T-REX Neutron Spectrometer Overview

• T-REX is a neutron time-of-flight spectrometer that uses Multi-Grid technology with 10B-coated aluminium blades to detect thermal neutrons.
• Advanced internal shielding with Ni-plated Al/B4C composites significantly reduces internal scattering, thus improving energy and momentum resolution.
• Comparative studies between VMM3A ASIC and CREMAT electronics reveal a trade-off between rapid digitization and uniform voxel response.

Searching arXiv for recent T-REX neutron spectrometer and Multi-Grid papers. T-REX is a neutron time-of-flight spectrometer being built at the European Spallation Source. Its detector system uses Multi-Grid technology, a voxelised proportional counter architecture based on thin $^{10}$B-enriched $B_4C$ coatings on aluminium blades for detection of scattered thermal neutrons. The technical development reported for T-REX has focused on two coupled requirements: suppressing internal neutron scattering generated by structural materials inside the detector, and establishing a readout chain that remains viable at ESS scale in channel count and rate while preserving acceptable pulse-height and efficiency uniformity across individual voxels (Backis et al., 4 Nov 2025, Backis et al., 2 Oct 2025).

1. Detector concept and neutron-conversion mechanism

The Multi-Grid detector in T-REX consists of modular grids: lattices of normal and radial aluminium blades forming the cathodes of a voxelised proportional counter. Thin vertical anode wires pass through the centers of the voxels. In the T-REX prototype geometry studied for readout, each voxel is an elongated rectangle with dimensions $23.5 \times 9.5~\mathrm{mm}^2$ in $(x,z)$ and $23.5 \times 24.0~\mathrm{mm}^2$ in $(x,y)$ (Backis et al., 2 Oct 2025).

Neutron detection relies on $^{10}$B-enriched $B_4C$ coatings on the normal blades. One description gives a thin film of approximately $1$–$2~\mu\mathrm{m}$ with $B_4C$0 $B_4C$1B enrichment; another describes thin $B_4C$2 coatings on the Al blades of the grids more generally. Thermal neutrons interact through the $B_4C$3 reaction, or equivalently

$B_4C$4

and one of the charged reaction products escapes the film and ionises the counting gas, given as Ar-CO$B_4C$5 in the prototype studies. The resulting charge is collected on the anode wire and read out as a proportional-counter signal (Backis et al., 2 Oct 2025).

Aluminium is used for structural elements, including the blades and containment vessel, because of its low neutron absorption cross-section. That material choice, however, does not eliminate neutron-induced background. Neutron scattering in Al and surrounding material produces background and measurement distortion by smearing time-of-flight and scattering angle, and therefore energy and momentum measurements. This makes detector mechanics and detector physics inseparable in T-REX (Backis et al., 4 Nov 2025).

2. Internal scattering as a detector-background problem

The principal materials issue identified for T-REX is “internal scattering”: scattering of neutrons within the detector structure, especially from aluminium and nearby components, which generates backgrounds that degrade spectrometer sensitivity and resolution. In the T-REX context, the problem is not simply loss of counts but distortion of the measured inelastic signal through multiple scattering inside the detector vessel and grid structure (Backis et al., 4 Nov 2025).

A Monte Carlo program was therefore used to study internal shielding. The simulations employed Geant4 enhanced with NCrystal for detailed thermal-neutron interaction modelling. The stated objective was to evaluate different shielding materials and placements within T-REX, including shielding on grid sides, at the rear, and on radial blades, and to quantify their effectiveness in reducing multiple scattering. These studies showed that internal shielding based on $B_4C$6 dramatically reduces internal neutron scattering (Backis et al., 4 Nov 2025).

For radially placed blades, the requirements were more restrictive than simple neutron absorption. The candidate material had to be mechanically robust, electrically conductive for use in the voxelised proportional-counter geometry, a strong neutron absorber, and manufacturable in thin sheets. Those constraints drove the comparative study of composite materials, coatings, and plated structures rather than pure absorber layers alone (Backis et al., 4 Nov 2025).

3. Experimental characterization of shielding materials

Following the Monte Carlo work, neutron transmission and scattering properties of twelve shielding-material samples were measured at the ISIS spallation neutron source. The samples targeted combinations of Al, $B_4C$7, Ni, epoxy, and different application techniques, including sputtering, drip coating, cold spray, and bulk Al/$B_4C$8 composites (Backis et al., 4 Nov 2025).

