betaFeedback

WorksPhysics and Astronomy

Constraints on quirks from a Geant4-based reinterpretation of the CMS search for heavy stable charged particles

Abstract

Quirks are new heavy fermions with electroweak charges that are also bound by a confining force whose scale ΛΛ is far below their mass. A produced pair therefore stays connected by a macroscopic string. We reinterpret the CMS search for heavy stable charged particles in 101 fb−1^{-1} of 13 TeV proton-proton collisions for Drell-Yan produced quirks. The quirk dynamics is simulated with Geant4 in a parametric CMS-like detector, and the event selection is emulated and combined with the public likelihoods. The emulation reproduces the published limits on a heavy charged lepton within 11%. For an SU(2)SU(2)-doublet quirk with one (two) infracolors, masses up to 870 (970) GeV are excluded for Λ≲70Λ\lesssim70 eV, where the quirks behave as independent heavy stable charged particles. For Λ≳10Λ\gtrsim10 keV, where the pair moves as one doubly ionizing object, the limit is 280 (320) GeV. In between, the quirks turn around inside the detector and the search has no sensitivity. For the SU(2)SU(2)-singlet benchmark of the FASER search, with two infracolors, the limits reach 990 GeV below 100 eV, where FASER has no sensitivity, and 300 GeV above 10 keV, beyond the FASER exclusion of 175 GeV.

Introduction

Quirks [1, 2] are new heavy fermions QQ with electroweak charges that are also charged under a confining gauge group, infracolor, whose confinement scale Λ\Lambda is much smaller than the quirk mass mQm_Q, as in folded supersymmetry [3]. Since no infracolored particle is lighter than Λ\Lambda, the infracolor flux tube between a pair of quirks cannot break: a quirk-antiquirk pair produced at a collider stays connected by a string of tension σ≃Λ2\sigma\simeq\Lambda^2 and oscillates around its center of mass until it annihilates. The amplitude of the oscillation is set by the kinetic energy of the quirks in the rest frame of the pair,

dmax⁡≃2(E∗−mQ)Λ2d_{\max} \simeq\frac{2(E^* - m_Q)}{\Lambda^2}

where E∗E^* is the energy of each quirk in that frame. For E∗−mQE^* - m_Q of order 100 GeV, the pair is meters wide at Λ≃100\Lambda\simeq100 eV, centimeters wide at 1 keV, and narrower than the pitch of a silicon detector at a few tens of keV (Table I).

Λ\LambdaΛ2/ℏc\Lambda^2/\hbar c [GeV/m]dmax⁡d_{\max}Signature
1 eV0.00540 kmindependent HSCPs
30 eV4.644 mHSCPs, deflected tracks
100 eV514 mturn in the calorimeters
300 eV46044 cmturn in the tracker
1 keV5×1035 \times10^34 cmcm-wide pair
3 keV5×1045 \times10^44 mmmm-wide pair
10 keV5×1055 \times10^50.4 mmpartly merged
30 keV5×1065 \times10^644 μ\mummerged, 2×\times ionization

Table I. String tension and largest separation of a quirk pair with E∗−mQ=100E^* - m_Q = 100 GeV, and the resulting detector signature.

At a general-purpose Large Hadron Collider (LHC) detector, quirks therefore appear as heavy stable charged particles (HSCPs) for Λ\Lambda of a few eV, as pairs that turn around inside the detector for Λ\Lambda between about 100 eV and 1 keV, and as one neutral object with twice the ionization of a single charge above a few keV. Reinterpretations of HSCP and monojet searches constrain Λ\Lambda of about 1 to 100 eV [4], the D0 experiment searched for quirks with Λ\Lambda from 10 keV to 1 MeV [5], and dedicated strategies target slow pairs whose tracks do not bend [6], the ionization of quirks in the tracker [7], quirks that stop in the detector [8], and the LHCb [9] and FASER [10, 11, 12] detectors.

