GADGET: Gas Detector with Germanium Tagging

• GADGET is a detector system designed to measure weak beta-delayed proton decays and resolve ambiguous decay branches via prompt gamma-ray tagging.
• It combines a low-material gas proton detector with a high-resolution, segmented HPGe array that achieves energy resolutions around 0.3% at 1332 keV and >95% proton detection efficiency.
• GADGET II upgrades feature a compact TPC with 3D track imaging and advanced readout electronics, enabling detailed topology-sensitive studies and robust decay reconstruction.

Searching arXiv for the cited GADGET papers to ground the article in current records. GADGET, the Gas Amplifier Detector with Germanium Tagging, is a purpose-built experimental system for the measurement of weak, low-energy $\beta$-delayed proton decays with simultaneous prompt $\gamma$-ray tagging. Its defining feature is the combination of a gaseous charged-particle detector, which suppresses low-energy $\beta$ background while retaining high proton efficiency, with a surrounding high-purity germanium array that resolves coincident $\gamma$ rays and thereby disambiguates decay branches. In its original implementation, GADGET coupled a gaseous Proton Detector to the Segmented Germanium Array (SeGA); in GADGET II, the Proton Detector was upgraded into a compact Time Projection Chamber (TPC) capable of 3D imaging and particle identification, extending the apparatus from calorimetric proton spectroscopy to topology-sensitive studies of single- and multi-particle decays (Friedman et al., 2019, Mahajan et al., 2023).

1. Scientific role and measurement problem

GADGET was developed for experiments in which low-energy $\beta$-delayed proton emission competes with $\gamma$ decay and where proton spectroscopy alone can leave the populated final state ambiguous. This problem is especially acute in nuclear astrophysics, because the inverse reaction of interest is often a radiative proton capture on the ground state rather than on excited states. In such cases, identifying the specific final state populated after proton emission is necessary for constraining astrophysical reaction rates (Friedman et al., 2019).

The detector concept addresses a limitation of traditional silicon detectors: although they are effective for proton spectroscopy, they suffer badly from $\beta$-particle background at low energies, precisely where weak astrophysical proton branches are most difficult to isolate. GADGET replaces the silicon charged-particle stage with a gas-based detector that strongly suppresses $\beta$ background while preserving high efficiency for low-energy protons, and adds germanium tagging to identify prompt $\gamma$ rays emitted in coincidence.

The original GADGET system was already used for calorimetric measurements of low-energy $\beta$-delayed protons when particle identification was not required, with applications including $\gamma$0Cl decay to constrain $\gamma$1P$\gamma$2S, $\gamma$3Al decay to probe $\gamma$4Na$\gamma$5Mg, $\gamma$6Si proton–$\gamma$7 coincidence spectroscopy, and exotic decay searches in $\gamma$8Be. The later GADGET II program extended this capability to 3D tracking and particle identification, motivated in part by rare and topologically complex decays such as

$\gamma$9

which is relevant to indirect study of

$\beta$0

This suggests a progression from low-background proton calorimetry to full topology-resolved decay reconstruction (Mahajan et al., 2023).

2. Original detector architecture and germanium tagging

The original GADGET apparatus consisted of three quasi-independent elements: the central Proton Detector, the surrounding SeGA HPGe detector array, and a beam-pipe cross with diagnostics and degrader upstream. The beam direction was along the detector axis, and rare-isotope beams produced in-flight were implanted into the gas volume, where they decayed at rest. Operation was organized in a beam-on / beam-off cycle: during implantation, a gating grid protected the Micromegas from intense ionization caused by the beam, while during measurement the delayed protons were recorded cleanly (Friedman et al., 2019).

The germanium component was SeGA in a barrel configuration comprising 16 coaxial HPGe detectors arranged in two rings of eight, with eight detectors upstream and eight downstream of the proton-detector center. SeGA provided an energy resolution of about 0.3% at 1332 keV. In this configuration, germanium tagging served several distinct functions: it identified the final state populated after proton emission, separated different branches in the decay scheme, confirmed or refuted previously ambiguous proton assignments, and helped diagnose beam contamination and beam-position distribution in the gas.

