Repositorio Institucional de Documentos

2013 78 Alfredo Tomás Alquézar Development of time projection chambers with micromegas for Rare Event Searches Departamento Física Teórica Director/es Villar Rivacoba, José Ángel García Irastorza, Igor Director/es Tesis Doctoral Autor Repositorio de la Universidad de Zaragoza – Zaguan http://zaguan.unizar.es UNIVERSIDAD DE ZARAGOZA Departamento Director/es Alfredo Tomás Alquézar DEVELOPMENT OF TIME PROJECTION CHAMBERS WITH MICROMEGAS FOR RARE EVENT SEARCHES Física Teórica Director/es Villar Rivacoba, José Ángel García Irastorza, Igor 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 Development of Time Projection Chambers with Micromegas for Rare Event Searches. memoria presentada por Alfredo Tomás Alquézar para optar al grado de doctor en Física Laboratorio de Física Nuclear y Astropartículas Área de Física Atómica, Molecular y Nuclear Departamento de Física Teórica Universidad de Zaragoza May 2013 Scope of the thesis. This thesis reports diverse research efforts developed in the frame of the T-REX project which intends the application of gaseous Time Projection Chambers (TPCs) to Rare Event Searches. The activities include pure R&D together with specific research lines for the experiments CAST (CERN Axion Solar Telescope) and NEXT (Neutrinoless double beta decay Experiment with a Xenon TPC). The Laboratorio de Física Nuclear y Astropartículas (LFNA) of the University of Zaragoza brings an exceptional opportunity to undertake this work because of its near three decades of experience in Rare Event Searches and a close relationship with the development of the Canfranc Underground Lab (LSC); besides its collaboration with the IRFU (Institut de Recherche sur les lois Fondamentales de l’Univers) at CEA/Saclay, which is the main authority on micromegas, one of the prominent Micro-Pattern Gaseous Detectors (MPGDs). Some of the Rare Event Searches that motivate the work are briefly reviewed in chapter 1, remarking how TPCs can contribute to the different scenarios. The main concepts required to follow the rest of the work are concisely introduced in chapter 2. The development of micromegas is quite recent, specially for the last micromegas types, bulk and microbulk. Therefore the exploration and characterization of their properties, even those that were already well known for classical layouts, is a first, basic but essential, stage of the research. Chapters 3 and 4 were aimed to be a practical reference (manual-like) about customary TPC features, despite focussing in micromegas particular issues. Part of the R&D original work on basic micromegas operation was inserted among these chapters. Chapter 5 can be seen as an appendix on TPCs and micromegas features of direct interest for particular applications to Rare Event Searches. A software package for simulation of TPCs and micromegas was developed for T-REX and is generically described in chapter 6. Chapter 7 contains two particular examples of applications of the software to dark matter and to double beta decay TPCs. The second part of the thesis naturally focusses on the CAST experiment, as the unique running Rare Event Search experiment that uses the state of the art micromegas. The experiment, that has a remarkable relation with the evolution of micromegas detectors, is presented in chapter 8. Chapters 9 and 10 report R&D on CAST micromegas background based on, respectively, simulations and experimental tests; and in chapter 11 their common conclusions are summarized as well as some reflections on the prospects of further background improvements. Contents 1 Rare Event Searches and TPCs. 1 1.1 Introduction. ............................... 1 1.2 Physical issues. .............................. 2 1.2.1 Neutrinoless double beta decay. ................. 2 1.2.2 Dark matter ............................ 6 1.2.3 Axions, ALPs and WISPs. .................... 12 2 TPCs and Micromegas. 19 2.1 The TPC concept. ............................ 19 2.2 New generation: MPGDs and micromegas. ............... 20 2.2.1 Bulk micromegas. ......................... 22 2.2.2 Microbulk micromegas. ...................... 24 2.2.3 InGrid. .............................. 25 2.2.4 GEMs ............................... 26 3 Interactions of particles in the micromegas chamber. 29 3.1 Ionization by charged particles. ..................... 30 3.1.1 Coulomb scattering. ....................... 30 3.1.2 Mean energy loss. Bethe-Bloch formula. ............ 30 3.1.3 Applicability of the Bethe-Bloch formula and thin absorbers. 33 3.1.4 δ-rays and restricted energy loss. ................ 33 3.1.5 Stopping Power of non relativistic ions. ............ 38 3.2 Other indirect mechanisms to generate ionization ........... 45 3.2.1 Interactions of photons. ..................... 46 3.2.2 Interactions of neutrons and WIMPs. .............. 50 3.2.3 Bremsstrahlung. ......................... 52 4 Working principles of the TPC-micromegas. 55 4.1 Primary Charge generation. ....................... 56 4.1.1 W, averaged energy per ion pair. ................ 56 4.1.2 F, the Fano factor. ........................ 58 4.2 Drift, diffusion and attachment. ..................... 58 4.2.1 Drift velocity. ........................... 60 4.2.2 Diffusion. ............................. 63 4.2.3 Attachment and recombination. ................. 65 4.3 Mesh transparency. ............................ 66 4.3.1 Parametrization of the typical transparency fall. ....... 69 4.3.2 Limits for the mesh design. ................... 71 4.3.3 Transparency quench. ...................... 72 4.3.4 Transparency and diffusion. ................... 73 iv Contents 4.4 Amplification: gain and energy resolution. ............... 75 4.4.1 The Townsend coefficient and the micromegas gain. . . . . . . 75 4.4.2 Avalanche fluctuations and energy resolution. ......... 78 4.4.3 Experimental examples: gain curves and energy resolution. . 82 4.5 Signal induction. The Ramo’s Theorem. ................ 90 5 Micromegas’ milestones for Rare Event Searches. 93 5.1 Behaviour at high pressure. ....................... 94 5.2 Use of pure noble gases. ......................... 97 5.3 Penning mixtures. ............................ 99 5.4 Radiopurity. ................................ 100 5.5 Electronics. T2Kelectronics library. ................... 102 5.5.1 T2K electronics set-up. ..................... 102 5.5.2 T2K electronics library. ..................... 103 5.5.3 TPC event reconstruction. .................... 104 5.5.4 Outlook. .............................. 108 6 Simulations for Rare Event Searches with TPCs 111 6.1 RESTSoft. ................................. 112 6.2 The RESTSoft GEANT4 application. .................. 113 6.2.1 Basic description of a simulated event in a TPC. ....... 113 6.2.2 Physical processes. ........................ 118 6.2.3 Geometry. ............................. 118 6.2.4 Event generation. ......................... 120 6.2.5 Simulation control. ........................ 120 6.2.6 Information management. Study of cases. ........... 122 6.2.7 Direct applications of the RESTSoft GEANT4 application. CAST micromegas quantum efficiency and HPXeTPC spectrum.124 6.3 The RESTSoft TPC simulation chain and Event Model. ....... 129 6.3.1 Primary charge generation and bridge classes ......... 130 6.3.2 Importation of a GEANT4 simulation. ............. 130 6.3.3 Simulation of drift effects and attachment. ........... 131 6.3.4 Mesh transparency and micromegas amplification. . . . . . . 134 6.3.5 Pixelization. ............................ 135 6.4 Simulation of electronic signals. ..................... 137 6.4.1 Generation of electronic pulses by RLC circuit solving. . . . . 138 6.4.2 Generation of electronic pulses by convolution with the response function. ......................... 139 6.4.3 TRestShaper. ........................... 141 7 Applications of RESTSoft on High Level Analysis Challenges. 143 7.1 Electron/recoil discrimination with a Dark Matter TPC (TREX-DM). 143 7.1.1 Project architecture. ....................... 143 7.1.2 Preliminary discussion of the scenario. ............. 145 1.2. Physical issues. 5 experiment based on bolometers, like CUORE[12]. When the source/detector is a fluid contained in a vessel, the scaling is more natural. The complexity of the detector is essentially related with the readout, which grows slowly than the mass (typically the mass-readout ratio would go as n2/3with nthe scaling factor). So a liquid xenon TPC can be an interesting option since 136Xe is a good ββ0νcandidate, with a high Qββ, relatively abundant and with a not very intense ββ2νmode. The latter is so much so that only recently the EXO (Enriched Xenon Observatory)[13] collaboration, with 200 kg of xenon enriched 80% in 136Xe, has reported the first observation of the ββ2νdecay of this isotope[14]. The observation has been confirmed more recently by the KamLAND-Z en (KamLAND Zero-Neutrino Double-Beta Decay) experiment[15]. That is a 300 kg xenon (enriched in 136Xe) calorimeter with capability to reconstruct the vertex and strong active shielding. The liquid TPC has modest energy resolution compared with germanium detectors and bolometers (σ/E < 2% is showed in [13]) and limited topological capabilities, compared with a gaseous TPC, for discrimination. However both features are enhanced by the operation in hybrid mode that allows the measurement of light and charge and the self-shielding is excellent. In conclusion the liquid Xe TPC seems very appealing because of being a very balanced solution. 1.2.1.2 Gaseous TPCs for ββ0νdecay. The double beta decay has a clear topological signature: that of two electron tracks with a common vertex (see section 7.2). That suggests that a gaseous TPC would be powerful for event discrimination. The NEMO (Neutrino Ettore Majorana Observatory)[16] collaboration already exploits that by introducing the target foils in a kind of narrow gaseous chambers. However, because of leaving the source/detector identification, the efficiency and energy resolution are degraded. Moreover the scaling possibilities of this approach are linear in the same sense as in the germanium and bolometer arrays. A gaseous enriched 136Xe TPC is highly appealing. In principle most of the benefits of the liquid TPC approach are kept or enhanced, the drift is easier, no cryogenics is needed, the energy resolution is better and the mass can scale just by increasing the pressure. The Gotthard TPC[17], with 3.3 kg of gas at 5.5 atm, was an encouraging attempt, pioneer of the application of TPCs to Rare Event Searches in the 90’s. Its main limitation was the energy resolution, which is the final requirement for all the detectors, since both neutrino and neutrinoless modes cannot be distinguished by topological means. However, the introduction of MPGDs (Micro-Pattern Gaseous Detectors, see section 2.2) makes modern TPCs overcome the Gotthard TPC’s limitations and they could exploit the promising results it actually showed. However a big enough high pressure gaseous xenon (HPGXe) TPC has still to be demonstrated (see some 6 Chapter 1. Rare Event Searches and TPCs. hints in chapter 5), and also that the discrimination capabilities (see section 7.2) would be powerful enough so as to compensate the total mass disadvantage (a first deep study about that is done in a Ph.D. work inside T-REX[18], which is extended in [19]). NEXT (Neutrinoless double beta decay Experiment with a Xenon TPC)[20] is expected to operate 100 kg of enriched xenon in the Canfranc Underground Laboratory. The NEXT project’s highlight is a full three dimensional reconstruction of the event topology, including its absolute position thanks to the combination of light and charge signals. NEXT has a R&D program dedicated to micromegas. The EXO project also supports a collateral R&D program for a HPGXe TPC[21]. 1.2.2 Dark matter The observational evidences that support the presence of Dark Matter (DM), accumulated since the 30’s, are quite numerous and variate (see a review in [22]). It is particularly significant that they appear at very different scales, from galaxies to clusters up to the largest observed astronomical structures. However the most valuable support comes from Cosmology; the last decade saw the birth of the Cosmology’s precision era, whose results[23,24,25] crystallised in the ΛCDM model. That means a new paradigm for the Universe’s History where its composition is quantified and rather constrained, giving an important role for DM. According to the customary notation in Cosmology the different contributions to the density of energy ρare denoted as Ωi≡ρi ρc (1.3) where ρcis the critical density that corresponds with a flat, i.e. with Euclidean metric, Universe. Today, the two relevant contributions are ΩM, the matter density, and ΩΛ, which is the Dark Energy (DE) that causes the accelerated expansion of the universe. Figure 1.4 shows eloquently how the three independent observations were coincident in a scenario dominated by DE, with an also very important contribution from matter that requires most of it to be DM, and probably a flat universe. Nevertheless the deductions taken from the Cosmic Microwave Background (CMB) registered by WMAP put the most precise constrains by far. Therefore the ΛCDM model suitability to fit the observational data has been recently confirmed and its accuracy be improved by the results of WMAP’s successor, the PLANCK[26] in March 2013[27]. The new 68% limits are[28]: ΩM= 0.314 ±0.020 (1.4) ΩΛ= 0.686 ±0.020 (1.5) From the 31.4% of matter density only the 4.2% corresponds to baryonic matter, constrained by the Big Bang Nucleosynthesis (only ≃0.3% is luminous); so 27.2% 1.2. Physical issues. 7 Figure 1.4: Three independent sets of observations: high-redshift supernovae, galaxy cluster inventories, and the CMB, converge nicely. The dark grey corner (down right) corresponds to a too recent Big Bang. Figure extracted from [24]. is non-baryonic DM and 68.6% is DE. The nature of the latter is even less known than that of DM, being the theoretical proposals rather diverse, from the Einstein’s cosmological constant to quintessence fields. Its study is, in principle, out of the scope of this work3. The PLANCK’s result for the Curvature Parameter today, that accounts for departures from spatial flatness, is ΩK≡1−PΩi=−0.0096+0.010 0.0082 (68% CL), what means a great hint towards a Universe with zero curvature. The observations that allow to reconstruct the DM distribution show it to be relatively disperse in big halos that surpass the galaxy and cluster luminous dimensions[29]. The collision at the Bullet cluster[30] evidences the nature of DM dynamics: very weak interactions with itself and the other types of matter involved in the collision (gas and galaxies). That observation confirms DM to be very weakly interacting, but also atomized, likely composed of simple particles. In ΛCDM, CDM stands for Cold DM, i.e. composed of non relativistic particles. A too large contribution of hot DM does not allow the structure formation as it is known to happen, from smaller to bigger structures. That rules out neutrinos as the only explanation (neither the main contribution, by far), and also axions that would had been generated thermically in the primordial plasma. However, a different mechanism can produce non relativistic axions, see section 1.2.3 about axions and axion-like particles. Therefore the DM candidates must be looked for beyond the Standard Model. If they are relatively heavy, more likely to be non relativistic, they are generically called WIMPs (Weakly Interacting Massive Particles). Also WISPs (Weakly Interacting 3There could be a link, nevertheless, regarding chamaleons, see section 1.2.3. 8 Chapter 1. Rare Event Searches and TPCs. Slim Particles), like axions, are candidates if produced so as not to be relativistic (see section 1.2.3). It was decided to talk only about WIMPs in this DM section because they are the traditional DM candidate, searched in most of DM experiments, and the experimental approaches of WIMP searches are usually very different from those searching WISPs. But WISPs are as good DM candidates as WIMPs; and WIMPs are as well motivated theoretically (independently on the DM problem) as WISPs. There is no lack of WIMPs in the Standard Model proposed extensions, like SUSY (Supersymmetry). Reviews on the zoo can be found in [22,31]. 1.2.2.1 Direct detection of WIMPs. WIMPs direct detection inherits the requirements and techniques from ββ0νdecay experiments, in fact many of the first limits where given as by-products of ββ0ν decay experiments[32]. The challenge becomes even more exigent because of the detection at low energies and the absence of a peak in the expected spectrum (see figure 1.5). In principle, no event rejection based on topology (like simple PSA) is powerful with low energy events beyond the rejection of multi-interaction events. The energy resolution cannot help to focus the search in a narrow range of interest. The discrimination of γ/β and αby other means is mandatory. Figure 1.5: Examples of WIMP-induced nuclear recoil expected rates and spectra for a spin-independent nucleon cross section of 10−42 cm2and both 100 GeV and 500 GeV WIMP masses for three example target materials: germanium (dot), xenon (solid) and argon (dashed). Energy deposits are given in electron equivalent energy (see section 3.1.5.1). The figure is extracted from [33]. The use of hybrid bolometers has been specially successfully since the comparison of different energy measurements depends on the interaction type. Thus the capability of registering both heat and light (CRESST II[34], ROSEBUD[35]) or 1.2. Physical issues. 9 heat and charge (CDMS II[36], EDELWEISS II[37]) brings the possibility of selecting the nuclear recoils that might be caused by a WIMP scattering. The way DM experiments are evolving is similar than that of the ββ0νdecay ones. Therefore an increase of the available information (for discrimination) is procured. As well as scaling masses up, but also segmentation (for self-shielding). The liquid TPCs have advantage in this sense since only their readout must be segmented, they guarantee the best self-shielding and the energy resolution limitations are not so critical in the DM context. They can be competitive because hybrid Liquid TPCs, reading both charge and light, have been developed (ZEPLIN III[38], XENON[39], LUX[40]) thanks to the combination of a liquid phase (which is the target itself) with a gaseous phase for charge amplification. For further scaling up, the use of argon, though apparently with worse scintillation/ionization properties, can be preferred because of its much lower cost. An argon-based liquid TPC, the ArDM project[41], will be soon operational at the Canfranc Underground Laboratory. 1.2.2.2 DM signatures and WIMP searches with TPCs. The kind of spectra plotted in figure 1.5, with pseudo-exponential shape and a energy range depending on the WIMP’s mass, makes very challenging the claim of a positive signal. To identify an excess of counts the background must be completely explained. The nuclear recoils caused by neutrons are practically not distinguishable from WIMP-induced ones. Underground sites usually have relatively high neutrons flux from fission in the rocks, which, in addition, is difficult to be characterized with accuracy. The recognition of an unambiguous extraterrestrial signature in the background signal could be indispensable to convince the DM community. The most famous and significant case is that of the DAMA experiment. DAMA was based in NaI(Tl) scintillators and claims to have found the annual modulation of the DM signal in its background evolution. The WIMPs interaction rate is expected to be modulated by the changing relative velocity of the Earth with respect to the galaxy (and so with respect to the DM halo) along its translational movement around the Sun. Currently DAMA/LIBRA (experiment upgrade to 250 kg of NaI) has accumulated 0.87 ton ×year along 6 complete annual cycles founding a positive signal with 8.9σ[42]. However the DM community has not accepted the result as a WIMPs direct detection evidence. It is also true that the claimed WIMP has been largely rejected by other experiments with, in principle, much better sensitivities (see figure 1.6). But the theoretical uncertainties still may have degrees of freedom to make compatible the DAMA positive results with the rest of negatives (see the next point about light WIMPs). The Canfranc Underground Laboratory will host the ANAIS[43] experiment that will try to confirm the DAMA result by means of the same basic technique, despite producing an independent result. Another DM signature would require the use of a gaseous TPC: the directional signature, which is actually related with the annual and daily modulations. In galactic coordinates the WIMPs wind always comes from a fixed direction, approximately 10 Chapter 1. Rare Event Searches and TPCs. Figure 1.6: WIMPs current exclusion plot summary for spin-independent WIMP-nucleon scattering and standard DM galactic halo[45] as published by XENON100[44], currently the most sensitive DM experiment. The remarked regions correspond to DAMA and other WIMP claims (by CoGeNT and CRESST), in principle excluded by XENON100. The XENON100 immediate prospects are exploring the SUSY favoured region (grey 1σ/2σ contours). The plot shows the amazing improvement in experiments sensitivity about 4 orders of magnitude during the last decade, since in the early 2000’s the DAMA region was not covered. from the Cygnus constellation in the Earth’s night sky view. Thanks to the correlation between the incident expected WIMP direction and the initial recoil’s one, a peak, that cannot be found in the energy spectra, should be identified in the direction distribution (see figure 1.7). In fact this method avoid the essential problem of neutron-induced background4. The experimental challenge is impressive: the reconstruction of low energy nuclear recoil tracks. It requires not only a gaseous TPC, but one working at low pressure, so decreasing the target mass. However, directionality allows a claim of a positive signal with less accumulation of statistics[46] which would be reinforced by an unmistakable galactic signature. A correct reconstruction of the recoil track is challenging even before considering the practical details. The ionization theory of the recoils in the range of energies of interest is situated in the knowledge frontier (see section 3.1.5). A fully confident Monte Carlo simulation that demonstrates potential for a high efficiency direction and sense recognition is not available. The performance of the TPC must be outstanding from all the points of view: high gain (so as to assure a low en4Another possibility the comparison of positive signals in different materials based on a different dependency on A, the atomic mass, of WIMPs and neutron cross sections. 1.2. Physical issues. 11 Figure 1.7: 100 GeV WIMPs induced nuclear recoils rate for cross section of 10−6pb and standard DM halo model, as they would be registered by an ideal directional detector in galactic (l, b)coordintates. Extracted from [46]. ergy threshold), and accurate imaging capabilities after the drift leaning on a highly granulated readout, in addition, a big chamber size (so as to procure a big target mass at low pressure). Nevertheless there are several projects that undertook the design and operation of prototypes. The DRIFT[47] collaboration has pioneered the field, proposing and proving non conventional TPCs (see section 4.2.2.1). However DRIFT prototypes’ performance has been limited in part due to the use of traditional wire planes readouts. The outstanding readout features demanded by the physical challenge make the present DM TPC projects opt for MPGDs: microdot in NEWAGE[48], GEM and micromegas in DMTPC[49] and micromegas in MIMAC[50]. These projects currently use fluorine as the target, instead of the classic option of argon or xenon. The reason is that only in the case of a WIMP cross section that were enhanced for non-zero spin nuclei, like fluorine, their discovery potential would be competitive. Furthermore the usage of a light nucleus brings practical advantages since the recoils are longer5. The community is growing, achieving milestones and generating expectations and gave birth to an international forum aiming to share R&D progress and also boost theoretical and simulation studies: CYGNUS (Cosmology with Nuclear Recoils)[51]. 5Unfortunately the use of 3He is practically forbidden at present because of its shortage. 12 Chapter 1. Rare Event Searches and TPCs. 1.2.2.3 Low mass WIMPs and TPCs. SUSY models prefer WIMP masses of ∼100 GeV; it is also true that sensitivity of common direct detection experiments sharply decreases for masses lower than ∼10 GeV because of the threshold of the discrimination techniques (see figure 1.6). Therefore the window of masses < 10 GeV is not strongly constrained, though not theoretically forbidden. A low threshold, at the sub-keV level, is the main requirement to be competitive, even with a raw counting detector and low target masses. The case for light WIMPs has been recently reinforced by the claim of an excess of counts in a novel low-noise germanium, CoGeNT[52], and analysis that proposes that a light neutralino could bring conciliation of CoGeNT excess and DAMA signal while the rest of experiments would be no sensitive to it[53,54]. TPCs could have a contribution to this question. First, simple recoil versus γ/β event discrimination based on event topology (track length) could be powerful. Even operating at atmospheric pressure, it could lead to relatively low discrimination thresholds ∼10 keV. A preliminary study on that possibility is presented in section 7.1. Second, TPCs equipped with micromegas readout have shown excellent discrimination capabilities as X-rays detectors below 10 keV. X-rays and nuclear recoils are not distinguished in this approach, but the discrimination for β,α, muons and noise rejection is efficient. The case is largely discussed in chapters 8and 9. The same detector type has been already operated underground producing encouraging results (see chapter 10). These detectors, thanks to the micromegas readout, seem to be already able of threshold below the keV. Moreover, when operating in optimum conditions, there is no sign of an increase of the background level at low energies (present software threshold = 2 keV) while software efficiency from analysis discrimination criteria remains high. The DM line of the T-REX project is oriented to this approach as a first goal. The detector mass of the prototypes already tried underground is easily scalable. The T-REX research summarized in section 5.1 even points to the possibility of increasing the gas pressure up to several bars without worsening the detector performance. 1.2.3 Axions, ALPs and WISPs. Axions are consequence of one of the most traditional and best motivated extensions of the Standard Model (SM). The Quantum Chromodynamics (QCD) lagrangian naturally accepts a term with the following form6: L¯ θ=¯ θαs 8πGµνa ˜ Gµνa (1.6) 6It is not only the possibility, but the therm had been invoked[55] to bypass another SM deficiency, the UA(1) problem[56]. 1.2. Physical issues. 13 where Gis the gluon field and ˜ Gits dual, αsthe strong coupling constant and ¯ θjust parametrizes the term’s strength. Such a term violates charge-parity (CP) symmetry, as electro-weak interactions actually do, and has observational consequences, like the prediction of an electric dipole moment for the neutron (nEDM). However experimental tests on nEDM constrain ¯ θ.10−10[57]. This extreme fine-tuning which has to be assumed by the SM is known as the strong CP problem[58]. The Peccei-Quinn (PQ) mechanism[59,60] proposes to see (1.6) as ∼aGµνa ˜ Gµνa, where ais a new field associated with a new U(1) global symmetry (the PQ symmetry), instead of a fixed parameter. The PQ symmetry is spontaneously broken at scale fa, which is not fixed by theory, then LQCD is CP-conserving as result of the dynamic relaxation to a minimum where L¯ θvanishes. The field awould manifest itself as a pseudo-Nambu-Goldstone boson[61,62]: the axion. Since fais higher than 105GeV, otherwise the axion would have been discovered in accelerators, its mass is small because is acquired by mixing with pions ma∼mπ fπ fa≃0.60 eV 107GeV fa (1.7) with mπ= 135 MeV and fπ≈92 MeV. The axion has a coupling with photons Laγ ≡gaγ ~ E·~ Ba, gaγ ∼1/fa(1.8) that leads to photon-axion oscillations in the presence of external magnetic fields[63]: the Primakoff7conversion of photons into axions (in the core of stars) and the inverse Primakoff conversion of axions into photons (in helioscopes, see later on). From (1.7) and (1.8) ma∼gaγ (1.9) and the three equations summarize the properties of the generic axion. Depending on the model an extra coupling with SM fermions can be postulated and the proportionality factors are varied[64]. They are called axion-like particles (ALPs) those that are light and share the axion phenomenology, i.e. the axion-like coupling with two photons (Primakoff effect), but are nor motivated by the strong CP problem neither linked with the PQ symmetry. In contrast to original QCD axions, for ALPs maand gaγ are independent parameters. They are typically consequence of new symmetries by SM extensions that are broken at high energy scales (like axions themselves), but also appear in string theory. They are an important key in the context of the low energy frontier of Particle Physics[65]. 7Named Primakoff by analogy with the Primakoff effects of pions. 14 Chapter 1. Rare Event Searches and TPCs. Several astrophysical and astronomical observations have been interpreted as hints for ALPs. To address the problem of the excessive transparency of the intergalactic medium to very high energy γ(&100 GeV), ALP-photon oscillations have been suggested. For instance as explanation for the observation of high energy γfrom active galactic nuclei[66,67,68] and for the detection of ultra high energy cosmic rays (&1018 GeV) correlated with very distant blazars[69]. The anomalies observed in the white-dwarf (WD) stars’ evolution could be a hint for an additional cooling mechanism via ALP emission[70,71,72]. A more generic category, that includes ALPs, is named WISPs (Weakly Interacting Slim Particles), because of the role they can play in DM problem as an alternative to WIMPs (see just later on). Examples of WISPs are paraphotons[73], minicharged particles[74] and chameleons[75]. The latter are related with quintessence field theories to explain the Dark Energy and, inside particular contexts still in early stages of development, could be classified as ALPs; thus an experiment aiming detection of ALP might carry out a first Particle Physics tests of Dark Energy theories[76]. 1.2.3.1 ALPs and DM. Due to its small mass the present relic from the thermally produced axions in the primordial plasma would contribute to the hot dark matter, so they cannot account for the contribution required by the ΛC(old)DM model. In fact the astronomical constrains to the hot DM contributions impose an upper bound for the axion mass ma<0.72 eV[77], since the thermal production mechanism is more efficient for larger masses. However, another two related mechanisms could produce non-relativistic axions[78,79]. The vacuum realignment relies in the oscillation of the afield around the CP-conserving minimum. At the same time domains with different populations can be created, together with the corresponding borders or topological defects, which can relax producing a new axions population. The impact on the present axion’s density depends on whether the inflation happens before or later than the PQ symmetry breaking[80]. In the second case the afield is homogenized by the inflation and no domains are left, only the particular primordial domain affects to the calculations. The bound imposed for maby the realignment mechanism, this time for accounting the whole DM contribution of the universe, is lower than the bound previously mentioned by the total amount of hot DM. Naturally a prediction for masses higher than µeV is deduced[80]. Smaller masses are not strictly forbidden, but must be argued by anthropic reasons. In the first case the calculation is much more complex, in principle a higher mass would be required, but for both cases axions could account for the cold DM. These mechanisms are even more interesting since could be generic for ALPs 2.2. New generation: MPGDs and micromegas. 21 Figure 2.2: Field configuration in a MSGC. The field lines are distorted from the chamber drift region and the readout pattern design makes them condensate towards the anode. There is a clear dependence of the field on the position. The introduction of micromegas by Ioannis Giomataris meant another stage in the readout evolution because the basic electric field layout is changed. The introduction of a flat electrode over the printed strips (gap ∼25 −130 µm ) changes the typical shape of the amplification field produced by wires, with cylindrical symmetry, to a much simpler near parallel planes geometry. While for the former the field intensity depends on the distance rto the wire as E∼1/r, for the latter a constant amplification field Eamp is produced in all the amplification region (see figure 2.3 and compare with 2.2). That disposition, where a high Eamp acts since the beginning of the avalanche process, reports benefits for the amplification statistics (see section 4.4.2). Another essential advantage is the fact that drift and amplification regions are practically decoupled. Thus there is Edrift that can be optimized regarding the drift properties of the gas while Eamp can be independently chosen so as to get a desired gain during the amplification process. The amplification does not depend on Edrift. It is needed, of course, that the drifted electrons enter the amplification region, so the flat electrode over the anode must be actually a grid, the mesh. The sharp increase of the electric field lines density below the mesh (see figure 2.3) can assure, in most of the situations, the electrons passage through the mesh not to be a practical limitation (see section 4.3). The main technological challenge is the suspension of the thin mesh, typically ∼3–5µm thick, assuring it to be parallel along all the area. Inhomogeneities of the gap would produce gain spatial dependence (though the problem is not so severe as it seems, see section 4.4.1.1), or even sparks in particulary bended zones; the quality of the mesh manufacturing is essential for the same reason. The solution adopted by all micromegas versions is the presence of periodic insulator spacers, the pillars, between the mesh and the anode. In early micromegas the pillars were glued to the anode and electroformed meshes were suspended on them. The electrostatical forces were mainly responsible of the mesh planarity, though peripheral tensors were also needed (as well as considerable skill and expertise). Latter developments[100] found ways to attach the pillars to the mesh. 22 Chapter 2. TPCs and Micromegas. Figure 2.3: Left: the TPC concept with a micromegas readout as it was first presented[94]. The mesh divides the chamber into drift-conversion volume and amplification volume. The primary electrons released by ionization by charged particles are driven to the mesh by a constant electric field, Edrift; the drift velocity is also constant (see section 4.2.1). Once in the amplification gap, where there is an also constant Eamp, the avalanche process occurs. Fast front-end electronics connected to the anode strips register signals which are ideally proportional to the projection of the ionization yield path. The mesh signal can provide a common trigger while the time differences between mesh and anodes signals allow the reconstruction of the drift coordinate. Right: electric field layout in a drift chamber with micromegas. Note both main regions have constant electric field, with only a fast transition in between. The ration between both fields strength, typically Eamp ∼100Edrift, produces a condensation of the electric field lines that makes the mesh near fully transparent. More recent micromegas manufacturing techniques build the complete micromegas amplification structure, anode-gap-mesh, as a unit, so as there is no freedom allowed for the mesh misalignment which relies in the manufacturing process accuracy. The robustness and homogeneity of these new micromegas, bulk and micrombulk types (see next sections), have really boosted their applications, in particular allowing a feasible scaling to larger surfaces. After classical micromegas, bulk and microbulk have been tested in the CAST experiment; a summary of CAST experience with them is reported in section 8.2.1. 