In this paper, we derive constraints on Drell-Yan (DY) produced quirks from the CMS search for HSCPs with large ionization energy loss in 2017 and 2018 data [13, 14], using only public information: the HEPData [15] record of the search [16], which contains the likelihoods (statistical models in the Combine format [17]), the selection efficiencies, and the trigger efficiency maps. The quirk propagation is simulated with a standalone Geant4 [18, 19, 20] application in a parametric CMS-like detector. The detector response and the event selection are emulated, and the emulation is calibrated and validated by reproducing the published limits on a heavy charged lepton τ′\tau' with charge ∣Q∣=1e|Q| = 1e.

Quirk production and dynamics

We consider a vector-like quirk with the electroweak quantum numbers of a fourth-generation lepton doublet partner (a τ′\tau' with ∣Q∣=1e|Q| = 1e, hypercharge as in the Standard Model), produced in pairs through qqˉ→γ∗/Z→QQˉq\bar{q} \to\gamma^*/Z \to Q\bar{Q}. Events are generated with PYTHIA 8.3 [21] at s=13\sqrt{s} = 13 TeV with the CP5 tune [22] and the NNPDF3.1 parton distributions [23], using the leading order (LO) cross section; with NICN_{\mathrm{IC}} infracolors the cross section is NICN_{\mathrm{IC}} times larger and the kinematics are unchanged. The cross section falls from 0.126 pb at mQ=200m_Q = 200 GeV to 7.2×10−67.2 \times10^{-6} pb at 1.4 TeV. Except at the lowest Λ\Lambda, the string tension (Table I) dominates every other force in the detector, already forty times the magnetic force on a relativistic unit charge in 3.8 T at Λ=100\Lambda= 100 eV, so the quirk transport has to be solved together with the interactions with matter.

Simulation

The quirk propagation uses quirksim (Appendix A), a standalone Geant4 application built on the same quirk transportation that the FASER search used for its detector simulation [12]. Each quirk is transported in short steps under the combined Lorentz and string forces, in alternation with its partner: a quirk is suspended after each step and resumed once its partner has caught up, so that the string force is always computed from the partner trajectory at the same time. Ionization, multiple scattering, and bremsstrahlung of the quirks are simulated with the standard electromagnetic models of Geant4 version 11.4.1. Being color neutral, the quirks have no hadronic interactions.

The detector is a set of cylinders and disks with the radii, thicknesses, materials, and magnetic field of the CMS detector [24], from the pixel detector to the muon stations, with 3.8 T inside the solenoid and 1.8 T in the return yoke. The double-sided strip modules of the first two inner and outer barrel layers are modeled as one sensor of twice the thickness. Varying the tracker material by ±50%\pm50\% or the calorimeter absorber by −20%-20\% changes the acceptance observables of quirk pairs by less than the statistical precision of 200-event samples.

Only the quirks are passed to Geant4, and only events that can pass the selection, as determined at the generator level, are simulated; the fraction kept enters the efficiency.

Emulated reconstruction and selection

The Geant4 hits are turned into clusters: hits of the two quirks within twice the pitch in a sensor form one cluster with the summed energy and the energy-weighted position, and hits more than 25 ns later than a particle at the speed of light are discarded. A track candidate in which at least half of the clusters contain both quirks is treated as one merged track of the pair. Each candidate is fitted with a circle in the transverse plane and a line in (r,z)(r,z), with the CMS hit resolutions and a multiple scattering term, which gives pTp_{\mathrm{T}}, χ2/dof\chi^2/\mathrm{dof}, the impact parameters dxyd_{xy} and dzd_z, and σpT/pT2\sigma_{p_{\mathrm{T}}}/p_{\mathrm{T}}^2.

