Prospects for long-range reactor monitoring with gadolinium-loaded water-Cherenkov detectors
PPramana – J. Phys. (2018) :
Prospects for long-range reactor monitoring with gadolinium-loadedwater-Cherenkov detectors
MICHAEL LEYTON and STEPHEN DYE Institut de F´ısica d’Altes Energies (IFAE), Barcelona Institute of Science and Technology Department of Physics and Astronomy, University of Hawaii * Corresponding author. E-mail: [email protected]
Abstract.
Antineutrino detectors are practical, non-intrusive tools capable of remotely monitoring the activity ofnuclear reactors. Here we explore the sensitivity of the Super-Kamiokande water-Cherenkov detector, followinggadolinium loading, to antineutrinos from a nuclear reactor complex at a distance of approximately 190 km. Thelivetimes required to observe the two currently operating cores in the reactor complex depend on the activity ofother reactors in the vicinity, as well as on estimates of detection e ffi ciency and background rates. Under reasonableassumptions, we find that gadolinium-loaded Super-Kamiokande could detect the flux of antineutrinos from bothcores at the Takahama reactor complex at 95% confidence level in 50 (10) live days 95% (50%) of the time, or theflux from one core in 397 (73) live days, provided that each core is operating at nominal power. Keywords.
Reactor, monitoring, non-proliferation, antineutrino, water-Cherenkov, Super-Kamiokande
PACS Nos1. Introduction
Reactor safeguards aim to detect the diversion of fis-sile materials from civil nuclear reactor facilities intoweapons programs. In comparison to current safeguards,which rely on bookkeeping and surveillance, reactormonitoring with antineutrino detectors can provide amore direct, and less intrusive, way to measure the op-eration of reactors and the evolution of their fuel. Astrategic goal of reactor monitoring with antineutrinosis, therefore, to remotely detect a change in the opera-tional status of a reactor.Water-Cherenkov detectors are well suited for thispurpose since the technology is scalable to the verylarge target masses ( ∼ Mtonne) needed for long-range( >
100 km) reactor monitoring. The Super-Kamiokandeneutrino observatory [1] is a water-Cherenkov detectorlocated 1 km underground at the Kamioka Observatoryin Japan, with a fiducial mass of 22.5 ktonne. It willsoon undergo a significant upgrade, adding a small frac-tion ( ∼ ffi ciently identified, provid-ing a significant boost in sensitivity to neutrino-inducedinverse beta decay reactions ( ν e p → e + n ).Due to the Fukushima-Daiichi disaster in 2011, onlya handful of reactors in Japan are currently in opera- tion: two at the Sendai reactor complex (829 km fromKamioka); one at Ikata (559 km from Kamioka); andtwo at Takahama (191 km from Kamioka). This mini-mal level of reactor activity in the vicinity of Kamiokaprovides a unique window in which to demonstrate thecapability of the upgraded Super-Kamiokande detector,here referred to as SuperK-Gd, to monitor reactors atlong range.In this paper, we examine the prospects of SuperK-Gd to monitor cores at Takahama, the reactor complexclosest to Kamioka currently in operation. We calculatethe livetime needed for SuperK-Gd to detect the fluxof antineutrinoss produced by reactors at Takahama at95% confidence level. The Takahama-Kamioka base-line distance of 191 km is approximately an order ofmagnitude larger than that of proposed reactor monitor-ing demonstration experiments using Gd-doped water-Cherenkov detectors [3], strengthening the argument,and widening the possible use cases, for reactor moni-toring with antineutrinos. Since preparation of the analysis, it was announced that two reac-tors at Ohi will restart in 2018. Reactors at Ohi are not consideredhere, but will be studied in a future publication.
Pramana – J. Phys. (2018) :
2. Antineutrino spectra
Nuclear power reactors generate heat via neutron-induced fissions of uranium (U) and plutonium (Pu)and the subsequent decays of unstable byproducts. Thefission fragments are produced in highly excited statesthat decay principally by beta ( β − ) decay, emitting ∼ ν e ) per fission on average. In a typicalreactor, more than 99.9% [4] of ν e s above the thresh-old of inverse beta decay on free protons (1.806 MeV)originate in the fission process of four isotopes: U, U, Pu and
Pu.We construct ν e emission spectra for each of thefour principal contributing isotopes using datasets andfits from ILL [5, 6, 7, 8], Mueller et al. [9], Huber [10]and Vogel et al. [11, 12]. Details of the calculation willbe published at a later date. The final ν e spectra for eachof the isotopes, with error bands, are shown in Fig. 1. (MeV) n E2 3 4 5 6 7 8 N eu t r i no s / f i ss i on / M e V U U U U Pu Pu Pu Pu Average Vogel et al.
