The DAMIC dark matter experiment

The DAMIC dark matter experiment J.R.T. de Mello Neto f ∗ on behalf of the DAMIC Collaboration

a

Universidad Nacional Autónoma de México, México D.F., México University of Michigan, Department of Physics, Ann Arbor, MI, United States c Centro Atómico Bariloche - Instituto Balseiro, CNEA/CONICET, Argentina d Fermi National Accelerator Laboratory, Batavia, IL, United States e Kavli Institute for Cosmological Physics and The Enrico Fermi Institute, The University of Chicago, Chicago, IL, United States f Universidade Federal do Rio de Janeiro, Instituto de Física, Rio de Janeiro, RJ, Brazil g Universität Zürich Physik Institut, Zurich, Switzerland h SNOLAB, Lively, ON, Canada i Facultad de Ingeniería - Universidad Nacional de Asunción, Paraguay j Northern Illinois University, DeKalb, IL, United States E-mail: [email protected] b

The DAMIC (Dark Matter in CCDs) experiment uses high resistivity, scientific grade CCDs to search for dark matter. The CCD’s low electronic noise allows an unprecedently low energy threshold of a few tens of eV that make it possible to detect silicon recoils resulting from interactions of low mass WIMPs. In addition the CCD’s high spatial resolution and the excellent energy response results in very effective background identification techniques. The experiment has a unique sensitivity to dark matter particles with masses below 10 GeV/c2 . Previous results have demonstrated the potential of this technology, motivating the construction of DAMIC100, a 100 grams silicon target detector currently being installed at SNOLAB. In this contribution, the mode of operation and unique imaging capabilities of the CCDs, and how they may be exploited to characterize and suppress backgrounds will be discussed, as well as physics results after one year of data taking. The 34th International Cosmic Ray Conference, 30 July- 6 August, 2015 The Hague, The Netherlands ∗ Speaker.

c Copyright owned by the author(s) under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike Licence.

http://pos.sissa.it/

PoS(ICRC2015)1221

A. Aguilar-Arevaloa , D. Amideib , X. Bertouc , D. Boleb , M. Butnerd, j , G. Cancelod , A. Castañeda Vázqueza , A.E. Chavarriae , S. Dixone , J.C. D’Olivoa , J. Estradad , G. Fernandez Moronid , K.P. Hernández Torresa , F. Izraelevitchd , A. Kavnerb , B. Kilminsterg , I. Lawsonh , J. Liaog , M. Lópezi , J. Molinai , G. Moreno-Granadosa , J. Penae , P. Priviterae , Y. Sarkisa , V. Scarpined , T. Schwarzb , M. Sofo Haroc , J. Tiffenbergd , D. Torres Machado f , F. Trillauda , X. You f and J. Zhoue

J.R.T. de Mello Neto f

The DAMIC dark matter experiment

1. Introduction

2. The DAMIC detectors The DAMIC CCDs feature a three-phase polysilicon gate structure with a buried p-channel. The CCDs are typically 8 or 16 Mpixels, with pixel size of 15 µm × 15 µm, with a total surface area of tens of cm2 . The CCDs are 675 µm thick, for a mass up to 5.2 g. A high-resistivity (10-20 kΩ cm) n-type silicon allows for a low donor density in the substrate (∼ 1011 cm−3 ), which leads to fully depleted operation at low values of the applied bias voltage (∼40 V for a 675 µm-thick CCD). Fig. 1 shows a cross-sectional diagram of a CCD pixel, together with a sketch depicting the WIMP detection principle. The substrate voltage also controls the level of lateral diffusion of the charge carriers as they drift the thickness of the CCD. The lateral spread (width) of the charge recorded on the CCD x-y plane may be used to reconstruct the z-coordinate of a point-like interaction [10]. Long exposures are taken in DAMIC (∼8 hours) in order to minimize the number of readouts and consequently the number of pixels above a given threshold due to readout noise fluctuations. The CCD dark current due to thermal excitations (< 0.1 e− pix−1 day−1 at the operating temperature of ∼140 K) contributes negligibly to the noise. During readout, the charge held at the CCD gates is measured by shifting charge row-by-row and column-by-column via phased potential wells to a low capacitance output gate. The inefficiency of charge transfer from pixel to pixel is as low as 10−6 . The readout noise for the charge collected in a pixel is ∼2 e− which corresponds to an uncertainty of ∼7 eV of ionizing energy in Silicon. Calibrations with a 55 Fe source, with fluorescence X-rays from a Kapton target exposed to the 55 Fe source and with αs from 241 Am were performed. As shown in fig. 2, the detectors present an excellent linearity and energy resolution (55 eV RMS at 5.9 KeV) for electron-induced ioniza2

