WorksPhysics and Astronomy
Collective Flow, Kinetic Freeze-Out, and a Data-Driven Baryon Chemical Potential in Au+Au Collisions at – GeV
Abstract
We study how transverse and longitudinal collective flow affect kinetic freeze-out conditions and the net baryon density in 0–5% central Au+Au collisions at – GeV, using a covariant statistical fireball model fitted to STAR identified-hadron transverse-momentum spectra at three fixed longitudinal velocities , , and . \rev{Longitudinal flow shifts the extracted temperature upward through a kinematic – degeneracy rather than by hardening the spectra; imposing – MeV bounds the longitudinal flow to .} \nw{A simultaneous , , fit with shared sharpens the extraction \fx{by a factor of – on relative to the single-species proton fit} and gives –~MeV, reproducing STAR's independent blast-wave values \fx{to within } and lying well below the chemical freeze-out temperature.} \fx{Further including the antiproton spectra, with a normalization shared between and , breaks the fugacity–normalization degeneracy and makes the baryon chemical potential data-driven: the ratio fixes , the meson spectra fix , and together they give –~MeV at kinetic freeze-out, satisfying –: the ratio is thus largely, but not exactly, set at the chemical surface, the – deficit being significant at up to . The resulting kinetic freeze-out trajectory in the plane shows that decoupling occurs in a dilute system, reaching only at ~GeV, \fx{a factor of } below the chemical-surface compression maximum, while longitudinal flow raises the inferred \fx{} by up to a factor of 4.}
Introduction
Mapping the phase diagram of strongly interacting matter, governed by quantum chromodynamics (QCD), across the space of temperature and chemical potentials (baryon, strangeness, etc.), and locating the critical point, represents one of the principal objectives of heavy-ion collision experiments and theoretical studies. Substantial progress has been made toward achieving this goal [1, 2, 3, 4, 5, 6, 7, 8, 9].
Within the QCD phase diagram [10, 11, 12], hadron-yield systematics [13, 14] show that chemical freeze-out, the stage at which inelastic reactions cease and the hadron abundances are fixed, occurs along a smooth curve in the plane [15, 16]. Near , this chemical freeze-out line lies close to the QCD crossover band at MeV established by lattice QCD [7, 17], thus providing an empirical “thermometer” anchoring the phase diagram at small baryon density [17]. After chemical freeze-out, the hadron gas continues to expand and cool until kinetic freeze-out, when elastic interactions also cease and the momentum spectra are frozen; kinetic freeze-out therefore occurs later and cooler, , and it is this later surface that transverse-momentum spectra probe. As decreases the baryon chemical potential increases, the chemical freeze-out line is expected to pass in the vicinity of any putative QCD critical point. This makes it the natural staging ground for investigating fluctuation observables and non-monotonic trends in Beam Energy Scan (BES) programs [18, 19]. Crucially, Randrup and Cleymans recast the chemical freeze-out systematics into thermodynamic variables constrained by conservation laws, the net baryon density and the energy density . They show that at freeze-out exhibits a clear maximum at intermediate energies (corresponding to –500 MeV), pinpointing the region of highest net-baryon compression accessible in heavy-ion collisions and guiding the choice of collision energies at FAIR and RHIC low-energy runs [15]. We extend that construction to the kinetic freeze-out surface by incorporating collective flow into the fireball description and re-evaluating the freeze-out mapping in the plane across . This allows us to assess how dynamical expansion and the later, cooler decoupling stage modify the net-baryon compression attained at decoupling and, consequently, the proximity of the freeze-out conditions to the conjectured critical region. At the kinetic surface, the net-baryon density is found to be largest at the lowest energy studied, placing the compression maximum at or below GeV, while its magnitude remains strongly sensitive to the assumed longitudinal flow.
The fireball formed in relativistic heavy-ion collisions is a spatially and temporally localized region of hot, dense matter composed of deconfined quarks and gluons, known as the Quark-Gluon Plasma (QGP). As the system expands and cools, it undergoes a transition to hadronic matter composed of various hadron species, well described by a Hadron Resonance Gas (HRG). There is ample evidence that both the QGP and HRG phases attain a state of local thermal equilibrium, characterized by a temperature and chemical potential . The HRG continues to expand until kinetic freeze-out, at which point elastic interactions cease and the momentum distributions of the constituent hadrons are fixed; it is these distributions that are ultimately recorded in detectors. As a conceptual baseline, one may treat the fireball as a static, thermalized source in the laboratory frame, from which hadrons are emitted with energies governed by the relativistic Boltzmann distribution,
where is a normalization constant and the factor of arises from the momentum-space density of states, . In reality, however, the fireball undergoes a dynamical evolution driven by internal pressure gradients, generating collective flow fields that superimpose a common expansion velocity on the underlying thermal motion. As a result, observables such as transverse-momentum spectra and particle yields carry the combined imprint of thermal and collective dynamics, and neglecting flow biases the freeze-out parameters extracted from thermal fits.
Concretely, we derive the invariant spectrum emitted by a fireball element having independent longitudinal and transverse flow velocities, and fit it to STAR identified-hadron spectra measured in central Au+Au collisions at –39 GeV. The fit is built up in three stages: single-species fits establish the – kinematic degeneracy; a simultaneous , , fit with shared fixes the kinetic freeze-out temperature; and the inclusion of with a normalization shared with renders data-driven through the ratio. This last step is what distinguishes the present analysis from a thermal-model fit: both and on the kinetic surface are determined by the measured spectra alone, with no chemical-freeze-out systematics used as input, so that the comparison with the chemical surface is a genuine comparison rather than a consistency check against an assumed parametrization. The resulting are then mapped onto using a hadron resonance gas, allowing the kinetic freeze-out trajectory to be compared directly with the chemical one.
The paper is organized as follows. In Section II, we present the theoretical formulation used to model a dynamically expanding fireball. Section III contains our main results, including fits to transverse-momentum spectra across a range of center-of-mass energies , from which we extract the freeze-out temperature and baryon chemical potential , and subsequently map these onto the energy density and net baryon density as functions of . Section IV presents our conclusions and outlook.
Throughout this paper, we adopt natural units, setting . We choose a mostly negative metric .
Model of a Dynamic Fireball
Before introducing the formulation for particle emission from a dynamically expanding thermal fireball, we first review the static fireball framework employed in Ref. [15], which serves as the baseline against which our approach is contrasted.
In the static fireball scenario, hadrons are assumed to be emitted from a thermal source following the Lorentz-invariant differential momentum distribution
where , , and are the energy, transverse momentum, and longitudinal momentum of the particle, respectively. A more convenient kinematic variable in high-energy collisions is the rapidity , defined as
from which it follows that
We further define the transverse mass , where is the particle mass, and the azimuthal angle . The Lorentz-invariant phase-space volume element can then be written as
with the energy expressed as . Substituting Eqs. (1) and (5) into Eq. (2), the differential particle spectrum for the static fireball becomes
where . The transverse-mass spectrum is obtained by integrating Eq. (6) over rapidity within the experimentally accepted window ,
This formulation is used in Ref. [15] to evaluate the net baryon density and energy density.
To model the dynamic fireball, we adopt the covariant statistical formulation of Ref. [20]. As noted in the introduction, the fireball undergoes dynamical evolution driven by internal pressure gradients, generating collective flow fields that superimpose a common expansion velocity on the underlying thermal motion. Consider a particle with four-momentum measured in the rest frame of a moving volume element of the fireball. We determine the transformation of the intrinsic thermal spectrum defined in the local rest frame of the moving source as seen by an arbitrary laboratory observer, expressed entirely in terms of Lorentz-covariant quantities. For a comoving volume element at rest in the local frame, the Touschek invariant phase-space measure reads [20]
where denotes the Dirac delta function restricted to positive energies, and is the four-volume element of the flowing fireball as observed in the laboratory frame, with the local four-velocity field of the fireball. A standard calculation yields
Adopting a cylindrical coordinate system consistent with the azimuthal symmetry of the collision geometry, the four-momentum of the particle is
We introduce a transverse rapidity for the particle, its longitudinal rapidity being the variable of Eq. (3), via
where is the particle’s transverse velocity in the frame reached from the laboratory by a longitudinal boost of rapidity , in which the particle has no longitudinal momentum; the two relations in Eq. (11) are consistent only with this reading, since gives . The same convention is used below for the flow rapidity . Equation (10) then becomes
Similarly, the four-velocity field of the fireball in cylindrical coordinates is
with . Primed quantities (, , , ) refer to the fireball velocity field, distinguishing it from the velocity of the particle .
