Emission beam geometry of selected pulsars derived from average pulse polarization data

By fitting the classical rotating vector model (RVM) to high-quality polarization data for selected radio pulsars, we find the inclination of the magnetic axis to the spin axis, alpha, as well as the minimum angle between the line of sight and the magnetic axis, beta, for 10 objects. We give a full treatment of statistical errors in the fitting process. We also present a dictionary and conversion table of various investigators' geometric definitions to facilitate future comparisons. We compare our results with other RVM fits and with empirical/geometrical (E/G) approaches, and we examine the strengths and weaknesses of RVM fits and E/G investigations for the determination of pulsar emission beam geometry. Our fits to B0950+08 show that it is an orthogonal rotator with the main and interpulse radiation emitted from opposite magnetic poles, whereas earlier RVM fits indicated that it is an almost-aligned, single-magnetic pole emitter. We demonstrate that low-level emission across a wide longitude range, when properly weighted in the RVM fit, conclusively favors the former scenario. B0823+26 is also an orthogonal rotator. We find that B1929+10 emits into its wide observed range of longitudes from portions of a single cone that is almost aligned with the spin axis. This result agrees with other RVM fits but conflicts with the E/G findings of Rankin & Rathnasree. We determine that convergent RVM solutions can be found only for a minority of pulsars: generally those having emission over a relatively wide longitude range, and especially those pulsars having interpulse emission. In pulsar B0823+26, our preferred fit to data at all longitudes yields a solution differing by several sigma from a fit to the main pulse/postcursor combination alone. For pulsar B0950+08, separate fits to the main pulse region, the interpulse region, and our preferred fit to almost all longitudes, converge to results differing by several times the formal uncertainties. These results indicate that RVM fits are easily perturbed by systematic effects in polarized position angles and that the formal uncertainties significantly underestimate the actual errors.