Repositorio Institucional de Documentos

2013 66 Clara Cuesta Soria ANAIS-0: Feasibility study for a 250 kg Nal(Tl) dark matter search experiment at the Canfranc Undreground Laboratory Departamento Director/es Física Teórica Sarsa Sarsa, María Luisa Director/es Tesis Doctoral Autor Repositorio de la Universidad de Zaragoza – Zaguan http://zaguan.unizar.es UNIVERSIDAD DE ZARAGOZA Departamento Director/es Clara Cuesta Soria ANAIS-0: FEASIBILITY STUDY FOR A 250 KG NAL (TI) DARK MATTER SEARCH EXPERIMENT AT THE CANFRANC UNDREGROUND LABORATORY Director/es Física Teórica Sarsa Sarsa, María Luisa Tesis Doctoral Autor 2013 Repositorio de la Universidad de Zaragoza – Zaguan http://zaguan.unizar.es UNIVERSIDAD DE ZARAGOZA Departamento Director/es Director/es Tesis Doctoral Autor Repositorio de la Universidad de Zaragoza – Zaguan http://zaguan.unizar.es UNIVERSIDAD DE ZARAGOZA ANAIS-0: Feasibility study for a 250 kg NaI(Tl) dark matter search experiment at the Canfranc Underground Laboratory Memoria presentada por ClaraCuestaSoria para optar al grado de Doctora en F´ısica Laboratorio de F´ısica Nuclear y Astropart´ıculas Departamento de F´ısica Te´orica ´ Area de F´ısica At´omica, Molecular y Nuclear Universidad de Zaragoza Marzo 2013 Contents 1 Introduction 1 1.1 UnderstandingoftheUniverse ......................... 2 1.1.1 TheStandardCosmologicalModel................... 2 1.1.2 Cosmologicalparameters ........................ 5 1.2 Darkmatterneed................................. 7 1.2.1 Darkmatterevidences.......................... 8 1.2.2 Darkmattercandidates ......................... 11 1.3 Darkmatterdetection .............................. 12 1.3.1 Indirectdarkmatterdetection ..................... 13 1.3.2 Directdarkmatterdetection ...................... 14 1.4 ANAIS experiment . . . . . . . . . ....................... 24 1.4.1 Scientificcase............................... 24 1.4.2 Technicalaspectsoftheproject..................... 26 1.4.3 Experimentalrequirements ....................... 29 2 ANAIS-0 experimental procedure 33 2.1 ANAIS-0 module assembly . . . . ....................... 34 2.1.1 NaI(Tl)crystal.............................. 34 2.1.2 NaI(Tl)crystalencapsulation...................... 34 2.1.3 Photomultipliertubes .......................... 37 2.1.4 Lightguides................................ 43 2.2 ANAIS-0 Setups ................................. 44 i ii 2.3 ANAIS-0 Shielding . . .............................. 46 2.4 CanfrancUndergroundLaboratory....................... 49 2.4.1 Old LSC facilities . . . . . . . . . .................... 49 2.4.2 NewinstallationsoftheLSC ...................... 50 2.4.3 Moving ANAIS-0 to the new LSC facilities . ............. 50 2.5 Signalprocessing................................. 51 2.5.1 Electronichardware ........................... 51 2.5.2 Acquisitionsoftware ........................... 56 2.5.3 Analysissoftware............................. 57 2.6 Slow-control.................................... 58 2.7 Detectorcalibrations............................... 63 2.8 Muon coincident events in ANAIS-0 . . .................... 69 3 Data analysis 73 3.1 Bulk NaI scintillation low energy events selection . . ............. 74 3.2 Neutroncalibration................................ 84 3.3 Asymmetricevents................................ 89 3.4 Energyestimators ................................ 94 3.5 Triggerefficiency.................................. 96 3.6 Precise determination of the NaI(Tl) scintillation constants. . . . ......100 3.7 Quenching factor for αparticles. ........................108 4 Background understanding 111 4.1 Background measurements and radioactive contaminants identification . . . 112 4.1.1 Backgroundmeasurements .......................112 4.1.2 Radioactive contaminants in ANAIS-0 crystal .............114 4.2 Backgroundsimulation..............................132 4.2.1 Geant4simulationdescription......................133 4.2.2 Distributionofcontaminants ......................134 4.2.3 Codevalidationandanalysisofsystematiceffects...........136 Indice iii 4.3 Backgroundmodel ................................140 4.3.1 Backgroundmodel............................141 4.3.2 Comparison between background simulations and measurements . . 144 4.3.3 Otherhypotheticalbackgroundsources ................148 4.4 Conclusionsfromthecomparison........................151 5 Towards the ANAIS experiment 155 5.1 UltrapureNaI(Tl)crystals............................156 5.1.1 MeasurementtechniquesandNaIreferencesamples .........157 5.1.2 UltrapureNaIpowdersamples .....................159 5.2 Photomultipliers .................................163 5.2.1 Gain....................................164 5.2.2 Gaindependencewithrate .......................166 5.2.3 SingleElectronResponse ........................168 5.2.4 Lightcollectionefficiency ........................169 5.2.5 Quantumefficiency............................171 5.2.6 Conclusions................................173 6 ANAIS-25 175 6.1 ANAIS-25 experimental procedure .......................176 6.1.1 GrowingandencapsulationatAlphaSpectra .............176 6.1.2 InstallationatLSC............................177 6.1.3 Datataking................................180 6.1.4 Energy calibration of ANAIS-25 modules . . . . . . . . . ......180 6.2 Background understanding . . . . .......................182 6.2.1 40Kbulkcontent.............................184 6.2.2 238Uand 232Th chainsisotopescontent ................188 6.2.3 Cosmogenicactivation..........................190 6.2.4 Backgroundmodel............................190 6.3 ANAIS-25 data analysis . . . . . . .......................192 iv 6.3.1 Scintillation events selection . . . ....................192 6.3.2 Asymmetry................................195 6.3.3 Triggerefficiency.............................195 6.3.4 Long scintillation constants in NaI(Tl) . . . . .............197 6.4 Lightcollectionefficiency ............................199 Conclusions 210 Resumen y conclusiones 225 Agradecimientos 226 Bibliography 227 1.2. Dark matter need parameters are: baryon density times h2,whereH0= 100hkm/s/Mpc; the CDM density times h2; the dark energy density; the curvature perturbations amplitude (Δ2 R)andthe spectral index of the corresponding power law (ns) are required to include a description of deviations from homogeneity; and finally, the ionization optical depth, τ, related with the probability that a given microwave photon scatters with ionized electrons in the interstellar medium, disturbing then the anisotropies pattern [15]. Curvature perturbations and the density perturbations spectral index have been measured as a complete model of the Universe should include a description of deviations from homogeneity, at least in a statistical way [16]. The universe is in accelerated expansion at present ΩΛ=0.72, and with a matter content of only Ωm=0.28, being negligible the present contribution from the radiation (Ωrad ≈10−4) to the universe energy content. Parameter Symbol WMAP data Fit parameters: Baryon density (×h2)Ω bh20.02264 ±0.00050 Cold dark matter density (×h2)Ω ch20.1138 ±0.00045 Dark energy density ΩΛ0.721 ±0.025 Curvature perturbations 109Δ2 R2.41 ±0.10 Density perturb. spectral index ns0.972 ±0.013 Ionization optical depth τ0.089 ±0.014 Derived parameters: Age of the Universe t013.74 ±0.11 Gyr Hubble parameter H070.0±2.2km/s/Mpc Matter density Ωm0.279 ±0.025 Cold dark matter density Ωc0.233 ±0.023 Baryon density Ωb0.0463 ±0.0024 Table 1.1: Some of the cosmological parameters derived from the WMAP measurements, obtained from [10]. 1.2 Dark matter need If we center our interest in the matter content of the Universe, it is easily found out that it is much higher than the baryon (neutrons and protons) density predicted by the Big Bang nucleosynthesis [17] and confirmed by WMAP data, Ωb=0.0463±0.0024 [10]. This implies the existence of other kind of matter, which neither emits nor absorb electromagnetic radiation, and represents an important contribution to the Universe density, Ωc=0.233± 0.023 [10]. 7 1. Introduction A modification or replacement of Einstein’s relativistic theory of gravity with some new, but not yet found, alternative could also solve the dark matter and, even, the dark energy issue, making unnecessary the introduction of further ingredients in the Universe content. However debate on this issue continues, the scientific community is more favorable to the dark matter hypothesis. 1.2.1 Dark matter evidences First observational evidence of the existence of dark matter in spiral galaxies go back till goes back till 1975, when Vera Rubin was measuring galactic rotation curves. According to Newtonian dynamics, the rotation speed of the bodies bound to a mass distribution M(r), assuming spherical symmetry, should show the following dependence in r: v2 rot(r)=GM(r) r(1.8) with Gthe Newton gravitational constant and M(r) the mass content in the sphere with radius r. However, Vera Rubin observed that rotation curves in the outskirts of the galaxies, where M(r) should be constant, did not decrease with r, but reached an almost constant value [18]. This implied that the galactic mass distribution did not follow the visible mass distribution. Figure 1.3 shows two examples of her observations. Afterwards, many other galactic rotation curves were measured, obtaining similar results, as shown in [19]. The existence of a dark matter halo, extending much farther than the visible galaxy is the more extended explanation of these rotation curves. (a) (b) Figure 1.3: Spectra and derived rotation curves for galaxies NGC 801 (a) and UGC 2885 (b). The curves at the right show the rotational velocities as a function of the distance to the galactic center, measured from the Doppler shift in the emission lines in the spectra. Images obtained from [18]. 8 1.2. Dark matter need In the galaxy cluster scale, many other observational evidences have accumulated since Zwicky used the virial theorem to infer the existence of unseen matter in the Coma galaxy cluster en 1933 [20]. He was the first to refer to such a matter as dunkle Materie.For instance, the gravitational lensing effect: the image of a far galaxy is distorted by a gravitational field generated by a very massive cluster, located between the telescope and the galaxy. The analysis of these images allows to determine the mass distribution of the interposed object, giving as general result much higher cluster masses than those estimated from the luminosity. Moreover, X-ray measurements reveal that the temperatures of the hot gas observed in the galaxy clusters can not be explained with the gravitational field caused by the visible matter of the clusters. Same argument would prevent clusters from being stable structures given the high velocities observed for the galaxies they contain. The Bullet cluster has provided very robust arguments for the existence of dark matter and about its nature, see Figure 1.4. This cluster consists of two large colliding clusters of galaxies, one of the most energetic events known in the Universe since the Big Bang [21]. Hot gas detected by Chandra by the X-rays emissions is shown as two pink clumps in the image and traces the distribution of most of the baryonic matter. The bullet-shaped clump on the right is the hot gas from one cluster, which passed through the hot gas from the other larger cluster during the collision. An optical image from Magellan and the Hubble Space Telescope shows the galaxies in orange and white. The blue areas in this image depict where astronomers find most of the mass in the clusters, traced by the gravitational lensing analysis previously explained. The combination of information derived from observations using three different techniques evidenced a spatial offset of the center of the total mass distribution (blue) from the center of the baryonic mass peaks (pink), whereas the galaxies follow the total mass distribution. This is explained by the slowing down of the hot gas during the collision by a drag force, due to the electromagnetic interaction. If most of the invisible matter of the cluster was not slowed down, it should be collisionless matter: it can only interact gravitationally or very weakly. Therefore, during the collision the dark matter clumps from the two clusters moved ahead of the hot gas, without hindrance, producing the separation of the dark and normal matter seen in the image. Observations of other galaxy cluster collisions, such as MACS J0025.4-1222, also show significant displacement between their hot gas distribution centers and their gravitational potential. However, there are other examples, like the merging galaxy cluster Abell 520, shown in Figure 1.5, formed by a violent collision of very massive clusters: dark matter, galaxies and hot gas distribution are shown [22]. Superimposed on the natural-color image 9 1. Introduction (a) Figure 1.4: This composite image shows the galaxy cluster 1E 0657-556, also known as the Bullet cluster. Hot gas detected by Chandra in X-rays is seen as two pink clumps in the image and contains most of the baryonic matter in the two clusters. The bullet-shaped clump on the right is the hot gas from one cluster, which passed through the hot gas from the other larger cluster during the collision. An optical image from Magellan and the Hubble Space Telescope shows the galaxies in orange and white. The blue areas depict where astronomers find most of the mass in the clusters. The concentration of mass is determined by analyzing the effect of gravitational lensing. Image obtained from [21]. of galaxies are false-colored maps: in orange the starlight from galaxies (as detected by the Canada-France-Hawaii Telescope), in green hot gas (as detected by NASA’s Chandra X-ray Observatory), and in blue the location of most of the mass in the cluster, which is dominated by dark matter. The dark matter map was derived from the Hubble Wide Field Planetary Camera 2 observations by gravitational lensing [22]. The dark matter is concentrated into a dark core containing far fewer galaxies than would be expected if the dark matter and galaxies mass distributions follow each other. Most of the galaxies apparently have sailed far away from the collision. This result could present a challenge to basic theories of dark matter, which predict that galaxies should be anchored to the invisible substance, even during the shock of a collision. 10 1.2. Dark matter need Figure 1.5: Composite image showing the distribution of dark matter (blue), galaxies (orange), and hot gas (green) in the core from the Abell 520 galaxy cluster, formed from a violent collision of massive galaxy clusters. The blend of blue and green in the center of the image reveals that a clump of dark matter resides near most of the hot gas, where very few galaxies are found. Image obtained from [22]. 1.2.2 Dark matter candidates In order to explain all the accumulated observational evidences from galaxy and galactic clusters dynamics, visible matter, baryonic content, cosmological data at large scales, etc. a very special kind of particle is required: stable (or very long lived, in the cosmological timescale), interacting only weakly with matter, non-relativistic at the decoupling moment, neutral and massive. Looking at the Standard Model of Particle Physics, only neutrinos fulfill partially these requirements, but they were relativistic at the decoupling moment, and they would not have been able to produce as much small scale structure in the universe as we observe. Strong hints at physics beyond the Standard Model have been found. In some of those scenarios, compelling dark matter candidates appear [23–25]. In particular, a new particle, the axion, was postulated in 1977 by Peccei and Quinn to solve the CP conservation problem in QCD (quantum chromodynamics). Axion-like particles have become interesting dark matter candidates and are being searched for by different experimental strategies. However, the axion can only explain the dark matter if its mass is tuned in a small range of values. The supersymmetric (SUSY) extension of the Standard Model provides good dark matter candidates, although in a very model dependent way. SUSY 11 1. Introduction extensions are an appealing concept because of their remarkable properties, for example, with respect to gauge coupling unification, to solve the hierarchy problem, and the embedding of gravity. As superpartners of the Standard Model particles, new particles appear including fields that are electrically and color neutral. Since they have not been detected at particle accelerators, these sparticles must be heavy and/or extremely weakly interacting. In particular, the ligthest neutralino, the gravitino or the axino could be the Lightest Supersymmetric Particle, stable in many SUSY models, able to explain the nonbaryonic dark matter. The neutralino is part of the Minimal Supersymmetric Standard Model which provides a solution to the hierarchy problem and allows for gauge coupling unification. Being the superpartner of the graviton and the gauge field associated with supergravity, the gravitino is equally well motivated with a mass that reflects the SUSY breaking scale. As the superpartner of the axion, also the axino appears naturally once the strong CP problem is solved with the Peccei-Quinn mechanism in a SUSY setting. The fact that neutrinos have mass is suggestive of the existence of right-handed, gauge singlet sterile neutrinos, which may constitute also dark matter if their masses are small. It is worth noting that the number of particles types contributing to dark matter is unknown and it does not restrict to one single. In general, dark matter candidates in a model independent way can be referred to as WIMPs, acronym for Weakly Interacting Massive Particles. These particles beyond the standard model of particle physics could be produced in accelerators [26], like the LHC, and many information about the nature of the dark matter could be obtained, because such a finding would imply new physics below the TeV scale. However, at the moment the data collected in pp collisions at √s= 7 TeV analyzed by the CMS [27,28] and ATLAS [29] collaborations, corresponding to integrated luminosities of 5.0fb−1and 4.6fb−1, respectively, do not show hints for new phenomena. As good agreement is observed between the data and the standard model predictions, the results are translated into exclusion limits on pair production of weakly interacting dark matter candidates. 1.3 Dark matter detection Although a large amount of dark matter is required to explain the observations of the universe at different scales, deciphering the nature of these particles is a challenge for astrophysics, cosmology, particle and nuclear physics. There are different approaches to detect the elusive dark matter particle and several experiments have been designed 12 1.3. Dark matter detection with such a goal. In the direct detection ones, the scattering of the WIMPs off a target nucleus in a convenient detector material is searched for, whereas in the indirect detection experiments the products of the WIMP annihilation are studied. 