Periodic Reporting for period 1 - JEC4HL-LHC (Jet Energy Corrections for High-Luminosity LHC)
Période du rapport: 2022-09-01 au 2025-02-28
The 2020 update of the European Strategy for Particle Physics puts the HL-LHC at the focal point, “together with continued innovation in experimental techniques”. Its primary goals include searching for new physics at high energy, exploration of the Higgs potential and precision tests of the Standard Model. At a hadron collider such as LHC, practically all measurements rely on jet energy corrections (JEC). Many important benchmark channels, such as top quark decaying to b jets and ud or cs jet pairs and inclusive jets produced with numerous gluon jets, have JEC as their limiting uncertainty. At present, the JEC uncertainties are at a level of 1% overall, with similar additional uncertainties from flavor-specific JEC from simulation only. However, evidence suggests that the simulation is biased at the level of these uncertainties, and that these biases will become the decisive factor for further progress. Here we show how to correct these biases with data-driven methods and bring the total uncertainties toward 0.1% level. We found that with integrated luminosity of more than 100/fb, the Z+flavor channel can be used to precisely measure all flavor-JEC, when combined with our novel techniques to address flavor-dependent biases from initial and final state radiation, underlying event and heavy-flavor neutrino production. Furthermore, we found that inclusive jet production can be used to monitor further drifts of JEC at 0.1% precision and that our observations can be motivated with a parameterized model of particle flow jet response. Our results plot a path to JEC at the HL-LHC high pileup environment, and show how to systematically control the leading simulation uncertainties with data and allow many physics analyses to improve their JEC-related systematic uncertainties. We anticipate our work will be a starting point to transition to 0.1%-level precision in JEC across the field and that it will lead to a number of precision jet physics measurements as a result.
In the first two years of the project, our team has assumed leading responsibility of Run 3 jet energy corrections (JEC). We process and combine Z+jet, photon+jet, dijet and multijet channels to derive JEC versus jet momentum (pT) and pseudorapidity (eta). We have also expanded our novel calibration methods to include derivations for jet energy resolution (JER) scale factors (SF) and to flag poorly functioning detector regions with jet veto maps in jet eta and azimuthal angle (phi). Our new calibration software has shortened the turn-around time for new calibrations from month to days at best, with rolling calibration accumulating statistics, making it possible to provide rapid feedback on detector performance and to flag rare detector problems at corners of phase space as soon as enough data has been collected to see them. This has enabled CMS to find, understand and fix several critical reconstruction issues in Run 3, as well as to significantly improve low-level detector calibration.
Working also on Run 2 legacy data, we have been able to demonstrate proof-of-principle on several new techniques important to the key topic of measuring flavor JEC: (1) we have derived pT-dependent working-point-based scale factors for quark and gluon tagging documented in a PhD thesis, (2) we have developed a novel method to measure gluon response relative to light quark jet response at high pT documented in a MSc thesis, and (3) we have developed a novel method to calibrate strange-quark jets relative to charm-quark jets, and strange and charm jets relative to ud-quark jets, documented in a MSc thesis.
Looking towards the future at HL-LHC, we have an early demonstration of methods to run nearly real-time jet energy corrections on prompt reconstruction immediately after Prompt Calibration Loop, which could enable running daily calibrations with 48h of collecting data. The automatized software has also an important role in crystallising and permanently documenting the accumulated knowledge in code, avoiding dispersion of and loss of knowledge at the end of the project.
Working also on Run 2 legacy data, we have been able to demonstrate proof-of-principle on several new techniques important to the key topic of measuring flavor JEC: (1) we have derived pT-dependent working-point-based scale factors for quark and gluon tagging documented in a PhD thesis, (2) we have developed a novel method to measure gluon response relative to light quark jet response at high pT documented in a MSc thesis, and (3) we have developed a novel method to calibrate strange-quark jets relative to charm-quark jets, and strange and charm jets relative to ud-quark jets, documented in a MSc thesis.
Looking towards the future at HL-LHC, we have an early demonstration of methods to run nearly real-time jet energy corrections on prompt reconstruction immediately after Prompt Calibration Loop, which could enable running daily calibrations with 48h of collecting data. The automatized software has also an important role in crystallising and permanently documenting the accumulated knowledge in code, avoiding dispersion of and loss of knowledge at the end of the project.
Our methods to track down reconstruction issues using JEC workflows are beyond the state-of-the-art of past calibrations, and have truly shown their power in Run 3, where we have already been able to find, understand and fix four critical reconstruction issues, one of which was mission critical. These issues were causing a loss up of to 50% of multi-TeV scale jets, or affected their response by up to 10-15%. In the case of forward region, the miscalibration of a fraction of jet energy at high jet |eta| reached up to 300%. Thanks to our rolling calibration, the impact of these issues was limited and/or can be addressed with re-reconstruction of data.
Working on Run 2 data, our novel methods go beyond the state-of-the-art for flavor-JES, as documented in one PhD and two MSc theses:
1) quark and gluon jet tagging scale factors for the first time at CMS as a function of pT and including advanced machine learning taggers, made possible by working-point based approach,
2) novel method to measures gluon jet JES relative to light quark jet JES using dijet events only, extending gluon jet JES to TeV scale jets for the first time,
3) novel method measures strange-quark jet JES relative to charm-quark jet JES and both relative to ud-quark jet JES in W>qq' events, marking the first time strange-jet JES is directly measured in data.
These novel methods are particularly important because they extend the direct data-based flavor-JES measurements to two previously uncovered corners of phase space: TeV scale for gluon jet JES and strange-flavor for light quark jets. These new windows to flavor JES could be critical for finding the fundamental cause of bad flavor response modelling in simulation.
Our third result beyond state-of-the-art is an early internal demonstration of a workflow that would enable daily calibrations within 48h of collecting data, which we call JEC4Prompt. The level of automatisation and speed that this represents for JEC would be unprecedented, with Run 2 JEC analyses typically taking months and teams of 10-20 people. Most importantly, the monolithic software would crystallise best practices and document accumulated knowledge permanently in code.
Working on Run 2 data, our novel methods go beyond the state-of-the-art for flavor-JES, as documented in one PhD and two MSc theses:
1) quark and gluon jet tagging scale factors for the first time at CMS as a function of pT and including advanced machine learning taggers, made possible by working-point based approach,
2) novel method to measures gluon jet JES relative to light quark jet JES using dijet events only, extending gluon jet JES to TeV scale jets for the first time,
3) novel method measures strange-quark jet JES relative to charm-quark jet JES and both relative to ud-quark jet JES in W>qq' events, marking the first time strange-jet JES is directly measured in data.
These novel methods are particularly important because they extend the direct data-based flavor-JES measurements to two previously uncovered corners of phase space: TeV scale for gluon jet JES and strange-flavor for light quark jets. These new windows to flavor JES could be critical for finding the fundamental cause of bad flavor response modelling in simulation.
Our third result beyond state-of-the-art is an early internal demonstration of a workflow that would enable daily calibrations within 48h of collecting data, which we call JEC4Prompt. The level of automatisation and speed that this represents for JEC would be unprecedented, with Run 2 JEC analyses typically taking months and teams of 10-20 people. Most importantly, the monolithic software would crystallise best practices and document accumulated knowledge permanently in code.