Liquid argon scintillation response to electronic recoils between 2.8–1275 keV in a highlight yield single-phase detector

M.Kimura,∗ K.Aoyama, M.Tanaka, and K.Yorita†

Waseda University, 3-4-1, Okubo, Shinjuku, Tokyo, 169-8555, Japan(Dated: November 23, 2020)

We measure the liquid argon scintillation response to electronic recoils in the energy range of 2.82to 1274.6 keV at null electric field. The single-phase detector with a large optical coverage used inthis measurement yields 12.8± 0.3 (11.2± 0.3) photoelectron/keV for 511.0-keV γ-ray events basedon a photomultiplier tube single photoelectron response modeling with a Gaussian plus an additionalexponential term (with only a Gaussian term). It is exposed to a variety of calibration sources suchas 22Na and 241Am γ-ray emitters, and a 252Cf fast neutron emitter that induces quasimonoenergeticγ rays through a (n, n′γ) reaction with 19F in polytetrafluoroethylene. In addition, the high lightdetection efficiency of the detector enables identification of the 2.82-keV peak of 37Ar, a cosmogenicisotope in atmospheric argon. The observed light yield and energy resolution of the detector areobtained by the full-absorption peaks. We find up to approximately 25% shift in the scintillationyield across the energy range and 3% of the energy resolution for the 511.0-keV line. The Thomas-Imel box model with its constant parameter ς = 0.033+0.012−0.008 is found to explain the result. Forliquid argon, this is the first measurement on the energy-dependent scintillation yield down to a fewkeV at null field and provides essential inputs for tuning the argon response model to be used forphysics experiments.

I. INTRODUCTION

A liquid argon (LAr) scintillation detector has severalfeatures that make it attractive for use in various physicsexperiments to detect ionization particles: it has efficientconversion of energy deposition into a scintillation lightsignal, powerful discrimination between electronic recoil(ER) and nuclear recoil (NR) events based on its scintil-lation pulse shape, and benefits from the fact that largequantities of argon are cheaply available. One promisingapplication of the detector is to search and identify theNR signal possibly induced by a dark matter candidate,weakly interacting massive particles (WIMPs) [1, 2]. Thetypical energy of the signal is in the range of a few keVto several hundreds of keV. Burdensome backgrounds inthis search are ER events caused by β rays from diffusedisotopes (such as 39Ar and 85Kr) in LAr and γ rays fromradioimpurities in detector components. Predicting themeasured signal from these background sources is neces-sary to estimate its contamination in the signal region ofinterest. In this context, characterization of the detec-tor response to ER events is crucial for achieving lowerenergy threshold, suppressing systematic uncertainty re-lated to background contamination, and hence enhancingphysics sensitivity of the search. Furthermore, recentlythe searches for new particles, such as bosonic dark mat-ter and axion-like particle, have been actively performedusing the ER events by xenon (e.g. [3–5]), where its scin-tillation response is well understood [6–8], while the onefor argon is not fully established yet. Therefore this workis essentially important for physics interpretation to ex-

∗ masato@kylab.sci.waseda.ac.jp† kohei.yorita@waseda.jp

tract physics quantity from observed scintillation signalwith LAr.

In the LAr detector, a charged particle interaction ex-cites and ionizes the detector medium, resulting in theformation of self-trapped exciton states, Ar∗2, throughthe collision and recombination processes. The excimer isformed in either a singlet or a triplet state, both of whichdecay radiatively with vast different lifetimes of approxi-mately 7 ns and 1.6 µs, respectively [9]. The scintillationlight spectra from both radiative decays lie in the vacuumultraviolet (VUV), peaked at 128 nm [10]. As direct de-tection of the VUV photon at LAr temperature (around87 K) is technically challenging, it is often downshiftedto the visible region where most cryogenic photosensorsexhibit peak sensitivity using a wavelength shifter suchas 1,1,4,4-tetraphenyl-1,3-butadiene (TPB) [11, 12]. Therecoiled particle and its energy are inferred from the ob-served photon signal waveform.

In this work, we measure the LAr scintillation responseto ER ranging from 2.82 to 1274.6 keV using a single-phase detector. The measurement is performed with avariety of calibration sources including the 2.82-keV lineof cosmic-ray induced 37Ar. Owing to a high light col-lection efficiency (LCE) of the detector, the low energy37Ar line in the scintillation signal is identified. Althoughthese kinds of measurement under finite electric field isimportant as well, we herein focus on the scintillationresponse at null electric field. We present the energydependence of the scintillation yield, as well as the ba-sic properties of this detector such as the observed lightyield and energy resolutions of the full-absorption peaks.The energy dependence of the scintillation yield down toa few keV is discussed by comparing a model prediction,which is allowed by the use of the 37Ar source.

arX

iv:2

003.

1424

8v3

[ph

ysic

s.in

s-de

t] 2

0 N

ov 2

020

mailto:masato@kylab.sci.waseda.ac.jpmailto:kohei.yorita@waseda.jp

2

LAr

LIQUID FILTER

CRYOSTAT

LIQUEFIER

HEAT

EXC

HA

NG

ER

CIRCULATION PUMP

GA

S FILTER

(MIC

RO

TOR

R)

GETTER

(PU

RER

ON

)

VACUUM PUMP Gas

Liquid

DETECTOR

CRYO- COOLER

LAr

GAr

FIG. 1. LAr handling system consisting of the filling line(left part of the schematic), the vacuum line (top center), therecirculation line (right), and the main cryostat (center). Inthe recirculation line, gaseous argon (GAr) extracted fromthe cryostat is pumped into the getters after passing througha heat exchanger. It then returns to the heat exchanger tobe cooled and is condensed in the liquefier. The cryostatcontaining the detector maintains GAr and LAr over the datacollection period in stable cryogenic conditions.

