Search for Low-energy Electron Antineutrinos in KamLAND Associated with Gravitational Wave Events

In 2015, gravitational waves (GWs) were first detected by the Advanced Laser Interferometer Gravitational-wave Observatory (LIGO) (Abbott et al. 2016). This event was shown to have originated from the merger of a binary black hole (BBH) system. Nearly two years earlier, the IceCube collaboration published the first observational evidence for high-energy astrophysical neutrinos (Aartsen et al. 2013). The gravitational and weak forces along with the electromagnetic were added to the astronomical observations, beginning a new era of extra-galactic multi-messenger astronomy.

In 2017, LIGO detected an event consistent with a comparably nearby binary neutron-star (BNS) merger (Abbott et al. 2017a). Within seconds of the GW, the electromagnetic counterpart was observed by the Fermi Gamma Ray Burst Monitor (Abbott et al. 2017b), making this the first GW multi-messenger event. The online neutrino telescopes—including IceCube, ANTARES, and the Pierre Auger Observatory—did not detect any directionally coincident high-energy (GeV–EeV) neutrinos or an MeV neutrino burst signal (Albert et al. 2017a). While no coincident neutrinos were found, this is consistent with model predictions for the merger (Kimura et al. 2017). In contrast with BBH mergers, BNS mergers are expected to emit neutrinos at both GeV and MeV energies (Mészáros 2017). MeV-scale neutrinos would be produced by the hot collapsing fireball at the beginning of a gamma-ray burst (Sahu & D'Olivo 2005), so we can be confident that they must be produced when there is collapsing matter outside of a black hole. These neutrinos are modeled in several ways, but have energies on the order of the dominant photon energy, and are considerably more numerous than the emitted photons (Halzen & Jaczko 1996). In the case of a post-merger neutron star remnant, thermal emission of MeV neutrinos is expected as the remnant cools (Foucart et al. 2016). The neutron rich environment also suggests a brighter ${\bar{\nu }}_{e}$ flux than the ν_e flux (Kyutoku & Kashiyama 2018).

Recently, the LIGO/Virgo collaboration published their event catalog (Abbott et al. 2019), including the full data set from their first and second observing runs, LIGO-O1 and LIGO-O2, respectively. During the third observing run, LIGO-O3, the LIGO/Virgo collaboration initiated the online GW candidate event database (GraceDB) (LIGO Scientific Collaboration 2020), providing public alerts and a centralized location for aggregating and retrieving event information. For such transient GW events, various neutrino detectors reported correlation searches: Super-Kamiokande (Abe et al. 2016, 2018), Borexino (Agostini et al. 2017), NOvA (Acero et al. 2020), Bikal-GVD Neutrino Telescope (Avrorin et al. 2018), Daya Bay (An et al. 2020), XMASS (Collaboration et al. 2020), and IceCube/ANTARES (Adrián-Martínez et al. 2016; Albert et al. 2017b; Aartsen et al. 2020). The Kamioka Liquid scintillator AntiNeutrino Detector (KamLAND) has also performed a search for electron antineutrinos coincident with gravitational waves GW150914 and GW151226, and then candidate event LVT151012 (Gando et al. 2016a).

In this paper, we present an updated coincidence search for MeV-scale electron antineutrinos in KamLAND associated with the observed GW events in LIGO-O2 (2016 November 30–2017 August 25) and LIGO-O3 (2019 April 1–March 27).

KamLAND is a large volume liquid scintillator neutrino detector located at the Kamioka mine, 1 km underground from the top of Mt. Ikenoyama in Gifu Prefecture, Japan. The KamLAND detector consists of a cylindrical 10 m radius × 20 m height water-Cerenkov outer detector for cosmic-ray muon veto, a 9 m radius stainless steel spherical tank that mounts 1325 17 inch and 554 20 inch photomultiplier-tubes (PMTs), and a 6.5 m radius Nylon/EVOH outer balloon filled with approximately 1 kton of ultra-pure liquid scintillator. The liquid scintillator is composed of 20% Pseudocumene (1,2,4-Trimethylbenzene, C₉H₁₂), 80% Dodecane (N-12, C₁₂H₂₆), and 1.36 g l⁻¹ PPO (2,5-Diphenyloxazole, C₁₅H₁₁NO). Further details of the KamLAND detector are summarized in Suzuki (2014).

