Searching for solar KDAR with DUNE

The observation of 236 MeV muon neutrinos from kaon-decay-at-rest (KDAR) originating in the core of the Sun would provide a unique signature of dark matter annihilation. Since excellent angle and energy reconstruction are necessary to detect this monoenergetic, directional neutrino flux, DUNE with its vast volume and reconstruction capabilities, is a promising candidate for a KDAR neutrino search. In this work, we evaluate the proposed KDAR neutrino search strategies by realistically modeling both neutrino-nucleus interactions and the response of DUNE. We find that, although reconstruction of the neutrino energy and direction is difficult with current techniques in the relevant energy range, the superb energy resolution, angular resolution, and particle identification offered by DUNE can still permit great signal/background discrimination. Moreover, there are non-standard scenarios in which searches at DUNE for KDAR in the Sun can probe dark matter interactions.


Introduction
There has been recent interest from the experimental community in detecting the neutrinos produced by kaon decay at rest (KDAR) [1,2]. One application for these techniques is the search for neutrinos produced when gravitationally-captured dark matter annihilates in the core of the Sun [3][4][5]. If dark matter annihilation produces u, d, and s quarks, then the result of subsequent hadronization and fragmentation would be a large number of K + which come to rest in the dense solar medium before decaying. 64% of these decays (K + → µ + ν µ ) produce monoenergetic ν µ with an energy of ∼ 236 MeV [6][7][8]. The oscillations of these neutrinos while passing through the dense solar medium and vacuum results in approximately comparable fluxes of active neutrinos in all three flavors at Earth [9]. Recent work has focused on developing new techniques for utilizing the excellent particle identification and energy and angular resolution of DUNE to identify the energy and direction of the incoming 236 MeV neutrino [10]. The identification of a flux of 236 MeV neutrinos arriving from the Sun would be an extraordinary signal of new physics, providing a new handle on dark matter interactions which could be a unique probe of non-standard dark matter models [11]. This work further develops techniques for measuring the monoenergetic neutrinos arising from KDAR in the Sun, with a focus on increasing the signal-to-background ratio. At water Cherenkov (WC) neutrino detectors, it is very difficult to determine the direction of an O(100) MeV neutrino because the charged lepton produced by a charged-current interaction is largely isotropic at these energies. But in a large fraction of neutrino-argon CC-interactions, a proton is ejected preferentially in the forward direction. Though this proton cannot be seen in a WC detector, its energy and direction can be well-measured in a liquid argon time projection chamber (LArTPC) detector, such as DUNE. Thus, although WC detectors will generally have a statistical advantage due to their size, LArTPC detectors can have an advantage in reducing some systematic uncertainties, due to a greater ability to reject background.
In [10], it was proposed that one search for DUNE events with exactly one proton and one charged lepton with a total energy of 236 MeV, and with the proton directed away from the Sun. It was found that this directionality strategy should improve DUNE sensitivity to dark matter annihilation in the Sun, while yielding a signal-to-background ratio as high as ∼ 40%. In this paper, we use the LArSoft package [14] to realistically model the detector response, including the asymmetric response due to the orientation of the detector with respect to the incoming neutrino, and we use the Pandora package [16] to perform track reconstruction. We also find that, although the charged lepton is produced roughly isotropically, its direction is correlated with that of the proton, providing a new method for rejecting background that can significantly improve the signal-to-background ratio.
At DUNE, the charged current interaction ν + 40 Ar → − + p + + 39 Ar produces an ejected proton and charged lepton which can be well-measured [12]. But the recoil of the remnant 39 Ar will not be well-measured, and although the kinetic energy of the remnant nucleus will be small, its momentum may be substantial. But given a hypothesis for the energy and momentum of the neutrino (i. e., a 236 MeV neutrino arriving from the Sun), the momentum of the remnant nucleus can be reconstructed using momentum conservation. We find that when the struck proton is very forward-directed, the remnant nucleus is typically backscattered (more on this in Section 2 and Fig. 10). Utilizing this correlation, we find that for models where evidence can be found at 90% C.L. with a 400 kT yr exposure of DUNE, the signal-to-background ratio can be as high as 2.2.
