LENA Detector development ANT2010, Santa Fe 2010/09/18 Marc Tippmann Technische Universität München

Most relevant topics at the moment Detector construction Cavern construction Tank design Liquid handling Scintillator purification Liquid scintillator Attenuation length vs. wavelength Decay time constants Emission spectra Photo sensors Photo sensor requirements PMT characterization Winston Cones Pressure encapsulations Electronics PMT arrays Software Geant4 simulation of optical model of detector Tracking with Liquid Scintillator Detectors

LENA (Low Energy Neutrino Astronomy) Physics objectives: Low energy: Neutrinos from galactic Supernova Diffuse Supernova neutrinos Solar neutrinos Geoneutrinos Reactor neutrinos Indirect dark matter search Higher energy: Proton decay Long-baseline neutrino / beta beams Atmospheric neutrinos Detector layout: Liquid scintillator ca. 50kt LAB Inner vessel (nylon) Radius (r) = 13m Buffer (inactive liquid scint.) Δr =2m Cylindrical steel tank, 48k PMTs (8“) with Winston Cones (2x area) r = 15m, height = 100m, optical coverage: 30% Water cherenkov muon veto 5000 PMTs, Δr > 2m to shield fast neutrons Cavern egg-shaped for increased stability Rock overburden: 4000 mwe

Detector construction

Detector construction LAGUNA: European site + design study for a next generation neutrino and p-decay detector 7 preselected sites Proposed experiments: GLACIER, LENA, MEMPHYS Includes cavern design study

Detector construction: Cavern construction New: Rockplan tank + excavation study for Pyhäsalmi Based on Technodyne study → substantial improvements Worked out excavation process and extra structures to fulfill safety requirements: 2 access tunnels, spherical work tunnel, 1 or 2 new shafts Long term rock stability simulations → elliptical horizontal cross-section and kink in vertical cross-section → Higher volume for Water Cherenkov detector

Detector construction Tank design Conventional Steel Tank + well known, straightforward to build, robust - expensive, single passive layer defense, a lot of elements and connections Sandwich Steel Tank + cost effective, room for cooling, fast install, high QA/QC, laser welds - a lot of welding, little used solution, mechanically challenging Sandwich Concrete Tank + well known, straightforward to build, robust, improved physics - steel plates and rebar prevent continuous casting, slow to build Hollow Core Concrete Tank + room for cooling, mechanically strongest, improved physics, quick build - little used solution, active leak prevention may lead to sustained pumping

Liquid Scintillator properties Extensively covered last year on ANT09 by Michael Wurm → only new results Liquid Scintillator properties are essentially understood + known Only particular measurements remain to be done Started to repeat measurements with higher precision

Liquid scintillator properties Attenuation length vs. wavelength Michael Wurm Wavelength (nm) Attenuation length (m) 10m Martin Hofmann Already measured with a 1m scintillator tube (10nm accuracy) + 10cm cell (1nm) → transparency considerably increases with wavelength → need longer tube → New experiment with ≈10m tube length Is being set up at new Munich underground lab (≈ 10mwe rock overburden) All parts ordered, most have arrived Should be able to start measurements within the next months 1m

Liquid Scintillator properties Decay time constants Paolo Lombardi, LENA Meeting 2010-07-05 Paolo Lombardi has set up a new experiment in Milan Measured time behavior of different scintillators (PC, LAB, DIN) with varying PPO concentration → Cross-check with Munich experiment possible Measurements on LAB in both experiments in good agreement in some variables, deviation in others → maybe due to different manufacturers / batches →will study this further preliminary

Liquid Scintillator properties Emission spectra Paolo Lombardi, LENA Meeting 2010-07-05 Measurements done in Perugia by Fausto Ortica and Aldo Romani → Cross-check with Munich experiment possible Very good agreement of results for LAB with 1.5 resp. 2g/l PPO Teresa Marrodan, PhD thesis

Photo sensors

Photo sensors: Photo sensor requirements PMT timing behavior Normal Pulses Early Pre-Pulses Late Pulses Transit Time Spread (FWHM, spe) Dark Noise

Photo sensors: Photosensor requirements Area Inner Detector: 10430 m² Targeted optical coverage: 30% → 3130 m² effective photosensitive area PMTs probably the only photosensor type Meeting the physical requirements Having a low enough cost /(area*photon detection efficiency) Durable for at least 30 years AND Utilizable until start of construction Important properties: Transit time spread, afterpulsing, gain, dynamic range, area, quantum efficiency, dark noise, peak-to-valley-ratio, early + pre-pulsing, late pulsing, pressure resistance, long term stability, low radioactivity, price

Photo sensors: Photosensor requirements Necessary parameters for PMTs Diameter Transit Time Spread (FWHM) Ion Afterpulses, Brems-strahlung AP Dark Noise Dynamic Range (pe) Gain Number PMTs in Inner Detector (using 2x Winston cones ) 3“ <3ns <1% (if not less!) <0.75 kHz 1 to > 10 some 106 343,000 5“ “ <2kHz 1 to > 40 124,000 8“ <5kHz 1 to >100 48,000 10“ <8kHz 1 to >150 31,000

Photo sensors: PMT characterization Selection of feasible PMT models for LENA Manufacturers available for 100k+ PMTs: Hamamatsu Photonics Electron Tubes Enterprises Ltd. Study most promising PMTs with diameters from 3“-10“ Measured until now: Hamamatsu: R6091(3“), R6594(5“), R5912(8“) and R7081(10“) ETEL: 9351(8“)

Photo sensors: PMT characterization Borexino PMT testing facility @LNGS: Excellent time resolution: 410nm laser diode light source, pulse FWHM <30ps, trigger output jitter <100ps; electronics jitter ≈45-90ps → total time resolution <140ps Can measure up to 32 PMTs simultaneously Repaired + reactivated setup Further improved setup: Use second channel to measure fast AP up to ≈25-30 ns after primary pulse, triggered by pulse in first channel Low noise preamp near PMT → increased p/V by a factor of 2

Photo sensors: PMT characterization Results: Comparison R6594(+) vs Photo sensors: PMT characterization Results: Comparison R6594(+) vs. R7081(+) 5“ R6594 (5“) R7081 (10“) Voltage +1670V +1520V Gain 1.0∙107 1.3∙107 pe/trigger (npe) 5.53% 2.91% TTS(FWHM) (acc. to HP) 1.91ns (1.5ns) 3.05ns (3.5ns) EP (all non-gaussian) 2.95% 0.57% LP (all non-gaussian) 34.16% 42.90% LP (after NP peak) 3.13% 3.09% 10“

Photo sensors: PMT characterization - Comparison PMTs R6091 (3“) with 1.8“ aperture R6594 (5“) R5912 (8“) R7081 (10“) ETL9351 (8“) no. 1732 ETL9351 (8“) average Voltage +1760V +1670V +1425V +1520V +1500V ≈+1450V Gain 1.0∙107 1.3∙107 pe/trigger (npe) 2.21% 5.53% 1.83% 2.91% 4.78% 5.19% TTS (FWHM) (manufacturer) 1.89ns (2.0ns) 1.91ns (1.5ns) 2.04ns (2.4ns) 3.05ns (3.5ns) 2.16ns 2.76ns EP (all non-gaussian) 0.14% 2.95% 1.93% 0.57% 1.23% 0.75%(3σ) LP (all non-gaussian) 27.29% 34.16% 55.01% 42.90% 8.70% - LP (after NP peak) 6.26% 3.13% 2.88% 3.09% 4.08% 7.90%(3σ) DN 0.192kHz (5.23kHz) 1.62kHz 2.64kHz 1.72kHz 2.48kHz DN/m² 120.9 kHz/m²(eff.) (462.6 kHz/m²) 50.5 kHz/m² 52.6 kHz/m² 53,0 kHz/m² 76,5 kHz/m² Ion AP 0.94% 6.62% 5.12% 2.57% 4.9% p/V 2.04 3.88 2.99 3.09 2.25 2.10

