The Mie Theory - Basics and Applications
The Mie Theory - Basics and Applications
The Mie Theory - Basics and Applications
Volume 169
Founded by
H. K. V. Lotsch
Editor-in-Chief:
W. T. Rhodes
Editorial Board:
Ali Adibi, Atlanta
Toshimitsu Asakura, Sapporo
Theodor W. Hänsch, Garching
Takeshi Kamiya, Tokyo
Ferenc Krausz, Garching
Bo A. J. Monemar, Linköping
Herbert Venghaus, Berlin
Horst Weber, Berlin
Harald Weinfurter, München
The Springer Series in Optical Sciences, under the leadership of Editor-in-Chief William T. Rhodes,
Georgia Institute of Technology, USA, provides an expanding selection of research monographs in all
major areas of optics: lasers and quantum optics, ultrafast phenomena, optical spectroscopy techniques,
optoelectronics, quantum information, information optics, applied laser technology, industrial appli-
cations, and other topics of contemporary interest.
With this broad coverage of topics, the series is of use to all research scientists and engineers who need
up-to-date reference books.
The editors encourage prospective authors to correspond with them in advance of submitting a man-
uscript. Submission of manuscripts should be made to the Editor-in-Chief or one of the Editors. See also
www.springer.com/series/624
Editor-in-Chief
William T. Rhodes
School of Electrical and Computer Engineering
Georgia Institute of Technology
Atlanta, GA 30332-0250
USA
e-mail: [email protected]
Editorial Board
Ali Adibi Bo A. J. Monemar
Georgia Institute of Technology Department of Physics and Measurement Technology
School of Electrical and Computer Engineering Materials Science Division
Atlanta, GA 30332-0250 Linköping University
USA 58183 Linköping, Sweden
e-mail: [email protected] e-mail: [email protected]
Ferenc Krausz
Ludwig-Maximilians-Universität München
Lehrstuhl für Experimentelle Physik
Am Coulombwall 1
85748 Garching, Germany and
Max-Planck-Institut für Quantenoptik
Hans-Kopfermann-Straße 1
85748 Garching, Germany
e-mail: [email protected]
Wolfram Hergert Thomas Wriedt
•
Editors
123
Editors
Wolfram Hergert Thomas Wriedt
Martin-Luther-Universität Stiftung Institut für Werkstofftechnik
Halle-Wittenberg Badgasteiner Str. 3
Fakultät für Naturwissenschaften II 28359 Bremen
Von-Seckendorff-Platz 1 Germany
06120 Halle/Saale
Germany
Reference
v
Contents
vii
viii Contents
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Contributors
xiii
xiv Contributors
Wolfram Hergert
Abstract The year 2008 saw the 100th anniversary of the publication of Gustav
Mie’s work ‘‘Contributions to the optics of turbid media, particularly colloidal
metal solutions’’ in Annalen der Physik. This event was an occasion to express
appreciation for his contribution to the theory of scattering of electromagnetic
waves. The achievements of Gustav Mie in the framework of the development
of an unified field theory and his contributions to the discussion around the
formulation of the theory of general relativity have been mentioned many times in
other contexts. A detailed description of Gustav Mie’s work has, until now, been
unavailable. This contribution undertakes the task of presenting a comprehensive
account of the life and scientific work of Gustav Mie.
1.1 Introduction
W. Hergert (&)
Department of Physics, Martin Luther University Halle-Wittenberg, Halle, Germany
e-mail: [email protected]
1
[Mie81], p. 738f.
1 Gustav Mie 3
First, the stages of Mie’s scientific career are reviewed. His scientific work is
then highlighted in the second section. The connection between experimental and
theoretical physics was very important for him. The following discussion will be
concentrated on his theoretical work. Classical electrodynamics, field theory,
theory of relativity and the quantum theory are his main fields of research. The
work on the theory of matter and the contributions to the theory of the scattering of
electromagnetic waves are particularly important. Gustav Mie’s role as a lecturer
along with a description of his disciples is the main theme of the next section.
Mie’s publications document also his desire, to present modern physical devel-
opments to a broader audience. These activities are treated together with his role as
a lecturer.
The scientific achievements and conflicts of his long life were also overshad-
owed by two world wars and the time of the National Socialist movement. In this
context, his reflections of social developments will also be discussed, as far as they
can be concluded from his diaries, letters and secondary sources. I attempt to list a
bibliography of the publications of Gustav Mie, as completely as possible. A
bibliography of biographic notes on Mie is also provided.
Biographic details are specified in footnotes for persons who were important to
Mie’s scientific career and life; otherwise only biographical data is given. The
origins of the figures and the photographs which did not originate from the author
are explicitly cited.
Gustav Adolf Feodor Wilhelm Ludwig Mie was born on 29 September 1868 in
Rostock (Fig. 1.2) to Hans Friedrich Ernst Amandus Mie (1828–1906), a salesman
active in the insurance business and his wife Caroline, nee Ziegler (1834–1901).
His parents were both children of Prostestant pastors. Many of his Mecklenburgian
relatives were Protestant pastors.2 His uncle, Fedor Mie (1840–1905), worked as a
Protestant pastor at St. Petri in Rostock.
Gustav Mie spent his childhood and youth in Rostock. He had three brothers and
one sister. His brother Friedrich (1865–1911) became a high school professor in
Berlin. His brother Johannes Mie worked as a merchant in Hamburg. His brother
Amandus carried on with the family tradition and worked as a Protestant pastor in
Scharnebeck.3 Gustav Mie attended High School at the Great City School (Große
Stadtschule) in Rostock (Fig. 1.3), an educational institution rich in tradition, from
2
The origins of the branched family can be followed up in [3].
3
H. Spehl, Mie, Gustav Adolf Feodor Wilhelm Ludwig in: [Bio6] p. 186ff., [4], [Mie81], p. 734.
4 W. Hergert
Fig. 1.2 Mie’s birthplace (house on the corner), Wokrenter Strasse 35 in Rostock. The house was
all but destroyed in the war; however, it was carefully rebuilt in 1981–1982
1877 to 1886.4 The Christian Protestant family tradition instigated his desire for
studies in theology after his Abitur. During high school, his interest shifted from
theology to natural sciences. In his memoirs, he wrote about this change:
It was not until I was a sixth former that I got the idea of becoming a natural scientist and
mathematician. Even today, I can recall the blessedness I felt when this idea came to mind
and my parents agreed upon it. I have never been happier in my whole life than during
those days and it felt like I was wandering around in paradise. The fact that I took the
exam in Hebrew for my Abitur in 1886, typical of someone aiming at studying theology,
proved how late I made my decision.5
However, it was not all plain sailing. Mie’s brother Friedrich struggled to leave
his career as a Protestant pastor imposed on him by his parents. In the year 1886,
Mie received his school leaving certificate (Abitur). In the Abitur certificate
(cf. Fig. 1.4), he was assessed to have a commendable behaviour. His diligence
and attention was reported as, ‘‘in former times [as] insufficient, in the last terms
good and very good’’. He finished the subject Mathematics with the grade ‘‘very
good’’. ‘‘In Mathematics, he demonstrated ...good knowledge. Solving problems
4
The Great City School was installed in 1580 under the superintendence of the humanist Nathan
Chyträus (1543–1598) in the former Dominician monastery St. Johannis. In the years 1864–1867
the new classicistic building was built by the Rostock architect Th.H. Klitzing. Between Easter
1860 and Michaelis 1901, 7 pupils named Mie completed the Great City School [5].
5
[Mie81], p. 734.
1 Gustav Mie 5
Fig. 1.4 Part of Gustav Mie’s Abitur certificate. (UAF, Nachlass Mie, C136/1—Zeugnisse)
never caused difficulties for him. The exam results were very good. Therefore, he
was excused from the verbal examination’’.6 He also achieved a good grade in
Physics. The school leaving certificate states: ‘‘In Physics he always had keen
interest, a good understanding of natural phenomena and acquired a good
6
Archives University of Freiburg (UAF), Nachlass Mie, C136/1—Zeugnisse.
6 W. Hergert
Mie completed his studies at the university in his hometown Rostock as well as in
Heidelberg. He first enrolled in Rostock in the winter semester of 1886–1887.
After three terms, he moved on to Heidelberg in the summer of 1888, ‘‘to become
acquainted with something else’’ as he writes in his memories.8 It might be, that
also the Rostock mathematician Martin Krause (1851–1920), a disciple of Leo
Königsberger (1837–1921) the prominent mathematician in Heidelberg at that
time, contributed to his decision. The summer term of 1889 saw him back in
Rostock. He finished his studies in Heidelberg, where he spent the remaining time
from the winter semester 1889–1890 up to the summer semester 1890.
Initially, he placed emphasis on chemistry, geology and mineralogy. In the
winter semester of 1886–1887, 18 students were enrolled in chemistry at Rostock
University.9 Chemistry was lectured by Oscar Jacobsen.10 Chemistry in Rostock
had a good reputation at the time. Mie attended lectures on mineralogy and
geology given by Eugen Geinitz.11
Prior to the beginning of Mie’s studies in Rostock, mathematics, physics,
geology and astronomy were represented by Hermann Karsten,12 born into an old
Mecklenburgian family, which produced a number of scholars. Karsten was, at the
same time, director of the Navigation School. Only late, in contrast to other
German universities, did an independent chair for physics become installed in
Rostock. The first full professor was Heinrich Matthiessen13 who started at 1873.
Physics was housed in the same building as chemistry for a long time. As a result
of considerable extensions to the university available in the years 1870–1890 a
7
ibid.
8
[Mie81] p. 735.
9
[6], p. 1003.
10
Oscar Jacobsen (1840–1889), study in Kiel, 1865 assistant at the Chemical Laboratory of the
University of Kiel, 1868 graduation, 1871 private lecturer, 1873 professorship in Rostock, see
also [7].
11
Eugen Geinitz (1854–1925), study in Dresden and Leipzig, 1876 graduation Leipzig, 1877
habilitation Göttingen, private lecturer in Göttingen and Heidelberg, 1878 professorship in
Rostock. Geinitz is the father of the geology of Mecklenburg (Book ‘‘Geologie Mecklenburgs’’).
12
Hermann Karsten (1809–1877), studies in Bonn, Berlin and in Königsberg with Bessel, 1830
private lecturer in mathematics and mineralogy in Rostock, 1832 extraordinary professor, 1836
professor.
13
Heinrich Friedrich Ludwig Matthiessen (1830–1906) studies of natural sciences in Kiel with
G. Karsten, grammar school teacher in Jever and Husum, 1873 professor in Rostock.
1 Gustav Mie 7
Fig. 1.5 Left View of the old Institute of Physics in the year 1885 (still without the front
extension). On the back of the original photo Prof. Matthiessen noted the following in August
1885: ‘‘The Physical Institute—1885 photo taken by Dr. phil. Mönnich, Privatdozent, and given
to the Institute. The persons presented on it, from left to the right, are: Institute servant Maass,
cand. König, stud. Black (IInd assistant), cand. Karnatz (assistant) and teacher Klingberg’’, Right
Interior of the old Institute of Physics, persons from left to right Dr. Groesser, Prof. Matthiessen,
Dr. Brüsch and Maass ([9], pp. 96 and 99.)
building within the yard of the main building became available for physics, which
had already served as a chemical laboratory from 1834 to 1844 [8] (see Fig. 1.5). This
was the situation the student Gustav Mie encountered at the physical institute.
In the year 1879, the mathematical–physical seminar was established; the first
directors were physicist Hermann Matthiessen and mathematician and astronomer
Johann Martin Krause.
Paragraph 1 of the statutes of this seminar reads as:
The mathematical–physical seminar has to give the students suggestions and guidance for
independent investigations and free lectures in the field of abstract mathematics and in
mathematical physics. Subject to further regulations, the control of this institute will be
transferred to the professors of mathematics and physics under the superintendence of the
Ministry, Department of Education. Independent from each other they offer topics for
small and larger written work and free lectures and give suggestion and guidance to the
members for execution of the tasks. The details of the contents and scheduling of the
seminar-like exercises are left to the directors.14
The directors could nominate three students for one award per term. Mie was
awarded in the year 1889.15 Chemistry and physics, however, quickly slipped into
the background of Mie’s studies.
In the middle of the nineteenth century, Heidelberg was oustandingly
represented in physics, chemistry and mathematics. Chemist Robert Bunsen
(1811–1899) taught there from 1852 until retirement in the year 1889. Hermann
von Helmholtz (1821–1894) a versatile physiologist and physicist, held the chair
for physiology in Heidelberg from 1858 to 1870. Gustav Kirchhoff (1824–1887)
14
[10], p. 223.
15
[11], p. 828.
8 W. Hergert
Fig. 1.6 Left Leo Königsberger (1837–1921), mathematician, Right Harry Rosenbusch
(1836–1914), mineralogist and petrographer. (Courtesy of UAH)
was a full professor for physics in Heidelberg from 1854 to 1875. His successor
was Georg Hermann Quincke (1834–1924). Mathematics was also outstandingly
represented in Heidelberg. Following Ludwig Otto Hesse (1811–1874), Leo
Königsberger16 taught for a short period between 1869 and 1874 (Fig. 1.6).
Succeeding Immanuel Lazarus Fuchs (1833–1902), Königsberger taught from
1884 to 1914 for the second time in Heidelberg. In 1869, a mathematical–physical
seminar was established at Heidelberg university, with Kirchhoff and Königs-
berger as the first directors.17 In the year 1890, the subjects physics, chemistry,
botany, zoology, mineralogy, mathematics and agriculture were separated from
the Philosophical Faculty and combined to form a new Scientific-mathematical
Faculty. The physicist Quincke was the first dean and the mathematician
Königsberger became his deputy.
In spite of Quincke and Bunsen being outstanding scholars in physics
and chemistry, Mie continued to concentrate on mathematics and mineralogy.
16
Leo Königsberger (1837–1921) Professor in Greifswald, Dresden Vienna and Heidelberg,
work on the theory of elliptical and hyperelliptical integrals and complex differential equations,
Helmholtz biography, see also [12].
17
Statutes of the seminar in, Archives of University of Heidelberg (UAH), Fak. -Akte H-IV-102/
71, Nr. 78, fol. 75–76, cf. [13]. Statutes of such seminars are regularly printed in ‘‘Zeitschrift für
mathematischen und naturwissenschaftlichen Unterricht’’, for example in vol. 5, 173 (1874) the
statutes of the mathematical–physical seminar of Halle University can be found.
1 Gustav Mie 9
One reason for this choice may have been the absence of lectures in theoretical
physics. Mie acquired his knowledge in theoretical physics mainly by self-
instruction. It may be, due to his engagement in mineralogy, that Mie was not a
regular member of the mathematical–physical seminar; only students having
mathematics and physics as their main subjects could be regular members.
In Heidelberg, the mathematician Leo Königsberger and the mineralogist and
petrographer Rosenbusch18 (Fig. 1.6) were his academic teachers. Königsberger
was a excellent academic teacher. A student reported on his lectures:
Königsberger performed with virtuoso control of the material rapidly, clearly, dragging
the listener along. His fresh, self-confident, strong nature, as well as his kindness and
fairness secured him the affection of the academic youth.19
Mie completed the summer semester of 1889, once again in Rostock, and
concluded his study in the summer of 1890 in Heidelberg. The emphasis of his
study was mineralogy and mathematics. His preoccupation with mineralogy was
so intense that Harry Rosenbusch, the founder of the systematic petrography,
employed him as a junior research assistant. Mie’s task was to organise and
catalogue the hand piece and thin section collection consisting of 7,000 pieces.
Leo Königsberger remembers his student Mie in his memories. He writes:
My lecturer activity ... turned out well beyond all expectation, excellent young men, who
later occupied many chairs of mathematics, physics, astronomy at German and foreign
universities, like Ph. Lenard, M. Wolf, G. Mie, K. Boehm among others, whom I call my
scholars.20
Max Wolf (1863–1932) later was the director of the National Observatory of
Baden in Heidelberg and Phillip Lenard (1862–1947), Nobel Laureate of the year
1905, occupied the physics chair in Heidelberg.
Gustav Mie registered for the state examination of the higher teaching pro-
fession in 1890. He had to take the exam at the Technical University of Karlsruhe.
He requested the examination in mathematics and physics as major subjects, and
took mineralogy and chemistry as minor subjects. He also took an exam in religion
through his own volition. His topic in physics was ‘‘to deduce, from the laws of
induction, the basic principles necessary for the construction of dynamo-electric
machines, as well as it’s mode of action in a mathematical way’’. He worked on
this topic with such enthusiasm and interest, leaving himself only 24 h to write his
report on the mathematical topic. The situation was reflected in the assessment on
his mathematical work:
The written examination in mathematics ‘‘On the geodetic lines on the surfaces of second
order’’, even if the developments in the individual points could be more detailed and more
18
Karl Heinrich Rosenbusch (1836–1914), professor for petrography and mineralogy at the
universities of Strasbourg and Heidelberg, 1903 Wollaston medal, cf. also [14, 15].
19
Report of the student O. Rausenberger in [16], p. 179.
20
[12] p. 187.
10 W. Hergert
clearly expressed and also, the order and care of elaboration could be better, but overall,
the contents and conception of the task can probably be rated sufficient.21
In the summer of 1891, Mie attained a doctorate with the thesis ‘‘On the
Fundamental Principle of the Existence of Integrals of Partial Differential Equa-
tions’’. In the personal record for the thesis he wrote:
During his study time he [the author] visited the lectures of the following Mr. professors
and associate professors: Krause, Matthiessen, Geinitz, Jacobsen, Königsberger, Cantor,
Schapira, Rosenbusch, Osann, Goldschmidt, Andrae. To all his admired teachers, most of
all Mr. Geh. Rat Königsberger and Mr. Geh. Bergrat Rosenbusch, the author expresses
devoted thanks [Mie1].
After graduation, the pursuit of a scientific career seemed unlikely. Mie found an
employment in Dresden at a private school in summer 1892.23 Anyway, Mie sent a
copy of his thesis to Otto Lehmann, his examiner in physics in the state exami-
nation, and was offered an assistantship at the Technical University of Karlsruhe.
Otto Lehmann24 was the successor of Heinrich Hertz (1857–1894), who was active
from 1885–1889 at the TH Karlsruhe (Fig. 1.7). Hertz had accomplished his
innovative experiments on the propagation of electromagnetic waves at that time.
His successor Lehmann became famous for the investigation of liquid crystals.
Gustav Mie had to lead the physical practical course25 and used this in order
to gain appropriate knowledge in experimental physics. Furthermore, he had to
21
UAF, Nachlass Mie, C136/1 Zeugnisse.
22
ibid.
23
From the copy of testimonials existing in the archives, it was revealed that this refers to the
Müller Gelinek Six-form High School. This school had a long tradition and before the
transformation into a private school on the 26 April 1819, it was the urban Friedrich August
School.
24
Otto Lehmann (1855–1922), Study Strasbourg, Teacher for Physics, Mathematics and
Chemistry in Mulhouse (Alsace), 1883 Lecturer in Physics TH Aachen, 1888 Professor for
Electrotechnology at the Polytechnic Institute Dresden, 1889 Ordinarius at Karlsruhe Technical
University Fridericiana with teaching assignment for Physics and Electrotechnology.
25
see Lehmann’s report in [17], p. 84.
1 Gustav Mie 11
support Lehmann with the experiment lectures. The large number of students in
Karlsruhe and the impression of the lectures of Kundt (1839–1894) during his
study in Strasbourg brought Lehmann to dedicate himself to the intensive devel-
opment of experiment lectures.26 Regarding the experiments shown in the lectures
of Lehmann it is said:
Also today most experimental physicists prefer the demonstration of physical experiments
in the large-scale setups, which could be noticed from all places of the lecture room. This
style was consequently pursued by Lehmann. The enormous size of his experimental
assemblies and the spryness of the experimental presentation led to largest acceptance and
concentration among his students. To demonstrate the forces that could be produced by
electric currents, Lehmann, according to the verbal report of a student, catapulted a lab
assistant to the lecture room ceiling using current power. Although such experimental
setups could not become completely accepted, they surely remained not without influence
on the modern presentation of physical experiments. They have been certainly treasured
by his students, who reported after the lecture grinning due to ‘‘Circus Lehmann’’.27
26
see [17], p. 88f., see also [18].
27
[19], p. 108, private communication K.A. Turban 1881.
28
see [17], p. 72.
29
see: W. König, Gewinner und Verlierer. Der Stellenwert der einzelnen Hochschulen im
Institutionalisierungsprozess der Elektrotechnik in Deutschland 1882 bis 1914, in [20], p. 171ff.
30
In 1892 he was appointed as first budgetary professor for electrotechnics in Karlsruhe. On 11
July 1896 Schleiermacher became regular professor for theoretical physics in the electrotechnical
department of Engelbert Arnold. In 1926 the chair was rededicated to theoretical electrotechnics
after Schleiermacher’s retirement. cf. [21].
31
cf. [21], p. 25.
32
Engelbert Arnold (1856–1911), 1874–1878 studies in Zürich, 1880–1891 Polytechnic Institute
Riga, joint founder of the Russian-Baltic Electrotechnical Factory 1888, 1891 chief electrician of
Maschinenfabrik Oerlikon, from 1894 Karlsruhe, see also [22].
12 W. Hergert
Fig. 1.7 Engelbert Arnold (1856–1911), Otto Lehmann (1855–1922). (Archives Karlsruhe
Institute of Technology, Karlsruhe)
33
O. Lehmann, Das absolute Maßsystem, Verh. des Naturwiss. Vereins zu Karlsruhe 12, 365
(1897), excerpts in: Zeitschr. für den physikal. und chem. Unterricht 10, 77 (1897).
34
University and regional Library (ULB) Halle: Dissertationes polytechn. Karlsruhanae 1896–
1899, No. 11, Entwurf einer allgemeinen Theorie der Energieübertragung von Dr. Gustav Mie,
Habilitationsschrift zur Erlangung der venia legendi für mathematische Physik an der
Großherzoglichen Technischen Hochschule in Karlsruhe. (Sonderdruck aus den Sitzungsberich-
ten der K. Akad. der Wiss. Wien [Mie6]).
1 Gustav Mie 13
In the year 1900 Mie was appointed professor with the teaching assignment
‘‘Modern Trends and Opinions on Electricity and the Science of the Electrical
Waves’’.35
In the scientific life of the university town Karlsruhe scientific associations
were very important. Such an association was already created in 1840 by
Alexander Braun (1805–1877), the director of the Naturalienkabinett at that time.
Starting from 1862, the association arose as ‘‘Naturwissenschaftlicher Verein in
Karlsruhe’’ with new statute. The ‘‘Naturwissenschaftlicher Verein Karlsruhe
e.V.’’ has presently been active for 150 years. The report on the meetings of the
years 1888–1895 proves that 12 to 15 meetings per year took place. The mean
number of listeners was approximately 30. In average, physical topics have been
discussed five times a year. The professors and members of the Physical Institute
participated actively in this association since its inception. Heinrich Hertz pre-
sented his experiments for the first time to the public during one of the meetings of
this association. From 1885 to 1889 Hertz presented eight times. Thus, he lectured
on 24 June 1887 on ‘‘New Relationship between Light and Electricity’’36 and on
22 February 1889 he gave a lecture on ‘‘Relations between Light and Electric-
ity’’.37 The original apparatuses used by Hertz were still present in the institute
with Lehmann’s assumption of office. Lehmann wrote:
I came into possession of the physical cabinet which belonged to Hertz that contained
several intrinsically worthless items, however, amongst them were, from a historical
perspective, very precious apparatuses. There are in particular items, which in occasion of
the International Electrotechnical Exhibition in Frankfurt a. M. in 1891 occupied an
important place ... and will be later relinquished to the German Museum in Munich on its
strong request ...’’38
Otto Lehmann spoke regularly at the association meetings. Lehmann met with
the other members for meetings mostly in the physics lecture hall, in order to be
able to demonstrate experiments. The ‘‘Experimental Lecture on Röntgen’s
X-rays’’ presentation by Lehmann was held on 7 February 1896.39 While Hertz
and Lehmann participated regularly and actively in the meetings of the scientific
association, Engelbert Arnold, a member of the association too, hardly contributed.
Gustav Mie became a member during the 432th meeting on 16 December 1892.
He gave his first lecture on 5 January 1894, entitled ‘‘On the Nature of Heat’’.
During his time in Karlsruhe Mie gave seven additional lectures in the meetings of
35
[17], p. 84.
36
Verh. des Naturwiss. Vereins zu Karlsruhe 10, pp. 150–51 (1883–1889).
37
ibid. 11, pp. 41–43 (1883–1895).
38
cf. [17], p. 71, in 1913 the Karlsruhe devices were inventoried in the German Museum in
Munich. For the history of the devices see also: J.H. Bryant, Heinrich Hertz’s Experiments and
Experimental Apparatus: His Discovery of Radio Waves and his Delineation of their Properties,
in [23], p. 39 ff.
39
Verh. des Naturwiss. Vereins zu Karlsruhe 13, p. 37 (1895–1900), ‘‘The lecture, which was
given by the speaker, over 3 weeks on five occasions, for different corporations and associations,
is printed in the papers’’., see also [Mie4].
14 W. Hergert
40
Verh. des Naturwiss. Vereins zu Karlsruhe 13, S. 98–103 (1895–1900).
41
UAF, Nachlass Mie, E12/69.
42
For the history of physics in Greifswald see [25–28].
43
cf. [2] vol. 2, chapter 15, p. 33 ff.
1 Gustav Mie 15
faculty insisted in transfering the chair to König. In the course of this personnel
decision the replacement of the professor extraordinarius should be settled
at the same time.44 The proposal list for the occupation of the extraordinary
professorship covered Pringsheim (1881–1963) (Berlin), Elster (1854–1920)
(Wolfenbüttel), Wiechert (1861–1928) (Göttingen), Mie (Karlsruhe) and Straubel
(1864–1943) (Jena). Mie was set only on fourth place and the judgement over him
was however farsighted and lead the Minister to entrust the position to a ‘‘coming
man’’:
Mie’s work shows the fact that his mathematical skill is linked also with the ability of
physical opinion and important achievements for the future are to be expected from him.
His publications are however up to now purely theoretical in nature.45
The ministry insisted on a more detailed rationale, and in the long run Mie got
the position. Both appointment procedures show that theoretical physics began to
be established that time. On 17 September 1905, Mie became a full professor
under the condition that he represented physics in its entire extent. At the same
time he was appointed the director of the Physical Institute and the Astronomical–
mathematical Institute. From winter semester 1836–1837 until 1887 the Astro-
nomical–mathematical Institute belonged to the Mathematics department. The
Astronomical–mathematical Institute was moved to Physics department during the
directorship of A. Oberbeck in the winter semester 1887–1888. Lectures in
astronomy were given by W. Ebert (1871–1916) from 1903 till 1905.47 During
Mie’s directorship astronomy was abandoned. It was reactivated in 1922.48 Gustav
44
Archive of the University of Greifswald (UAG), Phil.Fak I290.
45
ibid.
46
UAG Phil. Fak. I 298.
47
cf. A. Schnell, Gestrandet in Wien: Wilhelm Ebert (1871–1916), Acta Hist. Astr. 43, 318
(2011).
48
The building of the physical institute also contained an observatory. A modern astrodome on
the tower can be seen on Fig. 1.8, cf. H. Kersten, Astronomie in Greifswald in [25], p. 51ff.
16 W. Hergert
Fig. 1.8 The Physical Institute in Greifswald (built 1889–1891). A commemorative plaque is
posted on the wall of the building. The text reads: ‘‘Professor Dr. Gustav Mie (1868–1957)
cofounder of a uniform theory of particles and field, Greifswald 1902–1917’’, see also [24]
Mie was Dean of the Philosophical Faculty in Greifswald in the year 1912–1913,
and was elected rector of the university on 15 May 1916 for a one-year period. He
began his rectorship with the lecture ‘‘Law of Nature and Spirit’’ [Mie39].
In the year 1916, Mie received an offer to be the successor of Ernst Dorn in
Halle. After some hesitation and repeated negotiations with the ministry he
accepted the offer in December 1916. As candidates for his successor, he brought
Johannes Stark (1874–1957) (Aachen), as well as in second place, pari loco
Johannes Königsberger (1874–1946) (Freiburg) and Clemens Schäfer (1878–1968)
(Breslau) to discussion. Mie had already become acquainted with Stark during his
time in Greifswald where Stark was serving as a substitute professor, in the years
1907–1908, for professor extraordinarius Starke (1874–1960), who was later an
ordinarius at the RWTH Aachen from 1917 to 1940. The faculty at the end
followed Mie’s proposal.
Inferred from his diaries, it was a happy time for Mie personally in Greifswald,
despite difficulties due to the World War. During one evening on a sailing yacht
Stark reported:
He [Mie] had an assistant named Falkenberg, who possessed a beautiful large sailing yacht
with a habitable space. It lay on the Ryck opposite Wieck. Once we spent an amusing
evening on it. It had already become dark, when we rose to the deck, in order to go ashore.
It followed Guschen, as he was referred to by Mrs. Mie, deep in thought as usual, he
continued to rise when he fell into the water which was, fortunately, only as deep as his
1 Gustav Mie 17
height. Despite the unfortunate accident we had to laugh at the sudden disappearance of
the great philosopher. We pulled him out rapidly and he laughed, too.49
In 1890, at practically the same time as the Physical Institute in Greifswald, the
Physical Institute of Halle University also got a new modern functional building.
Also in Halle, special care was taken to set up a building that meets all the
requirements of the physical research at this time [30]. Here, Ernst Dorn50
acquired special merits. As successor of Oberbeck in Halle, he was considerably
involved in the construction of the new institute.
After 1870, theoretical physics itself began to be institutionalised in Halle.51
With Anton Oberbeck, Halle had a professor extraordinarius for theoretical
physics starting from 1878. In 1884, Oberbeck received an offer to Karlsruhe, his
position was subsequently converted to a personal chair in order to maintain a
position for him in Halle. The vacant chair in Karlsruhe was occupied by Heinrich
Hertz. Hertz had an offer from Greifswald, but preferred Karlsruhe due to
the excellent conditions in comparison to those at the institute in Greifswald.
Oberbeck’s personal professorship was occupied by Ernst Dorn later on. After the
death of the full professor of experimental physics, Carl Hermann Knoblauch,52
Dorn took over his chair for experimental physics and became the director of the
institute. The second position was downgraded to an extraordinary professorship
for theoretical physics and transferred to Karl Schmidt.53 This decision was
unfortunate, to that extent, since Schmidt’s field of interests had been more
technical and applied physics than theoretical physics.
The institute in Halle was larger than the institute in Greifswald, had a higher
reputation and even a higher remuneration. Theoretical physics, which was Mie’s
main focus, at least formally, was already institutionalised in Halle. Advancing
this institute must have appeared as an appealing task to Mie, so much so that he
accepted the offer from Halle. However, the hearings of the appeal were
49
[29], p. 35.
50
Ernst Dorn (1848–1916), studies in Königsberg, teacher in Königsberg and Berlin, 1873
habilitation in mathematics Greifswald, 1873 extraordinarius for physics in Breslau, 1881
professor in Darmstadt, 1886 professor in Halle as a successor of Oberbeck, 1895 director of the
physical institute, see also [30].
51
see [2] vol. 2, p. 33ff.
52
Carl Hermann Knoblauch (1820–1895), studies in Berlin, 1845 cofounder of the Berlin
Physical Society, 1848 habilitation in Berlin, 1849 professor in Marburg, 1853 professor in Halle,
1886–1871 rector of the University of Halle, 1878 President of the Leopoldina, see also [30, 31].
53
Karl Schmidt (1862–1946), studies in Göttingen and Berlin, assistant in Strasbourg and
Königsberg, 1889 habilitation Halle, 1895 extraordinary professorship for theoretical physics,
1912 personal tenured professorship.
18 W. Hergert
complicated. The special position held by Schmidt, being the cause for conflicts
over the next years, was a contributing factor.
During the regulation of the successor to Dorn, the faculty first mentioned Peter
Debye (1884–1966), then Max von Laue (1879–1960) and in third place, Gustav
Mie. As criteria for the selection the faculty stated:
Conserving the existing ordinaries is of outmost importance for the faculty. In addition, we
would like to engage a scientific personality capable of arranging the academic instruction
together with the present full professor in a harmonic and versatile manner. Another
decisive aspect is that the former contrast between experimental and theoretical research
has vanished and the present full professor’s effort has developed in the more experimental
direction. Therefore the faculty agreed on personalities with focus on theoretical aspects
without excluding experimental work.54
Debye and Laue were not to be attracted for Halle. Corresponding to the
qualities of Mie the faculty noted:
As a distant third choice following these two researchers we propose Gustav Mie,
Greifswald. ...The number of the Mie’s papers is not as large as those of the two physicists
selected first, nevertheless the treated problems and the kind of their solution marks Mie as
a researcher of outstanding gift, deep thoroughness and great originality.55
A large portion of the documents in the archive of Halle University from Mie’s
time in Halle reveals his conflicts with Karl Schmidt. Schmidt, who had large
support within the faculty and from the chancellor, steadily tried to increase his
influence in the institute. Mie’s plans for the modernisation of the institute in Halle
were thwarted by Schmidt.
During his time in Halle, Mie actively participated in the sessions of the
so-called ‘‘Naturforschende Gesellschaft’’. This society was founded already in
1779. On 20 June 1918 he spoke about ‘‘The study of the physics of the inner atom by
X-rays’’ and on 8 July 1919 about ‘‘The decay of atoms during radioactive pro-
cesses’’.56 Other scientific societies were also active in Halle. In 1917 the so-called
‘‘Hallescher Verband für die Erforschung der Mitteldeutschen Bodenschätze
und Ihrer Verwertung’’ was founded. Mie was also interested in such activities.
He was a member of this society until his departure to Freiburg. On 28 September
1917 he wrote:
On Wednesday, there was a great occasion: Foundation of Halle’s Federation for the Study
of the Mineral Resources. I hope that this will result in my increased contact with the
industrial circles, because I want to gain much knowledge of the technology. That’s my
largest desire.57
54
Archive of Halle University (UAH), PA11453: Peronalakte Mie 1916–1918.
55
ibid.
56
cf. Mitteilungen der Naturforschenden Gesellschaft zu Halle 5, 53f. (1920), [32].
57
UAF, Nachlass Mie, E12/71.
1 Gustav Mie 19
Although his theory of matter was conceived in Greifswald, the discussion and
exchange of letters about field theory and general relativity with Einstein mainly
occurred during his time in Halle. Mie’s time in Halle was scientifically fruitful,
although he could not realize all his ideas for the development of the institute.
On the other hand, Halle was intellectually exciting for him. Mie and his wife
regularly visited the theatre and concerts. He visited the art gallery near the institute
building, but did not get acquainted with expressionism, a main collection area of
the gallery. He met also ‘‘old Rostock guys’’ like the philologist K. Brockmann
(1868–1956) and the botanist G.H. Karsten (1863–1937). Both professors also
received their Abiturs from the Great City School in Rostock. While in Freiburg, on
26 October 1924, Mie wrote about that time in Halle:
I think back at Halle, as a time of my highest mental stimulation, in contrast to the
enormously rich industries and crowds of dissatisfied workers living in darkness, to
the mental contrasts of men of the industry, completely oriented on power and will, and
the religious powers, before with which I had never become acquainted. It was the most
interesting, spirited interplay of brightness and darkness, although somewhat though on
20 W. Hergert
the nerves at the time. Now, I withdrew myself here into peace, because I still want to
muster my strength, in order to produce something.58
Gustav Mie did not fully assimilate with Halle, probably due to the conflicts with
K. Schmidt. The political riots at the beginning of the 1920s in central Germany
surely was another reason for him to look for a position at a respected university.
His positive memories of the first years of his scientific career in Karlsruhe and
his wife’s family ties made a return to Baden seemed reasonable. In a letter to
Sommerfeld he wrote:
I have to say that I am looking forward not only to leaving here, but also to being at the
beautiful Freiburg institute and returning to the beautiful Baden country, with which I am
so familiar.59
58
UAF, Nachlass Mie, E12/72.
59
UAF, Nachlass Mie, E12/37 Nr.18.
60
Franz Himstedt (1852–1933), studies in Göttingen, graduation and habilitation University of
Göttingen, 1878 private lecturer Göttingen, 1880 private lecturer Freiburg, 1889 successor of
Röntgen in Gießen, 1895 professor in Freiburg.
61
Emil Warburg (1846–1931), studies in Heidelberg and Berlin, 1867 graduation, 1870
habilitation, 1872 Strasbourg, 1876 professor in Freiburg, 1894 professor in Berlin, 1905 director
of the Physikalisch-Technische Reichsanstalt.
62
Wolfgang Gaede (1878–1945), studies in Freiburg, 1901 graduation, 1909 habilitation in
Freiburg, 1913 professor in Freiburg, 1919 professor in Karlsruhe.
63
UAF, Generalakten, Bestand B1/1287.
1 Gustav Mie 21
to accept the offer from Freiburg. We did not mention him before, because we considered
it unlikely that he might be willing to accept our offer, however, it seems like he did not
get assimilated in Halle and would like to come to South Germany. Prof. M. Wien, Jena,
writes about him: ‘‘Among the living German physicists, I consider him among those of
highest rank, not only as a theoretician, but also as an experimentalist’’. Not only is he an
outstanding researcher, but also a very successful teacher, as proven by numerous
experimental and theoretical theses accomplished under his supervision. His style of
presentation is plain and clear. Mie is a subtle, highly educated man and a pleasant, kind
colleague.64,65
64
ibid.
65
Dissertations with purely theoretical content supervised by Mie could not be found.
66
UAK, Best. 27059, Sig. 1, 385.
67
ibid.
68
cf. UAF, B24/2448.
69
In honour of Mie’s time in Halle the new main physics lecture hall of Halle University was
named ‘‘Gustav-Mie-Hörsaal’’. Also, a so-called Mie prize is awarded to excellent students of
physics in Halle.
22 W. Hergert
The name of Gustav Mie is surely most frequently mentioned in connection with
the term ‘‘Mie scattering’’ and ‘‘Mie effect’’. His contributions to the development
of field theory and general theory of relativity are acknowledged in many reviews
[33, 34]. Other scientific contributions of Gustav Mie are still of great interest.
In a certain manner, Mie’s scientific papers are standard examples in the frame-
work of the sociology of scientific knowledge. Corresponding to R.K. Merton the
following thesis holds: ‘‘all scientific discoveries are, in principle, multiples,
including those that, on the surface, appear to be singletons’’.72 Stigler’s law of
eponymy states that the naming of an effect, a law, an equation rarely follows the
name of the discoverer.73 The scientific community has to be prepared also for the
breakthrough of a new theory. ‘‘If an early valid statement of a theory falls on deaf
ears, and a later restatement is accepted by the science, this is surely proof that the
science accepts ideas only when they fit into the then-current state of science’’.74 Mie
scattering can be discussed in such a context. In the Dictionary of the named Effects
and Laws in Chemistry, Physics und Mathematics, Mie does not appear as associated
with Mie scattering, instead, the ‘‘Mie-Grüneisen equation of state’’ is cited.75
70
Letter of Mie to Sommerfeld from 1 Februaray 1935, UAF, E12/37, Nr.23.
71
The value of those articles should not be discussed here. It was a subject of controversy among
the editors of the planned Mie edition Höhnl and Plötze. cf. UAF, Nachlass Hönl E14/2, see also
review to Mie’s ‘‘Naturwissenschaften und Theologie’’, Die Naturwissenschaften 20, 566 (1932).
72
[35], p. 298.
73
Stigler’s law of eponymy. [36], cf. also [37].
74
[38], p. 146.
75
[39], p. 212.
1 Gustav Mie 23
Table 1.1 Distribution of Gustav Mie’s publications on the stations of his scientific career and
on his scientific fields of activity
Period Station Publications
1892–1902 Karlsruhe 12
1903–1918 Greifswald 33
1919–1924 Halle 14
1925–1936 Freiburg 21
1937–1950 Retirement 7
Field of activity Publications
Field theory, theory of relativity 17
Electrodynamics, electrotechnics 10
X-rays 8
Thermodynamics, statistics 7
Quantum physics 6
Mathematics 4
Scattering theory, optics 3
Textbook /textbook contributions 4
Philosophy, religion 7
Popular science 8
Biographic contributions 4
Miscellanea 9
a b
/¼ þ ; ð1:1Þ
qm qn
76
H. Hertz, Untersuchungen über die Ausbreitung der elektrischen Kraft, Gesammelte Werke,
Leipzig 1894, Bd. 2, S. 234 und S. 294 and G. Helm, Die Energetik nach ihrer geschichtlichen
Entwicklung, Leipzig 1898.
77
[40], p. 187f.
78
[42], E. Grüneisen: Zustand des festen Körpers, p. 1 ff.
1 Gustav Mie 25
where the first term represents the attractive and the second term the repulsive force. a and
b are constants.79
Mie considered [Mie13] mainly single atomic gases, but Grüneisen [43]
extended the field of application of the potential to solids. Both authors did not
speculate about the origin of the forces. Experimental results for the coefficients m
and n are given for example by Fürth [44]. Usually the Mie potential is written in
the form
n n rm m rn o
VðrÞ ¼ D0 : ð1:2Þ
mn r mn r
The Mie potential is a precursor of the Lennard–Jones potential, which fixes the
exponents to m ¼ 12; n ¼ 6. The Lennard–Jones potential is written as
r 12 r6 rm 12 rm 6
V ¼ 4e ¼e 2 : ð1:3Þ
r r r r
79
ibid, p. 11. Grüneisen refers in the discussion on Mie’s paper [Mie13] and his own article
published considerably later [43].
80
z.B. W.C. de Markus, Planetary Interiors, S. 441 in [57].
81
s. [42], p. 22, [Mie13], [43].
26 W. Hergert
wire in 1899 [59]. In the year 1900, Mie’s article containing the complete theo-
retical solution of the Lecher problem appeared [Mie9]. Mie gave a complete
mathematical analysis, a style we encounter later in his treatment of the scattering
problem.
Sommerfeld answered on the consignation of the article on the parallel wire
system by Mie:
Dear colleague, Many thanks for the consignment of your treatise. It was, as you can
imagine, exceptionally interesting for me. I am personally not only glad about the sci-
entific progress, but also that my wire waves got such a competent and thorough reader.
...You have done an ample piece of work!82
82
Letter of Sommerfeld to Mie from 1 July 1900, UAF, E12/37 Nr.1.
83
Letter of Mie to Sommerfeld from 9 July 1900, UAF, E12/37 Nr.2.
1 Gustav Mie 27
conceptions used so far for the processes during the short circuit. Therefore the publication
of this work is still of interest.84
Apart from the development and construction of generators and motors, the
rating and improvement of distribution networks was an important issue of that
time.85 Mie studied heat conduction in stranded cables, a problem important for
the dimensioning of such cables [Mie17, Mie19].
84
cf. [Mie8], p. 97.
85
cf. [63], p. 132ff.
86
Instead of Mie theory, occasionally the terms Lorenz-Mie theory or Lorenz-Mie-Debye theory
are used.
87
Walter Steubing (1885–1965), Ph.D 1908 Greifswald, TH Aachen (Stark), 1927 professor for
applied physics Breslau, 1948 professor in Hamburg.
28 W. Hergert
factory Schott und Genossen in Jena. It was possible to correlate colour changes
with particle sizes and particle forms by means of ultramicroscopy. In 1905,
Zsigmondy published a review on the ultramicroscopic investigations on colloids
[70]. Zsigmondy wrote:
A series of examples should be used to demonstrate, how by an incremental fragmentation
of a solid—the metallic gold—the properties of the same will change. The fragmentation
of the solid should be continued as far as possible; if possible down to molecular
dimensions. On the other hand, size and properties of the single particles, obtained as a
result of the fragmentation, should be determined exactly.88
In 1906, the journal ‘‘Zeitschrift für Chemie und Industrie der Kolloide’’ was
founded.90 The introduction pointed to Zsigmondy’s book [70] and stated, ‘‘...that
we are at the doorstep of a new great area, where science has to develop and
industry to exploit—this is the colloids’’.91 In 1902, Mie studied Planck’s dis-
persion theory [71, 72]. This can be concluded from the exchange of letters with
Max Planck (1858–1947).92 The reasons that motivate the investigations of
Steubing were revealed clearly in the report ‘‘The optical properties of colloidal
gold solutions’’ [Mie23] published in ‘‘Zeitschrift für Chemie und Industrie der
Kolloide’’ in November 1907. The topic of Steubing’s and Mie’s investigation was
perfectly suitable for the new journal. Part of the reason why this journal was
chosen was probably due to its fast publication turnaround time. This article was a
report on a talk given by Mie on 18 September 1907 on the 79th ‘‘Versammlung
Deutscher Naturforscher und Ärzte’’ in Dresden.93 During the discussion of the
talk Siedentopf pointed out a paper by J. C. Maxwell Garnett (1880–1958) [73]
unknown to Mie that time. Siedentopf also discussed the possibility extending the
calculations to elliptical particles.94 Walter Steubing defended his thesis on 16
December 1907 in Greifswald [74]. In 1908, a corresponding publication appeared
in ‘‘Annalen der Physik’’ [74]. Steubing himself did not come back to the topic of
his Ph.D work in his future work. He started his scientific career with Johannes
88
[70], preface.
89
ibid., p. 112.
90
vol. 1 (1906)–vol. 12 (1913) Zeitschrift für Chemie und Industrie der Kolloide, vol. 13
(1913)–vol. 179 (1961) Kolloid-Zeitschrift, vol. 180 (1962)–vol. 251 (1973) Kolloid-Zeitschrift
& Zeitschrift für Polymere, than Colloid & Polymer Science.
91
Zeitschrift für Chemie und Industrie der Kolloide, 1 (1906).
92
UAF: E012/38, exchange of letters Mie-Planck, Nachlass Gustav Mie.
93
see also the publication in ‘‘Berichte der Dt. Physikal. Gesellschaft’’ [Mie24].
94
see also ‘‘Verhandlungen der Gesellschaft Deutscher Naturforscher und Ärzte’’, 79, 40 (1907).
1 Gustav Mie 29
Stark (1874–1957) in Aachen and published three papers with him on canal rays
from 1908 to 1910.
The first communications on the work done in Greifswald appeared at the end
of the year 1907. Mie sent his complete paper to Wilhelm Wien, editor of
‘‘Annalen der Physik’’, in January 1908. Mie wrote:
Dear colleague,
I have enclosed within an article of theoretical work on the optics of turbid media, to
request publication in the Annalen. Perhaps, I will send more soon, but hopefully with less
voluminous amount of material. ...Eventually it will be possible to meet you at Easter.
Maybe I intend to start with some kind of snow sport, because I see that I need some
physical fitness. If it is not possible to meet you in Würzburg, I would go to the Bavarian
mountains for a few days.95
I confirm the receipt of your work on optics of turbid media, which discusses a really
relevant topic. I would be very happy to welcome you here and to discuss a number of
things. At the end of the winter semester I have to go to Mittenwald immediately to stay
there for several weeks. I have to advise you strongly to start skiing. ...Because, as always,
when a number of theoreticians gather together, it is also a time for scientific
conversation.96
The submission of the paper was practically also the starting point for Mie’s
participation in the winter sports meetings in Mittenwald.
The publication of the article [Mie25] marked the conclusion of Mie’s work on
this topic. This was often considered as exceptional in retrospect. It is not
uncommon for a scientist to keep working on the same topic for years or decades
to become a specialist in a narrow field. Mie’s behaviour can be understood from
his style of work and the development of physics till that time. In his work on the
Lecher problem, a complete mathematical solution of a certain physical problem
was given. Also, in case of the seminal contribution to scattering theory, a com-
plete mathematical solution for spherical particles was given. An extension to
spheroids, when the expected difficulties in the numerical evaluation were taken
into consideration, did not look very promising. Mie was also busy with his
‘‘Lehrbuch der Elektrizität und des Magnetismus’’, which appeared 1910. His book
was dedicated to the electromagnetic program and in its intention exceeded the
solution of a single problem. Mie himself wrote in the preface:
It is without any doubt, that the actual foundation of the whole physics is electricity.
Electricity fills the place, that was occupied 100 years ago by mechanics. Mechanics itself
as well as optics becomes more and more a part of electricity.97
95
UAF: E01248, 16, exchange of letters Mie-Wien, Nachlass Gustav Mie.
96
UAF: E01248, 13, exchange of letters Mie-Wien, Nachlass Gustav Mie.
97
[Mie27], Vorwort, S. VII.
30 W. Hergert
Table 1.2 Summarizing timetable to the development of scattering theory, based on [77]
Year Author Ref. Achievement, Remark
1863 A. Clebsch [78] Elastic wave equation
1871 Lord Rayleigh [79] ‘‘Rayleigh’’ scattering
1872 Lord Rayleigh [80] Scattering of sound by a sphere
1881 Lord Rayleigh [81] Scattering by a dielectric cylinder
1881 H. Lamb [82] Elastic waves
1890 L.V. Lorenz [83] Results analogous to [Mie25, 84]
1893 J.J. Thomson [85] Scattering by a conducting sphere
1899/1900 G.W. Walker [86, 87] Scattering by a sphere
1908 G. Mie [Mie25] Complete reference solution
1909 P. Debye [84] Debye potentials
1910 J.W.N. Nicholson [88] Scattering by a sphere
1915 H. Bateman [89] Complete review
1917 A.J. Proudman et al. [90] Numerical results to [91]
1920 G.N. Watson [92] Theory of Bessel functions
1920 T.J. Bromwich [91] Complete solution (started in 1899)
Einsteins annus mirabilis was three years prior in 1908. The theory of special
relativity was intensively discussed and Mie found more excitement thinking about
a new foundation of a theory of matter instead of puzzling over details of a special
problem. A sign of such an attitude was his participation in the congress of natural
scientists in Salzburg in 1909. Einstein took part in such a large congress for the
first time. Mie participated in the discussions on talks about basic questions in
electromagnetism and the theory of special relativity. Contributions to the dis-
cussions on other topics are not documented.98 This demonstrated that his main
interest was already focussed on topics other than scattering theory.
The historical development of scattering theory will be now summarised.99
Table 1.2 lists a few milestones in the development of the scattering theory, that
were mainly, as it were, precursors to Mie’s paper. Clebsch (1833–1872) [78]
investigated boundary value problems in elastic media, especially of sound waves
impinging on spherical surfaces. Clebsch’s work is fundamental from several
points of view. It had already contained work on potentials that were later named
Debye potentials. Clebsch’s treatise was published before Maxwell’s theory. It
presented the theory in component form as opposed to the modern notation of
vector analysis. This might be the reason why the work of Clebsch is only rarely
cited. The work of Lord Rayleigh (1842–1919) on the exact solution of the
scattering problem of a dielectric cylinder [81], published in 1881, is based on
Maxwell’s theory. Lamb (1849–1934) [82] solved the vectorial wave equation
with methods, used also by Clebsch, and provideed a basis for a series of further
publications.
98
see Verhandlungen der Gesellschaft Deutscher Naturforscher und Ärzte, 80 (1908), 81 (1909).
99
The remarks are based on the the reviews of Kerker [75] and Logan [76, 77]. For a detailed
discussion see especially [77].
1 Gustav Mie 31
Ludvig Lorenz100 (Fig. 1.10) is mainly known as the ‘‘second Lorenz’’ from the
Lorentz-Lorenz formula and, incidentally, the so-called Lorenz gauge is often
erroneously attributed to H.A. Lorentz.101 Lorenz’s work from 1890 [83] is based
on the work of Clebsch and contains the exact solution of the scattering problem
for the sphere. The Lorenz theory is not based on Maxwell’s theory, but based on
the exact solution of the same boundary problem. This work has remained prac-
tically unknown. As already mentioned, Mie was pointed to a paper by Maxwell
Garnett at the conference of natural scientists in Dresden 1907. Mie cites Maxwell
Garnett [73] in his paper, but also an older publication of L. Lorenz [96]. In
contrast to Mie, to Maxwell Garnett, the work of Lorenz from 1890 was known.
He cited the Lorenz paper [83] together with publications of Lord Rayleigh and
H. Hertz in the first part of his own publication. Therefore, Mie could have found a
hint to the work by Lorenz. However, Lorenz’s original work in the Danish lan-
guage was most likely not known to him. Nicholson (1881–1955) [88] investigated
100
Ludvig Lorenz (1829–1891), studies in Kopenhagen, 1852 diploma as chemical engineer,
September 1858- Juli 1859 studies in Paris, 1866 teacher at the academy of military sciences in
Kopenhagen, 1887 working as a free scientist, funded by Carlsberg, 1876 professor, 1887 doctor
honoris causa of Uppsala University (see also [94, 95]).
101
see for instance H. Klingbeil, Elektromagnetische Feldtheorie, Teubner-Verlag Stuttgart,
2003; W. Nolting, Grundkurs Theoretische Physik 3—Elektrodynamik, Springer-Verlag, Berlin
2002.
32 W. Hergert
the light scattering from a metallic sphere and compared the results with those from
geometrical optics. He also developed a series of asymptotic formulae for Bessel
functions, necessary for the discussion of the scattering on large spheres. The same
formulae have been developed earlier by Lorenz and were part of his 1890 work.
The publications of Mie [Mie25] and Debye [97] remained largely unknown in
England as exemplified by the Nicholson paper [88]. This lack of knowledge of the
work of others is in strong contrast to the complete overview, given by Bateman for
the first time [89]. The bilateral appreciation and knowledge of scientific work
disembogued at the beginning of Word War I in appeals, that depressed the
atmosphere in the international scientific community for a long time (see also
Chap. 5). Finally, the work of Bromwich (1875–1929) and his co-workers
Proudman, Dodson and Kennedy should be mentioned. Bromwich published his
paper in 1920 [91], but the exact solution was already known to him in 1899.
