A GPU-Accelerated Monte Carlo Track-Structure Simulation Tool for Silicon Detectors
Abstract
Purpose
Track-structure Monte Carlo simulations are essential for modeling stochastic energy deposition at micro- and nanoscales relevant to radiobiology and for validating experiments in solid-state microdosimeters such as silicon-on-insulator (SOI) devices. However, the high computational cost of existing tools, e.g., Geant4-DNA MuElec, limits their use. This work develops gMicroMC_Si, an open-source efficient track-structure simulation tool for silicon materials to support SOI microdosimetric studies.
Methods
Inelastic interactions in silicon of protons and electrons were modeled using dielectric function theory, and elastic scattering of electrons was described using a screened Rutherford differential cross section. The gMicroMC_Si framework was implemented on NVIDIA GPUs using CUDA, with event-level parallelization of primary and secondary electron transport and table-based sampling of interaction cross sections and scattering kinematics. To validate our simulations, stopping powers were computed and compared with Geant4 MicroElec, and secondary electron energy spectra for 1 MeV protons and 10 keV electrons were compared with Geant4-DNA MuElec. An SOI microdosimeter was modeled, and microdosimetric spectra and lineal energy distributions in a 131 MeV pencil beam scanning (PBS) proton field were calculated and compared with experimental measurements.
Results
Stopping powers and secondary-electron energy spectra calculated by gMicroMC_Si showed good agreement with Geant4-DNA MuElec, with logarithmic root-mean-square deviations of 0.16 below 2 keV for electrons and 0.13 below 1.6 keV for protons. Benefiting from GPU-based parallelization, simulations achieved a throughput of approximately 10³ primary protons per minute for this condition. In SOI microdosimeter, simulated peak lineal energies closely matched the experimental values: 1.33 vs 1.20 keV/μm at the proximal region (110 mm), 3.8 vs 2.42 keV/μm at the Bragg peak center (122 mm), and 6.89 vs 6.79 keV/μm at the distal edge (129 mm).
Conclusion
gMicroMC_Si enables efficient and accurate track-structure simulations in silicon, providing a practical tool for microdosimetry in advanced radiation therapy applications.