3
Acceleration of Brownian particle diffusion parallel to a fast random field with a short spatial period A. N. Malakhov NizhniŃ Novgorod State University ~Submitted April 21, 1998! Pis’ma Zh. Tekh. Fiz. 24, 9–15 ~November 12, 1998! It is shown that a spatially periodic random field fluctuating rapidly near zero with a fairly small spatial period may substantially accelerate the diffusion of Brownian particles parallel to this field. © 1998 American Institute of Physics. @S1063-7850~98!00211-0# The diffusive spreading of the mean square of the coor- dinate x ( t ) of a Brownian particle ^ x 2 & 52 Dt undergoing free diffusion in a viscous medium parallel to the x axis with zero initial conditions is well-known. The diffusion coeffi- cient D determines the temporal rate of spreading of a Gauss- ian probability density W( x , t ) with zero average and vari- ance ^ x 2 & . The corresponding Langevin equation has the form dx ( t )/ dt 5j ( t ), where j ( t ) is stationary Gaussian white noise with ^ j ( t ) & 50 and ^ j ( t ) j ( t 1t ) & 52 D d ( t ). The diffusion coefficient D 5kT / h , where k is the Bolt- zmann constant, is determined by the equivalent temperature T and viscosity h of the medium. Let us now assume that the spreading of Brownian par- ticles from the initial distribution W( x ,0) 5d ( x ) takes place in this medium under the action of additional forces attribut- able to the random potential field F ( x ) z ( t ), where z ( t ) is a dimensionless Gaussian delta-correlated process with ^ z ( t ) & 50, ^ z ( t ) z ( t 1t ) & 52 D z d ( t ), which is statistically inde- pendent of the thermal noise j ( t ). In this case we can write the Langevin equation as dx ~ t ! dt 52 d F~ x ! hdx z ~ t ! 1j ~ t ! 52D d w ~ x ! dx z ~ t ! 1j ~ t ! , ~1! where we introduce the dimensionless potential profile w ( x ) 5F ( x )/ kT . We note that the random process x ( t ) is a continuous Markov process. The Fokker–Planck equation for W( x , t ) has the stan- dard form ~see, e.g., Ref. 1!: ] W~ x , t ! ] t 5 ] ] x @ K 1 ~ x , t ! W~ x , t !# 1 1 2 ] 2 ] x 2 @ K 2 ~ x , t ! W~ x , t !# , ~2! where K 1 ( x , t ) and K 2 ( x , t ) are the drift and diffusion coef- ficients which can be found from the Langevin equation. The general evolution equations for the first two mo- ments of the continuous Markov random process x ( t ) have the form ~Ref. 2, § 10.6! d ^ x & dt 5^ K 1 ~ x , t ! & , d ^ x 2 & dt 52 ^ xK 1 ~ x , t ! & 1^ K 2 ~ x , t ! & , ~3! where statistical averaging is performed over the probability density W( x , t ) determined by the Fokker–Planck equation ~2!. We can easily find ~see Ref. 1 and Ref. 2, p. 367! the following values of the drift and diffusion coefficients corre- sponding to the Langevin equation ~1!: K 1 ~ x ! 5 D 2 D z 2 l 2 d dx c 2 ~ x ! , K 2 ~ x ! 52 D F 1 1 DD z l 2 c 2 ~ x ! G . ~4! Here we introduce the dimensionless function c ( x ) 5ld w ( x )/ dx , where l is some scale. Thus, for an arbitrary profile w ( x ) the evolution equations for the moments have the form d ^ x & dt 5 D 2 D z 2 l 2 K d dx w 2 ~ x ! L , d ^ x 2 & dt 5 D 2 D z l 2 K x d w 2 ~ x ! dx L 12 D F 1 1 DD z l 2 ^ w 2 ~ x ! & G . ~5! If the potential function w ( x ) is even, by virtue of the symmetry of the situation, the probability density is also an even function W( 2x , t ) 5W( x , t ). In this case, we find K 1 ( x ) 50 ~and therefore ^ x & [0) and only the second equa- tion ~5! for the mean square of the Brownian particle coor- dinates remains. As a first example, consider the even sawtooth potential profile w ( x ) with the spatial period l shown in Fig. 1. We can easily see that w ( x ) 52 bx / l holds for 0 <x <l /2 and find c 2 ( x ) 54 b 2 for all x . We therefore obtain K 2 ( x ) 52 D @ 1 14 DD z ( b / l ) 2 # . Thus, for zero initial conditions the spreading law for the mean square of the coordinate of Brownian particles has the diffusion form ^ x 2 & 52 D eff t , where the effective diffusion coefficient is given by D eff 5D F 1 14 DD z S b l D 2 G .D . ~6! We note that this result is exact. If there is no fluctuating field ( D z 50), we arrive at D eff 5D , as should be the case. TECHNICAL PHYSICS LETTERS VOLUME 24, NUMBER 11 NOVEMBER 1998 833 1063-7850/98/24(11)/3/$15.00 © 1998 American Institute of Physics

Acceleration of Brownian particle diffusion parallel to a fast random field with a short spatial period

Embed Size (px)

Citation preview

TECHNICAL PHYSICS LETTERS VOLUME 24, NUMBER 11 NOVEMBER 1998

Acceleration of Brownian particle diffusion parallel to a fast random field with a shortspatial period

A. N. Malakhov

Nizhni� Novgorod State University~Submitted April 21, 1998!Pis’ma Zh. Tekh. Fiz.24, 9–15~November 12, 1998!

It is shown that a spatially periodic random field fluctuating rapidly near zero with a fairly smallspatial period may substantially accelerate the diffusion of Brownian particles parallel tothis field. © 1998 American Institute of Physics.@S1063-7850~98!00211-0#

o

-s

-th

tu

a

ut

fil

-

o

ilityon

-

ryve

an

-r-

tial

ethe

ng.

The diffusive spreading of the mean square of the codinate x(t) of a Brownian particle x2&52Dt undergoingfree diffusion in a viscous medium parallel to thex axis withzero initial conditions is well-known. The diffusion coefficientD determines the temporal rate of spreading of a Gauian probability densityW(x,t) with zero average and variance ^x2&. The corresponding Langevin equation hasform dx(t)/dt5j(t), where j(t) is stationary Gaussianwhite noise with^j(t)&50 and^j(t)j(t1t)&52Dd(t).

The diffusion coefficientD5kT/h, wherek is the Bolt-zmann constant, is determined by the equivalent temperaT and viscosityh of the medium.

Let us now assume that the spreading of Brownian pticles from the initial distributionW(x,0)5d(x) takes placein this medium under the action of additional forces attribable to the random potential fieldF(x)z(t), wherez(t) is adimensionless Gaussian delta-correlated process with^z(t)&50, ^z(t)z(t1t)&52Dzd(t), which is statistically inde-pendent of the thermal noisej(t). In this case we can writethe Langevin equation as

dx~ t !

dt52

dF~x!

hdxz~ t !1j~ t !52D

dw~x!

dxz~ t !1j~ t !,

~1!

where we introduce the dimensionless potential prow(x)5F(x)/kT. We note that the random processx(t) is acontinuous Markov process.

The Fokker–Planck equation forW(x,t) has the stan-dard form~see, e.g., Ref. 1!:

]W~x,t !

]t5

]

]x@K1~x,t !W~x,t !#1

1

2

]2

]x2@K2~x,t !W~x,t !#,

~2!

whereK1(x,t) andK2(x,t) are the drift and diffusion coefficients which can be found from the Langevin equation.

The general evolution equations for the first two mments of the continuous Markov random processx(t) havethe form ~Ref. 2, § 10.6!

d^x&dt

5^K1~x,t !&,d^x2&

dt52^xK1~x,t !&1^K2~x,t !&, ~3!

8331063-7850/98/24(11)/3/$15.00

r-

s-

e

re

r-

-

e

-

where statistical averaging is performed over the probabdensityW(x,t) determined by the Fokker–Planck equati~2!.

We can easily find~see Ref. 1 and Ref. 2, p. 367! thefollowing values of the drift and diffusion coefficients corresponding to the Langevin equation~1!:

K1~x!5D2Dz

2l 2

d

dxc2~x!, K2~x!52DF11

DDz

l 2c2~x!G .

~4!

Here we introduce the dimensionless functionc(x)5 ldw(x)/dx, where l is some scale. Thus, for an arbitraprofile w(x) the evolution equations for the moments hathe form

d^x&dt

5D2Dz

2l 2 K d

dxw2~x!L ,

d^x2&dt

5D2Dz

l 2 K xdw2~x!

dx L 12DF11DDz

l 2^w2~x!&G .

~5!

If the potential functionw(x) is even, by virtue of thesymmetry of the situation, the probability density is alsoeven function W(2x,t)5W(x,t). In this case, we findK1(x)50 ~and thereforex&[0) and only the second equation ~5! for the mean square of the Brownian particle coodinates remains.

As a first example, consider the even sawtooth potenprofilew(x) with the spatial periodl shown in Fig. 1. We caneasily see thatw(x)52bx/ l holds for 0<x< l /2 and findc2(x)54b2 for all x.

We therefore obtain K2(x)52D@114DDz(b/ l )2#.Thus, for zero initial conditions the spreading law for thmean square of the coordinate of Brownian particles hasdiffusion form ^x2&52Deff t, where the effective diffusioncoefficient is given by

Deff5DF114DDzS b

l D2 G.D. ~6!

We note that this result is exact. If there is no fluctuatifield (Dz50), we arrive atDeff5D, as should be the case

© 1998 American Institute of Physics

834 Tech. Phys. Lett. 24 (11), November 1998 A. N. Malakhov

FIG. 1. Sawtooth potential profile.

n

mio

eitean

n

aconhe

nd

po-n

the

e

.5

d

opeint of

ve

heresfu-

ven

the-

i-

We also note that for a sawtooth field, as for free diffusioW(x,t) is a strictly Gaussian probability density.

Thus, a periodic, sawtooth, rapidly fluctuating randofield acting on Brownian particles accelerates their diffusand this acceleration increases with increasingb/ l , i.e., withincreasing absolute value of the field slope per period.

Since this potential profilew(x) is even and the diffusionis symmetric in both directions, the probability flux at thpoint x50 is zero and so a reflecting boundary can be sat this point and diffusion from the reflecting boundary cbe considered only in the direction of positivex. The result~6! remains unchanged. From this it follows that diffusiocan be accelerated in one direction~away from the reflectingboundary!.

Let us now discuss the physical mechanism for theceleration of diffusion, confining our analysis to the motiof Brownian particles over a single period of the field in tdirection of increasingx. For simplicity we shall give thefield profile for two cases:z511 andz521 ~Fig. 2a!.

If z511 holds, the particles diffuse up the slope athis motion is slower than that along the horizontal axis. Wshall estimate this time with an absorbing boundary~a po-

FIG. 2. Diffusion of Brownian particles from the initial probability distrbution: a — slow motion up the slope and fast motion down the slope:1 —initial distribution, 2 — absorbing boundaries; b — general motion ofBrownian particles over a field period.

,

n

d

-

e

tential going vertically downward! situated at the pointl /2.In this case, the Brownian particles should overcome atential barrier of heightb. It can be shown that the time taketo overcome the barrier~see Ref. 3, for instance!, i.e., thetime taken for the particles to move up the slope bydistancel /2 is

Tup5l 2

4Db2~eb212b!. ~7!

For z521 the particles diffuse down the slope and the timtaken for them to cover the distancel /2 is

Tdown5l 2

4Db2~e2b211b!,Tup. ~8!

It follows from these formulas that forb50.5 the diffu-sion velocity of the Brownian particles down the slope is 1times higher than the upward diffusion velocity, forb52 thedownward diffusion velocity is almost four times higher, anfor b<4 the velocity ratio iseb/b@1.

Since the average of the random functionz(t) is zero, itssign varies continuously and the particles move up the slhalf the time and downward half the time. The differencethe velocities has the result that the particles cover mosthe path betweenx50 and x5 l /2 in the ‘‘downsloping’’state and this proportion increases with increasingb/ l . Themotion of the Brownian particles betweenl /2 andl is exactlythe same, the only difference being that they now modown the slope forz(t).0 and up the slope forz(t),0.

Thus, the changes in the sign ofz(t) make it possible forthe Brownian particles to cover most of the path in tdownsloping state~Fig. 2b! and they therefore move fastealong thex axis than under free diffusion, which acceleratthe diffusion process in accordance with the effective difsion coefficient~6!.

Let us now consider a second example, a harmonic efield w(x)5bcos(2px/l)/ with the same periodl . In this case,we have w2(x)54p2b2sin2(2px/l). However, the secondequation~5! cannot be solved accurately becauseW(x,t) isnon-Gaussian. Using a Gaussian approximation forW(x,t)with zero average and unknown variance, we can obtainfollowing nonlinear equation forx2& from the second equation ~5!:

d^x2&dt

54D2DzS b

l D2

a^x2&exp~2a^x2&!12D

3H 112p2DDzS b

l D2

@12exp~2a^x2&!#J , ~9!

n

ten

eidc-b

st

n-fufeeto

ralf affi-

stralon

n-–

o-

z.

h.

835Tech. Phys. Lett. 24 (11), November 1998 A. N. Malakhov

where a58p2/ l 2. An analysis of this nonlinear equatioshows that for times such that^x2&1/2 exceeds the half-periodof the field (a^x2&@1) we have

^x2&52Deff t, Deff5DF112p2DDzS b

l D2 G . ~10!

Thus, for a cosinusoidal fluctuating field this approximaeffective diffusion coefficient has the same order of magtude as that for the sawtooth field~6!, maintaining an accel-eration of the diffusion which increases with increasingb/ l .

Analysis shows that the diffusion acceleration effect dscribed above is not related to the sawtooth or cosinusoprofile of the field fluctuating rapidly about zero. The effewill evidently occur for any periodic or aperiodic field consisting of fairly steep up- and downsloping elements of artrary profile.

The so-called molecular motor mechanism or stocharatchet has recently been studied intensively~see Refs. 4–8and the literature cited therein!. This effect essentially in-volves the existence of a directional diffusion flux of Browian particles caused by the formation of asymmetric difsion conditions as a result of some type of modulation operiodic~asymmetric! field at a frequency determined by thdiffusion time on scales of the field period. The proposdiffusion acceleration effect is merely externally similar

i-

-alt

i-

ic

-a

d

the molecular motor effect. The difference is that a genedirectional transfer of material is observed in the case omolecular motor whereas in our case, the effective coecient of ‘‘ordinary’’ diffusion increases as a result of famodulation of the field, which is unrelated to the tempocharacteristics of the spreading of the probability densityscales of the field period.

This work was supported by the Russian Fund for Fudamental Research~Grants Nos. 96–02–16772-a and 9615–96718!.

1H. Risken,The Fokker-Planck Equation~Springer-Verlag, Berlin 1989!,472 pp.

2A. N. Malakhov, Cumulative Analysis of Random Non-Gaussian Prcesses and Their Transformations@in Russian#, Sovetskoe Radio, Moscow~1978!, 376 pp.

3N. V. Agudov and A. N. Malakhov, Izv. Vyssh. Uchebn. Zaved. Radiofi36~2!, 148 ~1993!.

4A. P. Nikitin and D. E. Postnov, Pis’ma Zh. Tekh. Fiz.24~2!, 47 ~1998!@Tech. Phys. Lett.24, 61 ~1998!#.

5D. E. Postnov, A. P. Nikitin, and V. S. Anishchenko, Pis’ma Zh. TekFiz. 22~9!, 24 ~1996! @Tech. Phys. Lett.22, 352 ~1996!#.

6M. Bier and R. D. Astumian, Phys. Rev. Lett.76, 4277~1996!.7M. Bier, Phys. Lett.211, 12 ~1996!.8J. Luczka, R. Bartussek, and P. Ha¨nggi, Europhys. Lett.31, 431 ~1995!.

Translated by R. M. Durham