Self-diffusion of spherical Brownian particles with hard-core interaction

Self-diffusion of spherical Brownian particles with hard-core interaction

1llA (1982) 181-199 Physica North-Holland Publishing Co. SELF-DIFFUSION OF SPHERICAL BROWNIAN WITH HARD-CORE INTERACTION Sarwat fir Physik, Fak...

898KB Sizes 9 Downloads 86 Views

1llA (1982) 181-199

Physica

North-Holland

Publishing

Co.

SELF-DIFFUSION OF SPHERICAL BROWNIAN WITH HARD-CORE INTERACTION Sarwat

fir Physik,

Fakulttit

HANNA,

Walter

Llniversittit

Received

The

N-particle

particles

interacting

self-diffusion constant. the

Oseen

Smoluchowski

coefficient

is found

The hydrodynamic approximation,

hydrodynamic

equation

through hard-core

interaction

HESS

Konstanz,

and Rudolf

D7750

for a system

of spherical

Brownian

to first order in the volume concentration where

is then taken into account

D, = Do(l -0.091#1) is found, due to Felderhof

Fed. Rep. Germany

9 July 1981

to be given by D, = Do(l - 24).

interaction

KLEIN

Konstanz,

is solved exactly

potentials

PARTICLES

gives D, = Do(

whereas

I-

in a perturbative the

4. The

DO is the free diffusion improved

way. form

Using for

the

1.894).

1. Introduction

In recent years there has been considerable interest in the concentration dependence of the collective (or mutual) diffusion coefficient D, and the self-diffusion coefficient D, for macroparticles suspended in fluids. A number of experiments by means of more conventional methods’) and by photon correlation spectroscopy*) have been performed which have led to considerable theoretical work in this field. From the theoretical point of view there is, for general interactions among the macroparticles, no reason why D, and D, should be equal, even to lowest order in the volume concentration C#Jof macroparticles. The diffusion coefficients D, and D, are defined by the hydrodynamic poles of the autocorrelation functions for the total concentration fluctuations 6c(k, t) and for one-particle fluctuations 6Ci(k, t), respectively. These two functions describe different physical situations and will therefore generally lead to different results for D, and D,. In this paper we will concentrate on the lowest-order concentration dependence of the self-diffusion coefficient D, for a system of particles interacting via a (repulsive) hard-core interaction and in addition through the so-called hydrodynamic interaction. The latter arises because of the velocity field which is produced in the solvent by the motion of the tagged particle and felt by neighboring particles. In the discussion we will come back to a comparison of D, with the collective diffusion coefficient D,. 0378-4371/82/0000-0000/$02.75

@ 1982 North-Holland

S. HANNA

182

Earlier starting

theoretical

from irreversible

equations.

Writing

free macroparticle, from the N-particle a hierarchy truncated a result

approaches

include

thermodynamics

D, = DO(l - CX~) where

et al

a calculation and using

of D, by Batchelor’) two-body

Do is the diffusion

hydrodynamic coefficient

of a

Batchelor obtains cy = 1.83. A different approach starts Smoluchowski equation. From it Felderhof4) has derived

of equations

for

s-particle

distribution

functions

which

for systems of low concentration. Felderhof has derived for the collective diffusion coefficient. This approach

can

be

in this way was later

extended by Jones’) to systems of two different kinds of particles. One can also apply the Mori-Zwanzig projection operator formalism to the N-particle Smoluchowski equation’.‘), obtaining in this way exact memory equations for the collective and the self-diffusion propagator. It can be shown that the memory function in the case of collective diffusion vanishes in the hydrodynamic limit faster than k*, so that memory effects are not important for the collective diffusion constant D,. In contrast, this is not the case for selfdiffusion. Since Felderhof’s method is essentially a mean-field theory without memory effects, his conclusions concerning D, are in agreement with those obtained by the Mori-Zwanzig method, whereas the extension to the selfdiffusion problem not included. The main the memory

leads to results

which disagree,

because

memory

effects

are

problem of the Mori-Zwanzig approach lies in the evaluation of functions. In an earlier paper’), concerning charged spherical

particles, a mode-mode coupling approximation for systems of small volume concentrations proximation’). It should used different approaches of course also dependent interaction.

