Characterization of water-based liquid scintillator for Cherenkovand scintillation separation

J. Caravaca1,2, B.J. Land1,2, M. Yeh3, and G.D. Orebi Gann1,2

1University of California, Berkeley, CA 94720-7300, USA2Lawrence Berkeley National Laboratory, CA 94720-8153, USA3Brookhaven National Laboratory, Upton, NY 11973-500, USA

Abstract

This paper presents measurements of the scintillation lightyield and time profile of a number of concentrationsof water-based liquid scintillator, formulated from lin-ear alkylbenzene (LAB) and 2,5-Diphenyloxazole (PPO).The separation between Cherenkov and scintillation lightis quantified using cosmic muons in the CHESS experi-ment for each formulation, and we discuss the prospectsfor large-scale detectors.

1 Introduction

Liquid-based optical detectors have been used to greatsuccess in particle physics, in particular for neutrinophysics and rare event searches (see Sec. 35.3.1 of [1]for a complete overview). Deployment of next-generationtechnology in future detectors has the potential to en-able further discovery. A broad community is pursu-ing the idea of a hybrid optical detector, that can lever-age both Cherenkov and scintillation signals simultane-ously [2–4]. Such a detector could achieve good energyresolution while also having sensitivity to particle direc-tion. The ratio of Cherenkov to scintillation light wouldalso offer a powerful handle for particle and event iden-tification, while a clean Cherenkov signal would facili-tate the ring-imaging necessary for long-baseline neutrinophysics. A detector with these capabilities would have fa-vorable signal to background ratio across a broad spec-trum of physics topics [5–8].

Given the much higher light yield for scintillation overCherenkov, the detection of a clean Cherenkov signalfrom pure scintillator is very challenging. Nevertheless, ithas been successfully demonstrated in the CHESS exper-iment[9] and other cases [10] using a technique that ex-ploits the differences in emission time profiles and topol-ogy.

The pursuit of water-based liquid scintillator (WbLS)[11], a mixture of liquid scintillator (LS) in water, is onepossible avenue to allow optical detectors to reach theserequirements. First, it provides a scintillator with an ab-sorption length close to water, increasing the photon de-tection efficiency in large scintillator detectors; second,it provides a better energy resolution than that of pureCherenkov detectors; and third it may enhance the iden-tification of Cherenkov light in scintillator detectors via areduced scintillation yield. The exact formulation can beoptimized to address particular physics goals, includingmodifying the choice and quantity of fluor to affect boththe scintillation yield and time profile.

The goal of this paper is to characterize the scintilla-tion light yield and time profile of several WbLS formu-lations in order to define a Monte Carlo (MC) model thatcan be used in the future to predict performance of WbLSin large detectors. We study 1%, 5% and 10% formula-tions of LS in water, using both β and γ sources, in allcases using linear alkylbenzene (LAB) as the base LS,with 2,5-Diphenyloxazole (PPO) as a fluor. This effortis complementary to [12], where the characterization ofWbLS is performed with a pulsed X-ray beam source andno Cherenkov radiation is produced. We also use vertical-

1

arX

iv:2

006.

0017

3v2

[ph

ysic

s.in

s-de

t] 1

7 Ju

n 20

20

going cosmic muons in order to demonstrate detectionof Cherenkov rings in WbLS and to further validate ourMC model. This manuscript is structured in the follow-ing way: details of the MC model of the apparatus andthe WbLS behaviour are presented in Sec. 2; the WbLSscintillation light yield and time profile measurements aredescribed in Sec. 3.1 and Sec. 3.2, respectively; evalua-tion of the Cherenkov and scintillation separation for cos-mic muons and a comparison to model prediction usingthe CHESS experiment is shown in Sec. 4; and we con-clude with Sec. 5 by holding a qualitative discussion ofthe performance of WbLS in large-scale detectors.

2 Monte Carlo modelA detailed MC simulation of photon creation, propagationand detection is used in this analysis, as included in theGEANT4-based [13] simulation package RAT-PAC [14],described in [4]. In short, it includes detector geometry,a 3D photomultiplier (PMT) model, the different opticalproperties of the materials and a model of the data acqui-sition. It uses the GLG4Scint model [14] to simulate thescintillation light emission, photon absorption and photonreemission, and default GEANT4 models for Cherenkovphoton production and Rayleigh scattering.

The inputs to the model from measurements presentedin this paper are the absolute, intrinsic scintillation lightyield and the scintillation emission time profile. Otherparameters are implemented from external measurementsor estimations, as described in the remainder of this sec-tion. This includes the scintillation emission spectrum,refractive index, the attenuation length, Rayleigh scatter-ing length, and the reemission probability.

2.1 Emission spectrumWe use the spectra reported in [12] for each WbLS con-centration. The different spectra have been measured tobe very close to that of pure LABPPO (2g/L) [12]. Thisis justified by the fact that the emission process is dom-inated by the radiative transition in the scintillator; ad-dition of non-scintillation components doesn’t affect theemission spectrum. Furthermore, the emission spectrumof LABPPO for PPO concentrations above 2g/L is verysimilar to that of the 2g/L [10, 12].

