WorksPhysics and Astronomy
Probing the Dependence of Partonic Energy Loss on the Initial Energy Density of the Quark Gluon Plasma
Abstract
Considerable evidence now exists for partonic energy loss due to interaction with the hot, dense medium created in ultra-relativistic heavy-ion collisions. A primary signal of this energy loss is the suppression of high transverse momentum hadron yields in A-A collisions relative to appropriately scaled collisions at the same energy. Measuring the collision energy dependence of this energy loss is vital to understanding the medium, but it is difficult to disentangle the medium-driven energy loss from the natural kinematic variance of the steeply-falling spectra across different . To decouple these effects, we utilize a phenomenologically motivated spectrum shift model to estimate the average transverse momentum loss imparted on high partons in A-A collisions, a proxy for the medium induced energy loss. We observe a striking correlation between and Glauber-derived estimates of initial state energy density , consistent across two orders of magnitude in collision energy for a variety of nuclear species. To access the path-length dependence of energy loss, we couple our model to geometric event shape estimates extracted from Glauber calculations to produce predictions for high- hadron elliptic flow that agree reasonably with data.
Introduction
Over the past two decades, strong evidence has accumulated for the creation of the Quark Gluon Plasma (QGP) in collisions of relativistic heavy-ions through extensive study at both the Relativistic Heavy-Ion Collider (RHIC) at Brookhaven National Laboratory, NY, USA and the Large Hadron Collider (LHC) at CERN in Geneva, Switzerland (see, for example, Ref. [1] and the references therein). Forefront among this evidence is partonic energy loss, termed “jet quenching.” This is revealed via the suppression of high transverse momentum (high-) hadrons in nucleus-nucleus (A–A) collisions relative to their production in collisions scaled by the mean number of binary nucleon-nucleon collisions so as to make the two comparable. Experimentally, this suppression is frequently identified via the nuclear modification factor
where is the nuclear overlap function determined from Glauber model calculations [2], proportional to via the inelastic nucleon-nucleon cross section at the relevant center of mass energy per colliding nucleon pair . and denote the charged particle yield per event in A–A collisions and the charged particle production cross section in collisions, respectively.
The observation of below unity indicates that partons (quarks and gluons) lose energy as they traverse the dense medium created in the collision [1]; data consistently show this signal across the range but especially at high-, where the effect is attributed to jet
quenching. Centrally, understanding the forces driving this partonic energy loss necessitates understanding the collision energy dependence. While the measured suppression values for different collision are comparable in principle, results contain a variety of convolved effects: medium-driven effects like collectivity and jet quenching, as well as kinematic restrictions of initial state parton composition and spectral shape. is strictly a ratio of yields, but is identified with energy loss because a uniform change in does not uniformly affect -differential yields for steeply falling spectra. LHC spectra, reflecting higher collision energy, are less steeply falling and dominated by gluon fragmentation when compared to the significantly softer RHIC spectra which primarily originate from the fragmentation of quarks [1, 3]. Hence, similar nuclear modification factors at RHIC and the LHC actually indicate more energy loss for partons traversing QGP created at the LHC. Additionally, observed differences between the for photon-tagged jets and inclusive jets, such as that reported by ATLAS in TeV Pb–Pb collisions [4], indicate that quarks and gluons interact differently with the QGP. These results highlight the importance of reliably deconvoluting medium effects from kinematic ones; the inherent hadron spectrum slope difference between RHIC and LHC alone can propagate to as much as a 10% difference in the .
The abundance of available jet quenching data from experimental collaborations such as high- measured below unity, alongside the challenge of calculations in QCD and related Effective Field Theories (EFTs), has motivated a wealth of phenomenological work attempting to bridge experimental observation with theoretical results. A non-exhaustive list of studies include EFT-motivated frameworks of jet quenching [5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20] and corresponding path-length dependent energy loss [12, 13, 17, 21, 22, 23, 24, 25], Bayesian extractions [20, 22, 26– 31], Boltzmann transport models [24, 26, 27, 28, 29, 32, 33, 34], quenched jet substructure [30, 35, 36], gauge boson-tagged phenomenology [33, 37], and modern computational techniques [20, 26, 35, 38, 39]; often a combination of these approaches is employed. For reviews on jet quenching and methods more broadly see [40, 41, 42, 43], for reviews on phenomenology in jet quenching see [42, 44].
On the experimental side, the effort to disentangle medium signatures from kinematic effects and better interpret results has led several collaborations to study the shift in —first proposed as by PHENIX [45]—needed to align the spectra in A–A with the binary scaled . Subsequent theoretical and phenomenological studies of jet quenching have also explored similar measures of energy loss related to horizontal shifts of spectra [9, 11, 14, 26, 31, 32, 34, 39, 46]; we will refer to this strategy broadly as methods. Unlike , these measures perform a direct fitting between hadron spectra, significantly reducing effects of distribution shape on the results. Measurements of and other similar observables show that the average energy loss at LHC energies exceeds that observed at RHIC, and by extension that gluon-dominated systems exhibit more energy loss than quark-dominated ones [4, 45, 47, 48, 49, 50, 51].
The goal of this analysis is to determine the degree of correlation between the medium-induced partonic energy loss and the initial energy density of the QGP. To quantify energy loss, we extract the average from experimentally observed charged particle spectra at high in and A–A collisions. This work differs from previous studies by assuming a fixed with the aim of estimating the average energy loss for a given centrality range and collision energy. While more rigorous first-principle calculations indicate that a fractional partonic energy loss is potentially preferred [52], the limited ranges studied in this work make our fixed approximation reasonable. In addition, given that high tracks, instead of full jets, are used as in this study to approximate the initial parton energy loss it is not clear that a fractional energy loss signature will be maintained through the natural smearing of the correlation between partonic energy and hadron that occurs during the fragmentation and hadronization process.
For QGP energy density, we use the reported particle yields at mid-rapidity and Glauber calculations to estimate the initial state energy density in the limit of Bjorken flow . By further utilizing Glauber event geometry, we predict the hadronic high- assuming a linear path-length dependent energy loss. The serves as an additional arena to compare degrees of freedom in our model and probes the relation between medium path length and energy loss.
This article is organized as follows: Section II details the data analysis including the determination of the transverse overlap area of the collisions (II A), the Bjorken energy density estimation (II B), the techniques used to extract (II C) and the high- predictions (IID). Section III presents our results. We conclude with a discussion in Section IV that summarizes our findings.
Analysis
The analysis proceeds along three axes. First, estimates of the initial transverse overlap area of the two colliding nuclei, , are made using Monte Carlo Glauber calculations [2]. The initial energy density is then approximated using and the charged particle multiplicities that have been determined experimentally. Second, the of each dataset is determined from the reported charged particle A–A and transverse momentum spectra. Once both the energy density and values have been calculated for each centrality bin of each collision system, the correlation between these quantities is determined. Finally, the initial overlap geometry is estimated from the Glauber simulation. Geometric quantities together with enable prediction of the high- to be made for each dataset, assuming a linear path-length dependent energy loss.
The principle data sets used in this analysis are Xe–Xe charged particle spectra from ALICE [53] and ATLAS [54] at TeV, ALICE [55] Pb–Pb results at TeV and 2.76 TeV, STAR [56] and PHENIX [57] charged particle spectra from Au–Au collisions at GeV, and STAR data from Cu–Cu [58] also at GeV. Corresponding spectra were obtained from the same A–A references for all datasets except STAR Au–Au, for which 200 GeV spectra were found in Ref. [59]. All data were obtained from publicly available HEPData archives.
Transverse Area Estimation
The transverse overlap area of the collision must be determined from simulation, and there is currently no definitive technique to perform this extraction. We therefore start this study by determining using several physically reasonable approaches, with the goal of determining if any exhibit similar behavior. First, we utilize the three grid-based calculations directly available from the Glauber code and described in Ref. [60]. We then present four new determinations based on averaged initial collision energy deposition.
The three built-in Glauber model [2, 60] estimates of transverse area—width-based, inclusive, and exclusive areas—are calculated on an event-by-event basis. The statistical width-based area, , is calculated from the (co-)variances of the nucleon distributions; the inclusive area, , is a grid-based set union including a disk around all participant nucleons; the exclusive area, , is a grid-based set intersection including only regions struck by both nuclei. The uncertainty on each area calculation is determined by the standard error of the mean of the output of the Glauber calculations. We label these three area calculations the event-by-event (EBE) methods.
The EBE methods are compared against estimates extracted from an averaged initial state energy density distribution in the transverse plane. In each event, the Glauber-generated nuclear thickness profiles are used to calculate local energy density via the geometric-mean scaling . This energy scaling enables the Glauber Monte Carlo to produce initial state energy distributions that agree qualitatively with more modern models like IP-Glasma [61], and is perhaps the Glauber energy scaling best supported by data [62]. The resulting single-event transverse energy profiles were then translated and rotated to align the center-of-mass position and the second-order participant plane angle . Finally the profiles were averaged to produce a single representative initial-state energy distribution in each centrality bin. Following established Glauber procedures [2], centrality was determined at simulation level by binning in the impact parameter as a proxy for final-state multiplicity. To reconcile this procedure with Glauber simulations used by the collaborations, which determine centrality by coupling Glauber Monte Carlo calculations to a simulation of a physical detector response [53, 54, 55, 57, 63] or otherwise bin in itself [56, 58, 64, 65], a systematic error is included on as follows. The Glauber Monte Carlo [2] is run with a variation of inelastic cross section with uncertainty as reported by the collaboration, and also with nuclear shape parameters disabled. The values reported by the collaboration are also tabulated. To obtain a systematic uncertainty on , we compare the nominal Glauber results—for which is always obtained from Ref. [2]—against each of the cases mentioned above. The largest difference against any variation in each centrality bin is symmetrized and taken as the systematic error on for that respective centrality bin. This error is used in the full procedure and extrapolated to a final uncertainty on the results.
