J . Phys. Chem.
1989, 93, 6897-6901
'0.
-..... .....
*.
Ob 0.b5 O.;O 0.\5 0.20 OL5 O!O
B arise, it is worthwhile inquiring what circumstances in a kinetic
Figure 7. Configurational entropy per lattice site, divided by ke, for the
case D = 3. 0 = 0.01.
properties. Furthermore the results indicate a shift of the smeared
transition to larger 0 (lower temperature): the heat capacity
maximum in Figure 4 occurs at 0 z 0.416.
Figures 6 and 7 respectively present heat capacity and entropy
curves versus 0 for another case, namely D = 3, B = 0.01. The
qualitative pattern agrees with the preceding two-dimensional
example, with the B = 0 sharp transition now smeared out and
shifted to larger ,13. The heat capacity maximum in this case occurs
=
at 0 0.182.
The entropy curves in Figures 5 and 7 are relevant to the "ideal
glass transition" controversy. Both display regions of rapid decline
toward zero. If kinetic sluggishness at lower temperatures were
to make observation of the remainder of the heat capacity curves
impossible, it would seem natural to extrapolate the entropy curves
to zero at a finite 0 (approximately 0.44 and 0.19 for D = 2 and
D = 3 ) . However, we see that the entropy curves smoothly turn
over and approach zero only as 0 approaches infinity. Conse-
quently, tiling model behavior with frustration present seems to
be inconsistent with the existence of an ideal glass transition.
Nevertheless, the presence of well-developed heat capacity peaks
. -.Darameters
in the tiling models with aDDroDriatelv chosen couDling
make the& useful in atiiining a qualitative understanding of
enhanced configurational heat capacities in fragile glass formers."
VI. Discussion
Time-deoendent DroDerties of the tiling models reauire sDeci-
fication of {he basic iraisition rates betwe& configuraiions. Even
after accounting for the constraints on these rates mentioned in
section 11, it appears that a wide range of relaxation behaviors
Shape Fluctuations in Ionic Micelles
Kyoko Watanabe and Michael L. Klein*
Department of Chemistry, University of Pennsylvania, Philadelphia, Pennsylvania 19104-6323
(Received: February 3, 1989)
Molecular dynamics calculations are reported for a microscopic model of a sodium octanoate micelle in aqueous solution.
The micelle, whose equilibrium structure is found to be a prolate spheroid, undergoes shape fluctuations with a characteristic
time scale of order 30 ps. The results are discussed in the light of recent theoretical and experimental studies on related
systems.
introduction
The rich variety of phase behavior exhibited by ionic surfactant
solutions' can ultimately be traced back to the competing inter-
actions between the constituent species: the repulsive interactions
(1) Ekwall, p. I,, Adoances in ~ i ~ ,-,ystuls;
~ i dB ~ G.H,,
~ Ed.;~
demic: New York, 1975; Vol. 1, p 1.
0022-3654/89/2093-6897$01.50/0
6897
is possible for a given equilibrium behavior, since the rates are
independently assigned.
A well-known relation due to Adam and Gibbs20describes the
temperature (7') dependence of average relaxation time ( 7 )in
glasses in terms of the configurational entropy S,:
(7)= 70 exp[A/TSc(T)] (6.1)
Here T~ and A are temperature-independent positive constants,
which can be expected to vary from substance to substance. While
the Adam-Gibbs relation appears to describe the glass-transi-
tion-region behavior of many substances quite satisfactorily?' both
experimentalzz and theoretical* exceptions have been noted.
In order to understand in principle how such exceptions could
tiling model should produce substantial violations of the Adam-
Gibbs relation. One qualifying scenario has
+
x / e = j~ 1)
where j is a large integer, so that the ground state consists of a
+
degenerate mixture of size-j and size-(j 1) tiles. In this event
lim Sc(7') = Sc(0) > 0
(6.3)
T-0
as a result of the degeneracy. At very low (but positive) tem-
peratures the system would be tiled almost exclusively with these
two large sizes, and relaxation rates would be controlled just by
the basic transition rates K(C-4') within this special tiling set.
