The ALICE experiment has successfully resumed physics data-taking in RUN 3 following a major upgrade of its detectors and data acquisition system during the LHC Long Shutdown 2 (LS2). While a portion of the delivered luminosity in 2022 was allocated to commissioning the upgraded experiment, ALICE achieved a groundbreaking milestone over the next two years, collecting an integrated luminosity of 75 pb⁻¹ (as of this article), representing more than five trillion minimum-bias proton-proton collisions stored on disk for offline trigger selection. This data sample already exceeds the total collected during RUN 1+2, validating the effectiveness of the upgrade strategy implemented during LS2.

Figure 1: Integrated luminosity recorded by ALICE in pp collisions at a centre-of-mass energy of 13.6 TeV.

The ALICE upgrades during LS2 introduced a complete shift in the data-taking paradigm, with a new continuous readout of the detector signals replacing the triggered readout of RUN 1+2. In this continuous readout mode, proton-proton collisions at a 650 kHz interaction rate are fully recorded, producing more than 200 GB/s of raw data. This input rate is reconstructed online and compressed to 30 GB/s thanks to the new data processing farm, consisting of 2800 graphics processing units (GPUs) and 50,000 central processing unit (CPU) cores. At such a rate, the ALICE experiment typically collects 10 petabytes of compressed data per week, which are sent to the CERN Data Centre for disk storage. Such a large amount of data cannot be fully kept on disk for long-term storage. In the case of proton-proton collisions, the compressed data undergoes a cycle of calibration, offline reconstruction, and quality assessment, which is completed a few weeks after the data is recorded. A small percentage of these reconstructed data are then selected based on software triggers designed for physics analysis. After a final quality check of the selected data, the originally compressed raw data set is permanently deleted.

At the end of 2023, the LHC delivered the first high-intensity lead beams of RUN 3. During this five-week heavy-ion run, ALICE recorded about 12 billion minimum-bias lead–lead collisions (see Figure 2 left)—40 times more than the total recorded by ALICE during previous periods of heavy-ion data-taking from 2010 to 2018. In lead-lead collisions, the number of reconstructed particles in a single event is considerably larger than in proton-proton collisions (as illustrated in the event display of Figure 2 right), resulting in a rate of up to 770 GB/s of raw data and 170 GB/s of compressed data. The total size of the compressed events from the 2023 PbPb run amounts to almost 30 PB, which is stored on tape for long-term preservation.

Fig 2: Left: Integrated luminosity recorded by ALICE for Pb-Pb collisions at √sNN = 5.36 TeV Right: Event display of a Pb-Pb collision.

The start of heavy-ion data taking proved challenging for both the ALICE and LHC teams. The first lead-lead collisions revealed a large, unexpected beam-related background in the upgraded Inner Tracking System (ITS2) of ALICE, leading to a significant loss of efficiency localized on the first layers of the detector. The issue was promptly identified within the first minute of data acquisition by the ALICE online quality control system and immediately addressed by the LHC team. The new ITS2 is positioned significantly closer to the interaction point (between 2.2 cm and 2.4 cm) than the ITS used in RUN 2 (3.9 cm for the Silicon Pixel Detector). It quickly became clear that the background was caused by a flux of particles in this innermost region, which was previously not instrumented. Through close collaboration between machine experts and the ALICE experiment, various mitigation strategies were explored over more than 30 hours of beam time until satisfactory machine settings were found that reduced the background in ALICE to an acceptable level for the physics program.

The large RUN 3 data sample of lead-lead and proton-proton collisions, combined with the excellent tracking and pointing performance of the upgraded ALICE detectors, will enable the exploration of uncharted territories in the study of quark-gluon plasma by providing precise measurements of rare observables that were previously inaccessible.

This is illustrated, for example, by the measurement of the relative production rates of two charmed baryon resonances Σ_c⁰,⁺⁺(2520), which has been measured for the first time at the LHC using the RUN 3 ALICE proton-proton data sample (see Figure 3). In lead-lead collisions, the unprecedentedly large data sample collected in 2023 already allows for an accurate measurement of a rare observable like the elliptic flow of anti-³He, as illustrated in Figure 4.

Figure 3: Invariant mass of Σ_c⁰,⁺⁺(2455) and Σ_c⁰,⁺⁺(2520) decay candidates in pp collisions at √s = 13.6 TeV.

Figure 4: Elliptic flow v₂ of anti-³He in Pb-Pb collisions at √s_NN = 5.36 TeV, compared to state-of-the-art model predictions.

ALICE is now preparing for a second heavy-ion physics run at the end of 2024. The first few days will be dedicated to collecting a reference sample of pp collisions at the same collision energy as for the lead beams, with the ambitious goal of recording 4.5 pb⁻¹ in approximately four days. This reference pp run will be followed by 17 days of lead-lead collisions, during which ALICE aims to roughly double the 2023 data sample.