was used. Another possibility is a weak-coupling ap-

also be noted that different authors have not only and various approximations, but that the results are on the particular form used for the hydrodynamic

In this paper we concentrate on the lowest-order concentration dependence of D,. We first solve the two-body problem with hard-core interactions exactly. Using this result we obtain without approximations the solution of the N-particle Smoluchowski equation with hard-core interactions to linear order in the volume concentration 4. This gives an expression for D, for hard-core systems, which includes the effect of the self-memory function to first order in 4 exactly. The hydrodynamic interaction is then treated by a perturbative method to first order in the strength of this interaction. Two different models, the Oseen tensor and Felderhof’s form for the hydrodynamic interaction”) are treated explicitly and are shown to lead to rather different results. Finally, we discuss our findings for D, in relation to earlier ones.

SELF-DIFFUSION

2. Exact solution

OF BROWNIAN

of the two-particle

183

PARTICLES

Smoluchowski

equation

for a hard-core

system The

Smoluchowski

hydrodynamic

equation

interaction,

for

two

interacting

particles,

but

without

has the form

$p(r,,t-2, t) = hl2p(rl,i-2, t), (2.1) ~,2=Do-$($.+P

I

~)+D,,$(&+p~),

I

where p(rl, r2, t) is the probability to find at time t the two particles at r, and r2, respectively, and U(r,,) denotes the two-particle potential, for which we assume a hard-core form 00 for r12 < d, for r12>d,

u(r’2)=(0

(2.2)

d is the particle diameter and r12 the relative distance t-r2 = r, - r2; fi = (knT)-’ and Da = ksT/c, J’ is the friction coefficient. Transforming (2.1) to relative coordinates

and center

of gravity

coordinates, aukI am

&=$D,&+?D,J&(&+/~-.

In the following

we want

to treat

only

particles and omit the center of gravity time 0 the centres of the two particles Then the initial distribution is

R = i(r, + r2), yields (2.3)

)

the

relative

motion

O(x) denotes 0

“(‘)=(I

the unit step function

forxOY

Because of the singular use the relation exp(-pfJ(rr9) Formally function

=

(2.5) nature

of the interaction

(2.2) we may also

(2.6)

eq. (2.1) may be integrated

to the initial

potential

@(r12).

p(r12, t ( r-0) = exp(fir2t)Wrr2subject

two

that at ro> d.

(2.4)

p(r12,O) = 8(rr2 - r-0)0 (r-12- d), where

of the

term in (2.3). Let us assume were separated by a distance

distribution

and gives for the conditional

ro)@(ru(2.4).

d),

distribution

(2.7)

et aI

S.HANNA

184 In spherical

f1 = (I‘+, q) expanded

coordinates

and

the s-function

f1,, = (19,). po) are

in spherical

the

can be expressed

spherical

angles;

as

6(R ~ {I,,)

can

be

harmonics,

(2.9) In spherical

coordinates

the two-particle

diffusion

operator

(2.3) has the form (2. IO)

where

iZ is the square of the angular

momentum

operator,

(2.11) The spherical

harmonics

i’Y;l(n,

are eigenfunctions

of L’, (2.12)

= I(/ + l)Y;“(O),

and therefore exp(d&)Y;l(O)

= Y;“(O)

(2.13)

exp(I&t),

where (2.14) Inserting

(2.8). (2.9) and (2.13) into (2.7) gives (2. IS)

where

X,(r. t 1r,,) is determined

Xl(r. t I rd = exp(RA) Since

we

penetration defining

know of

the

that two

a new function

by

$j 6(r - ro)O(rothe

hard-core

particles k,(r,

we

d).

(2.16)

potential take

this

allows

at no time

into

account

a by

t 1r,J as

X,(r, t 1ro) = O(r - d)JZ,(r, t I rd. Inserting

(2.2)

explicitly

this ansatz into (2.16) and differentiating

(2.17) with respect to time yields

SELF-DIFFUSION

the differential

OF BROWNIAN

PARTICLES

l?S

equation

+ 2D,$(r

- d) i

$(r,

(2.18)

t 1t-0).

Here we have used eq. (2.6) and 6(r - d) =

$

O(r - d).

