Electromagnetic Shower Reconstruction and Identification in FASER's Emulsion Detector for LHC Forward Neutrino Measurements

arXiv:2606.18517 · hep-ex · Submitted 2026-06-16 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: High Energy Physics: Experiment. The day in high-energy experiment: detectors, searches and limits.

Kai: We're looking at the paper: "Electromagnetic Shower Reconstruction and Identification in FASER's Emulsion Detector for LHC Forward Neutrino Measurements".

Mira: Here's the gist of the paper. Electromagnetic shower reconstruction and identification methods were developed for the FASERnu emulsion detector using test-beam data to validate a framework for electron neutrino identification at the LHC The reconstruction employs a clustering-based algorithm without energy-dependent tuning to determine shower axes The multi-level identification chain comprising track pre-selection, a cut-based selection, and a BDT classifier achieves combined background rejection rates of 99.99% (100 GeV) and 99.94% (200 GeV) The method reaches total reconstruction and identification efficiencies of 58.9% (100 GeV) and 70.8% (200 GeV) evaluated from simulated samples.

Kai: Experimental Setup and Data Quality Emulsion modules were exposed to electron and muon beams at CERN SPS H4 beamline in 2023, providing mono-energetic electron samples at 100 GeV and 200 GeV.

Mira: The experimental setup is shown in Figure 2 The target exposure density was approximately 8 electrons/cm squared.

Kai: Beam conditions achieved were characterized by an electron purity of >80% for the 100 GeV module and >90% for the 200 GeV module.

Title and authors: Kai: This paper's title is "Electromagnetic Shower Reconstruction and Identification in FASER's Emulsion Detector for LHC Forward Neutrino Measurements". It tells you exactly what the paper is about.

Mira: It’s essentially detailing the methods they developed for reconstructing and identifying those electromagnetic showers using their emulsion detector setup at FASER, which is key for studying neutrino interactions at the LHC.

Lev: What does that mean in plain terms? They are showing how to take raw tracks seen in the emulsion and turn them into a usable signal about whether a particle was an electron neutrino or something else >

Kai: Right. The core of it is the reconstruction pipeline they use, which involves clustering-based algorithms and multi-level identification steps to separate background from signal >

Mira: They’re highlighting that this framework works by determining shower axes without needing to tune parameters specifically for the energy of the shower, which is a big deal for flexibility >

Lev: So it’s about creating a general method that should be applicable across different energy scales at the LHC, which is what we need when scaling up detectors >

Kai: Precisely. They are showing how to use adaptive parameter selection based on local segment density during the reconstruction phase to handle those variations without prior knowledge of the shower energy >

Mira: That approach sounds flexible. It means that if you get a track with a very dense region locally, the algorithm adjusts its parameters automatically, instead of needing a specific tuning for that high-density situation >

Lev: But how much does that local density actually vary when you're looking at these forward showers? Is it stable enough to rely on this adaptation >

Kai: The paper doesn't detail the exact variation of local density across the whole volume, but they show that DBSCAN is applied independently to segments on each plate using their 2D spatial coordinates X and Y to group points in high-density regions > <ref:2606.18517#pg1>

Mira: So they’re relying on spatial clustering in a specific way, not on some energy-dependent tuning based on a calculated energy input for the shower >

Lev: That’s what I need to know. When we build real hardware, we need stability. If the clustering behavior changes drastically depending on the local environment, that makes things much harder to control >

Kai: They also use linear regression to determine the initial shower axis by fitting just the first three segments of each primary track independently in both Z-X and Z-Y projection planes >

Mira: That's a straightforward way to establish a starting point for the reconstruction, using just the very beginning of those tracks to define where they’re going >

Lev: So they don't need to wait for the whole shower to develop before they can start tracking its direction, which is good because it speeds things up >

Kai: They then define a preliminary cylindrical search volume around that initial axis, with a radius of one hundred µm and a length of eight cm >

Mira: That cylinder is where the next step happens: applying DBSCAN independently to segments on each plate using their X and Y coordinates to group points in high-density regions within that volume >

Lev: So they’re essentially creating localized groups of data based purely on how dense the points are, without needing an energy model to define what "dense" means >

Kai: That's the key feature they emphasize, which is that adaptive parameter selection is based on local segment density without requiring prior knowledge of the shower energy >

Mira: So, to summarize this section, they’re presenting a method for reconstructing shower axes and identifying them using a clustering algorithm that adapts locally to the data structure itself >

Lev: It sounds like they’ve solved one of the main hurdles in these emulsion experiments: getting robust geometry estimation without relying on complex energy modeling assumptions >

Kai: Exactly. It’s about making the reconstruction more self-sufficient when you don're dealing with a broad range of possible physics outcomes >

The paper's summary: Kai: Okay, so we’ve talked about the reconstruction method, now let’s look at what they actually found in terms of performance. This is about the results section of "Electromagnetic Shower Reconstruction and Identification in FASER's Emulsion Detector for LHC Forward Neutrino Measurements".

Mira: They are showing that their multi-level identification chain, which goes from track pre-selection to a cut-based selection and then the BDT classifier, yields very high background rejection rates.

Lev: What kind of numbers are we talking about here? Are we talking about something like ninety percent or ninety-five percent rejection?