We find that, with a 400 kT yr exposure, DUNE can probe O(10 3 ) m −2 s −1 fluxes of 236 MeV ν µ emanating from the Sun. As a specific example, we consider the case of lowmass dark matter (m 10 GeV) which scatters inelastically with nuclei. We estimate the sensitivity of DUNE to models which cannot be probed by direct detection experiments.
The plan of this paper is as follows. In Section 2, we describe our simulation framework and analysis cuts. In Section 3, we describe the resulting sensitivity to a flux of KDAR neutrinos, and as an example, interpret this as a sensitivity to a particular class of dark matter models which cannot be probed by direct detection experiments. We conclude with a discussion of our results in Section 4.

Event Simulation and Analysis Cuts
Dark matter annihilation at the core of the Sun can produce light mesons, whose decaysat-rest can produce monoenergetic neutrinos. KDAR (K + → µ + ν µ ) will produce a E ν = 236 MeV monoenergetic ν µ at the core of the Sun. On the other hand, K − and π − will tend to be Coulomb-captured by nuclei. Hence the flux of neutrinos from K − and π − is small [17]. π + decay-at-rest in the Sun will produce a monoenergetic 30 MeV neutrino. But this signal is less promising [8], because the background from atmospheric neutrinos is larger at these energies, while the ν − 40 Ar cross section is smaller. Moreover, the scattering of a 30 MeV neutrino is less likely to eject a proton, which is needed for directionality. Dark matter annihilation can also produce muons which decay at rest, but this signal is less promising because it does not yield a monoenergetic neutrino. As a result, we focus on the 236 MeV ν µ produced by KDAR in the Sun.
By the time this neutrino reaches Earth, it will have oscillated into all three flavors. But only ν µ and ν e can produce a charged-current interaction at this energy. In this analysis, we only consider ν µ . We are interested in charge-current events in which a muon is produced and a proton is ejected from the nucleus, since these particles can leave crisp tracks in DUNE, as shown in Fig. 1.   [18] generated proton and muon kinetic energies for the 1 proton + 1 muon = 2 total particles case. Generally, if we do not enforce a 2 particle cap, 13% of the CC events are multi-proton at generator level. 76% are single proton. 10% are without protons.

Event Generation
We use NuWro [18] to simulate neutrino-nucleus scattering events because it allows us to model the nuclear response using a spectral function to simulate the nucleus [20], rather than the Fermi Gas model . Final state interactions are modeled using an intra-nuclear cascade (INC) [21]. At 236 MeV, the NuWro neutrino event generator predicts a 4% MEC (meson exchange current) contribution, a 32% NCQE (neutral current quasi-elastic) contribution, and a 64% CCQE (charged current quasi-elastic) contribution to the neutrino-argon scattering cross section, with a negligible contribution for all other processes (pions are produced 0.04% of the time). However, neutral current interactions do not eject muons. We do not include NC in our analysis because we expect excellent muon identification in DUNE and hence very few NC events in which a muon is identified. This expectation is motivated by the success of the dE/dx vs. residual range method at ProtoDUNE-SP (as shown in [13]). At 236 MeV, neutrino charged-current interactions with nucleons are mostly quasi-elastic (CCQE), ν + n → − + p + . Fig. 2 shows the expected distribution, generated by NuWro, of the kinetic energies of the muons and protons produced by charged current interactions of a 236 MeV ν µ .
Thus, we are interested in charged-current quasi-elastic (CCQE) ν µ + 40 Ar interactions. We simulate CCQE signal events -236 MeV neutrinos arriving from the direction of the Sun -and background events (atmospheric neutrino events, assumed to be isotropic) in NuWro. We do not consider non-DM KDAR in the Sun as a background. True, cosmic rays impinge on the Sun and produce KDAR but this contribution is negligible [8].
For signal events, the neutrino is assumed to arrive from the direction of the Sun, but at a randomized time (which determines the orientation of the Sun with respect to the detector). For an atmospheric neutrino background event, the orientation of the neutrino with respect to the detector is randomized. The distribution of signal event directions relative to the detector are show in Fig. 3. In particular, and unlike atmospheric neutrinos, neutrinos arriving from the Sun cannot have an arbitrary orientation with respect to the detector wires, but must instead arrive from directions within the yellow band of Fig. 3.

