Theia: An advanced optical neutrino detector

New developments in liquid scintillators, high-efficiency, fast photon detectors, and chromatic photon sorting have opened up the possibility for building a large-scale detector that can discriminate between Cherenkov and scintillation signals. Such a detector could exploit these two distinct signals to observe particle direction and species using Cherenkov light while also having the excellent energy resolution and low threshold of a scintillator detector. Situated in a deep underground laboratory, and utilizing new techniques in computing and reconstruction techniques, such a detector could achieve unprecedented levels of background rejection, thus enabling a rich physics program that would span topics in nuclear, high-energy, and astrophysics, and across a dynamic range from hundreds of keV to many GeV. The scientific program would include observations of low- and high-energy solar neutrinos, determination of neutrino mass ordering and measurement of the neutrino CP violating phase, observations of diffuse supernova neutrinos and neutrinos from a supernova burst, sensitive searches for nucleon decay and, ultimately, a search for NeutrinoLess Double Beta Decay (NLDBD) with sensitivity reaching the normal ordering regime of neutrino mass phase space. This paper describes Theia, a detector design that incorporates these new technologies in a practical and affordable way to accomplish the science goals described above. We consider two scenarios, one in which Theia would reside in a cavern the size and shape of the caverns intended to be excavated for the Deep Underground Neutrino Experiment (DUNE) which we call Theia 25, and a larger 100 ktonne version (Theia 100) that could achieve an even broader and more sensitive scientific program.


I. INTRODUCTION AND THEIA OVERVIEW
The neutrino is the fundamental particle we would most expect to be ignored: they interact too weakly and are too light to directly affect most microscopic processes. Yet neutrinos access a breadth of science no other fundamental particle can: understanding the weak sector through direct measurements of neutrino properties; testing fundamental symmetries of Nature; probing near and distant astrophysical phenomena; peering into the interior of the Earth, and understanding the earliest moments of the Universe. That scientific breadth has been mirrored by the broad array of technologies used to detect and study neutrinos, with the strength of each technology typically focused on a narrow slice of neutrino physics. We discuss in this white paper a new kind of detector, called Theia (after the Titaness of light), whose aim is to make world-leading measurements over as broad range of neutrino physics and astrophysics as possible.
Long-baseline experiments T2K [1] and NOνA [2] operate off-axis in order to enhance the flux of neutrinos at low energies while suppressing backgrounds from higher energies. Theia is designed to operate with high sensitivity in an on-axis neutrino beam (such as the one being designed for the Long Baseline Neutrino Facility (LBNF) [3,4] and the Deep Underground Neutrino Experiment (DUNE) [5][6][7]. Theia will have good sensitivity to neutrino oscillation parameters and neutrino CP Violation (CPV) with a relatively modestly-sized detector. In addition to this long-baseline neutrino program, Theia will also contribute to atmospheric neutrino measurements and searches for nucleon decay, particularly in the difficult p → K + + ν and N → 3ν modes [8][9][10].
Theia will also make a definitive measurement of the solar CNO neutrinos, which to date have not been detected exclusively [11][12][13] but which would tell us important details about how the Sun has evolved [14]. Theia will also provide a high-statistics, low-threshold (∼ 3 MeV) measurement of the shape of the 8 B solar neutrinos and thus search for new physics in the MSW-vacuum transition region [15,16]. Antineutrinos produced in the crust and mantle of the Earth will be measured precisely by Theia with statistical uncertainly far exceeding all detectors to date. Should a supernova occur during Theia operations, a high-statistics detection of theν e flux will be made-literally complementary to the detection of the ν e flux in the DUNE liquid argon detectors. The simultaneous detection of both messengers-and detection of an optical, x ray, or gamma ray component will enable a great wealth of neutrino physics and supernova astrophysics. With a very deep location and with the detection of a combination of scintillation and Cherenkov light [17,18], Theia will have world-leading sensitivity to make a detection of the Diffuse Supernova Neutrino Background (DSNB) antineutrino flux [19][20][21]. The most ambitious goal, which would likely come in a future phase, is a search for Neutrinoless double beta decay (NLDBD), with a total isotopic mass of 10 tonnes or more, and with decay lifetime sensitivity in excess of 10 28 years [22,23].
Theia is able to achieve this broad range of physics by exploiting new technologies to act simultaneously as a (low-energy) scintillation detector and a (high-energy) Cherenkov detector. Scintillation light provides the energy resolution necessary to get above the majority of radioactive backgrounds and provides the ability to see slow-moving recoils; Cherenkov light enables event direction reconstruction which provides particle ID at high energies and background discrimination at low-energies. Thus, the scientific program benefits in many cases on the ability of Theia to discriminate efficiently and precisely between the "scintons" (scintillator photons) and "chertons" (Cherenkov photons).
Discrimination between chertons and scintons can be achieved in several ways. The use of a cocktail like water-based liquid scintillator (WbLS) provides a favorable ratio of Cherenkov/scintillation light [24]. Combining angular and timing information allows discrimination between Cherenkov and scintillation light for high-energy events even in a standard scintillator like LAB-PPO [24]. Slowing scintillator emission time down by using slow secondary fluors can also provide excellent separation [25]. Recent R&D with dichroic filters to sort photons by wavelength has shown separation of long-wavelength chertons from the typically shorter-wavelength scintons even in LAB-PPO, with only small reductions in the total scintillation light [26]. In principle, all of these techniques could be deployed together if needed to achieve the full Theia physics program. New reconstruction techniques, to leverage the multicomponent light detection, are being developed and with the fast timing of newly available PMTs and the ultrafast timing of LAPPDs (Large Area Picosecond Photon Detectors), allow effective tracking for high-energy events and excellent background rejection at low energies.

A. Detector configuration
The requirements for each of Theia's physics goals are different, although in nearly all cases increased detector mass unsurprisingly provides better sensitivity. We consider for our sensitivity studies two distinct size configurations: a 25-kt total mass detector with a geometry consistent with one of the planned DUNE caverns and which could be deployed relatively quickly (Theia 25); and a detector with 100-kt total mass in a right-cylinder geometry (Theia 100). Ultimately, the limitation on size is likely driven by the attenuation lengths of the scintillator mixture in the detector.
The variation in requirements in some cases is inclusive: the need for good energy resolution at low energies only improves the background rejection and reconstruction capabilities at high energies, and the same is true for requirements on direction reconstruction via Cherenkov photons. Other requirements may be exclusive or at least partially so: the presence of inner containment to hold a 0νββ isotope-loaded scintillator would make a long-baseline analysis more complex from an optical standpoint, or reduce fiducial mass.
A major advantage of Theia is that the target can be modified in a phased program to address the science priorities. In addition, since a major cost of Theia is expected to be photosensors, investments in Theia25 instrumentation can be transferred directly over to Theia100. Thus, Theia can be realized in phases, with an initial phase consisting of lightly-doped scintillator and very fast photosensors, followed by a second phase with enhanced photon detection to enable a very low energy solar neutrino program, followed by a third phase that could include doping with a 0νββ isotope and perhaps an internal container.
We have listed below the major high-level requirements for each physics target, and in the following sections we detail how Theia could meet these requirements for each physics case.