The problem

of solving

to the problem

of solving

$ JI,(r, together

t 1 rO) = y

(2.19) the two-particle

diffusion

equation

is now reduced

the equation ($

with the boundary

r2i-

I(1 + I))R,(r,

t 1r&

(2.20)

condition (2.21)

this condition expresses the fact that the probability current has to vanish at r = d. Since the probability to find a particle at infinity is zero we have as a second boundary

condition

A(m, t 1ro) = 0,

(2.22)

and from eqs. (2.16) and (2.17) the initial

condition

%(r, 0 1 rO) = i 6(r - ro). In order the initial dition.zero; %(r, r

to solve condition,

(2.23)

(2.20) we look for a particular and for a complementary

their sum has to satisfy I r0)

=

Udr,

t

I r0)

+

Wr,

the boundary t

solution solution

Ur which

satisfies

W, with initial

con-

conditions.

/ r0h

Udr, 0 I r0) = i W - r0),

(2.24)

W,(r, 0 ) ro) = 0. We make the ansatz

CJdr,t 1r-0)= j dq g,(r, q I ro) em2Doq2r 0

(2.25)

IX6

S. HANNA

and find that g,(r. y 1 r,,) satisfies

the Bessel

(

r’ < + 2r $ + (q’r’~ dr-

The

solutions

kind”),

are

the

I(1 + I)))g,(r.

spherical

et al

equation

q 1r,,) = 0.

Bessel

functions

(2.26) of the

first

and

second

j,(q r) and yr(q r), respectively. 1rdidq r) + &(q

gr(r. 4) = A(4 Expanding

the a-function

1rdydq r).

in spherical

(2.27)

Bessel-functions

of the first kind”)

1

(2.28)

2.

we find

AI(~) =

2y? .h(q rd. 77

(2.29)

H,(q) = 0. The Laplace

of Udr, t 1r,,) is

transform *

Udr. z 1rd =

1 dt

e “U,(r, t 1r,,)

0 (2.30) The complementary Laplace transformation

W(r, t I ro) has

solution yields

also

to

satisfy

eq.

(2.20).

zW,(r.zir,,)-++($r’$-l(l+l))W,(r.zlro) = Wl(r, t = 0 ) ro) = 0 because

(2.3 I)

of (2.24). This is again a Bessel

2 2a -$+;z(

[([+I) rC-+20,

2 ))

equation,

Wr(r, 2 I rd = 0,

(2.32)

the solutions are the modified spherical Bessel functions of the first kind i,(rgz/2&) and of the third kind kr(rV’z/2Do)“). Since the il(r%‘z/2Do) are divergent for r+cfl, they can be omitted because of (2.22). Then

+

CI(Z

I rdkdr~/zDDJ.

(2.33)

SELF-DIFFUSION

The

unknown

condition

Cr(z ) ro)

functions

(2.21), which

OF BROWNIAN

completes

finally

187

PARTICLES

determined

by the

boundary

the solution,

m

Cl(&)=-i]dqq’

’ z + 0

f;(b)

where

stands

With (2.19,

__-

jX4

d)h(q

r0)

(2.34)

2Doq2~z/2D,k;(d%‘/)’

for @,(x)/ax Ix+,.

(2.17)

(2.33) and (2.34) we find finally cc

p(r,zIr0)=O(r-d)f

Y~‘(R)YP(Ro)*~I~~~‘~~‘~~~~~

i

I=0

m=-I

0

(2.35)

3. The self-diffusion coefficient The

Fourier-Laplace

tagged

Brownian

C,(k, z) = (e-” Here

fi,

transform

particle

of the

self-diffusion

propagator

rl[~ - firi,lm e” “I).

is the diffusion

of one

is

operator

(3. I)

of an N-particle

system (3.2)

UN is the N-particle UN = f 5 i,j=

i#j

I

potential,

which we assume

as a sum of pair potentials,

LJ(rij).

(3.3)

The tensors Dii describe the hydrodynamic discuss in the next section. If hydrodynamic

interaction interaction

effects, which we will is neglected then

Dij = D()Gij1. ((. . .)) denotes

Furthermore, ((.

* .)> =

(3.4)

1

d{rl(.

a canonical

. ‘) exp(-puN)//

expectation at-1

value,

eW-PU,h

The self-diffusion constant D, is defined from the following self-diffusion propagator in the hydrodynamic limit C,(k, z) = (z + D,k*)-‘,

for k, z small.

