betaFeedback

WorksPhysics and Astronomy

Unveiling QCD Criticality with Cross-Rapidity Net-Baryon Cumulants

Abstract

Fluctuations of conserved charges are a primary tool in the search for the QCD critical endpoint, but their beam-energy dependence is complicated by global baryon-number conservation, whose influence changes as the experimental acceptance covers different fractions of the collision system. We propose using correlations of net-baryon fluctuations between two separated rapidity windows to exploit the distinct signatures of conservation and critical dynamics. Global conservation produces a negative cross-window correlation, whereas a common long-wavelength critical fluctuation produces a positive one. We show that, within a canonical independent-source framework, the conservation-induced background and the critical signal enter additively at leading order, allowing the leading conservation term to be estimated and subtracted. The resulting yield-scaled correlator follows the nonmonotonic enhancement of an Ising-mapped equilibrium correlation length along a freeze-out trajectory passing near a hypothetical critical endpoint, while wider rapidity windows reduce the response through thermal smearing. Cross-rapidity cumulants therefore provide a rapidity-differential strategy for reducing the leading conservation background and sharpening fluctuation-based searches for QCD criticality in beam-energy-scan experiments.

Introduction.—The search for the QCD critical endpoint (CEP) is a central goal of the heavy-ion beam-energy-scan programs [1, 2, 3, 4]. At small baryon chemical potential (μB\mu_B), lattice-QCD calculations establish that the transition from hadronic matter to the quark-gluon plasma (QGP) is a smooth crossover [5], while at larger μB\mu_B it may become first order and terminate at a critical endpoint [6]. Near the CEP, the growth of the correlation length is expected to enhance event-by-event fluctuations of conserved charges [7]. Measurements of net-proton cumulants as functions of the collision energy (sNN\sqrt{s_{NN}}) therefore provide one of the principal experimental approaches to the CEP search [8, 9, 10]. The RHIC Beam Energy Scan (BES) has reported nontrivial beam-energy dependence in higher-order net-proton cumulants, including nonmonotonic behavior in BES-I [11]. Recent BES-II measurements substantially improve the statistical precision and show significant departures from noncritical baselines near sNN=20\sqrt{s_{NN}} = 20 GeV [12].

The interpretation of this beam-energy dependence remains challenging because the measured cumulants reflect not only the thermodynamic fluctuations of the medium, but also the changing relation between the experimental acceptance and the full collision system. Current analyses employ comparable, though not strictly identical, midrapidity acceptance windows over a broad range of collision energies [13]. As sNN\sqrt{s_{NN}} decreases, however, the beam rapidity and the longitudinal extent of the baryon distribution decrease [14]. A similar laboratory rapidity interval therefore contains an increasingly large fraction of the total net-baryon number at lower collision energies. Global net-baryon conservation consequently becomes more important, suppressing fluctuations within the measured acceptance and generating correlations with the unobserved part of the system [15, 16, 17, 18]. Thus, even in the absence of critical dynamics, the conservation contribution itself changes across the beam-energy scan and can produce a nontrivial energy dependence in the measured cumulants. This effect is entangled with finite acceptance corrections [19, 20], event-by-event volume fluctuations [21, 22], variations of the baryon-stopping profile [14, 23], and the subsequent hadronic evolution [24, 25, 26, 27]. Separating these changing noncritical contributions from a possible critical signal is therefore essential for interpreting the measured beam-energy dependence.

In this Letter, we investigate connected fluctuations of net-baryon number measured in two nonoverlapping rapidity windows. Unlike cumulants constructed within a single acceptance, cross-rapidity cumulants retain information on how fluctuations are correlated across different regions of the final-state baryon distribution. Within a canonical independent-source framework coupled to a fluctuating critical mode, we show that the negative correlation from global net-baryon conservation and the positive contribution from a common long-wavelength critical fluctuation enter as separate additive terms at the leading order considered. This structure allows the leading conservation contribution to be estimated and motivates a conservation-subtracted, yield-scaled correlator that remains sensitive to the equilibrium critical correlation length. Rather than relying only on the beam-energy dependence of a single-window cumulant, the proposed approach uses the rapidity structure of fluctuations to distinguish a global constraint from a common long-wavelength mode.

Cross-rapidity cumulants.—In heavy-ion collisions, let B1B_1 and B2B_2 denote the event-by-event net-baryon numbers measured in two nonoverlapping rapidity windows, labeled A\mathcal{A} and B\mathcal{B} in Figure 1. We use ⟨⟨⋯ ⟩⟩\langle\langle\cdots\rangle\rangle to denote an average over events at fixed collision energy and centrality. The connected correlations between the two windows are characterized by the joint cumulants [28]

Image not extracted

Figure 1. Schematic cross-rapidity correlations between two nonoverlapping acceptance windows, A\mathcal{A} and B\mathcal{B}. Red and blue denote positive and negative deviations from the event-averaged baseline. (a) Global net-baryon conservation produces an anticorrelation between the two windows. (b) A common critical fluctuation can modify both yields in the same direction and generate a positive correlation. (c) In a heavy-ion collision, critical fluctuations coexist with the redistribution imposed by global conservation. The curves are illustrative and do not represent quantitative rapidity profiles.

κn,m≡∂n+m∂νn∂λmln⁡⟨⟨eνB1+λB2⟩⟩∣ν=λ=0(1)\kappa_{n,m} \equiv\left.\frac{\partial^{n+m}}{\partial\nu^n \partial\lambda^m}\ln\left\langle\left\langle e^{\nu B_1+\lambda B_2}\right\rangle\right\rangle\right|_{\nu=\lambda=0} \tag*{(1)}

This family retains information on correlations across different regions of the final-state rapidity distribution that is not contained in cumulants measured within a single acceptance window. We focus on its lowest but nontrivial member,

κ1,1=⟨⟨δB1 δB2⟩⟩,δBi≡Bi−⟨⟨Bi⟩⟩,(2)\kappa_{1,1}=\langle\langle\delta B_1\,\delta B_2\rangle\rangle,\qquad\delta B_i\equiv B_i-\langle\langle B_i\rangle\rangle, \tag*{(2)}

which is the covariance of the net-baryon yields in the two windows. It provides the most direct sign-level distinction between the mechanisms considered below and is experimentally more accessible than higher-order cross cumulants. In practice, the corresponding net-proton quantity can be constructed from event-by-event yields within specified rapidity and transverse-momentum cuts; we formulate the argument first for net baryon number, for which the global conservation law is exact.

Global net-baryon conservation produces anticorrelation between disjoint acceptances. An upward fluctuation in one window must be balanced by a depletion in the complementary phase space and therefore tends to reduce the baryon content available to the other window. This generates a negative contribution to κ1,1\kappa_{1,1} even in the absence of critical dynamics, as illustrated in Figure 1(a). The leading-order critical fluctuation produces a qualitatively different response. Near the critical region, a long-wavelength fluctuation of the scalar order-parameter field modifies the local baryon density and, through freeze-out emission, can affect particles observed in both rapidity windows. When the two windows respond with the same sign, their net-baryon yields fluctuate together, producing a positive contribution to κ1,1\kappa_{1,1}, as illustrated in Figure 1(b).

In an actual heavy-ion collision, both mechanisms act simultaneously, as illustrated in Figure 1(c). The measured covariance therefore contains the negative redistribution imposed by global conservation together with any positive correlation generated by a common long-wavelength fluctuation. Their contrasting signs motivate a quantitative separation of the two contributions. Cross-rapidity correlations thus retain information on the organization of fluctuations across rapidity that is not available from a single-window cumulant alone.

Conservation and critical contributions.—We now determine how global conservation and critical fluctuations

enter the measured cross-rapidity covariance. We first specify the critical degree of freedom that modulates baryon emission and then combine its event-by-event fluctuations with exact global net-baryon conservation.

Near the critical region, baryons couple to the scalar order-parameter field σ\sigma through the mass shift mhσ=mh+Ghσm_h^\sigma=m_h+G_h\sigma, where mhm_h and GhG_h denote the vacuum mass and coupling to σ\sigma for baryon species hh, respectively. The field fluctuates throughout the fireball, including along the longitudinal direction. Its equilibrium distribution is taken as P[σ]∝exp⁡(−E[σ]/T)P[\sigma]\propto\exp(-\mathcal{E}[\sigma]/T). Expanding around the spatial average σ‾=V−1∫d3x σ(x)\overline{\sigma}=V^{-1}\int d^3x\,\sigma(x), where VV is the volume of freeze-out element on the fireball, and neglecting thermodynamic gradients over a correlation domain gives the quadratic effective Hamiltonian [8, 10]

E[σ]=∫d3x(12(∇σ)2+mσ22(σ−σ‾)2),(3)\mathcal{E}[\sigma]=\int d^3x\left(\frac{1}{2}(\nabla\sigma)^2+\frac{m_\sigma^2}{2}(\sigma-\overline{\sigma})^2\right), \tag*{(3)}

where mσ=ξeq−1m_\sigma=\xi_{\rm eq}^{-1} the inverse of equilibrium correlation length. The softening of mσm_\sigma near the critical endpoint increases ξeq\xi_{\rm eq} and enhances long-wavelength fluctuations capable of correlating baryon production in different rapidity regions.

