Linear Polarimetry with γ rightarrow e^+e^- conversions

$\gamma$ rays are emitted by cosmic sources by non-thermal processes that yield either non-polarized photons, such as those from $\pi^0$ decay in hadronic interactions, or linearly polarized photons from synchrotron radiation and the inverse-Compton up-shifting of these on high-energy charged particles. Polarimetry in the MeV energy range would provide a powerful tool to discriminate among "leptonic" and "hadronic" emission models of blazars, for example, but no polarimeter sensitive above 1\,MeV has ever been flown into space. Low-$Z$ converter telescopes such as silicon detectors are developed to improve the angular resolution and the point-like sensitivity below 100 MeV. We have shown that in the case of a homogeneous, low-density active target such as a gas time-projection chamber (TPC), the single-track angular resolution is even better and is so good that in addition the linear polarimetry of the incoming radiation can be performed. We actually characterized the performance of a prototype of such a telescope on beam. Track momentum measurement in the tracker would enable calorimeter-free, large effective area telescopes on low-mass space missions. An optimal unbiased momentum estimate can be obtained, in the tracker alone, based on the momentum dependence of multiple scattering, from a Bayesian analysis of the innovations of Kalman filters applied to the tracks.