Benefits of MeV-scale reconstruction capabilities in large liquid argon time projection chambers
W. Castiglioni, W. Foreman, I. Lepetic, B. R. Littlejohn, M. Malaker, A. Mastbaum
BBenefits of MeV-Scale Reconstruction Capabilities in Large Liquid Argon Time ProjectionChambers
W. Castiglioni, W. Foreman, ˚ B. R. Littlejohn, : and M. Malaker Physics Department, Illinois Institute of Technology, Chicago, IL 60616, USA
I. Lepetic ; and A. Mastbaum Department of Physics and Astronomy, Rutgers University, Piscataway, NJ 08854, USA (Dated: June 29, 2020)Using truth-level Monte Carlo simulations of particle interactions in a large volume of liquid argon, wedemonstrate physics capabilities enabled by reconstruction of topologically compact and isolated low-energyfeatures, or ‘blips,’ in large liquid argon time projection chamber (LArTPC) events. These features are mostlyproduced by electron products of photon interactions depositing ionization energy. The blip identification capa-bility of the LArTPC is enabled by its unique combination of size, position resolution precision, and low energythresholds. We show that consideration of reconstructed blips in LArTPC physics analyses can result in substan-tial improvements in calorimetry for neutrino and new physics interactions and for final-state particles rangingin energy from the MeV to the GeV scale. Blip activity analysis is also shown to enable discrimination betweeninteraction channels and final-state particle types. In addition to demonstrating these gains in calorimetry anddiscrimination, some limitations of blip reconstruction capabilities and physics outcomes are also discussed.
I. INTRODUCTION
Large liquid argon time projection chambers (LArTPCs)have become one of the primary detector technologies usedfor performing neutrino physics experiments. At present,a suite of sub-kiloton scale LArTPCs, SBND [1], Micro-BooNE [2], and ICARUS [3], collectively called the FermilabSBN Program, are operating or being constructed along Fer-milab’s Booster Neutrino Beamline at sub-km baselines forthe primary purposes of probing short-baseline neutrino ap-pearance/disappearance [4] and measuring neutrino-nucleusinteraction cross sections on argon [5]. Within the nextdecade, the 40 kiloton Deep Underground Neutrino Experi-ment (DUNE) LArTPC [6] will be deployed underground inthe Sanford Underground Research Facility in South Dakotaalong a new Fermilab-based neutrino beamline [7] for the pri-mary purposes of measuring long-baseline neutrino oscilla-tions and leptonic CP-violation, probing a host of beyond-the-Standard-Model (BSM) physics models, and measuringneutrinos from astrophysical sources, such as core-collapsesupernovae [8].The primary technological advantage most exploited thusfar in existing large LArTPC measurements and physics sen-sitivity studies is arguably its millimeter-scale spatial reso-lution. The LArTPC’s uniform electric field, low electrondiffusion, and sub-centimeter charge readout element spac-ing enable the conservation and recording of initial ioniza-tion electron topologies produced by particle interactions inthe argon. For GeV-scale neutrinos, the rich topologies ofinteraction final-state tracks and showers can be used to dis-tinguish electron- and muon-type neutrino interactions, en-abling sensitive electron-type neutrino searches in conven-tional ν µ -dominated neutrino beamlines [9–11]. Simple but ˚ [email protected] : [email protected] ; [email protected] precise mm-scale topological analysis of interaction verticesand final-state tracks has also enabled LArTPCs to providesensitive BSM searches [12] and new insight into neutrino in-teraction models [13, 14]; major improvements on the latterfront are expected as a larger set of exclusive cross-sectionmeasurements are developed and published by MicroBooNEand other experiments. Small-angle muon scattering clearlyvisible in high-resolution images [15], as well as track length,have been used in MicroBooNE as primary methods of en-ergy reconstruction for BSM and GeV-scale neutrino interac-tion cross section measurements in LAr [12, 16]. Many soft-ware tools, based on a range of operational principles, havebeen developed that use LArTPC image topologies to identifytrack and shower objects and reconstruct their kinematics [17–19]. Many of the stated centerpiece goals of short-baselineand long-baseline LArTPC experiments will be achieved bycombining mm-scale resolution and calorimetric capabilitiesin analyzing charged particle interactions in LAr ranging fromthe tens to thousands of MeV scales. This recipe provides thekinematic and particle identification details necessary to per-form the long list of neutrino LArTPC physics goals givenabove.A technological advantage of the large neutrino LArTPCthat has received comparatively less attention is its low-energy-threshold detection capability. This capability is en-abled by the modest 24 eV mean ionization energy of liquidargon, the high ionization electron collection efficiency of theTPC, and the low levels of noise achievable in modern readoutelectronics. Through studies of Michel electrons [20–22], nu-clear de-excitation photons [23], and Ar β -particles [24], ithas been established that single-phase neutrino LArTPCs arecapable of identifying physics signatures at and well below theMeV energy scale. In one of these studies, performed by theArgoNeuT single-phase LArTPC experiment [25], a physicsdetection threshold of 200–300 keV was established by com-paring simulated and measured de-excitation photons gener-ated by final-state nuclei and neutrons from neutrino interac-tions [23]. This study specifically highlights the uniqueness of a r X i v : . [ phy s i c s . i n s - d e t ] J un LArTPCs among all demonstrated massive neutrino detectortechnologies in achieving a combination of mm-scale positionresolution and sub-MeV energy thresholds.The aim of this paper is to explore how these low-energy-threshold, high-position-resolution combined capabil-ities might be put to use in a broad variety of contexts relevantto large neutrino LArTPC physics goals. More specifically,we will describe how LArTPC signatures that are compact(sub-cm scale), low-energy (below the few MeV scale) and/ortopologically isolated (separated from larger topological ob-jects by cm or more) are produced in physics events of inter-est, and how these signatures can be used to enhance capabil-ities in calorimetry, energy calibration, and discrimination ofparticle type or interaction type. We will show that analysis ofthese low-energy, compact LArTPC signatures, which we re-fer to as ‘blips’ throughout the paper, can be broadly beneficialin supernova neutrino, solar neutrino, long-baseline oscilla-tion, and BSM studies in LArTPCs. All studies are performedusing a common framework of truth-level Monte Carlo simu-lations in a generic liquid argon environment.This paper will begin in Section II by describing the MonteCarlo methods used to define blip activity in LArTPC events,and to describe blip physics metrics of interest used through-out the paper. The benefits of considering reconstructed blipactivity in supernova and solar neutrino energy reconstructionand interaction channel identification (Section III), neutronidentification and calorimetry (Section IV), electromagneticshower calorimetry (Section V), particle discrimination (Sec-tion VI), BSM physics (Section VII), and single γ -ray spec-troscopy (VIII) will then be presented and discussed. Someprimary detector-related effects limiting the utility of blip ac-tivity for physics purposes will be studied in Section IX. Sum-marizing remarks will be given in Section X. II. STUDY DEFINITIONS AND PROCEDURES
In this section, we will summarize the physics processesthat define the LArTPC blip signals to be studied in this paper,and then describe the Monte Carlo simulation used to generatethe truth-level reconstructed physics quantities we will use todemonstrate the physics potential of blip signals.
A. Physics of Low-Energy Depositions in Liquid Argon
The blip features in LArTPC images that will receive mostof the focus of this paper are the product of ionization of ar-gon by low-energy ( „
50 keV to „ γ -rays generated via de-excitation of nu-clei following inelastic interactions with neutrinos, neutrons,pions, and muons, and via bremsstrahlung interactions of elec-trons.Isolated blip-like features can also be produced in LArTPCevents via proton-producing inelastic interactions of high-energy neutrons with argon. Given their higher depositiondensity profiles, proton-produced blips can have reconstructedenergies higher than what can be produced by an electron.However, MeV-scale proton-produced blips are largely indis-tinguishable from electron-produced blips. B. Blip Simulation and Truth-level Reconstruction
For this study, primary particles are generated in and prop-agated through a large, essentially infinite uniform volume ofliquid argon using the Geant4-based [32] LArSoft simulationpackage [33]. Primary electrons, neutrons, protons, pions,muons, and γ -rays are mainly generated using the standardGeant4 gun generator; for supernova-related studies, neutrinointeraction final-state products are generated using the MAR-LEY neutrino interaction software package [34]. In propagat-ing particles through the argon with LArSoft, care is taken toimplement the correct physics libraries and threshold settingsrequired for properly simulating high-energy physics pro-cesses, such as neutron inelastic scattering and pion and muoncapture. For high-energy physics processes, the NeutronHP and
QGSP BERT HP
Geant4 libraries are implemented. Torecord low-energy electron histories, Geant4/LArSoft track-ing thresholds are reduced to 10 keV for electromagnetic pro-cesses. During particle transport, a wide variety of particlehistory information is stored for further analysis, includingthe true starting and ending energies and locations for all par-ticles, their identities and the physics processes resulting intheir creation or destruction, and the properties of their parentand daughter particles.Rather than simulating the full readout, low-level process-ing, and higher-level object reconstruction pathways of a spe-cific LArTPC experiment to generate reconstructed blip ob-jects for analysis, we adopt a detector-agnostic truth-level ap-proach. To begin, any topologically-isolated electron deposit-ing more than 75 keV of energy in the liquid argon is con-sidered as an identified, reconstructed blip. This thresholdis lower than that achieved in previous blip analyses in Ar-goNeuT [27] to reflect the reduction in noise levels that areachievable in large LArTPCs using cold electronics [35]. Thereconstructed position and energy of each blip is taken to bethe true start location of and true energy deposited by the elec-tron interaction, respectively. The addition of a finite blip en-ergy resolution or higher detection threshold only marginallyimpacts the physics results to be described; these detector ef-fects will be explored in more detail in Section IX.The reconstructed blip quantities of interest for physicsanalysis in this paper are the total blip multiplicity and indi-vidual and summed blip energies. To consider the likely needto avoid inclusion of blip activity from radiogenic Ar β de-cays, only blips within a set proximity to points of interest inan event, such as a neutrino interaction vertex, will be con-sidered. In practice in this paper, this point of interest will bedefined as the true generation vertex of the relevant primaryparticle. In most physics cases considered in this paper, ref-erence points of interest will be easy to identify to mm-scaleprecision with conventional large-feature reconstruction algo-rithms. For this study, blip distances of 20, 30, and 60 cm areusually considered; for reference, the attenuation (interaction)length for a 1 MeV photon (10 MeV neutron) in LAr is ap-proximately 15 cm (30 cm). The presence and impact of Arcontamination of blip samples will not be considered in mostof the following sections, and is instead separately examinedin Section IX. Beyond blip proximity, other topological fea-tures of blips are not considered in our analysis.In this truth-level blip reconstruction scheme, care must betaken in summing energies for contiguous electron interac-tions and in considering electrons undergoing bremsstrahlunginteractions. These and other aspects of the blip reconstruc-tion procedure are illustrated in Figure 1 and Table I using anexample history of a simulated 3 MeV γ -ray. This γ -ray un- FIG. 1. Illustrated history of an example 3.0 MeV γ -ray interactionin liquid argon. Red (black) dotted (solid) lines indicate γ -ray (elec-tron) trajectories; energies of these particles are indicated in Table I.A blue circle illustrates a 30 cm proximity requirement with respectto the γ -ray generation point (dark blue dot). Four of five black elec-tron groups are identified as reconstructed blips, and only three ofthese, ‘1,’ ‘2,’ and ‘5,’ meet the proximity requirement.Item Type Creator Process E start E end E blip a γ Primary 3.00 3.00 -1 2 e ´ Compton scatter 1.50 0 1.50b γ Compton scatter 1.50 1.50 -2 e ´ Compton scatter 1.00 0 0.75c γ Bremsstrahlung 0.25 0.25 -3 e ´ Photoelectric Effect 0.25 0 0.25d γ Compton Scatter 0.50 0.50 -4 e ´ Compton Scatter 0.05 0 -e γ Compton Scatter 0.45 0.45 -5 e ´ Photoelectric Effect 0.45 0 0.45TABLE I. Particles followed in the recorded history of the example3.0 MeV γ -ray event shown in Figure 1, which is used to illustrate thedetails of the blip reconstruction procedure. This event would pro-duce four identified blips, with reconstructed energies of 1.5, 0.75,0.25, and 0.45 MeV. Of the primary γ -ray’s 3.00 MeV total energy,2.70 MeV of summed blip energy is considered after proximity re-quirements are applied. dergoes three Compton scatters and ends its life via the pho-toelectric effect. Reconstructed blip ‘1,’ formed at the firstCompton scatter vertex, must have an energy that includes thedepositions of both the initial Compton electron as well asthe hard scattered electron it produced. This is accomplishedby taking the Compton electron’s starting energy (E start in Ta-ble I) as the reconstructed blip energy. For a test sample ofsimulated 1.5 MeV γ rays, less than 1% of all tracked elec-trons were produced in hard scatters of a parent electron; thisfraction is likely to be higher for higher-energy simulated pri-mary particles.For reconstructed blip ‘2,’ energy lost via bremsstrahlunginteraction of the Compton electron must accounted for bysubtracting from E start the energies of any produced daugh-ter γ -ray. Both the first electron (‘2’) and that produced bythe bremsstrahlung photon interaction (‘3’) are considered asseparate candidate reconstructed blips. For the 1.5 MeV γ -raydataset mentioned above, bremsstrahlung photons will occa-sionally convert within a distance smaller than the positionresolution of a LArTPC ( „ γ -rayproduces a blip multiplicity of 3, with individual blip energiesof 1.50, 0.75, and 0.45 MeV, and a summed blip energy of2.7 MeV.As mentioned in the previous section, neutron inelasticinteractions in argon can produce final-state protons, whichwill also appear as blips in LArTPC images. To realisti-cally include these signatures in our truth-level reconstructionscheme, any proton blip below 3 MeV in total energy is treatedas a reconstructed electron blip. Any proton above this energywould produce a single-hit blip too high in energy to reason-ably be produced by an electron [27]; thus, these protons areexcluded from the set of considered blips. III. SUPERNOVA AND SOLAR NEUTRINOS