Having specified the critical-field fluctuations, we combine them with the conservation constraint through two nested statistical averages. The fireball represented by the thermal sources is treated as a closed system whose total net-baryon number BB, deposited by the participants, is fixed; event-by-event fluctuations of BB are not included. At the inner step, for each fixed-σ\sigma field configuration, ⟨⋅⟩ce[σ]\langle\cdot\rangle_{\mathrm{ce}[\sigma]} denotes the canonical average over the baryon and antibaryon degrees of freedom. The net-baryon numbers B1B_1 and B2B_2 in the two selected rapidity windows remain free to fluctuate subject to B1+B2+B3=BB_1+B_2+B_3=B, where B3B_3 is the net-baryon number in all momentum-space regions outside the two windows. The baryon and antibaryon multiplicities are not separately fixed; only their difference is constrained. At the outer step, the resulting field-dependent average is further averaged over the event-by-event configurations of σ\sigma. Thus, for any baryonic observable O^\hat{O}, ⟨O^⟩ce=⟨⟨O^⟩ce[σ]⟩σ\langle\hat{O}\rangle_{\mathrm{ce}}=\langle\langle\hat{O}\rangle_{\mathrm{ce}[\sigma]}\rangle_\sigma, where ⟨⋅⟩σ\langle\cdot\rangle_\sigma denotes the average with P[σ]P[\sigma]. The corresponding functional definitions are given in the Supplemental Material.

Applying the law of total covariance to this nested average yields an exact decomposition of the measured cross-rapidity covariance. With B^i\hat{B}_i denoting the net-baryon-number operator in window ii, we define Bi(σ)≡⟨B^i⟩ce[σ]B_i(\sigma)\equiv\langle\hat{B}_i\rangle_{\mathrm{ce}[\sigma]} as conditional mean at fixed σ\sigma and κ1,1ce[σ]≡⟨(B^1−B1(σ))(B^2−B2(σ))⟩ce[σ]\kappa^{\mathrm{ce}[\sigma]}_{1,1}\equiv\langle(\hat{B}_1-B_1(\sigma))(\hat{B}_2-B_2(\sigma))\rangle_{\mathrm{ce}[\sigma]} as the corresponding conditional covariance. The total covariance is then

κ1,1=⟨κ1,1ce[σ]⟩σ+C1,1σ,(4)\kappa_{1,1}=\left\langle\kappa^{\mathrm{ce}[\sigma]}_{1,1}\right\rangle_\sigma+C^\sigma_{1,1}, \tag*{(4)}

where C1,1σ≡⟨B1(σ)B2(σ)⟩σ−⟨B1(σ)⟩σ⟨B2(σ)⟩σC^\sigma_{1,1}\equiv\langle B_1(\sigma)B_2(\sigma)\rangle_\sigma-\langle B_1(\sigma)\rangle_\sigma\langle B_2(\sigma)\rangle_\sigma is the covariance of the field-dependent mean yields. Equation (4) is an exact statistical identity: the first term describes fluctuations at fixed σ\sigma, while the second arises from variations of the mean yields between field configurations. Identifying them with a conservation baseline and a leading critical contribution, respectively, requires the model approximations introduced below.

To evaluate the two terms, we adopt the continuous-source framework [13, 17]. Locally thermalized sources are continuously distributed along the longitudinal direction of the fireball. At fixed σ\sigma, each source emits independent Poissonian baryon and antibaryon multiplicities before the global canonical constraint is imposed; the constraint then correlate different rapidity regions. The local contribution per source-volume element at spacetime rapidity ysy_s to the baryon yield observed in window ii is

xˉi,s[σ]=∑hdh∫d3p(2π)3us⋅pp0eμs−us⋅pTsΘ(y∈yi),(5)\bar{x}_{i,s}[\sigma]=\sum_h d_h\int\frac{\mathrm{d}^3p}{(2\pi)^3}\frac{u_s\cdot p}{p^0}e^{\frac{\mu_s-u_s\cdot p}{T_s}}\Theta(y\in y_i), \tag*{(5)}

where dhd_h is the degeneracy of species hh; usμu_s^\mu, TsT_s, and μs\mu_s are the flow velocity, temperature, and baryon chemical potential; and pμp^\mu and yy are the emitted hadron energy-momentum and momentum rapidity. The momentum integration together with the restriction Θ(y∈yi)\Theta(y\in y_i) accounts for thermal smearing between source and momentum rapidity. The local σ\sigma dependence enters through mhσm_h^\sigma in the on-shell momentum. We approximate the spacetime rapidity by the longitudinal flow rapidity. The antibaryon contribution zˉi,s\bar{z}_{i,s} follows from μs→−μs\mu_s\to-\mu_s.

We first evaluate the covariance at fixed field. Let Ni=∫dVs xˉi,s[0]N_i=\int\mathrm{d}V_s\,\bar{x}_{i,s}[0] and Nˉi=∫dVs zˉi,s[0]\bar{N}_i=\int\mathrm{d}V_s\,\bar{z}_{i,s}[0] denote the source-integrated mean baryon and antibaryon yields in window ii, respectively, with dVs\mathrm{d}V_s denoting the source-volume element. We denote their full-phase-space counterparts by NN and Nˉ\bar{N}. As shown in the Supplemental Material, field-dependent corrections to the fixed-field covariance are suppressed by the weak mass dressing and, for the fluctuation-induced terms, by the large full-phase-space multiplicities. Neglecting the small mean-field near the CEP, σˉ≃0\bar{\sigma}\simeq0 [29], we therefore evaluate the covariance at the unshifted configuration σ=0\sigma=0 and obtain

⟨κ1,1ce[σ]⟩σ≃κ1,1ce[0]=−N1N2N−Nˉ1Nˉ2Nˉ.(6)\left\langle\kappa^{\mathrm{ce}}_{1,1}[\sigma]\right\rangle_\sigma\simeq\kappa^{\mathrm{ce}}_{1,1}[0]=-\frac{N_1N_2}{N}-\frac{\bar{N}_1\bar{N}_2}{\bar{N}}. \tag*{(6)}

The canonical constraint acts event-by-event on the full-phase-space multiplicities, fixing the net baryon number to BB in each event, whereas NN and Nˉ\bar{N} in Eq. (6) are the corresponding source-integrated mean yields. The negative sign represents the anticorrelation generated by distributing a fixed total net-baryon number among disjoint acceptances.

We next evaluate the covariance generated by fluctuations of the critical field. Define the response of window ii to a local change of σ\sigma at source rapidity ysy_s as χi(ys)≡∂(xˉi,s[σ]−zˉi,s[σ])/∂σ∣σ=0\chi_i(y_s)\equiv\left.\partial(\bar{x}_{i,s}[\sigma]-\bar{z}_{i,s}[\sigma])/\partial\sigma\right|_{\sigma=0}. Writing δσ=σ−σˉ\delta\sigma=\sigma-\bar{\sigma}, the field-dependent variation of the conditional mean is, to linear order, Bi(σ)−⟨Bi(σ)⟩σ≃∫dVs χi(ys)δσ(xs)B_i(\sigma)-\langle B_i(\sigma)\rangle_\sigma\simeq\int\mathrm{d}V_s\,\chi_i(y_s)\delta\sigma(x_s), where xsx_s is the position of source ss. For thermodynamic conditions that vary slowly over a correlation domain, the quadratic Hamiltonian gives

Gσ(x,y)≡⟨δσ(x)δσ(y)⟩σ=T4π∣x−y∣e−∣x−y∣ξeq.(7)\mathcal{G}_\sigma(x,y)\equiv\langle\delta\sigma(x)\delta\sigma(y)\rangle_\sigma=\frac{T}{4\pi\lvert x-y\rvert}e^{-\frac{\lvert x-y\rvert}{\xi_{\mathrm{eq}}}}. \tag*{(7)}

The leading field-induced covariance is therefore

C1,1σ=∫dVs1dVs2 χ1(ys1)Gσ(xs1,xs2)χ2(ys2).(8)C^\sigma_{1,1}=\int\mathrm{d}V_{s_1}\mathrm{d}V_{s_2}\,\chi_1(y_{s_1})\mathcal{G}_\sigma(x_{s_1},x_{s_2})\chi_2(y_{s_2}). \tag*{(8)}

The correlator is understood as an equal-time correlator evaluated in an appropriate common rest frame for each source pair; its implementation on the freeze-out hypersurface is specified in the Supplemental Material. Equation (8) has the form of a linear-response relation: the critical field correlates two source regions, and each rapidity window converts the local field variation into a change in its measured net-baryon yield. When the two windows respond with the same sign, C1,1σC^\sigma_{1,1} is positive.