Using this representative initial-state energy distribution, the average transverse area in a centrality bin was computed by extracting an “edge” azimuthal function . The area contained within this curve is the transverse area. Four methods were considered to determine : the average energy radius
the average pressure (energy gradient) radius
For simplicity, we choose , ; this can be thought of as translating the energy density maps to probability density distributions as in the Optical Glauber formalism. In addition, we consider the full-width-at-half-max contour which solves
and the surface of maximal gradient/pressure
Lastly, we also computed the (co-)variance “width” based area of the representative distribution, as was done on the event-by-event level. Confirming that the EBE and net-event averaged (co-)variance width areas agree is a validation of the event averaging procedure. These five methods are collectively labeled phenomenological areas, since while each edge-extraction method can be physically motivated (e.g. the average radius method represents the average total energy seen by a traversing high energy parton emitted at angle ), it is not immediately clear which method, if any, might best represent the effective relevant to this study.
The abundance of methods available for computing the transverse area motivates us to categorize them and search for defining properties of a given method’s predictions. For the purposes of this analysis, the primary behavior of interest is the centrality dependence of the transverse area, modulo any overall normalization in that dependence. For an illustrative example, we consider the centrality dependence of the various Glauber areas for Pb–Pb at TeV, shown in Figure 1(a), for each of the methods described above.
Figure 1. (a.) Computed transverse areas as a function of the event centrality for Pb–Pb collisions at TeV. Ratios of certain methods against (b.) and (c.) . The blue and orange markers denote the class of methods that scale like and respectively. The EBE exclusive calculations are shown as the black cross markers. See text for details.
A trend emerges when considering certain ratios. The eight transverse area calculations can be grouped into two classes and one outlier . The methods are considered as belonging to the same “class” if the ratio of their centrality dependences is roughly flat. Throughout Figure 1, consistent hue and marker shapes are used to denote class membership among methods. We denote the two classes as the inclusive class and the width class for methods that scale like the EBE inclusive area and EBE width-based calculations respectively. Figure 1(b) shows ratios against the EBE inclusive calculation, where the phenomenological Maximum Gradient and Full-Width at Half-Max contour methods are approximately flat. These two methods form the inclusive class together with . Figure 1(c) shows ratios against the EBE width method, where the remaining three phenomenological methods—the average energy radius, average gradient radius, and (co-)variance width—exhibit flat ratios, forming the width class . Note that the pre-averaged EBE width calculation and the post-averaged phenomenological width calculation exhibit ratios consistent with unity, a check on the event averaging procedure. In each case, exhibits markedly different scaling, shown by non-constant ratios. The EBE inclusive and width also do not have consistent scalings, meaning they cannot be merged to a single class.
While initially disconcerting, the disagreement between and other methods also coincides with observations in ALICE Pb, where computations using produce unusually high estimates of energy density [66]. Physically, the and methods reflect distinct pictures of Glauber modeling: considers entire struck nucleons as participants in traditional Glauber fashion, while regards only overlapping sub-nucleonic regions of participant nucleons as contributing.
Lastly, it is important to note that while Figure 1 shows the calculations for Pb–Pb collisions at TeV, the division of methods into these three distinct classes is persistent across all the energies and species studied, as detailed in Appendix § 2.1.
Bjorken Energy Density Estimation
The average initial energy density of the medium as a function of centrality is approximated in the limit of Bjorken hydrodynamics [67] as outlined in Ref. [1]; this results in the approximation
where is the Jacobian conversion from pseudorapidity to rapidity, is the charged hadron density at midrapidity, is the QGP formation time, and is defined by ; and are the initial entropy density and temperature respectively. As in Ref. [1], is assumed to be constant and independent of collision energy, species, and centrality with a value of . For highly Lorentz contracted nuclei the parameter is commonly chosen to be 0.6 fm/c, while at collision energies GeV, the time for the nucleons to fully cross is longer [1]. As all collision data used in this study have center-of-mass energy per nucleon GeV, we also fix fm/c.
The remaining input parameters were obtained from various sources. The Jacobian ranges from 1.25 for GeV to 1.09 at TeV [68, 69]. An uncertainty of 3% on is assumed for all beam energies, propagated forward to the uncertainty on . The data used in this analysis are those reported from Pb–Pb collisions at TeV and TeV from ALICE [55, 70], Xe–Xe collisions at TeV from ALICE [53], Au–Au collisions at GeV from STAR [71], and lastly Cu–Cu collisions at GeV from STAR [58]. The uncertainty used on is the quadrature sum of the reported statistical and systematic uncertainties in each measurement.
The transverse area is determined for each species, beam energy, and corresponding centrality class as described in the previous section. Uncertainties originating from , and are propagated differentially to the final energy density value. To fairly consider each of the three observed centrality scalings, we separately compute energy densities from each of the EBE methods with , , exclusive representing width-based , inclusive , and exclusive areas respectively. Details of the resultant values and uncertainties of and computed from the above inputs are given in Appendix § 2.1, along with the relevant data used to compute these quantities. The exclusive area is not included in the tables, as its especially sharp centrality scaling seen in Figure 1 leads to unphysical trends where the energy density exclusive is smaller in central collisions than peripheral ones for some systems; unexpected behaviors of the exclusive method have also been observed by ALICE in LHC pPb collisions [66], where predicts higher energy density in pPb collisions than in central Pb–Pb. Still, to enable a fair comparison among all methods, we do include in the final results of Sec. III.
Extracting from Transverse Momentum Spectrum Data
In previous analyses [4, 45, 46, 47, 48, 49, 50, 51] was determined for a fixed hadron . Using the transverse momentum spectra found in [53, 54, 55, 56, 57, 58, 59], this study explores if a common can be identified for the whole range of high- data reported. We identify such a via a spectrum shifting procedure described in this section.
The horizontal shifting of the spectra to describe mean parton energy loss is applicable only at high momentum, where parton fragmentation is the dominant source of particle production. A threshold value, , must therefore be chosen to ensure that other collective effects—such as radial flow—present at lower do not affect the extraction. Data below this threshold are not included in the determination of . To determine this threshold we consider the reported charged hadron . In all measurements studied, the in central collisions exhibit a local minimum near GeV/c, whereafter the data rise monotonically. While jet quenching effects are likely to still be present at momenta below this turnover, the region above should be reasonably free of collective effects; we therefore take these local minima as . The systematic uncertainty on due to this choice of threshold is determined by varying the threshold by an additional bin below and above each dataset’s local minima.
Since many of the peripheral datasets suffer from large uncertainties, preventing a clear determination of a local minimum, we decided to use the extracted in the most central events for all centralities reported for a given dataset. This likely results in a slight overestimate of the threshold in peripheral events since glancing collisions should produce a narrower collective region. As we are aiming to exclude collective phenomena, this overestimation should not affect our conclusions.
With the threshold chosen, we compute the by scaling and shifting the reference spectra to match the A–A data using the following procedure:
Reference spectra are first fit to a Tsallis distribution, given by
where is the transverse kinetic energy, is the high- power law scaling of the spectrum, is a temperature-like parameter controlling the width of the collective region, and is a normalization factor. We use the standard
pion mass in the kinetic energy for all particles. Motivation for this choice of fit function and details about our fitting procedure will be discussed shortly.
The resultant fit on the spectra is then scaled by a factor obtained from MC Glauber estimates [2], as they would be for an calculation.
The A–A data are shifted horizontally rightward toward high- (or equivalently, the scaled fit can be translated horizontally leftward) until the shifted baseline spectrum and A–A data agree as well as possible, according to the same fit metric used for the spectrum fitting. This optimal horizontal shift is .
This procedure is shown diagrammatically in Figure 2.
Figure 2. Cartoon illustrating the procedure used to determine .
The fit methods for the spectra and were chosen carefully to best extract the relevant parameters for our analysis: namely the power law scaling of the various spectra. The Tsallis distribution, also called the Hagedorn function, was chosen as the fit form following other work demonstrating that this function is an effective choice for spectra over the full range at both RHIC and LHC energies [72, 73, 74, 75, 76, 77]. The function smoothly connects a thermodynamic, low- region expected to scale as with the observed power law scaling at high . Since data below are excluded from the fit, in principle any function with power law asymptotics should yield comparable , but a known effective form was chosen as a safeguard. Fitting of the spectra is performed by binning a candidate Tsallis fit and computing a fit metric against the binned data; the set of parameters which minimize the metric are selected as the final fit. We found that the commonly used fit metric was reasonable but tended to be heavily biased by the first bin above the threshold, which can be many orders of magnitude larger than bins at higher , and therefore generally produced poor extractions of the power law, even in toy fits with infinite statistics. To avoid this bias, we selected a fit metric which compares logarithms of the bin contents. We found such a metric extracted the power law with significantly higher accuracy in all tests we performed; for details see Appendix § 2.2. The metric we chose takes the form
where and are the data and fit candidate; we label this metric the Mean Square Entropy (MSE). The form has some similarity to G-tests or likelihood tests in statistics. The use of this metric is analogous to performing a linear regression in the log-log plane, which allows it to reliably extract precise estimates of the power law scaling. For consistency, we used this fit metric for both the spectra fitting and the shift fitting.
In addition to the systematic uncertainty generated from the choice of , uncertainties in both and A–A spectra are propagated to the uncertainty on using a Monte Carlo method. Individual bins of the spectra are randomly varied using normal distributions with widths that matched the uncertainties, and is recalculated using these smeared spectra. The mean of 10000 variations is reported as the final shift, and the standard deviation becomes the propagated uncertainty. Uncertainty on described in § 2 A is propagated to the by performing the scaling component of the shifting procedure with replaced by and . The maximum difference between the default and these altered values is taken as the uncertainty from this source. All uncertainty sources are added in quadrature to obtain the final uncertainty.
High Estimation
As described above, the first part of this study approximates the initial energy density from the mid-rapidity hadron multiplicity and Glauber estimates of the initial geometry and overlap area . However, the correlation between these quantities is complex and merits further exploration; the link between energy density and charged particle multiplicity is expected to be sensitive to the medium shear viscosity [78], and the relationship between energy loss and event geometry informs the path-length dependence of medium-driven energy loss. We therefore turn to -differential azimuthal anisotropy.