II. DETECTOR CAPABILITIES
The Theia detector is made possible by the development of new technology, mainly in the areas of fast photosensors, novel scintillating liquids, and advanced image recognition techniques, especially those utilizing Machine Learning (ML) and other new techniques to discern underlying patterns from complex image data. In this section we present the status of these new technologies, and in addition discuss plans to further develop and incorporate them into the Theia design.

A. Water-based liquid scintillator
Theia will be a unique detector, designed as the first large realization of the Advanced Scintillation Detector Concept (ASDC) [27]. Thus far, large Water Cherenkov (WC) detectors such as Super-Kamiokande (SK) have suffered in sensitivity due to the inability to detect particles with energy below the Cherenkov threshold. For example, this limits sensitivity to the Diffuse Supernova Neutrino Background (DSNB) [19] due to enhanced backgrounds from low-energy atmospheric neutrino interactions, and reduced signal from the inability to detect positron annihilation, which enhances the prompt signal from the leading reaction ν e +p → e + +n. In the area of proton decay, the kaon from p → νK + is below the Cherenkov threshold, and in the area of solar neutrinos the 7 Be and CNO neutrinos are practically undetectable as much of the energy from the neutrino electron scattering reaction is invisible.
Organic liquid scintillators (LS) have been used to enhance sensitivity for below Cherenkov threshold particles. LS is currently being used in the KamLAND, Borexino, and SNO+ detectors, and is planned for use in the JUNO detector now under construction. While this is very effective at increasing sensitivity at low energies, it comes at the loss of the directional sensitivity and multi-track resolution that is a hallmark of WC detectors. Use of organic LS also introduces issues of high cost, short optical transmission lengths, and undesirable environmental and safety problems.
The recent development of Water-based Liquid Scintillator (WbLS) [28] has the potential to alter this situation. By introducing a small amount (typically 1%-10%) of liquid scintillator into water, the liquid yield can be adjusted to allow detection of particles below Cherenkov threshold while not sacrificing directional capability, cost, or environmental friendliness. First developed at Brookhaven National Lab (BNL), WbLS is a leading candidate for the main target medium for Theia, and will enhance the proposed scientific program significantly, as described in subsequent sections.
There is an active R&D effort to realize the novel WbLS liquid target being considered for Theia.
These efforts include a precision measurement of attenuation at long distances, demonstration of material compatibility with detector components, and accurate costs and production capabilities. Examples include WbLS development at BNL [29], purification and compatibility studies at UC Davis [30], characterization and optimization with the CHESS detector at UC Berkeley and LBNL [24], fast photon sensor development at Chicago and Iowa State [31], development of reconstruction algorithms, and potential nanoparticle loading in NuDot at MIT [32]. The WbLS R&D program for Theia also strongly leverages existing efforts and synergy with other programs, such as ANNIE [33], SNO+ [34], WATCHMAN [35], and others.
For the purposes of the studies presented in here we have made the following reasonable assumptions for WbLS performance: (1) absorption and scattering are simply weighted averages of pure water and LAB-based LS, and (2) a 10% of LS light yield can be achieved with good stability and reasonable costs.
For the studies of Theia performance for CP violation, the advantages of WbLS have not been incorporated into the analysis. For example, it is expected that this will likely provide better vertex resolution and enable detection of below Cherenkov threshold charged hadrons, but given the high light levels and a reasonable Cherenkov/scintillation photon separation, the tracking performance already achieved in existing WC detectors will not be degraded.

B. Photodetection
Complementary to the development of new chemical loadings and WbLS is the development of new advanced photodetection capabilities. Progress in photosensor technology will enable significant improvements in time resolution, improved light collection, spatial granularity at different scales, and even the ability to separate photons by production process, production point, and wavelength. Many of the technologies, once speculative, are now reaching maturity. The specific combination of these technologies that will optimize the physics reach of Theia is currently being explored. These are summarized below.
Since the construction of the last generation of large water optical neutrino detectors, significant progress has also been made in the advancement of conventional vacuum PMTs. High Quantum Efficiency (QE) PMTs with QE greater than 35% are now readily available from multiple suppliers.
Direct comparisons in the laboratory between these new devices and the 20-inch Hamamatsu PMTs installed in Super-Kamiokande have shown that they have a factor of 1.5 better photon collection efficiency per square centimeter [36]. Thus, a wall coverage of only 27% is needed to be equivalent to the 40% in Super-Kamiokande in terms of photon collection capabilities. In addition, timing is significantly better (e.g. 1.3 ns FWHM [36] versus 5.1 ns [37]), and other performance characteristics are also much improved [36,38]. In some studies below, PMTs with modern performance characteristics were used (e.g. solar neutrinos), while in others older performance characteristics have not yet been updated (e.g. CP violation search) and thus results are expected to improve.
Large Area Picosecond Photodetectors (LAPPDs) are 20 cm x 20 cm imaging photosensors with single photoelectron (SPE) time resolutions below 100 picoseconds and sub-cm spatial resolutions [31,39]. The combination of these capabilities makes it possible to better separate individual photons and develop reconstruction tools that fully capture correlations between the time and spatial patterns of light, rather than treating them independently. LAPPDs are now commercially available through Incom, Inc and will soon be deployed in their first neutrino application in the ANNIE experiment [33].
Studies done by ANNIE show a significant increase in reconstruction capabilities [33], albeit at rather short distances. Note that ANNIE has only one LAPPD for every twenty-five PMT's and it was determined from simulations that this was sufficient to meet the initial goals of ANNIE to measure neutron production from single track QE events. The addition of LAPPDs was shown to improve the vertex and tracking resolution by a factor of two over just using PMTs. The addition of more LAPPDs to handle multi-track events and further improve performance is still under study.
LAPPD costs are expected to drop with increasing production yields and market extent. Further work in developing new production techniques, such as the use of ceramic bodies, and the development of in situ photocathodes, once mature, could have a significant impact in further reducing these prices.
Another advancement in photon detection is the development of multi-PMT modules and mixed PMT coverage schemes. IceCube, KM3Net, and Hyper-Kamiokande are developing Digital Optical Modules (DOMs) [40][41][42] in place of conventional large area photomultipliers. These transparent modules each contain an array of smaller PMTs along with the readout electronics. These DOMs provide similar area coverage to more conventional large-area PMTs, but with finer spatial and time resolution, as well as the ability to resolve directionality. The JUNO collaboration is pursuing a detector design with a mixture of large-area and small-area PMTs to achieve increased coverage, and to provide different scales of spatial granularity [43].
Many of the key advances in photodetection look beyond the photosensors themselves and consider optics for using photodetectors more efficiently. Traditional light collectors such as Winston Cones and scintillating light guides have been rigorously pursued for the Long Baseline Neutrino Experiment (LBNE) [44] and Hyper-K [45]. Using arrays of lenses, plenoptic imaging would add directional information to detected light [46]. Novel designs using specular reflection off mirrors could also be used to enhance coverage. One particularly promising optical concept is the application of dichroic filters to separate light by wavelength. This would provide a strong additional handle for discriminating between the largely monochromatic scintillation light and broad-band Cherenkov light, as well as enabling better correction for chromatic dispersion of Cherenkov photons [26].
Moving forward, the Theia collaboration will leverage detailed simulations and reconstruction tools to evaluate the optimal suite of photosensors, light collection and sorting optics, and readout electronics. The challenge is to enable improved physics over a wide range of energies and to co-develop the photodetector systems with the optimization of the particular WbLS cocktail. The deployment of photosensors in Theia will likely be different than in other existing applications, so another key task will be the development of application-ready modules, in parallel with the process of technology down selection.

