CERN Accelerating science

Precise timing detectors for the LHC experiments

by Panos Charitos, with contributions from Tommaso Tabarelli de Fatis (CMS), Lindsey Gray (CMS), Ana Maria Henriques Correia (ATLAS) and Laurent Serin (ATLAS)

The ATLAS and CMS experiments physics programme at the HL-LHC will target a very wide range of measurements, including in-depth studies of the Higgs boson properties and direct searches for physics beyond the standard model (BSM). Timing detectors will help improve vertex identification, acceptance extension for isolated objects, improved missing transverse momentum resolution, and pileup jet rate reduction that make a significant impact on the their physics programme across several channels. The characterization of the Higgs boson properties, with precision measurements of the Higgs boson couplings to standard model (SM) particles, and the search for rare SM and BSM decays, will benefit from timing detectors. In addition, the proposed detectors will increase the sensitivity to several searches for new phenomena that are part of the HL-LHC scientific programme, including many SUSY models. Finally, the precise track-time reconstruction opens a new avenue in searches for neutral long lived particles (LLPs), postulated in many extensions of the Standard Model and ALP searches. 

CMS and ATLAS experiments have recently approved their technical proposal for timing detectors [CERN-LHCC-2017-027] and [CERN-LHCC-2018-023] and should submit their technical design reports by the first quarter of 2019. 

CMS timing detectors at HL-LHC and the pileup combat

The CMS collaboration submitted a technical proposal for a dedicated timing detector aimed at providing timing information with about 30 ps resolution for charged tracks. This MIP Timing Detector (MTD) will complement the timing capabilities for photons in the electromagnetic calorimeters and for hadron showers in the forward region of the upgraded CMS experiment. The MTD will consist in a single layer providing timing in the barrel and the end cap region, with an angular coverage up to a pseudorapidity |η|<3.0. Radiation tolerance, integration, services, and schedule constraints, as well as cost considerations, lead to the choice of LYSO crystals read out by silicon photomultipliers (SiPMs) for the barrel and silicon low gain avalanche detectors (LGADs) for the end caps.

A simplified GEANT geometry of the timing layer implemented in CMSSW for simulation studies comprises a LYSO barrel (grey cylinder), at the interface between the Tracker and the ECAL, and two silicon endcap (orange discs) timing layers in front of the CE calorimeter.

The CMS experimental program at the HL-LHC, which include the precision characterization of the Higgs boson as well as searches of particles and processes not included in the so-called standard model of particle interactions, will benefit greatly from the increased luminosity provided by the upgrade of the LHC accelerator complex. However, particle reconstruction and correct assignment to primary interaction vertices in the presence of as many as 200 concurrent collisions per beam crossing (pileup events) represents a formidable challenge to the LHC detectors that must be overcome in order to harvest that benefit.

Pileup mitigation typically relies upon the removal from relevant quantities of charged tracks inconsistent with the vertex of interest, and of neutral deposits in the calorimeters with ansatz-based statistical inference techniques like PUPPI. Time tagging of minimum ionizing particles (MIPs) produced in LHC collisions provides further discrimination of interaction vertices in the same 25 ns bunch crossing beyond spatial tracking algorithms. According to preliminary simulation studies, a timing resolution of about 30 ps, about one sixth of the time spread of the LHC luminous region, holds the promise to recover the track purity of vertices of current LHC conditions, offsetting the performance degradation due to event pileup experienced in several observables.

