The work, by Renjie Wang of the Institute of High Energy Physics in Beijing, focuses on the scotogenic model, one of the most elegant extensions of the Standard Model. In this framework, neutrinos acquire their tiny masses radiatively — through quantum loops rather than direct couplings — thanks to a hidden sector of new particles protected by an exact symmetry. That same symmetry also supplies a natural dark matter candidate, making the model unusually economical: one structural addition explains neutrino mass and dark matter at once.

The key insight is a numerical coincidence with profound experimental consequences. To reproduce the observed sub-electronvolt neutrino masses, the Yukawa couplings linking the hidden sector to ordinary leptons must be extraordinarily small, of order one thousandth. Such feeble interactions make the heavy neutral fermion known as N2 exceptionally long-lived: rather than decaying promptly, it survives for tens to hundreds of millimetres, travelling a macroscopic distance through the detector before it decays. Measuring that flight path would, in effect, be measuring the very coupling that generates neutrino mass.

Wang’s study explores a particularly challenging corner of the model’s parameter space, called the compressed-spectrum regime, in which a charged scalar particle and N2 are separated by only about 200 MeV — roughly the mass of a proton. In this regime, a second, even tinier Yukawa entry, of order one millionth, renders the charged scalar long-lived as well. The result is a nested cascade unique in collider physics: the charged scalar crosses the inner tracker as a high-momentum track that abruptly disappears when it decays into N2 plus a lepton too soft to be detected, and then the neutral N2 continues invisibly for another macroscopic distance before decaying into a pair of electrons or muons at a displaced vertex.

This nested disappearing-track-plus-displaced-vertex topology offers two decisive experimental handles. The first is spatial: because N2 is emitted at the endpoint of the vanishing track, the vector connecting that endpoint to the displaced vertex must align with the original track direction to within a few milliradians. No known Standard Model process produces a high-momentum charged track that disappears and is then followed, in the same flight direction, by a displaced dilepton vertex. The second handle is temporal: the two slow, sequential flights accumulate a combined time-of-flight delay of roughly 200 to 800 picoseconds, resolvable by the CMS MIP Timing Detector, a precision timing layer planned for the High-Luminosity LHC upgrade.

Together, these requirements suppress the expected Standard Model background to between one hundred-thousandth and one thousandth of an event over the full High-Luminosity LHC dataset of 3000 inverse femtobarns. Crucially, the discovery reach is robust: even if the background were underestimated by four orders of magnitude and reached ten events, a five-sigma discovery would still require only about twenty signal events. At the study’s benchmark point — an N2 mass of 150 GeV with proper decay lengths of 300 and 189 millimetres for the charged scalar and N2 respectively — the analysis yields roughly 1212 signal events, an overwhelming margin.

The analysis was performed with a deliberately conservative, reconstruction-level strategy. Every observable entering the selection, from the timing reference to the displaced-vertex position, is built from quantities a real detector would measure, with only one truth-level input: the requirement that the vanishing track traverse all four barrel pixel layers, decaying at a radius between 160 and 300 millimetres. Two complementary Level-1 trigger paths — one based on an energetic initial-state radiation jet plus missing energy, the other on a displaced muon — ensure the rare events can be recorded at all, since no hard charged object appears at the primary collision point.

Scanning a grid of 56 benchmark points across the two-dimensional plane of the two decay lengths, the study finds that all 40 points with measurable disappearing tracks exceed both the five-sigma discovery threshold and the 95 percent confidence exclusion limit, with peak sensitivity of about 1635 signal events at decay lengths of 150 and 60 millimetres. The remaining 16 points, where the charged scalar decays too early to leave a reconstructable track stub, fall into a blind zone that only a dedicated displaced-vertex search could address. No existing ATLAS or CMS analysis simultaneously targets this correlated cascade: standalone disappearing-track searches assume the downstream particle is effectively stable, while standalone displaced-vertex searches impose no upstream track requirement.

