Title: Characterization of systematic error in Advanced LIGO calibration in the second half of O3

Abstract: We present the probability distribution of the systematic errors in the most accurate, high-latency version of the reconstructed dimensionless strain $h$, at the Hanford and Livingston LIGO detectors, used for gravitational-wave astrophysical analysis, including parameter estimation, in the last five months of the third observing run (O3B). This work extends the results presented in Sun et. al (2020) [1] for the first six months of the third observing run (O3A). The complex-valued, frequency-dependent, and slowly time-varying systematic error (excursion from unity magnitude and zero phase) in O3B generally remains at a consistent level as in O3A, yet changes of detector configurations in O3B have introduced a non-negligible change in the frequency dependence of the error, leading to larger excursions from unity at some frequencies and/or during some observational periods; in some other periods the excursions are smaller than those in O3A. For O3B, the upper limit on the systematic error and associated uncertainty is 11.29% in magnitude and 9.18 deg in phase (68% confidence interval) in the most sensitive frequency band 20-2000 Hz. The systematic error alone is estimated at levels of < 2% in magnitude and $\lesssim 4$ deg in phase. These errors and uncertainties are dominated by the imperfect modeling of the frequency dependence of the detector response functions rather than the uncertainty in the absolute reference, the photon calibrators.