2.2.1 Bulk micromegas. The bulk type of micromegas is the most similar one to a commercial micromegas thanks to the use of an actually commercial woven wired mesh (available in several materials: copper, iron, nickel) and a fast, and inexpensive as well, manufacturing process (see figure 2.4 left). Such a mesh (see figure 2.4 right) is stretched and then encapsulated in an isolator material, named vacrel, to lately remove part of the vacrel, remaining the pillars together with additional protection structures as 2.2. New generation: MPGDs and micromegas. 23 border reinforcements or contact protections. The mesh is attached between the pillars, so they emerge over the mesh giving additional protection. This structure, together with the hard woven mesh of 30 µm thickness wires, result in a remarkably robust and flexible micromegas, that is not easily damaged by touching or by sparks, which can be glued to a conventional PCB readout. Figure 2.4: Left: sketch of the bulk manufacturing process[101]. Right: detail of the iron woven mesh showing one of the vacrel pillars. However the bulks performance is limited by the nature of their mesh, which is relatively thick and irregular (presenting nodes). To assure the geometrical desirable condition of the amplification gap being much longer than the mesh thickness, bulks typical gap is 128 µm . That relatively long gap limits the velocity of their signals and they operate at lower amplification fields than other micromegas types. Nevertheless they are able of very high gain and acceptable energy resolution. Bulks performance is shown and compared with microbulks (the other recent micromegas version, see next section) in chapter 4. The fact that they are used together with PCB readouts makes them non convenient from the radiopurity point of view. However it may exist the possibility of using readout bases of a different nature. The pillars, about 300 µm diameter, are relatively spaced, around 2 mm. Therefore the insulator amount of material between electrodes is minimized and bulks have low capacitance per unit area, what helps to build large areas without a limiting noise and there is no need to segment the mesh. Thus large areas are possible, like projected in MAMA (Muon Atlas Micromegas Activity) for the ATLAS detector upgrade for sLHC (LHC with enhanced luminosity)[102]. Bulks are the most popular micromegas type, in application to several experiments including the Rare Event Searches T2K[103] and MIMAC[50]. 24 Chapter 2. TPCs and Micromegas. 2.2.2 Microbulk micromegas. The microbulk type of micromegas[105] was added the prefix micro to emphasize the better accuracy of their manufacture with respect to bulks and the return to very thin meshes, 5µm , and narrower amplification gaps: 50, 25µm . However they could be said to be really bulk micromegas because the mesh is not separately built and attached, like in bulks; in contrast, in microbulks, the raw material is a double clad kapton1foil (copper-kapton-copper) from which the mesh, gap pillars and anode are built as an unit. The amplification structure is produced thanks to the state-of-the-art lithography techniques, particularly the etching of the kapton through the mesh holes to create the amplification volume. Thus the homogeneity of the gap has ∼1µm accuracy, so the energy resolution is outstanding even for a MPGD (see proofs of their performance in 4, particularly figure 4.22 right) and the whole detector is a robust block (see figure 2.5 left). Microbulks are more fragile than bulks to handling and sparks due to their delicate mesh, however more robust against continuous vibrations and tensions, and generally more stable. Figure 2.5: Left: sketch of the typical microbulk prototype (see a photo in figure 4.22 left). Right: microscope view of a CAST microbulk with a pixel-like anode’s pattern. The circles delimit the areas where the kapton has been removed by etching. Because the kapton etching is activated from the mesh printed holes, the amplification volume consists of a series of separated cells below the holes, rather than a common volume with pillars in between (see figure 2.5 right). That increases the capacity per unit area, with respect to bulks, what might mean a limitation for microbulks scaling to larger areas (they can be segmented, nevertheless). However a first experience with a relatively big microbulk is encouraging, see section 5.5.3, particularly figure 5.10. Moreover the manufacturing techniques are still progressing and new microbulk versions with true pillars are currently developed. On the other hand, there could be advantages for the isolated amplifications cells sketch related with geometrical quenching (see section 4.4.1.2). 1Kapton: flexible polyamide with very good insulating properties. 2.2. New generation: MPGDs and micromegas. 25 The microbulk basic structure can be glued to a complex readout plane. One advantage is that the readout layout can be also built by means of the lithography techniques using the same raw materials, therefore the whole microbulk is radiopure (see section 5.4). Readouts as complex as those used for pixel-like 2D CAST microbulk micromegas have been built in this way; the process, from the basic manufacturing of the mesh to the complete detector, is sketched in figure 8.14. Microbulks are the most recent micromegas approach (ca. 2007), their manufacturing is exclusive at CERN and they still need to be better known by the Rare Event Searches community. However microbulk may have the best potential applicability to the field. They are gathering merits in the running CAST experiment (three of its four detectors are microbuks at present, see chapter 8) and also tested by the NEXT-micromegas R&D line[107]. 2.2.3 InGrid. InGrid[108,109] are micromegas-like structures integrated in silicon chips. Thus a charge amplification stage and a high granularity readout are practically undistinguishable. They are descendants of the development line of high granularity pixel readouts based in CMOS technology that were commonly used as complement for MPGDs (usually GEM and micromegas) in small gaseous chambers. The MediPix chip[110] (whose name evidences the fact it was proposed for medical applications), was a matrix of photon counters. But the more recent TimePix[111] has the capability of measuring the arrival time of the avalanches, since it was already conceived to instrument a TPC. In InGrid, wafer post-processing based in SU8 photo-resistor allows the creation of a micromegas-like mesh and pillars over the chip. That technique produces an outstanding accuracy of mesh manufacturing and a very good alignment between the mesh and readout patterns (see photo in figure 2.6). The InGrid fabrication process and performance are studied in detail in [112]. Figure 2.6: Microscopic photograph of a InGrid integrated on a TimePix chip, extracted from [112]. 26 Chapter 2. TPCs and Micromegas. In spite of their impressing tracking capabilities, InGriD detectors still have to demonstrate they can fulfil some of the requirements to be competitive in Rare Event Searches: essentially scaling up and radiopurity. In principle, the fact they are based on Silicon technology is an advantage for any technical challenge and silicon chips are known to be radiopure, hence InGrids may be radiopure themselves by means of an optimized wafer post-processing. The upcoming installation of an InGrid detector in one of the CAST experiment’s detection lines will be a significant step towards the confirmation of InGrids as low-background detectors. 2.2.4 GEMs It is worth to comment on GEMs since they are, perhaps, the most known MPGD. At first sight, they can seem to be very similar to microbulks, since both are built from a double copper-clad kapton foil (50 µm thick) which is made holes in by etching. However, in the case for a GEM, the foil is completely perforated from face to face (see figure 2.7 right). The two copper faces are used as electrodes to produce an intense electric field whose lines concentrate through the holes that work as amplification regions. Figure 2.7: Left: microscope view of a GEM. Right: electric field configuration of a GEM. The main difference with micromegas is the fact that there is no a clearly defined anode, therefore, despite about 50% of the avalanche electrons being collected by the electrodes, the induced signal is more bipolar (see section 4.5). The electric field configuration is a bit less regular, since they are two transition zones. One advantage of micromegas is how they are naturally integrated with a readout, as a single and robust unit, with correlation between mesh and anode designs, in particular in the case for the microbulks. However the usage of GEMs as a floating pre-amplification stage over an independent readout brings a versatility that has made them very successful. They can work together with PCB readouts, CCDs or other MPGDs, typically MSGCs and micromegas. The idea can be generalized by the implementation of double or triple GEM (see figure 2.8) for higher amplification gains. 2.2. New generation: MPGDs and micromegas. 27 Figure 2.8: Sketch of a drift chamber operating a triple GEM[113]. The usage of a GEM cascade degrades the spatial resolution and implies construction difficulties; an alternative to avoid these problems can be the introduction of a thicker GEM (THGEM[114]) with 400 µm gap. Chapter 3 Interactions of particles in the micromegas chamber. Contents 3.1 Ionization by charged particles. ................. 30 3.1.1 Coulomb scattering. ....................... 30 3.1.2 Mean energy loss. Bethe-Bloch formula. ............ 30 3.1.3 Applicability of the Bethe-Bloch formula and thin absorbers. 33 3.1.4 δ-rays and restricted energy loss. ................ 33 3.1.5 Stopping Power of non relativistic ions. ............ 38 3.2 Other indirect mechanisms to generate ionization ...... 45 3.2.1 Interactions of photons. ..................... 46 3.2.2 Interactions of neutrons and WIMPs. ............. 50 3.2.3 Bremsstrahlung. ......................... 52 This chapter, and the next one, aim to be a brief report on the processes involved in the detection of particles by gaseous chambers equipped with micromegas. It is not a Physics lesson, not complete nor deep, but only focused in the good understanding of this kind of detectors’ phenomenology; the cases, always of direct interest for Rare Event Searches, are discussed in the context of a micromegas detector. The processes are described in chronological order, from the interaction of the particle to the amplification in the micromegas. Own research results are included, sometimes for illustration because the familiarization with this phenomenology, the creation of references, has been a goal of this work; others because our R&D lines were devoted to processes which are specific of micromegas or some particular cases interesting towards Rare Event Searches applications. This chapter treats the physical absorption of energy from particles and radiation by the gas which triggers the detection mechanism when they interact in the chamber’s sensitive volume. We will also pay some attention to processes which occur in the detector neighborhood, but can produce a secondary response of the detector. All the signal generated in the micromegas readout is finally due to the production of free charges by ionization by a charged particle (with exception of chambers equipped with light sensors); therefore this is the fundamental process we are going to start with and study in more detail. 30 Chapter 3. Interactions of particles in the micromegas chamber. 3.1 Ionization by charged particles. 3.1.1 Coulomb scattering. The ionization differential cross section is a previous result to obtain the ionization yield, however it is interesting by itself. It is important to define the physical picture and assumptions. The electron, the target which is susceptible to be hit resulting in ionization of the medium, is supposed to be free and at rest; thus no dependence on the medium can be expected from this scope. The interaction between the ionizing particle and the electron is purely Coulombian and the mass of the former makes the electron’s one negligible (therefore the particle mass will not appear explicitly in the result). That is practically the famous Rutherford’s problem, with just opposite projectile and target, and its deduction can be found in text books (semi-classical calculation by Bohr in [115] or by means of Fermi and Yukawa theories in [116]). When related with T, the energy lost by the particle and transferred to the electron in an individual Coulomb collision, it looks like dσ dT=2πz2α2 mec2β2 1 T2(3.1) with zthe net charge of the particle and β=v/c its relative velocity; being α= e2 4πε0~≈1 137 the fine-structure constant and methe electron mass. The expression (3.1) is valid while the ionizing particle, much heavier than an electron, is fast enough so as the electron can be considered to be at rest, and the energy transfer high enough to consider the electron to be free. The relevant remark is that dσ dT∝1/T2 what means a large tail in the dependence which assures that relatively high energy transfers are not so unlikely. 3.1.2 Mean energy loss. Bethe-Bloch formula. The mean energy lost in ionization when the particle passes through a volume element of material, named electronic stopping power, is computed from (3.1) by weighting all the possible energy transfers: dE dx=dN dxZTmax I dσ dTTdT(3.2) where the medium, with atomic and mass numbers Zand Aand mass density ρis described as a homogeneous density of free electrons at rest dN/ dx=NA(Z/A)ρ(3.3) being NAthe Avogadro’s number; and the integral is extended from I, a sort of mean ionization potential, and, Tmax, the maximum energy transfer allowed by the collision kinematics. Applying that the ionizing particle mass Mme: 3.1. Ionization by charged particles. 37 but we consider 2 cm projected range to estimate the case of xenon, were the track is shorter, therefore Emax = 250 keV is chosen using (3.12). It is shown in figure 3.5 how the implementation of a practical cut has a dramatic effect on the stopping power relative to the Bethe-Bloch formula, however the behavior is soft with Emax. In particular the relativistic rise is strongly moderated. The parametrization by ESTAR is in between our two examples of restricted energy, and the difference at high energies is due to the implementation of the density effect. In any case the increase is below the 50% compared to the minimum along a very wide range. The modified Bethe-Bloch formula with restricted energy loss by application of a custom-made Emax (3.9) seems to be a simple and understandable formula that allows a good description of the signal origin in the ionization chamber along a wide range of energies in most of the cases of interest for us. Even though it still has the disadvantages discussed in the previous section which are related with the mean adjective. Figure 3.6: Stopping power of positive muons on copper along 12 orders of magnitude in energy. Vertical bands indicate boundaries between different approximations. Extracted from [118]. The complete picture is of course more complicated (see figure 3.6). At higher energies specially the lightest particles like electrons lose energy mainly by radiation yield, which in principle has no effect on the chamber, but can generate indirectly signals (see section 3.2.3). On the other hand, shell-corrections must be included to attend to atomic binding energies to assure good accuracy in the low energy region. 38 Chapter 3. Interactions of particles in the micromegas chamber. In addition, it is clear that when the energy is very low the ionization should also decrease. Although the Bethe-Bloch formula produces a fall at very low β, the actual effects that produce the decrease have not been contemplated and a new approach must be developed. The Bethe-Bloch based formulas explained here are valid only down to β > 0.05. The low energy ions which are of interest for dark matter searches are deeply below this threshold, and therefore their stopping power must be discussed in a specific section. Regarding the distribution of energy depositions, The Landau’s approach can be a good estimator of the most probable value, but fails in the distribution shape description in a TPC, especially at energies of few keV, comparable with medium binding energies. A Monte Carlo simulation should be carried out, what is even more necessary if we are interested in the distribution of charge along the particle path. 3.1.5 Stopping Power of non relativistic ions. The boundary β≈0.05, pointed out as the condition for usage of Bethe-Blochbased stopping power expressions, can be expressed as ∼1MeV/nucleon for ions, what is far from the typical nuclear recoil energies involved in dark matter searches. Two factors make the stopping power more complicated below this mark: the loss of efficiency in ionizing the medium and the inclusion of a new component to the stopping power related with strong interactions with the medium nuclei of increasing contribution as the energy decreases dE dx=dE dxelectronic +dE dxnuclear (3.13) The energy lost in nuclear interactions is not directly involved in the ionization of the gas, but these elastic atom-atom collisions can produce secondary ions able to ionize by themselves (analogue to δ-electrons). The classical approach [125] is again Rutherford scattering in the screening field of the atoms, therefore the obtained stopping power looks very similar to a re-escaled Bethe-Bloch curve (see, for example, figure 3.8 left). The electronic contribution is well characterized by the non relativistic approach developed by Lindhard et al. (LSS-Theory) [125,126] where the media is described as a free electron gas whose interactions with the incident ion can be considered as small perturbations. Then, as a projectile moving in a viscous medium4, in the LSS-Theory the energy loss is found to be proportional to the velocity: dE dxLSS ∝vz2(3.14) Besides the dependency on the medium to be much more complicated than the 4This is why ballistic jargon is very extended in this issue. 3.1. Ionization by charged particles. 39 typical Z/A in Bethe-Bloch approach, the major difficulty here is how to define an effective ion charge zfor nuclei which are moving with velocities similar to those of the orbiting electrons. As the atom kinetic energy is increased it is stripped of its orbital electrons but, at the same time, the medium is polarized resulting in a screening of the ion charge5; both effects are reciprocally influenced and the result is a complicate dependency on the ion, the medium and the velocity. The LSS-theory is valid below β≈0.01 ≈50 keV/nucleon (see figure 3.6 and 3.7 right). For the range 0.01 .β.0.05 there is not an accepted comprehensible statistical theory. The use of more complicated models for atoms (both as incident particle and as medium) forced the use of computers (or, alternatively, computers brought the opportunity to try them) in the late 70’s[126]. The approach by Ziegler et al.[127] implemented in the free accessed, but not open source, SRIM (Stopping Range of Ions in Matter) software[128] has became the most similar to a paradigm. Figure 3.7: Left: Stopping power for He ions in He gas from SRIM tables. The nuclear stopping power has been amplified ten times. The lateral straggling is also represented for ions with this initial energy. Right: electronic stopping power as function of velocity for the same energy range. Note the linear regime around 0.01 .βpredicted by LSS-theories. Although SRIM was initially thought for ion transport in solids (for the study of ion implantation and similar applications), in our energy range of interest the atoms can be considered as free an SRIM is widely used by the gaseous TPC community (however, there is less experimental verification). Present version of SRIM produces tables for mean values of electronic and nuclear stopping ranges, also the projected range of the ion path and their longitudinal and lateral straggling. SRIM can also generate individual ion paths using Monte Carlo based on these tables, including cascades of secondary recoils caused by ion-atom collision, though only the primary path is described with detail registering, step by step, all the partial energy losses. The plots in figures 3.7 and 3.8, from SRIM tables, are used to illustrate the nature 5This is not exclusive of low energy ions, at higher energies difference between antiparticles stopping powers is known as the Barkas effect. An example of that is represented for µ+and µ− in figure 3.6 40 Chapter 3. Interactions of particles in the micromegas chamber. of the stopping power in this context. Similar plots can be found in [129] for C, F and S recoils in CF4and CS2gases, but here we will focuses in noble gases. In figure 3.7 αparticles (He ions in this context) are used to show the electronic stopping range at low energies. For such light ions moving in light media the electronic contributions dominates during practically all the energy range. It is shown how the stopping power has a transition from the Bethe-Bloch application range to the LSS regime known as the Bragg peak which explains totally the shape of the curve if figure 3.1. We can detect that SRIM adds some small corrections to Lindhard simple model at very low energies which are more evident for heavier ions. The straggling is strongly related with the nuclear stopping power (hence for He ions is always small), as collisions with particle of the same, or similar, masses can sharply change the ion-projectile momentum. In particular, the lateral straggling (final deviation orthogonal to the initial direction) means a fundamental difficulty for directional detectors. Plots for argon and xenon in figure 3.8 show that for medium and high ion masses the electric stopping power is always inside the LSS range and how the nuclear contribution enhances as the ion and the medium atoms masses increase. While for argon we can distinguish two stages where each contribution dominates, for xenon the nuclear stopping power is always the main effect. Like in the helium case, the straggling is more important where nuclear stopping power is higher and peaks up when electronic stopping power losses influence. Figure 3.8: Left: Stopping power and lateral straggling for argon ions in argon gas from SRIM tables. Right: the same for Xe ions in Xe gas. The electronic stopping power has been amplified ten times. In contrast to electrons, the projected range of low energy ions is approximately linear with energy. In figure 3.9, it is shown that this assertion is more accurate for medium size masses and fails at very low energies (.20 keV). In general, the lighter the ion and the medium atomic masses, the longer the range. The range, which scales with the inverse of the gas pressure, can also be stretched by addition of lighter atoms (figure 3.10), and the straggling is also improved this way, but very 3.1. Ionization by charged particles. 41 softly. Figure 3.9: Projected range in noble gases as function of the initial ion recoil energies, from SRIM tables. The line is a linear fit attempt. Figure 3.10: The line is a linear fit attempt. Right: projected range for Ar-He mixtures for several initial energies, from SRIM tables. The line is a linear fit attempt. The detail corresponds to the lateral straggling for 100 keV ions paths. 3.1.5.1 Quenching factor. As a result of the effect of a non negligible nuclear stopping power, low energy ions are not as efficient in ionizing the medium as electrons. The heavy ion can keep a significant amount of energy, though being unable to ionize the medium [130]. The primary ion produces a certain number of secondary recoils in a cascade process, each one leads to a loss of an appreciable amount of residual energy. Theoretically the quenching factor,Q, as it was introduced by Lindhard, is the 42 Chapter 3. Interactions of particles in the micromegas chamber. ratio between the energy given to the electronic excitation and the total energy. This is an experimental key parameter that is usually to be defined as Q≡Er Ee ;Eee ≡QEr(3.15) where Erand Eeare the measurements of the energy done by the detector for a nuclear recoil and an electron of the same energy, since for electrons practically all the energy is lost in ionization6. Since electrons (basically photoelectrons from X-ray or γsources) are commonly used to calibrate the detectors the electron equivalent energy,Eee, is defined for nuclear recoils as the energy of an electron that would give the same measurement of the energy. As Q < 1, this is a compromising parameter for Dark Matter detectors sensibility, because the recoils energy range of interest is invaded by lower energy (and so, surely, much many) γ. Lindhard et al.[125] already gave a parametrization for the case of recoils of medium own atoms7: Q=g(E) 1 + g(E)(3.16) and g(E)is approximated in the range from 10 keV to 1 MeV by: g(E)=0.66 Z5/18 A1/2E(keV)1/6(3.17) F.J. Iguaz wrote a method to get a detailed description of the ion cascade built from individual nuclear recoils simulated with SRIM for the RESTSoft framework[133] (see chapter 6). An energy threshold is defined to decide when the energy loss by nuclear stopping power yielded by a simulation step motivates the simulation of a secondary recoil. This threshold also defines the residual energy lost per recoil. We will try a validation of the method by a direct comparison between the simulation results and 3.16 and 3.17. Figure 3.11 shows the simulation of 200 argon recoils in argon per initial energy using 100 eV as residual-threshold energy. The primary charge is generated only from electronic energy losses, using NIPs =Eelec/26.4eV in average (see chapter 6). The comparison with the NIPs expected from the initial ion energy gives the simulated quenching factor. The results for Ar ions in Ar gas are plotted in figure 3.12 where it is shown how the simulation approaches Lindhard calculation only if secondary ions are added to the simulation. Secondary recoils are responsible of at least one half of the charge yield, so they should be taken into account for a detailed study of charge 6Here, we refer in general to the measurement of energy instead of talking about primary charge yield because the same concept applies for detectors based on light collection or semiconductors. Therefore, Eris the visible energy. 7The case of different ion and medium is discussed in [131] for directional detector purposes, and experimental data can be found in [132]. 3.1. Ionization by charged particles. 43 Figure 3.11: Charge yield for argon ions in argon obtained from SRIM plus RESTSoft ion cascade builder. The bars represent the distribution as FWHM. The little plot summarized the energy resolution as FWHM/E. generation along the recoil path. Regarding the theory-simulation comparison, we do not have a detailed study on energy residual-threshold dependency (which is not a well understood parameter[130]). On the other hand an obvious limitation of our cascade reconstruction method is that all the partial energy loss by nuclear stopping power does not have to be interpreted as produced by a single collision, and so, invested in only one secondary recoil. Figure 3.12: Quenching factor for Ar in Ar obtained from SRIM simulations, with and without addition of secondary recoils contribution, compared with 3.16. 44 Chapter 3. Interactions of particles in the micromegas chamber. Another remarkable point from these simulations is the very poor energy resolution visible in figure 3.11 which is consequence of the dispersion in the partition of the energy losses between recoils residual energy and electronic excitation. With the 100 eV residual energy-threshold up to 200 secondary recoils are simulated for each primary ion of 100 keV, with very big variations on this number. The asymptotic value of ∼30% FWHM is in agreement with results from [134] for F ions in CF4, even though secondary recoils are not used in those simulations. These results point to another fundamental difficulty for Dark Matter searches, despite it is not easy to be experimentally proven. 3.1.5.2 Head-tail effect. The head-tail effect is a common way of speaking that makes reference to the possibility of an asymmetry in the Bragg curve, like the one clearly shown by α-particles in figure 3.1, which would allow to distinguish the head (end of the track) from the tail (beginning of the track). In the case of α-particles and electrons, the head has much more intense ionization since these particles move on the right side of the Bragg peak (see figures 3.5 and 3.7). In contrast, for low energy ions we would expect the tail to dominate over the head, as they are situated on the left side of the Bragg peak. Note this consideration has not taken into account the effect of nuclear stopping power, however more deep discussion [131] leads to the same conclusion (only from mean values). Moreover the effect has been observed by the DMTPC collaboration using 14.1 MeV neutrons from a deuteron-triton generator[135] and also with a 252Cf source[136] for F recoils in the range from 200 to 800 keV. In [129], stopping power curves using SRIM for C, F and S in CF4and CS2are plotted. The most interesting, F and S, as medium size atoms are basically very similar to Ar (figure 3.8 left); there is a low energy stage, dominated by the nuclear stopping power. While a high energy stage, where Lindhard model for electronic stopping power is a good approximation, can be distinguished. In the high energy stage a tail dominance is expected. The DMTPC would have seen the effect because their F recoils have a long part of their paths along which the electronic stopping power is clearly dominant. In our preliminary study on argon recoils below 100 keV at 1 bar8, we proposed[133] the volumetric density of the charge as discriminator because simple Bragg-like curves often give non conclusive results. We detected that limitation is due to the straggling of the main path plus the presence of secondary recoils; since both are more common near the head, a higher charge yield around the end of the path can be produced. However, this additional charge is expected to be more disperse. The algorithm is illustrated in figure 3.13 for a complete simulation (including secondary recoils) of an argon recoil. Even though the success rate was promising, such a method would need for a really powerful topological capabilities 8The simulated path is scalable with the inverse of the pressure and so conclusions based on the path features should be universal, with the exception of drift diffusion effects. 3.2. Other indirect mechanisms to generate ionization 45 of the chamber. Figure 3.13: Illustration of a sense recognition algorithm over a 40 keV Ar ion cascade simulated with RESTSoft (based on individual SRIM recoil simulations). After the determination of the recoil direction, two volumetric regions (only readout projection is plot) are defined assuming that the head corresponds with the lowest density of charge. In the paper previously mentioned [129], following a simulation extension of SRIM very similar to the RESTSoft approach (see section 6.3.3) the DRIFT researchers admit the importance of the straggling caused by the nuclear collisions (despite neglecting secondary recoils). The straggling together with the necessity to project the ionization yield onto the readout axes, means a factor that competes with the pure Bragg curve to enhance the strength of the head respect to the tail. Depending on the ion and its initial energy the Bragg-like curves (projected onto the readout axes) can be basically flat or inclined to the head or to the tail (in agreement with our studies[133]). After studying simulations of the same nature[134], researchers from the MIMAC collaboration suggest not to consider the head-tail effect for Dark Matter searches with recoils below 100 keV. 3.2 Other indirect mechanisms to generate ionization Fast charged particles will always be the source of the signal readable in the micromegas chamber. Therefeore, at the end of every of the following subsections, we will be addressed to the previous section. However the following physical processes will enable the TPC to detect other kind of particles. 46 Chapter 3. Interactions of particles in the micromegas chamber. 3.2.1 Interactions of photons. In contrast to the ionizing particles (at least macroscopically), the photons interact with matter in a discrete way that suddenly changes the photon state and can even be completely absorbed. The stopping power concept is replaced by the differential cross section which describes the possibility of such a process to happen. The following expression illustrates the way that applies for a beam of photons with initial intensity I0crossing a thickness xof material with Nmolecules per unit volume, and it will be useful to show several common definitions related with the same concept: I(x) = I0e−σNx =I0e−µX =I0e−x/λ (3.18) where X=ρx is the mass thickness with ρthe mass density and µ=σN/ρ is the (frequently tabled) mass attenuation coefficient, and λ= 1/(ρµ)is the mean free path of the photons in the material. The mass attenuation coefficient is probably the most useful construction because it is just summed for different processes or simply composed for a material mixture: µmix: 1,2... , n =p1µ1+p2µ2+. . . +pnµn(3.19) with p1, p2µ2. . . , pnthe percentages, in weigth, of the 1,2, . .., ncompounds of the mixture. Photons can interact with matter through four processes: •photo-electric effect: absorption of the photon by the atom. •Compton effect: scattering of the photon by one shell electron. Also called Thompson scattering at very low energies. •pair creation: creation of an electron-positron pair. •Rayleigh scattering, or coherent scattering: scattering of the photon by the whole atom. All these effects are exposed in reference books as [115,137] and here only the ideas of direct interest will be remarked. Pair creation, as a typical process in High Energy Physics (see figure 3.14) will not be discussed. Tables for mass attenuation coefficients can be found in the XCOM database[138] from NIST. 3.2.1.1 Photo-electric effect (Fluorescence, escape peaks and Auger electrons). The photo-electric effect is the process that determines the quantum efficiency of the chamber for X-rays since is the one which can easily produce ionization yield 3.2. Other indirect mechanisms to generate ionization 53 The Bremsstrahlung (braking radiation) is experimented by charged particles when they are scattered by the electric field of the medium nuclei. That mechanism of radiative energy loss is in competition with the ionization produced by scattering with the electrons of the medium that was studied in the first sections of this chapter. dE dx=dE dxion +dE dxrad (3.28) The radiative stopping power is proportional to the particle energy (see, for example sections 11.5.2 and 11.5.3 of [145]) dE dxrad ≈ − E X0 (3.29) The last equation justifies the name of radiation length given to X0since, in average, E=E0exp(−x/X0). The radiation length dependency with medium (Z, A)and particle mass, m, is approximately (calculus and tables can be found in [146]) 1 X0≈Z2 A 1 m2(3.30) Then bremsstrahlung is more efficient for light particles that can suffer strong decelerations which act as photon shots. Since it is proportional to the particle energy, while ionization tends to saturate, it becomes dominant at high energies (see figure 3.6) and finally the energy loss by bremsstrahlung rises over the ionization stopping power at the critical energy,Ec(see figure 3.6). For our purposes, bremsstrahlung is only relevant for cosmic (high energy) muons or for electrons around the MeV. The following approximations (see [118] and references there in) are valid only for electrons: Ec=610 MeV Z+ 1.24 (solids and liquids); (3.31) Ec=700 MeV Z+ 0.92 (gases); (3.32) X0=716g/cm2·A Z(Z+ 1) ln 287/√Z(3.33) Even still far from their critical energy, electron’s radiation length in a xenon TPC at 10 bars is ∼1.5 m and so, along the few tens of cm paths of electrons between 1 or 2 MeV, it is not rare that some photons are shoe. They can generate secondary clusters or escape from the chamber with consequent danger for calorimetric detectors. On the other hand, the possibility is higher in metals, in particular in lead, and therefore relatively high energy electrons originated by natural radioactivity or cosmic muons can induce low energy events in the chamber by bremsstrahlung experimented in the detector’s shielding. Chapter 4 Working principles of the TPC-micromegas. Contents 4.1 Primary Charge generation. ................... 