Bayesian Optimisation of Breit-Wheeler Pair Production in Simulated Laser Experiments

Christopher Arran · 2025

Read Bayesian Optimisation of Breit-Wheeler Pair Production in Simulated Laser Experiments by Christopher Arran (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-16533.txt.

Introduction

Christopher Arran and co-authors use Bayesian optimization to maximize electron positron pair production in simulated laser experiments in a 2025 paper. Intense lasers can create pairs from gamma rays in strong fields. The most promising near term setup uses wakefield electron beams to make gamma rays, but shot to shot jitter in pointing and timing makes collisions rare and hard to arrange.

Monte Carlo simulations modeled those practical problems, with a sampling trick for rare events. A Gaussian process search then tuned laser spot size, beam energy, and stand off distance. The best stand off grew with jitter, and pair optima differed from energy optima.

Catalog record

CREATORChristopher Arran
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-16533

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Bayesian Optimisation of Breit-Wheeler Pair Production in Simulated Laser Experiments Christopher Arran Abstract High laser intensities enable the production of electron-positron pairs from bright gamma rays passing through strong fields. Potentially the most promising approach for all-optical experiments in the near term uses dense but higher divergence electron beams from laser wakefield acceleration to produce gamma rays through inverse Compton scattering. Achieving many-photon collisions between these gamma rays and the high intensity laser pulse in practice is extremely difficult, however, due to significant shot-to-shot jitter in laser pointing and timing. We model these practical difficulties using simulated Monte-Carlo experiments. By using a more efficient algorithm for sampling infrequent pair production with particle splitting, we enable the exploration of a multi-dimensional parameter space. Using Gaussian Process Regression we then efficiently find optimal conditions for maximising pair production by changing the laser spot size, the energy in the colliding beam, and the stand-off distance between the laser wakefield accelerator and the focus of the colliding laser pulse. We find that the optimal stand-off distance increases with the degree of laser jitter and that the best conditions for producing electron-positron pairs are not the same as the best conditions for maximising the energy in the gamma rays. With 100 J of laser energy, we estimate rates of pair production of around 1 pair per 100 electrons are achievable even with jitter of 10s of microns and 10s of femtoseconds. 1 Introduction Laser facilities continue to push the intensity frontier to new records, in recent years achieving peak powers of 10\,\mathrm{PW} [ 1 ] and peak intensities of 10^{23}\,\mathrm{Wcm^{-2}} [ 2 ] . Several studies have explored the possibility of creating matter from pure light in sufficiently strong laser fields [ 3 , 4 , 5 ] . In the centre-of-mass frame of this many-photon interaction, the field strength approaches the Schwinger field [ 6 ] , enabling the production of position-electron pairs in vacuum through the non-linear Breit-Wheeler process [ 7 ] . Previous experiments have been able to reach this regime using a 47\,\mathrm{GeV} electron beam from SLAC colliding with an 0.4\,\mathrm{TW} laser pulse to produce around 100 positron-electron pairs from around 10^{10} electrons [ 8 ] . This first produces bright \gamma -rays from inverse Compton scattering [ 9 ] , before the \gamma -rays themselves interact with the laser field. Another approach has been to use the strong atomic fields present within periodic crystal lattices [ 10 , 11 ] , with a 200\,\mathrm{GeV} electron beam driving trident pair production in a germanium crystal. Now, several laser facilities aim to utilise unprecedented laser powers and produce positron-electron pairs in vacuum by using laser-driven plasma wakefield accelerators [ 12 , 13 ] . Conducting these experiments in practice is significantly more complicated than may be imagined from idealised simulations, with the interactions requiring femtosecond- and micron-level synchronisation and alignment between the laser and electron beams. In order to successfully produce positron-electron pairs in vacuum in laser-driven experiments, this spatial and temporal jitter must be accounted for, typically by running many simulations over the range of possible configurations. However, running this many simulations is often computationally expensive, especially when the chances of producing a