Observation of the jet diffusion wake using dijets in heavy ion collisions

Abstract: Energetic quarks and gluons traversing a hot and dense quark-gluon plasma deposit energy and momentum into the medium before hadronizing to collimated sprays of particles, known as jets. This energy-momentum deposition is expected to produce medium responses, collectively known as jet wakes, with ``diffusion wake'' denoting a depletion of particles in the direction opposite to the propagating jet. These phenomena are studied by comparing dijet-hadron correlations measured in lead-lead (PbPb) and proton-proton (pp) collisions to assess jet-induced modifications of bulk particle production. The analysis uses PbPb and pp data recorded at a nucleon-nucleon center-of-mass energy $\sqrt{s_\mathrm{NN}}$ = 5.02 TeV with the CMS detector at the CERN LHC. By exploring how the dijet-hadron correlation distributions differ for various pseudorapidity separations of the two jets in the dijet, the presence of a jet diffusion wake is firmly established. The wake has a significance greater than 5 standard deviations for charged particles in the transverse momentum range 1 $\lt$ $p_\mathrm{T}$ $\lt$ 2 GeV. The measurements are compared with various model predictions with and without jet wake effects, providing new insights into quark-gluon plasma properties and the formation of jet-induced wakes.

• The paper observes the QGP jet diffusion wake at greater than five standard deviations in central PbPb collisions by subtracting small-rapidity-gap dijet correlations from large-gap samples.
• The analysis finds a stronger depletion for 1–2 GeV charged particles, central collisions, and larger dijet rapidity separations, matching expected signatures of medium response.
• The paper shows that HYBRID and CoLBT-hydro models reproduce the wake’s location and trends but overestimate its magnitude, highlighting unresolved modeling of jet–medium interactions.

Motivation and analysis strategy

Energetic partons traversing the quark-gluon plasma (QGP) deposit energy and momentum into the medium, producing jet-induced medium responses collectively known as jet wakes. While the Mach cone signature has historically been difficult to isolate—because both the Mach cone and medium-induced modifications to the jet shower appear in the direction of jet propagation—the "diffusion wake," a depletion of particles opposite to the propagating jet, offers a cleaner observable. Prior experimental attempts were inconclusive: CMS reported first evidence using Z-boson–hadron correlations, exploiting the colorless boson to isolate the wake, while ATLAS found no significant evidence in photon-jet events. This paper reports the first observation of the diffusion wake using dijet events, following the theoretical proposal that rapidity-asymmetric dijets spatially separate the jet modification from the diffusion wake signal.

The key insight is geometric. Dijets are required to be back-to-back in azimuth ($|\Delta\varphi| > 7\pi/8$) but separated in pseudorapidity ($\Delta\eta$). With an enforced $\eta$ ordering such that the leading jet is always at larger $\eta$, the subleading jet's diffusion wake appears displaced toward negative $\Delta\eta$ in the near-side correlation region. For small dijet $\eta$ gaps ($\Delta\eta < 0.5$), the wake is hidden beneath the jet fragmentation peaks; for large gaps, the away-side peak shifts away and the subleading-jet wake manifests as a dip in the near-side distribution. The difference between correlations for large- and small-gap selections therefore isolates the wake in a data-driven manner.

Data and methodology

The analysis uses PbPb (0.66 nb⁻¹, 2018) and pp (299 pb⁻¹, 2017) data at $\sqrt{s_{NN}} = 5.02$ TeV with the CMS detector. Events are selected by a calorimeter trigger requiring a jet with $p_T > 80$ GeV (fully efficient above 130 GeV). Leading jets must satisfy $p_T^{\text{jet}_1} > 130$ GeV, $\Delta\eta$0; subleading jets require $\Delta\eta$1 GeV, $\Delta\eta$2. Dijet samples are binned into small gaps ($\Delta\eta$3) and three large-gap intervals up to $\Delta\eta$4.

