---
title: 'babyMOSS: Prototype for ITS3 Pixel Sensors'
url: https://www.emergentmind.com/topics/babymoss
type: topic
---

# babyMOSS: Prototype for ITS3 Pixel Sensors

babyMOSS is a stitched monolithic pixel-sensor prototype developed within the ALICE ITS3 research-and-development program as a compact demonstrator of the repeated sensor architecture planned for the upgraded Inner Tracking System. It reproduces the functional concept of a single MOSS repeated sensor unit in a form factor that is easier to transport, irradiate, and operate in laboratory and beam-test conditions, while remaining technologically representative of the stitched 65 nm CMOS sensors intended for the ITS3 inner barrel. In beam tests and laboratory characterization, babyMOSS has been used to qualify detection efficiency, fake-hit rate, spatial resolution, and irradiation tolerance against the ITS3 performance targets [2510.01834].

## 1. Position within the ALICE ITS3 upgrade

During Long Shutdown 3, scheduled for 2026–2030, ALICE plans to replace the three innermost ITS2 layers with ITS3, a new low-material vertex detector based on wafer-scale stitched Monolithic Active Pixel Sensors fabricated in a 65 nm CMOS technology. The ITS3 concept uses very large-area monolithic sensors that are thinned and bent into true cylindrical or half-cylindrical layers around the beam pipe, with the three cylindrical layers positioned at 19, 25.2, and 31.5 mm from the beam pipe. The target average per-layer material budget is a reduction from 0.36% \(X_0\) in ITS2 to 0.09% \(X_0\) in ITS3, with additional benefit from air cooling replacing water cooling; this reduction is expected to improve tracking and vertexing, especially at low transverse momentum [2602.03504].

Within that program, stitching is the enabling fabrication technique. Standard lithographic reticle size limits a single die to approximately \(2.5 \times 3~\mathrm{cm}^2\), which is too small for the large seamless cylindrical sensors envisioned for ITS3. Stitching overcomes that limit by repeating a basic building block, the Repeated Sensor Unit, across multiple reticle fields so that a larger monolithic device can be produced. babyMOSS belongs to the Engineering Run 1 prototype family that also includes MOSS and MOST, and it occupies the role of a reduced-scale qualification vehicle for the stitched sensor concept [2510.01834].

## 2. Sensor concept and chip architecture

babyMOSS is described as one MOSS RSU implemented in compact form with end-cap structures for powering and readout. One source phrases this as “one MOSS RSU and a Right End Cap,” while another describes it as “a single MOSS RSU” plus “2 end-caps on the short edges for powering and readout.” Its size is about \(14\times 30~\mathrm{mm^2}\), whereas the full MOSS prototype has an active area of \(26 \times 1.4~\mathrm{cm}^2\) and is formed by ten RSUs plus end caps. The architectural point is that babyMOSS preserves the repeated stitched unit used in the larger sensor family while remaining experimentally manageable [2602.03504].

Each babyMOSS chip contains eight digitally read-out pixel matrices, called regions, arranged in two rows called Half-Units, labeled Top HU and Bottom HU. Each HU contains four regions. The segmentation is intentional: the eight regions implement different front-end electronic flavours, permitting side-by-side comparison on the same die under identical operating conditions. The two HUs also differ in pitch and matrix geometry. The top regions are \(256\times 256\) matrices with \(22.5~\mu\mathrm{m}\) pitch, while the bottom regions are \(320\times 320\) matrices with \(18~\mu\mathrm{m}\) pitch [2510.01834].

| Feature | Top HU | Bottom HU |
|---|---:|---:|
| Regions | 4 | 4 |
| Matrix size | \(256\times256\) | \(320\times320\) |
| Pixel pitch | \(22.5~\mu\mathrm{m}\) | \(18~\mu\mathrm{m}\) |
| Binary resolution | \(6.5~\mu\mathrm{m}\) | \(5.2~\mu\mathrm{m}\) |

A further architectural variable is the low-dose n-type blanket implant beneath the collection electrode, with gaps at the pixel edges. Wafer splits were produced with different gap widths to study design optimization. In the Top HU, split 1 uses \(2.5~\mu\mathrm{m}\) gaps and split 2 uses \(5~\mu\mathrm{m}\) gaps; in the Bottom HU, both splits use \(2.5~\mu\mathrm{m}\) gaps. This matters experimentally because the detailed irradiation comparison is performed on chips drawn from different wafer splits, and the bottom-region comparison is therefore the cleanest like-for-like case [2602.03504].

