arxiv: 2604.25489 · v1 · submitted 2026-04-28 · ⚛️ physics.acc-ph · cs.LG

Recognition: unknown

Adaptable phase retrieval for coherent transition radiation spectroscopy based on differentiable physics information

Alexander Debus, Andreas Doepp, Arie Irman, Jeffrey Kelling, Maxwell LaBerge, Michael Bussmann, Ritz Ann Aguilar, Ulrich Schramm, Zewu Bi

Pith reviewed 2026-05-07 14:08 UTC · model grok-4.3

classification ⚛️ physics.acc-ph cs.LG

keywords coherent transition radiationphase retrievalgradient descentdifferentiable forward modelelectron bunch profilelaser-plasma acceleratordiagnostics

0 comments

The pith

A differentiable physics model allows gradient descent to retrieve electron bunch profiles from coherent transition radiation spectra by optimizing phase under real-space constraints.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper aims to show that phase retrieval for coherent transition radiation can be made more adaptable by using a differentiable forward model in a gradient-based optimization. Instead of relying on explicit inverse steps like in older algorithms, the new method keeps the measured spectral amplitudes fixed and adjusts only the phases while applying physical priors on the bunch shape in real space. This setup matches the accuracy of standard methods on test cases with complex bunch profiles. Readers might care because it simplifies adding real experimental complications and supports future combinations of different measurements for better diagnostics in accelerators.

Core claim

The work establishes that a phase-only gradient descent algorithm, using a differentiable model of coherent transition radiation generation, successfully reconstructs the longitudinal electron bunch profile. It does so by treating the measured frequency-domain amplitude as an unchangeable constraint and varying the phase to minimize inconsistencies with chosen real-space physical priors, such as smoothness or support limits on the bunch.

What carries the argument

The phase-only gradient descent (GD-Phase) that enforces spectral amplitude as a hard constraint and optimizes Fourier phase with real-space priors through a differentiable forward model.

If this is right

The method achieves reconstruction fidelity comparable to Gerchberg-Saxton algorithms for synthetic multi-peaked and modulated profiles.
It permits direct inclusion of arbitrary differentiable experimental effects without needing custom inverse operators.
The framework serves as a starting point for integrating constraints from multiple diagnostics and performing uncertainty quantification.
This supports extension to higher-dimensional reconstructions and multimodal data in realistic accelerator settings.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The approach could make it easier to handle uncertainties in bunch length measurements for optimizing laser-plasma accelerators.
Similar differentiable frameworks might apply to other ill-posed inverse problems in radiation diagnostics where forward models are known.
Combining this with automatic differentiation libraries could speed up real-time analysis during experiments.

Load-bearing premise

The chosen real-space physical priors must be sufficient to resolve phase ambiguities, and the forward model must accurately represent the experiment in a differentiable way.

What would settle it

A case where the GD-Phase reconstruction deviates significantly from the true bunch profile on synthetic data that includes known experimental distortions, while traditional methods succeed.

Figures

Figures reproduced from arXiv: 2604.25489 by Alexander Debus, Andreas Doepp, Arie Irman, Jeffrey Kelling, Maxwell LaBerge, Michael Bussmann, Ritz Ann Aguilar, Ulrich Schramm, Zewu Bi.

**Figure 1.** Figure 1: Schematic of the CTR spectroscopy setup. A relativistic electron bunch (solid blue) impinges on a metallic foil, emitting CTR into a cone (orange). A parabolic mirror collects and directs the radiation; a turning mirror steers it into the CTR spectrometer for far-field spectral measurement. The electron beam (blue dashed trajectory) is deflected by a dipole magnet and transported to an electron spectromete… view at source ↗