Equation (4) is exact, whereas Eqs. (6) and (8) are leading-order model results. They rely on independent Poisson emission at fixed σ\sigma and on linear response, with the mean and fluctuating mass shifts gσˉg\bar{\sigma} and gδσg\delta\sigma small compared with mhm_h and TsT_s. These approximations apply in the critical region surrounding, but not exactly at, the CEP. The quadratic field theory captures the two-point correlation relevant here, while nonlinear critical interactions can modify its magnitude and become essential for higher-order cumulants. Additional correlations from participant and volume fluctuations, event-by-event baryon stopping, resonance decays, and hadronic transport are not included and may remain after subtraction of the leading conservation contribution.

Conservation-subtracted correlator.— The opposite signs of the conservation and critical contributions provide the central leverage of the cross-rapidity covariance. Since the leading conservation-induced anticorrelation is fixed, within the present model, by the mean baryon and antibaryon yields as in Eq. (6), it can be subtracted to expose correlations beyond this baseline. We further normalize by the mean net-baryon yields to remove the leading trivial dependence on the particle abundance and acceptance. This motivates

S1,1≡1B1B2(κ1,1−κ1,1ce[0]),(9)S_{1,1} \equiv\frac{1}{B_1 B_2}\left(\kappa_{1,1}-\kappa_{1,1}^{\mathrm{ce}}[0]\right), \tag*{(9)}

where Bi≡Ni−N‾iB_i \equiv N_i-\overline{N}_i is the mean net-baryon yield in window ii. Within the present framework and to the order retained, S1,1≃C1,1/(B1B2)S_{1,1} \simeq C_{1,1}/(B_1B_2). More generally, S1,1S_{1,1} is the residual cross-window correlation after subtraction of the leading modeled conservation contribution, rather than an exclusively critical quantity.

Experimentally, κ1,1\kappa_{1,1}, NiN_i, and N‾i\overline{N}_i are constructed within the selected acceptance windows, whereas the full-phase-space yields NN and N‾\overline{N} require data-constrained extrapolation, assisted by longitudinal hydrodynamic or transport calculations. Application to measured net protons further requires a controlled mapping from the conserved net-baryon quantity, including unobserved neutrons, isospin randomization, and decay feeddown [24, 25, 26, 27, 30].

To expose the correlation-length dependence analytically, we consider a constant-τ\tau cylindrical freeze-out hypersurface of radius RR, with a midrapidity plateau of constant TT and μB\mu_B and with spacetime rapidity approximated by the longitudinal flow rapidity. For adjacent narrow windows y1∈[−w,0)y_1 \in[-w,0) and y2∈[0,w)y_2 \in[0,w), the equal-time reduction and the geometric and thermal integrations detailed in the Supplemental Material give

S1,1≃πGp2mp28RK(T)T(ξeq2R2M1(2Rξeq)+ξeqR),(10)S_{1,1} \simeq\frac{\pi G_p^2 m_p^2}{8R}\frac{\mathcal{K}(T)}{T}\left(\frac{\xi_{\mathrm{eq}}^2}{R^2}M_1\left(\frac{2R}{\xi_{\mathrm{eq}}}\right)+\frac{\xi_{\mathrm{eq}}}{R}\right), \tag*{(10)}

where M1M_1 is the modified Struve function of the second kind [31], and GpG_p denotes the proton–σ\sigma coupling. After factoring out GpG_p, K(T)\mathcal{K}(T) contains the thermal species sums and the adopted relative baryon–σ\sigma couplings, as detailed in the Supplemental Material.

Illustrative critical response.— We now illustrate the equilibrium critical response predicted by the present framework. We adopt a hypothetical CEP at (μc,Tc)=(0.635,0.107) GeV(\mu_c,T_c)=(0.635,0.107)\ {\rm GeV}, motivated by the functional renormalization group calculation [32], and obtain ξeq\xi_{\mathrm{eq}} by mapping the universal three-dimensional Ising variables onto

the (T,μB)(T,\mu_B) plane following Ref. [33]. We take Gp=3.2G_p=3.2, motivated by the parity-doublet estimate near the chiral-restoration region [29]. The response is evaluated along the midrapidity chemical freeze-out trajectory of Refs. [34, 35], using R=6.38R=6.38 fm [36] and the adjacent-window configuration introduced above. Details of the Ising-to-QCD mapping and of K(T)\mathcal{K}(T) are given in the Supplemental Material. To examine the dependence on the rapidity acceptance, we consider w=0.01w=0.01, 0.10.1, and 0.50.5 in Figure 2. Along the adopted trajectory, ξeq\xi_{\mathrm{eq}} reaches its maximum near μB≃0.606\mu_B\simeq0.606 GeV, and the correlator develops a corresponding nonmonotonic enhancement. The peak occurs near the maximum of ξeq\xi_{\mathrm{eq}}, showing that the smooth thermal and geometric prefactors in Eq. (10) do not erase the underlying critical response.

Figure 2. (a) Conservation-subtracted correlator S1,1S_{1,1} along the adopted freeze-out trajectory for adjacent midrapidity windows with w=0.01w=0.01, 0.10.1, and 0.50.5. The gray dotted curve shows ξeq/ξ0\xi_{\mathrm{eq}}/\xi_0, where ξ0\xi_0 is the noncritical baseline correlation length; its maximum at μB=0.606\mu_B=0.606 GeV is marked by the vertical dashed line. (b) Freeze-out trajectory, hypothetical CEP, and boundary ξeq=ξ0\xi_{\mathrm{eq}}=\xi_0 of the mapped critical region.

The results for w=0.01w=0.01 and 0.10.1 are nearly identical, indicating that the narrow-window approximation remains effective for w=0.1w=0.1 in this setup. For w=0.5w=0.5, each momentum-rapidity window samples a broader range of source rapidities, whereas only the overlap of locally correlated source regions contributes strongly to the critical covariance. This overlap grows more slowly than the normalization B1B2B_1B_2, reducing the scaled response. Narrower windows therefore preserve a larger normalized critical signal, although experimental optimization must balance this against reduced particle statistics.

This figure is an equilibrium response study rather than a quantitative beam-energy prediction. The position and magnitude of the enhancement depend on the assumed 3-dimensional Ising mapping and freeze-out trajectory and may be modified by nonequilibrium critical evolution [37, 38, 39, 40, 41, 42, 43]. Interpreting an experimental residual would additionally require noncritical baselines and a controlled mapping from net baryons to measured net protons. Within the present setup, the result demonstrates that the leading-order conservation-subtracted correlator retains a pronounced response to an enhanced long-wavelength critical mode.

**Summary and outlook.—**Global conservation and criticality leave qualitatively different imprints on correlations across rapidity. Distributing a fixed net-baryon number among separated acceptances generates an anticorrelation, whereas a common long-wavelength critical mode shifts the mean yields in the two windows coherently and produces a positive correlation. By first evaluating baryon fluctuations at a fixed critical-field configuration and then averaging over the field fluctuations, we obtained an exact decomposition of the second-order cross-rapidity cumulant. Within the canonical independent-source model, this decomposition becomes, at leading order, the separation between the negative conservation baseline and the positive critical contribution. This sign contrast is the central physical advantage of the cross-rapidity observable.

Motivated by this structure, we introduced the conservation-subtracted correlator S1,1S_{1,1} in Eq. (9). Dividing by the mean net-baryon yields removes the leading dependence on the fireball lifetime, acceptance width, and net-baryon density, while preserving the response to the critical correlation length. In an equilibrium proof-of-principle calculation, the correlator develops a nonmonotonic enhancement along a freeze-out trajectory passing near a hypothetical critical endpoint, closely following the maximum of the input ξeq\xi_{\mathrm{eq}}. Narrow rapidity windows retain a larger normalized response, whereas thermal smearing over broad windows dilutes the overlap of the correlated source regions.

The proposed observable should therefore be viewed not as a standalone critical signature, but as a rapidity-differential diagnostic with several experimentally testable dimensions. Measurements versus beam energy, window width, and window separation can reveal not only whether an excess correlation is present, but also how far it extends in rapidity. Such a correlated pattern would provide substantially more information than a single nonmonotonic point and could help distinguish a long-wavelength critical mode from conservation, stopping, volume fluctuations, decays, and hadronic transport. Quantitative interpretation will require realistic noncritical baselines, detector effects, and a controlled mapping from conserved net baryons to measured net protons.

The next step is to embed the observable in dynamical multistage simulations with nonequilibrium critical evolution [44, 45, 46, 47, 48] and to determine the rapidity acceptances that optimize critical response relative to statistical and systematic uncertainties [49, 50, 51, 52]. With ongoing and future finite-density programs at RHIC [1, 3], NA61/SHINE [53], CBM [54], HIAF [55], and NICA [56], cross-rapidity cumulants open a complementary direction for the search for QCD criticality: using the sign, strength, and rapidity range of baryon correlations to resolve the underlying collective mode.

