Symmetry-Pinned Zero Gaps as Low-Threshold Dark Matter Targets

Searches for light dark matter in the halo have a blind spot that is a mirror image of the gaps in gamma-ray and neutrino astronomy. Large detectors such as liquid-xenon chambers have the mass to catch rare events, but their energy thresholds are too high to see the tiny kicks that a particle lighter than a proton gives a heavy nucleus. Tiny quantum sensors have very low thresholds but almost no mass. One way through is the Migdal effect, in which a nucleus that is struck suddenly also shakes its electrons and deposits detectable electronic energy. In crystals, the size of this signal depends on how easily the material's electrons absorb energy at very low energies. In silicon that ability stops at the 1.2 eV band gap, the minimum energy needed to free an electron, and the predicted rate climbs steeply just above it. This proposal asks whether crystals whose band gap is forced to be exactly zero by their symmetry, and can then be reopened to a few meV by controlled strain, give a much larger Migdal signal per kilogram than silicon. We would answer this with first-principles electronic-structure calculations, under conditions realistic enough that a negative answer would be informative.

The idea rests on a specific property. In "nonsymmorphic" crystals, whose symmetries combine a rotation or reflection with a fraction of a lattice step (glide planes and screw axes), group theory forces pairs of electronic bands to touch along whole lines or planes of momentum space. That touching does not depend on chemical details or on how well the approximate quantum theory reproduces them. This matters because the standard workhorse, density functional theory (DFT, which computes electronic structure from the ground-state electron density), is known to get band gaps wrong. A symmetry-enforced zero is one gap that DFT cannot get wrong. Breaking the symmetry with strain then opens a gap that grows smoothly from zero, which gives the detector threshold a physical tuning knob.

Core Claim

We claim that the near-threshold energy loss function of a nonsymmorphic nodal semimetal is the right quantity for deciding whether such crystals help the Migdal channel. The energy loss function, $\mathrm{Im}[-1/\epsilon(\mathbf{q},\omega)]$ with $\epsilon$ the momentum- and frequency-dependent dielectric function, measures how readily the crystal absorbs energy $\omega$ at momentum $\mathbf{q}$. It must be computed with three ingredients that prior Migdal work did not need for silicon: intraband (free-carrier) screening, crystal local-field effects, and a strain series that breaks the protecting symmetry.

The claim is falsifiable in two ways. First, we predict that the rate with a deposit threshold of tens of meV exceeds silicon's rate above its 1.2 eV gap by a large factor per unit target mass at dark matter masses where the impulse approximation is valid. We commit only to the direction of this effect. Second, we predict that most of that gain disappears in the same material once strain breaks the glide symmetry and the gap exceeds the deposit threshold. If free-carrier screening flattens the low-energy loss function so that the gain is absent, the idea is wrong, and we will say so.

Relation to Prior Work

Prior work supplies the method. Three papers compute Migdal rates in crystals from first principles. Migdal Effect in Semiconductors writes the rate in terms of the DFT energy loss function for silicon and germanium. It uses the impulse approximation, which treats the nucleus as recoiling freely and holds above roughly 70 MeV and breaks down below about 50 MeV, and it keeps only the diagonal part of the dielectric matrix. Describing the Migdal effect with a bremsstrahlung-like process and many-body effects recasts the process as virtual-photon emission screened by electrons. It computes this with Quantum ESPRESSO plus the random phase approximation (RPA, a standard treatment of collective electron screening) in YAMBO, and finds a rate rising as $\omega^{-4}$ toward low energy transfer, far above tight-binding estimates. Phonon-mediated Migdal effect in semiconductor detectors extends the treatment below 50 MeV through virtual-phonon exchange and finds agreement with the bremsstrahlung picture in silicon down to 10 MeV. It explicitly names narrow-gap Dirac materials as the untested next step. All three restrict themselves to cubic, gapped, essentially monatomic crystals. None faces a Fermi surface, a symmetry-enforced crossing, or strong anisotropy, and those are what this proposal adds.

