Blazars define a stable celestial reference frame

TL;DR

The study analyzes how photometric variability of AGN affects the stability of the celestial reference frame by constructing optical-radio ties from ICRF3 sources matched to Gaia DR3 and evaluating frame stability with a bootstrap-based vector spherical harmonic model across variability classes. It finds that frames defined by highly variable blazars are markedly more stable than those defined by quasars, with the $S/X$ frame showing up to a sixfold improvement; at higher frequencies, blazar selection remains beneficial though to a lesser extent. The authors introduce a simple variability-based covariance weighting, $F_{var}^{-x}$, with exponents $x_{S/X}=1.61$, $x_K=1.15$, and $x_{X/Ka}=0.87$, which reduces frame distortions by about a factor of two for $S/X$ and yields milder gains for the other bands. These results confirm the prediction that optimal linking of optical and radio reference frames can be achieved by accounting for photometric variability, and suggest a practical pathway to future frame construction using larger, variability-labeled samples.

Abstract

Recent work has shown that optical-radio position offsets and radio position variability are inversely correlated with the photometric variability of active galactic nuclei (AGN). A key prediction of these findings is that a reference frame constructed using highly photometrically variable AGN should be more stable than a frame that does not account for variability and that variability can be used to optimally weight all sources in order to maximize frame stability. Using ICRF3 matched to Gaia DR3, we employed a bootstrap method to estimate the multi-epoch stability of frames constructed using AGN selected at varying levels of photometric variability. We fit vector spherical harmonics to the coordinate differences between the three ICRF3 frames (S/X, K, and X/Ka) and Gaia and quantified the statistical dispersion as a function of blazar-like (high variability), quasar-like (low variability), and intermediate-variability class. An S/X reference frame constructed using blazars exceeds the stability of a frame constructed with quasars by a factor of 6 and is twice as stable as the ICRF3 defining sources. At K and X/Ka, a blazar-based frame matches or exceeds the stability of the defining sources by a factor of 1.4 in the case of X/Ka and exceeds the stability of a frame based on quasars by over a factor of 2 in both cases. The smaller improvement at K and X/Ka is likely because sources selected at higher frequency are more likely to be blazars. We derived a variability-based astrometric covariance scaling method that results in factor of 2 reduction in frame distortions and instabilities between S/X and Gaia, with a mild improvement for K but no difference for X/Ka, which is dominated by known distortions. Our results confirm the prediction that an optimal weighting of the link between the optical and radio celestial reference frames is enabled by accounting for photometric variability.

Figures (10)

• Figure 1: Ratio of the average bootstrap sample variance to the true population variance of the 16 terms in the $l\leq2$ VSH model we employed for simulated multi-epoch data in which position offsets vary with epoch. The position offsets were sampled from the empirical distribution of statistically significant ICRF3- Gaia offsets observed in the sample used by 2022ApJ...939L..32S and this work. The error bars denote the standard errors of the means.
• Figure 2: Distribution of $l=0$ (rotation) and $l=1$ (glide) values between ICRF3 $S/X$ and Gaia DR3, separating objects by fractional variability into quasar-like ($F_\mathrm{var} < 0.2$), intermediate ($0.2 < F_\mathrm{var} < 0.4$), and blazar-like ($F_\mathrm{var} > 0.4$) samples. A dramatic improvement in frame stability and rigidity is enabled by selection on blazars, exceeding even the sources used to define the ICRF3 frame.
• Figure 3: Distribution of $l=0$ (rotation) and $l=1$ (glide) values between ICRF3 $K$ and Gaia DR3, separating objects by fractional variability into quasar-like ($F_\mathrm{var} < 0.2$), intermediate ($0.2 < F_\mathrm{var} < 0.4$), and blazar-like ($F_\mathrm{var} > 0.4$) samples.
• Figure 4: Distribution of $l=0$ (rotation) and $l=1$ (glide) values between ICRF3 $X/Ka$ and Gaia DR3, separating objects by fractional variability into quasar-like ($F_\mathrm{var} < 0.2$), intermediate ($0.2 < F_\mathrm{var} < 0.4$), and blazar-like ($F_\mathrm{var} > 0.4$) samples. The large value of the $D_3$ term is a known issue and was discussed by 2020AA...644A.159C.
• Figure 5: Reduced chi-squared statistic of the 16-parameter VSH model as a function of fractional variability using the ICRF3 $S/X$ and Gaia DR3 positions, showing that the variability-based covariance scaling given by Eq. \ref{['eq: covariance weighting']} correctly weights the source positions.