Rayleigh scattering of light in glasses

E. E. Gorodetskii, A. M. Evtyushenkov, Yu. F. Kiyachenko, A. V. Kryukov, and

S. M. Rytov
All-Union ScientiJc-Research Institute of Physicotechnical and Radio Engineering Measurements,
Menedeleevo, Moscow Province
A. L. Mints Radio Engineering Institute, Academy of Sciences of the USSR, Moscow
(Submitted 18 June 1986)
Zh. Eksp. Teor. Fiz. 92, 1401-1413 (April 1987)
A comprehensive experimental study of the scattering of light in fused quartz showed that the
relaxation theory of the Rayleigh scattering developed for equilibrium systems is inapplicable
to glasses. This is because the equilibrium theory ignores the unlimited rise of the relaxation
time as a glass forms from the melt. A modified theory allowing for this aspect is developed
and it predicts a qualitative difference between the nature of the scattering in glasses from that
in equilibrium solids and liquids, and provides a quantitative description of all the
experimental results obtained by the authors.

In 1970 one of the present authors1developed a general eters, some of which are symmetric zero-trace tensors. The
relaxation theory of the Rayleigh scattering of light in iso- introduction of such tensor parameters is necessary because
tropic bodies, relating the Rayleigh scattering coefficient to they are responsible for the depolarized scattering, which
equilibrium thermodynamic properties of matter. The sub- always occurs in a liquid. On the other hand, relaxation of
sequent great improvement in the precision of the absolute these parameters gives rise to a shear modulus which distin-
measurements of the scattering coefficient2.' has made it guishes a glass from a liquid.
possible to check the validity of this theory in the case of A system of equations describing the dynamics of fluc-
different scattering media. tuations in such a medium includes the Navier-Stokes equa-
It has been demonstrated experimentally394 that in the tion, the equation of continuity, the heat conduction equa-
case of low-viscosity liquids the theory of Ref. 1 describes tion, and a set of equations describing relaxation parameters.
experimental results to within better than 1-2%. In the case Symbolically the system can be represented in the form
of solids this situation is much more complex. The existence
of a static shear modulus has the effect that the number of h

parameters governing scattering is five, whereas in the case where Lqis the operator matrix correspznding to the hydro-
of liquid it is two. Verification of the theory requires inde- dynamic and relaxation equations, i.e., Lurepresents opera-
pendent determination of all these parameters. Moreover, tors which are linear in time t and in the coordinatesx, y, and
the only type of isotropic solids is in the form of glasses. In z; pi are the macroscopic variables; fy' are the Langevin
spite of the many investigations of the scattering of light in fluctuation forces. The system ( 1) is complete if it is supple-
such media, the attempts to describe these results within the mented by the equation of state, i.e., by the dependence of
framework of any theory have so far been unsu~cessful.~-~ the free energy of the medium on the investigated macro-
Our main task was to investigate experimentally the scopic variables.
scattering of light in glasses. It should be pointed out imme- The system of equations ( 1) allows us to express nonsi-
diately that our results (obtained for fused quartz) were in multaneous correlation functions of the variables
strong conflict with the predictions of the theory of Ref. 1. (pi ( t ) p j ( t ') ) in terms of hydrodynamic responses and cor-
We shall show that this conflict is due to anomalous slowing relation functions of the fluctuation forces, which in turn
down of relaxation on transition of a liquid to a (according to the fluctuation-dissipation theorem) are pro-
Allowance for the anomalous increase in the relaxation time portional to k, T and to the corresponding transport coeffi-
together with some additional assumptions makes it possible cients (k,- and Tare the Boltzmann constant and the aver-
to modify the formulas of theory of Ref. 1. The results of age temperature of the system).
such a modified theory will be found to be in quantitative Fluctuations of the permittivity cap can be expressed
agreement with our experimental data. linearly in terms of deviations of macroscopic variables from
their equilibrium values. In the spectral representation, i.e.,
THEORETICAL SCHEME AND DESIGN OF EXPERIMENTS as a result of the Fourier transformation with respect tox, y,
We shall recount briefly the general scheme used in cal- z, and t , we find that E , ~( q , ~ can
) be described as follows in
culation of the coefficient representing the scattering of light terms by the spectral amplitudes of these variables:
in an isotropic medium based on the equations of hydrody-
namics and on the fluctuation-dissipation theorem (for de-
tails see Ref. 1) . The intensity of the scattered light is gov-
erned by fluctuations of the permittivity E , ~ ,which in turn
are related to deviations of macroscopic variables from their Here, X , and Y, are the mechano-optic coefficients of
average values. We shall assume that, in addition to the spectral amplitudes of shear strains (iiaB) and of fluctu-
quantities which are conserved (density, momentum, and ations of isotropic compression ( u = - Sp/po,po and& are
energy), a medium is described by a set of relaxation param- the average value and fluctuations of the density); Z , is the

