Y. Iiyama

This document is the final report of the ATLAS-CMS Dark Matter Forum, a forum organized by the ATLAS and CMS collaborations with the participation of experts on theories of Dark Matter, to select a minimal basis set of dark matter simplified models that should support the design of the early LHC Run-2 searches. A prioritized, compact set of benchmark models is proposed, accompanied by studies of the parameter space of these models and a repository of generator implementations. This report also addresses how to apply the Effective Field Theory formalism for collider searches and present the results of such interpretations.

We explore the use of graph networks to deal with irregular-geometry detectors in the context of particle reconstruction. Thanks to their representation-learning capabilities, graph networks can exploit the full detector granularity, while natively managing the event sparsity and arbitrarily complex detector geometries. We introduce two distance-weighted graph network architectures, dubbed GarNet and GravNet layers, and apply them to a typical particle reconstruction task. The performance of the new architectures is evaluated on a data set of simulated particle interactions on a toy model of a highly granular calorimeter, loosely inspired by the endcap calorimeter to be installed in the CMS detector for the High-Luminosity LHC phase. We study the clustering of energy depositions, which is the basis for calorimetric particle reconstruction, and provide a quantitative comparison to alternative approaches. The proposed algorithms provide an interesting alternative to existing methods, offering equally performing or less resource-demanding solutions with less underlying assumptions on the detector geometry and, consequently, the possibility to generalize to other detectors.

Abstract We introduce an automated tool for deploying ultra low-latency, low-power deep neural networks with convolutional layers on field-programmable gate arrays (FPGAs). By extending the hls4ml library, we demonstrate an inference latency of 5 µ s using convolutional architectures, targeting microsecond latency applications like those at the CERN Large Hadron Collider. Considering benchmark models trained on the Street View House Numbers Dataset, we demonstrate various methods for model compression in order to fit the computational constraints of a typical FPGA device used in trigger and data acquisition systems of particle detectors. In particular, we discuss pruning and quantization-aware training, and demonstrate how resource utilization can be significantly reduced with little to no loss in model accuracy. We show that the FPGA critical resource consumption can be reduced by 97% with zero loss in model accuracy, and by 99% when tolerating a 6% accuracy degradation.

Graph neural networks have been shown to achieve excellent performance for several crucial tasks in particle physics, such as charged particle tracking, jet tagging, and clustering. An important domain for the application of these networks is the FGPA-based first layer of real-time data filtering at the CERN Large Hadron Collider, which has strict latency and resource constraints. We discuss how to design distance-weighted graph networks that can be executed with a latency of less than 1$\mu\mathrm{s}$ on an FPGA. To do so, we consider a representative task associated to particle reconstruction and identification in a next-generation calorimeter operating at a particle collider. We use a graph network architecture developed for such purposes, and apply additional simplifications to match the computing constraints of Level-1 trigger systems, including weight quantization. Using the $\mathtt{hls4ml}$ library, we convert the compressed models into firmware to be implemented on an FPGA. Performance of the synthesized models is presented both in terms of inference accuracy and resource usage.

There is no unique way to encode a quantum algorithm into a quantum circuit. With limited qubit counts, connectivities, and coherence times, circuit optimization is essential to make the best use of quantum devices produced over a next decade. We introduce two separate ideas for circuit optimization and combine them in a multi-tiered quantum circuit optimization protocol called AQCEL. The first ingredient is a technique to recognize repeated patterns of quantum gates, opening up the possibility of future hardware optimization. The second ingredient is an approach to reduce circuit complexity by identifying zero- or low-amplitude computational basis states and redundant gates. As a demonstration, AQCEL is deployed on an iterative and effcient quantum algorithm designed to model final state radiation in high energy physics. For this algorithm, our optimization scheme brings a significant reduction in the gate count without losing any accuracy compared to the original circuit. Additionally, we have investigated whether this can be demonstrated on a quantum computer using polynomial resources. Our technique is generic and can be useful for a wide variety of quantum algorithms.

We present a first proof of concept to directly use neural network based pattern recognition to trigger on distinct calorimeter signatures from displaced particles, such as those that arise from the decays of exotic long-lived particles. The study is performed for a high granularity forward calorimeter similar to the planned high granularity calorimeter for the high luminosity upgrade of the CMS detector at the CERN Large Hadron Collider. Without assuming a particular model that predicts long-lived particles, we show that a simple convolutional neural network, that could in principle be deployed on dedicated fast hardware, can efficiently identify showers from displaced particles down to low energies while providing a low trigger rate.

