Pith. sign in

REVIEW 4 major objections 3 minor 4 cited by

A faithful GEANT4UCN forward simulation of ultracold neutron storage, extraction, and vertical time-of-flight detection makes possible the first simulation-based inference analysis in this field, recovering the loss parameters f and Γ′ from

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

GEANT4 simulations of the SuperSUN ultracold neutron source are paired with neural simulation-based inference to recover UCN loss parameters from time-of-flight spectra.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection Honest, first-of-kind SBI methods paper for UCN; forward simulation is credible, but the SBI validation is closed-loop and simulation fidelity is asserted rather than quantified. the 4 major comments →

arxiv 2509.02791 v1 pith:CTSSOH5O submitted 2025-09-02 nucl-th hep-phnucl-ex

Towards Precise Simulations and Inference for the Neutron EDM

Skyler Degenkolb , Luigi Favaro , Peter Fierlinger , Jennifer Franz , Husain Manasawala , Tilman Plehn This is my paper
classification nucl-th hep-phnucl-ex
keywords ultracold neutronsneutron electric dipole momentsimulation-based inferencevertical time-of-flightGEANT4UCNneutron loss parameterssuperfluid helium sourceneural posterior estimation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

Ultracold neutrons are the workhorse for precision neutron-property experiments such as the electric dipole moment search, but their low statistics and energy-dependent losses make measured data hard to interpret. This paper shows that the full chain—production and storage in a superfluid-4He converter, extraction through guides, and vertical time-of-flight detection—can be simulated faithfully with GEANT4UCN ray tracing. It then uses those simulations as the forward model for neural simulation-based inference (SBI), recovering joint posteriors for the wall-loss parameter f and the energy-independent loss rate Γ′ from vTOF counting matrices. If the simulation is faithful, this is the first demonstration that SBI can replace explicit likelihoods for ultracold neutron data, turning low statistics and correlated parameters into a tractable inference problem.

Core claim

The central claim is that a precise forward simulation of UCN storage, extraction, and vTOF detection is sufficient to perform simulation-based inference of physical loss parameters. The proof of principle uses a GEANT4UCN model upgraded to GEANT4-11-03.2, seeded with the analytic total-energy storage spectrum of Eq. (16), including gravity and full angle-dependent wall-reflection losses. Trained on roughly 9,000 simulated datasets, a conditional neural posterior estimator produces 1/2/3σ contours for (log f, log Γ′) that contain the true parameter values, with the expected f–Γ′ anti-correlation. This constitutes the first SBI application in UCN physics and directly addresses the longstandin

What carries the argument

The argument rests on two coupled objects. First, the vertical time-of-flight pseudo-2D count matrix C_ij—chopper frame i by TOF bin j—compresses the stored UCN spectrum and its time evolution into a data representation that preserves the long-time drain information lost by conventional frame-aggregated TOF spectra. Second, the forward simulator combines an analytically seeded initial spectrum (total-energy spectrum from Eq. (16), including the hypergeometric correction for gravity in the cylindrical converter) with GEANT4UCN ray tracing that applies the full quantum-mechanical reflection-loss amplitude |R(θ)|² at every wall interaction and simulates foil transmissions analytically. Inferenc

Load-bearing premise

The inference treats accumulation and holding as governed by the same loss parameters (f_acc = f_hold and Γ′_acc = Γ′_hold); if beam-induced heating or any other phase-dependent effect makes the losses differ, the recovered (f, Γ′) would be biased effective values rather than the clean physical parameters the analysis aims for.

What would settle it

Perform a coverage test on the trained neural posterior with held-out simulations: if the true (f, Γ′) values fall inside the nominal 68% and 95% contours substantially less often than 68% and 95%, the posterior is miscalibrated and the simulation-based inference claim fails.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Experimental vTOF datasets from SuperSUN can be analyzed with neural posterior estimation, replacing ill-posed chopper deconvolution with a forward convolution inside the simulation.
  • Inferred Γ′ values translate directly into limits on 3He contamination in the converter via Γ3He = |Γ′ − 1/878 s|, providing a diagnostic for source performance.
  • The same pipeline transfers to external UCN experiments because wall loss and guide transport are modeled with shared physical parameters rather than ad hoc efficiencies.
  • Moderate accumulation and holding times (e.g., 500 s and 100 s) preserve both energy-dependent and energy-independent loss information, whereas very long times wash it out—this informs how to schedule real measurements.
  • Simulation-based inference turns the binned vTOF dataset itself, rather than a deconvolved TOF spectrum, into the observable, avoiding the noise amplification of iterative deconvolution methods.

