattered radiation is passed into a spectrometer, a strong Rayleigh line is

detected at the unmodified frequency of radiation used to excite the sample.
As described previously, a very small percentage of photons will be scattered at frequencies higher and lower than the frequency of the Rayleigh line,
with the differences between the incident frequency of radiation and the
shifted frequencies corresponding to the frequency of molecular vibrations of
the molecules in the sample. As both infrared (IR) and Raman spectroscopies
are based on vibrational motions, it is not surprising that a Raman spectrum
and an IR absorption spectrum for a given compound would bear a strong
degree of resemblance. Owing to the different origins of each process, differences
in the character of functional groups cause the techniques to be complementary
and not duplicative. This is illustrated in Figure 1 where the IR and Raman
spectra of an active pharmaceutical ingredient (API) are shown.
SOME FUNDAMENTAL PRINCIPLES OF THE RAMAN EFFECT
In the previous discussion (chap. 7) a purely classical picture of vibrational
Raman scattering was developed, in which the oscillating dipole moment
induced by the electric field of the incident light was modulated by molecular
vibrational motion, resulting in scattering at frequencies higher and lower than
the frequency of the incident energy. Nevertheless, accurate descriptions of
any type of molecular spectroscopy must originate in the quantum theory,
because the states involved in such processes are derived from the quantization
of energy levels (7,8).
Raman Spectroscopy
273
Nonresonant Raman Scattering

Placzek (9) developed a quantum mechanical model for the Raman effect based
on the classical polarizability model, assuming that under appropriate conditions
a Raman transition can be treated in a manner analogous to an electric dipole
absorption. For those molecular systems having nondegenerate ground electronic states (the usual instance for organic molecules of pharmaceutical interest)
that are irradiated with electromagnetic radiation incapable of causing a transition to an excited state, the vibrational Raman intensities are given approximately by the vibrational matrix elements of the electronic polarizability.
Through the use of the polarizability theory, one may understand the intensity of
vibrational Raman scattering through a correlation with molecular polarizability
in a manner that parallels that of IR spectroscopy (where the intensity of absorption is related to the vibrational-coordinate dependence of the dipole moment).
Because in quantum theory the frequency shifts of Raman bands are determined by the energy differences of the vibrational states between which the transitions take place, it is therefore necessary to calculate the transition probabilities
of these transitions. As long as the atoms in a polyatomic molecule vibrate with
low degrees of amplitude, the motion as a whole can be taken as being a superimposition of numerous harmonic oscillators. As long as the harmonic oscillator
model applies, the selection rule of Dv + 1 applies, and therefore only molecular
transitions characterized by a change in polarizability will exhibit an allowed
Raman transition.
As developed previously in chapter 7 (section on Raman Spectroscopy of
the Fundamental Modes of Polyatomic Molecules), the family of scattering
bands observed at frequencies of (n0 nv) are known as the Stokes lines, and
the family of scattering bands observed at frequencies of (n0 nv) are known
as the anti-Stokes lines. According to the classical theory, the Stokes lines
should have the same intensity as the anti-Stokes lines, but in fact the antiStokes lines are much weaker in intensity than the Stokes lines. This situation
arises because the number of molecules occupying an initial state of vi 1 can
be only e 2hcn/kT times the number of molecules in the ground vi 0 state. The
intensity ratio of anti-Stokes to the corresponding Stokes likes is found to be
in agreement with the Boltzmann factor.
The intensity of scattered light depends on the magnitude of the induced
dipole moment, P, which for a transition between cN and cM is given by the
matrix formed from the integrals:
c N PcM d t
The time-independent part of Equation (1) is given by:
P0 NM cN P0 cM dt
(1)
(2)
274
Bugay and Brittain
where P 0 is the amplitude of P, and the intensity of the cN $ cM transition is

proportional to the square of [P 0]NM.
Knowing that the magnitude of the induced dipole moment equals the
product of the electric field strength and the polarizability, the components of
[P 0]NM are given by:
P0X NM EX0 cN aXX cM dt EY0 cN aXY cM dt EZ0 cN aXZ cM dt (3)
P0Y NM EX0 cN aYX cM dt EY0 cN aYY cM dt EZ0 cN aYZ cM dt (4)

P0Z NM
EX0
cN aZX cM dt
EY0
cN aZY cM dt
EZ0
cN aZZ cM dt (5)
In Equations (3), (4), and (5), EX0, EY0, and EZ0 are the components of the
amplitudes of the incident electromagnetic radiations along each Cartesian direction, and the various integrals
NM
aXX cN aXX cM dt
(6)
(7)
aXY NM cN aXY cM dt
(8)
aXZ NM cN aXZ cM dt
are the matrix elements of the polarizability tensor.
The diagonal matrix elements ([aXX]NM, [aYY]NM, and [aZZ]NM) correspond
to Rayleigh scattering, whereas the off diagonal elements ([aXY]NM, [aXZ]NM,
[aYZ]NM, [aYX]NM, [aZX]NM, and [aZY]NM) correspond to Raman scattering
associated with the cN $ cM transition. The cN $ cM transition will be
allowed Raman scattering if at least one of the six off-diagonal elements does
not equal zero.
As would be anticipated, molecular symmetry-based group theory can be
used to determine whether one of the off-diagonal elements will have a nonzero magnitude. It has been established before that as long as the ground state
of a molecule is not degenerate, then that state will be characterized by all
vibrational wave functions being in their v 0 level, and this state must transform as the totally symmetric representation in the molecular point group. To
achieve a symmetric end result (and a nonzero magnitude for an off-diagonal
element), the direct product of the representation for the excited state with the
representation of one of the components of the polarizability must contain the
totally symmetric representation of the point group. In other words, the representation of the excited state wave function must be the same as that of the
representation for one of the elements of the polarizability operator (10 13).
As discussed in a previous chapter, the elements of the polarizability tensor
Raman Spectroscopy
275
will transform as the squares and binary products of coordinates listed in area IV
of the point group character table.
Detailed consideration of the various character tables leads to the Exclusivity Rule. For molecules containing a center of symmetry, transitions that are
allowed in the IR absorption spectrum will be forbidden in the Raman spectrum,
and transitions that are allowed in the Raman spectrum will be forbidden in the IR
absorption spectrum. This rule is of great importance for the study of molecules
characterized by high symmetry (such as inorganic coordination complexes), but
rarely is of consequence in the low-symmetry world of organic molecules having
pharmaceutical interest.
Resonance Raman Scattering
In the nonresonant condition, the energy of the incident photons is not equal to
that of an allowed electronic transition, and therefore the wavelength of the incident light does not fall within the wavelength range of an electronic absorption
band of the molecule. A completely different situation arises when the wavelength of the incident photons lies within the absorption band of an electric
dipole-allowed electronic transition.
What has come to be termed the resonance Raman effect results from the
transition from the ground state into an excited vibronic state, which is subsequently accompanied by immediate relaxation into a vibrational level of the
ground state. What makes this resonance Raman process different from that of
ordinary photoluminescence is that the transition back to a vibronic level of the
ground state is not preceded by prior relaxation of the system to the lowest
vibrational level of the excited state. This distinction is illustrated in the
Jablonski-type diagram shown in Figure 2. The resonance Raman emission
process is essentially instantaneous, and the resulting spectra consist of narrow
bands whose energy difference from the exciting light is equal to the energy of
a ground state vibrational mode of the molecule.
The consequence of resonance scattering is that the contribution to the scattering from those electronic state(s) with transition energies equal to that of the
incident photons (i.e., the states with which the photons are in resonance)
becomes very large and dominates all others. This results in a selective enhancement in the intensities of vibrational transitions involving motions of those atoms
associated with the molecular orbitals that define the ground and excited states of
the electronic transition. The particular vibrational modes enhanced are those that
are coupled to the electronic transition responsible for the absorption band. In practical terms, this means that two classes of vibrational modes will produce intense
resonance-enhanced spectra, namely totally symmetric vibrations, and those
nontotally symmetric vibrations that vibronically couple two electronic states.
Because the molecular orbitals defining an electronic transition are often
somewhat localized to a specific functional grouping in a molecule, the resonance
Raman effect can provide a way to selectively enhance the Raman bands due to
276
Bugay and Brittain
Figure 2 Jablonski-type energy level diagram illustrating the resonance Raman process.
Excitation is drawn as the S0(v 0) ! S1(v 3) transition, and the resonance Raman
transition is the S1(v 3) ! S0(v 2) transition.
vibrations of the chromophore associated with the states of the functional group.
For instance, in heme-proteins, it has been shown that the resonance Raman
bands are solely due to vibrational modes of the tetrapyrrole chromophore and
that none of the vibrational bands associated with the protein gain resonance
enhancement. The utilities of ultraviolet (UV) resonance Raman spectroscopy
for analytical, physical, and biophysical applications have been described (14).
It has been pointed out that the sensitivity of resonance Raman spectroscopy
to only chromophore vibrational modes may be considered as being either strength
or weakness (15). The strength of the method is that the observed spectra are
greatly simplified, and a series of molecules containing slightly different chromophores will yield spectra that are easily distinguished. The main limitation is that if
a series of molecules contain the same chromophore but with a variety of aliphatic
side chains, the resonance Raman spectra would be nearly identical.
Surface-Enhanced Raman Scattering
In 1974, Fleischmann (16) observed intense Raman scattering from pyridine that
had been absorbed onto a roughened silver electrode. It was recognized that the
Raman Spectroscopy
277
large Raman intensities could not be merely a function of the number of scattering units and that the intensity enhancement had to have its origins in an interaction between the surface and the absorbed scattering molecules (17,18).
Subsequently it was found that the same degree of increase in the cross-section
of Raman scattering could be obtained if the molecules were allowed to absorb
onto the surfaces of colloidally dispersed silver particles (19). The methodology
has been reviewed (20), and questions related to scattering from single particles,
fractal clusters, and surfaces have been discussed (21).
As discussed earlier, the intensity of Raman scattering is determined by the
magnitude of the induced dipole moment, which equals the product of the electric
field strength and the polarizability. It follows that the enhancement of scattering
must be associated either with an increase in the polarizability of the molecule or
with an increase in the electric field experienced by the molecule. The most significant mechanism that could produce an increase in the polarizability would
entail some type of charge transfer interaction between the metal substrate and
the adsorbed molecule, or even formation of a chemical bond between the two.
However, it is now generally agreed that most of the surface enhancement is
associated with an amplification of the strength of the local electromagnetic field.
The most effective surface-enhancing systems consist of small metal particles or of rough surfaces of conductive materials. Light impinging on these
types of surfaces can excite conduction electrons in the metal, generating polarization in the substrate, and thus the electromagnetic field at the interior of the
particle becomes significantly larger than that of the applied field. The magnitude
of the enhancement is given by Lorenz-Mie theory and depends on the wavelength-dependent refractive index of the substrate, the size and shape of the
rough features or small particles, and the incident wavelength.
The theory predicts two key requirements in order to achieve a large degree
of surface enhancement of the Raman scattering. First, the imaginary component
of the refractive index of the metal should be small, and the real component
should be approximately equal to twice the wavelength of the incident light
source. Second, the frequency of the incident light should match the plasma resonance frequency of the substrate particle or the rough feature. For visible wavelength excitation sources, the first requirement leads to the conclusion that only
the noble metals would be good surface-enhancing substrates. The ability to
exploit the surface enhancement effect for these substrates is limited only by
instrumental considerations and by any technical difficulties associated with
the preparation of appropriate substrate morphologies.
The selection rules for conventional Raman spectroscopy also are applicable to surface-enhanced Raman scattering. Because the local electric field at
the surface is highest in the direction perpendicular to the surface, the vibrational
modes that involve changes in the polarizability of the absorbed molecule that are
perpendicular to the surface will be preferentially enhanced (22). In addition to
the fact that the electromagnetic field amplitude falls off rapidly with distance
from the surface, the polarizability dependence allows an investigator to
278
Bugay and Brittain
determine the orientation of the adsorbed molecule with respect to the surface
and to make some estimate of the distance between a given functional group
of the molecule to the surface.
INSTRUMENTATION FOR MEASUREMENT OF RAMAN

SPECTRA IN THE SOLID STATE
The type of systems used for the acquisition of Raman spectra will be briefly discussed, and readers seeking more detail are referred to the numerous reference
texts available (23 29).
Spectrometer Types
Although the Raman effect was discovered in 1928, the first commercial Raman
instruments did not start to appear until the early 1950s. As the laser had not yet
been invented, these instruments used elemental sources and arc lamps as the
irradiation source. In 1962, laser sources started to become available for
Raman instruments, and the first commercial laser Raman instruments appeared
in the mid-1960s. The first commercial FT-Raman instruments became available
in 1988, which was followed soon by the FT-Raman microscopy. A detailed
history of the development of Raman instrumentation is available (30).
The basic configuration and components of a dispersive Raman spectrometer are shown in Figure 3. The source of monochromatic radiation is a
laser, which could be helium-cadmium (325, 354, or 442 nm), air-cooled
argon-ion (488 or 514 nm), doubled continuous wave neodymium yttrium aluminum garnet (Nd:YAG or Nd:Y3Al5O12) (532 nm), helium-neon (633 nm), or
stabilized diode (785 nm). The stability of the laser radiation is a key attribute
of a good spectrometer, and good stability is essential for good function. Frequency stabilization of the laser under standard laboratory conditions (slight
temperature fluctuations, vibrational effects, etc.) is required. Laser lifetimes
and cost are also considerations of choice to use the laser.
One additional consideration associated with laser selection in dispersive
Raman systems concerns the use of wavelengths that could potentially generate
molecular fluorescence. As noted previously, the intensity of the Raman effect is
fairly weak, and photoluminescence can be so intense as to mask the scattered
Raman photons. If fluorescence does not pose a problem for a given sample,
lower frequency lasers can be used (532 or 514 nm) for enhanced sensitivity,
as the efficiency of Raman scatter is proportional to 1/l4 (where l is wavelength). If fluorescence is a problem when using these high-energy sources,
then lower energy sources, such as those used in Fourier transform (FT)-based
Raman spectroscopy, can be used to minimize the fluorescence effects.
In a dispersive Raman spectrometer, the sample is positioned in the laser
beam, and the scattering radiation is collected either in a 1808 (the backscattering
method) or a 908 (the right-angle method) scattering configuration. Subsequently,
a laser-line rejection filter is placed in the scattered beam path to filter out
Raman Spectroscopy
Figure 3
279
Schematic representation of a dispersive-type Raman spectrometer.
Rayleigh scattering, and the intensity of scattered light measured by a detector

positioned in the spectrometer. For dispersive systems, a charge-coupled
device (CCD) is typically utilized. Silicon CCD detectors are normally used
for Raman spectrometers in which visible wavelength lasers are used. At one
time, photomultiplier tubes were used for detection, but since the advent of
CCD detectors and their inherently better performance, photomultiplier tubes
are not normally used today. All commercial spectrometers are digitally controlled by a computer system.
The basic configuration and components of an FT-based Raman spectrometer is found in Figure 4. Advantages of an FT-Raman spectrometer are
wavelength accuracy, and the use of a near-infrared (NIR) laser that usually eliminates spurious effects due to fluorescence. In the FT-based system, a Nd:YVO4
laser (1064 nm output) is used to irradiate the sample; the sample is positioned in
the laser beam, and the scattering radiation is collected either in the 1808 backscattering or the 908 right-angle scattering configuration. By utilizing the
longer wavelength of the Nd:YVO4 laser in a FT-Raman spectrometer, fluorescence is minimized because it is unlikely that a change in electronic state
could be promoted through the use of a low-frequency laser.
The scattered photons are then passed into an interferometer with laser
line filtering. Detection of the scattered photons from systems that utilize
280
Figure 4
Bugay and Brittain
Schematic representation of a Fourier-transform type Raman spectrometer.
lasers emitting light with wavelengths greater than 1000 nm are of the singleelement type, being either high purity p-type germanium or indium/gallium/
arsenic (InGaAs) detectors. Such detectors are noisier than the CCD or PMT
detectors, but they do exhibit high quantum efficiencies. By cooling the Germanium detector to liquid nitrogen temperature (77 K), the frequency response can
be extended to 3400 cm21. Unfortunately, germanium detectors are subject to
interference by cosmic rays, and artifacts in their output may be generated.
However, corrections to these artifacts have been incorporated into hardware
and/or software methods.
Sampling and Data Acquisition
Sampling techniques for Raman spectroscopy are relatively simple because the
only requirements are that the monochromatic laser beam irradiates the sample
of interest and that the scattered radiation is focused upon the detector. Raman
spectroscopy may be performed on very small samples (samples as small as a
few nanograms). Powders do not need to be pressed into discs or diluted with
KBr as would be the case for most IR spectroscopic sampling procedures, but
instead the material merely needs to be irradiated by the laser beam. Solid
samples are often examined in stainless steel or glass sample holders that hold
approximately 25 50 mg of the sample.
Raman Spectroscopy
281
Liquid samples are typically analyzed in quartz or glass cuvettes, which

may have mirrored rear surfaces to improve the signal intensity. Water is a
good solvent for Raman studies, because the Raman spectrum of water consists
of essentially one broad, weak band at 3500 cm21. Glass does not scatter Raman
bands very efficiently, and therefore many samples (liquid and solid) can be
simply analyzed in situ in a bottle, or for convenience in a nuclear magnetic
resonance (NMR) sample tube. These spectral advantages have been used to
develop a method for the noninvasive identification of substances inside USP
vials using spectra acquired through the glass and a Raman spectral library (31).
One sample type that may pose a problem is any material that is darkly
colored. Often, these samples absorb excessive heat and burn, causing sample
and spectral degradation. Amorphous materials also have a tendency to absorb
heat in the laser beam. In order to avoid sample burning, it may be necessary
to dissipate the heat, which can be accomplished by reducing the laser power
or by using an accessory that spins the sample to avoid the irradiation of a
single point in the sample. Further reduction of the laser power can also be
accomplished through the use of a neutral density filter. Sometimes, the
sample can be diluted in KBr to aid in a reduction of sample burning.
The complete Stokes Raman spectrum spans in the range of 100
3500 cm21 and can be obtained under conditions where the intensity of Raman
scattering is directly proportional to the concentration of the scattering species
(an essential factor for quantitative analysis). Unlike IR absorption spectroscopy,
where the strong absorptivity of the bands enables analytes to be detected down
to concentration of approximately 0.01%, the weakness of the Raman effect leads
to the requirement that an analyte should be present at a concentration of at least
1% for accurate assessments.
As discussed previously, the presence of fluorescence, typically due to
additives in the glass sample tubes or impurities within the sample of interest,
is problematic for Raman studies. Data massaging techniques can sometimes
be used to blank out fluorescence or alternatively one may use photobleaching
as another means of suppressing fluorescence. The latter technique involves irradiating the sample for a prolonged period of time (seconds to hours) with the
laser. During this time, the fluorescence may decrease due to the destruction of
the fluorescing component due to the prolonged exposure to the laser irradiation,
and then the spectrum is acquired after photobleaching is complete.
In any Raman study, the possibility that the laser radiation may change the
component of interest must always be considered. For example, a solvated crystalline material may desolvate upon exposure to laser radiation. It is appropriate
and prudent to establish the integrity of the sample by comparing spectra acquired
with short data acquisition times to those acquired with long acquisition times.
Additionally, ancillary techniques, such as X-ray powder diffraction, thermal
analysis, or IR absorption spectroscopy, could be used to learn if the laser
irradiation caused changes in the quality of the sample as a result of the
conduct of the Raman spectroscopic analysis.
282
Bugay and Brittain
Variable temperature studies in Raman spectroscopy can be used to study

the mechanisms of thermally induced reactions. Because a Raman spectrum typically covers a wavelength range that extends beyond the usual mid-IR region
down to the far-IR region, information about the lattice vibrations of crystalline
organic compounds can be readily acquired. By varying the temperature of a
sample, the lattice energies of the compound become changed, which then facilitates interpretation of the nature of the crystal lattice. In addition, information on
crystal form changes and the nature of solvatomorph association, usually
obtained with other methods of thermal analysis or thermospectroscopic
studies, can be obtained through Raman investigations at variable temperatures.
Use of the Raman microprobe to measure spectra from small amounts of
material is of considerable interest. Utilizing the microscope, the Raman
scattered photons are collected in a 1808 backscattering configuration that
allows the operator to optically view the sample, focus the incident radiation,
and subsequently collect the Raman spectrum. Most commercial Raman microscope systems utilize confocal microscopy to increase axial resolution (along the
z-axis). Confocal points are defined as the point source, the in-focus sample
location, and the focused image of the sample point. Axial resolution, defined
as the distance away from the focal plane in which the Raman intensity from
the sample decreases to 50% of the in-focus intensity, can be approximated
from the numerical aperture used in the microscope. When utilizing a 0.95
numerical aperture (NA) objective on a confocal microscope system, the axial
resolution is proportional to the square of the NA, or in this case 0.9025 mm.
An additional advantage of Raman microscopy is that of spatial resolution,
as opposed to axial resolution. The spatial resolution (in the xy-plane) is dependent upon the NA of the collecting objective and the wavelength of the laser radiation. Larger NA values and shorter wavelengths provide higher degrees of
spatial resolution, which can be as low as 1 mm. One consideration is that
because the high intensity, monochromatic, laser radiation is focused upon a
small area of the sample, any sample degradation by the laser must be monitored.
With this caveat in mind, the Raman microprobe has been found to be ideal for
investigating polymorphism in single crystal samples, particulate contamination,
and small quantities of analytes. Using an apparatus similar to that used for IR
microscopy, variable temperature studies can be performed with a Raman
microprobe.
Fiber optics has been used in Raman spectroscopy since the early 1980s,
facilitating the analysis of solids and liquids by means of an arrangement of
optical fibers on the end of a probe. Today, much of the research being conducted
in the use of fiber optics in FT-Raman spectroscopy centers around fiber type and
fiber-bundle design. The number, type, and arrangement of the fibers in a fiber
bundle are key factors that are varied to produce fiber bundles for different applications. Fiber systems include single fibers (where the laser excitation and collected scattered radiation travel along the same fiber) and multifibers (where
laser excitation is transmitted along one or multiple fibers and the scattered
Raman Spectroscopy
283
radiation is transmitted to the detector along different fibers). The arrangements

of the fibers in a multifiber system can also vary. For example, one arrangement is
where one excitation fiber is surrounded by several collection fibers, and another
example is where several excitation and collection fibers are randomly mixed in a
bundle. The greatest single advantage of the use of fiber optics in Raman spectroscopy is the ability to sample analytes in a remote location, where fiber
optics can link the spectrometer to the sample, typically in a distances of tens
of meters. Common applications include monitoring process streams or hazardous reactions.
GROUP FREQUENCY CORRELATIONS IN RAMAN
SPECTROSCOPY AND EFFECTS OF THE SOLID STATE
As discussed previously, the concept that a given functional group would consistently exhibit vibrational frequencies within a narrow range is due to the fact that
the frequency of a vibrational mode in a molecule is determined primarily by the
force constant of the bond or bonds and by the reduced masses of the nuclei
involved in the motion. In a complex molecule, the characteristic frequency
range is modified by the interplay of many other factors, the resultant of which
determines the exact energy of the absorption band. Some of these factors are
associated with intramolecular origins, relating to changes in the molecular geometry, the masses of substituent groups, the occurrence of mechanical coupling
between one vibration and another, the effects of steric strain, and the electrical
influences of the substituent groups operating either along the bonds or across
space. Other factors that are of great interest to solid-state spectroscopy
concern the external environment of the vibrating group, including such
effects as details associated with the crystal form or the influence of hydrogen
bonding.
Despite these limitations, there is no doubt that correlation tables of group
frequencies continue to be of great value in vibrational spectroscopy. A number of
such compilations are available that summarize the frequencies associated with
the IR absorption of various functional groups (32 35), but it goes without
saying that the same principles of vibrational group frequencies must apply
equally to the scattering bands in a Raman spectrum. In fact, such a compilation
has been published specific to group frequencies in Raman spectroscopy (36).
Although excitation of a given vibrational mode must give rise to IR absorption and Raman scattering bands having identical frequencies, the fact that each
effect originates from fundamentally different principles causes the observation
of large differences in relative intensities. This is illustrated in Figure 5, where
the intensities of the C55O stretching, methyl deformation, and C22Cl stretching
bands in the IR absorption and Raman spectra of 2,5-dichloroacetophenone are
seen to exhibit very different intensities (37). Most significant is the aromatic
C55C stretching mode that appears prominently in the Raman spectrum but
which exhibits only moderate intensity in the IR absorption spectrum.
284
Bugay and Brittain
Figure 5 Comparison of the Raman (lower trace) and infrared absorption (upper trace)
spectra in the fingerprint region of 2,5-dichloroacetophenone, illustrating the difference in
relative intensities for each spectral type. Source: Adapted from Ref. 37.
As previously discussed, the vibrational modes that exhibit the strongest

intensities in a Raman spectrum are those that are associated with functional
groups that are characterized by high degrees of polarizability, because such
groups will be able to exhibit significant changes in the induced dipole
moment upon irradiation. Such functional groups include, for example, the
C22S, S22S, C22C, N55N, and C55C groups. Grasselli et al. have provided a
useful table summarizing the characteristic wavenumbers for a variety of
organic functional groups and have summarized the Raman and IR absorption
intensities of these (38); a selection of these can be found in Table 1.
Raman Spectroscopy
285
Table 1 Characteristic Frequencies of Some Organic Functional Groups and the

Corresponding Raman and Infrared Absorption Intensities
Functional group
Frequency region
(cm21)
2
2O2
2H
2
2N2
2H
;
;C2
2H
5
5C2
2H
2
2C2
2H
2
2S2
2H
2
2C5
5O
2
2C5
5C2
2
3000 3650
3300 3500
3300
3000 3100
2800 3000
2550 2600
1680 1820
1500 1900
2
2C5
5N2
2
2
2C5
5S
2
2C2
2O2
2C2
2
2
2O2
2O2
2
2
2S2
2S2
2
2
2C2
2Cl
2
2C2
2Br
1610 1680
1000 1250
1060 1150
845 900
430 550
550 800
500 700
Raman intensity
Weak
Medium
Weak
Strong
Strong
Strong
Strong weak
Very strong
medium
Strong
Strong
Weak
Strong
Strong
Strong
Strong
Infrared absorption
intensity
Strong
Medium
Strong
Medium
Strong
Weak
Very strong
Weak absent
Medium
Weak
Strong
Absent
Absent
Strong
Strong
Because of the polarizable nature of sulfur, Raman spectroscopy is particularly well suited for studies of the nature of sulfur bonding. Although the S22H
stretching mode around 2580 cm21 is weak in the IR absorption spectrum, it is
strongly observed in the Raman spectrum. Also prominent in the Raman
spectra of mercaptans are the C22S stretching modes (observed between 570
and 785 cm21) and the symmetric and antisymmetric C22S22C modes (also
observed between 570 and 800 cm21).
For most molecules in the gas state, a molecular vibration studied by
Raman (or IR absorption, for that matter) spectroscopy can be regarded as
being essentially free from the influence of other molecules. In the liquid state,
however, other molecules and the bulk solvent, which may influence its frequency either through the change they have produced in the dielectric constant
of the medium or through molecular association, surround the vibrating group.
For example, for carboxylic acids dissolved in a solvent, formation of dimers
through hydrogen bonding leads to shifts in carbonyl and hydroxyl frequencies
by as much as 50 cm21 and 500 cm21 relative to the vapor state, respectively.
The concept of group frequencies still has validity, however, but the range for
a given vibrational type becomes wider in the condensed phase.
The situation existing in a crystalline solid is different relative to either the
vapor or liquid phases. Frequency shifts can occur due to a further increase in
intermolecular forces, but these are usually relatively small unless the vibrations
are involved in hydrogen bonding. Owing to the increased order of a crystalline
286
Bugay and Brittain
structure, it is frequently observed that absorption bands associated with some

bonds will not be observed in the spectrum. This effect arises from the fact
that rotational isomerism cannot take place in an oriented crystal, which causes
absorption bands associated with any other conformer to disappear.
In other instances, additional vibrational bands will appear. Within the rigid
crystal environment, strong intermolecular forces exist so that the individual
group vibrations become strongly influenced by the nature of the unit cell. In suitable circumstances, a process analogous to linear combination takes place, where
in-phase and out-of-phase vibrations of the same groups in different molecules
are established, leading to splitting of the original vibrational band into two components. Alterations in the crystal lattice, associated with the existence of multiple polymorphic forms, can result in spectral changes due to the alterations,
which take place in the immediate environments of the vibrating groups. This
phenomenon is particularly wellsuited for Raman spectroscopic investigations,
because Raman can provide information in the low-frequency region where the
differences between the different polymorphs are appreciable.
APPLICATIONS OF RAMAN SPECTROSCOPY TO AREAS OF
The widespread use of Raman spectroscopy as a tool to solve problems of
pharmaceutical interest required the advent of the FT method, because prior to
that the technique was viewed as being too much of a research method and not
practical for use on a more routine basis. Part of the problem was that too
many compounds exhibited fluorescence when irradiated in the visible region
of the spectrum (i.e., argon-ion lasers operating at 488 or 514 nm). However,
because the technique is based on light scattering, samples can be investigated
in a noninvasive and nondestructive manner, and the analytes do not have to
be diluted with inert filler before being studied. A number of detailed reviews
are available regarding the use of Raman spectroscopy in areas of pharmaceutical
interest (4,5,39 41).
Qualitative Identification of Pharmaceutical Compounds
The identification of a substance on the basis of its Raman spectrum is completely
analogous to the identification of a compound on the basis of its characteristic IR
absorption spectrum. For example, the fingerprint region of the Fourier-transform
Raman spectrum of paracetamol is shown in Figure 6, with its diagnostic peaks
being marked. In practice, an analyst would choose five to six of the most intense
lines and enter these in a search library. Then, when a new spectrum of a compound purported to be paracetamol was obtained, one could match the observed
peaks with the library peaks to make the identification. Because the number of
Raman spectral libraries lags considerably behind the number of IR spectral
libraries, many workers compile their own libraries for identification purposes.
Raman Spectroscopy
287
Figure 6 Fingerprint region of the Fourier-transform Raman spectrum of paracetamol,

with the diagnostic peaks marked.
However, a compilation of spectra obtained on a wide variety of pharmaceutical

excipients is available (42).
The Raman and IR absorption spectra of a wide range of active pharmaceutical ingredients have been reported, and the corresponding spectra critically
compared (1). In all the compounds studied, ring breathing modes of monosubstituted phenyl groups, the 22C22C22C22 symmetric skeletal mode of tricyclic fused ring systems, the N55O symmetric stretch of aromatic nitro
groups, and the symmetric stretching vibrations of various double bonds were
found to be most prominent and useful for identification purposes. It was noted
that although the polar regions of drug substances could be easily studied by
IR absorption spectroscopy, the regions that are less polar and have little
dipole character (aliphatic chains, aromatic carboxylic groups, and heterocyclic
groups) are more amenable to study by Raman spectroscopy.
The complementarity between IR absorption and Raman spectroscopies
was demonstrated through a study of benzocaine hydrochloride (43). In addition
to the static spectra, measurements were also made at a variety of temperatures
and the results of all the spectral studies correlated with the results of semiempirical calculations. Because the amino group was known to play an important role
in the reactivity of the molecule, a detailed analysis of its torsional and wagging
288
Bugay and Brittain
modes was carried out. The band assignments thus obtained turned out to be the
most definitive of their type.
The IR absorption and Raman spectra of four compounds structurally
related to diazepam have been reported (44). A number of spectral characteristics
emerged that facilitated the distinction between diazepam and delorazepam, fludiazepam, flurazepam, and tetrazepam. All compounds exhibited a strong C55N
stretching mode near 1610 cm21, and the 1,4-benzo-diazepine nucleus was
characterized by a strong band at approximately 1160 cm21. It was concluded
that owing to the close similarities in their structures, a combination of both IR
absorption and Raman spectroscopies was required to obtain unequivocal differentiation of individual benzodiazepines.
The spectral sensitization associated with surface-enhanced Raman scattering has been used to develop an extremely sensitive identification method for
papaverine (45). Employing a sequential injection system, an automated apparatus was developed that enabled the reliable and rapid recording of data. The colloidal silver substrate (with or without analyte) was prepared in a cuvette by the
reduction of silver nitrate by hydroxylamine. It was shown that the Raman spectrum of papaverine, obtained by either surface-enhanced or conventional means,
was the same.
A FT Raman method has been developed to identify the two active components in formulated capsules (2). A spectral region in the Raman spectrum
of the formulated product was found where Raman bands unique to the active
ingredients, uracil and tegafur, were observed, and the presence of these bands
enabled a confirmation to be made that both components were present in a
given formulated product.
Raman microspectroscopy is well suited for in situ analysis of contaminants that may be found in pharmaceutical process materials. Because the technique is nondestructive in its nature, the analytes may be subjected to additional
confirmatory experiments, such as energy dispersive X-ray analysis or IR microspectroscopy. A positive aspect for contaminant analysis by Raman spectroscopy
is the favorable axial and spatial resolutions of the technique as compared with
IR microspectroscopy. In general, IR microspectroscopy is diffraction-limited
for investigating samples that are typically larger in size than 5 mm. When
working with a 0.95 NA objective, spatial resolutions as small as 1 mm can be
achieved with a Raman microscope, thus permitting the analysis of very small
contaminating particulates.
Raman Spectral Studies of Polymorphic Systems
When the differing crystal structures of polymorphs or solvatomorphs translates
into a perturbation of the pattern of molecular vibrations, then the techniques of
vibrational spectroscopy (such as Raman scattering) can be used to study the
solids. As a result, Raman spectroscopy has found widespread use for the qualitative and quantitative characterization of different solid-state forms of
Raman Spectroscopy
289
compounds having pharmaceutical interest. For example, the two polymorphs of

famotidine exhibit substantially different Raman spectra (46) in the fingerprint
region (Fig. 7) and can be readily differentiated from each other on the basis
of their characteristic spectra.
An advantage of using Raman spectroscopy for polymorphic investigations
is the ease of the technique and the ability to measure low-frequency vibrations
(500 50 cm21) that are associated with lattice vibrational modes. In many
cases, two different solid-state forms of a pharmaceutical entity will display spectral differences in the low-frequency region of the Raman spectrum. These
spectral differences will also be observed when comparing the spectra of
crystalline and amorphous states of the same substance, which can be used to
develop a method to determine the degree of crystallinity in a sample.
Raman spectroscopic studies were found to play a valuable role during
studies of the two polymorphs formed by losartan (47), and a consideration of
the spectral trends serves to illustrate the magnitude of a vibrational shift due
to crystal structure effects. The C22H out-of-plane motion in the biphenyl ring
of Form-I is observed at 763 cm21, whereas in Form-II the band is split into
two bands at 710 and 760 cm21. The ring-breathing mode associated with the
imidazole ring was observed at 803 cm21 in Form-II but is split into bands at
Figure 7 Raman spectra obtained within the fingerprint region for famotidine Form-A
(solid trace) and Form-B (dashed trace). Source: Adapted from Ref. 46.
290
Bugay and Brittain
807 and 819 cm21 in Form-I. Significant differences in the bands associated with
lattice modes were also reported.
The Raman spectra of two polymorphs of fluconazole were found to permit
the ready differentiation between the crystal forms (48). Numerous differences in
vibrational band energies are evident in the fingerprint and lattice regions, and
even the high-frequency spectral region contains well-resolved spectral features
that permit an easy characterization of the structural differences between the two
systems. Thirteen samples of spironolactone, obtained from different sources,
were evaluated by Raman spectroscopy, and four polymorphic forms of the
drug substance were identified in these (49). When IR absorption spectra were
obtained for samples in compressed KBr pellets, no differences attributable to
polymorphism could be detected, indicating interconversion during processing
of the sample. However, when the samples were processed using a nondestructive method, the existence of one (or sometimes more) of the polymorphs
could be readily detected.
It should not be automatically assumed that the Raman spectra of different
polymorphs will automatically exhibit significant differences. Stavudine is
known to exist in two anhydrous polymorphs and one hydrated form (50), and
the Raman spectra of these different forms are shown in Figure 8. Even
Figure 8
Raman spectra of the three forms of stavudine.
Raman Spectroscopy
291
through the crystal structures of the three forms (as evidenced by their X-ray
powder diffraction patterns) were found to be quite different, virtually no spectral
differences were noted among the Raman spectra, except for some bands in the
C22H stretching region. A similar finding was reached regarding the Raman
spectra of the polymorphs of roxifiban (51).
The ability of Raman spectroscopy to interrogate the surface of a sample
in a nondestructive manner was used to study the phase transformation of
cilostazol in its solid dose form (52). Exposure to dissolution media cause
the metastable forms of the drug substance to be converted into the stable
phase, and these changes can be spectroscopically followed without perturbing
the sample.
In those instances where the structural differences between the members
of a polymorphic system are sufficiently different so that the corresponding
Raman spectra are diagnostic, the degree of spectral simplification associated
with Raman data permits a more rapid evaluation of new chemical entities.
This feature has been successfully exploited in the cases of 2-f[4-(4-fluorophenoxy)phenyl]-methyleneg-hydrazine carboxamide (53), rac-5,6-diisobutyryloxy-2-methylamine-1,2,3,4-tetrahydro-naphthalene
hydrochloride
(54),
3-( p-thioanisoyl)-1,2,2-trimethyl-cyclopentanecarboxylic acid (55), and trans3,4-dichloro-N-methyl-N-[1,2,3,4-tetrahydro-5-methoxy-2-(pyrrolidin-1-yl)]naphth-1-ylbenzeneacetamide (56).
Raman spectroscopy has been shown to be a suitable technique for crystal
product triage in the high throughput method for polymorph screening (57). In
this work, Raman spectroscopy was used to measure the amount of dissolved
carbamazepine and then to determine whether Form-I or Form-III crystallized
upon thermal cooling of the solutions. The advantage inherent in use of the
Raman technique is that the work can be performed in 35- or 100-mL fluid
samples, where mixing was effected by diffusion. The advantage of miniaturized
techniques over the conventional macrotechniques is that equilibrium can be
attained much faster using less drug substance, facilitating higher degrees of
throughput.
Raman Spectral Studies of Solvatomorphic Systems
The characterization of solvates and hydrates by Raman spectroscopy also
requires that the differing crystal structures of the crystal forms cause a perturbation of the pattern of molecular vibrations. In many cases, the degree of crystallographic difference is sufficiently large, but in other instances the solvent of
crystallization only affects a small number of vibrations. Naproxen sodium has
been shown to crystallize in an anhydrate, a monohydrate, and a dihydrate
crystal form, and methods have been published regarding a number of their physical characteristics (58). The anhydrate and monohydrate forms of this compound
have been prepared (59), and these two solvatomorphs were found to exhibit very
similar Raman spectra (46). However, as evident in Figure 9, the two similar
292
Bugay and Brittain
Figure 9 Raman spectra obtained within the fingerprint region for the anhydrate form of
naproxen sodium (solid trace), and its monohydrate phase (dashed trace). Source:
forms can still be distinguished on the basis of a few characteristic Raman peaks
(such as the 1368/1388/1414 cm21 triplet of the anhydrate phase, compared
with the 1382/1410 cm21 doublet of the monohydrate phase).
The 1:1 methanol solvate of lasalocid has been found to be capable of crystallizing in two polymorphic forms, and vibrational spectroscopy has been used
to study the differences between these (60). In particular, the stretching frequency
of the carbonyl moiety in the ketone functional group (observable in both IR
absorption and Raman spectroscopies) was found to be a sensitive indicator of
the crystal form. The data were interpreted to demonstrate that the observed spectroscopic differences did not depend on the presence or absence of bound solvent,
and instead reflected the existence of small conformational changes of the
lasalocid molecule.
Raman spectroscopy has been used to study the lattice and intermolecular
vibrations of griseofulvin and various solvates (61). Only weak van der Waals
interactions between the substance and the solvent were identified for the
benzene solvate, whereas for the bromoform and chloroform solvates, weak
Raman Spectroscopy
293
hydrogen bonding existed between the solvent and the griseofulvin carbonyl
group. Using variable temperature methods, it was found that no intermediate
structures formed during the desolvation of the solvatomorphs and that the solvatomorph structure simply transformed into the structure of the nonsolvated
substance.
Torosemide is known to crystallize in two nonsolvated forms in addition to
a channel solvate form that can be formed by water or alcohol (62). Although all
three forms were readily distinguishable on the basis of their X-ray powder
diffraction patterns, the Raman spectra of the solvate and Form-II were barely
distinguishable. Both of these forms were found to be metastable with respect
to Form-I, and the phase transformations could be initiated by differing storage
conditions. The solvate structure was formed through a framework of hydrogen
bonds that established continuous channels of varying sizes along the b-axis of
the crystal but was found to be unstable when exposed to extreme relative humidity conditions.
Three nonsolvated polymorphs of tranilast have been reported, in addition
to the chloroform and methylene chloride solvatomorphs (63). The Raman
spectra contained strong bands derived from the methylene stretching modes,
the aromatic ring vibrations, and various low-frequency lattice modes. It was
reported that the intensity of the amide band in the Raman spectrum of FormII (observed at 1643 cm21) was particularly strong relative to the intensities of
the analogous band in the other forms. Owing to the ease of its measurement,
Raman spectroscopy was developed as an in-process method to evaluate any contamination by nonsolvated Forms II or III in bulk nonsolvated Form-I. The
pattern of lattice vibrational modes noted for nonsolvated Form-I and the chloroform solvate indicated the existence of a structural similarity, and a similar
conclusion was deduced for the nonsolvated Form-III and the methylene chloride
solvatomorph.
The Raman spectra of four different hydrate forms of risedronate were found
to be dominated by vibrations associated with the substituted pyridine ring (64).
Both the anhydrate and the hemipentahydrate forms yielded two intense peaks
derived from an in-plane pyridine ring deformation mode that was similar to
that of neat pyridine. On the other hand, the monohydrate and variable hydrate
forms were dominated by a single peak derived from this pyridine ring deformation. The spectroscopic differences appeared to originate from differing
types of intermolecular interactions originating from hydrogen bonding between
the pyridine ring and an adjacent phosphate group. Additional differences in
Raman spectra between the different hydrates were observed in the highfrequency symmetric and antisymmetric methylene C22H stretching modes.
Raman Spectroscopy Studies of Phase Transformations
Owing to its ease of acquisition, and noninvasive nature of sample handling,
Raman spectroscopy has proven to be an ideal technique for the in situ study
294
Bugay and Brittain
of chemical and physical reactions and has found particular use in the study of
phase transformations. For example, different sieve fractions of trehalose dihydrate were subjected to isothermal heating at 808C, and the thermally induced
effects were monitored through Raman spectra obtained during the heating
process (65). It was found that after heating a fine particle size for 210
minutes, the substance appeared to convert into an amorphous phase that was
equivalent to the material obtained after lyophilization. On the other hand,
coarse trehalose dihydrate appeared to convert into the crystalline anhydrate at
the end of the heating process. Analysis of the Raman spectra permitted an evaluation of the kinetics involved.
In an analogous study, the kinetics of the thermally induced solid-state
transformation of carbamazepine Form-III into Form-I was studied by Raman
spectroscopy (66). The rate of transformation was monitored through changes
in the relative intensities of two C22H bending modes, as this approach was
chosen to avoid experimental artifacts. The initial stages of the temperature
dependence of the phase conversion are illustrated in Figure 10, and the activation of the process was determined to be 344 368 kJ/mol using four kinetic
models that provided good fits to the data. The authors concluded that FTRaman spectroscopy would be a useful probe technique for in-process control
technique, especially for manufacturing steps that entailed a change in the
crystal form of a substance.
Temperature-dependent Raman spectroscopy has been used to study the
low-temperature phase transitions in a biphenyl-fullerene single crystal (67). It
was found that three biphenyl vibrational modes and the C22H stretching
mode were indicative of the phase transition, and changes in the planarity of
the biphenyl molecule appeared to accompany the phase change. The behavior
of most of the C60-fullerene vibrational modes could be correlated with either
of the transitions, indicating changes in its site symmetry. In contrast, Raman
spectroscopy was used to demonstrate that single crystals of different polymorphs
of pentacene did not undergo any phase transformation reactions over the temperature range of 79 to 300 K (68).
The melting behavior of dipalmitoyl-phosphatidylcholine has been studied
by Raman spectroscopy, with the C H and C55O stretching modes being
found to be particularly sensitive to lateral expansions of the lipid matrix and
increased motion of the sn-2 chain within the bilayer interface region (69). The
order disorder phase transition in three crystal modifications of oleic acid has
also been studied, where it was found that the phase transformations also
involved changes in molecular conformation (70). In addition, the Raman
spectra were used to deduce structural information about the nature of the
b-phase, for which no crystal structure was known at the time of the work.
Like many amino acids, glutamic acid is capable of existing in two polymorphic forms, and the production of a desired form in a batch crystallization
process requires in-process control. Raman spectroscopy has been used to
measure the polymorphic composition of glutamic acid suspensions, and good
Raman Spectroscopy
295
Figure 10 Early-stage time evolution of the phase transformation of carbamazepine

Form-III into Form-I as a function of the isothermal heating temperature. Source:
calibration curves were obtained for the quantitative analysis of suspensions

containing both the a- and b-forms (71). For example, it was shown that for crystallizations conducted at 258C, the a-form nucleated and completed the crystallization but that the b-form grew subsequently through a solution-mediated phase
transformation mechanism. These processes were successfully modeled in a
later work that required the appropriate combination of experiment and simulation
to adequately understand and predict the crystallization behavior (72).
Using a diamond anvil cell and an appropriate spectroscopic interface, it is
possible to study the effect of pressure on the physical state of a substance. For
example, different Raman spectra were obtained for 2-adamatanone at pressures
of 1 atm and 21 kbar, with the spectral pattern of the high-pressure phase strongly
resembling that measured for the ordered phase, which can be produced by
cooling the substance to less than 150 K (73). The lattice vibrations of crystalline
296
Bugay and Brittain
tetrafluoro-1,4-benzoquinone have been found to be particularly sensitive to the

effect of pressure (74), as have the lattice modes of crystalline hexamethylbenzene (75). In the latter study, the Raman spectra provided evidence that the
intramolecular methyl torsional mode played an important role in the solidstate phase transformation.
Raman Spectroscopic Studies of Manufacture and Processing
The real-time monitoring of pharmaceutically relevant processes continues to
represent a growth area for Raman spectroscopy. On-line monitoring by Raman
spectroscopy has been utilized to examine synthetic organic reaction schemes to
investigate kinetics, and to identify nonisolated reaction intermediates. Distinct
advantages for Raman spectroscopy in this area are: (i) the ability to work with
aqueous-based systems with little spectral interference from water, (ii) the utilization of a fiber optic probe for direct and/or remote sampling, (iii) collection
of the Raman spectrum directly through the reaction vessel with little or no
spectral interference and, (iv) the ability to quantitatively analyze the spectral
results.
Obviously, if information about the chemical or physical identity can be
measured rapidly in the presence of the entire product matrix, in situ process
monitoring is the next logical step. This is almost exclusively the realm of
fiber optic sampling (76,77), where the probe is inserted either into the process
stream so that it is in intimate contact with the products and reactants or where
it is mounted in a sight glass window for monitoring in a noncontact mode.
Several considerations dictate the instrument configuration, and often the
process area imposes restrictions on the physical state of the instrumentation.
Fiber optic sampling allows positioning of the laser, launch optics, spectrograph,
detector, and any conditioning systems in remote locations from the process.
Although a number of fiber optic probe designs are commercially available
that address temperature, pressure, and interface integrity in various ways, most
share a common geometry. Fibers are typically composed of low-hydroxide
silica to reduce the silicon oxide contributions from the incident and collection
fibers. Additional holographic filtering optics are housed in the probe head to
reject nuisance scatter from the incident laser and to prevent the Rayleigh
scatter from entering the collection fibers. Utilizing fiber optics, a number of processes have been either analyzed or controlled based upon data derived from
Raman spectrometers.
The degree of control exerted over a crystallization process and the avoidance of forming undesirable metastable by-products are important principles in
the manufacturing of crystalline substances (78). For example, polycarboxylic
acids are known to affect calcium carbonate crystallization, and Raman spectroscopy has been used to monitor the levels of different polymorphs (calcite,
vaterite, and aragonite) as a function of additive identity and concentration in a
crystallization study (79).
Raman Spectroscopy
297
The antisolvent crystallization of cortisone acetate from acetone or mixed

acetone solvents by water have been monitored using in situ Raman spectroscopy
(80). Once the Raman features of the solutes, solvent systems, and antisolvents
were profiled, characteristic scattering bands were subsequently used to follow
the levels of these compounds during conduct of the process. Given that
Raman spectroscopy could provide information on dissolved and crystalline substances, it was judged to be superior to IR absorption spectroscopy (obtained
using the attenuated total reflectance sampling mode) because the latter
method could only provide information on dissolved substances.
Some of the most interesting applications of real-time process monitoring
include the interconversion of polymorphic solids under slurry conditions, and
Raman spectroscopy has been an important tool in the characterization arsenal
(81,82). For example, the spectra in Figure 11 show the migration of a specific
peak of progesterone during a temperature-controlled solid-state conversion,
and a plot illustrating the disappearance of Form II and the appearance of
Form I as measured in real time for process optimization.
In another study of this type, Raman spectroscopy was used to determine
the rate of polymorph turnover for a multipolymorphic research compound,
known to exist in four anhydrate forms, two hydrates, and numerous solvates
(83). The penultimate and purification steps of the manufacturing process
involved a coupling reaction that generated a mixture of crystal forms, which
was followed by phase conversion to the desired form. Eventually, after the performance of empirical and simulation work, the use of thermodynamic data to
establish process boundaries, and the use of kinetic data to establish process
time cycles, a highly robust process that yielded the desired crystal form was
developed.
A variety of physical changes can take place in materials as they are manufactured in the dosage form, and Raman spectroscopy has been shown to be
useful for the monitoring of these processing steps (84). For example, the
effect of excipients on the formation of theophylline monohydrate from the anhydrate in wet massed products has been studied, and Raman spectroscopy was used
to evaluate the phase composition (85). It was found that lactose monohydrate
could not prevent the phase conversion, and silicified microcrystalline cellulose
(a material known to be able to adsorb large quantities of water) provided protection from the phase conversion only at low-moisture levels. The selectivity of the
Raman spectra enabled evaluation of the phase composition to be made even in
the presence of large amounts of excipients.
The ability to use Raman spectroscopy as an in situ probe for wet granulation processes has been demonstrated. In one work, anhydrous theophylline and
caffeine were processed using wet granulation, and the formation of the corresponding hydrate phases monitored through the coupled method of Raman scattering (CCD detector) and NIR absorption (86). In another study, Raman
spectroscopy was used to follow the properties of the drug substance during
the conduct of a high-shear wet granulation process (87). Knowledge of the
298
Bugay and Brittain
Figure 11 Interconversion of the two solid-state forms of progesterone, as monitored by

Raman spectroscopy. Source: Courtesy of Kaiser Optical Systems, Inc.
real-time transformation kinetics facilitated the modeling of the processing step,

which was used to deduce time scales for hydrate formation during the process.
Quantitative Analytical Applications
The ability to use Raman spectroscopy for the performance of quantitative analysis is a significant advantage of the technique. It has been shown that the Raman
scattering intensity is proportional to the number of molecules being irradiated,
and the intensity of the scattered radiation is also proportional to the intensity
of the incident radiation and to the fourth power of the difference in frequencies
between the laser frequency and the molecular vibrational frequency. It follows
that increased degrees of Raman scattering intensity and potentially lower limits
Raman Spectroscopy
299
of detection can be achieved by increasing the intensity of the laser radiation,

and/or the frequency of the laser irradiation. This quantitative relationship
between Raman scattered intensity and concentration can be expressed as:
IR (IL sK)PC
(9)
where IR is the measured Raman intensity (units of photons/sec), IL is the laser

intensity (units of photons/sec), s is the absolute Raman cross-section (units of
cm2/molecule), K is a constant composed of measurement parameters, P is the
sample path length (units of cm), and C is the concentration (units of molecules/cm3). The constant K represents measurement parameters, such as utilizing the same spectrometer (collection optics efficiency), sample positioning, and
the overall efficiency of the Raman spectrometer.
In the past, various problems inherent with dispersive systems precluded
the use of Raman spectroscopy for quantitative analysis. With the advent of
Raman spectrometers based on the FT approach, the feasibility of quantitative
applications greatly improved (88). Some of the advantages associated with
the FT approach that are conducive toward the conduct of quantitative applications are Jacquinots advantage, Fellgetts advantage, and Connes advantage.
Jacquinots advantage (also known as the throughput advantage) depends on the
fact that the entrance aperture of the interferometer is large, making the sampling
geometry less sensitive to absolute repositioning of the sample cell. Fellgetts
advantage (otherwise known as the multiplex advantage) concerns the increased
speed at which spectra can be collected, because all wavelengths of the spectrum
are collected simultaneously. This rapid collection implies that FT-Raman spectroscopy might be less sensitive to instrument drift. Finally, Connes advantage
(also called the precision advantage) depends on the fact that there is only one
moving part in an FT system (one mirror in the interferometer).
The consequence of these advantages is that an FT system is very rugged,
allowing for excellent day-to-day wavelength precision. Although there is not
presently an approved ASTM material for calibrating Raman spectrometers,
the very strong Raman peak at 520 cm21 of silicon has become the industry standard for calibrating intensity and wavelength.
Although FT-Raman spectroscopy is more applicable to quantitative applications than is dispersive Raman spectroscopy, there are still some drawbacks
concerning the optimization of sampling conditions. In FT-Raman spectroscopy,
one typically samples a relatively small area (typically a 1 2 mm spot size) of the
total sample. As such, it is important that any spectra collected be truly representative of the bulk sample. For solution-phase studies, homogeneity of multicomponent samples is not an issue, and one need not worry about whether the
measurement is representative of the bulk or not.
On the other hand, the typical inhomogeneous composition of a solid phase
can create significant difficulties in the conduct of quantitative analysis. One way
300
Bugay and Brittain
to improve sample uniformity is to prepare all solid mixtures by geometric

mixing. Slurry mixing is another good technique, if the sample can withstand
contact with the slurry solvent. For data acquisition, sample-spinning accessories
should be used to spin the sample cup containing the mixture during spectral
acquisition. Using this type of accessory results in the collection of a more representative sample spectrum (a ring of data is collected, instead of just at a single
point). To effectively minimize any sample heterogeneity concerns, different
rings of data can be collected for each sample cup. The individual spectra
from each sampled area can then be coaveraged into groups to create representative spectra for the sample in the cup.
Another factor to consider when using Raman spectroscopy for quantitative
analysis is that the power output of the Raman laser can vary from day-to-day,
affecting the intensity of spectral peaks. It is advisable to normalize any data
to a standard before using these for quantitative purposes. One possible
method for normalizing spectra is to ratio the spectral response of the analyte
against a peak response for a non-changing component (such as an excipient in
a drug product or an internal standard). Alternatively, a ratio can be measured
using a peak response of a component changing in an opposite direction to that
of the component being monitored.
The surface-enhanced Raman scattering of drug adsorbed onto silver colloids has been used for quantitative purposes. Using this approach, trace quantities of nitrogen-containing drug substances have been detected, where a
detection limit of 35 ng was determined for p-amino-benzoic acid and 7 ng for
2-aminofluorene (89). In another work, after sulfadiazine, sulfamerazine, or sulfamethazine were adsorbed onto colloidal silver dispersions, the compounds
could be detected in the low nanograms per milliliter level (90).
Quantitative Raman spectral analysis has been used to determine the amorphous-to-crystalline phase content in indomethacin samples (91). Some of the
highlighted aspects in this paper concerned the need for producing homogeneous
calibration/validation samples and the difficulties associated with collecting a
Raman spectrum that is truly representative of the concentration. A linear correlation curve was constructed, which could be used to detect and predict low
concentrations of both amorphous and crystalline material in mixtures. It was
concluded that the largest source of error in the measurements arose from
inhomogeneous mixing of the amorphous and crystalline components in the
blends.
Raman spectroscopy was used to quantitatively determine the level of
amorphous material in lactose samples, and excellent accuracy was obtained
when the amorphous phase content was less than 10% (92). In a related study,
amorphous lactose was prepared by lyophilization, crystalline lactose monohydrate by crystallization, and the Raman spectra of blended standards were used
to develop calibration curves (93). The best discrimination between the two
forms was obtained through measurements of depolarization ratios in the prominent Raman bands observed at 865 and 1082 cm21.
Raman Spectroscopy
301
With suitable calibration, the same Raman spectra that were used to identify the crystal form existing in a polymorphic mixture can also be used to develop
quantitative methods for the phase composition. For example, binary mixtures of
chlorpropamide forms A and B were analyzed for both qualitative and quantitative purposes, even though the Raman spectra of both crystal forms strongly
resembled each other (94). In another study, a combination of FTIR and
Raman spectroscopies was used for the identification and quantitation of the
orthorhombic and monoclinic polymorphs of paracetamol (95). The precision
of the latter method was found to be less than 5%, and the limit of detection
for the monoclinic form was estimated to be 0.012 in units of mole fraction.
A quantitative Raman method has been used to analyze various mixtures of
b- and d-mannitol, where it was found that levels of b-mannitol as low as 2%
could be accurately quantified (96). As illustrated in Figure 12, the C22C22O
stretching mode of b-mannitol was observed at 1037 cm21, the analogous peak
for d-mannitol was observed at 1052 cm21, and mixtures of the two polymorphs
exhibit peaks due to both forms. The effect of particle size was investigated, and
it was reported that when the particle size range of the sample was controlled to
be less than 125 mm, spectra more representative of the overall sample mixture
were obtained. In addition, it was also demonstrated that experimental artifacts
could be further reduced through the use of a rotating sample holder.
As long as one can differentiate the most intense Raman peaks of a drug
substance from those of its associated excipients, a quantitative method for determination of the API in its formulation can be developed. For example, the determination of acetylsalicylic acid in tablets containing corn starch has been
developed, where the calibration curve was constructed using intensity ratios
of aspirin and starch scattering peaks (97). In another study, the level of sulfasalazine in microspheres produced from Eudragit RS has been determined using the
characteristic Raman bands of the drug substance (98). In this latter study, it was
found that a quantitative relationship existed between the quantity of drug substance in the microsphere and the ratio between the area of a sulfasalazine
peak and the area of a Eudragit peak.
A chemometric Raman spectroscopic method has been reported for the
precise evaluation of the microcrystallinity of indomethacin in a model tablet
matrix (99). The calibration model was constructed by the partial least-squares
analysis based on the multiplicative scatter correction and second derivative
Raman spectra. The method was validated down to 2% amorphous or crystalline
material in the tablet, which in turn represented only 0.2% of the total mass of the
tablet. Not only could the amorphous content in a compressed tablet be evaluated,
but differences in microcrystallinity at the surface and interior of the tablet could
also be determined.
The relative amounts of the three polymorphic forms of benzimidazole in
a 5% formulation premix for animal use were determined by FT-Raman
spectroscopy (100). To achieve this end, after comparison to the reference
spectra of the three pure polymorphic forms, the raw intensities of 78 selected
302
Bugay and Brittain
Figure 12 Raman spectra for the C2

2C2
2O stretching modes of b-mannitol (dashed
trace) and d-mannitol (dashed trace), and mixtures of these having a b-phase content
of 20% ( - - -), 40% ( ), and 80% ( - - - - - -).
peaks were vector-normalized, and stepwise linear regression models were then
used to estimate the relative quantities of forms A, B, and C in the different
samples.
Raman Spectroscopic Studies of the Character of Drug

Substances in Formulations
One possibility that must always be considered during the development of an
active pharmaceutical ingredient is that even through the physical form of the
drug substance may have been carefully controlled during its manufacture, it
may change as a result of secondary processing when the dosage form is formulated. The potential of Raman spectroscopy as a technique to establish the solidstate form of an API in a tablet matrix has been evaluated, and it was found that in
many cases the drug substance could be studied even when it constituted less than
1% of the tablet mass (101). For example, as illustrated in Figure 13, the Raman
spectrum of ibuprofen contains scattering peaks at energies where lactose or
Raman Spectroscopy
303
Figure 13 Carbonyl region of the Raman spectrum of ibuprofen (solid trace), illustrating the lack of scattering peaks compared with those for lactose monohydrate (short
dashed trace) or microcrystalline cellulose (long dashed trace). Source: Adapted from
Ref. 101.
microcrystalline cellulose have no peaks, and therefore a method to monitor the

state of ibuprofen in tablet formulations is possible.
The determination of ambroxol content in tablets has been accomplished
using Raman spectroscopy, with a calibration being established in the range of
8.3 16.26% (102). The samples were rotated to eliminate effects associated
with compositional variation, baseline and noise effects were minimized by preprocessing the spectra using wavelet transformation, and then the spectra were
normalized prior to being subjected to partial least-squares regression analysis.
Four different geometric laser irradiation patterns were examined to study
the effect of tablet subsampling (103). Not surprisingly, it was found that the
best result in terms of prediction error was obtained by irradiating a large
tablet area.
The potential of Raman spectroscopy as a means to characterize the
relationship between the hydration state of a drug substance and the physical
stability of its tablet has been studied (104). Risedronate sodium was chosen as
304
Bugay and Brittain
the drug substance for the study, because its water of hydration is contained within
channels in the crystal structure and therefore, potentially mobile. The substance
was wet granulated and dried down to granulated moisture contents in the range of
1 7%. Raman spectra were obtained during the course of the drying step, providing information on the contraction of the crystal lattice that accompanied the
dehydration. The changes induced in the drug substance crystal structure were
found to be reflected somewhat in the stability of the dosage forms.
As discussed earlier, the favorable sampling and spectroscopic characteristics of Raman spectroscopy make it an ideal technique for monitoring pharmaceutical processing. For example, the on-line technique has been used to
monitor the blend uniformity of azimilide dihydrochloride formulations and to
assist in the optimization of the blending process (105). The methodology has
also been used to measure variabilities in tablet coatings, because Raman
spectral changes could be correlated with tablet exposure times in a pan
coater (106). It was found that the results for coating thicknesses could be
improved by preprocessing the spectral data with multiplicative scatter
correction, standard normal variate transformation, and Savitzky-Golay
second derivative smoothing.
Water that becomes associated with amorphous polymers is known to
affect their chemical and physical properties, and Raman spectra of polymers
under different environmental conditions have been used to study these effects.
In one study, the nature of the polymer-water hydrogen bond interactions in
two molecular weight grades of poly(vinylpyrrolidone) (PVP) and also in poly
(vinylacetate) and poly(vinylpyrrolidone-co-vinylacetate) (107). To assist in
interpretation of the Raman spectra, the vapor absorption isotherms and glass
transition temperatures of the polymers were also obtained, and it was determined
that the pyrrolidone group tended to interact more with water than did the acetate
group. Using principal components analysis, differentiation between polymers in
the rubbery state and those in the glassy state could be made. In a related study,
Raman spectroscopy was used in conjunction with electron microscopy to study
the interaction of water with different cellulose ethers and to monitor the swelling
and drying behavior of these materials when they went in and out of the gel
phase (108).
Raman spectroscopy has been used to evaluate the interactions between
drug substances and polymeric excipients. The molecular structure existing in
solid dispersions of indomethacin and PVP has been studied, with the carbonyl
stretching mode of the drug substance being used as the spectroscopic probe
(109). Addition of PVP to amorphous indomethacin increased the intensity of
the nonhydrogen bonded carbonyl groups, while the carbonyl-stretching band
of the PVP polymer decreased to lower energies. These findings led to the conclusion that indomethacin interacted with PVP through hydrogen bonds formed
between the drug hydroxyl and polymer carbonyl groups, resulting in disruption
of the indomethacin dimers that existed in the bulk drug substance.
Raman Spectroscopy
305
In another study, Raman spectra of promethazine, diclofenac, theophylline,

and indomethacin in polymeric diluents based on polyethylene oxide, sodium
alginate, and hydroxypropylmethyl cellulose were obtained (110). It was determined that the characteristic Raman spectra of each drug substance could be
used to quantitate its concentration in the formulations and also to provide information on the character of any drug excipient interactions and incompatibilities
that might exist in the formulation matrix. The latter aspect, and the ease of
obtaining Raman spectral results, clearly demonstrates the utility of this technique as one of the tools in the arsenal of the preformulation scientist.
The use of supercritical fluids in pharmaceutical manufacturing is becoming more widespread, and Raman spectroscopy has been used to study formulations obtained by this processing route. For example, the impregnation of
ibuprofen into PVP in supercritical carbon dioxide has been monitored using in
situ Raman and IR absorption spectroscopies (111). The spectral results led to
the conclusion that the ibuprofen became dispersed as molecules in the polymer
matrix through site-specific associations of the drug substance with the carbonyl
groups of PVP. It was also established that the incorporation of the drug substance
into the PVP perturbed the degree of water sorption by the polymer.
Raman spectroscopy has been used to evaluate the physical state of chlorhexidine, when it was formulated into viscoelastic, bioadhesive, semisolids
(112). The formulations were prepared by dispersing the drug-free base into
buffered polymer matrices consisting of hydroxyethylcellulose, PVP, and polycarbophil, and the Raman spectra of the drug substance were acquired within a variety
of formulation conditions. The existence of an acid base reaction between chlorhexidine and polycarbophil was demonstrated, forming the dictation of the drug
substance that was solubilized by the hydroxyethylcellulose component. The
utility of Raman spectroscopy as a process analytical technique during formulation of topical gels and emulsions has been demonstrated, with changes in the
spectral characteristics of the thickening agent carbopol and the emulsifying
agent tefose being monitored after performance of major processing steps (113).
Chemical Mapping and Imaging
One of the more exciting new applications of Raman spectroscopy concerns
chemical mapping. By incorporating a programmable, xyz-movement stage
into a Raman microscope, it is now possible to generate a chemical image of a
two-dimensional area of a sample. Through the use of a mapping stage, a
sample, such as a sliced tablet, can be moved in the x- and y- directions, obtaining
spectra at each step. If the sample requires refocusing at different locations, the
z-direction can be automated as well. The distance, which the stage moves in
the x- and y-directions is called the step size and usually can be as small as 1 mm.
There are three types of maps, and these are identified as point, line, and
area. A point map provides several different areas of a sample to be analyzed
306
Bugay and Brittain
consecutively, but the spectra are not related to each other spatially. A point map
can be considered as a type of auto-sampler. An example of a point map application is a 96-well plate that contains combinatorial beads in each well. The
second map type (the line map) defines a series of spectra to be obtained along
one dimension. In line maps, chemical changes that occur along this dimension
are investigated. An example of a line map sample is a cross-sectioned pharmaceutical beadlet that has several different layers exposed. One of the more practical reasons that line maps are so popular is that they can provide detailed
information regarding the chemical changes that occur across a sample without
the need to collect as many spectra as is required for area maps.
The final map type (the area map) defines a series of spectra to be collected
in two dimensions (i.e., over an entire region). This type of map provides a
Raman image that can be directly compared with the visual image, often allowing
nonvisible (to the eye) features to be identified. A common area map sample is a
tablet. The tablet can be mapped and the various ingredients can be monitored for
content uniformity. A series of images demonstrating the dispersion of several
excipients in a tablet is displayed in Figure 14. This particular area map was
obtained with a step size of 4 mm, a sampling spot size of approximately
1 mm, and a total sampling area of 87 52 mm. The time required to collect
this map was approximately 26 hours, but it should be noted that not all
Raman mapping experiments require this amount of time. For example, if a
map is being performed to search for a particular component, the step size
Figure 14 Peak area profiles (images) representing (A) mannitol, (B) aspartame, (C) cellulose, (D) magnesium stearate, (E) corn starch, and (F) monoammonium glycyrrhizinate.
A high concentration of material is represented by the gray color, whereas a low concentration is represented by the black color.
Raman Spectroscopy
307
need only be in the order of the particle size of that substance. A good approach
would be to first obtain a larger area map on the sample with larger step sizes
and shorter sampling times per point. Once an area of interest is defined by analyzing the data from the first map, a smaller, higher resolution map can then be
defined.
Area maps and imaging have experienced a recent surge in popularity,
mainly due to the use of CCD array cameras and liquid crystal tunable filter
(LCTF) technology, respectively. These detectors greatly reduce the amount of
time required to collect an area map (114). In one case, the time required to
collect an area map decreased by approximately four hours upon switching
from a CCD camera (longer than five hours for the experiment) to LCTF technology (less than one hour for the experiment).
Once line or area mapping experiments have been performed, profiles are
created that enable certain spectral features to be monitored spatially on the
sample. The data sets generated by mapping can consist of hundreds or even
thousands of spectra, and viewing all of these spectra concurrently is unrealistic.
Profiles can aid in this task by reducing the data set into a more easily viewed
format (an image). A profile is a representation of map data in which a measurement of spectral intensity or some other characteristic is shown for each sample
point. A profile will verify the presence, location, and extent of a defined spectroscopic feature in the sample. There are many types of profiles. Some profiles
(such as chemigram or component profiles) compare an entire reference spectrum
to every spectrum in the map, with the resulting image showing spectral similarity across the mapping region. Other images can be created based on profiling
a specific peak area, peak height, peak area ratio, or peak height ratio. Profiles can
also be performed based on a group of peaks specific for certain functional groups
(e.g., alkanes) or even on a quantitative method.
In recent Raman applications of chemical imaging, the content uniformity
of an active drug substance within a drug product was investigated (115,116).
A pharmaceutical tablet containing cyclobenzaprine hydrochloride was
shaved, such that a chemical image was obtained for each sampled depth
within the tablet (115). Because the Raman spectrum of cyclobenzaprine hydrochloride exhibited spectral features different from those of the excipients, a
chemical image specific to the drug substance could be obtained, or, for that
matter, for each excipient. Figure 15 displays the chemical image for one
plane (depth) of the tablet. It is clearly observed that the drug substance is
detected at the 0,0 spatial position within the sampled plane of the tablet.
The corresponding excipient chemical image also shows the lack of excipient
concentration (signal intensity) at the same spatial position. These results
demonstrate the excellent sensitivity and selectivity (spectrally and spatially)
for the spectroscopic, chemical imaging technique. In the second example of
Raman imaging of a drug product (116), a map of the ratio of ibuprofen/
PVP concentration for a 45 25 mm area was generated. The results showed
that ibuprofen was homogeneously distributed throughout the polymer PVP.
308
Bugay and Brittain
Figure 15 Raman spectral chemical images of a shaved surface of a tablet containing

cyclobenzaprine hydrochloride. Image (A) maps the intensity of the drug substance,
whereas (B) is an image of the same spatial area, mapping one of the excipients.
Determination that ibuprofen was equally distributed throughout the polymer

suggests high stability of the formulation (no localization of drug or polymer)
leading to desired dissolution characteristics.
Other works concerning mapping applications of interest to the pharmaceutical industry include crystal formation in hormone replacement therapy
Raman Spectroscopy
309
patches [Bug5], particle size analysis in mixtures (117), and pharmaceutical

matrix determination of dosage formulations (118).
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
Cutmore EA, Skett PW. Spectrochim Acta 1993; 48A:809.

Petty CJ, Bugay DE, Findlay WP, Rodriguez C. Spectroscopy 1996; 11:41.
Threfall TL. Analyst 1995; 120:2435.
Bugay DE, Williams AC. Vibrational spectroscopy, In: Brittain HG, ed. Physical
Characterization of Pharmaceutical Solids. New York: Marcel Dekker, 1995:59 91.
Findlay WP, Bugay DE. J Pharm Biomed Anal 1998; 16:921.
Armstrong CL, Edward HGM, Farwell DW, Williams AC. Vibrat Spect 1996;
11:105.
Eyring H, Walter J, Kimball GE. Quantum Chemistry. New York: John Wiley &
Sons, 1944:121 123.
Herzberg G. Infrared and Raman Spectra of Polyatomic Molecules. New York, NY:
Van Nostrand Reinhold Co., 1945:239 271.
Placzek G. Handbuch der Radiologie. Vol. 6. Leipzig: Akademische Verlagsgeselschaft, 1934:208.
Hochstrasser RM. Molecular Aspects of Symmetry. New York: W.A. Benjamin Inc.,
1966.
Orchin M, Jaffe HH. Symmetry, Orbitals, and Spectra. New York: Wiley-Interscience, 1971.
Cotton FA. Chemical Applications of Group Theory. 2d ed. New York: Wiley-Interscience, 1971.
Harris DC, Bertolucci MD. Symmetry and Spectroscopy. New York: Oxford University Press, 1978.
Asher SA. Anal Chem 1993; 65:59A; ibid 1993; 65:201A.
Morris MD, Wallan DJ. Anal Chem 1979; 51:182A.
Fleischmann M, Hendra PJ, McQuillan AJ. Chem Phys Lett 1974; 26:163.
Jeanmaire DL, van Duyne RP. J Electroanal Chem 1977; 84:1.
Albrecht MG, Creighton JA. J Am Chem Soc 1977; 99:5215.
Creighton JA, Blatchford CG, Albrecht MG. J Chem Soc Faraday Trans II 1979;
75:790.
Garrell RL. Anal Chem 1989; 61:401A.
Campion A, Kambhampati P. Chem Soc Rev 1998; 27:241.
Moskovits M, DiLella DP, Maynard K. Langmuir 1988; 4:67.
Durig JR, Harris WC. Raman spectroscopy. In: Weissberger A, Rossiter BW, eds.
Physical Methods of Chemistry, Part IIIB. Vol. 1. New York: Wiley-Interscience,
1972:85 205.
Grasselli JG, Snavely MK, Bulkin BJ. Chemical Applications of Raman Spectroscopy. New York: Wiley-Interscience, 1981.
Strommen DP, Nakamoto K. Laboratory Raman Spectroscopy. New York: WileyInterscience, 1984.
Hendra P, Jones C, Warnes G. Fourier Transform Raman Spectroscopy. Chichester:
Ellis Horwood Ltd., 1991.
Schrader B. Infrared and Raman Spectroscopy. Weinheim: VCH Publishers, 1995.
310
Bugay and Brittain
28. Mcreery RL. Raman Spectroscopy for Chemical Analysis. New York: Wiley-Interscience, 2000.
29. Lewis IR, Edwards HGM. Handbook of Raman Spectroscopy. New York: Marcel
Dekker, 2001.
30. Adar F. Evolution and revolution of Raman instrumentation, chapter 2. In: Lewis IR,
Edwards HGM, eds. Handbook of Raman Spectroscopy. New York: Marcel Dekker,
2001:11 40.
31. McCreery RL, Horn AJ, Spencer J, Jefferson E. J Pharm Sci 1998; 87:1.
32. Flett MC. Characteristic Frequencies of Groups in the Infra-Red. Amsterdam: Elsevier Pub., 1963.
33. Bellamy LJ. Advances in Infrared Group Frequencies. London: Methuen & Co.,
1968.
34. Colthup NB, Daly LH, Wiberley SE. Introduction to Infrared and Raman Spectroscopy, 2d ed. London: Academic Press, 1975.
35. Socrates G. Infrared and Raman Characteristic Group Frequencies Tables and
Charts. 3rd ed. Chichester: John Wiley & Sons, 2001.
36. Dollish FR, Fateley WG, Bentley FF. Characteristic Raman Frequencies of Organic
Compounds. New York: John Wiley & Sons, 1974.
37. Hendra PJ. American Laboratory, December issue, 17, 1996.
38. Reference 24, pp. 3839.
39. Huong PV. J Pharm Biomed Anal 1986; 4:811.
40. Frank CJ. Pharmaceutical applications of Raman spectroscopy. In: Pelletier MJ, ed.
Analytical Applications of Raman Spectroscopy. Oxford: Blackwell Science,
1999:224 275.
41. Bugay DE. Adv Drug Del Rev 2001; 48:43.
42. Bugay DE, Findlay WP. Pharmaceutical Excipients: Characterization by IR, Raman,
and NMR Spectroscopy. New York: Marcel Dekker, 1999.
43. Cutmore EA, Skett PW. Spect Lett 1993; 26:1395.
45. Leopold N, Baena JR, Bolboaca M, Cozar O, Kiefer W, Lendl B. Vibrat Spect 2004;
36:47.
46. Brittain HG. Unpublished Results for Raman spectra obtained at a resolution of
5 cm21, using an Ocean Optics model R-300-785 Raman spectrometer. Bulk
powder samples were contained in aluminum sample cups, and then excited at
785 nm to obtain the spectrum.
47. Raghavan K, Dwivedi A, Campbell GC, Johnston E, Levorse D, McCauley J,
Hussain M. Pharm Res 1993; 10:900.
48. Gu XJ, Jiang W. J Pharm Sci 1995; 84:1438.
50. Gandhi RB, Bogardus JB, Bugay DE, Perrone RK, Kaplan MA. Int J Pharm 2000;
201:221.
51. Maurin MD, Vickery RD, Rabel SR, Rowe SM, Everlof JG, Nemeth GA,
Campbell GC, Foris CM. J Pharm Sci 2002; 91:2599.
52. Stowell GW, Behme RJ, Denton SM, Pfeiffer I, Sancilio FD, Whittall LB,
Whittle RR. J Pharm Sci 2002; 91:2481.
53. Cheung EY, Harris KDM, Johnson RL, Hadden KL, Zakrzewski M. J Pharm Sci
2003; 92:2017.
Raman Spectroscopy
311
54. Giordano F, Rossi A, Moyano JR, Gassaniga A, Massarotti V, Bini M, Capsoni D,

Peveri T, Redenti E, Carima L, Alberi MD, Zanol M. J Pharm Sci 2001; 90:1154.
55. Terol A, Cassanas G, Nurit J, Pauvert B, Bouassab A, Rambaud J, Chevallet P.
J Pharm Sci 1994; 83:1437.
56. Raghavan K, Dwivedi A, Campbell GC, Nemeth G, Hussain M. J Pharm Biomed
Anal 1994; 12:777.
57. Anquetil PA, Brenan CJH, Marcolli C, Hunter IW. J Pharm Sci 2003; 92:149.
58. Kim T-S, Rousseau RW. Cryst Growth Design 2004; 4:1211.
59. The solvatomorphs of naproxen were supplied by Drs. Alan Salerno and Paul
Isbester, Albany Molecular Research, Albany, NY.
60. Friedman JM, Fousseau DL, Shen C, Chiang CC, Duesler EN, Paul IC. J. Chem. Soc.
Perkin II 1979; 835.
61. Bolton BA, Prasad PN. J Pharm Sci 1981; 70:789.
62. Rollinger JM, Gstrein EM, Burger A. Eur J Pharm Biopharm 2002; 53:75.
63. Vogt FG, Cohen DE, Bowman JD, Spoors GP, Zuber GE, Trescher GA, DellOrco
PC, Katrincic LM, Debrosse CW, Haltiwanger RC. J Pharm Sci 2005; 94:651.
64. Redman-Furey N, Dicks M, Bigalow-Kern A, Cambron RT, Lubey G, Lester C,
Vaughn D. J Pharm Sci 2005; 94:893.
65. Taylor LS, Williams AC, York P. Pharm Res 1998; 15:1207.
66. OBrien LE, Timmins P, Williams AC, York P. J Pharm Biomed Anal 2004; 36:335.
67. Palles D, Marucci A, Penicaud A, Ruani G. J Phys Chem B 2003; 107:4904.
68. della Valle RG, Venuti E, Farina L, Brillante A, Masino M, Girlando A. J Phys Chem
B 2004; 108:1822.
69. OLeary TJ, Levin IW. J Phys Chem 1984; 88:1790.
70. Kobayashi M, Kaneko F, Sato K, Suzuki M. J Phys Chem 1986; 90:6371.
71. Ono T, ter Horst JH, Jansens PJ. Cryst Growth Design 2004; 4:465.
72. Ono T, Kramer HJM, ter Horst JH, Jansens PJ. Cryst Growth Design 2004; 4:1161.
73. Harvey PD, Butler IS, Gilson DFR, Wong PTT. J Phys Chem 1986; 90:4546.
74. Ikuta A, Suzuki Y, Nibu Y, Shimada H, Shimada R. Bull Chem Soc Jap 1999; 72:963.
75. Takeshita H, Suzuki Y, Nibu Y, Shimada H, Shimada R. Bull Chem Soc Jap 1999;
72:381.
76. Wang F, Wachter JA, Antosz FJ, Berglund KA. Org Proc Res Dev 2000; 4:391.
77. Vergote GJ, DeBeer TRM, Vervaet C, Remon JP, Baeyens WRG, Diericx N,
Verpoort F. Eur J Pharm Sci 2004; 21:479.
78. Rodriguez-Hornedo N, Murphy D. J Pharm Sci 1999; 88:651.
79. Agerwal P, Berglund KA. Cryst Growth Design 2003; 3:941.
80. Falcon JA, Berglund KA. Cryst Growth Design 2003; 3:947.
82. Brittain HG. Solid-state analysis, chapter 3. In: Ahuja S, Scypinski S, eds. Handbook
of Pharmaceutical Analysis. New York: Marcel Dekker, 2001:57 84.
83. Starbuck C, Spartalis A, Wai L, Wang J, Fernandez P, Lindemann CM, Zhou GX,
Ge Z. Cryst Growth Design 2002; 2:515.
85. Airaksinen S, Luukkonen P, Jorgensen A, Karjalainen M, Rantanen J, Yliruusi J.
J Pharm Sci 2003; 92:516.
86. Jorgensen A, Rantanen J, Karjalainen M, Khriachtchev L, Rasanen E, Yliruusi J.
Pharm Res 2002; 19:1285.
312
Bugay and Brittain
87.
88.
89.
90.
Wikstrom H, Marsac PJ, Taylor LS. J Pharm Sci 2005; 94:209.

