Performance of the FARICH-based particle identification at charm superfactories using machine learning

M. Chadeeva; P. Rogozhin; T. Uglov

arxiv: 2506.14247 · v1 · submitted 2025-06-17 · ✦ hep-ex · physics.ins-det

Performance of the FARICH-based particle identification at charm superfactories using machine learning

M. Chadeeva , P. Rogozhin , T. Uglov This is my paper

Pith reviewed 2026-05-19 09:53 UTC · model grok-4.3

classification ✦ hep-ex physics.ins-det

keywords particle identificationFARICHBoosted Decision Treesmachine learningcharm physicsD0 decayspion muon separationaerogel Cherenkov

0 comments

The pith

Machine learning classifiers achieve high-efficiency particle identification with the FARICH detector and reduce uncertainties in D0 decay analyses.

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

The paper develops and tests Boosted Decision Tree classifiers for identifying particles detected by the Focusing Aerogel Ring Imaging Cherenkov subsystem in a simulated charm superfactory environment. It demonstrates that these classifiers maintain high efficiency for pion-muon separation across different assumptions about photosensor noise. The performance is validated by applying the classifiers to the analysis of D0 decays into a kaon, muon and neutrino, where the improved separation minimizes both systematic uncertainties and background contributions. A sympathetic reader would care because accurate particle identification directly improves the precision of measurements in charm physics and similar high-energy decay studies.

Core claim

A dedicated signal ring reconstruction algorithm implemented in the detector simulation, combined with two Boosted Decision Tree classifiers developed for varying photosensor noise levels, delivers high particle identification efficiency. This efficiency allows the systematic uncertainty and background contribution related to pion-muon separation to be minimised in the analysis of D0 to K mu nu decays.

What carries the argument

Two Boosted Decision Trees-based classifiers that use reconstructed Cherenkov ring information from the FARICH detector to distinguish particle types under different noise conditions.

If this is right

High particle identification efficiency minimises systematic uncertainty from pion-muon misidentification in D0 to K mu nu analyses.
Background contributions from misidentified particles are reduced in charm decay studies.
The classifiers maintain useful performance under different assumptions about photosensor noise levels.
The ring reconstruction plus classifier approach supports precise measurements of particle properties in future superfactory experiments.

Where Pith is reading between the lines

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

The same classifier strategy could be tested on other decay modes that require clean pion-muon or kaon-pion separation.
Results on noise dependence could guide choices of photosensor technology during detector optimisation.
The validation procedure on a specific decay channel offers a template for checking identification algorithms in additional physics analyses.

Load-bearing premise

The detector simulation accurately models the real FARICH detector response, ring reconstruction, and photosensor noise levels for the tested single particles and decay channels.

What would settle it

A side-by-side comparison of the simulated pion-muon separation efficiency against measurements from real data taken with a FARICH prototype or installed detector would show whether the predicted performance holds.

read the original abstract

A detailed study of the particle identification by the Focusing Aerogel Ring Imaging CHerenkov subsystem at the future charm superfactory detector is presented. The dedicated signal ring reconstruction algorithm is implemented in the detector simulation, the algorithm performance is tested with single particles generated within the Aurora framework. Two Boosted Decision Trees-based classifiers for the particle identification have been developed for various assumptions about photosensor noise levels. The approach is validated with the analysis of the D0->Kmunu decays, for which the systematic uncertainty and background contribution related to the pion/muon separation performance can be minimised due to high efficiency of the particle identification algorithm.

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 is a standard simulation study applying BDTs to FARICH PID in the Aurora framework, with validation on one decay channel, but all claims hinge on unverified modeling of the real detector response.

read the letter

The main thing to know is that the paper runs a targeted simulation of ring reconstruction and BDT-based pion-muon separation for a proposed FARICH at charm superfactories, then checks the output on D0 to K mu nu events to argue that PID efficiency can cut background and systematics in that analysis. It is not introducing new algorithms or frameworks, just applying standard boosted decision trees to this specific setup under varying photosensor noise assumptions. They implement the ring finder inside Aurora, generate single-particle samples, train two classifiers, and show the approach on the decay mode. That validation step on an actual physics channel is the part that gives it some practical value beyond pure detector metrics. The work is competent on its own terms for what it sets out to do. The numbers come from controlled simulation, the training and testing split is described, and the focus on a concrete decay helps tie the PID performance to an analysis use case. Credit for shipping a reproducible simulation study with explicit noise variations. The load-bearing issue is exactly what the stress-test note flags. Every efficiency, misidentification rate, and claim about reduced uncertainty lives inside the Aurora model of hit patterns, ring quality, and noise. No test-beam data, no prototype measurements, and no data-driven cross-checks appear to anchor how well the simulation matches a real FARICH. If the modeled response deviates from what the actual photosensors and aerogel will produce, the quoted performance and the asserted background reduction both move. That is a common limitation in pre-construction studies, but it is not minor here because the central physics argument rests on it. The paper is useful for the small group designing PID subsystems for the charm superfactory or similar future flavor factories. Readers who need concrete efficiency estimates under different noise assumptions or an example of how to close the loop with a decay-mode validation will find it worth reading. It is not broad enough or novel enough to interest a general detector or ML audience. I would send it to peer review. The simulation work is solid enough to deserve referee scrutiny on the modeling assumptions and the exact performance tables, even if the authors will need to strengthen the discussion of simulation uncertainties.

