MicroBooNE Measures MeV Energy Resolution in Liquid Argon

Low-energy reconstruction is a limiting detector capability for supernova-neutrino searches and other MeV-scale liquid-argon measurements. Yet LArTPC calibration is usually constrained by higher-energy tracks and showers, while isolated MeV deposits sit near reconstruction thresholds and amid radiological and cosmic backgrounds. The MicroBooNE collaboration measures that regime directly, using ambient $^{208}$Tl decays to create a recognizable internal calibration feature rather than relying only on simulation.

Core Contribution

The paper identifies a rare, nearly monoenergetic signature from pair production by 2.614 MeV gamma rays emitted in $^{208}$Tl decays near MicroBooNE's G10 detector struts. The reconstructed object is a small charge deposition, or blip, associated with the pair-production topology. The central measurement is the fractional width $p_2/p_1$ of a fitted Gaussian peak: $(7.52 \pm 0.78\,\mathrm{(stat)} \pm 0.92\,\mathrm{(syst)})\%$ in data at a fitted energy of $(1.54 \pm 0.01\,\mathrm{(stat)} \pm 0.02\,\mathrm{(syst)})$ MeV.

What is new is the use of an in-detector radiological line and its annihilation-photon topology as a MeV-scale LArTPC calibration handle. The authors describe this as the first direct characterization of MeV-scale energy resolution in a neutrino LArTPC. It is not an absolute energy-scale calibration at the 2.614 MeV gamma energy: the selected pair-production blip has a lower reconstructed energy, and the paper explicitly attributes the offset to detector response and the assumed linear charge-to-energy conversion.

Technical Approach

A gamma ray that undergoes pair production creates an electron-positron pair; the positron annihilates, producing two back-to-back photons that may Compton scatter or be photoabsorbed. MicroBooNE selects blips around known radiological hot spots and searches for the geometrical alignment between the annihilation point and deposits from those photons. This topology is intended to reject ordinary Compton-scattering backgrounds while retaining a calibration feature tied to the $^{208}$Tl gamma line.

The analysis applies cuts (1)--(11), then noise-reduction and background-reduction requirements. In simulation, the selected sample has a 4.9:1 ratio of pair-production blips to other processes, but only 2.6% selection efficiency. The low efficiency follows partly from the requirement that at least one collection-plane cluster be reconstructed for each annihilation gamma, as well as threshold losses and the chance that an annihilation gamma exits the active volume. This is a deliberately purity-first selection, appropriate for resolving a peak but not for collecting a large calibration sample.

Figure 8 makes the filtering visible: selection cuts alone leave only a suggestion of structure between 1 and 2 MeV, whereas the full selection exposes a peak in the highlighted 1.3--1.8 MeV interval. The result depends on a background estimate drawn from a control region and on fitting the background-subtracted distribution with a Gaussian plus a linear term.

Results and Analysis

The beam-off data set contains 653,367 events and 39,760,791 reconstructed blips. After selection, the authors identify 640 signal-blip candidates above a similarly sized background contribution. Fitting the 1.3--1.8 MeV region yields a data peak mean of 1.54 MeV and Gaussian width of 0.12 MeV, producing the reported 7.52% statistical resolution. The systematic uncertainty is 0.92 percentage points, formed from the largest effects in three categories: changing the fit range shifts the resolution by as much as 0.27%, replacing the linear background with an exponential shifts it by 0.26%, and varying the relative background magnitude contributes up to 0.60%.

The relevant comparison is a matched Monte Carlo sample. After all cuts and background subtraction, simulation gives $(9.70 \pm 0.65)\%$ resolution at a fitted peak mean of $(1.439 \pm 0.008)$ MeV. The data--simulation difference is $(2.2 \pm 1.4)\%$, which the authors characterize as agreement within $1.6\sigma$. Other simulated selections range from 8.75% when restricted to true pair-production requirements to 9.86% for all cuts, suggesting that energy-lowering processes are not the dominant contribution to the observed width.

The evidence supports a useful detector-performance measurement, not a general resolution model. It is based on real detector data, a control-region subtraction, and a simulation comparison, but only at one reconstructed energy and with a small selected sample. Its practical value is as a template for future LArTPCs with suitable internal radioactivity and similarly low reconstruction thresholds; extending it to sub-percent energy-scale work will require higher statistics and an explicit study of position dependence and nonlinearity.