The central Proton Detector was a gas-filled beam stopper and charged-particle detector. Its active volume was 40 cm long and 10 cm in diameter. It operated in P10 gas ($\beta$110% methane), typically at 780 Torr in the commissioning run, and could operate in a range of 0.05 to 2 atm. The detector vessel was designed with a low material budget, especially in the outer tube and field cage, to reduce $\beta$2-ray absorption.

Several internal subsystems were central to its operation. The field cage used a 50 $\beta$3m-thick Apical polyimide film, 37.9 cm long and 39.4 cm wide, rolled into a 12.2 cm-diameter cylinder. On its inner surface were 101 copper rings, each 3 mm wide and spaced by 750 $\beta$4m, connected through a resistor chain with total resistance 262.7 M$\beta$5. With nominal cathode bias $\beta$6 V, the resulting current limit was 22.85 $\beta$7A, according to

$\beta$8

The cathode itself was a 12.2 cm diameter stainless-steel disk with a 5.5 cm diameter beam hole covered by a 1.5 $\beta$9m aluminized Mylar window. A gating grid made of 60 gold-plated copper wires, each 20 $\gamma$0m diameter and spaced by 2 mm, could operate in unipolar mode to block beam-induced ionization electrons or in bipolar mode to collect part of the beam ionization charge.

Charge amplification and readout were provided by a Micromegas fabricated at CERN and integrated into the downstream PCB end-cap. The Micromegas used a stainless-steel micromesh with a 128 $\gamma$1m mesh-to-anode gap. The anode was segmented into 13 independent channels: a central circular pad of 14 mm radius, four $\gamma$2 annular sectors with 40 mm outer radius, and additional $\gamma$3 annular sectors extending to 50 mm outer radius, with 128 $\gamma$4m pad separation. At typical P10 conditions, the gas pressure was 800 Torr, the bias between mesh and anodes about 400 V, and the amplification field about 30 kV/cm (Friedman et al., 2019).

3. Commissioning of the original Proton Detector

Before beam operation, the Micromegas readout was tested with $\gamma$5Fe x rays in a small chamber called “the Pancake,” which had the same drift-field concept but only 1 cm length. In P10 gas at 780 Torr with collimated irradiation and 6 keV x rays, the individual pad resolution was 14%–16% FWHM, the veto pads yielded 20%–25% FWHM, and an example pad resolution of 14.5% FWHM at 6 keV was reported. Charged-particle response was then tested by introducing $\gamma$6Rn contamination from a $\gamma$7Th source. At 1000 Torr, the detector showed a 6.288 MeV $\gamma$8 peak with 2.8% FWHM, and the 6.778 MeV $\gamma$9 from $\beta$0Po was also observed (Friedman et al., 2019).

The full system was commissioned using the decay

$\beta$1

The $\beta$2Si beam was produced by fragmentation of a 150 MeV/u, 75 pnA $\beta$3Ar primary beam on a 1363 mg/cm$\beta$4 $\beta$5Be target, with purification through the A1900 fragment separator, a 300 mg/cm$\beta$6 Al wedge, and RFFS. The final beam was about 67% pure $\beta$7Si with rate about 4000 pps at the RFFS exit. The beam was pulsed with 500 ms beam on and 500 ms beam off; during commissioning, the detector operated in P10 at 780 Torr, with the gating grid in unipolar mode, grid voltage +150 V during implantation and -225 V during measurement, and about 15 ms delay around switching transitions.

Single-pad anti-coincidence spectra showed that low-energy $\beta$8 background was strongly suppressed, limited to roughly

$\beta$9

For the central pad, the main proton lines were 401 keV, 555 keV, and 943 keV. The corresponding resolutions were 6.3% FWHM at 401 keV, 5.4% at 555 keV, and 4.3% at 943 keV. Other pads ranged from 6.0% to 9.2% FWHM at 401 keV. In the combined detector spectrum, the 1040 keV proton line became more visible, although a larger continuum above 150 keV appeared and was interpreted as incomplete charge deposition from some protons.

