Machine Learning Simulation Techniques for the LZ Dark Matter Experiment

Large-scale particle physics experiments like the LUX-ZEPLIN (LZ) dark matter experiment rely on sophisticated simulation software to model behavior within the detector. LZ's full simulations model photon propagation and detector electronics response (DER), but their computational expense limits their use to targeted studies, mandating the use of faster but less detailed parametric simulations for background modeling. A significant speed-up in LZ full simulations would enable a new level of detail in background modeling, incorporating nuances such as the time profile of detected pulses and improving understanding of cuts aimed at background reduction using such quantities. We present efforts to accelerate LZ's DER simulations by training a machine learning generative model to quickly and accurately reproduce existing outputs.