56 4.1.1 W, averaged energy per ion pair. ................ 56 4.1.2 F, the Fano factor. ........................ 58 4.2 Drift, diffusion and attachment. ................. 58 4.2.1 Drift velocity. ........................... 60 4.2.2 Diffusion. ............................. 63 4.2.3 Attachment and recombination. ................. 65 4.3 Mesh transparency. ........................ 66 4.3.1 Parametrization of the typical transparency fall. ....... 69 4.3.2 Limits for the mesh design. ................... 71 4.3.3 Transparency quench. ...................... 72 4.3.4 Transparency and diffusion. ................... 73 4.4 Amplification: gain and energy resolution. .......... 75 4.4.1 The Townsend coefficient and the micromegas gain. ..... 75 4.4.2 Avalanche fluctuations and energy resolution. ......... 78 4.4.3 Experimental examples: gain curves and energy resolution. . 82 4.5 Signal induction. The Ramo’s Theorem. ........... 90 This chapter is devoted to the chain of processes that happen in the detector, which are triggered by the physical interactions described in the previous chapter and yield an electronic signal, and how they transform the original information. Since there exist several manuals that explains the working of a TPC, we will focus in micromegas particularities, and the discussion will be more detailed about the features that are being directly studied by the T-REX project. The next chapter is an special extension in this sense, dealing with no essential aspects, but which are critical for most or some applications of TPCs to rare event searches. 56 Chapter 4. Working principles of the TPC-micromegas. 4.1 Primary Charge generation. The creation of free electrons by the interaction of the gas atoms or molecules with the ionizing particles is the essential process that allows the working of the gaseous detectors. This process has, in principle, several dependences: on the type of gas molecule, on the type of the ionizing particle, on the strength of the particular energy deposition. The relationship between these issues brings diverse possibilities for the process end which, in principle, produce big statistical fluctuations. Fortunately the nature moderates these dependences and limits the fluctuations for the benefit of the gaseous detectors’ potential as proportional counters. There are two main parameters, empirically defined, that quantify the generation of the primary charge, which are the scope of this section. 4.1.1 W, averaged energy per ion pair. The original interactions that trigger the detection process are the inelastic collisions produced by an ionizing particle that were studied in section 3.1. These discrete energy deposits can ionize the gas molecules, but also excite higher energy electronic states (radiative excitations) or vibrational or rotational states1. Sometimes the energy invested in excitation can be reconverted into ionization (see section 3.2.1.1). Besides, even if the collision produces effective ionization, the free electron can retain a certain amount of energy and being unable to ionize again the gas. The ionization yield must be characterized by an empirical parameter that relates the total production of primary charges, ne, with the energy deposited in the active volume of the chamber E02 W≡E0/ne(4.1) Wis the average energy required to produce an electron-ion pair, often referred simply as the W-value. As a rule: Eexc (first energy of excitation) < Ei(first energy of ionization) < W. Recompilations of these values can be found in [147] and [115]. Roughly speaking Wis about 1.5–2 times Eifor noble gases, and 2–3 times Eifor more complex molecules. We will just remark WAr = 26.3eV and WXe = 21.9eV. A crucial result for the operation of gaseous detectors is the fact that Wis only slightly decreasing with the energy of the ionizing particle. Only when energies are of the same order than the ionization and excitation potentials Wrises[148]. 1Precisely the kind of gases used in the gaseous chambers usually belong to one of the two extreme cases: the simplest molecules, noble gases, unable of vibrational or rotational modes; or complex organic molecules with numerous and energetic excitation modes. In most of cases a mixture of both types is used because of reasons that will be shown along the chapter. 2The absolutely empirical character of the W-value causes some confusion about its exact definition. For practical reasons E0is often the initial total energy of the ionizing particle, what can produce a small overestimation of Wbecause of its rise at very low energies. 4.1. Primary Charge generation. 57 The dependence is also weak with the radiation type. In figure 4.1 the charge yield obtained in a chamber with micromegas is directly compared for 13.9 keV Xrays and 5.4 MeV α. Although the micromegas gain is plotted, it has been extracted from a comparison of the signal strength which already assumes Wα=WXray (see 4.4). These measurements were conceived to study other processes as well, and they would be revisited and discussed with more detail later; for the moment it is clear that the calculated gains could not match unless Wα≃WXray, even though <dE/ dx > is more than one order of magnitude higher for the α. Figure 4.1: Gains deduced for a micromegas in Ar with CH4for several CH4concentrations and pressures, using the set-up described in section 5.1. The closed markers (circle, squares and triangles) are calculated from 241Am 5.4 αwhile asterisks and crosses from 241Am 13.9 keV X-rays at the same time, i.e. during an only measurement run, using exactly the same gas and set-up conditions. The highest point of the series is the last one before the micromegas starts continuously sparking. The W-value for a gas mixture is given by a simple weighted average W= Y k Wk!/X k CkσkY i6=k Wi (4.2) with Ckthe concentration of each component and σkthe relative cross sections (Pσk= 1). Therefore no spectacular effect is expected from the addition of a new component in small concentrations. The rule 4.2 is not valid when there is an interplay between two components (see the case of Penning mixtures in section 5.3). 58 Chapter 4. Working principles of the TPC-micromegas. 4.1.2 F, the Fano factor. If the generation of primary charges followed Poisson Statistics, the variance of the charge yield for events that average newould be σ2=ne. Then that would imply an essential limitation for the energy resolution of the detector because of the variation of the primary ionization, which would be expressed as full width at half maximum (FWHM): R(FWHM)|pi = 2.35σ1 ne = 2.35r1 ne = 2.35rW E0 (4.3) that produces R|pi@Ar ≃15.6% FWHM for 6 keV X-rays in argon. Microbulk micromegas actually achieve 11% FWHM for the 55Fe peak3(see figure 4.22 right), an experimental value certainly affected by other additional variance sources (see section 4.4.2). Facing the evidence that the process is not purely Poissonian (besides the ionizing particle has a finite energy that imposes a maximum ne, the excitation and ionization possibilities are not independent) the actual variance is corrected using the Fanofactor, whose empirical definition could be: the Fso as to fit σ2=Fne(4.4) thus R(FWHM)|pi = 2.35rFW E0 (4.5) what is known as the Fano’s limit and yields R|pi@Ar ≃7% and R|pi@Xe ≃5% FWHM for argon (F= 0.17) and xenon (F= 0.13) respectively at 6 keV. A good compilation of Fvalues is done in [112]. It seems that Ffollows a composing rule equivalent to (4.2), and so small amounts of additives cannot bring impressive improvements, again with the possible exception of the Penning mixtures (section 5.3). 4.2 Drift, diffusion and attachment. The drift of the electrons along the chamber to the readout plane is obviously a crucial process for a TPC. The deviation suffered by the electrons from the projection of its initial position to the readout plane and the expected arrival time produces a diffusion of the topological information that affects the TPC imaging capabilities. 3That has been obtained with argon plus an small amount (2–5%) of isobutane, however the primary charge generation is a process where atoms basically can be considered independent (see equation (4.2), but keeping in mind the exception of the Penning mixtures), then the primary charge generation statistics are dominated by the interaction with the argon atoms. 4.2. Drift, diffusion and attachment. 59 The drift time limits its application to high rates scenarios and a slow drift velocity can produce undesired secondary effects. In spite of being continuously accelerated by the action of the drift field ~ E, the electrons get an steady drift velocity ~vd, which manifests macroscopically. That reminds the classical case of the motion with friction proportional to the velocity me d~vd dt=e~ E−K ~vd(4.6) with meand ethe electron mass and charge4. Note that the equation (4.6) suggests the definition of a characteristic time τ=m/K for reaching the steady state. From a microscopic point of view, from which the electrons have an instantaneous velocity ~vidifferent than ~vd, the macroscopic characteristic time is justified by the average time between collisions of the electrons with the gas molecules, 1/τ =σNvi, with N the molecules density of the gas and σthe cross section for interactions producing an effective momentum transfer between electrons and molecules. The following qualitative picture may help to construct comprehensive explanations for most of the formulas and plots recompiled in the section, though some of them can seem to go against intuition. Firstly, it must be noticed that the direction of ~vican change completely after a collision (in simple models the scattering is isotropic), due to the mass ratio between electrons and molecules. Besides, for elastic collisions the fraction of energy loss per collision, Λ, is of the order of 10−4. So, if only elastic collisions are frequent, we have basically a random walk affected by a continuously interrupted tendency to drift in the direction of ~ E. Under these circumstances it is understandable that the possibility of ~vibeing eventually reduced by a big energy loss could bring a benefit, since it would work like a kind of reset of the instantaneous velocity. Thus the deviations caused by the scattering are minimized and the electric field energy is invested in driving the charges only in the drift direction. The inelastic collisions causing high Λcan partially materialize such a reset. However, an excessive shortening of the electron’s mean free path is also negative for the drift velocity. In addition, each collision means a possibility of the electron being attached by the molecule. The interplay between both effects together with the cross sections that define the interactions between electrons and molecules explain the drift properties of the gases. There are two extreme cases for gases regarding the way they interact with free electrons5, both are often chosen for gaseous chambers. On one hand there are the noble gases, with the simplest molecular structures, inelastic collisions at low energies with electrons are not possible and they are very unlikely to absorb 4Equation (4.6), was first introduced for electrons transport by P. Langevin. With the addition of a magnetic force is known as the Langevin formula. 5Molecules with high electron affinity that tend to attach the electron are not considered here. 60 Chapter 4. Working principles of the TPC-micromegas. an electron and have low cross sections allowing long mean free paths. On the other hand there are molecules with complex structures, like organic vapours, whose rotational or vibrational modes can be activated at energies much lower than the electronic excitation energies, what requires a relatively high energy loss for the electron. Those possibilities also increase the cross section. The most common choice is a gas mixture based in a noble gas with addition of a, usually not too big, amount of a gas of the second group, called the quencher6. The mixture’s drift properties can be optimized depending on the composition and the applied drift field. There is a rigorous theory of the electrons transport in gases, which can calculate the velocities’ distribution, etc; it was just revised during the development of the first TPCs[149]. Usually the Monte Carlo approach is more popular, being the Magboltz code[150] the most common. We will just comment general tendencies and cases of particular interest. 4.2.1 Drift velocity. If the thermal energy is negligible, i.e. (1/2)mv2 i(3/2)kT, as in the case of the operational TPCs, the drift velocity as well as the mean instantaneous velocity can be accounted from the energy balance along the drift. The manifestation of a steady (macro picture) or a well defined mean (micro picture) drift velocity entails taht the energy won from the drift field equals the energy lost from collision to collision in average. The resulting velocities are function of the collision properties. The following expressions are due to Townsend[151] from a simple approach which can be also found in [116]: v2 d=eE meNσ(ε)rΛ(ε) 2(4.7) v2 i=eE meNσ(ε)s2 Λ(ε)(4.8) The shape of the velocities’ functions have been basically transferred to the dependence of Λ(ε)and σ(ε)with ε= (1/2)mv2 i;Nbeing the atomic density. As actually ε≃Eλ ∼v2 i, with λ∼N−1the average path drifted between collisions, both velocities are function of the ratio E/N, or E/P (Pthe gas pressure), called the reduced field, which is conveniently used to parametrize the drift properties. Note the opposite dependence in equations (4.7) and (4.8) on Λ; that is the manifestation of the reset effect of the collisions with big losses we referred to in the section’s introduction. 6This denomination more exactly comes from the role the quencher plays in the amplification process (see section 4.4), however in a TPC the gas must be selected regarding as much to the gain as to the drift properties. 4.2. Drift, diffusion and attachment. 61 From the drift velocity point of view, the optimum case is a gas that allows a long mean free path but causes very inelastic collisions. Thus the electrons are able to reach high velocities and are only driven by the drift field with small deviations. These are often conflicting features for a simple gas, and that is the reason why mixtures are frequently used. The average time between collisions (or the effective cross section) and the mean fraction of energy loss per collision can be constructed from the components following direct composing rules. The huge differences in the cross sections and the impact of a single inelastic collision implies that, in contrast to the W-value, impressive effects are produced by addition of small fractions of quencher gases. How it works is illustrated in figure 4.2 for argon gas and two typical quenchers. The shape of the drift curves can be justified in basis to the collision properties for the involved gases plotted in figure 4.3. Figure 4.2: Drift velocities calculated with Magboltz. Note the dependence on the E/P ratio, explicitly written in the axis; it was exactly reproduced in the Monte Carlo results. Although the pure argon gas has the lowest cross sections, the strategy of having only elastic collisions does not work satisfactorily because of the reason advanced in the section’s introduction: the dispersion of the direction of the electron’s velocity in each collision. The cross section has a dip due to a quantum interaction between the electron and the gas molecules wave lengths named Ramsauer effect, that is related with a maximum in the drift velocity, but it is placed at very low drift fields. The opposite strategy, that of the quencher here represented by the CF4gas7, certainly yields higher drift velocities, however a high Eis needed to compensate the short mean free path. 7Many times called freon. 62 Chapter 4. Working principles of the TPC-micromegas. Figure 4.3: Cross sections for effective momentum transfer of electron with Ar, isobutane and CF4, extracted from [150]. The solid line, which is continuous along the energy range and shows the Ramsauer dip around 0.1 eV, corresponds to the elastic cross section. The rest of solid lines are vibrational modes. The short dashed lines curves, which raise about 10 eV, are excitation cross sections. The long dashed ones, which are activated only with slightly higher energies, corresponds to the ionization cross sections. The last plot, taken from [116] shows a comparison of the inelasticity of the collisions of Ar and methane. When a mixture of gases that follow the opposite strategies is tried, the interplay between the mean free path and the efficiency of the reset mechanism makes the drift velocity curve develop a peak whose strength and position depends on the mixture composition: the higher the quencher fraction, the faster the gas can become, but stronger drift fields are needed. The effect of isobutane, iC4H10, in the drift properties is also plotted and compared with CF4. The isobutane is more appreciated for its benefits of the amplification properties, but its addition produces also an important improvement of the pure noble gases drift features. However, figure 4.3 shows that, in contrast to CF4, any amount of isobutane cannot make the inelastic collisions dominant, since the elastic cross section is also very high for this organic molecule. As a consequence, even with higher fractions of isobutane, drift velocities higher than ≃5cm/ µs cannot be surpassed in argon-isobutane mixtues. The butane, C4H10, whose Λ(ε)is plotted in figure 4.3 to illustrate generically the difference between a noble gas and 4.3. Mesh transparency. 69 The consequences of the loss of electronic collection efficiency go beyond a soft degradation of the signal strength. A limited transparency, like attachment and recombination, entails a difficulty affecting the primary charges, what degrades seriously the energy resolution of the micromegas. In figure 4.9 left is shown a perfect inverse correlation between the transparency and the energy resolution and how the latter degenerates suddenly when the end of the plateau is reached. For a detector without a wide plateau, the transparency becomes a matter of tuning, limiting the operational range and stability (see figure 4.9 right). Figure 4.9: Correlation between transparency and energy resolution. Left: for a microbulk from the group CM, classic type 1. Right: for a CAST microbulk, detector M10. 4.3.1 Parametrization of the typical transparency fall. The fall is well fitted to an exponential up to the transparency loss is around 50%, see for instance, figure 4.11. A simple fit for the first half of the fall is ηe= exp [−d(FR −FRend)] (4.14) where FRend marks the end of the plateau due to the loss of transparency and dis the slope of the fall. The results from fitting the transparency curves of figure 4.8 are gathered in table 4.2 and plotted in figure 4.10. micromegas # 1 2 3 4 5 FRend (×10−3)8.704 1.523 1.176 10.04 11.33 d56.35 185.3 298.5 49.26 25.56 Table 4.2: Fit results (fit 4.14) for the meshes described in table 4.1. The values are plotted versus the mesh’s optical transparency in figure 4.10. 70 Chapter 4. Working principles of the TPC-micromegas. Figure 4.10: Results from table 4.2 plotted versus the optical transparency of the mesh. An approximate linear dependency can be found only with the exception of the plateau extension of the bulk. The fall’s slope becomes softer below 0.5. Since it is the most reproduced design for microbulks, the transparency fall of mesh type 1 is characterized in detail in figure 4.11. In order to describe the whole fall the following function can be used: ηe=p1·exp [−p2(FR −p0)] + p3(FR −p4)−p5(4.15) taking the results for the parameters that are written in figure 4.11. Figure 4.11: Reference for a micromegas transparency (from curve number 1 in figure 4.8). The simple exponential fit (4.14) is drawn with a thin line. The complete fit (4.15) is drawn with a gross line; the functions it consists on are also drawn with non continuous lines and the fit parameters are written in the legend. 4.3. Mesh transparency. 71 4.3.2 Limits for the mesh design. The degradation of the electronic collection efficiency for low FR observed for the curve number 4 in figure 4.8 is difficult to be attributed to attachment regarding the testing conditions (high gas flow and short drift gap of only 5 mm). The gas hypothesis is completely ruled out by the precise reproduction of the curve with a different specimen in a different laboratory (see figure 4.12). Figure 4.12: Transparency curves registered for IKERLAN PCB micromegas (mesh type number 4 in figure 4.7). Black diamonds and solid line: PCB2, Zaragoza’s lab, Aug 2008. Empty circles and dashed line: PCB3, Saclay’s lab, Oct 2008. The absence of the plateau is a feature of the particular micromegas structure. Trying to explain the curve we come back to figure 4.6, in the sketch two zones with very well defined electric field are clearly distinguished above and below the mesh. These are Edrift and Eamp, and their sharp differentiation makes physics of micromegas be simple and effective. Nevertheless, it actually exists a transition zone originated by the deviation from the ideal parallel-plates geometry due to the required holes. Such a zone should be smaller as smaller are the holes with respect to the typical micromegas’ lengths: the pitch and the amplification gap. If the transition is too slow, or the transition volume starts being comparable with the amplification volume, the fact that the micromegas gain is independent on Edrift can stop being true. This should be the case for these IKERLAN PCB micromegas9with 50µm diameter holes and 50µm amplification gap. Therefore the left fall of the curves 9These micromegas were developed by the IKERLAN technology center (L. Fernández and D. Gassull) in collaboration with the LFNA, and are based on conventional lithography techniques. 72 Chapter 4. Working principles of the TPC-micromegas. cannot be due to electron transport problems neither to a transparency loss, but to a decrease of the gain through a dependence on Edrift (similar observations are made in [112] studying InGrid detectors). The peak must manifest a compromise between the gain and the transparency. These particular micromegas had a typical energy resolution of 16% FWHM at 6 keV which must be limited by the mesh pattern. 4.3.3 Transparency quench. The concept of quenching factor as a dependence of the signal strength on the kind of particle is usually referred to the physical interaction (see section 3.1.5.1). However any other difference on the drift or the amplification processes would lead to the same result in practice. It can be imagined that the high density of charges generated by the αparticles may disturb the electric field geometry in the transition region, between the drift and the amplification volumes, worsening the concentration of drift lines through the mesh’s holes and causing a loss of electronic transmission. That space-charge effect could be named transparency quench. It was already demonstrated in figure 4.1 that the effect is small when the micromegas is working inside the plateau. The measurements of figure 4.13 check whether the extension of the plateau itself is affected by the charge density. They show no distinction between the two kinds of events at high FR. On the contrary, for the dense electron clouds of αevents the recombination reduces collection efficiency at low FR. Figure 4.13: Electronic collection efficiency curves taken with a classic microbulk micromegas (type 1 in 4.1) in Ar with 5% iC4H10 at 4 bar, using the set-up described in 4.4. Circles are calculated from 241Am 5.4 αand asterisks from 241Am 13.9 keV X-rays at the same time (during an only measurement run, using exactly the same gas and set-up conditions). Eamp is fixed to 88 kV/cm and Edrift varies from 0.15 to 1.15 kV/cm. 4.3. Mesh transparency. 73 4.3.4 Transparency and diffusion. Most of reasonings about the mesh transparency are made on the basis of electrostatic sketches as the one in figure 4.6. We must be aware of the implicit assumption that the drift electrons will follow the electric field lines is an idealization, the contrary may lead to think that the mesh transparency is a purely geometrical aspect. A rigorous study of the mesh transparency requires the Monte Carlo simulation of the electrons’ trajectories, which are affected by collisions, and so it concerns the gas properties. Particularly the transversal diffusion can be seen as the degree of deviation from the ideal electrostatic model. We think that figure 4.14 shows evidence that diffusion affects the transparency. Measurements are made for both bulk and microbulk (classic type, number 1 in table 4.1) sharing the same gas (see figure 4.15) for argon with 5% isobutane (iC4H10) and CF4, whose diffusion properties have been studied in section 4.2.2. Figure 4.14: Transparency curves taken with a classic microbulk micromegas (ub), type 1 in 4.1, and a bulk micromegas (b), in Ar with 5% iC4H10 and CF10 at atmospheric pressure, and pure Xe at 3 bar. It seems that the improvement of diffusion by the CF4enlarges the end of the plateau and also softens the transparency’s fall slope. It is remarkable the fact that the microbulk, when using CF4, becomes more transparent than the bulk, if using iC4H10. A complete set of transparency curves for argon with several isobutane concentrations is reported in [154] and shows the same effect: the transparency plateau is extended as the isobutane concetration is increased. Similar observations with InGrids are made in [112]. 74 Chapter 4. Working principles of the TPC-micromegas. Figure 4.15: Saclay’s micromegas lab tests chamber, where most of the gas and micromegas characterization measurements of present and next sections were carried out (courtesy of A. Giganon). Particularly the photo shows the set-up prepared for the measurements of figure 4.14, a microbulk (left) and a bulk (right) with individual drift grids inside the same chamber, Dec 2007. A third transparency curve is added to figure 4.14 for a gas with much worse diffusion, pure xenon (only slightly better than argon in figure 4.4), using the same microbulk type, though in a different set-up (Zaragoza’s HP-test chamber) and pressure. The complete set of measurements with pure xenon is plotted in figure 4.16; there how the plateau extends to higher FR as higher is the gas pressure can be appreciated. Figure 4.16: Transparency curves for pure xenon at different pressures, the reference curve number 1 from figure 4.8 is also plotted for comparison. Figure extracted from [155]. 4.4. Amplification: gain and energy resolution. 75 Such a behaviour could be connected with the σ∼P−1/2general scaling rule expected from section 4.2.2. The effect of pressure in the plateau is much clearer in the set of measurements with xenon and 1.5% trimethylamine shown in [156]. The transparency curve at 1 bar is very similar to the one obtained in argon with 5% isobutane. Furthermore, as pressure is increased, the plateau’s end moves to higher FR and the transparency’s fall becomes softer. 4.4 Amplification: gain and energy resolution. In this section the avalanche process that produces the amplification of the primary charges will be studied for the particular case of a micromegas, i.e. in an uniform electric field. Some representative experimental results will be reported. 4.4.1 The Townsend coefficient and the micromegas gain. When an electron is accelerated by an electric field and becomes energetic enough, it can yield secondary ionization whose products will join the avalanche. The process of multiplication of the N electrons of the avalanche is described by the Townsend coefficient α, which is defined by the equation dN=N(s)αds(4.16) where sparametrizes the development of the multiplication process as the length the avalanche progresses in the direction of the electric field. Basically α(ε) = ρeσi(ε) depending on the electron energy from the ionization cross section and proportional to the electronic density of the gas. However the useful definition of the coefficient incorporates the energy distribution of the electrons for the electric field E,pE(ε), and works as a mean (for high N) α(E) = Z∞ 0 pE(ε)α(ε) dε(4.17) therefore the Townsend coefficient is determined experimentally or computed with a Monte Carlo (see figure 4.17). The discussion for a microscopic understanding of the avalanche development is made in similar terms of that of section 4.2, though the energies are higher and so the involved cross sections are different. The relevant processes are ionization, recombination, attachment by electronegative molecules, and excitation and desexcitation where photons in the UV range are implied. There are several popular semi-empirical expressions with parameters that depend on the gas, like the RoseKorff’s[157] α=P Ae−B P/E (4.18) 76 Chapter 4. Working principles of the TPC-micromegas. with Pthe gas pressure and where Aand Bare gas’ parameters. Diethorn proposed a Townsend coefficient proportional to E[158]. According to figure 4.17, the linear approximation seems quite accurate for electric fields higher than few tens of kV/cm. Figure 4.17: Townsend coefficient as function of the electric field according to Magboltz for pure argon and after the addition of two quencher gases. Detail in linear scale. For a uniform electric field, the integration of (4.16) along the micromegas amplification gap dproduces G≡N/N0= eαd (4.19) as the ideal expression for a micromegas gain. 4.4.1.1 The maximal condition of the micromgas gain. One of the beneficial basis of micromegas can be deduced by combination of (4.18) and (4.19): G= exp APd e−BP d V,(4.20) with V=Ed the voltage applied between the micromegas mesh and anode. It shows Gas a function of the product Pd. The differentiation of (4.20) with respect to d, ∂G ∂d =Gα 1−BP d V,(4.21) finds that a maximum can be tuned for a micromegas depending on the gas and the operation conditions. It can be shown[159] that for argon based mixtures operating near atmospheric pressure, the operative values of the pair Vand d(i.e. voltages 4.4. Amplification: gain and energy resolution. 77 that produce a high gain inside stability) are close to the condition BPd/V = 1 for micro-gaps around 50µm . Then, since working in the neighbourhood of a maximal point, the effect of the variations in the micromegas gap due to manufacturing imperfections are minimized, as well as the variations in the gas pressure. That brings benefits for the homogeneity of the micromegas response in space (energy resolution) and time (gain stability). The existence of the maximum is demonstrated in [159], not by testing micromegas with different gaps, but inducing equivalent variations of the pressure. In [160] three micromegas-like InGrid with gaps between 45 and 70 µm point to an optimum gap about 53 µm . That condition will not be mistaken for a condition for achieving a maximum gain with a micromegas. The condition was interpreted as an optimum electric field for a given gas and conditions. Then the combination of Eand dwill result in a certain gain, in case Eis applicable. Actually larger gaps usually achieve higher maximum gains despite working with lower fields. The limit on the gain of a micromegas has a different nature and is briefly commented in the following section. 4.4.1.2 Maximum achievable gain. The real avalanche has a geometry where ions and electron charges are not uniformly distributed. The internal electric field induced by the charges can produce a spark breakdown if they are too numerous. A phenomenological limit known as the Raether condition predicts the breakdown when G∼108, i.e. αd ∼20. In practice it is difficult to achieve G > 105because other aspects can also trigger the spark. The most important is the induction of a second avalanche started by a secondary photon that induces a photo-electron in a gas molecule or the mesh material. Avoiding this is the main purpose of the quencher gas, which must have good absorption cross section for these photons and lead to the excitation of non radiative states (rotating, vibrating). In the absence of any quencher, pure argon or xenon break down before achieving G > 103in typical microbulks[161]. The role of the quencher is then crucial in order to achieve high gain and stability; besides it can produce diverse effects that are left to section 4.4.3 for being illustrated in basis to some practical examples. Both spark breakdown trigger mechanisms mentioned above are understandably reinforced for high stopping power particles and high trigger rate, i.e. by the time and space density of avalanches, since the photons generated by an avalanche can affect another one. The maximum stable gain obtained in figures 4.1 and 5.3 can be compared with typical gains, for example from figure 4.24; they are much lower in presence of α-particles, even though the trigger rate was decreased to few Hz by collimation of the source. In this sense the microbulk geometry is thought to provide an advantage since the the amplification volume is divided in a kind of amplification cells isolated by ∼ 50–100µm of kapton, which can absorb the photons acting as a geometric quencher. 78 Chapter 4. Working principles of the TPC-micromegas. Finally the design lengths and tolerance of materials can imply a limitation to the voltage applied to the micromegas. Specially particular imperfections in the softness of the mesh and anode surfaces and borders, the presence of a piece of dust or humidity, which tends to be persistent inside the micromegas structure, can limit the operation of a micromegas because of reasons that are not related with the avalanche process. 4.4.2 Avalanche fluctuations and energy resolution. In order to characterize the avalanche fluctuations, an immersion in the microscopic picture associated to equation (4.16) is required (details are given in [116]). That means that α(ε)must be evaluated for the particular electron, what actually introduces a dependency α(s), even when Eis constant, related with the particular electron history or the avalanche development. A new electron, that is produced with nearly zero energy, needs to travel a distance longer than s0=Ui/E, with Uithe first energy of ionization of the gas, before being able to ionize as well. On the other hand, from (4.16) it can be defined an ionization mean free path for the electron λ= 1/α. The comparison of both lengths is related with the relevance of the particular position the electron was released, and parametrizes the distributions of avalanches χ(E)≡s0 λ=α(E)Ui E(4.22) χis named the relaxation parameter. Note that from the definition of α(E)in (4.17), s0already influences α(E)because is related with the electrons’ energy distribution and χ < 1. Values of χfrom Magboltz’s description of α(E)(see figure 4.17) for an argon-based mixture are plotted in figure 4.18. Values of Ui: 15.8 eV (Ar) and 12.1 eV (Xe). Figure 4.18: Relaxation parameter χ(E)plotted from expression 4.22 for argon based mixtures (Ui= 15.8eV) and α(E)of Ar/iC4H10 98/2 taken from figure 4.17. 4.4. Amplification: gain and energy resolution. 85 Figure 4.24: Comparison of gain curves for bulk (circles) and microbulk (triangles) micromegas in Ar/iC4H10 95/5 (filled markers) and CF4(empty markers). The maximum gain cannot be directly compared because different prototypes were used. Figure 4.25: Gain curves for bulk micromegas using different gas mixtures. The last (highest) point of each series corresponds to the maximum stable gain. avalanche (see section 4.4.1.2). However that fact is true for gas mixtures which actually have good quenching capabilities. Pure noble gases seems not to achieve gains higher than 103(see section 5.2) and it is very notable the improvement of Ar/CF490/10 with the addition of 2% iC4H10 in 4.25. 86 Chapter 4. Working principles of the TPC-micromegas. The quencher gases affect the dynamic of the avalanche at different levels, changing the Townsend coefficient and the photons and electrons absorption cross sections. Therefore, the different quencher characteristics can be complementary. That seems to be the case of isobutane and CF4. Isobutane has very good photon non radiative absorption, it allows very high gains at low electric fields and also good energy resolution. On the other hand CF4is more appreciated from the drift point of view (see section 4.2). Relatively high concentrations of CF4are needed to produce an efficient enough quenching effect to reach very high gains >104, as a consequence very intense electric fields are required (see figure 4.24). The ternary mixture Ar/CF4/iC4H10 88/10/2 (figure 4.25) balances both quenchers properties and seems to keep the isobutane and CF4mixtures individual advantages. Ar/CF4/iC4H10 mixtures are studied with more detail in section 4.4.3.1. Isobutane can also relax the very high operational voltages needed for pure CF4operation, though the effect saturates soon with the iC4H10 percentage (see figure 4.26). On the contrary, the effect of addition of isobutane to a CO2quenched gas is much more modest. Figure 4.26: Gain curves for microbulk micromegas showing the effect of adding iC4H10 to pure CF4. The 10% isobutane mixture’s curve is truncated because of microbulk’s damage. The resolution was also slightly improved from 17.1% to 16.7% (with 5% iC4H10). These resolution values are equivalent to the result in Ar/iC4H10 95/5 for that particular microbulk. Another interesting feature of the gain curve is the exactness of the exponential behaviour, i.e. the straightness of the series in the semi-logarithmic plots. It can be detected that the series show an over-exponential gain, which is more evident from some mixtures and becomes more clear at high gains. That behaviour could be due to the contribution of secondary avalanches triggered by photons (photon-feedback), and it is a prelude of the spark. This possibility implies new fluctuations that have 4.4. Amplification: gain and energy resolution. 87 Figure 4.27: Resolution vs gain, taken with aCM microbulk in Zaragoza. No very high gains were tried during that test, so as not to risk the microbulk. Equivalent curves showing further resolution degradation at high gain can be found, for instance, in [100]. not been accounted in the single avalanche model, and the energy resolution degrades in the over-exponential behaviour piece of the gain curve. A typical energy resolution curve, figure 4.27, improves with gain since the signal-to-noise-ratio (SNR) increases, but there is an optimum after which the growing extra-fluctuations cause resolution degradation. Therefore a straight gain curve can be a sign of good, or at least stable, energy resolution. Studying the measurements summarized in figure 4.25, one guesses that the prevention of the over-exponential behaviour and the tolerance of the secondary avalanches (not producing sparks) are different aspects of the gas mixture. The addition of quencher logically helps preventing the secondary avalanches, so the 10% isobutane curve is straighter than the 5%, thought the maximum achievable gain is similar. Comparing the energy resolution at high gain, ∼3×104, when the 5% gain curve is clearly over-exponential, the energy resolution is 25.9% (5% isobutane) while 21.8% (10% isobutane). Isobutane seems to have a good tolerance that makes these mixtures develop a long stable over-exponential stage that allow them to reach gains >104with small quencher amount. Ar/CF490/10 is also clearly over-exponential but, the bad photon quenching properties of CF4, makes the secondary avalanche too probable still at low gains. Energy resolution is worse than 30% along all the curve. On the contrary, both Ar/CO2curves (with and without iC4H10) are completely straight up to the end and show no resolution degradation, 23.1% in the last point of Ar/CO2 90/10 (it must be reminded Ar/iC4H10 95/5 performed 23.9% actually at a bit smaller gain); then, suddenly, the continuous sparking limit appears. Therefore CO2as quencher seems to prevent the secondary avalanches, but do not tolerate them. Ar/CO2/iC4H10 mixtures are studied with more detail in section 4.4.3.2 4.4.3.1 Ar/CF4/iC4H10 mixtures. The effect of adding iC4H10 to Ar/CF490/10 is studied in figure 4.28 for several amounts of isobutane. The effect of lowering the operational voltage saturates soon, but higher maximum gains and more straight curves are obtained as higher the isobutane proportion is. The energy resolution is improved for a given gain, as it is shown in figure 4.29 for G≈104with exception of the point with absence of iC4H10 88 Chapter 4. Working principles of the TPC-micromegas. (G≈2×103) and only 0.5% of iC4H10 (G≈4×103). A comparison of spectra at similar G≈4×103is down on the right14. Figure 4.28: Gain curves for bulk micromegas in different Ar/CF4/iC4H10 mixtures. These are independent measures of those of figure 4.25, though taken with the same bulk. Figure 4.29: Left: energy resolution at G≈ ×104for Ar/CF4/iC4H10 (90 −x)/10/x mixtures. Right: spectra at G≈4×103for x= 0.5(up) and x= 4.0(down). 14Unfortunately a systematic characterization of the energy resolution was not carried out, since the main purpose of the work[166] was just to find a high stable gain and fast gas mixture. 4.4. Amplification: gain and energy resolution. 89 To sum up, the addition of 4% iC4H10 has led to obtain a gas mixture that should keep the Ar/CF490/10 drift properties, but working at mesh-anode voltages which are 20% lower. Moreover it is able to achieve a maximum gain as high as Ar/iC4H10 95/5, but, due to the combination of both quenchers, with a softer over-exponential behaviour, what allows an improved energy resolution. The same has been tried with a mixture with higher concentration of CF4for excellent drift properties. The particular mixture Ar/CF4/iC4H10 73/25/2 was checked to have a transparency curve as good as pure CF4(see figure 4.14) and the energy resolution that is demonstrated in figure 7.14. 4.4.3.2 Ar/CO2/iC4H10 mixtures. The CO2concentration was increased to 20% to test whether the maximum gain could be enhanced and the energy resolution as well (see figure 4.30). The obtained energy resolution value was notable (see figure 4.31) achieving 20.1% at 625 V, G≃7×103, the best energy resolution got during this study. The result is even more significant as it was taken only 20 V before the sparking limit, again in a completely straight gain curve. But the maximum gain was hardly increased and sparks appeared without any warning signal; it seems that CO2alone cannot tolerate the secondary avalanches, even in high concentrations. Figure 4.30: Gain curves for bulk micromegas in different Ar/CO2/iC4H10 mixtures. Since the addition of isobutane to the Ar/CO290/10 mixture had modest results, the opposite approach was tried. Some CO2was added to the reference mixture Ar/iC4H10 95/5 in order to check if the straightness of the Ar/CO2gain curves, and so the good energy resolution, could be tranferred to it. The test was successful in producing a gas mixture with a maximum gain as high as Ar/iC4H10 95/5 at only 90 Chapter 4. Working principles of the TPC-micromegas. Figure 4.31: Spectrum taken with a bulk micromegas in Ar with 20% CO2at at 625 V, G≃7×103. ≃25 V higher voltages, but with notably more straight gain curve and enhanced energy resolution: 21.3% at 450 V, G≃6×103. 4.5 Signal induction. The Ramo’s Theorem. How the drifting charges induce electronic signals in the detector electrodes is described by the Ramo’s Theorem[167], which stands for the charge induced in a particular electrode, numbered n, by a charge qto depend only on the electric field created by the n-electrode itself and the trajectory follow by q. If ψn(~r)is the weighted potential, what means the potential induced by the n-electrode with voltage Vnwhile the rest of the electrodes are grounded and the charge qis not present, the current induced in the n-electrode is Iind n(t) = q Vn ~ ∇ψn(~r(t)) ·d~r(t) dt=q Vn ~ En(~r(t)) (·~v(t)(4.32) where ~ En(~r(t)) is the weighted field associated to the n-electrode with voltage Vn15 and ~r(t)is the trajectory followed by q. Pay attention to the fact that ~r(t)will be determined by the real electric field, instead of ~ En. Note also that the sign of the induced current depends on the sign of qbut on the relative orientation of ~v and ~ Enas well. In particular, for an infinitely long strip of width wand charges moving in the sdirection of the avalanche’s advance, perpendicular to the strip, it is obtained: Iind strip(t) = 4qw π[4s2(t) + w2] ds dt.(4.33) Equation (4.32) is the key in order to compute or simulate the electronic signals. However a corollary regarding the total induce charge is more often invoked in lab 15The weighted potential and fields are often defined specifying Vn= 1, thus Vndisappears from the equations. 4.5. Signal induction. The Ramo’s Theorem. 91 reasonings. To compute the total induced charge we must integrate (4.32) along the q’s trajectory. Since ~ En(~r(t)) is a conservative field Qind n(t) = Ztf ti Iind n(t) dt=q Vn [ψn(~rf)−ψn(~ri)] (4.34) the results depends only on the initial and final positions of q. If avalanche’s charges are created in pairs qand −qin the same initial point, that makes vanish the initial position dependence when both contributions are summed. Regarding the ψn(~rf), if the charges finally arrive to any electrode of the system, then by definition ψn= 0, if the electrode is different that our n-electrode (pixel or strip), and ψn=Vnif the charge is caught by it. Therefore the charge induced in a electrode is equal to the number of charges collected by it. That justifies the simple picture one often has in mind, but stands for the total induced charge after an undefined transitory time. The only thing that can be actually said for electrodes which do not collect net charge (for instance, a mesh being crossed by primary electrons in absence of amplification) is that the induced current will be strictly bipolar, i.e. the negative an positive areas of the pulse will be equal. Most of the electron-ion pairs are created in the last stages of the avalanche near the anode. That induces a fast rise in the signal due to the electrons, that is followed by a slower pulse produced by the drift of the ions along the amplification gap in the opposite sense (so inducing charge with the same polarity). Figure 4.32 contains a micromegas pulse taken with a fast current amplifier and shows the mentioned pulse structure for a single electron avalanche. The signal comes from the anode of a 100 µm gap micromegas working with He/iC4H10 90/10, a combination that allows a very high gain near 107[164]. The first fast rise, due to the electrons, had a distribution with mean around 3 ns and σ= 770 ps. The slower signal presents two peaks that are attributed to the presence of two kind of ions, from He and iC4H10, with different mobilities. Figure 4.32: Single electron pulse caught with a fast current preamplifier. From [164]. Chapter 5 Micromegas’ milestones for Rare Event Searches. Contents 5.1 Behaviour at high pressure. ................... 94 5.2 Use of pure noble gases. ..................... 97 5.3 Penning mixtures. ......................... 99 5.4 Radiopurity. .............................100 5.5 Electronics. T2Kelectronics library. ..............102 5.5.1 T2K electronics set-up. ..................... 102 5.5.2 T2K electronics library. ..................... 103 5.5.3 TPC event reconstruction. .................... 104 5.5.4 Outlook. .............................. 108 Many aspects of the performance that makes micromegas interesting for particle physics were shown in the previous chapter, and the principles that allow them as well. These operational principles can be put to the test because usually detector and target are the same entity: the experiment’s sensitive volume. Therefore the gas choice and its conditions are often strongly constrained by the physics case. Besides new requirements and features can be demanded by rare event searches, depending on the experiment type. Thus rare event searches sometimes force TPCs and micromegas to the frontier of their conventional operation. Some of the milestones demanded by rare event searches with TPCs are main aspects of the R&D program inside T-REX and are the subject of this chapter. The LFNA and close collaborators have already published several papers describing the status of these research lines and other Ph. D. works by LFNA researchers deeply study them, therefore here the challenges will be just addressed and we will only remark some general keys and cite the suitable references. As an exception, the last section 5.5 describes a work performed with the aim of adapting a last generation electronics, specifically designed for a rare event searches experiment with micromegas, for the T-REX researchers. For rare event searches high trigger rate capabilities would probably be not relevant, however high efficiency is, for sure, essential. Moreover the detector has usually to fulfil the requirements of both a tracker and a calorimeter. Larger TPC’s 94 Chapter 5. Micromegas’ milestones for Rare Event Searches. are needed to increase the target mass and/or the detection efficiency, what entails difficulties for the readout construction, longer drift distances, etc. Increasing the gas pressure can be a smart alternative, however the feasibility of the drift and the amplification stages must be proven (section 5.1). The definition of a TPC’s fiducial volume, i.e. a virtual volume fully contained inside the TPC active volume without physical borders, can be essential to avoid radioactivity from the innermost TPC’s surfaces. That includes the cathode and the readout plane, hence the reconstruction of the absolute drift coordinate, the t0, is required. It will be explain (section 5.2) that this new feature may imply the operation with pure noble gases in absence of quencher. Beyond the operation of the micromegas, higher pressure, longer drift distances and the lack of a quencher make the detector more susceptible to attachment due to gas impurities. The quality of gas becomes the first goal, an unavoidable requirement to explore the actual micromegas’ possibilities. Moreover, the high cost of xenon requires gas recovering or recirculation. Before the detector, the supporting gas system requires our attention. The T-REX works started adapting the HELLAZ setup at CEA/Saclay. Later a dedicated set-up was built in Zaragoza, implementing the learnt lessons. Both set-ups are described in the articles that will be cited, including consecutive upgrades of Zaragoza’s lab. Here only their effects will be shown shortly in section 5.2 to illustrate also how to detect and deal with attachment when it is present in the data. Specially for ββ0νdecay searches, the energy resolution is capital. The Penning effect is a via to improve the essential limitations in energy resolution that were analysed in section 4.4.2.2, it is presented in section 5.3. Finally, the TPC has to fulfil radiopurity requirements as all the low-background detectors. That entails restrictions in the detector materials that directly affect to very common materials, as stainless steel, and some common techniques, like soldering. We will focus in section 5.4 in the core of the TPC, the micromegas. 5.1 Behaviour at high pressure. Microbulk micromegas do not seem to suffer from practical limitations when increasing the pressure, as it is shown by figure 5.1 up to 10 bar using Ar/iC4H10 99/1. The gain curves[169] were taken at Zaragoza’s set-up using a microbulk micromegas prototype like the one described as classic type (number 1) in section 4.3 and photographed in figure 4.22. The same is true for the rest of the results showed in this chapter. The gain series were taken with a low energy source, 57Co 6.4 keV X-rays, and reaching the maximum achievable gain was not intended. The microbulk was not forced up to the breakdown limit. Nevertheless, a gain over 103was demonstrated for all the pressures and considered enough for illustration purposes. 5.4. Radiopurity. 101 raw materials used for their manufacturing are included in the group of the most radiopure: copper and kapton (double copper clad polyamide laminate Sheldahl G2300). However the intrinsic radiopurity of the final micromegas is not, in principle, guaranteed since several chemical baths are used for the kapton etching and adhesive (epoxy, Isola DE156) is needed to glue the micromegas foil to a mechanical support. It must be noticed that the manufacturing process has not been thought with the goal of minimizing the radioactive impurities. Moreover, if a readout layout is needed, additional Kapton-Cu layers can be glued to avoid conventional PCB technology, that is known to be too radioactive, what requires further epoxy and new processes to print the circuit. Therefore the LFNA, in collaboration with R. de Oliveira from the CERN MPGD’s Workshop, undertook a series of measurements with the aim of confirming the final intrinsic radiopurity of microbulks using a p-type germanium detector in the Canfranc Underground Laboratory (LSC)[175]. The measured samples included the microbulk raw materials (Cu-Kapton-Cu and Cu-Kapton structures), and two elaborated (up to different level) structures coming from CAST microbulk detectors (see section 8.2). The first one (figure 5.7 left) is a complete microbulk with a 2D readout, which requires two additional epoxy-Kapton-Cu layers. The second sample (figure 5.7 rigth) consists of the 2D readout layout of a classical CAST micromegas whose mesh-pillars structure has been removed. That readout was built from the same basic Cu-Kapton-Cu structure of posterior microbulks. The measurement of the microbulk raw foils checks whether these structures mantain the radiopurity of the isolated kapton and copper; the old 2D-readout should reflect the contribution from the basic processes for circuit printing of strips, vias, etc; the complete microbulk adds epoxy (though the old 2D-readout can obtain residual epoxy) and the results of a more aggressive kapton etching process. The measurements are summarized in table 5.1. Figure 5.7: Sketch of the two CAST micromegas from which the samples (squared) were taken. Left: CAST microbulk with 2D-readout. Right: CAST classic with 2D-Readout. The potential of microbulk micromegas is better appreciated if compared with the values of a light readout used by the XENON 100 experiment that is composed 102 Chapter 5. Micromegas’ milestones for Rare Event Searches. sample 232Th 235U238U40K60Co Kapton-Cu foil <4.6* <3.1* <10.8 <7.7 <1.6* Cu-Kapton-Cu foil <4.6* <3.1* <10.8 <7.7 <1.6* Old 2D-readout 4.6±1.6<6.2 <40.3 <46.5 < 3.1* CAST microbulk <9.3 <13.9 26 ±14 57 ±25 <3.1* Hamamatsu R8520-06[176]28 ±10 - <37.2 1700 ±310 93 ±16 Table 5.1: Radioactivity levels in µBq/cm2.*Level obtained from the minimum detectable activity of the detector. by an array of PMTs[176]. XENON100 is currently the most sensitive Dark Matter experiment, apart from other merits, because of a carefully selection of materials that includes the development of these particular low-background PMTs. The micromegas’ activity is specially lower for 40K and 60Co. Regarding our series of measurements, the only relevant increase of activity (despite being compatible with zero at 2σ) is registered for the complete microbulk, specially in the 40K. This facts points to the epoxy and/or the chemical baths (KOH is used during the etching process) as the origin of its radioactivity and, therefore, a potential improvement by optimization of the manufacturing process. A direct exploration of the practical limit of the micromegas background due to its intrinsic radiopurity has been carried out for an equivalent CAST microbulk by means of underground data taking with special shielding configurations and is summarized in section 10.3. 5.5 Electronics. T2Kelectronics library. The last strategy for decreasing the background is the development of discrimination criteria based on the detector’s topological capabilities, the main argument for a gaseous TPC. The readout potential must be enhanced by the use of state of the art electronics. One goal of this work has been the acquisition of experience with last generation electronics for TPCs and the development of tools that make them available for the test-bench activities of T-REX. 5.5.1 T2K electronics set-up. We focused on the electronics developed by the CEA/Saclay for a rare event search experiment using micromegas: T2K[177]. The electronics core is the AFTER ASIC[178], which is continuously recording every channel in a buffer with capacity for 512 samples with 12 bit precision at 100 MHz maximum sampling rate. Each channel has been pre-amplified and shaped, being the faster peaking time 100 ns, and with an also tunable gain. The basic T2K electronics unity is able to read 1728 channels which are delivered via optical link. 5.5. Electronics. T2Kelectronics library. 103 The basic unity of the electronics set-up is a light set of at least 2 cards (244 channels) and up to 7 cards (1728 channels), quite compact and robust (see figure 5.8), which needs a low voltage DC power supply. An intermediate commercial Data Concentrator Card is required to finally send the data to the PC via Ethernet. Figure 5.8: T2K electronics set-up sketch for reading a micromegas: the electronics unit is composed by a mother board (FEM) plus from 1 to 6 front-end cards (FECs). The trigger is produced by pre-amplification and shaping of the mesh signal, which must exceed the discriminator threshold. The logic signal may need of a level adaptation stage before be given to the Data Concentration Card (DCC) which triggers the DAQ pause and data transfer. The electronics must be provided with a logic TTL signal as external trigger. One advantage of micromegas is that the mesh signal is a suitable trigger signal because, since it is induced by the whole event charge, it is relatively intense and common to every event. Essentially the typical analogic electronic chain plus a discriminator will be needed (see figure 5.8 sketch). 5.5.2 T2K electronics library. T2Kelectronics library is the software tool developed to handle the data acquired with the T2K DAQ. It is written in C++ and makes use of ROOT classes (as customary in the LFNA, see RESTSoft chapter 6). It has four essential components: a low level class, T2Kreader (based on a program by David Attie from CEA/Saclay), which directly deals with the binary output files of the T2K DAQ, absorbs the T2K data format and can load in memory the recorded pulses and provide fast visualization; a general holder class, T2KrawEvent, to handle and store the data already as a ROOT TObject; the method to write T2KrawEvents ROOT trees from T2K binary files and a practical description of the set-up. The set-up description distinguishes between the electronics’ configuration and the detector’s architecture. T2Kreader requires the electronics structure which is read, number of cards, channels, etc. The auxiliary class T2Kconfig is able to deduce it from a simple configuration text file, together with other parameters which can be 104 Chapter 5. Micromegas’ milestones for Rare Event Searches. useful for other methods. An associated class, ReadoutDecoding, provides a model to implement the relationship between the electronic channels and their physical meaning for a particular detector and cabling (the decoding task). Several readoutdecodings, corresponding to real set-ups, have been written and can be taken as templates. T2KrawEvent is thought to be versatile and powerful, therefore simple in structure but relying on advanced objects; it is basically a collection of a variable number of ROOT histograms that hold the recorded pulses plus the event time. Every histogram is assigned a channel number, to address its origin, and two more optional numbers to describe the physical position of the signal, generally called row and column.T2KrawEvent can absorb the decoding result, but there is the alternative to use its pointer to a ReadoutDecoding object too. Therefore T2KrawEvent, when used by other methods, can provide the physical position of every signal either because a previous decoding process result was already recorded on it or directly by making it on-line. The variable size of T2KrawEvent, which is independent on the electronics configuration, allows off-line data selection methods. The main method that produces the T2KrawEvent can optionally perform a selection by threshold and the decoding during the format data transformation process. It also reads the DAQ time stamp and reconstructs the relative time. It is an especial case in the T2Ktools class, which is not conceived as a C++ class, but more exactly as a set of functions to handle T2KrawEvents using, if necessary, ReadoutDecoding and T2Kconfig. It includes very basic functions together with elaborate analysis and drawing methods; they are the basis for the programs which are written for practical use. Figure 5.9 contains representations of events recorded with a CAST micromegas detector using T2K electronics. They (and the rest of figures in this section) has been generated with T2Ktools methods and have the purpose to demonstrate the detector, the electronics and the software possibilities. 5.5.3 TPC event reconstruction. A full event reconstruction5method can only work with a pixelized readout. The microbulk micromegas in figure 5.10 left was specially designed for this purpose to fit Zaragoza’s lab high pressure chamber (see section 6.2.7.2 and [169]). It has got an active area with 12 ×12 pixels in 10 ×10 cm2, what was a world record at its time (Summer 2009) doubling CAST microbulks size. Its performance is remarkable because it supposed a milestone in scaling-up potential of microbulks. The resolution obtained using a non collimated 55Fe source was 12.6% FWHM, only slightly worse than the usual microbulk prototypes of almost 15 times smaller size (see section 4.4.3, in particular figures 4.22 and 4.23). Spectra like the one in figure 5.10 right, which has been taken from the mesh pulses, 5Here, for event reconstruction it is understood just the deduction of its three-dimensional structure, no physical interpretation is done. 5.5. Electronics. T2Kelectronics library. 105 Figure 5.9: Events as registered with CAST bulk micromegas B5 (strips readout): recorded pulses and readout projection (each bin amplitude is the product of the amplitudes of the corresponding strips). Top: a 109Cd 22 keV γ. Center: 241Am 5.5 MeV α. Bottom: cosmic muon. show that the increase of the electrodes capacity was not limiting the detector threshold and resolution. It is also noticeable the way other auxiliary detector aspects were developed at the same time. For these measurements a new feedthrough design by J. P. Mols from CEA/Saclay (figure 5.11), based in a SAMTEC high density connector[179], it was used to extract the signals from the high pressure vessel. The signals are driven by plane and flexible cables called limandes. 106 Chapter 5. Micromegas’ milestones for Rare Event Searches. Figure 5.10: Left: Microbulk micromegas with pixels readout. Right: 55Fe spectrum taken with this micromegas, collimating the source at its center. Energy resolution is 11.9% FWHM and the energy threshold is below 0.5 keV. The feedthrough leak tightness was tested at vacuum and high pressure obtaining no deviations from previous tests with the same vessel. Anyway the chamber with the feedthrough could operate perfectly with a 6 cm drift distance using pure argon at drift fields lower than 50 V/cm, as the following event reconstructions will show. Later versions of these feedthrough are currently extensively used in the NEXT-1MM prototype. Figure 5.11: Left: the feedthrough, a SAMTEC connector[179] integrated in a CF60 outlet, with limandes. Right: the feedthrough installed in the high pressure chamber and connection to the pixelized microbulk. The reconstruction algorithm written for the third dimension is very simple. The pulse is sampled around its maximum position with a period which should be related with the electronics peaking time; then, when a significant time extension of the signal is found (i.e., basically when the pulse’s width is longer than the shaping time) a secondary charge core, or 3D-pixel, is added to the event representation (see figures 5.12 and 5.13). A more advanced idea for the reconstruction could rely in a deconvolution of the signal using the electronics shaper typical response, following the inverse operation of that explained in section 6.4. 5.5. Electronics. T2Kelectronics library. 107 Figure 5.12: 241Am 5.5 MeV αevent in Ar with 2% iC4H10 at 1 bar and Edrift = 30 V/cm. The mean length of the αtracks is 5.3 cm. T2K electronics working at fastest mode: 100 MHz sampling rate and 100 ns shaping time. Right: individual pixels maximum amplitude. Top left: recorded pulses. Bottom left: event reconstruction. Figure 5.13: 241Am 5.5 MeV αevent in pure Ar at 1.23 bar and Edrift = 20 V/cm. The mean length of αtracks is 4.3 cm and longitudinal σfor 6 cm drift is 8.6 mm. T2K electronics configuration: 25 MHz sampling rate and 1 µs shaping time. 108 Chapter 5. Micromegas’ milestones for Rare Event Searches. 5.5.4 Outlook. Today T2K electronics are used in Zaragoza labs to read CAST detectors, the NEXT-1-MM prototype (see figure 5.15) and a prototype for high sensitivity radon detection and soon new T-REX prototypes for Dark Matter. Besides the CAST acquisition upgrade expected for 2013 is based on T2K electronics. The T2Kelectronics library was conceived as a BETA version (see figure 5.14), with absolutely general assumptions and only a basic functionality, however it produced a set of simple demonstration programs which are still intensively used by the former set-ups operators. It was thought as an outline for future more formal and specialized versions. Figure 5.14: Sketch of the classes involved in the T2Kelectronics library. There are at least three levels of data evolution, from the binary output to the reconstructed event (TRestPhysEvent class, see section 7.3) through the raw event. See section 7.3 to relate T2Kelectronics librare with RESTSoft package via the Tt2kManager class. Dashed lines represent the information flow via some practical programs based in T2Kelectronics library. That process has started thanks to the work by younger researchers with the establishment of stable particular versions for data management of CAST detectors, the possibility to emulate CAST classic DAQ (see section 10.4) and the development of new analysis based in T2KrawEvent for CAST and NEXT-MM. Furthermore see section 7.3 for a connection with RESTSoft libraries. A specially useful achieved goal has been setting up a combined DAQ which uses T2Kelectronics for reading the detector segmented anode and an oscilloscope[180]6 6Oscilloscope libraries by J. Galán and J. A. García, who also provide a on-line DAQ GUI. 5.5. Electronics. T2Kelectronics library. 109 Figure 5.15: Cosmic muon as registered with T2Kelectronics and drawn with T2Kelectronics library in the NEXT-1-MM chamber[107]. The readout was a 30 cm diameter bulk micromegas with 1252 pixels, though a microbulk version is currently in operation[19]. for the acquisition of the mesh pulse and the configuration of the common trigger. This relatively portable set-up is quite versatile thanks to the adaptability of the oscilloscope for any kind of signal and because the DAQ inherits from the oscilloscope the capability of naturally configuring complex trigger criteria, including coincidences between different signals. 6.2. The RESTSoft GEANT4 application. 117 In figure 6.4 the energy loss along the first step is plotted. The histograms with non filled area correspond to the energy deposited in the gas, that is to say, the energy that is assigned to the hit. We distinguish between normal steps and steps whose end was suddenly caused by the ionization process. The former cannot exceed the energy cut, which for our conditions is slightly higher than 200 eV, and the energy loss is equal to the energy deposition. For the second, the absorbed energy can be higher than the energy cut and lower than the energy loss if secondary particles (with initial energies higher than the energy cut-off) are generated. The L-edges of argon at 250 and 310 eV (above the energy cut-off) are clearly visible in that distribution (another peak at 3.2 keV, K-edge, is out of range). The second part of the distribution, caused by the presence of the L-edges, ends necessary at 510 eV since that is the sum of the energy cut-off and the most energetic absorption line. Both distributions are described (with exception of the atomic edges) by Gaussian statistics, as the sum of many independent energy depositions. Figure 6.4: Energy loss along the first step of 6 keV electron tracks. The difference between the energy loss and the deposited energy is the energy invested to secondary particles, which is represented by the solid area histogram. It can be fluorescence photons and Auger electrons (see section 3.2.1.1) related with the atomic bound or δ-electrons (see section 3.1.4). If we had chosen the energy cut-off higher than the binding energies, the related secondary particles would not have been simulated with consequent acceleration of the simulation. These are accumulated in peaks while the δ-electrons should draw a continuous distribution, however the distribution we see is not directly the δ-rays energy production distribution, that can be the sum of several secondary particles. Ignoring this last detail, a completely different distribution is expected here, according to section 3.1, the 118 Chapter 6. Simulations for Rare Event Searches with TPCs Coulomb scattering one (see equation 3.1), however ∆Eis still not much higher than the binding energies. The solid line, which is a fit with shape ∼1/Ex, is not able to reproduce such a long tail for a wider range than plotted in figure 6.4 and gives x= 1.59 instead of the expected 2. The distribution fits much better to a Landau tail (dotted curve). To sum up: the way the track is developed in the simulation is similar to the real case, in which discrete clusters are generated; however we should keep in mind the subtle difference existing between actual clusters and the hits of the simulated ionization track, due to the fact that ionization is a treated (most of times) as a continuous process. In contrast, when a photon is simulated there is no ambiguity to decide where a conversion into an electron by photoelectric effect or where it was supposed to suffer a Compton scattering. The basic description of the track inside the target, as a collection of hits with positions and energy depositions, is completed with the physical process that defined the new step (recorded following a numerical code) and the trackID, which numerates the concrete particle which caused the hit. The trackID can be useful, e.g. to distinguish between the primary ionizing particle track and secondary ones. 6.2.2 Physical processes. The Physical processes which can happen during a GEANT4 simulation are defined in its PhysicsList class for each particle. The RESTSoft GEANT4 application’s PhysicsList started from previous lists developed for simulations of the Dark Matter experiment ANAIS[43], which are described in [143,144], themselves based on the Underground Physics advanced example of GEANT4. Only relevant processes for low energy Physics are included (using the low energy versions of GEANT4 process and models). For electrons, in particular, the ionization and multiple scattering presented in the previous section plus Bremsstrahlung and annihilation for positrons. For X-rays and γ0s, all those presented in section 3.2.1; but optical photons are not considered since it is (still) not intended to simulate an optical readout. A special case is the Radioactive Decay process that can generate the natural decay of isotopes. In summary, the simulation is simple from the point of view of the amount of relevant particles and processes involved, but it must be detailed, therefore secondary processes as fluorescence, PIXE and δ-rays are activated with energy threshold related with the step length as was already discussed in the previous section. 