Where Pith is reading between the lines

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

  • We would test the accumulation-vs-holding assumption directly: infer (f, Γ′) for datasets with the same holding time but strongly different accumulation times; if the posterior shifts beyond statistical width, the single-loss-parameter reduction is invalid and the recovered values are effective parameters.
  • A natural extension is to include guide-system nuisance parameters—chopper offset, guide optical potential, gap sizes—as additional inference dimensions, since the simulation already contains them as fixed nominal values.
  • Future work could combine integral counting data with vTOF matrices in one posterior; the complementary information may break part of the f–Γ′ degeneracy and tighten the 3He limit.
  • A quantitative coverage test of the trained estimator—checking that the nominal 68% and 95% contours contain true parameters at the claimed frequencies—would establish whether the posterior widths can be read as calibrated uncertainties.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 3 minor

Summary. The paper presents GEANT4UCN-based simulations of ultracold neutron (UCN) production, storage, and extraction for the SuperSUN source, with a focus on vertical time-of-flight (vTOF) spectroscopy. It compares simulated vTOF spectra to measured SuperSUN data in a limited, qualitative way, then uses the simulator as a forward model for simulation-based inference (SBI) with conditional normalizing flows. A toy benchmark infers the accumulation and holding times (ta, th), and the main inference demonstration targets the wall-loss parameter f and the energy-independent loss rate Γ′. The paper claims this constitutes the first SBI application in UCN physics and argues that vTOF plus SBI can extract physically meaningful loss parameters from real SuperSUN data.

Significance. If the forward simulation is faithful and the SBI posteriors are calibrated, this would be a genuinely useful methodological advance: UCN experiments are statistics-limited and suffer from complex, energy-dependent transport, and a validated simulation-to-inference pipeline could help diagnose loss mechanisms in SuperSUN and similar sources. The paper has clear strengths: the GEANT4UCN implementation is detailed (gravity, non-specular reflections, wall-loss via the full reflection formula, chopper transmission forward convolution in Fig. 7), the analytic spectra of Sec. 2 are useful, and the SBI setup is technically sound as a proof of principle. The toy benchmark showing that a global time reference sharply improves the posterior for (ta, th) is informative. However, the evidence that the simulator is reliable enough for real data is currently qualitative, and the SBI demonstration is a closed-loop self-consistency test on one simulated test point. The paper itself acknowledges that a quantitative coverage study is missing, which tempers the strength of the central claim.

major comments (4)
  1. [Sec. 3, Figs. 5 and 7] The load-bearing claim that the GEANT4UCN forward simulation is 'precise and reliable' for vTOF is supported only by visual comparison with one experimental configuration. The text admits 'small discrepancies' (Fig. 5 caption) and attributes them to the static chopper, but no chi-square, residual, or any quantitative goodness-of-fit statistic is reported. Since the SBI posterior is only as trustworthy as the simulator, this qualitative validation is insufficient. I request a quantitative agreement metric (e.g., bin-wise residuals, chi-square/ndf, or a KS-type test) for the comparison in Figs. 5 and 7, and a discussion of which features of the data are and are not captured by the simulation.
  2. [Sec. 4, Inference and Fig. 10] The SBI demonstration is a closed loop: the test spectrum is generated from the same analytic Eq. (16) and the same GEANT4 code that define the training data and the parameters. Recovering the true value from such a test confirms self-consistency but does not establish that posterior is unbiased or well-calibrated for real SuperSUN data. The paper explicitly states 'a quantitative coverage study requires further investigation' (Sec. 4, Inference). In addition, only one test point is shown. I ask for a coverage or expected-calibration-error study over multiple test points, and, if possible, an application of the trained posterior to the measured vTOF data of Sec. 3, at least as a diagnostic of the simulator fidelity.
  3. [Sec. 2, after Eq. (17)] The central inference relies on the assumption that loss mechanisms during accumulation and holding are identical (facc = fhold, Γ′acc = Γ′hold). The authors note that beam-induced heating could make these differ. If this assumption fails, the recovered (f, Γ′) are biased effective values rather than the physically meaningful parameters promised in the abstract and outlook. This is a load-bearing assumption for the physics interpretation. I recommend adding a sensitivity study (e.g., letting facc differ from fhold in the simulator and testing the resulting posterior bias) or at least a quantitative discussion of the expected bias under plausible differences.
  4. [Eq. (16)] The stored-spectrum formulas in Eq. (16) are labeled dN(ta, ta)/du while the text defines a storage phase th; presumably this is a typo for dN(ta, th)/du. This matters because Eq. (16) is used to generate the initial spectra in the simulations and to fix the relative normalization in the SBI training data. Please correct the notation and verify that the implemented formula uses the same variables as written.
minor comments (3)
  1. [Introduction] Typo: 'an representation' should be 'a representation'. Also the phrase 'first precision simulations' (Sec. 2) is stronger than what is demonstrated; consider 'first detailed simulations' or similar.
  2. [Sec. 2, UCN Losses] Eq. (31): the absolute value |Γ′ − (878 s)−1| is introduced to guard against floating-point errors. This is unusual for a physical relation; a brief explanation of why negative differences can occur in the simulation would help the reader assess the robustness of the 3He interpretation.
  3. [Fig. 10] The text states the true value 'lies in the high-density region' but does not give the numerical credible interval. Reporting e.g. the 68% and 95% marginal intervals for the test point would make the posterior content quantitative and easier to interpret.

Circularity Check

0 steps flagged

No significant circularity: the SBI demonstration is an explicitly labeled benchmark on simulated data, and the forward model's external grounding is the direct comparison to measured vTOF spectra.

full rationale

The paper's derivation chain is not circular by the quoted-reduction standard. The forward simulation (GEANT4UCN) is seeded with analytic spectra from Eq. (16), and the SBI benchmark in Sec. 4 tests whether the neural posterior can recover parameters from a held-out simulation generated by the same simulator. This is a self-consistency check, but the paper presents it as a 'proof of principle' and 'benchmark' (Sec. 4), not as a measurement of f and Gamma' from real data. The claim that the simulator is reliable is grounded in the direct comparison to measured SuperSUN vTOF data in Sec. 3 (Figs. 5 and 7), which is an independent external benchmark. The paper explicitly acknowledges limitations: the static-chopper approximation leaves 'small discrepancies' (Fig. 5 caption), deconvolution is described as ill-posed, and 'a quantitative coverage study requires further investigation' (Sec. 4). These weaken the strength of the fidelity claim but are not circular reasoning. Citations to [25], [30], [33], and [34] are prior experimental, material-property, and software work by overlapping authors, but they are not invoked as uniqueness theorems or as substitutes for derivation; they provide independent measured inputs (e.g., CYTOP loss factor, GEANT4UCN code, experimental data). No equation is defined in terms of the quantity it purports to derive, and no fitted parameter is renamed as a prediction. The closed-loop SBI test is a standard calibration exercise, not a circular derivation of physical results.

Axiom & Free-Parameter Ledger

4 free parameters · 6 axioms · 0 invented entities

The paper introduces no new entities. Its load-bearing inputs are the physics model of UCN production and loss from prior literature, a set of hand-chosen material and guide parameters, and the explicit simplification that accumulation and holding losses are equal. The parameters f and Gamma' are the intended outputs of the SBI, not ad hoc fits.

free parameters (4)
  • f (converter wall loss parameter) = true test value log f = -7.57; training prior U(10^-5/pi, 2*pi*10^-4)
    Target parameter of the SBI; physical wall loss factor for the CYTOP-coated converter including gaps and absorbing areas. Not ad hoc, but a fitted output.
  • Gamma' (energy-independent loss rate) = true test value log Gamma' = -5.48 s^-1; training prior U(1.12e-3, 9.8e-3) s^-1
    Target parameter of the SBI; combines beta decay and helium-3 capture losses. Not ad hoc, but a fitted output.
  • f_guide = 0.0005
    Fixed by hand for the extraction guide system; affects transport efficiency and TOF spectra.
  • P_diffuse = 0.04
    Fixed diffuse reflection probability for guide walls; chosen as nominal value, affects the velocity remixing in transport.
axioms (6)
  • domain assumption UCN velocities are rapidly randomized by nonspecular reflections, so wall loss can be treated with angle-averaged kinetic theory
    Invoked in Sec. 2, 'UCN Losses', to justify Eq. (12) and mechanical equilibrium. If false, the loss rates are angle-dependent and the storage model changes.
  • domain assumption Loss mechanisms during accumulation and holding are identical (f_acc equals f_hold and Gamma'_acc equals Gamma'_hold)
    Stated explicitly after Eq. (17) as a proof-of-principle simplification. Directly couples the inferred (f, Gamma') to both phases.
  • domain assumption The analytic stored spectrum Eq. (16) accurately describes the UCN ensemble at the start of extraction
    Used in Sec. 3 to initialize GEANT4 without simulating the full production and storage sequence. Errors here propagate into all simulated TOF spectra.
  • domain assumption The static chopper approximation, with forward convolution of the measured chopper transmission function, adequately describes the TOF measurement
    Sec. 3 and Fig. 7; the static chopper neglects deconvolution and dynamical effects, acknowledged as the main source of simulation-data discrepancy.
  • domain assumption Neutron beta-decay lifetime is fixed at 878 s
    Sec. 3; used as a constant energy-independent loss. Slight lifetime variations are absorbed into Gamma'.
  • domain assumption Production spectrum scales as E^(3/2) via the single-phonon channel, with constant C from measured cold-neutron flux
    Sec. 2, Eq. (2); taken from prior literature (Golub-Pendlebury, Schmidt-Wellenburg et al.). Subleading channels add about 10% and are absorbed in normalization.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Precise Simulations and Inference for the Neutron EDM." pith.science (2026). https://pith.science/paper/CTSSOH5O

@misc{pith2026250902791,
  author       = {Pith},
  title        = {Pith review of: Towards Precise Simulations and Inference for the Neutron EDM},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CTSSOH5O}},
  note         = {Machine review of arXiv:2509.02791}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Precision measurements of neutron properties, like its permanent electric dipole moment, rely on understanding complex experimental setups in detail. We show how the properties of stored and transported ultracold neutron ensembles can be simulated reliably. In a second step, we illustrate how they can be used for simulation-based inference of the parameters associated with underlying physics processes such as neutron capture or beta decay. Our proof of principle for simulation-based inference confronts a longstanding challenge with ultracold neutrons: low measurement statistics coupled with a complex apparatus.

Figures

Figures reproduced from arXiv: 2509.02791 by Husain Manasawala, Jennifer Franz, Luigi Favaro, Peter Fierlinger, Skyler Degenkolb, Tilman Plehn.

