2766 J. Phys. Chem.

1994,98, 2766-2781

ARTICLES

Ultrafast Electron Diffraction. 4. Molecular Structures and Coherent Dynamics

J. Charles Williamsont and Ahmed H. Zewail'
Arthur Amos Noyes Laboratory of Chemical Physics,$ Calijornia Institute of Technology,
Pasadena, California 91 125
Received: January 13, 1994'

Ultrafast electron diffraction (UED) is developed, in this and the accompanying paper, as a method for studying
gas-phase molecular structure and dynamics on the picosecond (ps) to femtosecond (fs) time scale. Building
on our earlier reports (henceforth referred to as 1-3), we discuss theoretical and experimental considerations
for the approach. Specifically we show that the use of rotational and vibrational coherences can add a new
dimension to structural determination of gas-phase species. In addition to the internuclear separations of the
molecular sample, the spatial alignment reflected in the scattering pattern contains bond angles and rotational
constants for both excited-state and ground-state species. Vibrational coherence effects are also observable,
and the motion of the wave packet is revealed by the change of the diffraction pattern with time, thus yielding
the molecular dynamics. U E D provides the temporal evolution of the reaction coordinate directly and is well-
suited for studies of global structure changes on this time scale. Paper 5 details our experimental studies with
UED and the current time resolution of the apparatus.

I. Introduction many GED laboratories around the world. The traditional

experiment begins with a continuous, monoenergetic electron
The interaction between a monoenergetic X-ray beam and the beam (10-100 keV), typically produced froma hot cathodesource.
electronic distribution within a crystalline sample leads to the The electrons are collimated and directed toward a gaseous stream,
formation of a well-defined X-ray scattering pattern. The where they scatter and are detected with a photographic plate.
scattering pattern is highly sensitive to crystal structure; the
Analysis of the diffraction patterns, however, requires care,
inherent microscopic order of nuclei within the crystal, in terms
particularly when the molecule has many atoms. The advent of
of both spatial orientation and position, is projected into the computer fitting techniques has greatly accelerated the analysis
macroscopic environment by the scattering process. Peter Debye process.4
proposed that structural information about gaseous samples would
also be present in an X-ray diffraction pattern despite the lack The structures of transient intermediates, such as radicals,
of orientational and spatial order between mo1ecules.l The was originally studied on the millisecond to microsecond time
structural regularity of individual molecules is still present in a scale by flash photolysis combined with spectroscopicprobing.%'
gaseous sample, and manifests itself in the form of concentric In the past decade, several groups have invoked electron diffraction
rings (as opposed to the spotlike patterns from crystals). The as a method for examining these structures. Ischankoet al. created
ring pattern is related to a probability distribution of internuclear 1-ps electron pulses using an electromagneticchopper and studied
distances and reflects the relative locationsof each atom, although the structure of radical products formed in IR multiphoton
the bond angles and the exact three-dimensional structure are dissociation of CFJ8 (for a recent review see ref 9). Rood and
not immediately evident. Milledge conducted diffraction studies on the decompoeition of
ClOz using 100-l.tselectron pulses,1oand recently Bartell and his
Gas-phase X-ray diffraction was first demonstrated in 1929,
group studied the phase changes of clusters on the microsecond
when Debye and co-workers imaged the scattering pattern of
time scale as welLL1 Ewbanb et ai. improved the temporal
carbon tetrachloride vapor.2 The experiment was more chal-
resolution of electron diffraction to nanoseconds by combining
lenging than crystallographic studies because of the low density
a laser-initiatedelectronsourcewitha linear diodeanaydetector,12
of the gaseous sample, exposure times of several hours, and the
and their studies on the laser-decompositionof CS2 are supported
vacuum requirements. Mark and Wierl discussed the feasibility
by theoretical calculations. Mourou and Williamson demon-
of using electrons to study gas-phase molecular structure instead strated that the diffraction pattern from an aluminum film could
of X-rays. Electrons interact directly with the atomic nuclei
be recorded with a single 100-ps electron pulse and that a phase
such that electron scattering is several orders of magnitude
transformation of the film could be observed in real time.') With
stronger than X-ray scattering from molecules. Mark and Wierl
this resolution, Elsayed-Ali et ai. have successfully shown
were successful;in 1930 they produced a diffraction pattern from
reflection high-energy electron diffraction (RHEED) of laser-
carbon tetrachloride that was more distinct than similar X-ray
heated surfaced4 (Kitriotis and Aoyagi recently reported nano-
scattering exposures and required a fraction of the exposure time.3
second RHEED results's). Solution studies have also been
The utility of gas-phase electron diffraction (GED) was considered; Bergsma et ai. theoretically predicted the ultrafast
recognized by several investigators,including Linus Pauling here X-ray scattering patterns of solvated iodine; see section IV,part
at Caltech who was interested in using structures to examine the B.16
fundamental nature of the chemical bond. Currently there are
Structural changes of isolated molecular systems and reactions
tNational Science Foundation Pre-Doctoral Fellow. on the femtosecond-picosecond time scale have been probed with
*Contribution No. 8916. femtosecond transition-state spectroscopy (FTS), where the
.Abstract published in Aduance ACS Abstracts, February IS, 1994. experimental time resolution allows observation of the coherent
0022-3654/94/2098-2766so4.5o/o 0 1994 American Chemical Societv
Ultrafast Electron Diffraction The Journal of Physical Chemistry, Vol. 98, No. I I, 1994 2767
nuclear motion.17 With this in mind, we proposed some time ago II. Time-Independent CED Theory
that GED could be used as a variant of the FTS approach to
The general theory of gas-phase electron diffraction has been
follow changes in internuclear separations with time for systems
described elsewhere (see, for example, refs 4, 22, and 23). This
in a molecular The premise is similar to FTZ: a laser
section presents a theoretical framework with which to consider
with femtosecond temporal resolution initiates a chemical reaction
within a molecular beam sample, and a second pulse, in this case the effects of rotational and vibrational coherence. The incident
electron beam is defined by wave vector b,where the magnitude
an ultrashort electron burst, probes the structural change.18
of is related to the de Broglie electron wavelength by lbl =
Two years ago we reported on our first efforts in this direction
2a/X. The incident electron beam is perturbed by the electronic
with an electron time resolution of -1 ps.19 These electrons
potentials of the molecular sample and subsequently scatters with
(generated by the photoelectric effect) were accelerated to high wave vector k at an angle 0. If the scattering process is elastic
energy (15-17 keV) and directed toward the interaction region,
lkl), then the momentum transfer vector s may be written
where the femtosecond pump laser, electron beam, and molecular
beam intersect. The electrons were scattered by molecules in the sZ16-k (1)
interaction region, and the resulting diffraction pattern was
detected with a two-dimensional charge-coupled device (CCD) The magnitude of the momentum transfer is
operating in direct electron-'-?mbardment mode. The success of
these experiments relied on minimizing the space-charge effects 4a
Is1 = 21161 sin(6/2) = sin(8/2) (2)
to maintain picosecond time resolution, and therefore the beam
current was kept very low-up to 8 orders of magnitude below
conventional GED beam currents. Tocompensate, single-electron The total scattering intensity, I, is the sum of contributions
detection was invoked. from atomic (I*) and molecular scattering (fM). These intensities
In conventional FTS experiments the time delay is varied by vary with scattering angle and are usually expressed as a function
delaying the probe pulse relative to the initial t = 0 pulse in a of s rather than B
Michelson interferometer arrangement. A similar design is used
in UED, but the situation is more complicated because the electron
and photon beams have different velocities. This velocity
mismatch was recently shown to be important to the overall time If it is assumed that the electronic potentials of each atom in the
resolution and must becarefullyaccounted for in theexperimental molecule are independent (the independent atom model), then
design.20 Considerations of temporal and spatial resolutions in the atomic scattering intensity may be written as a sum of elastic
our apparatus are further discussed in the accompanying paper.2' and inelastic scattering contributions
An important feature of ultrafast electron diffraction is the
formation of a wave packet in the molecular sample, as occurs
with FTS. Both the rotational and vibrational motions of the
molecules are coherent; a degree of order is imposed on the
previously isotropic sample. It is therefore necessary to consider In this equation, N i s the number of atoms in the molecule,fi and
thestructural changes, both in this regimeof coherent elementary SIare the direct elastic and inelastic scattering amplitudes for
dynamics and when coherence is subsequently lost. Conversely, atom i, a0 is the Bohr radius, and Cis a proportionality constant.
it is possible to exploit the coherent motion to observe new The contributions from spin-flip scattering amplitudes (gi)have
structural features, as discussed below: not been included as they are generally neglected for high-energy
In this article, we turn our attention to the impact of rotational electron diffraction experiments24
and vibrational coherences on the scattering pattern in an ultra- The molecular scattering intensity is composed of interference
fast electron diffraction experiment. We wish to establish the terms which contain structural information about the scattering
nature of the structural and dynamical information that can be center. The intensity is written as a sum over all internuclear
obtained from an experiment which combines high temporal separations:
resolution with high structural resolution. It is shown that an N N
additional dimension of imaging can be achieved at times when
coherence is induced or recovered-in a sense, Debye's ring pattern
begins to approach the diffraction from a crystalline sample. We 7#I
will specificallydiscuss the diffraction patterns which result from
the temporal change of anisotropic distributions and show that where qris the complex phase term for the corresponding scattering
more structural information, including bond angles and rotational amplitude fr, and rij is the position vector between atom i and
constants, is available than in a conventional experiment. To
atomj. The interaction between incident electron and scattering
predict the impact of vibrational coherence on structural changes, molecule is extremely f a s t - o n the order of attoseconds-and
we present studies of the time-dependent scattering patterns for
the molecular scattering intensity must be averaged over all
both dissociative and bound-state systems with the help of populated vibrational and rotational states in the molecule.
molecular dynamics simulations. We illustrate our points by Vibrational averaging contributes an additional exponential term
first considering the structural changes in diatomics and then (the Debye-Waller factor) that is dependent upon I , the mean
examining more complex molecular systems of, for example, five amplitude of vibration
or eight atoms.
The paper begins by highligh ting relevant aspects of gas-phase
electron diffraction theory. This brief outline provides a basis
for much of the theoretical analysis presented later, when we
consider the effect of coherence on the diffraction pattern.
Rotational coherence is discussed in section 111, and we show
"snapshots" of the diffraction patterns that result from different
combinations of polarization and excitation. We also examine This factor represents the replacement of a specific internuclear
the impact of rotational coherence on the analysis of dissodiation separation with a weighted distribution corresponding to the
products and molecules with more structural complexity than a vibrational motion between the two atoms.
diatomic. The effects of vibrational coherence are considered in If the molecules have not been prepared in a specific rotational
section IV. Our conclusions are presented in section V. state, then the rotational average is equivalent to an average over
2168 The Journal of Physical Chemistry, Vol. 98, No. 11, 199'4 Williamson and Zewail

