Reconstruction of overlapping electromagnetic showers in calorimeters using Transformers

{"Abstract":[0],"Accurate":[1],"clustering":[2,36],"of":[3,117,178,277],"electromagnetic":[4],"energy":[5,30,41,149,250,278],"deposits":[6],"is":[7,145],"essential":[8],"for":[9,165],"reconstructing":[10],"photons":[11,196],"and":[12,22,28,42,62,73,83,122,150,155,187,197,252,259,274],"electrons":[13],"in":[14,55,85,279],"modern":[15],"hadron":[16],"collider":[17],"experiments,":[18],"where":[19,232],"boosted":[20,204],"topologies":[21],"pileup":[23],"often":[24],"cause":[25],"overlapping":[26,179],"showers":[27,180],"ambiguous":[29],"assignment.":[31],"We":[32],"present":[33],"deep":[34],"learning-based":[35],"approaches":[37,258],"that":[38,96],"reconstruct":[39,162],"particle":[40],"impact":[43],"position":[44,151],"directly":[45],"from":[46,103],"calorimeter":[47,121],"readout.":[48],"The":[49],"study":[50],"includes":[51],"a":[52,74,91,118,166],"two-step":[53,257],"strategy":[54],"which":[56,79],"candidate":[57,81],"seed":[58],"windows":[59],"are":[60,112],"identified":[61],"then":[63],"jointly":[64],"processed":[65],"via":[66],"distance-weighted":[67],"message":[68,189],"passing":[69],"or":[70],"attention":[71],"mechanism,":[72],"single-step":[75],"graph":[76,102],"transformer,":[77],"ClusTEX,":[78],"performs":[80],"selection":[82],"reconstruction":[84,177,230],"one":[86],"inference":[87],"stage.":[88],"ClusTEX":[89,242,263],"uses":[90],"novel":[92],"positional":[93],"encoding":[94],"scheme":[95],"separates":[97],"local":[98],"coordinates":[99],"within":[100],"the":[101,159,170,176,184,224,233,239,244,260],"global":[104],"detector":[105,269],"coordinates,":[106],"enabling":[107],"efficient,":[108],"geometry-aware":[109],"inference.":[110,202],"Models":[111],"trained":[113],"on":[114,194],"Geant4":[115],"simulations":[116],"simplified":[119],"toy":[120,171],"an":[123,127],"ECAL-inspired":[124,240],"topology":[125],"with":[126],"explicit":[128],"$$(\\\\eta":[129],",":[130],"\\\\phi":[131],")$$":[132],"<mml:math":[133,211],"xmlns:mml=\\"http://www.w3.org/1998/Math/MathML\\">":[134,212],"<mml:mrow>":[135,213],"<mml:mo>(</mml:mo>":[136],"<mml:mi>η</mml:mi>":[137],"<mml:mo>,</mml:mo>":[138],"<mml:mi>ϕ</mml:mi>":[139],"<mml:mo>)</mml:mo>":[140],"</mml:mrow>":[141,221],"</mml:math>":[142,222],"dependence.":[143],"Performance":[144],"evaluated":[146],"using":[147],"efficiency,":[148],"resolutions,":[152],"background":[153],"rejection,":[154],"“splitting":[156],"rate”":[157],"–":[158],"probability":[160],"to":[161,182,256],"two":[163],"objects":[164],"single-photon":[167],"shower.":[168],"In":[169,203,238],"calorimeter,":[172],"attention-based":[173,225],"interactions":[174],"improve":[175],"relative":[181],"both":[183],"standard":[185,234,261],"algorithm":[186,235],"distance-driven":[188],"passing,":[190],"while":[191],"maintaining":[192],"performance":[193],"isolated":[195],"reducing":[198],"splitting":[199,254],"without":[200],"multi-pass":[201],"$$\\\\pi":[205],"^0":[206],"\\\\rightarrow":[207],"\\\\gamma":[208,209],"$$":[210],"<mml:msup>":[214],"<mml:mi>π</mml:mi>":[215],"<mml:mn>0</mml:mn>":[216],"</mml:msup>":[217],"<mml:mo>→</mml:mo>":[218],"<mml:mi>γ</mml:mi>":[219,220],"events,":[223],"model":[226],"retains":[227],"di-photon":[228],"mass":[229],"capability,":[231],"becomes":[236],"inefficient.":[237],"topology,":[241],"provides":[243],"best":[245],"overall":[246],"performance,":[247],"yielding":[248],"improved":[249,272],"resolution":[251],"reduced":[253],"compared":[255],"algorithm.":[262],"also":[264],"remains":[265],"robust":[266],"under":[267],"localized":[268],"failures,":[270],"showing":[271],"stability":[273],"partial":[275],"recovery":[276],"non-responsive":[280],"channels.":[281]}

