betaFeedback

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 pTp_{\mathrm{T}} hadron yields in A-A collisions relative to appropriately scaled pppp 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 pTp_{\mathrm{T}} spectra across different sNN\sqrt{s_{\mathrm{NN}}}. To decouple these effects, we utilize a phenomenologically motivated spectrum shift model to estimate the average transverse momentum loss ΔpTΔp_{\mathrm{T}} imparted on high pTp_{\mathrm{T}} partons in A-A collisions, a proxy for the medium induced energy loss. We observe a striking correlation between ΔpTΔp_{\mathrm{T}} and Glauber-derived estimates of initial state energy density εBj\varepsilon_{\mathrm{Bj}}, 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-pTp_{\mathrm{T}} hadron elliptic flow v2v_2 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-pTp_{\mathrm{T}}) hadrons in nucleus-nucleus (A–A) collisions relative to their production in pppp collisions scaled by the mean number of binary nucleon-nucleon collisions ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle so as to make the two comparable. Experimentally, this suppression is frequently identified via the nuclear modification factor

RAA=1⟨TAA⟩d3NchAA/dpTdηdϕd3σchpp/dpTdηdϕ,(1)R_{\mathrm{AA}} = \frac{1}{\langle T_{\mathrm{AA}}\rangle}\frac{\mathrm{d}^{3}N_{\mathrm{ch}}^{\mathrm{AA}}/\mathrm{d}p_{\mathrm{T}}\mathrm{d}\eta\mathrm{d}\phi}{\mathrm{d}^{3}\sigma_{\mathrm{ch}}^{pp}/\mathrm{d}p_{\mathrm{T}}\mathrm{d}\eta\mathrm{d}\phi}, \tag*{(1)}

where ⟨TAA⟩=⟨Ncoll⟩/σinelNN\langle T_{\mathrm{AA}}\rangle= \langle N_{\mathrm{coll}}\rangle/\sigma_{\mathrm{inel}}^{\mathrm{NN}} is the nuclear overlap function determined from Glauber model calculations [2], proportional to ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle via the inelastic nucleon-nucleon cross section σinelNN\sigma_{\mathrm{inel}}^{\mathrm{NN}} at the relevant center of mass energy per colliding nucleon pair sNN\sqrt{s_{\mathrm{NN}}}. NchAAN_{\mathrm{ch}}^{\mathrm{AA}} and σchpp\sigma_{\mathrm{ch}}^{pp} denote the charged particle yield per event in A–A collisions and the charged particle production cross section in pppp collisions, respectively.

The observation of RAAR_{\mathrm{AA}} 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 pTp_{\mathrm{T}} range but especially at high-pTp_{\mathrm{T}}, 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 RAAR_{\mathrm{AA}} suppression values for different collision sNN\sqrt{s_{\mathrm{NN}}} are comparable in principle, RAAR_{\mathrm{AA}} 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. RAAR_{\mathrm{AA}} is strictly a ratio of yields, but is identified with energy loss because a uniform change in pTp_{\mathrm{T}} does not uniformly affect pTp_{\mathrm{T}}-differential yields for steeply falling pTp_{\mathrm{T}} spectra. LHC pTp_{\mathrm{T}} 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 RAAR_{\mathrm{AA}} at RHIC and the LHC actually indicate more energy loss for partons traversing QGP created at the LHC. Additionally, observed differences between the RAAR_{\mathrm{AA}} for photon-tagged jets and inclusive jets, such as that reported by ATLAS in sNN=5.02\sqrt{s_{\mathrm{NN}}}=5.02 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 RAAR_{\mathrm{AA}}.

The abundance of available jet quenching data from experimental collaborations such as high-pTp_{\mathrm{T}} RAAR_{\mathrm{AA}} 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 RAAR_{\mathrm{AA}} results has led several collaborations to study the shift in pTp_{\mathrm{T}}—first proposed as SlossS_{\mathrm{loss}} by PHENIX [45]—needed to align the spectra in A–A with the binary scaled pppp. Subsequent theoretical and phenomenological studies of jet quenching have also explored similar measures of energy loss related to horizontal shifts of pTp_{\mathrm{T}} spectra [9, 11, 14, 26, 31, 32, 34, 39, 46]; we will refer to this strategy broadly as ΔpT\Delta p_{\mathrm{T}} methods. Unlike RAAR_{\mathrm{AA}}, these ΔpT\Delta p_{\mathrm{T}} measures perform a direct fitting between hadron spectra, significantly reducing effects of pTp_{\mathrm{T}} distribution shape on the results. Measurements of SlossS_{\mathrm{loss}} and other similar ΔpT\Delta p_{\mathrm{T}} 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 ΔpT\Delta p_{\mathrm{T}} from experimentally observed charged particle spectra at high pTp_{\mathrm{T}} in pppp and A–A collisions. This work differs from previous studies by assuming a fixed ΔpT\Delta p_{\mathrm{T}} 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 pTp_{\mathrm{T}} ranges studied in this work make our fixed ΔpT\Delta p_{\mathrm{T}} approximation reasonable. In addition, given that high pTp_{\mathrm{T}} 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 pTp_{\mathrm{T}} 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 ϵBj\epsilon_{\mathrm{Bj}}. By further utilizing Glauber event geometry, we predict the hadronic high-pTp_{\mathrm{T}} v2v_{2} assuming a linear path-length dependent energy loss. The v2v_{2} 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 ΔpT\Delta p_{\mathrm{T}} (II C) and the high-pTp_{\mathrm{T}} v2v_{2} 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, ⟨A⊥⟩\langle A_{\perp}\rangle, are made using Monte Carlo Glauber calculations [2]. The initial energy density is then approximated using ⟨A⊥⟩\langle A_{\perp}\rangle and the charged particle multiplicities dNch/dη\mathrm{d}N_{\mathrm{ch}}/\mathrm{d}\eta that have been determined experimentally. Second, the ΔpT\Delta p_{\mathrm{T}} of each dataset is determined from the reported charged particle A–A and pppp transverse momentum spectra. Once both the energy density and ΔpT\Delta p_{\mathrm{T}} 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 ΔpT\Delta p_{\mathrm{T}} enable prediction of the high-pTp_{\mathrm{T}} v2v_{2} 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 sNN=5.44\sqrt{s_{\mathrm{NN}}}=5.44 TeV, ALICE [55] Pb–Pb results at sNN=5.02\sqrt{s_{\mathrm{NN}}}=5.02 TeV and 2.76 TeV, STAR [56] and PHENIX [57] charged particle spectra from Au–Au collisions at sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV, and STAR π±\pi^{\pm} data from Cu–Cu [58] also at sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV. Corresponding pppp spectra were obtained from the same A–A references for all datasets except STAR Au–Au, for which 200 GeV pppp 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 ⟨A⊥⟩\langle A_{\perp}\rangle must be determined from simulation, and there is currently no definitive technique to perform this extraction. We therefore start this study by determining ⟨A⊥⟩\langle A_{\perp}\rangle using several physically reasonable approaches, with the goal of determining if any exhibit similar behavior. First, we utilize the three grid-based ⟨A⊥⟩\langle A_{\perp}\rangle calculations directly available from the Glauber code and described in Ref. [60]. We then present four new ⟨A⊥⟩\langle A_{\perp}\rangle 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, AWA_{\mathrm{W}}, is calculated from the (co-)variances of the nucleon distributions; the inclusive area, AUA_{\mathrm{U}}, is a grid-based set union including a disk around all participant nucleons; the exclusive area, ANA_{\mathrm{N}}, 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 E(x,y)E(x,y) 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 E∝TATBE \propto\sqrt{T_A T_B}. 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 Ψ2\Psi_2. 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 bb 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 ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle itself [56, 58, 64, 65], a systematic error is included on ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle as follows. The Glauber Monte Carlo [2] is run with a variation of inelastic cross section with uncertainty σinelNN±δσinelNN\sigma_{\mathrm{inel}}^{\mathrm{NN}} \pm\delta\sigma_{\mathrm{inel}}^{\mathrm{NN}} as reported by the collaboration, and also with nuclear shape parameters β2,β4,γ\beta_2,\beta_4,\gamma disabled. The ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle values reported by the collaboration are also tabulated. To obtain a systematic uncertainty on ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle, we compare the nominal Glauber ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle results—for which σinelNN\sigma_{\mathrm{inel}}^{\mathrm{NN}} is always obtained from Ref. [2]—against each of the cases mentioned above. The largest difference δ⟨Ncoll⟩\delta\langle N_{\mathrm{coll}}\rangle against any variation in each centrality bin is symmetrized and taken as the systematic error on ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle 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 R(ϕ)R(\phi). The area contained within this curve is the transverse area. Four methods were considered to determine R(ϕ)R(\phi): the average energy radius

R⟨E⟩(ϕ)=1E0∫0∞rE(r,ϕ) dr,(2)R_{\langle E\rangle}(\phi)=\frac{1}{E_0}\int_0^\infty rE(r,\phi)\,\mathrm{d}r, \tag*{(2)}

the average pressure (energy gradient) radius

R⟨P⟩(ϕ)=1δE0∫0∞r∣∇E(r,ϕ)∣ dr.(3)R_{\langle P\rangle}(\phi)=\frac{1}{\delta E_0}\int_0^\infty r\lvert\nabla E(r,\phi)\rvert\,\mathrm{d}r. \tag*{(3)}

For simplicity, we choose E0=1 GeV/fmE_0=1\ \mathrm{GeV}/\mathrm{fm}, δE0=1 GeV/fm2\delta E_0=1\ \mathrm{GeV}/\mathrm{fm}^{2}; 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

E(RFWHM(ϕ),ϕ)=Emax⁡2,(4)E(R_{\mathrm{FWHM}}(\phi),\phi)=\frac{E_{\max}}{2}, \tag*{(4)}

and the surface of maximal gradient/pressure

RP,max⁡(ϕ)=arg max⁡r≥0∣∇E(r,ϕ)∣.(5)R_{P,\max}(\phi)=\underset{r\ge0}{\operatorname{arg\,max}}\lvert\nabla E(r,\phi)\rvert. \tag*{(5)}

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 AWA_{\mathrm{W}} 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 R⟨E⟩(ϕ)R_{\langle E\rangle}(\phi) represents the average total energy seen by a traversing high energy parton emitted at angle ϕ\phi), it is not immediately clear which method, if any, might best represent the effective ⟨A⊥⟩\langle A_{\perp}\rangle 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 sNN=2.76\sqrt{s_{\mathrm{NN}}}=2.76 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 sNN=2.76\sqrt{s_{\mathrm{NN}}}=2.76 TeV. Ratios of certain methods against (b.) A∪A_{\cup} and (c.) AWA_{\mathrm{W}}. The blue and orange markers denote the class of methods that scale like A∪A_{\cup} and AWA_{\mathrm{W}} 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 A∩A_{\cap}. 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 A∪A_{\cup} and the width AWA_W 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 A∪A_{\cup} 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 A∪A_{\cup}. Figure 1(c) shows ratios against the EBE width AWA_W 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 AWA_W. Note that the pre-averaged EBE width calculation AWA_W and the post-averaged phenomenological width calculation exhibit ratios consistent with unity, a check on the event averaging procedure. In each case, A∩A_{\cap} exhibits markedly different scaling, shown by non-constant ratios. The EBE inclusive A∪A_{\cup} and width AWA_W also do not have consistent scalings, meaning they cannot be merged to a single class.