The event display in the left panel of Fig.1, for instance, visually illustrates the power of space-time reconstruction in 200 pileup collisions, with a time-aware extension (4D) of the vertex reconstruction used by the CMS experiment. On average, the instances of vertex merging are reduced from 15% in space to 1% in space-time. Another measure of the performance improvement is illustrated in the right panel of Fig.1, showing the rate of tracks from pileup vertices incorrectly associated with the hard interaction vertex as a function of the line density of vertices. The rate of incorrect associations increases with the line density, as vertices start to overlap within a window for track-to-vertex association optimized for the 3D reconstruction. The addition of track-time information reduces the wrong associations at the typical vertex densities planned for HL-LHC operation (1.4-1.9 mm-1) to a level comparable to those observed without timing at the LHC vertex density of to about 0.3 mm-1  

Fig.1 - Left: Simulated (red-dots) and reconstructed vertices (blue) and tracks (black) in 200 pileup collisions using 4D tracking; vertices merged in 3D (yellow lines) are separated in 4D. Right: Rate of tracks from pileup vertices incorrectly associated with the primary vertex as a function of the vertex density.

Preliminary studies show that time cleaning substantially improves the reconstruction of final state observables relevant for the identification of processes with Higgs boson production and decays. The performance of b-jet identification, which relies on vertex reconstruction, is enhanced. The removal of pileup tracks from the isolation cones improves the identification efficiency for isolated leptons and photons. Similarly, the reconstruction of spatially extended objects and global event quantities that are vulnerable to the pileup, such as jets and the missing transverse momentum, is improved. Efficiency gains at the single-object level compound in multi-object final states – such as Higgs boson decays to four leptons, di-Higgs boson events or events where the Higgs boson is produced in association with other particles – providing potential gains, at constant rate of reducible backgrounds, of about 20-30% across all measurements. In addition, the ability to reconstruct the time of displaced vertices will provide enhanced capability in the search of long-lived particles (LLPs), with the ability of resonance reconstruction of the LLPs mass. The projected performance gains, for a 30 ps precision and hermetic coverage up to |η|<3, are summarized in Table 1.


The gain in sensitivity corresponds to a 20-30% gain in effective integrated luminosity and is equivalent to an additional 3 years of operation of the HL-LHC complex.

The barrel compartment of the MTD can be seen, in simplistic terms, as an optimization for charged track detection of existing detector and read-out chips developed for TOF-PET scanners, with specific customization of the geometry of the scintillation tiles and of the SiPMs. The high-rate and high-fluence environment of the HL-LHC, as well as the large detector area and the integration constraints, however, present a formidable challenge that will be addressed in a two-year long R&D and engineering phase, followed by three years of construction. The readout chip will be a high-rate evolution of the TOFPET2, including baseline subtraction to mitigate the impact of the radiation-induced SiPM noise.
The endcap compartment of the MTD, with LGADs, will be the first large-area timing detector is based on silicon sensors, with a total surface of 12 m2. The key reason for this choice is the required radiation hardness to withstand a fluence that is up to one order of magnitude larger than in the barrel. The development of multi-pad sensors of a few 10 cm2 size with the required radiation tolerance, response uniformity, and fill factor is the focus of the next three-year long R&D and engineering phase. This challenge is paired by, and the R&D tightly intertwined with, the development of an ASIC read-out chip of excellent timing capability, low power consumption, and good radiation tolerance. The chip will be designed in the 65 nm technology, leveraging on the success of the RD53 collaboration.
The barrel and endcap detectors are more complex than just their front-end components: cooling, clock distribution, mechanics, and power need to take up the gauntlet of designing a thin, noise-free, power-hungry set of services to make the MTD a success.
All in all, the R&D phase to demonstrate technologies and the detector concept is in full progress and is planned to continue throughout 2018, with the Technical Design Report (TDR) anticipated for the end of 2018. An engineering and prototyping phase, culminating in the delivery of the Engineering Design Reports (EDR), will follow. 

ATLAS plans for a new High Granularity Timing Detector (HGTD)

A High-Granularity Timing Detector (HGTD) using Low Gain Avalanche Diodes (LGAD) is proposed for the ATLAS phase 2 upgrade. Aiming to provide a time resolution of 30 ps /track in the region of 2.4<|η|<4.0 throughout the entire duration of the HL-LHC programme and provide a powerful way to mitigate pile-up that is one of the main experimental challenges.

