1: \section{Experimental Method} \label{Experimental_Methods}
2: \subsection{Neutron Scattering}
3: In the previous section we have presented a variety of experimental
4: glass transition signatures observed with conventional
5: spectroscopic methods.
6: What can we expect to learn about the glass transition from neutron
7: scattering?
8:
9: The energy of neutrons of the appropriate wavelength for structural and
10: dynamic studies corresponds to thermal energies for temperatures between
11: a few degrees K to well above room temperature.
12: In fact, neutrons produced from a reactor or spallation source have
13: their wavelength ``adjusted'' by the equilibration at a given
14: temperature; thus ``cold'' neutrons have a wavelength of several \AA ,
15: ``hot'' neutrons have a wavelength of a fraction of an \AA\ and ``thermal''
16: neutrons have a wavelength close to 1\,\AA .
17: Because of this wavelength--energy link, neutron inelastic scattering
18: allows the simultaneous study of spatial structure, on the one hand,
19: and dynamics, on the other, at the atomic and molecular level
20: \cite{Sch72,Squ78,Lov84,Bee88}.
21:
22: Neutron scattering experiments on viscous liquids
23: are motivated by the following observations:
24: In amorphous and crystalline solids vibrational excitation frequencies
25: are generally of the order of 1\,THz;
26: long wavelength phonon frequencies of sound waves are found in the GHz to
27: THz region depending on the wave number $Q$ \cite{Squ78,Lov84}.
28: Furthermore, the mean relaxation time in normal fluids is typically of
29: the order of 1\,ps \cite{Bee88}.
30: Thus, for the observation of the onset of the glass transition
31: -- when relaxational motions decouple from vibrational ones -- we have to
32: look at the GHz to THz range which is traditionally the domain of scattering
33: experiments.
34: Such high frequencies are nowadays accessible with many experimental
35: techniques %summarized in section~\ref{Other_Experimental_Approaches}.
36: like e.\ g.\ dynamic light scattering, inelastic X--ray scattering,
37: quasielastic neutron scattering, dielectric spectroscopy or impulsive
38: stimulated light scattering (section~\ref{Other_Experimental_Approaches}).
39: However, not all of these techniques have been applied to OTP.
40:
41: Neutron scattering remains one of the most versatile methods for the study
42: of the fast dynamics.
43: It can be used in almost every type of material since it does not depend on
44: optical transparency, dipole moments or polarizabilities \cite{Lov84}.
45: In neutron scattering one observes excitations at a well defined momentum
46: transfer $\bf Q$ of the order of several \Ar, {\it i.~e.} microscopic motion
47: on interatomic and intermolecular distances can be resolved.
48:
49: The observed quantity is the so called dynamic structure factor
50: $S({\bf Q},\omega)$ and is a well defined correlation function
51: \cite{Sch72,Squ78,Lov84,Bee88}
52: \begin{equation}
53: \label{Sqw-coh}
54: S({\bf Q},\omega)=
55: \frac{1}{N} \int \frac{{\rm d}\omega}{2\pi}e^{i \omega t}
56: \sum_{j,k=1}^{N} \langle e^{i{\bf QR}_j(0)}e^{-i{\bf QR}_k(t)}\rangle,
57: \end{equation}
58: which couples directly to the nuclear coordinates ${\bf R}_j(t)$.
59: It therfore provides an ideal link between experiment and theory as
60: $S({\bf Q},\omega)$ is the space--time Fourier transformation of the
61: well know van Hove correlation function $G({\bf r},t)$ \cite{Hov54}.
62:
63: Equation (\ref{Sqw-coh}) is based on the {\it coherent} superposition
64: of partial waves scattered by pairs of nuclei $j,k$.
65: The coherence can be diminished by spin flips except when both partial
66: waves are scattered by the same nucleus $j=k$.
67: This leads to an additional {\it incoherent} scattering
68: \begin{equation}
69: \label{Sqw_inc}
70: S_{\rm inc}({\bf Q},\omega)=
71: \frac{1}{N} \int \frac{{\rm d}\omega}{2\pi}e^{i \omega t}
72: \sum_{j=1}^{N} \langle e^{i{\bf QR}_j(0)}e^{-i{\bf QR}_j(t)}\rangle,
73: \end{equation}
74: which contains valuable information about tagged--particle motion.
