Measurement of isolated photon production in pp and PbPb collisions at sqrt(sNN) = 2.76 TeV

1 Introduction

• Isolated photon measurements test pQCD [1] and constrain proton [2] and nuclear [3] parton distribution functions.In nuclear collisions, comparing pp and nuclear production can estimate modifications relative to an incoherent superposition of nucleon PDFs.Source: §1 · p. 3
• Neutral-meson decays produce a large photon background, and isolation suppresses these decays and fragmentation photons while retaining most direct photons.In PbPb collisions, underlying-event activity must be subtracted before applying isolation criteria.Source: §1 · p. 3
• The measurement reports isolated photon production in pp and PbPb collisions at a nucleon-nucleon energy of 2.76 TeV with CMS.The paper describes the analysis sequence from detector and triggers through reconstruction, signal extraction, theory, and comparison of spectra.Source: §1 · p. 3

2 The CMS detector

• CMS uses silicon pixel and strip detectors for charged-particle tracking and ECAL and HCAL calorimeters for energy measurement.The tracking system covers |η| < 2.5, while the calorimeters cover |η| < 3.Source: §2 · pp. 3–4
• In the analyzed barrel region, |η| < 1.44, ECAL and HCAL granularity is 0.0174 × 0.0174 and 0.087 × 0.087 in η and φ, respectively.The calorimeters and tracking systems sit within the 3.8 T solenoid field.Source: §2 · p. 4
• The forward HF calorimeter covers 3 < |η| < 5.2 and determines PbPb centrality; BSC tiles support triggering and beam-halo rejection.Source: §2 · p. 4

3 Data samples, triggers and event selection

• The data samples correspond to integrated luminosities of 231 nb−1 for pp and 6.8 µb−1 for PbPb.The PbPb sample has a pp-equivalent luminosity of 294 nb−1.Source: §3 · p. 4
• PbPb photon-triggered events require an L1 electromagnetic cluster above 5 GeV and an HLT photon above 15 GeV; trigger efficiency exceeds 98% for corrected photon ET above 20 GeV.Figure 1 plots trigger efficiency against corrected photon ET using the minimum-bias PbPb sample, with statistical error bars.Source: §3 · p. 4, Figure 1 · p. 5
• The minimum-bias PbPb sample uses coincidences between the +z and −z sides of either the BSC or HF, with trigger and event-selection efficiency of (97 ± 3)% [4].Source: §3 · p. 4
• Offline selection requires a photon candidate with |ηγ| < 1.44 and corrected transverse energy above 20 GeV; PbPb events also pass vertex, HF-coincidence, and pixel-cluster quality selections.Source: §3 · p. 4

4 Monte Carlo simulation

• PYTHIA γ+jet, dijet, and W → eν events are detector-simulated with GEANT4 and embedded in recorded minimum-bias PbPb events.The mixed PYTHIA+DATA samples support signal-shape studies and energy and efficiency corrections.Source: §4 · p. 5
• Generator-level photon isolation uses a cone of radius 0.4 and requires less than 5 GeV of other hard-scattering particle ET inside it.Source: §4 · p. 5
• GEANT4 simulation determines isolated-photon energy and efficiency corrections.Source: §4 · p. 5

• Centrality uses the total HF energy in minimum-bias PbPb events to define 0–10% most-central, 10–30% mid-central, and 30–100% peripheral ranges.Source: §5 · p. 5
• N_coll counts incoherent nucleon-nucleon collisions and normalizes PbPb yields because it is directly proportional to high-pT particle-production yields.N_part counts nucleons that experience at least one collision, ranging from 2 for one nucleon-nucleon interaction to 2 × 208 for a head-on collision.Source: §5 · p. 6
• Photon-triggered events are more central than minimum-bias events, and the trigger-passing subset is about 3.3% of minimum-bias events.Source: §5 · p. 6, Figure 2 · p. 6

• In PbPb, island clustering grows ECAL clusters from seeds above 0.5 GeV and merges nearby clusters into superclusters.Source: §6 · p. 6
• Correction factors from γ+jet PYTHIA+DATA are tabulated by centrality and photon ET to remove underlying-event energy from reconstructed ET.Source: Table 1 · p. 7, §6 · p. 7
• 9% is the maximum photon energy resolution, in the lowest ET bin for 0–10% central events.Figure 3 shows relative resolution versus photon ET for three centrality intervals; horizontal bars indicate bin widths.Source: Figure 3 · p. 7, §6 · p. 7
• The anomalous-signal selections require timing within 3 ns and adjacent-cell energy of at least 10% of the central cell; 99.7% of photon signal candidates pass.Source: §6 · p. 7
• Candidates must satisfy |ηγ| < 1.44, avoid an electron-track match, have EγT > 20 GeV, and pass H/E < 0.2.Source: §6 · p. 7
• PbPb isolation sums ECAL, HCAL, and track activity within ΔR = 0.4 after underlying-event subtraction, requiring the combined value below 5 GeV.The underlying-event estimate uses a rectangular η–φ region centered on the photon and excludes the isolation cone.Source: §6 · p. 8
• Table 2 reports stepwise isolated-photon identification efficiencies relative to the preceding step, with selections more efficient at high photon ET.Source: Table 2 · p. 8, §6 · p. 8

7 Signal extraction, corrections, and systematic uncertainties

• The shower-shape variable σηη is a modified second moment of cluster energy around its mean η position; isolated photons have smaller means and narrower distributions than decay photons.The calculation uses a 5 × 5 crystal group centered on the highest-energy crystal.Source: §7 · pp. 8–9
• A binned maximum-likelihood fit to σηη estimates the isolated prompt-photon yield, with electron contamination subtracted separately.PbPb signal shapes use PYTHIA+DATA samples, while the background shape comes from an isolation sideband.Source: §7 · p. 9
• Raw yields receive energy-smearing corrections of 1.00–1.08, then become per-event PbPb yields or luminosity-normalized pp cross sections.Source: §7 · p. 9
• Total systematic uncertainties are 22–30% for PbPb and 14–16% for pp isolated-photon spectra.Table 3 summarizes the contributions and totals.Source: Table 3 · p. 10, §7 · p. 10
• Background modeling dominates the PbPb yield uncertainty, with an estimated 12–23% contribution from variations in the background shower-shape model.Source: §7 · p. 10
• Photon-shape, energy-scale, smearing, fit, electron-contamination, isolation-definition, and event-count or luminosity uncertainties are also propagated.The respective uncertainties are 1–5%, 9%, 4%, 5%, 3–6%, 2–7%, and 3% for PbPb event counts or 6% for pp luminosity.Source: §7 · pp. 10–11

8 Theoretical calculations

• JETPHOX 1.2.2 NLO pQCD spectra are compared with measured pp and PbPb isolated-photon spectra and reproduce earlier pp data at 7 TeV [11] [16].Source: §8 · p. 11
• pp calculations use CT10 [19], while PbPb calculations use EPS09 [20] with isospin effects; both use BFG-II fragmentation functions [21] and scales set to photon ET.Source: §8 · p. 11
• Parton isolation requires less than 5 GeV inside a cone of radius 0.4; scale uncertainty is ±(15–10)% and proton-PDF uncertainty is ±(7–5)% in the measured range.Source: §8 · p. 11
• Nuclear-PDF uncertainty is evaluated using 30 EPS09 eigenvalues, and PbPb spectra are also calculated with nDS [22] and HKN07 [23].Source: §8 · p. 11

9 Results

• TAA provides the normalization for comparing PbPb yields with pp cross sections and represents the NN-equivalent integrated luminosity at a given centrality.Source: §9 · p. 11
• The pp spectra and TAA-scaled PbPb yields agree with NLO JETPHOX calculations at all measured ET; Table 4 lists the cross sections and yields with uncertainties.The table covers |ηγ| < 1.44 and ET bins.Source: §9 · p. 11, Table 4 · p. 12
• RAA is compatible with unity across measured ET in 0–10% central PbPb and remains consistent with unity across centralities and photon ET bins.The centrality comparison uses Npart and spans five photon ET intervals.Source: §9 · p. 12
• The NLO PbPb-to-pp ratios from EPS09, nDS, and HKN07 differ by at most ±10%, and the EPS09 uncertainty band is consistent with measured RAA.Source: §9 · p. 12

10 Summary

• The measured pp and PbPb isolated-photon spectra are well reproduced by NLO perturbative QCD calculations with recent proton and nuclear PDFs.Figure 5 presents the spectra versus photon ET, including pp cross sections, TAA-scaled PbPb yields, and NLO predictions.Source: §10 · pp. 12–13, Figure 5 · p. 13
• PbPb yields scaled by TAA show no modification relative to pp across the measured centralities, consistent with unity in the RAA comparisons.Figure 6 shows central collisions; Figure 7 compares centrality and five photon ET intervals.Source: §10 · p. 12, Figure 6 · p. 13, Figure 7 · p. 14
• The result is consistent with nuclear parton densities not being significantly modified relative to proton PDFs in the explored kinematic range, dominated by large-Q2 and moderate-x photons.Source: §10 · p. 15
• Isolated photons are reported to be unaffected by the produced strongly interacting medium, unlike jets, and provide an initial-state probe and a baseline for studying in-medium parton energy loss in γ+jet events.Source: §10 · p. 15