HAL Id: jpa-00209153
https://hal.archives-ouvertes.fr/jpa-00209153
Submitted on 1 Jan 1979
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Light scattering and calorimetric study of the thermal diffusivity of glycerol, liquid and glass
C. Allain, P. Lallemand
To cite this version:
C. Allain, P. Lallemand. Light scattering and calorimetric study of the thermal diffusivity of glycerol, liquid and glass. Journal de Physique, 1979, 40 (7), pp.693-700. �10.1051/jphys:01979004007069300�.
�jpa-00209153�
Light scattering and calorimetric study of the thermal
diffusivity of glycerol, liquid and glass
C. Allain and P. Lallemand
Laboratoire de Spectroscopie Hertzienne de l’E.N.S., 24, rue Lhomond, 75231 Paris Cedex 05, France
(Reçu le 28 décembre 1978, accepté le 12 mars 1979)
Résumé.
2014On présente les résultats d’expériences de diffusion de la lumière en vue de déterminer la diffusivité
thermique de la glycérine dans diverses situations. Si le temps de relaxation interne, qui dépend de T, est long ou
court devant le temps de diffusion de la chaleur, qui dépend de q-2, on dira qu’on a un verre ou un liquide. On
décrit d’abord des mesures de diffusion spontanée de la lumière. On trouve une très faible dispersion du temps de relaxation de la raie Mountain, et une très faible intensité pour la composante Rayleigh dans le verre. On détermine
ensuite la dispersion du temps de relaxation Rayleigh à l’aide de la diffusion Rayleigh forcée. On présente les
résultats d’une expérience de calorimétrie. On compare ces divers résultats à des prédictions théoriques. On trouve
que le modèle viscoélastique généralisé permet d’obtenir une représentation cohérente des mesures acoustiques, thermiques et celles obtenues par diffusion de la lumière de part et d’autre de la transition liquide-verre dans la glycérine.
Abstract.
2014Light scattering experiments have been used to determine the thermal diffusivity of glycerol in a variety of situations, changing the temperature T and the wave-vector q. Depending on whether the internal relaxation time (depending on T) is very short or very long compared to the heat diffusion time (depending on q-2)
we say that we are studying either a liquid or a glass. Spontaneous light scattering experiments are described first.
They give essentially information concerning the dispersion of the Mountain relaxation time : the ratio of its values in the liquid and the glass. They also show that the intensity of the Rayleigh component is negligible in
the glass. Forced Rayleigh scattering experiments are then related. They allow a determination of the dispersion
of the Rayleigh relaxation time. Using a calorimetric technique, the dispersion of the thermal diffusivity is then
found to be the inverse of that of the Rayleigh relaxation time. These results concerning relaxation times and
amplitudes of the Rayleigh and Mountain components are then compared to theoretical models. It is shown that the generalized viscoelastic model gives a consistent description of acoustic, thermal and light scattering measu-
rements in glycerol.
Classification Physics Abstracts 62.60
-44.50
In this paper, we shall report experimental data
obtained by light scattering techniques concerning
the thermal properties of glycerol on either side of the glass transition, defined by the condition that the heat diffusion time is of the order of the structural relaxation time. The data, together with the results
of a calorimetric experiment, will then be analysed
in terms of phenomenological models that have been
discussed in detail in a paper referred to as 1 [1].
It was shown there, that much information can be obtained from the values of the decay rates and of the
relative amplitudes of the Rayleigh and Mountain components observed on either side of the glass
transition in spontaneous or forced Rayleigh light scattering experiments.
This transition, that we shall often call the Rayleigh-
Mountain coupling condition (by analogy with the
Brillouin-Mountain coupling which corresponds to
ultrasonic dispersion) takes place when the experi-
mental situation is such that the heat diffusion time TH is of the order of the internal relaxation time zo.
As Tu
=I/Kq2, where K is the thermal diffusivity and
q the wave-vector defined by the geometry of the experiment and as Io varies with the temperature proportionally to the viscosity [2], the coupling can
be observed by varying q at constant T or by varying T
at constant q. To indicate typical conditions for
glycerol, we mention that for a spatial wavelength (1
=2 n/q) of the order of 0.5 pm we have ru - ’to at 0 °C, and at - 60 °C for a wavelength of 100 pm.