C. Reconstruction techniques
While Cherenkov detectors have been very successful in reconstructing various properties of the particles involved in a neutrino event, liquid scintillation detectors have long been considered a source for calorimetric information only. However, in recent years it has become clear that the time information of the light in liquid scintillators can be used to access a wide range of information, similar or even superior to what a pure Cherenkov detector can deliver [47].
There are two complementary approaches to reconstruction in both detector types and, consequently, also in WbLS detectors. The first approach developed in MiniBooNE [48] and extended to the more complicated event topologies of Super-Kamiokande [49] follows a likelihood ansatz to find the optimal track parameters and compare different hypotheses. This technique can naturally accommodate Cherenkov and scintillation light, as was required for MiniBooNE, by combining Cherenkov and scintillation light predictions for each photosensor in the calculation of the likelihood. In contrast to this, three-dimensional topological reconstruction [50] tries to picture the spatial distribution of the energy deposition within the detector without using a specific hypothesis. This technique has been developed for the LENA [51] detector and also been implemented for the JUNO detector [52]. The application of these algorithms to Cherenkov detectors is straightforward. An example for a reconstructed stopping muon in LENA (a liquid scintillator detector) clearly showing the accessibility of the energy loss per unit length in shown in Figure 1.
Both methods have been improved considerably over the last couple of years. For example fiTQun [49], the reconstruction software used by Super-Kamiokande/T2K, is now able to reconstruct up to 6 Cherenkov rings produced by electron, muon, or pion particle hypotheses. This allows for a simultaneous determination of the identity and number of particles. Recently, the Super-Kamiokande collaboration also published results using fiTQun [53] with improved kinematic and particle identification capabilities. This allowed them to increase the volume accessible to the analysis by 32%.
Topological reconstruction offers large volume liquid detectors the same capabilities as highly segmented detectors (with all the resulting implications). This includes possibilities for particle identification at energies as low as a few MeV based on topological information. For example it is now possible to distinguish point-like events from multi-site events in liquid scintillator [54] using various techniques, including likelihood-based pulse shape discrimination methods [55]. Figure 2 shows an example of the separation of electrons from gammas, critical for separation of neutrino scatters, which produce electrons, from common gamma emitters in the uranium and thorium decay chains.
At low energies, reconstruction techniques have also improved remarkably. Reconstruction of events with energies down to 3-5 MeV using Cherenkov light in pure water has been done at the Super-Kamiokande experiment [56,57]. The ability to separate Cherenkov from scintillation light allows the use of additional information for event reconstruction. This was pioneered by the LSND experiment, which successfully used Cherenkov light in a diluted liquid scintillator to reconstruct electron tracks in the energy range of about 45 MeV [58]. More recently, the feasibility of directional reconstruction of few MeV electrons in a conventional high-light yield liquid scintillator has also been demonstrated in Reconstruction results for a simulated muon with 3 GeV initial kinetic energy in the cylindrical LENA detector projected along the symmetry axis (left) or a radial y-axis (right). The primary particle started at (0, 1000, 0) cm in the direction (1, −1, 0). Both the projected tracks of the primary particle (red) and of secondary particles (black) are shown. The cell content is given in a.u. and rescaled such that the maximum content is 1. The plots are taken from [50], which provides details on the reconstruction procedure.
simulations relevant to NLDBD experiments [59]. In both these examples, the use of timing information was crucial for separating directional Cherenkov light from the isotropic scintillation light.
Direction reconstruction based on Cherenkov-scintillation light separation continues to make rapid progress [54,60]. At their core, these new algorithms rely on selecting a sample of photoelectrons (PEs) with a favourable ratio of Cherenkov to scintillation light. Then the 'center-of-mass' of the PE distribution on the detector surface becomes aligned with the direction of the parent particle. An angular resolution of 0.5-0.8 radians can be achieved for a large spherical detector similar to KamLAND-Zen, but equipped with 100-ps time resolution photo-detectors. An example of direction reconstruction for 2.5-MeV electrons is shown in Figure 3.
Beyond directionality, topological features of events in the few-MeV energy range can also be extracted. A spherical harmonics analysis of PE distributions has been used to separate NLDBD from 8 B solar neutrino events [61]. This has been improved by introducing a Cherenkov-scintillation space-time boundary that allows for reliable and more general Cherenkov-scintillation light separation [60]. The Cherenkov PEs.

Reconstruction of various characteristics of candidate events in detectors with scintillation and
Cherenkov light can further benefit from the widely spreading use of Machine Learning techniques.
For example, separation of NLDBD from 10 C events using a convolutional neural network (CNN) has been explored in [62]. In a simulation using detector parameters similar to KamLAND-Zen, including photo-coverage, time and position resolution, 60% rejection of 10 C events has been achieved with a 90% signal efficiency for NLDBD candidate events.