75:
76: In an actual experiment one measures always the sum of both contributions
77: \begin{equation}
78: \label{DDCS}
79: S_{\rm exp}({\bf Q},\omega)=
80: \sigma_{\rm coh} S_{\rm coh}({\bf Q},\omega)
81: +\sigma_{\rm inc} S_{\rm inc}({\bf Q},\omega).
82: \end{equation}
83: Incoherent and coherent scattering can be separated by a polarization
84: analysis \cite{SchC93,Sch96} which is
85: based on the different scattering cross sections for
86: spin--flip $\sigma_{\uparrow \downarrow}=2/3 \sigma_{inc}$
87: and non--spin--flip $\sigma_{\uparrow \uparrow}=\sigma_{coh}+1/3 \sigma_{inc}$
88: \begin{equation} \label{sep}
89: \begin{array}{rcl}
90: S_{\uparrow \uparrow}({\bf Q},\omega) & = &
91: \sigma_{\rm coh} S_{\rm coh}({\bf Q},\omega)+
92: \frac{1}{3}\sigma_{\rm inc} S_{\rm inc}({\bf Q},\omega), \\
93: S_{\uparrow \downarrow}({\bf Q},\omega) & = &
94: \frac{2}{3}\sigma_{\rm inc} S_{\rm inc}({\bf Q},\omega).
95: \end{array}
96: \end{equation}
97: Although this implies a considerable loss in neutron flux in
98: high resolution experiments
99: such experiments are indeed feasible as shown below.
100:
101: In practice, most often one concentrates on either incoherent or
102: coherent scattering by choosing appropriate samples.
103: In organic samples like OTP, the total cross section is largely dominated
104: by incoherent scattering from hydrogen (table~\ref{cross-sections}).
105: Mainly coherent scattering can be measured on perdeuterated OTP-$d_{14}$.
106: Partial deuteration has been employed to hide the lateral phenyl rings
107: \cite{DeZB91} (section~\ref{t-squarerootsingularity}).
108:
109: % table cross-sctions %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
110: \input{Tables/cross-sections.tex}
111: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
112:
113: The spectrometers used for neutron scattering on OTP are listed in
114: table~\ref{instruments}.
115: They belong to basically five types \cite{Yel94} and are all
116: located at the Institute Laue--Langevin in Grenoble, France.
117: Three of them measure $S({\bf Q},\omega)$, one the so called intermediate
118: scattering function $S({\bf Q},t)$ which is the Fourier transform
119: of $S({\bf Q},\omega)$ and one spectrometer works without energy analysis:
120: \begin{itemize}
121: \item Triple--axis spectrometers (IN12) are instruments which can measure
122: the four dimensional dynamic structure factor $S({\bf Q},\omega)$.
123: In single crystals the phonon dispersions relations $\omega({\bf Q})$ can
124: explicitly be measured.
125: \item
126: Time--of--flight (TOF) spectrometers with multichopper (IN5) or
127: crystal (IN6, D7) monochromators where the energy of the scattered neutrons
128: is measured by the time they need to travel a certain distance are used
129: to measure vibrational dynamics as well as fast relaxation.
130: \item
131: Back--scattering (BS) spectrometers achieve high energy resolution by
132: (nearly) 180$^\circ$ Bragg scattering from both the monochromator and the
133: analyzer crystals.
134: They are used to measure the amplitude of elastic scattering and the onset
135: of the slow relaxation.
136: In order to measure full spectra the incident energy is changed either by
137: Doppler shifting the neutron wavelength (IN10, IN16) or by thermal expansion
138: of the monochromators (IN10, IN13).
139: \item
140: Neutron spin--echo (NSE) instruments (IN11) require by construction the use
141: of polarized neutrons.
142: They profit from the fact that the magnetic moment of the neutron undergoes
143: Larmor precessions in a homogeneous magnetic field.
144: This is used to encode the neutrons' energy in the number of Larmor
145: precessions.
146: The polarization analysis of the scattered neutrons amounts to an implicit
147: Fourier transform of the scattering function $\SQw$ and has a much larger
148: dynamic range as frequency domain spectrometers at the cost of a lower
149: contrast.
