The goal of this research was to develop an EnergyLossEstimator for COSI.
In order to develop reliable and accurate reconstruction techniques for the measured events, it is essential to first gain a detailed understanding of the photons' possible interactions in the detector. In keeping with the type of instrument discussed in this work, photo absorption is mentioned only in the context of fully contained Compton scattering events, and Rayleigh scattering is ignored since its effects in a MEGA-type telescope are negligible.
All relevant photon interactions — photo-absorption, Compton scattering, and pair production — can be detected only via interactions resulting in energetic electrons and positrons in the instrument. Using the idealized interactions of gamma-ray photons with matter as a starting point, each section subsequently describes the multitude of effects that occur in a real instrument.
When an electron (or positron) passes through matter, it interacts with the atoms' Coulomb potentials: It undergoes many small-angle scatterings (Molière scattering) and loses energy mostly via ionization and bremsstrahlung.
Six processes contribute to the energy loss of electrons and positrons, two of which dominate in a MEGA-type Silicon tracker: ionization at low energies, bremsstrahlung at high energies. The electron energy for which both loss rates are equal is called the critical energy. It can be approximated by Ec = (800 MeV)/(Z + 1.2) ≈ 53 MeV for Si (Berger and Seltzer, 1964). Electron and positron interactions in MEGA are thus dominated by ionization. A welcome consequence of this is that MEGA tracks are not going to be "polluted" by a significant amount of additional bremsstrahlung hits, making the tracking much easier than for higher energy gamma-ray telescopes like GLAST. Other energy loss mechanisms are Møller scattering for electrons, Bhabha scattering for positrons respectively, positron annihilation before the positron is completely stopped, and δ-rays (knock-on electrons) (see e.g. Particle Data Group, 2004). None of the latter play an important role for MEGA.
Essentially, the problem is to predict the inital energy of an electron from its final energy and the energy loss of the electron through various processes.
For data generation, we'll be using Cosima. Cosima is the Geant4-based simulator of the medium-energy gamma-ray astronomy library MEGAlib. Within the MEGAlib framework, Cosima is designed to be a simple, fool-proof interface to Geant4 and to enable the simulation of most of the measurement scenarios encountered by X-ray and gamma-ray detectors in space and on Earth.
Naturally, if we can learn the energy loss of an electron through many of these events, maybe we can predict the inital energy using some ML approaches.