PH YSICAL REVIEW A VOLUME 20, NUMBER 5 NOVEMBER l979

Phase separation and coalescence in critically quenched

isobutyric-acid —water and 2, 6-lutidine-water mixtures
Y. C. Chou and Walter I. Goldburg
Department of Physics and Astronomy, University of Pittsburgh, Pittsburgh, Pennsyluania 15260
(Received 29 May 1979)
Light scattering was used to study phase separation in critically quenched mixtures of isobutyric-
acid-water and 2, 6-1utidine —water. The measurements spanned the time interval 10(t 10 s and quench
depths from 0.7 to 9 mK. The parameters measured were k (t), i.e., the photon momentum transfer at the
angle of maximum scattering, and the ring intensity I(k,
t). The early-. stage measurements. of k vs t are in

domain growth, which gives k cr.

This implies that hydrodynamic
t,
very good agreement with the calculations of Kawasaki and Ohta and with the coalescence model of
with q = 1/3. In the late stage y increases to unity in both systems.
effects control the rate of growth of the nucleating domains of size
I k '. In both early and late stages, the intensity measurements are consistent with I(k,
t)~/'. The
scattering experiments were supplemented by direct observation of domain growth with a microscope.

I. INTRODUCTION odal decomposition was first observed in metals

but recently it has been studied in fluids. 4
This paper is concerned with the dynamics of The parameters most readily measured in spin-
phase separation in critical solutions. Because odal decomposition experiments are 8~ vs t and
the initial size of the nucleating grains can be the scattering intensity I(8m, t). In these so-called
as large as a micron near the critical point, the critical-quench experiments the fluid is driven
process is conveniently studied by light scattering through the critical temperature T, from its initial
and by direct microscopic observation, and we one-phase temperature T; to its final two-phase
have used both of these experimental tools. Mea- temperature Tf. The ring diameter is then mea-
surements were made on two mixtures: isobuiyric sured as a function of time and temperature dif-
acid and water (I-W) and 2, 6-lutidine and water ference &TI =- Tr —T, ~, there being only a weak
~

(I -W). The present results confirm that hydro- dependence on ATt = Tt —T, (. ~

dynamic effects are important in the late stage of The most extensive spinodal decomposition
the process, an early conjecture of Cahn and measurements to date are those of Wong and Knob-
Moldover. Droplet growth is speeded up by fluid ler (W-K), and we will frequently refer, to their
flow, which in turn is driven by the curvature of work. Investigating the I-W system they measured
irregular surfaces which interpenetrate the nu- h~(t), fitting their data to the form it (t) cc t ".
cleating fluid. To set the present study in per- The results were consistent with a time-indepen-
spective we review earlier related experiments, dent p, though this parameter did depend on quench
most of which were carried out on these same depth 4T&.
systems. More recently Kim et al. ' reported phase-sepa-
Using an ingenious pressure-jump technique to ration measurements in L-W using a microscope
quench the system, Wong and Knobler' made exten- to measure l(t) in the late' stage (100 sec 1000 (t (
sive light scattering measurements of phase sepa- sec) and light scattering in the early stage (10 sec
ration in isobutryic acid and water. When the & f & 100 sec). They observed a crossover from

critical mixture is quenched a few mK into the slow (tp = s) to rapid (p = 1. 7) coarsening. The
two-phase region it scatters light in the form of a results were not fully satisfying, however, be-
ring which brightens, sharpens, and decreases in cause neither experimental technique spanned
diameter as the system evolves toward its final both the slow and fast growth regimes. The pres-
equilibrium state. F rom elementary scattering ent experiments were undertaken to remove this
theory one expects that the ring diameter em, or deficiency. A secondary goal was to verify that
rather the corresponding photon momentum change the light scattering results were sample indepen-
k~, is inversely proportional to the characteristic dent and insensitive to both the quench technique
size l of the nucleating regions in the fluid. This and the quench rate. Because our light scattering
ring has been associated with spinodal decomposi- . measurements spanned a longer time interval
tion, a term introduced by Cahn. ' Cahn was the than those of W-K, a speedup in the droplet growth
first to consider and analyze the process of phase rate was more clearly seen and p was itself found
separation from an initially unstable state. Spin- to be time dependent.