While initially disconcerting, the disagreement between A∩A_{\cap} and other methods also coincides with observations in ALICE ppPb, where computations using A∩A_{\cap} produce unusually high estimates of energy density [66]. Physically, the A∩A_{\cap} and A∪A_{\cup} methods reflect distinct pictures of Glauber modeling: A∪A_{\cup} considers entire struck nucleons as participants in traditional Glauber fashion, while A∩A_{\cap} 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 sNN=2.76\sqrt{s_{\mathrm{NN}}}=2.76 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 εini\varepsilon_{\mathrm{ini}} as a function of centrality is approximated in the limit of Bjorken hydrodynamics [67] as outlined in Ref. [1]; this results in the approximation

εini∼εBj≈34(7JdNchdη⟨A⊥⟩biniτini)137JdNchdη⟨A⊥⟩τini,(6)\varepsilon_{\mathrm{ini}} \sim\varepsilon_{\mathrm{Bj}} \approx\frac{3}{4}\left(\frac{7J\frac{\mathrm{d}N_{\mathrm{ch}}}{\mathrm{d}\eta}}{\langle A_{\perp}\rangle b_{\mathrm{ini}}\tau_{\mathrm{ini}}}\right)^{\frac{1}{3}}\frac{7J\frac{\mathrm{d}N_{\mathrm{ch}}}{\mathrm{d}\eta}}{\langle A_{\perp}\rangle\tau_{\mathrm{ini}}}, \tag*{(6)}

where JJ is the Jacobian conversion from pseudorapidity to rapidity, dNch/dη\mathrm{d}N_{\mathrm{ch}}/\mathrm{d}\eta is the charged hadron density at midrapidity, τini\tau_{\mathrm{ini}} is the QGP formation time, and binib_{\mathrm{ini}} is defined by sini=biniTini3s_{\mathrm{ini}}=b_{\mathrm{ini}}T_{\mathrm{ini}}^3; sinis_{\mathrm{ini}} and TiniT_{\mathrm{ini}} are the initial entropy density and temperature respectively. As in Ref. [1], binib_{\mathrm{ini}} is assumed to be constant and independent of collision energy, species, and centrality with a value of bini=15.5b_{\mathrm{ini}}=15.5. For highly Lorentz contracted nuclei the parameter τini\tau_{\mathrm{ini}} is commonly chosen to be 0.6 fm/c, while at collision energies sNN<20\sqrt{s_{\mathrm{NN}}}<20 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 sNN≥200\sqrt{s_{\mathrm{NN}}}\ge200 GeV, we also fix τini=0.6\tau_{\mathrm{ini}}=0.6 fm/c.

The remaining input parameters were obtained from various sources. The Jacobian JJ ranges from 1.25 for sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV to 1.09 at sNN=5.44\sqrt{s_{\mathrm{NN}}}=5.44 TeV [68, 69]. An uncertainty of 3% on JJ is assumed for all beam energies, propagated forward to the uncertainty on εBj\varepsilon_{\mathrm{Bj}}. The dNch/dη\mathrm{d}N_{\mathrm{ch}}/\mathrm{d}\eta data used in this analysis are those reported from Pb–Pb collisions at sNN=2.76\sqrt{s_{\mathrm{NN}}}=2.76 TeV and sNN=5.02\sqrt{s_{\mathrm{NN}}}=5.02 TeV from ALICE [55, 70], Xe–Xe collisions at sNN=5.44\sqrt{s_{\mathrm{NN}}}=5.44 TeV from ALICE [53], Au–Au collisions at sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV from STAR [71], and lastly Cu–Cu collisions at sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV from STAR [58]. The uncertainty used on dNch/dη\mathrm{d}N_{\mathrm{ch}}/\mathrm{d}\eta is the quadrature sum of the reported statistical and systematic uncertainties in each measurement.

The transverse area ⟨A⊥⟩\langle A_{\perp}\rangle is determined for each species, beam energy, and corresponding centrality class as described in the previous section. Uncertainties originating from dNch/dη\mathrm{d}N_{\mathrm{ch}}/\mathrm{d}\eta, ⟨A⊥⟩\langle A_{\perp}\rangle and JJ 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 εBjwidth\varepsilon_{\mathrm{Bj}}^{\mathrm{width}}, εBjinclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{inclusive}}, exclusive representing width-based AWA_W, inclusive A∪A_{\cup}, and exclusive A∩A_{\cap} areas respectively. Details of the resultant values and uncertainties of εBjwidth\varepsilon_{\mathrm{Bj}}^{\mathrm{width}} and εBjinclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{inclusive}} computed from the above inputs are given in Appendix § 2.1, along with the relevant data used to compute these quantities. The exclusive area εBjexclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{exclusive}} 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 A∩A_{\cap} predicts higher energy density in pPb collisions than in central Pb–Pb. Still, to enable a fair comparison among all methods, we do include εBjexclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{exclusive}} in the final results of Sec. III.

Extracting ΔpT\Delta p_{\mathrm{T}} from Transverse Momentum Spectrum Data

In previous analyses [4, 45, 46, 47, 48, 49, 50, 51] ΔpT\Delta p_{\mathrm{T}} was determined for a fixed hadron pTp_{\mathrm{T}}. Using the transverse momentum spectra found in [53, 54, 55, 56, 57, 58, 59], this study explores if a common ΔpT\Delta p_{\mathrm{T}} can be identified for the whole range of high-pTp_{\mathrm{T}} data reported. We identify such a ΔpT\Delta p_{\mathrm{T}} via a pTp_{\mathrm{T}} spectrum shifting procedure described in this section.

The horizontal shifting of the pppp pTp_{\mathrm{T}} 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 pTp_{\mathrm{T}} value, pTmin⁡p_{\mathrm{T}}^{\min}, must therefore be chosen to ensure that other collective effects—such as radial flow—present at lower pTp_{\mathrm{T}} do not affect the extraction. Data below this pTp_{\mathrm{T}} threshold are not included in the determination of ΔpT\Delta p_{\mathrm{T}}. To determine this threshold we consider the reported charged hadron RAAR_{\mathrm{AA}}. In all measurements studied, the RAAR_{\mathrm{AA}} in central collisions exhibit a local minimum near pT∼5p_{\mathrm{T}} \sim5 GeV/c, whereafter the RAAR_{\mathrm{AA}} 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 pTmin⁡p_{\mathrm{T}}^{\min}. The systematic uncertainty on ΔpT\Delta p_{\mathrm{T}} due to this choice of threshold is determined by varying the pTp_{\mathrm{T}} threshold by an additional pTp_{\mathrm{T}} bin below and above each dataset’s local RAAR_{\mathrm{AA}} minima.

Since many of the peripheral datasets suffer from large uncertainties, preventing a clear determination of a local minimum, we decided to use the pTmin⁡p_{\mathrm{T}}^{\min} extracted in the most central events for all centralities reported for a given dataset. This likely results in a slight overestimate of the pTp_{\mathrm{T}} 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 pTp_{\mathrm{T}} threshold chosen, we compute the ΔpT\Delta p_{\mathrm{T}} by scaling and shifting the pppp reference spectra to match the A–A data using the following procedure:

  1. Reference pppp spectra are first fit to a Tsallis distribution, given by

12πpTd2NdpTdη∼Ed3Ndp3=C(1+ETnT)−n,(7)\frac{1}{2\pi p_{\mathrm{T}}}\frac{\mathrm{d}^{2}N}{\mathrm{d}p_{\mathrm{T}}\mathrm{d}\eta} \sim E\frac{\mathrm{d}^{3}N}{\mathrm{d}p^{3}} = C\left(1+\frac{E_{\mathrm{T}}}{nT}\right)^{-n}, \tag*{(7)}

where ET=m2+pT2−mE_{\mathrm{T}}=\sqrt{m^{2}+p_{\mathrm{T}}^{2}}-m is the transverse kinetic energy, nn is the high-pTp_{\mathrm{T}} power law scaling of the pTp_{\mathrm{T}} spectrum, TT is a temperature-like parameter controlling the width of the collective region, and CC is a normalization factor. We use the standard

pion mass m=mπm=m_{\pi} in the kinetic energy ETE_{\mathrm{T}} for all particles. Motivation for this choice of fit function and details about our fitting procedure will be discussed shortly.

  1. The resultant fit on the pppp spectra is then scaled by a factor ⟨TAA⟩\langle T_{\mathrm{AA}}\rangle obtained from MC Glauber estimates [2], as they would be for an RAAR_{\mathrm{AA}} calculation.

  1. The A–A data are shifted horizontally rightward toward high-pTp_{\mathrm{T}} (or equivalently, the scaled pppp fit can be translated horizontally leftward) until the shifted pppp baseline spectrum and A–A data agree as well as possible, according to the same fit metric used for the pppp spectrum fitting. This optimal horizontal shift is ΔpT\Delta p_{\mathrm{T}}.

This procedure is shown diagrammatically in Figure 2.

Figure 2. Cartoon illustrating the procedure used to determine ΔpT\Delta p_{\mathrm{T}}.