A DUNE-based analysis of solar neutrinos will improvethe solar-based measurement of ∆ m , enabling precise testsof the Standard Model neutrino mixing picture [36]. In thecase of supernova burst neutrinos, the primary reconstructedphysics metrics of interest are the independent energy andtime profile of fluxes for the different neutrino flavors, in addi-tion to the total number of detected neutrinos. Using our truth-level blip reconstruction technique on MARLEY- and Geant4-generated low-energy neutrino interaction final states, we willdemonstrate how blip activity can improve energy recoveryand resolution and can aid in separation of flavor-exclusive ν e charged current (CC) and flavor-inclusive neutrino-electronscattering (ES) channels. A. Neutrino Energy Calorimetry Improvements
To provide optimized low-energy neutrino energy recon-struction, one must perform calorimetry on all visible final-state particles. For supernova and solar neutrino interactionsin argon, ν e CC interactions represent the primary detectionchannel: ν e ` Ar Ñ e ´ ` K ˚ . (1)Thus, for this channel, an optimal reconstruction of energywill include depositions not just from the final-state e ´ , butalso from the products of de-excitation of the K nucleus, N u m be r o f P ho t on s De-excitationNeutronBremsstrahlungAnnihilation e n All
Energies of Individual Photons N u m be r o f P ho t on s De-excitationNeutronBremsstrahlungAnnihilation e n <15 MeV Energies of Individual Photons
FIG. 2. Top: γ -ray energies produced by a MARLEY-generated sam-ple of ν e CC supernova neutrino interactions. Bottom: Subset of su-pernova CC events below 15 MeV input ν e energy, roughly matchingthe maximum energy limit for B solar neutrinos. primarily γ -rays and neutrons. While current tools will likelybe capable of triggering on and reconstructing the former [8,20], inclusion of the latter in calorimetry has not been closelystudied in the literature. For the sub-dominant ν x -electron ESinteraction process, ν x ` e ´ Ñ ν x ` e ´ , (2)energy optimization will yield limited improvement due to thepresence of the invisible final-state ν x ; thus, the ES channelwill be ignored in the present sub-section.Figure 2 shows the energies of γ -rays produced as a resultof 10 ν e CC interactions generated with MARLEY for thefull energy range. Also pictured is the subset of supernovaCC events below 15 MeV input ν e energy, roughly matchingthe maximum energy limit for B solar neutrinos. Separateaccounting is given for photons produced by interactions offinal-state de-excitation γ -rays, bremsstrahlung photons pro-duced by the final-state electron, and de-excitation photonsproduced via inelastic scatters of final-state neutrons. It isclear that the average low-energy ν e event will feature farmore than one MeV-scale γ -ray in the final state, each ofwhich is quite likely to produce more than one reconstructedblip.Figures 3 and 4 demonstrate the impact of including blips O l d R e c on s t r u c t ed E ne r g y ( M e V ) Default Reconstruction
Old Reconstructed Energy vs True Neutrino Energy N e w R e c on s t r u c t ed E ne r g y ( M e V ) Blip-Including Reconstruction
New Reconstructed Energy vs True Neutrino Energy
FIG. 3. Reconstructed vs. true energies for supernova ν e undergoingCC interactions in liquid argon. Top: default reconstruction includ-ing only the primary final-state electron’s trunk. Bottom: improvedcase including reconstructed blips above 75 keV and within 30 cmof the neutrino interaction vertex. Events are generated using MAR-LEY simulations. The vertical black dotted line at 15 MeV denotesthe approximate endpoint of the B solar neutrino spectrum. in reconstructed ν e CC event energies. The former presentsthe reconstructed and true ν e energies for all events, while thelatter presents the one-dimensional profiled mean and RMSreconstructed energy versus true ν e energy. For the defaultscenario, we consider reconstruction of only the true energyof the primary electron topological object, or ‘trunk’; thisdefinition excludes energy lost by the primary electron tobremsstrahlung interactions. In the improved scenario, we in-clude reconstructed blips from the γ -rays in Figure 2, as longas they occur within 30 cm of the neutrino interaction vertex.For the improved scenario in Figure 3, a strong trend is visiblejust below the E rec “ E true ´ . MeV diagonal represent-ing the most complete possible extent of calorimetry. Thisdiagonal trend is substantially more smeared for the defaultcase, particularly at higher energies, where bremsstrahlunginteractions of the primary electron are more likely. In bothcases, substantial far-off-diagonal contributions are visible,which primarily arise from binding energy losses associatedwith final-state neutron production. As will be discussed inSection IV, it is possible but unlikely that this binding energyloss is recoverable in the reconstruction, given the challengeof positively identifying the presence of final-state neutrons inthese interactions. A v e r age R e c on s t r u c t ed / T r ue E ne r g y Blip-Including ReconstructionDefault Reconstruction
Average Reconstructed/True Energy F r a c t i ona l R M S R e s o l u t i on Blip-Including ReconstructionDefault Reconstruction
Resolution
FIG. 4. Top: Ratio of average reconstructed and true neutrino en-ergy versus neutrino energy for ν e CC interactions using the de-fault (dashed red) and blip-including (solid blue) reconstructionmethods shown in Figure 3. Bottom: RMS of the distribution of( E reco ´ E true q{ E true values for different energy bins. The verticalblack dotted line at 15 MeV denotes the approximate endpoint of the B solar neutrino spectrum.
In the default case pictured in Figure 4, the reconstructedenergy is significantly less than the true incoming ν e energy:on average, only 47% of the neutrino’s kinetic energy is re-constructed. When reconstructed blips are included, on aver-age 70% of the ν e kinetic energy is recovered. Increasing theradius to 60 cm improves this fraction to 79%. Much of the re-maining unaccounted energy in this case is due to the reactionthreshold and binding energy losses. The former is constantwith energy and thus does not result in broadened energy res-olution. The latter is energy-dependent, and does provide asubstantial energy resolution contribution.Improvements in the fractional RMS resolution of recon-structed energy distributions, defined as the RMS of theevents’ ( E reco ´ E true q{ E true distribution, are also illustratedin Figure 4. Resolutions for the blip-including case remainwithin „ ν e energies. If the proximityrequirement is loosened to 60 cm, resolutions are further im-proved by up to 4%. Thus, it is clear that blip reconstructioncan improve calorimetry for both solar and supernova neutrinoCC interactions.We note that the blip-including resolution achieved in thisstudy is comparable to that estimated in the DUNE Techni-cal Design Report (TDR) using existing charge signal recon-struction tools [8]. Given that we implement an optimistic75 keV blip threshold in this study, it seems unlikely that fur-ther improvements beyond that pictured in Figure 4 and refer-enced above are possible with DUNE charge signals. Thus,our study appears to indicate that a resolution approachingthe ‘physics limited’ scenario in the DUNE TDR is likely notachievable. B. Interaction Channel Identification
For low-energy neutrino signals, the multiplicity oftopologically-isolated signatures in an event will increase asthe assumed feature reconstruction threshold is decreased.This increase in multiplicity is likely to be sizeably differentfor CC and ES interactions of supernova and solar neutrinos,due to the lack of nuclear de-excitation activity in the lattercase.
FIG. 5. Top: LArTPC event display of a MARLEY-simulated super-nova ν e CC interaction in argon. Bottom: Similar simulation of a sin-gle electron in Geant4, with an energy comparable to that expectedfrom supernova ν x ES interactions. Both event displays are madewith a simulated LArTPC incorporating wire-based charge readout,and only collection plane signals are shown. Dimensional scales arelabeled, while color indicates the amount of charge collected.
Figure 5 shows two simulated supernova neutrino interac-tions in a LArTPC. The top event consists of a 30 MeV ν e CC event producing a 20 MeV electron and cascade of de-excitation photons. Compton scatters of these γ -rays producethe many blips seen in the event. The bottom event consists ofa single 20 MeV electron, as could be produced by a 30 MeV ν x ES interaction. We note that the extensive blip activityseen in the top event is absent in the bottom one. These topo-logical differences suggest that a capability to distinguish be-tween interaction channels is offered by blip reconstructionin LArTPCs. Such a capability would provide additional dis-crimination beyond that achieved by considering the forward- E ne r g y ( M e V ) CCabs Total Blip Energy vs Multiplicity 30 cm E ne r g y ( M e V ) Scattering Total Blip Energy vs Multiplicity 30 cm
FIG. 6. Multiplicity versus summed energy of blips above 75 keVand within 30 cm of the primary electron start point for ν e CC (top)and ν x ES (bottom) events. scattered kinematics of the final-state electron in the ES pro-cess [8, 36, 37].Figure 6 provides a demonstration of this discriminationcapability for supernova ν e CC and ν x ES events. The CCsample is the same as used in the previous section, whilethe ES electron sample is generated using the Geant4 parti-cle gun with an input energy profile matching that expectedfrom the default supernova energy parameterization describedin the DUNE TDR [8]. This figure shows the multiplicityand summed energy of reconstructed blips which meet the de-fault criteria described in previous sections. The two inter-action types produce characteristically different distributions,with higher blip multiplicities and summed energies in the CCsample.If we assume a CC:ES interaction ratio of „ “ , (3)Purity “ ` CCES ˆ . (4)A variety of cut scenarios are considered in Table II. Forthe default blip selection case, combined cuts on multiplicity Threshold Sphere Radius . ´ cm – ă MeV 44% 43%75 keV . ´ cm ă – 56% 35%75 keV . ´ cm ă ă MeV 40% 45%75 keV . ´ cm – ă MeV 59% 55%75 keV . ´ cm ă – 58% 48%75 keV . ´ cm ă ă MeV 51% 60%300 keV . ´ cm – ă MeV 46% 60%300 keV . ´ cm ă – 46% 60%300 keV . ´ cm ă ă MeV 40% 62%300 keV . ´ cm ă – 21% 67%TABLE II. Efficiency and purity in selecting ν x ES events from alarger sample of ν e CC events using only cuts on reconstructed blipactivity. Efficiency and purity definitions are given in the text. ( ă
4) and summed blip energy ( ă ă and ă MeV, respectively, purity is increased to60% with an improved 49% efficiency hit. Interaction chan-nel discrimination capabilities are not substantially degradedwhen considering a higher 300 keV blip energy threshold.These studies make it clear that even simplistic blip activitycriteria have substantial power to separate CC and ES chan-nels. It is likely that a more detailed study incorporating blipactivity as well as additional variables, such as primary elec-tron energy and momentum, would produce substantially im-proved performance with respect to that given above. Whileonly supernova neutrino fluxes have been considered here, wewould expect similar levels of discrimination from solar neu-trino fluxes, given the similar de-excitation γ -ray spectrum forhigher- and lower-energy CC interactions (Figure 2) as wellas the reduced production of bremsstrahlung photons at lowersolar ν e energies. IV. FINAL-STATE NEUTRON IDENTIFICATION ANDCALORIMETRY
Final-state and secondary neutrons play a key role in defin-ing the energy budgets of GeV-scale accelerator neutrinos andantineutrinos, and, to a lesser extent, supernova neutrinos.With a neutron separation energy below 10 MeV for Ar, thisshould not be at all surprising. Accelerator neutrino physicsexperiments have recently begun considering signatures fromfinal-state neutrons in neutrino measurements [38, 39]. Mostneutrons produced as a result of supernova, solar, and beamneutrino interactions will have kinetic energies in the sub-MeV to 10s of MeV range. Below, we consider and sum-marize the role of final-state neutrons in defining MeV-scaleand GeV-scale neutrino energy calorimetry in argon, and dis-cuss the extent to which blip activity can aid in the recoveryof neutron-related final-state information.
A. Neutron Signals in Liquid Argon
Recently-measured neutron interaction cross-sections on Ar in the „ γ -producing inelastic scattering andby neutron-producing reactions, which will in turn generate γ -producing inelastic scattering. These cross-sections corre-spond to effective neutron interaction lengths on the order oftens of cm. γ -ray energies produced by de-excitation of Arin response to inelastic interactions of 10 MeV neutrons arealso given in Figure 7. Produced γ -rays are in the 0-6 MeVrange, similar to those produced by supernova ν e CC interac-tion final-state K ˚ in Figure 2. Thus, deposition of somefinal-state neutron energy will be reflected in LArTPC eventsas electron- and positron-produced blips. These blips will tendto be more concentrated in the general vicinity of the neutron’spoint of production. This suggests the ability to estimate thepresence and/or energies of free neutrons in the final state ofa neutrino interaction based on the presence or multiplicity ofblips in its corresponding event. The remainder of this sub-section will demonstrate this level of calorimetric capabilityfor LArTPCs, as well as how this capability varies for differ-ent neutron energies.At energies below a few MeV, no excited states of Ar areaccessible to the incident neutron and interactions are domi-nated by elastic scattering. Due to their low kinetic energy,these recoiling nuclei are not visible in large neutrino LArT-PCs. This feature of LArTPC response to neutrons is demon-strated in Figure 8, which shows summed blip energies for pri-mary neutrons generated with kinetic energy between 0 and20 MeV (momentum between 0 and 195 MeV/c) using theprocedures described in Section II. At the lowest energies pic-tured in this figure, almost no blip activity is visible.Very low-energy neutrons will produce MeV-scale γ -rayactivity as they thermalize and capture on Ar. In Figure 8,the few events containing blip activity from neutron capturesexhibit a summed blip energy higher than the true kinetic en-ergy of the produced neutron. Due to very low predictedinteraction cross-sections in the 50-60 keV neutron energyrange [43], these blips are likely to be produced tens of me-ters or more from the point of neutron production. Beyondthis, neutron capture signals in pure argon are produced onaverage hundreds of µ s after other event activity, which, dueto charge drift effects in LArTPCs, is reflected in event dis-plays by an additional spatial separation of order 20-30 cm.Even in a DUNE-sized detector, many of these neutrons arelikely to escape the active LArTPC volume or acquired datawindow. The large escaping neutron fraction is illustrated inFigure 8 by the extremely small number of events above the E blip “ E true n diagonal. Due to the large distances betweenneutron production and capture locations, it will be difficultto use capture blip signals to provide more information aboutthe neutron-producing interaction, such as final-state neutronmultiplicities [44]. However, as will be discussed in Sec-tion VIII, neutron captures in LArTPCs can serve as a valu-able source of monoenergetic MeV-scale energy depositionsfor detector calibration purposes.As neutron energies rise above a few MeV, γ -producing in- - - - -
10 1 C r o ss S e c t i on ( ba r n s ) ) g (n,n (n,p)(n,2n) Neutron Interaction Cross Sections N u m be r o f P ho t on s Energies of Individual Photons
FIG. 7. Top: Cross sections of inelastic neutron scattering in liq-uid argon. Blue squares denote interactions with photon emission,green triangles denote neutron emission, and red circles denote pro-ton emission. Data retrieved from [42] for Refs. [40, 41]. Bottom:Geant4-reported true γ -ray energies created via inelastic scatteringof 10 MeV neutrons in LAr. elastic scattering off of Ar assumes the role as the domi-nant energy loss mechanism. Thus, it would be expected that,at this energy range and above, the summed energies of blipsproduced by a neutron should be proportional to that neutron’sinitial kinetic energy. This proportionality is demonstrated inFigure 8. Summed blip energies include all neutron-derivedblips passing the 75 MeV thresholding criterion; given themany-tens-of-cm neutron interaction lengths involved, a prox-imity criterion of 60 cm with respect to the neutron generationvertex is applied. Starting at roughly 2 MeV, a proportionalityis indeed visible a few MeV below the E blip “ E true n diagonal.Figure 8 provides further illustration of this trend by plot-ting the summed blip energy of a vertical slice of monoener-getic 10 MeV neutrons. The summed blip energy peak occursat 6.6 MeV, with a resolution of approximately 1.0 MeV, or „
15% of the reconstructed peak. If we fit a linear trend to themean of the main offset peak (Figure 8) for each energy slicefrom 3 to 12 MeV, we find a slope of 0.75 MeV of summedblip energy per MeV of true neutron energy, with an interceptof about -1 MeV. The offset between blip and true neutron en-ergies can primarily be explained by the cut-off of the ( n , nγ )process near the MeV scale due to the lack of available ex- R e c on s t r u c t ed E ne r g y ( M e V ) R=60cm N u m be r o f E v en t s Total Blip Energy 0.5-60 cm
FIG. 8. Top: Summed energy of blips versus generated primaryneutron energy. Bottom: Summed energy of blips produced by 10MeV primary neutrons. In both figures, only blips above 75 keV andwithin 60 cm of the neutron’s production point are included. citable Ar states.The described linear relationship continues up to the „ Ar [45]. Without positive iden-tification of an additional free final-state neutron, this recon-structed energy offset from binding energy loss will not be re-coverable. Highly offset trends will continue with increasingneutron energy, with additional further-offset bands appearingas multiple nucleons are freed from the final-state nucleus.Neutrons of even higher energies than that described above,above 100 MeV, will be produced in interactions of GeV-scalebeam neutrinos [38]. In liquid argon, these neutrons are muchmore likely to undergo proton- and/or neutron-releasing in-elastic collisions with an argon nucleus, as suggested by Fig-ure 7 and discussed and demonstrated in Ref. [46]. The formercase will result in the production of high-energy proton tracksthat can be reconstructed using standard tools [17] or high en-ergy density blips in the general vicinity of the neutron pro-duction point. The latter case will result in a multiplicationof neutrons and subsequent repetition of the various neutronscattering and binding energy loss processes, as discussed inRef. [47].In summary, we have described the energy loss mechanismsof neutrons across all relevant energy ranges in large neutrinoLArTPCs. In particular: • For high-energy ( ą MeV) neutrons, existing large-feature reconstruction algorithms may be sufficient forperforming final-state neutron calorimetry. • Below 100 MeV, blip activity will play an essential rolein determining the energy content of final state neu-trons. • Neutron energy recovery via blip identification will bemost complete in the „ • Kinetic energy deposited in neutrino LArTPCs by ă • Final-state neutron multiplicity determination via blipactivity will be extremely difficult, due to the highlydisplaced locations of neutron captures in LAr.Having outlined these neutron-related capabilities, we nowconsider how these capabilities can be leveraged for a few dif-ferent neutrino energy ranges of interest.
B. Supernova and Solar Neutrino Neutrons
Charged current supernova and solar neutrino interactionsare kinematically constrained to produce no more than oneor two final-state neutrons. These neutrons are sub-dominantcontributors to the event-averaged neutrino energy account-ing, as discussed in Section III. In the case of the MARLEY ν e CC events described in this section, we find that 15% ofall events have one or more produced neutrons, with a neutronenergy spectrum as pictured in Figure 9. Of all ν e CC in-teractions generated in Section III, final-state neutron kineticenergy accounted for 1.7% of the total kinetic energy of allinteracting neutrinos.All neutrons below „ „ γ -raysproduced in our simulated ν e CC dataset and depicted in Fig-ure 2, neutrons were responsible for 7.5% of the total, com-prising 13% of the total energy of all pictured γ -rays. Thesenumbers further illustrate the comparatively small calorimet-ric return from collecting neutron-produced blip activity.However, as illustrated in Figure 3, neutron-producing in-teractions are the source of a substantial increase in energyresolution due to the binding energy loss associated with thefreed neutron. If 7.80 MeV of energy is required to free a neu-tron from K in 15% of ν e CC interactions, this correspondsto at least 5.1% of the total kinetic energy of all interactingsupernova neutrinos; this energy fraction is three times higherthan that consumed by the kinetic energies of these neutrons. N u m be r o f N eu t r on s Individual Neutron Energies
FIG. 9. Energies of neutrons produced by supernova ν e CC interac-tions. Vertical line indicates the energy at which neutrons no longerinelastically scatter.
Thus, it would be beneficial to be able to positively identifythe presence of a neutron in a ν e CC interaction’s final state.As mentioned in the previous sub-section, identification vianeutron capture tagging will be challenging, if not impossi-ble, with existing LArTPC technology.As another method, one can attempt to exploit the relativelylonger interaction length of final-state neutrons in liquid argoncompared to final-state de-excitation γ -rays. This method isillustrated in Figure 10 by showing total multiplicity for blipsappearing within 30 cm of the neutrino interaction vertex, andwithin 30-60 cm of the interaction vertex. All CC and EScategories are normalized to one another, and are integratedover all interacting ν e energies. As mentioned in Section III,ES events are expected to have lower average multiplicity.For CC events, multiplicities at shorter vertex-blip distancesare smaller for events containing final-state neutrons, as alarger portion of the excess energy of the final-state nucleus isspent in liberating the neutron. In contrast, neutron-containingevents have a larger proportion of high-multiplicity events atlonger vertex-blip distances. It may be possible that a multi-variate approach (e.g. a boosted decision tree) including thisas well as other variables, such as individual blip energies andprimary electron kinematics, may yield some discriminationand attendant improvement in neutrino energy recovery andresolution. C. Accelerator Neutrino Neutrons
Final-state and secondary neutrons will also carry off alarge portion of energy from interacting GeV-scale neutrinos,with neutron energies ranging in energy from the sub-MeV tohundreds of MeV scales. Hundreds of MeV of neutrino or an-tineutrino energy will regularly be lost or deposited in visibleforms as a result of production or interaction of these primaryand secondary neutrons in a LArTPC. Final-state neutrons inthe lower-energy ( ă
50 MeV) range will have properties sim-ilar to those described in the previous section, primarily pro-ducing γ -rays and subsequent blips via inelastic scattering.0 N u m be r o f E v en t s With NeutronsWithout NeutronsScattering
Number of Blips Within 0.5-30 cm N u m be r o f E v en t s With NeutronsWithout Neutrons
Number of Blips Between 30-60 cm
FIG. 10. Number of blips above 75 keV and within 30 cm (top) andbetween 30-60 cm (bottom) of the primary electron for supernova ν e CC events with neutrons (red) and without neutrons (blue), and for ν x ES events (green). The scattering contribution to the bottom panelis negligible, so it is not included. Distributions are area-normalizedand integrated over all interacting neutrino energies.
