First detection of tracks of radon progeny recoils by MIMAC
Q.Riffard, D. Santos, G. Bosson, O. Bourrion, T. Descombes, C. Fourel, O. Guillaudin, J.-F. Muraz, P. Colas, E. Ferrer-Ribas, I. Giomataris, J. Busto, D. Fouchez, C. Tao, L. Lebreton, D. Maire
PPrepared for submission to JINST
First detection of radon progeny recoil tracks byMIMAC
Q. Riffard a D. Santos a O. Guillaudin a G. Bosson a O. Bourrion a J. Bouvier a T. Descombes a C. Fourel a J.-F. Muraz a L. Lebreton b D. Maire b P. Colas b E. Ferrer-Ribas c I. Giomataris c J. Busto d D. Fouchez d J. Brunner d C. Tao d,e a LPSC, Université Grenoble-Alpes, CNRS/IN2P3, Grenoble, France b LMDN, IRSN Cadarache, 13115 Saint-Paul Lez-Durance, France c IRFU, CEA Saclay, 91191 Gif-sur-Yvette, France d Aix Marseille Université, CNRS/IN2P3, CPPM UMR 7346, 13288, Marseille, France e Tsinghua Center for Astrophysics, Tsinghua University, Beijing 100084, China
E-mail: [email protected] , [email protected] Abstract : The MIMAC experiment is a µ-TPC project for directional dark matter search.Directional detection strategy is based on the measurement of the WIMP flux anisotropydue to the solar system motion with respect to the dark matter halo. The main purposeof MIMAC project is the measurement of nuclear recoil energy and 3D direction fromthe WIMP elastic scattering on target nuclei. Since June 2012 a bi-chamber prototypeis operating at the Modane underground laboratory. In this paper, we report the firstionization energy and 3D track observations of NRs produced by the radon progeny. Thismeasurement shows the capability of the MIMAC detector and opens the possibility toexplore the low energy recoil directionality signature.
Keywords:
Dark Matter directional detection; MIMAC; Radon progeny;
ArXiv ePrint : 1504.05865 a r X i v : . [ a s t r o - ph . I M ] J un ontents Rn pollution evidence 93.2 In coincidence events 10 Introduction
A large number of astrophysical and cosmological observations at different scales supportthe existence of a cold dark matter component in the Universe. At the Universe scale, thiscomponent represents roughly 26% of the total mass-energy density of the Universe [1]. TheWeakly Interacting Massive Particle (WIMP), a generic particle, is one of the leading darkmatter particle candidates: a massive particle interacting only through weak and gravita-tional interactions. This candidate is supported by the so-called "WIMP miracle" - the factthat such a particle, produced in the early Universe, would give the correct dark matterabundance - and naturally occurs in R-parity conserving supersymmetric models. At theMilky Way scale, dark matter forms a static halo surrounding our galaxy. The relative mo-tion of the solar system through the dark halo produces a flux of WIMP on Earth. Throughthe weak interaction, WIMP could interact with ordinary matter producing nuclear recoils(NR) by elastic scattering [2]. In the last two decades, a large experimental effort hasbeen deployed by international collaborations in order to probe a direct detection of NRfrom WIMP-nucleus interactions without success [3–5]. The main goal of these experimentsis to improve their sensitivity by increasing the exposure and/or by reducing the energythreshold. The major limitation of the direct search strategy arises from an irreducible– 1 –ackground: the neutrino-induced NR. Indeed, neutrinos can produce NRs with a kineticenergy at the 10 keV scale by coherent scattering limiting the detector sensitivity [6].The directional detection strategy, first proposed in 1988 [7] is based on the fact thatthe WIMP event distribution is expected to have an excess in the direction of the Solarsystem motion with respect to the galactic center. It induces a dipole feature in the WIMP-produced NR distribution [8], whereas the cosmic-ray and ( α, n ) induced background dis-tributions [9] are expected to be isotropic in the galactic rest frame or completely uncor-related with respect to the motion around the galactic center. Several directional featuresfrom different backgrounds would display unambiguous differences with the WIMP signal:dipole [8], ringlike [10] and aberrations [11]. They may be used to either exclude dark mat-ter [12, 13], discover galactic dark matter with a high significance [8, 14, 15] or constrainWIMP and halo properties [16–19], depending on the WIMP-nucleon cross section. It hasalso been recognized as the ultimate detection strategy to look for dark matter beyond theneutrino floor [6, 20, 21].The MIMAC experiment is a directional detection project initiated in 2007 at theLaboratoire de Physique Subatomique et de Cosmologie (LPSC Grenoble-France) [22] incollaboration with IRFU-Saclay, CPPM- Marseille, IRSN and recently the Tsinghua Univer-sity. As the other directional detection experiments [23, 24], the aim of the MIMAC projectis the measurement of the WIMP-induced NR energy and angular distributions in order toconstrain the WIMP properties. In this context, the MIMAC collaboration developed anoriginal detector and readout strategy allowing the measurement of NR tracks in 3D. SinceJune 2012, a bi-chamber prototype was installed at the Modane underground laboratory(Laboratoire Souterrain de Modane ) for preliminary tests. This first data taking pointsthe radon progeny as the current most important source of background for the MIMACexperiment. In this paper the first 3D tracks measurement of the daughter NRs from radonprogeny α -decay is presented. The main goal of the MIMAC experiment is to define the large gaseous TPC for directionaldetection of WIMP [22] by using a CF -based low pressure (at
50 mbar ) gas mixture. Theconcept of this experiment is a replication of active cells to reach a final active volume of m . Each cell consists of a gaseous TPC read by a pixelated Micromegas coupled witha self-triggered fast electronic system allowing the measurement of the NR energy and 3Dtracks. To demonstrate the feasibility of the chosen technology, a bi-chamber prototype,containing two active cells, has been deployed at the LSM since June 2012 [25]. Top panel of figure 1 represents a scheme of the bi-chamber prototype. The detector isfilled with a CF + 28%CHF + 2%C H gas mixture at 50 mbar and is composed of twomirroring symmetric cells sharing a fine aluminized mylar cathode. Each cell is equipped LSM – 2 – dCuFe e
