Sidereal On/Off Windows for Point-Source Boosted Dark Matter

Underground dark-matter detectors are usually analyzed as if they sit in a steady rain of particles: count events over hundreds of days, subtract a modeled background, and quote a limit. That picture is wrong for one popular class of signals, dark matter "boosted" to high speed by a specific astrophysical source such as a blazar. A blazar is a galaxy whose central black hole fires a relativistic jet toward us. A boosted flux arrives from a fixed point on the sky. As the Earth turns, the amount of rock between that point and the detector changes, and for part of every sidereal day (the 23.9-hour rotation period measured against the stars) the source is below the horizon, behind the whole Earth. Gamma-ray astronomers handle this kind of geometry with "on/off" analyses: they compare the region that contains the source with a region that does not, and they budget around limited observing windows. We propose to apply the same discipline to underground detectors. We would re-analyze the CDEX-10 germanium data as a sidereal-time-resolved on/off measurement, so that the hours when the source is shielded by the Earth serve as a measured background sample, and so that the strong-interaction end of the excluded region is tested by a time signature rather than by a fitted background shape.

The second half of the proposal asks when the same time-resolved idea gains a second handle, the orientation of the crystal itself. In a layered crystal, the electrons respond differently depending on the direction of the incoming momentum. Earth's rotation sweeps a point source across the crystal axes as well as through different rock paths. Predicting that crystal-axis modulation needs an electronic response function that standard first-principles calculations can get wrong in narrow-gap and correlated materials. Angle-resolved photoemission spectroscopy (ARPES), which maps electron energy against crystal momentum by measuring the angle and energy of electrons knocked out by light, is the measurement that can check those calculations. We treat ARPES as a calibration input with an explicit kill test, not as an assumed improvement.

Core Claim

For a point-source boosted dark-matter flux, the expected event rate in an underground detector is a known function of sidereal time once the cross-section is fixed. Write it as $R(t) = \epsilon \int dE\, \Phi(E; \sigma, L(t))\, S(E)$. Here $t$ is sidereal time, $L(t)$ is the rock-and-atmosphere path length toward the source computed from the laboratory's topography, $\sigma$ is the dark-matter–nucleon cross-section, $\Phi$ is the flux surviving that path, $S$ is the detector response above threshold, and $\epsilon$ is the efficiency-weighted exposure. When $\sigma$ is small the Earth is transparent and $R(t)$ is flat, so time resolution adds nothing; we say this plainly. When $\sigma$ is near the attenuation "ceiling," $R(t)$ becomes an on/off pattern locked to sidereal time. That pattern cannot be mimicked by radioactive backgrounds, which are constant or follow the solar day. Our claim is that a sidereal on/off likelihood makes the upper edge of the excluded region robust to background mismodeling. The size of the gain depends on how the published analysis handled the arrival geometry. It also yields a background-only control sample at no cost in exposure.

Relation to Prior Work

Two CDEX-10 analyses motivate this work and set its baseline. Constraints on the Blazar-Boosted Dark Matter from the CDEX-10 Experiment uses 205.4 kg·day of germanium data at a 160 eVee threshold, under roughly 2,400 m of marble, to exclude dark matter boosted by TXS 0506+056 and BL Lacertae. It accounts for overburden attenuation, which produces an upper ceiling near $10^{-26}$ cm². According to the research file, it treats the blazar flux in a time-integrated way. Constraints on dark matter boosted by supernova shock within the effective field theory framework from the CDEX-10 experiment does the same for the Monogem Ring remnant with an isotropic, static flux. Our method keeps their flux models and attenuation physics but makes the arrival direction, and therefore time, part of the likelihood. Xenon searches for dark matter boosted by cosmic rays (PandaX-4T, LUX-ZEPLIN) involve a diffuse galactic flux, so a point-source on/off geometry does not apply to them in the same way. They serve as context, not as competitors.

The on/off and window-budgeting logic comes from transient astronomy. Imaging Cherenkov telescopes coordinate across sites to work around limited observing windows when following up IceCube neutrino alerts (FACT, H.E.S.S., MAGIC, VERITAS follow-up). The Telescope Array mass-composition study (Telescope Array Surface Detector) shows the payoff of designing an analysis around when the instrument can actually see a signal. It used a detector running near 100% of the time instead of fluorescence telescopes limited to about 10%. We borrow the method, not the data.

The material side builds on Migdal Effect in Semiconductors. The Migdal effect is the knocking loose of an electron when a nucleus is jolted. That paper computes Migdal rates in silicon and germanium from a calculated energy loss function, which describes how the material's electrons absorb energy and momentum. Its authors note that this function omits off-diagonal terms and that their approximation fails below about 50–70 MeV. Phonon-mediated Migdal effect in semiconductor detectors extends the treatment to 10 MeV, but only for an idealized, near-rotationally-symmetric, zero-temperature crystal. Both papers constrain us: neither offers an anisotropic response, so any crystal-axis modulation we predict is new and must be validated. The ARPES studies of WTe₂ (Fermi arc in WTe2) and Sr₃Ru₂O₇ (heavy d-electron quasiparticles) motivate the calibration step. The WTe₂ study reported a hole-to-electron pocket balance that departed from the standard compensation picture. The Sr₃Ru₂O₇ study showed bands compressed far below local-density-approximation predictions, with effective masses approaching 1,000 times the electron mass.

Research Program

The first component is a geometric exposure model for the China Jinping Underground Laboratory. It maps topographic rock thickness versus azimuth and elevation and combines that map with the declinations of TXS 0506+056 and BL Lacertae to give $L(t)$ over the sidereal day. When the source is below the horizon, $L$ is the chord through the Earth, effectively infinite for cross-sections near the ceiling. We will reuse the attenuation and recoil-spectrum code paths of the CDEX blazar analysis. This makes any difference between our results and theirs a consequence of geometry and time binning, not of new physics assumptions.

