1: \section{Glass Transition in OTP} \label{GT}
2: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3: \subsection{Macroscopic Quantities}
4: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5: The calorimetric techniques used for measuring the heat capacity $c_{\rm P}$
6: of glasses are the same as used for other solid materials.
7: In particular, differential scanning calorimetry (DSC) \cite{ChBe72,GrTu67b}
8: or differential thermal analysis (DTA) \cite{AtAn79} are used to determine
9: the glass transition temperature \Tg .
10: When the heat capacity $c_{\rm P}(T)$ of a supercooled liquid is measured as
11: a function of temperature (or pressure) one notices
12: a significant change occurring at a temperature \Tg\
13: (figure~\ref{cp-GrTu67}).
14: % Figure %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
15: \input{Bilder/GT/cp-GrTu67.tex}
16: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
17: % Figure %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
18: \input{Bilder/GT/v_T_P-NaKo89.tex}
19: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
20: A similar effect is observed in volume--temperature--pressure measurements
21: \cite{NaKo89}, reproduced in figure~\ref{v_T_P-NaKo89}.
22: A "discontinuity" in the thermal expansion coefficient
23: $\alpha=\partial \ln V/\partial T|_{\rm P}$,
24: the derivative of the volume $V$ at constant pressure $P$
25: can be deduced from figure~\ref{v_T_P-NaKo89}
26: for the glass while nothing particular happens in the crystal.
27: Although these features are characteristic of a second--order
28: equilibrium phase transition, \Tg\ cannot be related unambiguously
29: to an equilibrium function, such as the free energy.
30: In particular, the position and the sharpness of the transition near
31: \Tg\ {\it depends on the cooling rate} at which for example
32: the heat capacity is monitored.
33: In figure~\ref{cp-GrTu67} results with
34: two different cooling rates \cite{GrTu67b,ChBe72} are compared.
35: Conventionally, cooling rates of about 10\,K/min are used to determine \Tg .
36: Operationally, \Tg\ is defined as the center of the temperature range
37: of the large change in the heat capacity in a DSC run with cooling
38: rate $\dot{T}$.
39: Note, that in this definition the glass transition is related with
40: a low--frequency excitation: e.\ g.\ for a typical rate of
41: $\dot{T}=10$\,K/min $\simeq 0.2$\,K/s and a width of the $c_{\rm P}$--jump
42: $\Delta T=10$\,K the excitation frequency is
43: $\omega=\dot{T}/\Delta T=0.02$\,s$^{-1}$, thus probing correlation
44: times of about 50\,s.
45: The glass transition temperature defined in this manner
46: can be understood as an isodynamic point
47: which allows for a comparison of different glasses.
48:
49: When pressure is applied the transition point shifts to higher temperatures
50: as can be seen in the volume--temperature--pressure relation in
51: figure~\ref{v_T_P-NaKo89}.
52: Note, that under pressure OTP shows a very strong density change.
53: Calorimetrically, the glass transition temperature in OTP as a function
54: of pressure, $T_{\rm g}(P)$, has been measured by DTA by Atake and
55: Angell \cite{AtAn79}.
56: Up to at least about 100\,MPa the variation of $T_{\rm g}(P)$ is linear
57: with a slope ${\rm d}T_{\rm g}/{\rm d}P \simeq 0.26$\,K/MPa (260\,K/GPa).
58: These values for OTP and its isomeric mixtures are the highest yet observed
59: for molecular liquids (though some polymers exhibit higher values).
60: It was suspected that ${\rm d}T_{\rm g}/{\rm d}P$ is larger the more
61: closely the molecular substance behaves repulsively like a hard sphere.
62: These facts have been further motivations for us to use OTP as a model
63: system also for pressure dependent neutron scattering experiments
64: (section~\ref{Experiments_under_Pressure}).
65:
66: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
67: \subsection{Primary $\alpha$--Relaxation: Stretching and Scaling}
68: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
69: The fluctuation--dissipation theorem connects a linear response
70: with the time evolution or spectral distribution of spontaneous
71: fluctuations.