**Figure 2.** Figure 2: Synthetic benchmark suite. Top: representative longitudinal profiles ρ(z): (1) double, (2) triple, (3) strongly modulated (spike train), and (4) Blackman–Nuttall window-like bunch, respectively. Middle: corresponding amplitude spectra |F(k)|. Bottom: corresponding intensity spectra Imeas(k) ∝ |F(k)| 2 in log-scale to emphasize the prominent features. The phase-retrieval task is to reconstruct ρ(z) given on… view at source ↗

**Figure 3.** Figure 3: Reconstruction stability analysis across R = 32 restarts for the relatively simple synthetic benchmark distributions (Blackman-Nuttall and Double-spike). Columns correspond to the three algorithms (GS, GD-Phase, GD-Amp), and rows represent the bunch profiles ρ(z) and corresponding spectra F(k). In each panel, the true profile (Truth) is shown in black. The Best reconstruction (lowest spectral loss) is plot… view at source ↗

**Figure 4.** Figure 4: Reconstruction for the complex synthetic benchmark distributions (Triple-spike and Highly Modulated bunch) similar to view at source ↗

**Figure 5.** Figure 5: Statistics of the ensemble quality for all bunch families. Histogram of final spectral loss Lspec over R = 32 restarts for all algorithms (top). GD-Phase (green) consistently achieves the lowest loss mode. GD-Amp with GS initialization (orange) shows slightly better spectral loss than GS (red). GD-Amp from random restarts (blue) consistently shows the highest low mode. Corresponding correlation C versus sp… view at source ↗

**Figure 6.** Figure 6: PCA landscape of reconstruction ensembles. For each bunch distribution, we project canonicalized longitudinal profiles ρ(z) onto the first two principal components (PC1–PC2) of a PCA model fit to the union of the ground truth profile (black cross), the top lowest-loss restarts from each algorithm, and the seeded GD-Amp run initialized from the best GS solution (orange diamond). Clouds correspond to the top… view at source ↗

**Figure 7.** Figure 7: Summary of the best reconstructions (lowest spectral loss) of all longitudinal bunch profiles (top), their corresponding normalized spectra in logarithmic scale (middle), and their phase gradients (bottom) for the GS and GD-Phase algorithms. The black curve shows the ground truth while the red and green curves represent the GS and GD-Phase plots, respectively. The shaded regions indicate k-intervals where … view at source ↗

**Figure 8.** Figure 8: Convergence dynamics for the triple-spike bunch. Left: Spectral loss vs. iteration for GS. Center: Breakdown of the penalty components (Positivity, Support, Smoothness, Imaginary) for GD-Phase. Note that GD-Phase prioritizes satisfying the positivity and support constraints early in the optimization. Right: Total loss vs. iteration for GD-Amp, highlighting instability spikes characteristic of the amplitude… view at source ↗

read the original abstract

Coherent transition radiation (CTR) spectroscopy is a critical diagnostic for characterizing the longitudinal structure of relativistic electron bunches in laser-plasma and conventional accelerators. In practice, recovering the bunch profile from a measured CTR spectrum is an ill-posed phase-retrieval problem. Traditionally, this is addressed using Gerchberg-Saxton (GS)-type iterative algorithms. However, these implementations often rely on explicit inverse propagators, making them difficult to adapt to sophisticated experimental forward models. In this work, we introduce a flexible gradient-based framework for CTR phase retrieval. By leveraging a differentiable forward model, we propose a phase-only gradient descent (GD-Phase) approach that enforces the measured spectral amplitude as a hard constraint while optimizing the Fourier phase under physical real-space priors. Using synthetic CTR spectra spanning multi-peaked and strongly modulated profiles, we benchmark GD-Phase against traditional GS and a real-space amplitude-parametrized gradient descent (GD-Amp) algorithm. Unlike traditional methods, this formulation allows for the seamless inclusion of arbitrary differentiable experimental effects into the reconstruction loop. We demonstrate that this physics-informed approach not only reproduces the fidelity of GS methods but also establishes a robust baseline for incorporating multi-diagnostic constraints and uncertainty quantification. This enables the systematic extension to higher-dimensional, multimodal, and uncertainty-aware diagnostics, facilitating fast and scalable phase retrieval in realistic experimental settings.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This paper gives a practical differentiable gradient-descent alternative to Gerchberg-Saxton for CTR phase retrieval that should make it easier to fold in real experimental effects.