1.3.1 Indirect dark matter detection The self-annihilation of WIMP particles in astronomical objects may leave observable imprints in several components of the cosmic rays. WIMPs are expected to annihilate efficiently in regions where they accumulate, such as the center of galactic halos or substructures as dwarf galaxies, because the annihilation rate depends on the square of the dark matter particles number density. Once they annihilate, they produce secondary particles, like quarks and gauge bosons, which subsequently fragment and decay in photons, protons, antiprotons, electrons, positrons and neutrinos. The typical energy of these final states is about a tenth of the dark matter particle mass, so to search indirectly for dark matter, an excess of photons (gamma rays), antimatter (positrons and antiprotons) or neutrinos in astrophysical data at energies between 1 GeV and 10 TeV have to be searched for. These experiments are complex instruments devoted to study gamma rays, cosmic rays or neutrinos and the study of dark matter is usually a byproduct of their main goal [30–33]. Gamma rays play an outstanding role among the possible messengers of a dark matter annihilation signal as they propagate almost unperturbed and directly point to their sources, hence, they could give distinctive spatial signatures, as well as sharp spectral signatures, in the case there is a direct annihilation channel to photons (a monoenergetic line at the WIMP mass would be expected) [34]. Gamma rays can be observed directly from space with satellites or indirectly, with ground-based telescopes via the Cherenkov light emitted by the showers of secondary particles produced by the interaction of very high energetic gamma rays hitting the upper part of the atmosphere. The Fermi Gammaray Space Telescope was launched in 2008 carrying the Large Area Telescope (LAT). The design life of the mission is 5 years. The LAT is an imaging high-energy gamma-ray telescope covering the energy range from about 20 MeV to more than 300 GeV. The Fermi-LAT collaboration has published upper limits at 95% C.L. from the absence of signal excess between 7 and 200 GeV [35]. However, there are recent claims of an observation of a 130 GeV line near the galactic center [36]. Due to the small area of detectors on satellites, higher energies can be only detected by ground-based Imaging Atmospheric Cherenkov Telescopes (IACTs) as HESS (High Energy Stereoscopic System), MAGIC with regular observations since 2009, and VERITAS, operational since 2007. None of them has found any excess and limits on the self-annihilation 13 1. Introduction cross section for WIMPS have been presented: by HESS from the Sculptor and Carina dwarf galaxies [37], by MAGIC from the dwarfs Willman [38] and Segue 1 [39], and by VERITAS also from Segue 1 [40], for example. The future Cherenkov Telescope Array (CTA) will provide a significantly improved sensitivity [41]. Positrons or antiprotons excess can be searched with cosmic rays telescopes [42]. The PAMELA (Payload for Anti-Matter Exploration and Light-Nuclei Astrophysics) satellite was launched in 2006, and two years later an excess of positrons was reported by the collaboration, which has been widely discussed [43,44]. AMS-01 (Alpha Magnetic Spectrometer) confirmed this excess and new results will be released in the next future by the AMS-02 spectrometer which was installed in 2011 in the International Space Station, confirming or not the positron excess detected by PAMELA. Cosmic rays more energetic should be detected via the showers of secondary particles, as it is done from ground in the Pierre Auger Observatory [45] or will be done by JEM-EUSO from the space [46]. Neutrinos can be detected with the IceCube Neutrino Observatory at the geographic South Pole [47], whose construction ended in 2010. Charged particles produced in neutrino interactions in about 1 km3of Antarctic ice are detected by the Cerenkov radiation, using a regularly spaced array of 5160 photomultiplier tubes buried between 1450 to 2450 meters of antarctic ice. Other experiment able to detect neutrinos is the ANTARES neutrino telescope, completed in 2008 [48]. It is located at a depth of 2475 m in the Mediterranean Sea, 42 km from the French coast. This detector consists of a tridimensional array of PMTs distributed along twelve detection lines, and is optimized for the detection of muons from high-energy astrophysical neutrinos. Moreover, the Galactic Center is accessible for ANTARES, being in the Northern Hemisphere. An important advantage of neutrino telescopes with respect to other indirect searches is that a potential signal would be very clean, since no other astrophysical explanations could mimic it (like pulsars can do in the case of the positron excess seen by PAMELA). As neutrino interactions are so rare, only the near annihilation sources could generate measurable fluxes: WIMPs gravitationally bound to the Sun or the Earth could produce an excess of neutrinos that could be detected by these experiments. No excess of neutrinos coming from the Sun, nor from the Earth, has been found for the moment 1.3.2 Direct dark matter detection Detecting dark matter particles is not a simple issue, as it has been explained in section 1.2, they are expected to interact with matter mainly by elastically scattering off nuclei. Small 14 1.3. Dark matter detection amounts of energy will be released in the target material in the form of nuclear recoil kinetic energy. The expected dark matter interaction rate of WIMPs having a mass mχin a target consisting of one type of nucleus of mass mN(and the total detector mass Mdet)canbe calculated as [7]: dR dER =ρ0Mdet mNmχvmax vmin vf(v)dσχN dER (v,ER)dv (1.9) where ρ0is the local WIMP density, f(v) the WIMP speed distribution in the detector reference frame normalized to unity. Both depend on the astrophysical observations, in the Standard Halo Model the halo is described as an isotropic and isothermal sphere of collisionless particles with ρr∝r−2and f(v) Maxwellian in terms of the WIMP velocity in the halo reference frame: f(v)= 1 v3 0π3/2 exp −v2 v2 0(1.10) where v0is the most probable speed value, related with the root mean square velocity of such a distribution by vrms =3/2v0. The differential cross section for the WIMPnucleus scattering is dσWN dER(v,ER). Parameter Typical value ρ00.3 GeV/cm3 vesc 544 km/s v0220 km/s vrms 270 km/s ν⊕28.9 km/s Table 1.2: Typical values used in the calculation of the WIMP interaction rate. Data obtained from [49, 50]. As WIMP-nucleon relative speed is of order of 100 km·s−1, the elastic scattering occurs in the non-relativistic limit, and the recoil energy of the nucleus, depends on the scattering angle θ(calculated in the center of mass reference frame): ER=μ2 χN v2(1 −cos θ∗) mN (1.11) 15 1. Introduction with μχN =mχmN (mχ+mN). The lower limit of the WIMP speed integration is vmin = mNER 2μ2 χN , the minimum WIMP speed that causes a recoil energy ERand the upper limit, vmax =vesc, the maximum WIMP speed to be bound to the Milky Way. Then, the total rate (usually expressed in cpd/kg) is calculated by integrating from the smallest recoil energy that the detector is able to measure, ETto the maximum recoil energy possible to detect (determined by kinematics): R=Emax ET dER ρ0 mNmχvesc vmin vf(v)dσWN dER (v,ER)dv(1.12) The WIMP-nucleus differential cross section dσWN dERis computed from the Lagrangian that describes the interaction of a given WIMP with ordinary matter and encodes the particle physics input. In general, it can be separated into a spin independent (SI) and a spin dependent (SD) contribution: dσWN dER =dσWN dERSI +dσWN dERSD =mN 2μ2 χN v2σSIF2 SI(ER)+σSDF2 SD(ER)(1.13) where σSI and σSD are the SI and SD components of the WIMP-nucleus cross sections at zero momentum transfer, and FSI and FSD are the form factors which account for the finite size of the nucleus, usually important at high momentum transfers. These nuclear point-like cross-sections can be expressed, in quite a general scenario, as: σSI =4μ2 χN π[Zfp+(A−Z)fn]2σSD =32μ2 χN πG2 F J+1 J[apSp+anSn]2(1.14) where fp,fn,ap,andanare the effective WIMP couplings to neutrons and protons in the spin-independent and spin-dependent case, respectively. They can be calculated using an effective Lagrangian of the theoretical model. They depend on the contributions of the light quarks to the mass of the nucleons and on the quark spin distribution within the nucleons, respectively, and on the composition of the dark matter particle. Spand Snare the expectation values of proton and neutron spin operators in the limit of zero momentum transfer, and must be determined using detailed nuclear model calculations. Usually, in the sensibility plots, SI coupling to protons and neutrons is not distinguished, and it is only consider the total spin-independent cross section; for the SD coupling they should be calculated individually depending on the nucleus considered. In general, in the spin-independent interaction is considered fp=fn, and then, the cross-sections depend 16 1.4. ANAIS experiment the projects ANAIS and DM-Ice. This work has been done in the frame of the ANAIS experiment, which will be briefly presented in the next section and the following chapters will detail the work carried out for the last years. The DM-Ice experiment will use also the same target and technique, with the particularity of the location: the South Pole. This experiment presents some advantageous features with respect to DAMA/LIBRA and ANAIS as are: the low operation temperature of the detectors, buried in the Antarctic ice, the high radiopurity and shielding offered by the surrounding ice and the southern position that inverts the possible correlation of any modulation in the signal with the temperature, allowing a better control of systematics. However, it has also some drawbacks as the impossibility to access to the detectors once buried, and hence, to calibrate them. The CUORE experiment deserves a special mention because, although its main goal is the double beta decay search, it has achieved a low energy (3 keV) threshold in a test run [74] with a background of 2-25 cpd/keV/kg and will be able to look for annual modulation first with CUORE-0 setup, and later with the whole of CUORE detectors (amounting 741 kg TeO2bolometers). A next generation of ton-target experiments has been planned. CRESST, EDELWEISS and ROSEBUD have rejoined efforts in the project EURECA [75], which could use Ge detectors (as those from EDELWEISS) or CaWO4scintillating bolometers (as those from CRESST), but could take into consideration the inclusion of other targets, in particular new scintillating bolometers being tested in the frame of ROSEBUD. The total mass of EURECA is planed to be 1 ton, and a multitarget approach could be envisaged. Combination of different target nuclei in the same set-up could bring interesting information in the case a positive signal is found. Moreover, the versatility of the target choice in the case of scintillating bolometers could allow to tune the target to the results of other experiments, in order to complement them and derive the parameters of the WIMPs, breaking degeneracies [76]. The XENON experiment will also be enlarged to 1 ton of fiducial mass. LUX and ZEPLIN will join efforts to build LZ with 7 t of liquid Xe. Also, SuperCDMS [77] will use improved Ge detectors, called iZIPs, to significantly improve the rejection of surface events, main background limitation in CDMS experiment. Moreover, EURECA and SuperCDMS have signed a Memorandum of Understanding in order to progress towards common goals and sharing know-how and results. Far into the future, two large scale experiments have been proposed: MAX [78] in the US using 40 t of liquid Ar and 20 t of liquid Xe and DARWIN [79] in Europe, using 10 t of liquid Xe and/or Ar. 23 1. Introduction 1.4 ANAIS experiment 1.4.1 Scientific case The ANAIS experiment will use NaI(Tl) as target material and will profit from its advantageous features in the search for dark matter. For instance, a low energy threshold can be achieved thanks to the high light yield, it is sensitive to Spin Dependent interactions because 100% of their isotopic content consists of nuclei having A-odd, and it combines light and heavy nuclei, that guarantees good sensitivity in the low and high mass WIMP ranges. Moreover, the state-of-the-art in the building of the detectors and the operation at normal temperature simplify the experimental challenge. ANAIS experiment will also profit from a large detection mass (it is easy to grow very large NaI(Tl) crystals) to analyze the annual modulation effect in the signal. Similar results to those obtained by DAMA/LIBRA collaboration (see previous section) should be found in ANAIS if WIMPs are responsible of such a signal. ANAIS experiment would confirm or refute that result in a model independent way. NaI(Tl) scintillators are background limited because no discrimination between nuclear and electron recoils is possible in the low energy region [80, 81]. However, the main disadvantage of the NaI(Tl) scintillators is the low relative efficiency factor (or quenching factor, Q) for the conversion of the energy deposited by nuclear recoils into visible signal. It is well known that the scintillation yield strongly depends on the type of particle. The relative efficiency factor between nuclear recoils and electrons, is defined as the ratio between the scintillation signal produced by a nuclear recoil depositing an energy in the scintillator and that produced by an electron depositing the same energy. Energies in the experiment are usually expressed as electron equivalent energies, because available calibration sources are beta/gamma sources. In the following chapters, we will refer to electron equivalent energies (keVee or just keV) unless it is otherwise stated. In order to compare with the expected dark matter interaction rates, conversion between nuclear recoil an electron equivalent energies is mandatory and Q should be well known. Qvaluesof∼0.3 for Na recoils and ∼0.1 for I recoils have been reported and were taken as constants in the low energy region [81–87]. Very recently, new measurements have been presented showing much lower values for that quenching factor for Na recoils in the low energy range, as well as strong energy dependence. Confirmation of such a result would imply important shift in the WIMP parameter space region within reach of ANAIS experiment [88]. Hence, a much better knowledge of this parameter is required in order to compare data from NaI experiments and those from other targets. 24 1.4. ANAIS experiment The expected dark matter signal in a NaI(Tl) target is shown in Figure 1.9 for different WIMP candidates: σSI =10 −5pb and WIMP masses of 10 and 50 GeV, and for different values of Q for Na and I nuclear recoils. The expected amplitude of the modulation is also shown in Figure 1.9. The relevance of the quenching factor considered is clearly observed in this Figure. However, this factor should not be important in the comparison between DAMA/LIBRA and ANAIS results, as far as the same target is used. By reaching 1 keVee threshold and a background low enough, compatibility with DAMA/LIBRA signal could be sought in the average rate, without waiting several cycles to analyze the annual modulation. Energy (keVee) 0123456 cpd/keVee/kg -7 10 -6 10 -5 10 -4 10 -3 10 -2 10 -1 10 1 10 2 10 Na Q=0.3 Na [Collar] Na [Tretyak] I Q=0.1 I [Collar] I [Tretyak] = 10 GeV W M = 1e-05 pb SI σ (a) Energy (keVee) 0123456 Sm (cpd/keVee/kg) -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 0.25 Na Q=0.3 Na [Collar] Na [Tretyak] I Q=0.1 I [Collar] I [Tretyak] = 10 GeV W M = 1e-05 pb SI σ (b) Energy (keVee) 0123456 cpd/keVee/kg -7 10 -6 10 -5 10 -4 10 -3 10 -2 10 -1 10 1 10 2 10 Na Q=0.3 Na [Collar] Na [Tretyak] I Q=0.1 I [Collar] I [Tretyak] = 50 GeV W M = 1e-05 pb SI σ (c) Energy (keVee) 0123456 Sm (cpd/keVee/kg) -0.35 -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 Na Q=0.3 Na [Collar] Na [Tretyak] I Q=0.1 I [Collar] I [Tretyak] = 50 GeV W M = 1e-05 pb SI σ (d) Figure 1.9: Expected dark matter rates in NaI, energy is shown in electron equivalent units. Different signals for WIMPS of Mw=10GeV and Mw=50GeV with σSI =10 −5pb and different values of Na and I nuclear recoils quenching factor (as reported on [85,88]) have been considered [89]. It is interesting to mention here the possibility of building a NaI scintillating bolometer, able to simultaneously measure the heat and light signals. With this technique, it would be possible to discriminate between nuclear and electron recoils [90], and the DAMA/LIBRA positive signal could be confirmed even with low exposures. This possibility has been studied in the frame of the ANAIS and ROSEBUD experiments [91]. With the long-term goal of developing a scintillating NaI bolometer, we have tested NaI(Tl) crystals coated 25 1. Introduction by vapor-deposited poly-p-xylylene (parylene) and studied their optical and mechanical behaviour in the mK range. 1.4.2 Technical aspects of the project The ANAIS experiment is the large scale conclusion of previous studies carried out with different prototypes. ANAIS was conceived to be made with 10 hexagonal NaI(Tl) crystals of 10.7 kg each made by BICRON (now Saint Gobain) (see Figure 1.10). These detectors were part of the 14 detectors used in an experiment which looked for the 76Ge ββ decay to the first excited state at the Modane Underground Laboratory first, and then at the Canfranc Underground Laboratory. They have been stored underground at the LSC since thelateeighties. After that, the NaI32 experiment searched for dark matter at the LSC with 3 of these detectors (amounting 32.1 kg) and accumulating two years of data taking. Bounds on WIMP masses and cross-sections were derived from the absence of positive hints, neither in the usual analysis of the total rate, nor in a pioneer modulation analysis [92–94]. Then, one of the NaI crystals was chosen to be further studied and modified: it was used to build ANAIS prototype I [95], and after decoupling the PMT and removing the original stainless-steel encapsulation (see Figure 1.10) it was used in prototypes II [89] and III. In prototype II a copper box was used to allow testing easily different light guides geometries, lengths, optical couplings and reflector/diffuser materials, see Figure 1.11.a, whereas in prototype III (PIII) a tight copper encapsulation was designed and PMTs and/or light guides were coupled in a second step, see Figure 1.11.b. As main result of the study of these prototypes the natural potassium content of the bulk crystal was measured to be too high to allow their use in a dark matter search experiment. Subsequently, all the available crystals were characterized, see section 4.1.2, resulting contaminations of the same magnitude. This result implied a significant change in the ANAIS experiment timeline. The ANAIS experiment had then to be redefined and, in order to reach the experimental goals, the final proposal consisted of 250 kg of ultrapure NaI(Tl) crystals to study the expected annual modulation in the galactic dark matter signal. An ultrapure powder provider and manufacturer of the ultra-low-background crystals had to be searched for, see section 5.1. The total NaI(Tl) mass will be divided into 20 modules of 12.5 kg each. Each module will be coupled to two photomultipliers to increase the light collection efficiency and reduce the photomultipliers events rate. A copper housing will close tightly 26 1.4. ANAIS experiment (a) (b) Figure 1.10: 10.7 kg NaI(Tl) crystal from BICRON. Original detector configuration: stainless steel encapsulation and PMT coupled (a), and bare NaI(Tl) crystal (b). (a) (b) Figure 1.11: (a) Prototype II, a 10.7 kg NaI(Tl) crystal from BICRON inside a Teflon cylinder used as light diffuser material and coupled to light guides, all inside a copper box tightly closed to keep a low relative humidity atmosphere. (b) Prototype III uses the same crystal enclosed in a copper encapsulation with the PMTs directly coupled to 1 cm quartz windows, in a second step. the crystals (NaI is highly hygroscopic), which will be covered by a Teflon layer as light diffuser to increase the light collection efficiency. In order to preserve the NaI(Tl) ANAIS detectors from the environmental background sources, the crystals must be conveniently shielded. The ANAIS shielding will be very similar to the shielding of the ANAIS-0 module, which will be described in section 2.3. The Figure 1.12 shows an artistic view of the ANAIS experiment, with the different parts of the shielding. It will consist of the following layers: •The first layer of the shielding consists of lead: 10 cm of ancient lead plus 20 cm of low activity lead to attenuate the γbackground. All the lead required for the 27 1. Introduction Figure 1.12: ANAIS experiment artistic view. whole ANAIS shielding is already available underground, at the LSC, ready for the mounting. •An anti-Rn box is required to allow tightly closing the shielding to prevent the entrance of airborne radon inside the experimental space (being continuously flushed with radon-free nitrogen furnished by the LSC). The box has been tested in ANAIS0 and an expansion of it could house the whole ANAIS shielding. However, other options are still being considered. •Plastic scintillators covering maximally the ANAIS shielding to act as active muon vetoes have been especially designed: six vetoes are ready to be used in the ANAIS experiment (three of them were used in the ANAIS-0 setup) and the other 10 have been already ordered. •Finally, a neutron shielding combining boron-loaded water and Polyethylene blocks which are also stored at the Canfranc Underground Laboratory. This neutron shielding is designed in such a way that a part could be easily removed to allow weekly calibrations. 