A central quantitative result of the commissioning experiment was low-background, low-energy $\gamma$0-delayed proton detection efficiency above 95%. The count rate of the 401 keV protons as a function of time after beam cycles gave

$\gamma$1

consistent with the literature value

$\gamma$2

for $\gamma$3Si. This demonstrated that the detector could perform reliable decay-curve measurements in addition to proton spectroscopy (Friedman et al., 2019).

4. Proton–$\gamma$4 coincidences and the function of germanium tagging

The SeGA $\gamma$5-ray spectra were calibrated pad-by-pad and combined. Gating on the proton detector or on $\gamma$6 decays separated $\gamma$7 rays associated with different daughter nuclei: the $\gamma$8-gated spectrum corresponded to $\gamma$9Al levels, while the proton-gated spectrum corresponded to $\beta$0Mg levels. The $\beta$1-gated $\beta$2 spectrum also enabled identification of residual beam contaminants, and no significant contaminant $\beta$3 lines were found; upper limits of about 1% were placed on each of the expected contaminant species (Friedman et al., 2019).

The commissioning run established a proton–$\beta$4 coincidence efficiency of 2.7% at 1.37 MeV. Because the proton signal has a longer drift time than the prompt germanium signal, the coincidence analysis used a software coincidence gate of 8 $\beta$5s backward in time from each detected proton to search for SeGA triggers. This allowed identification of proton–$\beta$6 coincidences up to about 2 MeV in proton energy. A prominent observation was the 1040 keV proton branch in coincidence with the 2.75 MeV $\beta$7, confirming both the existence of that proton branch and its assignment to the 4.12 MeV excited state of $\beta$8Mg.

The drift-time structure also provided position information along the beam axis. With proton drift velocity

$\beta$9

the relation

$\beta$0

maps drift time to position. Since the proton ranges are short compared with the implantation distribution, the time differences between $\beta$1 rays and proton drift signals reconstruct the beam implantation profile along the detector axis. This coupling of calorimetry, timing, and germanium coincidence is the distinctive operational meaning of “germanium tagging” in GADGET.

A common misconception is that the germanium system is merely an auxiliary trigger. In the GADGET design, the HPGe array is integral to state assignment, branch separation, contamination studies, and beam-profile diagnostics. The coincidence information is not incidental to the apparatus; it is part of the detector’s core spectroscopy logic (Friedman et al., 2019).

5. GADGET II: conversion to a compact TPC

GADGET II upgraded the original Proton Detector into a compact TPC in order to provide 3D track imaging, particle identification by range/energy topology, detection of multi-particle emissions, and improved discrimination of $\beta$2 background and wall-effect events. The new detector retained the germanium-tagging concept while replacing the low-channel Micromegas pad plane with a high-granularity readout suitable for volumetric event reconstruction (Mahajan et al., 2023).

The GADGET II TPC is a cylindrical gaseous detector in which radioactive ions are thermalized in the gas and then decay at rest inside the active volume. Its key nominal parameters are: drift region length 400 mm, pad plane diameter 10 cm, measurement pads 1016, veto pads 8, pad size $\beta$3, amplification gap 128 $\beta$4m, gas P10 = 90% Ar + 10% CH$\beta$5, pressure 800 Torr, drift field 150 V/cm, amplification field 30 kV/cm, electron drift velocity $\beta$6 cm/$\beta$7s, and typical gas gain 40 for $\beta$8Rn $\beta$9 events.