The observed dependent hadron multiplicity as a function of azimuthal angle admits a Fourier series decomposition,
where the angle is oriented relative to the -order collision event plane, measured on an event-by-event basis. is the particle’s azimuthal angle, and is that of the plane. Roughly speaking, the magnitude of the extracted high- has two distinct contributions at each : one from higher energy initial partons that lost more energy traversing a longer length of QGP, and another from lower energy initial partons that lost less energy traversing a shorter length. After determining the average for a given centrality as described above, we can use this principle to construct a model for estimating the charged particle high as follows:
Consider a Glauber derived estimate of the path length of an event-averaged collision region as a function of azimuth , discussed above as phenomenological area estimates. Fourier decompose this function, and take the second harmonic coefficient . The ratio over the average radius is the path length fraction of the maximal/minimal path against the average. Example values of for different collision systems and path length functions are given in Appendix § 2.1.
Take two copies of the shifted spectrum fit. Shift one copy up toward higher according to the proportion , and another copy down by the same proportion. These spectra enclose the original shifted spectrum. The upshifted and downshifted spectra are labeled and respectively.
Our model estimate for the high- differential is then the difference weighted to the original spectrum:
This estimation is similar to that used in Ref. [12] extrapolated under the assumption of uniform energy loss , or can alternatively be thought of as a relation between the Fourier series coefficients of path length and flow coefficients (9).
These steps are illustrated diagrammatically in Figure 3. The offers a new arena to compare the classes and observed in the transverse area. We will use the Full-Width-at-Half-Max contour from the width class and the average energy radius curve from the inclusive class ; the data for Glauber for these methods are shown in Table V of Appendix § 2.1.
Figure 3. Cartoon illustrating the steps of the estimation process.
| Collision System | Area Method and Class | Edge Coef. | 0–5% | 5–10% | 10–20% | 20–30% | 30–40% | 40–50% | 50–60% | 60–70% | 70–80% |
| Pb–Pb 2.76 TeV | Avg. Radius () | 3.83 | 3.55 | 3.23 | 2.85 | 2.53 | 2.25 | 2.00 | 1.77 | 1.53 | |
| Pb–Pb 2.76 TeV | Avg. Radius () | 0.14 | 0.26 | 0.38 | 0.48 | 0.53 | 0.55 | 0.56 | 0.54 | 0.50 | |
| Pb–Pb 2.76 TeV | Avg. Radius () | 0.03 | 0.07 | 0.12 | 0.17 | 0.21 | 0.24 | 0.28 | 0.31 | 0.32 | |
| Pb–Pb 2.76 TeV | FWHM Contour () | 5.17 | 4.70 | 4.13 | 3.49 | 2.96 | 2.51 | 2.15 | 1.87 | 1.49 | |
| Pb–Pb 2.76 TeV | FWHM Contour () | 0.32 | 0.47 | 0.62 | 0.73 | 0.77 | 0.75 | 0.73 | 0.72 | 0.68 | |
| Pb–Pb 2.76 TeV | FWHM Contour () | 0.06 | 0.10 | 0.15 | 0.21 | 0.26 | 0.30 | 0.34 | 0.38 | 0.45 | |
| Pb–Pb 5.02 TeV | Avg. Radius () | 3.83 | 3.55 | 3.22 | 2.84 | 2.52 | 2.24 | 1.99 | 1.75 | 1.51 | |
| Pb–Pb 5.02 TeV | Avg. Radius () | 0.14 | 0.27 | 0.38 | 0.48 | 0.53 | 0.56 | 0.56 | 0.54 | 0.50 | |
| Pb–Pb 5.02 TeV | Avg. Radius () | 0.03 | 0.07 | 0.12 | 0.17 | 0.21 | 0.25 | 0.28 | 0.31 | 0.33 | |
| Pb–Pb 5.02 TeV | FWHM Contour () | 5.17 | 4.69 | 4.12 | 3.48 | 2.94 | 2.50 | 2.14 | 1.84 | 1.45 | |
| Pb–Pb 5.02 TeV | FWHM Contour () | 0.31 | 0.47 | 0.62 | 0.73 | 0.77 | 0.75 | 0.72 | 0.71 | 0.67 | |
| Pb–Pb 5.02 TeV | FWHM Contour () | 0.06 | 0.10 | 0.15 | 0.21 | 0.26 | 0.30 | 0.34 | 0.38 | 0.46 | |
| Xe–Xe 5.44 TeV | Avg. Radius () | 3.22 | 3.02 | 2.77 | 2.48 | 2.24 | 2.02 | 1.82 | 1.62 | 1.39 | |
| Xe–Xe 5.44 TeV | Avg. Radius () | 0.18 | 0.24 | 0.31 | 0.39 | 0.44 | 0.47 | 0.47 | 0.46 | 0.44 | |
| Xe–Xe 5.44 TeV | Avg. Radius () | 0.05 | 0.08 | 0.11 | 0.15 | 0.19 | 0.23 | 0.26 | 0.28 | 0.31 | |
| Xe–Xe 5.44 TeV | FWHM Contour () | 4.13 | 3.77 | 3.32 | 2.85 | 2.47 | 2.17 | 1.94 | 1.62 | 1.20 | |
| Xe–Xe 5.44 TeV | FWHM Contour () | 0.34 | 0.40 | 0.50 | 0.57 | 0.59 | 0.61 | 0.65 | 0.67 | 0.59 | |
| Xe–Xe 5.44 TeV | FWHM Contour () | 0.08 | 0.10 | 0.15 | 0.20 | 0.24 | 0.28 | 0.33 | 0.41 | 0.49 | |
| Au–Au 0.20 TeV | Avg. Radius () | 3.70 | 3.45 | 3.15 | 2.80 | 2.51 | 2.25 | 2.02 | 1.80 | 1.58 | |
| Au–Au 0.20 TeV | Avg. Radius () | 0.16 | 0.26 | 0.36 | 0.45 | 0.50 | 0.53 | 0.54 | 0.53 | 0.50 | |
| Au–Au 0.20 TeV | Avg. Radius () | 0.04 | 0.07 | 0.11 | 0.16 | 0.20 | 0.23 | 0.26 | 0.29 | 0.31 | |
| Au–Au 0.20 TeV | FWHM Contour () | 4.95 | 4.51 | 3.99 | 3.40 | 2.91 | 2.51 | 2.18 | 1.92 | 1.57 | |
| Au–Au 0.20 TeV | FWHM Contour () | 0.33 | 0.45 | 0.58 | 0.69 | 0.73 | 0.72 | 0.70 | 0.70 | 0.67 | |
| Au–Au 0.20 TeV | FWHM Contour () | 0.06 | 0.10 | 0.14 | 0.20 | 0.25 | 0.28 | 0.32 | 0.36 | 0.42 | |
| Cu–Cu 0.20 TeV | Avg. Radius () | 2.63 | 2.50 | 2.33 | 2.13 | 1.96 | 1.80 | 1.64 | 1.47 | 1.29 | |
| Cu–Cu 0.20 TeV | Avg. Radius () | 0.18 | 0.22 | 0.27 | 0.33 | 0.38 | 0.40 | 0.40 | 0.39 | 0.39 | |
| Cu–Cu 0.20 TeV | Avg. Radius () | 0.07 | 0.09 | 0.11 | 0.15 | 0.19 | 0.22 | 0.24 | 0.26 | 0.30 | |
| Cu–Cu 0.20 TeV | FWHM Contour () | 3.24 | 3.00 | 2.71 | 2.39 | 2.14 | 1.94 | 1.71 | 1.41 | 1.11 | |
| Cu–Cu 0.20 TeV | FWHM Contour () | 0.35 | 0.39 | 0.42 | 0.47 | 0.52 | 0.58 | 0.63 | 0.64 | 0.54 | |
| Cu–Cu 0.20 TeV | FWHM Contour () | 0.10 | 0.13 | 0.15 | 0.19 | 0.24 | 0.30 | 0.36 | 0.45 | 0.49 |
Table V. Tabulated values for averaged event harmonics , from the and areas generated from Glauber simulations for the collision systems studied. Note that the Average Radius method represents the class, while the Energy Half-Max Contour represents in the class.
Note that this model contains an assumption of linear path-length dependent energy loss to assert that the path length proportion can be translated directly to an energy loss for extra path-length . Modeling a different dependence would require a more complicated relationship that reflects the nonlinear dependence on azimuthal angle: a different power would cause a mixing of Fourier components and other complications. We use the simple linear dependence for now, and relegate other powers to future study.
Results
Figures 4 and 5 present the Bjorken energy densities, and , respectively, as functions of centrality and for a variety of collision energies and species. Sensibly, increases monotonically with increasing or for both and , but this does not generally hold for , as will be discussed later. While the energy densities are similar for the most peripheral data, the width based rises more steeply with centrality, resulting in an approximately 50% larger energy density estimate for the most central Pb–Pb data relative to .
Following the procedure in Sec. IIC, illustrative results from ALICE Pb–Pb at TeV as well as PHENIX and STAR Au–Au at GeV datasets are shown in Figure 6, Figure 7 and Figure 8 respectively. In each figure, the top left panel (a.) shows the charged particle spectra for three sample A–A centrality bins alongside the corresponding appropriately -scaled inelastic data. The curves are the -scaled Tsallis fits to the reference spectra. The fit parameters and MSE metric for all resultant data fits are provided in Table I. The suppression of the A–A spectra with respect to the -scaled Tsallis fit is evident in all cases, in agreement with published values [53, 54, 55, 56, 57, 58]. The remaining three panels (b.–d.) show the -shifted A–A data for the three sample centrality bins alongside the Tsallis fit curve. The A–A spectra have been shifted by the which best aligns the two and minimizes the MSE metric above the determined momentum threshold denoted by the vertical line. Note that the line has, along with the A–A data points, been shifted rightward by in panels (b.–d.). The Tsallis fit curve is only shown in the relevant region above this threshold. The strong visual agreement between the shifted A–A spectra and -scaled reference Tsallis fit across the wide range of species, energy, and centrality further validates the MSE as a fit metric, and affirms the approach explored in this study in which a single energy loss is determined for each centrality bin. Note that the final MSE metric for each fit, including those shown in Figs. 6, 7 and 8 is provided for each collision system and centrality bin in Appendix § 2.1. As a further model check, the extracted -shifted Tsallis fits were substituted for the A–A data, and ratios against the corresponding baseline spectra were taken for comparison against published data. Rough agreement of
Figure 6. spectra of charged particles measured by the ALICE collaboration in Pb–Pb collisions at TeV for different centralities. The first panel shows the Tsallis fit appropriately -scaled to each Pb–Pb dataset. The remaining three panels show the -scaled Tsallis fit compared to the individual shifted Pb–Pb spectrum in the region beyond the shift lower bound. Uncertainty bars are present, but small relative to marker sizes.