Contracting Eqs. (12) and (13) gives a form symmetric in the particle and flow rapidities,
in which the unprimed factors belong to the particle four-momentum and the primed ones to the fluid four-velocity . Using the particle relations and from Eq. (11) for the first factor of each pair, and the corresponding flow relation for the second, this becomes
so that the transverse momentum is a particle variable while the transverse velocity is a property of the flow; the particle’s own does not appear. Substituting Eqs. (8) and (15) into the general spectrum, we obtain the invariant spectrum, which generalises Eq. (7),
where and . The integrand of Eq. (16) depends on the two azimuths only through their difference . For a single volume element, we may therefore choose the coordinate system so that , giving ; this is a relabeling of the azimuthal origin and involves no physical assumption. The transverse flow of the fireball as a whole is of course directed radially outward, so varies from element to element and cannot be set to zero globally. What makes Eq. (18) nonetheless applicable to the whole source is azimuthal symmetry: for the central collisions considered here, the elements are distributed uniformly in and carry a radial flow of common magnitude , so integrating over contributes only an overall factor , absorbed into in Eq. (19), and leaves the -dependence unchanged. Each element then contributes identically to the azimuthally integrated spectrum, and is to be read as a single effective radial flow velocity rather than a local one; we do not resolve a radial velocity profile, and azimuthal anisotropy is absent by construction. Since is even, we may restrict the integration to and multiply by a factor of 2. The -integral is then identified with the modified Bessel function of the first kind,
which we evaluate numerically. Performing the azimuthal integration in Eq. (16) yields
where .
Equation (18) gives the invariant particle spectrum emitted from a volume element of the dynamic fireball with transverse and longitudinal flow velocities and , as observed in the laboratory frame. The prefactor is an overall normalization,
in which collects the comoving volume and all -independent experimental factors, and is the fugacity associated with the conserved baryon number of the species ( for mesons). Only the product enters the spectrum; we return to the consequences of this in Sec. III. Since the fireball is characterised by four independent parameters (the temperature , the baryon chemical potential , the transverse flow velocity , and the longitudinal flow velocity ), we treat all four as free parameters and determine them by fitting Eq. (18) to experimental transverse-momentum spectra.
We note that Eq. (8) has the structure of a Cooper–Frye emission integral [21] for a freeze-out element whose surface normal is proportional to , and Eq. (18) may correspondingly be viewed as a covariant, single-surface-velocity variant of the blast-wave spectrum of Schnedermann, Sollfrank, and Heinz [22]. Unlike the standard blast wave, it retains an explicit longitudinal flow velocity and does not average over a transverse velocity profile; the fitted should therefore be compared with a surface, rather than profile-averaged, transverse velocity. We also note that the parametrization of Eq. (13) is a longitudinal boost of rapidity applied to a purely transverse flow: boosting by along gives . Hence is the longitudinal flow velocity in the laboratory frame, whereas is the transverse flow velocity measured in the longitudinally comoving frame, in which the flow has no longitudinal component. The laboratory-frame transverse flow velocity is correspondingly , smaller by one factor of ; the identification refers to the comoving transverse velocity.
A remark on notation. Throughout this section primes distinguish the fireball velocity field from the particle kinematics, the latter entering only through Eqs. (10)–(12). Since the particle transverse velocity of Eq. (11) plays no further role once the spectrum Eq. (18) has been obtained (the particle kinematics entering thereafter only through the measured , and ), we drop the primes from here on and write and for the transverse and longitudinal flow velocities of the fireball. All velocities quoted in Secs. III–IV, and in every figure and table, are fireball flow velocities in this sense.
Calculation of Energy Density and Net Baryon Density
The spectrum of Eq. (18) is controlled by the freeze-out temperature and, through the fugacity of Eq. (19), by the baryon chemical potential ; the fits that extract these from the measured spectra are presented in Sec. III. Since the freeze-out systematics of Ref. [15] are formulated not in but in the thermodynamic variables fixed by conservation laws, a mapping from the former to the latter is required. We set out that mapping here, before turning to the data. Specifically, we employ the grand-canonical partition function for each hadron species to express the energy density and the net baryon density of a hadron resonance gas at a given . The total partition function factorizes over all species as , where depends on the temperature , the volume element , and the baryon, electric charge, and strangeness chemical potentials , , and , respectively,
where the upper (lower) sign applies to bosons (fermions), is the mass of species , is the spin degeneracy, and is the fugacity
with , , and denoting the baryon number, electric charge, and strangeness of species , respectively. The subscripted of Eq. (21) is the species fugacity in the conventional hadron-resonance-gas notation, and is not to be confused with the fitted spectral normalization of Eq. (19).
The number density of species follows from the standard thermodynamic relation
which, upon substitution of Eq. (20), yields
For the thermodynamic densities reported below, we retain the full Bose–Einstein/Fermi–Dirac series (summed to convergence). The leading () term already provides an excellent approximation for the heavy baryons that dominate the net-baryon density, whereas the higher terms give a non-negligible contribution for light mesons (pions) and are kept for the energy density. The net baryon density is then obtained by summing the number densities weighted by the baryon number ,
where the sum runs over all hadron species. The energy density of species is given by , where . Evaluating this derivative yields
The coefficient 3 multiplying reflects the mean thermal energy per particle, ; equivalently . Keeping the leading term, the energy density simplifies to
The total energy density is . Equations (23), (24), and (26) thus determine and for any given . In Sec. III they are evaluated at the freeze-out parameters obtained from the fits, yielding the kinetic freeze-out trajectory in the plane across the full range of collision energies considered.
Results and Discussion
In this section, we present and discuss our main results. We begin by examining the best fits of Eq. (18) to the experimentally measured transverse-momentum () spectra of identified hadrons, specifically, positive pions and protons in 0–5% central Au+Au collisions measured by the STAR Collaboration at –39 GeV [23]. These fits determine the temperature , the transverse flow velocity and the normalization for three fixed values of the longitudinal velocity , 0.2, and 0.4; the baryon chemical potential is not accessible at this stage, for the reason given below, and is extracted from the data only in Sec. III D. We then compute particle yields by integrating the fitted spectra over the rapidity range and the transverse momentum range GeV/c, and examine their dependence on at each fixed in order to quantify the effect of collective flow on particle abundances.
For a single hadron species, enters the spectrum, Eq. (18), only through the fugacity , which is -independent and therefore multiplies a shape function of without distorting it. By Eq. (19) the fugacity is not separable from the volume factor : the data constrain only their product , and since is itself unknown, and the normalization are exactly degenerate. The mid-rapidity proton spectrum fixes the spectral shape, and hence and , but carries no independent information on . We verified this explicitly: scanning over a wide range at fixed and re-optimizing the normalization leaves the unchanged. We therefore do not treat as a free parameter. Instead we hold it fixed at each energy while fitting only the temperature , the transverse flow velocity , and the normalization to the measured proton spectra.
A value for the fixed is nonetheless required in order to report one, and for this we take the chemical freeze-out systematics of Cleymans et al. [24],
with GeV and GeV, together with the associated freeze-out temperature curve
with GeV, GeV, and GeV. These parametrizations were obtained from thermal model fits to particle yields at chemical freeze-out, prior to the BES program.
Their use in a kinetic freeze-out analysis therefore requires comment. Equation (27) is not a constraint on the fit but a placeholder for a quantity the single-species data do not determine, and the degeneracy established above guarantees that its value cannot propagate into any fitted quantity. We have verified this directly: rescaling the input over the range 0.25–4 times Eq. (27) shifts the fitted by less than MeV and by less than , leaving unchanged to one part in , at every energy, at each of the three values, and for both the single-species and three-species fits; the normalization absorbs the change exactly. The temperatures and flow velocities of this section are thus independent of the chemical-freeze-out input, and the choice affects only the column echoed back in Table I and any density evaluated from it. Equation (28) is likewise never used as a fit function; it enters only as the reference chemical-freeze-out line against which the fitted kinetic temperatures are compared, which is precisely the role a chemical-freeze-out parametrization should play here. Neither equation is used in the determination of in Sec. III D, where it serves at most as a starting value for the minimizer, and the trajectory reported there is built entirely on the data-driven .
A genuinely data-driven would require a simultaneous multi-species fit (e.g. , , , ), in which the relative normalization breaks the degeneracy; this is carried out in Sec. III D, where the inclusion of the antiproton spectra determines directly from the data. The resulting energy-dependent and (input) , together with the full set of fitted parameters and their uncertainties, are reported in Table I. The fitted kinetic freeze-out temperatures lie somewhat below the Cleymans chemical-freeze-out curve, as expected physically: kinetic freeze-out occurs later and cooler than chemical freeze-out, and the transverse flow (–) accounts for part of the spectral hardening that a purely thermal fit would otherwise absorb into a higher .
