In this paper, we assess whether Physics-Informed Neural Networks (PINNs) can capture spontaneous symmetry breaking (SSB) in the stationary Gross–Pitaevskii equation for double-well potentials. For attractive interactions ([Formula: see text]), the PINN successfully captures the qualitative features of macroscopic quantum self-trapping, yielding a robust asymmetric population imbalance ([Formula: see text]). For repulsive interactions ([Formula: see text]), the unregularized network converges to spurious asymmetric solutions, but a continuous parity regularization term effectively restores the symmetric ground state ([Formula: see text]). Benchmarking against a spectral imaginary-time reference, the post-hoc chemical potential agrees to within 2.3% across the full interaction range and to better than 0.6% for [Formula: see text], with density errors of order [Formula: see text]–[Formula: see text]. We reproduce the two-mode model parameters ([Formula: see text], B, [Formula: see text]) using full GPE eigenstates and show that the asymmetric branch of the two-mode model, although formally present ([Formula: see text]), corresponds to higher-energy self-trapped states for repulsive interactions, explaining why the ground state remains symmetric for [Formula: see text].
In this paper, we assess whether Physics-Informed Neural Networks (PINNs) can capture spontaneous symmetry breaking (SSB) in the stationary Gross–Pitaevskii equation for double-well potentials. For attractive interactions ([Formula: see text]), the PINN successfully captures the qualitative features of macroscopic quantum self-trapping, yielding a robust asymmetric population imbalance ([Formula: see text]). For repulsive interactions ([Formula: see text]), the unregularized network converges to spurious asymmetric solutions, but a continuous parity regularization term effectively restores the symmetric ground state ([Formula: see text]). Benchmarking against a spectral imaginary-time reference, the post-hoc chemical potential agrees to within 2.3% across the full interaction range and to better than 0.6% for [Formula: see text], with density errors of order [Formula: see text]–[Formula: see text]. We reproduce the two-mode model parameters ([Formula: see text], B, [Formula: see text]) using full GPE eigenstates and show that the asymmetric branch of the two-mode model, although formally present ([Formula: see text]), corresponds to higher-energy self-trapped states for repulsive interactions, explaining why the ground state remains symmetric for [Formula: see text].
Abstract We show that both single-component and two-component Bose-Einstein condensates’ (BECs) ground states can be simulated by a deep convolutional neural network. We trained the neural network via inputting the parameters in the dimensionless Gross-Pitaevskii equation (GPE) and outputting the ground-state wave function. After the training, the neural network generates ground-state wave functions with high precision. We benchmark the neural network for either inputting different coupling strength in the GPE or inputting an arbitrary potential under the infinite double walls trapping potential, and it is found that the ground state wave function generated by the neural network gives the relative chemical potential error magnitude below 10−3. Furthermore, the neural network trained with random potentials shows prediction ability on other types of potentials. Therefore, the BEC ground states, which are continuous wave functions, can be represented by deep convolutional neural networks.
Dynamics of the repulsive Bose–Einstein condensate (BEC) in a double-well trap is explored within the 3D time-dependent Gross–Pitaevskii equation. The model avoids numerous common approximations (two-mode treatment, time-space factorization, fixed values of the chemical potential and barrier penetrability, etc) and thus may provide a realistic description of BEC dynamics, including both weak-coupling (sub-barrier) and strong-coupling (above-barrier) regimes and their crossover. The strong coupling regime is achieved by increasing the number N of BEC atoms and thus the chemical potential. The evolution with N of Josephson oscillations (JO) and macroscopic quantum self-trapping (MQST) is examined and the crucial impact of the BEC interaction is demonstrated. At weak coupling, the calculations well reproduce the JO/MQST experimental data. At strong coupling with a significant overlap of the left and right BECs, we observe a remarkable persistence of the Josephson-like dynamics: the JO and MQST converge to a high-frequency JO-like mode where both population imbalance and phase difference oscillate around the zero averages.
We study the spontaneous symmetry breaking (SSB) of a superfluid Bose-Fermi (BF) mixture in a double-well potential (DWP). The mixture is described by the Gross-Pitaevskii equation (GPE) for the bosons, coupled to an equation for the order parameter of the Fermi superfluid, which is derived from the respective density functional in the unitarity limit (a similar model applies to the BCS regime, too). Straightforward SSB in the degenerate Fermi gas loaded into a DWP is impossible, as it requires an attractive self-interaction, and the intrinsic nonlinearity in the Fermi gas is repulsive. Nonetheless, we demonstrate that the symmetry breaking is possible in the mixture with attraction between fermions and bosons, like 40K and 87Rb. Numerical results are represented by dependencies of asymmetry parameters for both components on particle numbers of the mixture, NF and NB, and by phase diagrams in the (NF,NB) plane, which displays regions of symmetric and asymmetric ground states. The dynamical picture of the SSB, induced by a gradual transformation of the single-well potential into the DWP, is reported too. An analytical approximation is proposed for the case when the GPE for the boson wave function may be treated by means of the Thomas-Fermi (TF) approximation. Under a special linear relationship between NF and NB, the TF approximation allows us to reduce the model to a single equation for the fermionic function, which includes competing repulsive and attractive nonlinear terms. The latter one directly displays the mechanism of the generation of the effective attraction in the Fermi superfluid, mediated by the bosonic component of the mixture.
Models of two-dimensional (2D) traps, with double-well structure in the third direction, for Bose-Einstein condensates are introduced with attractive or repulsive interactions between atoms. The models are based on systems of linearly coupled 2D Gross-Pitaevskii equations, where the coupling accounts for tunneling between the wells. Each well carries an optical lattice (OL) (stable 2D solitons cannot exist without OLs). The linear coupling splits each finite band gap in the spectrum of the single-component model into two subgaps. The main subject of the work is spontaneous symmetry breaking (SSB) in two-component 2D solitons and localized vortices (SSB was not considered before in 2D settings). Using variational approximation (VA) and numerical methods, we demonstrate that, in a system with attraction or repulsion, SSB occurs in families of symmetric or antisymmetric solitons (or vortices), respectively. The corresponding bifurcation destabilizes the original solution branch and gives rise to a stable branch of asymmetric solitons or vortices. The VA provides for an accurate description of the emerging branch of asymmetric solitons. In the model with attraction, all stable branches eventually terminate due to the onset of collapse. Stable asymmetric solitons in higher finite band gaps and vortices with a multiple topological charge are found too. The models also give rise to first examples of embedded solitons and embedded vortices (the states located inside Bloch bands) in two dimensions. In the linearly coupled system with opposite signs of the nonlinearity in the two cores, two distinct types of stable solitons and vortices are found, dominated by either the self-attractive component or the self-repulsive one. In the system with a mismatch between the two OLs, a pseudobifurcation is found: when the mismatch attains its largest value $(\ensuremath{\pi})$, the bifurcation does not happen, as branches of different solutions asymptotically approach each other, but fail to merge.
We present analytical stationary solutions for the Gross-Pitaevskii equation (GPE) of a Bose-Einstein condensate (BECs) trapped in a double-well potential. These solutions are compared to those described by [Mahmud et al., Phys. Rev. A 66, 063607 (2002)]. In particular, we provide further evidence that symmetry preserving stationary solutions can be reduced to the eigenstates of the corresponding linear Schr\"odinger equation. Moreover, we have found that the symmetry-breaking solutions can emerge not only from bifurcations, but also from isolated points in the chemical potential--nonlinear interaction diagram. We also have found that there are some moving nodes in the symmetry-breaking solutions.
We report results of an analysis of the spontaneous symmetry breaking (SSB) in a basic (actually, simplest) model that is capable of producing the SSB phenomenology in a one-dimensional setting. It is based on the Gross-Pitaevskii-nonlinear Schrödinger equation with the cubic self-attractive term and a double-well potential built as an infinitely deep potential box split by a narrow (δ functional) barrier. The barrier's strength ɛ is the single free parameter of the scaled form of the model. It may be implemented in atomic Bose-Einstein condensates and nonlinear optics. The SSB bifurcation of the symmetric ground state (g.s.) is predicted analytically in two limit cases, viz., for deep or weak splitting of the potential box by the barrier (ɛ≫1 or ɛ≪1, respectively). For the generic case, a variational approximation (VA) is elaborated. The analytical findings are presented along with systematic numerical results. The stability of stationary states is studied through the calculation of eigenvalues for small perturbations and by means of direct simulations. The g.s. always undergoes the SSB bifurcation of the supercritical type, as predicted by the VA at moderate values of ɛ, although the VA fails at small ɛ, due to inapplicability of the underlying ansatz in that case. However, the latter case is correctly treated by the approximation based on a soliton ansatz. On top of the g.s., the first and second excited states are studied too. The antisymmetric mode (the first excited state) is destabilized at a critical value of its norm. The second excited state undergoes SSB bifurcation, like the g.s., but, unlike it, the bifurcation produces an unstable asymmetric mode. All unstable modes tend to spontaneously reshape into the asymmetric g.s.