MeV-scale performance of water-based and pure liquid scintillator detectors
B. J. Land, Z. Bagdasarian, J. Caravaca, M. Smiley, M. Yeh, G. D. Orebi Gann
MMeV-scale performance of water-based and slow liquid scintillators
B. J. Land,
1, 2, 3
Z. Bagdasarian,
2, 3
J. Caravaca,
2, 3
M. Smiley,
2, 3 and G. D. Orebi Gann
2, 3 University of Pennsylvania, Philadelphia, PA, USA University of California, Berkeley, CA 94720-7300, USA Lawrence Berkeley National Laboratory, CA 94720-8153, USA
This paper presents studies of the performance of water-based liquid scintillator in both 1-kt and50-kt detectors. Performance is evaluated in comparison to both pure water Cherenkov detectorsand a nominal model for pure scintillator detectors. Performance metrics include energy, vertex, andangular resolution, along with a metric for ability to separate the Cherenkov from the scintillationsignal, as being representative of various particle identification capabilities that depend on theCherenkov / scintillation ratio. We also modify the time profile of scintillation light, including boththe rise and decay times, to study the same performance metrics in slow scintillators. We go on tointerpret these results in terms of their impact on certain physics goals, such as solar neutrinos andthe search for Majorana neutrinos. We show that a 50-kt detector would be capable of better than10 (1)% precision on the CNO neutrino flux with a WbLS (pure LS) target, as well as sensitivity intothe normal hierarchy region for Majorana neutrinos, with half life sensitivity of T νββ / > . × years at 90% CL for 10 years of data taking with a Te-loaded target. I. INTRODUCTION
These are exciting times for neutrino physics, with anumber of open questions that can be addressed by next-generation detectors. Advances in technology and inno-vative approaches to detector design can drive the sci-entific reach of these experiments. A hybrid optical neu-trino detector, capable of leveraging both Cherenkov andscintillation signals, offers many potential benefits. Thehigh photon yield of scintillators offers good resolutionand low thresholds, while a clean Cherenkov signal offersring imaging at high energy, and direction resolution atlow energy. The ratio of the two components provides anadditional handle for particle identification that can beused to discriminate background events.One approach to achieving a hybrid detector is to de-ploy water-based liquid scintillator (WbLS) [1], a noveltarget medium that combines water with pure organicscintillator, thus leveraging the benefits of both scintilla-tion and Cherenkov signals in a single detection medium,with the advantage of high optical transparency and,thus, good light collection. There is significant effort inthe community to develop this technology, including tar-get material development [1–12], demonstrations of Cher-enkov light detection from scintillating media [13–16],demonstrations of spectral sorting [17, 18], fast and highprecision photon detector development [19–26], comple-mentary development of reconstruction methods and par-ticle identification techniques [27–33], and developmentof a practical purification system at UC Davis.Many experiments are pursuing this technology for arange of applications, including a potential ton-scale de-ployment at ANNIE at FNAL [34–36], possible kt-scaledeployments at the Advanced Instrumentation Testbed(AIT) facility in the UK [37–39] and in Korea [40], and,ultimately, a large (25–100 kt) detector at the Long Base-line Neutrino Facility, called
Theia . The
Theia programbuilds heavily on early developments by the LENA col- laboration [41]. Such a detector could achieve an incred-ibly broad program of neutrino and rare event physics,including highly competitive sensitivity to long-baselineneutrino studies, astrophysical searches, and even scopeto reach into the normal hierarchy regime for neutrinolessdouble beta decay [42–45].In this paper, we study the low-energy performanceof such a detector for a range of different target mate-rials, and compare the results to that for a pure waterCherenkov detector, and a pure liquid scintillator detec-tor, using linear alkyl benzene (LAB) with 2 g/L of thefluor 2,5-Diphenyloxazole (PPO) as the baseline for com-parison. Properties for the pure LS detector are takenfrom measurements by the SNO+ collaboration [46, 47].We start by considering three WbLS target materials,based on bench-top measurements of properties reportedin [14, 48]. Each cocktail is a combination of water withLAB+PPO, with differing fractions of the organic com-ponent: 1, 5 and 10% concentration by mass. The lightyield for each material is taken from [14], with the timeprofile and emission spectrum taken from [48]. Otherproperties are evaluated based on the composition of thematerial, as described in Sec. II. Measurements of theseparticular WbLS materials demonstrated a very fast tim-ing response: with a rise time consistent with 0.1 ns, anda prompt decay time on the order of 2.5 ns. These mea-surements were confirmed with both x-ray excitation [48]and direct measurements with β and γ sources [14]. Sincethis fast time profile increases the overlap between theprompt Cherenkov and delayed scintillation signals, wealso consider materials in which we delay the scintillationtime profile by some defined amount, to study the impactof a “slow scintillator”, for both pure LS and WbLS. Suchmaterials are under active development [4, 5].Metrics used for these performance studies include theenergy resolution (dominated by photon counting andquenching effects), vertex resolution, direction resolution,and a statistic chosen to represent the separability ofthe Cherenkov and scintillation signals. This is repre- a r X i v : . [ phy s i c s . i n s - d e t ] J u l sentative of low-energy performance capabilities such asparticle identification, which may rely on separating thetwo populations. The final choice of a detector materialfor any particular detector would depend on the physicsgoals, which will place different requirements on each as-pect of detector performance. In all cases, we focus onthe low-energy regime. Performance studies at the highenergies relevant for neutrino beam physics are underway,and will depend on a different combination of factors, somay yield different optimizations.We consider both a kt-scale and a 50-kt scale detector,as being representative of experiments currently underconsideration. We consider four options for photodetec-tors: a “standard” photomultiplier tube (PMT), with atransit time spread (TTS) of 1.6 ns, fast PMTs with a1 ns TTS, faster PMTs with a 500 ps TTS, and veryfast photon detectors, such as large-area picosecond pho-ton detectors (LAPPDs) [49], with a TTS of 70 ps. Ineach case we assume 90% coverage, with a representativequantum efficiency (QE) used for all four models.To understand the impact of the detector capabili-ties studied here, we discuss the impact for several low-energy physics goals, in particular considering scope fora precision measurement of CNO solar neutrinos, andnormal hierarchy sensitivity for neutrinoless double betadecay (NLDBD) [42, 45]. Large-scale scintillator detec-tors such as Borexino [50] and KamLAND-Zen [51] areleaders in the fields of solar neutrinos and searches forNLDBD, respectively, and new scintillator detectors suchas SNO+ [46] and JUNO [52, 53] are taking data or underconstruction. There is much interest in the community inusing new solar neutrino data for precision understand-ing of neutrino properties and behavior, as well as forsolar physics [54]. The proposed Theia experiment hasdiscussed and evaluated the potential of a multi-kiloton,high-coverage WbLS detector for the purposes of solarneutrino detection and NLDBD [42, 44], where the lat-ter would deploy inner containment for an isotope-loadedpure LS target, adapting techniques from SNO+ andKamLAND-Zen. Studies such as those presented herecan help to inform future detector design.Sec. II presents details of the scintillator model used.Sec. III describes the simulation and analysis methods,including the reconstruction algorithms applied. Sec. IVpresents results for performance of the measured WbLScocktails, including photon counting and reconstruc-tion capabilities. Sec. V presents the results for slowscintillators, considering both the pure LS and a 10%WbLS. Sec. VI discusses these performance results inlight of their impact on certain selected physics goals,and Sec. VII concludes.