20 2105 1979 The American Physical Society

 Y. C. CHOU AND %AI TER I. GOI DBURG

As a check on the above cited microscopic mea- The laser intensity didnot exceed 0. 3 m% and
surements of l(f) in L-W, we have used this same produced no observable local heating in the mix-
technique to measure domain growth in I -W and I- tures.
W, while simultaneously recording the ring dia- To monitor the intensity of the unscattered beam
meter, i. e. , k (t). Such experiments reveal a microscope slide (6) reflected a small fraction
whether k~ (t) is roughly equal to the visually de- of the beam into the photodiode D. Its output I~
tected given size I and whether both parameters provided a measure of the optical thickness of the
are characterized by the same exponent, y(t). sample and thereby the axnount of multiple scat-
The remainder of the paper is organized as tering from it.
follows. Section 0 gives experimental details. The remaining unscattered intensity, as well as
In Sec. III our findings are presented along with a the scattered radiation &(&), continued through
graphical comparison with the early-stage spinodal the microscope slide and fell on the rotating mir-
decomposition theories of I anger, Bar-on, and ror B. The mirror reflected the scattered light
Miller, and Kawasaki and Phta. Section IV is into the photomultipliers Pq a, nd P2 (RCA Model
devoted to a discussion of the present results and 1P28). A stepping motor rotated the mirror so
a comparison with earlier experiments. Emphasis that its angular position o!(t) varied in a sym-
will be placed on analysis of the late stage of the metrical sawtooth fashion. Its axis of rotation
process, due to Siggia. Finally, the work is coincided with that of the incident laser beam.
summarized in Sec. IV. The amplitude of n(t) was several degrees and its
period T was 4 sec. The photomultiplier recorded
II. EXPERIMENTAL &(&, f) only over a small fraction of T so that a
complete angular distribution was recorded rapidly
A. Apparatus enough such that &(&, f) changed very little during
a single sweep.
The experimental arrangement is shown in Fig.
At the same instant in each mirror cycle a
1. The sample (&) was located in a temperature-
controlled, rapidly stirred water bath (8) identical
to that used by Goldburg et al. (This reference,
"
stepping-motor pulse triggered a minicomputer
(PDPl1/10), which sampled the voltage outputs
of &q or I'2. These detectors were located at dif-
which contains many relevant experimental details,
ferent distances from the mirror and therefore
will be subsequently referred to as l. ) The bath
recorded the scattering over different angular
temperature, which was monitored by quartz
ranges and with correspondingly different angular
thermometer, was stable to within +0. 1 mK over
resolution &~g, 2. The closer photomultplier, &2,
at least 1800 sec, which was the time span of the
recorded &(&) during the first part of the run when
longest run. The incident laser beam, of vacuum
the ring was relatively large and relatively weak
wavelength &0 = 6328 A, went through an attenuator
and was, therefore, operated at greater sensitivity
(&), was reflected by a mirror (M), and then went
and lower angular resolution than P~(&62=2. 5
through a pinhole (S) before it reached the sample.
mrad).
When the scattered intensity exceeded the dy-
namic range of the computer input it was manually
He-Ne Laser
switched over to receive the signal from the
more distant detector &q whose angular resolution
was &8~ =1.5 mrad. With these two detectors
operated in tandem the scattered intensity could
be recorded over four decades. This scheme
8 requires that the relative gains of the two photo-
multipliers be accurately known. To minimize
LJc drift the gain ratios were remeasured after every
run.

B. Sample preparation
Both samples were prepared from commercially
available components, using multiply distilled
water. They were of critical composition as
evidenced by the volume equality of the two phases
FIG. 1. Experimental arrangement, described in Sec. on cooling -2 mK below &, . The measured critical
Il A. composition (c = c,) and critical temperatures
20 PHASE SEPARATION IN CRITICALLY QUENCHED. . . 2107

determined in this way mere close to the generally tiple scattering, gave the "corrected intensity"
accepted values for I-% and I -%. ' These
s(e, t) = [1(e, f) —I (8))yI„(f) .
are I-W: c, = 38. 6 wt. acid, &, =26. 08 'C;
'%%uo

L-W: c, = 28. 16 wt. 2, 6- lutidine T, =33. 37 'C.