II. EXPERIMENTAL APPARATUS

The measurement presented here is performed at thesurface laboratory at Waseda University. Figure 1 showsthe argon handling system used in this work. It mainlyconsists of a stainless-steel cryostat of diameter 50 cmand height 100 cm, in which a scintillation detector sits.The argon filled in the cryostat is cooled by the recircu-lation system, which extracts hot gas from the cryostatand passes it through the liquefier with a 200-W GM cry-ocooler (Sumitomo CH-110). The argon is maintained ata typical pressure of 1.4 atm and at a liquid level thatvaries by no more than 1 mm throughout the data col-lection period.

Impurities in the argon (such as water, oxygen, andnitrogen) affect the scintillation properties, resulting in areduced signal yield [13–15]. In order to remove adsorbedimpurities and outgassing from the detector components,the whole system is pumped to vacuum over about tendays before the measurement. The pressure of the cryo-stat reaches below 1.0×10−3 Pa. Then, commercial LArfills the system via a single path through a liquid filterconsisting of a molecular sieve and reduced copper whichremoves electronegative impurities. Additional purifica-tion is continuously performed by the getters (SAES Mi-croTorr MC1500-902 and PURERON GP-5) in the recir-culation system. Several measurements performed in thissystem confirm the concentrations of these impurities arenegligible in this measurement: water and oxygen con-taminations of sub-ppb level and nitrogen contamination

LAr

GAr

PTFE bulk

3-inch PMTs (x2) (TPB-evaporated window)

2-inch PMTs (x4) (TPB-evaporated window)

5.0 cm

6.4

cm

FiducialReflector (TPB-evaporated)

Lead (10 cm)OFC (2 cm)

241Am

137Cs, 22Na, 133Ba, 252Cf

NaI(Tl)

Field-shaper

FIG. 2. Schematic of the LAr scintillation detector (notscaled). The detector including the PMTs is immersed inLAr. Oxygen-free copper (OFC) of roughly 2 cm thick andlead of 10 cm thick surround the cryostat and act as a passiveshield against ambient γ rays. An 241Am source is installedat the outer surface of the PTFE bulk, and the other sources(137Cs, 22Na, 133Ba, and 252Cf) are placed on the outsidesurface of the cryostat wall.

of sub-ppm level.The scintillation detector shown in Fig. 2 is designed

to minimize the loss of scintillation photons in their pathand maximize LCE. The cylindrical fiducial volume of thedetector has a diameter 6.4 cm and a length 5 cm, con-tained within an approximately 3-cm-thick polytetraflu-oroethylene (PTFE) sleeve. The PTFE sleeve servesnot only as the main detector structure but also as aγ-ray emitter, as will be described in Sec. IV B. A mul-tilayer plastic-foil reflector (3M ESR) coated with theTPB wavelength shifter lines the inner surface of thePTFE sleeve. Each end of the cylindrical volume iscapped by a 3-in. Hamamatsu R11065 photomultipliertubes (PMTs), with around 30% quantum efficiency forblue light after wavelength conversion by the TPB. ThePMT windows are also coated with the TPB. Both theTPB layer on the reflector and that on the PMT windowsare deposited using a vacuum-evaporation technique, andtheir amounts are approximately 40 and 30 µg/cm2, re-spectively, corresponding to the deposited-layer thick-nesses of O(1 µm). These are confirmed by a quartzcrystal microbalance sensor and a stylus profiler, as witha procedure similar to that reported in Ref. [16]. The3-in. PMTs are operated with a negative bias voltage of−1570 V. Field-shaping rings with the same bias voltageare embed in the PTFE bulk and ensure electric field in-side the fiducial volume less than 1 V/cm to establish themeasurement under null electric field. The whole sleeveis immersed in a LAr bath contained in the cryostat.

Four 2-in. PMTs (Hamamatsu R6041-506) are imple-mented to view the LAr bath surrounding the fiducialvolume, as shown in Fig. 2. These PMTs are located20 cm above the fiducial volume and just below the liquidsurface so that additional energy deposition in the outerregion is tagged by a coincident scintillation signal. Thewindows of the PMTs are also coated with TPB. A pas-sive shield against ambient γ rays surrounds the cryostat,

3

sample)⋅Charge (counts 0 100 200 300

Eve

nts

1

10

210

310

410

LED Calibration

Data

Fit (Sum)

(0 p.e.) (1 p.e.)

(2 p.e.) (3 p.e.)

(4 p.e.)

Gaus. component

Exp. component

FIG. 3. A typical low-light charge distribution of a fiducial-viewing PMT from a LED calibration run. The charge is rep-resented in units of integrated digitizer count (count ·sample),where 1 count·sample corresponds to 9.8×10−15 C. The solidred line is the model fit as expressed in Eq.(1), and the col-ored lines represent its components. The dashed and dottedlines indicate the Gaussian and exponential terms of the singlephotoelectron response.

which consists of roughly 2-cm-thick oxygen-free copperand 10-cm-thick lead.