all spatial orientations at the inst--it of scattering. The molecular again: after several picoseconds the sample will again be isotropic,
scattering intensity for a spatially isotropic sample takes the form albeit a mixture of excited and unexcited molecules.
Rotational motion is quantized and the rotational frequencies
are commensurable (or nearly so), and barring any collisions,
there will be some later time in which all the molecules return
to their original spatial orientation. The short-lived anisotropic
distribution will therefore appear again and again aver the course
of many nanoseconds.26 This phenomenon, known as rotational
wherejo is the zero order spherical Bessel function, or sin(sr,)/ recurrence, has been extensively studied in the context of time-
srij- resolved spectroscopy experiments, where rotational recurrences
The scattering amplitudes,ft andfi, decrease with the square have proven useful for determining structures of large molec~les?~
of s (ref 25), so the total scattering intensity decays like d;as The period of the recurrence in a symmetric top molecule is equal
a result, a graph of the modified molecular scattering intensity, to 1/ ( 2 8 ) where 8,the rotational constant, is related to the speed
sM(s),is better suited for revealing the fine details of the scattering of light c, Planck's constant h, and the moment of inertia 16 by
signal than a plot of the total intensity B = h/8r2Zbc. For asymmetric tops, the key observables are A,
B, and C.
It has been shown that the temporal width At of the recurrence
is related to the temperature T and the rotational constant by27

where a and b correspond to two atoms in the molecule. As

mentioned above, the molecular scattering function contains all
of the structural informationabout the molecule, but an illustrative where B has units of cm-l and T i s in Kelvin. For example, the
interpretation of experimental results is provided by taking the first recurrence of tram-stilbene (B = 0.252 GHz), occurring
sine transformof sM(s) andexaminingflr), the radial distribution 1.95 ns after excitation, has a full width at half maximum (fwhm)
function of 21.Ops at 5 K and 6.8 ps at 50 K, and these calculations26have
been experimentally tested. In a diffraction study of rotational
f i r ) = JomsM(s)sin(sr) ds (9) recurrences, only spectroscopic information about the initial
excitation transition is required; the recurrences would be probed
The radial distribution function plots the relative density of by following the diffraction pattern in time, as opposed to further
internucleardistancesin the molecule. This transform relationship excitation to another state with a second ultrafast laser pulse.
is strictly true only in the case of pure elastic scattering from a Kohl and Shipsey showed theoretically that the diffraction
homonuclear diatomic molecule,23but f(r) is nevertheless very pattern from molecules in a specific rotational state, with a well-
useful for a qualitative understanding of the structure of more defined nutation axis, is not cylindrically symmetric and contains
complicated systems. The range of experimental scattering extra clues about themolecular structure.28 The molecular sample
intensity is limited by some maximum value smax (e.g., 10-15 A-1 in UED, however, gains a different type of rotational order.
for the UED experiments reported in the accompanying paper), Rotational states are coherently populated by the excitation laser,
and the experimental radial distributionfunction typically includes but the judicious combination of a dipole transition, quantized
an exponentialdamping term to filter out artificial high frequency rotations, and high temporal resolution creates an ordered spatial
oscillations orientation distribution. We might expect that the diffraction
pattern of a spatially anisotropic sample will contain more
f(r) = JosmusM(s)exp(-k,.,s2) sin(sr) d s structural information than a conventional GED pattern. Equa-
(10)
tion 5 must therefore be reexamined to ascertain what the impact
of an anisotropic spatial orientation distribution formed by
It should be noted that all data analysis and fitting procedures ultrafast laser excitation will be on diffracting electrons. These
are performed on the experimental molecular scattering function, effects must be considered at short times, immediately after
and not the radial distribution. excitation, and over longer times, when the recurrences take place
The scattering factors which appear in the above equations cf, and the anisotropic distribution reappears. To emphasize the
7,S ) are functions of the atomic number 2 of the scattering changes in the diffraction patterns, it will be assumed that the
atom, the energy of the incident electron beam, and s. These excitation probability is unity. Experimentally, however, the
factors may be obtained from published tables.25 All scattering probability will be one-half at best because of stimulated emission
intensities and molecular dynamics calculations reported here and saturation effects, unless the upper state is dissociative or
were computed with the C programming language on an IBM Rabi cycling is introduced.
PC compatible. The incident electron energy was assumed to be Fink et af., treated the alignment problem for molecules
40 keV, ,s was limited to 15 A-1, and the radial distribution spatially oriented in a hexapole field.29 They have suggested that
function damping constant (kd)was set to 0.01 A2. changes in electron scattering would be potentially useful for
diagnosing the degree of spatial alignment in molecular beams,
111. Coherent Spatial Orientation
and their most recent experimental work has shown agreement
In UED experiments, the molecular spatial-orientation dis- with the theoretical predictions.30 The geometrical framework
tribution is dependent on the polarization of the laser and the presented by Fink et af. describes their alignment studies well,
type of transition accessed. If Q is the angle between the molecular but more degrees of freedom are necessary for modeling the
transition moment and the polarization of a linearly-polarized experimental arrangement discussed here. We outline our
laser pulse, then immediately after a parallel dipole transition methodology in part A of this section. Part B follows with a brief
the number of excited molecules at angle Q is proportional to cos2 synposis of the diffraction patterns generated by molecules in
Q. If the excitation probability is unity, then one-third of the total spatial alignment; this synposis will assist in interpreting
sample is pumped into a higher state and the remaining unexcited the more complicated patterns from laser-excited samples. The
molecules must have a sin2 Q distribution. In effect, the pump theoretical diffraction equations for the molecular subset selected
laser converts the isotropic sample into an anisotropic mixture of by linearly-polarized lasers in parallel and perpendicular con-
excited and unexcited molecules with well-defined spatial ori- figurations are established in parts C and D, and we will discuss
entation distributions. The duration of the anisotropy, however, their application to UED rotational recurrence studies of
lasts only until rotational motion randomizes the spatial orientation molecular iodine in part E. Part F concludes this section by
Ultrafast Electron Diffraction The Journal of Physical Chemistry, Vol. 98, No. I I . 1994 1769

Fij(s) = (eiar'9Spa,ial (13)

Fjjcontains the spatial average of the scattering function for a

specificinternuclear separation. Note thatFjjisa functionofthe
momentum transfer vector s, and not just its magnitude.
Figure 1 presents the geometrical frame of reference we have
chosen to evaluate Fjp The incident electron beam travels along
the z-axis, and the position vector rjj is oriented relative to the
z- and x-axes by an altitude angle Q and an azimuthal angle $,
Scattering respectively. Diffracted electrons strike the two-dimensional
Center detector at a position determined by the scattering angle 8 and
the rotational angle @ (defined relative to the x-axis)." In this
coordinate system, the momentum transfer vector is written:

Figure 1. Electron scatteringgeometrydefined in the text. The incident s = k,(-cos @ sin 8, -sin @ sin 8, 1 -cos 8)
electron beam travels toward the scattering center along the z-axiswith
wavevector h.The position vector between atoms i and j , ru. is defined = s(-cos $ cos(8/2), *in @ cos(8/2), sin(8/2)) (14)
byaltitudeanglen(relativeto ther-axis) andazimuthalangle+((relative
to thex-axis). Electronsscatter at angle0 with wavevector k and intersect and the position vector is written
theplaneofthedetectorat angle$ (relativetothex-axis). Themomentum
transfer vector is s. In a UED experiment, the excitation laser travels rij = rij(sin Q cos $, sin Q sin $, cos 0 ) (15)
toward the scattering Center along the y-axis.
The dot product of these two vectors simplifies to

- = srij[sin(8/2) cos Q - cos(8/2) sin Q cos(@- $)I

siij (16)
Y
Todetermine Fil,thescattering function ei'"umnst beintegrated
over thesurfaceof thespheredefined by nand$. Theanisotropic
"8 spatial orientation is included by multiplying the scattering

b, 8s: function with an appropriatedistribntion function,P(Q,$), before

integration

0G."3 ec-
-#la
Mathematicaldetailson theevaluationofthis integral for several
Molecular Sample relevant distribution functions have been consigned to the
(Isotropic)
Detector Appendix. For the case of an isotropic distribution (P(Q,$) =
I), the results is:
FigureL Electronscatteringfromanisotropcsampleofmoleculariodine.
On theieftisaschematicoftheinterscnionbetween theincidentelectrons
and the molecular sample. On the right is a perspective view of the
resulting modified molecular scattering pattern, sM(s). The scattering
pattern consists of cylindrical oscillations with frequency r1.112~. asexpected. andthccorresponding two-dimensional plot ofsM(s)
for molecular iodinc is presented in Figure 2 uith a drawing of
evaluating the structural information present in the anisotropic the molecular sample. The frequency of the oscillations in this
diffraction pattern from a more complicated molecule, trifluor- familiar ring pattern is equal to q.t/2x. wherc rl.l = 2.666A.I2
nmethvl inrlid~. B. Diffraction from Total Alignment. With the theoretical
A. Theoretical Framework. The scattering intensity, I,, of framework defined in part A, electron diffraction patterns from
a laser-excited molecular sample must be extracted from eq 5 , a laser-prcparcd molccularsamplc may be predicted. It is useful.
and to simplify this derivation we will initially consider only the howevcr,tofirstconsider thescattering pattern froma molecular
scattering from position vectors rij which are parallel to the sample with tntal spatial alignment. The functional forms of
molecular dipole transition axis. This is sufficient for describing there pattcrnr require no integration and may bederivcd dircctl)
all diatomics and linear molecules; more complicated molecules from cqs 6and 16. Twodifferent fundamental arrangcmcntsare
with nonparallel position vectors will be considered in part F. possible: (1 J all molecules arc oriented parallel to the direction
of the incident electron beam, and ( 2 ) all molecules are oriented
The vibrational period of most chemical bonds is significantly
perpendicular to the direction of the electron beam.
faster than a few picoseconds, and thus the starting point for this
In the firrt case (0= 0, thc bond axis is parallel to the z-axis).
analysis is actually eq 6, where the vibrational motion has been
the Scattering function for iodinc is equal lo
separated from the average. In the context of UED, the average
over rotational states should be evaluated as an average over
spatial orientation since many rotational states are populated;
the important perturbation is the initial (and subsequent recur-
rence) spatialorientationdistributionselected with theexcitation The diffraction pattern is cylindrically symmetric and its
laser polarization. Equation 6 may he rewritten as: section is an oscillatory function that changes slowly with s at
small scattering angles (ka >> s). As shown in Figure 3, the