25 cm µ m X , Y strips
Pressure regulator O , H O filter
Flow controller
Pump
Bu↵er volume (1 bar)
Circulation µm cathode . c m
70% CF + 28% CHF + 2% C H @ 50 mbar e e ! E gain X ray generator ! E drift Electronic board
Mesh x x y y
200 µm x y Figure 1 : Top panel: Illustration of the bi-chamber prototype configuration and the ioniza-tion electron collection from a NR produced by the WIMP elastic scattering (not in scale).Bottom panel: Scheme of the X and Y strips arrangement of the Micromegas. Each pixelis 200µm wide with a 100µm gap corresponding to a 424µm pitch.– 3 –ith a of 10.8 cm wide pixelated Micromegas [26, 27] with × strips on X and Yaxis as represented in figure 1 bottom panel. When a WIMP (represented by a dashed line)interacts with a gas nucleus by elastic scattering, it transfers a part of its kinetic energyto the nucleus producing a NR (represented by a red dashed arrow). The NR releasesonly a part of its energy by ionization, creating electron-ion pairs. Ionization electrons arecollected to a mesh Micromegas with a . µm/ns drift velocity (measured by using themethod described in [28]) by means of an electric field E drift ∼
180 V . cm − . Passing throughthe mesh these electrons are amplified by avalanche thanks to a much higher electric field E gain ∼
18 kV . cm − (corresponding to an ∼
470 V amplification voltage). On the one hand,the deposited ionization energy is read by a charge preamplifier connected to the mesh.Due to the saturation of the ADC our energy dynamic range is ranging from 0 to 62 keVee.In addition, by coupling the pixelated Micromegas with a fast self-triggering electronicsystem [29, 30], the X and Y strips are read at 50 MHz allowing the measurement of the(X,Y) projection of the track as a function of time. Knowing the electron drift velocity, itis then possible to reconstruct the relative Z coordinate. The determination of the absoluteZ coordinate of the track is not directly possible with this readout. This aspect is discussedin section 5.1.This paper is focused on the analysis of the first two MIMAC data taking run in darkmatter search mode. The properties of these runs are summarized on the table 1. Thesame electric fields have been used all along the data taking. We can notice a differenceof total events rate between the two runs. This difference is due to a reduction of the gascontamination by radon isotopes as discussed in section 3.1.Run label Drift voltage Gain voltage Lived-time Total event rateRun 2012 4520 V 470 V 77 days . ± . − Run 2013 4520 V 470 V 103 days . ± . − Table 1 : Table summarizing the 2012 and 2013 data taking properties.
As shown in figure 1, an X-ray generator is permanently mounted in front of metal foilscoupled to the main vessel. During irradiation, metal foils and stainless steel vessel producefluorescence photons that can be used as calibration sources (photons from Cd, Fe, Cu, Crand Ni). The left panel of figure 2 shows a typical calibration spectrum from the chamber 1and the best-fit (red line). We obtained an excellent agreement between model and datawith a reduced χ of 1.09 and an associated p-value of 0.464. In addition, no structures areobservable on the fit residual (bottom panel), showing that the model is fully adequate todescribe the calibration spectrum. Each contribution from the metal foils is represented bydashed lines such as the 3.19 keV from Cd, the 5.4 keV from Cr, the 6.4 keV from Fe, the7.5 keV from the Ni, the 8.04 keV from Cu fluorescence photons.– 4 – [ADC] ioni E C oun t
10 Cd Cr Fe Ni Cu
Ionization energy [keV]
Ionization energy [ADC] C oun t - M od e l - - - Date [day]02/07/13 31/08/13 31/10/13 31/12/13 02/03/14 02/05/14 a [ k e V / A DC ] Chamber 1Chamber 2
Figure 2 : Left panel: X-ray generator calibration spectrum measurement. Energy peaks(dashed lines) produced using fluorescence photons from cadmium, iron and copper metalfoils. The red line represents the total fit of the spectrum. The bottom panel representsthe residual of the fit. Right panel: Slope of the linear calibration a as a function of timefor chamber 1 (black dots) and chamber 2 (red dots). The gas circulation system developed for this experiment is an important component of thedetector, ensuring the gas quality stability in a closed circuit. As schematically shown infigure 1, it includes a buffer volume, an oxygen filter, a dry and very low leak pump anda pressure regulator. This system has been designed to prevent the presence of impuritiesand O into the gas. These contaminants affect the gain of the Micromegas and then theenergy resolution. The detector is weekly calibrated in order to monitor the gain variationsthrough the slope of the linear calibration a . The right panel of figure 2 shows the variationof a as a function of time for both chambers (chamber 1 in black and chamber 2 in red).We can observe that the gains in both chambers are roughly constant with variations lowerthan 1% over several months, demonstrating the gain stability during the data taking. A background for dark matter detection arises from an intrinsic pollution of detector mate-rial by radioactive nuclei, such as U and Th . The decay chains of such nuclei produceelectron recoil (ER) background from β decays and γ de-excitation. In addition, we haveradon emanations from the surface of the materials, releasing Rn and Rn inside thegas of the detector. Radon Progeny Recoils denote NRs produced by α -decays from Rn and Rn decay chains. It includes α -particles and daughter nucleus recoils. These eventshave been extensively observed in dark matter detectors, see for example: [31–34]. The firstreport of RPR events in a dark matter directional detector was published by the DRIFTcollaboration in 2008 [31] [32, 33]. They measured the track length of α -particles from the RPR – 5 –arent T / Mode E kinα/β max Daughter E kinrecoil E ionirecoil [MeV] [keV] [keVee]From Rn Rn α . Po 100 . . Po α . Pb 112 . . Pb
