Progress of Theoretical Physics Supplement No.

166, 2007 179

Nonlinear Dynamics of Multi-Component Bose-Einstein

Condensates
Anti-Gravity Transport and Vortex Chaos

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

Katsuhiro Nakamura∗)
Department of Applied Physics, Osaka City University, Osaka 558-8585, Japan

Bose-Einstein condensate(BEC) provides a nice stage when the nonlinear Schrödinger

equation plays a vital role. We study the dynamics of multi-component repulsive BEC in
2 dimensions with harmonic traps by using the nonlinear Schrödinger (or Gross-Pitaevskii)
equation. Firstly we consider a driven two-component BEC with each component trapped
in different vertical positions. The appropriate tuning of the oscillation frequency of the
magnetic field leads to a striking anti-gravity transport of BEC. This phenomenon is a man-
ifestation of macroscopic non-adiabatic tunneling in a system with two internal(electronic)
degrees of freedom. The dynamics splits into a fast complex spatio-temporal oscillation of
each condensate wavefunctions together with a slow levitation of the total center of mass.
Secondly, we examine the three-component repulsive BEC in 2 dimensions in a harmonic
trap in the absence of magnetic field, and construct a model of conservative chaos based on
a picture of vortex molecules. We obtain an effective nonlinear dynamics for three vortex
cores, which represents three charged particles under the uniform magnetic field with the
repulsive inter-particle potential quadratic in the inter-vortex distance rij on short scale and
logarithmic in rij on large scale. The vortices here acquire the inertia in marked contrast
to the standard theory of point vortices since Onsager. We then explore “the chaos in the
three-body problem” in the context of vortices with inertia.

§1. Introduction

Recently a great number of theoretical and experimental efforts have been de-
voted to Bose-Einstein condensates (BECs).1) As well as single-component BECs,
the trapping techniques can create multi-component condensates which involve inter-
component nonlinear interactions. The multi-component BEC, far from being a triv-
ial extension of the single-component one, presents novel and fundamentally different
scenarios for its ground state and excitations. In particular, it has been observed that
BEC can reach an equilibrium state characterized by the separation of the species
in different domains.2) BEC provides a nice stage when the nonlinear Schrödinger
equation plays a vital role. We study the dynamics of multi-component repulsive
BEC in 2 dimensions with harmonic traps by using the nonlinear Schrödinger (or
Gross-Pitaevskii) equation.
In §2 we consider a driven two-component BEC with each component trapped
in different vertical positions, and indicate a striking phenomenon of levitation of
BEC against gravity with neither applied mechanical force nor associated classical
trajectory, which is a manifestation of macroscopic non-adiabatic tunneling in a
system with two internal(electronic) degrees of freedom. The Rabi oscillation is
coupled with the energy exchange between internal(electronic) and external(atomic)
degrees of freedom. In this scenario the dynamics splits into a fast complex spatio-
∗)
E-mail: [email protected]
180 K. Nakamura

temporal oscillation of each condensate wavefunctions together with a slow levitation

of the total center of mass.
In §3 we examine the three-component repulsive BEC in 2 dimensions in a
harmonic trap in the absence of magnetic field, and construct a model of conservative
chaos based on a picture of vortex molecules. Assuming the vortex solution with a
Gaussian tail for each component and applying the collective-coordinate method,

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

we obtain an effective nonlinear dynamics for three vortex cores. This dynamics
represents three charged particles under the uniform magnetic field with the repulsive
inter-particle potential quadratic in the inter-vortex distance rij on short scale and
logarithmic in rij on large scale. We show the novel feature of point vortices in
the multi-component BEC unseen in the single-component version, and explore “the
chaos in the three-body problem”.

§2. Levitation in the presence of oscillating magnetic field

We report how multicomponent BECs can be used to get a matter-wave analogue

of the Franck-Condon effect,3) a phenomenon well known in molecular physics. We
provide an analogy between (i)the interaction of electronic energy and vibrational
excitations of a molecule connected by absorption/emission of photons and (ii)the
interaction of spinor energy and oscillations in a parabolic trap of a spinor BEC con-
nected by absorption/emission of microwave photons by the two-body interactions.
Macroscopically, the phenomenon results in a levitation of the condensate, schemat-
ically illustrated in Fig. 1, and occurs when the spatial aspects of the dynamics
are put together with the transition between the components driven by a resonant
coupling field. To fix ideas, we consider a spinor BEC composed of two hyperfine
states of 87 Rb atoms4), 5) trapped at different vertical positions and coupled by a
time-dependent coupling field.
To simplify the treatment yet preserving the spatial aspects we assume the con-
densate to be magnetically tightly confined along one of the transverse directions to
two effective dimensions. In the mean-field approximation the system is described
by the equations5)
∂ψ1

i = L1 + U11 |ψ1 |2 + U12 |ψ2 |2 ψ1 + f (t)ψ2 , (2.1a)
∂t
∂ψ2

i = L2 + U22 |ψ2 |2 + U21 |ψ1 |2 ψ2 + f (t)ψ1 , (2.1b)
∂t
where f (t) = B cos (Ωt) and Lj = −∇2 /2 + Vj . Equations (2.1) are written in
dimensionless
 form: the coordinates r = (x, z) and time are measured in units of
 =
/mω and 1/ω, respectively, while the energies and frequencies are measured
in units of
ω and ω, ω being the trap frequency in the (x, z)-plane. The confining
potentials are given by
Vj ≡ Vj (x, z) = [x2 + (z − zj )2 ]/2 + gz + j , (2.2)
where j stands for the internal electronic energy of the j-th component and g is
the dimensionless gravitational constant. B and Ω are the dimensionless amplitude
 Nonlinear Dynamics of Multi-Component Bose-Einstein Condensates 181

V2

V1

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

gravity
initial location
of the atoms
z
z1* z2 *

Fig. 1. [Color online] System studied: A spinor BEC under the action of gravity and confining
potentials V1 (for |1
) and V2 (for |2
) with minima z1∗ and z2∗ , respectively. An oscillating
coupling field couples both states.

and frequency of the coupling field. The dimensionless nonlinear √ coefficients, for the
quasi-two-dimensional condensate, are given by Uij = Uji = 2 2πN aij /y where aij
are the scattering lengths for binary collisions, N is the total number of atoms, and
y is the oscillator length in the y-direction. The normalization of the wave function
is then 
ψ † ψd2 r = 1, (2.3)

where ψ =col(ψ1 , ψ2 ).  
Since Vj (x, z) has its minimum Vmin,j = zj g−g 2 /2+j at (x, z) = 0, zj∗ = zj − g ,
we can choose z2∗ = −z1∗ = z∗ (> 0) by shifting the origin of coordinates. Then
δz = 2z∗ is the distance between the trap minima. Due to the phase invariance we
can set 2 = −1 = (> 0).
Our model keeps the time-dependent nature of the coupling field. Although in
certain regimes this dependence can be ignored, we will explore a parameter regime
in which the time-dependence induces a new type of resonant response which has
not being studied previously and is responsible of an interesting phenomenology.
Initially we prepare a state with its center of mass at r 1 = (0, z1∗ ),
 
1 x2 + (z − z1∗ )2
ψ1 = √ exp −  (2.4)
π (1 + U11 /2π)1/4 2 1 + U11 /2π
182 K. Nakamura

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

Fig. 2. [Color online] Dynamics of the populations P1 (t) (dashed line) and P2 (t) (solid line) in the
case of resonance of Franck-Condon type for B = 50 and (a) δz = 1.0 (b) δz = 5.0.

which minimizes the energy E = d2 r(|∇ψ1 |2 /2+V1 |ψ1 |2 +U11 |ψ1 |4 /2) over a family
of Gaussian functions and approximates the ground state when all atoms are in |1.
We then apply an external microwave field, whose frequency is chosen to induce an
intercomponent Franck-Condon vertical transition: Ω = ∆V12 = V2 (0, z1∗ )−V1 (0, z1∗ ),
and study the dynamics governed by Eq. (2.1).
For the numerical integration we have used ADI, Crank-Nicholson, and pseu-
dospectral methods to cross-check the results. The values of the scaled parameters
used in this paper are: U11 = 100, U22 = 97, U12 = 94, g = 0.1, and δ = 2 = 500.
[Although  is of the order of GHz/Hz our results should be valid for any   1 as it
happens for Franck-Condon or standard resonances. This is so because in the RWA,
the unitary transformation making the oscillating field a static one produces in the
diagonal term the difference between Ω and 2. This difference is much less than
unity in the case of Franck-Condon resonance and vanishes in the case of standard
resonance. Since we are not working in the RWA, we use the oscillating field as it
stands and choose a large enough value of  = 250.]
First we have studied the population mixing between both components. In
the standard resonance case (Ω = 2) without coupling with the orbital degrees of
freedom, the Rabi formula implies that each population Pj (t) = |ψj |2 d2 r, should
oscillate with a frequency proportional to the coupling field amplitude.6) Although
our system is nonlinear, the period of the population oscillations in Fig. 2 is about
2π/B ≈ 0.125, close to the estimate of the Rabi equation. This happens because the
linear coupling in dimensionless units has an order of B = 50 while the amplitude
of the nonlinear coupling is less than 8.0.
The population fluctuations are suppressed for times much larger than the pe-
riod of oscillations (these times are smaller for larger δz). We find that the two
components |ψ1,2 |2 show a domain structure, a typically nonlinear phenomenon.
This behavior is remarkably similar to one observed in Refs. 4) and 5), where the
coupling was constant, and to the collapse of populations in the nonlinear tunneling
studied in Ref. 7) and attributed to formation of spatial structures due to instabili-
ties. The fact that the same phenomenology is present here could be an indication
that the development of spatial structures could be a generic property of the popu-
lation transfer in spatially extended systems resulting in the suppression of the Rabi
oscillations.
 Nonlinear Dynamics of Multi-Component Bose-Einstein Condensates 183

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

Fig. 3. Time evolution of the total center of mass for B = 50 and δz = 0.5 (broken line) and
δz = 5.0 (solid line). The dotted line stands for the case far from the standard resonance
(Ω = 1,  = 250) with δz = 5.

The spatio-temporal dynamics of the wavefunction |ψ2 |2 is as follows. In the

early stage region (t = 0 ∼ 0.5) the wavefunction oscillates near the minimum (z1∗ )
of the lower potential well. As time goes on (t = 1 ∼ 2), the central position of the
solutions moves towards the minimum of the upper well (z2∗ ). Finally, for t ∼ 3.1,
the wavefunction has a double-humped structure with its valley located near the
minimum of the upper well.
In Fig. 3 we show typical examples of the time evolution of the total center of
mass (CM), z̄ = ψ † zψd2 r. It exhibits a smooth and upward acceleration, reaching
a maximum height ∼ z2∗ for t ∼ 3.1 (corresponding to 3.1/τRB ∼ 25 Rabi periods) for
B = 50. This physical phenomenon is a levitation of the condensate against gravity.
We will explain later this effect as a result of a Franck-Condon resonance of matter
waves.
We have further explored the dependence on B of the initial upward acceleration
rate of the CM, α = α(0), evaluated by fitting the CM data for small times with
a quadratic function (z̄ ∼ z1∗ + αt2 /2). We can see that, as B is increased, α
monotonically grows and reaches a plateau value, which indicates the relevant role of
the Rabi oscillation for the energy conversion between electronic and orbital degrees
of freedom.
A direct algebra leads to the equation of motion for z̄
d2 z̄
+ z̄ = z∗ (P2 − P1 ), (2.5)
dt2
which is a manifestation of Ehrenfest’s theorem. Equation (2.5) together with the
initial condition z̄(0) = z1∗ , dz̄/dt|t=0 = 0 reveals the importance of the choice of
the resonant frequency and explains the features of Fig. 3. Indeed, (i) if Ω ∼ 1,
but δ  1, namely in the case of far-from resonance, we find little population
mixing: P1 ≈ 1 and P2 ≈ 0 throughout the time evolution. Then z̄(t) shows a small
oscillation with unit frequency around z1∗ (= −z∗ ), and α(0) ≈ −z̄(0) − z∗ ≈ 0. This
184 K. Nakamura

is the case of no levitation depicted by the dotted line in Fig. 3; (ii) If Ω ∼ δ with
B  Ω, namely in the case of Franck-Condon resonance, the right-hand side of
Eq. (2.5) vanishes owing to the rapidly-oscillating population difference. Then z̄(t)
shows a large oscillation (with amplitude z∗ ) of unit frequency around z̄ = 0 and
one obtains the acceleration rate, which in the leading order is independent neither
on Ω nor on B: α(0) = −z̄(0) = z∗ = 12 δz.

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

We have studied the dynamics of two-component Bose-Einstein condensates sub-
ject to an oscillating driving coupling field. In the case of Franck-Condon’s vertical
transition, the driving frequency Ω corresponds to the energy difference between the
trapping potentials V1 and V2 at z1∗ . Then the condensate is effectively driven to an
unstable state of V2 , moving towards the potential minimum z2∗ due to the restoring
force of V2 . In this situation, while the populations describe fast Rabi oscillations,
the total center-of-mass exhibits a smooth and upward acceleration with its rate (α)
showing a monotonic growth with increasing δz and B. This levitation phenomenon
of the center of mass against gravity provides a macroscopic manifestation of the
Franck-Condon effect of molecular physics, which should be tested experimentally.

§3. Vortex chaos: realization of 3-body problem

BEC has a dual aspect of waves and particles. The wave nature is highlighted
in the phenomenon of interference leading to fringe patterns.8), 9) On the other hand,
the particle nature of BECs can be seen in typical localized states like vortices and
solitons. In fact solitons were observed in the quasi-one dimensional BEC.10), 11)
Atomic BEC is a typical quantum fluid, and its flow can be described as potential
flow, resulting superfluidity of this system. The superfluid property of atomic BECs
is theoretically described by the macroscopic wave function. Because of the nonlin-
earity of the system caused by interaction between particles, the macroscopic wave
function can take a form of various solitons such as bright, dark, grey, and vortex
solitons, and these solitons were experimentally observed as soon as atomic BECs
were realized.
Solitons are non-diffusive and localized wave packets and behave like particles
described in classical mechanics. Recently, Kinoshita et al.12) measured the colli-
sional dynamics of two bright solitons in a trapped quasi one-dimensional 87 Rb BEC.
Two bright solitons with collisional interaction are integrable, and their dynamics is
similar to “Newton’s cradles”.
On the other hand, Martin et al.13) theoretically predicted that three bright
solitons in a one-dimensional BEC were non-integrable and showed the change of
their behavior from regular to chaos, being related to the three-body problem in
classical astrophysics. They further did a direct numerical simulation of the Gross-
Pitaevskii equation (GPE) which describes the dynamics of the macroscopic wave
function, and also confirmed such a change in the behavior of bright soliton dynamics.
In multi-component BECs, there are not only intra-component particle inter-
action but also inter-component particle interaction which is another origin of non-
linearity, so we expect novel soliton dynamics unseen in single-component BECs.
Motivated by this, Perez-Garcia et al.14) and Yamasaki et al.15) investigated bright
 Nonlinear Dynamics of Multi-Component Bose-Einstein Condensates 185

soliton dynamics in two or three component BECs in two dimensions, and discussed
their particle-like behavior. However, in two and higher dimensional systems, bright
solitons are unstable unless intra-component interaction oscillates between attraction
and repulsion or intra-component three-body interaction is strong enough, so exper-
imental realization of bright soliton in two-dimension seems to be very difficult. On
the other hand, topological vortex solitons as quantized vortices of the macroscopic

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

wave function can be stable even in two dimensions. Contrary to a bright soliton,
vortex solitons in multi-component BECs have already been experimentally realized,
and can become a good candidate to study solitons in 2-d multi-component BECs.
In this section, we consider the vortex soliton in multi-component BECs in two
dimensions. To consider the effective dynamics of point-like dynamics of vortices,
we extract some degrees of freedom by using the variational approach, and derive an
effective dynamics with finite degrees of freedom. For the case of an one-component
BEC, it is well known that effective hamiltonian with finite degrees of freedom be-
comes16)–18)

H = −8π ni nj log rij , (3.1)

j>i
in the limit of infinitesimal vortex cores in the vortex  point-like picture. Here ni
is the winding number of the i-th vortex, and rij = (xi − xj )2 + (yi − yj )2 is the
distance between cores of i-th and j-th vortices. Equation (3.1) shows that there
is no momentum degree of freedom and xi and yi are conjugate coordinates each
other. On the other hand, in the case of multi-component BECs, we can successfully
show that the system recovers momentum degrees of freedom and is described as
particle-like dynamics rather than vortex point-like dynamics.
Analyzing the multi-component GPE with vortex soliton, we now extract some
degrees of freedom of soliton by using the variational principle, and obtain the effec-
tive Hamiltonian with finite degrees of freedom for vortex solitons. BEC at zero tem-
perature is described by the GPE. A two-dimensional system of trapped n-component
macroscopic wave function Φ1 (t, x, y), Φ2 (t, x, y), · · · , Φn (t, x, y), obeys
 
∂

2 2 2
i Φi (t, x, y) = − ∇ + V (x, y) + gii |Φi (t, x, y)| + gij |Φj (t, x, y)| Φi (t, x, y),
∂t
j=i
(3.2)
2 2
for i = 1, . . . , n. V (x, y) = (x + y ) stands for the effect of trapping. gii and
gij are the coupling constant of intra-component and inter-component interaction,
respectively. We set the situation of repulsive interactions and assume the same inter-
component and the same intra-component interactions, i.e., g11 = g22 = · · · = g1 > 0
and g12 = g13 = · · · = g1n = g23 = g24 = · · · = g2 > 0. In the context of Eq. (2.1),
we here made the following replacement: 2t → t, 2Uij → gij .
In the absence of the inter-component interaction, each component has station-
ary states of a vortex. So, we consider the case in which each component has one
vortex and vortices interact with each other through the inter-component interaction
of BECs. Our goal is to derive from (3.2) the evolution equation for the collective
coordinates of trial vortices. The collective coordinates for a vortex are phase vari-
ables besides the center of mass. We Taylor-expand the phase with respect to space
186 K. Nakamura

coordinates relative to the center of mass. We choose a trial function for the vortex
with unit winding number (ni = ±1) as19)
Φi (t, x, y) ≡ fi (t, x, y) exp[iφi (t, x, y)]
 
1 x2 + y 2 (x − xi )2 + (y − yi )2
=√ exp −
2ξ 2 + (x − xi )2 + (y − yi )2

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

π∆ 2∆
   
y − y i
× exp i ni tan−1 + αi (x − xi ) + βi (y − yi ) , (3.3)
x − xi

which is the solution of (3.2). The meaning of collective coordinates is as follows:

(xi , yi ) is the center of mass and (αi , βi ) are the first-order coefficients of Taylor-
expansion of the phase φi (t, x, y) with respect to (x − xi , y − yi ) rather than (x, y). A
trivial constant phase has been suppressed. ∆ is the width of trapped condensates
and ξ is the healing length related to vortex core size. For the  case of a single-
component
 BEC, ∆ and ξ are approximately obtained as ∆ = 1 + g1 /4π and
ξ = 2π∆/g1 .
First of all we note: Equation (3.2) can be derived from the variational principle
that minimizes the action obtained from Lagrangian density for field variables,
 

i g1

∗ ∗ 2 2 2 2 4
L= (Φi Φ̇i − Φi Φ̇i ) + |∇Φi | + (x + y )|Φi | + |Φi | + g2 |Φi |2 |Φj |2 .
2 2
i j>i
(3.4)
In fact, the multi-component GPE is obtained from Lagrange equation:
∂ ∂L ∂L ∂L
− ∗ +∇ = 0. (3.5)
∗
∂t ∂ Φ̇i ∂Φi ∂∇Φ∗i
By regarding (xi , yi ) and (αi , βi ) as variational parameters, the Lagrangian density
(3.4) becomes
    2  2

∂φi ∂φi ∂φi ∂φi ∂fi ∂fi
2
L= fi ẋi + ẏi + α̇i + β̇i + +
∂xi ∂yi ∂αi ∂βi ∂x ∂y
i
 2  2  
∂φi ∂φi g1 4

2 2 2 2
+ + fi + fi + (x + y )fi + g2 fi2 fj2 . (3.6)
∂x ∂y 2
j>i

We now insert the trial Gaussian function (3.3) into (3.6) and obtain the effective
Lagrangian L for their collective coordinates by integrating L over space coordinates:

L = dxdyL. (3.7)

In the limit of X = ξ 2 /∆  1, the Lagrangian for collective coordinates is expressed

by

  c2 ri2 
L(xi , yi , αi , βi ) = (αi2 + βi2 − αi ẋi − βi ẏi ) c1 −
∆
i
 Nonlinear Dynamics of Multi-Component Bose-Einstein Condensates 187
ni c3
+ (xi ẏi − yi ẋi + 2αi yi − 2βi xi ) − c4 (xi α̇i + yi β̇i )
∆
1   2g2  
+ c 5 + c7 1 + + c9 ∆ 2
∆ g1
r 2   g2   c g

11 2
+ i2 c6 − c8 1 + + c10 ∆2 + 2 V (xij , yij ) . (3.8)
∆ g1 ∆ g1

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

j>i


Here, ri = x2i + yi2 , xij = xi − xj , and yij = yi − yj . Coefficients c1 ∼ c11 become

c1 = 1 − 2Xδ1 ,
c2 = 2X − (2X + 4X 2 )δ1 ,
c3 = 1 − 2Xδ1 ,
c4 = 1 − 2X + 4X 2 δ1 ,
3
c5 = + X + (1 − 4X − 2X 2 )δ1 ,
2
1
c6 = − 6X − 2X 2 − (1 − 4X − 14X 2 − 4X 3 )δ1 ,
2
1
c7 = + 2 − (4 + 8X)δ2 ,
2X
c8 = 12 + 16X − (8 + 64X + 64X 2 )δ2 ,
c9 = 1 − 2X + 4X 2 δ1 ,
c10 = 4X + 4X 2 − (2X + 12X 2 + 8X 3 )δ1 ,
1 8X 16X 2  64X 3 
c11 = − − + 16X 2 + δ2 , (3.9)
3 3 3 3
where δn = exp(2nX)Γ (0, 2nX) with the imperfect Gamma function Γ (ν, x). As-
ymptotic behavior for the interactions V (xij , yij ) between solitons becomes
 2
= −rij + V0 (rij  1)
V (xij , yij ) (3.10)
= −24∆2 X log rij + V0 , (rij  1)

with rij = x2ij + yij
2 . Equation (3.10) conveys that the repulsive inter-particle

potential is quadratic in the inter-vortex distance rij on short scale and logarithmic
in rij on large scale.
Lagrange equations of motion for the phase variables αi and βi ,
   
d ∂L ∂L d ∂L ∂L
− = 0, − =0
dt ∂ α̇i ∂αi dt ∂ β̇i ∂βi

lead to

αi
B1 ẋi − ni B2 yi ,
βi
B1 ẏi − ni B2 xi . (3.11)
188 K. Nakamura

Equation (3.11) shows that (αi , βi ) correspond to generalized momentum conjugate

to (xi , yi ) under the vector potential. Therefore we can rewrite other two equations
by eliminating αi and βi . Equation of motion for (xi , yi ),
   
d ∂L ∂L d ∂L ∂L
− = 0, − =0

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

dt ∂ ẋi ∂xi dt ∂ ẏi ∂yi

gives

∂V (xij , yij )
ẍi
−ni B3 ẏi − B4 xi − B5 ,
∂xi
j=i

∂V (xij , yij )
ÿi
ni B3 ẋi − B4 yi − B5 . (3.12)
∂yi
j=i

Coefficients B1 ∼ B5 are simple functions of c1 − c11 .

Equation (3.12) shows that the dynamics of vortices is very similar to charged
particles under the magnetic field in the harmonic potential. The corresponding
Hamiltonian to Eqs. (3.11) and (3.12) can be given as

1

H=  i ]2 + Wi +
[(αi , βi ) − ni A U (xij , yij ) , (3.13)
2
i j>i

with the vector potential A  i = −D1 (−yi , xi ), the harmonic potential Wi = D2 (x2 +
i
2
yi )/2, and interaction U (xij , yij ) = D3 V (xij , yij )/2, where D1 = B3 /(1 + B1 ), D2 =
B4 − B2 D1 − D12 , and D3 = 2B5 . Compared to Eq. (3.1), it is clear that the
system has momentum degrees of freedom, i.e., the inertia, and vortices have particle-
like behavior rather than vortex point-like behavior in conventional many vortices
system. The inertia of the present vortices has appeared due to the multi-component
nature of BEC. Parameters D1 ∼ D4 depend on ξ, ∆, and g, and can be controlled
by changing the number of particles, tightness of trapping and strength of interaction
which is tunable by Feshbach resonance.
We shall focus on the system of three vortices with equal winding numbers, and
find that chaos appears even in three vortices system. This feature is completely
different from that of point vortices system in a single component BEC in which
chaos appears in the case of more than three vortices.
For three vortices with the same winding numbers (n1 = n2 = n3 = 1), Hamil-
tonian (3.13) becomes

1
H = (α12 + α22 + α33 + β12 + β22 + β33 )
2
+D1 (x1 β1 + x2 β2 + x3 β3 − y1 α1 − y2 α2 − y3 α3 )
1
+ (D12 + D2 )(x21 + x22 + x33 + y12 + y22 + y33 )
2
+U (x12 , y12 ) + U (x23 , y23 ) + U (x31 , y31 ). (3.14)
 Nonlinear Dynamics of Multi-Component Bose-Einstein Condensates 189

With use of Jacobi coordinates

1
(xT , yT ) = (x1 + x2 + x3 , y1 + y2 + y3 ),
3
1
(xC , yC ) = (x1 − x3 , y1 − y3 ),
2
(xR , yR ) = (x1 + x3 − 2x2 , y1 + y3 − 2y2 ),

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

1
(αR , βR ) = √ (α1 + α3 − 2α2 , β1 + β3 − 2β2 ),
2 3
1
(αT , βT ) = √ (α1 + α2 + α3 , β1 + β2 + β3 ),
6
1
(αC , βC ) = (α1 − α3 , β1 − β3 ), (3.15)
2
Eq. (3.14) can be rewritten as
√ 3(D12 + D2 ) 2
H = αT2 + βT2 + 6D1 (xT βT − yT αT ) + (xT + yT2 )
2
2
+αC + βC2 + 2D1 (xC βC − yC αC ) + (D12 + D2 )(x2C + yC 2
)
D1 2
D + D2 2
2 2 2
+αR + βR + √ (xR βR − yR αR ) + 1 (xR + yR )
3 12
 2x + x 2y + y   2x − x 2y − y 
C R C R C R C R
+U , +U , + U (−2xC , −2yC ).
2 2 2 2
(3.16)
(xT , yT ), (xC , yC ), and (xR , yR ) represent the center of mass of all three compo-
nents, the relative displacement, and the bisector of the vertex (x2 , y2 ) , respectively.
Aside from the separable center-of mass degrees of freedom, the system of three vor-
tices with the same winding numbers has 4 degrees of freedom, but the independent
constants of motion is 2 (energy and angular momentum). The system is therefore
nonintegrable and can be chaotic. For three vortices with the different winding num-
bers (n1 = n2 = −n3 = 1), on the other hand, even the angular momentum is not
conserved, and one can expect chaos more easily.
We now carry out a simulation for time evolution of three vortices with the
identical winding number. Under the fixed center-of-mass coordinates, we construct
from (3.16) the equations of motion for xC , yC , xR , yR and their canonical-conjugate
variables, which is solved numerically. Poincaré cross section and power spectra in
Fig. 4 show the transition from high-dimensional tori to chaos, as the system’s energy
is increased. In fact the lowest panels show a collapse of torus and the broad power
spectrum, which are manifestations of chaos.

§4. Summary and discussions

We have considered a driven two-component BEC with each component trapped

in different vertical positions. The appropriate tuning of the oscillation frequency of
the magnetic field leads to anti-gravity transport. This phenomenon is a manifesta-
tion of macroscopic non-adiabatic tunneling in a system with two internal(electronic)
190 K. Nakamura

(a) D1 = D2 = 5, D3 = 1 (d)
2 1000
1.5
800
1
0.5 600

Ek(ω)
2 αC
0

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

-0.5 400

-1
200
-1.5
-2 0
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
2 xC ω

(b) D1 = D3 = 5, D2 = 1 (e)
2 2000
1.5
1 1500
0.5
Ek(ω)
2 αC

0 1000
-0.5
-1 500
-1.5
-2 0
-0.5 0 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
2 xC ω

(c) D1 = 1, D2 = D3 = 5 (f)
4 25
3
20
2
Ek(ω)[X 1000]

1 15
2 αC

0
-1 10

-2
5
-3
-4 0
-3 -2 -1 0 1 2 3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
2 xC ω

Fig. 4. [Color online] Poincaré cross section for βR = 0 (a)-(c), and power spectra (d)-(f). (a), (d):
D1 = D2 = 5, D3 = 1. (b), (e): D1 = D3 = 5, D2 = 1. (c), (f): D1 = 1, D2 = D3 = 5.

degrees of freedom. The dynamics splits into a fast complex spatio-temporal oscil-
lation of each condensate wavefunctions together with a slow levitation of the total
center of mass against gravity.
In the absence of driving magnetic field, we have explored vortex dynamics of
the multi-component BEC in the harmonic trap in the case that each component
has a single vortex. With use of collective coordinates for the vortex core and phase
gradients, we have obtained an effective nonlinear dynamics for vortex cores, which
represents three charged particles with inertia under the uniform magnetic field in the
harmonic potential. The repulsive inter-particle potential is quadratic in the inter-
vortex distance rij on short scale and logarithmic in rij on large scale. The inertia of
 Nonlinear Dynamics of Multi-Component Bose-Einstein Condensates 191

vortices is the most novel aspect that has not been found in the conventional theory
of point vortices since Onsager. As the energy is increased, the system of three
vortex cores shows the transition from regular high-dimensional torus to chaos as
illustrated by power spectra and Poincaré surface of section, which is a manifestation
of “chaos in the three-body problem”.
This short article is based on my talk at JSPS International Conference on

Downloaded from https://academic.oup.com/ptps/article-abstract/doi/10.1143/PTPS.166.179/1941300 by guest on 03 March 2019

Quantum Mechanics and Chaos at Osaka City University, September 19-21, 2006.
The author owes much to V. Perez-Garcia, V. Konotop, A. Kohi, D. Matrasulov,
H. Yamasaki and M. Fukushima for their immense contributions and to M. Kobayashi
for innumerable comments and discussions on numerical works. The full details will
be published elsewhere in separate papers.

References
1) L. P. Pitaevskii and S. Stringari, Bose-Einstein Condensation (Oxford University Press,
Oxford, 2003).
2) D. Hall et al., Phys. Rev. Lett. 81 (1998), 1539.
D. Hall et al., Phys. Rev. Lett. 81 (1998), 1543.
M. R. Matthews et al., Phys. Rev. Lett. 81 (1998), 243.
J. Stenger et al., Nature (London) 396 (1998), 345.
H.-J. Miesner et al., Phys. Rev. Lett. 82 (1999), 2228.
G. Modugno et al., Science 294 (2001), 1320.
3) J. Franck, Trans. Faraday Soc. 21 (1925), 536.
E. U. Condon, Phys. Rev. 28 (1926), 1182; ibid. 32 (1928), 858.
4) M. R. Matthews et al., Phys. Rev. Lett. 83 (1999), 3358.
5) J. Williams et al., Phys. Rev. A 61 (2000), 033612.
B. Deconinck et al., Phys. Rev. A 70 (2004), 063605.
6) J. J. Sakurai, Modern Quantum Mechanics (Addison Wesley, New York, 1994).
7) V. V. Konotop, P. G. Kevrekidis and M. Salerno, Phys. Rev. A 72 (2005), 023611.
V. A. Brazhnyi, V. V. Konotop and V. Kuzmiak, Phys. Rev. Lett. 96 (2006), 150402.
8) M. R. Andrews, C. G. Townsend, H. J. Miesner, D. S. Durfee, D. M. Kurn and W. Ketterle,
Science 275 (1997), 637.
9) M. R. Andrews, D. M. Kurn, H.-J.Miesner, D. S. Durfee, C. G. Townsend, S. Inouye and
W. Ketterle, Phys. Rev. Lett. 79 (1997), 553; ibid. 80 (1998), 2967.
10) K. E. Strecker, G. B. Partridge, A. G.Truscott and R. G. Hulet, Nature(London) 417
(2002), 150.
11) L. Khaykovich, F. Schreck, G. Ferrari, T. Bourdel, J. Cubizolles, L. D. Carr, Y. Castin
and C. Salomon, Science 296 (2002), 1290.
12) T. Kinoshita, T. Wenger and D. S. Weiss, Science 305 (2004), 1125.
13) A. D. Martin, C. S. Adams and S. A. Gardiner, Phys. Rev. Lett. 98 (2007), 020402.
14) G. D. Montesinos, V. M. Pérez-Garcı́a and H. Michinel, Phys. Rev. Lett. 92 (2004), 133901-
1.
15) H. Yamasaki, Y. Natsume and K. Nakamura, J. Phys. Soc. Jpn. 74 (2005), 1887.
16) L. Onsager, Nuovo Cim. Supple. 6 (1949), 279.
17) J. C. Neu, Physica D 43 (1990), 385.
18) A. Aftalion, Vortices in Bose-Einstein Condensates (Birkhäuser, Boston, 2006).
19) I. Aranson and V. Steinberg, Phys. Rev. B 53 (1996), 75.