Ids) = cccKjKl
N N

i=1 j=1.
exp(- ~i~sz)Re[ei'"'"JFij(s)l (I2)
modified molecular scattering pattern is an order of magnitude
more intense than electron scattering from an isotropic distri-
bution, but the slow onset of the oscillations makes it difficult to
I*J
extract structural information. This has a physical interpre-
tation-the molecules are struck by the electrons end on, and the
where second atom is "hidden" by the first atom along ko. Although
2770 The Journal ofphysical Chemistry. Vol. 98, No. 11, 1994 Williamson and Zewail

* 8'

88%
9
Si8

Molecular Sample
(Perpendicular) ..
Iletector
Figure 4. Electron scattering from a sample of molecular iodine with
total spatial alignment. The bond axis ofeach molecule is parallel to the
x-axis. On the left is a schematic of the intersection betwecn the incident
s (A-I) electrons and the molecular sample. On the right is a perspective view
Figure 3. Modified molecular scattering cross sections from diatomic afthecorrespandingmodifiedmolec"larscattering,sM(s).Theamplitude
iodine in three different spatial distributions: (1) isotropic, (2) total ofthe interference fringes is approximately an order of magnitude larger
alignment parallel to the incident electron beam (z-axis), and (3) total than the interferencefringes from the isotropic sample (see Figures 2 and
alignment perpendicular to the incident electron beam and parallel to the 3).
x-axis. The scattering intensity from the isotropic sample is an order of
magnitude lessthan thescattering intensity from spatially alignedsamples. shown in Figure4resultsfrom interference between theelectrons
The oscillation frequency of interference fringes from the parallel diffracted by these two scattering centers.
configuration is very slow for small s; the first cycle finishes at s = 22.2 Total alignment of the iodine molecules has affected both the
A-'. Scatteringfrom the perpendicular configurationis not cylindrically amplitude and the frequency of the interference fringes in the
symmetric. The interference fringes along the x-axis (parallel to the
internuclearaxis) arestrong,and the frequency issimilar tothe frequency diffraction pattern. The amplitude has increased by an order of
oftheisotropicfringes. Therearenooscillationsalong they-axis (normal magnitude, which implies that the signal-to-noise ratio for
to the internuclear axis). molecular scattering from spatially-oriented molecules is larger
than isotropic scattering. We might expected that this increase
the number of cycles increases with the square of s, there is little in signal-to-noise will counterbalance the decrease in the per-
information present in the range of s available to UED. centage of molecules contributing to the anisotropic diffraction
In the second case, when all molecules are spatially oriented pattern in an actual experiment.
perpendicular to the electron beam (R = 712, J. = 0, the bond The oscillation frequency of the diffraction signal from the
axis is parallel to the x-axis), the scattering function is first case (the parallel configuration) is significantly different
from the s a n d case (the perpendicular configuration). Struc-
tural information is more difficult to obtain from a parallel
configuration because the first few oscillations are very slow, and
the diffraction pattern would have to be measured out to large
values of s. The oscillation frequency of the fringes in the
perpendicular configuration, rt.I( 1 - (s/Zk0)911'/27, approaches
the frequency of the isotropic distribution,r1_1/2~,athigh electron
energies where ko is large. As will be emphasized in part F, the
Theprescnccofthccosotermdestroys thecylindricdlsymmetry. perpendicular configuration also contains information about bond
Thcreare nooscillationsa1ongthe)-axis, whereo = f ~2 (Figure / angles. Thediffraction pattern in Figure4 tells theexperimentalist
3). while the oscillations which exist along the x-axis have a that the iodine bond axis is parallel to the x-axis; if J. is nonzero,
period similar LO the intcrlercncc fringes from an isotropic then the interference fringes would be similarly rotated on the
distribution and are an order of magnitude larger. detector hyangleJ.relative to thex-axis. Thediffraction pattern
A tuo-dimensional plot of s M ( s ) for this second case. shown from more complicated molecules is a sum of contributions from
in Figure 4. clarifies the connection hctwecn the horizontal and individual atom pairs, and interference fringes in the diffraction
vertical cross sections presented in Figure 3. The diffraction pattern will reflect the differing spatial directions of the various
pattcrn from spatially-oriented molecules parallcl to the x-axis rjj.
is a series of linear fringes, as opposed to cunccntric rings. This C. Parallel Excitation. The laser beam, electron beam, and
figure is similar tothe interference pattcrn from Young's doublc. molecular beam are all mutually orthogonal in our UED
slit light bcattcrine- exocriment;
. in fact. the scattered electron e~periment.~ This
~ , ~experimental
~ design is motivated by ease
intensity of eq 20 may he wriken in a'form analogous to the of construction and a desire to reduce velocity mismatch
diffractedlight intensity froma pairofslitsseparated hya distance broadening in the total temporal resolution of the experiment.z0
d3' We will now consider the diffraction patterns which result from
molecules excited by a linearly polarized laser traveling toward
the scattering center along the y-axis, such that the polarization
vector is parallel to the wave vector b. For now it is assumed
(for 6 = 0) (21) that scattering takes place before the molecules have time to
rotate out of their initial distribution and that the entire subset
of molecules is excited. The scattering pattern at later times will
be considered in part E.
The spatial orientation distributions of both excited and
The potential wells of the two iodine atoms effectively function unexcited molecules will be cylindrically symmetric about b,
as a pair of "slits" for the incoming electrons, and the pattern and the diffraction pattern should reflect this symmetry. For a
Ultrafast Electron Diffraction The Journal of Physical Chemisrry. Val. 98, No. 11, 1994 2111

I ~ I*otroPic Excited ---- Unexcited 1 X l

L 2 -10 s,
Y

Molecular Sample
(Perpendicular 1 Excited) .-
Detector
Figure 5. Modified molecular scattering c r w scctioni of ground-rmtc
i d i n c from spatial urientation di,tribuuonscrcated by a linrirly pulanied Figure6. Electron scattering from thesubsetofa ground-statemolecular
laser. The polariration vector is parillel to ko and the molcculu dipole iodine sample excited by a linearly polarized laser. The polarization
trmsitiun i s a w m e d to be parallcl.sothc I~WCIEIIC muleculcsurwntcd vector is parallel to the x-axis and the molecular dipole transition is
porollpl to the electron beam Thc intcnrit) of the in1crfcren:c fringes assumed to be parallel. On the left is a schematic of the intersection
from thi,subsct ofthc molecular sample isan ,irdcruimagnitLdr.rmaller between the incident electrons and the perpendicular/ercited subset of
than the interfcrcncc fringcs from the isotropic simple. The 0,cillition
the molecular sample. On the right is a perspective view of the
f r c q x n c y is similar. but r 2 ridianr out ut phdrc. 'The intcrfcrencc corresponding modified molecular scattering, sM(s). The scattering
fringes from the unexcited sub,et are m x l a r i n inlcnsit) and frcqucnc'! pattern resembles a cross between diffraction from an isotropic sample
to the isotropic m t i r r m g pattern (Figure 2) and a totally aligned sample (Figure 4). The interference
fringes along the x-axis are similar in intensity and amplitude to the
interferencefringes from an isotropic sample. The interference fringes
parallel dipole tranbition. theorientation of the intcrnuclcar axis along the y-axis are weak and resemble the scattering pattern from the
ofthecxcitcd molecules will haveaspatialdisrributionofP(R.J) parallel/excited subset shown in Figure 5 .
=cosZR Thcspaual di5tribution ofthe unexcited molcculcsuill
therefore be P(R.$)=sinzl.l. Turciter~te,ueareonlyconsidering D. Perpendicular Excitation. Significantly different effects
position vectors which 3rc parallel to the cxsitation axis, and the are observed when the polarization vector of the laser is parallel
excitation probability IS unity. Integration of eq 17 using these to the x-axis and perpendicular to ko. As in the example of total
distribution5 yields the following functions for F,, (see Appendix) perpendicular alignment, the scattering pattern loses its cylindrical
symmetry because the spatial orientation distribution selected
parallel, excited with the polarized laser is not symmetric about the z-axis. With
an excitation probability of unity, the distribution of the excited
moleculesisdefinedby P(Q,$)=sin2acosz$andthedistribution
of the unexcited molecules is (1 - sin2 a cos2 $). Integration of
eq 17 with these distributions is discussed in the Appendix and
leads to the following functional forms for Fij:
parallel/unexcited perpendicularJexcited
j, b v 1
Fii(s) = jo[sr..] -~srii
'J
+ sin2(O/2)j2[srii] (24)
+
Fij(S) = j'[sriil cos2(S/2)j2[srii] cos26
~