HGTD will measure the time of individual tracks with a precision of 6 times better than the spread of the collision time, allowing to distinguish between collisions occurring very close in space but well-separated in time, as illustrated in Figure 2. This new detector will significantly improve the track-to-vertex association in the forward region, compensating for the reduced longitudinal impact parameter resolution of tracks reconstructed by the ITk tracker at large pseudorapidities. Rejecting pileup tracks with the new capability provided by HGTD improves the rejection of pileup jets by ~ a factor of 2 (for a hard-scatter jet efficiency of 98%), the lepton isolation efficiency increases by 14%, the light- jet rejection at a b-jet efficiency of 70% improves by a factor of 1.5 as seen in Figure 3.

Figure 2: Example of reconstructed times and z positions of the pT-weighted tracks associated to the hard-scatter vertex of an event with a VBF-produced Higgs boson decaying invisibly (H → Z(νν)Z(νν)) with ⟨μ⟩ = 200.  While superimposed along z, they two interactions are well separated in time.

Figure 3: Light-jet rejection versus b-tagging efficiency for the MV1 tagger using tt ̄ events at ⟨μ⟩ = 200 for different scenarios: start and end of the HL-LHC and in a pessimistic scenario where the time resolution of 30ps/track would be degraded by a factor of 2.

These improvements in object reconstruction performance translate into important sensitivity gains and enhance the reach of the HL-LHC physics program in the forward region. The signal strength determination for VBF-produced Higgs bosons decaying to H→WW and the signal significance for the tH(H → bb¯) are expected to improve by about 10 %. An improvement of 11% on the experimental uncertainty on the weak mixing angle sin(theta) can be achieved through the improved electron isolation performance in the forward region. Furthermore, the HGTD provides unique capabilities to measure the online and offline luminosity with high accuracy. It can also provide a minimum-bias trigger at L0 and possibilities for improved pileup mitigation in both the L0 and high-level trigger systems.

HGTD Layout

Reaching a 30 ps/track time resolution at mip amplitude in a harsh radiation environment is an ambitious goal and R&D on sensors and electronics have been quite intensive. In addition the design of the HGTD has limited available space allocated in the ATLAS experiment.

HGTD is a very thin, disk-shaped planar detector with 75 mm thickness in the Z direction and a radial coverage of 110 < R < 1000 mm, including the active area, peripheral electronics and the vessel needed to keep it at -30C, see Figure 4. It will be installed between the ITK and the endcap-forward calorimeters, in the place presently occupied by the Minimum Bias Scintillator trigger counters, at a distance of +- 3.5m from the interaction point.  

Figure 4: Illustration of the HGTD, showing the peripheral on-detector electronics (in green) and the active area (in blue).

In order to keep a 30 ps/track during its full life time, two double sided layers of LGAD will equip each forward region with an optimized overlap to achieve 3 hits per track in the inner ring (3.2<|η|<4.0) exposed to highest irradiation doses, and 2 hits/track in the external ring covering 2.4<|η|<3.2. The inefficiency due to non-instrumented zones, is expected to be less than 1% and the fraction of events with 0 hits should not exceed 2-3%.

A replacement of the inner ring (dark blue region in Figure 4), corresponding to 30% of the full active area, is needed at the middle life time of the HL-LHC to limit the radiation levels to a maximum of 3.7x1015 neq/cm2 and 4.0 MGy, including safety factors. The sensors will have a granularity of 1.3×1.3 mm2, a trade-off between occupancy and time resolution on one side (leading to small pad) and efficiency on the other side (leading to larger pad and fill factor). The sensor, of size of 2 x 4 cm2 ,  will be bump bonded to two 225 channels ASICs and read along X or Y axis with flexible printed board up to the large radius. The signals will then be transmitted at various speeds depending on the radial position of the ASIC through optical links to the ATLAS USA15 electronics cavern. This part of the electronic is quite similar to the one used for the ATLAS ITk strips upgrade (DC/DC converter and power supply, lpGBT, VL+ optical modules)   

A total of 6.3 m2 modules are needed to equip the full detector to be readout by 3.54 Million channels. This new detector proposal is the result of ~3 years of active R&D especially on sensors and front-end electronics, detailed below, undertaken by ∼20 Institutes and ~ 120 people from all around the word.

Ongoing R&D activities on HGTD sensors and ASICs

In close collaboration with RD50 and CNM/HPK manufacturers, an extensive R&D programme started ~ 3 years ago and is progressing quickly towards real size sensor pads and arrays. Those are being irradiated and tested in collaborating Institutes and then tested in beam tests, mostly at CERN/SPS H6 with pion beams [1].

Before irradiation, time resolution as good as 30 ps is obtained with gain up to 50 with 50 μm thickness LGADs, see Figure 5. Under irradiation, the observed degradation of the time resolution is attributed to a decrease of the gain caused by the loss of the effective doping concentration in the multiplication layer up to 1x1015 n/cm2. Beyond this fluence there is no more avalanche region but some gain from the bulk. Figure 6 shows the time resolution as a function of neutron fluence of LGAD single pads produced by HPK with a thickness of 35 and 50 μm thickness. The timing resolution per track at the maximum expected fluence  will be ~ 50 ps per hit. Therefore, maintaining a timing resolution of the order of 30 ps for the entire life of the HGTD should be possible with three effective layers at low radius and two layers elsewhere. Further R&D is on-going in close contact with CNM and HPK to improve the radiation hardness.

Figure 5: Time resolution vs.  gain for non irradiated single pad sensors (S) of 1.3x1.3 mmand arrays (A) from CNM with various doping (M, H). 

Figure 6: Time resolution vs. neutron fluence of LGAD single pad sensors produced by HPK with a thickness of 50 μm (50D) and 35 μm (B35).    

Another major R&D activity is the ASIC-ALTIROC development. The main requirements are driven by the targeted 30 ps time resolution per mip after irradiation. The electronics jitter contribution should be < 25 ps for a charge equivalent to a mip (10fC) while withstanding 4 MGy and have a power dissipation < 300 mW/cm2 to keep the cooling power budget to 25 KW maximum for the total detector. The ASIC in 130 nm CMOS technology from TSMC will contain 225 channels with a preamplifier, followed by a discriminator and Time-to-Digital Converters (TDC) for the digitization of Time-of-Arrival (TOA) and Time-Over-Threshold measurements. A Local FIFO memory is also included, storing the information until the trigger signal. The time information is transferred to the data acquisition system only upon L0/L1 trigger reception In order to measure the online bunch-by-bunch luminosity each ASIC of the outer ring will transmit at 40 MHz the total number of hits. A first prototype of the analog part has shown already promising jitter results on test bench (< 20 ps at 10 fC) and will be further tested with beam. A second prototype including the TDCs and the memory, looking as a complete pixel readout, was submitted middle of June. 

The clock distribution is also a challenge issue of this detector. A common working ATLAS/CMS group, led by CERN-EP-ESE-BE, has just started to evaluate the various contribution to the clock instability in term of jitter and wander. The current HGTD strategy is to bring the clock through the lpGBT and correct the T0 of each channel using the quite large bandwidth data information online and offline.  

In summary, an optimized baseline concept driven by the best compromise between performance and cost is summarized in the HGTD Technical Proposal and the Technical Design Report is expected at the end of March 2019. The intense R&D activities are expected to continue up to the end of 2020, followed by construction phase in 2021-2025 and final installation in the ATLAS cavern in summer 2025.


[1] Beam test measurements of Low Gain Avalanche Detector single pads and arrays for the ATLAS High Granularity Timing Detector”, submitted to JINST (archive link:


The author would like to thank Tommaso Tabarelli de Fatis and Lindsey Gray authors of the "CMS timing detectors at HL-LHC and the pileup combat" section and Ana Maria Henriques Correia and Laurent Serin who contributed the "ATLAS plans for a new High Granularity Timing Detector" part.