Large-Scale, Precision Xenon Doping of Liquid Argon
N. McFadden, S. R. Elliott, M. Gold, D.E. Fields, K. Rielage, R. Massarczyk, R. Gibbons
LLarge-Scale, Precision Xenon Doping of Liquid Argon
N. McFadden a , S. R. Elliott b , M. Gold a , D.E. Fields a , K. Rielage b ,R. Massarczyk b , R. Gibbons a a Department of Physics and Astronomy MSC07 4220, 1 University of NewMexico,Albuquerque NM 87131-0001 b Physics Division, Los Alamos National Laboratory MS H803, P-23, Los Alamos, NM,87545, USA
Abstract
The detection of scintillation light from liquid argon is an experimentaltechnique key to a number of current and future nuclear/particle physicsexperiments, such as neutrino physics, neutrinoless double beta decay anddark matter searches. Although the idea of adding small quantities of xenon(doping) to enhance the light yield has attracted considerable interest, thistechnique has never been demonstrated at the necessary scale. Here we reporton xenon doping in a 100 l crogenic vessel. We observed an increase in lightyield by a factor of 1.81 ± ± Keywords: neutrinoless double beta decay, xenon doping, liquid argon, Birk’s constant,Geant4
1. Introduction
Liquid noble-gas detectors (LNGD) are widely used or are planned to beused in many different particle physics applications, including dark matter(DARWIN [1]), neutrino tracking (DUNE [2]), and neutrinoless double betadecay (0 νββ ) (GERDA [3, 4], LEGEND [5]). When radiation interacts in aLNGD, it produces vacuum ultraviolet (VUV) scintillation light (Fig. 1) aswell as ionization electrons. These two energy deposition processes, combinedwith an electric field, enable TPC detectors to achieve energy resolution ofa few percent and full three-dimensional track reconstruction. As WIMPdark matter detectors, they can discriminate between nuclear recoils signals
Preprint submitted to Nuclear Instruments and Methods A June 18, 2020 a r X i v : . [ phy s i c s . i n s - d e t ] J un nd electronic recoil backgrounds with a rejection power of at least 10 [6, 7].Without an applied electric field, the scintillation light yield is enhancedwhich still allows for pulse shape discrimination to effectively discriminateagainst backgrounds [8].Although He and Ne have been considered [9], Ar and Xe monolithicLNGDs have advantages due to the large scintillation yields, longer scintil-lation wavelengths, and fast scintillation time constants. For the liquid Ar(LAr) case, the boiling temperature is near the preferred operating temper-ature of Ge detectors and Ar can provide substantial radiation shielding.These features led to the GERDA concept of operating Ge detectors barewithin a LAr environment [3]. By deploying bare Ge crystals, the mount-ing and readout infrastructure could be minimized reducing backgroundsassociated with that hardware. An added benefit is the response of LArto radiation. Since 0 νββ is a single-site energy deposit process, only back-ground can produce a coincident response between the Ge and LAr. GERDAused this anti-coincident technique to great advantage achieving the lowestbackground index of any 0 νββ experiment [4]. If the light detection of theLAr in GERDA could be significantly enhanced, it would result in improvedbackground rejection. Doping the LAr with Xe is one technique to improvethe light detection and is being considered for LEGEND-1000. However, aXe-doped LAr (XeDLAr) volume of sufficient size to demonstrate its use forLEGEND has not yet been demonstrated. It is important to demonstratethat the doping can be done at a large scale in a controlled manner, and todetermine if XeDLAr mixture is stable. Doping LAr with 0.1-1000 ppm of Xe [11], (10-1000ppm) [12, 13, 14, 15,16] shifts the wavelength of emitted scintillation light from 128 nm, whereLAr scintillates, to 175 nm where Xe scintillates. This wavelength shift isadvantageous for several reasons. Argon has a long attenuation length at175 nm [15] but only a 50-60 cm attenuation length at 128 nm [17, 16].Additionally, devices sensitive to detect both 128 nm and 175 nm are moresensitive at 175 nm. In order to detect the 128 nm light with conventionalphoto-multipliers (PMT), the light must first be wavelength shifted. Thecoupling of light detection devices to wavelength shifting materials (e.g. TPB[18]) typically occurs at the surface of the device, and therefore nearly halfthe light does not directly enter the PMT because the secondary re-emissionprocess is isotropic. In contrast, the XeDLAr wavelength shifting occurs2 igure 1: Scintillation spectrum for all liquid noble gases, excludingradon. Dotted lines represent opacity for several window materialsused to observed this light. Credit: [10] at the interaction point. Pulse shape discrimination with XeDLAr is stillpossible, though LAr is still superior. If all photons are collected, pulse shapediscrimination becomes worse as a function of Xe concentration [13, 11].If a fused silica filter is used to remove wavelengths below 160 nm, pulseshape discrimination improves with Xe concentration above 250 ppm [12]. Atconcentrations of 250 ppm Xe, a fast component emerges in the Xe spectrumthat helps recover the pulse shape discrimination.Obviously there are many advantages to a XeDLAr LNGD but in orderto accurately quantify them, XeDLAr needs to be demonstrated on a largevolume. The largest active volume previously used was 13.6 l [13], with mostexperiments using volumes of order 1 l or less. If XeDLAr is to replace pureLAr in any future experiment, Xe doping needs to be demonstrated on alarge scale. One concern is the possibility of clumping: reference [16] claimsto measure clumping at Xe concentrations of 3%, while a chemical analysisshows that Xe is 16% soluble in LAr at 87 K [19]. Another concern is thestability of the mixture, since Xe has a much higher freezing point than Ar(161.40 K versus 83.81 K).Xenon doping was performed in previous smaller experiments using oneof two methods. Either a warm argon-xenon mixture was injected into theliquid volume, or the mixture was prepared in the gaseous phase and then3ondensed into liquid. The warm gas injection technique will not work onlarge scale experiments because xenon homogeneity will take a long time toachieve, while a pre-mixed gas is not feasible due to the large gas storagevolumes required. Additionally, these doping techniques result in large total-concentration uncertainties making reported scintillation time constants dif-ficult to compare. The smallest doping concentration uncertainty reported todate is 23.5% [13]. New injection techniques need to be considering if xenondoping is to be pursued on a large scale.A large Xe doping test stand is also critical for understanding the effectsof the long attenuation length at 175 nm in XeDLAr. This long attenua-tion length can result in an increase in light yield as observed previously byRef. [13] ( × ±
80 ppm) and Ref. [14] ( × ±
547 ppm). The discrepancies in these increased light yield measure-ments can be attributed to different detectors sizes, with the larger increasebeing measured in the larger detector. Although the attenuation length wasmeasured by Ref. [16], it was predicted by Ref. [20] to be at least an order ofmagnitude longer. Additionally, the two attenuation length measurementsfor a Xe concentration of 3% made by Ref. [16], 170 ±
23 cm and 118 ±
10 cm,were not consistent with each other. Finally, some of these experiments [13]do not purify the Ar but rather rely upon the base purity of research gradeAr (99.9999%) provided by the vendor. Research grade Ar will produce lessscintillation light than pure Ar and thus measuring the increase due to Xedoping is complicated due to the presence of unknown contaminates. Giventhe uncertainties in the previous work, a novel liquid argon purification sys-tem that can precisely dope large volumes is needed before XeDLAr can beassessed for a large experiment like LEGEND.