The deeper significance lies in what the measurement would reveal. Because the N2 decay length is controlled by the same Yukawa couplings that generate sub-electronvolt neutrino masses, the covered region probes couplings between roughly six ten-thousandths and four thousandths — precisely the range associated with the neutrino mass scale. Mapping the nested cascade would therefore turn the High-Luminosity LHC into a direct collider probe of the interaction responsible for neutrino mass, complementing neutrino oscillation experiments and cosmological observations.

Challenges remain before the strategy can be implemented in practice. The dominant residual background arises from pileup combinatorics at the HL-LHC’s expected 200 simultaneous collisions per bunch crossing, and a definitive estimate will require data-driven control regions or full pileup-overlay simulation with four-dimensional tracking. Tau-lepton decay channels of N2, which would demand more complex vertex reconstruction, are deferred to future work. Nevertheless, the study provides a concrete roadmap for adding the scotogenic nested cascade to the Run 4 search programme — and a vivid reminder that the next neutrino-mass clue may not come from a deep underground detector, but from a track that vanishes in the dark of the CMS pixel barrel.

Subject of Research: A collider search strategy for long-lived particles in the scotogenic neutrino mass model at the High-Luminosity LHC

Article Title: Nested disappearing-track plus displaced-vertex topology at the HL-LHC: a dual long-lived-particle signature of the scotogenic model

Article References: Wang, R. (2026). Nested disappearing-track plus displaced-vertex topology at the HL-LHC: a dual long-lived-particle signature of the scotogenic model. The European Physical Journal C, 86(10), Article 1187. https://doi.org/10.1140/epjc/s10052-026-16416-2

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-16416-2

Keywords: scotogenic model, neutrino mass, long-lived particles, disappearing track, displaced vertex, HL-LHC, CMS MIP Timing Detector, dark matter, Yukawa coupling, particle physics, beyond the Standard Model, Nested

The new algorithm, dubbed MLPF, replaces the modular, physics-motivated heuristics of standard PF with a transformer neural network trained end-to-end on simulated collision events. Instead of linking tracks and calorimeter clusters through geometric proximity rules refined over years of tuning, the model learns those correlations directly from data-like simulation, processing the entire event in a single inference step rather than through sequential iterations. The architecture embeds each reconstructed track and calorimeter cluster into a 256-dimensional latent space and passes them through parallel transformer encoders for classification and regression, using scaled dot-product attention to learn pairwise correlations across the full event. With roughly 5.1 million trainable parameters, the model is compact by modern deep-learning standards, comparable in size to the unified particle transformer CMS already uses for jet flavor tagging.

Training such a model required solving a deceptively hard problem: defining what the network should learn. The collaboration constructed a particle-level training target from the Geant4 detector simulation, selecting only particles that genuinely leave detectable signals, excluding those that fall outside the detector acceptance or below energy thresholds, and treating charged hadrons that fail to produce a reconstructed track as neutral hadrons. A per-particle loss function assigns each target particle to exactly one input element, its primary track or cluster, combining a binary classification loss, a particle-identification loss computed with focal loss to handle class imbalance, and a regression loss on the four-momentum expressed in logarithmic ratios relative to the primary element. A transverse-momentum-dependent weighting factor emphasizes high-energy particles in the training tails, and all three loss components converge smoothly with no signs of overfitting.

The training itself ran on eight AMD MI250X GPUs at the LUMI supercomputer in Finland, taking about six hours per epoch for ten epochs using the PyTorch framework and a LAMB optimizer with cosine annealing. Simulated samples were generated under LHC Run-3 2023 conditions at a center-of-mass energy of 14 TeV, with testing events generated under 2024 conditions at 13.6 TeV, deliberately shifted from the training configuration to demonstrate that the model generalizes across detector conditions and collision energies rather than memorizing a single operating point. Events contained between 55 and 75 additional proton-proton interactions per bunch crossing, the pileup that contaminates real LHC collisions and makes reconstruction dramatically harder.

