A search for $hep$ solar neutrinos and the diffuse supernova neutrino background using all three phases of the Sudbury Neutrino Observatory

A search has been performed for neutrinos from two sources, the $hep$ reaction in the solar $pp$ fusion chain and the $\nu_e$ component of the diffuse supernova neutrino background (DSNB), using the full dataset of the Sudbury Neutrino Observatory. The $hep$ search is performed using both a single-bin counting analysis and a likelihood fit. We find a best-fit flux that is compatible with solar model predictions while remaining consistent with zero flux, and set a one-sided upper limit of $\Phi_{hep}<30\times10^{3}~\mathrm{cm}^{-2}~\mathrm{s}^{-1}$ (90% credible interval). No events are observed in the DSNB search region, and we also set an improved upper bound on the $\nu_e$ component of the DSNB flux.

A search has been performed for neutrinos from two sources, the hep reaction in the solar pp fusion chain and the νe component of the diffuse supernova neutrino background (DSNB), using the full dataset of the Sudbury Neutrino Observatory. The hep search is performed using both a single-bin counting analysis and a likelihood fit. We find a best-fit flux that is compatible with solar model predictions while remaining consistent with zero flux, and set a one-sided upper limit of Φ hep < 30 × 10 3 cm −2 s −1 (90% credible interval). No events are observed in the DSNB search region, and we also set an improved upper bound on the νe component of the DSNB flux. However, the highest energy branch in this cycle, the hep reaction ( 3 He(p, e + ν e ) 4 He), has yet to be directly detected. With a predicted branching ratio of ∼ 2 × 10 −7 , the flux expected on Earth in the BSB05(GS98) solar model is (7.93 ± 1.23) × 10 −3 cm −2 s −1 [7]. As the hep reaction has the highest endpoint energy of all solar neutrinos, and occurs at a relatively high radius in the Sun, an observation may provide sensitivity to non-standard solar models in addition to completing our picture of the pp chain neutrino fluxes.
Also expected in the energy range above the endpoint of the 8 B solar neutrino spectrum is the diffuse supernova neutrino background (DSNB), the isotropic neutrino flux from past core-collapse supernovae [8]. The DSNB provides information about the average neutrino luminosity and temperature at the flavor-dependent surface of last scattering, which would constrain models of supernova dynamics and provide context for nearby core collapse supernova events detectable on an individual basis, such as SN1987A [9]. This signal remains undetected, and the Sudbury Neutrino Observatory (SNO) experiment provides unique sensitivity to the ν e component of the flux.
A previous search for the hep and DSNB neutrinos with the SNO detector used data from the first operating phase, 306.4 live days with a heavy water (D 2 O) target [10]. The present work extends that counting analysis to the full SNO dataset across all operating phases, and additionally a spectral fit is performed. Section II briefly introduces the SNO detector. Next, Section III describes the data set, event selection, and the counting and fitbased analysis methods. Finally, results are presented in Section IV.

II. THE SNO DETECTOR
The SNO detector [11] consisted of a target volume enclosed within a transparent acrylic sphere 6 m in radius, viewed by 9456 inward-looking 8-inch photomultiplier tubes (PMTs) at a radius of 8.4 m, as illustrated in Figure 1. The acrylic vessel and the structure supporting the PMTs (PSUP) were suspended in a water-filled cavity, which was additionally instrumented with outwardlooking PMTs to provide an active veto system. In order to shield from cosmic ray muons and from the neutrons and radioisotopes resulting from muon interactions, the detector was located deep underground with a 5890 ± 94 meter water equivalent rock overburden at the Inco (now Vale) Creighton mine near Sudbury, Ontario, Canada.
The detector operated in three distinct phases, differing in the primary mechanism for neutron detection. In the first phase, the detector was loaded with a very low background heavy water (D 2 O) target. With the D 2 O target, SNO was sensitive to charged current (CC), We used a coordinate system with the center of the detector as the origin, and z direction as vertically upward.