If these rates have a common Arrhenius form, then (7)in this
very low temperature regime can be expected to adopt the
Adam-Gibbs form (5.1) with appropriate T~ and A values. At
somewhat higher temperatures the mean tile size will be sub-
stantially smaller, and concentrations pj and will be negligibly
small. Consequently a different set of basic transition rates K
will be involved that in principle can be assigned independently.
In particular these rates could be given Arrhenius form with a
very different activation energy. If the result for (7)were forced
into Adam-Gibbs form the constant A so obtained would differ
from that appropriate to the very low temperature regime.
Acknowledgment. F.H.S. acknowledgesbeneficial conversations
with Dr. S. Bhattacharjee concerning ground-state structures of
frustrated tiling models. We are also grateful to Dr. J. McKenna
for discussions involving material in section IV.
(20) Adam, G.; Gibbs, J. H.J . Chem. Phys. 1965, 43, 139.
(21) Angell, C. A.; Tucker, J. C. J . Phys. Chem. 1974, 78, 278.
(22) Laughlin, W. T.; Uhlmann, D. R. J . Phys. Chem. 1972, 76, 2317.
between polar head groups, the attractive interactions between
the surfactant tails (hydrophobic effect), and the solvation forces
associated with the hydrophilic head groups. Considerable the-
oretical progress has recently been made in understanding the
overall behavior of micelles and related microemulsion systems
~ ,in terms of mean-field theories and simplified or idealized mod-
els.2-'z However, the bulk of experimental and theoretical work
0 1989 American Chemical Society
6898 The Journal of Physical Chemistry, Vol. 93, No. 19, 1989 Watanabe and Klein

has centered on microemulsion Here, the equilibrium TABLE I: Molecular Dynamics Results for a Sodium Octanoate
structures and phase diagrams have been rationalized in terms Micelle in Aqueous Solution
of macroscopic quantities such as surface tension and the bending
time. (PE), (P),
energy of the i n t e r f a ~ e .Indeed,
~ the theory has been extended PS kJ/mol kbar (RHG),A (&),A (ntrana)
beyond equilibrium properties to include hydrodynamic fluctua-
0-8 -57.25 -0.12 10.2 5.8 3.4
tions of an ensemble of microemulsion droplets that take place 8-16 -57.21 0.00 10.4 5.8 3.6
subject to the constraint that the total volume of surfactant and 16-24 -57.23 -0.15 10.2 5.8 3.7
hence the total droplet area are c o n ~ e r v e d . ~Relatively little 24-32 -57.28 0.05 10.3 5.9 3.6
theoretical work has been devoted to ionic surfactant micelles.I0 32-40 -57.24 -0.1 1 10.6 5.9 3.7
Recent elegant neutron spin-echo experiments7 have demon- 40-48 -57.25 -0.12 10.8 6.2 3.9
strated the existence of such shape fluctuations in microemulsion 48-56 -57.16 -0.06 10.7 6.4 3.8
droplets, in this case reverse micelles, of radius ro = 27-70 A, with 56-64 -57.26 -0.05 10.6 6.0 3.9
a labeled layer of sodium bis(2-ethylhexy1)sulfosuccinate sur- 64-72 -57.29 -0.12 10.4 5.9 3.9
72-80 -57.13 -0.05 10.8 5.9 4.1
factant. The wave vector (Q) dependence of the scattering
80-88 -57.25 -0.11 10.6 6.2 3.8
spectrum from these droplets, whose internal (water rich) phase 88-96 -57.25 -0.10 10.4 5.9 3.6
had the same neutron-scattering-length density as the external
continuous phase (decane), exhibited a sharp peak in the relaxation of a molecular dynamics study, based on a completely microscopic
frequency spectrum at a wave vector Q,. Neutron spin-echo model, of a single sodium octanoate micelle in aqueous solution.