(25)
sr..'I
As expected, these functions are independent of 4. The modified
molecular scattering function ofthese two distributions for ground- perpendicularJunexcited
state iodine are plotted with the isotropic distribution in Figure
5 . The scattering contribution from the parallel/excited subset
of molecules is relatively weak; evidently the strong scattering
signal observed in part B (when all molecules are oriented parallel
to the electron beam) is rapidly washed out by interference with As in the parallel case, the sum of these two functions is jo[sr,].
scattering from molecules slightly off-axis to the electron beam. the iso:ropic scattering distribution, but note that the presence
Because this contribution is weak, the oscillation frequency of of the cos2 6 term means that each function is not cylindrically
thescattering intensity from the parallelJunexcited subset isnearly symmetric. A two-dimensional plot of sM(s) for the perpen-
identical to the isotropic distribution. The oscillation frequency dicular/excited subset of molecules is shown in Figures 6 and 7a,
from the parallel/excited subset is similar, but phase-shifted by and theperpendicular/unexcitedcaseisshowninFigure7b.These
~ 7 1 2 Sincealldistributionsarecalculatedfor
. thesamespecies, are calculated for ground-state iodine, Both patterns appear to
the sum of the two scattering signals is equal to the diffraction be a cross between isotropic scattering (Figure 2) and scattering
signal from the isotropic distribution. Consequently, if the excited from a sample with total perpendicular alignment (Figure 4),
species is structurally identical to the unexcited species, the and despite the fact that only one-third of the molecules are in
diffraction pattern will not change. the perpendicular/excited subset, the amplitudeof the interference
Structural information inthe diffraction pattern from a sample fringes along the x-axis is as strong as the isotropic distribution.
with total parallel alignment was identical to the information Furthermore, the 4 dependence of the fringes indicates the
from an isotropic distribution but more difficult to extract, and directionality of r,., as was discussed in part B.
thesameis trueforthediffractionsignal from theexcitedJparallel The two-dimensional radial distribution functions of the
subset. Like the isotropic pattern, the excitedJparalle1 signal isotropic distribution and the perpendicular/excited distribution
depends only on the magnitudes of the position vectors. It would are shown in Figure 7c,d. Admittedly, taking the sine transform
bedifficult, however, todetermine the molecular structureof the of the function in eq 25 to extract the radial distribution is a
excited-state species because the low amplitude signal is masked crude approximation, and this is reflected in the fluctuating
by the diffraction signal from the unexcited molecules. baseline of the surface (Figure 7d). We will consider more
2172 The Journal ofphysical Chemistry, Vol. 98, No. 11, 1994 Williamson and Zewail

Internuclear Separation (A) Internuclear Separation (A)

Figure 1. Diffraction from ground-state iodine. (a) sM(s) for the perpendicular/excited subset, This pattern was also shown in Figure 6. (b) sM(s)
for theperpendicular/unexcited subset. Theunexcited molecules are aligned parallel to the excitation laser (along they-axis), sotheinterference fringes
are strong along the y-axis and weak along the x-axis. (c) Two-dimensional radial distribution curye from an isotropic sample of ground-state iodine.
The ring is located at -2.1 A. Only one quadrant is shown. (d) Two-dimensional radial distribution curve from the perpendicular/excited subset
of the molecular sample. The intensity at 2.1 A is strong along the x-axis and weak along they-axis. This reflects the physical orientation of the iodine
sample in the labratory frame of reference. Only one quadrant is shown.

appropriate techniques for extractingf(r) in future publications. and electron diffraction is quite sensitive to small changes in
For now, however, the sine transform is sufficient to provide molecular structure. For example, Figure 8 shows that there is
excellent qualitative information, and when scattering from a clear difference between the radial distribution curves corres-
trifluoromethyl iodide is considered in part F, the radial ponding to the lowest vibrational levels of ground-state iodine
distribution function will provide a description of the structure andof B-stateiodine. In thenext few paragraphs, we willconsider
that is easier to interpret by visual inspection than the molecular how to combine rotational coherence effects with the diffraction
scattering function. As might be expected from the appearance patterns discussed in parts C and D such that more structural
of the perpendicular/excited molecular scattering function, the information about thescattering molecules can be extracted than
corresponding radial distribution function also indicates the in a conventional gas-phase electron diffraction experiment.
directionality of the iodine-iodine internuclear separation during With an isotropic distribution, it is possible to improve the
the experiment. The disappearance of the 2.7-A peak along the signal-to-noise ratio by summing the total scattering intensity a t
y-axis reflects the fact that most molecules oriented parallel to a constantvalueofson thedetector. When the laser polarization
the laser beam axis are not in the perpendicular/excited subset is parallel to the electron beam, the diffraction patterns from
aftera parallel dipole transition. The radial distribution function excited and unexcited subsets are cylindrically symmetric and
corresponding to the perpendicularlunexcitedsubset(not shown) may therefore beaveraged ina similar way over theentiredetector.
contains a peak along the y-axis and no peak along the x-axis. We haveseen, however, thatthisarrangementonly leads tosubtle
E. Experimental Applications: Ground- and Excited-State changes in the diffraction pattern, and that these patterns do not
Structures. Thediffraction pattern from laser-prepared molecules appear to contain any extra structural information. We will
differs from the isotropic pattern because excitation is a selective instead turn our full attention to excitation with the laser
process that depends on the spatial orientation of the molecular polarization perpendicular to ko.
dipole. Thisdifferencewillnot beseen, however, ifthemolecular In this case, the diffraction pattern loses cylindrical symmetry
structuresofthe twostatesaresimilar becausethetotaldiffraction and the signal cannot be averaged over the entire detector.
pattern is always a sum of scattering from the excited and the Variations in thediffraction pattern must be particularlydistinct,
unexcited subsets; if the two species have identical r,, then this and ideally there would be some way to separate the scattering
sum is just jo[srjj]. Fortunately, the excited-state structure is signal of the excited species from the unexcited species. The
usually changed from the ground-state species to some degree, scattering patternat early times,in theinstant after themolecular
 -b
Ultrafast Electron Diffraction The Journal of Physical Chemistry, Val. 98, No. 11. 1994 2113

Ground although they will have undergone several vibrations. Perhaps

the simplest experiment of this type is excitation of iodine from
Exclled State the X state to the bound B state (a parallel dipole transition). For
demonstration purposes, the influenceof Franck-Condon overlap
will be ignored, and it is assumed that the excitation laser places
ground-state molecules into the lowest vibrational level of the B
s state. The ground state molecules are also assumed to be in the

-
lowest vibrational level. (A more detailed investigation of the B
X transition must account for the impact of overlap and
vibrational motion and is delayed until section IV.) The resulting
diffraction pattern is a sum of the perpendicular/excited dis-
tribution (eq 25) evaluated for the B state (11.1 = 3.0267 A, 11.1
0 1 2 3 4 5 = 0.04598 A) plus the perpendicular/unexcited distribution (eq
26) evaluated for the X state (11-1 = 2.6664 A, 11.1 = 0.03519
Internuclear Separation (A) .&).I4 The modified molecular scattering pattern from this system
Figure 8. Radial distribution curves for ground-state (X) and excited- (Figure 9a) reflects the anisotropy of the sample. It is clear that
state (B) iodine at vibrational level Y = 0. the oscillation frequencies along the x- and y-axes are different,
but extracting informationon the excited-statespecies is nontrivial
because the background signal from the unexcited molecules also
depends on 4. Other possible approaches should be considered,
as discussed below.
Iodinemay beexcited toadissociativestate insteadofa bound
state. The dissociation time is very fast-within a few hundred
femtoseconds"-so shortly after excitation the molecular sample
will consist of atomic iodine in the perpendicular/excited
distribution and molecular iodine in the perpendicular/unexcited
distribution. The diffraction pattern will therefore he equivalent
to Figure 7b. The change in the diffraction pattern is certainly
distinct, and additional structural information about the ground
state species has been determined-the directionality of rl.l.
Over longer times, the anisotropicdistribution ofthe unexcited
iodine molecules will be lost due to rotational motion, and the
electron diffraction pattern will correspond to an isotropic
distribution. The amplitude oftherings will be two-thirdsofthe
0 early-time amplitudealong thex-axis,assuming that the excitation
5 10 laser has dissociated all of the molecules in the perpendicular/
5 10 excited subset. Approximately 1 / 2 8 picoseconds after laser
excitation, however, a rotational recurrence will take place and
the anisotropic distribution will appear once again. (A half-
recurrence will also occur earlier at 1/48 P S . ) ~By ~ taking a
series of diffraction patterns in time, it will be possible to mimic
the spectroscopic rotational recurrence experiments of Felker et
ai. and determine the rotational constant with UED. The
experimentalist looks for those times at which the isotropic
distribution changes to the anisotropic pattern. Rotational
recurrencestudiesoftheground-state species are therefore possible
with UED.
Observing rotational recurrences with UED is a time-domain
technique for measuringrotational constants, but these rotational
recurrences are useful for distinguishing bctween the structures

-
of ground-state and excited-state species. Recall that the early-
time diffraction pattern from excitation of the B X transition
was anisotropic hut did not distinguish well between the excited-
and ground-state species. This will not be the case at later times
0 10 because the rotational constants of the ground- and excited-state
5 species are different, and the recurrences will be separate and
5 10 distinct. As an example, with ground-state iodine (v = 0). EX =
Figure9. Modified molccularscatteringcuwesfromamiitureofground- 0.0373 cm-I and 1/2& = 447 ps; for iodine in the B state ( v =
state and B-state iodine. (a) Scattering pattern immediately after 0), BE = 0.0289 cm-' and 1/2& = 577 ps." These recurrences
excitation. The excited iodine is in a perpendicular/excited distribution are well-resolved in time, and the scattering pattern at each
and the ground-stateiodine 1s in a perpendicular/unexciteddistribution.
The oscillation frequency along the x-axis is slightly greater than along recurrence will correspond to a sum of one species in an isotropic
they-axis, which reflects that B-state molecules (q.1 = 3.027 A) in the distribution and the second species in its initial anisotropic
sample are oriented parallel to the x-axis and the unexcited molecules distribution. Studies of the excited-state species are best
@ . I = 2.666A) areoriented parallel to thcy-anis. (b) Scattering pattern conductedwithanunpolarizedlaser,~~ thatthediffractionsignal
a few picoseconds after excitation. The spatial anisotropy has been lost for the excited species is a sum of the parallel and perpendicular
due to rotational motion. The diffraction pattern now corresponds to an distributions. Although the parallel contribution does not add
isotropic sample of'/, ground-state icdine and I / , B-state icdine.
much to the scattering pattern, the isotropic background signal
sample has been excited by the polarized laser, should be from the unexcited molecules is reduced by a factor of 2. The
considered first. The excited and unexcited molecules have not excited pattern is easier to discern because only onc-third of the
hada chanceto rotateout oftheiranisotropicspatial distributions, sample remains unexcited.
1114 The Journal of Physical Chemistry. Vol. 98, No. 1 1 , 1994 Williamson and Zewail