2.2 Refractive index estimation

In order to estimate the refractive index for WbLS, n,Newton’s formula for the refractive index of liquid mix-tures [15] is used:

n =√

φLSn2LS +(1−φLS)n2

W , (1)

where φLS is the volume fraction of LABPPO in WbLS,and nLS and nW correspond to the measured refractive in-dexes for LABPPO [16] and water [17], respectively. Dueto the large water concentration, the WbLS refractive in-dex is very similar to that of pure water.

2.3 Absorption and scattering lengths

The absorption length, λ , of WbLS input to the MC modeldepends on the molarity c of each of the components and,thus, on the concentration of liquid scintillator in water. Itis computed as:

λ = (εlabclab + εppocppo + εwatercwater)−1 , (2)

where εlab, εppo and εwater are the molar absorption coeffi-cients of LAB, PPO [16] and water ([18] for wavelengthsover 380 nm and [19] for wavelengths below 380 nm).

In the same fashion, the Rayleigh scattering length, λ s,for WbLS is estimated as:

λs =(ε

slabclab + ε

sppocppo + ε

swatercwater

)−1, (3)

where εslab, εs

ppo and εswater are the molar scattering coef-

ficients for LAB, PPO [16] and water, respectively. As inthe previous case, the estimated values of these propertiesfor WbLS are very close to those of pure water.

This method might overestimate the attenuationlengths, in particular the component due to scattering,given the complex chemical structure and composition ofWbLS. This has negligible impact on the results presentedhere, due to the small (centimeter-) scale of the apparatus.The impact will be more significant for extrapolations ofthese results to large detectors. A planned long-arm mea-surement of absorption and scattering lengths is under-way.

2

2.4 Scintillation reemission

A photon absorbed by the scintillator volume has a non-zero probability of being reemitted. This reemission pro-cess becomes important at low wavelengths where the ab-sorption is strong, shifting photons to longer wavelengthswhere the detection probability is higher due to a smallerphoton absorption and a greater PMT quantum efficiency.The probability preem

i of a component i in the WbLS ab-sorbing a photon of frequency ω is determined as the con-tribution of the given component to the total absorptioncoefficient:

preemi (ω) = φiαi(ω)/α(ω), (4)

where the variables have been previously defined. After aphoton is absorbed, it can be reemitted with a 69% proba-bility for LAB and an 80% probability for PPO [16], fol-lowing the same primary emission spectrum.

3 Methods and measurements

Measurements of the scintillation light yield and time pro-file, using two distinct bench-top setups, are described inthe following section.

3.1 Light yield measurement

To determine the absolute scintillation light yield ofWbLS (number of photons produced per unit of energydeposited), we measure the number of photoelectrons(NPE) detected by a PMT located in front of a WbLS sam-ple excited with a known radioactive source. We fit ourMC model (see Sec. 2), which is previously calibrated, tothe NPE distribution, by tuning the intrinsic light yield pa-rameter in the model. Given that we simulate the relevantgeometry, optical properties of the materials, Cherenkovproduction, PMT response and data acquisition system,this technique takes into account the effects introduced bythose factors. In this way, the extracted light yield corre-sponds to an absolute property of the medium, rather thanbeing defined relative to scintillator standards. Details ofthe apparatus, calibration, analysis methods and resultsare given below.

3.1.1 Description of the apparatus

A minimal configuration is used to measure theabsolute light yield of the WbLS mixtures. A10 mm×10 mm×40 mm, 1-mm thick, high-performancequartz cuvette filled with WbLS is located 10 cm fromthe front face of a 10-inch R7081 Hamamatsu PMT (seeFig. 1.a). The PMT is fixed to a black aluminum struc-ture for mechanical stability to reduce possible geometry-related systematic uncertainties. A 90Sr source is locatedclose to the back of the cuvette so that the betas pene-trate the WbLS and the produced photons are collectedby the 10-inch PMT. The acquisition is triggered off a1-inch, cubic Hamamatsu PMT optically coupled to thebottom of the cuvette facing upwards. The portion of thetrigger PMT front face which is not coupled to the cu-vette is covered with a black mask in order to avoid re-flections on the otherwise exposed PMT glass. The setformed by the cuvette, the trigger PMT and the sourceis held by a 3D-printed black shelf, also attached to thealuminum structure and kept at a fixed distance from theR7081 PMT. Waveforms from both PMTs are digitized bya CAEN V1742.

3.1.2 Calibration and analysis method

The PMT gain is measured using a dataset with a deion-ized water target and the PMT located far from the cu-vette, in order to ensure single-photoelectron events. Thegain is measured by fitting a Gaussian model to the single-PE PMT charge distribution, as described in [9], resultingin (158.0±1.0)V×ns. The NPE is defined as the totalcharge collected in the PMT, corrected by the measuredPMT gain.

The PMT collection efficiency (CE) is measured us-ing the nominal configuration, with the cuvette filled withdeionized water. NPE distributions are compared to theones obtained using the MC and we fit for a CE parame-ter, defined as an additional weight to the modeled PMTquantum efficiency. A CE of 87± 8% is obtained. Thisvalue is used as the nominal value in the light yield mea-surements.