The fit methods for the pppp spectra and ΔpT\Delta p_{\mathrm{T}} were chosen carefully to best extract the relevant parameters for our analysis: namely the power law scaling of the various pTp_{\mathrm{T}} 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 pppp spectra over the full pTp_{\mathrm{T}} range at both RHIC and LHC energies [72, 73, 74, 75, 76, 77]. The function smoothly connects a thermodynamic, low-pTp_{\mathrm{T}} region expected to scale as ∼e−pT\sim e^{-p_{\mathrm{T}}} with the observed power law scaling at high pTp_{\mathrm{T}}. Since data below pTmin⁡p_{\mathrm{T}}^{\min} are excluded from the fit, in principle any function with power law asymptotics should yield comparable ΔpT\Delta p_{\mathrm{T}}, but a known effective form was chosen as a safeguard. Fitting of the pppp 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 χ2\chi^2 fit metric was reasonable but tended to be heavily biased by the first bin above the pTp_{\mathrm{T}} threshold, which can be many orders of magnitude larger than bins at higher pTp_{\mathrm{T}}, 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

MSE≡∑i(log⁡OiEi)2(8)\mathrm{MSE} \equiv\sum_i \left(\log\frac{O_i}{E_i}\right)^2 \tag*{(8)}

where OiO_i and EiE_i 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 pTp_{\mathrm{T}} spectra fitting and the ΔpT\Delta p_{\mathrm{T}} shift fitting.

In addition to the systematic uncertainty generated from the choice of pTmin⁡p_{\mathrm{T}}^{\min}, uncertainties in both pppp and A–A pTp_{\mathrm{T}} spectra are propagated to the uncertainty on ΔpT\Delta p_{\mathrm{T}} using a Monte Carlo method. Individual bins of the pTp_{\mathrm{T}} spectra are randomly varied using normal distributions with widths that matched the uncertainties, and ΔpT\Delta p_{\mathrm{T}} is recalculated using these smeared spectra. The mean of 10000 variations is reported as the final ΔpT\Delta p_{\mathrm{T}} shift, and the standard deviation becomes the propagated uncertainty. Uncertainty on ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle described in § 2 A is propagated to the ΔpT\Delta p_{\mathrm{T}} by performing the pppp scaling component of the shifting procedure with ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle replaced by ⟨Ncoll⟩−δ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle-\delta\langle N_{\mathrm{coll}}\rangle and ⟨Ncoll⟩+δ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle+\delta\langle N_{\mathrm{coll}}\rangle. The maximum difference between the default ΔpT\Delta p_{\mathrm{T}} and these altered values is taken as the uncertainty from this source. All uncertainty sources are added in quadrature to obtain the final ΔpT\Delta p_{\mathrm{T}} uncertainty.

High pTp_{\mathrm{T}} v2v_2 Estimation

As described above, the first part of this study approximates the initial energy density from the mid-rapidity hadron multiplicity dNch/dηdN_{\mathrm{ch}}/d\eta and Glauber estimates of the initial geometry and overlap area ⟨A⊥⟩\langle A_\perp\rangle. 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 η\eta [78], and the relationship between energy loss and event geometry informs the path-length dependence of medium-driven energy loss. We therefore turn to pTp_{\mathrm{T}}-differential azimuthal anisotropy.

The observed pTp_{\mathrm{T}} dependent hadron multiplicity as a function of azimuthal angle admits a Fourier series decomposition,

d2NdpTdΔiϕ=12πdNdpT[1+2∑n=1∞vn(pT)cos⁡(nΔiϕ)](9)\frac{\mathrm{d}^2N}{\mathrm{d}p_{\mathrm{T}}\mathrm{d}\Delta_i\phi} = \frac{1}{2\pi}\frac{\mathrm{d}N}{\mathrm{d}p_{\mathrm{T}}}\left[1+2\sum_{n=1}^{\infty}v_n(p_{\mathrm{T}})\cos(n\Delta_i\phi)\right] \tag*{(9)}

where the angle Δiϕ=ϕ−Ψi\Delta_i\phi=\phi-\Psi_i is oriented relative to the ithi^{\mathrm{th}}-order collision event plane, measured on an event-by-event basis. ϕ\phi is the particle’s azimuthal angle, and Ψi\Psi_i is that of the plane. Roughly speaking, the magnitude of the extracted high-pTp_{\mathrm{T}} v2v_2 has two distinct contributions at each pTp_{\mathrm{T}}: 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 ΔpT\Delta p_{\mathrm{T}} for a given centrality as described above, we can use this principle to construct a model for estimating the charged particle high pTp_{\mathrm{T}} v2(pT)v_2(p_{\mathrm{T}}) as follows:

  1. Consider a Glauber derived estimate of the path length of an event-averaged collision region as a function of azimuth R(Δϕ)R(\Delta\phi), discussed above as phenomenological area estimates. Fourier decompose this function, and take the second harmonic coefficient c2c_2. The ratio over the average radius c2/c0c_2/c_0 is the path length fraction of the maximal/minimal path against the average. Example values of c2/c0c_2/c_0 for different collision systems and path length functions R(ϕ)R(\phi) are given in Appendix § 2.1.

  1. Take two copies of the ΔpT\Delta p_{\mathrm{T}} shifted pTp_{\mathrm{T}} spectrum fit. Shift one copy up toward higher pTp_{\mathrm{T}} according to the proportion δpT=ΔpT⋅c2/c0\delta p_{\mathrm{T}} = \Delta p_{\mathrm{T}} \cdot c_2/c_0, and another copy down by the same proportion. These spectra enclose the original shifted spectrum. The upshifted and downshifted spectra are labeled dN/dΔpT+\mathrm{d}N/\mathrm{d}\Delta p_{\mathrm{T}}^{+} and dN/dΔpT−\mathrm{d}N/\mathrm{d}\Delta p_{\mathrm{T}}^{-} respectively.

  1. Our model estimate for the high-pTp_{\mathrm{T}} differential v2v_2 is then the difference weighted to the original spectrum:

v2(pT)=dNdΔpT+−dNdΔpT−4⋅dNdΔpT(10)v_2(p_{\mathrm{T}}) = \frac{\frac{\mathrm{d}N}{\mathrm{d}\Delta p_{\mathrm{T}}^{+}}-\frac{\mathrm{d}N}{\mathrm{d}\Delta p_{\mathrm{T}}^{-}}}{4\cdot\frac{\mathrm{d}N}{\mathrm{d}\Delta p_{\mathrm{T}}}} \tag*{(10)}

This estimation is similar to that used in Ref. [12] extrapolated under the assumption of uniform energy loss ΔpT\Delta p_{\mathrm{T}}, or can alternatively be thought of as a relation between the Fourier series coefficients of path length r(ϕ)r(\phi) and flow vnv_n coefficients (9).

These steps are illustrated diagrammatically in Figure 3. The v2v_2 offers a new arena to compare the classes A∪A_{\cup} and AWA_W observed in the transverse area. We will use the Full-Width-at-Half-Max contour from the width class AWA_W and the average energy radius curve from the inclusive class A∪A_{\cup}; the data for Glauber c2/c0c_2/c_0 for these methods are shown in Table V of Appendix § 2.1.

Figure 3. Cartoon illustrating the steps of the v2v_2 estimation process.

Collision SystemArea Method and ClassEdge Coef.0–5%5–10%10–20%20–30%30–40%40–50%50–60%60–70%70–80%
Pb–Pb 2.76 TeVAvg. Radius (AWA_W)c0c_03.833.553.232.852.532.252.001.771.53
Pb–Pb 2.76 TeVAvg. Radius (AWA_W)c2c_20.140.260.380.480.530.550.560.540.50
Pb–Pb 2.76 TeVAvg. Radius (AWA_W)c2/c0c_2/c_00.030.070.120.170.210.240.280.310.32
Pb–Pb 2.76 TeVFWHM Contour (AUA_U)c0c_05.174.704.133.492.962.512.151.871.49
Pb–Pb 2.76 TeVFWHM Contour (AUA_U)c2c_20.320.470.620.730.770.750.730.720.68
Pb–Pb 2.76 TeVFWHM Contour (AUA_U)c2/c0c_2/c_00.060.100.150.210.260.300.340.380.45
Pb–Pb 5.02 TeVAvg. Radius (AWA_W)c0c_03.833.553.222.842.522.241.991.751.51
Pb–Pb 5.02 TeVAvg. Radius (AWA_W)c2c_20.140.270.380.480.530.560.560.540.50
Pb–Pb 5.02 TeVAvg. Radius (AWA_W)c2/c0c_2/c_00.030.070.120.170.210.250.280.310.33
Pb–Pb 5.02 TeVFWHM Contour (AUA_U)c0c_05.174.694.123.482.942.502.141.841.45
Pb–Pb 5.02 TeVFWHM Contour (AUA_U)c2c_20.310.470.620.730.770.750.720.710.67
Pb–Pb 5.02 TeVFWHM Contour (AUA_U)c2/c0c_2/c_00.060.100.150.210.260.300.340.380.46
Xe–Xe 5.44 TeVAvg. Radius (AWA_W)c0c_03.223.022.772.482.242.021.821.621.39
Xe–Xe 5.44 TeVAvg. Radius (AWA_W)c2c_20.180.240.310.390.440.470.470.460.44
Xe–Xe 5.44 TeVAvg. Radius (AWA_W)c2/c0c_2/c_00.050.080.110.150.190.230.260.280.31
Xe–Xe 5.44 TeVFWHM Contour (AUA_U)c0c_04.133.773.322.852.472.171.941.621.20
Xe–Xe 5.44 TeVFWHM Contour (AUA_U)c2c_20.340.400.500.570.590.610.650.670.59
Xe–Xe 5.44 TeVFWHM Contour (AUA_U)c2/c0c_2/c_00.080.100.150.200.240.280.330.410.49
Au–Au 0.20 TeVAvg. Radius (AWA_W)c0c_03.703.453.152.802.512.252.021.801.58
Au–Au 0.20 TeVAvg. Radius (AWA_W)c2c_20.160.260.360.450.500.530.540.530.50
Au–Au 0.20 TeVAvg. Radius (AWA_W)c2/c0c_2/c_00.040.070.110.160.200.230.260.290.31
Au–Au 0.20 TeVFWHM Contour (AUA_U)c0c_04.954.513.993.402.912.512.181.921.57
Au–Au 0.20 TeVFWHM Contour (AUA_U)c2c_20.330.450.580.690.730.720.700.700.67
Au–Au 0.20 TeVFWHM Contour (AUA_U)c2/c0c_2/c_00.060.100.140.200.250.280.320.360.42
Cu–Cu 0.20 TeVAvg. Radius (AWA_W)c0c_02.632.502.332.131.961.801.641.471.29
Cu–Cu 0.20 TeVAvg. Radius (AWA_W)c2c_20.180.220.270.330.380.400.400.390.39
Cu–Cu 0.20 TeVAvg. Radius (AWA_W)c2/c0c_2/c_00.070.090.110.150.190.220.240.260.30
Cu–Cu 0.20 TeVFWHM Contour (AUA_U)c0c_03.243.002.712.392.141.941.711.411.11
Cu–Cu 0.20 TeVFWHM Contour (AUA_U)c2c_20.350.390.420.470.520.580.630.640.54
Cu–Cu 0.20 TeVFWHM Contour (AUA_U)c2/c0c_2/c_00.100.130.150.190.240.300.360.450.49

Table V. Tabulated values for averaged event harmonics c2c_2, c0c_0 from the AWA_W and AUA_U areas generated from Glauber simulations for the collision systems studied. Note that the Average Radius method represents the AWA_W class, while the Energy Half-Max Contour represents in the AUA_U class.

