University of Birmingham Production of charged pions, kaons and protons at large transverse momenta in pp and Pb–Pb collisions atsNN = 2.76 TeV

Transverse momentum spectra of π ± , K ± and p ( ¯ p ) up to p T = 20 GeV / c at mid-rapidity in pp, peripheral (60–80%) and central (0–5%) Pb–Pb collisions at √ s NN = 2 . 76 TeV have been measured using the ALICE detector at the Large Hadron Collider. The proton-to-pion and the kaon-to-pion ratios both show a distinct peak at p T ≈ 3 GeV / c in central Pb–Pb collisions. Below the peak, p T < 3 GeV / c , both ratios are in good agreement with hydrodynamical calculations, suggesting that the peak itself is dominantly the result of radial ﬂow rather than anomalous hadronization processes. For p T > 10 GeV / c particle ratios in pp and Pb–Pb collisions are in agreement and the nuclear modiﬁcation factors for π ± , K ± and p ( ¯ p ) indicate that, within the systematic and statistical uncertainties, the suppression is the same. This suggests that the chemical composition of leading particles from jets in the medium is similar to that of vacuum jets.


Introduction
Heavy-ion collisions at ultra relativistic energies produce a new form of QCD matter characterized by the deconfined state of quarks and gluons (partons). Measurements of the production of identified particles in Pb-Pb collisions, relative to pp collisions, provide information about the dynamics of this dense matter. In pp collisions, high transverse momentum (p T > 2 GeV/c) hadrons are produced from fragmentation of jets that can be calculated folding the perturbative QCD calculations for jets with universal fragmentation functions determined from data such as those reported here. The bulk production of particles at lower p T is non-perturbative and requires phenomenological modeling. In heavy-ion collisions the production can be affected by the medium in several different ways. In particular there is an intermediate transverse momentum regime, 2 < p T < 8 GeV/c, where the baryon-to-meson ratios, e.g. the proton yield divided by the pion yield, measured by experiments at RHIC revealed a, so far, not well understood enhancement [1-3]. This so-called "baryon anomaly" could indicate the presence of new hadronization mechanisms such as parton recombination [4-6] that could be significantly enhanced and/or extended out to higher p T at LHC due to larger mini-jet production [7]. For transverse momenta above 10 GeV/c one expects to be able to study the pure energy loss (jet quenching) of high p T scattered partons traversing the medium [8][9][10]. This affects * alice-publications@cern.ch. the inclusive charged particle p T spectrum as has been seen at RHIC [11,12] and over an extended p T range, up to 100 GeV/c, at the LHC [13,14]. The additional information provided by particle identification (PID) is of fundamental interest to study the differences in the dynamics of fragmentation between quarks and gluons to baryons and mesons [15], and also to study the differences in their interaction with the medium considering that, due to the color Casimir factor, gluons lose a factor of two more energy than quarks [16,17]. The results presented in this Letter address three open experimental questions: Are there indications that the kaons are affected by radial flow at intermediate p T ? Does the baryonto-meson ratio return to the pp value for high p T (>10 GeV/c) as suggested by the recent publication of the Λ/K 0 S ratio [18]? Are there large particle species dependent jet quenching effects as predicted in several models [19][20][21], where measurements at RHIC, in particular for baryons, are inconclusive due to the limited p T -range and the large systematic and statistical uncertainties [22-24]?