2. Argon and Xenon Scintillation
When ionizing radiation interacts with a noble gas, dimers form which,when relaxing back to the ground state, produce scintillation light. Scin-tillation light is emitted through the transition from one of the two lowestmolecular excited states, Σ + u (triplet) or Σ + u (singlet), to the ground state Σ + g . These states cannot be easily distinguished spectroscopically but dohave different relaxation times. The Σ + u state for liquid argon has a halflife of around 4-7 ns while the Σ + u state half lives for pure liquid argon havebeen measured to be between 1.2-1.6 µ s ([21],[22],[23]). This large differencein measured Σ + u state lifetimes for “pure” argon is likely due to small con-4entrations ( ∼ . The Σ + u state forliquid xenon has a half life of 2-4 ns and the Σ + u state 21-28 ns [25]. Themain feature of the argon and xenon scintillation spectra is the Gaussian fea-ture centered at 128 nm and 175 nm for argon and xenon respectively. Thesummed probability distribution function for the Σ + u and Σ + u componentscan be written as: I ( t ) = A s τ s e − t / τ s + A t τ t e − t / τ t (1)where τ s and τ t are the time constants for the Σ + u and Σ + u state respectivelyand A s and A t are the relative intensities of each components. The integralof I ( t ) over all time is normalized to unity and therefore A s + A t = 1. When Xe is inserted into LAr, it quenches the Ar dimers through colli-sional de-excitation. Unlike other contaminates (e.g. N), this collision allowsfor effective energy transfer between the Ar dimer and Xe, allowing the Xeatom to create a dimer of its own, as:Ar ∗ + Xe + collision → (ArXe) ∗ + Ar (2a)(ArXe) ∗ + Xe + collision → Xe ∗ + Ar (2b)Here collision refers to a molecular collision between a Xe atom and a Ar ∗ or ArXe ∗ dimer. The VUV spectroscopy of this reaction as a function of Xeconcentration can be seen in Fig. 2. As the Xe concentration increases, theAr dimer energy is more efficiently converted to the Xe dimer.Xenon and Ar have very different scintillation time profiles with thebiggest difference being their Σ + u state lifetimes ( ∼ µ s versus ∼
30 ns).When Xe is injected into Ar at low concentrations, the Σ + u state of the Aris quenched by collisions, thus the time distribution can be described [12]: I ( t ) = A τ f e − t/τ f + A τ s e − t/τ s − A τ d e − t/τ d (3)where τ f and τ s are the decay times of the fast and slow components respec-tively, τ d is a time characterizing the conversion of Ar ∗ to Xe ∗ . A , A and Atmospheric argon concentrations are 0.934% by volume in Earth’s atmosphere, whileatmospheric xenon concentrations are 87 ± igure 2: The VUV emission of electron-beam excited LAr dopedwith 0.1, 1, 10, 100, 1000 ppm Xe is shown. Note that at higher Xeconcentrations, another peak at 147 nm appears [11] A are the intensities of these three terms. In order to normalize Equation 3,the following relation is defined: A + A − A = 1. As the Xe concentrationincreases, the average collision rate between a Xe atom and an Ar dimer inthe Σ + u state approaches the Ar Σ + u decay time ( ∼ I ( t ) = A τ f e − t/τ f + A τ s e − t/τ s − A τ ds e − t/τ ds − A τ df e − t/τ df (4)where τ df and τ ds characterize conversion times for the fast and slow compo-nents separately [12]. Additionally the following relation, A + A − A − A =1, is needed to normalize Equation 4. These time constants have been mea-sured by several groups and the summary of these results can be found inRef. [12]. These measurements have unresolved tension, which is likely dueto the Xe doping techniques and/or residual Xe or contaminants in the Ar.