The data acquisition (DAQ) system used in this exper-iment consists of a 14-bit, 250-MS/s flash analog-digital-converter (ADC) (Struck SIS3316). The signals fromtwo fiducial-viewing PMTs and four outer-bath PMTsare digitized and recorded. The length of the digitizerrecords is set to 25 µs (5 µs before a trigger point and20 µs after), longer than the lifetime of the slow compo-nent of LAr scintillation light. The trigger is given bythe coincidence, within 1 µs, of the two fiducial PMTswith pulses above a threshold, which is set just abovethe baseline noise and below a typical single photoelec-tron (p.e.) pulse. The coincidence decision is internallymade by the flash ADC board itself. An inhibition timeof 100 µs is introduced after each trigger to prevent re-triggering of the afterpulse of the PMTs, which mainlyoccurs after events with far greater energies than the re-gion of interest (e.g., cosmic-ray events). A Monte Carlo(MC) simulation of the LAr data sample is generated toevaluate the trigger efficiency. By emulating the internaltrigger logic of the flash ADC board on these MC events,the efficiency is found to be consistent with unity for ERsignals larger than 25 p.e., as shown in Fig. 10.

III. EVENT ANALYSIS

A. PMT calibration

The gain of the fiducial-viewing PMTs is calibrated us-ing a blue light-emitting diode (LED) powered by a pulsegenerator. Light pulses from the LED characterized bya width of approximately 20 ns at tenth maximum are

injected into the fiducial volume through optical fiber,while the generator simultaneously triggers the DAQ sys-tem, and the corresponding waveforms from each PMTare recorded over a window of ±1 µs. A baseline ADCcount is determined by the first 0.6 µs of the window, andits subtraction is applied waveform by waveform. Thecharge response of the PMT is measured by integratingthe waveforms within a 48-ns window starting 20 ns priorto the photoelectron pulse arrival time. The gain valueis determined by fitting the charge distribution to modelfunctions. In this analysis, two models are considered todescribe the PMT response. One expression of the mod-els (gain-model A) as a function of the integrated chargeq is followed to that used in Ref. [17]:

f(q) =∑n

P (n;λ)× fn(q), (1)

fn(q) = ρ(q) ∗ ψn∗1 (q),ρ(q) = G(q;x0, σped),

ψ1(q) =pEτ

exp(−q/τ) + (1− pE)G(q;xm, σm)

where P (n;λ) is a Poisson distribution with mean λ,G(q;x, σ) is a Gaussian distribution with mean x andstandard division σ, ∗ denotes a convolution, ψ1(q) isthe PMT single photoelectron response, and ψn∗1 (q) is then-fold convolution of ψ1(q) with itself. This model con-sists of two components comprising the PMT response:a simple Gaussian term, which accounts for a photoelec-tron signal fully amplified by the dynode chain, and anexponential term characterized by a parameter τ , whichaccounts for underamplified photoelectrons and/or feed-back from the dynode photoemission signal. The frac-tion of the single photoelectron response found to be theunderamplified terms is pE . Another expression (gain-model B) is simpler, consisting of only the Gaussian term;i.e., the fraction pE in Eq. (1) is fixed to 0. This assumesthat there is no underamplified or dynode-feedback re-sponse in a PMT and that the photoelectron response isperfectly described by Gaussians.

Figure 3 shows the charge distribution and fit for aLED calibration run with the gain-model A (which has anonzero fraction pE), where 1 count · sample correspondsto an output charge of 9.8× 10−15 C. The mean chargefor a single photoelectron g defined as

g = pEτ + (1− pE)xm (2)

is approximately 2.0 × 106 e−/p.e. with a bias voltageof −1570 V. The fit with the gain-model B (i.e., simpleconvolution of Gaussian functions) returns a 12% highergain value than gain-model A. This difference is nearlyconsistent with the result reported in Ref. [17]. Whilewe do not have enough data to determine which modelis more appropriate to describe the PMT response, thegain-model A is adopted as the baseline, and the resultfrom the model is used in the later analysis. This calibra-tion is performed every 12 hours during a data collectionperiod lasting seven days. The overall stabilities of the

4

Detected Light (p.e.)0 2000 4000 6000 8000

PSD

Par

amet

er (

Slow

/Tot

al)

0

0.2

0.4

0.6

0.8

1

0

10

20

30

40

50

60

Na Back-to-Back Data22

ER 95%-Band

FIG. 4. Distribution of the PSD parameter (“Slow/Total”)versus the observed light signal. The data require the back-to-back tagging described in Sec. III C. The red dashed linescorrespond to the 95% containing band for ER events.

gain and observed light yield during the period are withinless than 0.5% from both the LED measurement and anenergy calibration mentioned below.

The nonlinearity of the PMT is studied by a pulsedlaser source, and we found that the effect is less than 1%(0.1%) at 1 MeV (below 200 keV) at the operation volt-age. The observed light yields are corrected accordingly,and its correction factors are considered as a systematicuncertainty.

B. Signal analysis and selection criteria

The analysis of the LAr scintillation signal is per-formed following a photon-counting algorithm. Foreach waveform, this algorithm first calculates the base-line from the pretrigger window; once that baseline issubtracted, all samples above a software threshold aregrouped with three neighboring samples (one bin beforeand two bins after). The software threshold is set basedon the baseline noise and is below a typical single photo-electron PMT pulse. The signal detection time is identi-fied as the first sampling time above a threshold of 50%peak amplitude. Detected scintillation light is defined asthe integrated charge in the time interval between −0.04and 7.0 µs. A pulse shape discrimination (PSD) param-eter is also defined as the fraction of light detected after0.1 µs of the scintillation signal (termed “slow/total”).

A set of data quality cuts is applied to remove instru-mental effects and event pileups. The selection criteriaare as follows: (1) Software imposes a 10-ms veto af-ter events that contain signals greater than ≈2.0 × 104(≈5.0 × 103 p.e.) for datasets taken with a γ-ray sourcewith >100 keV (

5

Detected Light (p.e.)0 5000 10000 15000 20000

Eve

nts

/ 40

p.e.

1

10

210

310

410

510

Na Spectrum22

511 keV 1274.6 keV

Self Trigger

BtB Tagging

1274.6 keV

/ ndf 2χ 72.61 / 72 µ 1.97e+01± 1.57e+04

(Corrected 1.56e+04)

µ / σ 0.00085± 0.02107

511.0 keV

/ ndf 2χ 96.51 / 48 µ 11.0± 6623

(Corrected 6520)

µ / σ 0.00187± 0.02887

FIG. 6. The observed light spectra from the 22Na sourcebefore and after requiring back-to-back coincidence (BtB tag-ging). The red and magenta lines represent the fit functionfor the 1274.6-keV peak in self-trigger data and the 511-keVpeak in back-to-back data, respectively.

with the 137Cs source. The full-absorption peak of the661.7-keV line of the 137Cs source is fit with a Gaussianwith mean µ and width σ. The continuous backgroundcomponents around the peak, mainly coming from theCompton edge and degraded tails, are modeled with errorand linear functions and added to the fit function. Thefit shown in Fig. 5 returns χ2/ndf = 62.5/56.

The observed light spectra obtained with the 22Nasource are shown in Fig. 6. In this measurement, an addi-tional NaI(Tl) scintillator (2×2 in.2 cylinder) is set withthe source at opposite sites of the cryostat to tag thebackward-traveling 511.0-keV γ ray (back-to-back tag-ging). The distance between the cryostat wall and thesource is set to 15 cm, and that between the source andthe scintillator to 25 cm. The black and blue spectrain Fig. 6 are the observed scintillation spectra before andafter requiring the coincidence detection of the 511.0-keVγ-ray signal in the NaI(Tl) scintillator. Since the 1274.6-keV γ ray is considered to have no angular correlationwith back-to-back γ rays, the corresponding peak ap-pears only in the former spectrum. Each peak is fit witha Gaussian plus background model function consisting oferror and linear functions. Values of χ2/ndf = 72.6/72and χ2/ndf = 96.5/48 are returned from the fits for1274.6- and 511.0-keV peaks, respectively.

These observed photoelectron signals contain extracharge from PMT afterpulses and systematic effect fromthe photon-counting algorithm. A correction for theseeffects is thus applied to reconstruct the observed lightsignal per ER energy. This correction is based on an in-dependent study of the PMT response as well as a MCsimulation of the LAr signal. It is relatively small, ap-proximately 1% for the 137Cs line and less than 3% for thewhole energy region of interest of this analysis, where theamount of afterpulse is estimated as 2%–4% of the pho-toelectron signal, and the algorithm can systematicallyslightly underestimate the charge signal. The observed

Detected Light (p.e.)0 1000 2000 3000 4000 5000 6000

Eve

nts

/ 40

p.e.

10

210

310

410

510

Ba Spectrum (Self Trigger)133

/ ndf 2χ 28.29 / 15 µ 5.7± 4715

(Corrected 4647)

µ / σ 0.00153± 0.02913

FIG. 7. The observed light spectrum from the 133Ba source.The red line represents the fit function.

light yields after the corrections are summarized in Ta-ble I with uncertainties. The uncertainty includes theestimation of PMT afterpulses, systematic error in thecorrections, and stability of the detector.

IV. MEASUREMENT OF SCINTILLATIONRESPONSE WITH CALIBRATION SOURCES

A. Barium-133 source

The detector is exposed to a 356.0-keV γ ray using a133Ba radioactive source with approximately 1 MBq. Thespectrum obtained with a 133Ba source is shown in Fig. 7.The peak around 4700 p.e. corresponds to the γ-ray lineand fitted with a Gaussian. An exponential function isadded to the fit function to model the overall backgroundcomponents; the main background sources are due to thedegraded γ-ray tail and the γ-ray spectra of the othertwo lines of the 133Ba source around the peak energy(those at 383.9 and 302.9 keV) that have relatively highintensity. The resulting fit function is overlaid in Fig. 7.

B. Californium-252 source exploiting γ raysthrough the (n, n′γ) reaction with fluorine-19

Measurements for the 109.8- and 197.1-keV quasimo-noenergetic lines are performed using γ rays emitted fromthe (n, n′γ) reaction with 19F [18]. As an external fastneutron source, a 252Cf source with a spontaneous fissionrate of approximately 1× 105 fission/s is used. The dis-tance between the center of the fiducial volume and thesource is set to 90 cm. The NaI(Tl) scintillator is placedbeside the source to detect associated γ rays from thespontaneous fission and to provide timing information.Fast neutrons from 252Cf generate (n, n′γ) reaction with19F in the PTFE bulk, producing quasimonoenergeticγ rays. Although the intensities of each quasimonoen-ergetic line depend upon their incident neutron energy,

6

Detected Light (p.e.)0 1000 2000 3000 4000

Eve

nts

/ 40

p.e.

210

310

410

Cf Spectrum (TOF Tagging)252

109.8 keV

197.1 keV

109.8 keV / ndf 2χ 2.594 / 9

1µ 1.4± 1457 (Corrected 1447)

1

µ / 1σ 0.00110± 0.04542 197.1 keV

/ndf 2χ 24.36 / 16

2µ 3.0± 2636 (Corrected 2605)

2

µ / 2σ 0.0013± 0.0358

FIG. 8. The observed light spectrum from the 252Cf sourceafter requiring the TOF to be consistent with fast neutrons.The magenta and red lines represent the fit functions for109.8- and 197.1-keV peaks, respectively.

Detected Light (p.e.)200 400 600 800 1000

Eve

nts

/ 10

p.e.