6.2.3 Geometry. For a simulation of a particular detector its geometry has to be implemented. Nevertheless all the simulations have some common features and many times only some dimensions and materials arrangements are enough in order to essentially reproduce 6.2. The RESTSoft GEANT4 application. 119 the situation. A configuration file is available and the application is able to interpret it and build automatically a simple but versatile geometry. Basically the general geometry can be described as a set of cylindrical layers, in onion-like scheme, plus two levels to define the electrodes (see figure 6.5). The target is, of course, contained in a gas volume which corresponds to the conversion volume of the TPC. The target is defined as the volume that is projected onto the micromegas’ surface (which is supposed to be at the bottom of the chamber) up to a certain level, usually that of the cathode position. The fact of distinguishing the target from the gas and conversion volume (volume between the electrodes, mesh and cathode) allow to consider realistic dead zones in our chamber. The simplest geometry is defined just by three parameters: the chamber’s height and radius, and the micromegas mesh radius. Figure 6.5: Sketch of the onion-like geometry of the general RESTSoft GEANT4 application. The optional shielding layers are not drawn for clarity. The following layers are defined by their thickness and are build so as to perfectly adjust to the immediate inner layer. The first one may represent the chamber structure and could be important in case the chamber would contribute with fluorescences. For the same reason, the micromegas mesh itself is represented as a thin layer of copper (in case of a microbulk) or stainless steel (if a bulk) and the readout behind with copper-kapton layers. The outer layer can be used to add a basic shielding in case we would like to study the effect of an external source. The gas inside the chamber is, by default, pure xenon or a mixture of argon and isobutane. The mixture proportion and the gas pressure can be set through the configuration file too. In the geometry class of the program, mMDetectorConstruction, there is a simple method to properly define the mixture of gases (whose proportions are given in volume), which can be generalized easily to other kind of gases. 120 Chapter 6. Simulations for Rare Event Searches with TPCs 6.2.4 Event generation. The eventGenerator class determines the initial conditions of the event to simulate: the number, nature, position and state (the initial vertex) of the particles that will originate an event that may be registered in the TPC. This class contains a collection of functions, some of them to set the position of the vertex and others to take care of the particles’ initial states, which are combined. These functions can make reference to geometrical parameters, therefore the generator class has access to the geometry builder, mMDetectorConstruction class, and the user-developer must be aware that a contradiction could be produced when the geometry is modified. The evenGenerator uses an auxiliary class, ProbGen, which generates some useful distributions, as isotropic directions, or homogeneous positions in simple geometrical volumes. This way methods to simulate some radioactive source commonly used in the lab, or some kinds of radiation fluxes or the result of key decays have been written (55Fe, 109Cd, general β-decays electron spectra, cosmic muons, etc.). The class is also able to interpret a Decay0[185] generated file for double beta simulations. The most typical uses can be chosen and configured through the configuration file. As usual, the time is used as a seed for the Random Numbers Generator (GEANT4 make use of CLHEP libraries[186]). However simulations are often performed as parallel processes which are sent to a cluster queue, and it is not rare that some of the programs are accepted, and start running, at the same time. To assure independent simulations an index can be given as argument to the RESTSoft GEANT4 application to be combined with the time in order to produce the seed. Thus a script can send simulation tasks with different arguments to the cluster’s queue in order to produce independent events. 6.2.5 Simulation control. Our main goals are to save events which deposited energy in the sensitive volume inside a predefined energy range of interest (RoI), but also to evaluate the probability of such an event happenning. Apart from that, we could also need to reconstruct the trigger rate (probability of events which deposit energy in the sensitive volume above a threshold). In principle, at the time we are going to launch the simulation, these probabilities are completely unknown. There are basically two types of conditions in order to decide if a simulation has been completed. The limitation of the number of stories to simulate is the usual one when there is a high probability to have a count (an event registered inside the RoI) per simulated story, what we call TIME end condition. This is usually the case when we are directly simulating the signal, for example, if we are simulating a calibration. In contrast, if there is no previous indication about the ratio of stories needed to have a count, as it happens typically when investigating the influence of any factor on a background, one would ask for a minimum acceptable amount of 6.2. The RESTSoft GEANT4 application. 121 counts before finishing the simulation. That is called STATISTICS end condition. By means of its configuration file both end conditions can be established and the application is able to execute the most restrictive one. The DataOut class is in charge to apply this criteria and reports the general progress of the simulation. However, many times the simulation is not finished by none of the previous conditions. During very long simulations it happens frequently that any external cause stops the program: a surprising shut down, an unexpected error or simply that our time is finished in a cluster queue. In consequence, DataOut must save in disk the selected events as they are produced, with all the information necessary for the partial simulation to be absolutely operational. The point is basically to save every time the simulated number of stories. The most robust manner to do it is to save together with each event the number of stories that were simulated between it and the previously saved event. This way the total amount of stories can be recovered at any time with absolute precision, even it the final number is very large. During the course of long simulations checking their correct development is also necessary. The application has an adjustable verbosity level and can print partial information every time a story is simulated, an event is saved or just periodically (or it can be even more detailed for debugging purposes). Every (also adjustable) stories period, a picture of the simulation partial result is saved in the form of a spectrum with the energies deposited by the events in the sensitive volume. The last bin accumulates the events which deposited energy over the maximum defined for the spectrum, which allows to reconstruct the trigger rate. When very few events are saved in the course of a simulation, our strategy of saving the number of simulated stories together with the saved event can fail. Even a simulation with no saved event is giving relevant information about the rareness of such an event. Because of that reason, every stories period the number of these periods which have been completed is saved in the same text file and, if necessary, it would allow to recover the statistics with the precision of the stories period. In contrast to the individual event data recording, the spectrum routine is not implemented in DataOut but in the main, since it does not need to follow the simulation evolution and just works periodically and at the end of the simulation. The range of interest (RoI) is defined in the configuration file by a lower and an upper limit. The simulation can be interrupted if there is no possibility to produce an event in the RoI, i.e. if the energy already deposited in the sensitive volume is over the upper limit or the remained available energy is not enough to reach the lower limit. This way the simulation is accelerated. That is specially siutable for the simulation of double beta decay, where there is a narrow RoI in a relatively high energy region. The same example is good to illustrate the following possibility: counts out of RoI can still be interesting in order to reconstruct the context. However the space needed in disk would be used for uninteresting events (which usually are many more than the interesting ones). In the configuration file a proportion of these events can be defined to be saved by chance, that is called the Russian roulette. 122 Chapter 6. Simulations for Rare Event Searches with TPCs Figure 6.6: Elemental sketch of the RESTSoft GEANT4 application. G4RunManager is the GEANT4 class that coordinates the rest of them. The configuration options loaded by the TPCManager are available at any point of the program. Together with TPCManager, ProbGen, TPCHits and DataOut are the only classes that do not inherit from a GEANT4 prototype. 6.2.6 Information management. Study of cases. The GEANT4 SteppingAction class gives the opportunity to write code that will be executed every time a new step of the Monte Carlo has been produced. It is at this level where most of the decisions about information management are taken. SteppingAction works in close collaboration with the TPCHits class, which holds the event information. It does not only contain the event description information, as it was explained in the section 6.2.1, but auxiliary information about the course of the event simulation. For instance, the total amount of energy deposited in and out of the target, relevant information about previous steps (as the initial vertex), and some number of flags that the user-developer can define. Combining this information and configuration settings, the program decides, for example, if to abort the event. Step by step, hit by hit, SteppingAction is gathering the event description and TPCHits holds it. Once the event simulation finishes, DataOut (see previous section) examines the TPCHits content and decides if it will be saved in disk, before resetting the class. In the case of an event generated by radioactive decay (or natural radioactivity simulation of detector materials) the game between the triangle SteppingAction, TPCHits and DataOut is more complicated because, in GEANT4, all the radioactive chain is simulated as a single event. The consecutive decays can be separated by 6.2. The RESTSoft GEANT4 application. 123 seconds or years and then (if producing a count in the target) they must be recorded as separate events. At the same time, it must be registered that they have been originated by the same primary decay and save at which level of the chain (the name of the father) this happened. Beyond the basic description as a hits collection (explained in section 6.2.1) and the vertex that originated the event (number, state and nature of initial particles), less obvious information can be relevant for a deeper understanding of the causeeffect relationships which drove to the result of the simulation. This information is usually related with what happens out of the target volume, however all the hits produced in the whole geometry cannot be systematically saved. The interesting cases must be classified in advance and recognized on-line by SteppingAction with the help of the information recorded in TPCHits. The idea is to put a label in the event that would mean something so complex as: “this event happened in the TPC because an incoming photon suffered a photoelectric in that part of the shielding and the resultant electron produced intense bremsstrahlung”. These labels, together with a brief description, allow to try a study of cases, which many times eloquently resumes our understanding of the Physics that dominates in the detector. Figure 6.7: RESTSoft GEANT4 event model. In analogy with the hit, that is used to describe the point-like energy depositions inside the sensitive volume which draw the track in the TPC; the interaction is defined to describe the highlights that happened in, or around, the sensitive volume, before or after the energy deposition on it, and draw the cause-effect chain which may explain the presence of the count in the spectrum. An interaction is described by the following fields: number codes for the particle, the place (a defined volume in the simulation) and the physical process of the step more closely related with the 124 Chapter 6. Simulations for Rare Event Searches with TPCs interaction and the kinetic energy of the particle. The circumstances which define an interaction can be more complex than simply a special step; an interaction type label will simplify notably the off-line analysis. Two more auxiliary fields are still available, and their meaning can be differently defined for each interaction type. Enabling this tool needs a considerable effort and skill by the user-developer. It has been far developed for the NEXT and CAST simulations and the current appearance of the general SteppingAction class is that of a template for building a new study of cases or labelling rule. Many of the types are quite general, in particular the ones which define the three common ways an event is produced inside a TPC in low energy physics: a photon produces a photoelectric in the gas (type 5), a photon suffers Compton scattering in the gas (type 6) and an electron directly bursts in the gas (type 1). 6.2.7 Direct applications of the RESTSoft GEANT4 application. CAST micromegas quantum efficiency and HPXeTPC spectrum. The RESTSoft GEANT4 application is the perfect example of an external program related with RESTSoft. In order to extract all the information the simulation gathered an interface (what we called bridge class) between its output and the RESTSoft classes has to be used. The GEANT4 application has been the origin of most of the simulation works done inside TREX; besides, it has shown enough versatility and the capacity to produce results independently. In many cases dealing just with the information we described in the former subsection, spectra and probabilities, can be enough. Researchers that may be not interested in complex simulations can take advantage of the GEANT4 application alone and deal just with simple objects like spectra saved as texts files. In this subsection two examples of use of the RESTSoft GEANT4 application, which have physical meaning by themselves, are showed. 6.2.7.1 CAST micromegas quantum efficiency. The efficiency of the detector is the probability of an event, in the case of CAST (see section 8), an X-ray coming from the magnet’s bore, being registered and positively identified. The most fundamental aspect of the efficiency is the probability of an X-ray interacting in the detector, it is called quantum efficiency because involves the cross section of the X-ray with the gas atoms. Moreover, the probability of the photon surviving from the cold-bore to the detector gas must be evaluated. In the case of the CAST sunrise micromegas docking place there is a trip longer than one meter and a half which is irrelevant, because the pipe is under vacuum. However the photon has to cross three windows that separate different volumes, and they compromise the quantum efficiency. Two of these windows are reinforced with metal grids, called strong-back, because they separate regions with different pressures: the cold window between the cold-bore (which in CAST phases II and III contained 4He and 3He) and the vacuum pipe, 6.2. The RESTSoft GEANT4 application. 125 and the drift window between the vacuum pipe and the detector gas, which is in overpressure (just for increasing the quantum efficiency). X-ray are blocked by the strong-back bars, so that introduces a geometrical loss for the efficiency. In addition there is the probability of the photon being absorbed or dispersed in the window material. The drift window is an aluminized mylar foil of 5 µm thickness (plus negligible aluminium thickness), the cold-bore window is 14 µm polypropylene thickness and the differential window, which is placed in the middle and has no strong-back, adds 4 µm extra of polypropylene. Finally the detector geometry and the gas conditions, basically the gas mixture (Ar with 2.3% iC4H10) and pressure (1350 mbar) are essential; but it regards also the definition of the sensitive volume. The latter concerns to border effects and the probability of secondary photons escaping the detector. They are important because events that deposited less than the photon initial energy will not be accounted for positive detections. Using the GEANT4 geometry model for the CAST micromegas detector exposed in section 9.1.1 with addition of the two windows of the vacuum system with the help of the RESTSoft GEANT4 application we can calculate the CAST micromegas’ quantum efficiency. A general method from application’s Event Generator class is used to launch the photons in perpendicular direction to the micromegas plane (as they would come if originated by solar axions) homogeneously scanning the cold-bore area with random energies between 0.5 and 10 keV. Imposing that the event deposited energy in the gas is the same as the photon initial energy, a graph proportional to the quantum efficiency is obtained directly from the spectrum output of the simulation. In figure 6.8 right several curves show the simulation step by step. First, a calculation of the chamber’s conversion volume efficiency. This curve is almost coincident with the one we can just calculate from NIST cross sections database[138] for the argon using the gap length and gas pressure (see the introduction of section 3.2.1). The only notable difference can be detected above 3.2 keV, where the K-layer of argon is placed. The difference is reflecting the probability of a secondary 3.2 keV photon from argon fluorescence escaping the chamber (see section 3.2.1.1). It can be observed how Quantum Physics limits the efficiency at high energies. Only replacing the gas, increasing its pressure or upgrading the chamber size could improve it. The rest of the curves are obtained by consecutive addition of the drift window and the other two. In the figure 6.8 left we have the contour map obtained for the complete simulation, showing the shadows of both superposed strong-backs. However, similarity between curves at high energies is telling us that the strong-backs do not represent a big disadvantage. We see now how the windows accumulated thickness is the factor which limits the efficiency at low energies. This way the GEANT4 application alone, just by drawing the output spectrum, has been able to produce one estimation that is capital for the CAST experiment. 126 Chapter 6. Simulations for Rare Event Searches with TPCs Figure 6.8: Left: readout projection of the interaction position where the shadows of the two consecutive windows grids (strong-back) are superposed. Right: “step by step” simulated quantum efficiency and comparison with NIST XCOM database[138]. 6.2.7.2 Reproduction of one spectrum in the HPXeTPC. Another example is the planning of a test in the Zaragoza’s high pressure test chamber. This vessel is designed for high pressure, leak-tight and low out-gassing, since it was thought to test pure noble gases[107] (HPXeTPC, high pressure xenon TPC). The idea is to measure energy resolution for medium energy gammas (higher than the usual 6 keV from 55Fe or 22 keV from 109Cd) in pure xenon at high pressure in order to do a more feasible extrapolation to the double beta decay Q value at 2448 keV. A limitation for that is the efficiency of the chamber whose conversion volume of 6 cm maximum drift distance is defined by rings of 8 cm inner diameter, while the size of the micromegas prototypes installed there is usually only 3 cm diameter. These dimensions can be used in the REST GEANT4 application to simulate such a chamber filled with pure xenon at 4 bar (see figure 6.9). To obtain the graphic in figure 6.10 a mono-energetic isotropic point-like source placed on the top of the conversion volume (see figure 6.9) has been simulated for γ energies between 10 and 100 keV. Like in the former example, the output spectrum is enough to define the efficiency. The green solid line in the plot corresponds to the calculation one easily would do (see section 3.2.1) from NIST cross sections database[138] considering the photoelectric probability, γtrajectories orthogonal to the read-out/cathode planes and the maximum drift distance. The simulation will also take into account the solid angle and the electrons finite path range . In addition, from the simulation we distinguish the detection efficiency (possibility of interacting in the sensible volume limited by the micromegas surface) and the photo-peak efficiency (possibility of deposition of the whole energy in the sensitive volume, so contributing to the photo-peak and having a really useful event). At low energies both efficiencies are almost coincident since the photoelectric cross section is by far dominant compared with Compton (see figure 3.14 right). On the contrary, 6.3. The RESTSoft TPC simulation chain and Event Model. 133 Figure 6.12: Evolution of a 60 keV electron event along the steps of the RESTSoft TPC simulation chain. Up: the electron path as registered by the GEANT4 application. All the hits that compose the track do not have to give rise to clusters, many of them are just marks to help to visualize the electron trajectory. The primary charge distribution is more evident after the addition of diffusion displacements (figure in the center). For instance, the charge accumulation at the track end, many times called blob, is clearly identified. Even an ideal read-out (1 mm pitch identically for the three dimensions), could not aspire to compose an event picture better than the representation of the figure at the bottom. The dots are 3D pixels (see section 6.3.5) that were activated and the boxes have a size proportional to the pixel signal. All the drawings were produced from a TRestPhysEvent. 134 Chapter 6. Simulations for Rare Event Searches with TPCs 6.3.4 Mesh transparency and micromegas amplification. Although the hits’ final position zafter the drift simulation was not explicitly drifted to the mesh, the electrons are supposed to pass through the micromegas mesh and, if they survived to the entrance, been amplified, one by one, in an avalanche process: that is the work done by the TRestMM (MicroMegas) class. The way that avalanche products are gathered inducing signals in the readout depends on the micromegas specific design, X, and should be implemented in a specific method for a new event container class, TRestDaqX, which can differ fundamentally from TRestPhysEvent. However, the main methods for amplification and mesh transparency are common and therefore TRestMM is a general auxiliary class. In fact TRestMM can transform a TRestPhysEvent into a second one which reflects the effects of the micromegas transparency and/or amplification. The mesh transparency is simulated by direct interpretation of the mesh transparency curve (see section 4.3) as the electron probability passing through the mesh into the amplification gap. This curve can be defined by the parameters of the associated fit formula (4.15); the values from figure 4.11 are loaded as default; or through file, by introducing experimental series that will be fitted. Once the fields ratio FR =Edrift/Eamplification is given, the probability would be equal to the unity if FR < FRend, being FRend the FR at which the transparency plateau ends (defined as the point for which the fit equals to the unity); while if FR > FRend it is evaluated from the fit. For simulations of the individual avalanches the Polya functions (see section 4.4.2) are implemented. The parameters they need, concerning the average gain and the relative variance for the individual avalanche, can be given directly or alternatively TRestMM can estimate them from a macroscopic reference, like the FWHM measured for a peak or a gain curve exponential fit. Another possibility is calculating them from the Townsend coefficient making use of the Legler-Alkhazov model (see section 4.4.2); TRestMM can reconstruct the electric field dependence of these parameters relying on a TMagtbolzGas and produce theoretical gain and resolution curves as those plotted in figure 6.13 (see section 4.4.2.1). Alternatively, the avalanche can be simulated with a Gauss distribution for which some primary electrons can even been collected in groups. This practice would dramatically reduce the simulation time consumption. The avalanche must be modelled with the Polya functions, but the question is when do they produce results that are essentially different from those obtained with the much “cheaper” Gaussian model. We tried to illustrate that in figure 6.14 which shows distributions of avalanche yields summed for several values of N, the number of initial electrons. The avalanches are obtained from Polya functions using m= 2, relative variance b= 1/m, which, we can see from figure 6.13 right, is actually an unfavourable case, i.e. with single avalanche distribution very different from a Gaussian, because typical values from argon with 2% isobutane lie in the range 2< m < 3.5. It can 6.3. The RESTSoft TPC simulation chain and Event Model. 135 Figure 6.13: Gain (left) and resolution (rigth) curves obtained by the TRestMM class using MAGBOLTZ resuls for Ar with 2% isobutano at 1 bar. Figure 6.14: Distributions of the sum of Navalanches yield with single avalanche simulated with a Polya(m= 2,¯ N= 100). be clearly appreciate how the Central Limit Theorem of Statistics works and the accumulated avalanches yield distribution approaches to a Gaussian as Ngrows up. It could be said that, if the number of primary electrons collected by a pixel (or any kind or readout segmentation unity) is higher than 20, the expected signal strength distribution has a Gaussian behavior. The only case it can be worth simulating the proper Polya distribution is that of a low energy signal which triggers a high multiplicity event, as can be the passing of MIPs or a directional Dark Matter TPC. 6.3.5 Pixelization. The last goal of the simulation is often to produce the information that would really be available, even though in an ideal TPC, in order to make a meaningful analysis. That would require a full description of the read-out working and electronics (see 136 Chapter 6. Simulations for Rare Event Searches with TPCs section 6.4 and chapter 9for an example of such kind of approach) which may have not being defined at all, or simply be beyond our interest. However, at least restrictions in 3D spacial resolution are usually imposed. This position assumes that the TPC read-out has ideally reacted to the stimulation (the drifted TRestPhysEvent) and, in addition, the off-line analysis has properly reconstructed the event as a 3D charge distribution, but with a limited resolution. The TRestPixel class is often the last link in the RESTSoft TPC simulation chain and, again, works performing a transformation in the necessary TRestPhysEvent. A 3D net is built from the definition of a 3D pixel which is delimited by the three independent dimension resolutions dx,dyand dz. The new hits have the position of the 3D pixel center and the charge contained inside them. Because small pixel sizes are frequently used, directly building such a 3D net would cause memory troubles, the algorithm works with a different philosophy that is slower but optimizes memory consumption. It starts with the first hit of the input TRestPhysEvent, it identifies the associated 3D pixel and creates the first hit of the output TRestPhysEvent. It goes on, hit by hit, gathering the charge for this pixel until it recognizes a new 3D pixel has been invaded, and then adds a new hit to the output event. Every time this happens, the algorithm has to check that the new 3D pixel had not already been activated and, in that case, just adds new charges to it. This method has also the advantage that the chronological order of the pixels is conserved, what certainly is not the kind of information available in a common TPC, but can be of interest for some kind of investigations. The resulting event is once again a TRestPhysEvent, a collection of hits containing elementary charges, but with regularly restricted 3D positions (see figure 6.12). The amount of hits has been dramatically reduced, and so the disk space and also the computing time in later analysis; but the effects of diffusion simulated by each independent electron are fully reflected. At this point it must be clear why working with TRestPhysEvent, or other classes which inherit from it, is desirable: this way the TPC simulation chain classes can directly deal with it. Finally, there must be a program where these methods sequentially act on the data; following the RESTSoft event model philosophy data from different origins can naturally be mixed and compared. The program can load data from some simulation and complete the data whole transformation directly for its analysis (see sketch in figure 6.15). Besides the chain can be implemented so as to generate data tree-like structures, for instance, from one simulated event a set of drifted events can be produced in different drift conditions or gases; and each subsequent one, can give rise as well to a new set where different read-out capabilities are tried (see figure 6.16). In each step all the original information can be transferred to the later one by using a big class (always containing a TRestPhysEvent) or, on the contrary, a different class can be used for description of each step event with only the relevant information at this point (perhaps only the TRestPhysEvent), and one can always browse between 6.4. Simulation of electronic signals. 137 Figure 6.15: Sketch of a program implementing a serial TPC simulation chain. the associated events. In the next chapter some examples of these kind of uses are described. Figure 6.16: Sketch of a program implementing a tree-like TPC simulation chain. 6.4 Simulation of electronic signals. Sometimes the experiment analysis works at lower level than that of reconstructed events, for instance to apply pulse shape analysis. That forces the simulation to perform a new step in order to generate the electronic signals induced in the read-out electrodes. This way a new reconstruction algorithm can be trained with simulated 138 Chapter 6. Simulations for Rare Event Searches with TPCs low level data, or an already existing analysis can be indistinctly used with both experimental or simulated data for a direct comparison (see chapter 9). The number and features of the induced signals will depend on the specific readout design and electronic set-up, X. That is the end of the TRestPhysEvent class domain, which must give place to a specific holder class TRestDaqXevent. However, the basic process is again common: a certain time distribution, which describes the arrival of the charges to or near the electrode, should generate an electronic signal, which could be integrated (giving a number) or recorded (producing a pulse). TPulseShaping is the class which is able to generate an electronic pulse from a given input intensity time signal, therefore it simulates the work done by a preamplifier followed by a time-amplifier, or shaper; i.e. the typical read-out electronic chain. The class was adapted to RESTSoft from an original program by Javier Galán. It has two alternative signal generation algorithms: pulse generation by solving the associated virtual RLC circuit or pulse construction by convolution with the typical amplifier response function. The former is faster and has the advantage that all their parameters are real numbers which can be continuously adjusted for fitting, while the latter is more intuitive and easily predictable and can be assigned to a generic type of shaper. 6.4.1 Generation of electronic pulses by RLC circuit solving. This method was already written in the original routine by J. Galán and is deeper explained in [187]. TPulseShaping deduces the output pulse v(t)by solving the second order differential equation that describes a virtual RLC circuit: 1 ωC d2v(t) dt2+dv(t) dt+ωLv(t) = Rdi(t) dt(6.6) where i(t)is the incoming intensity pulse, closely related with the distribution of the time arrivals of charges to the electrode, which is the input to TPulseShaping. The algorithm parameter Rcan be absorbed by the electronics gain, which is a configuration parameter of TPulseShaping. The other ωCand ωLcome from the characteristic times of the circuit and determine the pulse shape; the former has a direct effect on the rise of the pulse, while the second in its fall. J. Galán composed tables with the values of both frequencies which fit the pulses produced by 55Fe X-rays (in principle, point-like events with ideal i(t)) in a CAST micromegas detector for the different settings of an ORTEC 474 Timing Amplifier. These tables are included in TPulseShaping, but they are only valid for the same electronic chain used during the test: an ORTEC 142B Preamplifier followed by the previously cited amplifier. 6.4. Simulation of electronic signals. 139 6.4.2 Generation of electronic pulses by convolution with the response function. The output signal v(t)can be produced from the stimulation signal i(t), induced in the electrodes, from the characteristic response of a generic time amplifier, or shaper. In this section we follow chapter 6 of [116]. For a linear and causal system: v(t) = Zt 0 w(t−t0)i(t0) dt0(6.7) where w(t)is the delta response of the system, i.e. v(t) = w(t)if i(t) = δ(t)the Dirac function; in other words, it is the response of the shaper to an standard fast pulse. The delta response is customarily factorized as w(t) = gh(t)with h(t)the delta response normalized to have height the unity. Thus gis called the sensitivity of the charge amplifier because, if i(t)is narrow compared with h(t), then v(t) = gZt 0 h(t−t0)i(t0) dt0≈gh(t)Zt 0 i(t0) dt0=gh(t)·Q(6.8) and the peak value of the shaped signal is proportional to the total charge induced in the electrode. TPulseShaper allows to define the delta response functions of ideal n-degree unipolar and bipolar shapers (which are plotted inf figure 6.17): huni(t) = ent nτ n e−t τΘ(t)(6.9) hbi(t) = er √nn−t τt rτ n−1 e−t τΘ(t)(6.10) where τis the shaping time constant, r=n−√nand Θ(t)is the step function so as to assure the causality of the system. These functions peak at tp=nτ (unipolar) and tp=rτ (bipolar, with n > 1). It can be more convenient to parametrize them as a function of their peaking time constant (an alternative definition also considered in TPulseShaper): huni(t) = t tpn en(1−t/tp)Θ(t)(6.11) hbi(t) = 1 √nn−rt tpt tpn−1 er(1−t/tp)Θ(t)(6.12) The sensitivity, g, coming for the ideal circuits is not presented as it is, in practice, absorbed by the gain of our virtual shaper. Regarding the parameters’ settings, tpis obviously related with the rise time of the pulse and is the shorter semi-width of the response function; therefore it is 140 Chapter 6. Simulations for Rare Event Searches with TPCs Figure 6.17: Left: unipolar (n = 2) and bipolar (n = 4) delta responses. Rigth: unipolar delta responses for several values of n. the time reference to compare with the duration of the stimulation signal which, if comparable to tp, can make the approximation 6.8 not valid and the condition of charge amplification would be lost, what is called ballistic deficit (see figure 6.18 left). The asymmetry of the pulse is related with n, more Gaussian the peak as higher the n(see figure 6.17 right). Figure 6.18: Examples of i(t)and v(t)drawn together after simulating an unipolar shaper with n = 2 (see figure 6.17) with tp= 200 a.u. Gaussian distributions are used as the typical shape of signals induced by point-like clusters after drift diffusion. Left: shaper response to two gaussian clusters with the same area (charge) equal to 0.5 a.u., but different widths (σ= 50 and σ= 25). The second stimulation pulse is fast enough and the shaper pulse has the proper height value, while the first one is much shorter due to the ballistic deficit. Rigth: the same simulated shaper dealing with a more complex event. 6.4. Simulation of electronic signals. 141 6.4.3 TRestShaper. TPulseShaping is a relatively low level class, derived from a simple program, with just the pulses as input and output. It is desirable that it keeps being so simple, however it is convenient to have a more powerful interface to join the simulation chain. TRestShaper can control TPulseShaping incorporating more advanced options (virtual shaper description, noise addition, trigger position, pulse array preparation, etc.), using TRestConfig for shaping configuration by file, dealing with holder classes like TRestPhysEvent or TRestDaqXevent, instead of only pulses in arrays, or with an intermediate preamplifier pulse. Figure 6.19: Left: an initial 3D charge distribution (from a drifted TRestPhysEvent), with a long time structure and the resulting pulse (rigth) using TRestShaper. The drift time distribution of the primary charges takes only the first 1 µs long; TRestShaper configuration is producing a quite bad signal to noise ratio. With a short shaping time of 50 ns (optimized for X-rays), the simulated shaper suffers of severe ballistic deficit before that MIP signal, and only a casual charge accumulation in the center of the event 3D structure (likely a δ-ray) was able to generate a peak. 10.2. Preliminary tests at surface. 245 2009 campaigns. Procuring a high efficiency at low energies was aimed because of the interest in having the best sensibility for the underground background at very low energies. For tests carried out in different conditions the analysis settings will be specified. A Slow Control system (SC) has been developed based in the general purpose National Intruments I/O device via USB[213], which has been connected a temperature and pressure sensors used to continuous monitoring of the room temperature and the micromegas chamber pressure. Also intensities associated to the drift and mesh voltages are read with nA precision, and the SC can send an inhibit signal that would switch off the power supply when considered convenient. The SC can be completed with an AlphaGUARD radon monitor[214] that, by means of extracting a continuous flow with a little pump, is able to record the radon concentration inside the Faraday cage. Several existing differences with SRMM must be summarized. There is no cadmium layer in the test set-up and the lead of the internal shielding is not archaeological. Since there is no necessity of an outlet for connection to the magnet, the test set-up is basically closed in the 4πsolid angle with exception of a narrow gap for the calibrator, the unavoidable outlet for the detector’s raquette neck (also present in the sunrise) and a weaker zone in the join between the top an bottom layers with the rest of the shielding (that is easily appreciated in the photo of figure 10.1 right). While in the SRMM the immediate environment (with direct access to the chamber through the mylar window) is made of stainless steel, because there is the pipe to the magnet’s bore; in the test set-up the chamber is open to the innermost cavity inside the shielding, all made of copper since its walls are composed of the internal layers of the shielding and part of the Faraday cage. This difference is emphasized by the fact that the cathode, which is the frame of the calibration window, was made of copper for most of the tests done at Zaragoza’s test bench and at Canfranc Underground Laboratory. Furthermore, the contribution of a more radioactive cathode was measured lately and is reported in section 10.3.2.1. 10.2 Preliminary tests at surface. During the commissioning of the new set-up at Zaragoza’s lab, several tests were carried out with the goal of studying the dependence of the background on operational or environmental parameters. Despite the M13 microbulk detector used during that period had not the best performance and only a little range of operational parameters were allowed, its background level was comparable to CAST’s[215] (Run reference, R0) and its evolution completely coherent[216]. The set of changes which had no influence on the detector background, with a CAST-like set-up (only the shielding described in the previous section) installed at surface (a report was given in [217]) were: •Nitrogen flow interruption. 246 Chapter 10. Surface and underground experimental tests of the CAST micromegas background. •Micromegas gain. Increased about a factor 2; provided there always was a good signal-to-noise-ratio (SNR). •Variations in micromegas chamber pressure: 1.3–1.5 bar. •Changes in timing-amplifier integration times: 50–100 ns. •Upgrade of the external polyethylene shielding from 10 cm thickness and partial coverage to 20 cm thickness and 4πfull coverage. •Variability of environmental temperature: ∆T > 10◦C; it affects the gain stability, but not the background. Only settings that degenerated seriously the detector performance had a response in the background level. The increase of the drift voltage to 1000 V, far from the transparency plateau of the micromegas, led to an increase of the background level higher than 50%. The level decreased again when the mesh voltage was increased, partially compensating the lost of gain and transparency. The hypothesis of the drift voltage itself to originate background events by generation of micro-sparks in the cathode was in this way rejected (up to the sensitivity of CAST-like background levels). Many of these tests have been eventually reproduced underground in much more stable conditions and at lower background levels with the same result. 10.3 Underground operation. Since 1985 the Canfranc Underground Laboratory (LSC)[218] is scenario for Rare Event Searches mainly carried out by the Nuclear Physics and Astroparticles group (LFNA) of the University of Zaragoza. The laboratory is situated at 2500 m.w.e. (meter water equivalent) under the Tobazo peak, in the point with larger rock overburden of the railway tunnel. Recently the laboratory was upgraded; though the first part of the tests explained here were still carried out in the facilities of the (now called) old lab. The cosmic muons are known to be reduced by a factor of 10−4[219] in the LSC. Then an installation of the set-up should imply an evaluation of cosmic rays contribution to the background. The complete scenario is not so simple, because the transfer to the LSC implies more differences. The environmental γand neutron fluxes can vary notably, and the radon concentrations even orders of magnitude. Therefore we postpone that discussion until the next section 10.4. By the moment it is just commented that the direct transfer of the CAST-like set-up, equipped with the M13, from Zaragoza to Canfranc produced a very similar background level (see figure 10.2 comparing the runs R0 and R1), though the trigger rate was reduced from ∼1.0 Hz to ∼0.2 Hz. That fact confirms that cosmic muons produce most 10.3. Underground operation. 247 Figure 10.2: M13 background spectra taken at Zaragoza’s test bench (R0) and at LSC (R1). of the triggers and micromegas detectors to efficiently reject them. Undoubtedly cosmic muons were not the dominant contribution to the CAST background. Besides, from the calculations exposed in section 9.2.5.3 it is deduced that environmental radon contribution would still be smaller or even negligible compared with γ. Regarding radon very close to the detector, it can be avoided by flushing nitrogen and its contribution will be evaluated in section 10.3.2.2. What is clear is that environmental γflux and radon contributions can be practically blocked by a shielding upgrade and that contribution can be evaluated directly from a previous CAST-like underground reference, as it is done in the next section. The LSC environment provides more advantages for this study: the stable environmental conditions reject possible systematic effects and the environmental γ radiation has been measured by the new laboratory characterization program[219]. 10.3.1 Thicker shielding tests. Reaching set-up’s limitations due to intrinsic radioactivity. The measurement series summarized in this section have the goal of evaluating the relative weights from internal contamination and external radiation to the CAST background by making the CAST-like shielding evolve to an underground-like setup. The M10 detector, a microbulk micromegas with better performance than the M13, comparable with detectors at work in the CAST experiment, is used until almost the end of the series. The detector is mounted with a copper cathode with the same strong-back mesh design and mylar thickness window used in CAST. In two steps the original CAST-like set-up was surrounded by 20 cm of clean, but non archaeological, lead; first leaving the top with only the light shielding (R3) and 248 Chapter 10. Surface and underground experimental tests of the CAST micromegas background. finally closing all the solid angle (see figure 10.3). The progressive decrease of the background confirms the cause is the improvement of the coverage (see figure 10.4). Figure 10.3: Three different configurations of the external lead shielding: only CASTlike, i.e. 2.5 cm lead layer (the photo also shows the automatic calibrator external motor and electronic box and the feedthroughs used to extract the signals from the Faraday cage); 20 cm thick extra lead layer without closing the top of the lead castle, 5π/6coverage (the photo includes the electronics rack) and completing the external shielding, 4πcoverage, (the photo shows the addition of a plastic box and the nitrogen dewar). All the configurations are provided of continuous nitrogen flow. Figure 10.4: Evolution of the final background level (averaged from 2–9 keV, so including the copper fluorescence peak) of the M10 detector in the LSC along different set-up upgrades, from R2 and R3 to R4 and R5, including the period in between where probably a contamination of the Faraday cage limited the background. As a test done before the internal copper layer installation, the replacement of the 55Fe source by a 109Cd one during one week produced a background increase. Consecutive increases of the lead layer drove to an important reduction of the background (a factor of 6) but a impressive fall of the trigger rate (a factor of ∼40). Then the interactions in the chamber had almost exclusively an external origin. 10.3. Underground operation. 249 The spectra obtained during the first two phases of shielding completion shown a peak in the 5-6 keV region which remains independent of the external shielding (see plots in figure 10.5) and finally limits the background decrease to the level of 10−6cts keV−1cm−2s−1. The presence of such a peak introduced again the discussion about the origin of the 5-6 keV peak observed in CAST detectors. The simulations had explained the peak in CAST sunrise was produced by fluorescence of the stainless steel pipe (see chapter 9.2) and so it must have an external, but physical, origin no related with neither detector nor with analysis’ systematics. Weeks of tests were devoted to understand the origin of the peak at LSC[220] until, finally, a 1.5 mm copper layer was inserted so as to close tighter the detector cavity (see figure 10.9), then the background dropped demonstrating its external origin. Since such a weak shielding could only block soft X-rays emissions the analysis performance remained out of doubt. The origin of the peak was attributed to a contamination of the Faraday cage inner surface. Figure 10.5: Evolution of background spectra with shielding upgrades. An existing 5–6 keV peak, that was almost not visible at CAST-like background level, became relevant after the first upgrade of the external shielding and it started dominating the background level once the external shielding was completed. The peak disappeared when a thin copper layer was used as is shown in figure 10.9. 10.3.1.1 Ultra-low background underground. The addition of the thin copper layer led immediately to a sharp reduction of the background level (see last step in background progression plotted in figure 10.4) and a spectrum dominated by a copper fluorescence peak. The whole ultra-low background period (ULBP), contains R4+R5+R6+R7; see the spectrum in figure 10.7. The ULB was stable along almost six months of measurements, robust respect to electronic conditions and reproducible (see figure 10.6); it was twice interrupted 250 Chapter 10. Surface and underground experimental tests of the CAST micromegas background. to perform two tests (R9 and R10, see next section 10.3.2) and recovered again. Besides some details are worth to be remarked and analysed. Figure 10.6: Background evolution during the ULBP (R4+R5+R6+R7) showing two rises due to the installation of an aluminium cathode (R9) and the interruption of the nitrogen flow (R10). The levels are averaged in the 2–9 keV range, including the copper peak. Figure 10.7: Global spectrum obtained during the three ULBP’s (R4+R5, R6 and R7) showed in figure 10.6. It is mainly composed of copper and argon fluorescences. When plotting the position in the readout plane of the final events one by one, two accumulation zones, or hot spots, clearly appeared in the hit-map (see figure 10.8 left). It is understood that they are not physical events but due to micro-sparks related with any kind of defect or trouble of the micromegas in these particular points. Nevertheless they accounted for 1/3 of the total amount of counts during the first partial ULBP (see 10.8 right) in such an small detection area. They were 10.3. Underground operation. 251 rejected for the final background level calculation. In fact, after opening the chamber for the cathode replacement, they vanished. However, it is very interesting for the understanding of the detector background limitations the fact that these micro-sparks events can evade the CAST discrimination criteria. Figure 10.8: Partial results corresponding to the first ULBP. Left: events hit-map, where two accumulation zones (hot spots) are clearly detected. Right: background spectrum, the histogram distinguishes the events in the hot spots. These are distributed uniformly, while the rest of them tends to be accumulated in the copper fluorescence peak. Figure 10.9: Two inner (inside the internal shielding) configurations. First (on the left) the ionization chamber is closed by a 1.5 mm thick copper layer. Second (on the right) the internal cavity is closed from its top. The first one was adopted only during the first weeks of the first ULBP, until March the 17th. 252 Chapter 10. Surface and underground experimental tests of the CAST micromegas background. The mentioned thin copper layer was used to compare two different internal configurations, that divide the first partial ULBP into two periods, labelled R4 and R5. In a first configuration the copper layer was used to close the chamber’s window, leaving a small hole for calibrations (R4, figure 10.9 left). In a second one the copper layer was moved up to the calibrator level, this way the calibrator gap was blocked and the Faraday cage top screened while the window is kept open to the interior of the inner shielding cavity (R5 and the following runs, figure 10.9 right). The second configuration has been adopted for the rest of the measurements. A relatively larger copper peak found in the first configuration (see comparison in table 10.1) might be the unique difference between both internal configurations’ background (taken with similar exposure, see the summary table 10.7). This is very interesting because both set-ups are not physically equivalent at all. That fact points to the idea that the radioactivity which causes the final background should be present in the detector itself. If the radioactivity were present, for instance in the lead walls, one would expect the interactions of γemissions with the inner copper walls of the detector cavity to have an important contribution, because it has been shown that fluorescence are a very efficient via to produce final counts in this detector, since it is specialized in X-rays. However, when the copper layer closed the window, blocking the likely innermost shielding walls contribution, the effect on the background level was small2. Cu layer on strong-back Cu layer on top All the ULBP 2-7 keV 1.5±0.9 1.6±0.9 1.5±0.5 2-9 keV 1.8±0.8 1.7±0.8 1.7±0.4 Table 10.1: Comparison between both internal configurations during the first ULBP. Final background levels expressed in ×10−7keV−1cm−2s−1and statistical errors as 2σ. Besides, the copper peak seemed to be relatively higher in the first configuration, when the copper layer increases the copper surface directly exposed to the gas chamber. That geometry can be thought to be more favourable to the detection of copper fluorescence only if the exciting radiation is originated in the detector itself. This is, in principle, an argument that reinforces the idea that the detector’s intrinsic radioactivity is the main limitation for the background level evaluated in these tests. However, the same reasoning stands for the radiation originated in the electronic components (gassiplex cards, etc.) that can reach the detector through the unshielded neck of the plexiglass base. Considering that case, perhaps the background limitation is not put by the detector, but by the particular CAST-like set-up. This is a sensible possibility to be explored in future tests (see chapter 11). Last, but not least, during the beginning of the ULBP the analysis had to 2Lately, the lead internal shielding was replaced by a copper one, 2.5 cm thick, and no background improvement was observed. 10.3. Underground operation. 253 Figure 10.10: Evolution of the trigger rate during the first ULBP. Noise in the mesh line produced huge and variable increase of the trigger rate. The noise trouble is solved in March the 9th by the replacement of a damaged HV feedthrought of the Faraday cage; the background level kept stable. deal with severe electronic noise conditions (see plot in figure 10.10). That fact motivated several changes in the electronic settings and micromegas gain. None of them affected sensibly the background level. When the noise was finally improved the trigger rate decreased to a stable 5×10−3Hz, what was called Ultra-Low Trigger Rate, without reflection in the background level. This rate seems physical since a later improvement of the SNR, by means of a an increase of the integration time of the timing amplifier, in March the 24th had no effect. The same rate has been reproduced in all the later ULBP’s when the noise conditions were good enough. A very brief analysis report about this period can be found in figure 10.11. The table 10.2 summarizes the ULBPs and contains the final numbers. It might be noted a slow, but progressive, increase of the background level, though all the partial levels are compatible due to the few statistics. Such increase could be related with a slow degeneration of the M10 detector. Another possibility could be a progressive accumulation of radon progeny during the interruptions of nitrogen flow (mainly during the set-up interventions). Two arguments favorable to the radon accumulation hypothesis are that it looks like a faster increase in the 3rd ULBP, just after one week of radon intrusion in the detector environment; and the fact that the 2-7 keV averaged background increased respect to 2-9 keV, because a 3 keV peak is a kind of signature of radon induced background (see section 10.3.2.2). meas. time (days) Bkg ([2-7] keV) Bkg ([2-9] keV) 1st ULBP 41.41 1.5±0.5 1.7±0.4 2nd ULBP 25.86 2.0±0.6 2.1±0.6 3rd ULBP 19.65 2.8±1.0 2.7±0.8 Whole ULBP 86.92 1.9±0.3 2.1±0.3 Table 10.2: ULBP summary. Final background levels expressed in 10−7keV−1cm−2s−1 and statistical errors as 2σ. 254 Chapter 10. Surface and underground experimental tests of the CAST micromegas background. Figure 10.11: Evolution during the period of figure 10.6 of the software efficiencies (top), signal amplitudes (second plot) and some observables: mesh pulse’s width and strips cluster’s size (bottom). A 75% efficienciy for both 6 and 3 keV events could always be got. The gain plot reflects sharp changes of settings and conditions which are not correlated with the background steps, the latter are always explained only by physically relevant set-up changes. Mesh and strips correlated changes in gain are related with changes in mesh HV or gas. Changes that affect only to the mesh gain, are related with timing amplifier settings, clearly manifested in the pulse width evolution. At the middle of the third ULBP the timing amplifier integration time was again reduced without neither improving background level nor increasing the trigger rate. The cluster size is basically correlated with the strips gain, but also affected by the noise. 10.4. Surface vs underground and the cosmic rays contribution. 261 Therefore about 100 Bq m−3are needed to have a contribution comparable to the ULB. Such a concentration is certainly possible for an underground lab (in the old LSC rooms concentration spikes of 800 Bq m−3has been registered, typical values are 200-400 Bq m−3and 80-100 Bq m−3for the new Hall A) but not for a experimental hall at surface (CAST area typically registers 0-20 Bq m−3). In conclusion, this study definitely rejects the possibility of radon playing any significant role in the CAST detectors background at surface, though the use of nitrogen flow is recommended for detector safety and stability. Finally the background spectrum averaged during radon exposition days is plotted in figure 10.16 right. Its most relevant feature is what seems to be a strong 3 keV argon fluorescence, and the counts accumulation at low energies. The same shape has been detected more times when accidental radon intrusions had occurred. It reminds the aluminium cathode-induced spectrum (figure 10.14, what suggests such a peak could be a signature of a cathode contamination. 10.4 Surface vs underground and the cosmic rays contribution. The cosmic rays, together with the necessity of an outlet for X-rays detection, are the most fundamental handicap for the reproduction of the underground background in a surface helioscope. However the cosmic rays effect on the CAST micromegas detectors background is subtle at the CAST levels that were achieved before the 2012 sunset upgrade, i.e. about (5-7)×10−6keV−1cm−2s−1. A first direct comparison between surface and underground backgrounds, using the M13 detector, yielded two completely compatible results (see figure 10.2). Unfortunately the M13 detector had not the best performance and its backgrounds show signs of dependence of the observable’s distributions on the energy which surely affected to the software efficiency curve (see section 9.5, in particular figure 9.51). However, that is still the unique available direct comparison, i.e. using the same set-up, electronics and detector almost at the same time, up to date. Undoubtedly CAST detectors had great discrimination capabilities for cosmic muons discrimination. The typical trigger rate is ≃1Hz at surface, depending on the noise level and the threshold position. In Canfranc, with a CAST-like set-up, it is ≃0.2Hz. That stands for comic muons as responsible of most of the interactions in the detector. The typical muon flux at surface can justify the difference, thus most of the interactions are due to cosmic muons crossing the chamber themselves. The simulations have reported how high energy electron tracks can mimic Xray signals (see 9.5.2, in particular figure 9.50), despite being very unlikely. The same must be expected for cosmic muons as MIPs, furthermore they cause approximately 4 times more triggers than electrons. Besides there is the contribution of muon-induced secondary events. The most obvious would be the generation of fluorescences in the immediate detector environment: pipe or cavity walls, cathode, 262 Chapter 10. Surface and underground experimental tests of the CAST micromegas background. lead, etc. However the simulation of cosmic muons is more complex than that of environmental γ, since it implies much higher energy particles and more Physics are involved, specially if we are interested in secondary events. It is not only that much more resources are consumed, but it is also more difficult to plan a cross-check of the results. The experimental approach can produce results easier (providing the availability of an underground lab); and that is the subject of this section. 10.4.1 Comparison of the backgrounds of one CAST detector in three locations. The robustness of M10 microbulk detector have allowed the production of backgrounds in CAST-like set-ups in three different situations along more than four years, figure 10.17 and table 10.5 are a compilation. The oldest spectrum corresponds to the last period of the 2008 data-taking campaign, when the M10 was mounted in the sunrise side of the CAST magnet. The second one was registered at Canfranc old lab, using the CAST-like test set-up. For the most recent one, the CAST-like test set-up was moved to Zaragoza, and a new acquisition, based in T2K electronics and an oscilloscope (see section 5.5) was used6. A special electronic card was built for adapting the Gassiplex-like (SAMTEC) connectors to the T2K’s (ERNIE), being the set-up able to read 192 of the 212 strips. However an interface program was written so as to reduce all the information gathered by the T2K electronics to only that the classic CAST acquisition can register. Thus data were written in CAST raw data format so as the CAST analysis routines can normally deal with them. The motivation is to perform direct comparison as independent on DAQ and analysis as possible. set-up and date run time (days) Bkg (averaged [2-7] keV) CAST sunrise (Fall 2008) 23.6 7.1±0.4 Zaragoza (Nov 2011) 2.8 4.9±1.0 LSC old lab (Nov 2010) 6.0 3.8±0.6 Table 10.5: Attempt of evaluation of the effect of the set-up and environmental upgrades effect on M10 detector background from the spectra plotted in figure 10.17. The corresponding runs are R11, R2 and R12. Final background levels expressed in ×10−6keV−1cm−2s−1and statistical errors as 2σ. The 2008 background (R11) has the typical CAST sunrise spectrum shape, with two comparable peaks at 6 and 8 keV. In comparison, Canfranc’s background (R2) have a bigger 8 keV peak, but there is no 6 keV. Apart from this, the whole background level seems to be smaller below 7 keV. There are mainly two changes in the 6Nowadays the M10 detector is at work at the new Zaragoza T-REX lab in the CAST sunset replica set-up. 10.4. Surface vs underground and the cosmic rays contribution. 263 Figure 10.17: Three background spectra registered by the M10 microbulk detector: in CAST sunrise (R11) side and, with tests CAST-like set-up, underground (R2) and at surface (R12). set-up to be considered in order to explain the background differences. The absence of the stainless steel cathode and pipe to magnet’s bore in the test set-up explains the absence of the 6 keV peak in Canfranc, as the copper cathode and innermost copper walls of the detector’s cavity of the test set-up explains the bigger 8 keV peak. Because of the 4πclosure and the absence of the pipe in the test set-up, a sensible decrease of the background is expected; according to section 9.2, a reduction between 25% and 50% could be expected. There are also environmental changes. To start with, the absence of the cosmic rays contribution. In addition, we must be aware that in the LSC the environmental γflux is known to be more intense, perhaps twice than in the CAST area. Therefore a third background is added to the figure 10.5: R12 that represents the test set-up at surface (Zaragoza’s lab). The new background level below 7 keV is similar to CAST, with the exception of the 6 keV peak. It is higher than in Canfranc at low energies, but similar at medium energies, with exception of a little 5 keV peak. Regarding the expected effect of the better closure in the 8 keV, it cannot be evaluated since the introduction of the detector in an environment made of copper compensates and exceeds the improvement. Precisely that copper fluorescence peak is notably higher than in Canfranc and explains the 5 keV peak as a escape peak. It is tempting to attribute the increase of the copper peak to the cosmic muons. On the other hand the copper cathode used for the particular LSC run had a different strong-back design, with lower optical transparency, which had an influence on the collection of external copper fluorescence induced X-rays. 264 Chapter 10. Surface and underground experimental tests of the CAST micromegas background. Several conclusions can be drawn, basically from the comparison of the shape of the spectra. The 6 keV appears only in CAST set-ups, and can be attributed to stainless steel fluorescence. The underground background levels are systematically lower than at surface. The effect of cosmic-rays inducing fluorescences is suggested by comparison of Zaragoza’s and Canfranc’s spectra. The progression of table 10.5 looks coherent, it suggests that the improvement of the tightness (pipe and shielding outlet being removed) led to an improvement of the background about 30%, in agreement with the simulation estimation, and a non negligible cosmic rays contribution about 1–2×10−6keV−1cm−2s−1which would be very dependent on the set-up, since it seems closely related with the production of fluorescences. Unfortunately a robust quantitative estimation of the cosmic contribution cannot be extracted from these data, since the particular set-up, environmental differences and the elapsed time between runs (the M10 suffered one severe intervention between each run) introduce systematic uncertainties of the same order of the effect. Therefore, in order to determine the cosmic rays’ contribution, a direct attempt of shielding upgrade at surface must be carried out, so as to have a signal to background improvement. Moreover, the cosmic rays’ contribution may depend on the shielding thickness. That kind of tests are exposed in the next section. 10.4.2 Direct comparison of medium size shielding set-ups. Cosmics muons active veto. Several set-ups with medium size shielding, about 5 cm lead, were performed at surface at Saclay’s micromegas lab[225] and Zaragoza (like R13[221]) leading to a notable background decrease; the obtained backgrounds oscillate around 2×10−6keV−1cm−2s−1. The surface runs were completed by the 2012 sunset upgrade (R16, see section 8.2.1.3), the first heavy (> 5 cm thickness) shielding tested at surface. These surface runs were complemented with a 5 cm external shielding set-up carried out in the LSC (R15) as part of the progression for the establishment of the ULB of the M17 detector7only with 10 cm of extra lead thickness (R8) (section 10.3.1.2), see figure 10.18 and table 10.6. The direct comparison of both, surface and underground, set of runs done in figure 10.19 shows a clear saturation effect experimented by surface upgrades. On the other hand the underground values show a sharp decrease, that confirms the 10 cm thickness level as very close to be free of the external γflux influence. This underground curve is corroborated by a set of simulations of several shielding thickness[226] which consider the external γflux as the unique background source, so purely reflecting the screening effect of the shielding. 7For the M17 detector measurements the internal shielding (CAST-like) was made entirely in copper, so it is slightly lighter than for the M10 detector measurements. No clear background improvement was observed after the replacement of the internal lead with copper, using 10 cm 10.4. Surface vs underground and the cosmic rays contribution. 265 external shielding thickness run time (days) Bkg ([2-7] keV) Bkg ([2-9] keV) 0 (only CAST-like) 6.0 67 ±8 100 ±9 5 cm 11.6 3.6±1.4 4.8±1.7 10 cm 43.1 2.1±0.6 2.8±0.5 Table 10.6: Progressive upgrade of the M17 shielding in the new LSC Hall A; the progression of runs is R14, R15, R8. The corresponding spectra can be found in figure 10.18. Final background levels expressed in ×10−7keV−1cm−2s−1and statistical errors as 2σ. Figure 10.18: Progressive upgrade of the M17 shielding in the new LSC Hall A; the progression of runs is R14, R15, R8. The corresponding background levels can be found in table 10.6. Some treatment of the simulation curve must be explained. Firstly the simulation points have been corrected by a common factor so as to match the first one, which is basically the one obtained in section 9.2, with the experimental value; thus trying to correct systematic errors due to simulation efficiency and other uncertainties. Finally the Canfranc’s value for the intrinsic background level of the set-up (see 10.3.1), ≃2×10−7keV−1cm−2s−1, has been added to every point. The simulation curve has been fitted to a combination of two exponential decays over a fixed ultimate background level. The two exponential decays constants try to broadly describe the different attenuation rates for the set of natural emission energies. It is remarkable the fact that the longer obtained attenuation constant corresponds to a mean free path of 6 cm, which is needed to explain the similarity, external lead shielding. 266 Chapter 10. Surface and underground experimental tests of the CAST micromegas background. Figure 10.19: Underground vs surface. The improvement along the runs is exclusively due to external shielding upgrades. The red squares correspond to background measurements done in the LSC. The same strategy yields more modest results at surface (blue triangles). The difference should be attributed to cosmic rays. The thickest and lowest two points of the blue runs show the effect of the installation of a cosmic veto over CAST sunset detectors (see section 8.2.1.3). The green dots are simulation results used to draw the background decrease curve due to pure screening of environmental γflux: the green line that is a fit of these points. The plot shows that this behaviour explains the underground runs. but not coincidence, between the 5 and 10 cm results, that is three times the mean free path of 208Tl 2614 keV γ, the most energetic of the natural radioactivity. This fact suggests that the estimation done at the beginning of section 10.3.1.2 was certainly naive and secondary radiation can notably extend the penetration of the radiation through the shielding. The simulations would be needed to estimate it. Therefore, perhaps an upgrade of the 10 cm shielding still may drive to a moderate improvement and should be considered in future. To sum up, both set of values, the simulations and the underground measurements, match and describe the background behaviour in absence of cosmic rays as a pure γshielding effect. The disagreement with the surface series is notable and can only be attributed to the cosmic rays effect. The good news are that the cosmic rays contributions seems to be independent on the shielding thickness, since no background increase is detected, at least up to 10 cm thickness. A more direct proof should come from a set-up which also tests an strategy to avoid the cosmic muons: the installation of a plastic scintillator detector coupled to the acquisition used in Zaragoza’s lab (see figure 10.20). The scintillator signal is recorded and time coincidences with the micromegas triggers can be checked off-line. Using the M10 detector with a 5 cm lead thickness (but 10.4. Surface vs underground and the cosmic rays contribution. 