The CMS High-Granularity Calorimeter for Operation at the High-Luminosity LHC

The High Luminosity LHC (HL-LHC) will integrate 10 times more luminosity than the LHC, posing significant challenges for radiation tolerance and event pileup on detectors, especially for forward calorimetry, and hallmarks the issue for future colliders. As part of its HL-LHC upgrade program, the CMS collaboration is designing a High Granularity Calorimeter to replace the existing endcap calorimeters. It features unprecedented transverse and longitudinal segmentation for both electromagnetic (ECAL) and hadronic (HCAL) compartments. This will facilitate particle-flow calorimetry, where the fine structure of showers can be measured and used to enhance pileup rejection and particle identification, whilst still achieving good energy resolution. The ECAL and a large fraction of HCAL will be based on hexagonal silicon sensors of 0.5 to 1 cm$^2$ cell size, with the remainder of the HCAL based on highly-segmented scintillators with SiPM readout. The intrinsic high-precision timing capabilities of the silicon sensors will add an extra dimension to event reconstruction, especially in terms of pileup rejection. An overview of the HGCAL project is presented, covering motivation, engineering design, readout and trigger concepts, and expected performance.


Introduction
Starting from 2026 onwards, the HL-LHCs instantaneous luminosity will be increased by a factor 5 to 7 compared to LHC and will result in up to 200 collisions per bunch crossing. In this mode, LHC will run for 10 years and deliver an integrated luminosity of about 3000 fb −1 . The current CMS detector was designed for operation at 25 collisions per bunch crossing and up to 500 fb −1 [1].
To cope with the new environment and retain a good physics performance up to 3000 fb −1 , several upgrades to the CMS subdetectors are planned [2]. The endcap calorimeters are among the subdetectors that will be most exposed to high radiation levels. Fig. 1 shows the expected total dose and hadron fluences active thickness) will be within ±3 µm of the average for the wafer. This translates to an effective Gaussian spread of the diffusion depth between the pads in a wafer of 0.6% (1.7%) for wafers with 300 (100) µm active thickness, respectively. As part of the test and qualification protocol for wafers the depletion depth of cells will be characterized by voltage/capacitance measurements with a precision of better than 1%. A number of "longitudinal" towers will be put into test beams to calibrate the responses to electrons and hadrons before startup.

Radiation tolerance
At the HL-LHC the silicon sensors of the HGC will be exposed to hadron fluences ranging from about 2 ⇥ 10 14 up to about 10 16 1 MeV neutron equivalent per cm 2 (neq/cm 2 ) as shown in Fig. 3.35. These fluences are similar to those in the tracker and pixel volumes for the HL-LHC, and the basic parameters for the HGC sensor design are based on results obtained for the CMS Phase-II Tracker R&D (e.g. Fig. 3.36, showing charge collection as a function of fluence), and further dedicated measurements using neutron fluences up to 1.6 ⇥ 10 16 n/cm 2 .
The main difference between the tracker and the HGC is that whereas in the tracker case the fluence is dominated by charged hadrons, in the case of the HGC it is neutrons that dominate. A dedicated campaign is underway to determine if the performance of the sensors is affected differently by neutrons. This study includes both p-in-n and n-in-p sensors, with active thickness of 300, 200, 100 and 50 µm, exposed to fluences up to the highest to which the HGC will be exposed. First results from neutron irradiation are summarized in Fig. 3.37. At the start of life the collected charge is about 22, 15 and 9 ke for sensors with 300, 200 and 120 µm active thicknesses, respectively. These numbers are calculated using 73 e/µm for the MIP charge deposition in silicon. Based on the measurements shown above, the collected charge after 3000 fb 1 is estimated to be in the worst case 10, 6 and 4 ke for the three thicknesses. These numbers are somewhat lower than expected from purely proton irradiation and the reason is being investigated. Our current design specifies an active thickness of 100 µm for the region subjected to the highest flux, but in the light of these measurements we may wish to choose an active thickness of 120 µm. The measured leakage currents for irradiated sensors, at -20 C and at bias voltages of 600 and 800 V, are consistent with expectations, as can be seen in Fig. 3.38 The scintillator planes are expected to be constructed using doubly-doped scintillator, as described in Section 3.4.6. The detailed geometry of the individual tiles will depend on the expected local radiation dose. The map of the doses expected in the BH region is presented in Fig. 3.30. The radiation doses range from approximately 5 Mrad to below 1 krad. In the lowdose regions, the traditional sigma-shaped arrangement of the WLS will be used, while in the medium-dose regions, the towers will be structured as set of narrow tiles, each of which is read out by a single WLS fiber running along the length of the finger tile. Figure 3.31 shows a comparison between a sigma tile and a finger tile. This simple concept increases the radiation tolerance of the detector because it significantly shortens the average light path between the particle-scintillator interaction point and the closest WLS fiber.
The front-end electronics for the BH will be heavily based on the HCAL Phase-I upgrade electronics [5]. However, the reconstruction of the endcap and the integration of the HGC services will require changes to the mechanics and structure of the electronics, requiring a reconstruction of the system. The data link of the electronics will be upgraded from the Phase-I bandwidth of 5 Gbps to the standard Phase-II 10 Gbps, allowing for a higher data concentration in the readout and trigger electronics.