There is no unique way to encode a quantum algorithm into a quantum circuit. With limited qubit counts, connectivity, and coherence times, a quantum circuit optimization is essential to make the best use of near-term quantum devices. We introduce a new circuit optimizer called AQCEL, which aims to remove redundant controlled operations from controlled gates, depending on initial states of the circuit. Especially, the AQCEL can remove unnecessary qubit controls from multi-controlled gates in polynomial computational resources, even when all the relevant qubits are entangled, by identifying zero-amplitude computational basis states using a quantum computer. As a benchmark, the AQCEL is deployed on a quantum algorithm designed to model final state radiation in high energy physics. For this benchmark, we have demonstrated that the AQCEL-optimized circuit can produce equivalent final states with much smaller number of gates. Moreover, when deploying AQCEL with a noisy intermediate scale quantum computer, it efficiently produces a quantum circuit that approximates the original circuit with high fidelity by truncating low-amplitude computational basis states below certain thresholds. Our technique is useful for a wide variety of quantum algorithms, opening up new possibilities to further simplify quantum circuits to be more effective for real devices.

A new type of superconducting power transmission cable is proposed to fulfill both the conditions of low AC loss and rare leakage magnetic flux. A lot of superconducting tapes are arranged radially, and three-phase currents are applied one by another in the azimuthal direction. If the number of tapes is a multiple of three, the leakage flux can become very small. The AC loss in the superconducting tapes can also be suppressed because of the reduction of a magnetic field applied perpendicular to the tape face with increasing the number of tapes. The numerical estimation of AC losses in the proposed power cables is carried out by means of a finite element method as well as those in the conventional types of cables.

カメラ2台とプロジェクターを用いた三眼視による面的水面形状計測法を開発した．多数のカラードットを有するカラーパターン画像をプロジェクターから照射する．水を白濁させておき，各カラードットの水面近傍における散乱光をカメラ2台で撮影する．予めキャリブレーションにより取得するプロジェクターおよびカメラ座標系のパラメータと各画像座標から水面の3次元座標を計測する．本研究では座標の計算に必要となるカラードットのマッチングについて，色差ベクトルを用いた自動マッチング手法を提案した．また，水面近傍における散乱光を撮影することに伴い，撮影位置に生じるずれを補正する方法を検討した．小型平面水槽による計測実験により，本研究で検討した手法により概ね安定して良好な精度で面的に波浪形状を取得可能であることを示した．

Current distribution in Bi-2223 multifilamentary tapes with inter-filament barrier is numerically evaluated for the application of an alternating transport current. In the case of barrier materials with an infinite resistivity, the current distribution can be determined with a lumped parameter circuit model including the inductances and resistances of all the filaments. The mutual inductance between a pair of filaments is obtained by estimating their geometric mean distance. The power-law model is also assumed to take into account the effect of resistance in each filament. The influences of the silver ratio and filament number in the tape wire on the current distribution are discussed systematically as well as the power index and frequency.

Abstract Dynamo is a full-stack software solution for scientific data management. Dynamo’s architecture is modular, extensible, and customizable, making the software suitable for managing data in a wide range of installation scales, from a few terabytes stored at a single location to hundreds of petabytes distributed across a worldwide computing grid. This article documents the core system design of Dynamo and describes the applications that implement various data management tasks. A brief report is also given on the operational experiences of the system at the CMS experiment at the CERN Large Hadron Collider and at a small-scale analysis facility.

Quantum error correction is a crucial step beyond the current noisy-intermediate-scale quantum device towards fault-tolerant quantum computing. However, most of the error corrections ever demonstrated rely on post-selection of events or post-correction of states, based on measurement results repeatedly recorded during circuit execution. On the other hand, real-time error correction is supposed to be performed through classical feedforward of the measurement results to data qubits. It provides unavoidable latency from conditional electronics that would limit the scalability of the next-generation quantum processors. Here we propose a new approach to real-time error correction that is free from measurement and realized by using multi-controlled gates based on higher-dimensional state space. Specifically, we provide a series of novel decompositions of a Toffoli gate by using the lowest three energy levels of a transmon that significantly reduce the number of two-qubit gates and discuss their essential features, such as extendability to an arbitrary number of control qubits, the necessity of exclusively controlled NOT gates, and usefulness of their incomplete variants. Combined with the recently demonstrated schemes of fast two-qubit gates and all-microwave qubit reset, it would substantially shorten the time required for error correction and resetting ancilla qubits compared to a measurement-based approach and provide an error correction rate of $\gtrsim1$~MHz with high accuracy for three-qubit bit- and phase-flip errors.

