---
title: 'Trinity Demonstrator: Neutrino Pathfinder'
url: https://www.emergentmind.com/topics/trinity-demonstrator
type: topic
---

# Trinity Demonstrator: Neutrino Pathfinder

The Trinity Demonstrator is the first, proof-of-concept stage of the planned Trinity PeV Neutrino Observatory: a remotely operated, one-square-meter-class imaging atmospheric Cherenkov telescope deployed on Frisco Peak, Utah, to validate the Earth-skimming tau-neutrino detection concept, characterize backgrounds, and provide a technical bridge to Trinity One and the eventual observatory. In historical Los Alamos literature, the phrase “Trinity demonstrator” also appears for the Trinity high-explosive implosion system; in contemporary astroparticle physics, however, it denotes the neutrino pathfinder instrument [2503.11864, 2509.18236, 2103.05714].

## 1. Programmatic role and observatory context

The Demonstrator belongs to a broader Trinity program aimed at detecting astrophysical neutrinos above PeV energies by imaging the Cherenkov light from air showers initiated by tau decays in the atmosphere after Earth-skimming \(\tau\)-neutrino interactions. The earlier Trinity concept paper described a proposed ground-based air-shower imaging instrument optimized for earth-skimming ultrahigh-energy tau neutrinos with energies between \(10^7\) GeV and \(10^{10}\) GeV; later papers recast that concept as a three-stage development program consisting of the Trinity Demonstrator, Trinity One, and the full Trinity Neutrino Observatory [1907.08727, 2503.11864, 2509.18236].

Within that staged architecture, the Demonstrator is explicitly a pathfinder rather than the final science instrument. Stage one is the single Demonstrator used to prove that the concept works and to understand backgrounds. Stage two is Trinity One, the first complete Trinity telescope that can rotate in azimuth. Stage three is the full observatory, envisioned as an array of 18 wide-angle Cherenkov telescopes distributed across at least three mountain-top sites. This organizational role is central to the instrument’s interpretation: its purpose is to validate detection geometry, remote operation, triggering, calibration, and background rejection under field conditions, not to deliver a large neutrino event sample on its own [2509.18236, 2509.18237].

The Demonstrator was deployed in Fall 2023, with first light on October 3, 2023, and commissioning completed in June 2024; a companion status paper describes commissioning as ending in Summer 2024. The sequencing is significant because Trinity One is presented as following directly from the Demonstrator’s operational results and background studies [2503.11864, 2509.18236, 2509.18237].

## 2. Detection principle and horizon-pointing geometry

The detection concept is Earth-skimming \(\nu_\tau\) astronomy. A tau neutrino enters the Earth at a shallow angle, undergoes a charged-current interaction, produces a tau lepton in the crust, and that tau emerges and decays in the atmosphere. The resulting air shower emits Cherenkov light, which is imaged by a telescope pointed slightly below the horizon. Trinity is intended to complement IceCube, Pierre Auger, and other techniques by targeting the sparsely populated but scientifically rich \(>1\)\,PeV to \(>1\)\,EeV range [2503.11864].

The Demonstrator’s pointing geometry is tuned to that use case. It is installed on Frisco Peak, Utah, at 2930 m above mean sea level. In its normal observing position, the camera points \(1.56^\circ\) below the horizon. A later status paper states the same geometry in camera-centric terms: only \(0.7^\circ\) of the vertical field of view sees the sky above the horizon while \(3.1^\circ\) images the ground below it. The top part of the camera is therefore a natural veto region for downward-going cosmic-ray air showers, a geometrical feature that directly enters the background-rejection logic [2503.11864, 2509.18236].

This observing mode inherits the broader Trinity rationale developed in the concept literature. The instrument watches the horizon for showers produced when a neutrino enters the Earth at a very small angle, produces a tau lepton in the crust, and that tau emerges and decays in the atmosphere. Because such showers can occur at very large distances, the method relies on wide-angle Cherenkov imaging rather than direct in-ice or radio detection. A plausible implication is that the Demonstrator’s scientific value lies as much in validating this geometry experimentally as in any immediate event yield [1907.08727, 2503.11864].

## 3. Optical system, camera, and readout chain

The deployed instrument is a one-square-meter-class imaging atmospheric Cherenkov telescope with a tessellated light-collection surface built from 77 circular mirror facets of 15 cm diameter, for a combined light-collection area of 1.36 m\(^2\). The optical design is Davies-Cotton, with the facets placed on a sphere of radius 1.48 m, equal to the focal length. At the focal plane is a 256-pixel silicon photomultiplier camera; the commissioning paper reports a \(3.87^\circ\times3.87^\circ\) field of view and \(0.24^\circ\) angular resolution, while the status paper gives \(3.8^\circ\times3.8^\circ\) and the same \(0.24^\circ\) angular resolution [2503.11864, 2509.18236].

| Subsystem | Reported configuration | Source |
|---|---|---|
| Site | Frisco Peak, Utah, 2930 m | [2503.11864] |
| Optics | 77 facets, 15 cm diameter, 1.36 m\(^2\), Davies-Cotton | [2503.11864] |
| Camera | 256-pixel SiPM camera, \(3.87^\circ\times3.87^\circ\), \(0.24^\circ\) | [2503.11864] |
| Readout | 100 MS/s, 12-bit, switched capacitor array | [2503.11864] |
| Trigger threshold | Equivalent of 20 photoelectrons | [2509.18236] |

The camera and readout are modular. The camera paper describes a \(4\times4\) matrix module of 16 SiPM pixels, each \(6\,\text{mm}\times6\,\text{mm}\), mounted to a custom backplane. For the Trinity Demonstrator, the SiPMs are Hamamatsu S14161-6050HS devices; the status paper gives the more specific device designation Hamamatsu S14161-6050HS-04. Front-end electronics are based on the eMUSIC ASIC, used for amplification and shaping, leading-edge discrimination, and per-channel bias trim and current monitoring. Digitization is performed with the AGET system, with 64 channels per AGET ASIC, a 512-cell switched-capacitor array, 100 MS/s sampling, and 12-bit resolution [2406.08274, 2509.18236].

The commissioning paper gives the as-operated signal-chain performance in greater detail. Camera signals are amplified and shaped on front-end boards, sent to a 256-channel AGET-based digitizer, and stored in a 512-cell ring-capacitor sampler; when triggered, the stored traces are digitized with 12-bit resolution. After shaping, the digitized SiPM signals have a full-width at half maximum of 30 ns. The readout is linear up to 125 detected photoelectrons. Full \(5.12\,\mu\)s traces of all camera pixels are recorded on trigger, taking 1.44 ms, and with an average trigger rate of less than 10 events per second the effective dead time is below 1.4% [2503.11864].

The trigger architecture is intentionally simple. Because each eMUSIC ASIC provides only an OR of its discriminator channels, the Demonstrator does not resolve the exact firing pixel at the first trigger level. The camera/readout paper therefore states that any discriminator firing triggers a readout and that image rejection is deferred to offline analysis, where the event must show a spatially extended air-shower image. In the status paper, the operational threshold is described as a single-SiPM threshold equivalent to 20 photoelectrons, at which point all camera channels are digitized [2406.08274, 2509.18236].

Thermal control is part of the instrument concept rather than an auxiliary service. The camera uses a liquid-cooled thermal transport system in which heat from the eMUSIC ASICs is conducted into copper blocks, transported by heat pipes to the camera sides, and coupled there to liquid-cooled cold plates using a glycol-based antifreeze fluid. For the Trinity Demonstrator specifically, the liquid circulates through an OMTech 6L Industrial Water Chiller set to \(6^\circ\)C. This cooling strategy reflects the fact that the eMUSIC ASICs dominate camera power dissipation [2406.08274].

## 4. Deployment, alignment, calibration, and remote operation

The telescope was designed for field assembly and remote maintenance. The optical support structure was mounted into a support frame bolted into the building foundation, and the telescope optical axis was permanently aligned at an azimuth angle of \(280^\circ\pm0.06^\circ\). The optical support structure was intended to be field-assembled by at most three persons, and on-site assembly took about three hours. The telescope is housed in a compact building with a partially removable roof and a roll-up door, with remote network access through a VPN gateway and local power cycling and environmental monitoring via a weather station and cameras [2503.11864].