Prior work also supplies the candidate materials, and it constrains the choice. Node-surface and node-line fermions from nonsymmorphic lattice symmetries shows with DFT that screw axes in hexagonal Ba$MX_3$ force a node surface at $k_z=\pi/c$, and that spin-orbit coupling in BaTaS$_3$ reduces it to fourfold-degenerate node lines with Berry phase $\pi$. The same paper warns that BaVS$_3$ undergoes a metal-insulator transition near 70 K. We therefore exclude BaVS$_3$ as a cryogenic target and use BaTaS$_3$ and BaNbS$_3$. Strain-induced nonsymmorphic symmetry breaking and removal of Dirac semimetallic nodal line in an orthoperovskite iridate shows that the SrIrO$_3$ nodal line is protected by an $n$-glide rather than a mirror, and that a monoclinic shear of about 0.4° opens a gap of about 5 meV. This is the demonstrated knob our strain series copies. Their samples were 18 nm films, so scaling to bulk is an explicit open issue rather than an assumption we make. For screening and a cross-check, Catalogue of topological electronic materials provides symmetry diagnoses for 39,519 compounds, which we use to pick weakly correlated nonsymmorphic semimetals. None of these materials papers computes a dielectric response or a recoil rate.

Research Program

The first stage builds a calculation pipeline and checks it against silicon and diamond. Ground-state DFT runs in Quantum ESPRESSO, and the RPA dielectric matrix is computed in YAMBO. Both are the same tools as Describing the Migdal effect with a bremsstrahlung-like process and many-body effects, so differences against that work isolate the material rather than the code. Two changes are needed for semimetals. We include intraband transitions, meaning the free-carrier Drude term that dominates screening at small $\omega$ in any material with a Fermi surface. We also use adaptive k-point meshes dense enough to resolve the low-energy structure around node lines and surfaces. Rates are folded with the standard halo velocity distribution through the impulse-approximation expression of Migdal Effect in Semiconductors. We restrict the main study to dark matter masses of 50 to 500 MeV, where that approximation is trustworthy. At these masses the maximum electronic deposit is tens to hundreds of eV, so the low-$\omega$ part of the spectrum, where the $\omega^{-4}$ enhancement lives, is exactly what a lower threshold recovers.

The second stage computes the target materials and a strain series for each. For SrIrO$_3$ and BaTaS$_3$ we compute the loss function and the differential rate $dR/d\omega$ as a function of applied shear or uniaxial strain, from zero up to the strain that opens a gap above our lowest deposit threshold. The gap opened by the strain is symmetry-allowed only because the glide or screw is broken, so its size should grow smoothly from zero. We test the sensitivity to the treatment of electron correlation in the 5d iridate by repeating the calculations with DFT+U. The symmetry argument predicts that the zero gap at zero strain survives any such change, while the size of the gap at finite strain moves. Confirming this split is itself a check of the method's central premise.

The third stage adds two ingredients that matter specifically in these crystals. We compute local-field corrections, the off-diagonal elements $\epsilon_{\mathbf{G},\mathbf{G}'}$ that both silicon papers drop, and we compute the anisotropy of the rate with respect to the crystal axes. We have a concrete reason to expect local fields to matter here. In a nonsymmorphic crystal the two bands that are forced to touch are partners related by the fractional translation, so transitions between them carry weight at a reciprocal-lattice vector rather than at small $\mathbf{q}$. That is exactly where local-field mixing acts. The anisotropy calculation turns the quasi-one-dimensional chain structure of Ba$MX_3$ into a predicted daily modulation amplitude as Earth rotates relative to the dark matter wind. We treat this as a secondary output and do not rest the proposal on it.

The fourth stage extends the calculation below 50 MeV, where the low gap matters most. Following Phonon-mediated Migdal effect in semiconductor detectors, we generalize the phonon-mediated rate to polyatomic, anisotropic lattices, using DFT phonons in place of its monatomic, nearly isotropic model. We then test whether its agreement with the bremsstrahlung approach survives in this setting. We regard this stage as higher risk and report it separately.

Baselines

The naive reference point is silicon computed exactly as in Migdal Effect in Semiconductors, with the diagonal loss function, the impulse approximation and the 1.2 eV gap, evaluated at the realistic deposit thresholds silicon detectors use. Two checks show our pipeline reproduces this baseline: the characteristic loss-function feature near 13 eV in diamond and the $\omega^{-4}$ low-energy scaling, both reported in Describing the Migdal effect with a bremsstrahlung-like process and many-body effects. Matching both is a precondition for every comparison that follows.

The strongest existing method is the same first-principles RPA calculation applied to silicon with an aggressively lowered threshold. That is the realistic competing plan for the community. Migdal effect of Phonon-mediated neutrino nucleus scattering in semiconductor detectors reports that lowering the silicon threshold from 20 eV to 5 eV is worth roughly a 10-fold increase in exposure. We use this as the yardstick: beating silicon means a gain larger than what threshold engineering in silicon already buys. We found no reported absolute Migdal event rate per kg-year for silicon on Lacuna at a fixed dark matter mass. We therefore regenerate that number with the same code, the same velocity distribution, the same k-point convergence criterion and the same compute budget per material. The external limits in the file are search results rather than rate predictions. The best of them is PandaX-4T's dark-matter-electron limit of $1.5\times10^{-41}$ cm$^2$ at 200 MeV from Search for Light Dark Matter with 259 Days of Data in PandaX-4T, and we use it only to place projected reach in context. For the target material to count as a win, its rate per kilogram above its practical threshold must exceed silicon's rate above 5 eV by more than an order of magnitude at fixed dark matter mass, with free-carrier screening included.

Two ablations isolate the proposal's key ingredient. The first is the strain ablation: the same crystal with the protecting glide or screw broken and a gap opened above the deposit threshold. Rate gained at zero strain and lost at finite strain is attributable to the symmetry-enforced crossing. The second is a symmorphic control: the mirror-protected nodal-line metal PbTaSe$_2$ from Topological nodal-line fermions in spin-orbit metal PbTaSe2, computed identically. It tests whether nonsymmorphicity specifically matters or whether any nodal semimetal does equally well. A third switch turns local-field effects on and off to measure the size of that correction against the diagonal-only treatment.

Why not just…?

Why not just use boosted dark matter in existing xenon and germanium detectors? Searches such as Search for Cosmic-Ray Boosted Sub-MeV Dark-Matter–Electron Scattering in PandaX-4T and Constraints on the Blazar-Boosted Dark Matter from the CDEX-10 Experiment reach light dark matter, but only the small fraction that has been accelerated by cosmic rays or blazar jets. Their limits inherit those astrophysical flux models. They complement this proposal rather than replace it, because they cannot see the slow, ambient halo population that a low-threshold target is designed to detect.

Why not just push silicon to lower thresholds? This is the most serious practical competitor, and the baseline above is built around it. Silicon's limit is that no electronic deposit can fall below 1.2 eV, which is exactly where the $\omega^{-4}$ rate is largest. The decisive experiment is the direct per-kilogram comparison against silicon at 5 eV threshold. If that comparison shows no order-of-magnitude gain, silicon wins on maturity, and our conclusion will say so.

Why not just use any narrow-gap or accidentally gapless material? This is the real threat to the symmetry part of the claim. An accidental small gap could give the same loss function. Our answer is partly methodological. DFT routinely underestimates gaps, so a predicted accidental meV gap is not trustworthy, whereas a symmetry-enforced zero is exact and robust to disorder that preserves the space group. The symmorphic control and a scissor-shifted comparison test the physics directly. If an accidental or mirror-protected gap of the same size gives the same rate, the nonsymmorphic advantage reduces to predictability and strain tunability, and we will report it at that strength.

Why not just scale up superconducting nanowire sensors? First Sub-MeV Dark Matter Search with the QROCODILE Experiment Using Superconducting Nanowire Single-Photon Detectors reaches a 0.11 eV threshold and dark matter masses down to 30 keV, but with wire-scale sensing mass. A bulk crystal offers low-energy response throughout its volume. The cost is that its meV deposits must be read out with phonon-based, temperature-measuring sensors at millikelvin temperatures, and that readout problem is not solved here. Our output, the rate spectrum versus threshold per kilogram, states the threshold such a readout would have to reach.

Why not just use tight-binding models instead of DFT plus RPA? Describing the Migdal effect with a bremsstrahlung-like process and many-body effects shows that localized-orbital models badly underestimate low-energy rates because they miss collective screening. The topological bands we study also resist description in terms of localized orbitals. We keep the full calculation.

Decisive Checks

The study succeeds or fails on four results. The pipeline must reproduce the diamond 13 eV feature and the $\omega^{-4}$ scaling before any target result is trusted. With free-carrier screening included, the zero-strain target rate must exceed the silicon-at-5-eV rate per kilogram by more than tenfold at 100 MeV dark matter mass with a threshold of tens of meV. If Drude screening suppresses the low-$\omega$ loss function enough to erase this gain, the proposal's central claim is refuted. The strain series must show the gain disappearing as the symmetry-allowed gap opens past the threshold, and the symmorphic and accidental-gap controls must fall measurably short of the nonsymmorphic targets. Otherwise only the weaker robustness argument survives. Finally, the local-field correction must be reported with its size relative to the diagonal-only treatment, which tests whether the order-one bound established for silicon carries over to these crystals.