III. PHYSICS SENSITIVITIES AND DETECTOR REQUIREMENTS
Theia would address a broad program of physics, including: geo-neutrinos, supernova neutrinos, nucleon decay, measurement of the neutrino mass hierarchy and CP violating phase, and even a nextgeneration neutrinoless double beta decay search. The sections below discuss the progress in estimating sensitivities and detector requirements in the context of Theia 25 and Theia 100. These parameters have all been measured, with the exception of the value of the CP-violating phase, δ CP [64], although the ordering of the mass states is also not completely determined. Long-baseline neutrino oscillation experiments have significant sensitivity to the mixing parameters θ 23 , θ 13 , and δ CP , as well as to the mass splitting ∆m 2 32 , and the neutrino mass ordering via matter effects. The atmospheric parameters, θ 23 and ∆m 2 32 , have been measured by the existing long-baseline oscillation experiments T2K [1] and NOvA [2]. These experiments are also beginning to have sensitivity to δ CP . DUNE [6,7] is a next-generation long-baseline neutrino oscillation experiment being built at the Long Baseline Neutrino Facility (LBNF) [5] that will make a definitive determination of the neutrino mass ordering, will have sensitivity for a definitive discovery of CP-violation for much of the possible parameter space, and will make precise measurements of all the oscillation parameters governing longbaseline oscillation in a single experiment. DUNE plans to use liquid argon Time Projection Chambers (TPC) for the large detectors at the LBNF far site in the United States. Hyper-Kamiokande [45] is a next-generation long-baseline experiment using a water Cherenkov detector (WCD) at a far site in Japan. Both are anticipated to run on a similar timescale.
The long-baseline oscillation sensitivities of two potential configurations of Theia positioned at the LBNF far site have been considered. The full realization of Theia is a 100-kt right cylindrical volume (Theia 100), similar to the geometry of Super-Kamiokande and Hyper-Kamiokande, which corresponds approximately to 70-kt fiducial volume. A smaller 25-kt realization (Theia 25) with a 17-kt fiducial volume in one of the four LBNF caverns would be able to supplement the DUNE measurements with CP-violation sensitivity comparable to a 10-kt (fiducial) liquid argon detector [65].
The ability to measure long-baseline neutrino oscillations with a distinct set of detector systematic uncertainties and neutrino interaction uncertainties relative to the liquid argon detectors, would provide an important independent cross-check of the extracted oscillation parameter values.
For the studies presented here, we use GLoBES [66,67] to calculate predicted spectra for different oscillation parameter hypotheses and compare these to quantify experimental sensitivity. We make use of the publicly available LBNF beam flux and DUNE detector performance description [68]. For the DUNE sensitivity we assume a 10-kt fiducial mass, corresponding to a single DUNE far detector module. For the Theia sensitivity, we use Theia 's expected 70-or 17-kton fiducial mass and assume the detector can be designed to perform as well as and no better than a conventional WCD, using Monte Previous studies of a water Cherenkov detector in the LBNF beam were performed in the context of the predecessor experiment to DUNE: LBNE [71]. These studies were based on Super-Kamiokande event reconstruction techniques developed within the first several years of Super-Kamiokande data taking, and were restricted to single-ring events with no Michel electrons from stopped pion and muon decay. In the decade since, important advancements have been made in Cherenkov reconstruction that have substantially improved particle identification and ring counting. As described in Section II C, the FiTQun event reconstruction package used for Theia sensitivity studies has now been fully implemented in the most recent T2K analyses [1]. These improvements, when applied to the LBNF beam, enhance the sensitivity to neutrino oscillations in three important ways: 1. The improved ring counting has removed 75% of the neutral current background, relative to the previous analysis, due to improvements in the detection of the faint second ring in boosted π 0 decays; 2. The improved electron/muon particle identification has allowed for an additional sample of 1ring, zero Michel electron events from ν 3 -CCπ + interactions, without significant contamination from ν µ backgrounds 3. Multi-ring ν e event samples can now be selected with sufficient purity to further enhance sensitivity to neutrino oscillation parameters.
The long-baseline oscillation analysis now includes nine samples that are analyzed with independent systematic uncertainties within a single fit: one, two-, and three-ring events with either zero or one Michel electron in neutrino mode, and the corresponding zero Michel electron samples for anti-neutrino mode. To reduce the neutral current background, a boosted decision tree is employed, which uses the best fit likelihoods of all 1-, 2-, and 3-ring hypothesis fits, and the lowest reconstructed particle momentum in each fit. The resulting neutrino-mode samples are shown in Figure 4 and the antineutrino-mode sampes are shown in Figure 5. The two-and three-ring samples tend to have higher background than the single-ring samples, but do make significant contributions to the overall sensitivity.
We assign independent normalization uncertainties of 2%(5%) on each of the ν e and ν e appearance signal(background) modes. We do not explicitly include the ν µ disappearance samples, but the choice of uncertainty for the appearance samples assumes some systematics constraint from the disappearance samples. This treatment of systematic uncertainty is comparable to that in the DUNE CDR analysis, and assumes significant constraint of systematic uncertainty from the DUNE near detector. Under these assumptions, we find better than 3σ sensitivity to CP violation for much of parameter space for Theia 100, with the sensitivity of the Theia 25 WCD being comparable that of the DUNE 10-kt LArTPC, as shown in Fig. 6. It is expected that this may improve when the modern fast HQE PMTs, WbLS, and LAPPDs are considered. 2. Potential charged-current (CC) detection via isotope loading e.g. 7 Li [72]. The differential CC cross section for neutrino interaction on 7 Li is extremely sharply peaked. As a result, CC neutrino detection provides a high-precision measurement of the incoming neutrino energy, allowing extraction of the low-energy 8 B spectrum. This would provide a sensitive search for new physics via a probe of the transition region in the neutrino spectrum between vacuum-dominated and matter-enhanced oscillations. There is also the potential to separate the different components FIG. 6: Sensitivity to CP violation (i.e.: determination that δ CP = 0 or π) (left) and sensitivity to determination of the neutrino mass ordering (right), as a function of the true value of δ CP , for the Theia 70-kt fiducial volume detector (pink). Also shown are sensitivity curves for a 10-kt (fiducial) LArTPC (blue dashed) compared to a 17-kt (fiducial) WCD (pink dashed). Seven years of exposure to the LBNF beam with equal running in neutrino and antineutrino mode is assumed. LArTPC sensitivity is based on detector performance described by [68].
of the CNO flux via a shape analysis.