For each WbLS sample, a dataset of 100000 events iscollected. The absolute light yield parameter, Y , is definedin the MC model as the number of photons nγ emitted perunit of quenched deposited energy Eq given by Birk’s law

3

[20], so the total number of photons produced is:

nγ = Y ×Eq. (5)

We obtain Y by performing a binned maximum likeli-hood fit to the observed NPE distributions and minimizing−2logL .

10cm

a) Light yield configuration

Radioactivesource30cm

6.5cm

3cm

5cm

b) Time profile configuration

Figure 1: Experimental configurations used in this anal-ysis. Top: scintillation light yield setup; bottom: scin-tillation time profile configuration. See sec. 3.1.1 andsec. 3.2.1 for further details.

3.1.3 Results

The results for the light yield measurement are shown inTable 1 along with the NPE distributions in Fig. 2. The re-sult for pure LABPPO is included for completeness, andcomparison to previous results, with which it is in goodagreement. According to our simulation, the region be-low 8 PE is populated by betas that do not reach the WbLSvolume and only produce Cherenkov light in the quartz.Thus, the fit is constrained to the region above 8 PEs. Theobserved discrepancies in the low PE region are not con-sidered a cause of systematic uncertainty on this measure-ment given that it is due to events that do not generatescintillation light.

The measured light yield as a function of the LSconcentration is shown in Fig. 3. The LABPPO 2g/Lcase is also shown for completeness and corresponds tothe point at 100% concentration. The measured lightyield for LABPPO is in agreement with that providedby independent measurements [16], which gives confi-dence in our MC model and methods. The three WbLSpoints show a linear behavior with a slope of 127.9 +−17.0ph/MeV/%LS and an intercept value of 108.3 ±51.0ph/MeV. Given that the Cherenkov contribution istaken into account by using the full MC model, the non-zero intercept indicates that the light yield is not linearwith the LS content at very low concentrations (below1%). The measured light yield is noticeably larger thanthat expected from a naive linear extrapolation of theLABPPO 2g/L light yield. The considered sources of sys-tematic uncertainty affecting the light yield measurementare the PMT gain and the PMT CE, whose uncertaintiesare 0.6% and 9%, respectively, as evaluated from the mea-surements in Sec. 3.1.2.

We ran further tests to ensure our measurements werestable and no other systematic effects were present.Datasets for all the liquid mixtures, including water andLABPPO, were retaken by refilling the cuvette and re-coupling it to the trigger PMT in order to test for stability.No significant changes were observed. In addition, in or-der to test the performance of our MC model, datasets us-ing different gamma emitters (60Co, 137Cs, 22Na) in pureLABPPO 2g/L were collected and compared to the MC.The MC prediction using the measured Y was compatiblewith the data in the three cases.

4

WbLS Y[photons/MeV]1% 234±305% 770±7210% 1357±125

LABPPO 2g/L 11076±1004

Table 1: Measured absolute light yield parameters. Sta-tistical and systematic uncertainties are included in thequoted numbers.

#PEs20 40 60

Num

ber

of e

vent

s/bi

n

1

10

210

310

Data

MC

/d.o.f. = 8.0/18 WbLS 1%2χ

#PEs20 40 60

Num

ber

of e

vent

s/bi

n

1

10

210

310

Data

MC

/d.o.f. = 29.8/31 WbLS 5%2χ

#PEs20 40 60

Num

ber

of e

vent

s/bi

n

1

10

210

310

Data

MC

/d.o.f. = 42.4/47 WbLS 10%2χ

Figure 2: Distribution of the number of photoelectrons de-tected for the three different WbLS concentrations, com-pared to the result from the absolute light yield fit. Blackcrosses correspond to data and blue boxes to the best-fitMC. Only statistical uncertainties are shown.

3.2 Time profile measurementThe scintillation time profile of each of the WbLS mix-tures produced by excitation with a radioactive source ischaracterized by obtaining the time profile of the singlephotoelectrons (SPE) detected by a PMT array and fit-ting them using our MC model (see Sec. 2). By using acomplete MC simulation and by calibrating our detector’sSPE response and cable delays, we can account for effectson the time profile due to detector geometry, PMT pulseshape, multi-PE hits or the digitization process. Thus,the intrinsic time profile of the scintillation emission isextracted, unfolding any of those potential systematic ef-fects. Description of the apparatus, calibration, methodsand results can be found below.

3.2.1 Description of the apparatus

A WbLS sample deployed in a cylindrical acrylic con-tainer (referred as the vessel) is excited with betas from a90Sr source located on top (see Fig. 1.b). Light is collectedby an array of 12 1-inch PMTs (Hamamatsu H11934-200)and their waveforms are recorded by a digitizer (CAENV1742). An optically coupled acrylic block is placed be-tween the PMT array and the vessel to act as a light guideand avoid total internal reflections in the vessel. A PMT

100 101 102

WbLS concentration [%]102

103

104

Light

yie

ld [p

h/M

eV]

Linear fit with intercept constrained to zeroLinear fit with intercept floatedLinear referenceMeasured light yield

Figure 3: Measured absolute scintillation light yield as afunction of concentration (red dots, displaying the totaluncertainty), including pure LAB-PPO 2g/L (100% con-centration). Two linear fits to the three WbLS points areshown, one with the intercept constrained to zero (blue)and another one freely floating the intercept (orange). Forreference, a model linear with the pure LABPPO 2/L lightyield and with zero intercept is shown (gray).

5

(referred as the trigger PMT) is optically coupled to theside of the vessel and provides the data acquisition trig-ger. The setup features 4 scintillator panels that provide4π coverage to the vessel against through-going cosmicrays.

3.2.2 Calibration and analysis method

We collect a million triggered events using the 90Srsource. To veto cosmic ray events we reject those eventsin which any of the scintillator panels detect a signal. Wefurther reject events where any of the PMT waveformspresent large fluctuations in the pedestal region [4].

The time of a PE produced in one of the PMTs ofthe array, defined using a constant fraction discrimination(CDF) method at a 20% of the pulse height, is calculatedrelative to the trigger time. For the latter, we use a fixedthreshold approach, defined as the time at which the trig-ger waveform crosses 50% of the height of the SPE pulse,in order to ensure we are computing the time of the firstPE. This provides an optimal time resolution for high lightlevels, as we confirmed experimentally. PMT times arecorrected by photon time of flight and cable delays, previ-ously calibrated as described in [4]. Multi-PE contamina-tion on the PMT array is reduced below 5% by selectinghits with a charge between 0.5 PE and 1 PE. The PMTcharge is provided by integrating the digitized waveformand the PMT gains are calibrated by characterizing SPEcharge distributions, as described in [4].

Using a deionized water target, in order to guaranteea pure Cherenkov emitter, we measured the time resolu-tion to be (855±8) ps FWHM for SPE hits. We validateour MC model by simulating the full setup with the targetfilled with water and comparing it to our obtained reso-lution. The prediction underestimates it by 248 ps, so anadditional smearing is included in the MC as a correction.This small discrepancy is associated with PMT-to-PMTvariations in the transit-time spread and pulse shapes.

The scintillation time profile p(t) is modeled as the sumof two negative exponentials, to represent the decay pro-file, modulated by an exponential rise:

p(t) =1N(1− e−t/τr)

[R1e−t/τ1 +(1−R1)e−t/τ2

], (6)

where τr, τ1 and τ2 are the rise time, short and long decaytimes, respectively; R1 is the fraction of the short time

component over the total and N is chosen so that p(t) isnormalized to unity. This model is commonly used insimilar studies [9, 10] and it has been shown to fit thedata well. The parameters are independently determinedfor each WbLS mixture by implementing this time profilemodel into the MC and fitting the data.

The detected time profile is fit to the MC prediction us-ing a binned maximum likelihood method. We minimize−2logL by iteratively scanning the previously definedparameters. After the minimum has been found, the 1σ

uncertainties are computed by fitting a three-degree poly-nomial to the individual 1-dimensional scans. This un-certainty, σ∗

i , on parameter i does not include the correla-tions, ρi j, with other parameters, j. Given that scanningthe multi-dimensional likelihood space using MC simu-lations is extremely compute-intensive, ρi j are calculatedoff-line using the analytical form of the scintillation timeprofile model Eqn. (6). The total uncertainty, σi, on a pa-rameter i, is given by [1]:

σi = σ∗i ×∏

j(1−ρ

2i j)

−1/2. (7)

This procedure is validated using 100 toy MCs and thefinal scanned values are bias-corrected. The size of thebias is −0.28σ for τ1, −0.09σ for τ2, −0.03σ for R1 and−0.9σ for τr. The large bias in τr is due to its strongcorrelation with the normalization of the Cherenkov com-ponent, which is already reflected in the poor constraintson the measured values. According to this study, the un-certainties on the different parameters are overestimatedby 40% for τr, by 32% for τ1, by 47% for τ2 and by only2% for R1. In order to be conservative, the obtained un-certainties from the fit are not reduced by these factors.

3.2.3 Results

The fit results for the three different WbLS mixtures areshown in Table 2. The time profile distributions for dataand MC, using the best fit parameters, are shown in Fig. 4.

The considered sources of systematic uncertainty arethe hit-time resolution, the uncertainty in the Cherenkovcomponent, the WbLS light yield and the PPO concen-tration. The smearing added to the MC due to the under-estimation of the time resolution discussed in Sec. 3.2.2is included in this measurement. Since the size of theCherenkov component depends on the refractive index of

6

the medium and this has not been measured as a func-tion of the wavelength, we float an extra normalizationparameter for the Cherenkov component and treat it asa nuisance parameter. The WbLS light yield could po-tentially impact the amount of multi-PE hits we observe,biasing our measurement. We compared the MC time dis-tributions obtained for light yield parameters modified by1σ , as measured in Sec. 3.1. No significant variationswere observed, confirming that this effect is negligible inour setup. Potential uncertainties in the reemission prob-ability can affect the shape of the time profile, given thatabsorbed and reemitted photons get delayed. The reemis-sion probability is a function of the effective concentra-tion of PPO that contributes to the scintillator mechanism,which needs further study[10]. Given that this uncertaintycould potentially impact the measured time profile, wecompared MC time distributions modeling several highPPO concentrations, demonstrating a negligible impact.