The physics performance numbers are striking. In simulated top quark-antiquark events, MLPF improves jet energy resolution by 10 to 20 percent compared to standard PF for jets with transverse momentum between 30 and 100 GeV, an improvement that holds consistently in both top and quantum chromodynamics multijet samples. Because the model is trained only on individual particles and never explicitly optimized on jets, the better jet performance emerges purely from improved particle-level reconstruction. MLPF also achieves higher efficiency for reconstructing neutral hadrons while maintaining the same misidentification rate as standard PF, a combination that is difficult to achieve with rule-based linking, and the reconstructed missing transverse momentum distributions remain generally consistent with the standard algorithm.

Crucially, the team did not stop at simulation. The model was commissioned on a small sample of real CMS data collected in 2024 with a dijet trigger, corresponding to an integrated luminosity of 375 inverse picobarns. Running both PF and MLPF on the same data events, the collaboration found comparable jet distributions and consistent dijet asymmetry, an observable particularly sensitive to relative jet resolutions. The MLPF-reconstructed missing transverse momentum spectrum was somewhat harder in the tails, matching the trend seen in simulation. This data validation makes MLPF the first machine-learning-based full-event reconstruction pipeline ever applied to data from a hadron collider experiment, a milestone that could reshape reconstruction strategies across the field.

The computational gains are equally significant. Standard PF runs on CPUs and took approximately 110 milliseconds per event in the benchmark, while MLPF performed inference on an Nvidia L4 GPU with a median of 20 milliseconds per event, a roughly fivefold speedup. Because of the iterative nature of the baseline PF block algorithms, no performant GPU implementation exists for standard PF; MLPF unlocks accelerator use for this critical step. The FlashAttention kernel keeps memory and runtime scaling manageable despite the nominally quadratic cost of full-event attention, and a single L4 GPU supported 64 parallel inference streams using less than 10 GB of memory. The model also produces fully reproducible outputs, with repeated evaluations on the same GPU yielding identical results and CPU and GPU inference agreeing within numerical precision.

Integration into the CMS software framework was designed to be minimally invasive. The model is exported as a frozen Open Neural Network Exchange file and loaded through ONNX Runtime, taking reconstructed tracks and calorimeter clusters as inputs and producing particle candidates that downstream modules consume without modification. Upstream tracking and cluster-building algorithms remain untouched, making MLPF fully complementary to ongoing GPU-based developments elsewhere in the reconstruction chain. This separation of concerns means analyses can switch between PF and MLPF with a configuration flag, enabling the side-by-side comparisons that underpinned the validation campaign.

The motivation extends well beyond Run 3. At the high-luminosity upgrade of the LHC, expected to begin operation later this decade, average pileup will reach up to 200 simultaneous interactions per bunch crossing, roughly three times the conditions tested here. Reconstruction complexity grows rapidly with pileup, and hand-engineered algorithms whose combinatorics scale poorly with detector occupancy face diminishing returns. Machine-learning approaches offer favorable scaling, faster retraining when detector geometries change, and even the possibility of baking analysis-specific optimizations directly into reconstruction, as CMS already demonstrated by improving missing transverse momentum resolution for its W boson mass measurement. For proposed future colliders, ML-based particle-flow can accelerate detector design studies through rapid iteration and differentiable programming pipelines.

For now, the collaboration is careful about what has and has not been proven. Electron reconstruction sequences have not yet been retuned for MLPF, the forward detector region needs re-optimization ahead of the high-granularity endcap calorimeter upgrade, and deploying the algorithm in production requires further validation against simulation across a wide range of samples. But the demonstration is complete in its essentials: a transformer trained on simulation runs faster, generalizes across operating conditions, matches the physics performance of decades of hand-tuned software, and has already interpreted real collision data. The era in which the world’s most complex detectors are read out by learned models rather than rulebooks has effectively begun.

Subject of Research: Machine-learning-based particle-flow reconstruction in the CMS detector

Article Title: Full event interpretation with machine-learning-based particle-flow reconstruction in the CMS detector

Article References: Hayrapetyan, A., Makarenko, V., Tumasyan, A., Adam, W., Benato, L., Bergauer, T., Dragicevic, M., Hussain, P. S., Jeitler, M., Krammer, N., Li, A., Liko, D., Matthewman, M., Schieck, J., Schöfbeck, R., Shooshtari, M., Sonawane, M., Waltenberger, W., Wulz, C. E., … CMS Collaboration (2026). Full event interpretation with machine-learning-based particle-flow reconstruction in the CMS detector. The European Physical Journal C, 86(8), 1025. https://doi.org/10.1140/epjc/s10052-026-15754-5

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-15754-5

Keywords: CMS, particle-flow reconstruction, machine learning, transformers, CERN, LHC, jet energy resolution, GPU computing, high-energy physics, pileup, HL-LHC, graph attention

The study, published in The European Physical Journal C by Daniel Gillies and Andrea Banfi of the University of Sussex and Adam Martin of the University of Notre Dame, focuses on the production of a W-plus and W-minus pair in which both bosons decay leptonically, yielding an electron, a muon, and two invisible neutrinos. Diboson production is one of the most promising arenas for effective field theory because the higher-dimensional operators that encode new physics grow with energy, so their fingerprints should appear most clearly in the high-energy tails of kinematic distributions. The trouble is that the single most informative quantity, the invariant mass of the W pair, cannot be measured directly, because the neutrinos carry away energy that no detector records.

Experimentalists therefore work with proxies. The most common is the invariant mass of the electron-muon pair, which is fully measurable and has featured in numerous EFT analyses of the WW channel. But previous work had already shown that the dilepton mass correlates poorly with the true diboson mass, and the new paper quantifies just how dangerous that poor correlation is. The authors computed the conditional expectation value of the W-pair mass given each bin of the dilepton mass, separately for the Standard Model, for dimension-six operators, and for dimension-eight operators. The results are striking: in the Standard Model the average W-pair mass is roughly four-thirds of the dilepton mass, at dimension six it approaches twice the dilepton mass, and at dimension eight the relation degrades further, with the average W-pair mass reaching roughly four times the dilepton mass at low dilepton values.

That order-by-order drift in the correlation has a profound consequence. Higher-dimensional operators grow faster with energy, so events with very large W-pair masses can feed contributions into surprisingly low dilepton-mass bins. A physicist who clips the simulated dimension-six prediction at a W-pair mass below the new-physics scale, hoping to keep the theory in its valid regime, may still be fitting bins that are secretly dominated by dimension-eight effects from events far above the cutoff. The paper demonstrates this concretely: imposing the cut at exactly the new-physics scale of one teraelectronvolt leaves a window between 200 and 400 gigaelectronvolts in the dilepton distribution where dimension-eight squared contributions still dominate, and even a stricter cut corresponding to a scale of 1.65 teraelectronvolts fails to purge every contaminated bin.

The clipping procedure, in which the constraint is applied only to the EFT simulation and never to the data, carries a deeper conceptual risk that the authors highlight. Mathematically, multiplying a dimension-six Wilson coefficient by a step function of the momentum is equivalent to replacing the clean SMEFT operator with a modified interaction that switches off above a threshold. Replacing the step function with a smooth suppression would generate an infinite tower of higher-dimensional operators, meaning the analysis is no longer a truncation of the EFT at dimension six at all. The fitted coefficient therefore loses its interpretation as a local, model-independent parameter, which undermines the central selling point of EFT fits: their ability to characterise arbitrary new physics without committing to a specific model.

As a more rigorous alternative, the authors advocate a bin-by-bin comparison, in which the squared contributions of the leading dimension-six and dimension-eight operators are computed directly and only bins where dimension six dominates by at least a factor of two are retained. This procedure automatically guarantees that the fitted data lie within the convergent regime of the expansion, but it is cumbersome: it requires knowing the dominant dimension-eight operators in advance and re-evaluating the allowed bins for every assumed value of the new-physics scale. The ideal solution would be a measurable proxy for the W-pair mass good enough that a simple experimental cut on the data itself could carve out the valid region.

To find one, the team examined three transverse-mass observables, labelled M-T1, M-T2 and M-T3, which were originally designed to approximate the W-pair mass in the presence of missing neutrino energy. Each treats the event as a visible system, the electron-muon pair, plus an invisible system of neutrinos whose mass must be guessed. M-T1 assumes the invisible system is massless, M-T2 additionally assumes the visible system is massless, and M-T3 exploits the symmetry of the topology by assuming the visible and invisible masses are equal. Computing the conditional expectation of the W-pair mass for each variable, the authors found that M-T3 tracks the true diboson mass far better than the dilepton mass, and that the ratio of dimension-eight to dimension-six contributions follows the theoretically expected scaling with M-T3 much more closely than with the dilepton mass.

The payoff is a practical recipe. Because M-T3 is fully measurable, an experimental cut on it can be applied directly to the data, ensuring EFT validity without any reference to unobservable quantities or to the size of dimension-eight operators. In sensitivity studies projected for the High-Luminosity LHC, using fiducial cuts based on ATLAS analyses and predictions computed with the MCFM-RE framework at next-to-leading-logarithmic accuracy, the authors compared the cut-on-data approach against the clipping approach and the bin-by-bin method. The cut on M-T3 yields constraints comparable to those from clipping, and somewhat more conservative than the bin-by-bin method, but with a crucial advantage: the extracted Wilson coefficients retain a clear interpretation in terms of a local effective theory.

An intriguing subtlety emerged when the authors compared which observable delivers the strongest bounds once validity is guaranteed bin by bin. Despite tracking the W-pair mass so faithfully, M-T3 actually gives slightly weaker constraints than the dilepton mass. The reason is almost paradoxical: precisely because M-T3 correlates so well with the true energy scale, its low-energy bins receive far less contamination from high-energy events, and those low-energy bins are where sensitivity to higher-dimensional operators is weakest. The dilepton mass, by smearing high-energy events down into lower bins, accidentally concentrates BSM sensitivity, but only at the cost of an unpredictable mixing with uncontrolled higher-order effects.

The broader lesson resonates well beyond the WW channel. As the LHC community prepares increasingly ambitious EFT interpretations of differential measurements, the validity of the expansion cannot be taken for granted, and the choice of observable and validity prescription can quietly reshape the physics conclusions. The authors note that electroweak Sudakov corrections, not included in this analysis, will modify the high-energy tails of both signal and background, likely shifting the absolute reach of the analysis while preserving the qualitative differences between the validity methods. They also stress that the philosophy of checking EFT validity should apply to every production channel, not only the dominant ones, since assuming that gluon-fusion operators are suppressed relative to quark-antiquark operators is itself an assumption about the underlying new physics. Extending the analysis to the quark channel and to fermionic operators is the natural next step, and one that could determine whether the current generation of EFT constraints is as robust as it appears.

Subject of Research: Validity of Standard Model effective field theory methods in LHC W-boson pair production measurements

Article Title: Probing EFT breakdown in the tails of (W^+ W^-) observables

Article References: Gillies, D., Banfi, A., & Martin, A. (2026). Probing EFT breakdown in the tails of $$W^+ W^-$$ observables. The European Physical Journal C, 86(9), Article 1084. https://doi.org/10.1140/epjc/s10052-026-16081-5

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-16081-5

Keywords: effective field theory, SMEFT, W boson pair production, Large Hadron Collider, dimension-six operators, dimension-eight operators, invariant mass distributions, transverse mass, EFT validity, HL-LHC, particle physics phenomenology, new physics searches

The target of the proposed search is the minimal scotogenic model, an elegant extension of the Standard Model first proposed by Ernest Ma in 2006. In this framework, neutrino masses and the dark matter particle are linked in a single loop of quantum processes. The model adds a set of particles that are odd under a symmetry called Z2, including an inert scalar doublet and heavy right-handed neutrinos. When the lightest of these new fermions plays the role of dark matter, the observed abundance of dark matter in the universe, corresponding to a relic density of about 0.12 in standard cosmological units, can be naturally explained if it co-annihilates with a charged partner scalar whose mass lies only a few gigaelectronvolts above its own.

That compressed mass regime is precisely where the LHC struggles most. When the charged inert scalar and the dark matter fermion are separated by only about 2 GeV, the scalar decays into a dark matter particle plus a single very soft charged lepton, typically carrying a transverse momentum near 2 GeV. This sits right at the edge of standard track reconstruction, which at CMS requires transverse momenta of roughly 2 GeV or more. In the detector’s 3.8 tesla magnetic field, such a low-momentum lepton curves with a gyration radius of about 1.8 meters, exceeding the tracker’s radius and degrading conventional reconstruction. Meanwhile, the charged parent itself travels visibly through the inner detector as a heavy particle before abruptly vanishing, leaving what physicists call a disappearing track: a stub of hits that terminates with no matching continuation.

Existing searches were not built for this topology. ATLAS’s displaced-lepton searches typically demand lepton momenta above 30 to 60 GeV, a factor of fifteen or more beyond what this signal produces. Displaced-vertex searches require two or more tracks at a common vertex, impossible when the decay yields only one soft charged particle. Disappearing-track and soft-track analyses retain only partial, unoptimized acceptance. Wang’s simplified truth-level estimates quantify this gap dramatically: for a benchmark with a 200 GeV charged scalar and a 2 GeV mass splitting, the ATLAS disappearing-track search would retain roughly 8.5 percent of the signal, the CMS version about 3.3 percent, and a soft displaced-track search a vanishing 0.01 percent, while displaced-vertex searches capture none at all.

The proposed remedy adds time as an independent observable, capitalizing on a unique capability of the HL-LHC era. Both CMS and ATLAS are installing precision timing layers, the CMS MIP Timing Detector in the barrel and the ATLAS High-Granularity Timing Detector in the endcaps, each offering about 30 picoseconds of timing resolution for minimum-ionizing particles. These devices were primarily conceived to disentangle the roughly 200 simultaneous proton-proton collisions expected per bunch crossing. But the study shows they can do far more. The 2 GeV mass splitting makes the charged scalar long-lived: with a Yukawa coupling of roughly nine parts in ten million, its decay length reaches about 300 millimeters, corresponding to a lifetime of one nanosecond. The parent travels macroscopic distances, between roughly 100 and 1000 millimeters, before decaying.

That journey creates a measurable delay. Because the parent scalar moves at only about 35 percent of the speed of light, the decay products arrive at the timing layer tens to hundreds of picoseconds late relative to a prompt particle born at the collision point. For the 200 GeV benchmark, the expected delay is about 700 picoseconds, once a small helical path-length correction in the magnetic field is included, more than twenty times the detector’s timing resolution. By contrast, every prompt Standard Model background, from W boson decays to b-hadron semileptonic decays, produces arrival delays below about 0.2 picoseconds, a separation of more than three orders of magnitude. The signal and background thus occupy cleanly distinct regions of a two-dimensional map of momentum versus timing delay.

The proposed signal region, labeled SR-4DT, combines six requirements: missing transverse energy above 105 GeV at trigger level and 150 GeV in the final selection, a leading jet above 100 GeV from initial-state radiation, a delayed timing hit in the MTD barrel with a delay exceeding 200 picoseconds, a mandatory low-momentum disappearing-track stub between 0.7 GeV and 300 millimeters transverse displacement that points to the timing hit, a veto on any prompt lepton above 25 GeV, and compatibility with the hard-scatter primary vertex. Crucially, the analysis never demands full lepton reconstruction, sidestepping the tracking threshold that cripples conventional approaches. The stub-hit pointing requirement, with an angular window of 0.1 in radius, suppresses accidental coincidences by roughly a factor of a thousand or more.

