Christian W. Bauer. Zohreh Davoudi, A. Balantekin, Tanmoy Bhattacharya
et al.
It is for the first time that Quantum Simulation for High Energy Physics (HEP) is studied in the U.S. decadal particle-physics community planning, and in fact until recently, this was not considered a mainstream topic in the community. This fact speaks of a remarkable rate of growth of this subfield over the past few years, stimulated by the impressive advancements in Quantum Information Sciences (QIS) and associated technologies over the past decade, and the significant investment in this area by the government and private sectors in the U.S. and other countries. High-energy physicists have quickly identified problems of importance to our understanding of nature at the most fundamental level, from tiniest distances to cosmological extents, that are intractable with classical computers but may benefit from quantum advantage. They have initiated, and continue to carry out, a vigorous program in theory, algorithm, and hardware co-design for simulations of relevance to the HEP mission. This community whitepaper is an attempt to bring this exciting and yet challenging area of research to the spotlight, and to elaborate on what the promises, requirements, challenges, and potential solutions are over the next decade and beyond.
Searches for new physics at the LHC at CERN traditionally use advanced simulations to model Standard Model and new-physics processes in high-energy collisions and compare them with data. The lack of recent direct discoveries, however, has motivated the development of model-independent approaches in HEP to complement existing hypothesis-driven analyses, particularly Anomaly Detection. A review of the latest efforts in BSM searches with anomaly detection is presented in these proceedings, focusing on contributions within the ATLAS collaboration at LHC and discussing Variational Recurrent Neural Network, Deep Transformer and Graph Anomaly Detection applications.
Zichun Hao, Raghav Kansal, Abhijith Gandrakota
et al.
A common challenge with supervised machine learning (ML) in high energy physics (HEP) is the reliance on simulations for labeled data, which can often mismodel the underlying collision or detector response. To help mitigate this problem of domain shift, we propose RINO (Renormalization Group Invariance with No Labels), a self-supervised learning approach that can instead pretrain models directly on collision data, learning embeddings invariant to renormalization group flow scales. In this work, we pretrain a transformer-based model on jets originating from quantum chromodynamic (QCD) interactions from the JetClass dataset, emulating real QCD-dominated experimental data, and then finetune on the JetNet dataset -- emulating simulations -- for the task of identifying jets originating from top quark decays. RINO demonstrates improved generalization from the JetNet training data to JetClass data compared to supervised training on JetNet from scratch, demonstrating the potential for RINO pretraining on real collision data followed by fine-tuning on small, high-quality MC datasets, to improve the robustness of ML models in HEP.
The field of high energy physics (HEP) has seen a marked increase in the use of machine learning (ML) techniques in recent years. The proliferation of applications has revolutionised many aspects of the data processing pipeline at collider experiments including the Large Hadron Collider (LHC). In this whitepaper, we discuss the increasingly crucial role that ML plays in real-time analysis (RTA) at the LHC, namely in the context of the unique challenges posed by the trigger systems of the large LHC experiments. We describe a small selection of the ML applications in use at the large LHC experiments to demonstrate the breadth of use-cases. We continue by emphasising the importance of collaboration and engagement between the HEP community and industry, highlighting commonalities and synergies between the two. The mutual benefits are showcased in several interdisciplinary examples of RTA from industrial contexts. This whitepaper, compiled by the SMARTHEP network, does not provide an exhaustive review of ML at the LHC but rather offers a high-level overview of specific real-time use cases.
Designing the next generation colliders and detectors involves solving optimization problems in high-dimensional spaces where the optimal solutions may nest in regions that even a team of expert humans would not explore. Resorting to Artificial Intelligence to assist the experimental design introduces however significant computational challenges in terms of generation and processing of the data required to perform such optimizations: from the software point of view, differentiable programming makes the exploration of such spaces with gradient descent feasible; from the hardware point of view, the complexity of the resulting models and their optimization is prohibitive. To scale up to the complexity of the typical HEP collider experiment, a change in paradigma is required. In this contribution I will describe the first proofs-of-concept of gradient-based optimization of experimental design and implementations in neuromorphic hardware architectures, paving the way to more complex challenges.
Exciton polaritons are quasiparticles of photons coupled strongly to bound electron-hole pairs, manifesting as an anti-crossing light dispersion near an exciton resonance. Highly anisotropic semiconductors with opposite-signed permittivities along different crystal axes are predicted to host exotic modes inside the anti-crossing called hyperbolic exciton polaritons (HEPs), which confine light subdiffractionally with enhanced density of states. Here, we show observational evidence of steady-state HEPs in the van der Waals magnet chromium sulfide bromide (CrSBr) using a cryogenic near-infrared near-field microscope. At low temperatures, in the magnetically-ordered state, anisotropic exciton resonances sharpen, driving the permittivity negative along one crystal axis and enabling HEP propagation. We characterize HEP momentum and losses in CrSBr, also demonstrating coupling to excitonic sidebands and enhancement by magnetic order: which boosts exciton spectral weight via wavefunction delocalization. Our findings open new pathways to nanoscale manipulation of excitons and light, including routes to magnetic, nonlocal, and quantum polaritonics. Hyperbolic exciton polaritons (HEPs) are anisotropic light-matter excitations with promising applications, but their steady-state observation is challenging. Here, the authors report experimental evidence of HEPs in a van der Waals magnet, CrSBr, via cryogenic infrared near-field microscopy.
Short-lived hadronic resonances are unique tools for studying the hadron-gas phase that is created in the late stages of relativistic heavy-ion collisions. Measurements of the yield ratios between resonances and the corresponding stable particles are sensitive to the competing rescattering and regeneration effects. These measurements in small collision systems, such as pp and p-Pb, are a powerful method to reveal a possible short-lived hadronic phase. In addition, resonance production in small systems is interesting to study the onset of strangeness enhancement, collective effects,and the hadron production mechanism. On this front, the $φ$ meson is particularly relevant since its yield is sensitive to different production models: no effect is expected by strange number canonical suppression but its production is expected to be enhanced in the rope-hadronization scenario. In this presentation, recent measurements of hadronic resonances in different collision systems,going from pp to Pb-Pb collisions, are presented. These include transverse momentum spectra,yields, and yield ratios as a function of multiplicity. The presented results are discussed in the context of state-of-the-art phenomenological models of hadron production. The resonance yields measured in Pb-Pb collisions are used as an experimental input in a partial chemical equilibrium-based thermal model to constrain the kinetic freeze-out temperature. This is a novel procedure that is independent of assumptions on the flow velocity profile and the freeze-out hypersurface.
We revisit a method for determining the neutrino mass ordering by using precision measurements of the atmospheric $Δm^2$'s in both electron neutrino and muon neutrino disappearance channels, proposed by the authors in 2005 (hep-ph/0503283). The mass ordering is a very important outstanding question for our understanding of the elusive neutrino and determination of the mass ordering has consequences for other neutrino experiments. The JUNO reactor experiment will start data taking this year, and the precision of the atmospheric $Δm^2$'s from electron anti-neutrino measurements will improve by a factor of three from Daya Bay's 2.4 % to 0.8 % within a year. This measurement, when combined with the atmospheric $Δm^2$'s measurements from T2K and NOvA for muon neutrino disappearance, will contribute substantially to the $Δχ^2$ between the two remaining neutrino mass orderings. In this paper we derive a mass ordering sum rule that can be used to address the possibility that JUNO's atmospheric $Δm^2$'s measurement, when combined with other experiments in particular T2K and NOvA, can determine the neutrino mass ordering at the 3 $σ$ confidence level within one year of operation. For a confidence level of 5 $σ$ in a single experiment we will have to wait until the middle of the next decade when the DUNE experiment is operating.
Track reconstruction is a vital aspect of High-Energy Physics (HEP) and plays a critical role in major experiments. In this study, we delve into unexplored avenues for particle track reconstruction and hit clustering. Firstly, we enhance the algorithmic design effort by utilising a simplified simulator (REDVID) to generate training data that is specifically composed for simplicity. We demonstrate the effectiveness of this data in guiding the development of optimal network architectures. Additionally, we investigate the application of image segmentation networks for this task, exploring their potential for accurate track reconstruction. Moreover, we approach the task from a different perspective by treating it as a hit sequence to track sequence translation problem. Specifically, we explore the utilisation of Transformer architectures for tracking purposes. Our preliminary findings are covered in detail. By considering this novel approach, we aim to uncover new insights and potential advancements in track reconstruction. This research sheds light on previously unexplored methods and provides valuable insights for the field of particle track reconstruction and hit clustering in HEP.
Claire Savard, Nicholas Manganelli, Burt Holzman
et al.
With machine learning applications now spanning a variety of computational tasks, multi-user shared computing facilities are devoting a rapidly increasing proportion of their resources to such algorithms. Graph neural networks (GNNs), for example, have provided astounding improvements in extracting complex signatures from data and are now widely used in a variety of applications, such as particle jet classification in high energy physics (HEP). However, GNNs also come with an enormous computational penalty that requires the use of GPUs to maintain reasonable throughput. At shared computing facilities, such as those used by physicists at Fermi National Accelerator Laboratory (Fermilab), methodical resource allocation and high throughput at the many-user scale are key to ensuring that resources are being used as efficiently as possible. These facilities, however, primarily provide CPU-only nodes, which proves detrimental to time-to-insight and computational throughput for workflows that include machine learning inference. In this work, we describe how a shared computing facility can use the NVIDIA Triton Inference Server to optimize its resource allocation and computing structure, recovering high throughput while scaling out to multiple users by massively parallelizing their machine learning inference. To demonstrate the effectiveness of this system in a realistic multi-user environment, we use the Fermilab Elastic Analysis Facility augmented with the Triton Inference Server to provide scalable and high throughput access to a HEP-specific GNN and report on the outcome.