Proudman et al. [90] performed detailed numerical calculations, published in 1917.
Logan summarised the distinctive feature of Mie’s contribution, which resulted
in the paper to be generally accepted as a reference, in an apposite manner:
What Mie did that was different from the work of previous writers was to set out on an
ambitious computing program. He made calculations which involved summing the first
several partial waves, in order to be able to obtain numerical results for spheres, which
were too large to qualify for the criterion to be small enough for only the first partial wave
(Rayleigh scattering) to be important. Mie’s paper is complete in itself, and it served as the
basis for much of the work which has been done in this field since its publication. ...Mies
paper caught the attention of his and later generations because he employed the results of
his very thorough paper to study a very interesting practical problem, and others, who
came after him, found that his paper provided them with a good and a complete, guide to
follow in their applied work in light scattering.102
In the third edition from 1911 his intensive study of the subject is reflected in
the following reformulation of that passage:
That the colloidal metal solutions show totally different colors than the usual turbid media
is related to the special optical properties of metals. They show with respect to the
different types of light a very strong selective behavior, certain colors are strongly
absorbed, whereas other colors are strongly reflected. In the red gold solutions, for
example, a very strong absorption of green light by the gold particles occurs, but for blue
solutions one obtains mostly an abnormally strong lateral emission of yellow-red light.
102
Logan [77], p. 9 and footnote 16.
103
[Mie18], p. 57 1st edition, p. 61 2nd edition.
1 Gustav Mie 33
The theory of Lord Rayleigh, which we have already discussed, is valid only for particles
of such substances, which do not strongly absorb or reflect, that means substances which
would appear as a form of a coarse powder that is white or only weakly colored. Most
substances belong to this class.104
Even today, Gustav Mie’s publication ‘‘Contributions to the optics of turbid media,
particularly colloidal metal solutions’’ is the basis of many practical applications,
but his contributions to the theory of matter are interesting only in a historical
context. His three publications ‘‘Grundlagen einer Theorie der Materie’’ [Mie32,
Mie33, Mie36] have had strong influence on the development of field theory at that
time. Mie’s contributions in this area are appreciated and discussed in-depth in a
series of reviews and books.105
Up to his contribution to scattering theory, Mie’s scientific work was much
influenced by Maxwell’s theory. He contributed to the extension of the theory and,
propagation by way of his textbook [Mie29]. Therefore it is not surprising, that
Mie is seen, from a historical point of view, as a proponent of the electromagnetic
field program. The unified description of electric, magnetic and optic phenomena
based on Maxwell’s theory as well as the success of electron theory of H.A.
Lorentz (1853–1928) provided the electromagnetic field program a dominant role.
The program was formulated mainly by E. Wiechert (1861–1928) and J. Larmor
(1857–1942) in the 1890s and was at the turn of the century the dominant concept.
Significant contributions made by, to name only a few, M. Abraham (1875–1922),
W. Wien (1864–1928), A. Sommerfeld (1868–1951), J.J. Thomson (1856–1940),
H.A. Lorentz (1853–1928) and H. Poincaré (1854–1912). Vizgin states:
The radical form of the electromagnetic program advanced by Wiechert in 1894 and
somewhat later by Larmor declared the ether to be primary reality; its excited states gave
the charged particles (electrons), and the origin of their mass was explained on the basis of
the concept of electromagnetic mass developed in the 1880s and 1890s, primarily by
British scientists (J.J. Thompson, Heaviside, and also, somewhat later, Searle and Morton).
It was assumed that the laws of Newtonian mechanics could be deduced from the equa-
tions of the electromagnetic field.106
The persistent result of this effort was the introduction of the field concept and
the attempt to unify physical phenomena on this basis. Mie’s substantial work with
the field concept and the attempt to develop a new theory of matter in the
framework of the electromagnetic program is rather connected with the end of this
development. Late highlights in this sense are the unified theories of Hilbert
104
[Mie18], p. 66 3rd edition.
105
see Vizgin [98], Kohl [33], Renn (Ed.) [99, 100], Corry [101].
106
[98], p. 7.
34 W. Hergert
At the end Mie missed his grandiose goal, but his formalism has had a large
influence on the further development of unified field theories. A sign of the
importance of Mie’s work is that Weyl discusses the theory in his famous book
‘‘Raum-Zeit-Materie’’.108
107
UAF, Nachlass Mie, E12/68, diary note from 28 October 1915.
108
See [103] and also [104].
1 Gustav Mie 35
Hilbert included Mies’s work in his field theory and mentioned this in his
publication at the very beginning. Hilbert wrote:
The tremendous problems stated by Einstein and his subtle methods conceived to their
solution as well as the profound ideas and original conceptualisations whereby Mie for-
mulates his electrodynamics, opened new routes for the investigation of the foundations of
physics. I want in the following—in terms of the axiomatic method—essentially starting
from two simple axioms, to develop a new system of fundamental equations of physics,
which are of ideal beauty. I believe that the equation contains the solution of the problems
of Einstein and Mie simultaneously.109
Mie was very excited about the progress made by Hilbert. After the receipt of
Hilbert’s publication he noted:
Yesterday a separate print of Hilbert’s paper ‘‘The Fundamentals of Physics’’ arrived. In
this essay he combines my theory of matter with the principle of general relativity, which
Einstein aims at, but only Hilbert really fulfils. It is a completely wonderful work! With
what a strong fist this Mathematician strikes hard! He smashes the largest obstacles to
smithereens. My boldest dreams are exceeded. I believe that it will not take long, before
we really get the true world function and will be able to create matter like god in a
mathematical way. I just wrote a letter to Hilbert. I am extremely excited about this
event.110
Mie was also invited to the Wolfskehl lectures111 in 1917. In 1915, Einstein was
invited to give Wolfskehl lectures and M.v. Smoluchovski was the invitee in 1916.
Other renowned invitees have been Poincaré, Lorentz, Sommerfeld, Planck,
Debye, Born and others. Mie’s Wolfskehl lectures appeared in print in 1917
[Mie43, Mie44, Mie45]. After finishing the corrections in the proof sheets, Mie
commented:
I have now completed the last proof-sheet of my Einsteiniade and soon, the third and final
lecture will also be printed. I cannot help myself, I found it enormously interesting
reading.112
An important place for discussion and controversy was the regular meeting of
the ‘‘Gesellschaft deutscher Naturforscher und Ärzte’’. Mie and Einstein visited
the meetings regularly. They presented their own contributions or commented on
contributions from others during the discussion. A conflict between Mie and
Einstein appeared at the meeting in Vienna in 1913 (see Fig. 1.11). Einstein
presented an overview, ‘‘To the state of the art of the gravitational problem’’.
Einstein discussed his theory, developed together with Marcel Großmann, and also
109
D. Hilbert, Die Grundlagen der Physik. Erste Mitteilung. Vorgelegt in der Sitzung vom 20.
November 1915. in ‘‘Nachrichten von der Königlichen Gesellschaft der Wissenschaften zu
Göttingen. Mathematisch-physikalische Klasse’’, Heft 5, 395–407 (1915).
110
UAF, Nachlass Mie, E12/69, diary note from 13 February 1916.
111
Paul Wolfskehl (1856–1906) was an industrialist with mathematical interest. He bequeathed
100000 Mark to the Royal Academy of Sciences of Göttingen to be given to that person, who will
proof Fermat’s Theorem during the next 100 years. (Wolfskehl prize) The interest of the prize
money was used to organize the Wolfskehl lectures.
112
UAF, Nachlass Mie, E12/71, diary note from 8 December 1917.
36 W. Hergert
included attempts, especially Nordström’s theory, in his talk. The discussion was
led mainly by Mie and Einstein. Mie was angry that Einstein failed to mention his
theory. A controversy documented in the Physikalische Zeitschrift was the result.
In an exchange of letters from 1917 to 1919 both scientists discussed their different
points of view. Illy summarised the exchange of letters and discussion between
Mie and Einstein:
The case of Einstein and Mie underlines a rather sceptical thesis: there is no perfect
understanding of each other. Theories change in the hands of scientists, taking part in a
relay race called science with each member of the team leaving the fingerprints of his
commitments and beliefs on it. These fingerprints are often nonrational and incompre-
hensible to others. But this is the way science always grows...113
113
[105], p. 256.
1 Gustav Mie 37
1.4.1 Lectures
Gustav Mie was full professor for experimental physics and director of the insti-
tutes in Greifswald, Halle and Freiburg. The obligation which accompanied with
these positions was to cover teaching in experimental physics. A check of the
university calendars of the corresponding years demonstrates that Mie did not
regard teaching as an annoying obligation. On 2 November 1915, he wrote in his
diaries with reference to the beginning of the practical course of the students of
medicine and the difficult times during World War I:
In such a situation one is automatically kind and attentive to the young people and
immediately forgets to play the role of the schoolmaster, which is often, for me, so
inexpressibly distastefull. Yesterday, I took to being a teacher with great pleasure, and by
the way, there were also three ladies sprucing up the otherwise so martial audience.
...Then, I had my first lecture in this term ...and it was a great pleasure for me to tell the
young people, as an introduction to electricity, a little bit about my old friend, the world
ether to teach them respect.115
As an example, the canon of the courses that Mie lectured in Freiburg from
winter semester of 1925–1926 until the winter semester of 1934–1935 is given in
Table 1.3. The courses remained almost unchanged during the whole period.
Starting from the summer semester 1930 the half-day courses Advanced Practical
Course I and II, where Mie lectured the first part together with Kast (1896–1980),
replaced Physical Exercises for advanced Students. The workload from the lec-
tures was high and, apart from the large experimental physics lecture, also
included the work in practical courses.
Theoretical physics was lectured by Emil Cohn116 and Johann Georg Königs-
berger.117 Courses in theoretical physics were only rarely given by Mie.
114
For a summary see of the meeting see Weyl [106].
115
UAF, Nachlass Mie, E12/69, diary note from 2 November 1915.
116
Emil Cohn (1854–1944), 1879 Dr. phil. in Strassburg, assistant of A. Kundt, 1884
habilitation in theoretical physics, 1919 honorary professor in Rostock, 1920–1935 honorary
professor in Freiburg, 1938 together with Gans, Graetz, Jaffé, Kaufmann and other physicists of
Jewish descent withdrew out of DPG, 1939 emigrating to Switzerland.
117
J.G. Königsberger (1874–1946), 1904–1936 professor of physics at the University of
Freiburg, pioneer of semiconductor physics, coined 1911 together with Weiss the term
‘‘semiconductor’’ (sun of the mathematician Leo Königsberger).
38 W. Hergert
Table 1.3 Courses of Gustav Mie at the University of Freiburg from the WS 25/26 up to the WS
34/35
Lecture Amount
SS Experimental Physics I (Mechanics, Acoustics, Thermodynamics) 5 h/week
Mathematical Additions to Experimental Physics 2 h/week
Physical Exercises for Students of Natural Sciences 6 h/Wweek
Small Physical Practical Course for Physicians and Pharmacists 3 h/week
Physical Exercises for advanced students Daily
Guidance for independent work Daily
Physical Colloquium 2 h/week
WS Experimental Physics II (Electricity and Optics) 5 h/week
Mathematical Additions to Experimental Physics 2h/week
Physical Exercises for Students the Natural Sciences 6 h/week
Small Physical Practical Course for Physicians and Pharmacists 3 h/week
Physical Exercises for advanced students Daily
Guidance for independent work Daily
Physical Colloquium 2 h/week
see university calendars on-line in FREIMORE - FREIburger Multimedia Object REpository
http://freimore.uni-freiburg.de
118
Königsberger lectured in the same term Elements of the Quantum Theory.
119
[Bio10], p. 130.
1 Gustav Mie 39
1.4.2 Disciples
Gustav Mie wrote a series of popular scientific brochures and papers on actual
problems of physics for the general public. The brochures ‘‘Ions and Electrons’’
[Mie14] and ‘‘Molecules, Atoms, World Ether’’ [Mie18] are two examples. Mie
also published a popular description of Einstein’s theory in 1921 [Mie50]. His final
treatise ‘‘Foundations of Mechanics’’ dates back to 1950 [Mie86]. His book
entitled, ‘‘Textbook on Electricity and Magnetism. An Experimental Physics of the
World Ether for Physicists, Chemists an Electrotechnicians’’. [Mie29] is highly
120
See: Jahresverzeichnis der an den deutschen Universitäten erschienenen Schriften (1887–
1913), Jahresverzeichnis der an den deutschen Universitäten und Technischen Hochschulen
erschienenen Schriften (19-14-1925), Jahresverzeichnis der an den deutschen Universitäten und
Hochschulen erschienenen Schriften (1926–1936). Purly theoretical Ph.D thesis supervised by
Mie could not be found.
121
August Julius Herweg (1879–1936), studies in Würzburg, scholar of W. Wien, graduation
Würzburg 1905, habilitation Greifswald 1907, 1913 professor, 1914–1918 army service, 1921
extraordinary professor in Halle, from 1923, extraordinary professor at TH Hannover.
122
UAG, Acta of the Phil.Fak. of the Königl. Univ. Greifswald I 324 (Dekanat Mie), J.Nr. 1245.
123
Wilhelm Kast (1898–1980), studies in Halle, 1922 graduation, 1924 Freiburg, 1927
habilitation, 1933 appointment as extraordinary Professor, 1937 full professor and director of
institute in Halle, 1947 Bayer AG Leverkusen, 1954 honorary professor University Cologne,
1957 honorary professor university Freiburg.
40 W. Hergert
regarded, was printed in three editions 1910, 1941 and 1948, Mie was 80-years old
when he finished working on the final edition of his textbook.
Mie’s book was the first to contain a consequent classification of the field
quantities in quantities of intensity and quantities of quality. He also developed an
electromagnetic system of units, based on Volt, Coulomb, Centimeter and Second,
which is no longer relevant. The first edition starts from the physical phenomenon
rather than the mathematical approach.
Mie’s textbook, dedicated to Engelbert Arnold and Otto Lehmann in memory
of his Karlsruhe years of apprenticeship, appeared in the publishing house of
Ferdinand Enke, Stuttgart. Mie kept connections with this publisher for over 40
years. The exchange of letters between Enke and Mie highlights the relationship,
which was not entirely frictionless at times, between a scientist and the publisher,
who also had economic interests.124 Enke visited Mie in Halle at the beginning of
1924, to discuss details of a second edition after the success of the first edition.
They reached an agreement and settled on the deadline of April 1925. This
deadline was exceeded by more than 15 years in the end. Mie wrote in a letter to
Enke on 29 November 1925:
I started to rewrite my book to get a shorter version after your visit to Halle, but I cannot
do this without completely changing the character of the book. Therefore I finally decided
to follow the second possible way, namely to adapt it to a truly complete textbook which
includes all elements of the mathematical theory. ...I assume, that I will need nearly three
years to get it done. It will be a totally new book.125
Schäfer quoted in his recension of the second edition regarding the relevance of
Mie’s textbook as follows:
The first edition of Mie’s ‘‘Textbook of Electricity and Magnetism’’ appeared in 1910.
This was the first experimental textbook which took the introduction and realisation of the
field concept seriously into account, introducing the reader in the mindscape of Faraday
and Maxwell following a thoroughly thought out plan. Mie did not go back to ideas and
concepts of the theory of remote action. To the expert it is unquestionable that this book
has had a profound impact to the textbook literature after 1910, even if the authors did not
mention their source explicitly.126
The following remark of Schäfer about the style in Mie’s textbook seems to be
characteristic of Mie’s mode of practice: ‘‘The diction is sober, objective, clear,
and empathic. The formulation of the important theorems is presented in almost
pedantic accuracy to avoid any misunderstanding about the meaning of the
material: The learner cannot expect a better leader’’.127
124
vergl. UAF—Nachlass Mie, Bestand E12/30 Verlagsverhandlungen 1912–1946.
125
ibid.
126
Mie, Gustav: Lehrbuch der Elektrizität und des Magnetismus, Stuttgart 1941, Besprechung
durch Cl. Schaefer, Die Naturwissenschaften H. 17/18, S. 267f (1942).
127
ibid., S. 268.
1 Gustav Mie 41
Gustav Mie was strongly affected by the events of his times, in particular in his
work as a university teacher. His time in Greifswald coincides with World War I.
In these awkward times, he also officiated as rector of Greifswald University. The
time in Halle was accompanied by the revolution of 1918–1919, the Kapp Putsch
of 1920, the Central-German rebellion in March 1921 and the effects of hyper-
inflation in 1922–1923. During his last years at Freiburg University, Mie felt
the seizure of power of National Socialists and their increasing influence on the
University. Mie’s retirement was marked by the experience of the horrors of
World War II, as well as the beginning of the democratic reconstruction of
Western Germany. It is beyond the scope of this text to give a complete analysis
of the development of the academic, political and ideological opinions of Mie.
In addition, among other things, a detailed evaluation of the diaries would be
necessary.128
Gustav Mie was fully integrated into the network of theoretical physicists at
that time. He came into contact with Arnold Sommerfeld over his work on the
parallel wire system [Mie9]. At almost the same time, Mie began an exchange
of letters with W. Wien in 1900.129 Soon, Mie belonged to the community of
prominent German theoretical physicists, who met for skiing holidays, often in
Mittenwald. Mie was friends with W. Wien, A. Sommerfeld, M. von Laue and
P. Debye. The appreciation of his work was also expressed in the fact that his
colleagues recommended him for full professorships several times. He was sug-
gested for the post of the full professor in Frankfurt/Main by Sommerfeld
along with Laue and Debye in 1913, Wien suggested Mie for a full professor in
Tübingen as the second choice after Stark in 1916.
Mie’s extensive scientific exchange of letters as well as the numerous letters of
appraisal in appointment procedures130 show that Mie was strongly integrated into
the physics community of that time. S.L. Wolff places Mie in a network of
reactionary physicists during the time of the Weimar Republic [107, 108]. Gustav
Mie’s basic political orientation can be regarded as national conservative.131
Gustav Mie’s mainly conservative and somewhat nationalistic point of view can
be deduced, for example, from his publication ‘‘Werner Siemens as a Physicist’’
[Mie40] from 1916. At the end of the essay he characterised Siemens as a person.
Mie wrote:
If we consider Siemens as a person, as we can see him from the memoirs of his life, one
can conclude that he possess to a large degree the typical characteristics of a German. It is
128
UAF, inventory E012/68-E012/72 (copies), Nachlass Gustav Mie.
129
UAF, E012/48, Briefwechsel Mie-Wien, see also Archives Deutsches Museum Munich,
Nachlass Wilhelm Wien, Briefwechsel.
130
see UAF: E012/33, E012/34, Nachlass Gustav Mie, Berufungen.
131
Diary entries to current events prove Mie’s political point of view, UAF, inventory E012/68-
E012/72.
42 W. Hergert
definitely impossible to conceive this man to have grown up anywhere else but in
Germany. Also the peculiar connection of the desire for clear, scientific and objective
insight with a strong compulsion to influence the environment in an active and creative
way has to be counted as the typical properties which, in the end, makes the difference
between the German technology in general and that of other peoples. This unification of
energy and spirit of research in Germans, which allowed Siemens, with ingenious pecu-
liarity, to pace through his life as a hero, faced by the largest difficulties, which will
hopefully also contribute such that our native country will overcome the actual ordeal.132
World War I, involving all industrial nations, also gave rise to industrial con-
flicts, in which scientists were professionally mobilised [109]. At the beginning of
the war, scientists tried to justify the points of view of the respective belligerent
parties with invocations and corresponding replies. This development is charac-
terised by the term ‘‘War of the Spirit’’.133 After an article by Sir J.J. Thompson in
The Times on the 1 August 1914 on the 4 October 1914 the appeal ‘‘An die
Kulturwelt! Ein Aufruf’’ followed as a reply.
In this time also the Aufforderung, initiated by Wilhelm Wien appeared,
polemising against the ‘‘Ausländerei’’ and in particular the strong English influ-
ence. Finally it led to a politically motivated regimentation of the citations of
scientific work. Mie was one of the 16 signatories of the Aufforderung, 700 copies
of which were sent to all Universities in Germany and Austria.
To evaluate Mie’s attitude at the time of the National Socialism, one must
consider that he had already reached his retirement age in 1933 [110]. In the
conflict centred around ‘‘German Physics’’ the article ‘‘German and Jewish
Physics’’ by W. Menzel was published in the ‘‘Völkischer Beobachter’’ on 29
January 1936. Heisenberg answered in the same journal on the 28 February fol-
lowed by another reply by Stark. Thereupon, Heisenberg prepared a memorandum
to the Reichserziehungsminister Rust, which pointed out the difficulties of physics,
and in particular the importance of theoretical education and the cooperation
between theoretical and experimental physics. The memorandum was signed by 75
experimental and theoretical professors in the summer of 1936. Being asked by
Werner Heisenberg to sign the memorandum on 11 May 1936, Mie replied, that he
would do it with ‘‘great pleasure’’.134 In addition to the examples specified by
Heisenberg, Mie pointed out another example of fruitful collaboration between
theory and experiment:
A more impressive example, albeit from older times, ...is the discovery of the electro-
magnetic waves propagating freely in air by Heinrich Hertz. It is safe to state that without
Maxwell’s theory no one would ever have got the idea of performing such experiments.
H. Hertz considered them as an explicit verification of Maxwell’s theory. As I know
from a trustworthy source in Karlsruhe, any practical applications were far from his
132
[Mie40], p. 776.
133
After the title of the book ‘‘Krieg der Geister’’ from Hermann Kellermann, Dresden, 1915. In
this book corresponding calls and signatory lists are printed.
134
UAF, inventory E012/41, exchange of letters Mie-Heisenberg, 1a, letter Heisenberg at Mie of
11 1936, 1b, letter Mie at Heisenberg 15 May 1936, 2b text of the memorandum Leipzig, 30 May
1936.
1 Gustav Mie 43
considerations. What was more amazing is to see the world being reshaped by the elec-
tromagnetic waves.135
Single aspects of Mie’s scientific work have been illuminated in the past, often in
the context of different contributions to the history of physics, especially larger
historical studies about field theory and the theory of relativity [33, 98, 114]. Parts
of his extensive correspondence can be found in the edited correspondence of
Arnold Sommerfeld [115] as well as in the ‘‘Collected Papers of Albert Einstein’’
[116]. Gustav Mie’s archives136 are, to a large extent, unexplored.
In the 1970s an initiative existed to publish parts of scientific legacy, in par-
ticular the correspondence of Gustav Mie.137 This project was driven in particular
by Helmut Hönl138 in collaboration with Herbert Fröhlich139 and Ernst Plötze.
The Mie edition should have been published with the support of the Leopoldina.
The correspondence between Helmut Hönl and Heinz Bethge (1919–2001),
president of the Leopoldina in Halle at that time, documents this attempt. In the
135
ibid., 1b.
136
UAF, E0012, Nachlass Gustav Mie.
137
see UAF: C136, materials for the edition Gustav Mie, worked on by Rolf Steinmann 2001;
E14, Nachlass Helmut Hönl, worked on by David Geiger, 2003.
138
Helmut Hönl (1903–1981), Studies in Heidelberg, Göttingen and Munich, 1926 graduation
with A. Sommerfeld, 1933 habilitation in Stuttgart, 1940 extraordinary Professor in Erlangen,
1943 professor in Freiburg, 1971 emeritus.
139
Herbert Fröhlich (1905–1991), graduation 1930 with Sommerfeld, private lecturer in
Freiburg, as Jew forced to leave Germany, 1935 University of Bristol, 1948 University of
Liverpool.
44 W. Hergert
end, the project could not be realised. A detailed investigation of the life and
scientific work of Gustav Mie is thus still pending.
Acknowledgments I appreciate the support of the staff members of the university archives of
the Universities of Greifswald, Halle, Heidelberg and Freiburg, the archives of Karlsruhe Institute
of Technology as well as the archives of the National Academy of the Sciences Leopoldina
during my search. My special gratitude applies to Vincent Loke, Thomas Wriedt and Christian
Matyssek for the critical reading of the manuscript and Thomas Wriedt for suggestions and
references.
[Mie1] G. Mie, Zum Fundamentalsatz über die Existenz von Integralen partieller
Differentialgleichungen, Dissertationes Philosoph. Heidelbergensis 1892, Dresden, Druck
von B.G. Teubner, 1892
[Mie2] G. Mie, Fundamentalsatz über die Existenz von Integralen partieller Differentialgleichungen,
Zeitschrift für Mathematik 37, 151–170 (1892)
[Mie3] G. Mie, Fundamentalsatz über die Existenz von Integralen partieller Differentialgleichungen,
Zeitschrift für Mathematik 37, 193–211 (1892)
[Mie4] G. Mie Die Röntgen’sche Strahlung, S. V-IX, Sonderdruck aus: Beilage zur Nr. 7 der
Zeitschrift des Vereins für Finanzassistenten, 1896
[Mie5] G. Mie, Beweis der Integrirbarkeit gewöhnlicher Differentialgleichungen nach Peano,
Mathemat. Annalen 43, 553–568 (1893)
[Mie6] G. Mie, Entwurf einer allgemeinen Theorie der Energieübertragung, Sitzungsber. d.
K. Akad. d. Wiss. Wien, Math.-nat. Klasse 107 Abt. IIa, 1113–1180 (1898) (Habilitationsschrift)
[Mie7] G. Mie, Über mögliche Ätherbewegungen, Ann. der Physik 68, 129–134 (1899)
[Mie8] G. Mie und E. Arnold, Über den Kurzschluß der Spulen und die Commutation des
Stromes eines Gleichstromankers, Elektrotechn. Zeitschrift 20, 97–101, 136–138, 150–152
(1899)
[Mie9] G. Mie, Elektrische Wellen an zwei parallelen Drähten, Ann. der Physik 2, 201–249
(1900)
[Mie10] G. Mie, Die mechanische Erklärbarkeit der Naturerscheinungen: Maxwell-Helmholtz-
Hertz, Nat. Ver. Verh. (Karlsruhe) 13, 402–420 (1900)
[Mie11] G. Mie, Ein Beispiel für Poyntings Theorem, Zeitschrift für physikal. Chemie 34,
522–526 (1900)
[Mie12] G. Mie, Über die Bewegung eines als flüssig angenommenen Äthers, Physikal.
Zeitschrift 2, 319–325 (1901)
[Mie13] G. Mie, Zur kinetischen Gastheorie der einatomigen Körper, Ann. der Physik 11,
657–697 (1903)
[Mie14] G. Mie, Die neueren Forschungen über Ionen und Elektronen, Sonderausgabe aus der
Sammlung elektrotechnischer Vorträge Bd. 4, Enke Stuttgart, 1903, 40 S. (2. Aufl. 1906)
[Mie15] G. Mie, Der elektrische Strom in ionisierter Luft in einem ebenen Kondensator, Ann. der
Physik 13, 857–889 (1904)
[Mie16] G. Mie, Über eine Methode, das spezifische Gewicht sehr verdünnter Lösungen zu
bestimmen, Festschrift Ludwig Boltzmann gewidmet zum 60. Geburtstag, Ambrosisus
Barth,Leipzig 1904, 326–332
[Mie17] G. Mie, Das Problem der Wärmeleitung in einem verseilten elektrischen Kabel,
Neu-Vorpommern u. Rügen, Nat.-Ver. Mitteil. 36, 155–188 (1904)
[Mie18] G. Mie, Moleküle, Atome, Weltäther, Verlag von B.G. Teubner, Leipzig, 1. Auflage
1904, Aus Natur und Geisteswelt Sammlung wissenschaftlich-gemeinverständlicher
Darstellungen 58. Bändchen (2. Aufl. 1907, 3. Aufl. 1911, 4. Aufl. 1919)
1 Gustav Mie 45
[Mie19] G. Mie, Das Problem der Wärmeleitung in einem verseilten elektrischen Kabel,
Elektrotechn. Zeitschrift 137, 137–143 (1905)
[Mie20] G. Mie, Über die Kurzschlußstromkurve eines Gleichstromankers, Zeitschrift Math.
Phys. 53, 37–60 (1906)
[Mie21] G. Mie, Experimentelle Darstellung elektrischer Kraftlinien, Zeitschrift für d. phys. u.
chem. Unterricht 19, 154–156 (1906)
[Mie22] G. Mie, Erwiderung auf Riebesells Abhandlung: Kommutation des Stromes in
Gleichstromgeneratoren, Zeitschrift Math. Phys. 55, 143–146 (1907)
[Mie23] G. Mie, Die optischen Eigenschaften kolloidaler Goldlösungen, Zeitschrift für Chemie
und Industrie der Kolloide 2, H. 5, 129–133 (1907)
[Mie24] G. Mie, Die optischen Eigenschaften kolloidaler Goldlösungen, Berichte der dt. phys.
Ges. 5, 492–500 (1907)
[Mie25] G. Mie, Beiträge zur Optik trüber Medien, speziell kolloidaler Metallösungen, Ann. der
Physik 25, 377–445 (1908)
[Mie26] G. Mie, Sättigungsstrom und Stromkurve einer schlecht leitenden Flüssigkeit, Ann. der
Physik 25, 597–614 (1908)
[Mie27] G. Mie, Eine bequeme Methode zur Erzeugung ganz schwach gedämpfter elektrischer
Schwingungen von kleiner Wellenlänge, Verh. der dt. phys. Ges. 12, 860–866 (1910)
[Mie28] G. Mie, Hydratisierung und Molekularwärme der Ionen in sehr verdünnten wäss rigen
Lösungen, Ann. der Physik 33, 381–399 (1910)
[Mie29] G. Mie, Lehrbuch der Elektrizität und des Magnetismus. Eine Experimentalphysik des
Weltäthers für Physiker, Chemiker und Elektrotechniker, 736 S., Stuttgart 1910, 2. vollständig
umgearbeitete Aufl. 1941, 3. Aufl. 1948, span. Ausgabe Barcelona 1944
[Mie30] G. Mie, Antwort auf die Bemerkung des Hrn. Seibt zu der Arbeit des Hrn. K. Settnik:
Entstehung von sehr wenig gedämpften Wellen usw., Ann. der Physik 36, 207–208 (1911)
[Mie31] G. Mie, Die Materie, 32 S., Verlag von Ferdinand Enke, Stuttgart 1912 (Vortrag
gehalten am 27. Januar 1912 (Kaisers Geburtstag) in der Aula der Universität Greifswald)
[Mie32] G. Mie, Grundlagen einer Theorie der Materie I, Ann. der Physik 37, 511–534 (1912)
[Mie33] G. Mie, Grundlagen einer Theorie der Materie II, Ann. der Physik 39, 1–40 (1912)
[Mie34] G. Mie, Bemerkungen zu der Arbeit des Herrn W. Sorkau über Turbulenzreibung,
Physikal. Zeitschrift 14, 93–95 (1913)
[Mie35] G. Mie, Bemerkungen zu der Arbeit des Herren Gotthelf Leimbach: Energieaufnahme
elektrischer Sender von kleiner Wellenlänge, Physikal. Zeitschrift 14, 723–725 (1913)
[Mie36] G. Mie, Grundlagen einer Theorie der Materie III, Ann. der Physik 40, 1–66 (1913)
[Mie37] G. Mie, Bemerkungen zu Einsteins Gravitationstheorie I, Physikal. Zeitschrift 15,
115–122 (1914)
[Mie38] G. Mie, Bemerkungen zu Einsteins Gravitationstheorie II, Physikal. Zeitschrift 15,
169–176 (1914)
[Mie39] G. Mie, Prizip von der Relativität des Gravitationspotentials, Festschrift für J. Elster und
H. Geitel (1915), S. 251–268
[Mie40] G. Mie, Werner Siemens als Physiker, Die Naturwissenschaften 4, 771–775 (1916)
[Mie41] G. Mie, Naturgesetz und Geist, Deutsche Revue 41, 150–163 (1916)
[Mie42] Bilder aus Greifswalds Vergangenheit, Gewidmet von der Universität Greifswald ihren
Kriegsstudenten (Vorwort G. Mie), Verlag Brucken & Co. Greifswald, 1917
[Mie43] G. Mie, Die Einsteinsche Gravitationstheorie und das Problem der Materie, Physikal.
Zeitschrift 18, 551–555 (1917)
[Mie44] G. Mie, Die Einsteinsche Gravitationstheorie und das Problem der Materie, Physikal.
Zeitschrift 18, 574–580 (1917)
[Mie45] G. Mie, Die Einsteinsche Gravitationstheorie und das Problem der Materie, Physikal.
Zeitschrift 18, 596–602 (1917)
[Mie46] G. Mie, Einführung eines vernunftgemäßen Koordinatensystems in die Einsteinsche
Gravitationstheorie und das Gravitationsfeld einer schweren Kugel, Ann. der Physik 62,
46–74 (1920)
[Mie47] G. Mie, Die Gesetzmäßigkeit des Naturgeschehens, Bertelsmann, Gütersloh, 1920
46 W. Hergert
[Mie48] G. Mie, Das elektrische Feld eines um ein Gravitationszentrum rotierenden geladenen
Partikelchens, Physikal. Zeitschrift 21, 651–658 (1920)
[Mie49] G. Mie, Über die Abklingung der Lichtemission eines Atoms, Ann. der Physik 66,
237–260 (1921)
[Mie50] Die Einsteinsche Gravitationstheorie. Versuch einer allgemeinverständlichen Darstel-
lung der Theorie, 4+69 S., Leipzig, Hirzel 1921 (2. verbesserte Aufl. 1923)
[Mie51] Die Relativitätstheorie von Einstein. Zeitschrift des Vereins Deutscher Ingenieure 65,
26. Februar, S. 236 (1921)
[Mie52] Die Einsteinsche Gravitationstheorie, Deutsche Rundschau, 187, 171–184, 310–342
(1921)
[Mie53] G. Mie und J. Herweg, Die Zahl der von sehr schwachen Röntgenstrahlen ausgelösten
Elektronen, Ann. der Physik 68, 120–126 (1922)
[Mie54] G. Mie, Träge und schwere Masse, Ann. der Physik 69, 1–51 (1922)
[Mie55] G. Mie, Thèorie Einsteinienne de la Gravitation. Essai de vulgarisat. de la théorie,
Hermann, Paris 1922
[Mie56] G. Mie, Echte optische Resonanz bei Röntgenstrahlen, Zeitschrift für Physik 15, 56–57
(1923)
[Mie57] G. Mie, Echte optische Resonanz bei Röntgenstrahlen, Zeitschrift für Physik 18,
105–108 (1923)
[Mie58] G. Mie, Das elektrische Feld eines schweren elektrisch geladenen Kügelchens, das um
ein Gravitationszentrum kreist, Ann. der Physik 70, 489–558 (1923)
[Mie59] G. Mie, Abklingzeit und Verweilzeit angeregter Atome, Ann. der Physik 73,
195–208 (1924)
[Mie60] G. Mie, Theorie der Bremsstrahlung und der Comptonschen Streustrahlung (Vortrag auf
dem III. Dt. Physikertag, Danzig, 10.-16.9.1925) Physikal. Zeitschrift 26, 665–669 (1925)
[Mie61] G. Mie, Bremsstrahlung und Comptonsche Streustrahlung, Zeitschrift für Physik 33,
33–41 (1925)
[Mie62] G. Mie, Theorie der Bremsstrahlung und der Comptonschen Streustrahlung, Physikal.
Zeitschrift, 26, 665–669 (1925)
[Mie63] G. Mie, Das Problem der Materie und die Relativitätstheorie, Scientia (Mailand) 37,
149–156 (1925)
[Mie64] G. Mie, Das Problem der Materie und die Relativitätstheorie, Scientia (Mailand) 37,
225–234 (1925)
[Mie65] G. Mie, Das Problem der Materie (Öffentliche Antrittsrede, gehalten am 26.
Januar1925), Speyer & Kärner, Freiburg i. Br., 1925
[Mie66] G. Mie, Die Grundlagen der Quantentheorie (Vortrag in der Freiburger Naturfor-
schenden Gesellschaft), 38 S., Speyer & Kärner, Freiburg i. Br., 1925
[Mie67] G. Mie, Ein Linienspektrum bei Wellenlängen von mehreren Dezimetern (nach
Messungen von E. Frankenberger), Physikal. Zeitschrift 27, 792–795 (1926)
[Mie68] G. Mie, H. Staudinger, R. Signer, H. Johner und J. Hengstenberg, Der polymere
Formaldehyd, ein Modell der Zellulose, Zeitschrift für physikal. Chemie 126, 425–448 (1927)
[Mie69] G. Mie und J. Hengstenberg, Zur Arbeit: E. Ott, Röntgenometrische Untersuchungen an
hochpolymeren organischen Substanzen, Helvet. chim. Acta 11, 1052 (1928)
[Mie70] G. Mie, Probleme der Quantenelektrik, Ann. der Physik 85, 711–729 (1928)
[Mie71] G. Mie, Einleitung in die Physik, in: A. Eucken, O. Lummer, E. Waetzmann (Hrsg.) Müller-
Pouillets Lehrbuch der Physik, 11. Auflage, Erster Band, Erster Teil, 1–70, Friedr.Vieweg und
Sohn, Braunschweig 1929
[Mie72] G. Mie, Quantentheorie und Elektrodynamik, Verh. Dt. Phys. Ges. 12, 46 (1931)
[Mie73] G. Mie und E. Frankenberger, Über Präzisionsmessungen von elektrischen Brechungs-
exponenten nach der 2. Drudeschen Methode, Festschrift zur 10-j. Tätigkeit des
magnetischenLabors. Moskau (1931) S. 13–20
[Mie74] G. Mie, Elektrodynamik, in: W. Wien und F. Harms (Hrsg.) Handbuch der
Experimentalphysik Bd. 11, Teil 1, XII u. 502 S., Akad. Verlagsges. Leipzig 1932
1 Gustav Mie 47
[Mie75] G. Mie, Elektron, Bd. 2 S. 500–512, Materie, Bd. 6 S. 786–790, Weltäther, Bd.10 S. 649–655,
in: R. Dittler, G. Joos, E. Korschelt, G. Link, F. Oltmanns, K. Schaum (Hrsg.) Handbuch der
Naturwissenschaften, 2. Auflage, Jena, Gustav Fischer 1932–1935
[Mie76] G. Mie, Naturwissenschaft und Theologie, Akad. Verlagsgesellschaft, Leipzig 1932,
40S., Sonderdruck aus der Zeitschrift Die christliche Welt, Nr.22 und 23 (Leopold Klotz
Verlag, Gotha)
[Mie77] G. Mie, Geometrie der Spinoren, Ann. der Physik 17, 465–500 (1933)
[Mie78] G. Mie, Die geistige Struktur der Physik, Studien des apologetischen Seminars Bd. 38,
36 S., Bertelsmann, Gütersloh 1934
[Mie79] G. Mie, Zur Erinnerung an Ch.A. Coulomb und A.M. Ampère, Die Naturwissenschaften
24, 369 (1936)
[Mie80] G. Mie, Julius Herweg , Zeitschrift für techn. Physik 17, 321–323 (1936)
[Mie81] G. Mie, Raum und Zeit in der Physik, Festskrift til Professor Dr. Anathon Aall (1937)
[Mie82] G. Mie, Die Denkweise der Physik und ihr Einfluß auf die geistige Stellung des heutigen
Menschen, 37 S., Stuttgart 1937
[Mie83] G. Mie, Der Trägheitsfaktor, Zeitschrift für d. phys. u. chem. Unterricht 51, 90 (1938)
[Mie84] G. Mie, Die göttliche Ordnung in der Natur, 3 Aufsätze, 31 S., Tübingen 1946
[Mie85] G. Mie, Aus meinem Leben, Zeitenwende 19, 733–743 (1948)
[Mie86] G. Mie, Die Grundlagen der Mechanik, VI u. 80 S., Ferdinand Enke Verlag, Stuttgart 1950
[Mie87] G. Mie, Gott in der Natur, in: Wunder der Gnade Gottes in unserem Leben - Gesammelte
Zeugnisse, Verlagsbuchhandlung Bethel, Hamburg 1952, S. 94–97
[Bio19] Berühmter unbekannter Rostocker: Gustav Mie-in der Heimat vergessen, an der Uni
Freiburg und in Nachschlagwerken geehrt, Mecklenburg-Magazin: Regionalbeilage der
Schweriner Volkszeitung, Band 44, 22 (2005)
[Bio20] U. Kreibig, Optische Eigenschaften von Nanopartikeln: 100 Jahre Mie-Theorie, Phys. in
unserer Zeit 39 H. 6 (2008), S. 281–287
[Bio21] M.I. Mishchenko und L.D. Travis, Gustav Mie and the evolving discipline of
electromagnetic scattering by particles, Bull. Am. Met. Soc., 89, 1853–1861 (2008)
[Bio22] M.I. Mishchenko, Gustav Mie and the fundamental concept of electromagnetic scattering
by particles: A perspective, J. Quant. Spec. and Rad. Trans. 110, 1210–1222 (2009)
References
26. D. Alvermann, K.H. Spiess, Universität und Gesellschaft–Festschrift zur 550-Jahrfeier der
Universität Greifswald, Band I: Die Geschichte der Fakultäten im 19. Und 20. Jahrhundert
(Hinstorff, Rostock, 2006).
27. Festschrift zur 500-Jahrfeier der Universität Greifswald, Band II (Universität Greifswald,
Greifswald, 1956).
28. F. Schubel, Universität Greifswald (Wolfgang Weidlich, Frankfurt am Main, 1960)
29. A. Kleinert, Johannes Stark: Erinnerungen eines Deutschen Naturforschers (Bionomica-
Verlag, Mannheim, 1987)
30. W. Hergert, Martin-Luther-Universität Wissenschaftliche Beiträge 1990/33 (O32) (1990).
31. W. Hergert, scientia halensis (3), 37 (1995).
32. C. Stephan, Die Naturforschende Gesellschaft in Halle: eine lokale Gesellschaft und das
Engagement ihrer Mitglieder insbesondere auf den Gebieten der Physik und Chemie. Ph.D.
Thesis, Universität Marburg, Marburg (2006).
33. G. Kohl, Relativität in der Schwebe: Die Rolle von Gustav Mie. No. 209 in Preprint Series
(Max Planck Institut for the History of Science, Berlin, 2002).
34. W. Hergert, Scientia halensis (3), 13 (2005).
35. R.K. Merton, The Sociology of Science: Theoretical and Empirical Investigations (The
University of Chicago Press, Chicago and London, 1973)
36. S. Stigler, Trans. New York Acad. Sci. 39, 147 (1980)
37. J. Jackson, Am. J. Phys. 76, 704 (2008)
38. G. Stigler, Transactions of the New York Academy of Sciences 39, 143 (1980)
39. D. Ballentyne, D. Lovett, A Dictionary of Named Effects and Laws in Chemistry, Physics
and Mathematics (Chapmann and Hall, Londan, 1980)
40. M. Jammer, Der Begriff der Masse in der Physik (Wissenschaftliche Buchgesellschaft,
Darmstadt, 1981)
41. D. Plappert, Praxis der Naturwissenschaften-Physik in der Schule 57, 22 (2008)
42. H. Geiger, K. Scheel (eds.), Handbuch der Physik, Band X: Thermische Eigenschaften der
Stoffe (Springer, Heidelberg, 1926)
43. E. Grüneisen, Annalen der Physik 39, 257 (1912)
44. R. Furth, Proc. R. Soc. Lond. 183, 87 (1944)
45. J. Jones, Proc. R. Soc London A 106, 441 (1924)
46. J. Jones, proc. R. Soc. London A 106, 463 (1924)
47. J. Jones, Proc. R. Soc. London A 106, 709 (1924)
48. S. Erkoç, R. Sever, Phys. Rev. D 33, 588 (1986)
49. S. Erkoç, R. Sever, Phys. Rev. D 30, 2117 (1984)
50. E. Kasap, B. Gönül, M. Simsek, Chem. Phys. Lett. 172, 499 (1990)
51. R. Sever, C. Tezcan, M. Bucurgat, Ö. Yesiltas, J. Math, Chem. 43, 749 (2007).
52. R.S.S. Ikhdair, J. Mol. Struc. (Theochem) 855, 13 (2008)
53. S.Ö.A. Erkoç, E. Iltan, Chem. Phys. Lett. 135, 582 (1987).
54. H. Ledbetter, M. Lei, R. Rao, Physica B 159, 265 (1989)
55. K. Jayachandran, C. Menon, J. Phys. Chem. Sol. 60, 267 (1999)
56. T. Barakat, O. Al-Dossary, A. Alharbi, Int. J. Nanosci. 6, 461 (2007)
57. S. Flügge (ed.), Handbuch der Physik, Astrophysik III, Das Sonnensystem (Springer, Berlin,
1959)
58. E. Lecher, Annalen der Physik 41, 850 (1890)
59. A. Sommerfeld, Ann. der Physik und Chemie 67, 233 (1899)
60. A. Sommerfeld, Nuovo Cimento 9, 151 (1943)
61. D. Hondros, Über elektromagnetische Drahtwellen. Ph.D. Thesis, Ludwig-Maximilians-
Universität München (1909).
62. D. Hondros, Annalen der Physik 335, 905 (1909)
63. K. Jäger, Die Entwicklung der Starkstromtechnik in Deutschland, Teil 2: Von 1890 bis 1920
(VDE-verlag Berlin, 1991).
64. M. Cardona, W. Marx, Physik Journal 3(11), 27 (2004)
65. W. Marx, Phys. in unserer Zeit 38(1), 34 (2007)
50 W. Hergert
66. K. Degen, Beiträge zur Kenntnis kolloidaler Metalllösungen (Julius Abel, Greifswald,
1903)
67. E. Lischner, Über die elliptische Polarisation des Lichtes bei der Reflexion an Lösungen
anomal dispergierender Substanzen (Greifswald, 1903).
68. E. Lischner, Ann. der Physik 317, 964 (1903)
69. H. Siedentopf, R. Zsigmondy, Ann. der Physik Journal 10, 1 (1903)
70. R. Zsigmondy, Zur Erkenntnis der Kolloide (Verlag von Gustav Fischer, Jena, 1905)
71. M. Planck, Ann. d. Physik 306, 69 (1900)
72. M. Planck, Ann. der Physik 311, 818 (1901)
73. J.M. Garnett, Trans. Roy. Soc. CCIII A-370, 385 (1904).
74. W. Steubing, Über die optischen Eigenschaften kolloidaler Goldlösungen (Johann
Ambrosius Barth, Leipzig, 1908)
75. M. Kerker, Appl. Optics 30, 4699 (1991)
76. N. Logan, J. Opt. Soc. Am. 52, 342 (1962)
77. N. Logan, Proceedings of the IEEE 53, 773 (1965)
78. A. Clebsch, J. für Mathematik 61, 195 (1863)
79. L. Rayleigh, Phil. Mag. 41, 447 (1871)
80. L. Rayleigh, Proc. Math. Soc. (London) 4, 253 (1872)
81. L. Rayleigh, Phil. Mag. 12, 81 (1881)
82. H. Lamb, Proc. Math. Soc. (London) 13, 51 (1881)
83. L. Lorenz, Vidensk. Selsk. Skrifter 6, 1 (1890)
84. P. Debye, Ann. d. Physik 30, 57 (1909)
85. J. Thomson, Recent Researches in Electricity and Magnetism (University Press, Oxford, 1893)
86. G.W. Walker, Quart. J. Math. 30, 204 (1899)
87. G. Walker, Quart. J. Math. 31, 36 (1900)
88. J.W. Nicholson, Proc. Math. Soc. (London) 9, 67 (1910)
89. H. Bateman, The Mathematical Analysis of Electrical and Optical Wave Motion on the
Basis of Maxwell’s Equations (University Press, Cambridge, 1915)
90. A.J. Proudman, A. Doodson, G. Kennedy, Phil. Trans. Roy. Soc. 217, 279 (1917)
91. T.J. Bromwich, Phil. Trans. Roy. Soc. 220, 189 (1920)
92. G.N. Watson, Phil. Trans. Roy. Soc. 220, 189 (1920)
93. L. Lorenz, Oeuvres Sciéntifiques de L. Lorenz (Librairie Lehmann, Copenhagen, 1896)
94. H. Kragh, Applied Optics 30, 4688 (1991)
95. M. Pihl, Der Physiker L.V. Lorenz-Eine kritische Untersuchung (Einar Munksgaard,
Kopenhagen, 1939).
96. L. Lorenz, Wied. Ann. der Phys. 11, 70 (1880)
97. P. Debye, Zeitschrift für Physik 9, 775 (1908)
98. V. Vizgin, Unified Field Theories (Birkhäuser Verlag, Basel, Boston, Berlin, 1994)
99. J. Renn (ed.), The Genesis of General Relativity, vol. 4 (Gravitation in the Twilight of
Classical Physics, The Promise of Mathematics (Springer Verlag, Dordrecht, 2007)
100. J. Renn (ed.), The Genesis of General Relativity Vol. 3: Gravitation in the Twilight of Classical
Physics: Between Mechanics, Field Theory and Astronomy (Springer, Dordrecht, 2007).
101. L. Corry, Stud. Hist. Phil. Mod. Phys. 30, 159 (1999)
102. M. Born, L. Infeld, Proc. R. Soc. A 144, 425 (1934)
103. H. Weyl, Raum - Zeit - Materie, 5th edn. (Springer, Dordrecht, 1923)
104. E. Scholz (ed.), Hermann Weyl’s Raum-Zeit-Materie and a General Introduction to His
Scinetific Work (Birkhäuser Publishing Ltd, Switzerland, 2001)
105. J. Illy, in Einstein Studies Vol. 3, Studies in the History of General Relativity, ed. by J.
Eisenstaedt, A. Knox (Birkhäuser Publishing Ltd, Switzerland 1992), pp. 244–259.
106. H. Weyl, Jahresbericht der Dt. Mathematiker-Vereinigung 31, 51 (1922)
107. S. Wolff, Ber. Wissenschaftsgesch. 31, 372 (2008)
108. S. Wolff, Acta Historica Leopoldina 48, 41 (2007)
109. R. MacLeod, J. War Culture Studies 2, 37 (2009)
110. W. Heisenberg, Deutsche und Jüdische Physik (Piper Verlag, München, 1992)
1 Gustav Mie 51
Thomas Wriedt
Abstract In optical particle characterisation and aerosol science today light scatter-
ing simulations are regarded as an indispensable tool to develop new particle charac-
terisation techniques or in solving inverse light scattering problems. Mie scattering
and related computational methods have evolved rapidly during the past decade such
that scattering computations for spherical scatterers a few order of magnitudes larger,
than the incident wavelength can be easily performed. This significant progress has
resulted from rapid advances in computational algorithms developed in this field and
from improved computer hardware. In this chapter the history and a review of the
recent progress of Mie scattering and Mie-related light scattering theories and avail-
able computational programs is presented. We will focus on Mie scattering theories
but as there is much overlap to related scattering theories they will also be mentioned
where appropriate. Short outlines of the various methods are given. This review is of
course biased by my interest in optical particle characterisation and my daily reading.
2.1 Introduction
When Gustav Mie wrote his classic paper on light scattering by dielectric absorb-
ing spherical particles in 1908 he was interested in explaining the colourful effects
connected with colloidal Gold solutions. Nowadays, the interest in Mie’s theory is
much broader. Interests range from areas in physics problems involving interstellar
dust, near-field optics and plasmonics to engineering subjects like optical particle
characterisation. Mie theory is still being applied in many areas because scattering
particles or objects are often homogeneous isotropic spheres or can be approximated
in such a way that Mie’s theory is applicable.
T. Wriedt (B)
Institut für Werkstofftechnik, Badgasteiner Str., 328359 Bremen, Germany
e-mail: [email protected]
On the 100th anniversary of Mie’s theory a number of review papers, special papers
and conference proceedings were published [1–10] commemorating his 1908th
paper.
In this chapter I would like to review the history and the state of the art in Mie
scattering as it developed over the previous decades. I will give a short description
of the various extensions to Mie scattering available and provide information about
computation programs making reference to review papers where available. For more
profound reviews the interested reader is referred to the cited review articles.
In 1908, Gustav Mie published his famous paper on simulation of the colour effects
connected with colloidal Gold particles [19]. In this paper he gave a first outline of
how to compute light scattering by small spherical particles using Maxwell’s electro-
magnetic theory. With his first computations he managed to explain the colour of gold
colloids changing with diameter of the Gold spheres, which was later interpreted in
terms of surface plasmon resonances. First computations of scattering diagrams for
larger spheres of diameters up to 3.2λ are presented by Richard Gans (1880–1954)
[20] and Hans Blumer [21] in 1925. The early study of scattering by a sphere is
traced by Nelson A. Logan [22], who comments on Blumer’s results, that he missed
the regular undulations in the scattering diagram because of numerical mistakes.
This paper is interesting reading for all interested in the history of light scattering.
2 Mie Theory: A Review 55
who used the Debye potential in their derivation. Hendrik Christoffel van de Hulst
(1918–2000) [46] and Bohren and Hufman [47] follow the Stratton approach.
An up-to-date version of the derivation of Mie’s theories has been published in
Appendix H of the recent book by Le Ru and Etchegin [48] and in a book chapter
by Enguehard [49]. Both provide an outline on how to solve the electromagnetic
scattering problem in the case of a spherical particle using Mie’s theory which give
the most useful mathematical expressions of Mie theory and its derivatives and also
information about implementation.
Originally, the Mie theory was restricted to a nonabsorbing ambient medium although
some real-world applications require that the absorbing surrounding media be
accounted for in the theory. This is a topic in colloid science and computer graphics.
Apparently one of the first derivation of the Lorenz-Mie theory for a scattering
sphere immersed in an absorbing host medium is presented by Mundy et al. [66]
and by Bohren and Gilra [67]. Similar extension to the Mie theory was studied by
Quinten and Rostalski [68], Sudiarta [69] and Frisvad et al. [70]. It was found that
the absorption of the ambient medium can alter the scattering efficiency and the
scattering pattern of a sphere. The effect on the absorption efficiency is much less.