Fitting the Transverse Momentum Spectra
Figures 1 and 2 show the transverse momentum spectra of protons and positive pions, respectively, fitted to STAR measurements in 0–5% central Au+Au collisions at –39 GeV within the rapidity range . Symbols with error bars are the STAR data and the solid curves are the corresponding best fits, each drawn in the same color as the data set it describes. The curves are obtained by minimizing with respect to and then evaluating Eq. (18) at the resulting best-fit parameters on a continuous grid.
The fits describe the experimental data well across the full range of collision energies considered, as confirmed by the values for the proton fits reported in Table I for all three longitudinal velocities. With fixed to the Cleymans systematics, the free parameters are , so that the number of degrees of freedom equals the number of points minus three. The fit quality is essentially independent of , degrading by only 2–6% in between and at every energy. This is a direct expression of the fact, discussed further in Sec. III B, that mid-rapidity spectra carry almost no information on the longitudinal flow: the data cannot distinguish between the three values, and is accordingly treated as a fixed external parameter rather than fitted. The weights entering the are the total uncertainties published with the STAR invariant yields (the quadrature sum of the statistical and systematic components), rescaled by the same azimuthal/rapidity factor as the yields themselves. Over the windows fitted here these are strongly dominated by the systematic component: the median ratio of statistical to systematic uncertainty is 0.03–0.08 depending on species, so that is in effect the systematic error, amounting to 6–12% of the yield. This has a direct consequence for the reduced . Systematic uncertainties of this type (normalization, tracking and particle-identification efficiencies) are largely correlated from one bin to the next, whereas the diagonal used here treats them as independent. The reduced is therefore expected to fall well below unity, and indeed does so for the single-species fits of Table I (–); this reflects the error treatment rather than an anomalously accurate description of the data. The parameter uncertainties quoted throughout are the corresponding diagonal covariance errors, with no rescaling by applied anywhere in this work.
They are conservative in magnitude but do not carry a strict frequentist interpretation, and they should be read together with the strong – correlation discussed in Sec. III B rather than as independent one-parameter errors.
The fitted transverse flow is stable across the whole energy range, –, rising slightly with , while the temperature increases monotonically with at fixed energy; the associated parameter correlations are discussed in Sec. III B.
In evaluating Eq. (18), we integrate over the same rapidity window as the data, , rather than evaluating the spectrum at .
We now use the extracted parameters to compute the proton yield by integrating Eq. (18) over rapidity and transverse momentum GeV (the range adopted directly from the STAR data), for 0–5% central Au+Au collisions at –39 GeV. The resulting proton yields are tabulated in Table II.
Two notable features emerge from these results. First, the proton yield decreases monotonically with increasing collision energy, directly reflecting the energy dependence of baryon stopping. At lower , the colliding Au nuclei are relatively opaque to one another and a significant fraction of the incoming nucleons lose most of their longitudinal momentum through multiple inelastic scatterings and come to rest near mid-rapidity, producing a large net-baryon density in the central region. This is quantitatively captured by the large values of extracted from our fits at lower energies. As increases, the nuclei become progressively more transparent, and the incoming baryons largely retain their longitudinal momentum, being carried to forward and backward rapidities () while leaving a nearly baryon-free region at mid-rapidity. Consequently, at high energies, and the proton yield at mid-rapidity diminishes accordingly. The increasing availability of center-of-mass energy also drives copious production of quark-antiquark pairs, manifesting as a sharp rise in pion yields with , as clearly seen in Figure 2. Thus, as the collision energy increases, the character of particle production at mid-rapidity transitions from baryon-dominated to meson-dominated, a signature of the approach toward net-baryon transparency.
Figure 2. (Color online) Midrapidity () transverse momentum spectra for positive pions in Au+Au collisions at to 39 GeV for 0–5% centrality with longitudinal velocity of the dynamic fireball. Symbols with error bars are the STAR data [23]; the solid curves are the corresponding best fits of Eq. (18), each plotted in the same color as the data set it describes. The single-freeze-out Boltzmann spectrum reproduces the proton slope closely but describes the pions only approximately at low , where resonance feed-down (not included here) is significant.
Figure 1. (Color online) Midrapidity () transverse momentum spectra for positive protons in Au+Au collisions at to 39 GeV for 0–5% centrality with longitudinal velocity of the dynamic fireball. Symbols with error bars are the STAR data [23]; the solid curves are the corresponding best fits of Eq. (18), each plotted in the same color as the data set it describes.
| (GeV) | |
| 7.7 | 8.503 |
| 11.5 | 6.991 |
| 19.6 | 5.468 |
| 27 | 4.924 |
| 39 | 4.202 |
Table 2. Integrated proton yields at –39 GeV in 0–5% central Au+Au collisions. is the dimensionless per-event proton number obtained by integrating Eq. (18) over and GeV at the best-fit parameters of Table I; it is not . Values are quoted at ; they change by less than 0.02% at and 0.4, so a single column suffices (see text). The entering the fugacity is the Cleymans input of Eq. (27), not the data-driven value of Sec. III D.
Second, the proton yield is found to be essentially independent of the longitudinal fireball velocity . The cancellation is striking: going from to raises the fitted temperature by about 11% at every energy, yet changes the integrated yield by less than 0.02%. The normalization absorbs the change almost exactly, in the same way that it absorbs a rescaling of the input . This is why Table II quotes a single value per energy. This can be understood as follows. The net-baryon content at mid-rapidity is established during the very early, pre-equilibrium stage of the collision through baryon stopping, well before any collective flow develops. Collective longitudinal flow, characterized by , acts as a subsequent rigid rapidity boost of the expanding fireball. For the narrow rapidity window considered here, such a boost shifts the rapidity distribution by , but since the net-baryon rapidity distribution is broad and approximately flat near mid-rapidity at these energies, the integrated yield within this small window remains largely unaffected. In other words, the conserved baryon charge within is insensitive to the collective longitudinal motion of the fireball, as long as remains moderate, consistent with our chosen values of .
Each freeze-out point is formed by pairing the fitted kinetic temperature of Table I with the chemical-freeze-out baryon chemical potential of Eq. (27), which depends only on the collision energy and is therefore common to all three . This is done at each collision energy for 0–5% central Au+Au collisions and for each of the three fixed longitudinal velocities , , and . The resulting – freeze-out points are displayed in Figure 3 and listed in Table I.
Figure 3. (Color online) Fitted kinetic freeze-out temperature versus the (input) baryon chemical potential for three longitudinal velocities (black circles), (red squares), and (green triangles). Error bars are the fit uncertainties on . The shaded band marks the lattice QCD crossover temperature – MeV; all fitted points for lie at or below it. Gold stars are STAR’s published kinetic freeze-out temperatures from an independent blast-wave analysis, plotted at STAR’s measured chemical [23]; they fall just below our points, consistent with a kinetic freeze-out extraction. The upper axis gives the collision energy associated with each through Eq. (27), and the dotted vertical guides mark those five values, so that each column of points can be read directly as one energy. The small horizontal offset () between STAR’s measured and the Cleymans input used for our points is visible at the lowest energies as a displacement of the stars from the guides.
Several trends are immediately apparent. The input of Eq. (27) decreases monotonically with by construction, consistent with the picture of increasing nuclear transparency and diminishing net-baryon stopping at mid-rapidity discussed above. The fitted rises overall, spanning 119–136 MeV at , reflecting the larger energy density deposited in the collision zone at higher energies; the one exception is a 2 MeV dip at 11.5 GeV, well inside the MeV fit uncertainties of Table I (0.2) and discussed below in connection with . Together, these trends trace a freeze-out curve in the – plane that moves from the high-, low- region (low energies) toward the low-, high- region (high energies), a trend that qualitatively parallels the chemical freeze-out systematics of Cleymans et al. [24] while lying systematically below them in temperature, as expected for a later kinetic decoupling.