Event Simulation and Reconstruction
In each event, the particles generated with NuWro serve as input for LArSoft [14], which propagates the particles through argon (using GEANT4 [15]) and simulates the detector response to the drifted ionization electrons. LArSoft also searches the simulated TPC wire waveforms for regions of interest and deconvolves and fits them to a Gaussian. These cleaned up "hits" are 2D (each plane of wires is an image of ticks vs. wire) and are shown in Fig 1. Finally, Pandora [16], a pattern recognition software kit, maps the 2D hits from the 3 wire plane projections to 3D and then clusters the 3D positions into tracks and showers.

Energy and Angular Resolution
We can estimate the angular resolution with which proton and muon tracks can be reconstructed by comparing the direction of the outgoing particle at the event generator level to the direction of the fully reconstructed tracks. We find that roughly 50% of tracks are reconstructed to within 5 • of the true particle direction (Fig. 4). Furthermore, we infer the particle momenta via "range" (track length). Fig. 5 compares the true (GEANT4) and the reconstructed track lengths and gives us faith in this method. The true track length is the distance over which GEANT4 propagates the particle before it stops or decays, while the reconstructed track length is based on the hits generated by the ions created by this particle.  Figure 4. Cumulative distribution functions of the angular difference between the true and reconstructed track directions. φ p is the proton angular difference and φ µ is the muon angular difference.

DUNE Simulation
The charge read out on the LArTPC wires can be mapped to the kinetic energy of the culprit particle which caused the ionization. For events in which a proton and muon track are identified, we can measure the proton and muon energies, including the particle rest mass and the kinetic energy.
We reconstruct the ν µ energy using the expression

Event Selection
The atmospheric neutrinos are taken to have energies between 150 MeVand 400 MeV, with an angle-averaged energy spectrum * calculated for Homestake at the solar minimum [19]. We choose this background energy range in order to encompass 3 standard deviations of the reconstructed signal energy. We are justified in ignoring atmospheric neutrinos whose true energies lie outside this range, since they can be well distinguished from the signal by reconstruction of the neutrino energy.
NuWro reports the neutrino-nucleus CCQE cross section; for signal events it reports the cross section at E νµ = 236 MeV, and for atmospheric neutrinos it reports the average cross section weighted by the neutrino energy spectrum between 150 MeVand 400 MeV.
These cross sections are In simulating the CC cross section, we only have events with produced muons, and with neutrinos in the aforementioned energy range. The CCQE cross section is weighted and averaged only over this range. We have not simulated neutral current events, because such events do not produce a muon.
As an initial event selection cut, we consider events in which exactly two tracks are reconstructed, that of a muon and a proton. Although it is expected that DUNE will have excellent particle identification, for simplicity, we only require that Pandora identify exactly two tracks, and we assume that the longer track is a muon while the shorter track is a proton. At 236 MeV, GEANT4 predicts this to be the case 93% of the time. Out of these 93%, 97% are correctly reconstructed as the longer track. Also, a small number of events passing the cuts contain additional ejected nucleons at the event generator level, but for which only one nucleon track was reconstructed.
The requirement that we reconstruct the interaction with an interaction point within the fiducial volume justifies our assumption that the dominant background arises from atmospheric neutrinos. There are a variety of other cosmogenic backgrounds at DUNE, but these backgrounds are unlikely to produce an identified muon track which is reconstructed to begin within the detector. In other words, we have assumed that the analysis is based on a fiducial volume chosen such that the rate of such backgrounds is negligible.