read the letter

The main advance is the GD-Phase method: a phase-only gradient descent that keeps the measured spectral amplitudes as a hard constraint and optimizes the Fourier phase using a differentiable forward model plus real-space priors. This sidesteps the need for explicit inverse propagators that make standard GS hard to adapt when the experimental setup gets complicated. They benchmark it on synthetic CTR spectra with multi-peaked and strongly modulated profiles, showing it reaches similar fidelity to GS while also comparing against an amplitude-parametrized GD baseline. The ability to drop in arbitrary differentiable effects without rewriting the solver is the practical payoff, and the setup naturally supports adding multi-diagnostic constraints or uncertainty later. That part is useful and cleanly motivated. The abstract and stress-test description give no quantitative error bars or detailed metrics, so it is still hard to judge how large the practical gains are or how sensitive the method is to model mismatch. All the tests are synthetic, which is a reasonable first step but leaves open questions about real noise, calibration errors, and whether the chosen priors are always sufficient to break the phase ambiguities. The differentiability requirement itself could become a limitation if some experimental effects are only approximately differentiable. This is aimed at people doing beam diagnostics in laser-plasma or conventional accelerators who need phase retrieval that can grow with their setup. A reader working on physics-informed optimization or phase retrieval in general would get value from the framework and the direct comparison to GS. It deserves peer review because the core idea is new, the synthetic validation is internally consistent, and the limitation it targets is real even if the current evidence is still preliminary.

Referee Report

1 major / 0 minor

Summary. The manuscript introduces a flexible gradient-based framework for phase retrieval in coherent transition radiation (CTR) spectroscopy. It proposes a phase-only gradient descent (GD-Phase) algorithm that leverages a differentiable forward model to enforce measured spectral amplitudes as hard constraints while optimizing the Fourier phase under physical real-space priors. The approach is benchmarked on synthetic CTR spectra with multi-peaked and strongly modulated profiles against traditional Gerchberg-Saxton (GS) iterations and a real-space amplitude-parametrized gradient descent (GD-Amp) variant, with claims that GD-Phase reproduces GS fidelity while enabling seamless inclusion of arbitrary differentiable experimental effects, multi-diagnostic constraints, and uncertainty quantification.

Significance. If the quantitative validation holds, this work offers a valuable methodological advance for ill-posed phase-retrieval problems in accelerator beam diagnostics. The differentiable-physics formulation directly addresses the rigidity of GS-type methods that depend on explicit inverse propagators, providing a natural route to incorporate complex forward models. The emphasis on hard amplitude constraints combined with real-space priors, together with the synthetic benchmarking strategy, constitutes a solid baseline for extensible, uncertainty-aware diagnostics in laser-plasma and conventional accelerator settings.

major comments (1)

[Abstract] Abstract: the benchmarking claim that GD-Phase 'reproduces the fidelity of GS methods' on synthetic multi-peaked and modulated profiles is stated without any quantitative metrics, error bars, fidelity measures (e.g., RMS error, overlap integrals), or tabulated comparisons; this absence makes it impossible to evaluate whether the reproduction is statistically meaningful or merely qualitative.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their positive assessment of the manuscript and for the constructive comment on the abstract. We address the point below and will make the corresponding revision.

read point-by-point responses

Referee: [Abstract] Abstract: the benchmarking claim that GD-Phase 'reproduces the fidelity of GS methods' on synthetic multi-peaked and modulated profiles is stated without any quantitative metrics, error bars, fidelity measures (e.g., RMS error, overlap integrals), or tabulated comparisons; this absence makes it impossible to evaluate whether the reproduction is statistically meaningful or merely qualitative.