The energy loss per unit path length dE/dxdE/dx of each cluster is smeared by an effective resolution and compared to the distribution of a minimum ionizing particle (MIP), obtained in the same detector from charged particles at βγ=70\beta\gamma= 70 with the same smearing, separately for each sensor thickness. The probability PiP_i that a MIP gives a smaller energy loss than cluster ii enters the two discriminants of the search: the strip discriminant

GiStrips=3N[112N+∑i=1NPi(Pi−2i−12N)2],(1)G_i^{\mathrm{Strips}} = \frac{3}{N}\left[\frac{1}{12N}+\sum_{i=1}^{N}P_i\left(P_i-\frac{2i-1}{2N}\right)^2\right], \tag*{(1)}

with the strip clusters ordered by increasing PiP_i, and the pixel discriminant

FiPixels=1−Π∑k=0N−1(−ln⁡Π)kk!,(2)F_i^{\mathrm{Pixels}} = 1-\Pi\sum_{k=0}^{N-1}\frac{(-\ln\Pi)^k}{k!}, \tag*{(2)}

with Π=∏i(1−Pi)\Pi=\prod_i(1-P_i) over the NN pixel clusters outside the innermost layer, which is one minus the combined probability that a MIP gives a larger energy loss than every cluster. The effective resolutions, 40% for pixels and 15% for strips, are the only tuned parameters of the emulation: they are fixed by requiring that the fraction of the τ′\tau' signal in the FiPixels<0.9F_i^{\mathrm{Pixels}} < 0.9 (FAIL) region and in the last GiStripsG_i^{\mathrm{Strips}} bin of the PASS region match the published likelihoods (§ 2.3).

The relative mini-isolation and the track isolation are computed from the generator-level visible particles around the candidate direction, with the partner quirk added when it is inside the cone and not part of the candidate. The energy-to-momentum ratio adds the energy deposited by the quirks themselves in the calorimeters. The single-muon trigger is applied as an efficiency per quirk that reaches at least two muon stations, from the published maps for τ\tau leptons as a function of β\beta in six ∣η∣|\eta| bins [16]. For quirks, β\beta is the effective velocity to the first muon station (distance from the origin over arrival time), which coincides with the true β\beta for a free HSCP to better than 1%1\% above β=0.4\beta=0.4, and two quirks that arrive within ΔR<0.1\Delta R<0.1 count as one muon object. The up and down variations of the maps, which correspond to a 1.5 ns shift of the muon arrival time, provide the trigger uncertainty.

The candidate must pass the preselection of the search: pT>55p_{\mathrm{T}}>55 GeV, ∣η∣<1|\eta|<1, at least two pixel hits outside the innermost layer, more than 80% valid hits, at least ten dE/dxdE/dx measurements, χ2/dof<5\chi^2/\mathrm{dof}<5, dz<0.1d_z<0.1 cm, dxy<0.02d_{xy}<0.02 cm, mini-isolation below 0.02, track isolation below 15 GeV, E/p<0.3E/p<0.3, σpT/pT2<0.0008\sigma_{p_{\mathrm{T}}}/p_{\mathrm{T}}^2<0.0008 GeV−1^{-1}, and FiPixels>0.3F_i^{\mathrm{Pixels}}>0.3. The signal region requires pT>200p_{\mathrm{T}}>200 GeV and is binned in GiStripsG_i^{\mathrm{Strips}} (nine bins) separately for FiPixels>0.9F_i^{\mathrm{Pixels}}>0.9 (PASS) and 0.3<FiPixels<0.90.3<F_i^{\mathrm{Pixels}}<0.9 (FAIL). The emulated signal yields replace the signal templates of the published likelihoods, which are otherwise used unchanged: a binned PASS/FAIL model in which the background is constrained by the FAIL region. The signal shape uncertainties of the published model are replaced by an overall uncertainty of 15% from the closure test below, the trigger uncertainty, the MC statistical uncertainty of the emulated signal, and the 1.6% integrated luminosity uncertainty. Points with fewer than three effective MC events in the signal region are not used. Upper limits at 95% confidence level are computed with the CLs\mathrm{CL}_{s} criterion [25, 26] in the asymptotic approximation [27] with a standalone build of Combine [17].