Kai: They report combined background rejection rates of ninety-nine point nine nine percent at one hundred GeV and two hundred GeV, and then a slightly lower rate of ninety-nine point eight four percent at the higher energy > <ref:2606.18517#pg3>

Mira: That's extremely high, which is what you want when you are trying to distinguish genuine EM showers from the dominant muon background that mimics them in these detectors >

Lev: So they managed to keep that high rejection while still achieving decent signal efficiency, otherwise we wouldn't get much data to analyze later >

Kai: They also report the total reconstruction and identification efficiencies for simulated samples, showing fifty-eight point nine percent at one hundred GeV and seventy point eight percent at two hundred GeV >

Mira: So they’re saying that even with this high rejection, you still manage to pull out a significant fraction of the true signal that we care about >

Lev: And what about the energy measurement itself? How accurate are they when estimating the energy of these showers using just the segments?

Kai: They use the total number of segments across all plates in a reconstructed shower, Ntotal, as their estimator for calorimetric energy >

Mira: When they look at those numbers, relative biases are +zero point six percent at one hundred GeV and-zero point eight percent at two hundred GeV with resolutions of twenty-five point four percent and twenty-two point six percent, respectively >

Lev: That means the energy resolution is pretty good for a calorimetric estimate, but those biases are small enough that they’re not going to ruin our analysis significantly, right?

Kai: They also point out that systematic uncertainties on the reconstructed energy are dominated by single-film efficiency variations >

Mira: So it’s telling us where the detector imperfections are having the biggest impact on our energy estimates, which is good for understanding what to fix next >

The paper's improvements: Lev: Now that we know how they do it, let’s talk about what they suggest as improvements over previous work. What’s the actual upgrade path?

Kai: They refined the reconstruction algorithm by combining adaptive DBSCAN clustering, sequential centroid validation with a sixty µm deviation threshold, and Huber robust fitting to determine shower axes > <ref:2606.18517#pg3>

Mira: That combination of techniques is what allows them to get those reliable axes without relying on energy-dependent parameter tuning, which is a big methodological improvement >

Lev: So they’re essentially saying that you can get this robust geometric reconstruction using these specific mathematical tools instead of whatever complex model was used before >

Kai: They also mentioned that the ratio-based and angular variables are designed to remain stable across the broad FASERnu energy range, which is what enables a single trained classifier to be applied at fifty GeV and above without needing any retuning >

Mira: That's really practical. If you can have one classifier that works for both one hundred GeV and two hundred GeV, you save a ton of work on the analysis side > <ref:2606.18517#pg3>

Lev: It simplifies the downstream analysis significantly, but what about the limitations they explicitly state in this paper? Where does it stop working?

Kai: They do flag that their position resolution is evaluated with a track-extrapolation method for each film, and most plates show resolutions of zero point five to one point five µm >

Mira: And they also mention that some films get degraded up to about three µm because of things like variations in film alignment quality, emulsion distortions, and scanning conditions > <ref:2606.18517#pg1>

Lev: So the limitation isn't just the resolution itself; it’s the physical quality of the film setup which affects how well we can measure those positions >

Kai: And for angular resolution, they use a segment-difference method to evaluate track direction, and they note that this resolution is systematically larger in the horizontal projection in both modules because of the elliptical angular distribution of the test beam >

Mira: So to summarize these improvements, it’s about making the reconstruction more robust using adaptive clustering and robust fitting without needing energy-dependent tuning for axis determination >

Lev: It sounds like they’ve put together a solid set of tools for a broad energy range, but we still have to deal with the physical limitations of the emulsion itself >

Conclusion: Kai: So wrapping up on this paper, the main implication is that this detailed work establishes a validated framework for electron neutrino identification at the LHC using FASERnu data >

Mira: They’re essentially showing that their specific reconstruction and identification chain achieves background rejection rates of ninety-nine point nine nine percent at one hundred GeV and two hundred GeV > <ref:2606.18517#pg3>

Lev: That high level of rejection means we have a very clean way to search for the physics they are looking for in the forward region >

Kai: And they’ve shown that energy reconstruction using total reconstructed segments yields relative biases of +zero point six percent at one hundred GeV and-zero point eight percent at two hundred GeV >

Mira: So, overall, this work provides a validated pipeline for identifying these showers in the FASERnu detector for LHC forward neutrino measurements >

Lev: It’s a solid tool to use as we plan out the next phases of experimental design and data analysis based on what they’ve shown here >

Kai: This paper, "Electromagnetic Shower Reconstruction and Identification in FASER's Emulsion Detector for LHC Forward Neutrino Measurements", is a piece of work that shows how careful methodology can lead to solid results >

Mira: It validates the reconstruction and identification pipeline for electron neutrino identification at the LHC using test-beam data >

Lev: We’ll keep this as a useful reference when we start designing detectors that need to handle these types of challenges >