To extend such theories for light scattering by coated sphere immersed in an absorbing
medium is a small step [71].
As it is easy to consider spherical scatterers there are many extensions of the Mie
theory covering different aspects. A theory for a coated dielectric sphere was first
published by Aden and Kerker [72]. The Fortran code BART by Arturo Quirantes
[73] is based on the Aden-Kerker theory to calculate light-scattering properties for
coated spherical particles. In the program polydispersity is included for either core,
coating or entire particle.
A basic Fortan (BHCOAT) code is printed in the appendix of the book by Bohren
and Huffman [47]. An advanced algorithm is given by Toon and Ackerman [74].
An algorithm for a sphere having two coatings has been presented by Kaiser and
Schweiger [75]. These theories have also been extended to spherical particles con-
sisting of multiple layers by Li Kai [76]. Another algorithm for plane wave and
shaped beam scattering by a multilayered sphere has been published by Wu et
al. [77]. Even in recent times, improved algorithms have been published on this
subject [78].
applied to the surface of the sphere in angular functions identical to those used in the
spherical harmonic expansion. The approximation is that the particle be smooth and
not deviate far from sphericality.
With growing interest in magnetic nanostructures there is also interest in Mie scatter-
ing of magnetic spheres. A Mie theory allowing for magnetic media where the media
properties includes a nonzero permeability μ has been developed early by Kerker
et al. [81, 82]
A Fortran code for spherical scatterers with both a complex permittivity and
a complex permeability is listed in a report by Milham [83]. Tarento et al. [84]
considered Mie scattering of magnetic spheres extending the classical Mie scattering
approach to a media where the dielectric constant is no more a real number but a tensor
with a gyrotropic form. A usable code for scattering by a sphere having a different
magnetic permeability than the surrounding is included in the Matlab program by
Christian Mätzler [85].
During the previous decades, much attention has been focused on the light scatter-
ing interaction between chiral and anisotropic particles and electromagnetic waves,
as a result of numerous applications in the electromagnetic scattering and antenna
theory. The historical background, and a general description, of the subject of elec-
tromagnetic chirality and its applications can be found in the books by Lakhtakia
and Varadan [86] and Weiglhofer and Lakhtakia [87]. A isotropic chiral medium or
an optical active medium is rotationally symmetric but not mirror symmetric. Only
circularly polarised plane waves can propagate in the chiral medium without a change
in their state of polarisation.
A scattering sphere can also be chiral. The extension of Mie’s theory to an optically
active sphere was published by Bohren [88] and later extended to a sphere with a
chiral shell [89]. Bohren devised a transformation to decompose the problem such
that he could consider two independent modes of propagation in the chiral medium,
namely left- and right-circularly polarised waves. In the limit of no chirality the
solution is identical to the Mie solution.
Theories and programs for such types of scatterers have been published by Bohren
in his thesis for a chiral sphere [90].
Using the Bohren decomposition Hinders and Rhodes [91] investigate the problem
of scattering by chiral spheres embedded in a chiral host medium.
As colour pigments and crystals are often anisotropic there is some interest to
extend Mie’s theory to such kind of scattering particles. Stout et al. [92] established
2 Mie Theory: A Review 59
With a nanosized nobel metal particle of size lower than about 20 nm, various modi-
fications, extensions and corrections to Mie’s original theory are needed to take into
account that “sharp” boundary conditions do not hold in the nanoscale [100]. In his
recent survey paper [101] Kreibig lists among others the following supplementary
models to the Mie theory, i.e. non plane-wave incident field, non step-like boundary
condition and a particle size-dependent dielectric function. Applying these exten-
sions help to explain measured absorption spectra of Ag nanoparticles and plasmon
polaritons.
60 T. Wriedt
for the coefficients in the multipolar expansion of the incident field that results from
a beam passing through a high numerical aperture lens system.
With the recent rise of plasmonics there seems to be an increasing interest in com-
puting the near field or the internal field by scattering particles. There is especially
interest in morphological resonances and in plasmon resonances. There are not many
programs available which focus on near field computation. Most programs consider
the far field approximation of the radiating spherical vector wave functions to com-
pute scattering patterns in the far field. The T-Matrix programs on the disk accom-
panying the book by Barber and Hill [116] allow for the simulation of the internal
and external near field intensity distribution by a scattering sphere.
Near field and internal field computations of a spherical particle in a Gaussian
laser beam can be done using the Windows program GLMT Champ Internes by Loic
Mees [117].
To compute near fields by a number of scattering spheres you can use the extension
of Yu-lin Xu’s GMM program by Moritz Ringer. He extended GMM while working
on his PhD thesis [118]. His Fortran program GMM-Field allows for the computa-
tion of the near field. GMM-Dip gives the near field with a dipole as the incident
field. Apparently Giovanni Pellegrini did a similar extension of the GMM program
in this PhD thesis [119]. A paper on parallising multiple scattering and near field
computation by coated spheres was recently published by Boyde et al. [120].
The LightScatPro Matlab program by Sylvain Lecler used for his thesis [121] is
also suitable for near field computation for light scattering by a number of spheres.
Basically, in the Mie theory, transverse wave modes are dominant. However, there
is the question about what the theory would look like if there were also longitudinal
modes created inside the scattering sphere.
Ruppin [122] extended Mie theory to the case of small metal spheres whose
material sustains the propagation of longitudinal waves. This theory is also outlined
in the book by Borghese et al. [123]. Ruppin demonstrated that a slight shift of the
surface plasmon peak towards higher frequencies would occur. According to Quinten
in his recent book [102], all effects connected with longitudal modes have not yet been
identified experimentally. Travis and Guck [124] recently revisited the Mie theory
and also include longitudinal vector spherical wavefunctions in the expansion of the
internal field. The authors argue that in the optical wavelength range the longitudinal
wavenumber is almost purely imaginary such that for particles larger than about 2 nm
62 T. Wriedt
the longitudinal plasma modes are evanescent and the normal Mie approximation is
acceptable.
2.16 Parallelisation
Not all extensions of the Mie theory can be extensively covered in this chapter.
Therefore, in this section some interesting topics not treated above will be briefly
mentioned.
Travis and Guck revised the Mie theory [124] using a modern T-matrix formalism
and discuss inclusion of longitudinal components of the dielectric permittivity tensor.
Tagviashvili [141] considered the classical Mie theory for electromagnetic radia-
tion scattering by homogeneous spherical particles in the dielectric constant (ε) → 0
limit and TE field resonances in the visible spectrum are demonstrated.
With Mie scattering morphological resonances show up. This can be used to deter-
mine the diameter of a spherical droplet to a high precision. Ward et al. [142] used
a white light LED for illumination to study the broadband Mie backscattering from
optically levitated aerosol droplets to observe the morphological Mie resonances
simultaneously across a spectral range from 480 to 700 nm. Correlating the mea-
sured resonances to the mode order and mode number using Mie theory the droplet
size can be determined with an accuracy of ±2 nm.
As the scattering of a plane electromagnetic wave by a dielectric sphere is con-
sidered a canonical problem, Mie’s theory is still widely used as a standard reference
to validate methods intended for more complex scattering problems [143, 144]. For
example, Khoury et al. [145] compare COMSOL’s finite element method (FEM)
algorithm with the Mie theory for solving the electromagnetic fields in the vicin-
ity of a silica–silver core–shell nanoparticle. It is demonstrated that the COMSOL
FEM algorithm also generates accurate solutions of the near field. Takano and Liou
[146] developed a geometrical optics ray-tracing program to compute scattering for
concentrically stratified spheres and in the validation section show that it produces
the same general scattering features as the “exact” Lorenz–Mie theory for a size
parameter of 600.
There are various topics not considered in this review. For these topics I suggest some
further reading. For the basics on the Mie theory readers may refer to the classical
books by Stratton [43], van de Hulst [46] (available as a budget Dover edition), Born
and Wolf [45], Kerker [30] and Bohren and Huffman [47]. For a more recent and
extensive discussion on the topic refer to Grandy [27]. Morphological resonances in
Mie scattering are extensively covered in the book by Davis and Schweiger [28].
Scattering particles are commonly positioned on a plane interface and particle–
surface scattering interaction will have to be considered in the corresponding exten-
sion of Mie theory. This is for example a problem when simulation particle surface
scanners detect particle conterminants on wafers. For these who are interested in such
kinds of problems there are methods based on the T-matrix method taking account
64 T. Wriedt
of the plane surface. Such methods are fully outlined in the books by Doicu et al.
[147] and Borghese et al. [123]. There is a section on supported nanoparticles in the
recent book by Quinten [102]. For those interested in plasmonics and the optics of
nanosized noble metalparticles, Quinten’s book gives an overview of analytical and
numerical models for the optical response of nanoparticles and nanoparticle systems.
Nanooptics is the topic of the book by Novotny and Hecht [148] who also care for
the topic of particles on surfaces in an extensive way.
scattering pattern (Fig. 2.4) is for a sphere possessing a refractive index gradient gra-
dient with n = 1.33 at the centre of the sphere and n = 1.5 at the boundary of the
sphere. This gradient is approximated by 50 steps. With a chiral or optically active
sphere there are also great differences in the scattering pattern (Fig. 2.5).
If the sphere is positioned in a highly focused laser beam with Gaussian intensity
distribution pronounced differences to the first scattering diagram show up (Fig. 2.6).
A slightly perturbed sphere gives also only slight perturbation in the scattering dia-
gram (Fig. 2.7). With an unisotropic particle having three different refractive indices
in the three cartesian coordinates the scattering pattern is no longer rotationally sym-
metric (Fig. 2.8).
66 T. Wriedt
2.21 Conclusion
What is the current state of Mie’s theory? It seems to be as lively as ever. It is con-
tinuously used for code validation. Wherever particles can be considered spherical,
Mie scattering is applicable. It is applied everyday in diffraction-based instruments
to characterise particles. To this day, new approaches in computing Mie scattering
are developed and the methods are continuously extended to cover related scattering
problems such as coated spheres.
It is not only used for educating students and the validation of more advanced
theories but also it is the basis of radiative transfer, Lidar and optical particle char-
acterisation. Even today there is still progress in programming and new programs
which are based on Mie theory or which extend Mie theory in some respect show up
every year.
In this chapter an overview of the progress in developing light scattering programs
suitable for Mie-related scattering problems has been given. A short description of
each technique has been presented and suitable references have been provided such
that the reader can find more detailed information about the various methods and
sources of computer programs.
The state of the art in numerical light scattering modelling is progressing rapidly
especially with fast advancing research fields such as nanophotonics and near-field
optics. With almost all concepts there was much progress in the recent years.
The recent advances in computer hardware and the development of fast algorithms
with reduced computational demands and memory requirements have made the exact
numerical solution of the problem of scattering from large scattering particles highly
feasible. Today, scattering by spherical particles of realistic sizes can be computed in
an efficient way. Especially, the number of open source programs and development
projects is increasing continuously.
References
1. M.I. Mishchenko, L.D. Travis, Bull. Am. Meteorol. Soc. 89(12), 1853 (2008)
2. M.I. Mishchenko, J. Quant. Spectrosc. Radiat. Transf. 110(14–16), 1210 (2009)
3. U. Kreibig, Physik in unserer Zeit 39(6), 281 (2008)
4. W. Hergert, T. Wriedt (eds.), Mie Theory 1908–2008 (Universität Bremen, Bremen, 2008)
5. T. Stübinger, U. Köhler, W. Witt, U. Köhler, W. Witt, in Particulate Systems Analysis 2008
(Stratford-upon-Avon, UK, 2008)
6. T. Wriedt, in Mie Theory 1908–2008, ed. by W. Hergert, T. Wriedt (Universität Bremen.
Bremen, 2008), pp. 17–21
7. H. Horvath, J. Quant. Spectrosc. Radiat. Transf. 110(11), 783 (2009)
8. H. Horvath, J. Quant. Spectrosc. Radiat. Transf. 110(11), 787 (2009)
9. V. Garbin, G. Volpe, E. Ferrari, M. Versluis, D. Cojoc, D. Petrov, New J. Phys. 11(1), 013046
(2009)
10. M. Kolwas, Comput. Methods Sci. Technol. 16 Special Issue (2), 108 (2010)
11. T. Wriedt, Part. & Part. Syst. Charact. 15(2), 67 (1998)
12. F.M. Kahnert, J. Quant. Spectrosc. Radiat. Transf. 79–80, 775 (2003)
13. G. Veronis, S. Fan, in Surface Plasmon Nanophotonics (Springer, Dortrecht, 2007),
pp. 169–182
14. G.A. Niklasson, W.E. Vargas, Encyclopedia of Surface and Colloid Science, 2nd edn.
pp. 3346–3358 (2006)
15. J. Zhao, A.O. Pinchuk, J.M. McMahon, S. Li, L.K. Ausman, A.L. Atkinson, G.C. Schatz,
Acc. Chem. Res. 41(12), 1710 (2008)
16. C.J. Bouwkamp, Rep. Prog. Phys. 17(1), 35 (1954)
17. V. Myroshnychenko, J. Rodríguez-Fernández, I. Pastoriza-Santos, A.M. Funston, C. Novo,
P. Mulvaney, L.M. Liz-Marzán, F.J. F. Javier García de Abajo, Chem. Soc. Rev. (2008)
18. V. Veselago, L. Braginsky, V. Shklover, C. Hafner, J. Comput. Theor. Nanosci. 3, 189 (2006)
19. G. Mie, Annalen der Physik 330(3), 377 (1908)
20. R. Gans, Annalen der Physik 381(1), 29 (1925)
21. H. Blumer, Zeitschrift für Physik 32, 119 (1925)
22. N.A. Logan, Proc. IEEE pp. 773–785 (1965)
23. A. Clebsch, Journal für Mathematik, Band 61, Heft 3, 195 (1863)
24. L. Lorenz, Det Kongelige Danske Videnskabernes Selskabs Skrifter 6. Raekke, 6. Bind 1, 1
(1890)
25. L. Lorenz, Oeuvres scientifiques de L. Lorenz. Revues et annotées par H. Valentiner (Libraire
Lehmann & Stage, Copenhague, 1898), chap. Sur la lumière réfléchie et réfractée par une
sphère (surface) transparente., pp. 403–529
26. P. Debye, Annalen der Physik, Vierte Folge, Band 30. No. 1, 57 (1909)
27. W.T. Grandy, Scattering of waves from large spheres (Cambridge Univ. Press, Cambridge,
2000)
28. E. Davis, G. Schweiger, The Airborne Microparticle (Springer, Berlin, Heidelberg, 2002)
29. G. Burlak, The Classical and Quantum Dynamics of the Multispherical Nanostructures (Impe-
rial College Press, London, 2004)
30. M. Kerker, The Scattering of Light, and Other Electromagnetic Radiation: and other electro-
magnetic radiation (Academic Press, New York, London, 1969)
31. P. Pesic, Sky in a bottle (MIT Press, Cambridge, 2005)
32. O. Keller, Prog. Opt. 43, 257 (2002)
33. H. Kragh, Appl. Opt. 30(33), 4688 (1991)
34. M. Cardona, W. Marx, Physik Journal 11, 27 (2004)
35. W. Marx, Phys. Unserer Zeit, pp. 34–39 (2007)
36. B.F. Bowman, www.biochem.mpg.de/iv/AgFN_Bibliometrie.pdf. accessed 15. Aug. 2008.
(2008)
37. I. Fränz-Gotthold, M. von Laue, Annalen der Physik 425(3), 249 (1938)
2 Mie Theory: A Review 69
115. N.J. Moore, M.A. Alonso, Opt. Express 16(8), 5926 (2008)
116. P.W. Barber, S.C. Hill, Light scattering by particles: Computational methods (World Scientific
Publishing, Singapore, 1990)
117. L. Mees. GLMT Champ Internes (2008). http://www.scattport.org/index.php/programs-
menu/mie-type-codes-menu/130-glmt-champ-internes
118. M. Ringler, Plasmonische Nahfeldresonatoren aus zwei biokonjugierten Goldnanopartikeln.
(2008)
119. G. Pellegrini, Modeling the optical properties of nanocluster-based functional plasmonic mate-
rials. Ph.D. thesis, Tesi di dottorato, Università degli Studi di Padova, Padova (2008)
120. L. Boyde, K.J. Chalut, J. Guck, Phys. Rev. E 83(2), 026701 (2011)
121. S. Lecler, Etude de la diffusion de la lumière par des particules sub-microniques. Ph.D. thesis,
Thèse. l’Université Louis Pasteur, Strasbourg (2005)
122. R. Ruppin, Phys. Rev. B 11, 2871 (1975)
123. F. Borghese, P. Denti, R. Saija, Scattering from model nonspherical particles: theory and
applications to environmental physics (Springer- Berlin, New York, 2007)
124. K. Travis, J. Guck, Biophys. Rev. Lett. 2, 179 (2006)
125. Y. Okada, in Light Scattering Reviews, ed. by A.A. Kokhanovsky, 5th edn. Springer Praxis
Books (Springer Berlin, New York, 2010) pp. 3–35
126. W. Trinks, Annalen der Physik 414(6), 561 (1935)
127. S. Levine, G.O. Olaofe, J. Colloid Interface Sci. 27(3), 442 (1968)
128. J. Bruning, Y. Lo, IEEE Trans. Antennas Propag. 19(3), 378 (1971)
129. J. Bruning, Y. Lo, IEEE Trans. Antennas Propag. 19(3), 391 (1971)
130. K.A. Fuller, G.W. Kattawar, R.T. Wang, Appl. Opt. 25(15), 2521 (1986)
131. K.A. Fuller, G.W. Kattawar, Opt. Lett. 13(2), 90 (1988)
132. K.A. Fuller, G.W. Kattawar, Opt. Lett. 13(12), 1063 (1988)
133. D.W. Mackowski, J. Opt. Soc. Am. A 11(11), 2851 (1994)
134. D.W. Mackowski, M.I. Mishchenko, J. Opt. Soc. Am. A 13(11), 2266 (1996)
135. Y. lin Xu, Appl. Opt. 34(21), 4573 (1995)
136. Y.l. Xu, Appl. Opt. 36(36), 9496 (1997)
137. L. Liu, M.I. Mishchenko, W.P. Arnott, J. Quant. Spectrosc. Radiat. Transf. 109(15), 2656
(2008)
138. Y.l. Xu, J. Quant. Spectrosc. Radiat. Transf. 89(1–4), 385 (2004)
139. V. Schmidt, T. Wriedt, J. Quant. Spectrosc. Radiat. Transf. 110(14–16), 1392 (2009)
140. D. Mackowski, M. Mishchenko, J. Quant. Spectrosc. Radiat. Transf. 112, 2182 (2011)
141. M. Tagviashvili, Phys. Rev. A 81, 045802 (2010)
142. A.D. Ward, M. Zhang, O. Hunt, Opt. Express 16(21), 16390 (2008)
143. T. Grosges, A. Vial, D. Barchiesi, Opt. Express 13(21), 8483 (2005)
144. R.J. Zhu, J. Wang, G.F. Jin, Optik—Int. J. Light Electr. Opt. 116(9), 419 (2005)
145. C.G. Khoury, S.J. Norton, T. Vo-Dinh, Nanotechnology 21(31), 315203 (2010)
146. Y. Takano, K.N. Liou, Appl. Opt. 49(20), 3990 (2010)
147. A. Doicu, R. Schuh, T. Wriedt, in Light Scattering Reviews, 3 edn. by A.A. Kokhanovsky
(Springer-Berlin, Germany, 2008), pp. 109–130
148. L. Novotny, B. Hecht, Principles of Nano-Optics (Cambridge Univ. Press, Cambridge, 2006)
149. T. Wriedt, http://www.scattport.org
150. J. Hellmers, T. Wriedt, J. Quant. Spectrosc. Radiat. Transf. 110(14–16), 1511 (2009)
151. P.J. Flatau, http://code.google.com/p/scatterlib/
152. T. Wriedt, J. Quant. Spectrosc. Radiat. Transf. 109(8), 1543 (2008)
Chapter 3
From Theories by Lorenz and Mie
to Ontological Underdetermination
of Theories by Experiments
Gerard Gouesbet
3.1 Introduction
In light scattering theory, the most famous theory is likely to be the one published by
Gustav Mie in 1908 [1]. This theory describes the quasi-elastic interaction between
an electromagnetic plane wave and a homogeneous sphere defined by its (arbitrary)
diameter and its (arbitrary) complex refractive index. It allows one to calculate scat-
tered fields outside the sphere, internal fields, phase relations and various cross sec-
tions, including for radiation pressure forces. The theory is built using Maxwell’s
electromagnetism.
The year 2008 was the year of the 100th anniversary of Mie’s paper which has been
commemorated in several conferences (i) GAeF conference on “Light scattering:
Mie and More”, 3 and 4 July 2008, in Karlsruhe, Germany [2] (ii) 11th conference
on electromagnetic and light scattering, 7–12 September 2008, in Hatfield, UK [3]
G. Gouesbet (B)
LESP, CORIA, UMR CNRS, Rouen University,
6614 Rouen, France
e-mail: [email protected]
truth. But, even in this case, Descartes believed he would have done much if only all
the things that can be deduced from the causes he exhibited would entirely agree with
the experiments. Furthermore, such a false hypothesis, agreeing with experiments,
would be as useful to the life as a correct one, because it can be used in the same
way as the correct one to produce the effects that we want to produce.
There are actually many places in the principles of philosophy by Descartes where
such considerations are put forward. Let us simply mention another one, taken from
the fourth part of the principles [19], in which Descartes wrote that, “... in the same
way than an industrious watchmaker can produce two watches which indicate the
same time, without any difference which can be perceived for the external eye ... it
is certain that God possesses an infinity of various means, allowing him to make that
all the things of this world look as they now look, without however any possibility
for the human mind to know, among all the means it could use, which one he actually
used.” And Descartes reiterated the idea that, in some sense, it does not matter: ..and
I would believe having done enough if the causes I explained are such that all the
effects they can produce are like the ones that we see in the world, without informing
myself whether it is by them or by others that they are produced. I even believe that
it is equivalently useful in life to know the causes we have imagined or to know the
true ones.
Just after, Descartes, referring to Aristotle, also wrote that regarding the things
which our senses do not perceive, it is sufficient to explain what they could be, though
perhaps they are not as we describe them. For Van Fraassen [17], Descartes here
voiced a thoroughly empirical sentiment. Under a given interpretation, a physical
theory describes how things are - one way they might indeed be - but the story is
not unique, and for an empiricist it need not be unique. The idea that we may have
several stories is indeed in the spirit of Quine’s underdetermination.
In most cases, the history of Quine’s underdetermination, as has been reported
up to now, heavily relies on the invocation of the Supreme Being, namely God. The
process from a theological understanding of Quine’s underdetermination to its secular
interpretation is already achieved with Pascal. In a letter to the Révérend Père Noël
[20], he wrote: “In the same way than a same cause may produce several different
effects, a same effect can be produced by several different causes. Therefore, when
one discusses, with our human abilities, of the motion and of the stability of the
Earth, all phenomena of movements and retrogradations of the planets may perfectly
be derived from the hypotheses of Ptolemy, of Tycho, of Copernicus and of many
others, although from all of them only one can be correct”.
With Duhem, and his famous book on the structure of physical theories [16], we
arrive at one of the two thinkers who gave his name to a Duhem-Quine thesis (holism),
not to be confused however with Quine’s underdetermination [11], although the name
of Duhem can also be (and has been) associated with Quine’s underdetermination.
Duhem expressed the issue under discussion, in general terms, by comparing,
and opposing, the French and the English styles of making physics. As he stated, for
geometricians, like Laplace or Ampère (on the French side, but forgetting Descartes,
a theological precursor of Duhem), it would be absurd to give to a law of physics
two different theoretical explanations, and to express the idea that two such different
3 Ontological Underdetermination of Theories 77
because the levels of the hydrogen atom define a very weak domain of application.
Therefore, the problem is not only to define exactly what can be named a theory, but
also what are admissible domains of validity (or of application). There is indeed a
quasi-continuum of theoretical constructions, from a simple experimental law to a
consistent theory like relativity, and there is also a quasi-continuum of domains of
validity, from a certain domain, like the quantum mechanical domain, to subdomains,
and sub-subdomains, like the one concerning only the predictions of energy levels
of the hydrogen atom.
A pertinent analysis of Quine’s anthropology is available from Laugier-Rabaté
[24]. According to this author, it appears that Quine did not provide convincing
examples for his thesis of the underdetermination of the radical translation (another
famous thesis from Quine, not discussed in this paper). And, as other authors like
Duhem, the lack of convincing examples is also characterising the thesis of the
underdetermination of theories by experiments. Quine himself would certainly agree
with this statement if we refer to the mood expressed in his Erkentniss-paper [11].
Another author who is often quoted as relevant to the issue of Quine’s under-
determination is Van Fraassen, in particular in reference to his book “Laws and
Symmetries” [25]. Van Fraassen distinguished between theoretical equivalence and
empirical equivalence. For him, if two statements are logically (theoretically) equiv-
alent, then they are saying the same thing. However, if we consider a theory as a kind
of statement, say a meta statement, we have at least one example to demonstrate that
the affirmation of Van Fraassen is erroneous. This example concerns the Newton’s
and Hamilton-Jacobi’s formulations of classical mechanics (soon to be discussed).
As we shall see, both theories may be viewed as logically, theoretically, equivalent
by the simple fact that they are mathematically equivalent. But they are not saying
the same thing. Contrarily, they are proposing two drastically different visions of the
world, one with local trajectories, the other with local trajectories embedded into a
non-local field.
Also, Van Fraassen believed (as everyone he said!) that there exist numerous
theories, may be not yet formulated, but which are in agreement with the experimental
facts which are up to now known, and these theories are at least as explanatory as
the best theories that we possess nowadays. However, he added, the big majority
of these theories must be erroneous because they may disagree, in many different
manners, with experimental facts not yet known. He concluded that our best theory
is very unlikely to be correct. I would like to express this idea with a probabilistic
flavour: if there is only one truth, we are sure to be in error.
Elsewhere, Van Fraassen [17] discussed the lack of unicity asserted by Quine’s
underdetermination in association with the existence of different interpretations of
quantum mechanics. As he wrote, “why then be interested in interpretation at all?
If we are not interested in the metaphysical question of what the world is really like,
what need is there to look into these issues? Well, we should still be interested in the
question of how the world could be the way quantum mechanics–in its metaphysical
vagueness but empirical audacity–says it is. That is the real question of understand-
ing. To understand a scientific theory, we need to see how the world could be the
way that the theory says it is. An interpretation tells us that.” Later on, he added:
80 G. Gouesbet
“the answer is not unique, because the question ‘How could the world be the way
the theory says it is?’ is not the sort of question to call for a unique answer.”
The issue of interpretation is also considered by Cushing [26]: “Very loosely,
the formalism refers to the equations and calculational rules that prove empiri-
cally adequate (i.e. getting the numbers right) and the interpretation refers to the
accompanying representation the theory gives us about the physical universe (i.e. the
picture story that goes with the equations of what our theory ‘really’ tells us about the
world). Since a (successful) formalism does not uniquely determine its interpretation,
there may be two radically different interpretations (and ontologies) corresponding
equally well to one adequate formalism. This can be taken as an instantiation of the
Duhem-Quine thesis of undetermination of theory by an empirical base. Even if one
wants to restrict (and, arguably, that would be a mistake) the Duhem-Quine thesis
to different formalisms each handling equally well a given body of empirical infor-
mation, there nevertheless remains the interesting and important point of opposing
ontologies equally well supported by a common empirical base.”
The correct understanding of this Cushing’s quotation requires us to tell more
on the terminologies we may use. We have extensively discussed Quine’s thesis of
underdetermination. As underlined by Quine himself [11], this thesis “is not to be
confused with holism. It is holism that has rightly been called the Duhem thesis and
also, rather generously, the Duhem-Quine thesis. It says that scientific statements
are not separately vulnerable to adverse observations, because it is only jointly as
a theory that they imply their observable consequences.” In contradiction with this
quotation from Quine, but in agreement with other authors, Cushing used the termi-
nology “Duhem-Quine” to designate Quine’s underdetermination. Sometimes, the
expression of Duhem-Quine theorem (a philosophical theorem not demonstrated!)
may be found to express an enunciation of Quine’s underdetermination thesis.
Now, both van Fraassen and Cushing, driving us towards the issues of inter-
pretations and ontologies, allow one to put forward another point of view on the
interpretation of Quine’s underdetermination thesis, a point of view concerning
ontologies associated with theories and interpretations of theories. Let us enunciate
a Duhem-Quine theorem under the following form: we may have several conflicting
theories which are empirically equivalent. Now, such an enunciation is not complete
if we do not specify the meaning of the word “conflicting”. The meaning could be
as given by Quine, for example: two theories are conflicting when they are logi-
cally incompatible, and cannot be rendered logically equivalent by any reconstrual
of predicates. From now on, we use another definition: two theories are conflicting
when they exhibit different ontologies. The thesis of ontological underdetermination
then states that we may have empirically equivalent theories which do not exhibit the
same ontologies. It remains to define what is an ontology: the ontology of a theory
tells us which kinds of entities are populating the world. It describes the furniture of
the world as implied by the interpretation of the theory. For instance, the furniture of
classical physics is made out from localised objects and fields (possibly taking the
form of waves).
We are now going to provide explicit and well-defined examples of ontologi-
cal underdetermination, relying on theories which are rich enough, although not
3 Ontological Underdetermination of Theories 81
which are in contradiction insofar as they do not say the same thing on the world, i.e.
they do not exhibit the same ontologies. Indeed, Newton’s formulation only deals with
localised objects following trajectories, while Hamilton-Jacobi’s formulation deals
with localised objects following trajectories dressed by an unobservable, non-local,
action field filling the space. Because they are conflicting, these two formulations may
be counted as two different theories. They have to be taken as forming an example
of ontological underdetermination, for rich enough theories, except if we just shrug
our shoulders and rashly sweep S under the carpet.
A possible escape could be as follows. Since we are used to observe trajectories
of macroscopic objects in our everyday experience (macroscopic objects, not matter
points however), we are keen to believe that the field S of the Hamilton-Jacobi’s
formulation does not have any physical reality, but is simply an intermediary tool
for computations. After all, this field does not pertain to our sense data. Then, we
would have found a way to discriminate between Newton’s and Hamilton-Jacobi’s
formulations (theories) in favour of Newton’s formulation, by denying S to be an
ontological entity. As we shall see, this point of view, although highly reasonable,
will reveal itself to be erroneous.
For the second case study, we are going to compare two different kinds of interpre-
tations (actually two different theories) of quantum mechanics, the usual one, the
one in textbooks, which is conventional, standard, orthodox and the causal theories,
the heretical and unorthodox ones, mainly developed by Louis de Broglie and David
Bohm. As for the previous case, we shall avoid mathematical formulae to express
the essence in plain words.
The usual quantum mechanics, on the one hand, exhibits an intrinsic undeter-
minacy, associated with the Von Neumann’s projection postulate governing any
quantum measurement process. The causal theories, on the other hand, restore the
determinism in quantum mechanics by invoking a subquantum level containing
Newton-like trajectories of hidden particles. These causal theories are examples of
what is called “hidden variables theories”.
By definition, “causal theories” specifically refers to two theories initiated by
Louis de Broglie (double solution and pilot wave) intending to restore determinism
in quantum mechanics, hence producing a picture of the world much different from
the one provided by quantum mechanics. A good enough understanding of the issue
may be gained by reporting on a brief history of causal theories. Reporting on such a
history is a fairly simple exercise because we possess many invaluable testimonies,
in particular from the two main protagonists of the enterprise, namely Louis de
Broglie and Bohm, e.g. [28, 32–39] or [40], for a selection. Complementary historical
analyses and reports are also available from Vigier [41], Jammer [42, 43], Pinch
[44], Wheeler and Zurek [45], Holland [31] or Cushing [46], among others. A more
general point of view, not only on causal theories, but also on larger classes of hidden
3 Ontological Underdetermination of Theories 83
variables theories, is also discussed by Freistadt [47], Belinfante [48], Pipkin [49] or
d’Espagnat [50].
It happened that, for many reasons, Louis de Broglie was dissatisfied with quantum
mechanics, and then committed himself to the development of alternative theories.
Indeed, he certainly has been one of the most severe and most eloquent opponents
to the Copenhagen orthodoxy. He took many opportunities to claim his displeasure,
and even his reluctance, in particular with respect to the following items: the impos-
sibility to represent the motion of a quantum object in ordinary space and time and, in
contrast, the use of an abstract configuration space; more generally, the impossibility
of reconciling quantum mechanics with the clear ideas of classical physics; or the
fictitious and subjective character of the wave function (its probabilistic interpreta-
tion, or his lack of objectivity when it is interpreted as representing only a state of
knowledge). Indeed, for Louis de Broglie whose mind was educated with classical
physics, the ideas of the new physics, out of space, out of time and out of causality,
generated a head-on conceptual collision. Also, Louis de Broglie complained about
what has been called elsewhere the complacency to quantum mechanics, for instance
in [37] where he expressed his increasing conviction that the reasonings underlying
the present interpretation of quantum mechanics are not as decisive as they look, and
contain many small cracks. He later also expressed his feeling that most of the present
theoreticians, capitulating in face of exaggerated abstract tendencies, renounced too
easily to the building of an intelligible picture of quantum physical phenomena.
Very likely, the most important clash in Louis de Broglie’s mind arose from the
conflict between the concepts of particles and waves. This may be inferred from the
very writings of Louis de Broglie, in particular when he stated [34] that the duality
between waves and particles constituted the big tragedy of contemporary physics.
Quantum mechanics succeeded in a magnificent way to dissolve the duality between
waves and particles by a synthesis. Quantum objects are not classical waves and
they are not classical particles: not wave nor particle, but something else. For Louis
de Broglie, however, the price to pay was too heavy. In building hidden-variables
theories, he made the choice of another option: wave and particle [42]. The duality
became a complementarity, not however in the sense of Bohr. As soon as Louis
de Broglie conceived his first ideas on wave mechanics, there was no doubt for
him that it was necessary to find a new theory, synthesising the concepts of waves
and particles, and retaining as far as possible their traditional aspects. Concerning
particles, although quantum mechanics rejects the very concept of trajectory, Louis
de Broglie wanted to preserve it. For him, the two points of view (rejecting and
preserving) were not deeply conflicting because [35], from the fact that quantum
measuring processes do not allow us to simultaneously assign a position and a state
of motion to a corpuscle, it does not necessarily follow that, in reality , the corpuscle
does not have any position nor velocity. Concerning waves, for Louis de Broglie,
they had to be preserved too. In particular, Louis de Broglie was impressed by the
fact that the theory of atoms introduced integers to deal with the stationary motion
of electrons. But, in classical physics, such occurrences of integers only arose in
interference phenomena, and in oscillatory normal modes. Therefore, for Louis de
Broglie, corpuscles and waves were to become, in some way, two aspects of the same
84 G. Gouesbet
thing. As he expressed during the Solvay Congress of 1927 [51], the existence of
elementary corpuscles for matter and radiation being admitted as a matter of fact,
these corpuscles are supposed to be endowed with a periodicity. In such a vision, a
matter point is no more conceived as a static entity only concerning a tiny part of
space, but as the centre of a periodic phenomenon spreading around it. This idea that
a periodic phenomenon must be associated with a corpuscle, has been an obsession
to which Louis de Broglie dedicated most of his life.
The way to achieve the synthesis was to consider the corpuscle as a kind of per-
manent singularity embedded in an extended wavy phenomenon. This has led Louis
de Broglie to a theory known as the double solution. The name of the theory comes
from the idea (about 1925–1927) that the equation of motion of wave mechan-
ics (Schrödinger’s equation of course) is satisfied simultaneously by two coupled
equations. One solution, denoted u, represents a singularity, describing really the
corpuscle associated with the wave while the other solution, with a continuously
variable amplitude, describes only the statistical behaviour of the motion of a cloud
of particles. The first singular solution exhibits very high values of amplitudes, even
infinite in an early version, around a moving point of space. We are then facing a local
moving singularity which can be identified with a corpuscle. For Louis de Broglie,
although the motion of this corpuscle should admittedly satisfy new dynamical laws,
it should develop itself along a trajectory, with a well-defined velocity at each point
of it. The second solution formally identifies with the Ψ of quantum mechanics, but
with a different interpretation. For Louis de Broglie, the Ψ of the double solution, as
the Ψ of quantum mechanics, does not exactly describe the physical reality. It leads
indeed to a statistical theory, representing some kind of average, when the actual
trajectories followed by the corpuscles are totally unknown, with the exhibition of
two layers, the second layer, the deepest one, being of a classical nature and there-
fore deterministic. Then, at the level of the first layer , Ψ can be viewed as fictitious,
statistical, subjective, without however any epistemological damage.
The quantum mechanical indeterminacy then results from our lack of knowledge
concerning the initial conditions required to build the trajectories of the corpuscles in
the cloud. Therefore, determinism is restored in a classical way, with indeterminism
as the consequence of a lack of knowledge, instead of being of an intrinsic nature.
We obtain a hidden variables theory with hidden variables associated with quantities
describing the hidden trajectories. This is a brief summary of the first causal theory
of Louis de Broglie.
To gain a better understanding of the theory of the double solution, and more
important, of where it may come from, let us return to the Hamilton-Jacobi’s formu-
lation of classical mechanics. Let us recall that, in Hamilton-Jacobi’s formulation
of classical mechanics, we are facing trajectories which are orthogonal to iso-value
surfaces of the action S. Actually, we may consider a cloud of particles, generating
a compact set of trajectories, embedded in and orthogonal to a field phenomenon
extended in space. This is a classical structure that Louis de Broglie attempted to
generalise to a quantum mechanical structure. The motivation for this attempt was
that Louis de Broglie believed that it was wrong to abandon the notion of trajec-
tories, followed by classical particles. The Hamilton-Jacobi structure then provided
3 Ontological Underdetermination of Theories 85
vigorously attacked it, against Bohr and did not win. But more important and most
disappointing for Louis de Broglie is that Einstein, although in principle in favour of
a deterministic interpretation, did not support him. For details, we may return to the
printed discussion following the presentation of Louis de Broglie [53] which is not
very enthusiastic, and also to the final discussion following the bulk of the Congress
[51].
As a whole, besides the laconic indifference of Einstein, the pilot wave theory of
Louis de Broglie was severely criticised, possibly in a much more acerbic way than
the one we could draw by simply studying the accounts of the Congress [51]. You
know, during a congress, there is not only a conference room for public debates but
also coffee breaks, corridors and all sorts of informal gossips, particularly if, as was
the case, all participants shared the same hotel. Indeed, according to a testimony from
Heisenberg [56], the most animated discussions occurred during meals at the hotel.
This is the place where Einstein put forward, several times, his famous objection
that God does not play dices, to which Bohr could only reply that it is not to us to
prescribe how God should rule the world [56]. It might however be significant that
the name of Louis de Broglie is not mentioned at all by Heisenberg in this book [56]
where he reported so many details on what has been going on during the Solvay
Congress of 1927. This was however not a complete boycott of Louis de Broglie.
Indeed, Heisenberg let him express extensively in one of his books [57], although not
on the issue of hidden variables. A dramatic feature is that Pauli made an objection
that Louis de Broglie could not successively strike down. This objection was that the
pilot wave theory (or model) could not be applied in a coherent way to the case of
a two-body scattering process, specifically to the Fermi quantised rotator. Louis de
Broglie had become a lonesome warrior.
Returning from Brussels to Paris, thinking more about the pilot wave and on what
happened during the Solvay Congress (no support from Einstein, at least publicly,
Pauli’s objection, generally little support to him even from the parts of Schrödinger or
Lorentz ...), Louis de Broglie felt poignantly alone (Bohm’s testimony in [44]). It is
not a big step forward to believe that he must have felt depressed too. He thought that
Pauli’s objection could possibly be refuted, but more important, he found new objec-
tions by himself (such objections are discussed in [28]). The most significant one is
that, in the pilot wave model, the objective corpuscle was guided by a wave Ψ which,
insofar as it was reflecting our state of knowledge, was of a subjective nature. How
something subjective could possibly guide something objective? Another objection,
again associated with the status of Ψ , concerns the many-body problem because,
for this case, it is even more difficult to consider the pilot wave theory as giving a
genuine physical picture of phenomena, due to the fictitious and abstract character of
the propagation of a wave in the configuration space. Also, there was little comfort
with the double solution. It did not suffer from the above objections but was buried
in mathematical difficulties that, actually, Louis de Broglie never succeeded to solve.
For Louis de Broglie however, if the causal theories he started to develop were one
day or another day able to grow again from their ashes, it would be from the double
solution approach. Such a phoenix process would certainly be impossible from the
3 Ontological Underdetermination of Theories 87
pilot wave model (he erroneously thought), because this degenerated version was
absolutely indefensible.
Facing such a situation, Louis de Broglie felt so discouraged that he gave up
the causal theories, only a few years after he introduced them. He then capitulated,
defeated by the probabilistic interpretation, and even accepted to teach it, although,
we guess, reluctantly. Later on, Born will take notice of the acceptance by Louis de
Broglie of his defeat, and of his acknowledgment that the probabilistic interpretation
(the one of Born) is the only adequate one [58]. In 1932, five years after Brussels,
1927, von Neumann, in his celebrated book [59], produced a proof of the impossibility
of hidden variables which, seemingly, closed the issue forever. Apparently, this was
the final knock.
Ironically enough, it was not the double solution but, in contradiction with Louis
de Broglie’s expectation, the pilot wave theory which made the legendary Arabian
bird born anew from ashes. In 1952, Bohm published two papers on a pilot wave
approach [38, 60] in which he pointed out the fallacy of von Neumann’s proof of
impossibility, provided a rebuttal to Pauli’s objection and proposed a consistent solu-
tion to the measurement problem, among other issues. From an epistemological and
also methodological point of view, one of the most interesting features of Bohm’s
approach is the way he dealt with one of Louis de Broglie objections, namely the one
concerning the subjectivity of Ψ and the damning question to know how such a sub-
jective entity could pilot an objective corpuscle. For Bohm, the answer is incredibly
simple. By a kind of fiat, he just decreed from the beginning that Ψ is an objective
wave. This is however made consistent by the fact that Bohm provided an objective
solution to the measurement problem.
I shall not investigate deeply whether Bohm was aware or not of Louis de Broglie’s
work when he started to develop his theory, although we have the following testimony
from Bohm [39]: “then, in 1951 ... the author (that is to say Bohm) began to seek
such a model, and indeed shortly thereafter, he found a simple causal explanation of
the quantum mechanics which, as he later learned, had already been proposed by de
Broglie in 1927.” But, in any case, it is well attested that he was aware of it in 1951
because, one year before having his seminal papers published in 1952 in the Physical
Review, he sent preprints to Louis de Broglie. According to the latter, Bohm fully
picked up the pilot wave theory as it was exposed at the Solvay Congress of 1927,
including certain developments such as the introduction of a quantum potential. As a
matter of fact, he also acknowledged that Bohm was not aware of this work produced
25 years before, when he engaged himself on causal theories. He also acknowledged
that Bohm achieved a few more steps, in particular regarding Pauli’s objection and
the measurement problem [34]. The official reaction of Louis de Broglie has then
been to publish a C.R.A.S. to recall his indisputable priority in the topic, and also
the difficulties which made it repudiating the pilot wave after 1927 [33]. Objections
against the pilot wave are also summarised by Louis de Broglie in Refs. [28] and
[34].
According to a testimony of Bohm to Pinch [44], after having received the 1951
preprints, Louis de Broglie wrote back saying that he had already done this sort
of thing. Upon looking at the preprints, Pauli manifested a similar reaction, saying
88 G. Gouesbet
that “it was old stuff, dealt with long ago.” See Pauli [61] reporting on de Broglie-
Bohm story. But Louis de Broglie, in his reply to Bohm, also explicitly discussed
the objections against the pilot wave that made him give it up, and Bohm found that
he could answer them. The rebuttals are included in Appendix B of the second 1952
paper of Bohm [60] , saying: “all the objections of de Broglie and Pauli could have
been met if only de Broglie had carried his ideas to their logical conclusion.” Later
on, in many places, Bohm referred to Louis de Broglie as the primary father of the
filiation. For instance, in Bohm and Hiley [40], he wrote: “thus de Broglie proposed
very early what is, in essence, the germ of our approach,” and he considers his theory
as “‘an extension’ of de Broglie which answered all the objections.” But, as a whole,
the primary reception of Bohm’s ideas by Louis de Broglie was rather negative. In
the words of Belinfante [48], “when Bohm in his 1951 theory revived de Broglie’s
1927 theory of the pilot wave without attaching to it a theory of the double solution,
de Broglie objected against this by arguments that might have been presented by a
pupil of the Copenhagen school.”
Although Louis de Broglie’s reactions to Bohm’s work were not, at the beginning,
very enthusiastic (to say the least), he changed his mind rather fast, likely in part due
to the relevant answers of Bohm to his objections. According to Pinch [44], “these
answers were in part enough to convince de Broglie who then took up his original
interpretation again.” In the words of Cushing [46], “David Bohm’s 1952 paper was
to have a profound reconversion effect on de Broglie.” This was a new start for Louis
de Broglie who felt encouraged to take up his old ideas again, and to develop them in
various ways, with however his previous recurrent belief that the double solution was
on the right track. As testified in 1956 [35], Louis de Broglie has been wondering
since 1951 (the year when he received Bohm’s preprints) whether his first idea, the
double solution, after all, was not the correct one. In fact, already in 1952 [32], he
published a work in which he examined how the objective and causal theory of the
double solution could be taken up again.
Improvements and modifications are then introduced by Louis de Broglie. May
be the most innovative one concerns a new hypothesis telling that the equation of
propagation of the singularity wave u should be nonlinear, in contrast with the one for
Ψ which remains linear, although both equations are nearly everywhere identical.
Such a late focus on nonlinear equations was motivated by some developments
due to Einstein, in general relativity and unified field theory, in which besides the
gravitational field there also exist superimposed gravity singularities having to satisfy
nonlinear equations. We are here facing a strong similarity between the original ideas
of Louis de Broglie’s double solution and latest researches of Einstein, a similarity
pointed out by Vigier, and that Louis de Broglie tentatively exploited [35, 39, 41].
It is easy to pass from the double solution to the pilot wave. Essentially, what
we have to do is to forget anything concerning the second solution, that is to say
concerning the u-waves and their singularities, and to preserve the first solution Ψ
and the guidance formula driving the underlying hidden objective particles. Then,
these particles are piloted by the wave Ψ , hence the name of the theory. We are
still left with two versions, the pilot wave theory of Louis de Broglie, and the pilot
wave theory of Bohm. They essentially differ by the epistemological status of Ψ ,
3 Ontological Underdetermination of Theories 89
subjective for Louis de Broglie and objective for Bohm, and by the fact that Bohm
developed the theory much further, up to its logical conclusion, as he wrote. In
particular, objections which have been addressed to the version by Louis de Broglie
have been successfully enough answered by Bohm. From now on, when referring to
the pilot wave, we meant the 1952 version of Bohm [38, 60], which is enough for
further discussions.
In short, avoiding again, as in the previous case study, any mathematical devel-
opment, let us summarise by telling the reader that Bohm accepted the validity of
Schrödinger’s equation to describe the wave Ψ . However, Ψ is now viewed as an
objective field from the very beginning, exactly such as an electromagnetic field sat-
isfying Maxwell’s equations. We do not know from first principles why there should
be an electromagnetic field, nor do we know what is an electromagnetic field. There-
fore, why should we not adopt the same complacency for Ψ as the one we have for
E and H, or for their unification in an electromagnetic tensor Hi j ? In the words of
Bohm: “in the last analysis, there is, of course, no reason why a particle should not
be acted on by a Ψ -field, a gravitational field, a set of meson fields, and perhaps by
still other fields that have not yet been discovered.”
Elaborating on Schrödinger’s equation, Bohm demonstrated that the usual quan-
tum mechanics can be recast in another framework exhibiting Newton-like hidden
trajectories, forming a deterministic sublevel below the quantum level. In connection
with Hamilton-Jacobi’s formulation which inspired this approach, and to which it is
formally closely related, the velocity of the particles is orthogonal to the iso-phase
surfaces of Ψ which, therefore, indeed plays the role of a pilot wave.
Now, a most important fact is that the pilot wave is empirically equivalent to
quantum mechanics. Following Bohm, “all the results of the usual interpretation are
obtained from our interpretation if we make the following three special assumptions
which are mutually consistent:
(1) That the Ψ -field satisfies Schrödinger’s equation.
(2) That the particle momentum is restricted to p j = ∂ S/∂ x j .
(3) That we do not predict or control the precise location of the particle, but
have, in practice, a statistical ensemble with probability density P = |Ψ |2 . The
use of statistics is, however, not inherent in the conceptual structure, but merely a
consequence of our ignorance of the precise initial conditions of the particle.”
Therefore, we again have, in this example, two empirically equivalent theories
with conflicting ontologies: the orthodox quantum mechanics exhibiting an intrinsic
indeterminacy, and the heretical pilot wave relying on the deterministic motion of
hidden trajectories. In the pilot wave framework, indeterminacy is no more of an
intrinsic nature. It is simply the consequence of our lack of knowledge concerning
hidden trajectories. Probabilities are no more intrinsic, but now receive a classical
interpretation, like when tossing a coin, where our inability to predict the outcome
of the toss merely reflects our ignorance of initial conditions (and our inability to
carry out computations which are much too complicated and CPU-demanding).
90 G. Gouesbet
3.4.1 Undecidability
We can actually, at least in some cases, decide between undecidable physical theories.
When this is possible, it is achieved by using what is called ampliative arguments.
Harré [12], and Van Frassen quoting and rewording Harré [25] defined what an
acceptable theory is. Elaborating on these authors, I came to the following proposal.
A theory is satisfactory (acceptable, admissible...) if (1) it agrees with experimental
facts, (2) it is logically consistent and (3) it satisfies other demands. These other
demands, let us call them ampliative arguments. The word “ampliative” means that
these arguments enlarge our possibilities of choosing between several theories, the
enlargement being made with respect to the underdetermination by experiments. In
other words, ampliative arguments allow us to decide between undecidables. There-
fore, we may now build a new list of criteria as follows. A theory is satisfactory if
(1) it agrees with experimental facts, (2) it is logically consistent and (3) it satisfies
ampliative arguments.
It is interesting to remark that Einstein had something to tell us concerning Quine’s
underdetermination (without referring to it) and ampliative arguments (without using
92 G. Gouesbet
Similarly as for the optical rainbow, there exists a mechanical rainbow. To introduce
it, let us consider classical mechanics. A subtopic of classical mechanics is classical
scattering, e.g. [64, 65]. Following Nussenzveig [65], let us discuss the scattering of
a nonrelativistic particle of mass m by a central potential V (r ). The trajectory of the
incoming particle is deflected by an angle θ called the deflection angle which depends
on the impact parameter b. For a repulsive interaction, θ ranges from 0 to π but, for an
attractive interaction, it can take arbitrary large values (in modulus). There is a simple
relationship between the deflection angle θ and the more usual scattering angle θ ,
namely: θ + 2nπ = ±θ, n = 0, 1, 2, ... It is then established that the differential
−1
cross-section in the direction θ contains terms of the form dθ/db j , where the
index j arises from the fact that, in general, there exist several impact parameters
b j that lead to a same scattering angle θ . From this result, we see that we have a
divergence whenever dθ/db = 0, for at least one j, and for some angle θ R = θ , that
is to say when the deflection function, say θ (b), goes through an extremum. We then
have a stationary trajectory and the singularity is a rainbow singularity occurring at the
rainbow angle θ R . Other singularities involved in the phenomenon are glory, forward
scattering, orbiting, but we do not need to discuss them here. I then invoke a non-
singularity principle telling us that such singularities are inadmissible. From this, I
3 Ontological Underdetermination of Theories 93
conclude, without referring to any experiment (i.e. a priori), that classical mechanics,
which therefore contained the germ of its own destruction, is an approximation to a
more general theory. This more general theory must be a wave theory, say a wave
mechanics, in order to smooth out the rainbow singularity, and any other singularity,
as well. We therefore also conclude to the rational necessity of a wave mechanics.
I have used above a nonsingularity principle that I may call a special non-singularity
principle. This special principle might be an example of a more general nonsingularity
principle that I am not going, even tentatively, to formulate in this paper. Let us
however recall that there has been a long and venerable history concerning the concept
of infinity, both in mathematics and physics, even starting from the presocratic age.
And, nowadays, most physicists would certainly agree with the special nonsingularity
principle used above and with its consequences. Indeed, physicists are most usually
much reluctant to accept the actual existence of local infinite quantities in the world
they study. Whenever a physicist encounters a local infinity in physics, he instinctively
immediately recognises that something is going deeply wrong, or that he has been
dealing with convenient idealisations which cannot be actual, and possibly that he
has to think more. In other words, the non-singularity principle is indeed something,
at least implicitly, anchored in the mind of physicists, intimely associated with their
intuition. This does not imply that this intuition is correct, and does not mean that
it should be shared by everyone. Also, in any case, we possibly have to accept the
idea that any statement is revisable, even logical statements [66–68]. Nevertheless,
those who accept the special nonsingularity principle possess a decent ampliative
instrument to proceed further.
We are all used to the concept of trajectory, so easy to extract from classical mechan-
ics, in deep agreement with our intuition, and with our sense data: trajectories of cars,
of balls, or even of planets. There is no apparent difficulty to consider trajectories
of matter points. They at least constitute efficient models for the behaviour of so
many objects around us. Furthermore, physicists, when they are students, have been
trained, when learning classical mechanics, to the study of the mechanical behaviour
of matter points, before entering the realm of the mechanics of extended solids. There
is also this famous theorem telling us that the force between two uniform spheres
of masses M1 , M2 is equal to the force between two point masses M1 , M2 at their
respective centers [69].
