Dark Force Detection in Low Energy e-p Collisions

The useful thing here is a concrete, fully worked reach estimate for e-p -> e-p X -> e-p e+e- at a 100 MeV electron beam on a hydrogen gas target. Signal and background are computed from first principles, cross-checked against CompHEP and MadGraph, and the narrow-width approximation is justified by the eV-scale width. The matrix element method in App. D is a clean variational argument, and the comparison to e-e-, e+e-, and e+e- -> gamma X in Sec. 5 is the kind of thing that actually helps an experimentalist decide where to put beam time. The displaced-vertex analysis in Sec. 7 is a sensible complement. The soft spot is the one the paper half-names: the 1 ab⁻¹ number is S/sqrt(B)=5 with no systematic errors. At m_X ~ 50 MeV the background in a 1 MeV bin is ~10⁹-10¹⁰ events, so a 5-sigma excess needs the background known to ~10⁻⁵. The sideband fit in Eq. (6.3) is an ad hoc form, and Fig. 13 fits pseudo-data generated from the same model, a closure test rather than a demonstration that a real QED background can be controlled at that level. The authors say the background would have to be fit from data and that a trials factor applies, so they are not hiding it, but they do not quantify how the reach degrades once systematics enter. That is the number a referee should push on. Two smaller things. Eq. (4.3) is inverted: it says L(x MeV) = (1/x) L(1 MeV), but worse resolution should need more luminosity, and the surrounding text (improves linearly) implies L proportional to bin width. Typo-level, but it is the stated extrapolation. And the 10¹⁹ cm⁻² target thickness is a private communication, not a published measurement; the instantaneous luminosity claim rides on it. None of this undermines the physics. The calculation is sound, the idealization is stated, and the paper is honest about what it is: a reach estimate, not a discovery claim. Who should read it: anyone thinking about light dark bosons in the 10-100 MeV window, and especially anyone involved in the JLab FEL proposal. It is a useful reference for reach numbers and for the matrix element method. Recommendation: send it out. It deserves referee time, mainly to force a quantitative statement about systematics.

The central claim (abstract:s3) is that ~1 ab^-1 gives 5-sigma discovery. That number comes from S/sqrt(B) = 5 in a 1 MeV bin (Sec. 4.2), explicitly 'assuming no systematic errors.' The physics question is whether the QED background can be controlled to the required precision. From Fig. 7, at alpha_X = 10^-8 the signal is ~10^-2 to 10^2 pb and the background is ~10^4 larger, so in a 1 MeV bin at m_X ~ 50 MeV the background is ~10^9-10^10 events for 1 ab^-1 (1 ab^-1 = 10^6 pb^-1). To see a 5-sigma excess of S ~ 10^5-10^6 events, the background in the signal bin must be known to ~S/5, i.e. a relative precision of ~10^-5. The paper's sideband fit uses an 'ad hoc' functional form (Eq. 6.3), and the Fig. 13 demonstration fits pseudo-data generated from the same model, so it is a closure test, not evidence that a real, imperfectly known background can be controlled at 10^-5. This is the load-bearing soft spot: the reach is systematics-limited, not statistics-limited, and the paper's own 'assuming no systematic errors' concedes it without quantifying the degradation. A secondary, unacknowledged error: Eq. (4.3) states L(x MeV) = (1/x) L(1 MeV), i.e. worse resolution needs less luminosity; the correct scaling is L proportional to the bin width (as the reader's 'doubles' and Sec. 2's 'improves linearly' both imply), so Eq. (4.3) is inverted. This does not change the 1 MeV headline but undermines the stated extrapolation.

The paper is a careful, cross-checked feasibility study (CompHEP and MadGraph checks, narrow-width justification in App. B, matrix-element derivation in App. D). The central claim is a reach estimate, not a discovery, and the authors state its assumptions: 'assuming no systematic errors' (Sec. 4.2), 'we do not address experimental backgrounds' (Sec. 4.1), and the trials factor (Sec. 4.2). The reader's CONDITIONAL verdict correctly identifies the idealized detector and the private-communication target thickness (Ref. [34]) as the load-bearing inputs. I agree, and I sharpen the point: the binding issue is not detector resolution per se but the ~10^-5 relative background control that S/sqrt(B) = 5 implicitly requires given ~10^9-10^10 background events per MeV bin. The Fig. 13 closure test uses pseudo-data from the same model as the fit, so it does not establish that a real background can be controlled at that level. This is a genuine limitation of the reach estimate, but it is a limitation the authors largely acknowledge, and it does not make the calculation wrong; it makes the 1 ab^-1 number a best-case statistical reach. The Eq. (4.3) inversion is a real but secondary error that does not affect the 1 MeV headline. No unacknowledged fatal flaw; the verdict should remain CONDITIONAL.

If correct, the result identifies a cost-effective experimental venue for dark force searches, complementary to beam dump experiments and low-energy e+e- colliders, and provides falsifiable reach predictions in the m_X-alpha_X plane. The paper has real strengths: the signal and background cross sections are computed from first-principles QED and cross-checked against CompHEP 4.5.1 and MadGraph 4.4.17; the narrow-width approximation is justified by the eV-scale width; Appendix D gives a clean variational derivation of the optimal weighting function for the matrix element method; and the displaced-vertex strategy in Sec. 7 is a genuinely complementary probe. The authors are also explicit in the body that the S/sqrt(B) procedure is a crude estimate and that a sideband fit and trials factor would be needed in practice.