787 Sov. Phys. JETP 65 (4), April 1987 0038-5646/87/040787-08$04.00 @ 1987 American Institute of Physics 787
thermo-optic coefficient in the case of temperature fluctu- scattered wave K t lies in the xy plane making a scattering
ations ( T I ) ;gti' and f $ are the sets of scalar and tensor angle 9 with the x axis. The directions of the unit vectors p
relaxation parameters. The correlation function ( E , ~ E ; ~), and p' are described by the angles p and $ relative to the
governing the light scattering coefficient is then [see Eq. vertical axis z. The components are therefore described by
(22) in Ref. 1]
q=K'-K= {K(cos 6 - I ) , K sin 8, 01,
p= (0, sin cp, cos cp),
p'= {-sin $I sin 6.sin $I cos 6, cos $1.

Below we shall replace the vectors q, p, and p' with the angu-
Equation (3) contains five independent tensors composed of lar variables 8 , p, and $.
the components of the wave vector q and of the Kronecker In the experiments under discussion the quantities
deltas. The tensor structure of the correlation function (3) which were measured directly were the integrated values of
follows directly from the symmetry consideration^.'^ The the scattering coefficient related to Eq. ( 4 ) by
coefficients Mi(q,w) are combinations of the correlation OD

functions of macroscopic variables. The use of the equations R?(B) = J R ~ ( Bo)ao.

, (7)
of hydrodynamics and of the fluctuation-dissipation - m

theorem makes it possible to determine the explicit form of

Using Eqs. (3)-(6) and the explicit form of the coefficients
Mi (q,w) and, in particular, the dependence on the param-
Mi (q,w) given in Ref. 1, we obtain the integrated scattering
eters of the scattering medium. For example, the expression coefficient
for M I (q,w) has the form'

M,(q,o)=- k ~ T { X2(a)
p ( a )q2--poo2
(2~)~iw
- C.C.

where X(w ) and p (w) are the values of the mechano-optic

coefficientand of the shear modulus at the frequency w.
Considered as a function of q and w, the coefficients
Mi (q,w) represent the ratios of polynomials. Vanishing of
the denominators, i.e., the complex poles M i(q,w) deter-
mine the positions of the components of the spectrum of the
scattered light (Brillouin doublet, wings due to compres-
sion, shear, etc. ) .
.:
The light-scattering coefficient R (q,w) is related di-
rectly to the correlation function of Eq. (3):
The quantities Xo and Yoin Eq. (9) are the coefficients of the
expansion of E , ~in terms of ii, and u in the case when all
whereR is the wavelength ofthe incident light. Therefore, we the other variables reach equilibrium at fluctuations iia8and
are interested only in the convslutions of the tensors that u (static values). However, X, and Y, in Eq. ( 2 ) are the
occur in Eq. (3) and have the polarization unit vectors of the coefficients in the expansion of the permittivity in terms of
incident (p) and scattered (p') light: i, and u for fixed values of the remaining variable, i.e., they
correspond to the case when the other variables do not reach
equilibrium with fluctuations of the density and shear (high-
frequency values). The quantities KOand uo are the static
1 1 + cos 6 values of the bulk and shear moduli; A = r 2 k , /A 4. There-
---(¶P)'(~P')~= 8 sin2( c p + ~ )
Q" fore, the integrated scattering coefficient of an isotropic me-
dium can be expressed in terms of the set of angular func-
tions Y , 112, rI6, M, and 6' [Eqs. (6) and (6a) ] and a set of
five invariants Ni, which are described by equilibrium prop-
erties of matter. The expressions for the integrated scatter-
ing coefficients of Eqs. ( 8 ) and (9) can be obtained also
directly from thermodynamics allowing for the symmetry of
the fluctuating parameters. In other words, these relation-
ships do not depend on the actual system of the dynamic
equations ( 1) .
where Liquids do not support static shear stresses, so that pa,
KO,and X :/pa vanish1and of all the nonzero invariants only
a=sin cp sin ,I b=cos cp cos.I (6a)
two remain: N, and N,. Equation (8) then reduces to the
The angles 8 , p, and $ occurring in Eqs. ( 6 ) and (6a) familiar Einstein-Cabannes formula5 and the scattering co-
are related in the following way to the experimental geome- efficient depends only on 6, i.e., it depends only on the angle
try used in Ref. 5: the wave vector of the incident wave K is between the unit vectors p and p' [see Eq. (6) 1.
directed along the x axis and the corresponding vector of the In the case of a solid all the invariants N iare generally

