First results of a Monolithic Active Pixel Sensor with Internal Signal Gain Fully Integrated in a 180 nm CMOS Technology

This paper presents the first results of the CASSIA sensor, a novel monolithic active pixel sensor fabricated in 180 nm CMOS technology that utilizes fully integrated internal gain layers to achieve signal amplification, enabling operation in both low-gain proportional and high-gain single-photon avalanche modes for improved timing resolution and pile-up mitigation in high-luminosity particle physics experiments.

The Gist

Imagine you are trying to listen to a single whisper in a stadium filled with thousands of screaming fans. That is the challenge scientists face when trying to track tiny, fast-moving particles in high-energy physics experiments like those at CERN. The "whispers" are the tiny electrical signals created when a particle hits a sensor, and the "screaming fans" are the overwhelming background noise and the sheer volume of particles crashing into each other (called "pile-up").

This paper introduces a new kind of sensor called CASSIA (CMOS Active SenSor with Internal Amplification). Think of CASSIA not just as a microphone, but as a microphone that has a built-in, super-efficient amplifier right inside the diaphragm itself.

Here is the breakdown of what they did and why it matters, using everyday analogies:

1. The Problem: The "Weak Signal" Dilemma

Traditional sensors are like standard microphones. When a particle hits them, they produce a tiny, weak signal. To hear it, you need a big, external amplifier (a separate chip) to boost the volume.

• The Issue: Adding that external amplifier takes up space, uses more power, and adds "noise" (static). In the crowded, high-radiation environment of a particle collider, this makes it hard to distinguish the real signal from the noise.

2. The Solution: The "Internal Booster"

The CASSIA team built a sensor where the amplification happens inside the pixel itself, before the signal even leaves the sensor.

• The Analogy: Imagine a relay race. In the old way, the runner (the particle signal) hands the baton to a coach (the external chip) who then shouts the message to the crowd. In the CASSIA way, the runner has a megaphone attached to their chest. They shout the message while running, so it's loud and clear immediately.
• The Tech: They achieved this by adding a special "gain layer" (a specific type of electrical trap) right under the sensor's collection point. When a particle hits, it creates a few electrons. As these electrons try to escape, they get accelerated by an electric field and smash into other atoms, creating a chain reaction (avalanche) that multiplies the signal by thousands of times.

3. Two Modes: The "Dimmer Switch" and the "Flashbang"

The coolest part of this sensor is that it can do two different jobs just by turning a voltage knob (the bias voltage).

• Mode A: LGAD (Low-Gain Avalanche Diode) – The "Dimmer Switch"
  • How it works: You turn the voltage up just enough to get a gentle boost (amplification of 10x to 100x).
  • Use Case: This is great for tracking. It's like turning up the volume on a conversation so you can hear every word clearly without distorting it. It helps scientists see exactly where a particle went.
• Mode B: SPAD (Single-Photon Avalanche Diode) – The "Flashbang"
  • How it works: You turn the voltage up even higher, past a breaking point. Now, the sensor goes into overdrive. Even a single electron triggers a massive explosion of signal.
  • Use Case: This is great for timing. It's like a camera flash that goes off the instant a sound is made. It allows scientists to measure when a particle arrived with incredible precision (picoseconds), which is crucial for sorting out the "pile-up" of particles.

4. The "Swiss Army Knife" Design

Usually, to get a sensor that does timing well, you need a special, expensive manufacturing process. To get one that tracks well, you need a different one.

• The Innovation: CASSIA is built using a standard, mass-produced camera chip technology (180 nm CMOS)—the same kind used in your smartphone or digital camera.
• The Trick: They didn't change the factory recipe. Instead, they just tweaked the "ingredients" (the depth and size of the electrical layers) inside the standard chip. This means they can make these super-sensors cheaply and in huge quantities, just like regular camera sensors.