Roshan Mammen Abraham, Xiaocong Ai, Saul Alonso Monsalve, John Anders, Emma Kate Anderson, Akitaka Ariga, È Tomoko Ariga, f Jeremy Atkinson, Ɠ Florian U. Bernlochner, ᐫ Jianming Bian, Tobias Boeckh, Eli Bornand, Jamie Boyd, Lydia Brenner, Angela Burger, Franck Cadoux, Roberto Cardella, David W. Casper, Charlotte Cavanagh, Shiyang Chen, Xin Chen, Xing Cheng, Dhruv Chouhan, Andrea Coccaro, Fabio Cufino, Stephane Débieux Ansh Desai, œ Sergey Dmitrievsky, Radu Dobre, Monica D’Onofrio, Sinead Eley, Yannick Favre, Jonathan L. Feng, Carlo Alberto Fenoglio, Didier Ferrere, Max Fieg, Wissal Filali Elena Firu Haruhi Fujimori Edward Galantay Stephen Gibson Yuri Gornushkin Yotam Granov Jinjing Gu Carl Gwilliam Elie Hammou Daiki Hayakawa Michael Holzbock Shih-Chieh Hsu Zhen Hu Lekai Yao Daichi Yoshikawa Shunliang Zhang Yuxuan Zhang Xingyu Zhao, Eli Welch, Aaron White, Monika Wielers, Benjamin James Wilson Zhongyi Wu, Yue Xu Heng Yang Yi Liu Jiaxi Liu Jinfeng Liu Yi Liu Margaret S. Lutz Joern Mahlstedt Toni Mäkelä Yasuhiro Maruya Anna Mascellani Lawson McCoy Josh McFayden Andrea Pizarro Medina Hiroaki Menjo Toshiyuki Nakano Laurie Nevay Yuma Ohara Ken Ohashi Hidetoshi Otono Lorenzo Paolozzi Annabelle Parry Pawan Pawan Brian Petersen Titi Preda Markus Prim Junkai Qin Michaela Queitsch-Maitland Hiroki Rokujo André Rubbia Osamu Sato Paola Scampoli Kristof Schmieden Matthias Schott Cristiano Sebastiani Anna Sfyrla Davide Sgalaberna Mansoora Shamim Yosuke Takubo Kakeru Tanaka Noshin Tarannum Simon Thor Eric Torrence Serhan Tufanli Oscar Ivan Valdes Martinez Svetlana Vasina Emanuele Villa Benedikt Vormwald Chi Wang Yuxiao Wang, Eli Welch, Aaron White, Monika Wielers

Department of Physics and Astronomy, University of California, Irvine · School of Physics, Zhengzhou University · Institute for Particle Physics, ETH Zurich · University of Liverpool · É CERN

hep-ex

Submitted: 2026-06-16

Updated: 2026-08-24

Comments: 23 pages, 18 figures

Journal ref: JINST 21 (2026) P10044

DOI: 10.1088/1748-0221/21/10/P10044

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 87/100

The gist: Electromagnetic shower reconstruction and identification methods were developed for the FASERnu emulsion detector using test-beam data to validate a framework for electron neutrino identification at…

Key concepts

Adaptive Clustering
This is the technique used to find the shower axis. Instead of setting fixed parameters based on expected energy, the algorithm adjusts its search volume and grouping method based on how densely packed the actual segments are locally. This allows for accurate reconstruction regardless of whether a shower is at 100 GeV or 200 GeV.
Multi-level Identification Chain
This is a three-step process to separate genuine showers from background noise (like muons). It starts with filtering tracks, then uses simple cuts to remove most remaining muons, and finally employs a sophisticated Boosted Decision Tree (BDT) classifier. This layered approach ensures very high purity while keeping the signal efficiency high.
BDT Classifier Features
The BDT classifier distinguishes showers by analyzing ten specific shape variables: six related to the shower's length (longitudinal), three related to its width (transverse), clustering metrics, and one angular variable. These features are chosen because they remain stable across different energies, allowing one trained model to work effectively for 50 GeV and above.
Energy Reconstruction via Segments
The total number of reconstructed segments ($N_{total}$) in the shower is used as an estimator for its energy. This method provides relative energy biases that are predictable based on the beam energy (e.g., +0.6% at 100 GeV). The resolution of this estimate is relatively high, allowing for calibrated measurements despite detector imperfections.

Terminology

Summary

Electromagnetic shower reconstruction and identification methods were developed for the FASERnu emulsion detector using test-beam data to validate a framework for electron neutrino identification at the LHC

The reconstruction employs a clustering-based algorithm without energy-dependent tuning to determine shower axes The multi-level identification chain comprising track pre-selection, a cut-based selection, and a BDT classifier achieves combined background rejection rates of 99.99% (100 GeV) and 99. p. 1

Where the paper says it

17Institute of Space Science—INFLPR Subsidiary, Bucharest, Romania 18Theoretical Physics Division, Fermi National Accelerator Laboratory, Batavia, IL 60510, USA 19Royal Holloway, University of London, Egham, TW20 0EX, United Kingdom 20Department of Physics and Astronomy, Technion—Israel Institute of Technology, Haifa 32000, Israel 21Department of Physics, University of Washington, PO Box 351560, Seattle, WA 98195-1460, USA 22Department of Particle Physics and Astrophysics, Weizmann Institute of Science, Rehovot 76100, Israel 23Department of Physics & Astronomy, University of Sussex, Sussex House, Falmer, Brighton, BN1 9RH, United Kingdom 24Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8602, Japan 25University of Manchester, School of Physics and Astronomy, Schuster Building, Oxford Rd, Manchester M13 9PL, United Kingdom 26Dipartimento di Fisica “Ettore Pancini”, Universit`a di Napoli Federico II, Complesso Universitario di Monte S. Angelo, I-80126 Napoli, Italy 27National Institute of Technology (KOSEN), Niihama College, 7-1, Yakumo-cho Niihama, 792-0805 Ehime, Japan 28Particle Physics Department, STFC Rutherford Appleton Laboratory, Harwell Campus, Didcot, OX11 0QX, United Kingdom We present methods for electromagnetic shower reconstruction and identification in the FASERν emulsion detector using 100 GeV and 200 GeV electron test-beam data from the CERN SPS H4 beamline.