150: \item
151: Diffractometers (D7, D20) are used to measure the static structure factor
152: of liquids and amorphous solids.
153: The integration over all energies is intrinsically performed.
154: In addition to a time--of--flight option, D7 can perform a polarization
155: analysis and is thus able to measure
156: simultaneously incoherent and coherent scattering.
157: \end{itemize}
158: For isotropic samples like powders, liquids and glasses we can only access
159: $Q=|{\bf Q}|$.
160:
161: % table instruments%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
162: \input{Tables/instruments.tex}
163: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
164:
165: \subsection{Data Evaluation and Limitations}
166: To obtain the desired quantity $S(Q,\omega)$ from the measured
167: scattering intensities several standard corrections have to be performed
168: like normalization to the incoming flux and to the scattering of a
169: vanadium standard, subtraction of container scattering and background noise
170: and, if necessary, calculation of attenuation and absorption effects and
171: conversion to energy.
172:
173: Since all instruments measure at given scattering angles $\theta$ a
174: conversion to the physically interpretable wave number $Q$ has to be done.
175: In figure~\ref{Q-of-w-Matte} we show the dependence of the wave number
176: $Q$ on the energy transfer $\omega$ for several scattering angles
177: $2\theta=0 \ldots 180^{\circ}$.
178: % Figure %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
179: \input{Bilder/Exp/Q-of-w-Matte.tex}
180: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
181: For backscattering and neutron spin--echo instruments, where the energy
182: $E_i$ of the incident neutrons is much larger than the energy transfer
183: $\omega$ the momentum transfer $Q$ at a given scattering angle is
184: essentially constant (left panel).
185: However, when the energy transfer $\omega$ is of the order of $E_i$
186: -- for instance in time--of--flight spectroscopy -- then the dependence of
187: $Q$ on $\omega$ cannot be neglected (right panel) although sometimes still
188: done \cite{BuWR96}.
189: % and it is indispensable to
190: %interpolate the measured $S(2\theta,\omega)$ to constant $Q$ slices
191: %$S(Q,\omega)$.
192:
193: In any substance, one observes at intermediate temperatures significant
194: contributions to $S(2\theta,\omega)$ up to several 10\,meV.
195: Before calculating a Fourier transform to get $S(Q,t)$,
196: $S(2\theta,\omega)$ has to be interpolated to $S(Q,\omega)$.
197: As illustrated in figure~\ref{Q-of-w-Matte},
198: this imposes restrictions upon the
199: accessible energy transfers.
200: The Fourier transform $S(Q,t)$ can only be
201: evaluated if the frequency spectrum includes the elastic channel $\omega=0$.
202: The truncation problem at $\omega_{\rm max}$ is mathematically solved by a
203: multiplication with the step function and leads to a lower bound in
204: Fourier time $t$ below that
205: the Fourier transform is not reliable \cite{WuPC95}.
206: The truncation effect may also appear in
207: $S(Q,t=0)=2\int_0^{\infty}{\rm d}\omega S(Q,\omega)$ which is then less
208: than the true static structure factor $S(Q)$.
209: Calculating the density and tagged--particle correlation function
210: $\Phi(Q,t)$ from the dynamic structure factors
211: \begin{equation}
212: \label{corr-functions}
213: \Phi(Q,t)=S_{\rm coh}(Q,t)/S_{\rm coh}(Q) \quad \mbox{and} \quad
214: \Phi^s(Q,t)=S_{\rm inc}(Q,t).
215: \end{equation}
216: might then lead to a wrong normalization.
217: For incoherent scattering this problem can be circumvented by measuring
218: $S(Q,\omega)$ at low temperatures where the scattering is totally elastic
219: and $S(Q)=1$.
220: For coherent scattering this is not possible as $S(Q)$ strongly depends on
221: temperature and pressure.
222: Note, that the truncation effect depends on $Q$ and on the dynamics of the
223: studied system itself and that two competing effects are operative:
224: With decreasing $Q$ the accessible energy range gets narrower (see
225: figure~\ref{Q-of-w-Matte}), but at the
226: same time the quasielastic broadening and the inelastic intensity
227: decrease as well.