3. Experimental Setup
Our stainless steel cryostat consists of two nested cylinders (inner (IV)and outer (OV) vessels) (Fig. 3). The IV can hold approximately 100 lof liquid. The space between IV and OV is held near vacuum to preventconvective heating. An aluminized infrared-reflective wrapping (Fig. 3, right)surrounds the inner vessel to reduce radiative heating from the warm outer6 igure 3: (left) The cryostat design. (right) A photograph the inner vessel wrapped inaluminized infrared-reflective nylon. can. The inner vessel is attached to the lid of the outer vessel with fourG-10 [26] rods.The vessel is drip cooled using an AL60 Cryomech cold head [27] provid-ing approximately 60 W of cooling power at 80 K. The cold head is mountedon top of the OV and is connected to the IV through a thin walled, 2.75-inConFlat bellows. The cold head drips condensed Ar down a thin stainlesssteel rod directly to the bottom of the inner vessel. The cold head is sur-rounded by a Teflon funnel, which is used to separate the warm and coldAr gas. As Ar evaporates, it flows up the funnel toward the cold head, re-condenses, and drips back into the liquid. Warm Ar gas is inserted abovethe cold head and falls as it cools. The cold head is instrumented with a50 W heater and a thermocouple, which maintains the cold head at LArtemperature.The inner vessel was filled from the ullage of a research grade LAr cylinder(99.999% pure). Both inner and outer vessels were leak checked prior to eachfill. No thorough vessel pump and bake cycling was found necessary as aresult of our re-circulation system described below. Gas was flowed througha SAES PS4-MT3/15-R getter, which can purify noble gasses to less than1 ppb of all common contaminates, but not noble gases [28]. The quantityof Ar in the inner vessel is measured using a shipping scale accurate to ± The gas doping system consists of a small measured cylindrical volume,19 ± ∼ The inner vessel is instrumented with two 12-stage Hamamatsu, 3-inPMTs (R11065) [29] which are box&linear-focused, with Synthetic Silica3-in windows and Bialkali photo-cathodes designed to operate at LAr tem-peratures. The PMT windows have a mounted tetraphenyl butadiene (TPB)coated acrylic disk to wavelength shift the Ar scintillation light into the sen-sitive region of the PMTs. The average quantum efficiency over the TPBemission spectrum is ∼ ? ]. TPB is vapor deposited to a thickness of about 2 µ m. Toprevent the PMTs from floating in the LAr, the TPB disks are tied down tothe Al holders with stainless steel wire.After completing an initial test run, it was found that the TPB on theacrylic disks had dislodged and redeposited on the wall of the cryostat.Though TPB has been shown to be stable in liquid argon [31], it is spec-ulated that the volatile boundary between the gas and liquid phases of theAr can be abrasive to the vapor deposited TPB. It is not known what effect8 igure 4: Piping diagram for the gas insertions. this has on the light yield. To ensure maximum light for each successive run,new TPB coated disks are used.The PMT cathode ground scheme is chosen to allow signals and highvoltage to share a cable, reducing the heat load from the wires. Each PMTis instrumented with a fiber optic cable and a blue LED is used for lowlight calibration of the PMTs. The conflat feed-throughs are mounted onthe warm side of the cryostat to ensure thermal cycling did not damagethe ceramic/steel interface. The PMT signals are amplified using a PhillipsScientific Octal Variable gain amplifier Nim Model 777 [32]. These signalsare digitized using a Tetronix TDS 3052B digital oscilloscope and read outto a laptop using custom Python software. The scope has a slow read-outrate of ∼ µ s. A pulse finding method was adapted from Ref. [33], which uses a smoothedderivative method to identify signal pulses as large (3.5 σ ) fluctuations in thenoise. Four threshold crossings of the derivative define a pulse. An example9 igure 5: A single photon pulse found using the pulse finding algorithm [33]. The blue lineis the background noise while red line is the highlighted found pulse. The thick dashedgreen line is the derivative of the blue plus red lines (signal plus background). The twoyellow lines are the 3.5 σ rms derivative thresholds for the crossings, with two crossings perthreshold required for a pulse. of this pulse finding can be seen in Fig. 5. Local minima are identified usingthe derivative allowing individual pulses to be distinguished in the event ofpulse pileup. This feature proved essential for our measurement of the tripletlife of Ar. Prior to measuring the triplet lifetime for pure argon, a distribution ofthe number of dark pulses is generated in order to evaluate the single photonresponse of the PMT. A double Gaussian distribution is fit to the noise peakand single photon peak. The Gaussian mean and sigma of the single photo-electron (SPE) peak characterize the single photon response of the PMT.From this fit, a gain of (4 . ± . × is estimated after accounting forthe × The Σ + u lifetime can be measured by fitting equation 1 to a histogramof pulse arrival times weighted by their integrated charge. Pulse arrival timehistograms were created for 40 runs containing 5,000 events each and triplet10 igure 6: Pulse charge (V · s) versus pulse height (V) distribution. The two highly pop-ulated regions (yellow) show where the SPE and noise pulses typically occur. The SPEregion has a mean charge of about 0 . × − V · s and a mean peak height of 0.017 V. lifetime was measured for each run. After-pulsing was noticed in the databetween 1 µ s and 2.5 µ s. While effects of after-pulsing in PMTs are wellcharacterized [34], it cannot be reduced without decreasing the single photonresolution. Therefore, we investigated mitigating the after-pulsing effect onthe LAr triplet lifetime using two fitting methods. In the first method, thetriplet lifetime was found using a single exponential plus a constant fit. Astart time of 2.5 µ s was chosen in order to avoid the after-pulsing. A tripletlifetime of 1456 ±
38 ns was measured with this method. Alternatively, we fita Landau shape to the first after-pulse and a Gaussian shape to the secondafter-pulse. These functions were added to the first function and the sum wasfit starting from 1.75 µ s. This method yields a systematically larger value ofthe triplet lifetime of 1502 ± ± ± igure 7: Two fits to the triplet scintillation distribution in pure LAr. The top plot isthe fit with a single exponential and a constant from 2.5 µ s to 10 µ s and gives a value of1456 ±
38 ns. The bottom plot shows the fit with a Landau shape for the first after-pulseand a Gaussian shape for the second after-pulse with an exponential and a constant fitthe tail. The fit to the after-pulsing region is shown in the insert. This second fit givesa value of 1502 ± ± ±
4. Nitrogen Doping
In order to validate the gas doping and recirculation system, a small mea-sured quantity of N was injected using the methods described in Section 3.12ollowing the injection the N was filtered out using the getter. Past Xe dop-ing experiments [14] had large uncertainties ( ∼ τ (cid:48) ) and the true triplet life time ( τ ) follows Birk’s law: τ (cid:48) = τ τ · k · C (5)where k is Birk’s constant and C is the molar concentration of impuritiesmeasured in parts per million. Birk’s constant has been measured to be0.11 ppm − µ s − and 0 .
