How many pixels are there in a polarized pulsar timing array map?

The standard approach to searching for gravitational wave signatures in pulsar timing array (PTA) data has been to compare the theoretical Hellings and Downs (HD) curve with the observed correlations in pulsar timing residuals as a function of angular separation on the sky between pulsar pairs. While this approach has successfully produced evidence for the presence of nanohertz-wavelength gravitational waves, it does not, on its own, produce any directional information. It is also insensitive to the polarization of the gravitational waves. An alternative approach is to construct maps of the gravitational wave distribution on the sky. In this paper, we present a simple quadratic estimator of the gravitational wave power as a function of direction on the sky that is sensitive to the polarization state of the wave. In this way, we describe the full, $S_2 \times S_2$, state-space of a polarized gravitational wave background across the sky and the Poincar\'e sphere describing polarization. A natural question arises from this perspective: what is the resolution of a polarized sky-map, i.e. effectively how many independent pixels can a such a map contain? In other words, how many distinct gravitational waves can a PTA, in principle, distinguish? It turns out the answer is finite, and is approximately $N_{\rm res} = 16 \times 2 = 32$ per frequency, where 16 is the number of resolvable sky-positions and 2 is the number of distinct polarization states. This corresponds to an angular resolution of $58^\circ$, which can be achieved by a PTA with more than $N_{\rm pulsar} \gtrsim 20$ pulsars. We demonstrate that the variance of the map is equivalent to the HD significance, while for a single point source, a 3-$\sigma$ HD signal corresponds to a 5.2-$\sigma$ map significance.