positron-electron pair can be vanishingly small with current-generation laser facilities. In order to determine the best experimental parameters from a large domain of options, it is necessary to increase the efficiency both of calculating the pair production rates, and of exploring a multi-dimensional parameter space. Here, we find optimum parameters for electron-positron pair production in all-optical experiments by employing two techniques for performing many simulations with higher computational efficiency. First, we describe a particle splitting algorithm and show how it requires changes to the way we model pair production. This accurately calculates the rate of pair production for rare events without significantly increasing the runtime, allowing us to rapidly conduct many simulated experiments, including the effect of jitter. Second, we use a Bayesian optimisation scheme built on Gaussian process regression to efficiently explore the parameter space and find the maximum rate of pair production. This scheme is significantly faster than a brute force grid search and builds a fast surrogate model for estimating the rate of pair production across the domain. We show that more position-electron pairs can be produced by combining the tightest possible laser focus with a few-cm-scale stand-off distance between the laser wakefield accelerator and the interaction point. This ameliorates the effect of jitter while retaining the high field strength required for pair production. 2 Pair Production and Particle Splitting 2.1 Pair Production Algorithm In both Monte-Carlo and Particle-In-Cell codes, the emission of high energy photons and the production of positron-electron pairs are treated as discrete random and independent events which can occur in any fixed time interval. As such, they can be modelled using a Poisson distribution, where the average number of emissions per particle in a given timestep \delta t are given by: \lambda_{\gamma}=\frac{dN_{\gamma}(\chi_{e})}{dt}\delta t\\ \lambda_{e^{+},e^{-}}=\frac{dN_{e^{+},e^{-}}(\chi_{\gamma})}{dt}\delta t, (1) where the electron and photon quantum parameters are given by \chi_{e}=|F_{\mu\nu}p_{e}^{\nu}|/m_{e}E_{s} and \chi_{\gamma}=|F_{\mu\nu}\hbar k_{\gamma}^{\nu}|/m_{e}E_{s} , for a Schwinger field E_{s}=m_{e}^{2}c^{3}/e\hbar . The probability of emitting n particles in a given timestep is then P(n|\lambda)=\frac{\lambda^{n}e^{-\lambda}}{n!} (2) Often, in order to reduce the computational expense of sampling from a Poisson distribution at every timestep, codes use an optical depth approach. If the emission rate is sufficiently low, the chances of two or more emissions in a single timestep is negligible and the probability of a single particle being emitted on a timestep before t_{N} becomes P(t_{0}<t<t_{N})\approx 1-\exp\left[-\sum^{N}_{n=0}\lambda(t_{n})\right] . Now a uniformly distributed random number in the range 0\leq r<1 is chosen to represent P(t) and an emission occurs on the first timestep when 1-\exp\left[-\sum_{n}\lambda(t_{n})\right]\leq r , which is when the accumulated optical depth reaches \sum_{n}\lambda(t_{n})\geq-\ln(1-r) . Only one random number must be chosen for each emission event, calculated the moment a particle is created or the moment after an emission takes place, and a single running total tracks the optical depth. The optical depth method therefore efficiently models rare emissions which occur singly and consecutively. This model is particularly relevant for pair production, as a photon ceases to exist after emitting a positron-electron pair, so more than one emission is impossible. In fact, emitting a photon also changes an electron’s trajectory, due to radiation reaction, and so emitting multiple photons in a single timestep using the same value for \chi_{e} is also unphysical if the photon energy is sufficiently high. The optical depth method is therefore generally very well suited for the problems we are modelling. However, if emissions are very rare, such as for Breit-Wheeler pair production, a tremendously large number of particles must be used to model the likelihood of even a single emission, with the vast majority of these particles not participating in the interaction at all. Each of the particles must be tracked through the simulation even if they do not contribute at all to the creation of pairs. In this situation, despite the strengths of the optical depth method, modelling single rare emissions is extremely computationally inefficient. 2.2 Particle Splitting Algorithms We therefore propose introducing particle splitting into the calculation of Breit-Wheeler pair production in Monte-Carlo and PIC codes. In PIC codes,

Read the complete text