28 1.4. ANAIS experiment The electronics to be used in the ANAIS experiment is the same designed for ANAIS0, see section 2.5 and [96], and ANAIS-25 experiments, but scaled to 20 modules and 2 PMTs/module. It has been already tested with up to 2 modules simultaneously and further tests are in progress with a higher modularity. Fine-tuning of some of the components is still to be done. Acquisition and analysis programs have been already prepared and tested. Robust analysis procedures can be concluded after all the work carried out and presented mainly in chapters 3 to 6. The ANAIS experiment plans to start measuring as soon as the new ultrapure NaI(Tl) crystals are available. An stable measurement during several years is required to look for the annual modulation signal with a high enough confidence level. 1.4.3 Experimental requirements The ANAIS experiment has three main requirements: keeping energy threshold and radioactive background as low as possible, and operation in stable conditions. To study the feasibility of the ANAIS experiment, with the ANAIS-0 prototype we have worked in the corresponding directions: •Energy threshold below 2 keV should be achieved. This goal is required in order to improve sensitivity for the annually modulated WIMP signal (see Figure 1.13.a and b) and it has been addressed by testing new photomultiplier tubes: with high quantum efficiency PMTs and very low radioactive background PMTs that could allow to avoid the use of light guides, see section 5.2. Also low energy events filtering analysis has been developed in order to reject non-bulk-NaI scintillation events, that dominate the background below 10 keVee, see section 3.1. Reaching a threshold at the 1 keV range would really increase the ANAIS sensitivity for the dark matter signal, see Figure 1.13.c and d. •Background at low energy should be kept as low as possible. After the full background assessment of the hexagonal NaI(Tl) crystals, of 10.7 kg each, which were originally thought as ANAIS target, the measured 40Kbulk crystal contamination appeared as the most dangerous at the very low energies of interest for ANAIS (see section 4.4) and forced us to look for more radiopure crystals, implying the development of NaI powder purification techniques and the search for new providers (see section 5.1). The ANAIS-0 crystal contamination, and with a special care the 40Kactivity, has been thoroughly studied in this work, and the same analysis has been applied to the new radiopure ANAIS-25 crystals (see sections 4.1 and 6.2). 29 1. Introduction Wimp mass (GeV) 1 10 2 10 3 10 (pb) SD σ -2 10 -1 10 1 10 2 10 3 10 =0.1 I =0.3 Q Na Q Tretyak Collar Bkg: 2 c/keV/kg/d Energy window: 2.0-4.0 keV (a) Wimp mass (GeV) 1 10 2 10 3 10 (pb) SI σ -8 10 -7 10 -6 10 -5 10 -4 10 -3 10 -2 10 -1 10 =0.1 I =0.3 Q Na Q Tretyak Collar Bkg: 1 c/keV/kg/d Energy window: 2.0-4.0 keV (b) Wimp mass (GeV) 1 10 2 10 3 10 (pb) SD σ -2 10 -1 10 1 10 2 10 3 10 =0.1 I =0.3 Q Na Q Tretyak Collar Bkg: 2 c/keV/kg/d Energy window: 1.0-2.0 keV (c) Wimp mass (GeV) 1 10 2 10 3 10 (pb) SI σ -8 10 -7 10 -6 10 -5 10 -4 10 -3 10 -2 10 -1 10 =0.1 I =0.3 Q Na Q Tretyak Collar Bkg: 1 c/keV/kg/d Energy window: 1.0-2.0 keV (d) Figure 1.13: Annual modulation sensitivity curves estimated for NaI considering different energy windows, different backgrounds and quenching factors [89]. The materials to be used in the ANAIS experiment must be selected in terms of radioactivity. All of the ANAIS-0 module components have been screened with HPGe spectrometry at the LSC. A Geant4 simulation of the background has been carried out and a background model has been proposed, see section 4.3. This background model has allowed to understand much better the different background contributions present in ANAIS-0 and such a knowledge has been applied in the design of the new modules. The same simulation code will be applied, with few modifications, to the whole ANAIS experiment. The ANAIS shielding has been tested with ANAIS-0 module, showing good performance. In particular, the muon veto system (although only having partial shielding coverage) has been setup and muon related events have been studied with ANAIS-0 module. •Very stable operation conditions should be guaranteed. The ANAIS experiment plans to measure continuously during several years and any kind of instability could jeopardize the final result. Monitoring every relevant parameter has been implemented with the ANAIS-0 module at the LSC - Hall B (see sections 2.4 and 2.6) in order to discard environmental interferences in the events rate that could 30 1.4. ANAIS experiment affect the annual modulation analysis. Weekly calibrations at high and low energy have been carried out (see section 2.7) and the same will be done when running the ANAIS experiment. An stable electronic chain, and acquisition system are required. They have been upgraded, during the ANAIS-0 operation. The electronics design is compact and homogeneous consisting on VME and NIM modules, providing a solid performance. Some problems concerning channels cross-talk due to the HV power supply are still unsolved. For the acquisition system part, it has overcome satisfactory stress tests and long runs, see section 2.5. 31 2.1. ANAIS-0 module assembly Name Manufacturer Model Serial number ETL Electron Tubes Limited 9302B 1034 HQE Hamamatsu R6233-100 SEL 1062DA 1066DA LB Hamamatsu R6233-100 MOD ZE5331 ZE5336 ULB Hamamatsu R11065 SEL ZK5171 BA0086 BA0057 VLB Hamamatsu R6956 MOD ZK5902 ZK5908 Table 2.2: Different photomultipliers studied throughout this work: Low background PMTs from Electron Tubes Limited (ETL), Hamamatsu high quantum efficiency (HQE), low background (LB), ultra low background (ULB) and very low background (VLB) photomultipliers. Names shown in the first column have been given by the ANAIS research team for internal use, attending to the main properties of each model. Manufacturer, model and serial number are indicated for each of them. (a) (b) (c) (d) (e) Figure 2.2: Different photomultipliers studied: (a) Low background PMTs from Electron Tubes Limited (ETL), (b) Hamamatsu high quantum efficiency (HQE), (c) low background (LB), (d) ultra low background (ULB) and (e) very low background (VLB) PMTs. See Table 2.2 for more details about each model. Ham LB, ULB and VLB PMT models present flying leads. the main features for each PMT unit as given by Hamamatsu. The minimum quantum efficiency values requested and quoted by Hamamatsu are also shown in this table. A couple of units of each model were bought to study the corresponding performance, and, if first tests were good enough, they were installed in the ANAIS-0 module at the LSC to test them in a real situation. In a previous step, all the PMTs were characterized at the University of Zaragoza (results will be presented in section 5.2) and were screened for radiopurity with HPGe spectrometry at LSC. Results of such radioactive screening are 39 2. ANAIS-0 experimental procedure Units ETL HQE, LB ULB VLB 9302B R6233-100 R11065 R6956 GENERAL Diameter in 33 3 3 Spectral response nm 285-630 200-650 300-650 λof max. response nm 420 420 420 PK material SBA BA BA PK Min. Effective mm 70 64 70 Dynode structure L B+L B+L B+L Number of Stages 9 8 12 10 Operating Ambient T C-30/+60 -186/+50 -30/+50 Max. Supply Voltage V1300 1500 1750 1500 Max. Av. Anode Curr. mA 0.1 0.1 0.1 0.1 CHARACTERISTICS AT 25◦C Cathode Luminous Sens. μA/lm 75 130 90 100 QE @ peak % 30 35 25 QE @ 420 nm % 34 Cath. Blue Sens. Index 13.5 10 11 Anode Luminous Sens. A/lm 50 30 450 100 Gain ·10572.35 10 Anode Dark Curr. (max.) nA 0.5(5) 10(30) 10(100) 6(60) Dark Current Rate s−1500 Anode Pulse Rise Time ns 7.5 6.0 5.5 9.5 Electron Transit Time ns 125 52 46 60 Transit Time Spread ns 15 6.5 13 Pulse Linearity mA 30 (±5%) 20 (±2%) 30 (±2%) Table 2.3: Technical information of the different PMT models tested. The PMT model R6233-100 shares main features with HQE and LB PMTs, the main difference is the selection of special building materials in the LB PMT model that makes them suitable for low background applications. SBA stands for Super Bialkali and BA for Bialkali. B means Box, and L Linear focused. Transit time spread is given as the FWHM of the electron transit time distribution. presented in Table 2.5. On view of the high radioactivity level of the HQE PMTs, they were not tested in ANAIS-0. Copper caps were designed and specially manufactured to house the PMTs and the voltage dividers. They are made of copper specifically electroformed at the University of Zaragoza with such a purpose. Because of the different shape of Ham and ETL PMTs, both casings are different (see Figure 2.3). After the moving of ANAIS to the new LSC installations all the coupling and decoupling of the PMTs to the ANAIS-0 module was done inside the LSC clean room 40 2.1. ANAIS-0 module assembly Name Serial N. Sensitivity D.C. Q.E. Luminous Blue min. meas. C. A. C. Rate A. 420 nm 420 nm peak μA/lm A/lm Index s−1nA %%% ETL 1034 HQE 1062DA 125 16.5 13.7 3.12 40 1066DA 143 71.6 14 20.9 40 LB ZE5331 127 38.9 13.9 8.5 32 ZE5336 132 43.8 13.9 18 32 ULB BA0086 115 516 14.7 15 3028.68 BA0057 145 1120 11.9 8.1 3032.90 VLB ZK5902 111 225 14.4 577 2.6 33 34.90 38.20 ZK5908 99 66 14.3 57 0.1 33 34.30 37.20 Table 2.4: Main features of the PMT units tested in this work, as indicated by the manufacturer. C. stands for cathode, A. for anode and D.C. for dark current. The first column of the quantum efficiency indicates the minimum QE guaranteed by Hamamatsu, whereas the other two columns show particular values of the PMT units purchased (only given by the manufacturer for ULB and VLB units). Values shown between  indicate that only the average value for the purchased units had been guaranteed. Component 40K232Th 238U226Ra 60Co 137Cs mBq/unit PMTs: ETL 1034 420 ±50 24 ±4 220 ±12--- HQE 1062DA (185 ±1)103420 ±40 5100 ±30--- LB ZE5331 678 ±42 68 ±3 100 ±3--- ZE5336 647 ±56 75 ±4 109 ±5--- ULB BA0086 12 ±73.6±1.247±28 8.0±1.24.1±0.7BA0057 24 ±91.8±1.059±28 6.1±1.25.1±0.8VLB ZK5902 97 ±19 20 ±2 128 ±38 84 ±3- - ZK5908 113 ±13 20 ±2 150 ±34 88 ±3- - Light guide <21 <4.1<120 <4.7<1.4<1.1 SMD V.D. - - 22.0±1.5--- PMT C.E. <10 <1.0<26 <1.6<0.7<1.0 Table 2.5: Values or upper limits given at 95% C.L. for the content on the main radioactive isotopes for the PMTs, light guides, corresponding PMT copper encapsulations (PMT C.E.) and SMD voltage dividers (SMD V.D.). (class 10000), see section 2.4.3. A copper cleaning protocol (based on that of the CUORE 41 2. ANAIS-0 experimental procedure (a) (b) (c) (d) Figure 2.3: Images of the PMTs and voltage dividers inside their electroformed copper casings: ETL PMT (a) and Ham PMT (b). PMT copper casing during the cleaning process, using nitric acid inside an ultrasonic bath (c). Voltage dividers, both sides can be seen: showing the SMD components and the connectors for the PMT leads (d). experiment [107]) was applied to all the copper pieces after machining them. It consists of the following steps: 1. Mechanical cleaning: Grinding should be done when machining left-overs remained. 2. Soap cleaning: Pieces are put into a polyethylene bag together with a solution of ultra-pure (deionized) water and 5% of Micro 90 basic soap. The bag with the piece should be in the ultrasonic bath at 40 oC for 1 hour and then, be rinsed with ultrapure water. Pieces are put into another polyethylene bag together with a solution of ultra-pure water and 5% of Elma 60 acid soap. The bag with the piece should be in the ultrasonic bath at 60 oC for half an hour and afterwards, be rinsed again with ultra-pure water. 3. Acid etching: Pieces are put into a polyethylene bag together 0.5 M of nitric acid in ultra-pure water. The bag with the piece should be in the ultrasonic bath at 40 oC forhalfanhourandthen,berinsedwithultra-purewater. 42 2.1. ANAIS-0 module assembly 4. Passivation: The pieces are then put into a polyethylene bag with a solution at 10% of citric acid in ultra-pure water. The bag is kept in the ultrasonic bath at 60 oC for 1 hour. The piece should be rinsed with ultra-pure water and dried with clean nitrogen gas. A picture of the acid nitric bath step is shown in Figure 2.3.c. 5. After drying the pieces, they should be enclosed into tight plastic bags filled with nitrogen gas, radon-free quality, for storing till use. The same protocol will be applied to the cleaning of the copper pieces to be used in the encapsulation of the ultrapure ANAIS crystals. Figure 2.4: Scheme of a voltage divider circuit used for the VLB Ham PMTs, following Hamamatsu recommendations. PMTs voltage dividers were designed in Teflon boards with SMD components. The electric circuits used were adapted to each model, following the recommendations of the manufacturer (either Hamamatsu or ETL) [108]. The scheme of the voltage divider used for the Ham VLB PMT model is shown in Figure 2.4 as an example. Radioactivity screening results for the voltage divider mounted are also presented in Table 2.5 and some images in Figure 2.3.d. Ham LB, ULB and VLB PMT models present flying leads, min. 65 mm long, in order to reduce the voltage divider contribution to the background. 2.1.4 Light guides Two cylindrical light guides made of methacrylate (3” diameter and 10 cm length) were optionally used with the ANAIS-0 module, between the quartz window and the PMT, in order to evaluate, and eventually reduce, the contribution of the PMTs to the background. They were encapsulated individually in copper. Figure 2.5 shows a light guide and its encapsulation. 43 2. ANAIS-0 experimental procedure The light guide encapsulation consists of an electroformed copper cylinder and two commercial ETP copper clamps fixed under pressure on every side of the piece. Once the clamps were fixed on the cylinder, their joint was reinforced by depositing electroformed copper onto the juncture. All the electroformation work has been done at the University of Zaragoza. The radiopurity of the light guide together with its encapsulation was measured in the HPGe spectrometer, see results in Table 2.5. Only upper limits on the different radioactive isotopes searched for were derived. (a) (b) (c) Figure 2.5: Methacrylate cylindrical light guide 3” diameter and 10 cm long (a), copper casing for the light guide (b), view of the encapsulated light guide to be coupled in ANAIS-0 module (c). 2.2 ANAIS-0 Setups ANAIS-0 module was brought underground the 21/04/09 and up to the 27/11/12 it has been taking data continuously at the LSC. During this time it has been operated in various setups. As explained before, different photomultipliers coupled to ANAIS-0 module with or without light guides have been tested in order to choose the optimum configuration in terms of threshold and background. Table 2.6 summarizes those setups. A special set-up to measure the 40Kcontent of the crystal has been also taken data in that period, intersperse with the setups shown in Table 2.6, the results will be presented in section 4.1.2. The ANAIS-0 module measured in coincidence with the previously studied PIII, see section 1.4.2. This set-up had three phases because the PMTs were changed twice. Phases I and II took place between set-up 1 and 2 at the old LSC facilities, with the aim of determining the 40Kcontent of ANAIS-0, see section 4.1.2. Phase III took place at the Hall B of the new LSC, after the set-up 5, and the main goal was to measure 44 2.2. ANAIS-0 Setups Set−up PMTs Light guides Installation date Live time (days) 1 ETL No 21/04/09 31.32 2 LB Yes 22/01/10 178.9 3 ULB Yes 03/05/11 43.40 4 ULB No 21/06/11 126.4 5 VLB No 22/03/12 31.90 Table 2.6: Main features of the different setups of the ANAIS-0 module: PMT model used, whether light guides are or not coupled, beginning of the installation date and live time of measurement available. (a) (b) (c) Figure 2.6: Pictures of the ANAIS-0 module in the different setups shown in Table 2.6: (a) set-up 1, (b) set-up 3 (set-up 2 had the same external appearance), (c) set-up 4 (set-up 5 had the same external appearance). the long NaI scintillation constants also for the PIII, see section 3.6. Table 2.7 summarizes those 40K-coincidence setups. Some upgrades of the ANAIS-0 experimental layout have been implemented during this time: •During set-up 2, the 26/04/10, active vetoes anti-muons (see section 2.4) were installed. 45 2. ANAIS-0 experimental procedure Phase ANAIS-0 PMTs PIII PMTs Installation date Live time (days) I ETL ETL and LG 15/07/09 51.20 II LB ETL and LG 08/10/09 57.82 III VLB ULB 01/08/12 39.30 Table 2.7: The dedicated set-up to measure the 40Kactivity of the ANAIS-0 crystal had three phases. Light guides (LG) were used only with the PIII during phases I and II. Installation date, PMTs used at each detector module, and live time of each phase are shown. •The 18/02/11, also during the set-up 2 measurement, the ANAIS experiment was moved to the new installations of the LSC (see also next section). Live time of set-up 2 indicated in Table 2.6 refers only to measurements after the vetoes installation and before the moving to the new LSC facilities. •The 14/12/11, while measuring with set-up 4, the ANAIS electronic chain was upgraded (see section 2.6). 2.3 ANAIS-0 Shielding The ANAIS-0 module was placed underground to be shielded from cosmic radiation. Additional active and passive shielding was designed to reduce the environmental radioactivity contribution to the background. The Figure 2.7 is an artistic view of the ANAIS-0 module experimental layout at LSC, consisting of active and passive shielding. During the measurement time some improvements were done, as it has been summarized in the previous section. The ANAIS-0 shielding consisted of the following layers (from inside out): •The most inner layer of the shielding consists of 10 cm of ancient lead to attenuate the γbackground, see Figure 2.8. As explained in [109], lead is an excellent shielding material due to its high Z, price, mechanical properties, and low activation cross section for environmental neutrons. However, even for the highest purity lead, thepresenceof210Pb implies an important contribution to the background. The advantage of ancient lead, as the one found in wrecks of Roman ships sunk in the Mediterranean sea, is that 210Pb is expected to be absent because, as the lifetime is 22.3 years, it has completely decayed and the overburden of water has prevented the lead to be activated by cosmic ray neutrons. •Then, a second layer of low activity lead, 20 cm thick, see also Figure 2.8. 