Data analyses
In this Letter we present the measurement of the production of pions (kaons and protons) from a p T of a few hundred MeV/c up to p T = 20 GeV/c in √ s NN    The main focus in the following will therefore be on explaining the analysis details for the HMPID and the high p T dE/dx analysis.
The pp analyses use 40 × 10 6 and the Pb-Pb minimum bias analysis uses 11 × 10 6 collision events. The HMPID analysis used the 2011 centrality triggered Pb-Pb data with around 4.1 × 10 6 0-5% central collision events. Data were taken during 2010 and 2011 under conditions where pileup effects were negligible. Minimum bias interactions are triggered based on the signals from forward scintillators (V0) and, in pp collisions, the two innermost silicon pixel layers of the ITS (SPD). The trigger efficiency is 88.1% for pp inelastic collisions [28] and 97.1% for non-diffractive Pb-Pb collisions [29]. The Pb-Pb collision centrality is determined from the measured amplitude in the V0 detector [30] which is related to the number of participating nucleons and the nuclear overlap function (T AA ) through simulations based on a Glauber model [29]. The same event and track selection is used as in the inclusive charged particle analysis [31]. Track cuts are optimized in order to select primary charged particles in the pseudorapidity range |η| < 0.8 and all results presented in this paper are corrected for feed-down from weak decays. As listed in Table 1 the low p T analysis is done for |y| < 0.5, while the high p T analysis is done for |η| < 0.8, to take advantage of the full statistics, and the final spectra are then normalized to the corresponding rapidity intervals, see Eq. (1) below.
2.1. Identified particle spectra at low p T The pp low p T analysis relies on the combination of four almost independent PID techniques, named after the detectors involved: ITS-sa, TPC-TOF, TOF and HMPID. The techniques have complementary p T ranges listed in Table 1.
The ITS-sa analysis exploits stand-alone (sa) tracks reconstructed in the ITS to be able to go as low in p T as possible. The identification is done based on dE/dx measurements in up to 4 of the 6 silicon layers. This information is combined in a Bayesian approach using a set of priors determined with an iterative procedure, and the track identity is assigned according to the highest probability. The minor residual contamination due to misidentification is less than 10% in the p T -range reported in Table 1 and corrected for using MC.
The other three analyses all use global tracks reconstructed in both the ITS and the TPC. The TPC-TOF analysis is optimized to combine the information from the TPC and TOF. The identification is based on a three standard deviations agreement with the expected detector signal and resolution (3σ ) in the TPC dE/dx and for p T > 0.6 GeV/c a 3σ requirement is also applied for the timeof-flight provided by the TOF detector. The TOF analysis identifies particles comparing the measured time-of-flight from the primary vertex to the TOF detector, t tof , and the time expected under a given mass hypothesis, t exp i (i = π, K, p). The TOF standalone analysis is optimized for handling momentum regions where the separation is challenging. The precise signal shape for t tof −t exp i , including an exponential tail, is used, and the yield in a given p T interval is obtained by fitting.
The HMPID [32,33] is designed as a single-arm proximityfocusing Ring Imaging CHerenkov (RICH) detector where the radiator is a 15 mm thick layer of liquid C 6 F 14 (perfluorohexane). It is located at about 5 m from the beam axis, covering a limited acceptance of |η| < 0.55 and 1.2 • < ϕ < 58.5 • . The PID in the HMPID is done by measuring the Cherenkov angle, θ ch . In the reconstruction, the tracks are propagated to the HMPID detector and associated with a MIP signal. A Hough Transform Method (HTM) [34] is used to discriminate the signal from the background. For a given track, the mean Cherenkov angle is computed as the weighted average of the single photon angles selected by the HTM. The Cherenkov angle distribution is then fitted to obtain the yields, see Fig. 1 for an example of fits in the Pb-Pb and pp analysis.
The raw yields measured by each analysis are corrected for the reconstruction, selection, PID efficiency, and misidentification probability. The contamination due to particles from weak decays of light flavor hadrons and interactions with the material is subtracted using MC-template fits of the distance-of-closest approach distributions [26]. Finally the raw spectra are corrected for the detector acceptance, trigger selection, vertex and track reconstruction efficiency.
The systematic uncertainties for the ITS, TPC, and TOF analyses are obtained in essentially the same way as reported in [27,26]. The systematic uncertainty for the HMPID analysis has contributions from tracking and PID. These uncertainties have been estimated by changing individually the track selection cuts and the parameters of the fit function used to extract the raw yields by ±10%. In addition, the uncertainty of the association of the track to the MIP signal is obtained by varying the value of the distance cut required for the match.
The HMPID analysis in Pb-Pb collisions is analogous to the pp analysis except for the treatment of the background. In central Pb-Pb collisions, where the total number of hits in the HMPID chambers is large, it is possible that a Cherenkov ring is constructed based on hits incorrectly associated with the track. Fig. 1 gives examples of the reconstructed Cherenkov angle distributions in a narrow p T interval. In pp collisions (right panel) the reconstructed angle distribution is fitted by a sum of three Gaussian distributions, corresponding to the signals from pions, kaons and protons. In the case of Pb-Pb collisions (left panel) the additional background distribution is modeled with a 6th order polynomial found to minimize the reduced χ 2 of the fit. The shoulder in the background distribution starting at 0.7 rad is a boundary effect due to the finite chamber geometrical acceptance that is also observed in MC simulations. The fitting is done in two steps, where the width and the mean of each Gaussian distribution are free parameters in the first step and are then used to obtain a p T dependent parameterization. This parameterization is used to constrain the parameters in the second final fit. The means and widths constrained in this way are found to be independent of centrality. Finally we note that the background increases with the Cherenkov angle because the fiducial area used in the reconstruction becomes larger, making it more likely to associate spurious hits with the signal.
The PID efficiency has been evaluated from a Monte Carlo simulation that reproduces well the background in the data. A datadriven cross check of the efficiency has been performed using a clean sample of protons and pions from Λ and K 0 S decays identified in the TPC based on their topological decay.
To estimate the uncertainty due to the incomplete knowledge of the shape of the background distribution, an alternative background function, depending on tan(θ ch ) and derived from geometrical considerations in case of orthogonal tracks [32], has been used. The corresponding systematic uncertainty reaches the maximum value at low momenta for the most central collisions (∼15% for pions, and ∼8% for kaons and protons). The systematic uncertainty decreases with p T as the track inclination angle in the bending plane decreases so that the fiducial area for the Cherenkov pattern search is smaller.