A second important feature visible in Figure 3 is the systematic upward shift in the extracted as increases, at all collision energies. This effect is not a consequence of longitudinal flow hardening the spectra; longitudinal motion is along the beam axis and cannot directly transfer momentum into the transverse plane. Rather, the origin is kinematic: in our model, the distribution function depends on the Lorentz-invariant product , where is the four-velocity of the fluid element. When , the longitudinal component of is nonzero, and the effective energy of a particle as seen in the fluid rest frame receives a contribution from the longitudinal motion through the Lorentz factor . Consequently, at fixed observed and rapidity , the argument of the thermal distribution is modified relative to the case. To reproduce the same observed spectrum, the least-squares fit compensates by returning a higher value of . Because the rapidity window is narrow, across it, so the exponent of Eq. (18) depends on only through the ratio and the fit should return . Table I bears this out at every energy: exceeds by only 0.3% at and 1.5–2.0% at , the residual coming from the accompanying rise in . This represents a genuine degeneracy in the model between the longitudinal collective velocity and the thermal temperature, and shows the importance of stating explicitly: a temperature quoted without it is an effective temperature that has silently absorbed the longitudinal flow. Since the spectra themselves cannot discriminate between the three (the degrades by only 2–6% from to 0.4, as noted above), is held fixed and scanned here rather than fitted. The baryon chemical potential is -independent by construction here, since it is fixed to the Cleymans systematics of Eq. (27); this is consistent with the physical expectation that the net-baryon content at mid-rapidity is governed primarily by the initial stopping dynamics rather than by the subsequent collective expansion. Crucially, over the explored range the shift leaves the fitted at or below the crossover band –160 MeV at every energy; the extrapolation of this trend to larger is quantified in Sec. III B.
Before presenting the densities, a comment is in order on the meaning of the points extracted here. Ref. [15] evaluated the HRG densities at the chemical freeze-out point . Our spectral fits instead determine the kinetic freeze-out temperature, while the baryon chemical potential is supplied externally by the chemical freeze-out systematics, Eq. (27). Evaluating an equilibrium HRG at the hybrid point is therefore an approximation: between chemical and kinetic freeze-out the abundances are frozen and the hadron gas is only in partial chemical equilibrium, with effective chemical potentials that grow as the system cools [25, 26]. Since increases during this (approximately isentropic) hadronic cooling, the densities reported below should be read as conservative (lower) estimates of the net-baryon compression at kinetic decoupling; this coupling; this also explains naturally why our values lie below the chemical-surface maximum of Ref. [15]. A self-consistent treatment, enforcing conservation of the frozen yields (e.g. constant from the chemical surface down to ), is left for future work; a first step in this direction, the data-driven determination of the net-baryon chemical potential on the kinetic surface itself from the ratio, is presented in Sec. III D.
Using the extracted and at each collision energy, we compute the net baryon density via Eq. (24) and the energy density via Eq. (26), setting and employing the full Bose–Einstein/Fermi–Dirac series (summed to convergence). The results are displayed in Figure 4 and tabulated in Table III.
Figure 4. (Color online) Kinetic freeze-out points in the plane, obtained by evaluating the statistical model at the fitted kinetic freeze-out temperature and the input baryon chemical potential , for the different STAR energies and longitudinal fireball velocities. Each point is labelled by its collision energy in GeV. The low-lying point in each series is 11.5 GeV, where the fitted is at its minimum (Table I) while has already dropped well below its 7.7 GeV value; the excursion is once the uncertainty on is propagated, and it is absent from the fully data-driven evaluation of Figure 13.
The net baryon density is largest at GeV and decreases overall with increasing collision energy. This behavior directly reflects the energy dependence of : as increases, the decreasing degree of baryon stopping leads to a smaller net-baryon density at mid-rapidity, and correspondingly smaller . The mild non-monotonicity around –19.6 GeV mirrors the correspondingly lower fitted at 11.5 GeV, to which is sensitive through the Boltzmann factor . The underlying temperature difference ( MeV) is well within the fit uncertainties, so this dip is not statistically significant. At higher collision energies, freeze-out therefore occurs in a more baryon-dilute environment characterized by low and low net baryon density, but at a higher thermal temperature, paralleling the qualitative trend of the chemical freeze-out systematics at correspondingly lower (kinetic) temperatures.
The extracted freeze-out temperatures span the ranges 119–136 MeV for , 122–140 MeV for , and 132–152 MeV for , across the collision energy range –39 GeV. The systematic upward shift of approximately 30–40 MeV per unit increment in is consistent with the kinematic degeneracy between longitudinal collective motion and thermal temperature discussed above: the larger the longitudinal velocity of the fireball, the larger the compensation in required by the fit to reproduce the observed spectra. Importantly, for the explored range these temperatures all lie at or below the lattice-QCD crossover band –160 MeV, so the hadron-resonance-gas description underlying the density evaluation remains internally consistent at every energy and every considered here.
Parameter correlations and a bound on the longitudinal flow
Because the transverse-momentum slope can be reproduced either by a higher temperature with weaker flow or by a lower temperature with stronger flow, the fitted parameters are strongly correlated. We quantify this directly from the fit covariance matrix. Table IV lists the Pearson correlation coefficients between the free parameters at each energy for , and Figure 5 displays the full correlation matrices; the pattern is essentially energy- and -independent. Temperature and transverse flow are strongly anti-correlated, , the well-known thermal/flow – degeneracy of blast-wave-type descriptions, while the normalization is almost perfectly anti-correlated with (), which is the same channel that makes undetermined once its fugacity is freed (Sec. III).
Figure 5. (Color online) Fitted-parameter correlation matrices for the proton fit (, , ) at , for each STAR energy. The strong – anti-correlation () and – anti-correlation () are essentially energy-independent.
| (GeV) | |||
| 7.7 | −0.950 | −0.998 | +0.937 |
| 11.5 | −0.954 | −0.998 | +0.943 |
| 19.6 | −0.955 | −0.999 | +0.948 |
| 27 | −0.963 | −0.999 | +0.957 |
| 39 | −0.957 | −0.999 | +0.951 |
Table IV. Fitted-parameter correlation coefficients for the proton fit at . The strong – anti-correlation reflects the thermal/flow degeneracy; and must be quoted jointly.
The practical consequence is that the one-parameter uncertainties on and in Table I understate the true joint uncertainty: the allowed region in the plane is a narrow, steeply tilted ellipse, shown in Figure 6. and should therefore be regarded as jointly determined rather than independently measured, and claims about the absolute value of either should be made with the correlation in mind. We note in addition that the mid-rapidity spectra () carry very little information on itself, since a longitudinal boost acts as a near-constant rescaling within this narrow window; is therefore scanned as a fixed external parameter rather than fitted.
Figure 6. (Color online) Joint (solid) and (dashed) confidence regions in the plane for . The narrow, tilted ellipses are the geometric expression of the – anti-correlation: and are jointly, not independently, constrained. The contours are curves of constant from the block of the fit covariance, at the two-parameter levels and (68.27% and 95.45%); as joint regions they project onto the axis as times the one-parameter errors of Table I.
Although cannot be determined from the mid-rapidity spectra, its effect on the extracted temperature provides an indirect, thermodynamically motivated upper bound. Fixing and refitting over a fine grid in , we find that rises smoothly and monotonically with (Figure 7), crossing the crossover band –160 MeV at a critical value listed in Table V. Beyond the fitted kinetic temperature would exceed , i.e. the hadron-resonance-gas description on which the model rests would no longer be self-consistent. The bound is –0.61 at the lowest energies and tightens to –0.51 at 39 GeV, where the baseline temperature is already higher. The values explored above thus lie safely within the physically admissible region at all energies, while longitudinal flow velocities in excess of roughly half the speed of light are disfavored on the grounds that they would push kinetic freeze-out above the crossover temperature.
Figure 7. (Color online) Fitted kinetic freeze-out temperature as a function of the longitudinal flow velocity , for the five STAR energies. The grey band is the crossover temperature – MeV; the vertical shaded strip marks the range of critical values – (Table V) at which crosses . The color-matched downwards arrows near mark the kinematic ceiling for each energy, and the shaded region beyond is kinematically inaccessible; the thermodynamic bound is far more restrictive.
Two kinematic remarks complete this discussion. First, no point of the scan is superluminal by construction: with the rapidity parametrization of Eq. (13) the total laboratory speed of a fluid element obeys for any (relativistic velocity addition is built into the four-velocity), the largest value reached anywhere on the grid being . Second, the longitudinal collective velocity of matter produced at mid-rapidity cannot exceed the beam velocity in the center-of-mass frame, with , which ranges from 0.970 at 7.7 GeV to 0.999 at 39 GeV (Table V, arrows in Figure 7). Since – lies far below this kinematic ceiling at every energy, the requirement, not kinematics, is the operative constraint on the longitudinal flow.
| (GeV) | () | () | |
| 7.7 | 0.57 | 0.60 | 0.970 |
| 11.5 | 0.59 | 0.61 | 0.987 |
| 19.6 | 0.47 | 0.51 | 0.995 |
| 27 | 0.46 | 0.50 | 0.998 |
| 39 | 0.43 | 0.47 | 0.999 |
Table V. Critical longitudinal velocity at which the fitted kinetic freeze-out temperature reaches the lower (155 MeV) and upper (160 MeV) edges of the crossover band, obtained by interpolating the scan of Figure 7. The last column lists the kinematic ceiling: the beam velocity in the center-of-mass frame, with , which the fireball’s longitudinal collective velocity cannot exceed. The thermodynamic bound is far more restrictive than the kinematic one at all energies.