(3.5) form

of the

(3.6)

S. HANNA

188

Inserting

et al

(3.6) leads to (3.7)

z)) ‘(I - =C’,(k. 2)))

D, = lim lim{(k’C’,(k.

: -0 A 4

Using (3.1) we get

Employing

the expression

(3.2) for fl,.

and then taking the limit k +O, D,= where

k. (D,,)-

k-l;lrr((k

performing

some partial

integration5

D, is given by

.f,)[z

-- ri,]

(3.9)

‘(k .f,,).

k is a unit vector in the direction

of k and (3.10)

A

similar

operator

result

for

technique.

D,

was

Since

model of hydrodynamic

obtained

Ackerson

interaction,

by

Ackersonh),

restricted

a projection

his treatment

the Oseen-tensor.

absent in ref. [6], because the tensor divergence

using

to a special

the last term in (3.10) is

of D,j vanishes for the Oseen

tensor. In the present order

paper

concentration

corresponding first-order order

we restrict

correction

expression

assuming

term stems from

clusters

will

ourselves

to D,.

The

a noninteracting

N independent

contribute

to U(C’).

equivalent

it is sufficient

Neglecting

now also hydrodynamic

to the calculation zero-order

term

single particle,

two-particle

Since

clusters.

all particles

to treat one cluster and to multiply interaction,

of the first-

is given

are

by the Do. The

All higherstatistically

the result by N.

eq. (3.4), we obtain from (3.9)

and (3.10) D,=Do-lime

z-0

I

drl?(k ^ - fdrd)

x [z - fi(i,J’Ci *fdrd)

(3.11)

expt-pU(rdi,

where

fdrld

= HOP

&

U(rtd = DOB(T&U(r12))i12.

hii,, is now the same two-particle is the hard-core

potential

diffusion

operator

(2.2) and iI2 = r12/r12.

(3.12) as in (2.1),

U(rrz) = U(r,,)

SELF-DIFFUSION

OF BROWNIAN

We now use the properties of the hard-core exp(-pU(rr,)), eq. (2.6) and (2.19), which yields frArr2) exp(- PUG-r,)) = -Do?u &

189

PARTICLES

radial distribution

function

0 G-12- d)

= - Doi128(r12 - d) = -Doi1220(r12-

d)6(r12-

d).

(3.13)

The last equality should be considered as an operator equality, valid in an infinite integral over r12. The factor 2 arises from O(0) = l/2. This property can be verified by considering the O-function and the a-function as limits of continuous functions, e.g. (3.14a)

0 (x) = f + ljr&an-‘3,

(3.14b) Inserting now a a-function 8(r12- ro) behind the resolvent operator and using (2.7) we can write for the self-diffusion coefficient

.

aW12)

drr2 drok - i12 ~ x

in (3.11)

ah2

p(r12, z = 0 1ro)k - io6(ro- d) . >

(3.15)

Writing k^- i,2 = cos 19,~= ~/(47~/3)Y%fl,~) and analogously for k^- io, using the Laplace-transform of (2.15) and the orthogonality relations of the spherical harmonics, I

dRY ;I(fl)Y ;?‘(a)* = 6,,&,,,,,,,

(3.16)

D, becomes

x I drorfiXl(rlz, z = 0 1r,Jb(ro-

d)).

(3.17)

Finally, with (2.17), (2.6) and (2.19) D=D s

0

=

(~-&CD 3

Do( 1 - F

o I

dwdh

‘2

-

4

cd4D,,J?(d, z = 0 1d)).

j

dror%h,

.z =

0

1 ro)Wo

-

4)

(3.18)

S. HANNA

190

In

order

to determine

et al.

%(d, z = 0 1 d), we use (2.33) and (2.34) and perform

the

limit z + 0”)

id4 rd -

0

(p(r,,- r,z)++

$ idy

4 +

f$ qjdq d)) 1

(rl?-

II

rd+& I2

O(rr

(3.19)

-%_@(d-r")). For the special

case r. = r,? = d we find

%(d, z = 0 1d) = (8dDo)

Here again the property yields

the exact

II,= where

(3.20)

‘.

O(0) = l/2 has been used. Substitution

of (3.20) in (3.18)

result (3.21)

Do(l-24).

C#Jis the volume

cf~= $

concentration

of the Brownian

particles,

(d/2)‘c.