KamLAND began its data acquisition in 2002 March. The detector was upgraded in 2011 August to include a drop-shaped 1.5 m radius nylon inner balloon filled with approximately 400 kg of purified xenon loaded in a liquid scintillator (Gando et al. 2016b). In this configuration, known as the KamLAND-Zen 400 experiment, KamLAND searched for neutrinoless double-beta decay until 2015 December, at which point the inner balloon was removed. Subsequently in 2018 May, a further upgrade to the KamLAND-Zen 800 experiment ensued with the addition of a 1.9 m radius inner balloon, containing approximately 800 kg of purified xenon.

Electronic boards record the digitized PMT waveforms and provide the corresponding time stamp based on a 40 MHz internal clock. All internal clocks are synchronized to the Unix Time Stamp on every 32nd pulse per second (1 PPS) trigger from a Global Positioning System receiver, located at the entrance to the Kamioka mine. Uncertainties in the absolute trigger time stamp accuracy are less than ${ \mathcal O }(100)\,\mu {\rm{s}}$, derived from the signal transportation into the mine, optical/electrical signal conversion, and triggering electronics, which is negligibly small for this coincidence search.

The interaction vertex and energy deposition are reconstructed using the measured PMT charge and timing information. At low energies, the detector calibrations are performed using various radioactive sources: ⁶⁰Co, ⁶⁸Ge, ²⁰³Hg, ⁶⁵Zn, ²⁴¹Am⁹Be, ¹³⁷Cs, and ²¹⁰Po¹³C. At higher energies (>10 MeV), the energy response is calibrated using spallation-produced ¹²B/¹²N. Daily stability measurements are performed using the 2.2 MeV gamma-ray emitted from a spallation-neutron capture on a proton (Abe et al. 2010). The reconstructed energy and interaction vertex resolution are evaluated as $6.4 \% /\sqrt{E\,(\mathrm{MeV})}$ and $\sim 12\,\,\mathrm{cm}/\sqrt{E\,(\mathrm{MeV})}$ (Gando et al. 2013), respectively. The primary radioactive backgrounds found in the liquid scintillator are (5.0 ± 0.2) × 10⁻¹⁸ g g⁻¹ (93 ± 4 nBq m⁻³) of ²³⁸U and (1.8 ± 0.1) × 10⁻¹⁷ g g⁻¹ (59 ± 4 nBq m⁻³) of ²³²Th (Gando et al. 2015).

During the period in which LIGO-O2 and LIGO-O3 were collecting data, the KamLAND detector had an average livetime efficiency of _live = 0.878. For all but one GW event in LIGO-O2, GW170608, the KamLAND detector was actively taking physics data, whereas three GW events in LIGO-O3 (S191213g, S191215w, and S191216ap) overlapped with an unusual detector condition period. Tables 1 and 2 summarize the GW events used in this analysis during their respective observing runs, along with the KamLAND detector status.

Table 1. The Gravitational Wave Event List for LIGO-O2 (Abbott et al. 2019) and along with the KamLAND Detector Status

Gravitational Wave	Date and Time (UTC)	Distance (Mpc)	Source	KamLAND Status
GW170104	2017 Jan 4, 10:11:58.6	${990}_{-430}^{+440}$	BBH	running
GW170608	2017 Jun 8, 02:01:16.5	${320}_{-110}^{+120}$	BBH	unusual data condition
GW170729	2017 Jul 29, 18:56:29.3	${2840}_{-1360}^{+1400}$	BBH	running
GW170809	2017 Aug 9, 08:28:21.8	${1030}_{-390}^{+320}$	BBH	running
GW170814	2017 Aug 14, 10:30:43.5	${600}_{-220}^{+150}$	BBH	running
GW170817	2017 Aug 17, 12:41:04.4	${40}_{-15}^{+7}$	BNS	running
GW170818	2017 Aug 18, 02:24:09.1	${1060}_{-380}^{+420}$	BBH	running
GW170823	2017 Aug 23, 13:13:58.5	${1940}_{-900}^{+970}$	BBH	running

Note. The three events in which KamLAND has already published the results for a coincidence search (Gando et al. 2016a) are not included in this table.