Figure 7. spectra of charged particles measured by the PHENIX collaboration in Au–Au collisions at GeV for different centralities. The first panel shows the Tsallis fit appropriately -scaled to each Au–Au dataset. The remaining three panels show the -scaled Tsallis fit compared to the individual shifted Au–Au spectrum in the region beyond the shift lower bound.
Figure 8. spectra of charged particles measured by the STAR collaboration in Au–Au collisions at GeV for different centralities. The first panel shows the Tsallis fit appropriately -scaled to each Au–Au dataset. The remaining three panels show the -scaled Tsallis fit compared to the individual shifted Au–Au spectrum in the region beyond the shift lower bound.
| Experiment () | (c/GeV) | (GeV) | Reduced MSE Metric | |
| ALICE (5.44 TeV) | 18.540 | 5.566 | 0.214 | 0.002423 |
| ATLAS (5.44 TeV) | 18.540 | 5.555 | 0.209 | 0.001209 |
| ALICE (5.02 TeV) | 9.825 | 5.684 | 0.247 | 0.003396 |
| ALICE (2.76 TeV) | 10.720 | 5.970 | 0.227 | 0.004184 |
| STAR (0.2 TeV) | 16.578 | 8.165 | 0.154 | 0.01000 |
| PHENIX (0.2 TeV) | 16.578 | 9.108 | 0.169 | 0.00155 |
| STAR (0.2 TeV, ) | 95.266 | 8.169 | 0.112 | 0.000677 |
| Centrality Bin (%) | STAR reference GeV | STAR reference GeV | PHENIX reference GeV | PHENIX reference GeV |
| STAR Au–Au (GeV/) | PHENIX Au–Au (GeV/) | STAR Au–Au (GeV/) | PHENIX Au–Au (GeV/) | |
| 0–5% | 1.77 ± 0.23 | 1.55 ± 0.22 | 1.38 ± 0.13 | 1.27 ± 0.21 |
| 10–20% | 1.43 ± 0.22 | 1.40 ± 0.22 | 1.07 ± 0.16 | 1.12 ± 0.21 |
| 20–30% | 1.18 ± 0.20 | 1.04 ± 0.21 | 0.85 ± 0.15 | 0.81 ± 0.20 |
| 30–40% | 0.92 ± 0.20 | 0.62 ± 0.09 | 0.62 ± 0.17 | 0.63 ± 0.12 |
| 40–50% | – | 0.72 ± 0.13 | – | 0.56 ± 0.12 |
Table 1. TABLE I. cross section spectra fit parameters (, , ) and the reduced MSE/DoF metric for the Tsallis fit to each dataset used in the analysis.
Figure 4. Energy density using for a variety of collision species and beam energies, as a function of (upper) and centrality (lower).
Figure 5. Energy density using for a variety of collision species and beam energies, as a function of (upper) and centrality (lower).
this model ratio with was observed within for all collision systems studied, though it should be noted that the model deviates more strongly from in central collisions.
It should be noted that the Tsallis power law parameters fit to STAR and PHENIX data at GeV differ significantly, while all fit parameters for the ALICE and ATLAS data at TeV are consistent.
The disagreement in the RHIC fit results is likely due differences in how the charged particle cross section was determined by the two collaborations. PHENIX computed charged particle spectra by extrapolating the spectrum from Ref. [79]. STAR, on the other hand, determined the charged particle spectra by summing measured -differential , , and yields [59]. The STAR determination results in a more significant flattening of the higher data compared to the PHENIX extrapolation, reflected in the smaller power law .
Following the discrepancy in reference spectra, the values calculated for Au–Au collisions at GeV differ between the STAR and PHENIX databases. This is again in contrast to the Xe–Xe data reported by ATLAS and ALICE, where we see agreement within uncertainties for resultant values across all centrality bins. To examine whether this discrepancy in is purely due to the spectra differences observed in the fits, we compared values obtained from STAR and PHENIX Au–Au spectra with a common fixed refer-
ence spectrum. Table II shows the calculated when either the STAR or PHENIX charged particle spectrum is used as the common reference. In both cases, the computed values for each collaboration’s Au–Au data agree within uncertainties. This affirms that the observed differences in is dominantly an artifact of the differences in spectral shape between collaborations. For consistency, however, our analysis proceeds using the spectrum reported by each collaboration to calculate the for their respective A–A data in our final results.
With this understanding of and in hand, we can now move on to the central objective of this work: characterization of the relationship between these two quantities. The resulting as functions of Bjorken energy density for each area class are shown in Figure 9. For the inclusive and width-based areas, a clear direct correlation of the high- charged particle with the estimated initial energy density is observed, from peripheral Cu–Cu collisions at GeV to central Pb–Pb events at TeV. To evaluate whether these correlations are linear to leading order, a fit to a variable power is performed, from which the resulting powers are and . These correspond to of 58.77/46 and 57.56/46 for the inclusive and width-based methods respectively, strongly supporting the conclusion of a linear relationship at first order.
Figure 9. Energy density versus for a variety of collision species and beam energies. Estimates of energy density from each area scaling class , , and are shown separately. Note the STAR Cu–Cu data are reported spectra for calculating while other datasets use inclusive charged particle spectra.
Performing linear regressions, a slope of 0.054 0.002 fm/c and 0.099 0.003 fm/c for the width and inclusive area estimates respectively. The individual linear fits for each experiment-collision pair, as well as the corresponding reduced values are provided in Table III.
| Experiment | Beam Species, | Intercept (GeV/c) | slope (fm/c) | Intercept (GeV/c) | slope (fm/c) | Reduced | Reduced |
| ALICE | Xe–Xe, 5.44 TeV | ||||||
| ATLAS | Xe–Xe, 5.44 TeV | ||||||
| ALICE | Pb–Pb, 5.02 TeV | ||||||
| ALICE | Pb–Pb, 2.76 TeV | ||||||
| STAR | Au–Au, 200 GeV | ||||||
| PHENIX | Au–Au, 200 GeV | ||||||
| STAR | Cu–Cu, 200 GeV | ||||||
| Global |
Table III. Linear fit results for as a function of and . Fitting is performed using orthogonal distance regression to incorporate uncertainties in dependent and independent axes. Reduced computed using only independent axes uncertainties are included.
Such a strong correlation indicates that the initial energy density of the overlap region primarily drives partonic energy loss, and that other factors such as the shape of the overlap region are of secondary importance at best. In addition, this correlation is preserved through the fragmentation and hadronization process such that it can be detected in the single hadron spectra. It also appears that the manner of energy density generation is of little consequence; collisions of low species at high energy or high species at low energy result in the same , so long as QGP is formed and events are selected with equivalent initial energy densities. To explore further for evidence of subleading dependences, the residuals to the linear fits shown in Figure 9 are studied as a function of effective path length and path difference as well as in Appendix C.
Although presents itself as an outlier in Figure 1, we include the relationship between and in Figure 9 for completeness. We see that for collisions at RHIC energies, the calculated energy density using is approximately constant for all data points, indicating that the ratio of and does not vary over centrality or species. In the specific case of STAR Cu–Cu, slightly decreases with increasing impact parameter, suggesting nonphysically that peripheral collisions produce larger energy density than central collisions at
the same . For LHC data however, does rise with centrality and roughly increases with energy density, but these behaviors are different for each collision system. While in principle could provide a reasonable estimate of transverse area, our observations of the inconsistent scaling between RHIC and LHC data, the disagreement of centrality scaling with all other area methods shown in Figure 1 and Figure 12, and the previously reported strange behaviors in ALICE pPb [66] culminate to strongly disfavor this area scaling for the computation of .
Figure 12. Scalings of the various Glauber methods proposed to compute the transverse area across a range of collision species and energy. The panels are Pb–Pb at TeV (a.–c.), Xe–Xe at TeV (d.–f.), Au–Au at GeV (g.–i.), and Cu–Cu at GeV (j.–l.). Each column contains the explicit transverse areas produced by each method (a., d., g., j.), ratios against the EBE inclusive (b., e., h., k.) and ratios against the EBE width area (c., f., i., l.). EBE calculations produced by the Glauber code [2] are shown in solid markers while phenomenological area methods are shown in hollow markers.
Continuing to results, Figure 10 shows the prediction for TeV Pb–Pb high- using the procedure described in § 2 D. Model predictions for the and classes are compared to corresponding ALICE data [80] for three different centrality bins. Predictions for both area estimates are similar in shape but differ slightly in normalization, with the class favored by central data and performing better for peripheral collisions. Considering the simplicity of the modeling, a reasonable agreement is obtained across the available range, within for all data for -based . Stronger precision on high- data could better distinguish between the two models or clarify regimes of model failure, such as in peripheral collisions where there is more tension between the models and data.
Figure 10. High- estimations derived from as well as the avg. energy radius/avg. pressure radius area definitions, corresponding to the class (solid blue curve) and the FWHM contour area definition, corresponding to the area class (dashed orange curve), compared to ALICE experimental data from Pb–Pb collisions at TeV.
Encouraged by the general agreement with the magnitude of across the range shown in ALICE Pb–Pb data, Figure 11 presents our model’s predictions for the other collision systems studied. As it performs slightly better for the ALICE Pb–Pb, the predictions utilize the initial overlap regions’ harmonics , derived from the width-based area estimates, though the corresponding for are shown in Appendix D. The for the Au–Au data at GeV is the lowest at a fixed and centrality; although the ratios in Au–Au are similar to those for the LHC Pb–Pb and Xe–Xe, the extracted are significantly reduced due to the smaller , resulting in smaller in our modeling.
Figure 11. High predictions using the and based model for a variety of collision systems.
The trends of the Xe–Xe predictions are especially interesting, as the TeV Xe–Xe contains the largest among all species for the most central data, which falls to the smallest LHC in peripheral collisions. A similar flipping of the Xe–Xe relative to TeV Pb–Pb is also predicted by hydrodynamic calculations [81] and observed in data [82, 83], albeit at significantly lower . The enhancement of low- in central Xe–Xe is attributed primarily to nuclear shape, as the quadrupole deformation is expected to be much larger in than [84]. The nuclear size also plays a role; systematic analyses of observables have shown that statistical fluctuations in yields produce a universal enhancement of all in systems with fewer sources. The enhancement of in Xe–Xe relative to Pb–Pb across the centrality range is considered a signal of this effect [82].