Authors: We agree that the abstract would be improved by including quantitative support for the fidelity claim. The body of the manuscript (Section 3) already contains the detailed synthetic benchmarks, including RMS errors, overlap integrals, and direct comparisons of GD-Phase against GS and GD-Amp across the multi-peaked and modulated test cases. We will revise the abstract to incorporate the key quantitative metrics from those results so that the reproduction of GS fidelity is stated with explicit numerical evidence rather than qualitatively. revision: yes

Circularity Check

0 steps flagged

No significant circularity in methodological proposal

full rationale

The paper introduces GD-Phase as an explicit algorithmic construction: a phase-only gradient descent that hard-constrains measured spectral amplitudes while optimizing under real-space priors, using a differentiable forward model. This is benchmarked directly against GS and GD-Amp on synthetic multi-peaked and modulated CTR spectra. No load-bearing step equates a claimed first-principles result or prediction to its own fitted inputs or prior self-citations by construction; the framework is a self-contained optimization definition whose validity is tested via independent synthetic validation rather than tautological reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Abstract provides limited detail; main assumptions are differentiability of the forward model and sufficiency of real-space priors. No free parameters or invented entities are explicitly introduced.

axioms (2)

domain assumption The CTR forward model is differentiable and accurately represents experimental effects
Required for gradient-based optimization and seamless inclusion of arbitrary effects.
domain assumption Physical real-space priors are sufficient to constrain the phase retrieval solution
Invoked to resolve the ill-posed nature of the problem.

pith-pipeline@v0.9.0 · 5569 in / 1205 out tokens · 53920 ms · 2026-05-07T14:08:10.768007+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

35 extracted references · 33 canonical work pages · 1 internal anchor

[1]

Free-electron lasing at 27 nanometres based on a laser wakefield accelerator

Wentao Wang et al. “Free-electron lasing at 27 nanometres based on a laser wakefield accelerator”. en. In:Nature595.7868 (July 2021), pp. 516–520.issn: 0028-0836, 1476-4687. doi:10.1038/s41586-021-03678-x.url: https://www.nature.com/articles/s41586-021-03678-x

work page doi:10.1038/s41586-021-03678-x.url: 2021
[2]

Seeded free-electron laser driven by a compact laser plasma accelerator

Marie Labat et al. “Seeded free-electron laser driven by a compact laser plasma accelerator”. en. In:Nature Photonics17.2 (Feb. 2023), pp. 150–156.issn: 1749-4885, 1749-4893.doi: 10.1038/s41566-022-01104-w.url: https://www.nature.com/articles/s41566-022-01104-w

work page doi:10.1038/s41566-022-01104-w.url: 2023
[3]

Free-electron lasing with compact beam-driven plasma wakefield accelerator

R. Pompili et al. “Free-electron lasing with compact beam-driven plasma wakefield accelerator”. en. In:Nature605.7911 (May 2022), pp. 659–662.issn: 0028-0836, 1476-4687. doi:10.1038/s41586-022-04589-1.url: https://www.nature.com/articles/s41586-022-04589-1

work page doi:10.1038/s41586-022-04589-1.url: 2022
[4]

Greater than 1000-fold Gain in a Free-Electron Laser Driven by a Laser-Plasma Accelerator with High Reliability

S.K. Barber et al. “Greater than 1000-fold Gain in a Free-Electron Laser Driven by a Laser-Plasma Accelerator with High Reliability”. en. In:Physical Review Letters135.5 (July 2025), p. 055001.issn: 0031-9007, 1079-7114.doi:10.1103/vh62-gz1p.url: https://link.aps.org/doi/10.1103/vh62-gz1p

work page doi:10.1103/vh62-gz1p.url: 2025
[5]