II. WATER-BASED LIQUID SCINTILLATORMODEL
For Monte Carlo simulation of photon creation andpropagation in WbLS, we use the Geant4-based [55] RAT-PAC framework [56]. Cherenkov photon pro-duction is handled by the default Geant4 model,G4Cerenkov. Rayleigh scattering process is implementedby the module developed by the SNO+ collaboration [47].The GLG4Scint model handles the generation of scintil-lation light, as well as photon absorption and reemission.We utilize the light yield as measured in WbLS (1%,5%,and 10% solutions) in Ref. [14], and scintillation emis-sion spectrum and time profile as taken from Ref. [48].Other inputs to the water-based liquid scintillator opticalmodel have not yet been measured directly, and, hence,are estimated from those of water and LAB+PPO, asdescribed below.
A. Refractive index estimation
In order to estimate the refractive index for WbLS, n ,we use Newton’s formula for the refractive index of liquidmixtures [57]: n = (cid:113) φ labppo n labppo + φ water n water , (1)where φ denotes the volume fraction of a correspond-ing component, while n labppo and n water correspond tothe measured refractive indexes for LAB+PPO [47] andwater [58]. Due to the dominant fraction of water, theWbLS refractive index is very similar to that of pure wa-ter. B. Absorption and scintillation reemission
The absorption coefficient, α , of WbLS depends on themolar concentration, c , of each of the components as: α ( ω ) = c lab (cid:15) lab ( ω ) + c ppo (cid:15) ppo ( ω ) + c water (cid:15) water ( ω ) , (2)where (cid:15) lab , (cid:15) ppo and (cid:15) water are the molar absorption coef-ficients of LAB, PPO [47], and water (taken from Ref. [59]for wavelengths over 380 nm and from Ref. [60] for wave-lengths below 380 nm).A photon absorbed by the scintillator volume has anon-zero probability of being reemitted. This reemissionprocess becomes important at low wavelengths where theabsorption by scintillator is dominant. As a result, pho-tons are shifted to longer wavelengths where the detec-tion probability is higher due to a smaller photon ab-sorption and a greater PMT quantum efficiency. Theprobability p reemi of a component i absorbing a photonof frequency ω is determined as the contribution of thegiven component to the total WbLS absorption coeffi-cient: p reemi ( ω ) = φ i α i ( ω ) /α ( ω ) , (3)where φ i is the volume fraction of component i in WbLS.After a photon is absorbed, it can be reemitted with a59% probability for LAB and an 80% probability for PPO[47], following the primary emission spectrum. C. Scattering length
The Rayleigh scattering length, λ s , is estimated forWbLS as: λ s ( ω ) = (cid:0) φ lab λ − lab ( ω ) + φ water λ − water ( ω ) (cid:1) − , (4)where λ lab and λ water are the scattering lengths for LABand water, respectively, both taken from [47]. It wasnoted that the addition of PPO does not change λ s andthus it is omitted in Eq. 4.The resulting values of both absorption and scatter-ing lengths for WbLS are close to those of pure water.It is possible that this method overestimates the atten-uation lengths, in particular, the scattering, given thecomplex chemical structure and composition of WbLS.A long-arm measurement of WbLS absorption and scat-tering lengths is planned in the near future. III. SIMULATION AND ANALYSIS METHODS
The WbLS models developed in [61], and describedabove, can be used to evaluate the performance of thesematerials in various simulated configurations. Of interestare large, next-generation detectors such as
Theia [42],which could contain tens of kilotons of target material in-strumented with high quantum efficiency photodetectorsat high coverage, and proposed detectors in the range ofone to a few kt, such as AIT [37]. To evaluate these ma-terials, two detector configurations are simulated: a 1-ktdetector and a 50-kt detector, both with 90% coverageof photon detectors as a baseline. The different concen-tration WbLS materials studied in [14], 1%, 5% and 10%WbLS, are simulated and compared to both water andpure (100%) scintillator material LAB+PPO [47].
A. Monte Carlo simulation
Fully simulating next-generation detector sizes instru-mented with 3D models of photon detectors at the de-sired coverage of 90% requires significant computationalresources. This is especially true when studying multiplegeometries, as the simulation typically must be rerun foreach geometry. To avoid this redundancy, RAT-PAC [56]can easily simulate a sufficiently large volume of mate-rial and export the photon tracks to an offline geometryand photon detection simulation. Using this method,2.6-MeV electrons are simulated at the center of a largevolume of target material, isotropic in direction, and theresulting tracks are stored for later processing by a detec-tor geometry model and a photon detector model. Thisenergy is chosen as being representative of a number oflow-energy events of interest, including reactor antineu-trinos, low-energy solar neutrinos, and the end-point ofdouble beta decay for both
Xe and
Te.
1. Detector geometry
Each detector configuration is modeled as a right cylin-der with diameter and height of 10.4 m and 38 m forthe 1-kt and 50-kt sizes, respectively. Specifically, thiscalculation achieves a 1-kt and 50-kt total mass forthe LAB+PPO detector, with slightly modified targetmasses for the other target materials, based on differ-ent densities. The photon tracks from stored events thatare found to intersect with the cylinder representing thedetector boundary are stored as potential detected pho-tons (“hits”) for each event. In this way, the boundaryof each active volume acts as a photon-detecting surfacethat provides all information about each photon to a pho-ton detector model.
2. Photon detection
Photon detectors vary in their probability of detectinga photon as a function of wavelength (the QE) and theirtime resolution (TTS). Four photon detector models areconsidered for each material and geometry:1. “PMT” a modern large-area high-QE PMT suchas an R5912-100 [62] with 34% peak QE and 1.6-nsTTS.2. “FastPMT” a hypothetical PMT with a similar QEbut smaller TTS of 1.0 ns.3. “FasterPMT” a hypothetical PMT again with asimilar QE but even smaller TTS of 500 ps.4. “LAPPD” a next-generation device such as a large-area picosecond photodetector (LAPPD) [49] withsimilar QE but a 70-ps TTS.The same QE is used for all four models, assuming thatfuture LAPPDs can reach comparable QE to existingHamamatsu large-area PMTs.A coverage of 90% using these devices is simulated byaccepting only 90% of potential hits for the event. TheQE is accounted for by randomly accepting hits accordingto the value of the QE curve (shown in Fig. 1 with typi-cal wavelength spectra) at the wavelength of the hit. Forthe selected hits, the intersection position with the ge-ometry model is taken as the detected position. Finally,a normally distributed random number with a width cor-responding to the TTS of the photon detector model isadded to the truth time of the hit to get the detectedtime. These detected hit position and times can then bepassed to reconstruction algorithms for further analysis.
