Oscura Experiment: Low-Threshold Dark Matter Search

• Oscura Experiment is a direct-detection setup using multi-kilogram silicon Skipper-CCD sensors to probe sub-GeV dark matter–electron interactions with exceptional sensitivity.
• Its advanced design achieves single-electron resolution and ultra-low noise levels, ensuring background rates below one event per exposure.
• The project also targets millicharged particles, axion-like particles, and coherent neutrino scattering, setting a new standard for scalable low-threshold detection.

Oscura is a multi-kilogram direct-detection experiment optimized for the search for sub-GeV dark matter–electron interactions using the lowest-noise silicon Skipper-CCD technology. The experiment will deploy a 10 kg array of fully-depleted, high-resistivity silicon devices with single-electron sensitivity and stringent background controls, targeting less than one background event per exposure in the search region. Besides dark matter, Oscura provides a leading platform for studies of millicharged particles, axion-like particles (ALPs), and coherent neutrino scattering. It inherits core detector concepts and background models from SENSEI and DAMIC-M, extending these to the regime of maximized target mass, ultra-low instrumental backgrounds, and scalable readout at deep underground sites.

1. Scientific Motivation and Physics Targets

Oscura’s principal physics motivation is to explore light dark matter (DM) in the mass range $$0.5\,\mathrm{MeV} \lesssim m_\chi \lesssim 5\,\mathrm{GeV}$$, which is kinematically inaccessible to nuclear recoil searches. Silicon Skipper-CCDs allow the detection of electron recoils down to $\sim1$–$2\,e^{-}$, corresponding to an energy threshold $\sim3.6~\mathrm{eV}$, enabling direct searches for:

• Elastic DM–electron scattering: Provides sensitivity to $\bar{\sigma}_e$ as low as $10^{-44}\ \mathrm{cm}^2$ for $m_\chi \sim 1~\mathrm{MeV}$. Required exposure of $30~\mathrm{kg}\cdot\mathrm{year}$ is designed to yield $<$1 background event above a $2\,e^{-}$ threshold (Aguilar-Arevalo et al., 2022, Cervantes-Vergara et al., 2022).
• Bosonic dark matter absorption: Covers dark photons and axion-like particles with masses down to the silicon bandgap ($\sim1.1\,\mathrm{eV}$), with projected sensitivity to kinetic-mixing parameter $\varepsilon\sim 10^{-16}$–$10^{-14}$ (Aguilar-Arevalo et al., 2022).
• Millicharged particles (mCPs): Oscura and its precursor (the Oscura Integration Test, OIT) provide a world-leading probe for MeV-scale mCPs produced in photoproduction and Drell–Yan processes in accelerator and reactor environments, with reach down to $\epsilon\sim$ a few $\times 10^{-4}$ (Perez et al., 2023).
• Neutrino and new-physics probes: Single-electron sensitivity could also open channels for coherent elastic neutrino–nucleus scattering and searches for light mediators or new interactions in reactor or solar neutrino experiments (Cervantes-Vergara et al., 2022).

2. Detector Design and Technology

Skipper-CCD Sensors

The Oscura array will comprise approximately 20,000 Skipper-CCD sensors fabricated on 200 mm high-resistivity wafers, each $1.9~\mathrm{cm} \times 1.6~\mathrm{cm}$ and $675$–$725~\mu\mathrm{m}$ thick, fully depleted at voltages $\lesssim -60~\mathrm{V}$ (Cervantes-Vergara et al., 2022, Cervantes-Vergara et al., 2023). The design, originally developed at LBNL, emphasizes:

• Single-electron resolution: Achieved using non-destructive multi-sampling (“Skipper” readout with $N$ repeated measurements), yielding noise as low as $0.087\,e^-$ RMS ($N=1225$), with clean separation of integer $e^-$ peaks.
• Thermal dark current: Controlled to $\lesssim 10^{-6}\,e^-/\mathrm{pix}/\mathrm{day}$ by operating at $T \sim 130$–$140~\mathrm{K}$ in LN$_2$; prototype sensors achieved $(0.031\pm0.013)~e^{-}/\mathrm{pix}/\mathrm{day}$ at $140$ K (Cervantes-Vergara et al., 2022), with systematic underground rates $\sim1.8\times10^{-3}~e^-/\mathrm{pix}/\mathrm{day}$ so far dominated by exposure-dependent spurious charge from traps (Perez et al., 2024).

