ACTS Workshop 2026 @NIKHEF

Apr 20, 2026, 10:00 AM → Apr 24, 2026, 9:00 PM Europe/Zurich

NIKHEF

The 2026 Edition of the ACTS Workshop! will take place at NIKHEF from April 20th to April 23rd, followed by a hackathon-style April 24th.

As in past years, we will have a mix of topical presentations, presentations from experiments with plenty of time for discussion and time for hands-on collaborative work.

The entrance close to our room is here: Google Maps

Mattermost: https://mattermost.web.cern.ch/acts/channels/acts-ws-2026

 

 

Group photo 1

Group photo 2

Mon, April 20

10:00 AM → 10:30 AM Coffee

10:30 AM → 12:00 PM Presentations: Introduction

10:30 AM Introduction

Speaker: Pierfrancesco Butti (Nikhef National institute for subatomic physics (NL))

ACTS Workshop Introduction.pdf

10:45 AM Project Overview and Status

Speaker: Andreas Salzburger (CERN)

2026-04-20-ACTS-Project.pdf

11:05 AM ColliderML dataset overview

Speaker: Doğa Elitez (CERN)

colliderml_overview.pdf

11:15 AM pip install (py)acts

Presentation of the vision and status of the pip install acts project that providing a simplified way to install the ACTS software within the python ecosystem.

Speaker: Benjamin Huth (CERN)

pip_install_pyacts_acts_workshop_animations.pdf

pip_install_pyacts_acts_workshop.pdf

12:00 PM → 1:30 PM Box Lunch

1:30 PM → 3:30 PM Hands-on

3:30 PM → 4:00 PM Coffee

Tue, April 21

9:00 AM → 10:30 AM Experiments

9:00 AM SHiP

Speakers: James Maitland Webb (University of Freiburg (DE)), Oliver Lantwin (Universitaet Siegen (DE))

JamesWebb_ACTSWithSHIP.pdf

9:25 AM Experiment 320 experience with Acts planar geometry tracking

The SLAC Experiment 320 collides 10 TW laser pulses with a high-quality 10 GeV electron beam from the FACET-II RF LINAC. This configuration is expected to produce e⁺e⁻ pairs via the nonlinear Breit–Wheeler process in the strong-field tunneling regime, with an estimated yield of 0.01–0.1 pairs per collision. Such a small signal rate is accompanied by substantial backgrounds originating, for example, from dumping the high-charge primary beam, secondaries induced by the beam halo, and photons and low-energy electrons produced in the electron–laser interaction itself. These backgrounds can reach densities of O(100) charged particles per cm² at the surface of the pair detector, along with even larger neutral particle fluxes, posing a significant challenge for unambiguous single-particle detection.

Recent simulation studies have shown that high-precision tracking detectors based on radiation-hard, multi-layer pixel sensors, combined with the track reconstruction algorithms provided by the Acts framework, are strong candidates for the primary particle detection technology under such conditions. These studies indicate an energy resolution better than 2% and background suppression approaching two orders of magnitude relative to the signal reconstruction rate. A tracking detector system based on ALPIDE pixel chips was commissioned at Experiment 320 in August 2024. Analyses of recent data-taking campaigns demonstrate the promising performance of tracking detectors in strong-field QED experiments, achieving background suppression of multiple orders of magnitude relative to signal and a vertex reconstruction resolution of O(1 mm) at a distance of 17 m from the detector.

In this talk, we discuss how Acts enabled these results. Particular focus is placed on the use of Acts in telescope-style tracking detectors -- the geometry employed in this experiment. We present the track reconstruction pipeline, with emphasis on the pattern recognition, as well as the local and global alignment algorithms, which proved to be the most challenging aspects of the reconstruction. Finally, we describe the adaptations and extensions to the Acts codebase implemented to tailor its performance to the experiment’s needs.

Speaker: Roman Urmanov (Weizmann Institute of Science (IL))

e320_planar_tracking.pdf

9:45 AM ACTS at the proposed Lohengrin experiment in Bonn

The proposed Lohengrin experiment will search for light dark matter at the Electron Stretcher Accelerator (ELSA) in Bonn. It employs the fixed-target missing momentum based technique for searching for dark-sector particles. A beam of electrons is extracted from ELSA and is shot onto a thin target to produce mainly Standard Model bremsstrahlung and – in rare occasions – possibly new particles, like the dark photon, coupling feebly to the electron. A crucial part of the experiment are the two tracking detectors for tagging the electrons in front of the target and precisely reconstructing the momentum of the recoil electron after the target. ACTS was used for the track reconstruction in the initial feasibility study for the experiment. We currently implement the full detector simulation in the Key4hep/DD4hep framework, including ACTS for the track reconstruction.

We present how ACTS was used for the studies in the Lohengrin Whitepaper and how we plan to integrate ACTS in the full detector simulation for current studies and in the future.