measurements yield an effective diffusion constant for the droplets From this study, we are able to demonstrate explicitly the nature
which obtains when Q -
that contains two contributions: the real center-of-mass diffusion
0 plus a contribution from shape
fluctuations that give rise to the peak at Q,. In the case of the
of fluctuations that can arise in small ionic micelles. Our principal
finding is that such micelles undergo dramatic fluctuations in
shape, a result that is in excellent agreement with predictions of
experiment in ref 7, one might expect the shape change to be a considerably simpler model introduced by Pratt and co-workers."
dominated by the 1 = 2 spherical harmonic. However, because The presence of these fluctuations calls into question the utility
microemulsion droplets can exchange surfactant through collisions, of model calculations that employ constrained head groups and
the scattering experiment will also sample an ensemble distribution that ignore the solvent.I2 Our results also reiterate the view that
of droplet radii which, in effect, gives a nonvanishing 1 = 0 the current methods of analyzing small-angle neutron scattering
contribution, even for incompressible droplet^.^ (SANS) data from micelles are incomplete" because they in-
Fluctuation modes in droplets are customarily regarded as being variably ignore shape f l u c t u a t i o n ~ . l ~ ~ ~ ~
driven by surface tension or so-called capillary waves? Theoretical
arguments lead to the conclusion that fluctuation modes, driven The Model
by surface tension y, will have a relaxation rate T.,-I = y / v r o , The interaction potentials and the basic model we employ for
whereas for modes driven by the bending or curvature modulus the simulation are very similar to those that we used recently to
K,, the rate is given by T;' = K,/vr:, where v is the viscosity.' study structural aspects of the sodium octanoate mi~e1le.l~In
By measuring the variation in peak position as a function of droplet particular, we employ effective pair potentials for the water,17
radius and the ratio of peak height to line width (diffusion con- sodium ions,I8 and the polar carboxylate head groups.19 The
stant), it was concluded that the neutron spin-echo data are methylene (>CH2) and methyl ( X H J groups were treated as
consistent with a mode driven by bending elasticity and not surface fused pseudoatoms with no explicit treatment of the hydrogen
tension.' atoms. These groups interacted via pair potentialsI9 that yield
The above considerations for swollen microemulsion droplets good thermodynamic properties for liquid butane.20 All the
need not apply to ionic surfactant micelles. Indeed, a recent cross-interactions were constructed by using combining rules for
articlelo has been quite critical of the above approach that neglects the Lennard-Jones parameters. Hydrogen and oxygen atoms of
electrostatic interactions. The picture that emerged from this the SPC water model carry fractional charges 0.41 and -0.82 e,
theoretical studylo was that ionic micelles could adopt a wide range respectively, and the sodium ions have their full charge. The
of structures, including prolate and oblate spheroids as well as carbon and oxygen atoms of the head groups carry charges 0.4
quasi-spherical. Collective dynamical shape fluctuations were and -0.7 e, respectively. The relative positions of the head-group
postulated with time scales T 1 100 ps. The theory on which these atoms plus the first methylene grou were held rigid by con-
conclusions were based was, of necessity, highly approximate due
to the obvious complexity of the problem. It is natural therefore
K
straining the C-0 distance at 1.26 and the 0-C-0 and 0-
C-CH, angles at 122O and 119O, respectively. The hydrocarbon
to seek an alternative complementary approach to the problem chains were treated as flexible with a simple C-C-C-C dihedral
of shape fluctuations in micelles. One such approach is offered angle torsional potential and a C-C-C angle bending force con-
by computer The present article reports the results stant.,' There are five dihedral angles for each hydrocarbon tail
so that, in an all-trans configuration of an octanoate ion, the
(2) De Gennes, P. G.; Taupin, C. J. Phys. Chem. 1982, 86, 2294.
quantity ntrans = 5. The C-C distance was held fixed at 1.53 h;
(3) Safran, S.A. J. Chem. Phys. 1983,78,2073. Safran, S.A.; Turkevich, by using the method of constraint^.^^*^^
L. A. Phys. Rev. Lett. 1983, 1930.
(4) Milner, S. T.; Safran, S.A. Phys. Rev. A 1987, 36, 4371. Details of the Simulation
(5) Ben-Shaul, A.; Szleifer, I.; Gelbart, W. M. Proc. Nutl. Acad. Sci. The molecular dynamics cell, which contained 718 water
U.S.A. 1984, 81, 4601: J . Chem. Phys. 1985, 83, 3597, 3612.