Identified particle spectra at high p T
Particle identification is performed in the relativistic rise regime of the Bethe-Bloch (BB) curve where the dE/dx separation between particles with different masses is nearly constant [35]. The dE/dx is obtained as the truncated mean of the 0-60% lowest charge samples associated with the track in the TPC [36]. The dE/dx response depends on the track length so the analysis is done in four equally sized |η|-intervals, and a geometrical cut to remove tracks entering the gap in between the TPC readout chambers is applied to select tracks with the best dE/dx resolution. The separation in number of standard deviations (σ ) between pions and kaons (pions and protons) in pp and peripheral Pb-Pb collision is around 3.2 (4.6) at momentum p ≈ 6 GeV/c for 0.6 < |η| < 0.8 where the separation is largest. In central Pb-Pb collisions one finds a separation of 2.4σ (3.5σ ). In the worst case, |η| < 0.2, the separation is 11-15% smaller. Fig. 2 shows examples of the dE/dx spectra obtained for pp and Pb-Pb (central and peripheral) collisions for two momentum, p, intervals and |η| < 0.2 where p ≈ p T . The pion, kaon, and proton yields are extracted by fitting a sum of four Gaussian functions (including electrons) to the dE/dx spectra. 1 To reduce the degrees of freedom in the fits from 12 to 4, parameterizations of the BB ( dE/dx ) and resolution (σ ) curves as a function of βγ are extracted first using tracks from identified particles. Samples of secondary pions (30 < βγ < 50) and protons (3 < βγ < 7) are obtained through the reconstruction of the weak-decay topology of K 0 S and Λ, respectively; a similar algorithm is used to identify electrons resulting from photon conversions (fixing the dE/dx plateau: βγ > 1000). Finally, using information from the timeof-flight detector the relative pion content can be enhanced for sub-samples of the full datasets (16 < βγ < 50).
The dE/dx separation between kaons and protons in the high p T analysis is smallest for p ≈ 3 GeV/c and increases with p until both species are on the relativistic rise [35]. In central collisions the dE/dx separation is the lowest and the systematic uncertainties on the extracted yields are correspondingly large as discussed later, see Table 2. Hence, to improve the central val- 1 We note that muons from heavy flavor decays are subtracted from the pions based on the measured electron yields and that contamination from deuterons and tritons are negligible ( 1%).