The depletion of for peripheral TeV Xe–Xe collisions relative to TeV Pb–Pb is therefore unexpected on the basis of nuclear structure alone, as models indicate that the nuclear shape of has negligible effects in the peripheral region [81]. This signal must therefore be the result of differing medium properties. In the case of the low- measurements by ALICE [82] and ATLAS [83], the depletion is attributed to nonzero shear viscosity during hydrodynamic evolution; hydrodynamic calculations [81] reproduce the depletion only for nonzero , though calculations are similar for a range of [82]. In the absence of significant hydrodynamic effects for high- hadrons, the depletion is instead attributed to path-length dependent jet quenching [1]. While the ranges reported in Refs. [82, 83] are restricted due to the limited statistics from the short Xe–Xe run, future precision measurements at the LHC with a range of nuclear species could be very impactful to our understanding of the path length dependence of partonic energy loss.
Conclusions
In summary, we have demonstrated a strong linear correlation between the estimated initial energy density created in a heavy-ion collision and the average energy loss of high- charged particles generated by those collisions. This striking correlation is observed across a wide variety of collision systems, from Au–Au and Cu–Cu collisions at GeV to Pb–Pb and Xe–Xe events at TeV. The observed correlation appears independent of ion species and collision energy, and persists among disparate methods of Glauber modeling. This result suggests that the initial energy density is the driving factor in predicting the average energy loss of hard scattered partons to the QGP. Detailed knowledge of other factors such as the initial event eccentricity and parton composition are not critical for predicting the azimuthally averaged charged particle .
A variety of methods were studied to extract the transverse overlap area using Monte-Carlo Glauber calculations, needed to compute . These methods included three grid-based EBE calculations described in Ref. [60], along with five novel phenomenological methods. Two classes were identified based on the methods’ approximate scalings with centrality, with the EBE calculation being the sole outlier. The strange scaling for we observe is also seen in Glauber estimates for ALICE pPb [66]. Of the two identified classes, one scaled with centrality in a similar fashion to the EBE “inclusive” area, while the other class scaled with the “width-based” EBE calculation. This classification of scaling persisted across all species and energies studied. Generally, the class displayed a stronger dependence on centrality than the class, especially for more peripheral events. The estimated using from either or classes and the reported experimental show the expected trends with centrality, , and collision energy, while the same is not true for the outlier . Because of this, does not produce the strong – correlation observed for the other seven methods. The persistence of the correlation between both the and area classes, with no change in modeling, supports that the correlation is robust and not due to Glauber model self-correlation between inputs to and scaling of fit spectra. That the correlation then fails for demonstrates that the correlation is not a trivial consequence of the model construction. In addition, the uncertainty estimation procedure, applied only to the computation of , does not significantly increase uncertainty or blur the correlation.
Good descriptions of the A–A high- spectra can be obtained for all systems, collision energy, and centralities studied via shifting of the appropriately -scaled Tsallis function fits to the spectra. For simplicity a constant shift is assumed for each spectrum. While theoretical arguments suggest a fractional energy loss with parton type and dependence, the limited high- charged particle range currently available combined with the smearing in between parton initiator and final-state hadron appears to make this constant approximation reasonable. The effectiveness of this approximation over such a wide range of collision systems and energy may offer new insights into the deconvolution of medium-driven spectrum modifications from kinematic ones.
We extended our model to explore predictions of high- hadronic anisotropy by coupling Glauber estimates of event geometry to observed spectra in a novel approach. The model provides estimates of that are reasonably consistent with ALICE data [80] in the highest data of each centrality bin, but overestimates for lower data. As the breakdown region of the model is still well above the collective region, the deviation of from our model may indicate that our assumption of linear path length-dependent energy loss is ineffective for this mid- region. We also observe a flipping of the magnitude of the Xe–Xe at TeV relative to that in Pb–Pb at TeV when going from central to peripheral data. This observation indicates that our model is capable of accessing path-length dependent jet quenching and nuclear deformation through only initial state models and final-state spectra.
The interrelations between , , and related quantities merit further research. While the exclusive area exhibits strange behaviors when used for energy density computation, it may be valuable in other contexts. The division of the remaining area methods into two classes further complicates the question of whether a single calculation exists, or what might separate these two classes. While our -independent model serves as a reasonable approximation for interpreting high- hadron spectra, it would be interesting to explore whether this approximation holds for jet spectra which are understood to serve as better, but not perfect, approximations of the initial parton’s energy. Similarly, comparing our model’s predictions with -differential jet may provide a more precise exploration of path-length dependent energy loss, especially if non-linear energy loss dependence is incorporated. Increased precision of jet measurements from LHC Run3, first results from sPHENIX and improved precision from upgraded STAR data should allow such detailed comparisons in the jet regime to be made soon.
As the field begins to enter the precision era for both experimental measurements and theoretical modeling, simple models with few degrees of freedom such as this remain helpful for calibrating assumptions, (re)evaluating observables and highlighting the dominant necessary features that must be reproduced by more elaborate simulation frameworks and rigorous first-principle calculations.
ACKNOWLEDGMENTS
We thank Constantin Loizides for helpful discussions about the interpretation of the Glauber Monte Carlo calculations, and Raymond Ehlers for insights about spectra fitting and fit functions. We gratefully acknowledge Christine Nattrass and Brant Johnson for their help in obtaining PHENIX reference yields for analysis in this study. RH and HC are supported by DoE grant DE-SC004168.
Additional Information Concerning Energy Density, and Event Averaged Harmonic Calculations
This appendix provides further details on the area calculations and other input data used to compute the energy density and the averaged event harmonics , used in the elliptic flow predictions.
Figure 12 provides evidence that the categorization of area definitions into the and classes holds across collision systems studied here. The figure shows the Glauber area calculations for Pb–Pb collisions at TeV, Xe–Xe at TeV, and Au–Au and Cu–Cu at GeV as a function of centrality. In addition, the ratios against the EBE inclusive () and EBE width-based () are shown for each collision system. Solid markers indicate the area EBE calculations produced by the Glauber code [2] and hollow markers represent those from the phenomenological edge-area methods. For each system the ratio plots reveal the same two classes plus one outlier, the EBE exclusive (). These observations coincide with the same classifications in Pb–Pb at TeV, presented in Figure 1. Members of the same class are identified via their similar trends as a function of centrality and hence flat ratios. Throughout the plotting in Figure 1 and Figure 12, members of the same class share similar colors and marker shapes to better delineate them visually.
Table IV details the experimental data used to estimate the mid-rapidity initial energy densities : the Jacobian to translate from to , and the transverse areas and . Also included are the MSE/DoF fit metric results for the shifted spectra, as measured against the -scaled Tsallis reference. Note that for ATLAS Xe–Xe [54] and PHENIX Au–Au [57] the mid-rapidity from ALICE [53] and STAR [58] respectively were used for computation, as published ATLAS/PHENIX data in the same centrality binning were not available.
| Centrality Bin (%) | Transverse Area Width | Transverse Area Inclusive | Energy Density Width | Energy Density Inclusive | ALICE / STAR MSE/DoF, Eq. (8) | ATLAS / PHENIX MSE/DoF, Eq. (8) | |
| Pb–Pb at TeV. Jacobian from [69], from [70]. | ALICE [55] | ||||||
| 0.0–5.0 | 0.033 | ||||||
| 5.0–10.0 | 0.028 | ||||||
| 10.0–20.0 | 0.021 | ||||||
| 20.0–30.0 | 0.013 | ||||||
| 30.0–40.0 | 0.013 | ||||||
| 40.0–50.0 | 0.014 | ||||||
| 50.0–60.0 | 0.023 | ||||||
| 60.0–70.0 | 0.009 | ||||||
| 70.0–80.0 | 0.025 | ||||||
| Pb–Pb at TeV. Jacobian from [69], from [55]. | ALICE [55] | ||||||
| 0.0–5.0 | 0.058 | ||||||
| 5.0–10.0 | 0.046 | ||||||
| 10.0–20.0 | 0.032 | ||||||
| 20.0–30.0 | 0.021 | ||||||
| 30.0–40.0 | 0.014 | ||||||
| 40.0–50.0 | 0.006 | ||||||
| 50.0–60.0 | 0.004 | ||||||
| 60.0–70.0 | 0.014 | ||||||
| 70.0–80.0 | 0.010 | ||||||
| Xe–Xe at TeV. Jacobian from [69], from [53]. | ALICE [53] | ATLAS [54] | |||||
| 0.0–5.0 | 0.032 | 0.044 | |||||
| 5.0–10.0 | 0.054 | 0.059 | |||||
| 10.0–20.0 | 0.037 | 0.027 | |||||
| 20.0–30.0 | 0.030 | 0.013 | |||||
| 30.0–40.0 | 0.018 | 0.007 | |||||
| 40.0–50.0 | 0.040 | 0.007 | |||||
| 50.0–60.0 | 0.019 | 0.005 | |||||
| 60.0–70.0 | 0.019 | - | |||||
| 70.0–80.0 | 0.091 | - | |||||
| 60.0–80.0 | - | 0.004 | |||||
| Au–Au at GeV. Jacobian from [68], from [71]. | STAR [56, 59] | PHENIX [57] | |||||
| 0.0–5.0 | 0.109 | 0.237 | |||||
| 5.0–10.0 | 0.171 | - | |||||
| 0.0–10.0 | - | 0.129 | |||||
| 10.0–20.0 | 0.098 | 0.058 | |||||
| 20.0–30.0 | 0.086 | 0.047 | |||||
| 30.0–40.0 | 0.140 | 0.008 | |||||
| Cu–Cu at GeV. Jacobian from [68], from [58]. | STAR [58] | ||||||
| 0.0–10.0 | 0.010 | ||||||
| 10.0–20.0 | 0.006 | ||||||
| 20.0–40.0 | 0.0004 |
Table IV. Inputs for the energy density (mid-rapidity yields , Jacobian and transverse area for width and inclusive classes), resultant values of the energy density , and MSE/DoF fit metric results for the spectra shifts. Note that ALICE and STAR results were used for ATLAS and PHENIX computation respectively. The errors quoted for the area calculations correspond to the standard deviation of the area within a given bin. However, the standard error on the mean is instead used for propagation to our final uncertainty on the energy density.