We therefore feel very comfortable with the concept of trajectories in the Newton’s
formulation of classical mechanics and, as a consequence, we may feel very uncom-
94 G. Gouesbet
The fact that Newtonian trajectories of matter points do not exist is the ultimate
ampliative argument to discriminate between Newton’s and Hamilton-Jacobi’s for-
mulation of classical mechanics. In contrast with our naive expectation in which
the field S was viewed as a simple intermediary tool for computations, without any
physical significance, it is the Newtonian formulation of classical mechanics which
is to be rejected, and the Hamilton-Jacobi’s formulation which is to be given a due
privilege, let us say which is “closer to truth”. We are then left with a dressing field S
without any trajectory to be dressed. To understand such a weird situation, it is suffi-
cient to remark that, not only do Newtonian trajectories of matter points not exist but
also, it is classical mechanics as a whole which collapsed. Hence, Hamilton-Jacobi’s
formulation of classical mechanics is also to be rejected. Nevertheless, the field S still
remains physically meaningful. It actually appears to be related to the phase of Ψ in
quantum mechanics. Indeed [29], we may write Ψ under the form Ψ = exp(i S/) in
which, in the classical limit, the quantum field S of the previous expression identifies
with the classical field S of Hamilton-Jacobi’s formulation. The physical meaning of
the classical S of the inadmissible classical mechanics is therefore that it constitutes
a formal anticipation of the Ψ of quantum mechanics, more precisely of its phase.
The constant , which has the dimension of an action, has the virtue of changing S,
which has the dimension of an action too, to a dimensionless phase i S/.
Concerning the second case study (pilot wave versus usual quantum mechanics),
the relevance of ontological underdetermination for a discussion of Bohm’s pilot
wave has been noticed by Cushing [46, 70]. His point of view is that [70] “one
formalism, with two different interpretations, counts as two different theories” and
that [46] “the physical interpretation refers to what the theory tells us about the
underlying structure of ... phenomena, i.e. the corresponding story about the furniture
of the world.” This furniture of the world is that he called an ontology (the point of
view also adopted in this paper). Therefore, quantum mechanics and Bohm’s pilot
wave, although experimentally equivalent (implying that both of them should be
accepted by positivists), provided two different ontologies. Bohm himself was aware
of the issue at least implicitly [39].
However, we may ampliatively discriminate between the pilot wave and the usual
quantum mechanics. For this, we just need to remark that the same ampliative
arguments (non-singularity principle, inexistence of Newtonian trajectories of mat-
ter points) which have been used to discriminate between Newton’s and Hamilton-
Jacobi’s formulations of classical mechanics may be used to discriminate between
Bohm’s pilot wave and quantum mechanics. If objective, deterministic, Newtonian
trajectories of matter points (and matter points themselves) do not exist, it is non-
sensial to introduce them in quantum mechanics as done by Bohm. The attempt
to propulse the classical concepts of matter points and matter point trajectories in
the quantum domain leads definitively to a failure. Trajectories of matter points do
not exist in classical mechanics; hence, they should not be reintroduced in quantum
96 G. Gouesbet
mechanics, even if they are hidden. This statement applies to all causal theories (pilot
wave of Louis de Broglie and Bohm, and double solution of Louis de Broglie).
The rejection of inadmissible theories is not contradictory with the fact that they
can make very decent models, much useful in practice. And, after all, our best theories
are very far from being perfect and completed: from this point of view, they are
all models. In particular, the fact that singularities in the behaviour of Newtonian
trajectories are rare, occurring only occasionally, demonstrate that the concept of
trajectory of matter points still remains useful for many practical purposes, in the
same way that optical ray computing and tracing will forever remain invaluable
tools. But, if you need to dig deep into the mysteries of the world, it is definitely of
good advice, and even compulsory, to abandon inadmissible theories, when possible.
The idea, expressed by Bohm in its auto appraisals, and by other authors, that the
pilot wave could provide complementary insights in to a better, more thorough, and
deeper, understanding of quantum mechanics, is then erroneous and even dangerous,
for there could not be complementary insights in erroneous ideas, only an opportunity
to spoil the clarity of the mind. Any theoretician in quantum field theory, for whom
the description of reality in terms of particles being permanent entities with fixed
numbers is more than naive, would agree with this statement.
The two case studies discussed above are strongly correlated. One pertains to clas-
sical mechanics, the other to quantum mechanics, but both of them were anchored
on the existence of the Hamilton-Jacobi’s formulation of classical mechanics, and
undecidables were made decidable by relying on similar ampliative arguments. We
are now going to briefly discuss a third case study, related to another physical frame-
work, as discussed in the introduction, and solve a question that I asked myself about
30 years ago, a long time before I heard from Quine’s underdetermination and from
ampliative arguments, when I started to work on light scattering theories with the
building of a generalised Lorenz–Mie theory (GLMT) in mind [71].
For this third case study, we may again rely on an ampliative argument to dis-
criminate between Lorenz and Mie. This argument simply invokes the well-founded
rejection of ether by Einsteinian relativity. Hence, Mie was “closer to truth” than
Lorenz. Considering the effort accomplished by Lorenz, he certainly deserved to
have his name associated to what I always preferred to call Lorenz–Mie theory.
Nevertheless, the two theories, although empirically equivalent, are ontologically
different (conflicting), and better have to be viewed as two genuine different theo-
ries.
3.5 Conclusion
Quine’s underdetermination thesis, loosely speaking, states that theories are under-
determined by experiments. I also introduced an ontological version in which we may
have several empirically equivalent theories, with however different ontologies. We
have examined three exemplifying case studies. Two of them (Newton’s formulation
versus Hamilton-Jacobi’s formulation of classical mechanics, causal theories versus
3 Ontological Underdetermination of Theories 97
the orthodox interpretation in quantum mechanics) are closely related to the debates
on the foundations of quantum mechanics. The third case study (Mie’s theory versus
Lorenz’ theory) is borrowed from electromagnetic theory, more specificaly from
light scattering theory. Each case study exhibits a couple of theories which lead to
identical experimental predictions but are contradictory insofar as they do provide
conflicting visions of the world. The existence of conflicting empirically equivalent
theories implies, in principle, strong limitations to any realistic interpretation of
science. However, in each case study discussed in this chapter, we have been able to
invoke ampliative arguments allowing one to decide between undecidables.
References
1. G. Mie, Beiträge zur Optik trüber Medien speziell kolloidaler Metalösungen. Annalen der
Physik 25, 377–452 (1908)
2. H. Horvath (ed.), Light scattering, Mie and more. J. Quant. Spectrosc. Radiat. Transf.(special
issue) 2009
3. J.H. Hough (ed.), 11th conference on electromagnetic and light scattering by nonspherical
particles. J. Quant. Spectrosc. Radiat. Transf.(special issue) 2009
4. M.I. Mischenko, L.D. Travis, Gustav Mie and the evolving discipline of electromagnetic scat-
tering by particles. Bull. Am. Meteorol. Soc. 89(12), 1853–1861 (2008)
5. L. Lorenz, Lysbevaegelsen i og uden for en af plane lysbolger belyst kulge. Vidensk. Selk. Skr.
6, 1–62 (1890)
6. L. Lorenz, Sur la lumière réfléchie et réfractée par une sphère transparente. Lib. Lehmann et
Stage, Oeuvres scientifiques de L.Lorenz, revues et annotées par H.Valentiner, 1898
7. N.A. Logan. Survey of some early studies of the scattering of plane waves by a sphere, reprinted
in selected papers on light scattering by M. Kerker. SPIE, p. 951, 1965
8. N.A. Logan. Survey of some early studies of the scattering of plane waves by a sphere, ed.
by D. Hirleman. in Proceedings of the Second International Symposium on Optical Particle
Sizing, (Arizona State University, Arizona, 1990), pp. 7–15
9. H. Kragh. Ludwig Lorenz: his contributions to optical theory and light scattering by spheres,
ed. by D. Hirleman, in Proceedings of the Second International Symposium on Optical Particle
Sizing, (Arizona State University, Arizona, 1990), pp.1–6
10. H. Kragh, Ludwig Lorenz and the nineteenth century optical theory. the work of a great Danish
scientist. Appl. Opt. 30(33), 4688–4695 (1991)
11. W.V. Quine, On empirically equivalent systems of the world. Erkenntnis 9, 313–328 (1975)
12. R. Harré, Varieties of realism, a rationale for the natural sciences (Basil Blackwell, Oxford,
1986)
13. A. Koestler, Les somnambules, French translation of “The sleepwalkers” (Calmann-Lévy,
Paris, 1960)
14. O. Gingerich, Le livre que nul n’avait lu, à la poursuite du “De Revolutionibus” de Copernic.
French translation of: “The book nobody read-chasing the revolution of Nicolaus Copernicus”
(Dunod, Paris, 2008)
15. Galileo Galilei, Dialogue sur les deux grands systemes du monde, French translation of: Dial-
ogo sopra i due massimi sistemi del mondo tolemaico e copernicano (Editions du Seuil, Paris,
1992)
16. P. Duhem, La théorie physique, son objet, sa structure. English version: the aim and structure
of physical theory (Librairie philosophique J. Vrin, Paris, 1997)
17. B. Van Fraassen, Quantum mechanics: an empiricist view (Clarendon Press, Oxford, 2000)
18. R. Descartes, Les principes de la philosophie, troisième partie: pirouettes et tourbillons des
cieux (Paleo, France, 2000)
98 G. Gouesbet
19. R. Descartes, Les principes de la philosophie, quatrième partie: la Terre et son histoire (Paleo,
France, 1999)
20. B. Pascal, Oeuvres complètes, tome 1, chapter: Lettre de Blaise Pascal au très Révérend Père
Noël (Gallimard, Paris, 1998), pp. 377–386
21. W.V. Quine, Le mot et la chose, French translation of “Word and Object” (MIT
Press/Flammarion, Cambridge/Paris, 1960/1977)
22. W.V. Quine, La poursuite de la vérité, French Translation of “Pursuit of Truth” (Harvard
University Press/Editions du Seuil, Cambridge/Paris, 1990/1993)
23. Sous la direction de Pierre Wagner, editor. Les philosophes et la science (Gallimard, Paris,
2002)
24. S. Laugier-Rabaté, L’anthropologie logique de Quine, l’apprentissage de l’obvie (Librairie
philosophique J Vrin, Paris, 1992)
25. B. Van Fraassen, Lois et symétries, French translation of “Laws and symmetry” ( Oxford
University Press/Librairie philosophique J Vrin, Oxford/Paris, 1989/1994)
26. J.T. Cushing, Wave-particle duality, ed. by F. Selleri, Causal quantum theory, why a nonstarter?,
(Plenum Press, New York, 1992), pp. 37–63
27. E. Squires, The mystery of the quantum world (Adam Hilger, Bristol, 1986)
28. Louis de Broglie, Introduction à l’étude de la mécanique ondulatoire (Hermann et Cie, Paris,
1930)
29. D.I. Blokhintsev, Mécanique quantique (Masson et Cie, Paris, 1967)
30. L.D. Landau, E.V. Lifchitz, Mécanique, English translation: Mechanics ( Pergamon
Press/Editions Mir, Oxford,/Moscow, 1969/1966)
31. P.R. Holland, The quantum theory of motion, an account of the de Broglie-Bohm causal inter-
pretation of quantum mechanics (Cambridge University Press, Cambridge, 1993)
32. Louis de Broglie, Sur la possibilité d’une interprétation causale et objective de la mécanique
ondulatoire. Comptes-Rendus de l’Académie des Sciences 234, 265–268 (1952)
33. Louis de Broglie, Louis de Broglie, physicien et penseur Vue d’ensemble sur mes travaux
scientifiques, Collection dirigée par André George (Editions Albin Michel, Paris, 1953) pp.
457–486
34. Louis de Broglie, La physique quantique restera-t-elle indéterministe? (Gauthier-Villars, Paris,
1953)
35. Louis de Broglie, Une tentative d’interprétation causale et non linéaire de la mécanique
ondulatoire (la théorie de la double solution). English translation: Nonlinear Wave Mechanics
(Elsevier/Gauthier-Villars, Amsterdam/Paris, 1960/1956)
36. Louis de Broglie, La théorie de la mesure en mécanique ondulatoire ( interprétation usuelle
et interprétation causale) (Gauthier-Villars, Paris, 1957)
37. Louis de Broglie, Etude critique des bases de l’interprétation actuelle de la mécanique ondu-
latoire (Gauthier-Villars, Paris, 1963)
38. D. Bohm, A suggested interpretation of the quantum theory in terms of “hidden” variables.
part 1. Phys. Rev. 85, 166–179 (1952)
39. D. Bohm, Causality and chance in modern physics (Routledge and Paul Kegan, London, 1957)
40. D. Bohm and B.J. Hiley, The undivided universe, an ontological interpretation of quantum
theory (Routledge and Paul Kegan, London, 1993)
41. J.P. Vigier, Structure des micro-objets dans l’interprétation causale de la théorie des quanta
(Gauthier-Villars, Paris, 1956)
42. M. Jammer, The philosophy of quantum mechanics. The interpretations of quantum mechanics
in historical perspective (Wiley, New York, 1974)
43. M. Jammer, The conceptual development of quantum mechanics. The History of Modern
Physics, 1800–1950, vol 12 (Tomash Publishers, Los Angeles, 1989)
44. T.J. Pinch, in What does a proof do if it does not prove, a study of the social conditions and
metaphysical divisions leading to David Bohm and John von Neumann failing to communicate
in quantum physics, ed. by E. Mendelson, P. Weingart and R. Whitley. Social production of
scientific knowledge, sociology of the sciences, Vol 1 (D. Reidel, Dordrecht, 1977), pp. 171–
215
3 Ontological Underdetermination of Theories 99
45. J.A. Wheeler, W.H. Zurek (eds.), Quantum theory and measurement. Princeton ser. phys.
(Princeton University Press, Princeton, 1983)
46. J.T. Cushing, Quantum mechanics, historical contingency and the Copenhagen hegemony (The
University of Chicago Press, Hyde Park, 1994)
47. H. Freistadt, The causal formulation of the quantum mechanics of particles (the theory of de
Broglie, Bohm and Takabayasi). Supplemento al Nuovo Cimento, Ser. 10(5), 1–70 (1957)
48. F.J. Belinfante, A survey of hidden-variable theories (Pergamon, Oxford, 1973)
49. F.M. Pipkin, in Atomic physics tests of the basic concepts in quantum mechanics, ed. by D.R.
Bates and B. Bederson. Advances in Atomic and Molecular Physics (Academic Press, New
York, 1978) pp. 281–340
50. B. d’Espagnat, Traité de physique et philosophie (Fayard, Paris, 2002)
51. Electrons et photons, Rapports et discussions du cinquième congrès de physique tenu à Brux-
elles du 24 au 29 octobre 1927 (Gauthier-Villars, Paris, 1928)
52. P. Marage and G. Wallenborn, Les conseils Solvay et les débuts de la physique moderne (Uni-
versité libre de Bruxelles, Brussels, 1995)
53. Louis de Broglie, Electrons et photons, Rapports et Discussions du cinquième conseil de
physique tenu à Bruxelles du 24 au 29 octobre 1927, chapter: Nouvelle dynamique des quanta,
(Gauthier-Villars, Paris, 1928) pp. 105–141
54. N. Bohr Electrons et Photons, Rapports et Discussions du cinquième conseil de physique
tenu à Bruxelles du 24 au 29 octobre 1927, chapter: Le postulat des quanta et le nouveau
dèveloppement de l’atomistique, (Gauthier-Villars, Paris, 1928), pp. 215–287
55. E. Schrödinger, Electrons et Photons, Rapports et Discussions du cinquième conseil de physique
tenu à Bruxelles du 24 au 29 ocotbre 1927, chapter: La mécanique des ondes, (Gauthier-Villars,
Paris, 1928), pp. 185–213
56. W. Heisenberg, La partie et le tout, le monde de la physique atomique, French translation of:
Der Teil und das Ganze, Gespräche im Umkreis der Atomphysik (Editions Albin Michel, Paris,
1972)
57. W. Heisenberg, La nature dans la physique contemporaine, translated from German, collection
Idées (Gallimard, Paris, 1962)
58. M. Born, Louis de Broglie, physicien et penseur, collection dirigée par André George, chapter:
La grande synthèse, (Editions Albin Michel, Paris, 1953) pp. 165–170
59. J. von Neumann, Les fondements de la mécanique quantique, French translation of “Mathe-
matische Grundlagen der Quanten-mechanik”, 1932. English version: Mathematical founda-
tions of quantum mechanics (Princeton University Press/Librairie Félix Alcan, Princeton/Paris,
1955/1946)
60. D. Bohm, A suggested interpretation of the quantum theory in terms of “hidden” variables,
part 2. Phys. Rev. 85, 180–193 (1952)
61. W. Pauli, Louis de Broglie, physicien et penseur, collection dirigée par André George, chapter:
Remarques sur le problème des paramètres cachés dans la mécanique quantique et sur la théorie
de l’onde pilote (Editions Albin Michel, Paris, 1953) pp. 33–42
62. B. d’Espagnat, Le réel voilé, analyse des concepts quantiques (Fayard, Paris, 1994)
63. M. Born, The Born-Einstein letters, correspondence between Albert Einstein and Max and
Hedwig Born, from 1916 to 1955 with commentaries by Max Born, ed. by M.Born (MacMillan,
London, 1971)
64. R.G. Newton, Scattering theory of waves and particles (Dover Publications, New York, 2002)
65. H.M. Nussenzveig, Diffraction effects in semiclassical scattering (Cambridge University Press,
Cambridge, 1992)
66. W.V. Quine, Philosophie de la logique, présentation par Denis Bonnay et Sandra Laugier
(Aubier-Montaigne, Paris, 1975)
67. W.V. Quine, Relativit de l’ontologie et autres essais, présentation par Sandra Laugier (Aubier-
Montaigne, Paris, 1977)
68. W.V. Quine, Du point de vue logique, neuf essais logico-philosophiques, French translation
of: from a logical point of view, bibliothèque des textes philosophiques, (Editions Vrin, Paris,
2003)
100 G. Gouesbet
69. C. Kittel, W.D. Knight, and M.A. Ruderman, Mechanics (McGraw-Hill Book Company, New
York, 1962)
70. J.T. Cushing, Bohm’s theory: common sense dismissed. Stud. Hist. Philos. Sci. 24(5), 815–842
(1993)
71. G. Gouesbet, Generalized Lorenz-Mie theories, the third decade: a perspective, invited review
paper. J. Quant. Spectrosc. Radiat. Transf. 110, 1223–1238 (2009)
Chapter 4
Predicting the Appearance of Materials Using
Lorenz–Mie Theory
Jeppe Revall Frisvad , Niels Jørgen Christensen and Henrik Wann Jensen
Computer graphics systems today are able to produce highly realistic images. The
realism has reached a level where an observer has difficulties telling whether an
image is real or synthetic. The exception is when we try to compute a picture of a
scene that really exists and compare the result to a photograph of the real scene. In
this direct comparison, an observer quickly identifies the synthetic image. One of the
problems is to model all the small geometrical details correctly. This is a problem
that we will not consider. But even if we pick a simple experimental set up, where the
objects in the scene have few geometrical details, a graphics system will still have a
hard time predicting the result of taking a picture with a digital camera. The problem
here is to model the optical properties of the materials correctly. In this chapter, we
show how Lorenz–Mie theory enables us to compute the optical properties of turbid
materials such that we can predict their appearance. To describe the entire process of
predicting the appearance of a material, we include a description of the mathematical
models used in realistic image synthesis.
4.1 Introduction
The appearance of an object is a result of light reaching the eye after interacting with
matter. One way to capture the appearance of an object is to take a picture of it. This
means that a digital camera is not only an inexpensive consumer product, it is also a
fairly advanced device for measuring light. As we know from spectroscopy, light can
tell us a great deal about the material that it is illuminating. The question is then: what
can we learn about a material by taking a picture of it using a digital camera? The
W. Hergert and T. Wriedt (eds.), The Mie Theory, Springer Series in 101
Optical Sciences 169, DOI: 10.1007/978-3-642-28738-1_4,
© Springer-Verlag Berlin Heidelberg 2012
102 J. R. Frisvad et al.
immediate response is: not a lot. Unless we have a lot of pictures of similar materials
in the same lighting environment which we can use to make a statistical model (this
is image analysis). Or unless we have a physical model of the material and lighting
environment that we can use to predict the picture (this is computer graphics). The
purpose of this chapter is to show how to construct a model which predicts the result
of taking a picture of a material in a controlled lighting environment. In short, we
describe how to predict the appearance of materials. Once we have an appearance
model that predicts the result of taking a picture of a material, we can take a picture
of a similar material and fit the parameters in the model such that the picture of this
similar material is obtained. In this way, we can use a picture of a material to estimate
the properties of the material (analysis by synthesis).
The advantage of using a camera instead of a spectrometer is that it is inexpensive
and that it does not require small samples to be prepared for analysis inside a larger
device. Taking a picture using a digital camera is a simple, non-invasive process. This
means that it is particularly useful for food quality control and medical diagnosis.
To construct an appearance model that predicts the result of taking a picture of a
material, we need (a) a model of the camera, (b) a geometrical model of the material
sample, (c) a model of the light sources, (d) a model for light propagation and
(e) a model for light scattering. The problematic model is the one for light scattering
because the material properties are unknown. The macroscopic optical properties of
materials are so diverse that it is impossible to measure and tabulate the properties
of all materials directly. This is where Lorenz–Mie theory comes in handy. Lorenz–
Mie theory provides the link between the particle composition of a material and
its macroscopic optical properties. The particle composition of a material is a very
flexible starting point. There are many different ways to combine small particles
into one macroscopic bulk material. To limit the potentially large number of input
parameters, we use empirical models to describe the size distribution and optical
properties of the particles. When available, these empirical models enable us to
construct material appearance models that depend on only few essential inputs. A
list of ingredients, for example, where the relative contents are specified in weight
percent. The example used in this chapter is cow’s milk with fat and protein contents
of the milk (in weight percent) as input parameters.
The basics of realistic image synthesis have been described in many textbooks.
A useful reference which provides both detailed descriptions and source code is the
book by Pharr and Humphreys [5]. The following sections provide brief descriptions
of the five models that we need.
4.2.1 Camera
Think about taking a picture. The first thing you do is to position the camera and
orient it towards the objects that you wish to photograph. To model this set up,
we need a camera position p, a viewing direction v, and an up-vector u. A digital
camera is built around a photoactive charge-coupled device chip (CCD chip) which
has a limited number of bins across its area. This means that output images from a
digital camera have a specific resolution (image width W and height H ) measured
in numbers of pixels. Each pixel corresponds to a small area of the CCD chip and
has an associated vector which describes the amount of light (of different colours)
that reached this area during exposure. To capture an image is to expose the chip for
a short amount of time and record the vectors for all the pixels. We model the light
sensitive area of the CCD chip by a rectangle in the image plane, and we will call it
the film. The viewing direction v is normal to the image plane and the up direction of
the image plane is given by the up-vector u. To position the film in space, we centre
it around the camera position p and move it a distance d in the viewing direction.
The distance d depends on the lens system of the camera and is called the camera
constant. The size of the film (width by height, w × h) is usually specified by an
angle in the vertical direction called the field of view, fov and an aspect ratio a such
that
w = ah
h = 2d tan(fov/2).
Using the camera resolution (W × H ), we can divide the film into pixels. To compute
an image, we trace a number of rays from p through each pixel of the film.
The triangle mesh poses some problems. Unless we want to make objects with
sharp edges, we need an awfully large number of triangles to make an object look
smooth. Even more triangles are needed if we want smooth reflections and refractions
as well. This problem is usually solved by computing a normal for each vertex in
the triangle mesh as originally suggested by Phong [6]. Such a vertex normal is an
average of the normals associated with the neighbouring triangle faces. When we
want to find the surface normal where a ray intersects a triangle, we interpolate the
vertex normals across the triangle using trilinear interpolation. In this way, we have
a number of points (the vertices in the triangle mesh) which make out a discrete
representation of a smooth surface.
A light source is typically modelled by letting an object in the scene emit light equally
in all directions from every surface point. Another option is to have a spherical map
around the scene which acts as a background source or environment lighting. This
is called an environment map and it is modelled as if it were infinitely far away.
It is particularly useful for outdoor sceneries, where one can use a model of the
atmosphere to compute a realistic environment map [7–9].
Another important type of light source is laser. As has been discovered in bio-
medical optics [10], shining a laser into a turbid material (skin, for example) tells us
surprisingly much about the material. To model a laser source, we use a circular disk
which emits light weighted by a Gaussian function diminishing with the distance to
the centre of the disk. Laser is also nearly collimated. This is modelled by letting the
source emit light only in the direction normal to the surface of the disk.
Pictures most often capture light in the visible part of the electromagnetic spectrum.
Visible wavelengths are very short, ranging from 380 to 780 nm, and it was shown
by Sommerfeld and Runge [11] that plane electromagnetic waves propagating in a
non-absorbing, homogeneous dielectric are equivalent to rays of light for λ → 0,
where λ denotes the wavelength in vacuum. This means that rays of light are often
a good model of visible light. According to Fermat’s principle, rays of light follow
the path of least time [12, pp. 457–463]. We can use this principle to trace rays of
light through a digitally modelled scene. The main limitation is that wave phenomena
such as interference and diffraction will not be captured. The visual effects from such
phenomena must then be incorporated into the scattering properties of the materials
in the scene. Diffraction by particles is incorporated into the macroscopic scattering
properties of a material using Lorenz–Mie theory. This is covered in Sect. 4.2.5.
Interference and diffraction effects caused by microfacets in the surface of an object
4 Predicting the Appearance of Materials 105
are incorporated into surface reflection models. We will not discuss microfacetted
surface reflection models.
Rays of light are traced in straight lines as long as they travel in a homogeneous
medium, since, in this case, the shortest path is also the path of least time. Having
settled on a ray theory of light, we trace light backwards from the camera position p
through a pixel into our scene to see if it intersects one of the triangles that the objects
are composed of. The number of triangles is large, so some kind of spatial data
structure is needed to find the triangles that are most likely to be intersected [13–16].
An efficient ray-triangle intersection algorithm has been developed by Möller and
Trumbore [17]. Only the closest intersected triangle is considered, other intersected
triangles are hidden behind it. Once a point of intersection is identified and a surface
normal has been computed, the next step is to estimate the light that reaches this
point. Assuming a smooth surface, light reflects off the surface and refracts into the
material as described by Fresnel’s formulae for reflectance [18]. New rays are traced
from the point of intersection in the directions of reflection and refraction (as dictated
by the laws of reflection and refraction). Light reflects and refracts recursively until
it reaches a light source. This is called ray tracing, and this recursive version, where
new rays are traced from the point of intersection, was first described for graphics
by Whitted [19]. If light refracts into a turbid material, it scatters as described in
Sect. 4.2.5.
What we trace along a ray is electromagnetic energy. The rays follow the direction
of the time-averaged Poynting vector as closely as possible.
In the previous section, we found a way to trace energy through matter. This is all
we need to render materials with a continuous interior. However, if a material is
composed of millions of microscopic particles, it is infeasible to model the surface
of every particle. Therefore, we introduce a model for macroscopic scattering. The
model used to describe light scattering is the radiative transfer equation [3]:
(ω · ∇)L(x, ω) = −σt (x)L(x, ω) + σs (x) p(x, ω , ω)L(x, ω ) dω + L e (x, ω),
4π
(4.1)
where L(x, ω) is the radiance at x in the direction ω, the subscript e denotes emis-
sion, and σs , σa , and σt = σs + σa are the scattering, absorption and extinction
coefficients, respectively. The phase function p specifies the normalised distribu-
tion of the scattered light. Radiance is a radiometric quantity measured in energy
flux per solid angle per projected area. The equation splits the directional derivative
(left-hand side), that is, the change in radiance along a ray, into three terms (right-
hand side): The first term denotes the exponential attenuation, the second denotes
the in-scattering from all directions and the third is an emission term.
106 J. R. Frisvad et al.
where L i , L r and L t are the incident reflected and transmitted radiances and likewise
θi , θr , and θt are the angles of incidence, reflection, and refraction. In spherical
coordinates, the solid angles are defined by
Using the law of reflection θi = θr and the fact that both the reflected and transmitted
rays lie in the plane of incidence φi = φr = φt , the boundary condition becomes
To find the transmitted angle, we have to consider the direction of the transmitted
ray. This is given by the law of refraction. We find
n 1
sin θt = sin θi
n 2
n
cos θt dθt = 1 cos θi dθi .
n2
4 Predicting the Appearance of Materials 107
2
n
With this result, the boundary condition is L i = L r + n 1 L t . Or, in terms of
2
outgoing radiance,
2
n1
L 0 = R L i + (1 − R) Li , (4.2)
n 2
where the subscripts 1 and 2 denote two locations along a ray of light. This means
that the quantity L 1 /n 2 2
1 is (approximately) constant along the ray. If we store L 1 /n 1
before tracing a ray from one point in a medium, then the radiance at the destination
point is L 2 = n 2 2
2 L 1 /n 1 . Now that we know how radiance behaves as we move along
a ray, we are ready to evaluate the radiative transfer Eq. (4.1) using ray tracing.
Rendering realistic images using the radiative transfer equation, was first pro-
posed by Kajiya and Von Herzen [22]. Most rendering algorithms use approximate
evaluation schemes to gain speed. Here, we will only describe the general way of
evaluating the radiative transfer equation. The remainder of this section is a short
account of Monte Carlo path tracing, which is a sampling-based rendering algo-
rithm that works in general. More information is available in the book by Pharr and
Humphreys [5]. Unfortunately, Pharr and Humphreys stop their treatment at sin-
gle scattering. Evaluation of the general case has been described by Pattanaik and
Mudur [23].
The emission term in the radiative transfer equation is just an added constant. It
is not difficult to include it, but it makes the equations rather long. Therefore, we
leave out the emission term in the following. The general approach is as follows. We
first parameterise the radiative transfer equation using the distance s that light has
travelled along a path into a medium. We have (for a non-emitter):
dL(s )
+ σt (s )L(s ) = σs (s ) p(s , ω , ω)L(s , ω ) dω , (4.3)
ds
4π
where ω denotes the tangential direction of the path at the distance s along the path.
The parameterised Eq. (4.3) is a linear, first-order, ordinary differential equation
(where σt (s ) is a variable coefficient). One way to solve such an equation is by
means of an integration factor:
108 J. R. Frisvad et al.
⎛ ⎞
s
Tr (s , s) = exp ⎝− σt (t) dt ⎠ . (4.4)
s
Note that Tr (s, s) = 1. Then by integration along the ray from the surface s = 0 to
the considered location in the medium s = s, Eq. (4.5) attains the form:
s
L(s) = Tr (0, s)L(0) + Tr (s , s)σs (s ) p(s , ω , ω)L(s , ω ) dω ds . (4.6)
0 4π
The integration factor itself Tr (s , s) = e−τ (s ,s) is sometimes referred to as the beam
(or path) transmittance. Finally, the first term on the right-hand side of the formal
solution (4.6) for the radiative transfer equation (Eq. 4.1) is referred to as the direct
transmission term, whereas the second term is called the diffusion term. Realistic
rendering is all about evaluating these terms in various kinds of ways.
To evaluate Eq. (4.6) using Monte Carlo path tracing, we first take a look at the
direct transmission term Tr (0, s)L(0). For this term we need to estimate the value of
the optical thickness τ (0, s). If the medium is homogeneous, this optical thickness
is simply given by
τ (0, s) = σt s. (4.7)
For heterogeneous media, we use sampling. The optical thickness is found quite
efficiently by the estimator:
N −1
σt (ti )t, (4.8)
i=0
where t is the step size (given as user input) and ti are locations along the ray found
using a single uniform random variable ξ ∈ [0, 1]:
4 Predicting the Appearance of Materials 109
ξ +i
ti = s. (4.9)
N
The number N of locations along the ray is found using the depth s which is the
distance to the next surface, that is, N = s/Δt. Having estimated the optical
thickness τ (0, s), the beam transmittance Tr (0, s) is easily found and multiplied by
the amount of radiance L(0) to reveal the direct transmission term. The radiance
L(0) is the radiance which contributes to the ray at the surface of the medium.
Evaluating the diffusion term is more involved. What we need is an estimator of
the form
N −1
1
Tr (s j , s)σs (s j )J (s j )
, (4.10)
N pdf(s j )
j=0
where the probability distribution function (pdf) preferably cancels out the transmit-
tance. The source function
J (s ) = p(s , ω , ω)L(s , ω ) dω (4.11)
4π
s
ccdf(s j ) = σt (t)Tr (t, s) dt = Tr (s, s) − Tr (s j , s) = 1 − Tr (s j , s).
s j
Using the inverse transformation method [5, e.g.], we get the following equation for
sampling an interaction along the ray according to this pdf:
ξ j = 1 − Tr (s j , s) = 1 − e−τ (s j ,s) or ln(ξ j ) + τ (s j , s) = 0,
t1 in ti = s −iΔt, we step along the ray by incrementing i, and if ln(ξ j )+τ (ti , s) > 0,
we stop and compute the location of the interaction as follows [23]:
τ (ti−1 , s) + ln(ξ j )
s j = s − (i − 1)Δt + . (4.13)
σt (ti )
α(s j ) = σs (s j )/σt (s j ),
N −1 N −1 M−1
1
1
p(s j , ωk , ω)L(s j , ωk )
J (s j ) = (4.14)
N NM pdf(ωk )
j=0 j=0 k=0
for ξ < α(s j ) and 0 otherwise. For an isotropic phase function, sampling a uniform
distribution over the unit sphere leaves only L(s j , ωk ) in the sum.
To summarise the algorithm, a ray is traced from an observer through a scene,
when it refracts into a turbid material, we do the following:
1. Trace a refracted ray. The radiance carried along the refracted ray is corrected
according to the boundary condition (Eq. 4.2).
2. The tracing of the refracted ray gives the depth s to the next surface.
3. The radiance L(0) which contributes to the ray at the surface is found by tracing
new rays in the directions of reflection and refraction. If L(0) is not too small,
we evaluate the direct transmission term:
a. The optical thickness τ (0, s) is estimated using Eqs. 4.8 and 4.9 (or Eq. 4.7
for homogeneous media).
b. The direct transmission term Tr (0, s)L(0) is found using the optical thick-
ness τ (0, s) (see Eq. 4.4) and the radiance which contributes to the ray at
the surface L(0).
4. For every diffusion term sample j = 1, . . . , N , a sample depth s j is found using
Eq. 4.13 (or Eq. 4.12 for homogeneous media).
4 Predicting the Appearance of Materials 111
5. For the samples s j > 0, a Russian roulette is done using the scattering albedo
α(s j ). For ξ < α(s j ), where ξ ∈ [0, 1] is a random variable, there is a scattering
event.
6. For every scattering event, the phase function p(s j , ωk , ω) is evaluated in M
sampled directions ωk with k = 1, . . . , M. Likewise, M new rays are traced
at the position s j in the directions ωk to obtain the radiances L(s j , ωk ). Using
Eq. 4.14 this gives an estimate of the diffusion term.
7. Finally, the direct transmission term and the diffusion term are added to get the
radiance emergent at the surface L(s).
The number of samples chosen are often N = 1 and M = 1. Then, we get a very
noisy sample image rather quickly. Another sample is then rendered and this is
averaged with the previous one. The next sample is weighted by one-third and added
to the other two samples which are weighted by two-thirds and so on. In this way the
image will improve itself over time. Even so, the Monte Carlo path tracing procedure
spawns a formidable number of rays. It is very slow, but it converges to the intended
result in an unbiased manner. This is nice in a predictive rendering context, where
we want to predict the appearance of real-world materials.
In order to predict the result of taking a picture with a digital camera, we need to
digitally model the 3D scenery that is going to be in the picture as explained in
Sect. 4.2.2. For example, in order to compute a realistic image of a glass of milk,
we need a geometric model of the glass and the milk inside it. The most practical
way of modelling such geometry is to use a computer aided design (CAD) system
(e.g. AutoCAD® or Pro/ENGINEER® ) or a 3D modelling system (e.g. 3ds Max® ,
Softimage® or Blender™ ). A scene modelled using one of these systems can be
exported to a text file and imported into a rendering system which implements a
camera model (see Sect. 4.2.1) and the models for light propagation and scattering
described in Sects. 4.2.4 and 4.2.5. This could be our own rendering system or another
physically based rendering system such as Maxwell Render™ or Indigo Renderer.
Once the geometry of the scene is in place, we need information about the materials
and the light sources in the scene. More specifically, we need the optical properties
of the materials and the emission profiles of the light sources.
To acquire the optical properties of a material, one option is to measure them.
Another option is to compute them from the particle contents of the material. The
second option is made possible by the Lorenz–Mie theory, and it provides us with
a very flexible way of modelling how the appearance of a material changes when
we change its contents. Nevertheless, the Lorenz–Mie theory has only reluctantly
been adopted in graphics. In the two papers [24, 25] where the theory was first
considered for graphics applications, it was found to be either too complicated [24]
or too restricted [25] to be useful. The problematic restrictions are that the mate-
112 J. R. Frisvad et al.
The input parameters for the radiative transfer equation (Eq. 4.1) are the phase func-
tion p(x, ω , ω), the scattering coefficient σs (x) and the extinction coefficient σt (x)
(or the absorption coefficient since σt = σa + σs ). Together with the index of refrac-
tion n bulk (x), these parameters constitute the optical properties of a material. In the
following, we will omit the dependency on the position x in the material. Then we
just have to remember that this dependency should be inserted if the material is
heterogeneous.
The phase function and the scattering coefficient are collectively referred to as
the scattering properties of the material. The direction of the incoming light ω is
called the forward direction while ω is the direction of the scattered light. The plane
spanned by these two vectors is called the scattering plane and the angle between
them θ is called the scattering angle.
Scattering under the surface of a material is typically caused by particles. For
simplicity, let us assume that the particles are small, randomly distributed throughout
the material, not too densely packed, and approximately spherical. A particle is
considered small when it is not directly visible to the human eye from the distance
that it is observed. If every particle were visible, we would have to model the surface
of each individual particle. If the particles were not randomly distributed, we would
again need to make a very precise model taking into account the ordered placement
of the particles in the material. When we assume that the particles are not too densely
packed, we assume that the distance between them is considerably larger than the
wavelength of the light. Under this assumption, we say that the particles scatter light
independently of each other. The assumption about approximately spherical particles
ensures that scattering is symmetric around the forward direction and that the two
polarisation components do not affect each other. In other words, we only need two
scattering components to describe the scattering of a spherical particle: S1 (θ ) for
the polarisation component with the electric vector perpendicular to the scattering
4 Predicting the Appearance of Materials 113
plane and S2 (θ ) for the polarisation component with the electric vector parallel to
the scattering plane. For unpolarised light, these two scattering components define
the phase function of a single particle by
where Cs is the scattering cross section of the particle, k = 2π n med /λ is the wave
number and n med is the refractive index of the host medium. If the host medium is
absorbing, n med is a complex number. The scattering components (S1 and S2 ) and
the scattering cross section of a particle change depending on the radius r of the
particle. To indicate this, we let pr denote the phase function of a single particle of
radius r .
The scattering cross section Cs of a particle is the area that would receive the same
amount of energy as the particle scatters if we subtend it normal to the incident light.
If we multiply the scattering cross section of a particle with the number density of
this type of particle in the material, we get the amount of light that will be scattered
away from a ray of light per unit distance as it propagates through the material, that
is, we get the scattering coefficient σs . Since the particles may have many sizes with
different cross sections and scattering components, the scattering coefficient is an
integral over particle radii r :
rmax
σs = Cs (r )N (r ) dr. (4.16)
rmin
As the equation shows, another way to find the ensemble phase function is as a
scattering cross section weighted average of the single particle phase functions pr
(defined by Eq. 4.15).
Now we have an idea about how to find the scattering due to one type of particle in
a medium. To deal with several different types of particles, we let A denote the set of
homogeneous substances appearing as particles in the host medium. Then pi , σs,i and
114 J. R. Frisvad et al.
σt,i denote the phase function and the scattering and extinction coefficients for every
individual particle inclusion i ∈ A. Once the phase function has been determined for
each individual particle inclusion, the ensemble phase function is computed using a
scattering coefficient weighted average:
1
p(θ ) = σs,i pi (θ ). (4.17)
σs
i∈A
A more exact option is to use a multi lobed phase function where the Henyey-
Greenstein function replaces pi (θ ) in Eq. 4.17. The asymmetry parameter is defined
by the integral over all solid angles of the cosine weighted phase function:
g= p(θ ) cos θ dω.
4π
If the asymmetry parameter is computed for single particles and individual parti-
cle inclusions, we can compute the ensemble asymmetry parameter as a weighted
average using the same weights as for the ensemble phase function.
Because we assume that particles scatter light independently, not only scattering
cross sections are additive (see Eq. 4.16), but also scattering coefficients (and extinc-
tion coefficients) are additive. Finding the bulk scattering coefficient is straightfor-
ward:
σs = σs,i .
i∈A
Note that volume fractions are not included in this formula, because they are a part
of the number density distributions.
In a transparent medium, the extinction coefficient is defined by an equivalent
sum, but in an absorbing medium an important correction must be made. Since the
host medium is a part of the extinction process, a non-absorbing particle will reduce
the extinction of the bulk medium. This means that the extinction cross sections can
be negative [30]. The extinction cross section resulting from the Lorenz–Mie theory
is, in other words, relative to the absorption of the host medium and the necessary
correction is to include the host medium absorption σa,med in the sum. For this
purpose, we compute the bulk extinction coefficient for particles in an absorbing
medium by
4 Predicting the Appearance of Materials 115
σt = σa,med + σt,i ,
i∈A
rmax
Si,r,λ (0)
Re(n bulk (λ)) = Re(n med (λ)) + λ Im Ni (r ) dr,
k2
i∈A rmin
where Si,r,λ (0) = S1 (0) = S2 (0) is the amplitude in the forward direction of the
wave of wavelength λ scattered by a type i particle of radius r , Ni (r ) is the number
density distribution and k is the wave number. The imaginary part is found by its
relation to the bulk absorption coefficient:
λ
Im(n bulk (λ)) = σa (λ) . (4.18)
4π
Lorenz [32] and Mie [33] showed that when the scattering components are expanded
using spherical functions they are defined, for a homogeneous plane wave, by [31–34]
∞
2n + 1
S1 (θ ) = (an πn (cos θ ) + bn τn (cos θ )) (4.19)
n(n + 1)
n=1
∞
2n + 1
S2 (θ ) = (an τn (cos θ ) + bn πn (cos θ )) , (4.20)
n(n + 1)
n=1
where the functions πn and τn are related to the Legendre polynomials Pn as follows:
where the primes denote derivative. The spherical functions ψn (z) and ζn (z) are
known as Riccati-Bessel functions. They are related to the spherical Bessel functions
jn (z) and yn (z) as follows:
ψn (z) = z jn (z)
ζn (z) = z( jn (z) − i yn (z)).
The argument z is an arbitrary complex number, the arguments x and y used for the
Lorenz–Mie coefficients are related to particle and host media as follows:
Here An (z) and Bn (z) denote the logarithmic derivatives of ψn (z) and ζn (z), respec-
tively:
ψ (z) ζ (z)
An (z) = n and Bn (z) = n .
ψn (z) ζn (z)
The ratio An is only numerically stable with downward recurrence. Therefore, the
following formula is employed for its evaluation [37]
−1
n+1 n+1
An (z) = − + An+1 (z) . (4.25)
z z
This formula is also valid for the ratio Bn , but then it is unfortunately unstable for
both upward and downward recurrences [40]. Instead, we use a different formula for
Bn which has been developed by Mackowski et al. [41] in the field of multilayered
particles embedded in a non-absorbing medium. It is numerically stable with upward
recurrence for any complex argument [41]:
i
Bn (z) = An (z) + (4.26)
ψn (z)ζn (z)
n n
ψn (z)ζn (z) = ψn−1 (z)ζn−1 (z) − An−1 (z) − Bn−1 (z) . (4.27)
z z
It remains to give a recurrence relation for the ratio ψn (z)/ζn (z) in Eqs. 4.23 and 4.24.
Recent developments in the context of multilayered particles, provide a recurrence
relation that works well for small Im(z) [42, 43]:
Once A0 (z), . . . , A M (z) have been computed for both z = x and z = y, we are
able to find the ratios Bn (x) and ψn (x)/ζn (x) as well as the Lorenz–Mie coefficients,
an and bn , step by step. Note that there is no need to store Bn (x) and ψn (x)/ζn (x)
since they are computed using upward recurrences (4.26– 4.28). These recurrences
should be initialised by
B0 (z) = i
ψ0 (z)ζ0 (z) = 21 1 − ei2z
ψ0 (z)/ζ0 (z) = 21 1 − e−i2z .
Recall that there is a direct relationship between wavelength λ and the size parame-
ters x and y. This tells us that the Lorenz–Mie coefficients are spectrally dependent
and should preferably be sampled at different wavelengths. They also depend on
the particle radius r and are valid for spherical particles of arbitrary size as long
as they do not exhibit diffuse reflection (which is only possible if the particle size
greatly exceeds the wavelength and, even so, the surface of the particle might still be
smooth) [31]. Furthermore, the equations provided in this section reveal that the com-
plex refractive index of each particle inclusion, as well as that of the host medium,
are needed as input parameters for computing the optical properties of a scattering
material.
This robust way of computing the Lorenz–Mie coefficients enables us to compute
the scattering amplitudes S1 and S2 (using Eqs. 4.19 and 4.20). With these, we are
able to find the extinction and scattering cross sections as well as the phase function
of the particle. These are all well-defined quantities for particles in a non-absorbing
medium. For a particle in an absorbing medium, the scattering cross section is a
problematic quantity because the resulting formula depends on the distance to the
observer.
When particles are embedded in an absorbing host, the extinction cross section
Ct is the only well-defined observable quantity [30]. It is computed using an optical
theorem first presented by van de Hulst [31, 44]. The original theorem by van de
Hulst is valid for particles of arbitrary shape and size, but it only applies to a non-
absorbing host medium. To account for an absorbing host, we use a slightly modified
equation presented by Bohren and Gilra [30]:
S(0)
Ct = 4π Re , (4.29)
k2
where S(0) = S1 (0) = S2 (0) is the amplitude in the forward direction of the scattered
wave and k = 2π n med /λ is the wave number. Since the host medium was assumed
by van de Hulst to be non-absorbing, n med and therefore also k were assumed real
and moved outside the Re operator (which takes the real part of a complex number).
This is not allowed if the host medium is absorbing as the result would be a mean-
ingless complex extinction coefficient. Correction by discarding the imaginary part
of the result would not be a good approximation (except when particle absorption is
considerably stronger than that of the host medium [30]). Inserting the expression
4 Predicting the Appearance of Materials 119
∞
λ2
an + bn
Ct = (2n + 1)Re .
2π
n=1
n 2med
A form has not been found for the scattering cross section Cs which is independent
of the distance to the observer, but we still have to approximate Cs to evaluate the
radiative transfer Eq. (4.1). We use a far-field approximation which has been reported
to be consistent with measured data [45, 46]. The chosen formula is identical to the
scattering cross section for transparent media except for two correction terms: an
exponential term and a geometrical term γ . The formula is
∞
λ2 e−4πr Im(n med )/λ
Cs = (2n + 1) |a n | 2
+ |bn | 2
, (4.30)
2π γ |n med |2
n=1
where r in the exponential term is the uncertain part of the equation because it ought
to be the distance to where the scattered wave is observed. This distance is unknown,
and consequently it has been projected to the particle surface, such that r denotes
the particle radius.
The geometrical term γ accounts for the fact that the incident wave changes over
the surface of the particle as a consequence of the absorbing host medium. It is
defined by [39]
2(1 + (α − 1)eα )
γ = , (4.31)
α2
where α = 4πr Im(n med )/λ and γ → 1 for α → 0. Note that α is 0 when the medium
is transparent and close to 0 for small particles in a weakly absorbing medium.
To avoid numerical errors, one should use γ = 1 for α < 10−6 .
The precision of the far-field approximation (4.30, 4.31) has recently been
reviewed [47] and compared to experimental data [45, 46]. The conclusion is that it
(as expected) does not give entirely accurate results, but it does give physically plau-
sible results. It is also concluded that significant errors can result if the absorption
of the host medium is ignored (this is especially true when the size parameter x is
large).
In the same way that it is possible to formulate expressions for the scattering and
extinction cross sections of a spherical particle using Lorenz–Mie theory, it is also
possible to derive the following formula for the asymmetry parameter of a single
spherical particle [31]:
∞ n(n+2) ∗ ∗ ) + 2n+1 Re(a b∗ )
n=1 n+1 Re(an an+1 + bn bn+1 n(n+1) n n
g= ∞ ,
n=1 (2n + 1) |an | + |bn |
1 2 2
2
This concludes the robust scheme for computing the scattering properties of a
sphere in a host medium. When using the Lorenz–Mie theory to compute macroscopic
scattering properties of a medium (as described in Sect. 4.3.1), some information
about the particle shapes and sizes is needed. We will look at this in Sects. 4.3.3 and
4.3.4.
Particle size distribution is the common term for distributions that we can use to find
the number densities of particles of different sizes. One type of size distribution,
which is often encountered in the literature, is the volume frequency distribution
r 3 N (r ). Such distributions typically follow a log-normal distribution. Log-normal
distributions are often described by a mean particle size μ and a coefficient of vari-
ation cv = σ/μ, where σ is the standard deviation.
If we find that the volume frequency of some type of particle in a medium follows
the log-normal distribution with mean value μ and standard deviation σ , the volume
frequency distribution is given by
2
1 ln r −α
− 21
r N (r ) =
3
√ e β
, (4.32)
rβ 2π
2
1 σ σ2
α = ln μ − ln +1 and β = ln +1 .
2 μ2 μ2
N (r ) = N∗r −α ,
rmax
4π
v= r 3 N (r ) dr. (4.33)
3
rmin
Suppose we measure the particle size distributions for some sample of material. Then
we would have empirical functions or tabulated data that fit the volume fractions of the
particles in the original sample. Most probably the original volume fractions are not
the volume fractions we desire in our medium. Equation (4.33) is important because
it explains how we find the original volume fraction voriginal,i of particle type i.
If the volume fraction vi is desired rather than voriginal,i , the measured number density
distribution should be scaled by vi /voriginal,i .
Particles are not always spherical. There exist a number of theories for the scattering
of other perfect mathematical shapes like cylinders, hexagonal columns and plates,
etc. They are useful because some materials actually do consist of particles approxi-
mately of these shapes. Halos are, for example, the result of scattering by hexagonal
ice crystals in the atmosphere. Instead of the mathematical approach, let us use a
more practical approach for non-spherical particles.
With the theory we have already developed, we are able to approximate non-
spherical particles by an appropriate collection of spherical particles. It is not obvious
what set of spheres we should choose to model a non-spherical particle in the best
way. Many different concepts have been tried: equal-volume spheres, equal-area
spheres, etc. The best approach we are aware of is that of Grenfell and Warren [48].
They use volume-to-area equivalent spheres. As opposed to equal-volume and equal-
area spheres, the volume-to-area equivalent spheres have proven to be quite exact.
They have been tested for cylinders [48], hexagonal columns and plates [49], and
hollow columns and plates [50]. In most cases the error is less than 5 %. At least this
is true for scattering and extinction coefficients. The approximation is, as could be
expected, less accurate with respect to the phase function.
To represent a particle of volume V and surface area A by a collection of spheres,
the radius of the equivalent spheres is found simply using the volume to surface area
ratio of a sphere [48]:
V
req = 3 .
A
Since the number of equivalent spheres is not equal to the number of non-spherical
particles, the number density must be adjusted accordingly [48]:
Neq 3V
= 3
.
N 4πreq
122 J. R. Frisvad et al.
The equivalent radius req and the equivalent number density Neq are then used for
computing the optical properties of the material with Lorenz–Mie theory for com-
puting the cross sections of the equivalent spheres. This is a simple and practical
approach which gives rather good results. It has been used with the theory presented
here to develop an appearance model for natural ice. which contains both cylindri-
cally and spheroidally shaped particles [28, 51]. In the milk case study (Sect. 4.4),
the volume-to-area equivalent size distribution is reported directly in the literature,
so the radius and number density adjustments are not necessary.
In wave optics, a generalised version of the Lorenz–Mie theory [52, 53] would be
needed to model the scattering by a particle of a shaped beam such as laser. However,
we are using the scattering by particles in a geometrical optics context. The scattering
of a ray, which traces an infinitely thin part of the wavefront, is adequately modelled
by the scattering of a plane wave. Thus we do not employ the generalised Lorenz–Mie
theory, but model a Gaussian beam (a laser source) as described in Sect. 4.2.3.
When we render an image using the algorithm described in Sect. 4.2.5, the laser
source poses a problem. Rays are traced from the observer and new directions are
sampled at every scattering event. Since the laser source is small and collimated,
the chance of a ray hitting the laser source from the right direction is almost non-
existing. To solve this problem, we use bidirectional path tracing [54]. As described
in Sect. 4.2, rays are traced from the observer through each pixel in the image into
the scene. To account for a laser source, we also sample a position on the laser
source and trace a ray from this position to the first scattering event in a medium. For
every ray from the observer that reaches this medium, we compute the contribution
from the scattered laser light. Every time all pixels have been sampled once by rays
from the observer, a new sample ray is traced from the laser source. This is how we
accommodate a laser source in the path tracing algorithm. In the following, we will
use the theory to predict the appearance of milk.