94% (200 GeV) The method reaches total reconstruction and identification efficiencies of 58. p. 1
Where the paper says it

94% (200 GeV). The method reaches total reconstruction and identification efficiencies of 58.

9% (100 GeV) and 70. p. 1
Where the paper says it

17Institute of Space Science—INFLPR Subsidiary, Bucharest, Romania 18Theoretical Physics Division, Fermi National Accelerator Laboratory, Batavia, IL 60510, USA 19Royal Holloway, University of London, Egham, TW20 0EX, United Kingdom 20Department of Physics and Astronomy, Technion—Israel Institute of Technology, Haifa 32000, Israel 21Department of Physics, University of Washington, PO Box 351560, Seattle, WA 98195-1460, USA 22Department of Particle Physics and Astrophysics, Weizmann Institute of Science, Rehovot 76100, Israel 23Department of Physics & Astronomy, University of Sussex, Sussex House, Falmer, Brighton, BN1 9RH, United Kingdom 24Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8602, Japan 25University of Manchester, School of Physics and Astronomy, Schuster Building, Oxford Rd, Manchester M13 9PL, United Kingdom 26Dipartimento di Fisica “Ettore Pancini”, Universit`a di Napoli Federico II, Complesso Universitario di Monte S. Angelo, I-80126 Napoli, Italy 27National Institute of Technology (KOSEN), Niihama College, 7-1, Yakumo-cho Niihama, 792-0805 Ehime, Japan 28Particle Physics Department, STFC Rutherford Appleton Laboratory, Harwell Campus, Didcot, OX11 0QX, United Kingdom We present methods for electromagnetic shower reconstruction and identification in the FASERν emulsion detector using 100 GeV and 200 GeV electron test-beam data from the CERN SPS H4 beamline.

8% (200 GeV) evaluated from simulated samples. p. 1
Where the paper says it

9% (100 GeV) and 70. 8% (200 GeV) evaluated from simulated samples.

Experimental Setup and Data Quality

Emulsion modules were exposed to electron and muon beams at CERN SPS H4 beamline in 2023, providing mono-energetic electron samples at 100 GeV and 200 GeV. The experimental setup is shown in Figure 2 The target exposure density was approximately 8 electrons/cm squared. p. 4

Where the paper says it

5 • 200 GeV (module T1101): 40–60 particles per spill, electron purity >90%. The target exposure density was approximately 8 electrons/cm2 .

Beam conditions achieved were characterized by an electron purity of >80% for the 100 GeV module and >90% for the 200 GeV module. p. 3
Where the paper says it

A right-handed coordinate system is adopted throughout this paper: the Z-axis points along the beam direction, X is horizontal transverse, and Y is vertical transverse. The achieved beam conditions were: • 100 GeV (module T1202): 40–50 particles per spill, electron purity >80%;

Shower Reconstruction Pipeline

The reconstruction begins with tracks originating from the first emulsion plate, and the initial shower axis is determined by fitting the first three segments of each primary track independently in the Z-X and Z-Y projection planes using linear regression. A preliminary cylindrical search volume is then defined around the initial axis, with radius 100 µm and length 8 cm. p. 7

Where the paper says it

8 mrad) while minimizing deflections from early shower development and multiple scattering; using fewer plates gives insufficient constraints, while more plates increasingly biases the fit through shower scattering. A preliminary cylindrical search volume is then defined around the initial axis, with radius 100 µm and length 8 cm (approximately 18 radiation lengths, spanning 56 plates).

Within this volume, DBSCAN is applied independently to segments on each plate using their 2D spatial coordinates (X, Y) to group points in high-density regions. p. 7
Where the paper says it

Within the preliminary cylindrical volume, DBSCAN is applied independently to segments on each plate using their 2D spatial coordinates (X, Y). It groups points in high-density regions while classifying isolated points as noise, making it well-suited for separating the dense EM shower core from background tracks and peripheral shower particles.

The key feature of the implementation is adaptive parameter selection based on local segment density, without requiring prior knowledge of the shower energy. p. 7
Where the paper says it

The algorithm requires two input parameters: a neighbourhood radius ε, within which points are considered neighbours, and a minimum cluster size Nmin, the minimum number of neighbouring points required to form a cluster. The key feature of our implementation is adaptive parameter selection based on local segment density, without requiring prior knowledge of the shower energy.

Identification Strategy and Performance

A multi-level identification strategy is employed to distinguish genuine EM showers from the dominant muon background. After shower reconstruction, a cut-based selection (Loose ID) removes the bulk of remaining muon tracks that mimic shower topologies through accidental segment accumulation in the cylindrical volume. p. 5

Where the paper says it

SHOWER RECONSTRUCTION AND IDENTIFICATION Figure 3 provides an overview of the complete analysis chain. The pipeline proceeds in three stages: pre-selection applies track-level quality cuts to reject obvious muon backgrounds; shower reconstruction uses adaptive DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering [20] and Huber robust axis fitting to collect shower segments within a cylindrical volume; identification applies cut-based Loose ID followed by BDT-based Medium ID to distinguish genuine electrons from residual backgrounds.