228: Though, while it is straightforward to measure a plateau or a decay
229: of the correlation function it is rather difficult to determine its
230: absolute value within a few percent.
231: The truncation effect can be studied using several upper cut--off
232: frequencies and is in general negligible at least for intermediate and
233: large wave numbers $Q$.
234:
235: %Multiple scattering effects? \\
236: The determination of the dynamic structure factor in absolute units
237: is further complicated by the inevitable presence of multiple scattering.
238: Multiple scattering is basically a convolution of the dynamic
239: structure factor $S(Q,\omega)$ with itself.
240: The correction for it is far from being a standard data treatment.
241: In the inelastic spectrum, most multiple scattering intensity
242: comes from elastic--inelastic scattering histories \cite{Sea75}.
243: This contribution is nearly isotropic and therefore weakest at large
244: scattering angles.
245: It tends to smear out the characteristic $Q$ dependence of the scattering,
246: and dominates at small angles where the single--scattering signal
247: is expected to vanish as $Q^2$.
248:
249: There are several different approaches to correct for this effect.
250: One can perform a Monte-Carlo simulation of the neutron histories
251: in an effective way \cite{WuKB93}.
252: In order to calculate the fraction of multiply scattered neutrons at one
253: point ($Q,\omega$) one has, in principle, to know the ideal scattering in
254: the whole ($Q,\omega$)--space in absolute units.
255: As can be seen in figure~\ref{Q-of-w-Matte} only part of it is experimentally
256: covered.
257: Therefore, on needs a physical model for the ideal scattering including
258: $Q$-- and temperature dependences.
259: This makes the correction even more complicated for coherent scattering
260: where elastic, quasielastic and inelastic scattering are oscillating functions
261: of $Q$.
262:
263: A new experimental, model independent, however very time consuming approach
264: for multiple scattering correction in inelastic neutron scattering consists
265: in the use of a set of different incident neutron wavelengths
266: in connection with Monte Carlo simulations \cite{RuML00}.
267: Combining data at different wavelengths one can reliably establish an ideal,
268: single scattering function in a wide $Q$ range which then serves as an input
269: to the simulation followed by a self--consistent, iterative correction.
270:
271: Another possibility at least for incoherent scattering
272: is to expand the scattering function
273: $S(Q,\omega)=A(\omega) + Q^2 B(\omega) + \ldots$
274: which enables us to estimate the multiple scattering $A(\omega)$
275: by $Q^2 \to 0$ extrapolation
276: from the $S(2\theta(Q,\omega),\omega)$ spectra
277: however, loosing the elastic channels \cite{ToZF00}.
278: In practice, one often uses thin samples with high transmissions of the
279: order of 90\% and more.
280: Then multiple scattering is in general neglected at least at high scattering
281: angles \cite{WuKB93}.
282:
283: Recently, an alternative approach has been worked out.
284: By extensive Monte Carlo simulations on very simple models
285: some generic trends of multiple scattering were identified \cite{Wut00a}:
286: it distorts the wave number dependence much more than the
287: frequency dependence of $S(Q,\omega)$.
288:
289: Neutron scattering is always a compromise between %of several needs among these
290: the desired statistics, the wave number resolution and coverage,
291: the elastic energy resolution and the accessible energy window.
292: Apart from notoriously lacking statistics, one major problem in neutron
293: scattering is that practically all high resolution spectrometers lack dynamic
294: range.
295: It is known that in order to measure the dynamics of supercooled liquids and
296: glasses properly one has to cover several decades in frequency or
297: time.
298: Several approaches (partly) overcome this problem:
299: (i) the use of master functions obtained by scaling procedures like the
300: time--temperature superposition and / or
301: (ii) the combination of data from different spectrometers with overlapping
302: time ranges.
303: Over the years we have been using both approaches.
304: Elastic resolution effects can be corrected in the $S(Q,t)$ representation
305: by division with the Fourier transformed resolution function (Fourier
306: deconvolution) that has been measured at very low temperatures either on
307: the sample itself or on a Vanadium standard.
308:
309: Neutron spin--echo spectrometers which measure directly in time domain have
310: a rather large dynamic range of about 3 decades and more, depending on
311: the scattering angle and the incoming wavelength \cite{Yel94}.