46 2.3. ANAIS-0 Shielding Figure 2.7: Schematic drawing of the ANAIS-0 experimental layout. It can be seen the NaI(Tl) crystal, the copper encapsulation, light guides, PMTs, ancient lead, low activity lead, a polyethylene base, and three vetoes anti-muons. (a) (b) Figure 2.8: The ANAIS-0 lead shielding over the polyethylene base. In (a) the ANAIS-0 module over the lead and in (b) the complete lead shielding are shown. •An anti-Rn box allows closing the shielding to prevent the entrance of airborne radon inside the experimental space (by continuously flushing with radon-free nitrogen gas which keeps an slight overpressure). First, an anti-Rn bag was used, upgraded for 47 2. ANAIS-0 experimental procedure the set-up 5 to a box made of aluminum and PVC and expandable to house the whole ANAIS shielding. Figure 2.9 shows the two anti-Rn systems used. (a) (b) Figure 2.9: ANAIS-0 experimental layout in two different configurations: (a) set-up 2, at old LSC facilities using an anti-Rn plastic bag to close the shielding; (b) set-up 5 at the new LSC facilities using an anti-Rn box. In both setups, active vetoes anti-muons can be seen on top of the shielding. •The lead is placed over a polyethylene base (see Figure 2.9), which is part of the neutron shielding that will be used for the whole ANAIS experiment, but that was not completed during ANAIS-0 measurements. •3 plastic scintillators (0.5×1×0.05 m3) were installed, during the set-up 2 measurement, to monitor the muon rate in the laboratory at the shielding position, 48 2.5. Signal processing (low, high and very high energy) by a resistors set, conveniently adjusted to the desired energy ranges. •Charge Digital Converter (QDC). It is a device that converts the input charge signal into a channel number by integration in the chosen gate (1 μs in our configuration), keeping the proportionality with the charge in the dynamic range. A fixed value of charge (pedestal) is always injected by the module itself to the signal, in order to guarantee the proper working of the QDC. The channel range goes from 0 to 2048 for the old QDCs (Canberra ADC 8713) and from 0 to 4096 for the new QDCs (CAEN V792). The CAEN QDCs require a decoupling board (A992) in the input signal, which eliminates the DC level of the signal. •Time to Digital Converter (TDC). The information of the arrival time of each signal, is saved by the TDC modules (CAMAC 2229 and CAEN V775) to study the chance coincidences. The time lapsed from the beginning of the gate up to the arrival of the signal is saved as a digital number. •Digitizer. In order to obtain all the information from the shape of the pulse, each event is digitized. First, a scope from Tektronix series TDS5034 was used. It presents a bandwidth of 350 MHz, a sampling rate up to 5 GS/s with 8 bits of resolution, and records from 500 to 2 ·106points of temporal resolution. A choice of 0.8 ns/point sampling was done for the ANAIS-0 data taking. With the upgrading of the electronics, a MATACQ CAEN V1729 digitizer card was installed. It presents a bandwidth of 300 MHz, a sampling rate of 2 GS/s, 2520 points of temporal resolution and a dynamic range of 1 V (in this case, it is fixed) with 12 bits of resolution. •Time Counters. For the ANAIS experiment a precise determination of the live time is required. Dedicated counters (CAEN V560) have been installed with that purpose. Independent time counters are used for the registering of the time after a veto signal in the plastic scintillators. •FAN-IN/FAN-OUT. Analogic (CAEN N625) and Logic (CAEN N454) FANIN/FAN-OUT modules are used to obtain up to four outputs from the same input. They are also used to add the three vetoes signals first, and then to invert the subsequent preamplifier output before entering the CFD (see Figures 2.11 and 2.12). •NIM-ECL converter. This converter (CAEN N638) is needed to adapt the TDC and MALU input signals. •Delays. Analogical delays are needed to retard the QDCs input signals till the trigger is done and the QDC integration windows are open. This analogical delays 55 2. ANAIS-0 experimental procedure are implemented by using long cables and adjusting conveniently their length. The scope inputs also needed to be delayed, unlike the MATACQ signals. Digital delays are required for the TDC and PU inputs. A module has been specially designed for the ANAIS electronic chain with such a purpose [108]. 2.5.2 Acquisition software The acquisition program used in ANAIS-0 is easily scalable to a large number of detectors and it can be configured, i.e. some parameters can be changed without compiling the program. Data are saved in binary files (*.raw), organized in blocks. This program was described in [89,123]. The information saved for each event is: •General information: live time counter value, real time counter value, time after last veto event, mask informing about the trigger (useful if more than one detector is taking data). •Specific information for each PMT output signal (two signals for each detector triggering): three QDC values, one TDC value, and the digitized pulse. In parallel to the upgrading of the electronic chain, a completely new environment was developed in C++. It is fully configurable: with a single configuration file many different hardware configurations can be controlled. Sampling rate, number of detectors and signals, file size and pretrigger are some of the parameters that can be configured, but also the graphical interface can be chosen and some conditions for the acquisition can be imposed, as for example, in calibration mode digitized pulses are not saved to reduce the measurement dead time, or an energy window can be selected in order to only store pulses in that region of interest. All the information is directly saved into root files (see section 2.5.3) for off-line analysis. Drivers for the control of many VME modules have been specially developed, tested and optimized. The software was tested to avoid bottlenecks and stability problems. Dead time was estimated with pulser generated signals and measured with counters, resulting a value of 2 ms/event, mainly dominated by the digitization conversion time, improving the 4 ms/event obtained with the old acquisition [89]. 56 2.5. Signal processing 2.5.3 Analysis software The data analysis is done with ROOT [124]. It is a framework for data processing, born at CERN, to be used in the field of high energy physics. However, the versatility and functionality of the package has triggered a broad spread of its use, being nowadays the reference analysis code in different particle physics applications. It can be used to save, access and process data as well as to show results, either in interactive mode or by developing specially designed scripts. In the old version of ANAIS acquisition software, raw data were produced that had to be converted into root format, as first analysis step. The upgraded acquisition saves the data directly in root format, although an off-line analysis is still required in order to get the relevant pulse information without increasing the acquisition dead time. There are some configurable parameters in this off-line analysis procedure that depend on the particular acquisition configuration: analysis threshold to define the start of the pulse, pretrigger region where baseline is calculated, threshold for the peak search algorithm (and also peak width) related to the SER of the PMTs used, etc. Some of the more relevant parameters resulting from the analysis for each pulse are: •Baseline and root mean square. The mean value and standard deviation of the baseline noise are obtained from the first points (pretrigger region) of the pulse. •T0. The pulse onset is assigned to the point the point where the pulse overtakes the analysis threshold. •Pulse areas. The pulse area in a chosen pulse region is calculated as the addition of the points in that region, subtracting the corresponding DC contribution. Different parts of the pulse can be of interest and, hence, different pulse areas are calculated in the off-line analysis: total pulse area (from T0 to the end of the pulse), SER area (50 ns around the last peak in the pulse) or some partial areas, that are used to build pulse timing parameters, as for instance P1s (see section 3.1). •Maximum and minimum of the pulse. The minimum (and maximum) of the pulse are calculated and its positions in the pulse are also saved. Subtracting the DC level to the minimum of the pulse, the amplitude of the pulse is calculated. It is worth noting that the pulse is a negative voltage signal. •Number of peaks in the pulse. An algorithm that determines the number of peaks in the pulse has been developed. It is based on the TSpectrum ROOT class 57 2. ANAIS-0 experimental procedure and the Search method [125]. Peaks are considered gaussian with a minimum height and width, which are selected specifically for every data set. Information about the position of the peaks in the pulse is also saved. In section 3.1 this variable will be studied. This variable makes sense only for very low energy pulses where single photoelectrons can be distinguished. Figure 2.13 shows an example of a digitized pulse corresponding to a low energy event, in which some of the parameters derived from the off-line analysis are remarked: baseline (horizontal red line), T0 (green line), minimum position (blue line) and the peaks identified (red triangles). Figure 2.13: Example of digitized pulse with the MATACQ board and some of the parameters derived from the analysis: baseline (horizontal red line), T0 (green line), minimum position (blue line) and the peaks identified (red triangles). 2.6 Slow-control To monitor the stability of all the environmental and experimental parameters is mandatory for an experiment aiming at being sensitive to the annually modulated dark matter signal. Tuning of the probes and test measurements with some of them have been carried out since the beginning of ANAIS-0 operation. Environmental operation conditions at the old LSC facilities were not optimal, but they improved noticeably after moving at the new facilities. However, also some problems were detected, as it will be commented later. Continuous monitoring of different environmental parameters has being carried out since the moving of ANAIS-0 to the new LSC facilities (March 2011): radon air content, pressure, humidity, temperature in the ANAIS electronics rack, temperature in ANAIS experimental room, temperature in Hall B, evaporated nitrogen flux coming into ANAIS shielding, PMT High Voltage Supply values, intensity of the HV supply source and total trigger rate of the ANAIS-0 module. 58 2.6. Slow-control (a) (b) (c) (d) (e) Figure 2.14: Temperature probes: at ANAIS hut (a), at Hall B (b), and at the electronics rack (c). AlphaGUARD meter placed in Hall B (d). Nitrogen flux meter before the entrance of the gas pipeline into the ANAIS hut (e). Radon air content, pressure and relative humidity are measured with an AlphaGUARD system [126] every 10 minutes. The temperatures, controlled with PT100 probes, and the nitrogen flux that enters into the shielding are monitored every 3 minutes with a specially designed program [123]. Figure 2.14 shows the AlphaGUARD meter, the temperature probes and the flux meter in the new LSC facilities. During May of 2011, the radon activity was simultaneously measured in Hall B and inside the ANAIS hut. Average activities of 88.0±0.5Bq/m3and 86.5±0.5Bq/m3 were measured in Hall B and inside the ANAIS hut, respectively. Hence, similar radon activity is expected in both locations. The radon content inside the ANAIS shielding, in overpressure and continuously flushed with radon-free nitrogen gas, should be much lower to that of Hall B (upper limits are estimated in section 4.2.3, but they have not been directly measured). After 19 months of measurements, radon air content in Hall B of LSC averaged per day and month is shown in Figure 2.15, also pressure and relative humidity measurements are shown in Figure 2.16. Corresponding mean values and their standard deviations are reported in Table 2.8. A seasonal fluctuation can be clearly identified in the radon and humidity values along the measurement period and they have been fitted to a sinusoidal 59 2. ANAIS-0 experimental procedure function, letting as free parameters the amplitude of the fluctuation, the mean value, the period and the phase: y=y0+A·sin 2π T(x−xc)(2.1) The values obtained in the fit are shown in Table 2.9. They are compatible with an annual modulation, having the maximum around the 25th July (with an error of more than ten days). Both modulations are clearly correlated, although the amplitude is higher for humidity (30.9%) than for radon (15.5%). Nevertheless, more data are needed to confirm such a modulation and to identify its origin. (a) (b) Figure 2.15: Rn air content in Hall B from May 2011 to November 2012. Data are averaged on days (a) and months (b). Result of the data fit to a sinusoidal function is shown in blue, and the parameters of the fitted function are shown in Table 2.9. Parameter Mean value σ Radon 80.9±0.3Bq/m329.9±0.4Bq/m3 Pressure 882.84 ±0.02 mbar 5.41 ±0.02 mbar Relative humidity 44.8±0.7% 9.9±0.9% Table 2.8: Mean values of the radon air content, pressure and relative humidity monitored during the ANAIS-0 operation (from May 2011 to December 2012) at the new LSC facilities, measured with an AlphaGUARD system. However, the radon activity in Hall B (or inside ANAIS hut) does not affect directly to the ANAIS background, because the radon content inside the ANAIS shielding, in overpressure and continuously flushed with radon-free nitrogen gas, should be much lower to that of Hall B (upper limits are estimated in section section 4.2.3). Therefore, even in the case of confirmation of such an annual modulation effect in the radon air content, 60 2.6. Slow-control (a) (b) (c) (d) Figure 2.16: Pressure (a,b) and humidity (c,d) measurements obtained with an AlphaGUARD system in Hall B. Data are averaged on days (a,c) and months (b,d). Result of the relative humidity data fit to a sinusoidal function is shown in blue, and the parameters of the fitted function are shown in Table 2.9. Parameter Radon Relative humidity y077.70 ±1.34 Bq/m341.19 ±0.68% A12.07 ±1.89 Bq/m312.75 ±0.96% T379 ±20 days 392 ±11 days xc22nd April ±20 days 19th April ±10 days Maximum 26th July ±21 days 23th July ±11 days Table 2.9: Parameters obtained from the sinusoidal fit of the radon and humidity measurements. neither a relevant contribution to the ANAIS background, nor a modification in the ANAIS prospects are expected. The temperature is monitored in different locations around ANAIS experimental space: in ANAIS electronics, inside ANAIS hut and in Hall B, using PT100 probes. With the AlphaGUARD the temperature of Hall B is also measured, but with lower precision. In Figure 2.17 the behaviour of the monitored temperatures is shown. In particular, it has 61 2. ANAIS-0 experimental procedure to be remarked the presence of sudden changes in the Hall B temperature, some of them corresponding to changes in LSC thermostat configuration: from 17 oCto20oCthe3 rd of August and from 20 oCto19oCthe15 th of September, for instance. Modifications on the ventilation system operation conditions are responsible of other changes. The temperature at the ANAIS electronics decreased significantly after the installation of the new VME electronics (December 2011), thanks to the better ventilation system of the modules. Good agreement between both measuring systems, PT100 probes and AlphaGUARD, can be observed (see Figures 2.17.a and b) (a) (b) (c) (d) Figure 2.17: Measurements of the temperature in three different locations at LSC for the whole monitored period; data averaged in days are shown in (a) and averaged in months in (b). Variation of temperature in Hall B along one day, arbitrarily chosen, without averaging is shown in (c). It can be clearly seen the period of the thermostat regulation. Correlation between temperature at ANAIS electronics and trigger rate is shown in (d). In Figure 2.17.c a 24 h zoom of the temperature variations along the 11/08/11 is shown. It can be observed that the thermostat seems to work pretty well and can correct changes of 0.1 oC, in about one hour. However, there is only one thermostat for all the laboratory and is placed in Hall A, therefore, local temperature variations related, for instance, to the human presence in Hall B can not be corrected. 62 2.7. Detector calibrations Correlation of ANAIS-0 trigger rate with temperature at the electronics rack was observed in some periods of ANAIS-0 data taking, disturbing considerably the acquisition. An example of that correlation is shown in Figure 2.17.d. In this case, the origin of the effect was found in the QDC module DC input level and was avoided by the use of an specific decoupling board (see section 2.5). Similar effects were also observed in the Dual OR module. In this case, the module needed to be repaired by CAEN (two diodes and one integrated circuit malfunctioned). The nitrogen gas flux that enters into the ANAIS-0 shielding has also been monitored. At the moment, the LSC furnishes radon-free nitrogen gas (evaporated from liquid) to all the experiments, but previously ANAIS had its own nitrogen supply system. An average of 0.830 ±0.011 l/min can be reported. The PMT High Voltage Supply values and intensity of the HV supply source were monitored during set-up 2 at the old LSC and during set-up 4 at the new facilities. Fluctuations of only 0.25 V (<0.02%) have been observed. 2.7 Detector calibrations Radioactive sources have been used to calibrate the ANAIS-0 module, to check the gain stability, to get reference NaI(Tl) scintillation events for the data analysis, to search for systematic effects in the simulations, etc. An aluminized Mylar window was included in the detector design to allow calibration at very low energy (below 20 keV) with external gamma sources, it can be seen in Figure 2.18.a. The radioactive source is introduced into the shielding and placed just in front of the Mylar window with the help of a Teflon tray, see Figure 2.18.b. The radioactive sources used for high and low energy calibrations are listed in Table 2.10, together with their main emission lines and intensities. Typical calibration spectra for most of them are shown in Figure 2.19. Stability of the gain (through periodic calibration at high and low energy) along the ANAIS-0 data taking has been studied. A 57Co source has been chosen for weekly calibrations of the prototype because its low energy X and gamma lines (at 6.4 keV, 14.4 keV and 122.1 keV) reach the crystal through the calibration window built in the copper encapsulation. The deviation in % from the average channel value for each of these lines is shown in Figure 2.20.a. A 22Na source has been also used for calibration and the high energy stability is shown in Figure 2.20.b for the 511 keV and 1275 keV lines. 