Download table as:  ASCIITypeset image

In this analysis, we focus on KamLAND events induced by the electron antineutrino inverse beta-decay (IBD) reaction (${\bar{\nu }}_{e}+p\to {e}^{+}+n$) with 1.8 MeV neutrino energy threshold. The IBD candidate events can be selected by the delayed coincidence (DC) signature: scintillation light from the positron and its annihilation gamma-rays as a prompt signal, and a 2.2 MeV (4.9 MeV) gamma-ray from neutron capture on a proton (carbon-12) as a 207.5 ± 2.8 μs delayed signal (Abe et al. 2010). The incident neutrino energy (E_ν ) is computed from the reconstructed prompt energy (E_prompt) with energy and momentum conservation in the reaction as E_ν ≃ E_prompt + 0.78 MeV + T_n , where T_n represents the neutron kinetic energy.

The energy range of this analysis is selected to be E_prompt between 0.9 and 100.0 MeV, with a delayed neutron capture on a proton (carbon-12) energy between 1.8 and 2.6 MeV (4.4–5.6 MeV). Accidental backgrounds are suppressed by imposing a spatial and time correlation between the prompt and delayed signals. In particular, the reconstructed vertex and time difference between the prompt and delayed signals must be within 200 cm and 0.5–1000 μs of each other. All events must be reconstructed in the fiducial volume region 6 m radius from the center, corresponding to a total number of target protons of N_T = (5.98 ± 0.13) × 10³¹. Muon and spallation vetoes are applied after the interaction of a cosmic-ray muon, which occur at a rate of approximately 0.34 Hz in KamLAND. Further details regarding the event selection can be found in previous KamLAND analyses (Gando et al. 2011, 2013, 2016a; Asakura et al. 2015). A likelihood-based signal selection distinguishes electron antineutrino DC pairs from accidental coincidence backgrounds for a few to several MeV energy range. This has been updated from the previous analyses considering the accidental coincidence event rates, upgraded detector conditions of the outer detector refurbishment (Ozaki & Shirai 2017), inner balloon installation for KamLAND-Zen 800 (Gando 2020), and the activity of Japanese nuclear reactors.

From 2018 May onwards (the KamLAND-Zen 800 phase)—in order to avoid unexpected background contamination due to the xenon-loaded liquid scintillator, inner balloon body, and suspending ropes—the inner balloon region is vetoed for the delayed event. The inner-balloon cut regions are: a 2.5 m radius spherical volume centered in the detector and a 2.5 m radius vertical cylindrical volume in the upper-half of the detector. In this analysis, the effect of this additional inner balloon cut is considered as a selection efficiency suppression for the delayed event rather than a change in the number of target protons for the prompt event. Therefore, the total selection efficiencies are different between the KamLAND data sets corresponding to the periods operating during LIGO-O2 (without the inner balloon cut) and LIGO-O3 (with the additional inner balloon cut).

The selection efficiencies are shown in Figure 1, as a function of the reconstructed prompt energy (_s(E_prompt)). The structure of the efficiency suppression around E_prompt ≃ 2 MeV is primarily derived from the accidental background spectrum shape. Above E_prompt = 5.0 MeV, at which point the accidental background contamination becomes negligibly small, the selection efficiencies in each data set converge to 92.9% and 77.4%.

Zoom In Zoom Out Reset image size

The dominant neutrino sources below 8 MeV are the Japanese reactor power plants and geo-chemical radioactive decays in the Earth. After the Great East Japan Earthquake on 2011 March 11, most of the reactors in Japan were shut down and only a few have since been brought back online. Other backgrounds are DC pairs of accidental radioisotopes, spallation products ⁹Li, and ¹³C (α, n)¹⁶O reaction. Above ∼10 MeV, fast neutrons from cosmic-ray muons and atmospheric neutrino interactions are the dominant contribution to the background (Gando et al. 2012).

This analysis is performed using a coincident time window of ±500 s around each of the 60 GW events listed in Tables 1 and 2. The selected timing window is based on the largest expected time gap between GW events and neutrino events (Baret et al. 2011). This is sufficiently large to cover possible early neutrino emission scenarios as well as the neutrino time-of-flight delay from GW170729, the most distant GW source in this analysis. For example, assuming the sum of neutrino mass limits and cosmological constants from Aghanim et al. (2020), and the neutrino mass-squared splittings from Esteban et al. (2019), a neutrino with an energy of 1.8 MeV, upper mass state of 60 meV, traveling a distance of 2840 Mpc will be delayed by approximately 86 s relative to the GW.

