A Window in the Dark Matter Exclusion Limits

Short version: the mechanism is real, the window is probably where they put it, and the natural objection to it does not survive contact with the numbers. A one-mean-free-path estimate through 1400 m of SiO₂ gives ~88 μb per nucleus, which looks fatal until you remember the paper quotes σ_Xp. At σ_Xp = 1 μb, a 1 GeV particle on silicon has σ_XA ≈ 3×10³ μb after the (μ_A/μ_p)² A² factor, so the mean free path is ~100 m, not the full overburden. A dozen collisions at ~7% energy loss each is more than enough to fall under a 600 eV threshold. The turnover at μb is where it should be. If a reader told you the lower edge sits at 100 μb, that reader forgot the coherence factor; say so plainly if it comes up at group meeting. What is new: Rich et al. 1987 and Starkman et al. 1990 had the idea, and the title nods to Starkman. What this paper adds is the specific quantitative band — final XQC data on the upper side, a crustal propagation treatment of CRESST, ELEGANT, COSME and CDMS on the lower side — plus a genuine sensitivity study of the halo parameters. The standard-versus-conservative comparison in Figs. 1 and 2 is the right thing to do, and it shows the window does not close under large swings in v_esc and v_E. The spin-dependent section, the daily modulation signature, and the DMIC fraction constraint in §VI are useful, not padding. Where it is soft, in order of how much I care. First, the low-mass lower boundary rests on a CRESST fit whose background is described only as two falling exponentials plus a constant — no parameters, no binned spectrum, no tabulated rates. The authors say in §IV they are not aiming for precise signal values, which is honest, but this boundary is exactly what decides whether the window exists at m = 1 GeV, and no code ships. Second, XQC's f ≈ 0.5 and the ~10% in-window fraction are asserted rather than derived; moderate. Third, §VII's claim that 100 g on the ground would readily close the window is optimistic once real muon and environmental backgrounds are folded in — they quote the muon flux but do not do the calculation. None of this threatens the central claim. It is a competent parameter-space paper, honest about the systematics it can control and thin on the one it cannot. Anyone working on sub-GeV direct detection should read it; it is the reference for this window. Send it out, mainly to force the CRESST fit details and ideally the simulation into the record.

The central claim is a window at σ ≈ μb for m ≲ 2.4 GeV, bounded below by the CRESST turnover (σ_C,high) and above by XQC's lower edge (σ_X,low). §VII states σ_X,low = 2 μb for m = 1 GeV, and §VIII states σ ≈ μb; for a window to exist, σ_C,high must lie below σ_X,low. §IV computes σ_C,high by propagating DM through 1400 m of crust (ρ = 2.7 g/cm³, SiO₂) and fitting the CRESST spectrum with an unspecified background B₀; neither the fit parameters, the binned spectrum, nor the simulated rates are given. A simple optical-depth estimate is informative: with n ≈ 8.1×10²² nuclei/cm³ and L = 1.4×10⁵ cm, one mean free path corresponds to σ ≈ 1/(nL) ≈ 88 μb. Because a 1 GeV particle on A ≈ 20 loses only ~10–18% of its energy per elastic collision (Eq. 2, isotropic c.m.), several collisions are needed to drop below the 600 eV threshold, pushing the turnover toward ~100–1000 μb. If σ_C,high is at ~100 μb rather than below 2 μb, the window's lower edge moves up by roughly two orders of magnitude and may overlap XQC's exclusion, potentially closing the window at m = 1 GeV. The paper's own §IV says it does not aim for precise signal values, so this boundary carries weight it is not shown to bear. This is not a claim that the mechanism is wrong — the turnover is real — but that its quantitative placement, which determines whether a window exists at all, is unverified and in tension with a back-of-envelope estimate.

The reader correctly identifies the lower boundary as the weak point, and I agree that the CRESST background fit is unspecified and that the 600 eV threshold's derivation is not given. My concern is broader: the entire turnover location is unverifiable from the text, and a simple optical-depth estimate suggests it sits at tens to hundreds of μb rather than the μb range the text asserts. The paper is candid about not aiming for precision in the underground signal (§IV), which is appropriate, but the window's existence depends on the turnover lying below XQC's edge, and that is not demonstrated. The qualitative mechanism is sound, and the standard-versus-conservative comparison shows robustness to halo parameters, but not to the energy-loss model or the background fit. The spin-dependent treatment, daily-modulation signature, and DMIC-fraction constraint are useful additions and do not affect this concern. Hence CONDITIONAL remains the right verdict, and the specific check is to recompute the turnover and XQC edges from the stated inputs and compare to Fig. 2. I have not counted the paper's acknowledged limitations as new findings; the unresolved issue is quantitative placement, not the reality of the mechanism.

If the quantitative placement of the window holds, the result identifies a currently unexcluded region of light dark matter parameter space and gives a falsifiable target for a small dedicated experiment. The connection to baryon-number-carrying dark matter and the H dibaryon is a useful motivation. The simulation approach is reasonable and the paper provides a clear physical mechanism (crust energy loss producing a turnover in the underground rate) that is worth publishing. The main strengths are the explicit treatment of the velocity distribution, the separate spin-independent and spin-dependent analyses, and the proposed experimental signature with daily variation. The weakness is that the load-bearing lower boundary of the window is not tabulated, so the central claim cannot be checked from the text alone.