63 2. ANAIS-0 experimental procedure (a) (b) Figure 2.18: (a) Aluminized Mylar window in ANAIS-0 copper encapsulation and (b) Teflon tray to introduce the radioactive sources into the shielding and to position them precisely in front of the Mylar window. Deviations are small in all cases, but more stability will be required for the ANAIS experiment. It is worth noting that during the ANAIS-0 operation it was required to switch off and on the PMTs HV supply several times. That probably had an effect on gain stability. In data from setups 2 and 4, a gain drift can be observed. Its origin has not been established, but could be due to a malfunction of the old QDC modules or the HV power supply. One of the most important issues for ANAIS is to guarantee a good energy calibration at the lowest energies, down to the threshold. In Figure 2.21.a a linear fit (energy vs QDC channel) is shown; all the available energy lines below 150 keV have been used (from 57Co,55Fe,133Ba and 109Cd)aswellasthe3.2keVlinefrom40K, the origin of this line is explained is section 4.1.2. Residuals from the fit are also shown in Figure 2.21.a. In Figure 2.21.b it is shown a linear fit with all the lines used in 2.21.a except the 5.9 and 6.4 keV ones. It can be noticed a reasonable linearity down to 14.4 keV, but residuals at 6.4 and 5.9 keV are much larger (and both positive), specially considering them in percent, than those obtained for the other lines, including the 3.2 keV one. Although non-linearity effects had been previously reported in NaI(Tl) detectors, they cannot explain such a large effect as that seen in Figure 2.21.b. (5.9 keV line appears with an effective energy of 3.5 keV and 6.4 keV line with 4.0 keV), that however, could be explained by superficial effects either in the light yield or in the light collection efficiency (or both), that would not contribute for bulk energy depositions as that from 40K. Non-proportionality in the scintillation yield with respect to the deposited energy has been observed in inorganic scintillators [128, 129]. In particular, in NaI(Tl) scintillators some non-linear effects are expected at the K-shell Iodine binding energy (33.2 keV) because photoelectrons emitted following a K-shell photo-absorption do have less kinetic energy and ionize less. This fact has been observed experimentally [129–133] but it is 64 2.8. Muon coincident events in ANAIS-0 Figure 2.24: Rate of events in ANAIS-0 crystal in coincidence with muons in the plastic vetoes at the old and new LSC facilities in different setups. In Figures 2.25 and 2.26, the energy distribution of the events coincident with a muon interaction in the plastic scintillator vetoes besides the total background of the ANAIS-0 module can be seen for setups 3 and 4. Because of the partial coverage of the plastic vetoes used, about 30% of the events above the alpha region in the NaI crystal data (and, hence, attributable mostly to muon interactions in the crystal) were not coincident with the vetoes. A relatively high rate of coincident events is observed at the lowest energies, especially when using light guides, as commented before. The rate from 1 to 100 keV shows an 86% increase when light guides are present (comparing setups 3 and 4), pointing at a scintillation effect in the methacrylate. Such events are not compatible with NaI(Tl) typical scintillation, supporting again the hypothesis of scintillation in the light guides. In sections 3.3, and 3.6 this issue will be discussed again. It is worth noting that total event rate attributed to muon coincident events in the dark matter region of interest is clearly further down the rest of the background contributions (<0.1% in the 2-10 keV region. 71 2. ANAIS-0 experimental procedure Figure 2.25: Background spectrum measured with the ANAIS-0 module at very high energy, up to about 20 MeV. PMTs signals saturate (more in set-up 4 than in set-up 3), implying that the shape of the muon spectrum is modified and proportionality with energy is lost above 8 MeV. Figure 2.26: Distribution of events at low energy coincident with a muon in the active vetoes. Clear contribution from events in the light guides can be observed. 72 Chapter 3 Data analysis Low energy events are of utmost interest for the ANAIS experiment because dark matter particles are expected to produce very small energy depositions in the detector: below 100 keV of nuclear recoil energy, and depending on the candidate at much lower energies (see section 1.3.1). This implies electron equivalent energies below 30 keV for Na recoils and below 10 keV for I recoils. For this reason, a very good knowledge of the detector response function for real scintillation events in the active volume of the detector and a good understanding and characterization of other anomalous or noise event populations in that energy region is required. Efficiently filtering all the low energy events populations non attributable to dark matter interactions is one of the main issues for ANAIS. Among them, events having scintillation time constants other than NaI(Tl) one or events in coincidence with a signal in the plastic scintillator vetoes or in coincidence between at least two ANAIS modules should be rejected. Specific protocols to reject such events have been developed and applied to data from ANAIS-0 module corresponding to setups 4 and 5 (section 3.1). The low energy region has been carefully studied: old neutron calibration data from ANAIS Prototype III have been reanalyzed (section 3.2), asymmetry in the energy sharing has been investigated with ANAIS-0 and prototype III data (section 3.3), several low energy estimators have been compared to determine the one with the best resolution and performance (section 3.4) and, as trigger at the photoelectron level in the experiment is mandatory, the trigger efficiency in the different setups is discussed (section 3.5). The NaI scintillation constants have been determined for alpha, beta/gamma and neutron interactions very precisely. In particular, very slow scintillation in NaI(Tl), sparsely known previously, has been precisely measured in ANAIS-0 and PIII crystals (section 3.6). Finally, quenching factor for alpha particles vs. beta/gamma particles has been determined (section 3.7). 73 3. Data analysis 3.1 Bulk NaI scintillation low energy events selection Strong rejection of non bulk NaI scintillation events is required to reduce the effective threshold down to 2 keV. Dark matter particles are expected to interact in our detector by elastic scattering off Na and I nuclei in the NaI crystal. Hence, the energy is deposited through the interaction of the corresponding recoiling nucleus in the crystal. Dark matter events are expected to be very similar to those produced by neutrons, and to share some features with those having beta/gamma origin; both can be produced using calibration sources in order to have reference populations that allow a good NaI(Tl) bulk scintillation event characterization. However, dark matter events neither should be correlated with muon interactions in the vetoes, nor appear in coincidence between two or more modules, nor accumulate in short time periods, nor show anomalous scintillation time constants in the pulse shape, for instance. All these events populations, non attributable to dark matter interactions can be rejected before the dark matter analysis. The goal is to find a compromise between a high acceptance of bulk NaI scintillation events and low contribution of other events, not rejected by the filtering. In some of the cuts, the filtering implies an effective reduction in the acquisition live time, while in others, the efficiency of the cut to preserve the bulk scintillation events in NaI(Tl) active volume has to be estimated by studying populations of bulk scintillation events, specially at very low energies. 57Co,109Cd,and133Ba source events and the 3.2 keV events following 40K decay in the bulk (see section 4.1.2) have been used for that purpose. We describe below in detail the filtering procedure followed with ANAIS-0 data. Table 3.1 shows the number of events, as well as the available live time, before and after each cut for setups 4 and 5. All the cuts have been applied consecutively and in the order presented in the text to every set-up data. 1. Periods of anomalous high rate. Dark matter interaction rates are so low, that the probability of two events occurring in a short time period is negligible. Because of that, periods of anomalous high rate can be discarded. In this category of events are some electronic noise related events and events following high energy calibrations, see section 2.7. During the ANAIS0 operation period, some problems in the experimental conditions appeared (for instance, just after moving ANAIS-0 to the new LSC facilities or while testing the 74 3.1. Bulk NaI scintillation low energy events selection Cut Set-up 4 Set-up 5 Number of Events Live Time (s) Number of Events Live Time (s) 0 6924171 11216800 1272028 2755985 1 6642571 10918900 1272028 2755985 2 6470283 10649269 1272028 2755985 3 6326697 10648800 1259000 2755660 4 6323129 10648800 1259000 2755660 5 6322944 10648800 1173911 2755660 6 2121342 10648800 748640 2755660 Table 3.1: Total number of events and live time remaining after the different cuts described in the text. new VME electronics) and these data are not considered, and the corresponding time is discounted. 2. Periods of non-correlation between analog electronics and digitized data. Because in the analysis we combine information from the digitized pulse with QDC data, for instance, we have to reject such uncorrelated periods. They affected full blocks of data, that were discarded and their corresponding acquisition time discounted from the total live time. In set-up 4, there were 673 blocks of 256 events each rejected as uncorrelated, 97.53% of the live time remains after applying this cut, as it is shown in Table 3.1. In set-up 5, using the new acquisition, uncorrelated events are not present, which entails a significant improvement. 3. Cut on events after a muon (very high energy) event. When a very high energetic particle interacts in the ANAIS-0 module, due to the slow NaI(Tl) scintillation (see more details in section 3.6) photons in the tail of the pulse are able to trigger again the acquisition (because of the triggering at the photoelectron level) and the total acquisition rate increases (see Figure 3.1). For that reason, events triggering during 0.5 s after a high energy event (over 9 MeVee to guarantee to be well above the usual alpha and gamma backgrounds) are rejected and the corresponding live time deducted. As in set-up 5 PMTs signals saturated at energies much below 9 MeVee and then, muon events could not be tagged and discriminated from the high energy gamma and alpha background, it was decided (in a conservative way, because the muon veto coverage is not complete) to reject 0.5 s after the arrival of a muon at the plastic vetoes (see cut 4). The spectra of events rejected by this cut in setups 4 and 5 is shown in Figure 3.2. The different shape of the rejected events spectra is attributed to the different PMTs used in each 75 3. Data analysis set-up; in particular, VLB PMTs of set-up 5 present a higher quantum efficiency and can be more effective in triggering at those events. However, the different criterion used to tag the high energetic events could be blamed on it. (a) (b) (c) (d) Figure 3.1: Total acquisition rate (in blue) with ANAIS-0 module along a week (a), and in a zoom at different timescales (c) and (d). In red, very high energy events (above 9 MeVee) are marked. It can be observed the clear correlation between these events (mostly attributable to muon interactions in the NaI(Tl) crystal) and the increase in the trigger rate. In (b) the acquisition rate before (blue) and after applying cut 3 (gray) are shown. Figure 3.2: Spectrum of the events rejected by cut 3 (0.5 s after a very energetic event) for setups 4 (black) and 5 (blue). 76 3.1. Bulk NaI scintillation low energy events selection 4. Cut on events coincident with a a plastic veto signal. In section 2.8 muon related events in ANAIS-0 module coincident with a plastic veto signal were identified, both at high and low energies. Hence, these events are also rejected, without implying efficiency or live time corrections. For set-up 5 this cut is unnecessary, being done at the same time than cut 3, due to the PMT signals saturation. The spectrum of the events rejected by this cut has been shown in Figure 2.26. 5. Cut on events with anomalous baseline estimate. As explained in section 2.6, the baseline is calculated for each pulse with the first points before the beginning of the pulse. If a photon arrives in the pretrigger region, the baseline won’t be properly calculated and hence, neither other related parameters. These events are identified by their anomalous low baseline level and they are not considered for the analysis, see Figure 3.3. They can be attributed to tails of previous pulses or to dark current photons, see an example of each case in Figure 3.4. VLB PMTs present a higher dark current rate than ULB PMTs and hence, more events are rejected by this cut in set-up 5. (a) (b) Figure 3.3: Distribution of baseline parameter for PMT0 and PMT1 data corresponding to setups 4 (a) and 5 (b). Cut applied to reject events having anomalous estimated baseline is shown in red. Set-up 4 data were digitized with the scope, and set-up 5 with the MATACQ, implying different absolute values of the typical baseline values. 6. Cut on the peaks number of the event. Profiting from the good sampling rate of the digitized data, the discrete arrival of the scintillation photons to the PMT photocathode can be distinguished at low energies. An algorithm has been applied in order to count them, see section 2.5. The algorithm requires two parameters: a threshold voltage and a width for the peak. By choosing these parameters according to the S.E.R. of the PMT, peaks are properly found, 77 3. Data analysis (a) (b) Figure 3.4: Examples of events where the baseline (red) has not been properly calculated, due to the early arrival of a single photon (a) or to a previous pulse tail triggering (b). counted and their positions saved. In Figure 3.5, a low energy pulse is shown and the peaks identified by the algorithm are marked. Figure 3.5: Low energy pulse (corresponding to the population of 3.2 keV events after 40K decay, selected by coincidence, see section 4.1.2) showing peaks identified by the peak search algorithm corresponding to individual photoelectrons. They are counted independently for each PMT signal and their positions saved. The shown pulse corresponds to set-up 5 (data taken with the MATACQ). We reject events having less than 3 peaks in any of the PMTs. This implies an effective analysis threshold of about 1 keV (see section 5.2.4 where the photoelectrons/keV are calculated for each set-up). This cut allows to reject events due to chance coincidences (directly related to the dark current of the PMTs), or other events having their origin in the PMTs (possible Cerenkov light emission in the PMT glass, for instance), that are expected to produce a signal very similar to SER, except in amplitude/area. Events rejected by this cut cannot be attributed to NaI 78 3.1. Bulk NaI scintillation low energy events selection (a) (b) (c) Figure 3.6: Low energy spectra of set-up 4 (a) and 5 (b) corresponding to a 57Co calibration and low energy spectrum coincident with the 40Khigh energy gamma at low energy (showing the 3.2 keV line), corresponding to the phase III of the 40K-coincidence set-up (c), before (in black) and after (red) the application of cut 6. scintillation. The effect of this cut in data of a 57Co calibration is shown in Figure 3.6 for setups 4 and 5. This cut is only removing events below 2 keV, and 6.4 and 14.4 keV lines are not affected at all by the cut. Similarly, events from 40K corresponding to 3.2 keV, selected by the coincidence with the high energy gamma (see section 4.1.2), are not affected by the cut. They are shown before and after the application of the cut in Figure 3.6.c. On view of the results it does not seem necessary to add an efficiency to this cut above 2 keV. The low energy background spectra before and after cut 6 are shown in Figure 3.7 for setups 4 and 5. Also the spectra of the events rejected by this cut are shown. It can be seen that this cut allows to reject background events up to 20 keV equivalent energy. The spectra corresponding to the events rejected by this cut for setups 4 and 5, sharing some features, are different. As both setups use different PMT models, that supports our hypothesis that these events have a PMT origin. 79 3. Data analysis (a) (b) Figure 3.7: Low energy spectra of set-up 4 (a) and 5 (b) after cut 5 (black), after cut 6 (blue) and events rejected (red) because the number of photoelectrons is below 3 in at least one of the two PMT signals. 7. Cut on events faster than typical NaI(Tl) bulk scintillation. With the purpose of rejecting events clearly faster than typical NaI(Tl) bulk scintillation pulses, different parameters depending on the scintillation time (pulse decay constant) have been studied. P1 has been determined as the one that works better: P1 is the ratio between the pulse area from 100 to 600 ns after the pulse onset and the area from pulse onset till 600 ns after. In previous works [89], P1 was calculated for each PMT signal individually, but, in this work, in order to profit from the total number of photoelectrons at the lowest energies, a new parameter P1s has been built with the addition of both pulses, featuring much better discrimination ability. Hence, the P1s timing parameter has been defined: P1s=Area1(100 −600ns)+Area2(100 −600ns) Area1(0 −600ns)+Area2(0 −600ns)(3.1) This parameter is expected to be around 0.7 for the NaI(Tl) scintillation events, even though it shows an slight dependence on the energy (see Figure 3.8 where this parameter is shown for 57Co calibration data up to 100 keV.) In order to apply an additional cut based on this parameter, data from 57Co and 109Cd calibrations (set-up 4) have been used: mean values and standard deviation of P1s in different energy windows (1 keV width) have been calculated by fitting to a gaussian function the P1s parameter distribution (see Figure 3.9 and Table 3.2). During set-up 5 no 109Cd calibration was available, but data from 57Co and 133Ba were used instead. In principle, the values of this parameter should be characteristic of the NaI(Tl) scintillation and do not depend on the special set-up features. However, small differences have been observed in the P1s values corresponding to 80 3.3. Asymmetric events Figure 3.15: P1s parameter distribution in 1 keV regions (blue) and gaussian fits (red) for 252Cf neutron calibration data taken with prototype III. Although threshold effects are present below 10 keV, good scintillation events are found at those energies and we have decided to show also for them the P1s distribution. calibration data and some pulse tails and single photoelectron events are present with P1s ≈0, but they only amount a 0.2% of the total number of events, and they do not affect to conclusions derived below. For 109Cd and 57Co calibration data, cuts 1 to 6 have been applied. From the neutron calibration data analysis it can be concluded that an additional efficiency factor should be taken into account for the selection of possible dark matter events. Figure 3.16 shows clearly that nuclear recoil events would be rejected by the cut on the P1s parameter with an efficiency different than that estimated from beta/gamma calibrations. The cut value chosen with beta/gamma events (lower limit at 2 σconfidence level) is equivalent to a lower limit at 1.59 ±0.37 σfor neutron events, if data data from 5 to 30 keV are averaged. Hence, in the 5 to 30 keV energy window we should replace our acceptance efficiency for the selection of events using cut 7 (see section 3.1) by 0.944 instead of 0.977 in order to select conveniently nuclear recoils in the dark matter analysis. 