The expected number of uncorrelated background events per ±500 s time window are estimated using off-time windows from the GW and found to be 4.08 × 10⁻³ and 4.27 × 10⁻³ for the KamLAND periods corresponding to LIGO-O2 and LIGO-O3, respectively.

No IBD electron antineutrino events were found in the KamLAND data set within ±500 s of each GW event. Using the uncorrelated accidental background rates and zero observed signal events, the Feldman–Cousins method (Feldman & Cousins 1998) is used to derive the 90% confidence level (C.L.) upper limit on the number of detected electron antineutrinos. This is found to be N₉₀ = 2.435 for each GW event in the LIGO-O2 and N₉₀ = 2.435 for each GW event in the LIGO-O3. The upper limit (F₉₀) can then be used to place constraints on the neutrino fluence, as follows:

$$\begin{eqnarray}&&{F}_{90}=\displaystyle \frac{{N}_{90}}{{N}_{T}\,{\epsilon }_{\mathrm{live}}\,\int {\epsilon }_{{\rm{s}}}({E}_{\mathrm{prompt}}({E}_{\nu }^{{\prime} }))\,\sigma ({E}_{\nu }^{{\prime} })\,\lambda ({E}_{\nu }^{{\prime} })\,{{dE}}_{\nu }^{{\prime} }},\end{eqnarray} \tag{ 1 }$$

where σ(E_ν ) is the IBD cross section (Strumia & Vissani 2003) and λ(E_ν ) is the neutrino energy spectrum. In order to perform a model independent analysis from the neutrino emission mechanisms for various GW sources, we assume a monochromatic neutrino energy spectra. Hence, we calculate 90% C.L. fluence upper limits on the electron antineutrinos for each GW event in LIGO-O2 and LIGO-O3 with

$$\begin{eqnarray}&&{F}_{90}({E}_{\nu })=\displaystyle \frac{{N}_{90}}{{N}_{T}\,{\epsilon }_{\mathrm{live}}\,{\epsilon }_{{\rm{s}}}({E}_{\nu })\,\sigma ({E}_{\nu })},\end{eqnarray} \tag{ 2 }$$

as shown in Figure 2. The resulting upper limits on the electron antineutrino fluence are found to be between 10⁸ and 10¹³ cm⁻².

Zoom In Zoom Out Reset image size

We study the neutrino emission energy scales between two cases of GW sources for officially published and detail-known events during LIGO-O2: the BNS merger (BNS: GW170817), and six BBH mergers (BBHs: GW170104, GW170729, GW170809, GW170814, GW170818, GW170823). Because of the ±500 s coincidence search timing window for each event, the total number of expected background events are 4.08 × 10⁻³ for the BNS event and 2.45 × 10⁻² for the six BBH candidates. Using the Feldman–Cousins method again with the 90% C.L., for zero events observed, the upper limit on the number of neutrino events is ${N}_{90}^{\mathrm{BNS}}=2.435$ and ${N}_{90}^{\mathrm{BBHs}}\,=2.415$ for the BNS and the BBHs, respectively. According to Kyutoku & Kashiyama (2018), the assumption that the neutrino energy spectrum has a Fermi–Dirac distribution is reasonable for exploring the mechanism of neutrino emissions from the GW sources. Assuming the Fermi–Dirac distribution, the neutrino energy spectra can be written as

$$\begin{eqnarray}&&{\lambda }_{\mathrm{FD}}({E}_{\nu })=\displaystyle \frac{1}{{T}^{3}{f}_{2}(\eta )}\displaystyle \frac{{E}_{\nu }^{2}}{{e}^{{E}_{\nu }/T-\eta }+1},\end{eqnarray} \tag{ 3 }$$

$$\begin{eqnarray}&&{f}_{n}(\eta )={\int }_{0}^{\infty }\displaystyle \frac{{x}^{n}}{{e}^{x-\eta }+1}{dx},\end{eqnarray} \tag{ 4 }$$

where we assume zero chemical potential and pinching factor η = 0, the temperature is given as T = 〈E〉/3.15, and the average neutrino energy 〈E〉 = 12.7 MeV (Caballero et al. 2016). Integrating between the true electron antineutrino energy limits, E_ν = 1.8–111 MeV, following Equation (1) and assuming equal contribution from six neutrino species, we obtain upper limits on the total fluence (${{ \mathcal F }}_{90}^{\mathrm{BNS},\mathrm{BBH}}$) in the Fermi–Dirac distribution case with 90% C.L. as

$$\begin{eqnarray}&&{{ \mathcal F }}_{90}^{\mathrm{BNS}}\leqslant 2.04\times {10}^{10}\,{\mathrm{cm}}^{-2}\end{eqnarray} \tag{ 5 }$$

for the BNS and

$$\begin{eqnarray}&&{{ \mathcal F }}_{90}^{\mathrm{BBH}}\leqslant 2.02\times {10}^{10}\,{\mathrm{cm}}^{-2}\end{eqnarray} \tag{ 6 }$$

for the BBH. Considering the luminosity distances from the GW source, we convert the total fluence (${{ \mathcal F }}_{90}$) to the total energy (${{ \mathcal L }}_{90}$) radiated in neutrinos from single source as

$$\begin{eqnarray}&&{{ \mathcal L }}_{90}^{\mathrm{BNS},\mathrm{BBH}}=\displaystyle \frac{{{ \mathcal F }}_{90}^{\mathrm{BNS},\mathrm{BBH}}}{1/(4\pi {D}_{\mathrm{eff}}^{2}\langle E\rangle )},\end{eqnarray} \tag{ 7 }$$

where D_eff is the effective distance defined as $1/{D}_{\mathrm{eff}}^{2}\,\equiv {\sum }_{i}1/{D}_{i}^{2}$ for every ith GW event, and the central values are used to D_i . Hence, the upper limits on the total energy are obtained as

$$\begin{eqnarray}&&{{ \mathcal L }}_{90}^{\mathrm{BNS}}\leqslant 7.92\times {10}^{58}\,\mathrm{erg}\end{eqnarray} \tag{ 8 }$$

based on the 40 Mpc distance to the BNS event, and

$$\begin{eqnarray}&&{{ \mathcal L }}_{90}^{\mathrm{BBH}}\leqslant 8.22\times {10}^{60}\,\mathrm{erg}\end{eqnarray} \tag{ 9 }$$

for the BBHs based on the effective distance of 407.6 Mpc, without accounting for neutrino oscillation effects. The observed upper limits are found to be larger than the typical total energy radiated from supernovae ${ \mathcal O }({10}^{53})$ erg (Bethe 1990).

This paper searched for coincident IBD electron antineutrinos in KamLAND with the 60 GW events associated with the second and third observing runs of the LIGO detector. No coincident signal was observed within a ±500 s timing window around each GW event. The 90% C.L. electron antineutrino fluence upper limit for each GW, assuming a mono-energetic neutrino flux, was presented for neutrino energies between 1.8 and 111 MeV. We set the most strict upper limit on each GW event in the LIGO-O2 data set below 3.5 MeV neutrino energies. For the LIGO-O3 data set, this is the first result of an MeV-scale energy coincidence neutrino search.

The obtained upper limits on the total energy radiated from GW source class, BNS or BBH, in LIGO-O2 with the assumption of a Fermi–Dirac neutrino energy distribution, are found to be 7.92 × 10⁵⁸ erg and 8.22 × 10⁶⁰ erg, respectively. These results depend on the number of GW events and distances. This limit will be improved once the candidate events of LIGO-O3 are published.

In the future, the mechanism of neutrino emission may be constrained and explored by combining with multi-messenger astronomy: GeV/TeV neutrino detectors, X-ray/gamma-ray telescopes, and gravitational wave detectors. The KamLAND detector continues to take physics data while running in the KamLAND-Zen 800 configuration and is monitoring for transient astrophysical events. The recently implemented online monitor at KamLAND (Asakura et al. 2016) also readily searches for correlations with transient events and reports the results to the Gamma-ray Coordinates Network (GCN) and/or the Astronomer's Telegram (ATel).

The KamLAND experiment is supported by JSPS KAKENHI Grants 19H05803; the World Premier International Research Center Initiative (WPI Initiative), MEXT, Japan; Netherlands Organisation for Scientific Research (NWO); and under the U.S. Department of Energy (DOE) Contract No. DE-AC02-05CH11231, the National Science Foundation (NSF) No. NSF-1806440, NSF-2012964, as well as other DOE and NSF grants to individual institutions. The Kamioka Mining and Smelting Company has provided service for activities in the mine. We acknowledge the support of NII for SINET4. This work is partly supported by the Graduate Program on Physics for the Universe (GP-PU), and the Frontier Research Institute for Interdisciplinary Sciences, Tohoku University.