This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Regular and chaotic behavior of collective atomic motion in two-component Bose-Einstein condensates

Wei-Can Syu [email protected] Department of Physics, National Dong-Hwa University, Hualien, Taiwan, R.O.C.    Da-Shin Lee [email protected] Department of Physics, National Dong-Hwa University, Hualien, Taiwan, R.O.C.    Chi-Yong Lin [email protected] Department of Physics, National Dong-Hwa University, Hualien, Taiwan, R.O.C.
Abstract

We theoretically study binary Bose-Einstein condensates trapped in a single-well harmonic potential to probe the dynamics of collective atomic motion. The idea is to choose tunable scattering lengths through Feshbach resonances such that the ground-state wave function for two types of the condensates are spatially immiscible where one of the condensates, located at the center of the potential trap, can be effectively treated as a potential barrier between bilateral condensates of the second type of atoms. In the case of small wave function overlap between bilateral condensates, one can parametrize their spatial part of the wave functions in the two-mode approximation together with the time-dependent population imbalance zz and the phase difference ϕ\phi between two wave functions. The condensate in the middle can be approximated by a Gaussian wave function with the displacement of the condensate center ξ\xi. As driven by the time-dependent displacement of the central condensate, we find the Josephson oscillations of the collective atomic motion between bilateral condensates as well as their anharmonic generalization of macroscopic self-trapping effects. In addition, with the increase in the wave function overlap of bilateral condensates by properly choosing tunable atomic scattering lengths, the chaotic oscillations are found if the system departs from the state of a fixed point. The Melnikov approach with a homoclinic solution of the derived z,ϕz,\,\phi, and ξ\xi equations can successfully justify the existence of chaos. All results are consistent with the numerical solutions of the full time-dependent Gross-Pitaevskii equations.

pacs:
03.75.Lm, 03.75.Fi, 05.30.Jp, 05.45.Ac

I Introduction

The first experimental observation of interference fringes between two freely expanding Bose-Einstein condensates (BECs) has launched the fascinating possibility to probe new quantum phenomena on macroscopic scales related to the superfluid nature of the Bose condensates Andrews1997 . A chief effect is the Josepshon analog, in which the phase differences between two trapped BECs in a double-well potential can generate Josephson-like current-phase oscillations via collectively atomic tunneling through the barrier Smerzi1997 ; Raghavan1999 . Nevertheless, the nonlinearity of tunneling effects produces novel anharmonic Josephson oscillations and macroscopically quantum self-trapping (MQST) effects. The potentially existing such interesting phenomena indeed triggers a wide variety of theoretic investigations in the settings of two-component BECs systems in the double-well potential JuliaDiaz2009 ; Sun2009 ; Satija2009 ; Messeguer2011 , triple-well Richaud2018 and the single BEC system in the triple-well Liu2007 , four-well Liberato2006 and multiple-well potentials Cataliotti2001 ; Nigro2018 , to cite a few. The effective Josephson dynamics can also be observed in some systems without an external potential aba . More importantly, several experiments have successfully observed coherence oscillations through quantum tunneling of atoms between the condensates trapped in a double-well potential Albiez2005 ; LeBlanc2011 ; Pigneur2018 , whereas similar phenomena were also seen from attractive to repulsive atomic interactions by experimentally changing atomic scattering lengths through Feshbach resonances Spagnolli2017 . In addition to the regular coherent oscillations mentioned above, the studies on the possibility of the appearance of chaotic motions, by applying an externally time-dependent trap potential, are also a fascinating task with findings that deserve further experimental justification Abdullaev2000 ; Lee2001 ; Jiang2014 ; Tomkovic2017 . Chaotic behavior is also studied in the dynamics of of three coupled BECs fra .

In the present work, we consider the two-component BECs system with atoms in two different hyperfine states in a single-well harmonic trap potential. Such a two-component BEC system provides us an ideal platform for the study of intriguing phenomena, for example, mimicking quantum gravitational effects in Ref. Syu2019 . Using Feshbach resonances to experimentally tune the scattering lengths of atoms allows us to construct the diagram in Fig. 1, showing the typical ground-state wave function of the binary condensates Thalhammer2008 ; Tojo2010 ; Catani2008 ; Wacker2015 ; Moses2015 . The effective parameter characterizing the miscibility or immiscible regime of binary condensates is mainly determined by the value of Δ=g11g22g122\Delta=g_{11}g_{22}-g_{12}^{2}, where gijg_{ij} denotes the atomic interaction between ii and jj atoms Riboli2002 ; Papp2008 ; Sasaki2009 . The parameter regime for Δ>0\Delta>0 corresponds to the miscible distribution of two-component condensates, whereas for Δ<0\Delta<0 the different species of atoms repel each other and the distribution becomes immiscible Riboli2002 ; Papp2008 ; Sasaki2009 . In particular, we focus on the situations of the scattering lengths in the immiscible regime, where the condensate of one of the components is located at the center of the potential trap, and can effectively be regarded as an effective potential barrier between bilateral condensates of the second component of the system. Experimentally, spatially asymmetric bilateral condensates can be prepared by means of a magnetic field gradient McCarron2011 ; Eto2015 ; Eto2016 , allowing the possibility to observe collectively coherence oscillations of atoms between them. This is an extension of the works of Smerzi1997 ; Raghavan1999 , where the system of a single BEC in an external double-well potential is studied. The collective oscillations of atoms through quantum tunneling between the condensates centered at each of the potential wells have been observed where the full dynamics of the system can be further simplified, and turned into an analog pendulum’s dynamics in the so-called two modes approximation Burchianti2017 ; Smerzi1997 ; Raghavan1999 ; Ananikian2006 . In our setting, the dynamics of the system can also be approximated by analogous coupled pendulums dynamics using the variational approach as well as two modes approximation. One can then reproduce regular oscillations around the stable states studied in Refs. Smerzi1997 ; Raghavan1999 within some parameter regime by considering the effects from time-averaged trajectories of the central condensate. Additionally, the collective motion of atoms, if departing from the unstable states, could show chaotical behavior, oscillating between bilateral condensates due to the time-dependent driving force given by the dynamics of the condensate centered in the middle Abdullaev2000 ; Lee2001 ; Jiang2014 ; Tomkovic2017 . The scattering lengths in the parameter regime, which give the chaotic dynamics of the system, will be useful as a reference for experimentalists to observe the effects.

To be concrete, consider a mixture of Rb85{}^{85}\text{Rb} |1=|F=2,mF=2|1\rangle=|F=2,m_{F}=-2\rangle and Rb87{}^{87}\text{Rb} |2=|F=1,mF=1|2\rangle=|F=1,m_{F}=-1\rangle by ignoring the small mass difference Papp2008 . The scattering lengths we choose are estimated to be a22=50a0a_{22}=50a_{0} and a12=85a0a_{12}=85a_{0} with a0a_{0} the Bohr radius. The existence of the Feshbach resonance at 160G160\text{G} permits us to tune the scattering length a11a_{11} of Rb85{}^{85}\text{Rb} atoms from 50a050a_{0} to 150a0150a_{0} to be specified in each case later Papp2008 . Let us consider two-component elongated BECs in quasi-one-dimensional (1D) geometries with the trap potential Becker2008 ωx=2π×12Hz,ωy,z=2π×100Hz\omega_{x}=2\pi\times 12\text{Hz},\,\omega_{y,z}=2\pi\times 100\text{Hz}. We take the numbers of atoms N1=500N_{1}=500, and N2=1000N_{2}=1000 as an example. The experimental realization of the Josephson dynamics by two weakly linked BECS in a double-well potential in a single BEC system is confirmed in Ref. Albiez2005 . Both Josephson oscillations and nonlinear trapping are observed by tuning the external potential. Such phenomena are also explored in two-component BECs system in a double-well potential where apart from their own atomic scattering processes the Rabi coupling between the atoms from each components of the BECs is introduced. Tuning the scattering lengths using the associated Feshbach resonances and/or the Rabi coupling constant allow us to observe the transition from Josephson oscillations to nonlinear self-trappings zib . Our proposal also in the two-component BECs system but in a single-well potential suggests a novel setup, although involving more rich dynamical aspects of condensates, the only tunable parameters are atomic scattering lengths again through Feshbach resonances to achieve the miscible condensates that needs more experimental endeavor as compared with zib .

Refer to caption
Figure 1: The ground-state wave function of binary BECs system in a cigar-shape trap potential is shown for two species Rb85{}^{85}\text{Rb} |2,2|2,-2\rangle state and Rb87|1,1{}^{87}\text{Rb}\,|1,-1\rangle state with atomic coupling constants in miscible/immiscible regimes respectively. The numbers of atoms, say N1=500,N2=1000N_{1}=500,\,N_{2}=1000, and also the potential trap ωx=2π×12Hz,ωy,z=2π×100Hz\omega_{x}=2\pi\times 12\text{Hz},\,\omega_{y,z}=2\pi\times 100\text{Hz} are chosen to show the spatial distribution of wave functions in the immiscible (miscible) regime for Δ<0\Delta<0 (Δ>0\Delta>0) below (above) red dashed line determined by Δ=0\Delta=0, where Δ=g11g22g122\Delta=g_{11}g_{22}-g_{12}^{2}. The black solid line separates between symmetric and asymmetric condensate wave functions.

Our presentation is organized as follows. In Sec. II, we first introduce the model of the binary BECs system, and the corresponding time-dependent Gross-Pitaeviskii (GP) equations for each of condensate wave functions. In this two-component BECs system, we study the evolution of the wave functions within a strong cigar-shape trap potential so that it can effectively treated as a quasi one-dimensional system. We further assume that the condensates start being displaced from their ground state in the immiscible regime where one of the condensates is distributed in the trap center while the other condensate surrounds the central condensate on its two sides. Then we propose the ansatz of the Gaussian wave functions, which can reduce the dynamics of the system into few degrees of freedom, such as the center of the central condensate as well as the population imbalance and the relative phase difference between bilateral condensates in the two-mode approximation. We later obtain their equations of motion with the form of coupled oscillators equations. In Sec. III, we analyze the stationary-state solutions and their stability property. In Sec. IV, we examine the time evolution of the system around the stable stationary state, showing the regular oscillatory motion. In Sec. V, the study is focused on when the system starts from near unstable stationary-state solutions, exhibiting the chaotic behavior by numerical studies. The Melnikov homoclinic method is adopted for analytically studying the existence of chaos. Concluding remarks are in Sec. VI. In Appendix, we provide more detailed derivations or approximations to arrive at coupled oscillators equations from the time-dependent GP equations using the appropriate ansatz of Gaussian wave functions.

II Variational approach and analogous coupled pendulums dynamics

We consider the binary BECs for same atoms in two different hyperfine states confined in a strong cigar-shape potential with the size of the trap LxL_{x} along the axial direction,taken in the xx direction, which is much larger than LrL_{r} along the radial direction. This system can be effectively treated as the pseudo one-dimensional system with the Lagrangian described as Garcia1996 ; Salasnich2002

L1D=\displaystyle L_{\text{1D}}= j=1,2dx[i2(ψjψjtψjψjt)\displaystyle\sum_{j=1,2}\int dx\Bigg{[}\frac{i\hbar}{2}\left(\psi_{j}\frac{\partial\psi_{j}^{*}}{\partial t}-\psi_{j}^{*}\frac{\partial\psi_{j}}{\partial t}\right)
(22m|ψjx|2+Vext|ψj|2+gjj2|ψj|4)]\displaystyle\qquad\qquad-\left(\frac{\hbar^{2}}{2m}\left|\frac{\partial\psi_{j}}{\partial x}\right|^{2}+V_{\rm ext}|\psi_{j}|^{2}+\frac{g_{jj}}{2}|\psi_{j}|^{4}\right)\Bigg{]}
+𝑑xg12|ψ1|2|ψ2|2,\displaystyle\quad+\int dx\,g_{12}|\psi_{1}|^{2}|\psi_{2}|^{2}, (1)

where the external potential in the axial direction is given by Vext(x)=mωx2x2/2V_{\rm ext}(x)=m\,\omega_{x}^{2}\,x^{2}/2. The mass of atoms is mm and the coupling constants between atoms are given in terms of scattering lengths aija_{ij} as gij=22aij/mLr2g_{ij}=2\hbar^{2}a_{ij}/mL_{r}^{2}. The wave function ψj\psi_{j} is subject to the normalization

𝑑x|ψj(𝐱)|2=Nj\displaystyle\int d\,{x}\,|\psi_{j}(\mathbf{x})|^{2}=N_{j} (2)

with the numbers of atoms NjN_{j} in each hyperfine states jj. The time-dependent GP equations for each component of the BEC condensates are given by Pethick2008

iψ1t=(22m2x2+Vext+g11|ψ1|2+g12|ψ2|2)ψ1,\displaystyle i\hbar\frac{\partial\psi_{1}}{\partial t}=\left(-\frac{\hbar^{2}}{2m}\frac{\partial^{2}}{\partial x^{2}}+V_{\text{ext}}+g_{11}|\psi_{1}|^{2}+g_{12}|\psi_{2}|^{2}\right)\psi_{1}, (3)
iψ2t=(22m2x2+Vext+g22|ψ2|2+g12|ψ1|2)ψ2.\displaystyle i\hbar\frac{\partial\psi_{2}}{\partial t}=\left(-\frac{\hbar^{2}}{2m}\frac{\partial^{2}}{\partial x^{2}}+V_{\text{ext}}+g_{22}|\psi_{2}|^{2}+g_{12}|\psi_{1}|^{2}\right)\psi_{2}. (4)

The full dynamics of the condensates in a harmonic trap potential can be explored by solving the GP equations numerically with given initial wave functions. However, in order to extract their feature analytically, we will propose an ansatz of the wave functions with several relevant parameters. All results from solving the equations of motion for the introduced parameters given by the variational approach will be justified by comparing with those given by numerically solving the GP equations. Moreover, we will rescale the spatial or temporal variables into the dimensionless ones by setting tt/ωxt\rightarrow t/\hbar\omega_{x} and xx//mωxx\rightarrow x/\sqrt{\hbar/m\omega_{x}}. Henceforth, the dimensionless tt and xx will be adopted, and their full dimensional expressions will be recovered, if necessary.

Experimentally, the values of parameters gijg_{ij}, which are related to scattering lengths aija_{ij}, can be tuned by Feshbach resonances  Thalhammer2008 ; Tojo2010 ; Papp2008 . In fact, we might explore the evolution of condensate wave functions, in which ψ1\psi_{1} and ψ2\psi_{2} are initially prepared in the immiscible regime, determined by the parameter Δ=g11g22g122\Delta=g_{11}g_{22}-g_{12}^{2} when Δ<0\Delta<0 Riboli2002 ; Papp2008 ; Sasaki2009 as illustrated in Fig. 1. Let us now assume the general Gaussian wavefunction ψ2\psi_{2} centered at the potential trap center as Garcia1996 ; Salasnich2002 ; Lin2000

ψ2(x,t)=\displaystyle\psi_{2}(x,t)=
(N22πw(t)2)1/4exp[(xξ(t))22w(t)2+ixα(t)+ix2β(t)],\displaystyle\,\,\left(\frac{{N_{2}}^{2}}{\pi w(t)^{2}}\right)^{1/4}\exp[-\frac{(x-\xi(t))^{2}}{2w(t)^{2}}+ix\alpha(t)+ix^{2}\beta(t)], (5)

where four time-dependent variational parameters are the position of the center ξ(t)\xi(t), the width w(t)w(t), and α(t)\alpha(t) (slope) and β(t)\beta(t) ((curvature)1/2({\rm(curvature)}^{1/2}) of the wave function. We also assume that the ground-state configuration of ψ1(x,t)\psi_{1}(x,t) is of the two-mode form Smerzi1997 ; Raghavan1999 ; Burchianti2017 ; Ananikian2006

ψ1(x,t)=φL(t)ψL(x)+φR(t)ψR(x).\displaystyle\psi_{1}(x,t)=\varphi_{L}(t)\psi_{L}(x)+\varphi_{R}(t)\psi_{R}(x)\,. (6)
Refer to caption
Figure 2: The typical condensate wave functions are plotted for the stationary-state solutions of the GP equations for scattering length a11=110a0a_{11}=110a_{0}, a22=50a0a_{22}=50a_{0} and a12=85a0a_{12}=85a_{0} that give g11/g22=2.2g_{11}/g_{22}=2.2 and g12/g22=1.7g_{12}/g_{22}=1.7 (Δ<0\Delta<0 in the immiscible regime shown in Fig. 1), which can be fitted into a Gaussian function (dashed line).

The involved Gaussian-like spatial functions ψL(x)\psi_{L}(x) and ψR(x)\psi_{R}(x) together with the initial wave function ψ1(x,t=0)\psi_{1}(x,t=0) with width σ0\sigma_{0} will be determined numerically by the stationary-state solutions of the GP equations (3) and (4) with the forms shown in Fig. 2, which is symmetric with respect to x=0x=0. ψL(x)\psi_{L}(x) and ψR(x)\psi_{R}(x) presumably serve as a good basis to express the spatial part of the wave functions of bilateral condensates with normalization conditions 𝑑x|ψL,R|2=1\int dx\,|\psi_{L,R}|^{2}=1, and also negligible overlap 𝑑xψLψR0\int dx\,\psi_{L}\psi_{R}\approx 0 for the weak link between bilateral condensates in the immiscible regime. Note that the analytical approach based on the wave functions above gives transparent and consistent results as compared with those from numerically solving the full GP equations in the case of Josephson oscillations but for the motion of MQST discrepancy arise as will be seen later where one should look for the wave functions of bilateral condensates being not restricted to be symmetric with respect to x=0x=0. Thus, the time-dependent part of the wave function can be parametrized as

φL(t)=NL(t)eiϕL(t),\displaystyle\varphi_{L}(t)=\sqrt{N_{L}(t)}\,e^{i\phi_{L}(t)}, (7)
φR(t)=NR(t)eiϕR(t)\displaystyle\varphi_{R}(t)=\sqrt{N_{R}(t)}\,e^{i\phi_{R}(t)} (8)

with the phase ϕL,R(t)\phi_{L,R}(t) and the amplitude NL,R(t)\sqrt{N_{L,R}(t)} given by the number of atoms in each sides of the condensates. The total number of atoms (j=1j=1) is N1=NL+NRN_{1}=N_{L}+N_{R}. To realize quantum coherence between left and right condensates, and also atomic number oscillations, it finds more convenient to introduce the population imbalance

z(t)NL(t)NR(t)NL(t)+NR(t),\displaystyle z(t)\equiv\frac{N_{L}(t)-N_{R}(t)}{N_{L}(t)+N_{R}(t)}, (9)

and relative phase

ϕ(t)ϕR(t)ϕL(t).\displaystyle\phi(t)\equiv\phi_{R}(t)-\phi_{L}(t). (10)

When all atoms locate in the left(right) side of the central condensate, z=1(1)z=1(-1).

The time evolution of coherent oscillations between bilateral condensates surrounding around the central condensate is mainly determined not only by the self-interaction of atoms Smerzi1997 ; Raghavan1999 , but also the interspecies coupling constant g12g_{12} with the wave functions ψ1(x,t)\psi_{1}(x,t) and ψ2(x,t)\psi_{2}(x,t), which give an additional effect to influence the dynamics of coherent oscillations due to the wave-function overlap between them. Also, notice that although the initial wave function of the central condensate is centered at x=0x=0, which is the minimum of the external trapped potential, the later evolution of the wave function will find its stationary state with the center located at x=ξ0x=\xi_{0}, moving around that position with nonzero kinetic energy. Thus, in the presence of ψ2(x,t)\psi_{2}(x,t) shown in Fig. 2 together with the external trapped potential, the resulting effective potential experienced by the wave function ψ1(x,t)\psi_{1}(x,t) in the GP equation in (3) can be seen in the shape of the double-well potential. However, the dynamics of the central condensate leads to the effective potential time dependent, contributing its effects to quantum coherence oscillations between bilateral condensates. This will be the main purpose of studies in Secs. IV and V. This is an extension of the work in Refs. Smerzi1997 ; Raghavan1999 with additional degrees of freedom of the central condensate. When the initial conditions of the central or bilateral condensates are chosen near their stable-state configurations, the dynamics of coherence oscillations modifies slightly the behavior found in Refs. Smerzi1997 ; Raghavan1999 . Nevertheless, as their initial conditions are chosen near the unstable-state configurations, the chaos occurs. Although the analogous coupled pendulums dynamics, which is reduced from the full dynamics of the system and later gives essential information on the time evolution of the condensate wave functions, may produce relatively different trajectories from the solutions of the time-dependent GP equations, they are still very useful to provide us analytic analysis on the chaos dynamics by adopting the Melnikov Homoclinic method.

Substituting the ansatz of the wave function (5) and (6) into the Lagrangian (1), and carrying out the integration over space, the effective action then becomes the functional of time-dependent variables α\alpha, β\beta, ξ\xi, ww, zz, and ϕ\phi. Their time evolutions follow the Lagrange equations of motion derived from this effective Lagrangian, discussed in detail in the Appendix. When the wave function ψ2(x,t=0)\psi_{2}(x,t=0) is slight away from the stationary state, the dynamics of the condensate just undergoes small amplitude oscillations around that state. We will further assume that the displacement ξ\xi and the wave function width, deviated from σ0\sigma_{0} and defined by w=σ0+σw=\sigma_{0}+\sigma, are small as compared with σ0\sigma_{0}, namely ξσ0\xi\ll\sigma_{0} and σσ0\sigma\ll\sigma_{0}.

Retaining terms up to linear order in ξ\xi and σ\sigma of interest, we have found the Lagrange equations for the parameters by the variational approach as

z˙+2(k+κσσ0)1z2sinϕ=0,\displaystyle\dot{z}+2\left(k+\kappa\frac{\sigma}{\sigma_{0}}\right)\sqrt{1-z^{2}}\sin\phi=0, (11)
ϕ˙ΔEΛz\displaystyle\dot{\phi}-\Delta E-\Lambda z
2(k+κσσ0)z1z2cosϕ+2N2/N1ηξ=0,\displaystyle\quad-2\left(k+\kappa\frac{\sigma}{\sigma_{0}}\right)\frac{z}{\sqrt{1-z^{2}}}\cos\phi+2\sqrt{N_{2}/N_{1}}\,\eta\xi=0, (12)
ξ¨+ωξ2ξN1/N2ηzεσ=0,\displaystyle\ddot{\xi}+\omega_{\xi}^{2}\xi-\sqrt{N_{1}/N_{2}}\,\eta z-\varepsilon\sigma=0, (13)
σ¨+ωσ2σ2εξ=0.\displaystyle\ddot{\sigma}+\omega_{\sigma}^{2}\sigma-2\varepsilon\xi=0\,. (14)

The quantities that appear in the equations above are the integrals involving the condensate wave functions, namely

k0=𝑑x(12dψLdxdψRdx+x22ψLψR),\displaystyle k_{0}=-\int dx\left(\frac{1}{2}\frac{d\psi_{L}}{dx}\frac{d\psi_{R}}{dx}+\frac{x^{2}}{2}\psi_{L}\psi_{R}\right), (15)
k=k0g12N2πσ0𝑑xψLψRex2/σ02,\displaystyle k=k_{0}-\frac{g_{12}N_{2}}{\sqrt{\pi}\sigma_{0}}\int{dx\psi_{L}\psi_{R}\,e^{-x^{2}/\sigma_{0}^{2}}}\;, (16)
κ=g12N2πσ0𝑑xψLψR(12x2σ02)ex2/σ02,\displaystyle\kappa=\frac{g_{12}N_{2}}{\sqrt{\pi}\sigma_{0}}\int dx\psi_{L}\psi_{R}\left(1-2\frac{x^{2}}{\sigma_{0}^{2}}\right)e^{-x^{2}/\sigma_{0}^{2}}\;, (17)
ΔE=𝑑x[12(dψLdx)2+x22ψL2+g11N12ψL4]\displaystyle\Delta E=\int dx\left[\frac{1}{2}\left(\frac{d\psi_{L}}{dx}\right)^{2}+\frac{x^{2}}{2}\psi_{L}^{2}+\frac{g_{11}N_{1}}{2}\psi_{L}^{4}\right]
𝑑x[12(dψRdx)2+x22ψR2+g11N12ψR4],\displaystyle\qquad-\int dx\left[\frac{1}{2}\left(\frac{d\psi_{R}}{dx}\right)^{2}+\frac{x^{2}}{2}\psi_{R}^{2}+\frac{g_{11}N_{1}}{2}\psi_{R}^{4}\right], (18)
Λ=g11N12(𝑑xψL4+𝑑xψR4),\displaystyle\Lambda\,=\,\frac{g_{11}N_{1}}{2}\left(\int dx\psi_{L}^{4}+\int dx\psi_{R}^{4}\right)\;, (19)
η=g12N1N2πσ03𝑑x(ψL2ψR2)xex2/σ02,\displaystyle\eta=-\frac{g_{12}\sqrt{N_{1}N_{2}}}{\sqrt{\pi}\sigma_{0}^{3}}\int dx(\psi_{L}^{2}-\psi_{R}^{2})\,x\,e^{-x^{2}/\sigma_{0}^{2}}\;, (20)
ε=g12N1πσ03𝑑x(ψL2ψR2)(3xσ02x3σ03)ex2/σ02.\displaystyle\varepsilon=\frac{g_{12}N_{1}}{\sqrt{\pi}\sigma_{0}^{3}}\int_{-\infty}^{\infty}dx(\psi_{L}^{2}-\psi_{R}^{2})\left(3\frac{x}{\sigma_{0}}-2\frac{x^{3}}{\sigma_{0}^{3}}\right)e^{-x^{2}/\sigma_{0}^{2}}. (21)
ωξ2=1+g12N1πσ03𝑑x(ψL2+ψR2)(2x2σ021)ex2/σ02,\displaystyle\omega_{\xi}^{2}=1+\frac{g_{12}N_{1}}{\sqrt{\pi}\sigma_{0}^{3}}\int dx(\psi_{L}^{2}+\psi_{R}^{2})\left(2\frac{x^{2}}{\sigma_{0}^{2}}-1\right)e^{-x^{2}/\sigma_{0}^{2}}\;, (22)
ωσ2=1+3σ0+g22N22πσ03\displaystyle\omega_{\sigma}^{2}=1+\frac{3}{\sigma_{0}}+\frac{g_{22}N_{2}}{\sqrt{2\pi}\sigma_{0}^{3}}
+g12N1πσ03𝑑x(ψL2+ψR2)(210x2σ02+4x4σ04)ex2/σ02.\displaystyle+\frac{g_{12}N_{1}}{\sqrt{\pi}\sigma_{0}^{3}}\int dx(\psi_{L}^{2}+\psi_{R}^{2})\left(2-10\frac{x^{2}}{\sigma_{0}^{2}}+4\frac{x^{4}}{\sigma_{0}^{4}}\right)e^{-x^{2}/\sigma_{0}^{2}}\,. (23)

The expressions of kk (16) and κ\kappa (17) are determined by the wave function overlap between the left and right condensates, and thus kκ1k\sim\kappa\ll 1 by considering the weak link between bilateral condensates that depends not only on atomic scattering lengths but also on the number of the condensates with N2>N1N_{2}>N_{1}. Nevertheless, it will be seen that since the frequency of collectively atomic oscillations ωJξωxk\omega_{J\xi}\propto\omega_{x}\sqrt{k} with ωx=2π×12Hz\omega_{x}=2\pi\times 12\text{Hz} we adopt, the corresponding oscillation timescale t1/(ωxk)t\propto 1/(\omega_{x}\sqrt{k}) is constrained to be not larger than the typical lifetime of this type of the condensates of order 10s\sim 10\text{s} Becker2008 , giving k>104k>10^{-4} where N2N_{2} cannot be too large. The variation of the Gaussian width induces a time-dependent modification to the parameter k+κσ/σ0k+\kappa\sigma/\sigma_{0} in (11) and (12). In the single BEC experiment Smerzi1997 ; Raghavan1999 , the similar modification to kk can be achieved by the time-dependent laser intensity or the magnetic field with the possibility to induce the Shapiro effect Grond2011 , analog Kaptiza pendulum Boukobza2010 and even chaotic motions Abdullaev2000 ; Lee2001 ; Jiang2014 ; Tomkovic2017 . Here, since σσ0\sigma\ll\sigma_{0}, the time-dependent modification is relatively small compared with kk to be ignored in this work. Also, for simplicity, we consider that the spatial part of the wave function ψ1(x,t)\psi_{1}(x,t) is symmetric with respect to x=0x=0 shown in Fig. 2, thus giving ΔE=0\Delta E=0 in (12). The coupling of the population imbalance zz to the relative phase ϕ\phi in (12) is given by the strength Λ\Lambda in (19), a tunable value by changing the scattering lengths of atoms in bilateral condensates.

It is known that the dynamical equations of pair variables (z,ϕ)(z,\phi) in the single BEC system show an analogy to that of a nonrigid pendulum model, whose length depends on the angular momentum zz Raghavan1999 . Nevertheless, in our binary BECs system, the pair variables (z,ϕ)(z,\phi) are also coupled to the displacement of the wave function ψ2\psi_{2} with the coupling constant η\eta in (20) depending on the interspecies coupling constant g12g_{12} of atoms, and also the wave function overlap between the bilateral and central condensates. With the parameters shown in Fig. 1 and the chosen scattering lengths obeying Δ<0\Delta<0, which lead to the spatial parts of the wave functions schematically shown in Fig. 1, η\eta is of order one η1.0\eta\sim 1.0, and ε0.1\varepsilon\sim 0.1 due to cancellation between the contributions from the left and right wave functions. In the case of the initial deviation of the wave-function width σ(0)=σ˙(0)=0\sigma(0)=\dot{\sigma}(0)=0, the time-dependent σ\sigma is driven by ξ\xi, giving σεξ\sigma\sim\varepsilon\xi, which in turn leads to the term εσ\varepsilon\sigma in the equation of ξ\xi (13) of the order of ε2ξ\varepsilon^{2}\xi with ε2η\varepsilon^{2}\ll\eta to be safely ignored. From the viewpoint of the wave function ψ2\psi_{2}, the presence of ψR\psi_{R} and ψL\psi_{L} will make trap potential narrower and affects slightly the natural vibration frequency in ξ\xi and give a modified frequency ωξ\omega_{\xi} in (22). Since we just focus on various types of coherent oscillations between bilateral condensates with the dynamical variables z,ϕz,\,\phi, apart from their mutual coupling, they are also coupled to the displacement of the central condensate ξ\xi. Ignoring the εσ\varepsilon\sigma term in (13), the small deviation from the wave-function width σ\sigma is decoupled. In fact, in this approximation, once the time-dependence of ξ\xi is found, the equation of σ\sigma in (14) can be solved, and then plugging all solutions to the equations (86) in Appendix can find α\alpha and β\beta. Also notice that taking into account the time-dependent σ\sigma by involving (14) certainly can improve the agreement with the full numerical results. Thus, after the further replacement ξ2N2/N1ξ\xi\rightarrow\sqrt{2N_{2}/N_{1}}\xi, the set of the key equations can be simplified as

z˙+2k1z2sinϕ=0,\displaystyle\dot{z}+2k\sqrt{1-z^{2}}\sin\phi=0\,, (24)
ϕ˙Λz2kz1z2cosϕ+ηξ=0,\displaystyle\dot{\phi}-\Lambda z-2k\frac{z}{\sqrt{1-z^{2}}}\cos\phi+\eta\xi=0\,, (25)
ξ¨+ωξ2ξηz=0.\displaystyle\ddot{\xi}+\omega_{\xi}^{2}\xi-\eta z=0\,. (26)

These are the main results of this section and we will term them the coupled pendulums (CP) dynamics. The corresponding Hamiltonian then reads as

H=Λ2z22k1z2cos(ϕ)+12ξ˙2+ωξ22ξ2ηzξ.\displaystyle H=\frac{\Lambda}{2}z^{2}-2k\sqrt{1-z^{2}}\cos{\phi}+\frac{1}{2}\dot{\xi}^{2}+\frac{\omega_{\xi}^{2}}{2}\xi^{2}-\eta\,z\,\xi\,. (27)

In our binary BECs case, the above Hamiltonian also shows an analogy between the dynamics of coupled pendulums and the BEC dynamics as in Ref.  Raghavan1999 .

III Equilibrium Solutions and Stability analysis

In the two-component BEC system, the relevant equations of motion for describing quantum coherent oscillations between bilateral condensates in the presence of the central condensate are effectively described by the dynamics of two-coupled pendulums. In this case, the stationary states with the vanishing time-derivative terms now obey

2k1z02sin(ϕ0)=0,\displaystyle 2k\sqrt{1-z_{0}^{2}}\sin{\phi_{0}}=0, (28)
Λz0+2kz01z02cos(ϕ0)ηξ0=0,\displaystyle\Lambda z_{0}+2k\frac{z_{0}}{\sqrt{1-z_{0}^{2}}}\cos{\phi_{0}}-\eta\,\xi_{0}=0, (29)
ωξ2ξ0ηz0=0.\displaystyle\omega_{\xi}^{2}\xi_{0}-\eta\,z_{0}=0. (30)

As a result, the stationary relative phase is given by

ϕ0=0,orϕ0=±π.\phi_{0}=0\,,\qquad\text{or}\qquad\phi_{0}=\pm\pi. (31)

Also, Eq. (30) leads to a relation between the population imbalance z0z_{0} and the displacement of the central condensate ξ0\xi_{0}

ξ0=ωξ2ηz0.\xi_{0}=\omega_{\xi}^{-2}\eta z_{0}. (32)

Inserting the relation (32) into Eq. (29) together with ϕ0=0\phi_{0}=0 or ±π\pm\pi, the population imbalance is found to be

z0=0,orz0=±14k2Λeff2,\displaystyle z_{0}=0,\qquad\text{or}\qquad z_{0}=\pm\sqrt{1-\frac{4k^{2}}{\Lambda_{\text{eff}}^{2}}}, (33)

where Λeff=Λωξ2η2\Lambda_{\text{eff}}=\Lambda-\omega_{\xi}^{-2}\eta^{2}. The solution z00z_{0}\neq 0 exists as long as Λeff/2k1\Lambda_{\text{eff}}/2k\geq 1 or Λeff/2k1\Lambda_{\text{eff}}/2k\leq-1 in the MQST state for the atom system. Notice that Λeff\Lambda_{\text{eff}} can be positive or negative, and it is tunable by changing either g11g_{11} or g12g_{12} to vary kk (16), Λ\Lambda (19), and η\eta (20) . There are four different types of stationary states, (z0,ϕ0)=(z_{0},\,\phi_{0})= (0, 0)(0,\,0)-I, (0,±π)(0,\,\pm\pi)-II, (±14k2/Λeff2,  0)(\pm\sqrt{1-4k^{2}/\Lambda_{\text{eff}}^{2}},\,\,0)-III and (±14k2/Λeff2,±π)(\pm\sqrt{1-4k^{2}/\Lambda_{\text{eff}}^{2}},\,\,\pm\pi)-IV, summarized in Table 1. Furthermore, the corresponding energy for each stationary states can be computed through (27) using the relation of stationary-state solutions (32), where the results are listed in Table 2. In the case Λeff/2k>1\Lambda_{\text{eff}}/2k>1, the ground state is in configuration I where the bilateral condensates have the same populations z0=0z_{0}=0, thus the central condensate is centered at ξ0=0\xi_{0}=0, consistent with Fig. 1. As for Λeff/2k<1\Lambda_{\text{eff}}/2k<-1, the ground state shows the population imbalance for z00z_{0}\neq 0 and the wavefunction ψ2\psi_{2} is centered at ξ00\xi_{0}\neq 0 due to (32) for a given z0z_{0} also seen in Fig. 1.

We next study the stability of the system around the equilibrium solutions. To do so, we consider the small perturbations around ξ0\xi_{0}, z0z_{0}, and ϕ0\phi_{0} defined by

ξ(t)\displaystyle\xi(t) =ξ0+δξ(t),\displaystyle=\xi_{0}+\delta\xi(t), (34)
z(t)\displaystyle z(t) =z0+δz(t),\displaystyle=z_{0}+\delta z(t), (35)
ϕ(t)\displaystyle\phi(t) =ϕ0+δϕ(t).\displaystyle=\phi_{0}+\delta\phi(t). (36)

Substituting (34)–(36) into Eqs. (24)–(26), and using ξ0=ωξ2ηz0\xi_{0}=\omega_{\xi}^{-2}\eta z_{0}, one obtains the linearized equations of motion for δξ(t)\delta\xi(t) and δz(t)\delta z(t) as

(δξ¨(t)δz¨(t))=(ωξ2η2kη1z02cosϕ02kcosϕ0[(Λ+2Λz02ωξ2η2z02)1z022kcos(ϕ0)])(δξ(t)δz(t)).\displaystyle\begin{pmatrix}\delta\ddot{\xi}(t)\\[8.0pt] \delta\ddot{z}(t)\end{pmatrix}=\begin{pmatrix}-\omega_{\xi}^{2}\quad&\eta\\[8.0pt] {\displaystyle{2k\eta\sqrt{1-z_{0}^{2}}\cos\phi_{0}}}\quad&{\displaystyle 2k\cos\phi_{0}\left[\frac{(-\Lambda+2\Lambda z_{0}^{2}-\omega_{\xi}^{-2}\eta^{2}z_{0}^{2})}{\sqrt{1-z_{0}^{2}}}-2k\cos{\phi_{0}}\right]}\end{pmatrix}\begin{pmatrix}\delta\xi(t)\\[8.0pt] \delta z(t)\end{pmatrix}. (37)
Table 1: Stationary states
Label z0z_{0} ϕ0\phi_{0} ω±\omega_{\pm}
I 0  0 [ωξ2+ωJξ2±(ωξ2ωJξ2)2+8kη2]/2\Big{[}\omega_{\xi}^{2}+\omega_{J\xi}^{2}\pm\sqrt{(\omega_{\xi}^{2}-\omega_{J\xi}^{2})^{2}+{8k\eta^{2}}}\,\Big{]}/2
II 0 ±π\pm\pi [ωξ2+ωJξ2±(ωξ2ωJξ2)28kη2]/2\Big{[}\omega_{\xi}^{2}+\omega_{J\xi}^{2}\pm\sqrt{(\omega_{\xi}^{2}-\omega_{J\xi}^{2})^{2}-{8k\eta^{2}}}\,\Big{]}/2
III ±14k2/Λeff2\pm\sqrt{1-4k^{2}/\Lambda_{\text{eff}}^{2}}  0 [ωξ2+ωJξ2±(ωξ2ωJξ2)2+8kη21z02]/2\Big{[}\omega_{\xi}^{2}+\omega_{J\xi}^{2}\pm\sqrt{(\omega_{\xi}^{2}-\omega_{J\xi}^{2})^{2}+{8k\eta^{2}}\sqrt{1-z_{0}^{2}}}\,\Big{]}/2
IV ±14k2/Λeff2\pm\sqrt{1-4k^{2}/\Lambda_{\text{eff}}^{2}} ±π\pm\pi [ωξ2+ωJξ2±(ωξ2ωJξ2)28kη21z02]/2\Big{[}\omega_{\xi}^{2}+\omega_{J\xi}^{2}\pm\sqrt{(\omega_{\xi}^{2}-\omega_{J\xi}^{2})^{2}-{8k\eta^{2}}\sqrt{1-z_{0}^{2}}}\,\Big{]}/2
Table 2: Mean field extreme for different Λeff/2k\Lambda_{\text{eff}}/2k parameters
stable unstable energy
Λeff/2k>1\Lambda_{\text{eff}}/2k>1 I, IV II EIV>EIE_{\text{IV}}>E_{\text{I}}
1<Λeff/2k<1-1<\Lambda_{\text{eff}}/2k<1 I, II none EII>EIE_{\text{II}}>E_{\text{I}}
Λeff/2k<1\Lambda_{\text{eff}}/2k<-1 II, III I EII>EIIIE_{\text{II}}>E_{\text{III}}

Notice that the linear term in δϕ\delta\phi vanishes where ϕ0=0\phi_{0}=0 or ±π\pm\pi for the stationary-state configurations is evaluated in obtaining the linearized equations of motion. Considering the oscillation motion of δξ(t)\delta\xi(t) and δz(t)\delta z(t) with δξ(t),δz(t)exp(iωt)\delta\xi(t),\delta z(t)\propto\exp(i\omega t), we obtain the eigenfrequencies

ω±2=\displaystyle\omega_{\pm}^{2}=
12[ωξ2+ωJξ2±(ωξ2ωJξ2)2+8kη21z02cos(ϕ0)].\displaystyle\frac{1}{2}\left[\omega_{\xi}^{2}+\omega_{J\xi}^{2}\pm\sqrt{\left(\omega_{\xi}^{2}-\omega_{J\xi}^{2}\right)^{2}+{8k\eta^{2}}\,\sqrt{1-z_{0}^{2}}\cos{\phi_{0}}}\,\right]. (38)

In addition to the nature frequency ωξ\omega_{\xi} for the central condensate, there exists an oscillating frequency

ωJξ2=ωJ2+2kcosϕ0ωξ2η2z021z02\displaystyle\omega_{J\xi}^{2}=\omega_{J}^{2}+\frac{2k\cos\phi_{0}\,\omega_{\xi}^{-2}\eta^{2}z_{0}^{2}}{\sqrt{1-z_{0}^{2}}}\, (39)

with

ωJ2=2kcosϕ0[ 2kcos(ϕ0)+Λ(12z02)1z02]\omega_{J}^{2}=2k\cos\phi_{0}\,\left[\,2k\cos{\phi_{0}}+\frac{\Lambda(1-2z_{0}^{2})}{\sqrt{1-z_{0}^{2}}}\,\right] (40)

which manifests the nature frequency for quantum oscillations between two bilateral condensates ωJ\omega_{J} with the effects from the dynamics of the central condensate Smerzi1997 . However, the true eigenfrequencies are the mixture of these two fundamental frequencies ωξ,ωJξ\omega_{\xi},\,\omega_{J\xi} as they couple. With the parameters in the immiscible regime, typically for small kk where the wave-function overlap between bilateral condensates is small, this gives ωξωJξ\omega_{\xi}\gg\omega_{J\xi}. In the two-component BECs system, the existing oscillation frequency ωξ\omega_{\xi} of the central condensate will drive the dynamics of the pair variables (z,ϕ)(z,\phi) into the oscillatory motions with frequencies ω\omega_{-} and ω+\omega_{+} where ω+ω\omega_{+}\gg\omega_{-}.

Later we will manipulate the initial conditions of z(0),ϕ(0)z(0),\,\phi(0) as well as ξ(0),ξ˙(0)\xi(0),\,\dot{\xi}(0) to effectively switch off the fast varying mode of frequency ω+\omega_{+}, giving the results presumably to those in Ref. Smerzi1997 . Then, the presence of the fast varying mode for more general initial conditions will also be studied to see its effects on quantum coherent phenomena (z,ϕ)(z,\phi). We will learn that, in the case of relatively small k103k\sim 10^{-3} with very small wavefunction overlap, the time average over the evolution of the fast varying mode is adopted to find the deviation from the results in Ref. Smerzi1997 . Those trajectories are regular motion moving around the stable states of the system listed in Table 2. However, we also probe the parameter regime with relatively large wave-function overlap with relatively large value k102k\sim 10^{-2} for amplifying the influence of the fast varying mode that can even drive the dynamics of zz and ϕ\phi to the chaotic behavior, if the system starts from the unstable state also seen in Table 2. In our system, the dynamics of quantum coherence oscillations (z,ϕ)(z,\phi) is very sensitive to the tunneling energy kk. For even larger kk, the proposed wave-function ansatz fails to describe the evolution of (z,ϕ)(z,\phi) where the dynamics of the system can only be explored by directly solving time-dependent GP equations, which is beyond the scope of the present work.

IV Regular motion

IV.1 General solutions for small amplitude oscillations

After studying the stationary-state configurations and examining their dynamical stability, we step forward to discuss the time evolution of the system when zz and ξ\xi undergo small amplitude oscillations around the stationary states. Recalling the linearized equations of motion (37), the general solutions will be the superposition of two eigenfunctions,

δξ(t)=\displaystyle\delta\xi(t)= A++eiω+t+A+eiω+t\displaystyle A_{++}e^{i\omega_{+}t}+A_{+-}e^{-i\omega_{+}t}
+A+eiωt+Aeiωt,\displaystyle\;\qquad\qquad+A_{-+}e^{i\omega_{-}t}+A_{--}e^{-i\omega_{-}t}, (41)
δz(t)=\displaystyle\delta z(t)= B++eiω+t+B+eiω+t\displaystyle B_{++}e^{i\omega_{+}t}+B_{+-}e^{-i\omega_{+}t}
+B+eiωt+Beiωt.\displaystyle\;\qquad\qquad+B_{-+}e^{i\omega_{-}t}+B_{--}e^{-i\omega_{-}t}. (42)

Substituting Eqs. (41) and (42) into (37), one gets the relations among those coefficients

A+,±=P+B+,±,\displaystyle A_{+,\pm}=P_{+}\,B_{+,\pm}\;, (43)
A,±=PB,±,\displaystyle A_{-,\pm}=P_{-}\,B_{-,\pm}\;, (44)

where P±P_{\pm} depends on the stationary-state solutions obtained as

P±=\displaystyle P_{\pm}= 14kη1z02cos(ϕ0)[ωξ2ωJξ2\displaystyle\frac{-1}{{4k\eta\sqrt{1-z_{0}^{2}}\cos{\phi_{0}}}}\Bigg{[}{\omega_{\xi}^{2}-\omega_{J\xi}^{2}}
±(ωξ2ωJξ2)2+8kη21z02cos(ϕ0)].\displaystyle\qquad\pm\sqrt{\left(\omega_{\xi}^{2}-\omega_{J\xi}^{2}\right)^{2}+8k\eta^{2}\sqrt{1-z_{0}^{2}}\cos{\phi_{0}}}\,\Bigg{]}. (45)

According to the experimental feasibility where an application of a magnetic gradient trap may cause the displacements of condensates McCarron2011 ; Eto2015 ; Eto2016 , we then consider the initial conditions such as δϕ(0)=0\delta\phi(0)=0, and δξ˙(0)=0\delta\dot{\xi}(0)=0. The corresponding solutions are

δξ(t)=\displaystyle\delta\xi(t)= (P+PP+Pδz(0)+P+P+Pδξ(0))cos(ω+t)\displaystyle\left(\frac{-P_{+}P_{-}}{P_{+}-P_{-}}\delta z(0)+\frac{P_{+}}{P_{+}-P_{-}}\delta\xi(0)\right)\cos{\omega_{+}t}
+(P+PP+Pδz(0)PP+Pδξ(0))cos(ωt),\displaystyle+\left(\frac{P_{+}P_{-}}{P_{+}-P_{-}}\delta z(0)-\frac{P_{-}}{P_{+}-P_{-}}\delta\xi(0)\right)\cos{\omega_{-}t}, (46)
δz(t)=\displaystyle\delta z(t)= (PP+Pδz(0)+1P+Pδξ(0))cos(ω+t)\displaystyle\left(\frac{-P_{-}}{P_{+}-P_{-}}\delta z(0)+\frac{1}{P_{+}-P_{-}}\delta\xi(0)\right)\cos{\omega_{+}t}
+(P+P+Pδz(0)1P+Pδξ(0))cos(ωt).\displaystyle+\left(\frac{P_{+}}{P_{+}-P_{-}}\delta z(0)-\frac{1}{P_{+}-P_{-}}\delta\xi(0)\right)\cos{\omega_{-}t}. (47)

In the case that ψL\psi_{L} and ψR\psi_{R} are very spatially separated, resulting in small tunneling energy kk in (16), two frequencies will have very different values, ωξωJξ\omega_{\xi}\gg\omega_{J\xi}, since ωJξk\omega_{J\xi}\propto\sqrt{k} in (39). Then the eigenfrequencies (38) and the coefficients (45) can be further approximated respectively as

ω+ωξ,ωωJξ\displaystyle\omega_{+}\rightarrow\omega_{\xi}\quad\text{,}\quad\omega_{-}\rightarrow\omega_{J\xi} (48)

and

P+ωξ22kη1z02cos(ϕ0)+𝒪(k),Pηωξ2+𝒪(k).\displaystyle P_{+}\rightarrow-\frac{\omega_{\xi}^{2}}{2k\eta\sqrt{1-z_{0}^{2}}\cos{\phi_{0}}}+\mathcal{O}\left(k\right)\text{,}\,\,\,P_{-}\rightarrow\frac{\eta}{\omega_{\xi}^{2}}+\mathcal{O}\left(k\right). (49)

The general solutions for small kk now become

δξ(t)\displaystyle\delta\xi(t)\simeq
1ωξ4[ωξ2(ηδz(0)ωξ2δξ(0))cos(ωξt)\displaystyle\frac{1}{\omega_{\xi}^{4}}\bigg{[}-\omega_{\xi}^{2}\left(\eta\,\delta z(0)-\omega_{\xi}^{2}\delta\xi(0)\right)\cos{\omega_{\xi}t}
+η(ωξ2δz(0)+2kη1z02cos(ϕ0)δξ(0))cos(ωJξt)],\displaystyle+\eta\left(\omega_{\xi}^{2}\delta z(0)+2k\eta\sqrt{1-z_{0}^{2}}\cos{\phi_{0}}\delta\xi(0)\right)\cos{\omega_{J\xi}t}\bigg{]}, (50)
δz(t)\displaystyle\delta z(t)\simeq
1ωξ4[2kη1z02cos(ϕ0)(ηδz(0)ωξ2δξ(0))cos(ωξt)\displaystyle\frac{1}{\omega_{\xi}^{4}}\left[2k\eta\sqrt{1-z_{0}^{2}}\cos{\phi_{0}}\left(\eta\,\delta z(0)-\omega_{\xi}^{2}\delta\xi(0)\right)\cos{\omega_{\xi}t}\right.
+ωξ2(ωξ2δz(0)+2kη1z02cos(ϕ0)δξ(0))cos(ωJξt)].\displaystyle\left.\ +\omega_{\xi}^{2}\left(\omega_{\xi}^{2}\delta z(0)+2k\eta\sqrt{1-z_{0}^{2}}\cos{\phi_{0}}\delta\xi(0)\right)\cos{\omega_{J\xi}t}\right]. (51)

In this binary BEC system, it is important to explore the effects from the additional mode with frequency ωξ\omega_{\xi} on the coherent oscillations between bilateral condensates. To do so, we first choose the initial displacement of the central condensate and the initial population imbalance as Pδz(0)+δξ(0)=0{-P_{-}}\delta z(0)+\delta\xi(0)=0 in (46) and (47) where this choice of the initial conditions leads to the amplitude of the rapidly oscillatory motion vanishing, thus leaving with the slowly varying motion only. For the situations of small kk, using (49) we obtain a linear relation,

δξ(0)=ωξ2ηδz(0),\delta\xi(0)=\omega_{\xi}^{-2}\eta\,\delta z(0)\,, (52)

as in the stationary-state solutions (32). In this case, δξ(t)\delta\xi(t) and δz(t)\delta z(t) undergo slow oscillations and the linearized solutions (50) and (51) of the system read as

ξs(t)ξ0+ωξ2ηδz(0)cosωJξt,\displaystyle\xi_{s}(t)\simeq\xi_{0}+\omega_{\xi}^{-2}\eta\delta z(0)\cos\omega_{J\xi}t\;, (53)
zs(t)z0+δz(0)cosωJξt.\displaystyle z_{s}(t)\simeq z_{0}+\delta z(0)\cos\omega_{J\xi}t. (54)

Notice that together with (32) and (52), ξ(t)=ωξ2ηz(t)\xi(t)=\omega_{\xi}^{-2}\eta z(t) and Eqs.(24)–(26) reduce to the same set of the equations in the single BEC case in Ref. Smerzi1997 by letting

Λeff=Λη2ωξ2.\Lambda_{\text{eff}}=\Lambda-\eta^{2}\omega_{\xi}^{-2}\,. (55)

In addition, the solution (53) shows ξ˙sξs\dot{\xi}_{s}\ll\xi_{s}.

Another interesting initial condition is δξ(0)=0\delta\xi(0)=0 and in this case we can activate properly the rapidly oscillatory motion with the solutions

ξ(t)ξs(t)+Cξcosωξt,\displaystyle\xi(t)\simeq\xi_{s}(t)+C_{\xi}\cos\omega_{\xi}t\;, (56)
z(t)zs(t)+Czcosωξt,\displaystyle z(t)\simeq z_{s}(t)+C_{z}\cos\omega_{\xi}t\;, (57)

where

Cξ=ηωξ2δz(0),\displaystyle C_{\xi}=-{\eta\omega_{\xi}^{-2}\,\delta z(0)}\;, (58)
Cz=2ωξ4kη21z02cos(ϕ0)δz(0)\displaystyle C_{z}={2\omega_{\xi}^{-4}k\eta^{2}\sqrt{1-z_{0}^{2}}\cos{\phi_{0}}\delta z(0)}\, (59)

and CzCξC_{z}\ll C_{\xi} in the regime of small kk. Later we will explore the effects from the rapidly oscillatory mode on the dynamics of (z,ϕ)(z,\phi), in the system. However, this system involves two largely separate frequencies, namely ωξωJξ\omega_{\xi}\gg\omega_{J\xi}, for small kk with small wave-function overlap between bilateral condensates. Although numerical trajectories of the CP dynamics is straightforward, we find more convenient to represent the trajectories in Poincaré maps (stroboscopic plots at every period, 2π/ωξ2\pi/\omega_{\xi}). These trajectories in Poincaré maps can be analytically understood by considering the time-averaged effects over the evolution of the ωξ\omega_{\xi} mode in the timescale TT where 1/ωξT1/ωJξ1/\omega_{\xi}\ll T\ll 1/\omega_{J\xi}, with which we construct the corresponding effective potential in next section.

IV.2 The effective potential

In order to construct an effective potential for the (z,ϕ)(z,\phi) dynamics, we substitute (24) into (27) to obtain

4k2(1z2)z˙2\displaystyle 4k^{2}(1-z^{2})-\dot{z}^{2}
=(Λ2z2+ξ˙22+ωξ2ξ22ηξzH(0))2,\displaystyle\qquad=\left(\frac{\Lambda}{2}z^{2}+\frac{\dot{\xi}^{2}}{2}+\frac{\omega_{\xi}^{2}\xi^{2}}{2}-\eta\xi\,z-H(0)\right)^{2}, (60)

where H(0)H(0) is a constant value determined by the initial state (z(0),ϕ(0),ξ(0),ξ˙(0))(z(0),\,\phi(0),\,\xi(0),\,\dot{\xi}(0)). To get an effective potential for the small amplitude oscillations around z0z_{0} and ξ0\xi_{0} that includes the effects from the ωξ\omega_{\xi} mode of the general C2C^{2}, C3C^{3}, and C4C^{4} terms, we substitute (56) and (57) into (60) giving

4k2[1(zs+Czcos(ωξt))2](zs˙ωξCzsin(ωξt))2=\displaystyle 4k^{2}\left[1-(z_{s}+C_{z}\cos{\omega_{\xi}t})^{2}\right]-{(\dot{{z}_{s}}-\omega_{\xi}C_{z}\sin{\omega_{\xi}t})}^{2}=
[Λ2(zs+Czcos(ωξt))2+12(ωξCξsin(ωξt))2+ωξ22(ηωξ2zs\displaystyle\bigg{[}\frac{\Lambda}{2}({z_{s}}+C_{z}\cos{\omega_{\xi}t})^{2}+\frac{1}{2}(\omega_{\xi}C_{\xi}\sin{\omega_{\xi}t})^{2}+\frac{\omega_{\xi}^{2}}{2}(\eta\omega_{\xi}^{-2}{z_{s}}
+Cξcos(ωξt))2η(ηωξ2zs+Cξcos(ωξt))(zs+Czcos(ωξt))\displaystyle+C_{\xi}\cos{\omega_{\xi}t})^{2}-\eta(\eta\omega_{\xi}^{-2}z_{s}+C_{\xi}\cos{\omega_{\xi}t})\,(z_{s}+C_{z}\cos{\omega_{\xi}t})
H(0)]2,\displaystyle-H(0)\bigg{]}^{2}\,, (61)

where, since ξ˙s2ξs2\dot{\xi}^{2}_{s}\ll\xi_{s}^{2}, the term ξ˙s2\dot{\xi}^{2}_{s} is ignored for keeping the leading-order effects. Although CzCξC_{z}\ll C_{\xi} in the small kk approximation, we still retain the terms of CzC_{z}, which are small, for seeing how they contribute to the effective potential. Let us now introduce the time-averaged population imbalance zz over the timescale TT for 1/ωξT1/ωJξ1/\omega_{\xi}\ll T\ll 1/\omega_{J\xi} as

zs(t)1T0T𝑑tzs(t)=z¯(t),\displaystyle\langle z_{s}(t)\rangle\equiv\frac{1}{T}\int_{0}^{T}dtz_{s}(t)=\bar{z}(t), (62)

where z¯(t)\bar{z}(t) will be determined self-consistently from the resulting effective potential. It is now straightforward to find the expression

z¯˙2+Veff(z¯)=4k2H(0)2,\displaystyle\dot{\bar{z}}^{2}+V_{\text{eff}}(\bar{z})=4k^{2}-H(0)^{2}\,, (63)

from which the effective potential is obtained as

Veff(z¯)=V0(z¯)+δV(z¯).\displaystyle V_{\text{eff}}(\bar{z})=V_{0}(\bar{z})+\delta V(\bar{z}). (64)

Among them, V0(z¯)V_{0}(\bar{z}) is the well-known effective potential for a single BEC in a double-well trap potential Smerzi1997

V0(z¯)=z¯2(4k2H(0)Λeff+Λeff24z¯2),\displaystyle V_{0}(\bar{z})=\bar{z}^{2}\left(4k^{2}-H(0)\Lambda_{\text{eff}}+\frac{\Lambda_{\text{eff}}^{2}}{4}\bar{z}^{2}\right), (65)

and δV(z¯)\delta V(\bar{z}) is the corrections due to the high-frequency ωξ\omega_{\xi} mode with the terms of CzC_{z} and CξC_{\xi}, given by

δV(z¯)\displaystyle\delta V(\bar{z})
=12[4k2+ωξ2Λ(H(0)Λeff2z¯2)+Λeff2z¯2]Cz2\displaystyle=\frac{1}{2}\left[4k^{2}+\omega_{\xi}^{2}-\Lambda\left(H(0)-\frac{\Lambda_{\text{eff}}}{2}\bar{z}^{2}\right)+\Lambda_{\text{eff}}^{2}\bar{z}^{2}\right]C_{z}^{2}
+ωξ2(Λeff2z¯2H(0))Cξ2+η(H(0)Λeff2z¯2)CzCξ\displaystyle+\omega_{\xi}^{2}\left(\frac{\Lambda_{\text{eff}}}{2}\bar{z}^{2}-H(0)\right)C_{\xi}^{2}+\eta\left(H(0)-\frac{\Lambda_{\text{eff}}}{2}\bar{z}^{2}\right)C_{z}C_{\xi}
ηωξ22CzCξ3+18(2Λeffωξ2+5η2)Cz2Cξ2\displaystyle-\frac{\eta\omega_{\xi}^{2}}{2}C_{z}C_{\xi}^{3}+\frac{1}{8}\left(2\Lambda_{\text{eff}}\omega_{\xi}^{2}+5\eta^{2}\right)C_{z}^{2}C_{\xi}^{2}
3ηΛ8Cz3Cξ+3Λ232Cz4+ωξ44Cξ4.\displaystyle-\frac{3\eta\Lambda}{8}C_{z}^{3}C_{\xi}+\frac{3\Lambda^{2}}{32}C_{z}^{4}+\frac{\omega_{\xi}^{4}}{4}C_{\xi}^{4}\,. (66)

Furthermore, since ξ(t)=ωξ2ηz(t)\xi(t)=\omega_{\xi}^{-2}\eta z(t) holds true in the case of Cz=Cξ=0C_{z}=C_{\xi}=0, the whole dynamics of (z,ϕ,ξ)(z,\phi,\xi) can reduce to that of (z,ϕ)(z,\phi) with symmetry of ΛeffΛeff\Lambda_{\text{eff}}\rightarrow-\Lambda_{\text{eff}}, and ϕϕ+π\phi\rightarrow-\phi+\pi as in Ref.  Smerzi1997 . Nonetheless, since CzC_{z} and CξC_{\xi} are in general nonzero, the system then does not obey the above-mentioned symmetry. We will verify this in the later numerical studies.

IV.3 Results and Discussions

IV.3.1 Λeff/2k>2\Lambda_{\textsf{eff}}/2k>2 (Josephson oscillation, running phase and π\pi-mode self trapping)

Consider the numbers of atoms in this binary condensates with the atom numbers N1=200,N2=1000N_{1}=200,\,N_{2}=1000. In Fig. 3, we have chosen scattering lengths a11=97.5a0,a22=50a0,a12=85a0a_{11}=97.5a_{0},\,a_{22}=50a_{0},\,a_{12}=85a_{0}. As stated previously, we prepare the initial states of the condensates, the Gaussian-like spatial functions ψL(x)\psi_{L}(x) and ψR(x)\psi_{R}(x) together with the initial wave function of the central condensate ψ2(x,t=0)\psi_{2}(x,t=0) with width σ0\sigma_{0} determined numerically from finding the stationary-state solutions of GP equations (3) and (4) with the forms shown in Fig. 2. The resulting wave functions can in turn give the values of parameters k=0.0025k=0.0025\,, Λ=1.91\Lambda=1.91, and η=1.25\eta=1.25 via (16), (19), and (20), which lead to Λeff/2k=68>2\Lambda_{\text{eff}}/2k=68>2.

We first depict the trajectories by directly solving the equations of motions (24)–(26) where the initial conditions are slight away from the stationary-state of ϕ0=0\phi_{0}=0 and z0=ξ0=0z_{0}=\xi_{0}=0 shown in Fig. 3(a). The initial conditions are chosen as δϕ(0)=δξ˙(0)=0\delta\phi(0)=\delta\dot{\xi}(0)=0 together with various choices of δz(0)\delta z(0) and δξ(0)\delta\xi(0), obeying δξ(0)=ηωξ2δz(0)\delta\xi(0)=\eta\omega_{\xi}^{-2}\delta z(0), with which only the ωJξ\omega_{J\xi} mode becomes relevant. Thus, by varying δz(0)\delta z(0) and δξ(0)\delta\xi(0) accordingly, the plot shows the regimes of Josephon-like oscillations for z¯(0)=z0+δz(0)<z¯c(0)\bar{z}(0)=z_{0}+\delta z(0)<\bar{z}_{c}(0), and MQST with the running phase for z¯(0)>z¯c(0)\bar{z}(0)>\bar{z}_{c}(0) as in Ref. Smerzi1997 . The critical z¯\bar{z} in this case is z¯c(0)=0.24\bar{z}_{c}(0)=0.24, obtained from (LABEL:zc_lambda_ge2) below. Also, we see the consistency of the solutions (50) and (51) with those of the equations in (24)–(26). In the case of Cz=Cξ=0C_{z}=C_{\xi}=0, the effective potential (64) reduces to that in Ref. Smerzi1997 drawn in Fig. 3(c) of a double-well form. Thus, for z¯(0)<z¯c(0)\bar{z}(0)<\bar{z}_{c}(0) the system has large enough kinetic energy z¯˙2\dot{\bar{z}}^{2} to move forward and backward between two potential minima around the stationary state z0=0z_{0}=0 and ϕ0=0\phi_{0}=0 undergoing the Josephson oscillations. As z¯(0)\bar{z}(0) increases to z¯(0)=z¯c(0)\bar{z}(0)=\bar{z}_{c}(0), the initial kinetic energy of the system drives z¯\bar{z} rolling down toward one of the potential minima and then climbing up the potential hill to reach the state of z¯=0\bar{z}=0 just with z¯˙=0\dot{\bar{z}}=0, obeying the condition

z¯˙c(0)2+Veff(z¯c(0))=Veff(z¯=0).\displaystyle\dot{\bar{z}}_{c}(0)^{2}+V_{\text{eff}}(\bar{z}_{c}(0))=V_{\text{eff}}(\bar{z}=0)\,. (67)
Refer to caption
Figure 3: The phase portrait of the population imbalance zz and the phase difference between the bilateral condensates ϕ\phi as well as the plots of the effective potential with the parameter Λ=1.91,η=1.25,k=0.0026\Lambda=1.91,\,\eta=1.25,\,k=0.0026, giving Λeff/2k=68.02)\Lambda_{\text{eff}}/2k=68.02): (a) with the initial conditions, δz(0)=0.25,δξ(0)=0.31\delta z(0)=0.25,\,\delta\xi(0)=0.31 [blue (gray)], δz(0)=0.24,δξ(0)=0.30\delta z(0)=0.24,\,\delta\xi(0)=0.30 [red (dark gray)], and δz(0)=0.23,δξ(0)=0.29\delta z(0)=0.23,\,\delta\xi(0)=0.29 [green (light gray)] where the ωξ\omega_{\xi} mode is effectively switched; (b) with the initial conditions δz(0)=0.25,δξ(0)=0\delta z(0)=0.25,\delta\xi(0)=0 [blue (gray)], δz(0)=0.24,δξ(0)=0\delta z(0)=0.24,\,\delta\xi(0)=0 [red (dark gray)], and δz(0)=0.23,δξ(0)=0\delta z(0)=0.23,\,\delta\xi(0)=0 [green (light gray)] where the the ωξ\omega_{\xi} mode is effectively activated. Other initial conditions for both cases are δϕ(0)=0\delta\phi(0)=0, ξ˙(0)=0\dot{\xi}(0)=0. The plots (c) and (d) are drawn for the corresponding effective potential for (a) and (b), respectively.

Moreover, when z¯(0)>z¯c(0)\bar{z}(0)>\bar{z}_{c}(0), the relatively small kinetic energy constrains the system to move around one of the potential minima with z00z_{0}\neq 0, leading to the MQST, also shown in Fig. 3(c).

Considering the initial conditions mentioned above together with the population imbalance z¯(0)=z0+δz(0)\bar{z}(0)=z_{0}+\delta z(0) and the displacement of the central condensate ξ(0)=ξ0+δξ(0)\xi(0)=\xi_{0}+\delta\xi(0) obeying the relation ξ(0)=ηωξ2z¯(0)\xi(0)=\eta\omega_{\xi}^{-2}\bar{z}(0), H(0)H(0) becomes

H(0)=Λeff2z¯(0)22k1z¯(0)2cos(ϕ(0)).\displaystyle H(0)=\frac{\Lambda_{\text{eff}}}{2}\bar{z}(0)^{2}-2k\sqrt{1-\bar{z}(0)^{2}}\cos{\phi(0)}\,. (68)

Then, using Eq. (24) at the initial time to replace z¯˙c(0)\dot{\bar{z}}_{c}(0) in (67), for a given Λeff\Lambda_{\text{eff}}, one can find the value of z¯c(0)\bar{z}_{c}(0) with the effective potential (60) by setting Cz=Cξ=0C_{z}=C_{\xi}=0, which gives the known result in Ref. Smerzi1997 . For Λeff/2k>2\Lambda_{\text{eff}}/2k>2, if the initial relative phase 0|ϕ(0)|π/20\leq|{\phi(0)}|\leq\pi/2, then

z¯c2(0)=\displaystyle\bar{z}^{2}_{c}(0)= 4kΛeff2[Λeff2kcosϕ(0)\displaystyle\frac{4k}{\Lambda_{\text{eff}}^{2}}\bigg{[}\Lambda_{\text{eff}}-2k\cos\phi(0)
+Λeff(Λeff4k)cos2ϕ(0)+4k2cos4ϕ(0)],\displaystyle\,\,+\sqrt{\Lambda_{\text{eff}}\left(\Lambda_{\text{eff}}-4k\right)\cos^{2}\phi(0)+4k^{2}\cos^{4}\phi(0)}\bigg{]},

and when π/2|ϕ(0)|π\pi/2\leq|{\phi(0)}|\leq\pi we have

z¯c2(0)=\displaystyle\bar{z}^{2}_{c}(0)= 4kΛeff2[Λeff2kcosϕ(0)\displaystyle\frac{4k}{\Lambda_{\text{eff}}^{2}}\bigg{[}\Lambda_{\text{eff}}-2k\cos\phi(0)
Λeff(Λeff4k)cos2ϕ(0)+4k2cos4ϕ(0)].\displaystyle\,\,-\sqrt{\Lambda_{\text{eff}}(\Lambda_{\text{eff}}-4k)\cos^{2}\phi(0)+4k^{2}\cos^{4}\phi(0)}\bigg{]}. (70)

The critical value of z¯c(0)\bar{z}_{c}(0) as a function of ϕ(0)\phi(0) draws a boundary between the Josephson oscillations and the MQST mode to be discussed in the later numerical studies in Fig. 9. As for a negative Λeff\Lambda_{\text{eff}}, the solutions of z¯c2(0)\bar{z}^{2}_{c}(0) are given due to the following symmetry ΛeffΛeff\Lambda_{\text{eff}}\rightarrow-\Lambda_{\text{eff}} and ϕϕ+π\phi\rightarrow-\phi+\pi. The value of z¯c(0)\bar{z}_{c}(0) found in Fig. 3(a) and 3(c) can also been achieved from (LABEL:zc_lambda_ge2) and (70) with ϕ(0)=0\phi(0)=0 for the positive value of z¯c(0)\bar{z}_{c}(0).

Refer to caption
Figure 4: The evolution of the population imbalance zz (a) and the displacement of the central condensate (b) as a function of time in the case of the Josephson oscillations are compared between the results of the numerical solutions of the time-dependent GP equations (blue line), and the linearized equations in Eqs. (24)–(26) (red dashed line) with the same parameters described in the text and the initial conditions z(0)=0.1,ϕ(0)=ξ(0)=ξ˙(0)=0z(0)=-0.1,\phi(0)=\xi(0)=\dot{\xi}(0)=0.

In Fig. 3(b), we present the evolutions of z(t)z(t) and ϕ(t)\phi(t), which include the effects from the ωξ\omega_{\xi} mode with the same set parameters discussed previously (see figure caption for details). To make a comparison with the evolution in Fig. 3(a) of the slowly varying mode, it finds more convenient to represent the trajectories in Poincaré maps (stroboscopic plots at every period T=2π/ωξT=2\pi/\omega_{\xi}) by collecting the results from the section of ξ(0)=0\xi(0)=0 and ξ˙(0)>0\dot{\xi}(0)>0. It in turn can be further analyzed in terms of the effective potential as we will see later by considering the time-averaged effects from the fast oscillatory mode. The similar Josephson-like oscillations (z¯(0)<z¯c(0))(\bar{z}(0)<\bar{z}_{c}(0)) and the running phase MQST dynamics z¯(0)>z¯c(0)\bar{z}(0)>\bar{z}_{c}(0) are shown with a transition between them at the shifted critical value z¯c(0)=0.238\bar{z}_{c}(0)=0.238. The shifted critical value z¯c(0)\bar{z}_{c}(0) can be obtained from the conserved quantity

H(0)=\displaystyle H(0)= Λeff2z¯(0)22k1(z¯(0)+Cz)2cos(ϕ(0))\displaystyle\frac{\Lambda_{\text{eff}}}{2}\bar{z}(0)^{2}-2k\sqrt{1-(\bar{z}(0)+C_{z})^{2}}\cos{\phi(0)}
+(Λeff+ηωξ2)2Cz2+Λeffz¯(0)Cz+ωξ22Cξ2\displaystyle+\frac{(\Lambda_{\text{eff}}+\eta\omega_{\xi}^{-2})}{2}C_{z}^{2}+\Lambda_{\text{eff}}\bar{z}(0)C_{z}+\frac{\omega_{\xi}^{2}}{2}C_{\xi}^{2}
ηCzCξ,\displaystyle-\eta\,C_{z}C_{\xi}, (71)

where we have substituted the results of (53) and (54) into (27), and evaluated the Hamiltonian at an initial time. Again, substituting (24) at an initial time into (67) and with the help of effective potential (64), we can obtain the critical value of z¯c(0)\bar{z}_{c}(0) through the expression as follows

Λeff=\displaystyle\Lambda_{\text{eff}}= 1z¯c(0)2+4Czz¯c(0)Cz2{4k1(z¯c(0)+Cz)2cos((ϕ(0)))+ηCzCξη2ωξ22Cz2\displaystyle\frac{1}{\bar{z}_{c}(0)^{2}+4C_{z}\bar{z}_{c}(0)-C_{z}^{2}}\Bigg{\{}4k\sqrt{1-(\bar{z}_{c}(0)+C_{z})^{2}}\cos{(\phi(0))}+\eta C_{z}C_{\xi}-\frac{\eta^{2}\omega_{\xi}^{-2}}{2}C_{z}^{2}
+[(4k1(z¯(0)+Cz)2cos((ϕ(0)))+ηCzCξη2ωξ22Cz2)2+16k2z¯c(0)2(z¯c(0)2\displaystyle+\Bigg{[}\left(4k\sqrt{1-(\bar{z}(0)+C_{z})^{2}}\cos{(\phi(0))}+\eta\,C_{z}C_{\xi}-\frac{\eta^{2}\omega_{\xi}^{-2}}{2}C_{z}^{2}\right)^{2}+\frac{16k^{2}}{\bar{z}_{c}(0)^{2}}(\bar{z}_{c}(0)^{2}
+4Czz¯c(0)Cz2)[1Cz(2z¯c(0)+Cz)(1(z¯c(0)+Cz)2)cos2(ϕ(0))]]1/2},\displaystyle+4C_{z}\bar{z}_{c}(0)-C_{z}^{2})[1-C_{z}(2\bar{z}_{c}(0)+C_{z})-(1-(\bar{z}_{c}(0)+C_{z})^{2})\cos^{2}{(\phi(0))}]\Bigg{]}^{1/2}\Bigg{\}}, (72)

where CzC_{z} and CξC_{\xi} are given explicitly in terms of z0z_{0} and δzc(0)\delta z_{c}(0) via (59) and z¯c(0)=z0+δzc(0)\bar{z}_{c}(0)=z_{0}+\delta z_{c}(0) in (54). For a fixed Λeff\Lambda_{\text{eff}}, the shifted value of z¯c(0)\bar{z}_{c}(0) obtained from (IV.3.1) is consistent with the numerical result in Fig. 3(b).

In Fig. 4 we provide numerical comparison on the results of the coupled pendulums equations (24)–(26) and those by solving the full time-dependent GP equations (3) and (4). The initial wave functions for both condensates, as previously said, are prepared from the stationary solutions of GP equations with the form similar to Fig. 2. The wave function of the central condensate can be approximated by (5) with ξ(0)=α(0)=β(0)=0\xi(0)=\alpha(0)=\beta(0)=0 giving ξ˙(0)=0\dot{\xi}(0)=0. However, for the bilateral condensates, their initial wave functions are prepared for the functions of xx with the relative phase ϕ(0)=0\phi(0)=0 and then tune the wave functions to give z(0)0z(0)\neq 0. In Fig. 4, using the same parameters and the initial conditions we find consistency between the solutions from analogous CP dynamics and the time-dependent GP equations.

Refer to caption
Figure 5: The phase portrait with the parameters Λ=1.80\Lambda=1.80, η=1.31\eta=1.31, k=0.017(Λeff/2k=2.47)k=0.017\,(\Lambda_{\text{eff}}/2k=2.47). The π\pi mode trajectories circle the stationary point z0=0.91,ϕ0=±π,ξ0=1.19z_{0}=0.91,\,\phi_{0}=\pm\pi,\,\xi_{0}=1.19 shown in top panels: (a) with the initial conditions, δz(0)=0.15,δξ(0)=0.20\delta z(0)=0.15,\,\delta\xi(0)=0.20 [blue, (gray)], δz(0)=0.10,δξ(0)=0.13\delta z(0)=0.10,\,\delta\xi(0)=0.13 [red (dark gray)], and δz(0)=0.05,δξ(0)=0.065\delta z(0)=0.05,\,\delta\xi(0)=0.065 [green (light gray)] where the ωξ\omega_{\xi} mode is effectively switched: (b) with the initial conditions δz(0)=0.15,δξ(0)=0\delta z(0)=0.15,\delta\xi(0)=0 [blue (gray)], δz(0)=0.10,δξ(0)=0\delta z(0)=0.10,\,\delta\xi(0)=0 [red (dark gray)], and δz(0)=0.05,δξ(0)=0\delta z(0)=0.05,\,\delta\xi(0)=0 [green (light gray)] where the the ωξ\omega_{\xi} mode is effectively activated. Other initial conditions for both cases are δϕ(0)=0\delta\phi(0)=0, δξ˙(0)=0\delta\dot{\xi}(0)=0. In bottom panels, the plots (c) and (d) are drawn for the corresponding effective potential for (a), and (b) respectively.
Refer to caption
Figure 6: The evolution of the population imbalance zz (a) and the displacement of the center condensate (b) as a function of time in the case of the MQST are compared between the results of the numerical solutions of the time-dependent GP equations (blue line), and the linearized equations in Eqs. (24), (25), and (26) (red dashed line) with the parameters a11=87a0,a22=50a0,a12=85a0a_{11}=87a_{0},\,a_{22}=50a_{0},\,a_{12}=85a_{0} (Λ=2.0,k=0.007,η=1.4\Lambda=2.0,\,k=0.007,\,\eta=1.4) and the initial conditions are z(0)=0.1,ϕ(0)=π,ξ(0)=ξ˙(0)=0z(0)=0.1,\,\phi(0)=\pi,\,\xi(0)=\dot{\xi}(0)=0.

In Fig. 5, the scattering lengths are slightly changed to probe the dynamics of z,ϕz,\phi around the stationary state ϕ0=π,z0=0.91(0)\phi_{0}=\pi,z_{0}=0.91\,(\neq 0) obtained from (33) with the parameters Λ=1.63\Lambda=1.63, η=1.267\eta=1.267, k=0.005k=0.005, giving Λeff/2k=2.5>2\Lambda_{\text{eff}}/2k=2.5>2. We have the π\pi-mode MQST oscillations in this case. The initial conditions of δz(0)\delta z(0) are chosen being δz(0)1\delta z(0)\ll 1 with small perturbations around nonzero z0z_{0}, where the effective potential constructed in (64) can safely be applied. The plots of Figs. 5(c) and 5(d) of the respective effective potentials illustrate the single-well profile with the effects of turning on or off the ωξ\omega_{\xi} mode can satisfactorily interpret the dynamics of the population imbalance zz shown in Fig. 5(a) and 5(b). This indicates that the MQST mode will persist for all range of z(0)z(0) with 1<z(0)<1-1<z(0)<1. The effects from nonzero CzC_{z} and CξC_{\xi} seem changing slightly the turning points of z¯\bar{z}, as z¯\bar{z} oscillates around z0z_{0}. For this Λeff\Lambda_{\text{eff}} with ϕ(0)=π\phi(0)=\pi, z¯c(0)=0\bar{z}_{c}(0)=0 is found. This means that for 1<z(0)<1-1<z(0)<1, the MQST mode sets in. In general, we can find z¯c(0)\bar{z}_{c}(0) as a function of ϕ(0)\phi(0) using the effective potential obtained above to be shown in Fig. 5. Similar check is done using the time-dependent GP equation as in Fig. 4 for the case of MQST whereas in Fig. 6, large discrepancy is found due to the fact that the two modes approach in (2) may not provide a relatively good basis to parametrize the spatial parts of the wave functions of bilateral condensates.

Refer to caption
Figure 7: In top panels, the phase portrait with the parameters of Λ=1.63\Lambda=1.63, η=1.27\eta=1.27, η=0.005(Λeff/2k=1.07)\eta=0.005\,(\Lambda_{\text{eff}}/2k=1.07) is plotted for the Josephson oscillation trajectories around the stationary state: z0=0ϕ0=0,ξ0=0z_{0}=0\,\phi_{0}=0,\,\xi_{0}=0 (a) with the initial conditions, δz(0)=0.3,δξ(0)=0.38\delta z(0)=0.3,\,\delta\xi(0)=0.38 [blue (gray)], δz(0)=0.2,δξ(0)=0.25\delta z(0)=0.2,\,\delta\xi(0)=0.25 [red (dark gray)], and δz(0)=0.10,δξ(0)=0.12\delta z(0)=0.10,\,\delta\xi(0)=0.12 [green (light gray)] where the ωξ\omega_{\xi} mode is effectively switched: (b) with the initial conditions δz(0)=0.30,δξ(0)=0\delta z(0)=0.30,\delta\xi(0)=0 [blue (gray)], δz(0)=0.20,δξ(0)=0\delta z(0)=0.20,\,\delta\xi(0)=0 [red (dark gray)], and δz(0)=0.10,δξ(0)=0\delta z(0)=0.10,\,\delta\xi(0)=0 [green (light gray)] where the ωξ\omega_{\xi} mode is effectively activated. Other initial conditions for both cases are δϕ(0)=0\delta\phi(0)=0, δξ˙(0)=0\delta\dot{\xi}(0)=0. In bottom panels, the plots (c) and (d) are drawn for the corresponding effective potential for (a), and (b) respectively.
Refer to caption
Figure 8: In top panels, the phase portrait with parameters as in Fig.7 is plotted for the π\pi mode trajectories around the stationary state: z0=0.37,ϕ0=π,ξ0=0.47z_{0}=0.37,\,\phi_{0}=\pi,\,\xi_{0}=0.47 (a) with the initial conditions, δz(0)=0.15,δξ(0)=0.1905\delta z(0)=0.15,\,\delta\xi(0)=0.1905 [blue (gray)], δz(0)=0.100,δξ(0)=0.127\delta z(0)=0.100,\,\delta\xi(0)=0.127 [red (dark gray)], and δz(0)=0.050,δξ(0)=0.064\delta z(0)=0.050,\,\delta\xi(0)=0.064 [green (light gray)] where the ωξ\omega_{\xi} mode is effectively switched: (b) with the initial conditions δz(0)=0.150,δξ(0)=0\delta z(0)=0.150,\delta\xi(0)=0 [blue (gray)], δz(0)=0.100,δξ(0)=0\delta z(0)=0.100,\,\delta\xi(0)=0 [red (dark gray)], and δz(0)=0.050,δξ(0)=0\delta z(0)=0.050,\,\delta\xi(0)=0 [green (light gray)] where the the ωξ\omega_{\xi} mode is effectively activated. Other initial conditions for both cases are δϕ(0)=0\delta\phi(0)=0, δξ˙(0)=0\delta\dot{\xi}(0)=0. In bottom panels, the plots (c) and (d) are drawn for the corresponding effective potential for (a), and (b) respectively.

IV.3.2 1<Λeff/2k<21<\Lambda_{\textsf{eff}}/2k<2 (Josephson oscillation and π\pi-mode self trapping)

In the interval 1<Λeff/2k<21<\Lambda_{\text{eff}}/2k<2, there exist two solutions of z¯c2(0)\bar{z}^{2}_{c}(0), if π/2|ϕ(0)|π\pi/2\leq|{\phi(0)}|\leq\pi,

z¯c2(0)=\displaystyle\bar{z}^{2}_{c}(0)= 4kΛeff2[Λeff2kcosϕ(0)\displaystyle\frac{4k}{\Lambda_{\text{eff}}^{2}}\bigg{[}\Lambda_{\text{eff}}-2k\cos\phi(0)
±Λeff(Λeff4k)cos2ϕ(0)+4k2cos4ϕ(0)].\displaystyle\pm\sqrt{\Lambda_{\text{eff}}(\Lambda_{\text{eff}}-4k)\cos^{2}\phi(0)+4k^{2}\cos^{4}\phi(0)}\bigg{]}. (73)

In Figs. 7 and 8, the parameter Λeff/2k\Lambda_{\text{eff}}/2k is tuned to the value between 1<Λeff/2k<21<\Lambda_{\text{eff}}/2k<2. On the one hand, the Josephson oscillations are seen for ϕ(0)=0\phi(0)=0, and however, the running phase MQST does not exist in this case in Fig. 7 where there is no such a z¯c(0)\bar{z}_{c}(0) found with the single-well potential, shown in (IV.3.1) and (IV.3.2). On the other hand, the transition between the Josephson oscillations and the MQST can occur in ϕ(0)=π\phi(0)=\pi with the double-well potential, giving the critical value of z¯c(0)\bar{z}_{c}(0) either by (IV.3.2) for the positive z¯c(0)\bar{z}_{c}(0) with the ωJξ\omega_{J\xi} mode only, or by (IV.3.1) involving the contributions from the ωξ\omega_{\xi} mode.

IV.3.3 0<Λeff/2k<10<\Lambda_{\textsf{eff}}/2k<1 (Josephson oscillation)

As for 0<Λeff/2k<10<\Lambda_{\text{eff}}/2k<1, since there does not exist the solution of z¯c(0)\bar{z}_{c}(0) due to the fact that the corresponding effective potential for both ϕ(0)=0\phi(0)=0 and ϕ(0)=π\phi(0)=\pi shows a single well like the case in Fig. 7, we find the Josephson oscillations for both relative phase difference cases.

To summarize the discussion given above, we show the phase portraits in Fig. 9 from solving the double pendulum equations (24)–(26). The initial conditions we choose in the plots of the top row are again z(0)=z0+δz(0),ϕ(0)=ϕ0z(0)=z_{0}+\delta z(0),\phi(0)=\phi_{0} and ξ(0)=ξ0+δξ(0),ξ˙(0)=0\xi(0)=\xi_{0}+\delta\xi(0),\,\dot{\xi}(0)=0, where δξ(0)=ωξ2ηδz(0)\delta\xi(0)=\omega_{\xi}^{-2}\eta\delta z(0) so that the mode of ωξ\omega_{\xi} is effectively switched off. The bottom row figures correspond to the initial condition of δξ(0)=0\delta\xi(0)=0 instead with the solutions that the mode of ωξ\omega_{\xi} is effectively switched on for comparison. For top row results, in Fig. 9(e), there exists a transition between the Josephson oscillations and the running phase MQST at ϕ(0)=0\phi(0)=0 for Λeff/2k>2\Lambda_{\text{eff}}/2k>2, and in Fig. 9(d), the transition occurs between the MQST and Josephson oscillations at ϕ(0)=±π\phi(0)=\pm\pi instead apart from the Josephson oscillations at ϕ(0)=0\phi(0)=0 for 1<Λeff/2k<21<\Lambda_{\text{eff}}/2k<2, and further in Fig. 9(c) the Josephson oscillations at ϕ(0)=0,±π\phi(0)=0,\pm\pi respectively occur for 0<Λeff/2k<10<\Lambda_{\text{eff}}/2k<1. For Λeff/2k<0\Lambda_{\text{eff}}/2k<0, with Cξ=Cz=0C_{\xi}=C_{z}=0, all results for Λeff/2k>0\Lambda_{\text{eff}}/2k>0 are shifted from ϕϕ+π\phi\rightarrow-\phi+\pi. Nevertheless, including the effects from the ωξ\omega_{\xi} mode, such symmetry mentioned above does not exist.

Refer to caption
Figure 9: The phase portraits, choosing different η\eta with fixed parameters of Λ=1.35,k=0.005\Lambda=1.35,\,k=0.005 and ωξ=1\omega_{\xi}=1 by solving (24), (25), and (26). The top row figures correspond to the initial conditions: z(0)=z0+δz(0),ϕ(0)=ϕ0z(0)=z_{0}+\delta z(0),\phi(0)=\phi_{0}, and ξ(0)=ξ0+δξ(0),ξ˙(0)=0\xi(0)=\xi_{0}+\delta\xi(0),\dot{\xi}(0)=0, where δξ(0)=ηωξ2δz(0)\delta\xi(0)=\eta\omega_{\xi}^{2}\delta z(0) away from the various stationary states by changing δz(0)\delta z(0) with the solutions that the mode of ωξ\omega_{\xi} is effectively switched off. The bottom row figures correspond to the initial condition of δξ(0)=0\delta\xi(0)=0 in stead with the solutions that the mode of ωξ\omega_{\xi} is effectively switched on for comparison. The parameters we choose are as follows : (a) Λeff/2k=2.47\Lambda_{\text{eff}}/2k=-2.47, η=1.172\eta=1.172; (b) Λeff/2k=1.55\Lambda_{\text{eff}}/2k=-1.55, η=1.168\eta=1.168; (c) Λeff/2k=0\Lambda_{\text{eff}}/2k=0, η=1.162\eta=1.162; (d) Λeff/2k=1.55\Lambda_{\text{eff}}/2k=1.55, η=1.155\eta=1.155; (e) Λeff/2k=2.47\Lambda_{\text{eff}}/2k=2.47, η=1.151\eta=1.151.

V Chaotic dynamics

In previous sections, we mainly focus on various regimes of the regular motions. In particular, when two frequencies ω+\omega_{+} and ω\omega_{-} have large difference in magnitude, the dynamics of the rapidly varying mode is averaged out giving a mean effect to the mode of slow oscillations. However, as the difference in magnitude between two frequencies becomes small with a relatively larger overlap between bilateral condensates, the system exhibits chaotic phenomena with the scattering lengths in the regimes marked in Fig. 10.

Refer to caption
Figure 10: With same parameters of Fig. 1, the (blue and deep blue) regimes of the scattering lengths lead to the chaotic behavior of the system for Λeff/2k>1\Lambda_{\text{eff}}/2k>1 with the initial conditions near the hyperbolic fixed point z0=ξ0=0z_{0}=\xi_{0}=0, ϕ0=±π\phi_{0}=\pm\pi seen in Fig. 9(d) (1<Λeff/2k<2)1<\Lambda_{\text{eff}}/2k<2) and Fig. 9(e) (Λeff/2k>2\Lambda_{\text{eff}}/2k>2), whereas the (red and pink) regimes for Λeff/2k<1\Lambda_{\text{eff}}/2k<-1 with the initial conditions near the state z0=ξ0=0z_{0}=\xi_{0}=0, ϕ0=0\phi_{0}=0 in stead, which is also a hyperbolic fixed point in this range of Λeff/2k\Lambda_{\text{eff}}/2k in Fig. 9(a) (Λeff/2k<2\Lambda_{\text{eff}}/2k<-2) and in Fig. 9(b) (1<Λeff/2k<2-1<\Lambda_{\text{eff}}/2k<-2).

From experimental perspectives, this can be achieved by increasing the scattering length a11a_{11} of the bilateral condensates or decreasing the inter-species scattering length a12a_{12}, thus resulting in relatively large tunneling energy kk. In Fig. 11, we consider the scattering lengths giving Λeff/2k>2\Lambda_{\text{eff}}/2k>2 shown in Fig. 10 for the system to illustrate the chaotic behavior. The initial conditions are chosen near the solutions z0=ξ0=0z_{0}=\xi_{0}=0, ϕ0=±π\phi_{0}=\pm\pi, a hyperbolic fixed point, which is a unstable point by changing zz, realized from the effective potential in Fig. 3 and also in Table 2, but is a stable point by changing the phase ϕ\phi instead. We solve the time-dependent GP equations using the initial condition ψ2(x,0)\psi_{2}(x,0) with ξ(0)=0\xi(0)=0 and ξ˙(0)=0\dot{\xi}(0)=0. The bilateral condensate wave function ψ1(x,0)\psi_{1}(x,0) is constructed by spatial wave functions ψL\psi_{L} and ψR\psi_{R} with the initial population imbalance z(0)=0.09z(0)=0.09 and the initial relative phase difference ϕ(0)=π\phi(0)=\pi. The dynamics of zz is shown to run between the Josephson oscillations and the running phase MQST in Fig. 11(b). With the same initial conditions for z(0),ϕ(0),ξ(0)z(0),\phi(0),\xi(0) and ξ˙(0)\dot{\xi}(0), the numerical results of coupled pendulum equations (24), (25), and (26) also show the similar dynamics from which solutions we can compute the so-called Lyapunov exponents for each of the dynamic variables. A chaotic motion can be understood from the fact that, for two initial nearby initial variables q(II)(0)q(I)(0)q^{(II)}(0)\rightarrow q^{(I)}(0), their difference grows exponentially in time tt, obeying

|q(II)(t)q(I)(t)|=|q(II)(0)q(I)(0)|eλt,\displaystyle|q^{(II)}(t)-q^{(I)}(t)|=|q^{(II)}(0)-q^{(I)}(0)|e^{\lambda t},

where the rate λ\lambda is the Lyapunov exponent obtained from

λ=limtlimq(II)(0)q(I)(0)1tln(|q(II)(t)q(I)(t)q(II)(0)q(I)(0)|).\displaystyle\lambda=\lim_{t\rightarrow\infty}\;\lim_{{q}^{(II)}(0)\rightarrow q^{(I)}(0)}\frac{1}{t}\ln{\left|\frac{q^{(II)}(t)-q^{(I)}(t)}{q^{(II)}(0)-q^{(I)}(0)}\right|}. (74)

There are four Lyapunov exponents λz,λϕ,λξ\lambda_{z},\lambda_{\phi},\lambda_{\xi}, and λξ˙\lambda_{\dot{\xi}} shown in Fig. 11(c). The chaotic dynamics can be recognized when at least one of them is a nonzero positive value at large time, and it is found λz>0\lambda_{z}>0 in our choice of the initial conditions and scattering lengths.

Refer to caption
Figure 11: (a) The phase portrait for the parameters: N1=500,N2=1000,a11=137a0,a22=50a0N_{1}=500,\,N_{2}=1000,\,a_{11}=137a_{0},\,a_{22}=50a_{0}, and a12=85a0a_{12}=85a_{0} that result the effective parameters Λ=7.1,η=2.45\Lambda=7.1,\,\eta=2.45, and k=0.0345k=0.0345, (b) shows chaotic oscillation with initial conditions z(0)=0.09,ϕ(0)=π,ξ(0)=ξ˙(0)=0z(0)=0.09,\phi(0)=\pi,\,\xi(0)=\dot{\xi}(0)=0. The red dashed line corresponds to the linearized equation Eqs. (24), (25), and (26) and blue line corresponds to the simulation results of real-time GP equations. (c) The corresponding Lyapunov exponents obtained from (74) are λz\lambda_{z}, λξ˙\lambda_{\dot{\xi}}, λξ\lambda_{\xi}, λϕ\lambda_{\phi} from top to bottom.

In order to analytically understand the chaotic dynamic, we start from the CP dynamics with Eqs. (24)–(26), and substitute the solution of ξ(t)\xi(t) in (56) and (58) into (27) where again the full dynamics of the system reduces to that of (z,ϕ)(z,\phi) only. Then the corresponding Hamiltonian splits into two parts

H=Hz+HI(t),\displaystyle H=H_{z}+H_{I}(t), (75)

where HzH_{z} is the Hamiltonian to account for the dynamics of the slowly varying mode given effectively by

Hz=Λeff2z22k1z2cos(ϕ)\displaystyle H_{z}=\frac{\Lambda_{\text{eff}}}{2}z^{2}-2k\sqrt{1-z^{2}}\cos{\phi} (76)

In the case of relatively small difference between the values of ω\omega_{-} and ω+\omega_{+}, the dynamics of the fast varying mode ω+\omega_{+} can be treated as the time dependent perturbation contributing to the interaction Hamiltonian as

HI(t)ΔE¯(t)z(t)=η2ωξ2z(0)cos((ω+t))z(t).\displaystyle H_{I}(t)\equiv\Delta\bar{E}(t)z(t)=\eta^{2}\omega_{\xi}^{-2}z(0)\cos{(\omega_{+}t)}z(t). (77)

Nevertheless in Refs.  Abdullaev2000 ; Lee2001 ; Jiang2014 ; Tomkovic2017 , this time-dependent perturbation is implemented by introducing an external driving force. Here, assuming HIHzH_{I}\ll H_{z}, one can write the solution of zz as z=ζ+ζ~z=\zeta+\tilde{\zeta} where ζ\zeta is the solution of the unperturbed equation

ζ¨(ΛeffHz4k2)ζ+Λeff22ζ3=0,\displaystyle\ddot{\zeta}-(\Lambda_{\text{eff}}H_{z}-4k^{2})\zeta+\frac{\Lambda_{\text{eff}}^{2}}{2}\zeta^{3}=0, (78)

and ζ~\tilde{\zeta} is the first-order correction in zz due to the perturbation g(t)g(t) satisfying the equation

ζ~¨(ΛeffHz4k2)ζ~+32Λeffζ2ζ~=g(t),\displaystyle\ddot{\tilde{\zeta}}-(\Lambda_{\text{eff}}H_{z}-4k^{2})\tilde{\zeta}+\frac{3}{2}\Lambda_{\text{eff}}{\zeta}^{2}\tilde{\zeta}=g(t), (79)
g(t)=ΔE¯(t)Hz+HIΛeffζ32ΛeffΔE¯(t)ζ2.\displaystyle g(t)=\Delta\bar{E}(t)\,H_{z}+H_{I}\,\Lambda_{\text{eff}}\,\zeta-\frac{3}{2}\Lambda_{\text{eff}}\,\Delta\bar{E}(t)\,\zeta^{2}. (80)
Refer to caption
Figure 12: (a) The phase portrait for the parameters: N1=500,N2=1000,a11=77a0,a22=50a0N_{1}=500,\,N_{2}=1000,\,a_{11}=77a_{0},\,a_{22}=50a_{0}, and a12=77a0a_{12}=77a_{0} that result the effective parameters Λ=4.8,η=2.18\Lambda=4.8,\,\eta=2.18, and k=0.019k=0.019, (b) shows chaotic oscillation with initial conditions z(0)0,ϕ(0)π,ξ(0)=ξ˙(0)=0z(0)\simeq 0,\phi(0)\simeq\pi,\,\xi(0)=\dot{\xi}(0)=0. The red dashed line corresponds to the linearized equation Eqs. (24), (25), and (26) and blue line corresponds to the simulation results of real-time GP equations. (c) The corresponding Lyapunov exponents obtained from (74) are λz\lambda_{z}, λξ˙\lambda_{\dot{\xi}}, λξ\lambda_{\xi}, λϕ\lambda_{\phi} from top to bottom.

When ΛeffHz4k2>0\Lambda_{\text{eff}}H_{z}-4k^{2}>0, Eq. (78) presents a homoclinic solution with the double-well potential shown in Fig. 3 for Λeff/2k>2\Lambda_{\text{eff}}/2k>2 and in Fig. 8 for 1<Λeff/2k<21<\Lambda_{\text{eff}}/2k<2 when Hz=2kH_{z}=2k. The homoclinic solution might start from a particular initial condition, and the system will drive zz rolling down toward one of the potential minima, and then climbing up the potential hill of the state with z=0z=0, a fixed point, just with zero velocity. The solution ζ\zeta reads as

ζ(t)=2(Λeff2k)/Λeff2sech(Λeff2kt+𝒟),\displaystyle\zeta(t)=2\sqrt{(\Lambda_{\text{eff}}-2k)/\Lambda_{\text{eff}}^{2}}\,\sech(\sqrt{\Lambda_{\text{eff}}-2k}\,t+\mathcal{D}), (81)
𝒟=arcsech(z(0)2(Λeff2k)/Λeff2).\displaystyle\mathcal{D}=\text{arcsech}{\left(\frac{{z}(0)}{2\sqrt{(\Lambda_{\text{eff}}-2k)/\Lambda_{\text{eff}}^{2}}}\right)}. (82)

This homoclinic solution is also a separatrix, which is a boundary between the Josephson oscillations and the running phase MQST for Λeff/2k>2\Lambda_{\text{eff}}/2k>2, and also the Josephson oscillations and the π\pi phase MQST for 1<Λeff/2k<21<\Lambda_{\text{eff}}/2k<2. One can then construct the Melnikov function from the homogeneous solution of (79), which is ζ~=ζ˙\tilde{\zeta}=\dot{\zeta}, and the function g(t)g(t) as

M(0)=𝑑tζ˙(t)g(t).\displaystyle M(0)=\int_{-\infty}^{\infty}{dt\dot{{\zeta}}}(t)g(t)\,. (83)

The existence of zero of the Melnikov function shows the chaos Abdullaev2000 ; Lee2001 . Substituting Eq. (79) into (83) above turns out to be

M(0)=\displaystyle M(0)= 2πω+ΛeffΛeff2kΔE¯(0)\displaystyle-\frac{2\pi\omega_{+}}{\Lambda_{\text{eff}}}\sqrt{\Lambda_{\text{eff}}-2k}\Delta\bar{E}(0)
×[HzΛeff2k13Λeff2k(1+ν2)]\displaystyle\times\left[\frac{H_{z}}{\sqrt{\Lambda_{\text{eff}}-2k}}-\frac{1}{3}\sqrt{\Lambda_{\text{eff}}-2k}\,\left(1+\nu^{2}\right)\right]
×sech((πν2))sin((𝒟ν)),\displaystyle\times\sech{\left(\frac{\pi\nu}{2}\right)}\sin{(\mathcal{D}\nu)}\,, (84)

where ν=ω+/Λeff2k\nu=\omega_{+}/\sqrt{\Lambda_{\text{eff}}-2k}. For the case of the fixed point z0=0z_{0}=0, ϕ0=π\phi_{0}=\pi, the value of 𝒟\mathcal{D}\rightarrow\infty means that with such highly rapidly oscillations of the sine function, any finite ω+\omega_{+} including the frequency given by the fast varying mode will make the Melnikov function zero. Thus, the chaos occurs as shown in Fig. 11 for Λeff/2k>2\Lambda_{\text{eff}}/2k>2. Similar chaos occurs when 1<Λeff/2k<21<\Lambda_{\text{eff}}/2k<2, and is shown in Fig. 12. In this case, the dynamics of zz runs between the Josephson oscillations and the π\pi mode MQST instead, where its behavior can also be analyzed using the Melnikov Homoclinic method discussed above. Moreover, the chaos may appear also in Λeff/2k<1\Lambda_{\text{eff}}/2k<-1 near the hyperbolic fixed point z0=0,ϕ0=0z_{0}=0,\phi_{0}=0 (see Table 2) in the regime of the scattering lengths shown in Fig. 10. Notice that although the semiclassical (mean-field) GP equation can reliably signal the onset of the chaotic motion, it might fail to provide the details of the dynamics on the short timescales right after entering the chaotic regime, which can be dominated by quantum many-body effects. The exploration of chaos beyond the mean field approximation deserves further studies  ber ; Kelly2019 .

VI Conclusions

In summary, we have proposed a new setting with binary BECs in a single-well trap potential to probe the dynamics of collective atomic motion. In this setting, Rb85{}^{85}\text{Rb} atoms |2,2|2,-2\rangle and Rb87{}^{87}\text{Rb} atoms |1,1|1,-1\rangle are considered with tunable scattering lengths via Feshbach resonances so that the ground-state wave function for two types of the condensates are spatially immiscible shown in Fig.1. As such, the condensate of atoms for one of the hyperfine states centered at the potential minimum can be effectively treated as a potential barrier between bilateral condensates formed by atoms in the other hyperfine state. In the case of small wave-function overlap of bilateral condensates, one can parametrize their spatial part of the wave functions in the two-mode approximation together with time-dependent population of atoms and the phase of each of the wave functions. Besides, the wave function of the condensate in the middle is approximated by an ansatz of the Gaussian wave function. The full system can be reduced to the dynamics of the imbalance population of atoms in bilateral condensates zz, as well as the relative phase difference ϕ\phi between two wave functions together with the time-dependent displacement of the central condensate ξ\xi. For small wavefunction overlap of bilateral condensates shown in Fig. 10, all sorts of the regular trajectories, moving about the stable states, in Refs. Smerzi1997 ; Raghavan1999 can be reproduced. Moreover, the numerical results given by solving the equations of z,ϕz,\,\phi, and ξ\xi are in close agreement with the solutions of the full time-dependent GP equations. Nevertheless, with an increase in wave-function overlap also shown in Fig. 10, we study the possibility of the appearance of the chaotic oscillations driven by the time-dependent displacement of the central condensate. The application of the Melnikov approach with the homoclinic solutions of the z,ϕz,\,\phi, and ξ\xi equations successfully predicts the existence of the chaos, which are further justified from solving the full time-dependent GP equations. All of the findings in this work deserve further experimental investigations using advanced techniques for manipulation of atomic condensates.

Acknowledgements.
This work was supported in part by the Ministry of Science and Technology, Taiwan.

VII Appendix

Section II has discussed a variational approach to the dynamics of binary BEC system. In this appendix we provide more detailed derivations and approximations to arrive at the equations of the CP dynamics given in (24)–(26). Substituting the ansatz of the ground-state wave function (5) and (6) into the Lagrangian (1), and carrying out the integration over space, the effective Lagrangian then becomes a functional of the time-dependent variables α\alpha, β\beta, ξ\xi, ww, zz, and ϕ\phi as

L1D=\displaystyle L_{\text{1D}}= L0N12(zϕ˙+ΔEz+Λ2z22k01z2cosϕ)\displaystyle L_{0}-\frac{N_{1}}{2}\left(-z\dot{\phi}+\Delta Ez+\frac{\Lambda}{2}z^{2}-2k_{0}\sqrt{1-z^{2}}\cos\phi\right)
N2{(α˙ξ+β˙ξ2+β˙2w2)+12[12w2+2β2w2+(α+2βξ)2]+14(2ξ2+w2)+g22N222πw}\displaystyle-N_{2}\Bigg{\{}(\dot{\alpha}\xi+\dot{\beta}\xi^{2}+\frac{\dot{\beta}}{2}w^{2})+\frac{1}{2}\left[\frac{1}{2w^{2}}+2\beta^{2}w^{2}+(\alpha+2\beta\xi)^{2}\right]+\frac{1}{4}(2\xi^{2}+w^{2})+\frac{g_{22}N_{2}}{2\sqrt{2\pi}w}\Bigg{\}}
g12N1N22πw𝑑x[(ψL2+ψR2)+z(ψL2ψR2)+21z2cosϕψLψR]exp[(xξ)2w2],\displaystyle-\frac{g_{12}N_{1}N_{2}}{2\sqrt{\pi}w}\int dx\left[(\psi_{L}^{2}+\psi_{R}^{2})+z(\psi_{L}^{2}-\psi_{R}^{2})+2\sqrt{1-z^{2}}\cos\phi\psi_{L}\psi_{R}\right]\exp[-\frac{(x-\xi)^{2}}{w^{2}}]\;, (85)

where the effective parameters are tunneling energy k0k_{0} (15), difference of energies between the wells ΔE\Delta E (18) and self-interaction energy Λ\Lambda (19). Using Lagrangian equations for the parameters α\alpha, β\beta, ξ\xi, ww, zz, and ϕ\phi, we first obtain

α=ξ˙2ξβandβ=w˙2w.\displaystyle\alpha=\dot{\xi}-2\xi\beta\quad\text{and}\quad\beta=\frac{\dot{w}}{2w}. (86)

Then, after inserting Eqs. (86) into (VII), the equations of motion for population imbalance and relative phase between bilateral condensates become

z˙+(2k02g12N2πw𝑑xψLψRe(xξ)2w2)\displaystyle\dot{z}+\left(2k_{0}-\frac{2g_{12}N_{2}}{\sqrt{\pi}w}\int dx\psi_{L}\psi_{R}e^{-\frac{(x-\xi)^{2}}{w^{2}}}\right)
×1z2sinϕ=0,\displaystyle\hskip 113.81102pt\times\sqrt{1-z^{2}}\sin\phi=0, (87)
ϕ˙ΔEΛz\displaystyle\dot{\phi}-\Delta E-\Lambda z
(2k02g12N2πw𝑑xψLψRe(xξ)2w2)z1z2cosϕ\displaystyle-\left(2k_{0}-\frac{2g_{12}N_{2}}{\sqrt{\pi}w}\int dx\psi_{L}\psi_{R}e^{-\frac{(x-\xi)^{2}}{w^{2}}}\right)\frac{z}{\sqrt{1-z^{2}}}\cos\phi
g12N2πw𝑑x(ψL2ψR2)e(xξ)2w2=0.\displaystyle-\frac{g_{12}N_{2}}{\sqrt{\pi}w}\int dx(\psi_{L}^{2}-\psi_{R}^{2})e^{-\frac{(x-\xi)^{2}}{w^{2}}}=0. (88)

For the center condensate, the equations of motion of ξ\xi and ww are obtained as

ξ¨+ξ+g12N1πw3dx[(ψL2+ψR2)+z(ψL2ψR2)\displaystyle\ddot{\xi}+\xi+\frac{g_{12}N_{1}}{\sqrt{\pi}w^{3}}\int dx\left[(\psi_{L}^{2}+\psi_{R}^{2})+z(\psi_{L}^{2}-\psi_{R}^{2})\right.
+21z2cosϕψLψR](xξ)e(xξ)2w2=0,\displaystyle\left.\qquad+2\sqrt{1-z^{2}}\cos\phi\psi_{L}\psi_{R}\right](x-\xi)e^{-\frac{(x-\xi)^{2}}{w^{2}}}=0, (89)
w¨+w1w3g22N22πw2+g12N1πw2dx[(ψL2+ψR2)\displaystyle\ddot{w}+w-\frac{1}{w^{3}}-\frac{g_{22}N_{2}}{\sqrt{2\pi}w^{2}}+\frac{g_{12}N_{1}}{\sqrt{\pi}w^{2}}\int dx\Big{[}(\psi_{L}^{2}+\psi_{R}^{2})
+z(ψL2ψR2)+21z2cosϕψLψR]\displaystyle\quad+z(\psi_{L}^{2}-\psi_{R}^{2})+2\sqrt{1-z^{2}}\cos\phi\psi_{L}\psi_{R}\Big{]}
×[2(xξ)2w21]e(xξ)2w2=0.\displaystyle\qquad\times\left[\frac{2(x-\xi)^{2}}{w^{2}}-1\right]e^{-\frac{(x-\xi)^{2}}{w^{2}}}=0. (90)

It can be understood that the presence of bilateral condensates contribute a time-dependent deformation for the central condensate by coupling to population imbalance z(t)z(t) in the integrand of Eqs. (89) and (90). We introduce Gaussian functions (see Fig. 2)

ψL(x)=(1πλ2)1/4e(x+ζ)22λ2,\displaystyle\psi_{L}(x)=\left(\frac{1}{\pi\lambda^{2}}\right)^{1/4}e^{-\frac{(x+\zeta)^{2}}{2\lambda^{2}}}, (91)
ψR(x)=(1πλ2)1/4e(xζ)22λ2,\displaystyle\psi_{R}(x)=\left(\frac{1}{\pi\lambda^{2}}\right)^{1/4}e^{-\frac{(x-\zeta)^{2}}{2\lambda^{2}}}, (92)

and substitute them into Eqs. (87)–(90). In our cases, the displacement ξ\xi and the width defined as w=σ0+σw=\sigma_{0}+\sigma with σ0\sigma_{0} determined initially by solving time-independent GP equation for finding the ground state solution and σ\sigma driven by ξ\xi , satisfy the conditions

ξλ2+σ02andσλ2+σ02.\displaystyle\xi\ll\sqrt{\lambda^{2}+\sigma_{0}^{2}}\quad\text{and}\quad\sigma\ll\sqrt{\lambda^{2}+\sigma_{0}^{2}}\,. (93)

In the case of λσ0\lambda\sim\sigma_{0}, we have ξλ2+σ02σ0\xi\ll\sqrt{\lambda^{2}+\sigma_{0}^{2}}\sim\sigma_{0} and σλ2+σ02σ0\sigma\ll\sqrt{\lambda^{2}+\sigma_{0}^{2}}\sim\sigma_{0}. Back to Eqs. (87)–(90), it is allowed to expand the equations in terms of small σ/σ0\sigma/\sigma_{0} and ξ/σ0\xi/\sigma_{0} as

z˙+[2k02g12N2πσ0(1σσ0+𝒪(σ2σ02))𝑑xψLψRe(xξ)2σ02(12σ/σ0)]1z2sinϕ=0,\displaystyle\dot{z}+\left[2k_{0}-\frac{2g_{12}N_{2}}{\sqrt{\pi}\sigma_{0}}\left(1-\frac{\sigma}{\sigma_{0}}+\mathcal{O}\left(\frac{\sigma^{2}}{\sigma_{0}^{2}}\right)\right)\int dx\psi_{L}\psi_{R}e^{-\frac{(x-\xi)^{2}}{\sigma_{0}^{2}}(1-2\sigma/\sigma_{0})}\right]\sqrt{1-z^{2}}\sin\phi=0\;, (94)

where the integral above can be further expanded as

[1σσ0+𝒪(σ2σ02)]𝑑xψLψRe[(xσ0)2+2(xξσ02)+𝒪(ξ2σ02)](12σσ0+𝒪(σ2σ02))\displaystyle\left[1-\frac{\sigma}{\sigma_{0}}+\mathcal{O}\left(\frac{\sigma^{2}}{\sigma_{0}^{2}}\right)\right]\int dx\psi_{L}\psi_{R}e^{{\left[-\left(\frac{x}{\sigma_{0}}\right)^{2}+2\left(\frac{x\xi}{\sigma_{0}^{2}}\right)+\mathcal{O}\left(\frac{\xi^{2}}{\sigma_{0}^{2}}\right)\right]\left(1-2\frac{\sigma}{\sigma_{0}}+\mathcal{O}\left(\frac{\sigma^{2}}{\sigma_{0}^{2}}\right)\right)}}
=𝑑xψLψRex2/σ02[1σσ0(12x2σ02)]+𝑑xψLψRex2/σ02(2xξσ02)+𝒪(ξ2σ02,σ2σ02).\displaystyle=\int\,dx\psi_{L}\psi_{R}e^{-{x^{2}}/{\sigma_{0}^{2}}}\left[1-\frac{\sigma}{\sigma_{0}}\left(1-2\frac{x^{2}}{\sigma_{0}^{2}}\right)\right]+\int dx\psi_{L}\psi_{R}e^{-{x^{2}}/{\sigma_{0}^{2}}}\left(\frac{2x\xi}{\sigma_{0}^{2}}\right)+\mathcal{O}\left(\frac{\xi^{2}}{\sigma_{0}^{2}},\;\frac{\sigma^{2}}{\sigma_{0}^{2}}\right)\,. (95)

The term of order ξ/σ0\xi/\sigma_{0} vanishes due to the odd function in the integrand. Therefore Eq. (87) can be further simplified by keeping the terms of order σ/σ0\sigma/\sigma_{0} as

z˙+[2k02g12N2πσ0𝑑xψLψRex2/σ02+2g12N2πσ0𝑑xψLψRex2/σ02(12x2σ02)(σσ0)]1z2sinϕ=0,\displaystyle\dot{z}+\left[2k_{0}-\frac{2g_{12}N_{2}}{\sqrt{\pi}\sigma_{0}}\int{dx\psi_{L}\psi_{R}e^{-x^{2}/\sigma_{0}^{2}}}+\frac{2g_{12}N_{2}}{\sqrt{\pi}\sigma_{0}}\int dx\psi_{L}\psi_{R}e^{-x^{2}/\sigma_{0}^{2}}\left(1-2\frac{x^{2}}{\sigma_{0}^{2}}\right)\left(\frac{\sigma}{\sigma_{0}}\right)\right]\sqrt{1-z^{2}}\sin\phi=0, (96)

and reduces to (11) with the definition of the constants in (16) and (17). Following the same procedure to approximate Eq. (88), we have

ϕ˙ΔEΛz[2k02g12N2πσ0𝑑xψLψRex2/σ02+2g12N2πσ0𝑑xψLψRex2/σ02(12x2σ02)(σσ0)]\displaystyle\dot{\phi}-\Delta E-\Lambda z-\left[2k_{0}-\frac{2g_{12}N_{2}}{\sqrt{\pi}\sigma_{0}}\int{dx\psi_{L}\psi_{R}e^{-x^{2}/\sigma_{0}^{2}}}+\frac{2g_{12}N_{2}}{\sqrt{\pi}\sigma_{0}}\int dx\psi_{L}\psi_{R}e^{-x^{2}/\sigma_{0}^{2}}\left(1-2\frac{x^{2}}{\sigma_{0}^{2}}\right)\left(\frac{\sigma}{\sigma_{0}}\right)\right]
×z1z2cosϕ2g12N2πσ03dx(ψL2ψR2)xex2σ02ξ=0,\displaystyle\hskip 71.13188pt\times\frac{z}{\sqrt{1-z^{2}}}\cos\phi-\frac{2g_{12}N_{2}}{\sqrt{\pi}\sigma_{0}^{3}}\int dx(\psi_{L}^{2}-\psi_{R}^{2})\,x\,e^{-\frac{x^{2}}{\sigma_{0}^{2}}}\xi=0\;, (97)

which leads to (12).

Now we turn to linearize Eq. (89), which is

ξ¨+ξ+g12N1πσ03[13σσ0+𝒪(σ2σ02)]𝑑x[(ψL2+ψR2)+z(ψL2ψR2)+21z2cosϕψLψR]\displaystyle\ddot{\xi}+\xi+\frac{g_{12}N_{1}}{\sqrt{\pi}\sigma_{0}^{3}}\left[1-3\frac{\sigma}{\sigma_{0}}+\mathcal{O}\left(\frac{\sigma^{2}}{\sigma_{0}^{2}}\right)\right]\int dx\left[(\psi_{L}^{2}+\psi_{R}^{2})+z(\psi_{L}^{2}-\psi_{R}^{2})+2\sqrt{1-z^{2}}\cos\phi\psi_{L}\psi_{R}\right]
×(xξ)ex2/σ02[1x2σ02+2(xσ0)(ξσ0)+2(x2σ02)(σσ0)+𝒪(ξ2σ02,σ2σ02)]=0.\displaystyle\qquad\times(x-\xi)e^{-{x^{2}}/{\sigma_{0}^{2}}}\left[1-\frac{x^{2}}{\sigma_{0}^{2}}+2\left(\frac{x}{\sigma_{0}}\right)\left(\frac{\xi}{\sigma_{0}}\right)+2\left(\frac{x^{2}}{\sigma_{0}^{2}}\right)\left(\frac{\sigma}{\sigma_{0}}\right)+\mathcal{O}\left(\frac{\xi^{2}}{\sigma_{0}^{2}},\,\frac{\sigma^{2}}{\sigma_{0}^{2}}\right)\right]=0.

Considering the vanishing of the integral due to the odd function in the integrand, we conclude

ξ¨+[1+g12N1πσ03𝑑x(ψL2+ψR2)(2x2σ021)ex2/σ02]ξ+[g12N1πσ03𝑑x(ψL2ψR2)xex2/σ02]z\displaystyle\ddot{\xi}+\left[1+\frac{g_{12}N_{1}}{\sqrt{\pi}\sigma_{0}^{3}}\int dx(\psi_{L}^{2}+\psi_{R}^{2})\left(2\frac{x^{2}}{\sigma_{0}^{2}}-1\right)e^{-x^{2}/\sigma_{0}^{2}}\right]\xi+\left[\frac{g_{12}N_{1}}{\sqrt{\pi}\sigma_{0}^{3}}\int dx(\psi_{L}^{2}-\psi_{R}^{2})xe^{-x^{2}/\sigma_{0}^{2}}\right]z
+[g12N1πσ03𝑑x(ψL2ψR2)(2x3σ033xσ0)ex2/σ02]σ=0,\displaystyle\hskip 170.71652pt+\left[\frac{g_{12}N_{1}}{\sqrt{\pi}\sigma_{0}^{3}}\int dx(\psi_{L}^{2}-\psi_{R}^{2})\left(2\frac{x^{3}}{\sigma_{0}^{3}}-3\frac{x}{\sigma_{0}}\right)e^{-x^{2}/\sigma_{0}^{2}}\right]\sigma=0, (99)

that gives (13). Finally, we can linearize Eq. (90) as

σ¨+[1+3σ0+g22N22πσ03+g12N1πσ03𝑑x(ψL2+ψR2)(210x2σ02+4x4σ04)ex2/σ02]σ\displaystyle\ddot{\sigma}+\left[1+\frac{3}{\sigma_{0}}+\frac{g_{22}N_{2}}{\sqrt{2\pi}\sigma_{0}^{3}}+\frac{g_{12}N_{1}}{\sqrt{\pi}\sigma_{0}^{3}}\int_{-\infty}^{\infty}dx(\psi_{L}^{2}+\psi_{R}^{2})\left(2-10\frac{x^{2}}{\sigma_{0}^{2}}+4\frac{x^{4}}{\sigma_{0}^{4}}\right)e^{-x^{2}/\sigma_{0}^{2}}\right]\sigma
[2g12N1πσ03𝑑x(ψL2ψR2)(3xσ02x3σ03)ex2/σ02]ξ=0,\displaystyle\hskip 170.71652pt-\left[\frac{2g_{12}N_{1}}{\sqrt{\pi}\sigma_{0}^{3}}\int_{-\infty}^{\infty}dx(\psi_{L}^{2}-\psi_{R}^{2})\left(3\frac{x}{\sigma_{0}}-2\frac{x^{3}}{\sigma_{0}^{3}}\right)e^{-x^{2}/\sigma_{0}^{2}}\right]\xi=0, (100)

giving (14).

References