This Ph.D. thesis is a search for physics beyond the standard model (SM) of particle physics, which successfully describes the interactions and properties of all known elementary particles. However, n

The analyzed event samples are taken from the CMS pp collision data recorded in 2012. The data corresponds to an integrated luminosity of 19.7 fb−1 [1] at $$\sqrt{s} = 8\,\text{TeV}$$ . The analysis is performed in two channels, the eγ channel, where events are selected if at least one photon and one electron are present, and the μγ channel, where a muon instead of an electron is required in addition to the photon. In both channels, an excess beyond the SM prediction of events with large E T miss is searched for.

The standard model of particle physics (SM) is a Lorentz-invariant quantum field theory (QFT) that is highly successful in describing the properties of subatomic particles and their interactions.

Having determined all components of the background, the observed data are compared to the background estimation. Figures 6.1, 6.1, 6.3 show the observed distributions of the E T γ , lepton p T (p T ℓ ), E T miss, M T, H T, and the number of jets (N jet) in the eγ and μγ channels, together with the stacked background estimations. Two benchmark signal event distributions, one each from the TChiWg and T5Wg models, are also shown in the figures. The TChiWg point is for $$m_{\widetilde{\chi }} = 300\,\text{GeV}$$ , which has the nominal cross section of 0.146 pb, and the T5Wg point is for $$m_{\tilde{\mathrm{g}}} = 1000\,\text{GeV}$$ and $$m_{\widetilde{\chi }} = 425\,\text{GeV}$$ , which has the nominal cross section of 0.0122 pb. In the figures and the remainder of this section, the displayed uncertainties are a combination of the systematic and statistical uncertainties, added in quadrature. Details on the determination of the systematic uncertainties are given in Sect. 6.3.

The CERN Large Hadron Collider (LHC) is the world’s largest accelerator, storage ring, and proton–proton collider. Its two orbits are filled by oppositely circulating particle beams which collide at four interaction points (IP). Located at the interaction points are the particle detectors designed for different physics programs: A Toroidal LHC Apparatus (ATLAS) and Compact Muon Solenoid (CMS), general-purpose detectors for searches of physics beyond the standard model and precision measurements of QCD and electroweak interactions; A Large Ion Collider Experiment (ALICE), specializing in hadron physics; LHC beauty (LHCb), for precision heavy-flavor physics; and Large Hadron Collider forward (LHCf) and TOTEM (Total Cross Section, Elastic Scattering, and Diffraction Dissociation Measurement at the LHC), for hadron physics in the forward region and the proton cross section measurement. Note that LHC is designed to also circulate lead ion (208Pb82) beams, and the ion–ion and ion–proton collisions are important parts of the LHC physics program. However, the remainder of this thesis will focus on the proton–proton operation.

A search for anomalous production of events with a photon, lepton, and large E T miss using 19.7 fb−1 of pp collisions at $$\sqrt{s} = 8\,\text{TeV}$$ recorded in 2012 with the CMS detector at the CERN LHC has been presented. Signal candidate events with large E T miss and M T are counted in multiple bins of E T γ , H T, and E T miss. The amount of SM background present in the data is estimated using data-driven methods as well as MC simulation. No excess of events above the expected SM background is observed. The result is interpreted in the context of supersymmetric models with gauge-mediated supersymmetry breaking as limits on the cross section of the pair-production processes of gaugino-like particles that decay subsequently to a W boson and a photon along with nearly massless gravitinos. Assuming that the next-to-lightest supersymmetric particles are the charged and neutral winos (wino co-NLSP scenario), wino and gluino masses below 370 and 820 GeV, respectively, are excluded at 95% confidence level. Compared to the corresponding previous best limits of 221 and 619 GeV [1], this result sets a significantly more stringent constraint on such a scenario of supersymmetry.