## 3. Characterization program and beam-test apparatus

The babyMOSS characterization campaign had two components. Laboratory studies were used for systematic functional scans and exploration of front-end behaviour over a range of settings, with the objective of identifying acceptable efficiency/noise working points. The published summaries, however, provide little numerical detail on threshold-equalization strategy, scan variables, power figures, or detailed laboratory outputs. The second component was beam testing, where the observables most relevant to ITS3 tracking performance—detection efficiency, fake-hit rate, spatial resolution, and average cluster size—were measured under high-energy charged-particle beams [2510.01834].

The beam campaign described in detail was carried out at the CERN Proton Synchrotron in September 2024. One account reports a \(10~\mathrm{GeV}\) \(\pi^-\) beam, while another notes that the campaign used charged-pion beams of 7 and 10 GeV/\(c\), with the 7 GeV/\(c\) data arising from an initial beam misconfiguration before later running at the intended 10 GeV/\(c\). The setup was a babyMOSS telescope consisting of six babyMOSS tracking planes, a babyMOSS Device Under Test placed in the middle, and two scintillators for trigger generation; one description adds that each scintillator was coupled to a photomultiplier tube and that the distance between planes was \(2.5~\mathrm{cm}\). Readout used the compact DAQ-Raiser system, composed of a DAQ board and a Raiser board interfacing to the babyMOSS carrier [2602.03504].

Three DUTs were studied. The non-irradiated sample was `babyMOSS-2_1_W22C7` from split 2. Two irradiated samples, `babyMOSS-2_2_W02F4` and `babyMOSS-3_3_W02F4`, were from split 1 and had been pre-irradiated with neutrons to \(10^{13}~1~\mathrm{MeV}~n_{eq}/\mathrm{cm}^2\). That fluence was chosen as an early estimate of the ITS3 load and lies above the quoted ITS3 NIEL requirement of \(4\times10^{12}~1~\mathrm{MeV}~n_{eq}/\mathrm{cm}^2\); the TID requirement quoted in the later paper is \(400~\mathrm{krad}\) [2510.01834].

## 4. Reconstruction workflow and performance observables

The beam-test analysis was performed with Corryvreckan. The reported workflow comprises noisy-pixel masking, clusterization of hits, telescope pre-alignment, alignment with track reconstruction, and final alignment of all planes including the DUT. In the more detailed description, tracking-plane pixels with occupancy above 1000 times the average were masked; for the DUT, noisy-pixel masks were derived from off-beam fake-hit-rate measurements. The coordinate system was the standard right-handed Cartesian system with \(z\) along the beam. Pre-alignment shifted planes in \(x\) and \(y\) to minimize inter-plane correlations, and full alignment was run first without and then with the DUT included [2510.01834].

Track finding began from hit combinations in the first and last tracking plane, with additional associated hits or clusters accepted inside a distance window. The stated association windows were \(200~\mu\mathrm{m}\) for alignment without the DUT and \(300~\mu\mathrm{m}\) for alignment with the DUT. Good tracks were required to have at least seven associated hits, i.e. one per plane. In the DUT analysis, hits were associated within a \(100~\mu\mathrm{m}\) window. The later paper defines detection efficiency in words as the ratio of tracks with an associated DUT hit to the total number of tracks, and it states explicitly that the fake-hit rate was measured with the beam off, in units of hits/pixel/event [2602.03504].

For spatial resolution, the later paper gives the extraction formula explicitly:
\[
\sigma_{i,DUT} = \sqrt{\sigma_{i,fit}^2 - \sigma_{telescope}^2} \quad i=x,y
\]
where \(\sigma_{i,fit}\) is obtained from Gaussian fits to residual distributions and \(\sigma_{telescope}\) is the telescope tracking precision computed with the Telescope Optimizer tool. The reported spatial resolution is then taken as the average of the extracted \(x\)- and \(y\)-direction resolutions. Threshold scans were performed in DAC units, but no DAC-to-electron calibration is provided, and no explicit formulas are given for fake-hit rate, cluster-position reconstruction, threshold extraction, noise, or charge calibration [2602.03504].