The paper asks what happens to direct-detection limits on light dark matter — masses from about 0.4 to 10 GeV — when the particle's scattering cross section on nucleons is neither vanishingly weak nor hadronically strong, but sits in the middle, around a micro-barn. The authors point out that essentially all underground experiments quote limits under the assumption that the incoming dark matter arrives at the detector with its full galactic velocity. At intermediate cross sections that assumption fails: a particle scatters repeatedly in the few kilometres of crust above the detector, loses energy, and drops below the detector's recoil threshold, so the experiment goes blind precisely where it was thought to be most sensitive. The consequence is that the exclusion region does not simply extend upward forever; it closes from above at some cross section, and the space-based XQC observation closes it again from below at larger cross sections, leaving an unexcluded band. The authors put that band at masses below roughly 2.4 GeV and cross sections of order a micro-barn, and argue it is small enough to be probed by a modest dedicated experiment. The payoff, if the argument holds, is a genuinely untested corner of dark matter parameter space — one that overlaps with the mass range expected for a stable H dibaryon dark matter candidate.

What carries the argument. The carrying mechanism is energy loss in the Earth's crust, modelled as a Monte Carlo transport problem. A dark matter particle is drawn from a truncated Maxwell-Boltzmann halo distribution, propagated through crust of density 2.7 g/cm$^3$, and its successive collisions are sampled with an effective mean free path $1/λ_{eff} = Σ_i n_i σ_{XA_i}$ and a species-selection probability $P_i = n_i σ_{XA_i} / Σ_j n_j σ_{XA_j}$. Each collision deposits a nuclear recoil energy set by the scattering kinematics, $E_R = 4 m_A m_X (m_A+m_X)^{-2} (m_X v_X^2/2)(1-φ)/2$ with isotropic centre-of-mass scattering, and a nuclear form factor $F^2(q^2) = exp[-(qR)^2/10]$ handles momentum transfer for heavier masses. Particles that reach the detector with enough energy are then collided again inside the target, and the resulting spectral rate is compared with the published background of CRESST, DAMA, ELEGANT, COSME I and CDMS, with the XQC limit derived from its two observed events against an expected background of 1.3. The same machinery produces a second, independent signature: because only particles arriving from the sky above the detector can reach it with energy to spare, the rate depends on the angle between the detector and the direction of the Earth's motion through the Galaxy, and so varies over the course of a day.

Core claim. The paper establishes that the standard way of quoting dark matter cross section limits — assume the full halo flux reaches the detector, scale the rate linearly with cross section — breaks down for light particles with intermediate couplings. Because the mean free path of such a particle in ordinary rock is of order a hundred metres, it scatters many times on its way down to an underground laboratory, losing a fraction of its energy at each collision. By the time it arrives it may no longer be able to deposit a recoil above the experiment's threshold, so the measured rate falls as the cross section rises. The exclusion limit therefore turns over: above some cross section, the experiment stops excluding and starts being insensitive. Combining this turnover with the upper limit from the XQC sounding-rocket experiment, which sits above the atmosphere and so is not shielded, and with the older balloon limit of Rich et al. that closes the range above about 2.4 GeV, the authors find a surviving window for spin-independent interactions at $m_X < 2.4$ GeV and $n_X$-nucleon cross sections near $10^{-3}$ pb (a micro-barn). They repeat the analysis for purely spin-dependent interactions, where only the少量 of target nuclei carrying spin respond, and the window widens; they also map the allowed region when both interactions are present at once, where screening between the two target types permits larger cross sections for the same rate.

The premise it rests on. The lower edge of the claimed window rests entirely on the CRESST exclusion, which the authors obtain by fitting the measured spectrum with an empirical background function described only as two falling exponentials plus a constant, without publishing the fitted parameters or the binned data — so if that fit or the 600 eV nuclear threshold is wrong, the window moves or closes.

A dedicated low-threshold experiment on the ground — the authors suggest a microcalorimeter array of roughly 100 g, similar in spirit to XQC's — could either detect a signal or close the window outright, since the predicted rate at $σ_{Xp} = 2$ $μ$b is comparable to the unshielded cosmic-ray muon background.

The surviving window overlaps the mass range expected for a stable H dibaryon, so a baryon-number-carrying dark matter candidate is not excluded by existing direct-detection data.

For spin-dependent interactions the window is wider than for spin-independent ones, because only the small odd-neutron or odd-proton fraction of the XQC target responds, weakening its upper limit.

Because the rate depends on the angle between the detector and the local standard of rest, a light intermediate-cross-section signal would show strong daily modulation — a signature distinguishable from ordinary weakly interacting dark matter.

The XQC data can be turned around to bound the fraction of the local dark matter that may consist of such particles, as a function of cross section, with the bound falling linearly with increasing cross section in the regime where overburden interactions can be neglected.

The turnover argument is generic to any shielded detector, so the same reasoning should apply to future large-scale direct-detection experiments: a deeper or lower-threshold detector is not automatically more sensitive to light particles, and a cross-section ceiling on sensitivity should be quoted alongside every floor.

The daily modulation the authors flag is a stronger discriminator than the rate itself, because a modulations analysis does not require knowing the absolute background normalization — the same logic that motivates annual-modulation searches, but on a 24-hour timescale and with an amplitude set by geometry rather than by the Earth's orbital velocity.

The window's lower boundary is set by CRESST, which the authors treat as the only running experiment with a low enough threshold; a reanalysis that pushes that threshold down or fits the background differently would move the boundary and could shrink the window substantially.

If such a particle exists at micro-barn strength, it would also be stopped in the upper metres of any rock overburden, which suggests that shallow-site experiments and balloon or space-based instruments are complementary in a way that is not usually encoded in a single exclusion plot.

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