Table 4.1 The coefficients for the empirical formula (4.34) by Quan and Fry [57]
n 1 = 1.779 · 10−4 n 4 = −2.02 · 10−6 n 7 = −0.00423
n 2 = −1.05 · 10−6 n 5 = 15.868 n 8 = −4382
n 3 = 1.6 · 10−8 n 6 = 0.01155 n 9 = 1.1455 · 106
The host medium is water in which many different components are dissolved. For
the real part of the refractive index of the milk host, we use the refractive index for
pure fresh water (S = 0‰ in Eq. 4.34).
Quan and Fry [57] have developed an empirical formula for computing the real part
of the refractive index of pure water or brine as a function of salinity S, temperature
T , and wavelength λ. It is as follows [57]:
1 http://omlc.ogi.edu/spectra/PhotochemCAD/abs_html/riboflavin.html.
124 J. R. Frisvad et al.
Table 4.2 Imaginary part of the refractive index for pure water [61], riboflavin [59] (0.17 mg pr.
100 g solution), milk host n milk , and milk fat n fat
λ(nm) n water n riboflavin n milk n fat
375 3.393 · 10−10 2.927 · 10−7 2.93 · 10−7 4.0 · 10−6
400 2.110 · 10−10 2.603 · 10−7 2.60 · 10−7 6.4 · 10−6
25 1.617 3.363 3.36 8.6
50 3.302 4.096 4.10 1.1 · 10−5
75 4.309 3.323 3.33 1.1
500 8.117 · 10−10 1.076 · 10−7 1.08 · 10−7 1.0 · 10−5
25 1.742 · 10−9 5.470 · 10−8 5.64 · 10−8 4.7 · 10−6
50 2.473 5.772 6.02 4.6
75 3.532 7.554 7.91 4.7
600 1.062 · 10−8 6.889 · 10−8 7.95 · 10−8 4.9 · 10−6
25 1.410 7.169 8.58 5.0
50 1.759 7.563 9.32 5.0
75 2.406 4.967 7.37 5.1
700 3.476 · 10−8 7.937 · 10−8 1.14 · 10−7 5.2 · 10−6
25 8.591 4.683 1.33 5.2
50 1.474 · 10−7 7.287 2.20 5.2
775 1.486 · 10−7 8.626 · 10−8 2.35 · 10−7 5.2 · 10−6
The milk host spectrum is n water + n riboflavin . The milk fat spectrum is from Michalski et al. [62]
ε(266.5 nm)
ε(λ) = ln(10)D(λ).
D(266.5 nm)
The natural content in milk of riboflavin is 0.17 mg per 100 g milk [56]. Using the
molar mass of riboflavin which is 376.3682 g/mol, we find that the natural concen-
tration of riboflavin in milk is
0.17 mg 100 g
c= = 4.65 · 10−6 mol/L.
376.3682 g/mol 1.03 g/mL
Walstra and Jenness [63] have found experimentally that the real part of the refractive
index of milk fat approximately follows the function
(b(T ) + 2)λ2 − 0.03
n fat (λ, T ) = , (4.35)
(b(T ) − 1)λ2 − 0.03
4 Predicting the Appearance of Materials 125
where r43,fat is measured in micrometres. The relationship between r43,fat and req,fat
is [64]
req,fat = r43,fat /(cv,fat
2
+ 1).
The radius r43,fat is used since it can be estimated empirically with good accuracy [64].
The coefficient of variation cv,fat is usually between 0.4 and 1.2 in normal milk.
Reasonable limits for the range of fat globule radii are rmin,fat = 0.005 μm and
rmax,fat = 10 μm.
The refractive index of casein micelles is not readily available in the literature. For
comparison to goat’s milk it has been determined to be the following for cow’s
milk [66]:
n casein = 1.503.
This value is assumed to be constant in the visible range and absorption of the casein
micelles is neglected.
Structure and size distribution of casein micelles is still being disputed in the
literature. Recent research on the matter is discussed by Gebhardt et al. [67]. Most
investigations are based on either light scattering or electron microscopy. Light scat-
tering approaches find micelles of large average size while electron microscopy
report a large number of very small casein particles in addition to the larger micelles.
Sometimes, these very small particles are excluded from the reported size distribu-
tion since they are regarded to represent non-micellar casein or single submicelles.
No matter what we call these very small particles, they scatter light as do the larger
126 J. R. Frisvad et al.
Table 4.4 Densities for computing of milk fat and casein volume fractions using weight percents
ρfat ρprotein ρmilk
1.11 g/mL 0.915 g/mL 1.03 g/mL
Measured by Walstra and Jenness [63] at 20 ◦ C
aggregates and therefore should be included in the size distribution employed for the
Lorenz–Mie calculations.
A size distribution based on electron microscopy, which includes the single
submicelles in the distribution, was reported by Schmidt et al. [68]. They found
the mean req,casein = 43 nm and showed that a log-normal distribution (4.32) of
r/(rmax,casein − r ) is a good fit of the measured volume frequency distribution. The
limits for the casein micelle radii are rmin,casein = 0 nm and rmax,casein = 150 nm.
The microscopic properties of milk are summarised in Table 4.3.
To model the concentration of fat and protein we use wt% (g per 100 g milk), since this
value is used on content declarations on the side of milk cartons. In the remainder
of this chapter we let w f and w p denote the wt% of fat and protein respectively.
To translate wt% into volume fractions, we use the densities given by Walstra and
Jenness [63]. They are summarised in Table 4.4. Casein micelles make up about
80% · 95% = 76% of the protein volume fraction (see above). This means that
w f /ρfat w p /ρprotein
vfat = and vcasein = 0.76 .
100 g/ρmilk 100 g/ρmilk
This simple translation from fat and protein contents to volume fractions of the
particle inclusions in the milk means that we have an appearance model with the
following parameters:
• Fat content w f
• Protein content w p .
These two parameters are all we need to model most types of milk.
4 Predicting the Appearance of Materials 127
Fig. 4.1 Rendered images of the components in milk (top row) as well as mixed concentrations
(bottom row). From top left to bottom right the glasses contain: pure water, water and vitamin B2,
water and protein, water and fat, skimmed milk, regular milk, and whole milk
4.4.3 Results
The protein content of the milk we buy in a grocery store is usually around w p =
3.4 wt.% while the content of fat is what we use to distinguish between different
milk products: skimmed milk, w f = 0.1 wt.%; low fat milk, w f = 1.5 wt.%; whole
milk w f = 3.5 wt.%. This information, and the appearance model described above,
enables us to visualise different types of milk. We are also able to show the visual
significance of each component in the milk. This is a particular strength of the
approach that we take. Knowing the visual significance of the different ingredients
in a material is important if we would like to design the appearance of the material
or if we would like to interpret the appearance of the material. Figure 4.1 shows
the visual significance of different components in milk as well as the predicted
appearance of different milk products. Starting leftmost in the top row, the first glass
128 J. R. Frisvad et al.
contains pure water, the second includes the absorption of vitamin B2 and is the
host medium of the particles in the milk, the third glass contains casein micelles in
water, the fourth contains fat globules in water, and the three glasses in the bottom
row contain skimmed milk, low fat milk and whole milk, respectively. The contents
of these glasses have all been rendered using the appearance model described in
this section. Spectral optical properties (sampled at every 25 nm) were computed
as described in Sect. 4.3 for all the milk materials. For rendering, we computed
RGB representations of the optical properties by weighted averages using the RGB
colour matching functions of a standard human observer [69]. We used homogeneous
scattering properties for the milk and the approximative Henyey-Greenstein phase
function with an ensemble asymmetry parameter.
The results provided in Fig. 4.1 (bottom row) compare only qualitatively to pic-
tures of real milk. Our eyes are the instruments in such a comparison. To find out if
our model correctly predicts the appearance of real milk, some sort of quantitative
comparison is necessary. We can do a quantitative comparison by constructing a sim-
ple experimental setup which we can easily model and render. The set-up we choose
is photographed in Fig. 4.2. It consists of a digital camera on a tripod pointing at the
surface of milk in a cup, and some arbitrary device for hanging up a laser pointer over
the cup. The experiment is to take a picture of the milk while the laser pointer shines
light into it directly from above. The room should be darkened as much as possible,
but it is usually not possible nor practical to black it out completely. For this reason,
a picture was first taken with the laser pointer turned off and one was then taken with
the laser pointer turned on. Subtracting the first picture from the second removes the
background illumination to reveal the scattering of the laser in the milk alone. To
avoid scattering of light in the cup itself, the white cup in Fig. 4.2 was replaced by a
black, opaque cup when the pictures were taken. In addition, the digital camera was
and should be configured to process the picture as little as possible.
To model the experimental set-up digitally, all we need is a correctly sized cylinder
of milk, a laser source placed directly above it, and a camera placed at the right
distance from the milk surface. The laser pointer we used had a diameter of 3 mm,
power of Φ = 1 mW, and wavelength of λ = 650 nm. This means that all the laser
4 Predicting the Appearance of Materials 129
Laser in skimmed milk − photo Laser in skimmed milk − computed Diffuse reflectance: photo (blue), computed (green)
1
0.9
0.8
0.7
pixel value
0.6
0.5
0.4
0.3
0.2
0.1
0
0 200 400 600 800
pixel distance
Fig. 4.3 Photographed laser in skimmed milk (w f = 0.1 wt.%, w p = 3.4 wt.%), simulated laser in
skimmed milk, and comparison of the diffusion curves, that is, the middle lines through the images
light should go into the red colour band. The horizontal line in the image through
the laser spot centre is referred to as the diffuse reflectance curve. We only use the
diffuse reflectance curve of the red colour band in our quantitative comparison. This
line is perpendicular to the direction towards the camera, so only the distance to the
camera is important not the angle. For this reason, we place both camera and laser
pointer directly above the milk in our digital scene. To save computations, we only
render pixels from the laser spot centre to the right border of the film (half the diffuse
reflectance curve), and then we revolve this result around the centre to get a full
synthesised image of the laser spot.
The photographed milk, the rendered milk, and the quantitative comparison are
provided in Fig. 4.3. There are a few differences between theory and experiment
that we should discuss. The real camera is overexposed by the reflected laser light
while the synthetic camera is not (the top of the blue curve has been cut off). The
reason is the background illumination which was subtracted. In this experiment,
the background illumination was just enough to overexpose the laser spot in the
photograph. Another difference is that the photographed laser spot is not symmetric.
It is more elliptically shaped. The reason for this deviation in the rendered image
is that the laser source was modelled as a circular disc (see Sect. 4.2.3), and this is
not the shape of the light emitted by a standard laser pointer. Finally, the central
part of the photographed laser spot is not only red. The most likely explanation is
that light penetrates the colour filters in the CCD chip when it is overexposed. This
means that some of the red light spills into the other colour bands in the overexposed
area of the image. Apart from these explainable differences, there is surprisingly
good quantitative agreement between theory and experiment (the diffuse reflectance
curves are close to each other).
From a qualitative perspective, the rotational symmetry of the simulated image
is a giveaway which makes it easy for the human eye to spot the synthetic image in
comparison to the real one. This might not be as easy if we had not revolved a single
line in the image around the centre to save computations. Although it is not physically
130 J. R. Frisvad et al.
Laser in whole milk − photo Laser in whole milk − computed Diffuse reflectance: photo (blue), computed (green)
0.8
pixel value
0.6
0.4
0.2
0
0 200 400 600 800
pixel distance
Fig. 4.4 Photographed laser in whole milk (w f = 3.5 wt.%, w p = 3.4 wt.%), simulated laser in
whole milk, and comparison of the diffusion curves, that is, the middle lines through the images
accurate, the red light spilling into the other colour bands where the light is intense
in the photograph convinces the eye that the photograph is more real. This is because
the human eye is able to perceive a more intense red colour than what the camera
can capture (and what a standard computer screen can display). The added intensity
from the other colour bands therefore seems more realistic than a flat red colour. As
a final example, to illustrate that the appearance model is reliable for different types
of milk, Fig. 4.4 compares photographed and simulated laser in whole milk.
4.5 Conclusion
The milk example demonstrates that we are able to show the visual significance
of the different components in a material by computing synthetic images. It also
demonstrates that we are able to predict the appearance for various ratios between
the contents. This makes a number of interesting applications possible:
• If you want to design materials with a specific appearance, the appearance model
can help you choose the right components to obtain the desired appearance.
• If you want to detect whether a component is present in a material or not, the
appearance model can help you visualise the material as it would look with and
without the component.
In other words, synthesised images with a connection to the contents and the physical
conditions of a material enable us to learn a lot about the reasons for the appearance
of materials.
Considering the input parameters that brought us to the simulated diffuse
reflectance curves in Figs. 4.3 and 4.4, the presented method is definitely useful
for analysis by synthesis. If we simulate a large number of diffuse reflectance curves
for milks with different particle inclusions, we can construct numerical methods for
analysing the properties of these inclusions from photographed diffuse reflectance
4 Predicting the Appearance of Materials 131
curves. The slope of the diffuse reflectance curve, for example, reveals a lot about
the fat content of the milk (as is obvious from the figures). Further investigation can
lead to more precise models for determining the fat content of milk from a digital
photograph.
The Lorenz–Mie theory provides the link from the particle composition of a mate-
rial to the macroscopic scattering properties that we can use with radiative transfer
theory to compute a realistic image. In this way, a century after its formulation, the
Lorenz–Mie theory is the key component when we want to predict the appearance
of materials using a physical model.
References
1. R.A. Hall, D.P. Greenberg, IEEE Comput. Graph. Appl. 3(8), 10 (1983)
2. A.S. Glassner, Principles of Digital Image Synthesis (Morgan Kaufmann Publishers, Inc., San
Francisco, California, 1995). Two-volume set
3. S. Chandrasekhar, Radiative Transfer (Oxford, Clarendon Press, 1950). Unabridged and
slightly revised version published by Dover Publications, Inc., in 1960
4. R. Siegel, J.R. Howell, Thermal Radiation Heat Transfer, 4th edn. (Taylor & Francis, New
York, 2002)
5. M. Pharr, G. Humphreys, Physically Based Rendering: From Theory to Implementation
(Morgan Kaufmann Publishers, an imprint of Elsevier Inc., 2004)
6. B.T. Phong, Commun. ACM 18(6), 311 (1975)
7. R.V. Klassen, ACM Trans. Graph. 6(3), 215 (1987)
8. D. Jackèl, B. Walter, Comput. Graph. Forum 16(4), 201 (1997)
9. A.J. Preetham, P. Shirley, B. Smits, in Proceedings of ACM SIGGRAPH 1999 (ACM Press,
1999), pp. 91–100
10. L.V. Wang, H.I. Wu, Biomedical Optics: Principles and Imaging (John Wiley & Sons, Inc.,
Hoboken, New Jersey, 2007)
11. A. Sommerfeld, J. Runge, Annalen der Physik 340, 277 (1911)
12. P.d. Fermat, OEuvres de Fermat: Correspondance, vol. 2 (Gauthier-Villars et fils, 1894)
13. A. Glassner, IEEE Comput. Graph. Appl. 4(10), 15 (1984)
14. F.W. Jansen, in Data Structures for Raster Graphics: Proceedings of a Workshop Held at
Steensel, the Netherlands, from 24–28 June 1985. Eurographics seminars, (Springer, 1986),
pp. 57–73
15. A. Fujimoto, T. Tanaka, K. Iwata, IEEE Comput. Graph. Appl. 6(4), 16 (1986)
16. K. Sung, P. Shirley, in Graphics Gems III, ed. by D. Kirk (Academic Press, 1992), pp. 271–274
17. T. Möller, B. Trumbore, J. Graph. Tools 2(1), 21 (1997)
18. A. Fresnel, Mémoires de l’Académie des sciences de l’Institut de France 11, 393 (1832).
Presented 7 January 1823
19. T. Whitted, Commun. ACM 23(6), 343 (1980). Presented at SIGGRAPH 79
20. R.W. Preisendorfer, Radiative Transfer on Discrete Spaces (Pergamon Press, 1965)
21. A. Ishimaru, Wave Propagation and Scattering in Random Media (Academic Press, New York,
1978). Reissued by IEEE Press and Oxford University Press 1997
22. J.T. Kajiya, B.P. Von Herzen, Computer graphics (Proc. ACM SIGGRAPH 84) 18(3), 165
(1984)
23. S.N. Pattanaik, S.P. Mudur, J. Vis. Comput. Animat. 4(3), 133 (1993)
24. T. Nishita, Y. Miyawaki, E. Nakamae, Computer graphics (Proc. ACM SIGGRAPH 87) 21(4),
303 (1987)
132 J. R. Frisvad et al.
25. H. Rushmeier, in Realistic Input for Realistic Images. ACM Press (ACM SIGGRAPH 95 Course
Notes, 1995). Also appeared in the ACM SIGGRAPH 98 Course Notes—A Basic Guide to
Global Illumination
26. P. Callet, Comput. Graph. Forum 15(2), 119 (1996)
27. T. Nishita, Y. Dobashi, in Proceedings of Computer Graphics International 2001 (IEEE Com-
puter Society, 2001), pp. 149–156
28. J.R. Frisvad, N.J. Christensen, H.W. Jensen, ACM Trans. Graph. 26(3) (2007). Article 60
29. L.G. Henyey, J.L. Greenstein, Annales d’Astrophysique, 3, 117, (1940) Also in. Astrophys. J.
93, (1941)
30. C.F. Bohren, D.P. Gilra, J. Colloid Interface Sci. 72(2), 215 (1979)
31. H.C. van de Hulst, Light Scattering by Small Particles (John Wiley & Sons, Inc., New York,
1957), Unabridged and corrected version of the work published by Dover Publications, Inc.,
in 1981
32. L. Lorenz, Det kongelig danske Videnskabernes Selskabs Skrifter 6(1) (6. Række, naturviden-
skabeligog mathematisk Afdeling, 1890), pp. 2–62
33. G. Mie, Annalen der Physik 25(3), 377 (1908). IV. Folge
34. M. Kerker, The Scattering of Light and Other Electromagnetic Radiation (Academic Press,
New York, 1969)
35. J.V. Dave, IBM J. Res. Dev. 13(3), 302 (1969)
36. C.F. Bohren, D.R. Huffman, Absorption and Scattering of Light by Small Particles (John Wiley
& Sons, Inc., 1983)
37. G.W. Kattawar, G.N. Plass, Appl. Opt. 6(8), 1377 (1967)
38. W.J. Wiscombe, Appl. Opt. 19(9), 1505 (1980)
39. W.C. Mundy, J.A. Roux, A.M. Smith, J. Opt. Soc. Am. 64(12), 1593 (1974)
40. V.E. Cachorro, L.L. Salcedo, J. Electromagn. Waves Appl. 5(9), 913 (1991)
41. D.W. Mackowski, R.A. Altenkirch, M.P. Menguc, Appl. Opt. 29(10), 1551 (1990)
42. Z.S. Wu, Y.P. Wang, Radio Sci. 26(6), 1393 (1991)
43. W. Yang, Appl. Opt. 42(9), 1710 (2003)
44. H.C. van de Hulst, Physica 15(8–9), 740 (1949)
45. J. Randrianalisoa, D. Baillis, L. Pilon, J. Opt. Soc. Am. A 23(7), 1645 (2006)
46. J. Yin, L. Pilon, J. Opt. Soc. Am. A 23(11), 2784 (2006)
47. Q. Fu, W. Sun, J. Quant. Spectr. Radiat. Transf. 100(1–3), 137 (2006)
48. T.C. Grenfell, S.G. Warren, J. Geophys. Res. 104(D24), 31697 (1999)
49. S.P. Neshyba, T.C. Grenfell, S.G. Warren, J. Geophys. Res. 108(D15, 4448), 6 (2003)
50. T.C. Grenfell, S.P. Neshyba, S.G. Warren, J. Geophys. Res. 110(D17203), 1 (2005)
51. J.R. Frisvad, Light, Matter, and Geometry: The Cornerstones of Appearance Modelling (VDM
(Verlag Dr. Müller, 2008)
52. J.A. Lock, G. Gouesbet, J. Quant. Spectr. Radiat. Transf. 110(11), 800 (2009). Review
53. G. Gouesbet, J. Quant. Spectr. Radiat. Transf. 110(14–16), 1223 (2009). Review
54. E.P. Lafortune, Y.D. Willems, in Proceedings of the 7th Eurographics Workshop on Rendering
(1996), pp. 91–100
55. H.D. Goff, A.R. Hill, in Dairy Science and Technology Handbook: Principles and Properties,
vol. 1, ed. by Y.H. Hui (VCH Publishers, Inc., New York, 1993), Chap. 1, pp. 1–81
56. P.F. Fox, P.L.H. McSweeney, Dairy Chemistry and Biochemistry (Blackie Academic & Pro-
fessional, London, 1998)
57. X. Quan, E.S. Fry, Appl. Opt. 34(18), 3477 (1995)
58. P.D.T. Huibers, Appl. Opt. 36(16), 3785 (1997)
59. H. Du, R.C.A. Fuh, J. Li, L.A. Corkan, J.S. Lindsey, Photochem. Photobiol. 68(2), 141 (1998)
60. J. Koziol, Photochem. Photobiol. 5, 41 (1966)
61. G.M. Hale, M.R. Querry, Appl. Opt. 12(3), 555 (1973)
62. M.C. Michalski, V. Briard, F. Michel, Lait 81, 787 (2001)
63. P. Walstra, R. Jenness, Dairy Chemistry and Physics (John Wiley & Sons, New York, 1984)
64. P. Walstra, Neth. Milk Dairy J. 29, 279 (1975)
65. D.W. Olson, C.H. White, R.L. Richter, J. Dairy Sci. 87(10), 3217 (2004)
4 Predicting the Appearance of Materials 133
W. Hergert and T. Wriedt (eds.), The Mie Theory, Springer Series in 135
Optical Sciences 169, DOI: 10.1007/978-3-642-28738-1_5,
© Springer-Verlag Berlin Heidelberg 2012
136 L. K. Ausman and G. C. Schatz
5.1 Introduction
Surface enhanced Raman scattering (SERS) was discovered in the mid-1970s when
it was realised that Raman scattering from molecules like pyridine adsorbed in mono-
layer concentrations on anodised (roughened) silver electrode surfaces was enhanced
by a factor of 106 compared to the same scattering process for molecules in solution
[1–3]. Shortly after this discovery, it was postulated that this enhancement could be
understood based on the enhancement in local fields on the surfaces of the nanoparti-
cles being irradiated as a result of plasmon excitation in these particles [4]. Approx-
imate theories that include plasmon excitation were then formulated, as reviewed
in many places in the mid-1980s [5–10], in which the electromagnetic (EM) con-
tribution to the enhancement factor is given by |Eloc (ω)|2 |Eloc (ω )|2 , where Eloc is
the electric field at the position of the molecule and ω and ω are the incident and
Stokes shifted frequencies, respectively. This so-called plane wave (PW) approxima-
tion assumes that the incident and scattered field enhancements can both be derived
from the scattering of plane waves from the metal nanostructure; however, it was
noted in 1980 by Kerker [11] that this is only an approximation to the rigorous
EM enhancement factor. Kerker used concepts derived from Mie theory [12] which
he had extended in the mid-1970s to describe fluorescence from molecules inside
polymer nanoparticles [13–15] to show that although the incident photon enhance-
ment factor is given by |Eloc (ω)|2 , the emitted photon enhancement factor should
be derived from emission by the dipole induced in the adsorbed molecule at the
Stokes frequency. The emitted radiation interacts with the nanostructure, leading to
plasmon enhancement in the scattered field. Kerker demonstrated that this dipole
re-radiation (DR) enhancement factor matches the plane wave (PW) enhancement
factor for spherical particles in the quasistatic (small particle limit) approximation
for molecules in special locations. However, the more general evaluation of DR
effects has received scant attention over the years, with the few exceptions being
work done by Kerker and others in the mid-1980s [10, 16–18], and a recent study by
Ausman and Schatz [19]. Instead the vast majority of the SERS models are based on
the PW approximation [20–29]. A further restriction is often made where the differ-
ence between ω and ω is neglected, leading to the zero Stokes shift enhancement
factor |Eloc (ω)|4 .
The lack of interest in using the dipole re-radiation expression for the SERS
enhancement factor has arisen in part because of the technical difficulty associated
with evaluation of the enhancement factor (few codes provide this capability and the
computational effort is higher), and in part because Kerker’s early estimates of DR
effects [18] suggested that the PW approximation was generally adequate. However,
in the recent work by Ausman and Schatz [19], it has been found that the DR/PW
comparison can show a factor of 2–3 differences for large enough particles or for
dimers of particles when averaged over detector locations. In addition, there can be
much larger differences for specific molecule/detector locations. These differences
are typically due to interference effects that are important in the DR enhancements but
which are completely neglected by PW. These interference effects typically become
5 Dipole Re-Radiation Effects in Surface Enhanced Raman Scattering 137
important when multipole resonances are important, or when the molecule is located
in a position that does not lead to large SERS enhancement based on PW excitation.
Such effects can be significant for many kinds of SERS measurements, particularly
single molecule SERS measurements [27, 29–34] and for other measurements that
are designed to yield the highest possible SERS enhancements.
Many SERS substrates are currently being produced with the goal of being highly
sensitive and at the same time highly reproducible in applications to biomolecule
[29, 35–38] and chemical sensing [39–41]. Array structures are one substrate of
interest due to fabrication capabilities that are routinely available, along with the
added benefit that plasmon resonances in each particle can couple to give photonic
enhancement at wavelengths that can be specified by the array structure [20–25,
42–44]. Experiments on nanoparticle aggregates and array structures have generated
interest in the incorporation of dipole re-radiation into SERS models [19, 45–49];
however, the comparison between DR and PW results for these structures has not been
assessed. Therefore, knowledge concerning the limitations of the PW approximation
for these structures will aid in its proper employment and also hopefully spur new
advances in numerical methods (i.e. DDA, FDTD, etc) for describing SERS.
This chapter will highlight the incorporation of dipole re-radiation in the modeling
of SERS due to silver particles and silver particle array structures where all particles
are modeled as spheres so that Mie theory (with extensions) can be used. We will
only briefly consider applications to an isolated sphere and to a dimer of spheres as
this has been addressed in our previous work. Instead, most of our development will
refer to larger array structures, with the aim of determining the validity of the PW
approximation and to assess the conditions under which it breaks down. We do this
by considering SERS enhancements for several molecule locations and two different
detector locations. These applications will show that far-field phase interference
effects can result in SERS enhancement factors that differ significantly from the PW
approximation.
5.2 Methods
5.2.1 Theory
There is a vast literature that presents Mie theory and its extensions to coupled
spheres; however, our focus will be on the inclusion of dipole re-radiation. For
thorough treatments of single sphere plane wave scattering see Bohren and Huffman
[50]. The extension of this to multiple spheres using the T-matrix method can be found
in papers by Mackowski and Mishchenko [51–53], along with the accompanying
references.
The starting point for all of these theories is the single sphere scattering problem.
This is treated through an expansion of incident and scattered electric fields in a basis
of vector spherical harmonics. A general incident electric field has the form
138 L. K. Ausman and G. C. Schatz
l
∞
(1) (1)
Ei (r, ω) = plm Nlm (kr) + qlm Mlm (kr) , (5.1)
l=0 m=−l
l
∞
(3) (3)
Esc (r, ω) = alm Nlm (kr) + blm Mlm (kr) , (5.2)
l=0 m=−l
where Nlm and Mlm are the vector spherical harmonic basis functions that can be
found in [11, 50, 51, 54]. The coefficients alm and blm are related to the general
incident electromagnetic field coefficients plm and qlm through the Lorenz–Mie
single sphere coefficients αl and βl by the following:
Explicit expressions for the Lorenz–Mie single sphere coefficients αl and βl for a
sphere of radius a and permittivity ε1 in a medium with ε0 are
(1)
In (5.5) and (5.6) jl (x) is the spherical Bessel function and h l (x) is the spherical
Hankel function of the first kind. The coefficients plm and qlm when the incident
field is a plane wave (the traditional Lorenz–Mie theory), and the resulting scattered
field coefficients alm and blm can be found in [50], in [51] and in Ausman and
Schatz [54].
In order to describe dipole re-radiation, the coefficients plm and qlm in (5.1) are
modified from their usual plane wave expressions by making use of the fact that
the electric dipole field can be expressed as the product of the free-space tensor
Green’s function,
Gs , with the dipole
moment induced by plane wave scattering,
p = α R · Epw (r0 ) + Esc,pw (r0 ) , through the relation
4π ks2
Edp (r; ωs ) = Gs (r, r0 ; ωs ) · p. (5.7)
εs
5 Dipole Re-Radiation Effects in Surface Enhanced Raman Scattering 139
The terms in the induced dipole moment are the Raman polarisability tensor,
α R , and the local field at the molecule location, r0 , due to plane wave scattering.
The field Edp can be expressed in a vector spherical harmonic basis by noting that
the tensor Green’s function can be expanded in a basis of vector spherical harmonics
by way of the Ohm-Rayleigh method as described by Tai [55]. With the modified
plm and qlm for a molecule located at r0 , the resulting scattered field coefficients alm
and blm have the form [11, 16, 19, 54]
iks3 2l + 1 (3)
alm = (−1)m αl N (ks r0 ) · p, (5.8)
εs l(l + 1) l(−m)
iks3 2l + 1 (3)
blm = (−1)m βl M (ks r0 ) · p. (5.9)
εs l(l + 1) l(−m)
In the above equations εs is the dielectric function of the medium at the frequency
√ √
ωs , ks = ωs εs μs /c and εs μs is the complex refractive index of the medium.
Note that the expressions in (5.8) and (5.9) will generally be nonzero for all values
of m between −l and +l. In the case of plane wave scattering m will only take on the
values of ±1 which means that only specific symmetry components of any multipole
l can be induced in a sphere. This is no longer the case for the scattering of an electric
dipole field. As the l value increases the difference between accessible modes for a
plane wave compared to those of a dipole field increases.
Now we extend this theory to dipole re-radiation that arises from emission in the
presence of many interacting spheres, following the notation used in [51]. By making
use of (5.9) and (5.10), sphere i is then coupled to sphere j by use of the addition
theorem for vector spherical harmonics in the same way as plane wave scattering is
treated. These coefficients are [19, 51–53]
iks3 (3) i
i
alm = αli N ks r0 · p
εs l(−m) ⎞
∞
Ns l
⎟ (5.10)
(3) (3)
al m Alml m ks Rji + bl m Blml m ks Rji ⎟
j j
+ ⎠,
j=1 l =1 m =−l
j=i
iks3 (3) i
i
blm = βli M ks r0 · p
εs l(−m) ⎞
∞
Ns l
⎟ (5.11)
j (3) j (3) ji ⎟
+ bl m Alml m ks R
ji
+ al m Blml m ks R ⎠.
j=1 l =1 m =−l
j=i
140 L. K. Ausman and G. C. Schatz
(3) (3)
Alml m and Blml m are vector harmonic addition coefficients, expressions for which
can be found in [51–53]. The superscript (3) signifies that the coefficients are gener-
ated using spherical Hankel functions and Rji = r − r0 .
When the individual sphere coefficients are determined, the total field, Etot sc , at a
point r is generated by taking the sum of the individual scattered fields from each
sphere in Cartesian space
Ns
Esc =
tot
Eisc . (5.12)
i=1
This field is then added to the field Edp produced by the radiating dipole, and for a
sufficiently large rs is multiplied by rs / exp(ikrs ) to obtain the far-field amplitude
F R (θs , φs ) = rs (Edp + Etot
sc )/ exp(ikrs ). With this it is possible to obtain the dipole
re-radiation Raman enhancement factor, G, using
|F R (θs , φs )|2
G(θs , φs ) =
, (5.13)
F R,0 (θs , φs )
2
with F R (θs , φs ) being the far-field amplitude of the molecule/particle system and
F R,0 (θs , φs ) being the corresponding far-field amplitude of the scattered radiation
from an isolated molecule. We note that F implicitly depends on rs but we simply
assume that this distance is large with respect to the wavelength of light and fixed.
H (the y-axis). The dimer chains are similarly arranged, with the second sphere in
the dimer being offset along the axis that coincides with E. The molecule is located
along the E-axis (i.e., x-axis) a distance 0.5 nm from the surface of the sphere that
is at the origin. This is the central sphere in the chain,and the central dimer in the
chain of dimers. Note that the classical point dipole is assumed to have an isotropic
response. While there are likely to be many interesting effects due to anisotropic
responses, the choice of an isotropic α R is done for simplicity and we note that other
responses could be considered. Indeed, a quantum mechanical molecular response
can be included (still keeping the particle as a classical scatterer) by employing the
many-body theory of SERS developed by Masiello and Schatz [57].
The most extensive studies of dipole re-radiation have been done for isolated spheres.
In 1980, Kerker and coworkers examined SERS in the presence of Ag spheres and
dielectric spheres [11], considering distance dependence, frequency dependence and
sphere size dependence of the SERS enhancement factor. Later, Kerker studied sen-
sitivity of the estimated enhancement factors to the dielectric data used [17] and also
briefly compared the average enhancement over the surface of a sphere as calcu-
lated by dipole re-radiation with the average local electric field enhancement [18].
142 L. K. Ausman and G. C. Schatz
Raman Enhancement
from the sphere surface along
the axis parallel with the
incident E-field polarisation. 1×102
The solid black line is the
PW approximation result,
the dashed red line is the 1×101
backscattering DR result and
the blue dash–dotted line is
the DR result for a detector at 1×100
θs = 3π/4 and φs = π/2 350 400 450 500 550 600 650 700 750 800
Excitation Wavelength (nm)
However, Kerker fixed the detector location in these studies at a position that is
orthogonal to the kE-plane. Recently, we compared the dipole re-radiation G value
and the PW approximation for other detector locations [19]. In that work we found
that the agreement between the PW approximation and G is dependent upon both
molecule and detector location. Here, we only consider the influence of the detector
location.
In Fig. 5.2 we present an isolated sphere result so that a comparison with arrays
made up of a chain of spheres can be made later. The sphere under consideration is a
100 nm diameter Ag sphere in vacuum. We choose two different detector locations:
(1) a backscattering configuration at θs = π and (2) an off axis direction θs = 3π/4
and φs = π/2. Neither of these locations was considered in our earlier study.
The first location corresponds to a common detector direction while the second
is somewhat arbitrarily chosen but is typical of directions used for SERS measure-
ments. As was observed in [19] for other detector locations, we see that the PW
approximation is accurate for the dipole plasmon part of the spectrum but there are
significant differences for higher order multipoles. For the detector positions con-
sidered here there is no difference between the G values except in the blue region
of the spectrum. This behaviour is indicative of far-field interference effects due to
the angular structure of higher order multipoles. These effects have been explained
previously [19], and we will return to this in more detail later.
As can be seen in Fig. 5.2 the peak SERS enhancement factor from an isolated
sphere is rather small (approximately 3 × 103 ). However, the junction of a sphere
dimer can produce large local electric field enhancements that lead to considerably
larger SERS enhancement factors. Therefore, we also will briefly examine SERS
enhancements due to sphere dimers in order to make later comparisons with chains
of dimers analogous to the sphere chains. In Fig. 5.3 we present the PW enhancement
factor along with the dipole re-radiation enhancement factor at the same two detector
locations as used in the isolated sphere case for two 100 nm diameter Ag spheres
separated by 6.25 nm. It has been shown previously that the absolute maximum
enhancement factors occur when the interparticle axis of the dimer system lies parallel
5 Dipole Re-Radiation Effects in Surface Enhanced Raman Scattering 143
7
1×10
Raman Enhancement
6
1×10
5
1×10
4 4
1×10 |Eloc|
G(θs=π)
1×10
3 G(θs=3π/4, φs=π/2)
2
1×10
350 400 450 500 550 600 650 700 750 800
Excitation Wavelength (nm)
Fig. 5.3 SERS excitation profile for a dimer of 100 nm diameter Ag spheres with a gap distance
of 6.25 nm. The molecule is located in the junction between the two spheres at a distance of 0.5 nm
from the surface of the sphere that is at the coordinate origin and the inter-sphere axis lies along the
polarisation direction of the incident E-field. The solid black line is the PW approximation result,
the dashed red line is the backscattering DR result, and the blue dash-dotted line is the DR result
for a detector at θs = 3π/4 and φs = π/2
to the incident electric field polarisation. In this case we see in Fig. 5.3 that the peak
enhancement factor for the sphere dimer is around 3 × 107 which is approximately
4 orders of magnitude larger than the isolated sphere case.
For the dimer system phase interference effects in the far-field are small at the two
detector locations chosen in this work and therefore the SERS excitation profiles are
qualitatively similar to the PW enhancements. We note that this is not necessarily
true for other detector locations as has been described previously in [19]. What is
interesting about the dimer excitation profile is that the most intense peak occurs for
the quadrupole plasmon rather than the dipole plasmon, whereas the opposite is the
case for smaller spheres. This is likely due to the fact that the near-zone quadrupole
field varies as 1/r 4 which means that at close enough distances it can play a more
significant role than the electric dipole field.
Having introduced some ideas concerning the differences between dipole re-radiation
and the locally enhanced field approximation for an isolated sphere and a sphere
dimer we now turn our attention to systems that are more complex. Earlier work
has examined the role of linear arrays of spheres separated by various distances.
It was shown by Zou and Schatz [20–24] that 1D arrays with sufficiently many
spheres, specific sphere separations and sphecific choices of polarization and wave-
144 L. K. Ausman and G. C. Schatz
Qext
6
0
300 350 400 450 500 550 600 650 700 750 800
Wavelength (nm)
(b)
22
20 Single sphere
106.25nm
18 400nm
16 500nm
14
Qext
12
10
8
6
4
2
0
300 350 400 450 500 550 600 650 700 750 800
Wavelength (nm)
vector could exhibit long range interactions that can increase the isolated sphere
enhancement factor by at least one order of magnitude at specific wavelengths. These
photonic effects are characterised by a sharp peak in the extinction spectrum that
occurs at a wavelength that is roughly equal to the center-to-center particle spacing
(and higher order modes also exist). In this section, we examine the role of DR effects
for these structures.
Figure 5.4 presents the extinction spectra of 1D arrays of either 5 or 45 diameter
spheres measuring 100 nm, where the array axis is taken to be perpendicular to both
the k and E directions (and therefore is along the y-direction). The figure shows that
for short separation distances the dipole plasmon in the isolated sphere spectrum
near 390 nm is suppressed (by destructive interference of interacting dipole near-
fields) while it becomes enhanced at longer separations (due to the different dipolar
interactions for far-field interaction compared to near-field interaction). As has been
described [20–24], we see that longer separations with a larger number of spheres
can result in sharp peaks in the extinction spectrum at wavelengths that roughly
5 Dipole Re-Radiation Effects in Surface Enhanced Raman Scattering 145
Raman Enhancement
spheres with a center-to-center 1×10 3 G(θs=3π/4, φs=π/2)
separation of 106.25 nm.
The molecule is located 0.5 nm
from the surface of the sphere
at the coordinate origin along
1×10 2
the axis parallel with the
incident E-field polarisation.
The case of 5 spheres is
presented in a and the case of
45 spheres is presented in b. 1×101
In both a and b the solid black 350 400 450 500 550 600 650 700 750 800
line is the PW approximation Excitation Wavelength (nm)
result, the dashed red line is
the backscattering DR result, (b)
and the blue dash–dotted line |Eloc|
4
1×10 3
1×10 2
1×101
350 400 450 500 550 600 650 700 750 800
Excitation Wavelength (nm)
match the sphere separation. These are photonic resonance peaks, and we note that
the peaks here are not as sharp as were observed previously due to the fact that our
chains are truncated at 45 spheres (because of computational limitations), instead of
hundreds.
The trends observed in the extinction spectra are also seen in the frequency depen-
dence of the PW enhancement factor for a molecule that is located along the x-axis.
Figure 5.5 shows sharp peaks near 425 nm that do not correlate with features in the
extinction spectrum, reflecting the contributions of localised photonic enhancement
at the position of the molecule that are obscured by the more global nature of extinc-
tion. Figure 5.6 also shows a peak near 415 nm, but in this case the resonance peak
reflects a photonic resonance at that wavelength. Figure 5.7 shows similar photonic
resonance behaviour near 500 nm. The largest enhancement factors occur for sphere
to sphere separations around 400 nm. For backscattered polarisation, there is gener-
ally good agreement between DR and PW results except at very short wavelengths
where higher multipoles are expected to be important. This good agreement indicates
that for a detector that is located in a nearly symmetric position with respect to the
chain, mismatches between the phases of the scattered fields in the far-field, an effect
146 L. K. Ausman and G. C. Schatz
Raman Enhancement
Ag spheres with a center- G(θs=3π/4, φs=π/2)
to-center separation of 3
1×10
400 nm. The molecule is
located 0.5 nm from the sur-
face of the sphere at the 2
coordinate origin along the 1×10
axis parallel with the incident
E-field polarisation. The case 1
of 5 spheres is presented in 1×10
a and the case of 45 spheres
is presented in b. In both a 350 400 450 500 550 600 650 700 750 800
and b the solid black line is Excitation Wavelength (nm)
the PW approximation result,
the dashed red line is the (b)
5
backscattering DR result, and 1×10 |Eloc|
4
3
1×10
2
1×10
1
1×10
350 400 450 500 550 600 650 700 750 800
Excitation Wavelength (nm)
that we will explain later, are largely balanced and cancel out. This leaves the major
contribution to the enhancement factor being from the local environment near the
molecule.
The blue dash–dot lines in Figs. 5.5–5.7 are for the case when the detector is at
θs = 3π/4 and φs = π/2. For this detector location there are significant differences
between the PW and DR results, demonstrating far-field phase interference effects
that are significant for an array structure that is not symmetrically positioned with
respect to the detector. The enhancement factors for this detector location are gener-
ally lower than for the backscattering arrangement in the blue region of the spectrum
where the maximum enhancement occurs, and are larger in the red region of the spec-
trum if the minimum separation between the spheres in the chain is below 300 nm. The
change for longer separation distances arises because cavity-like photonic resonance
effects begin to occur in the visible region of the spectrum for separation distances
larger than 400 nm. One thing to note is that for the largest separations the local field
from each sphere closely approximates the radiation profile of a point dipole when
evaluated at the position of other spheres. In other words, the point-dipole approxi-
mation works well since the spheres are far-enough apart and the wavelengths long
5 Dipole Re-Radiation Effects in Surface Enhanced Raman Scattering 147
Raman Enhancement
Ag spheres with a center- G(θs=3π/4, φs=π/2)
to-center separation of
500 nm. The molecule is
located 0.5 nm from the sur- 2
1×10
face of the sphere at the
coordinate origin along the
axis parallel with the incident
E-field polarisation. The case 1
1×10
of 5 spheres is presented in
a and the case of 45 spheres
is presented in b. In both a 350 400 450 500 550 600 650 700 750 800
and b the solid black line is Excitation Wavelength (nm)
the PW approximation result,
the dashed red line is the (b) 4
1×10 4
backscattering DR result and |Eloc|
the blue dash–dotted line is G(θs=π)
Raman Enhancement
2
1×10
1
1×10
350 400 450 500 550 600 650 700 750 800
Excitation Wavelength (nm)
enough that higher multipolar contributions are unimportant. We will expand on this
point later when examining chains of sphere dimers.
Since the most interesting features of the excitation profiles are the maximum
enhancement factors, we now note the absolute errors in the PW approximation
for the two detector positions. For the backscattering position the errors in the
5–sphere chains are 1.74, 0.31 and 0.88% at separations of 106.25, 400 and 500 nm in
Figs. 5.5a, 5.6a, and 5.7a. The error is 0.52, 2.37 and 0.13% for these same separations
in Figs. 5.5b, 5.6b and 5.7b when the number of spheres is increased to 45. Of course
all these errors would be negligible in practical applications, which means that the
PW approximation can be used with confidence. When the detector is moved to the
second position at θs = 3π/4 and φs = π/2 the errors become considerably larger.
With the same separations as before the 5–sphere PW errors are 63.5, 416 and 64.5%
while the 45–sphere PW errors are 80.6, 16.2 and 334%. Looking at Figs. 5.5–5.7
we note that the wavelength at which the maximum enhancement occurs can change
depending on the detector location. In these results we have taken the maximum to
be the location at which the PW is largest.
148 L. K. Ausman and G. C. Schatz
The reason that we see significant differences between the PW approximation and
dipole re-radiation results can be explained by phase interference effects as follows.
We recall from Sect. 5.2.1 that the electric field in the far-field has the form
F exp(ikr )
E=
r
and therefore each field Ei for sphere i has an oscillatory term that behaves as
exp(ikrsi )/rsi , where rsi is the distance from the center of sphere i to the detector
location. Therefore, the overall far-field amplitude behaves as
rs
Ftot = F
i i
exp ik r i
s − r s ,
rs
i
where rs is defined in Fig. 5.1. The result of this is that for array structures where
spheres have larger separations there can be interference patterns in the far-field
between amplitudes of the form
Re F∗i · Fj cos k rs − rsi − I m F∗i · Fj sin k rs − rsi ,
j j
where i=j. These phase interference effects contribute to differences between the PW
and dipole re-radiation enhancement factors. In our earlier study, it was demonstrated
that phase interferences are smallest when the nanospheres are arranged symmetri-
cally with respect to the detector location such that emission from the different parti-
cles is all in-phase. This also applies in the present application to the backscattering
results, which explains the small errors we find for that geometry. In the nonsym-
metric geometry that we have studied (detector along the y-axis), the scattering from
different particles is no longer in-phase, so much larger errors are found.
As was mentioned earlier, the largest observed enhancement factors in SERS have
been found to occur at the junction between two particles. In an analogous fashion
to the construction of arrays made up of single sphere chains, arrays made of two
chains next to each other, in effect a chain of dimers, can be constructed. This
structure was also examined by Zou and Schatz [22] and it was found that long
range photonic effects similar to those observed with single sphere chains can be
combined with plasmonic hotspots between the particles to produce even higher
enhancement factors. The main difference is that the location of the most intense
photonic resonances are now red shifted from the isolated sphere case due to the
red shift in the dipole plasmon mode of a sphere dimer system. In this work we
study models similar to Zou and Schatz, but now including for DR effects in the
enhancement.
5 Dipole Re-Radiation Effects in Surface Enhanced Raman Scattering 149
Qext
to-center distances. The gap 6
between the two spheres in the
dimers is 6.25 nm 4
0
300 350 400 450 500 550 600 650 700 750 800
Wavelength (nm)
(b) 20
18 Single dimer
16 106.25nm
500nm
14 600nm
12
Qext
10
8
6
4
2
0
300 350 400 450 500 550 600 650 700 750 800
Wavelength (nm)
(c) 22
20 Single dimer
18 500nm
600nm
16
14
12
Qext
10
8
6
4
2
0
300 350 400 450 500 550 600 650 700 750 800
Wavelength (nm)
In Fig. 5.8 we present the extinction spectra of arrays of sphere dimer chains with
a dimer gap of 6.25 nm for various chain lengths and dimer-to-dimer separation dis-
tances. Note that the dimer chain is taken to be along the y-axis, but with each dimer
oriented along the x-axis. Also, the spheres have a diameter of 100 nm, and the mole-
150 L. K. Ausman and G. C. Schatz
Raman Enhancement
G(θs=3π/4, φs=π/2)
center-to-center separation of
1×10 6
106.25 nm. The gap between
the two chains is 6.25 nm.
1×10 5
The molecule is located 0.5 nm
from the surface of the sphere
1×10 4
at the coordinate origin along
the axis parallel with the
1×10 3
incident E-field polarisation.
The case of 5 dimers is
1×10 2
presented in a and the case of 350 400 450 500 550 600 650 700 750 800
15 dimers is presented in b. Excitation Wavelength (nm)
In both a and b the solid black
line is the PW approximation (b)
1×10 7 4
result, the dashed red line is |Eloc|
the backscattering DR result G(θs=π)
Raman Enhancement
1×10 4
1×10 3
350 400 450 500 550 600 650 700 750 800
Excitation Wavelength (nm)
cule is located 0.5 nm from the surface of one sphere in the gap of the central dimer.
An examination of the extinction spectra of the arrays of dimers shows trends similar
to the single sphere chains in Fig. 5.4. For small dimer-to-dimer separations the dipole
peak is suppressed and then becomes enhanced as the separation between the dimers
becomes larger. Increasing the number of components in the chains for both single
spheres and dimers increases the extinction, and an examination of Figs. 5.5–5.7 and
Figs. 5.9–5.11 indicates that this also leads to an increase in the PW enhancement
factor. At separations greater than 400 nm we see intense enhancements that are due
to long range photonic resonances. These become more prominent as the number
of dimers is increased, just as is observed for single sphere chains. There is little
change in the maximum observed enhancement factor when the number of dimers
is increased from 5 to 15, but the larger number of dimers results in a maximum in
the dipole plasmon region that is not as prominent in the 5–dimer case.
In Figs. 5.9a, 5.10a and 5.11a the SERS excitation profiles for chains made up of
5 dimers are shown. We see that chains made up of 5 dimers where the detector is
located in a backscattering configuration have SERS enhancement factors that are
accurately approximated by the PW enhancement factor. In fact comparisons with
5 Dipole Re-Radiation Effects in Surface Enhanced Raman Scattering 151
Raman Enhancement
Ag spheres with a dimer 6
center-to-center separation 1×10
of 500 nm. The gap between 5
the two chains is 6.25 nm. 1×10
The molecule is located 0.5 nm 4
4
from the surface of the sphere 1×10 |Eloc|
at the coordinate origin along G(θs=π)
the axis parallel with the 1×10
3 G(θs=3π/4, φs=π/2)
incident E-field polarisation.
2
The case of 5 dimers is 1×10
350 400 450 500 550 600 650 700 750 800
presented in a, the case of
Excitation Wavelength (nm)
15 dimers is presented in b,
and the case of 45 dimers is
presented in c. In a, b, and (b)
1×10 8
c the solid black line is the
Raman Enhancement
1×10 2
350 400 450 500 550 600 650 700 750 800
Excitation Wavelength (nm)
(c) 1×108
Raman Enhancement
1×107
1×106
1×105
1×104 4
|Eloc|
3 G(θs=π)
1×10
G(θs=3π/4, φs=π/2)
1×102
350 400 450 500 550 600 650 700 750 800
Exciataion Wavelength (nm)
Figs. 5.5a, 5.6a and 5.7a show that the PW approximation is better in the dimer case
than is observed in the case of single sphere chains. This is because for single sphere
chains, higher order multipoles are important at certain frequencies (towards the blue)
152 L. K. Ausman and G. C. Schatz
Raman Enhancement
Ag spheres with a dimer 1×10
6
center-to-center separation
of 600 nm. The gap between 5
1×10
the two chains is 6.25 nm.
The molecule is located 0.5 nm 4
from the surface of the sphere 1×10 4
|Eloc|
at the coordinate origin along G(θs=π)
3
the axis parallel with the 1×10 G(θs=3π/4, φs=π/2)
incident E-field polarisation.
The case of 5 dimers is 350 400 450 500 550 600 650 700 750 800
presented in a, the case of Excitation Wavelength (nm)
15 dimers is presented in b,
and the case of 45 dimers is (b)
presented in c. In a, b, and 7
1×10
Raman Enhancement
350 400 450 500 550 600 650 700 750 800
Excitation Wavelength (nm)
(c) 8
1×10
Raman Enhancement
7
1×10
6
1×10
5
1×10
4 4
1×10 |Eloc|
3
G(θs=π)
1×10 G(θs=3π/4, φs=π/2)
350 400 450 500 550 600 650 700 750 800
Excitation Wavelength (nm)
which are not accessed by plane wave scattering. However the dimer resonances are
dipolar in character so the influence of higher multipoles is less important.
Figures 5.9b, 5.10b and 5.11b show SERS excitation profiles for 15–dimer chains
at dimer-to-dimer separations of 106.25, 500 and 600 nm while Figs. 5.10c and 5.11c
show excitation profiles of 45–dimer chains with dimer-to-dimer separations of 500
5 Dipole Re-Radiation Effects in Surface Enhanced Raman Scattering 153
and 600 nm. Again there is good agreement of PW and DR results for the backscat-
tering arrangement. In Figs. 5.10b and 5.11b we see the influence of long range
photonic resonances which become more prominent in Figs. 5.10c and 5.11c. These
photonic resonances combine with the large field enhancement associated with the
dimer to give peak enhancement factors above 108 , thus confirming the previous
conclusions of Zou and Schatz concerning the effect of combining plasmonic and
photonic resonance contributions to the electromagnetic enhancement.
As seen in Sect. 5.4 with single sphere chains, Figs. 5.9b, 5.10b and 5.11b show
that changing the detector position so that it is no longer symmetric with respect
to the dimer chain results in significant differences between the PW approximation
and dipole re-radiation. These differences are due to phase interference effects as
discussed earlier. There are certain frequencies where the PW and DR results are in
good agreement for this detector location, especially for larger dimer separations and
for longer dimer chains. This is especially noticeable for shorter wavelengths, where
there are no important photonic resonances, so only localised plasmonic excitatiton
is important.
We now examine the absolute errors associated with the PW approximation at the
maximum enhancement for the dimer chains. The reported errors are for separations
of 106.25, 500, and 600 nm for the 5–dimer chains in Figs. 5.9a, 5.10a, and 5.11a, the
15–dimer chains in Figs. 5.9b, 5.10b, and 5.11b and separations of 500 and 600 nm
for the 45–dimer chains in Figs. 5.10c and 5.11c. In a similar fashion to the single
sphere chains, the backscattering detector location results in smaller errors which for
5–dimer chains are 0.077, 0.22 and 3.7 %. For 15–dimer chains these errors become
0.59, 0.23 and 3.84% and for 45 dimers the errors are 0.89 and 0.56%. Larger
errors are again seen when the detector location is at θs = 3π/4 and φs = π/2.