0

50

100

150

200

250

-Tagging)αAm Spectrum (241

/ ndf 2χ 15.48 / 23

L.Y. [p.e./keV] 0.02± 12.98

(Corrected 12.96)

µ / σ 0.0014± 0.0506

FIG. 9. The observed light spectrum from the 241Am sourceby requiring α-ray detection by the veto PMTs, along withthe MC fit spectrum (red line). The blue dashed vertical linesrepresent the fitting range.

109.8- and 197.1-keV lines are major channels for therange of neutron energy from 252Cf. Time differencesbetween the NaI(Tl) and fiducial signals (time of flight;TOF) are used to remove γ-ray events that come directlyfrom the fission. Figure 8 shows the spectrum and fittingresults for corresponding peaks. Each peak is fit by aGaussian plus exponential function.

C. Americium-241 source

To expose the detector to 59.5-keV γ rays, an 241Amsource of approximately 40 Bq is used. The radioactivesource is deposited on a 100-µm-thick platinum foil in-stalled at the outer surface of the PTFE bulk. It de-cays into an excited level of 237Np via α-ray transition,and subsequent deexcitation of the 237Np emits γ rayswith a major line of 59.5 keV. The scintillation signal

Detected Light (p.e.)0 20 40 60 80 100

Eve

nts

/ 1 p

.e.

0

20

40

60

80

100

120

Tri

gger

Eff

icie

ncy

0

0.2

0.4

0.6

0.8

1Ar Spectrum (Self Trigger)37

/ ndf 2χ 82.21 / 84 µ 0.58± 28.93

(Corrected 29.79)

µ / σ 0.0229± 0.1931

FIG. 10. The 37Ar spectrum obtained by requiring anticoin-cidence with the outer bath PMTs for the no external sourcedata. The cyan dashed line represents the estimated triggerefficiency, and the data are corrected based on this curve.

from the α ray from the primary disintegration is de-tected by the outer-bath PMTs, allowing the γ-ray in-teraction to be proved in the fiducial volume. Figure 9shows the observed light spectrum after requiring the de-tection of the α-ray signals in the outer region. Becauseof the relatively low energy of the γ ray from 241Am andthe passive components between the source and the fidu-cial volume, the spectrum does not exhibit a clear full-absorption peak. The tail of the peak comes from γ raysthat reach the fiducial volume via single or multiple scat-tering from any materials in their path.

The detector response to a 59.5-keV γ ray is evaluatedvia MC simulation of the experimental setup based onthe Geant4 toolkit [19, 20]. The MC simulation takesinto account the detector geometry and composition in-side the LAr bath, as well as the radioisotope mountingstructure. It proceeds by generating γ rays from 241Amwith a random momentum direction and calculating theenergy deposition in the fiducial volume. The observedspectrum is fitted by converting the energy depositionto the observed light yield with a constant scintillationyield, constant LCE, and Gaussian resolution. The bestfit spectrum is also shown in Fig. 9; although the fit isperformed only around the 59.5-keV peak (700–900 p.e.),reasonable agreement between the data and MC is founddown to around 400 p.e.

D. Argon-37 source

Measurement for ERs of a few keV is performed using37Ar, which is the second most abundant radioactive iso-tope in atmospheric argon, comprising an abundance of≈1.3 × 10−20 [21]. It decays via electron capture to theground state of 37Cl with a half-life of 35 days, producingx rays and Auger electrons with a total energy release of2.82 keV (for K-shell capture), 0.27 keV (for L-shell cap-ture), or 0.02 keV (for M-shell capture) [22, 23]. Since

7

(p.

e./k

eV)

γ/Eµ

10

10.5

11

11.5

12

12.5

13

13.5

12.8 p.e./keV

Energy (keV)1 10 210 310

µ/σ

2−10

1−10

Fit

Ar

37

Am

241

)γF

(n,

n'19

)γF

(n,

n'19

Ba

133 N

a22

Cs

137 N

a22

FIG. 11. Top: observed light yields obtained by the fittinganalysis for each calibration line divided by corresponding in-cident energy. The red dashed line represents the energy cal-ibration using 511.0-keV full-absorption peak. Bottom: en-ergy resolution of the detector measured with full-absorptionpeaks. The red dashed line represents the fit function withstochastic and constant terms (see text).

the production of 37Ar is mainly due to cosmogenic acti-vation of atmospheric argon [21], it is expected to reachequilibrium and the decay rate of 37Ar in the detectoris expected to be constant from the argon filling time tothe end of measurement.

The data used in this measurement come from approx-imately 27 hours of detector operation without any exter-nal sources. Figure 10 shows the observed light spectrumfor this measurement. The spectrum consists of eventsthat do not have associated scintillation signals in any ofthe four outer-bath PMTs. The peak around 25 p.e. isattributed to the energy release of 2.82 keV from 37Ar.No structures corresponding to the L- or M-shell cap-ture could be seen, probably due to the large amount ofrandom coincidence background and the lack of photo-statistics. The spectrum with 37Ar is fitted with the sumof the Gaussian, exponential, and constant terms thatdescribe the signal and low energy background model.The rate of 37Ar decays returned by the fit is approxi-mately 25 mBq/kg, which is compatible with literaturevalues [21, 24, 25]. The goodness of fit for the peak isχ2/ndf = 82.21/84.

V. SCINTILLATION YIELD AND ENERGYRESOLUTION

The upper panel of Fig. 11 summarizes the mean valuesof the number of detected photoelectron divided by cor-responding incident energies measured by the set of ra-

TABLE II. Observed coefficients and estimated contributionsof the stochastic (S) and constant (C) terms of the energyresolution. Although the origin of the constant term is notquantitatively estimated, almost all of which is believed tocome from the geometrical effect.