267 not very compact) the background detected in coincidence with the scintillator was (3.0±0.6)×10−6keV−1cm−2s−1, in 2 to 7 keV for a total background of (5.0±0.8)×10−6keV−1cm−2s−1[228]. It must be specified that the micromegas was mounted in horizontal orientation, i.e. with the micromegas’ surface transversal to the upward direction, in contrast to CAST detectors set-ups and the rest of the tests, thus the flux of muons crossing the detector area is higher. The muon-induced background accounted by the veto is of the same order than the surface background levels. Moreover the fact that it is higher than the one registered in vertical orientations, could be interpreted as a hint for the muons to cause directly the background, more likely than secondary radiation. Figure 10.20: CAST-like set-up with plastic scintillator veto installed at Zaragoza’s lab. Once again, the CAST experiment itself was the first set-up proving an effective background reduction thanks to a new advance when a scintillator was added to the sunset set-up for the 2012 data-taking (see figure 8.9). Because of practical limitations the scintillator could be mounted so as to provide a 44% geometrical coverage for cosmic muons8. During the Summer 2012 campaign the veto got a 25% background reduction in both sunset detectors[227]. The averaged background, with and without cosmic veto, is plotted for one of them (sunset1) in figure (10.21). Thedr two background levels are also represented in figure 10.19 as the two thickest surface points. As it was already advanced by the spectrum comparisons of the previous section, the cosmic effect (and so the reduction) is more intense in the copper fluorescence peak. A sunset set-up replica mounted in the Zaragoza’s T-REX lab with a scintillator veto which has been calculated a geometrical coverage of 74% registered background values of (1.8±0.3)/(1.18 ±0.28)×10−6keV−1cm−2s−1with/without veto application, that is to say, a 36 ±8% reduction[229]. However, in this particular set-up the detector (again the M10) had been mounted with the same aluminium cathode 8Calculated from cosmic muons angular distribution at sea level without considering muons interactions (so straight trajectories and no secondary ones) and assuming a 100% efficiency for the veto. 268 Chapter 10. Surface and underground experimental tests of the CAST micromegas background. Figure 10.21: Background spectrum averaged by sunset detector 1 during the 2012 data-taking, with and without (R16) vetoing events in coincidence with the scintillator signal[227]. studied in section 10.3.2.1. The intrinsic background level for such a set-up was there evaluated to be ≃7×10−7keV−1cm−2s−1. Taking into account this floor level, unavoidable by means of the cosmic veto, the reduction experimented for the rest of the unexplained (but attributed to cosmic rays) background was really ≃58%. In other words, the expected background for a geometrically equivalent set-up, but using a more radiopure copper cathode, should be about 7×10−7×10−6keV−1cm−2s−1. In conclusion, the preliminary tests done with cosmic veto (including an official data-taking campaign of the CAST experiment) confirm the cosmic origin of one important part of background at surface for heavy shielded set-ups. The clues point to the particular cosmic muons that directly cross the detector, or generate secondaries (as copper fluorescences) very close to it, to cause most of the background, because its contribution is almost independent of the detector shielding, but depends on the detector orientation. Therefore they should be strongly reduced by a cosmic veto, as the last test, with better geometrical coverage, has demonstrated. 10.5. Runs summary table. 269 10.5 Runs summary table. R Det. Shd. Lab Bkg. [2-7] keV Exp. Comments keV−1cm−2s−1days 0 M13 C-l Zgz ∼9·10−6∼60 Systematic tests, see 10.2. 1 M13 C-l LSC ∼8·10−615 2 M10 C-l LSC (3.8±0.6) ·10−66.0 3 M10 +20 LSC (2.4±0.2) ·10−624.8 Closure, only 5/6·4π 4 M10 +20 LSC (1.5±0.9) ·10−722.3 Window closed 5 mm Cu 5 M10 +20 LSC (1.6±0.6) ·10−719.6 6 M10 +20 LSC (2.0±0.6) ·10−725.9 After cathode test 7 M10 +20 LSC (2.8±1.0) ·10−719.7 After Rn intrusion 8 M17 +10 HA (2.1±0.6) ·10−743.1 9 M10 +20 LSC (7.1±1.2) ·10−735.0 Al cathode 10 M10 +20 LSC ∼1·10−67 Variable Rn ∼200 Bq/m3 11 M10 SR CAST (7.1±0.4) ·10−623.6 2008 campaign 12 M10 C-l Zgz (4.9±1.0) ·10−62.8 T2K+Oscillo DAQ 13 M10 +2 Zgz (2.5±1.0) ·10−64.9 T2K+Oscillo DAQ 14 M17 C-l HA (6.7±0.8) ·10−66.0 Cu shielding 15 M17 +5 HA (3.6±1.4) ·10−711.6 16 M19 SS CAST (1.7±0.1) ·10−692.3 2012 campaign Table 10.7: Summary of runs with CAST microbulks. Runs are ordered as they appeared in the text. Tabled fields are: R (run reference number), Det. (detector name), Shld. (Shielding or set-up configuration), Lab (laboratory where the set-up was placed), background (with statistical errors as 2σ), Exp. (exposure time) and Comments. Keys for the configurations: C-l (CAST-like test set-up), +X (lead extra thickness in cm on C-l), SR (sunrise set-up, see section 8.2.1.2), SS (sunset set-up after 2012 upgrade, see section 8.2.1.3). Keys for places: Zgz (Zaragoza’s chambers test bench), LSC (Canfranc old laboratory), HA (Canfranc new experimental Hall A), CAST (CAST experiment). The ULBP in the old LSC contains Runs 4-7. Runs 4 and 5 are also referred as the 1st ULPB, R6 as the 2nd ULBP (after test R9) and R7 as the 3rd ULPB (after test R10). R8 is also cited as the new ULBP in the new LSC. R14, R15 and R8 conform a shielding upgrade progression for the M17 detector; like R2, R3 and R4 for the M10. 11.2. Prospects for next CAST background improvements. 277 Figure 11.3: Left: mono-detector sunset replica in an intermediate montage stage. Right: ray-tracking simulation of cosmic muons rain over one CAST detector partially covered by a preliminary veto design. Computed coverage for the set-up: 95%. been reduced since the sunset detectors already incorporated radiopurity improvements which have not been tested yet in the LSC. The rest of the background is attributed to cosmic rays. Although the mechanism that produces the final events is not completely known (apart from the fluorescence peaks), the muons passing near the detector are thought to produce the main contribution. New ideas have been conceived in order to answer one more time to the same questions we did for the sunrise. Figure 11.4: Sunset micromegas background accumulated during the summer 2012 datataking campaign by detector 1, before and after applying the cosmic rays veto[227]. 278 Chapter 11. Conclusions and prospects from the study on the CAST micromegas background 11.2.1 About a future shielding upgrade. Is the present (sunset) shielding the optimum one? which parts of it are really useful? can an upgrade produce further improvements? The most obvious step is the implementation of a high efficient cosmic veto. A new events generator method for the Monte Carlo (see ??) has been written to emulate the cosmic muons energy and angular distribution. Up to know it is being used to evaluate the geometrical coverage of preliminary veto designs for the 2013 upgrade (see figure 11.3 right). If no other fundamental limitation arises, the LSC levels will, in principle, be reproducible at surface with a perfect cosmic veto. One limitation that makes a difference between sunset and LSC set-ups is the presence of outlets in the shielding of the CAST set-ups. We have quantitative indications of the weight of the contribution of the shielding’s outlets to the sunrise’s background. From simulations they are about 1/2 of the total γ-induced background and concentrates in the stainless steel fluorescence peak. It is expected that muons contributed in a similar way. All the counts in that peak can be assumed to be produced via the shielding’s main outlet in the front-side. So they represent roughly 1/2 of the total sunrise background, which is actualy higher than the present sunset background. On the other hand, great efforts have been devoted to minimize the effect of the outlets. The front outlet solid angle has been decreased by the introduction of the copper pipe in the new sunset. Preliminary simulation tests suggest the outlet’s influence on environmental isotropic γflux induced background to be very small[226]. In any way, testing that hypothesis is very easy with the sunset replica (see figure 11.3) just by trying to shield the outlet. Moreover the main channel to produce background events via the front outlet, the stainless steel fluorescence, has been strongly suppressed thanks to the replacement of steel with copper which, in addition, has been internally shielded with a teflon coating. In fact, the peak is not found in Summer 2012 spectrum (see figure 11.4). Furthermore, the new pipe is radiopure. It must also be remembered that the sunrise set-up was particularly vulnerable to this effect due to the long stainless steel pipe; the old sunset set-up, before 2012, already produced background spectra with a smaller stainless steel fluorescence peak. The possible use of a X-ray optics focussing device coupled to the micromegas to improve the signal to background ratio, would allow further minimization of the oultet because the chamber’s window could be significantly reduced. Regarding the second shielding outlet for the extraction of the micromegas’ signals, the Monte Carlo simulations evaluated its contribution to be ∼1/10 of the sunrise background, that is about 1/2 of the sunset level. However the solid angle of this outlet has also been strongly reduced due the extension of the lead shielding from 2.5 to 10 cm thickness. The problem has been already considered for the new detectors design[227](expected for 2013 sunrise upgrade) where this outlet is 11.2. Prospects for next CAST background improvements. 279 removed of the shielding because a the new detector’s raquette is made of copper, thus fully integrated in the innermost shielding. Other kind of upgrades, like a new shielding against neutrons, may be suggested by the underground study of the LSC’s level. If this level was demonstrated to be lower, a further increase of the lead thickness could be proposed. 11.2.2 About the intrinsic background level of the detector and the radiopurity. Is there an essential limit, intrinsic to CAST detectors, to the background potential improvement? The LSC’s ULB level is not definitely attributed to the microbulk radioactivity yet. Actually preliminary simulations of the radioactivity of the microbulk readout and other less fundamental detector pieces[230] yield background values <5×10−8keV−1cm−2s−1. The sunrise-like tests set-up, still at LSC, is sequentially replacing the non radiopure pieces of the detector: teflon nuts and washers instead of steel ones, copper gas connectors replacing brass ones and teflon o-rings to close the chamber avoiding viton; the internal lead shielding replaced with a copper one, and the copper cathode cleaned with acid etching. These improvements are already included in the previously mentioned new detector design, to be used in the sunrise side in 2013. Finally the microbulk itself can be replaced by other technologies, still under development, like cuflon (copper and teflon foils)-based manufacturing. Some precautions were adopted in advance in the sunset set-up: thicker copper walls and copper gas connectors. In addition all the copper pieces follow typical ultra-low background cleaning techniques, including acid etching. The ULB level can be determined not by the detector’s material budget, but by the particular set-up’s design. The CAST detector used still has a raquette whose neck which cannot be completely shielded. Despite the upgrade of the external walls thickness, the internal shielding remains the same to shield the radioactivity generated by the electronic elements in the other side of the Faraday cage. Their contribution could have became now the limiting factor. The internal shielding can be upgraded and/or the electronics moved farther by means of flat and flexible cables. That is already done in the new sunset-like tests set-up that uses T2K electronics which are connected with long flat cables. This set-up will be mounted in the LSC. 11.2.3 About the nature of the background, analysis and electronics upgrades. Firstly, it does not seem likely that sparks or any kind of rare noise are systematically limiting the background, since the background levels (always taken with good SNR) look almost independent on detector status and conditions. Another possibility is that the underground environment itself would mean a limitation, for example due 280 Chapter 11. Conclusions and prospects from the study on the CAST micromegas background to neutrons flux. The LSC set-up is planned to be provided with a shielding against neutrons. could the present (sunset) background be improved with a better analysis performance? could small changes in detector, system or environmental parameters produce relevant impacts on the background? While the answer to the same question in the previous section (regarding the sunrise) was that basically it was not possible; now (regarding the sunset), if the background, even after the application of a realistic cosmic veto, is still dominated by muon events, the new answer is: probably yes. The lessons learnt from γsimulations about the fast electrons contributing to the final background could surely be applied to muons as both are MIPs. The focus of the simulations has moved from γto muons. The kind of events they can generate must be systematically studied and the most potentially dangerous cases identified. Several R&D lines and expected upgrades are already intended to improve signal quality and analysis. In section 8.2.2 was generically asserted that the homogenization of signals and the increase the available information had led to the discrimination improvements. That idea regarded, until now, the X-Y strips’s signals equivalence. The new upgrades aim at the readout plane’s signals and the third dimension, i.e. the time signal1. There is still a kind of extra dimension where to generalize the analysis: the event energy. 11.2.3.1 The electronics upgrade: possibilities for a T2K-based DAQ. In the new detectors’ design[227] the readout remains essentially the same, the main upgrade comes from the chamber structure, which is provided with a field shaper and metallic walls. Such structure should improve the drift field homogeneity. Besides, the adoption of the T2K electronics means the upgrade of CAST detectors from a 2D readout with an additional fast time signal to a true TPC (see figure 11.5). This fact brings, at least the following possibilities: •to apply pulse shape analysis to every strip signal. The know-how accumulated from mesh pulse analysis could be transfered to the strips. •to achieve a better SNR. Thanks to the new electronics features, but also due individual pulse analysis that allows to subtract the instantaneous baseline to each pulse, instead of applying an averaged pedestal. According to the simulations a sensitive enough evaluation of the strength of the strips signals is the key parameter to reject MIP tracks, because the fundamental difficulty is, in most of cases, the identification of the weak ionization path of a MIP. •to improve energy resolution and lower the energy threshold. The pulse shape analysis applied to the strips signals has an advantage with respect to the mesh 1Some hints about the convenience of this kind of upgrade were already noted in section 7.1.2 from an apparently different context. 11.2. Prospects for next CAST background improvements. 281 Figure 11.5: Strips signals upgrade. 55Fe 6 keV X-ray signals in a CAST micromegas as seen by the classic Gassiplex card (left) and the T2K electronics (right). one: the electric capacity of the strips is much lower and so the noise. The strips signals should become the main reference for the energy determination. •a further specialisation of the mesh signal. Thanks to their new time capabilities, the strips signals alone are enough to reject most of the events that before needed the mesh signal analysis. In particular, multi-cluster events aligned in the zdirection or vertical tracks can be easily rejected using only the information from the strips. There is no more the necessity to record the whole chamber drift time for the mesh signal, then the mesh signal could be even faster. There is no more sense in integrating the mesh signal, entailing primary information sacrifice, since the valid reference for the energy will be given by the strips. The preamplifier pulse could be directly analysed, which will deliver a better time information. •In section 9.5.3 it was shown that diffusion of primary electrons in the gas is already affecting the shape of the signals. That would be more evident if the signals become faster, in particular if the preamplifier pulse is recorded. The application of more sensible discrimination routines could be limited by the diffusion or, perhaps, even a sort of z-dependent analysis could be developed. Regarding to the diffusion properties, new gas mixtures (e.g. the addition of CF4), could be tested for powering these possibilities. 11.2.3.2 The possibility of an energy-dependence analysis characterization of the detectors. It was already suggested in sections 8.3 and 9.5 how the study of the energy dependence of the observables could improve, more than the discrimination power, the relationship between signal detection efficiency and background. The precise determination of the efficiency versus energy curve is one of the most crucial works still pending for the CAST micromegas detectors and would enhance the reliability of 282 Chapter 11. Conclusions and prospects from the study on the CAST micromegas background CAST results. The CAST micromgas team has started a tests-bench at CERN[203] provided with an electron beam which, by means of PIXE (particle induced X-rays emission) can generate several emission lines depending on the selection of a target material: 1.49 keV (Al), 2.12 keV (Au), 2.29 keV (Mo), 4.41 keV (Ti) and 8.05, 8.9 keV (Cu). Note these lines are complementary with the typical CAST calibration at 5.9, 6.2 and 3 keV (55Fe). There is also the possibility of using the 241Am alphas to produce the PIXE in a simpler set-up. In conclusion, to answer again the same questions, the same strategies, that brought CAST detector here, can be invoked. They are still not exhausted, rather on the contrary they are expected to bring shortly a new step down. In order fulfil the expectations of the new proposed axion helioscope IAXO background levels of the order or 10−7×10−7keV−1cm−2s−1or even less are demanded, depending on the different (realistic or optimistic) considered scenarios[193]. We will try a very rough estimation. The present LSC level, 2×10−7keV−1cm−2s−1could be reduced to 1×10−7keV−1cm−2s−1, with the adoption of the radiopurity precautions already installed in the SSMM. Assuming that the rest of the background is due to cosmic muons and interpolating the performance of the cosmic muon veto from the tests already done, it could be estimated that the background levels in CAST will be around 2–6×10−7keV−1cm−2s−1already for the 2013 campaign. Chapter 12 Summary and conclusions. The Rare Event Searches is a heterogeneous field from the point of view of their physical motivations; ββ0νdecay experiments, direct detection of WIMPs as well as axions and other WISPs (candidates for the DM, but also motivated by other questions from Particle Physics) were considered in chapter 1. The field is rather defined by the requirements of these experiments, essentially a very sensitive detector with low background which is usually operated in underground laboratories. The availability of a rich description of the event registered by the detector is a powerful tool for the discrimination of the signal from the background. That is a goal that makes the rare event searches experiments evolve to become more sophisticated, in order to produce more information about the event by means of applying pulse shape analysis, segmenting their sensitive volumes, the specialization of different active zones, combining different signals, etc. The topological description of the interaction that can be delivered by a gaseous TPC, that includes the ionization path and dE/ dxalong it, is a useful source of information about the event. That is particularly clear for the ββ0νdecay, where the complete reconstruction of the ionization path would allow to identify the decay events that are composed of two electrons with a common vertex. Regarding the DM searches, there can be several approaches. Beyond some energy threshold, the nuclear recoils induced by the collisions with WIMPs or neutrons can be discriminated from γ/β events since the electron’s tracks are longer and less straight than nuclear recoils. If the nuclear recoils were be fully reconstructed, deducing their initial direction (and, perhaps, sense), the directional signature of DM could be invoked to claim a positive signal. Even with more modest topological capabilities, a TPC can discriminate nuclear recoils and X-rays from higher energy β, muons, α, besides with good noise rejection and avoiding border effects down to low energy thresholds, surely below the keV. This latter kind of operation is suitable also for X-rays detectors of helioscopes, where WISP into photon conversion is produced. The generic requirements for a gaseous TPC that is intended for rare event searches are the following. Very good imaging, that requires both good drift properties of the chamber and a good performance of the readout, probably with high granularity. The readout must yield also high gain so as to allow a low energy threshold (that is a main goal for DM) and a correct reconstruction of the ionization paths of MIPs (which can be essential for imaging of electrons or for discrimination of β and muons). In order to increase the detector’s efficiency, many times operation at high pressure is required. This can be needed to increase the detector’s quantum 284 Chapter 12. Summary and conclusions. efficiency for γand X-rays or in order to contain high energy tracks or to have an effective self-shielding to detect secondary interactions or avoid border effects. Radiopurity must be guaranteed for the detector, starting from the readout itself, to the whole chamber structure, including the gas. The source/target-equal-to-detector approach, very common in rare event searches, can impose the operation with particular gases. If a light readout is also required (to combine signals or in order to reconstruct the absolute position of the event) that may imply operation without quencher. In this case, the requirements are also strong for the supporting gas system and further constrictions apply to the material selection for the chamber. In addition these properties must be scalable, large TPCs will be needed to increase the source/target mass. That applies also to the readout and their electronics, which will have to be more compact (integration in cards), fast (usage of ASICs, optical links) and multiplied (low consuming, multiplexing). Last, but not least, stability and reliability during long time data-taking are needed. In the particular case of ββ0νdecay experiments, very good energy resolution is also demanded. Such requirements could be fulfilled by TPCs because they are equipped with MPGDs, which were introduced in chapter 2. This work focussed in micromegas that stand out on stable high gain, fast signals, spatial resolution and robustness. Chapters 3, 4 and 5 were intended to be an ordered exposition on micromegas phenomenology. Since that is still a recent technology (in particular the bulk and microbulk types of micromegas) many of the presented experiments and characterization efforts are original. The text had always present the particularities of rare event searches. The questions addressed above were proven, addresed or, at least, the state of the R&D was reported. The paths of low energy electrons were characterized by means of GEANT4 Monte Carlo simulations, and the results were compared with customary estimations. They are useful to estimate the strength of the signals in the micromegas, depending on the chamber and readout design parameters (gas type and pressure, readout pitch). They are also interesting to deduce if the length of the path can have an influence in the signals, so as they would be seen as different than point-like events for discrimination purposes. The much shorter paths of ions, of primary interest for directional DM searches, were studied as well using SRIM simulations. After giving a general view, the study focused in argon nuclei as WIMP target. The role of secondary recoils produced by elastic collisions seemed to be important, more complex simulations being required. It was shown how a collision followed by a secondary recoil can suddenly change the direction of the primary recoil. Besides they influence, and can even invert, the expected Bragg curve, making more difficult the event reconstruction. Moreover, according to our conclusions, the variable number of collisions would work as an additional degradation of the energy resolution for ion recoils, with relevant effects for low energy signals. It was presented a fast review on recent efforts on this issue from theory and simulations to experiments inside CYGNUS groups. Its conclusions, 285 together with ours (also presented in CYGNUS), seem to be rather pessimistic, since the head-tail effect (invoked to deduce the sense of the recoil) is not clear for ions heavier than helium below 100 keV, while, at the same time, identifying the former piece of the path seems important for a precise reconstruction of the initial direction. The signals in microbulk micromegas from X-rays and α-particle events have been systematically compared to conclude that there is no significant difference or quench-like effect; the ratio of the intensity of the αsignal was always at least 96% of the X-ray signal taken in the same conditions. Firstly, the W-value should be very similar; then recombination does not affect to α, if the drift field is intense enough (over 150 V/cm in Ar/iC4H10 98/2), and the transparency of the mesh behaves equally. The drift properties of typical gas mixtures have been discussed from plots generated with Magboltz for reference. The transparency of the mesh, an specific issue to micromegas, has been described in detail. The transparency curve has been defined, parametrized and discussed for different mesh types, and related with their geometrical layout. In particular the diffusion properties of the gas have been proven to influence the mesh transparency. A step by step procedure has been explained to generate the result of avalanches in micromegas using customary models, starting from either the gain curve of the micromegas or the Townsend coefficient dependence on the electric field of the gas. The gain curves have been studied for bulk and microbulk micromegas using different gas mixtures and conditions. The effect of several quencher gases where explored, in particular for bulk micromegas, with special attention to the highest part of the gain curve. It was shown that different quencher gases bring different properties that can be combined or balanced in ternary mixtures. Isobutane brings stability that allows very high gain at relatively low amplification fields; CO2inhibits the overexponential behaviour of the gain curve at high gain, what produces a better energy resolution that does not degrade at high gains; CF4contributes to the enhancement of the drift properties which also reflects in the mesh transparency. Many of these tests were carried out for the first time with this technology. Three series of microbulk prototypes were characterized, contributing to the progressive refinement of the technology. An energy resolution of 11.0% FWHM at 6 keV (world record for a MPGD) with Ar/iC4HC10 95/5 at atmospheric pressure was registered, with only small deviations for most of the rest of the microbulks. Furthermore a 100 cm2microbulk with a pixelized readout was tested and instrumented, reconstructing three dimensional αtracks and yielding a performance, in gain an energy resolution, that was not far from that of the small prototypes. The microbulk prototypes were also tested at high pressure. The maximum achievable gain decreases with pressure, roughly exponentially, however gains comfortably above 103were registered with Ar/iC4H10 99/1 at 10 bar. The energy resolution degrades with pressure progressively, but slowly. Operation with pure noble gases, argon and xenon, was also tested. Gain over 102was obtained for 286 Chapter 12. Summary and conclusions. pure xenon at 4 bar, which is not allowed by other MPGDs. The energy resolution measurements were encouraging towards ββ0νdecay prospects, 2.5% FWHM in pure xenon at 4 bar for 5.5 MeV α, despite systematic limitations. The importance of the supporting gas system was put in value, confirming significant improvements with the selection of the construction materials, accumulating benefits from cleaning procedures as pumping and bake-out, the usage of filters and recirculation. Germanium spectroscopy measurements on two microbulk samples and raw materials have been reported. They showed micromegas to be a radiopure readout in terms of activity per unit area, even when compared with PMT readouts specifically designed for rare event searches. Hints were found also that the manufacturing process of microbulk could be further optimized from the radioactive point of view. A library and some programs have been developed to adapt the T2K electronics for the T-REX researchers as a general purpose electronics for micromegas detectors with imaging capabilities. The data are dealt according to the T-REX customary philosophy and way of working and the library has been provided with simple analysis tools. A big part of the work has been devoted to the development of libraries and programs for generic Monte Carlo simulations on low energy TPCs. The package was called RESTsoft (Software for Rare Event Searches with TPCs) and is presented in chapter 6, also in a guide-like style. It includes a versatile GEANT4 application optimized for the simulation of gases in the low energy region. The nature of these kind of interactions, as they are simulated by GEANT4, is discussed with certain detail through several simulation tests. Two examples of works that were carried out with the RESTsoft GEANT4 application were discussed. The calculation of the CAST micromegas detectors quantum efficiency for X-rays and the reproduction of aγspectrum in an high pressure xenon TPC, that allowed its proper interpretation. Further libraries were written to simulate the specific processes of TPCs and micromegas (primary charge generation, drift processes, implementation of the readout, generation of the electronic signals, i.e the TPC simulation chain) and associated tools for information management and interpretation of the results. The T2Kelectronics library is linked to RESTsoft by means of a bridge class due a common conception of the ideal TPC physical event. The classes the RESTsoft library contains can be combined to write complex programs to develop an specific simulation followed by analysis at different levels. Two examples of these complete programs were shown in chapter 7. In the first example the case for a non directional dark matter TPC is studied. Several observables related with the event size and shape were explored in order to estimate the possibility of discriminating nuclear recoils from γ/β events. It was shown that a full three dimensional reconstruction of the events, yielded by a pixelized readout, is a considerable advantage. The simulation of a feasible chamber prototype with 10 cm height and a gas with diffusion coefficient σ= 100 µm cm−1/2produced a γ/β rejection power about 1/105for energies above 20 keVee (electron equivalent) using 293 ables, es decir, han de mantenerse cuando TPCs más grandes se construyan para incrementar la masa de blanco/fuente. Una ventaja de las TPCs es que las cámaras pueden crecer de forma bastante natural, el aumento de la complejidad es absorbida principalmente por el plano de lectura y su electrónica. Ésta última deberá tender a ser compacta (integrada en tarjetas), permitir un rápido procesado y transmisión de una cantidad creciente de datos (uso de ASICs y fibra óptica) y multiplicarse fácilmente (bajo consumo, multiplexado de señales). Por último, pero sin dejar de ser un aspecto esencial, el detector debe acreditar estabilidad y fiabilidad durante largos periodos de toma de datos. En el caso partícular de una TPC para desintegración ββ0ν, una muy buena resolución en energía es primordial. Existe la posibilidad de cumplir con todos estos requisitos gracias a que las TPCs modernas están equipadas con MPGDs (estructuras del orden de micras para la amplificación de electrones en gases que se utilizan como planos de lectura), una breve introducción de los cuales se hace en el capítulo 2. De todos ellos, este trabajo se ha centrado en micromegas que sobresalen por su rendimiento, en particular por el mantenimiento de altas ganancias con estabilidad, la generación de señales rápidas, la integración del ánodo que posibilita una muy buena resolución espacial y la robustez de su estructura. A lo largo de los capítulos 3, 4 y 5 se ha pretendido hacer una exposición ordenada de la fenomenología de los detectores micromegas, en la que se han tenido siempre presentes las particularidades del contexto de la física de sucesos poco probables, con la idea de crear una referencia práctica. Dada la juventud de esta tecnología (especialmente las versiones bulk y microbulk) muchas de las pruebas y observaciones incluídas para ilustración de sus propiedades son originales. Se han tratado de probar las cuestiones que se planteaban en el párrafo anterior o, al menos, de responder comentando el estado de la investigación y desarrollo sobre los puntos abiertos. Las trayectorias de los electrones de baja energía se caracterizaron por medio de simulaciones de Monte Carlo y los resultados se compararon con cálculos frecuentemente propuestos en la literatura. Estos datos resultan útiles para estimar la cantidad de carga que puede esperarse en la micromegas en función del diseño de la cámara, las condiciones del gas y la segmentación del plano del ánodo. También puede resultar interesantes para deducir el punto en el que la longitud de la trayectoria comenzará a influir en la forma de las señales, de manera que deje de poder considerarse como un evento puntual, lo que debe considerarse por los algoritmos de discriminación. Las trayectorias, mucho más cortas, de los iones, de un interés inmediato para búsquedas direccionales de DM, se estudiaron a su vez utilizando SRIM para las simulaciones. Después de dar algunas conclusiones generales, el estudio se centró en los retrocesos nucleares de argon, usado como blanco para los WIMPs. El papel jugado por las colisiones con otros núcleos que dan lugar a retrocesos secundarios pareció ser relevante, por lo que se necesitarían simulaciones algo más complejas que los contemplaran. Se vio que, además de que una de estas colisiones puede modificar 294 Chapter 13. Resumen y conclusiones. bruscamente la dirección del retroceso principal, la presencia de los retrocesos secundarios puede cambiar profundamente, hasta llegar a invertir, la curva de Bragg, lo que dificulta notablemente la tarea de