(3.22)

4. The effect of the hydrodynamic Neglecting

d)

0

the hydrodynamic

interaction interaction

on the self-diffusion has enabled

coefficient

us to find an exact

solution for the two-particle diffusion equation and an exact result for the first-order concentration correction to the self-diffusion coefficient. This result is however not of much use for the understanding of experiments because of the complete absence of the hydrodynamic interaction. the effect of this interaction in a perturbative way.

Therefore

we include

To lowest order in the concentration we need only the hydrodynamic interaction of a pair of particles. In its most general form the diffusion tensors can then be written as’,“)

@I = 022= &AI + h,(rdP 012 = &I = &CWrP

+ MM1

+ M-&1

The functions hn(r12) will be specified The tensor P projects an arbitrary

- IV, - P)).

(4. la) (4. lb)

later on by choosing particular models. vector on the relative vector r12. In

SELF-DIFFUSION

spherical

OF BROWNIAN

191

PARTICLES

coordinates (4.2)

PV = 1, and all other elements are zero. Using eqs. (4.1) in the general responding &=

two-particle

diffusion

form

of the

operator

diffusion

separates

operator

the

cor-

as

&+A&.

(4.3)

Here fi$ is the operator without hydrodynamic interaction, identical with (2.1) or (2.3), and Ah:, is the additional contribution from hydrodynamic interaction. Omitting again the terms describing the center of gravity motions of the two-particle

cluster

we get (4.4)

Inserting

(4. I), Afi(i:, becomes

+

2Do&.(Mrd

- h(r,z)Xl - P) -

&

=2D,(&(hdrd - h3(r,3))(~2+ P y)

+2Ddhdrd

+2c2ri2(&2 +P y)

- M-d)

- 2Ddhdr12)- h&d k2i’. We now consider

the action

which

to the solution

is identical

Afiim(n2,t 1rd = 2 I=0

(4.5)

of Afit* on the zeroth-order

i

found

in section

solution

2. With (2.19,

pdrr2, t 1ro), (2.17),

Y;"(fl)Yi"(fl~)

m=-I

x [ WI2 -

d)2Q($2 (h&-d - Wd) &

+ Oh - 42Ddhdr12) - Mrd

ia T

-

r12 arlz

a

62-ah2

- O(r,, - d)2DO(hAr,2) - h4(r,& $z [(I + 1) + 2DO(h,(r,2) - h3(r,2)M(r,2-

d) -$-j%(r,2,

t 1rd.

(4.6)

192

S. HANNA

Because

of eq. (2.21),

dfi(i:z&rl5

the fourth

et al

term vanishes and

t I rd (4.7)

with the operator

The

Laplace-transformed

equation

solution

with the operator

of

the

complete

two-partic!e

as given in (4.3) can be written

in a formal

diffusion way as

with

p”(r12. z ( rd = ([z -

dkl

Eq. (4.9) is the perturbative

‘A~i:#‘pdrI~. solution

interaction. The first-order

correction

is

z ( rd

of the problem

(4. IO) including

hydrodynamic

SELF-DIFFUSION

Inserting

OF BROWNIAN

193

PARTICLES

(2.15) and (2.17) gives

pr(r12, z

1 ro) = @(rt2-

x

4

Y;l(fl)Yi"(fl~) 2 2 I=0 m=mI

dr’r’2_%r(r12r z

1 r')&%(r',

(4.12)

z 1 ro).

I

For the calculation

of the diffusion

in section

eqs. (3.9), (3.10) we find

3. From

D,=Do+c -

d3rr2k^ .

I

we proceed

in the same way as

. i ew-pU(b2))

d3rr2(k^ . fl2(r12))[z -

lim c Z-r0

D’,,

coefficient

h721m

I

x (I .fdr12))

(4.13)

exp(-P~(r~2)),

D;, = /A, - Do1 and

where

(4.14)

The tensor

divergence

is”)

=

Do(&

(hdr12) -

+Ddhdrd =

defining the function This result shows

-

h(rd))i

- h(rd

- hdrd + Wd

+2i I

(4.15)

DoHdrdi,

Hl(rr2). that f12(rr2) is proportional

ffdr12) - H2(r12)P

to i

~

(4.16)

where I-Mrr2) = - (1 + hr(rr2) - Mrr2)). Using

a similar

procedure

as in eq. (3.13) we can write the last factors

(4.17) of the

194

S. HANNA

integrand

et al

in (4.13) as

k^* fdrd

expt-pUtrid)

= exp(-pU(r,z))D,,(H,(r,,) + 2H2(d)ti(rIT-

Inserting yields AD,=

a a-function

D,-

behind

the resolvent

Y:‘(0).