Photo sensors: PMT characterization Improve + extend Munich PMT testing setup PMT test setup at TUM, Garching (currently) FlashADC (10bit, up to 8 Gigasample/s = 125ps time resolution, 4 channels) → excellent Currently using LED (430nm peak-wavelength) with moderate relaxation time (≈5-10ns) driven by 5ns FWHM voltage pulse → bad → Build in fast light pulser developed by George Korga: fast LED driven by Avalanche Diode -> total time jitter FWHM ≈1ns or lower Medium term: build in ps laser diode → time jitter ≈30ps → Confirm results from Italy → Can measure in addition pulseshape + fast afterpulses (Δt<30ns) + prepulses Status: coincidence running; working with two students on improving readout + evaluation software

Photo sensors: PMT characterization HQE measurements Paolo Lombardi, LENA Meeting 2010-07-05 (Another) experimental setup of Paolo Lombardi in Milan to characterize PMTs: Picosecond laser (405nm) 8bit 1GS/s FADC Measured Hamamatsu super-bialkali PMTs (~34% peak-QE): R6594-SEL, R5912-SEL, R7081-SEL → Direct comparison with standard photocathode PMTs measured @ Borexino PMT testing facility possible

Photo sensors Winston Cones Attach Compound Parabolic Concentrators („Winston Cones“) to increase effective area of PMTs by a factor of 2-8(!) → increases number of detected photons / MeV deposited → increased resolution Limits field of view by introducing an acceptance angle (conservation of etendue) → can be used to limit field of view to fiducial volume …coming soon: study influence on detector performance with the Geant4 Monte Carlo simulation of LENA (software development finished + working) Borexino Winston Cone CTF Winston Cone

Photo sensors Pressure encapsulations Pressure at tank bottom might be too high for PMT glass envelopes → a) Try if thicker glass envelope (4mm) fulfills pressure tests b) Develop pressure encapsulations → Can use standard PMTs Integrate Winston Cones + µ metal shielding into design Starting points for development e.g. Borexino encapsulation with pressure-withstanding window instead of thin PET window with pressure-resistant window in front Development will commence soon

Electronics

Electronics: PMT arrays Employ PMT arrays with central on-site front-end electronics on chip → Reduce number of channels (only 1 cable for data + voltage); however voltage not adjustable for individual PMTs after installation + can read out pulse shape only for sum of PMT channels Currently under development in PMm² project PICS collaboration between LENA + MEMPHYS → can use existing R&D of PMm² (3 years) for LENA! PMm²: 16 PMs(12“) + frontend electronics (PARISROC) PMm² has studied this concept for the past 2 years, however mainly for MEMPHYS (partially in general for large experiments with PMs), scheduled end of R&D mid 2010 -> much experience, great part of R&Dalready done, PMm² PMA, collaboration of LENA/MeMPhys?? Concept: array of PMs with pre-evaluation electronics instead of single PMs and centralized electronics (chip) Schema(PMm²) + explanation (as short as possible) For LENA: Pros (main reason: lower costs) + Cons, pros>cons (how can cons be compensated) J.E. Campagne, LENA - PMm² meeting 07/04/2009

Electronics: PMT arrays PMm² frontend electronics: PARISROC Triggerless Individual channels Charge + time measurement Variable gain of preamplifier Common HV for PMs 1/3pe 100% efficiency 1ns time resolution 16 trigger channels or 1 OR-trigger-output 2 gains per channel (not showed here): 0-10pe & 0-300pe Scaleability

Electronics: PMT arrays Advantages: Data transfer digital → less attenuation Lower cost (less cables, voltage in cables smaller, bundled frontend electronics reduces data processing cost/channel) Disadvantages: Same HV for each PMT within module → counter measures: partial compensation via individual gain adjustment, combine similar PMTs for a module Different physical requirements for PMT Arrays → adopt R&D of PMm² as far as possible + adapt setup where neccessary Can‘t adapt voltage of each PM to achieve same gain, can‘t compensate aging effects Similar PMs -> need all more or less same voltage for same gain Basically a lot of great advantages and only one disadvantage which can be mostly compensated -> interesting concept Different Physical requirements(which pr are there??) from original purpose of PMm² => adopt concept and development of PMm² as far as possible + adapt setup where neccessary: PMs and electronics One especially important aspect: After pulses (what other aspects are important for apllication of PMAs to Lena??)

Software

Software Tracking in Liquid Scintillator Detectors Light front generated by GeV particles resembles a Cherenkov cone → directionality → can use arrival time of first photons on PMTs and total photon count for tracking → Can be used for p decay, neutrino beams and atmospheric neutrinos Have developed two event reconstruction programs for tracking: „Scinderella“ by Juha Peltoniemi Tracking script for the Geant4 LENA optical detector model by Dominikus Hellgartner J. Learned, arXiv/0902.4009, J. Peltoniemi, arXiv/0909.4974

Software: Tracking in Liquid Scintillator Detectors: „Scinderella“ Single-Particle Tracks (leptons, CC QES): excellent reconstruction Multi-Particle Vertices (deep inelastic neutrino scattering creates pions, 1-5GeV): new reconstruction algorithm fits a superposition of the light fronts generated by MC sample tracks to the overall PMT signal → for ≤ 3 pions good track reconstruction at ≤ 5% E resolution possible → LSD is good tracking detector at E > 1GeV: excellent flavour recognition, E resolution typically < 5%, very good for neutrino beams: sin²(2ϑ) > 10-2 – 10-3 if not lower If pulse-shape of PMT signals available even better Detector simulation software based on Java developed by Juha Peltoniemi Can do both event simulation and event reconstruction Event simulation → record observation time of photons in PMTs Track reconstruction by comparing light pattern produced by test track with that of track from simulation → minimalization of deviation Reco of 4GeV  deep-inelastic sample event J. Peltoniemi

Software: Tracking in Liquid Scintillator Detectors Tracking script for Geant4 optical model Developed by Dominikus Hellgartner based on work done by Michael Wurm on tracking in Borexino Procedure: minimize log-likelihood value of photon pattern depending on all track parameters → Look at even lower energies than 1GeV: First preliminary results: a) 100 500MeV single-track muons: very good track reconstruction with <≈1+/-2cm position uncertainty and 0.1+/-0.1ns time uncertainty for the starting point and 2.5° direction deviation b) 100 750MeV single-track electrons: all uncertainties comparable to muon events, however one problem occurring: for some events, the direction is reconstructed with wrong sign → starting point at end of track → outliers; are working on this right now Dominikus Hellgartner

Software: Tracking in Liquid Scintillator Detectors Proof of principle: Borexino Borexino is actually a low E calorimetry detector; not optimized for tracking at all LENA will be designed for tracking capability Reconstructed tracks from CNGS beam in Borexino No simulation: real data! Michael Wurm

Summary Pyhäsalmi site study finished → elliptical cavern, steel or concrete tank Liquid scintillator properties mostly known, setting up attenuation length precision measurement at the moment Photo sensor: 5“-10“ PMTs (which type will be decided shortly) with Winston Cones and pressure encapsulations Bundle PMTs to arrays to save channels Geant4 detector simulation is running, have to measure last missing properties for optical model Liquid Scintillator Detectors are actually quite good at tracking

Backup slides

Detector construction Excavation

Scattering Length Results no hints for Mie-scat. anisotropic scattering in good agreement with Rayleigh expectation correct wavelength- dependence found literature values for PC, cyclohexane correctly reproduced Results for l=430nm LS = 22±3 m after purification in Al2O3-column Michael Wurm

Liquid Scintillator properties Quenching factors Gamma Quenching light output of low-energetic electrons (E<200keV) by small-angle Compton scattering …in progress q Timo Lewke, Jürgen Winter Proton Quenching using neutron recoils (AmBe-source, n-generator) …coming soon scintillator sample

Proton decay Large impact on proton decay (into K+n) detection efficiency: Signal of kaon (t=13ns) and of its decay products (mostly muons) must be separated by rise time analysis Background Source: Atmospheric Muon-Neutrinos Kaon Muon Efficiency ranges from 56% to 69% for typical rise times of 7 to 10 ns

Proton decay

Software: Tracking in Liquid Scintillator Detectors: „Scinderella“ Single-Particle Tracks excellent flavour recognition Positional accuracy: a few cm Angular accuracy: a few degrees initial neutrino energy + momentum can be determined at 1%(?) accuracy from lepton track (CC QES) Multi-Particle Vertices - deep-inelastic scattering of neutrinos creates pions (1-5 GeV) - new reconstruction algorithm fits a superposition of the light fronts generated by MC sample tracks to the overall PMT signal - resolution of vertices featuring ≤ 3 ‘s is possible at % accuracy - recoil protons/neutrons quenched, not invisible, allowing a calorimetric measurement of the event energy deposited. - time-resolved response of individual PMTs needed for optimum result

Software: Tracking in Liquid Scintillator Detectors Tracking script for Geant4 optical model Developed by Dominikus Hellgartner based on work done by Michael Wurm on tracking in Borexino Procedure: For muon: separate track signal + decay signal: look for second peak in detected photons vs. time -> time cut Determine start values for fits: Energy → from number of photons of track; barycenter fit → position on track; first PMT hit → crude direction: vertical or horizontal Vertical: Fit point-like source from first hit times for PMT-ring containing minimum hit time PMT → starting point → with barycenter fit direction → with energy track length → track Horizontal: TOF-correction to barycenter → direction from most negative first hit times; table of position of charge barycenter vs track E (input from other simulations) → with energy track Main fit: minimize log-likelihood value depending on all track parameters → Look at even lower energies than 1GeV: First preliminary results: a) 100 500MeV single-track muons: very good track reconstruction with <≈1+/-2cm position uncertainty and 0.1+/-0.1ns time uncertainty for the starting point and 2.5° direction deviation b) 100 750MeV single-track electrons: all uncertainties comparable to muon events, however one problem occurring: for some events, the direction is reconstructed with wrong sign → starting point at end of track → outliers; are working on this right now Dominikus Hellgartner

Requirements on PMTs in LENA

Other types of photosensors Avalanche Photodiode (APD): solid state equivalent to PMTs; price/area too high; Dark Noise very high, if not cooled; low amplification SiPM: array of small APDs; excellent tts (FWHM < 0.5ns), excellent energy resolution, high QE(?), immense cost/area; huge dark count (some 100kHz up to several MHz) Cupid/ Hybrid Photo-Detector: Photocathode just like in PMTs, e- however focussed on APD in center instead of dynode chain; problems with huge DN? Still working on fixing that Microchannel Plate (MCP): Photocathode, e- accelerated along very thin etched channels perpendicular to cathode, coated with metal on inside, thickness of coating varies along axis -> increasing voltage along flight path, e- crashes on walls + knocks out more e- -> amplification

Other types of photosensors Quasar: Photocathode, solid scintillator block in center with a small PMT looking at it; apply very high voltages (typ.25-30kV) -> e- creates scintillation light in crystal -> light signal on PMT advantages: small jitter even for large photocathode area, excellent energy resolution, …, price=? Already used in BAIKAL, however needs further development Hybrid-Gas Photomultiplier: photocathode in vacuum, e- accelerated through mebrane into Thick Gas Electron Multiplier (THGEM) -> accelerated through holes in metal-covered plate, different voltage on upper+ lower side -> very high field inside holes -> e- scatters off gas atoms -> produces further e- -> amplification; use several THGEMs in series for higher amplification still under development, needs several more years to be ready for production

Photomultiplier Tubes (PMTs) Principle of operation Extremely sensitive light detectors capable of detecting single photons Photon → photo electron(pe) → amplified by ≈10^7 in dynode chain Gain = amplification factor

PMTs Ocurring effects + basic parameters Ion Afterpulses Dark Noise Primary Pulses (Next laser pulse) Afterpulse: delayed additional pulse caused by the primary pulse Ion Afterpulses: typically 0.3-15µs later Bremsstrahlung Afterpulses:typ. 30-60ns later Dark Noise: electrons emitted from cathode without incident photon; mostly thermionic emission

PMTs Ocurring effects + basic parameters Single photo electron (spe) spectrum: 3 components: normal pulses, underamplified pulses + noise Peak-to-valley-ratio (p/V): ratio between value of maximum and valley noise under- amplified pulses spe (normal pulses) Peak 2 pe Valley

PMTs Voltage behavior Two different modes of applying voltage: a) Photocathode at ground potential → positive voltage on anode → need capacitor to separate DC voltage from electronics b) Negative voltage applied to photocathode → anode at ground potential → signal can be sent directly to electronics Advantage: faster Disadvantage: DN higher Dynamic Range: number of photons incident at the same time on the cathode, which can be detected Raise voltage → gain rises, transit time decreases, jitter decreases, however: AP rise, DN rises if voltage is very high (field emission), dynamic range decreases → find optimum working point