Note that this model contains an assumption of linear path-length dependent energy loss ΔpT(L)∝L\Delta p_{\mathrm{T}}(L) \propto L to assert that the path length proportion c2/c0c_2/c_0 can be translated directly to an energy loss for extra path-length δpT=ΔpT⋅c2/c0\delta p_{\mathrm{T}} = \Delta p_{\mathrm{T}} \cdot c_2/c_0. 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, εBjwidth\varepsilon_{\mathrm{Bj}}^{\mathrm{width}} and εBjinclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{inclusive}}, respectively, as functions of centrality and ⟨Npart⟩\langle N_{\mathrm{part}}\rangle for a variety of collision energies and species. Sensibly, εBj\varepsilon_{\mathrm{Bj}} increases monotonically with increasing ⟨Npart⟩\langle N_{\mathrm{part}}\rangle or sNN\sqrt{s_{\mathrm{NN}}} for both εBjwidth\varepsilon_{\mathrm{Bj}}^{\mathrm{width}} and εBjinclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{inclusive}}, but this does not generally hold for εBjexclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{exclusive}}, as will be discussed later. While the energy densities are similar for the most peripheral data, the width based εBjwidth\varepsilon_{\mathrm{Bj}}^{\mathrm{width}} rises more steeply with centrality, resulting in an approximately 50% larger energy density estimate for the most central Pb–Pb data relative to εBjinclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{inclusive}}.

Following the procedure in Sec. IIC, illustrative ΔpT\Delta p_{\mathrm{T}} results from ALICE Pb–Pb at sNN=5.02\sqrt{s_{\mathrm{NN}}}=5.02 TeV as well as PHENIX and STAR Au–Au at sNN=200\sqrt{s_{\mathrm{NN}}}=200 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 pTp_{\mathrm{T}} spectra for three sample A–A centrality bins alongside the corresponding appropriately ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle-scaled pppp inelastic data. The curves are the ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle-scaled Tsallis fits to the pppp reference spectra. The fit parameters and MSE metric for all resultant pppp data fits are provided in Table I. The suppression of the A–A spectra with respect to the ⟨TAA⟩\langle T_{\mathrm{AA}}\rangle-scaled Tsallis pppp fit is evident in all cases, in agreement with published RAAR_{\mathrm{AA}} values [53, 54, 55, 56, 57, 58]. The remaining three panels (b.–d.) show the ΔpT\Delta p_{\mathrm{T}}-shifted A–A data for the three sample centrality bins alongside the Tsallis pppp fit curve. The A–A spectra have been shifted by the ΔpT\Delta p_{\mathrm{T}} which best aligns the two and minimizes the MSE metric above the determined momentum threshold pTmin⁡p_{\mathrm{T}}^{\min} denoted by the vertical line. Note that the pTmin⁡p_{\mathrm{T}}^{\min} line has, along with the A–A data points, been shifted rightward by ΔpT\Delta p_{\mathrm{T}} in panels (b.–d.). The Tsallis pppp fit curve is only shown in the relevant region above this pTmin⁡p_{\mathrm{T}}^{\min} threshold. The strong visual agreement between the shifted A–A spectra and ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle-scaled pppp 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 ΔpT\Delta p_{\mathrm{T}} is determined for each centrality bin. Note that the final MSE metric for each ΔpT\Delta p_{\mathrm{T}} 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 ΔpT\Delta p_{\mathrm{T}}-shifted Tsallis fits were substituted for the A–A data, and ratios against the corresponding pppp baseline spectra were taken for comparison against published RAAR_{\mathrm{AA}} data. Rough agreement of

Figure 6. pTp_T spectra of charged particles measured by the ALICE collaboration in Pb–Pb collisions at sNN=5.02\sqrt{s_{NN}} = 5.02 TeV for different centralities. The first panel shows the pppp Tsallis fit appropriately ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle-scaled to each Pb–Pb dataset. The remaining three panels show the ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle-scaled pppp Tsallis fit compared to the individual ΔpT\Delta p_T shifted Pb–Pb pTp_T spectrum in the region beyond the pTp_T shift lower bound. Uncertainty bars are present, but small relative to marker sizes.

Figure 7. pTp_T spectra of charged particles measured by the PHENIX collaboration in Au–Au collisions at sNN=200\sqrt{s_{NN}} = 200 GeV for different centralities. The first panel shows the pppp Tsallis fit appropriately ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle-scaled to each Au–Au dataset. The remaining three panels show the ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle-scaled pppp Tsallis fit compared to the individual ΔpT\Delta p_T shifted Au–Au pTp_T spectrum in the region beyond the pTp_T shift lower bound.

Figure 8. pTp_T spectra of charged particles measured by the STAR collaboration in Au–Au collisions at sNN=200\sqrt{s_{NN}} = 200 GeV for different centralities. The first panel shows the pppp Tsallis fit appropriately ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle-scaled to each Au–Au dataset. The remaining three panels show the ⟨Ncoll⟩\langle N_{\mathrm{coll}}\rangle-scaled pppp Tsallis fit compared to the individual ΔpT\Delta p_T shifted Au–Au pTp_T spectrum in the region beyond the pTp_T shift lower bound.

Experiment (s\sqrt{s})CC (c/GeV)nnTT (GeV)Reduced MSE Metric
ALICE (5.44 TeV)18.5405.5660.2140.002423
ATLAS (5.44 TeV)18.5405.5550.2090.001209
ALICE (5.02 TeV)9.8255.6840.2470.003396
ALICE (2.76 TeV)10.7205.9700.2270.004184
STAR (0.2 TeV)16.5788.1650.1540.01000
PHENIX (0.2 TeV)16.5789.1080.1690.00155
STAR (0.2 TeV, π0\pi^0)95.2668.1690.1120.000677
Centrality Bin (%)STAR pppp reference sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeVSTAR pppp reference sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeVPHENIX pppp reference sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeVPHENIX pppp reference sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV
STAR Au–Au ΔpT\Delta p_{\mathrm{T}} (GeV/cc)PHENIX Au–Au ΔpT\Delta p_{\mathrm{T}} (GeV/cc)STAR Au–Au ΔpT\Delta p_{\mathrm{T}} (GeV/cc)PHENIX Au–Au ΔpT\Delta p_{\mathrm{T}} (GeV/cc)
0–5%1.77 ± 0.231.55 ± 0.221.38 ± 0.131.27 ± 0.21
10–20%1.43 ± 0.221.40 ± 0.221.07 ± 0.161.12 ± 0.21
20–30%1.18 ± 0.201.04 ± 0.210.85 ± 0.150.81 ± 0.20
30–40%0.92 ± 0.200.62 ± 0.090.62 ± 0.170.63 ± 0.12
40–50%–0.72 ± 0.13–0.56 ± 0.12

Table 1. TABLE I. pppp cross section pTp_{\mathrm{T}} spectra fit parameters (CC, nn, TT) and the reduced MSE/DoF metric for the Tsallis fit to each dataset used in the analysis.

Figure 4. Energy density using AUA_U for a variety of collision species and beam energies, as a function of ⟨Npart⟩\langle N_{\mathrm{part}}\rangle (upper) and centrality (lower).

Figure 5. Energy density using AWA_W for a variety of collision species and beam energies, as a function of ⟨Npart⟩\langle N_{\mathrm{part}}\rangle (upper) and centrality (lower).

this model ratio with RAAR_{\mathrm{AA}} was observed within 2σ2\sigma for all collision systems studied, though it should be noted that the model deviates more strongly from RAAR_{\mathrm{AA}} in central collisions.

It should be noted that the Tsallis power law parameters nn fit to STAR and PHENIX pppp data at s=200\sqrt{s}=200 GeV differ significantly, while all fit parameters for the ALICE and ATLAS data at s=5.44\sqrt{s}=5.44 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 pTp_{\mathrm{T}} spectra by extrapolating the π0\pi^{0} spectrum from Ref. [79]. STAR, on the other hand, determined the pppp charged particle pTp_{\mathrm{T}} spectra by summing measured pTp_{\mathrm{T}}-differential π±\pi^{\pm}, K±K^{\pm}, pp and pˉ\bar{p} yields [59]. The STAR determination results in a more significant flattening of the higher pTp_{\mathrm{T}} data compared to the PHENIX extrapolation, reflected in the smaller power law nn.

Following the discrepancy in pppp reference spectra, the ΔpT\Delta p_{\mathrm{T}} values calculated for Au–Au collisions at sNN=200\sqrt{s_{\mathrm{NN}}}=200 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 ΔpT\Delta p_{\mathrm{T}} values across all centrality bins. To examine whether this discrepancy in ΔpT\Delta p_{\mathrm{T}} is purely due to the pppp spectra differences observed in the fits, we compared ΔpT\Delta p_{\mathrm{T}} values obtained from STAR and PHENIX Au–Au spectra with a common fixed pppp refer-

ence spectrum. Table II shows the ΔpT\Delta p_{\mathrm{T}} calculated when either the STAR or PHENIX pppp charged particle spectrum is used as the common pppp reference. In both cases, the computed ΔpT\Delta p_{\mathrm{T}} values for each collaboration’s Au–Au data agree within uncertainties. This affirms that the observed differences in ΔpT\Delta p_{\mathrm{T}} is dominantly an artifact of the differences in pppp spectral shape between collaborations. For consistency, however, our analysis proceeds using the pppp spectrum reported by each collaboration to calculate the ΔpT\Delta p_{\mathrm{T}} for their respective A–A data in our final results.

