ALICE ITS Upgrade for LHC Run 3: Commissioning in the Laboratory
AALICE ITS Upgrade for LHC Run 3: Commissioning inthe Laboratory D omenico C olella for the ALICE Collaboration Wigner Research Center for Physics, Konkoly-Thege Mikl´os ´ut 29-33, 1121 Budapest, HungaryE-mail: [email protected] (Received November 3, 2020)ALICE is the CERN LHC experiment optimised for the study of the strongly interacting matterproduced in heavy-ion collisions and devoted to the characterisation of the quark-gluon plasma.To achieve the physics program for LHC Run 3, a major upgrade of the experimental apparatusis ongoing. A key element of the upgrade is the substitution of the Inner Tracking System (ITS)with a completely new silicon-based detector (ITS 2) whose features will allow the reconstructionof rare physics channels not accessible with the previous layout. The enabling technology for sucha performance boost is the adoption of custom-designed CMOS Monolithic Active Pixel Sensor asdetecting elements. In this proceedings, an overview of the adopted technologies as well as the statusof the detector commissioning will be given.
KEYWORDS:
Tracking System, MAPS
1. ALICE Experiment upgrade
ALICE (A Large Ion Collider Experiment) [1] is a general-purpose, heavy-ion experiment atthe CERN LHC. Its main goal is to study the physics properties of the quark-gluon plasma (QGP).During LHC Run 1 and Run 2, the analysis of the data collected in heavy-ion collisions allowed theobservation of hot hadronic matter at unprecedented values of temperatures, densities and volumes.These studies confirmed the basic picture, emerged from the experimental investigation at lower ener-gies, of a QGP as an almost inviscid liquid. Moreover, precision and kinematic reach of all significantprobes of the QGP were extended with respect to previous measurements. The study of the strongly-interacting state of matter in the second generation of LHC heavy-ion studies in Run 3 and Run 4will focus on rare processes such as production of heavy-flavour particles, quarkonium states, realand virtual photons and heavy nuclear states [2]. The earlier methods of triggering will be limited formany of these measurements, particularly at low- p T . Therefore, the ALICE collaboration planned toupgrade the Run 1 / Run 2 detector by enhancing its low-momentum vertexing and tracking capabil-ity, allowing data taking at substantially higher rates and preserving the already remarkable particleidentification capabilities. The upgraded experimental apparatus is designed to readout all Pb–Pb in-teractions, accumulating events corresponding to an integrated luminosity of more than 10 nb − . Thisminimum-bias data sample will provide an increase in statistics by about a factor 100 with respectto data collected during Run 1 / Run 2 data taking. In summary, the ALICE upgrade consists of thefollowing sub-system upgrades: • Reduction of the beam-pipe radius from 29.8 mm to 19.2 mm. • Installation of two new high-resolution, high-granularity, low material budget silicon trackers: – Inner Tracking System (ITS 2) [3] in the central pseudo-rapidity ( − . < η < . – Muon Forward Tracker [4] covering forward pseudo-rapidity ( − . < η < − . a r X i v : . [ phy s i c s . i n s - d e t ] N ov Replacement of the endcap wire chambers of the Time Projection Chamber by GEM detectorsand installation of new readout electronics allowing continuous readout [5]. • Upgrades of the forward trigger detectors and the Zero Degree Calorimeter [6]. • Upgrades of the readout electronics of the Transition Radiation Detector, Time-Of-Flight, Pho-ton Spectrometer and Muon Spectrometer for high rate operation [6]. • Upgrades of online and o ffl ine systems (O project) [7] in order to cope with the expected datavolume.