(4.18)

in the last term of (4.13)

(4.19) for AD,. Restricting

ourselves

to

(2.17) and (4.12)

AD:“‘=+;_/ I

AD’,” = 7 x

operator

%

d)).

Using (4.9), this is the perturbation series p = p. + pI, we obtain AD, = AD/“’ + AD’,“.

x

J

D,,

x (Hdrd + 2HddP(ro-

With (2.!S),

d))

drr,r%r(rr:, CD;

I

drI?r:?(H,(rl*)0(r,z-d) + Hd46(r12-

I

z = 0 1 r&Hl(rcJ

+ 2HAdP(ro-

dr~~r%H~(r~~)@ (rt2 - d) + HAd)S(r,,

dr’rr2%,(r12,z = 0 1 r’)

I

drg~~,~,(r’,

d)) d)).

(4.20)

- d))

z = 0 1 ro)

x (Hdrd + 2HddF%r(~- d)).

(4.21)

The function .%,(rn, z = 0 1 r,,) has been given in (3.19), the differential operator 2, is defined in eq. (4.8), the only unknown functions are now the h,(r), describing

the specific

form of the hydrodynamic

interaction.

The rest of our

task is a tedious but elementary calculation of the integrals in (4.20), (4.21). We performed these calculations for two specific models of hydrodynamic interaction: the Oseen-approximation and Felderhof’s reflection series”). The Oseen-approximation is obtained from the solution of the linearized NavierStokes equation to lowest order in the ratio d/r. Therefore the Oseen approximation is valid if two particles are far separated. On the other hand, the hydrodynamic interaction will only be important if two particles are close together, and therefore the Oseen approximation must be considered as poor. Nevertheless, because of the simplicity of its analytical form, it is the most

SELF-DIFFUSION

OF BROWNIAN

195

PARTICLES

often used model of hydrodynamic interaction; for this reason we include it here. There have been many attempts to improve the Oseen approximation, solving the linearized Navier-Stokes equation to higher orders in d/r. Batchelo?) has obtained numerical results for the hydrodynamic interaction for all values of the interparticle separation and has given the analytical asymptotic form up to (d/r)4. In this paper we adapt results derived by Felderhof’@) who has presented the asymptotic expansion up to (d/r)‘. Felderhof’s result seems to be the most reliable analytical form of the hydrodynamic interaction for a system of low concentration (to Q(4), which is consistent with our calculations). In table I we specify the two forms of hydrodynamic interaction, and in table II we present the results for AD’,“, AD:“, J d3r& * D;, * h9(rlz- d), and a, where (Y is defined by D, = DO(l - (~4). The mean-field contribution _fd3r& * DiI * b(r12d) has already been calculated by Felderhof and is taken from his work.

5. Discussion

The N-particle Smoluchowski equation has been reduced to a two-particle in solution. equation for a system at low concentration of macroparticles

TABLE I Expressions for the function hi(x), i = 1,. ,4 entering definition (4.la, b) of the diffusion tensor D, which describe hydrodynamic interaction in Oseen- and Felderhof-approximation, respectively. Also given are the functions H)(x), eq. (4.15) HZ(X), eq. (4.17); x stands for d/r

Oseen

hl

0

hz

0

h, h4

3 -x 4 3 -x 8

HI

0

HZ

-1+:x

Felderhof 11 --x154 +mx6 64 --x6 17 1024 3 1 75 -x-jjx 3 +mx’ 4 3 1 -x+16x) 8

the the and

S. HANNA

196

et al

TAHI.E Results for the self-diffusion column

represents

contributions

(4.20)

coefficient

the mean-field and

(4.21).

tributions

D, to first

contribution. respectively.

II order in the volume

The The

second fourth

by giving the coefficient

and third column

concentration columns

summarizes

IY in D, = Do(

4. The first

are the memory the

three

con-

I ~ UC/J)

Hydrodynamic interaction neglected Oseenapproximation Felderhofapproximation

From

the

2+! x

0

- 1.736,

knowledge

of the

distribution

function

we have

calculated

the

self-diffusion propagator C,(k, t). The hydrodynamic pole of this correlation function defines the self-diffusion coefficient II,. The method to obtain D, which has been used in this paper, is equivalent to the approach which applies the Mori-Zwanzig projection operator formalism to the N-particle Smoluchowski equationh.‘). One obtains an exact memory equation for C,(k, t), which can be written as*)

6 C,(k, t) = - K,,(k)C,(k,

dt’M,(k,

t) +

t - t’)C,(k,

t’).

(5.1)

0 Here,

K,,(k)

is the first cumulant

of C,(k, t), which determines

the initial

slope

of log C,(k, t) versus t and M,(k, t) is the memory function. The quantity given in the first column of table II, which represents the mean-field contribution

to ID,, is closely

related

to the

first

K,r(k)/(D,k*) as k + 0. Taking the Laplace-transform the self-diffusion coefficient is

cumulant;

it is the

limit

of (5. I) and using

of

(3.6),

(5.2) It was first pointed out by Ackerson6) that M,(k, t)/k* does in general not vanish as k -+ 0. Therefore, memory effects enter the determination of D,. Although the approach used in this paper avoids a direct calculation of the memory function M,(k, t) the contributions AD!‘) and 406” to D, are just the memory effects. We note that in writing eqs. (3.7) and (5.2) for D, it is implicitly assumed that M&k, z) has no hydrodynamic pole.

SELF-DIFFUSION

OF BROWNIAN

PARTICLES

197

The memory function also enters the velocity-autocorrelation function one particle. Using the continuity equation it is easy to show14) that

!j(Ui(t)



Vi(O)) = 1’ ,‘lom & [2K,,(kM(t)

- K(k, t>l,

of

(5.3)

where Vi(t) is the velocity of particle i at time t. This result together with eq. (5.2) leads to co D,= i \ dt(ui(t) * vi(O)).

(5.4)

0

Since the mean-square displacement relation function one obtains’)

W(t) is related to the velocity-autocor-

W(t) = k ((R(t) - R(O))*) = lim k-O !$@

t-

dt’(t - t’)M,(k,

(5.5)

t’)]

I 0

and D,

= lim *+m

w(t).

(5.6)

t

When analyzing experimental data it might be easier to determine first cumulant. From it one can define a quantity’5*‘6) Ds,& k) =

.K,l(k) k

which is sometimes called an effective diffusion coefficient. and (5.6) one should note that this quantity is just D s,eR= lim wO. f-0 t

only the

(5.7) From eqs. (5.5)

(5.8)

Therefore, it neglects all memory effects and differs from the hydrodynamic definition of a self-diffusion coefficient as in eqs. (5.4) and (5.6). The importance of the memory function lies in taking the fluctuations of the surrounding particles into account. Since self-diffusion is a one-particle property these fluctuations are essential. As mentioned in the introduction, this is different in the case of collective diffusion, which is a macroscopic effect. Therefore we believe that an extension of Felderhof’s method4) to the self-diffusion problem is not justified.

198

S. HANNA

et al.

We now turn to a discussion of our results First, it should be noted that memory effects

for D, summarized in table II. in the hard-core case without

hydrodynamic

changing

24).

This

contribution interest

interaction substantial

are very important, shift

(first column

is the

large

is due in table

difference

Oseen-approximation and Whereas the hydrodynamic

to the

fact

that

D, from there

II) in the hard-core between

Felderhof’s interaction

the

two

system.