**Acknowledgments.—**We thank Navid Abbasi, Xin An, Hanwen Feng, Weiyao Ke, Jan M. Pawlowski, and Yi Yin for helpful discussions. J.L. appreciates the discussion and warm hospitality of Wei-jie Fu and Shanjin Wu during his visit to Dalian University of Technology. This work is partially supported by the National Key Research and Development Program of China under Contract No. 2024YFA1610700 (S.S.), NSFC by grant No. 12575143 (S.S.), Tsinghua University under grant Nos. 04200500123, 531205006, 533305009 (S.S.), and the DFG Collaborative Research Centre “SFB 1225 (ISO-QUANT)” (J.L.). An AI-assisted tool was used solely to improve the readability of the manuscript. All scientific content was reviewed and approved by the authors.

References

Supplemental Material

This Supplemental Material gives the derivation of the cross-rapidity cumulant used in the main text. The aim is to make explicit how the measured covariance of the net-baryon numbers in two separated rapidity windows separates into a fixed-field canonical contribution and a contribution generated by event-by-event fluctuations of the critical field, and to identify possible applications in experiments.

Detailed calculation of the cross-rapidity cumulant

We keep the notation of the Letter: the two observed windows are denoted by i=1,2i=1,2, the rest of phase space by i=3i=3, and the total net-baryon number of the closed system is fixed to BB. The second-order cross-rapidity net-baryon cumulant is

κ1,1≡⟨δB^1δB^2⟩ce.(11)\kappa_{1,1} \equiv\langle\delta\hat{B}_{1}\delta\hat{B}_{2}\rangle_{\mathrm{ce}} . \tag*{(11)}

Here δB^i≡B^i−⟨B^i⟩ce\delta\hat{B}_{i}\equiv\hat{B}_{i}-\langle\hat{B}_{i}\rangle_{\mathrm{ce}} is the event-by-event fluctuation of the net-baryon number in rapidity bin ii, and

B^=∫d3x∑hdhψˉhγ0ψh(12)\hat{B}=\int\mathrm{d}^{3}x\sum_{h}d_{h}\bar{\psi}_{h}\gamma^{0}\psi_{h} \tag*{(12)}

is the total net-baryon-number operator. Two physical ingredients are retained. First, every event is treated as a canonical ensemble system, so global net-baryon conservation correlates different rapidity regions. Second, baryons couple to the classical scalar critical field σ\sigma, whose long-wavelength fluctuations can coherently modulate the mean yields in both windows. The canonical ensemble average is therefore organized in two nested steps. For a fixed σ\sigma configuration, we average over baryon fields at fixed total net-baryon number,

⟨O^⟩ce[σ]≡∫∏hDψˉhDψhδB,B^e−SE[σ;ψ,ψˉ]O^∫∏hDψˉhDψhδB,B^e−SE[σ;ψ,ψˉ](13)\langle\hat{O}\rangle_{\mathrm{ce}[\sigma]}\equiv \frac{\int\prod_{h}D\bar{\psi}_{h}D\psi_{h}\delta_{B,\hat{B}}e^{-S_{E}[\sigma;\psi,\bar{\psi}]}\hat{O}} {\int\prod_{h}D\bar{\psi}_{h}D\psi_{h}\delta_{B,\hat{B}}e^{-S_{E}[\sigma;\psi,\bar{\psi}]}} \tag*{(13)}

with Euclidean action SE[σ;ψˉ,ψ]=∫01/Tdτ∫d3x∑hdhψˉh[γ0(∂τ−μB)−iγ⋅∇+mh+Ghσ]ψhS_{E}[\sigma;\bar{\psi},\psi]=\int_{0}^{1/T}\mathrm{d}\tau\int\mathrm{d}^{3}x\sum_{h}d_{h}\bar{\psi}_{h}[\gamma^{0}(\partial_{\tau}-\mu_{B})-i\boldsymbol{\gamma}\cdot\nabla+m_{h}+G_{h}\sigma]\psi_{h}. The projector δB,B^\delta_{B,\hat{B}} restricts the baryonic configurations to the canonical ensemble with fixed total net-baryon number BB. The full average then further averages the fixed-σ\sigma result over the field distribution,

⟨O^⟩ce≡⟨⟨O^⟩ce[σ]⟩σ=∫Dσ e−E[σ]/T⟨O^⟩ce[σ]∫Dσ e−E[σ]/T.(14)\langle\hat{O}\rangle_{\mathrm{ce}}\equiv\langle\langle\hat{O}\rangle_{\mathrm{ce}[\sigma]}\rangle_{\sigma} =\frac{\int D\sigma\,e^{-E[\sigma]/T}\langle\hat{O}\rangle_{\mathrm{ce}[\sigma]}} {\int D\sigma\,e^{-E[\sigma]/T}} . \tag*{(14)}

These definitions are the functional version of the statistical decomposition used in the Letter.

Let Bi(σ)≡⟨B^i⟩ce[σ]B_{i}(\sigma)\equiv\langle\hat{B}_{i}\rangle_{\mathrm{ce}[\sigma]} be the conditional mean net-baryon number in window ii at fixed σ\sigma. Since ⟨B^i⟩ce=⟨Bi(σ)⟩σ\langle\hat{B}_{i}\rangle_{\mathrm{ce}}=\langle B_{i}(\sigma)\rangle_{\sigma}, the total fluctuation could be decomposed as

δB^i≡B^i−⟨B^i⟩ce=(B^i−Bi(σ))+ΔBi(σ),ΔBi(σ)=Bi(σ)−⟨Bi(σ)⟩σ.(15)\delta\hat{B}_{i}\equiv\hat{B}_{i}-\langle\hat{B}_{i}\rangle_{\mathrm{ce}} =(\hat{B}_{i}-B_{i}(\sigma))+\Delta B_{i}(\sigma),\qquad \Delta B_{i}(\sigma)=B_{i}(\sigma)-\langle B_{i}(\sigma)\rangle_{\sigma}. \tag*{(15)}

Substituting this identity into Eq. (11) gives

κ1,1=⟨((B^1−B1(σ))+ΔB1(σ))((B^2−B2(σ))+ΔB2(σ))⟩ce=⟨(B^1−B1(σ))(B^2−B2(σ))⟩ce+⟨ΔB1(σ)ΔB2(σ)⟩ce+⟨(B^1−B1(σ))ΔB2(σ)⟩ce+⟨ΔB1(σ)(B^2−B2(σ))⟩ce=⟨⟨(B^1−B1(σ))(B^2−B2(σ))⟩ceψ⟩σ+⟨ΔB1(σ)ΔB2(σ)⟩σ.(16)\begin{aligned} \kappa_{1,1} &=\left\langle\left((\hat{B}_{1}-B_{1}(\sigma))+\Delta B_{1}(\sigma)\right)\left((\hat{B}_{2}-B_{2}(\sigma))+\Delta B_{2}(\sigma)\right)\right\rangle_{\mathrm{ce}}\\ &=\left\langle(\hat{B}_{1}-B_{1}(\sigma))(\hat{B}_{2}-B_{2}(\sigma))\right\rangle_{\mathrm{ce}} +\left\langle\Delta B_{1}(\sigma)\Delta B_{2}(\sigma)\right\rangle_{\mathrm{ce}}\\ &\quad+\left\langle(\hat{B}_{1}-B_{1}(\sigma))\Delta B_{2}(\sigma)\right\rangle_{\mathrm{ce}} +\left\langle\Delta B_{1}(\sigma)(\hat{B}_{2}-B_{2}(\sigma))\right\rangle_{\mathrm{ce}}\\ &=\left\langle\left\langle(\hat{B}_{1}-B_{1}(\sigma))(\hat{B}_{2}-B_{2}(\sigma))\right\rangle_{\mathrm{ce}}^{\psi}\right\rangle_{\sigma} +\left\langle\Delta B_{1}(\sigma)\Delta B_{2}(\sigma)\right\rangle_{\sigma}. \tag*{(16)} \end{aligned}

The cross terms vanish because, at fixed σ\sigma,

⟨(B^i−Bi(σ))ΔBj(σ)⟩ce=⟨⟨B^i−Bi(σ)⟩ceψΔBj(σ)⟩σ=0.(17)\left\langle(\hat{B}_{i}-B_{i}(\sigma))\Delta B_{j}(\sigma)\right\rangle_{\mathrm{ce}} =\left\langle\left\langle\hat{B}_{i}-B_{i}(\sigma)\right\rangle_{\mathrm{ce}}^{\psi}\Delta B_{j}(\sigma)\right\rangle_{\sigma}=0 . \tag*{(17)}

Defining the fixed-field covariance

κ1,1ce[σ]≡⟨(B^1−B1(σ))(B^2−B2(σ))⟩ce[σ],(18)\kappa^{\mathrm{ce}}_{1,1}[\sigma] \equiv\left\langle\left(\widehat{B}_{1}-B_{1}(\sigma)\right)\left(\widehat{B}_{2}-B_{2}(\sigma)\right)\right\rangle_{\mathrm{ce}[\sigma]}, \tag*{(18)}

and the covariance of the field-dependent mean yields,

C1,1σ≡⟨ΔB1(σ)ΔB2(σ)⟩σ=⟨B1(σ)B2(σ)⟩σ−⟨B1(σ)⟩σ⟨B2(σ)⟩σ,(19)C^{\sigma}_{1,1} \equiv\left\langle\Delta B_{1}(\sigma)\Delta B_{2}(\sigma)\right\rangle_{\sigma} = \left\langle B_{1}(\sigma)B_{2}(\sigma)\right\rangle_{\sigma}-\left\langle B_{1}(\sigma)\right\rangle_{\sigma}\left\langle B_{2}(\sigma)\right\rangle_{\sigma}, \tag*{(19)}

one obtains the identity Eq. (4) quoted in the main text,

κ1,1=⟨κ1,1ce[σ]⟩σ+C1,1σ.(20)\kappa_{1,1}=\left\langle\kappa^{\mathrm{ce}}_{1,1}[\sigma]\right\rangle_{\sigma}+C^{\sigma}_{1,1}. \tag*{(20)}

No dynamical assumption has entered this step. The model assumptions enter only when the two terms on the right-hand side are evaluated.

