ALICE ITS3 WP4 Research

The ALICE (A Large Ion Collider Experiment) at the CERN LHC investigates the properties of the Quark-Gluon Plasma (QGP), the extremely hot and dense state of matter believed to have existed microseconds after the Big Bang. A central handle on QGP is the behaviour of heavy-flavour (Charm and Beauty) quarks that traverse the medium produced in heavy-ion collisions.

Just like throwing a heavy bowling ball into a foggy room to determine the fog's density from its trail, these heavy quarks act as probes that allow us to figure out the properties of the QGP.

However, hadrons containing these heavy quarks have extremely short proper lifetimes (e.g. the D⁰ meson has cτ ≈ 123 μm), so reconstructing their decay vertices separately from the primary interaction vertex demands a precision tracker placed as close as possible to the beam pipe, with as little material as possible.

This role is played by the Inner Tracking System (ITS). During LHC Long Shutdown 3 (2026–2028), the three innermost layers of the present ITS2 (Inner Barrel) will be replaced by ITS3, a next-generation tracker built from wafer-scale, stitched MAPS fabricated in TPSCo 65 nm CMOS technology, thinned to below 50 μm and bent around the beam pipe into truly cylindrical half-layers.

A flat, single piece of silicon is very thin and weak, making it difficult to maintain its shape on its own under the influence of gravity without additional structures, but when rolled into an arch shape, it becomes sturdy. Utilizing this geometric stiffness allows the sensors to keep their cylindrical shape held only by carbon-foam supports and air cooling — without any liquid cooling pipes inside the active acceptance.

Thanks to this, the material budget in the innermost layer is reduced to about 0.05% X₀ per layer, roughly a factor of seven lower than the ITS2 Inner Barrel (~0.35% X₀).

As a result, the multiple scattering of low-momentum particles is strongly suppressed, and the pointing (impact-parameter) resolution is expected to improve by roughly a factor of two compared to ITS2.

Once thin silicon sensors that bend to gain geometric stiffness were devised, the next challenge was how to supply power and retrieve data from them. To solve this, the Flexible Printed Circuit (FPC) serves as the backbone for establishing physical and electrical connections.

As a member of Pusan National University's Heavy Ion Physics Experiment Lab (HIPEx), I participated in the ALICE Collaboration's ITS3 WP4 (Mechanics & Engineering) research. ALPIDE — the 180 nm CMOS MAPS already in service in ITS2 — has a chip area of ≈ 30 × 15 mm² with a 512 × 1024 pixel matrix (524,288 pixels, pitch ≈ 29 × 27 μm). As a precursor to ITS3's next-generation 65 nm wafer-scale stitched MAPS, our work focused on verifying that thinned ALPIDE chips (50 μm and 100 μm) tolerate the mechanical stress of being bent to the ITS3 target radii (18 mm, 24 mm, 30 mm) without measurable degradation in their pixel response.

To achieve this, the first problem to solve was, "How can we precisely bend and secure a paper-thin silicon sensor to the desired radius without breaking it?" I built the test environment by designing 'PNU Guide & Frame', a custom mechanical jig assembly in 3D CAD, and 3D printing it. This allowed us to bend the sensor-FPC assembly safely while keeping the bending radius repeatable across measurement runs.

During a sensor beamtime test at the Korea Multi-purpose Accelerator Complex (KOMAC) proton accelerator in Gyeongju, we observed persistent, unexplained anomalies (spurious hits) in the sensor data.

To diagnose the issue, I focused on the optical properties of silicon. Silicon at room temperature has an indirect bandgap of about 1.12 eV, which corresponds to a photon-absorption cutoff wavelength of about 1107 nm. In other words, near-infrared (Near-IR) photons with wavelength ≲ 1100 nm can be absorbed and generate electron-hole pairs that the pixel sees as ionization signal. Typical surveillance / night-vision IR LED CCTV cameras emit at 850 nm or 940 nm — well inside silicon's absorption window. I therefore hypothesized that the IR illuminator on the camera installed inside the irradiation room was reaching the thinned ALPIDE surface and producing fake hits and distortions in the threshold-scan distribution.

To verify this, I returned to the laboratory and intentionally illuminated the ALPIDE sensor with an IR LED inside a fully light-tight dark room. The same shifts in the threshold distribution and the same increase in fake hits seen at the accelerator were reproduced, confirming the IR-illumination hypothesis. This was a valuable experience of using first-principles physics knowledge to track down a non-obvious systematic effect in the field, rather than treating the anomalies as unexplained noise.

Within the bending radii we tested (18–30 mm), the thinned ALPIDE sensors themselves showed no significant change in noise or threshold between the flat and bent configurations — consistent with the ALICE published result of detection efficiency > 99.9% and spatial resolution ≈ 5 μm, independent of bending radius. Our practical conclusion was that controlling ambient light conditions (especially IR) is highly critical for the measurement itself. The PNU bending campaign was paused when I started my mandatory military service, so the final group-level write-up was completed by other collaborators.