II. THE SNO DETECTOR
The SNO detector [9], shown schematically in Fig. 1, consisted of an inner volume containing 10 6 kg of 99.92% isotopically pure heavy water ( 2 H 2 O, hereafter referred to as D 2 O) within a 12 m diameter transparent acrylic vessel (AV). Over 7 × 10 6 kg of H 2 O between the rock and the AV shielded the D 2 O from external radioactive backgrounds. An array of 9456 inward-facing 20 cm Hamamatsu R1408 photomultiplier tubes (PMTs), installed on an 17.8 m diameter stainless steel geodesic structure (PSUP), detected Cherenkov radiation produced in both the D 2 O and H 2 O. A nonimaging light concentrator [10] mounted on each PMT increased the effective photocathode coverage to nearly 55% of 4π . The PMT thresholds were set to 1/4 of the charge from a single photoelectron. The inner 1.7 × 10 6 kg of H 2 O between the AV and the PSUP shielded the D 2 O against radioactive backgrounds from the PSUP and PMTs. Extensive purification systems removed radioactive isotopes from both the D 2 O and the H 2 O [11].
The detector was located in Vale's Creighton mine (46 • 28 ′ 30 ′′ N latitude, 81 • 12 ′ 04 ′′ W longitude) near Sudbury, Ontario, Canada, with the center of the detector at a depth of 2092 m (5890 ± 94 meters water equivalent). At this depth, the rate of cosmic-ray muons entering the detector was approximately three per hour. Ninety-one outward-facing PMTs attached to the PSUP detected cosmic-ray muons. An offline veto based on information from these PMTs significantly reduced cosmogenic backgrounds.
The recoil electrons from both the ES and CC reactions were detected directly through their production of Cherenkov light. The total amount of light detected by the PMT array was correlated with the energy of the interacting neutrino. neutral current (NC), and elastic scattering (ES) channels: The hep and DSNB searches benefit in particular from the enhancement by a factor of about 100 of the CC cross section with respect to that for ES, and from the fact that in the CC interaction, the outgoing electron energy is strongly correlated with the incoming neutrino energy. Additionally, the NC channel allows rejection of atmospheric neutrino interactions by tagging of coincident neutrons. In SNO's second operational phase, the D 2 O was doped with 0.2% NaCl by mass, to take advantage of the improved neutron capture cross section on Cl and the higher energy and more isotropic de-excitation γ cascade. In the third phase, the NaCl was removed and an array of 3 He proportional counters (NCDs) was deployed to further improve neutron detection.

III. ANALYSIS
We performed a single-bin counting analysis in two different energy ranges, for the hep and DSNB neutrino signals. Additionally, a maximum likelihood fit was used to extend the sensitivity of the hep search. The following sections describe the data set, event selection criteria, and systematic uncertainties common to the counting and likelihood analyses, and then introduce those techniques.

A. Data Selection
This analysis makes use of the entire SNO dataset, across all three operational phases, with data collected between November 1999 and November 2006. Table I indicates the live time for each phase, corresponding to a total exposure of 2.47 kilotonne-years after fiducialization. We adopted a pseudo-blind approach in which the analysis was tuned on Monte Carlo simulations, then validated on one third of the data randomly sampled in short blocks of time uniformly distributed throughout the phases. Finally, with cuts and parameters having been fixed, the full dataset was re-opened for this analysis.
The set of signal candidate events follows from three stages of event selection. First, entire runs (approximately 8 hour blocks of live time) are accepted or rejected based on detector conditions. The same selection is applied as in Reference [12] for Phase I and Reference [5] for Phase III. For Phase II, the selection from Reference [13] is extended to include periods with higher than average levels of Rn or activated Na, which presented important backgrounds for the low energy threshold 8 B oscillation analyses but are insignificant for the higher-threshold hep and DSNB searches.
Next, a set of low-level cuts are applied, which address instrumental background events as well as coincidences with bursts of events or tagged muons. The instrumental backgrounds are caused by detector effects, for example high-voltage discharge of a PMT, or electronic pickup. Such events tend to have distinct signatures, such as correlations in the physical locations of electronics channels, which are very different from signal events. For each phase, the same set of low-level cuts are used as in previous work [5], as these have been extensively validated and tuned to optimize signal efficiency. For this analysis, signal-like events are further required to be isolated in time: any candidate event occurring within 250 ms of another candidate event is rejected. This includes coincidences with any event with a reconstructed vertex within a 6 m fiducial volume and a kinetic energy above 4 MeV, a trigger of the external veto, or (in Phase III only) a detected signal in the NCD array. This reduces background classes that produce coincident electrons, neutrons, or photons, and in particular targets Michel electrons following low-energy muons and atmospheric neutrino CC electrons with neutron followers.
Finally, a series of high level criteria have been developed based on reconstructed observables, which discriminate the signals of interest from other physics backgrounds. The signature of a signal hep or DSNB neutrino interaction is a single electron-like Cherenkov ring originating within 550 cm of the detector center. This fiducial volume is chosen to reduce backgrounds associated with γ rays and other backgrounds due to the materials surrounding the target volume. Signal Cherenkov rings are highly anisotropic, at a level quantified by the variable β 14 previously described in Reference [14]. The fraction of PMTs hit within a narrow prompt time window is calculated as the in-time ratio (ITR). This variable can discriminate between well-reconstructed single-ring events or multi-ring events due to a pile up of interactions or particles. To further discriminate single electron-like events, three Kolmogorov-Smirnov (KS) tests are used. The first simply tests the compatibility of the azimuthal distribution of hits around the reconstructed direction relative to a flat distribution. The second test is a two-dimensional extension that includes the polar angle and compares to a probability distribution derived from calibration data, accounting for energy dependence in the polar angle and solid angle effects in the azimuthal angle. A final test compares the time-of-flight corrected PMT hit times for hits inside the Cherenkov ring to a template distribution also extracted from calibration data. Cuts on these parameters have been adjusted relative to previous SNO analyses as described in Section III D, as both the energy regime (> 15 MeV) and the objectives (rejection of atmospheric neutrino backgrounds) differ. The distributions in these high-level observables are validated by comparing simulations to data in the low-energy sideband below the hep region of interest and to calibration data using a signal-like 8 Li source [15].

B. Monte Carlo Simulation
The detailed microphysical detector model used in previous SNO measurements, SNOMAN [5,11], was again employed for this analysis. SNOMAN was used to generate solar neutrino events, propagate final state particles through the detector geometry, and simulate the optical, triggering, and electronics response of the detector. The SNOMAN Monte Carlo contains run-by-run detector state information, tracking changes over time. All Monte Carlo was produced with at least 500 times the statistics expected in data.
For atmospheric neutrinos above 100 MeV, we use GE-NIE v2.12.2 [16,17] using the default model set, and the Bartol04 flux predictions [18], interpolated between the solar minimum and maximum according to the dates of each operational phase. The final state particles from GENIE are then input into SNOMAN for propagation through the full detector simulation. Atmospheric neutrino oscillations are applied using best-fit parameters in a model which samples an ensemble of baselines from the neutrino production height distributions of Gaisser and Stanev [19].
One potential background which is not simulated in GENIE is a 15.1-MeV γ produced in de-excitation of 12 C * following neutrino interactions on 16 O. Here, we take a sample of such untagged γ events following neutral current quasi-elastic (NCQE) interactions from a NUANCE (version 3r009) simulation, which uses the calculation of Ejiri [20], and scale according to the relative NCQE cross section in GENIE.
To model low-energy (E ν < 100 MeV) atmospheric neutrino interactions, we use the flux given by Battistoni et al. [21]; for this subdominant background only ν e and ν e are simulated, and the fluxes at the solar minimum (when the background is largest) are used. This simulation is performed directly in SNOMAN. We note that the low-and high-energy atmospheric neutrino fluxes are the same as those used in the 2006 SNO hep and DSNB search analysis [10].

C. Signals and Backgrounds
For the hep solar neutrino signal, we use the spectrum computed by Bahcall [22] and use the BSB05(GS98) flux of 7.93(1 ± 0.155) × 10 3 cm −2 s −1 [7] as a benchmark. The primary background for the hep search is due to electrons from 8 B solar neutrino interactions, at a level that depends on the shape of the spectrum near the endpoint. The spectral shape from Winter et al. [23] is used, and oscillations are applied according to a three-neutrino oscillation model using best-fit parameters [24]. The 8 B solar neutrino flux is based on a three-phase analysis of SNO 8 B solar neutrino data, identical to that presented in Reference [5] except that an upper energy threshold at 10 MeV was applied to eliminate any contamination from a possible hep signal. The extracted 8 B flux is Φ8 B = (5.26 ± 0.16 (stat.) +0.11 −0.13 (syst.)) × 10 6 cm −2 s −1 , consistent with the published value.
The DSNB signal is modeled as an isotropic ν e source using a benchmark energy spectrum and total flux. We use the model of Beacom and Strigari [25] with T = 6 MeV, which predicts a total flux of Φ DSNB νe = 0.66 cm −2 s −1 in the energy range 22.9 < E ν < 36.9 MeV.
Backgrounds due to isotropic light emission from the acrylic vessel [13] have also been studied using a dedicated event selection and Monte Carlo. The background contamination depends on the choice of fiducial volume, and is constrained to the negligible level of < 0.01 events within our energy regions of interest for the chosen cut of 550 cm. Atmospheric neutrinos and associated 12 C * backgrounds are modeled as described in III B.

