Domain, Specialties : Neutronique
Keywords:
Nuclear and particle physics, quantum mechanics, neutronics
Research Unit : DPhN / LEARN
Summary
Full description
The Monte Carlo simulation of particle transport has become central in many fields such as fundamental physics, medical physics or space applications just to name a few. In order to overcome the computation time inherent to this method, so-called “variance reduction” techniques are developed, whose objective is to “push” the particles towards the detector(s) without inducing bias on the computation results. Recently, a variance reduction method called Adaptive Multilevel Splitting (AMS) [1,2], which focuses on rare events, has been designed and implemented in different calculation codes, such as the Monte Carlo code TRIPOLI4 developed at CEA or the Monte Carlo code Geant4 developed at CERN by an international collaboration [3]. This method offers very interesting perspectives because it allows to preserve the correlations between the particles contributing to the measurement, which is not possible with standard variance reduction technics such as Importance Sampling.
This internship topic proposes to evaluate the interest of this approach in the context of fundamental physics, where simulations seek to detect rare events, as for example in experiments aiming to detect the coherent neutrino scattering at nuclear reactors (CEvNS). Indeed, these experiments regularly use coincidence detection methods to increase the signal-to-noise ratio (as an example, one can imagine the detection of a neutron in detector A and a photon in detector B). They represent a challenge in terms of numerical calculation because their simulation requires the sampling of an often-prohibitive number of events (so that a very small fraction of them contribute to the measurement).
The subject of this internship is therefore to use the AMS implemented in the Geant4 code and to optimize its parameters in so-called “multi-particle/multi-detector” configurations, in order to be able to simulate rare correlated events. The trainee will have to finalize the developments already undertaken (see [4]) and to verify their efficiency on different simple representative physical cases. Once this work is completed, two avenues can be pursued, depending on the candidate’s interest:
1) either the candidate will be interested in deploying this method in the Geant4 code itself, so that it can be made available to all physicists for inclusion in the official version,
2) or the candidate will set up an experimental setup to compare the results obtained with AMS with experimental data.
[1] F. Cérou and A. Guyader, Stochastic Analysis and Applications 25(2):417-443, 2007
[2] H. Louvin et al., EPJ Web of Conferences, Vol153, EDP Sciences, 2017
[3] https://geant4.web.cern.ch
[4] L. Thulliez, B. Mom et E. Dumonteil, Nuclear Inst. and Methods in Physics Research, A 1051 (2023) 168190
Location
Saclay
Internship conditions
- Internship duration: 6 months
- Level of study: Bac+4/5
- Training: Ingenieur/Master
- Continuation in PhD thesis: No
- Application deadline: 24 octobre 2024
Experimental skills
Language : English
Useful methods and technics:
Variance reduction method
Computer languages and software:
Python, C++, ROOT (est un plus/is an asset), Geant4 (est un plus/is an asset)