Probing unexplored spin-dependent dark matter-proton coupling with few-photoelectron threshold in COSINE-100

Abstract

We report new constraints on the spin-dependent scattering cross section between low-mass dark matter and protons using data collected by the COSINE-100 experiment. By implementing a specialized event selection process using a multi-layer perceptron and robust noise mitigation, this analysis pioneers a detection threshold of 3 and 4 isolated peaks, corresponding to the reconstructed photoelectrons, which is significantly lower than the 8 photoelectron threshold used in previous analyses. In this unstudied few-photoelectron regime, where PMT-induced noise and phosphorescence are prevalent, we utilize a phenomenological background model to search for the annual modulation signal expected from the Standard Halo Model. No statistically significant annual modulation is observed in our data. We derive new 90% confidence level upper limits for the spin-dependent DM-proton cross section, establishing the world's most stringent constraints in the 1.75-2.25 GeV/c$^2$ mass range. Furthermore, by incorporating the Migdal effect, we extend the experimental sensitivity to the sub-GeV/c$^2$ regime, setting world-leading limits in the 15-58 MeV/c$^2$ range. These results demonstrate the capability of NaI(Tl) target materials to probe previously unexplored regions of the dark matter parameter space.

Figures (3)

• Figure 1: Representative waveforms for NC-4 events in a NaI(Tl) crystal. (a) A signal-like event passing the MLP selection criteria, with clusters distributed within the characteristic 200 ns scintillation decay time. The inset provides a zoomed-in view of the 2.3–2.8 $\mu\text{s}$ region, clearly illustrating the individual photon clusters within the pulse. (b) A noise-like event rejected by the MLP selection due to clusters distributed over a broader time interval, characteristic of phosphorescence. (c) A noise-like event induced by Cherenkov radiation in the PMT glass, characterized by an anomalous single-cluster profile, rejected by the cluster charge cut.
• Figure 2: Event rates as a function of time for the five NaI(Tl) crystals used in the analysis, overlaid with the best-fit annual modulation model for single-hit events. Data are grouped in 15-day bins, and the fits are performed simultaneously across all crystals with a fixed phase of 152.5 days and a shared modulation amplitude $A$. The bottom panels show the residuals from the fits. (Left) The NC-3 dataset is shown with the blue shaded bands representing the systematic uncertainties ($\delta u_i$). (Right) The NC-4 dataset is shown.
• Figure 3: The derived 90% C.L. upper limits on the SD DM-proton cross section as a function of DM mass m$_\chi$. The thick solid line represents the result of this work. (Top) Constraints in the $1.2–3.0$ GeV/c$^2$ mass region for standard SD nuclear recoils. Our result provides new constraints in the $1.75–2.25$ GeV/c$^2$ range, a region not previously explored by the compared experiments: PICO picocollaboration2019, PandaX-4T zhang2025a, CRESST-$\text{LiAlO}_2$angloher2022, NEWS-G arora2025a, and the recent COSINE-100 3-year data result yu2025. (Bottom) Extended sensitivity in the $15–200$ MeV/c$^2$ region enabled by the Migdal effect. Our results establish stringent new limits in the $15–58$ MeV/c$^2$ range, significantly extending the reach beyond the 100 MeV/c$^2$ limit achieved in previous COSINE-100 3-year data result yu2025. This work surpasses the previous experimental reach of Collar collar2018a and XENON1T xenoncollaboration2019.