Charged-particle tracks with $\Delta\eta$5 GeV and $\Delta\eta$6 GeV within $\Delta\eta$7 are correlated with the leading jet axis in $\Delta\eta$8 space. Corrections include tracking efficiency, per-track misreconstruction, and pair acceptance via mixed events (30 events per pairing, matched in vertex position within 0.5 cm and centrality within 0.5%). Notably, corrections for jet position resolution and selection biases are deliberately omitted because they cancel in the large-minus-small gap subtraction. The long-range uncorrelated background is estimated from the positive $\Delta\eta$9 sideband ($\eta$0), where no wake is expected given the $\eta$1 ordering, and subtracted separately for each sample. Detector response is corrected using PYTHIA 8 embedded in HYDJET minimum-bias PbPb events, passed through GEANT4, with centrality reweighting applied since HYDJET simulates only minimum-bias collisions.

Results

The difference distributions $\eta$2 for large minus small gaps show approximately flat behavior consistent with zero in pp collisions across all bins, while PbPb collisions exhibit a pronounced dip in the region $\eta$3, most prominent for low-$\eta$4 particles in central collisions. The depletion strengthens monotonically from peripheral (50–80%) through mid-central (30–50%) to central (0–30%) collisions; the pp and peripheral PbPb results are statistically consistent, whereas central PbPb data for $\eta$5 GeV lie significantly below the pp reference.

Several features support the diffusion wake interpretation:

• Centrality dependence: the depletion grows with collision centrality, as expected if the effect scales with medium density.
• Momentum dependence: the signal is stronger for $\eta$6 GeV than for $\eta$7 GeV, consistent with theoretical expectations that the wake primarily displaces soft bulk particles.
• Gap dependence: the mean position of the depletion shifts toward more negative $\eta$8 as the $\eta$9 selection increases from $\eta$0 to $\eta$1, and the integrated depletion strength increases with gap size, reflecting reduced contamination from medium-induced jet modifications at larger separations.

Quantitatively, the correlated yields integrated over $\eta$2 deviate from zero by more than five standard deviations for $\eta$3 GeV in 0–30% central collisions, for both $\eta$4 and $\eta$5, based on a $\eta$6 test. Since the paper argues that no mechanism other than the diffusion wake is known to produce such structures in the underlying-event distribution opposite to the propagating jet, this constitutes an observation-level claim rather than mere evidence.

Comparison with models

Three model predictions are confronted with the data: PYTHIA+HYDJET without quenching, which shows no $\eta$7 dependence as expected in the absence of jet-medium interactions; the hybrid strong/weak coupling model (HYBRID) in its fully incoherent energy loss scenario; and CoLBT-hydro, coupling linear Boltzmann transport to (3+1)D hydrodynamics. Both physics models capture the qualitative trend and the predicted location of the dip's mean position, but both overpredict the magnitude of the depletion, particularly for $\eta$8 GeV, where the observed enhancement near $\eta$9 is much smaller than the HYBRID prediction. CoLBT-hydro predictions are available only for 0–10% centrality, limiting direct comparison to the 0–30% measurement. The systematic overprediction suggests current treatments of the interplay between medium response and jet-induced modifications may overestimate the wake amplitude or underestimate competing effects such as out-of-cone radiation filling the depleted region.

Limitations and open questions

The extraction method has inherent constraints. The away-side correlation also shows a wake-like depletion, but the different positions of the away-side jet peaks in large- and small-gap samples prevent data-driven signal extraction on that side, so only the near-side projection is used. Residual underlying-event level mismatch between gap selections dominates the systematic uncertainty (absolute values up to 0.04), followed by tracking efficiency (up to 0.025) and pair-acceptance effects (up to 0.01). Medium-induced jet peak modifications and out-of-cone radiation can contaminate the wake signal for smaller gaps, motivating the emphasis on large-gap selections. Whether the quantitative tension with HYBRID and CoLBT-hydro predictions reflects missing physics in the energy-loss description, hydrodynamic modeling of the wake, or treatment of soft-particle transport remains unresolved by this measurement.

Conclusion

This measurement establishes the jet diffusion wake in the QGP at greater than 5 standard deviations significance using pseudorapidity-separated dijets, confirming the rapidity-asymmetry observable proposed theoretically as a robust wake signature. The signal exhibits the expected dependencies on centrality, charged-particle momentum, and dijet rapidity separation, while models incorporating jet-medium interactions reproduce trends but overestimate magnitudes. The result provides a new constraint on jet-induced medium response and, by extension, on QGP transport properties, though precise extraction of transport coefficients will require reconciling the observed wake strength with theoretical predictions.