Figure 11. (a) Structure of trifluoromethyl iodide. (b) Spatial distri-

butions of the four CFlI pition vectors after a parallel molecular dipole
IransitionalongtheC-Ibondaris has beenexcited. The threcremaining
position vectors (TCF. rF.+. rF-1) sweep out cones definedby an angle Rtj.
where Rjj is the angle between IGI and rp
0 10 which were parallel to the molecular dipole transition axis. To
5 correct this, we willexamine thescattering pattern fromall position
s, 10
vectors in CFJ after the moleculehas been excited along a dipole
transition parallel to the C-I bond.
The structure of CFJ is shown in Figure 1la. Excitation of
a dipole transition parallel to the C-I bond is insensitive to the
angular position of the fluorine atoms around the C-I axis, and
thescatteringpattern must amount for an averageoverall possible
rotational orientations of the other internuclear separations. As
demonstrated in Figure 1Ih, each position vector r, sweeps out
a cone about the C-I axis; the angle between the C-I axis and
each cone is defined by RIP The geometry of this arrangement
is shown in Figure 12. The angle a defines the angular position
of rv about the excitation axis, and in the primed coordinate
system, rlj is

rij = rij(sin Rij cos a,sin Rlj sin a,0) (27)

where Olj is the angle between rij and the excitation axis. The
0 primed m r d i n a t e system may he transformed to the unprimed
5 . ' 10 coordinate system according to
s, 10
Figure 10. Modifiedmolecularscatteringcurvesfromamixtureofground- sin R cos $
state and B-state iodine during subsequent rotational recurrences. (a)
447 ps, ground-state rotational recurrence. The molecular sample has
El [
= rlj cos Ro sin R sin $
cos0 1 +
been excited witha linearly polarized laser polarization parallell ox-axis)
and is ' 1 3 B state, 2/3 ground state. Ground-state iodine returns to a
perpendicular/unexcited spatial distribution. B-state iodine remains
isotropic. (b) 577 ps. B-staterotationalrecurrence. The molecularsample
has been excited with an unpolarized laser and is therefore B state,
' 1 3 ground state. B-state iodinereturns to an unpolarized/excitedspatial
cos !

[
l $ -sin $ sin R cos J.
cos
cos R sin $ cos $ sin R sin $
+in O 0 cos
Application of this transform to eq 27 leads to
][I:] (28)

distribution. Ground-state iodine remains isotropic.

r.. =
The scattering patterns for molecular iodine a t 447 and 577
psarepresented in Figure IO. At 447 ps, theinterference fringes (sin R4 cos a ws D + cos Do sin D) cos *-sin Oil sin a sin J.
corresponding to ground-siaie iodine are enhanced along the
y-axis; similarly, a t 517 ps the interference fringes of exciied-
Sioie iodine are enhanced along the x-axis. In both figures, the
x- and y-axis scattering profiles are different in frequency and
[ +
rlj (sin O,j COS a cos R ws R, sin R) sin *-sin D,j sin a cos J.
cos R4 cos R - sin R, cos 01 sin J.
(2%
1
amplitude; separation of the ground- and excited-state scattering
patterns has been accomplished by exploiting the differences in Note that if rlj is parallel to the excitation axis (O,j = 0)then eq
the period of the rotational recurrences in UED. Note that the 29simplifiestotheearlierdefinitionofr4(eq15). Themomentum
temporal and spatial characteristics of this experiment enhance transfer vector, s, is not dependent on R, and a,but Fij must now
the capability of electron diffraction to give both ground-state include a third integration about the circle defined by a and r,
and excited-state structures. sin 0,
F. More Complicated Molecules, CF& The enhancement of
the interference fringes shown in Figure 10 is perhaps a little
misleading because iodine is a simple diatomic molecule. In
general, the scattering pattern of a molecule with many atoms
is a sum of different frequency components from the various It is important to note that the weighting distribution P is
internuclear separations, and therefore the total signal results independent of Rlj because P depends only on the angle between
from a combination of constructive and destructive interference the excitation axis (the C-I bond in the example discussed here)
between these different components. Furthermore, the scattering and the laser polarization axis. The inner integral may be
theory developed in part A does not adequately describe more evaluated analytically using an identity discussed in the Appendix.
complicated molecules because we only considered position vectors The resulting double integral was integrated numerically using
Ultrafast Electron Diffraction The Journal of Physical Chemistry, Vol. 98, No. 11, 1994 2775
Mathematica and analysis of these molecular scattering curves
led to the following equation for the perpendicular/excited
distribution when Qij is arbitrary

This equation agrees with the numerical integration results to at

least ten decimal places; note that eq 31 reduces to eq 25 when
n, = 0.
The structural constants rfj, 111,and nufor CF31are summarized
in Table 1.36 The two-dimensional radial distribution curves
corresponding to the perpendicular/unexcited subset of CFJ are
presented in Figure 13. The isotropic pattern consists of three
rings: the inner ring corresponds to the C-F atom pairs, the
middle ring is an overlap of the C-I and the F-F atom pairs, and
the outer ring corresponds to the Fa-I atom pairs. A cross section
of this cylindrically symmetric pattern is shown in Figure 14. All Figure 12. Scattering geometry for a position vector ri, that is not parallel
three rings are present in the pcrpendicular/unexcited distribution, to the excitationaxis. The excitation axis (2’) is defined by altitude angle
Q and azimuthal angle tj (see Figure 1). The angle between ru and I’
but now the amplitude of each ring varies with the angle 4. The is Q,. The position vector sweeps out a circle in the primed coordinate
order in the molecular sample (imposed by the anisotropic frame, where angle a is defined relative to the x’-axis. Calculation of
distribution) provides more structural information in the dif- the scattering pattern rquires a triple-integration over a,tj, and Q. To
fraction pattern by separating the atom pair components as a
function of 4. The amplitude of the middle ring is approximately
half as large as the isotropic distribution because the C-I and the
= ru co8 Q, and not at z’ 0.-
simplify the illustration, the x’and y’axes are shown intersecting at z’

Fa-F components are orthogonal. The C-I scattering contribution

TABLE 1: Structural Parameters for Trifluoromethyl Iodide
(Ref 36).
should be similar to the molecular iodine pattern in Figure 7b;
since rc-I I rF-F (QF-F = goo), the contribution from F.-F atom atom pairs no. r,lA hJlA Qijldeg
pairs is closer to the pattern in Figure 7a. The sum of the two C-F 3 1.3314 0.0510 110.6
contributions leads to a complete ring at -2.15 A. The area of c-I 1 2.1409 0.0465 0.0
a radial distribution curve peakis proportional to n&Zj/rij, where F-F 3 2.1535 0.0547 90.0
F--I 3 2.8916 0.08 12 154.5
nij is the number of identical atom pairs present in the molecule
and Zi b the atomic number of atom i. In many cases it would a Internuclear separation, ru, mean amplitude of vibration, 11,; angle

be possible to idenfity the two components of the middle ring between rij and C-I axis, Q,.
because their maximum peak areas would be different, but in this
The CF3 radicals will take on a perpendicular/excited distribution
example it happens that 2~21- 328.It is still relatively simple at each rotational recurrence; since the moment of inertia ofCF3
to identify the components, however, because the C-I interference
is not equal to the moment of inertia of CF31, the rotational
fringes are damped by the cos(qc - 71) phase term and the F-F
recurrence times of CF3 will be different (and resolvable) from
interference fringes are not damped. Consequently, the F-.F
the parent species.
peak is expected to be narrower and taller than the C-I peak, and
this is confirmed in the cross sections (Figure 14).
IV. Vibrational Coherence
The perpendicular/unexcited radial distributions in Figure 13
and in the profiles of Figure 14 provide qualitative information In the previous section it was shown that additional structural
about the angles QC-F and &...I. A quick glance at the inner ring and spectroscopic information can be obtained by observing the
shows that the C-F bond is almost orthogonal to the C-I bond diffraction pattern of a laser-excited molecular sample during
because the peak is larger on the x-axis than on the y-axis. rotationalrecurrences. The experimental factors that enable this
Conversely, the amplitude distribution of the outer ring shows are the ultrafast laser pulse, which creates a coherent superposition
that Q F - ~ must be closer to Oo (or 180O) than 90°. It will be of rotational states, and the ultrafast electron pulse, which has
possible to determine aij with great precision by carefully fitting sufficient time resolution (a few picoseconds) to record the
thex- and y-axis molecular scattering curves, and a more detailed diffraction pattern of the molecular sample during the rotational
discussion of this process will be presented in future publications. recurrence window.
In X-ray crystallography, the entire structure of the unit cell Here we investigate the electron scattering patterns from a
is mapped out by taking a series of diffraction patterns using molecular beam sample with vibrational coherence. A femto-
different crystal orientations. This process can be imitated in second laser pulse creates the coherence, a superposition of
UED to some extent by exciting different dipole transitions-the vibrational eigenstates. Traditionally, the time evolution of this
perpendicular/unexcited radial distribution of CF31 will be superposition, or wave packet, is monitored by a second femto-
different if the excitation axis is parallel to a C-F bond. These second laser pulse with FTS to follow the molecular dynamics of
diffraction patterns thus contain information about the nature of unbound, bound, and quasi-bound systems.17 We wish to show
the excited transition (perpendicular versus parallel, etc.), in that UED provides similar information about the dynamics and
addition to structural details. potential energy surfaces of molecular systems; the simulations
Finally, rotational recurrence UED provides a unique oppor- presented are important to test two regimes for- structural
tunity for determining the rotational constants and structure of changes: coherent motion and the kinetic regime. UED may be
radicals and other dissociation products. If the dissociation takes conducted even when little spectroscopic information is known
place more quickly than the duration of the rotational recurrence, about the system-the probe involves only the state of interest,
then the dissociation products will have anisotropic spatial and not a transition involving two states. Furthermore, the
orientation distributions. For example, CF31readily dissociates dynamics along the reaction coordinate can be extracted for
into CF3 and atomic iodine when excited by a 266-nm laser pulse. complex systems because UED records all structural changes. In
1116 The Journal of Physical Chemistry, Vol. 98, No. I I , 1994 Williamson and Zewail

