We describe the outcome of a data challenge conducted as part of the Dark Machines (https://www.darkmachines.org) initiative and the Les Houches 2019 workshop on Physics at TeV colliders. The challenged aims to detect signals of new physics at the Large Hadron Collider (LHC) using unsupervised machine learning algorithms. First, we propose how an anomaly score could be implemented to define model-independent signal regions in LHC searches. We define and describe a large benchmark dataset, consisting of >1 billion simulated LHC events corresponding to 10\, fb^{-1}10fb−1 of proton-proton collisions at a center-of-mass energy of 13 TeV. We then review a wide range of anomaly detection and density estimation algorithms, developed in the context of the data challenge, and we measure their performance in a set of realistic analysis environments. We draw a number of useful conclusions that will aid the development of unsupervised new physics searches during the third run of the LHC, and provide our benchmark dataset for future studies at https://www.phenoMLdata.org. Code to reproduce the analysis is provided at https://github.com/bostdiek/DarkMachines-UnsupervisedChallenge.

We present a complete phenomenological prospectus for thermal relic neutralinos. Including Sommerfeld enhancements to relic abundance and halo annihilation calculations, we obtain direct, indirect, and collider discovery prospects for all neutralinos with mass parameters M 1 , M 2 , |µ| < 4 TeV, that freeze out to the observed dark matter abundance, with scalar superpartners decoupled. Much of the relic neutralino sector will be uncovered by the direct detection experiments Xenon1T and LZ, as well as indirect detection with CTA. We emphasize that thermal relic higgsinos will be found by next-generation direct detection experiments, so long as M 1,2 < 4 TeV. Charged tracks at a 100 TeV hadron collider complement indirect searches for relic winos. Thermal relic bino-winos still evade all planned experiments, including disappearing charged-track searches. However, they can be discovered by compressed electroweakino searches at a 100 TeV collider, completing the full coverage of the relic neutralino surface.2

Massive dwarf galaxies that merge with the Milky Way on prograde orbits can be dragged into the disk plane before being completely disrupted. Such mergers can contribute to an accreted stellar disk and a dark matter disk. We present evidence for Nyx, a vast new stellar stream in the vicinity of the Sun, that may provide the first indication that such an event occurred in the Milky Way. We identify about 500 stars that have coherent radial and prograde motion in this stream using a catalog of accreted stars built by applying deep learning methods to the second Gaia data release. Nyx is concentrated within ±2 kpc of the Galactic midplane and spans the full radial range studied (6.5-9.5 kpc). The kinematics of Nyx stars are distinct from those of both the thin and thick disk. In particular, its rotational speed lags the disk by ∼ 80 km/s and its stars follow more eccentric orbits. A small number of Nyx stars have chemical abundances or inferred ages; from these, we deduce that Nyx stars have a peak metallicity of [Fe/H] ∼ −0.5 and ages ∼ 10-13 Gyr. Taken together with the kinematic observations, these results strongly favor the interpretation that Nyx is the remnant of a disrupted dwarf galaxy. To further justify this interpretation, we explicitly demonstrate that metal-rich, prograde streams like Nyx can be found in the disk plane of Milky Way-like galaxies using the Fire hydrodynamic simulations. Future spectroscopic studies will be able to validate whether Nyx stars originate from a single progenitor.

Applications of machine learning tools to problems of physical interest are often criticized for producing sensitivity at the expense of transparency. To address this concern, we explore a data planing procedure for identifying combinations of variables-aided by physical intuition-that can discriminate signal from background. Weights are introduced to smooth away the features in a given variable(s). New networks are then trained on this modified data. Observed decreases in sensitivity diagnose the variable's discriminating power. Planing also allows the investigation of the linear versus nonlinear nature of the boundaries between signal and background. We demonstrate the efficacy of this approach using a toy example, followed by an application to an idealized heavy resonance scenario at the Large Hadron Collider. By unpacking the information being utilized by these algorithms, this method puts in context what it means for a machine to learn.