With this understanding of ΔpT\Delta p_{\mathrm{T}} and εBj\varepsilon_{\mathrm{Bj}} in hand, we can now move on to the central objective of this work: characterization of the relationship between these two quantities. The resulting ΔpT\Delta p_{\mathrm{T}} 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-pTp_{\mathrm{T}} charged particle ΔpT\Delta p_{\mathrm{T}} with the estimated initial energy density εBj\varepsilon_{\mathrm{Bj}} is observed, from peripheral Cu–Cu collisions at sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV to central Pb–Pb events at sNN=5.02\sqrt{s_{\mathrm{NN}}}=5.02 TeV. To evaluate whether these correlations are linear to leading order, a χ2\chi^2 fit to a variable power ΔpT=A(εBj/ε0)p+b\Delta p_{\mathrm{T}}=A(\varepsilon_{\mathrm{Bj}}/\varepsilon_0)^p+b is performed, from which the resulting powers are pinclusive=1.006−0.112+0.115p_{\mathrm{inclusive}}=1.006^{+0.115}_{-0.112} and pwidth=1.016−0.080+0.083p_{\mathrm{width}}=1.016^{+0.083}_{-0.080}. These correspond to χ2/DoF\chi^2/\mathrm{DoF} 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 εBj\varepsilon_{\mathrm{Bj}} versus ΔpT\Delta p_{\mathrm{T}} for a variety of collision species and beam energies. Estimates of energy density from each area scaling class AUA_{\mathrm{U}}, AWA_{\mathrm{W}}, and AΠA_{\Pi} are shown separately. Note the STAR Cu–Cu data are reported π±\pi^{\pm} spectra for calculating ΔpT\Delta p_{\mathrm{T}} while other datasets use inclusive charged particle spectra.

Performing linear regressions, a slope of 0.054 ±\pm 0.002 fm3^3/c and 0.099 ±\pm 0.003 fm3^3/c for the width and inclusive area estimates respectively. The individual linear fits for each experiment-collision pair, as well as the corresponding reduced χ2\chi^2 values are provided in Table III.

ExperimentBeam Species, sNN\sqrt{s_{NN}}εBjwidth\varepsilon_{\mathrm{Bj}}^{\mathrm{width}} Intercept (GeV/c)εBjwidth\varepsilon_{\mathrm{Bj}}^{\mathrm{width}} slope (fm3^3/c)εBjinclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{inclusive}} Intercept (GeV/c)εBjinclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{inclusive}} slope (fm3^3/c)Reduced χwidth2\chi_{\mathrm{width}}^2Reduced χinclusive2\chi_{\mathrm{inclusive}}^2
ALICEXe–Xe, 5.44 TeV−0.164±0.069-0.164 \pm0.0690.057±0.0020.057 \pm0.002−0.689±0.121-0.689 \pm0.1210.106±0.0060.106 \pm0.0060.500.500.960.96
ATLASXe–Xe, 5.44 TeV−0.170±0.049-0.170 \pm0.0490.054±0.0020.054 \pm0.002−0.669±0.086-0.669 \pm0.0860.100±0.0040.100 \pm0.0040.170.170.330.33
ALICEPb–Pb, 5.02 TeV0.295±0.0380.295 \pm0.0380.049±0.0010.049 \pm0.001−0.095±0.025-0.095 \pm0.0250.090±0.0010.090 \pm0.0010.230.230.060.06
ALICEPb–Pb, 2.76 TeV0.466±0.0670.466 \pm0.0670.049±0.0020.049 \pm0.0020.144±0.0810.144 \pm0.0810.086±0.0040.086 \pm0.0040.380.380.360.36
STARAu–Au, 200 GeV0.027±0.1030.027 \pm0.1030.074±0.0060.074 \pm0.006−0.489±0.188-0.489 \pm0.1880.124±0.0130.124 \pm0.0130.070.070.120.12
PHENIXAu–Au, 200 GeV0.035±0.1100.035 \pm0.1100.055±0.0070.055 \pm0.007−0.410±0.203-0.410 \pm0.2030.096±0.0150.096 \pm0.0150.340.340.500.50
STARCu–Cu, 200 GeV−0.131±0.078-0.131 \pm0.0780.063±0.0090.063 \pm0.009−0.603±0.142-0.603 \pm0.1420.101±0.0140.101 \pm0.0140.170.170.160.16
Global0.047±0.0460.047 \pm0.0460.054±0.0020.054 \pm0.002−0.429±0.066-0.429 \pm0.0660.099±0.0030.099 \pm0.0031.471.471.781.78

Table III. Linear fit results for ΔpT\Delta p_T as a function of εBjwidth\varepsilon_{\mathrm{Bj}}^{\mathrm{width}} and εBjinclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{inclusive}}. Fitting is performed using orthogonal distance regression to incorporate uncertainties in dependent and independent axes. Reduced χ2\chi^2 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 pTp_{\mathrm{T}} spectra. It also appears that the manner of energy density generation is of little consequence; collisions of low AA species at high energy or high AA species at low energy result in the same ΔpT\Delta p_{\mathrm{T}}, 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 δ(ΔpT)\delta(\Delta p_{\mathrm{T}}) to the linear fits shown in Figure 9 are studied as a function of effective path length LeffL_{\mathrm{eff}} and path difference ΔLeff\Delta L_{\mathrm{eff}} as well as ⟨Npart⟩\langle N_{\mathrm{part}}\rangle in Appendix C.

Although AΠA_{\Pi} presents itself as an outlier in Figure 1, we include the relationship between ΔpT\Delta p_{\mathrm{T}} and εBjexclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{exclusive}} in Figure 9 for completeness. We see that for collisions at RHIC energies, the calculated energy density using AΠA_{\Pi} is approximately constant for all data points, indicating that the ratio of dNch/dηdN_{\mathrm{ch}}/d\eta and AΠA_{\Pi} does not vary over centrality or species. In the specific case of STAR Cu–Cu, εBjexclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{exclusive}} slightly decreases with increasing impact parameter, suggesting nonphysically that peripheral collisions produce larger energy density than central collisions at

the same sNN\sqrt{s_{\mathrm{NN}}}. For LHC data however, εBjexclusive\varepsilon_{\mathrm{Bj}}^{\mathrm{exclusive}} does rise with centrality and ΔpT\Delta p_{\mathrm{T}} roughly increases with energy density, but these behaviors are different for each collision system. While in principle A∩A_{\cap} 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 εBj\varepsilon_{\mathrm{Bj}}.

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 sNN=5.02\sqrt{s_{\mathrm{NN}}}=5.02 TeV (a.–c.), Xe–Xe at sNN=5.44\sqrt{s_{\mathrm{NN}}}=5.44 TeV (d.–f.), Au–Au at sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV (g.–i.), and Cu–Cu at sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV (j.–l.). Each column contains the explicit transverse areas ⟨A⊥⟩\langle A_{\perp}\rangle produced by each method (a., d., g., j.), ratios against the EBE inclusive A∪A_{\cup} (b., e., h., k.) and ratios against the EBE width area AWA_{W} (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 v2v_{2} results, Figure 10 shows the prediction for sNN=2.76\sqrt{s_{\mathrm{NN}}}=2.76 TeV Pb–Pb high-pTp_{\mathrm{T}} v2v_{2} using the procedure described in § 2 D. Model predictions for the AWA_{W} and A∪A_{\cup} classes are compared to corresponding ALICE v2v_{2} data [80] for three different centrality bins. Predictions for both area estimates are similar in shape but differ slightly in normalization, with the AWA_{W} class favored by central data and A∪A_{\cup} performing better for peripheral collisions. Considering the simplicity of the modeling, a reasonable agreement is obtained across the available pTp_{\mathrm{T}} range, within 3σ3\sigma for all data for AWA_{W}-based v2v_{2}. Stronger precision on high-pTp_{\mathrm{T}} v2v_{2} 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-pTp_{\mathrm{T}} v2v_{2} estimations derived from ΔpT\Delta p_{\mathrm{T}} as well as the avg. energy radius/avg. pressure radius area definitions, corresponding to the AWA_{W} class (solid blue curve) and the FWHM contour area definition, corresponding to the A∪A_{\cup} area class (dashed orange curve), compared to ALICE experimental data from Pb–Pb collisions at sNN=2.76\sqrt{s_{\mathrm{NN}}}=2.76 TeV.

Encouraged by the general agreement with the magnitude of v2v_{2} across the pTp_{\mathrm{T}} 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 c2c_{2}, c0c_{0} derived from the width-based area estimates, though the corresponding v2v_{2} for A∪A_{\cup} are shown in Appendix D. The v2v_{2} for the Au–Au data at sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV is the lowest at a fixed pTp_{\mathrm{T}} and centrality; although the c2/c0c_{2}/c_{0} ratios in Au–Au are similar to those for the LHC Pb–Pb and Xe–Xe, the extracted ΔpT\Delta p_{\mathrm{T}} are significantly reduced due to the smaller sNN\sqrt{s_{\mathrm{NN}}}, resulting in smaller v2v_{2} in our modeling.

Figure 11. High pTp_T v2v_2 predictions using the ΔpT\Delta p_T and AWA_W based model for a variety of collision systems.

The trends of the Xe–Xe predictions are especially interesting, as the sNN=5.44\sqrt{s_{\mathrm{NN}}}=5.44 TeV Xe–Xe contains the largest v2v_{2} among all species for the most central data, which falls to the smallest LHC v2v_{2} in peripheral collisions. A similar flipping of the Xe–Xe relative to sNN=5.02\sqrt{s_{\mathrm{NN}}}=5.02 TeV Pb–Pb is also predicted by hydrodynamic calculations [81] and observed in data [82, 83], albeit at significantly lower pTp_{\mathrm{T}}. The enhancement of low-pTp_{\mathrm{T}} v2v_{2} in central Xe–Xe is attributed primarily to nuclear shape, as the quadrupole deformation β2\beta_{2} is expected to be much larger in 129Xe{}^{129}\mathrm{Xe} than 208Pb{}^{208}\mathrm{Pb} [84]. The nuclear size also plays a role; systematic analyses of vnv_{n} observables have shown that statistical fluctuations in yields produce a universal enhancement of all vnv_{n} in systems with fewer sources. The enhancement of v3v_{3} in Xe–Xe relative to Pb–Pb across the centrality range is considered a signal of this effect [82].