'%%u~, The background term &s(e) contains a contribution
The 2, 6- lutidin. e-water syst, em, has an inverted co- from spontaneous composition fluctuations at
existence curve so this 7, is a lower critical tem- & = T;, but it was negligibly small compared to
perature. %'hatever impurities were present in the scattering produced by the phase-separating
our samples are expected to have little effect mixture.
other than to shift T, and perha. ps c, . Of the wealth of data. obtained from this experi-
Except where otherwise specified, the sample ment we present only our measurements of the
cells were made of glass and were of the reen- time dependence of the ring diameter k~(t) and ring
trant type. They accommodated roughly 1 ml of = S(e~, f), where e~ is defined above.
intensity S~(t) —
solution. The optical path was only 100 Q in The momentum transfer at ~= e~ is found from
order to minimize multiple scattering. 'To deter- '
0 = (4vn/Xo) sin —,8, (2)
mine if our results mere influenced by sample
thickness, a one-mm cell was used in some of the where n is the refractive index of the critical
microscope measurements. Both cells are mixture. For our systems, ideal-solution. calcula-
sketched in Fig. 3 of I (the thick cell shown there tions give n(I-W) =1.35 and N(L-W) =1.38. These
contains two electrodes which were absent here). refractive indices differ insignificantly from that
The samples were filled under atmospheric pres- of pure mater. for which n =1.33.
sure. *he microscope measurements utilized an in-
I -W is reported to attack glass containers, strument of large working distance so the sample
There- '
producing a slow change in &, with time. could be viewed at the center of the water bath.
fore, T, was remeasured frequently. On such Most of the measurements were made at a magni-
remeasurements &, was taken as the cloud-point fication of 25.
temperature. ) No drift in the critical temperature
was detected. III. RESULTS
A. km vst
C. Procedure
Figures 2 and 3 show km vs t in I-% and L-W,
Prior to quenching, the samples mere in equili- respectively. The quench depths span the range
brium at the initial one-phase temperature T, , 0. 3 mK~T&~9 mK and are shown in the figures.
which was typically several mK from 7;. For The runs were terminated at t -10 sec, when
I -% &~ is less than &„hence a "quench" was an e (f) became too small to measure. The initia. l
upward temperature jump. The heating was ac-
complished by applying a voltage pulse to a bare
Nichrome heater immersed in the rapidly stirred IO
water bath.
The I-W samples was quenched (cooled) by in- f( ~

jecting a few ml of ice water into the bath with a n Dys

syringe. With both quenching schemes the tem- )~
perature changed by &5% of ~T& T, in less than-
~
0 n nn ~
3 sec. In a few more seconds the temperature km/Cm v
0
0
nn n
I~
0
bath stabilized at & =7& so that the light scattering o
o u

measurements could commence at t 10 sec fol- IO— TT'v

n n
nn

lowing the quench at t= 0. At that time the com-

v o o n
v o o

puter commenced its repetitive recording of I(e)

while a chart recorder displayed &I (t) In the.
early stage of phase separation the angular scan-
~
h, Tf /mK
0.7
I. O
ning range (8) at Pq was + 7. 5,
while at P2 it was
~o l. 8
2. 7
+4. 5'. This photomultiplier recorded the ring n 34
5.8
intensity at such small scattering angles that it 9.0
became necessary to subtract spurious scattering l I I

arising from the walls of the mater bath and the IO IO 10

sample cell itself. This background contribution t /sec

Ie(e) was measured at 7 = &q prior to each run. FIG. 2. Ring diameter vs time in isobutyric acid-@ra-
Dividing I(e, f) —&s(8) by lg(t), to correct for mul- ter at the indicated quench depths. &T'~= &f Tg, .
 2108 Y. C. CHOU AND %ALTER I. GOLDBURG 20

IO
s 0
~e P 0
v
~ d v
e~g 0
0 0
~ d oo
~
~w o
d dv ~ 0.7 mK
cF I.O rnK
~ ~e o
v oo I— ~, o l.8 rnK
~g Da
~ ~~
0
O.
P.7mK
3.4 mK
d
km/crn o vv 5.8 mK
9.0mK
Vd d
~ V
QTf /mK d
K
0.3
~ 0. 6 v
O.oi
I. 3 I I I I I

1. 5
I IO IOO I 000
I. 9 L-W

FIG. 4. I-W data of Fig. 2 replotted in the dimension-

less length and time units defined in the text.
I I I I I
IO I

I0 IO IO
wh~re & = & —&, ~/T„and the high-frequency
~

t/sec
kinematic viscosity 'g* = 2. 42 10 P. The same &&

FIG. 3. Bing diameter vs time in 2, 6-lutidine-water. parameters in L-%, as measured by Gulari

et al. , ' are
conditions for I-% were similar to those of W-K
(see their Fig. 5) except that our runs lasted & = (2. 0+ 0. 2 A) ~
'"' '
(6c)
several times longer. and
In these two figures a solid line was drawn by
eye through the data at &7& —1. 6 mK (Fig. 2) and D=(0, 29+0. 02)&&10 s ' ' cm /sec. (6d)
&7& —0. 6 mK (Fig. 3) to draw attention to the The parameters in (6a)-(6d) were measured in the
curvature in k (t). This curvature appears in the one-phase region.
shallower quenches only. The measurements are The most important conclusion to be drawn
consistent with a power-law variation of the from Figs. 4 and 5 is that the ring-diameter mea-
coarsening rate, surements exhibit the expected property of criti-
(t) ~t", (3) cal-point scaling, i. e. , all the measurements
collapse quite satisfactorily to form a single
with y strongly dependent on &&& and slowly in- curved line. Furthermore, at each ~, q~ =q~
creasing with t. (The earlier experiments es- though the measured curve q' "(T) lies measurably
tablished that p is no more than weekly dependent below the corresponding curve for I -W.
on &T;. ) At small quench depths the measure- lf one extracts p(r) from Figs. 4 and 5 it is
ments at small t are consistent with y —3, while seen that
at the largest t all the measurements are encom-
passed by y=0. 95+0. 15. g w(r) = 0. 3+ 0. 1, 7 ~ r & 30,
yg w(T) =1. 1+0. 1, 200~'r ~1000,
m VSV