72: Thus, relaxation in the equilibrium state can be studied
73: by a variety of spectroscopic or scattering methods.
74:
75: Time--resolved measurements are performed for instance by
76: photon--correlation spectroscopy (PCS) which measures ultimately
77: the evolution of density correlations $\Phi(t)$ \cite{FyWL81,FyDW83,HwSh99}.
78: The PCS data in figure~\ref{pcs_t-FyWL81}
79: illustrate nicely the most prominent features of structural relaxation:
80: stretching, scaling and a strong temperature and pressure
81: dependence of the mean relaxation time.
82: % Figure %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
83: \input{Bilder/GT/pcs_t-FyDW83.tex}
84: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
85:
86: The decay of the time correlations $\Phi(t)$ is called {\it stretched}
87: since it extends over a much wider time range than a simple exponential
88: process.
89: An exponential decay is expected when the dynamics consists of
90: uncorrelated random steps.
91: This is the case for tagged--particle diffusion like motion in
92: the long time and long distance limit \cite{WuCR96}.
93: On microscopic scales, however, the motion of molecules is a
94: highly collective process marked by memory effects.
95:
96: In many cases the structural relaxation function $\Phi(t)$ is
97: well described by a Kohlrausch law \cite{Ko1854,Ko1863},
98: obtained by replacing $t/\tau$ in the Debye law by $(t/\tau)^{\beta}$:
99: \begin{equation}
100: \label{KWW}
101: \Phi(t)=\exp(-(t/\tau)^{\beta}), \quad 0<\beta<1.
102: \end{equation}
103: For this function, the mean relaxation time $\langle \tau \rangle$
104: can be calculated as
105: \begin{equation}
106: \langle \tau \rangle = \int_{0}^{\infty} {\rm d}t \Phi(t)
107: =\tau\beta^{-1}\Gamma(\beta^{-1})
108: \end{equation}
109: where $\Gamma()$ is the gamma function.
110: The exponent $\beta$ is {\it not} (!) a material constant;
111: it differs for different correlation functions or observables which
112: are probed by different experimental techniques \cite{CuHL98} (see also
113: figure~\ref{gt-beta}).
114: Such variations of $\beta$ demonstrate that an understanding of
115: structural relaxation will not be possible without accounting for the
116: different ways in which experimental probes couple to the microscopic
117: degrees of freedom.
118:
119: On cooling or compressing a viscous liquid, $\tm$ increases
120: rapidly as can be seen in figures~\ref{pcs_t-FyWL81} and \ref{die_w-WiHa71}
121: % and \ref{tau_T_P-FyDo83},
122: while the line shape of $\Phi(t)$, expressed through $\beta$
123: varies only weakly.
124: This {\it scaling} property, known as the {\it time--temperature--pressure
125: superposition}, implies that in a first approximation the
126: measured individual correlations $\Phi(t;T,P)$ can be collapsed
127: onto a master curve
128: \begin{equation}
129: \label{mastercurve}
130: \Phi(t;T,P)=\Phi(t/\tau(T,P))
131: \end{equation}
132: by rescaling the time axis to $\hat t = t/\tau(T,P)$.
133: This property turns out to be particularly useful in the analysis
134: of neutron scattering experiments, to be
135: discussed later in sections~\ref{t-arx} and \ref{p-arx}.
136:
137: In many spectroscopic experiments, the response of a viscous liquid
138: is probed as a function of frequency.
139: A complex susceptibility $\chi(\omega)$ is related to the relaxation
140: function $\Phi(t)$ through
141: \begin{equation}
142: \chi(\omega)= \chi_{\infty} +
143: (\chi_{\infty} - \chi_0)
144: \int_{0}^{\infty} {\rm d}t e^{i\omega t} \dot{\Phi}(t).
145: \end{equation}
146: For high frequencies, the system reacts as a solid
147: $\chi(\omega \gg \tau^{-1})=\chi_{\infty}$;
148: for very low frequencies the fluid--like response is given by
149: $\chi(\omega \ll \tau^{-1})=\chi_0$.