The second component is the likelihood. CDEX events, with timestamps and live-time records obtained through collaboration access, are binned in energy and sidereal phase. The signal template is $R(t)$ evaluated at each $(m_\chi, \sigma)$, where $m_\chi$ is the dark-matter mass. The background is constrained jointly by the published background model and by the below-horizon hours, which act as the "off" sample. Live-time gaps and calibration downtime are handled as exposure weights per phase bin, the same way telescopes treat weather losses.

The third component is a forecast for a layered target. We build tight-binding band models of WTe₂ fit to the ARPES dispersions in Fermi arc in WTe2. For the occupied bands we use the measured pockets. For states just above the Fermi level we use the data taken at 165 K, where heating populates those states. Higher unoccupied bands remain DFT-only, and we state this limitation openly; DFT here means density functional theory, the standard first-principles method for computing electronic structure. From these bands we compute a direction-dependent energy loss function in the random-phase approximation. We then fold it with both the point-source geometry and the standard galactic dark-matter "wind" to predict sidereal modulation amplitudes. Sr₃Ru₂O₇ serves only as a stress test of how strong band renormalization propagates into the response.

Baselines

The naive reference is the published time-integrated CDEX-10 result. On 205.4 kg·day at a 160 eVee threshold, Constraints on the Blazar-Boosted Dark Matter from the CDEX-10 Experiment excludes $\sigma_{\chi N}$ from $4.6\times10^{-33}$ to $1\times10^{-26}$ cm² for TXS 0506+056 and from $2.4\times10^{-34}$ to $1\times10^{-26}$ cm² for BL Lacertae, across dark-matter masses of 10 keV to 1 GeV.

That result is also the strongest existing method, so we will reproduce it ourselves before comparing. We will rerun the time-integrated likelihood on the same event list, flux model, attenuation code, and background model, and confirm that we recover the published band within numerical tolerance. The sidereal analysis then gets the same data and no additional tuning freedom; its only new input is the topographic map. Beating the baseline means one of two things. Either the upper edge of the excluded band rises measurably above $10^{-26}$ cm² for at least one source, or, at equal reach, the limit near the ceiling stays stable when the background model is deliberately perturbed while the integrated limit does not. We do not expect, and will not claim, any gain at the lower edge, where the Earth is transparent.

Two ablations isolate the key ingredient. The first replaces $L(t)$ with its daily average while keeping the time binning, which tests whether the gain comes from geometry or from binning. The second shuffles event timestamps within live periods, which gives the null distribution of any apparent sidereal structure.

For the material forecast, the baselines are the DFT-only anisotropic energy loss function and the isotropically averaged response in the style of Migdal Effect in Semiconductors. No reported result was found on Lacuna for anisotropic Migdal or electron-excitation modulation in any layered material.

Why not just…?

Why not keep the time-integrated analysis and simply collect more exposure? More exposure pushes the lower edge of the band down, but it barely moves the ceiling. The ceiling is set by attenuation in the overburden, not by event counts, as both CDEX analyses note. Near the ceiling, the dominant risk is that a mismodeled background shape absorbs or fakes a signal. The below-horizon hours provide a measured background drawn from the same detector, with no extra running time. The background-perturbation test described above shows directly whether this matters.

Why not correlate events with blazar flares or IceCube alerts instead of sidereal geometry? The research file suggests this route (Gap 2), but arrival-time spread largely rules it out. Over a distance $D$, particles with Lorentz factor $\gamma$ spread in arrival time by roughly

$$\Delta t \approx \frac{D}{2c\gamma^2},$$

which follows from expanding the velocity $v \approx c(1 - 1/2\gamma^2)$. For TXS 0506+056 at about a gigaparsec, keeping $\Delta t$ under a two-day flare requires $\gamma$ of order $10^5$–$10^6$. That is plausible only for the lightest keV-scale masses. At the MeV–GeV masses where CDEX is most sensitive, a flare smears over years or much longer. We will compute this cutoff explicitly and report the narrow mass window, if any, where flare timing survives. Sidereal modulation does not depend on emission time at all.

Why not use DFT alone for the anisotropic target and skip ARPES? This alternative is a real threat, and it may win. ARPES measures only occupied single-particle states from a thin surface layer, while the energy loss function is a bulk two-particle response. ARPES therefore corrects the band input; it does not measure the response directly. We settle the question with a kill test. If the ARPES-corrected and DFT-only predictions of sidereal modulation amplitude differ by less than the spread between standard DFT functionals, or by less than the statistical resolution of a kilogram-year exposure, then ARPES calibration is unnecessary for this purpose and we will say so.

Decisive Checks

The first check is a closure test. We inject simulated signals at cross-sections spanning $10^{-28}$ to $10^{-24}$ cm² into background-only pseudo-data with the real live-time pattern, and confirm that the sidereal likelihood recovers them with correct coverage. The second check probes systematics: we fit for modulation at the sidereal frequency, the solar frequency, and the anti-sidereal frequency (a control period of about 24.07 hours). Any real point-source signal must appear only at the sidereal frequency, while daily environmental effects such as radon or temperature will appear at the solar frequency. The third check compares limits. We report the TXS 0506+056 and BL Lacertae exclusion bands from the sidereal analysis against our reproduction of the published bands, along with the stability of each under a ±20% reshaping of the background model. The material check is the DFT-versus-ARPES modulation comparison for WTe₂, judged against the kill threshold above.

If the first three checks show no change in the ceiling and no gain in robustness, the conclusion is still useful: it would show that existing time-integrated limits already capture the geometry. The forecast would then rest on the crystal-axis handle alone.