Mirror alignment used the standard \(2f\) method originally used for Davies-Cotton systems. A screen was placed at \(2f\), or 2.96 m from the dish center, and the mirrors were adjusted so that the LED image projected by each facet coincided at the expected symmetric position. The alignment procedure took two people about two hours. Imaging of the \(2f\) screen with a Raspberry Pi HQ Camera yielded a radial brightness distribution with an 80% containment radius of \(4.3~\text{mm}\) at the \(2f\) point; simulations tuned to these measurements imply an 80% containment radius of \(2.1\)\,mm in the focal plane for a point source at infinity. The point spread function is explicitly stated to be well within one SiPM camera pixel, and a sunset image of the western mountain ridge confirmed that the ridge is clearly resolved on the camera [2503.11864].

Calibration is centered on a pulsed UV LED flasher derived from CTA calibration concepts. It emits at 400 nm with a 15-degree opening angle and is operated with 6 ns pulses during nightly calibration at 1 Hz. Uniformity tests found no intensity change across the focal plane within the 1% uncertainty of the measurement. For flat-fielding, 3,000 flasher events were recorded each at 42 V and 44 V SiPM bias, and linear response with bias voltage was assumed to derive per-pixel trim settings. After flat-fielding, the camera response was uniform within a 5% standard deviation, though during observing it broadened to 10%, partly due to temperature dependence of the SiPM breakdown voltage, which shifts by 0.1%/\(^\circ\)C [2503.11864].

Gain and trigger calibration were likewise quantified. From single-photoelectron dark-count structure, the calibration analysis derived a \(4.37\pm0.01\) mV/PE conversion factor; flasher measurements gave an average of \(40.5\pm0.2\) PEs per flash in the calibration SiPM, and a median camera amplitude of 975 DC yielded a factor of \(24.1\pm0.1\) DC/PE per camera pixel at \(12.7\pm0.1^\circ\)C. The trigger system was flat-fielded by scanning discriminator thresholds with flasher pulses at 100 cps, reducing the dispersion in trigger response from \(31.0\) DAC to \(4.0\) DAC. Using the flasher intensity of 48 PE per flash, the 50% trigger point of 22 DAC, and the noise threshold where trigger rates diverge at 242 DAC \(= 0\) PE, the observing threshold was set to 150 DAC, corresponding to 20 PE; this gave stable operation on dark, moonless nights with a trigger rate below 0.5 Hz [2503.11864].

Remote operation was an explicit design requirement and was reported as successful. The observer starts the system remotely, configures the camera and readout, verifies nominal operation, and then the telescope runs autonomously through the night and shuts itself down at dawn or under unsafe conditions. Safety logic checks sun and moon position, weather, storage, connectivity, SiPM current, temperatures, and trigger rates; independent scripts can force shutdown if sunrise is imminent; and the SiPM supplies are current-limited. The commissioning paper states that these procedures, together with standardized checklists and training, allowed a team that included undergraduate students to perform most observations [2503.11864].

## 5. Analysis workflow, backgrounds, and the null result

The first science data set discussed in the status paper consists of 346 hours of observations taken between October 2024 and May 2025, with the additional quality requirement that cloud bases be at least 300 m above the detector elevation. Over that period, the system triggered 979,957 times, and no neutrino candidate event was identified. The paper emphasizes that this is not unexpected at the Demonstrator’s current sensitivity; the analysis is designed primarily to find and understand backgrounds and to demonstrate that they can be controlled [2509.18236].

The event-processing chain follows the usual workflow for imaging atmospheric Cherenkov telescopes. Raw camera signals are gain-corrected using interleaved LED flashes that uniformly illuminate the camera, then calibrated into photoelectrons with additional correction for temperature-driven SiPM breakdown-voltage drift. Image cleaning proceeds by identifying the brightest pixel in the triggered sector, selecting core pixels with signal similar to that brightest pixel, retaining connected core pixels that touch another core pixel by a side or corner, and then keeping adjacent pixels with signal at least 30% of the core-pixel average. For further analysis, only events with a cleaned image containing at least three pixels are kept [2509.18236].