PoS(ICRC2015)1221

A well established body of evidence from astrophysics and cosmology supports the existence of cold dark matter as the major component of the material content of the universe. The leading candidate for this dark matter is a hypothetical weakly interacting massive particle (WIMP) [1, 2]. WIMPs could produce keV-energy nuclear recoils when scattering elastically off target nuclei in the detector. Minimal supersymmetric extensions to the standard model favor particles above 50 GeV/c2 , while other models relating dark matter with the baryon asymmetry prefer masses around 5 GeV/c2 [3, 4, 5]. Several experiments have reported statistically significant evidence of WIMPs scattering on light nuclear targets [7, 6]. The DAMIC (Dark Matter in CCDs) experiment uses the bulk silicon of scientific-grade charge-coupled devices (CCDs) as the target for coherent WIMP-nucleus elastic scattering. Due to the low readout noise of the CCDs and the relatively low mass of the silicon nucleus, CCDs are ideal instruments for the identification of the nuclear recoils with keV-scale energies and lower from WIMPs with masses < 10 GeV/c2 . The first DAMIC measurements were performed in a shallow underground site at Fermilab using several 1-gram CCD detectors developed for the Dark Energy Survey (DES) camera (DECam) [8]. With 21g-days DAMIC produced the best upper limits on the cross-section for WIMPs below 4 GeV/c2 [9]. DAMIC is now located in SNOLAB laboratory 2 km below the surface in the Vale Creighton Mine near Sudbury, Ontario, Canada.



y

3-phase CCD structure Poly gate electrodes n— — (10 kΩ-cm)

Buried p channel

x

WIMP

Energy calibration using X-rays Photosensitive volume _300µm) (200 (500 μm)

zx

y

z 55Fe

(a) A CCD pixel 4.2this keVwork. 103 Figure 1. Cross-sectional diagram of the CCD described in

2.

(b) WIMP detection in a CCD

K K⇥ Esc K CCD PHYSICSdiagram AND OPERATION Si 10 1:1. a) Figure a) Cross-sectional Cross-sectional µmµm ⇥ 15 in a fully back-illuminated CCD. Figure diagramofofa a1515 × µm 15 pixel µm pixel in a depleted, fully depleted, back-illuminated

2 FULLY-DEPLETED

1.7 keV ws a cross-sectional diagram the fully-depleted, back-illuminated CCD. A conventionally-processed, Esc K2 Theofthickness of the gate structure and the backside ohmic contact are µm. transparent rear window, ⇥ CCD. The thickness the gate structure and the backside ohmic contact are The ≤ 2µm. The transparent rear Al n-typeofsilicon CCD is fabricated on a high-resistivity, substrate. Webeen haveeliminated fabricated in CCD’s on bothCCDs. b) Dark matter detection in essential for astronomy applications, has the DAMIC 10 window has been eliminated in the DAMIC CCDs. b) Dark matter detection in a CCD. A WIMP scatters 150 mm diameter high-resistivity silicon substrates. The resistivity of 100 mm wafers is as high as awork CCD. A WIMP scatters with a silicon nucleus producing ionization in the CCD bulk. The charge carriers with a silicon in the region, producing 0 -cm, while the initial on 150 nucleus mm wafers has active been on 4,000–8,000 -cmionization silicon. from the nuclear recoil which drifts along

are then drifted along the z-direction and collected at the CCD gates.