reconstrucción. Además, de acuerdo con nuestras simulaciones, el número variable de colisiones secundarias sufridas supondría un mecanismo adicional de degradación de la resolución de la energía para retrocesos secundarios, que tendría efectos notables para señales de baja energía. Se han recogido algunos de los diversos esfuerzos recientes en esta materia, todos realizados en la órbita de CYGNUS, y que han abarcado formulaciones teóricas, bastantes simulaciones y algún resultado experimental. Las conclusiones que nos parece poder extraer del conjunto, incluyendo las nuestras (también presentadas en CYGNUS) se nos antojan algo pesimistas porque no parece claro que el efecto cabeza-cola (la decisión del sentido de la trayectoria basada en la curva de Bragg) pueda utilizarse eficazmente para iones más pesados que el helio con energías menores de 100 keV, mientras que el hecho de encontrar el extremo inicial de la trayectoria parece un punto de partida necesario para poder reconstruir la direción inicial con precisión. Se compararon sistemáticamente las señales producidas por rayos X y partículas αen microbulk micromegas para no encontrar ninguna diferencia práctica entre ellas, nada parecido a un efecto de quench; la intensidad de las señales αfue, en todo caso, al menos un 96% de la generada por rayos X. Esto significa, para empezar, que los valores W son muy similares, la recombinación no afectaba a las trazas mucho más densas de las αen cuanto el campo de deriva superaba los 150 V/cm (en Ar/iC4H10 98/2) y la transparencia de la mesh (el electrodo-rejilla que da paso a la región de amplificación de la micromegas) mostraba una evolución idéntica con el campo de deriva. Se han analizado las propiedas de deriva de varias mezclas arquetipo a través de los gráficos generados por el programa Magboltz, que se han mostrado con detalle para servir de referencia. La curva de transparencia de la mesh, por ser un aspecto específico de micromegas, se ha definido, parametrizado y discutido con especial ahínco, relacionándose con el diseño geométrico de la mesh. En particular se han proporcionado pruebas de que las propiedades de difusión del gas tienen una influencia no pequeña en la transparencia. Se ha explicado un procedimiento paso a paso para generar la producción final de carga de avalanchas en micromegas, apoyándose en los modelos habituales y a partir de la curva de ganancia de la micromegas o de la dependencia del coeficiente de Townsend del gas con el campo eléctrico. Las curvas de ganancia de micromegas, bulk y microbulk, se han estudiado probando varias mezclas gaseosas en diferentes condiciones. Con bulk micromegas se exploró la utilización de varios quenchers, poniendo una atención especial a la parte final, más alta, de la curva de ganancia. Se comprobó que los distintos quenchers proporcionan diferentes propiedades a la mezcla que pueden combinarse o compensarse hasta cierto punto utilizando mezclas ternarias. Se ha visto que el isobutano dota al gas de estabilidad, lo que permite alcanzar ganancias más altas, además con campos de amplificación relativamente ba- 295 jos. El CO2inhibe el comportamiento sobre-exponencial que las curvas de ganancia tienden a mostrar cuando la ganancia es muy alta, lo que, se ha comprobado, va unido a un deterioro de la resolución en energía; de modo que las mezclas con CO2 exhiben una mejor resolución en energías que no se deteriora apenas con la ganancia. El CF4se añade para mejorar las propiedades de deriva, lo que se refleja también en una mejora de la transparencia de la mesh. Muchas de estas observaciones se hacían por primera vez para esta tecnología. Tres conjuntos de prototipos microbulk, que fueron fabricados sucesivamente, se caracterizaron contribuyendo así al progreso de está técnica. Un espectro con una resolución de un 11.0% FWHM (anchura a media altura) se registró para rayos X de 6 keV en Ar/iC4H10 95/5 a presión atmosférica, lo que supone un record mundial para MPGDs. La mayoría del resto de las microbulk caracterizadas fueron peores sólo por pequeñas desviaciones respecto de este valor. Además se testeó un prototipo pixelizado de 100 cm2de área activa, en el cual se pudieron reconstruir trazas tridimensionales de partículas αy que mostró un rendimiento, en cuanto a ganancia y resolución, no muy alejado del de las versiones más pequeñas. Los prototipos de microbulk se probaron con gas a alta presión. Se ha visto que la ganancia máxima alcanzable decrece, aproximadamente de forma exponencial, con la presión; sin embargo ganancias sobradamente por encima de 103pudieron lograrse utilizando Ar/iC4H10 99/1 en todos los casos, al menos hasta 10 bar. La resolución en energía también se deteriora con el aumento de la presión, pero lo hace despacio y con tendencia a estabilizarse. Se probó también la operación con los gases nobles puros, argon y xenon. Con xenon puro, los prototipos microbulk registraron una ganancia superior a 102a 4 bar para una αde 5.5 MeV, algo que parece no estar al alcance de otros MPGDs. Los resultados en resolución fueron también alentadores para la perspectiva del ββ0ν, obteniéndose un 2.5% FWHM para las mismas condiciones, pese a constatar algunas limitaciones sistemáticas. Se ha puesto en valor la importancia del sistema de gas auxiliar, y se han confirmado mejoras significativas gracias a la selección de materiales, una mejora acumulativa por el uso sistemático y combinado del bombeo y el baqueado (calefactado del sistema para fomentar la expulsión de gases), el uso de filtros y la recirculación del gas. Se han descrito las medidas de espectrometría con germanio realizadas con dos muestras de microbulk micromegas, así como algunos de sus materiales primarios. Los resultados mostraron que se trata de un plano de lectura radiopuro en términos de activad por unidad de área, incluso cuando se compara con PMTs específicamente diseñados para experimentos de física de sucesos poco probables. También se hallaron indicios de que el proceso de manufactura de las microbulk podría optimizarse desde este nuevo punto de vista mejorando aún más la radiopureza. Una librería y algunos programas fueron escritos para adaptar la electrónica del experimento T2K a los intereses de T-REX, donde se adopta como electrónica de propósito general para TPCs con capacidades de reconstrucción topológica. De esta 296 Chapter 13. Resumen y conclusiones. manera los datos se tratan de acuerdo con la filosofía de T-REX y se integran en sus procedimientos habituales. Se ha dotado a la librería de funcionalidades básicas, herramientas sencillas de análisis y visualización que han sido frecuentemente utilizadas en el trabajo de laboratorio de T-REX. Gran parte del trabajo se ha invertido en el desarrollo de librerias genéricas y programas para simulación Monte Carlo de eventos de baja energía en TPCs. El paquete, llamado RESTsoft (software para física de sucesos poco probables con TPCs), se presentó en el capítulo 6, también tratando de conseguir un estilo de “guía de usuario”. Incluye una versátil aplicación de GEANT4 optimizada para la simulación de gases en la región de baja energía. La naturaleza de este tipo de interaccioes, tal GEANT4 las simula, en particular la ionización por parte de electrones, se discutió con cierto detalle a través de varios ejemplos utilizados como tests. Dos ejemplos de trabajos realizados con la aplicación GEANT4 de RESTsoft fueron presentados. El cálculo de la eficiencia cuántica de un detector micromegas de CAST y la reproducción, para su correcta interpretación, de un espectro producido con una fuente γen una cámara de xenon a alta presión. Se han desarrollado librerías para continuar la simulación con los procesos específicos de TPCs y micromegas (generación de carga primaria, procesos de deriva, implementación del plano de lectura, generación de pulsos electrónicos; todo en serie se denominó la cadena de simulación de una TPC) junto a herramientas para la administración de datos e interpretación de los resultados. La librería T2Kelectronics está conectada naturalmente con RESTsoft, auxiliada por una clase puente, gracias al concepto común sobre el típico evento físico registrado en una TPC. Las clases del paquete RESTsoft pueden combinarse para escribir programas complejos que desarrollen una simulación específica y analísis a distintos niveles de la cadena de transformaciones. Dos ejemplos de este tipo de programas de simulación más complejos se presentaron en el capítulo 7. El primero se usó para abordar el caso de una TPC no direccional para búsqueda de DM. Se exploraron algunas definiciones para el tamaño del evento que utilizar como observables para la discriminación de retrocesos nucleares frente a eventos γ/β. Se puso de manifiesto que una reconstrucción completa en tres dimensiones, producida por un plano de lectura pixelizado, conlleva ventajas considerables. Se simuló un prototipo completo de una cámara bastante viable de 10 cm de altura y llena de un gas con coeficiente de difusión σ= 100 µm cm−1/2, este detector fue capaz de desarrollar un poder de rechazo de eventos γ/β de alrededor de 1/105para energías por encima de los 20 keVee (electrón equivalente), utilizando criterios de discriminación aplicados secuencialmente, los cuales aceptaban sobre el 90% de los retrocesos nucleares simulados en la misma cámara. Se estudió una posible figura de mérito para el diseño de la cámara (la elección del gas y su altura) que maximizase la masa activa, siendo ésta la perteneciente al núcleo blanco que está encerrada en el volumen en el cual los eventos γ/β son eficazmente rechazados. Aplicándolo a mezclas de Ar/CF4se concluyó que el óptimo se encontraría entorno 297 a la proporción 70/30 con una cámara de aproximadamente 1 m de altura. Una mezcla muy similar a la propuesta, Ar/CF475/25, se testeó con una micromegas tipo bulk, observándose un buen rendimiento en ganancia, transparencia y resolución en energía. En un segundo ejemplo se hizó una demostración de las capacidades de discriminación de una TPC de alta presión para desintegración ββ0ν. En un estudio preliminar se mostró la naturaleza de los eventos más peligrosos para el fondo del experimento. El papel, de creciente importancia, que los detectores micromegas han jugado en la historia del experimento CAST fue analizado en el capítulo 8. Se describió a los detectores con detalle, espacialmente las últimas versiones basadas en la tecnología microbulk. Se discutió cuáles han sido las estrategias que han permitido mejorar el fondo de las micromegas más de un factor 50 desde la instalación del primer detector hasta 2011 (nivel típico de fondo 5–7×10−6keV−1cm−2s−1[2–7] keV): la capacidad discriminativa, que se fundamenta en un rendimiento extraordinario de las micromegas (sobretodo después de la introducción de las microbulk), un sistema de adquisición que combina el pulso generado en la mesh y las señales de las strips en que está segmentado el ánodo, y un sofisticado análisis; la selección de materiales radiopuros para la construcción del detector y el uso de blindajes pasivos contra la radiación ambiental. Se ha intentado evaluar cual ha sido la contribución específica de cada estrategia, y han sido comentadas con afán de crítica constructiva, especialmente el diseño del blindaje y algunos aspectos de las rutinas de discriminación. Precisamente el blindaje fue mejorado en 2012 consiguiendo un descenso del nivel de fondo en un factor 5–4 (1.3–1.7×10−6keV−1cm−2s−1[2–7] keV). Con el objetivo de concretar las líneas maestras para revitalizar la alentadora progresión de los fondos obtenidos por las micromegas en CAST y proyectar predicciones razonables para IAXO (el helioscopio que lo sucederá), debía emprenderse un estudio menos contemplativo, que atendiese a la importancia relativa de las diferentes fuentes que contribuyen al fondo así como a la naturaleza de los eventos que finalmente lo componen. Esto se llevó a cabo en los trabajos comprendidos en los capítulos 9 (simulaciones) y 10 (tests experimentales). Se utilizó RESTsoft para componer una completa cadena de simulaciones desde las interacciones de GEANT4 hasta la producción de datos en el mismo formato en el que lo hace la adquisición de CAST. La geometría del montaje sunrise se tomó como modelo porque esta disposición obtenía el fondo más bajo, hasta la remodelación del sunset ya en verano de 2012. Una serie de medidas de γse dedicó a caracterizar la actividad en el área experimental de CAST. Se justificó el uso de un flujo ambiental γcomo punto de partida para las simulaciones, deducido a partir de estas medidas. La contributión obtenida para el flujo ambiental γes del mismo orden que el fondo experimental total (∼1/3 fue obtenido directamente, ∼1/3–2/3 se estima apoyándonos en más observaciones). Además, el estudio detallado de la simulación mostró como este tipo de radiación produce la típica forma del espectro experimental global, con una importante contribución proveniente de la fluorescencia del acero inoxidable. 298 Chapter 13. Resumen y conclusiones. El modelo planteado para las micromegas se testeó in situ utilizando una fuente de 57Co que emité γcon energías entorno a los 100 keV, obteniéndose un excelente acuerdo simulación-estimación experimental. El modelo queda así establecido y disponible para la evaluacion de la contribución de distintas fuentes potenciales de fondo. Otro uso que podría darse al modelo sería el de herramienta para explorar la interrelación entre los diferentes tipos de eventos del fondo y el análisis. Algunas tendencias fueron ya indicadas en un primer tanteo para demostrar esta posibilidad. En particular algunas observaciones han puesto de manifiesto la importancia de una buena relación señal-ruido en las señales de las strips para rechazar MIPs, lo que se entendió como una advertencia de que los muones cósmicos podían ser una fuente relevante de fondo, ligada al rendimiento del detector. Los tests realizados en el Laboratorio Subterráneo de Canfranc con detectores CAST con blindajes pesados (de al menos 10 cm de espesor de plomo) y un cátodo de cobre, dieron lugar a un fondo ∼30 veces más bajo que el del montaje sunrise de CAST (record hasta entoces), ≃2×10−7keV−1cm−2s−1. Este nivel se ha mostrado estable y reproducible. También se evaluó la contribución del radon cerca del detector como (3.0±0.8) ×10−9keV−1cm−2s−1/(Bq m−3), lo cual produciría efectos despreciables en superficie. El nivel de ultrabajo fondo no fue reproducible al instalar blindajes pesados en superficie, lo que fue atribuído a la acción de los rayos cósmicos. Este hecho se confirmó cuando se pudo probar que la instalación de un veto cósmico producía una eficiente reducción del fondo. El éxito de la remodelación del montaje del sunset en 2012 ha supuesto la primera aplicación/confirmación de las conclusiones obtenidas de estos estudios, así como la realización de un test de indubitable valor añadido. El blindaje fue ampliado hasta un diseño más hermético de 10 cm de grosor de plomo, evitando el uso de acero inoxidable que fue reemplazado por cobre (en especial el cátodo del detector). Los antiguos blindajes exteriores de polietileno y cadmio fueron, al contrario, desestimados. Con esta configuración los detectores promediaron un fondo de 1.3– 1.7×10−6keV−1cm−2s−1(5–4 veces menor que el del sunrise) durante la campaña de verano de 2012, incluyendo un 25% de reducción extra debida a un veto cósmico cuya instalación provisional procuraba tan sólo un 44% de covertura geométrica de los muones cósmicos. Las conclusiones y predicciones que pueden deducirse de la combinación de los hechos e indicios acumulados por ambos estudios (simulaciones y tests experimentales) se expusieron en el capítulo 11. Concluyendo, las perspectivas para la aplicación de micromegas son prometedoras para los tres objetivos de física de sucesos poco probables que nos propusimos. Para empezar, la tecnología micromegas más brillante en su rendimiento, se ha confirmadio radiopura. Se han dado varios pasos hacia su escalado para conseguir planos de lectura de áreas mayores, con buenos resultados. Estos planos de lectura, además de mayores también más complejos, se leen eficazmente con la electrónica T2K, más compacta y que proporciona más información. Actualmente está siendo utilizada extensivamente en las actividades de T-REX: la TPC de alta presión NEXT-1-MM, la 299 nueva adquisición de CAST, un detector de alta sensibilidad de radón con capacidad de reconstruir las trazas α, la futura cámara T-REX-DM. También han dado resultados alentadores los tests realizados con micromegas sobre algunos aspectos de su operación requeridos por algunos escenarios de la física de sucesos poco probables: alta presión, uso de algunas mezclas en particular (incluyendo gases nobles puros), detección de luz. La investigación y el desarrollo continuán en estas direcciones, siendo el uso de mezclas Penning una de las soluciones más prometedoras. El paquete de librerías RESTsoft se muestra ya como una herramienta útil usada en T-REX para la comprensión y/o la concepción de pequeños experimentos o proyectos más ambiciosos. La simulación de los detectores CAST, en particular, ha tenido que alcanzar un alto nivel de detalle en el cual se han realizado un conjunto de comparaciones con el experimento que son valiosas para validar y mejorar el código. Los resultados producidos por las simulaciones de CAST, junto con las simulaciones sobre el poder de discriminación de una TPC para la desintegración ββ0νy la detección directa de DM son muy alentadores. La impresionante progresión del fondo de los detectores micromegas en CAST es, quizá, el resultado más significativo de las posibilidades de la aplicación de esta tecnología en física de sucesos poco probables. Se han hecho avances importantes en la comprensión de la naturaleza de su fondo y, por tanto, en la determinación potencial de las distintas estragegias aplicadas para reducirlo y el modo en que el rendimiento del detector y las rutinas de discriminación se interrelacionan con los distintipos tipos de eventos del fondo. Puede asegurarse que la evolución del fondo, que ha mejorado en más de dos órdenes de magnitud desde que la primera micromegas fue instalada en CAST, no se detendrá en el presente nivel, y que el futuro helioscopio IAXO contará con detectores micromegas más sensibles. El nivel de fondo tan bajo obtenido en el Laboratorio Subterráneo de Canfranc (el cual representa probablemente tan sólo una cota superior, y no una limitación realmente impuesta por la micromegas) es uno de los hechos que apoyan la última afirmación. Sin embargo, su relevancia va más allá de la aplicación de micromegas en helioscopios. Se ha demostrado la posibilidad de registrar fondos muy bajos bajo 10 keV con un umbral en energías también muy bajo (conseguirlo bajo el keV se antoja bastante inmediato), esto ha motivado el enfoque actual del proyecto T-REX-DM, cuyo prototipo sigue una línea tipo CAST. Bibliography [1] I.G. Irastorza et al., Status of R&D on Micromegas for Rare Event Searches: The T-REX project, Proceedings of the Third International Conference on Directional Detection of Dark Matter (CYGNUS 2011), EAS Pub. Ser. 53 (2012) 147-154. 1,143 [2] H.V. Klapdor-Kleingrothaus, Sixty Years of Double Beta Decay: From Nuclear Physics to Beyond Standard Model Particle Physics, World Scientific (2001). 3 [3] F.T. Avignone III, G.S. King III and Yu G. Zdesenko, Next generation double-beta decay experiments: metrics for their evaluation, New J. Phys. 7 (2005) 6. 3 [4] A. Gouvea, On Determining the Neutrino Mass, FNAL Theory Seminar, March 30, 2006. 3 [5] N. Fatemi-Ghomi, Mearurement of the Double Beta Decay Half-life of 150Nd and Search for neutrinoless Decay Modes with the NEMO3 Detector, PhD Thesis, Univerity of Manchester, 2009. 3 [6] P. Vogel, Neutrinoless double beta decay, (2008) arXiv:hep-ph/0611243.4 [7] C.E. Aalseth et al., Neutrinoless double beta decay of 76Ge: first results from the International Germanium Experiment (IGEX) with six isotopically enriched detectors, Phys. Rev. C 59 (1999) 2108. 4 [8] M. Gunther et al., Heidelberg-Moscow ββ experiment with 76Ge: full setup with five detectors, Phys. Rev. D 55 (1997) 54. 4 [9] H. Klapdor-Kleingrothaus et al., First Evidence for Lepton Number Violation and of the Majorana Character of Neutrinos, Modern Physics Letters A 16 (2002) 092007. 4 [10] The Majorana Collaboration, White Paper on the Majorana Zero-Neutrino Double-Beta Decay Experiment, (2003) Preprint arXiv:nucl-ex/0311013. 4 [11] I. Abt et al., GERDA Letter of Intent, A new 76Ge Double Beta Decay Experiment a LNGS, (2004) Preprint arXiv:hep-ex/0404039. 4 [12] C. Arnaboldi et al., CUORE: a cryogenic underground observatory for rare events, Nucl. Instrum. Meth. A 518 (2004) 775. 5 [13] EXO collaboration, Search for Neutrinoless Double-Beta Decay in 136Xe with EXO-200, Phys. Rev. Lett. 109 (2012) 032505. 5 302 Bibliography [14] EXO collaboration, Observation of Two-Neutrino Double-Beta Decay in Xe136 with EXO-200, Phys. Rev. Lett. 107 (2011) 212501. 5 [15] KamLAND-Zen Collaboration, Measurement of the double-beta decay halflife of 136 Xe with the KamLAND-Zen experiment, Phys. Rev. C 85, 045504 (2012). 5 [16] A.S. Barabash et al., NEMO3 and SuperNEMO double beta decay experiments, J. Phys. Conf. Ser. 39 (2006) 347. 5 [17] R. Luescher et al., Search for ββ decay in 136Xe: new results form the Gotthard experiment Phys. Lett. B 434 (1998) 407. 5 [18] F.J. Iguaz, Development of a Time Projection Chamber prototype with Micromegas technology for the search of the Double Beta Decay of 136Xe, PhD Thesis, Universidad de Zaragoza, (2010) http://zaguan.unizar.es/record/5731.6,65,95,97,99,154 [19] L. Seguí, Pattern Recognition in a High Pressure Time Projection Chamber prototype with a Micromegas readout for the 136Xe double beta decay, PhD Thesis, Universidad de Zaragoza, in preparation (2013). 6,97,99,109,154 [20] NEXT collaboration, NEXT, a HPGXe TPC for neutrinoless double beta decay searches, Letter of Intent to the LSC Scientific Committee, Preprint arXiv:[hep-ex] 0907.4054 (2009). 6 [21] K. Pushkin, R&D for the EXO-GAS experiment to search for neutrinoless double beta decay, poster presented i the 12th Pisa Meeting on Advanced Detectors, May 2012. 6 [22] G. Bertone, D. Hooper and J. Silk, Particle dark matter: evidence, candidates and constrains, Physics Reports 405 (2005) 279-390. 6,8 [23] E. Komatsu et al., Seven-Year Wilkinson Microwave Anisotropy Probe (WMAP) Observations: Cosmological Interpretation, Astrophys. J. Suppl. 192:18 (2011). 6 [24] S. Perlmutter (Supernova Cosmology Project), Supernovae, Dark Energy, and the Accelerating Universe, Physics Today, April 2003. 6,7 [25] S.W. Allen, A.E. Evrard, A.B. Mantz, Cosmological Parameters from Observations of Galaxy Clusters, ARA&A (2011) 49:409-470. arXiv:1103.4829. 6 [26] http://www.rssd.esa.int/index.php?project=planck 6 [27] Plank collaboration, Planck 2013 results. I. Overview of products and scientific results, arXiv:1303.5062.6 Bibliography 309 [122] E.J. Kobetich and R. Katz, Phys. Rev 170 (1968) 391. 35 [123] http://physics.nist.gov/PhysRefData/Star/Text/method.html 36 [124] J. Linhard et al.,Mat.-Phys. Medd. 33 10, (1963) [125] J. Lindhard, M. Scharff and H. E. Schiott, Mat.-Phys. Medd. 33, 14 (1963). 38,42 [126] J.F. Ziegler, Historical Review on The Stopping and Range of Ions in Matter http://www.srim.org/SRIM/History/HISTORY.htm 38,39 [127] J. F. Ziegler, et al,Stopping and Ranges of Ions in Matter series Pergamon Press, New York (1978-1984). 39 [128] http://www.srim.org 39 [129] P. Majewski et al., Simulations of the Nuclear Recoil Head-Tail Signature in Gases Relevant to Directional Dark Matter Searches, Astr. Ph., 35, 5(2010) 284. 40,44,45 [130] J.D Lewin and P.F. Smith, Review of mathematis, numerical factors, and corrections for dark matter experiments based on elastic nuclear recoil, Astr. Ph., (1996) 6:87. 41,43,51 [131] A. Hitachi, Quenching factor and electronic LET in gas at low energy, Proceedings of the 3rd Symposium on Large TPCs for Low Energy Rare event Detection, J. Phys. Conf. Ser. 65 (2007) 012013. 42,44 [132] J.A. Phipps, J.W. Boring and R.A. Lowry, (1964) Phys. Rev. 135, A.36 42 [133] A. Tomás, F.J. Iguaz and I.G. Irastorza, talk given in the First Workshop on Directional Detection of Dark Matter (CYGNUS 2007). http://www.hep.shef.ac.uk/conferences/cygnus2007/talks/zaragoza.pdf 42,44,45 [134] J. Billard, F. Mayet and D. Santos, Three-dimensional track reconstruction for directional Dark Matter detection. JCAP 04 (2012) 006 44,45 [135] D. Dujmic et al. (DMTPC Collaboration) Astr. Ph., 30:58, (2008). 44 [136] D. Dujmic et al. (DMTPC Collaboration), Observation of the “head-tail” effect in nuclear recoils of low-energy neutrons, Nucl. Instrum. Meth. A 584:327 (2008). 44 [137] G.F. Knoll. Radiation detection and measurement. John Wiley and Sons, 3rd edition 2000. 46 [138] M.J. Berger et al., XCOM: Photon Cross Sections Database. http://www.nist.gov/pml/data/xcom/index.cfm 46,47,125,126 310 Bibliography [139] S.Y.F. Chu, L.P. Ekstrom and R.B. Firestone. The Lund/LBNL Nuclear Data Search, http://nucleardata.nuclear.lu.se/nucleardata/toi 47 [140] G. Jungman, M. Kamionkowski and K. Griest. Supersymmertric dark matter. Phys. Rep., 267:198-373, 1996. 51 [141] Evaluated Nuclear Data File (ENDF) Retrieval & Plotting. http://www.nndc.bnl.gov/sigma 52 [142] I.G. Irastorza, Un telescopio subterráneo de axiones y WIMPs como instrumento de investigación astrofísica y detector de Materia Oscura, Universidad de Zaragoza, 2001. 52 [143] S. Cebrián. Estudio del fondo radioactivo en experimentos subterráneos de búsqueda de sucesos poco probables: CUORE y ANAIS, PhD Thesis, Universidad de Zaragoza, 2002. 52,113,118 [144] M. Martínez. Diseño de un prototipo para un experimento de detección directa de materia oscura mediante modulación anual con centelleadores de ioduro de sodio, PhD Thesis, Universidad de Zaragoza, 2006. 52,113,118 [145] D.H. Perkins. Introduction to High Energy Physics, Cambridge University Press, 4th edition (2000). 53 [146] Y.S. Tsai, Rev. Mod. Phys. 46, 815 (1974). 53 [147] F. Sauli, Principles of operation of Multiwire Proportional and Drift Chambers, CERN Yellow Reports 77-09, (1977). 56 [148] D. Combecher, Proceedings of the 3rd Symposium on Neutron Dosimetry in Biology and Medicine, ed. by G. Burger and H.G.Elbert (1977) 97.56 [149] N. Palladino and B. Sadoulet, Nuclear Instrum. Methods 128, 323 (1975). 60 [150] S.F. Biagi, Montecarlo simulation of electron drift and diffusion in counting gases under the influence of electric and magnetic fields, Nucl. Instrum. Meth. A 421 (1999) 234. Program and cross sections available in http://www.cern.ch/magboltz 60,62,132 [151] J. Townsend, Electrons in gases (Hutchinson, London, 1947). 60 [152] C.J. Martoff et al., Low Pressure Negative Ion TPC for Dark Matter, Nucl. Instr. and Meth. A 440 355(2000). 65 [153] Y. Giomataris, P. Rebourgeard, J.P. Robert and G. Charpak, MICROMEGAS: A high-granularity position-sensitive gaseous detector for high particle-flux environments, Nucl. Instrum. Meth. A 376 (1996) 29. 67 Bibliography 311 [154] F.J. Iguaz, E. Ferrer-Ribas, A. Giganon and I. Giomataris, Characterization of microbulk detectors in argonand neon-based mixtures. arXiv:1201.301. 73 [155] S. Cebrián et al., Micromegas readouts for double beta decay searches, JCAP 10 (2010) 010. 74,98,99 [156] S. Cebrián et al., Micromegas-TPC operation at high pressure in xenontrimethylaminte mixtures, (2012). Preprint arXiv:1210.3287. 75 [157] M.E. Rose and S.A. Korff, An Investigation of the Properties of Proportional Counters, Phys. Rev. 59 850-9 (1941) . 75 [158] W. Diethorn, A methan proportional counter system for natural radiocarbon measurements, UASEC Report NY06628 (1956). 76 [159] Y. Giomataris, Development and prospects of the new gaseous detector ’Micromegas’, Nucl. Instr. and Meth. A 419 (1998) 239-252. 76,77 [160] M. Chefdeville et al., Pulse height fluctuations of integrated Micromegas detectors, Nucl. Instr. and Meth. A 591 (2008) 147-150. 77 [161] T. Dafni et al., Energy resolution of alpha particles in a Micromegas detector at high pressure, Nucl. Instrum. Meth. A 608 (2009) 259. 77,95,96,97,98 [162] W. Legler, The influence of the relaxation of the electron energy distribution on the statistics of electron avalanches, Brit. J. Appl. Phys. 18 (1967). 79, 80 [163] J. Byrne, Statistics of the electron multiplication process in proportional counters, Proc. R. Soc. Edinburgh,XVI A 33 (1962). 79 [164] J. Derré et al., Fast signals and single electron detection with a MICROMEGAS photodetector, Nucl. Instrum. Meth. A 449 314-321 (2000). 79,91,97 [165] G.D. Alkhazov, Statistics of electron avalanches and ultimate resolution of proportional counters, Nucl. Instrum. Methods 89, 155 (1970). 80 [166] T. Alexopoulos et al., Micromegas study for the sLHC environments, JINST 5P02001 (2010). 88 [167] S. Ramo, Currents induced in electron motion, Proc. IRE 27, 584 (1939). 90 [168] J. Dolbeau et al., The solar neutrino HELLAZ project, Nucl. Phys. Proc. Suppl. 138 (2005) 94. [169] A. Tomás et al., Development of Micromegas for neutrinoless double beta decay searches, 2009 JINST 4P11016. 94,95,104 312 Bibliography [170] S. Cebrián et al., Micromegas-TPC operation at high pressure in xenontrimethylamine mixtures, 2013 JINST 8P01012. 95,99,100 [171] C. Balan et al., Micromegas operation in high pressure xenon: charge and scintillation readout, 2011 JINST 6P02006. 97,98,100 [172] B. Ramsey and P. Agrawal, Xenonbased Penning mixtures for proportional counters, Nucl. Instrum. Meth. A 278 (1989) 576. 99 [173] C.A.N. Conde and A.J.P.L. Policarpo, Nucl. Instr & Meth. 53 (1967) 7-12. 100 [174] D. Nygren, Can the intrinsic energy resolution in xenon be surpassed?, J. Phys.: Conf. Ser. 309 (2011) 012006. 100 [175] S. Cebrián et al., Radiopurity of micromegas readout planes, Astroparticle Physics 34 (2011) 354-359. 101,172 [176] E. Aprile, The XENON100 Dark Matter Experiment at LNGS: Status and Sensibility, J. Phys: Conf. Ser. 203 (2010) 012005. 102 [177] T2K Collaboration, The T2K experiment, arXiv:1106.1238. 102 [178] D. Calvet et al., AFTER, an ASIC for the Readout of the Large T2K Time Projection Chambers, Nuc. Science, IEEE Trans. on 55 3 (2008) 1744-1752. 102 [179] SAMTEC, GFZ-30-03-G-10-AD. 105,106 [180] LeCroy WR6050. 108 [181] ROOT team, ROOT framework for data processing, root.cern.ch 112 [182] S. Agostinelli et al., GEANT4 - a simulation toolkit, Nucl. Instrum. Meth. A506 (2003) 250-303 113 [183] J. Allison et al., GEANT4 developments and applications, IEEE Transcations on Nuclear Science 53 (2006) 270. 113 [184] GEANT4 Collaboration, Application Developer’s Guide, http://geant4.cern.ch 114 [185] O.A. Pokratenko et al., Event Generator DECAY4 for simulation of double beta processes and decays of radiactive nuclei, Phys. At. Nucl. 63 (2000) 1282-1287. 120 [186] CLHEP - A Class Library for High Energy Physics, http://proj-clhep.web.cern.ch 120 Bibliography 313 [187] J. Galán, Probing eV-mass scale Axions with a Micromegas Detector in the CAST experiment, PhD Thesis, Universidad de Zaragoza, 2011. 2011 JINST TH 003.138,161,167,173,174,175,177,179 [188] J. Billard, F. Mayet and D. Santos, Low energy electron/recoil discrimination for directional Dark Matter detection, JCAP 07 (2012) 020. 153 [189] S. Andriamonje et al.(CAST collaboration), An improved limit on the axionphoton coupling from the CAST experiment, J. Cosmol. Astropart. Phys 04 (2007) 10.159 [190] E. Arik et al., (CAST collaboration), Proving eV-scale axions with CAST, J. Cosmol. Astropart. Phys 02 (2009) 008.161 [191] M. Arik et al., (CAST collaboration), Search for Sub-eV Mass Solar Axions by the CERN Axion Solar Telescope with 3He Buffer Gas, Phy. Rev. Let., 107 26 (2011) . 161 [192] CAST collaboration, CAST final results with 3He Buffer Gas, in preparation (2013). 161 [193] I.G. Irastorza et al., The International Axion Observatory (IAXO), Proceedings of the 7th Patras Workshop on Axions, WIMPs and WISPs, Mykonos, Greece, 2011. arXiv:1201.3849.17,161,282 [194] D. Kang, Search for solar axions with the CCD Detector at CAST (CERN Axion Solar Telescope). PhD thesis Albert-Ludwings-universität, 36, 2007. 161 [195] D. Autiero et al., The CAST time projection chamber. New Journal of Physics, 9(6):171, 2007, 48 163 [196] G. Luzón et al., Background studies and shielding effects for the tpc detector of the CAST experiment, New Journal of Physics, 9(7):208, 2007, 49 163 [197] S. Cebrián et al., Background study for the pn-CCD detector of CERN Axion Solar Telescope, Astroparticle Physics,48 28:205-215 (2007). 163 [198] J. Galán et al., New Micromegas detectors in the CAST experiment,NIM A 604, 15 (2009). 161,165,166 [199] J.C. Santiard et al., Presented at the 6th Pisa Meeting on Advanced Detectors (1994) CERNECP-94-17.164 [200] T. Geralis , G. Fanourakis, Y. Giomataris and K. Zachariadou, The data acquisition of the Micromegas detector for the CAST experiment, IEEE Nucl. Sci. Symp. Conf. (2003) 53455-99. 164 314 Bibliography [201] D. Breton, E. Delagnes and M. Houry, Very high dynamic range and high sampling rate VME digitizing boards for physics experiments, IEEE Trans. Nucl. Sci. (2005) 52 2853. 164 [202] T. Dafni. A Search for Solar axions with the MICROMEGAS Detector in CAST. PhD thesis, Technischen Universität Darmstadt, 2005. 36, 54, 108. 173,176 [203] T. Vafeiadis, Contribution to the search for solar axions in the CAST experiment, PhD thesis, Aristotle University of Thessaloniki (2012), to be published. 173,176,182,282 [204] C. Yildiz, PhD thesis, Dogus University, Istanbul (2013), in preparation. 173 [205] P. Abbon et al., The Micromegas detector of the CAST experiment, New Journal of Physics 9(2007) 170. 164 [206] Dumond Gerard, Background Measurement for CAST experiment, Technical Note,CERN-SC-2004-027-RP-TN. 211 [207] T. Dafni et al., Background Measurement at the CAST experiment, CAST internal note (2012). 212,216,217 [208] A. Rodríguez, internal communication (2012). 212 [209] E. Ruiz Chóliz, Detección directa de materia oscura en el Laboratorio Subterráneo de Canfranc, TAD in the F’sica Atómica, Molecular y Nuclear Area of the University of Zaragoza (2012). 213 [210] G. Luz’on, internal communication (2012). 213 [211] A. Tomás, Brief report on simulations about 57Co source tests in SRMM, CAST internal note, Feb 2012. 227 [212] T. Vafeiadis, talk given in the 48th CAST Collaboration Meeting (Patras, May 2012). 240 [213] NI USB-6008, http://sine.ni.com/nips/cds/view/p/lang/es/nid/201986 245 [214] http://www.genitron.de/products/alpha_slides.html 245 [215] A. Tomás, CAST Microbulk Micromegas in the Canfranc Underground Laboratory, Proceedings of the 2nd International Conference on Technology and Instrumentation in Particle Physics (TIPP 2011), Physics Procedia 37 (2012) 478-482. 245 [216] J. Galán, talk given in the 41st CAST Collaboration Meeting (CERN, Feb 2010). 245 Bibliography 315 [217] A. Tomás, talk given in the 42nd CAST Collaboration Meeting (CERN, Jun 2010). 245 [218] LSC website, http://www.lsc-canfranc.es 246 [219] G. Luzón et al. Characterization of the Canfranc Underground Laboratory: status and future plans. Proceedings of the International Conference in the Identification of Dark Matter (IDM2006), pp 514-519. 246,247 [220] A. Tomás, talk given in the 43rd CAST Collaboration Meeting (CERN, Jan 2011). 249 [221] J.A. García, talk given in the 47th CAST Collaboration Meeting (CERN, Jan 2012). 264 [222] S. Cebrián and D.C Herrera, private communication. 258 [223] ILIAS radiopurity data base http://radiopurity.in2p3.fr [224] Taken from NEXT experiment radiopurity data base. 258 [225] F.J. Iguaz et al., CAST internal document (Jan 2011). 264 [226] J.G. Garza, talk given in the 47th CAST Collaboration Meeting (CERN, Jan 2012). 264,278 [227] Th. Papaevangelou for the CAST Collaboration, 2012 Status report of the CAST Experiment and Running in 2013-2014, CERN-SPSC-2012-028. 267, 268,272,277,278,280 [228] J.A. García, talk given in the 48th CAST Collaboration Meeting (Patras, May 2012). 267 [229] A. Diago, talk given in the 49th CAST Collaboration Meeting (CERN, Sep 2012). 267 [230] J.G. Garza, talk given in the 50th CAST Collaboration Meeting (CERN, Feb 2013). 279 316 Bibliography