Simultaneous multi-species fit and cross-check against STAR
The strong – anti-correlation of the single-species fit (Sec. III B) originates in the fact that the same spectral slope can be produced by a hotter, less rapidly flowing source or by a cooler, faster one. This ambiguity is lifted by the particle mass: under a common radial flow, heavier hadrons receive a larger transverse blue-shift, so the spectra of species of different mass constrain and in different combinations. We therefore perform a simultaneous fit of the , , and proton spectra at each energy, sharing a single kinetic freeze-out temperature and transverse flow velocity while allowing an independent normalization for each species, with again fixed to the Cleymans systematics. Because the fitted observables are the spectral shapes and the flow, this remains a determination of the kinetic freeze-out surface; the chemical information () is supplied externally. As shown above, rescaling that external by factors of four in either direction leaves the fitted and of this three-species fit unchanged to better than MeV and respectively, so the chemical input is inert here as well.
It is the sharing of a single and across species of different mass, not the freedom of the individual normalizations, that removes the near-singular – direction of the single-species fit: a change in can still be absorbed by the normalization of any one species, but the compensation required depends on the particle mass, so it cannot be made for , and at once. The joint constraint tightens accordingly: the resulting uncertainties on and shrink substantially relative to the proton-only fit, falling from 6.8–10.3 MeV to 1.6–2.0 MeV and from 0.019–0.025 to 0.008–0.009, gains of factors 3.7–5.3 and 2.2–2.9 respectively (Table I and Table VI). The shared kinetic temperature is correspondingly lower and more physical, –115 MeV at (rising to 120–129 MeV at ), and is nearly independent of . The residual – anti-correlation () persists, as it must for any blast-wave–type description, but the joint confidence region is far smaller.
| (GeV) | NDF | (GeV) | (GeV) | |||
| 7.7 | 26 | 0.0 | 0.422 | 0.031 | ||
| 0.2 | 0.422 | 0.031 | ||||
| 0.4 | 0.422 | 0.032 | ||||
| 11.5 | 25 | 0.0 | 0.316 | 0.040 | ||
| 0.2 | 0.316 | 0.041 | ||||
| 0.4 | 0.316 | 0.042 | ||||
| 19.6 | 26 | 0.0 | 0.206 | 0.051 | ||
| 0.2 | 0.206 | 0.052 | ||||
| 0.4 | 0.206 | 0.053 | ||||
| 27 | 20 | 0.0 | 0.156 | 0.075 | ||
| 0.2 | 0.156 | 0.076 | ||||
| 0.4 | 0.156 | 0.079 | ||||
| 39 | 19 | 0.0 | 0.112 | 0.028 | ||
| 0.2 | 0.112 | 0.028 | ||||
| 0.4 | 0.112 | 0.030 |
Table I. Fitted freeze-out temperature and transverse flow velocity (1 uncertainties from the fit covariance) obtained from the proton spectra using Eq. (18), at each collision energy and each fixed longitudinal velocity . The baryon chemical potential is an input fixed to Eq. (27), hence independent of ; the normalization is a nuisance parameter and is not tabulated. Parameter correlations are given in Table IV. NDF is -independent and is quoted once per energy.
| (GeV) | (this work) (MeV) | STAR (MeV) | (this work) | STAR | STAR (MeV) |
| 7.7 | 106.5(16) | 116(11) | 0.517(8) | 0.462(43) | 400 |
| 11.5 | 112.1(19) | 118(12) | 0.503(9) | 0.464(44) | 292 |
| 19.6 | 115.1(19) | 113(11) | 0.507(9) | 0.458(34) | 194 |
| 27 | 113.4(19) | 117(11) | 0.528(9) | 0.482(38) | 149 |
| 39 | 114.4(20) | 117(11) | 0.539(8) | 0.492(38) | 104 |
Table VI. Kinetic freeze-out temperature and transverse flow from the simultaneous fit at , compared with STAR’s published blast-wave and and measured chemical-freeze-out [23]. The Cleymans used as input (Table I) lies 5–8% above the STAR at every energy. Numbers in parentheses are uncertainties in units of the last digit. Ours are the one-parameter errors from the fit covariance, given unrounded so that they may be compared directly with the proton-only values of Table I; the STAR columns carry that collaboration’s published combined statistical and systematic errors.
Crucially, these values agree with STAR’s own published kinetic freeze-out parameters, obtained from an independent blast-wave analysis of the same data [23]. As shown in Table VI and Figure 9, our agrees with the STAR values at every energy: the differences run from +2.1 to −9.5 MeV and remain within of the combined uncertainties throughout, STAR’s blast-wave temperatures themselves carrying –12 MeV. Our lies well below the STAR chemical freeze-out temperature –160 MeV, exactly as expected if kinetic freeze-out occurs later and cooler than chemical freeze-out. Our fitted lies systematically above STAR’s mean transverse velocity by 0.04–0.06 (0.9–1.4), as expected for a single flow velocity compared with the average over a velocity profile. Finally, the external Cleymans we adopt lies 5–8% above STAR’s measured chemical-freeze-out at every energy. The simultaneous fit and its fit curves are shown in Figure 8.
Figure 8. (Color online) Simultaneous fit of the (blue), (green), and proton (red) spectra with a shared kinetic freeze-out temperature and transverse flow (curves), at for the five STAR energies (successively scaled by powers of 4 for clarity). is fixed to the Cleymans systematics.
Figure 9. (Color online) Left: the multi-species kinetic freeze-out temperature (this work, ) compared with STAR’s published blast-wave and chemical-freeze-out ; the grey band is – MeV. Right: our single flow velocity compared with STAR’s mean transverse velocity . STAR values from Ref. [23]. All error bars are : ours from the fit covariance, STAR’s their published combined statistical and systematic uncertainties (Tables IX and X of Ref. [23]). The shaded red bands span the scanned range and indicate the systematic associated with the longitudinal flow, which the mid-rapidity spectra do not constrain; the plotted points are .
Figure 9 displays the comparison. In the left panel our shared kinetic temperature tracks STAR’s blast-wave across the whole BES range, while both lie far below the chemical freeze-out temperature, which rises from 143.5 MeV at 7.7 GeV and levels off inside the crossover band from 19.6 GeV upward (157.6, 160.0 and 159.6 MeV). The separation consequently widens with energy, from 37 MeV at 7.7 GeV to 45–47 MeV at 27–39 GeV, as expected if the hadronic phase between the two freeze-out surfaces is longer-lived at higher collision energy. The right panel shows our and STAR’s rising together with , the offset between them staying nearly constant at 0.04–0.06. The shaded bands span the scanned range : within STAR’s uncertainties the comparison is insensitive to , so it corroborates the kinetic interpretation of the fit without independently constraining the longitudinal flow.
Applying the criterion of Sec. III B to the more robust multi-species temperature refines the bound on the longitudinal flow. Scanning the joint fit in (Figure 10), the shared reaches the crossover band at a critical value –0.65, somewhat larger than the proton-only estimate because the multi-species is lower. The explored range therefore lies comfortably within the physically admissible region at all energies, and longitudinal flow velocities above are disfavored on the grounds that they would drive kinetic freeze-out above the hadron-resonance-gas limit.
Figure 10. (Color online) Shared kinetic freeze-out temperature from the simultaneous fit as a function of . The grey band is – MeV; the vertical shaded strip marks the refined critical range –. Arrows near mark the kinematic ceiling per energy, as in Figure 7.