The principal concern of the data analysis in this SUSY search is to accurately estimate how many of the candidate events selected in the signal region as described in Sect. 4.6 can be attributed to known SM processes. Multiple SM processes, or backgrounds, can contribute events in the signal region. There is no way to distinguish the SUSY signal from the SM backgrounds on an event-by-event basis, since the signal region is by definition where all the cuts have been applied. Therefore, the background estimation must be done statistically.

Particle physics is an endeavor to explain the natural world from a single principle. It is often said that the discipline takes root in the classic question: “What are we made of?”

There is no unique way to encode a quantum algorithm into a quantum circuit. With limited qubit counts, connectivity, and coherence times, a quantum circuit optimization is essential to make the best use of near-term quantum devices. We introduce a new circuit optimizer called AQCEL, which aims to remove redundant controlled operations from controlled gates, depending on initial states of the circuit. Especially, the AQCEL can remove unnecessary qubit controls from multi-controlled gates in polynomial computational resources, even when all the relevant qubits are entangled, by identifying zero-amplitude computational basis states using a quantum computer. As a benchmark, the AQCEL is deployed on a quantum algorithm designed to model final state radiation in high energy physics. For this benchmark, we have demonstrated that the AQCEL-optimized circuit can produce equivalent final states with much smaller number of gates. Moreover, when deploying AQCEL with a noisy intermediate scale quantum computer, it efficiently produces a quantum circuit that approximates the original circuit with high fidelity by truncating low-amplitude computational basis states below certain thresholds. Our technique is useful for a wide variety of quantum algorithms, opening up new possibilities to further simplify quantum circuits to be more effective for real devices.

Super-resolution is a machine-learning technique in image processing which generates high-resolution images from low-resolution images. Inspired by this approach, we perform a numerical experiment of quantum machine learning, which takes low-resolution (low plane-wave energy cutoff) one-particle molecular wavefunctions in plane-wave basis as input and generates high-resolution (high plane-wave energy cutoff) wavefunctions in fictitious one-dimensional systems, and study the performance of different learning models. We show that the trained models can generate wavefunctions having higher fidelity values with respect to the ground-truth wavefunctions than a simple linear interpolation, and the results can be improved both qualitatively and quantitatively by including data-dependent information in the ansatz. On the other hand, the accuracy of the current approach deteriorates for wavefunctions calculated in electronic configurations not included in the training dataset. We also discuss the generalization of this approach to many-body electron wavefunctions.

Entangling gates in superconducting quantum computers suffer from imperfection of control and relaxation of qubits. The Toffoli gate, essential to various quantum algorithms and to certain fast error correction schemes, represents a particular challenge, as it is decomposed into at least six two-qubit gates, whose errors accrue on top of the increased probability of qubit relaxation due to the long gate time. While it is possible to significantly reduce the depth of the Toffoli gate by exploiting three-level, i.e., qutrit states internally, existing methods are susceptible to errors arising from frequency drifts due to charge fluctuation effects in the second excited state of the transmon. In this poster, we propose a novel qutrit-based implementation of the Toffoli gate that is inherently robust against such errors. The method is extensively validated through a pulse-level simulation and experiments performed on IBM Quantum machines. On one device, where the total Toffoli gate time is 2.510μs, we obtained average gate fidelity values of 0.928 ± 0.007 and 0.896 ± 0.036 from multiple measurements performed during the first one and twenty-four hours after calibrations, respectively, where the errors represent the standard deviations of the measurement results.

Quantum state preparation is an important class of quantum algorithms that is employed as a black-box subroutine in many algorithms, or used by itself to generate arbitrary probability distributions. We present a novel state preparation method that utilizes less quantum computing resource than the existing methods. Two variants of the algorithm with different emphases are introduced. One variant uses fewer qubits and no controlled gates, while the other variant potentially requires fewer gates overall. A general analysis is given to estimate the number of qubits necessary to achieve a desired precision in the amplitudes of the computational basis states. The validity of the algorithm is demonstrated using a prototypical problem of generating Ising model spin configurations according to its Boltzmann distribution.