Table 2
Systematic uncertainties, separated into the N ch , PID, and efficiency part, on the invariant yields from 3 < p T < 4 GeV/c (left quoted value) to 10 < p T < 20 GeV/c (right quoted value).
c Same for all centralities and all particle species.
ues for the kaons and protons, the K 0 S yields [18] are used as a proxy for the charged kaons to further constrain the BB curve in Pb-Pb collisions in a procedure which uses a two dimensional fit of dE/dx vs momentum. The effect of the K 0 S bias is only relevant in central collisions at low p T (<4 GeV/c). At 3 GeV/c the effect on the extracted kaon yield is an increase of 10% (<1%) for 0-5% (60-80%) collision centrality.
With the above information the BB and the resolution curves are determined for kaons and protons in the full momentum interval reported here and for pions with p < 7 GeV/c. For p > 7 GeV/c the pion dE/dx is restricted by the logarithmic rise until the dE/dx starts to approach the plateau. This lack of additional constraint currently limits the p T reach of the analysis to ∼20 GeV/c. From the fits in Fig. 2 the particle fractions, f π /K/p (p) are extracted. The fraction in a p T bin, f π /K/p (p T ), is obtained as the weighted average of the contributing momentum (p) bins. The p T -dependent fractions are found to be independent of η and so all four η regions are averaged.
Finally, the invariant yields are obtained using the p T spectrum for inclusive charged particles [31], d 2 N ch dp T dη , in the following way: d 2 N π /K/p dp T d y where ( ch ) π /K/p is the efficiency for (un)identified particles and J π /K/p is the Jacobian correction (from η to y). Normalizing to the p T spectrum of inclusive charged particles guarantees that only the systematic uncertainty due to PID is relevant when comparing the modification of the p T spectra of π/K/p to those for the unidentified particles. The p T resolution is around 5% at p T = 20 GeV/c and the p T spectra have been corrected for this resolution using an unfolding procedure for p T > 10 GeV/c [31,37]. This correction is less than 2% at p T = 20 GeV/c.

Systematic uncertainties
The systematic uncertainty on the invariant yields has three main components: event and track selection, efficiency correction of the fractions, and the fraction extraction. Contributions from the event and track selection are taken directly from the inclusive charged particle analysis [31]. Efficiency ratios ( ch / π /K/p ) are found to be nearly independent of p T (a small dependence is only observed for kaons), similar for all systems, and model independent within 3%. The largest systematic uncertainty in the extraction of the fractions comes from the uncertainty in the constrained parameters: the means ( dE/dx ) and the widths (σ ) used in the fits. The uncertainty on these parameters are estimated from the average difference between the final parameterizations and the data points obtained from the enhanced samples with identified particles. In addition, the statistical uncertainty on the extracted BB parameterization in peripheral Pb-Pb collisions is found to be of a similar magnitude and also taken into account in the following variations. The dE/dx spectra are then refitted, varying the means and the widths within the estimated uncertainties, and the variation of the fractions are assigned as systematic errors. In this way the correlations in the systematic uncertainty for the particle ratios can be directly included. A summary of the PID systematic uncertainties is shown in Table 2. The N ch systematic uncertainties cancel in the particle ratios.

Results and discussion
The measurement of charged pion, kaon and (anti-)proton transverse momentum spectra has been performed via several independent analyses, each one focusing on a sub-range of the total p T distribution, with emphasis on the individual detectors and specific techniques to optimize the signal extraction. The results were combined using the independent systematic uncertainties as weights in the overlapping ranges (a 3% common systematic uncertainty due to the TPC tracking is not in the weight but added directly to the combined spectrum). The statistical uncertainties are much smaller and therefore neglected in the combination weights. For p T > 4 GeV/c only the high p T analysis is used for all species. Fig. 3 shows the ratio of individual spectra to the combined spectrum for the 0-5% central Pb-Pb data, illustrating the compatibility between the different analyses. Fig. 4 shows the invariant yields measured in Pb-Pb collisions compared to those in pp collisions scaled by the number of binary collisions, N coll [29] obtained for the measured pp cross section [28]. For peripheral Pb-Pb collisions the shapes of the invariant yields are similar to those observed in pp collisions. For central Pb-Pb collisions, the spectra exhibit a reduction in the production of high-p T particles with respect to the reference which is characteristic of jet quenching.   At LHC energies the mini-jet activity is expected to be larger than at RHIC energies, which motivated ratio predictions in the framework of recombination models where shower partons in neighboring jets can recombine to be an order of magnitude larger than the measurements reported here [7]. Other predictions where recombination only occurs for soft thermal radially flowing partons are, as shown in the figure, more consistent with the data [4]. The surprising new result is that in central Pb-Pb collisions the (K + + K − )/(π + + π − ) ratio also exhibits a bump at p T ≈ 3 GeV/c. This has not been observed at RHIC (this could be due to limitations in precision in this p T region) but is also observed in the soft coalescence model [4]. The Kraków [38] hydrodynamical model captures the rise of both ratios quantitatively well, while a similar model, HKM [39] that is not shown, does slightly worse. The EPOS [40] event generator which has both hydrodynamics, but also the high p T physics and special hadronization processes for quenched jets [41] qualitatively well describes the data but tends to overestimate the peaks. The recent result [42] that for p T < 3 GeV/c the shape of the phi-to-pion ratio is consistent with the proton-to-pion ratio, reported here, taken together with the model comparisons shown in Fig. 5 indicate that the peak is mainly dominated by radial flow (the masses of the hadrons).
For higher p T (>10 GeV/c) both particle ratios behave like those in pp, suggesting that fragmentation dominates the hadron production. In this p T regime, the particle ratios in pp are not well described by the pQCD calculations in [43]. It was recently shown [44] that in general the fragmentation functions for gluons are badly constrained, leading to disagreement of up to a factor 2 with N ch spectra measured at LHC. Furthermore it was pointed out that data with p T > 10 GeV/c, as reported here, are needed to reduce the scale dependence that seems to be the origin of the disagreement. Fig. 6 shows the nuclear modification factor R AA as a function of p T defined as the ratio of the Pb-Pb spectra to the N coll scaled pp spectra shown in Fig. 4. The R AA for the sum of kaons and protons is included as it allows the most precise quantitative comparison to the R AA of pions. For p T < 10 GeV/c protons appear to be less suppressed than kaons and pions, consistent with the particle ratios shown in Fig. 5. At larger p T (>10 GeV/c) all particle species are equally suppressed; so despite the strong energy loss observed in the most central heavy-ion collisions, the particle composition and ratios at high p T are similar to those in vacuum.
The models cited in the introduction all suggest large differences, of 50% or more, between the suppression of different species that are either related to mass ordering or baryon-vsmeson effects. The differences are naturally large in these scenarios because they are directly related to the large suppression. To quantify the similarity of the suppression the double R AA ratios, e.g. R p+p AA /R π + +π − AA , are inspected. Fig. 7 shows the double ratios constructed using the particle ratios to properly handle that the dominant correlated systematic uncertainties are between particle species and not between different collision systems. We  note that a similar ratio for protons and pions made with the STAR data [22,23] would give a flat ratio for p T > 3 GeV/c of approximately 3 ± 2. The results disfavor significant modifications of hadro-chemistry within the hard core of jets, as predicted based on medium modified color flow which introduces a mass ordering of the fragmentation [19], or due to changes in the color structure of the quenched probe which could enhance baryon production [20]. The data also contradict predictions where fragmentation into color neutral hadrons, assumed to have no energy loss after formation, occurs in the medium and the formation time scales directly with the hadron mass [21].

Conclusions
The production of pions, kaons and protons has been measured in pp and central and peripheral Pb-Pb collisions up to high p T . From the invariant yields we derived the particle ratios and the R AA as a function of p T . We observe that the proton-to-pion and the kaon-to-pion ratios both exhibit a peak and that at low p T the rise of both ratios can be well described by hydrodynamic calculations. This rules out models where shower partons recombine and sets strong constraints for soft recombination models. At higher-p T , both ratios are compatible with those measured in pp collisions. From the nuclear modification factor R AA , we conclude that for p T > 10 GeV/c within the systematic and statistical uncertainties, pions, kaons and protons are suppressed equally. This rules out ideas in which the large energy loss leading to the suppression is associated with strong mass ordering or large fragmentation differences between baryons and mesons. The results presented here establish strong constraints on theoretical modeling for fragmentation and energy loss mechanisms.

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.
The ALICE Collaboration gratefully acknowledges the resources and support provided by all Grid centres and the Worldwide LHC Computing Grid (WLCG) Collaboration.
The ALICE Collaboration acknowledges the following funding agencies for their support in building and running the ALICE de-