Particle-yield modification in jet-like azimuthal di-hadron correlations in Pb-Pb collisions at $\sqrt{s_{\rm NN}}$ = 2.76 TeV

The yield of charged particles associated with high-$p_{\rm T}$ trigger particles ($8<p_{\rm T}<15$ GeV/$c$) is measured with the ALICE detector in Pb-Pb collisions at $\sqrt{s_{\rm NN}}$ = 2.76 TeV relative to proton-proton collisions at the same energy. The conditional per-trigger yields are extracted from the narrow jet-like correlation peaks in azimuthal di-hadron correlations. In the 5% most central collisions, we observe that the yield of associated charged particles with transverse momenta $p_{\rm T}>3$ GeV/$c$ on the away-side drops to about 60% of that observed in pp collisions, while on the near-side a moderate enhancement of 20-30% is found.

vertex is used to select primary track candidates and to constrain the p t of the track.
In this analysis 14 million minimum-bias Pb-Pb events recorded in fall 2010 at √ s NN = 2.76 TeV as well as 37 million pp events from March 2011 ( √ s = 2.76 TeV) are used. These include only events where the TPC was fully efficient to ensure uniform azimuthal acceptance. Events are accepted which have a reconstructed vertex less than 7 cm from the nominal interaction point in beam direction. Tracks are selected by requiring at least 70 (out of up to 159) associated clusters in the TPC, and a χ 2 per space point of the momentum fit smaller than 4 (with 2 degrees of freedom per space point). In addition, tracks are required to originate from within 2.4 cm (3.2 cm) in transverse (longitudinal) distance from the primary vertex.
For the measurement of I AA and I CP the yield of associated particles per trigger particle is studied as a function of the azimuthal angle difference ∆ϕ. This distribution is given by 1/N trig dN assoc /d∆ϕ where N trig is the number of trigger particles and N assoc is the number of associated particles. We measure this quantity for all pairs of particles where p t,assoc < p t,trig within |η| < 1.0 as a function of p t,assoc . Pair acceptance corrections have been evaluated with a mixed-event technique but found to be negligible for the yield ratios due to the constant acceptance in ϕ and the same detector conditions for the different data sets.
Corrections for detector efficiency (17-18% depending on collision system, p t and centrality) and contamination (4-8%) by secondary particles from particle-material interactions, γ conversions and weakdecay products of long-lived particles are applied for trigger and associated particles, separately. Addi-tional secondary particles correlated with the trigger particle are found close to ∆ϕ = 0 in particular due to decays and γ conversions. We correct for this contribution (2-4%). These corrections are evaluated with the HIJING 1.36 [20] Monte Carlo (MC) generator which was tuned to reproduce the measured multiplicity density [19] for Pb-Pb and the PYTHIA 6 [21] MC with tune Perugia-0 [22] for pp using in both cases a detector simulation based on GEANT3 [23]. MC simulations underestimate the number of secondary particles. Therefore, we study the distribution of the distance of closest approach between tracks and the event vertex. The tail of this distribution is dominantly populated by secondary particles and the comparison of data and MC shows that the secondary yield in MC needs to be increased by about 10% (depending on p t ). An MC study shows that effects of the event selection and vertex reconstruction are negligible for the extracted observables. The correction procedure was validated by comparing corrected simulated events with the MC truth. Figure 1a shows a typical distribution of the corrected per-trigger pair yield before background subtraction. The fact that the ∆ϕ distribution is flat outside the near-and away-side region gives us confidence that the background can be estimated with the zero yield at minimum (ZYAM) assumption [24]. This procedure estimates the pedestal value by fitting the flat region close to the minimum of the ∆ϕ distribution (|∆ϕ − π/2| < 0.4) with a constant. The validity of the ZYAM assumption has been questioned in cases where collective effects dominate [25,26]; however, for the high-p t correlations of this analysis, the narrow width and large amplitude of the correlated signal compared to the flow modulation drastically reduce the ZYAM bias. Therefore, we define the integrated associated yield as the signal over a flat background. Figure 1b illustrates the background determination. Also indicated is a background shape accounting for elliptic flow v 2 , the second coefficient of the particle azimuthal distribution measured with respect to the reaction plane. It is given by 2v 2,trig v 2,assoc cos 2∆ϕ where v 2,trig (v 2,assoc ) is the elliptic flow of the trigger (associated) particles. The v 2 values are taken from an independent measurement [27] of v 2 up to p t = 5 GeV/c. As an upper limit we use the measured v 2 for p t = 5 GeV/c also for larger p t where v 2 is expected to decrease. For the centrality class 60-90% no v 2 measurement is available, therefore, as an upper limit, v 2 is taken from the 40-50% centrality class. Since v 2 decreases from mid-central to peripheral collisions and the flat pedestal assumes v 2 = 0, this includes all reasonable values of v 2 .
Contributions from ∆η-independent correlations (e.g., due to flow harmonics at all orders) can also be removed on the near-side (where the jet peak is centered around ∆η ≈ 0) by calculating the per-trigger pair yield in the region |∆η| < 1 and subtracting the contribution from 1 < |∆η| < 2 normalized for the acceptance. This prescription, which we call the η-gap method, provides a measurement independent of the flow strength.
In Fig. 1c the flat-pedestal subtracted distributions of central and peripheral Pb-Pb collisions are compared to that of pp collisions. The integral over those distributions in the region where the signal is significantly above the background, i.e., within ∆ϕ of ±0.7 and π ± 0.7 results in the near-and awayside yields per trigger particle (Y ), respectively. This procedure samples the same fraction of the signal in Pb-Pb and pp collisions, since in the p t -range used for this study the width of the peaks is similar for both systems. The yields are used to compute the ratio

