Beam-test evaluation of pre-production Low Gain Avalanche Detectors for the ATLAS High Granularity Timing Detector

The High Granularity Timing Detector (HGTD) will be installed in the ATLAS experiment as part of the Phase-II upgrade for the High Luminosity-Large Hadron Collider (HL-LHC). It will mitigate pile-up effects in the forward region, and measure per bunch luminosity. The design of HGTD is based on Low Gain Avalanche Detector (LGAD) sensors. This paper presents the results of beam-test campaigns conducted at CERN and DESY in 2023 and 2024 on single LGADs from HGTD pre-production test structures, before and after neutron irradiation up to fluences of $2.5 \times 10^{15}~\mathrm{n_{eq}/cm^2}$. The tested LGADs can meet HGTD requirements in terms of charge collection, time resolution, and hit efficiency, even under HL-LHC end-of-life conditions, supporting their deployment in the final detector.

💡 Research Summary

The paper reports on a comprehensive beam‑test program carried out to qualify the pre‑production Low‑Gain Avalanche Detector (LGAD) sensors that will form the active elements of the ATLAS High Granularity Timing Detector (HGTD) in the Phase‑II upgrade for the High‑Luminosity LHC. The HGTD is intended to mitigate the severe pile‑up expected in the forward region (2.4 < |η| < 4.0) by providing per‑hit time measurements with a resolution of 40 ps (or better) and by contributing to per‑bunch luminosity determination.

Two design families, produced by the Institute of Microelectronics (IME) in China for the IHEP and USTC collaborations, were investigated. Both designs employ an n‑in‑p diode with a thin (≈50 µm) active layer and a p⁺ gain layer that is co‑implanted with carbon. Carbon doping is known to slow the acceptor removal process under irradiation, thereby preserving gain at high fluences. The pre‑production devices are single‑pad Quality‑Control Test Structures (QC‑TS) extracted from 8‑inch wafers. Two thickness variants were studied: a “core” pre‑production thickness of 300 µm and an “early” pre‑production thickness of 775 µm (the latter only for IHEP).

All sensors were irradiated with fast neutrons at the TRIGA reactor in Ljubljana to fluences of 0, 0.8, 1.5 and 2.5 × 10¹⁵ n_eq cm⁻², the latter corresponding to the end‑of‑life (EOL) requirement for the HGTD. Post‑irradiation I‑V measurements performed at –30 °C showed comparable leakage currents for sensors receiving the same fluence, confirming uniformity of the irradiation and the robustness of the devices up to the safe electric field limit of ≈11 V µm⁻¹ (≈550 V for a 50 µm active layer).

Beam tests were conducted at two facilities: the CERN SPS H6A line with a 120 GeV pion beam, and the DESY II test beam with 5 GeV electrons. The devices under test (DUTs) were mounted on custom read‑out boards that provided a first‑stage transimpedance amplifier (≈4.7 kΩ) followed by a 20 dB second‑stage commercial amplifier. The second stage could be placed either inside the cooling enclosure (to minimise temperature‑dependent gain variations) or outside; a systematic uncertainty of 10 % was assigned to the collected charge and timing results to cover this effect.

Time reference was supplied by a Micro‑Channel Plate Photomultiplier Tube (MCP‑PMT, HPK R3809U‑50). The MCP‑PMT was calibrated using coincidences with two LGADs, yielding an intrinsic resolution of 10.6 ± 2.2 ps at 2650 V. An eight‑channel oscilloscope (1 GHz bandwidth, 6.25 GS/s) recorded waveforms from both the DUTs and the MCP‑PMT.

The analysis extracted the collected charge by integrating the amplified pulse and the arrival time using a constant‑fraction discriminator algorithm. Charge‑versus‑bias curves showed that non‑irradiated sensors reach >20 fC already at 200 V, while the most heavily irradiated sensors (2.5 × 10¹⁵ n_eq cm⁻²) achieve the required >15 fC at ≈300 V. This satisfies the HGTD specification of >15 fC (initial) and >4 fC (EOL) for a 2 fC discriminator threshold of the ALTIROC front‑end ASIC.

Time resolution improves with bias voltage, reaching 25–35 ps across the 350–500 V range. Even after the highest fluence, the resolution remains below 30 ps, comfortably meeting the 40 ps target. Hit‑efficiency maps, built from reconstructed tracks, indicate central‑pad efficiencies of 97–99 % for unirradiated devices and 95–97 % after the highest fluence, exceeding the required >97 % (initial) and >95 % (EOL). Efficiency shows only a mild dependence on the incident angle (≤2 % loss up to 20°).

The results confirm that carbon‑enriched gain layers effectively mitigate acceptor removal, allowing the LGADs to operate at lower bias while preserving gain. The difference in wafer thickness (300 µm vs 775 µm) does not materially affect the timing or charge performance, and the absence of under‑bump metallisation (UBM) in early‑pre‑production devices does not degrade the measured metrics.

In conclusion, the pre‑production LGADs satisfy all HGTD performance criteria after exposure to the full HL‑LHC radiation budget. This validation paves the way for full‑scale production, integration with the ALTIROC ASIC, and system‑level tests of complete 15 × 15 LGAD arrays. Future work will focus on module‑level thermal management, long‑term stability under realistic operating conditions, and the final integration of the timing detector into the ATLAS forward region for the HL‑LHC run.