Low-Energy Excess (LEE): Detector and Nuclear Anomalies

• Low-Energy Excess (LEE) is an anomalous surplus of low-energy events observed across neutrino, cryogenic, and nuclear studies, highlighting domain-specific detection challenges.
• In neutrino experiments like MiniBooNE and MicroBooNE, LEE appears as a ~3σ excess of electron-neutrino-like events below 600 MeV, complicating flavor and topology discrimination.
• In cryogenic and semiconductor detectors, LEE manifests as a steeply rising event rate near detection thresholds that degrades sensitivity to low-mass dark matter, prompting targeted mitigation techniques.

“Low-Energy Excess” (LEE) is a context-dependent term used in several research programs to denote an anomalous surplus, enhancement, or unexplained population at comparatively low energy relative to an experiment’s nominal signal or background model. In accelerator neutrino physics, it commonly refers to the MiniBooNE excess of electron-neutrino-like or electromagnetic-like events at reconstructed energies below roughly $600\ \mathrm{MeV}$, and to the subsequent MicroBooNE program designed to test whether that excess is electron-like or photon-like (Yates, 2017). In low-threshold cryogenic, phonon, calorimetric, and semiconductor detectors, it denotes a steeply rising population of events near threshold, typically below a few hundred eV or at sub-eV to few-eV energies, that degrades sensitivity to low-mass dark matter and coherent elastic neutrino–nucleus scattering (Nordlund et al., 2024). In nuclear-structure studies, the same acronym is used for the low-energy enhancement of the magnetic-dipole $\gamma$-ray strength function in de-excitation spectra (Rodgers et al., 14 Nov 2025). The shared label therefore names a phenomenology of low-energy surplus, but not a single underlying mechanism.

1. Terminological scope and domain-specific definitions

Within short-baseline neutrino physics, the LEE is the excess reported by MiniBooNE in the Booster Neutrino Beam (BNB): an approximately $3\sigma$ excess of electron-neutrino-like events with reconstructed neutrino energy between $200$ and $600\ \mathrm{MeV}$ in the 2017 MicroBooNE status description, and more generally an anomalous excess of low-energy electromagnetic activity in the 2020 MicroBooNE proceedings (Yates, 2017). A later MiniBooNE reanalysis emphasized reconstructed quasielastic energy bins $[0.2,0.3]\ \mathrm{GeV}$, $[0.3,0.475]\ \mathrm{GeV}$, and $[0.475,1.3]\ \mathrm{GeV}$, with the largest visible discrepancy in the first bin (Giunti et al., 2019).

In cryogenic and semiconductor rare-event detectors, the LEE denotes a persistent population of events that rises steeply near threshold and cannot be explained by standard radiogenic or cosmogenic backgrounds. This definition is used for sapphire, silicon, germanium, CaWO$_4$, and related calorimetric or phonon-mediated systems, with the excess appearing below a few hundred eV in many devices and at sub-eV to few-eV energies in athermal phonon calorimeters (Mondal et al., 29 May 2026). The same terminology is also applied to low-energy electronic-recoil anomalies in liquid-xenon data, such as the XENON1T excess concentrated around $2$–$\gamma$0 and analyzed in the $\gamma$1–$\gamma$2 window (Szydagis et al., 2020).

A distinct usage appears in high-energy neutrino astronomy, where the LEE refers to an excess of events in IceCube and ANTARES relative to a hard astrophysical component, with papers discussing either a $\gamma$3–$\gamma$4 surplus or a broader $\gamma$5–$\gamma$6 excess depending on dataset and modeling choice (Chianese, 2017). In nuclear-structure theory, the term instead denotes an upbend of the de-excitation M1 $\gamma$7-ray strength function toward $\gamma$8, parameterized as $\gamma$9 in shell-model Monte Carlo calculations for actinides (Rodgers et al., 14 Nov 2025).

This multiplicity of meanings suggests that “LEE” functions primarily as a phenomenological label tied to a measurement regime rather than as a theory-specific concept.