The depletion of v2v_{2} for peripheral sNN=5.44\sqrt{s_{\mathrm{NN}}}=5.44 TeV Xe–Xe collisions relative to sNN=5.02\sqrt{s_{\mathrm{NN}}}=5.02 TeV Pb–Pb is therefore unexpected on the basis of nuclear structure alone, as models indicate that the nuclear shape of 129Xe{}^{129}\mathrm{Xe} 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-pTp_{\mathrm{T}} measurements by ALICE [82] and ATLAS [83], the depletion is attributed to nonzero shear viscosity η/s\eta/s during hydrodynamic evolution; hydrodynamic calculations [81] reproduce the depletion only for nonzero η/s\eta/s, though calculations are similar for a range of η/s\eta/s [82]. In the absence of significant hydrodynamic effects for high-pTp_{\mathrm{T}} hadrons, the depletion is instead attributed to path-length dependent jet quenching [1]. While the pTp_{\mathrm{T}} 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 εBj\varepsilon_{\rm Bj} created in a heavy-ion collision and the average energy loss ΔpT\Delta p_T of high-pTp_T 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 sNN=200\sqrt{s_{\rm NN}} = 200 GeV to Pb–Pb and Xe–Xe events at sNN>5\sqrt{s_{\rm NN}} > 5 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 RAAR_{\rm AA}.

A variety of methods were studied to extract the transverse overlap area using Monte-Carlo Glauber calculations, needed to compute εBj\varepsilon_{\rm Bj}. 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 A∩A_{\cap} calculation being the sole outlier. The strange scaling for A∩A_{\cap} 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” A∪A_{\cup} area, while the other class scaled with the “width-based” AWA_W EBE calculation. This classification of scaling persisted across all species and energies studied. Generally, the AWA_W class displayed a stronger dependence on centrality than the A∪A_{\cup} class, especially for more peripheral events. The εBj\varepsilon_{\rm Bj} estimated using ⟨A⊥⟩\langle A_{\perp}\rangle from either AWA_W or A∪A_{\cup} classes and the reported experimental dNch/dη\mathrm{d}N_{\rm ch}/\mathrm{d}\eta show the expected trends with centrality, ⟨Npart⟩\langle N_{\rm part}\rangle, and collision energy, while the same is not true for the outlier A∩A_{\cap}. Because of this, A∩A_{\cap} does not produce the strong ΔpT\Delta p_T–εBj\varepsilon_{\rm Bj} correlation observed for the other seven methods. The persistence of the correlation between both the A∪A_{\cup} and AWA_W area classes, with no change in ΔpT\Delta p_T modeling, supports that the correlation is robust and not due to Glauber model self-correlation between ⟨A⊥⟩\langle A_{\perp}\rangle inputs to εBj\varepsilon_{\rm Bj} and ⟨Ncoll⟩\langle N_{\rm coll}\rangle scaling of ΔpT\Delta p_T fit spectra. That the correlation then fails for A∩A_{\cap} demonstrates that the correlation is not a trivial consequence of the model construction. In addition, the ⟨Ncoll⟩\langle N_{\rm coll}\rangle uncertainty estimation procedure, applied only to the computation of ΔpT\Delta p_T, does not significantly increase uncertainty or blur the correlation.

Good descriptions of the A–A high-pTp_T spectra can be obtained for all systems, collision energy, and centralities studied via ΔpT\Delta p_T shifting of the appropriately ⟨Ncoll⟩\langle N_{\rm coll}\rangle-scaled Tsallis function fits to the pppp pTp_T spectra. For simplicity a constant pTp_T shift is assumed for each spectrum. While theoretical arguments suggest a fractional energy loss with parton type and pTp_T dependence, the limited high-pTp_T charged particle range currently available combined with the smearing in pTp_T between parton initiator and final-state hadron appears to make this constant ΔpT\Delta p_T 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 pTp_T spectrum modifications from kinematic ones.

We extended our model to explore predictions of high-pTp_T hadronic anisotropy v2v_2 by coupling Glauber estimates of event geometry to observed pTp_T spectra in a novel approach. The model provides estimates of v2v_2 that are reasonably consistent with ALICE data [80] in the highest pTp_{\mathrm{T}} data of each centrality bin, but overestimates v2v_2 for lower pTp_{\mathrm{T}} data. As the breakdown region of the model is still well above the collective region, the deviation of v2v_2 from our model may indicate that our assumption of linear path length-dependent energy loss is ineffective for this mid-pTp_{\mathrm{T}} region. We also observe a flipping of the magnitude of the Xe–Xe v2v_2 at sNN=5.44\sqrt{s_{\mathrm{NN}}}=5.44 TeV relative to that in Pb–Pb at sNN=5.02\sqrt{s_{\mathrm{NN}}}=5.02 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 pTp_{\mathrm{T}} spectra.

The interrelations between ΔpT\Delta p_{\mathrm{T}}, εBj\varepsilon_{\mathrm{Bj}}, v2v_2 and related quantities merit further research. While the exclusive area A∩A_{\cap} 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 ⟨A⊥⟩\langle A_{\perp}\rangle calculation exists, or what might separate these two classes. While our pTp_{\mathrm{T}}-independent ΔpT\Delta p_{\mathrm{T}} model serves as a reasonable approximation for interpreting high-pTp_{\mathrm{T}} 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 pTp_{\mathrm{T}}-differential jet v2v_2 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 pppp pTp_{\mathrm{T}} spectra fitting and fit functions. We gratefully acknowledge Christine Nattrass and Brant Johnson for their help in obtaining PHENIX pppp reference yields for analysis in this study. RH and HC are supported by DoE grant DE-SC004168.

Additional Information Concerning Energy Density, ΔpT\Delta p_{\mathrm{T}} and Event Averaged Harmonic Calculations

This appendix provides further details on the area calculations and other input data used to compute the energy density εBj\varepsilon_{\mathrm{Bj}} and the averaged event harmonics c2c_2, c0c_0 used in the elliptic flow v2v_2 predictions.

Figure 12 provides evidence that the categorization of area definitions into the AWA_{\mathrm{W}} and AUA_{\mathrm{U}} classes holds across collision systems studied here. The figure shows the Glauber area calculations for Pb–Pb collisions at sNN=5.02\sqrt{s_{\mathrm{NN}}} = 5.02 TeV, Xe–Xe at sNN=5.44\sqrt{s_{\mathrm{NN}}} = 5.44 TeV, and Au–Au and Cu–Cu at sNN=200\sqrt{s_{\mathrm{NN}}} = 200 GeV as a function of centrality. In addition, the ratios against the EBE inclusive (AUA_{\mathrm{U}}) and EBE width-based (AWA_{\mathrm{W}}) 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 (A∩A_{\cap}). These observations coincide with the same classifications in Pb–Pb at sNN=2.76\sqrt{s_{\mathrm{NN}}} = 2.76 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 εBj\varepsilon_{\mathrm{Bj}}: the Jacobian to translate from dNch/dη\mathrm{d}N_{\mathrm{ch}}/\mathrm{d}\eta to dNch/dy\mathrm{d}N_{\mathrm{ch}}/\mathrm{d}y, and the transverse areas AWA_{\mathrm{W}} and AUA_{\mathrm{U}}. Also included are the MSE/DoF fit metric results for the ΔpT\Delta p_{\mathrm{T}} shifted spectra, as measured against the ⟨TAA⟩\langle T_{\mathrm{AA}}\rangle-scaled Tsallis pppp reference. Note that for ATLAS Xe–Xe [54] and PHENIX Au–Au [57] the mid-rapidity dNch/dη\mathrm{d}N_{\mathrm{ch}}/\mathrm{d}\eta from ALICE [53] and STAR [58] respectively were used for εBj\varepsilon_{\mathrm{Bj}} computation, as published ATLAS/PHENIX dNch/dη\mathrm{d}N_{\mathrm{ch}}/\mathrm{d}\eta data in the same centrality binning were not available.