Data-driven extraction of from the ratio
The analyses above supply externally, because for any set of species fitted with independent normalizations the fugacity is absorbed into the factor of each baryonic species, Eq. (19), and remains undetermined. The antiproton resolves this. We extend the simultaneous fit of Sec. III C to the four species , , , and , sharing a single kinetic temperature , transverse flow , and baryon chemical potential , with independent normalizations for and but a common volume factor for and . Since the proton
and antiproton then differ only through their fugacities , the measured ratio determines directly, with no external chemical input ( as before); the spectra are fitted over the STAR window GeV, restricted to GeV at 7.7 GeV by the data coverage. The fit quality remains good, – (Figure 11), though the is not evenly distributed among the species: the spectra carry a per point of 1.5–3.0 and account for 63–76% of the total, the spectra 0.5–1.1, and the and spectra only 0.1–0.4. The temperature is therefore determined predominantly by the meson spectra, as anticipated in Sec. III C, while the baryon spectra—reproduced well within their systematics-dominated uncertainties—act mainly through the ratio. The shared and move by at most 2 MeV and 0.012 respectively with respect to the three-species fit, comparable to their uncertainties, confirming that the added spectra are consistent with the same kinetic surface.
Figure 11. (Color online) Simultaneous fit of the (blue), (green), (red), and (purple) spectra with shared and a common normalization, at for the five STAR energies. The splitting is controlled entirely by the fugacity factor , which fixes the ratio ; the simultaneously fitted and spectra, which are insensitive to but share the same , determine the temperature, and the two together yield .
The extracted chemical potential decreases from MeV at GeV to 63.2 MeV at 39 GeV (; Table VII), lying systematically below the chemical freeze-out values of Ref. [23] (–104 MeV, Table IX, strangeness-canonical yield fit) by approximately the ratio of the temperatures. This is precisely the behavior expected if the ratio is frozen at chemical freeze-out: since , conservation of the ratio during the hadronic cooling from to requires to remain constant, i.e.
Figure 12 shows that the extracted –0.56 indeed tracks STAR’s chemical –0.65 at every energy, at the 85–88% level. Propagating the uncertainties quoted by STAR on and together with our own fit covariance, this deficit is significant at 2.4–3.2 for GeV but only 1.5 at 39 GeV, where the chemical is itself known to ; the trend is therefore established by the lower energies. These significances treat STAR’s and uncertainties as uncorrelated, which is conservative: the positive correlation between them in a thermal fit would partially cancel in the ratio and increase the significance. The residual offset is consistent with the neglected weak feed-down (the inclusive STAR spectra receive different contributions for and , biasing the ratio slightly upward and hence slightly downward) and with .
Figure 12. (Color online) Left: the data-driven kinetic-surface (circles/triangles, three values) compared with the Cleymans chemical freeze-out systematics (dashed) and STAR’s measured chemical (squares) [23]. Right: the fitted () compared with STAR’s chemical ; the near-agreement reflects the conservation of the ratio between the chemical and kinetic surfaces. All error bars are : ours from the fit covariance, including the – correlation, which cancels strongly in the ratio; STAR’s their published combined statistical and systematic uncertainties (Table IX of Ref. [23]), propagated to treating and as uncorrelated. The reference uncertainties dominate the comparison, and their growth relative to the offset at the highest energy is why the deficit falls from below 27 GeV to at 39 GeV.
Two features of the data warrant comment here. First, each published STAR spectrum is a concatenation of two partially overlapping measurements covering different intervals; in the overlap region the same particles are therefore represented twice, and the two sets of points are treated here as independent, which mildly over-weights that region. Second, and more importantly, the sensitivity of the fit to the range exceeds what the covariance errors alone would suggest. Refitting with only the higher- branch of each spectrum raises by 2–5 MeV and changes by at most 4 MeV, within the quoted uncertainties; a fit restricted to the lower- branch alone, however, returns –97 MeV with correspondingly smaller . The latter is not an independent determination: that branch spans only GeV, precisely where resonance feed-down, which the present thermal source does not include, is most important, and where the residuals show a coherent trend rather than random scatter. It does establish, however, that the dominant uncertainty on the absolute scale of , and hence on , is the model description of the soft part of the spectrum rather than the fit statistics. The ratio , fixed by alone, is markedly more robust: across these fit variants it changes by at most 4.4%, against for and separately. This is the quantity the data determine most directly, and the stability of the comparison in Figure 12 should be read accordingly.
The extracted is independent of the assumed (Table VII), as it must be, since the longitudinal boost affects both species identically. In the language of partial chemical equilibrium, the fitted is the net-baryon (antisymmetric) chemical potential on the kinetic surface, while the symmetric over-abundance factors common to particles and antiparticles are absorbed into the species normalizations.
With both and now determined from the same spectra, the trajectory can be evaluated fully self-consistently on the kinetic surface, with uncertainties propagated from the covariance (including their correlation). The result (Figure 13) is a substantially more dilute trajectory than the hybrid evaluation of
Figure 13. (Color online) Fully data-driven kinetic freeze-out trajectory in the plane from the fit, with uncertainties propagated from the covariance, for the three longitudinal velocities.
| (GeV) | (MeV) | (MeV) | (STAR) | ||
| 7.7 | 2.46 | 2.79 | |||
| 11.5 | 1.71 | 1.93 | |||
| 19.6 | 1.05 | 1.23 | |||
| 27 | 0.81 | 0.93 | |||
| 39 | 0.56 | 0.65 |
Table 7. Simultaneous fit at : shared kinetic freeze-out temperature , transverse flow , and the data-driven baryon chemical potential extracted from the ratio, with uncertainties. The last two columns compare the fitted with STAR’s chemical-freeze-out ratio , taken from the strangeness-canonical yield fit (SCEY) of Table IX of Ref. [23]; the grand-canonical yield fit (GCEY, Table VIII of the same reference) differs by (largest at 27 GeV) and would not alter any conclusion drawn here.
Table III: at we find and MeV/ at 7.7 GeV, falling to at 39 GeV, with approximately energy-independent at – MeV/. The net baryon density now decreases monotonically with (the non-monotonicity of the hybrid evaluation, inherited from the statistically insignificant temperature dip at 11.5 GeV, is absent), while the dependence on the longitudinal flow persists through the fitted temperature:
| (GeV) | (fm) | (MeV/fm) | |
| 7.7 | 0.0 | 0.0535 | 120.1 |
| 0.2 | 0.0642 | 143.2 | |
| 0.4 | 0.1187 | 261.9 | |
| 11.5 | 0.0 | 0.0193 | 68.5 |
| 0.2 | 0.0236 | 81.3 | |
| 0.4 | 0.0465 | 147.8 | |
| 19.6 | 0.0 | 0.0216 | 120.3 |
| 0.2 | 0.0267 | 143.2 | |
| 0.4 | 0.0540 | 262.6 | |
| 27 | 0.0 | 0.0149 | 115.8 |
| 0.2 | 0.0186 | 137.9 | |
| 0.4 | 0.0388 | 254.2 | |
| 39 | 0.0 | 0.0115 | 125.4 |
| 0.2 | 0.0144 | 150.0 | |
| 0.4 | 0.0309 | 281.6 |
Table III. Net baryon density and energy density on the kinetic freeze-out surface, for each collision energy and each fixed longitudinal velocity . They are obtained by evaluating the hadron resonance gas, Eqs. (23), (24) and (26), at the fitted temperature of Table I and the Cleymans input of Eq. (27), with and the full Bose–Einstein/Fermi–Dirac series summed to convergence. depends only on , so the dependence enters entirely through . These are the points plotted in Figure 4. This is the hybrid evaluation discussed above; the fully data-driven densities are given in Sec. III D.
at , is larger by a factor of and by a factor of . These numbers characterize the dilute, cold endpoint of the hadronic phase: within the equilibrium-HRG evaluation (which, as discussed above, omits the symmetric partial-chemical-equilibrium over-abundances and therefore provides lower estimates), kinetic decoupling at the lowest BES energies occurs at a net baryon density of only a few percent of normal nuclear density, a factor of below the chemical-surface compression maximum of Ref. [15].
Conclusions and Outlook
In this study we have investigated kinetic freeze-out conditions in Au+Au collisions at –39 GeV using a covariant statistical fireball model [20] that explicitly incorporates transverse and longitudinal collective flow, extending to the kinetic freeze-out surface the mapping introduced by Randrup and Cleymans [15]. Fitting STAR transverse-momentum spectra in 0–5% central collisions within , the analysis proceeds in three stages of increasing constraint: a single-species proton fit and a three-species fit, in both of which the baryon chemical potential must be supplied externally, and finally a four-species fit including the antiproton, in which becomes data-driven and no external chemical input enters the analysis at all.