Figure 1
Figure 1. Figure 1: Time [s] 100 101 102 103 104 UCN counted per second ta=1500 s accumulate th=1600 s hold tc=2000 s count 0 1000 2000 3000 4000 5000 Time [s] −5 0 5 0.0 0.1 0.2 0.3 0.4 0.5 0.6 TOF [s] 1100 1200 1300 1400 1500 1600 TTL Start Time [s] 0 10 20 30 40 Counts [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Total-energy UCN spectra in-situ, calculated for SuperSUN’s converter properties. The trappable and extractable energy range lies between the gray grid-lines. Left: energy￾dependent UCN production rate (Full: closed-form integral Eq.(15) including all geometrical and gravitational effects / Simplified: kinetic energy spectrum shifted by the mean gravita￾tional potential in the converter, Eq.(14)). Right: S… view at source ↗
Figure 3
Figure 3. Figure 3: Diagram of SuperSUN [25] for the vertical time-of-flight configuration, with gravity vertical in the plane of the page. converter vessel vacuum housing polypropylene foil UCN extraction point chopper gratings detector entrance 636 mm motor beam guide aperture motor [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Simulation geometry and representative particle trajectories for the vertical time-of [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of measured data (left) and simulation (right) for [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Left: chopper transmission function used for deconvolution, provided by the ILL and based on optical measurements. The grey dashed box defines ∆T. A green dashed box defines the effective opening time, by matching a rectangular chopper function to the integral of the true one. Right: Deconvolution using Richardson-Lucy reconstruction, with a cubic spline based seed spectrum, for an accumulation mode TOF sl… view at source ↗
Figure 7
Figure 7. Figure 7: Left: Initial spectrum for (ta , th ) = (1000 s, 100 s). Right: Projected TOF spectrum showing a comparison of experimental data to simulation with and without forward convo￾lution of the chopper transmission function. The pale bands represent 1σ counting statistics errors. While the noise parameter fundamentally limits deconvolution by fast Fourier transform, smooth functional models provide a more stable… view at source ↗
Figure 8
Figure 8. Figure 8: Neural posteriors for (ta , th ) = (1010 s, 989 s). Left: using a different reference time for each simulation. Right: using a global reference time. Benchmark To test our method and gain control over our parametrization of physics effects, we consider the toy problem of inferring experimentally defined parameters which have relatively simple implications at the detector level. We select the accumulation a… view at source ↗
Figure 9
Figure 9. Figure 9: Left: spectra at beginning of UCN extraction. Right: spectra with much shorter (ta , th ). energy for extraction, and are mainly distinguished by their normalization. Those in the right panel with much shorter (ta , th ) exhibit stronger variation, with respect to each other and to the test point. This is what the inference picks up correctly. For the right panel of [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Neural posteriors for the inference of (log f , log Γ ′ ). The posterior is extracted from a test simulation with parameters (−7.57,−5.48). Contours show 1/2/3σ surfaces. relation is employed to determine energy-independent losses from storage in a bulk medium contaminated with 3He, Γ3He = [PITH_FULL_IMAGE:figures/full_fig_p020_10.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Neural Control Variates at LO and NLO

    hep-ph 2026-07 accept novelty 7.0

    Signed neural control variates from normalizing flows, combined with neural importance sampling, reduce weight ranges and negative weights for LO and NLO phase-space integration and event generation.

  2. One Generator, Any Process: LLM-Conditioning for the LHC

    hep-ph 2026-06 unverdicted novelty 7.0

    LLM embeddings condition generative networks for LHC events, yielding faster convergence, higher quality, and generalization to unseen processes.

  3. Agentic Re-Casting using Agentic Re-Simulations

    hep-ph 2026-07 conditional novelty 6.0

    An agentic AI system with a physicist in the loop re-casts an ATLAS ttZ measurement into a global top-quark SMEFT fit and recovers injected coloron Wilson coefficients in a repeatable benchmark.

  4. One Generator, Any Process: LLM-Conditioning for the LHC

    hep-ph 2026-06 unverdicted novelty 6.0

    LLM embeddings condition a generative transformer to enable faster convergence, better performance, and generalization to unseen LHC processes using a single model.

Reference graph

Works this paper leans on

41 extracted references · 25 canonical work pages · cited by 3 Pith papers · 6 internal anchors

  1. [1]

    Dubbers and M

    D. Dubbers and M. G. Schmidt, The Neutron and Its Role in Cosmology and Particle Physics, Rev. Mod. Phys.83 (2011) 1111, arXiv:1105.3694 [hep-ph]

    Pith/arXiv arXiv 2011

  2. [2]

    Navas et al., Review of particle physics, Phys

    Particle Data Group, S. Navas et al., Review of particle physics, Phys. Rev. D110 (2024) 3, 030001

    work page 2024

  3. [3]

