DAMIC-M: Silicon CCD Dark Matter Detector

• DAMIC-M is a silicon CCD-based experiment that targets low-mass dark matter and hidden-sector particles using eV-scale ionization thresholds.
• Its innovative skipper amplifier technology enables multiple non-destructive charge measurements, achieving sub-electron resolution critical for detecting minute energy deposits.
• The detector employs advanced low-background techniques such as thick passive shielding and three-dimensional event reconstruction to optimize sensitivity for sub-GeV dark matter searches.

DAMIC-M, usually expanded as DArk Matter In CCDs at Modane, is a silicon CCD-based direct-detection experiment developed to search for very low-energy interactions from low-mass dark matter and related hidden-sector scenarios at the Laboratoire Souterrain de Modane in France. It is the next-generation continuation of the DAMIC program at SNOLAB, retaining the use of fully depleted, high-resistivity silicon as both target and imaging detector while replacing conventional CCD readout with skipper readout that provides single-electron or sub-electron charge resolution. Across its design, prototype, and early-physics literature, DAMIC-M is defined by eV-scale thresholds, very low dark current, three-dimensional event reconstruction from charge diffusion, and a staged move from a prototype Low Background Chamber to a larger underground detector optimized for sub-GeV dark-matter searches through both electronic and nuclear recoils (Settimo, 2018, Lee et al., 2020, Arnquist et al., 2022, Smida, 2024).

1. Lineage within the DAMIC program

DAMIC-M emerged from the CCD-based dark-matter program established by DAMIC at SNOLAB. In that predecessor configuration, the detector used seven fully depleted CCDs with a total active mass of 40 g; each CCD was a 16-Mpixel device with pixel size $(15\times15)\,\mu\mathrm{m}^{2}$ and thickness 675 $\mu$m, housed in a copper box cooled to 140 K inside vacuum and shielded by 18 cm of lead, with the innermost 2 inches made of ancient lead, plus 42 cm of polyethylene for neutron suppression (Settimo, 2018). The SNOLAB apparatus exploited the low nuclear mass of silicon for sensitivity to WIMPs in the range 1–10 GeV/$c^{2}$, while the silicon band gap provided sensitivity to dark-matter–electron interactions depositing as little as 1.1 eV in the target (Settimo, 2018).

The early DAMIC literature had already established the central experimental logic that later defined DAMIC-M: thick, fully depleted CCDs can deliver very low ionization thresholds, low noise, and a scalable silicon target for dark matter below 10 GeV (Collaboration et al., 2013). Subsequent DAMIC analyses at SNOLAB reported leakage current as low as $10^{-3}$ e$^-$/pix/day at 140 K, Gaussian pixel noise of about $\sim 1.6$ electrons, and a likelihood-based image scan with discriminant

$\Delta LL = \mathcal{L}_n - \mathcal{L}_G$

to separate noise from signal-like clusters (Settimo, 2018). In that context, preliminary analysis suggested an energy threshold of 50 eV$_{ee}$ with an expected leakage of only 0.01 noise events over the full data set, and background levels of 2 DRU for the CCD sandwiched between ancient-lead blocks and 5 DRU for the others in the 0.5–14.5 keV range (Settimo, 2018).

DAMIC-M was formulated as the next step beyond that 40 g SNOLAB array. The 2018 DAMIC paper described DAMIC-M as a Modane installation with a total mass of 1 kg, new CCDs with 36 Mpixels each, 1 mm thickness, and about 20 g mass per CCD, combined with skipper readout to target a 2-electron ionization threshold and background levels reduced to fractions of a DRU (Settimo, 2018). Later overview papers reiterated that the purpose of the upgrade was not merely increased target mass but the combination of larger exposure, sub-electron charge resolution, and lower radioactive and instrumental backgrounds (Settimo, 2020, Castello-Mor et al., 2020).

2. Detector principle and skipper-CCD technology

The detector principle is inherited from DAMIC: the bulk silicon of a fully depleted scientific CCD acts simultaneously as the interaction target and the sensor. Ionization charge created by a particle interaction drifts through the silicon bulk to the pixel plane; during drift, the charge diffuses laterally, so the final pixel cluster encodes depth information through its spatial extent (Lee et al., 2020). This depth sensitivity is one of the defining features of the DAMIC/DAMIC-M approach, because it enables three-dimensional reconstruction and rejection of surface backgrounds (Castello-Mor et al., 2020).

