1: \begin{abstract}

2: In this paper, we propose a regularized Newton method for  computing ground

3: states of Bose-Einstein condensates (BECs), which can be formulated as

4:  an energy minimization problem with a spherical

5: constraint. The energy functional and constraint

6: are discretized by either the finite difference, or sine or Fourier

7: pseudospectral discretization schemes and thus the original infinite dimensional

8: nonconvex minimization problem is approximated by a finite dimensional constrained

9: nonconvex minimization problem. Then an initial solution is first constructed by using a feasible gradient type

10: method, which is an explicit scheme and maintains the spherical constraint automatically.

11: To accelerate the convergence of the gradient type method, we approximate the energy

12: functional by its second-order Taylor expansion with a regularized term at each

13: Newton iteration and adopt a cascadic multigrid technique for selecting initial data.

14: It leads to a standard trust-region subproblem and we

15:  solve it again by the feasible gradient type method.

16: The convergence of the regularized Newton method is established by adjusting

17: the regularization parameter as the standard trust-region strategy.

18: Extensive numerical experiments on challenging examples,

19: including a BEC in three dimensions with an optical lattice potential and

20: rotating BECs in two dimensions with rapid rotation and strongly repulsive interaction, show that

21: our method is efficient, accurate and robust.

22: 

23: \end{abstract}

24: