Search for a light Higgs boson decaying to long-lived weakly-interacting particles in proton-proton collisions at sqrt(s) = 7 TeV with the ATLAS detector

This is the first ATLAS search for h → π_v π_v with both pseudoscalars decaying inside the muon spectrometer, and the interesting part is not the limit itself but the machinery built to get there. The muon-RoI cluster trigger and the MS tracklet/vertex algorithm are genuinely new, and they buy roughly two orders of magnitude in proper decay length over the CDF and D0 searches cited. The exclusion windows in Table II reach cτ ~ 27 m. That reach is the result; the numbers are modest but the topology is one nobody had covered at the LHC before this, and the paper is honest that it is a benchmark Hidden Valley model with 100% branching ratio assumed and only two π_v masses tried (20 and 40 GeV). Those assumptions are stated, not buried, which is the right way to do it (abstract:s2). The background is fully data-driven, written out as a formula, with P_vertex and P_reco quoted with uncertainties, giving 0.03 ± 0.02 expected against zero observed. That is about as clean as a null result gets, and the limit is not going to move much under background-systematic shifts. Where I'd push: the trigger and vertex efficiencies are calibrated on punch-through jets, and the validation sample is barrel-only (|η| ≤ 1.4). But the event selection only requires one π_v to decay in the barrel; the second can be in the forward spectrometer, where the vertex algorithm is quoted at ~40% efficient and is not validated against data at all. The 15% data-to-MC ratio from Table I is then applied globally. The stress-test note is right that this is a real gap, though I'd call it a 10–20% effect on the efficiency, not something that changes the qualitative conclusion given zero events on 0.03 background. The 'at most ~1% relative change in P_reco' line is asserted without support and should either be shown or dropped. The malformed arXiv identifier in Ref. [27] is a typo, not a substantive problem. Who should read it: anyone working on long-lived particles, hidden sectors, or displaced-vertex reconstruction, and detector people interested in what the muon spectrometer can do beyond muon identification. It deserves a serious referee, mainly to press on the forward-region efficiency question and to get the P_reco claim either justified or removed.

The central claim (Table II exclusion windows) depends on the total signal efficiency, which combines the barrel trigger efficiency (calibrated on punch-through jets, 14% systematic) and the vertex reconstruction efficiency (16% systematic). The punch-through-jet validation sample is selected in the barrel calorimeter (|η| ≤ 1.4), and Table I compares data and MC only in that region. However, the signal selection requires one πv to decay in the barrel (trigger) and allows the second πv to decay in the forward spectrometer (8 ≤ |z| ≤ 14 m), where the vertex algorithm is quoted as ~40% efficient but is not validated against data. The 16% vertex systematic is derived from the barrel data/MC ratio and is applied globally. If the forward data/MC ratio differs from the barrel by more than 16%, the total signal efficiency is biased, and the limit shifts accordingly. The paper does not report a forward-specific validation or a forward-specific systematic. This is a concrete gap in the efficiency calibration, though it is likely a ~10-20% effect on the limit rather than a qualitative change.

The paper is a careful null-result search. The background is data-driven and tiny (0.03 ± 0.02), so the limit is driven by the zero-observation Poisson term and is insensitive to background systematics. The signal efficiency is the load-bearing quantity. The reader correctly identifies the punch-through-jet proxy as the weakest link. I sharpen this to the forward spectrometer, where the validation sample does not reach. The effect is bounded by the 16% systematic, so the verdict remains ACCEPT. The paper's own statements (Table I, Fig 1b, the 16% systematic) support the concern but also show it is quantified. No other concern rises to the level of threatening the central claim.

The result is a genuinely new reach statement: by using the muon spectrometer as the decay volume, the search extends sensitivity to proper decay lengths of order metres to tens of metres, well beyond the few-hundred-millimetre range accessible to the Tevatron displaced-vertex searches, and it does so with a background that is measured in data rather than modelled. The signature-driven trigger and the muon-spectrometer vertexing are non-trivial detector-level developments that are reusable for a broad class of long-lived-particle models. The exclusion windows in Table II are falsifiable predictions of the Hidden Valley benchmark and are quoted in a form that can be applied directly by model builders. The paper is candid about its main extrapolation, namely that the trigger and vertex efficiencies are calibrated on a punch-through-jet control sample; that candour is a strength, though it also defines where the residual questions lie.

Many extensions of the Standard Model predict that a Higgs boson lighter than about 140 GeV can decay into a pair of neutral, weakly-coupled particles that live long enough to travel metres before decaying. This paper reports the first ATLAS search for that topology, using 1.94 fb$^{-1}$ of 7 TeV proton-proton collisions collected in 2011 and requiring that both of the long-lived particles decay inside the muon spectrometer. The benchmark is a Hidden Valley model in which the Higgs goes to two pseudoscalars $h^0 to pi_v pi_v$, each of which decays to a boosted cluster of hadrons and photons in the outer detector. The point of the search is reach: because the muon spectrometer is large, the analysis can probe proper decay lengths from tens of centimetres to roughly 27 m, far beyond the few-hundred-millimetre reach of the earlier Tevatron searches that were confined to tracking volumes. No candidate events survive, against an expected background of $0.03 pm 0.02$ events, and that null result is turned into 95% confidence-level exclusions on Higgs production times branching ratio as a function of the pseudoscalar lifetime.

What carries the argument. The carrying mechanism is a signature-driven trigger plus a dedicated muon-spectrometer vertex reconstruction, both built for a decay topology that no standard ATLAS trigger was designed to see. A $pi_v$ decaying in the outer detector leaves roughly ten low-$p_T$ charged particles and about five $pi^0$s in a narrow cone, producing on average around 800 monitored drift tube hits, about three quarters of them from electromagnetic showers. The muon RoI cluster trigger fires on three or more level-1 muon regions of interest inside a $Delta R = 0.4$ cone, which selects decays between roughly 4 m and 7 m in radius with 30-50% efficiency; the vertex algorithm then groups single-chamber tracklets in $Delta R = 0.6$ cones and fits a common origin, reaching about 40% efficiency in the barrel ($4 le r le 7.5$ m) and in the forward region ($8 le |z| le 14$ m). Background is estimated fully from data through two measured probabilities, the chance that a random event contains a muon-spectrometer vertex and the chance of reconstructing a vertex given a trigger, so that cosmic, halo and noise contributions are folded in rather than modelled.

Core claim. The paper's central claim is a set of exclusions, not an observation. In 1.94 fb$^{-1}$ of $sqrt{s} = 7$ TeV data, no event contains two isolated, back-to-back vertices reconstructed in the ATLAS muon spectrometer, where the expected background from random vertex coincidences, cosmic showers, beam halo and detector noise is $0.03 pm 0.02$ events. Interpreting the null result in the Hidden Valley benchmark $h^0 to pi_v pi_v$ with 100% branching ratio, the collaboration excludes, at 95% confidence level, pseudoscalar proper decay lengths of $0.50 < ctau < 20.65$ m for $(m_{h^0}, m_{pi_v}) = (120, 20)$ GeV, $1.60 < ctau < 24.65$ m for $(120, 40)$ GeV, $0.45 < ctau < 15.8$ m for $(140, 20)$ GeV, and $1.10 < ctau < 26.75$ m for $(140, 40)$ GeV. The limits are quoted as a multiple of the Standard Model Higgs production cross section and, as the authors note, apply equally to models in which the Higgs decays to a pair of weakly-interacting scalars that themselves decay to heavy quark pairs.

The premise it rests on. The trigger and vertex efficiencies measured on punch-through jets are assumed to hold, to within the quoted 14-16% systematic uncertainties, for genuine long-lived-particle decays, even though a jet-induced shower and a boosted pseudoscalar decay need not populate the muon spectrometer in the same way.

Any model in which a 120-140 GeV Higgs decays with appreciable branching ratio to weakly-interacting particles with proper decay lengths between roughly half a metre and 27 m is now constrained, with the exact window depending on the masses involved.

The same limits transfer directly to Higgs decays into weakly-interacting scalars that subsequently decay to heavy quark pairs, extending the reach beyond the specific Hidden Valley pseudoscalar benchmark.

The techniques, a muon-region-of-interest cluster trigger and stand-alone muon-spectrometer vertexing, become reusable tools for any displaced-decay search reaching beyond the inner tracker.

The reach in lifetime is set by detector geometry rather than by luminosity, so modest additional data extends the sensitivity in cross section at fixed lifetime more than it extends the lifetime window.

Because the excluded lifetime window is bounded on both ends, the surviving parameter space for this benchmark sits at very short lifetimes, where the pseudoscalars decay before the calorimeter, and at lifetimes beyond tens of metres, where they mostly escape the detector; both ends are where a larger or denser detector, rather than more beam, would help most.

The calibration of trigger and vertex efficiency on punch-through jets assumes that a quark- or gluon-initiated shower mimics a boosted pseudoscalar decay in charged multiplicity, neutral fraction and shower content; a dedicated test comparing simulated signal-like clusters with jet-induced clusters at matched hit multiplicity would sharpen or break that assumption.

Since the background estimate is built from two measured probabilities rather than from simulation, the same machinery could be inverted into a search for an excess of single displaced vertices, which would be sensitive to models where only one of the two long-lived particles decays inside the spectrometer.

The null result sits in a region of parameter space that electroweak precision and flavour observables constrain only weakly, so the lifetime windows quoted here are likely to remain the primary direct handle on this topology until detectors with larger instrumented volume are available.

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review