B. Event reconstruction
To evaluate the performance of the different materialsunder different detector configurations, a fitter was devel-oped to reconstruct the initial vertex parameters based
200 300 400 500 600 700 800Wavelength (nm)0510152025303540 Q u a n t u m E ff i c i e n c y ( % ) Quantum Efficiency 0.00.20.40.60.81.01.2 I n t e n s i t y ( A r b . U n i t s ) Cherenkov (1 kt)Scintillation (1 kt)Cherenkov (50 kt)Scintillation (50 kt)
FIG. 1. The quantum efficiency (QE) used for photon detec-tor models considered here (digitized from [62]). Also shownare Cherenkov and scintillation spectra for the 1% WbLS ma-terial in the two detector sizes prior to application of QE.The relative normalization of the spectra have been preserved,with the maximum value normalized to 1.0. on detected hit information. Position and time recon-struction are both aided by the large number of isotropicscintillation photons, while direction reconstruction relieson identification of non-isotropic Cherenkov photons. AsCherenkov photons are prompt with respect to scintilla-tion photons, the reconstruction will first identify promptphotons, and then use them to reconstruct direction in astaged approach. Promptness is defined in terms of thehit time residual t resid distribution.The reconstruction algorithm used here has the follow-ing steps, which are described in detail in the followingsections:Step 1: Position and time are reconstructed using all de-tected hits.Step 2: Reconstructed position and time are used to com-pute the t resid distribution for detected hits.Step 3: Direction is reconstructed using only hits beforesome t cut on the t resid distribution with positionand time fixed.Step 4: Finally, the total number of hits is recorded asan estimate of the energy of the event.The approach is inspired by vertex reconstruction algo-rithms used in the SNO experiment [63]. The algorithmhas been tested and demonstrated to achieve similar posi-tion and direction resolution to SNO for equivalent eventtypes in a SNO-like detector—for example, for 5 MeVelectrons in a SNO-sized vessel, with TTS and photo-coverage set to relevant values (approximately 1.8 ns and55%, respectively) this algorithm achieves 27.4 ◦ angularresolution, compared to the SNO reported value of 27 ◦ .
1. Position and time
Reconstructing vertex position and time can be doneby maximizing the likelihood of t resid, i for each hit i inthe event: t resid, i = ( t i − t ) − | (cid:126)x i − (cid:126)x | cn , (5)where ( (cid:126)x i , t i ) are the position and time of a detected pho-ton, ( (cid:126)x, t ) represents the fitted vertex position and time,and cn is the group velocity typical of a 400-nm photon.This expression includes two important assumptions thatare made to approximate a realistic detection scheme.1. The travel time is calculated assuming a photonwavelength of 400 nm, since for a real detector thewavelength is typically not known. Fig. 1 showsthe expected spectra for both Cherenkov and scin-tillation light.2. Each photon is assumed to travel in a straight line,as photon detectors are typically not aware of theactual path the photon traveled.A result of these assumptions is that dispersion in thematerial will broaden the t resid distribution, as the traveltime will be overestimated (underestimated) for longer(shorter) wavelength photons. Additionally, scattered orreemitted photons will appear later than their true emis-sion time due to ignoring their true path. An example ofa t resid distribution using the true detection times, butwith these approximations, is shown for the 10% WbLSand LAB+PPO material in Fig. 2 for the 1-kt and 50-ktdetector geometries. In plots shown in this paper, the t resid is arbitrarily shifted such that the average t resid of Cherenkov photons across many events is 0 ns. Theintegral of these distributions is the number of detectedphotons per event on average, which highlights both thedifficulty of identifying Cherenkov photons in pure scin-tillators, and their prompt placement in the t resid distri-bution.For each material and detector configuration, a PDFfor t resid of all photons is produced using truth informa-tion from a subset of the simulated events. Reconstruc-tion is then done by minimizing the sum of the negativelogarithm of the likelihood for each hit with a two-stagedapproach: a Nelder-Mead [64] minimization algorithmwith a randomly generated seed is used to explore thelikelihood space and approximate the global minima, fol-lowed by a BFGS [64] minimization algorithm to find thetrue (local) minima using the minima from the previousstep as the seed. This method produces the best estimateof the true t resid distribution for each event, to be usedin the direction fit. Residual distributions are calculatedfor position and time, and fit to Gaussian distributions.Position residuals are fit in a reference frame where thez axis is aligned with the true event direction. The posi-tion resolutions reported here are the quadrature sum ofthe widths in all three dimensions. P h o t o n s / n s
10% WbLS (1 kt) CherenkovScintillationReemissionAll Photons P h o t o n s / n s LAB+PPO (1 kt) CherenkovScintillationReemissionAll Photons P h o t o n s / n s
10% WbLS (50 kt) CherenkovScintillationReemissionAll Photons P h o t o n s / n s LAB+PPO (50 kt) CherenkovScintillationReemissionAll Photons
FIG. 2. True hit time residual distributions for (top left) 10% WbLS in 1-kt detector, (top right) LAB+PPO in 1-kt, (bottomleft) 10% WbLS in 50-kt detector, and (bottom right) LAB+PPO in 50-kt detector. This uses the same QE as the photondetector models, but with zero TTS.
2. Direction
As Cherenkov light has a conical/ring geometry, Cher-enkov hits can be used to infer the event direction. Amethod for doing this is by maximizing the likelihood ofthe cosine of the angle, θ i , between the vector from thereconstructed event position, (cid:126)x , to each detected photonposition, (cid:126)x i , and a hypothesized direction ˆ d :cos θ i = ( (cid:126)x i − (cid:126)x ) · ˆ d | (cid:126)x i − (cid:126)x | . (6)For Cherenkov light, the PDF for this distribution ispeaked at the Cherenkov emission angle, θ c , of the ma-terial. Because non-Cherenkov photons do not carry di-rectional information, they will appear flat in this distri-bution, and will degrade the performance of the fit. Itis beneficial, therefore, to restrict this likelihood maxi-mization to only photons with t resid < t cut for some t cut ,as this should maximize the number of Cherenkov pho-tons relative to other photons. Here, the impact of dis-persion is typically beneficial, as the broad spectrum ofCherenkov light compared to typical scintillation spectraresults in long-wavelength Cherenkov photons appearing earlier in the t resid distribution compared to their trueemission times. We note that a photon detection schemethat can distinguish between long and short wavelengthphotons [18] could further enhance the ability to identifyCherenkov photons.PDFs for the cos θ i distribution are created using sub-sets of the simulated events for many t cut values between-1 ns and 10 ns, and event reconstruction is done for each t cut value for every event. Reconstruction proceeds in thesame way as the position-time minimizing the sum of thenegative logarithms of the likelihood of each selected hitwith a randomly seeded coarse Nelder-Mead [64] search,followed by a BFGS [64] method seeded with the resultof Nelder-Mead to find the best minima. The value cos θ is calculated for each reconstructed direction as ˆ d · ˆ d true ,where ˆ d true is the initial direction of the electron. Thecos θ distribution from each simulated configuration and t cut pair is integrated from cos θ = 1 until the cos θ valuethat contains 68% of events, and this value is defined asthe angular resolution for that pair. Finally, the angularresolution resulting from the t cut with the best angularresolution for each configuration is taken as the angularresolution for that configuration.
3. Energy
The distribution of the total number of hits is fit to aGaussian to determine the mean µ N and standard devi-ation σ N of detected hits for each condition. The frac-tional energy resolution is reported as σ N /µ N . IV. PERFORMANCE OF WATER-BASEDLIQUID SCINTILLATOR IN A LARGE-SCALENEUTRINO DETECTOR
The materials described in Sec. III were simulated inthe two detector geometries (1 kt and 50 kt) and four pho-todetector models (“PMT,” “FastPMT,” “FasterPMT,”and “LAPPD”) described in the same section. Between10,000 and 100,000 events were simulated for each ma-terial, with fewer events for the pure LS due to the highphoton counts (and accordingly slower simulation times).The following sections explore the true MC informationprovided by those simulations, as well as presenting thereconstruction results for all cases.