The CCDs used in DAMIC-M are described as high-resistivity n-type silicon devices with 15 $\mu$m $\times$ 15 $\mu$0m pixels and thicknesses around 670–675 $\mu$1m, fully depleted at substrate biases of at least 40 V or, in one prototype configuration, 70 V (Arnquist et al., 2023, Smida, 2024). DAMIC-M’s key innovation is the skipper amplifier, which permits multiple non-destructive charge measurements of the same pixel. In the design literature, the corresponding noise reduction is described by the familiar scaling

$\mu$2

or equivalently as noise decreasing approximately as the inverse square root of the number of repeated measurements (Arnquist et al., 2022, Arnquist et al., 2024).

This readout concept is what places DAMIC-M in the single-electron regime. The 2020 instrumentation paper reported prototype skipper performance with readout noise \SI{0.068}{\electron} and a demonstrated sub-electron value of

$\mu$3

stating that this performance, combined with low leakage current, would allow processes with collision energies as low as 1 eV to be observed (Lee et al., 2020). Another 2020 overview cited 0.07 e⁻ readout noise demonstrated with 4000 samples per pixel, and also gave the ionization calibration relation 3.8 eV_ee = 1e^- (Castello-Mor et al., 2020). In the first underground DAMIC-M runs, the collaboration used 650 skipper samples per pixel, reaching a charge resolution of about 0.2 $\mu$4 (Arnquist et al., 2023). In the later pattern-based prototype search, the quoted charge resolution was

$\mu$5

which is the regime needed to distinguish 1-, 2-, 3-, and 4-electron deposits (Collaboration et al., 18 Mar 2025).

The choice of silicon is central to the physics program. The experiment is explicitly designed to search for dark matter too light to produce conventional detectable nuclear recoils in many other detectors, making dark matter–electron scattering a primary channel (Collaboration et al., 18 Mar 2025). At the same time, the DAMIC literature consistently emphasizes that the same detector architecture remains sensitive to low-mass WIMPs through nuclear recoils, to hidden-photon absorption, and to other low-energy processes in silicon (Lee et al., 2020, Castello-Mor et al., 2020).

3. Experimental realizations at Modane

DAMIC-M is operated at the Laboratoire Souterrain de Modane (LSM), beneath roughly 4800 m.w.e. of overburden. The Low Background Chamber paper describes LSM as about 1700 m below the Fréjus Peak, with cosmic-ray muons suppressed to 5.4 $\mu$6/m$\mu$7/day, and notes that the DAMIC-M infrastructure includes an ISO 5 cleanroom and ISO 6 gowning room built for underground handling of CCDs and electronics (Arnquist et al., 2024). Environmental measurements reported there include a median radon concentration of $\mu$8 Bq/m$\mu$9, with 99% of values below 42 Bq/m$c^{2}$0, as well as gamma and neutron flux measurements relevant to low-background operation (Arnquist et al., 2024).

The first underground DAMIC-M prototype was the Low Background Chamber (LBC). In its original configuration, the LBC used two 6k $c^{2}$1 4k CCDs with 669 $c^{2}$2m thickness and active mass about 8.9 g per CCD setup (Arnquist et al., 2024). The first DAMIC-M dark-matter search, however, is described in the collaboration’s 2023 results paper as using two large, thick skipper CCDs in the LBC, each about 9 g and 670 μm thick, for a total integrated exposure of 85.23 g days split between SR1 and SR2 (Arnquist et al., 2023). The daily-modulation paper based on the same early period specifies 63 days of uninterrupted data starting on June 8, 2022, with a final clean exposure of 39.97 g-days over 8779 images (Arnquist et al., 2023).

The LBC later evolved toward the final DAMIC-M module format. A second configuration installed two CCD modules, each with four 6k $c^{2}$3 1.5k CCDs mounted on silicon pitch adapters; the LBC paper states that these modules are close to what will be used in the final detector (Arnquist et al., 2024). The 2025 benchmark-model search used this module-style prototype geometry: two CCD modules, eight CCDs total, and six CCDs retained for analysis after excluding two problematic devices. Each of those CCDs had

$c^{2}$4

pixel size $c^{2}$5, thickness $c^{2}$6, and mass $c^{2}$7 g (Collaboration et al., 18 Mar 2025).