Walder FT, Smith MJ. Spectrochim Acta 1991; 47A:1202.
Torres EL, Winefordner JD. Anal Chem 1987; 59:1626.
Sutherland WS, Laserna JJ, Angebranndt MJ, Winefordner JD. Anal Chem 1990;
62:689.
Taylor LS, Zografi G. Pharm Res 1998; 15:755.
Niemela P, Paallysaho M, Harjunen P, Koivisto M, Lehto V-P, Suhonen J,
Jarvinen K. J Pharm Biomed Anal 2005; 37:907.
Murphy BM, Prescott SW, Larson I. J Pharm Biomed Anal 2005; 38:186.
Tudor AM, Church SJ, Hendra PJ, Davies MC, Melia CD. Pharm Res 1993; 10:1772.
Al-Zoubi N, Koundourellis JE, Malamataris S. J Pharm Biomed Anal 2002; 29:459.
Roberts SNC, Williams AC, Grimsey IM, Booth SW. J Pharm Biomed Anal 2002;
28:1135.
Kontoyannis GG, Orkoula M. Talanta 1994; 41:1981.
Watts PJ, Tudor A, Church SJ, Hendra PJ, Turner P, Melia CD, Davies MC. Pharm
Res 1991; 8:1323.
Okumuar T, Otsuka M. Pharm Res 2005; 22:1350.
de Spiegeleer B, Seghers D, Wieme R, Schaubroeck J, Verpoort F, Slegers G,
van Vooren L. J Pharm Biomed Anal 2005; 39:275.
Taylor LS, Langlilde FW. J Pharm Sci 2000; 89:1342.
Hwang M-S, Cho S, Chung H, Woo Y-A. J Pharm Biomed Anal 2005; 38:210.
Johansson J, Pettersson S, Folestad S. J Pharm Biomed Anal 2005; 39:510.
Hausman DS, Cambron RT, Sakr A. Int J Pharm 2005; 299:19.
Hausman DS, Cambron RT, Sakr A. Int J Pharm 2005; 298:80.
Romero-Torres S, Perez-Ramos JD, Morris KR, Grant ER. J Pharm Biomed Anal
2005; 38:270.
Taylor LS, Langlilde FW, Zografi G. J Pharm Sci 2001; 90:888.
Fechner PM, Wartewig S, Kiesow A, Heilmann A, Kleinebudde P, Neubert RHH.
J Pharm Pharmacol 2005; 57:689.
Taylor LS, Zografi G. Pharm Res 1997; 14:1691.
Davies MC, Binns JS, Melia CD, Hendra PJ, Bourgeois D, Church SP, Stephenson
PJ. Int J Pharm 1990; 66:223.
Kazarian SG, Martirosyan GG. Int J Pharm 2002; 232:81.
Dias CS, Mitra AK. J Pharm Sci 2000; 89:572.
Islam MT, Rodriguez-Hornedo N, Ciotti S, Ackermann C. Pharm Res 2004;
21:1844.
Zugates CT, Treado PJ. Int J Vib Spec 1999; 2:59.
Houghtaling MA, Bugay DE. Raman Spectroscopic Imaging: A Potential New
Technique for Content Uniformity Testing. Paper presented at the American Association
of Pharmaceutical Scientists Annual Meeting, New Orleans, LA, Nov 1418, 1999.
Breitenbach J, Schrof W, Neumann J. Pharm Res 1999; 16:1109.
Theophilus A, Lancaster R. Particle Size Analysis of Binary or Tertiary Mixtures
Using Raman Image Analysis. Paper presented at the Federation of Applied Spectroscopy Conference, Nashville, TN, Sep 24 28, 2000.
Clarke F, Famieson M, Hammond S, Clard K. Chemical Images The Key to
Pharmaceutical Matrix Determination. Paper presented at the Federation of
Applied Spectroscopy Conference, Nashville, TN, Sep 24 28, 2000.
91.
92.
93.
94.
95.
96.
97.
98.
99.
100.
101.
102.
103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
114.
115.
116.
117.
118.
10
Near-Infrared Spectroscopy
Robert P. Cogdill and James K. Drennen, III
School of Pharmacy, Duquesne University,
Pittsburgh, Pennsylvania, U.S.A.
INTRODUCTION
The primary analysis methods, such as infrared (IR) absorption spectroscopy, are
most often applied within the controlled confines of analytical laboratories. Over
the years their performance frontiers have been expanded, giving researchers
unprecedented capabilities in the analysis of materials.
A high sensitivity is generally achieved, however, at the expense of robustness and flexibility. Only in select instances when sufficient demand is present to
justify the expense of technology transfer have primary analytical methods been
adapted for field application. In contrast, the development of near-infrared (NIR)
spectroscopy has followed a unique path relative to other spectroscopic methods
for analyzing pharmaceutical solids. As discussed in the following paragraphs,
the current level of knowledge in NIR spectroscopy can be mainly attributed
to its roots as a rugged, field-tested technique. This chapter is devoted to illustrating the intricacies and advantages of NIR spectroscopy for the analysis of
pharmaceutical solids, beginning with some perspectives on the historical
development of NIR technology.
In his famous work, Experiments on the Refrangibility of the Invisible
Rays of the Sun, presented to the Royal Society in 1800 (1), Sir William
F. Herschel first demonstrated the existence of optical radiation beyond the
limits of the visible spectrum. In a series of experiments probing the relationship
between color and heat, Herschel observed an increase in temperature from violet
to red. In a serendipitous moment, however, he discovered that the hottest
313
314
Cogdill and Drennen
temperature was actually beyond the red limit of visible light. The mysterious
invisible radiation, which he initially termed as calorific rays, is actually
what has come to be known as short-wave near-infrared (SW-NIR), or the
Herschel near-infrared.
As a number of texts dedicated to NIR spectroscopy have chronicled (26),
the NIR spectral region is largely ignored as an analytical asset for nearly 150 years
following Herschels discovery until Karl Norris, an agricultural engineer working
for the U.S. Department of Agriculture (USDA), saw its potential for rapid, quantitative analysis of complex biological samples. Norriss pioneering work eventually spawned the development of an NIR industry that produced analyzers
capable of rapidly determining the concentration of constituents in whole foodstuffs
with very little sample preparation. From the mid-1960s until 1986 the rate of publication for articles mentioning NIR spectroscopy increased at a dizzying pace, with
most articles concerned primarily with the rapid analysis of food and agricultural
samples. From 1800 to 1970 only about 50 papers concerning NIR analysis were
contributed (7). In 1987 Phil Williams and Karl Norris edited a comprehensive
text with nearly 1000 references on the subject of NIR technology in the food
and agricultural industries (5); their book, which is now in second edition (4),
should be considered a staple text for anyone (regardless of their industry)
interested in applied NIR spectroscopy for the analysis of solid materials.
Unlike primary analytical methods, which have pushed innovation in new
sensors and exotic optical and magnetic materials on their own behalf, NIR
technology has generally advanced by riding the wake of technology developments in seemingly unrelated industries. The interest in the NIR region of
the electromagnetic spectrum, generally accepted as the range from 780 to
2500 nm, or 12,820 to 4000 cm21, lagged behind the ultraviolet (UV), visible
(VIS), and mid-IR for most of the twentieth century. Unlike the sharp absorbance
bands of the mid-IR fingerprint region (frequencies below 1500 cm21), which are
utilized for qualitative analyses in synthetic organic chemistry, the absorbance
bands in the NIR, which consist primarily of overtones and combinations of
the C22H, N22H, and O22H fundamental vibrations, are very broad, overlapping,
and orders of magnitude weaker, thereby precluding its use as a primary qualitative analysis method. Furthermore, because of the propensity for matrix and
solvent interactions, as well as the great complexity in absorbance bands, early
chemists had difficulty specifying baseline effects in the NIR region, and they
viewed NIR spectroscopy as being less reliable for quantitative assays relative
to UV/VIS spectrophotometry.
The outlook for NIR spectroscopy eventually brightened, however, as
enabling technologies began to appear following World War II. Karl Norriss
early work, which was carried out using the Beltsville Universal Computerized
Spectrophotometer (BUCS), was revolutionary in that it utilized dual Cary
model 14 grating/prism monochromaters and new broad-band PbS detectors
(commercialized from military efforts to build heat-seeking weaponry), and
powerful digital computing capabilities (Karls computer had an immense
315
memory capacity of 256 kb) (8). The popularity of NIR spectroscopy accelerated
during the 1970s and 1980s on the heels of the development of low-cost personal
computers (PCs), consumer imaging devices based on silicon detectors, and
re-writable digital memory. Since the 1990s, NIR spectroscopy has benefited
from the significant public investments in the telecom industry, which brought
economies of scale by mass-introduction of InGaAs diode array detectors,
tunable NIR lasers, fiber optics, and advances in dielectric optical coatings.
Over the past decade, NIR spectroscopy has earned its reputation as a
powerful analytical technology in its own right. The aforementioned technological developments and the knowledge and understanding gained by the early
adopters of NIR have turned the negative aspects of the technology into positive
features. The NIR spectroscopy was ultimately found to be useful because the
relatively weak absorbance bands in the region allow for interrogation of
sample matrices to greater depth than is possible in the IR and UV/VIS
regions. Furthermore, in contrast to the IR, NIR spectroscopy has the unique
ability to quantify constituents in the presence of significant amounts of water.
The complex nature of NIR absorbance bands and baseline effects were initially
viewed as a puzzle; the study of chemometrics (one of the few scientific fields
owing much of its early development to NIR spectroscopy) has yielded the
ability to resolve NIR spectra for simultaneous multicomponent analyses.
With respect to its evolution from a nondestructive method for analyzing
whole grain samples, it could be said that the development of NIR spectroscopy
has been chiefly concerned with the analysis of solids from the very beginning.
As a mature technology, a number of books (2 4,6), reviews (9 11), and a
peer-reviewed journal dedicated to NIR technology (12) that discuss the application of NIR spectroscopy in specific industries are available. The following
sections of this chapter provide a synopsis of the unique aspects of the fundamental science of NIR spectroscopy, an overview of the equipment and methods
employed, and a discussion of the specific areas where NIR spectroscopy is
being used for the analysis of pharmaceutical solids.
FUNDAMENTAL PRINCIPLES
At this point it should be clear that the NIR is a unique locale on the continuum of
the electromagnetic spectrum. Some potential practitioners of NIR spectroscopy
have failed in the past by treating the NIR as an extension of the IR or UV/VIS
during method development or application. Indeed, the unique aspect of the
interaction between solid materials and NIR radiation at the molecular and
particle scales set NIR spectroscopy apart from other methods.
Molecular-Scale Phenomena
Most comprehensive texts on applied NIR spectroscopy begin with a discussion
of the anharmonic oscillator model of the fundamental molecular vibrations
316
Cogdill and Drennen
(and their overtones and combinations) induced by photonic absorption, which

have already been covered in earlier chapters of this book. More important to
the current discussion, however, is how the complexity of the NIR spectrum
contrasts with the IR, which is more readily described by fundamental models
and quantum theory.
The previous chapters have discussed the harmonic and anharmonic
oscillator models of molecular vibration. The concept of anharmonic oscillation
holds significant implications for the existence and complexity of absorption
bands in the NIR (13,14):
.
.
Overtone transitions, which involve an increase in a vibrational

quantum number of greater than one, are allowed to occur,
Combination modes, which involve a simultaneous increase in the
vibrational quantum numbers of two or more different vibrations
from absorption of a single photon, are allowed to occur, and
The separations of the vibrational energy states of a given vibration are
not equal in the NIR and are offset depending on the degree of
anharmonicity.
Thus, overtone and combination absorption bands would not exist, and the
NIR region would be virtually useless for analytical purposes, if the anharmonic
model did not apply. Anharmonic oscillation adds significant complexity to NIR
spectra; combinations of fundamental bands and overtones, as well as those of
fundamental and overtone bands, are possible. However, vibrations must
involve the same functional group and have the same symmetry. The most
common bands in the NIR involve C22H, N22H, and O22H stretching; C22H,
N22H, and O22H bending, as well as C22O, C22N, and N22O stretching that
are represented to some extent. It is well known that, given a molecule with N
atoms, up to (3N 2 6) overtone absorption bands are possible. A simple methyl
group, for example, will have up to six absorption bands. Hence, a typical
small-molecule active pharmaceutical ingredient (API) will have many possible
overtone bands, with significant opportunities for combination bands. Correlation
charts indicating the expected location of absorbance bands in the NIR spectrum
for common organic bonds and functional groups are available in many classic
spectroscopy texts (15 18).
The degree of anharmonicity determines the amount by which the frequency of overtone bands are displaced from exact integer multiples of the
fundamental vibrations and their intensity. As the mass of hydrogen in C22H,
N22H, and O22H bonds is so low, such vibrations tend to be very anharmonic
and will tend to shift further from the integer multiples of their fundamental
frequencies. The degree of anharmonicity for a molecule is further dependent
on the level of hydrogen bonding, temperature, and interaction with other
molecules in the sample matrix.
317
Resonance Effects
The NIR spectrum is further complicated by the existence of Fermi and DarlingDennison resonances. Fermi resonance is attributable to the cubic terms of the
anharmonic oscillator potential energy functions and may occur in cases when
two vibrational modes of the same functional group have the same symmetry.
If a fundamental transition is close to the energy of an overtone or combination
transition in the same functional group, the overtone or combination mode
borrows energy from the fundamental mode. This causes the intensity of the
overtone and combination bands to increase relative to the intensity of the fundamental band and the differential frequency shift between the bands will increase
as well.
Darling-Dennison resonances involve different overtone transitions within
a molecule and are attributable to quartic and higher terms in the potential energy
function. Such resonances are pronounced when interacting energy levels are
very close to one another and when the degree of anharmonicity is large.
Darling-Dennison resonances tend to increase along with the level of excitation.
Thus, as the level of excitation increases, resonance bands will tend to be more
pronounced and will shift even farther from the position described by the
normal mode transition. In terms of practical applications, however, the major
implication of resonance is that, in some cases, two absorbance bands might
exist where only one band is expected.
Third Overtone Decoupling
The NIR spectra tend to be relatively more featureless in the second and third
overtone regions. The reasons are: (i) the intensity of overtone absorption bands
decrease by orders of magnitude as the cumulative change in quantum numbers
increases, and (ii) the vibration of multiple bonds with greater degrees of anharmonicity, especially bands involving hydrogen, begin to decouple in the third
overtone region (780 1300 nm). As the energy level increases, vibrational
modes begin to decouple, and the bonds begin to move more independently of
one another, resulting in much broader, less intense absorbance bands. Furthermore, symmetry will be reduced, leading to a reduction in the occurrence of
resonance effects.
Functional Group Effects
The important attributes of functional groups that affect the position and
intensities of NIR bands include: atomic mass, bond strength, dipole moment,
symmetry, and anharmonicity; the latter three have the greatest effect on band
intensity. Although it is not a factor for fundamental modes, the effect of anharmonicity is greater with asymmetric stretching vibrations than for symmetric
stretching vibrations of the same group (18,19).
318
Cogdill and Drennen
Hydrogen Bonding
The interaction of a hydrogen bond acceptor with an O22H group makes the
O22H stretching vibration more harmonic, thus greatly reducing the intensity
of stretching overtone bands in the NIR spectrum. Hydrogen bonding is the
most significant effect in NIR spectroscopy that does not involve a change in
composition. The most common hydrogen bond donor groups are 22OH and
22NH; there are many possible hydrogen bond acceptor groups, including:
O22H, N22H, C22O22C, C55O and 22N, groups (13). A summary of the
effects of hydrogen bonding on band location and shape follows:
.
.
.
Stretching bands of groups involving the acceptor atom decrease in

frequency,
Donor-hydrogen bending bands increase in frequency, and
Donor-hydrogen stretching bands decrease in frequency and become
broader.
Hydrogen bonding causes a decrease in the intensity of overtone and combination bands involving the donor-hydrogen stretching vibration, which depend
greatly on anharmonicity to exist. Hence, overtone and combination bands
for O22H and N22H groups are more intense for the free (nonhydrogenbonded) forms than for the hydrogen-bonded forms (references).
Neighboring Group Effect
Even identical functional groups can produce very different spectra depending on
the composition and structure of the sample matrix. The exact position of NIR
absorbance bands in specific materials can be influenced by the composition of
the neighboring functional groups. This is especially important if the neighboring
functional groups are either strongly electron-withdrawing or electron-donating,
both of which greatly affect bond strength and dipole moment.
Water
Water is the most important and misunderstood absorber in the NIR region. It is
unique due to its low reduced-mass, strong dipole moment, and the fact that it is
both a hydrogen-bond donor and acceptor. Thus, water has great propensity for
hydrogen bonding. As a very anharmonic molecule, water absorption bands commonly exhibit Darling-Dennison resonance effects. Water has intense absorption
bands at approximately 1940-, 1420-, 1200-, and 970 nm. Indeed, determination
of moisture in solids was one of the first applications of NIR spectroscopy and
has provided many published references across diverse fields of application.
Pharmaceutical process analytical applications of NIR for moisture monitoring
are discussed at length in subsequent sections.
The varying presence and character of water molecules in pharmaceutical
solids will generate significant and complicated effects on their observed NIR
absorbance bands and baselines. Water absorption bands can be quite broad
319
and shifted by a considerable amount from their anticipated location because of

matrix interactions. Water is present in pharmaceutical solids in two principal
forms: (i) free water adhered to the surface and interstitial spaces of the particle
matrix by London dispersion forces, and (ii) bound water held tightly inside
individual crystals. The level of free and bound water in pharmaceutical solids
has implications for drug dosage form stability, solubility, and bioavailability.
Free and bound water exhibit different levels of hydrogen bonding and
unique intermolecular distances. Recent works have demonstrated that the
spectral signatures of the two forms can be determined via NIR spectroscopy
(20 23). Hogan and Buckton (22) demonstrated that the mono-, di-, tri-, tetra-,
and pentahydrate states of crystalline and amorphous raffinose could be
distinguished from NIR spectra (Fig. 1). Other groups (21,23) have utilized
NIR spectroscopy, along with multiple chemometric methods, to develop
process analytical techniques for effectively discerning surface and bound
water in various drugs and drug products. Each of these groups showed that
the presence of free surface water causes increased NIR absorbance intensity
near 1905 nm, and increasing levels of bound water lead to relatively higher
absorbance at lower frequency bands. For example, Zhou et al. (23), studying
an unnamed model drug compound, reported decreasing absorbance intensity
near 1936 nm as the relative levels of bound water (in the form of crystalline
hydrates) decreased (Fig. 2). Additionally, in two separate studies (20,21),
researchers from the Rantanen group showed that bound water solvated in
Figure 1 The NIR second derivative spectra corresponding to different hydrate levels.
Abbreviation: NIR, near-infrared. Source: Courtesy of Elsevier, 2001.
320
Cogdill and Drennen
Figure 2 Principle component scores plot of PC2 versus PC3 during the drying process
of a wet cake of model drug by in-line NIR monitoring. Abbreviation: NIR, near-infrared.
Source: Courtesy of Wiley-Liss, Inc. and the American Pharmaceutical Association, 2003.
caffeine is detected near 1960 nm, whereas bound water in theophylline is

detected near 1970 nm.
In 1998 Buckton et al. (24) published the results of their work investigating
changes in the form of crystalline and amorphous lactose using NIR spectroscopy. The researchers were able to produce samples of amorphous, form-a
and form-b lactose by exposing spray-dried amorphous lactose to conditions
of elevated relative humidity for time periods of up to three months. The crystalline form of the lactose samples was confirmed using differential scanning
calorimetry (DSC) and thermal gravimetric analysis (TGA). In contrast to
earlier studies on crystallinity, the NIR reflectance analyses of the samples
suggested that the major source of sensitivity was due to changes in the character
of water in the lactose matrix. As the amorphous lactose crystallized into the a
form and eventually converted to the b form, intense changes in the band
shapes near 1440- and 1920 nm were observed, which agree with the observations of other researchers analyzing different materials.
In more recent work, Rasanen et al. (25), utilized NIR spectroscopy along
with a novel multichamber microscale fluidized bed dryer to investigate the
effect of temperature and moisture content of the drying air on the dehydration
of disodium hydrogen phosphates with wet theophylline granules and three
different levels of hydrates. The removal of bound crystalline water was monitored by in-line Fourier transform near-infrared (FT-NIR) spectroscopy and
321
off-line X-ray powder diffraction. The real-time drying curves illustrated the
stepwise change in moisture content as hydrates were liberated from the solids.
Temperature Effect
The sample temperature variability produces a complex effect on the NIR spectra
of pharmaceutical solids, especially for solids with high moisture content, as is
typical during granulation and drying operations. As the temperature increases,
the degree of anharmonicity of absorbing groups will tend to increase, leading
to an upward shift in the frequency center and a broadening of absorbance
bands. Because of their reduced mass, molecules with significant hydrogen
bonding are most affected by temperature variation. Thus, the temperature
effect on NIR absorbance is most relevant for pharmaceutical solids when it is
varying in conjunction with high moisture concentration or when critical
analytical bands for a weakly absorbing constituent are heavily overlapped by
water absorbance bands.
Maeda et al. (26), utilized multiple chemometric and curve fitting tools to
investigate the relationship between water temperature and spectral changes as a
result of the temperature effect on hydrogen bonding. They observed a strong,
multifactor effect (Fig. 3) that could be used to accurately predict water temperature, which has been verified in more recent work by other groups (27). During
operations that cause substantial fluctuation of temperature and moisture, such
as high-shear granulation or drying, the evolution of the water spectrum with
shifting temperature will often induce significant nonlinearity in the NIR
Figure 3 FT-NIR spectra of water measured over a temperature range of 5 to 858C

in increments of 58C. Abbreviation: FI-NIR, Fourier transform near-infrared. Source:
Courtesy of NIR Publications, 1995.
322
Cogdill and Drennen
response curve. If the NIR analyses are to be conducted over wide temperature
and/or moisture concentration ranges, it will generally be necessary to increase
the complexity of the mathematical method for resolving the spectra by either
adding model factors, or by estimating nonlinear model terms.
Particle-Scale Phenomena
The discussions of molecular-scale phenomena in the previous section dealt
with the mechanisms of absorption and the confounding factors associated with
optical absorption in the NIR. Such phenomena are general to most sample
types analyzed by NIR spectroscopy. This section on particle-scale phenomena
will be focused on the unique interactions between NIR radiation and solid
particulate samples.
NIR Transmittance in the Solid State
One of the most important aspects of the interaction between NIR radiation and
particulate solids is the unique combination of relatively weak absorbance intensity with high scattering efficiency. This combination allows the NIR radiation
to probe the interior of many solid samples with little or no sample preparation.
The spectroscopic analyses of solids in the NIR region are performed using two
primary modes of sampling geometry: transmittance and diffuse reflectance (DR).
The transmittance measurement through particulate solids is a distinctive
capability of spectroscopy in the NIR relative to UV/VIS and IR. Sample transmittance, T, is estimated as the ratio of intensities for light transmitted through
an empty path (e.g., cuvette), I0, and light transmitted through an equal distance
of a particulate sample, Is. Transmittance data are most often reported in terms
of Beer-Lambert absorbance:

1
Is
A log
abc
(1)
log
T
I0
where, for a single wavelength, l:
A Beer-Lambert optical absorbance, AU
T transmittance ratio
a absorption coefficient, cm21
b pathlength (or sample thickness), cm
c concentration of absorbing species.
In contrast to UV/VIS (transmission) absorption spectrophotometry, which
is typically performed in solution, the transmittance of NIR radiation through
a sample matrix comprised of tightly packed solid particles is expected to
deviate nonlinearly from the Beer-Lambert law of absorption. The source of
these deviations becomes apparent when the derivation and simplifying assumptions of the Beer-Lambert law, or simply Beers law, are considered.
323
Figure 4 Illustration of optical absorption from an infinitesimally thick layer of a

particle matrix.
The derivation of Beers law is best visualized by considering a twodimensional plane of a sample material containing a homogeneous distribution
of an absorbing analyte (Fig. 4). For a collimated beam of light directed
normal to a sample plane of infinitesimal thickness, dt, there is a finite probability
that a fraction of the photons, d I, will be absorbed by the analyte, thereby reducing the intensity of the transmitted radiation. This probability is proportional
to the intensity of the photons, I0, the number of analyte molecules encountered,
N, and the effective absorption cross section, s, where:
d I sNI0 dt
(2)
The relationship can be extended to estimate the intensity of the beam remaining
after transmittance through an optically thick layer, b, by dividing each side by
the initial intensity, I0, and integrating from t 0 ! b, generating the following
expression:

I0
1
(3)
ln (I0 ) ln (Is ) sNb or ln
ln
sNb
T
Is
This result [Eq. (3)] is converted to the more user-friendly version [Eq. (1)] by
replacing the effective absorption cross-section, s, with the catch-all absorption
coefficient term, a, according to:
a
2:303 N
(4)
The most notable deviations from Beers law for NIR transmittance
through clear liquids are due to changes in absorption coefficients across wide
ranges of concentration. Otherwise the response between optical absorption, A,
and concentration c, is expected to be linear. The transmittance through particulate solids is less accurately described by the Beer-Lambert law, however,
because it assumes that all radiation encountering an absorbing particle will
324
Cogdill and Drennen
Figure 5 The optical interactions, of which the three fluxes are comprised of: (A) incident beam absorption, (B) specular reflectance, (C) transmittance, (D) forward scatter, (E)
back scatter, (F) anisotropic scattering intensity fields.
either be transmitted or absorbed. There is, however, finite probability

that incident photons will be scattered (inelastic or elastic), or reflected either
forward or backward relative to the direction of propagation of the incident
beam, I0 (Fig. 5).
Thus, the incident photons are not transmitted directly through particulate
samples, rather, they take a tortuous path through the material with multiple
opportunities for scattering and backward reflection (Fig. 6). Because the intensity of scattered radiation decreases along with absorptivity, the transmittance
through a solid sample will become less diffuse as the absorptivity decreases.
These effects tend to increase the path a photon will travel before being
emitted from the sample. Thus, the effective pathlength for transmittance
Figure 6 Schematic depiction of the most common modes of sample interface for NIR
spectroscopy of solids. Abbreviation: NIR, near-infrared.
325
through a plane of compressed particulates of thickness b will be some nonzero,

normally distributed amount greater than b. Furthermore, as b increases, the
transmitted radiation will become more diffuse until the diffuse thickness (28)
is achieved, at which point the maximum amount of incident radiation reflected
back toward the source occurs. Hence, solid-state transmittance in the NIR region
might more aptly be termed diffuse transmittance spectroscopy.
The absorptivity, reflectance, and scattering characteristics for a particular
sample are dependent on multiple material qualities, such as particle size and
morphology, packing density, and index of refraction. Hence, the effective
pathelength and the level of nonlinearity relative to Beers law are difficult to
predict in practical situations. When taken over narrow concentration ranges,
however, the simple log(1/T) approximation works fine for the majority of
applications for NIR analysis of pharmaceutical solids.
The transmittance NIR spectroscopy is usually performed in the third
overtone region between 780 1100 nm, where absorption bands are orders of
magnitude less intense than the lower-order overtones. Although scatter coefficients do increase nominally in the NIR with increasing frequency, the scattering
efficiency in the third overtone region is low because the amount of radiation loss
to backward reflection is reduced along due to the decreasing absorption coefficient, as described by the complex Fresnel equations (29).
DR Spectroscopy
The DR sampling geometry is arguably the defining quality of NIR spectroscopy
for process control applications. The DR sampling enables true noncontact, nondestructive analysis of solid materials. The DR analysis poses fewer constraints
on sample positioning for such applications as high-speed tablet analysis (30).
Remote-sensing via DR permits quantitative analysis of material qualities for granular or particulate samples in fluidized beds, mixing vessels, or on moving belts.
The sample reflectance, R, is most often estimated as the ratio of intensities
for light reflected from a nonabsorbing, diffusely reflecting surface (e.g., white
ceramic or inner surface of an integrating sphere), I0, and light reflected from a
sample, Is. Even though reflectance measurements bear little resemblance to
the scenario depicted in the derivation of Beers law (shown earlier), reflectance
data are most often reported in terms of absorbance units, log(1/R). Furthermore,
as is typical for absorbance spectra collected in transmittance, the absorbance
spectra collected in reflectance are treated as being linearly correlated with the
concentration according to Beers law. For most practical applications, nonlinearities are either ignored over a narrow concentration range or are approximated by
additional empirical model factors or nonlinear terms.
Scientists have been attempting to develop rigorous theoretical models of
DR for more than 200 years. These early development efforts are summarized
in Birth & Hecht (28). The earliest efforts to explain the mechanism of DR
treated it as strictly a surface phenomenon. In 1760, Bouguer posited that DR
326
Cogdill and Drennen
could be considered as the summation of mirror-type reflections from an assembly of microcrystalline faces statistically distributed over all possible angles.
Lambert published a mathematical relationship describing the link between the
intensity of an incident beam of light, angle of incidence, angle of observation
(known today as the Lambert cosine law) in the same year. Although the Lambert
cosine law has been shown to be theoretically sound, it fails in practice because
an ideal diffuse reflector, or true Lambertian reflector, has never been found.
It wasnt until more than 100 years after the theories of Bouguer and
Lambert were published that scientists began to consider that optical remission
occurred through interaction with the surface and the interior of samples
placed within an optical path (28,29). Seeliger described radiation as penetrating
the surface of materials, where it is either absorbed or returned to the surface via
reflection, refraction, or diffraction from the surfaces of the internal microstructure (28,29). Diffusely reflected radiation is now considered to be the
superposition of two components: (i) reflection from external surfaces, and (ii)
elastic scatter. Numerous theories of elastic scattering phenomenon have been
developed to explain the relationship between the frequency of radiation and
the scattering intensity. The major difference between the theories is their
treatment of particle size and morphology.
Mie theory, which relates primarily to the scattering of radiation by isolated, spherical particles, is suggested to be the most relevant to NIR analysis
of solids, because the Mie equation does not include a limitation on particle
size (31,32). According to Mies theory, scattered radiation is not distributed isotropically; rather, a complex scattering intensity pattern is produced, with forward
scatter preferred over scattering in the reverse direction (similar to Fig. 5):
Iuscat
l2
2 2 i1 i2
I0
8p R
(5)

Iuscat Intensity of scattered radiation at distance R and angle u
from the center of the scattering particle.
i1, i2 Complex functions of the angle of the scattered
radiation, the spherical harmonics, or their derivatives
with respect to the cosine of the angle of scattered
radiation, the refractive index of both the sphere and
surrounding medium, and the ratio of the particle
circumference to wavelength.
For practical NIR applications, however, particles are not isolated but are
generally in intimate contact with one another. Furthermore, Mies theory does
not address the consequences of multiple forward- or back-scattering events as
individual particles propagate through a solid matrix. An intuitive assessment
of the overall relationship between wavelength and scattering intensity can be
327
Figure 7 Pure-component NIR reflectance spectra for: (A) lactose, (B) theophylline, (C)
starch, (D) MCC PH 101, (E) magnesium stearate. Abbreviations: NIR, near-infrared;
MCC, microcrystalline cellulose.
gained from Equation (5): As the wavelength decreases, the intensity of scattered
radiation will increase, thereby decreasing the apparent absorbance in the third
overtone region. This phenomenon is one of the factors contributing to the
shape of the upward-sloping baseline that is typically observed in the NIR
reflectance spectra of solids (Fig. 7).
Experiments by Theissing, which concur with Births concept of diffuse
thickness described earlier, showed that as the number of times a photon is
scattered increases the distribution of scattered radiation will deviate further
from Mies theory, becoming more isotropic and with greater proportion of the
radiation being scattered in the reverse direction (33,34). In general, a
comprehensive theory of elastic radiation scattering is lacking for most NIR
applications; hence, complete coverage of the various theories of elastic scattering are beyond the scope of this chapter.
Theoretical Models of DR
Besides the theoretically flawed log(1/R) format in which most NIR absorbance
data are reported, the absence of a quantitative theoretical relationship explaining
multiple-scattering phenomena has led to the development of phenomenological
models from a generalized radiation transfer equation in some ways similar to
328
Cogdill and Drennen
Equation (2). A modification of the radiation transfer equation to accommodate

forward and reverse reflection, as well as anisotropic scattering and absorption,
leads to a rather messy integration problem that is difficult to employ in practice.
In 1905, Shuster reported a simplified solution of the radiation transfer equation
for the case of reflectance, summarized here from a later reference (33,34). By
assuming the total radiation flux is comprised of two components, one flux traveling in the forward direction, I, and a second flux traveling in the reverse direction,
J, Shuster derived the following two differential equations:
d I
(k s)I sJ
dt
dJ
k sJ sI
dt
(6)
(7)

2a
Absorption coefficient for a single-particle scattering
a sscat
event
sscat
s
Scatter coefficient for a single-particle scattering event
a sscat
sscat Effective absorbance cross-section due to scattering

a Effective molecular absorbance coefficient.
Schuster made the additional simplifying assumption that only singleparticle scattering occurs. By setting boundary conditions such that, at infinite
thickness (t ! 1), there is no further change in flux in the reverse direction
with increasing depth: DJ/Dt ! 0, Equations (6) and (7) can be solved for reflectance at infinite depth, R1:
R1
1 (k=(k 2s))1=2
1 (k=(k 2s))1=2
(8)
which can be rewritten as:

(1 R1 )2 k
2a

s sscat
2R1
(9)
Schusters result is virtually identical to a later, more well-known expression derived by Kubelka and Munk, published in 1931 (35). Whereas Schuster
derived his expression using k and s, which are applicable to single-particle
scattering events, Kubelka and Munk employed the more general versions, K
and S, which are defined as the absorption and scattering coefficients, respectively, for a densely packed sample layer as a whole. Although this change
in parameters leads to an entirely different path for derivation (33,34),
the result is (operationally) identical and is in a form that can be correlated
329
with concentration:
(1 R1 )2 K 2:303ac

S
S
2R1
(10)
where a is the absorptivity at wavelength l; the 2.303 scaling factor corresponds

to a conversion to the more-traditional base-10 representation of absorptivity.
Because the K/S is generally related to concentration in practice via linear
regression, the scaling factor can be omitted, and the absorptivity be treated as
an unknown.
Kubelka and Munk made several assumptions in deriving their solution
to the radiation transfer equation (some of which are shared by Schusters
derivation):
1. The total radiation flux is comprised of two components, one flux
traveling in the forward direction, I, and a second flux traveling in
the reverse direction, J.
2. The sample is illuminated with monochromatic radiation of intensity I0
at wavelength l; otherwise the next point (3) applies.
3. Particles must be assumed to be much larger than the wavelengths of
illumination (so the scattering coefficient will be independent of
wavelength).
4. Scattered radiation is distributed isotropically (implying that the
specular component of reflection is zero).
5. Particles are distributed heterogeneously (randomly) throughout the
sample as a whole.
6. Particles have much smaller diameter than the overall thickness of the
sample, t.
7. The sample is subject only to diffuse illumination.
8. The area of the sample layer is large compared to the sample thickness,
t (so as to avoid edge effects).
In light of the discussion from the preceding paragraphs, it is apparent that
many of the assumptions are in conflict with physical reality, resulting in deviation from linear response to concentration. Some early practitioners posited
that observed nonlinearity was due to the nonzero specular component of firstsurface reflection. Even after removing the specular component of reflectance
(via crossed polarizers), however, the Kubelka-Munk absorbance coefficient, K,
will not be a linear function of concentration. Other researchers have developed
augmentations or complete reformations of the theory of DR in an effort to relax
some of the assumptions of the Kubelka-Munk model, some of which have seen
significant application (36). It has been shown (34), however, that their refinements result in little more than a rescaling of the Kubelka-Munk function
stemming from variations in the assumed lighting conditions. The theoretical
shortcomings notwithstanding the Kubelka-Munk reflectance transformation,
330
Cogdill and Drennen
often denoted by f (R1), has been used extensively because it works reasonably
well when applied correctly and has relatively few parameters such that it can
easily be employed in practical situations.
Dahm and Dahm have shown in their recent work that both log(1/R) and
the Kubelka-Munk transformation are inherently nonlinear functions of concentration. The log(1/R) transformation is inherently nonlinear: (i) because it treats
the absorbance coefficient of a sample as being an additive sum of the absorbance
coefficients of all absorbing species in the sample, which is true only for weakly
absorbing constituents, and (ii) because it fails to account for the effects of
scatter, which results in nonconstant effective pathlength (29). They demonstrate
that the inherent nonlinearity of the Kubelka-Munk relationship is due to the
erroneous assumption that the scattering coefficient, S, is generally assumed in
practice to be constant, and independent of the level of absorption. Their work
(37) is an extension of the mathematical theories of plane-parallel layers introduced by Stokes in 1860 and subsequently refined by Benford in 1946, then
Wendlandt and Hecht in 1966 [references found in Dahm & Dahm (29)].
Their work contrasts with Kubelka-Munk theory in that it treats samples
not as a homogeneous mass of scatterers but rather as a series of parallel infinite
planes. The total radiation flux (from a source normal to the plane) encountered
by any layer is accounted for as the sum of fluxes transmitted through the layer,
T, remitted (reflected back toward the source), R, and absorbed, A. Hence,
the optical flux incident upon a sample comprised of two identical planes
(or layers), n and n 1, can be accounted for by the Benford equations:
Tn1
Tn T1
(1 Rn R1 )
Rn1 Rn
Tn2 R1
(1 Rn R1 )
(11)
An1 1 Tn1 Rn1

By considering the effect of doubling or halving the thickness of an
individual layer on each of the flux components, Dahm and Dahm arrived at
an iterative relationship for the estimation of absorbance, which is constant for
samples of arbitrary thickness. Their postulated theorem is known as the absorbtion/remission function, A(R, T ), or, simply, the Dahm equation:
A(R, T)
(1 R)2 T 2 (1 T R)A (2 A 2R)A
R
R
R
(12)
The Dahm equation is proportional to the ratio of the linear absorption, KAR, and
remission, BAR coefficients, as well as the fraction of light absorbed, A0, and
remitted, R0, by a sample of infinitesimal thickness:
A(R, T) /
2KAR 2A0
/
BAR
R0
(13)
Hence, in practical applications, it is necessary to know two of the three

flux components (A, R, or T) to evaluate the Dahm equation for an individual
331
sample. By doing so, the relationship between absorbance and scatter for a
sample of infinite thickness is implicitly defined. The Dahm equation relates to
the Kubelka-Munk relationship by:
2A(R1 , 0) / F(R1 )
K
S
(14)
Utilizing the Dahm equation to estimate the Kubelka-Munk absorbance and scattering coefficients, K and S, respectively, is theoretically advantageous because it
accounts more effectively for the wavelength-dependent relationship between
absorbance and scatter.
Although the Dahm equation and the Kubelka-Munk transformation (to a
lesser extent) can be shown to be superior to the log(1/R) transformation in
terms of linearity, the bulk of the DR data is likely to continue to be dealt with
using the Beers law representation for quantitative applications. This can be
attributed to the simple fact that most NIR methods are ultimately empirical in
nature and, in the case of NIR analysis of pharmaceutical solids, are expected
to cover a relatively narrow range centered about some nominal condition. As
the understanding of DR (and diffuse transmittance) phenomena continues to
mature, however, these advancements may become essential parts of routine
method development by enabling calibration using fewer training samples or
by extending the limit of detection (LOD) of very subtle material qualities.
Physical Sample Factors Affecting the NIR Spectroscopy of Solids
Based on the discussion in the preceding paragraphs, it should be clear that the
NIR spectra are the convolution of chemical and physical interactions between
the sample and NIR radiation. Although the value of the chemical information
in NIR spectra is obvious, the physical factors, which affect NIR spectra may
also yield useful information related to the quality (state) of the sample material.
In other cases, the presence of strong variations in concomitant physical
parameters may overwhelm the analytical signal of interest.
Particle Size
The particle size distribution (PSD) and/or mass-mean particle size (MMPS) are
critical quality control factors for raw pharmaceutical solids or for process
control during unit operations, such as milling and granulation. Drug and excipient PSD variability can affect dosage form performance by altering dissolution
rate or bioavailability, or it can affect drug product quality in manufacturing
by altering the flow properties of the materials. It is a well-known fact that the
angle of monochromatic optical diffraction for a single-particle event is inversely
proportional to particle size. The actual angle of diffraction will depend not
only on particle size and wavelength of incident light but also on particle morphology and refractive index, as described by Mie and Fraunhofer theories of
optical scattering (15). Thus, as particle size varies, the relative level of
332
Cogdill and Drennen
forward and backward scattering events will vary as well, causing a change in
apparent optical absorption. Larger particles, for example, will have a greater
proportion of forward scatter, which reduces the probability that a diffusely scattered photon from an incident beam will be returned to the surface of a sample in
DR (Fig. 8). The particle size variation will also affect particle packing characteristics, leading to more complex spectral effects. Hence, there has been great
interest in using NIR spectroscopy for rapidly determining MMPS or PSD of
pharmaceutical solids and granules.
Because much of the first widespread application of NIR spectroscopy was
in the area of rapid content analysis of ground grain samples, the effect of particle
size on NIR spectra was for a long time considered to be a detrimental factor
affecting model performance. Possibly the earliest study (38) of the effect of
particle size on NIR spectra was published in 1981 by Osborne, et al. While
their study was intended to develop an NIR calibration for wheat kernel
texture, their reference measurement of texture was actually a particle size
fraction of the flour from the ground wheat kernels used for NIR analysis.
Norris and Williams (8) published some of the first work directly investigating the effect of sample particle size on calibration model performance.
Their work, published in 1984, describes a method of optimal treatment of
NIR spectra to reduce the effect of MMPS variation on NIR analysis of hard
Figure 8 Comparison of NIR process spectra for two particle sizes of MCC: (A) PH
200180 mm, (B) PH 10150 mm. Abbreviations: NIR, near-infrared; MCC, microcrystalline cellulose. Source: Courtesy of Control Developments, Inc.
333
red spring (HRS) wheat for the prediction of Kjeldahl protein content. They
observed that increasing particle size tended to increase the slope and baseline
of NIR reflectance spectra of HRS wheat so that the absorbance values at
longer wavelengths were more inflated by an increase in particle size than was
observed at shorter wavelengths. Although they did not provide a detailed explanation for their observations, they deduced (via regression analysis) that the
primary relationship between particle size and the magnitude of the effect on
NIR spectra was due to the level of apparent absorption [log(1/R)] rather than
wavelength. Thus, for solids with many strong NIR absorbance bands, it
should be expected that the PSD variation will have a complex, nonlinear
effect on the log(1/R) spectrum. Although this suggests that Kubelka-Munk
transformation [Eq. (10)] of the reflectance spectra should be more effective in
reducing the effect of particle size variation on NIR analysis, Norris and
Williams went on to show that a ratio of two selected wavelengths in second
derivative spectra was superior.
Bull (39,40) continued to explore the observations of Norris and Williams
utilizing Kubelka-Munk theory (and the same dataset as Norris and Williams) in
her 1990 and 1991 publications. Bull ultimately proposed a ratio of KubelkaMunk functions at two wavelengths to mitigate the effect of particle size on
the measurement of protein and water concentration in milled samples of HRS
wheat (40). Although her result was not empirically superior in predictive performance, Bull suggested the solution would be less sensitive to particle size variation over a wider range of experimental parameters.
Ciurczak (41) published some of the first pharmaceutical work relating
the particle size of pharmaceutical raw materials to variation in NIR spectra.
Ciurczak demonstrated a linear relationship between the absorbance at any wavelength and the reciprocal of particle size as measured using low-angle laser light
scattering (LALLS). Subsequent references on the use of NIR for particle size
analysis include the application of NIR for real-time median particle size determination during granulation (42) and that of MMPS of an API in a binary mixture
(43). Pasikatan et al. (38), published a detailed review of NIR for particle size
analysis of powders and ground materials covering pharmaceutical and agricultural applications. Their review includes a theoretical discussion of the basics of
NIR spectroscopy related to particle size determination and the aspects of
sampling, reference methods, calibration development, and method validation.
The effect of varying particle size on the NIR spectra of solids is often
described as a sloping, multiplicative baseline offset. Although this is qualitatively correct, the true effect of particle size variation is convolved with
absorbance, morphology, density, and other sample factors. Thus, various chemometric treatments have been applied to extract MMPS information from
NIR spectra of pharmaceutical solids, including Kubelka-Munk transformation
(44), multiple linear regression (MLR) (45,46), principal component regression
(PCR) and partial least-squares (PLS), as well as artificial neural networks
(ANN) (47). Some form of laser diffraction or sieve analysis was used for
334
Cogdill and Drennen
reference testing in all cases. Although Frake et al., did not observe significant
advantages by using more complex modeling algorithms, they succeeded in
demonstrating that the ability to predict MMPS is not diminished when the
NIR spectra are preprocessed to remove the effects of scatter. Although they
did not discuss this aspect at length, this is an important concept because
certain aspects of instrument and interface stability may necessitate some preprocessing before the analysis of NIR spectra. Chemometric topics, such as preprocessing and model algorithm selection, will be discussed in later sections.
Beyond prediction of MMPS and single-parameter particle size indicators,
some researchers have explored methods of predicting entire cumulative particle
size distributions from NIR spectra (46). Their method involved correlating the
NIR reflectance spectra of a series of samples against 11 quantiles of their
respective cumulative PSDs measured using forward angle laser light scattering
(FALLS). Thus, 11 calibration equations were estimated to predict the 5, 10, 20,
30, 40, 50, 60, 70, 80, 90, and 95% cumulative size fractions using both MLR and
PCR. They observed that, for both algorithms, the lowest calibration error was
achieved near the MMPS. Although their results demonstrated an ability to
predict cumulative PSD from NIR spectra using a multitude of calibration
equations, a more detailed investigation of the correlation among PSD quantiles
might have yielded a simpler solution along with some sense of the specificity
(and the basis for specificity) of prediction at individual quantiles. Higgins
et al. (48), recently demonstrated the use of NIR spectroscopy and PLS
regression to accurately predict the volume-weighted particle size of nanoparticles in a high solid-content colloidal dispersion.
Hardness and Other Physical Parameters
Similar to particle size, variation in sample hardness has been observed to have a
measurable effect on NIR spectra. Tablet hardness is a common product release
parameter and is important for process control, because varying hardness might
affect product friability, disintegration, dissolution, and ultimately, bioavailability. The hardness and density (or porosity, solid fraction) are common
parameters for process control during roller compaction.
Drennen (49) published the first examples of using NIR spectroscopy and
chemometrics for nondestructive tablet hardness in 1991. Subsequent publications by Ciurczak and Drennen in 1992 (50,51) and Drennen and Lodder in
1993 (52) further explored the application using both quantitative and qualitative
methods. It was observed in all of these experiments that changes in dosage form
hardness are indicated by a sloping baseline shift of the apparent absorption
spectrum measured in reflectance (Fig. 9). As the tablet (or compact) hardness
increases, the apparent absorption baseline increases. Although the root cause
of the spectral effect is not conclusively established, some references do posit
hypotheses. The results of some of the work recently conducted in the
Drennen laboratory suggests that, as tablet hardness increases (and the total
335
Figure 9 Effect of increasing hardness on NIR Transmittance and reflectance spectra

(arrows indicate direction of increasing hardness). The transmittance spectra have been
scaled for comparison. Abbreviation: NIR, near-infrared.
surface area of interparticle contact increases), a lesser fraction of the radiation is

backscattered due to the air particle interface, reducing the magnitude of reverse
flux. This concurs with similar observations in earlier work by Kirsch and
Drennen (53). At the same time, a greater fraction of the radiation propagates
through points of interparticle intimate contact, increasing the magnitude of
forward flux. As fewer scattered photons reach the reflectance detector via
reverse flux, the apparent absorbance increases. This hypothesis is supported
by the observed relationship between the tablet hardness and NIR transmittance,
whereby increasing hardness reduces the apparent tablet absorbance (Fig. 9).
This intuitively suggests that as powders are compressed more firmly, the
resulting matrix becomes more glasslike. As is for the effect of particle size,
the rate of baseline increase is more pronounced near strong absorption bands.
Although baseline shifting is the most pronounced spectral change observed
with varying tablet hardness, other spectral changes (e.g., peak shifting) have
been observed (54).
The development of NIR calibration models for hardness have utilized
MLR and PCA (55), PLS (30,56,57), and more complicated algorithms, such
as ANN (58). A survey of the literature suggests that there is little accuracy to
be gained by using very complex chemometric methods, such as ANN, for prediction of hardness. Indeed, it has been shown that the hardness can be predicted
336
Cogdill and Drennen
with reasonable accuracy by simply fitting linear or nonlinear functions to the

NIR baseline (53,57,59) and then correlating hardness to selected coefficients
of the baseline fit. Because the NIR spectral baseline can be influenced by
tablet alignment variation (55) and certain instrumental drift factors (57), a
more robust calibration may be developed when preprocessing is used to
reduce unspecific baseline variations. Despite the subtlety of the wavelengthspecific nature of the hardness effect on NIR spectra, hardness calibrations
have demonstrated specificity in the presence of concomitant variation in
coating thickness (55), drug content (53,57), lubricant content (60), and both
the type and level of filler material (59).
Donoso et al. (61) recently published the results of experiments involving
the creation of calibrations for porosity and hardness of theophylline and
placebo tablets. Although the validation results suggest that some nonlinearity
was not accounted for in the model, the data indicate NIR sensitivity for
porosity. Because Donoso et al. neither supplied many details of the hardness
and porosity calibrations, nor did they present the regression coefficients or
model parameters, the specificity of the predictions is suspected. It is because
the porosity variation used for calibration was generated by varying compression
force, similar to the methods used for hardness calibration.
Otsuka et al. (62) published similar findings in applying NIR spectroscopy
for the evaluation of pharmaceutically relevant properties of antipyrine granules,
such as MMPS, PSD, true density, angle of repose, and compressibility, as well as
the porosity and hardness of the tablets produced from the granules. They
observed NIR sensitivity to each parameter, and the collinearity among the physical parameters were tested. This is not surprising because the variation induced
in all of the granulations was developed by varying the amount of water added
during granulation. Indeed, spectral decomposition of the physical parameter
data might have revealed that fewer independent factors were present in the
measurements. More detailed multivariate experiments with variation in
multiple production factors are likely to improve the level of understanding
and specificity of the NIR response.
Effective Mass Sampled During NIR Analysis
For any new measurement technology or application to be deployed with confidence (and validated), there must be at least a mechanistic understanding of the
measurement scale involved. This concept is critical for composition uniformity
measurements (such as blend and content uniformity release testing) because, as
is well known, variability between sample units increases as the size of the
sample unit decreases (63). Thus, a sample considered to be homogeneous at
the scale of a few hundred grams but would be quite heterogeneous at the
level of tens of grams.
The safety and efficacy of most drug therapies is dependent on a specific
level of uniformity among individual dosage forms. For this reason the uniformity of finished and in-process pharmaceutical materials is determined at a scale
337
determined by the magnitude of a unit dose, sometimes referred to as the scale

of scrutiny (64,65). For in-process measurements, such as blend uniformity
analysis, precedent (66) suggests that the effective sampling mass should be no
more than three times the mass of a single dosage form. Thus, while utilizing
larger collection optics may increase signal levels, smaller sensors may be desirable for quantifying blend uniformity to ensure the appropriate scale has been
assayed. The case is reversed for the analysis of individual dosage forms,
however, where it is desirable to include as much of the product mass in the
analysis as possible. Indeed, there has been some sentiment among early practitioners that NIR transmittance analysis of intact dosage forms would be superior
to reflectance measurements based on the supposition that more of the dosage
form would be effectively sampled in transmittance.
Besides being a qualification or validation issue, the effective mass
sampled during NIR analysis has a complex, nonlinear effect on the shape of
reflectance or transmittance spectra. Whereas the independent sample parameters
affecting a spectrum (composition, PSD, hardness, etc.) can be adjusted relatively
easily, the effective mass sampled is largely a convolution of all of such effects on
the NIR spectrum. Hence, effective mass sampled must be inferred from indirect
experiments, rather than measured directly. In her 1990 experiments investigating the effect of particle size variation on NIR spectra, Bull (39) developed
the following relationship to quantify the observed relationship between multiple
optical properties and the depth of penetration, dl, of the measurement into a
sample at a given wavelength (for a particular measurement geometry). This
relationship is directly related to effective mass sampled:
F(R1 )
K
x
S 2dl rl
(15)
where K/S is the Kubelka-Munk function of reflectance at infinite depth, x is the

mean particle size of the sample material, and rl is the reflectance at a single
air-particle discontinuity. The depth of penetration, dl, is specified as the 1/e
information depth; in other words, the sample depth at which the fractional
change in reflectance is less than 1/e (e 2.718) (40). This depth of penetration
is consistent with the depth of infinite thickness defined in earlier paragraphs.
Olinger and Griffiths (33,34) demonstrated a procedure for determining
information depth of DR NIR spectroscopy using mixtures of carbazole, graphite,
and sodium chloride. Using the standard absorbance values, known particle size
and absorptivity values for the mixtures, they were able to calculate the number
of particles that the NIR radiation reaching the detector had encountered while
passing through the matrix. Although they estimated an information depth of
1 mm, the researchers suggested that the true depth might be much less due
to the effect of scattering. Furthermore, they observed a strong relationship
between absorbance, particle size, and information depth.
In perhaps the first work investigating the information depth of penetration
for pharmaceutical applications of NIR reflectance spectroscopy, Cho et al. (63),
338
Cogdill and Drennen
conducted a series of experiments to estimate the effective mass sampled by NIR

fiber-optic probes for blend uniformity analysis (63). Their work was based on
theoretical models relating sampling variance to effective sample mass. By
making certain assumptions regarding dl and true blend homogeneity, they
were able to estimate two parallel regression models relating observed spectral
variance and actual sample mass for a series of samples (of known mass)
analyzed using a laboratory NIR reflectance spectrophotometer:
log (ss2 ) b0 b1 log (Ms ) e
1
ss2 b0 b1
e
Ms
(16)
(17)
where s 2s and Ms are observed spectral variance and true sample mass for a set of
seven known mixtures (of the same nominal composition), respectively, and e is
assumed to be normally distributed indeterminate error. Following the estimation
of the linear slope and intercept coefficients, b1 and b0, effective mass sampled
could be estimated for other NIR measurement systems by evaluating spectral
variance at individual wavelengths for repeat measurements at different locations
in the same sample. To simulate monitoring of blend uniformity, Cho et al.
collected 10 replicate spectra each for a series of blends (of the same nominal
composition), which were mixed for an increasing number of rotations. The
spectra were collected by carefully inserting a thin fiber-optic probe into multiple
locations within each of the blends at each time point.
Both models [Eqs. (16) and (17)] were quite consistent with one another in
the prediction of effective mass sampled for the fiber-optic probe system. The
experimenters observed effective mass sampled to be wavelength-specific,
ranging from 0.154 to 0.858 g with standard deviation ranging from 0.0379 to
0.1581 g. They further noticed that the effective mass sampled was significantly
lower for wavelengths where the major constituents of the mixture contribute
most to the spectral variance, though the experimenters were unable to determine
the root cause. Given this disparity and the fact that the wavelength dependency
of effective mass sampled was not monotonic, it seems likely that effective mass
sampled is dependent on the apparent absorption of the material rather than wavelength, which is in agreement with the relationship posited by Bull [Eq. (15)].
Thus, for reflectance, as absorbance increases (or scattering decreases) the effective mass sampled is reduced. This hypothesis suggests that, for NIR blend
uniformity analyses relying on spectral RSD, individual wavelengths might be
selected or weighted by NIR absorptivity to evaluate homogeneity at the most
representative scale.
A second group of researchers published results (64,65) from a series of
experiments aimed at estimating effective sample size for general applications
of NIR spectroscopy for the analysis of pharmaceutical solids. The studies
explored the wavelength-dependent effective sample mass curve for multiple
instruments and materials of diverse particle size and composition. Two
339
models were employed to estimate effective mass sampled: (i) variable layer
thickness (VLT) model, and (ii) equation of radiative transfer (ERT) model;
each model necessitated a unique experimental strategy.
The VLT experiments (65) were conducted using a modified reflectance
sample cup consisting of a strongly absorbing polyamide back plate upon
which a series of spacer rings were stacked. A uniform column of powder
was contained within the cavity at the center of the rings; the thickness of the
powder layer was adjusted by adding or removing spacer rings to vary the
height of the powder slug. Thus, as the powder layer thickness increased,
a larger fraction of incident radiation was diffusely reflected by the powder
while a smaller fraction of the photons reached the back plate to be absorbed.
By fitting a curve to the depth versus absorbance data at every wavelength, the
wavelength-specific information depth could be estimated via interpolation
(Fig. 10). The results were presented in relative units (mg/cm2); because the
effective measurement area was never provided, correlation of the authors
results with those of Cho et al., is not straightforward. The experiment was
repeated for four instruments (two dispersive, two FT-NIR) using each of
Figure 10 The effect of powder mass per unit area in sample cell on absorbance
(log 1/R) for DR measurements of MCC at: (A) 900, (B) 1300, (C) 1700, and (D)
2100 nm. Abbreviations: DR, diffuse reflectance; MCC, microcrystalline cellulose.
Source: Courtesy of Elsevier, 1998.
340
Cogdill and Drennen
seven samples of varying composition and particle size (microcrystalline

cellulose (MCC), film-coated pellets, polystyrene microspherestwo sizes,
nonpareil pelletsthree sizes). It was observed for MCC and the film-coated
pellets that at wavelengths longer than 1350 and 1600 nm, respectively, the effective sample information depth was too shallow for the quantification using the
VLT method due to the higher absorbtivity. Hence, the expected relationship
where increasing absorptivity reduces effective sample mass was observed.
The researchers noted that, across all the samples tested, increasing the particle
size resulted in greater effective mass sampled. This concurs with the hypothesis
related to the particle size effect on NIR spectra stated earlier. Interestingly,
bias shifts in the effective mass sampled profiles were observed among the
instruments tested. The authors suggest that the differences in lighting intensity
and geometry may be the root cause.
The ERT experiments (64) were based on a three-flux approximation of the
equation of radiative transfer described by Chandrasekhar (36), which is one
of the augmented theories on DR cited in earlier paragraphs discussing
Kubelka-Munk theory. The solution of the ERT model yields a set of functions
relating the effective depth of penetration and the effective mass sampled to
the observed reflectance and mass-specific scattering and absorption coefficients.
As it is quite involved, the derivation and solution of the ERT model is left to the
references cited by Berntsson et al. In order to determine the mass-specific
absorption and scattering coefficient profiles for the series of test samples, the
authors utilized directional-hemispherical NIR transmittance and reflectance
measurements. By applying the experimental data to the ERT model solution,
an effective mass sampled could be predicted for reflectance measurements
from samples of the same materials. Thus, relative to the VLT method and the
methods employed by Cho et al., the ERT method relies to a greater extent on
the mechanistic understanding of the phenomena than empirical observation.
Despite this distinction, the VLT and ERT methods were in very close agreement
(Fig. 11) in predicting effective mass sampled per area.
The only areas of noticeable disagreement between the two methods were
in spectral regions of high absorbance, where the VLT method was not sufficiently sensitive. One noted advantage of the ERT method is its ability to
estimate the fraction of the total reflected radiation contributed by each progressively deeper layer of powder. As would be expected, the nearest layers to the
surface of the powder contributed the majority of reflected radiation. The
authors emphasized that there is a distinction between the fraction of reflected
light and the fraction of total information contributed by a particular layer.
This point suggests that the actual effective mass sampled is probably somewhat
smaller than estimated by any of the methods discussed.
Clarke et al. (67), utilized a method similar to the VLT procedure of
Berntsson et al., to estimate the effective mass sampled during NIR microscopic
imaging. To accurately determine the spatial distribution of chemical and
physical effects in a dosage form or powder mixture, it is imperative that the
341
Figure 11 Effective sample size for film-coated pellets MCC calculated using the ERT
method (solid line) and the VLT method (dashed line). The left ordinate gives the sample
mass per unit area and the right ordinate the corresponding effective sample thickness
calculated using tapped density of 0.30 g/cm3. Abbreviations: MCC, microcrystalline
cellulose; ERT, equation of radiative transfer; VLT, variable layer thickness. Source:
Courtesy of American Chemical Society, 1999.
depth of penetration be controlled to a limited range of surface features, otherwise stronger absorbers located deeper within the matrix may overwhelm the
signal of an analyte of interest on the surface (or vice versa).
Because it was anticipated that the sampling depth and volume for
microscopy measurements would be much smaller (due to the very small detector
area for individual pixel measurements), the investigators devised a novel solution using layers of paper to achieve precise control over layer thickness.
Because paper is composed primarily of cellulose, they were able to show that
their results were applicable to systems using granular cellulose excipients,
such as MCC. Clarke et al., concluded that, depending on the wavelength
region for analysis, the effective mass sampled for individual pixel measurements
varied between .03 and 418 mg; they went on to suggest that, along with the judicious choice of wavelength, NIR microscopy data could be used reliably in conjunction with Raman imaging, which is known to have a very limited effective
mass sampled.
Whereas the efforts discussed in the preceding paragraphs have focused on
the effective mass sampled by NIR for the analysis of free-flowing powders
(during blend uniformity analysis), effective mass sampled for single-tablet
analysis poses a somewhat different problem. Unlike blend uniformity analysis,
the ideal nondestructive tablet analysis would sample as much of the tablet
342
Cogdill and Drennen
(or capsule) mass as possible. Furthermore, whereas the earlier works have
focused on effective mass sampled during NIR reflectance analysis, many
pharmaceutical tablets can be analyzed in both DR and transmittance modes.
Thus, early NIR tablet analysis efforts were confronted with the following questions: (i) is enough of the dosage form sampled for the measurement to accurately
capture the sample characteristics, and (ii) what is the relative difference in effective mass sampled between single-tablet transmittance and reflectance NIR analyses? Indeed, there was some early concern regarding the effect of
compositional heterogeneity within individual tablets on the accuracy of NIR
reflectance measurements (64,65). The initial conclusion was that the transmission measurement would be less sensitive to such sample inhomogeneities
and would therefore be more suitable for tablet analysis. Efforts to rigorously
answer these questions were undertaken by Iyer et al. (68), by performing two
series of physical pharmaceutical experiments at Duquesne University that
were published in 2002. The first series of experiments sought to determine the
effective mass sampled for both NIR transmittance and reflectance measurements.
In order to estimate effective mass sampled, a series of specially designed
model tablets were constructed to determine the effective information depth of
NIR reflectance and transmittance measurements and the effective optical path
diameter of transmittance measurements. Effective information depth was
determined by analyzing a series of tablets of increasing thickness; the experiment was repeated using four strengths of cimetidine (other constituents were
lactose, MCC PH101, and magnesium stearate). A polynomial function was fit
for each wavelength studied to describe the change in spectral signature as a
function of tablet thickness (Fig. 12); effective information depth of reflectance
was determined by locating the first zero of the derivative of the polynomial
function fit to the absorbance data.
Figure 12 Effective sampling depth for reflectance measurements: (A) Second derivative spectra at different thickness for the 4% Cimetidine tablets at the Cimetidine and
lactose bands (arrows indicate increasing tablet thickness). Tablet thickness ranged
from 0.36 to 6.44 mm. (B) Second derivative values at 1186 nm plotted against tablet
thickness.
343
The limiting tablet thickness for transmittance measurements was determined by monitoring the transmittance spectrum as tablet thickness is increased
until the transmittance signal showed signs of distortion due to stray light around
the tablets. The limiting tablet thickness for transmittance measurement was
observed to range between 3.5 and 4.8 mm. The effective information depth of
reflectance measurement varied according to wavelength, ranging from 1.9 to
2.7 mm. Effective information depth of reflectance at a wavelength of high
cimetidine absorbance was observed to decrease as the concentration of cimetidine increased; information depth at wavelengths of high lactose absorbance was
shown to increase with increasing lactose concentration. These results conflict
with all previous studies, where the depth of penetration (or effective mass
sampled) was directly related to the level of absorbance. The researchers
suggest that these findings are the result of the effect of particle size on the
NIR spectra because lactose has a much larger equivalent diameter than cimetidine. Hence, it was suggested that as the level of lactose increased (and cimetidine
decreased), the level of scatter decreased, reducing the level of observed
absorbance. The tablet composition had no observed effect on the limiting
tablet thickness for transmittance measurement. The results for the depth of penetration and limiting tablet thickness were confirmed by analyzing a set of doublelayer tablets with varying layer thickness. Although the observed results were in
general agreement with those for the single-layer studies, a more linear spectral
effect of tablet thickness was observed for the transmittance measurements.
In order to determine the effective optical path diameter, a series of cavities
were drilled 2 mm into the tablets at three distances from the center of the tablet:
0, 2, and 4 mm. After the cavities were drilled, the transmittance spectra of the
tablet were acquired and the cavities were subsequently filled with a mixture
of ibuprofen and orange dye. The transmittance spectra from the tablets after
the cavities were filled were compared with spectra prior to filling the cavities
to identify any detected ibuprofen signal in the NIR spectra. Ibuprofen signal
was observed in the spectra for the tablets with cavities at 0 and 2 mm but not
for the tablets with a cavity at 4 mm. Thus, the estimated effective diameter of
transmittance measurement was assumed to be 7 mm (the distance from the
tablet center to the edge of the cavity placed at 4 mm was 3.5 mm).
Based on these results, and the measurements of effective information
depth of reflectance, it is concluded that: (i) neither transmittance nor reflectance
measurement sampled the entire tablet mass in any case, and (ii) the estimated
mass sampled for transmittance and reflectance measurements were comparable
based on the assumptions employed. Depending on the wavelength of analysis,
the effective mass sampled for both measurements was observed to be near
200 mg.
The second series of experiments was aimed at determining the relative
abilities of NIR transmittance and reflectance measurements to detect tablet
heterogeneities in controlled locations. The experiments were performed by comparing reflectance and transmittance spectra of the double-layer tablets and the
344
Cogdill and Drennen
tablets with drilled cavities from the earlier studies. The double-layer tablets
simulated heterogeneity along the y-axis, while the tablets with drilled cavities
simulated heterogeneity along the x-axis. Qualitative examination of the data
revealed that both transmittance and reflectance measurements were affected
similarly by varying tablet heterogeneity in both directions. Thus, transmittance
measurement geometry does not necessarily improve the robustness of singletablet NIR measurement. Indeed, later studies published by Cogdill et al. (30),
demonstrate that, for a high-speed process analytical scenario, transmittance
measurement may have an inherent disadvantage due to the effect of tablet
positioning variation.
The fact that neither transmittance nor reflectance measurements sample
the complete volume of individual solid dosage forms will only generally
be a minor drawback, especially for process control applications. Rather,
because NIR spectroscopy offers so much more speed and flexibility in process
analysis, the ability to gather relevant information on orders of magnitude
more samples offsets such limitations. Finally, as mentioned earlier, the effective
mass sampled is hardly a constant for any given material. The effective mass
sampled will change along with packing and hardness parameters, sampling
geometry, sample moisture content, and ambient humidity. It is important that
for process analytical applications of NIR spectroscopy, where real-time
control or release decisions might be based on NIR analyses, such spectral
effects should be built into the instrument calibration as much as possible.
The preceding sections were meant to provide some of the historical background that differentiates NIR from other optical techniques and to better inform
the reader of the true power and complexity of the NIR spectrum. Although the
complexity of the NIR spectra from pharmaceutical solids may seem to be
drowning in overlapping, competing signals from complex absorption and scattering phenomena, NIR spectroscopy has repeatedly proven to be an indispensable tool when paired with adequate instrumentation, chemometrics, and rigorous
method understanding. The subsequent portion of this chapter is devoted to the
discussion of NIR instrumentation, mathematics, and details of responsible
method development. The third and final section of this chapter will cover
broad applications of NIR spectroscopy in the analysis of pharmaceutical solids.
PRACTICE OF NIR SPECTROSCOPY
In the past the NIR spectrum was usually either an option or an afterthought in the
marketing of optical instruments. For example, the Cary-14 recording spectrophotometer, with which much of the groundbreaking work of NIR spectroscopy
in solids was performed (8), was actually a UV VIS spectrophotometer with
optional capabilities for measurements in the NIR region. The convergence of
inexpensive computing power, improved digital detection systems, and new
data analysis methods, along with the years of fundamental research, have led
to dramatic growth in the demand for NIR instrumentation. The demand for
345
NIR technologies is growing quickly mainly due to process analytical applications in the pharmaceutical, chemical, and semiconductor industries. This
demand is fueling research into more advanced technologies, such as laserbased instrumentation and microspectrometers. This section of the chapter will
illustrate aspects of analytical instrumentation, data analysis, and method development relevant to NIR spectroscopic analysis of pharmaceutical solids.
Instrumentation
Because of its favorable position in the electromagnetic spectrum the development of NIR instrumentation has been relatively inexpensive compared to
other optical techniques. Unlike the fingerprint region of the IR, simple and inexpensive quartz glass optics, fiber-optics, and lamps have excellent efficiency in
the NIR region. Furthermore, the NIR region has been used for many military,
aerospace, and telecommunications tasks, such as night vision, remote sensing,
and dense-wave division multiplexing (DWDM). Thus, much of the fundamental
research in optics and broadband digital detector materials [essential for many
process analytical technology (PAT) applications of NIR spectroscopy] are
paid for by the needs of other industries.
Modern NIR instrumentation for the analysis of pharmaceutical solids is
best classified according to the demands of the application. According to the
PAT guidance, at-line, on-line, and in-line measurements are defined as
follows (69):
. At-line: Measurement where the sample is removed, isolated from, and
analyzed in close proximity to the process stream.
. On-line: Measurement where the sample is diverted from the manufacturing process and may be returned to the process stream.
. In-line: Measurement where the sample is not removed from the
process stream and can be invasive or noninvasive.
Within this framework, the demands for instrument speed and robustness
increase, beginning with at-line instrumentation and ending with in-line
analyzers. The need for flexibility, in terms of the capacity to be used for a diversity of measurement tasks, however, decreases from at-line through in-line
measurement.
At-Line and Off-Line Instruments
At-line (and off-line) applications of NIR for the analysis of pharmaceutical
solids include raw material identification, investigation of in-process and finished
materials, and data collection for method development. The instrumentation for
at-line analysis is generally self-contained (including power supply, lamp, optics,
and detector), may have an integrated computer platform for data handling, and is
designed for either table top or mobile (e.g., pushcart) applications. Many at-line
NIR analyzers have automated sample carousels and/or fiber-coupled probes.
346
Cogdill and Drennen
As they may be used for a wide variety of tasks, at-line analyzers generally offer
the widest spectral range (often covering the entire NIR region), the highest spectral resolution, and internal means for instrument background correction and performance qualification (PQ). At-line NIR analyzers can be both interferometric,
usually FT-NIR, or dispersive, typically utilizing a grating monochromator.
The FT-NIR spectrometers are most advantageous when high-resolution
capabilities are important or if the spectrometer needs to have many options
for sample interaction. The spectral resolution of an FT-NIR analyzer is constant
across the spectral range detected and is adjusted by varying the length of the
interferogram (cm) (Fig. 13); for a Michelson interferometer, this corresponds
to increasing the range of displacement for the moving mirror (Fig. 14) (15).
Thus, increasing the spectral resolution of an FT-NIR spectrometer will require
longer scan times for equivalent performance in terms of signal-to-noise ratio
(SNR). It is useful to keep in mind that the SNR of optical measurements is proportional to the square root of the number of photons collected, which scales
linearly with scan time and/or the number of exposures. The FT spectrometers
have a combination of three theoretical advantages, which make the technology
attractive. They are: (i) all frequencies (wavelengths) are detected simultaneously
(Fellgetts advantage), (ii) wavelength accuracy is constantly maintained by an
internal laser line (Connes advantage), (iii) the interferometer has much higher
optical throughput than other (dispersive) technologies (Jacquinots advantage).
Figure 13 Fourier interferogram, or power spectrum (upper panel ) and the corresponding reflectance spectrum (lower panel).
Figure 14
347
Michelson interferometer.
The FT spectrometer technology is extensively characterized in a number of

comprehensive reference texts (70 72).
The typical resolution for FT-NIR applications is 2 8 cm21 for explorative
or qualitative analysis, and 8 32 cm21 for routine applications. In contrast to the
analyses of gases or liquids, the resolution needs for most applications of NIR for
the analysis of pharmaceutical solids could be met with relatively low resolution
(8 64 cm21 or 6 12 nm). This is due to the extensive coupling of molecular
bond vibrations in pharmaceutical solids, which results in rather wide absorbance
bands. Indeed, it is not uncommon for practitioners to overestimate the need
for spectral resolution when it would have been more beneficial to increase the
SNR. The wavelength accuracy of FT-NIR spectrometers is guaranteed by
using a constant laser line reference. The spectral range of FT spectrometers is
controlled by adjusting the resolution of the interferogram (e.g., adjusting the
step length of the moving mirror). The spectral range of modern FT-NIR spectrometers is mainly limited by the sensitivity of the detector material, rather
than the mechanical stability or transmittance efficiency of the optical platform.
There are a variety of detector materials (72,73) suitable for the NIR region.
The earliest NIR detectors utilized materials, such as silicon (,1100 nm) or lead
sulfide (PbS, 850 3300 nm), among others. Modern NIR spectrometers often
use indium gallium arsenide (InGaAs, 850 1750 nm) or extended-InGaAs
(1100 2200 nm). The InGaAs detectors are capable of high performance at
348
Cogdill and Drennen
room temperatures and work well in digital systems; however, early extendedInGaAs detectors had the reputation for unpredictable lifespan and variable
detector linearity. Recently, more stable versions of mercury cadmium telluride
(MCT or HgCdTe) and lead sulfide detectors have become available in linear and
2-D diode array formats with excellent sensitivity over the entire NIR range
(850 2500 nm); these devices are likely to be useful for future instrumentation
designs. The choice of detector material will be determined by the price,
desired wavelength range, and performance requirements.
Numerous methods (15) (in addition to the Michelson interferometer) exist
to exploit the inherent Fourier relationships of optical mechanical systems. The
Mach-Zender, Sagnac, or polarization interferometers are some examples.
The polarizer interferometer relies on a two-piece birefringent crystal wedge
polarizer and birefringent compensator to modulate optical energy in the
pattern of a Fourier interferogram (15,74). As shown in Figure 15, randomly
polarized radiation passes through a linear polarizer and a birefringent crystal,
which separates the light into two orthogonally polarized parallel rays with a
(wavelength specific) phase shift, thereby slightly rotating the plane of polarization. The phase-shifted rays are transmitted through the two-piece wedge polarizer, which then rotates the plane of polarization in the opposite direction,
reducing the phase shift. The angle of polarization rotation is dependent on the
effective pathlength through the wedge polarizer, and wavelength. At the
initial position, the thickness of the wedge polarizer is set so that the polarization
rotation angle (at all wavelengths) perfectly offsets the compensator, resetting the
polarization state to that of the first polarizer. In this state, the maximum light
energy at all wavelengths will pass through the second polarizer (DC intensity).
As the thickness of the wedge polarizer is varied, the polarization angle incident
on the second polarizer will vary in a sinusoidal pattern between 0 and 100%
transmittance; the frequency of the sinusoidal pattern will vary according to
optical frequency. Therefore, the sum of the components incident on the detector
will produce an optical interferogram. The change in the thickness of the wedge
polarizer (and, therefore, the change in phase shift) is analogous to the distance
traveled by the moving mirror of a Michelson interferometer.
Figure 15
Simplified view of a polarization interferometer.
349
At first glance, the advantages of the polarization interferometer for process