Validation with a heavy charged lepton

The ∣Q∣=1e τ′|Q| = 1e\ \tau^\prime of the CMS search is what a quirk becomes when the string is negligible, and its published likelihoods contain the signal templates, so it fixes both the calibration and the precision of the emulation. We generate it with the generator settings of the search (published with the HEPData record), simulate it with the same chain as the quirks with the string switched off in practice (Λ=1 eV\Lambda= 1\ \mathrm{eV}, a string amplitude of tens of kilometers), and treat the published signal yields as the target.

Figure 1 compares the emulated and published 95% confidence level upper limits on the τ′\tau^\prime cross section, both computed on the published background model. The emulated limits agree with the published ones within 11% from 0.8 to 2.2 TeV, for both the observed and the expected limits. The total signal efficiency agrees within 13% up to 1.8 TeV, and the split between the FAIL and PASS regions and the fraction in the most sensitive bin (FiPixels>0.9F_i^{\mathrm{Pixels}} > 0.9, GiStrips>0.4G_i^{\mathrm{Strips}} > 0.4) agree within 0.03 and 0.06 (Figure 2). Above 1.8 TeV the emulation underestimates the FAIL fraction, most likely because the response of the CMS pixel detector to very slow particles, such as late charge collection and readout saturation, is not modeled; this mass range lies beyond the quirk masses considered here, and the limits still agree within 9%. Normalized to the events with a candidate within ∣η∣<1|\eta| < 1, the published τ′\tau^\prime cutflow is reproduced step by step within 5%: the full preselection keeps 92 to 97% of these events in CMS and 97 to 99% in the emulation, with the largest difference, 2 to 3%, at the pixel-hit requirement, since the modeled pixel detector has no inactive regions. We assign a 15% uncertainty to the emulated signal efficiency to cover these differences.

Figure 1. Upper limits at 95% confidence level on the cross section of a ∣Q∣=1e τ′|Q| = 1e\ \tau^\prime: published observed (filled points) and expected (open points, with 68% band) limits, and the limits with the emulated signal on the same background model (lines, band: MC statistical uncertainty). The lower panel shows the emulated-to-published ratio.

Figure 2. Total signal efficiency of the ∣Q∣=1e τ′|Q| = 1e\ \tau^\prime and its fractions in the FAIL region and in the most sensitive bin, from the published likelihood (points) and the emulation (lines, band: MC statistical uncertainty). The lower panel shows the published-to-emulated ratio; the gray band is the 15% uncertainty assigned to the emulated efficiency.

Quirk signatures

Table I summarizes the scales that decide the detector signature, and Figure 3 shows the same quirk pair simulated at three values of Λ\Lambda.

Figure 3. Transverse view of the trajectories of a quirk pair with mQ=200 GeVm_Q = 200\,\mathrm{GeV} and momenta (300,50,100)(300, 50, 100) and (250,−40,150) GeV(250, -40, 150)\,\mathrm{GeV} in the CMS-like detector at Λ=30 eV\Lambda= 30\,\mathrm{eV} (left), 100 eV100\,\mathrm{eV} (middle), and 300 eV300\,\mathrm{eV} (right). The dashed circles mark the outer radii of the tracker, the electromagnetic and hadron calorimeters, the solenoid, and the muon system.