Final-state neutrons in the higher-energy range will have splitenergy depositions between γ -produced and proton-producedionization, with the possibility of many follow-on generationsof neutrons and subsequent nucleon binding energy losses.Ref. [47] explains this neutron energy accounting in detailfor GeV-scale neutrino interactions. Thus, for the purposesof completeness of our description, we will only briefly sum-marize some relevant conclusions in this paragraph, while en-couraging the reader to carefully study that excellent paper.For FLUKA-simulated 4 GeV neutrino interactions in liquidargon, 30% of hadronic energy is lost, on average, to the pro-duction (binding energy) or interaction (inelastic or elasticscattering) of final-state neutrons. Of this neutron-related bud-get, less than a third is likely to be identified using standardlarge-feature reconstruction tools, such as Pandora [17]. Oneof the largest neutron-related energy loss categories is ion-ization below quoted DUNE CDR detection thresholds [48],i.e. electromagnetic and proton-produced blip activity. Properidentification and consideration of neutron-related blip activ-ity can provide a relative improvement in energy resolution oforder 25% for both GeV-scale neutrinos and antineutrinos.Ref. [47] also notes that binding energy represents the largest contributor to neutron-related energy losses. Thus, wemight expect an additional substantial improvement in energyresolution from accurate determination of the number of final-state primary and secondary neutrons in an event. As men-tioned in the previous section, precise capture-based neutronidentification will be extremely challenging in LArTPCs. Inaddition, given the large number of average primary and sec-ondary neutrons in a GeV neutrino event and the diffusenessof their produced activity, the blip proximity method intro-duced in the previous section also seems unlikely to provideeasy insight into true neutron counts.Beyond the concretely defined improvements in energy ac-counting and resolution described above, blip multiplicities,energies, and positions represent a new source of data for con-straining modelling of hadronic interactions and energy lossmechanisms in argon, as well as modelling of nuclear effectsin neutrino-nucleus interactions. While information regard-ing final-state neutron multiplicities, such as that describedrecently in NOvA oscillation analyses [44], may not be eas-ily leveraged in LArTPCs, proxies for total neutron energyand the presence of high-energy neutrons will certainly bepresent in LArTPC events. A reduction in modelling sys-tematics enabled by analysis of MeV-scale activity in beamneutrino events could have the potential to be more valuablethan energy resolution reductions in maximizing the oscilla-tion physics reach of DUNE. V. ELECTROMAGNETIC SHOWER RECONSTRUCTION
While neutrino energy resolution improvements broughtabout by reconstruction of blip activity were demonstratedin previous sections, it is worth briefly defining calorimetricgains specifically for electromagnetic showers, such as thoseproduced in interaction of solar, supernova, atmospheric, andbeam ν e and ν e in LArTPCs.Electromagnetic showers are composed of electrons andpositrons produced by hard electron-electron scattering andbremsstrahlung photons. Many electrons produced bybremsstrahlung photons will have energies at or below theMeV-scale regime and may be lost from shower energy re-construction in the absence of low thresholds and/or topolog-ically loose feature collection criteria. These bremsstrahlungcharge loss effects are stochastic, and contribute substantiallyto overall shower energy resolution in LArTPC reconstruc-tion. These effects have been previously described in the lit-erature: MicroBooNE reports Michel electron and π electro-magnetic shower resolutions of order 20% over a range of en-ergies, with much of this resolution arising from non-inclusionof charge below MicroBooNE hit-finding thresholds or out-side of defined shower topologies [20, 31]. Similarly, LAr-IAT [21, 49] has demonstrated a visible energy resolution ofapproximately 10% for the energy deposited by Michel elec-trons within their active volume, and an average overall en-ergy resolution of about 3% for fully-contained samples ofsimulated isolated electrons spanning a similar energy range.Using the blip reconstruction procedure outlined in Sec-tion III, we have conducted a similar study of reconstructed1 R e c on s t r u c t ed E ne r g y ( M e V ) Old Reco Method vs True Michel Energy
Trunk-only Reconstruction R e c on s t r u c t ed E ne r g y ( M e V ) New Reco Method vs True Michel Energy
Trunk + Shower Reconstruction
FIG. 11. Reconstructed energy plotted against the true energy forsimulated Michel electrons. The top plot uses only the main electrontrunk, while the bottom plot also incorporates blips within 60 cm ofthe electron’s starting point. The red line indicates the expected trendfor perfect reconstruction. energy resolution for Geant4-generated electrons as done inRef. [20] and Ref. [21]. The goal of this study is to demon-strate what the limits of electromagnetic shower resolutionmight be with the maximum achievable inclusion of charge(blip activity).Figure 11 shows reconstructed versus true energies for asample of Michel electrons, which range in energy from 0 to53 MeV. We consider both the ‘default’ reconstruction case(from Section III) in which only ionization from the primaryelectron trunk is included, as well as the case that includesthe electron trunk plus the summed energy from all displacedbremsstrahlung-produced shower products and blips within60 cm of the electron start point. As expected, when incorpo-rating displaced blip activity into the total energy reconstruc-tion, a significant improvement is visible in both the accuracyand energy resolution.For further illustration, the fraction of reconstructed energyand energy resolution are plotted in Figure 12 for a sampleof isolated electrons spanning a range of 0-50 MeV. Reso-lution is calculated by composing distributions of the energyvariance, p E reco ´ E true q{ E true , across a range of true electronenergy bins, and then taking either the RMS or the width pa-rameter from a Gaussian fit to the peak of each distribution.For the electron trunk-only reconstruction case, we see an en- M ean R e c on s t r u c t ed E ne r g y F r a c t i on Trunk-onlyTrunk + Shower, R = 60 cmTrunk + Shower, R = 100 cm
10 15 20 25 30 35 40 45 50True Electron Energy (MeV) - -
10 1 F r a c t i ona l R e s o l u t i on Trunk-only, RMSTrunk + Shower, RMS (R = 60 cm)Trunk + Shower, RMS (R = 100 cm)Trunk + Shower, Peak Fit (R = 60 cm)Trunk + Shower, Peak Fit (R = 100 cm)
FIG. 12. Top: Average fraction of energy recovered versus thetrue initial electron energy when including all reconstructed activ-ity within either 100 cm or 60 cm of the electron’s start point, as wellas in the case when only the primary electron trunk is considered.Bottom: Fractional energy resolution plotted for these same threescenarios, measured using either the RMS or a Gaussian fit to thepeak of the distribution of ( E reco ´ E true q{ E true for different energybins. ergy loss ranging from only „
10% at the lowest energies toas much as „
40% near 50 MeV, with an RMS energy reso-lution in the 10-20% range — in reasonable agreement withthat reported in Ref. [20] for a similar trunk-only case. If weinstead consider a case in which we include all blip activityabove 75 keV within a 60 cm (100 cm) radius of the electronstart point, we achieve a relatively flat average energy lossof only „
5% ( „ Thus, rather than substantially improving theresolution of the primary full-energy peak of the shower, in-creasing inclusion of blip activity serves primarily to reducenon-Gaussian off-diagonal energy smearing contributions. For energies below about 15 MeV, electrons are too far below the criticalenergy to produce significant bremmstrahlung activity, and the peaks intheir distributions of energy variance become highly non-Gaussian; datapoints from these cases are therefore excluded from Figure 12. σ { E « {? E [MeV] ‘
2% for 5-50 MeV electron show-ers, which equates to a resolution of „
5% at 5 MeV thatdrops to about 2.5% at 50 MeV [21]. This {? E dependencearises from the presence of smearing introduced by the detec-tor readout and by the reconstruction process. Our ‘best case’resolutions presented in this section do not include any simu-lated smearing, and are instead limited only by energy thresh-olding and by a simple blip proximity requirement; therefore,these resolutions are found to be largely flat for energies be-low 50 MeV. Further discussion of other detector-related res-olution contributions is given in Section IX.For shower energies greater than 50 MeV, the primaryelectron and many in subsequent generations will be wellabove the electron critical energy, producing a large num-ber of stochastic bremsstrahlung features. It is worth exam-ining whether the trend in energy resolution observed for sub-critical electrons holds for this higher-energy regime. For ex-ample, in Ref. [47], it is reported that for 4 GeV ν e and ν e , a1.5% electromagnetic shower energy resolution is producedby missed depositions below a 0.1 MeV blip identificationthreshold. We observe a ‘best-case’ Gaussian-fitted energyresolution of 0.3% for simulated 200 MeV electrons usinga 75 keV threshold and no proximity requirement. Increas-ing the electron energy to 500 MeV, we find the resolutionimproves even further to 0.2%, though is then degraded to1.2% and 5.8% when the energy deposition detection thresh-old is raised to 1 MeV and 10 MeV, respectively. For physicssensitivity studies, it therefore seems reasonable to assume anenergy resolution modeling for high-energy electron showersthat is substantially better than the 15%/ ? E [GeV] ‘
2% as-sumed in previous literature, which translates to about 21% atan electron energy of 500 MeV [4, 48].
VI. PARTICLE DISCRIMINATION CAPABILITIES
We will now study the role that MeV-scale reconstructioncan play in distinguishing the identity and charge of particlesin LArTPC events. Focus will primarily be given to the roleblip activity can play in distinguishing π ` , π ´ , µ ` , and µ ´ from one another, given the limitations of existing tools forlarge neutrino LArTPCs. A. Existing Particle Identification Methods in LArTPCs
A variety of approaches have been advocated or demon-strated to use LArTPCs’ excellent calorimetric capabilitiesand mm-scale resolution to provide discrimination betweendifferent types of particles. Most prominently, energy depo-sition density helps discriminate between charged particles of
Particle Decay (%) Capture (%) Other (%) π `
72 0 28 π ´ µ `
100 0 0 µ ´
26 74 0TABLE III. End-of-life processes for 100 MeV positively and nega-tively charged muons and pions as simulated in Geant4. Decay pro-cesses will produce Michel electrons, while capture and other pro-cesses (such as absorption and charge exchange) will not. substantially differing mass, such as protons, kaons, and pi-ons [50–53], as well as enabling discrimination between high-energy electrons and γ -rays [9, 31].However, pions and muons are too close in mass to producea highly-efficient density-based separation in a large LArTPC.If high purities are desired, discrimination must include otherparameters. The presence of a Michel electron signature ineither charge or light LArTPC data can be used to enrich asample in specific muon or pion types [21]. Specifically, con-tained π ´ are far more likely to undergo nuclear capture thanto decay to a Michel electron in a LArTPC via π ´ Ñ µ ´ ` ν µ ë e ´ ` ν µ ` ν e , (5)while all µ ` will decay to a Michel electron, µ ` Ñ e ` ` ν e ` ν µ . (6)The presence of hard or inelastic scatters along the path ofa track is more indicative of a strongly-interacting pion [53].When both a muon and a pion are possibly present and sharea vertex, relative track length is also used as a proxy for par-ticle identification [12, 16]. Sign selection is also an impor-tant consideration in understanding neutrino interaction im-ages, particularly in antineutrino-mode accelerator neutrinodata, which has an outsized wrong-sign contamination. In thiscase, Michel electron identification can also be considered asa possible tool in LArTPCs for muon or pion sign determina-tion, for the reasons stated above. For the case of muons, somecharge-sign discrimination may also be gained by exploitingthe longer characteristic decay time of µ ` in liquid argon withrespect to µ ´ [21].As an example of the difficulty of disambiguating pion andmuon identities, Table III illustrates the level of pion-muonor charge-sign purity that can be achieved by Michel elec-tron identification. This table considers the specific case ofprimary particles with kinetic energy of 100 MeV generatedin an effectively infinite volume of liquid argon using LAr-Soft and the QGSP BERT HP high-energy hadronic library inGeant4. This kinetic energy corresponds to that commonlyobserved for pions produced in interactions of GeV-scale neu-trino interactions [54, 55]. Most of these pions and muonswill end their lives either decaying or capturing at rest. Evenassuming 100% efficient charge-based Michel electron identi-fication, the contamination issues are apparent. For selectionof muons (pions) only, a requirement of finding one (zero)Michel still accepts 72% (74%) of contaminating π ` ( µ ´ ).3For exclusive charge-sign selection, a required Michel countof one will produce a π ` sample nearly free of π ´ , but a µ ` sample containing 26% of contaminating µ ´ . A Michel countrequirement of zero will produce a pure µ ´ sample, but a sign-contaminated π ´ sample, due to the non-negligible nuclearabsorption of 100 MeV π ` in flight. In the case of less than100% efficient Michel electron identification, the describedseparations above will be further degraded. B. Discriminating π ˘ and µ ˘ End States Using Blip Activity
Pion-muon discrimination should in principle be possiblebased on the level of MeV-scale activity at or near the end ofa candidate pion or muon track. Pion nuclear capture leavesthe entirety of the pion’s rest mass energy in the capturing nu-cleus, which will be released in the form of final-state protons,neutrons, and de-excitation γ -rays. In addition to having lessrest mass energy to begin with, a capturing muon will con-vert a substantial portion its rest mass into invisible final-state ν µ kinetic energy. Thus, muon capture should be expected toproduce comparatively less proton, neutron, and γ -ray activ-ity around its capture point. As pion and muon decay involveno direct nuclear interaction at all, the only visible activityat the particle end point should be a Michel electron and itsattendant bremsstrahlung photons.While a complete theoretical description of final-state en-ergy accounting for the case of pion and muon capture is verycomplex and does not exist, there are many existing nuclearphysics measurements of these processes [56, 57]. In Geant4,final-state pion and muon capture on argon are modelled pri-marily using parameterizations based on existing measure-ments on lighter and heavier nuclei. We use Geant4 simula-tions of 1 MeV π ´ and µ ´ to demonstrate the relevant truth-level and reconstructed blip activity differences.Figure 13 shows final-state proton and neutron multiplic-ities for capturing muons and pions; for decaying pions andmuons, obviously these multiplicities are zero. We find thatpion capture emits more protons and neutrons than muon cap-ture. In about 75% of muon nuclear captures, no proton isemitted, while this occurs only about 20% of the time in pionnuclear capture. The substantial difference in average neutronmultiplicity indicates an expected difference in blip multiplic-ities and summed blip energies for these various cases. Inaddition, de-excitation γ -ray production at the capture vertexmay also differ between pion and muon capture as differentdaughter nuclei are produced in different excited configura-tions.To judge the level of discrimination provided by blip sig-nals, we again consider the signal blip metrics examined inprevious sections. Figure 14 shows the 2D joint distributionof blip multiplicity and total energy as in Section III, but for µ ´ and π ´ capture, and for µ ´ decay. Distributions are shownfor blips within 60 cm of the primary particle end point. Wealso note that π ` capture (decay) distributions should be quitewell-represented by the π ´ capture ( µ ´ decay) cases; µ ` de-cay blip distributions should also be well-represented by µ ´ decay. We see that, of the three possible end-state processes, N u m be r o f E v en t s PionMuon
Number of Primary Protons N u m be r o f E v en t s PionMuon
Number of Primary Neutrons
FIG. 13. Number of protons (top) and neutrons (bottom) emittedfrom π ´ captures at rest (red dashed line) and µ ´ captures at rest(blue solid line). Distributions are area normalized.Radius N blip E blip E vert µ CAR µ Decay π CAR30 cm ą – – 34% 66% 85%30 cm – ě MeV – 15% 45% 76%30 cm ą ě MeV – 12% 41% 73%60 cm ą – – 6.0% 48% 85%60 cm – ě MeV – 1.7% 41% 76%60 cm ą ě MeV – 0.80% 32% 75%60 cm – – ą MeV 18% 0% 74%60 cm ą ě MeV ą MeV 0.17% 0% 52%TABLE IV. Selection efficiency for various applied blip activity andvertex activity cuts for µ ´ captures at rest (CAR), decaying µ ´ , and π ´ CAR. The vertex region is defined by a 0.5 cm radius spherecentered at the particle’s decay or capture point; only blips foundoutside of this region are considered. pion nuclear capture at rest produces by far the most blip ac-tivity within 60 cm of the capture point. Interestingly, theMichel electron bremsstrahlung blips in muon decay appearto be more numerous than those from muon nuclear capture atrest. Thus, blip activity can provide new discrimination capa-bility independent of whether other discrimination methods,such as Michel electron identification, are employed.To demonstrate more quantitatively the level of pion-muonand sign discrimination possible using blip information, weplace a variety of cuts on blip multiplicity and summed blipenergy. In Table IV, one can see the capabilities of these cutsalone to distinguish the pion capture, muon capture, and decay4 E ne r g y ( M e V ) - Capture Energy vs Blips Between 0.5-60 cm of Vertex µ μ capture E ne r g y ( M e V ) - Decay Energy vs Blips Between 0.5-60 cm of Vertex µ μ decay E ne r g y ( M e V ) - Capture Energy vs Blips Between 0.5-60 cm of Vertex π π capture FIG. 14. Summed blip energy versus blip multiplicity within 60 cmof the capture/decay point for µ ´ captures at rest (top), decaying µ ´ (middle), and π ´ captures at rest (bottom). end processes. To further illustrate, we focus on a hypothet-ical identification of π ´ in LArTPC events. For the 60 cmproximity case, we find that by placing a cut of ě MeV( ą
14) on summed blip energy (multiplicity), we are able tocorrectly identify a π ´
76% (85%) of the time, while reject-ing all but 41% (48%) of decaying muons and 1.7% (6.0%)of capturing muons. If these two cuts are combined, we re-ject 99.2% (68%) of all capturing (decaying) muons, with a75% π ´ selection efficiency. Since pion blips are primarilyneutron-generated, a similar selection based on a 30 cm prox-imity requirement, also given in Table IV, performs substan-tially less well, particularly in discriminating pion and muonnuclear capture. It should also be noted that the ‘other’ cat-egory of pion end-states in Table III is dominated by nuclear absorption in flight, which will produce even more blip activ-ity than nuclear capture at rest, due to the additional absorbedpion kinetic energy. Thus, this high π ´ selection efficiencyshould be realizable at kinetic energies higher than the simpli-fied 1 MeV case simulated here.For comparison to the 60 cm proximity blip selectiondescribed above, a Michel-rejecting selection with perfectMichel tagging would yield 97% π ´ selection efficiencywhile rejecting 0% (100%) of capturing (decaying) muons. Inthis case, blip-based discrimination excels where the Michel-based selection performs less well, and vice-versa. This em-phasizes the value of combining blip-based discriminationwith the other forms described in the previous sub-section;in this case, the combination of methods would yield multi-ple orders of magnitude reduction in muon contamination. Asimilar level of discrimination as that described above shouldbe achievable when considering an exclusive selection of µ ` .To demonstrate sign selection capability, we use the resultsof Tables III and IV to consider the case of 100 MeV π ` .97% of 100 MeV π ´ will decay and be rejected at a rate of75% if blip multiplicity and total energy cuts in Table IV areinverted. After adding a small contribution from π ´ decay,we obtain a „
72% reduction of wrong-sign π ´ backgroundwith a π ` efficiency of roughly 68%. This purification canobviously be substantially improved with a Michel electronrequirement. For wrong-sign purification of µ ` , µ ´ capturewith 74% probability and can be rejected at a rate of 80%if blip summed energy and multiplicity cuts are adjusted to ą ą
4, respectively. This would produce a 60%efficient µ ` selection while rejecting 60% of µ ´ .While outside the realm of blip reconstruction so far con-sidered in this paper, the proton final-state multiplicities inFigure 13 are also worth considering in the context of par-ticle discrimination. In particular, pion capture will producefinal-state protons, which will produce either tracks or excessionization at the capture vertex beyond that expected froma pion or muon Bragg peak. To estimate the discriminationpower provided by these protons, we consider the ambitiouscase where we are able to positively identify the presence of aproton with energy in excess of 5 MeV [58]. For this case, wesee that a cut on ą VII. BSM PHYSICS CAPABILITIES
A variety of BSM searches in large neutrino LArTPCs canbenefit from the identification of blip activity and classifica-tion of events based on the presence or absence of these fea-tures. We will briefly highlight a few promising scenarios hereand encourage the performance of more quantitative studiesin the future using full simulations of the BSM processes andfinal-state distributions in question.Many BSM physics processes discussed as possible points5of focus within the SBN and DUNE physics programs can becategorized by the variety of distinctive particle combinationsthey produce [5, 8, 59]. For example, high-energy di-leptonpairs can be expected from Standard Model neutrino tridentinteractions in liquid argon [60, 61], which have the poten-tial to uncover new physics — such as heavy sterile neutri-nos [62], a dark neutrino sector [62], or dark Higgs [63, 64] —if rates are divergent from Standard Model predictions. Otherspecific particle combinations have also been hypothesized:for example, pion-muon pairs would be expected from decaysof heavy neutral leptons [12] produced in accelerator neu-trino experiments, while pion-pion pairs could be producedin these experiments by decays of dark Higgs bosons [63] orup-scattered dark neutrinos [65], respectively.It is expected that the primary backgrounds to these dedi-cated BSM searches are different final-state particle combina-tions produced by common Standard Model neutrino interac-tions. For example, Ref. [60] provides an excellent overviewof the various expected background channels to neutrino tri-dent µ ` - µ ´ production, particularly 1 µ -1 π final states from ν µ CC interactions. The pion-muon discrimination capabilitydelivered by analysis of reconstructed blips, as described inSection VI above, may be a useful additional tool in reducingbackgrounds for this and other BSM analyses in large neutrinoLArTPCs.Other BSM signatures can be characterized primarily bythe unique topological distribution of blip signals they pro-duce in LArTPC events. An obvious example is searchesfor millicharged particles produced in neutrino beams, asrecently demonstrated by the ArgoNeuT experiment on theNuMI beamline [66]. The track of weak ionization producedby these particles would be visible in a LArTPC as two ormore reconstructed blips that can be connected by a line point-ing back to the neutrino beam’s target [67]. We note that low-ered LArTPC blip thresholds in these searches leads directlyto improvements in sensitivity. Other hidden sector particleinteractions in LArTPCs, such as up-scattering and decayingheavy neutral leptons [5], can produce two displaced eventvertices, one of which consists of a de-exciting nucleus ex-hibiting primarily or exclusively reconstructed blip activity.Thus, identification of these secondary low-activity verticesis likely only possible through the use of blip identificationcapabilities.Hidden sector physics scenarios may also be characteris-tic in the total level, rather than the spatial distribution, ofMeV-scale activity present in events. For example, decaysof hidden sector particles in LArTPCs, such as heavy neutralleptons, dark photons, or dark Higgs, need not include sub-stantial momentum exchange with an argon nucleus. Thesedecay vertices, unlike neutrino-argon interactions, will not in-clude the neutron and photon products of final-state nuclearde-excitation, resulting in an event with little or no blip ac-tivity near the interaction vertex. This lack of blip activityis another possible input for reducing neutrino-induced back-grounds to these BSM scenarios.
