From Spin Noise to Systematics: Stochastic Processes in the First International Pulsar Timing Array Data Release

TL;DR

This paper analyzes stochastic signals in pulsar timing data, demonstrating how detailed modeling of system and band-dependent effects enhances gravitational wave detection sensitivity.

Contribution

It introduces Bayesian model selection to distinguish various stochastic processes, improving the interpretation of pulsar timing residuals and GW sensitivity.

Findings

Identification of band-dependent noise in PSR J1643-1224

Improved GW background upper limit by 60% for PSR J0437-4715

Better separation of system effects due to overlapping data

Abstract

We analyse the stochastic properties of the 49 pulsars that comprise the first International Pulsar Timing Array (IPTA) data release. We use Bayesian methodology, performing model selection to determine the optimal description of the stochastic signals present in each pulsar. In addition to spin-noise and dispersion-measure (DM) variations, these models can include timing noise unique to a single observing system, or frequency band. We show the improved radio-frequency coverage and presence of overlapping data from different observing systems in the IPTA data set enables us to separate both system and band-dependent effects with much greater efficacy than in the individual PTA data sets. For example, we show that PSR J1643$-$1224 has, in addition to DM variations, significant band-dependent noise that is coherent between PTAs which we interpret as coming from time-variable scattering or refraction in the ionised interstellar medium. Failing to model these different contributions appropriately can dramatically alter the astrophysical interpretation of the stochastic signals observed in the residuals. In some cases, the spectral exponent of the spin noise signal can vary from 1.6 to 4 depending upon the model, which has direct implications for the long-term sensitivity of the pulsar to a stochastic gravitational-wave (GW) background. By using a more appropriate model, however, we can greatly improve a pulsar's sensitivity to GWs. For example, including system and band-dependent signals in the PSR J0437$-$4715 data set improves the upper limit on a fiducial GW background by $\sim 60\%$ compared to a model that includes DM variations and spin-noise only.