A fast algorithm for estimating actions in triaxial potentials

Abstract: We present an approach to approximating rapidly the actions in a general triaxial potential. The method is an extension of the axisymmetric approach presented by Binney (2012), and operates by assuming that the true potential is locally sufficiently close to some St\"ackel potential. The choice of St\"ackel potential and associated ellipsoidal coordinates is tailored to each individual input phase-space point. We investigate the accuracy of the method when computing actions in a triaxial Navarro-Frenk-White potential. The speed of the algorithm comes at the expense of large errors in the actions, particularly for the box orbits. However, we show that the method can be used to recover the observables of triaxial systems from given distribution functions to sufficient accuracy for the Jeans equations to be satisfied. Consequently, such models could be used to build models of external galaxies as well as triaxial components of our own Galaxy. When more accurate actions are required, this procedure can be combined with torus mapping to produce a fast convergent scheme for action estimation.