VIII. SINGLE γ -RAY CALIBRATION ANDSPECTROSCOPY Previous sections describing the utility of MeV-scaleLArTPC signals have focused on a handful of metrics re-lated to total blip energy or multiplicity. In addition, LArTPCphysics analyses may be enhanced by considering individualblips or blip sub-groups within an event. In this section, wewill focus primarily on the benefits of blip sub-grouping forperforming MeV-scale single γ -ray spectroscopy in LArTPCevents. This technique could be valuable for different pur-poses, such as low-energy LArTPC calibration [8] or taggingof final-state nuclei produced in neutrino or BSM interac-tions [68].We have attempted to resolve γ -ray spectrum features in anevent by iteratively forming sub-groups of blips produced byelectrons that are daughters of the same parent γ -ray. Prox-imity is our sole metric in determination of common parent-age, with grouping achieved by the following algorithm. First,we identify all of the blips in an event by tagging electronsthat deposit at least 75 keV of energy. Then, a candidate ‘re-constructed γ -ray’ is formed by grouping all identified blipslocated within a 30 cm spherical radius centered around thehighest-energy blip. The blips present in this reconstructed γ -ray candidate are then removed from consideration, and theprocess is repeated using the remaining blip of highest en-ergy. Formation of reconstructed γ -rays continues until noblips above our energy threshold of 75 keV remain in theevent. The primary reconstructed γ -ray metric investigatedhere will be total energy.To generate reconstructed γ -ray energy spectra moreclosely resembling those attainable from a LArTPC, we applya 50 keV energy smearing to each blip’s energy to simulate theimpact of electronics noise on LArTPC ADC signals; this en-ergy smearing choice is guided by measurements of raw wirewaveform noise in MicroBooNE [24, 35]. Further discussionof the limitations presented by electronics noise in blip analy-ses will be given in Section IX.It is likely that a more detailed analysis of blip sub-groupingwill yield algorithms with improved spectroscopic perfor-mance. In particular, optimal grouping criteria are likelyto be dependent on the exact signal in question. It alsoseems likely that additional spectroscopic information can begleaned through combined consideration of reconstructed γ -rays’ total energies and blip multiplicities [69]. Nonetheless,here we forego these considerations, as the method describedabove is sufficient to demonstrate the value of blip activity inperforming MeV-scale γ -ray spectroscopy in LArTPCs.The benefits of blip sub-group metrics are illustrated by ap-plying the blip reconstruction technique from Section II andthe blip grouping algorithm described above to LArSoft sim-ulation of individual γ ray and neutron samples of varioustypes. For each sample, 10 total events are produced. • A single 1.46 MeV γ -ray: This sample can be used todirectly characterize the impact of thresholding on γ -ray calorimetric capabilities at the MeV scale. This en-ergy reflects that of γ -rays preferentially produced in6neutron inelastic scattering off Ar, as visible in Fig-ure 7. • Two 1.46 MeV γ -rays generated in the vicinity of oneanother: γ -rays are simulated 30 cm apart, traveling atrandomized angles. This sample enables us to investi-gate the ability to separate secondary electrons from dif-ferent γ -rays, and to examine the effects of blip pile-upon reconstructed γ -ray spectra, resolutions and biases. • A single neutron capture in liquid argon: A largely mo-noenergetic 6.1 MeV signal from capture on Ar, com-prised of a cascade of γ -rays of varying energy [70],likely to be observed in large LArTPCs like DUNE. • A single 10 MeV kinetic energy neutron. These events,as described in Section IV, will generate many γ -rayswith a variety of true energies. We can attempt to re-construct spectral features within this realistic mass ofoverlapping Compton electron activity.The energies of individual blips in events containing a sin-gle simulated 1.46 MeV γ -ray, as well as energies of re-constructed γ -rays using the iterative sphere-based groupingmethod described above, are shown in Figure 15. In the indi-vidual blip spectrum, a Compton edge is observed at roughly1.25 MeV, as would be expected from a 1.46 MeV incident γ -ray. This edge is accompanied by dramatically increasingblip counts at lower energies. As blip sub-groups are formed,this low-energy tail is decreased in magnitude as a prominentpeak emerges at an energy above that of the Compton edgein the single-blip spectrum. The latter feature represents thereconstructed γ -ray full-energy peak, which is biased down-ward from the true γ -ray energy of 1.46 MeV due to the non-collection of energy in below-threshold blips. The full-energypeak is fit to a Gaussian function with a linear backgroundcomponent to account for the overlap from the distribution ofincomplete γ -ray candidate energies to the left of the peak.The Gaussian fit provides a mean of 1.32 MeV, biased -9.5%from the true energy, as well as a 1 σ resolution of 0.13 MeV,or 9.5%. Taking the integral of the Gaussian component ofthe fit and dividing by the total number of simulated events,we calculate a ‘peak efficiency’ of 75%. This indicates that theexisting algorithm is relatively efficient in its grouping of blipsoriginating from a common γ -ray. These performance met-rics are summarized in Table V. It should be noted that sinceenergy peaks reported on in this section are non-Gaussianto varying degrees, values reported in this table fluctuate atthe few-percent level based on exact fitting assumptions andranges.Similar metrics are provided for cases where altered blipgrouping settings have been applied. If thresholds are raisedto 150 keV, the resulting full-energy peak bias, resolution, andefficiency come out to -18%, 7.8%, and 55%, respectively. Forthis case it appears that total energy is biased further down-ward from the expected true energy, while peak efficiency isalso degraded significantly. Meanwhile, if we return to the75 keV energy threshold but reduce the sub-group proximityto 20 cm, these metrics are altered to -11%, 10.4%, and 62%, E n t r i e s E n t r i e s Energy (R = 30 cm)
FIG. 15. Reconstructed energies of individual blips (top) andgrouped-blip reconstructed γ -ray energies produced from a LArSoftsimulation of single 1.46 MeV γ rays (bottom). The γ -ray’s Comp-ton shoulder is visible in the single-blip spectrum, while the full-energy peak is the most prominent feature in the reconstructed γ -rayspectrum. E n t r i e s Energy (R = 30 cm)
Two 1.46 MeV gammasR = 30 cm
FIG. 16. Reconstructed γ -ray energies produced from a LArSoftsimulation of two 1.46 MeV γ -rays generated at random angles at aseparation distance of 30 cm. The full-energy peak and the pile-uppeak containing energies of both γ -rays are most prominent featuresin the spectrum. respectively. In this case, the energy bias and resolution re-main relatively stable compared to the 30 cm radius scenario,while peak efficiency is worsened.The energies of reconstructed γ -ray candidates identifiedin simulated events containing two mono-energetic γ -rays are7 Sample & Sphere Radius E γ Bias(%) 1 σ Res.(%) Peak Eff.(%) Pile-Up(%)1 γ , 30 cm -9.5 9.5 75 -1 γ , 30 cm (150 keV) -18 7.8 55 -1 γ , 20 cm -11 10.4 62 -2 γ , 30 cm -9.5 10.3 107 282 γ , 20 cm -11 11.2 123 7.0 n - Ar capture, 60 cm -8.2 5.0 58 1.810 MeV n , 30 cm -10.5 10.2 27 14410 MeV n , 20 cm -11.8 11.2 37 146TABLE V. Total energy bias, resolution, and efficiency metrics fordifferent γ -ray samples, each containing 100k simulated events, us-ing a blip energy threshold of 75 keV and proximity requirement of30 cm. A variety of alternate threshold and proximity cases are alsoshown. shown in Figure 16, with performance metrics also outlinedin Table V. The full-energy peak of the single γ -ray from Fig-ure 15 is again apparent in this sample’s energy spectrum, witha similar bias and resolution: -9.5% and 10.3%. Thus, single γ -ray spectroscopy can still be performed even when signalsfrom multiple γ -rays are present in the same event region. Apeak efficiency of 107% is produced, indicating that, on av-erage, one of the two simulated γ -rays will have its energyproperly reconstructed.We also note the additional peak at roughly twice the energyof the first peak; this feature is the result of grouping energiesfrom the two different γ -rays into one reconstructed γ -ray.We characterize the size of this effect by counting the num-ber of reconstructed γ -rays ą σ above the single full-energypeak and dividing by the total number of simulated two γ -rayevents; this metric is referred to as the ‘pile-up fraction.’ Ap-plying the default blip reconstruction and sphere-based group-ing methods to this two γ -ray sample, we observe a pile-upfraction of 28%. If the smaller 20 cm group proximity re-quirement is used on the two γ -ray sample, the single γ -rayenergy resolution is again modestly degraded as in the one γ -ray case. However, higher fidelity is achieved in γ -ray en-ergy grouping: peak efficiency increases to 123%, while thepile-up fraction reduces to 7.0%.The reconstructed γ -ray energy spectra following the simu-lation of single 1 eV thermal neutrons are shown in Figure 17for both a 30 cm and 60 cm proximity requirement, with per-formance metrics outlined in Table V. For this sample, theneutrons are not energetic enough to produce the 1.46 MeV γ -rays that are characteristic of our other samples; instead,our full-energy peak corresponds to 6.1 MeV, the total energyof γ -rays emitted during neutron capture on Ar. Applyingthe default γ -ray reconstruction to this sample yields a distri-bution that remains largely flat above 1 MeV, with a mutedfull-energy peak. Thus, it appears that the default 30 cm blipproximity requirement is better tuned to the identification ofindividual γ -rays as shown in previous samples, but it is insuf-ficiently wide to capture the energy of all γ -rays from the neu-tron capture cascade. At the same time, the spectrum does notreflect the rich underlying forest of true monoenergetic γ -raysproduced by the Geant4 simulation, highlighting the com-bined limitations of our technique and inherent LArTPC ca- E n t r i e s Energy (R = 30 cm)
Neutron captureR = 30 cm E n t r i e s Energy (R = 60 cm)
Neutron captureR = 60 cm
FIG. 17. Reconstructed γ -ray energies produced from LArSoft sim-ulation of captures of 1 eV primary neutrons, using either a 30 cm(top) or 60 cm (bottom) proximity requirement. For the 60 cm case,the 6.1 MeV peak from capture on Ar is clearly visible. E n t r i e s Energy (R = 30 cm)
10 MeV neutronR = 30 cm
FIG. 18. Reconstructed γ -ray energy spectrum produced from LAr-Soft simulation of 10 MeV primary neutrons, using a 30 cm blipproximity requirement. The 1.46 MeV peak corresponding to thefirst excited state of Ar is clearly visible. pabilities. If the proximity requirement is loosened to 60 cm,a clear peak is visible just below 6.1 MeV. This peak exhibitsa bias of -8.2%, a resolution of 5.0%, and a peak efficiency of58%. We also note the existence of a much smaller peak justbelow 8.8 MeV due to neutron capture on Ar.In Figure 18, we plot reconstructed γ -ray energies for mo-noenergetic 10 MeV fast neutrons, as might be producedby neutrino interactions, nuclear interactions of final-state8heavy charged particles, or on-surface cosmic rays. Anexponentially-decreasing spectrum is observed with a clearpeak in the vicinity of the 1.46 MeV first excited state of Ar.Using the default reconstruction, a combined Gaussian pluslinear fit yields a 1.46 MeV peak bias of -10.5%, and a reso-lution of 10.2%, comparable to that obtained from the single1.46 MeV γ -ray sample. Based on the area of the fitted Gaus-sian, we see that for every simulated 10 MeV fast neutron,we identify 0.27 well-reconstructed 1.46 MeV γ -rays. Thelarge pile-up fraction for this dataset is produced by overlapof γ -ray signals from multiple inelastic neutron scatters andfrom de-excitations of higher-lying states of Ar. The latteris likely responsible for the additional energy peak appearingat roughly 2.2 MeV.Both capture and inelastic scatter γ -ray signals will be nat-urally produced during operation of all LArTPC detectors,whether in signal neutrino interactions, or in background ra-diogenic and cosmogenic processes. Thus, these can servealongside Ar as additional naturally-occurring low-energycalibration signals in existing and future large LArTPCs.