spectrometry are not obvious. The optical efficiency of the device is limited by
absorption in the beam path, especially as the thickness of the wedge polarizer
reaches a maximum (which will limit maximum resolution). Spectrometers
utilizing the polarization interferometer have been available for many years.
Although early instrument models suffered from some wavelength stability
problems (74), recent upgrades of the technology have yielded a very robust
instrument with more than adequate resolution capability (8 cm21) (75). In
fact, the advantage of the polarization interferometer is its ruggedness; there is
only a single optical path through the instrument (compared with the dual path
of Michelson, Sagnac, or Mach-Zender interferometers). Thus, the polarization
interferometer is less affected by small misalignments in the beam path, the
sort of which might arise during at-line or in situ operation. Furthermore, the
simple design is more cost-effective and can be produced with a smaller footprint.
The grating monochromator is the dominant dispersive technology for
at-line NIR analyzers. Although there are many grating monochromator
designs, they all rely on mechanical modulation (rotation of a scanning mirror,
or rotation of the grating) of the instrument geometry to scan regions of the electromagnetic spectrum (Fig. 16). Although grating monochromators are not
subject to the theoretical advantages of FT-NIR (e.g., Jacquinot, Fellgett,
Connes), the high practical efficiency and robustness of their optical systems
(which often utilize nearly 100% reflective optics), as well as their relatively
low cost, have made them a popular choice for many NIR analytical applications.
These analyzers have a long history of NIR applications in food and agriculture.
Despite their mechanical limitations, they are capable of stable performance in
relatively harsh environments (4,5,72,73).
The spectral resolution of a grating monochromator is determined by the
slit function and the quality of the grating (15). Similar to FT-NIR, increasing
the spectral resolution of a grating monochromator instrument will involve an
SNR trade-off because the optical flux is reduced as the slit is narrowed. The
spectral range of grating monochromator instruments is limited by the geometry
of the grating and optical bench as well as the detector material. Some instruments utilize multiple gratings and detectors to increase spectral range and/or
resolution. The wavelength accuracy of rotating grating instruments is limited
by the quality of the grating and the mechanical stability of the optical system.
Grating monochromator instruments have a proven track record of stability
and transferability. In fact, the NIR grain analyzer networks exist where hundreds
or thousands of instruments utilizing the same calibration equations have been
established in many countries and trading regions around the world.
The stability and transferability of rotating grating instruments are due to
the following characteristics:
. Optical standardization utilizing a simple gain and bias adjustment
function can sometimes be employed, because each wavelength is
350
Cogdill and Drennen
Figure 16
Czerny-Turner rotating grating monochromator.
sampled by the same detector(s). If the standardization function is

wavelength dependent, a simple (e.g., quadratic) function is expected,
depending on the geometry of the grating.
Wavelength errors can typically be characterized by a linear shift or
stretch of the axis (76), which is related to variability in the position
of the detector relative to the monochromator.
Spectral resolution is generally very stable and will vary quadratically
across the spectral range according to the monochromator slit function.
Some instrument manufacturers utilize spectral convolution to increase
the effective bandpass of the monochromator to ensure a more constant
bandpass across the spectrum and between instruments. Because
resolution is rarely a limiting factor in NIR analysis, it is an efficient
method of increasing transferability.
One instrument manufacturer (77), in particular, undertook a study of the

instrumental factors affecting calibration transfer. Besides the factors mentioned,
their work showed that the precise control of the geometry between the sample,
monochromator, and detector is critical to calibration transferability among
at-line or off-line instruments.
351
On-Line Instruments
The choice of instrumentation for on-line applications will be determined by the
performance requirements in terms of sensitivity, robustness, and speed. Because
process analyzers are often located in hazardous or controlled environments,
their performance must be stable and robust to environmental variation.
The expense of frequent instrument standardization and maintenance may be
prohibitive due to access or scheduling restrictions. Furthermore, frequent
restandardization or calibration updates can become a regulatory burden.
Hence, after meeting the demands for sensitivity, the critical factor affecting
technology selection (and method development) for PAT implementations of
NIR technology should be robustness (78).
If the speed of analysis is not critical, an at-line type of spectrometer can be
adapted for on-line use by employing an automated sample handling mechanism.
However, speed will often be a critical factor for on-line analyses of in-process
materials or finished products. Thus, spectral range, resolution, sensitivity, or
SNR may be sacrificed to increase the measurement speed. For example, highspeed FT-NIR instruments have been developed for on-line applications by
significantly reducing the length of measured interferograms and utilizing step
scanning. Because on-line applications of NIR spectroscopy for solids analysis
will generally not require high spectral resolution, relatively few interferogram
points must be measured to generate a suitable broadband spectrum (79).
Although grating monochromators can be deployed effectively for some on-line
applications, two newer dispersive technologies have some advantages in such
cases: acousto-optic tunable filter (AOTF) and holographic grating diode array.
The AOTF modulator relies on an interesting quality of optical diffraction
in an anisotropic medium. AOTFs utilized for NIR spectroscopy consist of a birefringent tellurium dioxide (TeO2) crystal bonded to an acoustic transducer
(72,73,80 83). When the transducer (Fig. 17) is excited by an applied radio
frequency (RF) signal, the acoustic pressure waves generated propagate
through the crystal lattice, resulting in a periodic modulation of the refractive
index of the crystal. Thus, a narrow band of optical frequencies that satisfy the
phase matching conditions of the RF signal will be cumulatively diffracted,
while other frequencies will be transmitted without diffraction (the zero-order
beam). Because the crystal is birefringent, two diffracted beams are produced
with orthogonal polarization states. One of the beams is directed to the sample
for detection, and the other is either directed to a beam stop, or is detected directly
as a reference channel. In the latter case, the instrument is operating as a double
beam (in space) spectrometer. The center wavelength of the diffracted beam(s) is
varied by modulating the frequency of the RF driver. Because the frequency of
RF driver can be modulated very quickly, AOTF devices are capable of extremely rapid scanning. Furthermore, since the TeO2 crystal is mostly transparent
in the NIR region, AOTF devices have high optical efficiency. The spectral
range of AOTF devices is determined by the transmittance curve of the crystal
352
Figure 17
Cogdill and Drennen
Acousto-optical tunable filter (AOTF) operation.
and the sensitivity of the detector material. The bandpass of commercially available devices in the NIR is similar to that of a grating monochromator (6 10-nm
FWHM). The wavelength accuracy is reported to be less than 0.5 nm (84),
though the variation depends on a number of factors related to SNR, such as
tuning speed.
The AOTF spectroscopy (80) was popularized (beginning in the 1980s)
because it presented a real breakthrough in simplicity, speed, and efficiency.
The devices were considered to be inherently stable because there were no apparent moving parts and were, therefore, more practical for rugged process applications or high-speed reaction monitoring. For spectroscopists unfamiliar with
diode array technology, AOTF devices were desirable because, they utilized
simple, single-element detectors, like FT-NIR and grating monochromators
that were assumed to be more amenable to calibration transfer. Furthermore,
because diode array devices were viewed to be limited by the tradeoff between
resolution and spectral range, which is set by instrument geometry and the
number and size of detector elements, the broadband high-resolution capacity
of AOTF devices was an apparent advantage. In long-term applications,
however, Cogdill et al., have reported the occurrence of complex, unpredictable
patterns of baseline variation between measurements over time and those on
similar instruments (85). These baseline shifts are thought to be the result of
changes in the character of the bond between the acoustic transducer and the
353
TeO2 crystal. The time constant of these changes is thus far unknown and have
been observed over both short (weeks) and long (months) periods (unpublished
findings). Moreover, the complexity of the baseline variations are in stark contrast to the simple (linear or quadratic) baseline variations expected for competitive technologies. Although these phenomena do not necessarily rule out
successful implementation of AOTF technology for pharmaceutical applications,
their existence does necessitate a comprehensive strategy for instrument performance monitoring, calibration transfer, and calibration update (30,57,85).
The development of multielement detector arrays, such as the linear
charge-coupled device (CCD) and photodiode array (PDA, or diode array),
more than 30 years ago spurred the development of multichannel spectrophotometers based on imaging polychromators, known as linear diode array spectrometers (72,86,87). A polychromator (Fig. 18) is an augmentation over a
monochromator whereby the spectrum is acquired by capturing an image of
the holographic grating on a diode array (either CCD or PDA) in a parallel
fashion, rather than serial scanning the surface of the grating through mechanical
translation. Thus, though diode array spectrometers still utilize dispersive technology, they are subject to Fellgetts (multiplex) advantage and therefore offer
high SNR. Depending on the source intensity and the means of optical collection,
a typical diode array spectrometer can acquire a broadband NIR spectrum of
good SNR in less than 100 ms. Because diode array spectrometers have no physically modulated components and are constructed using a solid block of material,
diode array spectrometers are extremely robust to environmental conditions and
aging. The diode array spectrometers have only gained significant application
in NIR applications with the relatively recent availability of InGaAs photodiode
arrays, the development of which is largely attributable to demand by the telecommunications industry.
Figure 18
Fiber-coupled process diode array spectrometer.
354
Cogdill and Drennen
As mentioned in earlier paragraphs, diode array spectrometers were

initially received by some spectroscopists with a bit of skepticism. This is
largely attributable to the poor stability of early linear photodiodes, which
tended to make calibration transfer and maintenance rather complex. Although
simple standardization functions can often be estimated for most other spectrometer technologies (e.g., slope and bias correction), diode array devices
require that every detector is treated as an individual spectrometer. This is no
longer viewed as a significant limitation of the technology, especially given
the extreme robustness of the diode array optical platform.
The housing of a diode array spectrometer is typically constructed using a
single piece of aluminum or even glass/ceramic to reduce temperature-induced
flexure. Other concerns surrounding diode array spectrometers were related to
their limitations in terms of absolute resolution and wavelength range. The spectral resolution and range of a diode array device will typically vary quadratically
across the wavelength range, and is determined by the number of detector
elements, their size, instrument geometry, and the slit width. Typical diode
array devices have nominal spectral resolution of 6 nm, which is more than
adequate for process analysis of most pharmaceutical solids. The NIR diode
array spectrometers might have 128-, 256-, or 512-element arrays and will
cover the wavelength range from 850 1750 nm for standard InGaAs or 1100
2200 nm for extended-range InGaAs. In some cases the curvature of the polychromator grating is optimized to project a uniform wavelength axis on the
linear diode array. During the manufacture of a diode array spectrometer, an
internal lookup table is created to determine the actual wavelength sensed by
each photodiode. Even if the grating has been designed to have a linear wavelength axis, most diode array spectrometers utilize the lookup table along with
some internal preprocessing (interpolation) to correct for slight differences in
the wavelength axis between instruments.
The diode array spectrometers most often interact with samples via fiberoptic probes and might not have an integrated light source. Although this
might hinder the use of such instruments for at-line applications, it adds significant flexibility for on- and in-line use. Early versions of commercially available
diode array spectrometers were produced without internal means of photometric
correction (dark current subtraction, normalization). With such instruments, it is
necessary to manually collect background correction spectra using standard
reference materials, though the mathematical background correction procedure
is generally automated. One diode array instrument manufacturer (88) has
begun to offer an integrated process interface system with redundant internal
light sources, an automated background correction mechanism, and one or
more performance monitoring standard reference materials (Fig. 19).
In-Line Instruments
The instrumentation for in-line NIR applications is much the same as for on-line.
The main difference is the way the sensor interacts with the process. Whereas
355
Figure 19 Design of a self-referencing process interface for a fiber-coupled diode array

spectrometer. Source: Courtesy of Control Developments, Inc. (patents pending).
on-line measurements occur in a controlled side stream, where the sampling

geometry can be easily controlled and the flow can be stopped, the in-line
measurements are typically performed by noninvasive or noncontact sensing
directly through an optical window while the process is running. For example,
an NIR spectrometer for fluidized bed powder processing might be positioned
to acquire spectra of the fluidized material directly through a window in the
processor at the same height as the fluidized layer. Because the material being
analyzed by in-line sensors is typically in motion, very high-speed spectrometers,
especially diode array devices, are preferred. This is in spite of the fact that
slower-scanning instruments will often work surprisingly well for dynamic
processes, because total scan time is still likely to be much less than the
overall process time. Robust design of the sampling system is a critical aspect
of in-line sensor design. Flowing materials might be abrasive, wet, or sticky;
thus, the spectrometer window should be made of a hard material (e.g., sapphire)
and might have an automated cleaning or antifouling mechanism. Some
scientists have indicated that slight heating of the optical probe above the
process temperature will reduce the incidence of probe fouling (unpublished
conversations).
In-line instrumentation may often be smaller and less integrated than
at-line analyzers and is likely to resemble a sensor much more than an analyzer.
In-line instruments may be subjected to temperature extremes and vibration,
356
Cogdill and Drennen
or they may be deployed in remote, inaccessible locations. A fairly recent innovation in in-line instrumentation is RF wireless communication. By using wireless data communication and battery-powered source illumination a process
spectrometer can be attached to moving equipment, such as a V-blender. The
physically rugged, lightweight devices, such as diode array instruments, are
usually the most ideal for in-line use.
Because in-line applications often involve long-term measurement of a
single type of material, filter-type instruments (discrete photometers) are
another option. Filter instruments often employ a set of 10 20 interference
filters mounted on a high-speed rotating filter wheel positioned between collection optics and a single detector. High-speed measurement is possible (approximately as fast as a diode array spectrometer) by spinning the filter wheel at a very
high rate (10,000 rpm). Fast Fourier transform (FFT) of the detector signal can
be used to isolate the spectral intensity data from instrumental noise effects,
which, along with the simple, efficient design, make filter instruments very
robust devices for single-material measurement. Filter instruments have been significantly deployed as in-line moisture gauges, though some have been successfully calibrated for multiconstituent content prediction, especially in agricultural
applications (72,73). Because they provide such limited spectral information,
however, their applicability has begun to erode as low-cost diode array
systems have become widely available.
As the market demand for process spectrometers in the pharmaceutical
and other industries has continued to expand, new technologies are constantly
evolving. A current trend in sensor development is miniaturization. A new
type of NIR microspectrometer (89) has been developed, which has the
potential to greatly increase the pervasiveness of NIR process monitoring
by producing extremely small, relatively inexpensive devices, based on microscopic tunable Fabry-Perot interferometers that are built into a single microchip. Fabry-Perot interferometers (15,90) consist of a stack of partially
reflective parallel dielectric plates, or etalons, for wavelength selection. For
any given wavelength, the distance between an individual pair of plates, and
the refractive index of the space between the plates, will determine whether
or not the multiple reflections between the plates will lead to constructive
or destructive interference. Thus, by varying the thickness of the etalons, or
by modulating the refractive index of the void space [e.g., via liquid crystals
(90)], a tunable filter can be constructed. Because each etalon passes harmonics of each optical frequency, a stack of etalons is used to select individual
wavelength bands. The Fabry-Perot etalons are capable of very high efficiency
and resolution; however, the free spectral range of an etalon is inversely proportional to the resolution. Thus, a relatively narrow spectral range is required
to produce a high resolution device. Because the microspectrometer is so small
and efficient, however, a bank of multiple devices can be assembled to cover a
wide spectral range (if such flexibility is necessary) while maintaining a very
small instrument footprint.
357
Imaging Spectroscopy
Chemical imaging spectroscopy as a micro- or macroscopic analytical tool has
recently become available. Commercially available visible, NIR, Raman, IR
(FT-IR), and most recently, terahertz (91) imaging spectrometers, capable of analyzing a diversity of pharmaceutical solids, have been developed. The NIR
imaging spectroscopy, however, is unique in its combination of speed, analytical
capabilities, resolution, and efficiency. A variety of optical systems have been
created for high-resolution NIR imaging spectroscopy, including point-mapping
systems, push-broom type sensors utilizing holographic gratings or linearly variable filters (LVF, or wedge) or more often tunable filters, such as the AOTF and
the liquid crystal tunable Lyot filter (LCTF) (72,81 83,92 94). The current
dominant technology for pharmaceutical imaging spectroscopy is based on the
combination of an LCTF and an InGaAs or InSb (83) digital focal plane array
(Fig. 20). The LCTF-based systems operate in a manner whereby single-plane
images are collected of a sample (typically by reflectance) over a wide range of
individual wavelengths. The collection of single plane images forms a threedimensional matrix of data called a hypercube. Thus, a high-resolution spectrum
is collected for every pixel in a single-plane image of the sample (Fig. 21). Clarke
(92) and Tran (81) have recently published reviews on the methods, instrumentation, and applications of imaging spectroscopy.
Figure 20
Sapphire imaging spectrometer from spectral dimensions.
358
Figure 21
Cogdill and Drennen
Example of a spectral image data hypercube.
At the heart of the LCTF-based imaging system is the camera focal plane
array (FPA) and the tunable filter. The systems based on InGaAs cameras cover
the wavelength range from 850 to 1750 nm, whereas the InSb system is sensitive
over the range of 1200 to 2450 nm. The NIR FPAs are typically either 0.08
(320 256 pixels) or 0.20 (640 320 pixel) megapixel arrays. The highestvolume applications of NIR FPAs are either military/emergency (night vision,
ballistic detection, heat sensing), or industrial (preventative maintenance,
machine vision). Further maturation of the industry is likely to yield larger
imaging arrays; however, many pharmaceutical process analysis applications
of imaging spectroscopy can be solved with the 0.08 megapixel camera.
Imaging spectroscopy FPAs are available in 12-, 14-, and 16-bit versions.
Although one might typically expect a 16-bit system to have much higher performance, the true performance limit of an FPA may be limited by the electron
well depth, rather than the quantization limit. Because the pixels of an NIR
FPA are typically much smaller (30 mm) than the photodiode of an FT-NIR
or even diode array, their electron well capacity and, hence, their individual
SNR will be much lower. The SNR performance of an imaging spectrometer
can be greatly improved, though, by leveraging the three-dimensional correlation
structure of the data by using 3D convolution, 2D image processing, or spectral
preprocessing, such as smoothing, normalization, and derivatives. Furthermore,
compared to single-point spectra of pharmaceutical materials, which represent
the average spectral intensity over a large area of the sample, single-pixel
spectra are often dominated by the absorbance of a single constituent (Fig. 22).
Thus, by interrogating a much smaller area (volume or mass) of material, concentration variation between pixels is magnified. It is in this way that extremely
359
Figure 22 Spectral image of a directly compressed tablet consisting of theophylline,

MCC, and magnesium stearate. The upper left and right images were acquired at 1050
and 1570 nm, respectively. The lower left image is a pseudocolor PLS prediction image
(gray, theophylline; black, MCC; light gray, other/Mg stearate). The single plane
images clearly show the local sensitivity of pixel intensity to composition. The periphery
of the images demonstrates the focal problems associated with spectral imaging of samples
with irregular surface height (such as a convex tablet). Abbreviations: MCC, microcrystaline cellulose; PLS, partial least-squares.
high-quality spectral information about a sample can be gathered, even though

the individual pixel-level spectra are of comparatively low SNR. Furthermore,
very low-concentration constituents can be detected and quantified by NIR
imaging spectroscopy.
The most common LCTF design used for pharmaceutical NIR imaging is
based on a tunable Lyot filter (94), which exploits the index of refraction characteristics of birefringent liquid crystals. Birefringent liquid crystals exhibit an
index of refraction characteristic that depends on the polarization parameters
of incident light. Liquid crystals transmit either the left- or the right-handed
polarization state, depending on the helical ordering of the crystals, and reflect
the opposite polarization state. Application of an electric field parallel to the
optical axis causes the crystals to rotate in the plane perpendicular to the electric
360
Cogdill and Drennen
field. This rotation changes the index of refraction of the crystal by changing the
position of the crystal relative to the optical axis (90). Thus, tunable optical filter
devices that employ liquid crystals to vary the index of refraction can be tuned by
simply adjusting an applied voltage.
Lyot filters, also known as polarization interference filters, are constructed
in a sandwich arrangement of multiple stages. An individual stage consists of a
birefringent polarizer and an analyzer (linear polarizer). Similar to the polarization interferometer, the birefringent polarizer effects a wavelength-dependent
rotation of the plane of polarization. When the plane-polarized light impinges
on the analyzer, a fraction (from 0 to 100%) of the light in the plane of the analyzer passes, while the orthogonal component is blocked (94). Because each stage
passes all harmonics of a single frequency band, a complete Lyot filter is
constructed by stacking multiple stages which combine to pass only a single
optical frequency band. By adding a tunable liquid crystal layer to the Lyot
stages, the Lyot filter becomes an LCTF (Fig. 23).
The LCTF holds advantages over similar devices constructed using
Fabry-Perot etalons and AOTFs, because it is capable of tuning across very
wide regions of the electromagnetic spectrum, while maintaining reasonable
optical throughput with no discernible image shift or deterioration. The LCTFs
have been constructed with resolutions ranging from 0.25 60 nm, operating in
Figure 23 Exploded view schematic of a single Lyot filter stage. A complete filter would
be created by stacking multiple stages with increasing fixed compensator thickness.
361
the visible (400 720 nm), the SW-NIR (700 1100 nm), or the NIR (900
1800 nm) (93), and more recently 1200 to 2450 nm (95). As expected, increasing
the spectral resolution of an LCTF decreases the optical throughput. Commercially available NIR imaging spectrometers have nominal spectral resolution of
approximately 6 nm, which increases quadratically by wavelength. Although
LCTFs are relatively fast, capable of tuning between adjacent wavelengths in
150 ms, exceeding the maximum tuning speed can produce erratic results as
tuning relies on dynamically altering the orientation of the liquid crystals. A
major drawback of the LCTF, however, is that by placing the filter between
the camera and the sample, the depth of the field of the imaging optics can be
greatly limited. Thus, as the level of magnification is increased, the effective
aperture is greatly reduced (reducing the SNR). Furthermore, the focal distance
will vary greatly across the spectrum due to chromatic aberration (Fig. 24).
Depending on the surface characteristics of the sample, however, the focal
quality of linear combinations (e.g., principal component score images, concentration maps) of images will be somewhat better than the focal quality at
individual wavelengths.
One factor that has limited the use of imaging spectroscopy among diverse
industries is the cost of the equipment and skills needed for successful implementation of the technology. Among pharmaceutical applications, however, the NIR
imaging spectroscopy has the potential to increase the efficiency of formulation
and process development and root cause analysis, as well as in-process testing
and PAT method development. A common misconception about applying
imaging spectroscopy for routine applications, such as PAT, is that the data
flow would be too great to be feasible for on-line or in-line use. Because the magnification factor of imaging spectroscopy greatly increases sensitivity over
single-point spectroscopy (as discussed in earlier paragraphs), most beginning
Figure 24 Image of extruded particles at 10 magnification. The difference between the
left and right panels illustrates the effect of chromatic aberration on focal quality. Depthrelated blurring can also be seen in areas of major thickness variation.
362
Cogdill and Drennen
practitioners greatly overestimate the number of wavelengths that should be

sampled. For a typical multicomponent tablet analysis, reflectance imaging
(320 256 pixels) at ,64 wavelengths is likely to capture most relevant
spectral information and allow many opportunities for spectral preprocessing
and modeling, yet consume only 10 Mb/hypercube (assuming an unsigned
16-bit integer). Furthermore, the elimination of background pixels, image and
spectral data compression, and metadata extraction can reduce the data load
even further. Spectral image data do not have an intrinsically high information
density because much of the data are correlated in all three dimensions. Thus,
as with any instrumental PAT method, rigorous, knowledgeable method development is necessary to determine which data (whether spectral images, NIR spectra,
or other sensor data) are relevant to critical material properties. The NIR imaging
spectroscopy has been deployed for root-cause investigation, solid dosage form
analysis, and blend homogeneity analysis. As imaging spectroscopy continues
to be explored, and the equipment, methods, and understanding become more
refined, many more applications will be discovered. Because the field continues
to evolve quite rapidly, those interested should consult current literature for a
more in-depth discussion of imaging spectroscopy (81,92).
Information Technology
Whether deployed in a development laboratory or embedded within a complex
process, the efficient use of NIR technology in the future will require careful
consideration of the information technology (IT) infrastructure to be employed.
The NIR instruments have historically included layers of proprietary systems
developed by their manufacturer for data handling and calibration development.
For example, most early NIR spectrometers included an internal computer
and chemometrics software package and might have used a proprietary data/
model structure. Although such vertical integration is suitable for many at-line
analytical applications, on- and in-line measurement applications demand
real-time interaction with a larger data system (e.g., management execution
system, MES), and may require a common interface for many types and
brands of instrumentation. In the ideal scenario, a process NIR spectrometer
should be little more than a sensor, leaving data processing demands to remote
software applications. The modular approach increases flexibility and transparency and allows more customization of the system to adhere to any current
data management architecture involved.
As the demand for multivariate process analytics grows, systems integration hardware and software will play a larger role in managing NIR data
and those from other sensor technologies. Eventual linkage of the highest-level
planning systems (enterprise resource planning, ERP) with instrumentation and
process control systems (components of the MES layer) will increase the importance of rigorous method development. Finally, the pharmaceutical industry could
gain from a strategic investment in the development of a comprehensive set of
multivariate data/model standards for storage, secure communication, and
363
interoperability. Such an effort would fit into the scope of the ASTM (American
Society of Testing and Materials) E55 committee on pharmaceutical application
of process analytical technology.
From the onset of technology selection, and continuing through the process
of deployment and operation, careful consideration of sensor performance parameters is important (6). All too often it seems that NIR projects begin with
the purchase of equipment, followed by significant efforts to use what turns
out to be instrumentation that is not suitable for the application. Whether the
objective of the application is process analysis or materials research during
process development, the chemical and physical constraints imposed by the
materials and processes should guide the selection of an analyzer (Fig. 25).
Some questions to be considered during this phase include:
. Is the measurement required to be noncontact or noninvasive?
. What is the required speed of analysis?
. What environmental factors have the potential to affect instrument
performance?
. What is the acceptable level of analytical error?
. Is in situ maintenance feasible in a given deployment?
. Is instrument stability suitable such that the maintenance can be performed at feasible intervals (e.g., between production campaigns)?
Figure 25 The ideal path of development actions is similar across multiple levels of
method development. Costly errors in technology selection, model development, or
systems integration that might hinder the progress of deployment will be avoided by following a rigorous, science-based approach from the beginning of any PAT application.
Abbreviation: PAT, process analytical technology applications.
364
Cogdill and Drennen
.
.
What is the sensitivity and selectivity of the analytical device for the
analyte(s) of interest?
From an analytical perspective, does the instrument have a suitably
broad range of performance to be expected to serve effectively over
the expected composition range of the product(s) being monitored?
The first few items on the list will play a significant role in determining the
design of the instrument and data handling system to be chosen, whereas the latter
items will guide method development and the determination of performance
standards.
The United States Pharmacopeial (USP) convention has established a
monograph, USP ,1119., that sets forth rigorous procedures for NIR instrument performance qualification (96). The monograph describes the procedures
and materials for testing such instrument performance parameters as photometric
linearity, range, high- and low-flux noise, wavelength accuracy, spectral resolution, and wavelength repeatability. The monograph also suggests performance
qualification limits for each of the tests, frequency of performance verification,
and discusses methods for model transfer and validation. Although the guidance
serves as a good framework for qualification (and validation), the proposed performance limits are likely to be excessive for some applications, and insufficient
for others (30).
Some of the performance parameters, such as photometric linearity, spectral resolution, and wavelength accuracy, are often less critical in technology
selection because they are generally fixed parameters that can be accurately
described for a particular instrument. The most important parameters would
relate to measurement precision and stability (e.g., high- and low-flux noise,
wavelength repeatability, baseline stability) because they ultimately dictate the
maximum level of confidence in predictions and instrument suitability. For
example, depending on the optical system and detector material employed, the
NIR instruments would have vastly different performance in terms of spectral
noise, even though they are manufactured to cover the same wavelength
range (Fig. 26).
If the instrumental noise level is too great in the primary spectral regions
of sensitivity, the net analyte signal (NAS) for the component of interest will
be diminished, and an otherwise appropriate instrument might be determined
to have insufficient SNR. Alternatively, choices made during method development can greatly impact acceptable levels of various types of noise (30,57,85).
For example, by utilizing high levels of spectral smoothing, or by building wavelength variation into the analytical model, the influence of wavelength repeatability on prediction stability can often be diminished without degrading the
accuracy significantly. Whatever the application may be, in the same way that
process understanding reveals the product and process parameters that are critical
to quality, method understanding will help with the determination of appropriate
instrument performance requirements.
365
Figure 26 Relative repeat-scan noise profiles for three different instruments in normal
operation: gray, diffraction grating-based monochromator (at-line); black, AOTF
(on-line); light gray, diode array (in-line). All three sets of profiles were collected by
mean-centering of a matrix of repeat scans of a single pharmaceutical solid material.
The relative scan time was as follows (from longest to shortest): monochromator .
AOTF . diode array. Although sample presentation varied somewhat for each of the
instruments, all scans were collected without moving the sample. (The grating-based
instrument profile is a composite noise profile collected from three separate tablets,
which should not significantly affect the noise profile.) It should be noted that the performance of each of the instruments can be adjusted by altering the exposure time, light intensity, rescans, or sample positioning; furthermore, the scaling of the noise will be affected
by the magnitude of the reference scan used in calculating the absorbance spectrum.
Abbreviation: AOTF, acousto-optic tunable filter.
Data Analysis
Because of the complex chemical and physical factors affecting NIR spectra of
pharmaceutical materials, the NIR data analysis is a multidisciplinary exercise,
utilizing the aspects of chemistry, physics, digital signal processing, and multivariate statistics. Chemometrics, as defined by the International Chemometrics
Society (ICS), is the science of relating measurements made on a chemical
system or process to the state of the system via application of mathematical or
statistical methods. Chemometrics is a scientific field of study, and the credit
of much of its development can be attributed to the development of algorithms
366
Cogdill and Drennen
for NIR data analysis and calibration. Whether the goal of data analysis is to
interpret some observed patterns in the spectral data or to develop a predictive
calibration model, the same toolbox of chemometric methods is employed. Furthermore, almost all chemometric tools are based on statistical and mathematical
concepts of correlation and convolution. A chemometrician, however, seeks to
find a valid chemical or physical interpretation for the relationships estimated.
In recent years, a great deal of excitement and mysticism has been associated with chemometric data analysis. Powerful, highly automated tools, such as
the PLS and ANN, have made the development of a predictive equation from a
myriad of data types virtually trivial. However, the importance of truly understanding the chemical and physical phenomena underlying the spectral data
and the modus operandi of the chemometric algorithms cannot be understated.
No combination of chemometric procedures can ever increase the amount of
signal within a set of data. In the most abstract sense, chemometrics can only
serve to optimally isolate relevant signals from noise and background, it is the
duty of the scientist to extract information and knowledge from the signals.
The following paragraphs will outline some of the most important chemometric
tools for NIR method development and are not meant to be a rigorous treatment
of the subject. For more information on chemometrics, numerous reviews of the
subject are available (79,97 101). Three well-respected journals are dedicated to
the science and application of chemometrics (102,103) and the application of
chemometrics and NIR spectroscopy (12).
Data-Analysis Objectives
The selection of chemometric tools to be employed in solving a particular
problem is guided, mainly, by the data-analysis objective. There are two
general categories of data analyses, quantitative and qualitative, though the
two categories are, at times, indistinct.
Quantitative chemometric data analysis involves multivariate calibration
(79,98,104) for prediction and estimation. In a pattern recognition sense, quantitative data analysis is consistent with supervised learning, where a functional
relationship between a set of inputs (NIR spectra) and outputs is sought.
Hence, quantitative chemometric tools are typically related to regression
(e.g., multiple linear regression, PLS) and function estimation (e.g., ANN). As
discussed in section I of this chapter, functions can sometimes be estimated to
predict derived properties of pharmaceutical materials (e.g., dissolution rate)
and typical chemical and physical properties from the NIR spectra.
Qualitative chemometric methods are a collection of methods to explore
patterns observed (or hidden) within a set of NIR spectra. Conversely, qualitative
data analysis is comparable to unsupervised learning. Thus, while the results of
qualitative analyses may be applied in very quantitative ways, the primary distinction is that, rather than directing the algorithm toward a particular solution,
qualitative methods will reveal whatever relationships happen to be observed
in the data (depending on the mathematical treatment). Beyond exploratory
367
data analysis, examples of qualitative methods include curve resolution, discriminate analysis and classification, and conformity testing (105).
Preprocessing
The purpose of spectral preprocessing is to reduce the amount of noise and/or
background signal in a set of data so that a more robust model can be developed
and the basis of the model developed can be more easily understood. Spectral preprocessing associated with early spectrometer designs, which sampled relatively
few wavelengths, such as filter instruments, consisted of time-based noise filtering of spectral signals and the optimal selection of spectral filters. Indeed, a great
deal of effort was often spent to optimize wavelength selection, typically through
the use of stepwise multiple linear regression (MLR) (106,107).
With the advent of scanning spectrometers and full-spectrum regression
methods, such as PLS, the need for optimal wavelength selection has diminished.
However, for most pharmaceutical applications of NIR, it is a good practice to
utilize instrument performance testing and pure-component scanning to eliminate
regions of the spectrum where analyte signal is greatly reduced by instrument
noise (57) or where model performance may suffer due to concomitant signals
(e.g., temperature-induced shifting of water absorption bands). Model robustness
can be improved in some cases where limiting the spectral range will remove
interfering signals while preserving most of the net analyte signal. Finally, it is
often advantageous to retain a contiguous spectrum to ease model interpretation
and to facilitate full-spectrum preprocessing operations.
Most preprocessing operations (besides wavelength selection) are actually
a form of mathematical bandpass (convolution) filter. The most common NIR
preprocessing operations used to mitigate the effects of scatter and instrument
baseline variation are similar to a high-pass filter. These operations include
multiplicative scatter correction (MSC) (98), standard normal variate (SNV)
(108) transformation, and polynomial detrending (108). As its name suggests,
polynomial detrending is simply the operation of subtracting a polynomial fit
of a spectrum.
The SNV transformation is the simplest of the three, does not require the
storage of a reference spectrum, and is calculated according to the following
formula:
x^
x x
sX
where:
x Vector of NIR absorbance intensities (e.g., NIR spectrum)
x^ SNV transformed vector of NIR absorbance intensities
x Mean absorbance intensity for the current spectrum
sX Standard deviation of absorbance intensity for the current
spectrum.
(18)
368
Cogdill and Drennen
The MSC transformation is calculated by subtracting the linear fit of a spectrum against a reference spectrum (typically, the mean of the data set). Although
many experienced spectroscopists would suggest one or the other transformation
to be superior, it can easily be shown that the difference between the two algorithms is quite trivial. The MSC transformation can be derived from the SNV transformation by replacing sX and x from Equation (18) with slope and offset terms
from the linear fit, respectively. Only in certain, nontrivial circumstances will
there be a significant performance difference between the two algorithms (Fig. 27).
Spectral smoothing derivatives are an example of a common preprocessing
operation that resembles a signal bandpass filter; the derivative calculation suppresses low-frequency shapes, while the necessary smoothing calculation
suppresses high-frequency shapes in the NIR spectra. Smoothing derivatives
are used to enhance the subtle features in the NIR spectra and to reduce the
effect of baseline variation (Figs. 28 and 29).
Figure 27 Comparative illustration of the effect of SNV and MSC preprocessing transformations: (A) raw spectra and pure-component spectra from a blending experiment,
(B) SNV transformation of the raw spectra, (C) MSC transformation of the raw spectra,
(D) correlation between the SNV and MSC transformation at single wavelengths (offset
to ease viewing). The wavelength-dependent nonlinear relationship between SNV and
MSC is related to the absorbance intensity, magnitude of baseline variation, and the
source of baseline variation. Both MSC and SNV are expected to induce some small
nonlinear effects in practical application. Abbreviations: MSC, multiplicative scatter
correction; SNV, standard normal variate.
369
Figure 28 Illustration of the relationship between absorbance peak shape, derivative

order, and the resultant derivative spectrum. The first row of panels illustrate shapes in
raw data, the second row is the first derivative, and the third row is the second derivative
of the raw data. In panels A.1 3 it can be seen that increasing the width of an absorbance
band will tend to reduce the intensity of the derivative spectrum. Furthermore, it can be
seen that the zero crossing in the first derivative corresponds to the center of a pure
band (A.2), and the distance between the zero crossings of the second derivative are
related to the band width (A.3). Panels B C.1 3 illustrate the effect of intensity on derivative shape and the impact of derivatives on the ability to detect minor components. The
spectra in panel C.1 are the superposition of the spectra in B.1 and are intended to simulate
the effect of a minor constituent in a strongly absorbing matrix (e.g., drug in aqueous
solution).
A very simple way to calculate a spectral derivative at any given wavelength, l, is the first-difference approximation:
XG XG
2G
XG 2XG XG
second derivative:
2G
first derivative:
where:
G Derivative gap, nm or variables
XG Absorbance intensity at l
(19)
(20)
370
Cogdill and Drennen
Figure 29 Raw, first- and second-derivative pure component spectra of lactose, MCC
PH101, and MCC PH200. Abbreviation: MCC, microcrystalline cellulose.
371
XG Absorbance intensity at l G
XG Absorbance intensity at l 2 G
Thus, by repeatedly applying the derivative approximation [Eqs. (19) and
(20)] to a series of wavelengths from near the beginning of a spectrum to near the
end, a full-spectrum derivative can be calculated. The so-called derivative gap
parameter, G, can be adjusted to vary the bandpass frequency of the derivative.
A narrow setting of G will tend to enhance subtle features in the spectra while
mitigating low-frequency effects but also increase the sensitivity of the derivative
function to noise. A wide setting of G will attenuate high-frequency features in
the spectra while preserving larger-scale features. Hence, the ideal derivative
gap is related to the width of the absorbance bands of interest in the spectra.
Smoothing is accomplished by using a wider G, and by co-averaging (boxcar
smoothing) the absorbance intensity at multiple wavelengths near XG , XG ,
and XG .
The first-difference derivative is sometimes called Gap Derivative, or the
Norris Derivative, with respect to Karl Norris, who was the first to widely
publish the use of derivatives in NIR spectroscopy. The first-difference
formula is especially attractive for use with fixed-filter or imaging spectrometers
for which it is either impossible or too time consuming to measure NIR absorbance at many wavelengths. Norris and Williams have demonstrated that the
ratio of derivative functions at different wavelengths can be used for the prediction of constituent concentration in the presence of significant baseline variation
due to particle size or scatter effects (8).
The derivative function described first by Savitzky and Golay (109) in 1964
is more applicable to continuous, high-resolution NIR spectra and, hence, is used
more often in modern NIR methods. Their method utilizes least squares fitting of
piecewise polynomial functions to continuous signals to create a smoothed
approximation. Thus, the derivative at any given wavelength can be estimated
by evaluating the derivative of a piecewise polynomial fit centered at that wavelength. Rather than repeatedly fitting polynomials (and estimating their derivative) at every variable across a spectrum, Savitzky and Golay showed that
exact local polynomials can be calculated by convolving the spectrum by a convolution kernel derived using least squares criteria. Convolution is a much faster
operation than least squares estimation at every wavelength because only a single
multiplication and summation is required at every convolution step across the
spectrum. Furthermore, they showed that the exact derivative function can be
calculated by convolving the spectrum by the appropriate derivative of the convolution function. Savitzky and Golay include an explicit derivation, tables of
integer convolution filters, and FORTRAN code to implement their method in
the appendix of their paper.
The application of Savitzky-Golay derivatives to NIR spectra requires
proper specification of the polynomial order of the smoothing function, the
convolution kernel width, and the derivative. For most full-spectrum NIR
372
Cogdill and Drennen
applications, stable performance can be achieved by using a quadratic polynomial and a smoothing kernel width of 11 23 variables (assuming 2-nm
increment). As the wavelength increment is decreased, or as the noise level
increases, a wider smoothing function will be more appropriate. Increase in the
polynomial order of the least-squares approximation should also accompany
the increase in filter width. The use of a smoothing filter that is too wide (well
beyond the width of the absorbing bands) will actually produce artifacts in the
derivative trace. The ideal combination of derivative parameters, however,
is generally specific to the instrument and material at hand. Only first- and
second-derivative spectra are typically used for NIR spectroscopy, because each
successive derivative order has the effect of intensifying more high-frequency
noise. The order of baseline effect suppressed by the derivative is one less than
the order of derivative. For example, the first derivative transformation suppresses
zero-order baseline shift, and the second derivative mitigates the baseline slope. Just
as any convolution can be carried out by multiplication in Fourier-transformed
space, spectral derivatives can be calculated by FFT, or by discrete wavelet transform (DWT) with a derivative-of-Gaussian (DOG) mother wavelet.
It may seem somewhat redundant, yet combinations of multiple preprocessing operations are often helpful in developing a sensitive, yet robust NIR
method. For example, common combinations of operations include SNV first
derivative or SNV second derivative. Utilizing normalization after derivative
transformation can be useful, though it is seen less often in the literature. The
use of other convolution kernels, FFT, or wavelet functions (100,110), and a
wide variety of linear projection methods, such as orthogonal signal correction
(OSC) (111 113), generalized least-squares (GLS) weighting (104,114 118),
and direct orthogonalization (DO) (111), along with baseline correction and
derivatives, is not uncommon. Because these are more specialized algorithms,
they will not be further detailed in this chapter. Finally, even though the selection
and optimization of NIR spectral preprocessing seems to be much of an empirical
art form, it is important to keep in mind that each operation is targeted to either
suppress or enhance particular spectral shapes (or variance components) that
define the frequency bandpass of the operation. It is the job of the spectroscopist
(or chemometrician) to match the combination of preprocessing operations to
chemical, physical, and instrumental demands of the application.
Exploratory Data Analysis
Exploratory data analysis is performed to gain a qualitative sense of the significant patterns in the variance of the spectral data. The first steps of qualitative data
analysis are quite simple and straightforward. Once a matrix of new NIR spectra
has been assembled, it is important to begin by simply plotting the spectra (with
and without preprocessing). Although the process is not as glamorous, a careful
visual examination of the raw data will often reveal relationships that might be
obscured by more advanced data analysis methods. Furthermore, it is sometimes
important to look at spectra either one at a time or in small groups, because
373
aberrant scans can sometimes be buried within a plot of many traces. Once the
raw data have been visually characterized, chemometric tools dedicated to exploratory data analysis, such as bootstrap inference analysis (119121), principal components analysis, and multivariate curve resolution, can be applied with confidence.
Principal components analysis (PCA) is a form of eigenvalue decomposition [see singular value decomposition-SVD (122)] whereby a matrix, X, of p
correlated variables (e.g., NIR spectra) is reduced to a subset of latent variables,
P, scores, T, and some residual error, Ex. The PCA algorithm, first described by
Harold Hotelling in 1933 (123), has become an essential tool for multivariate data
analysis across a diverse range of fields. The power of the PCA model is its automatic data compression and orthogonalization capabilities. As a consequence of
the PCA model estimation process, the matrix of principal components, P,
describes an orthonormal basis set, which maximizes the amount of information
captured in the first few components (123):
X TPT EX
(21)
where:
X [n p] matrix of absorbance spectra
T [n r] matrix of principal component scores
PT [r p] matrix of principal component loadings
EX [n p] matrix of spectral residuals
r Number of principal components retained
n Number of samples (spectra) in the matrix.
Each latent variable, Pi, represents the linear combination of the original
variables that captures the maximum residual variance. Hence, for highly correlated systems, such as raw NIR spectra, the first principal component, P1, will
often capture .95% of the total spectral variance, and each subsequent latent
variable will capture a decreasing fraction of the variance until the total information has been captured. Once all of the viable information has been captured
in r principal components, the remaining p 2 r components will be correlated
mostly with random noise and can be omitted (as part of Ex). The scores are
simply the projection of each spectrum into the PCA basis, that is, the covariance
between the spectrum and each individual loading. When considered as a whole,
the scores for the X matrix used to derive the PCA model will be mutually orthogonal. However, the scores of new data (spectra) projected into the PCA model
will not necessarily be mutually orthogonal.
For NIR spectra, chemometricians often notice that the addition of preprocessing operations, such as scatter correction and derivatives, will reduce the
variance captured by the first few principal components. This is observed
because preprocessing suppresses low-frequency components of the spectra,
which typically account for a large fraction of the correlated variance. For
example, without scatter correction, one of the first principal components will
374
Cogdill and Drennen
Figure 30 Logarithm of median eigenvalues during PLS regression using SNV first
derivative preprocessing (solid line) and raw (dashed line) spectra. The whiskers extending above and below each point on the lines illustrates the upper (75%) and lower (25%)
quartiles of the data. The curves show that preprocessing has removed much of the lowfrequency variation that would otherwise dominate the first latent variables. The magnitude of the later eigenvalues (.5) for the preprocessed data are actually larger; this
may be because the preprocessing adds some nonlinear components to the spectra that
are difficult to approximate using a linear factor structure. Abbreviations: PLS, partial
least-squares; SNV, standard normal variate.
generally correspond to the shape (or slope) of the spectral baseline. The absolute
amount of information contained in the matrix does not change with preprocessing, though, so the eigenvalue plots will eventually converge (Fig. 30). It can be
helpful to plot the logarithm of eigenvalues to quickly compare the magnitude of
information captured. For factor models of NIR spectra (e.g., PCA, PLS, etc.),
logarithmic eigenvalues less than 25 are likely to be correlated primarily to spectral noise and will thus be uninformative or unreliable for modeling (unpublished
observations).
The principal component loadings, P, are a mathematical representation of the
unique correlation structures in the data. Thus, for a single factor (Figs. 31 and 32),
regions of large positive magnitude are correlated with one another, as are those
of large negative magnitude, and regions of positive magnitude are inversely correlated with those of negative magnitude. Furthermore, regions of low magnitude
375
Figure 31 Principal component analysis of NIR spectra (A) of moisture loss during tray
drying of granulated lactose. Since the two constituents are bound by closure, the first principal component accounts for 99% of the spectral variation (B). Analysis of the loadings
(C) and score (D) profiles suggests that the first component is correlated to the relative concentration gain of lactose as water is removed. Comparison of the first component loading
and the lactose pure-spectrum (C) demonstrates that the loading is actually the superposition of the lactose spectrum with the inverse of a moisture spectrum (along with the shape
of the baseline variation). The second principal component is related to the nonlinear
change in the water spectrum as moisture concentration is reduced. The vertical axis of
panel (D) is in arbitrary units. Abbreviation: NIR, near-infrared.
have little variance in that principal component direction. By observing the

shapes in a principal component loading it is sometimes possible to deduce the
chemical or physical mechanism(s) responsible for the variance approximated
by that principal component. Spectroscopists sometimes attribute a principle
component to a single constituent (e.g., the first PC appears to be attributable
to API concentration).
Although a single constituent (or physical factor) can dominate a principal
component, the divergent orthogonality and closure constraints of the PCA model
and sample matrix, respectively, strictly prevent such a relationship. In other
words, it is important to keep in mind that a principal component is nothing
376
Cogdill and Drennen
Figure 32 Principal component analysis of SNV preprocessed NIR spectra (A) of moisture loss during tray drying of granulated lactose. The magnitude of the first two eigenvalues (B) barely changed compared to the model without preprocessing (Fig. 31),
suggesting that either the preprocessing did not have a great effect or that it increased
the efficiency of information capture in the initial factors. The later eigenvalues remain
larger for the SNV model, which may be related to nonlinearity introduced by the transformation. The overall shape of the first principal component is similar to the raw
spectra model, but the direction is reversed (C). Furthermore, because the baseline variation has been suppressed, the concentration features in the loading vector are more
intense, and are centered around zero. The second principal component loading (dashed
line in C) is much sharper after SNV preprocessing, but the score profile is the same,
which suggests that the same process is being modeled. The derivative-shaped feature
in the second component (centered near 1950 nm) is similar to hydrate phenomena
described in the literature and is apparently related to the level of bound and unbound
water in the lactose matrix. The vertical axis of panel (D) is in arbitrary units. Abbreviations: NIR, near-infrared; SNV, standard normal variate.
more than an estimate of the linear combination of variables most correlated to a

variance factor in the space spanned by the original variables. It is not guaranteed
to have a direct chemical or physical interpretation. Indeed, experimental design
(of X ) has a major impact on the shape of principal component loadings (124).
Finally, it is important to keep in mind that, unlike PLS factors, which are positively correlated with a target (Y ) matrix, PCA loadings are nondirectional.
377
For example, it may be that if a data matrix is changed only slightly (e.g., some
samples are deleted), the direction of a loading will be reversed (with a corresponding change in the sign of the scores). In other cases, what appears to be
an absorbance band in a principal component loading may actually correspond
to a region of low sensitivity lying between two correlated bands. Thus, if it is
difficult to interpret the features in a particular principal component loading, it
may be helpful to plot the inverse of the vector.
The orthogonal basis sets created during PCA offer a convenient platform
for outlier detection, multivariate statistical process control, and matrix analysis
by investigating individual sample scores, variable influence and contribution
plots (125), spectral residuals, and leverage statistics. A very common application of PCA for qualitative analysis is to determine if there is any substructure
or grouping within the matrix. For example, score score plots are a rapid means
of identifying and analyzing subclasses within a set of samples (Fig. 33) (30).
The residual and leverage statistics for samples projected into a PCA model
are useful for determining the condition of the model (85) and for multivariate
statistical process control (MSPC) (122,125 128). For each 1 x p sample spectrum, xi, the PCA model residual lack-of-fit, or Q statistic, is calculated as the
sum of squared reconstruction error across all wavelengths:
Qik
(xi tik PTk )2
(22)
where:
Qik Sum of squared reconstruction error for ith sample spectrum using
model k
xi Sample spectrum
tik Latent variable scores for ith sample spectrum, using model k
Pk Model k loadings, or eigenvectors.
Sample leverage, or distance-to-model, is evaluated by using Hotellings
T 2 statistic, which is calculated as:
T
Tik2 tik (l1
k )tik
(23)
where:
Tik2 Hotellings T 2 for the ith sample spectrum using model k
tik Latent variable scores for ith sample spectrum, using model k
l1
k Diagonal matrix of normalized eigenvalues of the covariance
matrix for model k.
The normalized eigenvalue matrix is calculated for model k as:
lk
(TkT Tk )
nk
(24)
378
Cogdill and Drennen
Figure 33 Scores plot for a series of tablets produced at varying levels of compression
force, using laboratory-scale single-punch, laboratory-scale rotary, and production-scale
rotary tablet presses. The first principal component was correlated with hardness,
whereas the second component was correlated to tablet press scale. The plot illustrates
the capability of NIR spectroscopy to detect subtle process signatures that would not be
detected by other methods. Although the PCA model was estimated using nonproduction
samples (light gray circles), an independent data set of production-scale tablets projected
into the PCA model (gray squares, black triangles) showed separation. The separation was
attributed to the use of two different tablet presses in the process stream. No quality or performance difference in the dosage forms was observed. Abbreviations: NIR, near-infrared;
PCA, principal components analysis.
where:
Tk Matrix of calibration sample model scores for model k
nk Number of calibration samples used for the estimation of
model k.
As demonstrated by Equations (23) and (24), Hotellings T 2 is a measurement of statistical distance in score space. Hence, T 2 is not influenced by spectral
variation out of the plane (hyperspace) of the model (Fig. 34). A large T 2 indicates
that the sample has high leverage on the model and may exceed the confidence
limits of the model hyperspace, which is consistent with the conditions of
extrapolation. Confidence limits for the T 2 statistic are calculated using a chisquare distribution. Hotellings T 2 is very similar to Mahalanobis distance
(105,129 131). Although both are statistically based distance measurements,
379
Figure 34 Demonstration of the approximation of a three-dimensional dataset by a twocomponent PCA model (panel A). The ellipse illustrates the 95% confidence limits of the
Hotellings T 2 statistic. The projection shown in panel B demonstrates the variation out
of the plane of (orthogonal to) the model, which is measured by the Q residual statistic.
Abreviation: PCA, principal components analysis.
the Hotellings T 2 distance is used for distance measurement among orthogonal

variables (such as principal component scores), whereas Mahalanobis distance
takes into account covariance among the variables directly.
In contrast to T 2, Q describes the lack of fit, or spectral variation orthogonal
to the plane, which is unexplained by the model. Thus, Q and T 2 are completely
independent measurements. A large Q residual indicates that the sample is poorly
reconstructed by the model, which is an indication that either a new factor may be
present in the sample matrix (contamination, phase change, etc.) or that an
instrumental fault has occurred (e.g., sampling error, component failure, etc.).
Soft independent modeling of class analogies (SIMCA) is an extension of
principal component analysis, which was first described by Wold (99,132) in
1976. SIMCA can be used for discriminate analysis, multiclass classification
or as part of an exploratory cluster analysis routine. The SIMCA procedure,
which is in some ways similar to canonical correlation or Fishers linear discriminate analysis (122), begins by dividing the data matrix (e.g., spectra) into m
groups, based either on some sort of clustering routine or by user-supplied reference information on class membership. After splitting the data, a unique PCA
decomposition is performed separately for each of the m groups; each PCA
model can utilize different numbers of latent variables (or other model parameters), which are typically determined using cross-validation. After suitable
models have been estimated for each group, a new sample can be classified by
comparing a weighted combination of the T 2 and Q statistics in each of the
models. Thus, a sample is either attributed to the group with the lowest score
beneath a threshold or is attributed to none of the groups if it does not satisfy
the threshold (based on an F-test criterion) for any group.
Principal component analysis is one of the most basic and widely used
chemometric tools devoted to finding the number and direction of the relevant
380
Cogdill and Drennen
sources of variation in a matrix of bilinear data. A major drawback of PCA for the
analysis of mixtures and process data, however, is the orthogonality constraint of
the principal component loadings. As discussed earlier, since the factors must be
orthogonal, it is impossible to isolate the shapes of individual chemical components using PCA when the mixture is constrained by closure. Self-modeling
curve resolution (SMCR) (133), or self-modeling mixture analysis, is a rapidly
evolving family of algorithms which overcome this drawback through either
direct or iterative means. The goal of all resolution methods is to mathematically
decompose a set of mixed instrument responses into the pure contributions
attributable to each of the active components in a system. Similar to PCA,
SMCR analysis does not explicitly require any a priori knowledge of the
system or chemical reference values, though such knowledge can be incorporated
into the model. For example, given nothing more than a time-series of NIR
spectra of a chemical reaction, SMCR can be used to estimate the concentration
and identity of each of the constituents at any point in the reaction without developing a calibration (134). Alternatively, given a single NIR spectral image hypercube of a tablet or powder sample, SMCR can be used to resolve the data into a
set of pure-component intensity maps that describe the distribution of the chemical constituents in the sample (135).
The self-modeling resolution methods utilize a factor structure similar to
PCA, but the factors are allowed to be correlated to one another as shown:
X CST EX
(25)
where:
X [n p] matrix of absorbance spectra
C [n k] matrix of pure component concentration scores
S T [k p] matrix of pure component sensitivity vectors
EX [n p] matrix of spectral residuals
k number of resolvable absorbing species in the matrix.
There are two types of ambiguity, rotational and intensity, that confound
the process of resolving the true concentration and sensitivity profiles (133).
Rotational ambiguity refers to the fact that there are an infinite number of
pure-component sensitivity profiles (shapes) that can describe X and the true profiles; intensity ambiguity refers to the fact that there are an infinite number of
intensity levels at which C and S T can exist without changing the magnitude of
EX in Equation (25). Unique solutions to Equation (25) can be found, however,
by properly constraining the form of the solution.
The constraints employed during SMCR analysis are mathematical representations of the cumulative knowledge of the chemical, physical, and spectroscopic limitations on the system. For example, the pure component concentration
and sensitivity vectors may be constrained so that negative values are not
allowed, which reflects the fact that (in normal situations) it is impossible to
have negative real optical absorbance or concentration. For mixture systems, a
closure constraint can be used because it is known that, when the matrix is
381
considered as a whole, the concentration of all constituents at all times (or pixels,
locations, etc.) must sum to 100%.
There are numerous articles and reviews (133) on the growing topic of
self-modeling resolution for chemical analysis. Two exciting new areas of selfmodeling research are hard soft mixed model algorithms, which use specific
knowledge of the reaction, such as hard stoichiometric models, to add powerful
constraints to the soft-model solution (136), and multiway analyses (e.g., parallel
factor analysisPARAFAC), which are proving useful in the analysis of batch
systems (125,133). The field of self-modeling resolution analysis is only beginning to be utilized for mainstream analytical tasks and will likely be very useful
as an ever-greater volume of process analytical data must be efficiently analyzed.
NIR Calibration and Estimation
Multivariate calibration (79,98,104) is the essence of quantitative NIR data
analysis. The NIR calibration models can be used to estimate a model for qualitative purposes (e.g., to indicate regions of spectral sensitivity), to estimate
unknown sample qualities within a data set where some samples are of known
quality or to generate predictions of quality for unknown samples. Early
univariate instrumental calibrations utilized so-called classical least-squares
(137) (CLS) regression, whereby the model residual error is attributed to the
instrument data. The CLS method was extended for use with multivariate
systems, such as NIR spectroscopy years later (138 140). The CLS model has
some advantages in terms of interpretability and model transfer/update. A
serious drawback of the CLS method, however, is that the concentration of all
chemical constituents in the calibration matrix, X, must be known, otherwise
the estimation fails due to rotational ambiguity.
Although augmentations (27) of the CLS model have been formulated to
overcome this limitation, most spectroscopic calibrations are based on a firstorder general linear model for inverse least-squares (ILS):
Y XBT EY
(26)
where:
X [n p] matrix of absorbance spectra (dependent variable)
Y [n m] matrix of concentration profiles
B T [m p] matrix of regression coefficients
EY [n p] matrix of concentration residuals
m number of independent variables to be modeled.
It can be seen from Equation (26) that the ILS model is based on the
assumption that the spectra are free of error, are that all the errors are attributable
to the reference data. In practice, this is actually very close to reality, though there
is always some nonzero level of spectral error. Equation (26) is referred to as ILS
because, even though spectral absorbance is dependent on the composition of the
sample matrix, it is treated as being the independent estimator. The ILS model is
382
Cogdill and Drennen
unconstrained by any assumptions related to the pure spectral profiles and is

completely determined by the correlation structure of the data. As a result, the
ILS model can be estimated for a single chemical constituent without specifying
the concentrations of any of the other components in the matrix. Although this
added flexibility is convenient, it shifts a great deal of burden toward responsible
calibration dataset design. For example, depending on the calibration experimental design, the variance of some of the matrix constituents might be aliased with
the component of interest. Thus, although it is assumed that the model is sensitive
to gain in the target component, the model may actually be quantifying the loss of
a secondary component (124). If the ideal calibration design is infeasible, there is
a greater burden on the method developer to analyze the regression equation to
determine the specificity of prediction.
The ILS model, also commonly known as multiple least squares regression
(MLR), is estimated by solving the normal (106,107) equations for the following:
B (X 0 X)1 X 0 Y
(27)
For full-spectrum calibration problems, as is typical for NIR analyses,

Equation (27) suffers during the covariance matrix inversion because of the
multicollinearity among the spectral variables. Spectral preprocessing can
reduce the level of multicollinearity by reducing the baseline correlation, but
the full-spectrum MLR solution may, nevertheless, be unstable. Early efforts to
mitigate the effect of collinearity involved combining MLR and one of the
several methods of variable selection. More recently a number of regression techniques have been developed to address the problem of collinearity directly,
including ridge regression (106,107) and latent variable methods, such as the
PCR and PLS regressions.
The PCR (79,98,106,141 143) is an extension of the general linear model
whereby the matrix of absorbance spectra, X, is precompressed into a basis of
orthogonal factors by using the PCA. The calibration model is then estimated
using the principal component scores in lieu of the absorbance spectra as the
dependent ILS variable:
B P(T 0 T)1 T 0 Y
(28)
The latent variable compression adds complexity to the calibration process

by increasing the number of user-selected parameters. In general, it is best to
minimize the number of parameters required to optimize a method. As the
number of principal components (factors) included in the model is increased,
the significance (eigenvalue) of each additional factor would decrease, thereby
increasing the likelihood that the model will overfit the data. There are many
methodologies suggested to determine the ideal number of latent variables to
be included in a predictive model. Some suggestions include adding latent variables up to the point of a sharp change in the screen plot (Figs. 31 and 32), using
all latent variables prior to reaching a minimum absolute eigenvalue, retaining all
383
latent variables that capture .2% of the variance in the Y (or X ) matrix, or continuing adding latent variables until the loading vectors begin to appear unstable
(noisy). The most defensible tactic, when feasible, is to use independent test sets
or independent cross-validation (e.g., batch-wise cross-validation) (57) to determine the maximum number of informative latent variables to include in the
model. Unless there is a high degree of nonlinearity present, the ideal number
of latent variables to include in a calibration model will typically be within
one component more or less than the number of varying constituents in the
sample matrix. Because principal components are estimated in order of decreasing spectral variance and are not necessarily associated with the target, Y, it is
sometimes advantageous to omit components with larger eigenvalues while
retaining smaller eigenvalue components that are more directly correlated with Y.
The PLS regression (79,98,141 146) is quite similar to PCR in that they
are both biased regression methods utilizing latent variable compression prior
to estimation of a set of regression coefficients using ILS. In many nontrivial
situations, PCR and PLS regression analyses yield equivalent solutions
(142,143). The difference between the algorithms lies in the method of estimating
the basis vector set, P. In contrast to PCR, PLS utilizes an iterative procedure,
whereby an intermediate matrix of weight vectors, W, is formed in a manner
similar to CLS (141), utilizing information from the dependent variable to estimate the model basis vectors. There are two principal algorithms for the estimation of the PLS regression solution. They are: (1) NIPALS (nonlinear
iterative partial least-squares), and (2) the more recent SIMPLS (simple PLS).
Although both algorithms will yield identical results when the target matrix, Y,
is univariate, NIPALS is the preferred method when Y is multivariate (e.g., PLS2) (98,141). In contrast to PCR basis vectors, PLS regression produces a basis
vector set arranged in order of decreasing covariance between the spectral
data, X, and the analytical target, Y. Hence, although PLS and PCR will often
produce virtually identical results, PLS will generally arrive at the solution by
estimating fewer latent variables (142). Consequently, PLS is a very powerful,
user-friendly algorithm for empirical regression analysis and is considered by
many practitioners to be the preferred tool for chemometric regression problems
associated with NIR method development.
A number of recent publications have demonstrated the benefits of a new
hybrid multivariate calibration procedure based on the NAS work of Lorber,
Faber, and Kowalski (147,148). Given a linear system, the NAS of a single component, k, in a multicomponent sample matrix can be computed as the part of its
spectrum orthogonal to the contribution of other coexisting constituents (147).
Thus, NAS processing seeks to reduce the portion of the spectrum used for prediction to a univariate quantity. Although the NAS regression procedure follows
a unique course, it has been shown that NAS regression can produce a regression
solution identical to PLS or PCR (149).
The NAS regression method has been presented in the literature as a generalization of the spectroscopic calibration problem. The benefit of NAS
384
Cogdill and Drennen
calibration cited most often is the ability to calculate useful figures of merit
(FOM), such as selectivity, SNR, sensitivity, and LOD, which may be beneficial
for describing the performance limitations of an NIR method during validation.
The NAS processing has also been suggested as a useful means for variable selection (150) and outlier detection (149). Goicoechea and Olivieri (151) have shown
NAS to be useful for multivariate calibration using fewer training samples.
Unlike ILS regression methods, the unrestrained addition of calibration
training samples with only marginal spectral variation caused by the constituent
of interest should, intuitively, benefit the NAS calculation. This is true because
NAS projection works to remove portions of the hyperspace spanned by the
samples that are subject to spurious correlations. Such a situation is advantageous, because, for typical process analytical applications, production-scale
samples are relatively inexpensive, whereas samples having sufficient range for
CLS and ILS calibration methods may be expensive, difficult, or even impossible
to obtain. As NIR for process analysis becomes more pervasive in pharmaceutical
manufacturing, there will be an increased need for efficient methods of calibration. Methods, such as NAS and a new variant of GLS (104,114 118),
known as pure-component projection (152) (PCP) seek to address this need.
Although the brief introduction to multivariate calibration and data analysis
presented in this section is far from comprehensive, the section would not be
complete without some mention of linearizable and inherently nonlinear
methods of NIR data analysis. Calibration models for concentration-based
sample qualities are expected to follow a nominally linear model, such as
Beers law. Performance characteristics, such as dissolution or disintegration,
however, are likely to be inherently nonlinear. For solution of this type of data
analysis problem there are entire families of algorithms that may prove useful.
The simplest form is the polynomial augmentation of MLR (106,107), which
has been extended to work with latent variable regression (e.g., quadratic
PLS). When very large data sets are available, and if a static model equation is
not required, local-linear methods, such as locally weighted regression and
local-PLS (LOCAL) (153), as well as nonparametric boostrapping methods
(119 121), have been shown to be very powerful for predictive discriminate
analysis of NIR spectra.
The ANN approach has recently been explored for function estimation and
classification in many fields including NIR spectroscopy and pharmaceutical
science (153 158). An ANN prediction model consists of a collection of
nodes (or neurons) and activation functions, each connected by a series of
weight coefficients (Fig. 35). The ANN predictions progress, layer by layer,
from the input nodes, through the layer(s) of hidden nodes, to the output
nodes. At the input layer, the nodes can be raw data points or some other preprocessed form of the data. For NIR spectroscopy, the inputs are often a subset of
latent variable scores in order to avoid problems of multicollinearity. At each
subsequent layer, the node values are a function of the sum of the connected
nodes in the previous layer multiplied by their respective weight coefficients.
385
Figure 35 Graphical illustration of the form of an ANN prediction function.

Abbreviation: ANN, artificial neural network.
The node activation function can be linear or nonlinear, and must be selected
prior to the derivation of the ANN model.
An ANN function is trained by tuning its weight coefficients according to
the squared error of prediction until some stopping criterion is reached. Unlike
linear least squares regressions, such as PLS, which arrive at a unique, explicit
solution by making certain assumptions about the distribution of the calibration
data, ANN iteratively derives an implicit solution using error-gradient backpropagation (159) while making no assumptions about the distribution of the
calibration data. Because the ANN derivation process can be prone to generating
only locally minimal solutions, it is necessary to repeatedly train and test the
model using random starting weights to create the best possible model.
It is easy to see that the ANN model complexity grows quickly as nodes are
added. Due to the high number of trained weights, an ANN could memorize the
calibration set, leading to poor performance when making predictions from new
data. Various preprocessing techniques and training algorithms allow the
386
Cogdill and Drennen
avoidance of model overfitting the training data. Depending on the application,

an enormous amount of training data (e.g., .500 spectra for NIR calibration)
may be necessary to train an accurate, reliable ANN prediction model.
Because of the vast amount of peripheral knowledge required to competently
derive ANN models, they are generally only developed by expert practitioners.
A more in-depth discussion of the ANN methodology can be found by consulting
ANN literature (156,160).
The statistically based methods of nonlinear function estimation have been
developed. These methods may ultimately replace ANN techniques for NIR
method development in situations where complex nonlinear modeling is necessary. Known as support vector machines (161,162) (SVM), these methods utilize a
so-called nonlinear kernel function to transform a nonlinear regression or classification problem into a higher-order space in which a linear solution is feasible.
Indeed, with suitable constraints on the algorithm, it is possible to solve the SVM
optimization using a form of ridge regression (LS-SVM) (161) or even PLS
(163). Additionally, Cogdill and Dardenne (161) have shown that nonlinear
SVM functions can be developed with good predictive ability using far fewer
training spectra than for a comparable ANN.
Method Development
The specific details of NIR method development (164) will vary considerably
depending on the analytical task and with the performance requirements of the
greater system involved. The overall plan, however, is basically the same
whether NIR spectroscopy is being used for basic research or process analysis.
A multipart series of articles by Cogdill et al. (30,57,85) illustrates NIR
method development for a process analytical application. The remaining
portion of this section briefly describes the basic steps of the NIR method
development: (i) risk analysis and feasibility studies, (ii) design of experiments,
(iii) data analysis, (iv) validation, and (v) maintenance.
The selection of a particular type of instrumentation is determined by the
definition of the analytical task, which is one of the results of risk analysis and
feasibility studies. Once the analytical task has been defined (whether it is
process analytical or for research), the choice to use NIR or some other technology should be dictated by careful consideration of the risk/benefit profile of all
feasible methods. If NIR is indeed the technology to be employed, a quantitative
risk and performance evaluation of the various instrument types (and vendors)
should be undertaken. Some aspects to consider include speed and sensitivity
requirements, instrument size and accessability, cost, and the available knowledge base. The technology feasibility study should yield a clear choice of
technology, and an estimate of the scope of data that will be required for
calibration or model development.
The NIR model development (either quantitative or qualitative) can either
proceed through passive sample selection, where calibration samples are drawn
387
from routine operations, or through active experimentation, where calibration

samples are created according to a designed experiment, which may or may
not exceed the design space (165) of the process. Although passive sample
collection is typically useful for many qualitative analyses, well-controlled processes (which are the norm for pharmaceutical solid dosage manufacturing) offer
relatively little matrix variation. This lack of leverage can make it difficult to
interpret observations from qualitative models and can prevent the development
of quantitative models with acceptable robustness and specificity when using ILS
regression methods (152). Thus, some portion of the model development and
validation datasets must usually be generated from non-production samples.
The specificity of an ILS regression solution is contingent on the calibration of experimental design (DOE). Inadequate designs may produce spurious
or unspecific correlations that are ultimately unreliable in a predictive situation.
Spiking of the nominal, or production-scale, mixture with pure API is a common
practice that is often seen in published applications of NIR for solid dosage
content uniformity analysis. Although this method is suitable for feasibility
studies, it introduces an aliasing effect with the inactive excipients in the
dosage matrix. In order to reduce the potential for such effects, NIR calibration
models that are to be deployed in a process analytical environment are often
developed using hundreds, even thousands, of calibration samples having significant variation balanced in all controllable factors (e.g., composition, material
source, particle size, hardness, etc.). Current research is aimed at reducing the
number of nonproduction samples, and the associated cost of reference testing,
required for robust NIR method development (152).
Data analysis is one of the most important and engaging aspects of method
development. It is during the data analysis phases that hidden processes behind
sample variation begin to be understood and performance limits of the method
are defined. The NIR data analysis is largely an exercise of prudent data management and optimization through simulation. Problems with NIR calibration
methods that ultimately fail upon deployment will often stem from a failure to
properly segregate the training, testing, and validation data to adequately preserve independence and diversity in the respective datasets. Specifically, it is
absolutely critical that adequate independence is available in the calibration
dataset (e.g., multiple distinct batches) to avoid overfitting of the training data.
Nearly every qualitative or quantitative NIR method will involve the optimal
determination of model parameters, such as the number of PLS factors to
include. Truly independent test sets, or batch-wise cross-validation, should be
employed to ensure that parameter selections are optimal, if not conservative.
The performance limits of the NIR method can be determined by utilizing simulation (e.g., addition of synthetic noise sources and variation to the training data),
or by bootstrapping (119 121). Finally, the data analysis portion of method
development may ultimately reveal the need to augment the calibration data
set with a more extensive DOE, select alternative instrumentation, or revise
the analytical goals.
388
Cogdill and Drennen
Method validation is a critical part of the development process. However, if

the prior steps of method development have been completed in a rigorous manner,
validation is simply a formal description of the method and its performance limitations. This augments with the testing of method performance using appropriate
independent validation samples. Besides the validation of the model, effort is
necessary to test the capability of the entire system (e.g., sample presentation,
systems integration, etc.). Detailed descriptions of analytical method validation
are described in the International Conference on Harmonization (ICH) Q2A
and Q2B guidelines, USP monograph ,1119., and FDA guidance documents
(69), as well as agricultural guidances (164). Even though the basic validation parameters (accuracy, precision, specificity, linearity, range, and robustness) are
common among NIR methods and primary methods, such as UV VIS spectrophotometry, the definition, description, and means of determining each of these
parameters is somewhat unique for each application and for NIR spectroscopy
as a whole. As the use of NIR as a critical process analysis tool grows, literature
examples describing validation procedures will continue to define the requirements for NIR methods in diverse industrial pharmacy applications.
The final phase of NIR method development involves deploying, maintaining, and propagating the method. Even though some instrument manufacturers
have made considerable improvement in the uniformity and stability of NIR analyzers, calibration models that will be utilized over long periods of time, or in
numerous installations, will ultimately require transfer or updating (85). Calibration transfer describes either the propagation of the mathematical calibration
model to new instrumentation (or installations), or the redeployment of a model
after an instrument has been substantially modified. The goal of calibration transfer is to derive a function that changes either the analyzer response or the calibration model so that instrument output after transfer matches the method
performance observed during calibration development. Numerous mathematical
techniques are developed for the estimation of the calibration transfer function; a
more detailed description of these methods is available in the current literature
(166). A common problem for NIR calibration transfer is that standard reference
materials are typically not sufficient for the development of the transfer function.
This occurs because the shapes of the NIR spectra are partially the result of the
geometry of the instrument relative to the scattering pattern of the sample. Thus,
not only must standard reference materials have suitable absorption characteristics, it is believed that their scattering pattern must match that of the sample
material to be useful for calibration transfer. Although research in this area is
ongoing, the current solution is to maintain a dynamic supply of calibration
standard rescue samples for the purpose of maintaining the calibration in the
event of instrument failure (85). The composition of the rescue set, in terms of
the number of samples and spectral diversity, is anticipated to be very application
specific.
In some cases, the performance of an NIR method might be reduced
because of a slight change in the product over time as a result of process
389
equipment wear or due to drift in raw material characteristics. Due to the sensitivity of NIR spectroscopy, even though a drug product or material may change
very little in quality (even staying well within the space of acceptable product),
the performance of a calibration model may vary significantly. To mitigate the
effects of such method drift, the model should be periodically updated by
adding new sample data as part of a comprehensive parallel testing regime.
The addition of new calibration samples will alter the distribution of data in
the model, as a result of which the model performance might change.
However, as suggested in ICH Q2A (167), a new validation exercise should
not be required as long as the specifications for calibration update are set a
priori (the same applies for calibration transfer). Hence, it is important to adequately describe calibration transfer, update, and maintenance routines as they
fit into the overall analytical procedure prior to validation and deployment. If
calibration update or transfer significantly alters the performance of the
method or is a deviation from a validated procedure, it should be viewed as recalibration; in these cases it is most prudent to undertake a new validation
study. Indeed, knowledgeable use of chemometrics, and some method understanding, can build robustness into the model, greatly reducing the extent and frequency of effort in calibration update or transfer.
APPLICATIONS IN PHARMACEUTICAL SOLIDS ANALYSIS
The previous sections of this chapter have described the theory, equipment, and
methods for NIR analysis of pharmaceutical solid materials. The objective of this
section is to provide a brief summary of how NIR spectroscopy is applied in the
pharmaceutical sciences, highlighting the areas that are sometimes unique relative to other fields where NIR spectroscopy is used. Although this section is
devoted to covering a broad range of pharmaceutical applications and references
will be provided where appropriate and available, it is not meant to be a comprehensive review of the literature. Thorough reviews are presented elsewhere
(9,11,168). For the purpose of organization, the NIR applications discussed in
this section will be divided into three categories: (i) basic materials analysis,
(ii) in-process analysis, and (iii) analysis of finished products.
Basic Materials Analysis
For a number of reasons, NIR spectroscopy is well-suited for many fundamental
analyses of the performance of pharmaceutical materials. Such analyses, typically performed using at- or off-line instruments, are performed during formulation and process development stages. Many applications of the NIR for basic
material analysis are focused on the content and character of moisture in pharmaceutical materials. Additionally, the micrometric properties of pharmaceutical
materials are important relative to product performance characteristics. The
details of such analyses are already covered in the first section of this chapter.
390
Cogdill and Drennen
Beyond these applications, the following paragraphs will discuss the application
of NIR for the analysis of pharmaceutical polymer materials and the quantitative
and qualitative determination of crystalline morphology.
Analysis of Pharmaceutical Polymers
As modern pharmaceutical drugs and therapeutic regimens continue to grow in
complexity, the share of advanced solid dosage forms and delivery systems utilizing natural and synthetic polymers will increase. Polymeric materials,
especially cellulose derivatives, have served an important role in the manufacture
of immediate- and controlled-release tablets for many years and are nearly
ubiquitous in modern solid dosage forms as either fillers or coating agents.
Hydrophilic polymers, such as poly(ethylene glycol) (PEG), poly(lactic acid)
(PLA), or poly(DL-lactic-co-glycolide) (PLGA), are increasingly utilized for
sustained-release implant dosage forms (169) and for the delivery of proteins,
peptides, or DNA in the form of nano- or microspheres. For every application
of polymers in pharmaceutical drug formulation, there are associated polymer
quality parameters which have a significant impact on dosage form performance
or manufacturability. Furthermore, as is the case of compressed matrix tablets,
the quality, content, and distribution of individual polymers can have a significant impact on drug release. For all of these reasons, there is great need for
rapid, accurate and nondestructive means for measuring polymer quality and concentration in pharmaceutical raw materials and dosage forms.
As a class of materials, polymers are excellent subjects for NIR analysis. In
addition to the fact that most polymer residues are rich in 22CH, 22NH, and 22OH
intramolecular bonds, the critical performance parameters of polymers, such as
viscosity, yield strength, or glass transition, can be correlated to the long-range
bonding structure of the material (e.g., chain length, cross-linking, crystallinity),
residual monomer concentration, or the degree of substitution, which are readily
detected using NIR spectroscopy. Consequently, the examples of NIR applications for polymer analysis in the chemical, petroleum, and textile industries
abound. Lachenal published a review paper discussing many aspects of NIR
spectroscopy for the analysis of polymers, including instrumentation, process
quality control, monitoring water sorption characteristics, morphology, and crystallinity (170). Other general references on NIR spectroscopy of polymers
include a chapter in the general NIR text by Siesler et al. (171) and a book on
modern polymer spectroscopy by Zerbi et al. (172).
One of the first published applications of NIR spectroscopy for the direct
analysis of pharmaceutical polymers, undertaken by the Lodder group in 1994,
involved the determination of moisture content of intact gelatin capsules (173).
Studies indicated that repeated or prolonged exposure of gelatin capsules to
high humidity conditions resulted in decreased drug release upon dosing. It
was noted that gelatin stored in high-humidity conditions experienced changes
in conformation and cross-linking, which led to the formation of a swollen,
water-insoluble gelatin (pellicle), thereby reducing the in vitro dissolution rate.
391
Thus, NIR spectroscopy was explored as a rapid, nondestructive means of

determining the water content and, hence, the likelihood of decreased dissolution
rate. Similar studies were performed much later by Berntsson et al. (174). Later
studies by the Lodder group (175) utilized NIR spectroscopy to rapidly detect
formaldehyde-induced cross-linking in soft gelatin capsules.
As discussed in earlier paragraphs, it is well known that the quality of polymeric excipients has an important impact on the physical and performance attributes of compressed tablets. The NIR spectroscopy has become a valuable tool
for rapid identification and qualification of pharmaceutical excipients. Early
studies by Gemperline et al. (129 131) concerning raw material identification
and qualification demonstrated the successful application of NIR spectroscopy
and chemometrics (SIMCA) for correctly identifying cellulose derivatives with
varying levels of quality or impurities. Indeed, they showed that NIR spectroscopy was able to distinguish between individual batches of polymeric excipients from single and multiple suppliers. Svensson et al. (176), later demonstrated
the combination of NIR spectroscopy and similar chemometric methods to classify 11 cellulose derivatives among more than 400 batches of pharmaceutical
excipients with a high degree of accuracy.
Beyond simply identifying and qualifying polymer excipients based on a
qualitative NIR test, Gustaffson et al. (177) recently published their work utilizing NIR and FT-IR spectroscopy and chemometrics to accurately predict tablet
performance quality parameters from the spectra of raw hydroxypropyl methylcellulose (HPMC). Their work, which is a continuation of earlier work which
explored the relationship between HPMC quality parameters and tablet performance, analyzed 12 grades of HPMC in terms of methoxy- and hydroxylpropyl
concentrations, total degree of substitution, methoxy/hydroxypropyl ratio, Heywoods shape factor, apparent density, and specific surface area. Tablet performance parameters analyzed included radial tensile strength, axial tensile strength,
specific surface area, pore radius, elastic recovery, yield pressure, and drug
release after one hour. They utilized principal component analysis of the raw
material and tablet parameters to identifiy correlations. In contrast to earlier
studies by other researchers, Gustafsson, et al. observed that particle shape and
powder surface area were more critical in predicting compaction behavior than
the degree of substitution of the HPMC powders. Both NIR and FT-IR were
useful for predicting excipient quality parameters and tablet performance.
Although both techniques demonstrated high correlations, the predictive ability
of NIR was superior to FT-IR in all cases.
Although the opportunity to measure the physical and chemical characteristics of polymers with NIR is well established, the ability to optimize a pharmaceutical manufacturing process based on the quality of incoming raw materials
will require additional study. Such studies will ultimately allow pharmaceutical
production processes to be controlled rapidly, and with flexibility, to variable
quality of incoming raw materials via real-time process analytical measurements
and feed-forward controls. Alternatively, such functional modeling might be used
392
Cogdill and Drennen
to set evidence-based limits on the quality of incoming raw materials, thereby

replacing more arbitrary qualitative methods, which rely on pattern recognition
algorithms, such as SIMCA. The majority of published process analytical applications of NIR spectroscopy involve the adaptation of NIR spectroscopy to existing processes. However, by deploying NIR spectroscopy for exploratory studies
much earlier during formulation and process development, the establishment of a
functional product or process design space [as described in ICH Q8 (165)] might
be significantly augmented by increasing both the speed and value of data
collection.
Tablet coating analysis is yet another polymer analysis application for
which NIR spectroscopy is particularly well-suited. Coating thickness and uniformity impact dosage form performance, especially in terms of drug release.
A direct correlation can be observed between the thickness of the coating layer
and the NIR signal (Fig. 36). Precedence suggests, however, that analysis
below the coating surface is possible through judicious application of chemometrics (e.g., content analysis of coated tablets, quantification of multiple
coating layers). Kirsch and Drennen (55) demonstrated the effectiveness of
NIR spectroscopy for the prediction of coating thickness, dissolution rate
(t50%), and core hardness of theophylline tablets coated with ethylcellulose.
Their results proved that NIR spectroscopy can be used to quantify tablet
quality parameters through a polymer coating. Recently, Perez-Ramos et al.
(178), building upon the early work of Kirsch and Drennen, utilized NIR spectroscopy to estimate the rate constants of a semiempirical model for tablet
Figure 36 NIR spectra of coated solid dosage tablets. Tablets were coated for four
different times: 0, 20, 40, and 44 minutes. The effect of increased coating time can be
observed in the increasing baseline offset, and the sharp features near 1180 and
1680 nm. Abbreviation: NIR, near-infrared.
393
coating thickness. In developing a calibration model for the prediction of coating

thickness, they observed that the calibration regression vector utilized both the
positive correlation with the increase in coating thickness, and the negative correlation with the decreasing intensity of absorbance bands related to tablet
constituents.
Functional polymer coatings, such as controlled-release films, multilayer
films, and drug-laden polymer films are becoming more common means for
dosage form design. The spatial uniformity of functional coatings is a critical
quality parameter for process development or control. The NIR imaging spectroscopy is a powerful means for rapidly assessing coating thickness and uniformity during process development. By correlating NIR absorbance with
coating time, a sensitive indicator of coating thickness can be estimated
without complicated reference testing. By projecting NIR image spectra onto
the regression vector, a pseudocolor map of coating uniformity can be created
for visual interpretation of coating uniformity (Fig. 37). Indeed, a similar qualitative prediction vector might be estimated by subtracting the spectrum of the
uncoated tablet core from the pure coating spectrum.
Arguably, the future of drug delivery technology will involve sustainedrelease or programmable-release implants and dosage forms, as well as advanced
methods for delivering protein and peptide drugs parenterally and, eventually,
orally. The performance of these systems will rely heavily on the drug diffusion
characteristics of polymer matrices, membranes, or hydrogels. For example, the
uniformity of the programmable pulsatile release profiles of modular AccudepTM
Figure 37 Sample images of coated tablets at 2 (upper image) and 10 (lower image)
magnification. The coating thickness calibration was developed by regressing single-point
NIR spectra against the total time of coating. Tablets were precoated with Surelease before
being coated with Eudragit for 0, 20, 40, and 44 minutes. The images clearly show the
coating thickness variability for early time points. Abbreviation: NIR, near-infrared.
394
Cogdill and Drennen
dosage forms (179,180) relies on reliable characterization of the diffusivity of

polymer membranes. It is easy to imagine how NIR imaging spectroscopy
could be used to qualify the uniformity of the physical characteristics of the
film sheets used to create the diffusive films or to investigate the physicochemical
properties related to film performance. For polymer implants, the rate of drug
release can be controlled by dispersing inert, water-soluble excipients throughout
the polymer matrix. The dissolution of these excipients in vivo creates a pore
structure for drug diffusion and release. Hence, control of the concentration
and distribution of these excipients is an important quality control parameter in
the manufacture of such dosage forms. Brashear et al. (169), demonstrated the
use of NIR spectroscopy to quantify the content of PEG and loading of Lomefloxacin in poly(1-caprolactone) (PCL) matrix implants. They utilized ratios of
second derivative functions at two wavelengths, as described by Norris and
Williams (8), to develop calibration models with very high correlation for both
constituents.
Polymer hydrogels are increasingly being considered for sustained release
of injectable biotechnological therapeutics, scaffolds for release of growth factors
in tissue engineering, and oral delivery of gastric labile protein or peptide therapeutics. Continual concerns with these systems include the quantification of drug
loading and characterization of drug release potential via diffusion. Early work in
this area by Nerella and Drennen (158) utilized dispersive NIR spectroscopy with
a novel controlled aperture to develop a method for depth-resolved determination
of drug content in a polymer hydrogel. They utilized the NIR method to solve the
Fickian diffusion constants of salicylic acid through a hydrogel matrix, which
served as a practical in vitro model for a transdermal drug delivery system. In
much later work, Blanco and Romero (181) demonstrated the use of NIR transflectance spectroscopy for the determination of dexketoprofen in a hydrogel.
They noted that because hydrogels have poor flow characteristics and a highly
aqueous structure, they are difficult to characterize using either transmittance
or reflectance methods; however, transflectance, via a cuvette modified with a
gold-plated reflector allowed for repeatable sample presentation and for the
development of a robust, quantitative calibration model. As multicomponent
hydrogels see more routine use in critical drug delivery and tissue engineering
applications, NIR spectroscopy may enable in situ feedback control of drug
loading operations, perhaps to compensate for variability in hydrogel structure,
thereby stabilizing the delivery characteristics of the system in vivo.
Crystalline Morphology
The NIR spectroscopy has been applied extensively for detection and quantification of changes in crystalline forms of numerous pharmaceutical solids.
Although variation in polymorphic form might not be expected to have a significant NIR signature, because there is no change in composition, per se, the
realignment of crystal lattices associated with unique polymorphs does often
cause a noticeable, although sometimes subtle, variation in the NIR spectra.
395
This may often be due to variations in intermolecular hydrogen bonding. The

analysis of crystalline hydrates will not be discussed at length in this section
because they are covered in the fundamentals section related to the spectral
characteristics of water.
The first application of NIR spectroscopy for the analysis of crystalline
morphology was published by Gimet and Luong in 1987 (182). Their work
was concerned with the quantitative determination of two polymorphic forms
(a and b) of an unnamed drug compound in a multicomponent formulation
matrix. The more stable form, b, exists as fine acicular crystals (2 50 mm).
Certain manufacturing operations, however, cause enantiotropic conversion to
form a, which exists as larger rectangular plates (60 230 mm); thus, a rapid
method of accurately determining the concentration of the two forms in a drug
product matrix was needed. The baseline shapes of pure-component spectra of
the two forms were basically the same; however, the spectrum of form a had
many more sharp spectral features. Although the authors chose not to venture
a qualitative explanation for the differences in absorption between the two
forms, they did develop a simple quantitative calibration model for the prediction
of the content of both forms in a drug matrix.
Buchanan et al. (183) published their work investigating NIR spectroscopy
for the determination of enantiomeric purity of valine in 1988. The determination of optical isomers is typically performed in solution using circular
dichroism or optical rotatory dispersion spectroscopy. The authors noted that
an DR NIR spectroscopy of solids is inherently insensitive to optical rotation.
(DR efficiently randomizes the state of polarization.) However, absorbance
differences between the d or l and the racemic dl forms of valine arise due to
differences in the orientation (packing) of groups in the crystalline matrix
which affect hydrogen bonding.
Because the d and l forms of valine are structurally indistinguishable, a
calibration model was developed to predict the ratio of the two forms according
to the following equation:
R
x
1x
where x is the mole fraction of the minor component. Thus, for a racemic
mixture R would be equal to one, and for pure d or l, R would be equal to
zero. The fact that the major and minor components cannot be determined spectroscopically is of no consequence, because it would typically be known a priori.
A series of 14 physical mixtures of the d and l isomers, which spanned the range
of R values from 0 to 0.99, were prepared and analyzed by NIR reflectance. The
calibration model, which was optimized using wavelength selection and MLR,
used three wavelengths for the prediction of R with correlation greater than
0.99. A spectral reconstruction algorithm was used to interpret the calibration
model by comparing the relative shape of band intensities. The first two wavelengths, 1562 and 1576 nm, were not located in conjunction with an observed
396
Cogdill and Drennen
absorbance peak and were suggested to function as baseline correction factors in

the calibration model. The third wavelength (1660 nm) was selected just off of
an inflection point of the first overtone C22H stretch absorbance band. The
authors suggested that because the major regression coefficient was located
just near an absorbance intensity maximum, the spectral phenomenon being
detected was likely a band shift. However, no further investigation of the spectral
signature was discussed.
In 1989, Miller and Honigs (184) published what was, at that time, the most
detailed analysis of changes in NIR spectra due to differences in crystalline form.
They presented a detailed analysis of the spectral differences between the a and g
crystalline forms of glycine. Rather than developing a calibration model for predicting the content of one or the other forms, Miller and Honigs focused on comparing the relative spectral shapes with expected bands based on mathematical
combinations of the fundamental frequencies observed for the two forms.
Their results supported the earlier works, which suggested that the basis for
NIR discrimination of polymorphic forms was related to differences in hydrogen
bonding. Specifically, the authors noted intensity changes in O22H and N22H
absorbance bands and shifts of 4 5 nm in the location of C22H vibrations.
Their results correlated well with earlier observations using Raman spectroscopy
and with band locations predicted using mathematical combinations of the fundamental absorption bands.
Three studies of NIR spectroscopy for the quantification of crystalline form
for a variety of pharmaceutical substances were published in 2000 by the Luner
group of the University of Iowa in conjunction with Foss NIR Systems. Binary
mixtures of two forms of six different compounds were analyzed to create a
series of calibration models. They are as follows: Form-I and Form-III sulfathiazole (185,186), Form-I and Form-II sulfamethoxazole (185), ampicillin
anhydrate and trihydrate (185), b-lactose and a-lactose monohydrate (185),
amorphous and crystalline indomethacin (187), and amorphous and crystalline
sucrose (187). As described in the three papers, DSC, TGA, and X-ray powder
diffraction (XRPD) were used to verify the different crystalline forms. Although
some qualitative interpretations of the spectroscopic results were given, the
overall objective of the experiments was to determine the suitability of NIR spectroscopy for quantitative prediction of the content of particular crystalline forms.
Good predictive performance was observed in all cases, with concentration limit
of quantification (LOQ) as low as 0.3% (186). These results should not be surprising, though, given that the calibrations and predictions were generated using
simple binary mixtures. Other work by the Luner group published in 2000
demonstrated the ability of the NIR to predict the relative concentration of
Forms-I and II sulfamethoxazole in mixtures diluted with up to 90% of a pure
absorbing excipient; LOQ was observed to increase marginally to 1 2% (188).
Similar experiments using binary physical mixtures of Forms-I and III carbamazepine were performed (with similar results achieved) in more recent work by
Braga and Poppi (189).
397
Otsuka et al. (190) published the first assessment of NIR spectroscopy for
the determination of crystalline form of drug compounds in finished dosage forms
in 2003. The objective of their work was to develop an NIR calibration model to
predict the relative concentrations of a- and g-indomethacin in tablets. The
researchers developed a series of calibration samples by compressing (at constant
pressure) mixtures consisting of a 50:50-w:w combination of a multicomponent
excipient matrix and indomethacin. The composition of the indomethacin in the
mixture ranged from 0 100% g-indomethacin. The performance of the calibration was evaluated by analyzing a series of samples with unknown relative
levels of a and g indomethacin obtained by a metastable transformation
process; the reference test in all cases was XRPD.
In analyzing the calibration samples, Otsuka et al. attributed most of the
spectral features used to discriminate between the two forms to C22H stretch
vibrational absorbance bands. The researchers observed no loss in accuracy or
precision in the prediction of polymorph content in tablets compared with
powder mixtures. Furthermore, good correlation was observed between the
NIR and XRPD predicted values for the unknown samples, though NIR demonstrated lower standard deviation of repeatability. Thus, the authors suggested that
NIR should be useful (if not superior to XRPD) for the analysis of polymorphic
purity in finished pharmaceutical products.
Process Development and Scale-Up
The majority of the NIR applications referenced throughout this chapter have
been concerned with demonstrating the sensitivity of NIR to, or the development
of an NIR calibration for, certain drug product components or quality characteristics. Without detracting from the significance of these efforts, it should be noted
that the development of NIR methods and calibration models are small steps
toward a larger goal of utilizing NIR spectroscopy along with other tools to generate a greater understanding and control of pharmaceutical manufacturing processes. Thus, as the application of NIR spectroscopy for the analysis of
pharmaceutical solids continues to mature, NIR sensors will increasingly be utilized as a basis for mechanistic and control models of the processes being monitored. For example, Davis et al. recently demonstrated the use of real-time NIR
process spectroscopy in developing a time-based model of polymorphic phase
transformation during the granulation of g-glycine (191).
The advantage of this sort of methodology is that, because NIR analyses
can be performed independent of the scale of the operation, process controls
can be created using NIR spectroscopy during the process development phase
and applied during scale-up to production levels. Furthermore, the complex patterns in the time profiles of predicted values during in-process analyses will be
useful in determining the kinetics of certain operations. Methods and examples
of NIR spectroscopy for in-process analysis will be discussed in the following
paragraphs.
398
Cogdill and Drennen
In-Process NIR Analysis

Among analytical technologies, NIR spectroscopy is unique in its combination of
sensitivity, robustness, flexibility, and economy. Thus, it is not surprising that the
number of NIR process applications increased rapidly long before the PAT guidance was drafted. Process analytical installations of NIR spectroscopic instrumentation are expected to grow rapidly. Many industry observers have noted
that PAT is not about sensors, rather, it relates to continuous management of
quality through the reduction of variability. Sensors, such as NIR, however,
are necessary to gather the relevant quality data in a timely fashion and, hence,
are integral to PAT. On-line NIR instrumentation, such as diode array spectrometers, typically gather spectroscopic data at rates sufficient to continuously
characterize both the instantaneous mean and variance of many process
operations. By analyzing the mean and variance in quality parameters for a particular process over time, estimates of process risk and capability [e.g., CpK
(192)] can be monitored for trends. Additionally, because NIR spectroscopy
offers multivariate detection, the capability of multiple individual aspects of
operations can be assessed simultaneously. In addition to quality monitoring,
the most compelling applications of in-process NIR spectroscopy will be to
enable the sort of feed-forward control anticipated for future batch and continuous manufacturing operations. Without real-time product quality assessment at
each critical-to-quality (CTQ) operation, there is little information to be fed
forward through process transfer functions (192).
Although in-process NIR spectroscopy is applied in fundamentally the
same manner as for basic materials analyses, there are many unique aspects of
process-spectroscopy worth noting. The most significant differences are related
to the type of instrumentation used and the characteristics of the data analysis
methods. In-process NIR analyses utilize the types of on- or in-line instrumentation described in earlier sections. Furthermore, in-process calibration models are
often optimized for robustness, efficiency, and simplicity, sometimes at the
expense of sensitivity. Calibration models to be deployed for in-process testing
may be qualitative or quantitative and should be developed in a way that accounts
for a sufficient portion of the expected process variability (ref. PAT II). The
primary characteristic by which in-process NIR applications can be classified
is the method of sample interaction. The basic means of sampling pharmaceutical
operations by in-line NIR spectroscopy can be classified as either interactive or
remote sensing.
Interactive Sampling Systems
For our purposes, interactive sampling refers to the means and methods used to
acquire spectra where some part of the instrument is in contact with the sample
material. Interactive transmittance measurements of in-process pharmaceutical
solids are less common relative to other methods of interactive sampling, due
largely to the fact that it can be difficult to reliably transport solids through a
399
transmittance cell in a repeatable manner. Though absorptivity in transmittance is

sometimes more linear than DR over a wide range of concentrations (unpublished
observations), the increased complexity of sample handling erodes most tangible
benefits. Thus, in-process analysis of pharmaceutical solids using interactive NIR
transmittance will not be further discussed herein.
Interactive DR, on the other hand, is a very common technique for
in-process NIR analysis. For interactive DR, the in-process sample material is
typically in contact with some sort of window or lens. For example, fluidized
bed processes (e.g., granulation or drying) are monitored using NIR spectroscopy
by inserting a quartz or sapphire window in the wall of, or protruding into, the
processor at the static bed height. As the materials inside the processor tumble
in the airstream, a continuous fraction of the solids will be in dynamic contact
with the window, which allows a reflectance spectrum to be acquired by an
NIR spectrometer on the outside of the processor. Rantanen et al., have published
the details of many experiments demonstrating through-wall NIR spectroscopy
for monitoring and control of fluidized bed drying and granulation
(21,25,42,193 200). More recent NIR blend monitoring systems, based on
RF-coupled wireless diode arrays, use a fixed window for both illumination
and data collection. As the blender rotates, the spectra are acquired during the
periods when powder is flowing over the window (88).
One of the advantages of interactive DR (relative to remote sensing, discussed later) is that the sampling geometry is stable because the distance
between the spectrometer and sample is controlled by the window. There will
be, however, a certain amount of unavoidable spectral variability related to the
packing density of the material pressing against the window. There are also a
number of potential drawbacks to NIR measurement through a window into
the process. An obvious consideration is the path of illumination; for instruments
where the lamp is separated from the spectrometer, it can be difficult to align the
illumination source and instrument detector through the same window. For
instruments using a fiber-optic probe facing the window, reflection from the
window surface can dramatically reduce the dynamic range of the measurement
unless an antireflective (AR) coating is applied. Furthermore, variation in the illumination and detector geometry relative to the window can have complicated
effects on the spectral baseline (Fig. 38). Thus, if the through-window NIR
sampling system is not rigidly attached to the processor, some instrument positioning variation will be present in the sample spectra.
Another difficulty encountered during interactive DR (and other interactivetype sampling methods) is the interaction of the sampling window with the
in-process material. For example, quartz glass windows or AR coatings can be
degraded by abrasive, reactive, or high-temperature flowing solids. With
respect to NIR reflectance analysis of typical pharmaceutical materials, a more
significant problem is the buildup of a layer of sample materials on, or in front
of the window, also known as window fouling. One manufacturer of filterbased NIR sensors for process analysis has developed a system whereby an
400
Cogdill and Drennen
Figure 38 Scatter-corrected (SNV) NIR spectra of a mixture of acetaminophen, lactose,

and MCC in a laboratory scale blender. The spectra were acquired using a fiber-optic probe
pressed against an antireflective coated window in the side of the blender. Each spectrum is
the mean of 50 spectra gathered from the same blend after repositioning the probe. As
shown, there is variability in the spectral baseline for the wavelength region beyond
1500 nm. Although the observed effect may be related to the composition balance
between lactose and MCC, the effect has been observed to be repeatable. Supposing the
spectral variation to be the result of probe alignment with the coated window illustrates
the importance of considering all sources of error when developing NIR methods. (The
apparent variation near 1020 nm is probably the result of SNV preprocessing because,
as the baseline is reduced beyond 1500 nm, the baseline must increase in other areas of
the spectrum to compensate.) Abbreviations: SNV, standard normal variate; NIR, nearinfrared; MCC, microcrystalline cellulose.
automated wiping mechanism periodically clears the window of residue (201).

Although the ability to automatically clean in place (CIP) is attractive, there
must still be a determination of the ideal cleaning frequency and some means
of dealing with the level of residue between cleanings. Hence, it would be a
better to devise a means of avoiding the buildup of residue altogether. Rantanen
et al. (195) prevented fouling during fluidized-bed wet granulation by continuously heating the optical window to a temperature greater than that of the
in-process material. As other researchers have corroborated (unpublished correspondence), heating NIR windows or probes beyond the temperature of the
process is an effective means of reducing fouling in the presence of wet materials.
401
Figure 39 Spray ring for use with an NIR process window. The fluid spray ports located
around the periphery of the inner surface are designed to continuously wash the optical
window with an inert gas or solvent. Abbreviation: NIR, near-infrared.
Another alternative would be to continuously wash the surface of the window

with a layer of air or inert gas. For example, J.M. Canty Process Technology
(202) has developed an integrated spray ring system for use with their fused
silica process windows. By spraying a constant volume of air through a series
of ports around the periphery of the window bezel (Fig. 39), wet or dry materials
can be deflected from the window.
For obvious reasons, fiber-optic probes are one of the most common means
for interactive sampling and are an advantage that sets NIR (and Raman) spectroscopy apart from other analytical technologies. Fiber-optic probes significantly reduce the effort required for sampling materials while increasing the
repeatability of some measurements. Fiber probes consist of a set, or bundle,
of fibers to deliver illumination from a lamp source to the sample and one or
more fibers to collect and transmit the light reflected from the sample. Transflectance probes, used for NIR monitoring of liquid operations, such as reaction
monitoring or fermentation, have a nonabsorbing reflector positioned normal to
the tip of the probe, separated by a fixed pathlength. Ball lenses or other collimating optics can be used to increase the solid angle of light captured by the fiber
probe. For example, if only the bare end of a single fiber is used to collect
reflected light, the effective sampling volume will be very small (the acceptance
angle of bare fibers is approximately 40 558). Some fiber-probe manufacturers
402
Cogdill and Drennen
have claimed that effective sample volume can be increased and that the detection of residue on a process window can be avoided, by using lenses that set
the focal point of analysis at a depth further beyond the surface of the
process window (unpublished communications). Such claims have yet to be substantiated in the literature, however, and seem to be in conflict with the accepted
concepts of NIR absorption.
Some probes utilize a single collection fiber, whereas others use a bundle of
multiple fibers. Besides their reduced cost, it has been supposed that single fiber
probes are advantageous because they have more uniform delivery of light to the
diffraction grating (because it is difficult to align the beams emanating from multiple fibers). Whether or not this is true is debatable; on the other hand, singlefiber probes will have lower efficiency and will be more likely to experience
complete failure during operation if the collection fiber is damaged. Besides
the number and configuration of the optical fibers, the quality of fibers and connections are important in determining the performance of the system (203). The
NIR process spectrometers are generally paired with low-OH optical fibers,
which have better transmittance characteristics than telecommunication-grade
fibers in the NIR region (Fig. 40). The attenuation of transmitted light through
optical fibers is mainly related to absorption by the glass and losses to connections and transmission through fiber walls. Thus, the spectral range of fibercoupled process NIR spectrometers is typically limited to less than 2200 nm.
The NIR fiber-optic probes are typically applied in reflectance or interactance modes (or some combination of the two). For example, NIR fiber probes
are often employed using optical interactance for raw material qualification
and identification (204), whereby the probe tip is pressed against the wall of a
container (e.g. polyethylene bags) (177). In the interactance mode, the light collected by the detector has either been reflected from beneath the surface of the
Figure 40 Typical spectral transmittance of standard and low-OH optical fibers in the NIR
region. Abbreviation: NIR, near-infrared. Source: Courtesy of Romack Fiber Optics Inc.
403
sample or has been transmitted laterally from the point of incidence of the source
illumination. Fiber probes can also be inserted directly into a dry powder bed,
though care must be taken to avoid fouling of the probe tip. Many of the early
studies using NIR for blend uniformity testing were based on the premise of
replacing a sample thief probe with a spectroscopic fiber probe (205 207).
When used for monitoring chaotic operations, such as particle coating or granulation, fiber probes can be inserted directly into the material stream to gather
spectra in the reflectance mode (208); again, probe fouling and contamination
is a concern. Furthermore, it can be difficult to deliver sufficient illumination
for reliable reflectance measurements unless the probe is in close interaction
with the sample (i.e., remote sensing measurements are limited with fiber
probes).
Remote Sensing
The capability for noncontact, or remote-sensing, sampling is an advantage
unique mainly to NIR and Raman spectroscopy. For our purposes (in contrast
to military and agricultural definitions), remote sensing refers to methods of
spectroscopic sampling that do not require any contact or interaction between
sample materials and sensing components. Remote sensing is more convenient
and is less disruptive to the sample material because there is no window or
probe interaction. Because remote sensing spectrometers are typically positioned
at some distance away from in-process sample materials, ambient light contamination and sample misalignment can add significant noise factors to the spectral
data.
Remote sensing NIR measurement can be accomplished in both DR and
transmittance modes. As mentioned earlier, transmittance measurements of
pharmaceutical solids are becoming relatively less common because of difficulties in sample alignment and limitations on the spectral range. Remote NIR transmittance might be an ideal means for sampling certain manufacturing operations
where the materials are uniformly thin and presented in a manner amenable to
transmittance, such as thin-film or web processing of diffusion membranes and
transdermal dosage forms or granulation via roller compaction of powdered
solids. Remote NIR transmittance has been used for rapid analysis of solid
dosage tablets by using a two-piece conveyor belt with a channel down the
center to hold tablets and allow transmitted NIR radiation to pass. The NIR
transmittance analysis of solid dosage forms was initially suspected to be more
robust, because, as was theorized, a greater fraction of the tablet mass is
sampled. Recent research comparing transmittance and reflectance for in-process
tablet release testing via NIR spectroscopy demonstrated, however, that the effect
of sample alignment on tablet spectra is more pronounced for transmittance
measurements (30).
Remote NIR reflectance is a practical means for in-process spectral data
acquisition. Depending on the geometry of the illumination and detector optics
(relative to the sample stream), the effect of sample alignment on NIR spectra
404
Cogdill and Drennen
will be limited and easily mitigated by spectral preprocessing methods. For

example, scatter correction and spectral derivatives have been found to be
useful for mitigating sample alignment effects during NIR reflectance analysis
of tablets for real-time release (RTR) testing (30,57). Remote reflectance
sampling has been found to be especially advantageous for measuring the inprocess quality of wet or sticky products during wet granulation
(20,21,191,193,194,200) and pan coating (178). By positioning the NIR sampling
port in an appropriate location of the processing chamber, spectra can be acquired
real-time with no concern for sample window fouling. One disadvantage,
however, is that the access to all or part of the NIR instrumentation for maintenance might be restricted if it is positioned inside a containment barrier. For
certain operations where the in-process material is immobile, such as tray
drying, remote reflectance analysis can be carried out through a process window,
with no contact between the sample material and the optical window (209).
Remote sensing NIR reflectance spectrometers have been positioned above
conveyor belts for continuous moisture and quality measurement in agricultural
and other industries for decades. Even so, continuous conveying operations are
relatively scarce in pharmaceutical solid dosage production. As pharmaceutical
manufacturing begins to transit from exclusive batch production to a combination
of batch and continuous operations, however, remote sensing of conveyed
materials will be critical for control of operations.
Analysis of Finished Products
The analysis of intact, finished solid dosage forms was, at one time, considered to
be a very important goal for NIR analysis of pharmaceutical solids. The PAT
initiative has caused some debate regarding the value of developing methods
for finished dosage form analysis. The argument is that, rather than simply
using NIR (and other PAT methods) as a new tool for inspection, process analytics should be deployed further upstream in the processing chain as a tool for
process control. Although this is a valid point, the importance of rapid finished
product analysis in accelerating product release should not be diminished. Furthermore, in order to develop and verify the many interacting transfer functions
that will ultimately be required for feed-forward control of processes, real-time,
comprehensive measurement of product quality must be in place as feedback.
Thus, RTR methods for testing finished dosage form quality parameters, such
as identification, potency, content uniformity, and hardness, continue to be
explored (30,57,85).
The NIR spectroscopy is also demonstrated as a tool for predicting dosage
form performance (e.g., bioavailability) in terms of dissolution and disintegration time (55,210 212). Although this is an intriguing and worthwhile
concept, there has been too little discussion regarding the specificity of the dissolution prediction. In general, published examples of NIR calibrations for the
prediction of dissolution or disintegration are developed by varying specific
405
formulation factors, such as hardness, composition, or raw material quality. In

other words, although it is obvious that the researchers have demonstrated an
ability to correlate dosage form performance to composition, it has not been
rigorously demonstrated that the NIR signal reveals information related to dissolution not related directly to (more tangible) composition factors. Ultimately,
as there is currently some debate regarding the overall value of dissolution
release testing for certain dosage forms, it remains to be seen if NIR spectroscopy (or any other nondestructive test) will be uniquely useful in predicting
dosage form performance in truly relevant terms (e.g., bioavailability and
permeability).
Beyond the development lab and manufacturing floor, the combination of
chemometrics and NIR spectroscopy is increasingly being considered a hightech tool for postproduction analytical tasks. Thousands of solid dosage
forms are gathered from the marketplace for analysis every year, either
through product recall, regulatory investigation, or surveillance. The NIR spectroscopy and imaging can be valuable tools for nondestructively assessing the
quality of individual dosage forms before more complicated or destructive
tests are performed. One of the earliest applications of NIR spectroscopy of
intact dosage forms, published in 1987, was to detect adulteration and tampering of solid gelatin capsules. Lodder et al. (213), showed that NIR spectroscopy
and bootstrapping pattern recognition could detect various adulterants with concentrations down to 0.4% of capsule weight. In a later example, Bauer et al.
(214), utilized NIR spectroscopy of raw materials to investigate dissolution failures for erythromycin dihydrate tablet formulation. The spectral data indicated
that a dehydrated dihydrate of erythromycin is produced during formulation,
which gradually binds with Mg(OH)2, delaying the process of dissolution. Furthermore, the researchers were able to successfully predict that humidifying the
tablets would reverse the unfavorable binding and therefore increase the dissolution rate.
Finally, as the economic value of the pharmaceutical drug market continues
to expand, and as more consumers purchase their personal drug supply through
international trade via the internet, the counterfeit drug trade will increasingly
grow in sophistication. In response, NIR spectroscopy is currently being explored
as a high-speed method of validating drug product identification and integrity
throughout the supply chain, whether as part of an automated dispensing
system in a high-tech pharmacy or carried as a mobile investigation tool by
regulators. Indeed, the future of NIR technology in pharmaceutical science is
bright.
ACKNOWLEDGMENTS
The authors would like to thank Hua Ma, Zhenqi (Pete) Shiz, and Steve Short
for their efforts in locating and organizing current references and Ms. Ryanne
Forcht for her skillful scientific and grammatical editing.
406
Cogdill and Drennen
REFERENCES
1. Herschel SWF. Phil Trans Royal Soc 1800; 90:225.
2. Burns DA, Ciurczak EW, eds. Handbook of Near-Infrared Analysis. 1st ed.
New York, NY: Marcel Dekker, Inc., 1992. Practical Spectroscopy; No. 13.
3. Burns DA, Ciurczak EW, eds. Handbook of Near-Infrared Analysis. 2nd ed.
New York, NY: Marcel Dekker, Inc., 2001. Practical Spectroscopy; No. 27.
4. Williams P, Norris K, eds. Near-Infrared Technology in the Agricultural and Food
Industries. 2nd ed. St. Paul, MN: American Association of Cereal Chemists, Inc.,
2001.
5. Williams PC, Norris KH. Near-Infrared Technology in the Agricultural and Food
Industries. 1st ed. St. Paul: American Assoc. Cereal Chemists, 1987.
6. Ciurczak EW, Drennen JK. Pharmaceutical and Medical Applications of NearInfrared Spectroscopy. 1st ed. New York, NY: Marcel Dekker, 2002.
7. Hindle PH. Historical development. In: Burns DA, Ciurczak EW, eds. Handbook of
Near-Infrared Analysis. 2nd ed. New York, NY: Marcel Dekker, Inc., 2001:1 6.
8. Norris KH, Williams PC. Cereal Chem 1984; 61(2):158 165.
9. Kirsch JD, Drennen JK. Appl Spectrosco Rev 1995; 30(3):139 174.
10. Workman JJJ. Appl Spectrosco Rev 1996; 31(3):251 320.
11. Ciurczak EW. Appl Spectrosco Rev 1987; 23(1&2):147 163.
12. Journal of Near Infrared Spectroscopy.
13. Miller CE. Chemical principles of near-infrared technology. In: Williams P,
Norris K, eds. Near-Infrared Technology in the Agricultural and Food Industries.
2nd ed. St. Paul, MN: American Association of Cereal Chemists, Inc., 2001:19 38.
14. Murray I, Williams PC. Chemical principles of near-infrared technology. In:
Williams PC, Norris KH, eds. Near-Infrared Technology in the Agricultural and
Food Industries. 1st ed. St. Paul: American Assoc. Cereal Chemists, 1987:17 34.
15. Ingle JDJ, Crouch SR. Spectrochemical Analysis. Upper Saddle River, NJ: Prentice
Hall, 1988.
16. Colthup NG, Daly LH, Wiberly SE. Introduction to Infrared and Raman Spectroscopy. 3rd ed. New York: Academic Press, 1990.
17. Hirschfeld T, Zeev-Hed A. The Atlas of Near-Infrared Spectra. Philadelphia, PA:
Sadtler Research Laboratories, 1981.
18. Hollas JM. Modern Spectroscopy. Chichester, UK: John Wiley & Sons, 1987.
19. Bonanno AS, Olinger JM, Griffiths PR. The origin of band positions and widths in
near-infrared spectroscopy. In: Hildrum KI, Isaksson T, Ns T, Tandberg A, eds.
Near-Infrared Spectroscopy, Bridging the Gap Between Data Analysis and NIR
Applications. New York, NY: Ellis-Horwood, 1992:19 28.
20. Jorgensen A, Rantanen J, Karjalainen M, Khriachtchev L, Rasanen E, Yliruusi J.
Pharm Res 2002; 19(9):1285 1291.
21. Rasanen E, Rantanen J, Jorgensen A, Karjalainen M, Paakkari T, Yliruusi J. J Pharm
Sci 2001; 90(3):389 396.
22. Hogan SE, Buckton G. Int J Pharm 2001; 227:57 69.
23. Zhou GX, Ge Z, Dorwart J, et al. J Pharm Sci 2003; 92(5):1058 1065.
24. Buckton G, Yonemochi E, Hammond J, Moffat A. Int J Pharm 1998; 168:231 241.
25. Rasanen E, Rantanen J, Mannermaa J-P, Yliruusi J, Vuorela H. J Pharm Sci 2003;
92(10):2074 2081.
26. Maeda H, Ozaki Y. J Near IR Spectrosco 1995; 3:191 201.
407
27. Haaland DM, Melgaard DK. Appl Spectrosco 2000; 54(9):1303 1312.
28. Birth GS, Hecht HG. The Physics of near-infrared reflectance. In: Williams PC,
Norris KH, eds. Near-Infrared Technology in the Agricultural and Food Industries.
1st ed. St. Paul: American Association Cereal Chemists, 1987:1 15.
29. Dahm DJ, Dahm KD. The Physics of near-infrared scattering. In: Williams P,
Norris K, eds. Near-Infrared Technology in the Agricultural and Food
Industries. 2nd ed. St. Paul, MN: American Association of Cereal Chemists, Inc.,
2001:1 18.
30. Cogdill RP, Anderson CA, Delgado-Lopez M, et al. AAPS Pharm Sci Tech 2005;
In Press.
31. Ciurczak EW. Principles of near-infrared spectroscopy. In: Burns DA, Ciurczak EW,
eds. Handbook of Near-Infrared Analysis. 2nd ed. New York, NY: Marcel Dekker,
Inc., 2001:7 18.
32. Ciurczak EW. Principles of near-infrared spectroscopy. In: Burns DA, Ciurczak EW,
eds. Handbook of Near-Infrared Analysis. New York, NY: Marcel Dekker, Inc.,
1992:7 12.
33. Olinger JM, Griffiths PR. Theory of diffuse reflectance in the NIR region. In: Burns
DA, Ciurczak EW, eds. Handbook of Near-Infrared Analysis. New York, NY:
Marcel Dekker, Inc., 1992:13 35.
34. Olinger JM, Griffiths PR, Burger T. Theory of diffuse reflection in the NIR region. In:
Burns DA, Ciurczak EW, eds. Handbook of Near-Infrared Analysis. New York,
NY: Marcel Dekker, Inc., 2001:19 51.
35. Kubelka P, Munk F. Z Tech Phys 1931; 12:593 601.
36. Chandrasekhar S. Radiative Transfer. Oxford: Clarendon Press (reprinted by Dover
Publications, 1960), 1950.
37. Dahm DJ, Dahm KD. Appl Spectrosco 1999; 53:647 654.
38. Pasikatan MC, Steele JL, Spillman CK, Haque E. J Near IR Spectrosco 2001; 9:
153 164.
39. Bull CR. J Mod Optics 1990; 37(12):1955 1964.
40. Bull CR. Analyst 1991; 116:781 786.
41. Ciurczak EW. Spectroscopy 1986; 1(7):36.
42. Rantanen J, Yliruusi J. Pharm Pharmacol Commun 1998; 4:73 75.
43. Frake P, Luscombe CN, Rudd DR, Waterhouse J, Jayasooriya A. Anal Commun
1998; 35:133 134.
44. Otsuka M. Powder Technology 2004; 141:244 250.
45. ONeil AJ, Jee RD, Moffat AC. Analyst 1998; 123:2297 2302.
46. ONeil AJ, Jee RD, Moffat AC. Analyst 1999; 124:33 36.
47. Frake P, Gill I, Luscombe CN, Rudd DR, Waterhouse J, Jayasooriya UA. Analyst
1998; 123:2043 2046.
48. Higgins JP, Arrivo SM, Thurau G, et al. Anal Chem 2003; 75:1777 1785.
49. Drennen JK. A Noise in Pharmaceutical Analysis [Ph.D.]: College of Pharmacy,
University of Kentucky, 1991.
50. Ciurczak EW, Drennen JK. Near-infrared spectroscopy in pharmaceutical applications. In: Burns DA, Ciurczak EW, eds. Handbook of Near-Infrared Analysis.
2nd ed. New York, NY: Marcel Dekker, Inc., 2002:609 632.
51. Ciurczak EW, Drennen JK. Near-infrared spectroscopy in pharmaceutical applications. In: Burns DA, Ciurczak EW, eds. Handbook of Near-Infrared Analysis.
1st ed. New York, NY: Marcel Dekker, Inc., 1992.
408
Cogdill and Drennen
52. Drennen JK, Lodder RA. In: Patonay G, ed. Advances in Near-Infrared Measurements. Greenwich, CT: JAI Press, 1993:93 112.
53. Kirsch JD, Drennen JK. J Pharm Biomed Anal 1999; 19:351 362.
54. Kirsch JD, Drennen JK. Paper APQ 1177. Paper presented at: AAPS National
Meeting, 1996; Seattle, WA.
55. Kirsch JD, Drennen JK. J Pharm Biomed Anal 1995; 13:1273 1281.
56. Morisseau KM, Rhodes CT. Pharm Res 1997; 14(1):108 111.
57. Cogdill RP, Anderson CA, Delgado-Lopez M, et al. AAPS Pharm Sci Tech 2005; In
Press.
58. Chen Y, Thosar SS, Forbess RA, Kemper MS, Rubinovitz RL, Shukla AJ. Drug Dev
Ind Pharm 2001; 27(7):623 631.
59. Guo J-H, Skinner GW, Harcum WW, Malone JP, Weyer LG. Drug Dev Ind Pharm
1999; 25(12):1267 1270.
60. Ebube NK, Thosar SS, Roberts R, et al. Pharm Dev Technol 1999; 4(1):19 26.
61. Donoso M, Kildsig DO, Ghaly ES. Pharm Dev Technol 2003; 8(4):357 366.
62. Otsuka M, Mouri Y, Matsuda Y. AAPS Pharm Sci Tech 2003; 4(3):1 7.
63. Cho JH, Gemperline PJ, Aldridge PK, Sekulic SS. Anal Chim Acta 1997; 348:
303 310.
64. Berntsson O, Burger T, Folestad S, Danielsson L-G, Kuhn J, Fricke J. Anal Chem
1999; 71(3):617 623.
65. Berntsson O, Danielsson L-G, Folestad S. Anal Chim Acta 1998; 364:243 251.
66. FDAs Summary of Judge Wolins Interpretation of GMP Issues Contained in the
Courts Ruling in USA vs. Barr Laboratories. GMP Institute, ISPE. Available at:
www.gmp1st.com/barrsum.htm.
67. Clarke FC, Hammond SV, Jee RD, Moffat AC. Appl Spectrosco 2002; 56(11):1475
1483.
68. Iyer M, Morris HR, Drennen JKI. J Near IR Spectrosco 2002; 10:233 245.
69. FDA. PATA Framework for Innovative Manufacturing and Quality Assurance,
2004.
70. Ferraro JR, Basile LJ. Fourier Transform Infrared Spectroscopy: Applications to
Chemical Systems. Vol. 1 3. New York: Academic, 1978.
71. Griffiths PR, de Haseth JA. Fourier Transform Infrared Spectrometry. New York:
Wiley-Interscience, 1986.
72. Wetzel DL. Contemporary near-infrared instrumentation. In: Williams PC, Norris
KH, eds. Near-Infrared Technology in the Agricultural and Food Industries. 2nd
ed. St. Paul, MN: AACC, 2001:129 144.
73. McClure WF. Near-infrared instrumentation. In: Williams PC, Norris KH, eds. NearInfrared Technology in the Agricultural and Food Industries. 2nd ed. St. Paul, MN:
AACC, 2001:109 128.
74. Ciurczak EW. Spectroscopy 2005; 20(2):68 75.
75. www.Buchi.com. Buchi FT-NIR NIRFlex Technology.
76. Funk DB. J Near IR Spectrosco 1996; 4:101 106.
77. www.Foss.dk. XDS Series NIR Analyser.
78. Zeaiter M, Roger JM, Bellon-Maurel V, Rutledge DN. Trends Anal Chem 2004;
23(2):157 170.
79. Martens H, Ns T. Multivariate calibration by data compression. In: Williams PC, Norris
KH, eds. Near-IR Technol Agr Food Ind. 2nd ed. St. Paul, MN: AACC, 2001:59100.
80. Chang IC. Optical Eng 1981; 20(6):824 829.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
92.
93.
94.
95.
96.
97.
98.
99.
100.
101.
102.
103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
114.
115.
116.
117.
118.
409
Tran CD. Anal Letters 2005; 38(5):735752.

Tran CD. J Near IR Spectrosco 2000; 8:87 99.
Treado PJ, Levin IW, Lewis EN. Appl Spectrosco 1994; 48(5):607615.
Brimrose L-M, personal communication. Baltimore, MD: Brimrose Corp.
Cogdill RP, Anderson CA, Drennen JK. AAPS Pharm Sci Tech 2005; 6:E284 E297.
Milano MJ, Kim K-Y. Anal Chem 1977; 49(4):555 559.
Yates DA, Kuwana T. Anal Chem 1976; 48(3):510 514.
www.ControlDevelopment.com. CDI, Inc. Diode Array Technology.
www.Axsun.com. Axsun Microspectrometer Technology.
Schneller W, Noto J, Kerr R, Doe R. Proc SPIE 1998; 3355:877 883.
Taday PF, Bradley IV, Arnone DD, Pepper M. J Pharm Sci 2002; 92(4):831 838.
Clarke FC. Vib Spectrosco 2004; 34:25 35.
Miller PJ. Paper No. 95 3216: Liquid Crystal Tunable Filters (LCTF) for Multispectral Imaging. St. Joseph, MI: ASAE, 1995.
Miller PJ, Hoyt CC. Proc SPIE 1995; 2345:354 365.
www.SpectralDimensions.com. Spectral Dimensions Imaging Spectroscopy Instruments and Research.
United States Pharmacopeial Convention I. Near-infrared spectrophotometry.
USP-NF, Second Supplement. Rockville, MD USA: United States Pharmacopeial
Convention, Inc., 2003:3337 3344.
Brown SD, Sum ST, Despagne F, Lavine BK. Anal Chem 1996; 68(12):21 61.
Martens H, Ns T. Multivariate Calibration. New York, NY: John Wiley and Sons, 1989.
Massart DL, Vandeginste BGM, Buydens LMC, de Jong S, Lewi PJ, SmeyersVerbeke J. Handbooks of Chemometrics and Qualimetrics. Vol. Parts A and
B. New York: Elsevier, 1998.
Mobley PR, Kowalski BR, Workman JJ, Jr, Bro R. Appl Spectrosco Rev 1996;
31(4):347 368.
Workman JJ, Jr, Mobley PR, Kowalski BR, Bro R. Appl Spectrosco Rev 1996;
31(1 & 2):73 124.
Chemometrics and Intelligent Laboratory Systems: Elsevier.
Journal of Chemometrics: Wiley-Interscience.
Brown PJ. J Roy Stat Soc B 1982; 44:287 308.
Ritchie GE, Mark H, Ciurczak EW. AAPS Pharm Sci Tech 2003; 4(2):Article 24.
Draper NR, Smith H. Applied Regression Analysis. 2nd ed. New York: John Wiley
& Sons, Inc., 1981.
Draper NR, Smith H. Applied Regression Analysis. 3rd ed. New York: John Wiley &
Sons, Inc., 1998.
Barnes RJ, Dhanoa MS, Lister SJ. Appl Spectrosco 1989; 43(5): 772 777.
Savitzky A, Golay MJE. Anal Chem 1964; 36(8):1627 1639.
Teppola P, Minkkinen P. J Chemometrics 2000; 14:383 399.
Fearn T. Chemometrics Intell Lab Sys 2000; 50:47 52.
Goicoechea HC, Olivieri AC. Chemometrics Intell Lab Sys 2001; 56:73 81.
Wold S, Antti H, Lindgren F, Ohman J. Chemometrics Intell Lab Sys 1998; 44:175185.
Ns T. Technometrics 1985; 27:301 311.
Ns T. Biometrical J 1985; 27:265 275.
Ns T. Biomet J 1986; 28:99 107.
Skrede G, Ns T, Martens M. J Food Sci 1983; 48:1745 1749.
Spjtvoll E, Martens H, Volden R. Technometrics 1982; 24(3):173 180.
410
Cogdill and Drennen
119.
120.
121.
122.
Drennen JK, Lodder RA. Spectroscopy 1991; 6(8):34 39.

Lodder RA, Hieftje GM. Appl Spectrosco 1988; 42(8):1351 1365.
Lodder RA, Hieftje GM. Appl Spectrosco 1988; 42(8):1500 1512.
Johnson RA, Wichern DW. Applied Multivariate Statistical Analysis. 4th ed. Upper
Saddle River, JN: Prentice Hall, 1998.
Hotelling H. J Educ Psychol 1933; 24:417 441, 489 520.
Anderson CA, Cogdill RP, Drennen JKI. NIR News 2005; 16(3):7 8.
Westerhuis JA, Gurden SP, Smilde AK. Chemometrics Intell Lab Sys 2000; 51:
95 114.
Box GEP, Jenkins GM, Reinsel GC. Time Series Anal. Englewood Cliffs, N.J., USA:
Prentice Hall, 1994.
Gallagher NB, Wise BM, Butler SW, White D, Barna GG. Development and benchmarking of multivariate statistical process control tools for a semiconductor etch
process: improving robustness through model updating. Paper presented at: IFAC
ADCHEM 97, 1997; Banff, Canada.
Jackson JE, Mudholkar GS. Technometrics 1979; 21(3):341 349.
Gemperline PJ, Boyer N. Anal Chem 1995; 67:160 166.
Gemperline PJ, Webber LD, Cox FO. Anal Chem 1989; 61(2):138 144.
Shah NK, Gemperline PJ. Anal Chem 1990; 62:465 470.
Wold S. Pattern Recog 1976; 8:127 139.
de Juan A, Tauler R. Anal Chim Acta 2003; 500:195 210.
Garrido M, Larrechi MS, Rius FX, Tauler R. Chemometrics Intell Lab Sys 2005;
76:111 120.
de Juan A, Tauler R, Dyson R, Marcolli C, Rault M, Maeder M. Trends Analy Chem
2004; 23(1):70 79.
Ma B, Gemperline PJ, Cash E, Bosserman M, Comas E. J Chemometrics 2003;
17:470 479.
Krutchkoff RG. Technometrics 1967; 9(3):425 439.
Haaland DM, Easterling RG. Appl Spectrosco 1980; 34(5):539 548.
Haaland DM, Easterling RG. Appl Spectrosco 1982; 36(6):665 673.
Haaland DM, Easterling RG, Vopicka DA. Appl Spectrosco 1985; 39(1):73 84.
Haaland DM, Thomas EV. Anal Chem 1988; 60:1193 1202.
Helland IS. Commun Stat B Simulation Comput 1988; 17(2):581 607.
Helland IS. Scandinavian J Stat 1990; 17:97 114.
De Jong S. Chemometrics Intell Lab Sys 1993; 18:251 263.
Wold S, Ruhe A, Wold H, Dunn WJ, III. SIAM J Sci Stat Comput 1984; 5(3):735
743.
Geladi P, Kowalski BR. Anal Chim Acta 1986; 185:1 17.
Lorber A. Anal Chem 1986; 58:1167 1172.
Lorber A, Faber K, Kowalski BR. Anal Chem 1997; 69:1620 1626.
Ferre J, Brown SD, Rius FX. J Chemometrics 2001; 15:537 553.
Skibsted ETS, Boelens HFM, Westerhuis JA, Witte DT, Smilde AK. Appl Spectrosco 2004; 58(3):264 271.
Goicoechea HC, Olivieri AC. Anal Chim Acta 2002; 453:289 300.
Cogdill RP, Anderson CA. J Near IR Spectrosco 2005; 13:119 132.
Roussel S, Hardy CL, Hurburgh CRJ, Rippke GR. Appl Spectrosco 2001;
55(10):1425 1430.
Agatonovic-Kustrin S, Beresford R. J Pharm Biomed Anal 2000; 22:717 727.
123.
124.
125.
126.
127.
128.
129.
130.
131.
132.
133.
134.
135.
136.
137.
138.
139.
140.
141.
142.
143.
144.
145.
146.
147.
148.
149.
150.
151.
152.
153.
154.
411
155. Borggaard C. Neural networks in near-infrared spectroscopy. In: Williams PC,

Norris KH, eds. Near-IR Technology in the Agricultural and Food Industries. 2nd
ed. St. Paul, MN: AACC, 2001:101 108.
156. Despagne F, Massart DL. The Analyst 1998; 123:157R 178R.
157. Hussain AS, Zuanqiang Y, Johnson RD. Pharm Res 1991; 8(10):1248 1252.
158. Nerella NG, Drennen JKI. Appl Spectrosco 1996; 50(2):285 291.
159. Rumelhart DE, Hinton GE, Williams RJ. Learning internal representations by error
propagation. In: Rumelhart DE, ed. Parallel Distributed Processing. Cambridge, UK:
MIT Press, 1986:318 362.
160. Fiesler E, Beale R. Handbook of Neural Computation. Bristol and Oxford, UK: IOP
Publishing, Ltd. and Oxford University Press, Inc., 1997.
161. Cogdill RP, Dardenne P. J Near IR Spectrosco 2004; 12:93 100.
162. Vapnik V, Golowich S, Smola A. Support vector method for function approximation,
regression estimation, and signal processing. In: Mozer M, Jordan M, Petsche T, eds.
Advances in Neural Information Processing Systems. Cambridge, MA: MIT Press,
1997:281 287.
163. Walczak B, Massart DL. Anal Chim Acta 1996; 331:177 185.
164. AACC. Near-infrared methods: guidlines for model development and maintenance, AACC Method 3900. Approved Methods of the American Association of
Cereal Chemists. 10th ed. St. Paul, MN USA: AACC Press, 1999.
165. ICH. Harmonized Tripartite Guideline Q8: Pharmaceutical Development: International Conference on Harmonization, 2004.
166. Fearn T. J Near IR Spectrosco 2001; 9:229 244.
167. ICH. Harmonized Tripartite Guideline Q2A: Text on Validation of Analytical
Procedures: International Conference on Harmonization, 1994.
168. Reich G. Adv Drug Deliv Rev 2005; 57:1109 1143.
169. Brashear RL, Flanagan DR, Luner PE, Seyer JJ, Kemper MS. J Pharm Sci 1999;
88(12):1348 1353.
170. Lachenal G. Vib Spectrosco 1995; 9:93 100.
171. Siesler HW, Ozaki Y, Kawata S, Heise HM. Near-infrared spectroscopy: principles,
instruments, applications. Weinheim, Germany: Wiley-VCH GmbH, 2002:
213 246.
172. Zerbi G, ed. Modern Polymer Spectroscopy. Weinheim: Wiley-VCH GmbH, 1999.
173. Buice RGJ, Gold TB, Lodder RA, Digenis GA. Pharm Res 1995; 12(1):161163.
stling G. J Pharm Biomed Anal 1997; 15:895 900.
174. Berntsson O, Zackrisson G, O
175. Gold TB, Buice RGJ, Lodder RA, Digenis GA. Pharm Dev Technol 1998; 3(2):209
214.
176. Svensson O, Josefson M, Langkilde FW. Appl Spectrosco 1997; 51(12):1826 1835.
177. Gustafsson C, Nystrom C, Lennholm H, Bonferoni MC, Caramella CM. J Pharm Sci
2003; 92(3):460 470.
178. Perez-Ramos JD, Findlay WP, Peck GE, Morris KR. AAPS Pharm Sci Tech 2005;
6(1):E127 E136.
179. Friend DR. Pharm News 2002; 9:375 380.
180. Friend DR, Chrai SS, Kupperblatt G, et al. Accudep technology for oral modified
drug release. In: Rathbone M, ed. Modified-Release Drug Delivery Technology.
New York, NY: Marcel Dekker, Inc., 2002:89 100.
181. Blanco M, Romero MA. J Pharm Biomed Anal 2002; 30:467 472.
182. Gimet R, Luong AT. J Pharm Biomed Anal 1987; 5(3):205 211.
412
Cogdill and Drennen
183. Buchanan BR, Ciurczak EW, Grunke AQ, Honigs DE. Spectroscopy 1988;
3(5):54 56.
184. Miller CE, Honigs DE. Spectroscopy 1989; 4(3):44 48.
185. Luner PE, Majuru S, Seyer JJ, Kemper MS. Pharm Dev Technol 2000; 5(2):231
246.
186. Patel AD, Luner PE, Kemper MS. J Pharm Sci 2001; 90(3):360 370.
187. Seyer JJ, Luner PE, Kemper MS. J Pharm Sci 2000; 89(10):13051316.
188. Patel AD, Luner PE, Kemper MS. Int J Pharm 2000; 206:63 74.
189. Braga JWB, Poppi RJ. J Pharm Sci 2004; 93(8):2124 2134.
190. Otsuka M, Kato F, Matsuda Y, Ozaki Y. AAPS Pharm Sci Tech 2003; 4(2):1 12.
191. Davis TD, Peck GE, Stowell JG, Morris KR, Byrn SR. Pharm Res 2004; 21(5):860
866.
192. NIST/SEMATECH. e-Handbook of Statistical Methods. http://www.itl.nist.gov/
div898/handbook. Available at: http://www.itl.nist.gov/div898/handbook. Accessed
October 22, 2005.
193. Jrgensen A, Rantanen J, Karjalainen M, Khriachtchev L, Rasanen E, Yliruusi J.
Pharm Res 2002; 19(9):1285 1291.
194. Jrgensen AC, Luukkonen P, Rantanen J, Schfer T, Juppo AM, Yliruusi J. J Pharm
Sci 2004; 93(9):2232 2243.
195. Rantanen J, Antikainen O, Mannermaa J-P, Yliruusi J. Pharm Dev Technol 2000;
5(2):209 217.
196. Rantanen J, Jrgensen A, Rasanen E, et al. AAPS Pharm Sci Tech 2001; 2(4):21.
197. Rantanen J, Kansakoski M, Suhonen J, et al. AAPS Pharm Sci Tech 2000; 1(2):10.
198. Rantanen J, Lehtola S, Ramet P, Mannermaa J-P, Yliruusi J. Powder Technol 1998;
99:163 170.
199. Rantanen J, Rasanen E, Tenhunen J, Kansakoski M, Mannermaa J-P, Yliruusi
J. European J Pharm Biopharm 2000; 50:271 276.
200. Rantanen J, Wikstrom H, Turner R, Taylor LS. Anal Chem 2005; 77:556 563.
201. www.NDCInfrared.com. NDC Infrared Engineering.
202. www.JMCanty.com. J.M. Canty Process Technology.
203. www.romackfiberoptics.com. Romack Fiber Optics.
204. Kemper MS, Luchetta LM. J Near IR Spectrosco 2003; 11:155 174.
205. El-Hagrasy AS, Delgado-Lopez M, Drennen JKI. J Pharm Sci. 2006; 95:407 421.
206. El-Hagrasy AS, Drennen JKI. J Pharm Sci 2006; 95:422 434.
207. El-Hagrasy AS, Morris HR, DAmico F, Lodder RA, Drennen JKI. J Pharm Sci
2001; 90(9):298 1307.
208. Frake P, Greenhalgh D, Grierson SM, Hempenstall JM, Rudd DR. Int J Pharm 1997;
151:75 80.
209. Parris J, Airiau C, Escott R, Rydzak J, Crocombe R. Spectrosco 2005; 20(2):34 41.
210. Donoso M, Ghaly ES. Pharm Dev Technol 2005; 10:211 217.
211. Donoso M, Ghaly ES. Pharm Dev Technol 2004; 9(3): 247 263.
212. Freitas MP, Sabadin A, Silva LM, et al. J Pharm Biomed Anal 2005; 39:17 21.
213. Lodder RA, Selby M, Hieftje GM. Anal Chem 1987; 59:1921 1930.
214. Bauer JF, Dziki W, Quick JE. J Pharm Sci 1999; 88:1222 1227.
PART V.
NUCLEAR SPIN LEVEL SPECTROSCOPY
11
Solid-State Nuclear Magnetic
Resonance Spectrometry
Ales Medek
Pfizer Global R&D, Groton, Connecticut, U.S.A.
FUNDAMENTAL PRINCIPLES OF NMR

Introduction
Nuclear magnetic resonance (NMR) spectroscopy has many similarities with the
molecular spectroscopic techniques involving transitions between electron
energy levels. However, NMR stands apart in many details, which will be
briefly discussed in this introduction section. The obvious distinction is that
NMR involves transitions between nuclear energy levels. In the absence of external magnetic or electric fields, the energy levels of the ground state of an isolated
nucleus are degenerate. It is the external magnetic field that brings the distinction
between the nuclear energy levels. If we apply external magnetic field B0 on a
sample containing nuclei with nonzero spin angular momentum I (i.e., magnetically active spins), the 2I 1 nuclear energy levels will be
E gIz B0
(1)
in which the proportionality factor, referred to as gyromagnetic ratio, g is the

characteristic constant for a given isotope. Because the values of angular momentum are quantized with the magnetic quantum number m (IZ h m, where h is
the Plancks constant), so are the nuclear energy levels known as Zeeman
levels. Following the selection rule governing any magnetic dipole transitions
(Dm +1), the differences between the Zeeman energy levels are
DE h gB0 h v0
(2)
413
414
Medek
where v0 is the frequency (in units of radian per second) of the electromagnetic
radiation, corresponding to the difference between the spin energy levels,
known as Larmor frequency. Equation (2) also predicts linear dependence of
the energy level difference on the external magnetic field B0. For the currently
available magnetic fields, the energy difference falls into the radio frequency
(rf) portion of the electromagnetic radiation. When the system is at thermal
equilibrium, the population of the given energy level is given by the Boltzmann
distribution
Ei
Ni exp
kB T

(3)
where Ni is the number of nuclei occupying energy level Ei, kB is the Boltzmann
constant, and T is the absolute temperature. In contrast to the electronic transitions, at temperatures close to the room temperature, the product of kBT is
several orders of magnitude larger than the difference between the Zeeman
levels DE. Consequently, even random thermal fluctuations can very efficiently
induce the nuclear spin transitions, nearly equalizing the number of nuclei occupying each Zeeman energy level. Being directly proportional to the population
difference, the NMR signal strength is significantly weaker compared with other
atomic or molecular spectroscopies involving electron levels transitions. Out of
200,000 1H spins placed into the external magnetic field of 11.7 T (500 MHz 1H
frequency), approximately 100,001 spins will occupy the lower energy level
and 99,999 the higher one. Effectively, only 1/200,000 of the sample gives
rise to the observable NMR signal. When compared with other spectroscopies,
NMR is a relatively insensitive technique.
Another property of nuclear transitions stands apart from spectroscopies
involving electronic transitions. It is the relative isolation of the nuclear system
from its surroundings. This simplifies the NMR response of the system under
the study and, typically, one can understand and simulate the outcome of any
NMR experiment based on first principles considerations. When out of thermal
equilibrium, nuclear spins tend to spontaneously return to their equilibrium
states. This relaxation process is driven by random fluctuations of variables
affecting the nuclear transitions. In many cases, encountered especially in
solid-state NMR (SSNMR), the relaxation might be prohibitively long. It is not
uncommon for the return to equilibrium to take in the order of hours or longer,
prohibiting the practical observation of the signal. On the other hand, unlike
the electronic spectroscopies, the nuclear spins can form coherent states for
long periods of seconds or longer. Modern NMR multidimensional experiments
take excellent advantage of this phenomenon. The slow decay of coherent states
allows one to design rather complex pulse sequences manipulating the coherent
states of nuclear spins to obtain the targeted information about the system under
the study.
SSNMR Spectrometry
415
Nuclear Spin Interactions

To understand the applications of the NMR spectroscopy in both the solution and
solid phases, it is useful first to review the various interactions affecting the
nuclear spins. Nuclear spins can interact with each other or with internal or external electric and magnetic fields (1). The internal electric and magnetic fields arise
due to the effect of electrons surrounding the nuclei of interest in atoms or molecules. The possibility to selectively exploit the nuclear interactions makes NMR
an extremely powerful tool, finding its applications across many scientific
disciplines.
In quantum mechanics, all the interactions of a system of interest are represented by corresponding Hamiltonians. The state of the system is given by its
wave function and all the measurable quantities are represented by their operators. The most basic NMR interaction, the Zeeman interaction, was already
introduced in the previous section. Zeeman interaction causes the nuclear spins
to interact with the external magnetic field. This interaction is responsible for
observing the NMR phenomenon. Based on the Correspondence Principle, the
classical description from Equation (1) can be rewritten using the quantum mechanical Zeeman Hamiltonian HZ
HZ gI^Z B0
(4)
where I^Z is the spin angular momentum operator. The energy levels of the nuclei
are directly proportional to the applied external field B0. It turns out that the external magnetic field has to be extremely homogeneous to resolve the fine features
of NMR spectra. A great deal of attention is paid to this issue. Every modern
NMR spectrometer is equipped with a set of shimming coils correcting the
local imperfections of the field.
NMR would not be a very chemically useful technique if only Zeeman
interaction existed, because all spins from any compound would resonate at
the same frequency. The useful NMR properties arise from other nuclear interactions. All the interactions exist simultaneously and can be described by the
total Hamiltonian (2 5)
HT HZ HCS HJ HD HQ Hrf
(5)
where the subscripts CS, J, D, Q, and rf denote the chemical shift interaction,
indirect spin-spin (J-coupling), dipolar coupling, quadrupolar coupling, and rf
irradiation, respectively. It should be noted that these nuclear interactions do
not represent all the interactions that nature provides. They are, however, sufficient to facilitate the majority of discussions of NMR applications in this
review. In terms of magnitude, the Zeeman interaction is typically, but not
always, the strongest, followed by quadrupolar, dipolar, chemical shift, and
J-coupling. This is just a general order, which may change strongly depending
on the type and shape of molecules. For example, in many asymmetric molecules
with quadrupolar nuclei, the quadrupolar coupling can be larger than the Zeeman
416
Medek
interaction, especially when working in weaker magnetic fields. Such nuclei

may be difficult to observe by NMR but are a great subject to nuclear quadrupolar
resonance (NQR) studies, a spectroscopic technique, which does not require a
presence of an external magnetic field, because it deals only with the quadrupolar states. On the other hand, highly symmetrical molecules containing
quadrupolar nuclei can show vanishing quadrupolar contribution.
In general, in the laboratory frame, the Hamiltonian representing a particular interaction Hl can be characterized as a product of two spin operators with a
Cartesian tensor Alof rank two, described by the (3 3) matrix:
0
Alxx
Alxy
B
Hl Cl (I^x , I^y , I^z ) @ Alyx
Alyy
Alzx
Alzy
1
S^ lx
B C
Alyz C
A @ S^ l A
Alxz
1 0
Alzz
(6)
S^ lz
where Cl is a constant dependent on the interaction l, I (Ix, Iy, Iz) is the

nuclear spin angular momentum operator of spin I, and S (Sx, Sy, Sz) can be
either the nuclear spin operator of spin S or B0 in the special case of l CS
(the chemical shift interaction). From this notation, it is evident that the
nuclear interactions depend on properties that exhibit a dependence on a
particular orientation in the physical space and have to be described by
tensors rather than scalars (numbers). For example, the Alzx element of the
tensor A, describes the reaction of the system along the z direction when a
force represented by the operator Slx is applied in the x direction. In a
special frame of reference called the principal axis system (PAS), the symmetrical part of the second rank tensor takes a diagonal form, that is, the only nonzero
components lie along the diagonal:
0
Al11
B
Al (PAS) B
@ 0
0
0
Al0 (PAS)
Al22
Al33
1
0 0
C
1 0C
A
dl11
B
Al2 (PAS) B
@ 0
dlii Alii AlISO
C
l
l
0 C
A A0 (PAS) A2 (PAS)
AlISO B
@0
0
0
0
dl22
0
1
0
C
0 C
A
dl33
(7)
SSNMR Spectrometry
417
The orientation of the principal axis system of a tensor Al depends on the

local molecular symmetry and even for the same nucleus can, in general,
change depending on the type of interaction l. The PAS orientation for given
coupling l is fixed with respect to the molecular frame of the compound.
Because Tr(Al2 ) 0, only two parameters are necessary for a complete
description of Al2 :
dl dl33
hl
dl22 dl11
dl
(8)
where hl is usually referred to as the asymmetry parameter with values limited

to 0 hl 1. An arbitrary second rank tensor is then defined in its respective
PAS by its isotropic part Aliso, by its anisotropic contribution dl, and the asymmetry parameter hl.
The chemical shift interaction is a coupling of the nuclear spins I with the
external magnetic field B0 mediated through the electronic environment. Diamagnetic currents in the electron orbitals and partial unquenching of their paramagnetism generate local field Bind proportional to B0, Bind 2ACSB0. The chemical
shift Hamiltonian is defined as
^ ind gIA
^ CS B0
HCS gIB
(9)
By comparison of Equation (9) and Equation (6), the proportionality constant CCS
is simply equal to g. The values of the components of chemical shift tensor ACS
depend, in general, on the orientation and local symmetry of the molecule. The
remarkable property of the chemical shift tensor is that the sum of its diagonal
elements (known as its trace) does not vanish, thus giving rise to the isotropic
shift

1
diso Tr ACS = 0
3
(10)
This is why even the fast isotropic tumbling of molecules in solution does not
average out the chemical shift to zero. Different nuclei in the same molecule
experience different local fields. Based on the chemical shifts alone, many molecules can be uniquely identified.
At the atomic level, a given nucleus affects, and is also affected by, the
local dipolar fields of its neighbors. The indirect spin spin J-coupling can be
described by Hamiltonian in Equation (11).
HJ I^ AJ S^
(11)
It is called indirect, because it acts through the electronic environment, rather

than directly though space. Similar to the chemical shift, the trace of the
J-coupling tensor is not zero. Therefore we can observe and measure finite
J-coupling values even in solution. Moreover, for majority of practical
418
Medek
applications, the J-coupling tensor AJ can be approximated as being symmetrical with vanishing anisotropy and as such can be regarded as scalar (a number)
rather than tensor.
Nuclear spins can also interact directly through space. Such an interaction
is referred to as dipolar coupling (or dipole dipole coupling) and can be
described by the dipolar Hamiltonian:
"
(I^1 r)(I^2 r)
HD D I^1 I^2 3
r2
#
(12)
where r is the internuclear vector and D is the dipolar coupling constant in frequency units, defined as
D
m0 g1 g2 h
4p 2p r 3
(13)
m0 is the permeability of vacuum constant. The trace of the dipolar tensor

vanishes, which is why dipolar coupling has no direct effect on isotropic solution
NMR spectra apart from relaxation-induced processes.
Spin I 1/2 nuclei possess only magnetic dipole moments, because all the
higher magnetic and electric multipole moments vanish. Therefore, all their
^ However, apart from the magnetic couplings
Hamiltonians are linear in I.
encountered in spin I 1/2 isotopes, nuclei with I . 1/2 are distinguished by
a nonzero nuclear quadrupole moment Q, which gives origin to the interaction
with the surrounding electric field gradients (EFG). The quadrupolar coupling
between the quadrupole moment and the electric field gradient is described by
the quadrupolar Hamiltonian:
HQ
eQ
^ Q I^
IA
2I(2I 1)h
(14)
The symbol e designates the elementary charge. Similar to the dipolar coupling,
the trace of EFG tensor vanishes having no effect on solution NMR signals except
through the relaxation processes. The majority (72%) of all the magnetically
active nuclides in the Periodic Table possess half-integer spin higher than 1
(Fig. 1) (6). This clearly underlines the importance as to why we should study
these spins (7 9).
The chemical shift interaction, J, dipolar, and quadrupolar couplings are
called internal interactions, because they represent couplings between the
various fields present internally within a molecule. On the other hand, Zeeman
interaction and the rf irradiation are external interactions, because they represent
the interactions of the nuclear spins with the external magnetic fields.
In order to disturb the spin system from its thermodynamic equilibrium,
an oscillating rf field B1 has to be applied in the direction perpendicular to that
of B0. If the frequency of this oscillating field is v and its phase in the plane
SSNMR Spectrometry
419
Figure 1 Magnetically active isotopes in the Periodic Table. The majority (72%) of the magnetically active nuclides are half-integer quadrupolar nuclei. If one plans to use solid state NMR
for studying ceramics, semi-conductors, catalysts, glasses, superconductors or metal binding
processes it is necessary to deal with the spectroscopy of these spins. For elements with
several different nuclides, only those with the highest natural abundance are considered.
perpendicular to B0 is w, then the B1 vector can be fully described by:

0
1
cos w cos (vt)
B1 (t) 2B1 @ sin w cos (vt) A
0
(15)
The irradiation Hamiltonian representing the coupling between the oscillating rf

field and the particular components of the magnetic dipole is in the laboratory
frame given by:
Hrf v1 I^x cos(vt w) I^y sin(vt w)
(16)
where v1 gB1 is the nutational frequency directly proportional to the amplitude

of the irradiation field. This value, typically given in kHz, is used frequently in the
NMR literature to describe the strength of irradiation. The irradiation field is typically calibrated based on the experimentally observed duration of 908 pulse (t90),
which is related to v1 by v1 1/(4t90). In practice, its value depends on the available power delivered from the rf amplifier (v1 is directly proportional to the rf
voltage amplitude) and on the internal characteristics of the probe, such as coil
geometry. Of a great interest is the homogeneity of the B1 field throughout the
coil, strongly influencing performance of many pulse sequences.
420
Medek
SSNMR OF POWDER SOLIDS