Measurement	Beamline and range	Core method
Transmission	EMMA, $B_4C$9–$23.5 \times 9.5~\mathrm{mm}^2$0	sample-in/sample-out TOF ratio with a 2D neutron GEM detector
Scattering	Merlin, $23.5 \times 9.5~\mathrm{mm}^2$1, $23.5 \times 9.5~\mathrm{mm}^2$2, $23.5 \times 9.5~\mathrm{mm}^2$3, $23.5 \times 9.5~\mathrm{mm}^2$4	energy-transfer spectra and angular yields after sample-out subtraction

In the EMMA transmission study, the beam was collimated to $23.5 \times 9.5~\mathrm{mm}^2$5 and monitored with a low-efficiency beam monitor based on $23.5 \times 9.5~\mathrm{mm}^2$6 GS1 scintillating glass. Transmission was recorded with a $23.5 \times 9.5~\mathrm{mm}^2$7 2D position-sensitive neutron GEM detector with 128 x/y strips at $23.5 \times 9.5~\mathrm{mm}^2$8 pitch, corresponding to 16384 pixels, and a $23.5 \times 9.5~\mathrm{mm}^2$9B layer on the cathode. Transmission was defined through the sample-in/sample-out ratio

$(x,z)$0

using $(x,z)$1 barns, where $(x,z)$2 is the number density of $(x,z)$3B, $(x,z)$4 the sample thickness, and $(x,z)$5 the neutron wavelength in ångströms (Backis et al., 4 Nov 2025).

The Merlin scattering measurements were performed with an array of $(x,z)$6He tubes positioned on an arc $(x,z)$7 from the sample, with the sample at $(x,z)$8 relative to the incident beam. Scattered yields were compared over $(x,z)$9, and relative yields were normalized so that peak elastic scattering was $23.5 \times 24.0~\mathrm{mm}^2$0 for pure Al. This setup enabled separation of elastic/coherent contributions, including Bragg features, from broader non-elastic shoulders (Backis et al., 4 Nov 2025).

4. Comparative performance of candidate shielding materials

The material comparisons identify Al/$23.5 \times 24.0~\mathrm{mm}^2$1 composites, especially the Ni-plated BA31-1 variant denoted NiBA31-1, as the most effective compromise between neutron absorption, scattering suppression, electrical functionality, and manufacturability. In transmission, BA31-1 and particularly NiBA31-1 showed decay constants matching manufacturer’s specifications, very effective neutron absorption, and high material uniformity. The Ni-plated BA31-1 is electrically conductive and suitable for grid structures. Two-dimensional absorption images showed excellent uniformity for the composite, in contrast to dripped or sprayed coatings, which exhibited banding and irregularity (Backis et al., 4 Nov 2025).

In scattering, BA31-1 and NiBA31-1 showed suppressed coherent scattering, including reduced Bragg peaks relative to pure Al, especially at longer wavelengths. At the shortest measured wavelength, $23.5 \times 24.0~\mathrm{mm}^2$2, these samples, which contain more total Al, had slightly higher scattering than pure Al, and the Ni plating slightly increased yield in the elastic region, though not dramatically. Among the studied samples, BA25-3 performed best in terms of lowest neutron scattering yields (Backis et al., 4 Nov 2025).

Other material classes were less favorable for T-REX. Drip-coated $23.5 \times 24.0~\mathrm{mm}^2$3/epoxy samples had lower transmission but higher scattering, especially in non-elastic shoulders, attributed to the organic resin content. Sputtered $23.5 \times 24.0~\mathrm{mm}^2$4 on Al reduced scattering compared with Al at longer wavelengths, but sputtering was limited in achievable thickness to approximately $23.5 \times 24.0~\mathrm{mm}^2$5. Cold-spray and drip/spray approaches also showed more variable uniformity and more difficult thickness control than the composite materials (Backis et al., 4 Nov 2025).

The resulting material preference is explicit: Ni-plated Al/$23.5 \times 24.0~\mathrm{mm}^2$6 composite material is preferred for radial blades and other internal shielding because it combines high neutron absorption due to $23.5 \times 24.0~\mathrm{mm}^2$7B, low scattering background, electrical conductivity, structural robustness, and suppression of intrinsic alpha background through Ni plating. The reported operational implication is that internal shielding using Ni-plated Al/$23.5 \times 24.0~\mathrm{mm}^2$8 composite material on the radial blades drastically reduces backgrounds from internal neutron scattering in T-REX, improving energy/momentum resolution and sensitivity to weak inelastic signals. Thicker composite sheets can be used at the grid’s rear, for example $23.5 \times 24.0~\mathrm{mm}^2$9 BA31, for further suppression of backscattering from support structures, and this choice replaces the need for additional side shielding in most cases (Backis et al., 4 Nov 2025).

