Dark Matter Test Science Project

• The project defines a framework where dark matter theories are mapped to explicit experimental tests and quantifiable observables.
• It unites diverse methodologies—from direct detection to cosmological constraints and simulation-based inference—to tackle dark matter challenges.
• Results from initiatives like COSINUS and CMB-S4 demonstrate the impact of rigorous calibration and open-science reproducibility on dark matter searches.

A dark matter test science project is a structured research program that converts dark-matter hypotheses into falsifiable observables, instrument or survey requirements, inference pipelines, and explicit control of systematics. In current usage, the term spans open-science interoperability efforts such as ESCAPE, same-target direct-detection programs such as COSINUS and PICOLON, proof-of-principle cryogenic targets such as SWEET and ALETHEIA, beam-dump and mission-scale indirect searches such as BDX and GRAMS, cosmological tests with CMB-S4, and simulation-driven platforms such as DREAMS (Pearson et al., 26 Sep 2025, Angloher et al., 2016, Fushimi et al., 2021, Battaglieri et al., 2016, Aramaki et al., 2019, Dvorkin et al., 2022, Rose et al., 2024).

1. Conceptual basis

The modern form of a dark matter test science project is best understood as a coordinated falsification framework rather than as a single experimental technique. One explicit formalization is the “Ten-Point Test,” which requires a candidate to satisfy relic-density, coldness, neutrality, Big Bang nucleosynthesis, stellar-evolution, self-interaction, direct-detection, gamma-ray, broader astrophysical, and experimental-testability criteria (0711.4996). This framework is unusually important because dark-matter proposals now span thermal relics, light mediators, axions and ultra-light bosons, sterile neutrinos, self-interacting sectors, compact-object scenarios, and effectively millicharged or dissipative sectors.

A second defining feature is complementarity. The ESCAPE Dark Matter Test Science Project was created explicitly to connect astronomy, astroparticle physics, and high-energy particle physics, using FAIR publication of data, software, and workflows on the European Open Science Cloud so that collider, direct-detection, indirect-detection, and astrophysical constraints can be translated into one another (Pearson et al., 26 Sep 2025). A similar logic underlies CMB-S4 dark-matter forecasting, Via’s stellar-stream survey design, and DREAMS’s simulation-and-emulator program: each treats dark matter not as a single-parameter exclusion problem, but as a joint inference problem spanning particle microphysics, cosmology, baryonic physics, and survey systematics (Dvorkin et al., 2022, Collaboration, 16 Jun 2026, Rose et al., 2024).

This suggests that “test science project” functions as a design philosophy. The common element is not detector technology, but the requirement that a candidate dark-matter model be tied to specific observables, null tests, calibration strategies, and decision criteria.

2. Experimental and observational modalities

Dark matter test science projects now occupy four main domains. Direct-detection programs seek nuclear or electronic recoils, phonons, scintillation, bubbles, or coherent forces in terrestrial targets. Accelerator and mission-scale searches generate or intercept dark-sector particles with beam dumps, gamma-ray telescopes, or antimatter spectrometers. Cosmological programs use primary anisotropies, lensing, and secondary CMB observables to constrain energy injection, scattering, and small-scale structure. Astrophysical structure projects infer dark-matter microphysics from stellar streams, clusters, satellites, strong lensing, or hydrodynamic simulation suites (Angloher et al., 2016, Battaglieri et al., 2016, Dvorkin et al., 2022, Collaboration, 16 Jun 2026, Nieuwenhuizen, 2017, Rose et al., 2024).

The range of implementations is unusually broad. Same-target NaI projects attempt decisive material-controlled tests of the DAMA/LIBRA modulation claim; unconventional cryogenic targets investigate whether hydrogen-rich organics or liquid helium can become competitive low-mass detectors; beam-dump experiments probe MeV–GeV sectors through elastic and inelastic scattering; GRAMS combines MeV gamma rays with antideuteron searches in a single LArTPC; CMB-S4 turns damping-tail and lensing precision into sensitivity to annihilation, scattering, dark radiation, and ultra-light bosons; Via converts cold stellar streams into gravitational detectors of dark, starless subhalos (Angloher et al., 2016, Bento et al., 29 Sep 2025, Liao et al., 2021, Battaglieri et al., 2016, Aramaki et al., 2019, Dvorkin et al., 2022, Collaboration, 16 Jun 2026).

