Abstract
We report the first search result for the flux of astrophysical electron antineutrinos for energies in the gadolinium-loaded Super-Kamiokande (SK) detector. In 2020 June, gadolinium was introduced to the ultrapure water of the SK detector in order to detect neutrons more efficiently. In this new experimental phase, SK-Gd, we can search for electron antineutrinos via inverse beta decay with efficient background rejection thanks to the high efficiency of the neutron tagging technique. In this paper, we report the result for the initial stage of SK-Gd, during 2020 August 26, and 2022 June 1 with a 22.5 × 552 kton · day exposure at 0.01% Gd mass concentration. No significant excess over the expected background in the observed events is found for the neutrino energies below 31.3 MeV. Thus, the flux upper limits are placed at the 90% confidence level. The limits and sensitivities are already comparable with the previous SK result with pure water (22.5 × 2970 kton · day) owing to the enhanced neutron tagging. Operation with Gd increased to 0.03% started in 2022 June.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
1. Introduction
Astrophysical electron antineutrinos () are a unique probe to assess various physical phenomena in the universe, such as past supernovae (Beacom 2010), resonant spin flavor precession (RSFP) of solar neutrinos (Akhmedov & Pulido 2003; Díaz et al. 2009), and the annihilation of MeV-scale light dark matter (Palomares-Ruiz & Pascoli 2008), which are expected to appear at energies .
The core-collapse supernova (CCSN) is one of the transient processes with the highest neutrino production in our universe. A neutrino observation from CCSN has occurred only once so far, from the supernova SN1987A, by the Kamiokande (Hirata et al. 1987), IMB (Bionta et al. 1987), and Baksan (Alekseev et al. 1987) neutrino detectors. Although nearby CCSNe are rare, neutrinos from all past supernovae (diffuse supernova neutrino background, DSNB) should exist around us. The theoretical expectations for the DSNB flux depend on various parameters: the supernova rate introduced from the cosmic star formation rate depending on the redshift, the neutrino mass ordering, the equations of states for remnant neutron stars, the metallicity of the galaxy, the failed supernova rate, and the binary interaction effect of stars (Hartmann & Woosley 1997; Malaney 1997; Kaplinghat et al. 2000; Ando et al. 2003; Ando 2005; Horiuchi et al. 2009; Lunardini 2009; Galais et al. 2010; Nakazato et al. 2015; Horiuchi et al. 2018, 2021; Kresse et al. 2021; Tabrizi & Horiuchi 2021; Ekanger et al. 2022). Ashida & Nakazato (2022) have investigated the fraction for the failed CCSNe forming black holes from the DSNB flux upper limit.
The Sun is one of the most intense astrophysical electron neutrino sources at Earth. The solar can be produced by the combination of the Mikheyev–Smirnov–Wolfenstein (MSW) effect (Smirnov 2005) and the RSFP effect via the neutrino magnetic moment. The limit for the neutrino–antineutrino conversion probability of solar neutrinos due to RSFP was placed from the flux upper limit by Agostini et al. (2021) and Abe et al. (2022a, 2022c). In addition, a small flux of the from the Sun, via the beta decay of 40K, 238U, and 232Th, is predicted by Malaney et al. (1990), but also has not been observed yet.
Neutrino production from self-annihilation of MeV-scale light-dark-matter particles (χ χ → ν ν) is also predicted (Argüelles et al. 2021). An upper limit on the averaged cross section for the self-annihilation of the light dark matter is placed from the flux upper limit by Abe et al. (2022c).
Although neutrino detectors worldwide have been searching for these events, no significant signal has been found so far. The most stringent upper limits for the astrophysical flux are set by Super-Kamiokande above 13.3 MeV (Abe et al. 2021) and by KamLAND experiment below 13.3 MeV (Abe et al. 2022c). For further lower-energy regions, the Borexino experiment (Agostini et al. 2021) gave the lowest upper limit below 8.3 MeV. In this paper, we present a new search for astrophysical in the neutrino energy range of 9.3 to 31.3 MeV based on the observation made with the new Super-Kamiokande detector configuration.