The averaged event harmonics, , and the ratio , from the Glauber simulated and areas for the collision systems studied are presented in Table V. Note that the Average Radius method represents the class, while the Energy Half-Max Contour is in the class. These values are used in the model estimate for high- hadron elliptic flow .
Exploration of Fitting Procedures and Fit Metrics for Spectrum Fitting
At first glance, the choice of the MSE fit metric Eq. (8) over conventional choices like the or Likelihood appears arbitrary. This appendix provides details on the studies undertaken to select the MSE metric, where it performs better than the and Likelihood in extracting parameters of interest from spectra, particularly the high power law tail.
To begin, it is helpful to compare the MSE more directly to the from their respective definitions. The between an observed dataset and fitting expectation with corresponding error is given by
The MSE metric, defined by Eq. (8), admits the power series expansion in the case of reasonable fits
That is to say, an MSE fit is at leading order in exactly a fit with error , a generous but reasonable assumption for measured spectra for which statistical and experimental systematic uncertainties are folded together. From this point of view, it’s clear that the MSE metric is bounded below by the corresponding for the same fit, which assures that the fits discussed in Tabs. I, Table IV are of quality as good or better than for equivalent to the resulting MSE.
This error scaling additionally produces a unitless fit metric, and weights all bins equally against the fit form. This reduces the bias toward bins with larger magnitude, which bias traditional or Likelihood fits against the correct parameters for steeply falling power law spectra. To explore this quantitatively, we consider Tsallis distributions of the form Eq. (7), with parameters in ranges , GeV, and normalization chosen so that the distribution integrates to unity over the range GeV. The ranges are chosen using fit parameters from Refs. [72, 73, 74, 75, 76, 77], and roughly capture the range of fits obtained against data spectra in this study, shown in Tab. I. The distributions of input Tsallis parameters can be seen in panel (a.) of Figure 13.
Figure 13. Comparison of fitting results for the Pearson , Likelihood and MSE metrics for toy Tsallis fits against exact Tsallis distributions. Panel (a.) shows the distribution of Tsallis functions used to generate toy data over the 10,000 fits. Panels (b.), (c.) and (d.) show the distribution of resultant fit parameters , , respectively, relative to the thrown true parameters, as well as the uniformly scrambled inputs given to the fitter. Panel (e.) shows the average deviation of the ratio fit/truth from unity over the range for the different fitting methods, and panel (f.) displays the distributions of minimized fit metric values for each method.
These Tsallis distributions are then sampled to fill histograms in the same binning as in data spectra above the relevant threshold used in our analysis, either to the same statistics as in the data or to higher statistics. Then, the parameters are shuffled in a wide range, and passed in as initial guesses to ROOT’s TMinuit Migrad fitter [85, 86], set to use , Negative Log Likelihood or MSE fitting as the goodness-of-fit metric. The resultant parameters from fits are then compared against the thrown truth parameters, and the difference is tabulated in a histogram for each fit, which can be seen in panels (b.-d.) of Figure 13. In addition, the average absolute residual of the fit against the thrown distribution is tabulated in Figure 13 (e.) across the range to study the bias of fitting methods against . Finally, the distribution of resultant minimized fit metrics , , and MSE are shown for the fits in Figure 13 (f.) to compare the relative magnitudes for successful fits. A total of 10,000 sets of Tsallis parameters were generated and fit for this study.
Note that all fits are performed by comparing the sampled bin contents against the integral of the fit form in that bin. In the specific case shown in Figure 13, the histograms are extrapolated to infinite statistics, i.e. filled directly with the integral of the truth distribution, so that in principle exact parameters could be found by the fitter. This case most strongly exhibits the biases present in and Likelihood fits; despite full access to exact truth information, these methods still contain a strong bias against the correct parameters due to over-weighting bins at low . However, it is worth noting that similar behaviors were observed in fits with finite samplings, and in the case of samplings according to the statistics observed in corresponding datasets. The infinite statistics case is shown here for simplicity and brevity, as each distribution studied here and fit in Tab. I contains varying counts of hadrons.
The bias toward low in the and NLL fitters is most apparent in Figure 13 (e.), where the residuals in the ratio fit/truth exploded at high . Meanwhile, the MSE fit remains flat over the full range, indicating that it weights the full range equally. As a consequence, the resultant MSE fits extract the power law with two orders of magnitude smaller error on average than the and NLL, though the improvements for the remaining parameters , are smaller and likely a direct consequence of better extraction of . This precision extraction of the power law motivates the MSE for this study, where a good extraction of the power law tail is essential to correctly estimate .
The extracted MSE/DoF can be interpreted similarly to a , as shown in Eq. (12) and in Figure 13 (f.). Moreover, we found that when attempting to fit an incorrect form, the MSE explodes as expected for the . To show this, we fit an exponential against the spectra in the range GeV, which is of comparable order of magnitude in the fitting region but has an incorrect shape relative to the spectra. Then, we fit the Tsallis form against data sampled from this fit exponential using the , NLL and MSE fits. We found that the resultant fit metrics per degree of freedom were all larger than 5, and within of one another, which demonstrates that the MSE/DoF is comparable to a /DoF for bad fits as well as good ones.
Investigation of Fit Residuals for Subdominant Effects
The strong linear correlation observed in Figure 9 merits further exploration. Specifically, it is worthwhile to check if the residuals to the linear fits in obtained in Sec. III exhibit any clear dependence against other parameters such as , that might reveal a residual geometry or path-length component. Specifically, we consider an asymptotic expansion
with and (little-O), and investigate the form of this additional contribution
via the fit residuals
Here denotes an effective medium path length and an effective second order eccentricity/path-length difference scale ; these are expected to represent the strongest next-to-leading-order dependence though in principle may depend dynamically on a number of parameters.
The energy loss estimate and corresponding residuals plotted against are shown in Figure 14. Note that while and do not depend on the choice of area scaling method, the residuals do inherit a slight dependence on from the fit form against , which varies slightly between models as shown in Figure 9. Therefore, the residuals against both fits are shown in the breakout panels of Figure 9. The two are very similar, affirming that both the width based and inclusive models produce consistent linear fits, albeit with different slope.
Figure 14. (Top) Energy loss across for the various collision systems studied in this work. (Bottom) Residuals from the linear fit plotted against .
In the dependence of (top panel of Figure 14), a clear separation between the LHC and RHIC datasets is present, and attributable to the expected dependence of energy loss. This separation disappears in the residuals against the fit, indicating that the dependent evolution of can be well captured by . In particular, the ALICE Pb–Pb datasets at , 5.02 TeV [55] now sit between the Au–Au [56, 59] and Cu–Cu [58] datasets at GeV recorded by STAR. Note that the discrepancy between the STAR and PHENIX [57] Au–Au datasets is due to differences in the reference spectrum as discussed in Sec. III. In general, the residual data do not show a clear trend against , indicating that the system size dependence of is also generally well described by the fitting.
However, it can be seen in Figure 14 that the the larger systems (Au–Au, Pb–Pb) tend to sit above the smaller systems (Xe–Xe, Cu–Cu), which could indicate some dependence on nuclear shape, size or structure. To investigate these possibilities, the residuals are plotted against estimates of path length and path length difference in Figure 15. Both the EBE area and phenomenological methods are used to estimate , using for the EBE methods and for the phenomenological methods; for , the phenomenological is used. Results are shown for both the width class and the inclusive class in the left and right columns of Figure 15 respectively. Note that the coefficients , come from the Average Radius method for the class, and from the Energy Half-Max Contour for the class as used above, and can be seen in Tab. V. The data generally show identical trends between both EBE and phenomenological estimates of the path length , affirming the grouping of EBE and phenomenological methods into common classes performed in this study. In addition, the dependences on are reversed relative to those for , consistent with a picture of collision geometry where the most central collisions possess the most isotropic overlap area, and thus the smallest path length difference.
Figure 15. Energy loss as a function of as modeled by (first row) and the harmonic coefficient (second column), as well as the path length difference for the various systems studied. For each such dependence, the residuals from the fit are also shown in the corresponding breakout panel below. All plots in the left column correspond to the width-based area calculation , while those in the right use the inclusive area .
Because all the residual plots share the same data in the -axis, the overall ordering of datasets in the residuals is consistent among all plots of Figure 15. However, as for , the fitting against succeeds in capturing the general behavior of the datasets across both and , as well as , leaving residuals in each case that do not exhibit a clear dependence on , , or . Specifically, there is no clear leftover shape in the dependence of the residuals on or that follow a common trend among all systems studied. This strongly supports the claim made in the conclusions of Sec. IV: that the dominant behavior of can be well understood in terms of energy density alone, and further details of nuclear size, shape and collision geometry leave subdominant contributions to the partonic energy loss.
Additional predictions using
For the sake of completeness, and for comparison against the results using -based shown in Figure 11, we also show the using the in Figure 16.
Figure 16. High predictions using the and based model and for a variety of collision systems.
References
- [1]J. W. Harris and B. Müller, “QGP Signatures” revisited, Eur. Phys. J. C 84, 247 (2024).arxiv.org/abs/2308.05743
- [2]C. Loizides, J. Kamin, and D. d’Enterria, Improved Monte Carlo Glauber predictions at present and future nuclear colliders, Phys. Rev. C 97, 054910 (2018), [Erratum: Phys.Rev.C 99, 019901 (2019)], arXiv:1710.07098 [nucl-ex].arxiv.org/abs/1710.07098
- [3]K. J. Eskola, H. Honkanen, C. A. Salgado, and U. A. Wiedemann, The Fragility of high-pT hadron spectra as a hard probe, Nucl. Phys. A 747, 511 (2005), arXiv:hep-ph/0406319.DOI
- [4]G. Aad et al. (ATLAS), Comparison of inclusive and photon-tagged jet suppression in 5.02 TeV Pb+Pb collisions with ATLAS, Phys. Lett. B 846, 138154 (2023), arXiv:2303.10090 [nucl-ex].