27 min β − Bi - - Bi
20 min β − Po - - Po
164 µs α . Pb 146 . . Pb
22 years β − Bi - - Bi β − Po - - Po
138 days α Pb (stable) 103.7 40.28From Rn Rn
55 s α . Po 116 . . Po α . Pb 128 . . Pb β − Bi - - Bi (64%) 61 min β − Po - - Bi (36%) 61 min α . Tl 117 . . Po α . Pb 169 . . Tl β − Pb (stable) - - Table 2 : Details of the α and β decays from Rn and Rn decay chains. This tablecontains the half-life T / of each element and the decay mode. For α -decays, it summarizesenergies E kinα of the emitted α -particles and the kinetic E kinrecoil and ionization E ionirecoil energiesof daughter nuclei. Ionization energies of daughter nuclei were estimated with SRIM [35].For β -decays, it summarizes only β maximal energies E kinβ max .radon progeny and correlated the length of the α -particle track with its kinetic energy inorder to identify each nucleus decay.When the MIMAC detector is in dark matter search mode, it is not possible to measure α -particle energies or their track lengths. Indeed, at 50 mbar their tracks are not fullycontained in the active volume of one chamber. Instead, we can observe the daughter NRsproduced by the α -particle emission measuring their 3D tracks with their total ionizationenergy showing the ability of the detector to get a clear signature of low energy NR tracks. Table 2 presents the considered radio nuclei from Rn and Rn decay chains. It showsthe half-life of each element and the emitted particle with their maximum kinetic energy.From these radio-nucleus decays we can expect two different kinds of events: electrons from β and γ emissions, and RPR events from α -particles and daughter nucleus emissions.From radon chains we can distinguish two contributions to the ER background: β decays and electrons produced by Compton scattering of γ -rays inside the detector. For dark– 6 –atter searches, this background can be removed using a dedicated ER/NR discriminationmethod based on boosted decision trees (as discussed in section 4).In addition, RPR events occur at different positions inside the detector, which impactstheir tracks and energies and, in consequence, their discrimination. The different positionsare schematically described in figure 3. Red polygons represent α -decays, blue dots representthe daughter nuclei, plain arrows represent α -particles or daughter nucleus motion directionsand the dashed arrows represent the daughter nucleus migration due to the drift electricfield. We can distinguish five types of events as a function of their positions:
1) Volume events Rn and Rn can be present everywhere in the active volume.While an α -decay occurs in the gas volume, the energy deposited by the emitted α -particle with a 5.5 or 6.8 MeV kinetic energy (see table 2) saturates the preamplifier.These events can be easily discriminated using a cut on the saturation energy. Resultantdaughters (blue dots) are in general produced with a positive electric charged. It impliesthat they are collected on the cathode due to the drift electric field. Their drift velocitycan be estimated as three orders of magnitude lower than the electron drift velocity( v e − drift = 21 . µm / ns in MIMAC gas mixture [28]) allowing daughter nuclei to reach thecathode before the next α -decay of the chain. At the cathode, there is an accumulationof radon progeny elements.
2) Through cathode events
A SRIM [35] simulation shows that α -particles with ki-netic energies ranging from 5.5 to 8.8 MeV (see table 2) can pass through a 12 µmthickness mylar cathode and reach the other chamber with kinetic energies going upto hundreds of keV. For such events, we observe coincidences of signals from the twochambers.
3) Cathode events
As mentioned before in the case of RPR volume events, each daughternucleus produced in the drift electric field is collected on the cathode surface. Moreover,radon isotopes from the detector gas can also be fixed on the cathode surface by adsorp-tion. At the cathode surface there is an accumulation of radon progeny events. While an α -decay occurs at the cathode surface, there are two cases to consider: i) if the α -particleis emitted in the direction of the gas volume, as mentioned before, the ionization energydeposition saturates the preamplifier, ii) if the α -particle is absorbed in the matter, onlythe recoil of the daughter nucleus is detected. The recoils of the daughter nuclei from Rn and Rn progeny have kinetic energies from 100 to 170 keV, as shown in thetable 2, with simulated track lengths ranging from 650 to 900 µm. In addition, there isan important difference between the measurable ionization energy E ionirecoil and the kineticenergy E kinrecoil for low energy NRs defined by the Ionization Quenching Factor ( IQF ).The
IQF of a NR is defined by the ratio of the measured ionization energy E ionirecoil and itskinetic energy E kinrecoil or the ionization energy released by an electron of the same kineticenergy. This factor decreases rapidly with decreasing kinetic energies of NRs. It dependson the mass of the NR and on the gas and pressure, as shown in [36]. In our case, aSRIM simulation gives an IQF of about 40% for a heavy nucleus such as Po at 100keV. Taking into account this correction from SRIM, the RPR events should release an– 7 – E (stable) (stable) (3) Figure 3 : This schematic diagram illustrates the daughter RPR event spatial distribu-tion. The α -decays are represented by the red polygons. Daughter nuclei and their pathsare respectively represented by blue dots and plain arrows. Dashed arrows represent thedaughter nucleus migration. See section 5 for more details.ionization energy from 38 to 58 keVee as shown in the table 2. In this case, the daughterNR is associated with an α -particle (plain arrow) passing through the thin mylar cathodeand reaching the other chamber. These events will be in coincidence for the data analysis.When a radon progeny nucleus recoils in the active volume, it is collected again at thecathode surface by the drift electric field. Then, at the cathode surface, all the Rn and Rn progeny elements are collected.