Experimental Measurements of Electron-Bunch Trains in a Laser-Plasma Accelerator

O. Lundh et al. “Experimental Measurements of Electron-Bunch Trains in a Laser-Plasma Accelerator”. en. In:Physical Review Letters110.6 (Feb. 2013), p. 065005.issn: 0031-9007, 1079-7114.doi:10.1103/PhysRevLett.110.065005.url: https://link.aps.org/doi/10.1103/PhysRevLett.110.065005

work page doi:10.1103/physrevlett.110.065005.url: 2013
[6]

Temporal evolution of longitudinal bunch profile in a laser wakefield accelerator

M. Heigoldt et al. “Temporal evolution of longitudinal bunch profile in a laser wakefield accelerator”. en. In:Physical Review Special Topics - Accelerators and Beams18.12 (Dec. 2015), p. 121302.issn: 1098-4402.doi:10.1103/PhysRevSTAB.18.121302.url: https://link.aps.org/doi/10.1103/PhysRevSTAB.18.121302

work page doi:10.1103/physrevstab.18.121302.url: 2015
[7]

Revealing the three-dimensional structure of microbunched plasma-wakefield-accelerated electron beams

Maxwell LaBerge et al. “Revealing the three-dimensional structure of microbunched plasma-wakefield-accelerated electron beams”. en. In:Nature Photonics18.9 (Sept. 2024), pp. 952–959.issn: 1749-4885, 1749-4893.doi:10.1038/s41566-024-01475-2.url: https://www.nature.com/articles/s41566-024-01475-2

work page doi:10.1038/s41566-024-01475-2.url: 2024
[8]

Long-Range Persistence of Femtosecond Modulations on Laser-Plasma-Accelerated Electron Beams

C. Lin et al. “Long-Range Persistence of Femtosecond Modulations on Laser-Plasma-Accelerated Electron Beams”. en. In:Physical Review Letters108.9 (Mar. 2012), p. 094801.issn: 0031-9007, 1079-7114.doi:10.1103/PhysRevLett.108.094801.url: https://link.aps.org/doi/10.1103/PhysRevLett.108.094801

work page doi:10.1103/physrevlett.108.094801.url: 2012
[9]

Constraints on photon pulse duration from longitudinal electron beam diagnostics at a soft x-ray free-electron laser

C. Behrens et al. “Constraints on photon pulse duration from longitudinal electron beam diagnostics at a soft x-ray free-electron laser”. en. In:Physical Review Special Topics - Accelerators and Beams15.3 (Mar. 2012), p. 030707.issn: 1098-4402.doi: 10.1103/PhysRevSTAB.15.030707.url: https://link.aps.org/doi/10.1103/PhysRevSTAB.15.030707

work page doi:10.1103/physrevstab.15.030707.url: 2012
[10]

Coherent-Radiation Spectroscopy of Few-Femtosecond Electron Bunches Using a Middle-Infrared Prism Spectrometer

T. J. Maxwell et al. “Coherent-Radiation Spectroscopy of Few-Femtosecond Electron Bunches Using a Middle-Infrared Prism Spectrometer”. en. In:Physical Review Letters111.18 (Oct. 2013), p. 184801.issn: 0031-9007, 1079-7114.doi:10.1103/PhysRevLett.111.184801.url: https://link.aps.org/doi/10.1103/PhysRevLett.111.184801

work page doi:10.1103/physrevlett.111.184801.url: 2013
[11]

Diagnostics for plasma-based electron accelerators

M. C. Downer et al. “Diagnostics for plasma-based electron accelerators”. en. In:Reviews of Modern Physics90.3 (Aug. 2018), p. 035002.issn: 0034-6861, 1539-0756.doi: 10.1103/RevModPhys.90.035002.url: https://link.aps.org/doi/10.1103/RevModPhys.90.035002. 15 IOP PublishingJournal of Machine Learning: Science and Technologyvv(2026) aaaaaa Aguilaret al

work page doi:10.1103/revmodphys.90.035002.url: 2018
[12]