2. The Super-Kamiokande Experiment
Super-Kamiokande (SK; Fukuda et al. 2003) is a water-Cherenkov detector experiment located 1000 m underground in Kamioka, Japan. The detector is of cylindrical shape inside a tank with a diameter of 39.3 m and a height of 41.4 m. It is currently filled with 50 kton of gadolinium-doped ultrapure water (Abe et al. 2022b) as a target. It is optically separated into a main cylindrical volume (inner detector, ID), surrounded by the outer detector (OD) that extends up to the inner surface of the SK tank. SK observes signals of neutrino interactions via the detection of Cherenkov light produced by charged particles within the ID using the ID PMTs. The ID is 33.8 m in diameter and 36.2 m in height, with 11,129 20 inch photomultiplier tubes (PMTs) mounted pointing inwards to the inside of the ID tank. SK is sensitive to neutrinos with an energy range from several MeV to above 1 TeV.
The OD is concentrically placed outside the ID; it is about 2 m wide. It is instrumented with 1885 8 inch PMTs that are mounted on the outer side of the ID structure pointing outwards. The OD surface is covered by white Tyvek sheet to enhance the light reflection in the OD. The OD is primarily utilized to veto cosmic-ray muons.
Since the start of SK operation in 1996, several upgrades to the detector have been carried out. Notably, the electronics were upgraded in 2008 (Yamada et al. 2010) allowing the implementation of a new event trigger, super high energy (SHE), to identify events above ∼6 MeV. In addition, the upgrade enables us to record all PMT hits within 535 μs after the SHE-triggered event. The period for which the SHE trigger was implemented is called "SK-IV," and it operated until 2018.
In general, the triggers are issued using the number of ID or OD PMT hits within a 200 ns window. For the SHE trigger, the threshold is typically 60 ID PMT hits. The threshold for the OD trigger is typically 22 OD PMT hits.
Another important upgrade was the loading of Gd into the ultrapure water of SK in order to improve neutron-tagging efficiency due to Gd's high neutron-capture cross section and enhanced neutron-capture signal. The original idea was proposed by Beacom & Vagins (2004), and the loading was carried out starting in 2020 July. After dissolving gadolinium-sulfate-octahydrate in the ultrapure water (Abe et al. 2022b), the SK experiment started a new phase called SK-Gd. The initial running period of SK-Gd is called "SK-VI"; it operated with 0.011% Gd mass concentration until 2022 June. Under this condition, about 50% of neutrons are captured by the Gd with a typical time constant of 115 μs. Thermal neutron capture on Gd results in multiple gamma-ray emissions with a total energy of about 8 MeV, which can be easily distinguished from the random backgrounds from natural radioactivity and PMT dark noise. In this paper, 552.2 days of live time taken during SK-VI from 2020 August to 2022 June is used for the signal search.
3. Event Selection
In this analysis, inverse beta decay (IBD; ) signals from the astrophysical neutrinos are investigated. The signals consist of a positron-like event (prompt event) and a subsequent delayed neutron (delayed signal). We search for IBD signals in the data with the SHE-triggered and successfully recorded 535 μs event from the signal energy region of 7.5–29.5 MeV for the reconstructed kinetic energy of prompt events (Erec). In addition, the condition that the OD trigger is not issued is also required in order to remove incoming cosmic-ray muon events.
The event selection and reconstruction follow the previous SK-IV search (Abe et al. 2021), except for the delayed neutron identification. Prompt events with exactly one delayed-tagged neutron signal are selected as IBD signal candidates. Those neutron signals are searched for within a 535 μs window after the prompt event. In SK-Gd, the neutron signal from thermal neutron capture on the Gd nucleus is efficiently tagged by a simple signal selection without using the machine-learning-based cut used in previous analysis (Abe et al. 2021).
As a primary selection, an algorithm searching for PMT hit clusters within 200 ns windows is applied. In this stage, the clusters with 25 hits or more are selected as candidates for neutron signals. Then, basic event reduction cuts, used in common with the analysis for the scale in SK, are applied. The cuts include the goodness of reconstruction, the electron likeness of the Cherenkov ring hit pattern, and the fiducial volume cut that requires the event to be further than 2 m from the wall.
