Runge-Kutta type methods for orthogonal integration (Q5961742)

This paper is concerned with the numerical solution of matrix differential equations that possess orthogonal lows by means of onestep methods that preserve the orthogonality along the numerical solution. Such differential systems have the form \(Y'=F(Y)Y\), where \(F\) is a matrix function that maps orthogonal matrices to skew-symmetric matrices. The author shows that the linearly-implicit modification of the implicit midpoint rule given by \(Y_{n+1}=Y_n+(h/2)F(Y_n)\) \((Y_n+Y_{n+1})\) preserves orthogonality and has order one, while the implicit midpoint rule does not preserve this property. A second-order linearly-implicit method which preserves orthogonality is also given. Finally, some numerical experiments are presented to show the effects of the lack of orthogonality of some numerical methods in the numerical solution of two selected nonlinear matrix problems.