Our measured time profiles are compatible with thosein [12] in the prompt 5 ns region. This is the relevanttime window for Cherenkov and scintillation separationand also where most of the scintillation light is produced.Beyond that region, our model and the one in [12] diverge,both due to a different analytical form chosen, and lack ofsensitivity to a third exponential decay in our measure-ments. This difference is driven by large statistical uncer-tainties in the tail of the distribution, and our MC-basedfitting approach, which limits the size of the parameterspace (adding a third decay exponential would add an ad-ditional dimension to be scanned, complicating the fit pro-cedure to an extent that was deemed unnecessary givenour limited sensitivity to the long tail).

WbLS 1% 5% 10%τr [ns] 0.00±0.06 0.06±0.11 0.13±0.12τ1 [ns] 2.25±0.15 2.35±0.13 2.70±0.16τ2 [ns] 15.10±7.47 23.21±3.28 27.05±4.20

R1 0.96±0.01 0.94±0.01 0.94±0.01

Table 2: Time profile best fit parameters. From top tobottom: rise time, short decay time, long decay time, andfraction of the short decay time component over the to-tal. The quoted uncertainties include the statistical andsystematic uncertainties.

4 Cherenkov and scintillation sepa-ration with cosmic muons

The CHESS experiment[4, 9] is used in order to quan-tify the separation between Cherenkov and scintillationlight. Using vertical, downward-going cosmic muons wedetect the Cherenkov and scintillation light produced ina target filled with WbLS. The strong directional natureof the Cherenkov light projects ring-like structures in theupward-facing PMT array. By using the PMT hit-time in-formation and the number of prompt PEs, we are able toseparate the populations of PMTs that detect Cherenkov

Time Residuals [ns]0 50 100

Num

ber

of e

vent

s/bi

n

1

10

210

310

410

Data

MC Cherenkov

MC Scintillation

WbLS 1%

Time Residuals [ns]0 50 100

Num

ber

of e

vent

s/bi

n

1

10

210

310

410

Data

MC Cherenkov

MC Scintillation

WbLS 5%

Time Residuals [ns]0 50 100

Num

ber

of e

vent

s/bi

n

1

10

210

310

410

Data

MC Cherenkov

MC Scintillation

WbLS 10%

Figure 4: Scintillation time profile as detected by the PMTarray. Black dots represent data with statistical uncertain-ties, and colored histograms correspond to MC, which isdivided into Cherenkov (blue) and scintillation compo-nents (red). MC includes only statistical uncertainties.

7

light from those that do not, thus imaging clear Cherenkovring structures. This analysis also serves as a validationof our WbLS model in a different configuration, using anindependent source in a different energy regime (cosmicmuons).

4.1 Description of the apparatusTo the geometry described in Sec. 3.2.1, we add two cos-mic tags, one above the vessel and another below thePMT array, aligned vertically with the center of the vesselto provide the trigger for through-going vertical cosmicmuons (see Fig. 5). The original trigger PMT is removedfrom the side of the vessel to avoid possible reflectionsfrom its front face. The geometry of the setup is opti-mized so that a cosmic muon passing vertically down-wards through the vessel projects a distinct Cherenkovring in the middle PMTs while it creates isotropic scintil-lation light detected by the entire PMT array. The data ac-quisition is triggered off the bottom muon tag and a triplecoincidence between the two muon tags and the scintilla-tor panel below the vessel is required offline. This setupis the same as that used in previous studies and furtherdetails can be found in [4, 9].

4.2 Calibrations and analysis methodAfter the triple coincidence is required, we veto nonsingle-muon events (showers, secondary particles in sur-rounding materials, etc.) by requiring that there is no sig-nal in any of the other scintillator panels. Due to produc-tion of secondary particles that travel through the acrylicblock, some residual light is observed in some events inthe PMTs closer to the vertical. This effect is reducedby rejecting events whose total number of detected PEsin those PMTs is abnormally large. Events with unstablepedestals are also rejected. More detail on event selec-tion criteria can be found in [4]. After the event selectionis applied, 179 events are selected for WbLS 1%, 158 forWbLS 5% and 177 for WbLS 10%, for a 5-week exposureper sample.

In order to compare to our MC prediction, the detec-tion efficiency was calibrated for each PMT. The lightcollection of each individual PMT depends on the PMTCE and the quality of the optical coupling. The individualphoton detection efficiency is calibrated for each PMT

of the array by using the scintillation light producedby cosmic muons in a LABPPO 2g/L sample, whoseabsolute light yield is well known – as measured bySNO+ [16] and confirmed here (Sec. 3.1.3). The averagenumber of PEs per PMT is measured and compared tothat predicted by the MC model. A photon detectionefficiency parameter is defined for each PMT as the ratioof the average number of PEs in data over that of MC. Anindependent factor to calibrate the amount of Cherenkovlight detected in the PMT array is introduced to takeinto account systematic uncertainties on the exact ringposition and discrepancies of the refractive index withrespect to the estimation. A single factor that weights theamount of detected Cherenkov light is included in ourMC model. This factor is calibrated using the Cherenkovlight produced by cosmic muons in a water target. Theratio of the average number of PEs between data andMC provides the Cherenkov detection efficiency factor,which corresponds to 0.59 ± 0.04. This calibration isperformed after the individual PMT detection efficiencieshave been measured. None of these calibration factorshave a significant effect on our WbLS characterization

Uppercosmic tag

Lowercosmic tag

Target

Propagationmedium

PMT array

Cosmic muonCosmic muon

Figure 5: Configuration for the Cherenkov and scintil-lation separation study with cosmic muons. The detectedCherenkov rings are expected to lay on the middle PMTs,highlighted in blue. See Sec. 4.1 for a more detailed de-scription.