image

4200

50 pixels

4190

Var(E) / keV2

4180

Reconstructed energy / keV

1330

1320

1310

1300

1290

1280

1 z-direction and is collected at thewhen CCD gates. to conventional thinned ness of the CCD resultsthe in improved near-infrared sensitivity compared 1 of absorption 2 4 5 6 E [keV] image3 at photon energies is is due to the strong dependence length on wavelength approaching ndgap.4 Figure 2 shows measured 1330 quantum efficiency (QE) versus wavelength for30 a fully-depleted, Energy Calibration data to X-ray isX-ray? especially high at lines near-infrared wavelengths. The CCD resolution from X-ray lines ated CCD operated at 130 C. The QE ure 2 has a two-layer anti-reflection (AR)n,coating tuned for good red response. It consists of 60 nm 241 25 WIMP? pe from Si Am oxide (ITO) and 100 nm of silicon1320 dioxide (SiO2 ). fluorescence XDiffusion 16 10-2 lly-depleted CCD’s also greatly reduce the problem at near-infrared wavelengths.5 ray absorption 20 e of “fringing” 0. limited 4.2 keV = urs when the absorption depth incident light exceeds the CCD thickness. Multiple reflections 10 of the 55 1310 o Fe ging patterns that are especially a problem in 10–20 µm thick CCD’s used in spectrographs. an 15

4210

feature of the CCD shown in Figure 1 is the use of aCa-K substrate bias to Back fully deplete the substrate. α 1300 CCD fabricated on high-resistivity silicon the channel potential is to first order independent of the 10-3 Al-K Si-K 10 μ 1 This is because for typical substrate thicknesses and doping densities considered here only s. 1 from Mn Kα ion of the electric field lines from the O depleted channel terminate in the fully-depletedpesubstrate. 1290 rtical clock levels can be set to optimize operating features such as well capacity5 andX-ray CTEabsorption while C bias is used to deplete the substrate. 4220

1

Front

10

1

F

30eV noise 1.7 keV 10

30

25

20

15

10

5

trate bias also plays a role in the point-spread function of the CCD./ keV For light absorbed near the 1280 Energy / keV 4180 4190 4200Energy 4210 4220 0.3 0.6the0.9 1.2 1.5 1.8 2.1 of the CCD the lateral charge spreading during transit of 15 the photogenerated charges0 through 5 10 20 25 30 6 measured by pixel / keV d substrate to the CCD collection wells is described bymeasured an rmsbystandard Energy pixel / keV deviation given by1,Energy

Figure 2: a) Reconstructed energy of an X-ray line compared to is true energy. The labeled Kα markers are (a) Portion of a DAMIC image (b) Emission of Si fluorescence X-ray 2 fluorescence lineskTfrom yelements in the Kapton target and other materials in the CCD setup. The 55 Fe and D 15 Semin´ ario CBPF April 15, 2015 (1) 2 od 241 Am markers areq X-rays by the radioactive sources. Linearity in the measurement of ionization (Vsub emitted VJ ) energy is 2. demonstrated fromportion 0.3 kev 60 keV. b) Variance of the lines asata ground functionlevel. of energy. The Figure a) 50⇥50 pixel of to a CCD image, taken when theX-ray detector was Different oltzmann’s constant, T is absolute temperature, q is the electron charge, yD is the thickness of effective factor is 0.16, typical(see for aisCCD [11]. kinds ofFano particles are better potential contrast, only substrate, Vsub is the applied substrate biasrecognizable voltage, and VJtext). anFor average near pixels the with deposited energy >0.1 keV are represented b) Event with two nearby clusters detected after illuminating the CCD with a 55 Fe V is the voltage drop across the region where ial wells due to the channel potentials.inVcolor. sub J nerated holes are drifted source. by the The electric This result is a simplified asymptotic that isof a Si fluorescence X-ray, emitted 1.7 field. keV cluster is a photoelectron (pe) from the form absorption of the substrate doping and is valid for highwith electric fields inofthe Therefore casein a nearby tion, as measured X-ray sources [10]. The ionization efficiency of nuclear following photoelectric absorption the substrate. incident 5.9 keV Mnin Kathis X-ray site. recoils is signifi-

directly proportional to yD , T , and 1/ (Vsub VJ ). The PSF for a CCD of this type can be different than that of electrons. Previous measurements have been done down to energies of reducing the substratecantly thickness and operating the CCD at high substrate bias. PSF measurements in more detail in Section 5. 3-4 keVr [12, 13] in agreement with Lindhard theory [14]. From this, DAMIC’s nominal 50 eVee

