DEAP-3600: Liquid Argon DM Detector

• DEAP-3600 is a single-phase liquid argon dark matter detector at SNOLAB that utilizes pulse-shape discrimination and high photocathode coverage to target WIMP–nucleon scattering below 10⁻⁴⁶ cm² sensitivity.
• The detector features a 3.6-ton liquid argon target in a precision-machined acrylic vessel with a TPB wavelength shifter and multiple layers of active and passive shielding to reduce backgrounds.
• Advanced calibration, timing reconstruction, and multivariate background rejection techniques demonstrate DEAP-3600’s capability as both a sensitive dark matter search instrument and a precursor to next-generation detectors like ARGO.

DEAP-3600 is a single-phase liquid-argon direct-detection experiment for dark matter at SNOLAB in Sudbury, Canada. It was designed to search for spin-independent WIMP–nucleon scattering with a target sensitivity of $10^{-46}\,\mathrm{cm}^2$ at a WIMP mass of $100\,\mathrm{GeV}/c^2$, using a large spherical acrylic target, high photocathode coverage, strong passive shielding, active muon vetoing, and pulse-shape discrimination based on the time structure of liquid-argon scintillation (Cai, 2015). Across its design, commissioning, and operating phases, descriptions of the instrument cite a total liquid-argon mass of 3.6 t or an operating mass near 3.3 t, reflecting the distinction between design capacity and later fills; later operating papers quote $(3269 \pm 24)\,\mathrm{kg}$ or similar values for the active argon inventory (Seth, 28 Mar 2026).

1. Conception, site, and design objectives

DEAP-3600 emerged from the DEAP program’s single-phase liquid-argon strategy, in which scintillation-light timing rather than charge drift is used to discriminate nuclear recoils from electronic recoils. Early design papers describe a $3\,600\,\mathrm{kg}$ natural-liquid-argon target with a $1\,000\,\mathrm{kg}$ fiducial mass, housed in an acrylic vessel of inner radius $85\,\mathrm{cm}$ and intended to deliver a background-free exposure of $O(3\ \mathrm{tonne\cdot year})$ (Boulay, 2012). The experiment’s design requirement was that total expected background in the WIMP search window remain below $0.6$ events in $3\,000\,\mathrm{kg}\times\mathrm{yr}$, with surface $\alpha$ activity, neutron backgrounds, and residual $100\,\mathrm{GeV}/c^2$0 leakage each constrained to subdominant levels (Fatemighomi, 2016).

Its underground location at SNOLAB is integral to that design philosophy. Papers describing the detector place it approximately $100\,\mathrm{GeV}/c^2$1 underground, with overburden quoted as roughly $100\,\mathrm{GeV}/c^2$2, suppressing the cosmic-ray muon flux by more than six orders of magnitude or to values such as $100\,\mathrm{GeV}/c^2$3, depending on the analysis context (Kuźniak, 2021). Cleanliness control was treated as a detector-defining requirement rather than an auxiliary engineering constraint: assembly occurred in radon-reduced clean rooms, and radon-scrubbed nitrogen purges were used around argon-wetted volumes and during critical construction stages (Cai, 2015).

A common misunderstanding is that atmospheric-argon backgrounds would preclude a competitive liquid-argon WIMP search. DEAP-3600 was instead built on the premise that liquid argon’s scintillation timing permits sufficiently strong pulse-shape discrimination to render the intrinsic $100\,\mathrm{GeV}/c^2$4Ar activity manageable, provided that the light yield and radiopurity are high enough (Collaboration et al., 2014).

2. Detector architecture and instrumentation

The central detector element is a spherical acrylic vessel whose inner radius is consistently reported as about $100\,\mathrm{GeV}/c^2$5; construction papers describe a $100\,\mathrm{GeV}/c^2$6-thick acrylic sphere fabricated in multiple pieces, annealed, bonded, and precision-machined underground, with a cold radius of $100\,\mathrm{GeV}/c^2$7 (Amaudruz et al., 2017). Operational papers describe a liquid level above the equator and a gaseous-argon head space connected through an acrylic neck used for purification and cooling (Chen, 2020).

The vessel’s inner surface is coated with tetraphenyl butadiene, typically quoted as a $100\,\mathrm{GeV}/c^2$8 layer, to shift the $100\,\mathrm{GeV}/c^2$9 liquid-argon scintillation to visible light near $(3269 \pm 24)\,\mathrm{kg}$0 (Collaboration et al., 22 Jan 2025). The optical readout uses 255 inward-facing 8-inch Hamamatsu R5912 or R5912-HQE photomultiplier tubes, coupled through acrylic light guides of order $(3269 \pm 24)\,\mathrm{kg}$1–$(3269 \pm 24)\,\mathrm{kg}$2 length. These light guides perform several functions simultaneously: optical coupling, thermal decoupling of the room-temperature PMTs from the cryogenic argon, and moderation of radiogenic neutrons from the PMT assemblies (Amaudruz et al., 2017).

Descriptions of photocathode coverage cluster around $(3269 \pm 24)\,\mathrm{kg}$3–$(3269 \pm 24)\,\mathrm{kg}$4, with projected or measured light yields around $(3269 \pm 24)\,\mathrm{kg}$5 in design and commissioning documents and values near $(3269 \pm 24)\,\mathrm{kg}$6 or $(3269 \pm 24)\,\mathrm{kg}$7–$(3269 \pm 24)\,\mathrm{kg}$8 in later operating analyses, depending on calibration method and epoch (Fatemighomi, 2016). This spread reflects differences between projected performance, low-energy WIMP-search calibration conventions, and later energy-scale studies rather than a contradiction in the detector concept.

Outside the acrylic vessel, DEAP-3600 uses a stainless-steel enclosure and a large ultrapure-water tank instrumented as a Cherenkov muon veto. Water-tank dimensions are reported as about $(3269 \pm 24)\,\mathrm{kg}$9 diameter in design papers and as $3\,600\,\mathrm{kg}$0 diameter by $3\,600\,\mathrm{kg}$1 high in later analyses; the veto PMT count is reported as 48 in several operational descriptions (Adhikari et al., 2023). Construction papers also describe a dedicated neck-veto system using 100 Kuraray Y-11 wavelength-shifting fibers read out by 4 Hamamatsu R7600-300 PMTs to tag light produced in the neck region, which later became important because the neck was a persistent background locus (Amaudruz et al., 2017).

The readout chain evolved into a dual-gain waveform-digitization system using CAEN V1720 and V1740 modules, with custom signal-conditioning boards handling PMT high voltage and signal splitting (Collaboration et al., 22 Jan 2025). This architecture supported both the low-energy WIMP program and higher-energy calibration and rare-event analyses.

3. Scintillation physics and pulse-shape discrimination

DEAP-3600’s signal model is anchored in liquid argon’s excimer scintillation. Across the detector literature, the fast singlet lifetime is quoted near $3\,600\,\mathrm{kg}$2–$3\,600\,\mathrm{kg}$3 and the slow triplet lifetime near $3\,600\,\mathrm{kg}$4–$3\,600\,\mathrm{kg}$5, with nuclear recoils populating the singlet channel more strongly than electronic recoils (Lai, 2023). The corresponding PSD observable is the prompt-light fraction, usually denoted $3\,600\,\mathrm{kg}$6 and defined in DEAP-3600 analyses as

$3\,600\,\mathrm{kg}$7

where $3\,600\,\mathrm{kg}$8 is waveform charge in photoelectrons relative to trigger time (Collaboration et al., 22 Jan 2025).