Figures 4 and 5 show the measurements of Figs.

2 and 3 in dimensionless units, KP
~LBM
q. = n.(t) ((t z', ), (4) o~
'emq 0
r A&(d &,)/i'(t &,)]t (5) +&a
0
Here g is the correlation length in the critical AT)/mK
mixture at a temperature &7' C from the criti- O. I— o 0.3
cal point in theone-phase region, and D is the
)I ~

~ 0.6 V
~O
1
I.3
Mo
composition diffusivity. For I-% we used the I.5 V~
data of Chu et al. ': ~ 1.9 dd~

2, 6- lutidine 8
$ = (3. 57 + 0. 07 A) C (6a)
water

O. OI l I I I

and I IO IOO I 000

D=ksr. /6vrl*g cm'/sec, (6b) FIG. 5. L-W data of Fig. 3 in dimensionless units.
20 PHASE SEPARATION IN CRITICALLY Q UEN C HE D. . . 2109

5 K I I

h, Tf 0
yLW(~) =0. 4+0. 1, 10~ r ~40, ~ 0.7 mK 0
O

yL w(7) =0 95+0 1 300 ~r ~ 1000. +

a
2. 7
34 0
5.8 o~
Limited significance should be given to these nu- +g
4
merical values since they are determined from +0 4
k
less than a decade of data in both 7 and q . + ~
0
The solid line in Fig. 4 summarizes the results
of Wong and Knobler, using Eqs. (4) and (6) to
express their measurements of km vs t in reduced
units. The excellent agreement between the %-K K. O:- . I
measurements and ours indicates a gratifying in- O
sensitivity to sample thickness and purity and to la,
the quenching scheme. /Ne

The dashed lines labeled LBM and KO in Figs. 4 ga

and 5 are the theoretical calculations of Langer,
LBM
Bar-on, and Miller'(LBM) and Kawasaki and Ohta'
(KO). They both use the same basic computational
scheme (that of LBM); it avoids linearization
~
assumptions and takes fluctuations into account. 7
Kawasaki and Ohta, however, have included hydro- 00 ~
dynamic effects. Both calculations share the
I l I

IO IOO IOOO
fundamental assumption that the evolving system
can be characterized by a single characteristic
FIG. 6. Sealed ring intensity vs reduced time in l-%'.
length. They should therefore not apply in the late
stage [k~(t)» $(&T&)], i, e. , when the domain
size is much greater than the interface thickness. curve, as scaling agreements would suggest, but
Another of their assumptions-one which is rela- the slopes of lines drawn through the points differ
tively easily relaxed — is that the system is at small r; if we write 8 ~r then eq ~(10) =0. 6,
quenched instantaneously from an initial state of whereas eL „(10)=0. 9. However, at large r, the
complete disorder, i. e. , &&;=~. In the limited
range over which the present results and the cal-
culations should be compared, the agreement is
10
-very good, especially when the hydrodynamic in-
teraction is included. Theoretically, flow effects
should be unimportant at small ~ so that, as
0, the I BM and KO theories should coincide. L-W
IO
C. Ring intensity versus time

Figures 6 and 7 show the relative ring intensity

versus ~ in I-W and L-W, respectively. The
quench depths &&& are as designated. The inten-
sities have been expressed in reduced units, i. e. , IO
at each &&&, l~ of Eq. (1) has been further divided
by c ", where e = 1 —T&/T, I . The factor contains
I

the temperature dependence of the equilibrium

K.O. —.
j
composition susceptibility. The y values used
were the measured ones: y(1-W) =1. 24 a 0. 05 (Ref. 10
b, Tf/mK
11) and y(&-W) = 1.26+ 0. 02. 0.3 mK
The solid and broken lines labeled LBM and
M ~ 0.6
I.3
KO are the theoretical calculations of the scatter- ~
l. 5
I.9
ing function S(p„, 'r) (from which the previously
discussed function q (&) wa, s extracted). We have 0 I I
IO
arbitrarily normalized the two theoretical curves IOO 1000
to our measurements at & =20.
In Figs. 6 and 7 the data lie on a relatively smooth FIG. 7. Scaled ring intensity vs reduced time in L-%.
 2110 Y. C. CHQU AND %ALTER I. GQLDBURG 20