5. The Results: "The Whisper is Now a Roar"

The team tested their first prototype (CASSIA1) and found:

• It Works: They successfully created the internal chain reaction.
• It's Tunable: By changing the size of the internal "gain layer" (like changing the size of the megaphone), they could control how much the signal is boosted.
• It's Quiet: Even with the high gain, the sensor didn't generate too much "static" (dark noise), which is critical for seeing faint signals.
• It's Fast: The signal rises very quickly, meaning the sensor can react almost instantly.

Why Should You Care?

This technology is a game-changer for the future of particle physics.

• For Science: It allows experiments at the High-Luminosity LHC (and future colliders) to see through the chaos of millions of collisions, picking out the rare, interesting particles that might reveal new laws of the universe.
• For You: Because this uses standard camera chip manufacturing, the techniques learned here could eventually lead to better, faster, and more sensitive sensors in medical imaging (like better X-rays) or even in your phone's camera, allowing it to see in the dark or capture motion without blur.

In summary: The CASSIA team took a standard camera chip, gave it a built-in megaphone, and showed that it can hear the faintest whispers of the universe while running a marathon. It's a cheaper, faster, and smarter way to listen to the building blocks of reality.

Technical Summary

Here is a detailed technical summary of the paper "First results of a Monolithic Active Pixel Sensor with Internal Signal Gain Fully Integrated in a 180 nm CMOS Technology."

1. Problem Statement

High-energy physics (HEP) experiments at the High-Luminosity LHC and future facilities (FCC) require pixel detectors that provide not only spatial resolution but also precise timing information to mitigate pile-up effects. Current state-of-the-art solutions face a trade-off:

• Monolithic Active Pixel Sensors (MAPS): Offer low material budget and cost but typically rely on passive charge collection, resulting in lower signal amplitudes and poorer timing resolution compared to hybrid sensors.
• Low-Gain Avalanche Diodes (LGADs) and Single-Photon Avalanche Diodes (SPADs): Provide internal signal amplification and excellent timing resolution but are traditionally fabricated as separate chips using specialized processes (high-resistivity substrates, specific implant structures) incompatible with standard CMOS readout electronics. This necessitates expensive bump-bonding and increases material budget.

The challenge is to integrate internal charge multiplication (gain) directly into a standard, high-volume CMOS imaging process to create a monolithic sensor that combines the benefits of MAPS (compactness, low power) with the performance of LGADs/SPADs (high signal, fast timing).

2. Methodology

The authors developed the CASSIA (CMOS Active SenSor with Internal Amplification) sensor, a novel MAPS prototype fully integrated into a 180 nm TowerSemiconductor imaging process (originally used for the MALTA sensor family).

• Design Strategy: The sensor utilizes a depleted p-type epitaxial layer (30 μm thick) on a p+ substrate. Internal amplification is achieved by implanting a p-type gain layer beneath the central n+ collection electrode.
• Operation Modes: By adjusting the reverse bias voltage, the sensor operates in two modes:
  • LGAD Mode: Below hard breakdown, providing moderate gain (10–100) for low-noise tracking.
  • SPAD Mode: At or above breakdown, providing high intrinsic gain for single-photon sensitivity and ultra-fast timing.
• Prototype Variations (CASSIA1): The prototype chip contains 24 single pixels and four 3×3 pixel matrices with varying designs to optimize performance:
  • Gain Layer Depth: Deep p-well (DP) vs. Extra Deep p-well (XDP).
  • Gain Layer Diameter: Varied from 12 μm to 28 μm.
  • Electrode Depth: Standard, shallow, and deep n+ implants.
  • Spacing: Varied distance between the n+ electrode and the surrounding electronics deep p-well.
• Characterization: The sensors were tested using:
  • TCAD Simulations: To model electric fields, breakdown voltages, and avalanche multiplication regions.
  • DC Measurements: Current-Voltage ($I-V$) curves under dark and illuminated conditions.
  • Optical Scanning: Focused laser beams (532 nm and 785 nm) to map position-dependent gain and active area.
  • Pulsed Laser Tests: 1060 nm pulsed laser (10–100 kHz) to measure signal amplitude distributions and timing.
  • Dark Count Rate (DCR): Measurements at temperatures ranging from 0°C to 30°C to assess noise performance.