4) Anode events
The Micromegas PCB and strips contain the most important U and Th pollution. Consequently they are the major sources of radon internal emanations.In the 256 µm amplification space, the multiplication of ionization electrons depends ontheir positions and it affects the energy measurement of an event passing through theamplification space. The impact of the gain variation on the energy measurement is– 8 –iscussed in section 5.2. At the anode level we expect two types of α -particle and/orRPR events: bulk events from the decay of U and Th and surface events from theradon isotopes emanations. In the case of bulk events, the NR leaves the surface witha reduced energy and reaches the amplification space. In some cases, it is possible fora bulk event to pass through the mesh and reach the drift space. In the case of surfaceevents, the daughter NRs pass through the mesh reaching the drift space. In any ofthese cases, the ionization energy measurement misestimates the total ionization energydeposition due to the gain variation through the amplification space. The daughter NRfrom the α -decay is collected either at the mesh or at the cathode surface.
5) Mesh events
As mentioned before, radon isotopes can be fixed on the Micromegasmesh wire surfaces by adsorption. The mesh is a woven stainless steel thread of 18 µmdiameter wires. While α -decays occur at wire surfaces and daughter recoils enter on theactive volume, the associated α -particles have a probability of 10% to pass throughoutthe wire. The energy released by these α -particles in the amplification zone will be in thatcase added to daughter nuclei deposition. The daughter NR is collected at the cathodesurface.All RPR events in the detector contribute to the accumulation of RPR at the cath-ode surface. Cathode and passing through the cathode events can be identified using thecoincidence between both chambers. The main feature of the anode and mesh events isthe misestimation of the ionization energy either due to gain variation in the amplificationspace or due to the addition of an associated α -particle energy fraction. Rn pollution evidence The α -particle rate was monitored selecting the 3D tracks saturating the preamplifier inthe 2012 data set. This selection includes α -particles from the active volume and from theother chamber through the mylar cathode. Figure 4 shows the 2012 α -particles rate. FromJuly th to September th, the α -rate was rather constant at . ± .
11 min − .On October 3rd, the gas circulation was switched off (red dashed line). We observed anexponential reduction of the event rate. We model it by the sum of a constant distribution c and a decreasing exponential with a T / half-life: R ( t ) = R exp (cid:18) − tτ (cid:19) + c , where τ = T / / ln(2) (3.1)It indicates an external pollution of the gas mixture from the circulation gas system and anintrinsic pollution from the materials. By fitting the event rate (green line), we obtaineda T / = 3 . ± .
69 days
Half life which is compatible with the 3.8 days half-life of the Rn . The constant c = 1 . ± . − estimates the intrinsic saturation event ratefrom the materials pollution. We identified the external source of contamination in thegas circulation system as a small leak in the circulation pump. During this data taking,– 9 –
012 date ] - R [ m i n -1 ± R = 3.80 C i r c u l a t i on s t opp e d ± = 3.87 T Figure 4 : The saturation event rate covering the first data run and a zoom of the lasttwelve days. The circulation system was stopped on October 3 rd , 2012, as represented bythe orange dashed line. The green line represents the exponential fit of the event rate. Themeasured half-life is . ± .
69 days , it is compatible with the Rn half-life (3.8 days).we measured a . × − mbar . L / s global leak rate. This leak injected Rn and Rn from the air of the cavern into the circulation loop. According to calibration data, no gaindegradation was observed thanks to the presence of oxygen and humidity filters. AfterOctober 3 rd , the circulation pump leak was fixed and the global leak rate was reduced by afactor of 4.5. After this operation, we measured a saturation event rate of . ± . − which is compatible with the estimated intrinsic saturation event rate showing that ourbackground is now dominated by internal sources of contamination of the detector. Inaddition, the impact of this leakage is as well visible on the total event rate as shown bythe table1.The exponential reduction of the event rate was dominated by the contribution fromthe Rn half-life. From this measurement, we could not conclude about the contributionfrom the Rn half-life. In general, the Rn contribution to the RPR event progeny isneglected [37]. Due to its shorter period (55 s vs 3.8 days) the impact of emanating Rn from materials is much smaller than Rn contributions. This hypothesis was supportedby a dedicated one-month measurement performed with 700 mbar of pure CF . The α -particle spectroscopy showed no contribution coming from the . alpha particle fromthe Po decay in the α -particle spectrum of the intrinsic background. The chamber clocks were synchronized with a 40 ns precision, allowing chamber coincidencesearches. As explained in section 2, we are expecting two different types of events in– 10 – igure 5 : 2D distribution with contour levels of the energy seen in each chamber for in-coincidence events (bottom left panel) and 1D marginalization (top left - bottom right).The regions delimited by the blue dashed lines represent events associated with saturationin the other chamber. The black (red) spectrum represents the chamber 1 (2) measuredenergies and the filled area corresponds to events associated with saturation as mentionedin section 2.2.coincidence: cathode events and passing through cathode events. We consider two eventsin coincidence if there are separated by less than . µs. It corresponds to the time requiredto travel the distance between the cathode and the anode (25 cm) for primary ionizationelectrons. Indeed, in the case of a cathode event as described before (case 3), the α -particlecan pass through the gas volume releasing part of its energy close to the anode while thedaughter NR releases its energy close to the cathode. In that case, we measure a promptsignal from the α -particle and a delayed signal from the daughter nucleus with a . µsdelay.Considering that the Rn contribution is negligible with respect to the Rn contri-bution, we expect to find four peaks on the "in coincidence" event energy spectrum fromcathode events: Po , Pb , Pb and Pb NRs. In a first approximation, we can alsoneglect the Pb contribution due to the long half-life of Pb .At the cathode surface, the Po contribution comes from the attachment of Rn while the other contributions come from the collection of the α -decay daughter nuclei dueto the drift electric field. In this context, the Po population at the cathode surface comes– 11 – Po Pb Absorbed (lost) . . Pb Collection and -decays Po Absorbed (lost) . . Pb Pb ( N/ N/