The PSD performance reported for DEAP-3600 is exceptionally strong. Recent summaries state an electronic-recoil leakage probability of $3\,600\,\mathrm{kg}$9 at $1\,000\,\mathrm{kg}$0 for $1\,000\,\mathrm{kg}$1 nuclear-recoil acceptance, after Bayes-based deconvolution of after-pulses (Lai, 2023). Earlier status papers give $1\,000\,\mathrm{kg}$2 at a $1\,000\,\mathrm{kg}$3 threshold, while design studies and prototype extrapolations quote rejection levels better than $1\,000\,\mathrm{kg}$4 or even $1\,000\,\mathrm{kg}$5 under specific light-yield and threshold assumptions (Kuźniak, 2021).

A more detailed pulse-shape model was developed from DEAP-3600 data in the $1\,000\,\mathrm{kg}$6–$1\,000\,\mathrm{kg}$7 region. That model includes a singlet fraction $1\,000\,\mathrm{kg}$8 with $1\,000\,\mathrm{kg}$9, an intermediate recombination component with fraction $85\,\mathrm{cm}$0 and $85\,\mathrm{cm}$1, and a triplet fraction $85\,\mathrm{cm}$2 with $85\,\mathrm{cm}$3 (Collaboration et al., 2020). The same work identifies delayed TPB emission on $85\,\mathrm{cm}$4 timescales, finding that approximately $85\,\mathrm{cm}$5 of the wavelength-shifted light intensity resides in a long-lived TPB state. This delayed component contributes stray light to subsequent events, a detector effect with implications for energy reconstruction, low-PE PSD behavior, and pile-up handling (Collaboration et al., 2020).

This suggests that DEAP-3600’s PSD capability is not only a property of liquid argon itself but also of a coupled optical system involving TPB time response, PMT double-pulsing, afterpulsing, and the event-rate environment dominated by $85\,\mathrm{cm}$6Ar.

4. Radiopurity, background control, and contamination management

Background control in DEAP-3600 was organized around four principal classes: intrinsic $85\,\mathrm{cm}$7 backgrounds, radiogenic neutrons, surface and suspended $85\,\mathrm{cm}$8 emitters, and radon-related contamination. Material selection was based on HPGe $85\,\mathrm{cm}$9-spectrometry, radon-emanation measurements, and $O(3\ \mathrm{tonne\cdot year})$0-counting. Design and status papers specify that argon-wetted components were required to remain below a few $O(3\ \mathrm{tonne\cdot year})$1 of $O(3\ \mathrm{tonne\cdot year})$2Rn emanation, and that structural materials were screened for U/Th/K at the few $O(3\ \mathrm{tonne\cdot year})$3 level or lower (Fatemighomi, 2016).

The acrylic vessel was a special radiopurity focus. One design-era assay constrained bulk $O(3\ \mathrm{tonne\cdot year})$4Pb in acrylic to $O(3\ \mathrm{tonne\cdot year})$5, while resurfacing studies cite an upper limit of $O(3\ \mathrm{tonne\cdot year})$6 acrylic for the as-manufactured bulk (Cai, 2015). Because the vessel interior had been exposed to radon-bearing air, DEAP-3600 tracked integrated exposure histories and then deployed an in-situ resurfacer to remove the contaminated inner layer. Different papers quote removal depths of about $O(3\ \mathrm{tonne\cdot year})$7 or $O(3\ \mathrm{tonne\cdot year})$8, with post-machining residual surface activity reduced to about $O(3\ \mathrm{tonne\cdot year})$9 and a $0.6$0 reduction in surface alphas relative to pre-resurfacing estimates (Giampa, 2017).

The resurfacer itself was a substantial instrument. It used two sanding arms deployed through the detector neck on an 18 ft stainless-steel tube, spring-loaded against the acrylic surface and monitored with position sensors of $0.6$1 resolution (Giampa, 2017). Ultra-purified water flushing suppressed dust and transported acrylic slurry to external filtration and collection hardware. The motivation was explicit: the acrylic vessel inner surface was required to contribute no more than $0.6$2 background events in a three-year run, and the resurfacing campaign was intended to keep residual surface-$0.6$3 background below that budget (Giampa, 2017).