Trigger and Off-Detector Electronics
Trigger data will be generated from sums of adjacent channels, using every alternate active plane. The sums will be made with a granularity of 2 ⇥ 2 sensor pads for the both the EE and FH, and sent at the full rate of 40 MHz by the front-end electronics to the services cavern where trigger primitives will be generated. The total number of trigger sums will be about 600k for the EE and 250k for the FH. In the BH, each digitized sample will be transferred to the off-detector electronics, with no pipeline in the front-end.
Trigger primitives, to be sent to the Level-1 calorimeter trigger, will be constructed in the offdetector electronics from the single-plane sums and the BH samples. The calorimeter electronics will be responsible for forming local longitudinal clusters and projective "towers" for use in the calorimeter trigger. as a function of R and Z. In the innermost regions, the detector has to withstand 10 16 neq/cm 2 and 150 MRad. Under these conditions, the current endcap calorimeters would degrade very quickly in performance [2]. Therefore, they will be completely replaced by a silicon and scintillator based highly granular sampling calorimeter called HGCAL (High Granularity Calorimeter).

Detector Design
The CMS HGCAL consist of an electromagnetic part called EE and two hadronic parts called FH & BH. 3 The electromagnetic part will be 25 X 0 deep and will consists of 28 layers of silicon pad sensors as active elements with lead in a stainless steel envelope as absorber. The two hadronic parts are in total 8.5 λ I deep with 24 layers and steel absorbers. As active elements, silicon will be used in the high |η| regions and scintillating tiles with SiPM readout in the lower |η| regions. The full system will be maintained at -30 • C using evaporative CO 2 cooling to limit the leakage current of the silicon sensors.
With silicon pads and scintillating tiles, high granularity in transverse and longitudinal direction will be maintained throughout the calorimeter and will allow for particle flow analysis. High precision time measurement with better than 50 ps resolution on a cell level is aspired for vertex reconstruction and pile-up rejection.

Active Elements
One of the most relevant quantity for the detector performance is the signal-tonoise ratio. For silicon, it has been shown that the signal loss due to irradiation is decreased in thinner sensors and when operating at increased bias voltages [2,3]. The increased noise contribution from the leakage current can be mitigated by cooling. Both aspects are displayed in Fig. 2. Additionally, the intrinsic time resolution of silicon has been shown to be below 15 ps for signals above 20 MIPs [3].
In total, the system will consist of roughly 600 m 2 of silicon. The use of 6 or 8 inch wafers with hexagonal geometry is foreseen to reduce costs. The active thickness will be adapted to the expected radiation dose and will vary between 120, 200 and 300 µm. 4 The cell capacitance should be around 50 pF for all sensor thicknesses and therefore thinner sensors will be equipped with smaller cells. A granularity of 0.5 cm 2 for the 120 µm and 1 cm 2 for 200 and 300 µm thick sensors will be used. One of the key aspects of these sensors is the high-voltage sustainability to mitigate radiation damage. The goal is a breakdown voltage above 1 kV. It is also foreseen to use a few cells with smaller area than the regular ones on each sensor. The smaller area at unchanged thickness will reduce the noise contributions from capacitance and leakage current in these cells, so that they should still be sensitive to single MIPs after 3000 fb −1 .
At larger distances to the interaction point radiation levels are lower and plastic scintillating tiles with SiPM readout will be used, analogous to the CAL-ICE AHCAL [4]. The exact intersection between scintillator and silicon regions as well as the tile granularity will be evaluated in the coming months.

Modules, Absorbers and Mechanical Integration
Silicon modules start with a metallic baseplate (CuW in EE and Cu in FH/BH), which acts as an absorber and mechanical support, that has a polyimide goldplated foil glued to it. The silicon sensor is then glued onto that foil. The readout PCB hosting the front-end ASICs is in turn glued onto the sensor and wirebonds reaching through holes in the PCB connect to the sensor contact pads. The design of the active scintillator modules is currently being developed.
Modules will be mounted on cooling plates together with the front-end electronics to make up cassettes. The absorber structure that hosts the cassettes will be made in full disks to guarantee an optimal physics performance. In the EE case, self-supporting double sided cassettes are used while in the FH and BH case, the mixed cassettes will be directly mounted on the steel absorber.

Readout Electronics
The driving requirements for the front-end readout ASIC are a large dynamic range of 0.4 fC to 10 pC (15 bits), a noise level below 2000 electrons, timing information with below 50 ps accuracy and radiation hardness up to 150 MRad. The goal is to keep within a power budget of around 10 mW/channel for the analog part. To meet these requirements, a chip based on OMEGA's ROC family [5] is being developed. The baseline option includes two traditionl gain stages and a time-over-threshold stage, as well as a time-of-arrival path with 50 ps binning. The ASIC will be fabricated in TSMC 130 nm CMOS technology which has been qualified up to 400 MRad [2].
Information from HGCAL will also be used for the L1 trigger decision. A subset of the data is sent to a concentrator chip and, after clustering, combined with the track trigger. The trigger latency of 12.5 µs drives the requirement for large buffer sizes in the readout chip.
A first version of the readout chip will be submitted in the summer of 2017.

Expected Performance
The choice of lead as absorber with a small Moliere radius and a large ratio of interaction length to radiation length allows for a compact calorimeter with excellent particle separation capabilities. The narrow showers together with the high granularity and excellent time resolution will allow for a pile-up suppression in the first few layers of EE. The instrinsic energy resolution of the EE part for incident electrons is expected to have a stochastic term below 25%/ √ GeV and a constant term below 1% [2]. These values are sufficient as the energy resolution will be dominated by the confusion term in the particle flow algorithm rather than the intrinsic resolution of the calorimeter. Optimisation of these algorithms to the physics environment and detector design is currently ongoing.

Outlook
The CMS collaboration is making good progress towards the construction of a new generation of imaging calorimeter. The basic design has been validated in testbeam and design optimisation is ongoing. The technical design report is expected to be released by the end of 2017.