Authors

Institutions

Publication Details

Journal
The European Physical Journal C
Published
2026-07-26
DOI
https://doi.org/10.1140/epjc/s10052-026-16097-x
Primary Topic
Particle physics theoretical and experimental studies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
article

Reconstruction of overlapping electromagnetic showers in calorimeters using Transformers

F. Couderc, J. Rander, J. Malclès, Y. Maidannyk
The European Physical Journal C
Particle physics theoretical and experimental studies
article

Reconstruction of overlapping electromagnetic showers in calorimeters using Transformers

F. Couderc, J. Rander, J. Malclès, Y. Maidannyk
article en

Abstract

Abstract Accurate clustering of electromagnetic energy deposits is essential for reconstructing photons and electrons in modern hadron collider experiments, where boosted topologies and pileup often cause overlapping showers and ambiguous energy assignment. We present deep learning-based clustering approaches that reconstruct particle energy and impact position directly from calorimeter readout. The study includes a two-step strategy in which candidate seed windows are identified and then jointly processed via distance-weighted message passing or attention mechanism, and a single-step graph transformer, ClusTEX, which performs candidate selection and reconstruction in one inference stage. ClusTEX uses a novel positional encoding scheme that separates local coordinates within the graph from global detector coordinates, enabling efficient, geometry-aware inference. Models are trained on Geant4 simulations of a simplified toy calorimeter and an ECAL-inspired topology with an explicit $$(\eta , \phi )$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mo>(</mml:mo> <mml:mi>η</mml:mi> <mml:mo>,</mml:mo> <mml:mi>ϕ</mml:mi> <mml:mo>)</mml:mo> </mml:mrow> </mml:math> dependence. Performance is evaluated using efficiency, energy and position resolutions, background rejection, and “splitting rate” – the probability to reconstruct two objects for a single-photon shower. In the toy calorimeter, attention-based interactions improve the reconstruction of overlapping showers relative to both the standard algorithm and distance-driven message passing, while maintaining performance on isolated photons and reducing splitting without multi-pass inference. In boosted $$\pi ^0 \rightarrow \gamma \gamma $$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:msup> <mml:mi>π</mml:mi> <mml:mn>0</mml:mn> </mml:msup> <mml:mo>→</mml:mo> <mml:mi>γ</mml:mi> <mml:mi>γ</mml:mi> </mml:mrow> </mml:math> events, the attention-based model retains di-photon mass reconstruction capability, where the standard algorithm becomes inefficient. In the ECAL-inspired topology, ClusTEX provides the best overall performance, yielding improved energy resolution and reduced splitting compared to two-step approaches and the standard algorithm. ClusTEX also remains robust under localized detector failures, showing improved stability and partial recovery of energy in non-responsive channels.

The European Physical Journal CVol. 86(7)
Commissariat à l'Énergie Atomique et aux Énergies Alternatives (FR), Université Paris-Saclay (FR), CEA Paris-Saclay (FR)
Affordable and clean energy
Openalex Percentile: Top 11%
Particle physics theoretical and experimental studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.