2) ( N/
2) ( N/ N ) ( N ) ↵ -decay ↵ -decay (2 N ) (2 N ) ( N )( N ) ( N ) ( N/ Figure 6 : Illustration of the relative contribution of Pb and Pb and the evolution oftheir population at the cathode surface. This figure illustrates why the Pb contributionamplitude is 2 times smaller than the Pb one.from the collection of Po after α -decays. If we neglect the Rn attachment, two maincontributions remain: Pb and Pb .Figure 6 illustrates the relative contribution of Pb and Pb and the evolution oftheir population at the cathode surface. Considering N Po nuclei at the cathode surface.After Po α -decay, Pb daughter nucleus has a 0.5 probability to be absorbed by thesurface. The number of Pb emitted in the gas is N . Assuming a 100% efficiency for the Pb collection, the number of Po at the cathode surface is N after two β -decays. Aspreviously, after Po α -decay, Pb daughter nucleus has a 0.5 probability to be absorbedby the surface. Thus, the number of Pb emitted in the gas is N/ . In conclusion, the Pb contribution amplitude is 2 times smaller than the Pb one due to the populationreduction by the surface absorption.In conclusion, we can expect two well-defined peaks from RPR cathode events, one flatdistribution from the α -particles passing through the cathode and an important saturationfrom α -particles which release more than 62 keVee.The panels on the top left - bottom right diagonal of figure 5 present the "in coinci-dence" event energy spectra in both chambers measured in 2013. These spectra show twopeaks at roughly 33 and 45 keVee and saturation over 62 keVee.The left bottom panel presents the 2D distribution of the energy seen in each chamberfor "in coincidence" events. We can clearly identify four regions delimited by the bluedashed lines:• The "double saturation" region corresponds to events with E Ch . and E Ch . >
62 keVee .It represents of 44% of the total "in coincidence" event sample. The events belonging tothis region are α -particles passing through the 12 µm mylar cathode and saturating the– 12 – keVee] ioni E ] - . ke V ee - R [ d ay Chamber 1 [keVee] ioni E ] - . ke V ee - R [ d ay Chamber 2
Figure 7 : Zoom from 0 to 60 keVee of the "in coincidence" event energy spectra in thechamber 1 (top panel) and 2 (bottom panel). The plain green line corresponds to the fitresult by the sum of two gaussians and a constant (see equation 3.2). The dashed blue andMagenta lines represent the individual contributions.preamplifiers in both chambers.• There are two "one saturation" regions ( E Ch . or E Ch . >
62 keVee ). These events (52%of the total "in coincidence" event sample) correspond to RPR events associated withan α -particle detected in the other chamber or not fully detected α -particles in bothchambers (in/out-going α -particles). In these regions, we can clearly identify the contourlevels of the two peaks shown by the marginalized energy distributions at 33 and 45 keVee.• The non-saturation region corresponds to events with E Ch . and E Ch . <
62 keVee . Thisevent sample constitutes 4% of the total. These events correspond to RPR events witha non-saturating α -particle such as an in/out-going α -particles, or a not fully detectedRPR event.This figure shows that most of "in coincidence" events are associated with an energysaturation: i.e. with an α -particle. It is illustrated by filled areas which represent theevents associated with a saturated event. More than 95% of the "in coincidence" eventsare in the saturation regions.Figure 7 shows a zoom of these energy spectra from the threshold in ionization energyto 62 keVee. These energy spectra show only two main gaussian contributions as expectedbefore: Pb and Pb . This observation supports our hypothesis about the Rn ad-sorption contribution. – 13 –arameter Unit chamber 1 chamber 2 Cst [ day − . keVee − ] . ± .
01 0 . ± . A [ day − . keVee − ] . ± . . ± . µ [keVee] . ± . . ± . σ [keVee] . ± . . ± . A [ day − . keVee − ] 5.8 ± . . ± . µ [keVee] . ± . . ± . σ [keVee] . ± . . ± . Table 3 : Parameters of the "in coincidence" energy spectra fits in figure 7 by the equa-tion 3.2.The energy spectra were fitted using the sum of a constant and two gaussians from 15to 60 keVee: f ( E ioni ) = Cst + (cid:88) i =1 A i σ i √ π exp (cid:32) − ( E ioni − µ i ) σ i (cid:33) . (3.2)In figure 7, the green line represents the fit result, blue and magenta dashed lines the indi-vidual contributions. Table 3 presents the values of the fit parameters for both chambers.These two peaks, measured at . ± . and . ± . , taking into account themean values between the two chambers, correspond respectively to the ionization energyreleased by Pb and Pb NRs. The measured
IQF values are Q ( Pb) = 29 . ± . for Pb and Q ( Pb) = 31 . ± . for Pb showing respectively a and 22% SRIM IQF overestimation (see table 2 for simulated values). This is consistent with previous re-sults published by our team on
IQF measurements on several gas mixtures and for severalions [36].The ratios of the two peak amplitude are . ± . for the chamber 1 and . / − . for the chamber 2 supporting our statement described above about the Po and Pb contributions. In order to reject the ER background from dark matter searches data, we developed anoriginal ER/NR discrimination method described in [38]. We placed a MIMAC mono-chamber detector on a monochromatic neutron field allowing us to acquire two specific datasets: (ER and NR) and ER only. We applied a Boosted Decision Tree (BDT) algorithm onthese two data sets. It gives a ER rejection power. As discussed in [38], by using thismethod the detector efficiency is not directly accessible. Then we have developed a Monte-Carlo simulation of the MIMAC readout that is able to reproduce our observables. Theapplication of the BDT analysis on the Monte Carlo shows a . ± . % NR efficiencyconsidering the full energy range and . ± . % considering a 5 keV lower threshold withan ER rejection power. This method was applied on the 2013 data run (103 days) andfigure 8 presents the resulting NR energy spectra. After the application of this analysis, we– 14 – keVee] ioni E ] - . ke V ee - R [ d ay Ch. 1Ch. 2