slopes are almost equal, with ~~ = eL & = 3. In information that should not be overlooked when
the previous measurements of ring intensity in interpreting the scattering data.
I-W and I -W, a similar increase of 8 with time During the I-W measurements a laser beam
was observed. ' passed obliquely through the sample so that km(f)
In comparing our ring-intensity measurements could be recorded at the same time. Runs were
with the calculations of LBM and KO it has been made at &T& —0. 5, 2. 4, and 4. 5 mK, and over the
implicitly assumed that the I(&, f) is equal, within -
time interval 40 sec ~ t 1000 sec. With a 1-mm
a multiplicative constant, to the scattering func- cell the nucleating fluid always exhibited a granu-
tion S(k, f), which is the quantity of direct theoreti- lar interconnected structure, i. e. , isolated spher-
cal interest. The assumption is justified if (a) ical drops were not seen. This inter connectedness
the attenuation of the transmitted beam is small is expected since the fractional amounts of both
and (b) the restrictions imposed by the first Born phases are equal. We were surprised, however,
approximation are satisfied (these two conditions to find that the visually estimated l(t) grew at a
are not equivalent ). Here we discuss the beam faster rate than did k (t) at all three quench
attenuation problem and defer (b) to'Sec. IV. depths. For example, in the interval 30 sec &t
Our normalization of the scattering intensity by ~ 100 sec and at &7& 2. 4 m— K, l(t) ~ t~&, with
&~(f) gives only an approximate correction for p& —1. 3, whereas the ring diameter collapsed at
multiple scattering, which increases with increas- a rate characterized by p =0. 6 in the range 100
ing t. A proper treatment would take into account sec ~ t & 500 sec. Thus the two schemes for mea-
the shape of the sample and the functional form of suring grain size seemingly give different weights
&(e, f), which is not known, Let us assume that to the droplet-size distribution. Of course the
our multiple-scattering correction is adequate if, visual measurements contain a subjective element
say, the forward beam attenuation is less than that does not afflict the km data. It is likely, how-
10/0. To determine the time interval over which ever, that more sophisticated statistical techni-
our data are properly corrected we list in Table I ques of micrograph analysis would not remove
the actual and dimensionless times t =t„, v = 7„, this problem.
beyond which &g(t)lip(10) & 0. 9. Since the scatter- To examine the influence of sample thickness on
ing is an increasing function of quench depth, a visual grain size experiments on I -W were made
different entry appears for each AT& (see Figs. 6 with both thick (1 mm). and thin (0. 1 mm) cells.
and 7). Comparison of the entries in Table I ahd In the thin cell the interconnected structure broke
the scaled measurements of Figs. 6 and 7 suggest up into spherical droplets at I =20-30 P but the
that the scattering correction is adequate out to simultaneously recorded km(t) showed no evidence
7' =200. of this morphological transition, which occurred
at t =200 sec. The quench depth here was 1.4
D. Visual observa'tions mK. No such transi. tion occurred in thick-cell
In the late stage of phase separation the nucleat- measurements at the same quench depth. How-
ing domains can be seen with a low-power micro- ever, a hydrodynamic instability did intervene,
scope. The visual measurements described be- causing macroscopic fluid flow, as discussed in
low are only semiquantitative, but they do provide detail in Ref. 5. In the thin cell this instability
is presumably suppressed by the walls.
Finally, we note that in both systems the visual
grain size is always much larger than k (f);
TABLE I. Multiple-scattering data. Actual and re- typically l(t)/km(t) varies from 5 to 10. This re-
duced times t„and 7'„, beyond which the unscattered in- sult, however, is not surprising, in neutron scat-
tensity Iz(t) decreased by more than 10%. tering'from liquid argon the first diffraction peak
occurs at k~t'm = 7. 5, where x~ is the maximum
in the radial distribution function.
Solution t sec