(6) Ben-Shaul, A.; Gelbart, W. M. Annu. Reo. Phys. Chem. 1985,36, 179. molecules plus 15 sodium and octanoate ions, was a periodically
Szleifer, I.; Ben-Shaul, A.; Gelbart, W. M. J. Chem. Phys. 1986, 85, 5345;
1987, 86, 7094.
(7) Huang, J. S.; Milner, S. T.; Farago, B.; Richter, D. Phys. Rev. Left. (15) Hayter, J. B.; Zemb, T. Chem. Phys. Lett. 1982, 93, 91.
1987, 59, 2600. (16) Sheu, E. Y.; Chen, S.-H.;Huang, J. S. Preprint, 1988.
(8) Brown, D.; Clarke, J. H. R. J. Phys. Chem. 1988, 92, 2881. (17) Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.; Her-
(9) Lunggren, S.;Eriksson, J. C. J . Chem. SOC.,Faruduy Trans. 2 1984, mans, J. In Intermolecular Forces; Pullman, B., Ed.; Reidel: Dordrecht, 1981.
80, 489. (18) Chandrsekhar, J.; Spellmeyer, D. C.; Jorgensen, W. L. J. Am. Chem.
(10) Halle, B.; Landgren, M.; Jonsson, B. J . Phys. (Les Ulis, Fr.) 1988, SOC.1984, 106, 903.
49, 1235. (19) Jorgensen, W. L. J . Am. Chem. SOC.1981, 103, 335.
( 1 1) Owenson, B.; Pratt, L. R. J . Phys. Chem. 1984,88, 2905. Pratt, L. (20) Jorgensen, W. L. J. Am. Chem. SOC.1981, 103, 4721.
R.; Owenson, B.; Sun, Z. Adv. Colloid Interface Sci. 1986, 26, 69. (21) van der Ploeg, P.; Berendsen, H. J. C. J. Chem. Phys. 1982,76, 3271.
(12) Haile. J. M.; O'Connell, J. P. J. Phys. Chem. 1984,88, 6363. Woods, (22) Ryckaert, J. P.; Bellemans, A. Chem. Phys. Lett. 1975, 30, 123.
M. C.; Haile, J. M.; OConnell, J. P. J . Phys. Chem. 1986, 90, 1875. Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. J . Comput. Phys. 1977.23,
(13) Jonsson, B.; Edholm, 0.;Teleman, 0. J . Chem. Phys. 1986,85,2259. 327.
(14) Watanabe, K.; Ferrario, M.; Klein, M. L. J . Phys. Chem. 1988, 92, (23) Allen, M. P.; Tildesley, D. Computer Simulation of Liquids; Clar-
819. endon: Oxford, 1987.
Shape Fluctuations in Ionic Micelles
0.04 I I
c-
w
0.03 -
h
0
0.02 -
0.01 -
A,
0.00
0 2 4 6 8 10 12 14 16
TIA
Figure 1. Density distributions for carbon atoms (squares), water mol-
ecules (triangles), and counterions (circles) measured with respect to the
micelle center of mass.