A. Photon population statistics
Roughly speaking, energy resolution is limited by thetotal number of detected photons, position and time res-olution are limited by the number of direct photons (notabsorbed and reemitted, scattered, or reflected), and di-rection resolution is limited by the number of Cherenkovphotons and how visible they are within the brighter scin-tillation signal. The total population of photons can bebroken down into the following categories:1.
Cherenkov photons, which were not absorbed andreemitted by the scintillator.2.
Scintillation photons, which were not absorbed andreemitted by the scintillator.3.
Reemitted photons, regardless of their origin.These populations are shown in Fig. 3 for the materialsand detector sizes considered here. Since each consideredphoton detector model has the same QE and coverage,the populations are the same in each case.Higher scintillator fractions are very advantageousfrom an energy resolution perspective, having many moretotal photons. The same is true from the perspective ofposition and time resolution in a 1-kt detector. For alarger 50-kt detector, the population of reemitted pho-tons for LAB+PPO is greater than the scintillation pop-ulation, hinting that attenuation and reemission in purescintillators may be disadvantageous in larger detectors.Interestingly, simulations indicate that there are moreCherenkov photons detected in LAB+PPO than in theother materials, though fraction of detected hits that areCherenkov is much smaller. The larger refractive index of Scintillator Fraction10 P h o t o n s / E v e n t Cherenkov (50 kt)Scintillation (50 kt)Reemission (50 kt)Total (50 kt)Cherenkov (1 kt)Scintillation (1 kt)Reemission (1 kt)Total (1 kt)
FIG. 3. The number of detected photons for 2.6-MeV elec-trons simulated at the center of two detector geometries (50-ktand 1-kt) differing in size. These photon counts are shown asa function of material scintillator fraction. Water is artificiallyplotted at 10 − (due to log scale). LAB+PPO results in more primary Cherenkov photonscompared to water, and this competes with the greaterabsorption in pure scintillators. The lower scintillatorfractions have refractive indexes closer to that of water,and the additional absorption compared to water resultsin fewer Cherenkov photons detected.
B. In-ring photon counting
Without applying reconstruction algorithms, one caninspect the truth information for the detected hits tounderstand their origins and time distributions. Of in-terest here is how discernible the Cherenkov photons are,and how well they may be identified against a scintilla-tion background. Since Cherenkov photons are emittedat a particular angle θ c with respect to the track of thecharged particle, it is instructive to see how many hitsare detected in the region θ c ± δ (“in-ring”) with respectto the event direction. Further, since Cherenkov photonsare prompt with respect to scintillation photons, it is in-structive to see these populations as a function of howearly they arrive. As in the reconstruction algorithm,this is defined in terms of the hit time residual, t resid ,where smaller t resid values are more prompt.Fig. 4 shows the number of Cherenkov and other (scin-tillation and re-emitted) photons for photons with cos θ satisfying θ c ± ◦ using true detected times (TTS =0) and true origins, but including the effect of photode-tector coverage and QE, as a function of a promptnesscut on t resid . Of particular note is that there are more“in-ring” Cherenkov photons than other photons for suf-ficiently prompt t resid cuts for all materials using truthinformation.With the number of in-ring Cherenkov photons definedas S and the number of in-ring other-photons defined as B , a single metric, S/ (cid:112) ( S + B ), for the significance ofthe Cherenkov photons as a function of a prompt cut on t resid is shown in Fig. 5. The larger this significance, theeasier it should be to identify the Cherenkov topology ontop of the isotropic scintillation background. The highsignificance at early times in the LAB+PPO materialcan be understood as a combination of dispersive effects,differences in scintillation time profile, and the higherrefractive index in this material relative to the WbLSmaterials in general. Dispersion separates the narrowscintillation spectrum from the longer-wavelength tail ofthe Cherenkov photons in large detectors, pushing thelonger-wavelength Cherenkov hits earlier, while a higherrefractive index results in more Cherenkov production.The LAB+PPO material has both a larger refractive in-dex and more dispersion than the other materials. Addi-tionally, the measurements from [14] show the time pro-file of WbLS materials is faster than LAB+PPO, bring-ing scintillation light earlier in those materials. However,the greatest significance of Cherenkov detection in scin-tillating materials is achieved in WbLS, for slightly latercuts on t resid . C. Reconstruction results
Inspecting the truth information provides a detailedunderstanding of the information available, however, totruly evaluate these materials, it is necessary to apply re-construction algorithms and evaluate the impact on po-sition, time, and direction reconstruction. This is doneusing the reconstruction algorithm described in Sec. IIIand the results are shown in Fig. 6. An example viewof the fit residuals for LAB+PPO with a 1 . t resid distribution, are greaterin the larger geometry. In particular, the better trans-parency of WbLS compared to LAB+PPO is evidentin the relatively poorer position resolution seen withLAB+PPO when compared to 10% WbLS in the 50-ktdetector. Position and time resolutions unsurprisinglyimprove with the reduction in TTS from the PMT modelto the LAPPD model.For direction reconstruction, the water material actsas an excellent baseline with best resolution, having onlyCherenkov hits and excellent transparency. The addi-tional scintillation light from the WbLS materials de-grades this resolution by approximately a factor of two resid cut (ns)10 I n - R i n g P h o t o n s / E v e n t resid cut (ns)10 I n - R i n g P h o t o n s / E v e n t
50 kt Water1% WbLS5% WbLS10% WbLSLAB+PPOCherenkovOther
FIG. 4. The number of “in-ring” (see text) photons per eventdetermined using truth information from 2.6 MeV electronssimulated at the center of two detector geometries (top) 1kt and (bottom) 50 kt. The number of photons is shown asa function of a cut on the t resid distribution, selecting forprompt photons. Cherenkov photons are shown in solid lines,with all other photons shown with dashed lines. The coloredlegend applies to both Cherenkov and other photons. in the 1-kt detector, and by less than 50% in the 50-ktdetector for 10% WbLS. For LAB+PPO, dispersion (es-pecially in the 50-kt detector) and the relatively slowertime profile results in enhanced t resid separation betweenCherenkov and scintillation photons, enabling compara-ble or better angular resolution than the WbLS materials.Notably, the LAPPD model has sufficient time resolutionto easily identify a pure population of prompt Cherenkovphotons in LAB+PPO resulting from dispersion, allow-ing direction reconstruction comparable to water. This isnot seen with the PMT model, which lacks the time reso-lution to resolve this population. This indicates that thedispersion of a pure scintillator is a beneficial quality fordirection reconstruction, and that the faster timing pro-files of the WbLS materials relative to LAB+PPO maybe a hindrance to accurate direction reconstruction. Theformer point may be difficult to address in WbLS, giventhat the refractive index is very close to that of water resid cut (ns)012345678 S / S + B Water1% WbLS5% WbLS10% WbLSLAB+PPO resid cut (ns)012345678 S / S + B
50 kt
Water1% WbLS5% WbLS10% WbLSLAB+PPO
