betaFeedback

WorksPhysics and Astronomy

Polarization instability of unpolarized excited states in trapped two-component Bose–Einstein condensates

Abstract

A binary Bose–Einstein condensate in an elongated trap supports unpolarized excited states, in which both components share the same spatial profile. Within mean-field theory we show that the interspecies repulsion at which these states demix rises with the excitation level. Since each component keeps its atom number, the components separate by moving in opposite directions in the trap, and demixing sets in when the interaction energy released by the separation balances its cost in kinetic and trap energy. We show that the threshold also marks the onset of an exponential, non-oscillating growth of the separation, which divides the cloud into domains of opposite polarization, with walls at the density nodes and further walls interlaced with them. In the Thomas–Fermi regime of a quasi-one-dimensional condensate the rise of the threshold follows power laws in the ratio of the trap oscillator energy to the chemical potential. The trap raises the ground-state threshold, in units of the intraspecies repulsion, by 1.82 times the square of this ratio. Each density node of an excited state near the trap center raises the threshold further by 1.24 times the cube of this ratio. Because the steps between successive levels scale as the cube of this ratio, they fall as the inverse square of the atom number, and the dependence on the excitation level is a mesoscopic effect, appreciable only when the chemical potential is no more than a few times the oscillator energy. In a hyperfine mixture of 87^{87}Rb a Feshbach resonance tunes the interspecies repulsion across the thresholds of the lowest states.

Introduction

Binary Bose–Einstein condensates in elongated traps support stationary excited states in which both components occupy the same spatial profile; the simplest of them is the dark–dark soliton [1, 2, 3]. Since the two components have equal densities at every point, these states are unpolarized, and their components separate once the interspecies repulsion is strong enough. Dark solitons and their trains have been generated in elongated condensates [4, 5], and in a hyperfine mixture of 87Rb^{87}\mathrm{Rb} a Feshbach resonance has tuned the interspecies repulsion across the demixing transition [6, 7]. For a uniform mixture, mean-field theory places the separation of the mixed ground state where the interspecies repulsion exceeds the geometric mean of the intraspecies repulsions, g12>g11g22g_{12} > \sqrt{g_{11}g_{22}} [8, 9, 10, 11]. At what interspecies repulsion, then, do the unpolarized excited states of a trapped condensate demix, how does this threshold depend on the number of density nodes, and does the separation begin to grow in time at the same threshold?

Excited states of binary condensates have been studied mainly as solitary waves in which the two components occupy distinct modes, such as dark–bright solitons and their bound states [4, 12, 13, 14, 15, 16, 17], beating dark–dark solitons [2, 18], trapped dark–dark states with different node numbers in the two components [19, 20], dark–antidark structures [21, 22], dark solitons in one component of a mixed or phase-separated binary condensate [23, 24], and polarized dark solitons and domain walls in nonlinear optics and in binary condensates [25, 26, 27]. In all of these states the two components have different spatial profiles, and the states are polarized from the outset. The dark–dark soliton with coincident notches, whose components share one profile, has been analyzed on the miscible side, where windows of oscillatory instability extend up to an interspecies repulsion close to the intraspecies one, and for an attractive interspecies coupling, where its two notches separate through a pitchfork [3].

For the mixed ground state the trap already shifts the onset of separation away from the uniform criterion, since the separating components must pay kinetic and trap energy [11, 28, 29, 30, 31, 32, 33, 34], and in tight 1D confinement the kinetic energy keeps the components mixed beyond the uniform criterion [35]; in a quasi-2D trap the same quantum pressure shifts the transition and enhances the spin fluctuations on its miscible side [36]. The shift is controlled by the ratio of the oscillator quantum to the chemical potential, the polarized ground state emerges continuously through a supercritical pitchfork, and phase-separated states of increasing complexity branch off the mixed state at stronger couplings [37]. Since the mixed ground state is a minimum of the energy, its energetic threshold is also a dynamical one, at which the frequency of the spin-dipole mode, the out-of-phase oscillation of the two components, softens to zero [38]; this softening has been observed on the miscible side of the transition in a sodium mixture [39].

In this paper we determine the interspecies repulsion g12∗(n)g_{12}^{*}(n) at which the unpolarized state with nn nodes demixes. For equal intraspecies couplings g11=g22=gg_{11}=g_{22}=g, the interaction energy does not depend on the local composition at the SU(2)-symmetric point g12=gg_{12}=g, and for g12>gg_{12}>g separating the components lowers it. Each component keeps its atom number, however, and the two components therefore separate by moving in opposite directions in the trap, at a cost in kinetic and trap energy; demixing sets in when the released interaction energy balances this cost, and the threshold rises monotonically with nn at every coupling computed.

In the Thomas–Fermi (TF) regime, with μ\mu the chemical potential of the mixed state, the trap correction to the ground-state threshold approaches a universal multiple of (ℏω/μ)2(\hbar\omega/\mu)^2, a scaling that a Gaussian variational treatment also finds [37]. A dark notch near the trap center removes a narrow segment from the region available to the polarization, which raises the cost, and expels atoms, which raise μ\mu and lower it, and the net effect multiplies the correction by one plus a universal multiple of ℏω/μ\hbar\omega/\mu. The step between successive levels in g12/gg_{12}/g is thus of third order in ℏω/μ\hbar\omega/\mu and falls as the inverse square of the atom number, a mesoscopic effect that is appreciable only when μ\mu is no more than a few ℏω\hbar\omega, as for a hundred to a few hundred atoms per component in typical elongated traps; there the computed thresholds give the steps, and the asymptotic laws their order of magnitude. The scaling of the ground-state correction, the softening of the spin-dipole mode that underlies it and the counting of unstable modes through the Krein signature are known [37, 38, 40, 41]; the coefficients and the dependence on nn are new.

Because a moving dark soliton is shallower and carries less energy, the excited states, unlike the ground state, are saddles of the energy even below the threshold [3, 5], and the energy balance alone does not decide whether the separation grows in time. The linearized dynamics shows that the energetic threshold is also the onset of an exponential, non-oscillating growth of the separation, and the polarized state emerges continuously through a supercritical pitchfork; for the dark–dark soliton at the chemical potential of Ref. [3], the spectrum continued beyond the SU(2) point is stable up to this threshold. For the hyperfine pair ∣1,+1⟩\lvert1,+1\rangle and ∣2,−1⟩\lvert2,-1\rangle of 87Rb^{87}\mathrm{Rb} [6], a Feshbach resonance tunes the interspecies repulsion across the thresholds of the lowest states.

Energy Balance of Demixing

Model

We describe the mixture in an elongated trap by the quasi-1D Gross–Pitaevskii (GP) energy functional

E=∫[ℏ22m(∣ψ1′∣2+∣ψ2′∣2)+V(∣ψ1∣2+∣ψ2∣2)+g2(∣ψ1∣4+∣ψ2∣4)+g12∣ψ1∣2∣ψ2∣2]dx.(1)E=\int\left[\frac{\hbar^2}{2m}\left(\lvert\psi_1'\rvert^2+\lvert\psi_2'\rvert^2\right)+V\left(\lvert\psi_1\rvert^2+\lvert\psi_2\rvert^2\right)+\frac{g}{2}\left(\lvert\psi_1\rvert^4+\lvert\psi_2\rvert^4\right)+g_{12}\lvert\psi_1\rvert^2\lvert\psi_2\rvert^2\right]\mathrm{d}x . \tag*{(1)}

where mm is the atomic mass, V=12mω2x2V=\frac{1}{2}m\omega^2x^2 and g11=g22≡gg_{11}=g_{22}\equiv g. The atom numbers of the components are conserved separately, and each component is normalized to unity, ∫∣ψj∣2dx=1\int\lvert\psi_j\rvert^2\mathrm{d}x=1. The reduction to 1D assumes a transverse trap frequency ω⊥≫ω\omega_\perp\gg\omega, with the transverse motion frozen in its ground state. With NN atoms in each component and scattering lengths ajka_{jk} the couplings are gjk=2Nℏω⊥ajkg_{jk}=2N\hbar\omega_\perp a_{jk}. Numerical results are given in oscillator units, with energies in units of ℏω\hbar\omega, lengths in units of aho=(ℏ/mω)1/2a_{\mathrm{ho}}=(\hbar/m\omega)^{1/2} and couplings in units of ℏωaho\hbar\omega a_{\mathrm{ho}}, in which gjk=2Najkω⊥/(ωaho)g_{jk}=2Na_{jk}\omega_\perp/(\omega a_{\mathrm{ho}}). Stationary states are constrained critical points of Eq. (1), with the chemical potentials of the two components as Lagrange multipliers, and their wave functions can be taken real, with a change of sign at each node.

The unpolarized states have ψ1=ψ2=ϕn\psi_1=\psi_2=\phi_n, and both stationary equations reduce to

−ℏ22mϕn′′+Vϕn+geffϕn3=μnϕn,geff=g+g12,(2)-\frac{\hbar^2}{2m}\phi_n''+V\phi_n+g_{\mathrm{eff}}\phi_n^3=\mu_n\phi_n,\qquad g_{\mathrm{eff}}=g+g_{12}, \tag*{(2)}

the stationary equation of a single condensate with coupling geffg_{\mathrm{eff}}. Its nnth excited state ϕn\phi_n, with nn nodes and chemical potential μn\mu_n, is the unpolarized state with nn dark notches.

Polarization at fixed atom numbers

The local composition of the mixture, the fraction of each component at a point, is described by the spin density s=∣ψ1∣2−∣ψ2∣2s=\lvert\psi_1\rvert^2-\lvert\psi_2\rvert^2, the local magnetization of the mixture regarded as a pseudospin-1/2 condensate. With the total density ρ=∣ψ1∣2+∣ψ2∣2\rho=\lvert\psi_1\rvert^2+\lvert\psi_2\rvert^2 the interaction energy becomes

Eint=14∫[(g+g12)ρ2+(g−g12)s2]dx,(3)E_{\mathrm{int}}=\frac{1}{4}\int\left[(g+g_{12})\rho^2+(g-g_{12})s^2\right]\mathrm{d}x, \tag*{(3)}

in which the composition enters through the single coefficient g−g12g-g_{12}; at the SU(2)-symmetric point g12=gg_{12}=g the composition drops out, and for g12>gg_{12}>g a separation of the components at fixed total density lowers the energy. Let uu be a real profile, normalized, ∫u2dx=1\int u^2\mathrm{d}x=1, and orthogonal to the condensate, ∫ϕnudx=0\int\phi_nu\mathrm{d}x=0, and set

ψ1,2=ϕn±εu1+ε2.(4)\psi_{1,2}=\frac{\phi_n\pm\varepsilon u}{\sqrt{1+\varepsilon^2}} . \tag*{(4)}

The orthogonality condition keeps both components normalized for every ε\varepsilon, so that the ansatz conserves N1N_1 and N2N_2. The spin density s=4εϕnu/(1+ε2)s=4\varepsilon\phi_nu/(1+\varepsilon^2) then integrates to zero; where one component gains density the other loses it, and the two components move in opposite directions in the trap.

With the deformations that are symmetric in the two components, such as the change of the total density, made stationary at fixed ε\varepsilon, the energy depends on ε\varepsilon alone and is even in it, E(ε)=E0+αε2+βε4+⋯E(\varepsilon)=E_0+\alpha\varepsilon^2+\beta\varepsilon^4+\cdots, since exchanging the components maps ε→−ε\varepsilon\to-\varepsilon and leaves Eq. (1) invariant. This relaxation is of order ε2\varepsilon^2 and changes the energy only at order ε4\varepsilon^4, and α\alpha therefore follows from the energy along the path of Eq. (4), which Appendix A evaluates,

α[u]=2[Tu+Vu+(3g−g12)M−μn],(5)\alpha[u] = 2\left[T_u+V_u+(3g-g_{12})M-\mu_n\right], \tag*{(5)}

with Tu=(ℏ2/2m)∫(u′)2 dxT_u=(\hbar^2/2m)\int(u')^2\,\mathrm{d}x, Vu=∫Vu2 dxV_u=\int Vu^2\,\mathrm{d}x and M=∫ϕn2u2 dxM=\int\phi_n^2u^2\,\mathrm{d}x.

Threshold of a saddle

The threshold is the coupling at which a polarized branch splits off from the unpolarized state, and it follows from the spectrum of the quadratic form of Eq. (5). Integrating the kinetic term by parts, α=2λ[u]\alpha=2\lambda[u] with λ[u]=∫uA^u dx\lambda[u]=\int u\hat{A}u\,\mathrm{d}x and

A^=−ℏ22m∂x2+V+(3g−g12)ϕn2−μn.(6)\hat{A}=-\frac{\hbar^2}{2m}\partial_x^2+V+(3g-g_{12})\phi_n^2-\mu_n. \tag*{(6)}

Making λ[u]\lambda[u] stationary under the normalization and orthogonality conditions gives the constrained eigenvalue problem A^u=λu+cϕn\hat{A}u=\lambda u+c\phi_n with ∫ϕnu dx=0\int\phi_nu\,\mathrm{d}x=0, derived in Appendix B, whose eigenfunctions uku_k, labeled by increasing eigenvalue with k=0,1,2,…k=0,1,2,\ldots and the label nn of ϕn\phi_n omitted, diagonalize the quadratic energy,

E=E0+2∑kλkεk2+O(ε4),E=E_0+2\sum_k\lambda_k\varepsilon_k^2+O(\varepsilon^4),

with εk\varepsilon_k the amplitude along uku_k.

A polarized branch splits off where a constrained eigenvalue λk\lambda_k passes through zero. Along that direction the energy is E0+2λkε2+βε4E_0+2\lambda_k\varepsilon^2+\beta\varepsilon^4, and for β>0\beta>0 the unpolarized state turns, along this direction, from a minimum into a maximum flanked by two polarized minima at ε2=−λk/β\varepsilon^2=-\lambda_k/\beta, which grow continuously out of it; Figure 1(a) shows this demixing transition for n=2n=2 at g=1g=1. A constrained eigenvalue that is negative at every g12g_{12} creates no new state, and the unpolarized excited states are saddles at every g12g_{12}, since already at g12=0g_{12}=0, where the components do not interact, A^\hat{A} has nn negative constrained eigenvalues, which reflect that ϕn\phi_n is an excited state and are unrelated to miscibility.

Figure 1. Energy cost of polarization at g=1g=1. (a) Energy along the path of Eq. (4) with u=un+1u=u_{n+1} for n=2n=2 at four values of g12g_{12}, including the threshold g12∗g^*_{12}, in shades of the color of n=2n=2 in (b); circles mark the minima of the path energy above threshold, which differ from the polarized stationary states since the path leaves the total density unrelaxed. (b) Quadratic coefficient α\alpha against g12g_{12} for n=0,1,2n=0,1,2. Solid lines give the full calculation, in which the state ϕn\phi_n and the polarization mode un+1u_{n+1} are recomputed at each g12g_{12}; dashed lines give the estimate of Eq. (8), in which both are held at their form at the SU(2) point, so that α\alpha falls linearly with g12g_{12}. The two coincide at g12=gg_{12}=g, and circles mark the thresholds, where the solid lines cross zero.

With δg=g12−g\delta g=g_{12}-g the coefficient of ϕn2\phi_n^2 in A^\hat{A} is 3g−g12=geff−2δg3g-g_{12}=g_{\mathrm{eff}}-2\delta g, and

A^=B^−2δgϕn2,B^=−ℏ22m∂x2+V+geffϕn2−μn,(7)\hat{A}=\hat{B}-2\delta g\phi_n^2,\qquad\hat{B}=-\frac{\hbar^2}{2m}\partial_x^2+V+g_{\mathrm{eff}}\phi_n^2-\mu_n, \tag*{(7)}

where B^ϕn=0\hat{B}\phi_n=0 at every coupling by Eq. (2), so that B^\hat{B} is the single-particle Hamiltonian in the self-consistent potential V+geffϕn2V+g_{\mathrm{eff}}\phi_n^2, measured from μn\mu_n, whose nnth orbital is ϕn\phi_n. We denote by N−(⋅)\mathcal{N}_-(\cdot) the number of negative eigenvalues of an operator on the complement of ϕn\phi_n. Below the SU(2) point, where δg<0\delta g<0, A^\hat{A} lies above B^\hat{B}, whose count is N−(B^)=n\mathcal{N}_-(\hat{B})=n at every g12g_{12}, ϕn\phi_n being its zero mode with nn nodes, and the min–max principle gives N−(A^)≤n\mathcal{N}_-(\hat{A})\le n. We therefore define the threshold g12∗g^*_{12} as the smallest g12>gg_{12}>g at which N−(A^)\mathcal{N}_-(\hat{A}) exceeds nn; at every coupling and excitation level considered here the computed count equals nn at every coupling sampled between 00 and g12∗g^*_{12}, so that no constrained eigenvalue is found to cross zero below the threshold, and n+1n+1 immediately above it. In every case computed the eigenvalue that crosses zero belongs to the mode un+1u_{n+1} with n+1n+1 nodes, whose parity is opposite to that of ϕn\phi_n, as Appendix B shows; for such modes the orthogonality constraint is inactive, and the threshold is the zero of an ordinary eigenvalue of A^\hat{A} in that parity sector.

Estimate at the SU(2) point

At the SU(2) point the interaction energy does not depend on the composition, and the cost of a polarization is set by the single-component condensate. At g12=gg_{12}=g the operator A^\hat{A} of Eq. (7) reduces to B^\hat{B}, whose nth orbital is ϕn\phi_n. The direction ϕn\phi_n itself, ψ1,2∝(1±ε)ϕn\psi_{1,2}\propto(1\pm\varepsilon)\phi_n, changes the composition uniformly and would lower the energy as soon as g12g_{12} exceeds gg, which is the uniform criterion, but it transfers atoms from one component to the other and is excluded by the orthogonality condition. The nn orbitals below ϕn\phi_n are the negative directions of the polarization energy, and the lowest polarization that costs energy is along the next orbital, un+1(0)u_{n+1}^{(0)}, whose excitation energy λn+1(0)\lambda_{n+1}^{(0)} is the spacing between the nnth and (n+1)(n+1)th orbital energies in the self-consistent potential; both belong to the single-component problem. At fixed N1N_1 and N2N_2 the components separate instead by moving in opposite directions in the trap, which costs λn+1(0)\lambda_{n+1}^{(0)} and raises the threshold above gg; this cost is the origin of the trap correction g12∗/g−1g^*_{12}/g-1, which is already present for the ground state.

Away from the symmetric point the term −2 δg ϕn2-2\,\delta g\,\phi_n^2 lowers the energy of a polarization by the interaction energy that the separation releases; with B^\hat{B} and the polarization mode held at their values un+1(0)u_{n+1}^{(0)} at the symmetric point, λn+1\lambda_{n+1} falls linearly in δg\delta g, with the slope −2Mn-2M_n set by the overlap MnM_n, the value of MM of (5) for u=un+1(0)u=u_{n+1}^{(0)}, and the threshold is reached when this release balances the cost,

δg∗≃δgest≡λn+1(0)2Mn,Mn=∫(un+1(0))2ϕn2 dx.(8)\delta g^{*}\simeq\delta g_{\mathrm{est}}\equiv\frac{\lambda_{n+1}^{(0)}}{2M_n},\qquad M_n=\int\left(u_{n+1}^{(0)}\right)^2\phi_n^2\,dx. \tag*{(8)}

Both factors follow from the single-component solution at the symmetric point, and no parameter is adjusted. Figure 1(b) illustrates this balance at g=1g=1; the computed coefficient α=2λn+1\alpha=2\lambda_{n+1}, shown solid, coincides with the estimate, shown dashed, at g12=gg_{12}=g but falls faster, since the estimate holds ϕn\phi_n, μn\mu_n and un+1u_{n+1} at their values at the symmetric point, so that the estimate overestimates the threshold. The estimate is useful because it factorizes the threshold into an excitation energy and an overlap, which can be followed separately as functions of nn.

Threshold map and the two regimes

Thresholds and the estimate

The threshold rises monotonically with nn at every coupling; Table 1 lists g12∗(n)g_{12}^{*}(n) for six values of gg, and Figure 2(a) plots the shift δg∗=g12∗−g\delta g^{*}=g_{12}^{*}-g against n\sqrt{n}, which shows no sign of saturation up to n=16n=16. The estimate of (8) places the threshold too high but captures how it grows with nn. Figure 2(b) shows the ratio δg∗/δgest\delta g^{*}/\delta g_{\mathrm{est}} of the computed shift to the estimate; at each gg it changes by at most 0.16 between n=0n=0 and n=16n=16, whereas δg∗\delta g^{*} itself grows by a factor between 2.2 and 2.9 over the same range. The ratio lies between 0.74 and 0.96, below unity at every nn and gg computed. The estimate holds the condensate and the polarization mode at their form at the SU(2) point, whereas the computed λn+1\lambda_{n+1} includes their change as g12g_{12} rises to the threshold, and the solid lines of Figure 1(b) show that this change lowers it further. The condensate changes most for the low states at small gg, where δg∗\delta g^{*} exceeds gg and geff=g+g12g_{\mathrm{eff}}=g+g_{12} grows appreciably between the symmetric point and the threshold, and there the ratio is lowest, 0.74 for n=0n=0 at g=0.5g=0.5. In the TF limit the condensate shape hardly changes, since δg∗≪g\delta g^{*}\ll g, but the term −2δgϕn2-2\delta g\phi_n^2 of (7) still deforms the polarization mode, and this deformation alone leaves the ratio for the ground state at about 0.98.

nng=0.5g=0.5112244881616
02.08252.46933.29685.08088.864516.6811
12.48852.83503.59305.28408.983116.7445
22.79253.12043.84015.46639.093616.8042
33.04323.36024.05515.63339.198616.8615
43.26033.56994.24725.78789.299416.9171
63.62923.92934.58266.06809.489917.0243
83.94114.23514.87266.31869.668317.1274
104.21494.50475.13086.54669.836417.2270
124.46104.74785.36516.75699.995917.3236
144.68614.97035.58066.952710.147717.4175
164.89435.17655.78117.136510.292917.5090

Table 1. Demixing threshold g12∗(n)g_{12}^{*}(n) in oscillator units.

Figure 2. Demixing threshold for six values of gg. (a) Threshold shift δg∗=g12∗−g\delta g^{*}=g_{12}^{*}-g against n\sqrt{n}; lines are fits linear in n\sqrt{n} over n=4,6,…,16n=4,6,\ldots,16 at g≤2g\leq2, and thin lines join the points at g≥4g\geq4. (b) Ratio of the computed shift to the estimate of Eq. (8) against nn.

Since the ratio varies slowly with nn, the growth of the threshold is carried by the two factors of Eq. (8), and which factor carries it depends on how strongly the interaction modifies the trap, which we measure by the mean-field shift of the chemical potential at the symmetric point, μn−(n+12)ℏω\mu_n-(n+\frac{1}{2})\hbar\omega. At weak intraspecies coupling, where this shift is smaller than ℏω\hbar\omega, as for every n≤16n\leq16 at g≤1g\leq1 and for n≥2n\geq2 at g=2g=2, the self-consistent potential V+geffϕn2V+g_{\mathrm{eff}}\phi_n^2 differs from the trap by a small correction, and the thresholds lie far from the SU(2) point, at g12∗/gg_{12}^{*}/g between 1.9 and 10. In the TF regime, where the shift is much larger than ℏω\hbar\omega, the interaction flattens the potential inside the cloud, and the healing length is short compared with the size of the cloud, as for n≤4n\leq4 at g=16g=16, where the shift exceeds 5ℏω5\hbar\omega, although the notches of the stationary trains are then only four to five healing lengths apart; the couplings g=4g=4 and 8 lie between the two regimes. For trap frequencies of 2π×302\pi\times30 Hz and 2π×6002\pi\times600 Hz and a scattering length of 97.7a097.7a_0, g=16g=16 corresponds to 152 atoms per component, whereas g≤2g\leq2 means fewer than 20, so few that exact calculations describe the phase separation of such microscopic mixtures as a crossover [42], and the mean-field thresholds of this regime are idealizations.

Weak intraspecies coupling

The intraspecies repulsion is not needed for demixing in this regime; as g→0g \to0 the thresholds tend to finite values, from g12∗=1.720g_{12}^{*}=1.720 for n=0n=0 to 2.972 for n=4n=4 and 4.626 for n=16n=16, at which the interspecies repulsion alone balances the cost, as kinetic energy alone does in a box without intraspecies interactions [35]. Where δg∗\delta g^{*} exceeds gg, the balance is better evaluated at the threshold than at the SU(2) point. To first order in the interactions about the linear limit, with ϕn\phi_n and un+1u_{n+1} the normalized eigenfunctions hnh_n and hn+1h_{n+1} of the harmonic oscillator, hn(x)=(2nn!πaho)−1/2Hn(x/aho)e−x2/2aho2h_n(x)=(2^n n!\sqrt{\pi}a_{\rm ho})^{-1/2}H_n(x/a_{\rm ho})e^{-x^2/2a_{\rm ho}^2} with HnH_n the Hermite polynomials, and μn=(n+12)ℏω+geffPˉn\mu_n=(n+\frac{1}{2})\hbar\omega+g_{\rm eff}\bar{P}_n, the condition λn+1=0\lambda_{n+1}=0 gives

g12∗≃ℏω+g(3Mˉn−Pˉn)Mˉn+Pˉn,(9)g_{12}^{*}\simeq\frac{\hbar\omega+g(3\bar{M}_n-\bar{P}_n)}{\bar{M}_n+\bar{P}_n}, \tag*{(9)}

with Mˉn=∫hn2hn+12 dx\bar{M}_n=\int h_n^2h_{n+1}^2\,dx and Pˉn=∫hn4 dx\bar{P}_n=\int h_n^4\,dx, in which the interspecies repulsion lowers the energy of the polarization mode by g12Mˉng_{12}\bar{M}_n and raises the chemical potential, from which this energy is measured, by g12Pˉng_{12}\bar{P}_n.

Although the interaction energy at threshold is not small, g12∗Pˉng_{12}^{*}\bar{P}_n lying between 0.6 and 0.7 ℏω\hbar\omega at g=0g=0, Eq. (9) lies between 3% and 6% below the computed thresholds there and within 5% of Table I for n≥4n\geq4 at g≤2g\leq2; it lies further below for the lowest states, by 12% for n=0n=0 at g=0.5g=0.5 and by up to 29% at g=2g=2, where the interaction energy at threshold, (g+g12∗)Pˉn(g+g_{12}^{*})\bar{P}_n, exceeds ℏω\hbar\omega, and beyond its formal range it serves as an interpolation. Its denominator, which falls from 0.60 for n=0n=0 to 0.23 for n=16n=16 because hn2h_n^2 and hn+12h_{n+1}^2 spread over a cloud that grows as n\sqrt{n} and dilutes, carries the growth of the threshold with nn.

The same dilution appears in the two factors of Eq. (8). The orbital spacing λn+1(0)\lambda_{n+1}^{(0)}, which equals ℏω\hbar\omega in the trap alone, lies below ℏω\hbar\omega, since the mean-field potential flattens the trap, and approaches it as the cloud dilutes, as Figure 3(a) shows, whereas the overlap MnM_n falls, as Figure 3(b) shows, its inverse rising by 42% to 44% between n=4n=4 and 16 at g≤2g\leq2, affinely in the rms radius Rn=(∫x2ϕn2 dx)1/2R_n=(\int x^2\phi_n^2\,dx)^{1/2} of the cloud. Since RnR_n grows as n\sqrt{n}, the computed shifts are well described by δg∗≃κ0+κ1n\delta g^{*}\simeq\kappa_0+\kappa_1\sqrt{n} for the levels n=4,6,…,16n=4,6,\ldots,16 of Table I, with κ1=0.82\kappa_1=0.82, 0.80 and 0.77 at g=0.5g=0.5, 1 and 2, as the lines in Figure 2(a) show. This form describes the thresholds over this range of nn rather than asymptotically; in the linear limit MnRnM_nR_n grows logarithmically with nn, and the estimate grows as n/ln⁡n\sqrt{n}/\ln n far beyond n=16n=16.

Figure 3. The two factors of Eq. (8) for six values of gg. (a) Excitation energy at the SU(2) point; the dotted line marks ℏω\hbar\omega. (b) Overlap.

The Thomas–Fermi regime

At strong coupling this account fails on both counts, since the orbital spacing falls far below ℏω\hbar\omega and rises with nn, as Figure 3(a) shows, whereas at g=16g=16 the overlap hardly changes with nn, as Figure 3(b) shows, although the cloud doubles in size between n=2n=2 and 16. The growth of the threshold then comes from the spacing, and its dependence on nn changes; the exponent pp of a fit b0+b1npb_0+b_1n^p to the computed shifts over the same levels rises from 0.484(4) at g=0.5g=0.5 through 0.536(6) at g=2g=2 and 0.754(8) at g=8g=8 to 0.856(5) at g=16g=16, approaching a dependence linear in nn. Such a fit describes this range of nn rather than the asymptotic dependence. The spacing at strong coupling is set by the TF profile, in which the interaction flattens the potential inside the cloud and the spacing becomes a spin-wave energy of order (ℏω)2/μ(\hbar\omega)^2/\mu, which falls as gg increases.

For the ground state the TF profile reduces the threshold to a universal coefficient, since δg∗/g→0\delta g^{*}/g\to0 and the background keeps its TF form at the symmetric point, with geff=2gg_{\rm eff}=2g and μ=[3g/(22ℏωaho)]2/3ℏω\mu=[3g/(2\sqrt{2}\hbar\omega a_{\rm ho})]^{2/3}\hbar\omega the TF chemical potential at the midpoint, in units of the TF radius the threshold condition A~u1=0\tilde{A}u_1=0 becomes an eigenvalue problem without parameters, whose lowest odd solution, found in Appendix C, gives

g12∗(0)g−1≃1.822(ℏωμ)2.(10)\frac{g_{12}^{*}(0)}{g}-1\simeq1.822\left(\frac{\hbar\omega}{\mu}\right)^2. \tag*{(10)}

For g≥4g\geq4 the computed correction exceeds Eq. (10) by at most 3%, by 2% at g=16g=16, and approaches it as gg grows.

Expanding the Gaussian variational threshold, Eq. (20) of Ref. [37], for small ℏω/μ\hbar\omega/\mu gives the same form with the coefficient 25/825/8, about 70% above the asymptotic value. Reference [37] works at fixed chemical potential, and its GP calculations locate the onset of the polarized ground state at g12/g≃1.08g_{12}/g\simeq1.08 when the chemical potential of the mixed state at the onset is 5ℏω5\hbar\omega; this corresponds to g=10.0g=10.0, where the computed threshold is g12∗/g=1.080g_{12}^{*}/g=1.080 and Eq. (10), with the TF chemical potential 4.83ℏω4.83\hbar\omega, gives 1.078. Since μ∝g2/3\mu\propto g^{2/3} in 1D, the relative correction g12∗(0)/g−1g_{12}^{*}(0)/g-1 decays as g−4/3g^{-4/3}; it is appreciable only when μ\mu is a few ℏω\hbar\omega, and in a large condensate the uniform criterion is recovered.

For the excited states the spacing grows with nn because of the dark notches, where the flattened potential keeps narrow wells that bind the nn lower orbitals. On the scale of the cloud an excited state polarizes like the spin dipole of the ground state, and each notch changes the threshold in two ways. The density vanishes at the notch, and the polarization mode, which cannot use this region, keeps its slope across it and jumps by an amount proportional to this slope; for the mode the notch is thus an excluded segment of effective length 222\sqrt{2} healing lengths ξ=ℏ/(2mμ)\xi= \hbar/(2m\mu), which shortens the region available to it and raises its energy; near a central notch the mode reverses sign within 1.7ξ1.7\xi, so that each notch also carries a narrow domain of reversed polarization, as Figure 4(a) shows. The notch also expels atoms, which at fixed atom number raises the chemical potential by ℏω/2\hbar\omega/\sqrt{2} and lowers the threshold, since by Eq. (10) δg∗\delta g^* falls as μ−2\mu^{-2} at fixed gg. Matching the threshold mode near a notch to the solution of the TF problem, as Appendix C shows, gives the two contributions of a central notch to the relative change of δg∗\delta g^* as +2.092+2.092 and −1.414-1.414 times ℏω/μ\hbar\omega/\mu, so that each central notch enlarges δg∗\delta g^* by the fraction 0.678ℏω/μ0.678\hbar\omega/\mu and, for nℏω/μ≪1n\hbar\omega/\mu\ll1 and with μ\mu the TF chemical potential of Eq. (10),

Figure 4. Density notches in the TF regime. (a) Spin-density pattern ϕ1u2\phi_{1u2} of the threshold mode of the dark–dark soliton against the distance from its notch x0x_0 in units of the healing length ξ\xi at the chemical potential of the state, scaled to its largest value within 10ξ10\xi of the notch; the dotted line is the analytical solution near an isolated notch. (b) Relative threshold shift per notch, [δg∗(n)/δg∗(0)−1]/n[\delta g^*(n)/\delta g^*(0)-1]/n in units of ℏω/μ\hbar\omega/\mu, against ℏω/μ\hbar\omega/\mu, with μ\mu the TF chemical potential of Eq. (10), for g=16g=16 to 512 (symbols), with quadratic fits over g=32g=32 to 512 (lines); the star marks the coefficient of Eq. (11).

g12∗(n)g−1≃1.822(ℏωμ)2(1+0.678nℏωμ).(11)\frac{g_{12}^*(n)}{g}-1 \simeq1.822\left(\frac{\hbar\omega}{\mu}\right)^2\left(1+0.678\frac{n\hbar\omega}{\mu}\right). \tag*{(11)}

Equation (11) holds for notches close to the trap center and far apart. A notch displaced from the center contributes more, since it also removes density where the polarization mode, which vanishes at the center, has a finite amplitude. The single central notch of the dark–dark soliton meets the condition at every gg, and its computed fraction, 0.615ℏω/μ0.615\hbar\omega/\mu at g=16g=16, lies within 10% of the asymptotic value. The notches of the stationary trains approach the center only slowly, those of n=4n=4 lying within 0.390.39 times the TF radius RTFR_{\rm TF} at g=16g=16 and within 0.07RTF0.07R_{\rm TF} at g=512g=512, and only four to five healing lengths apart at g=16g=16; the computed fractions per notch of all levels nevertheless extrapolate to 0.678ℏω/μ0.678\hbar\omega/\mu as μ\mu grows, with an empirical leading correction between −0.25n-0.25n and −0.33n-0.33n times (ℏω/μ)2(\hbar\omega/\mu)^2, which makes the step between successive levels fall with nn, as Figure 4(b) shows. Equation (11) makes the step between successive levels, δg∗(n+1)−δg∗(n)≃1.24g(ℏω/μ)3\delta g^*(n+1)-\delta g^*(n)\simeq1.24g(\hbar\omega/\mu)^3, of third order in ℏω/μ\hbar\omega/\mu. Equations (10) and (11) are asymptotic in ℏω/μ\hbar\omega/\mu for the 1D model, which describes an elongated gas as long as μ≪ℏω⊥\mu\ll\hbar\omega_\perp; at g=16g=16, where μ=6.6ℏω\mu=6.6\hbar\omega, the step formula exceeds the computed steps of Table I by 9% to 24% for n≤4n\leq4 and by about 50% at n=16n=16. At fixed nn the threshold approaches that of the ground state as μ\mu grows, and since gg is proportional to the atom number in a given trap, the steps in g12/gg_{12}/g, and with them the steps in the interspecies scattering length, fall as the inverse square of the atom number.

The Polarized State

Supercritical pitchfork

Symmetry-breaking pitchforks of solitary waves in binary condensates can be supercritical or subcritical [16, 43, 44], and whether the polarized state grows continuously out of the unpolarized one is decided by the quartic coefficient β\beta of E(ε)E(\varepsilon). For β>0\beta>0 the polarized stationary points, ε2=−α/2β\varepsilon^2=-\alpha/2\beta, exist where α<0\alpha<0 and grow continuously from zero, in a supercritical pitchfork, whereas for β<0\beta<0 they fold back to the side α>0\alpha>0 and coexist with the unpolarized state, in a subcritical one.

Because the relaxation of the total density and of μn\mu_n contributes at order ε4\varepsilon^4, β\beta does not follow from the path of Eq. (4), and we obtain it instead from the polarized branch itself. Since the stationary equations are invariant under the exchange of the components combined with x→−xx\to-x, and ϕn\phi_n and un+1u_{n+1} have opposite parities, the branch lies in the invariant subspace ψ2(x)=(−1)nψ1(−x)\psi_2(x)=(-1)^n\psi_1(-x), and we continue it in its projection η=∫un+1ψ1 dx\eta=\int u_{n+1}\psi_1\,dx on the threshold mode, which equals ε\varepsilon to leading order, as Appendix § 2.4 describes.

Table II lists β\beta at sixteen points; it is positive at every point, and the pitchfork is supercritical in every case computed. Near threshold λn+1≃λn+1′(g12−g12∗)\lambda_{n+1}\simeq\lambda'_{n+1}(g_{12}-g_{12}^*), where λn+1′<0\lambda'_{n+1}<0 is the derivative with respect to g12g_{12}, and minimizing E0+2λn+1η2+βη4E_0+2\lambda_{n+1}\eta^2+\beta\eta^4 gives the leading-order laws

η2≃∣λn+1′∣β(g12−g12∗),Eunpol−Epol≃(λn+1′)2β(g12−g12∗)2.(12)\begin{aligned} \eta^2 &\simeq\frac{|\lambda'_{n+1}|}{\beta}(g_{12}-g_{12}^*),\\ E_{\mathrm{unpol}}-E_{\mathrm{pol}} &\simeq\frac{(\lambda'_{n+1})^2}{\beta}(g_{12}-g_{12}^*)^2. \tag*{(12)} \end{aligned}

The energies of the two branches thus meet at threshold with equal value and equal slope, and the transition is continuous. Figure 5 compares Eq. (12) with the branches continued in η\eta for n=0n=0, 1 and 2 at g=1g=1 and 16; the amplitude grows with the mean-field exponent 1/2 over more than two decades in g12−g12∗g_{12}-g_{12}^{*}, and the energy difference follows Eq. (12) to within 0.5%. Along all sixteen branches g12g_{12} increases monotonically with η\eta up to at least 2g12∗2g_{12}^{*}, without folding back at larger amplitude. The excited states thus demix through the same supercritical pitchfork as the ground state [37].

Supercritical pitchfork plots

Figure 5. Supercritical pitchfork. Projection amplitude η\eta of the polarized branch against the distance from threshold at (a) g=1g=1 and (b) g=16g=16. Symbols are computed branches; lines are Eq. (12) with β\beta and λn+1′\lambda_{n+1}^{\prime} from Table II; for clarity η\eta is multiplied by 2 for n=1n=1 and by 4 for n=2n=2.

Domain walls

Near threshold the polarized state divides the condensate into domains of opposite polarization, since its spin density s≃4ϵϕnun+1s \simeq4\epsilon\phi_n u_{n+1} vanishes at the nn nodes of ϕn\phi_n and the n+1n+1 nodes of un+1u_{n+1}. At threshold un+1u_{n+1} and ϕn\phi_n are zero modes of A^\hat{A} and B^\hat{B}, the former by the parity argument of Appendix B, and since A^=B^−2δg ϕn2\hat{A}=\hat{B}-2\delta g\,\phi_n^2 lies below B^\hat{B} wherever ϕn≠0\phi_n\ne0, the Sturm comparison theorem [45]

places at least one node of un+1u_{n+1} strictly inside each of the n+1n+1 intervals into which the nodes of ϕn\phi_n divide the axis, including the two outer intervals, on which both functions decay at infinity, so that their Wronskian vanishes there and the Wronskian proof of the theorem [46] applies unchanged, and exactly one, since un+1u_{n+1} has n+1n+1 nodes. The polarization therefore changes sign 2n+12n+1 times, as Figure 6 shows for n=0n=0 to 4, and each sign change is a domain wall.

Figure 6. Polarized states at g12=g12∗+0.1g_{12}=g_{12}^{*}+0.1 and g=16g=16, for n=0,…,4n=0,\ldots,4 from left to right. (a) Densities ∣ψ1∣2|\psi_1|^2 (line) and ∣ψ2∣2|\psi_2|^2 (shaded). (b) Local polarization s/ρs/\rho inside the cloud; circles mark its 2n+12n+1 sign changes, the nn walls at the density notches in red and the n+1n+1 walls at the nodes of un+1u_{n+1} in blue.

The two sets of walls differ in character, and Figure 6(b) marks them in different colors; at a node of un+1u_{n+1}, blue in the figure, the polarization crosses zero smoothly, as the perturbation does, whereas near a node x0x_0 of ϕn\phi_n, red in the figure, the components are ψ1,2≃ϕn′(x0)(x−x0)±ϵun+1(x0)\psi_{1,2}\simeq\phi_n^{\prime}(x_0)(x-x_0)\pm\epsilon u_{n+1}(x_0), which vanish at x0∓ϵun+1(x0)/ϕn′(x0)x_0\mp\epsilon u_{n+1}(x_0)/\phi_n^{\prime}(x_0). Each component thus keeps its dark notch, displaced opposite to that of the other, and between the two notches s/ρs/\rho runs from −1-1 to $+1; this wall is sharp, with a width proportional to ϵ\epsilon, and locally the pair of displaced notches is a polarized dark soliton of the kind found in uniform media [25, 26]. In the TF regime the polarized excited state is thus a spin dipole of the whole cloud with, at each notch, a polarized dark soliton flanked by a narrow domain of reversed polarization that ends at the adjacent node of un+1u_{n+1}, on the outer side of an off-center notch and on both sides of a central one. The count persists far from threshold; along the branches continued up to $2g_{12}^{*}$ for n≤2n\le2 at g=1g=1 and 16, the polarization changes sign 2n+12n+1 times at every amplitude.

Dynamical stability

Onset of the instability

Because the unpolarized excited states are saddles of the energy at every coupling, the energy balance of Sec. II does not decide by itself whether a polarization grows in time, just as a dark soliton, whose energy decreases as it moves and becomes shallower, still oscillates stably in a trap. Writing the perturbations of the two components as δψj=vj+iwj\delta\psi_j=v_j+iw_j, the linearized dynamics separates into a charge sector, δψ1=δψ2\delta\psi_1=\delta\psi_2, which carries no polarization, and a spin sector, δψ1=−δψ2\delta\psi_1=-\delta\psi_2, which carries it. In each sector the linearized GP equations read ℏ∂tv=B^w\hbar\partial_t v=\hat{B}w and ℏ∂tw=−A^charge/spinv\hbar\partial_t w=-\hat{A}_{\mathrm{charge/spin}}v, so that the real parts are acted on by A^charge\hat{A}_{\mathrm{charge}} or A^spin\hat{A}_{\mathrm{spin}} and the imaginary parts by B^\hat{B}, with frequencies

ℏ2Ωcharge/spin2=eig⁡(B^A^charge/spin),(13)\hbar^2\Omega_{\mathrm{charge/spin}}^2=\operatorname{eig}(\hat{B}\hat{A}_{\mathrm{charge/spin}}), \tag*{(13)}

where eig denotes the eigenvalues of the operator product, A^spin=A^\hat{A}_{\mathrm{spin}}=\hat{A} and A^charge=B^+2geffϕn2\hat{A}_{\mathrm{charge}}=\hat{B}+2g_{\mathrm{eff}}\phi_n^2. A frequency with a nonzero imaginary part signals an instability, whose perturbation grows as eΓte^{\Gamma t} with the growth rate Γ=∣Im⁡Ω∣\Gamma=|\operatorname{Im}\Omega|.

For linearizations of this form, the numbers of negative eigenvalues of the two operators fix a weighted count of the unstable modes and of the oscillation modes of negative energy in each sector [40, 41],

Figure 8. Real-time evolution of the n=2n=2 state at g=16g=16. (a) Polarization amplitude ∥s∥2\lVert s\rVert_2 against time above threshold, starting from 1.1×10−51.1\times10^{-5}, and below threshold, starting from 1.6×10−31.6\times10^{-3}; the dashed line grows at the linearized rate Γ\Gamma. (b) Local polarization inside the cloud against position and time above threshold, g12=g12∗+0.3g_{12}=g_{12}^{*}+0.3; dashed and dotted lines mark the nodes of ϕ2\phi_2 and u3u_3.

kr+2kc+2ki=N−(A^)+N−(B^),(14)k_r+2k_c+2k_i=\mathcal{N}_{-}(\hat{A})+\mathcal{N}_{-}(\hat{B}), \tag*{(14)}

where krk_r, kck_c and kik_i count the positive real eigenvalues of the linearization, its complex quartets and its oscillation modes of negative Krein signature, which lower the energy when excited, and both counts are taken on the complement of ϕn\phi_n, which removes the phase zero mode of B^\hat{B} and accounts for the conservation of N1−N2N_1-N_2, as in the constrained count of Refs. [40, 41]. The count applies where A^\hat{A} has no zero eigenvalue on this complement and QQ of Appendix B is finite and nonzero, that is, for g12≠gg_{12}\ne g and g12≠g12∗g_{12}\ne g_{12}^{*}. In the spin sector N−(B^)=n\mathcal{N}_{-}(\hat{B})=n,

since ϕn\phi_n is the zero mode of B^\hat{B} and has nn nodes, and N−(A^)\mathcal{N}_{-}(\hat{A}) equals nn below threshold and n+1n+1 above it. The right-hand side of Eq. (14) therefore changes from 2n2n to 2n+12n+1 at g12∗g_{12}^{*}, and since an odd count requires kr≥1k_r\ge1, the energetic threshold is also the onset of a non-oscillatory dynamical instability. Below threshold the count is even, and since the computed spectra contain no real unstable eigenvalue there, it is carried by the nn anomalous modes, ki=nk_i=n, the out-of-phase oscillations of the dark notches of the two components, unless an anomalous mode collides in frequency with a mode of positive signature and the pair forms a complex quartet, an oscillatory instability that arises in a Hamiltonian–Hopf bifurcation. At the SU(2) point A^=B^\hat{A}=\hat{B} and ℏ2Ω2=eig⁡(B^2)≥0\hbar^2\Omega^2=\operatorname{eig}(\hat{B}^2)\ge0, and the spin sector is stable there for every nn.

Growth rates

Figure 7(a) shows that at g=1g=1 the spin-sector growth rate Γ\Gamma becomes nonzero at threshold for every nn and rises as (g12−g12∗)1/2(g_{12}-g_{12}^{*})^{1/2}, reaching about 0.37ω0.37\omega at g12∗+0.5g_{12}^{*}+0.5. For n≥1n\geq1 the spin sector carries in addition Hamiltonian–Hopf instabilities near Ω≃ω\Omega\simeq\omega at small g12g_{12}, with rates up to 0.05ω0.05\omega, which for n=1n=1 arise from the resonance between the out-of-phase oscillation of the notches and the lowest spin mode [3] and lie close to the interval 0.019<g12/g<0.3930.019<g_{12}/g<0.393 found near the linear limit [20]; these windows close at g12<0.5g_{12}<0.5 for n≤3n\leq3, and the spin sector is stable from there up to threshold. The charge sector, shown dashed, carries for n≥2n\geq2 the oscillatory instability of multiple dark solitons in a single condensate [5], with rates of order 10−2ω10^{-2}\omega at g=1g=1, and is stable for n=0n=0 and 11, since at the degenerate frequency Ω=ω\Omega=\omega of the linear limit the mode of positive signature is the Kohn mode, whose frequency stays at ω\omega exactly, and no quartet forms.

Near threshold the unstable mode has v≃un+1v\simeq u_{n+1}, and since B^\hat{B} is invertible in the parity sector of un+1u_{n+1}, eliminating w=ℏΓB^−1vw=\hbar\Gamma\hat{B}^{-1}v and projecting on un+1u_{n+1} gives

ℏ2Γ2≃∣λn+1′∣(g12−g12∗)κn,κn=∫un+1B^−1un+1 dx,(15)\hbar^{2}\Gamma^{2}\simeq\frac{\lvert\lambda_{n+1}^{\prime}\rvert(g_{12}-g_{12}^{*})}{\kappa_n},\qquad\kappa_n=\int u_{n+1}\hat{B}^{-1}u_{n+1}\,\mathrm{d}x, \tag*{(15)}

where B^−1un+1\hat{B}^{-1}u_{n+1} is the pattern of relative phase that drives the polarization, and 1/κn1/\kappa_n equals λn+1(0)\lambda_{n+1}^{(0)} at the SU(2) point, where un+1u_{n+1} is an eigenfunction of B^\hat{B}. In the TF regime both factors are small and the rates are smaller; at g=16g=16, Figure 7(b) shows Γ≃0.21(g12−g12∗)1/2ω\Gamma\simeq0.21(g_{12}-g_{12}^{*})^{1/2}\omega for n=0n=0, with coefficients 0.230.23 and 0.240.24 for n=1n=1 and 22, as (15) gives with λn+1′\lambda_{n+1}^{\prime} from Table 2.

ggn=0n=0n=1n=1n=2n=2n=4n=4
β\beta
0.51.8161.8891.9051.905
11.7131.8491.8971.926
41.2441.4671.6211.810
160.6280.7120.7920.939
−λn+1′-\lambda_{n+1}^{\prime}
0.50.5220.4360.3850.325
10.4880.4260.3830.327
40.3800.3740.3600.331
160.2600.2670.2720.276

Table 2. Quartic coefficient β\beta of E(ϵ)E(\epsilon) at threshold, obtained from the polarized branch, and slope λn+1′\lambda_{n+1}^{\prime} of the constrained eigenvalue at threshold, in oscillator units.

Figure 7. Bogoliubov growth rates of the unpolarized states. (a) Spin sector (solid) and charge sector (dashed, n=2n=2 and 33, nearly coincident; stable for n=0n=0 and 11) at g=1g=1; circles mark the thresholds, and the inset magnifies the windows at small g12g_{12}. (b) Spin sector at g=16g=16 for n=0n=0, 11 and 22 against the distance from threshold.

Integrating the time-dependent GP equations that follow from Eq. (1) for the n=2n=2 state at g=16g=16 tests these rates; we measure the polarization by ∥s∥2=(∫s2 dx)1/2\lVert s\rVert_{2}=(\int s^{2}\,\mathrm{d}x)^{1/2}, which does not depend on the relative phase of the components. As Fig. 8(a) shows, at g12=g12∗+0.30g_{12}=g_{12}^{*}+0.30 a seed along the unstable Bogoliubov eigenvector, together with random noise of the same order, grows exponentially over four decades at the linearized rate Γ=0.13ω\Gamma=0.13\omega, whereas at g12=g12∗−0.30g_{12}=g_{12}^{*}-0.30, below the weak oscillatory window of this state near the threshold, a seed along un+1u_{n+1} oscillates without exceeding its initial value. Since the dynamics is conservative, ∥s∥2\lVert s\rVert_{2} overshoots its stationary value, reaching 0.530.53 at ωt≃87\omega t\simeq87, and returns toward the unpolarized state; whenever the polarization is appreciable it changes sign 2n+1=52n+1=5 times at the positions of the walls of the stationary state, as Fig. 8(b) shows, and the domains of the polarized state therefore form in the dynamics, as domains of opposite composition do when a binary condensate is driven across the miscibility transition [47].

The Bogoliubov spectrum at g=16g=16 also has weak oscillatory windows in the spin sector of n=1n=1 and 22 just below the threshold, as Figure 7(b) shows, narrow resonances between higher spin modes with rates below 0.01ω0.01\omega; for n=1n=1 the window extends up to the threshold, so that the non-oscillatory growth of this state is preceded in g12g_{12} by a weak oscillatory one. Such resonances recur at isolated values of gg; sampling gg at the even values from 8 to 26, and for n=1n=1 also at 14.7 and 15.4, we find them within 0.3 below the threshold for n=1n=1 at g=16g=16 and 24 and for n=2n=2 at g=16g=16, 18 and 26, so that the coupling g=16g=16 used in Sec. VII lies in such an interval. Far below the SU(2) point the spin sector carries stronger Hamiltonian–Hopf windows, with rates up to 0.26ω0.26\omega, which close at least 3.8 below the threshold.

For the trapped dark–dark soliton at μ=10ℏω\mu=10\hbar\omega, with equal chemical potentials of the two components, Ref. [3] reported stable and unstable windows alternating up to a value of g12/gg_{12}/g close to unity, closer to unity the larger the chemical potential, from a Bogoliubov spectrum computed for g12/g≤1g_{12}/g\leq1. At this fixed chemical potential the profile depends on the couplings only through g+g12g+g_{12}, and the spin sector of the n=1n=1 state has its last oscillatory window between g12/g=0.89g_{12}/g=0.89 and 0.930.93, is stable above it up to the threshold at g12∗/g=1.023g_{12}^{*}/g=1.023, and grows exponentially beyond it, which extends that spectrum through the SU(2) point and ends its stable region at the polarization threshold. At fixed atom number, with μ=10ℏω\mu=10\hbar\omega reached at the threshold, the threshold is the same and the last window lies between 0.950.95 and 0.960.96; Eqs. (10) and (11) give this threshold with μ=9.2ℏω\mu=9.2\hbar\omega, the TF chemical potential of Eq. (10) for this gg.

The polarized states themselves are linearly stable near threshold for n=0n=0 at both couplings, for n=1n=1 at g=1g=1 and for n=2n=2 at g=16g=16, whereas the polarized n=2n=2 state at g=1g=1 and n=1n=1 state at g=16g=16 inherit the weak oscillatory instabilities that the unpolarized states carry at threshold, the charge-sector one and the spin-sector window, with rates of 0.025ω0.025\omega and 0.012ω0.012\omega that vanish by 1.2g12∗1.2g_{12}^{*} and 1.004g12∗1.004g_{12}^{*}. Further from threshold the polarized ground state remains stable up to 2g12∗2g_{12}^{*}, whereas the states with n=1n=1 and 22 cross oscillatory windows with rates up to 0.2ω0.2\omega.

Unequal couplings and population imbalance

In real mixtures the intraspecies couplings differ slightly, g11=g+ζg_{11}=g+\zeta and g22=g−ζg_{22}=g-\zeta with gg their arithmetic mean, and the populations are not exactly balanced. For the 87^{87}Rb pair ∣1,−1⟩\lvert1,-1\rangle, ∣2,−2⟩\lvert2,-2\rangle, in which dark–dark and dark–antidark solitons have been observed [1, 2], ζ/g=7.1×10−3\zeta/g=7.1\times10^{-3} [1], and for the pair ∣1,+1⟩\lvert1,+1\rangle, ∣2,−1⟩\lvert2,-1\rangle, ζ/g=2.8×10−2\zeta/g=2.8\times10^{-2} [6]. The interaction energy becomes

Eint=∫{14[(g+g12)ρ2+(g−g12)s2]+ζ2ρs} dx,(16)E_{\mathrm{int}}=\int\left\{\frac{1}{4}\left[(g+g_{12})\rho^{2}+(g-g_{12})s^{2}\right]+\frac{\zeta}{2}\rho s\right\}\,\mathrm{d}x, \tag*{(16)}

in which the coefficient of s2s^2 is set by the arithmetic mean gg, and a new term couples the total density to the polarization and breaks the exchange symmetry of the components explicitly.

At ζ≠0\zeta\ne0 a common profile ψ1=ψ2=ϕ\psi_1 = \psi_2 = \phi no longer solves the stationary equations, since their difference requires 2ζϕ22\zeta\phi^2 to equal the difference of the two chemical potentials, a constant, which a trapped profile cannot satisfy. Continuing the ζ=0\zeta= 0 solution in ζ\zeta gives a nearby branch ψ1,2=ϕn±χ\psi_{1,2} = \phi_n \pm\chi with χ=O(ζ)\chi= O(\zeta), on which the component with the larger self-repulsion expands slightly. Since ϕn2\phi_n^2 is even, χ\chi carries the parity of ϕn\phi_n and the induced polarization is even, whereas the destabilizing mode un+1u_{n+1} has the opposite parity. The symmetry broken at threshold is therefore parity, which unequal couplings preserve, and the bifurcation survives as a pitchfork, in which the odd polarization that grows at threshold, with its 2n+12n+1 walls, adds to the even polarization of the deformed branch and sets the sign pattern once it dominates.

Exchanging the components maps ζ→−ζ\zeta\to-\zeta, and the threshold shift is therefore even in ζ\zeta. Figure 9(a) confirms the quadratic dependence at g=1g = 1, g12∗(ζ)−g12∗(0)≃0.104ζ2g_{12}^{*}(\zeta) - g_{12}^{*}(0) \simeq0.104\zeta^2 for n=0n = 0 in oscillator units, with a coefficient that falls with nn. The shift is controlled by the ratio of ζ\zeta to the threshold shift δg∗\delta g^* rather than by ζ/g\zeta/g, and at g=16g = 16, where δg∗\delta g^* is about half its value at g=1g = 1, its coefficient for n=0n = 0 is about twice as large.

Figure 9. Second-order threshold shifts at g=1g = 1. (a) Unequal intraspecies couplings g11,22=g±ζg_{11,22} = g \pm\zeta; dotted lines mark the 87Rb{}^{87}\mathrm{Rb} pairs ∣1,−1⟩\lvert1,-1\rangle, ∣2,−2⟩\lvert2,-2\rangle (left) and ∣1,+1⟩\lvert1,+1\rangle, ∣2,−1⟩\lvert2,-1\rangle (right). (b) Population imbalance N1,2=N(1±ν)N_{1,2} = N(1 \pm\nu). Lines are proportional to ζ2\zeta^2 and ν2\nu^2.

For a uniform mixture Eq. (16) places the demixing at the geometric mean of the intraspecies couplings, which differs from the arithmetic mean at second order, g11g22=g−ζ2/2g+O(ζ4)\sqrt{g_{11}g_{22}} = g - \zeta^2/2g + O(\zeta^4), and at that order gives the wrong sign in the trap; substituting it for gg in the symmetric threshold predicts that the n=0n = 0 threshold at g=1g = 1 and ζ=0.20\zeta= 0.20 falls by 0.016, whereas the computed threshold rises by 0.004. The shift originates instead in the deformation of the unpolarized branch, which neither mean contains. The quadratic law is an expansion in ζ/δg∗\zeta/\delta g^*, which in the TF regime equals (ζ/g)(μ/ℏω)2/1.822(\zeta/g)(\mu/\hbar\omega)^2/1.822 and reaches unity at μ≃8ℏω\mu\simeq8\hbar\omega for the pair ∣1,+1⟩\lvert1,+1\rangle, ∣2,−1⟩\lvert2,-1\rangle and at

the pair ∣1,+1⟩\lvert1,+1\rangle, ∣2,−1⟩\lvert2,-1\rangle and at μ≃16ℏω\mu\simeq16\hbar\omega for ∣1,−1⟩\lvert1,-1\rangle, ∣2,−2⟩\lvert2,-2\rangle; for the former it reaches about 0.65 at g=16g = 16, so that for this pair the shift has to be computed directly; when ζ\zeta exceeds δg∗\delta g^*, as in large condensates, the deformation of the unpolarized branch is no longer small, the TF profiles of the two components acquire different radii, whose edges require a treatment beyond the TF approximation [48], and in the limit of a uniform mixture the geometric-mean criterion is recovered.

A population imbalance N1,2=N(1±ν)N_{1,2} = N(1 \pm\nu), with the couplings defined with the mean number NN and the components normalized to 1±ν1 \pm\nu, enters in the same way, breaking the exchange of the components but not parity, and Figure 9(b) shows the quadratic dependence of the threshold, with coefficients that exceed those of ζ2\zeta^2 and grow with nn; an imbalance below 1% still shifts the threshold by less than 1.2×10−41.2 \times10^{-4} for n≤2n \le2. Since the exchange of the components reverses both ζ\zeta and ν\nu, a term linear in ν\nu is permitted once ζ≠0\zeta\ne0.

Realization with 87Rb{}^{87}\mathrm{Rb}

The hyperfine pair ∣F,mF⟩=∣1,+1⟩\lvert F,m_F\rangle= \lvert1,+1\rangle and ∣2,−1⟩\lvert2,-1\rangle of 87Rb{}^{87}\mathrm{Rb} combines nearly equal intraspecies couplings with a tunable interspecies scattering length. Its intraspecies scattering lengths are a11=100.4 a0a_{11} = 100.4\,a_0 and a22=95.00 a0a_{22} = 95.00\,a_0, and the interspecies one, abg=97.66 a0a_{\mathrm{bg}} = 97.66\,a_0 away from resonance, nearly equals their mean aˉ=97.70 a0\bar{a} = 97.70\,a_0 and is tuned by a mixed-channel Feshbach resonance near 9.10 G, near which values of Δa=a12−abg\Delta a = a_{12} - a_{\mathrm{bg}} between −6.5-6.5 and +8.0 a0+8.0\,a_0 have been reported [6], and an elongated mixture of this pair has been demixed near this resonance [7]. Since the dependence of the threshold on nn is appreciable only when μ\mu is a few ℏω\hbar\omega, we take ω=2π×30\omega= 2\pi\times30 Hz, ω⊥=2π×600\omega_\perp= 2\pi\times600 Hz and N=152N = 152 atoms per component, which gives g=16g = 16 for the mean scattering length aˉ\bar{a}.

At μ≃ℏω⊥/3\mu\simeq\hbar\omega_\perp/3 the transverse width of the cloud follows the local density, and we compute the states with the nonpolynomial Schrödinger equation (NPSE) [49], in which the transverse profile is a Gaussian whose width minimizes the local energy. The NPSE lowers the chemical potential of the unpolarized ground state at the SU(2) point from 6.64 to 6.25 ℏω\hbar\omega, and raises the threshold detunings of the 1D model by 19% and the steps between them by about 20%; axisymmetric 3D GP calculations place the thresholds of n=0n = 0, 1 and 2 higher than the NPSE by 0.11 a0a_0 and reproduce the steps between them to within 0.002 a0a_0.

Table III lists the thresholds, computed with the unequal couplings and expressed as Δa∗=aˉg12∗/g−abg\Delta a^* = \bar{a}g_{12}^*/g - a_{\mathrm{bg}}, together with the amount Δaξ∗\Delta a_\xi^* by which the unequal couplings raise them above the thresholds for equal intraspecies couplings; the ground state demixes at Δa∗=5.17a0\Delta a^* = 5.17a_0, and successive levels follow in steps of 0.39(1)a00.39(1)a_0, within the reported range of Δa\Delta a. The steps are uniform because Δaξ∗\Delta a_\xi^* falls with nn, and an imbalance ν=±0.05\nu= \pm0.05 changes Δaξ∗\Delta a_\xi^* by less than 0.05a00.05a_0. The unpolarized state can be prepared from a single-component condensate in its nnth stationary state, a dark soliton or a train of dark solitons imprinted at their stationary positions [4, 5], by a π/2\pi/2 pulse at the target field [50], and the polarization can be read out by state-selective imaging; for n=0n=0 the centers of mass of the two components separate by 3.4 μm3.4\,\mu\mathrm{m} at 2a02a_0 above threshold. At this atom number quantum fluctuations of the soft mode round the pitchfork; treating η\eta and the projection χ\chi of the relative phase on un+1u_{n+1} as conjugate variables, [η,χ]=i/4N[\eta,\chi]=i/4N, with the phase stiffness 1/κn1/\kappa_n of Eq. (15), gives an rms amplitude of 0.0640.064 at threshold, which the mean-field amplitude reaches 0.070.07 to 0.09a00.09a_0 above threshold, about a fifth of a step. The unequal couplings polarize the deformed branch itself, with an even pattern that changes sign twice, and the 2n+12n+1 walls set the sign pattern once the odd polarization dominates, from 0.120.12, 0.080.08 and 0.06a00.06a_0 above threshold for n=0n=0, 11 and 22.

nnΔa∗\Delta a^* (a0a_0)Δac∗\Delta a_c^* (a0a_0)fsf_s (Hz)Γ1\Gamma_1 (s−1\mathrm{s}^{-1})Γ2\Gamma_2 (s−1\mathrm{s}^{-1})
05.170.195.61623
15.570.116.41725
25.960.077.01826
36.350.047.61927
46.740.028.12028

Table III. Predictions for the 87Rb{}^{87}\mathrm{Rb} pair ∣1,+1⟩\lvert1,+1\rangle, ∣2,−1⟩\lvert2,-1\rangle with N=152N=152 atoms per component, ω=2π×30\omega=2\pi\times30 Hz and ω⊥=2π×600\omega_\perp=2\pi\times600 Hz (q=16q=16), computed with the NPSE. Threshold detuning Δa∗\Delta a^*, its rare Δac∗\Delta a_c^* above the value for equal intraspecies couplings, frequency fsf_s of the soft polarization mode at a12=aˉa_{12}=\bar{a}, and growth rates Γ1\Gamma_1 and Γ2\Gamma_2 at 1a01a_0 and 2a02a_0 above threshold, the last three computed with equal intraspecies couplings.

Two features of this resonance make resolving successive levels demanding. A step corresponds to a few mG of field, larger than the field noise below 11 mG reached with low-noise current sources [51], but the scattering lengths determined near the resonance carry a typical accuracy of ±1a0\pm1a_0 [6], so that the absolute thresholds are known only to a few steps, whereas the differences between levels do not depend on this calibration; at a fixed field the levels can also be compared through their critical atom numbers, which differ by 4%4\% to 6%6\%, since g12/gg_{12}/g does not depend on NN whereas the trap correction does. Two-body loss near the resonance, with rate coefficients of 11 to 3×10−13 cm3/s3\times10^{-13}\,\mathrm{cm}^3/\mathrm{s} [6], removes 8%8\% to 29%29\% of the atoms in 100 ms100\,\mathrm{ms}, the time over which the polarization grows at the rates of Table III, and raises the thresholds by a few steps during the hold, so that the separation of the onsets of successive levels has to be compared with calculations that include the measured loss.

The cost of the polarization can also be measured away from the resonance and its loss; at a12=aˉa_{12}=\bar{a} the soft-mode frequency fsf_s is the orbital spacing λn+1(0)/h\lambda_{n+1}^{(0)}/h of Eq. (8), and its spectroscopy, as for the spin-dipole mode of a sodium mixture [39], would test the growth of this factor with nn. The ground state and the dark–dark soliton, whose preparation involves at most one notch at the trap center, are the natural first targets.

Conclusions

The unpolarized excited states of a trapped binary condensate demix at an interspecies repulsion that rises with the excitation level, set by the balance between the single-particle cost of separating the components at fixed atom numbers and the interaction energy that the separation releases; evaluated at the SU(2) point, this balance factorizes the threshold into the energy of the polarization mode and its overlap with the condensate. In the quasi-1D TF limit the trap correction to the ground-state threshold approaches 1.822(ℏω/μ)21.822(\hbar\omega/\mu)^2, and each dark notch near the trap center enlarges it by the fraction 0.678 ℏω/μ0.678\,\hbar\omega/\mu of its value, the net effect of the segment that the notch removes from the polarization mode and of the rise of the chemical potential by the atoms that it expels; at μ\mu of a few ℏω\hbar\omega the computed thresholds give the steps, and the laws their order of magnitude.

The threshold is the onset of an exponential, non-oscillating growth of the polarization, distinct from the oscillatory instabilities that these states can carry, and in every case computed the polarized state emerges through a supercritical pitchfork with 2n+12n+1 domain walls, nn of which sit at the density nodes of the unpolarized state. Unequal intraspecies couplings and a population imbalance preserve parity and shift the threshold at second order, apart from a term linear in the imbalance once the couplings differ; for the proposed 87Rb{}^{87}\mathrm{Rb} pair the shift from the unequal couplings amounts to half a step for the ground state and makes the steps between levels uniform. In the hyperfine mixture ∣1,+1⟩\lvert1,+1\rangle, ∣2,−1⟩\lvert2,-1\rangle of 87Rb{}^{87}\mathrm{Rb} the thresholds of the five lowest states lie within the range of interspecies scattering lengths reported near the resonance, in steps of about 0.4a00.4a_0, and the ground state and the dark–dark soliton are the natural first targets. In 2D and 3D the nodal points of ϕn\phi_n become nodal lines and surfaces and the counting that yields 2n+12n+1 walls changes, whereas the factorization of the threshold into an excitation energy and an overlap is not tied to 1D.

Acknowledgments

This work was partially supported by the Hebei Natural Science Foundation (A2026502021), the Fundamental Research Funds for the Central Universities (2026MS219), and the Research and Practice Project of Higher Education Teaching Reform of Hebei Province (Nos. 2025GJJG412 and 2026GJJG415).

DATA AVAILABILITY

The data that support the findings of this article are openly available [52].

Expansion coefficients

Multiplying Eq. (2) by ϕn\phi_n and integrating gives

Tϕ+Vϕ+geffP=μn,(17)T_{\phi}+V_{\phi}+g_{\mathrm{eff}}P=\mu_n, \tag*{(17)}

with Tϕ=(ℏ2/2m)∫(ϕn′)2 dxT_{\phi}=(\hbar^2/2m)\int(\phi'_n)^2\,\mathrm{d}x, Vϕ=∫Vϕn2 dxV_{\phi}=\int V\phi_n^2\,\mathrm{d}x and P=∫ϕn4 dxP=\int\phi_n^4\,\mathrm{d}x. Inserting Eq. (4) into Eq. (1), the terms linear in ε\varepsilon cancel between the two components, and with Kϕ=Tϕ+VϕK_{\phi}=T_{\phi}+V_{\phi}, Ku=Tu+VuK_u=T_u+V_u and U4=∫u4 dxU_4=\int u^4\,\mathrm{d}x the energy along the path is

E(ε)=2(Kϕ+ε2Ku)1+ε2+geffP+(6g−2g12)ε2M+geffε4U4(1+ε2)2,(18)E(\varepsilon)=\frac{2(K_{\phi}+\varepsilon^2K_u)}{1+\varepsilon^2}+\frac{g_{\mathrm{eff}}P+(6g-2g_{12})\varepsilon^2M+g_{\mathrm{eff}}\varepsilon^4U_4}{(1+\varepsilon^2)^2}, \tag*{(18)}

exactly. The intraspecies term contributes g(P+6ε2M+ε4U4)g(P+6\varepsilon^2M+\varepsilon^4U_4) to the numerator of the second term, since (ϕn+εu)4+(ϕn−εu)4=2ϕn4+12ε2ϕn2u2+2ε4u4(\phi_n+\varepsilon u)^4+(\phi_n-\varepsilon u)^4=2\phi_n^4+12\varepsilon^2\phi_n^2u^2+2\varepsilon^4u^4, and the interspecies term contributes g12(P−2ε2M+ε4U4)g_{12}(P-2\varepsilon^2M+\varepsilon^4U_4), since (ϕn+εu)2(ϕn−εu)2=(ϕn2−ε2u2)2(\phi_n+\varepsilon u)^2(\phi_n-\varepsilon u)^2=(\phi_n^2-\varepsilon^2u^2)^2. Expanding the denominators gives E0=2Kϕ+geffPE_0=2K_{\phi}+g_{\mathrm{eff}}P and

α=2(Ku−Kϕ)+(6g−2g12)M−2geffP,(19)\alpha=2(K_u-K_{\phi})+(6g-2g_{12})M-2g_{\mathrm{eff}}P, \tag*{(19)}

and eliminating Kϕ+geffPK_{\phi}+g_{\mathrm{eff}}P with Eq. (17) gives Eq. (5).

Constrained eigenvalue problem

Introducing the multipliers λ\lambda and 2c2c for the normalization and orthogonality conditions on uu, we make stationary

L[u]=∫uA^u dx−λ(∫u2 dx−1)−2c∫ϕnu dx.(20)\mathcal{L}[u]=\int u\hat{A}u\,\mathrm{d}x-\lambda\left(\int u^2\,\mathrm{d}x-1\right)-2c\int\phi_nu\,\mathrm{d}x. \tag*{(20)}

Under u→u+δuu\to u+\delta u, the self-adjointness of A^\hat{A} gives the first variation δL=2∫δu(A^u−λu−cϕn) dx\delta\mathcal{L}=2\int\delta u(\hat{A}u-\lambda u-c\phi_n)\,\mathrm{d}x, which vanishes for arbitrary δu\delta u when

A^u=λu+cϕn,∫ϕnu dx=0.(21)\hat{A}u=\lambda u+c\phi_n,\qquad\int\phi_nu\,\mathrm{d}x=0. \tag*{(21)}

Multiplying Eq. (21) by uu and integrating gives λ=λ[u]\lambda=\lambda[u], and multiplying it by ϕn\phi_n gives c=∫uA^ϕn dxc=\int u\hat{A}\phi_n\,\mathrm{d}x. With the projector Q^f=f−ϕn∫ϕnf dx\hat{Q}f=f-\phi_n\int\phi_nf\,\mathrm{d}x, Eq. (21) is equivalent to Q^A^Q^u=λu\hat{Q}\hat{A}\hat{Q}u=\lambda u with Q^u=u\hat{Q}u=u, the eigenvalue problem of A^\hat{A} restricted to the complement of ϕn\phi_n. Its eigenfunctions uku_k are mutually orthogonal and orthogonal to ϕn\phi_n, and since A^uk=λkuk+ckϕn\hat{A}u_k=\lambda_ku_k+c_k\phi_n, one has ∫ujA^uk dx=λkδjk\int u_j\hat{A}u_k\,\mathrm{d}x=\lambda_k\delta_{jk}, which makes the quadratic energy diagonal, as § 2 C uses.

Away from the SU(2) point ϕn\phi_n is not an eigenfunction of A^\hat{A}, since Eq. (2) gives A^ϕn=−2δg ϕn3\hat{A}\phi_n=-2\delta g\,\phi_n^3, and the constrained spectrum differs from the unconstrained one. Because ϕn3\phi_n^3 has the parity of ϕn\phi_n, the multiplier c=−2δg∫ϕn3u dxc=-2\delta g\int\phi_n^3u\,\mathrm{d}x vanishes for every uu of the opposite parity, and the threshold mode un+1u_{n+1}, which has n+1n+1 nodes and this opposite parity, is an eigenfunction of A^\hat{A} itself at every g12g_{12}, with the same eigenvalue in the constrained and unconstrained spectra. The unconstrained spectrum contains one additional branch, continuously connected to ϕn\phi_n and nearly parallel to it, whose eigenvalue, ≃−2δg∫ϕn4 dx\simeq-2\delta g\int\phi_n^4\,\mathrm{d}x for small ∣δg∣|\delta g|, passes through zero at g12=gg_{12}=g; it is the uniform change of composition that the conservation of N1N_1 and N2N_2 excludes. The two counts are related through Q=∫ϕnA^−1ϕn dxQ=\int\phi_n\hat{A}^{-1}\phi_n\,\mathrm{d}x [40, 41], since the constrained count N(−A^)\mathcal{N}(-\hat{A}) equals the number of negative eigenvalues of A^\hat{A} on the whole space when Q>0Q>0 and falls short of it by one when Q<0Q<0. Below the SU(2) point Q>0Q>0 and the two counts coincide, whereas at g12=gg_{12}=g the operator A^\hat{A} annihilates ϕn\phi_n and QQ passes through a pole, and above this point Q<0Q<0 removes the additional branch from the count. A constrained eigenvalue with the parity of ϕn\phi_n can cross zero only where QQ does, since the zero mode u=cA^−1ϕnu=c\hat{A}^{-1}\phi_n must then be orthogonal to ϕn\phi_n; the computed count stays nn between the SU(2) point and the threshold, so that QQ stays negative there, and at the threshold QQ lies between −1.01-1.01 and −0.95-0.95 at g=1g=1 and between −4.3-4.3 and −3.6-3.6 at g=16g=16 for n=0n=0, 1, 2 and 4. The constrained count therefore remains nn across the SU(2) point and increases at g12∗g_{12}^*, where an eigenvalue of the opposite parity crosses zero.

Thomas–Fermi limit

In the TF limit ϕ02=(μ−V)/geff\phi_0^2=(\mu-V)/g_{\mathrm{eff}} inside the cloud, which ends at the TF radius RTF=(2μ/mω2)1/2R_{\mathrm{TF}}=(2\mu/m\omega^2)^{1/2}. Writing u=ϕ0yu=\phi_0y and using B^ϕ0=−(ℏ2/2m)(ϕ02y′)′/ϕ0\hat{B}\phi_0=-(\hbar^2/2m)(\phi_0^2y')'/\phi_0, and with x~=x/RTF\tilde{x}=x/R_{\mathrm{TF}} the eigenvalue problem B^u=λu\hat{B}u=\lambda u becomes

−ddx~[(1−x~2)dydx~]=Λ(1−x~2)y,Λ=4μλ(ℏω)2,(22)-\frac{\mathrm{d}}{\mathrm{d}\tilde{x}}\left[(1-\tilde{x}^2)\frac{\mathrm{d}y}{\mathrm{d}\tilde{x}}\right]=\Lambda(1-\tilde{x}^2)y,\qquad\Lambda=\frac{4\mu\lambda}{(\hbar\omega)^2}, \tag*{(22)}

with yy bounded at x~=±1\tilde{x}=\pm1. Its lowest solution, y=1y=1 with Λ=0\Lambda=0, is ϕ0\phi_0 itself, and the lowest odd solution, with Λ=4.622\Lambda=4.622, is the spin-dipole mode, whose energy at the symmetric point is λ1(0)=Λ(ℏω)2/4μ\lambda_1^{(0)}=\Lambda(\hbar\omega)^2/4\mu.

Odd modes are eigenfunctions of A^\hat{A} itself, as Appendix B shows, and for them the threshold condition A^u=0\hat{A}u=0 becomes

−ddx~[(1−x~2)dydx~]=σ(1−x~2)2y,σ=8μ2δggeff(ℏω)2.(23)-\frac{\mathrm{d}}{\mathrm{d}\tilde{x}}\left[(1-\tilde{x}^2)\frac{\mathrm{d}y}{\mathrm{d}\tilde{x}}\right]=\sigma(1-\tilde{x}^2)^2y,\qquad\sigma=\frac{8\mu^2\delta g}{g_{\mathrm{eff}}(\hbar\omega)^2}. \tag*{(23)}

Its lowest odd solution has σ∗=7.286\sigma^{*}=7.286, and with geff=2gg_{\mathrm{eff}}=2g the threshold δg∗/g=(σ∗/4)(ℏω/μ)2\delta g^{*}/g=(\sigma^{*}/4)(\hbar\omega/\mu)^2 gives Eq. (10). The estimate of Eq. (8) amounts to evaluating the Rayleigh quotient ∫(1−x~2)(dy/dx~)2 dx~/∫(1−x~2)y2 dx~\int(1-\tilde{x}^{2})(dy/d\tilde{x})^{2}\,d\tilde{x}/\int(1-\tilde{x}^{2})y^{2}\,d\tilde{x} of Eq. (23) with the spin-dipole mode, which gives 7.454 instead of σ∗\sigma^{*} and the coefficient 1.863. With μ\mu the TF chemical potential at geff=2gg_{\mathrm{eff}}=2g, the computed coefficient (g12(0)/g−1)(μ/ℏω)2(g_{12}^{(0)}/g-1)(\mu/\hbar\omega)^2 exceeds 1.822 by at most 3% for g≥4g\geq4, peaks at g=8g=8 and approaches it from above for larger gg, to within 0.2% at g=256g=256, with a deviation that falls faster than ℏω/μ\hbar\omega/\mu and slower than (ℏω/μ)2(\hbar\omega/\mu)^2 over this range; at smaller gg, where μ\mu approaches ℏω\hbar\omega and the cloud departs from the TF profile, Eq. (10) overestimates the correction.

The notches of the excited states add structure on the scale of the healing length ξ=ℏ/(2mμ)1/2\xi=\hbar/(2m\mu)^{1/2}. Near a notch at xℓx_{\ell} the density is ϕn2≃ϕ02(xℓ)tanh⁡2[(x−xℓ)/ξℓ]\phi_n^2\simeq\phi_0^2(x_{\ell})\tanh^2[(x-x_{\ell})/\xi_{\ell}], with ξℓ=2ξ(1−x~ℓ2)−1/2\xi_{\ell}=\sqrt{2}\xi(1-\tilde{x}_{\ell}^{2})^{-1/2} the local soliton width, and with u=ϕnyu=\phi_n y the threshold condition A^u=0\hat{A}u=0 reads (ϕn2y′)′=−(4m δg/ℏ2)ϕn4y(\phi_n^2y')'=-(4m\,\delta g/\hbar^2)\phi_n^4y, whose right side is negligible within a few ξℓ\xi_{\ell} of the notch. The inner solution is therefore y=c1(t−coth⁡t)+c2y=c_1(t-\coth t)+c_2, with t=(x−xℓ)/ξℓt=(x-x_{\ell})/\xi_{\ell}, which far from the notch behaves as c1(t∓1)+c2c_1(t\mp1)+c_2 for t→±∞t\to\pm\infty. On the scale of RTFR_{\mathrm{TF}} the outer solution of Eq. (23) thus keeps its slope across the notch and jumps by [y]=−2ξℓy′(xℓ)[y]=-2\xi_{\ell}y'(x_{\ell}), the jump produced by excising a segment of length 2ξℓ2\xi_{\ell} from the cloud. For the central notch of an odd state c2=0c_2=0 by parity, the outer solution vanishes at x=±ξℓx=\pm\xi_{\ell}, and the inner solution changes sign at t=coth⁡tt=\coth t, at x=1.697ξx=1.697\xi, so that between the notch and this node the polarization is reversed; Figure 4(a) shows that at g≥64g\geq64 the computed pattern ϕ1u2\phi_1u_2 of the dark–dark soliton coincides with the inner solution, and its node stays at 1.70ξ1.70\xi for gg between 16 and 512. For an off-center notch c2≠0c_2\ne0, and the node moves toward the notch, to between 0.350.35 and 1.35ξ1.35\xi for n=2n=2 and 4.

Multiplying Eq. (23) by its lowest odd solution y0y_0 without notches and integrating by parts on both sides of the notch shows that the jump shifts the eigenvalue by

δσσ∗=2ξℓRTF(1−x~ℓ2)y0′(x~ℓ)2∫(1−x~2)y0′2 dx~(24)\frac{\delta\sigma}{\sigma^{*}}=\frac{2\xi_{\ell}}{R_{\mathrm{TF}}}\frac{(1-\tilde{x}_{\ell}^{2})y_0'(\tilde{x}_{\ell})^2}{\int(1-\tilde{x}^{2})y_0'^2\,d\tilde{x}} \tag*{(24)}

to first order in ξℓ/RTF\xi_{\ell}/R_{\mathrm{TF}}, where the prime denotes the derivative with respect to x~\tilde{x}. For a notch off the center the inner region also lacks part of the right side of the threshold condition, since ϕn4≃ϕ04(xℓ)tanh⁡4t\phi_n^4\simeq\phi_0^4(x_{\ell})\tanh^4t there, and the missing weight, ∫(1−tanh⁡4t) dt=8/3\int(1-\tanh^4t)\,dt=8/3, adds

δσσ∗=8ξℓ3RTFσ∗(1−x~ℓ2)2y0(x~ℓ)2∫(1−x~2)y0′2 dx~,(25)\frac{\delta\sigma}{\sigma^{*}}=\frac{8\xi_{\ell}}{3R_{\mathrm{TF}}}\frac{\sigma^{*}(1-\tilde{x}_{\ell}^{2})^2y_0(\tilde{x}_{\ell})^2}{\int(1-\tilde{x}^{2})y_0'^2\,d\tilde{x}}, \tag*{(25)}

which vanishes at the center, where y0=0y_0=0; elsewhere the notch removes density while the polarization mode has a finite amplitude, and with it part of the interaction energy that the polarization releases.

For a central notch Eq. (24) gives 2Jξℓ/RTF=2Jℏω/μ2J\xi_{\ell}/R_{\mathrm{TF}}=\sqrt{2}J\hbar\omega/\mu, with J=y0′(0)2/∫(1−x~2)y0′2 dx~=1.4795J=y_0'(0)^2/\int(1-\tilde{x}^{2})y_0'^2\,d\tilde{x}=1.4795, that is 2.092 ℏω/μ\hbar\omega/\mu, the result of the Hadamard formula for a Dirichlet boundary moved to +ξℓ+\xi_{\ell}. The atoms missing from the notch, 2ξℓϕ02(xℓ)2\xi_{\ell}\phi_0^2(x_{\ell}) of the normalized density, raise μ\mu at fixed atom number by (1−x~ℓ2)1/2ℏω/2(1-\tilde{x}_{\ell}^2)^{1/2}\hbar\omega/\sqrt{2}, as the computed μ1−μ0=0.71ℏω\mu_1-\mu_0=0.71\hbar\omega at g=256g=256 and 512 confirms for a central notch, and since δg=σgeff(ℏω/μ)2/8μ\delta g=\sigma g_{\mathrm{eff}}(\hbar\omega/\mu)^2/8\mu at fixed σ\sigma, they lower the threshold by 2(1−x~ℓ2)1/2ℏω/μ\sqrt{2}(1-\tilde{x}_{\ell}^2)^{1/2}\hbar\omega/\mu. A central notch therefore contributes 2(J−1)=0.678\sqrt{2}(J-1)=0.678 times ℏω/μ\hbar\omega/\mu to the relative threshold shift, the coefficient of Eq. (11), and a notch at ∣x~ℓ∣=0.1|\tilde{x}_{\ell}|=0.1, 0.2 and 0.3 contributes 0.76, 0.98 and 1.23 times ℏω/μ\hbar\omega/\mu, so that a notch displaced from the center raises the threshold.

For the stationary trains x~ℓ\tilde{x}_{\ell} falls only as (ℏω/μ)ln⁡(μ/ℏω)(\hbar\omega/\mu)\ln(\mu/\hbar\omega), and these position effects, of order x~ℓ2\tilde{x}_{\ell}^{2}, carry logarithmic factors that are not small at the couplings computed, so that the approach of the computed shifts per notch to 0.678, shown in Figure 4(b), rests on extrapolation. Extrapolated quadratically in ℏω/μ\hbar\omega/\mu from g=32g=32 to 512, the computed ratio [δg∗(n)/g−δg∗(0)/g]/nℏω[\delta g^{*}(n)/g-\delta g^{*}(0)/g]/n\hbar\omega approaches 0.678 for n=1n=1 and 2 and between 0.676 and 0.677 for n=4n=4, drifting toward 0.678 as the fit window moves to larger gg. The same fits give the next order, a reduction of the ratio by 0.33, 0.60 and 1.0 times ℏω/μ\hbar\omega/\mu for n=1n=1, 2 and 4, between 0.25nn and 0.33nn, which makes the step between successive levels fall with nn and brings Eq. (11) to within 3% of the computed ratios for n≤2n\leq2 at g=16g=16. Adding to the fits a term (ℏω/μ)2ln⁡(ℏω/μ)(\hbar\omega/\mu)^2\ln(\hbar\omega/\mu) changes these coefficients by up to 17%, and a term (ℏω/μ)ln⁡(ℏω/μ)(\hbar\omega/\mu)\ln(\hbar\omega/\mu) instead leaves the limits between 0.677 and 0.681; the limit is thus robust, whereas the coefficients of the next order are empirical. Adjacent notches of the trains lie between 4 and 5 healing lengths apart at g=16g=16 and between 6 and 7 at g=512g=512, a spacing that grows only logarithmically with RTF/ξR_{\mathrm{TF}}/\xi, so that at g≤64g\leq64 their inner regions overlap and the expansion for separate notches does not apply.

Numerical methods

Stationary states are computed on a Fourier spectral grid by a bordered Newton method with damped line search, with the chemical potentials as unknowns alongside the wave functions, and satisfy the stationary equations to 10−1310^{-13} in the maximum norm; the box has a length of six times the larger of the linear turning point and the TF radius, and at least 12 ahoa_{\mathrm{ho}}, the grid spacing is at most a twentieth of the box divided by n+1n+1 and a quarter of the healing length, with 160 to 640 points, and the boundary densities are below 10−2010^{-20}. Distances quoted in healing lengths for computed states use the chemical potential μn\mu_n of each state in place of the TF value. The constrained spectrum of A^\hat{A} is obtained by projecting orthogonally to ϕn\phi_n before diagonalizing, and the threshold is located as the smallest g12g_{12} at which the number of negative eigenvalues exceeds nn, by a coarse warm-started sweep followed by bisection, to 10−810^{-8} or better. For every entry of Table 1 the count, evaluated at 41 couplings between g12=0g_{12}=0 and 0.999g12∗0.999g_{12}^{*} with ϕn\phi_n continued in geffg_{\mathrm{eff}}, equals nn at each of them. For unequal couplings the same counting is applied to the Hessian of the two-component energy on the deformed branch, constrained by both normalizations, whose number of negative eigenvalues rises from 2n2n to 2n+12n+1 at the threshold, also for the couplings of the 87Rb{}^{87}\mathrm{Rb} pair, where ζ/δg∗≃0.65\zeta/\delta g^{*}\simeq0.65.

For Appendix C the thresholds at g=32g=32 to 512512 are located by a secant iteration on λn+1(g12)\lambda_{n+1}(g_{12}), with the state followed in geffg_{\mathrm{eff}} from g=16g=16, on grids with a spacing of a third to a quarter of the healing length, and a grid with half again as many points changes them by less than 10−910^{-9}. For § 8 the same methods are applied to the NPSE, in which the interaction energy density

W=12(g11∣ψ1∣4+g22∣ψ2∣4)+g12∣ψ1∣2∣ψ2∣2W=\frac{1}{2}\left(g_{11}|\psi_1|^4+g_{22}|\psi_2|^4\right)+g_{12}|\psi_1|^2|\psi_2|^2

of Eq. (1) is replaced by ℏω⊥ρ[(1+2W/ℏω⊥ρ)1/2−1]\hbar\omega_{\perp}\rho[(1+2W/\hbar\omega_{\perp}\rho)^{1/2}-1], and fsf_s, Γ1\Gamma_1 and Γ2\Gamma_2 are computed with equal intraspecies couplings. The 3D check solves the axisymmetric GP equation with ω⊥=20ω\omega_{\perp}=20\omega on the half space x≥0x\geq0, with cell-centred finite differences on 240×104240\times104 points over 6aho6a_{\mathrm{ho}} axially and 5.85.8 transverse oscillator lengths radially, and imposes the parity of ϕn\phi_n or of the threshold mode through the boundary condition at x=0x=0; a grid with half again as many points in each direction changes Δa∗\Delta a^{*} by less than 0.003a00.003a_0.

On the invariant subspace ψ2(x)=(−1)nψ1(−x)\psi_2(x)=(-1)^n\psi_1(-x), on which the two components share one chemical potential, the stationary equation for ψ1\psi_1 is solved together with this chemical potential and g12g_{12} at fixed projection η\eta, with un+1u_{n+1} the threshold mode at g12∗g_{12}^{*}, which keeps the continuation regular at the branch point. The branch is continued at 16 amplitudes logarithmically spaced between 0.0120.012 and 0.140.14, and beyond in steps of 0.010.01 until g12>2g12∗g_{12}>2g_{12}^{*}, on the grid used for the threshold, which is located there by a secant iteration on λn+1(g12)\lambda_{n+1}(g_{12}). Fitted over the eight smallest amplitudes, the exponents of η\eta and of the energy difference against g12−g12∗g_{12}-g_{12}^{*} differ from 1/21/2 and 22 by less than 10−310^{-3} at every point of Table 2, by terms of order η2\eta^2. Near threshold ∥ψ1−ψ2∥2≃4η2\|\psi_1-\psi_2\|^2\simeq4\eta^2 with η2=−λn+1/β\eta^2=-\lambda_{n+1}/\beta, and the branch lies below the unpolarized state by λn+12/β\lambda_{n+1}^2/\beta, which gives

β=−4λn+1∥ψ1−ψ2∥2=−λn+12Epol−Eunpol.(26)\beta=-\frac{4\lambda_{n+1}}{\|\psi_1-\psi_2\|^2}=-\frac{\lambda_{n+1}^2}{E_{\mathrm{pol}}-E_{\mathrm{unpol}}}. \tag*{(26)}

where λn+1\lambda_{n+1} is the constrained eigenvalue of the unpolarized state at the same g12g_{12}. The two expressions of Eq. (26), extrapolated linearly in η2\eta^2 over the eight smallest amplitudes, agree to within 2×10−62\times10^{-6}, and a grid with half again as many points changes β\beta by less than 2×10−52\times10^{-5}.

The Bogoliubov rates of Eq. (13) change by less than 3×10−5ω3\times10^{-5}\omega when the grid is refined to 512 points. For the polarized states the branches are continued in η\eta up to 2g12∗2g_{12}^{*}, and the linearization about (ψ1,ψ2)(\psi_1,\psi_2) gives ℏ2Ω2=eig⁡(L^−L^+)\hbar^2\Omega^2=\operatorname{eig}(\hat{L}_{-}\hat{L}_{+}), where L^−\hat{L}_{-} is diagonal in the components, with entries −(ℏ2/2m)∂x2+V+gψj2+g12ψ3−j2−μ-(\hbar^2/2m)\partial_x^2+V+g\psi_j^2+g_{12}\psi_{3-j}^2-\mu, and L^+\hat{L}_{+} adds to the interaction matrix with entries 2gψj22g\psi_j^2 and 2g12ψ1ψ22g_{12}\psi_1\psi_2; rates below 10−4ω10^{-4}\omega come from the two phase zero modes and are discarded. The real-time integration uses a Strang-split Fourier propagator on 256 points over 16aho16a_{\mathrm{ho}} with a time step of 0.002/ω0.002/\omega, which over the full run conserves N1−N2N_1-N_2 to 2×10−132\times10^{-13}, and the growth rate fitted over several oscillations agrees with the linearized value to within 10−510^{-5}.

References

  1. [1]M. A. Hoefer, J. J. Chang, C. Hamner, and P. Engels, Dark-dark solitons and modulational instability in miscible two-component Bose-Einstein condensates, Phys. Rev. A 84, 041605(R) (2011).DOI
  2. [2]D. Yan, J. J. Chang, C. Hamner, M. Hoefer, P. G. Kevrekidis, P. Engels, V. Achilleos, D. J. Frantzeskakis, and J. Cuevas, Beating dark–dark solitons in Bose–Einstein condensates, J. Phys. B 45, 115301 (2012).DOI
  3. [3]I. Morera, A. Muñoz Mateo, A. Polls, and B. Juliá-Díaz, Dark-dark-soliton dynamics in two density-coupled Bose-Einstein condensates, Phys. Rev. A 97, 043621 (2018).DOI
  4. [4]C. Becker, S. Stellmer, P. Soltan-Panahi, S. Dönscher, M. Baumert, E.-M. Richter, J. Kronjäger, K. Bongs, and K. Sengstock, Oscillations and interactions of dark and dark–bright solitons in Bose–Einstein condensates, Nat. Phys. 4, 496 (2008).DOI
  5. [5]G. Theocharis, A. Weller, J. P. Ronzheimer, C. Gross, M. K. Oberthaler, P. G. Kevrekidis, and D. J. Frantzeskakis, Multiple atomic dark solitons in cigar-shaped Bose-Einstein condensates, Phys. Rev. A 81, 063604 (2010).DOI
  6. [6]S. Tojo, Y. Taguchi, Y. Masuyama, T. Hayashi, H. Saito, and T. Hirano, Controlling phase separation of binary Bose-Einstein condensates via mixed-spin-channel Feshbach resonance, Phys. Rev. A 82, 033609 (2010).DOI
  7. [7]E. Nicklas, H. Strobel, T. Zibold, C. Gross, B. A. Malomed, P. G. Kevrekidis, and M. K. Oberthaler, Rabi flopping induces spatial demixing dynamics, Phys. Rev. Lett. 107, 193001 (2011).DOI
  8. [8]T.-L. Ho and V. B. Shenoy, Binary mixtures of Bose condensates of alkali atoms, Phys. Rev. Lett. 77, 3276 (1996).pubmed.ncbi.nlm.nih.gov/10062180
  9. [9]E. Timmermans, Phase separation of Bose-Einstein condensates, Phys. Rev. Lett. 81, 5718 (1998).DOI
  10. [10]H. Pu and N. P. Bigelow, Properties of two-species Bose condensates, Phys. Rev. Lett. 80, 1130 (1998).DOI
  11. [11]P. Ao and S. T. Chui, Binary Bose-Einstein condensate mixtures in weakly and strongly segregated phases, Phys. Rev. A 58, 4836 (1998).DOI
  12. [12]Th. Busch and J. R. Anglin, Dark-bright solitons in inhomogeneous Bose-Einstein condensates, Phys. Rev. Lett. 87, 010401 (2001).pubmed.ncbi.nlm.nih.gov/11461449
  13. [13]C. Hamner, J. J. Chang, P. Engels, and M. A. Hoefer, Generation of dark-bright soliton trains in superfluid-superfluid counterflow, Phys. Rev. Lett. 106, 065302 (2011).pubmed.ncbi.nlm.nih.gov/21405475
  14. [14]S. Middelkamp, J. J. Chang, C. Hamner, R. Carretero-González, P. G. Kevrekidis, V. Achilleos, D. J. Frantzesakis, P. Schmelcher, and P. Engels, Dynamics of dark-bright solitons in cigar-shaped Bose-Einstein condensates, Phys. Lett. A 375, 642 (2011).
  15. [15]D. Yan, J. J. Chang, C. Hammer, P. G. Kevrekidis, P. Engels, V. Achilleos, D. J. Frantzeskakis, R. Carretero-González, and P. Schm elcher, Multiple dark-bright solitons in atomic Bose-Einstein condensates, Phys. Rev. A 84, 053630 (2011).DOI
  16. [16]G. C. Katsimiga, J. Stockhofe, P. G. Kevrekidis, and P. Schmelcher, Dark-bright soliton interactions beyond the integrable limit, Phys. Rev. A 95, 013621 (2017).
  17. [17]Superheavy dark-bright soliton as a signature of spatial symmetry breaking transition in harmonically trapped Bose mixtures, arXiv:2607.09064 (2026).
  18. [18]E. G. Charalampidis, W. Wang, P. G. Kevrekidis, D. J. Frantzeskakis, and J. Cuevas-Maraver, SO(2)-induced breathing patterns in multicomponent Bose-Einstein condensates, Phys. Rev. A 93, 063623 (2016).DOI
  19. [19]W. Wang, L.-C. Zhao, E. G. Charalampidis, and P. G. Kevrekidis, Dark–dark soliton breathing patterns in multi-component Bose–Einstein condensates, J. Phys. B 54, 055301 (2021).DOI
  20. [20]H. Xu, P. G. Kevrekidis, and T. Kapitula, Existence, stability, and dynamics of harmonically trapped one-dimensional multi-component solitary waves: The near-linear limit, J. Math. Phys. 58, 061901 (2017).DOI
  21. [21]I. Danaila, M. A. Khamehchi, V. Gokhroo, P. Engels, and P. G. Kevrekidis, Vector dark-antidark solitary waves in multicomponent Bose-Einstein condensates, Phys. Rev. A 94, 053617 (2016).DOI
  22. [22]G. C. Katsimiga, S. I. Mistakidis, T. M. Bersano, M. K. H. Ome, S. M. Mossman, K. Mukherjee, P. Schmelcher, P. Engels, and P. G. Kevrekidis, Observation and analysis of multiple dark-antidark solitons in two-component Bose-Einstein condensates, Phys. Rev. A 102, 023301 (2020).DOI
  23. [23]P. Öhberg and L. Santos, Dark solitons in a two-component Bose-Einstein condensate, Phys. Rev. Lett. 86, 2918 (2001).
  24. [24]A. Roy and D. Angom, Fluctuation- and interaction-induced instability of dark solitons in single and binary condensates, Phys. Rev. A 90, 023612 (2014).DOI
  25. [25]M. Haelterman and A. P. Sheppard, Bifurcations of the dark soliton and polarization domain walls in nonlinear dispersive media, Phys. Rev. E 49, 4512 (1994).pubmed.ncbi.nlm.nih.gov/9961747
  26. [26]A. P. Sheppard and Y. S. Kivshar, Polarized dark solitons in isotropic Kerr media, Phys. Rev. E 55, 4773 (1997).DOI
  27. [27]S. Coen and M. Haelterman, Domain wall solitons in binary mixtures of Bose-Einstein condensates, Phys. Rev. Lett. 87, 140401 (2001).pubmed.ncbi.nlm.nih.gov/11580632
  28. [28]B. D. Esry and C. H. Greene, Spontaneous spatial symmetry breaking in two-component Bose-Einstein condensates, Phys. Rev. A 59, 1457 (1999).DOI
  29. [29]M. Trippenbach, K. Goral, K. Rzazewski, B. A. Malomed, and Y. B. Band, Structure of binary Bose-Einstein condensates, J. Phys. B 33, 4017 (2000).DOI
  30. [30]F. Riboli and M. Modugno, Topology of the ground state of two interacting Bose-Einstein condensates, Phys. Rev. A 65, 063614 (2002).DOI
  31. [31]K. L. Lee, N. B. Jørgensen, I.-K. Liu, L. Wacker, J. J. Arlt, and N. P. Proukakis, Phase separation and dynamics of two-component Bose-Einstein condensates, Phys. Rev. A 94, 013602 (2016).
  32. [32]L. Wen, H. Guo, Y.-J. Wang, A.-Y. Hu, H. Saito, C.-Q. Dai, and X.-F. Zhang, Effects of atom numbers on the miscibility-immiscibility transition of a binary Bose-Einstein condensate, Phys. Rev. A 101, 033610 (2020).DOI
  33. [33]M. Pyzh and P. Schmelcher, Phase separation of a Bose-Bose mixture: Impact of the trap and particle-number imbalance, Phys. Rev. A 102, 023305 (2020).DOI
  34. [34]B. B. Baizakov, B. A. Malomed, and M. Salerno, Nonlinear management of the miscibility-immiscibility transition in binary Bose-Einstein condensates, Phys. Rev. E 112, 024204 (2025).arxiv.org/abs/2507.13683
  35. [35]L. Wen, W. M. Liu, Y. Cai, J. M. Zhang, and J. Hu, Controlling phase separation of a two-component Bose-Einstein condensate by confinement, Phys. Rev. A 85, 043602 (2012).DOI
  36. [36]R. N. Bisset, P. G. Kevrekidis, and C. Ticknor, Enhanced quantum spin fluctuations in a binary Bose-Einstein condensate, Phys. Rev. A 97, 023602 (2018).DOI
  37. [37]R. Navarro, R. Carretero-González, and P. G. Kevrekidis, Phase separation and dynamics of two-component Bose-Einstein condensates, Phys. Rev. A 80, 023613 (2009).DOI
  38. [38]C. Ticknor, Excitations of a trapped two-component Bose-Einstein condensate, Phys. Rev. A 88, 013623 (2013).DOI
  39. [39]T. Bienaimé, E. Fava, G. Colzi, C. Mordini, S. Serafini, C. Qu, S. Stringari, G. Lamporesi, and G. Ferrari, Spin-dipole oscillation and polarizability of a binary Bose-Einstein condensate near the miscible-immiscible phase transition, Phys. Rev. A 94, 063652 (2016).DOI
  40. [40]T. Kapitula, P. G. Kevrekidis, and B. Sandstede, Counting eigenvalues via the Krein signature in infinite-dimensional Hamiltonian systems, Physica D 195, 263 (2004).
  41. [41]D. E. Pelinovsky, Inertia law for spectral stability of solitary waves in coupled nonlinear Schrödinger equations, Proc. R. Soc. A 461, 783 (2005).DOI
  42. [42]M. A. García-March, B. Juliá-Díaz, G. E. Astrakharchik, T. Busch, J. Boronat, and A. Polls, Sharp crossover from composite fermionization to phase separation in microscopic mixtures of ultracold bosons, Phys. Rev. A 88, 063604 (2013).DOI
  43. [43]G. C. Katsimiga, J. Stockhofe, P. G. Kevrekidis, and P. Schmelcher, Stability and dynamics of dark-bright soliton bound states away from the integrable limit, Appl. Sci. 7, 388 (2017).arxiv.org/abs/1703.04721
  44. [44]E. T. Karamatskos, J. Stockhofe, P. G. Kevrekidis, and P. Schmelcher, Stability and tunneling dynamics of a dark-bright soliton pair in a harmonic trap, Phys. Rev. A 91, 046337 (2015).
  45. [45]P. Hartman, Ordinary Differential Equations, 2nd ed., Classics in Applied Mathematics, Vol. 38 (SIAM, Philadelphia, 2002).
  46. [46]B. Simon, Sturm oscillation and comparison theorems, in Sturm–Liouville Theory: Past and Present, edited by W. O. Amrein, A. M. Hinz, and D. P. Pearson (Birkhäuser, Basel, 2005) pp. 29–43.
  47. [47]J. Sabbatini, W. H. Zurek, and M. J. Davis, Phase separation and pattern formation in a binary Bose-Einstein condensate, Phys. Rev. Lett. 107, 230402 (2011).DOI
  48. [48]J. Polo, V. Ahufinger, P. Mason, S. Sridhar, T. P. Billam, and S. A. Gardiner, Analysis beyond the Thomas-Fermi approximation of the density profiles of a miscible two-component Bose-Einstein condensate, Phys. Rev. A 91, 053626 (2015).
  49. [49]L. Salasnich, A. Parola, and L. Reatto, Effective wave equations for the dynamics of cigar-shaped and disk-shaped Bose condensates, Phys. Rev. A 65, 043614 (2002).
  50. [50]I. Vidanović, N. J. van Druten, and M. Haque, Spin modulation instabilities and phase separation dynamics in trapped two-component Bose condensates, New J. Phys.DOI
  51. [51]M. Zhao, A. Restelli, J. Tao, Q. Liang, and I. B. Spielman, A 20 A bipolar current source with 140 μA noise over 100 kHz bandwidth, AIP Adv. 13, 065309 (2023).DOI
  52. [52]Z. Liu, S. Du, and M. Zhao, Data for “Polarization instability of unpolarized excited states in trapped two-component Bose–Einstein condensates”, Zenodo (2026).arxiv.org/abs/2611.01488

Paper details

vlnc:W_zyvvzqfr
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents