Abstract
In (Bernier in Exact splitting methods for semigroups generated by inhomogeneous quadratic differential operators. arXiv:1912.13219, (2019)), some exact splittings are proposed for inhomogeneous quadratic differential equations including, for example, transport equations, Fokker–Planck equations, and Schrödinger type equations with an angular momentum rotation term. In this work, these exact splittings are used combined with pseudo-spectral methods in space. High accuracy and efficiency of exact splitting methods are illustrated and comparison are performed with the numerical methods in literature. We show that our methods can be used to improve significantly some classical splitting methods for some nonlinear or non-quadratic equations.
Similar content being viewed by others
Notes
http://gpelab.math.cnrs.fr
References
Alphonse, P., Bernier, J.: Polar decomposition of semigroups generated by non-selfadjoint quadratic differential operators and regularizing effects. arXiv:1909.03662 (2019)
Ameres, J.: Splitting methods for Fourier spectral discretizations of the strongly magnetized Vlasov-Poisson and the Vlasov-Maxwell system. arXiv preprint arXiv:1907.05319 (2019)
Antoine, X., Bao, W., Besse, C.: Computational methods for the dynamics of the nonlinear Schrödinger/Gross-Pitaevskii equations. Comput. Phys. Commun. 184(12), 2621–2633 (2013)
Antoine, X., Duboscq, R.: GPELab, a Matlab toolbox to solve Gross–Pitaevskii equations I: computation of stationary solutions. Comput. Phys. Commun. 185(11), 2969–2991 (2014)
Antoine, X., Duboscq, R.: Gpelab, a matlab toolbox to solve Gross–Pitaevskii equations II: dynamics and stochastic simulations. Comput. Phys. Commun. 193, 95–117 (2015)
Bader, P.: Fourier-splitting methods for the dynamics of rotating Bose–Einstein condensates. J. Comput. Appl. Math. 336, 267–280 (2018)
Bader, P., Blanes, S.: Fourier methods for the perturbed harmonic oscillator in linear and nonlinear Schrödinger equations. Phys. Rev. E 83(4), 046711 (2011)
Bader, P., Blanes, S., Casas, F.: Efficient time integration methods for Gross–Pitaevskii equations with rotation term. preprint arXiv:1910.12097 (2019)
Bao, W., Du, Q., Zhang, Y.: Dynamics of rotating Bose-Einstein condensates and its efficient and accurate numerical computation. SIAM J. Appl. Math. 66(3), 758–786 (2006)
Bao, W., Marahrens, D., Tang, Q., Zhang, Y.: A simple and efficient numerical method for computing the dynamics of rotating Bose–Einstein condensates via rotating Lagrangian coordinates. SIAM J. Sci. Comput. 35(6), A2671–A2695 (2013)
Bao, W., Wang, H.: An efficient and spectrally accurate numerical method for computing dynamics of rotating Bose–Einstein condensates. J. Comput. Phys. 217(2), 612–626 (2006)
Bernier, J.: Exact splitting methods for semigroups generated by inhomogeneous quadratic differential operators. Preprint, arXiv:1912.13219 (2019)
Bernier, J., Casas, F., Crouseilles, N.: Splitting methods for rotations: application to Vlasov equations. SIAM J. Sci. Comput. 42(2), 1 (2020)
Besse, C., Descombes, S., Dujardin, G., Lacroix-Violet, I.: Energy preserving methods for nonlinear Schrödinger equations. IMA J. Numer. Anal. 1, drz067 (2020)
Besse, C., Dujardin, G., Lacroix-Violet, I.: High order exponential integrators for nonlinear Schrödinger equations with application to rotating Bose-Einstein condensates. SIAM J. Numer. Anal. 55(3), 1387–1411 (2017)
Besse, N., Mehrenberger, M.: Convergence of classes of high-order semi-Lagrangian schemes for the Vlasov–Poisson system. Math. Comput. 77(261), 93–123 (2008)
Caliari, M., Ostermann, A., Piazzola, C.: A splitting approach for the magnetic Schrödinger equation. J. Comput. Appl. Math. 316, 74–85 (2017)
Chen, B., Kaufman, A.: 3D volume rotation using shear transformations. Graph. Models 62(4), 308–322 (2000)
Coulaud, O., Sonnendrücker, E., Dillon, E., Bertrand, P.: J. Plasma Phys. 61, 435–448 (1999)
Dujardin, G., Hérau, F., Lafitte, P.: Coercivity, hypocoercivity, exponential time decay and simulations for discrete Fokker–Planck equations. arXiv preprint arXiv:1802.02173 (2018)
Hairer, E., Lubich, C., Wanner, G.: Geometric numerical integration: Structure-Preserving Algorithms for Ordinary Differential Equations, Springer Series in Computational Mathematics (2006)
Hérau, F., Sjöstrand, J., Hitrik, M.: Tunnel effect for the Kramers-Fokker-Planck type operators: return to equilibrium and applications. Int. Math. Res. Not. 57, 48 (2008)
Hérau, F., Sjöstrand, J., Hitrik, M.: Tunnel effect for the Kramers–Fokker–Planck type operators. Ann. Henri Poincaré 9, 209–274 (2008)
Hérau, F., Thomann, L.: On global existence and trend to the equilibrium for the Vlasov–Poisson–Fokker–Planck system with exterior confining potential. J. Funct. Anal. 271, 1301–1340 (2016)
Hochbruck, M., Ostermann, A.: Exponential integrators. Acta Numerica 19, 209–286 (2010)
Hörmander, L.: Symplectic classification of quadratic forms, and general Mehler formulas. Math. Z. 219, 413–449 (1995)
Hörmander, L.: The analysis of linear partial differential operators. III, classics in mathematics. Pseudo-differential operators. Springer, Berlin (2007). https://doi.org/10.1007/978-3-540-49938-1
Jin, S., Zhou, Z.: A semi-Lagrangian time splitting method for the Schrödinger equation with vector potentials. Commun. Inf. Syst. 13(3), 247–289 (2013)
Li, Y., He, Y., Sun, Y., et al.: Solving the Vlasov–Maxwell equations using Hamiltonian splitting. J. Comput. Phys. 396, 381–399 (2019)
McLachlan, R.I., Quispel, G.R.: Splitting methods. Acta Numerica 11, 341–434 (2002)
Marsden, J.E., Ratiu, T.S.: Introduction to mechanics and symmetry: a basic exposition of classical mechanical systems. Springer, Berlin (2013)
Raymond, N.: Bound States of the Magnetic Schrödinger Operator. EMS Tracts Math. (2017)
Nicola, F., Rodino, L.: Global Pseudo-Differential Calculus on Euclidean Spaces. Birkhäuser, Basel (2010)
Welling, J.S., Eddy, W.F., Young, T.K.: Rotation of 3D volumes by Fourier-interpolated shears. Graph. Models 68(4), 356–370 (2006)
Zeng, R., Zhang, Y.: Efficiently computing vortex lattices in rapid rotating Bose–Einstein condensates. Comput. Phys. Commun. 180(6), 854–860 (2009)
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The author J.B. was supported by the French National Research Agency project NABUCO, grant ANR-17-CE40-0025. The author Y.L. is supported by a scholarship from Academy of Mathematics and Systems Science, Chinese Academy of Sciences. The author N.C. has been supported by the EUROfusion Consortium and has received funding from the Euratom research and training programme 2019–2020 under Grant Agreement No. 633053. The views and opinions expressed herein do not necessarily reflect those of the European Commission.
Appendix
Appendix
1.1 Backward Error Analysis of the Strang Splittings for Rotations
We consider the ODE associated with a rotation in \({\mathbb {R}}^n\)
where \(B\in {\mathfrak {so}}_n({\mathbb {R}})\) is a real skew-symmetric matrix of size n. For \(t\in {\mathbb {R}}\), the Strang splitting naturally associated with the decomposition of this rotation in shear transforms is
where \(e_1,\dots ,e_n\) is the canonical basis of \({\mathbb {R}}^n\). It is indeed a Strang splitting because since \((e_j\otimes e_j) B\) is nilpotent of order 1 we have
The following proposition states that, up to a change of coordinate close to the identity, \(P_t\) is a rotation of which angles coincide with those of \(\exp (tB)\) up to an error of order \(t^3\).
