Inclusive Search for Anomalous Single-Photon Production in MicroBooNE

The MicroBooNE experiment conducted an inclusive search for anomalous single-photon production to investigate the MiniBooNE low-energy excess, finding overall agreement with predictions but observing a localized excess of approximately 2σ significance in events with no visible protons and shower energies below 600 MeV.

The Gist

Imagine a giant, ultra-sensitive underwater camera sitting deep underground, waiting to catch tiny flashes of light from invisible particles called neutrinos. This is the MicroBooNE experiment, a massive tank of liquid argon (frozen neon-like gas) that acts like a 3D movie camera for subatomic particles.

The story of this paper begins with a mystery from a neighbor. A previous experiment, MiniBooNE, which sat just down the road on the same particle beam, kept seeing a strange "glitch." It detected more flashes of light (electromagnetic showers) at low energies than physics textbooks predicted. Scientists called this the "Low Energy Excess" (LEE).

The big question was: What was causing these extra flashes?
Was it a new type of particle (like a "sterile neutrino")? Or was it just a standard particle, like a photon (a particle of light), that the detectors were misidentifying? MiniBooNE's camera was a bit blurry; it couldn't tell the difference between a flash caused by an electron and a flash caused by a single photon.

The MicroBooNE Mission: The High-Definition Detective
MicroBooNE decided to solve this mystery with a camera that has much higher resolution. Because it uses liquid argon, it can see the very beginning of a particle's path.

• The Electron vs. Photon Test: When an electron starts a flash, it leaves a thick, fuzzy trail immediately. When a photon starts a flash, it travels a tiny bit before turning into an electron, leaving a small gap. MicroBooNE can see this gap.
• The Goal: The team wanted to count only the events that looked like single photons (the "photon-like" events) to see if the "Low Energy Excess" was actually just a bunch of photons that MiniBooNE couldn't distinguish.

How They Searched: The "Blind" Hunt
To avoid bias, the scientists played a game of "Blind Man's Bluff."

1. The Setup: They built a complex filter (using computer programs called "Boosted Decision Trees") to sort through millions of particle collisions. They wanted to find events with exactly one photon shower and no other messy debris.
2. The Blindfold: They locked the data in the "signal region" (the area where the mystery events would be) so no one could look at it until the rules of the game were perfectly set.
3. The Calibration: Before looking at the mystery box, they checked their "side pockets" (sidebands). These were areas where they knew what should happen (like collisions with muons or pions). They used these known areas to tune their predictions, ensuring their "map" of what to expect was accurate.

The Results: A Slight Hint, But No Smoking Gun
When they finally lifted the blindfold and looked at the data:

• The Big Picture: In the full range of energies, the data matched the predictions almost perfectly. The "Low Energy Excess" didn't show up as a massive, obvious spike of single photons. The overall "goodness of fit" was good (a p-value of 0.11), meaning the standard physics model still held up well.
• The Subtle Clue: However, when they zoomed in on a specific, tricky subset of events—those with no visible protons (tiny particles that usually fly out of the collision) and low energy (under 600 MeV)—they found something interesting.
  • They saw 93 events in the data.
  • They expected only about 60 events based on their calculations.
  • This is a 2.2 sigma difference. In the world of particle physics, this is like hearing a faint whisper in a noisy room. It's noticeable, but not loud enough to shout "Eureka!" (which usually requires a 5-sigma shout).

What Does This Mean?
The paper concludes that while there is a small, intriguing bump in the data for low-energy, single-photon events with no protons, it is not a definitive discovery of new physics yet.

• The "excess" could be caused by standard physics processes that are slightly harder to model than expected (like photons coming from outside the main detection area or from specific types of particle decays).
• The team tried to see if this excess matched the "photon-only" version of the MiniBooNE mystery, but the numbers didn't line up perfectly.

The Bottom Line
MicroBooNE acted like a high-definition detective, clearing up the blurry picture from its neighbor. It found that the "Low Energy Excess" is not simply a flood of misidentified single photons. While there is a small, curious bump in the data that warrants further investigation, the paper does not claim to have found a new particle or a new law of physics. For now, the mystery remains unsolved, but the MicroBooNE camera has narrowed down the list of suspects significantly.

Technical Summary

Technical Summary: Inclusive Search for Anomalous Single-Photon Production in MicroBooNE

Problem Statement
This paper addresses the "Low Energy Excess" (LEE), a long-standing anomaly in neutrino physics reported by the MiniBooNE experiment. MiniBooNE observed an excess of electromagnetic shower events at low energies that cannot be explained by standard neutrino oscillation models. While MiniBooNE's Cherenkov detector cannot distinguish between electron-induced showers (potential $\nu_e$ appearance) and single-photon-induced showers, the MicroBooNE experiment, utilizing a Liquid Argon Time Projection Chamber (LArTPC), possesses the capability to differentiate these topologies based on ionization energy deposition ($dE/dx$) at the shower start and photon conversion distance. Previous MicroBooNE analyses focused on specific Standard Model (SM) processes or "electron-like" topologies; this work presents an inclusive search for anomalous single-photon production to determine if the LEE can be attributed to a broad class of photon-like events.