Centrality Bin (%)dNch/dηdN_{\mathrm{ch}}/d\etaTransverse Area Width AWA_WTransverse Area Inclusive A∪A_{\cup}Energy Density WidthEnergy Density InclusiveALICE / STAR MSE/DoF, Eq. (8)ATLAS / PHENIX MSE/DoF, Eq. (8)
Pb–Pb at sNN=2.76\sqrt{s_{\mathrm{NN}}}=2.76 TeV. Jacobian J=1.09J=1.09 from [69], dNch/dηdN_{\mathrm{ch}}/d\eta from [70].ALICE [55]
0.0–5.01601±601601 \pm60107.56±7.04107.56 \pm7.04149.18±7.51149.18 \pm7.5171.80±4.6071.80 \pm4.6046.42±2.9746.42 \pm2.970.033
5.0–10.01294±491294 \pm4993.21±7.0493.21 \pm7.04131.96±8.26131.96 \pm8.2665.42±4.2165.42 \pm4.2141.16±2.6541.16 \pm2.650.028
10.0–20.0966±37966 \pm3777.34±8.4277.34 \pm8.42109.77±10.61109.77 \pm10.6156.83±3.6956.83 \pm3.6935.63±2.3135.63 \pm2.310.021
20.0–30.0649±23649 \pm2361.24±7.9261.24 \pm7.9285.57±9.4685.57 \pm9.4645.65±2.8345.65 \pm2.8329.22±1.8129.22 \pm1.810.013
30.0–40.0426±15426 \pm1548.98±7.9248.98 \pm7.9265.91±8.6565.91 \pm8.6535.07±2.1635.07 \pm2.1623.61±1.4623.61 \pm1.460.013
40.0–50.0261±9261 \pm939.15±8.1739.15 \pm8.1749.34±7.9849.34 \pm7.9824.60±1.5024.60 \pm1.5018.08±1.1018.08 \pm1.100.014
50.0–60.0149±6149 \pm630.98±8.4230.98 \pm8.4235.27±7.4035.27 \pm7.4015.92±1.0715.92 \pm1.0713.39±0.9013.39 \pm0.900.023
60.0–70.076±476 \pm423.69±8.4223.69 \pm8.4223.34±6.8223.34 \pm6.829.28±0.759.28 \pm0.759.47±0.769.47 \pm0.760.009
70.0–80.032±232 \pm217.03±7.6717.03 \pm7.6713.91±5.8313.91 \pm5.834.55±0.424.55 \pm0.425.95±0.555.95 \pm0.550.025
Pb–Pb at sNN=5.02\sqrt{s_{\mathrm{NN}}}=5.02 TeV. Jacobian J=1.09J=1.09 from [69], dNch/dηdN_{\mathrm{ch}}/d\eta from [55].ALICE [55]
0.0–5.01910±491910 \pm49107.47±7.16107.47 \pm7.16153.41±7.42153.41 \pm7.4291.11±4.8091.11 \pm4.8056.68±2.9856.68 \pm2.980.058
5.0–10.01547±401547 \pm4093.04±7.0493.04 \pm7.04136.45±8.45136.45 \pm8.4583.37±4.4083.37 \pm4.4050.03±2.6450.03 \pm2.640.046
10.0–20.01180±31.61180 \pm31.677.07±8.4277.07 \pm8.42114.01±10.87114.01 \pm10.8774.68±4.0074.68 \pm4.0044.30±2.3844.30 \pm2.380.032
20.0–30.0786±20.8786 \pm20.860.95±7.9260.95 \pm7.9289.40±9.8089.40 \pm9.8059.41±3.1759.41 \pm3.1735.64±1.9035.64 \pm1.900.021
30.0–40.0512±15.5512 \pm15.548.62±7.9248.62 \pm7.9269.22±8.9769.22 \pm8.9745.34±2.5845.34 \pm2.5828.31±1.6128.31 \pm1.610.014
40.0–50.0318±12.5318 \pm12.538.78±8.1738.78 \pm8.1752.20±8.3752.20 \pm8.3732.49±2.1432.49 \pm2.1421.85±1.4421.85 \pm1.440.006
50.0–60.0183±8.2183 \pm8.230.54±8.4230.54 \pm8.4237.57±7.7937.57 \pm7.7921.38±1.5421.38 \pm1.5416.22±1.1716.22 \pm1.170.004
60.0–70.096±5.996 \pm5.923.24±8.4223.24 \pm8.4225.12±7.2125.12 \pm7.2113.02±1.1913.02 \pm1.1911.74±1.0711.74 \pm1.070.014
70.0–80.045±3.545 \pm3.516.57±7.5416.57 \pm7.5415.12±6.2515.12 \pm6.257.44±0.837.44 \pm0.838.41±0.938.41 \pm0.930.010
Xe–Xe at sNN=5.44\sqrt{s_{\mathrm{NN}}}=5.44 TeV. Jacobian J=1.09J=1.09 from [69], dNch/dηdN_{\mathrm{ch}}/d\eta from [53].ALICE [53]ATLAS [54]
0.0–5.01167±261167 \pm2677.47±6.1677.47 \pm6.16110.87±6.77110.87 \pm6.7773.07±3.6473.07 \pm3.6445.31±2.2645.31 \pm2.260.0320.044
5.0–10.0939±24939 \pm2468.33±6.4168.33 \pm6.4199.32±7.4299.32 \pm7.4264.65±3.4064.65 \pm3.4039.27±2.0639.27 \pm2.060.0540.059
10.0–20.0706±17706 \pm1757.72±7.5457.72 \pm7.5483.49±9.0583.49 \pm9.0555.35±2.8455.35 \pm2.8433.84±1.7433.84 \pm1.740.0370.027
20.0–30.0478±11478 \pm1146.78±7.7946.78 \pm7.7965.66±8.5265.66 \pm8.5243.56±2.2043.56 \pm2.2027.71±1.4027.71 \pm1.400.0300.013
30.0–40.0315±8315 \pm838.16±8.1738.16 \pm8.1750.81±8.1450.81 \pm8.1432.77±1.7232.77 \pm1.7222.37±1.1722.37 \pm1.170.0180.007
40.0–50.0198±5198 \pm530.96±8.5530.96 \pm8.5538.11±7.8038.11 \pm7.8023.32±1.2223.32 \pm1.2217.68±0.9217.68 \pm0.920.0400.007
50.0–60.0118±3118 \pm324.55±8.8024.55 \pm8.8027.26±7.4727.26 \pm7.4715.93±0.8415.93 \pm0.8413.86±0.7313.86 \pm0.730.0190.005
60.0–70.064.7±264.7 \pm218.65±8.2918.65 \pm8.2918.28±6.8618.28 \pm6.8610.32±0.5910.32 \pm0.5910.60±0.6110.60 \pm0.610.019-
70.0–80.032±1.332 \pm1.313.48±7.1613.48 \pm7.1611.57±5.6411.57 \pm5.646.22±0.426.22 \pm0.427.62±0.517.62 \pm0.510.091-
60.0–80.048±1.6548 \pm1.6516.07±8.5516.07 \pm8.5514.92±7.1214.92 \pm7.128.47±0.518.47 \pm0.519.35±0.569.35 \pm0.56-0.004
Au–Au at sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV. Jacobian J=1.25J=1.25 from [68], dNch/dηdN_{\mathrm{ch}}/d\eta from [71].STAR [56, 59]PHENIX [57]
0.0–5.0625±55625 \pm55101.84±7.04101.84 \pm7.04122.34±7.31122.34 \pm7.3124.37±3.0224.37 \pm3.0219.08±2.3619.08 \pm2.360.1090.237
5.0–10.0501±44501 \pm4488.78±6.7988.78 \pm6.79106.01±7.13106.01 \pm7.1321.77±2.6921.77 \pm2.6917.19±2.1317.19 \pm2.130.171-
0.0–10.0563±55563 \pm5595.31±7.6795.31 \pm7.67114.17±8.32114.17 \pm8.3223.14±2.5823.14 \pm2.5818.18±2.0318.18 \pm2.03-0.129
10.0–20.0377±33377 \pm3374.16±8.0474.16 \pm8.0486.56±8.9186.56 \pm8.9118.94±2.3418.94 \pm2.3415.41±1.9015.41 \pm1.900.0980.058
20.0–30.0257±23257 \pm2359.28±7.5459.28 \pm7.5465.97±7.7965.97 \pm7.7915.32±1.9315.32 \pm1.9313.28±1.6713.28 \pm1.670.0860.047
30.0–40.0174±16174 \pm1647.83±7.6747.83 \pm7.6749.51±7.0849.51 \pm7.0812.12±1.5612.12 \pm1.5611.58±1.4911.58 \pm1.490.1400.008
Cu–Cu at sNN=200\sqrt{s_{\mathrm{NN}}}=200 GeV. Jacobian J=1.25J=1.25 from [68], dNch/dηdN_{\mathrm{ch}}/d\eta from [58].STAR [58]
0.0–10.0176.3±12.7176.3 \pm12.749.57±6.1649.57 \pm6.1648.69±5.2048.69 \pm5.2011.76±1.2211.76 \pm1.2212.05±1.2512.05 \pm1.250.010
10.0–20.0121.5±8.7121.5 \pm8.741.29±6.6641.29 \pm6.6637.49±4.8337.49 \pm4.839.14±0.959.14 \pm0.9510.39±1.0810.39 \pm1.080.006
20.0–40.069.6±4.969.6 \pm4.931.95±8.0431.95 \pm8.0424.96±5.6924.96 \pm5.696.12±0.626.12 \pm0.628.50±0.878.50 \pm0.870.0004

Table IV. Inputs for the energy density (mid-rapidity yields dNch/dηdN_{\mathrm{ch}}/d\eta, Jacobian JJ and transverse area ⟨A⊥⟩\langle A_{\perp}\rangle for width and inclusive classes), resultant values of the energy density εBj\varepsilon_{\mathrm{Bj}}, and MSE/DoF fit metric results for the ΔpT\Delta p_{\mathrm{T}} spectra shifts. Note that ALICE and STAR dNch/dηdN_{\mathrm{ch}}/d\eta results were used for ATLAS and PHENIX εBj\varepsilon_{\mathrm{Bj}} 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, c0c_0, c2c_2 and the ratio c2/c0c_2/c_0, from the Glauber simulated AWA_{\mathrm{W}} and AUA_{\mathrm{U}} areas for the collision systems studied are presented in Table V. Note that the Average Radius method represents the AWA_{\mathrm{W}} class, while the Energy Half-Max Contour is in the AUA_{\mathrm{U}} class. These values are used in the model estimate for high-pTp_{\mathrm{T}} hadron elliptic flow v2v_2.

Exploration of Fitting Procedures and Fit Metrics for pTp_T Spectrum Fitting

At first glance, the choice of the MSE fit metric Eq. (8) over conventional choices like the χ2\chi^2 or Likelihood appears arbitrary. This appendix provides details on the studies undertaken to select the MSE metric, where it performs better than the χ2\chi^2 and Likelihood in extracting parameters of interest from pTp_T spectra, particularly the high pTp_T power law tail.

To begin, it is helpful to compare the MSE more directly to the χ2\chi^2 from their respective definitions. The χ2\chi^2 between an observed dataset OiO_i and fitting expectation EiE_i with corresponding error δEi\delta E_i is given by

χ2=∑bins i(Oi−EiδEi)2.(11)\chi^2 = \sum_{\mathrm{bins}\ i} \left(\frac{O_i-E_i}{\delta E_i}\right)^2 . \tag*{(11)}

The MSE metric, defined by Eq. (8), admits the power series expansion in the case of reasonable fits ∣Oi/Ei∣<2\lvert O_i/E_i\rvert< 2

MSE=∑bins i(Oi−EiEi−12(Oi−EiEi)2+⋯ )2=∑bins i(Oi−EiEi)2+O((Oi−Ei)/Ei)4.(12)\begin{aligned} \mathrm{MSE} = \sum_{\mathrm{bins}\ i} \left(\frac{O_i-E_i}{E_i} - \frac{1}{2}\left(\frac{O_i-E_i}{E_i}\right)^2 + \cdots\right)^2 \\ &= \sum_{\mathrm{bins}\ i} \left(\frac{O_i-E_i}{E_i}\right)^2 + O\left((O_i-E_i)/E_i\right)^4 . \tag*{(12)} \end{aligned}

That is to say, an MSE fit is at leading order in (Oi−Ei)/Ei(O_i-E_i)/E_i exactly a χ2\chi^2 fit with error δEi∝Ei\delta E_i \propto E_i, a generous but reasonable assumption for measured pTp_T 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 χ2\chi^2 for the same fit, which assures that the fits discussed in Tabs. I, Table IV are of quality as good or better than for χ2/DoF\chi^2/\mathrm{DoF} equivalent to the resulting MSE.

This error scaling additionally produces a unitless fit metric, and weights all bins ii equally against the fit form. This reduces the bias toward bins with larger magnitude, which bias traditional χ2\chi^2 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 n∈[6,12]n \in[6, 12], T∈[0.1,0.25]T \in[0.1, 0.25] GeV, and normalization AA chosen so that the distribution integrates to unity over the range pT∈[5,50]p_T \in[5, 50] 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 pppp 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 χ2\chi^2, 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 AA, nn, TT 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 pTp_{\mathrm{T}} 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 pppp data spectra above the relevant pTp_T 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 χ2\chi^2, 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 ∣{A,n,T}thrown−{A,n,T}fitted∣\lvert\{A,n,T\}_{\mathrm{thrown}}-\{A,n,T\}_{\mathrm{fitted}}\rvert 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 pTp_T range to study the bias of fitting methods against pTp_T. Finally, the distribution of resultant minimized fit metrics χ2\chi^2, χNLL2\chi^2_{\mathrm{NLL}}, 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 χ2\chi^2 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 pTp_T. 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 pppp distribution studied here and fit in Tab. I contains varying counts of hadrons.

