6533b7ddfe1ef96bd1275299

RESEARCH PRODUCT

Erratum: “Searches for Gravitational Waves from Known Pulsars at Two Harmonics in 2015–2017 LIGO Data” (2019, ApJ, 879, 10)

subject

description

Two analysis errors have been identified that affect the results for a handful of the high-value pulsars given in Table 1 of Abbott et al. (2019). One affects the Bayesian analysis for the five pulsars that glitched during the analysis period, and the other affects the 5n-vector analysis for J0711-6830. Updated results after correcting the errors are shown in Table 1, which now supersedes the results given for those pulsars in Table 1 of Abbott et al. (2019). Updated versions of figures can be seen in Figures 1-4. Bayesian analysis.-For the glitching pulsars, the signal phase evolution caused by the glitch was wrongly applied twice and was therefore not consistent with our expected model of the pulsar phase. This error did not affect the F/G-statistic or 5n-vector analysis. Analyses of the five pulsars PSR J0205+6449, PSR J0534+2200, PSR J0835-4510, PSR J1028-5819, and PSR J1718-3825 have been repeated after correcting for the error. There are small quantitative differences in the results, but the changes do not affect the main conclusions of the paper. The largest differences are for PSR J0835-4510 (the Vela pulsar), for which the updated upper limits from the Bayesian method are found to be between 1.1 and 2 times larger than those obtained when the error was present. This appears primarily to be due to the error leading to the decohering of a strong spectral line in the LIGO Livingston detector and thus lowering the amplitude limit. 5n-vector analysis.-An error was also identified in the settings of the 5n-vector analysis, which affected the upper limit computation at the rotation frequency for C21 95% of J0711-6830. Specifically, we found an incorrect choice for the range of amplitudes used to inject simulated signals in the O2 data. The updated upper limit is about 2.5 times worse than that obtained when the error was present. This error did not affect the Bayesian or F/G-statistic results. (Table Presented) (Figure Presented).