Introduction
Most NMR applications have traditionally involved solution samples. In solution,
all the valuable anisotropic information is lost by the fast tumbling of the liquid.
Unlike in solution, the full extent of the anisotropic interactions is observed in the
solid state. In essence, there is more NMR information available on the system
when in the solid state. However, SSNMR possess much greater technical challenge to mine this information, which typically requires selective turning on
and off the various strong nuclear interactions discussed in the preceding
chapter. The great advantage of SSNMR is the concentration of the samples.
One does not have to rely on dissolution properties of compounds to introduce
them into the NMR coil.
The NMR spectra of stationary (not spinning) polycrystalline or disordered
solids are known as powder patterns. It is possible to predict their line shapes by
numerical simulations based on the a priori known nuclear coupling parameters.
Figure 2 and Figure 3 show the typical line shapes of chemical shift and
quadrupolar interactions. In simulating these spectra, only a single spin contribution was considered. Because more than one magnetically inequivalent
nucleus is typically present in a realistic system, the static spectra can quickly
become overcrowded and difficult or impossible to interpret. The available
Figure 2 Examples of static powder patterns for chemical shielding as a function of

asymmetry parameter hCS. The magnitude of chemical shielding dCS was arbitrarily set
to 143 kHz, and the isotropic shift diso 0.
SSNMR Spectrometry
421
Figure 3 Examples of static powder patterns for quadrupolar coupling to the first- and
second-order as a function of asymmetry parameter hQ for I 7/2. The quadrupolar coupling constant e 2qQ/h was arbitrarily set to 22 MHz, the Larmor frequency (important for
second-order spectra) used in these simulations was v0 47.7 MHz (59Co at 4.7 T), and
no chemical shift was considered.
signal is also spread over broad range of frequencies, which diminishes the
achieved sensitivity in terms of the signal-to-noise ratio (S/N ). Therefore, a technique improving the resolution of the SSNMR spectra of crystalline solids has
long been sought. The two key factors precluding widespread use of SSNMR
in the past were sensitivity and resolution. A boost in solid-state applications
appeared after a combination of the line-narrowing technique called magic
angle spinning (MAS) (10 12) and sensitivity-enhancing technique called
cross-polarization (CP) (13) was introduced. Cross-polarization magic angle
spinning (CPMAS) (14) brought about the necessary sensitivity enhancement
to make SSNMR spectroscopy of natural abundance samples practicable. Spinning around the magic angle, which is the root of second-order Legendre polynomial, x 54.748 (Fig. 4), removes completely the second rank anisotropies if
these are much smaller than the rotational frequency or may otherwise lead to a
spinning sideband manifold. Example of this behavior is shown in Figure 5 on
carbon spectra DL -alanine. At spinning speed of 1.0 kHz, all three alanine
carbons show significant spinning sidebands. At 2.5 kHz, no appreciable
422
Medek
Figure 4 The most conventional example of spinning about a single axis, the magic
angle spinning. The time average of kP2(cos b)l over integer multiples of rotor period
vanishes.
Figure 5 The 11.7 T carbon CPMAS spectra of DL -alanine at variable spinning speeds of
1.0, 2.5, 4.0, 5.5, and 7.0 kHz. The spinning sidebands are spaced in integer multiples of
the spinning speed (highlighted by the arrowhead lines). Abbreviation: CPMAS, crosspolarization magic angle spinning.
SSNMR Spectrometry
423
spinning sideband intensities are observed for the methyl group. The methine
carbon still shows signs of spinning sidebands, albeit with greatly reduced intensities. Even at the maximum spinning speed of 7.0 kHz (the maximum specified
spinning speed of standard wall 7.0 mm Bruker-Biospin rotors), the carboxylic
carbon signal is still spread between spinning sidebands. The spinning sidebands
in this example arise due to the modulation of the carbon chemical shift anisotropies by the MAS spinning. Clearly, the anisotropic contribution of methyl group
is smaller than that of methine, which is still smaller than that of carboxylic
carbon. In general, the magnitude of the chemical shift anisotropy for different
carbons approximately follows the rank order of CH3,CH2,CH,aromaticcarbonyl, which happens to coincide with the typical rank order of their isotropic
shifts. The contribution of methyl groups to spinning sidebands is further reduced
due the fact that they exhibit fast wheel-like rotation around the C C axis, even
in the solid state at temperatures close to room temperature.
Spins higher than 12 show additional quadrupolar coupling contribution to
their spectral line shapes. Two cases should be distinguished. For half-integer
spins, the central transition (21/2 $ 1/2) is void of quadrupolar coupling
to the first order. For compounds with high molecular symmetry (small nuclear
quadrupole moment), MAS is likely to remove most of the quadrupolar
anisotropies. To get high resolution for compounds with moderately strong quadrupolar coupling, such techniques as double rotation (15,16), dynamic angle
spinning (17), or multiple quantum MAS (18,19) have to be applied. There is
no central transition for nuclei with integer spins. The full effect of the quadrupolar coupling shows in their spectra. There are at least two pharmaceutically
relevant spin 1 nuclei: 2H and 14N. Deuterium quadrupole moment tends to be
small and spinning around magic angle is typically sufficient to average out
the anisotropies, leaving only the extensive spinning sideband manifold. Due
to the low 2H natural abundance, isotope enrichment is required for any deuterium studies. The easiest way to introduce 2H labels is through deuterium
exchange of the exchangeable protons. Unlike 2H, 14N is a nitrogen isotope
with almost 100% natural abundance. However, its widespread applications
have been largely hindered by the presence of rather sizeable nuclear quadrupole
moment. This is why NQR technique has been successfully applied to detect
nitrogen-containing explosives. Several NMR approaches to get better 14N
resolution have been tried, but no simple method has been found yet (20 27).
Proton SSNMR
In contrast to 1H spectroscopy in solutions, proton NMR in the solid state has
faced significant technical challenges, mainly due to the extensive line broadening originating from strong, through-space dipole dipole coupling. This interaction is, to a very good approximation, completely averaged out in isotropic
solution as molecules tumble fast and completely randomly on the NMR timescale. Recently, it was shown that the dipolar coupling may not get completely
424
Medek
averaged out over long distances, if the interacting spins do not share the same
diffusion sphere (28 31). This contribution is, however, very small and for
most applications of practical interest can be disregarded. The only remaining
effect that the dipolar coupling has on isotropic solutions is through the incoherent relaxation. The distance measurements in solution rely on dipolar relaxationinduced nuclear Overhauser effect (NOE) (32). NOE-derived structural
constraints are the basis for protein structural determination by NMR. Due to
the type and magnitude of motions involved, NOE effect plays much smaller
role in the solid state.
In the crystalline solids, molecules are locked in their crystal lattices.
Equation (12) predicts the dipolar coupling to be directly proportional to the
product of the gyromagnetic ratios of the two involved nuclei (two protons in
this case) and inversely proportional to the cube of the distance between the
nuclei (r 23). The gyromagnetic ratio of protons is among the biggest of all
nuclei. Moreover, the typical organic solids contain many protons in the close
vicinity of each other. Both of these factors translate into very strong proton
dipolar coupling. In the absence of extensive molecular motions, the dipole
dipole interaction is usually significantly larger than the chemical shielding
effect. When out of the equilibrium state, the through-space dipole dipole
interaction induces fast exchange of magnetization between the protons, effective
over many molecules. This magnetization exchange is called a spin diffusion
(2,4). The term of diffusion is very appropriate as it refers to the diffusion
of the NMR signal over the net of nuclei, albeit without any physical
movement of the molecules. The signal transfer due to spin diffusion can be
described by conventional equations for diffusive motion [Equation (17)],
where D is the diffusion coefficient, t is the time during which the diffusion is
effective, and kLl is the average distance traveled by the NMR signal.
kLl
p
6Dt
(17)
Thus, domain sizes of phase-separated solids may be estimated from the

rates of spin diffusion. Unless the dipolar coupling is averaged out by significant
molecular motions or by very fast MAS rotation, the spin diffusion is fast on the
NMR timescale. A typical proton static line shape of organic molecules is a single
peak, up to 50 kHz wide, precluding resolution of the chemical shifts (2 4)
(Fig. 6; bottom trace). As a result of the spin diffusion, the anisotropic proton
lineshape broadening is referred to as being homogeneous. In contrast to the
inhomogeneous broadening arising from, for example, the chemical shift or
quadrupolar coupling (Figs. 2 and 3), selective excitation of arbitrarily narrow
part of the proton spectrum leads to saturation (or disappearance) of the whole
spectrum. The spin diffusion also counteracts the MAS averaging effect (Fig. 6;
second trace from bottom). Because MAS narrows down the proton line shape
only marginally, other line-narrowing techniques have been sought. Combined
rotation and multiple-pulse spectroscopy (CRAMPS) (33 39) offers a partial
SSNMR Spectrometry
425
Figure 6 Solid-state proton spectra of L -alanine. Differences between proton spectra

acquired in static, fast magic angle spinning, and FSLG-CRAMPS modes. The top is a
projection of the indirectly detected domain of 2D constant time 1H 1H correlation experiment. Abbreviation: FSLG-CRAMPS, frequency-switched Lee-Goldburg experimentcombined rotation and multiple-phase spectroscopy. Source: Modified from Ref. 38.
solution to this problem (Fig. 6; top two traces). It combines conventional MAS
with rotation of proton magnetization around the magic angle in spin space. To
avoid interference between the MAS averaging and the rf pulses, successful
application of CRAMPS requires the duration of the complete train of pulses
(typically 24 or more) to be much shorter with compared one cycle of the
MAS rotation. This significantly restricts the applicable MAS rotation speeds.
Typically, spinning at speeds in access of 3 kHz is prohibitive. The CRAMPS
experiment also puts a large demand on the NMR hardware, especially on
NMR probes, because high-power rf pulses are applied in between each acquisition point. High homogeneity of the irradiation field B1 and careful set-up of
experimental variables are required to avoid distortions of the proton peaks.
Typical line widths afforded by the CRAMPS experiment are approximately
1 ppm, limiting the application of the CRAMPS experiment to compounds
with a small number of well-resolved protons. The more modern variants of
CRAMPS are based on frequency-switched Lee-Goldburg (FSLG) experiment
426
Medek
(40 43). The frequency switching removes the pedestals of the proton peaks
and better resolution can be afforded, especially when applied in the indirect
domain of a 2D 1H-1H experiment (44,45). Both CRAMPS and FSLG work in
spite of the MAS spinning. The currently available very fast MAS spinning
cannot be applied, because the MAS spatially induced averaging would interfere
with the similar averaging in the spin space. Both techniques work under the
assumption that the full cycle of pulses is much shorter than one rotor cycle.
More details on the homonuclear decoupling can be found in the section describing decoupling techniques. In spite of all the challenges, there is a big incentive to
exploit the proton spectroscopy in the solid state, mostly due to the ubiquitous
presence of protons in all organic compounds, their 100% natural abundance,
and high gyromagnetic ratio, both translating into excellent sensitivity.
Heteronuclear SSNMR
The difficulty to get highly resolved proton spectra of crystalline solids largely
precluded widespread applications of proton spectroscopy. The SSNMR
spectra of magnetically active nuclei other than protons are typically void of
the limitations encountered in proton applications. Carbon, nitrogen, fluorine,
and phosphorus are the examples of spin 1/2 nuclei present in pharmaceutically
relevant systems.
Carbon and nitrogen are isotopicaly diluted nuclei, having 1.1 and 0.36%
natural abundance, respectively. The spin dilution basically eliminates the
homonuclear dipole-dipole coupling. For example, the probability of having
two 13C carbons next to each other is 1/10,000. Even a moderate MAS speed
will remove such dipolar coupling completely. Though a strong dipole dipole
coupling of carbons to protons still prevails, it is relatively easy to decouple
by a high-power proton irradiation (3,4). Before the development of the
signal-enhancing CP technique, the two serious limitations that precluded extensive use of carbon and nitrogen solid-state spectroscopies were low sensitivity
and long T1 relaxation times of these nuclei. Both of these limitations are
largely removed by CP of NMR magnetization from protons. CP relies on a
strong pool of proton magnetization that is transferred during the CP step by
applying spin-locking pulses on both proton and the heteronuclear channels
(13,14). The simultaneous spin-locking pulses on the proton and the insensitive
nucleus channel must be of the specific rf amplitudes required by the HartmanHahn matching condition (13). The theoretical maximum enhancement factor is
given by the ratio of gyromagnetic ratios of protons and the insensitive nuclei.
The enhancement factor is up to 4 for carbons and up to 10 for nitrogens.
At the beginning of the CP experiment the magnetization starts at protons. As
such, it is the proton T1 relaxation that determines the recycle delay between
successive scans. Due to the efficient dipole dipole relaxation mechanism, the
proton relaxation (T1H) is typically substantially faster than that of carbon
(T1C). The recycle delay between successive scans of CP-based experiment
SSNMR Spectrometry
427
can be shortened significantly when compared with carbon direct polarization

acquisitions without the CP signal enhancement. Therefore, the CP experiment improves carbon sensitivity and, at the same time, shortens the necessary
recycle delays.
Fluorine and phosphorus are both almost 100% naturally abundant. The
obvious advantage is great sensitivity. However, if more than one of these
nuclei are present simultaneously in the molecule of interest, the homonuclear
dipolar coupling may be significant, calling for CRAMPS-like type of linenarrowing solution. Even when a single fluorine or phosphorus is present, the
distances between the two nuclei, each residing on different molecules, may
still be short enough to give rise to significant homonuclear dipolar coupling.
Arguably, carbon is the most important of the aforementioned nuclei.
Depending on the crystallinity and residual dynamics of the compound, the
13
C SSNMR peaks can be very narrow. Compounds with high internal mobility,
such as adamantane with its fast rotation of the whole molecule around its center
of gravity, narrows down the line widths below 0.01 ppm. Coupled to its favorable relaxation properties, adamantane is frequently used not only as a chemical
shift standard but also for optimizing experimental variables and for shimming
probes. A typical line width encountered in crystalline organic compounds is
in the order of 0.5 ppm, but it is strongly dependent on the compounds mobility
and the available proton decoupling field. It is not unusual to resolve the majority
of carbon resonances in compounds containing 40 or more carbons.
Oxygen is also present frequently in pharmaceutically relevant materials,
but the low natural abundance of magnetically active 17O of 0.038% makes
NMR applications on nonisotopically enriched samples impractical. Moreover,
17
O is a quadrupolar spin (S 25/2) bringing additional issues of quadrupolar
interactions to cope with.
It is worth noting here that in many cases the nature of the samples studied
does not allow for the incorporation of spin labels (e.g., analysis of production
lots). Even when dealing with research samples, studied for development purposes, it is rather difficult to obtain isotopic enrichment of the small organic molecules in quantities suitable for SSNMR.
In addition to the nuclei discussed previously, other NMR spins may be
occasionally present in pharmaceutical materials and may be relevant for
NMR studies. Because many of these more exotic nuclei, such as chlorine,
bromine, sodium, potassium, or manganese, possess spin higher than 12, special
experimental arrangements discussed earlier have to be applied.
PHARMACEUTICALLY RELEVANT APPLICATIONS OF SSNMR

Introduction
The goal of this review is to give the readers a balanced overview of all the
currently available pharmaceutical applications of SSNMR. Less emphasis is
428
Medek
put in the comprehensiveness of reviewing all the original literature. On the other
hand, some nonpharmaceutical applications of SSNMR are also included (e.g.,
SSNMR characterization of polymers, food and cosmetic science, or quantification of natural organic matter). The reason for including these seemingly nonrelated topics is the potential of these applications for pharmaceutical use. The
common thread is that they are applied to natural abundance samples to study
crystalline and amorphous phases and, therefore, are directly applicable to
characterization of pharmaceutical materials with little or no additional modifications. It is envisioned that the inclusion of these topics will bring new ideas
and perspectives to the pharmaceutical SSNMR field. The primary emphasis is
given to literature appearing after year 2000, but whenever appropriate, older
literature is also included. The pharmaceutical SSNMR applications reviewed
(in no particular order) are:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
Crystal form characterization (polymorphism) and determination of

the number of molecules per asymmetric unit.
Differentiation of crystalline and disordered (amorphous) phases.
Identification and characterization of excipients and nonpolymorphic
impurities.
Identification of crystal-bound solvents.
Identification and quantification of the different phases when in
mixtures.
Structural characterization based on fitting the experimental chemical
shift and lineshape data.
Structural characterization based on measurement of interatomic
distances.
Characterization of domain sizes.
Characterization of sample dynamics.
Characterization of drug drug or drug excipient interactions.
Physical and chemical stability and solid-state reactivity studies.
Characterization of hydrogen bonds and proton transfer.
Worth noting is that any of these SSNMR applications can be used to generate
and claim an intellectual property. In a sense, generating the intellectual property
is yet another pharmaceutical application of SSNMR.
Previous Reviews
Only reviews detailing pharmaceutical applications or containing some pharmaceutically relevant aspects are discussed. Brittain et al. (46) published a review on
physical characterization of pharmaceutical solids containing a short chapter on
SSNMR. Bugay (47) reviewed the theory and pharmaceutical applications
of SSNMR. Holzgrabe et al. (48) published review on NMR spectroscopy in
pharmacy, mostly dealing with solution application but containing a section on
SSNMR. Burn et al. (49) published a chapter devoted to SSNMR in their book
Solid State Chemistry of Drugs. Duer (50) published a review on the SSNMR
SSNMR Spectrometry
429
studies of molecular motion. Sections on powder line shape analysis, relaxation

time measurements, exchange experiments, and applications are included. Smith
published a review of SSNMR as applied to various fields, such as organic, organometallic, bio-organic and inorganic materials, liquid crystals and membranes,
silicates, microporous materials, and glasses and ceramics (51). This reviewcontains almost 1000 references to the original literature. Potrzebowski (52) presented a review of SSNMR studies of carbohydrates and their analogs. Many
examples of 13C CPMAS, 31P MAS, 2H MAS, 2D correlation experiments and
single-crystal NMR studies are shown. Ye et al. (53) presented a concise review
of modern SSNMR experiments. Stephenson et al. (54) published a review on
quantitative issues of spectroscopic techniques (including SSNMR) when
applied to pharmaceutical solids. Bugay (55) published a review of spectroscopic
techniques used to characterize pharmaceutical materials in the solid state. A
review of SSNMR is included. Harris (56) published a review on recent advances
of SSNMR containing a clear and concise overview of important SSNMR experiments. Shapiro and Gounarides (57) published a rivew of high-resolution MAS
(HRMAS) applications to combinatory chemistry. Bryce et al. (58) presented a
review on practical aspects of modern routine 1D SSNMR spectroscopy. Ando
et al. (59) and Kolodziejsi and Klinowski (60) reviewed the quantification
aspects of CPMAS technique. Dybowski et al. (61) presented a short review of
SSNMR, mainly considering years 2000 and 2001. Auger (62) published a
short review on biological and pharmaceutical applications of SSNMR. The
review contains mostly biological applications (structure and dynamics of lipid
and proteins of biological membranes).
The short review by Reutzel-Edens and Bush (63) described the pharmaceutical applications of SSNMR in characterization of small molecules. The
phase identification capability of SSNMR and its suitability for pharmaceutical
formulations are stressed. The aspects of intellectual property and regulatory
issues are also discussed. Schmidt-Rohr (64) published a review on dynamics
in polymer. Although not centered exclusively around pharmaceutical applications, the review by Brown and Emsley (65) brings a concise and easy
to-read account of modern SSNMR experiments and applications. Potrzebowski
(66) published a short review titled What high resolution SSNMR can offer to
organic chemist. Medek (67) published a review of pharmaceutical applications
of SSNMR. Rather than being an exhaustive review of existing literature, this
chapter concentrates on review of the possible SSNMR applications to characterize pharmaceutical solids. Tishmack et al. (68) presented a comprehensive
and well-balanced review of pharmaceutical applications. Offerdahl and
Munson (69) published a short review on SSNMR of pharmaceutical materials.
Vinogradov et al. (45) published a review on strategies for high-resolution
proton spectroscopy in SSNMR. It compares the available techniques and evaluates their applicability under different conditions (spinning speeds, decoupling
power, etc.). Watts (70) published a review on SSNMR in drug design and discovery for membrane-embedded targets. Aliev and Law (71 75) published a
430
Medek
series of comprehensive but concise reviews of SSNMR spectroscopy in the

years 2000 to 2005.
Review of Basic SSNMR Experiments
NMR can probe various aspects of materials of interest. By exploiting the nuclear
interactions discussed in the previous chapter, one can, in general, gain understanding about the chemical and physical nature of the system, the 3D arrangement of its atoms and its dynamics. Before discussing the pharmaceutical
applications in detail, it is useful to review the conventional SSNMR experiment
arsenal. Only experiments commonly referenced throughout this review are
discussed. Pulse sequences designed for specialized, one-of-type of studies, or
even more conventional ones, but typically applied to other than pharmaceutically relevant studies are excluded.
The simplest SSNMR experiment is a direct signal observation (one pulseacquire) on static samples (no sample rotation applied). Because all the nuclear
interactions will show their full strength, this is applicable only on systems
with a few isolated nuclei. For example, it would be of interest to describe the
shielding tensor of a nitrogen in system where the 15N 15N homonuclear
dipolar coupling can be neglected. Typically the static line shape would be
acquired in the spin echo mode (Fig. 7) to eliminate the dead time of the electronics and to keep the pulse voltage ring down. It should be noted that similar
information could be obtained with better sensitivity while spinning. This spin
echo experiment can be also combined with MAS.
Decoupling Techniques
The most typical SSNMR experiment is the direct polarization (one pulseacquire) coupled to MAS (Fig. 8). Of the pharmaceutically relevant spins,
it is typically applicable only to highly abundant nuclei, such as 1H, 19F, and
31
P. A high-power proton decoupling during acquisition is applied to remove
the heteronuclear dipolar coupling. A caution has to be exercized when combining the decoupling with MAS. The MAS can interfere with decoupling if the frequency of spinning is in the order of the decoupling field.
An interference with decoupling can also arise if the frequencies of the
random internal sample motions approximately match that of the decoupling
field (76). Both types of interferences might substantially degrade the decoupling performance.
Figure 7
Spin echo experiment.
SSNMR Spectrometry
431
Figure 8 DP experiment, typically coupled with high-power proton decoupling and

magic angle spinning. Abbreviation: DP, direct polarization.
Two-pulse phase modulation (TPPM) decoupling sequence rather than the

simple continuous wave (CW) decoupling is typically used (77). The TPPM
scheme uses two pulses of duration corresponding to slightly less than 1808
flip angle with a phase difference of 15 to 308. It shows significantly better
off-resonance performance than CW, especially when combined with a moderately fast MAS. More recently, novel heteronuclear decoupling schemes were
designed by various groups, outperforming TPPM under special conditions
(e.g., very fast MAS) (78 83). Noteworthy are the XiX and SPINAL-64
sequences. The XiX decoupling consisting of windowless rf irradiation with a
repeat of two pulses of equal width, phase shifted by 1808 (81,84). Gains in
peak height up to 29% over TPPM were demonstrated but realized mostly
when coupled to high-spinning speeds (.30 kHz) and decoupling fields
(.110 kHz). Interestingly, for very high MAS spinning speed exceeding the
proton-proton dipolar coupling (i.e., .40 kHz for rigid solids), it has been
shown that much lower decoupling power actually achieves almost the same
decoupling efficiency as the optimized decoupling at 15 times higher power
(Fig. 9) (83,85,86). SPINAL-64 decoupling, originally developed for liquid
crystal applications, consists of windowless irradiation with changing of phase
in a stepwise manner and then combining the basic cycles into supercycles
(79). Better TPPM performance has been shown for crystalline solids, especially
for low rf power levels (87). More recently, DePaepe et al. (88) devised a methodology of optimizing the performance of the heteronuclear decoupling. This
generic approach, which is referred to as eDUMBO, consists of an iterative
optimization of the decoupling performance based on experimental feedback.
The eDROOPY decoupling sequence (with a simple cosine modulation of the
decoupler RF phase) was designed using the eDUMBO approach and was
shown to outperform the optimized TPPM and XiX sequences (Fig. 10) (89,90).
A quite different class of decoupling schemes is used for homonuclear
decoupling (i.e., removing the strong proton proton dipolar interaction while
preserving chemical shift or heteronuclear coupling) (34 41,43,91 103). A
recent comprehensive account of homonuclear decoupling was given (45). The
strong homonuclear dipolar coupling of protons is the direct result of their
100% natural abundance and high gyromagnetic ratio. Due to the homogeneous
character of this coupling, MAS narrows down the proton peaks substantially
432
Medek
Figure 9 Comparison of low power CW decoupling at various spinning speeds y r to

optimized TPPM. Relative intensities and line widths of the a-carbon resonance of
[13C-2] glycine are shown. Abbreviations: CW, continuous wave; TPPM, two-pulse
phase modulation. Source: Adapted from Ref. 85.
only at speeds equivalent of exceeding the strength of the dipolar coupling. To

efficiently narrow down the proton line widths, MAS must be combined with
similar averaging in the spin space. The early pulse sequences to achieve this
were Lee-Goldburg (LG) (5,104) and Waugh, Huber, and Hacberlen (WHH4)
(5). Because they were applied to static samples, the proton lines were broadened
by the chemical shift interaction. CRAMPS (30,32) (combined rotation and
MAS) combined both the spin space and real-space averaging techniques.
Based on the averaging principle, there are four classes of homonuclear decoupling techniques: (i) solid echo-based sequences [WHH4 (5), MREV8
(105,106), BR24 (92), CORY24 (94), BLEW12 (93), and DUMBO (99), (ii)
magic echo sandwich-based sequences [TREV8 (107) and MSHOT3 (37,108)],
(iii) Lee-Goldburg based sequences [LG (104), FSLG (40 42), PLMGn
(42,43)], and (iv) rotor synchronized sequences [semi windowed WHH4, CNkx
Figure 10 Comparison of the decoupling performance of different decoupling techniques. Relative intensities and line widths of the a-carbon resonance of [13C-2] glycine
are shown. Source: Modified from Ref. 89.
SSNMR Spectrometry
433
Figure 11 The irradiation field profile of Bruker 7-mm rotor as a function of the rotor
z-axis disctance. The profile was obtained by measuring NMR signal of 1-mm thickness
rubber disk placed at different z-positions inside the rotor. The irradiation field profile is
shown at the bottom. Note that this profile is far from the ideal flat response. It can also
hamper quantification of multiple components if there is substantial component
segregation inside the rotor. Abbreviation: NMR, nuclear magnetic resonance. Source:
Modified from Ref. 173.
(109) and RNkx (110)]. PLMG and DUMBO sequences were found to be preferred
due to their relative ease of implementation and their adaptability to fast
MAS spinning speeds (45). The rf irradiation inhomogeneity (Fig. 11) and the
contribution of unaveraged higher order terms remain the main reasons for
limiting the attainable resolution.
CP Techniques
Because the majority of the pharmaceutical SSNMR applications involve 13C
acquisitions, arguably the most typical pulse sequence is the CP between
protons and carbons (Fig. 12), which was already discussed in the previous
section. In order to set up the CPMAS experiment parameters, it is useful to
know the various relaxation times of the system under study. As discussed
next, the CP dynamics is a function of the relaxation times of both nuclei
involved (T1H, T1rH, T1C, and T1rC) and the CP constant TCH. Due to the long
carbon relaxation times, CP dynamics can often be approximated by employing
just the proton relaxation. The most sensitive (and fastest) way to determine
proton relaxation times is through direct proton detection. However, if a
434
Medek
Figure 12 Cross-polarization magic angle spinning (CPMAS) experiment. After the

initial 908 proton pulse, spin-locking fields on both channels are applied matched according the Hartmann-Hahn condition, causing a net signal flow from the abundant proton
reservoir to rare nucleus (schematically shown as the dotted arrowhead line). The rare
nucleus signal is detected usually with high-power proton decoupling. The last proton
908 pulse (usually referred to as the proton flip-back pulse) is optional. It returns the
remaining proton transversal magnetization into the IZ direction, thus speeding up the
return to equilibrium between the successive scans.
mixture with multiple components is involved, heteronuclear detection may offer

advantage in resolving the components. Figure 13 shows the CP-based pulse
sequence for measurement of the proton longitudinal relaxation time (T1H).
The equilibrium proton magnetization is inverted by the initial 1808 pulse, followed by the variable time relaxation delay trelax, during which the magnetization
returns back to equilibrium with the rate corresponding to the inverse of T1H.
After the proton 908 pulse, the proton magnetization is transferred to the xy
plane, where it can be either directly detected by proton acquisition, or, as actually shown in Figure 13, it is cross-polarized and detected heteronuclearly.
Figure 14 shows the pulse sequence for measurement of the proton relaxation
time in the rotating frame (T1rH). The equilibrium magnetization is transferred
to the xy plane followed by the spin-locking pulse of variable lengths during
which the magnetization decays by the rate corresponding to the inverse of
T1rH. The magnetization can again be detected directly by proton acquisition
or, as shown in Figure 14, after cross-polarizing it to heteronuclei. It should be
noted that the value of the measured T1rH, depends strongly on the strength of
the applied irradiation field (the stronger the rf, the longer the T1r).
In many cases, especially at higher magnetic fields, the maximum rotor
spinning speed is smaller than (or on the order of) the chemical shift anisotropy,
Figure 13
CPMAS-based proton T1H relaxation sequence.
SSNMR Spectrometry
Figure 14
435
CPMAS-based proton T1rH relaxation sequence.
leading to a spinning sideband manifold, which may (i) compromise resolution

by introducing peak overlap, (ii) complicate quantification, and (iii) lead to
misassignments of peaks in the absence of variable spinning speed repetitions.
In 1982, Dixon designed an ingenious and simple way to suppress the spinning
sidebands by introducing the total suppression of spinning sidebands (TOSS)
sequence (111,112). TOSS cancels the spinning sidebands by destructive
interference by adding four (or in general 2n 2) of the properly spaced 180
carbon pulses right after the CP step (Fig. 15). An added advantage of TOSS
is the long (100 ms) delay before the start of acquisition, eliminating the
receiver dead time problems. However, if many spinning sidebands are
present, TOSS may lead to slight phase distortions, incomplete suppressions,
or even negative intensities of the spinning sidebands.
Spectral Editing
A very important class of SSNMR pulse sequences represents the spectra editing
techniques, because they simplify spectra according to some factors (usually
proton multiplicities) to facilitate the desired peak assignment. As evidenced
by many examples of the reviewed literature that follows, the typical 1D
SSNMR peak assignment consists of comparison of the chemical shift with solution assignments in combination with spectra editing. In many cases, a close-tocomplete peak assignment can be achieved by using this approach. The most
simple of the editing techniques is the so-called interrupted decoupling (also
referred to as dipolar dephasing or nonquarternary carbon suppression), which
consists of switching off the decoupling field for short period of time of
Figure 15 CPMAS sequence with TOSS (total suppression of spinning sidebands) and
proton flip back pulse.
436
Figure 16
Medek
Interrupted decoupling spectral editing.
approximately 40 100 ms after CP and before acquisition (Fig. 16) (113). The
decoupling delay dephases the strongly proton-coupled carbons (CH and CH2
groups) but leaves the qaurternary, and also CH3, groups largely unchanged.
The methyl group is not dephased due to the partial averaging of their 1H 13C
dipolar coupling as a result of fast, wheel-like rotation in the solid state. Complementary to this technique is the short CPMAS (also referred to as quaternary
carbon suppression), with CP limited to only a short period of approximately
100 ms. During this short CP, unlike the protonated carbons, the quaternary
carbons do not have enough time to develop appreciable magnetization and are
suppressed. When the CP is combined with short period of polarization inversion
with 1808 change of phase (CPPI; Fig. 17A), CH signals are cancelled, CH2 are
negative with intensity of 1/3, and methyl and quaternary carbons are attenuated
by 62 and 86%, respectively (114). Methylene only spectra can be obtained
by the simultaneous phase inversion (SPI) variant of this pulse sequence
(Fig. 17B), (115,116). Later, this technique was improved to give better
suppression of artifacts and better S/N. It was termed CPDDRCP for CP
Figure 17 CPPI. The X and X symbols denote 0 and 1808 phases of the pulses, respectively. (A) Standard CPPI affording negative CH2, positive CH3 and quaternary and no CH
carbons. (B) SPI affording only (negative) CH2 sub-spectra. Abbreviations: CPPI, Crosspolarization with polarization inversion; SPI, Simultaneous phase inversion.
SSNMR Spectrometry
Figure 18
437
Edited monoethyl fumarate spectra. Source: Modified from Ref. 117.
dipolar dephasing re-crosspolarization, and it affords CH and CH2 only spectra

(Fig. 18) (117). Different approaches to select methylene only subspectra were
used by Rossi et al. (118). This DQCP approach is based on first generating
proton double-quantum coherence, which develops fast for CH2 groups, filtering
it, converting back to single quantum coherence, followed by CP and detection
on carbon channel. As an example, an application of these sequences to select
side-chain carbons of peptides was shown by Kumashiro et al. (119). De Vita
and Frydman (120) utilized the editing power of the amplified dipolar dephasing
coupled with the fact that at moderate spinning speeds (.10 kHz) the CH and
CH2 groups behave like isolated spins, that is, they are isolated from the surrounding proton reservoir. The technique distinguishes between CH, CH2, and
quarternary/methyl groups. No homonuclear decoupling has to be applied
unlike the windowless isotropic mixing for spectral editing sequence (WIMSE)
(121). Various other variants of dipolar coupling-based spectral editing appeared
in the literature (122 127). Selection of carbons bound to 14N was shown by
Schmidt-Rohr and Moa and named saturation-pulse-induced dipolar exchange
with recoupling (SPIDER) (128). For quadrupolar nuclei, editing out the
narrow line components with smaller quadrupolar coupling strength was demonstrated (129)
The editing techniques discussed so far are all based on through-space
dipolar coupling. Lesage et al. (130) used hetoronuclear J-coupling editing
approach. The result of this through-bond transfer is the solid-state equivalent
438
Medek
Figure 19 SS-ATP. FSLG is frequency-shifted Lee-Goldburg decoupling, and t 1/

(2JCH) with typical values ranging from 4 to 6 ms. Abbreviation: SS-ATP, Solid-state
attached proton test.
of the solution sequence-attached proton test, named SS-ATP (Fig. 19). This
sequence affords spectra with CH2 groups having an opposite phase to the CH
and CH3 groups (quaternary carbons are not observed; Fig. 20). Although excellent editing efficiency is obtained, narrow carbon lines of less than approximately
50 Hz are required to resolve the J-coupling (1JCH, scaled to approximately 70 Hz
by the FSLG scaling factor), restricting the use of this technique to highly
crystalline samples. The J-coupling editing approach was further developed to
afford pure subspectra of carbons according to the proton multiplicity. It is
based on multiple quantum filtering of the proton coherences (J-MQF; Fig. 21)
(131). Unlike the dipolar coupling-based editing, this method is insensitive of molecular motions and MAS spinning. However, it suffers from rather low sensitivity.
Although it is not the spectral editing technique, a spin selection based on
the solution variant of IPAP experiment was shown to enhance resolution of
multiple spin-labeled systems, because the contribution of J-coupling to the
line width is removed (132,133). A selectivity in coherence transfer for 2D
Figure 20 SS-APT spectral editing. Abbreviation: SS-APT, solution sequence-attached

proton test. Source: Modified from Ref. 130.
SSNMR Spectrometry
439
Figure 21 J-MQ-filtered 13C CPMAS spectra of the tripeptide Boc-Ala-Ala-Pro-Obzl.

Abbreviations: MQ, multiple quantum; CPMAS, cross-polarization and magic angle spinning. Source: Modified from Ref. 131.
recoupling experiments, which can lead to enhanced resolution was also reported
by Carravetta et al. (134). An 1D technique to separate spinning sidebands
according to their order was proposed (135).
Correlation Techniques
To take the full advantage of the wealth of information SSNMR offers, multidimensional correlation techniques should be employed. In general, different
nuclear spin interactions are selectively switched on and off during the evolution,
mixing, and acquisition periods of these experiments. The purpose of using multidimensional experiments is to have more information available (e.g., better
resolution) by using the higher dimensions. On the other hand, many of the
multi-dimensional techniques could be reduced to their 1D variants. In these
cases, the mixing period of the originally multidimensional experiment serves
as a selective information filter. Some examples of the correlation techniques
include chemical shift correlations (tracing through space or through bond
atom connectivities and facilitating a site-specific chemical shift assignments
and confirming the identity of molecules), recoupling techniques (reintroducing
the MAS averaged anisotropic interactionstypically dipolar coupling but also
the chemical shift anisotropy or the quadrupolar couplingto gain structural
information, such as distances and torsional angles or also to help assignments),
and chemical exchange techniques (studying slow dynamics or chemical reactions). Again, as in other sections, the emphasis is to preferentially review
440
Medek
Figure 22 WISE (wideline separation) experiment separates the proton MAS line
shapes by the carbon chemical shift of close-by carbons. Because it involves CP transfer,
it is a through-space correlation technique. It is typically applied to get information on
mobility of the different phases in the sample. Abbreviations: MAS, magic angle spinning;
CP, cross-polarization.
techniques applicable to nonlabeled materials conducive with most pharmaceutical applications.

WISE (wide line separation; Fig. 22) is a 2D technique separating the
proton MAS lines by the carbon chemical shifts (136). The free proton evolution
periods follows the initial proton 908 pulse, after which the magnetization is
cross-polarized and detected on the carbon channel. Because WISE involves
CP-based polarization transfer, the correlation is dipolar coupling driven and,
consequently, takes place through space. If multiple phases with limited spin diffusion exist, the mobility of the protons in the different phases can be judged by
the line width of the corresponding proton slices (wide lines correspond to rigid
domains with low mobility and vise versa). This technique is also frequently
applied in its reduced, 1D version, with the proton evolution representing the
proton T2 or mobility filter. It is especially suited for polymer science to
filter out the rigid crystalline polymer domains. By setting the proton evolution
time to longer values, the signals originating from the rigid components decay
to zero and only the mobile portion of the sample is detected.
ODESSA (one-dimensional exchange spectroscopy by side-band alteration) is a rotor-synchronized MAS exchange technique, with preparation period
fixed to one-half of a rotation period and a mixing time equal to an integer
number of rotation periods (Fig. 23) (137). After the half-rotor period and just
before the mixing, the spinning sidebands are polarized in the alternate fashion.
This creates an initial difference in polarization. Dipolar coupling-driven spin
Figure 23 ODESSA sequence. Abbreviation: ODESSA, one-dimensional exchange

spectroscopy by sideband atteration.
SSNMR Spectrometry
441
Figure 24 The 1D CODEX characterizes slow motions through chemical shift anisotropy modulations. After the cross-polarization step, the magnetization evolves under
chemical shift anisotropy (recoupled by the series of 1808 pulses), followed by mixing
time tm and second evolution period under chemical shift anisotropy. At the end of this
period, the signal is refocused by stimulated echo. The amplitude of the echo, reduced
by the motions, is detected after the TOSS segment as magic angle spinning peak intensity.
Abbreviation: TOSS, total suppression of spinning sidebands; CODEX, center band only
detection of exchange.
exchange then results in a redistribution of the polarization and the corresponding

decay in the sideband intensities. Internuclear distances can thus be measured.
High abundance nuclei (19F or 31P) or isotope enrichment is required.
A technique related to ODESSA is the center band only detection of
exchange experiment (CODEX) (138 140). CODEX enables the observation
and characterization of slow molecular motions with the high resolution of 13C
chemical shifts (Fig. 24). The chemical shift anisotropy serves as the actual
probe of the motions. The type of motions can be established from a series of
1D CODEX experiments.
The 2D HETCOR (heteronuclear correlation) experiment correlates abundant nucleus (protons) chemical shift through space to nearby carbons (or other
hetreronulei) based on the dipolar coupling mechanism (Figs. 25, 26) (141 143).
For short CP contact times, the one bond H C correlations are preferably
selected, whereas at longer contact time this through-space experiment also
often includes intermolecular correlations, which, depending on the application,
may or may not be desired. Mao et al. (144) applied 2D HETCOR and 2D
SUPER (undistorted powder-patterns by effortless recoupling) to characterize
Figure 25 HETCOR (hetoronuclear correlation) correlates proton and carbon chemical

shifts by through-space dipolar interaction. The homonuclear frequency-shifted LeeGoldburg (FSLG) decoupling is used during proton t1 evolution to decouple protons
from each other. MA stands for pulses of magic angle (54.78) flip angle.
442
Medek
Figure 26 Dipolar coupling-based, trough space 2D 1H 13C FSLG (homonuclear

frequency-shifted Lee-Goldburg) heteronuclear correlation on [U 13C] L -tyrosise
hydrochloride. Source: Modified from Ref. 141.
biocolloidal organic matter (145,146). Sozzani et al. (147) applied HETCOR to

study polymer multi-phase (nanocomposite) materials.
DeLacroix et al. (148) developed 2D TOSS-based experiment separating
spinning sidebands by their order (Fig. 27). This technique allows for separation
of the isotropic and anisotropic chemical shifts along two dimensions, making it
possible to analyze the spinning sideband manifolds at moderate spinning speed.
Smith et al. (149) applied this experiment to study conformational polymorphism
of ROY compounds (Fig. 28). Chemical shift anisotropy of ROY compounds
were obtained. Several multidimensional techniques based on magic angle
turning and hopping, correlating carbon isotropic shifts to their anisotropic
powder patterns, were developed in David Grants lab. When coupled to ab-initio
calculations, FIREMAT (five p-replicated magic angle turning) and its preceding
versions, provide an excellent tool to study molecular conformations and polymorphism (150 159). An improvement of FIREMAT enabling faster spinning,
Figure 27 The 2D TOSS: spinning sidebands are separated according to their orders.
The t1 evolution is typically sampled over only one rotor period. The number of t1 increments must be chosen to correspond to the number of the sidebands. Abbreviations: TOSS,
total suppression of spinning sidebands.
SSNMR Spectrometry
443
Figure 28 The 2D TOSS spectrum of ROY-orange form (top). Simulated spectrum

using a best fit to the CSA tensor values (bottom). Abbreviation: TOSS, total suppression
of spinning sidebands. Source: Modified from Ref. 149.
and hence avoiding the need for ultrastable slow spinning speeds, was designed at
the same lab and is referred to as SPEED MAS (160). The technique uses a series
of p-pulses to effectively reduce the apparent spinning speed. Elena et al. (161)
developed an experiment separating fast and slow spinning sidebands, dubbed
ROSES. The principle of the chemical shift anisotropy amplification was
applied to increase the number of sidebands in the indirect domain of the 2D
correlation (162).
Gullion and Schaefer (163 174) introduced an important technique to
measure internuclear distances, which is referred to as rotational echo double
resonance (REDOR, Fig. 29). Inserting several p pulses within the rotor period
of MAS interferes with its ability to average out the heteronuclear dipolar
coupling (recouples the averaged dipolar coupling), and as such, decreases the
amplitude of the rotational spin echoes (Fig. 30). The attenuation of the rotor
echoes is proportional to the heteronuclear dipolar coupling and hence to the distance between the nuclei. REDOR was originally devised for measuring 13C 15N
distances specifically in 13C- and 15N-labeled samples. By judicious choices of
444
Medek
Figure 29 Basic 13Cf15Ng REDOR sequence. Inserting the p pulses within the rotor
periods of MAS interferes with MAS ability to average out the 13C 15N heteronuclear
dipolar coupling and, as such, decreases the amplitude of the rotational spin echoes
(Fig. 30). The attenuation of the rotor echoes is proportional to the heteronuclear
dipolar coupling and hence to the distance between the 13C 15N nuclei. Abbreviations:
REDOR, rotational echo double resonance; MAS, magic angle spinning.
Figure 30 13Cf19Fg REDOR dephasing curves as a function of the dipolar evolution time
for complexes of LY329332 (chloroeremomycin) with whole cells and cell walls. The cell
walls with LY329332 binding occupancy of 33% (open circles) and 16% (closed circles)
were used (hence the different dephasing corresponding to the same average fluorine
carbon distance). Source: Modified from Ref. 174.
SSNMR Spectrometry
445
the placement of the pairs of labels, a structure of a molecule can be determined

one distance at a time. Multiple improvements of the original REDOR
scheme appeared in literature (175 178). The notable one, u-REDOR, enables
simulataneous measurements of many distances in multiple-labeled molecules
(179 182). The methodology was also extended to measure distances between
quadrupolar nuclei (SEDOR, REAPDOR, and TRAPDOR) (183 191). Fu
measured 15N 1H bond lengths based on REDOR-type experiment (192). FSLG
was used during proton evolution. A simple compensation for pulse angle error
in REDOR experiments was also devised (193).
Saalwachter and Spiess (194) designed REDOR-based heteronuclear correlation sequences (HMQC, HSQC, HDOR, and RELM), which work optimally at
very high MAS spinning speeds (.20 kHz). They use the REDOR-like pulse
sequence to excite and re-convert the heteronuclear coherences. Another set of
these four heteronuclear correlation experiments designed in the same lab is
based on the recoupled polarization transfer (REPT) (195). Weak heteronuclear
dipolar coupling constants can be extracted by means of the spinning sideband
analysis extracted from the indirectly detected domain, and structural information
can be obtained. Comparison of the REPT performance with that of HETCOR,
MAS-J-HMQC is shown in L -tyrosine, which is abundant in nature. Applications
to spectral editing are also shown. All these sequences work best at high-spinning
speeds and are applicable to natural abundance samples encountered in most of
the pharmaceutical applications. Saalwachter and Schnell (196) further built on
the REDOR scheme, making it more sensitive to extract the 1H 13C couplings
for up to medium range distances (REREDOR).
The same authors also introduced 1H 13C and 1H 15N correlation experiments with highly improved sensitivity by inverse detection on protons, applicable also to natural abundance 15N samples (197,198). Ishii et al. (199,200)
described the merits of inverse versus direct detection of 1H 13C correlation.
Applications to unlabeled polymers are shown (Fig. 31). Indirect detection was
also applied to peptides and proteins (201 204) and small molecules (205).
Figure 31 The standard carbon (left) and inverse proton dectection (right) of 2D
13
C 1H heteronuclear correlation spectra of poly(methyl methacrylate). Note the sensitivity enhancement of the inverse-detected spectrum. Source: Modified from Ref. 200.
446
Medek
Figure 32 The MAS-J-HMQC (magic angle spinning, J-coupling based heteronuclear

multiple quantum correlation) through bond correlation experiment. The homonuclear
frequency shifted Lee-Goldburg (FSLG) decoupling is used during the JCH coupling
evolution period t, as well as during proton t1 evolution to decouple protons from each
other while preserving the JCH coupling during period t. MA stands for pulses of
magic angle (54.78) flip angle. The carbon 1808 pulse refocuses carbon chemical shift.
Building on the previously proved concept of HETCOR by J-coupling

cross-polarization (206), Lesage and Emsley (207,208) introduced throughbond MAS-J-HSQC (J-coupling-based heteronuclear single quantum correlation)
and MAS-J-HMQC (J-coupling-based heteronuclear multiple quantum
correlation; Fig. 32) experiments, which correlate proton chemical shift by the
J-coupling mechanism. Both sequences afford similar proton resolution on
unlabeled samples, but MAS-J-HMQC offers slightly better sensitivity and less
artifacts. For 13C-labeled samples, MAS-J-HSQC provides a resolution advantage. The same lab showed applications of these techniques to the complete
13
C and 15N resonance assignments of naturally abundant peptides in the
absence of any other solution chemical shift or structural information (Fig. 33)
(209). Based on their through-bond nature, these experiments assign the short
and long range correlations very reliably without any artifacts, because no
inter-molecular correlations are present in the spectra. The authors also
devised a variant of the MAS-J-HSQC suitable for detection of sample mobilities. In essence, it is the WISE experiment coupled to the through bond,
J-coupling-based 1H 13C polarization transfer, named J-WISE (210). The advantage of this experiment over WISE stems from the fact that J-WISE can
distinguish bonded and nonbonded interactions. It was applied by the authors
to localize bound water in the onion cell architecture. Total through-bond correlation spectroscopy (TOBSY) for solids was also devised (211 213). Alonso and
Massiot (214) used different variants of through-bond heteronoclear correlation
based on solution INEPT experiment to characterize mesostructured materials.
Another type of J-coupling-based, through-bond correlation is the solidstate analog of the well-known solution experiment 2D INADEQUATE
(Fig. 34) (215). It is typically applied to carbons to trace the through-bond
carbon carbon connectivities. It is a very powerful technique for carbon assignments. It is based on the excitation and evolution of DQ carbon coherences and as
such requires the presence of two connected 13C nuclei. In unlabeled samples, on
SSNMR Spectrometry
447
Figure 33 Through bond 2D MAS-J-HMQC spectra of the natural abundance tripeptide

Boc Ala Ala Pro O Bzl. The 1H 13C one bond and multiple bond correlation spectra
were recorded with evolution times of 1.6 and 16 ms, respectively. The 1H 15N correlation
spectrum was recorded with evolution time of 4.32 ms. Source: Modified from Ref. 209.
average, only one out of 10,000 carbons conforms with this scenario. Spin enrichment is, therefore, preferred, but the authors also show an INADEQUATE
application to naturally abundant L -alanine. This experiment was also applied
to observe NHN hydrogen bond directly, through 15N 15N J coupling
mediated by the H-bond (216). The refocused variant of this experiment
was applied to carbon assignment of disordered solids, such as wood and
448
Medek
Figure 34 The 2D INADEQUATE for tracing carbon-carbon connectivities by throughbond 1JCC-coupling based correlation. The carbon double quantum coherence is formed
after the first carbon 908 pulse, evolved during t1, and detected as an antiphase coherence
after the second carbon 908 pulse.
cellulose (217). From the experimental T2C values, which were in the order of 10

times longer than the apparent T2C
based on the inverse of line widths, authors
conclude that the contribution to the line width must be through dispersion of isotropic chemical shifts. Moreover, authors noted that the distribution of chemical
shift was highly correlated (i.e., if one carbon chemical shift changes due to the
local change in the structure of the amorphous compound, the other bonded
carbon reacts to the structural change in very well-determined way) and a
much improved high resolution can be obtained in the sheared projection of
the 2D spectra (218,219). The origin of the disorder (its type) can be characterized by reproducing the experimental spectra using simulations (220). Kono et al.
(221 226) applied INADEQUATE and MAS-J-HMQC to study the polymorphism of cellulose triacetate and other biopolymers. Kaji and Schmidt-Rohr (227)
devised a filtered INADEQUATE approach, where the observed magnetization is
dephased by the difference of the recoupled anisotropic chemical shifts. It was
applied to estimate trans/gauche ratios in double 13C-labeled polymers.
Spiess group developed a correlation sequence based on DQ excitation
with the use of back-to-back pulses (BABA). Initially, it was developed as a heteronuclear correlation sequence (228,229). In the indirect domain of such 2D
spectrum, each proton resonance forms spinning sideband pattern with each sideband exhibiting narrow lines. Later, this sequence was generalized for any DQ
excitation and typically applied to homonuclear correlations between abundant
nuclei (230). Brown et al. (231) showed the application of BABA to investigate
hydrogen-bonding structure of bilirubin. Analysis of the proton DQ spinning
sideband patterns allowed for quantitative determination of proton proton distances and the geometry. The shortest distances (the largest dipolar couplings)
were obtained to a very high degree of accuracy (.0.01 nm).
Countless other recent multidimensional solid-state experiments can be
found in the literature, especially in the field of structural characterization of proteins, which has received lots of attention recently. It is beyond the scope of this
review to cover all the sequences applicable to the SSNMR characterization of
proteins, since, in general, the techniques require samples with isotope enrichment and, as such, are of limited interest to pharmaceutical field. Typically,
SSNMR Spectrometry
449
different types of recoupling of the rare nuclei (13C or 15N) anisotropic interactions are applied to reintroduce the structural information that are otherwise
lost by the averaging effect of fast MAS (128,173,180,195,232 258).
However, some of these recoupling schemes are applied to the abundant
proton nuclei only and as such could be applied to pharmaceuticals. Dusold
and Sebald (241) published a review of the different recoupling schemes.
Some selected references to correlation experiments requiring isotope enrichment are given (181,212,259 298).
Chemical Shift Assignments
The knowledge of the site-specific chemical shift assignments of peaks to the
different nuclei present in the molecule is not absolutely essential in order to
extract the pharmaceutically relevant information, such as to differentiate
crystal forms. However, the known assignments provide additional valuable information, so that one can relate the chemical shift differences between polymorphs
to conformational changes at particular molecular sites. Additional structural
insights can be gained by comparing the experimental assignments to calculated
shifts. Lesage et al. (209) recently demonstrated that spectral assignments can
be made by relying solely on the SSNMR data from the natural abundance
samples. All carbon, nitrogens, and proton peaks of a tripeptide sample were completely assigned in the solid state. However, for small molecules with a limited
number of resonances and good spectral dispersion, most of the assignments
can be done by direct comparison of the liquids and solid-state spectra. The
SSNMR resonances usually fall within several parts per million from solution
shifts. Unlike the solution, no solvent is present to affect the chemical shifts of
solids, and the dynamics of molecules in the solid state is typically much slower
so that only one kinetically frozen conformer (polymorph) typically exists, structure of which may substantially differ from that in solution. There is a solid statespecific contribution to chemical shifts that does not exist in solutionthe effect of
crystal packing (the intermolecular contribution). The solution-solid state chemical shift differences based on the different average molecular conformations typically do not exceed 10 ppm. A smaller difference, typically not exceeding 3 ppm,
comes from the packing of molecules in the crystal lattice (299).
The assignments of the well-resolved resonances can be inferred by
comparing the solid-state spectra with their assigned solution counterparts. The
remaining peaks can usually be assigned from the application of the spectraediting techniques discussed previously. Due to the relative ease of implementation and wide range of applicability, the short CP and dipolar dephasing
editing are preferred. If assignment ambiguities still exist, such as when methylene carbons resonate close to methines, more complex dipolar or J-couplingbased editing techniques should be applied. Finally, application of the various
multidimensional correlation techniques discussed earlier should be able to
remove any prevailing assignment ambiguities. Further check can be provided
450
Medek
by the comparison of experimental shifts to empirical or ab initio quantum

mechanical calculations of chemical shifts, which is likely to become an important tool for resonance assignments. Specifically, the ab-initio type calculations
are likely to prove useful in predicting conformations of polymorphs by comparing the 3D molecular structure-dependent simulated chemical shifts with
experimental data.
Molecular Structure from Chemical Shifts and Molecular Modeling
A knowledge of the experimentally determined principal values of the chemical
shift interaction tensor, or at least the isotropic value, enables one to extract a
wealth of structural information about the sample. This is true, especially when
coupled to calculations. The speed of modern computers allows for relatively
accurate predictions of chemical shielding molecules with multiple heavy
atoms, such as carbons, oxygens, and nitrogens. By reproducing the experimentally determined chemical shift tensor values by means of ab initio or empirical
calculations, the 3D structure of the compound can be determined. It should be
noted that due to the presence of the crystal lattice, intermolecular interactions
are considered to obtain accurate results. Some software packages allow for
incorporation of the crystal lattice effect. In principle, chemical shift variations
between the different molecules per asymmetric unit and different crystal
forms can be predicted in this way. The importance of this modeling approach
is likely to grow as the calculation methods improve.
Ando et al. (300) examined the 13C chemical shifts of aromatic carbons as a
probe for estimating the conformation of aromatic groups in the solid state. The
chemical shifts of the biphenyl group show systematic correlations with the
inter-ring torsion angles, when the substituent shielding effects are removed.
The correlations can be used to estimate the conformation of biphenyl groups
incorporated into larger molecules. The ab initio GIAO-CHF calculations
agreed well with experimental results.
Maciejewska et al. (301) studied structure of 4-(4-methoxyphenyl)hexahydro-1H,3H-pyrido[1,2-c]pyrimidine by 13C CPMAS and ab initio GIAO-CPHF
chemical shielding calculations. The studied compound showed two molecules
per asymmetric unit. To assign carbons specifically to the two molecules, the
shielding constants were calculated using the X-ray geometry. Good fits with
R 20.9988 (molecule A) and R 20.9994 (molecule B) were obtained
between the experimental and calculated values (Fig. 35).
Buchanan et al. (302) compared SSNMR and X-ray to study the stereochemistry of 4-acetyl-, formyl-, and carboxy-benzo-9-crown3 ether. The differences between chemical shifts were rationalized by LORG calculations as a
function of the aryl-O-CH3 torsion angle. A complementary role of the two techniques was illustrated.
Dega-Szafran et al. (303) studied some N-(v-carboxyalkyl)morpholine
hydrohalides by spectroscopic techniques and PM3 calculations. Contrary
to the solution, 13C CPMAS spectra showed nonequivalence of the ring carbon
SSNMR Spectrometry
451
Figure 35 Chemical shift assignments of 13C CPMAS spectrum (middle) of a compound

shown at top. This compound contains two molecules per asymmetric unit. The correlation
between the calculated shielding constants and the experimental chemical shifts for both
asymmetric molecules is shown at bottom. Abbreviation: CPMAS, cross polarized magic
angle spinning. Source: Modified from Ref. 301.
atoms, most likely due to the interactions of the chloride ion with ring
hydrogen atoms. Good linear relationship (r 0.997) was found between the
experimental and calculated shifts (using GIAO-CPHF) based on both the
X-ray and the semi-empirical PM3 geometries. This was used to confirm
the chemical shift assignments originally based on comparison to solution shifts.
Harper et al. (304) studied the relative stereochemistry and molecular conformation using SSNMR tensors. This new SSNMR methodology involved a
comparison of experimental 13C tensor principal values (measured by using the
FIREMAT experiment) with ab initio computed values for all possible stereoisomers. An accurate computation of the tensor values was required. About
3 15% and 1 2% error of the tensor values spaning for sp 3 and sp 2 carbons,
respectively, was achieved. The method was tested on terrain, chosen as
relatively rigid molecule with limited conformational diversity that would
otherwise influence the calculation of the tensor values. Spectra-editing experiments and comparisons with calculated tensor value were used for carbon
assignments.
Cyranski et al. (305) studied structure of 1-(2-hydroxy-4-bromophenyl)4-methyl-4-imidazolin-2-ones. Comparison of solution and solid chemical
452
Medek
shifts allowed for the identification of rigid and flexible structure fragments.
Solution to solid differences were found significant for aromatic carbons,
which are subject to the largest changes of the environment during reorientation
of the imidazolin-2-one moiety. Due to the insufficient 1H resolution in the solid
state, the hydrogen bonds were also studied by FTIR spectroscopy. GIAO-HF/631 G chemical shift calculations were used.
Olejniczak et al. (306) studied 2-(30 ,40 -dihydroxyphenyl)-7-b-D -Glucopyranos-1-O-yl-8-hydroxychroman-4-one using 2D NMR and DFT calculations.
In order to correlate the chemical tensor elements to the molecular structure,
the authors chose 2D PASS experiment to extract the tensor parameters by
fitting the spinning sideband patterns of the F1 domain projections (Fig. 36), a
very good match between the experimental and simulated data was achieved.
B3PW91 functional DFT and 6-311G basis set GIAO calculations were
used. Because the calculations were done in the vacuum (no intermolecular interactions were considered), the experimental versus calculated chemical shift
values showed linear relationship with considerable scatter (R 2 between 0.8
and 0.96 for the various tensor components).
Rajeswaran et al. (307) demonstrated the feasibility of an integrated
approach to determine 3D structure of medium-sized organic molecules
without using single crystals on N-( p-tolyl)-dodecylsulfonamide. The approach
combined PXRD data, Monte Carlo, and ab initio computations, and SSNMR
data. GIAO with DFT method using B3LYP/MIDI was used to predict the
shielding constants that were converted to chemical shifts relative to tetramethylsilane using linear equation developed over a diverse set of 37 organic compounds with RMS error of 3.6 ppm. A monomer geometry extracted based on
PXRD data using PowderSolve software was used for the chemical shielding calculations. Good matches between the calculated and observed data was achieved.
To confirm the right geometry, the computations were repeated for several combinations of torsion angles. The authors propose the following work flow for
crystal structure determination of polycrystalline samples: (i) collect PXRD
and assess the crystallinity, (ii) obtain SSNMR and assess the presence of
solvates, multiple phases, and number of molecules per asymmetric unit, (iii)
confirm chemical composition using elemental analysis, (iv) measure sample
density and compare with the X-ray calculation, (v) perform PXRD structural
computations, and (vi) validate the structure using computed SSNMR shifts or
select the most likely PXRD structure.
Aimi et al. (308) studied conformation of aromatic imide compounds containing diphenyl ether and benzophenone and diphenyl sulfide, diphenyl sulfone,
and diphenylmethane (309) moieties. Dihydral angles were estimated using
13
C CPMAS approach combined with RHF/6-31G(d) ab initio chemical shift
calculations. Chemical shift assignments were based on solution studies and
interrupted decoupling solid-state spectral editing. TOSS mode was used to
minimize spinning sidebands.
SSNMR Spectrometry
Figure 36
453
The 2D PASS experiment on flavanone. Source: Modified from Ref. 306.
Claramunt et al. (310) studied 1-benzoylazoles (imidazole, pyrazole, indole,

benzimidazole, and carbazole). To understand the 9-benzoylcarbazoles unexpected 13C CPMAS spectrum with additional splitting, GIAO/RFH/6-311G
ab initio calculations were performed. The splitting was proposed to be due to
the conglomerate structure of the two enantiomers.
454
Medek
Marek et al. (311) studied 15N naturally abundant and substituted purine
derivatives. Geometric parameters obtained by using RHF/6-31G were used
to calculate chemical shielding (GIAO and IGLO). The predicted shielding
was used to assign the experimental spectra to determine the parameters of the
chemical shift anisotropy (CSA) tensor. The experimental CSA values were
obtained by fitting the 15N CPMAS spinning sidebands. Differences between
solution and solid shifts were as large as 73 ppm.
Goward et al. (312) combined advanced SSNMR experiments and DFTbased chemical shift calculations to elucidate the supramolecular structure of
benzoxazine oligomers. About 30 kHz spinning 1H MAS spectra were compared
with DFT-based calculation to assist with assignments. Good agreement was
achieved. N H protons were found to be spatially separated based on 2D
1
H 1H double quantum (DQ) spectra of dimers and tetramers, but not trimers
(Fig. 37), consistent with the proposed conformational model. The 15N 1H
distances were also measured on 15N-labeled sample using pulse-field gradient-enhanced REPT-HDOR (313) experiment.
Helluy and Sebald studied structure and dynamic properties of solid
13
L -tyrosine-ethylester. Variable temperature spectra on the
C natural
abundance sample showed considerable dynamics attributed to the p-flips of
the phenyl ring. Zero quantum 13C recoupling experiments on fully labeled
sample afforded distance constraints that faithfully reproduced the molecular
structure defined by short-range interactions but failed to uniquely characterize
the complete conformation defined by longer range interactions.
Zolek et al. (314) recorded 13C CPMAS spectra of a series of 11 coumarin
samples and, in combination with GIAO-CHF calculations, characterized their
conformations. Single resonance was found for each carbon of each compound.
Assignments were done based on combination of liquid spectra and interrupted
decoupling editing and were confirmed by the ab initio calculations. The calculations were performed for isolated molecules on their equilibrium geometries.
Figure 37 The 2D 1H DQ spectra of benzoxazine oligomers. The double letter marked

peaks in the double quantum domain describe connectivities between the peaks in the SQ
domain. Source: Modified from Ref. 312.
SSNMR Spectrometry
455
When available, structures from single-crystal data were used. A general relationship between calculated shielding constants and experimental chemical shifts
was developed for the coumarin series.
Mao et al. (146) studied biosolids-derived biocolloidal organic matter by
TOSS-CPMAS, 2D HETCOR, and SUPER techniques. Two 2D HETCOR
spectra were compared, one with Lee-Goldburg CP and one with standard
Hartmann-Hahn CP. The LGCP showed correlations of protons and carbons
separated by three or less bonds, whereas the HHCP showed correlations on
scale. Chemical shifts anisotropies were measured with SUPER technique
6A
to help to distinguish carbons from esters, amides, and carboxylates (144).
Harper et al. (315) compared the measured and computed chemical shift
tensor principal values for all diastereomers of ambuic acid to predict the
relative stereochemistry. No crystal structure was available for this compound.
The tensor principal values were measured using the FIREMAT method. Eight
model structures were prepared for the computation, differing in relative stereochemistry, hydrogen bonding, and conformations. Only tensors of carbons
directly affected by the structural changes were considered. A dimeric structure
was found to fit the experimental data best. This method of predicting a relative
stereochemistry is of general applicability and especially useful for characterizing materials for which it is difficult to grow a good quality single crystal.
Olsen et al. (293,316,317) used a through-bond 13C 13C correlations to refine
dynamic regions in the crystal structures of naturally abundant vitamin D3. A scalar
coupling-driven UC2QF COSY (double quantum correlation spectroscopy) developed by authors was used to completely assign the 54 carbon peaks of vitamin-D3
(27 carbons per molecule and two molecules per asymmetric unit; Fig. 38). This
required two weeks of acquisition on 500 MHz instrument. The ab-initio calculated
chemical shifts using B3LYP functional and 6-311 G(2d) basis were compared
with the assigned shifts. A good agreement was observed.
Sun and Oldfield (318) showed that the structural constraints can be
obtained based on SSNMR carbon chemical shifts. This was demonstrated
experimentally by assigning a series of tryptophan-containing peptides with
known X-ray structures, and comparing the experimental shifts with quantum
chemical calculations. Nonprotonated aromatic carbons correlated well with
the side-chain torsion angles.
Mehta et al. (319) studied the structure of alanine dipeptide in solution and
the solid state. The 13C chemical shift of alanine dipeptide in polycrystalline,
lyophilized, and solvated state was compared. The structure was determined
using DQNMR. The same secondary structure was observed with lyophilized
form of H-bonding (H2O) and non-H-bonding (CHCl3) solvents.
Giavani et al. (320) devised experimental strategies for collecting 14N MAS
spectra of some amino acids exhibiting quadrupole coupling constant in the order
of 1.2 MHz. At spinning frequency of 6 kHz, the spectra consisting of approximately 600 spinning sidebands (spectral width in access of 1 MHz) were fitted
to extract the quadrupolar parameters. For some amino acids two different
456
Medek
Figure 38 The 2D uniform-sign cross-peak double-quantum correlation spectroscopy

(UC2QF COSY) on natural abundance level vitamin D3. Source: Modified from Ref. 316.
14
N spinning manifolds were observed in accord with two sites per asymmetric
unit predicted from X-ray structures.
Quantification Issues
The NMR has a great advantage over other techniques when it comes to quantification. Because NMR signal is directly proportional to the number of spins present
in the sample, no external reference needs to be applied to calibrate the system
response. In this sense, NMR is an absolute technique. However, in most pharmaceutical applications, only relative concentrations of mixtures of two or more
polymorphs need to be determined even in complex dosage forms. Two cases
should be distinguished: (i) direct polarization using one pulse-acquire sequence,
and (ii) CPMAS-type of experiments. Direct polarization experiments require no
special precautions, but for the sensitivity reasons, they typically apply only to
such high abundance nuclei as 1H, 19F, or 31P. Due to the CP step, the CPMAS
signal intensities are not necessarily quantitative and some precautions discussed
next have to be exercised in order to quantify accurately. In general, the basic
requirement for successful quantification by SSNMR is that the NMR signals corresponding to the two or more components of the sample have to be discriminated
SSNMR Spectrometry
457
according to one or more of their NMR parameters. An existence of differences in

the chemical shifts (isotropic and anisotropic) of the components is the obvious
differentiation. A good peak resolution simplifies the quantification substantially.
A chemometric approach to peak deconvolution can be applied if partial peak
overlap is present. Even in the case of totally overlapping resonances, differences
in other NMR parameters, such as the dipolar coupling, J-coupling, or relaxation
properties, could frequently be used to selectively filter in or out the signals from
single components and the sample could be still quantified successfully. In particular, relaxation-based filters can turn out to be instrumental for identifying
and quantifying the presence of an amorphous phase. Many pharmaceutical
applications involve the quantification of an amorphous phase in the presence
of a crystalline phase. Even though, amorphous spectra could show differences
in chemical shifts when compared with their crystalline counterparts, it may
still be difficult to precisely quantify small amounts of the amorphous phase.
The amorphous peaks present in low concentration can be lost in the baseline
due to their broad line character. Because the mobility of amorphous and crystalline phases might differ substantially, the mobility-induced relaxation differences
allow for the relaxation filters to be employed (321,322). Another type of a filter
can be employed specifically to discriminate amorphous phases. It is based on the
fact that the glass transition temperature (Tg) of amorphous compounds is likely to
be lower than the melting temperature of the crystalline phase. If CPMAS detection scheme is used, heating the sample above its Tg results in observation of only
the crystalline phase. The signal from the amorphous phase is lost, because this
phase reaches a liquidlike mobility, and therefore the dipolar coupling, which
the CPMAS experiment relies on, is averaged out to zero. This may, however,
be an invasive technique not suitable for all applications.
The easiest and usually the most accurate quantification of polymorphs is
achieved when a sensitive nucleus, such as fluorine or phosphorus, is present
in the molecule, and its spectrum shows good chemical shift differences
between polymorphs. Direct polarization experiments of these nuclei are sensitive, and therefore the polymorphs can be quantified directly from the integrated
peak ratios. However, attention must be paid to let the magnetization return fully
to equilibrium between the consecutive scans. Good practice for an unknown
sample is to run first a T1 relaxation experiment to set the proper recycle
delay. The limit of quantification when using sensitive nuclei can be well
below 1% of one form in the presence of other forms. In the case of well-resolved
peaks and fast relaxation, LOD of 0.01% is not unthinkable. As an added advantage, the spectra of these nuclei are usually very simple to analyze, because only a
very limited number of resonances are typically present.
When no abundant nuclei (other than protons) are present in the sample,
carbon-based SSNMR quantification is often the only remaining option. One
might still choose to use a direct polarization acquisition. However, due to the
long carbon T1C relaxation times of typical crystalline organics, very long acquisition times will result. In spite of this, many literature examples of quantification
458
Medek
using carbon direct polarization exist and are reviewed next. An alternative way
to partially circumvent this issue of long experimental time is to calibrate the T1C
values on pure standards of the different forms in the mixture to be quantified.
The intensities can then be corrected for incomplete relaxation without waiting
for the full intensity equilibration (dictated by the slowest relaxing component)
between successive scans. One has to make sure that the sameness of carbon
relaxation times of the standards and unknowns is justified.
If the acquisition times for carbon direct polarization are prohibitive (rather
typical case for pharmaceuticals), depending on the availability of pure form of
standards, the analyst can choose to either construct a calibration graph or map
the kinetics of CP. When pure forms are available, SSNMR spectra of mixtures
with different known concentrations of polymorphs can be acquired, and a calibration curve can be constructed. To diminish the dependence on a precise setting
of the experimental parameters, the calibration curve can be plotted as relative
signal ratio of the two forms rather than as absolute intensities. The advantage
of this approach is that the quantification of formulations is not hampered by
the presence of excipients, assuming none or only very limited overlap of the
excipient peaks with the peaks of interest. The calibration graph constructed
with known API form mixtures can be directly applied for quantification of
formulations. Logarithmic or other transformation of the resulting nonlinear
calibration curves is desirable to optimize the fit. Because only a single spectrum
has to be acquired for every unknown, the calibration curve method is usually
faster than quantification based on mapping the CP kinetics, especially if large
number of unknown samples have to be quantified. Unlike the CP kinetics
approach, the only consideration in selecting the peaks for this analysis is that
they originate from only one component at a time. Any overlap of peaks belonging to the same component (but well resolved from the other component) is tolerated, whereas, strictly, the peaks corresponding to single carbons are required
using the CP kinetics approach. However, it should be noted that the calibration
method automatically assumes the sameness of the samples (the standards and
the unknowns) in terms of the NMR parameters, such as relaxation rates or CP
dynamics. For well-defined crystalline samples this assumption may hold well,
but it may be more troublesome when dealing with amorphous phases. It is
well known that the mobility (and hence all the NMR parameters mentioned
earlier) is a strong function of the relative humidity and the samples thermal
history. Even in the case of crystalline compounds, formulations processing as
milling or compacting can dramatically change the relaxation and other NMR
properties. Hence, the calibration method has to be applied judiciously.
CP Kinetics Approach
The CP is based on the pulse sequence depicted in Figure 12. The polarization
transfer mechanism depends on the existence of strong dipolar coupling
between the two cross-polarizing nuclei. According to Equations (12) and (13),
the effective dipolar coupling, in turn, depends on the distance between the
SSNMR Spectrometry
459
coupled nuclei and the orientation (with respect of the external magnetic field) of
the vector connecting them and on the local dynamics of the molecular segment.
The higher the frequency and the more isotropic character of the motions, the
smaller the effective dipolar coupling is. Beside these internal sample-dependent
factors, external conditions, such as temperature or MAS spinning speed, will
also influence the effective dipolar coupling. The best cross-polarizing efficiencies are therefore achieved on rigid and static (nonspinning) samples. It also
follows, that within the same molecule, different nuclei may cross-polarize in
different rates, rendering the resulting peak intensities nonstochiometric. When
a quick and only an approximate estimate of the polymorph content is sufficient,
the intensity ratio of two corresponding carbon peaks, each belonging to the
different form, can be used for quantification. Any additional knowledge of the
system could prove useful in selecting the peaks, as the selected carbons
should have comparable conformation and dynamics. For true polymorphs, this
may be a reasonable assumption. The CP kinetics is then likely to be similar,
reflecting the approximate quantitative ratio of the two forms.
In order to use CPMAS spectra for any accurate quantification purposes,
one has to characterize the dynamics of the CP process. Examples of using
CPMAS for quantification while ignoring the CP kinetics still abound in the literature. The dynamics of the CP was recently described in two excellent and
exhaustive reviews describing cross-polarizing from abundant to rare nuclei
(60) and CP between two abundant nuclei (59,323). An earlier detailed discussion
of CP in terms of quantification was also described by Sullivan and Maciel (324)
and by Harris (325). In the context of the quantification, only the abundantrare
CP kinetics is considered in this section, because in the case of the abundant
abundant system, it is assumed that the quantification can be achieved by
direct polarization acquisition without the need for CP. The CP can be described
by one of the two kinetic models: the simplified classical I-S model (Figs. 39
and 40) or the more rigorous I-I -S model (Fig. 41). The thermodynamic property
of spin temperature is well suited to explain the CP process. In the classical
model, the lattice is defined as having an infinite heat capacity and the abundant
spins I have much larger heat capacity than the rare spins S. The spin temperature
TIS
T1I
T1S
Lattice
Figure 39
Classical I-S CP kinetics model.
460
I(t)
Medek
(D)
0.8
0.6
(C)
0.4
(B)
0.2
(A)
0
0
lnI(t)
0.2
10
(D)
0.4
0.6
(C)
0.8
1
(A)
0
(B)
4
6
Contact Time [ms]
10
Figure 40 Calculated dependence of CPMAS intensity on the contact time using regular
intensity scale (top) and logarithmic scale (bottom). Four different equations were used to
generate these curves: (A) Equation 19, (B) Equation 20, (C) Equation 21, and (D)
Equation 22. The parameters used to generate the curves were: I0 1, 1 0.1,
TIS 0.2 ms, T1rS 1 ms and T1rH 10 ms. Note that for typical carbon CPMAS applications (1 0.01 and T1rS . 15 ms) the curves (A), (B) and (C) would be almost exactly
equivalent and the values of 1 0.1 and T1rS 1 ms were chosen to illustrate the differences between Equations 19, 20, and 21. In the logarithmic representation, it is apparent
that extrapolating the linear decaying part of the curve to zero contact time does not reproduce the correct intensity [ln I(0) 0]. When quantifying using the graphical extrapolation, a correction for the finite relaxation rates using the prefactors of Equations 19,
20, and 21 have to be applied. Abbreviation: CPMAS, cross polarization and magic
angle spinning.
Figure 41 More rigorous I-I -S CP kinetics model. Abbreviation: CP, crosspolarization.
SSNMR Spectrometry
461
b is the so-called inverse spin temperature in the rotating frame, that is, the
higher the b, the higher is the polarization difference between the energy
levels, corresponding to stronger NMR signal. Just before CP, the abundant
spin inverse temperature b0I is high, whereas the rare spin b0S is zero. The
Hartman-Hahn match brings the two spin reservoirs into thermal contact,
allowing the energy to flow from abundant to rare spins, building up the rare
spin polarization with the rate proportional to the inverse of the CP time constant
TIS. This process continues until equilibrium is reached, when the spin temperatures of both reservoirs equalize. However, simultaneously with the CP process,
the inverse spin temperatures of both I and S reservoirs are cooling down due
to their contact with the lattice brought about by relaxation mechanism. The
resulting rate of signal loss is proportional to the inverse of the rotating frame
relaxation times T1rS and T1rI. The following differential equations describe
the overall CP process.
d bI
b bS
b
1a2 I
I
dt
TIS
T1rI
d bS bI bS
bS

dt
TIS
T1rS
(18)
where 1 is the spin population ratio (1 NS/NI), and a is the Hartman-Hanh

matching coefficient (a gSB1S/gIB1I 1). It follows that for a rare abundant
spin system, 1 is close to zero. Since the detected NMR CP signal I(t) is
proportional to the inverse temperatures, it is described by the general solution
of Equation (18) as
I0
(ea t=TIS ea t=TIS )
a a
r!
d
a c 1 + 1 2
c

TIS
TIS
TIS
d
1
1 a2
T1rI
T1rS
T1rS

1
TIS
TIS
c 1 1 a2
2
T1rS T1rS
I(t)
(19)
This general equation is rather complicated and for the majority of

practical applications involving CP to rare nuclei, it can be simplified using
1 a2 0
I(t)

I0
et=I1rI et(1=TIS 1=T1rS )
TIS
TIS
1

T1rS T1rI
(20)
462
Medek
Because the relaxation of rare nuclei is typically slow due to the lack of relaxation
pathways, further approximation can be made (TIS/T1rS 0) leading to
I(t)
t=T1rI

I0
et=TIS
TIS e
1 T 1rI
(21)
The typical double exponential character of the CP kinetics is clearly evident

from the last two equations. According to Equation (21), the signal grows with
the rate proportional to inverse of TIS and decays with the rate proportional to
inverse of T1rI. It should be noted that there is a prefactor in both Equations
(20) and (21). Instead of fitting the experimental curves to one of these equations,
several literature studies suggested extracting the intensity I0 by graphical means,
by plotting the intensities in the logarithmic scale and extrapolating the linear
portion of the curve to zero contact time. A care has to be taken, because I0
scaled up by the factor of (1 2 TIS/T1rI) is actually obtained in this way.
If, in addition, the relaxation of the abundant spins is very slow with respect
to the CP time constant (TIS/T1rI 0), the Equation (21) further simplifies to
I(t) I0 (1 e1=TIS )
(22)
The last equation predicts that, in the absence of any significant relaxation, the CP
signal grows exponentially to reach the I0 intensity for long enough crosspolarizing times. Many of the small organic molecule crystalline compounds
fall into the category with relatively short T1rI, and therefore either Equation
(21) or even Equation (20) have to be used. It should be noted that the rotating
frame relaxation times strongly depend on the strength of the B1 irradiation
field. In the slow motional regime, the stronger the irradiation, the longer the
rotating frame relaxation is.
The asterisk in the I-I -S model (Figs. 41 and 42) denotes the physical
proximity of the I spin (the abundant spin involved directly in the CP transfer)
to the S spin. This model was designed to account for the oscillatory CP kinetics
behavior found in many samples. It takes into consideration the fact that the spin
temperature of all I spins might not be equivalent, that is, the spin diffusion
between the abundant nuclei might be slow compared with the CP kinetics.
The I S spins can be viewed as isolated spin pairs involved in cross-polarizing
magnetization in the oscillatory manner, while the spin diffusion to the remaining
surrounding abundant nuclei dampens these oscillations. For an isolated SIn
cluster, the cross-polarizing kinetics predicted by this model is
I(t) I0 e1=T1r1
h
i
2
2
1 l et=Tdf (1 l) e3t=2Tdf et =2T2
(23)
where l 1/(n 1), Tdf is a spin diffusion constant, and the oscillatory behavior
is replaced by constant T2, characterizing the initial decay of the S magnetization.
Unlike the classical model predicting a two-stage process (signal growth
SSNMR Spectrometry
I(t)
463
1
(D)
0.8
(C)
0.6
(B)
(A)
0.4
0.2
0
0.5
1.0
1.5
2
Contact Time [ms]
2.5
3.0
Figure 42 Calculated dependence of CPMAS intensity on the contact time using the
I-I -S CP kinetics model [Equation (23)]. The parameters used to generate the curves
were: I0 1 and T1rH 10 ms (all curves); (A) l 0.5, T2 0.2 ms, Tdf 5 ms; (B)
l 0.5, T2 0.2 ms, Tdf 1 ms; (C) l 0.3, T2 0.1 ms, Tdf 1 ms; (D) l 0.2,
and T2 0.1 ms, Tdf 0.2 ms. Note that the initial slope is controlled mostly by the parameter T2, the maximum achieved after the initial raise by parameter l and the double exponential character of the raising part of the curve by parameter Tdf. Abbreviations: CPMAS,
cross polarization and magic angle spinning; CP, cross-polarization.
followed by decay), the I-I -S model predicts the CP kinetics to proceed in three
stages: fast signal growth with possible oscillations, slow growth, and decay.
The I-I -S model applies when the heteronuclear I -S dipolar coupling is
sufficiently strong as compared with the homonuclear I I interaction. Samples
exhibiting substantial degree of internal mobility comply with the classical
model. In more rigid organic molecules, the I-I -S model typically applies to
CP kinetics of CH and CH2 groups. Even for rigid compounds, the quarternary
groups and fast rotating CH3 groups usually conform with the classical model.
To discriminate between the kinetic CP models, the early part of the CP kinetics
curve must be accurately sampled to detect any oscillatory behavior or sudden
intensity changes characteristic of the I-I -S model. An example of fitting the
CP kinetics of CAV11 molecule using Equation (23) is given in Figure 43.
Carbonyl and methyl carbons were fitted using this model.
For accurate quantitative results, the CP kinetics should be mapped for
every unknown sample. Clearly, this is a very time consuming proposition,
especially for samples exhibiting long proton T1H relaxation times or for formulations with low API loads. The alternative is to calibrate the two (or more) CP
characteristic constants, T1rH and TCH, on pure standards of all polymorphs
present in the mixture and then use the known CP parameters to quantify the
unknown mixtures. However, this scheme suffers the same insufficiencies
discussed earlier in the context of the quantification by a calibration graph,
namely the potential differences between relaxation and CP kinetic properties
of the standards and the unknowns.
464
Medek
Figure 43 An example of fitting the CP kinetics of CAV11 molecule to Equation (23).

Abbreviation: CP, Cross-polarization. Source: Modified from Ref. 496.
Quantification of Pharmaceutical Materials

When quantifying relative amounts of anhydrous carbamazepine and carbamazepine dihydrate, Suryanarayanan and Wiedmann combined the CPMAS approach
together with T1 filtering and an internal standard addition (326). The spectra of
both anhydrate and dihydrate were practically identical, but the anhydrous carbamazepine showed very long T1H relaxation. Setting the recycle delay to 10 seconds
resulted in saturation of the anhydrous signal and acquisition of only the dihydrate. The authors used addition of known amount of glycine as an internal standard. By variable contact time CPMAS experiment, they determined TCH and
T1rH for both glycine and the dihydrate. Based on calibration mixture containing
variable ratios of anhydrate and dihydrate with fixed amount of glycine (Fig. 44),
they constructed calibration graph of peak area ratio (dihydrate/glycine) versus
the mole fraction of dihydrate. This was done by choosing a single contact
time (optimized for sensitivity) and correcting the intensities of both glycine
and dihydrate peaks by TCH and T1rH values.
Likar et al. (327) studied keto-enol equilibrium of trospectomycin sulfate
bulk drug by 13C CPMAS. Drying the bulk drug or formulation to low water
levels dehydrated trospectomycin sulfate from the diol to the keto form. Variable
contact time experiment was used to select the optimum contact time suitable for
the quantification of the two forms. The quantification of the keto form was
accomplished by monitoring the ratio of the carbonyl (keto form only) to the
methyl group peaks (common for both forms).
Gao quantified three different delavirdine mesylate polymorphs and pseudopolymorphs by 13C CPMAS (Fig. 45, top) (328). The CP dynamics was fully
considered in the quantification as the data from the variable contact time experiment were fitted according to Equation (21) taking into account T1rH and TCH
SSNMR Spectrometry
465
Figure 44 The CP kinetics of carbazepine dihydrate, Form-II, and that of a-carbon and
carbonyl carbon of glycine (left). 13C spectra of mixtures of anhydrate and dihydrate with
constant content of glycine. Zero percent of dihydrate corresponds to 100% of anhydrate.
Abbreviation: CP, cross-polarization. Source: Modified from Ref. 326.
(Fig. 45, bottom). An empirical detection limit of Form-VIII in Form-XI was

found to be 2%. A discussion of the respective merits of using peak areas and
peak intensities (heights) is provided. For forms with partially overlapped
peaks, a difference spectrum (mixturepure major form) was used to obtain
the peak areas of the minor form. Sources of the potential experimental error
were identified as (i) poor S/N, (ii) improper phase adjustments, and (iii) improper baseline correction. Peak heights were found to be less sensitive to the phase
adjustments, but the quantification using peak heights were found to be less
precise. In the follow-up study, Gao further characterized these three crystalline
forms in addition to an amorphous form (329). The assignments of the carbon resonances were performed by combination of solution data, interrupted decoupling,
and cross-polarization and polarization inversion (CPPI) experiments. Form-XII
was found to consist of two molecules per asymmetric unit. Carbon shifts of
methyl groups were found to be sensitive indicators of the intra-molecular geometries. T1H, T1rH, and T1C relaxation times were also measured. A decrease in
T1rH values correlated with increased disorder, which is rationalized by the
dependence of the spin diffusion on the proximity of the neighboring protons.
T1C relaxation times of methyl groups (1 second) were found to be approximately two orders of magnitude shorter than those of other carbons (100
seconds), consistent with the fast rotation of the methyl groups. The rank order
466
Medek
Figure 45 The 13C CPMAS spectra of three forms of delavirdine mesylate (top). The CP
kinetics of these three forms are also shown (bottom). Abbreviations: CPMAS, crosspolarization and magic angle spinning; CP, cross-polarization. Source: Modified from
Ref. 328.
of relaxation times T1rH and T1C, but not T1H for nonmethyl groups was consistent. T1H rank order was, however, consistent with that of methyl group T1C,
confirming the role of the methyl groups as T1H relaxation sinks.
Gustafsson et al. (330) compared SSNMR and isothermal microcalorimetry
to quantify the amorphous component (the degree of disorder) of lactose. Two principal least squares (PLS) models established from the NMR data, one for the whole
range of concentrations 0100% and the second one for 010%, accurately determined the amorphous component even when in mixture with acetylsalysilic acid
SSNMR Spectrometry
467
Figure 46 The 13C CPMAS spectra of crystalline and amorphous lastose (left). PLS
correlation plot for amorphous lactose (right), including a validation mixture containing
acetylsalycilic acid (ASA). Abbreviations: CPMAS, cross polarization and magic angle
spinning; PLS, principal least squares models. Source: Modified from Ref. 330.
(Fig. 46). The results from microcalorimetry and SSNMR were in agreement. Both
techniques were able to detect lower than 0.5% amorphous component.
Lee et al. (331) applied 13C CPMAS SSNMR to quantify impurities
and excipients in ecstasy tablets. Chemical shift assignments were based
on comparison with solution values in combination with spectral-editing
experiments. Smaller than 2 ppm solution solid chemical shift differences
were observed. Various CP contact times were screened to find the optimal parameters affording quantitative data. Some of the excipients (lactose and sucrose)
were found not to be fully relaxed even when 60-second recycle delay was
applied.
Paris et al. (332,333) has used a line shape decomposition of 13C CPMAS
spectra to quantify the types of the glycosidic linkages in treated amorphous
starchy substrates. The 2D WISE spectra were recorded to investigate the mobility. Variable contact time curves for all the deconvoluted peaks were found to fit
the two proton reservoirs CP model [Equation (23)]. The extracted parameters of
Tdf and T1rH are correlated with the hydration level and to the local structure
(domains larger than 1 nm are found in one of the structures). However, no
attempt is shown to use the measured CP dynamics for the actual quantification
of the types of the glycosidic linkages.
468
Medek
Kono et al. (334 336) used quantitative 13C CPMAS to help in assignment
of two native cellulose allomorphs (a and b). The assignment was achieved by
using selectively labeled cellulose biosynthesized by two different routes. To
determine the precise transition of the 13C labeling to each position of cellulose,
quantitative measurements were performed. A quantitative contact time was
chosen based on variable contact time experiment.
Tozuka et al. (337) quantified two clarithromycin polymorphs by PXRD
and SSNMR. A calibration curve method was used, plotting the carbonyl peak
area ratio of one form with respect to the known content of this form.
Vickery et al. (338) quantified two crystalline polymorphs of roxifiban. A
calibration graph method was also used. The detection limit was estimated to be
9% of Form-II in Form-I.
Apperley et al. (339) quantified mixtures of anhydrate and dihydrate of formoterol fumarate by 13C CPMAS. One molecule per asymmetric unit cell was
found. The protonated carbons of one of the two phenyl rings were not observed
at room temperature but gave an observable signal at 2508C that was consistent
with the fast rotation of the phenyl ring at room temperature. Differences between
some solution versus solid chemical shifts were justified by forming hydrogen
bonding in the solid state. The experimental limit of detection in lactose was
found to be 0.45% (15 hours acquisition). Dipolar dephasing experiment was
chosen to minimize the intensities of lactose and its spinning sidebands. To facilitate the quantification, T1H was measured indirectly through CPMAS by proton
saturation recovery, and TCH and T1rH were estimated from variable contact
time experiment. Similar CP dynamics allowed for using the intensity of the carbonyl signal of the two forms quantitatively at one contact time. A correction was
applied to account for different T1H relaxation rates.
Fu et al. (340) devised a new LG-based CP approach affording nearly quantitative intensities across the CPMAS spectra. During the contact time, proton
magnetization is spin-locked by the LG sequence and the irradiation of the insensitive spins (15N or 13C) is modulated sinusoidally and with a constant rf
amplitude. LG sequence lengthens the T1rH spin locked at the magic angle and
the frequency modulation shortens the CP time for nonprotonated spins, while
it has only limited effect on the protonated spins. Longer CP times can be used
to afford a uniform CP signal for both protonated and non-protonated spins,
making quantitative CP measurements possible. The approach is illustrated on
lyophilized sample of 15N-labeled L -histidine (Fig. 47). The two tautomeric
forms of L -histidine (nonprotonated and protonated) are observed in the 15N spectrum. T1rH increased from 18 ms to 37 ms when spin locked using LG. The I-I -S
model (Fig. 41) was used to evaluate the CP dynamics for protonated nitrogens,
because the LG sequence attenuated the 1H 1H dipolar coupling, and therefore
made the 1H 15N dipolar coupling stand out.
Byard et al. (341) quantified a presence of an amorphous component in
micronized pharmaceutical development drug substance by SSNMR and
PRXD. The limit of detection of PRXD for the amorphous component was
SSNMR Spectrometry
469
Figure 47 Differences between the standard continuous wave CWCP, continuous wave
with frequency modulation CW-FMCP and Lee-Goldburg frequency-modulated LGFMCP 1H 15N cross-polarization kinetics curves on L -Histidine. The solid circles and
squares represent the nonprotonated and protonated forms of histidine, respectively.
Based on direct polarization 15N experiment, the ratio of nonprotonated to protonated
intensities should be 1.94. Note that only LG-FMCP with contact time of 8 ms approximately reproduces this intensity ratio. Source: Modified from Ref. 340.
found to be .5% with an error of +3%. NMR was used to quantify crystal
defects at levels .3% with error +2%. The defects arose due to the conformational differences attributed to the removal of ethanol and had only small
effects on the crystal lattice and, hence, on PXRD spectra. Excellent linearity
is demonstrated using standards. The stability study of the drug at stressed conditions showed formation of crystalline component. NMR is suggested to be used
as a complementary technique alongside of PXRD.
Offerdahl et al. (69) quantified three crystalline and amorphous forms of
anhydrous neotame by 13C CPMAS. When in binary mixture, the detection
limit for the crystalline forms was 1 2%. When mixed with amorphous phase,
the detection limit of the amorphous phase was 5%. When quantifying using
CPMAS, the authors rightfully stress the importance of recording the contact
time dependence on each sample, as opposed to just standards. This is due to
the possible differences in the CP and relaxation parameters of each new sample.
The importance of long T1rH (relatively to TCH) is also discussed. For example, for
contact time of 2 ms and T1rH of the amorphous component of 11 ms, the signal
has already relaxed by 25%. The differences between the T1rH of amorphous
and crystalline components were used to partially filter out the crystalline
one. Two intermittent acquisitions, one with standard contact time and the
second one with long contact time, were subtracted by altering the phase of
the receiver by 1808 (Fig. 48). Even though a large penalty in S/N is incurred,
the increased relative ratio of the amorphous component allows for better
deconvolution.
Quantification of Natural Organic Matter, Wood, and Tar
This section on quantification of natural organic matter, wood, and tar is included
due to the similarities of the NMR techniques used and the availability of only
470
Medek
Figure 48 CP dynamics of different forms of neotame (left). 13C CPMAS spectra of a

mixture of small amounts of amorphous component (11%) in the crystalline form
G. Standard CPMAS spectrum (right, top) and dual contact time (2 ms and 8 ms) spectrum
(right, bottom). The deconvoluted peak areas (180 for crystalline and 100 for amorphous
forms) have to be corrected by the CP dynamics to get the correct quantification. Abbreviations: CPMAS, cross polarization and magic angle spinning; CP, cross-polarization.
Source: Modified from Ref. 69.
unlabeled materials. In many respects, the quantification of these materials is

actually more challenging due to their multicomponent nature, where some of
the components may show very fast relaxation and hence, do not cross-polarize
efficiently.
Mao et al. (321) used direct polarization quantification corrected for incomplete relaxation. To avoid the time-consuming carbon relaxation experiment,
they obtained the correction factor for relaxation in a clever way by comparing
two simple 1D CPMAS experiments: The first one was a standard CPMAS,
and the second one was CPMAS with T1-filter placed before the carbon acquisition. The T1-filter length was of the same duration as the recycle delay in the
direct polarization experiment. These two CPMAS measurement allowed for
correcting the direct polarization intensities.
Smernik and Oades (342) used spin counting to quantify natural organic
matter. A detailed discussion of CP dynamics is provided. A number of materials
including cellulose, pectin, lignin, and palmitic acid afforded quantitative spectra
using both direct polarization and CPMAS. Some materials, including chitin, collagen, and HF-treated soil, gave better results with direct polarization. Both
experiments performed poorly to quantify the charcoal, commercial humic
acid, and whole soil. Signal losses in the CPMAS spectra were attributed to
rapid relaxation rates (T1rH) due to paramagnetic species and slow CP magnetization build-up rates. The same authors pointed out an existence of carbon background signal problems when quantifying soil organic matter (SOM) (343). The
background signal was especially significant when using direct polarization
SSNMR Spectrometry
471
acquisition. They attributed this background to the Kel-F material of the probe
head. They suggested a pulse sequence, which suppresses the Kel-F signal by
acquisition of spin echo while decoupling on proton, but not fluorine channels.
Due to the strong 19F 13C coupling, the carbon signal originating from Kel-F
is fully dephased before the start of the acquisition.
Direct polarization MAS technique was found to be more suitable as
compared with CPMAS when quantifying tar pitches (344). Variable contact
time experiments indicated large contact time-dependent signal differences for
different carbons.
Smernik et al. (345) evaluated a performance of three different techniques
to determine T1rH: variable contact time (VCT), variable spin-lock (VSL) followed by CP and 1H observed experiments. The authors discuss the consequences
for quantification of charred and uncharred wood. VCT is shown to overestimate
the T1rH, mainly due to the presence of slowly cross-polarizing carbons. VSL was
found to be better performing when analyzing the rapidly relaxing components.
Directly proton-observed T1rH detected even faster relaxing components that
were virtually undetectable by the indirect VCT and VSL techniques. The
same group also studied impact of the remote protonation on CPMAS quantification of charred and uncharred wood (346). They used simultaneous fitting of
the VCT and VSL curves to minimize the effect of slow and fast T1rH on the
quantification. Spin-accounting term is proposed to account for all the wood
carbon species by correcting the differential intensities by known CP parameters
when quantifying at one preselected optimal CP time. It improved on the established spin-counting technique by correcting for the rapid T1rH relaxation and
inefficient CP. Another editing technique called RESTORE (restoration of
spectra via TCH and T1r editing), generated three sub-spectra (well-represented
CPMAS signal, under-represented signal due to rapid T1rH relaxation, and
under-represented signal due to inefficient CP) based on the differences in the
T1rH and TCH (347,348). Acquisition of three CP spectra with different combinations of spin lock and contact time is required. RESTORE can identify the
types of structures underestimated in CPMAS and the cause for their underestimation. Rapid T1rH mostly affected carbonyl and carbohydrate carbons, whereas
inefficient CP mostly affected aromatic carbons. Smernik (349) also studied an
effect of the magnetic field strength on the quantification (200 and 400 MHz).
No improvement in spectral resolution was found at the higher field, but the spinning speed had to be doubled to keep the same spacing of the sidebands and the
CP efficiency was reduced by 35 40%.
Similar analysis was applied by Bardet et al. (350) for quantitative characterization of an archeological wood. VCT CPMAS experiment was used to
characterize the parameters of the CP and to choose an optimum contact time
for further quantitative analysis of unknowns. The CPMAS intensities were
then subjected to corrections to render quantitative results.
Mao et al. (351) studied suitability of five different 13C SSNMR techniques
to characterize humic acids of various origins and locations. The techniques,
472
Medek
performance of which was compared at different magnetic field strengths,

included direct polarization at 13 kHz spinning speed, a conventional CPMAS
at 5 kHz, ramped CPMAS at 8 kHz, TOSS-CPMAS at 4.5 kHz, and direct polarization corrected by a factor obtained from TOSS-CPMAS-detected T1C relaxation. The ramped CPMAS and TOSS-CPMAS gave consistently better results
than those of the conventional CPMAS.
Direct polarization was also the preferred choice of Keeler and Maciel
(352) for quantifying soil and organic soil fractions. An internal calibration
sample of silicone rubber, placed in a capillary along the rotational axis of the
spinner, was used and a correction due to different T1C applied. Internal standard
was also used for CPMAS characterization. Based on the measured values of the
CPMAS parameters (T1H, T1rH, and TCH), corrected intensities were used for the
quantification. Paramagnetic components were identified to be largely responsible for the invisible carbon portion when CPMAS was used. The paramagnetic components shorten the T1rH of some proton, rendering the CP inefficient.
Van Lagen and de Jager (353) used steady state and well-defined data processing to improve CPMAS quantification of humic acids. By testing the reproducibility of the variable contact time experiments, the authors found the signal
consistently decreased and leveled only after the first five experiments (total 2500
scans). Achieving the steady-state magnetization was found important for quantification. A well-defined processing procedure including spectra with large spectral widths (to improve baseline), linear prediction, and same phasing correction
are recommended. Authors also recommend extrapolating the decaying part of
the variable contact time experiment to zero time. Similar to a number of other
studies, they seem to omit the necessary correction of the intensities I0 extracted
by this procedure by the relaxation factors of 1 2 TIS/T1rI [Equation (21)]. Nonprotonated carbons are therefore overestimated (342).
Peuravuori et al. (354) assessed a practical accuracy for quantitative determination of natural organic matter (NOM) by carbon SSNMR. Different magnetic fields, MAS speeds, and applications of single and ramped amplitude
CPMAS with TOSS was reviewed. Little difference between single and
ramped CPMAS was found. When accompanied with faster spinning speeds,
better performance was found at higher magnetic fields. A poor spectral repeatability performance and errors contributed by TOSS are discussed. The overall
conclusion reached was that it is still difficult, if not impossible, to make
SSNMR of NOM quantitative. Only qualitative analysis is recommended. On
the contrary, Conte et al. (355) found CPMAS affording accurate representation
of whole carbon content in NOM when compared with elemental analysis.
Khalaf et al. (356) compared sorption domains in molecular weight fractions of a soil humic acid. Integration of the 13C CPMAS spectra of unfractioned
and fractioned humic acids was used to show the relative carbon-type distribution. The 19F MAS was used to characterize the absorptive uptake of hexafluorobenzene (HFB). At least three sorptive sites were observed, each showing
different HFB mobility
SSNMR Spectrometry
473
Figure 49 Failure of the standard-editing techniques using interrupted dipolar coupling

coupled to high MAS spinning speeds (middle two traces). Interrupted decoupling with
inclusion of carbon 1808 pulse (causing recoupling of the 1H 13C dipolar coupling)
properly dephases the CH carbons. Source: Modified from Ref. 357.
Mao and Schmidt-Rohr (357) devised accurate quantitation of aromaticity

and nonprotonated aromatic carbon fraction in NOM. Direct polarization is used
for the quantification. An improved dipolar dephasing sequence for high MAS
spinning speeds is used for spectral editing (Fig. 49). The problem of overlap
between aromatic and alkyl carbon resonances at 90120 ppm is solved by 13C
chemical shift anisotropy filtering technique, selecting sp 3-hybridized carbons.
Because the sp 3-hybridized carbons of OCO groups have smaller chemical
shift anisotropy than the aromatic carbons, inclusion of the 13C chemical shift
anisotropy filter dephased the carbons with larger anisotropies. In addition nonprotonated OCO groups can be distinguished by adding dipolar dephasing step.
Polymorphism
One of the most important contributions of SSNMR in characterization of pharmaceutical solids is its exquisite sensitivity to differentiate the different crystalline
474
Medek
and amorphous forms and, hence, detect the polymorphism. Polymorphs are typically differentiated based on their chemical shift differences (including the
chemical shift anisotropy) but, in principle, differences in any other NMR property, such as dipolar coupling, J-coupling, or relaxation, can be used as well.
A whole separate section describes the various applications of relaxation as
selective filteres or as the mobility probes. Chemical shifts are very sensitive
to molecular conformations and crystal packing. In order for the polymorphs
to be distinguished based on their chemical shifts, they must show differences
in either the conformations or the crystal packing.
If a single crystal of suitable qualities can be grown, the single crystal X-ray
diffraction provides a detailed description of the crystal structure, including molecular conformation and crystal packing. However, even the availability of single
crystal data does not negate the need for additional studies. For example, the
single crystal X-ray data contain only very limited information about the
sample dynamics. SSNMR can fill the voids by probing molecular motions
occurring at very different timescales. Also, pharmaceutical solid formulations
are based on polycrystalline samples (not single crystals), and may actually
contain mixtures of many crystal forms or other impurities. The industry
golden standard for powder characterization remains to be the powder X-ray
diffraction (PXRD). SSNMR and PXRD should be viewed as complementary
techniques, as they take advantage of different phenomena to analyze polymorphism. PXRD is sensitive to differences in unit-cell dimensions, whereas SSNMR is
sensitive to conformational changes. One can think of two purely hypothetical
situations: (i) If two crystal forms differ exclusively in their unit-cell dimensions
but the conformation of the molecule is preserved, the PXRD patterns are likely
to be substantially different, whereas SSNMR might not register any changes. (ii)
If the unit-cell dimensions are preserved while the conformation changes substantially, SSNMR is likely to pick up the differences, whereas the PXRD patterns will be identical. The real cases are always somewhere in between these
two hypothetical situations, but based on the potentially biased authors personal
experience with numerous crystal form studies by both SSNMR and PXRD techniques, at least several cases were encountered for which no differences in
PXRD spectra were observed but showed characteristic SSNMR spectra. No
case has yet been encountered for which the opposite would hold true.
The SSNMR has a number of important advantages over other spectroscopic techniques when it comes to determining polymorphism. One of the
major advantages is its high degree of resolution making it a sensitive probe of
conformation. Even minor changes of conformation or crystal packing between
polymorphs produce different local environments and consequently different
chemical shifts. SSNMR is usually more sensitive for polymorph detection
than other techniques, including PXRD, IR, NIR, or Raman spectroscopies. In
contrast to these more established spectroscopic techniques, SSNMR spectra
are largely void of any dependence on physical sample properties as particle
size, homogeneity, or residual solvent content. Therefore, pharmaceutical
SSNMR Spectrometry
475
solids can be studied by NMR without a need for a special sample preparations.
Note that if severe segregation of a multicomponent sample occurs, the quantification of the components might be effected due to their uneven excitation
(Fig. 11). Samples viable for SSNMR analysis include a whole range of pharmaceutical formulations, such as tablets (typically crashed into powders; however,
spinning of whole tablets has been also shown), lyophilized powders, capsules,
suspensions, ointments, and others. SSNMR does not suffer from the preferred
orientation phenomena, which often lead to an incorrect identification or quantification of polymorphs by PXRD. In addition, SSNMR is a nondestructive technique, which allows other analyses to be performed after the NMR spectrum is
acquired. Another advantage of SSNMR is that the observed chemical shift
differences between polymorphs can be related to particular molecular sites
based on known assignments. The study of polymorphic transitions can be performed by variable temperature experiments. Probably, the most important
advantage of SSNMR for pharmaceutical applications is its suitability for the
analysis of complex formulations. Typically, excipients found in drug formulations resonate in limited spectral regions. In carbon spectra, most excipient
peaks can be found between 60 and 100 ppm and as such do not completely
overlap the active ingredient resonances, whereas fluorine spectra are always
completely void from excipient contributions. Spectral regions containing only
drug signals can often be found even in complex formulations.
A caution has to be exercised when evaluating the number of molecules per
asymmetric unit. Usually the presence of multiple molecules per asymmetric unit
is detected by SSNMR by comparing the number of expected resonances (based
on the structure of the compound) and experimentally observed peaks. If more
than the expected number of peaks is counted, presence of multiple molecules
per asymmetric unit or a physical mixture is evident. However, even if the
total count of the peaks does not exceed the expected number (due to the
overlap of some peaks), the resolved areas of the spectrum can be used for the
comparison, and the excess lines can still be detected. After the possibility of
the presence of a physical mixture of polymorphs is excluded, at least one
more factor has to be checked. The residual dipolar coupling to quadrupolar
14
N or 35,37Cl nuclei, which is not completely averaged out by MAS, can
cause carbon lines to split, typically exhibiting an asymmetric doublet pattern.
Because this second-order effect is indirectly proportional to the magnetic
field, it is the largest at lowest fields, and typically not important for magnetic
field in excess of 11.7T (500 MHz proton frequency). This residual dipolar coupling is a well-known phenomenon and can even be used to extract information
about the structure of the compound (157,358 374). It depends on the quadrupolar coupling parameters of the particular 14N nucleus and on the orientation of the
quadrupolar and dipolar tensors.
If more than the expected number of resonances are observed and the possibility of attributing it to the residual dipolar coupling to quadrupolar nuclei is
eliminated, then the multiple lines should be interpreted as either the evidence
476
Medek
for a mixture of several polymorphs or, conversely, for the presence of multiple
asymmetric sites per unit. There are several ways to differentiate between these
two instances. As a first approximation, when pure form of standards are absent,
one can measure the relative peak intensities of the split resonances. If the peak
ratios of all the resonances appearing as multiplets are approximately equal to
one, multiple sites are likely to be present. This is, however, not a proof of multiple sites because, by coincidence, one could have an equal mixture of different
polymorphs. In some cases, the build-up dynamics of CP, which is a dependence
of CPMAS signal intensity on CP time, can be used to differentiate polymorphs
from multiple sites (54). However, it is not straightforward to analyze the CP
build-up intensity differences between carbons from the same molecular sites
in terms of polymorphs versus multiple sites. Another way to distinguish mixtures from multiple sites is to measure proton T1H or T1rH relaxation. Due to
the efficient proton spin diffusion, the proton relaxation as detected indirectly
on the split resonances will be either the same if multiple molecules
per asymmetric unit are present or different if a mixture of forms is present.
The difference of the proton relaxation values is the proof of a presence of
mixture, but the opposite is not the proof of multiple molecules per asymmetric
unit, because by coincidence, the components of the mixture may relax at the
same rate. A positive proof can be gained by observation of cross-peaks
between the split resonances in a through space correlation experiment. A positive proof for the existence of polymorphs can also be gained by varying the relative concentration of the two forms and observing the varying intensity ratios of
corresponding carbons. This, in turn, requires sample manipulation, such as
varying the temperature, humidity, or other conditions to change the ratio of
polymorphs.
Stephenson et al. (375) studied polymorphic, isomorphic, and solvated
forms of dirithromycin. Three different crystal forms out of the 12 known
forms (two anhydrous forms, nine stoichiometric solvated forms, and one amorphous form) were studied. Carbon chemical shift assignment was achieved by
comparison with solution shifts in combination with interrupted decoupling spectral editing technique. The resonances of the incorporated solvents were observed
in the carbon spectra. More than one molecule per asymmetric unit was found in
some of the forms. Steric hindrance of the solvent molecules was related to the
N,N-dimethylamine peak coalescence in the variable temperature spectra.
Carbon chemical shifts of the lactone carbon were related to the presence or
absence of hydrogen bonding to this group. Combined spectroscopic approach
(PXRD, IR, and SSNMR) was found to be advantageous to characterize crystalline forms in the absence of single crystal X-ray data.
McGeorge et al. (376) studied conformational analysis of three polymorphs
of a disperse azobenzene dyestuff by 13C and 15N SSNMR in an attempt to
understand the effect of polymorphism. Carbon chemical shifts indicated the
possibility of forming an intramolecular hydrogen bond resulting in two conformations coexisting in the crystal structure of each polymorph and differing by
SSNMR Spectrometry
477
internal rotation of a sidechain acetamido group. The conformer interchange can

occur even in the solid state and can be explained by the open nature of the crystals as a result of steric effects of substituents.
Schmidt et al. (377) measured the temperature width of first-order single
crystal to single crystal phase transition of three forms of 2-(2,4-dinitrobenzyl)-3-methylpyridine by 15N (15N enriched samples) and 13C CPMAS. For molecular crystals, the classical first-order phase transition description is claimed to
be too simplistic, because the different crystal phases can exhibit gradual onset of
physical properties, such as heat capacity and volume over broad range of temperatures. This is due to fine microscopic details or nonequilibrium conditions. To
shed light on the poorly understood contributions of the intermolecular potentials
to the thermodynamic functions, the authors studied photoactive crystals by
SSNMR. A comparison with solution, interrupted decoupling, REDOR, and
TEDOR experiments were used for chemical shift assignments of all phases.
Attention has been paid to proper temperature calibration. Temperature hysteresis of the phase transition of 7K was observed. The observations are rationalized
in terms of two different phases having very different timescales. The first process
constitutes a fast redistribution of the mole ratio of the coexisting phases, whereas
the much slower second process involves macroscopic relaxation of the system.
Existence of mesoscopic domains having different surface areas and different
interfacial and strain energy contributions is therefore postulated.
Glasser and Shiftan (378) studied the solvatomorphism of (-)-scopolamine
hydrobromide/hydrochloride salts. Regular CPMAS and spectral-editing experiments, such as dipolar dephasing, CPPI and T1C inversion recovery, were used.
The last experiment was used to filter out the fast relaxing CH3 signals after 30
seconds delay. (T1C of 50 60 seconds was observed on nonmethyl carbons.) The
authors discuss in detail the carbon assignments, the observed line widths, and the
differences between solution and solid carbon shifts. The otherwise commonly
observed rapid topomerization was not observed. A p-flip of aromatics rings
was observed in some of the solvatomorphs, and not in others. Its presence
appeared to be correlated with the void-space above and below the ring. RHF
GIAO-DFT chemical shielding calculations matched the observed chemical
shift differences caused by g-gauche effect. Correlations between torsion
angles and the methylol and carbonyl chemical shifts were found, which could
be utilized as diagnostic markers of the ester conformation. The authors conclude
that SSNMR can be a powerful tool to unravel the contributions to weighted
time-average structures observed in solution.
Apperley et al. (379) studied five polymorphs of sulfathiazole (Forms-I
to -V) by 13C CPMAS. Carbon spectra of all five forms were noticeably different,
making SSNMR an excellent monitoring technique (Fig. 50). A combination of
dipolar dephasing, comparison with solution shifts, and the broadening of the
carbons close to 14N was used to unambiguously assign nearly all the carbon
resonances. Forms I III and IV V were found to have two and one molecule
per asymmetric unit, respectively. The rotation of the aromatic group about the
478
Medek
Figure 50 The 13C CPMAS spectra of five polymorphs of sulfathiazole. Abbreviation:

CPMAS, cross-polarization and magic angle spinning. Source: Modified from Ref. 379.
S C bond was found to be low on the NMR timescale. The complex behavior of
variable temperature spectra of Form-I was consistent with a strong anisotropic
lattice expansion rather than the effect of the internal reorientation. Differences
between chemical shifts of the different forms are discussed in detail and
related to their intra- and intermolecular packing and H-bond patterns.
Dong et al. (380) described crystal structure and physical characterization
of neotame methanol solvate. A single crystal X-ray structure is presented. The
13
C peaks were assigned based on the interrupted decoupling and previous
studies with aspartame (381). Unlike the monohydrate, two crystallographically
non-equivalent sites were observed for the methanol solvate. Padden et al. (381)
compared the relative capabilities of carbon SSNMR and PXRD for analyzing
mixtures of polymorphs of neotame. Altering the crystallization and drying
conditions generated mixtures of solid forms. A systematic study was performed
to observe conversion under vacuum of the monohydrate to a mixture of forms
and then reconversion to the monohydrate upon exposure to moisture under
SSNMR Spectrometry
479
ambient conditions. Up to seven forms were observed on the basis of the presence
of distinct resonances for quaternary phenyl carbon. PXRD did not register any
significant differences. SSNMR indicated a presence of many forms. The insensitivity of the PXRD to detect the different forms is explained by conformational
changes of neotame not being accompanied by changes in the unit-cell parameters. The authors concluded that it is important that PXRD no longer is the
exclusive technique for polymorph screening. Zell et al. (382,383) investigated
polymorphism in aspartame and neotame. Aspartame was found to exist in
three forms at room temperature, two of which had more than three molecules
per asymmetric unit. Different 2D exchange experiments on 13C-labeled aspartame based on spin diffusion and RFDR were run for assignment purposes. For
crystal forms with multiple molecules per asymmetric unit, it was possible to
assign the crystallographically inequivalent carbons. To obtain an optimal resolution on the 3C-labeled aspartame, proton decoupling power combined with
TPPM was found to be more critical than high-spinning speed to average out
the 13C 13C and 13C 1H dipolar couplings. High-spinning speed MAS extending a period of several days was found to covert the neotame forms. PXRD
was found to be insensitive to follow the interconversions. This was again rationalized as mostly the conformational changes, which SSNMR is sensitive to, are
happening without significant changes in the unit-cell parameters. Dong et al.
(384) studied seven neotame anhydrate polymorphs. Carbon SSNMR clearly distinguished between forms A, D, F, and G, the only forms produced in bulk. Form
D was identified as having two molecules per asymmetric unit. Based on the
different 13C chemical shifts between the forms, a conformational polymorphism
is suggested. Dong et al. (385) studied dehydration kinetics of neotame monohydrate. A carbon chemical shift of one of the aromatic neotame carbons was found
to be characteristic of each crystal form and used to differentiate them. SSNMR
data clearly indicated a partial dehydration of neotame monohydrate at 508C to
produce neotame anhydrate Form A.
Sack et al. (386) studied 15N 1H dipolar coupling and 15N chemical shielding in two polymorphs of polyglycine by 1H, 2H, and 15N SSNMR. Partial deuteration eliminated the largest intramolecular 1H 1H dipolar coupling, and it was
possible to get structural information by orienting the dipolar vector in the molecular frame using the chemical shielding.
Lee et al. (387) studied polymorphism of phenylpyruvic acid. Partial
assignments of the 13C CPMAS is given for both Forms-I and -II. The Form-II
was found to partially convert into Form-I upon the NMR analysis. The
authors suspect that the high pressure produced by the spinning (4 kHz), and
the long acquisition times (four hours) speeded up the form of transformation.
Chen et al. (388) characterized three hydrated, one water/methanol, and
one amorphous forms of nedocromil by SSNMR and FTIR. The chemical shift
assignments were accomplished based on combination of solution NMR, interrupted decoupling, and knowledge of crystal structures. More than one molecule
per asymmetric unit was observed in some forms. FTIR and SSNMR are shown
480
Medek
to be powerful complementary tools for probing the chemical environment. FTIR

was found more useful for probing the interactions of water molecules, whereas
SSNMR was superior in probing the interactions and comformations of the nedocromil ion.
Henck et al. (389) studied polymorphism of tedisamil dihydrochloride. The
13
C CPMAS spectra of the three enantiotropically related polymorphs were
obtained at different temperatures: Form-II at 508C (most stable form at
ambient conditions), Form-I at 1908C, and Form-III at 2608C. The carbon
spectra of the three forms were substantially different (Fig. 51). Even though
single-crystal X-ray spectroscopy predicted only one molecule per asymmetric
unit, splitting of most carbons was observed. This is explained by the reduced
symmetry due to the disordered cyclopropyl groups. The authors noted increased
resolution at higher temperatures and the significant temperature dependence of
chemical shifts of some carbons, whereas others were largely unchanged.
Stephenson (390) presents a structure determination based on powder diffraction data obtained from conventional laboratory sources by application of
Monte Carlo/simulated annealing methods. The 13C CPMAS is used to
confirm the predicted structure of two forms of acetohexamide.
Medek and Frydman (391) studied polymorphic forms of vitamin B12
recrystallized from various solvents by 13C, 15N, 31P, and 59Co SSNMR. The
majority of the carbon resonances were assigned based on combination of
solution assignments and solid-phase spectral-editing experiments (interrupted
decoupling, short contact time CP, and INEPT-based experiment; Fig. 52).
Variable temperature 13C indicated reversible phase transformations. Some of
Figure 51 The 13CPMAS solid-state and solution spectra of tedisamil dihydrochloride.

Abbreviation: CPMAS, cross polarization and magic angle spinning. Source: Modified
from Ref. 389.
SSNMR Spectrometry
Figure 52
481
Spectral editing on vitamin B12. Source: Modified from Ref. 391.
the nitrogens were not observed, presumably due to dipolar coupling to the 59Co
nucleus. The parameters of chemical shift and quadrupolar tensors were extracted
by fitting the 59Co static line shapes. The combined knowledge from this multinuclear approach allowed for structural predictions that correlated well with the
known enzymatic role of B12.
Variankaval et al. (392) characterized crystal forms of b-estradiol. SSNMR
was used to monitor the changes in estradiol crystals as a function of temperature.
Temperature-dependent 1D 1H and 2D 1H 13C WISE experiments were utilized
to follow the molecular mobility of estradiol during heating and cooling. The 2D
WISE confirmed that the increased mobility at high temperatures observed in 1D
1
H MAS spectra was ensuing not from the steroidal skeleton but from the water in
the sample. Differences between 13C CPMAS spectra of each of the three forms
(EA, EC, and ED) were observed. The broad nature of the ED peaks was attributed to improper crystallization. The presence of two molecules per asymmetric
unit was found in forms EC and ED.
482
Medek
Wenslow et al. (393) studied polymorphism in an aza-steroid. Remarkably

well-resolved 13C CPMAS peaks of Form-I were observed, with some showing
FWHM of 5 Hz, which was still limited only by the upper limit on acquisition
time due to decoupling cycle constraints. In contrast to Form-I, Form-II showed
multiple peaks per carbon, indicating two molecules per asymmetric unit.
Lee et al. (394) studied 3,4-methylenedioxyamphetamine hydrochloride
and seven related compounds and their mixtures with lactose. Differences
between solution- and solid-phase carbon spectra of the compounds are related
to their crystal structures. The chemical shift differences are attributed to the variations in crystal packing. Lactose monohydrate was added to mimick the illicit
production of Ecstasy tablets. An interaction between the 3,4-methylenedioxy-N-methylamphetamine hydrochloride and lactose monohydrate,
as evidenced by chemical shift changes in several carbons, was found to be
specific for this compound. It is attributed to lattice packing brought about by
thorough dry-mixing with lactose at room temperature. It is concluded that
lactose affects the crystal packing by reducing conformational rigidity, so that
the molecule more closely resembles that in solution.
Crowley et al. (395) characterized mesomorphism of oleic acid and propranolol oleate. Spectral differences were observed upon g-to-a phase transition.
Carbon resonances were assigned by means of interrupted decoupling and
solution data. Variable temperature 13C CPMAS spectra and T1C times were
obtained to study molecular mobility and conformational disorder. T1C relaxation
for the a-phase indicated considerable conformational changes supporting the
a-oleic acid classification as a conformationally disordered crystalline phase.
Low values of T1C times (fast relaxation) of propranolol oleate-I and -II
suggested the absence of a rigid crystalline molecular lattice. The 13C SSNMR
is deemed as a powerful technique for pharmaceutical lipid characterization.
Harper and Grant (396) characterized carbon chemical shift tensors in
polymorphs of verbenol. The verbenol sample was found to be a mixture of
polymorphic materials containing two distinct lattice types. The major component contained three molecules per asymmetric unit, whereas the minor
form contained four molecules per asymmetric unit. FIREMAT technique was
used to correlate the isotropic with anisotropic carbon chemical shifts. The
chemical shift tensorial parameters for each carbon were extracted from the
anisotropic part of the spectrum. Only four different carbons showed the error
bars of chemical shift calculations to be smaller than the predicted chemical
shift differences upon varying the geometries. Verbenol has a rigid bicyclic
structure. The experimental and ab initio computed chemical shift tensors
demonstrated that most of the structural variations arose from differences in
the dihedral angle of the hydroxyl hydrogen (H O C C). The inability to
differentiate the different molecules of the asymmetric unit by NMR (each of
the three peaks for given carbon could not be specifically assigned to a given
molecule in the asymmetric unit) prevented association of the calculated
values to the specific molecule in the asymmetric unit.
SSNMR Spectrometry
483
Strohmeier et al. (397) investigated the polymorphs of dimethyl-3,6dichloro-2,5-dihydroxyterephthalate. From previous X-ray study, the major
structural difference between the white and yellow forms was the torsional
angle between the ester group and the aromatic ring. The conjugation of the
p-electronic system and the H-bond patterns are greatly affected. GIAO-DFT
and HF quantum mechanical calculations (Gaussian 98) were coupled to experimentally derived carbon chemical shift tensorial values (extracted using
FIREMAT technique) and to the carbon detected chlorine electric field gradient
(EFG) values. The 35,37Cl 13C dipolar coupling parameters were extracted
by fitting the variable magnetic field spectra (2.35, 4.7, and 9.4T) spectra
(Fig. 53). Chlorine EFG parameters were taken from previous NQR study.
Strong effect of the 35,37Cl 13C dipolar coupling was observed on the a- and
b-carbon positions (with respect to the chlorine) at the lower fields, whereas
only the a-carbon resonance was split at 9.4T. The conformational differences
between the two polymorphs were found to have a great effect on the electronic
properties of the molecules and, hence, on the chlorine EFG. Differences in the
isotropic chemical shift values of up to 10.3 ppm were observed between the two
forms. The conformational differences were found to be even better characterized
by chemical shift tensor principal values (up to 14 ppm differences) than the
isotropic chemical shifts alone. Good correlation between the experimental and
calculated chemical shift tensorial values was obtained, especially when a
stack of three molecules incorporating the intermolecular H-bonds, rather than
just one molecule, was used for the calculation.
Figure 53 The 13C CPMAS spectra at three different fields. Simulation of the residual
35,37
Cl 13C dipolar coupling is included. Abbreviation: CPMAS, cross polarization and
magic angle spinning. Source: Modified from Ref. 397.
484
Medek
Hanai et al. (398) compared vibrational and NMR spectroscopies to study

cis-cinnamic acid polymorphs. Unlike the cis-588 crystal form, carbon NMR
spectrum of the cis-688 crystal form showed splitting of several carbons, consistent with multiple molecules per asymmetric unit.
Fojud and Jurga (399) studied phase transitions in a series of n-alkylamonium chlorides by variable temperature 13C CPMAS. The temperature of the
transitions depended on the length of the alkyl chain and involved a change in
the motional freedom of the methylene groups.
Yoshinari et al. (400) studied the moisture-induced polymorphic transition
of mannitol and its morphological transformations. The 13C CPMAS spectra of
three polymorphs afforded distinctly different chemical shifts that could be
used for the polymorphic characterization. The d to b polymorphic transition
was confirmed by SSNMR upon exposure of form d to 97% RH. T1H and T1C
relaxation was employed to estimate the molecular mobility characteristic of
each structure. Both carbon and proton relaxation times were, on average,
approximately twice longer in form d. This was consistent with the d-structure
having rigidly linked H-bond network. The authors conclude that water acts as
a molecular loosener by breaking the crystals H-bonds to facilitate d- to bform conversion as a result of multi-nucleation.
Bauer et al. (401) studied the extraordinary example of ritonavirs conformational polymorphism. Only one crystal form was originally identified during
development. Two years later, several lots failed the dissolution tests, because a
new polymorph appeared. During the following several weeks, new Form-II began
to appear in both the bulk drug and formulation areas. Due to the much lower
solubility of Form-II, the formulation was unmanufacturable. The authors are not
very specific about the SSNMR analysis, except for stating that 13C CPMAS and
1
H MAS were used to distinguish the two forms. No spectra are shown.
Liang et al. (402) studied solvent-modulated polymorphism of sodium
stearate crystals. Ternary crystalline dispersions of 10% sodium stearate in mixtures of water and propylene glycol were investigated. The conformation and
mobility of the alkane chains in a- and b-forms were studied by 13C CPMAS.
The rigidity of the alkyl chains was compared for the freshly prepared and
aged samples by measuring T1rH relaxation. Interestingly, longer T1rH values
indicated more mobility. The authors quote a reference of Pursch et al. (403)
in which a similar phenomena on alkyl chains attached to silica (chromatographic
phases) is observed. A greater percentage of trans-chain conformation with
higher rigidity was found in the a-form. Propylene glycol resonances appeared
in the CPMAS spectra of both forms, indicating a strong association with the
sodium stearate.
Giordano et al. (404) studied polymorphism of CHF1035 (Forms-I, -II, and
-III). Comparison with solution NMR was used for carbon assignment purposes.
All forms showed more than one molecule per asymmetric unit. Unlike in
solution, distinct signals for each isobutyryl methyl carbon were observed,
suggesting highly restricted motion in the solid state. Reversible changes were
SSNMR Spectrometry
485
observed in variable temperature spectra of Form-I, suggesting some conformational changes.

Kimura et al. (405) characterized a, b, and g polymorphs of a quinolinone
derivative. Solution data were used to assign the 13C CPMAS spectra of the three
forms. Spectral differences were found between the three polymorphs, suggesting
differences in the molecular conformation and crystal packing. Form-a showed
two molecules per asymmetric unit.
Wiegerinck et al. (406) studied polymorphism of 13C-labeled Org OD14
and Org 30659 drugs during drug processing. Different low API loading tablet
formulations were examined by 13C CPMAS. No form conversion was observed
immediately after formulating and after seven days at 258C/60%RH. Carbon
SSNMR was found to be a powerful tool for the quantitative analysis of polymorphic pharmaceutical compounds in different stages of manufacturing of the
drug product.
Lee et al. (407) characterized paclitaxel in its dihydrate, anhydrous, and
amorphous forms. Significantly different 13C CPMAS spectra of the compounds
studied in terms of peak splitting and peak sharpness were observed.
Smith et al. (408) determined the crystal structure of anhydrous theophylline by PXRD and 13C and 15N SSNMR. SSNMR confirmed existence of only a
single molecule in the crystallographic asymmetric unit. Dipolar dephasing was
used to partially assign both nitrogen and carbon resonances. Nitrogen chemical
shift supported only one type of H-bonds out of the several ab initio calculated
models. This paper is a nice example how to use the synergies of PXRD and
SSNMR for structure determination.
Wang et al. (409) characterized solubility, metastable zone width, and
racemic mixture of the beta-blocker drug propranolol hydrochloride. The structural differences between the (R)- and (S )-enantiomers were investigated by
PXDR, IR, and 13C SSNMR. Carbon spectra of the (S)-enantiomer and the
(R,S)-racemate were markedly different in both chemical shifts and the lineshapes, suggesting that the racemic mixture belongs to the racemic compound.
Zhang et al. (410) studied crystallization and transitions of sulfamerazine
polymorphs. Tentative assignments of 13C CPMAS spectra of Forms-I and -II
were made based on solution data and interrupted decoupling. Multiple sites
per asymmetric unit were observed in Form-I. Peak broadening of some
carbons indicated significant molecular motion. The temperature induced from
transformation is studied by variable temperature 13C CPMAS.
Moynihan and OHare (411) studied monoclinic and orthorhombic forms
of paracetamol. The 13C CPMAS spectra of the monoclinic form and a spectrum
of a mixture of forms prepared from a melt were obtained. A spectrum of the
orthorhombic form was reconstructed as a difference of the two experimental
spectra. Well-resolved differences in chemical shifts of both forms were observed.
Asymmetric doublets were observed for carbons bound to nitrogen due to the
residual 13C 14N dipole dipole interaction not completely averaged out by
MAS.
486
Medek
Garcia et al. (412) showed differences between polymorphism and desmotropy on examples of 3-phenyl- and 5-phenyl-1H-pyrazoles and 3-phenyl-1Hindazole. In addition to polymorphism, the tautomeric compounds can also
present desmotropy. The term desmotropy refers to the crystallization of a
compound in two different tautomers. The compounds were characterized by
15
N and 13C CPMAS. The 15N chemical shifts were found to be more sensitive
to environmental effects, such as hydrogen bonds, than to tautomerism. The
13
C chemical shifts were found to be similar to the solution and constituted a
precise tool to identify and determine the purity of the tautomers. In the two
closely related compounds studied, the authors found a case of desmotropy and
a case of concominant (one pot) polymorphism.
Novoselsky and Glaser (413) studied conformational polymorphism
of sertraline hydrochloride, an antidepressant drug. The carbon isotropic
chemical shift values of Forms-I and -III were correlated with g-gauche effect
resulting from the respective antiperiplanar and synclinal torsion angles. RHF
GIAO calculations of the magnetic shielding tensors were used to support the
conclusions.
Zimmermann et al. (414) studied mesomorphism, isomerization, and
dynamics of a series of pyramidic liquid crystals. Variable temperature and
variable spinning speed carbon NMR was used to study the structure dynamics
and mesomorphic properties of the crown and saddle mesophases of nonaalkanoyloxy tribenzocyclononene. Rotor synchronized carbon 2D exchange
spectra were recorded, identifying the presence of dynamics by the off-diagonal
cross-peaks (415). Chemical exchange (cross-peaks between different sidebands)
and physical reorientation (cross-peaks linking sidebands of the same manifold)
could be distinguished. The spectra of the crown mesophase exhibited dynamic
features consistent with physical reorientation of planar threefold molecular
jumps. In contrast, the saddle isomers did not show dynamic effects.
Potrzebowski et al. (416) devised a new method to distinguish enantiomers
and racemates and to determine an enantiomeric purity. Chiral discrimination can
be achieved through differences in 1H 1H, 1H 13C, and 13C 13C dipolar coupling, chemical shift anisotropies, and 2H quadrupolar splitting. Typically, enantiomers and racemates crystallize in lattices belonging to different point groups
and as such can show differences in the chemical shifts. The authors show that
by using 1D ODESSA, differences between enantiomers and racemates can be
recognized even when no differences in isotropic chemical shifts (magnetically
equivalent nuclei) between the two exists. An advantage is taken of the fact
that the molecular symmetries and packing of enantiomers and racemates are
usually significantly different, giving rise to different angles and distances
between the nuclei and hence the different dipolar coupling. By following the
intensity of the entire 31P spinning sideband set of cyclophosphamide, isophosphamide, and bromophosphamide as a function of the mixing time, the racemates and enantiomers of these compounds were distinguished. Good correlation
with enantiomeric purity is shown.
SSNMR Spectrometry
487
Maurin et al. (417) studied polymorphism of roxifiban, which was found to

exist in two polymorphic forms. No differences between the two forms were
observed by DSC, IR, and Raman spectroscopy. Slight difference was detected
by isothermal microcalorimetry. The PXRD and 13C CPMAS SSNMR spectra
of the two forms were substantially different. Carbon NMR observed lower
symmetry in Form-I by splitting several peaks. In contrast, one molecule per
asymmetric unit was observed for Form-II.
Wang et al. (418) studied polymorphic behavior of NK1 receptor antagonist by, among other techniques, 13C CPMAS. Four anhydrous polymorphic forms
I to IV were discovered. The compound can exist as Forms-I or -II at RT and as
Forms-III and -IV at elevated temperatures. Polymorphic transitions were studied
by variable temperature experiments. At 1158C, Form-I transforms into Form-III,
spectrum of which shows less peaks, indicating less molecules in the asymmetric
unit. At 808C, Form-II transforms into Form-IV. Carbon spectrum of Form-IV
differs from that of Form-III mainly in three peaks, shifted approximately by
0.5 ppm.
Pisklak et al. (419) combined 13C CPMAS and molecular modeling to
study three new derivatives of buspirone. Variable contact time experiments
were recorded to optimize intensities of all carbons. More than one molecule
per asymmetric unit was observed. GIAO-RHF ab initio calculations were
performed to facilitate the resonance assignments. Pisklak et al. (419) studied
coumarine anticoagulants, warfarin, and sintrom, by 13C CPMAS and ab initio
calculations (420). Two enol-ketal tautomers were found to exist for each. The
solution solid chemical shifts differences for the solution minor and major
components were found to be of similar size, and no conclusions as far as the
solid-state configuration and conformation could be reached. Good correlations
between the experimental and GIAO-DFT calculated shifts were established.
Komber et al. (421) studied tautomerism of cyclo-tris(4-R-2,6-pyridylformamidine). Dipolar dephasing was used to aid the assignment of 13C CPMAS
spectra. Three molecules of DMSO were found to be incorporated into the
crystal lattice. Sharp signals from sp 2-hybridized pyridine and imino nitrogens
and a broad signal from the amino nitrogens were observed in 15N CPMAS
spectra. The broadening of NH signals was caused by the residual dipolar coupling with the proton. Signal splitting of carbon and nitrogen resonances indicated
a nonsymmetric structure. No signal coalescence was observed in the 13C and
15
N CPMAS spectra, proving the absence of very low rate of proton exchange
and therefore the existence of only one tautomer in the solid state.
Harper et al. (422) investigated solid-state polymorphism in 10-deacetyl
baccatin III. Carbon chemical shift tensors were measured by using the
FIREMAT technique for the DMSO solvate and the unsolvated forms. Both
forms represented well-defined lattices, as indicated by narrow 13C CPMAS
peaks. The origin of the chemical shift differences between the two forms is
studied by the aid of ab initio calculations. The carbon assignment was done
by combination of spectra editing and comparison with solution and the
488
Medek
calculated shifts. The near invariance of all nonaromatic signals in the variable
temperature spectra suggested that only the benzoyl moiety was mobile. The
authors discussed in detail whether the chemical shift differences between the
forms were caused by lattice or conformation. In general, for molecules with
moderate polarity, the lattice does not contribute significantly to chemical
shifts. Conversely, in charged compounds and polar structures, the inclusion of
lattice is necessary to explain the shifts. The incorporation of the DMSO molecule was found not to cause the shift differences. Only three out of 10
carbons showed evidence for the lattice contribution. The observed tensor differences appeared to arise primarily from conformational variations in ring substituents and the cyclohexyl ring.
Meejoo et al. (423) studied the structural disorder of b-polymorph of (E)-4formylcinnamic acid. In the absence of the single crystal X-ray data, the structure
of this form was determined from powder X-ray. The 13C CPMAS was used to
investigate structural disorder of the orientation of the formyl group. One molecule in the asymmetric unit was observed. The CHO group shows peak splitting.
Because there is no NMR interaction that could give rise to such splitting,
authors conclude that it must be due to the structural disorder of this group.
The relative intensities of the two CHO peaks did not depend on the CP contact
time, reflecting the relative populations of the two orientations. The peak that is
shifted downfield is proposed to correspond to the orientation engaged in short
intermolecular hydrogen bond, on the basis that intermolecular H-bonding to
carbonyl groups is known to increase the 13C isotropic chemical shift.
Iuliucci et al. (424) studied ring chain tautomerism of erythromycin A. The
13
C CPMAS resonances were assigned using spectral editing and chemical shift
calculations with DFT B3LYP/D95 method. Previously acquired values of
chemical shift anisotropy for all carbons are also shown (425). A chemical
shift assignment using 2D correlation techniques was also previously performed
(Fig. 54) (426). The calculated chemical shift tensors correlated well with the
experimental shifts. Formation of cyclic hemiketal was observed upon heating
or desolvation by desiccation. The principal values of the hemiketal carbon
chemical shift tensor are reported.
Reutzel-Edens et al. (427) characterized the five crystal forms of
LY334370, a 5HT1f agonist, three anhydrates, a dihydrate, and an acetic acid
solvate. Single crystal data of only Form-I were available. The 13C CPMAS
spectra for these five forms showed the differences typically observed for polymorphs and solvates (Fig. 55). Splitting due to the residual dipolar coupling to
quadrupolar 14N and the distorted intensities of the carbons J-coupled to 19F
were observed. Single molecular conformation was detected in each crystal
form. Some carbon resonances were found to show significant differences compared with the solution (up to 3 ppm), possibly due to the formation of H-bonds.
Zolek et al. (428) combined 1H and 13C SSNMR spectroscopy with GIAOCPHF calculations to characterize chloramphenicol, thiamphenicol, and their
pyrrole analogs. Unexpectedly, good resolution in the 1H MAS (32 kHz spinning)
SSNMR Spectrometry
489
Figure 54 The 2D 13C 13C chemical shift correlation spectrum of [U-13C]-labeled

erythromycin. Source: Modified from Ref. 426.
spectrum was observed for chloramphenicol, resolving the signals of ortho- and
meta-protons separated by 0.5 ppm. The observed differences of the carbon shifts
between solution and solid were up to 6 ppm. Broadening or splitting of the
carbon resonances of carbons linked to 14N or to 35,37Cl were observed as a result
of the residual dipolar coupling to quadrupolar nuclei. Good correlation was observed
between the GIAO-CPHF calculated and observed carbon chemical shifts. Some of
the differences between the calculated and observed values are ascribed to
H-bonding, because the calculations were performed on isolated molecules.
Mirmehrabi et al. characterized the enamine and nitronic acid tautomeric
forms of ranitidine hydrochloride. The 13C CPMAS identified a molecular
disorder in Form-II, whereas Form-I was more ordered.
Vanhaecht et al. (429) studied cocrystallization phenomena in piperazinebased copolyamides by SSNMR and other techniques. The 1H SSNMR and T1rH
relaxation was used to investigate the influence of the composition on the
490
Medek
Figure 55 The 13C CPMAS spectra of LY334370 hydrochloride crystal forms. Abbreviation: CPMAS, cross polarization and magic angle spinning. Source: Modified from
Ref. 427.
percentage of the rigid phase of the copolyamides and delivered additional indications for cocrystallization.
Portieri et al. (430) studied the effects of polymorphic differences on three
sulfanilamide forms observed by 13C and 15N SSNMR. The relaxation times were
measured at different temperatures and analyzed in terms of mobilities. Nitrogen
shielding parameters were obtained from static and MAS spectra of labeled
samples and compared with calculations.
Kameda et al. (431) studied 13C chemical shifts of the triclinic and monoclinic crystal forms of valinomycin. The 2D FIREMAT technique was used to
characterize carbon chemical shift tensors (Fig. 56). The isotropic chemical
shifts and the corresponding tensor components gave useful information about
SSNMR Spectrometry
491
Figure 56 The 2D 13C FIREMAT of triclinic valinomycin. Abbreviation: FIREMAT,

five p-replicated magic angle turning. Source: Modified from Ref. 431.
the molecular structure of valinomycin in the solid state. An utility of the chemical shift tensors for the understanding of secondary structures of peptides was
demonstrated.
Remenar et al. (432) determined the physical form of 2-tert-butyl-4methoxy-phenol (BHA), 13C labeled on the methoxy carbon, and wet granulated
onto common pharmaceutical excipients. At 0.1% loading, BHA was shown to be
amorphous and mobile (by mobility-induced loss of the 13C CPMAS signal due
to smaller average dipolar coupling) signal in the freshly prepared blends. At
0.5% loading, it was found amorphous on cellulose and crystalline on lactose,
mannitol, calcium phosphate dehydrate, and croscarmellose sodium.
Sheth et al. (433) studied mechanochromism of piroxicam accompanied by
intermolecular proton transfer. A proton transfer accompanied both solid-state
disorder, and a change in color induced by mechanical stress, a phenomenon
referred to by authors as mechanomorphism. Variable temperature and variable
contact time 13C CPMAS spectra of polymorphs I, II, amorphous and monohydrate were acquired (Fig. 57). Deprotonation of the acidic enol group provided
more carbonyl-like character, and therefore a large downfield shift occurred.
The protonated and deprotonated species could thus be quantified by integration
of their respective peaks. Most of the amorphous piroxicam was found to consist
492
Medek
Figure 57 The 13C CPMAS spectra of piroxicam polymorphs. (A) (E) show Form-I,
Form-II, Form-I cryoground for 60 minutes, Form-II cryoground for 60 minutes, and
piroxicam monohydrate, respectively. Source: Modified from Ref. 433.
of neutral molecules. The charged species comprised of only about 8% of the

amorphous phase. The ability to quantify the fractions of neutral and charged
molecules in the amorphous phase highlighted the unique capability of
SSNMR to quantify mixtures in the absence of standards.
Booy et al. (434) investigated the use of 13C labeling to enhance the carbon
sensitivity to study the effect of tabletting on the polymorphism of a steroidal
drug at dosage levels of 0.5 2.5% active. The crystalline form could be
readily analyzed in the tablets, and no change of form was observed. Labeling
of neighboring carbons, which causes broadening due to the homonuclear
dipolar coupling, should be avoided.
Schmidt (435) characterized the polymorphism of local anesthetic drugs of
falicaine HCl and dyclonine HCl by 13C and 15N CPMAS. Differences between
SSNMR Spectrometry
493
Figure 58 Line widths of 13C CPMAS spectra of ibuprofen (IBP) when in physical
mixture with the talk. Different IBP peaks correspond to different symbols. Abbreviation:
CPMAS, cross polarization and magic angle spinning. Source: Modified from Ref. 437.
the polymorphs of both drugs were readily observed by SSNMR. Dipolar dephasing and comparison with solution help with assignments. An internal rotation of
part of the drug molecule was confirmed by variable temperature study.
Suzuki and Kawasaki (436) evaluated forms of troglitazone by SSNMR.
The SSNMR distinguished the hydrated and anhydrous forms better than
PXRD. Diastereomers were found to coexist as a physical mixture. Unlike
PXRD, SSNMR was able to characterize the crystal forms when formulated in
tablets. A stable amorphous form was found in tablets.
Barich et al. (437) investigated SSNMR line widths of ibuprofen in drug
formulations. The line widths were measured for different sample preparations,
including crystallization from different solvents, melt-quenching, manual grinding, cryogrinding, compacting, and blending with excipients. Very interestingly,
when in physical mixtures with most excipients, the line width decreased, affording better 13C resolution. The exception was mixing with talc, for which the line
width increase is attributed to the increased magnetic susceptibility anisotropy of
talc (Fig. 58). The line width can thus provide information on physical characteristics of the sample.
Fluorine Applications
As a result of the structure-based rational drug dosing, many active pharmaceutical ingredients contain fluorine atoms. Fluorine groups may improve the
potency or offset the infavorable pharmacological properties of the compound.
494
Medek
From the perspective of the SSNMR spectroscopist, the presence of fluorine

atoms is very desirable since 19F nuclei are almost as sensitive as 1H, but do
not face the same limitations as protons. Because the number of fluorine atoms
per molecule is usually small, homonuclear 19F 19F couplings are relatively
weak and can be removed by conventional MAS. However, a proton decoupling
and/or a very fast MAS is desirable to cancel the strong 1H 19F dipolar interaction, sometimes necessitating the use of special NMR hardware. If CF3
groups are present, a very fast MAS spinning should be applied to remove the
homonuclear 19F 19F coupling, already scaled down considerably by the fast
internal rotation of the CF3 group. The proton decoupling is less important in
this case, because the residual 1H 19F coupling is well removed just by the
fast MAS. The fluorine line widths were found to be determined by mostly
inhomogeneous contributions (438,439). The typical fluorine lines are approximately 1 ppm wide. Due to the large fluorine overall chemical shift range, this
is typically sufficient to resolve the different polymorphs. The limit of detection
based on sensitivity considerations is well under 1%, and in favorable cases LOD
of 0.01% is not unusual. Fluorine NMR can be thought of as an ideal filter when
applied to formulations. Because no typical excipient contains fluorine atoms, the
only signal is due to the fluorinated API. Also the same detection limits apply
when corrected for the dilution by excipients.
The presence of fluorine also enables measurements of fluorine carbon
distances by REDOR experiment without the need for isotope enrichment.
Multidimensional homonuclear or heterunuclear experiments are also possible
serving as an additional tool for structural investigations. It should be noted,
however, that the presence of fluorine atoms complicates the observation of resonances in carbon spectra for carbons that are in physical proximity of the
fluorines. It is very difficult to efficiently decouple fluorine and protons simultaneously. The 19F 1H dipolar coupling makes the proton decoupling of
nearby carbons inefficient, because the 19F 1H coupling effectively increases
the proton chemical shift range and the proton off-resonace affects then
limits the 1H 13C decoupling efficiency (440).
Campbell et al. (441) studied two polymorphs of a piperidone-substituted
benzopyran with a fluorinated sidechain by 13C and 19F SSNMR and other techniques. More than one molecule per asymmetric unit was observed for Form-II. A
splitting of the carbons bound to 14N due to the residual dipolar coupling was
observed and calculated with literature values of the quadruplar tensor. An
increased carbon resolution was achieved by the application of double (1H and
19
F) decoupling. Low-temperature spectra did not show any extra information.
Proton and fluorine relaxation parameters were obtained. The proton decoupled
fluorine line widths were approximately 500 Hz.
Harris and Crowe (442) studied fluorinated diazadiphosphetidines to obtain
information on structure and dynamics. Chemical shift anisotropy tensor were
characterized by analysis of the spinning sideband manifolds. Proton and fluorine
relaxation times were measured.
SSNMR Spectrometry
495
Brouwer et al. (443) studied supramolecular solids by carbon and fluorine

NMR. The 19F MREV-8 experiment, minimizing the homonuclear fluorine
contribution, was used to characterize the shielding tensor of a static sample.
The static CP spectra (no MREV-8 applied) were then simulated using fluorine
inter-nuclear distances.
Fuchs and Scheler (444 447) and others used high-resolution 19F SSNMR
to quantitatively investigate structural changes like branching and cross-linking
of irradiated polymers.
Holstein et al. (448) studied the semicrystallinity and polymorphism in
poly(vinylidene fluoride) by 1H 19F double resonance SSNMR. Amorphous and
crystalline regions provided distinct chemically shifted signals. Application
of relaxation filters suppressed either the amorphous or crystalline parts.
The polymorphism of the crystalline polymer was reflected in the observation
of several fluorine resonances. Isolated spectra of the crystalline phase of different polymorphs could be obtained. Especially the immobile crystalline chains
required application of high-power proton decoupling. Ando et al. (449,450)
studied the crystalline and amorphous phases of the semicrystalline poly(vinylidene fluoride) using variants of 1H 19F and 19F 1H CPMAS experiments. An
1
H 19F CPMAS inversion recovery experiment, in which, after a short CP
period, the phase of the fluorine spin-lock pulse is changed by 1808 and the
inverted fluorine signal decays and goes through zero, was used to observe
details of the initial CP processes as the coherent magnetization oscillations
(Fig. 59). The CP drain experiment was used to measure the contact time dependence of the residual proton magnetization after the proton background signal
was removed by the double CP (Fig. 60). The spin lock served to remove any
residual fluorine magnetization before the CP drain step. Significant differences
were observed between the crystalline and amorphous phases in the CP and relaxation behavior and related motions in the amorphous domain. Su and Tzou (251)
determined the relative content of the various domains of poly(vinylidene fluoride). The phase separation was analyzed by 2D 19F exchange spectroscopy and
chemical shift anisotropy was extracted by fitting the spinning sideband patterns.
Wormald et al. further studied this polymer using combination of relaxation
Figure 59 The 1H 19F CPMAS sequence with variable contact inversion recovery to
study details of the initial part of the CP process. Abbreviations: CPMAS, cross polarization and magic angle spinning; CP, cross-polarization.
496
Medek
Figure 60 Double CPMAS (19F 1H and 1H 19F) experiment to study contact time
dependence of proton magnetization after the removal of proton background. Abbreviation: CPMAS, cross polarization and magic angle spinning.
filters and two-dimensional RF-driven re-coupling (RFDR) and spin diffusion

experiments (252).
Ando et al. (253) studied an amorphous perfluorated polymer by a combination of high-speed 19F MAS, 19F CRAMPS, and 19F 13C CPMAS experiments.
The peak assignment was confirmed by GIAO shielding calculation. A single
phase was confirmed based on relaxation measurements. The 19F 13C CP kinetics
was followed by variable contact time studies and analyzed theoretically by the
quantum mechanical master equation. CF distances, CPMAS time constants,
and parameters of homonuclear spin-diffusion among fluorines were determined.
Wenslow (454) studied crystalline and amorphous forms of a selective
muscarinic M3 receptor antagonist in both bulk and pharmaceutical dosage
form samples (Fig. 61). Ambient and elevated temperature spectra were taken.
A temporal line narrowing was observed at high temperatures, caused by the
solvent escape, resulting in a high mobility of the drug. Observing the same
effect on drug product, the conversion of crystalline to amorphous form was
unambiguously confirmed in the low dosage tablets.
Solis et al. (455) studied the fluorine chemical shifts of organic molecules
and analyzed them in terms of surface charge representation of the electrostatic
embedding potential of crystalline inter-molecular effects. Different charge
models (GRID and SCREED) and a cluster model were evaluated in order to
include intermolecular interactions in the calculation of 19F chemical shift
tensors. The charge models allowed determination of the point charge distribution that mimicked the crystal field in the shielding calculations.
Glaser et al. (456) studied orientation of a peptide in lipid membranes based
on fluorine dipolar coupling of the 4-CF3-phenylglycine labels. A simple pulse
experiment allowed for simultaneous measurement of both the anisotropic
chemical shift and the homonuclear dipolar coupling of the CF3 group in an
oriented membrane sample, uniquely defining the effective tilt angle of the
CF3-axis with respect to the external field. The sign of the dipolar coupling
was deduced from the anisotropic chemical shift of the oriented sample. The
dipolar coupling provided an accurate measure of the orientation of the amino
acid sidechain in the peptide backbone.
SSNMR Spectrometry
497
Figure 61 The 19F MAS spectra (15 kHz spinning) of different forms of the compound
shown at the top. No proton decoupling was applied. Abbreviation: MAS, magic angle
spinning. Source: Modified from Ref. 454.
Park et al. (457) investigated crystalline structure of poly(vinylidene fluoride) hybrid composites. The long T1rF of the crystalline phase, which was significantly different from that of amorphous phase, was used to observe the crystalline
domains selectively after spin-lock filter. The selective fluorine spectra were
found to be very sensitive to the crystalline structures, and the relative populations of each morphological component could be extracted quantitatively.
Budarin et al. (458) studied surface energy and surface area of inorganic
materials indirectly by measuring chemical shifts of adsorbed trifluoroacetic
acid. The advantage of this technique over the standard surface area measurement
by using N2 stemmed from the fact that the real-life catalysis involves larger
molecules that may not be able to access the entire surface. By NMR, the
surface area can be measured specifically with fluorine analogs of the desired
molecules.
Antonioli and Hodgkinson (440) studied 13C 19F interactions in carbon
spectra of spinning solids. Carbon resonances of carbons close to fluorine
nuclei typically showed significant line broadening. This is explained by the ineffieciency of proton decoupling by the 1H off-resonance effect caused by the
1
H 19F dipolar coupling (Fig. 62). Sharper lines could be obtained by applications of more sophisticated decoupling schemes, such as SPINAL-64 (79).
The lines were found to have an essentially inhomogeneous character ascribed
498
Medek
Figure 62 Topsimulation of CW-decoupled 13C spectra in 13C 19F 1H coupled

system under MAS for different combinations of chemical shift anisotropy (dH) and
19
F 1H dipolar coupling (dHF). Bottomexperimental and simulated 13C CPMAS
spectra of flurbiprofen at different decoupling and MAS spinning conditions. Abbreviations: CW, continuous wake; MAS, magic angle spinning; CPMAS, cross polarization
and magic angle spinning. Source: Modified from Ref. 440.
to a large anisotropy of bulk magnetic susceptibility. The line widths were also
found to be very sensitive to the precise setting of magic angle. A misset of
the magic angle resulted in asymmetric broadening of the 1JCF doublet components, degrading resolution. Another factor influencing the carbon line width
was the effect of the finite lifetime of the 19F states. If 19F relaxes fast
(1JCF , 1/2pT2F; often true for rigid solids), the carbon 1JCF doublet collapses
SSNMR Spectrometry
499
to a single line (it is self-decoupled). If the fluorine relaxation is slower,

the carbon J-doublet is broadened. The dipolar interaction with fluorine also
often makes the appearance of the J-doublet asymmetrical, due to the different
intensity distribution in the spinning sideband manifolds from the two 19F
spin states.
Amorphous Phase Characterization
Gustafsson et al. (330) compared SSNMR and isothermal microcalorimetry in the
assessment of the amorphous component of processed lactose. The PLS mathematical models established from NMR spectra accurately predicted results
even when the lactose contained an additional component, acetylsalicylic acid.
Both microcalorimetry and SSNMR could detect as low as 0.5% amorphous
lactose. SSNMR has the advantage to provide additional structural information
about the mixture. Since the authors used recycle delay of 2.5 seconds, the crystalline component was partially saturated due to its long T1H. Moreover, this
relaxation is likely to change from sample to sample, limiting the potential
applications.
Utz (450 462) applied NMR to investigate structure and dynamics of
amorphous solids. A global and local order that develops under plastic deformation is investigated by 2D DECODER. This technique correlates two different
orientations of the sample to each other, while the sample undergoes flips or slow
rotation. Comparison of the experimental results with molecular simulations indicated that thermal relaxation was active up to the entanglement length scale, even
at temperatures far below the glass transition.
Cadars et al. (220) studied the disorder of amorphous compounds in chemical shift-correlated 2D experiments. This study is a follow-up on the previous
publications from the same lab discussed earlier in the context of INADEQUATE
experiment (Fig. 63) (217 219). A conditional probability of chemical shift distribution can be extracted from the experimental data. The tranverse dephasing
times are important in determining the resolution of the distribution. The structural information contained in the distribution is important for the understanding
of disorder.
Relaxation and Dynamics
NMR is an excellent technique to study dynamic processes, such as molecular
motions, in solid samples. The sensitivity of NMR to motions arises from the
orientational dependence (the anisotropy) of the nuclear spin interactions
described in the introduction section. Types of motions and their frequencies
can be inferred. Motions can be sampled by NMR directly, through coherent processes, or, indirectly, through incoherent processes, such as relaxation. The direct
sampling of motions through changes in the nuclear spin interactions include
monitoring 1D spectral line shapes or designing multidimensional experiments,
such as chemical exchange experiments. The indirect sampling of motion stems
500
Medek
Figure 63 The 2D 13C INADEQATE spectrum of 10% 13C-labeled cellulose. Two

different structural conformations were found. The chemical shift correlations are highlighted by the lines. The 1D stick spectra reproduce the chemical shifts of the two structural conformations. The regular 1D 13C CPMAS spectrum of cellulose is shown at top.
Abbreviation: CPMAS, cross polarization and magic angle spinning. Source: Modified
from Ref. 219.
from the fact that the relaxation processes are driven by the random fluctuations in
the nuclear spin interactions from molecular motions. Both the coherent and incoherent processes are sensitive to the motions in broad frequency range. As such,
NMR can detect motions with frequencies ranging from near zero to 109 Hz.
Relaxation is a response of the system to perturbations, such as a series of rf
pulses by driving the system back to its thermodynamic equilibrium. There are
different types of NMR relaxation, including longitudinal spin-lattice (T1), transversal spin spin (T2), longitudinal relaxation in the rotating frame (T1r), and
cross-relaxation (the NOE effect). Longitudinal relaxation corresponds to
SSNMR Spectrometry
501
restoring the spin levels to their Boltzmann equilibrium distribution described by

equation (3). The spin spin relaxation induces the loss of the transversal coherence (the loss of NMR signal) due to gradual randomization of the phases of individual spins in the whole-spin ensemble. The longitudinal relaxation in the
rotating frame represents the loss of coherence under the application of a continuous irradiation by a spin-locking pulse. If strong enough, the spin-locking pulse
cancels the evolution due to coherent contribution of the various nuclear interactions (mainly dipolar coupling in the case of organic molecules), and the
only loss of signal is due to the incoherent T1r relaxation. The actual values of
the relaxation times depend on the type of the nuclear interaction causing the
relaxation and on the motions present locally in the molecule. Typically, in the
absence of paramagnetic species or quadrupolar nuclei, the dominant relaxation
mechanism for organic molecules is the dipolar coupling. The motional frequency dependence of the relaxation process is expressed by so-called spectral
density functions. They reflect the likelihood of the given nuclear interactions
(e.g., dipolar coupling or combination of interactions) that molecular motion
with a particular frequency causes the relaxation. This frequency dependence
of the spectral density is very different for the T1, T2, and T1r types of relaxations.
In general, T2, T1r, and T1 are sensitive to very low frequency motions, mid-kHz
(tens to hundreds kHz; on the order of the spin-locking field) and MHz (tens to
hundreds MHz; on the order of the Larmor frequency) range, respectively. The
motions are typically described by one or more correlation times, which are
indirectly proportional to temperature. Figure 64 represents a schematic depiction
of the temperature dependence of the T1 and T2 relaxation times for two different
values of external magnetic field. The extreme narrowing regime is defined by the
positive slope of the relaxation with increasing temperature and is also
B0a > B0b
T1 at B0a
lnT1
or
lnT2
T1 at B0b
Extreme
Narrowing
Regime
T2 at B0a
Slow Motional
Regime
c ~ 1/T
Figure 64 Schematic representation of temperature dependence of the T1 and T2 relaxation times. The vertical dotted line corresponds to the correlation time in the order of the
inverse of the Larmor frequency at the field B0a. This line divides the temperature dependence into two regions: the extreme narrowing regime with fast motions (with respect to
the Larmor frequency) and the slow motion regime.
502
Medek
characterized by fast molecular motions (as compared with the Larmor frequency). Conversely, the slow motional regime is defined by the negative
slope of the relaxation with increasing temperature.
Torchia and Szabo (463) presented a formalism for extracting spin-lattice
relaxation in static and spinning polycrystalline solids in their important and
highly quoted publication. General equations were developed for orientationdependent frequencies and T1. Various jump and diffusive models are considered.
Explicit equations for dipolar and quarupolar values of T1 are presented for twoand three-site jumps and the continuous diffusion. Torchia (464) also developed a
CP-based sequence for determination of T1C, which is routinely applied to
measure carbon longitudinal relaxation.
Schaefer et al. (465) studied motion in glassy polymers of poly(methyl
methacrylate), polycarbonate, poly(phenylene oxide), polystyrene, polysulfone,
poly(ether sulfone), and poly(vinyl chloride) by 13C NMR. T1C, T1rC, and TCH
have been measured for individual carbons. From adiabatic demagnetization in
the rotating frame experiments, it was found that T1rC is dominated by spin
lattice rather than spin spin processes, rendering the T1rC sensitive to the lowto mid-kHz frequency range of motions. T1C contained information about
motions in the Larmor frequency range. The TCH/T1rC ratio was correlated to
toughness and impact strength for all seven polymers.
Kitamaru et al. (466) used relaxation filters to detect crystalline-amorphous
interphase of polyethylene. The T1C relaxation times as long as 2750 seconds for
the crystalline component were observed. In contrast, the T1C relaxation times of
the amorphous component were less than 1 second. Only little differences in T1C
were found. The crystalline components were eliminated by either short recycle
delay (T1C filter) or by T2C filter (carbon direct polarization with interrupted
proton decoupling). Further, the rubbery and interfacial amorphous phases
could be distinguished based on their T2C differences (2.4 vs. 0.044 ms). Two
amorphous phases were found by the virtue of their different T1H times. No
proton spin diffusion between the two phases was observed, most likely due to
a significant difference in mobility. The fully relaxed direct polarization
spectra were quantified in terms of the different crystalline and amorphous components by means of line shape analysis.
Oksanen and Zografi (467) studied molecular mobility in mixtures of
absorbed water and poly(vinylpyrrolidone) (PVP). At least two water populations
with different timescales were identified as the water content decreased based
on NMR diffusion measurements and T1H (Fig. 65). Significant rotational and
translational mobility of water existed even at very low water content. T1C
measurements of PVP showed that water, even at very low content, had a profound effect on the molecular mobility of PVP. The authors conclude, that
more consideration should be given to the state of polymer and less to the
extent to which water is bound or unbound.
Van den Dries et al. (468) studied the molecular mobility of water and
carbohydrate protons in maltose samples as a function of water content and
SSNMR Spectrometry
503
Figure 65 Effective T1Heff of absorbed water in PVP as a function of temperature for

divergent water contents. Abbreviation: PVP, polyvinylpyrrolidone. Source: Modified
from Ref. 467.
temperature using 1H NMR line shapes analysis and T2H relaxation. Temperature-dependent mobile (mostly due to water, especially when below Tg of
sugar) and immobile (sugar) proton fractions were observed. The mobility of
the mobile water increased with water content and temperature. A small break
in water mobility was observed at Tg, indicating that water molecules sensed
the glass transition. The plastisizing effect of water is interpreted as disrupting
the hydrogen bond network of the sugar molecules that is formed when the
sugar is cooled under its Tg. Authors used the second moment analysis (as a
measure of the strength of the dipolar coupling) of the broad immobile proton
line shapes to learn about their mobility.
Yoshioka et al. (469) studied the effect of excipients on the molecular mobility of lyophilized formulations as measured by Tg and Tmc (NMR-based critical
mobility temperature, defined as the temperature of appearance of the Lorentzian
relaxation process due to liquid protons). The Tmc of dextran (PHEA), PVP, carboxymethylcellulose sodium salt (CMC-Na), hydroxypropylmethylcellulose
(HPMC), and methylcellulose (MC) have been determined as a function of the
plasticizing effect of the increasing water content. The formulations containing
PHEA, MC, and HPMC exhibited comparable Tmc values and a significant
increase in T2H values of water peak (corresponding to higher water mobilities)
504
Medek
at lower water levels compared with dextran and CMC-Na. They became microscopically liquidized (containing liquid protons as detected by NMR) at lower
temperatures, suggesting higher molecular mobility of water. Higher mobility
of water did not necessarily correspond to greater reduction in Tmc, because T2H
reflected the average of both bound and mobile water. The determined Tmc
values were lower than the corresponding Tg values by 23 328C, indicating
that the formulations became microscopically liquidized below their Tg.
Yoshioka et al. (470) linked the molecular mobility of lyophilized protein
formulations to the molecular mobility of polymer excipients by using 13C NMR.
In order to calculate the correlation time tc, T1C was measured and related to tc
assuming existence of a single correlation time, a slow motional regime, and
either the dominance of the 1H 13C dipole dipole interaction [Equation (24)]
or chemical shift interaction [Equation (25)] for the relaxation.
6
1=T1C (1=10)g2C g2H h2 (2p)2 rCH
1=(vC vC )2 3=v2C
6=(vC vC )2 =tc
(24)
The correlation time of dextran was found to follow the linear temperature
dependence described by the Adam-Gibbs-Vogel equation (Arrhenius-like
relationship) below Tg and the nonlinear temperature dependence by the
Vogel-Tamman-Fulcher equation above Tg. The discontinuity was likely due
to glass-rubber phase transition and was found to
1=T1C (6=40)g2C B20 d2Z (1 h2 =3)(2=v20 tc )
(25)
correspond to Tmc as determined by T2H.