Figure 8 : Energy spectra measured in 2013 by the chamber 1 (black line) and the chamber 2(red line). It should be pointed out that each event represented on these spectra has its own3D track associated. These distributions were obtained applying the low energy ER/NRdiscrimination.can consider the ER event contamination in our data as negligible. These energy spectrashow two peaks at 33 and 46 keVee, in both chambers (red line chamber 1 and blackline chamber 2) as already observed in the "in coincidence" energy spectra. However, theobserved shapes are different from the "in coincidence" spectrum shapes, especially below25 keVee. These differences are due to the fact that these spectra contain all contributionsfrom RPR events described in section 2.2. We measure a total RPR event rate of . ± . − in chamber 1 and . ± . − in chamber 2.The fact that some progeny events are passing our BDT cuts shows that their reductionis needed for dark matter searches. The reduction of this background can be done by:• A screening of the materials in order to reduce their radioactivity and radon emana-tions.• A rejection of the RPR by the coincidence between the different chambers. Thismethod is limited by the thickness of the cathode that can prevent the coincidence.• A fiducialisation on z axis of the detector. As these events are located at the cathodeand anode levels, a cut on the z-coordinate of the events must allow to suppress theircontribution. A new signal, from the cathode, produced by the primary electronsdrift will be added [39].In section 5.1, we present a new observable exploiting the electron diffusion to identify theRPR events position along the z-coordinate.– 15 – igure 9 : Projections of a 41.1 measured NR track in the (X,Z), (Y,Z) and (X,Y) planesand 3D reconstructions. The Z axis is in units of time slice (20 ns) and the X and Y axisin strip number (424µm pitch). The color scale corresponds to the number of strips firedon the time slice. Figure 10 : Projections of a 36.4 keVee measured NR track in the (X,Z), (Y,Z) and (X,Y)planes and 3D reconstructions. The Z axis is in units of time slice (20 ns) and the X andY axis in strip number (424µm pitch). The color scale corresponds to the number of stripsfired on the time slice. The horizontal arrows represent the width along the X/Y axis ofthe time slice i : ∆ X i / ∆ Y i . The vertical arrow represents the slot duration ∆ t slot .– 16 – Radon progeny recoil position identification
As seen in section 3, the RPR events occur at different positions in the detector. Theposition determination in the (
X, Y ) plane is made using the pixelated Micromegas readout,while the z coordinate identification needs to use the information given by the electrondiffusion. Indeed, the electron diffusion in the drift space is directly related to the z coordinate via the probability density function of charge distribution on the anode plan.The transverse/longitudinal standard deviation of the charge dispersion σ T/L follows asquare root dependency on the distance of the track to the anode ( z ): σ T/L = D T/L √ z ,where D T/L is the diffusion coefficient. In order to obtain the distance of the track to theanode, we take the distance with respect to the center of the track. The diffusion coefficients D T/L at such pressure and electric field have been calculated using Magboltz [40], givingthe following values: (cid:40) D T = 237 . µm / √ cm D L = 271 . µm / √ cm Figures 9 and 10 show respectively the NR tracks of a 41.1 keVee cathode event anda 36.4 keVee anode event. These figures illustrate the impact of the ionization electrondiffusion on the track topology. It affects the measurement of the track length and thetrack width ( ∆ X/ ∆ Y ) depending on the z track coordinate. As illustrated in figure 10,close to the anode, the electron diffusion is negligible. However, as illustrated by figure 9,close to the cathode the electron diffusion is maximal. In that case, the edges of the primaryelectron density from cathode events are not completely sampled by the pixelated anodedue to the strip thresholds. It implies that the measured track length after a diffusiondeconvolution is misestimated.The 2D projections on the (X,Z) and (Y,Z) planes of the bottom panel show thedefinition of ∆ X i / ∆ Y i , the width of the i st time slice along the x/y axis. Then, in order toestimate the transverse contribution of the diffusion, we defined an observable called MeanProjected Diffusion as: M P D = log (cid:0) ∆ X × ∆ Y (cid:1) , (5.1)where ∆ X and ∆ Y are respectively the mean value of ∆ X i and ∆ Y i ( i = 1 , N ) . As anexample, the events show by figures 9 and 10 have respectively a 4.03 and a 1.21 MPDvalues.Figure 11 presents a simulated distribution of 112.3 keV Pb events in the ( z, MPD) plane. This simulation is based on Pb tracks generated by SRIM with a kinetic energy of112.3 keV on MIMAC gas mixture at 50 mbar. It included the ionization electrons diffusionusing a diffusion coefficient estimation from Magboltz as mentioned before. We can see thatMPD increases as the square root of the track z coordinate. The diffusion variation alongthe track is negligible for small tracks ( ∼ . − mm). As track length and mean widthdepends on the NR energy, the MPD depends on energy too. In conclusion, using this MPD – 17 – igure 11 : Simulated MPD distribution as a function of the track z coordinate. Thissimulation shows 112.3 keV Pb tracks generated by SRIM [35] and coupled to the MIMACdetector simulation. The black plain and dashed lines represent respectively the mean andthe standard deviation of the MPD as a function of the ionization energy.observable, we can identify the track position of the events in the detector. This observableallows one to discriminate anode events from volume or cathode events. For Pb NRsof 112.3 keV, cathode and volume events have a MPD value higher than 2.3, while anodeevents have a value lower than 2.3 for the lengths of NRs being smaller than 1 mm, seeinsert in figure 11.