At fixed σ\sigma, the canonical average defined in Eq. (13) can be evaluated equivalently through the generating-function method [R1]. The corresponding generating function is

gBσ(ν,λ)=ln⁡(∑B1,B2=−∞∞eνB1eλB2Pσ(B1,B2∣B)).(21)g^{\sigma}_{B}(\nu,\lambda)=\ln\left(\sum_{B_{1},B_{2}=-\infty}^{\infty}e^{\nu B_{1}}e^{\lambda B_{2}}P_{\sigma}(B_{1},B_{2}\mid B)\right). \tag*{(21)}

The joint probability is built from independent baryon and antibaryon emission into the three bins, followed by the exact net-baryon constraint,

Pσ(B1,B2∣B)=∑Ni,N‾i≥0i=1,2,3δB,B1+B2+B3∏i=13(Piσ(Ni)Pi‾σ(N‾i)δBi,Ni−N‾i).(22)P_{\sigma}(B_{1},B_{2}\mid B)=\sum_{\substack{N_{i},\overline{N}_{i}\geq0\\i=1,2,3}}\delta_{B,B_{1}+B_{2}+B_{3}}\prod_{i=1}^{3}\left(P^{\sigma}_{i}(N_{i})P^{\sigma}_{\overline{i}}(\overline{N}_{i})\delta_{B_{i},N_{i}-\overline{N}_{i}}\right). \tag*{(22)}

Here bin 3 denotes all momentum-space regions outside the two observed windows, and overbars denote antibaryons. The Kronecker factor is the source of the conservation-induced anticorrelation: once the total net-baryon number is fixed, an excess assigned to one bin constrains the remaining bins.

The fixed-field cumulants follow from

κn,mce[σ]=dn+mgBσ(ν,λ)dνndλm∣ν=0,λ=0.(23)\kappa^{\mathrm{ce}}_{n,m}[\sigma]=\left.\frac{\mathrm{d}^{n+m}g^{\sigma}_{B}(\nu,\lambda)}{\mathrm{d}\nu^{n}\mathrm{d}\lambda^{m}}\right|_{\nu=0,\lambda=0}. \tag*{(23)}

For the covariance, this is equivalent to

κ1,1ce[σ]=∑B1,B2=−∞+∞(B1−B1(σ))(B2−B2(σ))Pσ(B1,B2∣B)∑B1,B2=−∞+∞Pσ(B1,B2∣B).(24)\kappa^{\mathrm{ce}}_{1,1}[\sigma]=\frac{\sum_{B_{1},B_{2}=-\infty}^{+\infty}(B_{1}-B_{1}(\sigma))(B_{2}-B_{2}(\sigma))P_{\sigma}(B_{1},B_{2}\mid B)}{\sum_{B_{1},B_{2}=-\infty}^{+\infty}P_{\sigma}(B_{1},B_{2}\mid B)}. \tag*{(24)}

Under independent emission in the continuous source model, the baryon and antibaryon multiplicities in bin ii are Poisson distributed [R2],

Piσ(Ni)=PPois(Ni;zi[σ]),Pi‾σ(N‾i)=PPois(N‾i;z‾i[σ]),i∈{1,2,3}.(25)P^{\sigma}_{i}(N_{i})=\mathcal{P}_{\mathrm{Pois}}(N_{i};z_{i}[\sigma]),\qquad P^{\sigma}_{\overline{i}}(\overline{N}_{i})=\mathcal{P}_{\mathrm{Pois}}(\overline{N}_{i};\overline{z}_{i}[\sigma]),\qquad i\in\{1,2,3\}. \tag*{(25)}

Here, zi[σ]z_{i}[\sigma] denotes the single-particle grand-canonical partition function for baryons emitted into bin ii at fixed σ\sigma. Since we include only baryons carrying unit baryon number, zi[σ]z_{i}[\sigma] also gives the mean baryon number in that bin. The corresponding source-integrated baryon yield is then

zi[σ]=∫dVs zi,s[σ],(26)z_{i}[\sigma]=\int\mathrm{d}V_{s}\,z_{i,s}[\sigma], \tag*{(26)}
zi,s[σ]=∑hdh∫d3p(2π)3us⋅pp0eμs−us⋅pTsΘ(y∈yi)=∑hdhTs(2π)2∫yilowyiupperdy eμs−mhσcosh⁡(y−ys)Ts((mhσ)2+2mhσTscosh⁡(y−ys)+2Ts2cosh⁡2(y−ys)).(27)\begin{aligned} z_{i,s}[\sigma]&=\sum_{h}\mathrm{d}_{h}\int\frac{\mathrm{d}^{3}p}{(2\pi)^{3}}\frac{u_{s}\cdot p}{p^{0}}e^{\frac{\mu_{s}-u_{s}\cdot p}{T_{s}}}\Theta(y\in y_{i})\\ &=\sum_{h}\frac{\mathrm{d}_{h}T_{s}}{(2\pi)^{2}}\int_{y_{i}^{\mathrm{low}}}^{y_{i}^{\mathrm{upper}}}\mathrm{d}y\,e^{\frac{\mu_{s}-m_{h}^{\sigma}\cosh(y-y_{s})}{T_{s}}}\left((m_{h}^{\sigma})^{2}+\frac{2m_{h}^{\sigma}T_{s}}{\cosh(y-y_{s})}+\frac{2T_{s}^{2}}{\cosh^{2}(y-y_{s})}\right). \tag*{(27)} \end{aligned}

Here, zˉi,s[σ]\bar z_{i,s}[\sigma] denotes the contribution from source ss to the mean baryon yield emitted into rapidity bin ii at fixed σ\sigma. The antibaryon quantities zˉˉi[σ]\bar{\bar z}_i[\sigma] and zˉˉi,s[σ]\bar{\bar z}_{i,s}[\sigma] are obtained by charge conjugation. In the main text we denote the unshifted integrated yields by Nˉi≡zˉi[0]\bar N_i \equiv\bar z_i[0], Nˉˉi≡zˉˉi[0]\bar{\bar N}_i \equiv\bar{\bar z}_i[0] and Bi=Nˉi−NˉˉiB_i = \bar N_i-\bar{\bar N}_i, with Nˉ≡zˉ[0]\bar N \equiv\bar z[0], Nˉˉ≡zˉˉ[0]\bar{\bar N} \equiv\bar{\bar z}[0] and B=Nˉ−NˉˉB = \bar N-\bar{\bar N} for the full phase space.

Using the Poisson form, Eq. (22) becomes

Pσ(B1,B2∣B)=(∏i=13(zˉizˉˉi)Bi2e−zˉi−zˉˉiIBi(2zˉizˉˉi))B3=B−B1−B2,(28)P_\sigma(B_1,B_2|B)=\left(\prod_{i=1}^{3}\left(\frac{\bar z_i}{\bar{\bar z}_i}\right)^{\frac{B_i}{2}}e^{-\bar z_i-\bar{\bar z}_i}I_{B_i}\left(2\sqrt{\bar z_i\bar{\bar z}_i}\right)\right)_{B_3=B-B_1-B_2}, \tag*{(28)}

where In(x)I_n(x) is the modified Bessel function of the first kind. Let

zˉ[σ]≡∑i=13zˉi[σ],zˉˉ[σ]≡∑i=13zˉˉi[σ].(29)\bar z[\sigma]\equiv\sum_{i=1}^{3}\bar z_i[\sigma],\qquad\bar{\bar z}[\sigma]\equiv\sum_{i=1}^{3}\bar{\bar z}_i[\sigma]. \tag*{(29)}

Applying Graf’s addition theorem [R3] twice gives

egBσ(ν,λ)=e−(zˉ[σ]+zˉˉ[σ])(zˉ[σ]zˉˉ[σ])B2(qσ{ν,λ}qˉσ{ν,λ})B2IB(2zˉ[σ]zˉˉ[σ]qσ{ν,λ}qˉσ{ν,λ}).(30)e^{g_B^{\sigma}(\nu,\lambda)}=e^{-(\bar z[\sigma]+\bar{\bar z}[\sigma])}\left(\frac{\bar z[\sigma]}{\bar{\bar z}[\sigma]}\right)^{\frac{B}{2}}\left(\frac{q_\sigma^{\{\nu,\lambda\}}}{\bar q_\sigma^{\{\nu,\lambda\}}}\right)^{\frac{B}{2}}I_B\left(2\sqrt{\bar z[\sigma]\bar{\bar z}[\sigma]q_\sigma^{\{\nu,\lambda\}}\bar q_\sigma^{\{\nu,\lambda\}}}\right). \tag*{(30)}

