The Pursuit of Dark Matter at Colliders - An Overview

This is a solid review of collider dark matter searches, and its value is exactly where the author says it is: the synthesis. Table 2, which lines up direct, indirect and collider sensitivity across vector, axial-vector, scalar and pseudoscalar couplings, is the kind of thing people actually photocopy, and Section 5's summary plots with the CMS/ATLAS numbers pulled together are genuinely handy. The physics it reports checks out against the sources: the Higgs-to-invisible combination of 0.24 observed and 0.23 expected matches Ref. [58], the GAMBIT neutralino bound of 2.5 TeV matches Ref. [176], the dijet reach to nearly 3 TeV at g_q = 0.25 matches Ref. [158]. Nothing here is circular, and there is no derivation to be circular about. The author is also honest about scope: the Fig. 22 caption says outright that the exclusions do not transfer to other couplings or models. That caveat is stated, so I won't press it as a hidden assumption. What I will press is that the benchmark the caveat refers to cannot be traced from inside the review. Section 2.2 carries a block with equation number 2.3, a figure labelled 'Figure 1', page markers '–4–', and citations [11, 12, 32] that in this bibliography land on Martin's SUSY primer, Feng's non-WIMP review, and Haisch–Pani–Polesello's CP study. None of those is the source of the MSDM width formula. The stress-test is right that this and the benchmark issue are one item, not two: the one place a reader would go to check the model behind Table 2 and Figs. 22–24 is the one place the document is not self-consistent. That is a traceability failure, not a physics error, but a review's product is accurate mapping, so it matters. The smaller ones are the same species: Section 2.3.3 credits the muon g-2 discrepancy to [66], a nucleosynthesis paper; the operator normalization in 2.1 is printed without the square on the mediator mass; the Fig. 8 caption cites [132] for both SIMPLE and COUPP. All fixable in a copy-edit pass. Who should read it: someone entering the subfield, or a neighbouring-field person who needs the complementarity argument in one place. Worth a referee, mainly to force the citation mapping to be cleaned up.

The central claim (abstract:s4) is a synthesis claim: colliders are a necessary, complementary part of a multi-pronged dark matter programme. For it to hold, the cross-method comparison must be made inside a common model, and the review must let a reader verify that model. That model is the leptophobic s-channel simplified model with g_q = 0.25, g_DM = 1.0, g_l = 0, and a mediator width computed from the other four parameters. The reader correctly names this benchmark as the weakest assumption and correctly flags the §2.2 block as an execution defect, but treats them as two separate items. They are one item. The benchmark's actual definition — the width formula (2.3), the g_l = 0 rationale, and the sentence 'the lepton couplings play no role (at tree level) in the phenomenology in either hadron collider and direct detection searches' — sits inside a passage carrying equation number 2.3, a figure labelled 'Figure 1', page markers '–4–', and citations [11, 12, 32] that in this bibliography resolve to Martin's supersymmetry primer, Feng's non-WIMP review, and Haisch–Pani–Polesello's CP-study of spin-0 mediators. None of those three is the source of the MSDM width formula. So the one place a reader would go to check the benchmark on which Table 2, Fig. 22, Fig. 23 and Fig. 24 all rest is the one place the document is not self-consistent. This is not a physics error: the benchmark is the standard DMWG choice and the numbers quoted in §5 match the CMS and ATLAS summary plots. It is a traceability failure, and it matters because the review's product is precisely accurate mapping. The paper is candid about the benchmark's scope in the Fig. 22 caption ('not applicable to other choices of coupling values or model'), so the caveat itself is acknowledged and I do not press it; what is not acknowledged is that the definition the caveat refers to cannot be traced from within the review.

The physics is sound and the reader is right that correctness_risk is low. The complementarity argument of §2.5 and the summary plots of §5 are consistent with the CMS and ATLAS public results they cite, and the paper is explicit in the Fig. 22 and Fig. 24 captions that the exclusions, relic-density contours and unitarity curves do not transfer to other coupling choices or models. That acknowledgement is genuine and I do not re-raise it as a hidden assumption. Where I part company with the reader is on bookkeeping. The reader lists the benchmark dependence as the weakest assumption and the §2.2 imported passage as an execution defect, then treats the execution defects as not touching the central claim. I think that separation is wrong in one specific respect: the imported block is where the benchmark is defined, so the defect sits directly on the load-bearing joint rather than beside it. The same pattern recurs elsewhere and corroborates it — §2.3.3 attributes the ~3.5σ muon anomalous magnetic moment discrepancy to [66], which is Pospelov's Big Bang nucleosynthesis catalysis paper and contains no g-2 result; the Fig. 8 caption cites [132] for both SIMPLE and COUPP; and the operator normalization in §2.1 is printed as 1/Λ² = g_q g_DM m_med, which is dimensionally inconsistent and should carry m_med² in the denominator. Each of these is individually minor, but together they show that the citation layer of this review is not reliable at exactly the points where a reader would check the model assumptions. None of this changes the verdict: the reader's CONDITIONAL is the right landing spot, and my concern reinforces rather than moves it. I would only ask that the revision treat the §2.2 block as a substantive fix — remap its citations and restore the missing derivation — rather than as copy-editing, because the benchmark that the entire quantitative half of the paper rests on is defined there and nowhere else.