FIG. 5. With S defined as Cherenkov photons and B definedas other photons, these figures plot S/ √ S + B , or the signif-icance of the population of “in-ring” Cherenkov photons, forthe data shown in Fig. 4, with the two detector geometries(top) 1 kt and (bottom) 50 kt. and it is hardly tunable without significantly altering thematerial. However, the time profiles of liquid scintilla-tors can be adjusted [4, 5], and this is explored in thefollowing section. V. IMPACT OF SCINTILLATION TIMEPROFILE IN A LARGE-SCALE NEUTRINODETECTOR
Two properties are explored here: the rise time of theprofile, τ r , and a single decay constant, τ , using theform: p ( t ) = 1 N (1 − e − t/τ r ) e − t/τ , (7)where N is a normalization constant. Both theLAB+PPO and 10% WbLS materials have their timeprofiles adjusted, and reconstruction metrics are shownusing the methodology described in Sec. III. We consider Scintillator Fraction0102030405060708090 1 kt
PMTFastPMTFasterPMTLAPPDAngular Res. (deg)Position Res. (cm)Energy Res. (%) Scintillator Fraction0102030405060708090 50 kt
PMTFastPMTFasterPMTLAPPDAngular Res. (deg)Position Res. (cm)Energy Res. (%)
FIG. 6. Reconstruction resolutions of 2.6 MeV electronssimulated at the center of two detector geometries (top) 1-kt and (bottom) 50-kt, differing in size, and four photondetector models (“PMT,” “FastPMT,” “FasterPMT,” and“LAPPD”), differing in TTS. These resolutions are shown asa function of scintillator fraction. Water is artificially plottedat 10 − (due to log scale). Angular resolution is shown forthe best prompt cut. See legend for units. both a scan of the decay constant for two chosen risetimes, and a scan of the rise time for two chosen decaytimes. In all cases, all other properties of the materials(light yield, refractive index, absorption and scattering,emission) are kept constant at the values presented inSec. II. This allows us to decouple the effect of the timeprofile from other properties of the scintillator, whichmay be useful input for guiding future material devel-opment. A. Decay time
The decay constant is scanned from 2.5 ns (typicalof current WbLS) to 10 ns (typical of slow scintilla-tors [4, 5]), and the simulation and reconstruction meth-
400 200 0 200 400Fit Position (mm)0100200300400 x = 0.1±0.1=7.6±0.2 cmy = 0.2±0.1=8.4±0.2 cmz =7.5±0.1=7.7±0.2 cm t =0.023±0.004=0.231±0.006 ns =61.3 deg N =2064.1±0.7=73.4±1.1 nhits FIG. 7. The upper left panel shows the position fit residu-als in three dimensions, where Z is always aligned with theinitial event direction. The top right panel shows the fittedtime residuals. The cos θ fitted event direction distributionis in the bottom left, with the bottom right being the totalnumber of detected photons, from which the energy resolu-tion is calculated. This is shown for the LAB+PPO materialin the 1 kt detector geometry using the “PMT” photon de-tector model and a t resid < . ods described in Sec. III are used for each combination.This scan is repeated for two choices of rise time: a fastrise time of 100 ps is used, characteristic of the WbLScocktails explored in this paper, and a slow rise time of1 ns, more representative of LAB+PPO.As before, this is done for 2.6-MeV electrons withboth the 1-kt and 50-kt detector geometries. Only theLAPPD photon detector model is explored here, to sim-plify the presentation of results. Resolution metrics arepresented for position and direction with the 10% WbLSand LAB+PPO materials in Fig. 8. Energy resolution isunaffected by changes to the time profile.Slower decay constants in 10% WbLS appear to im-prove angular resolution quite significantly in the 1-ktgeometry, more so for the faster rise time, but degradethe resolution in the 50-kt geometry. This is likely dueto a difference in the impact of dispersion at the differentscales (see Fig. 2). If dispersion has already separated theCherenkov and scintillation photons sufficiently in time,there is no further advantage to slowing the scintillator.Since, in this limit, slowing the decay constants reducesposition and time resolution without better separatingthe Cherenkov population, the angular resolution is de-graded.Notably for LAB+PPO the effects are small, and scan-ning the decay time has little effect on the angular reso-lution, and only a small impact on vertex reconstructionover this range. This suggests that the time profile ofLAB+PPO may be close to optimal for these purposes. Simulated hit time residuals in Fig. 2 show that the un-modified LAB+PPO material has a clear prompt Cher-enkov population in the 50-kt detector (c.f. 10% WbLS),which is primarily due to dispersion, and is the dominantfactor in the good performance of LAB+PPO. B. Rise time
The rise time is scanned for values from 100 ps to 1 ns,for both a 2.5 ns and 5 ns decay time, characteristic ofWbLS and LAB+PPO, respectively. As before, this isdone for 2.6-MeV electrons with both the 1-kt and 50-ktdetector geometries. Results are shown in Fig. 9.In all cases, slowing the rise time improves the angularresolution, but slightly degrades the position and timeresolution. Slower rise times in 10% WbLS degrade theposition and time resolution more than in the LAB+PPOmaterial. 10% WbLS demonstrates significant gains inangular resolution for slower time constants, though forthe same position or time resolution, LAB+PPO resultsin better angular resolution.
VI. IMPACT FOR PHYSICS REACH
We now briefly examine how the energy and angularresolutions evaluated in the previous sections affect thecapability for rejection of the B solar neutrino back-ground in NLDBD searches, and identification of signalevents for CNO solar neutrino detection. In both cases,identification (as either signal or background) of the di-rectional solar neutrino events is the capability understudy.In order to do so, we again make use of the RAT-PACframework [56], including the neutrino-electron elasticscattering generator and the radioactive decay generatorused by SNO [63] and SNO+ [65] as well as an imple-mentation of Decay0 [66]. In simulation, the neutrino-electron elastic scattering differential cross section [67] isweighted by the neutrino energy spectrum [68] for thedifferent fluxes from the Sun and then sampled in outgo-ing electron energy and scattering angle, for both ν e and ν µ . Solar neutrino fluxes are taken from [69]. The decayenergy spectra are also found for various backgroundsassociated with the CNO energy region of interest. Thesolar neutrino interactions and decays are then simulatedaccordingly to extract the expected energy deposition inthe target materials under consideration. After the sim-ulation, solar neutrino event samples are weighted fol-lowing the survival probability calculated from the B16GS98 Standard Solar Model [70].The extracted angular resolution parameters fromSecs. IV and V are used to smear the scattering anglefor solar neutrino events using a functional form takenfrom [44], while radioactive and cosmogenic backgroundevents, as well as double beta decay events, are assumedto be isotropic.0 r = 0.1 ns)
10% WbLS Position Res (cm)LAB+PPO Position Res (cm)10% WbLS Ang Res (best cut; deg)LAB+PPO Ang Res (best cut; deg) r = 1.0 ns)
10% WbLS Position Res (cm)LAB+PPO Position Res (cm)10% WbLS Ang Res (best cut; deg)LAB+PPO Ang Res (best cut; deg) r = 0.1 ns)
10% WbLS Position Res (cm)LAB+PPO Position Res (cm)10% WbLS Ang Res (best cut; deg)LAB+PPO Ang Res (best cut; deg) r = 1.0 ns)
10% WbLS Position Res (cm)LAB+PPO Position Res (cm)10% WbLS Ang Res (best cut; deg)LAB+PPO Ang Res (best cut; deg)
FIG. 8. Reconstruction resolutions for a scan of the scintillation decay time with a rise time of (left) 100 ps and (right) 1 nsin the (top) 1-kt detector geometry and (bottom) 50-kt detector geometry. Results are shown for the LAPPD photon detectormodel for the 10% WbLS and LAB+PPO materials. Angular resolution is shown for the best prompt cut. See legend for units.
A. NLDBD sensitivity
For the NLDBD study, we consider LAB+PPO loadedwith 5% natural Te (34.1%
Te), and assume the ex-pected 3% / √ E energy resolution from [42], since theisotope-loaded scintillator will behave differently fromthose studied here. We make the same assumptionsabout location and background rates as in the previousstudy. The purpose of this study is to explore the im-pact of the angular resolutions determined in Sec. IV.No assumption on angular resolution was directly madein [42], so we use the angular resolution found here forunloaded scintillator to extend the previous analysis, asbeing representative of reasonably achievable time pro-files. Energy cuts are applied to restrict the study tothe 0 νββ region of interest for Te, as outlined in [42].We further apply cuts as a function of reconstructed di-rection relative to the Sun, cos θ (cid:12) , in order to reducethe background from directional B solar neutrinos. Thefraction of ν e and ν µ samples for B neutrinos surviving these analysis cuts are scaled according to expected eventrates on LAB+PPO in order to maintain the correct ra-tio of ν e and ν µ interactions and properly calculate theoverall efficiency for rejecting solar neutrino backgroundevents and accepting isotropic events such as radioactivedecays or 0 νββ .The efficiencies for the cut values are then propa-gated through the box analysis procedure of [42] to se-lect an optimal cut that yields the best sensitivity. Toquote an example, we find an expected sensitivity of T νββ / > . × years at 90% CL in the 50-kt,LAPPD-instrumented pure LAB+PPO detector with de-cay time of 2 . . m ββ < . − . ◦ . This result is achieved by cuttingon a solar angle corresponding to cos θ (cid:12) = 0 .
5, which re-jects over 60% of the B background while keeping 75%of the signal. This increases confidence in assumptions of1 = 2.5 ns)
10% WbLS Position Res (cm)LAB+PPO Position Res (cm)10% WbLS Ang Res (best cut; deg)LAB+PPO Ang Res (best cut; deg) = 5.0 ns)
10% WbLS Position Res (cm)LAB+PPO Position Res (cm)10% WbLS Ang Res (best cut; deg)LAB+PPO Ang Res (best cut; deg) = 2.5 ns)
10% WbLS Position Res (cm)LAB+PPO Position Res (cm)10% WbLS Ang Res (best cut; deg)LAB+PPO Ang Res (best cut; deg) = 5.0 ns)
10% WbLS Position Res (cm)LAB+PPO Position Res (cm)10% WbLS Ang Res (best cut; deg)LAB+PPO Ang Res (best cut; deg)
FIG. 9. Reconstruction resolutions when the scintillation rise time is scanned for a decay time of (left) 2.5 ns and (right) 5.0ns in the (top) 1-kt detector geometry and (bottom) 50-kt detector geometry. This is done using the LAPPD photon detectorfor the 10% WbLS and LAB+PPO materials. Angular resolution is shown for the best prompt cut. See legend for units. rejection capability used in [42]. Notably, improving theangular resolution to 30 ◦ and performing the same anal-ysis does not yield changes to sensitivity to the leadingdecimal.Several other configurations for the 50-kt, LAPPD de-tector give results with similar sensitivity, to this preci-sion. We see that the impact of scanning the decay timefor values from 2.5 to 10 ns for LAB+PPO changes thesensitivity by less than 0 . × years, and the sensi-tivity improves for slower rise times, but the impact ofthe change from a rise time of 100 ps to 1 ns is less than0 . × years. As such variation of the decay and risetime of the scintillation time profile at the scale examinedare not thought to have a large impact on sensitivity toNLDBD. It should be noted that this conclusion is spe-cific to our particular choice of direction reconstructionmethodology, and conclusions may differ for other ap-proaches. B. Precision CNO measurement
We also evaluate scenarios for CNO solar neutrino de-tection in a manner akin to the large-scale WbLS de-tector studies presented in [44] and [42]. We make thesame assumptions about location and background ratesas in those studies. Instead of the hit-based lookup recon-struction scheme applied in those studies, we employ aGaussian smearing based on the expected number of hits,as determined in Sec. IV. Since quenching effects are fullysimulated, we take only the part of the width that is dueto photon counting, so as not to double count that effect.The resolution is scaled with energy according to photonstatistics. The rest of the fitting procedure remains thesame as that described in the mentioned analyses, thoughwe consider the use of a constraint on the pep flux at 1.4%from the global analysis of [73], which leverages the infor-mation afforded by the full pp -chain and solar luminosityon experimental data. This constraint is considered in[74, 75].2We find the expected energy spectra in different LS andWbLS detector configurations according to the energyresolution parameters extracted, using photon statisticsonly, so as not to double count the uncertainty due toquenching, which is fully simulated. Since the angularresolution evaluated at 2.6 MeV is expected to be muchfiner than at energies more relevant to the CNO search,for this study, we instead use resolution values deter-mined using simulated electrons at 1.0 MeV. At this en-ergy, we find that in the 50 kt, LAPPD-instrumented de-tector, the angular resolution achieved by the fitter is 70 ◦ for 1% WbLS and 65 ◦ for LAB+PPO, as opposed to 40 ◦ and 36 ◦ respectively at 2.6 MeV. The energy resolutionis assumed to vary ∝ / √ E and the angular resolutionis assumed to be flat. This does not fully incorporateexpected improvements in resolution at higher energies,and degradation at lower energies. A more sophisticatedstudy implementing the full energy dependence is under-way. This result is intended to guide the reader as to thecapabilities of this style of detector. Energy cuts are ap-plied to the CNO solar neutrino fit region, following theapproach in [42]. We consider a threshold of 0.6 MeV inall cases.We find that in 5 years of data taking, the CNO fluxcould be determined to a relative uncertainty of 18% (8%)in the 50-kt, LAPPD-instrumented 10% WbLS detectorwhen the pep flux is unconstrained (constrained to 1.4%),and to 1% in the same detector filled with LAB+PPO,with the pep flux either constrained or unconstrained. Wenote that the result for the pep -constrained case is notvery sensitive to the fraction of scintillator in WbLS (1–10% perform similarly) whereas in the pep -unconstrainedcase the performance degrades with reduced scintillatorfraction. This is understood because the angular resolu-tion is found to be similar for different WbLS materials at1 MeV (approximately 70 ◦ ), so the light yield becomesthe critical component in determining performance. Amore comprehensive study of these effects will be forth-coming in a future publication. VII. CONCLUSIONS