replicated truncated octahedron24with a volume V = 0.5 X (37
= 25326.5 A3.The simulation was carried out at constant
volume and constant temperature (298 K) using an Ewald method
to handle the long-range electrostatic interactions.23 The equations
of motion were integrated by using a Verlet algorithm and a time
step of 2.5 fs. As in our previous work, the simulation was carried
out in three stages.I4 First, the micelle core, Le., the tails of the
15 octanoate ions, was equilibrated by carrying out a long mo-
lecular dynamics run with the head-group carbon atoms con-
strained to lie on the surface of a sphere whose radius was
gradually decreased from 15 8,to the final value of 10.8 8,. Next,
the micelle core was inserted into an equilibrated box of water
along with the 15 counterions, and a run of 35 ps was then carried
out with the head groups still constrained to lie on the surface
of a sphere. This was followed by a further equilibration run of
23 ps, with the head-group constraint removed. Finally, a pro-
duction run of 96 ps was used to generate the structural and
dynamical results that are described below and listed in Table
I.
Structural Results
The variation of the potential energy ( P E ) and pressure ( p )
of the micelle-water system, throughout the 96-ps production run,
is given in Table I. For the purpose of analysis, the total run has
been partitioned into 12 equal segments for which subaverages
are recorded. (Each segment is actually a convenient overnight
run on our IBM 3090/200 VF computer.) The overall structure
of the micelle can be characterized through various probability
distribution functions. Figure 1 shows the density distribution
of carbon atoms in the micelle core, the sodium counterions, and
the solvent water with respect to the center of mass of the micelle,
calculated during the last 64 ps of the run. The extent of water
penetration into the core region of the micelle is somewhat greater
than we found in our previous calculation on this system. On the
average, the number of water molecules found within a 4-8, radius
from the methylene and methyl groups is, from head to tail, 5.0,
4.6, 3.9, 3.7, 3.3, 3.2, and 3.4. The total number of water molecules
within 4 8, of any methyl or methylene group in the micelle is
127. Thus, on average, there are 8.5 water molecules in contact
with the hydrophobic tail of each monomer comprising the micelle.
This number is significantly larger than the result found in our
previous simulation which gave 4.6 water molecules in contact
with each hydrocarbon chain. In the latter case, the average
pressure was actually quite negative, being (p) = -1.1 kbar
compared with the present value of (p) = -0.07 kbar. In the
present near-zero-pressure calculation, unlike in the previous one,
the carbon atom density in the micelle core region is close to that
of a liquid hydrocarbon. As found previ~usly,'~*'~ the water content
of the micelle core is large and disagrees with a recent analysis
of new SANS data. This new work estimates the number of water
(24) Adams, D. J. Chem. Phys. Lett 1979,62, 329.
The Journal of Physical Chemistry, Vol. 93, No. 29, 1989 6899
0.4
h
0.3
r
*'&
9.
0.2
0.1
0.0
0 2 4 6 8 10 12 14 16
TIA
Figure 2. Probability distributions for carbon atoms measured with
respect to the micelle center of mass. The curves labeled by squares, open
circles, triangles, and dots refer to carbon atoms 1, 3, 5 , and 8 of the
hydrocarbon chains, starting at the head group, respectively.
molecules in the core to be less than one per monomer.16
In previous molecular dynamics calculations, the sodium ions
were seen to localize in the region of the head groups (the Stern
r e g i ~ n ) . ' ~In
? ' ~the present calculation, about four of the ions are
actually in contact with the head groups. The net charge on the
micelle is thus about -1 1 e, which is a little larger than our previous
value of -9 e.I4 The value derived from a fit to the original SANS
data is -10 f 2 e.Is
The distribution functions for carbon atoms at positions 1, 3,
5, and 8 (counting from the head-group carbon atom along the
chain backbone) are shown in Figure 2. The average distances
of the carbon atoms, C1 to C8, from the micelle center of mass
are 10.6 f 1.6,9.4 f 1.7, 8.6 f 1.7, 7.6 f 1.7, 7.1 f 1.7, 6.4 f
* *
1.8, 6.2 1.9, and 6.0 2.2 8,. Overall, the carbon atom dis-
tributions are rather similar to those reported previously by other
but the peak positions occur at distances that are
12-19% smaller than in our previous sim~1ation.l~ The main point
to note in the carbon atom distributions is that the terminal methyl
group distribution is very broad.2s*26 The mean location, with
respect to the micelle center of mass, of the terminal methyl group
( R T )is given in Table I for each segment of the run. This value
of ( R T )appears to be rather constant as is the analogous quantity
for the carbon atom of the head groups (RHG).The values of
(nb,,) listed in Table I indicate that, on average, there is at least
one gauche defect in each chain.