The 5–dimer chain has errors of 1205, 636 and 5.78% for this detector location
while the 15–dimer chain has errors of 56.2, 1192 and 7.72%. The errors associated
with this detector location in the 45–dimer chain are 986 and 1555%.
5.6 Conclusions
Dipole re-radiation effects can play a significant role in the modeling of SERS
enhancement factors. In single particle chains and in dimer chains we observe that
the main differences between dipole re-radiation calculations and the PW approxi-
mation are the result of phase interference effects that result from scattering of the
emitted dipole wave from multiple spheres. Such interferences are relatively unim-
portant for cases where the detector location is in a location that is symmetric with
respect to the array structure; however, they can produce major differences in peak
SERS enhancement factors for other detector geometries. Differences between PW
and DR also occur in certain cases for isolated spheres due to multipole excitation
effects, as multipole resonances can be more easily excited by a dipole field than by
plane waves.
154 L. K. Ausman and G. C. Schatz
The maximum enhancement factors for single spheres are around 103 , and for
sphere dimers around 107 . Photonic resonances in sphere chains and dimer chains
with the right orientation can increase these enhancement factors by 1–2 orders of
magnitude when the center-to-center separation distance is close to the wavelength
of light. We note, however, that these effects will be smaller when nonzero Stokes
shifts are included in the Raman intensity evaluation.
References
1. M.G. Albrecht, J.A. Creighton, J. Am. Chem. Soc. 99, 5215 (1977)
2. D.L. Jeanmaire, R.P. van Duyne, Electroanal. Chem. 84, 1 (1977)
3. M. Fleischmann, P.J. Hendra, A.J. McQuillian, Chem. Phys. Lett. 26, 163 (1974)
4. M. Moskovits, J. Chem. Phys. 69, 4159 (1978)
5. M.R. Philpott, J. Phys. Colloques 44, 295 (1983)
6. H. Metiu, P. Das, Ann. Rev. Phys. Chem. 35, 507 (1984)
7. G.C. Schatz, Acc. Chem. Res. 17, 370 (1984)
8. M. Moskovits, Rev. Mod. Phys. 57, 783 (1985)
9. A. Wokaun, Mol. Phys. 56, 1 (1985)
10. M. Kerker, Enhanced Raman scattering in colloidal systems. No. 45 in Studies in Physical and
Theoretical Chemistry (Elsevier, Amsterdam, 1987)
11. M. Kerker, D.S. Wang, H. Chew, App. Opt. 19, 4159 (1980)
12. G. Mie, Ann. Phys. 25, 377 (1908)
13. H. Chew, P.J. McNulty, M. Kerker, Phys. Rev. A 13, 396 (1976)
14. H. Chew, M. Kerker, P.J. McNulty, J. Opt. Soc. Am. 66, 440 (1976)
15. H. Chew, M. Kerker, D.D. Cooke, Phys. Rev. A 16, 320 (1977)
16. R. Ruppin, J. Chem. Phys. 76, 1681 (1982)
17. M. Kerker, J. Opt. Soc. Am. B 2, 1327 (1985)
18. M. Kerker, J. Colloid Interface Sci. 118, 417 (1987)
19. L.K. Ausman, G.C. Schatz, J. Chem. Phys. 131, 084708 (2009)
20. S. Zou, N. Janel, G.C. Schatz, J. Chem. Phys. 120, 10871 (2004)
21. S. Zou, G.C. Schatz, J. Chem. Phys. 121, 12606 (2004)
22. S. Zou, G.C. Schatz, Chem. Phys. Lett. 403, 62 (2005)
23. S. Zou, G.C. Schatz,Coupled Plasmonic Plasmon/Photonic Resonance Effects in SERS. No.
103 in Topics in Applied Physics (Springer, Berlin, 2006)
24. S. Zou, G.C. Schatz, Nanotechnology 17, 2813 (2006)
25. S. Zou, G.C. Schatz, Isreal J. Chem. 46, 293 (2006)
26. L. Quin, S. Zou, C. Xue, A. Atkinson, G.C. Schatz, C.A. Mirkin, Proc. Natl. Acad. Sci. USA
103, 13300 (2006)
27. J.P. Camden, J.A. Dieringer, Y. Wang, D.J. Masiello, L.D. Marks, G.C. Schatz, R.P. van Duyne,
J. Am. Chem. Soc. 130, 12616 (2008)
28. W. Wei, S. Li, J.E. Millstone, M.J. Banholzer, X. Chen, X. Xu, G.C. Schatz, C.A. Mirkin,
Angew. Chem. Int. Edit. 48, 4210 (2009)
29. J.A. Dieringer, K.L. Wustholz, D.J. Masiello, J.P. Camden, S.L. Kleinman, G.C. Schatz, R.P.
van Duyne, J. Am. Chem. Soc. 131, 849 (2009)
30. K. Kneipp, W. Wang, H. Kneipp, L.T. Perelman, I. Itzkan, R.R. Dasari, M.S. Feld, Phys. Rev.
Lett. 78, 1667 (1997)
31. S. Nie, S. Emory, Science 275, 1102 (1997)
5 Dipole Re-Radiation Effects in Surface Enhanced Raman Scattering 155
32. M. Moskovits, L.L. Tay, J. Yang, T. Haslett, SERS and the single molecule. No. 82 in Topics
in Applied Physics (Springer, Berlin, 2002)
33. B. Vlčková, M. Moskovits, I. Pavel, K. Šišková, M. Sládková, M. Šlouf, Chem. Phys. Lett.
455, 131 (2008)
34. Y. Fang, N.H. Seong, D.D. Dlott, Science 321, 388 (2008)
35. K.E. Shafer-Peltier, C.L. Haynes, M.R. Glucksberg, R.P. van Duyne, J. Am. Chem. Soc. 125,
588 (2003)
36. X. Zhang, M.A. Young, O. Lyandres, R.P. van Duyne, J. Am. Chem. Soc. 127, 4484 (2005)
37. N.C. Shah, O. Lyandres, C.R. Yonzon, X. Zhang, R.P. van Duyne, ACS Symposium Series
963, 107 (2007)
38. S.D. Hudson, G. Chumanov, Anal. Bioanal. Chem. 394, 679 (2009)
39. D.A. Stuart, K.B. Biggs, R.P. van Duyne, Analyst 131, 568 (2006)
40. A.V. Whitney, F. Casadio, R.P. van Duyne, App. Spect. 61, 994 (2007)
41. C.L. Brosseau, A. Gambardella, F. Casadio, C.M. Grzywacz, J. Wouters, R.P. van Duyne, Anal.
Chem. 81, 3056 (2009)
42. S. Malynych, G. Chumanov, J. Am. Chem. Soc. 125, 2896 (2003)
43. A.S. Kumbhar, M.K. Kinnan, G. Chumanov, J. Am. Chem. Soc. 127, 12444 (2005)
44. M.K. Kinnan, G. Chumanov, J. Phys. Chem. C 111, 18010 (2007)
45. Y.D. Suh, G.K. Schenter, L. Zhu, H.P. Lu, Ultramicroscopy 97, 89 (2003)
46. L. Zhu, G.K. Schenter, M. Micic, Y.D. Suh, N. Klymyshyn, H.P. Lu, Proc. SPIE 4962, 70
(2003)
47. P.I. Geshev, K. Dickmann, J. Opt. A: Pure Appl. Opt. 8, S161 (2006)
48. T. Shegai, Z. Li, T. Dadosh, Z. Zhang, H. Xu, G. Haran, Proc. Natl. Acad. Sci. U.S.A. 105,
16448 (2008)
49. D. Pristinski, E.C. la Ru, S. Tan, S. Sukhishvili, H. Du, Opt. Express 16, 20117 (2008)
50. C.F. Bohren, D.R. Huffman, Absorption and Scattering of Light by Small Particles (Wiley-
VCH, Weinheim, 2004)
51. D.W. Mackowski, Proc. R. Soc. London Ser. A 433, 599 (1991)
52. D.W. Mackowski, J. Opt. Soc. Am. A 11, 2851 (1994)
53. D.W. Mackowski, M.I. Mishchenko, J. Opt. Soc. Am. A 13, 2266 (1996)
54. L.K. Ausman, G.C. Schatz, J. Chem. Phys. 129, 054704 (2008)
55. C.T. Tai, Dyadic Green’s Functions in Electromagnetic Theory (Intext Educational, Scranton,
San Francisco, Toronto, London, 1971)
56. P.B. Johnson, R.W. Christy, Phys. Rev. B 6, 4370 (1972)
57. D.J. Masiello, G.C. Schatz, Phys. Rev. A 78, 042505 (2008)
Chapter 6
Optical Force and Torque on Single and
Aggregated Spheres: The Trapping Issue
Abstract Radiation force and radiation torque stem from the conservation theorems
governing the interaction between radiation and particles. As a result, even simple
plane waves may yield force and torque that are nonnegligible when studying the
dynamics of particles. We show how these forces and torques can be put into formulas
of practical use starting with the expansion of the electromagnetic field in a series
of vector multipole fields, and adapt the formalism for particles that are fairly well
modeled as single or aggregated spheres. Then, we extend the formalism to deal with
the case the field is a highly focalized laser beam, that, as is well known, may trap
particles within the focal region. Finally, we present our calculations performed by
the theory we exposed, finding a convincing concordance as long as comparison can
be made with available experimental data.
6.1 Introduction
The interaction of radiation field with particles obeys the conservation of energy as
well as the conservation both of the linear and of the angular momentum. As a result,
particles illuminated by a radiation field may experience both a force FRad and a
torque Γ Rad that contribute in determining their dynamical behavior. For instance,
in an astrophysical context, the radiation force may determine the space distribution
of the cosmic dust grains [1], whereas the radiation torque is considered as a non-
W. Hergert and T. Wriedt (eds.), The Mie Theory, Springer Series in 157
Optical Sciences 169, DOI: 10.1007/978-3-642-28738-1_6,
© Springer-Verlag Berlin Heidelberg 2012
158 R. Saija et al.
negligible agent of their partial alignment [2]. In laboratory applications both FRad
and Γ Rad proved useful in manipulating microsized and nanosized particles under
controlled conditions.
The actual calculation of FRad and Γ Rad requires the knowledge of the electro-
magnetic field in the presence of the particles. This amounts to say that besides the
incident radiation fields EI and BI one has to know the fields ES and BS scattered by
the particles. The latter fields can be calculated by several methods among which we
prefer the one based on the expansion of the electromagnetic field in a series of vec-
tor spherical multipole fields in the framework of the transition matrix approach [3].
Actually, this is the approach on which Mie theory is based [4], although in 1908 the
concept itself of transition matrix was still to be devised. The advantages of resorting
to the transition matrix stem from its independence of the direction of incidence and
of the polarization, even for nonspherical particles, and from its well-defined trans-
formation properties under rotation of the coordinate frame. As a consequence, the
elements of the transition matrix need to be calculated only once for an arbitrarily
chosen orientation of the particle with respect to the incident field.
In this contribution we will describe how the multipole fields expansion can be
applied to calculate the force and torque exerted by a radiation field on particles that
either are spherical or can be modeled as aggregates of spheres that may be radially
inhomogeneous and need not be identical to each other. In any case, most of the
theory can be formulated without referring to the shape of the particles or to their
dielectric properties. It is only when one has to perform the calculation of ES that
the shape and properties of the particles enter into play.
In particular, the theory we are going to describe will prove suitable to deal
with the problem of trapping of microsized and nanosized particles and to study the
orientation they assume under the effect of the radiation torque.
Anyway, we do not present here a review of the theories of trapping and manipu-
lation of small particles but only a description of our own approach and some of its
applications.
The geometry of the problem we deal with is depicted in Fig. 6.1, which also applies
when the incident radiation is a focalized laser beam. We choose a reference frame
Σ fixed in the laboratory, with respect to which the particle, i.e., its center of mass,
has the vector position R O . Of course, the coincidence of O with the center of
mass is non-necessary, but it is useful when one is interested both in the translational
and in the rotational dynamics of the particle. We also attach to the particle a local
frame of reference Σ with origin at O and whose axes are parallel to the axes of
Σ. Then, the radiation force and the radiation torque on the particle are given by the
integrals [5]
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 159
FRad = r 2
r̂ · TM d (6.1)
and
Γ Rad = −r 3 r̂ · TM × r̂ d , (6.2)
respectively. In (6.1) and (6.2) the integration is over the full solid angle, r is the
radius of a large (possibly infinite) sphere with center at R O surrounding the particle,
and
1 1
TM = Re n 2 E ⊗ E∗ + B ⊗ B∗ − n 2 |E |2 + |B |2 I (6.3)
8π 2
is the time averaged Maxwell stress tensor in the form of Minkowski [5]. In (6.3),
the fields E and B are considered in the frame Σ , ⊗ denotes dyadic product, the
asterisk denotes complex conjugation, and I is the unit dyadic. In turn, n is the real
refractive index of the homogeneous, non absorbing, isotropic medium surrounding
the particle. Of course, choosing for TM the form of Minkowski rather than the form
of Becker is somewhat arbitrary as discussed by Pfeiffer et al. [6], who also stress
that no experiment has been able, till date, to encourage the choice of the Minkowski
form over the Becker form. Actually, the form of Minkowski is nonsymmetric in an
anisotropic medium, but since the particles we deal with are embedded in an isotropic
medium, no inconvenience arises because of our choice.
160 R. Saija et al.
According to our statement in Sect. 6.1, the fields E and B that enter (6.3) are the
superposition of the incident field and of the field scattered by the particle. We, for
now, assume that the incident radiation is the plane wave
EI = E 0 ûI eikI ·r = E 0 ûI eikI ·(r +R O ) = E 0 ûI eikI ·r = EI (6.4)
of wavevector kI = k̂I k = k̂I nkv , where k̂I is the unit vector in the direction of
incidence, kv = ω/c, and ûI is the (unit) polarization vector that may be either real
(linear polarization) or complex (elliptic polarization). Obviously, in (6.4), |E 0 |2 =
|E 0 |2 = II , the intensity of the incident field. In view of our choice of the incident
field and of the large radius of the integration sphere, the integral in (6.1) can be
calculated by resorting to the asymptotic expansion of a plane wave [7]
2πi
+ |BS | + 2B∗
2
I · BS
d . (6.5)
Of course, choosing v̂ζ to coincide in turn with the unit vectors that characterize the
axes of Σ one gets the cartesian components of the radiation force, whereas, choosing
v̂ζ along k̂I , one gets the radiation pressure in the form derived by Mishchenko [8]
in terms of the customary asymmetry parameter. The components of FRad in a plane
orthogonal to k̂I can also be calculated and imply the definition of two new asymmetry
parameters [9].
Obviously, since the incident field is a plane wave, the integral (6.5) gets no con-
tribution from the terms EI · E∗ ∗
I , and BI · BI that, accordingly, have been omitted.
However, it is less obvious, although conceptually expected, that the integral (6.5)
gets no contribution from the same terms even when the incident field is a superpo-
sition of plane waves with different directions of propagation. This statement will be
proved in Appendix 1.
We expand both the incident and the scattered field in a series of vector multipole
fields [10]. Accordingly,
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 161
( p) ( p)
EI = E 0 Jlm r , k Wlm ûI , k̂I , (6.6)
plm
( p) ( p)
ES = E 0 Hlm r , k Alm ûI , k̂I , (6.7)
plm
hence the multipole expansion of the magnetic field can be inferred through the
equation
i
B = − ∇ × E.
kv
In (6.7) the amplitudes of the scattered field bear the arguments ûI and k̂I to recall that
in general they depend both on the polarization and on the direction of propagation
of the incident field. The multipole fields in (6.6) are
where
(1) (2)
Zlm k̂ = Xlm k̂ , Zlm k̂ = Xlm k̂ × k̂
are transverse vector harmonics [10, 11], and the amplitudes of the scattered field
( p)
Alm are calculated by imposing the customary boundary conditions across the sur-
face of the particle. The actual calculation of the A-amplitudes in the framework of
the transition matrix method will be discussed in Appendix 2.
Once the scattered field has been calculated, we perform the integration in (6.5)
by exploiting the asymptotic expansion of the multipole fields [10] up to terms which
give a contribution of order 1/r . Actually, these expansions for the incident and the
scattered field give
162 R. Saija et al.
( p) ( p) (−)
p−1
EI → E 0 Zlm r̂ WI lm sin kr − (l + 1 − p)π/2 , (6.8a)
kr
plm
( p) ( p) exp ikr
ES → E 0 Zlm r̂ Alm i−l− p (6.8b)
kr
plm
(Sca) (Ext)
= −FRad ζ + FRad ζ , (6.9)
where Y1μ (v̂ζ ) denotes spherical harmonics whose arguments are the polar angles
of v̂ζ , and
( pp ) ( p)∗ ( p )
K μ;lml m = 16π 2 il−l Y1μ r̂ i p− p Zlm r̂ · Zl m r̂ d .
( pp )
The integrals K μ;lml m can be performed in closed form [9, 10] with the result
( pp ) 3 ( pp )
K μ;lml m = 16π 2
C 1, l , l; μ, m , m i l−l Oll , (6.10)
4π
where C 1, l , l; μ, m , m are Clebsch-Gordan coefficients [14] and
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 163
( pp ) 1 ( pp ) (l − 1)(l + 1)
Oll = −√ 1 − δ pp , Ol,l−1 = δ pp ,
l(l + 1) l(2l + 1)
( pp ) l(l + 2)
Ol,l+1 =− δ pp .
(l + 1)(2l + 1)
K ζ(11) (22)
lml m = K ζ lml m , K ζ(12) (21)
lml m = K ζ lml m ,
(Sca) (Ext)
which help us to get a more compact expression for FRad ζ and FRad ζ . In fact, from
(6.9) we get
(Sca) II ( p)∗ ( p ) ( pp )
FRad ζ = 2
Re Alm Al m i l−l Iζ lml m , (6.11a)
8π kv
plm p l m
II ( p)∗ ( p )
(Ext) l−l ( pp )
FRad ζ =− Re W I lm A l m i I ζ lml m . (6.11b)
8π kv2
plm p l m
Note that in (6.11b) above the apparent undue dependence on r remaining in (6.9)
disappeared. The meaning of the superscripts (Ext) and (Sca) will be explained in
Sect. 6.2.3.
In order to calculate the radiation torque, we remark that, since r̂ · I × r̂ = 0, the
last two terms in TM give no contribution to the integral in (6.2), which, therefore,
reduces to
Γ Rad = −r 3
n 2 r̂ · E E∗ × r̂ + r̂ · B B∗ × r̂ d . (6.12)
Since the radius of the integration sphere is large, possibly infinite, we can resort
to the asymptotic expression of the fields. Nevertheless, the reader is warned that,
as regards the scattered field, the customary far zone expression in terms of the
scattering amplitude
exp (ikr )
ES = E 0 f k̂S , k̂I
r
cannot be used lest to get a vanishing result [8] because r̂ · f = 0. The correct result
is obtained by solving the problem of scattering and then by expanding the fields for
164 R. Saija et al.
large r and retaining all terms that give contributions of order 1/r 3 to the integrand
in (6.2).
Assuming the scattered field has been calculated, let us search for its expression
in the far zone. For the H-multipole fields we get [15]
(1) eikr (1)
Hlm → (−i)l+1 Z ,
kr lm
(2) (−i)l ikr (−i)l+1 ikr (2) (−i)l ikr (2)
Hlm → l(l + 1)e Ylm r̂ − e Zlm − e Zlm . (6.13)
k 2 r 2 k 2 r 2 kr
Analogously, for the J-multipole fields we get [15]
(1) 1 (1)
Jlm → sin kr − lπ/2 Zlm ,
kr
(2) i 1 (2)
Jlm → 2 2 l(l + 1) sin kr − lπ/2 Ylm r̂ − 2 2 sin(kr − lπ/2)Zlm
k r k r
1 (2)
− sin kr − (l − 1)π/2 Zlm . (6.14)
kr
Now, the H-fields and J-fields enter the integrand in (6.2) through the dot products
ikr
(2) l e
r̂ · Hlm = l(l + 1)(−i) 2 2 Ylm = cr2l Ylm ,
k r
(2)
sin kr − lπ/2
r̂ · Jlm = i l(l + 1) Ylm = cr1l Ylm ,
k 2 r 2
and through the cross products
( p̄)∗ ¯ e−ikr ( p̄)∗ ( p̄) ( p̄)∗
Hl¯m̄ × r̂ = il+ p̄
Zl¯m̄ × r̂ = ct2l¯ Zl¯m̄ × r̂ ,
kr
( p̄)∗ sin kr − (l¯ + p̄ − 1)π/2 ( p̄)∗ ( p̄) ( p̄)∗
Jl¯m̄ × r̂ = Zl¯m̄ × r̂ = ct1l¯ Zl¯m̄ × r̂ ,
kr
where we neglected the further terms that would come from (6.13) and (6.14) because
they vanish at infinity to an order higher than the order we have to retain.
At this stage we found it convenient to focus on the spherical components of
the radiation torque Γμ , that are related to the rectangular components of Γ Rad
through [14]
1 i
ΓRad x = Re √ (Γ−1 − Γ1 ) , ΓRad y = Re √ (Γ−1 + Γ1 ) ,
2 2
ΓRad z = Re (Γ0 ) . (6.15)
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 165
Accordingly, we note that the integrand in (6.12) gets contributions that can be
written as
( p p̄) ( p̄)
K μ;α ᾱlml¯m̄ = δ p2 d αlm vμ;ᾱl¯m̄ , (6.16)
where
(2) (2)
d1lm = r̂ · Jlm , d2lm = r̂ · Hlm ,
( p̄) ( p̄)∗ ( p̄) ( p̄)∗
vμ;1l¯m̄ = Jl¯m̄ × r̂ , vμ;2l¯m̄ = Hl¯m̄ × r̂ .
μ μ
The subscripts α and ᾱ in (6.16) denote terms coming either from the incident field
(α, ᾱ = 1) or from the scattered field (α, ᾱ = 2). Thus, performing in (6.16) the
appropriate substitutions and integrating according to (6.2), we get
(21) 2l + 1
(1) ¯ 0, 0)δm,m̄−1
I1;α ᾱlml¯m̄ = icrαl ctᾱl¯ C(1, l, l;
2l¯ + 1
× C(1, l, ¯ l;
¯ 1, m̄ − 1)C(1, l, l;
¯ 0, m, m̄ − 1)
−C(1, l, ¯ l;
¯ 0, m̄)C(1, l, l;
¯ 1, m, m̄) ,
(21) (1) 2l + 1 ¯ 0, 0)δm,m̄
I0;α ᾱlml¯m̄ = icrαl ctᾱl¯ C(1, l, l;
2l¯ + 1
× C(1, l, ¯ l;
¯ 1, m̄ − 1)C(1, l, l;
¯ −1, m, m̄ − 1)
−C(1, l, ¯ l;
¯ −1, m̄ + 1)C(1, l, l;¯ 1, m, m̄ + 1) ,
(21) (1) 2l + 1 ¯ 0, 0)δm,m̄+1
I−1;α ᾱlml¯m̄ = icrαl ctᾱl¯ C(1, l, l;
2l¯ + 1
× C(1, l, ¯ l;
¯ 0, m̄)C(1, l, l;
¯ −1, m, m̄)
−C(1, l, ¯ l;
¯ −1, m̄ + 1)C(1, l, l;¯ 0, m, m̄ + 1) ,
and
(22) (2)
¯ μ, m̄ − μ)δ ¯δm,m̄−μ ,
Iμ;α ᾱlml¯m̄ = crαl ctᾱl¯C(1, l, l; ll (6.17)
where the C’s denote Clebsch-Gordan coefficients [14]. Finally, collecting all the
terms one obtains
n 2 r 3 (2 p̄) (2) ( p̄)∗ (1) ( p̄ )∗
Γμ = − Iμ;α ᾱlml¯m̄ aαlm aᾱl¯m̄ + aαlm aᾱl¯m̄ , (6.18)
8π
α ᾱ p̄ lm l¯m̄
( p) ( p) ( p) ( p)
where p̄ = p̄, a1lm = WI lm , and a2lm = Alm .
At this stage we remark that (6.18) could be used for a brute force calculation of
the spherical components of the radiation torque. However, we can obtain a more
166 R. Saija et al.
efficient formula through a detailed inspection of the terms present. In fact, (6.18) is
built as a sum of contributions with different α, ᾱ, and p̄ or p̄ .
• The contributions with α = ᾱ = 1 are just vanishing both for p̄ = 2 or p̄ = 1
and for p̄ = 1 or p̄ = 2. This result, that is identical to the one of Marston
and Crichton [16], comes from the fact that these terms describe the flux of the
momentum of the Maxwell stress tensor through a closed surface in the absence
of any scatterer.
• The contribution with α = ᾱ = 2 vanishes both for p̄ = 1 and for p̄ = 2.
• The contributions for α = ᾱ depend on r , i.e., on the radius of the spherical surface
of integration. Nevertheless, those with p̄ = 1 are identical but of opposite sign
and cancel each other; on the other hand, the sum of those with p̄ = 2 turns out
to be independent of r . Analogous considerations hold true for the contributions
with p̄ = 2 and p̄ = 1, respectively. This result is a consequence of the fact that
the torque cannot depend on the choice of the surface of integration.
Now, separating in (6.18) the sum over α = ᾱ from the one over α = ᾱ, we can
write
Γμ = Γμ(Ext) − Γμ(Sca) , (6.19)
with
II ( p) ( p)∗
Γμ(Sca) = − s μ;lm Al,m−μ Alm , (6.20a)
8π nkv3
plm
II ( p) ( p)∗
Γμ(Ext) = s μ;lm WIl,m−μ Alm , (6.20b)
8π nkv3
plm
where
(l − m)(l + 1 + m) (l + m)(l + 1 − m)
s−1;lm =− , s1;lm = ,
2 2
s0;lm = −m .
We want to stress that the radiation torque does not vanish for a homogeneous
sphere provided the polarization is elliptical and the sphere is absorbing. This result
has been first proved by Marston and Chricton [16] just in terms of multipole fields
expansion. The radiation torque due to an elliptically polarized plane wave does not
vanish either for an absorbing scatterer with axial symmetry provided the incidence
is along the cylindrical axis.
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 167
Although in our approach the radiation force and torque have been separated into
two contributions labeled (Sca) and (Ext), there is no similarity with the customary
separation into a field gradient contribution and a scattering contribution. The latter
contributions, in fact, arise when exploiting the dipole approximation [17], whereas
the separation effected in (6.11a) and (6.11b) as well as in (6.20a) and (6.20b) can
be tracked back to (6.1) and (6.2).
Equation (6.1), due to the structure of the Maxwell stress tensor, includes |ES |2
and E∗
I · ES as well as the corresponding terms from the magnetic field. When these
(Sca)
terms are expanded as a series of multipole fields we get FRad ζ , that depends on
(Ext)
the multipole amplitudes of the scattered field only, and FRad ζ that depends on the
multipole amplitudes both of the incident and of the scattered fields, in analogy to
what one finds for the scattering cross section and for the extinction cross section of
a particle, respectively [10]. As a consequence, FRad ζ can loosely be related to the
absorptivity of a particle.
In turn, (6.2) gets from the Maxwell tensor the terms (r · ES )(E∗
S × r ) and
(r · EI )(E∗
S × r ), and the corresponding terms from the magnetic field. Again, the
multipole expansion of these terms yields just ΓRad Sca and Γ Ext which depend on the
μ Rad μ
multipole amplitudes of the scattered field only and by the multipole amplitudes of
the incident and of the scattered field, respectively. Thus, considerations analogous
to those made above for the radiation force hold true also for the radiation torque
[18, 19]: in particular, the torque exerted by an elliptically polarized plane wave on a
spherical scatterer can be explicitly written in terms of the difference of the extinction
and of the scattering cross section [16].
The theory that we summarized in Sect. 6.2 can be applied to particles of any shape,
provided we are able to calculate their transition matrix and the incident field is
well represented by simple plane waves. This is often the case in an astrophysical
environment, because the cosmic dust grains are well represented either by single
spheres or by aggregates of spheres [20]. The component of the radiation force
along the direction of propagation is the well-known radiation pressure, whereas the
transverse components, although less well-known, produce dynamical effects whose
relevance has been recognized in the last few years [21]. In turn, the radiation torque
may produce the so-called superthermal spin-up whose relevance for the alignment
of the cosmic dust grains has been investigated, e.g., by Purcell [22] and by Draine
and Weingartner [23]. Although we dealt with the calculation of the radiation force
and torque on reliable models of cosmic dust grains [24], in the present context we
prefer to expound a more illuminating study on clusters of selected geometry. Our aim
168 R. Saija et al.
is, indeed, to highlight the dependence of radiation force and torque on the structure
and refractive index of particles of known geometry and dielectric properties.
We apply the theory to binary aggregates of latex spheres whose dielectric function,
according to Ma et al. [25] has the value ε = 2.490087 + i0.001578 at λ = 600 nm.
More precisely, we considered a reference sphere with radius 200 nm and the aggre-
gate composed by two mutually contacting spheres whose radius is 158.77 nm so that
their total volume equals that of the reference sphere. The frame Σ attached to the
aggregate is chosen to coincide to the laboratory frame Σ and the point of contact
of the spheres always coincide with the origin, whereas the axis of the aggregate
lies in the z -x plane. We considered four configurations each characterized by the
angle ϑ between the axis of the aggregate and the z axis: (i) ϑ = 90◦ , (ii) ϑ = 60◦ ,
(iii) ϑ = 30◦ , and (iv) ϑ = 0◦ . We assume the incident field to be linearly polar-
ized and that the plane of reference is the meridional plane through the z = z axis
and the incident wavevector kI whose direction is individuated by the polar angles
ϑI and ϕI . We decompose the radiation force along the triplet formed by k̂I = v̂3 ,
ϑ̂ I = v̂1 , and ϕ̂ I = v̂2 . The component Fk = FRad · k̂I is the well-known radiation
pressure, whereas the transverse components are FT η = FRad · v̂η , η = 1, 2. Actu-
ally, the quantities that we report in the following figures are Fk = 8π Fk /II and
1/2
FT = 8π FT2 1 + FT2 2 /II , as well as their axial averages around the z axis of Σ.
All these quantities are given in μm2 and, although they depend on the polarization
of the incident field, are reported only for polarization parallel to the plane of scat-
tering. The interested reader can find a more complete set of figures in [9, 10]. As
our calculations are based on the multipole expansion of the fields, their convergence
must carefully be checked. Actually the size parameter of the smallest mathematical
sphere that includes a whole binary aggregate is x = 3.1, so that convergence to 4
significant digits required lM = 8.
We start by stating that the radiation pressure on the reference sphere is
0.125027 μm2 , whereas the transverse components vanish due to the spherical sym-
metry. On the contrary, the results reported in Fig. 6.2 show that, as expected, the
transverse components on the binary aggregate do not vanish for any value of ϕI
and their value is a significant fraction of the radiation pressure. These results are
confirmed by the axial averages we report in Fig. 6.3 for all the configurations we
considered. Perhaps the most interesting results are those reported in Fig. 6.4 where
we study the dependence of the axial averages on the absorptivity of the aggregates.
In other words, we considered particles whose real part of the dielectric function is
that of Ma et al. [25], whereas the imaginary part changes from 0 to 0.1. In fact,
the transverse components of the radiation force show a weak dependence on Im(ε),
whereas the radiation pressure has a comparatively stronger dependence. This behav-
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 169
(a) (b)
0.07 0.11
0.06
0.05 0
20 0.09
40
0.04 60
80
0.03
0.07
0.02
0.01
0 0.05
0 30 60 90 0 30 60 90
Fig. 6.2 FηT in a, and Fηk , in b, in μm2 for the binary cluster in configuration (i), for polarization
of the incident wave parallel to the plane of reference (η = 1), as a function of ϑI for the values of
ϕI indicated in the inset in a
(a)0.06 (b)
90° (i)
60° (ii)
30° (iii) 0.1
0° (iv)
0.04
0.08
0.02
0 0.06
0 30 60 90 0 30 60 90
Fig. 6.3 Axial averages of FηT in a, and of Fηk in b in μm2 for the binary cluster in configurations
(i), (ii), (iii), and (iv), for polarization of the incident wave parallel to the plane of reference (η = 1),
as a function of ϑI . The averages over random orientation give Fηk = 0.08543 μm2 both for η = 1
and for η = 2
ior may be important for understanding the possible expulsion of cosmic dust grains
from galaxies and protoplanetary disks [26].
The absorptivity of the particles is important also for understanding their rotational
behavior when driven by the radiation torque. To show this we report in the following
figures the adimensional vector
170 R. Saija et al.
(a) (b)
0.01
Im(ε)=0
Im(ε)=0.001 0.11
0.008 Im(ε)=0.01
Im(ε)=0.1
0.006 0.1
0.004
0.09
0.002
0 0.08
0 30 60 90 0 30 60 90
Fig. 6.4 Dependence on Im(ε) of the axial averages of FηT in a, and of Fηk in b in μm2 for the
binary cluster in configuration (iii) for polarization parallel to the plane of reference (η = 1) as a func-
tion of ϑI . The averages over random orientation of Fηk (in μm2 ) are 0.08595, 0.08621, 0.08499,
and 0.1095 for Im(ε) = 0, 0.001, 0.01, and 0.1, respectively, both for η = 1 and for η = 2. Note
that in a the curves for Im(ε) = 0.0, 0.001, and 0.01 are almost coincident on the scale of the figure
8π k
Tη = Γ Rad η , (6.21)
n 2 II σTη
where Γ Rad η and σTη are the radiation torque and the extinction cross section calcu-
lated for incident light with polarization η, and n = 1. The incident field is circularly
polarized and we show the results for η = 1 only, because, on account of the sym-
metry of the scatterer, the results for η = 2 are identical, except for the sign.
First, we consider a binary aggregate of identical spheres with radius 50 nm,
composed of astronomical silicates [27]. The origin of the frame of reference lies at
the point of contact of the spheres. Since, according to Sect. 6.2.2, the torque depends
on the absorptivity of the particles, we report in Fig. 6.5 the quantity Tηk = Tη · k̂I
as a function of the wavelength for the binary cluster with its axis perpendicular to
kI both in the case in which the imaginary part of the dielectric function is set to
zero or assumes its actual value. As expected, the transverse components of Γ Rad η ,
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 171
1.4 0.2
1.2 0
−0.2
1
−0.4 T1y
0.8 T1x −0.6
0.6 −0.8 T2y
−1
0.4
T2x −1.2
0.2 −1.4
0 −1.6
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
λ (μm) λ (μm)
1.5
1 T1z
0.5
−0.5
−1 T2z
−1.5
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
λ (μm)
Fig. 6.6 Cartesian components of T1 and T2 for the five-spheres cluster whose geometry is given
in Table 6.1. The direction of the incident wavevector is along the z axis
i.e., those in the plane orthogonal to kI , were found to be zero. We do not report the
results for the case in which the axis of the aggregate is parallel to kI , because, in
this case, Γ Rad η · k̂I is nonvanishing only for complex refractive index, and changes
its sign with the change of polarization. Even in this case the transverse components
are rigorously zero for symmetry reasons. Anyway, the results in Fig. 6.5 show the
great importance of the absorptivity of the particles on the value of the torque, and
in our opinion do not deserve further comments.
As a second example we now consider a more complex cluster composed of five
spheres identical to each other and to the spheres we used above to compose the binary
172 R. Saija et al.
cluster. The geometry is chosen so that no symmetry is present, and the coordinates of
the centers of the spheres are listed in Table 6.1. The incident wavevector kI is parallel
to the z axis (k̂I ≡ êz ) and circular polarization is assumed. In Fig. 6.6 we report the
cartesian components of Tη both for η = 1 and for η = 2. We assume here that the
refractive index is just that of astronomical silicates [27]. We first notice that the x and
y components of Γ Rad do not change their sign with changing polarization, unlike
the z component which does change sign. Of course, the lack of symmetry prevents
the curves of each component to coincide with changing polarization. Nevertheless,
the most striking result is the coincidence of the axial average around the z axis,
Γ Rad η · k̂I , with Γ Rad η · k̂I . This result would be quite evident for the axial average
of the binary cluster with its axis along kI . In the present case, because of the lack of
symmetry, the occurrence of this coincidence deserves a few additional comments.
It is not easy to extract any conclusion from the results on the basis of the formulas
of Sect. 6.2.2. Nevertheless, a few heuristic remarks can be drawn with the help of
a complete set of calculations, in the sense that they were also performed using
linearly polarized incident radiation. Both when we assume circular and when we
assume linear polarization, the components of Γ Rad η in a plane orthogonal to kI are
found, in general, nonzero and different from each other, whereas their axial averages
around kI do vanish. Actually, averaging around kI makes the average particle akin
to an axially symmetric particle, for which the mentioned result is to be expected.
Furthermore, assuming circular polarization, we found Γ Rad η · k̂I = Γ Rad η · k̂I
that follows at once, if one considers the behavior of the field of a circularly polarized
wave. On the other hand, in case linear polarization is assumed, the axial average
of Γ Rad η around kI turns out to be independent of η. In fact, as regards the axial
averaging, the different polarization is seen only as a different orientation of the
particle around z . Of course, in general, the axial average Γ Rad η · k̂I = Γ Rad η · k̂I
for linear polarization.
At this stage, since Γ Rad depends on the orientation of the particle, the problem
arises out of the stability of the rotational motion driven by the electromagnetic
torque. Actually, we performed a study of this problem using as a model just the
aggregate whose geometry is given in Table 6.1. The interested reader is referred to
[28] for a detailed discussion of the problem.
The theory of Sect. 6.2 can be extended to deal with the issue of trapping and manip-
ulation of microsized and nanosized particles. The extension is necessary because
the so-called optical trapping occurs when a laser beam is focalized through a high
numerical aperture lens. The radiation force may then trap a particle within the focal
region. At the same time, the torque exerted by the same laser beam may force the
particle to assume particular orientations or to perform rotational motions. Now, a
laser beam is quite different from a plane wave, nevertheless, whatever its transverse
profile, it can be represented as a superposition of plane waves with the same fre-
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 173
quency but with a different direction of propagation. This is the so-called angular
spectrum representation of an optical field. Moreover, when the angular spectrum
representation is considered in the far field zone, one may exploit the geometrical
optics to describe the focalization in terms of rays.
Let us consider a monochromatic field E(r) that propagates within a region of space
that contains no sources and is filled by a homogeneous isotropic medium of real
refractive index n. The propagation of such a field is described by the Helmholtz
equation. Let us now choose a rectangular system of axes, the direction of the z axis
being quite arbitrary, and perform the bidimensional space Fourier transform of E(r)
in a plane orthogonal to z,
∞
1
Ê(k x , k y ; z) = E(x, y, z) exp[−i(k x x + k y y)] dxdy
4π 2
−∞
with inverse
∞
E(x, y, z) = Ê(k x , k y ; z) exp[i(k x x + k y y)] dk x dk y . (6.22)
−∞
it is an easy matter to see that the space Fourier transform in the plane z = 0 is
related to the one in the plane at z = 0 by
∞
E(x, y, z) = Ê(k x , k y ; 0) exp[i(k x x + k y y ± k z z)] dk x dk y , (6.24)
−∞
that is, by definition, the angular spectrum representation of the field E(r). Note that
the ± sign in front of k z is due to the fact that the field can propagate either in the
positive or in the negative z direction. However, henceforth we consider only beams
propagating along the positive z axis, so that only the positive sign will be retained
in front of k z . We stress that, although n is real, k z may be imaginary, so that (6.24)
174 R. Saija et al.
We will now show that the geometrical optics stems from the angular spectrum
representation when we consider the field in the far zone, i.e., at a large distance
from its source, that is assumed to be of finite extent. Let us thus consider an optical
field in the plane z = 0 and let us use the angular spectrum representation to describe
how the field transforms when it is considered in a plane at z = z ∞ at large distance
from the chosen origin. To this end we need to find the limiting expression of (6.24)
for r → r∞ . Let us introduce the unit vector
E∞ (sx , s y , sz ) =
lim Ê(k x , k y ; 0)eikr (k x sx /k+k y s y /k+kz sz /k) dk x dk y . (6.25)
kr →∞
(k x2 +k 2y )≤k 2
We remark that in (6.25) the evanescent waves, due to their fast decay, do not con-
tribute to the field at infinity so that the limits of integration have accordingly been
set. The integral in (6.25) can be performed through the stationary phase method
[29] with the result
eikr
E∞ (sx , s y , sz ) = −2π iksz Ê(ksx , ks y ; 0) . (6.26)
r
The field (6.26) has the typical form of the scattered field in the far zone with Ê
playing the role of the scattering amplitude. Note, however, that the argument of Ê
is parallel to r∞ , so that a single plane wave, the one that propagates along ŝ, enters
the description of the far field. This is enough to justify the use of the geometrical
optics for any optical field. As a consequence, on account that ŝ is fully determined
by k x and k y only, we can write
ir e−ikr
Ê(k x , k y ; 0) = E∞ (k x , k y )
2π k z
Equation (6.27) can now be interpreted as a description of the field in terms of rays.
NA = n sin ϑMax ,
where n is the refractive index of the medium that fills the image space. We consider,
as an example, the focalization of a gaussian TEM00 beam, whose waist, of radius w0 ,
is taken to coincide with the entrance pupil of the lens. Then, according to Novotny
and Hecht [17], the field at any point within the focal region can be written in the
angular spectrum representation as
E(r) = E PW (k̂)ûk̂ eik·r dk x dk y , (6.28)
k x2 +k 2y ≤k⊥
2
where k⊥ = k sin ϑMax , k has polar angles ϑk and ϕk , ûk̂ = û(ϑk , ϕk ); the limits of
integration ensure that only the rays that actually traverse the exit pupil of the optical
system are considered. In (6.28)
eik f nI
E PW (k̂) = E 0 i f (cos ϑk )−1/2 f w ,
2π k n
where n I is the refractive index of the object space and f w is the apodization function
1 sin2 ϑk
f w = exp − 2 2 , (6.29)
f 0 sin ϑMax
w0
f0 = .
f sin ϑMax
The apodization function (6.29) is common also to the higher gaussian modes TEM10
and TEM01 , and when f w = 1, i.e., for w0 → ∞, one recovers the description of
the field of Richards and Wolf [30].
Often, the image space is not filled by a single homogeneous medium but rather
by two homogeneous media, of refractive indexes n and n F , separated by a plane
interface orthogonal to the optical axis. We assume the interface to be located at
z s = −D between the exit pupil and the nominal focus. Hereafter, the quantities
considered in the region z > z s will be characterized by the index F, even when,
strictly speaking, this notation would not be necessary. For instance, since k x and
k y are unaffected by the refraction, we have k x = kFx and k y = kFy , and thus also
k⊥ = kF⊥ . On the contrary, k z is affected by the refraction according to
2 1/2
1/2 nF
kFz = kF2 − k x2 − k 2y = k 2 − k x2 − k 2y .
n
The refraction of the rays through the interface introduces a spherical aberration
and a polarization-dependent transmission that can be taken into account by the
Fresnel coefficients Tη [5]; η = 1 stands for polarization parallel and η = 2 for
polarization perpendicular to the plane of incidence that, for each ray, coincides with
the meridional plane defined by the k or by the kF vector, and the z axis (optical
axis). The decomposition of the polarization vectors of each of the plane waves in
(6.28) into their components parallel and perpendicular to the plane of incidence can
be effected by introducing for each plane of incidence a pair of unit vectors ûηk̂ .
Thus, we have
ûk̂ = ûk̂ · ûηk ûηk = cη ûηk̂ (6.30)
η η
where
E FPW k̂F = exp[iz s (k z − kFz )]E PW (k̂) .
We stress that, although ϑk appears in place of ϑkF in (6.31), Snell’s law grants
an immediate relation between these angles. Moreover, the same law makes the
apodization function (6.29) insensitive to the use of ϑk or of ϑkF (Fig. 6.7).
We are now able to calculate the radiation force that the field exerts on a particle
at O . To this end we rewrite (6.31) as
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 177
E (r ) = E FPW k̂F eikF ·R O
2 +k 2 ≤k 2
kFx Fy F⊥
× cη Tη (ϑk )ûηk̂ eikF ·r dkFx dkFy , (6.32)
F
η
Since the J multipole fields depend on the magnitude of kF only, they can be carried
outside the integral with the result
( p) ( p)
E = Jlm (r , kF )Wlm (R O ), (6.33)
plm
where
( p)
Wlm (R O ) = cη E FPW k̂F eikF ·R O
η 2 +k 2 ≤k 2
kFx Fy F⊥
( p)
×Tη (ϑk )Wlm ûηk̂F , k̂F dkFx dkFy . (6.34)
Of course, (6.33) refers to the case in which there is a plane of separation between
two media of different refractive index, so that the consequent refraction must be
taken into account. It is an easy matter to show that, when n F → n, (6.33) simplifies
into
178 R. Saija et al.
( p) ( p)
E = Jlm (r , k)Wlm (R O ), (6.35)
plm
where
( p) ( p)
Wlm (R O ) = E PW k̂ eik·R O Wlm ûk̂ , k̂ dk x dk y . (6.36)
k x2 +k 2y ≤k⊥
2
According to the case we deal with, (6.33) and (6.35) show that the multipole expan-
sion of the field in the focal region resembles the expansion of a plane wave, whose
( p)
amplitudes Wlm (R O ), according to (6.34) or (6.36) depend on the position of the
particle. Thus, when calculating the radiation force or the radiation torque exerted on
( p)
a particle, we only have to substitute the newly defined amplitudes Wlm into (6.38)
( p)
of Appendix 2 to get the amplitudes of the scattered field Alm , and into (6.11a) and
(6.11b) to get FRad ζ , and into (6.20a) and (6.20b) to get Γμ . In practice, the required
result is still given by (6.11a) and (6.11b) and by (6.20a) and (6.20b) provided that
( p) ( p) ( p) ( p)
E 0 WIlm → Wlm (R O ) , E 0 Alm → Alm .
( p)
The preceding considerations highlight the importance of the quantities Wlm (R O )
that describe the lens, as they depend on the characteristics of the latter. Then, the
integrals (6.34) or (6.36) can be calculated numerically once for all at the nodes of
a suitably chosen grid for a given lens and stored for further use.
The theory we expounded in Sect. 6.2 is, in a sense, nonconventional. In fact, although
both the radiation force and torque were separated into a scattering contribution and
an extinction contribution, these parts have no relation with the customary gradient
contribution and scattering contributions, that, as stated in Sect. 6.2.3, come from the
application of the dipole approximation [17]. Our theory is exact, as it goes beyond
the dipole approximation, its accuracy depending on the use of a sufficiently large
lM to ensure the convergence of the multipole field expansions.
We calculate the trapping position and the stiffness constants (see later) of homo-
geneous spheres of latex (n p = 1.57) of various sizes and compare our results with
the experimental data existing in the literature. To this end, we assumed an aplanatic
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 179
optical system with numerical aperture NA = 1.2 both in case the image space is
filled of water with refractive index n = 1.33 and in case the exit pupil is immersed
in oil with refractive index n = 1.52, whereas the trapping particle is immersed in
water. These media are separated by a cover slip orthogonal to the optical axis located
at z s = −20 μm with respect to the origin at the nominal focus of the lens. Since
NA < 1.33 no plane wave is totally reflected at the separation interface. As the pres-
ence of two media of different refractive indexes introduces a spherical aberration,
the maximum field intensity occurs at a point F with z F ≈ −4.0 μm. The incident
field is assumed to be a TEM00 Gaussian beam, linearly polarized along the x axis,
with filling factor f 0 = 2 and wavelength λ0 = 1064 nm in vacuo, i.e., λ = 800 nm
in water. The contour plot of the focalized field [28] shows that the effect of the
aberration manifests as a large shift of the field maximum and a more accentuated
lack of cylindrical symmetry around the optical axis due to the linear polarization of
the incident beam. These results are in substantial agreement with those reported by
Rohrbach and Stelzer [31].
We now go to consider the trapping of single spheres both in the unaberrated
and in the aberrated field. According to the experimentalists we define the so-called
trapping efficiency as
Q(r) = FRad (r)c/(Pn),
where P is the power of the trapping beam and n is the refractive index of the medium
surrounding the particle. Thus, n = 1.33 in all the cases we deal with. The argument
r = R O (see Fig. 6.1) denotes the position of the particles that is assumed to coin-
cide with their center. In Fig. 6.8 we report, as a representative example of spheres to
which, according to our experience, the Rayleigh or the Born approximations do not
apply, the components of the trapping efficiency for single spheres with diameter of
d = 850 nm both for the case of an unaberrated field and in the case spherical aberra-
tion is present. The trapping occurs where the components of the trapping efficiency
vanish with a negative derivative. Figure 6.8 shows that the spheres we considered
undergo trapping on the optical axis. In fact, both Q x (x, 0, 0) and Q y (0, y, 0) vanish
on the optical axis for evident symmetry reasons. As regards Q z (0, 0, z), we see that
there exist at least one value of z that satisfies the trapping condition. The values of
z at which the trapping occurs both for unaberrated (z 0 ) and for the aberrated field
(z 0a ) in Table 6.2 for spheres of diameter d = 220–1900 nm.
Since the behavior of the components of the trapping efficiency is almost linear in
the vicinity of the trapping point, one is able to define the stiffness of the optical trap
by introducing the constants κx , κ y , and κz such that the components of the radiation
force, in the absence of aberration, can be written as
in the neighborhood of the point where they vanish. In case aberration is present we
must put
Fz = −κz (z − z 0a ) .
180 R. Saija et al.
80 40
d= 850 nm d = 850 nm
60 30
Trapping Efficiency
Trapping Efficiency
40 20
20 10
0 0
−20 −10
−40 −20
−60 −30
−80 −40
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2
Lateral Position (μm) Lateral Position (μm)
Fig. 6.8 Q x (x, 0, 0) (blue solid line), Q y (0, y, 0) (red dashed line), and Q z (0, 0, z) (green dotted
line) for single spheres with diameter d = 850 nm both in unaberrated gaussian (left panel) and
aberrated gaussian beam (right panel). Note that the origin of the z axis in the right panels is shifted
by −z F = 4.0 μm
A good parameter to be compared with the experimental data is the so-called stiffness
asymmetry factor sl = 1 − κx /κ y that we report, for all the spheres we dealt with,
in Table 6.2 both for the unaberrated field (sl ) and for the aberrated field (sla ). The
experimental values of the asymmetry factor were taken from the paper of Rohrbach
[32] for the unaberrated field and from the paper of Zakharian et al. [33] for the
aberrated field. We note that the size parameter x = π nd/λ0 for the spheres con-
sidered in Table 6.2 goes from 0.6 to 5.61 in vacuo, i.e., from 0.80 to 7.46 in water,
so that we had to use up to lM = 12 to get full convergence of all the calculated
values [34]. Table 6.2 seems to show a fair agreement between theory and experi-
ment. This agreement seems confirmed by Fig. 6.9, where we report the curves of
the asymmetry factor calculated both for unaberrated and for the aberrated field,
together with the experimental values for the spheres used by Rohrbach (circles) and
by Zakharian et al. (dots). Nevertheless, we call the attention on the values of the
asymmetry parameters for the spheres with d = 850 and 1660 nm. Actually, these
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 181
1−κx/κy
et al. [33] (red dots) 0
−0.1
−0.2
−0.3
−0.4
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Sphere diameter (μm)
values are favorably located on at least one of the two theoretical curves we draw
in Fig. 6.9. We stress, however, that Rohrbach declares the experimental values at
d = 850 and 1660 nm as obtained in the absence of aberration [32], whereas in
Fig. 6.9 they appear to be on the curve for the aberrated field. On the other hand,
the experimental values of Zakharian et al. appear very close to our curve for the
aberrated field, that has been used in [33]. In this respect, it may be interesting to
notice that the values of sla measured by Zakharian et al. are not located on their
theoretical curve (see Fig. 6.12 of [33]). This discrepancy, in our opinion, may be due
to an inadequate consideration of the effect of aberration, and/or to the fact that with
so large a diameter any approximation used to calculate the Lorenz force density
does not apply. Moreover, the diameter of these spheres is larger than the size of the
trap, that, according to our calculations, spans ≈λ = 800 nm so that one must be
very careful both in computations and in the interpretation of the experimental data
when dealing with such large objects.
In this section we calculate the optical forces on gold nanospheres with the purpose of
checking up to what size our theoretical approach foresees the trapping of spherical
particles of noble metals. To this end, we adopt even in this section the geometry that
is sketched in Fig. 6.1.
A critical point in the calculation of the radiation force, whatever the approach
one chooses to use, is the knowledge of the dielectric properties of the particles.
Here, we use the dielectric function tabulated by Johnson and Christy [35], which
yields a good agreement between calculated and experimental extinction spectra of
such nanoparticles. The main feature for gold spheres is the structure of the plasmon
resonances whose position and complexity depends on the radius of the particles. In
this respect, we notice that at λ = 1064 nm, the wavelength used by Hansen et al.