D. Counting Analysis
Within each energy region of interest (ROI) for the single-bin counting analysis, 1D cuts on high level features are simultaneously tuned to optimize the search sensitivity in Monte Carlo, with further adjustments to minimize the impact of the systematic uncertainties on the shapes of the observable distributions. The hep energy ROI of 14.3 < T eff < 20 MeV and DSNB ROI of 20 < T eff < 40 MeV are chosen to optimize signalto-background ratio while maximizing signal acceptance, following the procedure described in Reference [10]. The signal efficiency of the high-level cuts is validated using calibration data sets as shown in Figure 2. Within the hep ROI, the high level and burst cuts together reduce the atmospheric neutrino backgrounds by 97%, with a signal efficiency of ∼ 99%.
For the purposes of this cut-based analysis, confidence intervals are constructed using a Bayesian framework in which we construct intervals from a Poisson likelihood function marginalized over the expected background distribution. This function is defined as − log L(µ, b|n,b, σ b ) = µ + b + log Γ(n + 1) − n × log(µ + b) where µ is the true signal mean, b the true background rate, n the observed number of events,b the mean background expectation, and σ b the Gaussian uncertainty on b. Integrating over the background parameter b yields − log L(µ|n), which is treated as a posterior PDF for µ and used to construct intervals. For a confidence level α, a two-sided interval C is defined by the highest posterior density region, i.e. adding points µ in order of their posterior probability density until C L(µ|n) ≥ α. Onesided intervals are constructed by direct integration of L to determine the smallest µ such that µ 0 L(µ|n) ≥ α. We note that in this construction we implicitly assume a uniform step function prior, constant for µ > 0.

E. Likelihood Analysis
In order to leverage the energy dependence of the signal spectra and lower the threshold for the hep search, an unbinned maximum likelihood fit was also performed. The fit considers all three phases simultaneously, with the 8 B and hep fluxes held constant across time, as well as the overall atmospheric neutrino flux normalization after accounting for differences across the solar minimum and maximum. The dominant systematic uncertainties are varied in the fit using Gaussian pull terms, and include the oscillation parameters θ 12 and ∆m 2 12 as well as the energy scale and resolution model parameters and angular and β 14 resolutions, which are treated as uncorrelated. The fit uses three-dimensional probability distribution functions (PDFs), binned in reconstructed energy (T eff , 10 bins, 10 -20 MeV), the angle relative to the Sun (cos θ sun , 10 bins, -1 -1), and the isotropy parameter (β 14 , 15 bins, -0.12 -0.95). PDFs are constructed for 8 B CC electrons, 8 B ES electrons, hep CC electrons, hep ES electrons, and atmospheric neutrino interactions for each phase. The relative normalizations of the CC and ES components for each signal are fixed. The cuts described previously are applied to data and Monte Carlo prior to PDF construction and fitting; these include the fiducial volume, ITR, three KS probability figures of merit, and low-level cuts. In contrast to the counting analysis, energy and isotropy are observables in the fit.
The full negative log likelihood function optimized in the fit is of the form [27]: where the first term corresponds to the total normalization constraint, the second the unbinned likelihood given the PDFs, and the final two terms represent Gaussian uncertainties on rate and systematic parameters. In Equation 2, P are PDFs for each signal, binned in observables x. These PDFs are constructed from Monte Carlo events, modified according to the set of s systematic parameters ∆, with associated Gaussian uncertainties σ ∆ . Rate parameters r are related to an expected number of eventsÑ by an efficiency matrix such that N i = j i r j , and have Gaussian uncertainties σ r . M is simply the number of rate parameters which are externally constrained. The efficiency matrix relates the fit parameters to the signal rates, e.g. with a single hep flux parameter controlling the hep ES and CC components. These matrix elements also include a weighting with a factor of |{x i |S(x i , ∆) ∈ V }| to account for the number of events (described by a set of observables x) shifting across the boundary of the analysis window (a volume V in the observable space) after application of a systematic transformation S.
The fit was performed using purpose-built Markov Chain Monte Carlo (MCMC) code. Two specific tasks were accelerated using general-purpose GPUs: the building of PDFs from Monte Carlo performed at each iteration of the fit to account for varying systematic parameters, and the computation of the unbinned likelihood function, where the sum over events is performed in parallel. The use of GPU acceleration reduces the time required per iteration by about a factor of 1000 relative to a single CPU. A number of metrics were used to evaluate fit quality and convergence, including a check of statistical compatibility of parameter distributions within subdivisions of the MCMC random walk, and a toy Monte Carlo to evaluate the goodness of fit through a χ 2 hypothesis test. Additional validation included signal injection tests varying the hep flux from 0.01 -10 times model predictions.

F. Systematic Uncertainties
A number of systematic effects are important within these analyses. The primary background to the hep search is electrons from 8 B solar neutrino interactions, where the spectrum is affected by the energy response modeling as well as the flux normalization and intrinsic shape. The flux uncertainty is taken from the three-phase fit to low-energy SNO data described in Section III C, and the shape is varied within the uncertainties provided by Winter et al. [23]. For solar neutrinos, the uncertainties in the oscillation parameters and the ν − d CC cross section are also included. To address the energy response modeling, which affects all signals and backgrounds, uncertainties are derived from fits to deployed calibration sources and samples of Michel electrons; this procedure is described in Section III F 1. Uncertainties impacting atmospheric neutrino backgrounds are detailed in Section III F 2. The major systematic uncertainties impacting the analyses are summarized in Table II.

Detector Response
In order to calibrate the response in the detector across an energy range up to 40 MeV, several event samples were compared against SNOMAN Monte Carlo predictions. The vertex reconstruction is described in References [28] (Phases I and II) and [29] (Phase III), and based on this we include a 2.4 cm uncertainty on reconstructed position resolution, and an overall 2.9% fiducial volume uncertainty. Additionally, a 2% uncertainty on the angular resolution for ES events is modeled as a scal- ing via a parameter ∆ θ [5]: where (cos θ) outside the interval [−1, 1] are assigned a random value within that interval.
In each of the three phases, a large sample of 6.13-MeV γ rays from a deployed 16 N source provided the primary calibration. Additionally, a pT source in Phase I provided a sample of 19.8-MeV γ rays [26]. To extend the model to higher energies, samples of Michel electrons from decays of stopping cosmic ray muons were selected and fit to a response model allowing an energy-dependent fractional energy scaling (∆ (i) S ) and shift in resolution (∆R): The parameters were extracted using a maximum likelihood fit to the Michel electron samples for each phase, subject to prior constraints based on the deployed source measurements. The extracted parameters are given in Table III. We find that the parameters are consistent with zero, confirming that the initial 16 N-based energy calibration provides a reasonable estimate of energy across the regions of interest, and the correlated errors in each phase indicate the magnitude of systematic shifts that remain compatible with the higher-energy calibration samples. This provides a data-driven constraint on the smearing of the spectrum of electrons produced by 8 B solar neutrino interactions, which forms a dominant background for the hep search.
Additionally, a similar model including a linear scaling and resolution was applied to the shape of the isotropy parameter β 14 , with ∆ (0) S = ∆ R = 4.2×10 −3 for all three phases, based on measurements with the 16 N calibration source [5]. Finally, the contribution of any non-Gaussian (flat) tails in the energy response was constrained to the level of 10 −3 events in the energy region of interest based on samples of events from the deployed 8 Li source [15], which has a β spectrum similar to that of the 8 B solar neutrinos.

Atmospheric Neutrinos
Two main classes of uncertainty affect the atmospheric neutrinos: the flux uncertainty, which is taken to be 25% [21] and 10% [18] for low (< 0.1 GeV) and high (0.1 − 10 GeV) energies, respectively, and the cross sections. The cross section uncertainties are evaluated through event reweighting, by simultaneously varying the parameters in the default GENIE model set (see Ref. [17]) within their respective uncertainties to produce an ensemble of weights corresponding to different model hypotheses.
To validate the modeling of atmospheric neutrino interactions, a sample of events coincident with a fully contained atmospheric neutrino predecessor event was selected. The predecessors are required to have 200 -5000 PMTs hit, no activity in the veto region, and not to follow a through-going µ event. The follower events, which consist of Michel electrons and neutrons, must pass all analysis cuts and have an energy 5 < T eff < 100 MeV. For the selected events, we compare the multiplicity and timing of coincidences as well as the energy, position, isotropy, and other high-level observables between the atmospheric Monte Carlo and data. The distributions are shown in Figure 3 for the follower multiplicity distributions for each phase, and the joint distribution of reconstructed energy T eff and the time difference between the predecessor and follower event (∆t). The two populations correspond to neutron captures and stopped muon decays. The distributions show reasonable agreement with the GENIE Monte Carlo processed through SNOMAN, consistent within the uncertainties due to statistics, flux, and cross section modeling.

A. Counting Analysis
Within the sensitivity-optimized energy regions of interest for the hep and DSNB signals, we performed a single-bin counting analysis as introduced in Section III D. The energy spectra for selected events are shown in Figure 4. The total signal and background expectations in the 14.3 -20 MeV hep energy ROI are 3.09±0.12 and 13.89 ± 1.09, respectively, with 22 events observed; the distribution across phases is given in Table IV. The uncertainties on the total three-phase signal and background expectations are correlated (r hep = 0.83, r DSNB = 0.12), and are obtained using an ensemble of 500 three-phase pseudo-experiments with systematic parameters randomly sampled according to their correlated uncertainties. The dominant source of uncertainty in the hep region is the energy response modeling, due to the steeply-falling tail of the 8 B solar neutrino backgrounds. This model is constrained using data spanning the energy range as described in Section III F 1.
The majority of candidate events, 13 of 22, occurred during Phase III. These events appear signal-like in all respects, and consistency with background is observed in sidebands with respect to energy and all other highlevel observables. According to toy Monte Carlo studies, the probability of observing a statistical fluctuation of at least this magnitude in any one phase is approximately 8%.
Applying the Bayesian procedure described in Section III D yields an 68.3% credible interval (CI) of Φ hep = (9.6 − 33) × 10 3 cm −2 s −1 ; however, as the probability of  [10]. With 0.08 signal and 2.58 ± 0.26 background events expected in the DSNB ROI, the median experiment in a Monte Carlo ensemble provides 90% CI sensitivity to signals at least 52 times larger than the benchmark Beacom & Strigari T = 6 MeV model. With an apparent downward fluctuation, zero events are observed, and we set an upper limit of 29 times the model prediction, corresponding to DSNB ν e flux of Φ DSNB νe < 19 cm −2 s −1 (90% CI) in the energy range 22.9 < E ν < 36.9 MeV. The dominant source of systematic uncertainty in the DSNB ROI is the 10% normalization uncertainty for the flux of atmospheric neutrinos with E ν > 100 MeV.

B. Likelihood Analysis
For the hep search, we additionally performed a likelihood fit as described in Section III E. One-dimensional projections of the best fit in the observable dimensions T eff , β 14 , and cos θ sun are shown in Figure 6. The quality of the fit was evaluated using a χ 2 test based on an effective test statistic distribution derived using a toy Monte Carlo, yielding a p-value of 16.0% considering statistical errors only.
A Bayesian credible interval for the hep flux is obtained by marginalizing over all other parameters, and constructing the highest posterior density region (HPDR) containing a given probability, as shown in Figure 5. The intervals and best fit value obtained with the Bayesian HPDR approach are consistent with quantities obtained by directly analyzing the likelihood space sampled by the MCMC.
In agreement with the counting analysis up to differences introduced by the statistical treatments, this result  is compatible with the BSB05(GS98) model prediction and is consistent with zero hep flux. The fit yields a 68.3% HPDR credible interval for the hep flux parameter corresponding to Φ hep = (5.1 − 23) × 10 3 cm −2 s −1 ; as in the counting-based analysis, we define a one-sided upper limit: Φ hep < 30 × 10 3 cm −2 s −1 (90% CI).

V. CONCLUSIONS
Data from the full SNO dataset, representing an exposure of 2.47 kilotonne years with a D 2 O target, has been analyzed to search for neutrinos from the hep reaction in the Sun's pp chain and ν e from the diffuse supernova neutrino background. In addition to increasing the exposure by a factor of 3.8 relative to the previous SNO search for these signals [10], a new spectral fit has been employed to improve the sensitivity to the hep flux.
We have performed the most sensitive search to date for the hep solar neutrino flux, the final unobserved branch of the pp fusion chain. This measurement is compatible with the BSB05(GS98) model prediction, while remaining consistent with zero hep flux, and we extract a one-sided upper limit of Φ hep < 30 × 10 3 cm −2 s −1 90% CI. In a search at energies above the solar neutrino endpoints, we observe no evidence for the DSNB ν e flux, and set an upper limit on this flux; our results suggest that a ν e flux larger than about thirty times the current predictions is disfavored. Future experiments sensitive to DSNBν e through inverse beta decay, in particular SuperK-Gd [30], are expected to achieve sensitivity to the model predictions in the near future. There also exists great potential to further explore both the hep and DSNB ν e in upcoming experiments such as DUNE [31,32], where the CC ν e − 40 Ar cross section provides similar advantages to ν e − d in SNO, together with low threshold and large exposure.