The readout plane is a custom resistive-anode MICROMEGAS board fabricated at CERN. Of its 1024 total pads, 1016 are central measurement pads and 8 are surrounding veto pads. The resistive anode uses surface resistivity 10 M$\gamma$0/square and consists of a thin diamond-like carbon (DLC) layer on top of a 50 $\gamma$1m polyimide layer glued to the segmented readout plane. The polyimide and glue act as a dielectric in a 2D RC network. The charge dispersion in this geometry is small and effectively confined to a single pad, with a footprint of about

$\gamma$2

The stainless-steel micromesh has 18 $\gamma$3m wire diameter, 45 $\gamma$4m opening, 30 $\gamma$5m calendared thickness, and 45% optical transparency; the mesh–anode separation remains 128 $\gamma$6m, with avalanching in a field of roughly 30 kV/cm.

The gating grid is retained in the TPC geometry. It again contains 60 gold-plated copper wires, 20 $\gamma$7m diameter, with 2 mm spacing, and is used during implant-decay cycles: beam-on mode suppresses beam-induced ionization, while beam-off mode records delayed decay particles. The paper characterizes the resulting apparatus as one of the first generation of micro pattern gaseous detectors to utilize a resistive anode applied to low-energy nuclear physics, and states that the GADGET II TPC will be the first TPC surrounded by a high-efficiency array of high-purity germanium $\gamma$8-ray detectors (Mahajan et al., 2023).

6. Readout, validation, simulation, and learned representations

Because the TPC has 1024 channels, GADGET II uses the Generic Electronics for TPCs (GET) system. The GET chain comprises AsAd front-end cards, each carrying 4 AGET ASICs and handling 256 channels; CoBo boards, which collect data from AsAds, distribute the clock, remove subthreshold data, and manage data flow; and MuTanT, which provides the internal clock and triggering based on multiplicity or external triggers. In the FRIB setup, 4 CoBos were used, even though one CoBo can in principle read all 1024 channels, in order to reduce throughput bottlenecks. The AsAd boards were placed in a copper Faraday cage, the “AsAd box,” and connected by T-Zap boards directly to the Micromegas to minimize pickup noise (Mahajan et al., 2023).

A trigger is derived from the mesh signal. Ionization electrons entering the Micromegas gap induce signals on both mesh and resistive anode; the mesh signal is routed through a low-impedance charge amplifier, then a fast amplifier and leading-edge discriminator produce the trigger. The mesh and anode pulses have opposite polarity but equal pulse height. With the $\gamma$9 source, the measured mesh signal-to-noise ratio was

$\beta$0

Detector validation used both $\beta$1Rn $\beta$2 decays and cosmic-ray muons. For $\beta$3Rn in P10 at 800 Torr, after gain matching, the energy resolution was

$\beta$4

The analysis used outlier rejection by the Hotelling statistic and Squared Prediction Error (SPE), and used Principal Component Analysis (PCA) for track fitting, with the first principal component taken as the track length axis and the second as the track width axis. Range-versus-energy discrimination showed a main population of fully stopped $\beta$5 particles and a wall-effect population in which tracks terminated in inactive regions near the cathode or anode. Cosmic-ray muons, measured with two BC408 scintillators in coincidence, demonstrated performance for minimum-ionizing particles and allowed diffusion and drift-velocity studies. The empirical diffusion-width model was

$\beta$6

and the drift velocity measurement yielded

$\beta$7

with maximum drift time

$\beta$8

Simulation and reconstruction were performed with ATTPCROOT, a ROOT/FairRoot-based framework for active-target TPCs that provides event generation, custom detector geometries, Geant4-based transport, digitization, pulse-shape processing, and pattern recognition support. Simulated cases included $\beta$9Rn alpha decay, $\gamma$00Mg $\gamma$01O events, and efficiency studies for $\gamma$02Ga decays at FRIB. For the $\gamma$03Mg case, the proton energy was expected around 1.2 MeV, the proton range in P10 at atmospheric pressure about 3.7 cm, and the $\gamma$04 carried 506 keV with a track of less than 4 mm. The paper also introduced a VGG16-based convolutional neural network using early data fusion into a 2D composite image consisting of the 2D pad-plane projection with charge density, the time projection, and the integrated charge as an “energy bar.” On simulated $\gamma$05Mg proton, alpha, and proton–alpha coincidence events, it achieved 100% training and testing accuracy (Mahajan et al., 2023).