Modular Assembly and Readout

• Module architecture: Skipper-CCDs are arranged in Multi-Chip Modules (MCMs, 16 sensors each) and Super-Modules (16 MCMs per SM), with modules integrated onto low-radioactivity copper frames and read out in parallel (Perez et al., 2023).
• Cold readout electronics: Analog front-ends are implemented using low-noise ASICs (e.g., MIDNA), engineered for sub-$0.2\,e^-$ RMS noise at $120$–$140~\mathrm{K}$, with full array readout time $\lesssim2$ hrs to prevent pileup from dark current (Aguilar-Arevalo et al., 2022).
• Cryogenic/Mechanical: The array is housed in a $1~\mathrm{m}$-scale pressure vessel submerged in LN$_2$ at $450$ psi for temperature stabilization, with layered shielding (Pb, Cu, HDPE) for passive background suppression.

3. Backgrounds: Sources, Measurements, and Control

Major Sources

Source	Typical Target Value	Mitigation Strategy
Thermal dark current	$<10^{-6}\ e^-$/pix/day	130–140 K operation; short exposures; optimized clocking
Spurious charge (clocking)	$\kappa_{\rm SC} < 4\times10^{-11}\ e^-$/pix/transfer (goal)	Binning, clock shaping, low-SC designs
Trap-induced deferred charge	$<0.12$ traps/pix, $\tau_e < t_{\rm pix}$	Fabrication control, pocket-pumping, masking
Radiogenic/External	$<0.01$ dru in region of interest	Radiopure Cu, underground siting, thick passive shields
Surface NIR photons	$<10^{-2}$ dru	Thin backside passivation, light-tight packaging
Cosmogenic activation	$<5$ d above-ground exposure	Controlled logistics, tritium bakeout

Trap Characterization and Impact

Deferred charge due to deep-level traps (mainly at $E_t\sim0.31$–$0.34~\mathrm{eV}$ with $\sigma\sim0.7$–$3.5\times10^{-15}\,\mathrm{cm}^2$) was systematically studied by pocket-pumping, revealing strong batch dependence (Perez et al., 2024). Dominant trap populations induce “tails” of delayed single-electron hits, inflating the measured single-electron rate (SER) by an order of magnitude above intrinsic thermal dark current—Monte Carlo simulation shows that even with efficient masking, trap contributions dominate SER at $$(1.5\pm0.2)\times10^{-3}~e^-/\mathrm{pix}/\mathrm{day}$$ under current underground conditions.

Mitigation centers on batch process control (e.g., gettering, contamination management), trap-map calibration, dynamic masking, and possibly tuning $T$ and $N_{\rm samples}$ to optimize the tail profile and minimize uncorrectable events.

4. Experimental Program: Scaling, Prototypes, and Performance

Integration Test (OIT)

As an intermediate milestone, the Oscura Integration Test (OIT) will deploy $\sim$1 kg of Skipper-CCDs at Fermilab for both detector validation and physics output (Perez et al., 2023). With single-electron noise floors of $\lesssim0.12~e^-$ and heavy lead shielding, OIT will deliver leading limits on MeV-scale millicharged particles ($\epsilon\gtrsim5\times10^{-5}$ for $m_\chi < 100~\mathrm{MeV}$), leveraging the NuMI beam environment.

Projected Array Performance

Full-array tests with 150–200 mm wafer-scale devices show:

• 71% yield of packaged sensors supporting robust single-electron counting (Cervantes-Vergara et al., 2022).
• Readout noise: $0.087~e^-$ RMS ($N=1225$); target for $0.16~e^-$ at moderate $t_{\rm pix}$ for array scaling (Cervantes-Vergara et al., 2023).
• Dark current: surface measurements already achieve $0.03~e^-/\mathrm{pix}/\mathrm{day}$ at $140~\mathrm{K}$ (higher than expected underground due to cosmic rays); demonstrated $95\%$ light suppression with thin aluminum layers.

Projected background over $30~\mathrm{kg}\cdot\mathrm{year}$ with measured spurious charge and DC parameters yields $<$10 events above $4\,e^-$, maintaining essentially zero-background performance above $4\,e^-$ (Cervantes-Vergara et al., 2023).

5. Sensitivity and Physics Reach

Oscura’s anticipated $30~\mathrm{kg}\cdot\mathrm{year}$ exposure at a $2\,e^-$ threshold will probe DM-electron cross sections to $\bar{\sigma}_e \sim 10^{-44}\ \mathrm{cm}^2$ (heavy mediator) and DM absorption to mixing angles $\epsilon \sim 10^{-16}$ for bosonic DM. Performance at these thresholds is enabled by both the low instrumental background and the suppression of the irreducible dark current to $\mathcal{O}(10^{-6})~e^-/\mathrm{pix}/\mathrm{day}$ (Aguilar-Arevalo et al., 2022, Cervantes-Vergara et al., 2022).

Auxiliary programs include:

• Millicharged particle searches: OIT and the full experiment set world-leading laboratory exclusions for $1~\mathrm{MeV} < m_\chi < 200~\mathrm{MeV}$, especially using track-based doublet/triplet searches insensitive to noise backgrounds (Perez et al., 2023).
• Reactor ALP search via plasmon excitation: For a $30~\mathrm{kg}\cdot\mathrm{yr}$ exposure at $L=10~\mathrm{m}$ from a $4~\mathrm{GW}_\mathrm{th}$ reactor and a $10~\mathrm{eV}$ threshold, Oscura is projected to surpass current NEON experiment ALP–photon coupling limits by an order of magnitude, with $$g_{a\gamma}\lesssim3\times10^{-8}\ \mathrm{GeV}^{-1}$$ at $m_a\sim10~\mathrm{keV}$ (Gong et al., 12 Jan 2026).

6. Technical Challenges and R&D Pathways

Key open challenges include:

• Sensor fabrication yield: Ongoing process optimization (buried-channel dose control, contamination tracking) aims to move from 71% to $>$90% yield (Cervantes-Vergara et al., 2022).
• Trap-induced backgrounds: Emphasis on minimizing defect populations by refining cleanroom protocols and gettering, with continual in situ calibration via pocket-pumping to enable dynamic mitigation strategies (Perez et al., 2024).
• Readout scalability and multiplexing: Integrated ASICs, optimized for low radioactivity and cable mass, are in advanced development to read out $>24,000$ channels at cryogenic temperatures with sub-$0.2~e^-$ noise (Cervantes-Vergara et al., 2023).
• Material radioassay and logistics: Strict selection and underground electroforming of Cu, radiopure adhesives, and minimized exposure timelines for tritium control are critical to meeting background goals (Aguilar-Arevalo et al., 2022).

7. Future Prospects and Broader Impact

Oscura is scheduled for completion at SNOLAB, targeting full-scale construction post-2026 and three years of stable science data (Aguilar-Arevalo et al., 2022). Its fundamental approach—leveraging single-electron Skipper-CCD detectors for zero-background rare-event searches at scale—paves a path for broader application across low-threshold neutrino physics, ALP searches, and new techniques for cosmogenic background control.

The project’s methodology, especially in detector scaling, advanced surface and bulk event rejection, and model-independent eV-scale electron counting in macroscopic arrays, is poised to establish a benchmark for the next generation of rare-event silicon experiments and may directly inform future designs for nucleon- and light-mediator dark matter searches, as well as new approaches in quantum sensor development (Aguilar-Arevalo et al., 2022, Gong et al., 12 Jan 2026, Perez et al., 2023).