Methodology
The analysis utilizes a dataset corresponding to $6.34 \times 10^{20}$ protons on target (POT) from Fermilab's Booster Neutrino Beam (BNB), collected between February 2016 and July 2018. The study employs a blind analysis strategy, keeping the signal region data unblinded until the selection criteria and background models were finalized.

• Signal Definition: The signal is defined as any final state containing one "reconstructable" photon-initiated electromagnetic shower (starting $\ge 3$ cm from the TPC boundary with true energy $> 20$ MeV) and any number of charged particles below the MiniBooNE Cherenkov threshold. This inclusive definition encompasses five specific SM processes modeled in the GENIE neutrino event generator (v3.0.6):

  1. Neutral Current (NC) $\pi^0$ events with exactly one reconstructable photon ($NC\ \pi^0\ 1\gamma$).
  2. NC $\Delta$ radiative decay ($NC\ \Delta\ 1\gamma$).
  3. NC processes producing a single photon from other resonances (e.g., $\eta$, $\rho$) ($NC\ \text{Other}\ 1\gamma$).
  4. $\nu_\mu$ Charged Current (CC) interactions where the muon kinetic energy is $< 100$ MeV ($\nu_\mu\ CC\ 1\gamma$).
  5. Interactions occurring outside the fiducial volume (FV) but producing a photon entering the FV ($\text{out of FV}\ 1\gamma$).

• Reconstruction and Selection: The Wire-Cell framework is used for 3D imaging and event reconstruction. A multi-stage selection process is applied:

  1. Pre-selection: Removal of cosmic-ray backgrounds ($>99.99\%$ rejection) and requirement of a reconstructed vertex $>5$ cm from the TPC boundary.
  2. BDT Classification: Four Boosted Decision Trees (XGBoost) target specific backgrounds: $\nu_\mu$ CC, NC $\pi^0$, "other" backgrounds (cosmic/outside FV), and $\nu_e$ CC.
  3. Final Cut: Requirement of exactly one reconstructed electromagnetic shower.

• Background Constraints: To reduce systematic uncertainties, the analysis uses a data-driven approach. Sideband samples enriched in $\nu_\mu$ CC and NC $\pi^0$ interactions are used to constrain the central values and systematic uncertainties of the signal region predictions via a conditional constraint formalism. Systematic uncertainties (flux, cross-section, detector response, MC statistics, and "Dirt" interactions) are incorporated through a covariance matrix.

Key Results

• Inclusive Sample: After applying all selection criteria, 678 data events were observed in the signal region. The constrained Standard Model prediction was $564 \pm 24\ (\text{stat.}) \pm 51\ (\text{syst.})$ events. This corresponds to a global excess of approximately 20%.
• Goodness-of-Fit: Across the full reconstructed shower energy range (0–1500 MeV), the data and prediction are consistent, yielding a goodness-of-fit $p$-value of 0.11 ($\chi^2/\text{n.d.f.} = 23/16$).
• Sub-sample Analysis (Zero Protons, Low Energy): A specific region of interest (ROI) was defined containing events with no reconstructed protons (kinetic energy $> 35$ MeV) and shower energy $\le 600$ MeV.
  • In this ROI, the data shows an excess of $93 \pm 22\ (\text{stat.}) \pm 35\ (\text{syst.})$ events above the prediction.
  • This corresponds to a local significance of $2.2\sigma$.
  • The excess is concentrated at shower energies below 600 MeV.
• Model Compatibility:
  • The shape of the excess is not compatible with the $NC\ \Delta\ 1\gamma$ hypothesis, which peaks at higher energies.
  • The shape is potentially compatible with mismodeled $NC\ \pi^0\ 1\gamma$ (in-FV) or "out of FV" events; however, the scaling factors required to fit the excess would push these background predictions significantly beyond their uncertainties in the constraining sidebands.
  • A comparison with a "LEE-$\gamma$" model (derived from unfolding MiniBooNE's excess under the assumption of single-photon production) shows the observed excess (93 events) is larger than the model's scaled prediction (33 events).

Significance and Claims
The paper reports the first inclusive search for anomalous single-photon production in MicroBooNE. While the full dataset shows consistency with the Standard Model ($p=0.11$), a mild excess ($2.2\sigma$) is observed in the specific phase space of low-energy showers ($<600$ MeV) with no visible protons.

The authors conclude that they have not identified a simple and complete explanation for this observed excess. While the shape of the excess could theoretically arise from mismodeled Standard Model processes (specifically $NC\ \pi^0$ or out-of-FV interactions), the magnitude of the required corrections is inconsistent with the constraints provided by the sideband data. Consequently, the result does not confirm the MiniBooNE LEE as a single-photon phenomenon but motivates further investigation using the full MicroBooNE dataset and the broader Short-Baseline Neutrino (SBN) program. The paper emphasizes that the LEE remains an open question, requiring continued scrutiny of both electron-like and photon-like topologies.