788 Sov. Phys. JETP 65 (4), April 1987 Gorodetskil et a/. 788
different from zero, so that a comprehensive experimental results with just these qualitative features makes it necessary
investigation of the Rayleigh scattering in an isotropic solid to modify the theory and, consequently, the expressions for
should include determination of all these invariants. Equa- the scattering invariants.
tion ( 8 ) allows us, if we know a sufficient number of the
measured scattering coefficients, to determine the whole set EXPERIMENTS
Nl-N,. If the number of measured values R $ ( 6 )is greater
We investigated experimentally the Rayleigh scattering
than five, we can then check the self-consistency of the main
in fused quartz, which was selected for the following rea-
theoretical result, which is Eq. (8). On the other hand, some
sons. Samples of this glass can be of high optical quality with
of the invariants N, can be found independently from the
weak stray scattering by foreign inclusions.536 Fused quartz
data on elastic and elasto-optic properties of the investigated is characterized by high values of the ratio Xo/Yo (Ref. 11),
substance, which makes it possible to check the quantitative which results in a major relative contribution of the invar-
predictions of the theory. iants N,, N,, and N3 to the overall scattering and this in turn
Selection of the method of carrying out a comprehen- provides an opportunity for the experimental determination
sive series of experiments is to some extent arbitrary. For of these invariants at the precision level feasible at present.
example, we can determine five values of the scattering coef-
Finally, fused quartz has been investigated thoroughly: a
ficient [the sets of angles that give a nondegenerate system of sufficiently full set of data is available for this material when
equations can be deduced from Eq. (6) 1. However, this ap- calculations and comparisons are made.
proach is subject to the influence of various distorting fac-
Our measurements were carried out on four samples of
tors (glints, surface defects, etc.). It seems to us it is more fused quartz of the KV grade both immersed in carbon te-
reliable to determine the angular dependences of the scatter- trachloride and without such immersion. Measurements
ing coefficient and to find the invariants N, by approxima- were made at wavelengths of 488 and 515 nm of an argon
tion of the results obtained with Eq. ( 6 ) . We shall use the laser. An FEU-79 photomultiplier, operating in the photon-
following set of the angular dependences counting regime, was used as the photodetector.12 Appara-
tus was calibrated by scattering in benzene and the values of
the scattering coefficient of this substance R,=R
(90") + R ;(90") = 28.8 X cm-' (A = 515 nm) and
36.7x lop6 cm-' (A = 488 nm) were taken from Ref. 3.
Possible distortions due to luminescence and Raman scatter-
ing were avoided by the use of an interference filter. A con-
The first equation gives N, and N4; the second gives N, tribution of stray scattering by the cell walls and the sample
and the ability to check the values of N, and N4 by adding could reach 15% of the measured value. Such scattering was
them; the third gives N, and the fourth gives N,. The indices eliminated by a method similar to that described in Ref. 13
Vand H in Eq. ( 10) correspond to the polarization angles of and based on the fact that the useful signal represented single
0 and 90", respectively. The second of the equations in the scattering and the stray scattering was multiple. Measure-
system ( 10) is derived on the assumption that either g, = a ments were made at room temperature of 20 + 2°C.
or $ = a. In accordance with the system ( l o ) , we measured the
The scattering of light in a solid described by Eq. ( 8 ) dependences R ;(it), R ;(it), R ; ( a ) , R *, (90°),R :(90"),
has the following qualitative features. and R (90"). The results of the measurements on all four
1. The indicatrix of the depolarized scattering R ; (9)is samples in the range of scattering angles 9250" agreed with
extended in the forward direction [R F ( 0 ) > R (: 180")1. one another to within 5%. The value of the scattering coeffi-
This follows directly from the positive definite nature of the cient R v at A=515 nm was (0.104+0.005)Rb,
invariant N, and the explicit form of the angular function v . = (3.0 + 0.15) X l o p 6 cm-', where R b, is the scattering
2. The Krishnan ratio obeys p h = R ;(90")/R :: (90") coefficient of benzene; the value of the depolarization coeffi-
> 1. This condition can be obtained by substituting, in the cient was A, =R ;(9Oo)/R L(90") = (4.2 f 0.2). 10p2.
determination of the values ofp, ,the explicit expressions for These values agreed with those reported elsewhere, namely
the scattering coefficients R F(90") and R (90") and using with R v = (2.5-2.9) X l o p b cm-' given in Refs. 5-7, 14,
the fact that the Poisson ratio is O ( o < 1/2. Hence, we find and 15 (reduced to the wavelength A = 5 15 nm) and A ,
that N, - 2N2>,0. = 0.04-0.05 given in Refs. 5-7.
1. The indicatrix of the depolarized scattering R ;(it) is Our experimental results are plotted in Figs. 1-4. The
extended in the forward direction [R ;(O) > R (; 180") 1. graph R L (9)in Fig. 1 shows that for the scattering angles
This follows directly from the positive definite nature of the 9 < 50" the scattering coefficient is independent of 9 to with-
invariant N, and the explicit form of the angular function v. -
in 5%, which is in agreement with Eq. (8). Some increase
2. The Krishnan ratio obeys p h = R ;(90")/R (90") in R L(9) in the range 9 < 50", observed also in Ref. 6, is
> 1. This condition can be obtained by substituting, in the possible due to the large-scale structure of fused quartz ob-
determination of the values ofp, , the explicit expressions for served earlier.16An estimate of the characteristic size of the
::
the scattering coefficients R F(90") and R (90") and using homogeneities deduced from the scattering angles at which
the fact that the Poisson ratio is O < a < 1/2. Hence, we find -
the effect was observed gave a value of 1000 A, in agree-
that N, - 2N2>0. ment with the results obtained in Ref. 16. Therefore, in the
3. All the invariants N,, with the exception of N3, are theoretical interpretation we decided to ignore the data for
nonnegative. the scattering at angles 9 < 50".
We shall below that a comparison of the experimental The curves plotted in Figs. 2 and 3 are in conflict with

