Differentiable programming for particle physics simulations

TL;DR

This paper introduces differentiable programming techniques using adjoint sensitivity methods for particle physics simulations, enabling efficient inverse problem solving without approximating transport dynamics, implemented in a C++ library.

Contribution

It presents a novel application of adjoint sensitivity methods to backward Monte-Carlo particle simulations, facilitating derivative-based inverse problem solutions.

Findings

Developed algorithms for differentiable particle transport simulations.

Implemented these algorithms in the NOA C++17 library.

Demonstrated potential for improved inverse problem solving in particle physics.

Abstract

We describe how to apply adjoint sensitivity methods to backward Monte-Carlo schemes arising from simulations of particles passing through matter. Relying on this, we demonstrate derivative based techniques for solving inverse problems for such systems without approximations to underlying transport dynamics. We are implementing those algorithms for various scenarios within a general purpose differentiable programming C++17 library NOA (github.com/grinisrit/noa).