2. The MiniBooNE and MicroBooNE low-energy excess

MiniBooNE, a mineral-oil Cherenkov detector in the BNB at Fermilab, observed an approximately $3\sigma$0 excess of electron-neutrino-like events with reconstructed neutrino energy between $3\sigma$1 and $3\sigma$2 (Yates, 2017). A later summary of the MiniBooNE observation reported that, in neutrino mode with $3\sigma$3 POT, the excess was $3\sigma$4 events, and in antineutrino mode with $3\sigma$5 POT it was $3\sigma$6 events; the excess was concentrated at reconstructed CCQE energies below roughly $3\sigma$7 and with forward-peaked lepton angles (Kamp et al., 2023). Because MiniBooNE is a Cherenkov detector, single electromagnetic showers from electrons and photons are topologically ambiguous.

MicroBooNE was built in the same beam at a similar baseline, but with liquid-argon time projection chamber technology. The 2017 status report defined the signal as contained $3\sigma$8 topologies with neutrino energy in the $3\sigma$9–$200$0 range, with lepton kinetic energy $200$1 and proton kinetic energy $200$2 (Yates, 2017). The reconstruction chain combined PMT pre-cuts, cosmic pixel tagging, region-of-interest finding, SSNet semantic segmentation, 3D vertex reconstruction, outgoing-charge clustering, and CNN-based particle identification. PMT pre-cuts were tuned to maintain more than $200$3 of neutrino events in simulation while rejecting more than $200$4 of background events in off-beam detector data. On detector data, SSNet agreed with a human expert’s manual labeling more than $200$5 of the time, and the single-particle CNN PID correctly identified $200$6 with $200$7, $200$8 with $200$9, $600\ \mathrm{MeV}$0 with $600\ \mathrm{MeV}$1, $600\ \mathrm{MeV}$2 with $600\ \mathrm{MeV}$3, and $600\ \mathrm{MeV}$4 with $600\ \mathrm{MeV}$5 (Yates, 2017).

By 2020, MicroBooNE had organized the search into complementary electron-like and photon-like channels, using $600\ \mathrm{MeV}$6 POT from the first three years out of approximately $600\ \mathrm{MeV}$7 POT collected over five years (Caratelli, 2020). The first round of LEE analyses remained blind, with $600\ \mathrm{MeV}$8 POT of unbiased open data for development. Common reconstruction elements included beam-coincident triggering from scintillation light, charge-to-light matching, 3D shower reconstruction, and calorimetric electron–photon separation via shower-start $600\ \mathrm{MeV}$9. Cosmic backgrounds were reduced by about $[0.2,0.3]\ \mathrm{GeV}$0, yielding $[0.2,0.3]\ \mathrm{GeV}$1–$[0.2,0.3]\ \mathrm{GeV}$2 expected cosmics in the final selections, while $[0.2,0.3]\ \mathrm{GeV}$3 calibration studies validated the electromagnetic energy scale to better than $[0.2,0.3]\ \mathrm{GeV}$4 (Caratelli, 2020).

MicroBooNE’s physics reach depends on what the MiniBooNE excess actually contains. One study showed that if the excess is modeled as $[0.2,0.3]\ \mathrm{GeV}$5 charged-current activity, MicroBooNE’s analyses directly constrain it, but if it is sourced instead by $[0.2,0.3]\ \mathrm{GeV}$6, then liquid-argon detectors have poor sensitivity because of suppressed $[0.2,0.3]\ \mathrm{GeV}$7–$[0.2,0.3]\ \mathrm{GeV}$8Ar cross sections at low energy and missing neutron energy in calorimetric reconstruction (Kamp et al., 2023). For the Wire-Cell analysis, the test statistic at unit signal strength was $[0.2,0.3]\ \mathrm{GeV}$9 for all $[0.3,0.475]\ \mathrm{GeV}$0, $[0.3,0.475]\ \mathrm{GeV}$1 for a $[0.3,0.475]\ \mathrm{GeV}$2 antineutrino fraction, and $[0.3,0.475]\ \mathrm{GeV}$3 for all $[0.3,0.475]\ \mathrm{GeV}$4; under the assumptions stated in that study, the all-$[0.3,0.475]\ \mathrm{GeV}$5 scenario remained consistent at the $[0.3,0.475]\ \mathrm{GeV}$6 confidence level (Kamp et al., 2023).