789 Sov. Phys. JETP 65 (4), April 1987 Gorodetskil et a/. 789
FIG. 1. FIG. 3.

the above qualitative characteristics of the scattering in equi- which the relaxation time of the system is of the same order
librium solids in the following respects. of magnitude as the observation time. We can understand
1. The indicatrix of the depolarized scattering is ex- better the sense of the changes which we shall have to make
tended in the backward direction (Fig. 2), which is possible in the equilibrium theory by considering, by way of illustra-
only in the case of a negative value of the invariant N,. tion, the simplest model of a liquid, for which one relaxation
2. The Krishnan ratio is less than unity (Fig. 3). time can be considerably greater not only than all the other
3. The invariant N, deduced from the experimental data characteristic times of the system but also than t , .
is also negative. For sin~plicity,we shall assume that the thermodynam-
It is clear from just these results that it is pointless to try ic state of such a liquid is described by just two variables: the
to compare quantitatively the experimental results with the density p and the scalar relaxation parameter 6, character-
theory of Ref. 1. ized by a relaxation time which rises anomalously. Follow-
ing the general approach presented at the beginning of this
MODIFICATION OF THE THEORY OF THE RAYLEIGH paper, we shall write down the explicit system of equations
SCATTERING AND INTERPRETATIONOF THE ( 1) for this case:
EXPERIMENTAL RESULTS
We shall assume that the most important property of
glass samples resulting in such a major disagreement
between the experimental data and the theory of Ref. 1 is the
anomalous slowing down of the processes of establishment
of an equilibrium on approach to the glass-formation region
(see, for example, Refs. 8 and 9). The times needed to estab- where Sp and SP are the deviations of the density and pres-
lish an equilibrium increase as a result of cooling in accor- sure from the equilibrium values; 6 is the total viscosity of
dance with a nearly exponential law and in a narrow range of the liquid; a is a transport coefficient; 2, is the tensor of
temperatures AT< T the values of these times may change fluctuation stresses; Z is a fluctuation force with is the conju-
from negligible, compared with the measurement time t, ,to gate of {; \I! is the free energy of the system.
values many times greater than a reasonable t , . The glass- The pressure Pand the quantity h = d\I!/df considered
transition temperature is essentially the temperature at in an approximation linear in fluctuations are

where Pic' = (dP /dp ) c etc. Substituting Eq. ( 12) into Eq.

