Speaker
Description
Liquid argon time projection chambers (TPCs) are widely used in particle detection. High quality physics simulators have been developed for such detectors in a variety of experiments, and the resulting simulations are used to aid in reconstruction and analysis of collected data. However, the degree to which these simulations are reflective of real data is limited by the knowledge of the parameters of the corresponding detector system, which may be difficult to directly measure. We present a differentiable simulation of the DUNE near detector liquid argon TPC, and use it to perform a data-driven, gradient-based calibration of several relevant detector parameters. We describe the advantages and limitations of such an approach and highlight the challenges and procedures of making such a differentiable simulator while retaining the physics quality of the standard, non-differentiable version. Finally, we discuss applications of the differentiable simulator beyond gradient-based calibration.
Consider for long presentation | No |
---|