For further reduction, we apply a "distance" cut and an "energy" cut. The reconstructed vertex of the delayed signal tends to be close to the prompt event for the IBD interactions of neutrinos. Thus the distance between the prompt and delayed signal candidate is an efficient discriminator to select neutron events. In this analysis, the candidates that have a distance of over 3 m from the prompt event vertex are eliminated. Also, the reconstructed energy of multiple gamma-ray events from Gd capture is typically around 4–5 MeV, as shown in Harada (2022). After the 3 m distance cut, the remaining background events are low-energy accidental backgrounds due to noise hits. Hence a 3.5 MeV energy threshold is assigned in order to reduce the accidental coincidence rate to .
Neutron tagging efficiency and the misidentification probability are estimated by applying the above selection procedure to neutron signals simulated by a Monte Carlo (MC) simulation based on Geant4.10.5p1 (Allison et al. 2006), and SK random trigger data, respectively. The Gd concentration in the MC simulation is tuned to reproduce the time constant of the measured data. For the gamma-ray emission from thermal neutron capture on Gd in MC simulation, the ANNRI-Gd model (Hagiwara et al. 2019; Tanaka et al. 2020) is utilized. The selection efficiency of the Gd capture signal is evaluated to be 73.2% ± 0.2%. Given that the capture fraction on Gd is 47.8% ± 0.2% and systematic uncertainty estimated based on Harada (2022), the total neutron tagging efficiency is estimated to be 35%.6 ± 2.5%. The misidentification probability of noise candidates εmis is found to be (2.8 ± 0.1) × 10−4 per event. It is sufficiently low to remove most accidental coincidences. In the previous analysis (Abe et al. 2021), typical neutron tagging efficiency and the misidentification probability using a boosted decision tree (Chen & Guestrin 2016) were about 20% and also , respectively.
Figure 1 shows the total IBD signal efficiency for each 2 MeV Erec bin. The efficiencies after each of the main selection criteria applied are shown together.
Figure 1. IBD signal efficiency for the signal energy region. The 100% efficiency line corresponds to data after trigger requirements and noise reduction cuts. These lines show the cumulative efficiency at each cut stage, performed in the order shown in the legend. More detailed explanations for each reduction step, except for the neutron tagging, are described in Abe et al. (2021).
Download figure:
Standard image High-resolution imageThe signal efficiency after neutron tagging (the red line in Figure 1) is about twice that of the previous SK search (shown in the upper panel of Figure 18 in Abe et al. 2021), especially below 15.49 MeV, thanks to the higher neutron detection efficiency.
A large part of the accidental coincidence background originates from pairs of events from the decay of isotopes produced by muon spallation induced by the cosmic-ray muon and the low-energy noise events. Therefore, efficient spallation event reduction and low εmis in neutron tagging are important to reduce the accidental coincidence background. Thus, the spallation cut criteria are optimized in each energy range to remove various isotope decays at the corresponding energy. Because of the lower εmis than for SK-IV, we reoptimized the spallation cut criteria below 15.49 MeV, where we have large enough statistics of the spallation sample. As a result, the spallation cut condition was loosened to achieve higher signal efficiency. On the other hand, the spallation cut criteria cannot be optimized above 15.49 MeV because there are fewer spallation samples. In order to avoid possible systematics due to misestimating the spallation backgrounds in this region, harder spallation cut with lower signal efficiency was applied to almost completely eliminate spallation events above 15.49 MeV.
The side-band region above 29.49 MeV is used as the reference for the atmospheric neutrino events. The efficiency in this region is stable and almost the same as for the 27.49–29.49 MeV bin.
4. Background Estimation
Major background sources in the signal energy region are atmospheric neutrino interaction events, reactor neutrino events, and decays of spallation isotopes produced by cosmic-ray muon events. The solar neutrinos, which predominantly interact via electron scattering in SK, have no subsequent neutron signal. Thus, requiring one neutron makes the solar neutrino background negligible. Another background source is accidental background events, which are a pair of prompt SHE events and a misidentified delayed signal due to PMT noise hits or radioactivity.