A separate MiniBooNE reanalysis argued that enhanced single-$[0.3,0.475]\ \mathrm{GeV}$7 backgrounds could explain part of the excess. Using a revised $[0.3,0.475]\ \mathrm{GeV}$8 escape probability in carbon and adding coherent photon emission, incoherent higher-resonance production, and non-resonant nucleon production, the single-$[0.3,0.475]\ \mathrm{GeV}$9 background was increased by factors of $[0.475,1.3]\ \mathrm{GeV}$0–$[0.475,1.3]\ \mathrm{GeV}$1 across energy bins and beam modes. In that treatment, the MiniBooNE-only oscillation significance decreased from $[0.475,1.3]\ \mathrm{GeV}$2 to $[0.475,1.3]\ \mathrm{GeV}$3 (Giunti et al., 2019). This does not remove the anomaly, but it narrows the range of interpretations that any successor experiment must distinguish.

3. Low-threshold cryogenic and semiconductor detectors

In cryogenic calorimeters and semiconductor detectors, the LEE is a threshold-proximate background population rather than an appearance-like event excess. A broad review-level simulation study described it as a population of apparent energy-release events below a few hundred eV, reported across cryogenic calorimeters or phonon detectors such as CRESST, EDELWEISS, and SuperCDMS, as well as silicon and germanium CCD or Skipper-CCD ionization detectors such as DAMIC and SENSEI (Nordlund et al., 2024). The excess grows toward very low energies, masks rare-event signals, and persists at cryogenic temperatures.

Several experiments report material- and device-specific manifestations. In the MINER sapphire detector, the LEE was strongest near $[0.475,1.3]\ \mathrm{GeV}$4, with simulation/data mismatch confined to $[0.475,1.3]\ \mathrm{GeV}$5–$[0.475,1.3]\ \mathrm{GeV}$6 and good agreement above $[0.475,1.3]\ \mathrm{GeV}$7; after each non-operational warm-up to $[0.475,1.3]\ \mathrm{GeV}$8 from a base of $[0.475,1.3]\ \mathrm{GeV}$9, the event rate in $_4$0–$_4$1 increased sharply and then decayed over time (Mondal et al., 29 May 2026). In NUCLEUS, the shared LEE in an Al$_4$2O$_4$3 double-TES calorimeter was defined through

$_4$4

and the time evolution across datasets was best described by a common power law

$_4$5

with $_4$6 fixed at the moment the detector reached $_4$7 and $_4$8 (Abele et al., 8 Mar 2026).

A two-channel low-threshold silicon calorimeter resolved the LEE into two populations: “shared” multichannel events with a pulse shape consistent with substrate athermal phonon events, and “singles” sub-eV events coupling nearly exclusively to a single channel with a significantly faster pulse shape (Anthony-Petersen et al., 2024). The same system measured a world-leading baseline phonon energy resolution $_4$9, but also found excess correlated and uncorrelated noise linked to below-threshold analogs of the two LEE populations. A later silicon study with $2$0 and $2$1 substrates reported that both correlated shot noise and shared LEE relaxed with time since cooldown and scaled linearly with substrate thickness; the $2$2 device exhibited approximately $2$3 the correlated noise and approximately $2$4 the shared LEE rate of the $2$5 device (Chang et al., 22 May 2025).

In CRESST, the LEE appears as a featureless, steeply rising excess below about $2$6 in the phonon channel (Angloher et al., 2024). Above-ground doubleTES prototypes demonstrated that this excess contains at least two components: a sensor- or TES-proximal population appearing in only one TES, and a bulk-like component seen equally in both TESs. In a silicon-on-sapphire module, the absorber-band LEE showed a fast exponential decay with time constant $2$7 days for $2$8–$2$9 events, compatible with the underground fast component previously reported by CRESST (Angloher et al., 2024).

SuperCDMS-HVeV provided another mechanism-specific case. A $\gamma$00 silicon detector operated at $\gamma$01, $\gamma$02, and $\gamma$03 showed an excess at tens of eV in the zero-bias data. Cross-voltage comparisons, timing structure, and response-matrix studies indicated that the dominant contribution was consistent with photon-induced events produced by luminescence of the printed circuit boards used in the detector holder; the converted high-voltage spectra best matched the measured $\gamma$04 spectrum for an average pair-creation energy $\gamma$05–$\gamma$06 (Collaboration et al., 2022).

