All Research

Master Project (PDM) · 2026

Radiation Damage Analysis for Irradiated SiPMs using Emission Microscopy

Overview

Silicon Photomultipliers (SiPMs) are the detector of choice in many high-energy physics experiments, including LHCb. In radiation-harsh environments they accumulate damage from high-energy particles, which dramatically increases their dark count rate (DCR) — a noise floor that can eventually overwhelm physics signals. Understanding where inside each microcell that damage sits, and how strongly each defect site contributes to noise, is crucial for developing next-generation radiation-tolerant sensors.

This project applied Emission Microscopy (EMMI) — infrared camera imaging through a microscope — to directly visualise avalanche emission hotspots in both non-irradiated and neutron-irradiated SiPMs from Hamamatsu and FBK. By combining EMMI imaging with IV measurements, I established a device-independent linear relationship between emission intensity and dark current, and developed a 2D Gaussian reconstruction pipeline to extract the amplitude, size, and position of individual radiation damage points within single microcells.

---

SiPM Mechanism & Infrared Emission

A SiPM is an array of Single-Photon Avalanche Diodes (SPADs), each biased above the breakdown voltage (V_bd). An incoming photon — or a thermally generated carrier — triggers an avalanche in the high-field multiplication region near the silicon surface. The accelerating charge carriers lose energy via Bremsstrahlung, emitting infrared photons in the process. These photons can be captured by a cooled scientific camera mounted on a microscope, giving a direct spatial map of where avalanches are occurring.

Radiation damage (from neutrons or protons) creates trap states in the silicon bulk via Non-Ionising Energy Loss (NIEL). These traps generate electron–hole pairs through Shockley–Read–Hall (SRH) recombination, increasing the DCR and the associated IR emission. The emission intensity is therefore a proxy for local trap density — and for local dark current.

---

Devices Under Study

Two SiPM technologies were studied across four neutron fluence levels plus a non-irradiated reference, giving ten samples in total.

Hamamatsu H2024 (50 µm pitch)

Name	Pixel Size	Fluence (neq/cm²)	Vbd ch118
H2024	50 µm	1×10¹³	51.69 ± 0.02 V
H2024	50 µm	3×10¹²	50.97 ± 0.02 V
H2024	50 µm	1×10¹²	51.45 ± 0.07 V
H2024	50 µm	3×10¹¹	51.31 ± 0.02 V
H2024	50 µm	0 (reference)	51.10 ± 0.11 V

FBK W3m / W1m (31 µm pitch)

Name	Pixel Size	Fluence (neq/cm²)	Vbd ch118
W3m	31 µm	1×10¹³	31.07 ± 0.02 V
W3m	31 µm	3×10¹²	31.00 ± 0.02 V
W3m	31 µm	1×10¹²	31.11 ± 0.02 V
W3m	31 µm	3×10¹¹	31.19 ± 0.02 V
W1m	31 µm	0 (reference)	30.94 ± 0.04 V

The V_bd values show minimal systematic shift with fluence for both technologies, confirming that bulk damage does not significantly alter the junction doping balance at these fluence levels.

---

Emission Microscopy (EMMI) Technique

EMMI works by imaging the IR photons spontaneously emitted by avalanching microcells. A cooled scientific camera is mounted on a microscope above the biased SiPM. Two images are taken:

• Reference image — taken under visible-light illumination to provide a structural map of the microcell grid.
• Emission image — taken in complete darkness with the SiPM biased at a chosen over-voltage (OV = V_bias − V_bd), typically for a 10-minute exposure to accumulate sufficient photon statistics.

The two images are then co-registered and overlaid, with the emission intensity rendered as a false-colour heat map on top of the structural reference.

---

IV Scan Results

Before EMMI acquisition, each sample was characterised with a current–voltage (IV) scan. The dark current above breakdown — directly proportional to DCR — scales systematically with neutron fluence for both technologies.