Project	Modality	Distinctive feature
CMB-S4 (Dvorkin et al., 2022)	Cosmological test	≈1 μK-arcmin, <1.5 arcmin FWHM, ≈70% sky
COSINUS (Angloher et al., 2016)	Cryogenic NaI direct detection	Same NaI target as DAMA/LIBRA with phonon-plus-light readout
SWEET (Bento et al., 29 Sep 2025)	Cryogenic organic target	0.96 g sucrose crystal, ~19 h operation, base temperature below 7 mK
ALETHEIA (Liao et al., 2021)	Dual-phase liquid-helium TPC	30 g prototype cooled to 4.5 K; dark current <10 pA up to 17 kV/cm
BDX (Battaglieri et al., 2016)	Electron beam-dump search	~1 m^3 segmented CsI(Tl), up to 10^22 electrons-on-target
GRAMS (Aramaki et al., 2019)	MeV gamma ray and antimatter	LArTPC covering 0.1–100 MeV and low-energy antideuterons
Via (Collaboration, 16 Jun 2026)	Near-field cosmology survey	>2,000,000 stars and <100 m s^-1 radial-velocity stability
ESCAPE (Pearson et al., 26 Sep 2025)	Open-science integration	VRE, Data Lake, and OSSR for cross-experiment workflows

3. Theoretical observables and inference frameworks

A dark matter test science project is typically organized around a compact set of theory-to-observable maps. For direct detection, the standard recoil-rate formalism recurs across projects: $$\frac{dR}{dE_R} = \frac{\rho_\chi}{m_\chi m_N}\int_{v>v_{\min}(E_R)} f(\vec v)\,\frac{d\sigma_{\chi N}}{dE_R}\,d^3v,$$ with

$$v_{\min}(E_R) = \sqrt{\frac{m_N E_R}{2\mu_{\chi N}^2}}.$$

This appears in hydrogen-rich sucrose projections, NaI calorimetry, helium TPC planning, and superheated-emulsion limit setting, although each project couples it to different thresholds, quenching, target responses, or readout observables (Bento et al., 29 Sep 2025, Angloher et al., 2016, Liao et al., 2021, Kumar et al., 31 Aug 2025).

Cosmological projects instead map dark-matter microphysics into recombination, damping-tail, and lensing observables. For annihilation, the central parameter is

$$p_{\rm ann}=f_{\rm eff}\,\frac{\langle\sigma v\rangle}{m_\chi},$$

while dark-matter–baryon scattering is commonly written as

$\sigma(v)=\sigma_0 v^n.$

The 2022 CMB-S4 white paper translates these parameterizations into explicit forecasts, including a projected $95\%$ C.L. bound $\sigma_0<6\times10^{-27}\,{\rm cm}^2$ for velocity-independent scattering with $m_\chi=1$ GeV, a thermal-annihilation reach corresponding to $m_\chi>30\text{–}50$ GeV for $f_{\rm eff}\approx0.2$, $\sigma(N_{\rm eff})\approx0.03$, $$v_{\min}(E_R) = \sqrt{\frac{m_N E_R}{2\mu_{\chi N}^2}}.$$0 MeV for thermal relics, $$v_{\min}(E_R) = \sqrt{\frac{m_N E_R}{2\mu_{\chi N}^2}}.$$1 for dark-matter–dark-radiation models, and $$v_{\min}(E_R) = \sqrt{\frac{m_N E_R}{2\mu_{\chi N}^2}}.$$2 for ultra-light axions in the range $$v_{\min}(E_R) = \sqrt{\frac{m_N E_R}{2\mu_{\chi N}^2}}.$$3 (Dvorkin et al., 2022). The broader 2019 CMB-S4 project plan situates these signatures within a reference-design and forecasting framework based on TT/TE/EE/BB and lensing covariance (Abazajian et al., 2019).

Indirect and accelerator programs use distinct but equally explicit flux models. GRAMS frames annihilation and decay through the usual $$v_{\min}(E_R) = \sqrt{\frac{m_N E_R}{2\mu_{\chi N}^2}}.$$4- and $$v_{\min}(E_R) = \sqrt{\frac{m_N E_R}{2\mu_{\chi N}^2}}.$$5-factor relations for MeV gamma rays, while antimatter searches use coalescence and Galactic propagation for antideuterons and antiprotons (Aramaki et al., 2019, Aramaki et al., 2020). BDX models dark bremsstrahlung in an electron beam dump and converts the resulting $$v_{\min}(E_R) = \sqrt{\frac{m_N E_R}{2\mu_{\chi N}^2}}.$$6 flux into calorimetric event counts through full GEANT4-based acceptance (Battaglieri et al., 2016). Astrophysical structure projects increasingly use simulation-based inference rather than closed-form likelihoods: Via adopts Approximate Bayesian Computation and Sequential Monte Carlo on summary statistics such as velocity-track power spectra and gap statistics, whereas DREAMS trains emulators and CNNs on thousands of Arepo+IllustrisTNG simulations that vary $$v_{\min}(E_R) = \sqrt{\frac{m_N E_R}{2\mu_{\chi N}^2}}.$$7, cosmology, and feedback parameters (Collaboration, 16 Jun 2026, Rose et al., 2024).

4. Representative project classes

A major class consists of same-target or material-controlled direct tests. COSINUS develops a cryogenic scintillating calorimeter with undoped NaI so that any modulation seen by DAMA/LIBRA can be tested with the same nuclei but with event-by-event phonon/light discrimination (Angloher et al., 2016). PICOLON and the earlier PICO-LON project attack the same problem through radiopure NaI(Tl), emphasizing suppression of $$v_{\min}(E_R) = \sqrt{\frac{m_N E_R}{2\mu_{\chi N}^2}}.$$8Pb and $$v_{\min}(E_R) = \sqrt{\frac{m_N E_R}{2\mu_{\chi N}^2}}.$$9K; PICOLON reports $$p_{\rm ann}=f_{\rm eff}\,\frac{\langle\sigma v\rangle}{m_\chi},$$0Pb $$p_{\rm ann}=f_{\rm eff}\,\frac{\langle\sigma v\rangle}{m_\chi},$$1, natK $$p_{\rm ann}=f_{\rm eff}\,\frac{\langle\sigma v\rangle}{m_\chi},$$2 ppb, and a measured background of $$p_{\rm ann}=f_{\rm eff}\,\frac{\langle\sigma v\rangle}{m_\chi},$$3 day$$p_{\rm ann}=f_{\rm eff}\,\frac{\langle\sigma v\rangle}{m_\chi},$$4 keV$$p_{\rm ann}=f_{\rm eff}\,\frac{\langle\sigma v\rangle}{m_\chi},$$5 kg$$p_{\rm ann}=f_{\rm eff}\,\frac{\langle\sigma v\rangle}{m_\chi},$$6 in $$p_{\rm ann}=f_{\rm eff}\,\frac{\langle\sigma v\rangle}{m_\chi},$$7 keVee for a single underground module, while PICO-LON reports $$p_{\rm ann}=f_{\rm eff}\,\frac{\langle\sigma v\rangle}{m_\chi},$$8Ra $$p_{\rm ann}=f_{\rm eff}\,\frac{\langle\sigma v\rangle}{m_\chi},$$9, $\sigma(v)=\sigma_0 v^n.$0Th $\sigma(v)=\sigma_0 v^n.$1, $\sigma(v)=\sigma_0 v^n.$2Pb $\sigma(v)=\sigma_0 v^n.$3, and $\sigma(v)=\sigma_0 v^n.$4 keV$\sigma(v)=\sigma_0 v^n.$5kg$\sigma(v)=\sigma_0 v^n.$6day$\sigma(v)=\sigma_0 v^n.$7 at $\sigma(v)=\sigma_0 v^n.$8 keVee (Fushimi et al., 2021, Fushimi et al., 2015).