(a)

4.0 4.0

0 1 2 3 4
Internuclear Separation (A)
Figure 14. Radial distribution curve cross sections of isotropic CFjI and
of the perpendicularfunexcited CFjI subset. The curves have k e n
baseline-corrected to emphasize thechanges in each pcak (wmpare with
the uncorrected surface in Figure 13b). As discussed in the text, the C-I
and F-I pcaks arc most intcnse along the y-axis and least intense along
the x-axis. The C-F and F-F peaks are most intense along the x-axis.
scattering sample and assume that the excited molecules have an
isotropic distribution The diffraction patterns will therefore be
cylindrically symmetric and consist of sums of the zero-order
spherical Bessel function, jo. This will a m u n t for the rotational
averaging shown in eq 5 , but vibrational averaging must be
considered more carefully. Our classical trajectory calculation
consists of a molecular ensemble in which the initial conditions
of each trajectory are randomly determined with a Monte Carlo
technique according to some probability distribution, The total
number of trajectories, NT. must he sufficiently large such that
the macroscopic behavior emerges by averaging the behavior of
the individual trajectories. Consequently, if the molecular
scattering curve is calculated from a large numberoftrajectories,
Internuclear Separation (A) than the total scattering should automatically a m u n t for
Figure 13. Two-dimensional radial distribution curves of CF& (a) vibrational averaging.
Isotropic sample. The inner ring wrresponds to the C-F separation, the Themolecular scattering intensity fromaset ofNTtrajectories
middle ring is a sum of the C-I and F-F separations, and the outer ring
wrresponds to the F-I separation. Only one quadrant is shown. (b) for molecular iodine simplifies to
Perpendicularfunexcited subset of the molecular sample after laser
excitation along the C-I band axis. Only one quadrant is shown. The
inner ring is most intense along thc x-axis because ~ G F is almost
perpendicular to rc.1. The outer ring is most intense along the y-axis
becauserF-Iisalmost parallel torc.1. The twocomponentsofthemiddle
ring have been separated along the x- and y-axes because rF.+ IrGI. where r.(f) is the evolution of the internuclear separation of the
Scattering from TCI is found along the y-axis. This peak is less intense nth trajectory in time. Two-dimensional diffraction plots with
than therp-Fpeakalong thex-axis becausescatteringfromrc-lisdamped the modified molecular scattering along the abscissa (in A-I) and
by thews(qc-s) term. Distortions in the baseline (seesection II1,part
D) actually exaggerate this effect. but it can still be ObseNed in the time along the ordinate (in fs) will be created. Generating radial
corrected profiles (Figure 14) distribution curves from these plots is straightforward.
B. Molecular Dynamics Calculations. With molecular dy-
this section we will comparethediffraction patterns from electron namics simulations of solutions, Bergsma et al. have shown the
pulse/velocity mismatch temporal resolutions (re)of both 50 and calculated Vibrational dynamics of iodine in different solvents
500 fs. The experimental limit of time resolution is discussed in following X-ray diffraction.'6 Here, the trajectory calculation
the accompanying paper.21 was designed to mirror the UED experimental excitation process,
The diffraction patterns are determined as a function of time where the pump laser excites isolated molecular iodine, creating
by calculating the electron scattering from classical trajectories a wave packet whichevolvesin timeon theB state. Thisevolution
ofmolecular iodineexcited totheBstate. Asaresult,thissection was recorded for each trajectory inaone-dimensional array,r.(f),
begins with a discussion of the scattering equations involved in and the diffraction pattern at a specific time 1, was obtained by
trajectory calculations, and the trajectory calculations themselves summing the scattering contributions from all rn(f,) (eq 32). We
are then described in part B. We discuss the diffraction patterns wish to compare the diffraction patterns from several different
when B-state iodine dissociates (part C) and when it is trapped excitation energies, 2000,3000, and 4000 cm-I, which are below
within the potential well (part D), and conclude this section by the dissociation energy, and a t 4750 cm-I, which is above the
examining the experimental applications of UED to molecular dissociation energy. These energies are relative to the bottom of
dynamics studies on several different time.scales. the B-state potential
A. Theoretical Framework. To emphasize the role of vibra- One thousand trajectories were used for the bound systems,
tional coherence in the diffraction patterns, we will ignore the and tenthousandtrajectorieswereusedforthedissociativesystem.
effects of anisotropic spatial orientation distributions in the It wasassumed that theinitialground-state molecular iodinewas
Ultrafast Electron Diffraction The Journal of Physical Chemistry, Vul. 98. Nu. 11,1994 2111

(a) ,
I
I ~ .
..~~
~

i
\-._

s (A') - 10

Internuclear 10 0 2 4 6 8 1 0
Separation (A) Internuclear Separation (A)
Figure 15. Molecular dynamics of B-state iodine excited to 4750 cm-I Figure 16. Molecular dynamics of B-state iodine excited to 4750 Em-'
above the bottom of the potential well. Iodine dissociates at this energy. above the bottom of the potential well. Iodine dissociates at this energy.
Electron pulse temporal resolution: r. = 50 fs. (a) Modified molecular Electron pulsetemporal resolution: 7. = 500 fs. (a) Modified molecular
scattering curve as a function of time. (b) Radial distribution curve as scatteringcurveas a functionof time. Theascillationsdampoutquickly
a function of time. The hump at 400 fs appears because the wave packet because of the low temporal resolution. (b) Comparison of the radial
slows dawn as it climbs the outer potential wall before dissociation. distribution CUNS after I ps for = 50 fs and 7, = 500 fs. The radial
distribution C U N ~at time zero ( T ~= 50 fs) is shown as a refercnce.
vibrationally cold, with the initial energy and probability
distributions corresponding to Y = 0. The excitation laser pulse L ~ c . 3New
~ positions and velocities were calculated every 5 fs
had a Gaussian temporal profile (fwhm = 50 fs), and the energy using an embedded Runge-Kutta integration scheme designed
bandwidth was determined by assuming that the pulse was to conserve the total energy of each traject0ry.3~The calculation
transform limited. began a t 4 . 5 ps (with the static ground-state distribution) and
The goal of the trajectory calculation was to generate a two- continued out to +2.5 ps. The zero of time, f = 0, corresponded
dimensional array containing internuclear separation probabilities to excitation of the molecular sample with the pump laser.
as a function of time. The size of each position element was 0.01 Theduration of the pumplaser pulse was included by offsetting
A and the size of each temporal element was 5 fs. The initial time zero for each trajectory by a random amount determined
conditions of each trajectory were determined using the Monte witha MonteCarlotechniqueon thepumplasertemporal profile.
Carlo technique. The initial energy was equal to the excitation Prior to to.., the two-dimensional array was filled with the
energy (e.&, 3000 cm-I) plus an offset determined by the laser probability distribution ofground-state iodine normalized to unit
bandwidth, and the initial position was determined from the area. Evolution of the trajectory after was modeled by adding
probability distribution ofground-stateidine. Several iterations one to the array element corresponding to the current time and
were sometimes necessary to establish a valid combination of calculated position.
initial energy and position, particularly a t lower energies where The temporal profile of the electron pulse (including velocity
the Franck-Condon overlap between the X state and the B state mismatch effects) was assumed to be Gaussian. The duration of
is poor. The initialvelocitywas calculated from the initial energy the electron pulse was included in the calculation by convoluting
and the B-state potential surface, and a randomly determined the temporal behavior of a specific internuclear distance (one
direction for the velocity (Le., bond contracting or expanding) row in the array) with the temporal profile of the electron pulse.
was assigned. The final result was an array of internuclear distances ranging
After defining the initial conditions, the evolution of each in time from 0 to 2 ps. The modified molecular scattering curve
trajectory on the B-state potential surface was calculated. The was calculated from 0 to 15 A-' (in 0.02-A--1 steps) at each time
program employed a Hulburt-Hirschfelder potential energy interval using eq 31, and the radial distribution curve was
surface" with Dunham coefficients provided by Gerstenkorn and calculated from sM(s) with kd = 0.01 A2.
2118 The Journal of Physical Chemistry, Vol. 98, No. 11. 1994 Williamson and Zewail

(a) 1
, ._
-.__
, ..
'-,
'
..

