Trigger and readout for the LHC experiments for Run 3 and beyond LHCb

LHCb

Trigger and readout for the LHC experiments for Run 3 and beyond

Carla Marin

Valencia, 20.03.23

Outline

● The LHCb experiment and the Run 3 upgrade

● The LHCb trigger system

● First data and performance

The LHCb experiment at the LHC

LHCb: Large Hadron Collider Beauty experiment

Distribution of produced b-quarks

● Precision measurements of heavy flavor hadrons

● Core physics: CPV and rare decays

● Much more: spectroscopy, QCD, heavy ions...

Experimental setup

Typical decay signature

JINST 3 (2008) S08005

Tracking system

Reconstruct trajectories of charged particles

Identify pp and b-decay vertex Measure particle momentum from bending in magnetic field

Run 2 performance:

Δp / p = 0.5 - 1.0%

Particle identification system

● Cherenkov detectors: identify π^±, K^±, p

● Calorimeters: identify γ, π⁰, e^±

● Muon chambers: identify μ^±

Cherenkov detectors

ΔE/E_ECAL = 1% + 10% / √ (E[GeV])

Electron ID ~90% for ~5%

Muon ID ~ 97% for 1-3%

π→μ mis-id probability

Kaon ID ~ 95 % for ~ 5 % π→K mis-id probability

LHCb dataset in Run 1 and 2

Total recorded luminosity ~9 fb^-1:

● Run 1 (2010-2012) ~ 3 fb^-1

All b-hadron species! [PRD100(2019)031102]

● B_s:

● Λ_b:

and more: Ξ_b, Ω_b, B_c, B^* …

x2 b-quark production from 7 to 13 TeV pp collisions

LHCb Upgrade: a quasi-new detector

Motivation: run at x5 instantaneous luminosity: L_inst = 0.4 → 2·10³³

● read full detector at 30 MHz → 4 TB/s

LHCb Upgrade

New VeloPix detector

New tracker detectors

New RICH detectors

Removal of SPD/PS, new electronics

Removal of M1, new electronics

The LHCb trigger system

Run 3: physics & constraints

Heavy flavour physics at LHC:

● L_inst = 2·10³³, x5 that of Run 2

● huge b and c productions

Trigger goals:

● Select interesting events: O(10^5-6 Hz)

● Reduce bandwidth: 4 TB/s → 10 GB/s

The Run 2 trigger model limitation

Classical model: hardware + software trigger

● Limitation: HW trigger rate limit (1 MHz) saturates fully hadronic & e⁺e^-/γ modes

● Solution: read full detector at 30 MHz and apply all selections in software

J. Phys.: Conf. Ser. 878 012012

Run 2 , Φ → μ ⁺μ^-

Run 3: a trigger-less readout

The challenge

The turbo model

BW(kB/s) = rate(Hz) x event size(kB):

4 TB/s input → 10 GB/s offline limit

● huge signal rate → reduce evt size

Flexible persistence model:

● Turbo (35 kB): signal only

● Full (70 kB): all reco’ed objects

● Selective: signal + selection of reconstructed objects and raw banks

JINST 14 P04006

LHCb Run 3 trigger overview

DAQ architecture

Hybrid architecture:

● HLT1: GPUs installed in Event Builder servers

● HLT2: CPUs in Event Filter Farm

HLT1

Core based on tracking:

● VELO: tracking, vertex reconstruction

● UT: tracking, p estimate, fake rejection

● SciFi: track reconstruction, momentum measurement

PID from muon stations & Calo (extra) Highly parallel tasks → exploit GPUs:

HLT1 performance: tracking

Same performance at x5 luminosity: high efficiency, good δp, low fake rate

LHCB-FIGURE-2020-014

HLT2

Full reconstruction of tracks and neutrals, and PID with offline-quality

LHCB-FIGURE-2022-005

HLT2 performance

Full reconstruction of tracks and neutrals, and PID with offline-quality

Commissioning and first data

F. Blanc @ICHEP22

First Run 3 data

First Run 3 data

HLT1 commissioning

LHCb DAQ running in parallel to detector commissioning since July

● Triggering on ECAL clusters @20 MHz!

● ~200 GPUs installed and HLT1 running in global partition

● Trackers included from September

HLT1 reconstruction and selections @20 MHz

HLT1 commissioning

Alignment

VELO & Scifi alignments determined with first data

LHCB-FIGURE-2022-016 LHCB-FIGURE-2022-018

Calibration

ECAL calibration from π⁰ → γγ events

Calibration

LHCB-FIGURE-2022-019

HLT2 commissioning

Data processed through HLT2, using results from Alignment & calibration

● without (left) and with (right) PID selection

Plans for Run 4 and beyond

LHCb trigger in Run 4 and beyond

Run 4

Same integrated luminosity that Run 3 New prototype detectors

Same trigger strategy as in Run 3

● possibility to renew GPUs

Run 5

Full upgrade for LHCb, L_inst x 10 Exploring scalability of the system New alternatives under study:

● Full HLT1 & HLT2 in GPUs

● FPGA reconstruction

Conclusions

Wide range of physics at LHCb → flexible trigger Unique trigger-less readout and Turbo model allow to cope with x5 lumi. Special gain for:

● hadronic final states

● electron final states

● long-lived particles

Strong progress with commissioning but it is a new

detector so it takes time!

Stayed tuned!

Thanks for the attention

BACK-UP

GPU choice

HLT1 performance: Velo

Same performance at x5 luminosity: high efficiency, good δp, low fake rate

LHCB-FIGURE-2020-014

HLT1 performance

Same performance at x5 luminosity: high efficiency, good δp, low fake rate

LHCB-FIGURE-2020-014

HLT1 without UT

Same signal efficiency but larger fake rate

LHCB-FIGURE-2022-007

LHCb Upgrade

New VeloPix detector

● 3.5mm to beam (5mm Run1/2)

● 41M pixels of 55x55 μm

● improved PV and IP resolutions

CERN-LHCC-2013-021

LHCb Upgrade

● Scintillating fibers 2.4m x 250μm

● higher granularity for higher occupancy

New tracker detectors

CERN-LHCC-2014-001

LHCb Upgrade