PMTs in LENA Transit time spread must be low → tracking Bremsstrahlung Afterpulses (fast AP) can blur out proton decay coincidence Ion Afterpulses (slow AP) can prevent position reconstruction of neutrons → decreases detection rate of cosmogenics (C11): C11 is main background for CNO-ν-flux Michael Wurm, TUM, LENA - PMm² meeting 07/04/2009

PMTs in LENA High Dark Noise: random coincidences of DN pulses + bad for tracking; DN approx. Q cathode area Gain must be high enough or part of pulses overlaps with noise in spe spectrum → less pulses usable Dynamic range: Detector must be able to detect single photons as well as HE events (muon, proton decay) Assuming 1GeV event, 500pe/MeV, illuminating only 10% of PMTs → 1pe - 1700pe/(m² effective photosensitive area) = 1pe – 100pe for 8“ PMT with 2x Winston Cone upper boundary x10 if only 1% illuminated p/V: the bigger it is, the better spe events can be distinguished from noise

PMTs in LENA EP + PreP + maybe LP bad for tracking (uses first pulses = first photons) Area per sensor: Smaller → better tracking, dynamic range of detector increases, transit time spread smaller Bigger → less sensors + channels → cost/area lower (if not too big) → generally sensor area as small as affordable Quantum efficiency: higher → less PMTs needed for same energy resolution + threshold Pressure resistance: At least 10 (100m head of liquid scintillator) +1 (vacuum in tube) + 3 (safety margin for implosions) = 14bar at the moment no PMTs meet the pressure requirements → maybe use encapsulations for lower PMTs

Measurements at the Borexino PMT testing facility @LNGS (INFN), Italy

Measurements @LNGS Setup + Measurements Setup can measure: Transit time (TDC) → TTS, EP, PreP, LP AP/Primary Pulse (PP) transit time (MTDC) → slow AP (+some fast AP), DN, pe/laser trigger (npe) In second channel: transit time of first AP (TDC) → fast AP For both channels: charge spectrum (ChargeADC) -> p/V Measurements: Measured all PMTs with positive + negative HV at a gain of 1·107 Find new operating point for LENA → Sampled over complete reasonable voltage range for pos. voltages → scanned PMT properties over complete feasible range of gain for single photon electrons

Measurements @LNGS Measurements Also: Developed fast LED light source with emission time spread (fwhm) <≈ 1ns Measured R6233-100SEL: 3“ UBA photocathode, max. QE ≈43%(!) Measured several ETL9351 PMTs for comparison → gave one to Paolo Lombardi to calibrate new flash ADC setup in Milano → planned to exchange data of measured PMTs; e.g. R7081 ↔ R7081-100SEL (UBA) Evaluated up to now: Runs with ≈107 gain + positive voltage

Borexino PMT testing facility Block diagram - original configuration

Measurements @LNGS Results: Comparison R6594(+) vs. R7081(+) Voltage +1670V +1520V Gain 1.0∙107 1.3∙107 npe 5.53% 2.91% DN (5.23kHz) 2.64kHz DN/m² (462.60 kHz/m²) 52.59 kHz/m² Ion AP 0.94% 5.12%

Measurements @LNGS Results: Comparison R6594(+) vs. R7081(+) Voltage +1670V +1520V Gain 1.0∙107 1.3∙107 npe 5.53% 2.91% DN (5.23kHz) 2.64kHz DN/m² (462.60 kHz/m²) 52.59 kHz/m² Ion AP 0.94% 5.12%

Measurements @LNGS Results: Comparison R6594(+) vs. R7081(+) Voltage +1670V +1520V Gain 1.0∙107 1.3∙107 npe 5.53% 2.91% p/V 3.88 3.09

Measurements @LNGS Preliminary Results Most promising candidate – judging from only one sample each: Hamamatsu R6594 (5“ diameter) using positive HV (dark noise rate much higher for negative HV): transit time spread 1.9ns, total afterpulse rate 0.9%, peak-to-valley ratio 3.9 → Need more samples to be able to determine mean parameters