182 R. Saija et al.
−1
10
−2
10
κz/P −3
10
Au nanosphere into a
steam bubble
−4
10
−1
10
κx/P
−2
10
−3
10 Au nanosphere into a
steam bubble
−4
10 1 2
10 10
Radius (nm)
Fig. 6.10 Experimental (from [36]) and calculated stiffness κz /P and κx /P in pN/(nm · W) for
gold spheres, both for water immersion lens (red points and lines) and for oil immersion lens (blue
points and lines). The points marked by an arrow refer to our calculations for unaberrated field on
particles embedded into a steam bubble
radiation force both in the case of unaberrated field and, in the case of an aberrated
field, for different values of D, in order to see to what extent the aberration plays a
significant role in degrading the trapping in the axial direction. In agreement with
[36], we found that the more the field is aberrated the more its ability to trap relatively
large particles weakens, perhaps by reduction of the trapping region. As regards the
behavior of the radiation force within that region, it strongly suggests that a given
model with parameters chosen once for all cannot explain the trapping of the larger
particles, i.e., of spheres with r > 77 nm: the theory is linear in power, indeed. We
must thus consider the possibility that the parameters of the model, essentially the
refractive index, may change with increasing power. In this respect, it is quite natural
to suspect the heating both of the medium and of the particles yielded by the laser
beam. Therefore, we first considered the study of Liz-Marzán and Mulvaney [38] who
measured the change of the dielectric properties of Au colloids, more precisely of the
surface plasmon absorption, in the temperature range of 14–70◦ C. These authors,
after carefully considering all the physical effects that stem from the increase in
temperature, were led to the conclusion that the most important factor is the change
of resistivity of gold with a consequent change of its refractive index. Nevertheless,
the experiments of Liz-Marzán and Mulvaney were performed on Au spheres with
an average diameter d = 15 nm. When the corrections that they suggest on the basis
of the findings of Doremus [39, 40] and of Kreibig [41] (these corrections depend
on the size of the particles) are calculated for spheres with a radius r ≥ 77 nm, the
change of the refractive index turns out to be negligible and leads to no increase in
the radius of the trappable particles.
184 R. Saija et al.
Table 6.3 Temperature change with respect to room temperature for the Au nanospheres used in
[36] at their trapping position. r is the radius (in nm) of the spheres used in [36], σabs (in μm2 ) is
the absorption cross section, and T /P is the change of temperature (in K/W) due to heating by
the trapping beam. T has been estimated using the formulas in [42] at 5 nm from the surface of
the nanospheres
r σabs T /P
15 3.48 × 10−6 36.43
25 1.87 × 10−5 130.46
35 6.01 × 10−5 313.75
45 1.47 × 10−5 614.78
50 2.15 × 10−4 817.74
77 1.05 × 10−3 1679.90
98 2.50 × 10−3 5072.50
127 4.61 × 10−3 7303.20
Next, we considered that, according to Hansen et al., the power needed to trap
the larger Au spheres was of 135 mW, and that Seol et al. [42] in their study of
the heat developed by the trapping beam, estimate an increase in temperature at the
surface of Au nanospheres as high as 260◦ C/W. Much lower temperatures were
estimated by Peterman et al. [43] at the surface of dielectric particles, but in a recent
paper Lapotko [44] demonstrated that the laser-induced heating around plasmonic
nanoparticles, in particular Au nanospheres, may excite detectable vapor bubbles.
These studies suggested us to use the formulas reported by Seol et al. to estimate the
change in temperature, with respect to room temperature, at 5 nm from the surface
of the Au nanospheres considered in [36]. The results of our estimate are reported in
Table 6.3 where we also reported the absorption cross sections that, unlike [42], we
calculated without resorting to the dipole approximation. The other parameters were
taken as reported by Seol et al. because they are compatible even with our focalized
beam.
The increases in temperature reported in Table 6.3 should be taken with some
caution as they are due to a simplified model of the thermal equilibrium of the naked
metal spheres and the surrounding water, yielding perhaps too large values for the
largest Au spheres. Anyway, these increases turn out to be large enough to justify
the working hypothesis that the particles, especially the largest ones, may become
embedded into a steam bubble. In turn, the presence of a steam bubble, provided
it is stable enough, will attenuate the effect of the metallic nature of the particles.
Of course, we kept the thickness of the steam layer, which we assumed to have the
refractive index n steam = 1, as small as possible, for the sake of stability, and made
also the assumption that the particle remains steadily at the center of the bubble. The
spheres with r = 98 nm are surrounded by a steam bubble with thickness 100 nm,
and those with r = 127 nm are surrounded by a bubble with thickness 130 nm. We
stress that the total diameters (particle plus bubble) are smaller than the width of the
trapping spot and of the resolution of the imaging system used in the experiments.
Thus we calculated the field scattered by the stratified particle that includes the metal
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 185
nanosphere surrounded by the steam layer and embedded into water. This has been
done by exploiting the procedure that we developed to study the resonances of layered
metal spheres [45]. The resulting scattered field, when introduced into the Maxwell
stress tensor, yields through (6.1) the radiation force acting on the center of mass of
the particle. We also calculated the trapping position for spheres with r = 77 nm
surrounded by a steam layer with thickness 5 nm in order to see whether the presence
of the bubble improves the trapping. All these calculations were performed for the
unaberrated field, the one able to trap the largest particles we are considering. In fact,
in our calculations the steam layer grants stable trapping (at negative values of z)
also for the r = 98 and r = 127 nm spheres. We collect in Table 6.4 all the trapping
positions for the unaberrated field together with the thickness of the steam layer
enveloping the largest spheres. As regards the entries in Table 6.4 one may wonder
about the relevance of z E . For the Au spheres we considered here that absorption
and extinction prevail over scattering. Thus, according to Sect. 6.2.3 we found that
(Sca)
for the spheres with r ≤ 77 nm FRad z as a function of z within the focal region is, in
(Ext)
general, two orders of magnitude smaller than FRad z . As a consequence, it is mainly
(Ext)
FRad z that determines the trapping near the point where it vanishes. However, for
r = 77 nm at that point the zero is in no way a sharp one, and for r > 77 nm for
(Ext)
FRad z there is no zero point at all. Adding the steam layer results in a slightly less
(Ext)
stable trapping of the r = 77 nm spheres but makes a reshaping of FRad z for the
(Ext)
r = 98 and for the r = 127 nm spheres, so that FRad z has again a zero point for
stable trapping. In conclusion, the z E entry shows that for the Au spheres the trapping
(Ext)
is due to the behavior of FRad z.
The stiffnesses of the trap for spheres into a steam bubble are reported and marked
by arrows in Fig. 6.10. We see at once that our calculated values of κz /P, rather
surprisingly, coincide with the experimental values. In particular, the addition of the
bubble to the 77 nm spheres produces little change of κz /P. As for κx /P, we see that
for 77 nm spheres it lies on the calculated curve, whereas, for the largest spheres, it
lies on a line parallel to the experimental data.
186 R. Saija et al.
Water
Oil
High NA Lens
In the last few years there has been a growth of interest in trapping and manipulating
elongated nanostructures such as nanowires and fibers. Indeed, these structures are
characterized by a large aspect ratio and a subwavelength diameter that makes them
suitable for several applications such as nanoprobe microscopy and spectroscopy
[46]. However, for these applications to take hold it is of fundamental importance
to understand the optical trapping properties of quasi-1D nanostructures, to evaluate
quantitatively radiation force and torque acting on them and eventually to shed light
on their dynamics within the optical trap.
In this section we study the torque exerted on optically trapped linear nanostruc-
tures that we model as linear aggregates of identical spheres of diameter d and length
L. The geometry of the problem is sketched in Fig. 6.11 that is a specialized version
of Fig. 6.1.
The setup we assume is the aberrated one that we considered in Sect. 6.6.1. The
L and d values of the particles we deal with were chosen so as to highlight the
orientational behavior of the chains, as we will show below. We also investigate long
chains (up to 20 spheres) with d = 50 nm in order to compare our calculated results
with some experimental data available for semiconductor nanowires [46, 47]. In fact,
the d = 50 nm spheres are much smaller than the wavelength, so that these chains
are a good model for quasi-1D nanostructures [10, 48].
We consider linear chains composed of latex spheres (refractive index n p = 1.57),
with diameter 220, 100, and 50 nm, and for each chain we calculate the radiation
force FRad (r), the argument r denoting the position of O , the center of mass of
the chain. The trapping occurs on the optical axis [31] where all the components of
FRad (r) vanish with a negative derivative, and in the vicinity of the trapping point
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 187
undergo trapping, whatever their tilt angle, and that the trapping position is almost
independent of ϑ. Moreover, Fig. 6.13a, in which we report Ty as a function of ϑ
for these chains, shows that for the 2-spheres chains Ty > 0, whereas Ty < 0 for
the 6-spheres chains. Thus, the shorter chains tend to orient along the polarization
axis, whereas the longer ones tend to align with the optical axis. The chains with
L = 400 nm mark a change in the sign of Ty , hence a change in the orientational
stability.
The behavior of the chains with diameter 50 nm is quite analogous, i.e., whatever
their length they undergo trapping. According to Fig. 6.13b, in which we report Ty as
a function of ϑ, for the shorter chains Ty > 0, whereas for the longer ones Ty < 0.
The point of inversion lies somewhere between the 8 and the 9-spheres chains, i.e.,
at L ≈ 400 nm.
The results discussed above show that the trapping of our model linear chains
presents features that are akin to those observed for trapped nanowires. In fact,
force constants and torque we calculated for several materials for long chains of
spheres modeling nanowires are in qualitative agreement with the few available
and comparable experimental results. For instance, calculations confirmed trapping,
alignments along the optical axis, and linear dependence of κz on laser power, as
expected according to [47].
6.8 Appendix 1
As anticipated in Sect. 6.2, with the help of the asymptotic multipole expansions it
is easily proved that the dyadic terms
r̂ · n 2 (E ⊗ E∗ ) + (B ⊗ B∗ ) = n 2 r̂ · E E + r̂ · B B
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 189
give a vanishing contribution at the radiation force. In fact, a look to (6.8a) and
(6.8b) shows that the asymptotic multipole expansions of EI and ES contain the
( p)
transverse harmonics Zlm (r̂ ) which, according to their definition are orthogonal,
in the ordinary vector sense, to r̂ . This decrees the vanishing of the contribution of
( p)
the dyadic terms for whatever form of the incident amplitudes WIlm , even when the
( p)
latter are substituted by the Wlm (R O ).
Now, we show how it happens that the terms EI · E∗ ∗
I and BI · BI give a vanishing
contribution to the radiation force even when the incident field is not a single plane
wave but rather a superposition of plane waves with the same magnitude of k but
different direction of propagation, i.e., different k̂.
Let us thus assume that the incident electric field is a superposition of plane waves
of the kind of (6.28). Thus a typical term that would enter (6.1) is
⎡ ⎤
I E = Re ⎣r 2 n 2 EI (k̂) · E∗
I k̂ r̂ d
⎦
,
where EI k̂ = E PW
k̂ ûk̂ exp(ik · r ) = EPW k̂ exp(ik · r ). An analogous
term I B comes from BI . Since r is large, we can use the asymptotic form of a plane
wave [7, 8] so that I E becomes
4π n 2
I E = Re EPW k̂ · E∗
PW k̂
k2
A quite similar expression, except for the absence of n, is obtained for I B , so that
collecting all the terms we get
I = IE + IB
4π
= Re 2 n 2 EPW k̂ · E∗ PW k̂ + B
PW k̂ · B ∗
PW k̂
k
×k̂ δ k̂ + k̂ exp(2ikr ) − exp(−2ikr )
190 R. Saija et al.
which is easily seen to vanish, even when k̂ = −k̂, on account that
BPW (k̂) = −in k̂ × EPW k̂ .
Similar conclusions can be reached, starting from (6.31), for the case of an aberrated
laser beam.
( p)
Finally we note that, on account of the definition of the amplitudes Wlm (R O )
( p)
and of the related amplitudes Alm , in analogy to Sect. 6.2.2, the radiation torque
( p) ( p )∗
Γ Rad does not get contributions from terms of the form Wlm (R O )Wl m (R O ).
In conclusion, neither a plane wave field nor a focal field gives direct contributions
to the radiation force or torque, as they are just the fields in the absence of particles.
6.9 Appendix 2
In Sect. 6.1 we stated that the calculation of FRad and Γ Rad requires the knowledge of
the field scattered by the particles. We solve the scattering problem for the particles
concerned by expanding both the incident and the scattered field in a series of vector
multipole fields (see (6.6) and (6.7)). Then, according to Waterman [3] we relate the
( p) ( p)
multipole amplitudes, the incident field WIlm , to those of the scattered field Alm , by
the equation [10]
( p) ( pp ) ( p )
Alm ûI , k̂I = Slml m WI l m ûI , k̂I , (6.38)
p l m
that defines the elements of the transition matrix of the particle. Let us stress that, in
principle, the sums in (6.6) and (6.7) and thus also (6.38) include an infinite number
of terms. In practice, for computational reasons, these sums must be truncated to
some suitable lM chosen so as to ensure a fair description of the fields. Therefore, all
sums over the multipole order l should be understood to extend up to l = lM . Since
the particles we deal with either are, or can be modeled as aggregates of spheres,
( pp )
we calculate Slml m for such aggregates by inverting the matrix of the linear system
that is obtained by imposing to the fields the boundary conditions across each of the
spherical surfaces [10, 49]. The order of the matrix to be inverted is 2NlM (lM + 2),
where N is the number of the spheres of the aggregate. The convergence of such
kinds of calculations, i.e., the choice of the appropriate lM is studied in Refs. [34].
A comprehensive treatment of all the topics mentioned above related to the calcula-
tion of the transition matrix can be found in Ref. [10]. It may be useful to note that
for a homogeneous sphere
( pp ) ( p)
Slml m = Sl δ pp δll δmm ,
6 Optical Force and Torque on Single and Aggregated Spheres: The Trapping Issue 191
(1) (2)
and Sl = bl , Sl = al , al and bl being the well-known Mie coefficients [4, 10].
The transition matrix is related to the scattering amplitude of the particle f(k̂S , k̂I ; ûI )
through the equation
1
( p) ( p)
f k̂S , k̂I ; ûI = (−i)l+ p Zlm k̂S Alm ûI , k̂I (6.39)
k
plm
where the amplitudes of the scattered field are given by (6.38). Equation (6.39) thus
allows to express the scattering cross section and the extinction cross section in terms
of the elements of the transition matrix.
The usefulness of the transition matrix stems from the its transformation properties
under rotation of the coordinate frame. In fact, if the frame Σ attached to the particle
rotates to an orientation characterized by the Eulerian angles α, β, and γ , for short,
the elements of the transition matrix become
( pp )
( pp ) (l )
Slml m = (l)∗
Dmμ ()S¯lml m Dm μ () (6.40)
μμ
( pp )
where S¯lml m denotes the elements of the transition matrix calculated in the frame
Σ so that they are independent of the orientation of the particle. As a result, the ele-
ments of the transition matrix depend on the orientation through the rotation matrices
D (l) only. Then, the averages over the orientation of the particles for all the physically
significant quantities result in integrals of products of up to four elements of rotation
matrices further multiplied by P(), the latter being the appropriate function that
describes the distribution of orientations. There are relevant cases, e.g., the case of
random orientations (random average), or the case of random orientation around a
fixed axis (axial average), in which the integration above can be performed analyt-
ically. These averaging procedures, although straightforward, yield rather complex
formulas so that we refer the interested reader to the details that are fully expounded
elsewhere [9, 10, 50].
References
1. R.Y. Chiao, N.C. Wickramasinghe, Mon. Not. R. Astr. Soc. 159, 361 (1972)
2. J. Cho, A. Lazarian, Astrophys. J. 631, 361 (2005)
3. P.C. Waterman, Phys. Rev. D 3, 825 (1971)
4. G. Mie, Ann. Phys. 25, 377 (1908)
5. J.D. Jackson, Classical electrodynamics, 2nd edn. (Wiley, New York, 1975)
6. R.N.C. Pfeifer, T.A. Nieminen, N.R. Heckenberg, H. Rubinsztein-Dunlop, Rev. Mod. Phys.
79, 1197 (2007)
7. D.S. Saxon, Phys. Rev. 100, 1771 (1955)
8. M.I. Mishchenko, J. Quant. Spectrosc. Radiat. Transfer 70, 811 (2001)
9. R. Saija, M.A. Iatì, A. Giusto, P. Denti, F. Borghese, J. Quant. Spectrosc. Radiat. Transf. 94,
163 (2005)
192 R. Saija et al.
10. F. Borghese, P. Denti, R. Saija, Scattering from model nonspherical particles, 2nd edn.
(Springer, Heidelberg, 2007)
11. E. Fucile, F. Borghese, P. Denti, R. Saija, O.I. Sindoni, IEEE Trans. Antennas Propag. AP 45,
868 (1997)
12. F. Borghese, P. Denti, R. Saija, M.A. Iatì, Opt. Express 15, 11984 (2007)
13. F. Borghese, P. Denti, R. Saija, M.A. Iatì, Opt. Express 15, 14618 (2007)
14. M.E. Rose, Elementary theory of angular momentum (Wiley, New York, 1957)
15. M. Abramowitz, I.A. Stegun, Handbook of mathematical functions (Dover Publications, New
York, 1972)
16. P.L. Marston, J.H. Crichton, Phys. Rev. A 30, 2508 (1984)
17. L. Novotny, B. Hecht, Principles of nano-optics (Cambridge University Press, New York, 2007)
18. F. Borghese, P. Denti, R. Saija, M.A. Iatì, Opt. Express 14, 9508 (2006)
19. F. Borghese, P. Denti, R. Saija, M.A. Iatì, Opt. Express 15, 6946 (2007)
20. G. Wurm, M. Schnaiter, Astrophys. J. 567, 370 (2002)
21. J. Klačka, M. Kocifai, J. Quant. Spectrosc. Radiat. Transf. 70, 595 (2001)
22. E.M. Purcell, Astrophys. J. 231, 404 (1979)
23. B.T. Draine, J.C. Weingartner, Astrophys. J. 470, 551 (1996)
24. R. Saija, M.A. Iatì, A. Giusto, F. Borghese, P. Denti, S. Aiello, C. Cecchi-Pestellini, Mon. Not.
R. Astr. Soc. 341, 1239 (2003)
25. X. Ma, J.Q. Lu, R.S. Brock, K.M. Jacobs, P. Yang, X.H. Hu, Phys. Med. Biol. 48, 4165 (2003)
26. M.J. Greenberg, F. Ferrini, B. Barsella, S. Aiello, Nature 327, 214 (1987)
27. B.T. Draine, H.M. Lee, Astrophys. J. 285, 89 (1984)
28. F. Borghese, P. Denti, R. Saija, M.A. Iatì, Opt. Express 15(14), 8960 (2007)
29. L. Mandel, E. Wolf, Optical coherence and quantum optics (Cambridge University Press, New
York, 1995)
30. B. Richards, E. Wolf, Proc. Roy. Soc. (London) 253, 358 (1959)
31. A. Rohrbach, E.H.K. Stelzer, Appl. Opt. 41, 2494 (2005)
32. A. Rohrbach, Phys. Rev. Letters 95, 168102 (2005)
33. A.R. Zakharian, P. Polynkin, M. Mansuripur, J.V. Moloney, Opt. Express 14(8), 3660 (2005)
34. R. Saija, M.A. Iatì, P. Denti, F. Borghese, A. Giusto, O.I. Sindoni, Appl. Opt. 42, 2785 (2003)
35. P.B. Johnson, R.W. Christy, Phys. Rev. B 6(12), 4370 (1972)
36. P.M. Hansen, V.K. Bhatia, N. Harrit, L. Oddershede, Nano Letters 5, 1937 (2005)
37. S.N.S. Reihani, L.B. Oddershede, Opt. Lett. 32, 1998 (2007)
38. L.M. Liz-Marzán, P. Mulvaney, New J. Chem. 22, 1285 (1998)
39. R.H. Doremus, J. Chem. Phys. 40, 2389 (1964)
40. R.H. Doremus, J. Chem. Phys. 41, 3259 (1964)
41. U. Kreibig, J. Phys. (Paris) C2, 97 (1977)
42. Y. Seol, A.E. Carpenter, T. Perkins, Opt. Lett. 31, 2429 (2006)
43. E.J.G. Peterman, F. Gittes, C.F. Schmidt, Biophys. J. 1984, 1308 (2003)
44. D. Lapotko, Opt. Express 17, 2538 (2009)
45. F. Borghese, P. Denti, R. Saija, G. Toscano, O.I. Sindoni, J. Opt. Soc. Am. A 4, 1984 (1987)
46. Y. Nakayama, P.J. Pauzauskie, A. Radenovic, R.M. Onorato, R.J. Saykally, J. Liphardt, P. Yang,
Nature 447, 1098 (2007)
47. P.J. Pauzauskie, A. Radenovic, E. Trepagnier, H. Shroff, P. Yang, J. Liphardt, Nature Materials
5, 97 (2006)
48. R. Saija, M.A. Iatì, P. Denti, F. Borghese, O. Sindoni, Appl. Opt. 40, 5337 (2001)
49. R. Saija, M.A. Iatì, F. Borghese, P. Denti, S. Aiello, C. Cecchi-Pestellini, Astrophys. J. 559,
993 (2001)
50. F. Borghese, P. Denti, R. Saija, M.A. Iatì, O.I. Sindoni, J. Quant. Spectrosc. Radiat. Transf. 70,
237 (2001)
Chapter 7
Rainbows, Coronas and Glories
Philip Laven
Abstract Rainbows, coronas and glories are examples of atmospheric optical phe-
nomena caused by the scattering of sunlight from spherical drops of water. It is
surprising that the apparently simple process of scattering of light by spherical drops
of water can result in this wide range of colourful effects. However, the scattering
mechanisms are very complicated. Eminent scientists (such as Descartes, Newton,
Young, Airy and many others) offered various explanations for the formation of
rainbows—thus making major contributions to our understanding of the nature of
light. The basic features of rainbows can be explained by geometrical optics but, in the
early 1800s, supernumerary arcs on rainbows provided crucial supporting evidence
for the wave theory of light. In 1908, Mie provided a rigorous (but very complicated)
solution to the problem of scattering of light by spherical particles. More than 100
years later, Mie’s solution can now be used to produce excellent full-colour simula-
tions. Examples of such simulations show how the appearance of these phenomena
vary with the size of the water drops, as well as describing the scattering mechanisms
that are responsible for their formation.
7.1 Introduction
The rainbow, corona and glory are examples of atmospheric optical phenomena
caused by the scattering of sunlight from spherical drops of water. Eminent sci-
entists (such as Descartes, Newton, Young, Airy and many others) offered various
explanations for the formation of rainbows—thus making major contributions to our
understanding of the nature of light [1, 3, 7, 37, 61].
Mie’s rigorous solution [46] for the scattering of plane waves from homoge-
neous spheres has been available since 1908 but, for many years, its computational
P. Laven (B)
9 Russells Crescent, Horley RH6 7DJ, United Kingdom
e-mail: [email protected]
W. Hergert and T. Wriedt (eds.), The Mie Theory, Springer Series in 193
Optical Sciences 169, DOI: 10.1007/978-3-642-28738-1_7,
© Springer-Verlag Berlin Heidelberg 2012
194 P. Laven
Fig. 7.1 Mie calculations for scattering of unpolarised monochromatic red light (λ = 650 nm) by
a spherical water drop with radius r = 0.1 µm to r = 1,000 µm for refractive index n = 1.33257
complexity was a formidable obstacle to its practical application, especially for large
spheres. For scattering of light by water drops in the atmosphere, we are concerned
with water drops of radius r < 10 µm in fog to r > 1,000 µm in heavy rain. Rapid
advances in computer power now allow Mie’s solution to be used for such calcula-
tions [15–18, 29, 30, 32, 36, 58–60] and, indeed, computer software for full-colour
simulations of rainbows, coronas and glories is now freely available [11, 35].
Figure 7.1 shows the results of Mie calculations of scattered intensity for light
of wavelength λ as a function of scattering angle θ for monochromatic red light
for spherical water drops of different radius r. Note that θ = 0◦ implies forward
scattering (i.e. in the original direction) and θ = 180◦ implies backscattering (i.e.
back towards the source of the light). The primary and secondary rainbows are fairly
obvious in Fig. 7.1 for large values of r, whereas the corona and the glory appear on
the curve for r = 10 µm as a series of maxima and minima at θ → 0◦ and θ → 180◦
respectively.
Given the availability since 1908 of Mie’s rigorous solution for scattering of light
by homogeneous spherical particles, it might be assumed that everything is now
known about the rainbow, corona and glory. However, as Mie’s solution is based
on the summation of a large number of terms (which converges when the number
of terms is slightly greater than the size parameter x = 2πr/λ), it provides little
information about the scattering mechanisms causing the rainbow, corona and glory.
At this point, it is appropriate to refer to the Debye series (also published in 1908)
7 Rainbows, Coronas and Glories 195
[13, 23] which is essentially a reformulation of Mie’s solution separating the contri-
butions made by different scattering mechanisms of order p, where:
Figure 7.2 shows an example of the various ray paths, as indicated by different
values of p. Unlike calculations based on geometrical optics, the Debye series is
not an approximation: the summation of the Debye series for all integer values of p
from zero to infinity gives the same result as Mie’s solution. Figure 7.3 compares
the results of Mie and Debye series calculations for scattering of red light from a
water drop of radius r = 100 µm. The results of Mie calculations are not easy to
understand, whereas calculations using the Debye series reveal the intricacy of the
scattering processes. For example, the Debye series calculations in Fig. 7.3 show that
p = 2 rays are responsible for the primary rainbow at θ ≈ 139◦ and that p = 3
rays are responsible for the secondary rainbow at θ ≈ 128◦ . Although such results
are well known from geometrical optics, geometrical optics fails to predict that the
appearance of rainbows varies with the radius r of the water drops (as is evident from
the Mie results shown in Fig. 7.1). Similarly, Fig. 7.3 shows that the large variations
in intensity in the Mie results at θ ≈ 68◦ coincide with the crossing point of the
p = 0 and p = 1 contributions in the Debye series calculations. In this area, the
intensities of the two components are similar, but their relative phases vary rapidly
as a function of θ thus producing the complicated interference pattern at θ ≈ 68◦ .
196 P. Laven
Fig. 7.3 Comparison of Mie calculations (top) and Debye series calculations (bottom) for scattering
of monochromatic red light (λ = 650 nm) from a spherical water drop with radius r = 100 µm for
refractive index n = 1.33257 (perpendicular polarisation)
7 Rainbows, Coronas and Glories 197
7.2 Rainbows
Rainbows have long played an important role in the development of scientific knowl-
edge. Descartes and Newton used geometrical optics to explain the formation of rain-
bows. In discussing Figs. 7.1 and 7.3, the term rainbow has been used to denote the
concentration of light in a narrow angular zone (such as 138◦ < θ < 140◦ ) despite
the fact that these graphs are concerned with the scattering of monochromatic light.
To produce simulations of natural rainbows caused by the scattering of sunlight from
water drops, Mie calculations must be performed at many wavelengths across the
visible spectrum, taking account of the spectrum of sunlight and of the variations in
the refractive index n of water.
According to geometrical optics, the rainbow angles θrainbow for p = 2 and p = 3
can be calculated according to the following formulae [2]:
Fig. 7.4 Two p = 2 geometric rays contributing to scattering at θ = 141◦ by a water drop
with refractive index n = 1.33257—in this ◦
case, the geometric rainbow angle θrainbow = 137.9
corresponds to an impact parameter b = arccos[(n 2 − 1)/( p 2 − 1)] = 0.861 (i.e. between the
two rays shown)
θ > θrainbow have different optical path lengths, there is a phase difference between
the two rays. Constructive interference occurs when this phase difference is a mul-
tiple of 360◦ (leading to a maximum of intensity), whilst destructive interference
occurs when the phase difference is an odd multiple of 180◦ (leading to a minimum
of intensity). The maxima and minima in the scattered intensity as a function of θ
7 Rainbows, Coronas and Glories 199
Fig. 7.5 Scattering of sunlight by a spherical drop of water with radius r = 100 µm
occur because the phase difference varies with θ . From a historical perspective, it is
noteworthy that the appearance of supernumerary arcs on natural rainbows was used
by Young to support the wave theory of light [61].
Figure 7.5 shows Mie calculations for scattering of sunlight by a spherical drop
of water with radius r = 100 µm. The graph shows that the intensity of the pri-
mary rainbow is about one order of magnitude greater than that of the secondary
rainbow, whilst the horizontal coloured bars immediately above the graph show sim-
ulations of the rainbows—with the top bar showing perpendicular polarisation, the
middle bar showing parallel polarisation and the bottom bar showing unpolarised
light. Both rainbows are strongly polarised, with perpendicular polarisation being
dominant in both cases. Note that the maxima of the primary rainbow’s supernu-
merary arcs for parallel polarisation coincide with the minima for perpendicular
polarisation (and vice versa) [27]. The reason for this extraordinary behaviour is
that, for parallel polarisation, an abrupt phase change of 180◦ (due to the Brewster
angle) occurs when the impact parameter b = sin[arctan[n]]. For n = 1.33257, this
critical value occurs when b = 0.7998. Figure 7.4 shows that, for θ = 141◦ , one
of the rays is at b < 0.7998, whilst the other is at b > 0.7998. This means that
the phase difference between the two rays for parallel polarisation is 180◦ greater
than the phase difference for perpendicular polarisation—thus explaining the coinci-
dence of the maxima and minima of the supernumerary arcs. In fact, the real coinci-
dence is that critical impact parameter b = sin[arctan[n]] and the impact parameter
b = arccos[(n 2 − 1)/( p 2 − 1)] corresponding to θrainbow are very similar for
p = 2 and for the refractive index n of water.
Figure 7.6 shows a series of simulations of the rainbows caused by the scattering
of sunlight by spherical water drops of various values of r between 10 and 500 µm.
For very small water droplets (e.g. for r = 10 µm), the rainbow (or fogbow) is
almost white and not well-defined. For r = 50 µm, the primary rainbow is predom-
200 P. Laven
Fig. 7.6 Simulations of the primary and secondary rainbow caused by scattering of sunlight by
spherical water drops of specified radius r
inantly white with a hint of red at its outer edge. The colours of the primary rainbow
become more obvious for larger drops, together with the fainter secondary rain-
bow and Alexander’s dark band. Supernumerary arcs can be seen inside the primary
rainbow for r = 50 µm, r = 100 µm and r = 200 µm, as well as on the outside
of the secondary rainbow for r = 100 µm and r = 200 µm. However, the supernu-
merary arcs disappear for larger water drops (e.g. r = 500 µm) because the angular
separation between these arcs is less than the sun’s apparent angular diameter of 0.5◦ .
The Lee diagram [36] in Fig. 7.7 illustrates more generally how the appearance of
rainbows varies with the radius r of the water drops. Each coloured point in Fig. 7.7
7 Rainbows, Coronas and Glories 201
Fig. 7.7 Lee diagram showing variations in the appearance of primary and secondary rainbows
caused by scattering of sunlight by spherical water drops as a function of drop radius from r = 10 µm
to r = 1,000 µm
Fig. 7.8 Left simulation of the primary and secondary rainbows caused by spherical drops of water
with radius r = 100 µm. Right simulation of primary and secondary rainbows caused by spherical
drops of fixed refractive index n = 1.33 across the visible spectrum with radius r = 100 µm
maximum luminance for that value of r . Figure 7.7 shows that the angle of maximum
intensity for the primary rainbow decreases from θ ≈ 142◦ when r = 10 µm to
θ ≈ 138◦ when r = 1,000 µm. Similarly, the angle of maximum intensity for the
secondary rainbow increases from θ ≈ 124◦ when r = 10 µm to θ ≈ 128◦ when
r = 1,000 µm.
As shown in Table 7.1, the variations in the refractive index n of water across the
visible spectrum are relatively small (≈1%). However, these slight variations in n
are responsible for the colours of the natural rainbow. The availability of software
for simulation of rainbows allows us to isolate the effects of these variations in n.
For example, Fig. 7.8 compares natural rainbows caused by scattering of sunlight by
water drops with the hypothetical rainbows that would be caused by drops with a
fixed value of n = 1.33—the hypothetical rainbows are essentially white with tinges
of colour on the supernumerary arcs.
7.3 Coronas
Fig. 7.9 Simulations of the corona (top) and glory (bottom) caused by scattering of sunlight by
spherical water drops of radius r = 10 µm
and that, in general, the outer rings of glories are relatively bright compared with
those of coronas.
Although similar in appearance, these two phenomena are caused by completely
different mechanisms. The corona is generally considered to be due to diffraction
by small droplets of water in the clouds [10, 47, 54, 55, 58, 59]—and thus the
intensity of the corona I (θ ) is calculated using the Fraunhofer diffraction equation1 :
2
1 + cos θ J1 (x sin θ )
I (θ ) ∝ x 2
(7.3)
2 x sin θ
where:
• θ is the scattering angle;
• J1 is the first-order Bessel function;
• x = 2πr/λ where r is the radius of the scattering sphere and λ is the wavelength of the
incident light.
1Such equations are frequently attributed to Fraunhofer, but Craig Bohren has pointed out in a pri-
vate communication that, without diminishing the importance of Fraunhofer’s pioneering work in
experimental optics, there is no evidence to suggest that Fraunhofer developed any theoretical treat-
ment of diffraction. Consequently, Fresnel-Fraunhofer-Airy-Schwerd might be a more appropriate
designation. Schwerd’s contribution is described in [22].
204 P. Laven
Fig. 7.10 Lee diagrams showing the corona caused by scattering of sunlight by spherical water
drops as a function of drop radius from r = 0.1 µm to r = 10 µm. a–c show diffraction calculations.
d–f show Mie calculations. The brightness of (b) and (e) has been increased by a factor of ten to
reveal the colours of the corona. The results shown in a and b have been processed to remove the
brightness information, leaving only the saturated colours
The rings of the corona correspond to maxima in the last term of Eq. 7.3: the first
4 maxima occur when x sin θ = 5.136, 8.418, 11.62 and 14.796—corresponding
to sin θ = 0.8174λ/r , 1.3398λ/r , 1.8494λ/r and 2.3549λ/r . If θ is measured
in degrees and r is measured in µm, the inner rings of the corona for red light
(λ = 0.65 µm) are therefore defined by θ1 = 30.5/r , θ2 = 50/r , θ3 = 69/r and
θ4 = 88/r .
There are significant differences between the diffraction model and Mie’s solution
for the corona, as highlighted by van de Hulst [57] and by Lock and Yang [38].
The Lee diagrams for the corona in Fig. 7.10 compare the diffraction model with
Mie calculations for scattering of sunlight from water droplets for 0.1 < r < 10 µm.
Figure 7.10a shows the results of diffraction calculations. As the brightest scattering
7 Rainbows, Coronas and Glories 205
occurs at θ = 0◦ , Fig. 7.10a contains little information because the corona is very
much darker than the forward scattered light (i.e the sun or the moon) and computer
displays cannot reproduce the necessary large dynamic range. To overcome this
problem, the brightness of Fig. 7.10b has been increased by a factor of 10 so that
the top part of is “over-exposed”, allowing the colours of the corona to become
visible. The data from Fig. 7.10a have been re-plotted in Fig. 7.10c so as to remove
the brightness information, instead showing the saturated colour of each pixel. This
form of presentation shows that, according to the diffraction model, the sequence
of colours in the corona is independent of r. Figure 7.10c suggests that no more
than two or three rings of the corona will be visible even under optimum viewing
conditions. Although diagrams based on saturated colours (such as Figs. 7.10c and f)
are useful for comparing colours, it must be emphasised that they do not represent
the appearance of scattered light.
Looking now at the equivalent diagrams (Figs. 7.10d–f) produced using Mie calcu-
lations, the key difference between Fig. 7.10a and d is that, for 0.5 µm < r < 2 µm,
Mie calculations predict uniform bands of colour at scattering angles θ < 8◦ : for
example, red for r = 0.8 µm and for r = 1.6 µm and violet for r = 1 µm.
Comparison of Fig. 7.10b and e shows that Mie calculations produce very com-
plex patterns for r < 3 µm. Comparison of Fig. 7.10c and f indicates that Mie and
diffraction calculations produce similar results for r > 5 µm, but the diffraction
model is totally inadequate for r < 3 µm. Figure 7.10f also confirms the findings of
Gedzelman and Lock [18] who reported:
The sequence of corona colors changes rapidly for small droplets but becomes fixed once
droplet radius exceeds about 6 µm.
The irregular patterns shown in Fig. 7.10 are caused by interference between p = 0
(diffraction) and p = 1 (transmission through the sphere) contributions. Obviously,
the p = 1 contributions are not taken into account in diffraction calculations using
Eq. 7.3.
More generally, the Lee diagrams in Fig. 7.10 for scattering of sunlight show
that the size of the corona is inversely proportional to the radius r of the water
droplets: Fig. 7.10c suggests that the inner 3 red rings occur at θ1 ≈ 16/r , θ2 ≈ 31/r
and θ3 ≈ 47/r . These results for scattering of sunlight should be compared with
those obtained above for monochromatic red light: the red ring at θ1 ≈ 16/r does
not coincide with a maximum of the Fraunhofer equation. The “main beam” of the
diffraction pattern is wider for red light than for the rest of the visible spectrum—thus
producing a red ring that has no equivalent in the diffraction pattern of monochromatic
red light. However, the next two red rings at θ2 ≈ 31/r and θ3 ≈ 47/r for sunlight
are close to the values of 30.5/r and 50/r obtained for monochromatic red light.
The diffraction model is appealing because of its simplicity compared with Mie
calculations, but the above results show that it can give misleading results for the
corona—because the corona is not solely due to diffraction. In any event, the diffrac-
tion model relies on the evaluation of Bessel functions, thus requiring the user either
to use published tables of Bessel functions or to adopt a numerical solution—thus
negating some of its apparent simplicity. Given the ready availability of software for
Mie calculations, it is suggested that the diffraction model should be avoided.
206 P. Laven
Fig. 7.11 Left A glory as seen from a commercial aircraft. Right Mie simulation of this glory
assuming scattering of sunlight by spherical water droplets with radius r = 4.3 µm
7.4 Glories
In the past, sightings of glories were rare and usually associated with the Brocken
Spectre, which refers to the elongated shadows of mountaineers appearing on fog or
clouds beneath them. In such circumstances, a mountaineer might observe that the
shadow of his head was surrounded by a glory (consisting of a series of concentric
coloured rings). Strangely, such coloured rings appear around the shadow of your
head, but not apparently around the shadows of your companions! Nowadays, glories
are seen much more frequently surrounding the shadow of an aircraft on clouds: the
author has taken more than 4,000 pictures of glories whilst travelling on commercial
aircraft within the last 10 years. Figure 7.11 shows one such picture, together with a
matching simulation based on Mie calculations.
The effects of viewing a glory through a vertical polariser (equivalent to the view
seen through polarised sunglasses worn in the usual orientation) are simulated in
Fig. 7.12. The distinctive dark spots immediately above and below the antisolar point
in the right-hand image in Fig. 7.12 were initially predicted by Können [28]. These
theoretical simulations have been confirmed by the publication of several images of
polarised glories [32]. Figure 7.13 offers an explanation of the formation of polarised
glories.
Figure 7.14 is a Lee diagram showing how the appearance of the glory varies with
the radius r of the water drop. The sequence of ring colors is essentially independent
of r . Glories with large rings imply that they are caused by scattering from small
droplets, whereas glories with small rings imply large droplets. To a first approxi-
mation, the radius R of a given ring from the anti-solar point θ = 180◦ is inversely
proportional to r ; for example, the four inner red rings have radii of R1 ≈ 24/r ,
R2 ≈ 37/r , R3 ≈ 56/r and R4 ≈ 75/r , where R is measured in degrees and r is
measured in µm. This very simple relationship allows the values of r of the clouds
to be estimated very easily from images of glories [33].
7 Rainbows, Coronas and Glories 207
Fig. 7.12 The left image shows a simulation of a glory caused by scattering of sunlight from
spherical water droplets with r = 10 µm (the width of the image is about ±5◦ ). The right image
shows the same glory as viewed through a polariser with its transmission axis vertical
(b) (d)
(a) (f)
(c) (d)
Fig. 7.13 An unpolarised glory, shown as image (a), can be separated into two components (b)
and (c), which have polarisations perpendicular and parallel to the scattering plane, respectively.
When viewed through a vertical polariser, (b) is transformed into (d), while (c) is transformed into
(e). Image (d) shows that the polariser suppresses perpendicular polarisation along the vertical line
through the antisolar point, but has no effect along the horizontal line. This is similar to the effect
of viewing the primary rainbow through a polariser, because the primary rainbow is dominated
by perpendicular polarisation. Image (e) shows that the polariser suppresses parallel polarisation
along the horizontal line through the antisolar point. When (d) and (e) are combined, the resulting
image (f) shows the polarised glory with its distinctive dark spots above and below the antisolar
point. Note that the colors of the polarised glory along the horizontal line through the antisolar point
correspond to perpendicular polarisation, whereas those along the vertical correspond to parallel
polarisation. Other orientations obviously involve a mixture of the two polarisations
Most glories seem to be caused by water droplets with 4 < r < 25 µm, thus
overlapping with the white fogbows described in Sect. 7.2. In fact, fogbows can
usually be observed when glories are visible—and vice versa. In practice, it can
be difficult to observe a glory and a fogbow simultaneously, especially as glories
208 P. Laven
Fig. 7.14 Lee diagram showing variations in the appearance of glories caused by scattering of
sunlight by spherical water drops as a function of drop radius r from r = 5 µm to r = 20 µm
7 Rainbows, Coronas and Glories 209
Fig. 7.15 Left image of a glory and a cloudbow. Right Mie simulation assuming scattering of
sunlight by water drops of radius r = 20 µm
are now most frequently seen from inside commercial aircraft where the field of
view is generally very limited. Nevertheless, Fig. 7.15 shows a layer of cloud which
produced a glory at θ > 176◦ and a primary rainbow at θ ≈ 142◦ —it would be more
appropriate to call the latter a “cloudbow”. In practice, such effects are not easy to
observe because white clouds are typically very bright, but one useful tip is to look
for the dark band just inside the fogbow/cloudbow (at θ ≈ 144◦ in Fig. 7.15). Having
found this dark band, it is much easier to identify the white band corresponding to
the primary fogbow/cloudbow.
The Lee diagram in Fig. 7.16 shows that, although the angular size of the coloured
rings of the glory decreases as r increases, the angular size of the fogbow or primary
rainbow (as measured from the anti-solar point) increases with r . Although the rings
of the glory and the fogbow (and its supernumerary arcs) are easily separable when
seen in Fig. 7.16, the simulation in Fig. 7.17 shows a puzzling mixture of glory rings,
supernumerary arcs and fogbows that can be observed when r = 5 µm. Confronted
with such a complex display, it is almost impossible for observers to identify which
are glory rings and which are supernumerary arcs!
Mie’s solution can be used to simulate glories, but it does not offer any expla-
nation for their formation. In his seminal paper A theory of the anti-coronae [56],
van de Hulst suggested that the glory could be explained as the diffraction pattern
corresponding to a toroidal wave front and noted that back-scattering θ = 180◦ by
geometrical p = 2 rays (apart from the trivial case√ of central rays with impact para-
meter b = 0) would require refractive index n > 2 ≈ 1.414. Given that n ≈ 1.333
for water, this explanation left something to be desired! However, in his 1957
book [57], van de Hulst extended his treatment of the glory to propose that it is
caused by p = 2 rays with b = ±1 and that the 14.4◦ gap between θ = 165.6◦ and
θ = 180◦ could be bridged by surface waves, as illustrated in Fig. 7.18. This gap of
14.4◦ is depicted in Fig. 7.18 for simplicity as the last part of the ray path before it
emerges from the sphere, but this gap could be covered by three separate segments
of surface waves covering a total of 14.4◦ . Because of symmetry of the sphere, ray
210 P. Laven
Fig. 7.16 Lee diagram showing variations in the appearance of glories and fogbows caused by
scattering of sunlight by spherical water drops as a function of drop radius from r = 3 µm to
r = 30 µm
paths of the form shown in Fig. 7.18 generate a toroidal wavefront with a diameter
nearly equal to that of the sphere, propagating in the direction θ = 180◦ .
7 Rainbows, Coronas and Glories 211
Fig. 7.17 The left image shows a Mie simulation of scattering of sunlight by spherical water drops
of radius r = 5 µm as would be captured by a camera with a very wide-angle lens: the width of
this image is ≈±50◦ measured from the anti-solar point. The right image shows the same glory as
viewed through a polariser with its transmission axis vertical
include use of the Debye series to analyse the relative importance of various scattering
processes in the formation of the glory. His results indicated that p = 2 contributions
are dominant for droplets with effective radius x = 150, corresponding to r ≈ 13 µm
for scattering of white light.
Despite the success of van de Hulst’s theory for the formation of the glory, various
authors have bemoaned the lack of a simple physical model for the mechanisms
causing the glory. For example, Lynch and Livingston [43] remark:
Although the glory pattern is correctly predicted by Mie theory, a good physical expla-
nation is, in our opinion, lacking. In some way light is back-scattered after traversing the
periphery of the droplet. Examined in detail, each drop is found to shine uniformly around
its edge with an annulus of light that is coherent (the waves are in phase).
In a paper [52] entitled Does the glory have a simple explanation? Nussenzveig
tried to respond to the above requests, but he concluded that:
Mie theory describes the glory by the sum of a large number of complicated terms within
which the physical mechanisms cannot be discerned. CAM [Complex Angular Momentum]
theory brings out the dominant physical effects and provides an accurate representation for
each of them. That it does so by analytic continuation seems inevitable. I know of no other
way of quantitatively representing tunneling.
The above extracts indicate that the glory cannot be explained (even by eminent
scientists), except by “scientific arm waving”. Fortunately, the availability of software
for Mie and Debye series calculations allow us to investigate the glory in considerable
detail. For example, Fig. 7.19 shows that the Mie results are almost identical to the
Debye series p = 2 contributions (i.e. those that have suffered one internal reflection
in the sphere). This indicates that p = 2 scattering is the dominant mechanism for the
formation of the glory—except in the vicinity of θ = 180◦ where p = 11, p = 7 and
p = 6 scattering makes a significant contribution. The coloured rings of the glory
(e.g. between θ = 175◦ and θ = 178◦ ) are obviously due to Debye p = 2 rays. Note
that the colours of the plotted lines represent the saturated colours. In particular,
the p = 0 contribution (caused by reflection from the exterior of the sphere and
by diffraction) is essentially white, but all of the contributions from higher orders
(e.g. p = 6, p = 7, p = 11 and so on) show colours which are essentially identical
to those for p = 2.
Figure 7.20 shows two simulations of the glory caused by scattering of sunlight
from a spherical drop of water with r = 10 µm: the left side shows a simulation
7 Rainbows, Coronas and Glories 213
Fig. 7.19 Mie and Debye series calculations showing backscattering of sunlight by a spherical water
drop of radius r = 10 µm. The labels on the curves indicate the value of p with the symbols ⊥
indicating perpendicular polarisation and // denoting parallel polarisation
using Mie calculations, whilst the right side considers only p = 2 scattering. The
key difference is that, as indicated in Fig. 7.19, Mie calculations predict a bright white
zone near the centre (i.e. θ → 180◦ ). More importantly, both simulations produce
essentially identical sequences of coloured rings, thus confirming that the coloured
rings of the atmospheric glory are primarily caused by light that has suffered only
one reflection within the water drop (i.e. p = 2 scattering).
Nussenzveig also suggested that resonances are important in forming the glory;
for example, in 2003, he wrote [53]:
The glory provides direct and visually stunning experimental evidence of the importance of
resonances and light tunneling in clouds.
214 P. Laven
If the short path involves clockwise deviation, the long path involves counter-
clockwise deviation (and vice versa). The interference pattern is determined by the
phase difference between the short and long paths. Surface waves travel at a speed
determined by the refractive index of the medium (not by the refractive index of
the sphere) [34]. For water droplets in air, the medium has a refractive index of
about 1.003 ≈ 1). Consequently, the difference in path length between the two
p = 2 contributions at θ = 180◦ − δ is approximately the length of the arc of
radius r corresponding to the angle 2δ. Maxima will occur when the path difference
2πr (2δ/360◦ ) = kλ where k is an integer and λ is wavelength of the light. Rear-
ranging this equation gives δ = 90kλ/(πr ). Taking a numerical example, when
λ = 0.65 µm, δ = 18.62k/r (where r is measured in µm and δ in degrees).
7 Rainbows, Coronas and Glories 215
Fig. 7.21 Ray paths contributing to scattering at θ = 175◦ by a water drop with refractive index
n = 1.333. Each of the two edge rays with impact parameters b = 1 and b = −1 suffer one
internal reflection and generate surface waves: in this case, the short path surface wave travels about
10◦ around the circumference of the sphere before exiting at θ = 175◦ , whereas the long path
surface wave travels about 20◦ around the circumference of the sphere, thus being deviated by 185◦
(equivalent to θ = 175◦ )
Fig. 7.22 Debye p = 2 calculations for scattering of light of wavelength λ = 650 nm from a
spherical water drop of radius r = 10 µm assuming an incident Gaussian beam of width w0 = 2 µm
with x0 = 0 and z 0 = 0. The top diagram shows results for y0 = 10 µm (corresponding to
impact parameter b = 1), the middle diagram shows results for y0 = 10 µm and y0 = −10 µm
(corresponding to b = 1 and b = −1), whilst the bottom diagram shows the vector sum of the
results for y0 = 10 µm and y0 = −10 µm
due to the vector sum of these two components is shown in the bottom graph of
Fig. 7.22.
Various authors have examined the effects of illuminating a spherical particle with
very short pulses of light [44, 45]. Bech and Leder [4, 5] extended such work by
producing various diagrams showing how the impulse response of a sphere varies
7 Rainbows, Coronas and Glories 217
Fig. 7.23 Mie calculations of the impulse response of a spherical water drop of radius r = 100 µm
when illuminated by a 50 femtosecond pulse of light of wavelength λ = 650 nm. This false-colour
map shows the intensity of the scattered signal as a function of scattering angle θ and time delay
218 P. Laven
Fig. 7.24 Debye series p = 2 calculations of the impulse response of a spherical water drop of
radius r = 10 µm when illuminated by a 10 femtosecond pulse of light of wavelength λ = 650 nm.
The zone marked A shows scattering of geometric “central” rays with impact parameter b < 0.1.
As b increases, the scattered signal moves through B until it reaches the primary rainbow at C
caused by rays with b ≈ 0.86. As b → 1, the scattered signal moves from C towards D. Between D
and E, the scattered signal is due to short-path surface waves. Between E and F, the scattered signal
is due to long-path surface waves, which decline in intensity towards G. The maxima and minima
near E correspond to the glory, which can be considered as the interference pattern caused by the
short path and long path surface waves
7 Rainbows, Coronas and Glories 219
7.5 Conclusions
hypothesis that surface waves are responsible for the glory, it is necessary to call on
other calculation techniques, such as:
• Mie and Debye series calculations for spherical particles illuminated by Gaussian
beams;
• Mie and Debye series calculations of the impulse response of spherical particles
illuminated by femtosecond pulses of light.
Although Mie’s rigorous solution for the scattering of plane wave light by an
homogeneous sphere has been available for more than 100 years, it is surprising to
realise that rainbows, coronas and glories are still not entirely understood.