87 3. Data analysis Energy region 252Cf calibration Mean σ 0 - 1 keV 0.469 ±0.018 0.213 ±0.027 1 - 2 keV 0.480 ±0.017 0.208 ±0.029 2 - 3 keV 0.429 ±0.008 0.232 ±0.017 3 - 4 keV 0.449 ±0.005 0.170 ±0.008 4 - 5 keV 0.457 ±0.004 0.128 ±0.004 5 - 6 keV 0.469 ±0.003 0.120 ±0.003 6 - 7 keV 0.505 ±0.002 0.104 ±0.002 7 - 8 keV 0.520 ±0.002 0.093 ±0.002 8 - 9 keV 0.545 ±0.002 0.087 ±0.001 9 - 10 keV 0.558 ±0.002 0.085 ±0.001 10 - 11 keV 0.572 ±0.002 0.080 ±0.001 11 - 12 keV 0.579 ±0.002 0.077 ±0.001 12 - 13 keV 0.587 ±0.002 0.076 ±0.001 13 - 14 keV 0.598 ±0.002 0.073 ±0.001 14 - 15 keV 0.605 ±0.002 0.072 ±0.001 15 - 16 keV 0.606 ±0.001 0.068 ±0.001 Table 3.3: Parameters obtained from the gaussian fit of the P1s parameter distribution corresponding to 252Cf calibration data taken with prototype III. (a) (b) Figure 3.16: (a) P1s parameter versus energy for 252Cf calibration data taken with prototype III. The upper red line represents the P1s mean value obtained in 1 keV width windows and the lower red line the cut value chosen to define the acceptance region (the mean minus 2σ). (b) 252Cf calibration (black), 109Cd and 57Co calibrations (blue), P1s mean value and the cut value chosen to define the acceptance region (the mean minus 2 σ), for the 252Cf (red) and 109Cd and 57Co (green) calibrations. No filtering has been applied to the 252Cf calibration data. Cuts 1 to 6 have been only applied to 109Cd and 57Co calibration data. 88 3.3. Asymmetric events 3.3 Asymmetric events Asymmetric events have been observed in background and calibration data from ANAIS-0 by looking at the different sharing of light between the two PMT channels, see Figure 3.17. Figure 3.17: Energy sharing between the two PMT signals at low (a) and high (b) energy in background data taken with the ANAIS-0 module set-up 4. Similar data have been obtained with the other ANAIS-0 setups. Two different kinds of asymmetric events have been identified in the data and are further studied: 1. Bulk scintillation events are observed to produce different sharing of the light collected by the two PMT channels depending on the interaction position along the crystal axis. This effect can be caused by the spatial dependence of the light collection efficiency for every PMT channel and has been characterized with radioactive sources placed in different positions along the crystal axis in ANAIS-0 and in the PIII, observing important differences, as it will be reported later in this section. 2. Very strongly asymmetric events are identified in both channels at low energy (see Figure 3.17.a). These events are not compatible with bulk NaI scintillation events. They show a much larger pulse area in one PMT channel than in the other, hinting at chance coincidences or at light emission at or very near one of the PMTs, hindering a significant amount of light reaches the other PMT. In order to understand the asymmetric bulk scintillation events, calibration sources have been used, placing them in different positions along the crystal axis (see Figure 3.18). Figure 3.19 shows the QDC values obtained for each PMT signal and and the corresponding average values when calibrating with a 137Cs source in different positions along the crystal axis: near the PMT 0 (position 1), near the PMT 1 (position 7) and centered (position 4). By looking at the average of the two PMT signals, spatial dependence is strongly 89 3. Data analysis reduced but does not disappear and we have characterized it by studying a parameter A, measuring the asymmetry of the light sharing: A=Q0−Q1 Q0+Q1 (3.2) Figure 3.18: Calibration source positions along the axis of ANAIS-0 module used to characterize the asymmetric bulk scintillation events. Figure 3.19: Spectra obtained from each PMT signal and the corresponding average for ANAIS-0 (top) and PIII (bottom) modules when calibrating with a 137Cs radioactive source at different positions along the crystal axis: centered (position 4), near PMT 0 (position 1) and near PMT 1 (position 7). 90 3.3. Asymmetric events being Q0and Q1the mean channel associated to a determined gamma line as seen by PMT0 and PMT1, respectively. This parameter has been studied with the 662 keV gamma line from a 137Cs source, which was moved along the crystal axis, both in ANAIS0 and in PIII modules. Because of the different gain in the PMT 0 and PMT 1 channels, see Figure 3.19, the parameter A is not distributed around zero, as should be expected if it had been defined in energy instead of QDC channel, and is corrected by subtracting the mean value (<A>). This A−<A>parameter is plotted as function of the source position along the axis in Figure 3.20, for ANAIS-0 and PIII modules. A very different behavior for the two detectors, probably with a geometrical origin, is observed. This effect is very small (<2%) but clearly observable in both detectors. The PMT which is nearest to the energy deposition is expected to see more light in a very plain light propagation model, however PIII data shows the opposite behavior, requiring a deeper understanding of the light collection efficiencies. The crystal shape (ANAIS-0 has a square section while PIII has an hexagonal section and both crystals have different ratio length over cross-section) could be the origin of this effect, that will be also studied with ANAIS-25 modules as soon as the background assessment is finished. Figure 3.20: Parameter A, after subtracting the average value, calculated as a function of the 137Cs radioactive source position for ANAIS-0 (black points) and PIII (red points) modules. Figure 3.21 shows the different sharing of the collected light between the two PMT signals for events from a 137Cs calibration source in the center of the crystal (position 4), near PMT 0 (position 1), and near PMT 1 (position 7). The different behavior already commented with the parameter A is also observed in this representation. As the observed asymmetry in background data is very similar to that obtained from the combination of calibration data shown in Figure 3.20, and comparing data from setups 1 and 2, the fraction of events in the lateral bands reduce from 15 to 10% when using light guides 91 3. Data analysis in the low energy region, we hypothesize that the contribution from the PMTs to the background is responsible of the high and low bands seen in Figure 3.17. Contaminations homogeneously distributed at or around the crystal are not expected to generate such asymmetric sharing of the light. (a) (b) Figure 3.21: Repartition of the energy between PMT 0 and PMT 1 in data from ANAIS-0 set-up1(a)andPIII(b)witha137Cs calibration source placed in position 4, in the center of the crystal (black); in position 1, near PMT 0 (red); and in position 7, near PMT 1 (blue). This asymmetric events population can not be rejected because they are NaI crystal scintillation events. However, it could be possible to define an active volume region, rejecting the most asymmetric events, but at the expense of losing sensitivity in the search for dark matter, by reducing the active mass. Moreover, it has to be noticed that the asymmetry reduces at the low energy regime. On the contrary, the second asymmetric events population is due to a scintillation faster than that of NaI. Light guides have been observed to contribute to these population, as can be seen in Figure 3.22, corresponding to the energy sharing between the two PMTs for a background measurement with the set-up 3 of ANAIS-0 module. It can also be observed a high asymmetry in veto-coincident events at low energy, specially in the case of set-up 3 data, which used light guides, see Figure 3.23. Tagging events as asymmetric when one PMT detects more than the double of the light signal than the other, asymmetric events rate has been quantified to be 9 cpd in set-up 3 and 3 cpd in set-up 4. In section 2.8, it was stated that set-up 3 presented 8 cpd in coincidence with active vetoes more than set-up 4, being the use of light guides the only difference between both setups. Here, it is established that the majority of these events (6 cpd) are asymmetric, as expected to correspond to interactions in the light guides. On view of these results, scintillation in light guides and quartz windows due to muon interactions 92 3.4. Energy estimators Figure 3.22: Energy sharing between PMT 0 and PMT 1 for background data from the setup 3 of ANAIS-0. A population of strongly asymmetric events, having energy equivalent up to 200 keV, is clearly identified and attributed to scintillation in the light guides, because these events were not present in data from setups 4 and 5 (see Figure 3.17). could be responsible of such asymmetric events. One quartz window could scintillate more than the other (the one near PMT 0), because in Figure 3.23.b there are more asymmetric events with few energy in PMT 1 than in the inverse band. (a) (b) Figure 3.23: Energy sharing between PMT 0 and PMT 1 for the events coincident with a muon in the plastic vetoes. In (a) data from set-up 3 (43.4 days, 2184 events), using light guides, and in (b) data without light guides, from set-up 4 (126.4 days, 5689 events), are shown. Although it had been studied, an specific cut for asymmetric events at very low energy is finally not required in the filtering of ANAIS-0 data because it adds nothing to the cut 7, previously described. All the strongly asymmetric events have been filtered by cuts 6 and 7, see Figure 3.24. 93 3. Data analysis Figure 3.24: Energy sharing for PMT 0 and PMT 1 in set-up 4 before (black) and after (red) the cuts described in section 3.1. It can be observed how the strongly asymmetric events are efficiently rejected by the cuts proposed there (in particular by cuts 4, 6 and 7). 3.4 Energy estimators Different low energy estimators have been studied: Charge converted signal (QDC), area of the digitized pulse (area), number of peaks in the pulse (n), and area of the peaks in the pulse (area-p). The aim is to reconstruct the event energy with the best accuracy down to the threshold, which implies having the best resolution for monoenergetic events. For the high energy events, QDCs values are taken as energy estimators. In this regime, n and area-p are not defined, and even total area is usually useless because of the adjustment of the digitization scale, up to a few hundreds of keV, that prevents to record full pulses above some hundreds of keV in most of the setups. •QDC: The PMT charge output signal is integrated in a microsecond window at the QDC module and converted into a digital channel. QDCs inject a small charge amount (pedestal), resulting in a non-zero value for a zero energy deposition. The pedestal should be very stable to guarantee a very low and stable threshold. •Area: It is the integral of the pulse from the pulse onset (T0) up to the end of the digitized window subtracting the baseline, see section 2.5. •n: The number of peaks in the pulse, see section 2.5, could be a low energy estimator. However, this parameter in general underestimates the energy because two photons can arrive too close in time producing only one peak. •Area-p: This parameter is constructed by integrating the pulse around every identified peak (50 ns window around the peak position), and the baseline contribution is subtracted afterwards. This estimator has been tested trying to avoid the energy underestimate associated to the previous parameter. 94 3.4. Energy estimators Usually, QDC output is the best energy estimator and the most widely used. However, due to the excellent resolution of the MATACQ digitizer, the other parameters have been found to be competitive, not being the case with the previous scope data. Figure 3.25 shows the spectra obtained with 57Co (a), 109Cd (b) sources and the 3.2 keV population selected by the coincidence measurement (c) using the different energy estimators proposed: QDC channel, area and area-p. Parameter n will be studied later on. (a) (b) (c) Figure 3.25: Comparison of spectra built with different energy estimators at low energy: QDC channel, area and area-p, for (a) 57Co calibration data, (b) 109Cd calibration data, and (c) 3.2 keV population selected by the coincidence technique. In (a) the QDC signal saturates at about 120 keV and above this energy the spectrum is distorted. The three parameters show similar results in terms of energy resolution, see data in Table 3.4. Hence, area can be used as a good energy estimator in the energy region where the pulse is digitized, making the use of QDC modules in the same dynamic range redundant. As conclusion, low energy range QDCs could may be skipped in next ANAIS phases. As expected, data shown in Table 3.4 confirm that area-p is the best energy estimator in the very low energy window (below 10 keV). However, it can be observed in Figure Figure 3.25.c that the distribution of baseline noise is clearly broader than for the other energy estimators: this is due to the fact that the peak search algorithm sometimes identifies 95 3. Data analysis noise fluctuations as peaks. In the absence of real peaks attributable to photoelectrons, area-p integrates only the positive noise fluctuations, without compensating them with the negatives, as occurs in area or QDC estimates. Energy (keV) Resolution(%) QDC Area Area-p 3.2 22.98 ±1.68 22.58 ±2.34 20.60 ±1.71 6.4 23.26 ±7.56 23.70 ±8.09 18.74 ±5.85 14.4 14.6 ±0.80 14.05 ±0.72 14.55 ±0.61 22.6 11.79 ±0.10 11.72 ±0.06 11.70 ±0.06 88.0 5.74 ±0.04 5.71 ±0.05 5.71 ±0.04 122.1 5.07 ±0.03 5.04 ±0.05 5.13 ±0.05 Table 3.4: Resolution values for the different lines studied with the energy estimators: QDC, area and area-p. Moreover, the linear dependence between these energy estimators has been checked with 57Co calibration data and the 3.2 keV events population selected by the coincidence measurement (see Figure 3.26). All the three energy estimators present a linear dependence between them. The number of peaks in the pulse has been also studied as an energy estimator at the lowest energies, where single peaks can be distinguished. However, this number is not a good energy estimator, as can be seen in Figure 3.27: 3.2 keV events population is shown in the number of peaks vs energy plot and the n distribution for the same data. Very poor energy reconstruction is achieved with this parameter, although it has proven to be very useful to reject non-bulk NaI scintillation events (as explained in section 3.1). 3.5 Trigger efficiency. As it has been already explained, lowering the threshold as much as possible is mandatory in any experiment devoted to the direct search for dark matter. The lowest achievable threshold requires to trigger at the single photoelectron level in each PMT signal. To guarantee the trigger at this level is really an important experimental issue. We have studied the trigger level in the different ANAIS-0 setups by different methods: first the distribution of amplitude of the S.E.R. is compared to the amplitude of the first peak (at the trigger position), and second the 3.2 keV events, selected by the coincidence, have been profited to study how many of them trigger effectively our acquisition. 96 3.6. Precise determination of the NaI(Tl) scintillation constants. Digitizer sampling rate N of x points vertical scale N of y points Tektronix Scope 207 25 MS/s 8 ·10650 mV 256 Tektronix Scope 208 250 MS/s 104200 mV 256 MATACQ 2 GS/s 2520 1 V 4096 Table 3.7: Summary of the configuration values for the main digitization parameters relevant in the analysis of the NaI(Tl) scintillation constants: sampling rate, number of points in the horizontal scale, dynamic range and number of points in the vertical scale. In the following, we will show events having pulse area in a region corresponding from about 2.5 to 3 MeVee (equivalent to the region from 4.2 to 5 MeV for alphas). In Figure 3.32.a, the average pulses built for alpha, gamma and muon interactions in a 1 μs temporal scale, recorded by the MATACQ digitizer, with ANAIS-0 module are shown. No difference between gamma and muon events has been found, as expected, and gammas and muons will be considered in the following analysis as the same population, as far as it is not expected different behavior in the longer timescales. We will work with average pulses for the two populations in the cases of the fast and medium timescales, see Figure 3.32. For the longest timescale, a different approach has been followed: photoelectrons have been identified individually, at a given position of the pulse, and an histogram has been produced with the corresponding temporal distribution for all the events in the R.O.I., separately for alpha and beta/gamma/muon events, for ANAIS-0 and PIII modules. The corresponding pulses can be seen in Figure 3.33, conveniently normalized to the same fast pulse area and averaged according to the number of events in the ROI for each population. In order to derive the different scintillation time constants, a fitting procedure has been followed in three steps. First, a fit has been done in the longest timescale to two exponential decays using the range from 4 to 320 ms after the onset of the pulse. Fits have been carried out for alpha and gamma/muon events, both in independent and dependent ways, for ANAIS-0 and in independent way for PIII data and are shown in Figure 3.34. According to the chi-squared value of the fits for ANAIS-0 data, the same time constants fit properly pulses from both populations, although with different relative amplitudes, being the significance of the fit limited by the poor statistics in the gamma/muon population. The results of the fits are summarized in Table 3.8, showing very similar results in ANAIS0 and PIII for the fitted time constants. The most relevant difference between alpha event pulses and gamma/muon event ones in this long timescale is the total number of photoelectrons in these slow components: gamma or muon events excite much more efficiently 103 3. Data analysis (a) (b) Figure 3.32: (a) Pulses corresponding to the interaction of alpha particles (black), gammas (blue) and muons (red) in a 1 μs window, recorded by the MATACQ digitizer. (b) Pulses corresponding to alpha particles (black) and gammas/muons (blue) in a 40 μs window, recorded by the Tektronix 208. Only data taken with ANAIS-0 are shown. All the pulses shown have been normalized to the same pulse area. (a) (b) Figure 3.33: Pulses corresponding to the interaction of alpha particles (black) and gammas and muons (blue) in a 320 ms window, registered with the Tektronix 207 for the ANAIS-0 (a) and PIII (b) crystals. Events from both populations correspond to the same pulse area, dominated by the fast scintillation components. the long-lived states contributing to this very slow light emission. The distributions of the total number of photoelectrons for events of the alpha and gamma/muon populations are shown in Figure 3.35. In ANAIS-0, there is more than a factor of two different average values (0.050 ±0.016 for alpha particles vs. 0.118 ±0.029 for gammas/muons), however in PIII this difference is not so high (0.055 ±0.011 for alpha particles vs. 0.068 ±0.011 for gammas/muons). This is also evidenced in the different relative amplitudes for the slow scintillation components (see Table 3.8): gammas and muons produce more scintillation in a factor of 2 (3.5) for the 30 ms (95 ms) components. Because of the quenching factor for alpha particles vs beta/gamma events in NaI(Tl) and the fact that we have normalized 104 3.6. Precise determination of the NaI(Tl) scintillation constants. to the pulse area and not to the energy deposited, in terms of photoelectron production per unit of energy, the referred