Published descriptions of the full detector evolved across stages. Early design documents described a kg-scale array of 50 CCDs with a total active mass of about 1 kg, using 1 mm-thick, 36 Mpixel sensors of about 20 g each (Settimo, 2018, Castello-Mor et al., 2020). A 2020 instrumentation paper described a final detector of 50 large-area skipper CCDs with more than 36 million pixels in each CCD (Lee et al., 2020). By 2022 and 2024, status papers described a detector with about 200 or 208 large-format skipper CCDs of roughly 3.3–3.5 g each, totaling about 700 g, while abstracts still referred to “about 1 kg” of silicon target mass (Arnquist et al., 2022, Smida, 2024). The first production paper then reported the fabrication of 28 CCD modules, each containing four 9-megapixel skipper CCDs, with 26 modules—about 350 g active mass—selected for the first underground deployment planned for early 2026 (Lin et al., 8 Sep 2025).

4. Background control, imaging, and detector characterization

Background suppression in DAMIC-M combines passive shielding, material radiopurity, optical and thermal control, and event-topology reconstruction. The LBC design uses a cold copper housing, an inner shield of at least 6 cm of lead around the CCD box, the innermost 2 cm made of ancient Roman lead, and an external shield of 15–20 cm lead and HDPE, depending on location (Arnquist et al., 2024). In the prototype program, external shielding reduced the event rate by a factor of 50; the background between 1 and 6 keV excluding silicon K lines decreased from $c^{2}$8 dru to $c^{2}$9 dru, closely matching the Geant4 prediction of 5.46 dru for the mitigation step that was applied (Arnquist et al., 2024).

Material assays and cosmogenic-activation control are treated as central engineering constraints. The LBC paper reports assay values in mBq/kg for CCD flex, OFHC copper, electro-formed copper, and other components, and identifies the replacement of OFHC copper lids by electro-formed copper as a major background improvement (Arnquist et al., 2024). The module-production paper makes the same point at the scale of detector fabrication: the design budget required an overall background below

$10^{-3}$0

with each major source contributing at most about

$10^{-3}$1

to the region of interest (Lin et al., 8 Sep 2025). That paper also quantified residual contamination from production, including a cosmogenic tritium increase of

$10^{-3}$2

front-surface $10^{-3}$3Pb of

$10^{-3}$4

and particulate activities from wire bonding that were reported as acceptable relative to the DAMIC-M budget (Lin et al., 8 Sep 2025).

At the analysis level, DAMIC-M relies on the fact that a bulk interaction in a fully depleted CCD generally diffuses into a small cluster rather than an isolated pixel. The 2025 prototype search therefore abandoned a purely single-pixel strategy for its main analysis and instead looked for two or three consecutive pixels in the same row, with total charge between 2 and 4 electrons. The searched patterns were

$10^{-3}$5

with pattern-identification variables $10^{-3}$6 and $10^{-3}$7, and thresholds $10^{-3}$8 and $10^{-3}$9 that yielded about 90% pattern-classification efficiency (Collaboration et al., 18 Mar 2025). The same paper imposed further conditions—no additional charge above $^-$0 in adjacent rows, no corresponding charge in the other CCDs of the same module, and no extra $^-$1 pixel in candidate columns—to reduce random coincidences and correlated noise backgrounds (Collaboration et al., 18 Mar 2025).

Temporal stability and non-ionizing signatures have also been studied as detector observables. A 2025 daily-modulation analysis found a correlation between the single-electron rate and the external system temperature, modeled as

$^-$2

and attributed to infrared photons from room-temperature components (Aggarwal et al., 17 Nov 2025). Separately, a 2022 radiation-damage study irradiated a 24-megapixel DAMIC-M CCD with an AmBe neutron source and reported, for the first time, that individual defects produced by nuclear recoils can be identified in a DAMIC-M CCD. That paper found that at least 80% of nuclear recoils above $^-$3 were spatially correlated with a defect appearing after irradiation, suggesting a possible route to nuclear/electron recoil discrimination in some energy ranges (Lee, 2022). This suggests an additional analysis channel beyond ionization topology alone.