3. Key Contributions

• Process Integration: Demonstrated that internal charge multiplication can be achieved in a standard 180 nm CMOS imaging process without modifying process parameters or adding exotic layers. The gain layers utilize existing deep p-well and extra deep p-well implants.
• Dual-Mode Operation: Successfully showed that a single pixel design can function as both an LGAD (for tracking) and a SPAD (for timing) simply by changing the bias voltage.
• Design Optimization: Systematically investigated the impact of gain layer depth, diameter, and electrode geometry on breakdown voltage, gain uniformity, and noise.
• Scalability: Proved that large-area avalanche multiplication can be achieved in monolithic pixels, overcoming the limitations of previous hybrid LGAD approaches.

4. Key Results

• Breakdown and Gain:
  • Sensors exhibited distinct transition voltages: $V_{LGAD}$ (onset of amplification) and $V_{BR}$ (hard breakdown).
  • Voltage Tuning: Breakdown voltages varied significantly based on design. For example, M2 (DP gain layer) broke down at ~54 V, while M3 (XDP gain layer) broke down at ~99 V.
  • Gain Magnitude: Gains exceeded 5,000 in SPAD mode. In LGAD mode, gains of 10–100 were achieved over a voltage range of ~12 V.
  • Layer Depth Impact: Deeper gain layers (XDP) resulted in higher breakdown voltages and lower electric field peaks compared to shallower layers (DP).
• Uniformity and Fill Factor:
  • Light emission tests confirmed uniform avalanche multiplication across the active region, with no premature edge breakdown.
  • Position-dependent laser scans showed that the gain region extends beyond the physical gain layer implant. For instance, a 10 μm radius gain layer provided significant gain up to 15–20 μm, effectively increasing the pixel's "active" fill factor.
  • Increasing the gain layer diameter (e.g., from 12 μm to 28 μm) lowered the onset voltage for amplification and widened the uniform gain region.
• Dark Count Rate (DCR):
  • The lowest DCR ($\approx 0.01$ Hz/$\mu m^2$ at room temperature) was achieved with XDP gain layers combined with standard or deep n+ electrodes.
  • Shallow n+ electrodes combined with XDP layers resulted in significantly higher DCR (up to 30 Hz/$\mu m^2$), attributed to surface/interface defects.
  • DCR showed exponential dependence on temperature, consistent with thermally generated carriers.
• Signal Characteristics:
  • Pulse amplitude distributions were Gaussian for LGAD mode.
  • A smooth transition from LGAD to SPAD mode was observed, with a coexistence of signal types in the transition region.
  • The signal-to-noise ratio was significantly improved due to internal amplification, allowing for the use of simpler front-end electronics.

5. Significance

The CASSIA project represents a major step forward in the development of 4D tracking detectors (3D space + time) for High Energy Physics.

• Performance: It offers a path to sub-100 ps timing resolution (target) with high detection efficiency, essential for future collider experiments.
• Cost and Complexity: By integrating gain directly into standard CMOS, it eliminates the need for hybrid assembly and expensive specialized substrates, potentially lowering costs and material budget.
• Radiation Hardness: The use of a standard 180 nm process (proven by the MALTA sensor family) suggests inherent radiation hardness, though future studies (CASSIA2) will specifically optimize gain layer doping for non-ionizing energy loss (NIEL) tolerance.
• Versatility: The ability to switch between LGAD and SPAD modes allows the same sensor technology to be optimized for different applications, from high-rate tracking (LGAD) to low-light astrophysics or X-ray detection (SPAD).

In conclusion, the paper validates the feasibility of monolithic sensors with internal gain in standard CMOS, providing a robust foundation for the next generation of particle tracking and timing detectors.