Speaker: Cedric Carl Breuning (Universität Bonn)

Lohengrin-ACTS-2026.pdf

10:05 AM CMS

Speaker: Lorenzo Damenti (Universita e INFN, Firenze (IT))

CMSTrackingWithACTS.pdf

10:30 AM → 10:50 AM Coffee

10:50 AM → 12:00 PM Presentations

10:50 AM traccc integration

Speaker: Neza Ribaric (Duke University (US))

TracccIntegration_ACSTworkshop_21042026.pdf

11:10 AM Integration of new Graph Based Seeding Algorithm in ACTS

Over the last few years, ACTS has made substantial progress toward a detector-independent software toolkit for charged-particle tracking. Looking ahead to future experiments (most notably the Phase-II upgrade of the ATLAS detector) the much higher hit occupancies and track densities will demand seeding algorithms that are both more efficient, more selective and faster. This contribution presents ongoing work to integrate a Graph Based Track Seeding (GBTS) algorithm into ACTS. By exploiting a graph representation of hit triplets (or higher-order combinations) and their mutual compatibility, GBTS produces significantly purer seed candidates. It also reduces the downstream workload of track finding by limiting the number of duplicates and fake seeds that must be rejected. The algorithm itself also runs much faster than current methods. We present performance results from the ACTS standalone simulation chain, together with its implementation within the ATLAS - Athena framework.

Speaker: Jasper Preston (University of London (GB))

ACTS workshop 2026.pdf

ACTS workshop 2026.pptx

11:30 AM Vertex fitting in the variable magnetic field

Speaker: Dr Vadim Kostyukhin (Universitaet Siegen (DE))

ACTS26_MagFld.pdf

12:00 PM → 1:30 PM Lunch

1:30 PM → 3:30 PM Hands-on

3:30 PM → 4:00 PM Coffee

4:00 PM → 5:30 PM Hands-on

4:00 PM ACTS Governance Discussion

Speaker: Andreas Salzburger (CERN)

Wed, April 22

9:00 AM → 10:30 AM Presentations: Alignment

Convener: Andreas Salzburger (CERN)

9:00 AM ACTS ❤️ MILLEPEDE

Speaker: Maximilian Emanuel Goblirsch-Kolb (Deutsches Elektronen-Synchrotron (DE))

Millepede_ACTS_ACTSWS.pdf

9:30 AM Global alignment with tracks fitted with a Kalman filter

Speaker: Wouter Hulsbergen (Nikhef National institute for subatomic physics (NL))

lhcbalignment.pdf

9:50 AM Alignable (super) structures

Speakers: Andreas Salzburger (CERN), Knut Zoch (CERN)

2026-04-22-ACTS-WS-alignment-discussion.pdf

10:30 AM → 10:50 AM Coffee

10:50 AM → 12:00 PM Hands-on

12:00 PM → 1:30 PM Lunch

1:30 PM → 3:30 PM Hands-on

3:30 PM → 4:00 PM Coffee

4:00 PM → 6:00 PM Experiments: Afternoon 2

4:00 PM sPHENIX

Speaker: Joe Osborn (Brookhaven National Laboratory)

sPHENIX_ActsWorkshop_2026.pdf

4:25 PM EPIC

Speakers: Barak Schmookler, Wouter Deconinck (University of Manitoba (CA))

ePICActs2026.pdf

4:50 PM LDMX

Speaker: Mathew Graham (SLAC)

ldmx-acts-ws-april-22-2026.pdf

5:15 PM Muon Collider

Speaker: Samuel Ferraro

Samuel_Ferraro_ACTS_Nikhef_26.pptx.pdf

5:40 PM ATLAS

Speaker: Noemi Calace (CERN)

ACTS_Dev_WS_ATLAS-1.pdf

7:30 PM → 9:30 PM Social dinner

https://maps.app.goo.gl/4U9By4hXCvAxdBQ87

Thu, April 23

9:00 AM → 10:15 AM Experiments: Morning

9:00 AM STCF

Speakers: Hao Li (Zhengzhou University (CN)), Xiaocong Ai (Zhengzhou University)

STCF-2026ACTSWorkshop.pdf

9:25 AM LUXE

Speakers: Thomas Madlener (Deutsches Elektronen-Synchrotron (DESY)), Yufeng Wang

LUXE_acts_nikhef_260423.pdf

LUXE_acts_nikhef_260423.pptx

9:50 AM ALICE

Speakers: Igor Altsybeev (Technische Universitaet Muenchen (DE)), Pierfrancesco Butti (Nikhef National institute for subatomic physics (NL))

ACTS_ALICE_2026.pdf

10:15 AM → 10:35 AM Coffee

10:35 AM → 12:00 PM Presentations: Machine Learning

10:35 AM Transformer tracking

Speaker: Sebastien Rettie (Nikhef National institute for subatomic physics (NL))

2026.04.23_SebastienRettie_ACTSWorkshop_v1.pdf

11:05 AM Hybrid pattern recognition for charged particle tracking: Hough transform and convolutional neural networks

Charged particle track reconstruction is among the most computationally expensive tasks in event reconstruction in high-energy physics experiments. A major contribution to this cost arises from the track seeding stage, which provides the initial candidates for subsequent track fitting algorithms. Track seeding is particularly sensitive to combinatorial effects, and in the high-occupancy environment expected at the High-Luminosity Large Hadron Collider, the large number of simultaneously produced particles leads to a rapid growth in fake seed candidates, significantly degrading computational performance.