The electronic-recoil model was likewise elaborated in detail. A Bayesian template fit to a $0.6$4 dataset was used to decompose contributions from $0.6$5Ar, $0.6$6Ar/$0.6$7K, radon daughters in liquid argon, $0.6$8Pb at the LAr–TPB interface, and external $0.6$9 emitters in the acrylic vessel, light guides, PMTs, and stainless-steel shell (Ajaj et al., 2019). In that work, the specific activity of $3\,000\,\mathrm{kg}\times\mathrm{yr}$0Ar in atmospheric argon was measured through the daughter $3\,000\,\mathrm{kg}\times\mathrm{yr}$1K to be $3\,000\,\mathrm{kg}\times\mathrm{yr}$2 (Ajaj et al., 2019).

Later analyses identified more challenging backgrounds than the originally emphasized $3\,000\,\mathrm{kg}\times\mathrm{yr}$3 leakage. Status and recent-results papers describe the WIMP sensitivity as limited by $3\,000\,\mathrm{kg}\times\mathrm{yr}$4-induced scintillation in the neck region, produced when $3\,000\,\mathrm{kg}\times\mathrm{yr}$5Po on neck flowguides interacts with condensed argon films and generates partially shadowed light patterns that can leak into the nuclear-recoil band (Kuźniak, 2021). More recent profile-likelihood work identifies $3\,000\,\mathrm{kg}\times\mathrm{yr}$6 decays from a small number of dust particulates circulating within the liquid argon target as the dominant source of background events in that analysis (Collaboration et al., 14 Mar 2026). These developments are central to the detector’s later upgrade path.

5. Calibration, reconstruction, and analysis methodology

DEAP-3600 employed a layered calibration program combining optical systems, radioactive sources, and intrinsic isotopic activity. Construction and commissioning papers describe a laserball system using picosecond laser pulses at 375, 405, and $3\,000\,\mathrm{kg}\times\mathrm{yr}$7 to map channel gains, timing offsets, and optical properties, as well as LED-fiber systems at $3\,000\,\mathrm{kg}\times\mathrm{yr}$8 for routine monitoring (Amaudruz et al., 2017). Weekly timing and stability checks were also carried out with 20 acrylic-fiber AARF LEDs at $3\,000\,\mathrm{kg}\times\mathrm{yr}$9 (Cai, 2015).

For $\alpha$0 calibration, DEAP-3600 used a tagged $\alpha$1Na system deployed externally in calibration tubes around the steel shell. A detailed energy-calibration study reports nine fixed source positions, dual- and single-tag LYSO coincidence modes, a linear relation

$\alpha$2

and a weighted-average light yield corresponding to $\alpha$3, or $\alpha$4, with $\alpha$5 consistent with zero within $\alpha$6 (Luzzi, 17 Jan 2026). After applying position-dependent multiplicative corrections of $\alpha$7–$\alpha$8 to the Monte Carlo PE scale, the $\alpha$9Na full-energy peak positions agreed between data and simulation to better than $100\,\mathrm{GeV}/c^2$00 at all nine positions (Luzzi, 17 Jan 2026).

Position reconstruction became increasingly important as neck and surface backgrounds came to dominate the sensitivity budget. A dedicated photon-timing method used the first $100\,\mathrm{GeV}/c^2$01 hits in the first $100\,\mathrm{GeV}/c^2$02 after pulse start and constructed a likelihood in $100\,\mathrm{GeV}/c^2$03 from time-of-flight PDFs over trial cells in $100\,\mathrm{GeV}/c^2$04 and $100\,\mathrm{GeV}/c^2$05 (Chen, 2020). In the WIMP search region of 95–200 PE, corresponding to recoil energies of roughly $100\,\mathrm{GeV}/c^2$06–$100\,\mathrm{GeV}/c^2$07, that algorithm achieved a radial resolution of $100\,\mathrm{GeV}/c^2$08–$100\,\mathrm{GeV}/c^2$09 near the $100\,\mathrm{GeV}/c^2$10 fiducial cut and an angular resolution of about $100\,\mathrm{GeV}/c^2$11 (Chen, 2020).

The timing-based fit was used alongside a PE-based algorithm. Their differing biases near surfaces allowed consistency cuts such as $100\,\mathrm{GeV}/c^2$12 for bulk events; in the published 231-day search, this removed more than $100\,\mathrm{GeV}/c^2$13 of neck-$100\,\mathrm{GeV}/c^2$14 backgrounds, leaving a residual expectation of $100\,\mathrm{GeV}/c^2$15 events in a 231 live-day exposure (Chen, 2020). More recent detector summaries add a feed-forward neural-network position estimator that takes the 255-component PE vector as input, alongside maximum-likelihood charge-pattern fits using precomputed optical probability maps (Seth, 28 Mar 2026).

The statistical treatment of WIMP-search data also evolved. The 2026 profile-likelihood-ratio analysis used an unbinned likelihood in estimated energy, PSD parameter, and reconstructed radius, with additional binning in an auxiliary neck-discrimination variable, $100\,\mathrm{GeV}/c^2$16 (Collaboration et al., 14 Mar 2026). That approach was intended to exploit an increased fiducial volume and improved event-selection acceptance relative to earlier cut-based analyses.

6. Physics results and broader measurement program

The first DEAP-3600 dark-matter result, based on 4.44 live days and a fiducial exposure of $100\,\mathrm{GeV}/c^2$17 tonne-days, observed no candidate events in an ROI of 80–240 PE and set a $100\,\mathrm{GeV}/c^2$18 C.L. upper limit of $100\,\mathrm{GeV}/c^2$19 for a $100\,\mathrm{GeV}/c^2$20 WIMP (Collaboration et al., 2017). That initial analysis already demonstrated electronic-recoil leakage below $100\,\mathrm{GeV}/c^2$21 between 16 and $100\,\mathrm{GeV}/c^2$22 at $100\,\mathrm{GeV}/c^2$23 C.L., with $100\,\mathrm{GeV}/c^2$24 nuclear-recoil acceptance (Collaboration et al., 2017).

The later 231-day WIMP search, summarized in subsequent results papers, used a 1.9 t·yr exposure with 3260 kg fiducial mass, observed no events in the WIMP ROI after unblinding, and reported a total expected background below 0.1 events (Lai, 2023). Under the standard halo model, that analysis set

$100\,\mathrm{GeV}/c^2$25

at $100\,\mathrm{GeV}/c^2$26 C.L. (Lai, 2023). The 2026 profile-likelihood analysis, using 790.8 live-days of data with $100\,\mathrm{GeV}/c^2$27 of liquid argon and a $100\,\mathrm{GeV}/c^2$28 fiducial mass, reported an observed limit of $100\,\mathrm{GeV}/c^2$29 at $100\,\mathrm{GeV}/c^2$30 and improved exclusion limits between 20 and $100\,\mathrm{GeV}/c^2$31 on liquid argon (Collaboration et al., 14 Mar 2026).

DEAP-3600 results were also reinterpreted beyond the standard spin-independent isoscalar scenario. In a non-relativistic EFT framework, the collaboration constrained operators $100\,\mathrm{GeV}/c^2$32, $100\,\mathrm{GeV}/c^2$33, $100\,\mathrm{GeV}/c^2$34, $100\,\mathrm{GeV}/c^2$35, and $100\,\mathrm{GeV}/c^2$36 for isoscalar, isovector, and xenon-phobic couplings, and examined the influence of halo substructure models motivated by Gaia and SDSS (Adhikari et al., 2020). Later summaries state that DEAP-3600 set the world’s best exclusion limit for xenon-phobic dark-matter scenarios (Lai, 2023).

The detector also supported searches outside the canonical WIMP program. A dedicated multi-scatter analysis probed dark matter masses from $100\,\mathrm{GeV}/c^2$37 up to $100\,\mathrm{GeV}/c^2$38, identifying events with multiple peaks and low $100\,\mathrm{GeV}/c^2$39 due to successive delayed recoils. With zero candidates in 813 live-days, DEAP-3600 excluded composite “nugget” dark matter with $100\,\mathrm{GeV}/c^2$40 over $100\,\mathrm{GeV}/c^2$41–$100\,\mathrm{GeV}/c^2$42 (Lai, 2023).

A separate branch of the program used the detector’s very large atmospheric-argon inventory for isotope measurements. Using 167 live-days, DEAP-3600 measured the specific activity of atmospheric $100\,\mathrm{GeV}/c^2$43Ar to be

$100\,\mathrm{GeV}/c^2$44

(Adhikari et al., 2023). A later direct half-life measurement from 3.4 years of data reported

$100\,\mathrm{GeV}/c^2$45

for the $100\,\mathrm{GeV}/c^2$46Ar half-life (Collaboration et al., 22 Jan 2025). These results indicate that DEAP-3600 functioned not only as a dark-matter detector but also as a high-statistics liquid-argon isotope observatory.

7. Upgrades, residual limitations, and the transition toward ARGO

By 2021–2023, the principal limitation on DEAP-3600’s full design sensitivity was no longer electronic-recoil leakage but residual $100\,\mathrm{GeV}/c^2$47 backgrounds associated with the neck region and dust (Kuźniak, 2021). The corresponding hardware-upgrade program replaced the acrylic neck flowguides with guides coated in pyrene-doped polystyrene or pyrene-based slow wavelength shifter. Operational summaries report measured pyrene fluorescence lifetimes of $100\,\mathrm{GeV}/c^2$48 at $100\,\mathrm{GeV}/c^2$49 and $100\,\mathrm{GeV}/c^2$50 excitation, while more recent updates quote a slow wavelength-shifter timescale of $100\,\mathrm{GeV}/c^2$51 (Kuźniak, 2021). The purpose was to distinguish $100\,\mathrm{GeV}/c^2$52-induced neck scintillation from prompt nuclear recoils by PSD.

The same upgrade program moved the cooling function outside the neck or introduced external cooling in order to suppress liquid-argon film formation on the flowguides, and added liquid-argon extraction, recirculation, and filtration hardware to remove dust particulates (Seth, 28 Mar 2026). Multivariate background rejection was also developed. Proceedings describe a boosted-decision-tree discriminator combining reconstructed position, PMT-hit spatial asymmetry, $100\,\mathrm{GeV}/c^2$53, and PMT time structure; preliminary studies projected a further factor-of-10 reduction in neck-$100\,\mathrm{GeV}/c^2$54 leakage to below $100\,\mathrm{GeV}/c^2$55 events per year in the WIMP ROI at $100\,\mathrm{GeV}/c^2$56 nuclear-recoil acceptance (Lai, 2023).

DEAP-3600’s later role is also transitional. As part of the Global Argon Dark Matter Collaboration, it now serves as an operational and methodological precursor to ARGO, a next-generation liquid-argon detector under development with a 300-tonne fiducial mass (Seth, 28 Mar 2026). Design studies described in the 2026 update compare cylindrical and spherical acrylic-vessel geometries, with neutron-leakage simulations over a 3000 tonne·year exposure. In those studies, the spherical vacuum-cryostat design, denoted Geometry B, yielded a total expected neutron leakage of $100\,\mathrm{GeV}/c^2$57 events and was identified as meeting the $100\,\mathrm{GeV}/c^2$58 event requirement at 3000 t·y after cumulative cuts (Seth, 28 Mar 2026).

A plausible implication is that DEAP-3600’s long-term significance lies not only in its direct exclusion limits but also in its demonstration that a large single-phase liquid-argon detector can combine high light yield, precise timing, radiopure acrylic construction, and increasingly sophisticated reconstruction to support both rare-event searches and the design logic of a substantially larger successor.