ing of the 3-phase gates (“parallel clocks”), while higher frequency clocks (“serial clocks”) move threshold corresponds to ∼0.5 keVr . the charge of the last row horizontally to a charge-to-voltage amplifier (“output node”). The inThe total charge and shape of each hit is extracted using dedicated image analysis tools. In efficiency of charge transfer from pixel to pixel is as low as 10 6 and the252 readout noise for the fig. 3 a sample of tracks recorded during a short exposure at sea level to a Cf source is shown. charge collected in a pixel is ⇠2 e [2]. Since on average 3.6 eV is required to ionize an electron Clusters from typescorresponds of particlestomay be observed. Low electrons andThe nuclear in silicon, thedifferent readout noise an uncertainty of ⇠7 eV energy in deposited energy. imrecoils, whose physical track length is 0.1keV are colored. b) Event with two nearby clusters ee kinds of particles are recognizable (see text). For better potential contrast, only near pixels the with deposited energy >0.1 keV sub is the applied substrate bias voltage, and VJ is an average 55 detected after illuminating thewith CCDtwo with a Feclusters source. The 1.7 keV cluster is a photoelectron from are represented inVcolor. nearby after illuminating the CCD(pe) with a 55 Fe Vb)J Event is the voltage drop across the detected region where to the channel potentials. sub absorption a Si fluorescence emitted following form photoelectric of the incident 5.9 keV are drifted source. by the the electric This result is a X-ray, simplified asymptotic that isofabsorption The 1.7 field. keV of cluster is a photoelectron (pe) from the absorption a Si fluorescence X-ray, emitted Mn K X-ray in a nearby site. α high electric fields in the substrate. Therefore in this case ate doping and is valid for following photoelectric absorption of the incident 5.9 keV Mn Ka X-ray in a nearby site. ortional to yD , T , and 1/ (Vsub VJ ). The PSF for a CCD of this type can be substrate thickness and operating the CCD at high substrate bias. PSF measurements spatial extent of the cluster is dominated by charge diffusion. Higher energy electrons (e), from ail in Section 5.

either Compton scattering or βclocks”), decay, lead to extended tracks. α particles in the bulk or from ing of the 3-phase gates (“parallel while higher frequency clocks (“serial clocks”) move the back of the CCD produce large round structures due to the plasma effect [16]. Cosmic muons the charge of the last row horizontally to a charge-to-voltage amplifier (“output node”). The in(µ) pierce through the CCD, leaving a straight track. The orientation of the track is immediately efficiency of charge transfer from pixel to pixel is as low as 10 6 and the readout noise for the evident from its width, the end-point of the track that is on the back of the CCD is much wider than charge collected in a pixel is ⇠2 e [2]. Since on average 3.6 eV is required to ionize an electron the end-point at the front due to charge diffusion. in silicon, the readout noise corresponds to an uncertainty of ⇠7 eV in deposited energy. The im-

3. The DAMIC experiment at SNOLAB

Fig. 4 shows the infrastructure already installed in SNOLAB. A packaged CCD (2k×4k, 8 Mpixel, to a high-purity silicon support piece. The 500 µm-thick) is shown in fig 4a. The device–is3epoxied – Kapton signal flex cable bring the signals from the CCDs up to the vacuum interface board (VIB). The cable is also glued to the silicon support. A copper bar facilitates the handling of the packaged CCD and its insertion into a slot of an electropolished copper box (fig 4b). The box is cooled to ∼140 K inside a copper vacuum vessel (∼ 10−6 mbar). An 18 cm-thick lead block hanging from the vessel-flange shields the CCDs from radiation produced by the VIB, also located inside the vessel (fig 4c). The CCDs are connected to the VIB through Kapton flex cables, which run along the side of the lead block. The processed signals then proceed to the data acquisition electronic boards. The vacuum vessel is inserted in a lead castle (fig 4b) with 21 cm thickness to shield the CCDs from ambient γ-rays. The innermost inch of lead comes from an ancient Spanish galleon 4

PoS(ICRC2015)1221

4210

e CCD shown in Figure 1 is the use of a substrate bias to Back deplete the substrate. αchannel potential is to firstfully 1300 the ed on high-resistivity silicon order independent of the 10 μ densities considered because for typical substrate thicknesses and doping here only from Mn Kα ectric field lines from the depleted channel terminate in the fully-depletedpesubstrate. 1290 evels can be set to optimize operating features such as well capacity5 andX-ray CTEabsorption while to deplete the substrate.