150:
151: For intermediate frequencies $\omega \simeq \tau^{-1}$ the Kramers--Kronig
152: relation relates the step in the real part $\chi^{\prime}(\omega)$ to
153: the maximum in the dissipative part $\chi^{\prime\prime}(\omega)$.
154: An exponential decay of $\Phi(t)$ leads to
155: %the simplest interpolation between $\chi_0$ and $\chi_{\infty}$
156: \begin{equation}
157: \label{chi-Debye}
158: \chi(\omega)= \chi_{\infty} +
159: \frac{\chi_{\infty} - \chi_0}{1-i\omega\tau},
160: \end{equation}
161: which in the context of structural relaxation is often named
162: after Debye, who derived it for Brownian motion of uncorrelated dipoles
163: in a viscous medium \cite{Deb29}.
164:
165: For a stretched relaxation function, the step in $\chi^{\prime}(\omega)$
166: and the peak in $\chi^{\prime\prime}(\omega)$ are much broader than
167: a Debye peak.
168: To describe $\chi(\omega)$ one can either decorate (\ref{chi-Debye})
169: with one or two exponents \cite{CoCo41,DaCo51,HaNe67} or
170: use the Fourier transform of (\ref{KWW}) \cite{WiWa70}.
171:
172: Frequency dependent dielectric susceptibilities for pure OTP have rarely
173: been studied \cite{JoGo70,NaEM87,WuNa92,HaRi97,HaSB97}
174: probably due to the extremely weak dipole moment.
175: Instead, in many studies the reorientation of polar host molecules is measured
176: \cite{WiHa71,WiHa72,ShWi73,DaHW73,NaTH74,NaKT74,NaMa83,DiWN90}.
177: Typical temperature dependent dielectric spectra and the corresponding
178: master curve of 4.2 weight-\% anthrone in OTP are shown in
179: figure~\ref{die_w-WiHa71}.
180: For comparison, a Debye peak is also shown.
181: Molecular reorientation of different probe molecules has also been studied by
182: optical \cite{HyEE90,CiBE95}, ESR \cite{AnSG97,EaMP97}
183: and NMR spectroscopy \cite{RoBT91}.
184: % Figure %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
185: \input{Bilder/GT/die_w-WiHa71.tex}
186: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
187:
188: % Figure %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
189: \input{Bilder/GT/gt-beta.tex}
190: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
191: In figure~\ref{gt-beta} some published values of the Kohlrausch stretching
192: exponent $\beta$ are compared.
193: For a given experimental probe, $\beta$ increases systematically with
194: temperature in this temperature range.
195: Thus, the time--temperature superposition (\ref{mastercurve}) only
196: holds in a first approximation and in a limited temperature range.
197: In addition, for any given temperature, different experiments see different
198: exponents $\beta$ although approximate
199: values of $\beta$ seem to be correlated with other characteristics
200: of the glass transition dynamics \cite{BoNA93,Nga98}.
201:
202: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
203: \subsection{Temperature and Pressure Dependence of the
204: $\alpha$--Relaxation Time}
205: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
206: We have seen that the characteristic time (or frequency) of structural
207: relaxation strongly depends on temperature and pressure.
208: Cooling OTP by only 5\,K can slow down the decay of the correlation function
209: by a factor of ten or more, as shown in figures~\ref{pcs_t-FyWL81} and
210: \ref{die_w-WiHa71}.
211:
212: Figure~\ref{tau_T_P-FyDo83} demonstrates that over many decades
213: the characteristic time $\tau$ observed by different spectroscopic
214: techniques follows the same temperature and pressure dependence,
215: although at a given temperature or pressure the absolute values of
216: $\tau$ may differ.
217: This finding is confirmed by measurements on many other glass formers.
218: % Figure %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
219: \input{Bilder/GT/tau_T_P-FyDo83.tex}
220: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
221:
222: The famous Maxwell relation $\eta = G_{\infty} \tm$
223: connects the shear viscosity $\eta$ with the relaxation time over
224: the shear modulus at infinite frequency $G_{\infty}$.