After cleaning and the three-pixel requirement, 497 events remained, corresponding to about 1.4 events per hour. Visual inspection divided them into four classes: 461 clean muon tracks, 15 muon spillover events, 2 broad/fuzzy events, and 19 events from a misbehaving camera module. The dominant background is therefore direct muon hits in the active SiPM volume. These appear as line-like tracks, and the paper describes the camera as acting like a muon tracker with about 3 mm position resolution in the camera plane. Because the active SiPM thickness is only about 100 \(\mu\)m, a muon must pass nearly parallel to the SiPM surface to produce a long track, which explains the low rate of about 1.4 track-like events per hour [2509.18236].

The commissioning paper independently examined related background behavior. Using coincidence-triggered muon paddles in vertical and horizontal configurations, it showed that direct muons can produce signals above the 20 PE trigger threshold during normal data taking, but that because these signals are confined to individual pixels they can be rejected during image cleaning by requiring at least two adjacent camera pixels. It also reported total RMS signal fluctuations of 1.9 PE for pixels viewing above the horizon and 1.3 PE for pixels below the horizon, with night-sky background dominating the noise budget; the same noise table lists electronic noise at 0.32 PE, SiPM dark counts at 0.47 PE, total noise of 1.25 PE for ground-viewing pixels, and 1.89 PE for sky-viewing pixels [2503.11864].

The null result is interpreted in the status paper as encouraging rather than adverse. Nearly all triggered events can be explained as background or hardware artifacts, and a simple qualitative analysis leaves only two events out of 346 hours—about one per 200 hours—that remain unclassified. Those two events do not resemble the “gold-plated” neutrino-shower signature shown in simulation. The paper therefore argues that backgrounds can be sufficiently suppressed and that a plastic-scintillator veto around the camera focal plane should reliably eliminate the muon-track background during future data taking [2509.18236].

## 6. Performance assessment and transition to Trinity One

The commissioning paper concludes that the Trinity Demonstrator met its design specifications after one year of operation. It reports a \(2.1~\text{mm}\) containment radius for a point source at infinity, a signal chain calibrated to a 100 PE dynamic range, an intrinsic noise floor including SiPM dark counts of 0.57 PE, and camera and trigger flat-fielding to within 10% and 5%, respectively. Operationally, the telescope observed 515.3 hours between June 1, 2024 and February 28, 2025, including 18.7 hours on TXS 0506+056 and 7.9 hours on NGC 1068 within the field of view [2503.11864].

Some commissioning results were explicitly diagnostic rather than directly usable for science analysis. In a separate cosmic-ray test, the telescope was pointed upward to a zenith angle of about \(30^\circ\), recorded one hour of data, and successfully triggered on air showers. The resulting images were split into two lobes because of sagging and slack in the mirror mounts when the telescope is rotated upward, producing a “cross-eyed” optical configuration; the authors therefore stopped analyzing upward-pointing air-shower images. This episode is important because it shows how the Demonstrator functioned as an engineering testbed for operational envelopes and mechanical tolerances [2503.11864].

The Demonstrator’s principal legacy is its role in de-risking Trinity One. The status paper states that the most important lesson is that the main background comes from charged particle interactions in the SiPMs, especially direct muons, and that these can be handled with image-based rejection plus a dedicated veto. Trinity One is then presented as the first full Trinity telescope: a wide-angle Cherenkov telescope on Frisco Peak, Utah, with a segmented 60 m\(^2\) light-collection surface, a \(60^\circ\) horizontal and \(5^\circ\) vertical field of view, a 3,328-pixel SiPM camera, \(0.3^\circ\) angular resolution, 250 MS/s sampling, 12-bit resolution, and the ability to rotate in azimuth from \(60^\circ\) to \(300^\circ\) [2509.18236, 2509.18237].

More broadly, the Demonstrator anchors the transition from proof of principle to observatory-scale neutrino astronomy. The full observatory is envisioned as 18 telescopes across at least three mountain-top sites, with sensitivity scaling with the combined field of view and observing time of all telescopes. In that progression, the Demonstrator’s significance lies not in the absence of detected neutrino candidates, but in the empirical validation of horizon-pointing Cherenkov imaging, remotely operated field deployment, and a background model dominated by identifiable and rejectable non-neutrino events [2509.18236, 2509.18237].

Source: https://www.emergentmind.com/topics/trinity-demonstrator