The bias toward low pTp_T in the χ2\chi^2 and NLL fitters is most apparent in Figure 13 (e.), where the residuals in the ratio fit/truth exploded at high pTp_T. Meanwhile, the MSE fit remains flat over the full range, indicating that it weights the full pTp_T range equally. As a consequence, the resultant MSE fits extract the power law nn with two orders of magnitude smaller error on average than the χ2\chi^2 and NLL, though the improvements for the remaining parameters AA, TT are smaller and likely a direct consequence of better extraction of nn. 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 ΔpT\Delta p_T.

The extracted MSE/DoF can be interpreted similarly to a χ2/DoF\chi^2/\mathrm{DoF}, 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 χ2\chi^2. To show this, we fit an exponential ApTe−BpT+CA p_{\mathrm{T}} e^{-B p_{\mathrm{T}}+C} against the pppp spectra in the range pT∈[5,50]p_{\mathrm{T}} \in[5, 50] GeV, which is of comparable order of magnitude in the fitting region but has an incorrect shape relative to the pppp spectra. Then, we fit the Tsallis form against data sampled from this fit exponential using the χ2\chi^2, NLL and MSE fits. We found that the resultant fit metrics per degree of freedom were all larger than 5, and within ±3\pm3 of one another, which demonstrates that the MSE/DoF is comparable to a χ2\chi^2/DoF for bad fits as well as good ones.

Investigation of ΔpT\Delta p_{\mathrm{T}} 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 ΔpT(εBj)\Delta p_{\mathrm{T}}(\varepsilon_{\mathrm{Bj}}) obtained in Sec. III exhibit any clear dependence against other parameters such as ⟨Npart⟩\langle N_{\mathrm{part}}\rangle, ⟨A⊥⟩\langle A_{\perp}\rangle that might reveal a residual geometry or path-length component. Specifically, we consider an asymptotic expansion

ΔpT=αεBj+F(Leff,ε2,⟨Npart⟩,⋯ ),(13)\Delta p_{\mathrm{T}} = \alpha\varepsilon_{\mathrm{Bj}} + F(L_{\mathrm{eff}},\varepsilon_{2},\langle N_{\mathrm{part}}\rangle,\cdots), \tag*{(13)}

with α=const.\alpha= \mathrm{const.} and F(Leff,ε2,⟨Npart⟩,⋯ )=o(εBj)F(L_{\mathrm{eff}},\varepsilon_{2},\langle N_{\mathrm{part}}\rangle,\cdots) = o(\varepsilon_{\mathrm{Bj}}) (little-O), and investigate the form of this additional contribution

via the fit residuals

δ(ΔpT)=ΔpT−αεBj=F(Leff,ε2,⟨Npart⟩,⋯ ).(14)\delta(\Delta p_{\mathrm{T}}) = \Delta p_{\mathrm{T}} - \alpha\varepsilon_{\mathrm{Bj}} = F(L_{\mathrm{eff}}, \varepsilon_{2}, \langle N_{\mathrm{part}}\rangle, \cdots). \tag*{(14)}

Here LeffL_{\mathrm{eff}} denotes an effective medium path length and ε2\varepsilon_{2} an effective second order eccentricity/path-length difference scale ΔLeff\Delta L_{\mathrm{eff}}; these are expected to represent the strongest next-to-leading-order dependence though in principle FF may depend dynamically on a number of parameters.

The energy loss estimate ΔpT\Delta p_{\mathrm{T}} and corresponding residuals δ(ΔpT)\delta(\Delta p_{\mathrm{T}}) plotted against ⟨Npart⟩\langle N_{\mathrm{part}}\rangle are shown in Figure 14. Note that while ΔpT\Delta p_{\mathrm{T}} and ⟨Npart⟩\langle N_{\mathrm{part}}\rangle do not depend on the choice of area scaling method, the residuals δ(ΔpT)\delta(\Delta p_{\mathrm{T}}) do inherit a slight dependence on ⟨A⊥⟩\langle A_{\perp}\rangle from the fit form against εBj\varepsilon_{\mathrm{Bj}}, which varies slightly between ⟨A⊥⟩\langle A_{\perp}\rangle 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 ΔpT\Delta p_{\mathrm{T}} across ⟨Npart⟩\langle N_{\mathrm{part}}\rangle for the various collision systems studied in this work. (Bottom) Residuals from the ΔpT(εBj)\Delta p_{\mathrm{T}}(\varepsilon_{\mathrm{Bj}}) linear fit plotted against ⟨Npart⟩\langle N_{\mathrm{part}}\rangle.

In the ⟨Npart⟩\langle N_{\mathrm{part}}\rangle dependence of ΔpT\Delta p_{\mathrm{T}} (top panel of Figure 14), a clear separation between the LHC and RHIC datasets is present, and attributable to the expected sNN\sqrt{s_{\mathrm{NN}}} dependence of energy loss. This separation disappears in the residuals against the εBj\varepsilon_{\mathrm{Bj}} fit, indicating that the sNN\sqrt{s_{\mathrm{NN}}} dependent evolution of ΔpT\Delta p_{\mathrm{T}} can be well captured by εBj\varepsilon_{\mathrm{Bj}}. In particular, the ALICE Pb–Pb datasets at sNN=2.76\sqrt{s_{\mathrm{NN}}} = 2.76, 5.02 TeV [55] now sit between the Au–Au [56, 59] and Cu–Cu [58] datasets at sNN=200\sqrt{s_{\mathrm{NN}}} = 200 GeV recorded by STAR. Note that the discrepancy between the STAR and PHENIX [57] Au–Au datasets is due to differences in the pppp reference spectrum as discussed in Sec. III. In general, the residual data do not show a clear trend against ⟨Npart⟩\langle N_{\mathrm{part}}\rangle, indicating that the system size dependence of ΔpT\Delta p_{\mathrm{T}} is also generally well described by the εBj\varepsilon_{\mathrm{Bj}} 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 ΔpT\Delta p_{\mathrm{T}} residuals are plotted against estimates of path length LeffL_{\mathrm{eff}} and path length difference ΔLeff\Delta L_{\mathrm{eff}} in Figure 15. Both the EBE area and phenomenological methods are used to estimate LeffL_{\mathrm{eff}}, using Leff=⟨A⊥⟩/πL_{\mathrm{eff}} = \sqrt{\langle A_{\perp}\rangle/\pi} for the EBE methods and Leff=c0L_{\mathrm{eff}} = c_{0} for the phenomenological methods; for ΔLeff\Delta L_{\mathrm{eff}}, the phenomenological c2/c0c_{2}/c_{0} is used. Results are shown for both the width class AWA_{\mathrm{W}} and the inclusive class AUA_{\mathrm{U}} in the left and right columns of Figure 15 respectively. Note that the coefficients c0c_{0}, c2c_{2} come from the Average Radius method for the AWA_{\mathrm{W}} class, and from the Energy Half-Max Contour for the AUA_{\mathrm{U}} 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 LeffL_{\mathrm{eff}}, affirming the grouping of EBE and phenomenological methods into common classes performed in this study. In addition, the dependences on ΔLeff\Delta L_{\mathrm{eff}} are reversed relative to those for LeffL_{\mathrm{eff}}, 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 ΔpT\Delta p_{\mathrm{T}} as a function of LeffL_{\mathrm{eff}} as modeled by ⟨A⊥⟩/π\sqrt{\langle A_{\perp}\rangle/\pi} (first row) and the harmonic coefficient c0c_0 (second column), as well as the path length difference ΔLeff∼c2/c0\Delta L_{\mathrm{eff}} \sim c_2/c_0 for the various systems studied. For each such dependence, the residuals δ(ΔpT)\delta(\Delta p_{\mathrm{T}}) from the εBj\varepsilon_{\mathrm{Bj}} fit are also shown in the corresponding breakout panel below. All plots in the left column correspond to the width-based area calculation AWA_{\mathrm{W}}, while those in the right use the inclusive area A∪A_{\cup}.

Because all the residual plots share the same data in the yy-axis, the overall ordering of datasets in the residuals is consistent among all plots of Figure 15. However, as for ⟨Npart⟩\langle N_{\mathrm{part}}\rangle, the fitting against εBj\varepsilon_{\mathrm{Bj}} succeeds in capturing the general behavior of the datasets across both LeffL_{\mathrm{eff}} and ΔLeff\Delta L_{\mathrm{eff}}, as well as sNN\sqrt{s_{\mathrm{NN}}}, leaving residuals in each case that do not exhibit a clear dependence on ⟨A⊥⟩\sqrt{\langle A_{\perp}\rangle}, c0c_{0}, or c2/c0c_{2}/c_{0}. Specifically, there is no clear leftover shape in the dependence of the residuals on LeffL_{\mathrm{eff}} or ΔLeff\Delta L_{\mathrm{eff}} 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 ΔpT\Delta p_{\mathrm{T}} 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 v2v_2 predictions using AUA_{\mathrm{U}}

For the sake of completeness, and for comparison against the results using AWA_{\mathrm{W}}-based v2v_2 shown in Figure 11, we also show the v2v_2 using the AUA_{\mathrm{U}} in Figure 16.

Figure 16. High pTp_T v2v_2 predictions using the ΔpT\Delta p_T and AUA_{\mathrm{U}} based model and for a variety of collision systems.

References

  1. [1]J. W. Harris and B. Müller, “QGP Signatures” revisited, Eur. Phys. J. C 84, 247 (2024).arxiv.org/abs/2308.05743
  2. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [39]A. Takacs and K. Tywoniuk, Quenching effects in the cumulative jet spectrum, JHEP 10, 038, arXiv:2103.14676 [hep-ph].DOI
  40. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [63]S. Acharya et al. (ALICE), Centrality determination in heavy ion collisions, CERN Document Server (2018).
  64. [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. [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. [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. [67]J. D. Bjorken, Highly Relativistic Nucleus-Nucleus Collisions: The Central Rapidity Region, Phys. Rev. D 27,DOI
  68. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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. [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].
  84. [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
  85. [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).

Paper details

vlnc:W_xhhjbzqn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents