The ATLAS liquid argon calorimeter: upgrade plans for the HL-LHC

The ATLAS detector was designed and built to study proton-proton collisions produced at the LHC at centre-of-mass energies up to 14 TeV and instantaneous luminosities up to 1034 cm−2s−1. Liquid argon (LAr) sampling calorimeters are employed for electromagnetic and hadronic calorimetry. The luminosity for the proposed High Luminosity LHC phase (HL-LHC) will increase up to 5×1034 cm−2s−1 with the goal of accumulating an integrated luminosity of 3000 fb−1. This is well beyond the values for which the detectors were designed. The electromagnetic and hadronic calorimeters will be able to tolerate the increased particle flux, but the performance of the forward calorimeter (FCal) will be affected. Two possible solutions for keeping the current performance are being discussed. The readout electronics will also need to withstand larger radiation environment. In the hadronic endcap calorimeter (HEC) cold GaAs preamplifiers are located inside the endcap cryostats. The properties of these devices have been investigated in recent proton and neutron irradiation tests to determine whether they must be replaced. In addition, the entire front-end readout system is not expected to survive the integrated luminosity at the HL-LHC and will be replaced. The description of the new readout system is presented.


Introduction
The ATLAS detector [1] at the LHC was designed for measurements of proton-proton collisions at centre-of-mass energies up to 14 TeV and instantaneous luminosities up to 10 34 cm −2 s −1 .
The ATLAS LAr calorimeter, shown in Figure 1, is a sampling calorimeter with liquid argon as an active medium. It consists of barrel and endcap parts enclosed in corresponding cryostats. The barrel cryostat contains only the electromagnetic calorimeter (EMB), covering a pseudorapidity range of | η | < 1.475. The endcap cryostat contains the electromagnetic endcap (EMEC), covering a range of 1.375< | η | < 3.2, the hadronic endcap (HEC), covering the region of 1.5 < | η | < 3.2 and a forward calorimeter (FCal) covering the region 3.1 < | η | < 4.9.
The EMB and EMEC were constructed using an accordion geometry for full ϕ coverage, with copper and kapton electrodes and lead as the absorbing material. The electrodes are positioned between the absorber plates by honeycomb spacers. The calorimeter is segmented into three longitudinal layers for | η | < 2.5 and two coarser-granularity layers for | η | > 2.5.
The HEC is segmented into four layers constructed in parallel-plate geometry. It has copper absorbers and copper and kapton electrodes. The FCal consists of three modules constructed of electrode rods parallel to the beam pipe, sitting in a metal matrix absorber. The LAr gap between the rods and matrix are smaller than in the rest of the LAr detector in order to endure the high particle fluxes in the forward regions. The first module of the FCal is for electromagnetic measurements and consists of a copper matrix and 269 µm liquid argon gaps. The other two modules are for hadronic measurements and consist of tungsten matrices with 376 and 508 µm liquid argon gaps.
Although the nominal LHC experimental program is still in progress, plans for a High Luminosity LHC (HL-LHC) are already being developed for operation of the collider and associated detectors at luminosities of up to 5×10 34 cm −2 s −1 , with the goal of accumulating an integrated luminosity of 3000 fb −1 .
The LHC baseline plan for 2013-2030 [2] includes long shutdowns (LS) in which upgrades will be made for the next Phase. In the next long shutdown (LS2) during 2018-2019, an upgrade for Phase-I, it is planned to upgrade the injector system and secure the ultimate LHC luminosity. And the long shutdown (LS3) over 2023-2025, upgrade for Phase-II or HL-LHC phase, will increase an instantaneous luminosity up to 5×10 34 cm −2 s −1 .
The proposed instantaneous and integrated luminosities are both well beyond the values for which the detectors were designed. The electromagnetic and hadronic calorimeters will be able to tolerate the increased particle flux, but the performance of the forward calorimeter (FCal) will be affected. Two solutions for mitigation of the response degradation are under study: one option is to build and install a replacement FCal with smaller LAr gaps. A second approach involves the installation of a small Mini-FCal calorimeter in front of the FCal. These options are discussed in section 4. The currently preferred Mini-FCal option is a LAr/Cu sampling calorimeter for which detailed design and performance studies are underway.
The electronics readout, originally qualified for an integrated luminosity of only 700 fb −1 , will also potentially be affected. This concerns especially the hadronic endcap calorimeter (HEC) that employs cold GaAs preamplifiers located inside the endcap cryostats. Individual ASIC positions are located at the outer circumference of the HEC with r = 204 cm and 430 cm < z < 600 cm. The properties of these devices have been further investigated in recent proton and neutron irradiation tests to determine whether they must be replaced. The measured gain and linearity degradation is used to determine the performance expected for calorimetric measurements in HL-LHC conditions, discussed in section 3.
Beside the specific HEC radiation concerns, the entire front-end readout system is not expected to survive the integrated luminosity at the HL-LHC. Furthermore, the requirements of long-latency data buffering and multi-level hardware triggering by the future ATLAS trigger system are not compatible with the current front-end electronics. As these electronics are much more accessible than the HEC preamplifiers, a decision has been taken to replace the complete  readout chain. The new readout system must be able to process about 180000 channels at 40 MHz in real time, with 16 bit dynamic range and a total data rate of about 140 Tbit/s. To meet these challenging environmental and performance parameters, dedicated ASICs for lowlatency, low-power analog-to-digital conversion and for fast, low-latency optical transmission are being developed, along with FPGA-based high-bandwidth on-line data processing modules, discussed in section 2.

Readout Upgrade
The front-end crates (FEC) are installed at the two ends of the barrel cryostat at the outer ends of each endcap vessel. The crates house the front-end electronics which amplify, process and digitize the calorimeter signals. There are three points that can be addressed by an upgrade of the FEC: the radiation damage of the on-detector electronics, natural ageing effects of electronic components and the possibility to improve the ATLAS trigger system. The FEC electronics for HL-LHC phase will be upgraded in steps. First, additional new trigger boards will be installed in LS2. Then there will be the replacement of the primary readout and calibration boards in LS3. The FEC will contain four types of boards, implementing four different functionalities: the upgraded front-end boards (FEB2) reading out signals from 128 calorimeters cells, the LAr trigger digitizer boards (LTDB and LTDDB), the calibration boards and the control boards for the clock and configuration signals handling [3]. Figure 2 shows the front-end architecture planned for the HL-LHC phase. The calibration boards, which are not used during collisions, are not shown. The control board functionality may or may not be integrated directly into the new generation front-end boards. The upgrade benefits from the high granularity of the calorimeter to provide more information for the triggering system. This also requires development of highspeed data transmission elements. The new readout, described above, will have an optical transmission of all data to the back-end. Also front-end components are planned to be more radiation-hard and with trigger buffers located off-detector. The upgraded trigger readout foreseen in the LS2 will be used as a low-latency Level-0 trigger in the HL-LHC Phase. This also requires upgrading the back-end electronics.

The Upgraded Front-End Board
The new front-end boards are designed to improve radiation tolerance and to increased the size of the trigger pipelines which will allow more advanced trigger algorithms to be applied. The main functional blocks are: analog preamplification and shaping, production of (summed) analog signals for the LTDB, analog-to-digital conversion of all signals at a rate of 40 MHz, multiplexing and serialization of digital data, and transmission over high-speed optical links. The analog preamplification and shaping stages will be integrated in a single ASIC. The Liquid Argon Preamp and Shaper (LAPAS) test-chip, fabricated in 2009 in IBMs 8WL SiGe process, validated the design approach of implementing a wide dynamic range single ended preamp followed by low power differential shaping stages with multiple gains to achieve the required 16 bit resolution. Less expensive SiGe process alternatives are currently being explored: IHP SG25H3P and IBMs 7WL. These studies will evolve into LS3 applicable ADCs. As an alternative, the qualification of commercial 12-bit ADCs for LS2/LS3 upgrade applications was done. The best performance was shown by a Texas Instruments chip (ADS5272) which tolerated radiation up to 1 kGy, has a low power consumption (113 mW) and a low latency (163 ns).
Data of 8 ADC channels are encoded, multiplexed, serialized and send out by the Link On Chip (LOC). A development in Silicon-on-Sapphire (SoS) 0.25 µm technology allows 8 Gbps and will be applied at 5 Gbps in the LS2 upgrade. In addition, there are planned a custom interface and serializer VCSEL pairs from the CERN Versatile Link [4]. All together for LS3, a 12×10 Gbps CERN VCSEL arrays and an improved SoS process or a next-generation CERN GBTx [5] need to be developed. The data processing will be done on the basis of LS2 upgrade R&D. There are 14 FEB2s connected to one Pre-processor board covering 32 towers of size ∆η × ∆ϕ = 0.1 × 0.1. The incoming data rates per board are estimated to be 14×90 Gbps of digitized input data. At the output, low-latency data transmission of about 250 Gbps and 5 to 10 Gbps is expected to the Level-0 and Level-1 trigger systems, respectively. This requires high-bandwidth ATCA Pre-Processor boards implemented on high performance FPGAs and treating 1.2 Tbps input from the FEB2s and delivery of summed signals to the trigger system. The new FEB2 should fulfill the required tasks: energy reconstruction, bunch-crossing identification, data pipeline and delivery of summed signals to the trigger system.

HEC Low Voltage Power Supplies and HEC Cold Electronics
The radiation doses in ATLAS including at the positions of the HEC ASICs have been studied originally for the design luminosity of the LHC [3]. To extract the Si NIEL fluences for the HL-LHC from these simulations the following assumptions have been made: 14 TeV protonproton collisions with σ inelastic = 79.3 mb and an integrated luminosity of 3000 fb −1 after 10 years of running the HL-LHC. A safety factor of 2 was used to account for uncertainties in the simulation. This was the result of dedicated measurements in critical areas of the detector in order to validate the simulated particle fluxes. Figure 3 (left) shows the expected ATLAS Si-NIEL fluences induced in the ASICs as a function of the location of the readout cell in the HEC which will be impacted by the corresponding ASIC. Total ionizing dose (TID), single event effects (SEE) and neutron backgrounds measured in the 2012 run agree with background simulations using FLUKA and GEANT4. The electrical power required by the FE electronics is delivered by a Low Voltage Power Supply (LVPS). The expected neutron dose (from simulation) for HL-LHC for HEC LVPS exceeds 1.8×10 12 n eq /cm 2 including a safety factor two. For the HEC cold electronics it reaches up to 10 14 n eq /cm 2 . Due to the limited radiation tolerance of the current electronics, the 8 HEC LVPS should be replaced. Radiation qualification of power MOSFETs and logic devices has thus been started.
All together, 16 types of power MOSFETs were evaluated by measuring power loss as function of fluence (up to 1.4×10 14 n/cm 2 ), an example is shown in Figure 3 (right). All tested transistors qualify for DC/DC converter application with the new NIEL requirement. The updated radiation requirements may allow use of radiation-tolerant FPGAs in the digital part. Detector control electronics (Embedded Local Monitor Board -ELMB) also need be developed in radiation tolerant technology [6].
The readout electronics of the HEC will have to withstand larger radiation environment at the HL-LHC compared to their design values. The preamplifier and summing boards (PSBs) are equipped with GaAs ASICs. These ASICs were irradiated with neutrons and protons with fluences surpassing several times ten years of operation of the HL-LHC. The setup and measurements are described in [7].
The three different transistor technologies that were tested are Si CMOS FET in SGB25V 250 nm technology from IHP, SiGe Bipolar HBT (IHP SGB25V 250 nm and IBM 8WL-BiCMOS 130 nm), and the GaAs FET currently used in ATLAS, either the Triquint CFH800 250 nm transistors themselves or integrated into the HEC BB96 Preamplifiers (PAs) and Systems [8]. Later, the irradiated boards were re-tested in liquid nitrogen to simulate the conditions inside the HEC cryostats.
The investigation of the GaAs ASICs performance was studied as a function of neutron fluence with two parameters: the forward transmission coefficient of the Systems and the linearity of the PAs. The linearity of the PAs was characterized by a power-law fit to their response to triangular input pulses of varying amplitudes. Figure 4 shows the result of the warm and cold investigations for the currently installed BB96 ASICs. The warm results show only a small degradation of the HEC electronics up to HL-LHC conditions, whereas the cold measurements indicate a much more severe performance The red dashed line indicates the HL-LHC limit (Si) (including the safety factor of 5, recently lowered to two), taking into account that 4.1×10 14 n eq /cm 2 NIEL in ATLAS corresponds (Si) to 3.0×10 14 n eq /cm 2 NIEL at NPI (for GaAs devices) [9].
degradation. The red line corresponds to HL-LHC conditions with safety factor 5. This has recently been reduced to 2, after further validation of the background simulations. With this reduced SF, results show that the GaAs ASICs are good for HL-LHC operation.
In addition to the problems associated with the increased radiation levels, the current electronics has to survive until the end of the HL-LHC phase and work for approximately 30 years. Therefore there is a need to estimate the longevity of the pre-amps [3]. Activities to measure mechanical and electrical stability of the chips and system are ongoing.

Forward Calorimeters
The Forward Calorimeter (FCal) was designed with very narrow LAr gaps in order to avoid problems due to ion build-up that would distort the electric field. At HL-LHC luminosities and at high η region, the gaps in the first FCal sections are no longer sufficiently narrow. In this area the pulse shapes are expected to degrade due to Ar + build-up, which leads to field and signal distortion, and due to increased current draws over the 1 or 2 MΩ protection resistors used to deliver the high voltage currents, which leads to a voltage drop. Due to high expected particle flux at 5×10 34 cm −2 s −1 , a temperature increase of a liquid argon may lead to LAr bubble formation and has to be studied. Figure 5 shows the FCal cylindrical electrodes, parallel to the beamline, located within an absorber matrix. In the right figure the schematic structure of a unit cell, including the absorber matrix is shown. It comprises a cylindrical copper tube into which a metal absorber rod of slightly smaller diameter has been inserted. Each electrode consists of a rod at high voltage, a tube at ground, and a PEEK fibre that maintains the gap between the rod and tube. In the Figure 6 (left) the simulated combined effect of Ar + build-up and reduced HV on the nominally triangular physics pulses for the LAr cells is illustrated for cells at | η | = 4.7, for a range of instantaneous luminosities. The amplitude is reduced with increased luminosities. The main effect is coming from the HV drop due the higher currents flowing across the 1 or 2 MW HV protection resistors, which are also located within the cold volume of the endcap cryostats.
At a luminosity of 1×10 34 cm −2 s −1 about 30 Watts of heat is expected to be deposited in each FCal and especially 18 W in first FCal module (FCal1), as shown in Figure 7  endcap cryostat cold vessel. The greatest impedance to this heat flow is at the 12 mm LAr-filled gap between the FCal support tube and the inner bore of the HEC. Heat simulations for FCal for 2012 luminosities agree well with temperature probe data. The measured temperature increase is 0.1 K. The simulation for 6×10 34 cm −2 s −1 shows however a relative temperature increase of about 2.1 K, as shown in Figure 7 (right), which may lead to a bubble formation. To study the heat transport in the region between FCal and HEC a mock-up module is being prepared. The test module is displayed in Figure 6 (right). At the left is a heater plane which heats an aluminum plane representing the wall of the FCal support tube. Then the heat flow will go through an changeable liquid argon gap to an irregular copper surface constructed to resemble the inner bore of the HEC. The HEC will be cooled from the right side with liquid argon, since the copper structure of the HEC serves as main cooling source. In between temperature probes will be installed to measure temperature differences which will be used to determine the effective thermal conductivity of LAr gap. Finally the mock-up module can be rotated in the cryostat to simulate the different slices of the HEC inner ring. The results of this test will be used to validate the simulation used to predict the module heating.
To solve the problems of pulse degradation and bubble formation two options are considered as discussed below.

Upgrade Scenarios for the ATLAS Forward Region
The ATLAS Hilum Collaboration [10] has experimentally investigated the performance of the three endcap LAr calorimeters in a high-intensity test beam in Protvino. The LHC luminosity of 10 34 cm −2 s −1 minimum bias events energies were translated to Protvino energies and obtained a corresponding beam intensity of 6.7×10 8 p/s for the FCal(269), which is affected by large systematic errors. Their results indicate that while the present FCal1 performance would degrade significantly at high luminosity, a similar structure with an argon gap of 119 µm shows no signal degradation up to intensities an order of magnitude greater than those anticipated for the HL-LHC. This is shown in Figure 8 (right). There is no critical intensity observed within the intensity range studied in contrast to other FCal module.
High particle flux tests showed that a liquid argon calorimeter with this electrode structure can function at the rates expected in a Mini-FCal at the HL-LHC [10]. These narrow gap FCal electrode can therefore be employed in the sFCal and Mini-FCal designs.  This option is easier due to known technology to design and optimize the sFCal. It is basically a modified version of FCal with smaller LAr gaps (∼ 100 µm in the FCal1), new cold electronics boards (which hold the HV resistors), and cooling loops. This option is described in [3]. But the installation of the sFCal requires the opening of the cold volume of the endcap cryostats which are welded shut, so protection of the detector is also a major issue. This procedure is complicated and can be done only in a long shutdown. In addition, radiation protection issues are to be taken into account. to installers. Simulations of this option are ongoing to show the reduction both the heat load and the instantaneous ionization rate in the FCal1 to acceptable levels. A schematic view of one Mini-FCal design, located between an inner warm tube and the cold wall of the cryostat, is shown in Figure 8 (left). The Mini-FCal detector must be extremely radiation hard and several technologies have been considered. Currently preferred is an FCal1-like LAr calorimeter with copper absorbers and the same 100 µm gaps proposed for the sFCal. The disadvantage of this option is that it will require additional material upstream of the detector due to the necessity of an additional cryostat for Mini-FCal and by the associated services. This has the potential to degrade the overall detector performance. Design optimization and performance simulations for the LAr Mini-FCal option are ongoing.

Conclusions
In order to survive the expected luminosity at the HL-LHC, the ATLAS experiment will be upgraded to meet the experimental challenges: increased data rates, occupancy and radiation fluence. The goal of HL-LHC upgrade is to meet the challenging pile-up and radiation requirements, to ensure longevity of the system and to maintain or improve the existing detector performance. The LAr front-end and back-end electronics will be replaced. The current R&D for LS2 trigger readout developments is ongoing and will naturally continue. There are many activities ongoing to clarify the longevity and performance of critical components in the LAr endcap area, such as, HEC cold electronics, FCal detector and its possible upgrade with new sFCal or Mini-FCal detectors. A decision on which upgrade path will be selected for the forward region and for the endcap calorimetry is planned to be completed by 2015 in order to be prepared for a construction phase well ahead of the LS3 and the beginning of the HL-LHC phase.