Neutrino directionality
Since the momentum transfer to 39 Ar is non-negligible, one cannot use p µ and p p to reconstruct the direction of the incoming neutrino. † Instead we note that, given a hypothesis for the direction of the incoming neutrino, one can use momentum conservation to determine * Besides an angle-averaged spectrum, [19] provides direction-dependent fluxes binned in the cosine of the zenith angle, Z, and azimuth, φ. The fractional variance of the direction-dependent fluxes compared to the angle-averaged flux decreases with energy for the energies relevant to this study. At 236 (600) MeV, it is 0.34 (0.19). In using the average, the maximum overestimate at 236 MeV is a factor of 3.7. This happens between (-0.8,-0.9) in cos(Z) and between (90,120) degrees in φ. The maximum underestimate is by a factor of 2.2. This happens between (0, 0.1) in cos(Z) and between (270, 300) degrees in φ. † Note, for higher energy neutrinos, the momentum transfer to the remnant nucleus is negligible compared to the energy of neutrino, in which case the momentum of the charged lepton and of the hadronic ejecta is sufficient to reconstruct the neutrino direction effectively. These techniques were used in [22]. the momentum transfer to the remnant nucleus. We define the kinematic variable wherep is a unit vector pointing from the Sun to the detector. If the incoming 236 MeV neutrino were actually arriving from the Sun, then p39 Ar would be the reconstructed momentum of the remnant nucleus. As noted in [10], the ejected proton tends to emerge preferentially in the forward direction. As such, the angle θ p between the proton and the direction from the Sun, defined by cos θ p = (p · p p )/| p p |, is one of the kinematic variables upon which we will impose cuts (Fig. 7). We also find that a useful kinematic variable is θ N , defined by cos θ N ≡p · p39 Ar /| p39 Ar |. If the neutrino does indeed arrive from the direction of the Sun with an energy of 236 MeV, then θ N would evaluate to the angle between the reconstructed remnant nucleus momentum and the direction of the Sun. We plot a generator level (reconstruction level) 2D histogram of cos θ p vs. cos θ N in Fig. 8 (Fig. 9).
Unsurprisingly, both signal and background distributions contain events in which cos θ N is close to 1, since the definition of p39 Ar biases it in the forward direction. Perhaps more surprisingly, the signal distribution contains a significant population of events in which cos θ p ∼ 1, while cos θ N ∼ −1. There is no similar population of events in the background distribution, implying that a good way to reject background is to select events in which the proton is ejected in the direction away from the Sun, while p39 Ar points back to the Sun.
After reconstruction (Fig. 9), the discrimination between signal and background is poorer. Although the angular distribution for the charged lepton is isotropic, it is nevertheless correlated with that of the proton; for events where the ejected proton and recoiling nucleus are (anti-)collinear with the neutrino, the charged lepton track also lies upon the same line. In this case, the reconstruction algorithm may be unable to distinguish the proton and charged leptons tracks, leading to an event reconstructed with just a single track, which would be rejected by the event selection cuts. However, we will see that the shift in dark matter sensitivity due to this reconstruction failure is less than O(10). It may seem counterintuitive that the remnant nucleus should be backscattered in CCQE events. But an examination of the corresponding events at generator level provides an explanation; in the majority of events in which the proton is forward-directed and the remnant nucleus is backward-directed, the nucleon struck by the neutrino had an initial momentum in the direction away from the Sun (Fig. 10). When the struck nucleon is already moving away from the Sun, the outgoing nucleon is also typically very forward-directed, while the remaining nucleons have a net momentum in the opposite direction, leading to a backward directed remnant nucleus.   Figure 11. Signal 2D histograms of the reconstructed muon and proton momenta projected along the incoming neutrino direction. The left plot has only event selection cuts, whereas the right plot only includes such events with very forward protons ( cos θ p < 30 • ). The white line (( p ν − p p − p µ ) ·p ν = 0) separates forward/backscattered remnant nuclei. Momentum conservation means that p39 Ar = p ν − p p − p µ . Hence, the nucleus backscatters when p39 Ar ·p ν < 0 (to the right of the line). This figure emphasizes that the remnant nucleus tends to be backscattered if the proton is very forward scattered.
To illustrate this point, we plot the distribution of signal events in the ( p p ·p , p µ ·p ) plane (Figs. 11). The left panel is the distribution of all signal events passing event selection cuts, while the right panel is the distribution of such events for which cos θ p < 30 • . In both panels, the white diagonal line indicates ( p p + p µ ) ·p = 236 MeV; events to the right of this line have cos θ N < 0, while events to the left have cos θ N > 0,. We define the signal efficiency η µ S to be the fraction of signal neutrino events which pass the event selection cuts as well as the the energy and directionality cuts we impose. Similarly, we define the background efficiency η µ B to be the fraction of atmospheric neutrino events with a neutrino energy between E bgd min = 150 MeV and E bgd max = 400 MeV which pass these cuts. Only a negligible fraction of atmospheric neutrinos outside the range 150 − 400 MeV pass the cuts.
Motivated by the reconstructed energy resolution of the signal events, we impose an energy cut by selecting only events with reconstructed neutrino energy in the range 236 ± 30 MeV. Also, since protons often fly out forward, we require them to lie within an angular cone centered on the direction pointing from the Sun. A similar approach for the leptons is fruitless. At such energies, their ejection is largely isotropic. Finally, we impose cuts on cos θ N .