312: However, the concomitant shortcuts are low countrates and low $Q$ resolution
313: compared to multidetector instruments.
314: Recently, a multidetector (42 detectors) has been installed on
315: the NSE instrument IN11 covering about 40$^{\circ}$.
316: Now, by binning several detectors at least the statistical quality can be
317: improved.
318: The $Q$ resolution is still implicitly limited by the transmission band
319: of the velocity selector of
320: typically $\Delta \lambda/\lambda \simeq 10\,\%$ leading directly to
321: $\Delta Q/Q \gtrsim 10\,\%$.
322: One should be aware that in such a case one measures averages
323: $\langle \Phi(Q,t) \rangle_Q$ over correlation functions.
324: Increasing the $Q$ resolution by a narrower band width using for example
325: a graphite monochromator leads to an unacceptable reduction in intensity.
326:
327: Improvements of neutron sources \cite{FRM2,ESS} to increase the incoming flux
328: and of instrumental performance \cite{Milleniumprogramm} to increase the
329: efficiency are currently under way.
330: In order to test the new developments and extentions of the theory
331: (section~\ref{Extentions_of_MCT}) experimentally or to attack
332: even new questions the dynamic range
333: and concomitantly the $Q$--resolution and the $Q$--coverage
334: have to be enhanced.
335:
336: Let us demonstrate to which extent it is desirable to push
337: improvements for a real progress. % in glass science.
338: As an example we show in figure~\ref{Coh-inc} the tagged--particle and the
339: density dynamic structure factor of OTP measured in the only proper way
340: by polarization analysis on the instrument D7 at the ILL.
341: Currently, D7 is the only multidetector instrument in the world which is able
342: to separate the energy resolved incoherent and coherent scattering
343: from one sample.
344: Some triple--axis spectrometers can do polarization analysis
345: as well, however, are equipped with a single detector only.
346: On each other instrument one always measures both contributions which are
347: differently weighted depending on the spectrometer type.
348: % Figure %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
349: \input{Bilder/Exp/Coh-inc.tex}
350: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
351: This possibility is particularly interesting for some important questions
352: not only in glass science which have not been studied up to now.
353: For example, it would be possible to study exclusively the center--of--mass
354: (COM) motion of OTP.
355: Measuring the incoherent scattering of OTP-$d_{10}$ where only the lateral
356: phenyl rings are deuterated gives access to the tagged--particle correlation
357: function of the COM motion.
358: Conversely, measuring the coherent scattering of central ring deuterated
359: OTP-$d_{4}$ would lead to the density correlation function of
360: the COM motion.
361: At the same time, important information can be gained from the
362: coherent scattering and the incoherent
363: scattering of the lateral phenyl rings in OTP-$d_{10}$ and OTP-$d_{4}$,
364: respectively, about e.~g.\ relaxational modes and their time scales
365: or vibrational excitations of intramolecular motions.
366:
367: Furthermore, it would also be possible to compare both structural relaxation
368: times and their $Q$ dependence in the context of reccurent used models
369: like ``de Gennes'' narrowing \cite{Gen59}. % or ``Vinyard'' approximation.
370: The advantage is evident: such studies could be performed on one and the
371: same sample under exact identical experimental conditions.
372:
373: The spectra shown in figure~\ref{Coh-inc} were obtained under fairly
374: optimized experimental conditions.
375: Nevertheless, although the transmission of 67\% was extremely low and the
376: energy resolution rather moderate data accumulation took about 42\,h!
377: Such test experiments make it clear that, for example, increasing the
378: incoming flux by a factor of two while leaving the instruments
379: as they are is far too less.
380: The total efficiency which is a product of the source strength, the
381: transport efficiency and the instrument efficiency has rather to be
382: increased by orders of magnitudes \cite{Milleniumprogramm}.
383:
384: \subsection{Other Experimental Approaches}
385: \label{Other_Experimental_Approaches}
386: In section~\ref{GT} we have already presented many experimental techniques
387: probing the glass transition phenomena and the slow structural relaxation
388: of OTP.
389: Therefore, we limit ourselfs to a short enummeration of those other
390: experimental approaches that probe the fast dynamics in the GHz -- THz
391: band and have been applied to OTP.