This data set is an output of a simulation of electrons and pions shot at a chunk of an imaging calorimeter. It is used in the case study for GarNet-on-FPGA, documented in arXiv:2008.03601. Each HDF5 file contains the following arrays: Name | Shape | Description cluster | (10000, 128, 4) | Samples for training and inference. Outermost dimension is the event (cluster). Each cluster has maximum 128 hits, each of which has four features: x, y, z, and energy. The coordinates of the hits are in cm. The energy is in GeV. The x and y coordinates are relative to the seed hit, while the z coordinate is with respect to the calorimeter front face. size | (10000) | Number of hits in each cluster. The cluster array is zero-padded when the cluster size is below 128. truth_pid | (10000) | Identity of the primary particle (0: electron, 1: pion). truth_energy | (10000) | True energy of the primary particle. raw | (10000, 4375, 2) | Raw data (actual output of the simulation). For each event (outermost dimension), hit energy and primary fraction (innermost dimension indices 0 and 1) are given for each of the 4375 sensors. Energy is in MeV. coordinates | (4375, 3) | The x, y, and z coordinates of the 4375 sensors, to be used to interpret the raw data. See the paper for the details of the simulation.

This data set is an output of a simulation of electrons and pions shot at a chunk of an imaging calorimeter. It is used in the case study for GarNet-on-FPGA, documented in arXiv:2008.03601. Each HDF5 file contains the following arrays: Name | Shape | Description cluster | (10000, 128, 4) | Samples for training and inference. Outermost dimension is the event (cluster). Each cluster has maximum 128 hits, each of which has four features: x, y, z, and energy. The coordinates of the hits are in cm. The energy is in GeV. The x and y coordinates are relative to the seed hit, while the z coordinate is with respect to the calorimeter front face. size | (10000) | Number of hits in each cluster. The cluster array is zero-padded when the cluster size is below 128. truth_pid | (10000) | Identity of the primary particle (0: electron, 1: pion). truth_energy | (10000) | True energy of the primary particle. raw | (10000, 4375, 2) | Raw data (actual output of the simulation). For each event (outermost dimension), hit energy and primary fraction (innermost dimension indices 0 and 1) are given for each of the 4375 sensors. Energy is in MeV. coordinates | (4375, 3) | The x, y, and z coordinates of the 4375 sensors, to be used to interpret the raw data. See the paper for the details of the simulation.

Connecting multiple machine learning models into a pipeline is effective for handling complex problems. By breaking down the problem into steps, each tackled by a specific component model of the pipeline, the overall solution can be made accurate and explainable. This paper describes an enhancement of object detection based on this multi-step concept, where a post-processing step called the calibration model is introduced. The calibration model consists of a convolutional neural network, and utilizes rich contextual information based on the domain knowledge of the input. Improvements of object detection performance by 0.8-1.9 in average precision metric over existing object detectors have been observed using the new model.

There is no unique way to encode a quantum algorithm into a quantum circuit. With limited qubit counts, connectivities, and coherence times, circuit optimization is essential to make the best use of near-term quantum devices. We introduce two separate ideas for circuit optimization and combine them in a multi-tiered quantum circuit optimization protocol called AQCEL. The first ingredient is a technique to recognize repeated patterns of quantum gates, opening up the possibility of future hardware co-optimization. The second ingredient is an approach to reduce circuit complexity by identifying zero- or low-amplitude computational basis states and redundant gates. As a demonstration, AQCEL is deployed on an iterative and efficient quantum algorithm designed to model final state radiation in high energy physics. For this algorithm, our optimization scheme brings a significant reduction in the gate count without losing any accuracy compared to the original circuit. Additionally, we have investigated whether this can be demonstrated on a quantum computer using polynomial resources. Our technique is generic and can be useful for a wide variety of quantum algorithms.

The usefulness and value of Multi-step Machine Learning (ML), where a task is organized into connected sub-tasks with known intermediate inference goals, as opposed to a single large model learned end-to-end without intermediate sub-tasks, is presented. Pre-optimized ML models are connected and better performance is obtained by re-optimizing the connected one. The selection of an ML model from several small ML model candidates for each sub-task has been performed by using the idea based on Neural Architecture Search (NAS). In this paper, Differentiable Architecture Search (DARTS) and Single Path One-Shot NAS (SPOS-NAS) are tested, where the construction of loss functions is improved to keep all ML models smoothly learning. Using DARTS and SPOS-NAS as an optimization and selection as well as the connections for multi-step machine learning systems, we find that (1) such a system can quickly and successfully select highly performant model combinations, and (2) the selected models are consistent with baseline algorithms, such as grid search, and their outputs are well controlled.