Benchmarking coherent radiation spectroscopy as a tool for high-resolution bunch shape reconstruction at free-electron lasers

Bernhard Schmidt et al. “Benchmarking coherent radiation spectroscopy as a tool for high-resolution bunch shape reconstruction at free-electron lasers”. en. In:Physical Review Accelerators and Beams23.6 (June 2020), p. 062801.issn: 2469-9888.doi: 10.1103/PhysRevAccelBeams.23.062801.url: https://link.aps.org/doi/10.1103/PhysRevAccelBeams.23.062801

work page doi:10.1103/physrevaccelbeams.23.062801.url: 2020
[13]

Multioctave high-dynamic range optical spectrometer for single-pulse, longitudinal characterization of ultrashort electron bunches

Omid Zarini et al. “Multioctave high-dynamic range optical spectrometer for single-pulse, longitudinal characterization of ultrashort electron bunches”. en. In:Physical Review Accelerators and Beams25.1 (Jan. 2022), p. 012801.issn: 2469-9888.doi: 10.1103/PhysRevAccelBeams.25.012801.url: https://link.aps.org/doi/10.1103/PhysRevAccelBeams.25.012801

work page doi:10.1103/physrevaccelbeams.25.012801.url: 2022
[14]

Temporal profile monitor based on electro-optic spatial decoding for low-energy bunches

Wei Wang et al. “Temporal profile monitor based on electro-optic spatial decoding for low-energy bunches”. en. In:Physical Review Accelerators and Beams20.11 (Nov. 2017), p. 112801.issn: 2469-9888.doi:10.1103/PhysRevAccelBeams.20.112801.url: https://link.aps.org/doi/10.1103/PhysRevAccelBeams.20.112801

work page doi:10.1103/physrevaccelbeams.20.112801.url: 2017
[15]

Few-femtosecond time-resolved measurements of X-ray free-electron lasers

C. Behrens et al. “Few-femtosecond time-resolved measurements of X-ray free-electron lasers”. en. In:Nature Communications5.1 (Apr. 2014), p. 3762.issn: 2041-1723.doi: 10.1038/ncomms4762.url:https://www.nature.com/articles/ncomms4762(visited on 01/28/2026)

work page doi:10.1038/ncomms4762.url:https://www.nature.com/articles/ncomms4762(visited 2014
[16]

Phase retrieval algorithms: a comparison

J. R. Fienup. “Phase retrieval algorithms: a comparison”. en. In:Applied Optics21.15 (Aug. 1982), p. 2758.issn: 0003-6935, 1539-4522.doi:10.1364/AO.21.002758.url: https://opg.optica.org/abstract.cfm?URI=ao-21-15-2758

work page doi:10.1364/ao.21.002758.url: 1982
[17]

Phase problem associated with the determination of the longitudinal shape of a charged particle bunch from its coherent far-ir spectrum

R. Lai and A. J. Sievers. “Phase problem associated with the determination of the longitudinal shape of a charged particle bunch from its coherent far-ir spectrum”. en. In: Physical Review E52.4 (Oct. 1995), pp. 4576–4579.issn: 1063-651X, 1095-3787.doi: 10.1103/PhysRevE.52.4576.url: https://link.aps.org/doi/10.1103/PhysRevE.52.4576

work page doi:10.1103/physreve.52.4576.url: 1995
[18]

Determination of bunch asymmetry from coherent radiation in the frequency domain

R. Lai and A. J. Sievers. “Determination of bunch asymmetry from coherent radiation in the frequency domain”. en. In:AIP Conference Proceedings. Vol. 367. Upton, New York (USA): AIP, 1996, pp. 312–326.doi:10.1063/1.50316.url: https://pubs.aip.org/aip/acp/article/367/1/312-326/753981

work page doi:10.1063/1.50316.url: 1996
[19]

Longitudinal electron bunch profile reconstruction by performing phase retrieval on coherent transition radiation spectra