In this paper we have considered the low-energy per-formance of detectors ranging from 1- to 50-kt in size,with a range of target materials. We focus on new mea-surements of WbLS, and their impacts on detector per-formance, but consider both pure water and pure scin-tillator (LAB+PPO) detectors for comparison. We alsoconsider the impact of slowing the scintillation light inboth the pure LS and the WbLS. We consider four mod-els for photon detectors, with time resolution of 1.6 ns,1 ns, 500 ps, and 70 ps. We study detector performancein terms of energy, vertex, and angular resolution, and goon to the interpret the results in terms of sensitivity tothe CNO solar neutrino flux, and a search for NLDBD.Performance is determined by the number of detectedphotons, and their distribution in both time and space. Scintillator Fraction051015202530 R e l a t i v e un c e r t a i n t y t o t h e C N O f l u x ( % ) pep unconstrained pep constrained to 1.4% FIG. 10. Precision achieved for a measurement of the CNOflux in a 50-kt detector, as a function of the percentage ofLS in the target material, where a value of 10 refers toLAB/PPO. Detector performance is based on that found inSec. IV, assuming the as-measured properties of WbLS andLAB/PPO, without considering possible delays to the scintil-lation profile. These results use the LAPPD photon detectormodel. This depends on both the generated photon distribu-tions, and on effects due to optical propagation. Dueto the higher refractive index, more Cherenkov photonsare generated in pure scintillator than in water or WbLS,which competes with increased absorption and scatteringin this material. Effects of absorption and reemission canbe seen in the large detector, where more reemitted pho-tons are detected than direct scintillation photons.We evaluate energy resolution using the width of thedetected hit distribution. As expected, this increaseswith fraction of scintillator in the target, with minimalimpact from the photon detector model.We employ a likelihood-based evaluation of vertex anddirection reconstruction. The scintillation component ofWbLS improves the vertex resolution but degrades theangular resolution relative to pure water. The fastertime profile of WbLS compared to LAB+PPO makes theidentification of the Cherenkov population more challeng-ing, thus hindering direction reconstruction. Dispersioneffects play a significant role in the ability to separateCherenkov photons, particularly in the larger detector.We see that the impact of faster timing photon detec-tors on low-energy reconstruction performance is impor-tant in the larger detector size in order to fully leveragethis effect for reconstruction. The higher refractive indexof LAB+PPO increases the effects of dispersion for thismaterial. The optimal low-energy angular resolution ina scintillating detector is achieved for LAB+PPO, un-der the assumption of 70-ps time resolution. For timeresolutions of 1 ns or worse, water and WbLS performbetter. The difference in performance between WbLSand LAB+PPO is much less significant in the larger de-tector, where 5% and 10% WbLS perform similarly to3LAB+PPO. Differences are largely due to the fast timeprofile of WbLS and lower refractive index. It is worthnoting that studies of direction reconstruction at high en-ergies may yield different conclusions, given much higherphoton statistics.The fast time profile of WbLS motivated considerationof delaying the time profile, to understand the impact ondetector performance. Slow scintillators are under activedevelopment, in part for their potential to offer improvedangular resolution for low-energy events. This possibilitywas studied for both 10% WbLS and for LAB+PPO. Weobserve minimal impact on either position or directionreconstruction for LAB+PPO, but the angular resolu-tion of WbLS can be significantly improved by slowingthe scintillation light, to that equivalent to LAB+PPOor even slower, with relatively small impact on vertexresolution.We consider the impact of the 50-kt detector perfor-mance on both CNO solar neutrino detection, and po-tential for deployment of a containment vessel of Te-loaded LAB+PPO in a larger WbLS detector, for asearch for Majorana neutrinos via NLDBD. We find sen-sitivity to the CNO neutrino flux of better than 20%under conservative assumptions with no constraint onthe pep flux, better than 10% in a lightly loaded WbLSdetector when considering a constraint on the pep flux,and 1% for a pure LAB+PPO detector, instrumentedwith fast photon detectors. We find a half life sensi-tivity of T νββ / > . × years at 90% CL for 10years of data taking, which equates to a mass limit of m ββ < . − . ACKNOWLEDGMENTS
The authors would like to thank Minfang Yeh, TannerKaptanoglu, Michael Wurm, Josh Klein, Bob Svoboda,Matthew Malek and Adam Bernstein for useful discus-sions. The authors would like to thank the SNO+ col-laboration for providing a number of simulation modules,as well as data on the optical properties of LAB/PPO,including the light yield, absorption and reemission spec-tra, and refractive index.This material is based upon work supported by theDirector, Office of Science, of the U.S. Department ofEnergy under Contract No. DE-AC02-05CH11231 Theproject was funded by the U.S. Department of Energy,National Nuclear Security Administration, Office of De-fense Nuclear Nonproliferation Research and Develop-ment (DNN R&D). This material is based upon worksupported by the U.S. Department of Energy, Office ofScience, Office of High Energy Physics, under AwardNumber DE-SC0018974 [1] M. Yeh, S. Hans, W. Beriguete, R. Rosero, L. Hu, R. L.Hahn, M. V. Diwan, D. E. Jaffe, S. H. Kettell, andL. Littenberg, Nucl. Instrum. Meth. A , 51 (2011).[2] L. Bignell, D. Beznosko, M. Diwan, S. Hans, D. Jaffe,S. Kettell, R. Rosero, H. Themann, B. Viren, E. Worces-ter, M. Yeh, and C. Zhang, J. Instrum. , P12009(2015).[3] C. Buck and M. Yeh, J. Phys. G , 093001 (2016).[4] Z. Guo, M. Yeh, R. Zhang, D.-W. Cao, M. Qi, Z. Wang,and S. Chen, Astroparticle Physics , 33 (2019).[5] S. D. Biller, E. J. Leming, and J. L. Paton, Nucl. In-strum. Meth. A , 164106 (2020).[6] E. Graham, D. Gooding, J. Gruszko, C. Grant,B. Naranjo, and L. Winslow, J. Instrum. , P11024(2019).[7] C. Aberle, J. J. Li, S. Weiss, and L. Winslow, J. Instrum. , P10015 (2013).[8] T. Marrodn Undagoitia, F. von Feilitzsch, L. Oberauer,W. Potzel, A. Ulrich, J. Winter, and M. Wurm, Reviewof Scientific Instruments , 043301 (2009).[9] J. Ashenfelter et al. , J. Instrum. , P03026 (2019).