Systematic Uncertainties
The uncertainty from the pedestal determination has been estimated by comparing different pedestal evaluation strategies (see Fig. 1b). The constant-fit region has been shifted and an average of the 8 (out of 36) lowest ∆ϕ points has been used. The integration window for the near-and away-side has been varied between ±0.5 and ±0.9. The effect of detector efficiency and track selection has been studied by systematically varying the track cuts. Track splitting and merging effects were assessed by studying the tracking performance as a function of the distance of closest approach of the track pairs in the detector volume. A bias due to the p t resolution on the extracted yields was evaluated by folding the detector resolution with the extracted associated spectrum and found to be negligible. The sensitivity of the corrections to details of the MC has been studied by varying the particle composition, the material budget and the MC generator (using AMPT [28] for Pb-Pb and PHOJET [29] for pp). Uncertainties in the centrality determination were evaluated by comparing results obtained with the different centrality estimates from the VZERO, the SPD and ZDCs. Table 1 lists the size of the different contributions to the systematic uncertainties for I AA and I CP as well as their sum in quadrature. Figure 2a shows the yield ratio I AA for central (0-5% Pb-Pb/pp) and peripheral (60-90% Pb-Pb/pp) collisions using the three background subtraction schemes discussed. The fact that the only significant difference between the different background subtraction schemes is in the lowest bin of p t,assoc confirms the assumption of only a small bias due to flow anisotropies in this p t region. The influence of higher flow harmonics [27] on the background shape can be explicitly estimated: including v 3 , v 4 and v 5 from [27] changes the extracted jet yield by less than 1%, except for the first bin in p t,assoc in the most central collisions where it is about 8%. This is consistent with the difference between the data points labeled v 2 bkg and η-gap where the latter includes flow at all orders. In central collisions, an away-side suppression (I AA ≈ 0.6) is observed which is evidence for in-medium energy loss. Moreover, there is an enhancement above unity of 20-30% on the near-side which has not been observed with any significance at lower collision energies at these momenta [8]. In peripheral collisions, both the near-and away-side I AA measurements approach unity, as expected in the absence of significant medium effects. Figure 2b shows the yield ratio I CP . As for I AA , the influence of the flow modulation is small and only significant in the lowest p t,assoc bin. I CP is consistent with I AA in central collisions with respect to the near-side enhancement and the away-side suppression.

Results
Comparing this measurement and R AA to models simultaneously will constrain energy-loss mechanisms and model parameters. Robust conclusions can only be drawn with a systematic comparison of multiple observables with calculations spanning the parameter space and cannot be done with current calculations (e.g. [30]). Such a study is beyond the scope of this letter.
Comparison to RHIC Similar measurements have been performed at RHIC. Although the same range in p t,trig does not necessarily probe the same parton p t region at different √ s, we assess changes from RHIC to LHC in the following. The STAR measurement [8] (which includes only statistical uncertainties) of the near-side I AA is consistent with unity, albeit with a large uncertainty (18-40%). On the away-side the result from STAR is about 50% lower than the results shown in Fig. 2. We also calculated I AA for the 20% most central events to compare to PHENIX [7] (only v 2 -subtracted data on the away-side available). For p t,assoc < 4 GeV/c, the flow influence in this centrality interval is about 75%, too large to provide a reliable measurement. For 4 < p t,assoc < 10 GeV/c, the v 2 -subtracted I AA is 0.5 − 0.6 ± 0.08. This result is slightly larger than results from PHENIX in a similar p t,trig -region of 7 < p t,trig < 9 GeV/c: 0.31 ± 0.07 and 0.38 ± 0.11 for p t,assoc ≈ 3.5 GeV/c and 5.8 GeV/c, respectively. Based on an analysis in a lower p t -region, where collective effects are significantly larger than in the measurement presented here, the STAR collaboration mentions a slightly enhanced jet-like yield in Au-Au compared to d-Au collisions, but does not assess the effect quantitatively [31]. In conclusion, the observed away-side suppression at the LHC is less than at RHIC (I AA is larger), while the single-hadron suppression R AA is found to be slightly larger (R AA is smaller) than at RHIC [9].
Near-Side Enhancement These measurements represent the first observation of a significant near-side enhancement of I AA and I CP in the p t region studied. This enhancement suggests that the near-side parton is also subject to medium effects.
I AA is sensitive to (i) a change of the fragmentation function, (ii) a possible change of the quark/gluon jet ratio in the final state due to the different coupling to the medium and (iii) a bias on the parton p t spectrum after energy loss due to the trigger particle selection. If the fragmentation function (FF) is softened in the medium, hadrons carry a smaller fraction of the initial parton momentum in Pb-Pb collisions as compared to pp collisions. Therefore, hadrons with a given p t originate from a larger average parton momentum which may lead to more associated particles and I AA > 1. An increased fraction of gluon (quark) jets has a similar effect than softening (hardening) of the FF and leads to I AA > 1 (< 1).
A different parton distribution in pp and Pb-Pb collisions can modify I AA even if fragmentation of a given parton after energy loss is unmodified. In particular, in the same transverse momentum region, we see a strong suppression of the trigger particles (R AA ≈ 0.2) and the rising slope of R AA (p t ) [9]. A similar suppression should apply to partons, leading to a parton distribution after energy loss which is biased towards higher parton p t . Therefore, for a fixed trigger p t , the mean parton p t would be larger in Pb-Pb than in pp, leading to an increase in I AA . This argument can be quantified with the hadronpair suppression factor J AA [32]. J AA (p t,trig , p t,assoc ) = R AA (p t,trig )I AA (p t,trig , p t,assoc ) is approximately R AA (p t,trig + p t,assoc ) in this case, and with a rising R AA leads to I AA > 1.
It is likely that all three effects play a role, and following the above arguments, we note that the combined measurement of R AA and I AA is sensitive to the interplay of energy loss and the change of the fragmentation pattern in the medium.
In summary, the modification of the per-trigger yield of associated particles, I AA and I CP , has been extracted from di-hadron correlations in Pb-Pb collisions at √ s NN = 2.76 TeV. In central collisions, on the away-side, suppression (I AA ≈ 0.6) is observed as expected from strong in-medium energy loss. On the near-side, a significant enhancement (I AA ≈ 1.2) has been reported for the first time. Along with the measurement of R AA , I AA provides strong constraints on the quenching mechanism in the hot and dense matter produced.

Acknowledgements
The ALICE collaboration would like to thank all its engineers and technicians for their invaluable contributions to the construction of the experiment and the CERN accelerator teams for the outstanding performance of the LHC complex.