225: Under the assumption that $G_{\infty}$ varies only weakly with external
226: control parameters like temperature and pressure,
227: it follows $\eta \propto \tm$.
228: % as confirmed in figure~\ref{tau_T_P-FyDo83} or figure~\ref{tau-otp}.
229:
230: Therefore, another operational definition of the glass transition
231: temperature \Tg\ is given via the temperature, where the viscosity $\eta$
232: reaches $\eta(T_{\rm g}) = 10^{12}$\,Pa$\cdot$s.
233: With shear moduli of solids, typically of the order
234: $G_{\infty} \simeq 10$\,GPa, this corresponds to a relaxation time of
235: the order of 100\,s.
236: This is the same time scale on that structural relaxation falls out of
237: equilibrium in a caloric experiment (see above).
238:
239: The Maxwell relation makes it possible to determine the temperature or
240: pressure dependence of the structural relaxation time beyond the frequency
241: range of any spectroscopic technique.
242: In fact, the viscosity of OTP has been measured over more than 14 decades
243: as a function of temperature and pressure
244: \cite{McUb57,GrTu67a,LaUh72,CuLU73,SchKB98}.
245: The temperature dependent data at ambient pressure are shown in
246: figure~\ref{Angell-plot} together with viscosities of other glass formers
247: in a $\log\eta\ vs.\ T_{\rm g}/T$ representation (Angell plot) motivated by the
248: na$\ddot{\rm i}$ve expectation of thermally activated transport
249: \cite{Ang91}.
250: Although never observed in any real glass former the Arrhenius law
251: $\eta \propto \exp(-E_A/T)$ is taken as a reference to classify the
252: temperature dependence of $\eta(T,P)$.
253: % Figure %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
254: \input{Bilder/GT/Angell-plot.tex}
255: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
256:
257: This classification according to Angell is physical since the curvature in
258: the Arrhenius plot is correlated with the structure of a glass former:
259: {\it strong} liquids, which deviate only slightly from the Arrhenius law are
260: typically networks like SiO$_2$ or GeO$_2$ where the transport is
261: controlled by thermally activated breaking of covalent bonds.
262: Largest deviations are found for molecular liquids, ionic melts and some
263: polymers.
264: They are termed {\it fragile} liquids.
265: In this classification OTP is one of the most fragile liquids, which is
266: another motivation for its recurrent use as a model system.
267:
268: From free-volume or entropy arguments \cite{CoTu59,CoGr79}
269: it is possible to derive the popular Vogel--Fulcher--Tamman fit formula
270: \begin{equation}
271: \label{VFT}
272: \eta(T) \propto \exp \left( \frac{A}{T-T_0} \right)
273: \end{equation}
274: with some arbitrary reference temperature $T_0 \ll T_{\rm g}$.
275: % Figure %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
276: \input{Bilder/GT/vis-otp-T-MCT.tex}
277: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
278: In any case, (\ref{VFT}) and similar functions are not able to fit $\eta(T)$
279: data over the entire temperature range, but they work much better when
280: applied to a restricted range.
281: A critical review of these and other approaches can be found in
282: Ref.~\cite{CuLH97}.
283: The differential analysis of $\eta(T)$ in the inset of
284: figure~\ref{vis-otp-T-MCT} is one way to show that the temperature dependence
285: of $\eta(T)$ cannot be described by a single physical model
286: over the whole temperature range.
287: In particular, one recognizes a change of regime at about 290\,K.
288: Within mode--coupling theory this is interpreted as a cross--over from
289: liquid--like to hopping dynamics (section~\ref{Mode_Coupling_Theory}).
290: For OTP the value $T_{\rm c} \simeq 290$\,K was first determined by strict
291: elastic neutron scattering \cite{BaFK89}
292: (section~\ref{t-squarerootsingularity})
293: and in the meantime it reappeared
294: in many other measurements and simulations,
295: as will be shown throughout this review.
296: In its simplest version, mode-coupling theory predicts for $T>T_{\rm c}$
297: a power law divergence
298: \begin{equation}
299: \eta \propto (T-T_{\rm c})^{\gamma}, \quad 1.76 < \gamma < 4.83.