Micelle Shape
A natural interpretation of the data in Table I is that the micelle
is indeed stable for the duration of the simulation. Accordingly,
it seems worthwhile to examine the shape of the micelle in a more
quantitative fashion, particularly in view of the remarks in the
Introduction.lo Figure 3a shows the variation of R, the micelle
volume, throughout the run, where the quantity R = (4/3)ir( R H G ~ )
is defined in terms of the head-group positions at each time step.
The volume defined in this fashion varies between extremes of
about 4200 and about 6000 A3with a mean value of roughly 5200
A3over the last 64 ps of the run. An analogous calculation for
the surface area yields a mean value for each head group of about
94 A2. As mentioned in the Introduction, micelles need not
necessarily be spherical. Indeed, the calculations of Pratt and
co-workers on an idealized model suggest dramatic departures
from sphericity." We have examined representative micelle
configurations generated in the molecular dynamics run in some
detail using an IBM 5080 work station. The inescapable con-
clusion from these visualizations is not only that the micelle is
undergoing large volume fluctuations but also that it is very
(25) Cabane, B.; Duplessix, R.; Zemb, T. J . Phys. (La.Ulis, Fr.) 1985,
46, 2161.
(26) Dill, K. A.; Naghizadeh, J.; Marqusee, J. A. Annu. Reu. Phys. Chem.
1988, 39, 425.
6900 The Journal of Physical Chemistry, Vol. 93, No. 19, 1989 Watanabe and Klein
20

1.6 10

& 1.4 5 0
2.
1.2 -10

1 -20

6000 10
n
n
in 5 0 96 ps
4000

-10

) , I , 1 / 1 ,
2000 1 ' 1 ' 1 ' 1 '
0 20 40 60 80 1 I -20
-,
-20 -10 10 20
tips
Figure 3. (a) Time evolution of the micelle volume, defined in terms of
x ;A)
the head-group carbon atom positions (see text). (b) Time evolution of Figure 4. Instantaneous view of the micelle after 32 and 96 ps. The head
the ratios of principal moments of inertia of the micelle. groups are shown as circles and the tails as squares. The simulation cell
boundaries are indicated by lines and the sodium ions by crosses. Views
from orthogonal directions (not shown) confirm that at 32 ps the micelle
TABLE II: Spherical Harmonic Expansion Coefficients is spherical.