We map the parameter space for MSSM neutralino dark matter which freezes out to the observed relic abundance, in the limit that all superpartners except the neutralinos and charginos are decoupled. In this space of relic neutralinos, we show the dominant dark matter annihilation modes, the mass splittings among the electroweakinos, direct detection rates, and collider cross-sections. The mass difference between the dark matter and the next-to-lightest neutral and charged states is typically much less than electroweak gauge boson masses. With these small mass differences, the relic neutralino surface is accessible to a future 100 TeV hadron collider, which can discover inter-neutralino mass splittings down to 1 GeV and thermal relic dark matter neutralino masses up to 1.5 TeV with a few inverse attobarns of luminosity. This coverage is a direct consequence of the increased collider energy: in the Standard Model events with missing transverse momentum in the TeV range have mostly hard electroweak radiation, distinct from the soft radiation shed in compressed electroweakino decays. We exploit this kinematic feature in final states including photons and leptons, tailored to the 100 TeV collider environment.

Determining the best method for training a machine learning algorithm is critical to maximizing its ability to classify data. In this paper, we compare the standard "fully supervised" approach (which relies on knowledge of event-by-event truth-level labels) with a recent proposal that instead utilizes class ratios as the only discriminating information provided during training. This so-called "weakly supervised" technique has access to less information than the fully supervised method and yet is still able to yield impressive discriminating power. In addition, weak supervision seems particularly well suited to particle physics since quantum mechanics is incompatible with the notion of mapping an individual event onto any single Feynman diagram. We examine the technique in detail -both analytically and numerically -with a focus on the robustness to issues of mischaracterizing the training samples. Weakly supervised networks turn out to be remarkably insensitive to a class of systematic mismodeling. Furthermore, we demonstrate that the event level outputs for weakly versus fully supervised networks are probing different kinematics, even though the numerical quality metrics are essentially identical. This implies that it should be possible to improve the overall classification ability by combining the output from the two types of networks. For concreteness, we apply this technology to a signature of beyond the Standard Model physics to demonstrate that all these impressive features continue to hold in a scenario of relevance to the LHC. Example code is provided on GitHub.

Adding a fermion to the standard model particle content which is a fiveplet under SU (2)L gives a dark matter candidate. The lightest state in the multiplet is neutral and automatically stable. The charged component of the multiplet is only slightly heavier and can travel a finite distance in the LHC detectors before decaying, leaving a charged track which disappears before the edge of the detector. We use the recent ATLAS and CMS searches to exclude a Majorana fiveplet with a mass up to 267 GeV. We estimate that with 3 ab −1 of √ s = 14 TeV data this could be pushed to a mass of 520 GeV. These exclusions are ∼ 10% greater than what is achieved for wino-like dark matter. We also discuss how the doubly charged states could be used to distinguish a disappearing track signal from that given by a triplet such as the pure wino.

A new, strongly-coupled "dark" sector could be accessible to LHC searches now. These dark sectors consist of composites formed from constituents that are charged under the electroweak group and interact with the Higgs, but are neutral under Standard Model color. In these scenarios, the most promising target is the dark meson sector, consisting of dark vector-mesons as well as dark pions. In this paper we study dark meson production and decay at the LHC in theories that preserve a global SU (2) dark flavor symmetry. Dark pions -like the pions of QCD -can be pair-produced through resonant dark vector meson production, pp → ρ D → π D π D , and decay in one of two distinct ways: "gaugephobic", when π D → ff generally dominates; or "gaugephilic", when π D → W + h, Z + h dominates once kinematically open. Unlike QCD, the decay π 0 D → γγ is virtually absent due to the dark flavor symmetry. We recast a vast set of existing LHC searches to determine the current constraints on (and future opportunities for) dark meson production and decay. When m ρ D is slightly heavier than 2m π D and ρ ±,0 D kinetically mixes with the weak gauge bosons, the 8 TeV same-sign lepton search strategy sets the best bound, m π D > 500 GeV. Yet, when only the ρ 0 D kinetically mixes with hypercharge, we find the strongest LHC bound is m π D > 130 GeV, that is only slightly better than what LEP II achieved two decades ago. We find the relative insensitivity of LHC searches, especially at 13 TeV, can be blamed mainly on their penchant for high mass objects or large missing energy. Dedicated searches would undoubtedly yield substantially improved sensitivity. We provide a GitHub page to speed the implementation of these searches in future LHC analyses. Our findings for dark meson production and decay provide a strong motivation for model-independent searches of the form pp → A → B + C → SM SM + SM SM where the theoretical prejudice is for SM to be a 3rd generation quark or lepton, W, Z, or h.