Our principal findings are as follows. Because a single-species mid-rapidity spectrum determines only the product of the normalization and the fugacity, is not constrained by the fit; we therefore fix it to the Cleymans systematics and fit only the temperature , the transverse flow , and the normalization . That choice is inert: rescaling the fixed by a factor of four leaves the fitted and unchanged to numerical precision, so the temperatures quoted here carry no chemical-freeze-out assumption. Longitudinal flow shifts the extracted temperature systematically upward, the fitted spanning 119–136 MeV for , 122–140 MeV for , and 132–152 MeV for . This is not longitudinal flow hardening the spectra, which it cannot do, but a kinematic degeneracy: the longitudinal component of the fluid four-velocity modifies the Lorentz-invariant factor in the distribution function, and the fit compensates by returning a higher . Temperature and transverse flow are strongly anti-correlated ( at all energies), so they are jointly, not independently, determined.
A simultaneous fit of the , , and proton spectra with a shared temperature and transverse flow sharpens the extraction and tests the kinetic interpretation. The particle-mass ordering breaks the –normalization degeneracy of the single-species fit and tightens the parameter uncertainties by factors of 3.7–5.3 on and 2.2–2.9 on , yielding a lower and more physical kinetic freeze-out temperature –115 MeV at . These agree with STAR’s independently measured blast-wave temperatures at every energy: the differences run from +2.1 to −9.5 MeV and remain within of the combined uncertainties, STAR’s own values carrying ±11–12 MeV. They sit well below the STAR chemical freeze-out temperature, confirming that the present analysis probes the kinetic rather than the chemical freeze-out surface.
Throughout the scanned range the extracted temperatures remain at or below the QCD crossover temperature –160 MeV estimated from lattice QCD, so that the hadron-resonance-gas description underlying the model is internally consistent at every energy considered. Scanning locates the critical value at which would reach (Sec. III B): –0.61 at the lowest energies, falling to –0.47 at 39 GeV in the proton-only fit. The multi-species temperature is both lower and better determined, and it is the corresponding bound that we adopt; longitudinal velocities above that would drive kinetic freeze-out above the crossover band, into a regime where the hadron-resonance-gas picture no longer applies, and are on that basis physically disfavored. A natural refinement would impose as a hard constraint directly in the fit, yielding a posterior on rather than the fixed-grid scan used here.
Including the antiproton spectra with a normalization shared between and removes the last external input: the ratio fixes the baryon chemical potential directly from the data, giving –63 MeV at kinetic freeze-out across the BES range. The extracted agrees with STAR’s chemical at the 85–88% level and is independent of , exactly as expected if the ratio is frozen at the chemical surface and is conserved during the subsequent hadronic cooling. Across all , the freeze-out trajectory in the – plane follows the parametrization of Cleymans et al. [24], with decreasing monotonically with and rising overall, subject to a mild and statistically insignificant non-monotonicity near 11.5 GeV (Sec. III); it lies below both and the Cleymans chemical-freeze-out line, as expected physically since kinetic freeze-out occurs later and cooler than chemical freeze-out.
The resulting fully data-driven trajectory, with uncertainties propagated from the fit covariance, places kinetic decoupling at the lowest BES energies in a dilute system, (with ), a factor of below the chemical-surface compression maximum of Ref. [15]. The net baryon density falls monotonically as rises while remains nearly energy-independent, placing the maximum of net-baryon compression at or below the lower edge of the BES-I range, as found in Ref. [15]. Collective flow raises these densities substantially through the – degeneracy: at , grows by a factor of and by , the corresponding enhancement in the hybrid evaluation—which pairs the fitted with the chemical —being 2.2–2.7. The inferred thermodynamic conditions therefore depend significantly on the assumed collective motion, and the longitudinal flow must be constrained before the compression can be pinned down precisely.
The integrated proton yields are essentially independent of : going from to raises the fitted temperature by about 11% at every energy yet changes the integrated yield by less than 0.02%, because the normalization absorbs the change almost exactly—the same compensation that renders the fit insensitive to the input . Pion production, by contrast, rises steadily with , consistent with the transition from baryon-dominated to meson-dominated particle production as nuclear transparency increases with energy.
Several extensions of the present work are envisaged. In the near term, the most pressing is the inclusion of additional hadron species, in particular strange hadrons such as , , and , in the simultaneous fit. The required measurements already exist at exactly these five energies: the spectra are published in the same data set used here [23], and , , and spectra are available from STAR’s strange-hadron survey [27]. This would allow the strangeness chemical potential and the electric charge chemical potential , both set to zero here, to be self-consistently determined, following the same ratio-based strategy used for in Sec. III D; indeed Ref. [27] already extracts and at freeze-out from antibaryon-to-baryon ratios of precisely these species. Correcting the and spectra for weak
feed-down would likewise remove the dominant known bias in the extracted . The latter is expected to be non-negligible at lower energies, where baryon stopping and the neutron-to-proton imbalance of the Au nucleus introduce an appreciable isospin asymmetry. A global multi-species fit would also test the assumption of a single common freeze-out surface, which may break down if different species decouple at different times.
Both flow velocities enter the present description as single constants rather than as profiles: is a single surface-like transverse velocity, as the comparison with STAR’s profile-averaged in Sec. III C makes explicit, and is the magnitude of the longitudinal flow velocity of the emitting elements. Only that magnitude enters: because cosh is even and the acceptance window is symmetric about , the rapidity-integrated spectrum of Eq. (18) is an exact even function of , so a source element moving forward and its backward-moving mirror contribute identically and no forward–backward asymmetry is implied. Extending the analysis to wider rapidity windows would accordingly be valuable, though what such data would constrain is not the single magnitude but the underlying distribution of flow rapidities that it stands in for. Within the two are indistinguishable, since may be replaced by its value at throughout the window, precisely the origin of the – degeneracy above, whereas a boost-invariant continuum of flow rapidities spanning predicts a markedly different shape. Rapidity-differential spectra, together with net-baryon rapidity distributions, would discriminate between these descriptions and so replace the effective longitudinal parameter used here by a measured flow profile.
Finally, the present approach captures the freeze-out conditions through a parametrized velocity field on a single surface element. The emission itself is already of Cooper–Frye form, as noted below Eq. (8), and would remain so in a more complete treatment: what a full dimensional viscous hydrodynamic evolution with a baryon-density-dependent equation of state (EoS) supplies is not a different hadronisation prescription but the freeze-out hypersurface and the flow field upon it, determined dynamically by the conservation laws rather than imposed, together with the shear and bulk corrections to the equilibrium distribution that the present emission function omits. The parameters of Table I would then be replaced by their local counterparts , and , and the values quoted here read as emission-weighted averages over that surface. In such a framework the EoS, and with it the adiabatic sound speed at the finite net-baryon densities relevant here, is an input rather than an output: it is supplied to the evolution, which then propagates the pressure gradients that build the collective flow. Its value would therefore be confronted with the measured spectra and freeze-out conditions rather than extracted from them.
The ingredients for such a programme are already in place: lattice-matched equations of state at finite net-baryon density [28], families of equations of state carrying a critical point in the three-dimensional Ising universality class mapped onto them [29], and dimensional viscous simulations with explicit baryon transport applied across the BES energy range [30], assembled together with critical fluctuations and freeze-out prescriptions into a common framework [31]. What remains open is not whether such calculations can be carried out but where or whether the critical point lies, since its location is not fixed by the lattice and enters as a parameter of the constructed equation of state. The coordinates extracted here are complementary in a specific sense: they carry no hydrodynamic input, so the kinetic freeze-out surface of Table VII, and its -fold displacement below the chemical-surface compression maximum, are benchmarks that such calculations should reproduce at their own de-coupling surface independently of the equation of state assumed at earlier times.
ACKNOWLEDGMENTS
V.R. acknowledges financial support from Anusandhan National Research Foundation (ANRF), India through Core Research Grant , CRG/2023/001309.
Covariant Statistical Mechanics and the Four-Temperature
The covariant formulation of statistical mechanics was developed to resolve a fundamental ambiguity, how does temperature transform under a Lorentz boost, i.e., what is the temperature of a body in motion in Special Relativity? Two early proposals gave contradictory answers.
If one uses the ideal gas law, , to define temperature, and notes that volume contracts as under a boost while the equation of state holds in every inertial frame, one obtains
where is the rest-frame temperature and .
Alternatively, from the second law of thermodynamics, , the entropy counts microstates and is a Lorentz scalar, while transforms as the time-component of the heat four-vector, giving . Demanding that remain invariant then yields
The resolution of this apparent paradox was addressed in a more systematic way by Israel and van Kampen using a fully covariant formulation by replacing the scalar inverse temperature with the four-vector
where is the fluid four-velocity satisfying . The Gibbs ensemble for a relativistic gas is then constructed by maximizing the number of microstates subject to four-momentum conservation and particle number conservation . Introducing Lagrange multipliers and for these constraints respectively, the stationarity condition yields
from which the Boltzmann factor follows immediately
The partition function is a Lorentz scalar by construction, since is a scalar contraction. The Lorentz-invariant phase-space factor appearing throughout the covariant fireball model (see § 1.1) arises directly from this structure with the identification (A3).
