Highly-parallelized simulation of a pixelated LArTPC on a GPU

Abstract

The rapid development of general-purpose computing on graphics processing units (GPGPU) is allowing the implementation of highly-parallelized Monte Carlo simulation chains for particle physics experiments. This technique is particularly suitable for the simulation of a pixelated charge readout for time projection chambers, given the large number of channels that this technology employs. Here we present the first implementation of a full microphysical simulator of a liquid argon time projection chamber (LArTPC) equipped with light readout and pixelated charge readout, developed for the DUNE Near Detector. The software is implemented with an end-to-end set of GPU-optimized algorithms. The algorithms have been written in Python and translated into CUDA kernels using Numba, a just-in-time compiler for a subset of Python and NumPy instructions. The GPU implementation achieves a speed up of four orders of magnitude compared with the equivalent CPU version. The simulation of the current induced on 10^3 pixels takes around 1 ms on the GPU, compared with approximately 10 s on the CPU. The results of the simulation are compared against data from a pixel-readout LArTPC prototype.

Differentiable simulation of a liquid argon time projection chamber

“…The spatial resolution in the drift direction is below 2 mm. Further description of the LArTPC configuration used in this work can be found in [24].…”

“…The simulator developed for the DUNE LArTPC prototype is called larnd-sim [24]. Energy depositions are represented as line segments ('particle segments') defined by 3D start and end positions, (x, y, z), and an associated energy, E. The simulator models the production of ionization electrons due to these energy depositions, as well as the drifting of electrons to the pixel readout at each anode plane.…”

“…Peak memory usage in our simulator occurs during the computation of the signal on each pixel (the Current accumulation stage of the simulation), which is also the most computationally intensive step of our snapshot of larnd-sim [24]. In our case, the high memory usage is due to the construction of a very large, dense matrix.…”

“…A simultaneous correction of all detector processes and physics models would be an ideal new paradigm for detector calibration. However, this has not been previously achieved due to limitations of the existing This work specifically targets the improvement of parameter-based calibration for the simulator in [24], and is restricted to optimization of the models included there. We therefore emphasize a few caveats of our results.…”