2. Inner Tracking System during Run 3
The main goals of the ITS upgrade (ITS 2) are to achieve an improved reconstruction of theprimary vertex as well as decay vertices originating from heavy-flavour hadrons, and an improvedperformance for the detection of low- p T particles. The design objectives are to improve the impact-parameter resolution by a factor of 3 and 5 in the r φ - and z-coordinate, respectively, at a p T of500 MeV / c [3]. The tracking e ffi ciency and the p T resolution at low p T will also improve. Additio-nally, the readout rate will be increased to 50 kHz in Pb–Pb and 400 kHz in pp collisions. In order toachieve this performances, the following measures were taken: • Granularity increased by an additional seventh layer and by equipping all layers with pixel sen-sors having cell size of 29.24 µ m × µ m. • Innermost detector layer moved closer to the interaction point, from 39 mm to 22.4 mm. • Material budget reduced down to 0.3% X per layer for the innermost layers and to 1.0 % X forthe outer layers.A schematic representation of the ITS 2 is shown in Fig. 1. Based on geometrical position and fewmechanical characteristics, detector can be pictured as assembled in two parts: the innermost threelayers form the Inner Barrel (IB); the middle two and the outermost two layers form the Outer Barrel(OB). Layer radial positions (listed in table in Fig. 1) were optimised to achieve the best combinedperformance in terms of pointing resolution, p T resolution and tracking e ffi ciency in the expected hightrack-density environment of a Pb–Pb collision. The detector pseudo-rapidity coverage is | η | < . andcontaining about 12.5 × pixels with binary readout. It is operated at room temperature, whichis maintained between 20 ◦ C and 30 ◦ C using water cooling. The radiation load at the innermostlayer is expected to be 270 krad of Total Ionising Dose (TID) and 1.7 × eq / cm of Non-Ionising Energy Loss (NIEL). In order to meet the material budget requirements, the silicon sensorsare thinned down to 50 µ m and to 100 µ m in the IB and OB, respectively. The ITS 2 will be the first large-area silicon tracker based on the CMOS Monolithic ActivePixel Sensor (MAPS) technology operating at a collider. The ALPIDE, the pixel chip developed forthe ITS 2, has been designed to fulfil the requirements of both the IB and OB. It is manufacturedusing the TowerJazz 180 nm CMOS Imaging Sensor Process [8], has dimensions 15 ×
30 mm andhosts 512 × ff er. The pixels are arranged in double columns and read out by a priority encoder which sendsthe addresses of the pixels that recorded a hit in the chip peripheral readout circuitry [9]. Pixels withno hits are not read out, making the readout procedure faster and reducing the power consumption,which for every pixel is about 40 nW. In triggered mode, a short strobe of the order 100 ns is usedto store the hits in the in-pixel memories. In continuous-integration mode, the strobe is active forthe full integration window (of the order 10 µ s) except for the switch over (about 100 ns) from one ig. 1. (Left) Schematic layout of the upgraded ITS. (Right) Summary table containing the detector layersgeometric characteristics. to the other in-pixel hit bu ff er. There are several parameters, adjustable via on-chip 8-bit digital-to-analogue converters (DAC), that allow fine tuning of signal shaping and gain. The largest impact onthe discrimination threshold is obtained by adjusting the parameter current threshold, I T HR . Detectione ffi ciency has been measured as a function of this parameter, and found to be above the required99% for a large range of values (30 to 140 e − ) [10]. In the same range, the fake-hit rate stays at thesensitivity limit (10 − hits / pixel / event), after masking 0.015% of the pixels. After NIEL irradiationup to the ten times nominal value, the detection e ffi ciency drops below 99% at about 120 e − , whilethe fake-hit rate is not influenced by NIEL irradiation. The TID irradiation ( ∼
350 krad) leads to ane ff ective change of charge threshold of the front-end circuitry resulting in a higher fake-hit rate atlow I T HR and detection e ffi ciency close to 100% over the full range. In laboratory it was verified thata second internal DAC parameter, V CAS N , allows to compensate for the above mentioned e ff ectivechange of charge threshold. As a consequence, the operational range in terms of detection e ffi ciencyand fake-hit rate is not reduced.The basic element of a layer is the stave, which consists of a carbon space frame, supporting thechips, to which the cold plate and the cooling ducts are attached. Pixel chips (9 for the IB and 14 forthe OB) glued to a Flexible Printed Circuits (FPC) constitute the Hybrid Integrated Circuit (HIC). TheHIC(s) are then glued to the space frame: 1 HIC for the IB and 8 or 14 HICs, actually subdivided intwo adjacent half-staves, for the two innermost and the two outermost layers of the OB, respectively.The FPC consists of a polyimide with a low thermal expansion coe ffi cient plus aluminium and copperas conductor for the IB and OB, respectively. The chips are electrically connected to the FPC, forpowering and data stream, using wire bonding to pads distributed over the entire chip surface. Apower bus distributes the power to the HICs on the OB (half-)stave.The data readout chain is segmented in three stages. The digital periphery of the ALPIDE chipproduces the digitized and zero-suppressed hit data and sends them to the Readout Units (RUs)through high-speed serial data links running at either 1.2 Gbps for the IB sensors or 400 Mbps forthe OB sensors. Chip periphery interfaces also with a clock input, and a bi-directional control busfor configuration, control, and monitoring. The RUs, FPGA-based boards [11], are located outsidethe detector acceptance in a moderate radiation environment of less than 10 krad of TID. The RUsreceive the triggers from the Central Trigger Processor (CTP) via optical link and ship the data to theCommon Readout Units housed in the First Level Processor computers [7], located in control rooms,where the event reconstruction starts. The same data path is used to control the detector. Power to thestave is provided through a Power Board (PB), generating two 1.8 V supply voltages (for the analogand digital circuits of the pixel chip) and a negative voltage output for the reverse bias. The RUs andPUs are powered by a CAEN distribution system based on EASY3000 crates. ig. 2. Picture of the CERN clean room where the detector is seating while in commissioning at the surface,before installation at LHC P2 cavern. Each of the four half-barrel is connected to the corresponding crate,hosting the power and readout units.
The electrical and functional tests of the ALPIDE chips were performed at CERN (CH) for the IB50 µ m thinned chips and at Yonsei and Pusan / Inha (KR) for the OB 100 µ m thinned chips, betweenSeptember 2017 and May 2019. A total of almost 4 × chips were tested with a detector-gradequality yield of 63.7%.The full production of HICs and staves for the IB was carried out at CERN and completed inmiddle of 2019. A total of 95 staves, enough to build two copies of the three inner barrel layers, wereassembled with a yield of 73%. The OB HICs were produced in five sites (Bari (IT), Liverpool (UK),Pusan / Inha, Strasbourg (FR) and Wuhan(CN)), while the FPCs were tested in Trieste and Catania(IT). It took 80 weeks to assembly the needed 2500 HICs; this quantity includes the HICs neededto cover the whole detector acceptance plus spares and assumes an overall production yield of 74%(convolution of 82% for the HIC and 90% for the stave). The HIC production was completed duringSeptember 2019, achieving a yield of 84%. The OB staves were assembled in five sites (Berkeley(US), Daresbury (UK), Frascati (IT), NIKHEF (NL) and Torino (IT)) by October 2019, reaching ayield close to 90% for enough objects, qualified as detector grade, to cover the full OB acceptance.Additional spare staves were also assembled. Accelerated ageing tests both for HICs and staves,verified the assembly procedure and validated the use of the appropriate components [12].Production and qualification of the full set of 192 RUs (CERN, Bergen (NO), NIKHEF) and 142PBs (Berkeley), plus spares, were completed by mid 2019.A large clean room (Fig. 2) was built at CERN to allow the full detector assembly and the on-surface commissioning activities, before the installation in the ALICE cavern starting from January2021. Here the same backend system that will be used in the experiment is available, including pow- ring system, cooling system, full readout and trigger chains. Integration of the staves in the layerstructure and connection to the services progressed with staves availability at CERN. Layers are as-sembled in half-barrels and are currently divided in 4 parts: two (top and bottom) of the IB (layers 0,1and 2) and two (top and bottom) of the OB (layers 3,4,5 and 6). Each of them is connected to a cratehosting RUs and PBs. First completely assembled and connected half-barrel has been the IB top byJuly 2019; last one has been the OB bottom by January 2020. Installation of the di ff erent detector components in the clean room was a long process. Commis-sioning activities become more and more specific with the availability of hardware and developmentof firmware / software, starting from basic functionality verifications to detailed and quantitative mea-surements of system performance. First studies of the behaviour of staves assembled in layer usingfinal powering and readout schema were performed using a lower quality half-layer 0 (Debug Layer).Continuous operation of the detector started in May 2019, supported by crews of two shifters alter-nating along the day every 8 hours. Coordinated by shift leader, the shift crew collects the data, assessits quality and monitors the status of the detector. Standard data taking schedule foresees hourly exe-cution of a threshold scan, a fake-hit rate run and a readout test.As anticipated, the charge threshold for pixel hit generation can be modified acting on two inter-nal DACs (I T HR and V
CAS N ) [10]. Dedicated calibration runs are executed every time the poweringschema of the chips is changed. The threshold value is always a trade-o ff between pixel e ffi ciencyand amount of noise. It has been measured that a good target threshold value for calibration is 100 e − .The threshold scan allows to verify the correctness and uniformity over the pixel chip matrix of thedetector calibration. A fake-hit rate run allows to measure the amount of noise. A readout test is usedto verify the quality of the data transmission between the chips and the RUs.Most of the activities of the shifters have been performed on the di ff erent IB pieces available withtime, starting from the Debug Layer to the final full IB (top and bottom parts). OB took longer forthe final assembly and required more time for the complete integration in the system, due to largernumber of elements. Most of the verifications performed on the OB staves have been done by experts.First versions of Detector Control System (DCS) and Data Acquisition System (DAQ) as wellas Quality Control (QC) and track reconstruction softwares are available and running on machineshoused in a control room adjacent to the clean room. One of the first tasks was the verification of the threshold calibration procedure. E ff ectiveness ofthe DAC tuning can be determined by looking at the change of the pixel threshold distribution beforeand after the calibration step as in Fig. 3. As anticipated, target threshold value from the calibrationis 100 e − ; values reported in the plots are expressed in DAC units, and the conversion factor to e − is10. The level of homogeneity reached between the di ff erent chips in the layer can also be seen fromthe same figure. The large cumulated statistics of more than 5000 runs, recorded over tens of hoursof running, have confirmed that the threshold value after calibration remained stable in time. This hasbeen verified both with back-bias supply voltage equal to 0 V and -3 V.During fake-hit rate runs, random triggers are sent to the chips and all the hits created in coinci-dence are collected. These hits come from the noise and also from cosmic rays crossing the detectorin coincidence with the trigger. In the left plot of Fig. 4, the measured fake-hit rate is plotted as afunction of the threshold value for di ff erent numbers of noisier masked pixels. It can be seen that,for the target threshold value of 100 e − , masking a total of 10000 pixels, less than the 0.009% of theconsidered acceptance, a fake-hit rate value of the order of 10 − hits / pixel / event can be achieved.This is an important result, considering the really low fraction of masked acceptance and the fact thatthe target value for the fake-hit rate was 10 − hits / pixel / event [3]. The result is consistent with what ig. 3. Internal DAC tuning e ff ectiveness: pixel threshold maps (left) and threshold value distribution (right)before and after the tuning. Threshold values are expressed in DAC units; conversion factor to e − is 10. Resultsobtained for the Layer 2 top (L2T). was observed with single chip characterisation.Due to the very low recorded chip noise, hits coming from cosmic particles passing through thedetector can be clustered together and used to reconstruct the associated tracks. Starting from tracksone can perform an alignment study to determine the displacement of staves in layers. Preliminaryanalysis gives residuals of the order of 100 µ m. The code developed using this alignment study canalso be used to determine the final alignment, once the ITS 2 will be installed in the cavern, withinthe ALICE detector. The overlap of the staves at the edges facilitate the reconstruction of a track from6 aligned clusters in three layers, as seen in the right plot of Fig. 4.Quality selection criterion to consider a stave good for installation in the IB was to have a totalamount of not working pixels below 50 over the 9 chips. After installation, the total number of pixelsfound to be dead were less than 1000 out of a total 226 × pixels of the full IB. This number isactually reduced with respect to the measurements done on the staves before installation in the barrel,thanks to improved stuck pixel identification and masking mechanism.Readout tests are used to verify the quality of the data transmission from the chips to the associ-ated RU performing a measurement of the Bit Error Rate (BER) through counting of 8b10b decodingerrors. During the test, a fixed number of pixels are stimulated and the chip matrix is read out ata given rate. Multiple injected patterns (256, 512, 1008, 2048 and 4096 pixels) and readout rates(44.9 kHz, 67.3 kHz, 101 kHz, 123 kHz, 202 kHz and 247 kHz) were explored. From [3], a total of ∼
250 activated pixels are expected in case of 100 kHz Pb–Pb minimum-bias collisions, adding QEDbackground and pixel noise of the order of 10 − hits / pixel / event, for an integration time of 30 µ s. Sim-ulating an occupancy similar to the one described, a BER sensitivity of about 10 − was measured forall the readout rates mentioned above. The BER sensitivity was measured to be of the order of 10 − in a large statistics sample for the readout rate of 44.9 kHz. In a few extreme conditions, actuallynot corresponding to possible scenarios during data taking with beams, errors have been recorded ona limited amount of staves and with a rate of error of the order of ∼