a)

c)

2 km of rock

d)

CCD Si support

VIB

Polyethylene

Kapton signal cable

b)

Lead block

Lead

Kapton signal cable

Cu box with CCDs

42 cm

Cu vacuum vessel

21 cm

V e s s e l

Polyethylene

Figure 3. a) A packaged DAMIC CCD. b) The copper box housing the CCDs. c) Components of the DAMIC Figure 4: a) Atopackaged DAMIC CCD. b)vessel. The copper box housing thethe CCDs. c) Components of the DAMIC setup, ready be inserted in the vacuum d) The vessel inside lead castle, during installation of setup, ready to be inserted in the vacuum vessel. d) The vessel inside the lead castle, during installation of the polyethylene shield. the polyethylene shield. 210 Pb content, strongly suppressing the background from bremsstrahlung γs and hascoating negligible (ITO) deposited on the backside after thinning the CCD to 500 µm . The third CCD was 210 Bi decays in the outer lead shield. A 42 cm-thick polyethylene shielding is used to produced optimizedbyfor DAMIC by maximizing its mass (the CCD is un-thinned, 675 µm-thick) and minimoderate and absorb environmental mizing radioactive contamination (theneutrons. ITO layer containing b -radioactive 115 In is eliminated). The 500-µm CCDs are inserted in adjacent slots of the copper box, with copper plates above and below. 675-µm CCD is of in aradioactive lower slot of the box, separated from the other CCDs by ⇠1 cm of copper. 4.TheMeasurements contamination The dark matter search will be performed with DAMIC100, a detector with 100 g of sensitive mass, consisting of 18 CCDs, each of pixels andis675 µm thickness. The ultimate sensitivity of 4k⇥4k the experiment determined by the rate of the radioactive background that mimics the nuclear recoil signal from the WIMPS. The SNOLAB underground labo2.3 CCD image reduction and datadue samples ratory has low intrinsic background to its 6000 m.w.e. overburden. Dedicated screening and selection of detector shielding materials, as well as radon-suppression methods, are extensively Clusters of energy deposits are found in the acquired images with the following procedure. First, employed to decrease the background from radioactive decays in the surrounding environment. the pedestal of each pixel is calculated as its median value over the set of images. The pedestals are The measurement of the intrinsic contamination of the detector is fundamental. For silicon-based then subtracted from every pixel value in all images. Hot pixels or defects are identified as recurrent experiments the cosmogenic isotope 32 Si, which could be present in the active target, is particularly patterns over many images, and eliminated (“masked”) from the analysis (>95% of the pixels were relevant β decay spectrum extendsas to any the lowest energies andpixels may with become an irreducible deemed since good).its Pixel clusters are selected group of adjacent signals greater background. Thetheanalysis used to establish the contamination levels exploit the unique than four times RMS ofmethods the white noise in the image. The resulting clusters are considered spatial resolution of theinteractions. CCDs. candidates for particle Relevant variables (e.g. the total energy by summing over all The identification of α-induced clustersFor is the the studies first step in establishing limitsweonrequired uranium pixel signals) are calculated for each cluster. presented in this paper, theand thorium contamination [15]. Radiogenic αs lose most of their energy by ionization, creating cluster energy to be >1 keV, which guarantees a negligible probability of accidental clusters from a dense column electron-hole that tosatisfy the plasma condition For interactions readout noise. ofSelection criteriapairs specific the different analyses will be[16]. described in Sections deep 3 in the substrate, the charge carriers diffuse laterally and lead to round clusters of hundreds of

5 –5–

PoS(ICRC2015)1221

Copper bar

Pb

4876 4878 4880 4882 4884 4886 4888 4890

0

5000

15000

20000

25000

30000

Cluster #79

0

2000

4000

Δt = 35 days

1330

1335

1340

1345

1350

1355

4870

4872

Cluster #51

4874

1360

4876

6000

10000

8000

12000

14000

16000

18000

(xo, yo) 20000

22000

E1 = 114.5 keV

10000



4868 4866 1355

1356

1357

1358

1359

1360

1361

1362

1364

1363

1365

PoS(ICRC2015)1221

E2 = 328.0 keV

Figure 8. Candidate b decay sequence found in data. The first cluster was detected in an image taken on Figure 5: Candidate β decay sequence found in data. The first cluster had 114.5 keV of energy. A second 2014/08/05 andwith deposited 114.5 energy. second withBoth energy was observed in cluster, energy 328.0 keV,keV was of observed in anAimage takencluster, 35 days later. tracks328.0 appearkeV, to originate an imagefrom taken days Both appear to originate from the same point (yellow star) in the CCD the35 same pointlater. (yellow star)tracks in the CCD x-y plane. x-y plane.

