ENSEÑANZA REVISTA MEXICANA DE FÍSICA E 52 (2) 182–187 DICIEMBRE 2006

Brownian motion of a charged particle in a magnetic field

J. I. Jiménez-Aquino, M. Romero-Bastida
Departamento de Fı́sica, Universidad Autónoma Metropolitana–Iztapalapa,
Apartado Postal 55-534, D.F., México,
e-mail: rbm@xanum.uam.mx, ines@xanum.uam.mx
Recibido el 31 de enero de 2006; aceptado el 18 de abril de 2006

In this work we show that Taylor’s description of Brownian motion in a magnetic field is equivalent to the situation in which the constant
magnetic field is allowed to point along any direction. This can be achieved by means of a rotation of the Langevin equation given in the
space of coordinates r, to another space of coordinates r0 , where the description of the problem is quite similar to that studied by Taylor.
We use the over-damping approximation to show why, at equilibrium, the oscillatory behavior inherent in the system is not reflected in the
diffusion process across the magnetic field, a fact not studied by Taylor. We also use the large-time approximation to study the effects of
colored noise (small correlation time) on the diffusion processes across and along the magnetic field.

Keywords: (MSD) Mean Square Displacement.

En este trabajo mostramos que la descripción teórica de Taylor del movimiento Browniano en un campo magnético, es equivalente a la
descripción teórica en la que el campo magnético constante apunta en una dirección arbitraria. Esto de logra mediante una rotación de la
ecuación de Langevin, dada en el espacio de coordenadas r a otro espacio de coordenadas r0 , donde la descripción teórica del problema
es totalmente similar al problema de Taylor. Usamos la aproximación de sobre-amortiguamiento, para mostrar porqué en el equilibrio, el
comportamiento oscilatorio, inherente en el sistema, no se ve reflejado en el proceso de difusión transversal al campo magnético, situación
que no fue estudiado por Taylor. Usamos también la aproximación en el lı́mite de tiempos largos para estudiar los efectos del ruido de color
(tiempos de correlación pequeños), en los procesos de difusión transversal y a lo largo del campo magnético.

Descriptores: (DCP) Desplazamiento Cuadrático Promedio.

PACS: 05.40.-a; 02.50.-r

1. Introduction
The stochastic diffusion of a plasma across a magnetic field
arising from the fluctuations of the electric field was solved
by Taylor in 1961 using a theoretical Langevin descrip-
tion [1]. In that work, a situation is considered in which
the density gradient of charged particles exists only in the
direction perpendicular to the external magnetic field. Due
to this physical situation, the diffusion process is described
by two coupled stochastic differential equations associated
with two components of the ion velocity vector. The diffu-
sion process is then characterized through the Mean Square
Displacement (MSD) across the magnetic field. In the next
year, the same problem on the Brownian motion in a mag-
netic field was solved by Kurs.unoǧlu [2] by making an ex-
tension to the Chandrasekhar [3] treatment of ordinary Brow-
nian motion in terms of the probability distribution function
associated with the magnitude of the velocity. In his work,
Kurs.unoǧlu assumed that the surrounding medium in which
the charged particles are diffused is the electromagnetic field
which is envisaged by a set of classical harmonic oscillators.
By assuming the existence of a fluctuating electric field in the
plasma, the charged particles in the field can be assumed to
undergo a large number of collisions per second with the os-
cillators. Therefore, the diffusion is described as a stochastic
process where some kind of dynamical friction proportional
to the velocity of the particle acts. Consequently, the usual
expressions for the diffusion processes across and along the
external magnetic field were obtained. Almost forty years
later, the problem has again become of interest to other sci-
entists, cf. [4-7]. In particular, in Ref. 4 the full descrip-
tion of the Brownian motion in the magnetic field is given
through the transition probability densities for the velocity-
space, phase-space, and the Smoluchowsky configuration-
space. In Refs. 1, 2, and 4 the theoretical developments were
pursued by assuming that the constant magnetic field vector
explicitly points along the z-axis, that is B = (0, 0, Bz ). Our
purpose in this work is to study three theoretical extensions to
Taylor’s proposal. The first one consists in showing that Tay-
lor’s problem is equivalent to a situation in which the constant
magnetic field is allowed to point along any direction, that is
B = (Bx , By , Bz ). This can easily be achieved by means of
a rotation of this magnetic field along the z 0 axis of the trans-
formed space of coordinates (x0 , y 0 , z 0 ). The second one, al-
though it would seems to be obvious, was not considered by
Taylor and consists in the following: a charged particle in a
constant magnetic field is a strictly rotational phenomenon
and the conditions under which the Brownian motion in a
magnetic field have been studied (small fluctuations of the
electric field) also correspond to a rotational phenomenon.
The question then is why, at equilibrium (large time-limit),
the rotational effects of the charged Brownian particle are not
reflected in the diffusion process across the magnetic field as
effectively shown by Taylor and Kurs.unoǧlu [1, 2]? The an-
swer to this question can easily be clarified if we focus on
the Langevin equation in the large-friction force approxima-
tion, also known as the over-damping problem. The third
case is the following: because the large-time limit or diffu-
 BROWNIAN MOTION OF A CHARGED PARTICLE IN A MAGNETIC FIELD 183

sive regime is the regime of interest, we take advantage of
the results of the over-damped problem to study the effects of
colored noise on the MSD across and along the magnetic field
for a small correlation time. Two appendices are included to
support our theoretical results, and conclusions are given at
the end of the work.
2. The Langevin equation for a plasma
The Brownian motion of a charged particle can be described
by an equation similar to the Langevin equation associated
with the velocity column vector v, which reads as [2]
e
mv̇ = −αv + v × B + eE(t) , (1)
c e 0 (s) ds .
where −αv, with α positive, is the friction force added to
the Lorentz force for a charged particle in an electromagnetic
field, and refers to a dynamical friction between the charges;
it is not independent of the fluctuating force E(t), and it ac-
tually gives rise to the stochastic fluctuations of this electric
field. The above equation can also be written as
e ,
v̇ = −β v + W v + E(t) (2)
where β = α/m, W is a real antisymmetric matrix given by
 
0 ω3 −ω2
W = −ω3 0 ω1  , (3)
ω2 −ω1 0
whose elements, defined as ωi = eBi /mc, are known as the
Larmor frequency, Bi being each component of the magnetic
field vector B where the subindex i may take on the val-
such that ω 2 = ω12 + ω22 + ω32 = e2 B 2 /m2 c2 , with
B 2 = B12 + B22 + B32 being the square modulus of the exter-
nal magnetic field B. The second term of Eq. (5) can also be
written as the cross product W 0 v0 = (e/mc)v0 × B0 , where
B0 can be visualized as an external magnetic field given by
B0 = B k̂0 , k̂0 being the unitary vector along the z 0 axis.
Therefore, in the transformed space of coordinates, the exter-
nal magnetic field points, in a natural way, along the z 0 axis.
The Brownian motion of a plasma will be better described
through the Langevin equation (5), whose solution may be
written as
v0 (t) = e−β t R(t) v0 (0)
Zt
+ R(t) e−β(t−s) RT (s) E (7)
0
To calculate the correlation function for the velocity v0 (t)
at two different times, we have to impose the initial condition
on velocity v(0). We assume that it is determined by the
Maxwell distribution function
µ ¶3/2 · ¸
m m
P1 (v(0))= exp − v(0) · v(0) , (8)
2πkB T 2kB T
in which v(0)·v(0) denotes a scalar product, kB is the Boltz-
mann constant, and T is the temperature of the surrounding
medium. It is also assumed that there is no correlation be-
e
tween v(0) values and E(t) for any t ≥ 0. Because it will
be necessary to average over v0 (0) using the Maxwell dis-
tribution, as well as averaging stochastically with respect to
the noise E e 0 (t), we use the notation {· · · } for the averages
over v (0). In this case, due to the transformations of v0 (t)
0

and Ee 0 (t), it can also be shown that there is no correlation

ues 1, 2, 3 representing the coordinates x, y, z respectively.
e
The fluctuating electric field E(t) = (e/m)E(t) satisfies the between v0 (0) and E e 0 (t). Also
properties of Gaussian white noise with zero mean value and
hv0 (t)i = e−β t R(t) v0 (0) and {hv0 (t)i} = 0 . (9)
correlation function
ei (t)E
ej (t0 )i = 2q δij δ(t − t0 ) , The correlation function at two different times for the com-
hE (4) ponents of velocity v0 (t) reads as
q being the noise intensity. {hvi0 (t1 )vj0 (t2 )i} = e−β (t1 +t2 ) Rik (t1 )Rjl (t2 ) {vk0 (0)vl0 (0)}
T T
If we make the change of variable v0 = R v, where R
is the transpose of the rotation matrix R given in Appendix A, + Rik (t1 )Rjl (t2 )hhk (t1 ) hl (t2 )i, (10)
then Eq. (2) is becomes
where the stochastic average is
0 0 e 0 (t) ,
v̇ = −β v + W v + E 0 0
(5) Zt1 Zt2
0 0

e 0 (t) = R T E(t)
e T hhk (t1 ) hl (t2 )i = e−β(t1 +t2 −t1 −t2 )
where E and W 0 = R WR is another an-
0
0 0
tisymmetric matrix such that R(t) = eW t is in general an
T e 0 (t0 )E
×(RT )km (t01 )(RT )ln (t02 ) hE e 0 (t0 )i dt0 dt0 . (11)
orthogonal rotation matrix satisfying R (t) = R−1 (t), i.e. m 1 n 2 1 2
0
the transpose is its inverse and therefore R−1 (t) = e−W t . Using the Maxwell distribution (8), it can be shown that
Such matrices, as shown in Appendices A and B, are given
by kB T
{vk0 (0)vl0 (0)} = δkl (12)
    m
0 ω 0 cos ωt sin ωt 0
and, using the correlation function (4), we can also show that
W 0 = −ω 0 0 , R(t)=− sin ωt cos ωt 0 , (6)
0 0 0 0 0 1 e 0 (t0 )E
hE e 0 (t0 )i = 2q δmn δ(t0 − t0 ) . (13)
m 1 n 2 1 2

Rev. Mex. Fı́s. E 52 (2) (2006) 182–187

184 J. I. JIMÉNEZ-AQUINO AND M. ROMERO-BASTIDA

Therefore, the stochastic average (11) reduces to cross correlations can be calculated using Eq. (15). For the
x0 (t) component we have, after some algebra, that
Zt1 Zt2
h(∆x0 )2 i ≡ h[x0 (t) − x0 (0)]2 i
0 0
hhk (t1 ) hl (t2 )i = 2q e−β(t1 +t2 −t1 −t2 )
0 0 · Zt Zt ¸
q
×(R T
)kn (t01 )(RT )ln (t02 ) δ(t01 − t02 ) dt01 dt02 . (14) = cos ω(t1 − t2 ) e−β |t1 −t2 | dt1 dt2 , (19)
β
0 0
So, according to Eqs. (12) and (14), the correlation func-
tion (10) is then which can be readily evaluated, yielding:
µ ¶
D q
kB T −β (t1 +t2 ) h(∆x0 )2 i = 2 t− (1 − e−Λ1 t )
{hvi0 (t1 )vj0 (t2 )i} = e Rik (t1 )Rjk (t2 ) 1 + (ω/β) 2 β Λ21
m
q £ ¤ q
+ Rik (t1 )Rjk (t2 ) e−β |t1 −t2 | − e−β (t1 +t2 ) . − (1 − e−Λ2 t ) (20)
β β Λ22
(15)
where D=q/β 2 =kB T /mβ, which is consistent with the
The constant q can be calculated from the fluctuation- value of the Einstein coefficient, Λ1 =β−iω, and Λ2 =β+iω.
dissipation relation [8], which is related to the average kinetic The same expression is obtained for the y 0 (t) component, that
energy at equilibrium. This average can be calculated from is h(∆y 0 )2 i≡h[y 0 (t) − y 0 (0)]2 i=h(∆x0 )2 i. The MSD for the
Eq. (15) at the time t1 = t2 = t, yielding z 0 (t) component is simply

1 3k T 3q m £ ¤ h(∆z 0 )2 i ≡ h[z 0 (t) − z 0 (0)]2 i

m {hv 0 2 (t)i} = B e−2β t + 1 − e−2β t , (16)
2 2 2β · Zt Zt ¸
q −β |t1 −t2 |
where v 0 2 (t) is the square modulus of velocity v0 (t). As time = e dt1 dt2 , (21)
β
goes to infinity, the Brownian particle attains thermal equilib- 0 0
rium with the surrounding medium in which it is immersed. which leads to the following result
Consequently, this average kinetic energy should be 3kB T /2.
Eq.(15) agrees with this value if and only if q = β kB T /m, 2q
h(∆z 0 )2 i = 2 D t − (1 − e−β t ) , (22)
which is precisely the fluctuation-dissipation relation. Once β3
this relation is used in (16), it is seen that, for all time, which is also an expected result. On the other hand, it can be
shown that cross correlation functions
1 3 kB T
m {hv 0 2 (t)i} = , (17)
2 2 m h(∆x0 )(∆y 0 )i = h(∆y 0 )(∆x0 )i = 0,
which exhibits one aspect of the stationarity of the process. h(∆x0 )(∆z 0 )i = h(∆z 0 )(∆x0 )i = 0,
Therefore, in the transformed space of velocities, the plasma
diffusion also satisfies the same fluctuation-dissipation rela- and
tion as that of the usual Brownian motion.
h(∆y 0 )(∆z 0 )i = h(∆z 0 )(∆y 0 )i = 0.
2.1. The Mean-Square Displacement From the expressions given in Eqs. (20) and (22), we can
make the analysis for short and large times. For short times,
In the transformed space of velocities v0 , it is clear that such that βt ¿ 1 and ωt ¿ 1, we get the following expres-
T
v0 (t) = dr0 (t)/dt, where r0 (t) = R r(t). So, if at time sions for the three components:
t = 0 the particle is at r0 (0), then the MSD for the vector
r0 (t) is given by q
h(∆x0 )2 i = h(∆y 0 )2 i = h(∆z 0 )2 i = t2 ; (23)
β
3
X therefore,
h[r0 (t) − r0 (0)]2 i ≡ h[ri0 (t) − ri0 (0)]2 i
i=1 3
X µ ¶
0 0 2 kB T
h[r (t)−r (0)] i= h[ri0 (t)−ri0 (0)]2 =3 t2 . (24)
3 Zt Zt
X m
l=1
= {hvi0 (t1 )vi0 (t2 )i} dt1 dt2 . (18)
i=1 0 This result shows that, in this regime of approximation,
0
the MSD is the same for the three components and propor-
To calculate this quantity, we must calculate the MSD tional to t2 , which corresponds to the behavior of a free par-
for each component of vector r0 (t). If we define the com- ticle. This means that, in this limit of approximation, the
ponents of this vector as r10 (t) ≡ x0 (t), r20 (t) ≡ y 0 (t) and plasma is not sensitive to the surrounding medium in which
r30 (t) ≡ z 0 (t), then the MSD for each component and their it is immersed. For large times, such that βt À 1, ωt À 1

Rev. Mex. Fı́s. E 52 (2) (2006) 182–187

 BROWNIAN MOTION OF A CHARGED PARTICLE IN A MAGNETIC FIELD 185

and also ω < β we have, for the x0 and y 0 components, the The large friction-force approximation (30) is well known
following expression: as the over-damped problem, in which W 0 −1 can be called
µ ¶ a “diffusion coefficient matrix”. It is now clear that, in the
D
h(∆x0 )2 i = h(∆y 0 )2 i = 2 t, (25) dynamical evolution of Eq. (30), there is no rotational effect.
1 + (ω/β)2 Here, the MSD can simply be calculated by solving Eq. (30)
and for the z 0 component, in terms of the vector r0 (t). In this case,

h(∆z 0 )2 i = 2 Dt . (26) 3
X
h[r0 (t) − r0 (0)]2 i = h[ri0 (t) − ri0 (0)]2 ]i
Finally, the MSD will be i=1
·µ ¶ ¸ 3
X Zt Zt
0 0 2 2 f 0 −1 )2ik
h[r (t) − r (0)] i = + 1 2D t . (27) = 2q (W δ(t1 − t2 ) dt1 dt2 . (32)
1 + (ω/β)2 i,k=1 0 0
The expressions given by Eqs. (25) and (26) are the MSD
So, for each component we conclude that
across and along the magnetic field respectively, and they are
µ ¶
proportional to t. Therefore, in this regime of approximation, 0 2 0 2 D
which corresponds to the diffusive regime, the plasma enters h(∆x ) i = h(∆y ) i = 2 t, (33)
1 + (ω/β)2
in contact with the surrounding medium through the inces-
sant collisions with the oscillators. We can also see, from and
Eqs. (25) and (26), that the MSD’s are related by h(∆z 0 )2 i = 2 D t . (34)

h(∆z 0 )2 i β 2 h(∆z 0 )2 i The rest of the cross correlation functions are equal to zero.
h(∆x0 )2 i=h(∆y 0 )2 i= = , (28)
1 + (ω/β)2 β 2 + ω2
2.3. The colored noise problem
and the maximum value of the MSD across the magnetic field
occurs at β = ω, that is For this problem, the fluctuating electric field E(t) e satis-
µ ¶ fies the properties of Gaussian colored noise with zero mean
0 2 0 2 kB T
h(∆x ) imax = h(∆y ) imax = ct , (29) value and correlation function [13, 14]
eB
hE ej (t0 )i = q δij e−|t−t0 |/τ ,
ei (t)E (35)
where T is the temperature of the diffusing particle at equi- τ
librium.
As can be seen, for intermediate times the MSD with τ the correlation time of noise.
across the magnetic field, given by Eq. (20), con- Because the large-time limit is the time interval of inter-
tains the rotational effects of the system through est, we study the colored noise problem in the over-damped
the factors e−Λ1 t = (cos ωt + i sin ωt) e−β t and approximation given by Eq.(30). In this case, the MSD is
e −Λ2 t
= (cos ωt − i sin ωt) e−β t , which clearly disappear now written as
in the large time limit. For this reason the expression given 3
X
by (25) does not contain these rotational effects. h[r0 (t) − r0 (0)]2 i = h[ri0 (t) − ri0 (0)]2 ]i
i=1
3 Zt Zt
2.2. The over-damped problem q X f 0 −1 2 0
= (W )ik e−|t−t |/τ dt1 dt2 . (36)
Expressions (25) and (26) describe the plasma diffusion at τ
i,k=1 0 0
equilibrium and they do not properly exhibit the rotational
character of the system. This physical situation can be un- By evaluating the integral we have, for the x0 component
derstood if we pay attention to the solution to the Langevin µ ¶· µ ¶¸
D
equation in the large frictional force approximation. In this h(∆x0 )2 i = 2 t − τ e −t/τ
− 1 , (37)
1 + (ω/β)2
limiting case, the time derivative of Eq.(5) can be neglected,
resulting in the following approximation: and the same expression for the y 0 component. The MSD for
e 0 (t) ,
f 0 −1 E the z 0 component is
v0 = W (30) · µ ¶¸
0 2 −t/τ
f 0 −1 is defined as the inverse of the matrix
where W h(∆z ) i = 2 D t − τ e −1 . (38)
f = βI − W 0 , such that
W 0
 β  The rest of the cross correlation functions are equal to
ω
β 2 +ω 2 β 2 +ω 2 0 zero. In the limit of the small correlation time such that
 
  t À τ , we get

f 0 −1 =  2−ω 2 β 
W β +ω β 2 +ω 2 0 . (31) µ ¶
  D
  0 2 0 2
h(∆x ) i = h(∆y ) i = 2 (t + τ ) , (39)
0 0 1 1 + (ω/β)2
β

Rev. Mex. Fı́s. E 52 (2) (2006) 182–187

186 J. I. JIMÉNEZ-AQUINO AND M. ROMERO-BASTIDA

two equations. So, for a large-time limit, the MSD across the
magnetic field (25) is similar to that obtained by Taylor and
Kurs.unoǧlu, except for the Larmor frequency ω. It reduces
to the Taylor’s and Kurs.unoǧlu’s results if B1 = B2 = 0,
for which ω = eB3 /mc. According to this result, the dif-
fusion process across the magnetic field defines an effective
diffusion constant given by De = D/[1 + (ω/β)2 ].
We have also shown that, in the large-time limit, the dif-
fusion process is essentially equivalent to the over-damped
problem, which yields to the Langevin approximation (30).
The solution to this equation is clearly not rotational, as
shown by Eq.(32), and therefore the plasma diffusion across
the magnetic field is described without any oscillatory behav-
F IGURE 1. MSD across the magnetic field for each of the charged ior, as can be corroborated in the numerical simulation results
particle position components in the transformed space. The theoret- displayed in Fig. 1.
ical result given by Eq. (33) is the dashed line, and the simulation Finally, we take advantage of the over-damped approx-
results are the circles and squares. imation to calculate the effect of colored noise on the dif-
fusion processes across and along the magnetic field for
and small τ , yielding to the approximations (39) and (40). The
h(∆z 0 )2 i = 2 D(t + τ ) . (40) non-markovian contribution to the MSD’s in this approxima-
Obviously, for τ = 0 the expressions (39) and (40) reduce tion is a time translation for small τ values.
to the white noise case.
Acknowledgments
2.4. Numerical simulation of the over-damped problem
The authors wish to thank R. M. Velasco for her valuable
In order to corroborate the plasma diffusion across the mag- comments and suggestions. Financial support from the Con-
netic field, we perform a numerical study of the problem in sejo Nacional de Ciencia y Tecnologı́a (CONACYT) México
the over-damping case for white noise. If we rescale the po- is also acknowledged. acknowledgments
sition and time variables as r0 → r0 /lm and t → ωt, where
lm = (mc2 /B 2 )1/3 and ω is the aforementioned Larmor fre-
quency, we obtain a dimensionless expression for Eq.(5), in A Appendix: Transformation of matrix W
which we take ω = 30. These equations were written as a
Given the antisymmetric matrix
multivariate Langevin equation for which a suitable integra-  
tion algorithm was proposed in Ref. 9. The simulation pa- 0 ω3 −ω2
rameters were taken as ∆t = 0.001 for the integration time- W = −ω3 0 ω1  , (A.1)
step, β = 100.0 and q = 5 000, which guarantee that we are ω2 −ω1 0
simulating the over-damping problem, and the position vari-
ables represent a driftless Wiener process [10]. The results it can be transformed into another antisymmetric matrix W 0 ,
of the theoretical expression (33) agree with the simulation which defines a new reorientation of the magnetic field B.
results, as can be appreciated in Fig. 1. The transformation can be achieved by means of a rotation
matrix R which is composed by the unitary eigenvectors of
matrix W [11]. Therefore, the rotation matrix will be
3. Conclusions  −ω1 ω3 
√ √ ω22 2 ωω1
ω ω12 +ω22 ω1 +ω2
By means of a rotation of the Langevin equation given by  
 
Eq. (2), we have shown that the Brownian in a constant mag-  −ω ω −ω1 ω2 
 √ 2 3 √ 
netic field, allowed to point along any direction, is equivalent R =  ω ω2 +ω2 ω12 +ω22 ω  , (A.2)
 1 2 
to that studied by Taylor, as effectively shown by Eq. (5).  
 ω2 +ω2 
In our case, the two x0 and y 0 velocity components are cou- √1 2 2 2 ω3
ω ω +ω
0 ω
pled and are very similar to those proposed for Taylor, ex- 1 2
cept by the expression of the Larmor frequency, which in
where the first and second column are the real and imaginary
our case is ω = eB/mc, where B 2 = B12 + B22 + B32 .
part of one of the two complex eigenvectors of W. So, the
This is the frequency with which the charged particle ro- T
following transformation R WR leads to
tates around the z 0 -axis of the transformed space of coordi-
 
nates before it reaches a state of equilibrium state as time 0 ω 0
goes to infinity. The z 0 velocity component satisfies the usual W 0 = R T WR = −ω 0 0 , (A.3)
Brownian motion equation and is independent of the other 0 0 0

Rev. Mex. Fı́s. E 52 (2) (2006) 182–187

 BROWNIAN MOTION OF A CHARGED PARTICLE IN A MAGNETIC FIELD 187
T
where ω 2 = ω12 + ω22 + ω32 and R is the transpose of matrix where I is the unit matrix. Collecting both the odd and even
T
R, such that R R = I . terms and taking into account that
 
1 0 0
B Appendix: The rotation matrix R(t) Mz2n = S = 0 1 0 , Mz2n+1 = Mz , (B.3)
0 0 0
0
To show that the relation eW t = R(t) is a rotation matrix,
we follow the proposal of Ref. 12. For this purpose, we define where S is a symmetric matrix and iMz2n+1 = A, it can be
the rotation angle φ = ωt and the matrices shown that Eq.(B2) reduces to
    0
0 −i 0 0 1 0 eW t
= I + (cos φ − 1)S + sin φ A; (B.4)
Mz =  i 0 0 , A = iMz = −1 0 0 , (B.1) 0

0 0 0 0 0 0 therefore, by defining R(t) = eW t , we finally have that

 
cos ωt sin ωt 0
such that A is real and antisymmetric and therefore
R(t) = − sin ωt cos ωt 0 , (B.5)
W 0 t = iφMz . Using the property of the exponential, we have
0 0 1
0 (iφMz )2 (iφMz )3
eW t =eiφMz =I +iφMz + + + · · · (B.2) T
is an orthogonal rotation matrix because R(t)R (t) = I .
2! 3!

1. J.B. Taylor, Phys. Rev. Lett. 6 (1961) 262. 9. D.T. Guillespie, Am. J. Phys. 64 (1996) 1246.
2. B. Kurs.unoǧlu, Ann. Phys. 17 (1962) 259.
10. D.T. Guillespie, Phys. Rev. E 54 (1996) 2084.
3. S. Chandrasekhar, Rev. Mod. Phys. 15 (1943) 1.
4. R. Czopnik and P. Garbazewsky, Phys. Rev. E 63 (2001) 11. E. Piña, Acta Mexicana de Ciencia y Tecnologı́a VII (1973) 55.
021105.
12. E. Piña, Dinámica de Rotaciones (Universidad Autónoma
5. T.P. Simões and R.E. Lagos, Physica A 355 (2005) 274. Metropolitana, México, 1996).
6. H. Hayakawa, Physica D bf 205 (2005) 48.
13. G.E. Uhlenbeck and L.S. Ornstein, Phys. Rev. 36 (1930) 823.
7. L. Ferrari, J. Chem. Phys. 118 (2003) 11092.
8. N.G. Van Kampen, Stochastic Processes in Physics and Chem- 14. M.C. Wang and G.E. Uhlenbeck, Rev. Mod. Phys. 17 (1945)
istry (North-Holland, Amsterdam, 1981). 323.

Rev. Mex. Fı́s. E 52 (2) (2006) 182–187