Internuclear Internuclear
Separation (A) Separation (A)
(c)
-.
.
1-
(d) L.\\..,.
-. 1..
...~.
//
,/ \~

Internuclear Internuclear
Separation (A) Separation (A)
Figure 17. Molecular dynamics of B-state iodine excited below the dissociation energy (4381.8 cm-I): (a) 2000 cm-'. = 50 fs; (b) 3000 cm-', 1.
$I

= 50 fs; (c) 4000 cm-1, *I = 50 fs; (d) 4000 cm-1, . T =

~ 500 fs. In all graphs, the peak intensity at the inner and outer turning pints decreases as a
function of time, which shows that the classical wave packet is spreading and approaching a Static distribution.
C. Above theDissociationEnergy. The first systemof interest curve for re = 500 fs in Figure 16a. The interference fringes
was iodine excited to 4750 cm-1 above the bottom of the B-state are almost completely washed out after the first 400 fs and the
notentialwell. ThedissociationthresholdoftheB-stateislocated
at 4381.8 cm-l, and it is expected that the diffraction patterns
-
molecularscatterineintensitvisvervsmall. Acomoarison ofthe
radial distribution curves a t 1 ps for re = 50 fs and re = 500 fs
will reflect a continuous increase in the internuclear separation (Figure 16b) shows that the peak flattens significantly as the
of the molecule with time. The modified molecular scattering temporal resolution changes by an order of magnitude.
curve and radial distribution curves with re = 50 fs are presented D. Below the Dissociation Energy. We also performed
in Figure 15. The radial distribution curve shows that iodine moleculardynamicscalculationsfor idineatthreeenergies (2000,
molecules traverse the potential well within the first 500 fs and 3000,and4000c11-~)below thedissociationenergyoftheB-state.
subsequently dissociate. The position of the radial distribution The radial distribution curves are plotted as a function of time
peak increases in distance with time, and the peak also flattens in Figure 17. When the temporal resolution is high (7. = 50 fs),
out as the wave packet gradually disperses across the potential thevibrational motion is clearly present, even for fast oscillations
surface. The modified molecular scattering curve presents the at 2OOOcm-1. Closeexamination shows thatvibrationaldephasing
same information, but in the inverse-distance domain. The occurs, and one can expect that vibrational quantum coherence
frequency of oscillation increases with time as 11.1 increases. The would be apparent if a quantum mechanical wave packet
amplitude of the oscillations diminish rapidly as the wave packet calculation was conducted. As with FTS,a Fourier analysis of
spreads; scattering contributions from a wide range of internuclear this oscillatory motion would yield the spacing between the
separations lead to a low scattering intensity. Vibrational energy levels.41
The calculated behavior of iodine is similar to observations of The oscillation period a t 4000 cm-I is long enough that the
dissociating I2 and ICN made with FTWMand to the corres- vibrational motion may still be resolved when re = 500 fs (Figure
ponding quantum mechanical wave packet calculations of the 17d). although there appears to be rapid dephasing. The
FTS ex~eriment.".~2 UED Drovides the actual dvnamical vibrational motion a t lower energies is not resolved, and the
trajectory of the fragments. Aiso, by taking diffraction patterns diffraction pattern is static after t h e first 500 fs. This static
at various times after laser excitation and calculating the change pattern represents a weighted sum of the position probability
in the internuclear separation, it would he possible to determine amplitudes of the vibrational states excited within the laser
the kinetic energy distribution of the fragments and deduce the bandwidth, and therefore these diffraction patterns are useful
dissociation energy. Such measurements require femtosecond for determining the shapeofthe potentialenergysurface. Figure
time resolution, however, as is shown by the molecular scattering 18 shows the 2-ps radial distribution functions (with rC= 500 fs)
Ultrafast Electron Diffraction The Journal of Physical Chemistry, Vol. 98, No. 1I, 1994 2779
5000
I
4000

c
E 3000
2 0I M 4 0 6 0 8 0 1 0 0 I
*
w
r
C
2000
Time (DS)

1000

+--c--cI

1 2 3 4 5 6
Internuclear Separation (A)
Figure 18. Radial distribution curves at 2 ps from molecular dynamics
of B-state iodine excited below the dissociation energy. The RD curves
are shown superimposed over the B-state potential energy surface.
Electron pulse temporal resolution: T~ = 500 fs. At this temporal Internuclear - 5 6
resolution, the trajectories for 2000 and 3000 cm-l are static after the
first 500 fs, although vibrational oscillations of B-state iodine molecular Separation (A)
dynamics at 4000 cm-I may still be resolved (see Figure 17d). Figure 19. Reaction dynamicsof CzF412. Thefirst iodine atom dissociates
directly with 71 I0.5 ps. The second iodine atom dissociates with 7 2 =
for all three energies overlaid on the B-state potential energy 32 ps. (a) Ground-state iodine atom signal as a function of time (ref 43).
surface. Note that the curves for 2000 and 3000 cm-1 are static (b) Radial distribution curve as a function of time. At time zero the
and represent the average vibrational motion, but the 4000 cm-1 molecular sample is pure C2F412. After time zero the molecular sample
curve is still evolving in time. This type of experiment has no is a mixture of C2F.J radical and F2C=CF2. The composition was
need of ultrafast time resolution and is actually better suited to obtained from the experimental results for 71 and 72.
longer time scales in which the bandwidth of the excitation laser
is narrow enough to pick out specific vibrational states-state nential; see Figure 19a). The corresponding time evolution of
selective diffraction. theradialdistributioncurveisshowninFigure 19b. Themolecular
E. Experimental Applications to More Complicated Systems. scattering curve is simply a weighted sum of the three components
Here, we discuss ways UED might be applied to dynamical (C2F412, C2F41,and F&=CF2); the composition of the molecular
experiments where reaction intermediates are formed on longer sample was calculated as a function of time using measured values
time scales. The most important characteristic of gas-phase of T~ and 72. Structural parameters for C2F4I2 and F2C==CF2
electron diffraction is that the scattering patterns are extremely were obtained from the l i t e r a t ~ r e ~and
~ pthe
~ ~structure of the
sensitive to small changes in internuclear separation. Conven- radical component was assumed to be identical to C2F4I2 with
tional GED experiments typically report internuclear separations one iodine atom missing.
to within 0.005 or less;4 so far, a limited range of s is detected The radial distribution curve clearly reveals the dynamical
inUED(lSA--',asopposedto -4OA-linGED),andtheprecision changes of the fragmentation process. The primary bond breakage
is more on the order of 0.05 A. This sensitivity is still sufficient is shown by the rapid disappearance of the peak at 5.0 A, which
to distinguish between different vibrational states (Figure 18) as corresponds to the I-.I internuclear separation. The subsequent
well as between different electronic states (Figure 8). UED transformation of the radial distribution curve over the next 100
therefore provides another detection scheme for monitoring ps characterizes the secondary bond breakage; note that the
ultrafast decay processes, either when the products dissociate or intensity of the first peak at 1.3 A (C-C, C-F) does nor change
when they merely change to a different vibrational or electronic because all carbon and fluorine atoms remain in the molecule,
state, Such an avenue of detection is, of course, particularly although the position of the peak shifts inward slightly as the
useful when spectroscopic data about higher excited states is not C-C bond changes to a double bond. The ability of UED to
available or is too complicated. monitor this type of dynamical change has already been
As an example of how UED might be applied to dynamical demonstrated19 and is detailed in the accompanying paper.2'
studies, we can consider the three-body photofragmentation The primary and secondary bond breakages in C2F4I2 are very
process of CzF412 studied in real time.43 The photofragmentation different processes and help illuminate the distinctions between
is a two-step process in which both iodine atoms dissociate, leaving monitoring vibrational coherence versus following dynamical
tetrafluoroethylene decays. Loss of the first iodine atom is a direct process and very
fast; if the molecule is excited with a femtosecond laser pulse, a

-
'I
C2F412* C2F41* I* + (33)
wave packet is formed on the dissociative potential energy surface.
With a 50-fs electron pulse, it would be possible to monitor this
dissociation process in time, and the diffraction patterns would
'2 be similar to those shown in Figure 15. The 1-1 separation would
C2F41*+ F2C=CF2 +I (34) rapidly move outward from 5.0 A, and one set of C-I, C-I, and
Fa-I peaks would shift outward as well. Loss of the second iodine
The primary C-I bond breakage is very fast ( r l 5 0.5 ps), while atom, however, is not a coherent process. The individual C2F41
the secondary bond breakage results from internal energy radicals dissociate independently of each other, with the behavior
redistribution and is significantly slower (72 = 32 ps at a given of the entire population characterized by the decay constant 7 2 .
energy). The dynamics of the two steps depended on the total Trajectory calculations are unnecessary to model this type of
energy and were probed by measuring the rises of excited-state dynamical process because the loss of the second iodine is very
atomic iodine (I*) and ground-state atomic iodine (a biexpo- fast compared to the uncertainty in the onset of the dissociation.
2780 The Journal of Physical Chemistry, Vol. 98, No. 11, 1994 Williamson and Zewail

TABLE 2: Summary of Spatial Scattering Functions F, for Several Spatial Orientation Probability Distributions fin,$) Where
n,= 0.
spatial orientation fraction of
distribution molecules P(Q,+) Fij(5)

isotropic 1 1 Jo[sr,jl

parallel excited '13 cos2 a

parallel unexcited 2/ 3 sin2 Q

perpendicular excited 'I3 sin2 f? cos2 +

perpendicular unexcited 2 13 1- sin2Q cos2 +

unpolarized excited 1 - sin2Q sin2IL

unpolarized unexcited Ii3 sin2 f? sin2 +

perpendicular excited
S'ij
- (
sin2 Qij + (2-3 sin2 Q,j)

The bottom row shows the functional form of the perpendicu1arJexcited Fij for arbitrary