Characterizing the Micelle Shape
set 1 set 2 Since there are relatively few monomers in the micelle, char-
4
3 (0.8) (s(a,,)2)1'2 (a,$) (6(0,@)2)1'*
acterization in terms of the spherical harmonic expansion yields
only qualitative information. We therefore also used an alternative
00 36.85 0.79 36.85 0.79 scheme based on variations in the moment of inertia tensor of the
10 0.00 0.00
11 0.00 0.00 micelle throughout the run. We focused on the moment of inertia
20 7.40 3.17 2.73 6.26 tensor because it is dominated by the contributions from the head
21 -0.13 3.40 0.59 3.32 groups, and hence this quantity provides a simple way to char-
22 -0.15 2.58 0.98 3.44 acterize the micelle shape. The values of the moments Zl, Z2, and
30 -2.02 8.05 -0.78 7.22 13,corresponding to the principal axes, leave no doubt that, on
31 -1.68 5.83 1.49 5.06 average, the micelle is roughly a prolate spheroid, with Z, C Z2
32 1.07 5.48 -0.38 5.44 = 13.The variation throughout the run of the ratios of the principal
33 0.75 4.32 -0.62 5.29 moments of inertial R = Z2/11 or Z3/Z1 is shown in Figure 3b. The
40 -1.94 8.84 1.22 1.43 values of the two ratios vary quite widely but, nevertheless, track
41 0.43 5.28 -1.72 5.91
42 -1.15 5.37 1.33 5.24 each other quite closely as the run proceeds. At 32 ps, R = 1
43 1.38 5.41 0.95 5.90 which implies that the micelle is spherical. Figure 4 shows a view
44 1.75 5.40 -1.05 5.80 of the micelle at exactly this point in the trajectory. As expected,
the picture confirms that the micelle shape is quite spherical. The
nonspherical. In order to better characterize the micelle shape, mean value of R over the latter part of the run is about 1.33, which
we first attempted a spherical harmonic expansion of the radial indicates a significant spheroidal distortion. A second configu-
function R(t,8,$) which defines the surface of the head-group ration shown in Figure 4, which is the last one of the run, clearly
carbon atoms:* shows that at 96 ps the micelle is nonspherical. The fact that,
on average, we observe substantial deviation from a spherical
R(t9894) = Ca/m(t) Y/m(e*4) shape, even for such a small micelle, is in excellent accord with
/m
the results of Pratt and co-workers." However, the actual nu-
The expansion coefficients aIm(t)were calculated from the equation merical agreement for the predicted shape of the octanoate system
is only semiquantitative.
a/m(t) = ( 4 ~ / n ? C R ( t , @ i , 4Yi )d e i , A )
1
Shape Fluctuations
where i runs over the N = 15 head-group carbon atoms at positions The apparent success of the analysis used to interpret the
R(t,8,&) with respect to the center of mass of these atoms. Table structural features of the octanoate micelle makes it worthwhile
I1 lists two sets of values for the quantities (al,(t)) and their rms to explore the time evolution of the system in more detail. In the
fluctuations; the brackets denote an average over the run. The Introduction, we mentioned some current theories of shape
first set of coefficients are calculated by defining 8 with respect fluctuations of microemulsion droplets4v9and micelles10and the
to the axis corresponding to the smallest principal moment of recent elegant neutron spin-echo experiments that confirmed the
inertia. The second choice of coordinate frame sets the z axis along presence of shape fluctuations in the former case.' The results
the principal axis with the largest a20coefficient. In both cases, shown in Figures 3 and 4 already strongly suggest that periodic
( a20)gives the largest nonspherical contribution which suggests volume and shape fluctuations are occurring. In order to highlight
a prolate micelle shape. However, contributions from higher order this phenomenon more clearly, we have examined two different
spherical harmonics are appreciable and indicate the complexity autocorrelation functions, Cn(t)= (6Q(t) sQ(O))/(sQ(O) sQ(0))
of the actual micelle shape. and CR(t)= ( 6 % ( t ) b R ( 0 ) ) / ( 6 2 ( 06R(O)),
) where the brackets
Shape Fluctuations in Ionic Micelles
1 Discussion
0 on the time scale that is appropriate to micelle diffusion, rotation,
n parameters, especially the use of combining rules for water-
0 hydrocarbon interactions and the neglect of hydrogen atoms in
es
v
n
-1
0 \
0 posing a constrained starting configuration and observing the
-1
0 20 40 60 80
t/ps
Figure 5. (a) Autocorrelation function of fluctuations in the ratios of
moments of inertia for the principal axes. The bold and dashed curves
have the same meaning as those in Figure 3b (see text). (b) Autocor-
relation function of fluctuations in the micelle volume (see text).
denote an average over the run and time origins and 6n(t) = n(t)
- ( Q ( t ) ) .Figure 5 gives rudimentary information on the shape
and volume fluctuations that are occurring in an octanoate micelle
in aqueous solution. Although the statistics on the time correlation
functions are inevitably rather poor, Figure Sa suggests that the
time scale for the shape fluctuations is T R = 30 ps. On the other
hand, it is evident from Figure 5b that the time scale for the
volume fluctuations is considerably longer and is of the same order
of magnitude as the run length, namely, 70 ir 100 ps.