Figure 1 . Reference isotopic ν e spectra per fission , shown with1 σ error band, for U (blue),
U (green),
Pu (red) and
Pu(magenta). Also shown for comparison are ν e spectra from Vogel etal. [11, 12] for U (blue circles),
U (green squares),
Pu (redtriangles) and
Pu (magenta diamonds).
We then construct reactor ν e emission spectra forthree categories of reactors, grouped according to theircooling and moderating materials: I) pressurized waterreactors (PWRs), boiling water reactors (BWRs), fastbreeder reactors (FBRs), light water-cooled graphite-moderated reactors (LWGR) and gas-cooled reactors(GCRs); II) pressurized heavy-water reactor (PHWRs);or III) reactors burning mixed oxide fuel (MOX), as-sumed to provide 30% of the total power, while the re-maining 70% is produced by standard (category I) fuel.The reactor ν e emission spectra are proportional to thetotal thermal power output ( P th ) of the reactor and thefraction of the total thermal power produced by eachof the four contributing isotopes, summed over all fourisotopes. Although power fractions are generally time-dependent quantities that evolve over the reactor oper-ation cycle, here we assume power fractions [13, 14] that correspond to the midpoint of the operation cyclefor simplicity. Fig. 2 shows the calculated ν e emissionspectra for the three reactor categories considered here,per unit of thermal power output. (MeV) n E2 3 4 5 6 7 8 / M e V t h N eu t r i no s / M e V - - - - -
10 1
III) MOX x 100II) PHWR x 10I) PWR/BWR/FBR/LWGR/GCR (MeV) n E2 3 4 5 6 7 8 R a t i o Figure 2 . Reactor ν e emission spectra per unit of thermal poweroutput for I) pressurized water (PW), boiling water (BW), fastbreeder (FB), light water-cooled graphite-moderated (LWG) andgas-cooled (GC) reactors (red); II) pressurized heavy-water (PHW)reactors (blue); and III) reactors burning mixed oxide fuel (MOX).Error bands include uncertainties on isotope spectra, thermal energyreleased per fission and a ±
5% uncertainty on mid-cycle power frac-tions from [13, 14]. The bottom plot shows the 1 σ error band (blue)on the ν e emission spectrum for category I reactors and the ratioof ν e emission spectra for category I (blue), II (green) or III (red)reactors over that for category I reactors.
3. SuperK-Gd simulation
We simulate the expected performance of the upgradedSuperK-Gd detector, based on existing data from thefourth phase of Super-Kamiokande (SK-IV) and tak-ing into account the impact of adding 0.2% (by mass)gadolinium sulfate, Gd (SO ) , to the ultra-pure water.Gd has a neutron capture cross section ∼ γ ) cascade of 8 MeV, allowing neutrons to be easilydetected and low-energy ν e s to be e ffi ciently identifiedvia the delayed coincidence of neutrons produced ininverse beta decay reactions. The energy spectrum ofthe γ rays emitted by neutron capture on Gd has beenmeasured in a test vessel to have a mean energy of4 . ± . ramana – J. Phys. (2018) : γ s. The triggere ffi ciency from the Super Low Energy (SLE) triggerof SK-IV is then applied by fitting the values givenin [16] (84% between 3.49 and 3.99 MeV, 99–100%above 3.99 MeV). The reconstructed positron kineticenergy T reco e is calculated by smearing the total positronenergy E e according to the resolution parameterizationfrom SK-IV [16]: σ ( E e ) = − . + . √ E e + . E e .All events with T reco e < . ffi ciency, ranging from 80% at 3.0 MeV, upto 95% above 5.0 MeV. A Gd capture e ffi ciency of 90%and neutron reconstruction e ffi ciency of 90% are alsoapplied [17]. Finally, an acceptance of 80% [17] is ap-plied in order to simulate the e ff ect of a spallation cut,used to remove β and / or γ decays of radioactive ele-ments produced by cosmic-ray muons that break up anoxygen nucleus in the target.We assume a fiducial volume of 22.5 ktonne [16] toreject events near the wall of the detector and ignorethe small reduction in fiducial volume for low-energyevents.