[10] J. Cumming, S. Hans, and M. Yeh, Nucl. Instrum. Meth.A , 1 (2019). [11] A. Abusleme et al. (Daya Bay and JUNO), “Opti-mization of the juno liquid scintillator compositionusing a daya bay antineutrino detector,” (2020),arXiv:2007.00314 [physics.ins-det].[12] X. Zhou, Q. Liu, M. Wurm, Q. Zhang, Y. Ding, Z. Zhang,Y. Zheng, L. Zhou, J. Cao, and Y. Wang, Review ofScientific Instruments , 073310 (2015).[13] J. Caravaca, F. B. Descamps, B. J. Land, M. Yeh, andG. D. Orebi Gann, Eur. Phys. J. C , 811 (2017).[14] J. Caravaca, B. Land, M. Yeh, and G. Orebi Gann,“Characterization of water-based liquid scintillatorfor Cherenkov and scintillation separation,” (2020),arXiv:2006.00173 [physics.ins-det].[15] J. Gruszko, B. Naranjo, B. Daniel, A. Elagin, D. Good-ing, C. Grant, J. Ouellet, and L. Winslow, J. Instrum. , P02005 (2019).[16] M. Li, Z. Guo, M. Yeh, Z. Wang, and S. Chen, Nucl.Instrum. Meth. A , 303 (2016).[17] T. Kaptanoglu, M. Luo, and J. Klein, J. Instrum. ,T05001 (2019).[18] T. Kaptanoglu, M. Luo, B. Land, A. Bacon, and J. R.Klein, Phys. Rev. D , 072002 (2020).[19] A. Lyashenko et al. , Nucl. Instrum. Meth. A , 162834 (2020).[20] B. W. Adams, A. Elagin, H. J. Frisch, R. Obaid,E. Oberla, A. Vostrikov, R. G. Wagner, J. Wang, andM. Wetstein, Nucl. Instrum. Meth. A , 1 (2015).[21] E. Ramberg (PSEC), in Technology and instrumentationin particle physics. Proceedings, 1st International Con-ference, TIPP09, Tsukuba, Japan, March 12-17, 2009 ,Vol. 623 (2010) pp. 316–317.[22] E. Oberla and H. J. Frisch, Nucl. Instrum. Meth. A ,19 (2016).[23] O. H. W. Siegmund et al. , in
Proceedings, 2011 IEEENuclear Science Symposium and Medical Imaging Con-ference (NSS/MIC 2011) (2011) pp. 2063–2070.[24] N. Barros, T. Kaptanoglu, B. Kimelman, J. Klein,E. Moore, J. Nguyen, K. Stavreva, and R. Svoboda,Nucl. Instrum. Meth. A , 15 (2017).[25] J. Brack et al. , Nucl. Instrum. Meth. A (2012),10.1016/j.nima.2013.02.022.[26] J. Dalmasson, G. Gratta, A. Jamil, S. Kravitz, M. Malek,K. Wells, J. Bentley, S. Steven, and J. Su, Phys. Rev. D , 052006 (2018).[27] B. S. Wonsak et al. , J. Instrum. , P07005 (2018).[28] B. Wonsak et al. , in Solar Neutrinos: Proceedings of the5th International Solar Neutrino Conference , edited byM. Meyer and K. Zuber (World Scientific, Singapore,2019) pp. 445–463.[29] J. Dunger and S. D. Biller, Nucl. Instrum. Meth. A ,162420 (2019).[30] C. Aberle, A. Elagin, H. J. Frisch, M. Wetstein, andL. Winslow, J. Instrum. , P06012 (2014).[31] A. Elagin, H. Frisch, B. Naranjo, J. Ouellet, L. Winslow,and T. Wongjirad, Nucl. Instrum. Meth. A , 102(2017).[32] R. Jiang and A. Elagin, “Space-Time Discriminant toSeparate Double-Beta Decay from B Solar Neutri-nos in Liquid Scintillator,” (2019), arXiv:1902.06912[physics.ins-det].[33] A. Li, A. Elagin, S. Fraker, C. Grant, and L. Winslow,Nucl. Instrum. Meth. A , 162604 (2019).[34] A. R. Back et al. (ANNIE), “Accelerator NeutrinoNeutron Interaction Experiment (ANNIE): Prelimi-nary results and physics phase proposal,” (2017),arXiv:1707.08222v2 [physics.ins-det].[35] I. Anghel et al. (ANNIE), “Expression of Interest: TheAtmospheric Neutrino Neutron Interaction Experiment(ANNIE),” (2014), arXiv:1402.6411 [physics.ins-det].[36] I. Anghel et al. , “Letter of intent: The AcceleratorNeutrino Neutron Interaction Experiment (ANNIE),”(2015), arXiv:1504.01480 [physics.ins-det].[37] A. Bernstein,
AIT-WATCHMAN Conceptual Design Re-view Report , Tech. Rep. (Lawrence Livermore NationalLab.(LLNL), Livermore, CA (United States), 2019).[38] M. Askins et al. (WATCHMAN), “The Physics and Nu-clear Nonproliferation Goals of WATCHMAN: A WA-ter CHerenkov Monitor for ANtineutrinos,” (2015),arXiv:1502.01132 [physics.ins-det].[39] D. L. Danielson et al. , “Directionally accelerated de-tection of an unknown second reactor with antineutri-nos for mid-field nonproliferation monitoring,” (2019),arXiv:1909.05374 [physics.ins-det].[40] S.-H. Seo, “Neutrino Telescope at Yemilab, Korea,”(2019), arXiv:1903.05368 [physics.ins-det].[41] M. Wurm et al. , Astroparticle Physics , 685 (2012).[42] M. Askins et al. , Eur. Phys. J. C (2020), 10.1140/epjc/s10052-020-7977-8.[43] J. R. Alonso et al. , “Advanced Scintillator Detector Con-cept (ASDC): A Concept Paper on the Physics Po-tential of Water-Based Liquid Scintillator,” (2014),arXiv:1409.5864 [physics.ins-det].[44] R. Bonventre and G. D. Orebi Gann, Eur. Phys. J. C ,435 (2018).[45] S. D. Biller, Phys. Rev. D , 071301 (2013).[46] S. Andringa et al. (SNO+), Adv. High Energy Phys. , 6194250 (2016).[47] The SNO+ collaboration, Private communication.[48] D. R. Onken, F. Moretti, J. Caravaca, M. Yeh, G. D.Orebi Gann, and E. D. Bourret, Mater. Adv. , 71(2020).[49] B. W. Adams et al. (LAPPD), “A Brief Technical Historyof the Large-Area Picosecond Photodetector (LAPPD)Collaboration,” (2016), arXiv:1603.01843 [physics.ins-det].[50] M. Agostini et al. (Borexino), Nature , 505 (2018).[51] A. Gando et al. (KamLAND-Zen), Phys. Rev. Lett. ,082503 (2016).[52] S.-F. Ge and W. Rodejohann, Phys. Rev. D , 093006(2015).[53] F. An et al. (JUNO), J. Phys. G , 030401 (2016).[54] P. Bakhti and A. Y. Smirnov, Phys. Rev. D , 123031(2020).[55] S. Agostinelli et al. (GEANT4), Nucl. Instrum. Meth. A , 250 (2003).[56] “RAT-PAC User’s Guide,” https://rat.readthedocs.io/en/latest/ .[57] J. C. R. Reis et al. , ChemPhysChem , 3722.[58] “International association for the properties of water andsteam,” , release of September1997.[59] R. M. Pope and E. S. Fry, Appl. Opt. , 8710 (1997).[60] R. C. Smith and K. S. Baker, Appl. Opt. , 177 (1981).[61] J. Caravaca, F. B. Descamps, B. J. Land, J. Wallig,M. Yeh, and G. D. Orebi Gann, Phys. Rev. C , 055801(2017).[62] Hamamatsu Photonics K.K., “Photomultiplier tubes andassemblies,” (2017).[63] B. Aharmim et al. (SNO), Phys. Rev. C , 045502(2007).[64] P. Virtanen et al. , Nature Methods , 261 (2020).[65] M. Anderson et al. (SNO+), Phys. Rev. D , 032008(2019).[66] O. Ponkratenko, V. Tretyak, and Y. Zdesenko, Phys.Atom. Nucl. , 1282 (2000).[67] J. N. Bahcall, M. Kamionkowski, and A. Sirlin, Phys.Rev. D , 6146 (1995).[68] W. Winter, S. Freedman, K. Rehm, and J. Schiffer, Phys.Rev. C , 025503 (2006).[69] J. N. Bahcall, A. M. Serenelli, and S. Basu, Astrophys.J. Lett. , L85 (2005).[70] N. Vinyoles et al. , Astrophys. J. , 202 (2017).[71] T. R. Rodriguez and G. Martinez-Pinedo, Phys. Rev.Lett. , 252503 (2010).[72] J. Barea, J. Kotila, and F. Iachello, Phys. Rev. C ,014315 (2013).[73] J. Bergstrom, M. Gonzalez-Garcia, M. Maltoni, C. Pena-Garay, A. M. Serenelli, and N. Song, J. High Energ.Phys. , 132 (2016).[74] M. Agostini et al. (Borexino), “Sensitivity to neutri- nos from the solar CNO cycle in Borexino,” (2020),arXiv:2005.12829 [hep-ex].[75] M. Agostini et al.et al.