Various cuts and their effect on DUNE's sensitivity to a 236 MeV flux of ν µ emanating from the Sun are listed in Table 1.

Solar KDAR ν µ Flux
We will first determine the number of background atmospheric neutrino events which are expected to pass our cuts over a given exposure of DUNE.
where η µ B , E bgd min and E bgd max are defined as in the previous section. d 2 Φ µ /dE ν dΩ is the differential flux of atmospheric ν µ , and T is the exposure time. The effective area of DUNE effective is the product of the neutrino-nucleus scattering cross section with the number of nuclei in the fiducial volume. We take DUNE's effective area to atmospheric ν µ ,Ā where σ (µ)bgd. ν-Ar is the ν µ -Ar charged-current scattering cross section, weighted by the atmospheric neutrino spectrum in the energy range E bgd min , E bgd max , as described in Section 2. Given the background acceptances η µ B listed in Table 1, we can then determine the number of background events expected to pass the cuts, also listed in Table 1.
We assume that the number of signal and background events seen by DUNE will be drawn from Poisson-distributions whose means are given by the expected number of signal and background events, denoted by N µ S and N µ B , respectively. To estimate the sensitivity of DUNE, we assume a representative ("Asimov" [32]) data set in which the number of observed neutrinos is taken to be the number of expected background neutrinos, rounded to the nearest integer (that is, N µ O = round(N µ B )). We denote by N µ,90 S the number of expected signal events such that the likelihood of an experimental run observing a number of total events larger than round(N µ B ) is 90%. A model for which the expected number of signal events satisfies N µ S > N µ,90 S lies in the region to which we estimate DUNE would be sensitive. Given N µ,90 S and η µ S , we can then straightforwardly determine Φ 236 MeV , the maximum flux of 236 MeV neutrinos emanating from the core of the Sun which would be allowed (at 90% CL), given that DUNE observed only a number of events consistent with atmospheric neutrino background.
and σ (µ) ν-Ar (E ν = 236 MeV) = 2.6 × 10 −38 cm 2 . Φ 236 MeV is our primary result, and represents the minimum flux of 236 MeV ν µ emanating from the core of the Sun to which DUNE would be sensitive with any given exposure. This result is independent of the the specific model of new physics which generates this excess flux of neutrinos, but is determined only by the efficiency with which 236 MeV neutrinos from the core of the Sun and atmospheric background neutrinos pass the cuts.
We plot Φ 236 MeV in Figure 12, as a function of the exposure, for several different choices of cuts (see Table 1). In each case, the reconstructed neutrino energy is required to be in the range 236 ± 30 MeV. In one case, cuts on θ N and θ p are chosen to optimize signal significance (solid lines), while in the other case, these cuts are chosen to optimize the signal-to-background ratio (that is, η S /η B ) (dashed lines). To illustrate the effect of possible improvements in track reconstruction, we also apply this analysis framework directly to the muon and proton tracks produced by the event generator; these curves are presented as green lines. All four of the angular cut choices, along with their efficiencies, sensitivities, signal-to-background ratios, and number of expected signal and background events, are listed in Table 1. For the cuts (applied to reconstructed events) which maximize the S/B, the sensitivity varies discontinuously. This is because, in this case, the number of assumed events observed is small, and the jumps are where they vary discontinuously.

Application: Search for Inelastically Scattered Dark Matter
To place this result in context, we consider a dark matter scenario which can be constrained by data from DUNE, but which would be difficult to constrain with direct detection experiments. In particular, we consider the case of low-mass dark matter (m X 10 GeV)  Table 1 and the green lines correspond to the generator level quantities in the third and fourth rows. The dashed lines are for maximum S/B and the filled lines are for maximum DM sensitivity. The discontinuities are due to the limit of small numbers of events; noticeable when the number of observed events jumps by one.
which scatters inelastically with nuclei, with the emerging dark particle being δ = 50 keV heavier than the incoming dark matter particle. In this case, dark matter inelastic scattering is kinematically inaccessible for detectors on Earth, because there is insufficient energy to produce the excited state. But because dark matter accelerates as it approaches the Sun, it may have sufficient kinetic energy to scatter inelastically against solar nuclei, leading to its gravitational capture [23][24][25][26].
