Improved Measurements of IXPE Crab Polarization

IXPE X-ray polarization provides an important new probe of the geometry of the pulsar emission zone and of particle acceleration in the surrounding pulsar wind nebula (PWN). However extracting the pulsar signal in the presence of bright polarized PWN emission and mapping the detailed PWN polarization is a challenge, given IXPE’s modest HPD~20-30” spatial resolution. The conventional method defines an ‘off’ phase window as pure nebular emission, and subtracts this to isolate the pulsar variation (“on-off fitting”). We describe a more sensitive method that uses external measurements of the nebula structure and pulsar light curve to isolate their contributions to the phase- and spatially-varying polarization via a least-squares regression (“simultaneous fitting”). Analysis of simulation data shows that this method can reduce the image/pulse phase polarization uncertainties by ~2x while maintaining a ~30% better match to the true signal (goodness of fit). Applying “simultaneous fitting” to early IXPE Crab data results in a substantially improved nebular polarization map, and modestly improved phase-resolved polarimetry. The polarization map is affected by ‘polarization leakage’, due to correlation between the reconstructed event position and its polarization direction (EVPA). This introduces polarized fringes in the image on the scale comparable to the HPD — for compact sources like the Crab, this can substantially affect the nebular polarization measurements. We also introduce an improved method of correcting for these effects, using the anisotropic mirror PSFs and an iterative method to isolate the true image polarization. Applied to the first IXPE Crab data, we get better correction for polarization leakage, isolating the true polarization structure. Together, these methods improve our measurements of the polarization properties of the Crab pulsar and its surrounding PWN. These techniques will be applied to upcoming deeper IXPE Crab exposures, as well as other PSR/PWN data sets, such as Vela and MSH 15-5(2), and used to confront PSR/PWN models.