4. Interaction and event rates
We calculate the cross section of quasi-elastic neutrino-proton scattering, ν e p → e + n , also referred to as in-verse beta decay (IBD), from the V-A theory of weakinteractions, neglecting energy-dependent recoil, weakmagnetism and radiative corrections [18]. Assumingthe nucleon mass is infinite, then the energy of the in-cident neutrino E ν relates to the energy of the positron E e by E ν = E e + ∆ , where ∆ = m n − m p is the neutron-proton mass di ff erence [18].The number of IBD interactions induced by a re-actor with thermal power P th , as a function of stando ff distance L and antineutrino energy E ν , is calculated as: N ( L , E ν ) = n p τ P th π L (cid:90) σ IBD ( E ν ) dRdE ν P e → e ( L , E ν ) dE ν , (1)where n p is the number of free protons (e.g. hydrogennuclei), τ is the exposure time, dR / dE ν is one of the re-actor ν e emission spectra from Fig. 2, P e → e ( L , E ν ) is theneutrino survival probability after traveling a distance L and σ IBD ( E ν ) is the quasi-elastic neutrino-proton scat-tering cross section.Thermal power outputs ( P th ) of reactor cores cur-rently in operation worldwide are taken from 2016 datafrom [4, 19], updated with three additional reactors that have come online since the beginning of 2017. The dis-tance of each reactor core to Kamioka is calculated as-suming a spherical Earth.Figure 3 shows the expected interaction rate from asingle reactor core ( P th = th ) at Takahama, asa function of true positron kinetic energy T e , comparedwith that from all background nuclear reactor cores.The SuperK-Gd simulation, detailed in the previous sec-tion, is then applied on the interaction rates in order tocalculate the expected event rates, shown in Fig. 3 as afunction of reconstructed positron kinetic energy T reco e .The calculations here estimate 7.36 events / month be-tween 2 < T reco e < / month from all background nu-clear reactors. These estimates assume that each reactorcore is operating continuously at nominal power. (MeV) e(reco) T0 1 2 3 4 5 6 7 8 R a t e ( / m on t h / M e V ) Takahama-3 genTakahama-3 recoReactor background genReactor background reco gen x 100000 e n Atmospheric reco x 100000 e n Atmospheric
Figure 3 . Interaction and event rates per month at SuperK-Gdfor a single reactor core (2660 MW th ) at Takahama (red), comparedwith background from worldwide nuclear reactors (blue) and atmo-spheric ν e s (green). Interaction rates (solid lines) from Equation 1are calculated including the e ff ect of neutrino oscillation and shownas a function of true positron kinetic energy T e . Event rates (dashedlines) assume the detector performance discussed in Section 3 andare shown as a function of reconstructed positron kinetic energy T reco e . Atmospheric ν e rates have been scaled by a factor of 100,000.
5. Non-reactor backgrounds
In addition to the ν e background due to other reactors,both neutrino and non-neutrino sources can contributeto irreducible backgrounds in SuperK-Gd.5.1 Neutrino backgrounds
We estimate the rates of IBD reactions due to geo- ν e s,supernova relic neutrinos and atmospheric neutrinos viathe charged current channel, as well as the rate of neu-tral current interactions from atmospheric neutrinos.Geo-neutrinos are emitted by radioactive decays inthe Earth’s crust and mantle, but have energies below Pramana – J. Phys. (2018) : ν e s atKamioka from [20], we calculate the expected eventrate at SuperK-Gd to be < × − / month.Supernova relic neutrinos (SRNs), emitted by corecollapse supernovae throughout the universe, are ex-pected to contribute [17] 0.5–8.1 events / year between9 . < T e < . ffi ciency. Scaling to the energy range of interestfor this analysis, 2 < T e < / year at SuperK-Gd, or 0 . ± .
09 events / month.Charged current (CC) interactions of atmospheric ν e s are indistinguishable from reactor ν e s event-by-event,but their spectra are very di ff erent since the atmosphericneutrino flux is suppressed at low energies ( E ν (cid:46) E ν . We estimate the backgrounddue to atmospheric ν e CC interactions at SuperK-Gdusing 3D flux calculations from Bartol [21], averagedover all directions. Assuming average survival proba-bility after oscillation, P e → e = (1 − . θ ) = . < . T e < ν e s are shown in Fig. 3 as a function of T reco e .Atmospheric ν µ s can create a muon via a CC inter-action, whose decay electron may not be correlated tothe muon if the muon is below the Cherenkov thresh-old. However, the peak of the decay electron energyspectrum is closer to half the muon mass and the con-tribution below 8 MeV is expected to be negligible.Neutrino-nucleus neutral-current (NC) interactionsof atmospheric neutrinos pose an additional backgroundto the IBD signal from reactor ν e s. Such interactionscan readily be identified in water-Cherenkov detectorssince they produce nucleons or γ rays originating fromnuclear deexcitation. Quasi-elastic (QE) nucleon knock-out, e.g. ν x + O → ν x + p + N ∗ or ν x + O → ν x + n + O ∗ , becomes important for E ν (cid:38)
200 MeV [22].The latter reaction, in particular, can emit a neutronwith associated γ rays with energies between 1 and 10MeV [23], closely mimicking the IBD + Gd captureevent signature.To calculate the expected number of neutrino-nucleus interactions, we use the partial cross sectionsfor neutral-current neutrino-induced reactions on Ocalculated in [24] and the flux spectra of atmosphericneutrinos from Bartol [21]. The expected interactionrate at SuperK-Gd, summed over electron and muonflavors, is 0.055 events / month, or less than 0.035 events / month after applying Gd capture, neutron reconstruc-tion and spallation cut e ffi ciencies. 5.2 Non-neutrino backgrounds
To mimic the IBD + Gd neutron capture event signa-ture, non-neutrino backgrounds must produce twoflashes of Cherenkov light within ∼ µ s. Here we es-timate the rate of non-neutrino backgrounds, includingcosmogenic Li, fast neutrons, accidental coincidencesand spontaneous fission of
U.Cosmic-ray-muon spallation byproducts can fakean inverse beta decay reaction if they decay and emit β s and γ s. The largest contributor to cosmic-ray-muon-induced background for our energy range of interest is Li, whose decay mode with β − n emission can producedetectable energy up to ∼
12 MeV. The background fromcosmic-ray-muon-induced Li events was measured inSK-IV [17] down to T β = . Li, assuming 80% neutron tagging e ffi -ciency and 0.5% probability of muon-induced Li eventleakage, is expected to be 0 . ± . ± . / year be-tween 9 . < T β < . β − spectrum of Li decay modes leading to β − n emission, applying positron threshold, trigger ef-ficiency and energy resolution, as per the SuperK-Gdsimulation described in Section 3, and scaling such thatthe integral of the region 9 . < T β < . / year. The estimated rate below 8 MeV atSuperK-Gd is 0 . ± .
28 events / month.Cosmic ray muons can also produce fast neutronswhen they traverse the detector or external rock of theunderground cavern. Here we assume that neutronsproduced by muons traversing the detector can be taggede ff ectively using the active veto and removed from theevent sample. Although the outer detector volume canserve as a passive shield, neutrons produced in the ex-ternal rock are more di ffi cult to tag in coincidence withthe primary muon due to the hard energy spectrum andlong propagation range.We estimate the rate of events due to cosmogenicfast neutrons that could fake the IBD event signature byproducing two flashes of Cherenkov light within ∼ µ s, here referred to as ‘di-neutron’ events. We use pre-dictions of the neutron flux and energy spectrum at therock / cavern boundary at Kamioka by Mei & Hime [25].Assuming a dome-shaped cavern with a diameter andheight of 40 m, we estimate an incident rate of 16,590neutrons / day, for E n > . ± . / monthwill create a prompt and delayed pair in the SuperK-Gd fiducial volume within ∼ µ s, of which 8 . ± . / month will be reconstructed between 2 and 8 ramana – J. Phys. (2018) : ffi cult toestimate properly without a full simulation since it de-pends on the composition of the detector and environ-ment, vertex position in the detector and event selectioncriteria. Here we assume that the singles rate abovethreshold after all selection is (cid:46) ff ectively be ignored, buta more detailed calculation is warranted.Finally, we estimate the background due to sponta-neous fission of U, resulting from the addition ofGd (SO ) to the ultra-pure water target [26]. Assum-ing a background of 0.11 events per year per mBq / kg ofGd (SO ) and an activity of 20 mBq / kg, we calculatean upper limit of 0.183 events / month at SuperK-Gd.