Type Source Coefficient (α)

S(σµ

= α√Eγ

)

Data 0.37± 0.03Photostatistics ≈0.3Multiple scattering

8

TABLE III. Summary of the systematic uncertainty sources for the measurements of the light yields for each full-absorptionpeak and energy resolution.

SystematicScintillation yields Energy resolution

Dataset Fraction Dataset FractionPMT afterpulse All 2.0%PMT gain nonlinearity All

9

1 10 210 310Energy (keV)

0

10

20

30

40

50

Yie

ld (

phot

on/k

eV)

Data (This Work)

)-0.008

+0.012 = 0.033ςTIB (

(Jpn.J.App.Phys. 41 1538)Doke et al.

Doke-Birks' Law

(PRC 81 045803)CLEAN

(AstropartPhys 49 44)DarkSide-10

(PRD 97 112005)-ray)γARIS (

(PRD 97 112005)ARIS (Compt.Electron)

(arXiv:1909.02207)et al.Xiong

FIG. 13. Measured scintillation yield as a function of the inci-dent energy Eγ (black solid circle). The absolute yield is de-termined by referring the measurement by Doke et al. (blackopen circle) [29]. The TIB model function with a parameterfound by the 2.82-keV point is shown with its uncertainty (redband). The results from other experiments, CLEAN (violetstar point) [30], DarkSide-10 (green filled square) [17], ARIS(blue open square and blue filled rhombus) [31], and Xionget al. (green open rhombus) [32] are also shown where eachyield is normalized at 511 keV referring the Doke-Birks’s law(gray solid line) [29].

where Eer is the recoiled electron energy, W = 19.5 eV isthe effective work function [27], Nex and Ni are the num-bers of produced excitons or electron-ion pairs, respec-tively, α = 0.21 is the initial ratio of the average of Nex toNi [33], and ς is a constant parameter of the model. Con-sidering the facts that the number of the interaction pointof the 37Ar events can be approximated to be one due toits low energy deposition and decay mode mainly con-sisting of Auger electrons [22], and that the TIB modelis fully applied for liquid xenon at corresponding energywhere the electron track length is smaller than the ther-malization distance of the ionization electron [6, 34], wedetermine the parameter ς from the 37Ar data. It is cal-culated as ς = 0.033+0.012−0.008 and represented with the redband in Fig. 13. Further studies, such as additional mea-surements around 10 keV and discussion on the stitchingbetween the TIB model and Doke-Birks’s law, shouldbe performed in future work. This result also would bepractically essential input for tuning the response modelimplemented, for instance, in the NEST package [35].

VII. CONCLUSION

The energy dependence of the scintillation yield forelectronic recoils ranging from 2.82 to 1274.6 keV is mea-sured using a single-phase detector with high LCE ex-posed to a variety of calibration sources. The scintil-

lation detector with the TPB wavelength shifter is im-mersed in purified LAr and yields 12.8 ± 0.3 p.e./keV(11.2±0.3 p.e./keV) for a 511.0-keV γ-ray full-absorptionevent based on the PMT calibration assuming a PMTsingle photoelectron response model with an additionalexponential term (with only a Gaussian term), and itsenergy resolution is 3% for the γ-ray line. The scin-tillation response is investigated by the full-absorptionpeaks of external γ-ray sources, as well as an 37Ar sourcewith a 2.82-keV line. These measurements demonstratethat the scintillation yield decreases in the low energyregion. We interpret it by analogy with the LXe scintil-lation detector response, where the ionization electron-ion recombination probability is attributed to the energydependence of the yield. By referring the previous mea-surement of the scintillation yield at 1 MeV, the TIBmodel parameter ς is calculated by the 2.82-keV point asς = 0.033+0.012−0.008.

This work is primarily intended for use in the directWIMP dark matter search. In this field, low energyelectronic background is one of the most severe sourcesdisturbing the lower energy threshold, hence, reducingWIMP sensitivity. The result presented here makes useof the precise estimation of background contaminationin the low energy region and suppression of the system-atic uncertainty. The measurement of the scintillationresponse under nonzero electric field, which is the mat-ter for a double-phase detector (e.g., [1, 36]), is left forfuture work. In addition, the measurement of the en-ergy resolution for the keV to MeV range in this workprovides useful information for applying the LAr detec-tor to other fields, such as astrophysical MeV gamma-rayobservation [37]. The results presented here would helpwith the design, operation, and analysis of a wide varietyof astrophysical and particle physics experiments in thenear future to enhance their physical reach.

ACKNOWLEDGMENTS

This work is a part of the outcome of research per-formed under the Waseda University Research Institutefor Science and Engineering (Project No. 2016A-507)supported by Japan Society for the Promotion of ScienceGrant-in-Aid for Scientific Research on Innovative Areas(Grants No. 15H01038 and No. 17H05204), Grant-in-Aid for Scientific Research(B) (Grant No. 18H01234),and Grant-in-Aid for Japan Society for the Promotion ofScience Research Fellow (Grant No. 18J13018). The au-thors would like to thank the Material CharacterizationCentral Laboratory at Waseda University for granting usaccess to their stylus profiler. The authors acknowledgethe support of the Institute for Advanced Theoretical andExperimental Physics, Waseda University.

10

[1] P. Agnes, I. F. M. Albuquerque, T. Alexander, A. K.Alton, G. R. Araujo, M. Ave, H. O. Back, B. Baldin,G. Batignani, K. Biery, et al. (DarkSide Collabora-tion), DarkSide-50 532-day dark matter search with low-radioactivity argon, Phys. Rev. D 98, 102006 (2018).

[2] R. Ajaj, P.-A. Amaudruz, G. R. Araujo, M. Baldwin,M. Batygov, B. Beltran, C. E. Bina, J. Bonatt, M. G.Boulay, B. Broerman, et al. (DEAP Collaboration),Search for dark matter with a 231-day exposure of liquidargon using DEAP-3600 at SNOLAB, Phys. Rev. D 100,022004 (2019).

[3] D. S. Akerib, S. Alsum, C. Aquino, H. M. Araújo, X. Bai,A. J. Bailey, J. Balajthy, P. Beltrame, E. P. Bernard,A. Bernstein, et al. (LUX Collaboration), First Searchesfor Axions and Axionlike Particles with the LUX Exper-iment, Phys. Rev. Lett. 118, 261301 (2017).

[4] E. Aprile, J. Aalbers, F. Agostini, M. Alfonsi, L. Al-thueser, F. D. Amaro, V. C. Antochi, E. Angelino,F. Arneodo, D. Barge, et al. (XENON Collaboration),Light Dark Matter Search with Ionization Signals inXENON1T, Phys. Rev. Lett. 123, 251801 (2019).

[5] E. Aprile et al. (XENON Collaboration), Observation ofExcess Electronic Recoil Events in XENON1T, (2020),2006.09721.

[6] M. Szydagis, N. Barry, K. Kazkaz, J. Mock, D. Stolp,M. Sweany, M. Tripathi, S. Uvarov, N. Walsh, andM. Woods, NEST: a comprehensive model for scintilla-tion yield in liquid xenon, J. Instrum. 6, P10002 (2011).

[7] B. Lenardo, K. Kazkaz, A. Manalaysay, J. Mock,M. Szydagis, and M. Tripathi, A Global Analysis of Lightand Charge Yields in Liquid Xenon, IEEE Trans. Nucl.Sci. 62, 3387 (2015), ISSN 0018-9499.

[8] M. Szydagis, J. Balajthy, J. Brodsky, J. Cutter, J. Huang,E. Kozlova, B. Lenardo, A. Manalaysay, D. McKinsey,M. Mooney, et al., Noble Element Simulation Techniquev2.0, 10.5281/zenodo.1314669, (2018).

[9] A. Hitachi, T. Takahashi, N. Funayama, K. Masuda,J. Kikuchi, and T. Doke, Effect of ionization density onthe time dependence of luminescence from liquid argonand xenon, Phys. Rev. B 27, 5279 (1983).

[10] T. Heindl, T. Dandl, M. Hofmann, R. Krücken, L. Ober-auer, W. Potzel, J. Wieser, and A. Ulrich, The scintilla-tion of liquid argon, Europhys. Lett. 91, 62002 (2010).

[11] W. M. Burton and B. A. Powell, Fluorescence ofTetraphenyl-Butadiene in the Vacuum Ultraviolet, Appl.Opt. 12, 87 (1973).

[12] G. Porter and M. R. Topp, Nanosecond flash photolysis,Proc. R. Soc. A 315, 163 (1970).

[13] R. Acciarri, M. Antonello, B. Baibussinov, M. Baldo-Ceolin, P. Benetti, F. Calaprice, E. Calligarich, M. Cam-biaghi, N. Canci, F. Carbonara, et al., Oxygen contami-nation in liquid Argon: combined effects on ionizationelectron charge and scintillation light, J. Instrum. 5,P05003 (2010).

[14] R. Acciarri, M. Antonello, B. Baibussinov, M. Baldo-Ceolin, P. Benetti, F. Calaprice, E. Calligarich, M. Cam-biaghi, N. Canci, F. Carbonara, et al., Effects of Nitro-gen contamination in liquid Argon, J. Instrum. 5, P06003(2010).

[15] B. J. P. Jones, T. Alexander, H. O. Back, G. Collin, J. M.Conrad, A. Greene, T. Katori, S. Pordes, and M. Toups,

The effects of dissolved methane upon liquid argon scin-tillation light, J. Instrum. 8, P12015 (2013).

[16] B. Broerman, M. Boulay, B. Cai, D. Cranshaw, K. De-ring, S. Florian, R. Gagnon, P. Giampa, C. Gilmour,C. Hearns, et al., Application of the TPB WavelengthShifter to the DEAP-3600 Spherical Acrylic Vessel InnerSurface, J. Instrum. 12, P04017 (2017).

[17] T. Alexander, D. Alton, K. Arisaka, H. Back, P. Bel-trame, J. Benziger, G. Bonfini, A. Brigatti, J. Brodsky,L. Cadonati, et al., Light yield in DarkSide-10: A pro-totype two-phase argon TPC for dark matter searches,Astroparticle Phys. 49, 44 (2013).

[18] V. C. Rogers, Inelastic neutron scattering in 19F, Phys.Rev. C 9, 527 (1974).

[19] S. Agostinelli, J. Allison, K. Amako, J. Apostolakis,H. Araujo, P. Arce, M. Asai, D. Axen, S. Banerjee,G. Barrand, et al., Geant4—a simulation toolkit, Nucl.Instrum. Methods Phys. Res., Sect. A 506, 250 (2003),ISSN 0168-9002.

[20] J. Allison, K. Amako, J. Apostolakis, H. Araujo, P. A.Dubois, M. Asai, G. Barrand, R. Capra, S. Chauvie,R. Chytracek, et al., Geant4 developments and applica-tions, IEEE Trans. Nucl. Sci. 53, 270 (2006).

[21] R. Saldanha, H. O. Back, R. H. M. Tsang, T. Alexan-der, S. R. Elliott, S. Ferrara, E. Mace, C. Overman, andM. Zalavadia, Cosmogenic production of 39Ar and 37Arin argon, Phys. Rev. C 100, 024608 (2019).