S. I. Bajlekov et al. “Longitudinal electron bunch profile reconstruction by performing phase retrieval on coherent transition radiation spectra”. en. In:Physical Review Special Topics - Accelerators and Beams16.4 (Apr. 2013), p. 040701.issn: 1098-4402.doi: 10.1103/PhysRevSTAB.16.040701.url: https://link.aps.org/doi/10.1103/PhysRevSTAB.16.040701

work page doi:10.1103/physrevstab.16.040701.url: 2013
[20]

Reconstructing light field by two intersities in Fourier plane

RW Gerchberg and WO Saxton. “Reconstructing light field by two intersities in Fourier plane”. In:Optik35.1 (1972), pp. 23–30

1972
[21]

Advanced Methods for Temporal Reconstruction of Modulated Electron Bunches

Omid Zarini et al. “Advanced Methods for Temporal Reconstruction of Modulated Electron Bunches”. In:2018 IEEE Advanced Accelerator Concepts Workshop (AAC). Breckenridge, CO, USA: IEEE, Aug. 2018, pp. 1–5.isbn: 978-1-5386-7721-6.doi: 10.1109/AAC.2018.8659388.url:https://ieeexplore.ieee.org/document/8659388/

work page doi:10.1109/aac.2018.8659388.url:https://ieeexplore.ieee.org/document/8659388/ 2018
[22]

Adaptive Machine Learning for Robust Diagnostics and Control of Time-Varying Particle Accelerator Components and Beams

Alexander Scheinker. “Adaptive Machine Learning for Robust Diagnostics and Control of Time-Varying Particle Accelerator Components and Beams”. en. In:Information12.4 (Apr. 2021), p. 161.issn: 2078-2489.doi:10.3390/info12040161.url: https://www.mdpi.com/2078-2489/12/4/161

work page doi:10.3390/info12040161.url: 2021
[23]

Machine learning for beam dynamics studies at the CERN Large Hadron Collider

P. Arpaia et al. “Machine learning for beam dynamics studies at the CERN Large Hadron Collider”. en. In:Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment985 (Jan. 2021), p. 164652. issn: 01689002.doi:10.1016/j.nima.2020.164652.url: https://linkinghub.elsevier.com/retrieve/pii/S01...

work page doi:10.1016/j.nima.2020.164652.url: 2021
[24]

Phase Space Reconstruction from Accelerator Beam Measurements Using Neural Networks and Differentiable Simulations

R. Roussel et al. “Phase Space Reconstruction from Accelerator Beam Measurements Using Neural Networks and Differentiable Simulations”. en. In:Physical Review Letters130.14 (Apr. 2023), p. 145001.issn: 0031-9007, 1079-7114.doi:10.1103/PhysRevLett.130.145001. url:https://link.aps.org/doi/10.1103/PhysRevLett.130.145001. 16 IOP PublishingJournal of Machine L...

work page doi:10.1103/physrevlett.130.145001 2023
[25]

Machine learning for combined scalar and spectral longitudinal phase space reconstruction

Jan Kaiser et al. “Machine learning for combined scalar and spectral longitudinal phase space reconstruction”. en. In: (). Ed. by Assmann, Ralph et al. Artwork Size: 4464-4467 pages, 0.75 MB ISBN: 9783954502318 Medium: PDF, 4464–4467 pages, 0.75 MB.issn: 2673-5490.doi: 10.18429/JACOW-IPAC2023-THPL019.url: https://jacow.org/ipac2023/doi/jacow-ipac2023-thpl019

work page doi:10.18429/jacow-ipac2023-thpl019.url:
[26]

Differentiable simulations for particle tracking in accelerators: Analysis, benchmarking, and optimization

