Abstract: We present a novel methodology to search for intranuclear neutron-antineutron transition ($n\rightarrow\bar{n}$) followed by $\bar{n}$-nucleon annihilation within an $^{40}$Ar nucleus, using the MicroBooNE liquid argon time projection chamber (LArTPC) detector. A discovery of $n\rightarrow\bar{n}$ transition or a new best limit on the lifetime of this process would either constitute physics beyond the Standard Model or greatly constrain theories of baryogenesis, respectively. The approach presented in this paper makes use of deep learning methods to select $n\rightarrow\bar{n}$ events based on their unique features and differentiate them from cosmogenic backgrounds. The achieved signal and background efficiencies are (70.22$\pm$6.04)% and (0.0020$\pm$0.0003)%, respectively. A demonstration of a search is performed with a data set corresponding to an exposure of $3.32 \times10^{26}\,$neutron-years, and where the background rate is constrained through direct measurement, assuming the presence of a negligible signal. With this approach, no excess of events over the background prediction is observed, setting a demonstrative lower bound on the $n\rightarrow\bar{n}$ lifetime in $^{40}$Ar of $\tau_{\textrm{m}} \gtrsim 1.1\times10^{26}\,$years, and on the free $n\rightarrow\bar{n}$ transition time of $\tau_{\textrm{n-\bar{n}}} \gtrsim 2.6\times10^{5}\,$s, each at the $90\%$ confidence level. This analysis represents a first-ever proof-of-principle demonstration of the ability to search for this rare process in LArTPCs with high efficiency and low background.
Introduction: The Matter-Antimatter Mystery
One of the most profound questions in modern physics is the origin of the "Matter-Antimatter Asymmetry"—the observation that the universe is composed almost entirely of matter, despite the Big Bang's expectation of equal production for both. Explaining this requires processes that violate "Baryon Number Conservation."
The spontaneous transition of a neutron to an antineutron ($n \rightarrow \bar{n}$) is a predicted process that would violate this conservation by two units ($\|\Delta B\| = 2$). If this occurs within an $^{40}$Ar nucleus, the resulting antineutron immediately annihilates with a neighboring nucleon, releasing a characteristic "star-like" burst of pions.
The $n \rightarrow \bar{n}$ Signal Signature
The primary topological "tell" for an $n \rightarrow \bar{n}$ event is its spherical, multi-pronged shape, which contrasts sharply with the long, straight tracks left by cosmic muons. This signature is the result of various annihilation channels, each producing a specific combination of pions.
As shown in the branching ratios above, the annihilation of an antineutron with a proton or neutron typically results in multiple charged and neutral pions. Modeling these specific final states is critical for training the deep learning algorithms used to isolate signal events from the "blizzard" of cosmic ray backgrounds encountered at surface-level detectors like MicroBooNE.
Methodology: Deep Learning in LArTPCs
The MicroBooNE collaboration developed a multi-stage selection pipeline to identify these rare events:
- Topological Filtering: Initial selection used Boosted Decision Trees (BDT) based on geometric features like the "extent" of particle clusters. This stage rejected 91% of the background while retaining 86% of the signal.
- Sparse CNN (VGG16): The core of the analysis utilized a Convolutional Neural Network. Because LArTPC data is mostly empty space, a sparse CNN architecture was employed. This allowed the network to focus exclusively on pixels containing data, drastically reducing memory requirements and enabling high-resolution 2D projections of 3D clusters.
- Data-Driven Background Characterization: To ensure the model could handle real-world detector noise, the team used "off-beam" data (collected when the neutrino beam was inactive) to characterize cosmic ray backgrounds, rather than relying solely on simulations.
Experimental Results
The analysis achieved a signal selection efficiency of (70.22 ± 6.04)%, a significant improvement over the <10% efficiency typical of older water Cherenkov technologies. Simultaneously, the background efficiency was suppressed to (0.0020 ± 0.0003)%, representing a rejection rate of nearly 99.998%.
Upon unblinding the data, 2 events were observed, consistent with the expected background of 3.2 events. This result allowed the team to set a demonstrative lower bound on the $n \rightarrow \bar{n}$ lifetime in $^{40}$Ar of $\tau_m \gtrsim 1.1 \times 10^{26}$ years and a free neutron transition time of $\tau_{n-\bar{n}} \gtrsim 2.6 \times 10^5$ seconds at the 90% confidence level.
Conclusion and Future Prospects
While MicroBooNE's exposure was limited by its size and surface location, this work serves as a vital proof-of-principle for the Deep Underground Neutrino Experiment (DUNE). DUNE will feature massive kilotonne-scale LArTPCs located deep underground, shielded from cosmic rays. By applying these deep learning techniques, DUNE's sensitivity to baryon number violation could increase seven-fold over current projections, potentially providing the key to understanding the matter-dominated universe.
Summary updated