The simulated performance is remarkable. Across 4.2 million simulated prompt Standard Model background events, spanning W plus jet, Z plus jet and top pair production at a combined equivalent luminosity of 20.6 inverse femtobarns, not a single event survived the SR-4DT selection. Signal efficiencies range from 12 to 36 percent across the probed mass range, with 14 percent at the 200 GeV benchmark, translating to roughly 2,331 signal events at the full 3,000 inverse femtobarns of the HL-LHC program, fifty-four times above even the conservative exclusion threshold. Under deliberately conservative background assumptions derived from the finite simulated sample, the projected 95 percent confidence-level exclusion reaches charged scalar masses of about 670 GeV at a decay length of 300 millimeters; in the optimistic zero-background limit, the reach extends to roughly 1,100 GeV. The sensitivity holds across the full lifetime scan from 10 to 3,000 millimeters, with the yield peaking near 300 millimeters where both the timing delay and the stub reconstruction are optimal.

Wang is careful about the study’s limits. The simulation uses Delphes-level fast detector response with a parametrized timing layer, and instrumental and combinatorial backgrounds, including fake stubs, random timing hits, hadronic interactions in detector material and beam-induced background, were not modeled and would require full detector simulation to quantify. Physics arguments suggest each category is strongly suppressed, but the definitive residual background remains an open question for the experimental collaborations. Pile-up contamination is controlled by the mandatory stub-to-hit association, since out-of-time pile-up vertices will rarely point to a legitimate disappearing-track stub tied to the hard-scatter primary vertex.

If confirmed experimentally, the strategy would open a discovery window onto exactly the co-annihilation corridor that cosmology favors and collider searches have historically missed. The same small Yukawa coupling that renders the charged scalar long-lived automatically evades constraints from lepton-flavor-violating muon decays and direct-detection experiments, where the scotogenic candidate’s spin-independent cross section is loop-suppressed and consistent with the latest LUX-ZEPLIN limits. With the HL-LHC’s timing detectors coming online, dark matter may soon be caught not by where it goes, but by when it arrives.

The timing-based strategy also connects to a broader effort across collider physics to exploit temporal information in event reconstruction. Timing layers like the CMS MIP Timing Detector use silicon sensors with internal gain to timestamp minimum-ionizing particles, and their performance has been validated with test-beam measurements. What the new study illustrates is that such instrumentation, designed with pile-up separation in mind, can be repurposed as a discovery tool for long-lived particles whose macroscopic lifetimes imprint a characteristic time-of-flight signature.

From a theoretical standpoint, the compressed co-annihilation corridor occupies a special place in thermal dark matter calculations. When the mass gap between a particle and its charged companion shrinks to a few gigaelectronvolts, Boltzmann suppression of the heavier species weakens, and co-annihilation processes efficiently deplete the relic abundance without requiring large annihilation cross sections for the dark matter particle itself. This same kinematic compression, however, suppresses the visible energy release in collider decays, creating the tension between cosmological viability and experimental accessibility that motivates the timing approach.

The framework may generalize beyond the scotogenic model. Compressed spectra with macroscopic decay lengths and soft visible daughters arise in supersymmetric wino and higgsino scenarios, in simplified dark matter models, and in other radiative neutrino-mass constructions. A timing-assisted search requiring only a disappearing-track stub and a delayed hit is largely model-independent, depending mainly on the parent lifetime and the presence of a timing-visible charged daughter.

Verification will require the experimental collaborations to confront full detector simulation, including the combinatorial and instrumental backgrounds absent from this fast-simulation study. If the projected zero-background regime survives that scrutiny, the HL-LHC could deliver sensitivity to a class of dark matter models whose thermal production history makes them cosmologically favored yet collider-hidden.

Subject of Research: Timing-assisted collider search for ultra-compressed scotogenic dark matter at the High-Luminosity LHC

Article Title: Probing ultra-compressed scotogenic dark matter at the HL-LHC via 4D spacetime tracking

Article References: Wang, R. (2026). Probing ultra-compressed scotogenic dark matter at the HL-LHC via 4D spacetime tracking. The European Physical Journal C, 86(9), Article 1062. https://doi.org/10.1140/epjc/s10052-026-16304-9

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-16304-9

Keywords: dark matter, scotogenic model, HL-LHC, disappearing tracks, precision timing, long-lived particles, neutrino masses, co-annihilation, CMS MTD, ATLAS HGTD, particle physics, displaced leptons