References
1. J.A. Adam, The mathematical physics of rainbows and glories. Phys. Rep. 356, 229–365 (2002)
2. J.A. Adam, Geometric optics and rainbows: generalization of a result by Huygens. Appl. Opt.
47, H11–H13 (2008)
3. G.B. Airy, On the intensity of light in the neighbourhood of a caustic. Trans. Camb. Philos.
Soc. 6, 397–403 (1838)
4. H. Bech, A. Leder, Particle sizing by ultrashort laser pulses–numerical simulation. Optik 115,
205–217 (2004)
5. H. Bech, A. Leder, Particle sizing by time-resolved Mie calculations—A numerical study.
Optik 117, 40–47 (2006)
6. C.F. Bohren, D.R. Huffman, Absorption and Scattering of Light by Small Particles (Wiley,
New York, 1983)
7. C.B. Boyer, The Rainbow: From Myth to Mathematics. (Princeton University, Princeton, 1987),
reprint of 1959 Thomas Yoseloff edn
8. H.C. Bryant, A.J. Cox, Mie theory and the glory. J. Opt. Soc. Am. 56, 1529–1532 (1966)
9. H.C. Bryant, N. Jarmie, The glory. Sci. Am. 231, 60–71 (1974)
10. L. Cowley, P. Laven, M. Vollmer, Rings around the sun and moon: coronae and diffraction.
Phys. Educ. 41, 51–59 (2005)
11. Iris software http://www.atoptics.co.uk/droplets/iris.htm (Cited 31 October 2009)
12. J.V. Dave, Scattering of visible light by large water spheres. Appl. Opt. 8, 155–164 (1969)
13. P. Debye, Das elektromagnetische Feld um einen Zylinder und die Theorie des Regenbogens.
Physikalische Zeitschrift 9, 775–778 (1908) [N.B. An English translation of this paper entitled
The electromagnetic field around a cylinder and the theory of the rainbow is available in
Selected Papers on Geometrical Aspects of Scattering, SPIE Milestone Series Volume MS 89
(1993)]
14. T.S. Fahlen, H.C. Bryant, Direct observation of surface waves on droplets. J. Opt. Soc. Am.
56, 1635–1636 (1966)
15. S.D. Gedzelman, Simulating glories and cloudbows in color. Appl. Opt. 42, 429–435 (2003)
16. S.D. Gedzelman, Simulating rainbows in their atmospheric environment. Appl. Opt. 47,
H176–H181 (2008)
17. S.D. Gedzelman, Simulating halos and coronas in their atmospheric environment. Appl. Opt.
47,
H157–H166 (2008)
18. S.D. Gedzelman, J.A. Lock, Simulating Coronas in Color. Appl. Opt. 42, 497–504 (2003)
19. G. Gouesbet, B. Maheu, G. Grehan, Light scattering from a sphere arbitrarily located in a
Gaussian beam, using a Bromwich formalism. J. Opt. Soc. Am. A 5, 1427–1443 (1988)
7 Rainbows, Coronas and Glories 221
20. W.T. Grandy, Scattering of Waves from Large Spheres (Cambridge University, Cambridge,
2001)
21. R. Greenler, Rainbows (Halos and Glories. Cambridge University, Cambridge, 1980)
22. R.B. Hoover, F.S. Harris, Die Beugungserscheinungen: a Tribute to F. M. Schwerd’s Monu-
mental Work on Fraunhofer Diffraction. Appl. Opt. 8, 2161–2164 (1969)
23. E.A. Hovenac, J.A. Lock, Assessing the contributions of surface waves and complex rays to
far-field Mie scattering by use of the Debye series. J. Opt. Soc. Am. A 9, 781–795 (1992)
24. H. Inada, New calculation of surface wave contributions associated with Mie backscattering.
Appl. Opt. 12, 1516–1523 (1973)
25. V. Khare, Short-wavelength scattering of electromagnetic waves by a homogeneous dielectric
sphere. Ph.D. thesis (University of Rochester, Rochester, 1976). This reference may not be
readily available, but the calculation method is summarized in [23]
26. V. Khare, H.M. Nussenzveig, Theory of the glory. Phys. Rev. Lett. 38, 1279–1282 (1977)
27. G.P. Können, J.H. de Boer, Polarized rainbow. Appl. Opt. 18, 1961–1965 (1979)
28. G.P. Können, Polarized light in nature (Cambridge University Press, Cambridge, 1985)
29. P. Laven, Simulation of rainbows, coronas, and glories by use of Mie theory. Appl. Opt. 42,
436–444 (2003)
30. P. Laven, Simulation of rainbows, coronas and glories using Mie theory and the Debye series.
J. Quant. Spectrosc. Radiat. Transf. 89, 257–269 (2004)
31. P. Laven, How are glories formed? Appl. Opt. 44, 5675–5683 (2005)
32. P. Laven, Atmospheric glories: simulations and observations. Appl. Opt. 44, 5667–5674 (2005)
33. P. Laven, Noncircular glories and their relationship to cloud droplet size. Appl. Opt. 47, H25–
H30 (2008)
34. P. Laven, Effects of refractive index on glories. Appl. Opt. 47, H133–H142 (2008)
35. MiePlot software, http://www.philiplaven.com/MiePlot.htm (Cited 31 October 2009)
36. R.L. Lee, Mie theory, Airy theory, and the natural rainbow. Appl. Opt. 37, 1506–1519 (1998)
37. R.L. Lee, A.B. Fraser, The Rainbow Bridge: Rainbows in Art, Myth and Science (Pennsylvania
State University Press, Pennsylvania, 2001)
38. J.A. Lock, L. Yang, Mie theory model of the corona. Appl. Opt. 30, 3408–3414 (1991)
39. J.A. Lock, Contribution of high-order rainbows to the scattering of a Gaussian laser beam by
a spherical particle. J. Opt. Soc. Am. A 10, 693–706 (1993)
40. J.A. Lock, Improved Gaussian beam-scattering algorithm. Appl. Opt. 34, 559–570 (1995)
41. J.A. Lock, Role of the tunneling ray in near-critical-angle scattering by a dielectric sphere.
J. Opt. Soc. Am. A 20, 499–507 (2003)
42. J.A. Lock, G. Gouesbet, Generalized Lorenz—Mie theory and applications. J. Quant. Spectrosc.
Radiat. Transf. 110, 800–807 (2009)
43. D.K. Lynch, W. Livingston, Color and Light in Nature (Cambridge University, Cambridge,
2001)
44. L. Méès, G. Gouesbet, G. Gréhan, Time-resolved scattering diagrams for a sphere illuminated
by plane wave and focused short pulses. Opt. Commun. 194, 59–65 (2001)
45. L. Méès, G. Gouesbet, G. Gréhan, Scattering of Laser Pulses (Plane Wave and Focused Gaussian
Beam) by Spheres. Appl. Opt. 40, 2546–2550 (2001)
46. G. Mie, Beitrage zur Optik trüber Medien, speziell kolloidaler Metallosungen. Ann. Phys.
Leipzig 25, 377–445 (1908)
47. M. Minnaert, The Nature of Light and Colour in the Open Air (Dover Publications, New York,
1954)
48. H.M. Nussenzveig, High-frequency scattering by a transparent sphere. I. Direct reflection and
transmission. J. Math. Phys. 10, 82–124 (1969)
49. H.M. Nussenzveig, High-frequency scattering by a transparent sphere. II. Theory of the rainbow
and the glory. J. Math. Phys. 10, 125–176 (1969)
50. H.M. Nussenzveig, Complex angular momentum theory of the rainbow and the glory. J. Opt.
Soc. Am. 69, 1068–1079 (1979)
51. H.M. Nussenzveig, Diffraction Effects in Semiclassical Scattering (Cambridge University,
Cambridge, 1992)
222 P. Laven
52. H.M. Nussenzveig, Does the glory have a simple explanation? Opt. Lett. 27, 1379–1381 (2002)
53. H.M. Nussenzveig, Light tunneling in clouds. Appl. Opt. 42, 1588–1593 (2003)
54. J.A. Shaw, P.J. Neiman, Coronas and iridescence in mountain wave clouds. Appl. Opt. 42,
476–485 (2003)
55. R.A.R. Tricker, Introduction to Meteorological Optics (American-Elsevier, New York, 1970)
56. H.C. van de Hulst, A theory of the anti-coronae. J. Opt. Soc. Am. 37, 16–22 (1947)
57. H.C. van de Hulst, Light Scattering by Small Particles ( Dover, New York, 1981), reprint of
1957 Wiley edition
58. M. Vollmer, Effects of absorbing particles on coronas and glories. Appl. Opt. 44, 5658–5666
(2005)
59. M. Vollmer, Lichtspiele in der Luft (Elsevier, München, 2006)
60. R.T. Wang, H.C. van de Hulst, Rainbows: Mie computations and the Airy approximation. Appl.
Opt. 30, 106–117 (1991)
61. T. Young, Experiments and calculations relative to physical optics. A Bakerian Lecture read
on November 24, 1803. Phil. Trans. Roy. Soc. 1–16 (1804)
Chapter 8
The Extension of Mie Theory to Multiple
Spheres
Daniel Mackowski
8.1 Introduction
As the very presence of this text makes clear, analytical solutions to light scattering
by objects are both rare and historically significant. Indeed, it can be argued that
Mie’s solution for the sphere is the only analytical light scattering solution for a
3-D object that is routinely applied in scientific and engineering disciplines. Of
course, analytical solutions do exist for other shapes—most notably the ellipsoids—
yet a review of the literature will reveal that most of the theoretical examinations
D. Mackowski (B)
Mechanical Engineering Department, Auburn University,
Auburn, AL 36849, USA
e-mail: [email protected]
W. Hergert and T. Wriedt (eds.), The Mie Theory, Springer Series in 223
Optical Sciences 169, DOI: 10.1007/978-3-642-28738-1_8,
© Springer-Verlag Berlin Heidelberg 2012
224 D. Mackowski
of scattering by such objects are conducted using the numerical extended boundary
condition method (EBCM), a.k.a., T matrix or discrete dipole approximation (DDA)
procedures. That such is the case has much to do with the relative inaccessibility
of the analytical solutions for ellipsoids as well as the relative ease, accuracy, and
computational efficiency of the numerical procedures.
Mie, in developing his solution, was concerned primarily with the scattering
behavior of a single sphere when exposed to a plane–wave source of illumination.
His solution, however, is recognized today as representing a general relationship
between the incident and scattered field for a sphere. That is, the solution will provide
a description of the scattered field for an arbitrary incident field; the only requirement
being the ability to represent the incident field in a vector wave harmonic (VWH)
expansion. One specific extension of Mie theory in this regard is the description of
scattering of light by spheres when exposed to focussed, Gaussian–profile beams;
the effort behind this theory has not been spent to determine the response of the
sphere to the incident field—as this is already known from Mie’s solution–rather, it
has been directed to approximating the Gaussian beam in a VWH expansion [1, 2].
first derived by Stein [3] and, shortly thereafter, improved by Cruzan [4]; all subse-
quent contributions to the multiple sphere scattering theory can trace their lineage,
either directly or indirectly, back to these two seminal works. Liang and Lo, and
Brunning and Lo [5, 6], were the first to combine the addition theorems with Mie
theory to develop an exact solution for a pair of spheres. Significant advances in,
and applications of, the formulation were subsequently developed by Borghese et al.
[7, 8], Fuller et al. [9, 10], and Mackowski et al. [11–13]. Comprehensive reviews
of the historical development of the method can be found in [14, 15].
At the time of this report the multiple sphere solution can be considered well
established and mature; little remains to be done in the way of mathematical work to
uncover hidden features of the formulation, and most of the current work is aimed
at applying the solution method to predict the scattering and absorption proper-
ties of aggregated and cluster particles. A measure of this progress can be seen
in the recent compendia of references related to the T matrix method [16, 17];
listed are over 200 citations under the category of multiple sphere scattering, and the
vast majority of these works deal with the application of the solution method—in
a diversity of fields ranging from nanotechnology to astrophysics—as opposed to
mathematical/theoretical developments. Indeed, the accessibility and efficiency of
the multiple sphere formulation and codes—combined with the increasing capabili-
ties of modern computational machinery—are now enabling the direct simulation (or,
alternatively, numerical experiments) of radiation propagation in particulate media
via calculations involving on the order of 103 spheres [18, 19].
In keeping with the historical occasion and context of this volume, my report will
revisit the mathematical side of the multiple sphere scattering solution. One objective
of the report is to show that the entire solution—including the details of the VWH
addition theorem—can be derived using the basic properties of the VWH functions.
The basic procedures for applying the solution to obtain the observable properties of
a cluster of spheres, in either a fixed orientation with respect to the incident field or in
a random orientation, will be reviewed. The report will finish with a brief discussion
of future directions and applications of the multiple sphere formulation.
A general configuration for the multiple sphere problem is illustrated in Fig. 8.1. The
ensemble consists of NS nonoverlapping spheres, each characterized by a radius ai ,
refractive index mi , and position ri relative to a common origin, for i = 1, 2, . . . NS .
Incident upon the ensemble is a monochromatic, time harmonic electromagnetic field
of wavelength λ, typically in the form of a plane, linearly polarized wave.
The objective of the problem, in the most general sense, is to obtain a solution to
Maxwell’s time harmonic equations which satisfies the continuity conditions on all
226 D. Mackowski
z
ri rj aj
y
x
aC
surfaces. This problem would likely be intractable if the ensemble was mathemati-
cally represented as a single particle with a single—albeit discontinuous—surface.
On the other hand, the problem becomes relatively simple by adopting a superposi-
tion strategy. In this approach, the field external to the spheres is represented by the
superposition of the incident field and the fields scattered from each sphere in the
ensemble;
NS
Eext = Einc + Esca = Einc + Esca,i (8.1)
i=1
And as was done for the single sphere problem, the incident and scattered fields,
at the ith sphere in the cluster, can be represented by regular and outgoing vector
spherical harmonic (VWH) expansions, centered about the origin of the sphere;
Li
n
2
(1)
Einc = pimnp Nmnp (r − ri ) (8.2)
n=1 m=−n p=1
Li
n
2
(3)
Esca,i = i
amnp Nmnp (r − ri ) (8.3)
n=1 m=−n p=1
In the above, Nmnp denotes the VWH of either type 1 (regular) or 3 (outgoing), of
order n, degree m, and mode p = 1 (TM) or 2 (TE). The VWH functions used in this
report are defined by
1/2
(ν) 1 (ν)
Nmn2 (r) = ∇ × rψmn (r) (8.4)
n(n + 1)
(ν) 1 (ν)
Nmn1 (r) = ∇ × Nmn2 (r) (8.5)
k
where ψ denotes the scalar wave harmonic;
8 The Extension of Mie Theory to Multiple Spheres 227
(ν) jn (kr) Ymn (cos θ, φ) ν = 1
ψmn (r) = (8.6)
hn (kr) Ymn (cos θ, φ) ν = 3
The order truncation limit Li in Eq. (8.3) is chosen to provide an acceptable precision
in the calculated scattering properties of the cluster; Li can typically be set using the
Lorenz/Mie criterion, although in certain cases the neighboring spheres can have a
significant effect on the convergence of Eq. (8.3) [12]. The incident field expansion
coefficients pimnp , for sphere i, will depend on the characteristics of the incident field
i
and the relative position of sphere i, whereas the scattering coefficients amnp for each
sphere are initially unknown and are sought from the analysis.
Lorenz/Mie theory provides a means of relating the scattering coefficients for sphere
i to those representing the exciting field at i, yet this exciting field will consist not
only of the incident field, but also the scattered fields from the other spheres. In this
respect, what is needed to close the problem is a means of casting the outgoing VWH
about some origin j—which describes the scattered field from j—into an expansion
of VWH about origin i. This feat is accomplished with an addition theorem.
In a general sense, an addition theorem transforms an eigenfunction for a given
differential equation, referenced to a given coordinate origin, into an expansion of
eigenfunctions satisfying the same DE about a different coordinate origin. The exis-
tence of such a transformation can be anticipated simply from the fact that the vector
harmonics form a complete basis for representation of fields which satisfy (with the
exception of singular points) the vector wave equation. Consequently, a harmonic
centered about one origin describes a field which must be representable—within a
given radius of convergence—by an expansion of harmonics centered about a dif-
ferent origin.
With regard to an outgoing VWH centered about an origin j, the form and conver-
gence of the addition theorem is dictated by whether the singularity at origin j lies
inside or outside the radial bounds of the translated VWH expansion at some new ori-
gin i. For the latter case, which corresponds to the region defined by |r − ri | < |ri−j |,
where r and ri−j denote the evaluation point of the harmonics and the position vec-
tor between origins i and j (case 1 in Fig. 8.2), the field generated by the outgoing
harmonic will be represented entirely within the region by an expansion of reg-
ular VWH centered about i. Accordingly, the addition theorem for this case will
228 D. Mackowski
r < ri-j
(case 1)
i
ri-j
j
appear as
∞
l
2
(1)
(3)
Nmnp (r − rj ) = Hklq mnp (ri−j ) Nklq (r − ri ), ri−j > |r − ri | (8.8)
l=1 k=−l q=1
n
2
(1)
Eexi,i = i
gmnp Nmnp (r − ri )
n=1 m=−n p=1
with
NS
l
2
i−j j
i
gmnp = pimnp + Hmnp klq aklq
j=1 l=1 k=−l q=1
j=i
The scattering coefficients for sphere i will then be given by the Mie formula,
i
amnp = ainp gmnp
i
(8.9)
in which ainp denote the Mie coefficients of the sphere and depend on the sphere size
parameter xi = k ai and the refractive index mi . Equation (8.9) is recognized as a T
8 The Extension of Mie Theory to Multiple Spheres 229
matrix relationship; made specific for the special case of the sphere in which the T
matrix is diagonal and azimuthal–degree independent.1
By combining the previous two equations, one arrives at the interaction equations
for the sphere ensemble;
NS
l
2
i−j j
i
amnp − ainp Hmnp klq aklq = ainp pimnp (8.10)
j=1 l=1 k=−l q=1
j=i
Equation (8.10), in conjunction with Eqs. (8.1–8.3), represents the formal solution
for the scattered field produced by the sphere ensemble. In the case of equal-
sized spheres with equal truncation limits LS , Eq. (8.10) forms a system of 2NS LS
(LS + 2) linear equations for the set of scattering coefficients. The matrix H i−j will
be fully populated for an arbitrary translation between j and i, and correspondingly
all orders/degrees/modes of the scattered field from a sphere j will (in general) influ-
ence a particular order/degree/mode of the field at i. This is in stark contrast to the
isolated sphere case, in which each scattering order/degree/mode is excited only by
the same harmonic component for the incident field. And it is in this respect that
the multiple sphere solution departs—in a practical sense—from the single sphere
Mie theory: the latter provides an explicit formula for the scattered field, whereas the
former gives only an implicit relationship. That is, numerical methods (in the form
of linear equation solvers) are needed to produce a useable solution.
Important properties of the interaction equations, and the calculation of relevant
quantities (scattering matrix, cross sections) from the solution, will be presented
in a subsequent section. Before getting to this material, it is necessary to develop
formulas and properties for the translation matrix elements.
The same arguments that were used to pose Eq. (8.8) can be applied to the transfor-
mation of a regular VWH, centered about origin j, into an expansion of harmonics
about some new origin i. Since the regular harmonic contains no singularities, its
expansion about a new origin will not be constrained to a specified convergence
radius. Accordingly, the addition theorem for regular harmonics will appear as
1 The anp coefficients defined by Eq. (8.9) are the negative of those formulated in previous works,
e.g., Bohren and Huffman.
230 D. Mackowski
∞
l
2
(1) (1)
Nmnp (r − rj ) = Jklq mnp (ri−j ) Nklq (r − ri ) (8.11)
l=1 k=−l q=1
1
(3)
Nmnp (r) = Pk Hk11 mnp (r) (8.12)
k=−1
1
(1)
Nmnp (r) = Pk Jk11 mnp (r) (8.13)
k=−1
with
1
√
Pk = √ δ|k|−1 (−1)(k+1)/2 x̂ − i ŷ + 2 δk ẑ (8.14)
2 3π
in which δk = (0, 1) for k(=, =)0 is the Kronecker delta function. By using the above
formulas along with the definitions of the vector and scalar harmonics in Eqs. (8.4)
and (8.6), it can be deduced that ψ (3) (rij ) and ψ (1) (rij ) are the basis functions for H
and J, respectively. Since N(3) = N(1) + iN(2) , it follows that H + J + i Y , where
N(2) and Y are the VWH and the translation matrix based on the spherical Neumann
function yn . Consequently, the only difference in the formulas for the H and J must
be the basis functions ψ (3) and ψ (1) .
The remaining form of the addition theorem that needs to be identified concerns
the translation of an outgoing harmonic from j to i for the condition |r − ri | > |rij |,
i.e., when the spherical surface, centered about i, upon which the harmonics are
evaluated encloses the singularity at origin j (case 2 in Fig. 8.2). In order to maintain
the correct far-field behavior (k |r − ri | → ∞), the translated harmonic must appear
as an expansion of outgoing harmonics about the new origin. The translation must
also yield an identity relation when |rij | → 0, which implies that the translation
matrix will be generated from the regular scalar harmonic basis functions ψ (1) . And
since the regular components of the outgoing harmonics must translate according to
Eq. (8.11), the translation matrix must be identical to that appearing in Eq. (8.11).
Therefore, the addition theorem for this case will appear as
8 The Extension of Mie Theory to Multiple Spheres 231
r - r1
r - r3
r - r2
1 r2-1 2 r3-2 3
r3-2 r2-1 r3-1
∞
l
2
(3)
(3)
Nmnp (r − rj ) = Jklq mnp (ri−j ) Nklq (r − ri ), ri−j < |r − ri | (8.15)
l=1 k=−l q=1
Equations (8.8), (8.11), and (8.15) can be used to infer the following translation and
factorization properties of the translation matrices,
∞ n
2
Jmnp klq (ri−j ) = Jmnp m n p (ri−i ) Jm n p klq (ri −j ) (8.16)
n =1 m =−n p =1
∞ n
2
Hmnp klq (ri−j ) = Jmnp m n p (ri−i ) Hm n p klq (ri −j ),
n =1 m =−n p =1
ri−i < ri −j (8.17)
∞ n
2
Hmnp klq (ri−j ) = Hmnp m n p (ri−i ) Jm n p klq (ri −j ),
n =1 m =−n p =1
ri−i > ri −j (8.18)
Care must be exercised when applying Eqs. (8.8) and/or (8.15) recursively via
the above three formulas. In particular, it is not possible to extend, or to bypass, the
convergence radius criteria of the translations via a successive set of translations. One
such fallacious approach is illustrated using Fig. 8.3. In this example, an outgoing
harmonic centered about origin 1 and evaluated at point r is first translated to origin
2. Because |r −r2 | > |r2−1 |, the expansion will be in the form of Eq. (8.15), i.e., with
outgoing VWH centered about origin 2. One might then attempt to apply a second
translation, in which the outgoing VWH, centered about 2, are expanded about origin
3. And since |r − r3 | > |r3−2 |, it would appear that Eq. (8.15) would again apply,
leading to outgoing VWH centered about 3. The net result would be
232 D. Mackowski
(3)
Nμ (r − r1 ) = Jν μ (r2−1 ) Nν(3) (r − r2 )
ν
(3)
= Jν μ (r2−1 ) Jν ν (r3−2 ) Nν (r − r3 )
ν ν
(3)
= Jν μ (r3−1 ) Nν (r − r3 ) −→ ±∞ (8.19)
ν
In the above Greek subscripts are shorthand for order/degree/mode, i.e., ν = (klq).
Although the first equation in the above is valid, the second and third are not because
the implied result would contradict the convergence criterion of Eq. (8.15): the trans-
(3)
lations shown in Fig. 8.3 have |r − r3 | < |r3−1 |, and the translation of Nμ (r − r1 )
can only appear as an expansion of regular harmonics centered about 3.
It was demonstrated in [14] that the translation coefficients can be derived from the
expansion of a plane wave in regular VWH. Alternatively, a more compact derivation
can be obtained by consideration of the far-field behavior of outgoing VWH. This
derivation begins with the translation of an outgoing VWH from origin j to i for
|r − ri | > |ri−j | given in Eq. (8.15). Referring to Fig. 8.4, and recognizing that the
translation coefficients are independent of the location of the evaluation point r, the
distances k|r − ri | and k|r − rj | can be made arbitrarily large so that the VWH attain
their far-field asymptotic forms, which are represented by
1
(3)
Nmnp (r − rj ) → eik |r−rj | Π mnp (cos θj , φj ) (8.20)
k|r−rj |→∞ ik |r − rj |
(3)
and likewise for Nklq (r − ri ). In the above Π represent the vector spherical har-
monic (VSH) functions, which are formally defined by
1/2
1
Π mn1 (cos θ, φ) = (−i)n+1 r ∇ Ymn (cos θ, φ) (8.21)
n(n + 1)
Π mn2 (cos θ, φ) = i r̂ × Π mn1 (cos θ, φ) (8.22)
8 The Extension of Mie Theory to Multiple Spheres 233
The origin separation distance |ri−j | can also be made small with respect to both
|r − ri | and |r − rj | so that θi ≈ θj = θ , φi ≈ φj = φ, and |r − ri | ≈ |r − rj |. By
incorporating these limits into Eq. (8.15), and accounting for the phase shift due to
the displacement of the origin, one obtains
∞
l
2
ei k·ri−j Π mnp (cos θ, φ) = Jklq mnp (ri−j ) Π klq (cos θ, φ) (8.26)
l=1 k=−l q=1
Inversion of Eq. (8.26) can be accomplished by use of the VSH orthogonality rela-
tions;
2π 1
Π mnp (x, φ) · Π ∗klq (x, φ) dx dφ = δn−l δm−k δp−q (8.29)
0 −1
and the formula for the regular VWH translation matrix becomes
2π 1
Jklq mnp (rij ) = eik·ri−j Π mnp (cos θ, φ) · Π ∗klq (cos θ, φ) d cos θ dφ (8.30)
0 −1
Evaluation of the integrals in Eq. (8.30) can be performed by utilizing the linearization
expansion for VSH functions,
(2n + 1)(2l + 1) 1/2
Π mnp · Π ∗klq = il−n
4π
w2
1
× √ w
C−kl mn C−1l 1n Ym−k w (cos θ, φ)
w
(8.31)
w=w ,2
2w + 1
1
234 D. Mackowski
It is implicitly understood that Cklw mn is zero for w < ws = Max(|n−l|, |m+k|). In this
sense, the evaluation of the sum in Eq. (8.33) should begin with w2 = n+l−|p−q| and
step downwards by two until ws is reached. The formula for the outgoing harmonic
translation matrix is obtained simply by replacing ψ (1) with ψ (3) in the above;
Equation (8.26) shows that the VSH functions and the exponential phase factor rep-
resent the eigenvectors and eigenvalues to the regular VWH translation matrix. This
feature, along with the translation properties of the J matrix, can be used to identify
the coefficients for the VWH expansion of a plane, linearly polarized wave. By taking
the magnitude of both sides of Eq. (8.26) and applying Eq. (8.29), it can be shown
that the J matrix is unitary,
l
2
∗
Jklq mnp (rj−i ) Jklq m n p (rj−i ) = δm−m δn−n δp−p (8.35)
l=1 k=−l q=1
8 The Extension of Mie Theory to Multiple Spheres 235
and by using the recursive translation properties of J in combination with the above
formula, it follows that
∗
Jklq mnp (rj−i ) = Jmnp klq (ri−j ) (8.36)
∞
l
2
i k·ri−j
e Π ∗mnp (cos θ, φ) = Jmnp klq (ri−j ) Π ∗klq (cos θ, φ) (8.37)
l=1 k=−l q=1
(1)
This formula is now multiplied by Nmnp (r − ri ), where r denotes an arbitrary point,
and summed over (m, n, p). Application of the translation properties in Eq. (8.11)
results in
∞
n
2
ei k·ri−j Π ∗mnp (cos β, α) Nmnp
(1)
(r − ri )
n=1 m=−n p=1
∞
l
2
(1)
= Π ∗klq (cos β, α) Nklq (r − rj ) (8.38)
l=1 k=−l q=1
The angles β and α, which represent the polar and azimuthal direction of the vector
k, have been introduced to avoid confusion with the coordinates of the vector r.
Equation (8.38) indicates that the VWH expansions centered about origins i and j
must each be proportional to exp(k · (r − ri )) and exp(k · (r − rj )). In this context,
Eq. (8.38) represents a generalized plane wave VWH expansion for a wave that is
propagating in the θ = β, φ = α direction as illustrated in Fig. 8.5. The vector Π
in this expansion defines the direction of the electric field vector in the transverse
plane; the two cases of β̂ · Π and α̂ · Π would represent plane wave polarization
parallel and perpendicular to the ẑ − k plane. A unit amplitude can established by
multiplying the coefficients in the expansion by a constant, which can be obtained
evaluating the right-hand side in the limit of |r − rj | → 0.
In the general case, the electric field vector can be defined by the angle γ relative
to the ẑ − k plane. The expansion coefficients for the field, referenced to the target
origin, will then be given by
with
As illustrated in Fig. 8.5, the angles α, β, and γ , define an incident field frame,
obtained by Euler rotation of the target frame, so that z denotes the propagation
direction.The pmnp, 1 and pmnp, 2 expansion coefficients describe an incident plane
wave that is polarized either in the x or y directions.
When a pair of spheres are widely separated, so that the scattered field from one
attains the far-field form about the other, the translation operation in Eq. (8.8) can
be reduced to a much simpler form. The approach is to replace the Hankel function
appearing in the scalar wave harmonic in Eq. (8.34) with the asymptotic limit
(−i)n ikr
hn (kr) ∼ e , k r n2 (8.42)
ikr
By employing the linearization relation in Eq. (8.31), the outgoing wave translation
matrix will reduce to
4π ikri−j
Hklq mnp (ri−j ) ∼ e Π mnp (cos θi−j , φi−j ) · Π ∗klq (cos θi−j , φi−j ) (8.43)
ikri−j
All the necessary details underlying Eq. (8.10) are now in place, and we can now
turn to the problem of solving Eq. (8.10) for the scattering coefficients and, from
8 The Extension of Mie Theory to Multiple Spheres 237
the solution, determining the observable scattering and absorption properties for
the given cluster of spheres. Linear system of equations such as Eq. (8.10) can be
solved using either iteration or direct (i.e., inversion) methods; the optimum strategy
will depend on the properties of the system (such as the condition number which
will affect the iterative convergence rate), the requirements of the solution (e.g.,
solutions for one versus multiple right-hand sides), and the experience of the user
(perhaps the most important criterion of all). Iteration methods will typically be more
computationally efficient than direct inversion when a solution is needed for a small
number of specified incident directions. Our experience, and that of others, is that
the biconjugate gradient method provides the most reliable, and fastest, solution as
compared to over/under relaxation and order-of-scattering methods [20]. The number
of iterations required for a solution depends on a host of parameters; i.e., the number,
size parameters, and refractive indices of the spheres, and the proximity of the spheres
to each other. In general, as the spheres become more widely separated the solution
will converge faster. An important factor affecting convergence is whether any of the
spheres is at or near a resonance mode; such conditions can lead to extremely small
convergence rates and may be more effectively solved using direct methods [21].
In situations where the scattering properties are needed for a relatively large
number of incident directions, or where the random-orientation properties of the
cluster are sought, it becomes most efficient to calculate a T matrix for the cluster.
The T matrix will provide the linear relationship between the VWH coefficients
representing the incident and the scattered field, and the matrix can be obtained via
direct inversion of Eq. (8.10). It turns out, however, that the multiple-spheres problem
allows for a computationally efficient, iteration-based method for calculating the
cluster T matrix.
There are three levels of T matrix relationships for a cluster of spheres.The first,
referred to as the sphere–sphere T matrix, represents the formal solution to Eq. (8.10)
and appears as
NS
i
aμ = i−j j
Tμν pν (8.44)
j=1 ν
1 i−j
NS
i−j j −j
Tμ ν − Hμ ν Tν ν = δi−j δμ−ν (8.45)
aμ
j =0 ν
j =i
238 D. Mackowski
In the above and what follows, Greek subscripts are a shorthand for order/degree
/mode, i.e., μ = (m, n, p). The T j−i matrix describes the scattered field produced at
sphere i due to an incident field at j; all the effects of multiple scattering are embedded
within the matrix.
The next T matrix relationship describes the scattered field at sphere i due to the
entire effect of the incident field on the ensemble.This description begins with an
expansion of the incident field about the coordinate origin of the target;
L
n
2
(1)
Einc = pmnp Nmnp (r) (8.46)
n=1 m=−n p=1
The order truncation limit L in this expansion—which is chosen so that the expansion
will yield an acceptable description of the incident field on each sphere in the ensem-
ble—will typically depend on the size parameter kaC based on the circumscribing
radius aC (illustrated in Fig. 8.1).As was the case with the individual sphere limits
Li , the target L can typically be estimated using the Mie criterion applied to k aC .
The sphere centered incident field expansion coefficients can then be obtained from
the target coefficients via the translation operation,
L
l
2
i−0
pimnp = Jmnp klq pklq (8.47)
l=1 k=−l q=1
where origin 0 denotes the target origin. Replacing this into Eq. (8.44) results in
NS
i−j j−0
i
aμ = Tμν Jν ν pν
i=1 ν ν
= i
Tμν pν (8.48)
ν
In general, the sphere-cluster matrix Tμν i will have more columns than rows; the
1 i NS
i−j j
Tμν − Hμ ν Tν ν = Jμν
i−0
(8.49)
aiμ j=1 ν
j=i
This system can be solved by iteration for successive values of ν = (klq) to build
up the T i matrices for the spheres. Such a method is often more computation-
8 The Extension of Mie Theory to Multiple Spheres 239
L
n
2
(3)
Esca = amnp Nmnp (r) (8.50)
n=1 m=−n p=1
NS
aμ = 0−i i
Jμν aν (8.51)
i=1 ν
NS
0−i i
aμ = Jμν Tν ν pν
i=1 ν ν
= Tμν pν (8.52)
ν
Calculation of the full polarimetic scattering properties of the ensemble for a fixed
incident direction requires the solution of Eq. (8.10) for two mutually perpendicu-
lar linear polarization states of the incident wave. Referring again to Fig. 8.5, the
polarization state of the incident beam is specified by the angle γ . Since the incident
field coefficients defined in Eq. (8.39) are a linear combination of cos γ and sin γ
components, a solution to Eq. (8.10) for an arbitrary γ can be obtained via
240 D. Mackowski
i
amnp, = amnp,1
i
cos γ + amnp,2
i
sin γ (8.53)
i
where amnp,1 i
and amnp,2 denote the two solutions to Eq. (8.10) for incident field
coefficients of pmnp, 1 and pmnp, 2 given by Eqs. (8.40) and (8.41), i.e., γ = 0 and π/2.
The parallel () notation will be described shortly. The scattering plane used to define
the amplitude matrix will now be adopted as the x −z plane, and following the usual
convention, the parallel and perpendicular incident polarization states will correspond
to the electric field pointing in the x and the −y directions, respectively. The solution
corresponding to the parallel case will therefore be that given in Eq. (8.53), and the
perpendicular solution will be that obtained by replacing γ with γ − π/2;
i
amnp,⊥ = amnp,1
i
sin γ − amnp,2
i
cos γ (8.54)
The amplitude and scattering matrix of the ensemble can be calculated using
either the superposition representation in Eq. (8.1), which would retain the scattering
coefficients for the individual spheres, or the single-origin expansion in Eq. (8.50).
In the far-field limit, the former method is made complicated by the need to explic-
itly account for the relative phase shift among the different scattering origins as a
function of the scattering angle, yet this method can also be more computationally
efficient in situations in which the spheres are relatively far apart (i.e., large k aC ),
for which a large value of L would be required for convergence of Eq. (8.50) [22].
However, the latter method retains a spherical harmonic representation of the scat-
tered field, and as such allows for a more direct identification of various moments
of the scattered intensity (such as the scattering efficiency and the asymmetry para-
meter). Our experience is that advantages of the single-expansion method outweigh
the disadvantages; it should be noted as well that convergence or loss-of-precision
problems associated with the numerical implementation of Eq. (8.51) have not been
encountered in our codes.
Equations (8.53) and (8.54) will hold equally for the total scattered field coeffi-
cients amnp,1 and amnp,2 that are obtained from Eq. (8.51). The expansion in Eq. (8.50)
that is obtained from these coefficients would describe the scattered field in a target-
based reference frame; to conform to the conventional, incident-based frame it is
necessary to perform a rotation transformation on Eq. (8.50). The incident frame is
obtained by a Euler rotation of the target frame through (α, β γ ), and the correspond-
ing rotation transformation on the harmonic functions is given by
n
(n) γ (3)
(3)
Nmnp (kr, θ, φ) = ei m α Dm m (cos β) ei m Nm np (kr, θ , φ ) (8.55)
m =−n
where θ , φ denote the polar and azimuthal angles in the incident frame; θ = 0
(corresponding to θ = β) now represents the forward scattering direction. The
(n)
generalized spherical functions Dmk are related to the Jacobi polynomials; defini-
tions and properties of the functions are presented in the Appendix. Application of
Eq. (8.55) to (8.50) will transform the expansion so that the VWH become based on
8 The Extension of Mie Theory to Multiple Spheres 241
the incident frame (i.e., r replaced with r ); the total scattering coefficients appearing
in the rotated expansion will be given by
n
(n) α
amnp,s = ei m γ Dmm (cos β) ei m am np,s (8.56)
m =−n
where Einc, and Einc,⊥ denote the magnitudes of the incident electric field ampli-
tude parallel and perpendicular to the scattering plane. The explicit formulas for the
amplitude matrix elements are obtained by applying to the rotated Eq. (8.50) the
far-field asymptotic form of the outgoing VWH in Eq. (8.20), and noting that the
scattering plane corresponds to φ = 0.This gives
L
n
2
S1 = (−i)n amnp,⊥ τmn3−p (cos θ ) (8.58)
n=1 m=−n p=1
L
n
2
S2 = (−i)n+1 amnp, τmnp (cos θ ) (8.59)
n=1 m=−n p=1
L
n
2
S3 = (−i)n+1 amnp,⊥ τmnp (cos θ ) (8.60)
n=1 m=−n p=1
L
n
2
S4 = (−i)n amnp, τmn3−p (cos θ ) (8.61)
n=1 m=−n p=1
Elements of the scattering matrix can be obtained directly from those of the amplitude
matrix following the formulas presented in Bohren and Huffman [23].
The absorption cross section of sphere i is defined so that Cabs,i I0 is the rate at which
the sphere absorbs energy from the incident wave of irradiance I0 . This quantity
can be obtained by integration of the Poynting vector over the inside surface of the
particle. By employing the formulas of Lorenz/Mie theory to relate the internal and
external fields, the expression for the absorption cross section is
242 D. Mackowski
2π
Li n
2
i i 2
Cabs, i = bnp amnp (8.62)
k2 m=−n
n=1 p=1
i
in which bnp is a positive (or zero) real valued property solely of sphere i and is
defined by
i 1
bnp = −Re +1 (8.63)
ainp
The scattering coefficients in the above formula would correspond to either the paral-
lel or perpendicular polarization values; the absorption cross section for unpolarized
incident radiation would be the average of the two. By simple conservation-of-energy
considerations, the absorption cross section of the entire ensemble is the sum of the
individual sphere cross sections;
NS
Cabs = Cabs,i (8.64)
i=1
As before, the scattering and incident field coefficients would correspond to the
particular polarization state. The total ensemble extinction cross section would also
be obtained from the sum of the parts;
NS
Cext = Cext,i (8.66)
i=1
Unlike the absorption cross section, the extinction cross section for the individual
sphere would be difficult—if not impossible—to experimentally measure. The def-
inition of this quantity relies on the superposition model of the scattered field, and
although this model serves perfectly well as a means to solve Maxwell’s equations for
the ensemble, it is not obvious how the ‘partial’ fields scattered from the individual
spheres could be discriminated in an experiment.
The individual sphere extinction cross section is also not bounded by the usual
behavior for an isolated particle; although the total cross sections must satisfy Cext ≥
Cabs , it is entirely possible for Cext,i < Cabs,i and even for Cext,i to be negative. The
precise balance of radiant energy, at the individual sphere level, can be identified
8 The Extension of Mie Theory to Multiple Spheres 243
by multiplying Eq. (8.10) through by 2π/k2 amnp i∗ /ai , summing over m, n, p, and
np
rearranging to isolate Eqs. (8.62) and (8.65). This results in
2π i 2 2π
NS
Cext,i − Cabs,i = 2 aμ + 2 Re i∗
aμ Hμi−jν aνj
k μ k μ ν j=1
j=i
= Ci−sca,i + Cd−sca,i (8.67)
The quantities appearing on the right-hand side are referred to as the independent
and the dependent scattering cross sections of the sphere.The former will always
be positive, yet the latter—which accounts for energy transfer via the interference
between the fields scattered from i and the neighboring spheres—can take on positive
or negative values. And since Cabs,i + Ci−sca,i > 0, the above balance implies that
Cext,i > Cd−sca,i .
By using H i−j = J i−j + iY i−j in Eq. (8.67), and applying Eq. (8.104) to both J
and Y , it follows that the total dependent scattering cross section is given by
NS
Cd−sca = Cd−sca,i
i=1
2π
NS
NS
= Re a i∗ i−j j
H
μ μν ν a + a j∗ j−i i
H
μ μν ν a
k2 μ,ν
i=1 j=i+1
2π
NS NS
= 2 Re aμ i∗
Jμi−jν + i Yμi−jν aνj
k
i=1 j=i+1 μ,ν
∗
+ aμ Jμ ν − i Yμ ν aνj
i i−j i−j
2π
NS
NS
= i∗ i−j j
aμ Jμ ν aν (8.68)
k2 μ,ν
i=1 j=1
j=i
Note that it is no longer necessary to indicate the real part in the last line, because
the quantity will be identically real. The regular–wave translation matrix can now be
reexpanded about the target origin, via
i−j
L
n
2
0−j
i−0
Jmnp klq = Jmnp m n p Jm n p klq
n =1 m =−n p =1
L n
2
∗
0−j
= Jm0−i
n p mnp Jm n p klq (8.69)
n =1 m =−n p =1
244 D. Mackowski
Substituting the above into the previous equation and using Eq. (8.51) gives the
expected result for energy conservation,
NS
Cext − Cabs = Ci−sca,i + Cd−sca,i = Csca (8.70)
i=1
2π 2
Csca = aμ (8.71)
k2 μ
The random orientation cross sections can be obtained by using the matrix rela-
tionships for the scattered and incident field (Eqs. 8.44, 8.47, 8.48) in Eqs. (8.62)
and (8.65) and integrating the incident field over all propagation and polarization
directions. For a plane wave, this integration results in
∗
piν piν = δν−ν (8.72)
2π i i 2
Cabs,i = 2 b T
k μ ν μ μν
2π i
NS
NS ∗
i−j j−j i−j
= b T J T (8.73)
k2 μ ν μ μν
μν ν ν
j=1 j =1 ν
2π
Cext,i = − 2 Re Jμ0−i i
ν Tν μ
k μ ν
2π NS
= − 2 Re Tμi−jν Jνj−i
μ (8.74)
k μ ν j=1
A careful distinction must be made when interpreting the sphere absorption cross
section in the context of Kirchoff’s law. Consider the situation in which each sphere
in the ensemble is at a uniform temperature of Ti and the ensemble is surrounded
by a black environment at Te . As per the definition of the absorption cross section,
and considering the isotropic environmental radiation, the rate at which sphere i
absorbs radiant energy from the environment about a wavelength interval dλ will
be qe−i = 4π Cabs,i Ib,λ (Te ) dλ, in which Ib,λ denotes the blackbody
intensity
function. Consequently, Kirchoff’s law states that qi−e = 4π Cabs,i Ib,λ (Ti ) dλ
will be the
rate at which emission from sphere i is transferred to the environment.
That is, Cabs,i does not directly measure the rate of energy emission from sphere i;
only how much of this rate escapes the confines of the ensemble.On the other hand,
the detailed transfer of radiant
energy among the particles should be embedded
within the formula for Cabs,i . This structure is not obvious in the definition given
in Eq. (8.73), and an alternative definition is needed to elucidate the interactive form
of the absorption cross section.
The first step in this process is to combine Eqs. (8.97) and (8.101), so that
i−j j−i
Hmnp klq = (−1)m+k H−klq −mnp (8.76)
It can then be shown from Eq. (8.45) that T i−j will also have this property, and that
246 D. Mackowski
1
NS
i−j j −j
j
Tμi−jν − Tμ ν Hν ν = δij δμν (8.77)
aν j =1 ν
j =i
This equation is now multiplied by (Tμν )∗ and summed over j and ν. By employing
i−j
the contraction properties of the translation matrices used in Eqs. (8.68) and (8.69),
it follows that
NS 2
j i 2
bν Tμi−jν + Tμ ν = −Re Tμi−iμ (8.78)
j=1 ν ν
2π i
2 NS 2
i j
− bν Tμi−iν − bν Tμi−jν (8.79)
ν j=1 ν
j=i
Each of the terms appearing in this formula can be interpreted according to a radiant
exchange relationship among the spheres. When multiplied by 4π Ib,λ (Ti ) dλ, the first
and second terms on the right represent the source of emissive power from sphere i
and the rate of reabsorption by sphere i; the net emissive power being the sum of the
two. The third term accounts for the absorption of the net emissive power from i by
all the other spheres in the ensemble. The net emissive power from a sphere, and the
emissive exchange between a pair of spheres, can be used to define emission cross
sections via
2π i i 2
i−i
Cemis,i = − 2 b Re Tμ μ +
i−i
bν Tμ ν (8.80)
k μ μ ν
2π i j i−j 2
Cemis,i−j = 2 b b T (8.81)
k μ μ ν ν μν
8 The Extension of Mie Theory to Multiple Spheres 247
The exchange cross section is defined so that net rate of monochromatic emissive
transfer between a pair of spheres is given by
qi−j = 4π Cemis,i−j Ib,λ (Ti ) − Cemis,j−i Ib,λ (Tj ) dλ (8.82)
As per the symmetry property of the T i−j matrix, the exchange cross section will
automatically satisfy the required reciprocity condition of
it follows that
2
2π
Cemis,1 N = 2 −Re aμ − a1μ
1
S =1 k μ
2
2π
L1
2
= 2 (2n + 1) −Re anp − a1np
1
k
n=1 p=1
= Cext,1 − Csca,1 = Cabs,1 (8.85)
In other words, the emission cross section for an isolated sphere becomes identical
to the absorption cross section for the sphere.
Unlike the case for the absorption cross sections of the individual spheres, the
magnitude of Cemis,i and Cemis,i−j , as well as the number of orders required for
convergence of Eqs. (8.80) and (8.81), is strongly dependent on the separation dis-
tance between the spheres [24, 25].Indeed, as the gap between a pair of spheres
shrinks to zero, Eqs. (8.80) and (8.81) individually will fail to converge, although
their difference—which is the sphere absorption cross section per Eq. (8.79)—will
converge. This behavior is a consequence of the isothermal-sphere assumption that
formed the basis of the formulas; for two spheres each at uniform yet distinct temper-
atures, the radiative heat transfer between the spheres would become unbounded as
the spheres approached contact. In this limit, however, the spheres could not maintain
isothermal conditions as conduction heat transfer, within each sphere, would act to
maintain temperature continuity at the point of contact.
248 D. Mackowski
The focus of the presentation so far has been on the extension of Mie theory to
multiple sphere systems. In this sense, we have not examined a fundamentally new
solution to a boundary value problem: the only real boundaries in the problem are at
the sphere surfaces, and the continuity conditions at these points are provided by Mie
theory. The general formulation provided by the multiple sphere solution, however,
can be extended into topics well beyond that of light scattering by sphere clusters.
Figure 8.6 illustrates the two fundamental directions that such extensions can
take. One path concerns the description of EM wave propagation in discretely inho-
mogeneous media; i.e., a medium consisting of particles dispersed in an otherwise
continuous matrix. This situation can be viewed as the application of the multiple
sphere solution to a semi–infinite system of spheres, and an objective of this applica-
tion is to identify an effective propagation constant of the medium which describes
the attenuation of the average (or coherent) field via absorption and scattering by
the particles. In the diagram Fig. 8.6, this extension takes the situation on the left
and models it as that on the right, i.e., representing an inhomogeneous medium as
effectively homogeneous.
A robust theory exists for describing the propagation of EM waves in particulate
media, and the reader is referred to the seminal works of Ishimaru [26], Varadan et al.
[27], and Waterman and Pedersen [28] for the derivations and details. Most typically,
the theoretical models address the reflection and propagation of a plane wave by a
half space of particles that are either randomly distributed with a known concentra-
tion and pair distribution function, or arranged onto a lattice. Development of the
models begins with the interaction equations for the sphere scattering coefficients,
Eq. (8.10), applied to the semi–infinite expanse of particles. Closure of the equa-
tions is made possible by simplifying assumptions regarding the high–order correla-
tions among the particle positions and scattered fields—such as the quasi–crystalline
approximation—from which the configuration–averaged scattering coefficients can
be represented by a functional, plane-wave model [29–31].
A simplified presentation of the key elements in an effective medium theory will
be given in the following section. Before doing so, it will be noted that the second
basic extension of the multiple sphere solution can be viewed as the reverse of the
effective medium problem. The task here is to represent a homogeneous—albeit
8 The Extension of Mie Theory to Multiple Spheres 249
Tμi−jν ≈ αμ Hμi−j
ν αν , Hμi−j
ν = Hμ ν (ke ri−j ) (8.87)
in which
8 The Extension of Mie Theory to Multiple Spheres 251
3f rc2
HαJ =
4π a3 (ke2 − k2 )
∂ H(k(−r)) ∂ J (ke r)
× α J (ke r) − H(k(−r)) α d
4π ∂r ∂r r=rc
(8.90)
Of course, many additional details would be needed to develop the basic con-
cepts, presented above, into a working computational scheme.Analytical formulas
for the effective propagation constant can be derived for spheres in the dipole limit
[28, 31, 34]. However, calculation of ke for sphere size parameters of order unity
252 D. Mackowski
8.6 Appendix
Equations (8.33) and (8.34) provide a compact and relatively simple means of calcu-
lating the translation matrices. More computationally-efficient methods have been
devised, such as recurrence relations for the matrix elements, yet in the overall
scheme of things, the execution time required to evaluate the translation matrices
will typically be small compared to that required to solve the interaction equations.
8 The Extension of Mie Theory to Multiple Spheres 253
Calculation of the Bessel, Hankel, and Legendre functions that appear in the scalar
harmonic ψ should pose no problem to those familiar with Lorenz/Mie theory. The
Clebsch–Gordon coefficients, on the other hand, are not as common. These quantities
appear most often in problems relating to angular momentum in quantum mechanics
and the representation of the rotation group, and they are part of the intrinsic function
set in Mathematica.In numerical codes they are most efficiently calculated using
recurrence relations. The downwards recurrence is stable up to the maximum L
values encountered in our codes, and is given by
w−2 w−1
Cmn,kl = a b Cmn,kl − c Cmn,kl
w
(8.93)
with
1/2
(2w − 3)
a = 2(w − 1)
(w − m − k − 1)(w + m + k − 1)(l − n + w − 1)
1/2
(2w − 1)
×
(n − l + w − 1)(n + l − w + 2)(n + l + w)
(m − k)w(w − 1) − (m + k)(n(n + 1) − l(l + 1))
b=
2w(w − 1)
1 (w + m + k)(w − m − k)(n − l + w) 1/2
c=
2w (w + w + 1)
(l − n + w)(n + l − w + 1)(n + l + w + 1) 1/2
× (8.94)
(w + w − 1)
The index w in Eq. (8.93) begins with w = n+l, and the Clebsch–Gordon coefficients
for indices w < ws = Max(|n − l|, |m + k|) are implied to be zero. Starting values
for the C coefficients are obtained from
⎛ ⎞1/2
n+l+m+k n+l−m−k
⎜ l+k l−k ⎟
n+l
Cmn,kl =⎜
⎝ ⎟
⎠ (8.95)
2(n + l)
2l
with
n n!
= (8.96)
l l!(n − l)!
In the three previous equations, it is understood that the conjugate operation does
not apply to the Bessel function jw (kr). This has no bearing for translation in a non-
absorbing medium, for which k and jw (kr) will be real. The three previous formulas
can also be applied to the H translation matrix, yet again the conjugate operator does
not apply to the Hankel function hw (kr).
A time–intensive step during the iterative solution to Eq. (8.10) is the matrix–vector
product,
L l 2
i−j j
j,i
amnp = Hmnp klq aklq (8.105)
l=1 k=−l q=1
n
(n) j
j,1
amnp = Dmk (cos θi−j ) exp(ikφi−j ) aknp (8.106)
k=−n
2. Axial translation to i,
8 The Extension of Mie Theory to Multiple Spheres 255
L
2
j,1
j,2
amnp = Hmnp mlq (krij , θ = 0, φ = 0) amlq (8.107)
l=1 q=1
3. Inverse rotation,
n
(n) j,2
j,i
amnp = exp(−imφi−j ) D−m−k (cos θi−j ) aknp (8.108)
k=−n
(n)
The generalized spherical functions Dmk (x) are defined by
(n) (n − k)!(n + k)! 1/2
Dmk (x) = (−1)m+k
(n − m)!(n + m)!
1 + x (m+k)/2 1 − x (k−m)/2 k−m, k+m
× Pn−k (x) (8.109)
2 2
α,β
in which Pn is the Jacobi polynomial.
(n)
A stable recurrence formula for calculation of the Dmk functions starts with
1/2
(n) (n − m)!
D0k (x) = Pnk (x) (8.110)
(n + m)!
and, for m = 1, 2, . . . n,
(n) 1 1 + x (n)
Dmk (x) = [(n + k)(n − k + 1)]1/2 Dm−1k−1 (x)
[(n − m + 1)(n + m)]1/2 2
1 − x (n)
−[(n − k)(n + k + 1)]1/2 Dm−1k+1 (x)
2
(n)
−k (1 − x 2 )1/2 Dm−1k (x)
(8.111)
and
(n) (n)
D−m−k (x) = (−1)m+k Dmk (x) (8.112)
256 D. Mackowski
References
A D
Absorbing host medium, 113, 114, 116, Debye scalar potentials, 62
118, 119 Debye series, 59, 195
Absorption coefficient, 105, 112, 115 Debye, Peter, 55
Arnold, Engelbert, 11 Diffusion term, 108, 111
Asymmetry parameter, 114, 119 Direct transmission
ensemble, 114, 128 term, 108, 110, 111
Dorn, Ernst, 17
B
Bateman Harry, 55 E
Beam transmittance, 55, 109 Environment map, 104
Born, Max, 55 Extinction
Brewster angle, 199 bulk coefficient, 114
Bulk coefficient, 105, 110, 112, 114
absorption coefficient, 115 cross section, 118, 119
extinction coefficient, 114
refractive index, 115
scattering coefficient, 114 F
Field of view, 103
Film, 103
C Focused laser beam, 60
Camera, 101–103, 111, 128 Forward direction, 112
Camera constant, 103 Fourier series, 59
Clebsch, Alfred, 55 Fourier Lorenz–Mie theory, 59
Cluster Freiburg
of spheres, 62 Naturforschende Gesellschaft, 22
arbitrary shaped particles, 62 Freiburger Konzil, 43
of rotationally symmetric particles, 62
Coefficient of variation, 120
Colloidal gold, 53 G
Computer graphics, 101–103, 105, Generalized multi-sphere
111, 112 Mie-solution, 62
Corona, 202 Glory, 206
W. Hergert and T. Wriedt (eds.), The Mie Theory, Springer Series in 257
Optical Sciences 169, DOI: 10.1007/978-3-642-28738-1,
Ó Springer-Verlag Berlin Heidelberg 2012
258 Index
M P
Matlab, 58 Particle, 54
Mie coefficients, 56 nonspherical, 158
Mie potential, 25 Path tracing, 107, 111, 122
Mie theory bidirectional, 122
algorithms, 56 Path transmittance. See beam transmittance,
history, 54 108
programs, 56 Phase function, 105, 110––114, 118,
review papers, 53 121, 128
Mie translation project, 55 ensemble, 114
Mie, Berta, 14 Henyey–Greenstein, 114
Mie, Gustav, 53, 55 single particle, 113
Index 259
S
Sandia laboratories, 55 W
Scattering, 104, 105, 110–114, 119, 128, 131 Wolf, Emil, 55
albedo, 110
angle, 129