8

measurements given that we estimate the light yieldwith a completely independent setup Sec. 3.1 and thetime profile measurements are not affected by multi-PEeffects nor by changes in the Cherenkov component, dueto the fact that the Cherenkov normalization is freelyfloated during the parameter scans. Furthermore, it wasconfirmed through simulation that the impact of theseparameters on the time distributions is negligible.

We evaluate the Cherenkov and scintillation separationas a function of two observables: the PE hit-time and thefraction of prompt light.

The time of the first PE in each PMT hit is calculatedusing a fixed threshold of 25% of the SPE pulse height,as measured during our calibration campaign for the in-dividual PMTs. The time residuals are defined as the hittime relative to the event time, calculated as the averageof the first 4 earliest PMT hit times. The hit times arecorrected by cable delays and time of flight. A measure-ment of our time resolution is performed by using cosmicmuons in a water target, given that the fast nature of theCherenkov emission provides a sharp light pulse with awidth under tens of picoseconds. The measured time res-olution is (275±2) ps for data and (262±2) ps for MC.(This is improved over that reported in Sec. 3.2.2 due tothe multi-PE nature of these events). In order to accountfor the slight discrepancy, an 82 ps Gaussian smearing isapplied to the MC time distributions.

The fraction of early light collected in a PMT, Qr, is de-fined as the ratio of charge integrated in a prompt 5 ns win-dow around the event time to the total integrated charge[4]. PMT hits due to pure Cherenkov photons would havea larger Qr value than those due to slower scintillationlight. Given that the SPE PMT pulses for the CHESSPMTs are tens of nanoseconds wide, the value of Qr evenfor prompt photons is smaller than unity, and the exactvalue depends on the precise shape of the PMT wave-form. We model PMT pulses using an analytical formthat describes the average behaviour for all the electronicchannels [4], but small deviations in the pulse shape withrespect to this form can affect the predicted values of Qr.We quantify this effect by calibrating Qr using a datasetof cosmic muons in water, which provides a source ofprompt Cherenkov light. A correction factor is calculatedas the ratio of the average Qr in data over that in MC,which yields a factor of 1.34. This is applied as a correc-

tion to the MC for the WbLS samples.

4.3 ResultsThe hit-time residual distributions are shown in Fig. 6 forthe three WbLS concentrations, broken down by PMT ra-dial position. We expect the Cherenkov ring to fall on themiddle PMTs (in-ring), while the rest would detect onlyscintillation (off-ring). The in-ring PMTs show a hit-timeresidual distribution shifted towards earlier times with re-spect to the off-ring distribution, as expected given thefaster nature of Cherenkov light. The average of the hit-time residuals are also shown in Fig. 6 for each PMT po-sition, displaying clear ring structures.

Qr distributions are presented in Fig. 6, which alsoshow a clear separation between the two PMT popu-lations. By plotting Qr values averaged over events,Cherenkov ring structures can be identified (Fig. 6).Residual disagreement between data and MC is attributedto individual PMT variations in the exact pulse shape, asdiscussed in Sec. 4.2. Since the MC is not used in theevaluation of the separation of Cherenkov and scintilla-tion signals, these discrepancies do not affect the resultsquoted in this paper. For future detectors, this highlightsthe importance of a detailed understanding of PMT pulseshape in such analyses.

Given a time residual value t, we define the Cherenkovdetection efficiency εC(t) as the fraction of in-ring PMThits that occur before that time, t, such that it representsthe fraction of detected Cherenkov hits. The scintillationcontamination fs(t) within the prompt time window is de-fined as the fraction of hits occurring prior to t that areobserved on the off-ring PMTs, such that it represents thescintillation contribution to this selection. We identify atime residual cut tc that maximizes εC(t) while reducingthe scintillation fraction fs(t), by maximizing the figureof merit:

εC × (1− fs). (8)

The same technique is applied to Qr. The Cherenkov de-tection efficiency εC(Qr) is now defined as the fraction ofin-ring PMT hits that fall above Qr, and the scintillationfraction fs(Qr) is defined as the fraction of hits above Qrthat are on off-ring PMTs. An optimal value Qc

r is foundby maximizing the figure of merit.

9

Table 3 shows the εC and fs values for reference cutvalues of t and Qr and for the optimal cut values. Ahigh Cherenkov detection efficiency with a reasonablylow scintillation light fraction is achieved. The non-linearbehavior with concentration is due to the slightly differenttime profile measured for the different mixtures. In par-ticular, we observe that the WbLS 10% time constants arelonger than the other WbLS concentrations, which makesthe Cherenkov and scintillation separation perform simi-larly to the other cases despite the higher light yield.

