Prolongation-collocation variational integrators

DOI10.1093/IMANUM/DRR042MaRDI QIDQ2902208zbMATH OpenOpenAlexFDO

Publication date 17 August 2012

Published in IMA Journal of Numerical Analysis (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1101.1995

collocation error control Euler-Lagrange equations Euler-Maclaurin quadrature Hamiltonian ODEs Hermite interpolation Lagrangian variational integrators numerical examples variational order

Mathematics Subject Classification ID

Discretization methods and integrators (symplectic, variational, geometric, etc.) for dynamical systems (37M15) Numerical methods for Hamiltonian systems including symplectic integrators (65P10) Variational methods for problems in mechanics (70G75) Lagrange's equations (70H03)

Abstract: We introduce a novel technique for constructing higher-order variational integrators for Hamiltonian systems of ODEs. In particular, we are concerned with generating globally smooth approximations to solutions of a Hamiltonian system. Our construction of the discrete Lagrangian adopts Hermite interpolation polynomials and the Euler-Maclaurin quadrature formula, and involves applying collocation to the Euler-Lagrange equation and its prolongation. Considerable attention is devoted to the order analysis of the resulting variational integrators in terms of approximation properties of the Hermite polynomials and quadrature errors. A performance comparison is presented on a selection of these integrators.

Review text

An approach for solving differential systems that arise from variational Hamiltonian equations is proposed. In such a system, the time evolution of the generalized coordinates \( q = q(t)\) in the interval \([0,T]\) from a starting value \( q(0)=q^0\) to an end value \( q(T)= q^1\) is known to minimize the action integral \( S(q) \equiv \int_0^T L (q(t), \dot{q}(t) ) \; dt\), where \( L=L( \dot{q}, q)\) is the Lagrangian function. NEWLINENEWLINENEWLINENEWLINEFor a given discrete sequence of nodes \(t_k = k h\), \(k=0, \dots ,N\), \(N = T/h\) in the integration interval and a discrete curve \( \{ q_k \}_{k=0}^N\), where \( q_k \simeq q(t_k)\), the exact curve solution \( q = q(t)\) at each \( t \in [t_k, t_{k+1}]\) is approximated by a two-point Hermite interpolant of suitable order that ensures not only the continuity of positions and velocities but also some error control on higher order derivatives at the nodes. NEWLINENEWLINENEWLINENEWLINEIn this setting, the exact action in each interval \([t_k, t_{k+1}] \) is substituted by a discrete action \( L_d (q_k, q_{k+1})\) so that the action \( \int_{t_k}^{t_{k+1} }L(q(t), \dot{q}(t) ) \; dt\) is approximated by using the Euler-Maclaurin quadrature formula that only involves values of the derivatives at the end points of the integration interval. For separable systems, where \( L = (1/2) \dot{q}^T \dot{q} - V(q) \) with some potential \( V = V(q)\), several derivatives of the potential are required depending on the order of the Euler-MacLaurin approximation for the definition of the discrete action together with collocation conditions at both end points. NEWLINENEWLINENEWLINENEWLINEAfter obtaining the discrete action, a set of discrete variational equations is derived so that it defines implicitly the coordinates \( \{ q_k \}_{k=0}^N\) at the given nodes. It is proved that if the Lagrangian \(L\) is time independent then the proposed numerical solvers are self-adjoint and therefore the methods have an even order of accuracy. A study of the order is carried out and finally the results of some numerical examples with elementary test problems are presented to show the accuracy and the preservation of geometrical properties.

Recommendations

Cited in

(19)

This page was built for publication: Prolongation-collocation variational integrators

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2902208)