FIG. 2. FIG. 4.

790 Sov. Phys. JETP 65 (4), April 1987

( 11) and going over to the Fourier representation, we obtain Eq. ( 16) by a term proportional to ( 121 2). We shall call it
a system of two linear algebraic equations: the fast subsystem. The second (slow) subsystem is associat-
ed with a relaxation mode of frequency of the order of 7- '. It
is represented in Eq. ( 16) by the term with (12 1'). If fluctu-
(0) ations of the density are written down as the sum of the
- i a ( P ~ / p a ) 6p (q, o)+GEE (q, o)=-iaE, rapidly relaxing (Sp,) and slowly relaxing (Sp,) parts, then
where the spectral density Sp, is equal to the first term in Eq. ( 15),
whereas Sp, is equal to the second term. In accordance with
Eq. (16), the fast and slow contributions are uncorrelated,
(Sp,Sp:), = 0, and this means that their fluctuations can be
calculated independently.
The solution of this system is In our model the fast subsystem is always in equilibrium
(t, > y-'), so that the mean-square value of the fluctu-
ations in the fast subsystem can be found by integration of
the first term in Eq. ( 16) with respect to all the frequencies:

As long as the slow subsystem is also in equilibrium

where A (q,w) = G,, Gc - i a P ~ ) 2 q 2 / is
p ~the determinant with the thermostat, the mean-square value of Sp, can also
of the system ( 13). be found in a similar manner:
The expressions in Eq. ( 14) allow us to find the spectral
densities of macroscopic variables. For example, the spectral
density of fluctuations Sp is given by
[in the derivation of Eqs. ( 17) and ( 18) we have allowed for
the fact that, in accordance with the fluctuation-dissipation
where theorem, we have ( / E l 2 )= < k B T / r and (1=l2) = k,T/
r~ 1. The mean-square values of the density fluctuations
can be described, as expected, by the thermodynamic expres-
sion
Equation (15) is derived on the assumption that
(I;, Z * ) is obeyed in the thermodynamic equilibrium state
of a medium in the presence of random forces.
In a medium at equilibrium the mean-square value of However, if the relaxation timer rises without limit, we
fluctuations of the density is equal to the integral of Eq. ( 15) unavoidably reach a situation described by w; '$t, . We
with respect to all the frequencies. An equilibrium then oc- shall show later that this corresponds to the transition from
curs when all the times for establishment of the equilibrium the liquid to the glassy state. The slow subsystem cannot
values are much shorter than the measurement time t, . The then attain equilibrium during the measurement time and
time for the establishment of an equilibrium is governed by the corresponding part of the density fluctuations 6p, is
the imaginary parts of the roots of the equation A(q,w) = 0. found to be static (freezes), so that this part is not in equilib-
This equation has three roots: two acoustic roots o,,, rium at lower temperatures.
= f w,, +iy, and one relaxation root w, = iw,ir-' (we Integration of the second term in Eq. ( 16) with respect
to the frequency (equivalent to averaging with respect to
shall assume that the absorption of sound is weak, w,, $ y,
which is normally true in the case of real systems). In the time) does not, because the resultant nonergodicity, neces-
case of slow relaxation characterized by y >r-', the quanti- sarily give the same result as averaging over an ensemble.
ties governing the roots of the equation A(q,w) = 0 are as Therefore, we cannot calculate ( ISp 1 ,) rigorou~ly.~' How-
follows ever, in view of the anomalous increase in the relaxation time
as a result of cooling, we can assume that this quantity re-
(5) lh) (I) (P)
o:,=Pp g2, y=%g2/2po, o,=t-'P, lP, =ah; . mains constant throughout the range T < T, and equal to its
value at the glass-transition temperature T = Tg , i.e.,
'
If the inequalities w,, % y %r- are obeyed, Eq. ( 15)
simplifies greatly. In the main approximation with respect to
the parameters l/w,, r and l/yr, it reduces to the following
expression: The glass-transition temperature T, is found from w, ( Tg )
= t ; ' (Ref.9). WethusseethatifT<T,, then

The structure of Eq. ( 16), obtained from the theory of

the equilibrium case, corresponds to the separation of fluctu- We stress once again that if t, < r , then the rapidly re-
ations of macroscopic variables into two subsystems. The laxing subsystem remains in equilibrium. The thermody-
first subsystem, associated with a rapidly relaxing acoustic namic description is valid for this system and the usual rela-
wave [or in the more general case, associated with all the tionships between thermodynamic derivatives apply. The
'
modes of frequencies obeying Im(wi ) % r- is represented in frozen subsystem makes no contribution to the thermody-