Thus the diffraction pattern may be calculated as a sum of acknowledges the National Science Foundation for providing a
scattering from each of the three components weighted by their Graduate Student Fellowship.
relative populations.
Appendix
V. Conclusion Many of the definite integrals generated from eq 17 are not
readily available in mathematical references. This appendix
Debye's work on X-ray diffraction showed that the scattering provides a general description of the integration process needed
pattern from a gas is rich with structural information, despite the to obtain the closed form solutions summarized in Table 2.
randomness in the position and orientation of the molecular Integration of eq 17 over J, is accomplished with one of the
sample. One focus of this paper was to show that ultrafast following identities:
preparation and ultrafast electron diffraction can be used to
introduce and observe order in gas phase samples, thus yielding
a new dimension in molecular structure analysis. The formation
of a coherent superposition of rotational and vibrational eigen-
states, which can be controlled by the experimental time delay, or
are transient impositions of this order on an otherwise isotropic
sample. Rotational recurrence UED provides a detailed structural (A-2)
picture of the ground-state species: internuclear distances,
moment of inertia (from the rotational constant), and bond angles.
Vibrational coherence studies reveal transition-state dynamics where y is defined by
and provide a map of the potential energy surface; this information
is very similar to results from FTS. Applications were discussed y = (A2 + B2)'/* ('4-3)
for small systems (two atoms), more complex systems (five or
eight atoms), and reactions involving direct dissociation or reactive Equation A-1 may be verified by expanding the exponential as
intermediates. The following paper details our apparatus, its a series and employing definite integrals listed in Gradshteyn
temporal and spatial resolutions, and experimental applications. and Ryzhik.468 Equation A-2 may be verified by expanding the
exponential, expanding each term in the exponential sum, and
Acknowledgment. This work was supported by a grant from integrating term by term.6b As anexampleof how theseidentities
the U S . Air Force Office of Scientific Research and the National are applied, eq A-1 simplifies F',,for the isotropic distribution to
Science Foundation. The authors thank Dr. Scott Kim and Dr.
Marcos Dantus for their stimulating discussions on ultrafast F,,(s) = iJ:sin R Jo[ sr cos sin R] eisrsin(e/2)oos*
dR (A-4)
electron diffraction, and Hakno Lee and Dr. Juen-Kai Wang for 2
their help. We thank Dr. Ralf Friichtenicht for assisting in the The final closed form is obtained by integrating over R using
translation of some of the referenced (German) articles. J.C.W. one of the following identities:
Ultrafast Electron Diffraction The Journal of Physical Chemistry, Vol. 98, No. 11, 1994 2781
(13) Mourou, G. A.; Williamson, S . Appl. Phys. Lett. 1982, 41, 44.
Jor(sin (p)"+'J,,[Bsin v]eiAcoscP
dv = 2 --j,[y]
B" (A-5) Williamson, S.;Mourou, G.; Li, J. C. M. Phys. Rev. Lett. 1984, 52, 2364.
Y (14) Elsayed-Ali, H. E.; Herman, J. W. Reo.Sci. Instrum. 1990,61,1636.
Herman, J. W.; Elsayed-Ali, H. E.; Murphy, E. A. Phys. Rev. Lett. 1993,71,
400.
or (15) Kitriotis, D.; Aoyagi, Y. Jpn. J. Appl. Phys. 1993, 32, L441.
(16) Bergsma, J. P.; Coladonato, M. H.; Edelsten, P. M.; Kahn, J. D.;
Wilson, K. R.; Fredkin, D. R. J. Chem. Phys. 1986,84,6151.
Jorcos2 cp(sin cp)"+'J,[B sin cp]eAmcPdv = (17) Zewail, A. H. Science 1988,242, 1645. Khundkar, L. R.; Zewail,
A. H. Annu. Rev. Phys. Chem. 1990,41,15. Zewail, A. H. Faraday Discuss.
Chem. SOC.1991,91,207. Zewail, A. H. J. Phys. Chem. 1993,97, 12427.
(18) Williamson, J. C.; Zewail, A. H. Proc. Natl. Acad.Sci. U.S.A.1991,
88, 5021.
(19) Williamson, J. C.; Dantus, M.; Kim, S. B.; Zewail, A. H. Chem.
Phys. Lett. 1992, 196, 529.
These integrals may be verified by expanding the Bessel function (20) Williamson. J. C.: Zewail. A. H. Chem. Phvs. Lett. 1993.209. 10.
as a series and making a change of variables to t = cos cp. (21) Dantus, M.; Kim, S. B.; Williamson, J. C.;-Zewail, A. Hi J . Phys.
Integration over t generates another Bessel function,* which is Chem., following paper in this issue.
(22) Karle, J. In Determination of OrganicStructures by Physical Meihods;
also expanded. The double expansion that results for eq A-5 is Nachod, F. C., Zuckerman, J. J., Eds.;Academic Press: New York, 1973;
P 1.
(23) Schiifer, L. Appl. Spectrosc. 1976, 30, 123.
(24) Yates, A. C. Phys. Rev. 1968,176, 173.
(25) Ibers, J. A,, Hamilton, W. C., Eds. International Tables ForX-Ray
Crystallography; Kynoch Press: Birmingham, 1974; Vol. 4, p 176. Sellers,
H. L.; Schiifer, L.; Bonham, R. A. J. Mol. Struct. 1978,49, 125.
(26) Felker, P. M.; Zewail, A. H. J. Chem. Phys. 1987,86, 2460.
and the double sum for eq A-6 is similar. The final step is to (27) Baskin, J. S.;Felker, P. M.; Zewail, A. H. J . Chem. Phys. 1987,86,
change the summation limits by defining i = m k and j = k + 2483. Baskin, J. S.;Zewail, A. H. J. Phys. Chem. 1989, 93, 5701. Felker,
P. M.;Zewail, A. H. In Femtosecond Chemistry; Manz, J., Wbste, L., Eds.;
VCH: New York, 1994. Baskin, J. S.;Zewail, A. H. J . Phys. Chem. 1994,
in..aress.
.
(28) Kohl, D. A.; Shipsey, E. J. Z . Phys. D 1992, 24, 3 3 . Kohl, D. A.;
Shipsey, E. J. Z . Phys. D 1992, 24, 39.
(29) Fink, M.: Ross, A. W.; Fink. R. J. Z . Phvs. D 1989,ll. 231. Mihill.
A.;'Fink, M. Z . Phys. D 1989, 14, 77.
Summation over j leads to a y2i term, and the sum over i (30) Volkmer, M.; Meier, Ch.; Mihill, A.; Fink, M.; BBwering, N. Phys.
Rev. Lett. 1992, 68, 2289.
corresponds to the final Bessel function. (31) Note that Fink et al initially define +relative to the x-axis in Figure
+
1 of their first paper (ref 29). It appears, however, that is defined relative
References and Notes to the y-axis in their text and in subsequent articles.
(32) Spiridonov, V. P.; Gershikov, A. G.; Butayev, B. S.J. Mol. Struct.
(1) Debye, P. Ann. Phys. 1915, 46, 809. 1979, 51, 137.
(2) Debye, P.; Bewilogua, L.; Ehrhardt, F. Phys. Z . 1929,30,84. Debye, (33) Halliday, D.; Resnick, R. Physics, 3rd ed.; J. Wiley & Sons: New
P. Phys. Z . 1930, 31, 419. York, 1978; p 994.
(3) Mark, H.; Wierl, R. Naturwissenschaften 1930,18,205. Wierl, R. (34) Luc, P. J. Mol. Spectrosc. 1980, 80, 41.
Ann. Phys. 1931, 8, 521. (35) Bowman, R. M.; Dantus, M.; Zewail, A. H. Chem. Phys. Lett. 1989,
(4) Hargittai, I., Hargittai, M., Eds. Stereochemical Applications Of 161; 297.
Gas-Phase Electron Dij/raction; VCH: New York, 1988, and references (36) Typke, V.; Dakkouri, M.; Oberhammer, H. J . Mol. Struct. 1978.49,
therein. 85.
(5) Norrish, R. G. W.; Porter, G. Nature 1949, 164, 658. Porter, G. (37) Hulburt, H. M.; Hirschfelder, J. 0. J. Chem. Phys. 1941, 9, 61.
Proc. R . Soc. London, A 1950, 200, 284. Hulburt, H. M.; Hirschfelder, J. 0. J . Chem. Phys. 1961, 35, 1901.
(6) Herzberg, G.; Ramsay, D. A. Faraday Discuss., Chem. SOC.1950, (38) Gerstenkorn, S.;Luc, P. J. Phys. (Paris) 1985, 46, 867.
9, 80. (39) Press, N. H.; Teukolsky, S.A.; Vetterling, W. T.; Flannery, B. P.
(7) Marshall, R.; Davidson, N. J. Chem. Phys. 1953, 21, 659. Numerical Recipes In FORTRAN, 2nd ed.; Cambridge University Press:
(8) Ischenko, A. A.; Golubkov, V. V.; Spiridonov, V. P.; Zgurskii, A. V.; Cambridge, 1992; p 701.
Akhmanov, A. S.;Vabischevich, M. G.; Bagratashvili, V. N. Appl. Phys. B (40) Dantus, M.; Rosker, M. J.; Zewail, A. H. J. Chem. Phys. 1988,89,
1983, 32, 161. 6128.
(9) Ischenko, A. A.;Spiridonov, V. P.;Schiifer, L.; Ewbank, J. D.J.Mol. (41) Gruebele, M.; Zewail, A. H. J. Chem. Phys. 1993, 98,883.
Siruct. 1993, 300, 115. (42) Williams, S.0.;Imre, D. G. J . Phys. Chem. 1988, 92,6648.
(10) Rood, A. P.; Milledge, J. J . Chem. SOC.,Faraday Trans. 2 1984,80, (43) Knee, J. L.; Khundkar, L. R.; Zewail, A. H. J. Chem. Phys. 1985,
1145. 83, 1996. Khundkar, L. R.; Zewail, A. H. J. Chem. Phys. 1990, 92, 231.
(1 1) Bartell, L. S.;French, R. J. ReLiSci. Insfrum. 1989,60,1223. Dibble, (44) Thomassen, H.; Samdal, S.;Hedberg, K. J. Am. Chem. SOC.1992,
T. S.;Bartell, L. S.J . Phys. Chem. 1992, 96, 8603. Bartell, L. S.;Hovick, 114,2810.
J. W.; Dibble, T. S.; Lemon, P. J. J . Phys. Chem. 1993, 97, 230. (45) Carlos, J. L.; Karl, R. R.; Bauer, S. H. J. Chem. Soc., Faraday
(12) Ewbank, J. D.; Faust, W. L.; Luo, J. Y.; English, J. T.; Monts, D. Trans. 2 1974, 70, 177.
L.; Paul, D. W.; Dou, Q.;Schiifer, L. Rev. Sci. Insrrum. 1992, 63, 3352. (46) Gradshteyn, I. S.;Ryzhik, I. M. Tables Of Integrals, Series, and
Ewbank, J. D.; Luo, J. Y.; English, J. T.; Liu, R.; Faust, W. L.; Schiifer, L. Products; Academic Press: New York, 1980 (a) p 382, eq 3.661-1.2; (b)
J. Phys. Chem. 1993,97, 8745. p 131, 2.512; (c) p 321, 3.387-2.