References
References
- [1]Hitansh Shah, Mauricio Hippert, Jorge Noronha, Claudia Ratti, and Volodymyr Vovchenko. Locating the QCD critical point through contours of constant entropy density. Phys. Rev. C, 113:L012201, Jan 2026.
- [2]Agnieszka Sorensen and Paul Sorensen. Locating the critical point for the hadron to quark-gluon plasma phase transition from finite-size scaling of proton cumulants in heavy-ion collisions. arXiv:2405.10278 [nucl-th], 2024.
- [3]Gökçe Başar. QCD critical point, Lee-Yang edge singularities, and Padé resummations. Phys. Rev. C, 110:015203, Jul 2024.
- [4]Roy A. Lacey. Indications for a critical end point in the phase diagram for hot and dense nuclear matter. Phys. Rev. Lett., 114:123201, Apr 2015.
- [5]D. A. Clarke, P. Dimopoulos, F. Di Renzo, J. Goswami, C. Schmidt, S. Singh, and K. Zambello. Searching for the QCD critical end point using multipoint Padé approximations. Phys. Rev. D, 112:L091504, Nov 2025.
- [6]Kazuyuki Kanaya, Ryo Ashikawa, Shinji Ejiri, Masakiyo Kitazawa, Hiroshi Suzuki, and Naoki Wakabayashi. Phase structure and critical point in heavy-quark QCD at finite temperature. arXiv:2211.08631 [hep-lat], 2022. Proceedings of the 39th International Symposium on Lattice Field Theory (Lattice 2022).arxiv.org/abs/2211.08631
- [7]Y. Aoki, Z. Fodor, S. D. Katz, and K. K. Szabó. The QCD transition temperature: Results with physical masses in the continuum limit. Physics Letters B, 643(1):46–54, 2006.DOI
- [8]Anyi Li, Andrei Alexandru, and Keh-Fei Liu. Critical point of Nf=3 QCD from lattice simulations in the canonical ensemble. Phys. Rev. D, 84:071503, Oct 2011.DOI
- [9]Szabolcs Borsányi, Zoltán Fodor, Jana N. Guenther, Paolo Parotto, Attila Pásztor, Claudia Ratti, Volodymyr Vovchenko, and Chik Him Wong. Lattice QCD constraints on the critical point from an improved precision equation of state. Phys. Rev. D, 112:L115105, Dec 2025.
- [10]Adiba Shaikh, Ranjita K. Mohapatra, and Saumen Datta. QCD phase diagram and the finite volume fireball: A model study. Nuclear Physics A, 1054:122981, 2025.DOI
- [11]Gergely Endrődi. Critical point in the QCD phase diagram for extremely strong background magnetic fields. JHEP, 07:173, 2015.arxiv.org/abs/1504.08280
- [12]Eduardo S. Fraga, Letícia F. Palhares, and Paul Sorensen. Finite-size scaling as a tool in the search for the QCD critical point in heavy ion data. Phys. Rev. C, 84:011903, Jul 2011.DOI
- [13]S. Acharya et al. Closing in on critical net-baryon fluctuations at LHC energies: Cumulants up to third order in Pb–Pb collisions. Physics Letters B, 844:137545, 2023.DOI
- [14]B. E. Aboona et al. Precision measurement of net-proton-number fluctuations in Au+Au collisions at RHIC. Phys. Rev. Lett., 135:142301, Sep 2025.
- [15]J. Randrup and J. Cleymans. Maximum freeze-out baryon density in nuclear collisions. Phys. Rev. C, 74:047901, Oct 2006.DOI
- [16]A. Andronic, P. Braun-Munzinger, K. Redlich, and J. Stachel. Decoding the phase structure of QCD via particle production at high energy. Nature, 561:321–330, 2018.
- [17]A. Bazavov, T. Bhattacharya, C. DeTar, H.-T. Ding, S. Gottlieb, R. Gupta, P. Hegde, U. M. Heller, F. Karsch, S. Mukherjee, P. Petreczky, C. Schmidt, R. A. Soltz, W. Soeldner, R. Sugar, M. Wagner, P. Vranas, et al. Equation of state in (2+1)-flavor QCD. Phys. Rev. D, 90:094503, 2014.DOI
- [18]M. S. Abdallah et al. Measurements of proton high-order cumulants in √sNN = 3 GeV Au + Au collisions and implications for the QCD critical point. Phys. Rev. Lett., 128:202303, May 2022.DOI
- [19]Travis Dore, Jamie M. Karthein, Isaac Long, Debora Mroczek, Jacquelyn Noronha-Hostler, Paolo Parotto, Claudia Ratti, and Yukari Yamauchi. Critical lensing and kurtosis near a critical point in the QCD phase diagram in and out of equilibrium. Phys. Rev. D, 106:094024, Nov 2022.DOI
- [20]B. Touschek. Covariant statistical mechanics. Nuovo Cimento B, 58:295–307, November 1968.
- [21]Fred Cooper and Graham Frye. Single-particle distribution in the hydrodynamic and statistical thermodynamic models of multiparticle production. Phys. Rev. D, 10:186, 1974.DOI
- [22]Ekkard Schnedermann, Josef Sollfrank, and Ulrich Heinz. Thermal phenomenology of hadrons from 200A GeV S+S collisions. Phys. Rev. C, 48:2462, 1993.pubmed.ncbi.nlm.nih.gov/9969103
- [23]L Adamczyk et al. Bulk properties of the medium pro duced in relativistic heavy-ion collisions from the beam energy scan program. Phys. Rev. C, 96:044904, Oct 2017.
- [24]J. Cleymans, H. Oeschler, K. Redlich, and S. Wheaton. Comparison of chemical freeze-out criteria in heavy-ion collisions. Phys. Rev. C, 73:034905, Mar 2006.
- [25]H. Bebie, P. Gerber, J. L. Goity, and H. Leutwyler. The role of the entropy in an expanding hadronic gas. Nucl. Phys. B, 378:95, 1992.DOI
- [26]Tetsufumi Hirano and Keiichi Tsuda. Collective flow and two-pion correlations from a relativistic hydrodynamic model with early chemical freeze-out. Phys. Rev. C, 66:054905, 2002.DOI
- [27]J. Adam et al. Strange hadron production in au+au collisions at √sNN = 7.7, 11.5, 19.6, 27, and 39 gev. Phys. Rev. C, 102:034909, Sep 2020.DOI
- [28]Akihiko Monnai, Björn Schenke, and Chun Shen. Equation of state at finite densities for QCD matter in nuclear collisions. Phys. Rev. C, 100:024907, 2019.DOI
- [29]Paolo Parotto, Marcus Bluhm, Debora Mroczek, Marlene Nahrgang, Jacqulyn Noronha-Hostler, Krishna Rajagopal, Claudia Ratti, Thomas Schäfer, and Mikhail Stephanov. QCD equation of state matched to lattice data and exhibiting a critical point singularity. Phys. Rev. C, 101:034901, 2020.DOI
- [30]Chun Shen and Björn Schenke. Dynamical initial state model for relativistic heavy-ion collisions. Phys. Rev. C, 97:024907, 2018.DOI
- [31]Xin An, Marcus Bluhm, Lipei Du, Gerald V. Dunne, Hannah Elfner, Charles Gale, Joaquin Grefa, Ulrich Heinz, Anping Huang, Jamie M. Karthein, Dmitri E. Kharzeev, Volker Koch, Jinfeng Liao, Shiyong Li, Mauricio Martinez, Michael McNelis, Debora Mroczek, Swagato Mukherjee, Marlene Nahrgang, Angel R. Nava Acuna, Jacquelyn Noronha-Hostler, Dmytro Oliinychenko, Paolo Parotto, Israel Portillo, Maneesha Sushama Pradeep, Scott Pratt, Krishna Rajagopal, Claudia Ratti, Gregory Ridgway, Thomas Schaefer, Björn Schenke, Chun Shen, Shuzhe Shi, Mayank Singh, Vladimir Skokov, Dam T. Son, Agnieszka Sorensen, Mikhail Stephanov, Raju Venugopalan, Volodymyr Vovchenko, Ryan Weller, Ho-Ung Yee, and Yi Yin. The BEST framework for the search for the QCD critical point and the chiral magnetic effect. Nucl. Phys. A, 1017:122343, 2022.DOI