Simulation Methods

Overview: Monte Carlo simulations are critical tools for the increasingly complex questions being asked by current and planned experiments, and touch all aspects of experimental work in nuclear and particle physics. Accelerators and experimental apparatus are designed and optimized using sophisticated simulations. The extraction of fundamental parameters from experimental data uses simulations of the theoretical models for interactions between fundamental particles as well as simulations of the interactions between particles and the detector.

Large gaps exist with respect to needs for next-generation experiments. Given the significantly increased data volumes and physics needs, critical improvements are required to keep the total computational cost within available budgets and to improve the physics fidelity of the simulations. The challenge here is that scale of Monte Carlo simulation needed scales with the actual data recorded.

Significant research is going into improving current tools, either in their computational techniques (eg, use of GPUs for detector simulation) or in their physics completeness (eg, NNLO for matrix element generators). The incorporation of new machine learning and data science approaches and methodologies into the software ecosystem also holds great promise.

This is a vast area of research. We are looking for interested researchers (both potential mentors and potential students) on cross-experimental topics around software including:

  • Physics generators
  • Geant4 modernization
  • Fast simulation, fast generation approaches (eg, based on modern machine learning techniques)
  • Novel simulation approaches (eg, simulation based inference, differentiable programming techniques, etc)