6. Sensitivity analysis
We calculate the livetime needed for SuperK-Gd to de-tect the flux of reactor ν e s produced by one or two coresat Takahama at 95% confidence level (CL), for each ofthe following three scenarios:A two cores at Takahama operating at ≥ nominalpower;B one core at Takahama operating at ≥ nominalpower and the other core completely o ff (e.g. formaintenance or refueling);C same as scenario A, but one core treated as ‘sig-nal’ and the other as ‘background’.All other (452) reactor cores currently in operation [4,19] are treated as background and conservatively as-sumed to be operating at nominal power. Compilingmonthly data from 2003–2010 from [19], we find thatthe two Takahama cores were operating in scenario Aor C (B) for 47% (29%) of calendar months. For 21% ofcalendar months, one core was operating at ≥ nominalpower and the other was partially on (nominal power < ff . For no calendarmonths were both cores entirely o ff .We use the methodology developed in [20]. Foreach scenario and for a given livetime, we run a set of1000 pseudo-experiments according to the event ratesand spectra calculated in Sections 4 and 5. The num-ber of events in each pseudo-experiment is fluctuatedaccording to the statistical error on the predicted eventrates and scaled to the total livetime of the experiment.For each set of pseudo-experiments, we use a profile likelihood statistic to calculate the 95% confidence in-tervals, assuming that the data contain both signal andbackground events. We then assess the livetime at which95% (50%) of pseudo-experiments exclude the null(background-only) hypothesis at 95% CL. An examplepseudo-experiment for scenario A is shown in Fig. 4. (MeV) recoe T2 3 4 5 6 7 8 N u m be r o f E v en t s Pseudo-dataTakahama-3+4Fast neutron bkgndLi bkgnd Global reactor bkgndJapanese reactor bkgnd
Scen. A, 2 months
Figure 4 . Positron kinetic energy spectrum from IBD interac-tions and correlated backgrounds at SuperK-Gd (scenario A).
Signal events induced by ν e s from both Takahama reactor cores (redsolid) are shown together with background events induced by ν e sfrom Japanese reactors (purple fine hatched) and all other globalreactors (green hatched). All reactor cores are assumed to be oper-ating continuously at nominal power. Di-neutron background dueto cosmic-ray-induced fast neutrons (blue striped) and leakage ofbackground from Li decaying via β − n (orange weave) are alsoshown. Pseudo-experimental data (black squares) are shown withstatistical and systematic uncertainties added in quadrature. Distri-butions have been normalized to a livetime of 2 months, includingthe e ff ect of neutrino oscillation. Systematic uncertainties on both signal and back-ground distributions are applied when calculating thelikelihood statistic. To cover measurement uncertain-ties, we include the (symmetrized) energy-uncorrelatedsystematic uncertainties measured by SK-IV [16], whichinclude uncertainties on trigger e ffi ciency, reconstruc-tion quality, event selection and cross section, decreas-ing from 5.0% below 4 MeV to 0.9% for 7.5–8.0 MeV.We also include uncertainties on energy scale ( ± . ± . . < E ν < . < E ν < ν e fluxes uncertainties on oscillation param-eters, conversion of reactor power to flux, power valuereported by plant operators and seasonal changes in thereactor power output due to load requirements on theuser side, refueling or long-term shutdown.The results of the sensitivity analysis are shown inFig. 5. Pramana – J. Phys. (2018) :
Livetime (days)10 20 30 40 50 60 70 80 90 100 S i gna l po w e r ( G W ) Scenario A: Takahama-3+4 95% CL intervalNominal power output S i gna l s t r eng t h (a) scenario A Livetime (days)50 100 150 200 250 300 S i gna l po w e r ( G W ) Scenario B: Takahama-3 (4 off) 95% CL intervalNominal power output S i gna l s t r eng t h (b) scenario B Livetime (days)50 100 150 200 250 300 350 400 450 500 S i gna l po w e r ( G W ) Scenario C: Takahama-3 (4 bkgnd)95% CL intervalNominal power output S i gna l s t r eng t h (c) scenario C Figure 5 . Confidence interval at SuperK-Gd vs. livetime.
Mean 95% confidence interval for signal + background pseudo-experimentsfor three scenarios. In scenario A (a), both reactor cores at Takahama are on, operating at nominal power and treated as signal. In scenarioB (b), one core is operating at nominal power and the other core is completely o ff . In scenario C (c), both cores are on and operating atnominal power, but one is treated as signal and the other as background. The star- (diamond-) shaped markers represent the exposures atwhich 95% (50%) of pseudo-experiments exclude the null (background-only) hypothesis at 95% confidence level. Each curve has been fitwith a line plus exponential.
7. Conclusion
We study the sensitivity of Super-Kamiokande, follow-ing gadolinium loading, to antineutrinos from the Taka-hama reactor complex at a distance of approximately190 km. Assuming modest detector performance andreasonable estimates of background rates, we find thatgadolinium-loaded Super-Kamiokande could detect theflux of ν e s from both cores operating at nominal power(2.66 GW th each) at 95% confidence level in 50 (10)live days 95% (50%) of the time. If one core is o ff for refueling or maintenance, the flux of antineutrinosfrom the other core operating at nominal power couldbe detected in 240 (48) live days at 95% confidencelevel 95% (50%) of the time. Finally, the flux of an-tineutrinos from a single core could be detected in thepresence of antineutrinos from the other core at 95%confidence level in 397 (73) live days 95% (50%) ofthe time, assuming both cores are operating at nominalpower. These results suggest that gadolinium-loadedwater-Cherenkov detectors are capable of observing theoperational status of a single core within a reactor com-plex of known power at a range of ∼
200 km.
Acknowledgements
M.L. acknowledges support from the Marie Skłodowska-Curie Fellowship program, under grant 665919 (EU,H2020-MSCA-COFUND-2014), Ministerio de Econo-mia, Industria y Competitividad (MINECO), AgenciaEstatal de Investigaci´on (AEI) and Fondo Europeo deDesarrollo Regional (FEDER), under grants FPA2014-77347-C2-2 and SEV-2012-0234.
References [1] Y. Fukuda et al.
Nucl. Instrum. Meth.
A501:418–462 (2003).[2] J. F. Beacom and M. R. Vagins.
Phys. Rev. Lett.
Preprint ins-det / (2015).[4] M. Baldoncini et al. Phys. Rev.
D91:065002 (2015).[5] F. von Feilitzsch, A. A. Hahn, and K. Schreckenbach.
Phys.Lett. B
Phys. Lett. B
Phys. Lett. B
Phys. Rev. Lett.
Phys. Rev.
C83:054615 (2011).[10] P. Huber.
Phys. Rev.
C84:024617 (2011). [Erratum: Phys.Rev. C85:029901(2012)].[11] P. Vogel, G. K. Schenter, F. M. Mann, and R. E. Schenter.
Phys. Rev.
C24:1543–1553 (1981).[12] P. Vogel and J. Engel.
Phys. Rev.
D39:3378 (1989).[13] G. Bellini et al.
Phys. Lett.
B687:299–304 (2010).[14] G. Bellini et al.
Phys. Lett.
B722:295–300 (2013).[15] H. Watanabe et al.
Astropart. Phys.
Phys. Rev.
D94(5):052010 (2016).[17] Y. Zhang et al.
Phys. Rev.
D93(1):012004 (2016).[18] P. Vogel and J. F. Beacom.
Phys. Rev.
D60:053003 (1999).[19] Power Reactor Information System. (2017).[20] M. Leyton, S. Dye, and J. Monroe.
Nat. Commun.
Phys. Rev.
D70:023006 (2004).[22] A. M. Ankowski et al.
Phys. Rev. Lett.
Phys. Rev.
D90(7):072012 (2014).[24] E. Kolbe, K. Langanke, and P. Vogel.
Phys. Rev.
D66:013007(2002).[25] D. Mei and A. Hime.
Phys. Rev.
D73:053004 (2006).[26] A. Renshaw.