I-W 0.7 &930 &120

IV. DISCUSSION
1.8 340 170 A. Domain growth
3.4 110 180
5.8 45 190 The present experiments confirm previous
L-% 0.3 300 60 observations of a crossover from slow to rapid
0.6 140 100 domain growth in critically quenched systems.
1.3 40 110 In the early stage of phase separation, say, 0. 2
1.5 40 140 ~ q+ ~ 1, our results agree with the calculation
1.9 30 170
of Kawasaki and Ohta, who included hydrodynamic
20 PHASE SEPARATION IN CRITICALLY QUENCHED. . . 2lll

effects in the underlying computational scheme Siggia used a deformable spherical-drop approxi-
of I BM. These authors assume that the system mation in his calculation of A and a tube approxi-
is quenched from an initially disordered state mation to estimate &. To compare his results
(&Tq/T, »1), whereas in the experiments &T~/ with ours we take k~ = l = a, where a is his drop
~, =10 . One could, therefore, argue that this radius. We thus ignore the difference in grain
experimental test of the above results is not a size as measured by scattering and by eye, as
conclusive one. On the other hand, earlier experi- well as the nonspherical shape of the domains.
ments showed no initial- state effects. To estimate & and &, as well as the character-
As we will discuss below, our early-stage results istic cross-over length /q2 =—(&'/3B)'I, we follow
' dynes/cm in both sys-
also support the coalescence picture of domain Siggia in taking o = 100&
growth, which gives l =&t tems. Other needed parameters, also assumed
To explain the accelerated rate of domain growth to characterize both I-W and L-W, are $ = )pc
at large ~, a different theoretical approach i.s (in the two-phase region), (c -c,)/c, = a ~, p = 1
required. Recently, Siggia has carefully analyzed g/cm', 7) =0. 24 P, and 4p= 0. g/cm'.lc'"
the problem of late-stage phase separation, taking For critically quenched systems Siggia estimates
into account both diffusion and hydrodynamic flow.
To sort out various competing mechanisms it is
a=(,' k, T-/vn)"'t" '=At" ', n I, (»)
useful to classify experiments as to whether the a = (P o/'I}) t = Bt in II,
quenches are critical or off critical. In the off-
where pis adimensionless constant and A turns out
critical experiments of W-K and Goldburg et al. to be 10 'cm. The expression for A was obtained by
one expects spherical droplets of the minority
Siggia using a coalescence argument. We have
phase to develop in the background of the mother further assumed that a is much larger than the
phase. In the critical case two interpenetrating initial droplet size, which we might estimate to be
phases of near- equal volume should form. Siggia several times $(&T&). A more primitive esti-
emphasizes that in this regime domain growth (c)
mate 17 of domain size in the coalescence re-
should proceed through three stages towhichwe will
gime gave l =(ksT/6m&)' t
frequently refer: I, early stage (including coales- Equation (7} can easily be written in dimension-
cence), t &1 & 5$(l ~At' '); ll, flow stage (l =Bt), less units qm(7), where qm= $/a=k $, and & is
(ksT/cr)'~'» l & (o/g&p)' ', and III, gravitational defined in Eq. (5). Approximating a =ksT/$'
stage, l~ (o/g&p) '. To the left we have noted =(k&T/g)e ", Eq. (V) becomes
q-=( ,', )"'r "'
the mechanism which is responsible for domain
growth, and to the right the range of grain sizes I —h-q -1, (8a)
in which the corresponding mechanism dominates. II = I/6mpr, 1
(8b)
qm q23 qm
Here o is the surface tension (which approaches
where qq, —= tg&p/0&7&&10 'for 0T&& 1 mK in both
zero at the critical point), &p is the density dif-
ference of the two phases, and g is the gravita- L-W and I-W. Since all of our measurements
tional constant. were at larger values of qm, gravitational effects
In stage I, theories of the LBM-KO type are appro- never come into play.
priate. The coalescence picture also applies, The above calculations will now be compared
but only qualitatively because the coalescing re- with our measurements. Consider first the early
gions do not yet have mell-defined compositions stage, where Eq. (8a) predicts that at, say, &=10,
and boundaries. With the same caveat one also q —0. 24. At this time Figs. 4 and 5 shows that
can invoke the evaporation-condensation mecha- qm (10) =qm (10) =0. 5. The same level of agree-
nism of Lifshitz and Slyozov. ' Both mechanisms ment is seen in the unscaled data of Figs. 2 and 3.
are diffusion controlled and give rise to a t For example, in I-W at &T& —1. 8 mK and L-W at
growth rate. &T& 0. 6 mK, k '(100— sec) =10 cm '. Equating
In stage II, grain. growth is controlled by fluid flow. a with km, Eq. (Va) gives km(100 sec) =2. 4&&10
This flow is driven by pressure gradients which cm, whereas measurements give km(100 sec)
arise from the curvature of the surfaces which =4&& 10 cm in both systems. Better agreement
interpenetrate the fluid. Cahn and Moldover and with such a simple calculation of domain growth
Siggia have independently observed that, from would hardly be expected. The coalescence pic-
dimensional arguments alone, E should increase ture is also supported by the previously cited ex-
linearly with time in this regime. Finally, in III, periments of Refs. 2 and 17(c).
the domains become so large that gravitational Next consider the measurements at large ~ where
effects dominate and no simple power-law behavior Eqs. (7b) and (8b) apply. Siggia's calculations
is anticipated. The boundaries between the re- give p=0. 1 but, as he notes, this greatly over-
gimes are, of course, defined only very roughly. estimates a as measured by WK. It is also
2112 Y. C. CHOU A1VD WALTER I. GOLDBURG 20

an overestimate for the present measurements.