Figure 12 shows the measured MPD distribution as a function of the ionization energy E ioni in the chamber 2. In this figure, two regions can clearly be identified. Using thein-coincidence data, we performed a cut on MPD keeping 99% of the in-coincidence eventsin the upper region represented by the red line in figure 12. This cut was optimized inorder to take into account the energy dependence of the MPD. The upper and lower re-gions correspond respectively to volume and cathode events and to anode and mesh events.Figure 13 shows the energy spectra of the two chambers with an MPD above the cut (leftpanel) and with an MPD below the cut (right panel).In the region above the MPD cut the energy spectra show two peaks observed at 33and 46 keVee. It confirms that these events are related to RPR events from the cathode,shown previously on coincidence spectra, with a measured event rate of . ± . − inchamber 1 and . ± . − in chamber 2. These energy spectra were fitted from 25to 60 keVee using the same model as in section 3.2. In table 4 the fit parameter values aregiven. – 18 – nergy [keV] M P D [ A r b . ] M P D c u t Figure 12 : MPD distribution as a function of the event ionization energy in the cham-ber 2. These distributions were obtained applying the ER/NR discrimination. The red linecorresponds to the cut on the MPD. [keVee] ioni E ] - . ke V ee - R [ d ay Ch. 1Ch. 2 [keVee] ioni E ] - . ke V ee - R [ d ay Ch. 1Ch. 2
Figure 13 : The energy spectra measured in 2013 by the chamber 1 (black line) and 2(red line) after ER/NR discrimination with a cut on the event MPD value. The left panelrepresents the energy spectra of volume event and cathode events (above the MPD cut),and the right panel the energy spectra of anode and mesh events (below the MPD cut).The fitted peak positions match with the peak positions previously fitted in section 3.2.It confirms that the observed events are related to RPR events from the Rn decay chain.The ratios of the two peak amplitudes are . ± . for the chamber 1 and . ± . forthe chamber 2. These values are still compatible with a ratio of 2 supporting our statement.The right panel of figure 13 shows the ionization energy distribution of events belowthe MPD cut, events from the mesh and anode. In these spectra, no clear peak structuresare observable, this is due to an incomplete electron avalanche of events crossing the ampli-– 19 –arameter Unit chamber 1 chamber 2 Cst [ day − . keVee − ] . ± . ± . A [ day − . keVee − ] . × ± . × . × ± . × µ [keVee] . ± . . ± . σ [keVee] . ± . . ± . A [ day − . keVee − ] . × ± . × . × ± . × µ [keVee] . ± . . ± . σ [keVee] . ± . . ± . Table 4 : Fit parameters of the cathode energy spectra obtained after the ER/NR appli-cation and a cut on the MPD value.fication space. This anode and mesh backgrounds contribute to the global background witha measured event rate of . ± . − in chamber 1 and . ± . − in chamber 2.In conclusion, using the MIMAC observables, it is possible to estimate the track positionof the main background event sources inside the detector. In this paper, we have presented the first results of the analysis of the MIMAC data runsat the LSM. We have shown, for the first time, the observation of low energy NR 3D tracksfrom daughter nuclei of the Rn decay chain. Finally, using a new MIMAC observablecalled MPD, we have shown that it is possible to separate anode events from volume andcathode events. We have used this observable in order to show the NR ionization energyspectra. This measurement shows the capability of the MIMAC detector and opens thepossibility to explore the low energy recoil directionality signature.Even if the RPR measurement is a validation of the MIMAC detection strategy, itremains a background for dark matter directional detection. The next step will be thediscrimination of this background using the MIMAC observables, the coincidence betweenthe chambers and their directionality. This analysis will be described in a dedicated futurepaper.The radon gas emanation monitoring and its reduction are major topics for rare eventexperiments such as dark matter or double- β decay search experiments. The new degreesof freedom, offered by the observation of 3D low energy NR tracks describing these eventsshed a new light on them improving their localization and discrimination. Acknowledgements
We thank the Laboratoire Souterrain de Modane (CNRS/CEA) and its team for their sup-port all along our operation. We acknowledge F. Mayet, J. Lamblin, J. Brunner, C. Cou-turier and A. Naver-Agasson for many helpful discussions, and O. Méplan and M. Ramd-hane for radioactivity material measurements performed at the LBA. MIMAC collaborationacknowledges the ANR-07-BLAN-0255-03 funding.– 20 – eferences [1] P. A. R. Ade, N. Aghanim, M. Arnaud, M. Ashdown, J. Aumont, C. Baccigalupi et al.,Planck 2015 results, Astronomy & Astrophysics (oct, 2016) A13, [ ].[2] M. W. Goodman and E. Witten, Detectability of Certain Dark Matter Candidates, Phys.Rev.
D31 (1985) 3059.[3] P. Agnes, L. Agostino, I. F. M. Albuquerque, T. Alexander, A. K. Alton, K. Arisaka et al.,Results from the first use of low radioactivity argon in a dark matter search, PhysicalReview D - Particles, Fields, Gravitation and Cosmology (2016) 1–7, [ ].[4] D. S. Akerib, H. M. Araújo, X. Bai, A. J. Bailey, J. Balajthy, P. Beltrame et al., Improvedlimits on scattering of weakly interacting massive particles from reanalysis of 2013 lux data,Physical Review Letters (apr, 2016) 161301, [ ].[5] XENON100 Collaboration, E. Aprile, J. Aalbers, F. Agostini, M. Alfonsi, F. D. Amaro et al.,XENON100 Dark Matter Results from a Combination of 477 Live Days, .[6] J. Billard, L. Strigari and E. Figueroa-Feliciano, Implication of neutrino backgrounds on thereach of next generation dark matter direct detection experiments, Phys.Rev. D89 (2014)023524, [ ].[7] D. Spergel, Motion of the Earth and the detection of weakly interacting massive particles,Phys. Rev. D (Mar., 1988) 1353–1355.[8] J. Billard, F. Mayet, J. Macias-Perez and D. Santos,Directional detection as a strategy to discover galactic Dark Matter, Phys.Lett. B691 (2010)156–162, [ ].[9] D. Mei and A. Hime, Muon-induced background study for underground laboratories,Phys.Rev.