392:
393: Light scattering spectroscopy has been often employed in OTP
394: and many other supercooled liquids and glasses.
395: Raman scattering has primarily been performed in connection
396: with the Boson peak and the study of the vibrational dynamics
397: of the glassy and polycrystalline state \cite{StZP94,CrBA94,KiVP99}.
398: Brillouin scattering explores the longitudinal (LA) and transverse (TA)
399: acoustic modes in the GHz range and their interaction with the
400: structural relaxation \cite{HiWa81,WaZS87,MoCF98b}.
401: Inelastic X-ray scattering extends the $Q$--$\omega$ range of the
402: conventional Brillouin technique and makes it possible to study
403: sound waves on intermolecular distances and in the THz range
404: \cite{MaMR98,MoMR98,MoFM99}.
405: Depolarised light scattering using geometries in which the usual
406: LA and TA Brillouin components are forbidden is the most often used
407: method to study the sub GHz to THz dynamics
408: \cite{CuHL98,StPM92,StMP93,StPG94,PaSM94,CuLD97}.
409: Although the scattering mechanism and the connection to density fluctuations
410: are still debated \cite{PaSN97} the experimental results have often
411: been compared to mode--coupling theory.
412: A promising development covering about 5 orders of magnitude in time
413: is actually taking place using the optical Kerr effect in terms of a
414: pump--and--probe experiment \cite{HiBG00a,HiBG00b}.
415:
416: %Dielectric loss spectroscopy is now able to measure susceptibility
417: %spectra $\chi^{\prime\prime}(\omega)$ up to 380\,GHz \cite{LuPD96,LuPL97}.
418: %However, due to the weak dipole moment of OTP it has rarely been applied to
419: %pure OTP \cite{JoGo70,NaEM87}.
420:
421: Computer simulation yield $\Phi(t)$ as primary output.
422: In recent Monte--Carlo \cite{Bas94} and molecular dynamics
423: \cite{KoAn94,GaST96} studies the time range
424: was extended to such long times, that they now deal with the same
425: windows as achieved by the above mentioned modern experiments.
426: In molecular dynamics simulations many different observables can be
427: studied which are not accessible by any other experimental methods.
428: The data quality is so good that the results can be compared
429: with more sophisticated mode--coupling models and with numerical
430: solutions of the mode--coupling equations,
431: where the static structure factor is the only input.
432: However, simplified models have to be used approximating the real
433: molecule \cite{MoLR00,LeWa93,LeWa94,KuWi95,WaLe97,ScTa99}.
434: The need for extending the simulations to times long enough to follow the
435: relaxation processes in the supercooled and glassy state still necessitates
436: a compromise with respect to the complexity of the model and
437: the number of interaction sites.
438: Lewis and Wahnstr\"om used a three--site model mimicking OTP,
439: where each phenyl ring was treated as a single interaction site
440: \cite{LeWa93,LeWa94,WaLe97}.
441: Although simple, this model essentially reproduces the dynamics of OTP,
442: in particular the anomalous behaviour of the Debye--Waller (\Mob--Lamb)
443: factor and the existence of a fast motional process.
444: Later, Scortino et al.\ extended the calculations of Lewis and Wahnstr\"om
445: by more than one decade towards longer times \cite{ScTa99}.
446: In addition to the translational correlation functions the rotational ones
447: were calculated and compared with theory.
448: However, the planar nature of model makes it easier for the molecules
449: to slide past one another.
450: The 18--site, three--ring model of Kudchadkar and Wiest \cite{KuWi95}
451: and later of Mossa \etal \cite{MoLR00} takes into account the
452: three--dimensional nature of the OTP molecule and
453: may provide a better representation of the reorientational dynamics.
454: In addition it retains some internal torsional motions of the lateral
455: rings with respect to the central one.
456: However, with the extended complexity only a few number of molecules
457: could be simulated and the time range is rather limited.
458:
459: %The microscopic vibrational dynamics of crystalline OTP
460: %was investigated by the use of lattice dynamics calculations by
461: %Criado et al.\ \cite{CrBG93,CrBA94} where the motions of
462: %several internal modes are considered in detail.
463: %using the crystal structure determined by neutron \cite{BrLe79}
464: %and x--ray scattering \cite{AiMa78}.
465: