Efficient numerical methods for computing ground states and dynamics of dipolar Bose-Einstein condensates

1 Introduction

• Experimental advances in dipolar quantum gases have renewed interest in studying their ground states and dynamics.At temperatures much below the critical temperature, dipolar BECs are described by a macroscopic wave function governed by a 3D GPE.Source: §1 · p. 2
• Fourier-based methods can mismatch whole-space and bounded-domain transforms, producing locking, and may omit the zero-mode term in practical computation.The paper proposes a new formulation to avoid these drawbacks.Source: §1 · p. 4
• Decoupling the dipolar potential into short- and long-range interactions reformulates the GPE as a Gross-Pitaevskii-Poisson type system and supports ground-state and dynamical analysis.The paper reports proofs of ground-state existence, uniqueness, and nonexistence, and analyzes dynamics across parameter regimes.Source: §1 · p. 4
• A backward Euler sine pseudospectral method is proposed for ground states, and a time-splitting sine pseudospectral method for dynamics.The paper reports 3D numerical results to demonstrate the methods’ efficiency and accuracy.Source: §1 · p. 4

2 Analytical results for ground sates and dynamics

• The dipolar interaction is split into short- and long-range terms, yielding a Gross-Pitaevskii-Poisson type system.The short- and long-range terms are attractive and repulsive, respectively, when λ > 0, with the roles reversed when λ < 0.Source: §2 · p. 5

2.1 Existence and uniqueness for ground states

• The analysis establishes existence and uniqueness of a positive ground state in some parameter regimes, and nonexistence in others where the constrained energy has infimum −∞.Source: §2.1 · pp. 6, 8
• Stationary states satisfy a virial identity relating kinetic, trap, interaction, and dipolar energies.Source: §2.1 · pp. 8–9

2.2 Analytical results for dynamics

• Under the stated smooth, nonnegative trap assumptions, the Cauchy problem has a unique maximal solution, conserves mass and energy, and is global in the stated parameter regime.Source: §2.2 · p. 9
• A finite-time blow-up result is proved for specified parameter regimes when the trap satisfies 3V(x) + x·∇V(x) ≥ 0.Source: §2.2 · p. 9

3 A numerical method for computing ground states

• The paper proposes a backward Euler sine pseudospectral method, based on the new energy formulation, for computing dipolar BEC ground states.Source: §3 · p. 10
• For computation, the whole-space problem is truncated to a bounded domain with homogeneous Dirichlet boundary conditions.Source: §3 · p. 10
• A commonly used constrained approach applies steepest descent to the energy and projects the solution back onto the unit sphere after each time interval.Source: §3 · pp. 10–11
• The backward Euler sine spectral discretization is implemented on grid points, approximating its integrals by quadrature rather than computing them directly in phase space.The phase-space discretization is described as impractical because its integrals are difficult to compute.Source: §3 · pp. 11–12
• The resulting linear system can be solved iteratively in phase space using the discrete sine transform.Source: §3 · p. 12

4 A time-splitting sine pseudospectral method for dynamics

• Each time interval advances the Gross-Pitaevskii-Poisson system in two steps, with the intermediate equations preserving |ψ| and ϕ.Source: §4 · pp. 12–13
• The proposed second-order time-splitting sine pseudospectral method uses standard Strang splitting to compute dipolar BEC dynamics.Source: §4 · pp. 12–13
• The method is explicit and unconditionally stable, with memory cost O(MKL) and computational cost O(MKL ln(MKL)) per time step.Source: §4 · p. 13
• The TSSP method conserves normalization.Source: §4 · p. 13

5.1 Comparison for evaluating the dipolar energy

• The discrete-sine-transform method is reported to be much more accurate than the standard discrete-Fourier-transform method for evaluating dipolar energy.The comparison covers different mesh sizes.Source: Table 1 · p. 14, §5.1 · p. 14

5.2 Ground states of dipolar BECs

• At fixed trapping potential and dipolar direction, increasing β and λ while keeping λ/β fixed raises energy, chemical potential, potential and interaction energies, and z-width, while kinetic and dipolar energies and central density fall.Source: §5.2 · p. 15
• At fixed trapping potential, dipolar direction, and β, increasing λ/β raises kinetic energy, z-width, and central density, while energy, chemical potential, potential energy, and interaction energy decrease.Source: §5.2 · pp. 15–16
• The new method is reported to compute ground states accurately and efficiently across the presented cases.Source: §5.2 · p. 16

5.3 Dynamics of dipolar BECs

• Global existence is observed after changing dipole direction or weakening z-confinement, whereas finite-time blow-up is observed after strengthening dipolar interaction or making β negative.Source: §5.3 · p. 16
• Total energy is numerically conserved well without blow-up and before blow-up, but not near or after blow-up; the method is reported to compute dynamics accurately and efficiently.Source: §5.3 · pp. 17–18

• Decoupling the dipolar interaction into short- and long-range parts reformulates the GPE as a Gross-Pitaevskii-Poisson type system.Source: §6 · p. 18
• The paper proves existence, uniqueness, and nonexistence results for ground states and discusses dynamics across parameter regimes.Source: §6 · p. 18
• The backward Euler and time-splitting sine pseudospectral methods address ground-state and dynamics computations, respectively, and avoid Fourier-transforming the highly singular nonlocal interaction potential.Source: §6 · p. 18
• Comparison with existing methods showed better performance; future work is proposed on experimentally relevant setups and rotating dipolar BECs.Source: §6 · p. 18

• For a unit vector n, the proof reduces equality (2.2) to proving an equivalent statement.Source: Appendix Proof of the equality (2.2) · p. 19
• The calculation uses integration by parts, a change of variables, and an orthonormal basis chosen using n.Source: Appendix Proof of the equality (2.2) · p. 19
• The resulting identity establishes equality (2.2) by the definition of the derivative in the distribution sense.Source: Appendix Proof of the equality (2.2) · p. 20