The review is timely and useful. The comparative summary figures, which place collider, direct-detection, and dijet limits on common axes, are a genuine service to the field, and Table 2's breakdown by interaction type is a helpful organising device. The paper is also honest about model dependence: the Fig. 22 caption explicitly states that the exclusions are not applicable to other coupling choices, and §6 acknowledges that present models are among the simplest available. The paper reports no new results; its value is as a synthesis and reference. That value depends on the reader being able to verify the benchmark model, which is where the principal problem lies.

This review argues that particle colliders are not an optional extra in the dark matter hunt but a necessary third approach alongside underground scattering experiments and astrophysical annihilation searches, because a collider can produce dark matter in the laboratory and can do so through all four basic interaction types at once. The author's case is that colliders cover the low-mass window of a few GeV that current direct-detection experiments cannot reach, carry systematic uncertainties of a completely different origin from the astrophysical ones that dominate the other two methods, and therefore add information rather than duplicating it. The review surveys the machinery that makes this possible: the missing-transverse-momentum signature, the mono-$X$ searches built on it, dijet resonance searches for the mediator, Higgs-to-invisible decays, and the low-energy and beam-dump experiments that reach where the LHC cannot. It then collects the current exclusions, which for vector and axial-vector mediators extend to roughly 2 TeV in mediator mass and about 1 TeV in dark matter mass with 36 fb$^{-1}$ of 13 TeV data, stretching to about 2.5 TeV when dijet results are reinterpreted. The reader should care because the argument fixes where the next decade of dark matter effort ought to go, and because it says what a collider could still deliver that no other experiment can.

What carries the argument. The load-bearing construct is the mono-$X$ signature combined with simplified models. Because the colliding protons carry no net momentum transverse to the beam, the vector sum of all visible transverse energy must vanish; anything that escapes leaves an imbalance $E_T^{miss} = -sum_i E_T^{i}$ that is the primary dark matter handle. If a dark matter pair is produced and recoils against a hard photon, gluon, gauge boson or Higgs radiated in the initial state, the visible object and the invisible pair point in nearly opposite directions, giving $E_T^{miss} + X$ with $X$ a single high-momentum standard-model particle. Simplified models replace the older effective-field-theory description, which breaks down once the momentum transfer approaches the mediator mass, with an explicit mediator carrying four parameters: the mediator mass, the dark matter mass, and the couplings $g_q$ and $g_{DM}$. The same four parameters predict the direct-detection scattering rate, which is what makes the comparison between collider and underground limits possible at all. Dijet resonance searches supply the complementary handle on the mediator itself when it decays visibly, and the Higgs portal supplies a third, where the Higgs boson is the mediator and the signature is an invisible Higgs branching fraction.

Core claim. The claim the paper sets out to establish is a claim about strategy rather than a measurement: collider searches are an essential component of an interdisciplinary dark matter programme, not a secondary one. The supporting structure is a survey of what has actually been done. Dark matter produced at a collider escapes the detector and shows up as missing transverse momentum, $E_T^{miss}$, balanced against a visible photon, jet, $W$, $Z$, Higgs boson or heavy quark, giving the mono-$X$ signature; the same models predict a mediator that can decay back to quarks and appear as a narrow bump in the dijet invariant mass. The author reports that no significant excess has been seen, and that with 36 fb$^{-1}$ of 13 TeV data vector and axial-vector mediators are excluded up to almost 2 TeV in mediator mass and up to almost 1 TeV in dark matter mass, with dijet reinterpretation pushing the mediator reach to about 2.5 TeV largely independent of dark matter mass. For spin-0 mediators the searches are just beginning to bite below about 500 GeV. The comparison the paper draws is that colliders win at low dark matter mass and in spin-dependent scattering, while direct and indirect searches win at intermediate and large masses for spin-independent interactions, so the three approaches cover different regions of the same parameter space and all three are needed.

The premise it rests on. Every ranking and every exclusion contour in the review is computed for one benchmark scenario, a single mediator exchanged in the s-channel with quark coupling 0.25, dark matter coupling 1.0 and no couplings to leptons, and the paper itself warns that the contours and relic-density lines do not carry over to other coupling choices or other models.

If the argument holds, a dark matter signal seen first underground or in cosmic rays can be followed up at a collider, where the mediator's mass, couplings and decay modes could be measured rather than merely bounded.

The low-mass window of a few GeV stays a collider specialty, so the next generation of light-mediator and beam-dump experiments inherits the part of parameter space that direct detection cannot reach.

Spin-dependent scattering remains a place where collider limits beat direct-detection limits at larger dark matter masses, because direct searches lose the coherent enhancement over nucleons there.

The absence of a signal so far constrains the simplest single-mediator scenarios but leaves most of the LHC programme's data unanalysed, so the reach quoted in the review is a floor rather than a ceiling.

A future higher-energy collider becomes the natural next step if dark matter sits above the reach of the current machine, since only production can access it directly.

The review's comparative tables are only as stable as the benchmark coupling choice behind them, so a reader who wants to know which experiment is 'winning' should treat the ranking as conditional rather than settled.

Because collider searches are generically sensitive to the mediator rather than the dark matter particle, the field's near-term discovery potential is arguably better described as a search for a new force carrier that happens to connect to dark matter.

The same missing-momentum machinery that carries the dark matter searches also carries long-lived-particle and dark-sector searches, so improvements in triggering and reconstruction for one programme transfer directly to the others.

If the dark sector contains more than one mediator, the mono-$X$ and dijet channels stop being independent probes of a single coupling and start constraining a network, which would make the current exclusion plots systematically over-confident.

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