IX. LIMITING FACTORS IN LARTPC MEV-SCALERECONSTRUCTION
In summarizing the uses of blip activity in the previous sec-tions, we have deliberately overlooked the discussion of someof the possible limitations of this method. In this section wewill summarize what we see as the most obvious possible lim-itations, and will then either quantitatively assess their impactor suggest avenues for future assessment. A. Ar Contamination
Blip activity is or will be ubiquitous in the event displays ofall current and future planned large LArTPCs due to the natu-ral presence of 1 Bq/kg specific activity of Ar in the liquidargon [71]. While there are some benefits to the presence ofthis signal for detector response calibration [8], its β decayelectrons are an irreducible background for the purposes ofuncovering physics with non-radiogenic blip activity. Due toour knowledge of Ar specific activity in LArTPCs, however,it is straightforward to estimate the impact.To do so, we have used the existing radioactivity generatorin LArSoft to simulate the expected density of Ar β decaysin our generic LArTPC volume. Using random points withinthe center of the argon volume as candidate vertices, we applythe blip selection requirements described in Section II whilevarying the extent of the applied proximity requirement from10 to 150 cm. Blip activity metrics obtained using this pro-cedure and dataset are shown in Figure 19. As would be ex-pected, as the volume considered for blip reconstruction in-creases, blip multiplicity, summed energy, and RMS energyspread of Ar blips increase. When a proximity of more thana meter is considered, energy biases of order 10 MeV are pro-duced, with multi-MeV RMS spreads in energy contribution.For this reason, we have considered only sub-meter proxim-
20 40 60 80 100 120 140Sphere Radius (cm) −
10 1 E ne r g y R M S ( M e V ) E > 75 keVE > 300 keV
20 40 60 80 100 120 140Sphere Radius (cm) − − −
10 110 M ean E ne r g y B i a s ( M e V ) E > 75 keVE > 300 keV
20 40 60 80 100 120 140Sphere Radius (cm) − − −
10 110 M ean B li p M u l t i p li c i t y E > 75 keVE > 300 keV
FIG. 19. The contribution to blip summed energy resolution (top),bias (middle), and multiplicity (bottom) due to the presence of blipsproduced by background Ar β decays for varying proximity re-quirements and blip energy thresholds. ity in the physics analyses shown above. As the Q-value ofthe Ar β decay is 0.565 MeV, these contributions are onlymodestly reduced at a higher threshold of 300 keV, as shownin Figure 19.With a blip energy threshold of 75 keV and a proximityrequirement of 30 cm (60 cm), an average energy bias ofroughly 0.1 MeV (1 MeV) is expected, with an RMS spread9of 0.2 MeV (0.6 MeV). In particular, this RMS spread can becompared directly to the calorimetric resolutions and distribu-tions reported in the previous sections. In Sections III, IV,and V, reported energy resolutions are found to be largerthan this: for supernova neutrinos, resolution is „
20% witha 30 cm proximity cut, for 10 MeV neutrons, resolution is1 MeV with a 60 cm cut, and for low-energy electromag-netic showers, resolutions are „
10% with a 60 cm cut. In thecase of single γ -ray spectroscopy, all described sources hadfull-energy peak resolutions of order 0.15-0.25 MeV. Thus,for physics processes depositing above roughly 5 MeV, Aractivity can negligibly impact calorimetric capabilities; blipthresholding plays a much more important role for these eventclasses. For MeV-scale physics processes, such as very-low-energy supernova neutrinos, solar neutrinos, and single γ -rayspectroscopy, Ar blips will likely play a non-negligible role,and should be considered when modelling achievable energyresolution.In Sections III and VI, blip multiplicity and summed en-ergy were used for interaction channel and particle discrimi-nation. Smearing of these distributions due to Ar contam-ination was not considered. However, it can be seen that thelevel of smearing from this source is substantially smaller thanthe binning of the figures used to demonstrate the discrimina-tion capability. Thus, it seems that any reduction in discrimi-nation capability is not likely to be greater than of order 10%.
B. Electronics Noise
While we have implicitly acknowledged in this study thatelectronics noise will define achievable blip reconstructionlow-energy thresholds, we have in most cases not addressedthe contribution of electronics noise to the energy resolutionof reconstructed blip information. On the contrary, we haveassumed perfect correspondence between reconstructed blipenergy and true electron energy deposition.For wire-based charge readout systems, an electronics noiseof ă
400 and ă
700 electron-equivalent noise charge ( e ´ enc)has been achieved on all MicroBooNE and ProtoDUNE wireplanes, respectively [8, 35]. Noise floor performance is ex-pected to be enhanced with a pixel-based charge readout sys-tem, with „ e ´ enc [28], while somewhat degraded ina dual-phase DUNE module, with „ e ´ enc expected[29]. These values should be compared to an expected muonminimum ionizing particle e ´ enc of order 15,000 to 20,000for MicroBooNE and DUNE. Based on these numbers and asimple scaling argument, one would expect electronics noiselevels in various large LArTPCs to range from approximately20-80 keV. As mentioned earlier, when considering integratedMicroBooNE waveforms over a wire-time tick area compa-rable to that expected from Ar blips, an average blip noiselevel of „
50 keV is observed [24].A direct comparison of this 50 keV single-blip noise levelto the results in previous sections indicates that noise contribu-tions are likely to be sub-dominant for many of the calorimet-ric and discrimination use cases discussed above. For exam-ple, the 1 MeV calorimetric resolution for 10 MeV neutrons
10 20 30 40 50 60 70Per-Blip Smearing [keV]02468101214161820 F W H M E ne r g y R e s o l u t i on ( % )
75 keV Threshold150 keV Threshold
Single 1.46 MeV gamma
FIG. 20. The resolution of the full-energy peak for simulated1.46 MeV γ -rays, over a range of different blip smearing levels,for both 75 keV and 150 keV energy thresholds. A proximity re-quirement of 30 cm is used. Resolution is calculated based on theFWHM of the peak using the relationship to standard deviation: σ “ FWHM {p ? q . shown in Figure 8 is more than an order of magnitude largerthan this estimated noise level per blip; noise levels appearsimilarly small compared to the 10+% supernova neutrino en-ergy resolutions shown in Figure 4. To provide context forinteraction and particle discrimination capabilities, we notethat Figure 6 and Figure 14 are binned in 1 MeV increments,much more coarsely than any additional smearing one mightexpect from noise.The lowest observed resolutions discussed in this paper ap-pear in Sections V and Section VIII, where shower calorime-try and single-gamma spectroscopy are discussed, respec-tively. These sections show full-energy peak Gaussian res-olutions of order 50-500 keV (Figure 12) and 120-300 keV(Table V), respectively, much closer to expected single-blipnoise levels. Thus, in these cases, it seems likely that simula-tion of noise effects will be important in determining realisticcapabilities.As a demonstration of the impact of noise, we show in Fig-ure 20 the variations of observed full-energy peak resolutionfor the single 1.46 MeV γ -ray dataset described in the previ-ous section as per-blip noise smearing is varied from 10 keVto 70 keV. To reduce fitting dependencies for this relativelyclean peak, we calculate the fractional FWHM resolution withrespect to the FWHM window midpoint; then, to enable amore direct comparison to Table V, we scale the result by therelation between FWHM and the 1 σ width expected from aGaussian distribution: σ “ FWHM {p ? q . We remindthe reader that a default per-blip noise smearing of 50 keVwas applied to produce the results in Table V. In Figure 20,resolution is observed to decrease modestly as the appliednoise smearing is decreased: when smearing is reduced to10 keV, resolution is improved from 11.5% to roughly 8.5%.A similarly-sized decrease is observed if a higher blip thresh-old of 150 keV is chosen. Thus, for the purposes of gammaspectroscopy, it is clear that per-blip noise smearing has amodest but non-negligible impact on achieved resolutions.Using our current truth-based simulation method, it is more0difficult to convey the impact of noise on electromagneticshowers. For this signal type, the average energy per topo-logical feature (trunk or blips) is substantially higher than theprevious single-gamma case, meaning that many features willconsist of more than one or two above-threshold charge col-lection elements (wires or pads). In these cases, each element(rather than each blip) generates a fixed noise contribution. Aswe do not simulate individual charge collection elements, weare not able to comment accurately on this case: given that itwill generate far more hit elements than blips, noise smear-ing contributions are certain to be higher than what would beestimated using our truth-based methods. As a remedy, wewould encourage readers to examine the discussion of noisecontributions to low-energy electron showers from LArIATand ICARUS given in Refs. [21] and [22]. C. Other Detector Response Features
Our truth-level study procedures also do not account for avariety of other features of LArTPC detector response. Dueto the variability of many of these features between LArTPCsor to their indirect relationship to the blip physics studies pre-sented here, we will only comment briefly on them.Triggering LArTPCs to capture primary electron and blipsignals from low-energy neutrino events is a challenge thatmust be addressed by future LArTPC experiments. DUNEsupernova and solar studies have identified triggering scenar-ios producing high ( ą D. Pile-Up of Blip Activity From Many Physics Sources
Most of the physics capabilities afforded by blip recon-struction described in this paper have been demonstrated inotherwise empty LArTPC environments. In reality, this willrarely be the case. For on-surface LArTPCs, cosmic rays willbe a source of constant activity totally unrelated to any inter-esting physics events, including blip activity. For example,Refs. [66] and [24] provide measurements of cosmogenic-related blip activity for ArgoNeuT and MicroBooNE, respec-tively, before and/or after applying various forms of trackproximity-related blip rejection. Even absent cosmic ray ac-tivity, whether by being deep underground or by using of-fline data filtering, high-energy physics processes will pro-duce multiple final-state particles producing different kinds ofblip activity in different locations. These different populationscan overlap, an effect that has the potential to make targetedcalorimetry and identification tasks much more difficult.Unlike the first two limitations discussed, the level to whichthis effect limits the utility of blip reconstruction is completelydependent upon the application being considered. Thus, wedo not attempt to quantify this limitation for all scenarios, andinstead highlight two cases that represent the large range ofpossible impacts. For the case of supernova neutrino or solarneutrino detection, pile-up from separate physics events (i.e.different supernova or solar neutrino interactions) should haveessentially no impact on the calorimetric or interaction chan-nel identification information discussed in Section III. For a10 kpc distance supernova, even during the moment of highestinteraction vertex density at the arrival of the neutronizationburst flux, no more than a few dozen interactions are expectedin an entire event for the 40 kt DUNE detector. On the otherhand, consider the case of using blip activity to perform pion-muon discrimination on final-state tracks from GeV-scale neu-trino interactions. Blip activity is likely to provide the mostutility here for low-energy pions, which have a higher prob-ability of capturing at rest and producing an appearance verysimilar to a stopping muon. In these cases, the pion will of-ten end its life within 30 cm of the neutrino interaction ver-tex, resulting in a large degree of blip activity overlap from γ -rays and neutrons produced both at the pion end point and the neutrino interaction vertex. Detailed simulation and studyof these substantially-overlapped cases will be essential to un-derstanding the usefulness of blip-based information in them. E. Imperfect Nuclear Physics Simulation in Argon
For many of the studies in the paper, we have relied onGeant4 and MARLEY modelling of final-state neutron and γ -ray multiplicities and energies for complex nuclear inter-actions on argon. With the exception of neutron scatteringand capture, many of the processes discussed have never beenmeasured in argon. Thus, we stress that our studies are meantto highlight the potential for using blip-related information,rather than to provide authoritative predictions of attainablecapabilities. Before full use of some of these methods forhigh-level physics analysis, it would be prudent to assess1Geant4 and neutrino generator modelling of these final-stateproducts with dedicated measurements or systematics studiesusing LArTPC test beam experiments, meson decay-at-restneutrino LArTPC experiments, and neutrino beam LArTPCexperiments. X. SUMMARY
Using truth-level MC simulations in a generic liquid argonvolume, we have demonstrated how the unique combinationof excellent position resolution and low energy thresholds canbe leveraged to provide new information about particle inter-actions from the MeV to GeV scale in large neutrino LArT-PCs. The reconstructed positions and energies of compact,topologically isolated energy depositions of MeV-scale elec-trons, or blips, have been shown to enable better understand-ing of the identities and energies of the ancestors that createdthem. This paper has outlined the following uses for blip ac-tivity in large neutrino LArTPCs: • Improved calorimetry and interaction channel discrimi-nation for supernova and solar neutrino interactions • Calorimetry of final-state uncharged particles (such as γ -rays and neutrons) produced in high-energy interac-tions of neutrinos and other particles with argon nuclei • Improved calorimetry for electromagnetic showers • Improved discrimination and sign selection capabilitiesfor pions and muons • Improved sensitivity for BSM searches by enabling im-proved background rejection and/or identification ofinteraction-specific topological features. • Spectroscopy of single MeV-scale γ -rays.This list of use cases is certainly non-exhaustive: we fore-see broad applications including using nuclear decays and fi-nal state nucleus tagging, low-energy particle identification,and detector calibration, and anticipate further possibilitieswill be identified in the future. These capabilities should begenerally applicable to all existing and future LArTPC experi-ments, such as the SBND, MicroBooNE, and ICARUS experi-ments in the Fermilab SBN Program and the ProtoDUNE andDUNE experiments. Many of these concepts and use casesare equally relevant to other particle detector technologiespossessing excellent positioning and threshold combinations,such as opaque scintillator detectors [76] or optical TPCs [77].In demonstrating these capabilities, we have also identi-fied notable limitations of blip-based information. Whilecalorimetry of final-state neutrons in LArTPCs is enabled byblip reconstruction, this capability is degraded by the likelyunrecoverable loss of primary and secondary neutron bindingenergy; further, final-state neutron multiplicity determinationwill be difficult, if not impossible. The ubiquitous presence of Ar decays in argon limits the scope of blip reconstruction; fortunately, for all but the lowest considered MeV-scale ener-gies, Ar blip contamination is likely to play a sub-dominantrole with respect to to blip energy thresholding. Finally, forsome physics analysis scenarios, overlap of blip activity fromdifferent physics processes is likely to degrade the capabilitiesdescribed above.Analyses focused on blip activity have already been per-formed using the ArgoNeuT LArTPC [23, 66], and studiesare now also underway in other LArTPC experiments, such asMicroBooNE and ProtoDUNE. However, blip reconstructionshould not be relegated solely to the realm of dedicated stud-ies. Hopefully, we have convinced the reader that blip activitycan play a role in many of the centerpiece LArTPC physicsanalyses expected in the coming decades, such as beam and at-mospheric oscillation measurements, BSM searches, and so-lar and supernova neutrino studies. Indeed, blip activity pro-vides valuable information often orthogonal to that providedby the larger topological features in LArTPC events.As we see it, there are two technical roadblocks that slowa more complete implementation of blip activity in LArTPCphysics analyses. The first is the lack of a standard soft-ware toolkit focused on reconstruction of low-energy fea-tures in LArTPCs. Such a toolkit could standardize low-levelthresholding, identification, and position/energy reconstruc-tion tasks that should be relatively common across LArTPCexperiments, and provide for the end user blip physics ob-jects in a format similar to that currently provided fortracks/showers/particles by Pandora [17]. This will lower thebarrier to entry for new analysis by making low-energy fea-tures as easily accessible as high-energy ones. Inclusion ofblips in mainline physics analyses is also hampered by the un-certainty in the underlying nuclear modelling that determinesthe appearance of blip activity in most of the use cases con-sidered. As mentioned in the previous section, this limitationmust be resolved via dedicated measurements and subsequentmodel tuning, as has been done recently for argon-neutroninteractions [46, 70], argon-pion hadronic interactions [53],and more generally for neutrino interactions on heavy nu-clei [78, 79].