Figure 4 shows how the trigger and preselection efficiency changes with Λ\Lambda. At Λ=1 eV\Lambda= 1\,\mathrm{eV}, 0.40 of the generated pairs with mQ=600 GeVm_Q = 600\,\mathrm{GeV} pass the trigger and the preselection, a value close to that of the τ′\tau^\prime at similar mass; the remaining difference comes from the slower and more forward kinematics of the doublet quirk. The efficiency falls steeply between 30 and 100 eV (0.32 and 0.07 at 600 GeV600\,\mathrm{GeV}), long before the quirks turn around inside the tracker. The trigger is not the reason: the quirks still reach the muon system with a similar efficiency (0.48 and 0.31). The selection step that removes the events is the track χ2/dof<5\chi^2/\mathrm{dof} < 5 requirement: already at Λ=30 eV\Lambda= 30\,\mathrm{eV} the string tension, 4.6 GeV/m4.6\,\mathrm{GeV}/\mathrm{m}, exceeds the magnetic force on a unit charge, and whenever the string is not aligned with the quirk momentum (a boosted pair, or a pair recoiling against initial-state radiation) it deflects the trajectory transversely by more than the hit resolution over the length of the tracker.

Figure 4. Fraction of generated quirk pairs that pass the trigger and the preselection, as a function of Λ\Lambda for four quirk masses.

From 300 eV300\,\mathrm{eV} to 1 keV1\,\mathrm{keV} no event passes the preselection: the quirks are centimeters to tens of centimeters apart inside the tracker, so each leaves an oscillating sequence of hits that no helix fit accepts, while their clusters are too far apart to merge. From 11 to 3 keV3\,\mathrm{keV} the lightest pairs, up to about 150 GeV150\,\mathrm{GeV}, start to pass as merged tracks, and at 1010 and 30 keV30\,\mathrm{keV} most selected candidates at all masses are merged tracks of the pair, with GiStripsG_i^{\mathrm{Strips}} close to one. Their acceptance is limited by the trigger, since the pair is heavy and slow (β\beta between 0.2 and 0.6 for most central pairs, where the muon trigger efficiency is small), and by χ2/dof<5\chi^2/\mathrm{dof} < 5, since the energy-weighted cluster positions still follow the residual oscillation of the pair.

Results

Figure 5 shows the GiStripsG_i^{\mathrm{Strips}} distribution of the PASS region (pT>200 GeVp_T > 200\,\mathrm{GeV}) from the CMS search, observed and predicted, with the emulated signal of four quirk benchmarks. In every regime, the quirk signal is concentrated in the last bin, GiStrips>0.4G_i^{\mathrm{Strips}} > 0.4, where one event is observed for a predicted background of 0.43 events; the residual signal at lower GiStripsG_i^{\mathrm{Strips}} comes from fast quirks at small Λ\Lambda.

Figure 5. Distribution of GiStripsG_i^{\mathrm{Strips}} in the PASS region (FiPixels>0.9F_i^{\mathrm{Pixels}} > 0.9, pT>200p_T > 200 GeV) of the CMS search: observed events (points) and predicted background (gray), from the HEPData record [16], with the emulated signal of an SU(2)SU(2)-doublet quirk (NIC=1N_{\mathrm{IC}} = 1, LO cross section) for four values of (mQ,Λ)(m_Q, \Lambda).

Figure 6 shows the upper limit on the quirk pair cross section divided by the LO cross section for NIC=1N_{\mathrm{IC}} = 1, for each Λ\Lambda. Below about 150 GeV150\,\mathrm{GeV} the limits are not monotonic in mass: a free quirk with pT>200 GeVp_T > 200\,\mathrm{GeV} becomes

Figure 6. Observed upper limit at 95% confidence level on the DY quirk pair cross section, divided by the LO cross section of an SU(2)SU(2)-doublet quirk with NIC=1N_{\mathrm{IC}} = 1, as a function of the quirk mass for several values of Λ\Lambda. Values below one are excluded; for NICN_{\mathrm{IC}} infracolors the threshold is NICN_{\mathrm{IC}}.

too fast (β≈0.97\beta\approx0.97 at 75 GeV75\,\mathrm{GeV}) to ionize much more than a minimum ionizing particle, while the cross section keeps rising, so that σ95/σ\sigma_{95}/\sigma peaks between 50 and 100 GeV, at a mass that depends on Λ\Lambda and is affected by the limited MC statistics in this region. Figure 7 translates these limits into the excluded region in the (mQ,Λ)(m_Q, \Lambda) plane for NIC=1N_{\mathrm{IC}} = 1, 2, and 3, interpolating log⁡(σ95/σ)\log(\sigma_{95}/\sigma) linearly in mass and in log⁡Λ\log\Lambda between grid points and smoothing it mildly against the statistical scatter of the simulated points.