The paper asks whether a light boson $X$ — the messenger of a hypothetical 'dark force' with mass below 100 MeV and a preference for decaying to $e^+e^-$ — can be found in low-energy electron-proton collisions rather than at a beam dump or a lepton collider. The proposed setting is a 100 MeV electron beam fired into a diffuse hydrogen gas target, where the process $e^-p$ produces $e^-p+X$ and the $X$ decays promptly to an electron-positron pair. Because the recoiling proton and the scattered electron are both measured, all four final-state particles can be reconstructed, which gives kinematic cross-checks that a traditional fixed-target experiment cannot provide. The authors compute the signal and the irreducible QED background for the $e^-pe^+e^-$ final state and find the background is roughly four orders of magnitude larger, so the search is statistics-limited: reaching a $5$ $\sigma$ peak in a 1 MeV mass bin requires about $1~ab^{-1}$ of integrated luminosity. They argue that the JLab Free Electron Laser, with an average current of 10 mA on a target of thickness $10^{19}~cm^{-2}$, would deliver approximately $1~ab^{-1}$ per month, making the search feasible with a month to a year of running.

What carries the argument. The load-bearing device is a signal-versus-background comparison in the four-body final state, organized around the narrow-width approximation for the $X$ propagator, $D(q) = 1/[(q^2-m_X^2)^2+Γ_X^2 m_X^2]$, replaced by $π/(m_X Γ_X)$ times $δ(q^2-m_X^2)$ and combined with a cluster decomposition of four-body phase space into a three-body production piece and a two-body decay piece. This lets the authors keep the full angular correlations of the $X$ decay (which matter for vector and axial-vector couplings, where the amplitude does not factorize into production times decay but contracts through the propagator) while still getting a tractable cross section. On top of that sits the matrix element method: the ideal weight function $w_{best}(Φ) = S(Φ)/B(Φ)$ maximizes $S_{eff}/δB_{eff}$ and yields $[S_{eff}/δB_{eff}]_{best} = √{∫ dΦ S(Φ)^2/B(Φ)}$, which is what produces the factor-of-three improvement over a plain counting experiment.

Core claim. The central claim is a feasibility statement rather than a detection: a light, weakly coupled boson with $10~MeV < m_X < 100~MeV$ and coupling $10^{-8}$-ish, which is allowed by both anomalous-magnetic-moment and beam-dump constraints, could be discovered in low-energy $ep$ scattering with around $1~ab^{-1}$ of data, assuming a 1 MeV $e^+e^-$ invariant mass resolution. The authors establish this by explicit calculation of the signal cross section (narrow-width approximation, full polarization information, with the combinatoric ambiguity of pairing the positron with the wrong electron retained) and of the twelve-diagram QED background, both checked against CompHEP. For $m_X = 50~MeV$ and $m_X = 20~MeV$ benchmark points they show the expected $e^+e^-$ mass distribution, the four-orders-of-magnitude gap between signal and background, and a simulated sideband fit in which the vector-coupling benchmarks appear as visible bumps while the scalar ones, lying below the reach line, do not. They further show that an optimal weighting of phase space — the matrix element method — buys about a factor of three in reach, and that a displaced-vertex strategy could probe couplings two orders of magnitude smaller still.

The premise it rests on. The reach figure depends on an idealized detector with 1 MeV electron-positron mass resolution and no systematic errors, and on a hydrogen target thickness of $10^{19}~cm^{-2}$ that is taken from a private communication rather than a published measurement; if either the resolution or the achievable luminosity falls short, the required $1~ab^{-1}$ moves accordingly.

If the reach estimate holds, the JLab FEL on a hydrogen gas target becomes a competitive venue for dark-force searches in the 10–100 MeV mass range, using roughly a month of running rather than a year at a colliding-beam facility.

For the same integrated luminosity and sufficient effective beam energy, $ep$, $e^-e^-$, and $e^+e^-$ four-fermion searches have reach within a factor of two of each other, so the deciding advantage of $ep$ is the much higher instantaneous luminosity achievable in a fixed-target geometry.

A matrix element analysis using full kinematic information improves the coupling reach by about a factor of three relative to a simple mass-bin counting experiment, and the detector's angular acceptance already acts as a crude version of that weighting.

If vertex displacements of 1 mm or 0.1 mm can be reconstructed, a displaced-vertex search extends sensitivity to couplings about two orders of magnitude below the bump-hunt reach, complementing rather than replacing the prompt-decay analysis.

Pushing the beam energy from 100 MeV to the maximum sustainable 140 MeV improves reach near the kinematic limit $m_X \lesssim 131$ MeV, though it slightly worsens reach at low masses already excluded by beam dumps.

The paper's real comparative advantage is not raw cross section but the combination of a diffuse gas target with four-particle reconstruction: it inherits the luminosity of a beam dump while recovering the kinematic closure that beam dumps lack, which suggests the same logic could be transferred to other light, prompt-decaying particles produced in electron scattering.

Because the required luminosity scales linearly with mass resolution, the entire reach estimate is a statement about detector technology as much as about physics; a calorimetric or tracking upgrade that reached 0.2 MeV resolution would multiply the accessible coupling range by roughly a factor of five at fixed running time.

The authors treat scalar and vector couplings as the representative cases because pseudoscalar and axial-vector reach coincides with them in the small-$m_e$ limit; the real discriminator between coupling structures would be the decay-angle distribution, which they note is flat for scalar and structured for vector, but which sits far below the background in their benchmarks.

The proposal and the beam-dump program are genuinely complementary rather than redundant: beam dumps win at small coupling and long lifetime, the $ep$ experiment wins at moderate coupling with prompt decays, and a null result at $1~ab^{-1}$ would close a specific wedge of the $m_X$–$α_X$ plane that neither technique covers alone.

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