Zumbulyadis et al. (471) elucidated mixtures of polymorphs by 13C
CPMAS in combination with direct exponential curve resolution algorithm
(DECRA). DECRA is a chemometric method for self-modeling mixture analysis.
It employs a generalized rank annihilation method (GRAM), which requires two
proportional data sets for the analysis, that was accomplished by splitting a single
data set into two. The key to the analysis is the exponential dependence of the
signal from the individual components on the physical property being measured
(diffusion, T1 or T2 relaxations). Under optimal conditions, DECRA can quantitatively separate spectra of components whose relaxation differ only by 20%. It
is, however, critical to keep the spectral artifacts originating from instrumental
instabilities to a minimum. The authors show that pure-component carbon
spectra of each of the two polymorphs can be unambiguously extracted from a
spectrum of a mixture based on their different T1H (Fig. 66). This technique
should have a general application, because in majority of the cases, there are
relaxation differences between polymorphs. This technique would not work to
distinguish microphases with domains smaller than 30 nm, because spin diffusion
is expected to equalize the proton relaxation. DECRA is relatively insensitive to
poor S/N or spectral overlap.
SSNMR Spectrometry
505
Figure 66 Example of resolving components of mixtures using the DECRA algorithm

based on different relaxation properties of the components. The top spectrum was recorded
with inversion recovery time of 0.2 s. Abbreviation: DECRA, direct exponential curve resolution algorithm. Source: Modified from Ref. 471.
Silva et al. (472) studied a multicomponent polymer blend by 1H relaxation

(carbon-detected T1H and T1rH) and spin diffusion. The 2D WISE sequence was
also used as a T1H filter. Some phase separation was detected based on the proton
relaxation. This was confirmed by the WISE experiment. Merits of relaxation
studies versus 2D WISE experiment to determine phase separation are discussed.
Aso et al. (473) explained the crystallization rate of amorphous nifedipine and
phenobarbital from their molecular mobility was determined by carbon relaxation
times, T1C and T1rC. T1C of hydrogen-bound carbons are determined by molecular
mobility within nanoseconds to microseconds timescale, and T1C decreases with
increasing molecular mobility in the slow motional regime. Molecular motions in
the mid-kHz frequency range (as determined by T1rC, which was sensitive to the
fluctuations of the CH vector; T1rC decreased if motions existed in the mid-kHz
frequency range) became significant at temperatures between Tg-20 and Tg. The
506
Medek
temperature dependence of the crystallization rate on Tg was found to be coincident

with that of the relaxation rate calculated according to Adam-Gibbs-Vogel equation,
indicating a correlation of the crystallization with molecular mobility.
Glaser et al. (474) studied eight-membered ring solid-state conformational
conversion of nefopam methohalide salts via the atom flip mechanism. Editing
experiments, such as interrupted decoupling, CPPI, and inversion recovery
(20 second delay) were used for assignment purposes. Solid-state T1C was
measured and interpreted in terms of distance of the carbons from the closest
protons. From variable temperature CPMAS spectra, the temperature-dependent
occupancy factors for boatboat and twisted chairchair conformers are interpreted in terms of medium ring atom-flip facile interconversion between the two
low energy conformations in crystals containing an appropriate size interaction
void.
Kuwabara et al. (475) studied crystalline phase transitions of 1,20-eicosanediol. The 13C CPMAS and 1H CRAMPS revealed two magnetically different
structures in the chain ends associated with hydrogen bonding in the M and M0
phases. The 13C CPMAS line widths at intermediate temperatures showed broadening, which was explained by the authors as occurance of the molecular motion
with the order of magnitude of the decoupling field (about 100 kHz). The differences in the chemical shifts as a function of temperature are explained in terms of
the guache conformation effect. The T1C relaxation was measured to compare the
mobility of different carbon groups in each crystalline phase. The T1C relaxation
time as long as 2290 seconds was observed. Variable temperature 1H CRAMPS
data afforded detailed information about the hydroxyl groups.
Varner et al. (476) characterized molecular motions in the solid state by
measuring T1C at two different spinning angles. Because the C H dipolar interaction has the same orientational dependence as C D, the spectral densities
responsible for the dipolar and the quadrupolar relaxation are the same. The
orientational dependence of all three spectra densities [J0(0), J1(0), and J2(0)]
are thus determined.
Kakou-Yao et al. (477) showed hindered rotation of tert-butyl group in
2,2,5,5,-tetramethyl-3,4-haxandione monohydrazone by 13C CPMAS. The broadening of carbon resonances of the tert-butyl group at lower temperatures was
rationalized by higher activation energy of rotation, as compared with the free
conformers as a result of packing constraints.
Rheingold et al. (478) studied superlattices and polymorphs of 2,6-di-tertbutylnaphthalene. Two polymorphs were identified which differed by a factor of
12 in the number of crystallographically independent tert-butyl group environments. The 13C CPMAS showed only small and random chemical shift differences of no more than 1.6 ppm and only minor differences in line width, in
spite of the averaging of 12 tert-butyl groups. Low-field variable temperature
T1H relaxation data were acquired. Polymorph E relaxation exhibited very
complex behavior.
SSNMR Spectrometry
507
Beckmann et al. (479) studied relaxation properties of two polymorphs of

2,6-di-tert-butylnaphthalene. Single crystal X-ray structures of both forms A and
E were obtained. One molecule per asymmetric unit was found by 13C CPMAS.
Methyl group and tert-butyl group re-orientations in solids are usually the only
motions on the NMR timescale. All other motions, such as intramolecular
vibrations, occur on a fast scale to effect NMR relaxation. Temperature-dependent T1H relaxation was measured at two low Larmor frequency fields (8.5 and
22.5 MHz). In polymorph A, a very good fit of the temperature and Larmor frequency-dependent relaxation curves was achieved, predicting the three methyl
groups to reorient at the same rate as the tert-butyl group. The polymorph A
was found to be a textbook case of the effects of intramolecular reorientation
on NMR relaxation rates. In contrast, the relaxation of form E was found to be
significantly more complex. A unique fit could not be obtained since too many
parameters were involved. The temperature- and Larmor frequency-dependent
relaxation curves were very different and extremely sensitive to the differences
between the two forms.
Vittadini et al. (480) studied mobility in cellulose by 1H and 2H peak intensities, relaxation times, and spectral line shapes as a function of moisture content.
At higher moisture level, water molecules reoriented anisotropically, but the
intensity of the mobile 1H increased along with increases in deuterium T1 and
T2 values, approaching the fast exchange regime (Fig. 67). At the moisture
content below the monolayer value, the majority of the protons were immobile.
Perera et al. (481) characterized binary and ternary polymer blends by 1H
13
and C NMR. TCH CP constant and T1C and T1rC relaxation times were interpreted in terms of changes in molecular motions. TCH reflects the near static
motions, whereas T1C and T1rC reflect the MHz and mid-kHz regimes. T1H and
T1rH relaxation times were interpreted in terms of homogeneity of the polymer
matrices. Heterogeneity of the blends on 1 to 6-nm length scale were confirmed
Figure 67 Deuterium T1 and T2 relaxation of cellulose as a function of moisture content

(% dry basis; left side). 1H MAS spectra of cellulose at different moisture contents (right
side). Abbreviation: MAS, magic angle spinning. Source: Modified from Ref. 480.
508
Medek
based on calculation using equation (17), where the average literature values of
diffusion coefficient of D 0.5 0.8 nm2/msec and D 0.05 nm2/msec were
used for rigid and mobile polymers, respectively, and the time of diffusion was
substituted with the two observed relaxation times.
Tang et al. (482) studied dynamics of a-D -galacturonic acid monohydrate
by 13C T1r and chemical shift anisotropy. The comparison of solution shifts
allowed only for partial chemical shift assignment due to considerable spectral
overlap. The 13C T1 relaxation was slow on a timescale of 100 seconds, indicating
that 13C T1 is much longer than 100 seconds. T1r of all carbons showed a decreasing trend with temperature, implying that motion was on the low temperature side
of the T1r minimum and the motional correlation time was longer than 1025
seconds. The motional averaging of carbon chemical shift tensor as a function
of temperature was observed. The tensor parameters were extracted based on
fitting the 1 kHz MAS spinning sideband line shapes. Motions with frequency
of the order of 104 Hz were detected.
Higgins et al. (483,484) studied the temperature dependence of T1H and
T1rH proton relaxation of bisphenol-A tetramethyl polycarbonate/polystyrene
mixtures. Both pure components showed a maxima of T1H around the respective
Tg, indicating motions at frequencies inefficient for T1H relaxation. The temperature dependence of the relaxation of the blend was predicted as an average of the
single-component relaxations [Equation (26)].
blend
1=T1H
X na
a
T1H
a
(26)
The experimentally observed relaxation matched well with the predictions based
on Equation (26). T1rH was found very sensitive to motions in temperature in the
range 380 420 K. Some significant deviations were observed from the predicted
T1rH values of the blend, most likely due to time-varying concentration fluctuations leading to strong deviations from homogeneous relaxation behavior.
Lopes et al. (485) developed a mathematical method of obtaining 13C
CPMAS subspectra of single components based on differences in relaxation.
Application of this technique, which removes the need for obtaining the subspectra by trial-and-error approach, to cork cell walls is shown.
Tang et al. (486) introduced a proton relaxation-induced (PRICE; essentially a set of carbon detected, proton relaxation filters) technique, which used
T1H, T2H, and T1rH proton relaxation properties to relate the dynamics of the
plant cell walls to their domain structural details.
Nogueira and Tavares (487) studied behavior of alpha-methylstyreneco-acrylonitrile. The variable contact time CP and T1rH relaxation were found
to be the determinant factors to evaluate the dynamic molecular behavior and
the homogeneity of the comonomer distribution. T1rH values were not homogeneous, reflecting a random sequencing of the comonomer insertion in the
chains.
SSNMR Spectrometry
509
Lai et al. (488) studied poly(vinyl alcohol) gels by 13C NMR. Information
on the dynamics in the kHz range obtained from T1rH and T1rC data indicated a
homogeneous arrangement of the amorphous or swollen polymeric chains.
Adriaensens et al. (489) studied polymeric mixtures by 13C relaxation. No
indication for phase separation was found as both the T1H and T1rH relaxation
behavior were indicative of homogeneous mixture on the nanometer length scale.
Brachais et al. (490) studied a heterogeneous character of polymeric mixtures interpreted based on interpenetrating networks. The rate of 1H 13C CP was
composition-dependent, indicating that the chains of the polymeric components
were interpenetrating at a very low spatial scale. This was also supported by T1H
and 13C line width measurements as a function of temperature. The maxima in13C
line widths of pure components were coincident with Tg. For the blend, 13C line
widths were sensitive to the composition, supporting the interpenetrating
character.
Alamo et al. (491) measured T1C of noncrystalline regions of semicrystalline polymers. Because the T1C values of some carbons were found shorter than
T1H values, the relaxation showed multiexponential behavior due to the significant contribution of NOE effect. To eliminate this transient NOE contribution,
weak proton saturation pulses were added during the relaxation delay. The
authors used a modified CP-based saturation recovery sequence to filter out
regions with long T1C. It is based on alternating, in successive scans, the
carbon magnetization in IZ and 2IZ directions immediately after the CP step
and keeping the relaxation delay short.
Aso et al. (492) studied the effect of water on molecular mobility of sucrose
and PVP in a colyophilized formulation. Due to its much higher Tg, PVP reduced
the molecular mobility of sucrose as determined by T1C and T1rC. Contrary to the
colyophilized dispersion, the decreasing T1C with increasing relative humidity
(RH) for pure PVP indicated that the MHz-order motions of PVP sidechains
increased with RH. No RH dependence of T1C (in 0 22% RH range) was
found for both sucrose alone and the colyophilized dispersion. The presence of
water apparently did not increase MHz-order motions. In the colyophilized dispersion, the formation of hydrogen bonds between PVP and sucrose prevented
the increase in mobility with increasing RH. T1rC decreased with increasing
RH for both sucrose and the sucrose-PVP dispersion, but the decrease was
much larger for sucrose itself, suggesting that water molecules enhanced the
mid-kHz-order motions by the plastisizing effect of water. PVP acted as an antiplastisizing agent, which offsets any plastisizing effects of water.
Yoshioka et al. (493) studied the different motions in lyophilized protein
formulations as determined by the temperature and RH dependence of T1H,
T1rH, T1C, and T1rC relaxation times. Unlike for the T1H values, a minimum
was reached in the temperature dependence of T1rH at higher humidities,
above which the system was in the fast motional regime. Based on the temperature dependence of T1H, T1rH of dextran protons in the lyophilized formulations,
the dextran correlation time decreased from 1026 seconds at 21008C to 1027
510
Medek
seconds at 08C. Multiple correlation times were found. Equations (27) and (28)
were used to calculate each correlation time based on either T1H or T1rH. v0
and v1 are Larmor frequency and spin-locking field frequency, respectively.
The factor A in Equation (28) was calculated from the observed temperature
minimum of T1rH. Above 08C, the molecular motion of methylene groups was
found to be too fast to be reflected in T1rH, but it is still reflected in T1H.
Instead, motions of methine groups began to be reflected in T1rH. Above the
glass/rubber transition, both motions are greatly enhanced, substantially shortening the relaxation times.
6
1=T1H (9=8)g2H g2H h2 (2p)2 4tc =15rHH
(1 v20 t2c )
6
(1 4v20 t2c )
16tc =15rHH
(27)
4v21 t2c )
(28)
1=T1rH Atc =(1
Sidhu et al. (494) studied polymorphism, conformation, and dynamics of

the inclusion compound of 1,2-dichloroethane with tris(5-acetyl-3-thienyl)
methane. The 13C CPMAS spectra confirmed the phase purity of the slow
cooling sample as being the triclinic form, whereas the sample prepared by
slow evaporation resulted in a mixture of triclinic and monoclinic forms. Variable
temperature solid-state 2H spectra showed a dynamical conformational equilibrium of 1,2-dichloroethane consistent with either the trans-guache model with
1208 jumps about the C22C bond or the twofold flip axis model. The 2H T1 relaxation rate measurements as a function of temperature supported the twofold flip
axis model.
Prout (495) published a review titled Molecules in Motion: A Study in the
Synergy of X-ray Diffraction and SSNMR. PXRD was found to give only an insufficient information on infrequent activated processes in the crystal or about some
motional behavior related to phase changes. Multiple examples are shown,
including pharmaceutically relevant compounds, such as phenoxypenicillin, for
which a phenyl ring flip was detected only by SSNMR and not predicted by
PXRD and deoxycholic acid clathrate with camphor, for which an order
disorder transition was detected by SSNMR.
Kuzmicz et al. (496) studied dimethylacetamide solvate of tetra(C-undecyl)calix[4]resorcinarene, a compound of interest for transport across membranes. Four resonances for each carbon were expected based on the crystal
structure, but only two were observed, most likely due to the almost identical disposition of the molecules. CP kinetics was studied to provide information on the
molecular mobility. The system was found to behave according to model I-I -S
with different proton spin temperatures, described by Equation (23). Based on
the fitted parameters for different carbons, the mobility of the compound was
found greater than that of the crystallized solvent. The solvent molecules were
found rigid, because they were trapped between layers of the compound.
Harris et al. (497) presented chemometric analysis of morphology and
dynamics in irradiated cross-linked polyolefin by means of DECRA. The 13C
SSNMR Spectrometry
511
DP and T1C-filtered CPMAS experiments revealed peaks arising form the amorphous, crystalline, and interfacial components. The relative abundances of each
phase were determined by subtracting DP spectra with 10-second recycle delay
from those with 2000 second recycle delay. Corrections for incomplete relaxation, partial decay of crystalline component, and spinning sidebands were incorporated. The 1H spin diffusion with 13C detection was used to probe the thickness
of the crystalline regions. The DECRA analysis was used to resolve the intermediate and amorphous phases.
Mao et al. (498) correlated poly(methylene)-rich amorphous aliphatic
domains in humic substances. T1H filter was applied using proton inversion
recovery with CPMAS detection to filter out the components with short protonrelaxation times. The relaxation delay was chosen such that the signal of short
T1H components just passed the zero level. Only the long relaxing components
are then detected. Direct polarization of 13C MAS spectra were used for
quantification.
Kitchin et al. (499) studied effects of polymorphism on functional group
dynamics of a- and b-phases of L -glutamic acid by deuterium NMR. The contrasting dynamic properties of the a- and b-phases were investigated by 2H line shapes,
and 2H T1 relaxation times combined with simulations. The 2H line shapes, were
consistent with three-site jump 1208 motions of the 2ND
3 group. The variable
temperatures of 2H T1 relaxation curves show minima at different temperatures
for the two phases. The differences in the dynamics between the two polymorphs
are rationalized in terms of different local environment in the crystal structures due
to the different H-bonding geometries. An ability of 2H to elucidate differences in
functional group dynamics was demonstrated, even for cases when the structure
differences between the polymorphs may be comparatively minor.
Leisen et al. (500) measured sorption isotherms by SSNMR. Instrumentation for automated recording of sorption isotherms at precisely controlled levels
of temperature and RH was designed. The specially designed static probe was
used to acquire the isotherm for a cotton fabric. Static proton DP spectra, T1H,
and T2H times were recorded automatically as a function of temperature and
RH. An analysis of time-domain data, Bloch decay, in terms of fast (rigid part)
and slow (moisture exposed mobile part) 1H T2 relaxing components was
made. An excess of the slowly relaxing component, when compared with gravimetric data, was attributed to plasticized segments of the formerly rigid cellulose
matrix. Hysteresis similar to gravimetric sorption isotherm was observed. The
NMR relaxation parameters were only dependent on the moisture content and
independent of the exposure history. Thus, material properties, as defined by
molecular mobilities, were found to be governed by moisture content alone.
Witkowski and Wawer (501) studied conformational dynamics in a-tocopherol esters. Differences in solution-solid carbon chemical shifts up to 6.9 ppm
were observed. Two molecules per asymmetric unit were observed. By variable
contact time studies, the CP time constant TCH was found to be in the range of
0.55 to 1.22 ms for quarternary carbons, 0.12 ms for CH2, and 0.32 to 0.37 ms
512
Medek
for CH3. The authors concluded that the a-tocopherol esters are flexible in two
fragments of the molecule.
Lupulescu et al. (484) developed 2D relaxation-assisted separation (RAS)
experiment using the different T1 relaxation of powdered or disordered systems
to get resolution in the indirect domain. Applications to 23Na-, 35Cl-, and
25
Mg- containing inorganic salts and 1H of glycine sample were shown.
Harris et al. (502) devised a generally applicable numerical methodology to
obtain more efficient relaxation filters to selectively retain or remove components
based on relaxation times. The procedure uses linear combination of spectra with
various recycle delays to obtain components that are both quantitative and pure.
Exponential behavior is assumed. The intensity of NMR signal in DP experiments with a recycle delay t and a relaxation time T1 is
I 1 et=T1
(29)
A Taylor expansion of this equation provides

1
1
1
I t=T1 (t=T1 )2 (t=T1 )3 (t=T1 )4
2
2
24
(30)
In a system with two components with relaxation times of T1a and T1b, the selectivity of the component a can be expressed as Ia/Ib

T1b
1 T1b 1
1
Ia =Ib
t

(31)
2 T1a T1a T1b
T1a
The linear dependence of the selectivity on the T1b/T1a ratio prevents efficient
and quantitative selection of the faster relaxing component in most systems.
The maximum selectivity of the component a is when t T1a T1b is
T1b/T1a. A combination of multiple spectra can improve this selectivity. If, for
example, one subtracts the spectrum acquired with twice the recycle delay
from that with a single recycle delay, the first-order terms vanish:
2I(t) I(2t)
T1a
2 2 3 3
t
t
t

T1b
T1a
T1b
32
and the maximum selectivity of the component a when t T1a T1b is (T1b/
T1a)2. Figure 68 shows the calculated signal intensity as a function of the relaxation time for different recycle delays and their linear combinations [Equation
(32)]. Figure 69 shows the relative filtering efficiency of slow relaxing component with respect to the fast relaxing one. To selectively observe the faster
relaxing component a, allowing for 3% of the undesired component b, the
ratio of the relaxation times would have to be 1:115 for the standard DP experiment, but only 1:23 if a combination of two recycle delays are used, and even
1:13 if combination of three recycle delays are used. The limitation of this technique is reduction in S/N. Assuming the same experimental time, the two recycle
delay methods will afford only 26% of the DP S/N.
SSNMR Spectrometry
I(t)
513
(A)
0.8
(B)
0.6
(C)
0.4
(D)
0.2
0
(E)
0
0.5
1.5
2.5
Relaxation Time T1 [s]
Figure 68 Calculated signal intensity as a function of the longitudinal relaxation time

(T1) for different recycle delays and linear combinations thereof. (A), (B) and (C)
curves were calculated using Equation (29) with the recycle delays set to 3, 2, and 1
second, respectively. (D) curve is a linear combination of 2 (C) 2 (B) [Equation (32)],
and (E) curve is a linear combination of 3 (C) 2 2 (B) (A). Note that the linear combinations of the recycle delays show faster cutoff and, as such, better filtering efficiency.
Similar procedures can be designed for filtering out the fast relaxing component with T1 relaxation filter preceding acquisition. For example, a long relaxing component with T1 100 seconds will lose 10% intensity when subjected
to 10-second filtration which removes the fast relaxing components. It will
I(t)
(C)
0.8
(B)
(A)
0.6
0.4
0.2
0
4
6
Relaxation Time T1 [s]
10
Figure 69 A relative filtering efficiency of a slow relaxing component (T1 . 0.2 s) with
respect to a fast relaxing component with T1 0.2 seconds. (A) One relaxation delay of
1 second. (B) A linear combination of two relaxation delays, 1 and 2 seconds. (C) A
linear combination of three relaxation delays 1, 2, and 3 seconds. Note that in all cases
the fast relaxing component is observed quantitatively (RD . 5 T1). The filtering efficiency of 1 corresponds to 100% suppression of the slow relaxing component while quantitatively retaining the fast relaxing one.
514
Medek
only lose 1% of its sensitivity when linear combination of two experiments with
two filters (10 and 20 seconds) is used, and, at the same time, better filtration of
the fast relaxing component is achieved. In addition, a combination of experiments can be designed shape modulation profiles, as for example, to filter out
very fast and very slow relaxing components.
Lim et al. (503) studied dynamics in poly(ethylcrylate-co-sodium acrylate)
ionomer 13C CPMAS. T1rC relaxation times were studied as a function of temperature to address chain mobility, correlation times, and activation energies.
Arhenius plots of correlation time versus inverse of temperature were constructed. Protonated carbons displayed much faster T1rC relaxation than the nonprotonated carbons. Molecular motions in the poly(ethylcrylate) homopolymer
needed a higher activation energy compared with the ionomer.
Krushelnitsky et al. (504) studied hydration dependence of backbone and
sidechain polylysine dynamics by naturally abundant carbon SSNMR. Temperature- and hydration level-dependent carbon relaxation experiments (T1C and
T1rC) were performed. The relaxation data were simultaneously analyzed using
the correlation function formalism and model-free approach. Unlike the backbone, the sidechains showed strong hydration response.
The dynamics of t-butyl group of 2-t-butyl-4-methylphenol was studied by
low-frequency SSNMR proton relaxometry (505). From X-ray data, there were
two inequivalent t-butyl groups in the asymmetric unit. Variable temperature
and variable field proton T2 relaxation data were fitted to two dynamics
models, both of which were consistent with the X-ray data. Comparison of the
t-butyl dynamics to other literature examples of a van der Waals organic solids
is made. The synergies between X-ray and NMR were pointed out, where Xray provided the boundary conditions for a set of dynamical models and NMR
provided information on local angular anisotropies in the intra- and intermolecular potentials.
Masuda et al. (506) compared the molecular mobility in the glassy state
between amorphous indomethacin and salicin based on T1C relaxation times at
temperatures below Tg. Salicin molecules exhibited multiple-exponential relaxation and were found to behave heterogeneously, showing fast local and slow
mobilities. In contrast, the amorphous indomethacin was homogeneous and
restricted, particularly in backbone carbons. The difference in mobilities of
these two drugs were correlated with their different stabilities towards crystallization. At room temperature, 60% of amorphous indomethacin crystallizes after
two months, whereas 100% of salicin crystallizes in 12 days.
Lubach et al. (507) investigated the effects of pharmaceutical processing on
relaxation and its implications for formulation stability. Two crystalline lactose
forms were subjected to pharmaceutical processing, including compaction, lyophilization, spray-drying, and cryogrinding. The compaction resulted in a
decrease of the lactose monohydrate T1H from 243 seconds to 79 seconds but
did not produce any amorphous phase. Lyophilization and spray-drying produced
amorphous lactose with T1H of four seconds. Cryogrinding produced mixtures of
SSNMR Spectrometry
515
Figure 70 The 13C CPMAS spectra and T1H relaxation times of crystalline lactose forms
cryoground ( from top) for 0, 2, 10, 30, and 60 minutes, respectively. Abbreviation:
CPMAS, cross polarization and magic angle spinning. Source: Modified from Ref. 507.
crystalline with amorphous lactose (Fig. 70). A short cryogrinding time

(2 minutes) reduced the T1H by an order of magnitude without significant
effect on the carbon spectrum. The content of the amorphous component
increased with grinding time. After 60 minutes, a mostly amorphous sample
resulted with some crystalline material left, showing T1H of two seconds
(shorter than the fully amorphous compound). The cryoground material
showed some annealing at ambient conditions (Fig. 71). The authors concluded
that the reduction of particle size, introduction of crystal defects, and
formation of amorphous material all reduced the apparent T1H. The reduction
in relaxation rates suggest that processing can have significant negative impact
on formulation stability. The authors propose replacing the lengthy and laborintensive stability studies with solid-state relaxation measurements.
Stability
Markovic et al. (508) studied amorphous to crystalline drug transition in solid
dispersion of potent cholesterol absorption inhibitor SCH 48461 with a polyethylene glycol. Carbon assignment was accomplished by comparison with solution.
The broader than expected line width of the amorphous spectrum is explained by
an incomplete averaging of the chemical shift tensor due to the presence of an
516
Medek
Figure 71 The 13C CPMAS spectra and T1H relaxation times upon annealing of anhydrous lactose cryoground for 60 minutes. Source: Modified from Ref. 507.
intermediate rate of molecular motion. To prove that the sharp peaks observed in
the capsule formulation were due to crystalline drug (as opposed to liquidlike plastic or waxy solid in rapid motion), interrupted decoupling and crossdepolarization experiments were applied. The rigid, crystalline lattice was confirmed by selective dephasing of protonated carbons in these spectra. The
amount of crystallinity was found to increase under different storage conditions
(258C/60%RH , 308C/60%RH , 408C/80%RH).
Watanabe et al. (509) studied stability of amorphous indomethacin (IM)
compounded with silica. Higher stability of amorphous IM compounded with
SiO2 was achieved by a cogrinding rather than a melt-quenching approach.
The analysis of the 29Si CPMAS line shapes indicated mechanochemical reaction
between SiO2 and IM by cogrinding. The 13C CPMAS spectra after short
cogrinding indicated a significant crystalline component, but fully amorphous
line shapes were observed at longer cogrinding times. The observed decrease
in stability of the short-ground samples were interpreted as due to a seeding
effect of the remaining crystalline component.
Forster et al. (510) investigated physical stability of amorphous drug and
drug/polymer melts using variable temperature SSNMR. Because the direct
comparison of the absolute relaxation values between different compounds is
not possible, the authors compared the temperature dependence of nifedipine
and indomethacin T1rH instead. The relaxation data were found to reflect an
overall freeing up of molecular mobility as the samples softened with
SSNMR Spectrometry
517
increasing temperature. Based on the temperature-dependent relaxation behavior,

the stability of the two drugs were predicted to be different, with the indomethacin samples showing a higher stability towards recrystallization.
Kalachandra et al. (511) examined a drug stability during film casting of
five different drugs in ethylene vinyl acetate. This system promotes longer oral
drug delivery. Based on the similarity of peaks between the pure drug and in
the cast film, one of the drugs was found to be dispersed in the film as small crystallites rather than as a true solid solution. The phase separation most likely
occurred during the drying process.
Domain Characterization
Spin diffusion properties of the abundant spins present an ideal tool to identify
and characterize the different domains present in heterogeneous samples. By
measuring the rate of spin diffusion, domain sizes can be estimated. Relaxation
and dipolar coupling filters are typically employed to selectively observe the
signal arising from a single phase.
Robyr studied local order in disordered solids by homonuclear (13C and
15
N) and heteronuclear (2H 13C) techniques (512). Information on conformations of polymer chains and torsional angles along polymer backbone was
obtained by these techniques.
Hu and Schmidt-Rohr (513) characterized crystallinity, domain sizes, and
the highly mobile second amorphous phase of ultradrawn polyethylene fibres.
The crystallinity was determined by the 1H line shape deconvolution and by
direct polarization carbon MAS with the knowledge of a correcting factor for
incomplete relaxation obtained from the 13C CPMAS T1 filter measurement.
The presence of a low-level, highly mobile, second amorphous phase was determined by 1H NMR. Based on spin diffusion, domain sizes of the crystalline and
amorphous regions were determined.
Colhoun et al. (514) studied orientation relaxation of polymers by 2D 13C
SSNMR. The 2D DECODER technique allowed for determination of the complete orientation distribution function for both amorphous and crystalline
domains. The authors present a numerical method for extracting a biaxial
molecular orientation distribution function of polystyrene from slow spinning
speed MAS DECODER, based on the chemical shift anisotropy.
Asano and Takegoshi (515) studied free volume of amorphous polymers by
13
C line width experiments. A temperature dependence of carbon line width of
six different amorphous polymers was examined around their glass-transition
temperatures (Tg). The observed temperature dependence was explained by
motional averaging over a dispersion of isotropic chemical shifts and an interference between local anisotropic motions and the dipolar decoupling. The latter
was responsible for the line width versus temperature function to show
maximum at one particular temperature. The derived fractional free volume at
518
Medek
Tg was much larger compared with that from rheological experiments. The difference was attributed to the differences in the amplitude of motions detected by
both experiments. The rheological experiments observed an overall motion of
polymers (cooperative arrangements of polymer segments), whereas the carbon
line width was governed mainly by local fluctuations in dipolar interaction.
The void space required for the local fluctuations is smaller than that for cooperative segmental motions.
Valtier et al. (516) used Goldman-Shen dipolar filter to remove magnetization from the rigid section of latex polymer and proton spin diffusion to quantify
the sizes of inner core and shell domains.
Palmas et al. (517) studied curing and thermal ageing of elastomers by 13C
1
and H NMR. A combination of high-resolution 13C CPMAS, T2H (protondetected) and T1rH (carbon-detected) measurements and 2D WISE experiments
provided identification of cross-linking, scission, and oxidation at low concentrations. A good spectrometer stability was required for the 2D experiments
under slowly accelerated aging conditions.
Lin et al. (518) studied phase heterogeneity in the miscible blend of amorphous poly(benzyl methacrylate) with semicrystalline poly(ethylene oxide).
Carbon-observed rotating frame proton relaxation (T1rH) was measured to
probe on possible molecular scales of heterogeneity. Depending on composition,
one to three phases were observed.
Huster et al. (519) studied membrane protein topology by the rate of 1H
spin diffusion from the mobile lipids to the rigid polymer. Magnetization transfer
in the mobile lipid was a slow and inefficient process. The slope of the spin diffusion build-up curves unambiguously determined the presence or absence of membrane-embedded protein domains. Quantitative analysis of the spin diffusion
yielded an approximate insertion depth of the protein in the membrane. 1H T2
filter was used to select the mobile proton spins. The technique was also
applied on unlabeled protein by detection of the lipid signals.
Brus et al. (520) studied mobility, structure, and domain sizes in polyimidepoly(dimethylsiloxane) networks of different composition by 13C, 1H, and 29Si
NMR. Large differences of chain mobilities were determined by 1D and 2D
WISE. A correlation of the structure and morphology with mechanical and
thermal properties is discussed.
Kretschmer et al. (521) quantified composition and domain sizes of industrial block copolymers using different 1H SSNMR methods. Static measurements
using a dipolar coupling filter step, destroying the magnetization of the rigid
regions, to separate the mobile and rigid parts of the sample were used. A combination of high-spinning speed MAS (25 kHz) and high magnetic field
(700 MHz) was used to get the structural information. Proton spin diffusion
was used to determine the domain sizes.
Nozirov et al. (522) studied the rigid and mobile amorphous phases and the
crystalline phase of poly[(R)-3-hydroxybutyric] acid. A spectral deconvolution
of the 13C MAS spectra showed different relative content of the various phases
SSNMR Spectrometry
519
at different temperatures. A large difference in molecular mobility between the

crystalline and amorphous regions was observed.
Murata et al. (523,524) studied a conformational stability of poly(L alanine), poly(D -alanine), poly(L -isoleucine), polyglycine and poly(L -valine),
and their blends using 13CPMAS. The 13C CPMAS and T1rH experiments were
employed. The conformations of the polypeptides in their blends were found
to be strongly influenced by the intermolecular hydrogen bonding interactions,
which caused their miscibility at the molecular level. The T1rH relaxation
times were measured and the domain sizes were estimated based on their values.
Jia et al. (525) presented an independent calibration of proton spin diffusion
coefficients in amorphous polymers. The quantitative values of spin diffusion
coefficients were obtained by using appropriately selected internal reference distances for spin diffusion within amorphous homopolymer. The 2D 1H 13C
HETCOR with added spin diffusion evolution period was used due to the excellent resolution provided by carbon chemical shifts.
Qui and Ediger (526) studied the length scale of dynamic heterogeneity in
supercooled D -sorbitol using multidimensional 13C NMR. The experimental
values of the dynamic heterogeneity length scale, obtained from 2D echo and
4D3CP experiments, were compared with various models.
Munson et al. (527) studied the stereo defect locations in the polylactide by
carbon SSNMR. They synthesized a polymer with spin-labeled stereo defects.
The carbonyl region of the spectrum was used to deconvolute the spectrum
into amorphous and crystalline components. The locations of the stereodefect
sites in either the crystalline or amorphous domains were determined by
spectra subtraction of the corresponding unlabeled spectrum from the labeled
spectrum. The surprising result was that the stereodefects were found to incorporate into both the crystalline and amorphous domains with almost equal probability. Only with increased defect concentration did the probability of the
incorporation into the crystalline region decrease.
Werkhovan et al. (528) determine the domain sizes of a block copolymer
by 1H spin diffusion using WISE experiment. Two values of T1H were obtained,
consistent with two phases. No differences in mobilities were determined by
WISE. By the spin diffusion measurements, the domain sizes were in the
20 nm range.
Akita et al. (529) studied the reversible order disorder phase transition
between g- and a-phases of oleic acid. Discontinuous changes of 13C CPMAS
chemical shifts (as a function of temperature) were observed at the phase transition temperature. Variable temperature T1C data showed the same tendency.
The T1C time became discontinuously short above the transition temperature.
Drug Drug and Drug Excipient Interactions
The sensitivity of the spin interactions, such as chemical shift or dipolar coupling,
to the local environment can be exploited to characterize drug drug or
520
Medek
drug excipient interactions. The sample mobility or relaxation can also be

applied to characterize these interactions. In many cases, an observation of a
proton transfer or a formation of H-bonds can point to the drug drug or drug
excipient interactions. If applicable, multidimensional correlation experiments
provide a positive proof of the presence of these interactions.
Rohrs et al. (530) studied the delavirdine mesylate tablet dissolution
affected by a moisture-mediated solid-state interaction between drug and disintegrant. The 13C CPMAS spectra of stressed tablets showed progressive conversion
to the free base, which was quantified based on knowledge of the relaxation constants of both forms (328). The free base is known to be much less soluble in
water. The authors postulated that the released methane sulfonic acid protonated
the croscarmellose sodium, the cross-linked form of carboxymethylcellulose
sodium used as disintegrant. Water content was found critical for the onset
of this reaction and is assumed to have either a dissolution medium or a plasticizer role.
Puttipipatkhachorn et al. (531) studied the salicylic acid release from chitosan matrix dependent on the drug polymer interaction and the physical state
of the drug. An analysis of the 13C CPMAS peaks revealed the drug polymer
interaction between salicylic acid and chitosan at an amino group, resulting in
salicylate formation.
Watanabe et al. (532) studied the apparent equilibrium solubility of indomethacin compounded with silica by carbon SSNMR. The solubility of
indomethacin was increased by cogrinding with silica. The change of shortrange disorder was examined by carbon NMR. Variations in conformation,
bond distortion, or nearest-neighbor distance upon a mechanical stress gives
rise to peak broadening. The peak broadening was used as a quantitative
measure of any changes in the molecular arrangement of the solid after prolonged
griding. Combined PXRD and SSNMR data showed that the mechanical energy
applied at early stages of grinding was used mainly to disrupt the long-range
order, whereas the rest of the mechanical energy disrupted the short-range
order. A high degree of correlation was found between the standardized peak
width (SFWHM) and the solubility plotted on logarithmic scale. Therefore,
carbon spectra could be used instead of the dissolution tests.
Interactions between the ibuprofen guest molecule and mesoporous silica
host matrix was investigated by 1H, 29Si, and 13C SSNMR (533). The DP 29Si
MAS experiments were used to quantitatively characterize the silica materials.
Extremely high mobility of the ibuprofen molecules when the matrix was not functionalized were found. Very poor CPMAS efficiency, sharp 1H spectra (both due to
the reduction of dipolar coupling), and short T1C (typical for liquids) data were all
indicative of ibuprofen exhibiting liquidlike behavior. In contrast, when the silica
matrix was functionalized by amino groups, the ability to acquire regular CPMAS
spectra pointed to strong restrictions in mobility, suggesting interaction between
the amino groups and the carboxylic groups. Chemical shift assignments were
made by comparison with ibuprofen solution shifts.
SSNMR Spectrometry
521
Boland and Middleton (534) studied interactions between a drug and a

membrane protein target by 1H 19F CPMAS. When the fluorinated ligand was
not bound to its membrane target, no CP signal could be observed due to the averaging of the dipolar coupling by fast tumbling of the solution. However, when the
ligand bound its target, CPMAS signal was observed. The binding could in turn
be characterized based on the variable contact time CPMAS curves.
Kawada and Marchessault (535) studied crystalline complexes of amylose
with lauric acid, n-butanol, n-pentanol, dimethyl sulfoxide, and thymol by 13C
CPMAS. The complexes formed single helices (V-polymorph) stable up to
908C. Upon humidification (two days at 100% RH), a partial transformation to
a double helix (B-polymorph) was observed. The authors concluded that the complexing agents were trapped in a matrix with crystal surfaces and dislocations
accommodating the complexing. Amylose complexes were found to be present
as partially ordered hydrophobically bonded structures. Smits et al. (536)
studied interactions of crystalline amylose and crystalline and amorphous amylopectin with plasticizers, such as glycerol or ethylene glycol, in the absence of
water by 13C DP and CPMAS experiments. The appearance of the plasticizer
in the CPMAS spectra suggested strong enough interaction to immobilize the
plasticizer molecules. It was suggested that the amylopectin/glycerol interaction
involved chemical exchange mechanism, in which the glycerol was continuously
moving through the mixture and interacting alternatively with different sites.
Heat treatment caused further immobilization of both glycerol and ethylene
glycol.
Kozerski et al. (537) studied genistein (isoflavone) complexes with amines
(morpholine and piperazine) by 13C CPMAS. Carbon chemical shift assignments
were accomplished by comparison with solution and by short contact time
CPMAS. The extent of proton transfer between the free acid and complex was
followed by carbon chemical shifts in the solid state. Most of the carbons exhibited rather small differences between the free acid and the complex carbon chemical shifts, typically smaller than +2 ppm. However, large shifts of 6.8 and
12.6 ppm were observed on one particular carbon for the two complexes, reflecting full proton transfer. This carbon was identified as being both donor and acceptor in the hydrogen bond network.
Zech et al. (538) characterized protein ligand interactions by 1D 13C
spectra and 2D 13C 13C correlation experiments using 13C-labeled protein.
Differential chemical shifts (free protein versus protein with bound ligand) are
used to distinguish between direct protein ligand contacts and small conformational changes of the protein induced by ligand binding. Dipolar-assisted
rotational resonance (DARR) (539) experiment was used to record the 2D
13
C 13C correlation spectra. DARR uses rotational resonance modified with
13
C 1H dipolar interaction (1H rf field fulfils rotary-resonance condition).
Lubach et al. (540) studied pharmaceutical solids in polymer matrices. Two
systems were studied: bupivacaine (free base and HCl salt) incorporated into tristearin and encapsulated using a solid protein matrix and 13C-labeled asparagine
522
Medek
1:250 w/w in PVP and poly(vinyl alcohol). Different polymorphic forms of

bupivacaine were generated upon heating or recrystallization. No polymorphic
changes of bupivacaine were observed upon formulating in tristearin, but the
tristearin itself was partially crystalized. The estimated limit of detection was
1:400. No 13C CPMAS spectrum was observed for the protein matrix, indicating
that the high mobility prevented efficient CP. Amorphous asparagine was
identified when lyophilized in the polymer matrices. The polymer peaks broadened as well.
Schachter et al. (541) studied drug-poly(ethylene oxide) (PEO) solid solutions by 13C CPMAS and variable temperature 1H MAS. Ketoprofen was
chosen as a model drug with poor water solubility. The hydrogen bonds
between the carboxylic group of ketoprofen and the ether oxygen of PEO were
the likely source of the high miscibility and formation of molecular dispersion.
Gjerde et al. (542) studied the anti-psychotic drug chlorpromazine (CPZ)
interaction with phosphatidylserines by 13C and 31P SSNMR. 31P spectra indicated the effect of acyl chain unsaturation on the anisotropic motion of the
charged serine head group, implying restricted motions by intermolecular
rather than intramolecular effects.
Apperley et al. (543) studied interaction of sildenafil citrate (Viagra) with
water vapor. Unlike SSNMR, PXRD data showed no significant differences
between samples equilibrated at different RH. The overall pattern of the
SSNMR spectra as a function of RH remains unchanged, strongly suggesting
only small changes of unit cell as water is absorbed or removed from the
crystal lattice. Since only one set of 13C CPMAS peaks is observed, water was
said to move rapidly inside the lattice. The major changes were observed in
the citrate anion and the drugs propyl group (Fig. 72). The shift patterns were
consistent with formation of weak hydrogen bond. The site of protonation was
studied by 15N CPMAS. The lack of observing the characteristic 14N 15N
residual dipolar splitting of the two neighboring nitrogens in the equilibrium
state at high humidity (88% RH) was interpreted as a signature of the water
dynamics-induced 14N-enhanced relaxation. (The splitting is observed for
anhydrous samples.) Comparison with solution- and solid-state dipolar dephasing
editing was used to assign nitrogen resonances.
Bogdanova et al. (544) studied experimentally and by modeling the interactions of PVP with ibuprofen (IBP) and naproxen (NAP). A spontaneous conversion of PVP/IBP physical mixtures into a stable glass-like form (solid
dispersion) was observed after storage. The enantiomer was found to react
more strongly than the racemate. NAP did not interact with PVP. This was
related to the differences in the drug crystal structures.
b-Cyclodextrin Complexes
A special class of drug excipient interactions is the interaction of API with
inclusion compounds, such as b-cyclodextrin (bCD). It is well known that the
SSNMR Spectrometry
523
Figure 72 The 13C CPMAS spectra of the sildenafil citrate stored at different conditions.
Abbreviation: CPMAS, cross-polarization magic angle spinning. Source: Modified from
Ref. 543.
structure of bCD and its polymers give rise to remarkable ability to form
inclusion complexes with organic molecules, especially aromatic compounds,
through host guest interactions.
Schneider et al. (545) wrote a review on NMR studies of cyclodextrins and
their complexes. It concentrates mostly on solution NMR studies, but a brief
review of SSNMR of cyclodextrins is also included.
Kitchin and Halstead (546) studied methyl group dynamics in aspirin and
aspirin bCD complexes by 2H SSNMR. Variable temperature 2H spectra
and variable temperature 2H inversion recovery experiments were recorded,
and the activation energies of the aspirin-d3 methyl group rotation and its correlation times were calculated. The authors found the aspirin-d3 methyl group 2H
electric field gradient (EFG) tensor unaffected by the inclusion in bCD. They
found the activation energy and the spin-lattice relaxation anisotropy smaller
than for pure aspirin-d3, suggesting that the rotation of the methyl group in the
complex is less hindered than in the aspirin-d3.
Crini et al. (547) studied bCD polymers cross-linked with epichlorohydrin.
The 13C MAS and CPMAS spectra were recorded as a function of temperature to
524
Medek
overcome the spectral overlap of the polymerized epichlorohydrin with bCD

peaks. The MAS spectra showed only the mobile components.
Garbow et al. (548) studied structure, dynamics, and stability of bCD
inclusion complexes of aspartame and neotame. This was, in part, a solutionphase study showing that the inclusion complex formation accounts for the
observed stability enhancement in solution. Studies of inclusion complexes in
lyophilized solids provided an additional insight. The 13C CPMAS line widths
of the inclusion complex were broad, suggesting more amorphous environment
of the sweetener, which was consistent with the formation of a true inclusion
complex. Single T1rH relaxation time of the sweetener was observed for the
inclusion compound, which was much shorter than for the pure sweetener and
consistent with a presence of the single phase.
Saalwachter (549) studied chain dynamics and order of gCD inclusion
compound with poly(dimethylsiloxane) by 1H and 13C SSNMR. Even at
30 kHz MAS spinning, the 1H resonances of the individual protons were not
resolved, indicating essentially rigid dipolar coupling environment. In contrast,
signals from OH/H2O hydrogen bonding region were sharp due to the
dynamics-induced dipolar decoupling. This was confirmed by 1D DQ-filtered
1
H MAS. Assignment of the coupling partners was done by 2D SQ/DQ
correlation. REPT-HDOR experiment was used to quantitatively determine
1
H 13C dipolar coupling constant of the methyl groups. Two complimentary
ways of extracting the coupling constant were employed: following of the
REPT-HDOR intensity build-up and its sideband analysis. The dipolar couplings
obtained from the REPT-HDOR technique were found to be influenced less by
the remote couplings in the abundant proton spin bath than dipolar couplings
obtained from BABA analysis of spinning sidebands. The order parameter for
the methyl group, which is related to the ratio of the experimentally observed
and static (rigid limit) dipolar couplings, was found to be 0.72. All methyl
groups were found to be in the same rotation state. The definite validation of
the rotation model was not possible from the NMR data alone. This work represents the first quantitative SSNMR investigation of local segmental fluctuations
and order without the need for isotopic enrichment.
Wulff et al. (550) studied indomethacin complexes with cyclodextrins dispersed in PEG 6000 carrier by 13C CPMAS. Complexation was identified for
bCD and gCD but not for aCD. The distribution of chemical shift for the
bCD complex was attributed to the possible formation of different types of complexes. Different parts of the indomethacin molecule could bind to bCD and,
thereby, give several types of complexes. The change in chemical shifts of the
gCD in the dispersion indicated a bond between gCD and the drug. A crystallinity of the dispersion was measured by 1H SSNMR as a ratio between the broad
peak of the amorphous PEG and the total area of the peak (broad amorphous plus
sharp crystalline). The dispersions were found to be less crystalline compared
with the pure polymer carrier.
SSNMR Spectrometry
525
Lu et al. (551) studied molecular motions in the supramolecular complexes

between poly(1-caprolactone) PEO poly(1-caprolactone) and aCD and gCD.
Guest host magnetization exchange was observed by T1H, T1rH, and 2D heteronuclear correlation experiments. The CP dynamics of the inclusion complex was
studied. The identical T1H relaxation of the components of the inclusion complex
. Spin
pointed to the homogeneity of the domains on the length scale of 334A
diffusion was studied with 2D heteronuclear correlation experiment, using
BLEW-12/BB-12 homonuclear decoupling during 1H evolution and WIM-24
during CP to quench the 1H 1H spin diffusion. T1C was used to study high-frequency motions (MHz) and line shape analysis of 2D WIM/WISE was used to
investigate low-frequency motions (kHz). Temperature dependence of T1rC
was also measured, but no T1rC minimum was observed, making the interpretation for kHz range dynamics difficult.
Yannakopoulou et al. (552) studied complex of (Z)-tetradec-7-en-1-al with
bCD. Two different guest environments were detected by splitting of carbonyl
and methyl groups. The guest molecules were found to be tightly trapped in
the bCD torus and, therefore, resist oxidation. Two bCD molecules were
found to form head-to-head dimmers that enclose one guest molecule disordered
over two sites and packed in channels.
Cunha-Silva and Teixeira-Dias (553) studied inclusion of nonionic surfactant in bCD by 13C CPMAS. Carbon multiplicity was reduced for the inclusion
compounds studied, pointing to appreciable symmetrization of the bCD macrocycle. Two groups of spectral line shapes were observed when carbon spectra
were recorded as a function of humidity. The spectra of sample equilibrated
below 20% RH were different than those above 20% RH. Water seemed to contribute in a relevant way to the bCD macrocycle symmetrization.
Braga et al. (554) studied an inclusion complex of bCD with p-hydroxybenzaldahyde. The free bCD exhibited multiple resonances for each type of
carbon, which correlated with different torsion angles of the sugar linkages.
The multiplicity was reduced to single peaks upon complexation, suggesting a
symmetry increase upon inclusion. The same lab studied inclusion compound
of S-IBP in bCD by 13C CPMAS (555). Tight encapsulation was found in
which the bCD adopted a more symmetrical conformation in the inclusion
compound with each glucose unit in a similar environment. The same lab
studied encapsulation of sodium nimesulide and its precursors in bCD (556).
13
C CPMAS confirmed that no chemical modification of the guests occurred
upon formation of the inclusion complexes. The same lab also studied the humidity
effect on 2-phenoxyethanol bCD inclusion compound (557). At RH higher
than 20%, decreased multiplicities of the carbon resonances and diminished dispersion of chemical shifts was found, pointing to an improved symmetrization
of the bCD macrocycle in the channel structure of this inclusion compound.
Below 20% RH, the carbon spectra showed broad lines, characteristic of amorphous phase.
526
Medek
Lai et al. (558) studied imazalil bCD inclusion complex by 13C CPMAS.
Authors also observed a convergence of the multiple resonances of the carbons
of the glucose units to the single peak between bCD and the complex. The imazalil peaks were broadened, following encagement into the hydrophobic cavity.
Some of them split into multiple signals, indicative of a pronounced structural
rearrangement of the imidazole and aryl rings. Carbon-detected rotating frame
proton relaxation T1rH revealed the sample to be homogeneous. No change of
mobility of the sugar rings in the bCD skeleton was detected by comparison of
the free bCD T1rH with those of the inclusion compound. However, T1rH of imazalil decreased remarkably between free and complexed imazalil.
Excipient Characterization
Excipients can be characterized by SSNMR in a similar way a drug substance.
The same suite of techniques is available and similar types of problems can be
answered. Sugars appear as components of many excipient blends. Many
examples of characterization of carbohydrates can be found in the literature
(52,559,560).
Tanner et al. (561) studied crystalline forms of a- and b-chitin. Remarkably narrow carbon lines of a variety of chitin (polysaccharide similar to cellulose) samples were observed. Unlike previous reports, only a single molecule
per asymmetric unit cell was observed. Spectra of at least three different forms
of b-chitin were identified, differing in the hydration states. Spectra of the
poorly ordered squid pen b-chitin were interpreted by using the spectra of the
highly crystalline samples. The a-chitin spectra were found to be broad and
hard to interpret. T1C filter was used to filter out fast relaxing peaks, proving a
mixture character of some samples.
Kunze et al. (562) characterized cellulose and cellulose ethers. Qualitative
and quantitative analysis is shown, including quantitative characterization of
the homogeneity and structural selectivity of substitution. Spectra of the cellulose
polymorphs, cellulose-I (Ia and Ib), cellulose-II, Na-cellulose-I, NA-cellulose-II,
and amorphous are shown, as are the interconversions between the cellulose
forms upon the addition of NaOH. Mixtures of the forms are quantified
without taking the CP dynamics into account. A degree of substitution could
be quantified by choosing a resolved pair of carbon lines and a particular
contact time, assuring equal intensity response.
Lefort et al. (563) assessed SSNMR and DSC methods for quantifying the
amorphous content in solid dosage forms of ball-milled trehalose (Fig. 73).
Samples with different amorphous fractions were prepared as physical mixtures
of pure amorphous (obtained by mechanical milling) and crystalline components.
After calibration on the physical mixtures, both methods were applied to study
the trehalose amorphization during milling. NMR spectra for various lengths of
milling could be reproduced as linear combinations of the pure crystalline and
amorphous components. However, a different response was seen using DSC. The
SSNMR Spectrometry
527
Figure 73 The 13C CPMAS spectra of trehalose with different amorphous and crystalline contents. Abbreviation: CPMAS, cross polarization magic angle spinning. Source:
Modified from Ref. 563.
DSC peaks spread over a very broad range of temperatures, and their enthalpies
were smaller than expected. This is attributed to differences on a nanometer scale
or larger, which do not affect the NMR spectra. NMR can, therefore, remain a
successful technique for estimating the amorphous content, even if DSC might fail.
Atomic Distance Measurements
A determination of a three-dimensional structure of molecules by SSNMR relies
on the knowledge of distances between atoms. The 1/r 3 dependence of dipolar
coupling [Equation (13)] can be used to directly extract the distances with high
precision. Because the dipolar coupling is averaged out by fast MAS, it has to
be reintroduced by any of the numerous recoupling schemes. The various
recoupling techniques can restore either the heteronuclear or homonuclear
dipolar coupling and the pairwise distances between atoms can be extracted.
An isotope enrichment is typically required limiting the range of applications
to pharmaceutical materials. A selective labeling scheme can be used to
measure distances between the labeled nuclei. Procedures to measure multiple
distances simultaneously in uniformly labeled samples were also devised. Until
very recently, measurement of distances between protons were not successful,
mainly due to the poor available resolution in proton spectra. With the advancement of modern homonuclear decoupling techniques, this becomes less of an
issue. This opens up the possibility of solving crystal structures of small molecules based solely on SSNMR without the need for isotope enrichment (564).
528
Medek
Tong and Schaefer (565) characterized the interface of heterogeneous

blends of fluorinated and selective carbon-labeled polymers by measuring
13
C 19F distances using REDOR.
Fyfe et al. (566,567) compared 19F 29Si distances obtained from REDOR
and TEDOR techniques on static and MAS spun samples of clathrasil octadecasil. The distances, which were obtained from fitting of the oscillatory CP behavior, agreed well with those known from X-ray, and the distance accuracy of
0.004 nm was achieved. Fyfe applied the methodology to study 19F 29Si distances in other compounds. Bertani et al. used spin counting u-REDOR approach
to study octadecasil (568).
Middleton et al. (569) determined the molecular conformations of four
polymorphic forms of cimetidine from 13C SSNMR distance and angle measurements. The doubly 13C-enriched cimetidine was examined by 13C CPMAS. The
carbon spectrum of one of the four forms was consistent with two molecules per
asymmetric unit. Splitting of the carbon resonances due to the residual dipolar
coupling was evident. The distance between the two labeled carbons was determined from analysis of the rotational resonance magnetization exchange curves
resulting from the reintroduction of the weak 13C 13C dipolar coupling. Eightfold dilution by nonlabeled cimetidine was used to eliminate the inter-molecular
dipolar coupling. Difference spectroscopy was used to remove the natural abundance background. The carbon distances of two forms for which single crystal
structures were available were consistent with distances obtained from the
single crystal structure. The DQ heteronuclear local field (2Q-HLF) experiment
was used to determine the relative orientation of the 13C 1H bonds at the two
labeled carbons. The experimental distances and the angles were used to constrain the molecular conformation calculations. This approach demonstrated
that structures of polycrystalline compounds could be investigated in the
absence of diffraction techniques.
Gilchrist et al. (438,439) measured inter-fluorine distances (from 0.5 to
1.2 nm) in organic compounds by 19F recoupling methods. The RFDR method
was used and the exchange curves were simulated by Monte Carlo type algorithm (Fig. 74). An accuracy of 0.1 nm for short distances and 0.2 nm for long distances was achieved. The model compounds were selected based on four criteria:
(i) resolved resonances, (ii) known (rigid) distances, (iii) short relaxation times,
and (iv) dilution by matrix to prevent intermolecular dipolar coupling. The
observed fluorine peaks were approximately 1 ppm wide, dominated by inhomogeneous contributions and not limited by proton decoupling (.100 kHz).
Using modified CPMG sequence, Salgado et al. (570) studied membranebound structure of a protein derived from angular and distance constraints from
homonuclear 19F dipolar coupling. A good agreement with the known structure
was achieved. The chemical shift anisotropy of the macroscopically oriented
sample enabled to determine the orientation of the peptide in the membrane bilayer.
Middleton et al. (571) used conformational analysis by SSNMR to aid
restrained structure determination form PXRD. A molecular conformation is
SSNMR Spectrometry
529
Figure 74 The 19F 19F distance measurements using RFDR experiment. Experimental
and simulated RFDR exchange curves are shown. Abbreviation: RFDR, radiofrequencydriven re-coupling. Source: Modified from Ref. 438.
first determined approximately from a set of interatomic distances measured by

REDOR. The conformation is then optimized against PXRD data using simulated
annealing protocol. The 15N 13C REDOR distances were obtained on
15
N-labeled (and naturally abundant carbon) cimetidine. CPMAS with selective
CH/CH2 transfer was used to reduce spectral overlap (117). Restrained molecular dynamics was applied to obtain the molecular conformation from the SSNMR
data, followed by PXRD optimization. With the combined SSNMR/PRXD
approach, the problem of solving structures is transformed into a rigid
body search.
Mehta et al. (572) applied REDOR sequences to measure carbon fluorine
and deuterium fluorine distances on lyophilized bacterial whole cells
(selectively 13C-labeled), complexed with 4-fluorobiphenyl derivative of chloroeremomycin. The position of the drug was determined relative to the peptidoglycan stem and, therefore, did not require an absolute measurement of the
REDOR full echo.
Grage et al. (573) applied 2Hf19Fg REDOR to study distances in biological
solids using double resonance probes without proton decoupling. The lack of
proton decoupling was found only to influence the accuracy of the obtained distances. The maximum distances, thus have to be smaller than 0.8 nm. Composite
530
Medek
pulses were used on the fluorine channel, which greatly reduced the fluorine offresonance effects.
Elena and Emsley (564) applied 1H 1H spin diffusion CRAMPS experiment to characterize structure and the crystallographic unit-cell parameters of
naturally abundant small molecules. The spin diffusion was recorded using
high-resolution 2D 1H 1H correlation experiment with eDUMBO-1 (102) homonuclear decoupling in both t1 evolution and t2 detection periods. The eDUMBO-1
afforded high enough resolution that a kinetic rate matrix model could be used to
characterize the spin diffusion build-up curves between the single proton sites
(Fig. 75). The comparison between the experimental data and simulations on
Figure 75 The 2D 1H 1H CRAMPS correlation experiment with variable time spin

diffusion mixing period using eDUMBO-1 homonuclear decoupling (top left). The spin
diffusion build-up curves together with their best fits (top right). The intra- and intermolecular dipolar coupling network of a single alanine CHa (bottom). Abbreviation:
CRAMPS, combined rotation and multiple-phase spectroscopy. Source: Modified from
Ref. 564.
SSNMR Spectrometry
531
model dipeptide depended strongly on the crystal structure parameters, such as

unit cell and the orientation of the molecule within it.
H-Bonds and Proton Transfer

The knowledge of H-bonds and proton transfer offer an important information
about structure of the compound. Single-crystal X-ray diffraction is considered
to provide the ultimate knowledge of H-bonds. However, because the positions
of protons are inferred from X-ray data only indirectly, much remains to be
learnt even when the single crystal structure is available. SSNMR is an excellent
tool to study H-bonds and proton transfer. There are several different ways to gain
knowledge on proton transfer by SSNMR. Chemical shielding is very sensitive to
local distribution of electron densities, and the presence of H-bond is well
reflected by sizeable chemical shifts. The calibrated chemical shift scales can
be applied to precisely estimate the bond lengths of unknown samples. The
dipolar coupling interaction provides another way to directly measure distances
between nitrogens and protons with high precision. A picometer precision has
been achieved (574). A definite confirmation of the presence of H-bonds can
be obtained by observation of cross-peaks in correlation experiments based on
J-coupling mechanism. The values of the H-bond J-coupling correlates well
with the H-bond strength.
Limbach published a chapter on Studies of Elementary Steps of Multiple
Proton and Deuteron Transfers in Liquids, Crystals and Organic Glasses (575).
The focus of this study was put on NH. . .N proton transfer because (i) it is slower
than OH. . .O, and (ii) 15N is spin 12 that can be followed easily by NMR. One of
the major conclusions was that the proton transfer systems in the crystalline and
amorphous states are subject to static perturbations of the reaction energy profiles
of the proton motion.
McGregor et al. (76) studied the fast proton transfer in a series of substituted tetra-azaannulenes. The number of observed 13C CPMAS peaks for all compounds except one was in accord with their symmetry as determined by crystal
structures. The exception was the tetra-azaannulene, for which less than the
expected number of lines were observed as a consequence of fast proton transfer
accompanied with rapid molecular rotation. On cooling, a peak broadening was
observed for this sample, as a result of the interference of the proton transfer with
proton decoupling frequency of 60 kHz. The evidence from carbon spectra is in
full agreement with previously reported 15N data.
Hickman et al. (576) studied protonic conduction in imidazole by 15N
SSNMR. The 2D 15N exchange study demonstrated that the conduction mechanism did not involve the reorientation of the imidazole ring.
Wu et al. (577) studied the intermolecular H-bonding effects on the amide
17
O electric field gradient and chemical shift tensor of benzamide. The 17O MAS
and stationary spectra were fitted to extract the 17O quadrupolar and chemical
532
Medek
shift parameters. Quantum mechanical calculation provided a way to relate these

tensors to molecular structure.
Song and McDermott (578) studied proton transfer dynamics and N H
bond lengthening in N H. . .N model systems by variable temperature 15N
CPMAS and 2D exchange experiments. Based on the nitrogen line shapes,
short N. . .N distances and fast proton transfer through the N H. . .N bridge
was observed.
Zhao et al. (579) devised a new method for quantification of H-bonds by
measurements of NH bond lengths using 2D 1H 15N dipolar-chemical shift
correlation (with recoupling of the H N dipolar couplings) combined with fast
MAS. From the experimental 1H 15N dipolar coupling, the HN lengths were
calculated with accuracy of less than 0.01 nm.
Kiss et al. (580) studied H-bonded networks of methoxy-substituted
a-phenylcinnamic acids by spectroscopic and computational methods. It was
possible to detect the aromatic C H. . .O H-bonds responsible for the long-range
ordering. An absence of olefinic H-bonds was confirmed by 13C CPMAS spectra.
Molecular modeling provided an additional proof for existence (or absence) of
H-bonds.
Huelsekopf and Ludwig (581) correlated structural information, NMR, and
IR techniques of N-methylacetamide. Nonadditive contribution of H-bonds was
found to play a significant role. Calculated predictions were compared with
experimental data and reasonable agreement was obtained.
Henry et al. (582) studied solid-state acid base and tautomeric equilibria
of lyophilized L -histidine and determined pH and pK values. A separate observation of all three acid base pairs in the successive deprotonation of the carboxylic group was possible. The calculated solid-state pK values based on the
13
C CPMAS observed acid/base ratios were found similar to solution. Noninvasive in situ measurement of solid-state acidity was therefore possible. Differences
in the 1H CRAMPS spectra were also obsereved.
Foces-Foces et al. (583) studied H-bond structure and dynamics of pyrazole-4-carboxylic acid chains using 15N CPMAS and other techniques. At low
temperatures, NMR showed a proton disorder-order transition with localized
protons. The proton transfer equilibrium detected by nitrogen NMR was found
to be a sensitive tool for studying molecular structure that could not easily be
detected by X-ray methods.
Totz et al. (584,585) studied the mechanism of proton conductivity in
quasi-one-dimensional H-bonded crystals by 1D and 2D exchange 2H NMR.
The slow chemical exchange processes between H-bridges were studied over a
wide temperature range. The ionic conductivity was realized by jumps of the
protons between the different H-bonds. The proton mobility measured from the
2D experiments was related to the measured conductivity.
Takeda and Tsuzumitami (586) noted the slowing down of the proton transfer in H-bonds of benzoic acid crystals. Transfer rates were investigated by a
temperature dependence of T1H and T1D. The results indicated that H and D
SSNMR Spectrometry
533
pairs transfer in a concerted manner within the timescale of NMR (1026 to 10210
seconds).
Lorente et al. (587) studied proton and nitrogen chemical shielding (by line
shape simulation), dipolar coupling (providing distances), and H-bond geometry
correlations in a series of H-bonded acid base complexes of collidine with carboxylic acid. The isotropic 15N chemical shifts were found to be characteristic of
N H distances and could be used for evaluating H-bond geometries and solidstate acidities.
Goward et al. (588) investigated NH H-bond in solid benzoxazine dimer by
1
H 15N correlations under fast MAS. The proton was found to be shared between
the nitrogen and oxygen atoms. Different designs of REDOR-type of recoupling,
including the sensitive inverse detection on proton channel, of the heteronuclear
1
H 15N dipolar interactions were used. The spinning sidebands yielded the
values of dipolar coupling and hence the N. . .H distances. Relatively long
distances could be measured.
Ono et al. (589) studied H-bonding of amino acids and peptides by variable
temperature 2H static NMR. Parameters of the quadrupolar tensor were extracted
by simulations using the 2H dynamic NMR theory and correlated with N. . .O
H-bond length.
Colsenet et al. (590) devised a straightforward detection of deprotonated
conformers of malonic acid. By using slow spinning 13C CPMAS spectra of
samples lyophilized from solutions of different pH, different signals were
observed for protonated and deprotonated species as a result of dramatic slowdown of the inter- and intramolecular proton exchanges. A planar intermolecularly H-bonded monocation was observed for the first time.
Brown et al. (216,591) determined the H-bond strengths by a quantitative
measurement of homonuclear J-couplings. The existence of the through H-bond
correlation peak in 15N INADEQUATE confirmed the first ever directly detected
solid-state H-bonding. The 2D MAS echo technique is used to accurately determine the through H-bond JNN coupling strength, even in the cases in which no
peak splitting is actually observed and, hence, the strength of the H-bond can
be determined.
Lesage et al. (592) observed a H-agostic bond in a heterogeneous catalyst
using 2D SSNMR. The authors used a variant of the SS-APT through bond
J-coupling based technique, where t1 evolution was incorporated to make it an
2D technique. By correlating the carbon chemical shifts with 1JCH coupling,
the direct evidence for the H-agostic bond was provided.
Ilczyszyn and Ilczyszyn (593,594) studied betain-sulfamic acid (2:1)
crystal by 13C CPMAS and other techniques. It was found for the first time
that the Raman and shielding tensors are strongly correlated. Very short and symmetric H-bonds were formed by acidic protons and betaine carboxylate groups.
Pawsey et al. (595) studied H-bonding interactions in self-assembled
monolayers by 1H fast MAS SSNMR. The 2D HETCOR and 2D 1H double
quantum correlations were extensively used.
534
Medek
Emmler et al. (596) devised a simple method for the characterization of

OHO hydrogen bonds by 1H rotor-synchronized echo SSNMR, which afforded
higher resolution than the standard fast MAS spectra and the proton chemical
shifts could be faithfully extracted. Using empirical valence bond order model,
the experimental 1H shifts were correlated with the H-bond geometries for a
broad range of substances. For strong H-bonds, the deviations from the correlation curve are explained in terms of proton tautomerism. By applying the
proton echo technique, it is now possible to routinely determine the position of
the proton in a H-bond.
Leppert et al. (597) identified NH. . .N H-bonds in double-stranded RNA by
2D 15N dipolar-chemical shift correlation spectra. The presence of H-bond was
reflected in the cross-peaks between the donor and acceptor.
Schilf et al. (598) used 15N and 13C NMR to study intramolecular H-bonds
in eight Schiff bases. The nitrogen chemical shifts of OH and NH tautomers
were found especially useful to characterize the tautomerism. Four compounds
were found to exist as OH tautomers, whereas the remaining four existed
as NH tautomers with different stages of H transfer. This was confirmed by
13
C CPMAS. The position of protons in intramolecular H-bonds were found
to be strongly determined by substituents in aromatic ring. In other studies of
Schiff bases from the same lab, six out of eight compounds were found
to form intramolecular H-bonds (599 604). The 15N shifts and 1JNH were used
to estimate the tautomeric equilibrium.
Li et al. (605) studied acid base characteristics of bromophenol blue-citrate
buffer systems in the lyophilized amorphous state. Raman and 13C CPMAS
SSNMR spectroscopies were used to monitor the ratio of ionized to unionized
citric acid under various conditions as a function of pH. Downfield solid shifts
were observed with increasing solution pH especially for the carboxylic carbons.
Chekmenev et al. (606 608) correlated nitrogen chemical shielding in
labeled cyclic tripeptides (determined from MAS and stationary spectra) with
X-ray structure. 1H 15N and 13C 15N dipolar couplings were used to accurately
orient the chemical shielding tensor in the molecular frame (within 18). The
isotropic shifts were found dependent on the H-bonding and local conformation.
Diez-Pena et al. (609) studied the effect of H-bonding on the collapse of
PMAA hydrogels by advanced 1H SSNMR techniques. The 1H MAS and 2D
BABA-based correlation experiments are shown. Proton distances are obstained
from the fitting of BABA spinning sidebands. Dependence on pH was also
studied. Different populations of three COOH species (free mobile acid and
two H-bonded dimers with different stabilities) were found to depend on pH
and temperature.
Kawashima (610) studied coals soaked in pyridine by 15N NMR. Two
states of pyridine were observed: one with somewhat limited mobility and
weak H-bonded pyridine and a second with even less mobility and strong
H-bonded pyridine. The interaction between pyridine and coal was not strong
enough to be observed by CPMAS.
SSNMR Spectrometry
535
Benhabbour et al. (611) studied imidazole-based proton-conducting composite materials to provide an insight into the mechanism of proton transfer.
Application of variable temperature 1H high-speed MAS and DQ filtered 1H
NMR using BABA sequence identified mobile protons. Comparison of macroscopic and microscopic mobilities are presented.
Gobetto et al. (612) studied H-bonded supramolecular adducts using 1H
MAS and 15N CPMAS. Proton and nitrogen chemical shifts allowed a distinction
between the intermolecular H-bonds of the N H. . .O2 (with proton transfer)
and N. . .H O (no proton transfer). Correlations between the chemical shifts
and N O distances were found. The localization of H-bonds by DFT calculations
was in reasonable agreement with experimental data. The authors predict this
approach to be especially fruitful for studies of partially disordered systems
that cannot be studied by X-ray methods.
Schulz-Dobrick et al. (574) determined the geometry of H-bonds in solids
with picometer accuracy by combination of NMR and quantum chemical calculations. The 1H 15N dipolar recoupling techniques (REREDOR- and HDORbased) can determine HN distances up to 250 pm with accuracy of +1 pm and
+5 pm for short (100 pm) and long distances, respectively. The structures were
computationally optimized based on the HN distance constraints and checked
for agreement between the experimental and predicted chemical shifts. With a
precision of 1 pm, the vibrational averaging had to be considered, which reduced
the measured dipolar coupling and, thus, the apparent distances derived from it.
Li et al. (613) studied acid base interactions in complexes of heterocyclic
bases and dicarboxylic acids. Combination of 15N CPMAS and ab initio predictions showed the sensitivity of nitrogen chemical shifts to serve as a reliable
probe for the degree of proton transfer in realistic systems (as opposed to
models), ranging from no proton transfer (free base) through partial proton
transfer (neutral crystal complexes or cocrystals) to full proton transfer
(salts; Fig. 76).
Food and Cosmetic Applications
This short section on SSNMR applications in food and cosmetic sciences is
included based on similarities of the problems studied and the similar limitations
on obtaining isotopically labeled materials. A distinction between crystalline and
amorphous phases and domain characterizations are among the common SSNMR
applications to study the heterogeneous samples typically found in food and cosmetics materials. Because water plays a significant role in the physical and
chemical properties of these samples, the distinction between free and bound
water, mobility, and relaxation studies are frequent.
Waver and Zielinska (614) studied flavonoids by 13C CPMAS to characterize their conformations. Carbon resonances were assigned based on solution
shifts and solid spectral-editing techniques. Two molecules per asymmetric
unit were observed in one of the compounds. An orientation of the hydroxyl
536
Medek
N3
N4
N4
N3
N3
N4
N1 N5
N2
Freebase
N1, N5
H3C
N2
N4 N3
Sesqisuccinate
N4
CH3
N3
N4
N1
1
N2
4 NH
H3
N
O CH3
N5
Dimalonate
N2
N5
N1 N4
N3
N2
N1
N4
N3
Dimaleate
-50
N3
N1
-100
-150
-200
-250 ppm
-50
-100
-150
-200
-250 ppm
Figure 76 The 15N CPMAS spectra of CP-724,714 free base and salts (left). The progressive proton transfer (from free base to dimaleate) is evident from the chemical shifts of nitrogens N1 and N2. Short-contact time CPMAS (protonated nitrogens only; right). Abbreviation:
CPMAS, cross polarization magic angle spinning. Source: Modified from Ref. 613.
groups and the effects of hydrogen bonding were evaluated from the solution
solid differences of the carbon shifts.
Utility of SSNMR to study lyotropic materials formed from food
grade monoglyceride mixtures was demonstrated (615). Carbon chemical shift
anisotropy, as determined by MAS spinning sideband analysis, was used to
distinguish the phase after addition of glucose guest molecule. Deuterium DQ
spectroscopy was used to differentiate water in isotropic and anisotropic
environments.
Table 1
Nuclear Spin Hamiltonians

Cl
Interaction
Zeeman
2g
gI^ 1 B0
CS
gI^ ACS B0
I^ AJ S^
Dipolar
m0

2p g1 g2 h
Quadrupolar
eQ
2I(2I1)h
eQ
^ Q I^
IA
2I(2I1)h
rf irradiation
rf
2g
gB1 (t)I^
Chemical shift
J-coupling
Hl
m0
I^1
2p g1 g2 h
AD I^2
SSNMR Spectrometry
Table 2
Isotope
1
H
C
15
N
19
F
31
P
2
H
14
N
35
Cl
37
Cl
79
Br
17
O
23
Na
25
Mg
39
K
13
537
Pharmaceutically Relevant NMR Nuclei

Spin
Natural abundance (%)
Larmor frequency at 11.75T (MHz)
1/2
1/2
1/2
1/2
1/2
1
1
3/2
3/2
3/2
25/2
3/2
25/2
3/2
99.99
1.1
0.37
100.0
100.0
0.02a
99.6
75.8
24.2
50.7
0.04a
100.0
10.0
93.3
500.0
125.7
50.7
470.3
202.3
76.7
36.1
49.0
40.8
125.2
67.8
132.2
30.6
23.3
Isotope enrichment is necessary for NMR studies.