Several alternative seeding approaches have been investigated to address this challenge, including the Hough transform, which features favourable scaling with detector occupancy. However, when applied to track reconstruction in central detectors at high pile-up, the Hough transform typically produces a large number of false positives.

In this work, we propose a hybrid approach in which the output of the Hough transform is post-processed using a compact machine learning model trained to efficiently reject fake track candidates while retaining high efficiency for genuine seeds. The neural network operates directly on image-like representations of the Hough accumulator, avoiding costly data transformations. This design naturally supports efficient implementations on heterogeneous computing architectures, in particular GPU-based systems, making the approach well suited for high-throughput and real-time tracking applications.

Speaker: Carlo Varni (AGH University of Krakow (PL))

HTMLPresentation.pdf

11:35 AM GUNTAM: A Transformer for seed reconstruction

Reconstructing particle trajectories is a significant challenge in most particle physics experiments and a major consumer of CPU resources. In classical tracking, most of the intensive computational workload stems from the seed identification process, while track finding is well understood.
To address the complexity of seed reconstruction, we introduce GUNTAM, a transformer-based algorithm for seed reconstruction. This algorithm takes as input all the 3D space points in different phi bins through the detector and returns, for each, a list of possible seeds. To achieve this, we first perform an innovative embedding of the 3D space points using Fourier series. The embedded space points are then binned, with each bin being fed separately to the transformer. At the end of each transformer chain, we extract the attention matrix of the last transformer layer. By using an adequate loss function, this matrix can be made to correspond to an adjacency matrix between the 3D space points. It is then possible to reconstruct seeds from that matrix. Once those seeds have been reconstructed, the rest of the reconstruction can be performed on either CPU or GPU using a regular tracking chain.
These algorithms have been first presented at the CTD 2026 conferences (https://indico.cern.ch/event/1499357/contributions/6628608/), much progress has been achieved since then, reaching a decent seeding efficiency, but more discussion on this approach with the rest of the ACTS community would be interesting.

Speaker: Corentin Allaire (IJCLab, Université Paris-Saclay, CNRS/IN2P3)

GUNTAM_ACTS.pdf

12:00 PM → 1:30 PM Lunch

1:30 PM → 3:30 PM Hands-on: Hands-on time

3:30 PM → 4:00 PM Coffee

4:00 PM → 5:20 PM Presentations: Presentations: Displaced Tracking

Convener: Benjamin Henry Hooberman (Univ. Illinois at Urbana Champaign (US))

4:00 PM Large Impact Parameter Track Reconstruction in ATLAS Using ACTS (25+10)

Searches for long-lived particles in ATLAS require efficient reconstruction of charged-particle tracks with large transverse impact parameters, which generally fall outside the acceptance of standard tracking configurations. To address this, ATLAS employs a secondary tracking pass known as Large-Radius Tracking (LRT), which reconstructs tracks using hits not assigned in the primary tracking pass and applies significantly loosened impact-parameter requirements. As ATLAS transitions to a fully Acts-based tracking chain for the HL-LHC, the LRT pass must be reimplemented and reoptimized to satisfy both physics performance and computational requirements.

This talk presents the current status of large-radius tracking in the Acts-based ATLAS track reconstruction. I will discuss the main algorithmic challenges associated with reconstructing displaced tracks and the strategies adopted to address them. I will then present tracking efficiency results for benchmark long-lived particle scenarios and compare them to the performance of the legacy, non-Acts-based LRT pass. I will conclude with a discussion of open questions and areas for future improvement.

Speaker: Jackson Carl Burzynski (University of Oklahoma (US))

ActsWorkshopLRT.pdf

4:35 PM Combined primary and secondary vertexing at future colliders (25+10)

Speakers: Dr Vadim Kostyukhin, Dr Vadim Kostyukhin (Universitaet Siegen (DE)), Vadim Kostyukhin (Bonn)

ACTS26_FuturePVSV.pdf

5:10 PM Close-out

Speaker: Andreas Salzburger (CERN)

2026-04-23-Close-Out.pdf

Fri, April 24

9:00 AM → 12:00 PM Hackathon

12:00 PM → 1:30 PM Lunch

1:30 PM → 4:00 PM Hackathon