[22] B. T. Cleveland, T. Daily, J. Raymond Davis, J. R. Dis-tel, K. Lande, C. K. Lee, P. S. Wildenhain, and J. Ull-man, Measurement of the Solar Electron Neutrino Fluxwith the Homestake Chlorine Detector, Astrophys. J.496, 505 (1998).

[23] M.-M. Bé, V. Chisté, C. Dulieu, X. Mougeot, V. Chechev,F. Kondev, A. Nichols, X. Huang, and B. Wang, Tableof Radionuclides, Vol. 7 (Bureau International des Poidset Mesures, Sèvres, 2013).

[24] R. Purtschert, M. Kalinowski, P. Bourgouin, E. Wies-lander, X. Blanchard, R. Riedmann, L. Raghoo,J. Kusmierczyk-Michulec, A. Gheddou, and C. S. adnYutaka Tomita, in Proceedings of the CTBT Science andTechnology 2017 Conference (CTBT, Vienna, 2017),URL https://ctnw.ctbto.org/ctnw/event/3239/slides/5908D5735D9438FFE05308351BAC687C.

[25] P. Agnes, I. F. M. Albuquerque, T. Alexander, A. K.Alton, G. R. Araujo, D. M. Asner, M. Ave, H. O. Back,B. Baldin, G. Batignani, et al. (DarkSide Collaboration),Low-Mass Dark Matter Search with the DarkSide-50 Ex-periment, Phys. Rev. Lett. 121, 081307 (2018).

[26] J. Thomas and D. A. Imel, Recombination of electron-ion pairs in liquid argon and liquid xenon, Phys. Rev. A36, 614 (1987).

[27] T. Doke, H. J. Crawford, A. Hitachi, J. Kikuchi, P. J.Lindstrom, K. Masuda, E. Shibamura, and T. Takahashi,Let dependence of scintillation yields in liquid argon,Nucl. Instrum. Methods Phys. Res., Sect. A 269, 291(1988).

[28] M. Berger, J. Hubbell, S. Seltzer, J. Chang, J. Coursey,R. Sukumar, D. Zucker, and K. Olsen, XCOM: Photoncross sections database (2010), URL https://www.nist.gov/pml/xcom-photon-cross-sections-database.

https://ctnw.ctbto.org/ctnw/event/3239/slides/5908D5735D9438FFE05308351BAC687Chttps://ctnw.ctbto.org/ctnw/event/3239/slides/5908D5735D9438FFE05308351BAC687Chttps://www.nist.gov/pml/xcom-photon-cross-sections-databasehttps://www.nist.gov/pml/xcom-photon-cross-sections-database

11

[29] T. Doke, A. Hitachi, J. Kikuchi, K. Masuda, H. Okada,and E. Shibamura, Absolute Scintillation Yields in LiquidArgon and Xenon for Various Particles, Jpn. J. Appl.Phys. 41, 1538 (2002).

[30] W. H. Lippincott, S. B. Cahn, D. Gastler, L. W. Kastens,E. Kearns, D. N. McKinsey, and J. A. Nikkel, Calibrationof liquid argon and neon detectors with 83Krm, Phys.Rev. C 81, 045803 (2010).

[31] P. Agnes, J. Dawson, S. De Cecco, A. Fan, G. Fiorillo,D. Franco, C. Galbiati, C. Giganti, T. N. Johnson, G. Ko-rga, et al. (ARIS Collaboration), Measurement of theliquid argon energy response to nuclear and electronicrecoils, Phys. Rev. D 97, 112005 (2018).

[32] W.-X. Xiong, M.-Y. Guan, C.-G. Yang, P. Zhang, J.-C.Liu, C. Guo, Y.-T. Wei, Y.-Y. Gan, Q. Zhao, and J.-J.Li, Calibration of liquid argon detector with 83mKr and22Na in different drift field, (2019), 1909.02207.

[33] M. Miyajima, T. Takahashi, S. Konno, T. Hamada,S. Kubota, H. Shibamura, and T. Doke, Average energy

expended per ion pair in liquid argon, Phys. Rev. A 9,1438 (1974), and Phys. Rev. A 10 (1974) 1452.

[34] A. Mozumder, Free-ion yield in liquid argon at low-LET,Chem. Phys. Lett. 238, 143 (1995).

[35] NEST noble element simulation technique, URL http://nest.physics.ucdavis.edu.

[36] C. E. Aalseth, F. Acerbi, P. Agnes, I. Albuquerque,T. Alexander, A. Alici, A. Alton, P. Antonioli, S. Arcelli,R. Ardito, et al., DarkSide-20k: A 20 tonne two-phaseLAr TPC for direct dark matter detection at LNGS, Eur.Phys. J. Plus 133, 131 (2018), 1707.08145.

[37] T. Aramaki, P. O. H. Adrian, G. Karagiorgi, andH. Odaka, Dual MeV gamma-ray and dark matter ob-servatory - GRAMS Project, Astropart. Phys. 114, 107(2020).

http://nest.physics.ucdavis.eduhttp://nest.physics.ucdavis.edu

Liquid argon scintillation response to electronic recoils between 2.8–1275 keV in a high light yield single-phase detectorAbstractI IntroductionII Experimental apparatusIII Event analysisA PMT calibrationB Signal analysis and selection criteriaC Determination of photoelectron per keV with sodium-22 and cesium-137 sources

IV Measurement of scintillation response with calibration sourcesA Barium-133 sourceB Californium-252 source exploiting rays through the (n, n') reaction with fluorine-19C Americium-241 sourceD Argon-37 source

V Scintillation yield and energy resolutionVI TIB model interpretation on scintillation responseVII Conclusion Acknowledgments References