- [5]M. Djordjevic and U. W. Heinz, Radiative energy loss in a finite dynamical QCD medium, Phys. Rev. Lett. 101, 022302 (2008), arXiv:0802.1230 [nucl-th].DOI
- [6]M. Djordjevic, Theoretical formalism of radiative jet energy loss in a finite size dynamical QCD medium, Phys. Rev. C 80, 064909 (2009), arXiv:0903.4591 [nucl-th].DOI
- [7]K. C. Zapp, F. Krauss, and U. A. Wiedemann, A perturbative framework for jet quenching, JHEP 03, 080, arXiv:1212.1599 [hep-ph].DOI
- [8]K. M. Burke et al. (JET), Extracting the jet transport coefficient from jet quenching in high-energy heavy-ion collisions, Phys. Rev. C 90, 014909 (2014), arXiv:1312.5003 [nucl-th].
- [9]M. Spousta and B. Cole, Interpreting single jet measurements in Pb + Pb collisions at the LHC, Eur. Phys. J. C 76, 50 (2016), arXiv:1504.05169 [hep-ph].DOI
- [10]C. Andrés, N. Armesto, M. Luzum, C. A. Salgado, and P. Zurita, Energy loss versus centrality dependence of the jet quenching parameter q̂ at RHIC and LHC: a new puzzle?, Eur. Phys. J. C 76, 475 (2016), arXiv:1606.04837 [hep-ph].DOI
- [11]F. Arleo, Quenching of Hadron Spectra in Heavy Ion Collisions at the LHC, Phys. Rev. Lett. 119, 062302 (2017), arXiv:1703.10852 [hep-ph].DOI
- [12]D. Zigic, I. Salom, J. Auvinen, M. Djordjevic, and M. Djordjevic, DREENA-B framework: first predictions of RAA and v2 within dynamical energy loss formalism in evolving QCD medium, Phys. Lett. B 791, 236 (2019), arXiv:1805.04786 [nucl-th].DOI
- [13]D. Zigic, I. Salom, J. Auvinen, M. Djordjevic, and M. Djordjevic, DREENA-C framework: joint RAA and v2 predictions and implications to QGP tomography, J. Phys. G 46, 085101 (2019), arXiv:1805.03494 [nucl-th].
- [14]Y. Mehtar-Tani, D. Pablos, and K. Tywoniuk, Cone-Size Dependence of Jet Suppression in Heavy-Ion Collisions, Phys. Rev. Lett. 127, 252301 (2021), arXiv:2101.01742 [hep-ph].
- [15]D. Zigic, I. Salom, J. Auvinen, P. Huovinen, and M. Djordjevic, DREENA-A framework as a QGP tomography tool, Front. in Phys. 10, 957019 (2022), arXiv:2110.01544 [nucl-th].DOI
- [16]B. Karmakar, D. Zgic, I. Salom, J. Auvinen, P. Huovinen, M. Djordjevic, and M. Djordjevic, Constraining η/s through high-pT theory and data, Phys. Rev. C 108, 044907 (2023), arXiv:2305.11318 [hep-ph].
- [17]B. Karmakar, D. Zgic, M. Djordjevic, P. Huovinen, M. Djordjevic, and J. Auvinen, Probing the shape of the quark-gluon plasma droplet via event-by-event quark-gluon plasma tomography, Phys. Rev. C 110, 044906 (2024), arXiv:2403.17817 [hep-ph].arxiv.org/abs/2403.17817
- [18]C. Faraday and W. A. Horowitz, Collisional and radiative energy loss in small systems, Phys. Rev. C 111, 054911 (2025), arXiv:2408.14426 [nucl-th].DOI
- [19]C. Faraday and W. A. Horowitz, A unified description of small, peripheral, and large system suppression data from pQCD, Phys. Lett. B 864, 139437 (2025), arXiv:2411.09647 [hep-ph].DOI
- [20]D. Pablos and A. Takacs, Bayesian Constraints on Pre-Equilibrium Jet Quenching and Predictions for Oxygen Collisions, (2025), arXiv:2509.19430 [hep-ph].DOI
- [21]C. Andres, N. Armesto, H. Niemi, R. Paatelainen, and C. A. Salgado, Jet quenching as a probe of the initial stages in heavy-ion collisions, Phys. Lett. B 803, 135318 (2020), arXiv:1902.03231 [hep-ph].DOI
- [22]J. Wu, W. K. Ke, and X.-N. Wang, Bayesian inference of the path-length dependence of jet energy loss, Phys. Rev. C 108, 034911 (2023), arXiv:2304.06339 [hep-ph].arxiv.org/abs/2304.06339
- [23]F. Arleo and G. Falmagne, Probing the path-length dependence of parton energy loss via scaling properties in heavy ion collisions, Phys. Rev. D 109, L051503 (2024), arXiv:2212.01324 [hep-ph].arxiv.org/abs/2212.01324
- [24]A. Ogrodnik, M. Rybar, and M. Spousta, Flavor and path-length dependence of jet quenching from inclusive jet and γ-jet suppression, Eur. Phys. J. C 85, 899 (2025), arXiv:2407.11234 [hep-ph].DOI
- [25]Y. Mehtar-Tani, D. Pablos, and K. Tywoniuk, Jet suppression and azimuthal anisotropy from RHIC to LHC, Phys. Rev. D 110, 014009 (2024), arXiv:2402.07869 [hep-ph].
- [26]Y. He, L.-G. Pang, and X.-N. Wang, Bayesian extraction of jet energy loss distributions in heavy-ion collisions, Phys. Rev. Lett. 122, 252302 (2019), arXiv:1808.05310 [hep-ph].DOI
- [27]S.-L. Zhang, J. Liao, G.-Y. Qin, E. Wang, and H. Xing, Unraveling gluon jet quenching through J/ψ production in heavy-ion collisions, Sci. Bull. 68, 2003 (2023), arXiv:2208.08323 [hep-ph].DOI
- [28]W.-J. Xing, S. Cao, and G.-Y. Qin, Flavor hierarchy of parton energy loss in quark-gluon plasma from a Bayesian analysis, Phys. Lett. B 850, 138523 (2024), arXiv:2303.12485 [hep-ph].DOI
- [29]S.-L. Zhang, E. Wang, H. Xing, and B.-W. Zhang, Flavor dependence of jet quenching in heavy-ion collisions from a Bayesian analysis, Phys. Lett. B 850, 138549 (2024), arXiv:2303.14881 [hep-ph].DOI
- [30]R. Ehlers et al., Bayesian inference analysis of jet quenching using inclusive jet and hadron suppression measurements, Phys. Rev. C 111, 054913 (2025), arXiv:2408.08247 [hep-ph].arxiv.org/abs/2408.08247
- [31]A. Falcão and K. Tywoniuk, Constraining jet quenching in heavy-ion collisions with Bayesian inference, JHEP 02, 069, arXiv:2411.14552 [hep-ph].arxiv.org/abs/2411.14552
- [32]Y. He, S. Cao, W. Chen, T. Luo, L.-G. Pang, and X.-N. Wang, Interplaying mechanisms behind single inclusive jet suppression in heavy-ion collisions, Phys. Rev. C 99, 054911 (2019), arXiv:1809.02525 [nucl-th].DOI
- [33]S.-L. Zhang, X.-N. Wang, and B.-W. Zhang, Quenching of jets tagged with W bosons in high-energy nuclear collisions, Phys. Rev. C 105, 054902 (2022), arXiv:2103.07836 [hep-ph].DOI
- [34]R. A. Soltz, D. A. Hangal, and A. Angameri, Simple model to investigate jet quenching and correlated errors for centrality-dependent nuclear modification factors in relativistic heavy-ion collisions, Phys. Rev. C 111, 034911 (2025), arXiv:2412.03724 [nucl-th].arxiv.org/abs/2412.03724
- [35]Y.-T. Chien and R. Kunnawalkam Elayavalli, Probing heavy ion collisions using quark and gluon jet substructure, (2018), arXiv:1803.03589 [hep-ph].arxiv.org/abs/1803.03589
- [36]S.-Y. Chen, B.-W. Zhang, and E.-K. Wang, Jet charge in high energy nuclear collisions, Chin. Phys. C 44, 024103 (2020), arXiv:1908.01518 [nucl-ex].DOI
- [37]G.-Y. Neufeld, I. Vitev, and B. W. Zhang, The Physics of Z0/γ*-tagged jets at the LHC, Phys. Rev. C 83, 034902 (2011), arXiv:1006.2389 [hep-ph].
- [38]J. Brewer, J. Thaler, and A. P. Turner, Data-driven quark and gluon jet modification in heavy-ion collisions, Phys. Rev. C 103, L021901 (2021), arXiv:2008.08596 [hep-ph].DOI
- [39]A. Takacs and K. Tywoniuk, Quenching effects in the cumulative jet spectrum, JHEP 10, 038, arXiv:2103.14676 [hep-ph].DOI
- [40]Y. Mehtar-Tani, J. G. Milhano, and K. Tywoniuk, Jet physics in heavy-ion collisions, Int. J. Mod. Phys. A 28, 1340013 (2013), arXiv:1302.2579 [hep-ph].
- [41]G.-Y. Qin and X.-N. Wang, Jet quenching in high-energy heavy-ion collisions, Int. J. Mod. Phys. E 24, 1530014 (2015), arXiv:1511.00790 [hep-ph].DOI
- [42]S. Cao, A. Majumder, R. Modarresi-Yazdi, I. Soudi, and Y. Tachibana, Jet quenching: From theory to simulation, Int. J. Mod. Phys. E 33, 2430002 (2024), arXiv:2401.10026 [hep-ph].arxiv.org/abs/2401.10026
- [43]J.-F. Paquet, Applications of emulation and Bayesian methods in heavy-ion physics, J. Phys. G 51, 103001 (2024), arXiv:2310.17618 [nucl-th].DOI
- [44]A. Majumder and M. Van Leeuwen, The Theory and Phenomenology of Perturbative QCD Based Jet Quenching, Prog. Part. Nucl. Phys. 66, 41 (2011), arXiv:1002.2206 [hep-ph].