The variety of conditions in which the coupling can
take place is a clear indication of the relaxational character of the liquid glass transition as revealed by changes in the thermal properties.
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:01979004007069300
694
We shall describe successively the various types of experiments which we have performed [3] : spon- taneous light scattering, forced Rayleigh scattering
and calorimetry. We shall give results obtained outside the coupling region, that is for ’rH > 10,ro and
’tH 10-1 To- Results obtained in the coupling region using the forced Rayleigh scattering (which yields the
best experimental data) will be presented elsewhere.
1. Light scattering experiments.
-A variety of spontaneous light scattering spectra have been obtained using different experimental methods. Some of this work has already been published, so in some
cases we shall just recall the main features of the results to give as complete a picture of glycerol as possible,
to be able to make a choice among the models pre- sented in I.
A first type of experiment [4] was performed using
a 90° scattering angle geometry and spectral analysis
of the scattered light with a Fabry-Perot. We recall
the ess,ential result that was obtained in this way.
We found that when lowering the temperature, the spectrum below - 10°C (down to - 80°C) exhibits only one very narrow spectral component (apart
from the Brillouin lines). We thus concluded that the
intensity of the Rayleigh line, in the glass state, is very
small, so that (IR/Im). - 0. Here, we use the same
notation as in I to indicate in which condition a
quantity has been determined. If the time scale of the
experiment is very long compared toro, as in a liquid,
we use the index 0, whereas if it is very short compared
to To as in a glass, we use the index oo.
In addition to this result, using the depolarized light scattering (of time constant ’tdep) as a way to determine accurately the internal relaxation time To, it was possible to determine [5] the time constant of the
Mountain line on either side of the coupling. Working
at fixed scattering angle (90°), and comparing the polarized and depolarized spectra for various values of the temperature, we determine L&/Ldep for T > 30 °C and ’rm- /.rdep for T O°C. Assuming that the reorien- tational motions of glycerol molecules are not coupled
to the long wavelength hydrodynamic excitations of the fluid (which is plausible as this fluid does not exhibit any structure in the depolarized spectrum such
as that attributed to the coupling between molecular
alignment and vorticity) we concluded from the experimental data that iM/iM
=1.0 ± 0.2.
As this important result was obtained quite indi- rectly and by changing the temperature of the fluid,
we have performed a more direct measurement at fixed temperature, but changing the wave-vector.
The set-up is shown in figure 1. It involves usual light scattering techniques, so that we shall give details only concerning the peculiarities of the set-up. As
we want to make measurements at fixed temperature and different wave-vectors, it was found convenient to make recordings simultaneously for a large q (900 scattering) and a small g (small scattering angle 0).
Fig. 1.
-Schematic diagram of the experimental set-up used to record the correlation function of the spontaneously scattered light simultaneously for near forward and 90° scattering.
This allowed us to exclude any difficulty due to possible temperature drift and to work on a single sample, so avoiding any change in the properties of glycerol from sample to sample due to variation of the water content.
The temperature was chosen (T ~- 40°C) such that,ro is not too small due to limitations in the speed
of the correlators that were available and to signal/
noise considerations that were quite critical, as glycerol does not scatter much light. The heat diffusion time TH for 900 scattering was short enough (- 60 ns)
to satisfy the condition r, » TH. Now, to satisfy
the other condition Io « TH for small angle scattering,
we just needed to take a small angle 0, as
(n being the index of refraction of glycerol and kL
the wave-vector of the input laser beam). In practice 0
could not be taken too small due to parasitic light
scattered by the optics (essentially by the windows of the cryostat). As we did not use an elaborate set-up such that described by Lastovska [6], and as our laser intensity was not well stabilized, we had to use 0 - 10.
This is because we had to keep the amount of elastically
scattered light to a small enough level so that the component of the photoelectric current due to the beating between scattered and laser light was larger
than that due to the intensity noise of the laser. In
addition, it tums out that it is quite difficult to prepare
a dust and bubble free glycerol sample, so that it
would have been out of reach to prepare a very large sample compatible with a wide beam experiment such
as that of Lastovska.
Very significant improvements were obtained through the use of very thick optical windows for the cryostat, schematically shown in figure 2. The glass cylinders were chosen to have an index of refraction close enough to that of glycerol so that there was
almost no stray scattering at the glass-glycerol inter-
faces. We verified by making measurements of the
Mountain line scattered at 900 at various places in the
cell, that the heat leak due to the windows did not
Fig. 2.
-Details of the cryostat used for near forward spontaneous light scattering experiments.
produce severe temperature gradients inside the cell.
(We recall that around - 40 °C a 5 °C change leads
to a variation of the intemal relaxation time by a
factor 10.) With suitable temperature control devices,
we were thus able to perform near forward scattering experiments at temperatures of the order of - 40 OC, fluctuating by less than ± 10-2 °C over several hours.
For 90° scattering, we used the same equipment as previously [2]. Our first problem was to find out
whether these experiments were disturbed by stray elastically scattered light. We thus performed a series
of experiments on the Mountain line at lower tempe-
ratures so that both for near forward (10) and 90°
scattering, the heat diffusion time Tu was much shorter than To. By careful analysis of the data, we established with great confidence that the spectra recorded at 90°
were of the homodyne type, as we verified that the corresponding correlation function
was proportional to the square of {G1o(t) - G1o(oo)}
which was of the heterodyne type.
We also looked for the best conditions to obtain
a heterodyne detection in the forward direction.
In particular we made a theoretical analysis of the
influence of a finite amount of homodyne signal in
addition to the heterodyne signal. This problem is
well treated in the case of a simple exponentially decaying correlation function [7]. However, here,
our correlation functions are by no means exponential,
due to the existence of a very broad distribution of relaxation times. We thus considered the following
correlation function :
where gl(t) and g2(t) are the normalized first and second order correlation functions of the scattered
light, and ILO and Is > the mean intensities of the local oscillator and scattered light respectively. Taking
for gl(t) and g2(t) expressions calculated from a Cole- Davidson distribution of relaxation times, with 03B2
=0.5, we found that we have to take ILO/ Is > 20
to get a determination of TcD with an accuracy better
than 1 %, Tcd being the longest time involved in the Cole-Davidson distribution. We also looked for the influence of the small amount of depolarized light
which is always present and that was not considered when analysing previous experiments [2]. In fact it
seems that due to the small depolarization ratio of Rayleigh scattering in glycerol (~ 0.1), the presence of depolarized light could only lead to a very small
error ( 1 %) in the determination of the Cole- Davidson parameter fi.
We then recorded a series of correlation functions with the following conditions T
= -40°C, 0
=1 °.
As glycerol exhibits a very broad distribution of relaxation times, it was necessary to determine the
experimental correlation function over a very wide time domain. This was done by combining several recordings, as our correlators give information for
only 100 equally spaced values of the time.
We show in figure 3 a correlation function recorded at 90° with a clipped digital correlator. It was analysed
as by Allain-Demoulin et al. [2]. Setting p
=0.4 we
obtained TS)
=12.7 us, where we put the index o0
Fig. 3.
-Homodyne correlation function of the light scattered by glycerol at - 40 oC, 0
=900 and fit to a Mountain line using a
Cole-Davidson function.
as Th~ 60 ns LM. Note that in the case of a Cole- Davidson distribution of relaxation times, the mean relaxation time, that we can call Tm, is equal to 03B2TCD,
if TCD is the longest relaxation time of the distribution.
We show in figure 4 a correlation function recorded at 0
=1.09° for the same sample and the same tem- perature as for figure 3. To analyse this recording,
we tried to fit it to
where f(T) is a Cole-Davidson distribution with
03B2
=0.4 and 1:R the Rayleigh relaxation time. To increase the accuracy in the determination of 1:CD, we
set the value of TR from known values of the thermal
diffusivity and of the wave-vector. The thermal
diffusivity was derived by extrapolating the results
696
Fig. 4.
-Heterodyne correlation function of the light scattered by glycerol at - 40 °C ; 0
=1.09° and fit to a sum of a Rayleigh
line and a Mountain line using a Cole-Davidson function.
of a series of measurements performed at various temperatures (20 OC, 0 OC, - 25 oC, - 35 OC) using
the same set-up. For these temperatures ’t’cD is so short that the signal observed involves only the Rayleigh
line. For the conditions in which figure 4 was recorded,
ÏR
=133 us. We could thus adjust TCD 0 using a least
square fit procedure. We found ’t’go
=11.3 us and
(IR/IM)o
=0.26 for the ratio of the intensities. As in these experiments Tp/ïM is significantly larger than 10,
we estimated that the data had been collected beyond
the coupling region, so that we could use the index 0 to
indicate the conditions in which these results were
obtained.
A series of such experiments were performed, to
determine TCÎ/TCOD for the same sample, at the same temperature (so that we do not need to know the
exact value of the temperature). The average value of this ratio, which we set to be equal to the ratio of the
Mountain times (as fi is the same for the two measu- rements) leads to
The rather large error bar comes from the dispersion
of the measurements and takes into account possible
uncertainties due to non perfect homodyne or hete- rodyne detection.
As discussed in I, some models lead to a value close to 2 for the ratio ’tM /7:&, so that we also show in figure 4
the best fit obtained by setting TO - zcDI2
=5.7 ils.
The corresponding curve represents poorly the experi-
mental correlation function. It is thus obvious that such a value of ’tM/’tMo. is unacceptable. This work thus confirms the results indicated above conceming the
small difference of the Mountain relaxation times on
either side of the Rayleigh-Mountain coupling. The implications of this result will be discussed later.
We also made several recordings of the experimental
correlation function at small scattering angles, for
lower temperatures. All the recordings could be fitted very well using just a Mountain line. This means
that the intensity of the Rayleigh component becomes
very small in the glass state. This confirms the result that we had obtained using a Fabry-Perot interfero-
meter :
Apart from the small value of this intensity ratio, these measurements showed that spontaneous light scattering does not seem to be applicable to determine
the thermal diffusivity of glycerol in the glass state.
To overcome this difficulty we have performed forced Rayleigh scattering experiments which allowed us to
obtain complementary results.
2. Forced Rayleigh scattering.
-The experimental set-up is similar to that used in the early work on forced Rayleigh scattering [8]. It involves two lasers operated
at different wavelengths Âp and Âs, such that one beam
can be absorbed to heat-up the medium, whereas
the other one, used to monitor index changes, is such
that the medium is transparent at Âs. We used an argon ion laser, whose output beam is mechanically chopped
to give pulses as short as 20 ps, with a duty cycle of 1/2 000. It is then split into two beams that interfere
inside the sample. A real image of the interference pattern can be made on a screen to determine the
fringe spacing 1
=2 n/q. In order to get sufficient heating of the glycerol, we added a small quantity
of iodine such that the absorption coefficient at the argon ion wavelength was of the order of 1 cm-1.
We shall show later that this does not affect signifi- cantly the thermal properties of glycerol. The resulting phase grating is detected by using a He-Ne laser whose output is diffracted by the grating. The amplitude of
the diffracted light was small enough so that the
detection is of the heterodyne type. The local oscillator is provided by stray light elastically scattered by the
windows of the cryostat, or by some dust particles
inside the sample. It is worth mentioning that the temperature fluctuations induced by the heating beam,
of the order of 10 - 3 to 10 - 2 °C, are very large compar- ed to those due to thermal agitation. As a result the signals are quite large so that there is no need to be as
careful as for spontaneous light scattering experiments
to get rid of bubbles or dust particles. In addition, glycerol being very viscous at low temperature, there is hardly any motion of the defects of the sample so
that there is no fluctuation in the relative phase of the
diffracted light and local oscillator, provided the device
used to generate the interference pattern at the blue frequency is stable. One can then sum in a multi- channel analyser the signal over many successive
cycles, usually a few hundred to a few thousand.
The MCA that we used includes 1 024 channels, with
a smallest time increment of 10 gs. In some cases it was operated with a variable clock, so that we could
record the early parts of the heat-decay signal with
more detail than its later parts. The output S(t) of the
MCA was recorded on paper tape and with a strip-
chart recorder either as S(t) or as Log (S(t) - A ),
where A was adjusted manually.
For each temperature we made a series of measu-
rements for various values of the interfringe 1. We
show in figure 5 the results of such measurements far from the coupling region, plotting Ti/2 vs. l. When using the largest values of 1, the number of fringes in
the heating beam was quite small. To decide whether this could lead to difficulties [9], we made a theoretical
analysis of the possible errors from this. Taking a sinusoidally modulated Gaussian beam, we found
that provided there were more than 5 fringes within
the waist of the beam, the distortions could be
neglected in view of the usual signal to noise ratio accessible in our experiments.
We see in figure 5 that we can determine the thermal
diffusivity with very good accuracy. The value of 1.03 x 10-3 cm2/s is in good agreement with that which we determined by spontaneous light scattering, using either a Fabry-Perot interferometer or light beating techniques.
Fig. 5.
-Square root of the Rayleigh relaxation time, vs. interfringe spacing obtained by forced Rayleigh scattering : + at 20 °C ; 0 at - 74 oC ; D derived from Fabry-Pérot analysis at 20°C and A
derived from near forward spontaneous scattering at 20 OC.
We show in figure 6 the determination of the relaxation time of the Rayleigh component for various temperatures. The important point to note is that
away from the coupling region to be discussed later, the limiting values of high and low temperatures are quite different. An analysis of the data leads to
We obtain a variation of the Rayleigh relaxation time that is much larger than that of Mountain relaxation time.
In the coupling region, it is no longer possible to separate S(t) into a Mountain and a Rayleigh line.
However, it turns out that the long time behaviour of the signal can be fitted rather well to an exponential,
so that it is still possible to define a time constant for
all values of the temperature. The corresponding
Fig. 6.
-Relaxation time of the Rayleigh line measured by forced Rayleigh scattering. 0 : with a 70 J.1m interfringe ; + : with a 120 gm interfringe ; A : value derived from the calorimetric experiment.
values are plotted in figure 6, they correspond to a sharp increase in the coupling region. To get more information out of the data, it is necessary to compare the detailed shape of the relaxation function with that calculated using the models developed in I.
This will be presented elsewhere.
In addition to the results conceming the time cons-
tant of the Rayleigh component, we have used the signal S(t) to determine the ratio of the amplitudes of
the Mountain and Rayleigh components at high temperature such that Tm - 0.1 TR. Again, due to
limitations in the time scales of the MCA, we had to take into account the finite duration of the heating part of the experimental cycle. As was indicated in our
previous work on the subject [10], a uniform heating during time à leads to the following change in the signal. If the undistorted signal is
the measured quantity will be :
This leads to significant distortions if the Oi are not large compared to ô. This occurs in our experiments
as the distribution function of relaxation times of
glycerol includes very short relaxation times. It will therefore be somewhat difficult to derive precise
values for the ratio of the amplitudes of the Mountain
and Rayleigh contributions. Using the generalized
viscoelastic model with 5 relaxation times, as discussed in I, we find that if the mean relaxation time is 1 ms
(near - 60 OC), taking a heating time ô
=20 ps, the signal just at the end of the heating pulse (for
t
=ô) is only 5 % smaller than if the same amount of heat had been dissipated in the fluid in a very
short time. The experimental situation is probably
less favorable as we should take into account the
698
jitter of the electronics, so that the first measured
point may very well be taken 30 ps after the beginning
of the heating pulse, due to the fact that our MCA records one point only every 10 us. In that case the correction will be as much as 10 %. We can therefore only get an upper limit for the ratio of the amplitudes
of the Mountain and Rayleigh contributions, and
estimate that (AmIAR)’ - 0.3. We thus get a
negative ratio, but it is not accurate enough to allow
us to determine the ratio Co2 /C2, as we showed in 1
that (AM/AR)°
=1
-C20/C2 both for the purely
structural and the generalized viscoelasticity models.
We may remark that in our experiments, it is in principle straightforward to take into account the finite duration of the heating pulse, due to the fact
that we consider only linear differential equations for
the evolution of the fluid. This is a trivial example
of the influence of the past history of a glass upon its
properties, which is very important in usual experi-
ments on glasses when the departures from equilibrium
are large so that it is necessary to consider nonlinear differential equations.
We now summarize the results obtained on either side of the coupling :
- in spontaneous scattering :
for transition temperatures of
+ 200 (polarized/depolarized scattering)
-
45 OC (spontaneous scattering)
-