Francisco Huhn and Francesco M. Velotti. “Differentiable simulations for particle tracking in accelerators: Analysis, benchmarking, and optimization”. en. In:Physical Review Accelerators and Beams28.11 (Nov. 2025), p. 114603.issn: 2469-9888.doi:10.1103/cbds-lst1.url: https://link.aps.org/doi/10.1103/cbds-lst1

work page doi:10.1103/cbds-lst1.url: 2025
[27]

Efficient six-dimensional phase space reconstructions from experimental measurements using generative machine learning

Ryan Roussel et al. “Efficient six-dimensional phase space reconstructions from experimental measurements using generative machine learning”. en. In:Physical Review Accelerators and Beams27.9 (Sept. 2024), p. 094601.issn: 2469-9888.doi: 10.1103/PhysRevAccelBeams.27.094601.url: https://link.aps.org/doi/10.1103/PhysRevAccelBeams.27.094601

work page doi:10.1103/physrevaccelbeams.27.094601.url: 2024
[28]

Development and applications of differentiable coherent optical transition radiation simulations

Ryan Roussel et al. “Development and applications of differentiable coherent optical transition radiation simulations”. In:Proceedings of NAPAC2025. California, USA: JACoW Publishing, Aug. 2025, pp. 110–112.doi:doi:10.18429/JACoW-NAPAC2025-MOP029

work page doi:10.18429/jacow-napac2025-mop029 2025
[29]

Phase imaging with an untrained neural network

Fei Wang et al. “Phase imaging with an untrained neural network”. en. In:Light: Science & Applications9.1 (May 2020), p. 77.issn: 2047-7538.doi:10.1038/s41377-020-0302-3. url:https://www.nature.com/articles/s41377-020-0302-3

work page doi:10.1038/s41377-020-0302-3 2020
[30]

PhaseGAN: a deep-learning phase-retrieval approach for unpaired datasets

Yuhe Zhang et al. “PhaseGAN: a deep-learning phase-retrieval approach for unpaired datasets”. en. In:Optics Express29.13 (June 2021), p. 19593.issn: 1094-4087.doi: 10.1364/OE.423222.url:https://opg.optica.org/abstract.cfm?URI=oe-29-13-19593

work page doi:10.1364/oe.423222.url:https://opg.optica.org/abstract.cfm 2021
[31]

Deep learning based on parameterized physical forward model for adaptive holographic imaging with unpaired data

Chanseok Lee et al. “Deep learning based on parameterized physical forward model for adaptive holographic imaging with unpaired data”. en. In:Nature Machine Intelligence5.1 (Jan. 2023), pp. 35–45.issn: 2522-5839.doi:10.1038/s42256-022-00584-3.url: https://www.nature.com/articles/s42256-022-00584-3

work page doi:10.1038/s42256-022-00584-3.url: 2023
[32]

Transition radiation in the pre-wave zone

VL Ginzburg and IM Frank. “Transition radiation in the pre-wave zone”. In:Sov. Phys. JETP16.15 (1946), p. 222

1946
[33]

Theory of coherent transition radiation generated at a plasma-vacuum interface

C. B. Schroeder et al. “Theory of coherent transition radiation generated at a plasma-vacuum interface”. en. In:Physical Review E69.1 (Jan. 2004), p. 016501.issn: 1539-3755, 1550-2376. doi:10.1103/PhysRevE.69.016501.url: https://link.aps.org/doi/10.1103/PhysRevE.69.016501

work page doi:10.1103/physreve.69.016501.url: 2004
[34]

Fast Change of Windows Properties

M. Jaskula. “Fast Change of Windows Properties”. In:2006 IEEE Instrumentation and Measurement Technology Conference Proceedings. 2006, pp. 1043–1046.doi: 10.1109/IMTC.2006.328341

work page doi:10.1109/imtc.2006.328341 2006
[35]

Adam: A Method for Stochastic Optimization

Diederik P Kingma. “Adam: A method for stochastic optimization”. In:arXiv preprint arXiv:1412.6980(2014). 17

work page internal anchor Pith review arXiv 2014