You are on page 1of 7

PHYSICAL REVIEW A

VOLUME 47, NUMBER 6

JUNE 1993

Quantum

motion of a charged particle in a Paul trap


Fu-li Li

Institut fiir Theoretische Physik I, Universitiit Miinster, 4400 Mu nste'r, Germany and Department of Physics, Xian jiaotong University, Xian 7I0049, People's Republic of China (Received 3 August 1992)

A numerical method to solve the Schrodinger equation for a charged particle in a Paul trap is suggested. The quantum secular motion approximation is examined by comparing the exact numerical solution with the approximate result. The stability of quantum motion of the particle is discussed.

PACS number(s): 32.80.Pj, 03.65.Ge, 33.80.Ps, 42. 50.Vk

I. INTRODUCTION
The Paul trap has been recognized as a powerful tool in the study of the dynamics of systems that are composed of a few or even a single charged particle [1]. To constrain charged particles in a Paul trap effectively, it is necessary in experiments to reduce the kinetic energy of stored particles. Recent experiments make it possible to place a single ion in the quantum ground state in the trap by means of the laser-cooling technique [2]. Therefore, a treatment of motion of completely quantum-mechanical particles stored in a Paul trap is of great interest. A number of works have been devoted to this subject in recent years [3 6]. For the classical motion of a charged particle in a Paul trap, the secular motion approximation has been commonly used [7]. Cook, Shankland, and Wells [4] proposed a quantum version of the classical secular motion approximation and showed that it should be valid if the frequency of the time-dependent quadrupole potential of the trap is high on the basis of the first-order perturbation theory. Because a rigorous examination of the quantum secular motion approximation is absent, its validity has been questioned [3]. When a charged particle is stored, in other words, in a stable motion state in the trap, it means in classical mechanics that the amplitude of the motion is limited and in quantum mechanics that the width of wave packet of the particle is finite at all times. Combescure [3] noticed that the Heisenberg equation for the coordinate operator is formally the same as the classical equation of motion for the coordinate and then concluded that the stability regions for the quantum motion are exactly the same as the stability regions for the solution of the associated Mathieu equation. That argument is not sufficient because the same form of the linear equations for both the coordinate and the related operator only promises that the motion of the center mass of a wave packet in the trap is completely the same as the classical motion of a charged particle in the trap, and only by virtue of that observation one cannot say anything about the diffusion of a wave packet in the trap. On the other hand, the exact solution of the Schrodinger equation for a charged particle in a Paul trap seems to have been found in previous studies [3,5, 6], but in fact only the relation between solutions of the Schrodinger equation and the associated Mathieu equation was estab1050-2947/93/47(6)/4975(7)/$06. 00
47

lished. In order to acquire the explicit quantum solution and evaluate the physical quantities, one first of all has to solve the Mathieu equation and then do some auxiliary calculations. It is not convenient for the practical application. Therefore, directly solving the Schrodinger equation for a charged particle in a Paul trap would be worthwhile. To tackle the problems as mentioned above, we will in this article develop a numerical approach to finding the exact solution of quantum motion of a charged particle in a Paul trap and examine the validity of the quantum secular motion approximation by directly comparing the exact numerical solution with the approximate result. We will also discuss the stability of quantum motion of a charged particle in a Paul trap.

II.

NUMERICAL APPROACH

AND SECULAR MOTION APPROXIMATION

In a Paul trap the quadrupole force for constraining a charged particle results from the time-dependent potential

4(t)=[V + V

cos(Qt)]

x +y 2z 0+2zp

with ro

=2zo

where 2ro and 2zp are the internal ring diameter and the distance of the two caps of the trap, Uo and Vp denote amplitudes of the dc and ac voltage between the ring and the end caps, and Q is the driving frequency of the ac voltage [1]. In Cartesian coordinates, the quantum motion of a particle of charge e and mass m in the trap is governed by the Schrodinger equations
aq
l

(x, t) at

2m

qr

(x, t)

eUo
2rp~

1+

Vp

Up

cos(Qt) x qr(x, t),

(2)

+ evo 1+ 2
2rp2

Vp

Up

cos(At) y y (y, t),

(3)

4975

1993

The American Physical Society

4976

FU-LI LI
1

a~, (z, t)
Bt

(z, t) 2m P, y,
e Uo

r2

1+

Vo

where ~n~ ) is an eigenstate of the number operator a a. with the eigenvalue n . In the algebraic representation, Eqs. (2) can be (4) identically written as

cos(At)

y, (z, t) .

(4)

Bt

=Hie(t) &,
(14)

For the convenience of numerically solving Eqs. (2) we first transfer them into an algebraic represen(4), tation. To this end, we denote Hamiltonians for the
motion along the three Cartesian axes by

where

H,

2m'
g,

P~

1 + A, [1+Pocos(At ) ]g, m

j =x,y, z,

(S)

where and

g,

and g, represent
ZeUO

x, y, and z, respectively,

and (10), and b (t) and bo(t) are real continuous functions of time. In order to solve this equation we introduce an evolution operator 0(t) such that
~%(t) ) = U(t) ~%(0) ) with U(0)

=b(t)(J++J )+bo(t)Jo, and Jo obey the commutators J+,

J,

=1,

eUo
ozro

mro

Po= I'o~Uo .

(6)

We then define the operators

Qm

co&

In terms of these annihilation and creation operators the Hamiltonians can be rewritten as
CO.

where %(0) ) is the initial state vector of the system. Oband Jo also form a closed algebraic set viously, like J+ and Jo. Since the Hamiltonian is a linear function of and Jo, the evolution operator can be represented locally by a product of three exponential and Jo [8]. Because the evolution operators of operator must be unitary, we write it in the form
~

J++J

J++J

J++J

U(t) =exp[ia, (t )Jo]exp[ip, (t )(J++J )]


X exp[y, (t )( J+

[1+/3ocos(At)] (J+

+J,

J )],

(16)

+Q) .
where

1+

[1+/3ocos(At ) ]

Joi,
(9)

where a, (t ), /3, (t), and y, (t) are real functions of time to be determined. Substituting (16) and (15) into (13) we obtain the three differential equations for a, (t), p, (t), and

y, (t):

bo(t) 2y, a, = sinh(2P,

),
),

(17)

In Eqs. (5) and (8) the upper sign is for the x and y axes and the lower one for the z axis. It can be shown that the operators (9) satisfy the commutation relation

b(t)cosa, , P, = b(t)sina, cosh j, =


with the initial conditions

(18)

'(2p,
0

(19)

Therefore the Hamiltonians are linear functions of operators of the su(1, 1) algebra. Suppose that at t =0 the particle is in a state specified by the wave functions

2Jo~, [Joj,J+i]=+J+, . [J+J,J ~]= Hence J+ and Jo. form an su(1, 1) algebra.

(10)

ai~~=o

Pi~~=o

Xt ~=o

(20)

y ($, , 0) =

g aNexp( , m co~gj )M(Qmcoj g, ), j j


'

linear superposition coeKcients, Hand J Nare Hermite polynomials and normalization factors, l and co are the frequencies of harmonic oscillators whose wave functions are used as the basis vectors for the above expansions. In the algebraic representation, the initial wave functions become
where
~y,

aare j

&= ga~n, &,

(12)

By solving these equations we can determine the evolution operator and obtain the vector ~%(t)) if the initial vector ~%(0)) is known. In this way, the expectation values of various dynamical quantities of the system can be evaluated. Equations (17) (19) are easily solved by use of the Runge-Kutta method. In the coordinate representation, it can be seen that the Hamiltonian (14) represents a harmonic oscillator with time-dependent mass and frequency. The solution of the Schrodinger equation for this time-dependent harmonic oscillator has been extensively investigated in recent years [9, 10]. When the frequency is a linear function of time and the mass is a constant, Agarwal and Kumar [10] found the exact analytical result on the time evolution of the density matrix and the Wigner function for the timedependent harmonic oscillator. When the frequency and the mass are arbitrary functions of time, previous authors [9] employed the evolution operator method developed by Wei and Norman [8] to treat the problem, but they chose the realization of the su(1, 1) algebra different from

47

QUANTUM MOTION OF A CHARGED PARTICLE IN A PAUL TRAP


1

4977
eVO
2 2 p'o

(9). In comparison with the realization of the su(1, 1) algebra which was adopted by the previous authors, the present realization has a few advantages. First, since an operator that is a function of the coordinate and momentum operators can be expressed in terms of the creation and annihilation operators (7), except for having to solve (19) numerically, other calculations such as Eqs. (17) finding the expectation value of an operator are only some simple and direct algebraic manipulations if the initial state is given in the form of (ll). Second, it is clearly observed from (17) (19) that if bo(t) and b(t) are bound, apand y, will take limited values at any given time since apand y, are finite at all times. Therefore, the present approach allows one to study the time evolution of the system within any given time period by means of calculation. In addition, numerical only the first differentiation and onefold integral are involved in solv(19). In the next section, it will be seen that the ing (17) above advantages are absent in the realization of the su(1, 1) algebra, which was commonly adopted in literature [9]. As an example of application of the present approach, we now consider the case in which the dc voltage U~ is absent and the particle is initially in the ground state of a harmonic oscillator of frequency ~o and mass m. Under this initial condition the expectation values of x, y, and z will be zero at all times. Hence the variances ((bx ) ), ((Ay ) ), and ((hz ) ) are employed to measure the stability of the motion of the particle along the three axes. In our calculations we take (m coo) ' and coo ' as variance and time unity, respectively. Because the motion along the y axis is the same as along the x axis in the present case, we here only give the results for the x and z axes. In Figs. 1 and 2 the time evolution of the variances ((bx ) ) and ((hz) ) for 0/coo=6. 0 with different values of q =eVo/mrocoo is depicted. In these figures it is clearly observed that the oscillations of ((b,x) ) and ((b,z) ) are composed of a fast micromotion and a slow secular motion. As the parameter q increases the amplitude of the secular motion decreases but the frequency rises. When q reaches a certain value the frequency of the secular motion is close to that of the micromotion and the secular motion disappears, and then ((b,x ) ) and ((b,z ) ) oscillate fast with the small amplitudes [see Figs. 1(d) and 2(d)]. However, with further increase of q the amplitudes of ((b,x) ) and ((b,z) ) rise again as shown in Figs. 1(e) and 2(e). These results indicate that for a fixed 0/coo we can constrain the particle in the trap by choosing an adequately large amplitude of the ac voltage. In our calculations we find that for a given value of q the amplitude of the secular motion increases with the value of 0/ct)o and for a larger value of 0/coo the secular motion exists in a larger varying region of q . Therefore, the amplitude, frequency, and existence of the secular motion are effectively controlled by the parameters Q/coo

2m

P, qI(x, t ) +

cos( Qt )x qI, (x, t

),

(21)

p, q,1(z, t)

e~o

cos(IIt)z 2 Ip, (z, t) .

By setting [4]

(a)
A

g. p

6.0

0. 0 0.0

(b)
8.0
&Vc

4. 0

p p

0.0

10.0

20. 0

30.0

40. 0

50.0

(c)
A

2. 0
V

p p

0.0

10,0

20. 0

30.0

40.0

50.0

80.0

1.P
V

0.5 0.0 0.0

(e)
A

9.0
6.0

IIILJI

llew)

andq . In the above we have seen that the secular motion is very regular. Therefore it is possible to find an effective time-independent potential being responsible for the secular motion. If the dc voltage is absent Eqs. (2) and (4) become

FIG. 1. Time evolution of variance ((bx ) ) with Q/coo=6. 0. The solid and dashed lines represent the exact and secular approximate results, respectively. (a) q = 2. 0; (b) q =3.0; (c) q =5.0; (d) q =10.0; (e) q =16.0.

4978

FU-LI LI

47

pf( x, t )

=p

(x, t)exp

eVp i
eVp

20rp

x sin(At

(23)

BP

(x, t)
Bt

2m

p p(x t)
eVp
Qyp
2

y, (z, t)=((I, (z, t)exp i


and inserting
we

Qrp

z sin(At)

(24)

+ 2m
+i.
eVo

x P, (x, t)sin(At)
a

obtain

, (z, t):
I
I

(23) and (24) into (21) and (22), respectively, the equations of motion for P (x, t ) and
l

m Qrp

+ P

(x, t)sin(At),
(25)

ay, (z, t) at

2m

p, p, (z,
1

t)

12.0

2e Vp

9.0
&H

2m

Qrp
z

P, (z, t)sin (At)


.
(26)

6.0

3.0
0.0 0.0
80.0

2e Vp
m
30.0
40.0

Qyp2

()z

+ P, (z, t)sin(At) 2

50.0

(b)
A

frequency 0 in these equations is large, one can discard terms that oscillate rapidly with average value zero. In this case there result the ordinary Schrodinger equations for P, (x, t) and P, (z, t):

If the

I V

2. 0

1.0
p
I

l
I

BP(x,t)
Bt
2m

PP, (x, t)+ V,s(x)P(x, t),


P,'P, (z, t)+ V, ~(z)P, (z, t),
potentials

(27)

0.0

10.0

20.0

30.0

40.0

50.0

eo. o

l
40
I

ay, (z, t) at

2m

(28)

(c)
3.0
A
(H

with the effective time-independent


2

1.0
o.o

V, (t(x) =
10.0
20.0

eVp

4m

~y~
2e Vp

(29)

0.0

30.0

40.0

50.0

6o.o

V, s(z) =
I
I

4m

(30)

Qyp2

1.5
A

I V

1.0

p p

0.0

80.0

30.0

60.0

40.0
3O. O

80.0

.. III' Ill)Ill/IIIii
.

lA. Afi XIII


30.0
40.0

IiiiiJI

Flex. 2. Time evolution of variance ((iffz)2) with 0/coo=6. 0. The solid and dashed lines represent the exact and secular approximate results, respectively. (a) q =1.0; (b) q =2. 0; (c)
q2

5 2

(d) q2

3 0 (e) q2 0 8

The numerical results which are obtained from (27) and (28) are represented by the dashed curves in Figs. 2(c). It is observed that the approxi1(c) and 2(a) 1(a) mate results are nicely in agreement with the exact numerical calculations. For a fixed parameter 0/cop the approximation gradually becomes bad as the parameter q increases. For a larger value of A/Mp we find that the region in which the approximation is valid becomes larger. These results show that the validity of the approximation is actually determined by the parameter eVo/(mroA ). when becomes The approximation very good eVo/(mroA ) is small, i.e. , A(eVo/mro)' It should be emphasized that under the condition the whole motion of the particle is A(eVolmro)' well composed of the micromotion and the secular motion and the static effective potentials (29) and (30) describe the secular motion very well, but the validity of the secular approximation never means that the wave functions obtained from the static potentials (29) and (30) should be very well equal to the exact wave functions at any time [3].

QUANTUM MOTION OF A CHARCTED PARTICLE IN A PAUL TRAP

4979

III. STABILITY OF QUANTUM MOTION


IN A PAUL TRAP

In this section, we will construct the solution of Eqs. (2) in terms of the real solution of the Mathieu equa(4) tion and then discuss the stability of quantum motion of a charged particle in a Paul trap. To this end, let us consider the general form of Eqs. (2) (4):

. a~(, t)
at
2m

[a+b cos(IIt )]g y(g, t ), P y(g, t )+


(31)

tegral has to be evaluated to determine y2. Therefore, in comparison with the approach given in the previous section, the present method is not practicable if one fails to find the analytic solution of these equations. However, the present method has the advantage which allows one to construct the formal solution of (31) in terms of the solution of the Mathieu equation. Suppose that F(t) is a real function of time and satisfies the Mathieu equation

F(t)+

cos(At)]F(t)=0, [a+b
F(t)

(41)

where a and b may be any constants and g may represent one of the three Cartesian coordinates. By de5ning the three operators [11]

with the initial conditions F(0) =0 and F(0) = l. In this (39) can be expressed as case, the solution of (37)

J+ = P
the Hamiltonian

J =

Pg

Jo=

az(t) =m
/3z(t)

(Pg+ kPg
(33)

(42)

= lnlF(t) l, 2

(43)
(44)

in (31) can be written as

Il=a, J +a2J+,
where

y, (t) =

a, =

,az= cos(Qt)] . i[a+b

(34)

It can be shown that the operators (32) also satisfy the commutator relation (10). Therefore, they form another realization of the su(1, 1) algebra. To solve the Schrodinger equation (13) with the Hamiltonian (33), we choose the evolution operator defined in (15) as
U(t) =exp[az(t) J+ ]exp[/3z(t) Jo ]exp[yz(t) J

These results indicate that the evolution operator (35) can be formally determined only by one classical orbit which initial conditions. satisfies (41) and the corresponding Substituting (42) (44) into (35), we obtain the formal solution of (31)

Iy(g t) ~ =exp[az(t)J+ ]exp[/3z(t )Jo]

Xexp[yz(t)J
In the coordinate form

]lp($, 0))

(45)

representation,

Eq. (45) takes the


d
dglz

],

(35)

where az(t), /3z(t), and yz(t) are functions of time to be Since the evolution operator must be unidetermined.

y(g, t ) =exp

/3z+

t i yz azg exp
2

qr(g',

0),
(46)
we

tary, i.e. ,

0 (t ) =exp[ J ]exp[ Jo]exp[ J+ ] az /3z yz ]exp[ = U '(t)=exp[ yzJ azJ+ ], PzJo]exp[


(36) az, /3z, and yz are real functions of time. Substituting (35) and (15) into (13) and equating the coefficients of the same operators in the two sides of Eq. (13), we obtain the differential equations for az, /3z, and yz.

where g'=(exp( , '/3z). To complete the differentiation write the initial wave function as

V(k'

o)=

&(4" k')V (4 0)d4",

"

(47)

and choose the 6 function representation

5(g" = P)

f
1

exp[ik(g"

g')]dk

(48)

az= [a+b cos(Qt)]


~

(37) (38) (39)

Substituting (47) and (48) into (46) and completing the integral with respect to the variable k, we obtain

/3z

= 2 az,
exp(Pz),

y( g, t ) =

exp

+i
2

y2=

with the initial conditions


z

X
yzl
~

exp
oo

2/2

g') (g"

y(g", 0)dg" .
(49)

at=

l0 o,

/ zl,

=o=o

o=0=

(40)

It is clear that numerically solving these equations is difficult since ai, /3i, and y, in (37) (39) can reach to infinity and then ai, /3i, and yi could become infinity in a limited time period. On the other hand, the twofold in-

In order to find a more explicit expression for qr(g', t), we assume that the initial wave function is a linear superposition of wave functions of a harmonic oscillator with mass m and frequency coo such as Eq. (11). In this way,

4980
we can complete the integral in (49) and obtain
' 'P2+ i ,cc2$ y( g, t ) = V' 2 ( t ) exp( ,

FU-LI LI

47

'mcooA(t)e , Xexp[
X
n=0

'g ]
,

aNexp[i(n+ ')(!)]
XH(+mcooA(t)e ' g),
(50)

where

p(t ) =arctan[m
A

coo7

2(t )]

(51)
(52)

(t)=

[1+[mcooy~(t)] ] 'i
]. /2

The form of the solution (50) is different from that obtained by Brown [5] and Combescure [3]. The reason for this difference is that in the present study the real solution of the Mathieu equation (41) with the special initial conditions is chosen, but in their studies a general complex solution of (41) was taken to construct the solution of (31). The present choice has the advantage that allows us directly to relate the quantum solution to the orbit of the classical motion. In this way, the behavior of the quantum motion can be investigated by virtue of the properties of the classical motion. Now let us study the stability of quantum motion of a charged particle in a Paul trap. The probability density of a charged particle in the trap is given by

p(g,

&)

=q*(g, &)q(g, &)=


X

m co(t)

exp [

co( t m

g]'
2

1=0 k =0

g g

at*ak(2"+'k!l!)'~ exp[i(k

l)P)HI(&mco(t)g)Hk(&mco(t)g),

where
co(t )

IV. CONCLUDING REMARKS

=,' F'(r)

I+coo

f F'(r)
o

1/2
(54)

For the stable motion in the trap, the probability density p(g, t) must expand only in a limited space at any time. From (53), it is clearly seen that the distribution of p(g, t) in the trap is completely controlled by the timedependent frequency co(t). When the amplitude of the classical motion specified by F(t) monotonously increases in time, co(t) will become more and more small. In this case, p(g, t ) will diffuse in the whole space if time is long enough. Therefore, the quantum motion is unstable if the classical motion is unstable. When the amplitude of F(t) is limited at all times, co(t) will take finite values in the whole motion. In this case, p(g, t) will expand only in a limited space. Therefore the stability regions for the quantum motion are exactly the same as for the classical motion. Combescure, based on the fact that the quantum coordinate operator satisfies the same equation as the classical coordinate, declared the same conclusion as the above. We think that the same form of the equations for both the coordinate operator and its classical counterpart only guarantees that the motion of the mass center of a wave packet in the trap is just as the classical motion of the charged particle, and it does not tell us any information on the diffusion of a wave packet in the trap. Therefore, the present discussion about the quantum motion stability is much more transparent than the discussion given by Combescure [3].
[1] W. Paul, Rev. Mod. Phys. 62, 531 (1990). [2] F. Diedrich, J. C. Bergquist, W. M. Itano, and D. J. Wineland, Phys. Rev. Lett. 62, 403 {1989).
[3] M. Combescure, Ann. Inst. Henri Poincare 44, 293 (1986).

The Hamiltonian of a charged particle in a Paul trap for each of three rectangular coordinates can be written as a linear function of operators of the su(1, 1) algebra. The two realizations of the su(1, 1) algebra have been presented. The first realization given in the second section is convenient for the numerical calculation. As shown in the third section, the second realization is useful for the analytical discussion. We have shown that the quantum and classical motion of a charged particle in a Paul trap have the same stability regions. The numerical results have shown that for a given frequency Q of the time-dependent quadrupole potential of the trap the secular motion is well pronounced when the amplitude Vo of the ac voltage is not large so that the condition eVO/Iron'1 holds. In comparison with the exact numerical calculation with the approximate result, we have found that in the region where the secular motion exists, the quantum secular approximation is valid.
ACKNOWLEDGMENTS

The author is very much grateful to Professor A. Weiguny and Professor S. J. Wang for carefully reading and revising this manuscript and many valuable suggestions. The author also wishes to acknowledge the hospitality at the Institut fiir Theoretische Physik I of Miinster University. This work was supported in part by the University of Munster and by the National Science Foundation of China.
[4] R. J. Cook, D. G. Shankland, and A. L. Wells, Phys. Rev. A 31, 564 (1985). [5] L. S. Brown, Phys. Rev. Lett. 66, 527 (1991). [6] V. N. Gheorghe and F. Vedel, Phys. Rev. A 45, 4828

QUANTUM MOTION OF A CHARGED PARTICLE IN A PAUL TRAP

4981

(1992). [7] H. G. Dehmelt,

in Aduances

Physics, edited by D. R. ic, New York, 1967), Vol. 3, p. 53; R. F. Wuerker, H. Shelton, and R. V. Langmuir, J. Appl. Phys. 30, 342 (1959). [8] J. Wei and E. Norman, J. Math. Phys. 4, 575 (1963). [9] C. F. Lo, Phys. Rev. A 43, 404 (1991); 45, 5262 (1992), and

in Atomic and Molecular Bates and I. Estermann (Academ-

references therein; C. M. Cheng and P. C. W. Fung, J. Phys. A 21, 4115 (1988). [10] G. S. Agarwal and S. A. Knmar, Phys. Rev. Lett. 67, 3665

(1991). [11) G. Dattoli, S. Solimeno, and A. Torre, Phys. Rev. A 34, 2646 (1986).

You might also like