effect is still much more important. It has to be noted that, by cause of the scope sampling and peak search algorithm required for building the photoelectron distribution, we can not rely on the absolute number of photoelectrons assigned to a given energy depositions. However, relative values between the both analyzed populations should be trustworthy. (a) (b) (c) (d) Figure 3.34: Pulses corresponding to the interaction of alpha particles, (a) and (c), and gammas and muons, (b) and (d), in a 320 ms window, registered with the Tektronix 207 for the ANAIS-0, (a) and (b), and PIII, (c) and (d) crystals are shown in black. The fit of the data shown in Table 3.8 are shown in blue. Events from both populations correspond to the same pulse area, dominated by the fast scintillation components. Second, in the medium timescale, important effects of the RC time constant of the PMT readout circuit can be observed, limiting the conclusions derived for possible additional time constants in the microseconds range. Corresponding pulses are shown in Figure 3.32.b. Clearly pulses overshoot the baseline being not recovered in the 40 μs studied range, but in the first milliseconds. This effect sweep any relevant scintillation temporal behavior. However, it can be clearly seen that it is much more important in alpha pulses than in gamma/muon ones of the same fast pulse area. We will modify in 105 3. Data analysis ANAIS-0 PIII Independent fit Dependent fit Independent fit α β/γ/μ α β/γ/μ α β/γ/μ τ1(ms) 10.7±0.322.7±2.329.0±2.034.2±4.719.8±10.33 τ2(ms) 79.0±0.392.3±1.694.4±1.581.7±4.175.77 ±2.7 A1/A20.44 0.35 0.73 0.38 0.59 0.16 Table 3.8: Results of the fit between 4 and 320 ms after the pulse onset for ANAIS-0 and PIII. Fits have been carried out for alpha and gamma/muon events, both in independent and dependent ways, for ANAIS-0 and in independent way for PIII data. (a) (b) Figure 3.35: Distribution of the number of photoelectrons per area unit identified in the longest timescale for alpha particles (black) and gammas/muons (blue) interacting in ANAIS-0 (a) and PIII (b). the future the PMT readout circuit to solve this problem, that it also slightly affect the shortest timescale pulses (see below). And finally, the shortest time scale pulses from ANAIS-0 have been fitted in the range from 0 to 1000 ns after the onset of the pulse to a combination of exponential decays. Fits have been done only in independent mode for both kind of events: from alpha and gamma/muon origin. In the case of the alpha pulse, the best fit is achieved when using only two exponential components: one for the rise and other for the decay, this fit is shown in Figure 3.36.a However, a slower component seems to be present in the pulses, that the fitting procedure is not able to take it out from the data; other possibility is an alteration of the pulse shape due to the pulse overshoot commented before. The same procedure has been applied for the fitting of the average gamma pulse, being most of the previous comments valid in that case also. The results of the fits are shown in Table 3.9 and the slow component no fitted can be seen in Figure 3.36.b. It is very probable that a 106 3.6. Precise determination of the NaI(Tl) scintillation constants. time constant in the microsecond range is present in the data, but the effect of the signal overshoot in this range prevents from deriving any estimate. (a) (b) Figure 3.36: (a) Pulses produced by the interaction of alpha particles (a) and of gammas/muons (b) in the ANAIS-0 module in a 1 μs temporal scale, registered with the MATACQ card, shown in black. The fit of the data shown in Table 3.9 are shown in blue. Events from both populations correspond to the same pulse area. α β/γ/μ τ1(ns) 17.98 ±0.12 28.23 ±0.14 A13119 ±11 2275 ±6 τ2(ns) 219.28 ±0.52 291.85 ±0.60 A22551 ±6 287.35 ±0.51 Table 3.9: Results of the fit from 0 to 1000 ns after the pulse onset for ANAIS-0 module. The fit takes every data set in an independent way. The first values correspond to an exponential growth and the second to an exponential decay. We can conclude that for alpha particles, the scintillation constant derived from our analysis is 220 ns, compatible with results from the bibliography. The rise time of the pulses is compatible with the rise time of the S.E.R. pulse, being an indication that no delay in the light emission is taking part in the scintillation mechanism. However, for the gamma pulse, some more complicated mechanisms are coming into play: an slower rise time of the pulse is found, that could be attributed to a delayed scintillation by an intermediate non-scintillating state decay, for instance; in addition, the decay time constant value of 290 ns, could be also explained by a combination of a faster constant and a longer one, but the fit is not able to converge in this scenario. This constant is somehow longer than typical values in the bibliography, but it is worth remarking that in many of the published results, the mean time of the pulse and not a decay constant is given. 107 3. Data analysis Summarizing, the fast components of the NaI(Tl) scintillation pulses for alpha and beta/gamma particle interactions have been fitted, given results compatible with those published [128, 140–144]. However, for the first time, the slow scintillation constants in NaI(Tl) are measured with this accuracy. The effect is observed in ANAIS-0 and PIII crystals, although there are differences between them, specially in the amplitude of the effect, pointing to impurities or defects in the crystalline lattice as the origin of the scintillation. 3.7 Quenching factor for αparticles. Quenching factor for the conversion into visible energy of the energy deposited by alpha particles with respect to that converted by gammas has been determined with ANAIS-0 data. Alpha particle events are selected by Pulse Shape Analysis, and the corresponding spectrum has been autocalibrated, because no external calibration is possible (see section 4.2). Hence, from the position of the alpha peaks in electron equivalent energy (using the gamma calibration at very high energy) it is possible to determine the quenching factor for alpha particle interactions with respect to gamma ones. The alpha lines are observed with electron equivalent energies shown in Table 3.10. Eα(MeV) Eee (MeVee) Eee/Eα 4.2 2.56 0.610 4.78 3.016 0.631 5.3 3.54 0.668 6 4.04 0.673 6.29 4.33 0.688 6.78 4.71 0.695 Table 3.10: Quenching factor determination for the different αlines observed with the ANAIS-0 module. Both, visible energy (in terms of electron equivalent energy) for the alpha lines observed in the ANAIS-0 module and quenching factor are plotted as a function of the alpha energy in Figure 3.37 and fitted linearly. Results of the fit are shown in equations 3.3 and 3.4. The alpha particle quenching factor derived is compatible with the one obtained by the DAMA/LIBRA experiment [97]: Eα/Eee =(0.0257 ±0.010) ·Eα(MeV)+(0.467 ±0.006) and with the predicted by Tretyak in [85]. 108 3.7. Quenching factor for αparticles. (a) (b) Figure 3.37: (a) Visible energy (in terms of electron equivalent energy) for the alpha lines observed in the ANAIS-0 module background as a function of the alpha energy. (b) Quenching factor for alpha events with respect to beta/gamma events is shown as a function of the alpha energy. Linear fits for both and corresponding residuals are shown and results of the fits are in equations 3.3 and 3.4. Eee =(0.855 ±0.014) ·Eα(MeV)−(1.050 ±0.085) (3.3) Eee/Eα=(0.032 ±0.003) ·Eα(MeV)+(0.483 ±0.017) (3.4) 109 Chapter 4 Background understanding This chapter presents the background model for the ANAIS-0 module, described in chapter 2. First, the ANAIS-0 prototype background in different set-up conditions is shown and compared (section 4.1.1). Then, we summarize the main bulk contaminants contributing to the background, that have been precisely identified and quantified by different techniques: discrimination of alpha particles vs beta/gamma background by Pulse Shape Analysis (PSA) for the uranium and thorium natural radioactive chains and coincidence technique for the 40Kbulk contamination (section 4.1.2). Cosmogenic isotopes present in the NaI crystal have been also identified and estimates are presented (section 4.1.2). Secondly, we describe the Monte Carlo simulations using Geant4 package, that have been accomplished for the different contributions and the tests carried out to validate the code (section 4.2). Then, we show the comparison between simulated and experimental data. Only a few assumptions are required in order to explain most of the measured background at high energy, supporting the goodness of the proposed model for ANAIS-0 whose background is dominated by 40Kbulk contamination. At low energy, some non-explained background components are still present and additional work is required to improve background understanding, but some plausible background sources contributing in this range have been studied and, under some hypotheses, its contribution to the background is considered in an improved background model (section 4.3). 111 4. Background understanding 4.1 Background measurements and radioactive contaminants identification ANAIS-0 module has been operated in various set-ups. Different photomultipliers coupled to ANAIS-0 module with or without light guides have been tested in order to choose the optimum configuration in terms of threshold and background, as it will be detailed in section 5.2. The background spectra measured for each configuration have been thoroughly studied in order to understand the origin of the main background contributions, specially in the low energy region. As result, it can be stated that bulk contaminations of the NaI crystal are responsible of most of the background. 4.1.1 Background measurements During the ANAIS-0 operation the energy spectra have been registered in three different energy ranges using different QDC channels: low energy (from 2 keV up to 100 keV), high energy (from 100 keV up to 2 MeV) and very high energy (from 2 MeV up to 40 MeV), as it was stated on section 2.5. In Table 2.6 main features of the different ANAIS-0 set-ups were summarized (see section 2.1.2 for more details). The main goal of ANAIS-0 was to determine the best configuration in terms of light collection and background at low energy, to be applied in the final design of the modules of the ANAIS experiment. Different types of photomultipliers were tested and their contribution to the background was studied. PMTs are complex devices that, traditionally were built with components non-specially selected in terms of radiopurity, so they used to be an important contribution to the radioactive background of any experiment using them. Recently, R+D in this field has allowed to develop ultralow background models that are changing this framework [153]. The normalized background spectra for the different ANAIS-0 setups are shown in Figure 4.1 for the high energy region and in Figure 4.2 for the low energy region. The set-up 1 data were taken just after the installation underground of the ANAIS-0 module and cosmogenic lines are observed (see section 4.1.2). It is worth remarking that these spectra correspond to raw data, and no filtering at low energies has been applied. Hence, no threshold comparison between the different set-ups can be derived from these plots. 112 4.1. Background measurements and radioactive contaminants identification Detector Phase Threshold (Channel) ANAIS-0 I 115 ANAIS-0 II 115 PIII I 125 PIII II 120 Table 4.4: Threshold (in channels) considered for the selection of 40Kevents from noise events in the low energy spectra for the different phases and detectors. highest window should only present fortuitous coincidences and could allow us to estimate their contribution in the other windows. Figure 4.8: Low energy spectra in coincidence with high energy windows in the other detector (all of them having the same width, ±σ) are shown for the different phases and detectors: centered at 1461 keV peak (μ) in black, centered 2σabove (μ+2σ)inblue, centered μ−2σbelow (μ−2σ) in red, and centered further above (μ+11σ)ingreen. Detector Phase μμ+2σμ−2σμ+11σ Events ANAIS-0 I 726 95 312 8 ANAIS-0 II 713 142 253 26 PIII I 706 146 358 22 PIII II 766 137 205 12 Table 4.5: Number of events selected by the coincidence above the threshold, shown in Table 4.4, in the different windows studied (all of them having the same width ±σ). In Table 4.5 the results are presented considering only events above the thresholds shown in Table 4.4. As expected, events at low energy coincident with the μ+11σwindow correspond to chance coincidences. The μ+2σwindow presents results compatible with the expected 16% of the total number of coincident events, but slightly underestimated. Events found in coincidence with the μ−2σwindow are much more than expected for 119 4. Background understanding a pure gaussian peak contribution, but this is probably due to the presence of compton events with partial energy deposition from the 40Kgamma line. The setup has been simulated with Geant4, version geant4.9.1.p02 [155], in order to evaluate the probability that, after a 40Kdisintegration in one crystal, the 1461 keV photon escapes and releases the full energy in the other detector. 500000 photons of 1460.8 keV have been simulated assuming homogeneous distribution of the contaminant in the bulk in ANAIS-0 and PIII crystals. The absolute branching ratio for the 40KK-shell EC followed by the emission of the 1460.8 keV photon is 0.080325729, as given by Geant4 [155]. The number of events with the full gamma energy absorbed in PIII or ANAIS-0 crystals when emitted in ANAIS-0 and PIII, are 8258 and 6998, respectively. Thus, the efficiencies for the respective coincidences are 1.33 ·10−3and 1.13 ·10−3. The 40Kactivity of every crystal is estimated with the area of the 3.2 keV peak (Area), the total available live time (LT), the crystal mass (m), efficiency of the coincidence (eff) and the fraction of the total number of events selected by the coincidence window (F). Activity(Bq/kg)= Area(counts) LT(s)·m(kg)·eff ·F(4.1) The area of the 3.2 keV peak (Area) is obtained by fitting to a gaussian the events above the threshold. The fits are shown in Figure 4.9. The 40Kactivity is calculated for each phase individually and using all the available data for the three different coincidence window widths (1σ,2σand 3σ) around 1461 keV energy. Results for each phase and detector are shown in Table 4.6. Detector Phase 40KActivity (mBq/kg) 1σ2σ3σ ANAIS-0 I14.3±0.815.1±0.917.2±1.1 II 11.1±0.512.4±0.513.4±0.6 IandII 12.7±0.513.6±0.515.2±0.6 PIII I13.5±0.916.8±1.13 20.1±1.4 II 13.9±0.916.1±1.119.0±1.4 IandII 13.7±0.616.4±0.719.5±1.0 Table 4.6: 40Kactivity calculated for ANAIS-0 and PIII using different width coincidence windows. Combined values derived for each detector are also shown. 120 4.1. Background measurements and radioactive contaminants identification (a) (b) (c) Figure 4.9: Low energy coincident events for the 1σ(a), 2σ(b) and 3σ(c) coincidence windows, and gaussian fits of the events above the threshold. Results derived from the different windows and phases analyzed are mostly compatible. As expected, larger windows have a larger contribution from chance coincidences and Compton events. Hence, results of the 1 σwindow have been taken in the following as the most reliable. Phase II presented a non gaussian shape for the 1460 keV gamma line of PIII which might have been caused by gain instabilities. However, phase I showed a higher discrepancy between the results determined with the different sigma windows and in Figure 4.9 it can be seen that the 3.2 keV peak is wider in phase I and more contribution from chance coincidences is expected. In PIII both phases were similar. Then, the average 121 4. Background understanding of phase I and II results for every detector has been taken as final result of our analysis, being shown in Table 4.7. Detector 40KActivity (mBq/kg) ANAIS-0 12.7±0.5 PIII 13.7±0.6 Table 4.7: 40Kactivity estimated for ANAIS-0 and PIII crystals by the coincidence technique. We checked that these results were compatible with the intensity of the 1461-1464 keV gamma line seen at ANAIS-0 background. There are contributions to this line coming from K-shell EC decay of 40K, producing 1464 keV (1460.8 keV + 3.2 keV = 1464.0 keV) total energy release, but also from L and M-shell EC decays, with energy depositions that can not be distinguished from the 1461 keV line. For this comparison we chose data from ANAIS-0 set-up 4 to minimize contributions from external components contaminated in 40Kto the photopeak and fitted it to a gaussian, comparing its area with the prediction of the previously introduced simulation. For 500000 isotropic 1460.8 keV photons simulated, the ANAIS-0 crystal detects 135294 photons in the photopeak (27.1%). Taking into account that only in 10.55% of the 40Kdecays a high energy gamma is emitted, if the ANAIS-0 set-up 4 result for this 1461-1464 keV peak is 32.38 ±0.32 cpd/kg,theactivity of 40Kderived is 12.32 ±0.18 mBq/kg assuming that only 40Kin the crystal bulk is contributing to this peak. This result is compatible with the activity derived from the coincidence measurement and implies that the ANAIS-0 background is dominated by 40K in the bulk. Figure 4.10: High energy spectrum for the ANAIS-0 set-up 4 (not a coincidence set-up) used to derive the 40Kbulk content from the intensity of the background line at 14611464 keV, assuming negligible contributions from external 40Ksources. 122 4.1. Background measurements and radioactive contaminants identification At last, the temporal distribution of 40Kevents at low energy selected by the coincidence above the threshold is shown in Figure 4.11. Average values of 9.32±3.63 counts/day for ANAIS-0 and of 8.46±3.11 counts/day for PIII crystals are reported without significant fluctuations. Figure 4.11: Rate of the 3.2 keV events selected by the coincidence above the threshold. Average values of 9.32 ±3.63 counts/day for ANAIS-0 and of 8.46 ±3.11 counts/day for PIII crystals are shown as horizontal lines in both plots. 