When compared to pure LAB and LABPPO 2g/L [9],an overall improvement on the Cherenkov and scintilla-tion separation is achieved. The Cherenkov detection ef-ficiency for WbLS is larger than that of pure LABPPO2g/L (70% and 63% for time and Qr respectively), and atthe level of that of pure LAB (83% and 96% for time andQr respectively). The scintillation fraction is smaller thanthat of pure LABPPO 2g/L (36% and 38% for time andQr respectively), demonstrating the potential to select amore pure sample of Cherenkov hits using separation intime and prompt charge.

5 Conclusion and discussion of theresults

We have measured the absolute scintillation light yieldand emission time profile of three mixtures of WbLS atdifferent concentrations of LS in water. Moreover, wehave demonstrated improved performance for separationof Cherenkov and scintillation signals with respect to thatachieved in LABPPO 2g/L for cosmic muons in a small-scale target (centimeters). This publication is also fol-lowed by the release of the first data-driven MC modelof the different characterized WbLS mixtures, compatiblewith the software RAT-PAC[14] and intended for publicuse.

The observables measured by this work comprise sev-eral of the critical factors for determining separation ofCherenkov and scintillation signals, and detector perfor-mance of liquid scintillator detectors at large scales (me-ters to 10s of meters). Other critical parameters for pre-dicting performance at large scales are the absorption andscattering lengths, and dispersion (as determined by therefractive index). Precise evaluation of WbLS perfor-

mance requires a complete understanding of these effectsand extrapolation using a detailed MC model includingdetector size and geometry. This complex study is out ofscope of the current work. The effort is ongoing, and aseparate publication will be prepared. A qualitative dis-cussion of large-scale performance in light of our resultsis included below.

When compared to LABPPO 2g/L, a standard liquidscintillator used in detectors like SNO+ [21], two pointsmust be remarked. First, the measured light yield issmaller than that of pure scintillator, which eases theCherenkov and scintillation separation but reduces thelight production and, thus, the energy resolution. Nev-ertheless, the light yield is larger than expected from anaive scaling of the scintillator content, by a factor be-tween 1.5 and 2.5, which is an advantage for physics top-ics for which energy resolution is important. Second,the extracted WbLS time profiles indicate a faster pho-ton emission rise time and first decay component with re-

TimeLS% 1% 5% 10%

Reference cut 0.2nsεC 75±4% 84±4% 80±4%fs 22±2% 19±2% 25±2%

Optimal cut 0.8ns 0.2ns 0.2nsεC 88±4% 84±4% 80±4%fs 27±2% 19±2% 25±2%

QrLS% 1% 5% 10%

Reference cut 0.1εC 86±4% 85±4% 65±4%fs 30±7% 26±8% 22±10%

Optimal cut 0.1 0.1 0.05εC 86±4% 85±4% 96±5%fs 30±7% 26±8% 45±5%

Table 3: Results of Cherenkov and scintillation separa-tion for cosmic muons and the three different versions ofWbLS for the time residual (top) and Qr (bottom) meth-ods. Values for the Cherenkov detection efficiency εC andthe scintillation contamination fs are shown for a com-mon cut value (reference cut) and for the optimal cut thatmaximizes the figure of merit (Eqn. (8)).

10

WbLS 1%

Hit Time Residuals [ns]-2 0 2 4 6 8 10

Hits

/PM

T/E

vent

/0.2

4[ns

]

-310

-210

-110

PMTs In-Ring DATA MC

PMTs Off-Ring DATA MC

Hit

Tim

e R

esid

ual A

vera

ge [n

s]-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

QRatio0 0.2 0.4 0.6 0.8

Hits

/PM

T/E

vent

/0.0

3

0

0.05

0.1

0.15

0.2

0.25

PMTs In-Ring DATA MC

PMTs Off-Ring DATA MC

QR

atio

Ave

rage

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

WbLS 5%

Hit Time Residuals [ns]-2 0 2 4 6 8 10

Hits

/PM

T/E

vent

/0.2

4[ns

]

-310

-210

-110

PMTs In-Ring DATA MC

PMTs Off-Ring DATA MC

Hit

Tim

e R

esid

ual A

vera

ge [n

s]

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

QRatio0 0.2 0.4 0.6 0.8

Hits

/PM

T/E

vent

/0.0

3

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

PMTs In-Ring DATA MC

PMTs Off-Ring DATA MC

QR

atio

Ave

rage

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

WbLS 10%

Hit Time Residuals [ns]-2 0 2 4 6 8 10

Hits

/PM

T/E

vent

/0.2

4[ns

]

-310

-210

-110

PMTs In-Ring DATA MC

PMTs Off-Ring DATA MC

Hit

Tim

e R

esid

ual A

vera

ge [n

s]

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

QRatio0 0.2 0.4 0.6 0.8

Hits

/PM

T/E

vent

/0.0

3

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

PMTs In-Ring DATA MC

PMTs Off-Ring DATA MC

QR

atio

Ave

rage

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

Figure 6: Cherenkov / scintillation separation in hit-time residuals and charge ratio for (top) 1%, (middle) 5% and(bottom) 10% WbLS. (Left-most): PMT hit-time residual distributions for PMTs in the Cherenkov ring and outsideit; (left-middle): PMT hit-time residuals vs PMT position, averaged across the data set. (Right-middle): Qr (ratio ofthe charge integrated in a 5ns window over the total charge) distributions for PMTs in the Cherenkov ring and outsideit; and (right-most) Qr vs PMT position averaged across the data set.