---

Camera Noise Subtraction

For non-irradiated SiPMs the emission signal is extremely faint — typically <200 ADU/px above background — making camera noise the dominant contribution. A dedicated dark image is taken under identical conditions (10-minute exposure, same temperature, SiPM unbiased). This dark frame is pixel-wise subtracted from the emission image before analysis, revealing the true photon emission structure.

---

Non-Irradiated SiPMs: Defect Localisation

Even in pristine, unirradiated SiPMs, EMMI reveals discrete bright hotspots at high over-voltages. These spots are reproducible, device-specific, and grow in brightness but not in number as OV is increased from 1.0 V to 10.0 V.

This pattern was observed consistently across multiple channels (ch101, ch102, ch117, ch118) and in the FBK W1m device at 12.0 OV. The hotspots always appear at, or very close to, the inter-cell trench walls. While this behaviour has been noted qualitatively in prior EMMI literature (Acerbi et al., Sensors 2021; Spatially-resolved DCR, Eur. Phys. J. C 2018), no clear mechanism has been established. Three candidate explanations are:

1. Etch-induced crystal damage: The deep reactive-ion etching (DRIE) used to form trench isolation creates lattice damage at trench walls, leaving permanently active trap states.
2. Field leakage at high OV: At elevated bias, the trench dielectric is no longer fully effective at isolating the electric field between adjacent cells, creating enhanced-field volumes at the trench wall.
3. Field line crowding at trench corners: The abrupt geometry at trench walls concentrates field lines, producing locally amplified electric field that triggers enhanced avalanche activity.

---

Irradiated SiPMs: Emission vs. Fluence

In irradiated devices the emission character changes fundamentally. Instead of isolated hotspots, every cell glows — the active area fills uniformly with emission that brightens rapidly as OV increases above breakdown. At 3.0 OV the cells are already near saturation and the emission intensity across all 160 visible cells follows a near-Gaussian distribution.

---

Emission Intensity ∝ Dark Current

A key physics question is whether the total optical emission from a cell is simply proportional to its dark current, or whether the relationship depends on fluence, bias, or device technology. To test this, the mean emission intensity of the active channel was plotted against the simultaneously measured dark current for each bias point and each fluence level.

The physical interpretation is straightforward: each SRH generation event (whether from a manufacturing defect or a radiation-induced trap) produces one avalanche, which in turn emits a fixed average number of IR photons. The proportionality constant (~57 ADU/px per µA in the active channel) reflects the camera sensitivity and the collection geometry, not the device physics. The single universal line spanning all fluences rules out any significant fluence-dependence of the per-avalanche photon yield.

---

Damage Point Analysis via Gaussian Reconstruction

At low to moderate over-voltages in irradiated FBK SiPMs, individual cells show discrete bright spots — localised emission peaks corresponding to single damage clusters. To extract quantitative information about each damage point (amplitude, spatial sigma, x–y position), a two-stage algorithm was developed:

1. Peak finding: A brightness mask (top 50% of pixel values) isolates candidate regions; connected-component labelling then finds peaks with an area threshold of ≥40 px to suppress noise spikes.
2. 2D Gaussian reconstruction: Each identified peak is fit with a 2D elliptical Gaussian to extract amplitude, centroid (x, y), and widths (σ_x, σ_y). The sum of all Gaussians is compared against the original image as a validation step.

The Gaussian parameters (amplitude, sigma, position) for each identified damage point are then catalogued across all cells and bias voltages. This enables spatial statistics on damage point density, comparison between fluence levels, and — in future work — correlation with specific neutron track signatures from Monte Carlo simulations.

---

Summary

This work forms part of a broader effort at LPHE, EPFL to understand and mitigate radiation damage in SiPMs destined for the LHCb Upgrade II VELO and SciFi detectors, which will operate at luminosities up to 1.5×10³⁴ cm⁻² s⁻¹.

All Research