20 hours. Exploration of theseextreme conditions as well as repetition of campaign with back-bias voltage at -3 V and investigationof boundary behaviour varying the digital supply voltage at the chip are in progress at the moment. ig. 4. (Left) Fake-hit rate as a function of the threshold value for di ff erent numbers of masked pixels.Results obtained for the full IB top. (Right) Cluster correlation distribution from cosmics tracks. In the eventdisplay, an example of tracks reconstructed from 6 aligned hits, obtained thanks to the overlap of the chips atthe edges. At the arrival at CERN, before the final installation in the detector, each OB stave went throughbasic tests consisting of verification of the power consumption, the control communication and thehigh speed links.Handling of very long and fragile objects, as the staves during the assembly of the layers and half-barrel, required the definition of a rigorous and yet delicate procedure. In spite of the complexity of thestave installation procedure, there was no increase in the number of dead chips, as ascertained fromelectrical and functional checks. The total number of dead chips over the full OB is 32, correspondingto the 0.14% of the acceptance; no overlap in the radial direction is present between the dead chips inthe di ff erent layers.A calibration campaign allowed to characterise the OB staves, and showed that a tuning to 100 e − threshold target with a 2 e − precision is possible, having a chip-to-chip spread within the stave of20 e − . An example of threshold map for a full OB stave, containing 196 chips, is shown in top plot ofFig. 5.Due to the very large amount of chips installed in the OB staves ( ∼ × chips), the selectioncriteria were released with respect to the chip used in the IB. As a consequence we do expect reducedperformance in terms of noise. Nonetheless a preliminary study of the fake-hit rate as a function ofthe number of the noisier masked pixel, as done for the IB, demonstrated that a fake-hit rate value ofthe order of 10 − hits / pixel / event is reachable masking up to the 0.014% of the full OB acceptance.This value is well below the requirement of fake-hit rate to be less than 10 − hits / pixel / event [3]. Anexample of noise map for an OB stave, with points enlarged by a factor 50 to make them visible, isshown in bottom plot of Fig. 5.A long term continuous powering campaign, to bring out events of early infant dead of staves andany service components (e.g. power supply modules), was carried out and no suspicious cases havebeen found. This campaign required the improvement of the monitoring processes of a lot of detectorcomponents.Next steps in the OB staves commissioning, foresee a chip to RU communication verificationcampaign, as done for the IB, and collection of tracks from cosmic muons to allow alignment study.The first data collected using cosmic rays was used to improve the code for the track reconstruction. ig. 5. Outer Barrel stave threshold map (top) and fake-hit rate map (bottom). Colour scale for the thresholdvalue is reported on the right. Points in the fake-hit rate map are enlarged by a factor of 50 to make them visible.
3. Conclusions
The ongoing replacement of the ITS will extend the physics reach of ALICE to lower transversemomentum, allowing the characterisation of the QGP via measurements of unprecedented precision.The new ITS is now fully constructed after a huge e ff ort. The working of readout chain for both IBand OB are demonstrated. IB is fully characterised and boundary limits are under exploration. Thecharacterisation of the OB is ongoing and a campaign to gather particle tracks from cosmic muons isready to begin. The commissioning at the surface will continue until the end of 2020. In the beginningof 2021 the detector will be prepared to be transferred to ALICE P2. The installation in the cavern,testing of functionalities and other procedures will take about two months. After that, ITS 2 will jointhe other detectors in the global ALICE commissioning, in order to be ready to take the very first datawith collisions in the foreseen LHC pilot run during October 2021. References [1] The ALICE Collaboration et al. , JINST S08002 (2008).[2] The ALICE Collaboration et al. , J. Phys. G et al. , J. Phys. G et al. , CERN-LHCC-2015-001 (2015).[5] The ALICE Collaboration et al. , CERN-LHCC-2013-020 (2013).[6] The ALICE Collaboration et al. , CERN-LHCC-2013-019 (2013).[7] The ALICE Collaboration et al. et al. , JINST C02042 (2016).[10] M. ˇSulji´c (on behalf of the ALICE collaboration), in proceedings of theInternational Workshop on Semi-conductor Pixel Detectors for Particles and Imaging (Pixel 2016), JINST C11025 (2016).[11] J. Schambach et al. , 2018 IEEE Nuclear Science Symposium and Medical Imaging Conference Proceed-ings (NSS / MIC) pp. 1 − //