The BDT-based Medium ID provides additional discrimination using ten shower shape variables. p. 5
Where the paper says it

SHOWER RECONSTRUCTION AND IDENTIFICATION Figure 3 provides an overview of the complete analysis chain. The pipeline proceeds in three stages: pre-selection applies track-level quality cuts to reject obvious muon backgrounds; shower reconstruction uses adaptive DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering [20] and Huber robust axis fitting to collect shower segments within a cylindrical volume; identification applies cut-based Loose ID followed by BDT-based Medium ID to distinguish genuine electrons from residual backgrounds.

The BDT classifier achieves an area under the ROC curve (AUC) of 0. p. 1
Where the paper says it

The reconstruction employs a clustering-based algorithm without energy-dependent tuning to determine shower axes. A multi-level identification chain comprising track pre-selection, a cut-based selection, and a BDT classifier achieves combined background rejection rates of 99.

997 when trained on 2517 electrons from MC simulation and 3517 muon candidates from test beam data. p. 9
Where the paper says it

The same visual confirmation is applied when measuring muon background rejection rates in test beam data (Section 4), where tracks surviving the identification chain are inspected to verify whether they are genuine muon backgrounds or misidentified EM showers. Both selections are performed by trained scanners familiar with emulsion event topologies.

Energy Reconstruction and Uncertainties

The total number of segments across all plates in the reconstructed shower, Ntotal, is used as the calorimetric energy estimator. Energy reconstruction using the total number of reconstructed segments yields relative biases of +0. p. 1

Where the paper says it

8% (200 GeV) evaluated from simulated samples. Energy reconstruction using the total number of reconstructed segments as the calorimetric estimator yields relative biases of +0.

6% (100 GeV) and −0. p. 1
Where the paper says it

17Institute of Space Science—INFLPR Subsidiary, Bucharest, Romania 18Theoretical Physics Division, Fermi National Accelerator Laboratory, Batavia, IL 60510, USA 19Royal Holloway, University of London, Egham, TW20 0EX, United Kingdom 20Department of Physics and Astronomy, Technion—Israel Institute of Technology, Haifa 32000, Israel 21Department of Physics, University of Washington, PO Box 351560, Seattle, WA 98195-1460, USA 22Department of Particle Physics and Astrophysics, Weizmann Institute of Science, Rehovot 76100, Israel 23Department of Physics & Astronomy, University of Sussex, Sussex House, Falmer, Brighton, BN1 9RH, United Kingdom 24Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8602, Japan 25University of Manchester, School of Physics and Astronomy, Schuster Building, Oxford Rd, Manchester M13 9PL, United Kingdom 26Dipartimento di Fisica “Ettore Pancini”, Universit`a di Napoli Federico II, Complesso Universitario di Monte S. Angelo, I-80126 Napoli, Italy 27National Institute of Technology (KOSEN), Niihama College, 7-1, Yakumo-cho Niihama, 792-0805 Ehime, Japan 28Particle Physics Department, STFC Rutherford Appleton Laboratory, Harwell Campus, Didcot, OX11 0QX, United Kingdom We present methods for electromagnetic shower reconstruction and identification in the FASERν emulsion detector using 100 GeV and 200 GeV electron test-beam data from the CERN SPS H4 beamline.

8% (200 GeV), with resolutions of 25. p. 1
Where the paper says it

6% (100 GeV) and −0. 8% (200 GeV), with resolutions of 25.

4% and 22.6%, respectively. p. 1
Where the paper says it

8% (200 GeV), with resolutions of 25. 6%, respectively.

Systematic uncertainties on the reconstructed energy are dominated by single-film efficiency variations, contributing +10. p. 1
Where the paper says it

6%, respectively. Systematic uncertainties on the energy reconstruction are dominated by variations in emulsion film detection efficiency, with totals of +10.

8−8.1 % and +10.2−6.8 % at the two energies. p. 6
Where the paper says it

8 GeV, while the upper bound removes muon tracks traversing most of the detector. The different upper bounds at the two energies reflect a module-dependent tracking quality effect and are determined by MC simulation samples.

p. 6
Where the paper says it

8 GeV, while the upper bound removes muon tracks traversing most of the detector. The different upper bounds at the two energies reflect a module-dependent tracking quality effect and are determined by MC simulation samples.

Conclusion

The reconstruction algorithm combines adaptive DBSCAN clustering, sequential centroid validation with a 60 µm deviation threshold, and Huber robust fitting to determine shower axes without energy-dependent parameter tuning. A multi-level identification chain comprising track pre-selection, cut-based Loose ID, and BDTbased Medium ID achieves combined background rejection rates of 99. p. 1

Where the paper says it

The reconstruction employs a clustering-based algorithm without energy-dependent tuning to determine shower axes. A multi-level identification chain comprising track pre-selection, a cut-based selection, and a BDT classifier achieves combined background rejection rates of 99.

