Stochastic problems in pulsar timing

Abstract

Langevin stochastic differential equations provide a dynamical description of pulsar timing noise and gravitational wave background (GWB) signals. They are also central to state space algorithms that have gained traction in pulsar timing array analysis due to their linear computational scaling with the number of observations. In this work, we utilize established methods in diffusion theory to derive analytical time-domain solutions (means, covariances, and probability density functions) to Langevin equations relevant to red noise and the GWB signal in pulsars. The solutions give direct physical insight on the dynamics of pulsar timing signals. As a canonical example, we show that the pulsar spin frequency modeled as an Ornstein-Uhlenbeck process is mathematically inconsistent with a stationary GWB signal when the timing residual is the direct observable. The nonstationarity can be partially dealt with by marginalizing over long time deterministic trends in the data. Then, we show that a random process based on an overdamped harmonic oscillator supports both a stationary spin frequency and phase residuals, consistent with a stationary GWB signal. We also turn our attention to a phenomenological model of a neutron star -- a two-component model with spin wandering -- that has been motivated to explain observed timing noise in radio pulsars. We derive analytical expressions for the means, covariances, and cross-covariances of the crust and superfluid rotational states driven by white noise. The associated constant deterministic torques are linked to the quadratic spin-down of pulsars. The solutions reveal the physical origin of nonstationarity in the residual model: the coexistence of damped and diffusive eigenmodes of the system.