First search for dark-trident processes using the MicroBooNE detector

The Big Picture

Imagine shining a flashlight into a pitch-black room and looking not for light reflecting off objects, but for tiny flickers that only appear at the very edge of your beam. That’s roughly what physicists at Fermilab’s MicroBooNE experiment just attempted. They hunted for dark matter not by catching it directly, but by watching for the brief, distinctive trace it leaves as it passes through a tank of ultra-cold liquid argon.

Dark matter makes up roughly 27% of everything in the universe. It shapes galaxies, bends light, and holds together the vast cosmic web of large-scale structure. Yet it has never once been directly detected in a laboratory. Decades of searching with increasingly sensitive detectors have yielded only silence.

A theoretical idea called the dark-trident process offers a different approach. Rather than wait for dark matter to crash into a nucleus head-on, scientists can look for it briefly emitting a burst of ordinary particles as it passes through a detector. The MicroBooNE collaboration has now conducted the world’s first experimental search for this process, using a neural network to sift through hundreds of thousands of particle collisions for an extremely faint signal.

Key Insight: The dark-trident search is a new strategy for hunting light dark matter. It exploits the distinctive signature of a fleeting dark-photon emission rather than a direct nuclear collision, reaching parameter space no previous experiment could probe.

How It Works

The experiment unfolds in two stages, separated by hundreds of meters of rock and steel.

Stage one: dark matter production. At Fermilab’s Main Injector, protons slam into a fixed target, generating a flood of secondary particles including neutral pions (π⁰) and eta mesons (η). In standard physics, π⁰ mesons decay almost instantly into two photons. But in certain dark-sector models, a small fraction could instead decay as π⁰ → γ χ χ̄, producing a pair of dark-matter particles χ through a virtual dark photon A’. Think of A’ as a ghostly cousin of the ordinary photon, one that couples to a hidden sector. These dark-matter particles, with energies in the 0.1–3 GeV range, stream invisibly through hundreds of meters of material straight toward the detector.

Stage two: the trident interaction. When a χ particle reaches MicroBooNE, it occasionally undergoes the dark-trident process: χ + Ar → χ + Ar + A’. The dark-matter particle radiates a real dark photon A’ while bouncing off an argon nucleus. This is analogous to bremsstrahlung (German for “braking radiation”), where a charged particle emits a photon as it decelerates. The A’ then promptly decays into an electron-positron pair (e⁺e⁻) with essentially 100% probability, since no other decay routes are open at these energies. The dark-matter particle itself exits unseen. What’s left behind is an apparently sourceless pair of electromagnetic tracks materializing from nowhere in the argon.

The search chain involved:

• Simulating signal events for two benchmark scenarios (mass ratios Mχ/MA’ = 0.6 and 2, representing cases where the dark matter is lighter or heavier than the dark photon it emits) across dark-photon masses from 10 to 400 MeV
• Identifying backgrounds dominated by ordinary neutrino interactions that can mimic the e⁺e⁻ signature
• Training a convolutional neural network (CNN) to distinguish dark-trident events from background, using the LArTPC’s inherently image-like 2D wire readout

The liquid-argon time projection chamber, or LArTPC, is well suited to this search. MicroBooNE’s 85-tonne active volume records particle tracks with millimeter-scale precision, and the resulting 2D images of particle interactions are natural input for CNNs.

The network learned to recognize the opening angle (the angle at which the electron and positron fly apart) and energy distributions of e⁺e⁻ pairs from dark-photon decays. These distributions shift with the mass ratio Mχ/MA’: the heavier the dark photon relative to the dark-matter particle, the wider the pair tends to spread. The full dataset contained 7.2 × 10²⁰ protons on target (POT), the standard measure of total beam exposure accumulated during MicroBooNE’s operation.

Why It Matters

No dark-trident signal appeared in the data. A null result here still carries real scientific weight, though: it tells us where dark matter is not, ruling out combinations of properties that were genuinely open before this measurement. The collaboration set 90% confidence level upper limits on ε², the squared kinematic mixing parameter that controls how strongly the dark photon couples to ordinary electromagnetism, as a function of dark-photon mass across 10–400 MeV. These limits cover previously unconstrained regions for dark fine-structure constants αD between 0.1 and 1.

This is also a proof of concept. The dark-trident channel can be pursued with far larger exposures at the DUNE near detector and Fermilab’s Short-Baseline Neutrino (SBN) program detectors, both of which will collect orders of magnitude more protons on target. The CNN-based analysis technique is now demonstrated and can be ported to those future experiments without starting from scratch.

Bottom Line: MicroBooNE ran the world’s first search for the dark-trident process, setting new limits on light dark matter in parameter space no previous experiment could reach. The neural-network-powered technique is ready to scale with the next generation of neutrino detectors.