**Title of Presentation:** Event Analysis for the Gamma-ray Large Area Space Telescope

**Primary (Corresponding) Author:** Robin D Morris

**Organization of Primary Author:** Research Institute for Advanced Computer Science

**Co-Authors:** Johann Cohen-Tanugi

**Abstract**: GLAST, the Gamma-ray Large Area Space Telescope is a next-generation gamma-ray observatory, to be launched in late 2007. It will observe the gamma-ray sky in the range 20MeV-300GeV. It's main instrument is the Large Area Telescope (LAT). One subsystem of the LAT is a pair-production tracker, interleaving tungsten conversion layers with silicon microstrip layers. Incident photons convert in the tungsten layers, producing electron-positron pairs. These charged particles traverse the detector, producing "hits" on the silicon microstrips. The charged particles are also subject to multiple scattering as they traverse subsequent tungsten layers, and produce secondary charged particles and photons. The secondary charged particles also produce hits, and further electrons and photons. The photons may also produce hits and further charged particles.

By analysing the elementary interactions of a particle with a single foil, we can construct the probabilities of each outcome. For example, we can construct the probability of an incident electron producing zero secondaries, one secondary, etc, together with the distribution of the energy of the secondary. By analysing the hits at subsequent layers, we can enumerate a list of hypotheses as to the physical processes that occurred in the tungsten foils. By combining the elementary distributions with the physical hypotheses, we can use Markov chain Monte Carlo (MCMC) methods to compute the posterior distribution over the hypotheses, and consequently, the distribution over the energy and direction of the incident particle.

We describe work-in-progress, showing results for muons incident on the lat, and electrons and photons for restricted physics, and compare the estimates from our new algorithm with the conventional Kalman filter-based estimates.