Figure 7. Excluded region in the plane of the quirk mass and the confinement scale Λ\Lambda (shaded: excluded for NIC=1N_{\mathrm{IC}} = 1; hatched: 68% band of the expected exclusion; the gray region is not simulated): observed for NIC=1N_{\mathrm{IC}} = 1, 2, and 3 (black lines) and expected with its 68% band for NIC=1N_{\mathrm{IC}} = 1.

For NIC=1N_{\mathrm{IC}} = 1, quirk pairs are excluded from mQ=140m_Q = 140 to 870 GeV (expected 910 GeV) for Λ=1 eV\Lambda= 1\,\mathrm{eV}, where the quirks are free HSCPs, 830 GeV at 30 eV30\,\mathrm{eV}, 740 GeV at 50 eV50\,\mathrm{eV}, and 620 GeV at 70 eV70\,\mathrm{eV}. From 10 to 50 eV the string force, 0.5 GeV/m at 10 eV, is comparable to the magnetic force on a slow quirk and can straighten its track, so that slow, highly ionizing quirks (β≈0.4\beta\approx0.4) are reconstructed with pT>200p_T > 200 GeV and the exclusion reaches down to 50 GeV. For Λ\Lambda from 100 to 800 eV, the search has no sensitivity above 40 GeV. For Λ\Lambda above 1 keV, where the pair is increasingly reconstructed as one merged track, the excluded mass grows from 130 GeV at 2 keV to 280 GeV at 30 keV. For NIC=2N_{\mathrm{IC}} = 2 (3), the limits at 1 eV and 30 keV rise to 970 (1030) and 320 (350) GeV.

For a direct comparison with the FASER search [12], we repeat the analysis for an SU(2)SU(2)-singlet quirk with hypercharge −1-1 and NIC=2N_{\mathrm{IC}} = 2, the benchmark of that search. Its couplings to the photon and the ZZ boson are purely vector, the same as those of the τ′\tau' model of the CMS search, so it is generated with the generator settings of the τ′\tau' and simulated and analyzed in the same way as the doublet. FASER generated the same Drell-Yan process at LO with MadGraph5_aMC@NLO [28] interfaced to Pythia 8 (with an LHCb tune and a forward-ZZ normalization correction) at s=13.6\sqrt{s} = 13.6 TeV, while we use the LO Pythia 8 matrix element at 13 TeV; both are LO with NIC=2N_{\mathrm{IC}} = 2. Figure 8 compares the excluded masses. From 4 keV on, where the pair is reconstructed as one merged track, the CMS search excludes the singlet quirk up to 280 GeV at Λ=10\Lambda= 10 keV and 300 GeV at 30 keV, beyond the FASER exclusion of 100 to 175 GeV in this range. From 1 to 3 keV the two searches overlap: FASER excludes masses up to 110 to 160 GeV, this recast up to 50 GeV at 1 to 1.5 keV and 145 to 155 GeV at 2 to 3 keV. From 300 to 600 eV, where the quirk pair is too wide to be reconstructed as a track, FASER provides the only constraint. Quirk masses below mZ/2m_Z/2 are in any case excluded by the ZZ width measured at LEP [29]; we simulate masses down to 30 GeV only to cover the mass range of FASER. Below 100 eV, outside the range covered by FASER, the CMS search excludes the singlet up to 990 GeV at Λ=1\Lambda= 1 eV and 720 GeV at 70 eV.

Figure 8. Excluded region of an SU(2)SU(2)-singlet quirk with NIC=2N_{\mathrm{IC}}=2 in the (mQ,Λ)(m_Q,\Lambda) plane (shaded: excluded), from this reinterpretation of the CMS search at s=13\sqrt{s}=13 TeV (blue) and from the FASER search at s=13.6\sqrt{s}=13.6 TeV [12] (orange): observed (solid lines) and expected (dashed) with its 68% band (hatched). The gray region is not simulated in this work.

Discussion

For the SU(2)SU(2)-singlet quirk with two infracolors and the same definition of Λ\Lambda, Ref. [4] excluded masses up to 650 GeV at 1 eV and 150 GeV at 100 eV with 12.9 fb−1^{-1} of 2016 data, a looser selection (pT>65p_T > 65 GeV, ∣η∣<2.1|\eta| < 2.1), and a track model with 30 μ\mum hit resolution and without multiple scattering or energy loss in matter. With 101 fb−1^{-1} the exclusion at Λ=1\Lambda= 1 eV extends to 990 GeV.

The main limitations of this reinterpretation are the following. The trigger efficiency of a bound pair is taken from the efficiency maps of a single slow charged particle, evaluated at the effective velocity of the pair to the muon stations. The ionization response is a parametrization tuned on the published heavy charged lepton likelihood, not a simulation of the CMS silicon readout; saturation of the strip and pixel readout, which matters for charges above 1e, is not modeled, and a bound pair deposits the ionization of a charge 2e only in the sense of the total energy loss of two unit charges along the same path. The isolation uses the generator-level event without pileup, a small effect for a heavy charged lepton according to the published efficiency table. Photon-initiated production, which adds to the DY cross section at high mass, is not included, so the limits are conservative in this respect, as they are with respect to higher-order corrections to the DY cross section.

The approach generalizes to other signatures in which the propagation through the detector, rather than the production, is non-standard: the same chain of a standalone Geant4 simulation in a parametric detector, an emulation calibrated on a benchmark of the original search, and the published likelihoods can be applied whenever a search publishes its statistical model and the signal templates of at least one benchmark.

Summary

We have reinterpreted the CMS search for heavy stable charged particles with large ionization energy loss for Drell-Yan produced quirks, with a standalone Geant4 simulation of the quirk dynamics, an emulation that reproduces the published heavy charged lepton limits within 11%, and the public likelihoods of the search. For an SU(2)SU(2)-doublet quirk with one (two) infracolors, masses are excluded up to 870 (970) GeV at Λ=1\Lambda= 1 eV, 620 (730) GeV at 70 eV, and 260 to 270 (290 to 320) GeV for Λ\Lambda of 10 to 30 keV, where the pair moves as one merged, doubly ionizing track, while for Λ\Lambda from 100 to 800 eV the search has no sensitivity. For the SU(2)SU(2)-singlet benchmark of FASER, the reinterpretation alone covers Λ\Lambda below 100 eV, up to 990 GeV; above 4 keV it excludes up to 280 to 300 GeV, beyond the 100 to 175 GeV of FASER, while FASER gives the stronger constraint around 1 keV and the only one from 300 to 600 eV.

Code and data availability

The quirk simulation (quirkcore and quirksim), the emulation of the CMS selection, the statistical analysis, and the scripts that produce every figure of this paper are available at https://github.com/tvamii/QuirkPheno. The public likelihoods and efficiency maps of the CMS search are taken from HEPData [16] without modification.

Acknowledgments

We would like to thank Joe Incandela for useful discussions. We used the UCSB computational facilities administered by the Center for Scientific Computing at the California NanoSystems Institute and Materials Research Laboratory (an NSF MRSEC; DMR-1720256) and purchased through NSF CNS-1725797. UCSB is supported by the US Department of Energy under grant DE-SC0011702. Support for UCSB is also made possible by the Joe and Pat Yzurdiaga endowed chair in experimental science. We acknowledge the use of Anthropic’s Claude Opus 5.5 model for code development and manuscript editing. All outputs were checked by the authors who take full responsibility for the content.