ES measurements
The sensitivity to CNO and pep solar neutrinos of an unloaded WbLS detector via the ES interaction has been studied in [73]. By performing a two-dimensional binned maximum likelihood fit in energy and direction relative to the Sun, cos θ , neutrino fluxes are separated from each other, as well as from certain sources of radioactive background. For each signal the cos θ distribution was determined fully analytically. All non-neutrino signals were assumed to be flat. For the solar signals the electron direction relative to the Sun was determined based on the differential cross sections. This was then convolved with a chosen angular resolution. The energy response was determined semi-analytically, based on chosen detector parameters such as target light yield and photocathode coverage. Here we summarize the assumptions made regarding detector configuration and performance, and the resulting sensitivity to solar neutrinos. Full details of the analysis are described in [73].
While the default detector sizes for this paper are 25 kT (Theia 25) and 100 kT (Theia 100), the baseline detector configuration was chosen to be 50 kT for historical reasons. The results presented here can easily be scaled by the exposure in kT-years to one of the default configuration sizes, as the TABLE I: Background assumptions for the baseline configuration. a The 40 K level in water is taken to be 0.1x the Borexino measurement [78] b The 85 Kr, 39 Ar, and 210 Bi levels in water are taken to be the Borexino measured level in scintillator [76], although levels increased by several orders of magnitude are explored light collection changes slowly with volume on this scale. The baseline detector has 90% PMT coverage, a 5% WbLS target, 0.6 MeV threshold, and 25 • angular resolution, with baseline background levels given in Table I. The dominant cosmogenic background, from 11 C, was conservatively taken to be at the Borexino level, scaled by the respective target mass. If located at the proposed site at LBNF this background would in fact be significantly lower. All results assume a five year live time.
The impact of each choice of detector configuration and background level was studied, including the target mass, the percentage loading of LS in the WbLS target, photocathode coverage, angular resolution, and background levels. Energy reconstruction was performed semi-analytically, and the resolution was determined from a combination of the target light yield and photocathode coverage.
The effect of systematic uncertainties in both energy scale and resolution were considered. A full reconstruction of event direction was not attempted; rather, the impact of certain values of angular resolution was considered. In practice, the achievable angular resolution would be correlated with other detector parameters, such as the percent LS loading -a higher fractional loading makes separation of the prompt Cherenkov signal from the isotropic scintillation more challenging, thus limiting the angular resolution. This separation could be further enhanced by deploying fast photon sensors, such as LAPPDs [31].
The sensitivity for each signal (solar neutrinos and dominant backgrounds) with the baseline detector configuration and background assumptions is shown in Table II.
The impact on the CNO solar neutrino sensitivity of detector size, LS fraction, and angular resolution for the baseline background assumptions is shown in Table III and Fig. 7.
A study of both energy scale and resolution systematic uncertainties shows that these can be constrained by the data to sub-percent levels, making them sub-dominant in the final flux sensitivities.
The sensitivity was shown to be extremely robust to background level variations of several orders

CC measurements
The potential for CC measurements in Theia was studied in [27]. That work is summarised here.
Several factors contribute to the choice of isotope for loading into a scintillator detector. 37 Cl and 71 Ga have been used successfully by radiochemical experiments from the late 1960s to the present day. 7 Li was also proposed as an additive to a water detector in [72]; such a detector would have excellent sensitivity to the high end of the 8 B spectrum, but would be limited in threshold. As seen in (1). 7 Li + ν e → 7 Be + e − (Q = 862 keV) (1) 7 Li has only two significant transitions: a mixed Fermi and Gamow-Teller transition to the ground state of 7 Be with a threshold of 0.862 MeV; and a super-allowed Gamow-Teller transition to the first excited state at ∼430 keV, which decays with a lifetime of τ ∼200 fs. The scattering is very hard, transferring almost all incident energy to the scattered electron. If one could differentiate between the electron of the ground state and the e − + γ of the first excited state one would have a high-precision reconstruction of neutrino energy. The two states also have precisely known angular distributions, which could then be used as an additional handle to differentiate signal from background. Even without the use of particle ID to differentiate between states, the contribution of the two is known precisely from theory, so the difference in threshold can be used to demonstrate that the two are being seen in the correct proportions. There is also the potential to observe NC interactions on 7 Li (Fig. 8), exciting the analog 478 keV first excited state of 7 Li, which then decays with a lifetime of τ ∼105 fs.
On preliminary investigation 7 Li would thus appear to be the preferred isotope. However, other factors may be important, such as the effect of isotope loading on detector optics.  [79], [80], [81], and [82], respectively. Although the differential uncertainties are not shown, the uncertainty on the lithium cross section is roughly 1% [72]. (Bottom) Predicted solar neutrino event spectra for 5 years of data-taking, with 1% loading by mass of candidate isotopes in a 30-kT WbLS-filled ASDC detector. Solid lines show the standard solar neutrino oscillation prediction. Dashed lines are for a flat neutrino spectrum to low energies, indicative of new physics interactions. 7 Li is the most favorable choice due to a high cross section for neutrino absorption. Five years of data taking results in over 17 σ separation in the integral flux, and correspondingly high precision (several σ significance) on the extracted spectrum.

C. Supernova neutrinos
The neutrino burst detected from the next galactic Supernova (SN) will provide us with a wealth of information on the dynamics of the core collapse (neutronization, reheating, proto-neutron star cooling) and the properties of the neutrinos themselves (mass hierarchy, absolute mass scale, collective oscillations) (e.g. [84]). Since the first detection of SN neutrinos in 1987, there has been a continuous stream of new features predicted for the SN neutrino signal, hinting at new stellar or particle physics [85]. So while we are uncertain what (superposition of) signatures to expect from the next event, it is beyond doubt that only a concerted effort of all neutrino observatories available will enable us to extract the full information contained in the burst signal (e.g. [86]), then to be combined with electromagnetic and gravitational wave observations (e.g. [17]).
If a SN neutrino burst would pass by the Earth today, the largest event statistics would be collected by the two large Cherenkov detectors, Super-Kamiokande (SK) and IceCube [87,88]. Ten years from now, we may expect that additional information will be added by JUNO's liquid scintillator and DUNE's liquid argon neutrino targets [52,89]. In a simplified picture, SK, JUNO and IceCube will dominate the information onν e flux and energies, while DUNE has the potential for a high-statistics ν e measurement. JUNO will provide information on the combined flux of ν µ and ν τ and antineutrinos (denoted commonly as ν x ).
What will Theia add to the global picture of SN neutrino observations? To answer this, we assume  and re-decays of 16 N, the presence of γ-lines from NCO reactions as well as the directional signature for ES will allow to resolve the integrated SN neutrino signal into its individual spectral components (Fig. 11). This will enable unambiguous spectroscopy of the ν e (ES+ν e O) andν e   shifted by cosmic expansion, these neutrinos constitute a faint isotropic background flux. The discovery and subsequent spectroscopy of the DSNB will provide unique information on the redshift-dependent SN rate, the relative frequency of neutron star and black hole formation and the equation of state of the emerging neutron stars (e.g. [94]).
The primary detection reaction for theν e component of the DSNB flux is the Inverse Beta Decay (IBD). With an expected event rate of 0.1 per year per kiloton of detector material, overwhelming backgrounds have to be faced. However, with SK-Gd and JUNO there are now two contenders with a realistic chance to obtain a first (3σ) evidence of the DSNB within the next 5−10 years [52,95].
Theia may play a pivotal role in the discovery and exploration of the DSNB signal. With a target mass several times the size of SK or JUNO, Theia 100 will obtain a ∼5σ discovery of the DSNB in less than 1 year of data taking and reach O(10 2 ) DSNB events within ∼5 years. Even the smaller Theia 25 will profit considerably from the dual detection of Cherenkov and scintillation signals that offer a background discrimination capability unparalleled by Gd-doped water or pure organic scintillator: For instance, a signal efficiency of 95 % can be maintained while reducing the most crucial background from atmospheric neutrino NC interactions to a residual ∼1.7 % [21].