13 ppm − µ s − for N in LAr by the two previouslymentioned authors. They obtained τ = 1.26 µs compared to our value of 1.48 µs . Reference [23] was aware that their τ was smaller than that typicallymeasured ([22], 1.46 µ s) and speculated that if they used the fitting methodof Ref. [22] and extrapolated their τ using Eqn. 5 to pure LAr, their valuebecomes τ = 1 . µ s, which is in good agreement with our measurement.Before doping, the recirculation and purification system was run and τ measured frequently. Once the τ measurement was stable, it was assumedthat the Ar was pure. The nitrogen contaminant was then inserted and τ decreased according Eqn. 5. As the Ar was purified, it is assumed that thecontamination concentration decreased as: C ( t ) = C · e − t / τ filter (6)where C is the initial concentration measured in molar parts per million and τ filter is a fitted time constant characterizing the filtration process.We performed one N doping run injecting 1.06 ± . Thetriplet lifetime dropped from 1.479 µs to 1.239 µs as seen in Fig. 8. Fromthis reduced lifetime a Birk’s constant of 0.12 ± − µ s − was found,consistent with the previous measurements. Using this measured value ofBirk’s constant, Eqn. 5 was fit to the triplet purification curve. Initial mea-surements of the filtration time constant found it to be 9.5 ± . × − Torr to 1 . × − Torr.Additionally the cold-head set point was raised from -190 C to -187 C thusincreasing the internal pressure of the cryostat from 550 Torr to 640 Torr. Amodification to Eqn. 5 was made to account the for non-zero start time to13 igure 8: Triplet lifetime ( µ s) versus run time (days) before and after the nitrogen injec-tions. It can be seen that the nitrogen thoroughly mixed after about three hours. obtain: τ (cid:48) = τ (cid:48) τ (cid:48) · k · C · e − ( t − t ) / τ filter (7)where t was the non zero start time of the fit. The filtration time constantmeasured by fitting Eqn. 7 to the triplet purification curve after these changeswas measured to be 4.5 ± ± ± ± ± ± ± ± ± Co source.
5. Xenon Doping
Having validated our doping procedure with nitrogen, we then used xenon.Xe doping was done in four steps of 1.00 ± ± ± ± igure 9: Nitrogen doping in pure LAr. The left fitted red line shows the fit obtainedwhen the heat leak for the system was minimized, yielding a filtration rate of 9.5 ± ± sired molar concentration was reached. Each 0.5 ppm injection took about30 min. High concentrations of Xe ( ∼
100 ppm) were not considered. Oncethe concentration was stable and fully mixed, 50,000 random backgroundevents were collected at each concentration to create the scintillation distri-butions (waveforms) seen in Fig. 10. The waveforms were fit from 1.05 µ s to10 µ s to use as much of the waveform as possible while avoiding the singletcontribution. The start time of the fit was varied by ±
25 ns in order toestimate the systematic uncertainty associated with the fit window.A modified version of Eqn. 3 was used to measure the time constants ofthe mixture: I ( t ) = A τ s e − t/τ s − A τ d e − t/τ d + A τ long e − t/τ long (8)where τ s is the slow component, τ d is the energy transfer time between theXe and triplet state of the Ar, A , A and A are the intensities, and τ long isan additional time constant to account for the exponential tail at late times.This artifact is consistent with light coming from the cold gas in the ullage[36].The slow component, τ s , represents the time constant for the combinationof unquenched Ar dimer light and Xe dimer light created through the mech-anism outline in Section 2.1. Though Xe decays rather quickly comparedto Ar, the population of Xe dimers is pumped by the Ar-Xe mixed state,15 igure 10: Scintillation distributions created by binning pulse arrival times weighted by thepulse charge for various concentrations of Xe from 0–10 ppm. Distributions are normalizedand then scaled to have the same maximum value. thus is able to have a much longer decay time compared to pure liquid Xe.The total concentration of Xe dimers and Ar dimers in the triplet state is A e − t/τ s − A e − t/τ d . The fitted waveforms are seen in Fig. 11 and the resultsfrom these fits are summarized in Table 1.Previous measurements [12, 14] of τ s for 1 ppm Xe reported values be-tween 1 µ s and 2.5 µ s, though the uncertainty on the measured concentrationis also 1 ppm. At 10 ppm Xe, τ s has been measured as 280 ns and 750 ns,with the concentration uncertainty at least 3 ppm. τ d has not been measuredfor 1 ppm Xe, but for 10 ppm the value ranges from about 250 ns to about700 ns. Past measurements of these time constants have a much larger Xe-concentration uncertainty making them difficult to compare with our mea-surements. Disagreements between the previous measurements is likely duethe concentration uncertainty and the unknown initial xenon concentrationin what is assumed to be pure liquid argon.
