Neural network-based deconvolution for GeV-Scale Gamma-Ray Spectroscopy

TL;DR

This work tackles the problem of precise spectral reconstruction for GeV-scale gamma rays by coupling a Monte Carlo–optimized spectrometer with a two-stage neural network. A denoising autoencoder suppresses statistical noise in lepton spectra, followed by a U‑Net deconvolution that recovers the incident gamma spectrum from denoised data, mitigating the ill-posedness of the inverse problem defined by the MC-generated response matrix $\,mathcal{A}$. Evaluation on synthetic and simulated data shows superior RMSE, PSNR, and SSIM performance compared with traditional methods, with predictive uncertainty quantified via a 95% Bayesian credible interval. The approach advances gamma-ray diagnostics for SFQED experiments and compact photon sources, and lays groundwork for future multi-parameter, double-differential spectrometry in high-energy photonics.

Abstract

High-energy gamma-ray spectroscopy is crucial for studying and advancing the application of high-energy photons in areas like strong-field physics, high-energy-density science, and laboratory astrophysics. However, high-energy gamma-ray spectroscopy in the multi-MeV to GeV range faces significant challenges in precise spectral reconstruction. This study presents a machine learning-based inversion approach that combines a spectrometer design with advanced deconvolution algorithms. We develop a gamma-ray spectrometer optimized through Monte Carlo simulations for maximum positron yield and minimal noise. A two-stage neural network framework is proposed based on the structure of the spectrometer: a denoising autoencoder suppresses statistical noise in measured positron spectra, while a U-Net architecture solves the ill-posed inverse problem to reconstruct incident gamma spectra. This approach establishes a new methodology for gamma-ray diagnostics in strong-field QED experiments and high-energy photon sources.

Figures (8)

• Figure 1: Top view schematic of the gamma-ray spectrometer, consisting of the converter, shielding, collimator, magnet, and detector.
• Figure 2: (a) The normalized spectra of positrons produced by a 1 GeV monochromatic gamma-ray beam after propagating through different 500 $\mu$m thick materials; (b) Spectra of positrons generated when a pencil-shaped monochromatic gamma-ray beam of energy 1 GeV penetrates tungsten targets of thicknesses 100 $\mu$m, 500 $\mu$m, 1 mm, 3 mm, and 5 mm. The x-axis represents the incident photon energy (in logarithmic scale), while the y-axis shows the number of positrons produced per bin of incident gamma-ray photon energy. As the tungsten layer thickness increases, the full width at half maximum (FWHM) of the peak decreases to 1 mm point, and then starts to increase.
• Figure 3: Time-integrated fluence distribution for (a) photons, (b) positrons, and (c) electrons resulting from the propagation of a monoenergetic gamma-ray beam. (d) The transverse distribution of electrons, positrons, and photons at the back of the spectrometer. The color bar represents particles/gamma photons per square centimeter.
• Figure 4: Schematic of the overall reconstruction workflow, organized into two sequential stages: (1) Denoising—a supervised autoencoder is trained on a hybrid dataset to suppress measurement noise while retaining key spectral features; (2) Deconvolution—the trained network, built on a U-Net architecture. This figure can be viewed as illustrating the forward and inverse processes of Eq.\ref{['eq:4']}.
• Figure 5: Multiple response functions were generated by simulating the monochromatic gamma-ray response using a gamma-ray spectrometer system configured as shown in Fig.\ref{['fig:1']} within the Geant4 framework. Vertical lineouts of this figure show the positron spectrum generated at a specific gamma-ray energy. (a) Ideal response of monochromatic gamma-rays interacting with a 1 mm tungsten converter. (b) Response function of a gamma-ray spectrometer with a 1 mm tungsten converter.