L-W
To see this we insert the above estimates for cr,
p, and q into Eq. (8b) giving qm(r) =0. 05/P'r. To
bring this equation into agreement with the late- ii I| II
stage results in Figs. 4 and 5 at, say ~=400,
requires that P =2& 10:
A more detailed theory
is obviously required to calculate the numerical
factor &.

8. Ring intensity E

nE )o
Figures 6 and 7 compare the reduced ring inten- I-W
sity 8~(r) with the theoretical calculations of the
structure factors S(k, f) by LBM and KO. Implicit
in this comparison is the assumed validity of the
first Born approximation. This approximation
should be applicable (i. e. , the Born series con- I

IO OOOO

verges and may be terminated at the first term)

when the phase shift & of light across a domain of FIG. 8. Product of scaled ring intensity and q~ as a
size I (measured relative to the background fluid) function of& for I-%' (O) and L-W ().
is much less than 2m.
To determine when the Born condition is satis-
fied in these experiments we estimate I (&&z), -
equilibrium value &c(~) c . Also we have invoked
where / is the domain size for which & = 1. Near the scaling relation 3v' —2P —Y' =0, ' where v
the critical point the phase shift is adequately and p are, respectively, the correlation length
approximated by o.'=&ne 2&l(t)l&0, where &n is and susceptibility exponents in the two-phase re-
the difference in refractive indices of the pure gion.
components and P is the critical exponent defining In Fig. 8 the data in Figs. 3-6 are combined to
the coexistence curve. Taking P = 0. 33 (Ref. 28) form 8~m vs r for both I-W (open circles) and
and &n(I-W) =0. 06, &n(L-W) =0. 16, one finds L-W (closed circles). Only a few representative
I, ~(9 mK) =90 cm and IL' &(2 mK) =300 cm ~. data points are shown. As predicted, both sets
Since all measurements were made at quench of measurements lie roughly on a straight line of
zero slope. Considering the crudeness of our
depths no larger than these, and since k (f) was
always much larger than above values of 1
we
feel justified in identifying the scattering intensity
l, arguments and the inadequacy of the multiple-scat-
tering correction when ~ & 200, better agreement
with S(k~, f} over most of the measuring time inter- could hardly be expected.
val.
Our finding that 8 ~ t' at large t can be accounted
V. SUMMARY
'

for by the following very crude model. The nu-

cleating fluid is regarded as a random array of I ight scattering was used to study phase sepa-
spheres, each of radius a{t). In this case the ration in critical binary mixtures of isobutyric
Born approximation (often called the Rayleigh- acid-mater and 2, 6-lutidine-water. The quantities
Gans approximation ) gives I(k, f) ~ a, since the measured were k~(t) (ring diameter versus time)
radiated intensity from each sphere is propor- and the ring intensity f(km, t) vs t. The measure-
tional to a, and the number of spheres in the in- ments, which exhibited critical-point scaling,
cident beam is proportional to a '.
Assuming the were fitted to power-law form, k (f) ~ f' "' and
a~f in the late stage gives f(k~, f) CC which f, 8(k, f) ~t (g) . In the early stage of phase separa-
agrees with our measurements. tion the results were in very good agreement with
The above argument implies that f(k, t)a ' the calculations of Kawasaki and Ohta and with a
=I(k, t)k should be a constant over the full mea- coalescence picture, which gives p = 3. In the
suring interval, 10 sec &t & 1000 sec. In reduced late stage p =1, as expected from dimensional
units [8 (r } = measured intensity &" and &&
qm
—km(] and semiquantitative agreements which invoke
=
this equation becomes 8m(~)qm(r) const. indepen- hydrodynamic effects. However, the calculated
dent of ~. To obtain. this result if. is assumed that grain size a(t) exceeds the measured values of
I(k, f) is proportional to &c, where 4 c(t) is the km(t) by roughly 10 .
difference in composition between the background The ring intensity exponent 8 increased more
and nucleating phases, and that &e(t) has reached rapidly with time than expected from the early-
 PHASE SEPARATION IN CRITICALLY, QUENCHED. . . 2113

stage theories of LBM and KO. In both systems M. W. Kim, C. M. KnobIer, S. Krishnamurthy,
the measurements are consistent with 8 (large r) J. S. Langer, T. Ohta, P. N. I'usey, A. J.
=3. Schwartz, E. Siggia. , and N. C. Wong. This work
ACKNOWLEDGMENTS
was supported by the National Science Foundation.
One of us (W. G. ) wishes to acknowledge the kind
We have profited from discussions and corre- hospitality of the Physics Department of Cornell
spondence with K. Binder, J.
Cahn, K. Kawa, saki, University, where this manuscript was written.

~J. W. Cahn and M. R. Moldover (unpublished). edited by M. S. Green (Academic, New York, 1973);
2Ning-Chih Wong and Charles M. Knobler, J.
Chem. (c) W. I. Goldburg and J. S. Huang, in Fluctuations,
Phys. 69, 725 (1978). Instabilities, and Phase Transitions, edited by T. Ris-
There is a wealth of metallurgical literature on spino- te (Plenum, New York, 1975), p. 87.
dal decomposition. A particularly readable introduc- ~8A. J.
Schwartz (private communication); (unpublished).
tion to the subject is the paper by J. W. Cahn, Trans. A word of caution: not all authors have defined q and
Metall. Soc. AIME 242, 166 (1968). 7' identically. In this paper we use the same defini-
J. S. Huang, Walter I. Goldburg, and Allan W. Bjer- tions as in I and KO, but this choice differs from that
kaas, Phys. Rev. Lett. 32, 921 (1974). of LBM. While the numerical results of LBM are
~M. W. Kim, A. J. Schwartz, and W. I. Goldburg, Phys. plotted correctly here, in I (and in KO) the curves la-
Rev. Lett. 41, 657 (1978). beled LBM shouM be shifted to the left by a factor of 2;
J. S. Langer, M. Bar-on, and Harold D. Miller, Phys. see the erratum by Goldburg et al. in J.
Chem. Phys.
Rev. A 11, 1417 (1975). 69, 5214 {1978).
K. Kawasaki and T. Ohta, Prog. Theor. Phys. Jpn. 59, See, for example, H. C. Van de Hulst, Light Scattering
362 (1978). by Small Particles (Wiley, New York, 1957).
Eric D. Siggia, Phys. Rev. A 20, 595 (1979). 2fsee, for example, A. J.Bray and R. F. Chang, Phys.
Walter I. Goldburg, Ching-Hao Shaw, John S. Huang, Rev. A 12, 2594 (1975).
$6
and Michael S. Pilant, J.
Chem. Phys. 68, 484 (1978). J. J.
L. Yarnall, M. Katz, R. G. Wenzel, and S. H.
-digital Equipment Corp. , Maynard, Massachusetts. Koenig, Phys. Rev. A 7, 2130 (1973).
J.
B. Chu, F. Schoenes, and W. I. Kao, J. Am. Chem. See, for example, (a) K. Binder and D. Stauffer, Phys.
Soc. 90, 3042 (1968). Rev. Lett. 33, 1006 (1974); (b) K. Binder, Phys. Rev.
A. Stein, S. J. Davidson, J.
C. Allegra, and G. F. Al- B 15, 4425 (19VV).
len, J. Chem. Phys. 56, 6164 (1972). 2
Walter I. Goldburg, Arthur J.
Schwartz, and Mahn
E. Gulari, A. F. Collings, R. L. Schmidt, and C. J. Won Kim, Prog. Theor. Phys. Suppl. No. 64, 477 (1978).
Pings, J.Chem. Phys. 56, 6169 (1972). I. M. Lifshitz and V. V. Slyozov, J.
Phys. Chem. Sol-
C. S. Bak and W. I. Goldburg, Phys. Rev. Lett. 23, ids 19, 35 (1961).
1218 (1969). In Ref. 8 the symbol D denotes the diffusion constant of
W. I. Goldburg and P. ¹ Pusey, J. Phys. 33, C1-105 a sphere of radius a rather than the composition dif-
-
{1ev1). fusivity as in this paper.
~6A. W. Loven and O. K. Rice, Trans. Faraday Soc. 59, VYu. N. Barabanenkov, Yu. A. Kravtsov, S. M. Rytov,
2v23 (1e63). and V. I. Tamarskii, Usp. Fiz. Nauk 102, 3 (1970) [Sov.
~7(a) H. E. Stanley, Introduction to Phase Transition and Phys. Usp. 13, 551 {1971)].
Critical Phenomena (OxfordUniversity, London, 1971); Sandra C. Gteer, Acc. of Chem. Res. 11, 427 (1978).
2
(b) Critical Phenomena, Proceedings of the Enrico Handbook of Chemistry and Physics, 46th ed. (Chemi-
Fermi International School of Physics, Course LI, cal Rubber, Cleveland, 1965).