6. Optical Simulations
The software package Geant4 [37] was used to simulate the light collec-tion efficiencies of scintillation events in our apparatus. In order to accuratelymodel optical physics in Geant4, it is required to input the optical properties,including surface reflectivity, attenuation length, scintillation yield, absorp-16a) 1 ppm (b) 2 ppm(c) 5 ppm (d) 10 ppm
Figure 11: Fit of Eqn. 8 for Xe doped waveforms with Xe concentrations of 1, 2, 5, and10 ppm. The solid red line is the total fit, the dashed blue line is the long exponentialcomponent associated with τ l , and the magenta dashed line is the Xe dimer and Ar tripletcomponent ( A e − t/τ s − A e − t/τ d ). Deviations from the fit at short times (1.5-2.5 µ s) aredue after-pulsing. Xenon Doping Time ConstantsConcentration τ s (ns) τ d (ns) τ long (ns)1 ppm 1243 ± ± ± ± ± ± ± ± ± ± ± ±
55 ppm 503 ± ± ± ± ± ± ± ± ± ± ± ± Table 1: Summary of Xe doping time constants, measured in ns, for various molar concen-trations from 1 to 10 ppm Xe. The first uncertainty is a systematic uncertainty estimatedby varying the fit start time from 1.025 µ s to 1.075 µ s and the second uncertainty isstatistical. Simulated light yield for liquid argon detectors requires running radioac-tive decay simulation (i.e. K , Uranium, or Thorium) and optical sim-ulations simultaneously. This requires an extensive amount of computingresources due to the many discrete steps an optical photon can take. Forefficiency, the simulations are broken into two separate steps. Since scintil-lators emit light isotropically and have a linear yield over a wide range ofenergies, an optical “heat” map can be created mapping the photon detec-tion probability as a function of initial position throughout the scintillatorvolume. Once these maps are created, they provide the average light yieldas a function of position without having to run an optical simulation in con-junction with the radioactive decay. Though generating an optical map isquite computationally intensive, it only needs to be done once.The average photon yield, N , is calculated as: N = Y scint × E dep ( r ) × P detection ( r ) × (cid:15) detector (9)where Y scint is the scintillator light yield, E dep ( r ) is the energy deposition inthe scintillator as a function of position, P detection ( r ) is the detection prob-ability as a function of position, and (cid:15) detector is the light detector efficiency.The average yield is calculated per step for each particle track and summedat the end of the event in order to estimate the average light yield of theevent.Liquid argon optical heat maps are created by generating VUV photonswith a mean wavelength of 128 nm and Gaussian sigma of 2.93 nm and prop-agating them through the volume. Photon detection probability is calculatedper (5 mm) voxel as the number of photons detected divided by the totalnumber of VUV photons generated in that voxel. Random vertices in theliquid argon are chosen and the photon direction is generated isotropically. The optical volumes include gaseous Ar, LAr, stainless steel, TPB, acrylicand the PMT windows. 10 ppm of Xe was also simulated. The PMTs aremodeled as stainless steel tubes. The 3-in cathode window, embedded onthe top, is simulated as the sensitive optical element with known quantum18fficiency. An ⁄ -in thick acrylic disk covers the cathode, while a 2- µ m thickTPB disk covers the acrylic. Figure 12: (top left) A Z-X projection of the pure Ar optical map. (top right)A Y-X pro-jection of the Ar optical map. (bottom left) A Z-X projection of the XeDLAr optical map.(bottom right)A Y-X projection of the XeDLAr optical map.The color indicates detectionprobability. Note the counting well centered at Y = 0 and the liquid/gas boundary at Z= +180.0 mm. The PMT is centered at (X = 0, Y = 125 mm, Z = -100 mm).
The generated LAr and XeDLAr optical maps can be seen in Fig. 12.In order to project this three dimensional map into two dimensions, thethird dimension is averaged. The XeDLAr attenuation length was extendedfrom 60 cm, appropriate to pure argom, to 1000 m, a value much larger19han the cyrostat dimension, since the XeDLAr attenuation length has notbeen measured at these concentrations.The scintillation central wavelengthwas shifted from 128 nm to 175 nm and the sigma has been shifted from2.93 nm to 8.60 nm. The low vapor pressure of Xe in cold gaseous Arresults in a neglibible concentration of Xe in the cold gas of the ullage(+180 mm < Z < +300 mm). We note that XeDLAr allows for improvedlight collection farther from the PMTs, which can be attributed to the in-creased attenuation length. We used cosmic ray muons to calibrate the optical maps. Cosmic muonsare triggered on by placing two 0 . × . × .
50 in scintillating panels aboveand below the cryostat rotated so that their long axes are perpendicular. Thisdouble coincidence requirement selects muons that are preferentially verticalthus limiting the variance in the traversal of the Ar (Fig. 13). Addition-ally, muons are minimum ionizing particles with a roughly constant dE/dxof about 2.11 MeV/cm for a 266 MeV muon [39]. This energy is depositeduniformly along the muon path, thus allowing for the effects of longer atten-uation lengths to be explored. The long path length also mitigates low lightcollection and non-uniform light response.Muon kinetic energies were generated according to a power law, withdownward momentum chosen from a cos ( θ ) distribution and uniform hori-zontal momentum. The predicted photo-electron (PE) spectrum for Ar andXeDLAr can be seen in Fig. 14. These spectra have been convolved with thePMT single photon resolution ( σ SP E = 0.3125 PE). Both spectra were fit toLandau functions and have a most probable value (MPV) at 4235 ± ±
24 PE for XeDLAr and pure LAr respectively, predicting an in-creased light yield of 1.98 ± (cid:15) total ∼ e − x/λ atten · (cid:15) T P B · (cid:15) P MT · Y scint (10)where x is the distance of the interaction in the liquid argon to the detector, λ atten is the LAr attenuation length, (cid:15) T P B is the TPB quantum efficiency,20 igure 13: (left) A ray tracing drawing of the geometry implemented in Geant4. Theminimum and maximum muon path lengths are indicated by green dotted lines. (right)The z direction of muons that pass through both panels. The blue histogram is themomentum for all simulated muons, while the red histogram is for muons that passed thedouble coincidence cut.Figure 14: Simulated muon spectrum after the double muon panel coincidence is appliedfor pure LAr (left plot in red) and XeDLAr (right plot in blue). The red line in bothplots is a Landau function with a MPV of 2138 and 4269 PE for the pure liquid argon andxenon doped liquid argon respectively. (cid:15)
P MT is the PMT quantum efficiency, and Y scint is the scintillator yield.The attenuation length in XeDLAr is not known at these concentrations.In order to understand the effect that the attenuation length has on theshape of the distribution, several XeDLAr optical maps were made withvarying attenuation lengths. The shape of the distribution is defined as the“sigma” of the Landau distribution divided by the MPV, where sigma is thescale parameter reported by the CERN ROOT Landau fit. The shape ofthe distribution (sigma/MPV) is measured to be 0 . ± . ±
14 898 ± ± ±
17 1120 ±
12 0.305 ± ±
18 1169 ±
12 0.299 ± ±
19 1201 ±
13 0.293 ± ±
19 1229 ±
13 0.294 ± ±
19 1248 ±
13 0.303 ± ±
20 1291 ±
14 0.302 ± Table 2: Summary of muon peak shape versus attenuation length obtained from thesimulation. Each muon peak was fit with a Landau function and the MPV and sigmawere obtained. The errors are the statistical errors from the fits. change. We use the shape parameter to give a lower limit on the attenuationlength in XeDLAr from the data in Section 7 below.
7. Cosmic Ray Muon Data
We used two scintillator panels, placed above and below our cryostat,to collect muon cosmic ray data. The double-coincident triggered data in10 ppm XeDLAr is shown in Fig. 15 (left). Our trigger coincidence alloweda significant number of accidental events to be collected as well. Data fromseparately collected accidental triggers was subtracted from our spectrum toobtain corrected spectrum Fig. 15 (right). The data was fit to a Landaudistribution with a most probable value (MPV) of 604.8 ± . ± . . ±
18 PE, which is 6 . ± .
08 timeslarger than what is measured in 10 ppm XeDLAr. When the simulated muonspectrum is scaled by the ratio of the measured and simulated MPVs ( × igure 15: (left) The blue spectrum is from cosmic muons in double coincidence in 10 ppmXeDLAr (signal); the red spectrum is from accidental double coincidence events. (right)The subtracted spectrum in 10 ppm XeDLAr. The fit is to a Landau function with a MPVof 604.8 ± ± σ uncertainty isthe red shaded band. Below an attenuation length of 3.0 m, the simulation is inconsistentwith the data. This determines our lower limit on the attenuation length of 175 nm lightin 10 ppm XeDLAr. In the simulation of the cosmic ray muons we have used a value of5.0 m. igure 17: (Top Left) The red spectrum is cosmic muons in double coincidence in 10 ppmXeDLAr.(Top right) The blue spectrum predicted from simulation and scaled by 6.47.The red curves in both top plots are Landau fits with MPVs of 604.8 ± ± and fit with a Landau function, then a MPV of 595 ± ± Second, the nominal value of 40 photons/keV light yield, maynot be a valid. The authors of Ref. [25] state that they measure a yield ofabout 20 photons/keV even though they have pure Ar. The absolute lightyield of XeDLAr remains to be measured. Third, the quantum efficiency ofthe PMTs may have been reduced after contamination with He. Though itis hard to quantify how much these effects are reducing the light yield, byselecting the correct attenuation length and accounting for this factor of 6.47,the simulation reproduces the data. These results validate the optical heatmap technique described in Section 6.
8. Light Yield Comparison
The total increase in light yield resulting from Xe doping was measuredby a direct comparison of the muon peak seen in pure Ar data to the 10 ppmXeDLAr data.
The cosmic muon peak should be visible in the pure LAr random triggerspectrum, even without the double coincidence requirement. Although thepeak is dominated by cosmic muons, there is a contribution from unknownradioactive backgrounds. Additionally, cosmic muons without the scintilla-tor trigger have different paths than with the trigger. Figure 18 shows therandom trigger spectrum in pure (un-doped)liquid (red histogram) and thesimulated double coincident muon spectrum (blue markers). The simulatedspectrum has been scaled by the factor × ± ± ± ± ± ± The cryostat remains sealed for several more months due to COVID-19. After opening,we will be able to verify this assumption. igure 18: A comparison of simulated and measured muon peaks in pure (undoped) LAr.The red histogram is data, while the blue markers are simulation. The two curves areLandau fits with the red curve fit to data and the blue curve fit to the simulation. Thedata was taken using the random trigger which accounts for the difference in shape withthe simulation (see text). The same scale factor ( × ± ±
9. Nitrogen Quenching in Xenon Doped Argon
We expect that contaminants will quench the total light output less inXeDLAr than in pure LAr. The total scintillation distribution is shiftedto shorter times, and so one expects that the probability of colliding withcontaminants (via collisional damping) is less likely. To evaluate this conjec-ture, the shift of the muon peak from a double coincidence was measured.This quenching effect cannot be modeled using Birk’s law because the en-ergy transfer between the Xe and Ar dimer creates a mixed state that does26enon Doping Time ConstantsN Concentration T s (ns) T d (ns) T long (ns)0 ppm 447 ± ± ± ± ± ±
141 ppm 439 ± ± ± ± ± ±
210 ppm 415 ± ± ± ±
12 2321 ± ± Table 3: Summary of xenon doping time constants (ns) for 10 ppm XeDLAr and variousconcentrations of N . not emit light but still can be quenched. This energy transfer process isalso quenched, thus a comparison of quenching constants between pure LArand XeDLAr cannot be made. We considered two molar concentrations ofnitrogen: 1 . ± .
06 ppm and 9 . ± . . These nitrogen concen-trations were chosen because commercially available research grade LAr canbe obtained at 99.999% (10 ppm) or 99.9999% (1 ppm) purity.The N contaminated XeDLAr scintillation time constants were measuredand summarized in Table 3. The scintillation distributions for all three N concentrations (0 ppm, 1 ppm, and 10 ppm) can be seen in Fig. 19. All threedistributions are normalized and then scaled by the maximum value of the0 ppm N measurement. It can be seen that at 10 ppm of N , significant after-pulsing is distorting the scintillation time distribution. The total light yieldin 10 ppm of N is much lower (see below) thus after-pulsing contributessignificantly more, especially around 1.5 µ s, to the total light seen in thedistribution.At 1 ppm of N and 10 ppm of Xe in LAr, the muon peak from a doublecoincidence trigger was found to shift from 604.8 ± ± ± ± and 10 ppm XeDLAr the muonpeak shifts to 294 ± ± is injected [23].These doping measurements suggest that XeDLAr is less sensitive to thetotal contamination concentration of N . Unlike N , Xe will quench theAr triplet in some characteristic time T d , and will then re-emit light at alonger wavelength. This Xe state is less likely to be quenched because, at10 ppm Xe, T d is measured to be 213 ± and10 ppm Xe, has nearly the same total light yield from cosmic muons as pureLAr (294 PE versus 338 PE). Thus, if one were to buy five 9’s LAr from a27 igure 19: Time distribution intensities for 10 ppm of Xe and 0 (blue), 1 (red), and10 ppm N (green) in LAr. The area of the blue curve is fixed to 1, the red curve contains1 ± .