micrometers in diameter, whereas α particles that strike the front of the CCD lead to mostly vertical clusters according to a phenomenon known as “blooming” [11]. Simple criteria are sufficient 210 Po to by efficiently selectofand classifyinαs. of plasma αsiscan be used to establishand limits decays or emission a g-ray 4%Spectroscopy of the decays. itself radioactive decays by a 210 238 232 on Pb, U and Th contamination in210 the bulk of the CCD. In special DAMIC runs, with emission. The possible contamination from Po in the CCD has been discussed in Section 3.2. a dynamic range optimized for α energies, four plasma αs whose energies are consistent with 32 P and 210 Bi, are expected to remain in the same lattice site as their The210 intermediate nuclei, Po were observed. One of them cannot be 210 Po, as it coincides spatially with two higher parent nuclei throughout their lifetimes. Therefore, the b slikely produced eachsequence. decay pair should energyand αs recorded in different CCD exposures, and is therefore part ofby a decay 210 210 6 ) on the x-y other(out threeof as 8⇥10 bulk contamination of plane Po (or Pb),CCD. an upper limit of a< search for originateWhen frominterpreting the samethe pixel of the Through −1 −1 238 234 230 226 37 kg d (95% CL) is derived. In the U chain, the isotopes U, Th 32 Ra decay by electron-like tracks starting from the same spatial position, individual andSi –32 P and 210 Pb –210 Bi emission of αs with energies 4.7-4.8 MeV. Since the isotopes’ lifetimes are much longer than the decay sequences can be selected with high efficiency. We performed this search with the lowest CCD exposure time, their decays are expected to be uncorrelated. No plasma αs were observed in 238 U the background data set in the µmlimit CCD. Given background electrons per the 4.5-5.0 MeV(Table energy1) range, and 675 an upper on the contamination of < level 5 kg−1(⇠10 d−1 (95% 238 is correspondingly (secular equilibrium of the among isotopes with U was assumed). A small for day in aCL) CCD), the numberderived of accidental coincidences uncorrelated tracks are −1 −1 232 210 Bi. similar results intoa upper limit of < 15 CL) A on candidate Th contamination in the periods of timeanalysis comparable the half-lives ofkg32 Pdand(95% decay sequence found CCD bulk [15].

in the data is shown in Figure 8 to illustrate the search strategy.

A search for decay sequences of two β tracks was performed to identify radioactive contamination from 32 Si and 210 Pb and their daughters. 32 Si leads to the following decay sequence:

4.1 Search procedures for spatially correlated b decay sequences 32

32

Si −→ P + β with τ1/2 = 150 y, Q − value = 227 keV The first step in the search for decay sequences is to find the end-points of the b tracks. The 32 P −→ 32 S + β − with τ = 14 d, Q − value = 1.71 MeV procedure is illustrated in Figure 9. First, we1/2 find the pixel with the maximum signal in the cluster, and we use it as a seed point. Then, for every pixel of the cluster we compute the length of the A total of 13 candidate pairs were observed in the data. With detailed Monte Carlo simulations shortest the path to the seed point, whereofthe taken only along pixels are included in the 32 Sipath overall efficiency for detection –32 P is decay sequences in the data set that was determined cluster. We this asThe thenumber “distance” from pairs the seed point. The pixel the greatest to berefer εSi =to 49.2%. of accidental was also determined withwith simulations. The distance +110 −1 −1 32 wasend-point estimated to 80−65 kg Finally, d (95% CI)recompute for Si in the bulk [15]. With apixel from is taken decay as therate first ofbethe track. we theCCD distance of every 210 similar procedure the upper limit on the Pb decay rate in the CCD bulk has been deduced as the first end-point, and take the pixel with the largest distance as the second end-point of the cluster.