References

References

  1. [1]L. B. Okun, Nucl. Phys. B 173, 1 (1980).
  2. [2]J. Kang and M. A. Luty, JHEP 11, 065 (2009), 0805.4642.
  3. [3]G. Burdman, Z. Chacko, H.-S. Goh, R. Harnik, and C. A. Krenke, Phys. Rev. D 78, 075028 (2008), 0805.4667.
  4. [4]M. Farina and M. Low, Phys. Rev. Lett. 119, 111801 (2017), 1703.00912.
  5. [5]D0 Collaboration, Phys. Rev. Lett. 105, 211803 (2010), 1008.3547.
  6. [6]S. Knapen, H. K. Lou, M. Papucci, and J. Setford, Phys. Rev. D 96, 115015 (2017), 1708.02243.
  7. [7]J. Li, T. Li, J. Pei, and W. Zhang, Phys. Rev. D 102, 056006 (2020), 1911.02223.
  8. [8]J. A. Evans and M. A. Luty, JHEP 06, 090 (2019), 1811.08903.
  9. [9]X. Cid Vidal, M. Fernández Gómez, M. Low, A. Novo Cal, Y. Tsai, and C. Vázquez Sierra, JHEP 06, 037 (2026), 2601.09023.
  10. [10]J. Li, J. Pei, L. Ran, and W. Zhang, JHEP 12, 109 (2021), 2108.06748.
  11. [11]J. L. Feng, J. Li, X. Liao, J. Ni, and J. Pei, JHEP 06, 197 (2024), 2404.13814.
  12. [12]FASER Collaboration, (2026), accepted by Phys. Rev. Lett., 2607.26195.
  13. [13]CMS Collaboration, JHEP 04, 109 (2025), 2410.09164.
  14. [14]T. A. Vami, J. Subatomic Part. Cosmol. , 100029 (2025).
  15. [15]E. Maguire, L. Heinrich, and G. Watt, J. Phys. Conf. Ser. 898, 102006 (2017), 1704.05473.
  16. [16]CMS Collaboration, “HEPData record for CMS-EXO-200,” https://www.hepdata.net/record/ins2840007 (2024), version 1.
  17. [17]CMS Collaboration, Comput. Softw. Big Sci. 8, 19 (2024), 2404.06614.
  18. [18]S. Agostinelli et al. (GEANT4), Nucl. Instrum. Meth. A 506, 250 (2003).
  19. [19]J. Allison et al., IEEE Trans. Nucl. Sci. 53, 270 (2006).
  20. [20]J. Allison et al., Nucl. Instrum. Meth. A 835, 186 (2016).
  21. [21]C. Bierlich et al., SciPost Phys. Codeb. 2022, 8 (2022), 2203.11601.
  22. [22]CMS Collaboration, Eur. Phys. J. C 80, 4 (2020), 1903.12179.
  23. [23]R. D. Ball et al. (NNPDF), Eur. Phys. J. C 77, 663 (2017), 1706.00428.
  24. [24]CMS Collaboration, JINST 3, S08004 (2008).
  25. [25]T. Junk, Nucl. Instrum. Meth. A 434, 435 (1999), hep-ex/9902006.
  26. [26]A. L. Read, J. Phys. G 28, 2693 (2002).
  27. [27]G. Cowan, K. Cranmer, E. Gross, and O. Vitells, Eur. Phys. J. C 71, 1554 (2011), 1007.1727.
  28. [28]J. A. Allwall et al., JHEP 07, 079 (2014), 1405.0301.
  29. [29]ALEPH, DELPHI, L3, OPAL, SLD Collaborations, LEP Electroweak Working Group, SLD Electroweak and Heavy Flavour Groups, Phys. Rept. 427, 257 (2006), hep-ex/0509008.

Paper details

vlnc:W_ya7gt2v9
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Claude Opus 5.5

View statement
Looking for related papers…
Contents