## 5. Beam-test results and irradiation response

The central result is that babyMOSS satisfies the ITS3 sensor requirements. The required operating point is
\[
\varepsilon > 99\%, \qquad \mathrm{FHR} < 10^{-6}\ \text{hits/pixel/event}, \qquad \sigma_{\text{spatial}} < 6~\mu\mathrm{m},
\]
and the beam tests found an operating margin in all investigated regions where detection efficiency exceeds 99% while the fake-hit rate remains below \(10^{-6}\) hits per pixel per event. The later paper summarizes the achieved spatial resolution as around \(5~\mu\mathrm{m}\), and the earlier one states that the measured resolution in the compared bottom regions remains below the \(5.2~\mu\mathrm{m}\) binary limit; both formulations are below the ITS3 requirement [2510.01834].

The results hold for both non-irradiated and irradiated devices. For the non-irradiated device, all four regions in both HUs exhibit an operational window satisfying the efficiency and fake-hit criteria. For the irradiated device, the same conclusion holds and a sufficient operational margin is preserved even after \(10^{13}~1~\mathrm{MeV}~n_{eq}/\mathrm{cm}^2\). The detailed comparison concentrates on the bottom regions because the top-region designs differ between split 1 and split 2, whereas the bottom-region design is common to both [2602.03504].

Irradiation does alter the margin and the threshold dependence. The papers state that detection-efficiency curves are comparable between irradiated and non-irradiated DUTs, but the irradiated chip has a somewhat narrower operating window. At low thresholds, below about 20 DAC, fake-hit-rate behaviour remains similar; above 20 DAC, the irradiated DUT shows measurable fake-hit rate, reaching approximately \(10^{-8}\) hits per pixel per event, while no fake-hit noise was detected in the non-irradiated babyMOSS. The reported interpretation is increased leakage current after irradiation [2510.01834].

Clustering and resolution trends are consistent with that picture. At low threshold, charge sharing increases, average cluster size rises, and spatial resolution improves. After irradiation, the average cluster size decreases and the spatial resolution becomes slightly worse at very low thresholds, which is interpreted as a consequence of non-ionizing radiation damage creating trapping centers in the silicon and reducing charge sharing or charge collection. At high thresholds, no appreciable differences in spatial resolution and cluster size are observed between irradiated and non-irradiated sensors [2602.03504].

## 6. Relation to full MOSS, significance, and limits of the present evidence

babyMOSS is not the final ITS3 sensor and not a direct bent full-wafer demonstrator. Its function is to validate, under realistic operating and irradiation conditions, the repeated stitched building block that underlies MOSS and, by extension, the large stitched sensors planned for ITS3. In that specific sense, the most important result is consistency: the published characterization states that babyMOSS beam-test results are consistent with full MOSS and confirm that babyMOSS devices meet the ITS3 requirements [2510.01834].

That consistency is significant because ITS3 ultimately depends on wafer-scale stitched MAPS that will be thinned to about \(50~\mu\mathrm{m}\) and bent into true cylinders. A compact one-RSU demonstrator cannot by itself prove full-detector behaviour, but it directly tests the RSU-based architecture, stitched 65 nm CMOS technology, and post-irradiation operating margin on which the larger concept depends. This suggests that babyMOSS functions as a de-risking stage between earlier small 65 nm test structures and full stitched sensors, rather than as an isolated prototype [2602.03504].

The published scope also has clear boundaries. The papers do not provide a full transistor-level or state-machine-level description of the readout logic, do not give a detailed threshold tuning or calibration algorithm, and do not report bias-voltage or incidence-angle scans. They do not provide numerical laboratory-scan results, power-consumption values, or babyMOSS yield numbers. They also do not discuss seam effects, edge effects, dead/noisy macroscopic regions, or stitching-boundary artifacts in detail, and they do not provide specific defect maps. Accordingly, the absence of such effects should not be inferred from the present reports; rather, those issues were not the focus of the characterization contributions [2510.01834].

A recurrent misconception is to treat babyMOSS as interchangeable with the final ITS3 sensor. The published evidence supports a narrower and technically more precise statement: babyMOSS is a compact one-RSU demonstrator of the stitched MOSS architecture, and its measured performance supports the viability of the stitched 65 nm monolithic sensor concept for the ALICE ITS3 inner barrel, including operation after irradiation beyond the nominal ITS3 NIEL requirement [2602.03504].

Source: https://www.emergentmind.com/topics/babymoss