    Chupp, P

    T . Chupp, P . Fierlinger, M. Ramsey-Musolf, and J. Singh,Electric dipole moments of atoms, molecules, nuclei, and particles, Rev. Mod. Phys.91 (2019) 1, 015001, arXiv:1710.02504 [physics.atom-ph]

    Pith/arXiv arXiv 2019

  4. [4]

    Abel et al., Measurement of the Permanent Electric Dipole Moment of the Neutron, Phys

    C. Abel et al., Measurement of the Permanent Electric Dipole Moment of the Neutron, Phys. Rev. Lett.124 (2020) 8, 081803, arXiv:2001.11966 [hep-ex]

    Pith/arXiv arXiv 2020

  5. [5]

    Degenkolb, N

    S. Degenkolb, N. Elmer, T . Modak, M. Mühlleitner, and T . Plehn,A Global View of the EDM Landscape, arXiv:2403.02052 [hep-ph]

    Pith/arXiv arXiv

  6. [6]

    Musedinovic et al., Measurement of the free neutron lifetime in a magneto-gravitational trap with in situ detection, Phys

    R. Musedinovic et al., Measurement of the free neutron lifetime in a magneto-gravitational trap with in situ detection, Phys. Rev. C111 (2025) 4, 045501, arXiv:2409.05560 [nucl-ex]

    Pith/arXiv arXiv 2025

  7. [7]

    A. P . Serebrovet al., Neutron lifetime measurements with a large gravitational trap for ultracold neutrons, Phys. Rev. C97 (2018) 5, 055503, arXiv:1712.05663 [nucl-ex]. 22 SciPost Physics Submission

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  8. [8]

    A. T . Yue, M. S. Dewey , D. M. Gilliam, G. L. Greene, A. B. Laptev, J. S. Nico, W . M. Snow, and F . E. Wietfeldt,Improved Determination of the Neutron Lifetime, Phys. Rev. Lett.111 (2013) 22, 222501, arXiv:1309.2623 [nucl-ex]

    Pith/arXiv arXiv 2013

  9. [9]

    Byrne and P

    J. Byrne and P . G. Dawber,A Revised Value for the Neutron Lifetime Measured Using a Penning Trap, EPL 33 (1996) 187

    work page 1996

  10. [10]

    UCNA, M. P . Mendenhallet al., Precision measurement of the neutronβ-decay asymmetry, Phys. Rev. C87 (2013) 3, 032501, arXiv:1210.7048 [nucl-ex]

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  11. [11]

    Cronenberg, P

    G. Cronenberg, P . Brax, H. Filter, P . Geltenbort, T . Jenke, G. Pignol, M. Pitschmann, M. Thalhammer, and H. Abele, Acoustic Rabi oscillations between gravitational quantum states and impact on symmetron dark energy, Nature Phys. 14 (2018) 10, 1022

    work page 2018

  12. [12]

    Jenke, J

    T . Jenke, J. Bosina, J. Micko, M. Pitschmann, R. Sedmik, and H. Abele,Gravity resonance spectroscopy and dark energy symmetron fields: qBOUNCE experiments performed with Rabi and Ramsey spectroscopy, Eur. Phys. J. ST 230 (2021) 4, 1131, arXiv:2012.07472 [hep-ph]

    Pith/arXiv arXiv 2021

  13. [13]

    Test of Lorentz invariance with spin precession of ultracold neutrons

    I. Altarev et al., Test of Lorentz invariance with spin precession of ultracold neutrons, Phys. Rev. Lett.103 (2009) 081602, arXiv:0905.3221 [nucl-ex]

    work page internal anchor Pith review Pith/arXiv arXiv 2009

  14. [14]

    qBounce, A. N. Ivanov, M. Wellenzohn, and H. Abele,Tests of Lorentz-Invariance Violation in the Standard-Model Extension with Ultracold Neutrons in qBounce Experiments, in 8th Meeting on CPT and Lorentz Symmetry. 2020

    work page 2020

  15. [15]

    N. J. Ayres, G. Bison, K. Bodek, V . Bondar, T . Bouillaud, E. Chanel, P .-J. Chiu, B. Clement, C. B. Crawford, M. Daum, et al., Search for an interaction mediated by axion-like particles with ultracold neutrons at the PSI, New J. Phys. 25 (2023) 10, 103012

    work page 2023

  16. [16]

    A search for neutron-to-hidden-neutron oscillations in a ultra-cold neutron beam

    G. Ban et al., Search for Neutron-to-Hidden-Neutron Oscillations in an Ultracold Neutron Beam, Phys. Rev. Lett.131 (2023) 19, 191801, arXiv:2303.10507 [hep-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  17. [17]

    Degenkolb, P

    S. Degenkolb, P . Fierlinger, and O. Zimmer,Approaches to high-density storage experiments with in-situ production and detection of ultracold neutrons, Journal of Neutron Research 24 (2022) 2, 123

    work page 2022

  18. [18]

    Plehn, A

    T . Plehn, A. Butter, B. Dillon, T . Heimel, C. Krause, and R. Winterhalder,Modern Machine Learning for LHC Physicists, arXiv:2211.01421 [hep-ph]

    Pith/arXiv arXiv

  19. [19]

    Cranmer, J

    K. Cranmer, J. Brehmer, and G. Louppe,The frontier of simulation-based inference, Proc. Nat. Acad. Sci. 117 (2020) 48, 30055, arXiv:1911.01429 [stat.ML]

    Pith/arXiv arXiv 2020

  20. [20]

    M. Dax, S. R. Green, J. Gair, J. H. Macke, A. Buonanno, and B. Schölkopf, Real-Time Gravitational Wave Science with Neural Posterior Estimation, Phys. Rev. Lett.127 (2021) 24, 241103, arXiv:2106.12594 [gr-qc]

    Pith/arXiv arXiv 2021

  21. [21]

    M. Dax, J. Wildberger, S. Buchholz, S. R. Green, J. H. Macke, and B. Schölkopf,Flow Matching for Scalable Simulation-Based Inference, arXiv:2305.17161 [cs.LG]

    Pith/arXiv arXiv

  22. [22]

    Bellagente, A

    M. Bellagente, A. Butter, G. Kasieczka, T . Plehn, and R. Winterhalder,How to GAN away Detector Effects, SciPost Phys. 8 (2020) 4, 070, arXiv:1912.00477 [hep-ph]

    Pith/arXiv arXiv 2020

  23. [23]

    Bellagente, A

    M. Bellagente, A. Butter, G. Kasieczka, T . Plehn, A. Rousselot, R. Winterhalder, L. Ardizzone, and U. Köthe, Invertible Networks or Partons to Detector and Back Again, SciPost Phys. 9 (2020) 074, arXiv:2006.06685 [hep-ph]. 23 SciPost Physics Submission

    Pith/arXiv arXiv 2020

  24. [24]

    Bieringer, A

    S. Bieringer, A. Butter, T . Heimel, S. Höche, U. Köthe, T . Plehn, and S. T . Radev, Measuring QCD Splittings with Invertible Networks, SciPost Phys. 10 (2021) 6, 126, arXiv:2012.09873 [hep-ph]

    Pith/arXiv arXiv 2021

  25. [25]

    Degenkolb et al., High-Density Ultracold Neutron Source for Low-Energy Particle Physics Experiments, arXiv:2504.13030 [physics.ins-det]

    S. Degenkolb et al., High-Density Ultracold Neutron Source for Low-Energy Particle Physics Experiments, arXiv:2504.13030 [physics.ins-det]

    Pith/arXiv arXiv

  26. [26]

    Golub and J

    R. Golub and J. Pendlebury ,The interaction of ultra-cold neutrons (UCN) with liquid helium and a superthermal UCN source, Physics Letters A 62 (1977) 5, 337

    work page 1977

  27. [27]

    Schmidt-Wellenburg, K

    P . Schmidt-Wellenburg, K. Andersen, and O. Zimmer,Ultra cold neutron production by multiphonon processes in superfluid helium under pressure, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 611 (2009) 2, 259. Particle Physics with Slow Neutrons

    work page 2009

  28. [28]

    J. M. Pendlebury and D. J. Richardson, Effects of gravity on the storage of ultracold neutrons, Nucl. Instrum. Meth. A 337 (1994) 504

    work page 1994

  29. [29]

    nEDM, M. W . Ahmedet al., A New Cryogenic Apparatus to Search for the Neutron Electric Dipole Moment, JINST 14 (2019) 11, P11017, arXiv:1908.09937 [physics.ins-det]

    Pith/arXiv arXiv 2019

  30. [30]

    Neulinger, D

    T . Neulinger, D. Beck, E. Connolly , S. Degenkolb, P . Fierlinger, H. Filter, J. Hingerl, P . Nordin, T . Saerbeck, and O. Zimmer,Ultracold neutron storage in a bottle coated with the fluoropolymer CYTOP, Eur. Phys. J. A 58 (2022) 7, 141

    work page 2022

  31. [31]

    Steyerl, Ultracold Neutrons

    A. Steyerl, Ultracold Neutrons. WSP , 2020

    work page 2020

  32. [32]

    Agostinelli et al., GEANT4 - A Simulation Toolkit, Nucl

    GEANT4, S. Agostinelli et al., GEANT4 - A Simulation Toolkit, Nucl. Instrum. Meth. A 506 (2003) 250

    work page 2003

  33. [33]

    Atchison, T

    F . Atchison, T . Bry´s, M. Daum, P . Fierlinger, A. Fomin, R. Henneck, K. Kirch, M. Ku´ zniak, and A. Pichlmaier, The simulation of ultracold neutron experiments using GEANT4, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 552 (2005) 3, 513

    work page 2005

  34. [34]

    Neulinger, H

    T . Neulinger, H. Filter, and O. Zimmer,Vertical time-of-flight spectroscopy of ultracold neutrons, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 1059 (2024)

    work page 2024

  35. [35]

    Time-of-flight spectroscopy of ultracold neutrons at the PSI UCN source

    G. Bison et al., Time-of-flight spectroscopy of ultracold neutrons at the PSI UCN source, Eur. Phys. J. A 59 (2023) 9, 215, arXiv:2301.11668 [physics.ins-det]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  36. [36]

    Vertical time of flight measurement

    M. Jentschel, D. Beck, E. Chanel, S. Degenkolb, H. Desalme, C. Filter, T . Jenke, H. Manasawala, T . Neulinger, and O. Zimmer, “Vertical time of flight measurement.” Institut laue-langevin (ILL) doi:10.5291/ill-data.test-3368, 2024

    work page doi:10.5291/ill-data.test-3368 2024

  37. [37]

    Atchison, B

    F . Atchison, B. Blau, A. Bollhalder, M. Daum, P . Fierlinger, P . Geltenbort, G. Hampel, M. Kasprzak, K. Kirch, S. Köchli, B. Kuczewski, H. Leber, M. Locher, M. Meier, S. Ochse, A. Pichlmaier, C. Plonka, R. Reiser, J. Ulrich, X. Wang, N. Wiehl, O. Zimmer, and G. Zsigmond, Transmission of very slow neutrons through material foils and its influence on the ...

    work page 2009

  38. [38]

    L. Babin, Caractérisation de HOPE un piège magnéto-gravitationnel à neutrons ultra-froids destiné à la mesure du temps de vie du neutron sur SUN2 une source superthermale de neutrons ultra-froids. PhD thesis, Université Grenoble Alpes, 2019

    work page 2019

  39. [39]

    W . H. Richardson,Bayesian-based iterative method of image restoration∗, J. Opt. Soc. Am. 62 (Jan, 1972) 55

    work page 1972

  40. [40]

    L. B. Lucy, An iterative technique for the rectification of observed distributions, Astronomical Journal 79 (June, 1974) 745

    work page 1974

  41. [41]

    How to Unfold Top Decays

    L. Favaro, R. Kogler, A. Paasch, S. Palacios Schweitzer, T . Plehn, and D. Schwarz,How to Unfold Top Decays, SciPost Phys. Core 8 (2025) 053, arXiv:2501.12363 [hep-ph]. 25

    work page internal anchor Pith review Pith/arXiv arXiv 2025

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.