5. Searches and published physics results

The first underground DAMIC-M search for dark matter interacting with electrons used an integrated exposure of

$^-$4

split into 45.26 g-days in SR1 and 39.97 g-days in SR2 (Arnquist et al., 2023). The analysis searched pixel charges up to 7e$^-$5, calibrated the charge scale using the 0e, 1e, 2e peaks, and performed a joint binned likelihood fit over four charge distributions, one for each amplifier in each run (Arnquist et al., 2023). No preference for a dark-matter signal was found, and the collaboration set 90% C.L. limits on dark-matter–electron scattering over the mass range 0.53 to 1000 MeV/$^-$6, excluding unexplored parameter space in [1.6,1000] MeV/$^-$7 for ultralight mediators and [1.5,15.1] MeV/$^-$8 for heavy mediators (Arnquist et al., 2023).

A distinct line of analysis used time dependence rather than total rate. The 2023 daily-modulation study searched the SR2 sample for a sidereal modulation of the 1$^-$9 event rate, motivated by the possibility that sufficiently strongly interacting dark matter may scatter in Earth’s bulk before detection (Arnquist et al., 2023). Its model-independent fit used

$\sim 1.6$0

and found no evidence for modulation at the sidereal period, $\sim 1.6$1 h (Arnquist et al., 2023). Interpreted in a dark-photon-mediated DM–electron scattering model, the same 39.97 g-day dataset yielded exclusion limits for masses in the range 0.53 to 2.7 MeV/$\sim 1.6$2, improving the collaboration’s previous total-rate limit by about 2 orders of magnitude and, according to the paper, constituting the current strongest limit on DM-electron scattering via ultralight mediators around $\sim 1.6$3 MeV/$\sim 1.6$4 (Arnquist et al., 2023).

The 2025 prototype search substantially increased the exposure and lowered the single-electron background rate. Using 84 days of data between October 2024 and January 2025, the collaboration obtained D1: 0.139 kg-day and D2: 1.257 kg-day, for a total of about 1.3 kg-day, while reporting a factor of 50 reduction in the single-electron rate relative to the earlier DAMIC-M search (Collaboration et al., 18 Mar 2025). The measured single-electron rate in D2 was

$\sim 1.6$5

equivalent to roughly 350–460 $\sim 1.6$6/g/day (Collaboration et al., 18 Mar 2025). In the blinded D2 sample, the analysis found 144 candidates for $\sim 1.6$7, 1 candidate for $\sim 1.6$8, and 0 candidates for $\sim 1.6$9, compared with expected backgrounds of 141.5 for total $\Delta LL = \mathcal{L}_n - \mathcal{L}_G$0-like backgrounds and 0.071 for $\Delta LL = \mathcal{L}_n - \mathcal{L}_G$1 (Collaboration et al., 18 Mar 2025). The paper concluded that there was no evidence for a dark matter signal and set 90% C.L. upper limits for masses between 1 and 1000 MeV/$\Delta LL = \mathcal{L}_n - \mathcal{L}_G$2 (Collaboration et al., 18 Mar 2025).

That null result was then mapped onto benchmark hidden-sector models. For an ultra-light mediator, the paper states that DAMIC-M excludes dark matter as the dominant component of cosmological dark matter via freeze-in for masses in the range

$\Delta LL = \mathcal{L}_n - \mathcal{L}_G$3

For the heavy-mediator benchmark of complex scalar dark matter freezing out in the early Universe, DAMIC-M excludes the scenario for masses between

$\Delta LL = \mathcal{L}_n - \mathcal{L}_G$4

and notes that only a narrow region around 25 MeV/$\Delta LL = \mathcal{L}_n - \mathcal{L}_G$5 remains viable in that benchmark when all relevant constraints are combined (Collaboration et al., 18 Mar 2025).

A later daily-modulation analysis used the 1.257 kg-day D2 dataset collected with the LBC and searched for periodicity in the single-electron rate over periods from 1 to 48 h (Aggarwal et al., 17 Nov 2025). After temperature-dependent background subtraction and periodogram analysis, it reported no significant modulation and emphasized the detector’s temporal stability (Aggarwal et al., 17 Nov 2025). In a complementary model-dependent analysis of Hidden Sector dark matter with masses in [0.53,2] MeV/$\Delta LL = \mathcal{L}_n - \mathcal{L}_G$6, the paper states that the resulting limits improve over previous DAMIC-M constraints by up to about two orders of magnitude below 1.2 MeV/$\Delta LL = \mathcal{L}_n - \mathcal{L}_G$7, and are the strongest published limits for Galactic halo DM in the mass range 0.53–1.22 MeV/$\Delta LL = \mathcal{L}_n - \mathcal{L}_G$8 (Aggarwal et al., 17 Nov 2025).

6. Theoretical interpretation and position in sub-GeV dark-matter phenomenology

Recent phenomenology has treated DAMIC-M as a benchmark experiment for sub-GeV direct detection through electron recoils. A 2025 comparative analysis states that the DAMIC-M collaboration had “recently reported impressive bounds on sub-GeV dark matter” that “robustly test both thermal and non-thermal models for the very first time” (Cheek et al., 21 Jul 2025). In that paper, PandaX-4T S2-only data are found to compete with the DAMIC-M results, providing the best constraints for scalar and asymmetric thermal dark matter models between 20 and 200 MeV, while DAMIC-M remains stronger for light or massless mediators (Cheek et al., 21 Jul 2025). The comparison makes explicit that DAMIC-M occupies the low-threshold, low-mass end of the current direct-detection landscape rather than a merely ancillary role.

The formalism used in DAMIC-M analyses is also standard within semiconductor dark-matter searches. The collaboration’s electron-scattering papers write the differential rate in the schematic form

$\Delta LL = \mathcal{L}_n - \mathcal{L}_G$9

with

$_{ee}$0

where $_{ee}$1 corresponds to a heavy mediator and $_{ee}$2 to an ultra-light mediator (Arnquist et al., 2023, Collaboration et al., 18 Mar 2025). Because DAMIC-M can resolve individual or few-electron deposits, it is especially relevant to models in which the visible signal is concentrated in the lowest ionization multiplicities.

Model-building papers from 2025 and 2026 sharpened this role. One study of light thermal dark matter in the light of DAMIC-M 2025 constraints concluded that the new electron-recoil limits are among the most important tests of sub-GeV thermal dark matter, but do not eliminate all viable models; instead they leave only narrow, predictive regions in the examples considered, such as a resonant $_{ee}$3 band with

$_{ee}$4

and typically

$_{ee}$5

(Borah et al., 19 Sep 2025). Another 2026 study used a projected DAMIC-M sensitivity to intermediate-mass mediators and argued that, for freeze-in-sized couplings, DAMIC-M is sensitive to dark-photon masses up to about

$_{ee}$6

highlighting that the experimentally relevant distinction between light and heavy mediators spans a broad intermediate regime rather than a single crossover scale (Stratman et al., 11 May 2026).

In cosmological applications, DAMIC-M constraints have been recast into statements about production mechanisms and early-Universe history. A 2026 freeze-in analysis states that recent DAMIC-M and PandaX results exclude the standard freeze-in production of dark matter for masses in the range

$_{ee}$7

in an ultra-light $_{ee}$8 gauge-boson framework, and further argues that direct-detection limits on $_{ee}$9 can translate into lower bounds on the reheating temperature in low-$\mu$0 freeze-in scenarios (Bertou et al., 10 Jun 2026). This suggests that DAMIC-M has become relevant not only for detector-scale sensitivity studies but also for testing the cosmological consistency of hidden-sector constructions.

A recurrent misconception is to treat DAMIC-M as simply a larger DAMIC. The published record does not support that reduction. From the earliest design papers onward, the defining changes are the move to skipper readout, the attempt to reach single-electron sensitivity, the redesign of packaging and materials to reduce backgrounds, and the use of imaging information—cluster morphology, diffusion-based depth reconstruction, and, in some studies, time-domain modulation or defect formation—as part of the signal discrimination strategy (Settimo, 2018, Lee et al., 2020, Arnquist et al., 2024). In that sense, DAMIC-M is best understood as the CCD rare-event program’s transition from low-threshold proof of principle to a dedicated sub-GeV dark-matter experiment whose prototype data already constrain benchmark thermal and non-thermal models.