238Uand 232Th chains Activities of the different branches in the 238Uand 232Thchains have been quantified after identifying their alpha emissions. Tables 4.8 and 4.9 show the type of decay, lifetimes and energies of the different alpha emissions for all the isotopes of both chains. The visible energy corresponding to these disintegrations should be: Eα·Qα+Erecoil ·Qrecoil because all the energy is absorbed in the detector. However, as Qrecoil for so heavy nuclei should be really low [85] we have not considered this contribution in the following. The acquisition dead time per event (≈4 ms, see section 2.5) complicates the observation of some alpha emissions, as 214Po and 212Po. However 212Bi −Po events sequence can be identified in the same digitization window, because of the very short lifetime of 212Po. In Figure 4.12, one example of such a pulse sequence is shown, corresponding to an special run of set-up 4 in which the digitization dynamic range was adjusted to match 123 4. Background understanding T1/2Isotope Eα(MeV) I (%) Activity 4.468 ·109y238U ↓α4.18 99.9 A0 24.1d234Th ↓β 1.17 m234mPa ↓β 2.455 ·105y234U ↓α4.75 99.8 A1 7.538 ·104y230Th ↓α4.66 99.7 A2 1600 y226Ra ↓α4.78 94.4 A3 3.8d222Rn ↓α5.49 99.9 A3 3.10 m218Po ↓α6.00 99.9 A3 26.8m214Pb ↓β 19.9m214Bi ↓β 164.3μs(*) 214Po ↓α7.69 99.9 A3 22.3y210Pb ↓β 5.01 d210Bi ↓β 138.4d210Po ↓α5.30 100 A4 Stable 206Pb Table 4.8: Summary of the 238Uchain general information with more details in alpha decays [127]. The type of decay and lifetime of every isotope are shown, as well as the energies of the alpha emissions. If two or more alpha lines are closer than the energy resolution, the weighted average energy is shown. A0 - A4 are the possibly independent activities, if secular equilibrium in the chain is not imposed. (*) The acquisition dead time per event (≈4 ms, see section 2.5) complicates the observation of this line, that should be strongly inhibited. high energy events. This population can be identified by the difference of the positions of the minimum and pulse onset. As αevents are faster in NaI(Tl) than β/γ ones (see section 3.6), they can be discriminated by Pulse Shape Analysis (PSA) techniques. The digitalization scale is optimized for the study of the low energy events and in most of the available data the alpha events 124 4.1. Background measurements and radioactive contaminants identification T1/2Isotope Eα(MeV) I (%) Activity 1.405 ·1011 y232Th ↓α3.99 99.9 A5 5.75 y228Ra ↓β 6.15 h228Ac ↓β 1.9116 y228Th ↓α5.37 99.4 A6 3.66 d224Ra ↓α5.69 94.9 A6 55.6s220Rn ↓α6.29 99.9 A6 0.145 s216Po ↓α6.78 99.9 A6 10.64 h212Pb ↓β 60.55 m212Bi ↓α36% 6.06 35.9 0.36·A6 ↓β64% 0.299 μs(*) 212Po ↓α8.78 100 0.64·A6 3.053 m208Tl ↓β Stable 208Pb Table 4.9: Summary of the 232Th chain general information with more details in alpha decay [127]. The type of decay and lifetime of every isotope are shown, as well as the energies of the alpha emissions. If two or more alpha lines are closer than the energy resolution, the weighted average energy is shown. A5 and A6 are the possibly independent activities to be determined if secular equilibrium is not imposed. (*) The acquisition dead time per event (≈4 ms) prevents the observation of this line, but 212Bi −Po events are clearly identified in our data (see Figure 4.12). are out of the digitizer dynamic range. Hence, we decided to choose as discrimination parameter a polygonal cut in two variables: QDC energy estimate in the very high energy window and pulse area, see Figure 4.13. The parameter pulse area is clearly saturated at the alpha events region. However, the signal arriving to the QDC has been attenuated and after integration in a microsecond window does not saturate. For this reason, a faster pulse with the same QDC value presents a smaller area. The total high energy spectrum and the alpha events selected with the referred polygonal cut (see Figure 4.13) are shown in Figure 4.14. The alpha events have as origin bulk crystal contaminations since αparticles have a very short mean free path in NaI. 125 4. Background understanding Figure 4.12: Example of 212Bi −Po event. The position of the minimum of the pulse is shown as a vertical red line and the pulse onset is shown as a vertical blue line. From the difference between both positions, events associated to this 212Bi−Po sequence can be identified. Figure 4.13: Plot of the pulse area vs. electron equivalent energy (estimated with the QDC and calibrated with gamma lines) in the very high energy range for set-up 2 of ANAIS-0. Only the PMT 1 signal is shown because PMT 0 signal was not stable, see Figure 4.15. Alpha events (selected by a polygonal cut in these two parameters and shown in red in the figure) have a smaller area for a given QDC than beta/gamma (black) ones. The data studied here were taken with the set-up 2 (see Table 2.6), i.e. ANAIS-0 module with Hamamatsu LB PMTs coupled to 10 cm light guides, measuring at the old Canfranc Underground Laboratory facilities. Only data from one of the two PMTs (PMT 1) have been considered in the following because the other (PMT 0) showed fluctuations in gain along the measurement (see Figure 4.15). If radioactive secular equilibrium in both chains is not imposed, there are seven possibly independent activities (A0 - A6) to be determined (see Tables 4.8 and 4.9). If there are two or more alpha lines from the same isotope closer than the energy resolution, the weighted average is shown. 126 4.1. Background measurements and radioactive contaminants identification Figure 4.14: The total high energy spectrum (black) of ANAIS-0 and the alpha spectrum built with the events selected by the polygonal cut shown in Figure 4.13 (red). 212Bi −Po events (between 5 and 9 MeV) are only partially identified by this cut, because they are not pure alpha events. The energy is shown in electron equivalent energy units. Figure 4.15: Energy vs time for alpha selected events for each PMT signal for data corresponding to ANAIS-0 setup 2. PMT 0 signal (top) was not stable and has not been considered for the analysis. As the energy scale of the alpha events is different from that of β/γ events, because of the different quenching, specific calibration of the alpha spectra is mandatory, and external sources cannot be used with such a purpose. Because of that, an autocalibration procedure has been followed, using only information from background data. First, 220Rn and 216Po lines have been identified in the following sequence of fast alpha decays: 55.6s0.145s 220Rn −→ 216Po −→ 212Pb (4.2) 127 4. Background understanding Looking at the time interval between one alpha event and the following (Δpost)or previous (Δpre) alpha event, 220Rn (if Δpre <55 s and Δpost <0.15 s) and 216Po (if Δpre <0.15 s) lines are identified in the alpha spectrum (see Figure 4.16). Figure 4.16: Total alpha spectrum is shown in black. The events identified as coming from the decays of 220Rn and 216Po are shown in red and blue, respectively. Then, we built a preliminary calibration function using these 220Rn and 216Po identified lines and, taking it as a guide, the expected alpha energy values from 238Uand 232Th chains are assigned to the seven main peaks observed in the spectrum, see the position of the peaks and the energy assigned in Table 4.10 and in Figure 4.17 the linear fit. Channel Eα(MeV) 75 4.20 94.5 4.78 114 5.30 136.5 6.00 146.55 6.29 164.2 6.78 198.58 7.69 Table 4.10: Position of the main peaks in the alpha spectrum (shown in Figure 4.16) with the energy associated to each of them. The result is Eα(MeV)=(0.0284 ±0.0003) ·Ch+(2.0885 ±0.0464) (4.3) 128 4.2. Background simulation •Activities of the different isotopes in the 238Uand 232Th chains, without imposing secular equilibrium, have been quantified after identifying their alpha emissions by means of Pulse Shape Analysis (PSA) technique (also detailed in section 4.1.2 and results shown in Table 4.11). •Concerning the activity of 129I, cosmogenic isotope usually found in iodine compounds in a very broad range of activity values (depending on the ore origin), the value deduced for DAMA/LIBRA crystals [97], produced (as the ANAIS-0 module crystal) by Saint Gobain company has been used: 129I/natI=(1.7±0.1) ×10−13 and T1/2=1.57 ×107y, which corresponds to a specific activity of 0.96 mBq/kg. •Other isotopes such as 227Ac,87Rb,85Kr or 235Uhave also been considered and studied, but no evidence of their presence has been found in the data and they have not been included in the ANAIS-0 background model. Isotope Activity (mBq/kg) 40K12.7±0.5 129I0.96 ±0.06 232Th 0.013 ±0.005 228Th 0.035 ±0.003 238U/234U0.075 ±0.005 230Th 0.023 ±0.007 226Ra 0.098 ±0.004 210Pb 0.188 ±0.005 Table 4.12: Isotopes and activities considered in the simulation distributed in the bulk of the NaI crystal of the ANAIS-0 module. External radioactive contaminants Besides the NaI crystal, PMTs are usually the main background contribution in scintillation detectors. For the rest of the components only upper limits for the different radioactive isotopes searched for have been obtained. For most of these external components High Purity Germanium (HPGe) spectrometry at LSC has been used to screen the radioactive contaminations (results are reported in Table 2.1). •Several low background Electron Tubes and Hamamatsu PMT models have been tested with the ANAIS-0 module. All of them were previously measured at the HPGe test bench at the LSC, see section 2.1.3. The activities of 238U,226Ra,232Th, 135 4. Background understanding 60Co and 40K, shown in Table 2.5, have been taken into account. Equilibrium in the natural chains has been observed except if otherwise is explicitly stated. •For quartz windows, light guides, optical coupling grease, and copper encapsulation the upper limits (95% C.L.) on the activities considered in the simulations are shown in Table 4.13 and have been obtained at the HPGe test bench at the LSC. Upper limits for 238Uare worse than those obtained for other isotopes because of the low intensity of its gamma emissions; in this case, equilibrium in the natural chains has not been imposed. •For archaeological lead, upper limits on 210Pb,232Th and 238Uactivities quoted in [158] have been used. •For radon content in the air filling the inner volume of the shielding, there is no real estimate. Radon content in the laboratory air is being continuously monitored, and the inner volume of the shielding is flushed with boil-off nitrogen to guarantee its radon-free quality. An arbitrary value for the radon content in the inner volume air of about one hundredth of the external air radon content has been assumed in our background model (0.6Bq/m3), compatible with the absence of lines coming from radon daughter isotopes in the measured background. Nevertheless, this contribution should be considered as an upper limit, as no direct evidence of the presence of radon in the data has been found. Contributions from neutrons and environmental gamma background have been also simulated, being negligible at the present level of sensitivity. Contribution from muons interacting in the crystal (and other muon related events) can be vetoed by the coincidence with a signal in the plastic scintillators covering the ANAIS-0 shielding. Although the coverage was only partial in the experimental layout, the corresponding residual background contribution has been estimated to be negligible, see section 2.3. 4.2.3 Code validation and analysis of systematic effects Energy spectra for different calibration sources have been simulated to validate the code. Systematic effects, specially at very low energy, are expected to be higher for external than for internal sources and are not easy to evaluate. With such a purpose, experimental and simulated spectra for all the radioactive sources measured with ANAIS-0 have been compared. The energy sources used for the high energy range are: 22Na,152Eu,133Ba and 137Cs, whereas low energy ones are: 57Co,109Cd,133Ba and 137Cs.IntheFigures 136 4.2. Background simulation Detector component Isotope Activity Copper encapsulation 40K<11 mBq 232Th <4.1mBq 238U<140 mBq 226Ra < 2mBq 60Co < 0.94 mBq Quartz optical window 40K<12 mBq/kg 232Th <2.2mBq/kg 238U<100 mBq/kg 226Ra < 1.9mBq Light guides 40K<21 mBq/guide 232Th <4.1mBq/guide 238U<120 mBq/guide 226Ra < 4.7mBq/guide Optical coupling grease 40K<200 mBq/kg 232Th <200 mBq/kg 238U<2000 mBq/kg 226Ra < 30 mBq/kg Archaeological lead 210Pb < 20 mBq/kg 232Th <0.3mBq/kg 238U<0.2mBq/kg Inner volume air 222Rn < 0.6m3 Table 4.13: Upper limits given at 95% C.L. for the content on the main radioactive isotopes in the different components of the ANAIS-0 module layout. 4.23 and 4.24 the experimental calibrations are compared with the simulations normalized to the activity of the sources (known only at a level of ±10% accuracy, except for 152Eu, whose activity is better known at ±1.5%). Response functions to take into account energy resolutions effects are different at high and low energy ranges, as it has been previously explained. With the same goal, radon was allowed to enter into the shielding in order to have data for an homogeneously distributed contamination in the air around the ANAIS-0 module. In Figure 4.25 experimental (after subtracting the background) and simulated spectra are shown for radon contamination in the inner cavity air, normalizing both spectra to the same total number of counts. The radon content required to match both spectra is 58.2Bq/m3. The radon level in the laboratory has been continuously monitored along the corresponding period with an AlphaGUARD detector (mean value, in this period, of 65 Bq/m3), see section 2.6. 137 4. Background understanding Figure 4.23: Measured and simulated spectra for different calibration sources in the high energy range: 22Na,133Ba,137Cs and 152Eu. Simulations have been normalized to the nominal activity of the sources. From the comparison between simulations and measured data, it can be concluded that simulation reproduces qualitatively very well all the measurements. Concerning the quantitative estimate of the activity, some systematics have been identified for external gamma calibration data, specially at the lowest energies; they could be related with our imprecise knowledge of the sources geometry and possible misplacement of the source with respect to the Mylar window, but we cannot discard a reduced scintillation yield or bad light collection efficiency for events corresponding to interactions of very low energy X-rays in the first microns of the crystal surface, that could explain, for instance, the non-linear response evidenced for the iron X-rays in Figure 4.24 and already brought forward in section 2.7. Comparing simulation and measurement (see Table 4.14) we can state that simulations agree with experimental data at the accuracy level of our knowledge of activity and other source details when using for the normalization the total number of detected counts, except for the 152Eu calibration source. However, some differences in Compton vs. photopeak relative event rates have been identified in the simulations (see Table 4.14): this suppression of Compton vs. peak event rates in the simulation can be possibly ascribed to 138 4.3. Background model Figure 4.24: Measured and simulated spectra for different calibration sources in the low energy range: 57Co,133Ba,109Cd and 137Cs. Simulations have been normalized to the nominal activity of the sources. some energy loss mechanisms in the detector performance, f.i. because of inhomogeneity in the light collection efficiency, resulting in less events contributing to the peaks but the same total rate. This Compton/peak defect between experimental and simulated data has been quantified as 75% with the 22Na calibration (using the 1275 keV gamma line intensity and the number of events in the region from 600 to 1100 keV). The lines associated to X-rays are systematically badly reproduced by the simulation (see Figure 4.24); this could be explained by a malfunction of Geant4 mechanism for the generation of the atomic electromagnetic deexcitation cascades after a nuclear process. There is also some error present in the 152Eu simulation: it does not follow the same tendency as the other sources do. However, no explanation has been found for the large overestimate, in particular, of the high energy gamma lines. 139 4. Background understanding Figure 4.25: Comparison of simulation (red) vs. experimental data (black) for a homogeneously distributed radon contamination in the air around the detector, inside the shielding. Source Activity Estimated activity (A) Estimated activity (B) 22Na (1 ±0.1) μCi 1.08 μCi 0.95 μCi 137Cs (1 ±0.1) μCi 1.07 μCi 0.99 μCi 152Eu (1.07 ±0.02) μCi 0.78 μCi 0.87 μCi 57Co (1 ±0.1) μCi 1.18 μCi 1.09 μCi 133Ba (1 ±0.1) μCi 0.83 μCi 0.90 μCi 222Rn (63 ±3.4) Bq/m358.2 Bq/m337.2 Bq/m3 Table 4.14: Comparison between nominal and derived from the simulation activities for the different calibration sources used with the ANAIS-0 module. Activities are derived using the total number of counts in the spectrum (A) and a single line (B): 1275 keV for 22Na, 662 keV for 137Cs, 344 keV for 152Eu, 122 keV for 57Co, 81 keV for 133Ba and 609 keV for 222Rn. 4.3 Background model The total energy spectrum in ANAIS-0 has been simulated for the different contaminations reported in section 4.2.2 and the different available experimental set-ups shown in Table 2.6. 140 4.3. Background model 4.3.1 Background model Internal background The contributions of internal radioactive contaminants, clearly dominant in the ANAIS0 module background, are shown separately for alpha particles (see Figure 4.26) and beta/gamma ones, in this case together with the corresponding uncertainties at 1 σlevel (see Figure 4.27). For these contaminants, the same contribution is expected for all the set-ups. Figure 4.26: Simulation of the alpha spectrum, for the bulk crystal contaminations in U and Th chains. The Geant4 version used in this work presents a very relevant modification, related with the shape of the 40Kbeta spectrum, because the proper shape factor for the third unique forbidden beta decay has been used, see Figure 4.28. It can be seen a much better accordance with the measured background spectrum shape at 1 MeV for the latter version: in both cases 12.7±0.5mBq/kg of 40Khave been considered, as deduced from the analysis of coincidences. External background In Figure 4.29 the contributions of the external radioactive contaminants (placed in PMTs, optical windows, light guides, optical coupling grease, copper encapsulation, inner volume 141 4. Background understanding (a) (b) Figure 4.27: Contribution to the background from internal radioactive contaminants for the ANAIS-0 crystal at high (a) and low energy (b). Uncertainties in the estimates of the contamination values used at 1 sigma level are shown in the figure. air, and archaeological lead shielding) to the background are shown for the different setups. In Figure 4.30 the total contributions to the background coming from external components are shown for every simulated set-up. It has to be noticed that for most of 142 4.3. Background model Figure 4.28: Simulated spectra for bulk 40Kcontamination in the ANAIS-0 NaI crystal are shown both for an allowed beta spectrum shape (Geant4.9.3 version) in red and taking into account the proper shape factor for the third unique forbidden beta decay (Geant4.9.4 version) in blue. The latter is the Geant4 version used throughout this work. these contaminations only upper limits have been derived, so the contributions to the background shown should be also taken as upper limits at 95% C.L. Figure 4.29: Background contribution from every external component of the ANAIS-0 set-ups. 143 4. Background understanding Figure 4.30: Total contribution to the background for every simulated ANAIS-0 set-up coming from contaminations in external components. 4.3.2 Comparison between background simulations and measurements The background measurements for the ANAIS-0 set-ups (summarized in Table 2.6) have been compared with the addition of the simulated contribution of the different components. First, the simulated alpha spectrum is compared with the alpha spectrum from the setup 2 (the alpha contribution to the background should be the same for all the set-ups). Then beta/gamma spectra at high and low energy are compared for set-ups 1 - 5. The energy spectrum corresponding to alpha particle energy depositions, discriminated by PSA as stated in section 4.1.2 has been compared with the corresponding simulation, assuming a homogeneous bulk distribution of 238Uand 232Th, and daughter isotopes in the NaI crystal with the activity values deduced and shown in Table 4.12. Above 6 MeVee the experimental spectrum shown in Figure 4.31 is the total spectrum in order to show all the 212Bi −Po events. Effect of acquisition dead time on 214Poemission detection efficiency and the summing of alpha and beta emissions for the 212Bi−Po sequence in the charge integration window have been properly taken into account in the simulated data (see Figure 4.31). Quenching factor for alpha particles vs. electrons has been fitted to properly match experimental data and simulation in the Bi-Po events (above 9 MeV in alpha energy). 144