5. Readout architecture and voxel-response uniformity

The readout study for T-REX examined two signal-processing schemes on a Multi-Grid prototype: the VMM3A ASIC and CREMAT electronics. The VMM3A is described as a fast, high-density ASIC with $(x,y)$0 maximum peaking time and prior use on Multi-Blade detectors at ESS. CREMAT provides charge-sensitive preamplification and shaping with $(x,y)$1 shaping times and serves as a performance reference for near-full charge integration, but it is not scalable to the 12,480 required channels of T-REX (Backis et al., 2 Oct 2025).

Readout	Peaking or shaping time	Role in the T-REX study
VMM3A ASIC	$(x,y)$2 maximum peaking time	scalable candidate for T-REX
CREMAT	$(x,y)$3 shaping times	reference for fuller charge integration

The measurements used the TRP-1 prototype, comprising 12 grids and 120 anode wires, exposed to a collimated neutron beam at the ISIS EMMA beamline. The beam size was $(x,y)$4 and was scanned along $(x,y)$5 from $(x,y)$6 to $(x,y)$7 across a voxel. Pulse-height spectra, waveforms, and relative voxel efficiencies were recorded with both readout chains (Backis et al., 2 Oct 2025).

The observed position dependence follows directly from voxel electrostatics. Voxel corners have low electric-field gradients and therefore slower charge collection. For VMM3A, the short peaking time produces sharp pulses with well-defined maxima near the voxel center but more variable pulse shapes with extended tails near voxel edges. Correspondingly, VMM3A shows loss of pulse height and pulse-height resolution as the beam approaches the voxel edge. CREMAT, by contrast, yields pulse shapes that are comparatively invariant with position and pulse-height distributions that remain much more uniform across the voxel, including the characteristic $(x,y)$8He bump at the high end of the spectrum (Backis et al., 2 Oct 2025).

A potentially misleading feature of the measurements is that VMM3A showed an approximately $(x,y)$9 higher Multi-Grid to beam-monitor counting ratio than CREMAT in the voxel center for $^{10}$0. The study does not interpret this as superior intrinsic charge integration by VMM3A. The reported explanation is dead-time and data-transfer limitations of the CREMAT external digitizer in the high-rate, pulsed-beam environment, resulting in substantial data loss and possible pile-up. Simulations predict slightly higher efficiency for CREMAT than for VMM3A, as expected theoretically, while VMM3A still exhibits a more rapid decrease in efficiency at voxel corners because of incomplete integration (Backis et al., 2 Oct 2025).

6. Simulation framework, validation, and operational implications

Two complementary simulation chains underpin the T-REX detector-development program. For shielding and internal-scattering studies, Geant4 with NCrystal was used to model thermal-neutron interactions in candidate structural and shielding materials. For the readout study, Geant4 simulated neutron capture and the energy and emission-angle distributions of the $^{10}$1He and $^{10}$2Li products emerging from the $^{10}$3 film, while Garfield++ simulated gas ionisation, drift, and full time-dependent charge collection $^{10}$4 in the actual electric-field configuration, including stochastic ionisation-track energy loss from SRIM (Backis et al., 4 Nov 2025, Backis et al., 2 Oct 2025).

Pulse shaping in the readout simulations was applied in post-processing through

$^{10}$5

For VMM3A, the parameters were $^{10}$6, $^{10}$7, $^{10}$8, and $^{10}$9; for CREMAT, $B_4C$0, $B_4C$1, $B_4C$2, and $B_4C$3. In the shielding study, the transmission model

$B_4C$4

was validated, and simulated and measured decay constants and angular distributions agreed within uncertainties. In the readout study, simulated Multi-Grid to beam-monitor counting ratios agreed well with the VMM3A data once systematics of approximately $B_4C$5 were considered, and simulated waveforms reproduced the delayed, broader pulses at voxel corners (Backis et al., 4 Nov 2025, Backis et al., 2 Oct 2025).

The validated design implications are specific. For the detector structure, Ni-plated Al/$B_4C$6 composite material is the preferred internal shielding and radial-blade material because it suppresses multiple-scattered neutron background while preserving conductivity and mechanical integrity. For the electronics, the short VMM3A peaking time is not optimal for full charge integration from all parts of the voxel, and longer peaking times would improve uniformity; however, the VMM3A architecture is favored because of fast onboard digitization, large channel density, high data-transfer bandwidth, minimized dead-time and pile-up in intense pulsed beams, and scalability to the T-REX channel count. The resulting T-REX detector concept is therefore defined by an explicit trade-off: modest voxel-edge non-uniformity is accepted in order to obtain a readout system compatible with ESS rate and scale, while internal material optimization is used to suppress structural backgrounds at their source (Backis et al., 2 Oct 2025, Backis et al., 4 Nov 2025).