Multi-detector characterization of gravitational-wave burst tensor polarizations with the BayesWave algorithm

Einstein's theory of general relativity predicts that gravitational waves (GWs) are tensor-polarized, with two modes of polarization: plus ($h_+$) and cross ($h_\times$). The unmodeled GW burst analysis pipeline, \textit{BayesWave}, offers two tensor-polarized signal models: the elliptical polarization model ($E$) and the relaxed polarization model ($R$). Future expansion of the global GW detector network will enable more accurate studies of GW polarizations with GW bursts. Here a multi-detector analysis is conducted to compare the performance of $E$ and $R$ in characterizing elliptical and nonelliptical GW polarizations, using nonprecessing and precessing binary black holes (BBHs) respectively as representative synthetic sources. It is found that both models reconstruct the elliptical nonprecessing BBH signals accurately, but $E$ has a higher Bayesian evidence than $R$ as it is has fewer model parameters. The same is true for precessing BBHs that are reconstructed equally well by both models. However, for some events with high precession and especially with three or more detectors, the reconstruction accuracy and evidence of $R$ surpass $E$. The analysis is repeated for BBH events from the third LIGO-Virgo-KAGRA observing run, and the results show that $E$ is preferred over $R$ for existing detections. Although $E$ is generally preferred for its simplicity, it insists on elliptical polarizations, whereas $R$ can measure generic GW polarization content in terms of Stokes parameters. The accuracy of $R$ in recovering polarization content improves as the detector network expands, and the performance is independent of the GW signal morphology.