99% at 100 GeV and 99. p. 1
Where the paper says it

17Institute of Space Science—INFLPR Subsidiary, Bucharest, Romania 18Theoretical Physics Division, Fermi National Accelerator Laboratory, Batavia, IL 60510, USA 19Royal Holloway, University of London, Egham, TW20 0EX, United Kingdom 20Department of Physics and Astronomy, Technion—Israel Institute of Technology, Haifa 32000, Israel 21Department of Physics, University of Washington, PO Box 351560, Seattle, WA 98195-1460, USA 22Department of Particle Physics and Astrophysics, Weizmann Institute of Science, Rehovot 76100, Israel 23Department of Physics & Astronomy, University of Sussex, Sussex House, Falmer, Brighton, BN1 9RH, United Kingdom 24Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8602, Japan 25University of Manchester, School of Physics and Astronomy, Schuster Building, Oxford Rd, Manchester M13 9PL, United Kingdom 26Dipartimento di Fisica “Ettore Pancini”, Universit`a di Napoli Federico II, Complesso Universitario di Monte S. Angelo, I-80126 Napoli, Italy 27National Institute of Technology (KOSEN), Niihama College, 7-1, Yakumo-cho Niihama, 792-0805 Ehime, Japan 28Particle Physics Department, STFC Rutherford Appleton Laboratory, Harwell Campus, Didcot, OX11 0QX, United Kingdom We present methods for electromagnetic shower reconstruction and identification in the FASERν emulsion detector using 100 GeV and 200 GeV electron test-beam data from the CERN SPS H4 beamline.

94% at 200 GeV. p. 1
Where the paper says it

17Institute of Space Science—INFLPR Subsidiary, Bucharest, Romania 18Theoretical Physics Division, Fermi National Accelerator Laboratory, Batavia, IL 60510, USA 19Royal Holloway, University of London, Egham, TW20 0EX, United Kingdom 20Department of Physics and Astronomy, Technion—Israel Institute of Technology, Haifa 32000, Israel 21Department of Physics, University of Washington, PO Box 351560, Seattle, WA 98195-1460, USA 22Department of Particle Physics and Astrophysics, Weizmann Institute of Science, Rehovot 76100, Israel 23Department of Physics & Astronomy, University of Sussex, Sussex House, Falmer, Brighton, BN1 9RH, United Kingdom 24Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8602, Japan 25University of Manchester, School of Physics and Astronomy, Schuster Building, Oxford Rd, Manchester M13 9PL, United Kingdom 26Dipartimento di Fisica “Ettore Pancini”, Universit`a di Napoli Federico II, Complesso Universitario di Monte S. Angelo, I-80126 Napoli, Italy 27National Institute of Technology (KOSEN), Niihama College, 7-1, Yakumo-cho Niihama, 792-0805 Ehime, Japan 28Particle Physics Department, STFC Rutherford Appleton Laboratory, Harwell Campus, Didcot, OX11 0QX, United Kingdom We present methods for electromagnetic shower reconstruction and identification in the FASERν emulsion detector using 100 GeV and 200 GeV electron test-beam data from the CERN SPS H4 beamline.

The ratio-based and angular variables are designed to remain stable across the broad FASERnu energy range, enabling a single trained classifier to be applied at 50 GeV and above without retuning. p. 1
Where the paper says it

The reconstruction employs a clustering-based algorithm without energy-dependent tuning to determine shower axes. A multi-level identification chain comprising track pre-selection, a cut-based selection, and a BDT classifier achieves combined background rejection rates of 99.

Energy reconstruction using the total reconstructed segments yields relative biases of +0. p. 1
Where the paper says it

8% (200 GeV) evaluated from simulated samples. Energy reconstruction using the total number of reconstructed segments as the calorimetric estimator yields relative biases of +0.

6% at 100 GeV and −0. p. 1
Where the paper says it

17Institute of Space Science—INFLPR Subsidiary, Bucharest, Romania 18Theoretical Physics Division, Fermi National Accelerator Laboratory, Batavia, IL 60510, USA 19Royal Holloway, University of London, Egham, TW20 0EX, United Kingdom 20Department of Physics and Astronomy, Technion—Israel Institute of Technology, Haifa 32000, Israel 21Department of Physics, University of Washington, PO Box 351560, Seattle, WA 98195-1460, USA 22Department of Particle Physics and Astrophysics, Weizmann Institute of Science, Rehovot 76100, Israel 23Department of Physics & Astronomy, University of Sussex, Sussex House, Falmer, Brighton, BN1 9RH, United Kingdom 24Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8602, Japan 25University of Manchester, School of Physics and Astronomy, Schuster Building, Oxford Rd, Manchester M13 9PL, United Kingdom 26Dipartimento di Fisica “Ettore Pancini”, Universit`a di Napoli Federico II, Complesso Universitario di Monte S. Angelo, I-80126 Napoli, Italy 27National Institute of Technology (KOSEN), Niihama College, 7-1, Yakumo-cho Niihama, 792-0805 Ehime, Japan 28Particle Physics Department, STFC Rutherford Appleton Laboratory, Harwell Campus, Didcot, OX11 0QX, United Kingdom We present methods for electromagnetic shower reconstruction and identification in the FASERν emulsion detector using 100 GeV and 200 GeV electron test-beam data from the CERN SPS H4 beamline.

8% at 200 GeV, with resolutions of 25. p. 1
Where the paper says it

6% (100 GeV) and −0. 8% (200 GeV), with resolutions of 25.