D73 (2006) 053004, [ astro-ph/0512125 ].[10] N. Bozorgnia, G. B. Gelmini and P. Gondolo,Ring-like features in directional dark matter detection, JCAP (2012) 037, [ ].[11] N. Bozorgnia, G. B. Gelmini and P. Gondolo,Daily modulation due to channeling in direct dark matter crystalline detectors, Phys.Rev.
D84 (2011) 023516, [ ].[12] J. Billard, F. Mayet and D. Santos,Exclusion limits from data of directional Dark Matter detectors, Phys.Rev.
D82 (2010)055011, [ ].[13] S. Henderson, J. Monroe and P. Fisher,The Maximum Patch Method for Directional Dark Matter Detection, Phys.Rev.
D78 (2008)015020, [ ].[14] J. Billard, F. Mayet and D. Santos,Assessing the discovery potential of directional detection of Dark Matter, Phys.Rev.
D85 (2012) 035006, [ ].[15] A. M. Green and B. Morgan,The median recoil direction as a WIMP directional detection signal, Phys.Rev.
D81 (2010)061301, [ ].[16] J. Billard, F. Mayet and D. Santos, – 21 – arkov chain Monte Carlo analysis to constrain dark matter properties with directional detection,Phys. Rev. D (Apr., 2011) 075002, [ ].[17] D. S. Alves, S. E. Hedri and J. G. Wacker, Dark Matter in 3D, .[18] S. K. Lee,Harmonics in the Dark-Matter Sky: Directional Detection in the Fourier-Bessel Basis, JCAP (2014) 047, [ ].[19] C. A. J. O’Hare and A. M. Green, Directional detection of dark matter streams, Phys.Rev. D90 (2014) 123511, [ ].[20] P. Grothaus, M. Fairbairn and J. Monroe,Directional Dark Matter Detection Beyond the Neutrino Bound, Phys.Rev.
D90 (2014)055018, [ ].[21] F. Ruppin, J. Billard, E. Figueroa-Feliciano and L. Strigari,Complementarity of dark matter detectors in light of the neutrino background, Phys.Rev.
D90 (2014) 083510, [ ].[22] D. Santos, O. Guillaudin, T. Lamy et al.,MIMAC: A Micro-TPC Matrix of Chambers for direct detection of Wimps, J. Phys. Conf.Ser. (Apr., 2007) 012012, [ ].[23] S. Ahlen, N. Afshordi, J. Battat et al.,The case for a directional dark matter detector and the status of current experimental efforts,Int.J.Mod.Phys. A25 (2010) 1–51, [ ].[24] J. B. R. Battat et al., Readout technologies for directional WIMP Dark Matter detection,Phys. Rept. (2016) 1–46, [ ].[25] D. Santos, J. Billard, G. Bosson et al.,MIMAC : A micro-tpc matrix for dark matter directional detection, J. Phys. Conf. Ser. (Oct., 2013) 012007, [ ].[26] F. J. Iguaz, D. Attié, D. Calvet et al., Micromegas detector developments for dark matterdirectional detection with mimac, J. Instrum. (July, 2011) P07002–P07002, [ ].[27] I. Giomataris, R. De Oliveira, S. Andriamonje et al., Micromegas in a bulk,Nucl.Instrum.Meth. A560 (May, 2006) 405–408.[28] J. Billard, F. Mayet, G. Bosson et al., In situ measurement of the electron drift velocity forupcoming directional dark matter detectors, J. Instrum. (Jan., 2014) P01013–P01013,[ ].[29] J. Richer, G. Bosson, O. Bourrion et al.,Development of a front end ASIC for Dark Matter directional detection with MIMAC,Nucl.Instrum.Meth. A620 (2010) 470–476, [ ].[30] O. Bourrion, G. Bosson, C. Grignon et al., Data acquisition electronics and reconstructionsoftware for directional detection of dark matter with mimac, Nucl.Instrum.Meth.
A662 (2010) 207–214, [ ].[31] S. Burgos, J. Forbes, C. Ghag et al., Track reconstruction and performance of driftdirectional dark matter detectors using alpha particles, Nucl.Instrum.Meth.
A584 (Jan.,2008) 114–128, [ ]. – 22 –
32] J. Brack, E. Daw, A. Dorofeev et al.,Long-term study of backgrounds in the DRIFT-II directional dark matter experiment, J.Instrum. (July, 2014) P07021–P07021, [ ].[33] J. B. R. Battat, J. Brack, E. Daw et al., Radon in the drift-ii directional dark matter tpc:emanation, detection and mitigation, .[34] D. C. Malling, S. Fiorucci, M. Pangilinan et al.,Dark Matter Search Backgrounds from Primordial Radionuclide Chain Disequilibrium,Astropart. Phys. (May, 2013) 21, [ (Feb., 2012) 119–127, [ ].[37] J. Beringer, J. F. Arguin, R. M. Barnett et al., Review of Particle Physics, Phys. Rev. D (July, 2012) 010001.[38] Q. Riffard et al.,MIMAC low energy electron-recoil discrimination measured with fast neutrons, JINST (2016) P08011, [ ].[39] C. Couturier, D. Santos, Q. Riffard et al.To be published in JINST (2007) .[40] S. Biagi, Monte carlo simulation of electron drift and diffusion in counting gases under theinfluence of electric and magnetic fields, Nucl.Instrum.Meth. A421 (Jan., 1999) 234–240.(Jan., 1999) 234–240.