A second class is proof-of-principle low-threshold target development. SWEET demonstrates that a monocrystalline sucrose absorber can operate as a phonon calorimeter with coincident scintillation readout; the first module used a $\sigma(v)=\sigma_0 v^n.$9 g crystal, ran for $95\%$0 h below $95\%$1 mK, achieved a $95\%$2 mV phonon baseline resolution, and observed a significant population of sugar–light coincidences (Bento et al., 29 Sep 2025). ALETHEIA treats liquid helium as a low-background, kinematically favorable target for low-mass WIMPs, with a $95\%$3 g prototype cooled to $95\%$4 K and dark current $95\%$5 pA up to $95\%$6 kV/cm (Liao et al., 2021, Liao et al., 2022). The semiconductor superlattice superstructure proposal instead engineers $95\%$7 in the few-hundred-meV range so that sub-MeV dark matter scattering on electrons could yield micrometer-wavelength photons, potentially read out with quantum cascade lasers (Bora, 2022). At still lower masses, coherent scattering with macroscopic objects becomes the signal channel: the asymmetric torsion-balance proposal exploits geometry-dependent form factors and projects the strongest $95\%$8 limits in the range $95\%$9 eV, using equal-mass tungsten cubes and shells whose differential acceleration sensitivity is quoted as $\sigma_0<6\times10^{-27}\,{\rm cm}^2$0 cm s$\sigma_0<6\times10^{-27}\,{\rm cm}^2$1 at $\sigma_0<6\times10^{-27}\,{\rm cm}^2$2 C.L. (Luo et al., 2024).

A third class consists of beam-dump and messenger programs. BDX proposes a downstream segmented CsI(Tl) calorimeter at Jefferson Lab, receiving up to $\sigma_0<6\times10^{-27}\,{\rm cm}^2$3 electrons-on-target in $\sigma_0<6\times10^{-27}\,{\rm cm}^2$4 days and targeting MeV–GeV dark matter through $\sigma_0<6\times10^{-27}\,{\rm cm}^2$5-electron and inelastic scattering, with a characteristic electromagnetic shower of a few hundreds of MeV and surrounding veto quietness (Battaglieri et al., 2016). The LHC proton beam-dump concept extends the same logic to $\sigma_0<6\times10^{-27}\,{\rm cm}^2$6 TeV protons, exploiting the IR6 dump to search for elastic and inelastic dark-sector signatures downstream (Kumar et al., 2016). GRAMS pushes test science to mission scale: a balloon/satellite LArTPC simultaneously targets the “MeV gap” in gamma-ray astronomy and low-energy antimatter, with antideuterons singled out as an essentially background-free channel at low kinetic energies (Aramaki et al., 2019, Aramaki et al., 2020).

A fourth class uses cosmological and astrophysical structure as the detector. CMB-S4 treats primary anisotropies and lensing as probes of annihilation, scattering, light relics, dark radiation, and ultra-light bosons (Dvorkin et al., 2022). Via converts stream perturbations into constraints on subhalos below the galaxy-formation threshold, forecasting sensitivity to $\sigma_0<6\times10^{-27}\,{\rm cm}^2$7 perturbers through $\sigma_0<6\times10^{-27}\,{\rm cm}^2$8 m s$\sigma_0<6\times10^{-27}\,{\rm cm}^2$9 radial-velocity stability and multi-epoch rejection of binaries (Collaboration, 16 Jun 2026). The “cluster test” uses merging clusters to infer self-interactions and relaxed spherical clusters to compare Maxwell–Boltzmann, Bose–Einstein, Fermi–Dirac, and NFW-like dark-matter phase-space models; in the cited A1835 analysis, all but the fermionic fit are formally rejected at over $m_\chi=1$0, and the acceptable fermionic interpretation corresponds to $m_\chi=1$1 eV/$m_\chi=1$2 for $m_\chi=1$3 (Nieuwenhuizen, 2017). DREAMS complements such observational tests with 1024 uniform-box and 1024 Milky Way zoom-in hydrodynamic simulations, explicit variation of $m_\chi=1$4 and IllustrisTNG feedback parameters, and machine-learning emulators that separate warm-dark-matter suppression from baryonic degeneracies (Rose et al., 2024).

5. Calibration, systematics, and reproducibility

The distinction between a proof of principle and a credible test science project is usually set by calibration and nuisance control. SWEET illustrates the problem sharply: the first sucrose run had no internal line source, no absolute energy scale, and pronounced $m_\chi=1$5 Hz noise, so the collaboration set a conservative $m_\chi=1$6 mV trigger threshold and used optimum filtering, pulse-quality cuts, and amplitude-correlation analysis rather than competitive exclusion limits (Bento et al., 29 Sep 2025). ALETHEIA similarly places R&D emphasis on high-voltage stability, electron extraction, TPB wavelength shifting, and low-temperature SiPM behavior before quoting physics sensitivity as more than a placeholder forecast (Liao et al., 2021, Liao et al., 2022). InDEx, a superheated $m_\chi=1$7 emulsion at the Jaduguda Underground Science Laboratory, uses Am–Be neutron calibration, acoustic $m_\chi=1$8, and FFT-based discrimination; its first run operated for $m_\chi=1$9 days at a $m_\chi>30\text{–}50$0 keV threshold with $m_\chi>30\text{–}50$1 kg-days exposure and reported a best spin-independent fluorine sensitivity of $m_\chi>30\text{–}50$2 at $m_\chi>30\text{–}50$3 GeV/$m_\chi>30\text{–}50$4 (Kumar et al., 31 Aug 2025).

Large-scale programs formalize the same issue differently. BDX anchors cosmogenic backgrounds with an INFN-LNS prototype and treats neutrinos as an irreducible beam-related floor in the high-threshold electron channel (Battaglieri et al., 2016). CMB-S4 dark-matter analyses rely on component separation, lensing reconstruction, and control of polarization calibration and beam systematics (Abazajian et al., 2019, Dvorkin et al., 2022). Via treats unresolved binaries, photospheric radial-velocity jitter, wavelength zeropoints, and baryonic perturbers as first-order nuisances, using three-epoch cadence and stream selection to suppress them (Collaboration, 16 Jun 2026). DREAMS explicitly quantifies the degeneracy between $m_\chi>30\text{–}50$5 and feedback parameters such as $m_\chi>30\text{–}50$6, $m_\chi>30\text{–}50$7, and $m_\chi>30\text{–}50$8, showing that astrophysical marginalization broadens, but does not erase, the dark-matter signal in satellite counts (Rose et al., 2024).

Open-science infrastructure has itself become a component of dark-matter test science. ESCAPE defines reproducibility through containerized notebooks, versioned datasets, provenance, the Virtual Research Environment, the Data Lake, and the Open-source Scientific Software and Service Repository, and it applies this to ATLAS reinterpretation, DarkSide exclusion curves, Fermi-LAT dwarf analyses, brown-dwarf gamma-ray searches, and KM3NeT–CTA multimessenger workflows (Pearson et al., 26 Sep 2025). This makes the workflow, not merely the final limit, part of the experimental object.

6. Scientific role, controversies, and outlook

Several prominent controversies now function as organizing targets for dark matter test science. The DAMA/LIBRA annual modulation remains the clearest example of a same-target imperative: COSINUS uses undoped NaI at millikelvin temperatures to add event-by-event identification, while PICOLON and PICO-LON pursue ever lower NaI(Tl) radioactivity to test whether a NaI signal survives once $m_\chi>30\text{–}50$9Pb- and $f_{\rm eff}\approx0.2$0K-dominated backgrounds are reduced (Angloher et al., 2016, Fushimi et al., 2021, Fushimi et al., 2015). The cluster self-interaction problem is another: merging systems have been interpreted as favoring $f_{\rm eff}\approx0.2$1, yet the cited cluster-test paper also notes criticisms of that interpretation and uses relaxed-cluster lensing plus X-ray data to challenge standard NFW-like fits (Nieuwenhuizen, 2017). These are not merely astrophysical disputes; they determine whether compact-object, self-interacting, or fermionic phase-space pictures remain viable.

The broader significance of dark matter test science lies in how it redistributes evidentiary burden. CMB-S4 can exclude or detect recombination-era energy injection and interaction signatures without relying on local halo uncertainties (Dvorkin et al., 2022). Via and DREAMS move the small-scale-structure program from qualitative “missing satellites” arguments to forward-modeled, uncertainty-marginalized inference (Collaboration, 16 Jun 2026, Rose et al., 2024). GRAMS and BDX emphasize orthogonal messengers and controlled sources, while torsion-balance and superlattice proposals open mass ranges in which conventional recoil detectors are intrinsically insensitive (Battaglieri et al., 2016, Aramaki et al., 2019, Luo et al., 2024, Bora, 2022).

This suggests that the future of the field will be increasingly project-ecological. The next generation is unlikely to be defined by a single dominant technology, but by interoperable programs in which cosmology, direct detection, indirect detection, accelerator searches, survey astrophysics, and simulation-based inference constrain one another. In that sense, the dark matter test science project has become both an experimental format and a method for organizing the search itself.