300: \end{equation}
301: Figure~\ref{vis-otp-T-MCT} shows a three parameter fit of $\eta(T)/T$
302: yielding a critical temperature $T_{\rm c}=290$\,K and an exponent
303: $\gamma=2.59$. %2.586
304:
305: \subsection{Diffusion and Decoupling of Time Scales}
306: It is interesting to compare the translational and rotational diffusion
307: coefficients, $D_t$ and $D_r$, which are integrals over single particle
308: correlation functions, with the shear viscosity $\eta$.
309: The shear viscosity depends upon intermolecular momentum transfer
310: through cross correlation functions of neighbouring molecules.
311: In simple \vdW\ liquids, the validity of the Stokes--Einstein and the Debye
312: relations,
313: \begin{equation}
314: \label{Stokes-Einstein-Debye}
315: D_t=\frac{k_{\rm B}T}{6\pi\eta r_S}, \quad
316: D_r=\frac{k_{\rm B}T}{8\pi\eta r_D^3},
317: \end{equation}
318: originally derived for a Brownian particle in a viscous liquid, is well
319: established in the fluid regime, although it is a non--trivial fact.
320: The lengths $r_S=0.21$\,nm \cite{McDF69} $\approx r_D=0.23$\,nm
321: \cite{DrFK88,SchFS92} are temperature independent
322: and smaller than the geometrical \vdW\ radius $r_W=0.37$\,nm \cite{Bon64}.
323: It is most remarkable that $D_t$ and $D_r$ are proportional to $\eta^{-1}$
324: in the whole fluid regime.
325:
326: Rotational self diffusion in neat OTP has been measured by $^2$H--NMR
327: stimulated echos \cite{Gei94,GeFS98}, translational self diffusion by
328: $^1$H-NMR static gradient stimulated echos \cite{Gei94,FuGS92}.
329: Towards lower temperatures, tracer diffusion coefficients of two photochromic
330: tracer molecules having a size comparable to OTP and dissolved in OTP at
331: low concentrations were measured by forced Rayleigh scattering \cite{LoSi91}.
332: Results are shown in figure~\ref{Decoupling}.
333: % Figure %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
334: \input{Bilder/GT/Decoupling.tex}
335: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
336: Rotational self diffusion could be measured almost down to $T_{\rm g}$
337: where Debyes' relation (cf.~\ref{Stokes-Einstein-Debye}) still holds.
338: Apparently, momentum transport is closely connected with molecular
339: reorientations.r
340: Translational diffusion, however, {\it decouples} from rotational
341: diffusion and viscosity and crosses over around
342: $T_{\rm c}=290\pm10$\,K to a weaker temperature dependence
343: approximately given by $D_t \propto \eta^{-0.75}$.
344: Thus, $D_t$ and $\eta$ behave differently as the mechanism of molecular
345: motion changes from fluid to glass like behaviour.
346: On the other hand, $D_r$ remains proportional to $\eta^{-1}$ at temperatures
347: down to $T_{\rm g}$ indicating that rotational motion remains more
348: efficiently coupled to structural relaxation on approaching the glass
349: transition, at least, in fragile glass forming liquids like OTP,
350: toluene \cite{HiSi96}, salol, glycerin and others \cite{ChSi97}.
351:
352: Indications for a cross--over between two distinct dynamic regimes
353: at a temperature $T_{\rm c}$ above the glass transition temperature
354: $T_{\rm g}$ have been found for many simple supercooled liquids
355: \cite{Ros90}.
356: A discontinuity for the temperature coefficient of $\tau/\eta$ or
357: $(D\eta)^{-1}$ demonstrates the failure of the Stokes--Einstein relations
358: for $T < T_{\rm c}$.
359: The existence of the cross--over temperature $T_{\rm c}$ is further
360: substantiated by a plot of the correlation times on a reduced
361: temperature scale $T_{\rm g}/T$, which then shows a bifurcation of the
362: two relevant $\alpha$ and $\beta$ processes for $T < T_{\rm c}$.
363: