Generalized Non-linear Bayesian Pulsar Timing with Enterprise
In this study, we use the Bayesian methods in the Enterprise package to examine the fully general parameterization of pulsar timing models in tandem with noise. We investigate four pulsars, PSR J1600$-$3053, PSR J2043+1711, PSR J0740+6620, and PSR J1640+2224, through the lens of Bayesian timing. These four are selected as they are well-studied, but exhibit interesting characteristics under the lens of Bayesian timing. Our new pulsar mass constraints (medians and 68\% confidence intervals) for our fully general non-linear Bayesian timing models are $m_{\mathrm{p}}=1.6(1)~\mathrm{M}_{\odot}$ for PSR J2043+1711 and $m_{\mathrm{p}}=2.3^{+0.9}_{-0.7}~\mathrm{M}_{\odot}$ for PSR J1600$-$3053 both using the NANOGrav 12.5-yr data release, and $m_{\mathrm{p}}=2.06(6)~\mathrm{M}_{\odot}$ for PSR J0740+6620 using the data from Fonseca, et al., 2021. We investigate the effects on placing physical priors on timing model parameters, including restricting the upper limit on the pulsar mass for PSR J1640+2224, which has a mass often estimated to be greater than $3~\mathrm{M}_{\odot}$. We find \ark{that restricting the allowed sampling space of the pulsar mass for PSR J1640+2224 to} $m_{\mathrm{p}}<3~\mathrm{M}_{\odot}$ results in a pulsar mass of $m_{\mathrm{p}}=2.2(5)~\mathrm{M}_{\odot}$ for PSR J1640+2224 using the NANOGrav 12.5-yr data release. For the first time, we find evidence for intrinsic red noise in PSR J2043+1711. We show how fully general Bayesian timing can better model the interplay of the intrinsic noise and the timing parameters.