TB: ATLAS Jet Calibration and the Global Sequential Calibration (Run 1)
Overview
ATLAS reconstructs jets (anti-\(k_t\), $R=0.4$/\(0.6\), from calorimeter topo-clusters) and brings them to the truth-particle-jet scale through a multi-step calibration chain. The distinctive Run-1 step is the Global Sequential Calibration (GSC): after the absolute jet energy scale (JES) fixes the mean response, GSC removes the residual jet-by-jet response fluctuations that arise because a non-compensating calorimeter responds differently to different kinds of jets (quark- vs. gluon-initiated, differing shower depth, punch-through). It improves resolution and flavour-uniformity without changing the mean response.
The Run-1 calibration chain (in order)
- Pile-up correction — jet-area/\(\rho\) subtraction plus a residual correction parameterised in \(N_\mathrm{PV}\) (in-time pile-up) and \(\mu\) (out-of-time pile-up).
- Absolute JES + \(\eta\) calibration — a single MC-derived correction restoring the jet energy to the truth-particle-jet scale on average, and removing the \(\eta\) bias at calorimeter transition/boundary regions. Fixes the mean response; leaves response fluctuations.
- Global Sequential Calibration (GSC) — the focus of this page (below).
- In-situ calibration — residual data/MC differences measured with $Z$+jet, $γ$+jet, multijet \(p_T\) balance, and dijet $η$-intercalibration.
What GSC corrects
The JES (step 2) sets the average response correctly, but the response still depends on how the energy was actually deposited, which varies jet-to-jet through fragmentation. Two physical problems:
- Flavour dependence (quark vs. gluon jets). Gluon-initiated jets have higher particle multiplicity, softer fragmentation, and broader/shorter showers; quark jets are narrower with harder leading particles that penetrate deeper. Because the calorimeter is non-compensating (\(e/h > 1\)), the two jet types give different responses at the same true \(p_T\). A JES derived on a flavour mixture is therefore biased per-jet.
- Punch-through. Very high-\(p_T\) jets are not fully contained — energy leaks out the back of the calorimeter into the muon system, biasing the measured response low.
GSC reduces these response fluctuations (improving resolution) while holding the inclusive-sample average response fixed at unity.
Why "sequential"
GSC applies a chain of independent multiplicative corrections, one per observable. For each observable \(x\), the response \(R\) is measured in MC as a function of \((p_T, \eta, x)\), and a correction is applied so \(R\) no longer depends on \(x\) — with the average response over the full sample held fixed. The corrections are applied one after another rather than as a single multi-dimensional fit: the observables are correlated, so a joint correction is impractical; applying them in sequence (re-deriving each on top of the previous one) captures most of the correlation while staying tractable, and each step removes its observable's dependence without spoiling the mean established by the earlier steps.
The five Run-1 (8 TeV) observables, in order
| # | Observable | Subsystem | Targets |
|---|---|---|---|
| 1 | \(f_\mathrm{Tile0}\) — energy fraction in the 1st layer of the hadronic Tile calorimeter | Calorimeter (longitudinal) | shower depth / hadronic component |
| 2 | \(f_\mathrm{LAr3}\) (\(f_\mathrm{EM3}\)) — energy fraction in the 3rd layer of the EM LAr calorimeter | Calorimeter (longitudinal) | shower start / EM-vs-hadronic fraction |
| 3 | \(n_\mathrm{trk}\) — number of tracks (\(p_T > 1\) GeV) ghost-associated to the jet | Inner-detector tracking | charged multiplicity → quark/gluon flavour |
| 4 | track width \(W_\mathrm{trk}\) — $pT$-weighted average \(\Delta R\) of tracks about the jet axis | Inner-detector tracking | jet transverse width → quark/gluon flavour |
| 5 | \(n_\mathrm{segments}\) — number of muon-spectrometer track segments behind the jet | Muon spectrometer | punch-through at high \(p_T\) |
Steps 1–2 (longitudinal calorimeter fractions) and 3–4 (track multiplicity + width) attack the flavour / shower-shape dependence; step 5 (muon segments) is the punch-through correction, relevant only at very high \(p_T\).
How it is derived
- Corrections derived from Monte-Carlo simulated dijet events; validated against 2012 data at \(\sqrt{s} = 8\) TeV.
- Derived across the full \(\eta\) and \(p_T\) range.
- Each per-observable correction is normalised so the mean JES is unchanged — GSC is a pure resolution / uniformity improvement, not a scale shift.
Performance
- Jet \(p_T\) resolution: improved by up to ~35%, growing with \(p_T\) and depending on \(\eta\) and on whether EM- or LCW-scale clusters were used.
- Flavour dependence: residual quark/gluon response difference reduced by roughly ~20% at low \(p_T\) up to ~80% at high \(p_T\).
Sources
- 7 TeV (2010/2011) — foundational Run-1 jet calibration paper, introduces GSC: "Jet energy measurement and its systematic uncertainty in proton–proton collisions at \(\sqrt{s} = 7\) TeV with the ATLAS detector," Eur. Phys. J. C 75 (2015) 17, https://arxiv.org/abs/1406.0076
- 8 TeV (2012) — full Run-1 calibration & resolution, most detailed GSC treatment: "Determination of jet calibration and energy resolution in pp collisions at \(\sqrt{s} = 8\) TeV using the ATLAS detector," Eur. Phys. J. C 80 (2020) 1104, https://arxiv.org/abs/1910.04482
- 8 TeV GSC public note: "Jet global sequential corrections with the ATLAS detector in pp collisions at \(\sqrt{s} = 8\) TeV," https://cds.cern.ch/record/2001682
- ATLAS Open Data — hadronic/jet calibration documentation: https://opendata.atlas.cern/docs/documentation/physic_objects/jets_calibration
Related Concepts
- Particle Physics and the Standard Model — the experimental program these jets feed into