Referee Report

1 major / 2 minor

Summary. The manuscript presents a simulation-based study of particle identification performance for the Focusing Aerogel Ring Imaging CHerenkov (FARICH) subsystem at future charm superfactories. A dedicated signal ring reconstruction algorithm is implemented within the Aurora framework; two Boosted Decision Tree classifiers are trained and tested on single-particle samples for different photosensor noise assumptions; the approach is then validated through a full analysis of simulated D0 → Kμν decays, with the central claim that the resulting high PID efficiency permits minimization of systematic uncertainties and background contributions associated with pion/muon separation.

Significance. Should the Aurora simulation prove sufficiently faithful to the eventual detector response, the quantitative PID efficiencies and the demonstrated reduction in analysis systematics would supply practical benchmarks for detector optimization and physics analyses at proposed charm-tau factories. The explicit use of machine-learning classifiers for ring-imaging data is a relevant methodological contribution for handling realistic noise and reconstruction effects.

major comments (1)

[Validation on D0 → Kμν decays (and single-particle performance sections)] The entire performance evaluation, including the quoted efficiencies and the asserted minimization of pion/muon misidentification systematics in the D0 → Kμν analysis, rests on the unvalidated assumption that the Aurora framework correctly reproduces real FARICH ring patterns, photosensor noise, and reconstruction behavior. No comparison to test-beam data, prototype measurements, or data-driven cross-checks is provided to bound the modeling uncertainty; this is load-bearing for the central claim.

minor comments (2)

[Abstract and methods] The abstract and methods would benefit from explicit numerical values (or a table) for the photosensor noise levels explored, rather than the generic phrase 'various assumptions'.
[Results figures] Figure captions and axis labels should state whether efficiencies are quoted with statistical uncertainties only or include any estimate of simulation systematics.

Simulated Author's Rebuttal

1 responses · 1 unresolved

We thank the referee for the careful and constructive review of our manuscript. The central concern regarding validation of the Aurora simulation framework is addressed below.

read point-by-point responses

Referee: [Validation on D0 → Kμν decays (and single-particle performance sections)] The entire performance evaluation, including the quoted efficiencies and the asserted minimization of pion/muon misidentification systematics in the D0 → Kμν analysis, rests on the unvalidated assumption that the Aurora framework correctly reproduces real FARICH ring patterns, photosensor noise, and reconstruction behavior. No comparison to test-beam data, prototype measurements, or data-driven cross-checks is provided to bound the modeling uncertainty; this is load-bearing for the central claim.

Authors: We agree that the reported performance metrics and the claimed reduction in systematics are obtained from simulation studies within the Aurora framework. As the manuscript concerns a proposed future detector at charm superfactories, no real FARICH data from the target configuration exists for direct comparison. The Aurora simulation employs established models for aerogel focusing, photon transport, and photosensor response, informed by parameters from prior aerogel RICH prototype work. To address the referee's point, we will revise the manuscript by adding an explicit discussion of simulation assumptions, key modeling parameters, and an estimate of associated uncertainties on the PID efficiencies. This addition will clarify the scope and limitations of the results for the reader. revision: yes

standing simulated objections not resolved

Provision of direct test-beam or prototype data comparisons for the specific FARICH configuration, as the relevant detectors have not yet been built.

Circularity Check

0 steps flagged

Performance evaluation based on simulation and ML training; no derivations that reduce to fitted parameters or self-referential definitions.

full rationale

The paper presents a simulation study of FARICH PID using the Aurora framework to generate single-particle events and D0->Kmunu decays. A ring reconstruction algorithm is implemented, BDT classifiers are trained and tested for varying photosensor noise assumptions, and performance is reported as efficiencies that allow reduced systematics in the decay analysis. No equations, fitted parameters renamed as predictions, self-definitional loops, or load-bearing self-citations appear in the derivation chain. Results follow directly from the simulation outputs and ML training procedure without reducing to tautological inputs by construction. The accuracy of the simulation model is a substantive external assumption rather than an internal circularity.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central performance claims rest on the accuracy of the Aurora simulation framework and the chosen noise level assumptions for photosensors; no new physical entities or free parameters fitted to real data are introduced.

free parameters (1)

photosensor noise levels
Various assumptions about noise levels are used to develop and test the two BDT classifiers.

axioms (1)

domain assumption Aurora framework simulation faithfully reproduces FARICH detector response and ring formation for single particles and D0 decays
All algorithm testing and validation depends on this simulation fidelity.

pith-pipeline@v0.9.0 · 5637 in / 1052 out tokens · 29572 ms · 2026-05-19T09:53:12.677167+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

15 extracted references · 15 canonical work pages · 1 internal anchor

[1]

Raimondi, D.N

P. Raimondi, D.N. Shatilov and M. Zobov,Beam-Beam Issues for Colliding Schemes with Large Piwinski Angle and Crabbed Waist, 2007

work page 2007
[2]

Achasov, V.E

M.N. Achasov, V.E. Blinov, A.V. Bobrov, D.A. Bodrov, A.E. Bondar’, V.S. Vorobiev et al., Experiments at the Super Charm-Tau factory, Phys. Usp.67(2024) 55

work page 2024
[3]

Achasov, X.C

M. Achasov, X.C. Ai, L.P. An, R. Aliberti, Q. An, X.Z. Bai et al.,STCF conceptual design report (Volume 1): Physics & detector,Frontiers of Physics19(2023) 14701

work page 2023
[4]

SuperCharm Tau Factory CONCEPTUAL DESIGN REPORT PART ONE (physics program, detector)

V.Anashin, A. Anisenkov, V. Aulchenko, E. Baldin, A. Barladyan, A. Barnyakov et al., “SuperCharm Tau Factory CONCEPTUAL DESIGN REPORT PART ONE (physics program, detector).” https://ctd.inp.nsk.su/wiki/images/4/47/CDR2_ScTau_en_vol1.pdf, 2018

work page 2018
[5]

A.Y.Barnyakovetal., R&DstatusofFARICHoptionforPID ,Int.J.Mod.Phys.A 39(2024)2442012

work page 2024
[6]

Belozyorova, D

M. Belozyorova, D. Maksimov, G. Razuvaev, A. Sukharev, V. Vorobyev, A. Zhadan et al.,Software framework for the Super Charm-Tau factory detector project,EPJ Web Conf.251 (2021) 03017

work page 2021
[7]

Huang, H

W. Huang, H. Li, H. Zhou, T. Li, Q. Li and X. Huang,Design and development of the core software for STCF offline data processing, Journal of Instrumentation18 (2023) P03004

work page 2023
[8]

Frank, F

M. Frank, F. Gaede, M. Petric and A. Sailer,AIDASoft/DD4hep, Oct., 2018. 10.5281/zenodo.592244

work page doi:10.5281/zenodo.592244 2018
[9]

Agostinelli, J

S. Agostinelli, J. Allison, K. Amako, J. Apostolakis, H. Araujo, P. Arce et al.,Geant4—a simulation toolkit,Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment506 (2003) 250

work page 2003
[10]

Barnyakov, V.S

A.Y. Barnyakov, V.S. Bobrovnikov, S.A. Kononov, P.D. Rogozhin and M.V. Chadeeva,Calibration and Reconstruction Algorithm for FARICH System of the Detector at Super Charm-Tau Factory,Bull. Lebedev Phys. Inst.50 (2023) 534

work page 2023
[11]

SiPM Solutions realized by KETEK

W. Hartinger, F. Wiest, P. Iskra and T. Ganka, “SiPM Solutions realized by KETEK.” https://indico.cern.ch/event/284070/sessions/53907/attachments/524514/723388/ 2013.11_-_KETEK_-_EPS-TIG_-_Ravenna_Italy_WHA_-_public.pdf , 2013

work page 2013
[12]

Shipilov, A

F. Shipilov, A. Barnyakov, V. Bobrovnikov, S. Kononov and F. Ratnikov,What Machine Learning Can Do for Focusing Aerogel Detectors, Phys. Atom. Nucl.86(2023) 864 [2312.02652]

work page internal anchor Pith review arXiv 2023
[13]

xgboost: Extreme Gradient Boosting

T. Chen, T. He, M. Benesty, V. Khotilovich, Y. Tang, H. Cho et al., “xgboost: Extreme Gradient Boosting.”https://github.com/dmlc/xgboost, 2025. – 13 –

work page 2025
[14]

Particle Data Groupcollaboration, Review of particle physics, Phys. Rev. D110(2024) 030001

work page 2024
[15]

BESIII collaboration, Study of the𝐷0 → 𝐾 − 𝜇+𝜈𝜇 Dynamics and Test of Lepton Flavor Universality with 𝐷0 → 𝐾 −ℓ+𝜈ℓ Decays, Phys. Rev. Lett.122(2019) 011804. – 14 –

work page 2019

[1] [1]

Raimondi, D.N

P. Raimondi, D.N. Shatilov and M. Zobov,Beam-Beam Issues for Colliding Schemes with Large Piwinski Angle and Crabbed Waist, 2007

work page 2007

[2] [2]

Achasov, V.E

M.N. Achasov, V.E. Blinov, A.V. Bobrov, D.A. Bodrov, A.E. Bondar’, V.S. Vorobiev et al., Experiments at the Super Charm-Tau factory, Phys. Usp.67(2024) 55

work page 2024

[3] [3]

Achasov, X.C

M. Achasov, X.C. Ai, L.P. An, R. Aliberti, Q. An, X.Z. Bai et al.,STCF conceptual design report (Volume 1): Physics & detector,Frontiers of Physics19(2023) 14701

work page 2023

[4] [4]

SuperCharm Tau Factory CONCEPTUAL DESIGN REPORT PART ONE (physics program, detector)

V.Anashin, A. Anisenkov, V. Aulchenko, E. Baldin, A. Barladyan, A. Barnyakov et al., “SuperCharm Tau Factory CONCEPTUAL DESIGN REPORT PART ONE (physics program, detector).” https://ctd.inp.nsk.su/wiki/images/4/47/CDR2_ScTau_en_vol1.pdf, 2018

work page 2018

[5] [5]

A.Y.Barnyakovetal., R&DstatusofFARICHoptionforPID ,Int.J.Mod.Phys.A 39(2024)2442012

work page 2024

[6] [6]

Belozyorova, D

M. Belozyorova, D. Maksimov, G. Razuvaev, A. Sukharev, V. Vorobyev, A. Zhadan et al.,Software framework for the Super Charm-Tau factory detector project,EPJ Web Conf.251 (2021) 03017

work page 2021

[7] [7]

Huang, H

W. Huang, H. Li, H. Zhou, T. Li, Q. Li and X. Huang,Design and development of the core software for STCF offline data processing, Journal of Instrumentation18 (2023) P03004

work page 2023

[8] [8]

Frank, F

M. Frank, F. Gaede, M. Petric and A. Sailer,AIDASoft/DD4hep, Oct., 2018. 10.5281/zenodo.592244

work page doi:10.5281/zenodo.592244 2018

[9] [9]

Agostinelli, J

S. Agostinelli, J. Allison, K. Amako, J. Apostolakis, H. Araujo, P. Arce et al.,Geant4—a simulation toolkit,Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment506 (2003) 250

work page 2003

[10] [10]

Barnyakov, V.S

A.Y. Barnyakov, V.S. Bobrovnikov, S.A. Kononov, P.D. Rogozhin and M.V. Chadeeva,Calibration and Reconstruction Algorithm for FARICH System of the Detector at Super Charm-Tau Factory,Bull. Lebedev Phys. Inst.50 (2023) 534

work page 2023

[11] [11]

SiPM Solutions realized by KETEK

W. Hartinger, F. Wiest, P. Iskra and T. Ganka, “SiPM Solutions realized by KETEK.” https://indico.cern.ch/event/284070/sessions/53907/attachments/524514/723388/ 2013.11_-_KETEK_-_EPS-TIG_-_Ravenna_Italy_WHA_-_public.pdf , 2013

work page 2013

[12] [12]

Shipilov, A

F. Shipilov, A. Barnyakov, V. Bobrovnikov, S. Kononov and F. Ratnikov,What Machine Learning Can Do for Focusing Aerogel Detectors, Phys. Atom. Nucl.86(2023) 864 [2312.02652]

work page internal anchor Pith review arXiv 2023

[13] [13]

xgboost: Extreme Gradient Boosting

T. Chen, T. He, M. Benesty, V. Khotilovich, Y. Tang, H. Cho et al., “xgboost: Extreme Gradient Boosting.”https://github.com/dmlc/xgboost, 2025. – 13 –

work page 2025

[14] [14]

Particle Data Groupcollaboration, Review of particle physics, Phys. Rev. D110(2024) 030001

work page 2024

[15] [15]

BESIII collaboration, Study of the𝐷0 → 𝐾 − 𝜇+𝜈𝜇 Dynamics and Test of Lepton Flavor Universality with 𝐷0 → 𝐾 −ℓ+𝜈ℓ Decays, Phys. Rev. Lett.122(2019) 011804. – 14 –

work page 2019