ACKNOWLEDGMENTS
This work was supported by DOE Office of Science, underaward No. de-sc0008347, as well as by the IIT College of Sci-ence and Rutgers University–New Brunswick School of Artsand Sciences. We thank Russell Betts, Avinay Bhat, Erin Con-ley, Ryan Dorrill, Roni Harnik, Zhen Liu, and Pedro Machadofor insightful discussions that helped motivate this study, andespecially thank David Caratelli, Steven Gardiner, Shirley Li,Ornella Palamara, and Tingjun Yang for their insights on por-tions of the presented analysis. We also acknowledge the Ar-goNeuT Collaboration for the use of its LArTPC event displayand LArSoft implementations, which were primarily used forthis analysis.2 [1] SBND Collaboration, (2020), https://sbn-nd.fnal.gov/.[2] R. Acciarri et al. (MicroBooNE), JINST , P02017 (2017).[3] ICARUS Collaboration, (2020), http://icarus.fnal.gov/.[4] M. Antonello et al. (MicroBooNE, SBND, ICARUS), (2015),arXiv:1503.01520 [physics.ins-det].[5] P. A. Machado, O. Palamara, and D. W. Schmitz, Ann. Rev.Nucl. Part. Sci. , 363 (2019).[6] B. Abi et al. (DUNE), (2020), arXiv:2002.02967 [physics.ins-det].[7] Proton Improvement Plan II, (2020), https://pip2.fnal.gov/.[8] B. Abi et al. (DUNE), (2020), arXiv:2002.03005 [hep-ex].[9] R. Acciarri et al. (ArgoNeuT), Phys. Rev. D , 072005 (2017).[10] ArgoNeuT Collaboration, (2020), arXiv:2002.01956 [hep-ex].[11] MicroBooNE Collaboration, (2018), MICROBOONE-NOTE-1054-PUB.[12] P. Abratenko et al. (MicroBooNE), Phys. Rev. D , 052001(2020).[13] R. Acciarri et al. (ArgoNeuT), Phys. Rev. D , 012008 (2014).[14] C. Adams et al. (MicroBooNE), Eur. Phys. J. C , 248 (2019).[15] P. Abratenko et al. (MicroBooNE), JINST , P10010 (2017).[16] P. Abratenko et al. (MicroBooNE), Phys. Rev. Lett. ,131801 (2019).[17] R. Acciarri et al. (MicroBooNE), Eur. Phys. J. C , 82 (2018).[18] R. Acciarri et al. (MicroBooNE), JINST , P03011 (2017).[19] C. Adams et al. (MicroBooNE), Phys. Rev. D , 092001(2019).[20] R. Acciarri et al. (MicroBooNE), JINST , P09014 (2017).[21] W. Foreman et al. (LArIAT), Phys. Rev. D , 012010 (2020).[22] S. Amoruso et al. (ICARUS), Eur. Phys. J. C , 233 (2004),arXiv:hep-ex/0311040.[23] R. Acciarri et al. (ArgoNeuT), Phys. Rev. D , 012002 (2019).[24] MicroBooNE Collaboration, (2018), MICROBOONE-NOTE-1050-PUB.[25] C. Anderson et al. , JINST , P10019 (2012).[26] ESTAR: Stopping Power and Range Tables for Electrons,(2020), https://physics.nist.gov/.[27] I. Lepetic, Ph.D. thesis, IIT, Chicago (2020).[28] D. Dwyer, M. Garcia-Sciveres, D. Gnani, C. Grace, S. Kohn,M. Kramer, A. Krieger, C. Lin, K. Luk, P. Madigan, C. Mar-shall, H. Steiner, and T. Stezelberger, Journal of Instrumenta-tion , P10007 (2018).[29] DUNE Collaboration, (2018), arXiv:1807.10340v1[physics.ins-det].[30] D. Caratelli, Ph.D. thesis, Columbia University (2018).[31] C. Adams et al. (MicroBooNE), JINST , P02007 (2020).[32] S. Agostinelli, et al. (GEANT4 Collaboration), Nucl. Instrum.Meth. A506 , 250 (2003).[33] E. Snider and G. Petrillo, Journal of Physics: Conference Series , 042057 (2017).[34] S. Gardiner,
Nuclear Effects in Neutrino Detection , Ph.D. the-sis, University of California, Davis (2018).[35] R. Acciarri et al. (MicroBooNE), JINST , P08003 (2017).[36] F. Capozzi, S. W. Li, G. Zhu, and J. F. Beacom, Phys. Rev. Lett. , 131803 (2019).[37] Q. R. Ahmad et al. , Phys. Rev. Lett. , 011301 (2002).[38] M. Elkins et al. (MINERvA), Phys. Rev. D , 052002 (2019).[39] M. Acero et al. (NOvA), Phys. Rev. Lett. , 151803 (2019).[40] S. MacMullin et al. , Phys. Rev. C , 064614 (2012).[41] C. Bhatia, S. W. Finch, M. E. Gooden, and W. Tornow, Phys. Rev. C , 041602 (2012).[42] N. Otuka et al. , Nucl. Data Sheets , 272 (2014).[43] D. Brown et al. , Nucl. Data Sheets , 1 (2018).[44] M. Acero et al. (NOvA), Phys. Rev. D et al. (CAPTAIN), Phys. Rev. Lett. , 042502(2019).[47] A. Friedland and S. W. Li, Phys. Rev. D , 036009 (2019).[48] R. Acciarri et al. (DUNE), (2015), arXiv:1512.06148[physics.ins-det].[49] R. Acciarri et al. (LArIAT), JINST , P04026 (2020).[50] C. Adams et al. (MicroBooNE), Eur. Phys. J. C , 673 (2019).[51] C. Adams et al. (MicroBooNE), JINST , P03022 (2020).[52] R. Acciarri et al. (ArgoNeuT), Phys. Rev. D , 052002 (2018).[53] E. Gramellini, Ph.D. thesis, Yale University (2018).[54] C. McGivern et al. (MINERvA), Phys. Rev. D , 052005(2016).[55] K. Abe et al. (T2K), Phys. Rev. D , 012010 (2017).[56] D. Measday, Physics Reports , 243 (2001).[57] H. J. Weyer, Physics Reports , 295 (1990).[58] MicroBooNE Collaboration, (2018), MICROBOONE-NOTE-1048-PUB.[59] C. Argelles et al. , (2019), arXiv:1907.08311 [hep-ph].[60] W. Altmannshofer, S. Gori, J. Martn-Albo, A. Sousa, andM. Wallbank, Phys. Rev. D , 115029 (2019).[61] P. Ballett, M. Hostert, S. Pascoli, Y. F. Perez-Gonzalez,Z. Tabrizi, and R. Zukanovich Funchal, JHEP , 119 (2019).[62] P. Ballett, S. Pascoli, and M. Ross-Lonergan, JHEP , 102(2017).[63] B. Batell, J. Berger, and A. Ismail, Phys. Rev. D , 115039(2019).[64] J. M. Berryman, A. de Gouvea, P. J. Fox, B. J. Kayser, K. J.Kelly, and J. L. Raaf, JHEP , 174 (2020).[65] E. Enrico Bertuzzo, S. Jana, P. A. Machado, and R. Z. Funchal,Physics Letters B , 210 (2019).[66] R. Acciarri et al. (ArgoNeuT), Phys. Rev. Lett. , 131801(2020).[67] R. Harnik, Z. Liu, and O. Palamara, JHEP , 170 (2019).[68] K. Abe et al. (T2K), Phys. Rev. D , 112009 (2019).[69] J. Wang, “Pulsed Neutron Source for Liquid Argon TPC Cal-ibration,” Workshop on Calibration and Reconstruction forLArTPC Detectors (2018).[70] V. Fischer et al. (ACED Collaboration), Phys. Rev. D ,103021 (2019).[71] P. Benetti et al. (WARP), Nucl. Instrum. Meth. A , 83(2007).[72] J. Crespo-Anadn (MicroBooNE), J. Phys. Conf. Ser. ,012006 (2019).[73] MicroBooNE Collaboration, (2019), MICROBOONE-NOTE-1030-PUB.[74] W. Walkowiak, Nucl. Instrum. Meth. A , 288 (2000).[75] S. Amoruso et al. , Nucl. Instrum. Meth. A , 275 (2004).[76] A. Cabrera et al. , (2019), arXiv:1908.02859 [physics.ins-det].[77] E. Oberla and H. Frisch, Nucl. Instrum. Meth. A , 19 (2016).[78] L. Alvarez-Ruso et al. (NuSTEC), Prog. Part. Nucl. Phys. ,1 (2018).[79] K. Mahn, C. Marshall, and C. Wilkinson, Ann. Rev. Nucl. Part.Sci.68