791 Sov. Phys. JETP 65 (4), April 1987 Gorodetskilet aL 791

namic derivatives because it is static. When such a divisionof integration of the relevant terms in Eq. (20) with respect to
the time scales is made, then Eqs. ( 17) and ( 18), like Eqs. all the frequencies:
( 8 ) and ( 9 ) , follow directly from thermodynamics and are
N,'" = A T X m 2 / p , ,
independent of the nature of the system of equations ( 11) or
the spectral representation ( 16). Equation ( 19) follows
IL-po
+ ---
( l ~ ~ - p opop,
)
from Eqs. ( 17) and ( 18) if we additionally assume that the
slow subsystem freezes and, therefore, it is also independent
hence, in full agreement with Eq. ( 9 ) , we obtain
of the actual form of the initial system of equations ( 11) . In
the liquid range the low-frequency bulk modulus measured
in a static experiment K A') is governed by the thermodynam-
ic derivative Pih':K;') In the case of a glass the If t , < T and the assumptions of the simplified model
low-frequency bulk modulus K Ag' measured in a similar ex- apply, we obtain
periment is now governed by the derivative Pic': KAg'
=p&';". On the other hand, the derivative Pic' gives the
high-frequency value of the bulk modulus of a liquid ( K ), :' We must bear in mind that in the frozen contribution of the
i.e., slow subsystem N 12'(Tg) we have to include not only an
explicit temperature dependence in the form of the factor
k , T, which originates from the correlation function of the
Similar relationships apply also to the other parameters fluctuation forces, but also contributions due to possible
of the system. The relationship between the high-frequency temperature dependences of X,, X, ,p,, and p ,.
characteristics of a liquid and the low-frequency characteris- In this way we obtain expressions also for the other in-
tics of a glass is necessary in the interpretation of the experi- variants:
mental results.
We shall now turn from the above example to the more
general situation when the state of a system is described by Since a system in the equilibrium region at tempera-
the full set of variables of Eq. ( 2 ) . Once again, we shall make tures T > Tg represents a liquid for which the first three in-
the separation into rapidly and slowly relaxing subsystems. variants vanish, it follows that in the glassy region, (i.e., at
The only thing that we need is a considerable difference temperatures T < T, ), these invariants are of the form
between the scale of the characteristic times. Then, the ex-
pressions for the coefficients M i ( q , u ) in Eq. ( 3 ) [see also
Eq. (22) in Ref. 1] split into two parts, one of which is asso-
Neglecting temperature dependences of X _ , Y , , KO,K _ ,
ciated only the fast subsystem and the other with the slow
and p ,, we obtain
subsystem. The contribution of the fast subsystem to the in-
tegrated scattering intensity of Eq. ( 7 ) is governed, as point-
ed out above, by modes of frequencies which obey Im(w, )
>) 7- l. The contribution of the slow subsystem is related to Therefore, the sign of the first three invariants for glass is
the modes of frequencies w, a 7-', i.e., the expressions for opposite to the sign of the invariant for the fast subsystem.
the coefficient M, (q,w) have a structure similar to that of Since in the case of the fast subsystem we find that the equi-
Eq. ( 16). For example, the quantity M I (q,w) [see Eq. ( 4 ) ] librium expressions of Eq. ( 8 ) are valid, it follows that both
considered in the same approximation as that used to derive I*)
N I" and N GO. Consequently, in the case of a glass the
Eq. (16) is two invariants are N, and N,>O. We can readily show that
the inequalities governing the values of N , - 2N2 and ph in
the case of a glass are also opposite to those in the case of an
equilibrium isotropic solid. These qualitative features of
glasses, liquids, and equilibrium isotropic solids are listed in
Table I.
In considering the general case corresponding to Eq.
( 1) with the full set of variables of Eq. ( 2 ) , we were assum-
where ( 18'12) = k g Tr]/n-and ( IZ 12) = k , TT/T are quanti- ing that the slow variable is a tensor relaxation parameter of
ties obtained from the correlations of the fluctuation forces an arbitrary form. Relaxation of just this tensor parameter
2, and gik,which are the conjugates of the strain tensor results, as is demonstrated by the equations of hydrodyna-
and of the slowly relaxing tensor parameter'; r] is the shear mics ( 1), in the appearance of dynamic shear stresses in a
viscosity; system (representing high-frequency shear sound"). How-