11

spect to LABPPO 2g/L. A faster scintillation componentis beneficial for vertex resolution, but induces a greateroverlap with the prompt Cherenkov emission peak, result-ing in a less efficient separation. These two points refer tocompeting effects that need to be carefully evaluated forlarge-scale detector scenarios. Given the flexible nature ofthe composition of WbLS, a mixture to ensure an optimalseparation and scintillation light yield for large detectorswill be pursued.

6 AcknowledgementsThis material is based upon work supported by the U.S.Department of Energy, Office of Science, Office of HighEnergy Physics, under Award Number DE-SC0018974.Work conducted at Lawrence Berkeley National Labora-tory was performed under the auspices of the U.S. Depart-ment of Energy under Contract DE-AC02-05CH11231.The work conducted at Brookhaven National Laboratorywas supported by the U.S. Department of Energy undercontract DE-AC02-98CH10886. The project was fundedby the U.S. Department of Energy, National Nuclear Se-curity Administration, Office of Defense Nuclear Nonpro-liferation Research and Development (DNN R&D).

The authors would like to thank the SNO+ collab-oration for providing data on the optical properties ofLAB/PPO, including the light yield, absorption and ree-mission spectra, and refractive index.

References[1] M. Tanabashi et al. “Review of Particle Physics”.

In: Phys. Rev. D 98.3 (2018), p. 030001.[2] C. Aberle et al. “Measuring Directionality in

Double-Beta Decay and Neutrino Interactionswith Kiloton-Scale Scintillation Detectors”. In:JINST 9 (2014), P06012. arXiv: 1307 . 5813[physics.ins-det].

[3] Andrey Elagin et al. “Separating Double-BetaDecay Events from Solar Neutrino Interactionsin a Kiloton-Scale Liquid Scintillator Detec-tor By Fast Timing”. In: Nucl. Instrum. Meth.A849 (2017), pp. 102–111. arXiv: 1609 . 09865[physics.ins-det].

[4] J. Caravaca et al. “Experiment to demonstrate sep-aration of Cherenkov and scintillation signals”. In:Phys. Rev. C 95.5 (2017), p. 055801.

[5] J. R. Alonso et al. “Advanced Scintillator DetectorConcept (ASDC): A Concept Paper on the PhysicsPotential of Water-Based Liquid Scintillator”. In:(2014). arXiv: 1409.5864 [physics.ins-det].

[6] M. Askins et al. “Theia: An advanced optical neu-trino detector”. In: (2019). arXiv: 1911 . 03501[physics.ins-det].

[7] R. Bonventre and G. D. Orebi Gann. “Sensitivity ofa low threshold directional detector to CNO-cyclesolar neutrinos”. In: Eur. Phys. J. C78.6 (2018),p. 435. arXiv: 1803.07109 [physics.ins-det].

[8] Christian Buck and Minfang Yeh. “Metal-loadedorganic scintillators for neutrino physics”. In: J.Phys. G 43.9 (2016), p. 093001. arXiv: 1608 .04897 [physics.ins-det].

[9] J. Caravaca et al. “Cherenkov and ScintillationLight Separation in Organic Liquid Scintillators”.In: Eur. Phys. J. C77.12 (2017), p. 811.

[10] Ziyi Guo et al. “Slow Liquid Scintillator Candi-dates for MeV-scale Neutrino Experiments”. In:Astropart. Phys. 109 (2019), pp. 33–40. arXiv:1708.07781 [physics.ins-det].

[11] M. Yeh et al. “A new water-based liquid scintilla-tor and potential applications”. In: Nucl. Instrum.Meth. A660 (2011), pp. 51–56.

[12] Drew R. Onken et al. “Time response of water-based liquid scintillator from X-ray excitation”. In:Mater. Adv. 1 (1 2020), pp. 71–76.

[13] S. Agostinelli et al. “GEANT4: A Simulationtoolkit”. In: Nucl. Instrum. Meth. A506 (2003),pp. 250–303.

[14] RAT-PAC is an Analysis Tool. https : / / rat .readthedocs.io/en/latest/.

[15] Joao Carlos R. Reis et al. “Refractive Index ofLiquid Mixtures: Theory and Experiment”. In:ChemPhysChem 11.17 (2010), pp. 3722–3733.eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/cphc.201000566.

[16] The SNO+ collaboration. Private communication.[17] International Association for the Properties of Wa-

ter and Steam. http://www.iapws.org/. Re-lease of September 1997.

[18] Robin M. Pope and Edward S. Fry. “Absorptionspectrum (380–700 nm) of pure water. II. Inte-grating cavity measurements”. In: Appl. Opt. 36.33(Nov. 1997), pp. 8710–8723.

[19] Raymond C. Smith and Karen S. Baker. “Opticalproperties of the clearest natural waters (200–800nm)”. In: Appl. Opt. 20.2 (Jan. 1981), pp. 177–184.

12

[20] John B. Birks. The Theory and practice of scintil-lation counting. 1964.

[21] S. Andringa et al. “Current Status and FutureProspects of the SNO+ Experiment”. In: Adv. HighEnergy Phys. 2016 (2016), p. 6194250.

13