Spallation isotope decays other than the β + n decay are efficiently removed by the spallation event cut and neutron tagging. 8He, 11Li, 16C, and 9Li are the representative decay isotopes that undergo β + n decay. However, 11Li is negligible since it can be removed by the spallation cut efficiently due to its short life (T1/2 < 0.01 s), and its production yield is low (10−9 μ−1 g−1 cm2; Li & Beacom 2014). Also, 8He and 16C have a low production yield of about 0.23 × 0.16 and 0.02 × 10−7 μ−1 g−1 cm2 (Li & Beacom 2014). In contrast, 9Li has a relatively long life (T1/2 ∼ 0.18 s), so it is difficult to identify its parent muon. In addition, the production yield is sufficiently higher, about 1.9 × 0.51 × 10−7 μ−1 g−1 cm2 (Li & Beacom 2014), where 0.51 comes from the branching ratio for this decay mode. Thus, we consider only 9Li as the remaining background from the spallation isotope decays. The production yield of 9Li event is measured to be 0.86 ± 0.12 (stat. ) ± 0.15 (sys. ) kton−1 · day−1 by Zhang et al. (2016), and a prediction of the spectrum is given in Abe et al. (2021).
Reactor inevitably remains in the final signal candidates since the observed event is IBD, which is the same as our target signal. The spectrum and yield are evaluated by the SKReact code (Goldsack 2022) based on the reactor neutrino model from Baldoncini et al. (2015) in the first 7.5–9.5 MeV Erec bin. Their contribution above the 9.5 MeV Erec bin is negligible.
The number of atmospheric neutrino background events is estimated by the simulation based on the HKKM flux (Honda et al. 2007, 2011) as the neutrino flux and NEUT 5.4.0.1 (Hayato 2009; Hayato & Pickering 2021) as the neutrino interaction simulator. Below 16 MeV of Erec, nuclear deexcitation gamma-rays from neutral-current quasi-elastic (NCQE) interactions dominate (Ankowski et al. 2012). NCQE interactions induce prompt gamma-rays and nuclei production. Above 16 MeV, it is dominated by charged-current quasi-elastic (CCQE) interactions with neutron production. Most of the events with Erec from 29.5 to 79.5 MeV consist of the decay electron from an unobserved muon originating from CCQE interacting atmospheric neutrinos. Since its energy distribution is the well-known Michel spectrum, the flux of atmospheric neutrino CCQE events in our MC is scaled by fitting the spectrum of the 29.5 < Erec < 79.5 MeV side-band region to the data.
The number of accidental coincidence background events Bacc is estimated as
where εmis is the neutron misidentification probability described in Section 3, and represents the number of remaining observed events after all selection criteria except neutron tagging.
Systematic uncertainties are estimated for only signal energy regions. The uncertainties on the NCQE events, spallation 9Li, and reactor neutrinos are taken as estimated by Abe et al. (2021), as 68% below 15.49 MeV and 82% above 15.49 MeV, 60%, and 100% for the NCQE, 9Li, and reactor neutrino backgrounds, respectively. Other components, such as non-NCQE events and accidental coincidence events, are newly estimated from the observed data in SK-VI based on the same method as Abe et al. (2021), 44% and 4%, respectively.
5. Results
After all event selection criteria are applied, 16 events remain within the signal energy region in 552.2 day data. In this analysis, we adopt five separate bins of Erec, of widths 7.5–9.5, 9.5–11.5, 11.5–15.5, 15.5–23.5, and 23.5–29.5 MeV for the signal window. Also, the side-band region is separated into bins for each 10 MeV. Figure 2 shows the Erec spectrum of those events. This is also listed in Table 1.
Figure 2. Reconstructed energy spectra of the observed data and the expected background after data reductions with a linear (top) and a logarithmic (bottom) scale for the vertical axis. These include the signal energy region and the side-band region above 29.5 MeV. Each color-filled histogram shows the expected backgrounds. The error bars in the data points represent the statistical uncertainty estimated by taking the square root of the number of observed events. These background histograms are stacked on the other histograms. The hatched areas represent the total systematic uncertainty for each bin. The size of uncertainty for each background is mentioned in the main text. The red dotted–dashed line shows the DSNB expectation from the Horiuchi+09 model (Horiuchi et al. 2009), which is drawn separately from the stacked histogram of the estimated backgrounds.
Download figure:
Standard image High-resolution imageTable 1. Summary of Observed Events, Expected Background Events, and p-value for Each Erec Bin
| Erec (MeV) | Observed | Expected | p-value |
|---|---|---|---|
| 7.5–9.5 | 5 | 7.73 ± 2.54 | 0.798 |
| 9.5–11.5 | 5 | 4.14 ± 1.23 | 0.398 |
| 11.5–15.5 | 3 | 2.13 ± 0.59 | 0.359 |
| 15.5–23.5 | 2 | 0.98 ± 0.35 | 0.258 |
| 23.5–29.5 | 1 | 0.98 ± 0.41 | 0.597 |
Note. Errors for the expected background represent only the systematic uncertainty.
Download table as: ASCIITypeset image
The probabilities of finding the observed number of events due to the fluctuation of the background events (p-value) are evaluated for each bin. It is done by performing 106 pseudo experiments based on the number of observed events and expected background events and the systematic uncertainties of the latter. The obtained p-values are listed in Table 1. We conclude that no significant excess is observed in the data over the expected background since even the most significant bin has a p-value is 25.8%.
We set the upper limit for the number of signal excess over the expected background with a 90% confidence level (C.L.; N90). It is evaluated by the pseudo experiments using the number of observed events with these 1σ statistical uncertainties and the number of expected background events with their systematic uncertainties. Then we estimate the flux upper limit based on N90 of the observed event. Assuming there is no signal event, the upper limit on the flux for each bin is calculated as
Here, is the averaged total cross section of IBD for each energy bin, Np is the number of protons as a target in the 22.5 kton of the fiducial volume of SK, T is the live time of observation (552.2 days), is the averaged signal efficiency for each energy bin after all event selection criteria are applied as shown in Figure 1, and dE is the bin width at each bin. The neutrino energy Eν is calculated by Eν = Erec + 1.8 MeV. The total cross section is given by the calculation in Strumia & Vissani (2003).
The expected upper limit from the background-only hypothesis at 90% C.L., , is evaluated using the number of expected background events and their statistical uncertainty. Then we extract the expected flux sensitivity by replacing N90 with in Equation (2).
Figure 3 shows the upper limit of the flux extracted in this search with the range of expectations of modern DSNB models. The most optimistic expectation is Kaplinghat+00 (Kaplinghat et al. 2000), and the most pessimistic one is Nakazato+15 (Nakazato et al. 2015) with the assumption of normal mass ordering in whole energy ranges, respectively. The upper limit of the flux for each bin is summarized in Table 2.
Figure 3. Upper limits on the flux, calculated by Equation (2). The red lines show the observed (solid) and expected (dotted–dashed) 90% C.L. upper limit for SK-VI. The blue lines show the observed (solid) and expected (dotted–dashed) 90% C.L. upper limit for SK-IV Abe et al. (2021). The green line represents the 90% C.L. observed upper limit placed by KamLAND Abe et al. (2022c). The gray-shaded region represents the range of the modern theoretical expectation. The expectation drawn in the figure includes DSNB flux models (Hartmann & Woosley 1997; Malaney 1997; Kaplinghat et al. 2000; Ando et al. 2003; Horiuchi et al. 2009; Lunardini 2009; Galais et al. 2010; Nakazato et al. 2015; Horiuchi et al. 2018, 2021; Kresse et al. 2021; Tabrizi & Horiuchi 2021; Ekanger et al. 2022). Ando+03 model was updated in Ando (2005).
Download figure:
Standard image High-resolution imageTable 2. Summary Table of Upper Limits, Sensitivity, and Optimistic and Pessimistic DSNB Expectation from Kaplinghat et al. (2000) and Nakazato et al. (2015), Respectively
| Neutrino Energy | Observed upper limit | Expected sensitivity | Averaged theoretical expectation of DSNB | ||
|---|---|---|---|---|---|
| (MeV) | (cm−2 s−1 MeV−1) | (cm−2 s−1 MeV−1) | (cm−2 s−1 MeV−1) | ||
| SK-IV | SK-VI | SK-IV | SK-VI | ||
| 9.29–11.29 | 37.30 | 34.07 | 44.35 | 50.78 | 0.20–2.40 |
| 11.29–13.29 | 20.43 | 18.43 | 11.35 | 15.12 | 0.13–1.66 |
| 13.29–17.29 | 4.77 | 3.76 | 2.05 | 2.71 | 0.67–0.94 |
| 17.29–25.29 | 0.17 | 0.90 | 0.21 | 0.50 | 0.02–0.30 |
| 25.29–31.29 | 0.04 | 0.33 | 0.11 | 0.33 | <0.01–0.07 |
Download table as: ASCIITypeset image
6. Future Prospects
In 2022 June, the SK-Gd experiment was upgraded to the SK-VII phase, in which additional Gd was introduced into the detector, providing a mass concentration of approximately 0.03%. In this phase, neutron tagging efficiency is expected to be over 55% while having comparable εmis with SK-VI, leading to 1.5 times higher sensitivity for the in the case of the same live time as for SK-VI. Furthermore, more efficient noise reduction by neutron tagging will enable a lower energy threshold. Hence we can search in a lower-energy region, which will increase signal acceptance for DSNB, solar antineutrinos, and light-dark-matter searches.
7. Conclusions
We searched for astrophysical , using the SK-VI data below 29.5 MeV for Erec between 2020 August and 2022 May, with 0.01% Gd mass concentration. This is an independent data set from the previous SK-IV search (Abe et al. 2021), using the data taken with pure water. In this analysis, a brand-new method for tagging neutrons using the signal of neutron capture on Gd is utilized so that the efficiency of neutron tagging is twice as high while keeping a low-misidentification probability. No significant excess above the expected backgrounds at greater than 90% C.L. level for five separate energy bins between 9.3 and 31.3 MeV is found. Thus, we placed the upper limits on the flux for the observed upper limit and the expected sensitivity. The sensitivity of this work is comparable to the previous SK-IV search (Abe et al. 2021) with a live time of 2970 days, which is the world's most sensitive search above 13.3 MeV, even though the live time of 552.2 days is about 5 times smaller. The result was achieved by neutron tagging with Gd signal and lowered probability of accidental coincidences thanks to the benefit of introducing Gd.
We gratefully acknowledge the cooperation of the Kamioka Mining and Smelting Company. The Super-Kamiokande experiment has been built and operated from funding by the Japanese Ministry of Education, Culture, Sports, Science and Technology, the U.S. Department of Energy, and the U.S. National Science Foundation. Some of us have been supported by funds from the National Research Foundation of Korea (NRF-2009-0083526 and NRF 2022R1A5A1030700) funded by the Ministry of Science, ICT, the Institute for Basic Science (IBS-R016-Y2), and the Ministry of Education (2018R1D1A1B07049158, 2021R1I1A1A01042256, the Japan Society for the Promotion of Science, the National Natural Science Foundation of China under grants No. 11620101004, the Spanish Ministry of Science, Universities and Innovation (grant PID2021-124050NB-C31), the Natural Sciences and Engineering Research Council (NSERC) of Canada, the Scinet and Westgrid consortia of Compute Canada, the National Science Centre (UMO-2018/30/E/ST2/00441) and the Ministry of Education and Science (DIR/WK/2017/05), Poland, the Science and Technology Facilities Council (STFC) and GridPP, UK, the European Union's Horizon 2020 Research and Innovation Programme under the Marie Sklodowska-Curie grant agreement No. 754496, H2020-MSCA-RISE-2018 JENNIFER2 grant agreement No. 822070, and H2020-MSCA-RISE-2019 SK2HK grant agreement No. 872549.