4+2.1−2.4% and 22.6+1.5−1.2%. The systematic uncertainties on the reconstructed energy are dominated by single-film efficiency variations, contributing +10. p. 1
Where the paper says it

6%, respectively. Systematic uncertainties on the energy reconstruction are dominated by variations in emulsion film detection efficiency, with totals of +10.

8−8.1 % and +10.2−6.8 % at the two energies. Background contamination remains below 0. p. 7
Where the paper says it

For each selected cluster, the spatial centroid is computed: (X, ¯ Y , ¯ Z¯) = 1 Ncluster N Xcluster i=1 (Xi , Yi , Zi) (4) Figure 5 shows an example of DBSCAN clustering on a shower from the 200 GeV electron MC sample. To suppress contamination from peripheral shower particles that may create spurious centroids far from the true shower axis, a sequential validation procedure is applied before the final axis fit.

1% under test-beam conditions. p. 3
Where the paper says it

A right-handed coordinate system is adopted throughout this paper: the Z-axis points along the beam direction, X is horizontal transverse, and Y is vertical transverse. The achieved beam conditions were: • 100 GeV (module T1202): 40–50 particles per spill, electron purity >80%;

Acknowledgments

We thank the CERN EP department for the refurbishment of the CERN dark room, used for the preparation and development of the FASERnu emulsion We thank Saya Yamamoto for supporting the preparation of emulsion films We thank the CERN beam group, especially Nikolaos Charitonidis, for excellent beam operation at the SPS-H4 beamline. This work was supported in part by Heising-Simons Foundation Grant Nos. 2019-1179 and 2020-1840 Simons Foundation Grant No. 623683, JSPS KAKENHI Grant Nos. 19H01909, 22H01233, 20K23373, 23H00103, 20H01919, and 21H00082. The joint research program of the Institute of Materials and Systems for Sustainability, ERC Consolidator Grant No. p. 6

Where the paper says it

After track reconnection and pre-selection, approximately 29% of tracks originating from the first plate survive at both energies in test-beam data. Figure 4 shows the joint distributions of θs and nseg for electron MC and test-beam data in both modules: muon tracks populate the region of large nseg and small θs, well separated from the short, large-scattering tracks characteristic of EM showers.

101002690, DFG grant SCHO 1527/13-1, Royal Society Grant No. URF R1 201519, UK Science and Technology Funding Councils Grant No. p. 1
Where the paper says it

2 10D´epartement de Physique Nucl´eaire et Corpusculaire, University of Geneva, CH-1211 Geneva 4, Switzerland 11Nikhef National Institute for Subatomic Physics, Science Park 105, 1098 XG Amsterdam, Netherlands 12L2IT, Universit´e de Toulouse, CNRS/IN2P3, UPS, Toulouse, France 13Department of Physics, Tsinghua University, Beijing, China 14INFN Sezione di Genova, Via Dodecaneso, 33–16146, Genova, Italy 15University of Oregon, Eugene, OR 97403, USA 16Affiliated with an international laboratory covered by a cooperation agreement with CERN. 17Institute of Space Science—INFLPR Subsidiary, Bucharest, Romania 18Theoretical Physics Division, Fermi National Accelerator Laboratory, Batavia, IL 60510, USA 19Royal Holloway, University of London, Egham, TW20 0EX, United Kingdom 20Department of Physics and Astronomy, Technion—Israel Institute of Technology, Haifa 32000, Israel 21Department of Physics, University of Washington, PO Box 351560, Seattle, WA 98195-1460, USA 22Department of Particle Physics and Astrophysics, Weizmann Institute of Science, Rehovot 76100, Israel 23Department of Physics & Astronomy, University of Sussex, Sussex House, Falmer, Brighton, BN1 9RH, United Kingdom 24Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8602, Japan 25University of Manchester, School of Physics and Astronomy, Schuster Building, Oxford Rd, Manchester M13 9PL, United Kingdom 26Dipartimento di Fisica “Ettore Pancini”, Universit`a di Napoli Federico II, Complesso Universitario di Monte S.

ST T505870/T 4. The National Natural Science Foundation of China, Tsinghua University Initiative Scientific Research Program, and the Swiss National Science Foundation. p. 1
Where the paper says it

2 10D´epartement de Physique Nucl´eaire et Corpusculaire, University of Geneva, CH-1211 Geneva 4, Switzerland 11Nikhef National Institute for Subatomic Physics, Science Park 105, 1098 XG Amsterdam, Netherlands 12L2IT, Universit´e de Toulouse, CNRS/IN2P3, UPS, Toulouse, France 13Department of Physics, Tsinghua University, Beijing, China 14INFN Sezione di Genova, Via Dodecaneso, 33–16146, Genova, Italy 15University of Oregon, Eugene, OR 97403, USA 16Affiliated with an international laboratory covered by a cooperation agreement with CERN. 17Institute of Space Science—INFLPR Subsidiary, Bucharest, Romania 18Theoretical Physics Division, Fermi National Accelerator Laboratory, Batavia, IL 60510, USA 19Royal Holloway, University of London, Egham, TW20 0EX, United Kingdom 20Department of Physics and Astronomy, Technion—Israel Institute of Technology, Haifa 32000, Israel 21Department of Physics, University of Washington, PO Box 351560, Seattle, WA 98195-1460, USA 22Department of Particle Physics and Astrophysics, Weizmann Institute of Science, Rehovot 76100, Israel 23Department of Physics & Astronomy, University of Sussex, Sussex House, Falmer, Brighton, BN1 9RH, United Kingdom 24Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8602, Japan 25University of Manchester, School of Physics and Astronomy, Schuster Building, Oxford Rd, Manchester M13 9PL, United Kingdom 26Dipartimento di Fisica “Ettore Pancini”, Universit`a di Napoli Federico II, Complesso Universitario di Monte S.

Appendix A

The position resolution is evaluated with a track-extrapolation method for each film. Most plates exhibit resolutions of 0. p. 5

Where the paper says it

Overview of the EM shower reconstruction and identification pipeline, showing the three main stages: pre-selection, reconstruction, and identification. Preliminary Selection on Tracks Muon tracks traversing the full detector depth are characterized by large segment counts and small scattering angles, whereas EM shower primary tracks are short (initiating cascades within a few radiation lengths) and exhibit larger apparent scattering due to early pair production and bremsstrahlung.

5–1.5 µm, with several films degraded up to ∼3 µm due to variations in film alignment quality, emulsion distortions, and scanning conditions. p. 2
Where the paper says it

Schematic view of the FASERν detector [4], consisting of tungsten plates interleaved with OPERAtype emulsion films. Each emulsion film comprises two 65 µm emulsion layers on each side of a 210 µm polystyrene base [6].

The angular resolution is evaluated with a segment-difference method for the track direction. p. 7
Where the paper says it

A preliminary cylindrical search volume is then defined around the initial axis, with radius 100 µm and length 8 cm (approximately 18 radiation lengths, spanning 56 plates). Segments are included in the preliminary shower candidate if they belong to a track with at least 3 segments (Nseg ≥ 3), have a maximum perpendicular distance from the axis ≤ 100 µm, a maximum angular deviation between the segment direction and the initial axis direction ≤ 10 mrad, and a minimum distance of closest approach between the segment trajectory and the initial axis ≤ 50 µm.

The resolution is systematically larger in the horizontal projection in both modules, reflecting the elliptical angular distribution of the test beam. p. 4
Where the paper says it

Measured data quality metrics for both test beam modules, evaluated using the methods described in Ref. Parameter 100 GeV (T1202) 200 GeV (T1101) Position resolution 0.

For each target film i, the single-film efficiency is defined using four films bracketing the trajectory. p. 4
Where the paper says it

45 mrad Angular resolution (σtan θy ) 1. 44 mrad Single-film efficiency (peak) ∼85% ∼95% 2.

The 100 GeV module peaks at ∼ 85% with a spread of σ = 6. p. 3
Where the paper says it

A right-handed coordinate system is adopted throughout this paper: the Z-axis points along the beam direction, X is horizontal transverse, and Y is vertical transverse. The achieved beam conditions were: • 100 GeV (module T1202): 40–50 particles per spill, electron purity >80%;

7%; the 200 GeV module peaks at ∼ 95% with a spread of σ = 3. p. 4
Where the paper says it

5 • 200 GeV (module T1101): 40–60 particles per spill, electron purity >90%. The target exposure density was approximately 8 electrons/cm2 .

9%.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Adaptive clustering for shower reconstruction allows systems to determine shower axes without energy-dependent parameter tuning, enabling reliable identification across a broad energy range from 50 GeV to multi-TeV without requiring separate models for different energies.

  2. A multi-level identification chain, comprising track pre-selection, a cut-based selection, and a BDT classifier, achieves extremely high background rejection rates of 99.99% at 100 GeV and 99.94% at 200 GeV while maintaining high signal efficiency.

  3. The BDT classifier uses ten shower shape variables: six longitudinal, three transverse and clustering, and one angular, which are specifically designed to remain stable across the broad FASERν energy range, enabling a single trained classifier to be applied at 50 GeV and above without retuning.

  4. Energy reconstruction using the total number of segments yields relative biases of +0.6% at 100 GeV and -0.8% at 200 GeV, providing calibrated energy estimates with resolutions of 25.4 +/- 2.1% at 100 GeV and 22.6 +/- 1.5% at 200 GeV.

  5. The system can quantify the impact of detector imperfections by measuring how single-film efficiency variations systematically shift the observed shower multiplicity and consequently the reconstructed energy, allowing for precise systematic uncertainty modeling.

Abstract

We present methods for electromagnetic shower reconstruction and identification in the FASERnu emulsion detector using 100 GeV and 200 GeV electron test-beam data from the CERN SPS H4 beamline. The reconstruction employs a clustering-based algorithm without energy-dependent tuning to determine shower axes. A multi-level identification chain comprising track pre-selection, a cut-based selection, and a BDT classifier achieves combined background rejection rates of 99.99% (100 GeV) and 99.94% (200 GeV). The method reaches total reconstruction and identification efficiencies of 58.9% (100 GeV) and 70.8% (200 GeV) evaluated from simulated samples. Energy reconstruction using the total number of reconstructed segments as the calorimetric estimator yields relative biases of +0.6% (100 GeV) and-0.8% (200 GeV), with resolutions of 25.4% and 22.6%, respectively. Systematic uncertainties on the energy reconstruction are dominated by variations in emulsion film detection efficiency, with totals of (+10.9%/-8.2%) at 100 GeV and (+10.3%/-6.9%) at 200 GeV. The methodology provides a validated framework for electron neutrino identification with the FASERnu detector at the LHC.

Sources

Related papers