One of the questions that naturally comes to mind when one
attempts to characterize the micelle shape is the overall translation
and rotational motion of the entity. The diffusion coefficient for
the micelle center of mass was calculated from the trajectory and
was estimated to be 0.3, in units of cmz s-', which is to be
compared to the values 3.3 and 1.2 for water molecules and sodium
ions, respectively, in the same units. The diffusion coefficient for
individual monomers is about 0.6 in these units. Diffusion of water
molecules in the hydration shells of the hydrophobic methyl and
methylene groups and solvent associated with the head groups and
sodium ions is somewhat slowed down compared to bulk water,
the respective diffusion coefficients being 2.9 and 2.2 in the
above-mentioned units. Finally, the diffusion of monomers on
the micelle surface was examined by calculating the autocorre-
lation function (u(t).u(O)),where u(t) is a unit vector in the
direction of the radial vector connecting the center of mass to the
head-group carbon atom. The characteristic time for the decay
of this function was found to be in the region of 500 ps, assuming
an exponential decay.
The Journal of Physical Chemistry, Vol. 93, No. 19, 1989 6901
A number of questions immediately come to mind concerning
the observations reported above. For example, to what extent is
the calculation influenced by the periodic boundaries and the fact
that there is only one micelle? Moreover, the run is rather short
and intermicelle collisions, etc. Also, there are the usual concerns
about the dependence of the results on the choice of potential
the construction of the model. Also, to what extent are the
fluctuations we observed in Figure 5 a consequence of our initial
conditions? It is obviously difficult to rule out the influence of
any of these factors. However, arguments similar to those used
to rationalize Onsager's regression hypothesis suggest that im-
resulting relaxation toward equilibrium is not an unreasonable
way to probe the spontaneous fluctuations of the system of interest.
This is in effect what we have done. In so doing, we have explicitly
demonstrated that shape fluctuations and volume fluctuations
occur on a time scale less than that of the exchange of octanoate
particles. That is, a run of 100 ps is apparently long enough to
sample shape changes that involve considerable monomer motion
but no actual loss of monomers from the micelle. The actual time
scales we seem to find have implications for larger more commonly
studied micelles such as those formed by lithium and sodium
dodecyl sulfate. Here, the time required to follow shape fluctu-
ations will likely be much longer than r0 = 30 ps. Indeed, a study
of such systems will likely present a real challenge to simulations.
The situation may be saved, however, because the magnitude of
the fluctuations in a larger micelle might be smaller than in the
case considered herein.
In summary, we have used molecular dynamics calculations
to demonstrate the nature of fluctuations that can arise in a small
ionic surfactant micelle. We find that a sodium octanoate micelle
in aqueous solution is best described as being a prolate spheroid
that undergoes shape fluctuations on a time scale of 30 ps. It
remains to be seen whether any of these findings can be carried
over to larger micelles of wider interest. Our results appear to
be at odds with a recent analysis of new SANS data on sodium
octanoate.I6 Here, the usual approximations were invoked to
analyze the data.2s These include an assumption of sphericity
and total neglect of shape fluctuations plus effects of polydispersity.
With these rather drastic assumptions, it is perhaps not too
surprising that theory" and experimentI6 are found to differ
somewhat on the nature of ionic micelles in aqueous solution.
Acknowledgment. We have benefited from discussions with
John Huang, Sam Safran, John Shelley, and Eric Sheu. The
research described herein was supported by the National Institutes
of Health under Grant GM-40712-02 and is dedicated to Bob
Zwanzig on the occasion of his 60th birthday. The calculations
were carried out on an IBM 3090-200/VF supported by the School
of Arts and Sciences at Penn and the National Science Foundation
under CHE-8815130.
Registry No. Sodium octanoate, 1984-06-1.