The new truly cylindrical tracker for the ALICE ITS3

This paper presents the design and R&D progress of the new cylindrical Inner Tracking System (ITS3) for the ALICE experiment, highlighting its ultra-lightweight wafer-scale Monolithic Active Pixel Sensors, air-cooled architecture, and demonstrated high performance in spatial resolution, detection efficiency, and time resolution.

The Gist

Imagine the Large Hadron Collider (LHC) as a massive, high-speed racetrack where particles zoom around at nearly the speed of light. Inside this track, the ALICE experiment acts like a super-high-speed camera, trying to take pictures of what happens when these particles crash into each other.

The paper describes a major upgrade to the "lens" of this camera, specifically the part closest to the crash site, called the Inner Tracking System (ITS). Here is the story of how they are building a brand new, ultra-thin, cylindrical tracker called ITS3.

1. The Goal: A Thinner, Closer Lens

Currently, the camera has a bulky lens that sits a bit far from the action. The team wants to replace the three innermost layers of this lens with something much thinner and closer to the collision point.

• The Analogy: Think of the old detector as a thick winter coat. It protects the sensors but blocks some of the view. The new detector is like a single, ultra-thin silk sheet. By making it thinner, the "view" becomes clearer, allowing scientists to see the tiniest details of the particle crashes with twice the precision.

2. The Material: Bending Silicon Like Paper

The biggest challenge is that silicon, the material used to make computer chips, is usually hard and brittle. If you try to bend it, it snaps.

• The Innovation: The team figured out how to shave silicon down until it is only 50 micrometers thick (about half the width of a human hair). At this thickness, the silicon becomes flexible, like a piece of paper.
• The Result: They can now wrap this silicon around the central pipe like a cylinder, creating the world's first "truly cylindrical" tracker. They tested this by bending the chips and shooting electrons at them; the chips survived the bend and kept working perfectly.

3. The Size: Stitching a Giant Puzzle

Standard computer chips are small, like postage stamps. But to cover the entire cylinder, the ALICE team needs sensors that are huge—up to 27 centimeters long (about the length of a ruler).

• The Problem: You can't print a chip that big in one go because the "printing plate" (called a reticle) used in factories is too small.
• The Solution: They invented a "stitching" technique. Imagine tiling a floor where you have to lay down small tiles to make a giant mural. They print the pattern in small sections and stitch them together on the silicon wafer so precisely that the electrical connections flow seamlessly across the seams.
• The Prototype: They built a "Monolithic Stitched Sensor" (MOSS) that is 26 cm long. It works perfectly, detecting particles with over 99% efficiency, even after being blasted with radiation.

4. Cooling: No Water, Just Air

The old detector needed a complex system of water pipes to keep it cool, which added weight and "clutter" (material) that interfered with the particles.

• The Change: The new design is so light and thin that it doesn't need water. Instead, it uses air cooling.
• The Metaphor: Think of it like a laptop. Old models needed heavy fans and liquid cooling loops. This new sensor is so efficient that a gentle breeze (air flowing at 8 meters per second) is enough to keep it from overheating.
• The Test: They built a model and blew air at it. The sensors stayed cool and didn't wiggle or vibrate enough to ruin the picture.

5. The "Super-Speed" Sensor

Inside these chips, there are tiny pixels that catch the particles. The team improved the design of these pixels to make them faster and better at catching signals.

• The Timing: They tested a special version of the chip to see how fast it could react. It turned out to be incredibly quick, with a time resolution of about 63 picoseconds (that's 63 trillionths of a second).
• The Analogy: If a regular camera shutter opens in a blink, this new sensor opens in the time it takes for a snail to move a microscopic distance. This speed helps them pinpoint exactly when a particle passed through.

6. The Bottom Line

The paper concludes that the ALICE collaboration has successfully proven that:

1. Silicon can be bent into a cylinder without breaking.
2. Huge sensors can be "stitched" together from smaller pieces.
3. Air cooling is sufficient to keep the system stable.
4. The sensors are incredibly efficient and fast.

This new ITS3 detector is ready to be installed during the next long shutdown of the LHC (2026–2030), promising to give scientists the sharpest, clearest view of the subatomic world ever achieved.

Technical Summary

Technical Summary: The New Truly Cylindrical Tracker for the ALICE ITS3

Problem Statement
The ALICE collaboration is preparing a major upgrade to its Inner Tracking System (ITS) for installation during the LHC's third long shutdown (LS3, 2026–2030). The primary objective is to significantly improve momentum resolution, particularly for particles with momenta below 1 GeV/c. The current ITS2 relies on a water-cooled system with a material budget of approximately 0.36% $X_0$ per layer. To achieve the required performance gains, the collaboration proposes replacing the three innermost layers with a "truly cylindrical" tracker (ITS3) composed of large-area, bent silicon sensors. This design necessitates overcoming several critical challenges:

1. Mechanical Feasibility: Demonstrating that thin silicon sensors (targeting 50 µm thickness) can be bent to a radius of 19 mm without breaking or losing functionality.
2. Thermal Management: Replacing water cooling with air cooling to reduce material, while ensuring the sensors remain mechanically stable against airflow and maintain thermal uniformity despite non-uniform heat dissipation from readout electronics.
3. Sensor Fabrication: Transitioning from the established 180 nm CMOS technology to a 65 nm process to utilize 300 mm wafers, and overcoming the standard lithographic mask size limit (approx. $2 \times 3$ cm²) to create sensors with active areas up to 27 cm in length.

Methodology
The R&D program addressed these challenges through a multi-stage validation process:

• Bending Validation: Thin silicon chips were subjected to mechanical stress tests to determine breaking points. Reference ALPIDE sensors were bent along both short and long edges to radii of 18–22 mm. Performance was evaluated using electron beams (5.4 GeV) and test beams at DESY and CERN SPS, comparing bent sensors against flat references.
• Technology Migration and Sensor Design: The sensor design evolved from the ALPIDE sensor to a 65 nm CMOS process. Three variants were developed:
  • Standard: Baseline design.
  • Modified: Introduced a low-dose n-type implant beneath the collection electrode to improve charge collection speed.
  • Modified-with-gap: Added a gap at pixel borders to optimize lateral electric fields.
    Small-scale test structures (MLR1) were fabricated to validate pixel pitches (10–25 µm) and doping profiles. Specific structures included Digital Pixel Test Structures (DPTS) for efficiency/fake-hit testing and Analogue Pixel Test Structures with Operational Amplifiers (APTS-OA) to isolate sensor timing performance from front-end electronics.
• Large-Area Stitching: To exceed reticle limits, a "stitching" technique was employed to replicate mask patterns across the wafer while maintaining electrical continuity. A large-area prototype, the Monolithic Stitched Sensor (MOSS), was developed (1.4 × 25.9 cm², 6.7 megapixels).
• Mechanical and Thermal Engineering: Carbon foam structures (K9 for thermal conductivity, RVC for mechanical support) were designed to support the sensors at the edges. Simulations and experimental mock-ups were used to determine the airflow velocity required to maintain thermal gradients below 10 K and to measure sensor displacement under airflow.

Key Contributions and Results

• Bending Feasibility: Silicon chips with thicknesses below 50 µm were successfully bent to the target radii without fracture. Bent ALPIDE sensors maintained a detection efficiency greater than 99% (inefficiency < $10^{-4}$) for thresholds below 100 $e^-$, even at varying incident angles.
• Sensor Performance:
  • Resolution and Efficiency: The modified-with-gap sensor variant achieved a spatial resolution of approximately 5 µm. Detection efficiency exceeded 99% with a fake hit rate lower than $10^{-6}$ hits/pixel/event.
  • Radiation Tolerance: Sensors maintained performance after irradiation up to $4 \times 10^{12}$ $1\text{MeV}_{\text{neq}}/\text{cm}^2$ (NIEL) and 4 kGy (TID).
  • Timing: The APTS-OA structure, utilizing an on-chip Operational Amplifier to preserve signal edges, demonstrated a time resolution of 63 ps on average. For tracks passing directly under the electrode, the resolution improved to 50 ps.
• Large-Area Prototyping: The MOSS prototype successfully demonstrated the feasibility of manufacturing sensors with active areas >90% and lengths of 26 cm. It met design goals for efficiency and fake hit rates post-irradiation.
• Thermal and Mechanical Stability: Simulations and mock-up tests indicated that an air flux of 8 m/s is sufficient to maintain a temperature gradient of less than 10 K along the detector. Under these conditions, the displacement of the outer layer was measured at 1 µm peak-to-peak, which is negligible compared to the sensor's spatial resolution.

Significance
The paper claims that the ITS3 upgrade represents the first implementation of a truly cylindrical tracking detector using wafer-scale Monolithic Active Pixel Sensors (MAPS). The successful R&D validates the transition to a 65 nm CMOS process and the use of stitching to create large-area sensors. By replacing water cooling with air cooling and utilizing a self-supporting carbon foam structure, the material budget per layer is reduced from ~0.36% $X_0$ (ITS2) to 0.09% $X_0$. This reduction, combined with moving the first tracker layer closer to the interaction point (from 22.4 mm to 19.0 mm), is projected to improve impact parameter resolution by a factor of two. The paper concludes that the technology is mature, with the first stitched sensors tested and the final prototype currently in production for deployment in LHC Run 4.