- [45]K. Adcox et al. (PHENIX), Formation of dense partonic matter in relativistic nucleus-nucleus collisions at RHIC: Experimental evaluation by the PHENIX collaboration, Nucl. Phys. A 757, 184 (2005), arXiv:nucl-ex/0410003.DOI
- [46]J. Brewer, J. G. Milhano, and J. Thaler, Sorting out quenched jets, Phys. Rev. Lett. 122, 222301 (2019), arXiv:1812.05111 [hep-ph].DOI
- [47]A. Adare et al. (PHENIX), Scaling properties of fractional momentum loss of high-pT hadrons in nucleus-nucleus collisions at √sNN from 62.4 GeV to 2.76 TeV, Phys. Rev. C 93, 024911 (2016), arXiv:1509.06735 [nucl-ex].DOI
- [48]N. R. Sahoo (STAR), Measurement of γ+jet and π0+jet in central Au+Au collisions at √sNN = 200 GeV with the STAR experiment, PoS HardProbes2020, 132 (2021), arXiv:2008.08789 [nucl-ex].arxiv.org/abs/2008.08789
- [49]A. Adare et al. (PHENIX), Neutral pion production with respect to centrality and reaction plane in Au+Au collisions at √sNN = 200 GeV, Phys. Rev. C 87, 034911 (2013), arXiv:1208.2254 [nucl-ex].DOI
- [50]A. Adare et al. (PHENIX), Evolution of π⁰ suppression in Au+Au collisions from √sNN = 39 to 200 GeV, Phys. Rev. Lett. 109, 152301 (2012), [Erratum: Phys.Rev.Lett. 125, 049901 (2020)], arXiv:1204.1526 [nucl-ex].
- [51]S. S. Adler et al. (PHENIX), Detailed Study of High-pT Neutral Pion Suppression and Azimuthal Anisotropy in Au+Au Collisions at √sNN = 200 GeV, Phys. Rev. C 76, 034904 (2007), arXiv:nucl-ex/0611007.
- [52]R. Baier, Y. L. Dokshitzer, A. H. Mueller, and D. Schiff, Quenching of hadron spectra in media, JHEP 09, 033, arXiv:hep-ph/0106347.
- [53]S. Acharya et al. (ALICE), Transverse momentum spectra and nuclear modification factors of charged particles in Xe–Xe collisions at √sNN = 5.44 TeV, Phys. Lett. B 788, 166 (2019), arXiv:1805.04399 [nucl-ex].
- [54]G. Aad et al. (ATLAS), Charged-hadron production in pp, p+Pb, Pb+Pb, and Xe+Xe collisions at √sNN = 5 TeV with the ATLAS detector at the LHC, JHEP 07, 074, arXiv:2211.15257 [hep-ex].
- [55]S. Acharya et al. (ALICE), Transverse momentum spectra and nuclear modification factors of charged particles in pp, p–Pb and Pb–Pb collisions at the LHC, JHEP 11, 013, arXiv:1802.09145 [nucl-ex].
- [56]J. Adams et al. (STAR), Transverse momentum and collision energy dependence of high pT hadron suppression in Au+Au collisions at ultrarelativistic energies, Phys. Rev. Lett. 91, 172302 (2003), arXiv:nucl-ex/0305015.
- [57]S. S. Adler et al. (PHENIX), High pT charged hadron suppression in Au+Au collisions at √sNN = 200 GeV, Phys. Rev. C 69, 034910 (2004), arXiv:nucl-ex/0308006.
- [58]B. I. Abelev et al. (STAR), Spectra of identified high-pT π± and p(p̄) in Cu+Cu collisions at √sNN = 200 GeV, Phys. Rev. C 81, 054907 (2010), arXiv:0911.3130 [nucl-ex].
- [59]G. Agakishiev et al. (STAR), Identified hadron compositions in p+p and Au+Au collisions at high transverse momenta at √sNN = 200 GeV, Phys. Rev. Lett. 108, 072302 (2012), arXiv:1110.0579 [nucl-ex].
- [60]C. Loizides, Glauber modeling of high-energy nuclear collisions at the subnucleon level, Phys. Rev. C 94, 024914 (2016), arXiv:1603.07375 [nucl-ex].
- [61]B. Schenke, P. Tribedy, and R. Venugopalan, Fluctuating Glasma initial conditions and flow in heavy ion collisions, Phys. Rev. Lett. 108, 252301 (2012), arXiv:1202.6646 [nucl-th].
- [62]J. E. Bernhard, J. S. Moreland, and S. A. Bass, Bayesian estimation of the specific shear and bulk viscosity of quark–gluon plasma, Nature Phys. 15, 1113 (2019).
- [63]S. Acharya et al. (ALICE), Centrality determination in heavy ion collisions, CERN Document Server (2018).
- [64]C. Adler et al. (STAR), Centrality dependence of high pT hadron suppression in Au+Au collisions at √sNN = 130 GeV, Phys. Rev. Lett. 89, 202301 (2002), arXiv:nucl-ex/0206011.
- [65]B. B. Back et al. (PHOBOS), Centrality dependence of the charged particle multiplicity near mid-rapidity in Au + Au collisions at √s (NN) = 130-GeV and 200-GeV, Phys. Rev. C 65, 061901 (2002), arXiv:nucl-ex/0201005.
- [66]S. Acharya et al. (ALICE), System-size dependence of the charged-particle pseudorapidity density at √sNN = 5.02 TeV for pp, pPb, and Pb–Pb collisions, Phys. Lett. B 845, 137730 (2023), arXiv:2204.10210 [nucl-ex].
- [67]J. D. Bjorken, Highly Relativistic Nucleus-Nucleus Collisions: The Central Rapidity Region, Phys. Rev. D 27,DOI
- [68]A. Adare et al. (PHENIX), Transverse energy production and charged-particle multiplicity at midrapidity in various systems from √sNN = 7.7 to 200 GeV, Phys. Rev. C 93, 024901 (2016), arXiv:1509.06727 [nucl-ex].
- [69]S. Chatrchyan et al. (CMS Collaboration), Measurement of the Pseudorapidity and Centrality Dependence of the Transverse Energy Density in Pb–Pb Collisions at √sNN = 2.76 TeV, Phys. Rev. Lett. 109, 152303 (2012).
- [70]K. Aamodt et al. (ALICE), Centrality dependence of the charged-particle multiplicity density at mid-rapidity in Pb–Pb collisions at √sNN = 2.76 TeV, Phys. Rev. Lett. 106, 032301 (2011), arXiv:1012.1657 [nucl-ex].
- [71]I. G. Bearden et al. (BRAHMS), Pseudorapidity distributions of charged particles from Au+Au collisions at the maximum RHIC energy, Phys. Rev. Lett. 88, 202301 (2002), arXiv:nucl-ex/0112001.
- [72]C.-Y. Wong and G. Wilk, Collisional Fits to pT Spectra for pp Collisions at LHC, Acta Phys. Polon. B 43, 2047 (2012), arXiv:1210.3661 [hep-ph].
- [73]G. Wilk and Z. Wlodarczyk, On the interpretation of nonextensive parameter q in Tsallis statistics and Levy distributions, Phys. Rev. Lett. 84, 2770 (2000), arXiv:hep-ph/9908459.
- [74]G. Aad et al. (ATLAS), Charged-particle multiplicities in pp interactions measured with the ATLAS detector at the LHC, New J. Phys. 13, 053033 (2011), arXiv:1012.5104 [hep-ex].
- [75]B. B. Abelev et al. (ALICE), Energy Dependence of the Transverse Momentum Distributions of Charged Particles in pp Collisions Measured by ALICE, Eur. Phys. J. C 73, 2662 (2013), arXiv:1307.1093 [nucl-ex].
- [76]B. I. Abelev et al. (STAR), Strange particle production in p+p collisions at √s = 200 GeV, Phys. Rev. C 75, 064901 (2007), arXiv:nucl-ex/0607033.
- [77]A. Adare et al. (PHENIX), Identified charged hadron production in pp collisions at √s = 200 and 62.4 GeV, Phys. Rev. C 83, 064903 (2011), arXiv:1102.0753 [nucl-ex].
- [78]G. Giacalone, A. Mazeliauskas, and S. Schlichting, Hydrodynamic attractors, initial state energy and particle production in relativistic nuclear collisions, Phys. Rev. Lett. 123, 262301 (2019), arXiv:1908.02866 [hep-ph].
- [79]S. S. Adler et al. (PHENIX), Mid-rapidity neutral pion production in proton proton collisions at √s = 200 GeV, Phys. Rev. Lett. 91, 241803 (2003), arXiv:hep-ex/0304038.
- [80]S. Acharya et al. (ALICE), Energy dependence and fluctuations of anisotropic flow in Pb–Pb collisions at √sNN = 5.02 and 2.76 TeV, JHEP 07, 103, arXiv:1804.02944 [nucl-ex].
- [81]G. Giacalone, J. Noronha-Hostler, M. Luzum, and J.-Y. Ollitrault, Hydrodynamic predictions for 5.44 TeV Xe+Xe collisions, Phys. Rev. C 97, 034904 (2018), arXiv:1711.08499 [nucl-th].DOI
- [82]S. Acharya et al. (ALICE), Anisotropic flow in Xe–Xe collisions at √sNN = 5.44 TeV, Phys. Lett. B 784, 82 (2018), arXiv:1805.01832 [nucl-ex].
- [83]G. Aad et al. (ATLAS), Measurement of the azimuthal anisotropy of charged-particle production in Xe+Xe collisions at √sNN = 5.44 TeV with the ATLAS detector, Phys. Rev. C 101, 024906 (2020), arXiv:1911.04812 [nucl-ex]. P. Möller, A. J. Sierk, T. Ichikawa, and H. Sagawa, Nuclear ground-state masses and deformations: FRDM(2012), Atom. Data Nucl. Data Tabl. 109-110, 1 (2016), arXiv:1508.06294 [nucl-th].
- [85]R. Fletcher, A new approach to variable metric algorithms, The Computer Journal 13, 317 (1970), https://academic.oup.com/comjnl/article-pdf/13/3/317/988678/130317.pdf.DOI
- [86]F. James and M. Roos, Minuit: A System for Function Minimization and Analysis of the Parameter Errors and Correlations, Comput. Phys. Commun. 10, 343 (1975).