04 ppm N and has an area of 0.908, and the green curve contains 9 . ± .
13 N and has an area of 0.770. commercial vendor and dope it with 10 ppm of Xe, the total light outputwould be similar to pure LAr. It is also worth noting that five 9’s commercialgrade Ar has other contaminates like oxygen, air, Co and CO , all of whichhave different quenching factors. We speculate that it still might be possibleto see less quenching at higher Xe concentrations because T d will be shorter.
10. Conclusion
It has been demonstrated that a small concentration of gas can be reliablyand precisely injected into a cryostat containing ∼
120 kg of LAr. Nitrogenat a concentration of 1.06 ± ± − µ s − , consistentwith past literature. This demonstrates that small quantities of gas can beprecisely injected. The N contaminant was then removed using a ullagerecirculation and purification system with a filtration time constant as lowas 3.6 d.We investigated small XeDLAr concentrations between 1 and 10 ppm.It was shown that up to a factor of ∼ . heat maps as an alternative tophoton tracking. These optical heat maps contain the probability of photondetection as a function of position inside the simulated detector. Using thesemaps, a simulated double coincidence muon spectrum was created for pureLAr and XeDLAr. The data was consistently modeled once the attenuationlength was determined and an overall light yield correction factor (6.47)was applied. This demonstrates that the optical heat map technique canaccurately model the detector response. Acknowledgements
We thank Stefan Sch¨onert for useful conversations, Steven Boyd for his in-finite wisdom concerning our cryogenics system, Dinesh Loomba for his helpinterpreting our data, and Matt Green and Brain Zhu for their expertise inGeant4. In view of the optical simulations, we acknowledge the developmentof the computationally efficient ”heat-map” approach by the LEGEND sim-ulations group (chaired by Matt Green) and the fruitful discussions we hadwith Luigi Pertoldi and Christoph Wiesinger from the GERDA collabora-tion. We would like to thank the Department of Physics and Astronomyat the University of New Mexico for the partial support of one of our au-thors (N.M.) and allowing us to move the experiment exceptionally early intothe new physics building. This material is based upon work supported bythe U.S. Department of Energy, Office of Science, Office of Nuclear Physicsunder award number LANLE9BW. We gratefully acknowledge the supportof the U.S. Department of Energy through the LANL/LDRD Program forthis work. Finally, we would like to acknowledge the U.S. Department ofEnergy, Office of Science, Office of Workforce Development for Teachers andScientists, Office of Science Graduate Student Research (SCGSR) program.The SCGSR program is administered by the Oak Ridge Institute for Scienceand Education (ORISE) for the DOE. ORISE is managed by ORAU undercontract number DESC0014664. 29 eferences [1] J. Aalbers et al. DARWIN: towards the ultimate dark matter detector.
JCAP , 1611:017, 2016.[2] B. Abi et al. The DUNE Far Detector Interim Design Report Volume1: Physics, Technology and Strategies. arXiv , 2018. 1807.10334.[3] M. Agostini et al. LArGe: active background suppression using ar-gon scintillation for the Gerda 0 νββ -experiment.
Eur. Phys. J. ,C75(10):506, 2015.[4] M. Agostini et al. Improved limit on neutrinoless double beta decay of Ge from GERDA Phase II.
Phys. Rev. Lett. , 120:132503, 2018.[5] N. Abgrall et al. The Large Enriched Germanium Experiment forNeutrinoless Double Beta Decay (LEGEND).
AIP Conf. Proc. ,1894(1):020027, 2017.[6] J. Calvo et al. Backgrounds and pulse shape discrimination in the ArDMliquid argon TPC.
JCAP , 1812(12):011, 2018.[7] Andrew Hime. The MiniCLEAN Dark Matter Experiment. In
Particlesand fields. Proceedings, Meeting of the Division of the American PhysicalSociety, DPF 2011, Providence, USA, August 9-13, 2011 , 2011.[8] R. Ajaj et al. Search for dark matter with a 231-day exposure of liquidargon using DEAP-3600 at SNOLAB.
Phys. Rev. , D100(2):022004, 2019.[9] V. Chepel and H. Araujo. Liquid noble gas detectors for low energyparticle physics.
JINST , 8:R04001, 2013.[10] V. M. Gehman, S. R. Seibert, K. Rielage, A. Hime, Y. Sun, D. M.Mei, J. Maassen, and D. Moore. Fluorescence Efficiency and VisibleRe-emission Spectrum of Tetraphenyl Butadiene Films at Extreme Ul-traviolet Wavelengths.
Nucl. Instrum. Meth. , A654:116–121, 2011.[11] A. Neumeier, T. Dandl, T. Heindl, A. Himpsl, L. Oberauer, W. Potzel,S. Roth, S. Schnert, J. Wieser, and A. Ulrich. Intense vacuum ultravioletand infrared scintillation of liquid ar-xe mixtures.
EPL (EurophysicsLetters) , 109(1):12001, jan 2015.3012] D. Akimov, V. Belov, A. Konovalov, A. Kumpan, O. Razuvaeva,D. Rudik, and G. Simakov. Fast component re-emission in Xe-dopedliquid argon. arXiv , 2019. 1906.00836.[13] P Peiffer, T Pollmann, S Schnert, A Smolnikov, and S Vasiliev. Pulseshape analysis of scintillation signals from pure and xenon-doped liquidargon for radioactive background identification.