One example of a scenario in which inelastic scattering can dominate is the case in which the dark matter is charged under a spontaneously-broken U (1) gauge symmetry. In this case, a dark matter vector current couples to the dark photon, which can be mediate dark matter-nucleon scattering. The tree-level scattering process is necessarily inelastic, because the vector current for a single real particle vanishes. Elastic scattering is instead subleading, mediated either by multiple dark photon exchange or by other mediators with small couplings. Although the size of this subleading elastic scattering cross section is model-dependent, it can be well below current direct detection sensitivity.
After the initial inelastic scatter, the dark matter is gravitationally captured, and continues to orbit the Sun. As the dark matter passes through the Sun many times, subsequent inelastic or elastic scatters result in an even greater loss of dark matter kinetic energy, until the particle settles in the core of the Sun [27,28]. Once the dark matter has lost enough kinetic energy, inelastic scattering is no longer kinematically possible, but since the dark matter continues to pass through the Sun many times during the Sun's lifetime, even the subleading elastic scattering cross section can be sufficient to deplete the dark matter kinetic energy enough for it to settle in the core.
After gravitational capture, we assume dark matter annihilation to first generation quarks, with dark matter capture and annihilation being in equilibrium. Even though dark matter annihilation produces only first-generation quarks, a substantial number of kaons are produced by subsequent fragmentation and hadronization processes. If the dark matter mass is O(5 GeV), then the center of mass energy is large compared to the kaon mass, and the up, down, and strange quarks can all be treated as light quarks. We assume that dark matter scattering with nuclei is spin-independent and velocityindependent, with an equal coupling to protons and neutrons. Because δ m X , the dark matter-nucleon scattering matrix element is largely independent of δ. The dependence of the dark matter-nucleus scattering cross section on δ arises from the final state phase space. Thus, we will parameterize the dark matter model by σ 0 , which is the total cross section for dark matter-nucleon scattering, extrapolated to δ = 0. From this quantity, the differential cross section for scattering against any nucleus at δ = 50 keV can be determined.
In this scenario, the DM annihilation rate (Γ A ) is equal to one-half of the dark matter capture rate (Γ C ). The capture rate is directly proportional to σ 0 , with Γ C = C δ (m X ) × σ 0 . The proportionality constant C δ (m X ) is determined entirely by the dark matter mass, by solar physics, and the assumption that dark matter has a nominal Maxwell-Boltzmann velocity distribution with a density of 0.3 GeV/ cm 3 . Relevant values for the C δ (m X ) can be found in [29].
In this scenario, we can relate Φ 236 MeV to σ 0 , finding where F µ = 0.27 is the fraction of 236 MeV neutrinos which arrive at the detector as ν µ , assuming a normal hierarchy. While an experimental data analysis requires a full treatment of neutrino oscillations to obtain neutrino spectra and flavor ratios for specific times of detector operation, for this analysis it is sufficient to assume an annual averaged flavor ratio taken from [9] (if one assumed an inverted hierarchy F µ would increase by at most 25%). r ⊕ = 1.5 × 10 11 m is the distance from the Sun to the Earth, and r K (m X ) is the fraction of the center of mass energy of the dark matter initial state which is converted into stopped K + through dark matter annihilation, the hadronization and fragmentation of the outgoing particles, and the interactions of those particles with the dense solar medium (values for r K (m X ) can be found in [8]). The factor 0.64 is the branching fraction for K + decay to produce a monoenergetic 236 MeVν µ . We can thus relate Φ 236 MeV to a 90% CL exclusion contour in the (m X , σ 0 )-plane. In Figure 13, we plot the 90% CL sensitivity of DUNE (400 kT yr) in the (m X , σ 0 )-plane for the case where WIMPs annihilate solely to first generation quarks, assuming a search for monoenergetic neutrinos at 236 MeV from stopped K + decay. We plot sensitivity curves for each of the four cuts strategies given in Table 1.
There are a variety of other theoretical uncertainties which can have a significant effect on DUNE's sensitivity. For example, we have assumed that dark matter annihilates to first generation quarks. If dark matter annihilates instead to second generation quarks, the average number of K + produced per annihilation (and, thus, the flux of 236 MeV neutrinos) would increase by about a factor of 2. Furthermore, we have modeled neutrino-nucleus scattering at this energy with NuWro. Although there are experimental measurements of this Op tim al DM Se ns itiv ity (R eco ns tru cte d) Opt ima l DM Sen sitiv ity (Ge ner ator ) Figure 13.
Projected 90% sensitivity curves for DUNE (400 kT yr) for inelastic dark matter scattering. All the curves are for the stopped K + channel. The relevant cuts are listed in Fig. 12 and Table 1. Table 1. The angular cuts (including the energy cut of 236 ± 30 MeV) and the resulting signal and background efficiencies, the expected number of signal and background events, the expected signal to background ratio at DUNE, and the maximum flux of 236 MeV neutrinos emanating from the Sun which would be allowed (at 90% CL). The first two rows are cuts on reconstructed events and the last two rows are cuts on generator level events (no detector simulation/reconstruction). We include the generator level information to illustrate the optimistic case of perfect reconstruction. cross section, there are still significant uncertainties, both in the magnitude of the chargedcurrent cross section and in the angular dependence. But any stopped pion experiment also acts as a stopped kaon experiment [30], and a variety of future KDAR measurements are under consideration [31], and would serve as a calibration for this type of analysis. Importantly, DUNE itself can provide calibration data, by searching off-axis. Future improvements in reconstruction techniques that could enable electron channel to be used effectively, would lead to a significant improvement in sensitivity. The electron channel is generally expected to be more sensitive than the muon channel for three reasons [10]. First, the atmospheric neutrino background flux is smaller. Second, the effective area of DUNE is larger for 236 MeV ν e than for ν µ , because the charged-current scattering cross section for ν µ is suppressed by the reduced phase space of the outgoing muon. Third, the flux of 236 MeV ν e arriving at Earth from KDAR in the Sun is expected to be larger than the flux of 236 MeV ν µ as a result of oscillation effects in the dense medium of the Sun (assuming a normal hierarchy) [9].

Conclusion
In this work, we have estimated DUNE's potential to detect the monoenergetic 236 MeV neutrinos arising from kaon-decay-at-rest in the core of the Sun. Although the charged leptons produced from a charged-current interaction of a 236 MeV neutrino are roughly isotropic, many such interactions produce an ejected proton which is forward-directed. Moreover, the remnant nucleus tends to be backward-directed, and observable kinematic variables can be used as a proxy for the remnant nucleus momentum, allowing for better discrimination of signal from background.
We have used these observables in a realistic manner, with the response of the detector modelled numerically. Although we have found that the discrimination of signal from background, S/B, can be as large as 2.2 for a model where there are enough signal events to exclude, a realistic treatment of the detector results in reduced sensitivity with respect to earlier estimates.
Foreseeing future improvements in reconstruction (for example, via machine learning), we calculated the expected number of signal and background events which pass our cuts at the generator level (see Table 1). We've also plotted the generator level dark matter sensitivity curves in green in Fig. 12. These are the limits in the optimistic case of perfect reconstruction, and we find that this optimal sensitivity matches estimates made previously [10].
There are a variety of non-standard scenarios for dark matter particle physics and astrophysics in which the sensitivity of direct detection experiments is suppressed, and the flux of 236 MeV neutrinos produced in the Sun's core may provide an excellent indirect probe of dark matter interactions. In this case, DUNE's ability to identify 236 MeV neutrinos arriving from the direction of the Sun, while rejecting background, can provide unique control over systematic uncertainties. As an example, we have estimated DUNE's sensitivity to lowmass dark matter which scatters inelastically, with a mass splitting of δ = 50 keV. This is an example of a dark matter process which is kinematically inacessible for direct detection experiments on Earth, but for which a search for neutrinos at DUNE may lead to a discovery.
The search for direct evidence of non-gravitational interactions between dark matter and Standard Model matter has thus far yielded no conclusive positive signals. This has led to broader theoretical and experimental approaches to dark matter searches, and KDAR neutrinos can play an important role. It would be interesting to further study the theoretical scenarios in which searches for KDAR neutrinos provide a competitive advantage.
On the experimental side, it would also be interesting to study in more detail how the particle identification and track reconstruction at DUNE could be improved in the energy range relevant for KDAR searches. A possible DUNE module-of-opportunity may use a wireless design with an isotropic response and could improve the sensitivity to dark matter annihilation in the Sun, by reducing the loss of efficiency associated with the orientation of the Sun with respect to the DUNE wires.