Representation learning for the GADGET II TPC was subsequently studied with sparse convolutional neural networks. Raw pad-level signals were represented as four-dimensional hit data $\gamma$06, with spatial coordinates voxelized and charge treated as a feature channel. Using voxel size 0.05 and Minkowski Engine batched coordinates, events were encoded as sparse tensors without padding or truncation. A sparse ResNet14 backbone with global max pooling and a fully connected head was pretrained on a binary proton–alpha classification task, and embeddings were taken from the penultimate layer. The task can be written as

$\gamma$07

with a linear probe

$\gamma$08

and training objective

$\gamma$09

In PCA space, even the randomly initialized sparse ResNet produced partial class separation, while pretraining yielded much cleaner separation of 800 keV protons, 1600 keV protons, and 2 MeV alphas, despite training only on a binary task. On a GADGET three-class probing task, a linear SVM reached 0.97 accuracy and 0.97 F1 with pretrained embeddings, versus 0.85 accuracy and 0.85 F1 with random weights and 0.33 accuracy and 0.17 F1 for the naive baseline. Transfer to AT-TPC data likewise improved from 0.58 accuracy and 0.53 F1 for the untrained model to 0.74 accuracy and 0.70 F1 for the GADGET-pretrained encoder on a grouped track-counting task, suggesting that sparse convolutional features capture nontrivial detector-agnostic structure in TPC data (Wheeler et al., 14 Nov 2025).

7. Scope, related detector philosophies, and nomenclature

Within nuclear astrophysics and radioactive-beam spectroscopy, “GADGET” refers specifically to the gas-amplifier detector system with germanium tagging described above and to its TPC successor, GADGET II. The acronym denotes the coupling of a gaseous charged-particle detector to a high-resolution germanium array for coincidence spectroscopy (Friedman et al., 2019).

A related but distinct usage appears in work on very low-energy neutrino scattering detectors. That work proposes and bench-tests a gaseous proportional counter with a sub-keV electron-equivalent threshold, motivated by measurement of the neutrino magnetic moment and coherent neutrino-nucleus scattering. The paper explicitly describes its detector philosophy as a future GADGET-like configuration in which the gas detector provides ultra-low-threshold recoil measurement and a germanium system can be used to tag or benchmark associated radiation or backgrounds, but it does not use the acronym “GADGET” directly. The underlying idea is similar in that it pairs low-threshold gaseous detection with germanium-based tagging or benchmarking, but the target observables, detector geometry, and experimental program are different (Kopylov et al., 2014).

That low-threshold concept emphasizes a threshold of about 5 eV, gas gain about $\gamma$10, operation at about 1 MPa, and pulse-shape discrimination that suppresses noise by about a factor of $\gamma$11 in the 5–100 eV region. It proposes an array of 16 counters arranged in 4 planes of 4 counters each inside shielding of 30 cm of iron, 40 cm of borated polyethylene, 50 cm of water, and an external plastic scintillator veto, with reactor antineutrino flux taken as

$\gamma$12

The relevant recoil and scattering formulas were given schematically as

$\gamma$13

with

$\gamma$14

and the $\gamma$15 magnetic-moment term rising as $\gamma$16 at low recoil energy. The significance of this literature for GADGET is therefore conceptual rather than historical identity: it illustrates that “gaseous detector with germanium tagging” can also denote a broader detector philosophy in which gas amplification, waveform-based discrimination, and germanium-assisted background control are combined for rare-event measurements (Kopylov et al., 2014).

Taken together, the literature defines GADGET in two layers. In the narrow and standard sense, it is the NSCL/FRIB decay-spectroscopy apparatus built around a gaseous proton or TPC detector and HPGe coincidence tagging. In a broader detector-physics sense, it also names, or closely anticipates, a class of gas-plus-germanium strategies in which the gas medium provides favorable threshold or topology sensitivity and the germanium system provides spectroscopic or background-discriminating context.