Proposition 2
There exists \(t_0>0\) and two analytic functions \(t\in (-t_0,t_0) \mapsto (B_t,Q_t) \in \mathfrak {so}_n({\mathbb {R}}) \times \mathrm {S}_n({\mathbb {R}})\cap \mathrm {GL}_n({\mathbb {R}})\), where \(B_t\) is real skew symmetric and \(Q_t\) is real symmetric and invertible such that \(B_0 = B\), \(Q_0 = \mathrm {I}_n\) and
Proof
Observing that since
it follows from the Bakker-Campbell-Hausdorff formula that there exist \(t_1>0\) and an analytic function \(t \in (-t_1,t_1)\mapsto S_t \in \mathrm {S}_{n}({\mathbb {R}})\) such that \(S_0= \mathrm {I}_n\) and
Since the Strang splitting is reversible, we have \(P_t = P_{-t}^{-1}\). Thus, since the exponential map is injective in a neighborhood of the identity, we deduce that \(t\mapsto S_t\) is an even function. Consequently, there exists an analytic function \(t\mapsto K_t\) such that \(K_{t^2} = S_t\). Furthermore, since \(K_0= \mathrm {I}_n\), there exists \(t_0\in (0,t_1)\) such that for all \(t\in (-t_0,t_0)\), \(K_t\) is positive-definite. Finally, we deduce that,
where \(Q_t = \sqrt{K_{t}}\) and \(B_t = \sqrt{K_{t}}B\sqrt{K_{t}}\). \(\square \)
1.2 2D Magnetic Schrödinger Equation
where \(\mathbf{x} = (x_1,x_2) \in {\mathbb {R}}^2\), \(\mathbf{A} = \frac{1}{2}(A_1, A_2)\), \(A_1= -x_2\), \(A_2 = x_1\). The above system can be split into three systems:
The solutions of the above three subsystems can be obtained by operators \(e^{it\frac{\epsilon }{2} \Delta }\), \(e^{t\text {Rot}}\), and \(e^{tV}\) respectively. Since the second is nothing but a 2D rotation, we call the associated solution \(e^{t\text {Rot}}\). Then we have the following second order splitting method
from which we derive two variants according to the treatment of \(e^{\Delta t\text {Rot}}\). Indeed, ESR denotes the splitting method (42) when \(e^{\Delta t \text {Rot}}\) is solved by exact splittings for transport equation in Proposition 1. Strang denotes (42) when \(e^{\Delta t \text {Rot}}\) is approximated by Strang directional splitting.
1.3 2D Rotating Gross–Pitaevskii Equation
The rotating Gross-Pitaevskii equation (GPE) [9, 11] is
where \(\psi (\mathbf{x}, t)\) is the macroscopic wave function, \(\mathbf{x} = (x_1, x_2)\), \(L_{x_3} = - i (x_1\partial _{x_2} - x_2 \partial _{x_1})\). Two operator splittings are presented to approximate (43).
1.3.1 BW Method
Here we recall the splitting method introduced in [11] to approximate (43). We will call it BW in the sequel. BW splitting for rotating GPE (43) is based on the following two-steps splitting
Then, the authors in [11] noticed that (44) can be split further as
The solutions of subsystems (45), (46) and (47) can be obtained by operators \(e^{tN}, e^{tX}\) and \(e^{tY}\) respectively, the second order BW method is then derived from the following composition
Combined with Fourier pseudo-spectral method in space, we can see that in each time step, we need six calls to FFT.
1.3.2 Lagrangian Method
Here we recall the main step of the method introduced in [10] to approximate (43). We will call it Lagrangian in the sequel. First, a change of coordinates is considered
where J is the \(2\times 2\) symplectic matrix. Then, \(\phi \) satisfies the following equation
with the initial condition \(\phi (\tilde{\mathbf{x}}, 0)=\psi _0(\mathbf{x})\) and where W is defined by
so that a time dependency is created. However, if V is isotropic, \(W(\tilde{\mathbf{x}}, t) = V(\tilde{\mathbf{x}})\) is time-independent. The main advantage of (49) does not involve the angular momentum rotation term and the following splitting is used
For harmonic potential V, each step can be solved exactly (which is not the case for general potential V). Combined with a Strang splitting and with Fourier pseudo-spectral method in space, we can see that in each time step, this method needs four calls to FFT.
1.4 3D Time-Periodic Quadratic Linear Schrödinger Equation
For (23) with \(f=0\) and B and V are specified in (32) and (33), we consider two numerical methods: ESQM and a standard Strang operator splitting.
1.4.1 Exact Splitting
The coefficients for ESQM (21) are given by
1.4.2 Strang Method
Classically, we use the following operator splitting
The solutions of the above three subsystems can be obtained by operators \(e^{-it\frac{1}{2}\Delta }\), \(e^{t \text {Rot}}\), and \(e^{-it \text {V}}\) respectively so that we have the following second order splitting method
Strang denotes (52) when \(e^{\Delta t \text {Rot}}\) is also approximated by a Strang directional splitting.
1.5 3D Magnetic Schrödinger Equation
From (35), where \(\mathbf{A}(\mathbf{x}) = \mathbf{x} \times \mathbf{B}\), and \(V_{nq}\) is given by (36), we can use the following operator splitting
The solutions of the above three subsystems can be obtained by operators \(e^{-it\frac{1}{2}\Delta }\), \(e^{t \text {Rot}}\), and \(e^{t \text {VA}}\) respectively and we can derive a second order splitting method:
ESR denotes the splitting method (53) when \(e^{\Delta t \text {Rot}}\) is solved by exact splittings for transport equation in Proposition 1. Strang denotes (53) when \(e^{\Delta t \text {Rot}}\) is approximated by Strang directional splitting.
The coefficients when \(\Delta t = 0.1\) for ESQM (25) are as follows
1.6 Proof of the Period 360
Lemma 1
The function \(t\mapsto U_t = e^{ i t ( \Delta /2 - V(x)) - t Bx \cdot \nabla }\), where V and B are given by (32) satisfies
Proof
Since \(t\mapsto U_t\) is a group, we just have to prove that
We recall that by construction, we have \( U_t = e^{-t q_{(\hbox {QM})}^w}\) where
Step 1: To conjugate \(q_{(\hbox {QMS})}^w\) to a sum of harmonic oscillators. We are going to prove that there exists \(V\in {\mathcal {U}}(L^2({\mathbb {R}}^n))\) such that
where \((\omega _1,\omega _2,\omega _3) = \frac{\pi }{180}(20,75,132)\). Assuming first this decomposition, we deduce that
But, in dimension 1, the eigenvalues of the harmonic oscillator \(x^2-\partial _x^2\) being the odd positive integers, we know that \(\exp (i\pi (x^2-\partial _x^2) ) = -I_{L^2({\mathbb {R}}^3)}\). Thus, we deduce that
In order to prove (54) we are going to apply the following theorem due to Hörmander.
Theorem 2
(Hörmander, Theorem 21.5.3 in [27]) Let \(Q\in S_{2n}^{++}({\mathbb {R}})\) be a real symmetric positive matrix of size 2n. There exists a real symplectic matrix \(P\in \mathrm {Sp}_{2n}({\mathbb {R}})\) of size 2n such that and some positive numbers \(\omega _1,\dots ,\omega _n\) such that
where \(D(\omega ) = \mathrm {diag}(\omega _1,\dots ,\omega _n,\omega _1,\dots ,\omega _n)\) is the diagonal matrix such that, for \(j=1,\dots ,n\), \(D(\omega )_{j,j} = D(\omega )_{j+n,j+n} = \omega _j\) .
Indeed, here, it can be checked that \(Q_{(\hbox {QM})}\) (the matrix of \(q_{(\hbox {QM})}\)) is a symmetric positive matrix (computing, for example, an approximation of its eigenvalues). Thus, applying Theorem 2, we get a symplectic matrix P and some positive numbers \(\omega _1<\omega _2<\omega _3\) such that
Consequently, since P is symplectic, we have
where J is the symplectic matrix of \({\mathbb {R}}^{2n}\). Now, applying the monoid morphism (Theorem 3.1 in [12]) introduced also by Hörmander in [26], we get a function \(t\mapsto \sigma _t\in \{\pm 1\}\) such that
where \(\pm V\) is the Fourier Integral Operator associated with P. Note that V is unitary. Furthermore, by a straighforward argument of continuity we deduce that \(\sigma _t=1\) for all \(t\in {\mathbb {R}}\). Thus, to conclude, we just have to prove that \((\omega _1,\omega _2,\omega _3) = \frac{\pi }{180}(20,75,132)\).
Step 2: To determine \(\omega \). First, we observe that the matrices \(JQ_{(\hbox {QM})}\) and \(JD(\omega )\) are similar. Indeed, since \(P\in \mathrm {Sp}_6({\mathbb {R}})\), we have and applying (55) we deduce that
A fortiori, \(JQ_{(\hbox {QM})}\) and \(JD(\omega )\) have the same eigenvalues. Thus, the eigenvalues of \(JQ_{(\hbox {QM})}\) are
Consequently, to determine \(\omega \) we just have to determine the roots of the characteristic polynomial of \(JQ_{(\hbox {QM})}\), denoted \(\chi _{(\hbox {QM})}\). By a straightforward calculation, we observe that
But, by construction \(\lambda _1< \lambda _2 < \lambda _3\) are the roots of the polynomial
Thus, \(\lambda _1+\lambda _2+\lambda _3\), \(\lambda _1\lambda _2+ \lambda _1\lambda _3+\lambda _2\lambda _3\) and \(\lambda _1\lambda _2\lambda _3\) are some explicit rational numbers and we deduce that
Finally, we verify by an explicit computation that
So, we deduce of (56) that \((\omega _1,\omega _2,\omega _3) = \frac{\pi }{180}(20,75,132)\). \(\square \)
Rights and permissions
About this article
Cite this article
Bernier, J., Crouseilles, N. & Li, Y. Exact Splitting Methods for Kinetic and Schrödinger Equations. J Sci Comput 86, 10 (2021). https://doi.org/10.1007/s10915-020-01369-9
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10915-020-01369-9