TABLE I
w, = ~ - ~ p , / p.,As in Ref. 15, the term with (12'12) repre-
sents the contribution of the fast subsystem and the term
with ( I Z 12) represents the contribution of the slow system
I ETilibrium
solid
I Liquid / Glass

M,(q,m).
The equilibrium values of the fast ( N i l ' and slow
( N 12') contributions to the invariant N, can be found by

792 Sov. Phys. JETP 65 (4), April 1987 Gorodetskil et a/. 792
 listed in various handbooks, namely 1350 f 100 K (Refs. 7,
Experiment Calculation 20, and 22).
i
N$').IO' I Ni,108cm
Since in the liquid state (when T > T, ) the value of the
invariant N, does not vanish (in contrast to N,, N,, N,), we
had to know the value of the modulus K A') at temperatures
T Z T,. We took the value found by x-ray diffraction mea-
surementsZ3:KA" ( T 2 T, ) = ( 1.5 f 0.15) X 10'' J/m3.
It is clear from Table I1 that, within the limits of the
total experimental error, the calculated and experimental
values of the invariants are in agreement. ~ a t u r a l l our
~,
ever, if t, <T, these shear stresses are retained throughout modified theory is not based on general principles of statisti-
the measurement period, i.e., a finite static shear modulus cal physics, but simply on physically reasonable assump-
appears in the system. However, this implies the transition tions. However, we must stress that the good (not only quali-
from a liquid to a solid (glass). tative but also quantitative) agreement between our
The relationship between the shear stresses and the formulas and the experimental data, achieved without any
slow tensor relaxation parameter gives the following rela- fitting, supports strongly the assumptions and our descrip-
tionship between the invariants N, and N,: tion of the Rayleigh scattering of light in glasses.
It is worth pointing out an interesting observation
which applies directly to fused quartz. It is clear from Table
(the contributions of the slow subsystem balance out in this I1 that the inequality for N, can be replaced (within the
expression). limits of the experimental error) by the equality. This means
that the only tensor parameter of fused quartz which experi-
It follows from the experimental results that qualitative
ences relaxation is that which is responsible for the forma-
characteristics of the scattering in fused quartz agree with
tion of a glass. In this sense we can regard fused quartz as the
those of glass obtained in our modified theory. We shall now
"minimal" model of a glass.
carry out a quantitative comparison. Table I1 gives the ex-
The authors are grateful to N. V. Andreev and N. A.
perimental and calculated values of the invariants for fused
Bokov for valuable discussions, and to M. M. Mazur and N.
quartz. The experimental values are obtained by approxima-
V. Orekhova for their technical help.
tion of the data (Figs. 1-4) with the dependences in Eq.
(10). The errors listed in Table I1 include the confidence
limit of the approximation as well as the error in the deter- "The term with 22,in N, was omitted in Ref. 1 because of the incorrect
mination of the absolute values of the scattering coefficient statement that "for all the intensities of Eq. (24) we have g ( m ) /
using the selected apparatus. The contributions of the fast f(m ) = 0." In fact, the term with 22,which is a coefficient of ( T , T : ) ,
[see Eqs. ( 19) and (24) in Ref. 1] differs from zero in the limit 0-0.
subsystem were calculated using the following values of the "If T < T,, we have a time-independent field of frozen inhomogeneities
elastic and elasto-optic characteristics: Xo = 0.66 + 0.04; 6p2(r),which contribute to the intensity of the unshifted component in
Yo = - 1.01 + 0.06; &' = (3.0 + 0.2) x 101° J/m3; K Ag' the spectrum of scattered light in the form of the S function of the fre-
quency o.The time averaging of this field simply does not alter anything.
= (3.7 f 0.4) x 10" J/m3. These values were obtained
In this case it would have been natural to assume the existence of spatial
from the Pockels constants p,, = 0.27 + 0.01 and p,, ergodicity at which the statistical average is equal to the volume average.
= - 0.072 + 0.004 (Refs. 11 and 18), from the refractive However, we cannot carry out the correct averaging of the volume, be-
index n = E'/* = 1.462 (Ref. 19), from the velocities of lon- cause we do not know the statistics of the field 6p2(r).
gitudinal V, = 5900 50 m/sec and transverse V , = 3700 'S. M. Rytov, Zh. Eksp Teor. Fiz. 58,2154 (1970) [Sov. Phys. JETP31,
+ 100 m/sec s o ~ n d , ~ and
~ ' ~from
' ~ the density po = 2.206 1163 (1970)l.
'E. R. Pike, W. R. M. Pomeroy, and J. M. Vaughan, J. Chem. Phys. 62,
g/cm3 (Ref. 19). All the values apply at room temperature 3188 (1975).
T = 293 K at the wavelength of A = 5 15 nm. The relation- 'E. E. Gorodetskii, A. M. Evtyushenkov, V. S. Esipov, and Yu. F. Kiya-
ships between the initial values and those in Eq. (8) are as chenko, Zh. Eksp. Teor. Fiz. 81,588 ( 1981) [Sov. Phys. JETP 54,314
(1981)l.
follows: 4M. A. Anisimov, E. E. Gorodetskil, A. M. Evtyushenkov, and Yu. F.
Kivachenko. Oat. Spektrosk. 54,505 ( 1983) [Opt. . Spectrosc.
. (USSR)
54:297 (1983)j.
51. L. Fabelinskil. Molecular Scatterinn - o- f Linht,
- Plenum Press, New
York (1968).
6N. A. Bokov and N. S. Andreev, Fiz. Khim. Stekla 7, 509 (1981).
Since the data on the contribution of rapidly relaxing '5. Schroeder, Treatise Mater. Sci. Technol. 12, 157 (1977).
'N. Ostrowsky, in: Photon Correlation and Light Beating Spectroscopy
parameters were not available in the interval corresponding red. bv H. Z. Cummins and E. R. Pike), Plenum Press, New York
to N:" and Xi1),only the inequalities could be given for i1974j, p. 539.
these invariants. 9J. Wong and C. A. Angell, Glass: Structure by Spectroscopy, Dekker,
New York (1976), p. 707.
In calculating the total values of the invariants we need- '"B.Ya. Zel'dovich, Zh. Eksp. Teor. Fiz. 63,75 ( 1972) [Sov. Phys. JETP
ed the values of the moduli and the glass-transition tempera- 36,39 (1973)l.
ture and the temperature itself. Following the data presented "W. Primak and D. Post, J. A ~ p lPhys.
. 30,779 (1959).
o , I. Olefirenko, and I. K.
"A. M. Evtyushenkov, Yu. ~ . * i < i ~ a c h e n kG.
in Fig. 7 of Ref. 20, we assumed that the elastic moduli in- Yudin, Prib. Tekh. Eksp. No. 5, 157 (1981).
creased in the interval between T = 293 K and Tg = 1400 K 13A. M. Evtyushenkov and Yu. F. Kiyachenko, Opt. Spektrosk. 54, 692
by 10%. This was in agreement with the studies of the hyper- (1983) [Opt. Spectrosc. (USSR) 54,409 (1983)l.
14M.Goldstein, J. Appl. Phys. 30, 493 (1959).
sonic properties reported in Ref. 21. The glass-transition ''A. R. Tynes, A. D. Pearson, and W. R. Northover, J. Am. Ceram. Soc.
temperature T, was taken to be the average value of those 62,324 (1979).

793 Sov. Phys.JETP 65 (4), April 1887 Gorodetskilet a1 793

 I6M.A. Kozlova, V. V. Korneev, and 0.V. Korytin, Fiz. Khim. Stekla 10, Electric Vacuum Devices [in Russian], Energoizdat, Moscow ( 1981).
471 (1984). "J. Pelous and R. Vacher, Solid State Commun. 18, 657 ( 1976).
"I. L. Fabelinskii, L. M. Sabirov, and V. S. Starunov, Phys. Lett A29,414 22V.P. Pryanishnikov, Silicon System [in Russian], Izd. Lit. po Stroi-
(1969). tel'stvu, Leningrad ( 1971).
I8N. F. Borrelli and R. A. Miller, Appl. Opt. 7, 745 ( 1968). 23E.A. Porai-Koshits, Fiz. Khim. Stekla 3,292 (1977).
I9I. K. Kikoin (ed.), Tables of Physical Quantities (Handbook) [in Rus-
sian], Atomizdat, Moscow ( 1976).
'OL. P. Khodkevich and V. K. Leko, Quartz Glass in Manufacture of Translated by A. Tybulewicz

794 Sov. Phys. JETP 65 (4), April 1987 Gorodetskilet al. 794