# The 31-year Rotation History of the Millisecond Pulsar J1939+2134 (B1937+21)

## TL;DR

This study combines 31 years of timing data for pulsar J1939+2134, revealing its exceptional stability, a sinusoidal timing noise pattern, and potential explanations including a planetary companion or precession.

## Contribution

It provides a unified analysis of decades of pulsar timing data, identifying a 31-year sinusoidal noise pattern and exploring possible physical causes.

## Key findings

- Pulsar clock stability at 1 part in 10^15 over 31 years
- Detection of a 31-year sinusoidal timing noise pattern
- Possible planetary companion or precession as causes of timing noise

## Abstract

The timing properties of the millisecond pulsar PSR J1939+2134 -- very high rotation frequency, very low time derivative of rotation frequency, no timing glitches and relatively low timing noise -- are responsible for its exceptional timing stability over decades. It has been timed by various groups since its discovery, at diverse radio frequencies, using different hardware and analysis methods. Most of this timing data is now available in the public domain in two segments, which have not been combined so far. This work analyzes the combined data by deriving uniform methods of data selection, derivation of Dispersion Measure (DM), accounting for correlation due to "red" noise, etc. The timing noise of this pulsar is very close to a sinusoid, with a period of approximately $31$ years. The main results of this work are (1) The clock of PSR J1939+2134 is stable at the level of almost one part in $10^{15}$ over about $31$ years, (2) the power law index of the spectrum of electron density fluctuations in the direction of PSR J1939+2134 is $3.86 \pm 0.04$, (3) a Moon sized planetary companion, in an orbit of semi major axis about $11$ astronomical units and eccentricity $\approx 0.2$, can explain the timing noise of PSR J1939+2134, (4) Precession under electromagnetic torque with very small values of oblateness and wobble angle can also be the explanation, but with reduced confidence, and (5) there is excess timing noise of about $8$ $\mu$s amplitude during the epochs of steepest DM gradient, of unknown cause.