These observations establish that, in low-threshold detectors, the LEE is not a single empirical object. It includes bulk-like and sensor-localized components, above-threshold events and sub-threshold shot noise, and both time-dependent and cooldown-history-dependent behavior.

4. Proposed microscopic origins and mitigation strategies

Several papers advance explicit microscopic or mesoscopic origin models. For semiconductor detectors, one atomistic study proposed long-term recombination of radiation-induced complex defect pockets as the source of the excess (Nordlund et al., 2024). In that picture, keV-scale recoil cascades create nanometric disordered regions storing configurational energy; low-barrier local rearrangements with $\gamma$07 can trigger avalanche-like relaxation that releases much larger energy. The predicted energy spectrum follows

$\gamma$08

with $\gamma$09–$\gamma$10 across temperatures from $\gamma$11 to $\gamma$12. A representative quantum-thermal-bath result gave $\gamma$13, in agreement with an experimental SuperCDMS Si low-energy tail $\gamma$14 (Nordlund et al., 2024).

For low-threshold calorimeters with aluminum structures, another proposal attributes the LEE to relaxation of stressed aluminum films via dislocation motion (Romani, 2024). In that model, cooldown induces a biaxial strain $\gamma$15 and stress $\gamma$16 in Al films, dislocations pin in metastable locks, and quantum tunneling mediates depinning at mK temperatures. For a broad ensemble of pinned sites, the total depinning rate becomes

$\gamma$17

producing a slow $\gamma$18 decay after cooldown. One week after cooldown, the model predicts rates of $\gamma$19 for $\gamma$20 and $\gamma$21–$\gamma$22 for $\gamma$23–$\gamma$24 (Romani, 2024).

A related but interface-focused mechanism invokes relative thermal-contraction mismatch between the absorber and the amorphous SiO$\gamma$25 layer underneath TESs (Zema et al., 28 May 2026). In CRESST-type modules, the mismatch strain is written as

$\gamma$26

and the dislocation nucleation rate follows

$\gamma$27

Cooling from $\gamma$28 to $\gamma$29 was estimated to release $\gamma$30 for c-axis orientation and $\gamma$31 for a-axis orientation in a $\gamma$32-scale CaWO$\gamma$33 absorber, values described as being in the same ballpark as the integrated LEE energy measured after a $\gamma$34 warm-up (Zema et al., 28 May 2026).

Experimental mitigation proposals reflect these mechanistic differences. NUCLEUS found that slower cooldowns from room temperature to $\gamma$35 reduced the initial LEE amplitude by up to an order of magnitude while preserving a common power-law decay with $\gamma$36 (Abele et al., 8 Mar 2026). MINER used a CVAE-guided pulse-shape analysis and then a rise-time cut, with the best wavelet filter threshold $\gamma$37 yielding signal acceptance $\gamma$38 and LEE rejection $\gamma$39 (Mondal et al., 29 May 2026). In the double-channel silicon calorimeter, mitigation targets include reducing aluminum film stress, adjusting fin geometry, improving TES uniformity, and extending to more channels for localization and rejection of sensor-film events (Anthony-Petersen et al., 2024).

A plausible implication is that the LEE problem in cryogenic instrumentation is not exhausted by one mechanism. Bulk defect recombination, interfacial thermoelastic stress, aluminum relaxation, and holder-material luminescence are all supported in specific datasets, and the dominant contribution appears experiment-dependent.

5. Related low-energy excesses in liquid xenon and astrophysical neutrinos

The XENON1T low-energy electronic-recoil excess is conceptually different from the cryogenic-detector LEE but belongs to the same broader class of low-energy anomalies. XENON1T reported a $\gamma$40 excess of electronic recoils with the largest deviation around $\gamma$41–$\gamma$42 in a $\gamma$43–$\gamma$44 analysis window (Szydagis et al., 2020). A NEST-based reanalysis argued that this feature can be reproduced by adding $\gamma$45 $\gamma$46Ar decays over the $\gamma$47 tonne-year exposure. In that model, the relevant lines are the $\gamma$48 K-shell and $\gamma$49 L-shell de-excitations, and the line response is non-Gaussian and positively skewed. Including $\gamma$50Ar reduced the discrepancy to $\gamma$51 in non-PLR tests (Szydagis et al., 2020).

In IceCube and ANTARES analyses, the phrase “low-energy excess” denotes a surplus in astrophysical neutrino data relative to a hard power-law component. One study focused on a $\gamma$52–$\gamma$53 excess motivated by the six-year up-going muon-neutrino sample, where the high-energy fit above $\gamma$54 gave $\gamma$55, in $\gamma$56 tension with earlier softer diffuse-flux fits (Chianese, 2017). Another paper treated the LEE as a $\gamma$57–$\gamma$58 excess seen independently by IceCube and ANTARES and tested bright and choked gamma-ray bursts as candidate sources. Under its unified GRB model, the best fits were poor: for bright plus choked jets with internal-shock acceleration, the best-fit point was $\gamma$59 with $\gamma$60; choked-only scenarios gave $\gamma$61 or $\gamma$62 depending on the acceleration prescription (Denton et al., 2018).

These examples show that “low-energy excess” can also signify an inconsistency internal to a fitted astrophysical spectrum, without any threshold artifact or detector-noise connotation.

6. Low-energy enhancement in nuclear $\gamma$63-ray strength functions

In nuclear structure, the term refers not to a detector anomaly but to an intrinsic feature of the de-excitation strength function. Shell-model Monte Carlo calculations for six actinides found a pronounced low-energy enhancement in the M1 $\gamma$64-ray strength function, visible as a strong peak at $\gamma$65 in $\gamma$66 and as an exponential increase of $\gamma$67 as $\gamma$68 decreases (Rodgers et al., 14 Nov 2025). The low-energy part was parameterized as

$\gamma$69

For the actinides studied, the extracted slopes and normalizations were of similar scale but with moderate isotope dependence. For example, $\gamma$70Th gave $\gamma$71 and $\gamma$72, while $\gamma$73Pu gave $\gamma$74 and $\gamma$75 (Rodgers et al., 14 Nov 2025). The slope $\gamma$76 was found to be independent of the initial excitation energy within uncertainties, and the calculations simultaneously displayed a scissors resonance near $\gamma$77–$\gamma$78 and a spin-flip mode near $\gamma$79.

The formalism is defined through the thermal M1 strength

$\gamma$80

with the de-excitation $\gamma$81-ray strength obtained from the finite-temperature response through the relations given in that work (Rodgers et al., 14 Nov 2025). In this subfield, the “LEE” is therefore a microscopic spectral enhancement of nuclear M1 strength, not an unexplained background.

7. Comparative perspective and unresolved issues

Across all usages, the LEE denotes a low-energy surplus relative to an expected baseline, but the causal structure differs sharply. In the MiniBooNE/MicroBooNE program, the central issue is flavor and topology identification in a beam experiment, particularly electron versus photon discrimination and the possibility of alternative $\gamma$82, $\gamma$83, or single-$\gamma$84 interpretations (Caratelli, 2020). In cryogenic and semiconductor detectors, the main problem is an instrumental or materials background that intrudes directly into the signal window for low-mass dark matter or CE$\gamma$85NS searches (Nordlund et al., 2024). In nuclear $\gamma$86-strength studies, the effect is a genuine property of the many-body de-excitation spectrum (Rodgers et al., 14 Nov 2025).

Several misconceptions are therefore ruled out by the literature. The LEE is not universally a neutrino anomaly, nor universally a detector artifact, nor universally evidence for new physics. In some settings, known-physics reinterpretations substantially reduce the anomaly, as with enhanced MiniBooNE single-$\gamma$87 backgrounds (Giunti et al., 2019) or $\gamma$88Ar in XENON1T (Szydagis et al., 2020). In other settings, the excess is robustly detector-related but still mechanistically unsettled, with competing support for bulk-defect, interfacial, film-relaxation, and luminescence mechanisms (Romani, 2024).

The most stable cross-cutting lesson is methodological. LEE studies have increasingly relied on sidebands, control samples, time dependence, topology splitting, dual-channel readout, calibration lines, and data-driven anomaly characterization rather than on a single global fit. This suggests that, although the acronym is shared, progress on any specific LEE requires domain-specific observables: $\gamma$89 and shower topology in liquid argon, pulse-shape and cooldown-history diagnostics in cryogenic calorimeters, line-shape modeling in liquid xenon, and finite-temperature many-body response in nuclear structure.