Journal of Instrumen-tation , 3(08):P08007–P08007, August 2008.[14] C G Wahl, E P Bernard, W H Lippincott, J A Nikkel, Y Shin, andD N McKinsey. Pulse-shape discrimination and energy resolution of aliquid-argon scintillator with xenon doping.
Journal of Instrumentation ,9(06):P06013–P06013, June 2014.[15] A. Neumeier, T. Dandl, A. Himpsl, L. Oberauer, W. Potzel, S. Schn-ert, and A. Ulrich. Attenuation of vacuum ultraviolet light in pureand xenon-doped liquid argon —an approach to an assignment of thenear-infrared emission from the mixture.
EPL (Europhysics Letters) ,111(1):12001, July 2015.[16] N. Ishida, M. Chen, T. Doke, K. Hasuike, A. Hitachi, M. Gaudreau,M. Kase, Y. Kawada, J. Kikuchi, T. Komiyama, K. Kuwahara, K. Ma-suda, H. Okada, Y.H. Qu, M. Suzuki, and T. Takahashi. Attenuationlength measurements of scintillation light in liquid rare gases and theirmixtures using an improved reflection suppresser.
Nuclear Instrumentsand Methods in Physics Research Section A: Accelerators, Spectrome-ters, Detectors and Associated Equipment , 384(2):380 – 386, 1997.[17] A. Neumeier, T. Dandl, A. Himpsl, M. Hofmann, L. Oberauer,W. Potzel, S. Schnert, and A. Ulrich. Attenuation measurements ofvacuum ultraviolet light in liquid argon revisited.
Nuclear Instrumentsand Methods in Physics Research Section A: Accelerators, Spectrome-ters, Detectors and Associated Equipment , 800:7081, Nov 2015.[18] V. M. Gehman, S. R. Seibert, K. Rielage, A. Hime, Y. Sun, D. M.Mei, J. Maassen, and D. Moore. Fluorescence Efficiency and VisibleRe-emission Spectrum of Tetraphenyl Butadiene Films at Extreme Ul-traviolet Wavelengths.
Nucl. Instrum. Meth. , A654:116–121, 2011.3119] A. Buzulutskov. Photon emission and atomic collision processes in two-phase argon doped with xenon and nitrogen. arXiv , 117(3):39002, Feb2017. 1702.03612.[20] A. Neumeier, Hofmann M., Oberauer L., Potzel W., S. Schnert,T. Dandl, T. Heindl, A. Ulrich, and Wieser J. Attenuation of vacuumultraviolet light in liquid argon.
EPL (Europhysics Letters) , 72(10):2190,Oct 2012.[21] Akira Hitachi, Tan Takahashi, Nobutaka Funayama, Kimiaki Masuda,Jun Kikuchi, and Tadayoshi Doke. Effect of ionization density on thetime dependence of luminescence from liquid argon and xenon.
Phys.Rev. B , 27:5279–5285, May 1983.[22] W. H. Lippincott, K. J. Coakley, D. Gastler, A. Hime, E. Kearns, D. N.McKinsey, J. A. Nikkel, and L. C. Stonehill. Scintillation time depen-dence and pulse shape discrimination in liquid argon.
Phys. Rev. C ,78:035801, Sep 2008.[23] R. Acciarri et al. Effects of Nitrogen contamination in liquid Argon.
JINST , 5:P06003, 2010.[24] Shuen-Cheng; Robert D. Lein; Daniel A. Morgan Hwang.
NobleGases,Kirk-Othmer Encyclopedia of Chemical Technology (5th ed.) . Wi-ley, 2005.[25] D. S. Akerib et al. Liquid xenon scintillation measurements and pulseshape discrimination in the LUX dark matter detector.
Phys. Rev. ,D97(11):112002, 2018.[26] Laminated Plastics. G10 technical data sheet. Material Reference, June2010.[27] Cryomech. Al60 with cp820 cryorefrigerator specification sheet. MaterialReference, December 2007.[28] Saes Getter. Ps4-mt3/15-r/n specifications specification sheet. MaterialReference, May 2002.[29] Hamamatsu Photonics. R11065 preliminary data sheet. Material Ref-erence, May 2009. 3230] R. Acciarri et al. Demonstration and Comparison of Operation of Pho-tomultiplier Tubes at Liquid Argon Temperature.
JINST , 7:P01016,2012.[31] K. H. Ackermann et al. The GERDA experiment for the search of 0 νββ decay in Ge.
Eur. Phys. J. , C73(3):2330, 2013.[32] Phillips Scientific. Octal variable gain amplifier nim model 777 datasheet. Material Reference, October 2019.[33] P. ugec et al. Pulse processing routines for neutron time-of-flight data.
Nucl. Instrum. Meth. , A812:134–144, 2016.[34] A. Butcher, L. Doria, J. Monroe, F. Retire, B. Smith, and J. Walding.A method for characterizing after-pulsing and dark noise of PMTs andSiPMs.
Nucl. Instrum. Meth. , A875:87–91, 2017.[35] S. Himi, T. Takahashi, J. z. Ruan, and S. Kubota. LIQUID AND SOLIDARGON, AND NITROGEN DOPED LIQUID AND SOLID ARGONSCINTILLATORS.
Nucl. Instrum. Meth. , 203:153–157, 1982.[36] Michael Akashi-Ronquest et al. Triplet Lifetime in Gaseous Argon.
Eur.Phys. J. , A55(10):176, 2019.[37] M. G. Pia, T. Basaglia, Z. W. Bell, and P. V. Dressendorfer. Geant4 inScientific Literature. arXiv e-prints , page arXiv:0912.0360, Dec 2009.[38] Erik Dietz-Laursonn. Peculiarities in the Simulation of Optical Physicswith Geant4. arXiv , 2016. 1612.05162.[39] et. al. M. Tanabashi. Review of particle physics.
Phys. Rev. D ,98:030001, Aug 2018.[40] Mellissa Boswell et al.
MaGe - a geant4 -based monte carlo applicationframework for low-background germanium experiments.