results

Do to D,,(l ~

is no mean-field Of particular

obtained

form of hydrodynamic taken in Oseen-approximation

for

the

interaction. reduces

the concentration dependence of D, compared to the hard-core system rather drastically, the more reliable form given by Felderhof brings the result almost back to the hard-core case. This may be understood as follows: The Oseenterm represents the long-range part of hydrodynamic interaction and therefore reduces the effects of the very short-range direct interaction rather dramatically. The velocity field produced by the moving macroparticle removes neighboring particles from the collision trajectory leading to an increase of D, compared to the pure hard-core result. Taking into account the short-range term of the hydrodynamic interaction one also includes the reflected velocity fields from other macroparticles. These have the effect of slowing down the motion of the tagged particle, thereby decreasing D, again. In comparing our results with those Batchelo?) obtained D, = Do(l - 1.834),

of earlier papers we first which should be compared

note that with our

mean-field value in the first column of table II. More recently, Marqusee and Deutch? have calculated D, = Do( 1 - a$) from the memory equation approach by using a weak-coupling approximation in evaluating M,(k, t). For the hard-core system cy = 1.33 is obtained compared to cu = 2 in our treatment. The weak-coupling approximation replaces a two-particle propagator by the product of two one-particle propagators. This is an expansion in terms of the strength of the interaction. Such expansion seems to be questionable for the hard-core case, which would explain the rather large difference of the results. These have also included hydrodynamic interaction in Oseen-approximation cy = 0.07 compared to our (Y= 0.09. As mentioned above the Oseen-term collisions

drastically.

Therefore

it is really not so important,

authors finding reduces

how the hard-core

is

treated if hydrodynamic interactions are included. This seems to be the explanation why their result and ours are so close in the Oseen-approximation. Finally, we want to compare D, with the collective diffusion coefficient D,. As noted earlier the collective memory function M,.k. t) vanishes faster than k’ as k +O. Therefore, memory effects vanish for D, to first order in C#LOnly three-particle correlations would contribute if hydrodynamic interaction is included. For a hard-core system one has the well-known result”) D, = DO(I + 84). Defining the collective friction coefficient fC by writing in the

SELF-DIFFUSION

OF BROWNIAN

PARTICLES

199

solvent fixed frame (5.9)

where n is the osmotic pressure, one has fC = fo, f. being the friction coefficient of a free particle. The self-friction coefficient fS is defined by D, = kaT/f,. Therefore, for a hard-core system we have fS = fo( 1 + 24) Z fC. Including now Felderhof’s form for the hydrodynamic interaction, D, is given by4) D, = DO(1 + 1.564), and from (5.9) one finds fC = fO(1 + 6.444). This is to be compared with our result for D, which is D, = Do( 1 - 1.894) and with fS = f,Jl + 1.894). Therefore in all cases considered here, one always has D, # D, and fC # fS. There is no reason why these different transport coefficients should coincide besides in the trivial limit of infinite dilution.

References 1) See, for instance, K.H. Keller, E.R. Canales and S.I. Yum, J. Phys. Chem. 75 (1971) 379. These authors used a diaphragm diffusion cell. 2) A recent comprehensive review of light scattering methods is given by P.N. Pusey and R.J.A. Tough, in: Dynamics of Light Scattering and Velocimetry: Application of Photon Correlation Spectroscopy, R. Pecora, ed. (Plenum, New York, 1981) to be published. 3) G.K. Batchelor, J. Fluid Mech. 74 (1976) I. 4) B.U. Felderhof, J. Phys. All (1978) 929. 5) R.B. Jones, Physica 97A (1979) 113. 6) B.J. Ackerson, J. Chem. Phys. 64 (1976) 242, 69 (1978) 684. 7) W. Dieterich and I. Peschel, Physica 95A (1979) 208. 8) W. Hess and R. Klein, Physica 105A (1981) 552. 9) J.A. Marqusee and J.M. Deutch, J. Chem. Phys. 73 (1980) 5396. 10) B.U. Felderhof, Physica 89A (1977) 373. Functions (Dover, New York, 11) M. Abramowitz and I.A. Stegun, Handbook of Mathematical 1965). Methods for Physicists (Academic Press, New York, London, l-3 G. Arfken, Mathematical 1970). R.B. Bird, R.C. Armstrong and 0. Hassager, Dynamics of Polymeric 13) See, for instance, Liquids, Vol. 1 (Wiley, New York, 1977). in Liquids and Macromolecular Solutions, V. Degiorgio, M. 14) W. Hess, in: Light Scattering Corti and M. Giglio, eds. (Plenum, New York, 1980). 15) P.R. Wills, J. Chem. Phys. 70 (1979) 5865. 16) H.M. Fijnaut, J. Chem. Phys. 74 (1981) 6857. B.J. Berne and R. Pecora, Dynamic Light Scattering (Wiley, New York, 17) See, for instance, 1976).