Abbreviation: NMR, nuclear magnetic resonance.
Emulsifier phases of cosmetic emulsions were studied by 1H spin diffusion

and P CPMAS (616). The 31P CPMAS was used to suppress the dissolved
ingredients. Even though the centrifugal forces were found to destroy the
emulsion, leading to changes in chemical shift, new peaks not present in the
direct polarization experiment were observed. The 1H spin diffusion was used
to identify bound and free water. A related previous paper (617) applied
dynamic 31P methods to the interfacially bound emulsifier to obtain details
about the molecular dynamics at the interface. The 31P T1 and T2 times indicated
a restricted motion, independent of the droplet size in the emulsions.
Calucci et al. (618) studied hydration effects on gluten dynamics by
proton and carbon relaxation. A resolution of the carbon spectra was affected
by the heterogeneous character of the gluten samples. Hence, detailed
interpretations of the spectra were not possible. Comparison between DP
and CPMAS experiments were used to qualitatively assess the differences in
the local chemical environment and in mobility. The signals enhanced in
CPMAS, when compared with DP, were considered to originate from rigid
domains. Further discrimination was achieved on the basis of different T1C
times, using different DP recycle delays, and on the basis of different T2H,
using delay between the proton 908 pulse and the CP transfer. The latter experiment was carried out with 40 ms delay, giving origin to a signal of only the
carbons with strong enough dipolar coupling to give CP but coupled to protons
with spin spin relaxation longer than tens of microseconds. Hydration was
found to cause an increased mobility of all protein moieties. Line shape of
31
538
Medek
static 1H spectra was fitted by the sum of Gaussian and exponential functions
and analyzed based on the premise that Gaussian and exponential functions
can be associated with 1H nuclei in very rigid and relatively mobile environments, respectively. No liquidlike water was observed as no peaks with few
Hertz line widths were seen. Partial assignment of the 1H MAS spectra was
made. T1H with both direct proton and indirect carbon detection, and T1rH
with indirect carbon detection, were used to assess the rigidity of the
various sample components and the domain sizes. In dry gluten, the proteins
scale by T1H,
were found not to mix with residual starch on the 200A
scale.
whereas T1rH showed heterogeneities within protein on 10 20 A
Hydration was found to mix the different protein fragments and the protein
with starch.
REFERENCES
1. Slichter CP. Principles of Nuclear Magnetic Resonance. 3rd ed. New York: SpringerVerlag, 1990.
2. Abragam A. The Principles of Nuclear Magnetism. Oxford: Oxford University Press,
1985.
3. Mehring M. High Resolution NMR in Solids. 2nd ed. Berlin: Springer, 1983.
4. Schmidt-Rohr K, Spiess HW. Multidimensional Solid-State NMR and Polymers.
New York: Academic Press, 1994.
5. Haeberlen U. Advances in Magnetic Resonance. In: Waugh JS, ed. New York:
Academic Press, 1976.
6. Harris RK, Mann BE. NMR and the Periodic Table. New York: Academic Press,
1970.
7. Maciel GE. Science 1984; 226:282.
8. Oldfield E, Kirkpatrick RJ. Science 1985; 227:1537.
9. Turner GL, Kirkpatrick RJ, Risbud SH, Oldfield E. Am Ceram Soc Bull 1987;
66:656 663.
10. Andrew ER, Bradbury A, Eades RG. Nature 1958; 182:1659.
11. Lowe IJ. Phys Rev Lett 1959; 2:285.
12. Maricq MM, Waugh JS. J Chem Phys 1979; 70:3300 3316.
13. Hartmann SR, Hahn E. Phys Rev 1962; 128:2042.
14. Pines A, Gibby MG, Waugh JS. JCP 1973; 59:569.
15. Llor A, Virlet J. Chem Phys Lett 1988; 152:248 253.
16. Chmelka BF, Mueller KT, Pines A, Stebbins J, Wu Y, Zwanziger JW. Nature 1989;
339:42 43.
17. Mueller KT, Sun BQ, Chingas GC, Zwanziger JW, Terao T, Pines A. J Magn Reson
1990; 86:470 487.
18. Frydman L, Harwood JS. J Am Chem Soc 1995; 117:5367 5368.
19. Medek A, Harwood JS, Frydman L. J Am Chem Soc 1995; 117:12779 12787.
20. Pratum TK, Klein MP. J Magn Reson (1969 1992) 1989; 81:350 370.
21. Rajamohanan PR, Thangaraj A, Suryavanshi PM. Ganapathy S. Magn Reson 1991;
272 279.
22. Hill EA, Yesinowski JP. J Chem Phys 1997; 107:346 354.
SSNMR Spectrometry
539
23. Jeschke G, Jansen M. Angewandte Chemie, International Edition 1998; 37:

1282 1283.
24. Wi S, Frydman L. J Chem Phys 2000; 112:3248 3261.
25. Geffroy PM, Mabilat F, Bessada C, Coutures JP, Massiot D. Mater Sci Forum 2000;
325 326, 319 324.
26. Wi S, Frydman L. J Am Chem Soc 2001; 123:10354 10361.
27. Jakobsen HJ, Bildsoe H, Skibsted J, Giavani T. NATO Sci Ser, II: Mathematics,
Physics and Chemistry 2002; 76:43 55.
28. Bachiller PR, Ahn S, Warren WS. J Magn Reson Ser A 1996; 122:94 99.
29. Ahn S, Lee S, Warren WS. Mol Phys 1998; 95:769 785.
30. Warren WS. Science (Washington, D.C.) 1998; 280:398 399.
31. Ahn S, Lisitza N, Warren WS. J Magn Reson 1998; 133:266 272.
32. Ernst RR, Bodenhausen G, Wokaun A. Principles of Nuclear Magnetic Resonance in
One and Two Dimensions. Oxford: Clarendon, 1987.
33. Gerstein BC, Chow C, Pembleton RG, Wilson RC. J Phys Chem 1977; 81:595 599.
34. Taylor RE, Pembleton RG, Ryan LM, Gerstein BC. J Chem Phys 1979; 71:4541.
35. Bronnimann CE, Hawkins BL, Zhang M, Maciel GE. Analytical Chemistry 1988;
60:1743 1760.
36. Hafner S, Spiess HW. Solid State Nucl Magn Reson 1997; 8:17 24.
37. Hohwy M, Bower PV, Jakobsen HJ, Nielsen NC. Chem Phys Lett 1997;
273:297 303.
38. Lesage A, Duma L, Sakellariou D, Emsley L. J Am Chem Soc 2001;
123:5747 5752.
39. Brus J, Petrickova H, Dybal J. Monatshefte Fur Chemie 2002; 133:1587 1612.
40. Bielecki A, Kolbert AC, Levitt MH. Chem Phys 1989; 155:341 346.
41. Vinogradov E, Madhu PK, Vega S. Chem Phys Lett 2002; 354:193 202.
42. Vinogradov E, Madhu PK, Vega S. Chem Phys Lett 1999; 314:443 450.
43. Bosman L, Madhu PK, Vega S, Vinogradov E. J Magn Reson 2004; 169:39 48.
44. Madhu PK, Zhao X, Levitt MH. Chem Phys Lett 2001; 346:142 148.
45. Vinogradov E, Madhu PK, Vega S. Top Curr Chem 2004; 246:33 90.
46. Brittain HG, Bogdanowich SJ, Bugay DE, DeVincentis J, Lewen G, Newman AW.
Pharm Res 1991; 8:963 973.
47. Bugay DE. Pharm Res 1993; 10:317 327.
48. Holzgrabe U, Diehl BWK, Wawer I. J Pharm Biomed Anal 1998; 17:557 616.
49. Burn SR, Pfeiffer RR, Stowell JC. Solid-State Chemistry of Drugs. 2nd ed. SSCI,
Inc. 1999.
50. Duer Ann Rep NMR Spectrosc 2001; 43:1 58.
51. Smith ME. Nucl Magn Reson 2000; 29:251 315.
52. Potrzebowski MJ. In: Atta-ur-Rahman, ed. New Advances in Analytical Chemistry
2000, P1. 359 404.
53. Ye C, Ding S, McDowell CA. Annu Rep NMR Spectrosc 2000; 42:59 113.
54. Stephenson GA, Forbes RA, Reutzel-Edens SM. Advanced Drug Delivery Rev 2001;
48:67 90.
55. Bugay DE. Advanced Drug Delivery Rev 2001; 48:43 65.
56. Harris RK. Spec Publ-R Soc Chem 2001; 262:3 16.
57. Shapiro MJ, Goundarides JS. Biotechnol Bioeng 2001; 71:130 148.
58. Bryce D, Bernard GM, Gee M, Lumsden MD, Eichele K, Wasylishen RE. Canadian J
Anal Sci Spectrosc 2001; 46:46 82.
540
Medek
59.
60.
61.
62.
63.
64.
65.
66.
67.
Ando S, Harris RK, Scheler U. Encyclopedia Nucl Magn Reson 2002; 9:531 550.
Kolodziejski W, Klinowski J. Chem Rev 2002; 102:613 628.
Dybowski C, Bai S, Bramer SV. Anal Chem 2002; 74:2713 2718.
Auger M. J Anal Sci Spectrosc 2002; 47:184 189.
Reutzel-Edens SM, Bush JK. Am Pharm Rev 2002; 5:112 115.
Schmidt-Rohr K, Spiess HW. Annu Rep NMR Spectrosc 2002; 48:1 29.
Brown SP, Emsley L. Handbook of Spectroscopy 2003; 1:269 326.
Potrzebowski MJ. Eur J Org Chem 2003; 8:1367 1376.
Medek A. Pharmaceutical Applications of SSNMR. Vol. 5. Academic Press, Elsevier
Science, 2003.
Tishmack Patrick A, Bugay David E, Byrn Stephen R. J Pharm Sci 2003; 92:
441 474.
Offerdahl TJ, Munson EJ, Offerdahl TJ, Salsbury JS, Dong Z, Grant DJW, Schroeder
SA, Prakash I, Gorman EM, Barich DH, Munson EJ. In press, 109 112 .
Watts A. Nature Reviews Drug Discovery 2005; 4:555 568.
Aliev AE, Law RV. Nucl Magn Reson 2001; 30:214 310.
McGregor AC, Lukes PJ, Osman JR, Crayston JA. J Chem Soc, Perkin Trans 2 1995;
809 813.
Bennett AE, Rienstra CM, Auger M, Lakshmi KV. J Chem Phys 1995; 103:
6951 6958.
Gan Z, Ernst RR. Solid State Nucl Magn Reson 1997; 8:153 159.
Fung BM, Khitrin AK, Ermolaev K. J Magn Reson 2000; 142:97 101.
Ernst M, Zimmermann H, Meier BH. Chem Phys Lett 2000; 317:581 588.
Detken A, Hardy HE, Ernest M, Meier BH. Chem Phys Lett 2002; 356:298 304.
Khitrin AK, Fung BM, McGeorge G. Encyclopedia Nucl Magn Reson 2002;
9:91 98.
Ernst M, Samoson A, Meier BH. J Magn Reson 2003; 163:332 339.
Ashida J, Asakura T. J Magn Reson 2003; 165:180 183.
Ernst M. J Magn Reson 2003; 162:1 34.
Ernst M, Meier MA, Tuherm T, Samoson A, Meier BH. J Am Chem Soc 2004;
126:4764 4765.
Brauniger T, Wormald P, Hodgkinson P. Monatshefte fur Chemie 2002; 133:
1549 1554.
Paepe GD, Hodgkinson P, Emsley L. Chem Phys Lett 2003; 376:259 267.
Paepe GD, Giraud N, Lesage A, Hodgkinson P, Boeckmann A, Emsley L. J Am
Chem Soc 2003; 125:13938 13939.
Paepe GD, Lesage A, Emsley L. J Chem Phys 2003; 119:4833 4841.
Waugh JS, Huber LM, Haeberlen U. Phys Rev Lett 1968; 20:180.
Burum DP, Rhim WK. J Chem Phys 1979; 71:944 956.
Burum DP, Linder M, Ernst RR. J Magn Reson 1981; 44:173 188.
Cory DG. J Magn Reson 1991; 94:526 534.
Demco DE, Hafner S, Spiess HW. J Magn Reson Ser A 1995; 116:36 45.
Nanz D, Ernst M, Hong M, Ziegeweid MA, Schmidt-Rohr K. J Magn Reson Ser A
1995; 113:169 176.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
92.
93.
94.
95.
96.
SSNMR Spectrometry
541
97. Hafner S, Spiess HW. J Magn Reson Ser A 1996; 121:160 166.
98. Hafner S, Spiess HW. Concepts Magn Reson 1998; 10:99 129.
99. Sakellariou D, Hodgkinson P, Lesage A, Emsley L. Chem Phys Lett 2000;
319:253 260.
100. Yamauchi K, Kuroki S, Ando I. J Mol Struct 2002; 602 603:9 16.
101. Igumenova TI, A EM. J Magn Reson 2003; 164:270 285.
102. Elena B, de Paepe G, Emsley L. Chem Phys Lett 2004; 398:532 538.
103. Lesage A, Sakellariou D, Hediger S, Elena B, Charmont P, Steuernagel S, Emsley L.
J Magn Reson 2003; 163:105 113.
104. Lee M, Goldburg WI. Phys Rev Lett 1965; 20:180.
105. Rhim WK, Elleman DD, Vaughan RW. J Chem Phys 1973; 59:3740 3749.
106. Rhim WK, Elleman DD, Schreiber LB, Vaughan RW. J Chem Phys 1974; 60:
4595 4604.
107. Takegoshi K, McDowell CA. Chem Phys Lett 1985; 116:100 104.
108. Hohwy M, Nielsen NC. J Chem Phys 1997; 106:7571 7586.
109. Eden M, Levitt MH. J Chem Phys 1999; 111:1511.
110. Carravetta M, Eden M, Zhao X, Brinkmann A, Levitt MH. Chem Phys Lett 2000;
321:205 215.
111. Dixon WT. J Chem Phys 1982; 77:1800.
112. Dixon WT, Schaefer J, Sefcik MD, Stejskal EO, McKay RA. J Magn Reson
(1969 1992) 1982; 49:341 345.
113. Opella SJ, Frey MG. J Am Chem Soc 1979; 101:5854 5856.
114. Wu X, Zilm KW. J Magn Reson Ser A 1993; 102:205 213.
115. Wu X, Zilm KW. J Magn Reson Ser A 1993, 104, 119 122.
116. Wu X, Burns ST, Zilm KW. J Magn Reson Ser A 1994; 111:29 36.
117. Burns ST, Wu X, Zilm KW. J Magn Reson 2000; 143:352 359.
118. Rossi P, Subramanian R, Harbison GS. J Magn Reson 1999; 141:159 163.
119. Kumashiro KK, Niemczura WP, Kim MS, Sandberg LB. J Biomol NMR 2000;
18:139 144.
120. DeVita E, Frydman L. J Magn Reson 2001; 148:327 337.
121. Burum DP, Bielecki A. J Magn Reson 1991; 95:184 190.
122. Sethi NK. J Magn Reson 1991; 94:352 361.
123. Sangill R, Rastrup-Andersen N, Bildsoe H, Jakobsen HJ, Nielsen NC. J Magn Reson
Ser A 1994; 107:67 78.
124. Hu JZ, Harper JK, Taylor C, Pugmire RJ, Grant DM. J Magn Reson 2000;
142:326 330.
125. Hu JZ, Taylor CMV, Pugmire RJ, Grant DM. J Magn Reson 2001; 152:7 13.
126. Schmidt-Rohr K, Mao J-D. J Am Chem Soc 2002; 124:13938 13948.
127. Mao J-D, Schmidt-Rohr K. J Magn Reson 2003; 162:217 227.
128. Schmidt-Rohr K, Mao J-D. Chem Phys Lett 2002; 359:403 411.
129. Caldarelli S, Ziarelli F. J Am Chem Soc 2000; 122:12015 12016.
130. Lesage A, Steuernagel S, Emsley L. J Am Chem Soc 1998; 120:7095 7100.
131. Sakellariou D, Lesage A, Emsley L. J Magn Reson 2001; 19:40 47.
132. Duma L, Hediger S, Brutscher B, Boeckmann A, Emsley L. J Am Chem Soc 2003;
125:11816 11817.
133. Duma L, Hediger S, Lesage A, Emsley L. J Magn Reson 2003; 164:187 195.
134. Carravetta M, Zhao X, Bockmann A, Levitt MH. Chem Phys Lett 2003;
376:515 523.
542
Medek
135. Khitrin AK, Canlet C, Fung BM. Solid-State NMR Spectrosc 2001; 19:63 72.
136. Schmidt-Rohr K, Clauss J, Spiess HW. Macromolecules 1992; 25:3273 3277.
137. Gerardy-Montouillout V, Malveau C, Tekely P, Olender Z, Luz Z. J Magn Reson A
1996; 123(1):7 15.
138. Deazevedo ER, Hu W-G, Bonagamba TJ, Schmidt-Rohr K. J Am Chem Soc 1999;
121:8411 8412.
139. Deazevedo ER. J Chem Phys 2000; 112:8988 9001.
140. Reichert D, Baonagamba TJ, Schmidt-Rohr K. J Magn Reson 2001; 151:129 135.
141. Rossum BJv, Forster H, Groot HJMd. J Magn Reson 1997; 124:516 519.
142. Yao XL, Schmidt-Rohr K, Hong M. J Magn Reson 2001; 149:139 143.
143. Schmidt-Rohr K, Mao J-D. J Magn Reson 2002; 157:210 217.
144. Liu S-F, Mao J-D, Schmidt-Rohr K. J Magn Reson 2002; 155:15 28.
145. Mao J-D, Xing B, Schmidt-Rohr K. Environ Sci Technol 2001; 35:1928 1934.
146. Mao J-D, Hundal JS, Schmidt-Rohr K, Thompson ML. Environ Sci Technol 2003;
37:1751 1757.
147. Sozzani P, Simonutti R, Bracco S, Comotti A. Polym Prepr (Am Chem Soc, Div
Polym Chem) 2003; 44:297 298.
148. DeLacroix SF, Titman JJ, Haemeyer A, Spiess HW. J Magn Reson 1992;
97:435 443.
149. Smith J, MacNamara E, Raftery D, Borchardt T, Byrn S. J Am Chem Soc 1998;
120:11710 11713.
150. Harper JK, McGeorge G, Grant DM. J Am Chem Soc 1999; 121:6488 6496.
151. Barich DH, Orendt AM, Pugmire RJ, Grant DM. J Phys Chem A 2000; 104:8290
8295.
152. Barich DH, Facelli JC, Hu JZ, Alderman DW, Wang W, Pugmire RJ, Grant DM.
Magn Reson Chem 2001; 39:115 121.
153. Barich DH, Pugmire RJ, Grant DM, Iuliucci RJ. J Phys Chem A 2001; 105:6780
6784.
154. Strohmeier M, Orendt AM, Alderman DW, Grant DM. J Am Chem Soc 2001;
123:1713 1722.
155. Grant DM. Encyclopedia Nucl Magn Reson 2002; 9:73 90.
156. Barich DH, Hu JZ, Pugmire RJ, Grant DM. J Phys Chem A 2002;106:6477 6482.
157. Strohmeier M, Alderman DW, Grant DM. J Magn Reson 2002; 155:263 277.
158. Clawson JS, Strohmeier M, Stueber D, Orendt AM, Barich DH, Asay B, Hiskey MA,
Pugmire RJ, Grant DM. J Phys Chem A 2002;106:6352 6357.
159. Ma Z, Barich DH, Solum MS, Pugmire RJ. J Agric Food Chem 2004; 2:215 221.
160. Strohmeier M, Grant DM. J Magn Reson 2004; 168:296 306.
161. Elena B, Hediger S, Emsley L. J Magn Reson 2003; 160:40 46.
162. Shao L, Crockford C, Geen H, Grasso G, Titmann JJ. J Magn Reson 2004;
167:75 86.
163. Gullion T, Schaeffer J. J Magn Reson 1989; 81:196 200.
164. Guillon T, Schaefer J. Adv Magn Reson 1989; 13:57 83.
165. Gullion T, McKay RA, Schmidt A. J Magn Reson (1969 1992) 1991; 94:362 369.
166. Gullion T, Schaefer J. J Magn Reson (1969 1992) 1991; 92:439 442.
167. Li Y, Evans JNS. J Chem Phys 1994; 101:10211 10216.
168. Mueller KT. J Magn Reson Ser A 1995; 113:81 93.
169. Tong G, Pan Y, Dong H, Pryor R, Wilson GE, Shaefer J. Biochemistry 1997;
36:9859 9866.
SSNMR Spectrometry
543
170. Tong G, Schaefer J. Macromolecules 1997; 30:7522 7528.

171. Mehta AK, Hirsh DJ, Oyler N, Drobny GP, Schaefer J. J Magn Reson 2000;
145:156 158.
172. Jaroniec, Younge, Tienstra, Herzfeld, Hriffin. J Magn Reson 2000; 146:132 139 .
173. Nishimura K, Fu R, Cross TA. J Magn Reson 2001; 152:227 233.
174. Kim SJ, Cegelski L, Studelska DR, Oconnor RD, Mehta AK, Schaefer J.
Biochemistry 2002; 41:6967 6977.
175. Bennett AE, Rienstra CM, Lansbury PT, Griffin RG. J Chem Phys 1996;
105:10289 10299.
176. Chan JC. Chem Phys Lett 2001; 335:289 297.
177. Vogt, Mattingly, Gibson, Mueller. J Magn Reson 2000; 147:26 35.
178. Petkova AT, Tycko R. J Magn Reson 2002; 155:293 299.
179. Gullion T, Pennington CH. Chem Phys Lett 1998; 290:88 93.
180. Jaroniec CP, Tounge BA, Herzfeld J, Griffin RG. J Amer Chem Soc 2001;
123:3507 3519.
181. Jaroniec CP, Filip C, Griffin RG. J Am Chem Soc 2002; 124:10728 10742.
182. Vogt FG, Gibson JM, Mattingly SM, Mueller KT. J Phys Chem B 2003;
107:1272 1283.
183. Grey CP, Veeman WS. Chem Phys Lett 1992; 192:379 385.
184. Blumenfeld AL, Coster DJ, Fripiat JJ. Chem Phys Lett 1994; 231:491 498.
185. Yap AT-W, Forster H, Elliot SR. Phys Rev Lett 1995; 75:3946 3949.
186. Grey CP, Vega AJ. J Am Chem Soc 1995; 117:8232 8242.
187. Kao H-M, Grey CP. J Phys Chem 1996; 100:5105 5117.
188. Deng F, Yue Y, Ye C. Solid State Nucl Magn Reson 1998; 10:151 160.
189. Ba Y, Kao H-M, Grey CP, Chopin L, Gullion T. J Magn Reson 1998; 133:104 114.
190. Hughes E, Gullion T, Goldbourt A, Vega S, Vega AJ. J Magn Reson 2002;
156:230 241.
191. Sack I, Balzs YS, Rahimipour S, Vega S. J Am Chem Soc 2000; 122:12263 12269.
192. Fu R. Chem Phys Lett 2003; 376:62 67.
193. Sinha N, Schmidt-Rohr K, Hong M. J Magn Reson 2004; 168:358 365.
194. Saalwachter K, Spiess HW. J Chem Phys 2001; 114:5707 5728.
195. Saalwachter K, Graf R, Spiess HW. J Magn Reson 2001; 148:398 418.
196. Saalwachter K, Schnell I. Solid State Nucl Magn Reson 2002; 22:154 187.
197. Schnell J, Langer B, Sontjens SHM, Gendereen MHPV, Sijbesma RP, Spiess HW.
J Magn Reson 2001; 150:57 70.
198. Schnell I, Saalwachter K. J Am Chem Soc 2002; 124:10938 10939.
199. Ishii Y, Tycko R. J Magn Reson 2000; 142:199 204.
200. Ishii Y, Yesinowski JP, Tycko R. J Am Chem Soc 2001; 123:2921 2922.
201. Hong M, Yamaguchi S. J Magn Reson 2001; 150:43 48.
202. Schmidt-Rohr K, Saalwachter K, Liu SF, Hong M. J Am Chem Soc 2001;
123:7168 7169.
203. Reif B, Griffin RG. J Magn Reson 2003; 160:78 83.
204. Paulson EK, Morcombe CR, Gaponenko V, Dancheck B, Byrd RA, Zilm KW. J Am
Chem Soc 2003; 125:15831 15836.
205. Khitrin AK, Fung BM. J Magn Reson 2001; 152:185 188.
206. Verhoeven RA, Verel R, Meier BH. Chem Phys Lett 1997; 266:465 472.
207. Lesage A, Sakellariou D, Steuernagel S, Emsley L. J Am Chem Soc 1998;
120:13194 13201.
544
Medek
208. Lesage A, Emsley L. J Magn Reson 2001; 148:449 454.

209. Lesage A, Charmont P, Steuernagel S, Emsley L. J Am Chem Soc 2000;
122:9739 9744.
210. Hediger S, Lasage A, Emsley L. Macromolecules 2002; 35:5078 5084.
211. Hardy EH, Detken A, Meier BH. J Magn Reson 2003; 165:208 218.
212. Hardy ED, Verel R, Meier BH. J Magn Reson 2001; 148:459 464.
213. Baldus M, Iuliucci RJ, Meier BH. J Am Chem Soc 1997; 119:1121 1124.
214. Alonso B, Massiot D. J Magn Reson 2003; 163:347 352.
215. Lesage A, Auger C, Caldarelli S, Emsley L. J Am Chem Soc 1997; 119:7867 7868.
216. Brown SP, Perez-Torralba M, Sanz D, Claramunt RM, Emsley L. J Am Chem Soc
2002; 124:1152 1153.
217. Lesage A, Bardet M, Emsley L. J Am Chem Soc 1999; 121:10987 10993.
218. Fayon F, Saout GL, Emsley L, Massiot D. Chem Commun 2002; 16:1702 1703.
219. Sakellariou D, Brown SP, Lesage A, Hediger S, Bardet M, Meriles CA, Pines A,
Emsley L. J Am Chem Soc 2003; 125:4376 4380.
220. Cadars S, Lesage A, Emsley L. J Am Chem Soc 2005; 127:4466 4476.
221. Murata K, Kono H, Katoh E, Kuroki S, Ando I. Polymer 2003; 44:4021 4027.
222. Kono H, Erata T, Takai M. Macromolecules 2003; 36:5131 5138.
223. Kono H, Numata Y. Polymer 2004; 45:4541 4547.
224. Kono H. Biopolymers 2004; 75:255 263.
225. Kono H, Numata Y, Erata T, Takai M. Polymer 2004; 45:2843 2852.
226. Kono H, Numata Y, Erata T, Takai M. Macromolecules 2004; 37:5310 5316.
227. Kaji H, Schmidt-Rhor K. Macromolecules 2002; 35:7993 8004.
228. Sommer W, Gottwald J, Demco DE, Spiess HW. J Magn Reson Ser A 1995;
113:131 134.
229. Schnell J, Lupulescu A, Hafner S, Demco DE, Spiess HW. J Magn Reson 1998;
133:61 69.
230. Feike M, Demco DE, Graf R, Gottwald J, Hafner S, Spiess HW. J Magn Reson 1996;
122:214 221.
231. Brown SP, Zhu XX, Saalwaechter K, Spiess HW. J Am Chem Soc 2001;
123:4275 4285.
232. Tycko R, Dabbagh G. Chem Phys Lett 1990;173:461 465.
233. Lee YK, Kurur ND, Helmle M, Johannessen OG, Nielsen NC, Levitt MH. Chem
Phys Lett 1995; 242:304 309.
234. Karlsson T, Helmle M, Kurur ND, Levitt MH. Chem Phys Lett 1995; 247:534 540.
235. Ernst M, Bush S, Kolbert AC, Pines A. J Chem Phys 1996; 105:3387 3397.
236. Gregory DM, Wolfe GM, Jarvie TP, Sheils JC, Drobny GP. Mol Phys 1996;
89:1835 1849.
237. Gross JD, Warshawski DE, Griffin RG. J Am Chem Soc 1997; 119:796 802.
238. Baldus M. Solid State Nucl Magn Reson 1998; 11:157 168.
239. Gross JD, Costa PR, Griffin RG. J Chem Phys 1998; 108:7286 7293.
240. Hohwy M, Jakobsen HJ, Eden M, Levitt MH, Nielsen NC. J Chem Phys 1998;
108:2686 2694.
241. Dusold S, Sebald A. Annu Rep NMR Spectrosc 2000; 41:184 264.
242. Antzutkin NO, Levitt MH. J Magn Reson 2000; 147:147 151.
243. Takegoshi K, Nakamura S, Terao T. Chem Phys Lett 2001; 344:631 637.
244. Raya J, Bianco A, Furrer J, Briand J-P, Piotto M, Elbayed K. J Magn Reson 2002;
157:43 51.
SSNMR Spectrometry
245.
246.
247.
248.
249.
250.
251.
252.
253.
254.
255.
256.
257.
258.
259.
260.
261.
262.
263.
264.
265.
266.
267.
268.
269.
270.
271.
272.
273.
274.
275.
276.
545
Verel R, Ernst M, Meier BH. J Magn Reson 2001; 150:81 99.

Fujiwara T, Khandelwal P, Akutsu H. J Magn Reson 2000; 145:73 83.
Ishii Y. J Magn Reson 2001; 114:8473 8483.
Goobes G, Vega S. J Magn Reson 2002; 154:236 251.
Kristiansen PE, Mitchell DJ, Evans JNS. J Magn Reson 2002; 157:253 266.
Williamson PTF, Verhoeven A, Ernst M, Meier BH. J Am Chem Soc 2003;
125:2718 2722.
Takegoshi K, Nakamura S, Terao T. J Chem Phys 2003; 118:2325 2341.
Thieme K, Schnell I. J Am Chem Soc 2003; 125:12100 12101.
Karlsson T, Popham JM, Long JR, Oyler N, Drobny GP. J Am Chem Soc 2003;
125:7394 7407.
Bjerring M, Nielsen NC. Chem Phys Lett 2003; 382:671 678.
Dvinskikh SV, Zimmermann H, Maliniak A, Sandstrom D. J Magn Reson 2004;
168:194 201.
Matsuki Y, Akutsu H, Fujiwara T. Magn Reson Chem 2004; 42:291 300.
Verhoeven A, Williamson PTF, Zimmermann H, Ernst M, Meier BH. J Magn Reson
2004; 168:314 326.
Dvinskikh SV, Zimmermann H, Maliniak A, Sandstroem D. J Chem Phys 2005;
122:044512/044511 044512/044512.
Ramamoorthy A, Gierasch LM, Opella SJ. J Magn Reson Ser B 1995; 109:112 116.
Peersen OB, Groesbeek M, Aimoto S, Smith SO. J Am Chem Soc 1995; 117:7228
7237.
Fujiwara T, Sugase K, Kamosho M, Ono A, Akutsu H. J Am Chem Soc 1995;
117:11351 11352.
Gu Z, Ridenour CF, Bronnimann CE, Iwashita T, Mcdermott A. J Am Chem Soc
1996; 118:822 829.
Feng X, Eden M, Brinkmann A, Luthman H, Eriksson L, Graslund A, Antzutkin ON,
Levitt MH. J Am Chem Soc 1997; 119:12006 12007.
Verdegem PJE, Helmle M, Lugtenburg J, Groot HJMD. J Am Chem Soc 1997;
119:169 174.
Hong M, G Griffin, R. J Am Chem Soc 1998; 120:7113 7114.
Terao T. J Mol Struct 1998; 441:283 294.
Middleton DA, Robins R, Feng X, Levitt MH, Spiers ID, Schwalbe CH, Reid DG,
Watts A. FEBS Lett 1997; 410:269 274.
Ishii Y, Hirao K, Terao T, Terauchi T, Oba M, Nishiyama K, Kainosho M. Solid
State Nucl Magn Reson 1998; 11:169 175.
Tian F, Song Z, Cross TA. J Magn Reson 1998; 135:227 231.
Wilhelm M, Feng H, Tracht U, Spiess HW. J Magn Reson 1998; 134:255 260.
Williamson PTF, Grobner G, Spooner PJR, Miller KW, Watts A. Biochemistry
1998; 37:10854 10859.
Nishimura K, Naito A, Tuzi S, Saito H, Hashimoto C, Aida M. J Phys Chem B 1998;
102:7476 7483.
Baldus M, Petkova AT, Herzfeld J, Griffin RG. Mol Phys 1998; 95:1197 1207.
Verel R, Beek JDv, Meier BH. J Magn Reson 1999; 140:300 303.
Verdegem PJE, Bovee-Geurts PHM, Grip WJD, Lugtenburg J, Groot HJMD.
Biochemistry 1999; 38:11316 11324.
Hong M. J Magn Reson 1999; 136:86 91.
546
Medek
277. Rossum B-Jv, Groot CPd, Ladizhansky V, Vega S, Groot HJMd. J Am Chem Soc
2000; 122:3465 3472.
278. Robyr P, Meier BH. Chem Phys Letts 2000; 327:319 324.
279. Mcdermott A, Polenova T, Bockmann A, Zilm KW, Paulsen EK, Martin RW, Montelione GT. J Biomol NMR 2000; 16:209 219.
280. Rienstra CM, Hohwy M, Hong M, Griffin RG. J Am Chem Soc 2000; 122:10979
10990.
281. Ishii Y, Tycko R. J Am Chem Soc 2000; 122:1443 1455.
282. Heindrichs ASD, Geen H, Titman JJ. J Magn Reson 2000; 147:26 35.
283. Huster D, Yamaguchi S, Hong M. J Am Chem Soc 2000; 122:11320 11327.
284. Yang J, Parkanzky PD, Khunte BA, Canlas CG, Yang R, Gabrys CM, Weliky DP. J
Mol Graphics 19.
285. Carravetta M, EdeC
n M, Johannessen OG, Luthman H, Verdegem PJE, Lugtenburg
J, Sebald A, Levitt MH. J Am Chem Soc 2001; 123:10628 10638.
286. Chan JCC, Brunklaus G. Chem Phys Lett 2001; 349:104 112.
287. Alia, Matysik J, Soede-huijbregts C, Baldus M, Raap J, Lugtenburg J, Gast P,
Gorkom HJV, Hoff AJ, Groot HJMD. J Am Chem Soc 2001; 123:4803 4809.
288. Heindrichs, Geen, Giordani, Titman Chem Phys Lett 2001; 335:89 96.
289. Lange A, Luca S, Baldus M. J Am Chem Soc 2002; 124:9704 9705.
290. Asakura T, Yao J, Yamane T, Umemura K, Ulrich AS. J Am Chem Soc 2002;
124:8794 8795.
291. Vosegaard T, Nielsen NC. J Biomol NMR 2002; 22:225 247.
292. Rienstra CM, Hohwy M, Mueller LJ, Jaroniec CP, Reif B, Griffin RG. J Am Chem
Soc 2002; 124:11908 11922.
293. Mueller LJ, Elliott DW, Kim K-C, Reed CA, Boyd PDW. J Am Chem Soc 2002;
124:9360 9361.
294. Wei Y, Lee D-K, McDermott AE, Ramamoorthy A. J Magn Reson 2002; 158:23
35.
295. Boer ID, Bosman L, Raap J, Oschkinat H, Groot HJMD. J Magn Reson 2002;
157:286 291.
296. Ladizhansky V, Griffin RG. J Am Chem Soc 2004; 126:948 958.
297. Katoh E, Takegoshi K, Terao T. J Am Chem Soc 2004; 126:3653 3657.
298. Carravetta M, Zhao X, Johannessen Ole G, Lai Wai C, Verhoeven Michiel A,
Bovee-Geurts Petra HM, Verdegem Peter JE, Kiihne S, Luthman H, de Groot
Huub JM, et al. J Am Chem Soc 2004; 126:3948 3953.
299. Padden BE. In: Graduate School. University of Minnesota, 2000:221.
300. Ando S, Hironaka T, Kurosu H, Ando I. Magn Reson Chem 2000; 38:241 250.
301. Maciejewska D, Herold F, Wolska I. J Mol Struct 2000; 553:73 77.
302. Buchanan GW, Astegar MF, G PAY. J Mol Struct 2001; 561:43 54.
303. Dega-Szafran Z, Gaszczyk I, Maciejewska D, Szafran M, Tykarska E, Wawer I. J
Mol Struct 2001; 560:261 273.
304. Harper JK, Mulgrew AE, Li JY, Barich DH, Strobel GA, Grant DM. J Am Chem Soc
2001; 123:9837 9842.
305. Cyranski MK, Wawer I, Zielinska A, Mrozek A, Koleva V, Lozanova C. J Phys Org
Chem 2001; 14:323 327.
306. Olejniczak S, Ganicz K, Tomczykowa M, Gudej J, Portzebowski MJ. J Chem Soc,
Perkin Trans 2 2002:1059 1065.
SSNMR Spectrometry
547
307. Rajeswaren M, Blanton TN, Zumbulyadis N, Giensen DJ, Conesa-Moratilla C,

Misture ST, Stephens PW, Huq A. J Am Chem Soc 2002; 124:14450 14459.
308. Aimi K, Fujiwara T, Ando S. J Mol Struct 2002; 602 603:405 416.
309. Aimi K, Fujiwara T, Ando S. J Mol Struct 2002; 602 603:417 428.
310. Claramunt RM, Cornago P, Sanz D, Foces-Foces C, Alkorta I, Elguerdo J. J Mol
Struct 2002; 605:199 212.
311. Marek R, Brus J, Tousek J, Kovacs L, Hockova K. Magn Reson Chem 2002;
40:353 360.
312. Goward GR, Sebastiani D, Schnell I, Spiess HW, Kim H-D, Ishida H. J Am Chem
Soc 2003; 125:5792 5800.
313. Helluy X, Sebald A. J Phys Chem B 2003; 107:3290 3296.
314. Zolek T, Paradowska K, Wawer I. Solid State Nucl Magn Reson 2003; 23:77 87.
315. Harper JK, Barich DH, Hu JZ, Strobel GA, Grant DM. J Org Chem 2003;
68:4609 4614.
316. Olsen RA, Struppe J, Elliott DW, Thomas RJ, Mueller LJ. J Am Chem Soc 2003;
125:11784 11785.
317. Mueller LJ, Elliott DW, Leskowitz GM, Struppe J, Olsen RA, Kim K-C, Reed CA.
J Magn Reson 2004; 168:327 335.
318. Sun H, Oldfield E. J Am Chem Soc 2004; 126:4726 4734.
319. Mehta MA, Fry EA, Eddy MT, Dedeo MT, Anagnost AE, Long JR. J Phys Chem B
2004; 108:2777 2780.
320. Giavani T, Bildsoe H, Skibsted J, Jakobsen HJ. J Magn Reson 2004; 166:262 272.
321. Mao JD, Hu WG, Schmidt-Rohr K, Davies G, Ghabbour EA, Xing B. Soil Sci Soc
Am J 2000; 64:873 884.
322. Hu W-G, Mao J, Schmidt-Rohr K, Xing B. Spec Publ-R Soc Chem 1999; 247:63
68.
323. Hazendonk P, K Harris, R, Galli G, Pizzanelli S. Physical Chemistry and Chemical
Physics 2002; 4:507 513.
324. Sullivan MJ, Maclel GE. Anal Chem 1982; 54:1615 1623.
325. Harris RK. Analyst 1985; 110:649 655.
326. Suryanarayanan R, Wiedmann TS. Pharm Res 1990; 7:184 187.
327. Likar MD, Taylor RJ, Fagerness PE, Hiyama Y, Robins RH. Pharm Res 1993;
10:75 79.
328. Gao P. Pharm Res 1996; 13:1095 1104.
329. Gao P. Pharm Res 1998; 15:1425 1433.
330. Gustafsson C, Lennholm H, Iversen T, Nystrom C. Intern J Pharm 1998;
174:243 252.
331. Lee GSH, Craig DC, Kannangara GSK, Dawson M, Conn C, Robertson J, Wilson
MA. J Forensic Sci 1999; 44:761 771.
332. Paris M, Bizot H, Emery J, Buzare JY, Buleon A. Int J Biol Macromol 2001;
29:127 136.
333. Paris M, Bizot H, Emery J, Buzare JY, Buleon A. Int J Biol Macromol 2001;
29:137 143.
334. Kono H, Yunoki S, Shikano T, Fujiwara M, Erata T, Takai M. J Am Chem Soc 2002;
124:7506 7511.
335. Kono H, Erata T, Takai M. J Am Chem Soc 2002; 124:7512 7518.
336. Kono H, Erata T, Takai M. Macromolecules 2003; 36:3589 3592.
548
Medek
337. Tozuka Y, Ito A, Seki H, Oguchi T, Yamamoto K. Chem Pharm Bull 2002;
50:1128 1130.
338. Vickery RD, Nemeth GA, Maurin MB. J Pharm Biomed Anal 2002; 30:125 129.
339. Apperley DC, Harris RK, Larsson T, Malmstrom T. J Pharm Sci 2003;
92:2487 2494.
340. Fu R, Hu J, Cross TA. J Magn Reson 2004; 168:8 17.
341. Byard SJ, Jackson SL, Smail A, Bauer M, Apperley DC. J Pharm Sci 2005;
94:1321 1335.
342. Smernik RJ, Oades JM. Geoderma 2000; 96:101 129.
343. Smernik RJ, Oades JM. Solid State Nucl Magn Reson 2001; 20:74 84.
344. Prauchner MJ, Pasa VMD, Menezes SMCd. J Wood Chem Technol 2001;
21:371 385.
345. Smernik RJ, Baldock JA, Oades JM, Whittaker AK. Solid State Nucl Magn Reson
2002; 22:50 70.
346. Smernik RJ, Baldock JA, Oades JM. Solid State Nucl Magn Reson 2002;
22:71 82.
347. Smernik RJ, Oades JM. Eur J Soil Sci 2003; 54:103 116.
348. Smernik RJ, Oliver IW, Merrington G. J Environ Qual 2003; 32:1523 1533.
349. Smernik RJ. Geoderma 2005; 125:249 271.
350. Bardet M, Foray MF, Tran Q-K. Anal Chem 2002; 74:4386 4390.
351. Mao J-D, Hu WG, Schmidt-Rohr K, Ding G, Davies G, Ghabbour EA, Xing B. Int J
Environ Anal Chem 2002; 82:183 196.
352. Keeler C, Maciel GE. Anal Chem 2003; 75:2421 2432.
353. Van Lagen B, de Jager PA. Fresenius Eniron Bull 2003; 12:1211 1217.
354. Peuravuori J, Ingman P, Pihlaja K. Talanta 2003; 59:177 189.
355. Conte P, Piccolo A, Lagen Bv, Buurman P, Hemminga MA. Solid State Nucl Magn
Reson 2002; 21:158 170.
356. Khalaf M, Kohl SD, Klumpp E, Rice JA, Tombacz E. Environ Sci Technol 2003;
37:2855 2860.
357. Mao JD, Schmidt-Rohr K. Environ Sci Technol 2004; 38:2680 2684.
358. Hexem JG, Frey MH, Opella SJ. J Am Chem Soc 1981; 103:224 226.
359. Hexem JG, Frey MH, Opella SJ. J Chem Phys 1982; 77:3847 3856.
360. Olivieri AC, Frydman L, Diaz LE. J Magn Reson (1969 1992) 1987; 75:50 62.
361. Olivieri AC, Fyrdman L, Grasselli M, Diaz LE. Magn Reson Chem 1988; 26:281
286.
362. Olivieri AC. J Magn Reson (1969 1992) 1989; 82:342 346.
363. Olivieri AC. J Magn Reson (1969 1992) 1989; 81:201 205.
364. Olivieri AC. J Chem Soc, Perkin Trans 2: Phys Org Chem (1972 1999) 1990;
85 89.
365. Frydman L, Olivieri AC, Diaz LE, Frydman B, Schmidt A, Vega S. Mol Phys 1990;
70:563 579.
366. Grasselli M, Diaz LE, Olivieri AC. Spectrosc Lett 1991; 24:895 907.
367. Harris RK, Olivieri AC. Prog Nucl Magn Reson Spectrosc 1992; 24:435 456.
368. Olivieri AC. J Magn Reson, Ser A 1993; 101:313 316.
369. Alarcon SH, Olivieri AC, Jonsen P. J Chem Soc, Perkin Trans 2: Phys Org Chem
(1972 1999) 1993; 1783 1786.
370. Olivieri AC, Elguero J, Sobrados I, Cabildo P, Claramunt RM. J Phys Chem 1994;
98:5207 5211.
SSNMR Spectrometry
549
371. Odgaard L, Bak M, Jakobsen HJ, Nielsen NC. J Magn Reson 2001; 148:
298 308.
372. Hughes CE, Pratima R, Karlsson T, Levitt MH. J Magn Reson 2002; 159:25 35.
373. Strohmeier M, Grant DM. J Am Chem Soc 2004; 126:966 977.
374. Takegoshi K, Yano T, Takeda K, Terao T. J Am Chem Soc 2001; 123:10786 10787.
375. Stephenson GA, Stowell JG, Toma PH, Dorman DE,Greene JR, Byrn SR. J Am
Chem Soc 1993; 116:5766 5773.
376. McGeorge G, Harris RK, Chippendale AM, Bullock JF. J Chem Soc, Perkin Trans 2
1996; 2:1733 1738.
377. Schmidt A, Kababya S, Appel M, Khatib S, Botoshansky M, Eichen Y. J Am Chem
Soc 1999; 121:11291 11299.
378. Glaser R, Shiftan D, Drouin M. J Org Chem 1999; 64:9217 9224.
379. Apperley DC, Fletton RA, Harris RK, Lancaster RW, Tavener S, Threlfall TL.
J Pharm Sci 1999; 88:1275 1280.
380. Dong Z, VG Young, J, Padden BE, Schroeder SA, Prakash I, Munson EJ, Grant
DJW. J Chem Crystallography 1999; 29:967 975.
381. Padden BE, Zell MT, Dong Z, Schroeder SA, Grant DJW, Munson E. Anal Chem
1999; 71:3325 3331.
382. Zell MT, Padden BE, Grant DJW, Chapeau MC, Prakash I, Munson E. J Am Chem
Soc 1999; 121:1372 1378.
383. Zell MT, Padden BE, Grant DJW, Schroeder SA, Wachholder KL, Prakash I,
Munson EJ. Tetrahedron 2000; 56:6603 6616.
384. Dong Z, Padden BE, Salsbury JS, Munson EJ, Shroeder SA, Prakash I, Grant DJW.
Pharm Res 2002; 19:330 336.
385. Dong Z, Salsbury JS, Munson EJ, Shroeder SA, Zhou D, Prakash I, Vyazovkin S,
Wight CA, Grant DJW. J Pharm Sci 2002; 91:1423 1431.
386. Sack I, Macholl S, Wehrmann F, Albrecht J, Limbach HH, Fillaux F, Baron MH,
Bunkowsky G. Appl Magn Reson 1999; 17:413 431.
387. Lee H-H, Kimura K, Takai T, Senda H, Kuwae A, Hanai K. Spectrochim Acta, Part
A 1999; 55:2877 2882.
388. Chen LR, Padden BE, Vippagunta SR, Munson E, Grant DJW. Pharm Res 2000;
17:619 624.
389. Henck JO, Finner E, Burger A. J Pharm Sci 2000; 89:1151 1159.
390. Stephenson GA. J Pharm Sci 2000; 89:958 966.
391. Medek A, Frydman L. J Am Chem Soc 2000; 122:684 691.
392. Variankaval NE, Jacob KI, Dinh SM. J Cryst Growth 2000; 217:320 331.
393. Wenslow RM, Baum MW, Ball RG, Mccauley JA. J Pharm Sci 2000;
89:1271 1285.
394. Lee GSH, Taylor RC, Dawson M, Kannangara GSK, Wilson MA. Solid State Nucl
Magn Reson 2000; 16:225 237.
395. Crowley KJ, Forbes RT, York P, Apperley DC, Nyqvist H, Camber O. J Pharm Sci
2000; 89:1286 1295.
396. Harper JK, Grant DM. J Am Chem Soc 2000; 122:3708 3714.
397. Strohmeier M, Orendt AM, Alderman DW, Grant DM. J Am Chem Soc 2001;
123:1713 1722.
398. Hanai K, Kuwae A, Takai T, Senda H, Kunimoto K-K. Spectrochimica Acta Part A
2001; 57:513 519.
399. Fojud Z, Jurga S. Mol Phys Rep 2001; 33:172 174.
550
Medek
400. Yoshinari T, Forbes RT, York P, Kawashima Y. Int J Pharm 2002; 247:69 77.
401. Bauer J, Spanton S, Henry R, Quick J, Dziki W, Porter W, Morris J. Pharm Res 2001;
18:859 866.
402. Liang J, Ma Y, Chen B, Munson EJ, Davis HT, Binder D, Chang H-T, Abbas S,
Hsu F-L. J Phys Chem B 2001; 105:9653 9662.
403. Pursch M, Sander LC, Albert K. Anal Chem 1996; 68:4107 4113.
404. Giordano F, Rossi A, Moyano JR, Gazzaniga A, Massarotti V, Bini M, Capsoni D,
Peveri T, Redenti E, Carima L, et al. J Pharm Sci 2001; 90:1154 1163.
405. Kimura N, Fukui H, Takagaki H, Yonemochi E, Terada K. Chem Pharm Bull 2001;
49:1321 1325.
406. Wiegerinck P, Booy K-J, Kellenbach E, Lambregts D, Vromans H, Vadar J,
Kaspersen F. Synthesis and Applications of Isotopically Labelled Compounds.
Proceedings of the International Symposium, 7th, Dresden, Germany, June 18 22,
2000, 2001, Vol. 7, 185188.
407. Lee JH, Gi U-S, Kim J-H, Kim Y, Kim S-H, Oh H, Min B. Bull Korean Chem Soc
2001; 22:925 928.
408. Smith EDL, Hammond RB, Jones MJ, Roberts KJ, John JBO, Price SL, Harris RK,
Apperley DC, Cherryman JC, Docherty R. J Phys Chem B 2001; 105:5818 5826.
409. Wang X, Wang XJ, Ching CB. Chirality 2002; 14:318 324.
410. Zhang GGZ, Gu C, Zell MT, Burkhardt RT, Munson EJ, Grant DJW. J Pharm Sci
2002; 91:1089 1100.
411. Moynihan HA, OHare IP. Int J Pharm 2002; 247:179 185.
412. Garcia MA, Lopez C, Claramunt RM, Kenz A, Pierrot M, Elguero J. Helvetica
Chimica Acta 2002; 85:2763 2776.
413. Novoselsky A, Glaser R. Magn Reson Chem 2002; 40:723 728.
414. Zimmermann H, Bader V, Poupko R, Wachtel EJ, Luz Z. J Am Chem Soc 2002;
124:15286 15300.
415. De Jong AF, Kentgens APM, Veeman WS. Chem Phys Lett 1984; 109:337 342.
416. Potrzebowski MJ, Tadeusiak E, Misiura K, Ciesielski W, Bujacz G, Tekely P.
ChemEur J 2002; 8:5007 5011.
417. Maurin MB, Vickery RD, Rabel SR, Rowe SM, Everlof JG, nemeth GA, Campbell
GC, Foris CM. J Pharm Sci 2002; 91:2599 2604.
418. Wang Y, Wenslow RM, McCauley JA, Crocker LS. Int J Pharm 2002; 243:147 159.
419. Pisklak M, Perlinski M, Kossakowski J, Wawer I. Acta Poloniae Pharmaceutica
2002; 59:461 465.
420. Pisklak M, Maciejewska D, Herold F, Wawer I. J Mol Struct 2003; 649:169 176.
421. Komber H, Limbach H-H, Bohme F, Kunert C. J Am Chem Soc 2002;
124:11955 11963.
422. Harper JK, Facelli JC, Barich DH, McGeorge G, Mulgrew AE, Grant DM. J Am
Chem Soc 2002; 124:10589 10595.
423. Meejoo S, Kariuki BM, Kitchin SJ, Cheung EY, Albesa-Jove D, Harris KDM.
Helvetica Chimica Acta 2003; 86:1467 1477.
424. Iuliucci RJ, Clawson J, Hu JZ, Solum MS, Barich D, Grant DM, Taylor CMV. Solid
State Nucl Magn Reson 2003; 24:23 38.
425. Alderman DW, McGeorge G, Hu JZ, Pugmire RJ, M.Grant D. Mol Phys 1998;
95:1113 1126.
426. Rienstra CM, Hatcher ME, Mueller LJ, Sun B, Fesik SW, Griffin RG. J Am Chem
Soc 1998; 120:10602 10612.
SSNMR Spectrometry
551
427. Reutzel-Edens SM, Kleemann RL, Lewellen PL, Borghese AL, Antoine LJ. J Pharm
Sci 2003; 92:1196 1205.
428. Zolek T, Paradowska K, Krajewska D, Rozanski A, Wawer I. J Mol Struct 2003;
646:141 149.
429. Vanhaecht B, Devroede J, Willem R, Biesemans M, Goonewardena W, Rastogi S,
Hoffmann S, Klein PG, Koning CE. J Polym Sci, Part A: Polym Chem 2003;
41:2082 2094.
430. Portieri A, Harris RK, Fletton RA, Lancaster RW, Threlfall TL. Magn Reson Chem
2004; 42:313 320.
431. Kameda T, McGeorge G, Orendt AM, Grant DM. J Biomol NMR 2004;
29:281 288.
432. Remenar JF, Wenslow R, Ostovic D, Peresypkin A. Pharm Res 2004; 21:
185 188.
433. Sheth AR, Lubach JW, Munson EJ, Muller FX, Grant DJW. J Am Chem Soc 2005;
6641 6651.
434. Booy K-J, Wiegerinck P, Vader J, Kaspersen F, Lambregts D, Vromans H,
Kellenbach E. J Pharm Sci 2005; 94:458 463.
435. Schmidt AC. Eur J Pharm Sci 2005; 25:407 416.
436. Suzuki N, Kawasaki T. J Pharm Biomed Anal 2005; 37:177 181.
437. Barich DH, Davis JM, Schrieber LJ, Zell MT, Munson EJ. In press.
438. Gilchrist LM, Monde K, Tomita Y, Iwashita T, Nakanishi K, McDermott AE. J
Magn Reson 2001; 152:1 6.
439. Monde K, Tomita Y, ML Gilchrist, J, Mcdermott AE, Nakanishi K. Isr J Chem 2000;
40:301 306.
440. Antoniloli G, Hodgkinson P. J Magn Reson 2004; 168:124 131.
441. Campbell SC, Harris RK, Hardy MJ, Lee DC, Busby DJ. J Chem Soc, Perkin Trans 2
1997; 2:1913 1918.
442. Harris RK, Crowe LA. J Chem Soc, Dalton Trans 1999; 24:4315 4323.
443. Brouwer EB, Challoner R, Harris RK. Solid State Nucl Magn Reson 2000;
18:37 52.
444. Fuchs B, Scheler U. Macromolecules 2000; 33:120 124.
445. Lappan U, Fuchs B, Geibler U, Scheler U, Lunkwitz K. Polymer 2002;
43:4325 4330.
446. Dargaville TR, George GA, Hill DJT, Scheler U, Whittaker AK. Macromolecules
2003; 36:7138 7142.
447. Gleason KK, Hill DJT, Lau KKS, Mohajerani S, Whittaker AK. Nucl Instrum
Methods Phys Res, Sect B: Beam Interactions With Materials and Atoms 2001;
185:83 87.
448. Holstein P, Scheler U, Harris RK. Polymer 1998; 39:4937 4941.
449. Ando S, Harrison RK, Reinsberg SA. Magn Reson Chem 2002; 40:97 106.
450. Ando S, Harris RK, Holstein P, Reinsberg SA, Yamauchi K. Polymer 2001;
42:8137 8151.
451. Su T-W, Tzou D-LM. Polymer 2000; 41:7289 7293.
452. Wormald P, Apperley DC, Beaume F, Harris RK. Polymer 2003; 44:643 651.
453. Ando S, Harris RK, Hirschinger J, Reinsberg SA, Scheler U. Macromolecules 2001;
34:66 75.
454. Wenslow RM. Drug Dev Ind Pharm 2002; 28:555 561.
455. Solis D, Ferraro MB, Facelli JC. J Mol Struct 2002; 602 603:159 164.
552
Medek
456. Glaser RW, Sachse C, Durr UHN, Wadhwani P, Ulrich AS. J Magn Reson 2004;
168:153 163.
457. Park J-W, Seo Y-A, Kim I, Ha C-S, Aimi K, Ando S. Macromolecules 2004;
37:429 436.
458. Budarin VL, Clark JH, Tavener SJ. Chem Commun (Cambridge, United Kingdom)
2004; 5:524 525.
459. Utz M. Annu Techn Conf-Soc Plast Eng 2002; 2:2015 2019.
460. Utz M, Atallah AS, Robyr P, Widmann AH, Ernst RR, Suter UW. Macromolecules
1999; 32:6191 6205.
461. Utz M, Robyr P, Suter UW. Macromolecules 2000; 33:6808 6814.
462. Utz M, Eisenegger J, Suter UW, Ernst RR. J Magn Reson 1997; 128:217 227.
463. Torchia DA, Szabo A. J Magn Reson (1969 1992) 1982; 49:107 121.
464. Torchia DA. J Magn Reson 1978; 30:613 616.
465. Schaefer J, Stejskal EO, Buchdahl R. Macromolecules 1977; 10:384 405.
466. Kitamaru R, Horii F, Murayama K. Macromolecules 1986; 19:636 643.
467. Oksanen CA, Zografi G. Pharm Res 1993; 10:791 799.
468. Dries IJvd, Dusschoten Dv, Hemminga MA. J Phys Chem B 1998;
102:10483 10489.
469. Yoshioka S, Aso Y, Kojima S. Pharm Res 1999; 16:135 140.
470. Yoshioka S, Aso Y, Kojima S, Sakurai S, Fujiwara T, Akutsu H. Pharm Res 1999;
16:1621 1625.
471. Zumbulyadis N, Antalek B, Windig W, Scaringe RP, Lanzafame AM, Blanton T,
Helber M. J Am Chem Soc 1999; 121:11554 11557.
472. Silva NMd, Tavares MIB, Stejskal EO. Macromolecules 2000; 33:115 119.
473. Aso Y, Yoshioka S, Kojima S. J Pharm Sci 2000; 90:798 806.
474. Glaser R, Novoselsky A, Shiftan D. J Org Chem 2000; 65:6345 6353.
475. Kuwabara K, Horii F, Ogawa Y. J Mol Struct 2000; 525:163 171.
476. Varner SJ, Vold RL, Hoatson GL. J Magn Reson 2000; 142:229 240.
477. Kakou-Yao R, Pizzala H, Pietri N, Aycard J-P. J Chem Crystallography 2000;
30:593 598.
478. Rheingold AL, Figueroa JS, Dybowski C, Beckmann PA. Chem Commun 2000;
651 652.
479. Beckmann PA, Burbank KS, Clemo KM, Slonaker EN, Averill K, Dybowski C,
Figueroa JS, Glatfelter A, Koch S, Liable-Sands LM, Rheingold AL. J Chem Phys
2000; 113:1958 1965.
480. Vittadini E, Dickinson LC, Chinachoti P. Carbohydr Polym 2001; 46:49 57.
481. Perera MCS, Ishiaku US, Ishak ZAM. Eur Polym J 2001; 37:167 178.
482. Tang HR, Belton PS, Davies SC, Hughes DL. Carbohydr Res 2001; 330:
391 399.
483. Higgins JS, Hodgson AW, Law RV. Polym Mater Sci Eng 2001; 84:990 991.
484. Lupulescu A, Kotecha M, Frydman L. J Am Chem Soc 2003; 125:3376 3383.
485. Lopes MH, Sarychev A, Neto CP, Gil AM. Solid State Nucl Magn Reson 2000;
16:109 121.
486. Tang H-R, Wang Y-L, Belton PS. Solid State Nucl Magn Reson 2000;
15:239 248.
487. Nogueira RF, Tavares MIB. J Appl Polym Sci 2002; 138 143.
488. Lai S, Casu M, Saba G, Lai A, Husu I, Masci G, Crescenzi V. Solid State Nucl Magn
Reson 2002; 21:187 196.
SSNMR Spectrometry
553
489. Adriaensens P, Carleer R, Storme L, Vanderzande D, Gelan J. Polymer 2002;

43:7003 7006.
490. Brachais L, Laupretre F, Caille J-R, Teyssie D, Boileau S. Polymer 2002;
43:1829 1833.
491. Alamo RG, Blanco JA, Carrileo I, Fu R. Polymer 2002; 43:1857 1865.
492. Aso Y, Yoshioka S, Zhang J, Zografi G. Chem Pharm Bull 2002; 50:822 826.
493. Yoshioka S, Aso Y, Kojima S. J Pharm Res 2002; 91:2203 2210.
494. Sidhu PS, Enright GD, Ripmeester JA. J Phys Chem B 2002; 106:8569 8581.
495. Prout K. Croatica Chemica Acta 2002; 75:817 833.
496. Kuzmicz R, Dobrzycki L, Wozniak K, Benevelli F, Klinowski J, Kolodziejski W.
Physical Chemistry and Chemical Physics 2002; 4:2387 2391.
497. Harris DJ, Alam MK. Polymer 2002; 43:5147 5155.
498. Mao J-D, Hundal LS, Thompson ML, Schmidt-Rohr K. Environ Sci Technol 2002;
36:929 936.
499. Kitchin SJ, Ahn S, Harris KDM. J Phys Chem A 2002; 106:7228 7234.
500. Leisen J, Beckham HW, Benham M. Solid State Nucl Magn Reson 2002;
22:409 422.
501. Witkowski S, Wawer I. J Chem Soc, Perkin Trans 2 2002; 433 436.
502. Harris DJ, Azevedo ERd, Bonagambab TJ. J Magn Reson 2003; 162:67 73.
503. Lim AR, Kwark Y-J, Kim J-S. J Appl Phys 2003; 94:7351 7355.
504. Krushelnitsky A, Faizullin D, Reichert D. Biopolymers 2004; 73:1 15.
505. Beckmann PA, Paty C, Allocco E, Herd M, Kuranz C, Rheingold AL. J Chem Phys
2004; 120:5309 5314.
506. Masuda K, Tabata S, Sakata Y, Hayase T, Yonemochi E, Terada K. Pharm Res 2005;
22:797 805.
507. Lubach JW, Xu D, Segmuller B, Munson EJ. In Press 2005.
508. Markovich RJ, Evans CA, Coscolluela CB, Zibas SA, Rosen J. J Pharm Biomed Anal
1997; 16:661 673.
509. Watanabe T, Wakiyama N, Usui F, Ikeda M, Isobe T, Senna M. Int J Pharm 2001;
226:81 91.
510. Forster A, Apperley D, Hempenstall J, Lancaster R, Rades T. Pharmazie 2003;
58:761 762.
511. Kalachandra S, Lin DM, Stejskal EO, Prakki A, Offenbacher S. J Mater Sci: Mater
Med 2005; 16:597 605.
512. Robyr P. Polym Mater Sci Eng 2000; 82:135.
513. Hu WG, Schmidt-Rohr K. Polymer 2000; 41:2979 2987.
514. Colhoun FL, Armstrong RC, Rutledge GC. Polym Mater Sci Eng 2000; 82:
136 137.
515. Asano A, Takegoshi K. J Chem Phys 2001; 115:8665 8669.
516. Valtier M, Drujon X, Wilhelm M, Spiess HW. Macromolecules 2001; 202:
1262 1272.
517. Palmas P, Campion LL, Bourgeoisat C, Martel L. Polymer 2001; 42:7675 7683.
518. Lin R-H, Woo EM, Chiang JC. Polymer 2001; 42:4289 4297.
519. Huster D, Yao X, Hong M. J Am Chem Soc 2002; 124:874 883.
520. Brus J, Dybal J, Sysel P, Hobzova R. Macromolecules 2002; 35:1253 1261.
521. Kretschmer A, Drake R, Neidhoefer M, Wilhelm M. Solid State Nucl Magn Reson
2002; 22:204 217.
522. Nozirov F. Solid State Nucl Magn Reson 2002; 21:197 203.
554
Medek
523.
524.
525.
526.
527.
Murata K, Kuroki S, Kimura H, Ando I. Biopolymers 2002; 64:26 33.

Murata K, Kuroki S, Ando I. Polymer 2002; 43:6871 6878.
Jia X, Wolak J, Wang X, White JL. Macromolecules 2003; 36:712 718.
Qiu X, Ediger MD. J Phys Chem B 2003; 107:459 464.
Munson EJ, Carlson LK, Jorving JE, Zelt MT, Abbott J, Hillmyer MA. Polym Prepr
2003; 44:325 326.
Werkhoven TM, Mulder FM, Zune C, Jerome R, M.deGroot HJ. Macromol Chem
Phys 2003; 204:46 51.
Akita C, Kawaguchi T, Kaneko F, Yamamoto H, Suzuki M. J Phys Chem B 2004;
108:4862 4868.
Rohrs BR, Thamann TJ, Gao P, Stelzer DJ, Berger MS, Chao RS. Pharm Res 1999;
16:1850 1856.
Puttipipatkhachorn S, Nunthanid J, Yamamoto K, Peck GE. J Controlled Release
75:143 153.
Watanabe T, Hasegawa S, Wakiyama N, Kusai A, Senna M. Int J Pharm 2002;
248:123 129.
Babonneau F, Camus L, Steunou N, Ramila A, Vallet-Regi M. Mater Res Soc Symp
Proc 2003; 775:77 82.
Boland MP, Middleton DA. Magn Reson Chem 2004; 42:204 211.
Kawada J, Marchessault RH. Starch/Staerke 2004; 56:13 19.
Smits ALM, Kruiskamp PH, Soest JJGv, Vliegenthart JFG. Carbohydr Polym 2003;
53:409 416.
Kozerski L, Kamienski B, Kawecki R, Urbanczyk-Lipkowska Z, Bocian W,
Bednarek E, Sitkowski J, Zakrzewska K, Nielsen KT, Hansen PE. Org Biomol
Chem 2003; 1:3578 3585.
Zech SG, Olejniczak E, Hajduk P, Mack J, McDermott AE. J Am Chem Soc 2004;
126:13948 13953.
Takegoshi K, Nakamura S, Terao T. Chem Phys Lett 344:631 637.
Lubach JW, Padden BE, Winslow SL, Salsbury JS, Masters DB, Topp EM, Munson
EJ. Anal Bioanal Chem 2004; 378:1504 1510.
Schachter DM, Xiong J, Tirol GC. Int J Pharm 2004; 281:89 101.
Gjerde A, Holmsen H, Nerdal W. Biochimica et biophysica acta 2004; 1682:
28 37.
Apperley DC, Basford PA, Dallman CI, Harris RK, Kinns M, Marshall PV, Swanson
AG. J Pharm Sci 2005; 94(3):516 523.
Bogdanova S, Pajeva I, Nikolova P, Tsakovska I, Mueller B. Pharm Res 2005;
22:806 815.
Schneider H-J, Hacket F, Rudiger V. Chem Rev 1998; 98:1755 1785.
Kitchin SJ, Halstead TK. Appl Magn Reson 1999; 17:283 300.
Crini, Bourdonneau, Martel, Piotto, Morcellet, Richert, Vebrel, Torri, Morrin. J Appl
Polym Sci 2000; 75:1288 1295.
Garbow JR, Likos JJ, Schroeder SAJ. Agric Food Chem 2001; 49:2053 2060.
Saalwachter K. Macromol Rapid Commun 2002; 23:286 291.
Wulff M, Aden M, Egenfeldt J. Bioconjugate Chem 2002; 13:240 248.
Lu J, Mirau PA, Shin ID, Nojima S, Tonelli AE. Macromol Chem Phys 2002;
203:71 79.
Yannakopoulou K, Ripmeester JA, Mavridis IM. J Chem Soc, Perkin Trans 2 2002;
1639 1644.
528.
529.
530.
531.
532.
533.
534.
535.
536.
537.
538.
539.
540.
541.
542.
543.
544.
545.
546.
547.
548.
549.
550.
551.
552.
SSNMR Spectrometry
555
553. Cunha-Silva L, Teixeira-Dias JJC. J Phys Chem B 2002; 106:3323 3328.

554. Braga SS, Aree T, Imura K, Vertut P, Boal-Palheiros I, Saenger W, Teixeira-Dias
JJC. J Inclusion Phenomena and Macrocyclic Chemistry 2002; 43:115 125.
555. Braga SS, Goncalves IS, Herdtweck E, Teixeira-Dias JJC. New J Chem 2003;
27:597 601.
556. Braga SS, Ribeiro-Claro P, Pillinger M, Goncalves IS, Pereira F, Fenandes AC,
Romao CC, Correia PB, Teixeira-Dias JJC. Org Biomol Chem 2003; 1:873 878.
557. Cunha-Silva L, Teixeira-Dias JJC. New J Chem 2004; 28:200 206.
558. Lai S, Locci E, Piras A, Porcedda S, Lai A, Marongiu B. Carbohydr Res 2003;
338:2227 2232.
559. Anulewicz R, Wawer I, Piekarska-Bartoszewicz B, Temeriusz A. Carbohydr Res
1996; 281:1 10.
560. Temeriusz A, Piekarska-Bartoszewicz B, Wawer I. Carbohydr Res 1997;
304:335 340.
561. Tanner SF, Chanzy H, Vincendon M, Roux JC, Gaill F. Macromolecules 1990;
23:3576 3583.
562. Kunze J, Ebert A, Fink HP. Cellul Chem Technol 2000; 34:21 34.
563. Lefort R, De Gusseme A, Willart JF, Danede F, Descamps M. Int J Pharm 2004;
280:209 219.
564. Elena B, Emsley L. J Am Chem Soc 2005; 127:9140 9146.
565. Tong G, Schaefer J. Macromolecules 1997; 30:7522 7528.
566. Fyfe CA, Lewis AR, Chezeau J-M. Canadian J Chem 1999; 77:1984 1993.
567. Fyfe CA, Skibsted J, Grodney H, Altenschildesche HMz. Chem Phys Lett 1997;
281:44 48.
568. Bertani P, Raya J, Hirschinger J. Solid State Nucl Magn Reson 2002; 22:
188 203.
569. Middleton DA, Duff CSL, Peng X, Reid DG, Saunders D. J Am Chem Soc 2000;
122:1161 1170.
570. Salgado J, Grage SL, Kondejewski LH, Hodges RS, Mcelhaneyh RN, Ulrich AS.
J Biomol NMR 2001; 21:191 208.
571. Middleton DA, Peng X, Saunders D, Shankland K, David WIF, Markvardsen A.
J Chem Commun (Cambridge, United Kingdom) 2002; 17:1976 1977.
572. Mehta AK, Cegelski L, OConnor RD, Schaefer J. J Magn Reson 2003;
163:182 187.
573. Grage SL, Watts JA, Watts A. J Magn Reson 2004; 166:1 10.
574. Schulz-Dobrick M, Metzroth T, Spiess HW, Gauss J, Schnell I. Chem Phys Chem
2005; 6:315 327.
575. Limbach H-H. Intermolecular Forces. Berlin Heidelberg: Springer-Verlag,
1991:280 295.
576. Hickman BS, Mascal M, Titman JJ, Wood IG. J Am Chem Soc 1999;
121:11486 11490.
577. Wu G, Kazuhiko Y, Dong S, Grondey H. J Am Chem Soc 2000; 122:
4215 4216.
578. Song X-J, McDermott AE. Magn Reson Chem 2001; 39:S37 S43.
579. Zhao X, Sudmeier JL, Bachovchin WW, Levitt MH. J Am Chem Soc 2001;
123:11097 11098.
580. Kiss JT, Felfodi K, Hannus I, Palinko I. J Mol Struct 2001; 565 566:463 468.
581. Huelsekopf M, Ludwig R. Magn Reson Chem 2001; 39:S127 S134.
556
Medek
582. Henry B, Tekely P, Delpuech J-J. J Am Chem Soc 2002; 124:2025 2034.
583. Foces-Foces C, Echevarria A, Jagerovic N, Alkorta I, Elguero J, Langer U, Klein O,
Minguet-Bonvehi M, Limbach H-H. J Am Chem Soc 2001; 123:7989 7906.
584. Totz J, Michel D, Ivanov YN, Sukhovsky AA, Aleksandrova IP, Petersson J. Magn
Reson Chem 2001; 39:S50 S58.
585. Michel D, Totz J, Ivanov YN, Sukhovsky AA, Aleksandrova IP, Petersson J.
Ferroelectrics 2002; 267:303 310.
586. Takeda S, Tsuzumitani A. Magn Reson Chem 2001; 39:S44 S49.
587. Lorente P, Shenderovich IG, Golubev NS, Denisov GS, Buntkowsky G, Limbach
H-H. Magn Reson Chem 2001; 39:S18 S29.
588. Goward GR, Schnell I, Brown SP, Spiess HW, Kim H-D, Ishida H. Magn Reson
Chem 2001; 39:S1 S17.
589. Ono S, Taguma T, Kuroki S, Ando I, Kimura H, Yamauchi K. J Mol Struct 2002;
602 603:49 58.
590. Colsenet R, Gardiennet C, Henry B, Tekely P. Angew Chem Int Ed 2002;
41:4743 4745.
591. Brown SP, Perez-Torralba M, Sanz D, Claramunt RM, Emsley L. Chem Commun
(Cambridge, United Kingdom) 2002; 1852 1853.
592. Lesage A, Emsley L, Chabanas M, Coperet C, Basset J-M. Angew Chem Int Ed
2002; 41:4535 4538.
593. Ilczyszyn MM, Ilczyszyn M. J Raman Spectrosc 2003; 34:693 704.
594. Godzisz D, Ilczyszyn MM, Ilczyszyn M. J Mol Struct 2002; 606:123 137.
595. Pawsey S, Mccormick M, Paul SD, Graf R, Lee YS, Reven L, Spiess HW. J Am
Chem Soc 2003; 125:4174 4184.
596. Emmler T, Gieschler S, Limbach HH, Buntkowsky G. J Mol Struct 2004;
700:29 38.
597. Leppert J, Urbinati CR, Haefner S, Ohlenschlaeger O, Swanson MS, Goerlach M,
Ramachandran R. Nucleic Acids Res 2004; 32:1177 1183.
598. Schilf W, Kamienski B, Szady-Chelmieniecka A, Grech E. J Mol Struct 2004;
700:105 108.
599. Schilf W, Kamienski B, Kolodziej B, Grech E. J Mol Struct 2004; 708:33 38.
600. Rozwadowski Z, Schilf W, Kamienski B. Magn Reson Chem 2005; 43:573 577.
601. Schilf W, Kamienski B, Szady-Chelmieniecka A, Grech E. J Mol Struct 2005;
743:237 241.
602. Szady-Chelmieniecka A, Grech E, Rozwadowski Z, Dziembowska T, Schilf W,
Kamienski B. J Mol Struct 2001; 565 566:125 128.
603. Schilf W, Kamienski B, Dziembowska T, Rozwadowski Z, Szady-Chelmieniecka A.
J Mol Struct 2000; 552:33 37.
604. Schilf W, Kamienski B, Szady-Chelmieniecka A, Grech E. Solid State Nucl Magn
Reson 2000; 18:97 105.
605. Li J, Chatterjee K, Medek A, Shalaev E, Zografi G. J Pharm Sci 2004; 93:697 712.
606. Chekmenev EY, Zhang Q, Waddell KW, Mashuta MS, Wittebort RJ. J Am Chem
Soc 2004; 126:379 384.
607. Chekmenev EY, Xu RZ, Mashatu MS, Wittenbort RJ. J Am Chem Soc 2002;
124:11894 11899.
608. Waddell KW, Chekmenev EY, Wittebort RJ. J Am Chem Soc 2005; 127:
9030 9035.
SSNMR Spectrometry
557
609. Diez-Pena E, Quijada-Garrido I, Barrales-Rienda JM, Schnell I, Spiess HW.

Macromol Chem Phys 2004; 205:430 437.
610. Kawashima H. Energy & Fuels 2005; 19:538 543.
611. Benhabbour SR, Chapman RP, Scharfenberger G, Meyer WH, Goward GR. Chem
Mater 2005; 17:1605 1612.
612. Gobetto R, Nervi C, Valfre E, Chierotti MR, Braga D, Maini L, Grepioni F, Harris
RK, Ghi PY. Chem Mater 2005; 17:1457 1466.
613. Li ZJ, Abramov Y, Bordner J, Leonard J, Medek A, Trask AV. In press 2005.
614. Wawer I, Zielinska A. Magn Reson Chem 2001; 39:374 380.
615. Hughes E, Frossard P, Sagalowicz L, Nouzille A, Raemy A, Watzke H. Spec Publ-R
Soc Chem 2002; 286:144 150.
616. Plass J, Emeis D. Magnetic Resonance in Colloid and Interface Science 2002;
375 382.
617. Plass J, Emeis D, Blumich B. Journal of Surfactants and Detergents 2001;
4:379 384.
618. Calucci L, Forte C, Galleschi L, Geppi M, Ghiringhelli S. Int J Biol Macromol 2003;
32:179 189.

Spectroscopy of Pharmaceutical Solids PDF

Uploaded by

Copyright:

Available Formats

Spectroscopy of Pharmaceutical Solids PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Spectroscopy of Pharmaceutical Solids PDF

Uploaded by

Copyright:

Available Formats

DK5783_half-series-title.

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

DRUGS AND THE PHARMACEUTICAL SCIENCES

Center for Pharmaceutical Physics

University of Southern California

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

University of Kansas School of

1. Pharmacokinetics, Milo Gibaldi and Donald Perrier

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

26. Drug Dynamics for Analytical, Clinical, and Biological Chemists,

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

74. Oral Mucosal Drug Delivery, edited by Michael J. Rathbone

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

96. Freeze-Drying/Lyophilization of Pharmaceutical and Biological Products,

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

119. Dermatological and Transdermal Formulations, edited by

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

140. Microbial Contamination Control in Parenteral Manufacturing, edited by

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

New York London

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

DK5783_Discl.fm Page 1 Monday, November 14, 2005 11:32 AM

Library of Congress Cataloging-in-Publication Data

Visit the Taylor & Francis Web site at

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

Spectroscopic techniques have been widely used in the pharmaceutical sciences

2006 by Taylor & Francis Group, LLC

2006 by Taylor & Francis Group, LLC

Electromagnetic Radiation and Spectroscopy . . . . . . . . . . . . . . . . 1

Core Electron Spectroscopy

Core Electron States and X-Ray Absorption Spectroscopy . . . . . 33

X-Ray Photoelectron and X-Ray Fluorescence Spectroscopy

2006 by Taylor & Francis Group, LLC

X-Ray Photoelectron Spectroscopy . . . . 69

Valence Electron Spectroscopy

Molecular Orbital Theory and the Electronic Structure

UV/VIS Reflectance Spectroscopy . . . . . . . . . . . . . . . . . . . . . . 121

Luminescence Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

2006 by Taylor & Francis Group, LLC

Vibrational Level Spectroscopy

Molecular Motion and Vibrational Spectroscopy . . . . . . . . . . . 205

Infrared Absorption Spectroscopy . . . . . . . . . . . . . . . . . . . . . . 235

Raman Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271

Near-Infrared Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 313

2006 by Taylor & Francis Group, LLC

Nuclear Spin Level Spectroscopy

Solid-State Nuclear Magnetic Resonance Spectrometry . . . . . . 413