DSNB signal and background levels
For IBD events, the prompt energy of the positron signal translates almost directly to the incident neutrino energy, while the delayed neutron capture provides a fast coincidence tag to reduce the ample single-event backgrounds. Figure 13a)   can be greatly reduced by ring counting, Cherenkov/scintillation ratio and delayed decay cuts to obtain a signal-to-background ratio >1.
in case a neutron is released from the nuclear break-up. First recognized by the KamLAND experiment for its special importance in organic scintillators [96], it dominates the DSNB signal by more than one order of magnitude. For scintillator detectors like JUNO, the quenched nuclear signal constitutes a major challenge to be overcome by pulse shape discrimination [52]. SK-Gd (or HK-Gd) will be much less affected but features this background as well below ∼16 MeV [97].

Background discrimination in WbLS
The main virtue of Theia lies with the excellent background discrimination capabilities of the WbLS. In the context of the DSNB search, the following have been investigated: • A Fiducial Volume cut to reject surface background events, especially fast neutrons created by muons in the surrounding rock. In current configuration, Theia 25 will feature 20 kt of fiducial mass, Theia 100 about 80 kt.
• Distance cuts relative to muon tracks traversing the Fiducial Volume may be used to veto decays of the βn-emitter 9 Li while reducing live exposure by about 1 %.
• Ring counting: The reconstruction of Cherenkov rings from individual final-state particles provides a handle to discriminate one-ring positron events from multi-ring atm-NC events. From simulation, we find that about 40 % of the latter background events can be discriminated based on this feature 1 , introducing a negligible loss in signal efficiency.
• Delayed decays: In almost 50 % of the relevant atm-NC events, the residual nucleus is the β +emitting isotope 15 O with a lifetime of 2.2 min. Given the low energy threshold, Theia will be able to tag the delayed decay in order to reject half of the atm-NC background without any noticable loss in signal efficiency.
• C/S ratio: Uniquely, a WbLS detector offers the possibility to discriminate events by the magnitude of the Cherenkov signal accompanying the scintillation light. While e ± signals feature a high C/S ratio, that of non-relativistic nuclear recoils is practically zero. As demonstrated by the left panel of Fig. 14  While the first column represents the rates before cuts, the following columns apply delayed decay, C/S ratio and ring-counting cuts. The cited fast neutron rate assumes a 2.5 m fiducial volume cut or presence of corresponding active shielding surrounding the target volume.
Tab. V illustrates the impact of the aforementioned discrimination techniques on the signal and background rates within the observation window. While all background components including the atm-NC events are greatly reduced, the DSNB signal acceptance is hardly affected. The corresponding energy spectra are shown in Fig. 13b), demonstrating a clear dominance of the signal over the entire energy window.
Theia 25 will require about 6 years of data taking to achieve a 5σ discovery of the DSNB signal (assuming standard predictions for flux and spectral energy) [21]. Combined with SK+Gd and JUNO, the three detectors will acquire about 5-10 DSNB events per year (with ∼40% of statistics from Theia), so that a spectroscopic analysis of the DSNB based on O(10 2 ) events will become feasible over 10-20 years. At the same time, the C/S signatures of the large atm-NC event sample recorded in WbLS will enable an in-depth study of this most relevant background and will help to reduce the corresponding systematic uncertainties as well for SK-Gd and especially JUNO.
Theia 100 will take a clear lead in the exploration of the DSNB signal. DSNB discovery is expected in a bit more than 1 year of data taking, and first spectral analyses only a few years later: in the long run, spectroscopy will provide access to the astrophysics of SNe, ranging from the equation of state of neutron stars to the fraction of dark SNe resulting from black-hole collapses [21].

E. Geological and reactor neutrino measurements
Global antineutrinos emerge from nuclear beta-minus decays, which produce a characteristic energy spectrum in the few MeV range for each parent isotope. While the mixture of isotopes decaying within a source uniquely determines the energy spectrum of the emitted antineutrinos, neutrino oscillations distort the spectrum of the detected antineutrinos in a pattern determined by the distance from the source. Spectral distortion is pronounced for point-like nuclear power reactors and subtle for diffuse sources within Earth.
The rate and energy spectrum of global antineutrino interactions varies dramatically with surface location [98]. The following discussion assumes antineutrino interactions by inverse beta decay on hydrogen (E ν > 1.8 MeV) in a water target located at SURF and with an energy-independent detection efficiency of 90%, which depends heavily on the scintillator loading and light collection. Thus thes eplots need to be scaled for the different phases of Theia Figure 15 shows the detected energy spectrum of the predicted rate of antineutrinos from the nuclear power reactors and Earth.
Observations of Earth antineutrinos, or geo-neutrinos, probe the quantities and distributions of terrestrial heat-producing elements uranium and thorium. The quantities of these elements gauge global radiogenic power, offering insights into the origin and thermal history of Earth [99]. Spatial distributions reveal the initial partitioning and subsequent transport of these trace elements between metallic core, silicate mantle, and crust types [100]. Ongoing observations at underground sites in Japan [101] and Italy [102] record the energies, but not the directions, of geo-neutrinos from uranium and Theia would precisely measure the uranium and thorium components of the energy spectrum with the potential to test models of differential partitioning and transport of these trace elements between silicate mantle and crust types.
Theia also expects ∼ 20 reactor antineutrino events per kT-year. This detected rate allows Theia to demonstrate techniques for remote reactor discovery [104], including measuring range and direction at distances greater than 1000 km. Figure 16 plots the intensity of nuclear power as a function of distance from SURF. These distances are encoded by neutrino oscillations in the reactor antineutrino spectrum, which is shown in Fig. 15. An analysis using Rayleigh statistical methods can actually unfold the distance of the reactor clusters, as shown by the blue overlay.
In summary, with sufficient detection efficiency and exposure Theia located at SURF has the potential to make significant advances in applied antineutrino physics. Theia could provide the first evidence for surface variation of the geo-neutrino flux and make a precise measurement of the thorium and uranium components of the energy spectrum. Theia could demonstrate basic techniques for remote discovery of nuclear reactors by making antineutrino measurements of range and direction.

F. Neutrinoless double beta decay
The Theia search for neutrinoless double beta decay (NLDBD) aims for sensitivity to the nondegenerate normal hierarchy parameter space within the canonical framework of light Majorana neutrino exchange and three-neutrino mixing, at the level of m ββ ∼ 5 meV. This is achieved through the loading of a very large mass of a NLDBD candidate isotope into an ultra-pure liquid scintillator target, together with coincidence and topological particle identification techniques. PMT hits per deposited MeV is assumed (corresponding to about 3%/ √ E energy resolution). This value includes the reduction in the light yield due to the addition of the isotope, estimated to be around 30% at 5%, or higher, loadings. For the specific case of Xe-loaded scintillator this light yield is likely an underestimation, as the KamLAND-Zen experiment predicts a reduction of only 15% at 3% loading, which at 90% coverage would correspond to 1530 Nhits/MeV [105]. Figure 17a shows the impact of light yield on the Theia sensitivity. Simulations are obtained with the Geant4-based RAT-PAC software package 2 ; when available radioactive decays are simulated using the Decay0 code [106], otherwise the decay mode is input in RAT-PAC. Reconstructed energy is approximated by assuming the Poisson limit of photon counting: the true deposited energy, accounting for quenching, is smeared by a Gaussian resolution function corresponding to the light yield.

Backgrounds
The main sources of background included in the this analysis are summarized in Table VI   Cosmogenic Production: These backgrounds are due to activation of nuclei by muons (during data taking) or protons and neutrons (during material production and handling at Earth's surface).
The production rates of various radionuclides by cosmogenic neutron and proton spallation reactions in Xe and Te have been investigated in [113][114][115][116][117][118][119]. Among the most important nuclides produced there are 60 Co (Q = 2.8 MeV, T 1/2 = 5.27 y) and 110m Ag (Q = 3.1 MeV, T 1/2 = 250 d). Mitigation of these background sources requires minimal exposure at sea level, a deep underground cool-down period, chemical purification processes [120], and, to limit in-situ production during data taking, the use of a water shield. In these studies it is assumed that proper measures are taken to handle the target material, reducing the background to a negligible level.
The most dangerous nuclide for the NLDBD study from in-situ muon induced events is 10 C (Q = 3.65 MeV, T 1/2 = 19.3 s), produced by muon interactions with the carbon atoms of the liquid scintillator. In this study the detector is assumed to be located in the Homestake mine, at a depth of 4300 m.w.e. The estimated event rate is about 300 events/kt/yr [121] for a muon flux of 4.2 × 10 −9 cm −2 s −1 and an average muon energy of 293 GeV [122]. A reduction of 92.5% of the 10 C background has been demonstrated by Borexino [123] using a three-fold coincidence technique [109,124]. A machine learning approach, as described in Sec. II C, could be used in addition to the three-fold-coincidence method to further improve the rejection of 10 C events.
Solar Neutrinos: 8 B solar neutrino elastic scattering in the target material results in a background that is approximately flat across the NLDBD energy ROI, but that can be rejected using reconstructed event direction. Figure 17b shows how the sensitivity scales with this solar neutrino rejection efficiency. A rejection of at least 50% is necessary in order to reach the sensitivity goal.
Monte Carlo simulations show that, for 2.5 MeV electrons in water and about 50% coverage, about 80% of the 8 B events can be rejected while keeping 75% of the NLDBD signal (assumed to be isotropic). In the case of Theia, the confounding effect due to the liquid scintillator can be compensated by use of high QE PMTs, together with high coverage. Other options might include the use of slow scintillators. In [125] it is shown that more than 50% rejection in 8 B can be achieved for an LAB target, retaining more than 70% of the signal. In the following sensitivity calculations it is assumed a 8 B neutrino rejection of 50% with 75% signal efficiency. External Sources: Decays from U-and Th-chain impurities present in the balloon material, the external WbLS, the shielding water, and in the PMTs also contribute to the background. External background events can be reduced using a fiducial volume cut, and PMT hit-time information.
In the following study a rejection factor of 50% on the top of the fiducial volume is assumed.
However, at the assumed balloon purity, a smaller reduction factor has a small impact on the sensitivity.

NLDBD sensitivity: counting Analysis
To estimate the sensitivity, a single-bin counting analysis is employed. Since several backgrounds do not scale with isotope mass (e.g. solar neutrinos and external γ backgrounds), we use the Monte Carlo to evaluate the background expectation, establish a confidence region using the Feldman-Cousins frequentist approach, and derive an expected limit on the NLDBD half life: where N is the number of atoms of active NLDBD isotope, is the efficiency, t the live time, and b the expected background. 'FC' refers to a Feldman-Cousins interval at confidence level α.
The expected event rates per year for a nat Te or enr Xe loaded Theia detector are given in Table   VI, for a fiducial volume radius cut of 7 m (67% acceptance) and an asymmetric energy region, from −σ/2 → 2σ, to maximize signal acceptance ( = 66.9%) while removing much of the steeply-falling two-neutrino DBD background. Figure 18 shows the background spectra near the endpoint in the Te ( Figure 18a) and Xe (Figure 18b) cases. A 75% signal efficiency, following the 50% reduction in the 8 B solar neutrino events, is applied.
The expected sensitivity (90% CL) for 10 years of data taking, using phase space factors from [131] and matrix element from [132] (g A =1.269) is: Te : T 0νββ 1/2 > 1.1 × 10 28 y, m ββ < 6.3 meV Xe : T 0νββ 1/2 > 2.0 × 10 28 y, m ββ < 5.6 meV It should be noted that for the case of Xenon, the use of a more realistic light yield of about 1500 nhits/MeV, as obtained from [105], would increase the half-life limit to 2.1×10 28 years, corresponding to m ββ < 5.4 meV. Unfortunately, the required mass of Xe to reach the normal hierarchy is about 10 times the world annual production, which makes the use of Xe likely impractical.  The enrichment option for Se is less promising due to the smaller G 0ν M 2 0ν value, and the larger costs (although a larger world annual abundance than Xe is available). The limit in this case is T 1/2 = 1.6×10 27 yr (m ββ = 18 meV).
where C i is a dimensionless coefficient and O (k) i are local operators of mass dimension k [133]. Since Λ is large, operators of increasingly higher dimension are heavily suppressed, with the lowest order terms (dimension 5) resulting in p →ν i K + ,ν i π + , e + K 0 , µ + K 0 , e + π 0 , µ + π 0 , e + η, µ + η; i = e, µ, τ, where the dominant decay mode depends on the particular model of interest. More exotic modes can also exist, but require a means to suppress the above modes via some mechanism. Extra dimensional theories for example can create constraints on dimension 5 operators such that they are forbidden allowing for decays of higher order operators to dominate leading to decays such as n → 3ν [8]. Next generation detectors will be able to probe deeper into the phase space of such detectors, possibly reaching the sensitivity to measure such a process and give evidence for the type of physics beyond the Standard Model. The Theia detector has the size and resolution to contribute to this effort, and in certain modes, provide the dominant experimental measurement. In the case of modes like p → e + π 0 , the efficiency of Theia would be similar to current detector like Super-Kamiokande and future detectors like Hyper-Kamiokande, and therefore Theia 25 would add only marginally to the DUNE program, and Theia 100 would just add in proportion to the exposure. For modes like p → νK + the contribution to the global sensitivity would be significant, and for modes like n → 3ν Theia would be world-leading. These are discussed in more detail below.

p →νK and related modes
Unlike the pion, the kaon is less effected by intranuclear effects, which means that the primary efficiency will come down to how well any given detector will be able to distinguish this very specific signature. Water Cherenkov detectors are able to identify the number of rings corresponding to the kaon and its decay products: • K + → µ + ν µ (63.42%) • K + → π + π 0 (21.13%) • K + → π + π + π − (5.58%) • K + → π 0 e + ν e (4.87%) • K + → π + π 0 π 0 (1.73%) whereas a scintillator detector (JUNO) must rely on the timing of the decay to separate the kaon from its daughters, and a tracking detector (DUNE) can separate the particles using their tracks. The signal itself has roughly 3 unique energy depositions in each detector: the kinetic energy deposited by the kaon; the decay products (either a muon or pions); and the subsequent decay of those muons and pions. JUNO's primary loss in efficiency comes from the overlap in the kaon energy deposition and its subsequent decay (using a 7-ns prompt window for the 12-ns kaon lifetime) [52]. Since the prompt light would make identifying the Cherenkov ring from the decay product extremely difficult, it is safe to assume that Theia could do no better in separating these two and would have a similar efficiency of 55%. DUNE on the other hand would be able to identify the kaon track regardless of the decay time, and is expected to have a signal efficiency of 97% [134]. A pure water detector is only sensitive to the decay products, as the kaon is itself below Cherenkov threshold, making this particular mode more difficult to measure; however, in lieu of the kinetic energy deposition from the Kaon, water detectors are sensitive to the deexcitation γs emitted by the daughter nucleus (e.g. 16   Detection efficiency and background rates for p →νK + . The efficiency DUNE is due to its excellent kaon particle identification from the liquid argon TPC [135]. Theia and Juno are assumed to have the same relative efficiency, which is dominated by the short lifetime of the kaon.
Hyper-K efficiency is the lowest due to being unable to detect the kaon since it is below the Cherenkov threshold.
expects an efficiency of 23% to see the muon (or charged pion), and the subsequent decay. In all cases, the primary background is from atmospheric neutrino interactions, with the expected background contribution given in Table VII. The Theia signal efficiency and backgrounds are estimated to be similar to JUNO for this mode, since the observation of the Kaon energy will dominate deexcitation γs and the dominant selection criterion is inter-event timing. Figure 19 shows the expected sensitivity  FIG. 19: Sensitivity for p →νK + is highest for Theia, closely followed by the Hyper-K detector, whereas JUNO and DUNE will perform similarly. The inclusion of Theia in the fourth Dune cavity would provide an enhancement to this mode over the full 40-kt Dune. A subset of theories predict modes of nucleon decay where the decay products themselves do not leave a direct visible signature with the detector. It is in this set of potential decay modes that Theia would represent a potential increase sensitivity by over two orders of magnitude. An example of this would be the decay of a neutron into three neutrinos. For a bound nucleon this would leave the nucleus in an altered state, which would have observable deexcitation gamma rays and low-energy emitted nucleons. For large-scale detectors, the real difference between this decay mode and those mentioned above is a matter of energy scale and isotope. Water Cherenkov detectors such as SNO and SNO+ have looked for invisible decay from the oxygen nucleus, which has a relatively high branching ratio (44 %) to emit a 6.18-MeV γ. The signal is a single event, so the detector is required to have very low background in order to perform this search. Super-K, for example, is limited by the production of cosmogenic activation of oxygen to 16 N, which decays with a 7.13-s half-life, emitting a 6.13-MeV gamma 67 % of the time. This limitation also exists for Hyper-K due to its shallow depth, which means Theia would be the only large-scale water Cherenkov detector available to look for this decay. Leading limits on invisible neutron and dineutron decay are set by KamLAND [9], and on invisible proton and diproton decay by SNO+ [10]. Liquid scintillator detectors have both advantages and disadvantages in the search for invisible neutron decay. While the branching ratio for carbon is much lower (5.8 %), the signal itself is a triple coincidence signal. The primary deexcitation of 11 C emits secondary particles (p, n, d, α, γ) providing a secondary signal, which is followed by the radioactive decay of 11 C-half-life of 19.3 s. KamLAND identifies no signal that can directly mimic this signal, and so its primary source of background is from accidental coincidences. Background to the first two components of the signal is dominated by cosmogenic radioisotopes (particularly 9 Li), which requires a long cut of 2 s after each muon. Since the KamLAND muon rate is ∼ 0.34 Hz, this cut is tolerable; however, JUNO would likely be insensitive to this decay mode due to its larger size and slightly shallower depth, which results in a factor of 10 more muons through the detector.
For Theia, at a depth equivalent to DUNE, the primary backgrounds would come from 8 B solar neutrinos, production of 15 N * through atmospheric neutrino interactions, and cosmogenic production of 16 N. Compared with the SNO+ results, Theia would have much less impact from internal radioactivity and a greater signal efficiency due to enhanced energy resolution. Direction reconstruction would still play an important role, as the event direction is used as the primary means for rejecting solar neutrino events. The resulting backgrounds would be ∼ 100 /kt·y for Theia, with the expected sensitivity, well above existing limits, shown in Figure 20 for a 30 kt fiducial volume.

IV. CONCLUSIONS
The Theia detector design represents an important step forward in the realization of a new kind of large optical detector: a hybrid Cherenkov/scintillation detector. Such an instrument is made possible by the convergence of three technological breakthroughs: (1) the development of ultra-fast and/or chromatically sensitive photosensors, (2) the ability to make novel target materials that produce detectable levels of both kinds of light, and (3) highly sophisticated pattern recognition and data analysis techniques that have moved well beyond the relatively simple methods of a decade ago. In addition, the coming availability of the new Long Baseline Neutrino Facility enables a broad program due to the deep depth and powerful neutrino beam available at the site.
This paper has detailed the exciting possibilities of Theia for new scientific discovery across a broad spectrum of physics, including long baseline neutrino measurements of oscillations and CP violation searches, solar neutrino measurements of unprecedented precision and scope, and the potential to extend the reach of neutrinoless double beta decay searches to the mass scales implied by a Normal Ordering of neutrino masses. In addition, Theia represents a major step forward for advancement in other fields -ranging from detection of the Diffuse Supernova Background flux to a measurement of geo-neutrinos with unprecedented statistics.