PHYSICAL REVIEW E 77, 051105 2008

Brownian motion of a classical harmonic oscillator in a magnetic field

J. I. Jimnez-Aquino,* R. M. Velasco, and F. J. Uribe

Departamento de Fsica, Universidad Autnoma MetropolitanaIztapalapa, Apartado Postal 55-534, Codigo Postal 09340, Mxico,
Distrito Federal, Mexico
Received 20 December 2007; revised manuscript received 27 February 2008; published 8 May 2008
In this paper, the stochastic diffusion process of a charged classical harmonic oscillator in a constant
magnetic field is exactly described through the analytical solution of the associated Langevin equation. Due to
the presence of the magnetic field, stochastic diffusion takes place across and along the magnetic field. Along
the magnetic field, the Brownian motion is exactly the same as that of the ordinary one-dimensional classical
harmonic oscillator, which was very well described in Chandrasekhars celebrated paper Rev. Mod. Phys. 15,
1 1943. Across the magnetic field, the stochastic process takes place on a plane, perpendicular to the
magnetic field. For internally Gaussian white noise, this planar-diffusion process is exactly described through
the first two moments of the positions and velocities and their corresponding cross correlations. In the absence
of the magnetic field, our analytical results are the same as those calculated by Chandrasekhar for the ordinary
harmonic oscillator. The stochastic planar diffusion is also well characterized in the overdamped approxima-
tion, through the solutions of the Langevin equation.

DOI: 10.1103/PhysRevE.77.051105 PACS numbers: 05.40.a, 02.50.r

I. INTRODUCTION lated to the so-called nonequilibrium work relations 1923,

In Chandrasekhars celebrated paper 1, the study of
Brownian motion of a free particle was very nicely described
through a mathematical treatment relying upon both the
Langevin and Fokker-Planck FP equations, and some gen-
eral lines were suggested toward solving the same problem
in the presence of an external field of force. The proposal has
been extended to those situations for which the Brownian
particle is considered to be electrically charged and also un-
der the influence of an external magnetic field, in different
situations 28. In particular, in Ref. 8, the problem of
stochastic diffusion of heavy ions in the presence of an elec-
tromagnetic field was solved through an alternative method
of solution of the FP equation. This method was extended to
solve the phase-space FP equation, describing the stochastic
diffusion of heavy ions in the presence of an external electric
field only 9. Also in Chandrasekhars paper, the solution to
the problem of a one-dimensional classical harmonic oscilla-
tor describing a Brownian motion was given through the
solutions of the Langevin and FP equations. After this work
on the stochastic harmonic oscillator, many related topics
have been reported in the literature; for example, we can
mention that the study of a harmonic oscillator with random
frequency is a subject that has been extensively investigated
in different fields, including physics 10,11, biology 12,
economics 13, and so on; the study of a harmonic oscillator
with random damping 14 was also used for the problem of
water waves influenced by a turbulent wind field 15; the
study of an underdamped harmonic oscillator for additive
and multiplicative noise has been described in Ref. 16,
showing the phenomenon of stochastic resonance 14,17,18.
Another branch that has been developed recently is that re-
*ines@xanum.uam.mx
rmvb@xanum.uam.mx
paco@xanum.uam.mx
also termed fluctuation theorems. Very recently, in Ref. 21,
Jayannavar and Sahoo reported the analytical calculation of
the distributions of work expended by a charged particle in
the presence of a magnetic field in a two-dimensional har-
monic well. This physical model is used to verify the Jarzyn-
ski equality JE 22, which relates nonequilibrium quanti-
ties with equilibrium free energies, in two different cases: in
the first case i, the center of the harmonic trap is dragged
with a uniform velocity whereas in case ii it is subjected to
an ac force. In both cases, the particle is moving in the pres-
ence of a time-dependent harmonic potential and the analyti-
cal results of 22 complement the Bohrvan Leeuwen theo-
rem on the absence of diamagnetism in a classical system.
An important point we would like to note here is that the JE
is calculated through the explicit solution of the Langevin
equation in the overdamped regime, not through the com-
plete solution of this equation. This is so because in the
large-friction limit or diffusive regime the fluctuations are
dominant. However, as far as we know, the explicit and com-
plete solution of the stochastic diffusion of a charged har-
monic oscillator in the presence of an external constant mag-
netic field is a problem that has not been reported yet in the
literature, and our purpose in this work is to give the com-
plete solution of such a problem in a three-dimensional
space-dependent harmonic potential well. Our proposal can
be extended to a situation for which the charged particle is in
a time-dependent harmonic potential well, or used for other
possible applications on single nanosized systems in physical
environments where fluctuations play a fundamental role
2428. As usual, in our work we assume that the magnetic
field is allowed to point along the z axis of the Cartesian
reference frame, and as a result two independent processes
are taking place. One takes place along the z axis or along
the magnetic field but is not affected by this field, and the
particle is in a one-dimensional harmonic well. This case is
exactly the same as that described by the ordinary one-
dimensional harmonic oscillator describing a Brownian mo-
tion 1. The other one takes place on the x-y plane across or

1539-3755/2008/775/05110513 051105-1 2008 The American Physical Society

JIMNEZ-AQUINO, VELASCO, AND URIBE
perpendicular to the magnetic field and the particle is in a
two-dimensional harmonic well. In this case, the Langevin
equations associated with the x and y axes form a coupled
pair of equations describing the planar-diffusion process,
which is an interesting but not an easy problem to describe.
To solve this planar harmonic oscillators Brownian motion
analytically, we use Landaus strategy 29, which relies
upon a transformation, on the complex plane, of the Lange-
vin equation. On the complex plane, the resulting complex
Langevin equation strongly resembles that of the free har-
monic oscillator and, therefore, its explicit solution is easily
calculated using Chandrasekhars ideas. Following the same
strategy, we also study the overdamped limiting case corre-
sponding to the large-friction limit. This case is easier than
the preceding description, because the complex linear Lange-
vin equation is not already of second order but linear and of
first order. In fact, in the same large-friction limit, the same
Landau strategy is used by Jayannavar and Sahoo 21 to
solve the problem in a two-dimensional time-dependent har-
monic well, and applied to calculate the JE. This work is
organized as follows. In Sec. II, we establish the Langevin
equation for the two independent process and solve only that
associated with the planar-diffusion process in a two-
dimensional harmonic potential well. In the absence of a
magnetic field, our analytical results are compared with those
calculated by Chandrasekhar. In Sec. III, we study the prob-
lem in the overdamped approximation and the conclusions
are given in Sec. IV. Finally, we introduce three appendixes
for the explicit calculations in our work.
II. THE LANGEVIN EQUATION OF A CLASSICAL
HARMONIC OSCILLATOR IN A CONSTANT
MAGNETIC FIELD
The Langevin equation of a charged particle embedded in
a fluid describing a Brownian motion in a space-dependent
potential Ur and in the presence of a constant magnetic
field, can be written for the velocity vector u as
q 1
u = u + u B grad Ur + At, 1 0.8
mc m
where 0 is the friction constant, q is the charge of the
particle and m its mass, grad U is the gradient operator of
potential U, and At is the fluctuating force per unit mass,
which satisfies the properties of Gaussian white noise with
zero mean value Ait = 0 and a correlation function given
by
AitA jt = 2ijt t. 2
is a constant that measures the noise intensity and, accord-
ing to the fluctuation-dissipation theorem, is related to the
friction constant by = kBT / m with kB the Boltzmann con-
stant and T the temperature of the surrounding medium. The
overdot in Eq. 1 denotes the derivative with respect to time.
Here we use A1 Ax, A2 Ay, and A3 Az. If we assume a
three-dimensional harmonic potential well Ur = kx2 + y 2
+ z2 / 2 with k a constant, and the magnetic field for simplic-
ity pointing along the z axis of the Cartesian reference frame,
051105-2
PHYSICAL REVIEW E 77, 051105 2008
that is, B = 0 , 0 , B with B a constant, then Eq. 1 can be
described by means of two independent processes. One is
described on the x-y plane perpendicular to the magnetic
field and the charged particle is in a two-dimensional har-
monic well Ux , y = kx2 + y 2 / 2; the other is along the z axis
or along the magnetic field in a one-dimensional harmonic
well Uz = kz2 / 2. In these cases, Eq. 1 can be written in
terms of its components as follows:
x + x + 2x y = Axt, 3
y + y + 2y + x = Ayt, 4
z + z + 2z = Azt, 5
where = qB / mc is the Larmor frequency and = k / m is 2
the characteristic frequency of the oscillator. As we can see,
Eq. 5 is exactly the same as that of the ordinary classical
harmonic oscillator describing a Brownian motion, which
has already been solved by Chandrasekhar 1. Thus we will
focus on the two stochastic differential equations Eqs. 3
and 4, which describe the diffusion process on the x-y
plane perpendicular to the magnetic field in a two-
dimensional harmonic well. Due to the Gaussian character-
istics of the processes here considered, there exist two ways
of describing this planar-diffusion process explicitly. One is
through the calculation of the first two moments of the vari-
ables x, y, ux, and uy, which can be achieved through the
explicit solution of the Langevin equations 3 and 4. The
other one is through the explicit solution of the Fokker-
Planck equation associated with those Langevin equations.
Here we follow the proposal of solving those Langevin equa-
tions and as a consequence the calculation of the moments of
the relevant variables. Once those moments are calculated,
all the transition probability densities TPDs Px , tx0,
1.0
0.6
A
0.4
0.2
0 2.0 4.0 6.0 8.0 10.0
FIG. 1. Reduced correlation A xx xx / kT / m2 vs the re-
duced time t for / = 1 and different values of / . Solid
line, = 0; dashed line, / = 1; dotted line, / = 2; long-dashed
line, / = 3.
BROWNIAN MOTION OF A CLASSICAL HARMONIC PHYSICAL REVIEW E 77, 051105 2008

D
1.0
B 0.2

0.1

0.8
0 2.0 4.0 6.0 8.0 10.0

FIG. 4. Reduced correlation D / kT / m vs the re-

yux yux
duced time t for / = 1 and different values of / . Solid
line, = 1; dashed line, / = 2; dotted line, / = 3.

2.0 4.0 6.0 8.0 10.0 j. In our case i = x , y , ux , uy. The two differential equa-
B / kT / m tions 3 and 4 are clearly coupled and seem to be some-
FIG. 2. Reduced correlation uxux vs the re-
uxux what complicated to solve analytically. To avoid this math-
duced time t for / = 1 and different values of / . Solid ematical difficulty, we use an alternative method of solution
line, = 0; dashed line, / = 1; dotted line, / = 2; long-dashed
to the one proposed initially by Landau and Lifshitz 29, by
line, / = 3.
solving the simple classical harmonic oscillator in the pres-
ence of a constant magnetic field. The proposal consists in
Py , ty 0, Pux , tx0 , ux0, Puy , ty 0 , uy0, Px , ux , tx0 , ux0, mapping Eqs. 3 and 4 on the complex plane, by defining
Py , uy , ty 0 , uy0, Px , uy , tx0 , uy0, Py , ux , ty 0 , ux0, and the complex function
Px , y , ux , uy , tx0 , y 0 , ux0 , uy0 can readily be calculated from
the general Gaussian distribution 30,31 = x + iy. 7
In this case, Eqs. 3 and 4 can be written as
1
P,t0 =
2n/2Det ij1/2 + + 2 = At, 8

exp 1
1iji i j j ,
2 i,j where = + i, At = Axt + iAyt. From now on, we will
write any complex number and any complex function with
6 an overbar, except for the correlations which appear in Figs.
16. As we can see, on the complex plane, Eq. 8 has a
where P , 00 = 0 represents the initial con- structure very similar to that of the ordinary one-dimensional
dition, = 1 , . . . , n, and the variance ij = i i j harmonic oscillator describing a Brownian motion and,

0.3 0.8
D.

0.6
10

0.2
C,
C

0.4
B,

0.1
A,

0.2

0 2.0 4.0 6.0 8.0 10.0 0 0.2 0.6 1.0 1.4

/ kT / m vs the re- FIG. 5. Different reduced correlations see Figs. 14 for their
FIG. 3. Reduced correlation C xux xux
definitions vs the reduced time t for / = 1 and / = 10.
duced time t for / = 1 and different values of / . Solid
line, = 0; dashed line, / = 1; dotted line, / = 2; long-dashed Solid line, A xx; dashed line, B u u ; dotted line, C
;
xux
x x
line, / = 3. long-dashed line, 10D 10 .
yux

051105-3
JIMNEZ-AQUINO, VELASCO, AND URIBE PHYSICAL REVIEW E 77, 051105 2008

1.0 0
2 u 0
1 u
a10 = 0
, a20 = 0
, 13

1
2
1
2
and therefore the solution of Eq. 8 and its corresponding
0.8
velocity can be written, respectively, as
1
t + 2 u e 1t 0
0 1 u e 2t
0.6
1
2 0 0

t
A

= Attdt , 14
0.4 0

1
ut + 2 u e 1t
10 1 u e 2t
20
0.2
1
2 0 0

=
0
t
At
tdt , 15
0 2.0 4.0 6.0 8.0 10.0

5000
where the complex functions t and
t read as follows:

FIG. 6. Reduced correlation A xx xx / kT / m2 vs the re- 1

duced time t divided by 5000 for / = 0.01 and different
t = e 1tt e 2tt, 16

1
2
values of / . The solid line corresponds to the exact value given
by Eq. 26 and = 0; the dashed line is the exact result for = 1;
1
the dots the exact result for = 10. The circles correspond to the t =
1e 1tt
2e 2tt. 17
overdamped approximation given by Eq. 50.
1
2
The solution for the complex conjugate t and its corre-
therefore, its solution is easy to calculate. In fact the solution sponding velocity u are also calculated in a completely
of Eq. 8 is calculated in Appendix A, yielding the follow- similar way in Appendix A. In order to analyze the meaning
ing result: of the solutions given in Eqs. 14 and 15 in terms of the

t t
original coordinates x, y, ux, and uy, we must separate the
1 real and imaginary parts of Eqs. 14 and 15, as well as the
t = e 1t e 1tAtdt e 2t e 2tAtdt

1
2 0 0 real and imaginary parts of the roots 1 and 2, in such a
way that the complex number 1 can be written as
+ a10e 1t + a20e 2t , 9
= + i, 18
where 1

where

1 = + 1,
2 = 1,
10
=
1 a2 + b2 + a, =
1 a2 + b2 a, 19
2 2 2 2 2 2
and and the parameters a = 2 2 42 and b = 2. We show

1

2 42 = 2 2 42 + i2. 11
in Appendix B the solution and the complete expressions for
the averages x, y, ux, and uy for arbitrary initial conditions,
and only as an example we will write here the results for the
On the other hand, the complex velocity is defined as u particular case when x0 = y 0 = uy0 = 0 and ux0 0. In this case,
= u + iu . Therefore
x y
x = et/2B0sinht/2cost/2cost/2

ut =

1
1
2
1e 1t
0
t
e 1tAtdt
+ cosht/2sint/2sint/2
+ D0cosht/2sint/2cost/2

2e 2t
0
t
e 2tAtdt +
1a10e 1t +
2a20e 2t .
sinht/2cost/2sint/2, 20

y = et/2B0cosht/2sint/2cost/2
12
sinht/2cost/2sint/2
The constants a10 and a20 can be calculated in terms of the
initial conditions 0 0 = x0 + iy 0, u0 u0 = ux0 + iuy0 D0sinht/2cost/2cost/2
and are given by + cosht/2sint/2sint/2, 21

051105-4
BROWNIAN MOTION OF A CLASSICAL HARMONIC PHYSICAL REVIEW E 77, 051105 2008

ux = et/2ux0cosht/2cost/2cost/2 4 k BT t
xux = e sinh2t/2 + sin2t/2, 28
+ sinht/2sint/2sint/2 m2 + 2

et/2F0sinht/2cost/2cost/2
2kBT
+ cosht/2sint/2sint/2 yux = et sinht sint, 29
m2 + 2
+ H0cosht/2sint/2cost/2
and xy = 0, uxuy = 0.
sinht/2cost/2sint/2, 22 Now, let us define the reduced correlations as the

corresponding dimensionless quantities, that is, xx

uy = et/2ux0cosht/2cost/2sint/2 = xx / kBT / m ,
2
u u = uxux / kBT / m, xu
x x x
= xux / kBT / m, and yu = yux / kBT / m. Also the re-
sinht/2sint/2cost/2 x
duced time is = t and in these new variables the behavior
+ et/2H0sinht/2cost/2cost/2 of correlations is shown in Figs. 14 for certain values of the
magnetic field represented by the variable / and of / .
+ cosht/2sint/2sint/2
Several comments can be made about the behavior shown in
F0cosht/2sint/2cost/2 Figs. 14. First we recall that the value = 0 corresponds to
Chandrasekhars solution. In this case, in Fig. 4 the correla-
sinht/2cost/2sint/2, 23
tion yu = 0 = 0, and it is not shown. Second, the relax-
x
where now ation when t is slower when 0 than in the case
without the magnetic field. This effect is enhanced as the
2ux0 2ux0 normalized field / grows. In Figs. 13 the small-time
B0 2, D0 , 24 behavior is essentially the same no matter the field value.
2 + 2 + 2 Lastly, in Fig. 5 we can observe the behavior of the reduced
correlations for / = 1 and / = 10. In this case, due to the
u x0 2 + 2 u x0 2 2 magnitude of the / parameter, oscillations can be seen.
F0 , H0 . 25 All our analytical results for the averages x, y, ux,
2 + 2 2 + 2 and uy, and the correlation functions given in Eqs.
2629, can be compared in the absence of magnetic field
Equations 2023 represent the deterministic solution of
= 0 with those obtained by Chandrasekhar for the ordi-
the problem since the average of the noise vanishes. The
nary one-dimensional harmonic oscillator. In this case, the
quantity is such that 0 and the leading terms in
parameter b = 0 and thus = 0, a = 2 42, which leads to
Eqs. 2023 when t decrease exponentially with a re-
laxation time 2 / , in such a way that the presence of = a. If we define 1 a, then = 1. Under these condi-
the magnetic field makes the relaxation slower than in the tions, we can clearly verify that

case where B = 0. As a second step in the analysis of this

problem we calculate the two-time auto- and cross correla- x = x0et/2 cosh1t/2 + sinh1t/2
tion of the relevant variables. In Appendix B, we provide the 1
explicit expressions for all the correlations at different times 2ux0
t1 and t2. For equal times t1 = t2 = t, Eqs. B47B52 can be + et/2 sinh1t/2, 30
straightforwardly reduced to 1

xx =
k BT
m 2 1e
t 2 + 2
1+ 2
+ 2
ux = ux0et/2 cosh1t/2

1
sinh1t/2

2 sinh2t/2 +


sinht
2x02 t/2
e sinh1t/2. 31

1
+
2 2
+
2
2 2 sin t/2 + sint
2

, 26 For the variances, we have

uxux =
k BT
+
1 e t 1 + 2
2 2 xx =
k BT
m 2
1 e t
2
2
2 sinh 1t/2 +
1
2
1
sinh1t + 1 ,
m + 2 32

2 sinh2t/2 sinht
2

k BT
1 et 2 2 sinh21t/2 sinh1t + 1

uxux = ,
2 2 m 1 1
2 2 sin t/2 sint 27
2
+ 2 ,
+ 33

051105-5
JIMNEZ-AQUINO, VELASCO, AND URIBE PHYSICAL REVIEW E 77, 051105 2008

4 k BT e Rt
xux = et sinh2t/2. 34 ux = 2y 0 x0sinIt 2x0
m21 2 + 2
The averages y and uy have the same expressions as Eqs. + y 0cosIt, 44
30 and 31, respectively, just changing the initial condi-
tions x0 y 0 and ux0 uy0; both correspond exactly to Chan- e Rt
drasekhars results. The auto- and cross correlations also re- uy = 2x0 + y 0sinIt + 2y 0
+ 2
2
duce immediately to the corresponding expressions, when
we recall that in this case xuy = yux = 0, which means that in x0cosIt. 45
the absence of the magnetic field the processes x , uy and Similarly as in the preceding section, we now evaluate the
y , ux become statistically independent as expected. integrals given in Eqs. C30C34 to obtain the following
III. THE OVERDAMPED LIMITING CASE expressions for the variances at two times, that is,

In the overdamped approximation, or in the limit of large kBT t t t +t

xx = e R 1 2 e R 1 2 cos It1 t2, 46
friction, where is large, we can neglect the inertial term in m2
Eqs. 3 and 4 and therefore the resulting equation on the
complex plane can be written as kBT t t t +t
xux = e R 1 2 e R 1 2
At me
+ =

, 35

cos It1 t2



sin It1 t2 , 47
where
2 2 yux =
kBT t t t +t
e R 1 2 e R 1 2
= R i I, = R + iI , 36
me

and


cos It1 t2 + sin It1 t2 , 48
2 2
R , I , 37
2 + 2 2 + 2 kBT t t t +t
xy = e R 1 2 e R 1 2 sin It1 t2, 49
which represent the real and imaginary parts of . The solu- m2
tion of Eq. 35 and its corresponding velocity are immedi-
where e 1 + 2 and, according to Eq. 37, R = 2 / 2e
2

ately calculated, yielding the following results:

and I = 2 / e. Again, if t1 = t2 = t then the variances re-
= + 0
t
Attdt , 38
duce to
k BT 2
xx = 1 e2 t/e, 50
m2
u = u
+ 0
t
Attdt +
At

, 39
k BT 2
xux = 1 e2 t/e, 51
where the deterministic solutions and u
are given by me

= et0, = et0 .
u 40 k BT 2
yux = 1 e2 t/e. 52
The complex functions t and t are defined as m e
Here e accounts for a redefinition of the friction coefficient
t =
1 tt
e , t = ett . 41 when the magnetic field is present. This situation can be
understood in the following way. In the absence of the mag-
In a similar way, the solutions for the complex conjugates of netic field, it is evident that e = and therefore expressions
t and ut, as well as for xt, yt, uxt, and uyt are 5052 reduce, respectively, to
given in Eqs. C24C27 of Appendix C. So the statistical k BT 2
properties of the variables of the system will be calculated in xx = 1 e2 t/, 53
m2
a similar way as before. The x, y, ux, and uy averages are
shown to be
k BT 2
x = eRtx0 cosIt y 0 sinIt, 42 xux = 1 e2 t/, 54
m
y = eRty 0 cosIt + x0 sinIt, 43 and xuy = yux = 0. Thus, the description of the Brownian

051105-6
BROWNIAN MOTION OF A CLASSICAL HARMONIC PHYSICAL REVIEW E 77, 051105 2008

motion through Eqs. 5052 suggest a redefinition of the ACKNOWLEDGMENT

friction coefficient by the value e.
Partial support from PROMEP under Grant No. UAM-I-
The reader may notice that we did not give the expres-
CA-45 is acknowledged.
sions for the correlation of the velocity; actually is easy to
see that this correlation must have a term proportional to
Diracs . This follows from Eq. 39 since the velocity is APPENDIX A: SOLUTIONS FOR (t)
proportional to a combination of the Gaussian white noise, AND (t) FUNCTIONS
which is correlated. A Dirac in the correlations for the To calculate the solution of Eq. 8, we follow the same
velocity is a direct result of the approximation done, and method used by Chandrasekhar to solve the one-dimensional
does not correspond to what we obtained without the ap- harmonic oscillator 1; thus, we first calculate its homoge-
proximation. It is to be expected that, although the correla- neous solution, which is given by
tions for x-ux and y-ux remain finite, the comparison with the
exact results may not show agreement. This is actually the h = a1e 1t + a2e 2t , A1
case if one considers the large-time behavior for the exact
where a1 and a2 are constant complex numbers,
1 and
2
result and the overdamped approximation. In Fig. 6 a com-
are the roots of the homogeneous equation associated with
parison between the exact and the overdamped approxima-
Eq. 8, i.e.,
tion results for xx is shown for / = 0.01 and different
values of / . As is evident from the figure the agreement is 2 +
+ 2 = 0, A2
rather good. Notice that for the free-field case xx is given
equally well by the exact result and the overdamped approxi- and therefore the roots are given by
mation; in contrast there is a noticeable difference in the

values for xu given by the exact result and the overdamped
1 1
x
1 = + ,
2 = , A3
approximation. 2 2 2 2
and
IV. CONCLUDING REMARKS

In this work, we have been able to solve the problem of 1

2 42 = 2 2 42 + i2. A4
Brownian motion of a charged particle in the presence of a Now, we suppose that the solution of Eq. 8 is also of the
constant magnetic field in a time-independent harmonic well. form of Eq. A1, but now assuming that a1 and a2 are time-
The magnetic field is allowed to point along the z axis of a dependent functions, that is,
Cartesian reference frame and, as a consequence of this fact,
two independent processes for the charged harmonic oscilla- t = a1te 1t + a2te 2t , A5
tor occur. One process takes place along the magnetic field and also satisfy the following condition:
and is described by the Langevin equation 5. This process
was exactly solved by Chandrasekhar 1 in 1943. The other 1 t da
da 2 t
one takes place on the x-y plane perpendicular to the mag- e 1 + e 2 = 0. A6
dt dt
netic field and is described by two coupled Langevin equa-
tions. To solve these equations, we use an alternative method Substituting Eq. A5 into Eq. 8, we have
of solution relying upon a transformation of these two equa-
tions through the change of variables given in Eq. 7. This 1 t
da
da

1 e 1 + 2 2 e 2t = At. A7
method allows us to establish Eq. 8, which is then easier to dt dt
solve than Eqs. 3 and 4. We have established the general The solution of the system of equations A6 and A7 leads
auto- and cross correlations at two different times, which are to
given in Eqs. B47B52. At equal times t1 = t2 = t Eqs.
B47B52 reduce, respectively, to Eqs. 2629 whereas
xy = 0 and uxuy = 0. Our analytical results have been com- a1t =

1
2
1
0
t
e 1tAt + a10 , A8
pared with those calculated by Chandrasekhar, in the absence

of the magnetic field. In this case, the planar-diffusion pro- t
1
cess, on the x-y plane, takes place in an independent way for a2t = e 2tAt + a20 , A9
each one of the processes, along the x and y axes, as ex-
1
2 0
pected. In the overdamped approximation we have calculated
the same statistical properties, which is easier than in the full where a10 and a20 are constant. So the solution of Eq. 8 is
description. Our complete solution can be extended to that then

situation for which the particle is in a two-dimensional time- t t
1
dependent harmonic well, similar to that studied in Ref. 21 t = e 1t e 1tAtdt e 2t e 2tAtdt
in the overdamped approximation. Finally, in this large-
1
2 0 0
friction limit the explicit solution of the Smoluchowski equa-
+ a10e 1t + a20e 2t . A10
tion associated with the Jayannavar and Sahoo problem is in
progress. In a completely similar way, we can calculate the solution of

051105-7
JIMNEZ-AQUINO, VELASCO, AND URIBE PHYSICAL REVIEW E 77, 051105 2008

the Langevin equation for the conjugate variable t x 0 + 2u

iy, which is given by = 0et/2 cosh1t/2 + 0
et/2 sinh1t/2,

1
+ + 2 = At, A11 A22
where the asterisk stands for the complex conjugate. In this
case, the complex function t and its corresponding veloc-

= u et/2 cosh1t/2
u
ity ut = ux iuy are given by
0

+ 202
u

t +
1



u e 1t 0
1

u e 2t 0
et/2 sinh1t/2, A23
1 2 0 2
0 0 1

= 0
t
Attdt A12 = 0e
t/2
cosh1t/2

0 + 2u t/2
and + 0
e sinh1t/2, A24

1
1
ut +
u e 1t 1 u e 2t
20
1
2 1 0 2
0 0 t/2
= u e
u cosh1t/2
t 0

= At
tdt , A13 + 202
u t/2
0
0
e sinh1t/2. A25

where now the roots 1

1 =
+ 1, 2 =
1, A14 APPENDIX B: SOLUTIONS FOR x(t)
2 2 2 2
AND y(t) FUNCTIONS
and
By defining the complex functions t Rt + iIt,
1

= 2 42 = 2 2 42 i2. A15 with Rt and It the real and imaginary parts of t,
respectively, and similarly for
t Rt + iIt, and also
Here the functions t and
t are now
t Att,
t Att,
B1
t = 1 1tt
2tt

e e , A16
1
2 t At
t, t At
t,
B2

1 we can easily show that

t =
1tt
1e
2e 2tt.
A17
1

2 t = Re
t + iIm
t, B3
Under these conditions, the aforementioned solutions can
be written in shorthand as t = Re
t iIm
t, B4

= + 0
t
Attdt , A18 t = Re
t + iIm
t, B5

t
t = Re
t iIm
t, B6
= + Attdt , A19
0
with

and t = A A ,
Re B7
x R y I

u = u
+ 0
t
At
tdt , A20 t = A + A ,
Im x I y R B8

t = A A , B9

Re x R y I
t
u = u
+ At
tdt , A21
0 t = A + A .
Im B10
x I y R

where However, from Eqs. 16 and 17, we can also show that

051105-8
BROWNIAN MOTION OF A CLASSICAL HARMONIC PHYSICAL REVIEW E 77, 051105 2008

1 x = et/2x0cosht/2cost/2cost/2
Rt = Rt,t + It,t, B11
+ 2
2
+ sinht/2sint/2sint/2
+ et/2y 0cosht/2cost/2sint/2
1
It = 2 2 It,t Rt,t, B12 sinht/2sint/2cost/2 + et/2B0 + C0
+
sinht/2cost/2cost/2

Rt =
1
Rt,t + It,t, B13 + cosht/2sint/2sint/2 + et/2D0 E0
+ 2
2
cosht/2sint/2cost/2

1 sinht/2cost/2sint/2, B23
It = It,t Rt,t, B14
+ 2
2
y = et/2y 0cosht/2cost/2cost/2
where
+ sinht/2sint/2sint/2
1 1
Rt,t = e 1tt e 2tt + e 1tt e 2tt, et/2x0cosht/2cost/2sint/2
2 2
sinht/2sint/2cost/2 et/2D0 E0
B15
sinht/2cost/2cost/2
i i + cosht/2sint/2sint/2 + et/2B0 + C0
It,t = e 1tt e 2tt + e 1tt e 2tt,
2 2 cosht/2sint/2cost/2
B16 sinht/2cost/2sint/2, B24
and
ux = et/2ux0cosht/2cost/2cost/2
1
1e 1tt
Rt,t = 2e 2tt + sinht/2sint/2sint/2
2
1 tt
+ et/2uy0cosht/2cost/2sint/2
+
1e 1 2e 2tt,
B17
2 sinht/2sint/2cost/2 et/2F0 + G0
sinht/2cost/2cost/2
i
1e 1tt
It,t = 2e 2tt + cosht/2sint/2sint/2 et/2H0 I0
2
cosht/2sint/2cost/2
i tt
+
1e 1 2e 2tt.
B18 sinht/2cost/2sint/2, B25
2
If we now take into account that = x + iy, = x uy = et/2uy0cosht/2cost/2cost/2
= ux iuy, then, from Eqs.
= ux + iuy, and u
iy, u
A18A21 and B1B6, we can conclude that the solu- + sinht/2sint/2sint/2
tions of Eqs. 3 and 4 will be given, respectively, by
et/2ux0cosht/2cost/2sint/2

xt = x + 0
t
tdt ,
Re B19 sinht/2sint/2cost/2 + et/2H0 I0
sinht/2cost/2cost/2

t + cosht/2sint/2sint/2 et/2F0 + G0
yt = y + tdt ,
Im B20 cosht/2sint/2cost/2
0
sinht/2cost/2sint/2, B26

uxt = ux + t
tdt ,
Re B21 where

0
x0 2
B0 + 2 + 2ux0 , B27

t +
2 2

uyt = uy + tdt ,
Im B22
0

where the averages for x, y, ux, and uy will be

C0
+
2

2 y0 2


2 + 2uy0 , B28

051105-9
JIMNEZ-AQUINO, VELASCO, AND URIBE PHYSICAL REVIEW E 77, 051105 2008

D0
2 + 2

x0 2

2 + 2ux0 , B29 It1,t2 =
2
2
t1

0 0
t2
Rt1,t1Rt2,t2

E0

+
y0 2

+ 2 + 2uy0 , B30
+ It1,t1It2,t2dt1dt2 , B41

2 2
t1 t2
2
Jt1,t2 = Rt1,t1Rt2,t2

F0 2

+
2
u x0

2 + 2 + 2 2x 0 , B31
2 0 0

+ It1,t1It2,t2dt1dt2 , B42

G0

2 + 2
uy0

+ 2 y 0 ,
2 2 2
B32 Kt1,t2 =
2
2

0
t1

0
t2
Rt1,t1Rt2,t2

+ It1,t1It2,t2dt1dt2 , B43
u x0
H0 2 2 + 2 2x 0 , B33
2 + 2
Lt1,t2 =
2
2
t1 t2
Rt1,t1It2,t2

0 0
uy0
I0 2 + 2 + 2 2 y 0 . B34 It1,t1Rt2,t2dt1dt2 , B44
+
2 2

On the other hand, for the two-time auto- and cross correla-
tions we use the definition established in the text. According
to Eqs. B7B14 and B19B22, it is possible to show
Mt1,t2 =
2
2

0
t1

0
t2
Rt1,t1It2,t2

that the variances at two times are It1,t1Rt2,t2dt1dt2 , B45

xx = xt1xt2 xt1xt2
= yy = yt1yt2 yt1yt2 = It1,t2, Nt1,t2 =
2
2

0
t1

0
t2
Rt1,t1It2,t2
B35
It1,t1Rt2,t2dt1dt2 , B46
uxux = uxt1uxt2 uxt1uxt2 where 2 2 + 2. These integrals can be evaluated by us-
= uyuy = uyt1uyt2 uyt1uyt2 = Jt1,t2, ing Eqs. B15B18; the result is the following:

B36
xx =
2
2 E1t1,t2

cos1t1 t2 +
E2t1,t2
+
xux = xt1uxt2 xt1uxt2
= yuy = yt1uyt2 yt1uyt2 = Kt1,t2,
cos2t1 t2
2
2 F B FIBI ,
+ 2 R R

B37 B47

xuy = xt1uyt2 xt1uyt2 uxux =

2

E t ,t
1 1 2 cos1t1 t2
2 1 1

= yux = yt1uxt2 yt1uxt2 = Lt1,t2,
E2t1,t2 2
B38 + 2
2 cos2t1 t2 2 Re 2
1
+ + 2
xy = xt1yt2 xt1yt2
= yx = yt1xt2 yt1xt2 = Mt1,t2,
FRBR FIBI Im 2FRBI + FIBI , B48
1
uxuy = uxt1uyt2 uxt1uyt2
B39
xux =
2


2
E1t1,t2
2

cos1t1 t2 +

= uyux = uyt1uxt2 uyt1uxt2 = Nt1,t2, sin1t1 t2 + E2t1,t2 +

2 +

sin2t1 t2

where
B40
cos2t1 t2 1
F C FICI ,
+ 2 R R
2 B49

051105-10
BROWNIAN MOTION OF A CLASSICAL HARMONIC PHYSICAL REVIEW E 77, 051105 2008

yux =
2
2

2 E1t1,t2
cos1t1 t2
CR sint1 t2/2 sinht1 t2/2 + cosh
t1 t2/2 cost1 t2/2 cosht1 t2/2

+ sin1t1 t2
E2t1,t2 +
2 +

cos2t1 t2 + sinht1 t2/2, B64

+ sin2t1 t2 + 1
F D + FIDR , B50
+ 2 R I
2 CI sint1 t2/2 cosht1 t2/2 + sinh
t1 t2/2 + cost1 t2/2 cosht1 t2/2

+ sinht1 t2/2, B65
2 E1t1,t2 E2t1,t2
xy = 2 sin1t1 t2 +
+ DR sint1 t2/2 cosht1 t2/2 + sinh

sin2t1 t2 + 2
2
F E FREI ,
+ 2 I R
t1 t2/2 cost1 t2/2 cosht1 t2/2
+ sinht1 t2/2, B66
B51
DI sint1 t2/2 cosht1 t2/2 + sinh
2

uxu y = 2

1
E t ,t
1 1 1 2 sin1t1 t2

t1 t2/2 + cost1 t2/2 sinht1 t2/2
+ cosht1 t2/2, B67
E2t1,t2 2
+ 2
2 sin2t1 t2 + 2 Re 2
1
+ + 2 ER cost1 t2/2sinht1 t2/2, B68

FREI + FIER + Im 2FREI FIER , B52

1 EI sint1 t2/2cosht1 t2/2. B69

where 2 2 + 2, 1 21 , 2 21 + , and
APPENDIX C: SOLUTIONS FOR x(t) AND y(t)
E1t1,t2 et1t2/2 et1+t2/2, B53
FUNCTIONS FOR LARGE FRICTION

E2t1,t2 e+t1t2/2 e+t1+t2/2, B54 The Langevin equation for the complex function t
reads
F1t1,t2 eit1t2/2 eit1+t2/2, B55 At
+ = . C1

+it1t2/2 +it1+t2/2
F2t1,t2 e e , B56
Thus its solution and its corresponding velocity u read
ImF
FR ReF 1
,
1 FI ReF
+ ImF
1
,
1

B57
= + 0
t
Attdt , C2

= et1t2/2 cos t t /2 et1+t2/2 cos t + t /2,

ReF 1 1 2 1 2
B58
u = u
+ t

0
Attdt +
At

, C3

such that
= et1t2/2 sin t t /2 et1+t2/2 sin t + t /2,
ImF
1 1 2 1 2
= e t0, = e t0 ,
u C4
B59

1 1 t =
1 tt
e , t = ett . C5
Re 2 = 2 2 2 2,
1 B60
4 4
Again, if we define the complex functions t = R + iI and
1 t = + i , and also
2 = ,
R I
Im
1 B61
2 t Att, t Att,
C6

BR cost1 t2/2cosht1 t2/2, B62

t Att,
t Att,
C7
BI sint1 t2/2sinht1 t2/2, B63 we can show that

051105-11
JIMNEZ-AQUINO, VELASCO, AND URIBE PHYSICAL REVIEW E 77, 051105 2008

t = Re
t + iIm
t, C8 yt = y + 0
t
tdt ,
Im C25
t = Re
t iIm
t, C9 and
t = Re
t + iIm
t, C10 uxt = ux + t

0
tdt + t,
Re x C26

t = Re
t iIm
t, C11
where uyt = uy + t

0
tdt + t,
Im y C27
t = A A ,
Re C12
x R y I
where
t = A + A ,
Im C13 Axt + Ayt
x I y R
xt = , C28
2 + 2
t = A A ,
Re C14
x R y I
Ayt Axt
yt = . C29
t = A + A .
Im x I y R C15 2 + 2

Here the real and imaginary parts of and can be written Also, from Eqs. C12C19 and C24C29, we can show
as analogously as in Appendix B that at two times xx = yy
= Pt1 , t2, uxux = uyuy = Qt1 , t2, xux = yuy = Rt1 , t2, and
1 xuy = yux = St1 , t2, and xy = yx = T,where now
Rt = Rt,t + It,t, C16
+ 2
2

1
Pt1,t2 =
2
2

0
t1

0
t2
Rt1,t1Rt2,t2
It = 2 It,t Rt,t, C17
+ 2 + It1,t1It2,t2dt1dt2 , C30
1
Rt,t = t,t + It,t, C18 2
+ 2 R
2 Qt1,t2 = t1 t2 Rt1,t1 + Rt2,t1
2

It =
1
t,t Rt,t,
+ 2 I
2 C19 +
t1

0 0
t2
Rt1,t1Rt2,t2

where + It1,t1It2,t2dt1dt2, C31

1


Rt,t = ett + e tt, C20 2 t1 t2
2 Rt1,t2 = Rt1,t1Rt2,t2
2 0 0
i
It,t = ett e tt, C21 + It1,t1It2,t2dt1dt2 , C32
2

1
2

Rt,t = ett + e tt, C22
St1,t2 =
2
2
t1

0 0
t2
Rt1,t1It2,t2

It1,t1Rt2,t2dt1dt2 , C33
i
It,t = ett e tt. C23
2
Also, in a similar way as was done in Appendix B, the solu-
Tt1,t2 =
2
2
t1

0 0
t2
Rt1,t1It2,t2

tions for the real functions xt, yt, uxt, and uyt are then
It1,t1Rt2,t2dt1dt2 , C34

xt = x + 0
t
tdt ,
Re C24 where 2 2 + 2. These integrals can be evaluated with
the help of Eqs. C20C23.

051105-12
BROWNIAN MOTION OF A CLASSICAL HARMONIC
1 S. Chandrasekhar, Rev. Mod. Phys. 15, 1 1943.
2 T. P. Simes and R. E. Lagos, Physica A 355, 274 2005. 19 D. J. Evans, E. G. D. Cohen, and G. P. Morris, Phys. Rev. Lett.
3 R. Czopnik and P. Garbaczewski, Phys. Rev. E 63, 021105
2001.
4 I. Holod, A. Zagorodny, and J. Weiland, Phys. Rev. E 71,
046401 2005.
5 A. Zagorodny and I. Holod, Condens. Matter Phys. 3, 295
2000.
6 J. I. Jimnez-Aquino and M. Romero-Bastida, Phys. Rev. E
74, 041117 2006.
7 J. I. Jimnez-Aquino and M. Romero-Bastida, Phys. Rev. E
76, 021106 2007.
8 L. Ferrari, Physica A 163, 596 1990.
9 L. Ferrari, J. Chem. Phys. 118, 11092 2003.
10 S. C. Venkataramani, T. M. Antonsen, Jr., E. Ott, and J. C.
Sommerer, Physica D 96, 66 1996
11 R. Graham, M. Hohnerbach, and A. Schenzle, Phys. Rev. Lett.
48, 1396 1982.
12 M. Tureli, Theoretical Population Biology Academic, New
York, 1977.
13 H. Takayasu, A. H. Sato, and M. Takayasu, Phys. Rev. Lett.
79, 966 1997.
14 M. Gitterman, Phys. Rev. E 67, 057103 2003.
15 B. West and V. Seshadri, J. Geophys. Res. 86, 4293 1981.
16 M. Gitterman, Physica A 352, 309 2005.
17 M. Gitterman, Phys. Rev. E 69, 041101 2004.
18 L. Gammaitoni, P. Hanggi, P. Jung, and F. Marchesoni, Rev.
051105-13
PHYSICAL REVIEW E 77, 051105 2008
Mod. Phys. 70, 223 1998.
71, 2401 1993.
20 G. Gallavotti and E. G. D. Cohen, Phys. Rev. Lett. 74, 2694
1995.
21 A. M. Jayannavar and Mamata Sahoo, Phys. Rev. E 75,
032102 2007.
22 C. Jarzynski, Phys. Rev. Lett. 78, 2690 1997; Phys. Rev. E
56, 5018 1997.
23 F. Douarche, S. Joubaud, N. B. Garnier, A. Petrosyan, and S.
Ciliberto, Phys. Rev. Lett. 97, 140603 2006.
24 J. Liphardt, S. Dumont, S. B. Smith, I. Tinoco, Jr., and C.
Bustamante, Science 296, 1832 2002.
25 F. Douarche, S. Ciliberto, A. Petrosyan, and I. Rabbiosi, Eu-
rophys. Lett. 70, 593 2005.
26 G. M. Wang, E. M. Sevick, E. Mittag, D. J. Searles, and D. J.
Evans, Phys. Rev. Lett. 89, 050601 2002.
27 O. Narayan and A. Dhar, J. Phys. A 37, 63 2004.
28 E. H. Trepagnier, C. Jarzynski, F. Ritort, G. E. Crooks, C. J.
Bustamante, and J. Liphardt, Proc. Natl. Acad. Sci. U.S.A.
101, 15038 2004.
29 L. Landau and E. Lifshitz, The Classical Theory of Fields, 4th
English ed. Pergamon, New York, 1975.
30 H. Risken, The Fokker-Planck Equation: Methods of Solution
and Applications Springer-Verlag, Berlin, 1984.
31 N. van Kampen, Stochastic Processes in Physics and Chemis-
try North-Holland, Amsterdam, 1981.