The counting factors are contained in

qσ{ν,λ}≡eνzˉ1[σ]zˉ[σ]+λzˉ2[σ]zˉ[σ]+zˉ3[σ]zˉ[σ],(31)q_\sigma^{\{\nu,\lambda\}}\equiv e^{\nu\frac{\bar z_1[\sigma]}{\bar z[\sigma]}+\lambda\frac{\bar z_2[\sigma]}{\bar z[\sigma]}+\frac{\bar z_3[\sigma]}{\bar z[\sigma]}}, \tag*{(31)}

with qˉσ{ν,λ}\bar q_\sigma^{\{\nu,\lambda\}} defined analogously for antibaryons. Differentiating the generating function yields

κ1,1ce[σ]=d2gBσ(ν,λ)dν dλ∣ν=0,λ=0=−zˉ1[σ]zˉ2[σ]zˉ[σ]−zˉˉ1[σ]zˉˉ2[σ]zˉˉ[σ].(32)\kappa_{1,1}^{\mathrm{ce}}[\sigma]=\left.\frac{d^2g_B^\sigma(\nu,\lambda)}{d\nu\,d\lambda}\right|_{\nu=0,\lambda=0}=-\frac{\bar z_1[\sigma]\bar z_2[\sigma]}{\bar z[\sigma]}-\frac{\bar{\bar z}_1[\sigma]\bar{\bar z}_2[\sigma]}{\bar{\bar z}[\sigma]}. \tag*{(32)}

In this differentiation we use the recurrence relations for In(x)I_n(x) [R4] together with

zˉ[σ]+zˉˉ[σ]zˉ[σ]zˉˉ[σ]=IB+1(2zˉ[σ]zˉˉ[σ])+IB−1(2zˉ[σ]zˉˉ[σ])IB(2zˉ[σ]zˉˉ[σ]).(33)\frac{\bar z[\sigma]+\bar{\bar z}[\sigma]}{\sqrt{\bar z[\sigma]\bar{\bar z}[\sigma]}}=\frac{I_{B+1}\left(2\sqrt{\bar z[\sigma]\bar{\bar z}[\sigma]}\right)+I_{B-1}\left(2\sqrt{\bar z[\sigma]\bar{\bar z}[\sigma]}\right)}{I_B\left(2\sqrt{\bar z[\sigma]\bar{\bar z}[\sigma]}\right)}. \tag*{(33)}

Equation (32) is the microscopic origin of the negative conservation term in the Letter: the two observed windows are anticorrelated because they draw from a fixed total conserved charge.

We now evaluate the second term in Eq. (20). We define the baryon and antibaryon responses of bin ii to a local change of σ\sigma as

χiB(ys)=∂zˉi,s[σ]∂σ∣σ=0=−∑h∫yiloweryiupperdyidhGhmh2(2π)2eμB−mhcosh⁡(y−ys)Tcosh⁡(y−ys),\chi_i^B(y_s)=\left.\frac{\partial\bar z_{i,s}[\sigma]}{\partial\sigma}\right|_{\sigma=0}=-\sum_h\int_{y_i^{\mathrm{lower}}}^{y_i^{\mathrm{upper}}}dy_i\frac{d_hG_hm_h^2}{(2\pi)^2}e^{\frac{\mu_B-m_h\cosh(y-y_s)}{T}}\cosh(y-y_s),
χiBˉ(ys)=∂zˉˉi,s[σ]∂σ∣σ=0=−∑h∫yiloweryiupperdyidhGhmh2(2π)2eμB+mhcosh⁡(y−ys)Tcosh⁡(y−ys),\chi_i^{\bar B}(y_s)=\left.\frac{\partial\bar{\bar z}_{i,s}[\sigma]}{\partial\sigma}\right|_{\sigma=0}=-\sum_h\int_{y_i^{\mathrm{lower}}}^{y_i^{\mathrm{upper}}}dy_i\frac{d_hG_hm_h^2}{(2\pi)^2}e^{\frac{\mu_B+m_h\cosh(y-y_s)}{T}}\cosh(y-y_s),
χi(ys)=χiB(ys)−χiBˉ(ys).(34)\chi_i(y_s)=\chi_i^B(y_s)-\chi_i^{\bar B}(y_s). \tag*{(34)}

Thus χi(ys)\chi_i(y_s) is the linear response of the net-baryon yield in bin ii to the order-parameter field at source rapidity ysy_s. The field-dependent mean net-baryon yield in bin ii is, to linear order in σ/mh\sigma/m_h,

Bi(σ)≈Bi+∫σ(ys)χi(ys) dVs.(35)B_i(\sigma)\approx B_i+\int\sigma(y_s)\chi_i(y_s)\,dV_s. \tag*{(35)}

Writing σ=σˉ+δσ\sigma=\bar\sigma+\delta\sigma and using ⟨δσ⟩σ=0\langle\delta\sigma\rangle_\sigma=0, we obtain

⟨Bi(σ)⟩σ=Bi+σˉ∫χi(ys) dVsi+O((σmh)2),(36)\langle B_i(\sigma)\rangle_\sigma=B_i+\bar\sigma\int\chi_i(y_s)\,dV_{s_i}+O\left(\left(\frac{\sigma}{m_h}\right)^2\right), \tag*{(36)}
⟨B1(σ)B2(σ)⟩σ=∏i=12(Bi+σ‾∫χi(ys) dVsi)+∫⟨δσ(ys1)δσ(ys2)⟩σχ1(ys1)χ2(ys2) dVs1 dVs2.(37)\left\langle B_1(\sigma) B_2(\sigma) \right\rangle_\sigma= \prod_{i=1}^{2} \left(B_i+\overline{\sigma}\int\chi_i(y_s)\,\mathrm{d}V_{s_i}\right)+\int\left\langle\delta\sigma(y_{s_1})\delta\sigma(y_{s_2})\right\rangle_\sigma\chi_1(y_{s_1})\chi_2(y_{s_2})\,\mathrm{d}V_{s_1}\,\mathrm{d}V_{s_2}. \tag*{(37)}

The regular pieces cancel in the covariance definition, leaving only the connected critical-field correlator. With

Cσ(x,y)≡⟨δσ(x(ys1))δσ(y(ys2))⟩σ=T4π∣x−y∣e−∣x−y∣ξeq,(38)\mathcal{C}_\sigma(x,y)\equiv\left\langle\delta\sigma(x(y_{s_1}))\delta\sigma(y(y_{s_2}))\right\rangle_\sigma=\frac{T}{4\pi\lvert x-y\rvert}e^{-\frac{\lvert x-y\rvert}{\xi_{\mathrm{eq}}}}, \tag*{(38)}

which reduces to Eq. (7) of the main text for a constant-temperature source; the leading critical contribution becomes

C1,1σ=∫Cσ(xs1,xs2)χ1(ys1)χ2(ys2) dVs1 dVs2.(39)C^\sigma_{1,1}=\int\mathcal{C}_\sigma(x_{s_1},x_{s_2})\chi_1(y_{s_1})\chi_2(y_{s_2})\,\mathrm{d}V_{s_1}\,\mathrm{d}V_{s_2}. \tag*{(39)}

The correlator Cσ\mathcal{C}_\sigma is an equal-time correlator in the local rest frame. To apply it on the freeze-out hypersurface, one must specify how two source cells are compared in a common rest frame. We assume local thermal equilibrium and neglect transverse flow. For a cell ii at xiμ=(ti,xi,yi,zi)x_i^\mu=(t_i,x_i,y_i,z_i) with four-velocity uiμ=γi(1,vi,x,vi,y,vi,z)u_i^\mu=\gamma_i(1,v_{i,x},v_{i,y},v_{i,z}), where γi=1/1−vi2\gamma_i=1/\sqrt{1-v_i^2} and vi2=vi,x2+vi,y2+vi,z2v_i^2=v_{i,x}^2+v_{i,y}^2+v_{i,z}^2, the Lorentz transformation to the rest frame of the cell is

Λi,νμ=(γi−γivi,x−γivi,y−γivi,z−γivi,x1+(γi−1)vi,x2vi2(γi−1)vi,xvi,yvi2(γi−1)vi,xvi,zvi2−γivi,y(γi−1)vi,xvi,yvi21+(γi−1)vi,y2vi2(γi−1)vi,yvi,zvi2−γivi,z(γi−1)vi,xvi,zvi2(γi−1)vi,zvi,yvi21+(γi−1)vi,z2vi2).(40)\Lambda^\mu_{i,\nu}= \begin{pmatrix} \gamma_i & -\gamma_i v_{i,x} & -\gamma_i v_{i,y} & -\gamma_i v_{i,z}\\ -\gamma_i v_{i,x} & 1+(\gamma_i-1)\frac{v_{i,x}^2}{v_i^2} & (\gamma_i-1)\frac{v_{i,x}v_{i,y}}{v_i^2} & (\gamma_i-1)\frac{v_{i,x}v_{i,z}}{v_i^2}\\ -\gamma_i v_{i,y} & (\gamma_i-1)\frac{v_{i,x}v_{i,y}}{v_i^2} & 1+(\gamma_i-1)\frac{v_{i,y}^2}{v_i^2} & (\gamma_i-1)\frac{v_{i,y}v_{i,z}}{v_i^2}\\ -\gamma_i v_{i,z} & (\gamma_i-1)\frac{v_{i,x}v_{i,z}}{v_i^2} & (\gamma_i-1)\frac{v_{i,z}v_{i,y}}{v_i^2} & 1+(\gamma_i-1)\frac{v_{i,z}^2}{v_i^2} \end{pmatrix}. \tag*{(40)}

It satisfies Λi,νμuiν=(1,0,0,0)\Lambda^\mu_{i,\nu}u_i^\nu=(1,0,0,0). In the absence of transverse flow, vi,x=vi,y=0v_{i,x}=v_{i,y}=0, the transformation reduces to the longitudinal boost

Λi,νμ=(γi−γivi,z−γivi,zγi).(41)\Lambda^\mu_{i,\nu}= \begin{pmatrix} \gamma_i & -\gamma_i v_{i,z}\\ -\gamma_i v_{i,z} & \gamma_i \end{pmatrix}. \tag*{(41)}

We parametrize the freeze-out coordinates as ti=τicosh⁡yit_i=\tau_i\cosh y_i, zi=τisinh⁡yiz_i=\tau_i\sinh y_i, and use vi,z=tanh⁡yiv_{i,z}=\tanh y_i, corresponding to the usual approximation that the spacetime rapidity of the source is the longitudinal flow rapidity. This approximation is the same one used in the continuous-source construction of the main text: a source at ysy_s emits thermally smeared particles into momentum rapidity yy, and the difference y−ysy-y_s is already retained in Eq. (27). The approximation is therefore not an identification of source rapidity with observed particle rapidity, but an assumption that the local longitudinal flow labels the source position on the freeze-out hypersurface.

Requiring equal time in the rest frame of cell ii gives

Λi,ν0(xiν−xjν)=0,(42)\Lambda^0_{i,\nu}(x_i^\nu-x_j^\nu)=0, \tag*{(42)}

which leads to

τi−cosh⁡(yi−yj)τj=0.(43)\tau_i-\cosh(y_i-y_j)\tau_j=0. \tag*{(43)}

For a constant-proper-time hypersurface, τi=τj≡τ\tau_i=\tau_j\equiv\tau, the unique solution is yi=yjy_i=y_j. Thus the equal-time condition collapses the double source-rapidity integral in Eq. (39) to a single ysy_s integral. The remaining integration is over transverse positions at the same source rapidity.

For a cylindrical transverse profile with radius RR, the transverse integral depends only on the separation u=∣ρ1−ρ2∣u=\lvert\boldsymbol{\rho}_1-\boldsymbol{\rho}_2\rvert. The overlap area of two disks of radius RR separated by u∈[0,2R]u\in[0,2R] is

A(u)=2R2arccos⁡(u2R)−u24R2−u2.(44)\mathcal{A}(u)=2R^2\arccos\left(\frac{u}{2R}\right)-\frac{u}{2}\sqrt{4R^2-u^2}. \tag*{(44)}

Using this geometric factor, Eq. (39), i.e., Eq. (8) in the main text, becomes

C1,1σ=∫02RA(u)Cσ(x,y) 2πuτ2 du∫χ1(ys)χ2(ys) dys=πTRτ2ξeq22(M1(2Rξeq)+Rξeq)∫χ1(ys)χ2(ys) dys.(45)\begin{aligned} C^\sigma_{1,1}&=\int_0^{2R}\mathcal{A}(u)\mathcal{C}_\sigma(x,y)\,2\pi u\tau^2\,\mathrm{d}u\int\chi_1(y_s)\chi_2(y_s)\,\mathrm{d}y_s\\ &=\frac{\pi T R\tau^2\xi_{\mathrm{eq}}^2}{2}\left(\mathcal{M}_1\left(\frac{2R}{\xi_{\mathrm{eq}}}\right)+\frac{R}{\xi_{\mathrm{eq}}}\right)\int\chi_1(y_s)\chi_2(y_s)\,\mathrm{d}y_s. \tag*{(45)} \end{aligned}

The function M1M_1 is the modified Struve function of the second kind. This result is the geometric input for the analytic estimate of the conservation-subtracted correlator in the main text.

However, the fixed-field result in Eq. (32) still carries a σ\sigma dependence through the dressed masses. Near the CEP we keep only the leading corrections generated by the small mass shift. Effective model estimates give Ghσ≈Ghˉσ≪mhG_{h\sigma} \approx G_{\bar{h}\sigma} \ll m_h [29]. We also write

δzi[σ]≡∫χiB(ys)σ(xs)dVs,δzˉi[σ]≡∫χiBˉ(ys)σ(xs)dVs,(46)\delta\mathfrak{z}_i[\sigma] \equiv\int\chi_i^B(y_s)\sigma(x_s)\mathrm{d}V_s,\qquad\delta\bar{\mathfrak{z}}_i[\sigma] \equiv\int\chi_i^{\bar{B}}(y_s)\sigma(x_s)\mathrm{d}V_s, \tag*{(46)}

with δz[σ]≡∑i=13δzi[σ]\delta\mathfrak{z}[\sigma] \equiv\sum_{i=1}^{3}\delta\mathfrak{z}_i[\sigma] and analogously for antibaryons. Expanding Eq. (32) gives

⟨κ1,1ce[σ]⟩σ=κ1,1ce,(0)+⟨κ1,1ce,(1)[σ]⟩σ+⟨κ1,1ce,(2)[σ]⟩σ+O((Ghσmh)3),(47)\left\langle\kappa_{1,1}^{\mathrm{ce}}[\sigma]\right\rangle_\sigma= \kappa_{1,1}^{\mathrm{ce},(0)} + \left\langle\kappa_{1,1}^{\mathrm{ce},(1)}[\sigma]\right\rangle_\sigma+ \left\langle\kappa_{1,1}^{\mathrm{ce},(2)}[\sigma]\right\rangle_\sigma+ \mathcal{O}\left(\left(\frac{G_{h\sigma}}{m_h}\right)^3\right), \tag*{(47)}

where

κ1,1ce,(0)=−z1[0]z2[0]z[0]−zˉ1[0]zˉ2[0]zˉ[0],\kappa_{1,1}^{\mathrm{ce},(0)}=-\frac{\mathfrak{z}_1[0]\mathfrak{z}_2[0]}{\mathfrak{z}[0]}-\frac{\bar{\mathfrak{z}}_1[0]\bar{\mathfrak{z}}_2[0]}{\bar{\mathfrak{z}}[0]},
⟨κ1,1ψ,(1)[σ]⟩σ=z1[0]z2[0]δz[σˉ]z[0]2−z1[0]δz2[σ]z[0]−δz1[σ]z2[0]z[0]+[charge conjugation].(48)\left\langle\kappa_{1,1}^{\psi,(1)}[\sigma]\right\rangle_\sigma= \frac{\mathfrak{z}_1[0]\mathfrak{z}_2[0]\delta\mathfrak{z}[\bar{\sigma}]}{\mathfrak{z}[0]^2} -\frac{\mathfrak{z}_1[0]\delta\mathfrak{z}_2[\sigma]}{\mathfrak{z}[0]} -\frac{\delta\mathfrak{z}_1[\sigma]\mathfrak{z}_2[0]}{\mathfrak{z}[0]} +[\text{charge conjugation}]. \tag*{(48)}

For mh≫Tm_h \gg T, one has δzi≈(Gpσ/T)zi\delta\mathfrak{z}_i \approx(G_{p\sigma}/T)\mathfrak{z}_i. The first-order correction is therefore parametrically small,

⟨κ1,1ce,(1)[σ]⟩σ≈κ1,1ce,(0)GpσT≪κ1,1ψ,(0),(49)\left\langle\kappa_{1,1}^{\mathrm{ce},(1)}[\sigma]\right\rangle_\sigma\approx\kappa_{1,1}^{\mathrm{ce},(0)}\frac{G_{p\sigma}}{T} \ll\kappa_{1,1}^{\psi,(0)}, \tag*{(49)}

because Gpσ≪TG_{p\sigma} \ll T in the vicinity of the CEP [29].

At second order, the terms relevant for correlations between two narrow and separated windows are

⟨κ1,1ce,(2)[σ]⟩σ≈−⟨δz1[σ]δz2[σ]⟩σz[0]−⟨δzˉ1[σ]δzˉ2[σ]⟩σzˉ[0]=−∬dVs1dVs2(σˉ2+Cσ(xs1,xs2))(χ1B(ys1)χ2B(ys2)z[0]+χ1Bˉ(ys1)χ2Bˉ(ys2)zˉ[0]).(50)\begin{aligned} \left\langle\kappa_{1,1}^{\mathrm{ce},(2)}[\sigma]\right\rangle_\sigma\approx-\frac{\left\langle\delta\mathfrak{z}_1[\sigma]\delta\mathfrak{z}_2[\sigma]\right\rangle_\sigma}{\mathfrak{z}[0]}-\frac{\left\langle\delta\bar{\mathfrak{z}}_1[\sigma]\delta\bar{\mathfrak{z}}_2[\sigma]\right\rangle_\sigma}{\bar{\mathfrak{z}}[0]} \\ &= -\iint\mathrm{d}V_{s_1}\mathrm{d}V_{s_2}\left(\bar{\sigma}^2+\mathcal{C}_\sigma(x_{s_1},x_{s_2})\right)\left(\frac{\chi_1^B(y_{s_1})\chi_2^B(y_{s_2})}{\mathfrak{z}[0]}+\frac{\chi_1^{\bar{B}}(y_{s_1})\chi_2^{\bar{B}}(y_{s_2})}{\bar{\mathfrak{z}}[0]}\right). \tag*{(50)} \end{aligned}

Here we have used the narrow-window hierarchy z1,2≪z\mathfrak{z}_{1,2}\ll\mathfrak{z} and zˉ1,2≪zˉ\bar{\mathfrak{z}}_{1,2}\ll\bar{\mathfrak{z}}, together with

⟨δzi[σ]δzj[σ]⟩σ=∬dVs1dVs2 χiB(ys1)χjB(ys2)(σˉ2+Cσ(xs1,xs2)).(51)\left\langle\delta\mathfrak{z}_i[\sigma]\delta\mathfrak{z}_j[\sigma]\right\rangle_\sigma= \iint\mathrm{d}V_{s_1}\mathrm{d}V_{s_2}\,\chi_i^B(y_{s_1})\chi_j^B(y_{s_2})\left(\bar{\sigma}^2+\mathcal{C}_\sigma(x_{s_1},x_{s_2})\right). \tag*{(51)}

The first term in the parenthesis of Eq. (50) is the regular mean-field dressing contribution, while the second is generated by critical fluctuations. The antibaryon response is suppressed by the fugacity factor e−2μB/Te^{-2\mu_B/T} relative to the baryon response. Moreover, the entire contribution in Eq. (50) is suppressed by the large full-phase-space multiplicities, z[0]=N≫1\mathfrak{z}[0]=N\gg1 and zˉ[0]=Nˉ≫1\bar{\mathfrak{z}}[0]=\bar{N}\gg1. However, C1,1σ\mathcal{C}_{1,1}^\sigma is not divided by the full-phase-space multiplicity. It is therefore the leading field-induced contribution to the cross-rapidity covariance, while the analogous fluctuation pieces inside κ1,1ce[σ]\kappa_{1,1}^{\mathrm{ce}}[\sigma] are suppressed by NN and Nˉ\bar{N}. Hence these field-dependent corrections to the fixed-field covariance are subleading compared with C1,1σ\mathcal{C}_{1,1}^\sigma. To the order retained in the Letter, one then has Eq. (6)

⟨κ1,1ce[σ]⟩σ≃κ1,1ce,(0)=−N1N2N−Nˉ1Nˉ2Nˉ.(52)\left\langle\kappa_{1,1}^{\mathrm{ce}}[\sigma]\right\rangle_\sigma\simeq\kappa_{1,1}^{\mathrm{ce},(0)}=-\frac{N_1N_2}{N}-\frac{\bar{N}_1\bar{N}_2}{\bar{N}}. \tag*{(52)}

Narrow-window limit, the prefactor K(T)K(T) and the numerical implement

We then derive the analytic expression that connects Eq. (45) to the scaled correlator Eq. (9) in the main text. Consider two adjacent narrow windows y1∈[−w,0)y_1\in[-w,0) and y2∈[0,w)y_2\in[0,w) around midrapidity. In the narrow-bin limit, smooth functions can be pulled out of the window integrals,

∫−w0dx f(x)≈wf(0),∫0wdx f(x)≈wf(0).(53)\int_{-w}^{0}\mathrm{d}x\,f(x)\approx wf(0),\qquad\int_{0}^{w}\mathrm{d}x\,f(x)\approx wf(0). \tag*{(53)}

Since the freeze-out rapidity span is much larger than ww, the source-rapidity integral may be extended to (−∞,∞)(-\infty,\infty) for this analytic estimate. Then

∫χ1(ys)χ2(ys) dys≃w24π4sinh⁡2(μBT)∫−∞+∞dys∏i=1,2∑hidhiGhimhi2cosh⁡(ys)e−mhiTcosh⁡(ys)=Gp2mp2w24π4sinh⁡2(μBT)∑h1,h2dh1dh2mh1mh2(K0(mh1+mh2T)+K2(mh1+mh2T)),B1B2≃w2τ24π4sinh⁡2(μBT)T2R4∏i=1,2∑hi∫−∞+∞dys dhi(mhi2+2mhiTcosh⁡(ys)+2T2cosh⁡2(ys))e−mhiTcosh⁡(ys)=w2τ2π4sinh⁡2(μBT)T2R4∏i=1,2∑hidhimhi2K2(mhiT).(54)\begin{aligned} \int\chi_1(y_s)\chi_2(y_s)\,dy_s &\simeq\frac{w^2}{4\pi^4}\sinh^2\left(\frac{\mu_B}{T}\right)\int_{-\infty}^{+\infty}dy_s\prod_{i=1,2}\sum_{h_i}d_{h_i}G_{h_i}m_{h_i}^2\cosh(y_s)e^{-\frac{m_{h_i}}{T}\cosh(y_s)} \\ &=G_p^2m_p^2\frac{w^2}{4\pi^4}\sinh^2\left(\frac{\mu_B}{T}\right)\sum_{h_1,h_2}d_{h_1}d_{h_2}m_{h_1}m_{h_2}\left(K_0\left(\frac{m_{h_1}+m_{h_2}}{T}\right)+K_2\left(\frac{m_{h_1}+m_{h_2}}{T}\right)\right), \\ B_1B_2 &\simeq\frac{w^2\tau^2}{4\pi^4}\sinh^2\left(\frac{\mu_B}{T}\right)T^2R^4\prod_{i=1,2}\sum_{h_i}\int_{-\infty}^{+\infty}dy_s\,d_{h_i}\left(m_{h_i}^2+\frac{2m_{h_i}T}{\cosh(y_s)}+\frac{2T^2}{\cosh^2(y_s)}\right)e^{-\frac{m_{h_i}}{T}\cosh(y_s)} \\ &=\frac{w^2\tau^2}{\pi^4}\sinh^2\left(\frac{\mu_B}{T}\right)T^2R^4\prod_{i=1,2}\sum_{h_i}d_{h_i}m_{h_i}^2K_2\left(\frac{m_{h_i}}{T}\right). \tag*{(54)} \end{aligned}

Here Bi=Ni−N‾iB_i=N_i-\overline{N}_i is the mean net-baryon yield in window ii, as in the Letter, and Kα(x)K_\alpha(x) denotes the modified Bessel function of the second kind. Combining these expressions with Eq. (45) gives the prefactor

K(T)≡∑h1,h2dh1dh2mh1mh2(K0(mh1+mh2T)+K2(mh1+mh2T))∑h1,h2dh1dh2mh12mh22K2(mh1T)K2(mh2T).(55)\mathcal{K}(T)\equiv\frac{\sum_{h_1,h_2}d_{h_1}d_{h_2}m_{h_1}m_{h_2}\left(K_0\left(\frac{m_{h_1}+m_{h_2}}{T}\right)+K_2\left(\frac{m_{h_1}+m_{h_2}}{T}\right)\right)}{\sum_{h_1,h_2}d_{h_1}d_{h_2}m_{h_1}^2m_{h_2}^2K_2\left(\frac{m_{h_1}}{T}\right)K_2\left(\frac{m_{h_2}}{T}\right)}. \tag*{(55)}

For simplicity, we choose a phenomenological inverse-mass scaling, Gh=Gpmp/mhG_h=G_pm_p/m_h, so that heavier baryons have a smaller relative mean-field mass dressing, Ghσ‾/mhG_h\overline{\sigma}/m_h, and a weaker linear response to critical fluctuations. The resulting normalized critical response is

S1,1≃πGp2mp2K(T)8R(ξeq2R2M1(2Rξeq)+ξeqR).(56)S_{1,1}\simeq\frac{\pi G_p^2m_p^2\mathcal{K}(T)}{8R}\left(\frac{\xi_{\mathrm{eq}}^2}{R^2}M_1\left(\frac{2R}{\xi_{\mathrm{eq}}}\right)+\frac{\xi_{\mathrm{eq}}}{R}\right). \tag*{(56)}

which is Eq. (10) of the main text. Here, the normalization B1B2B_1B_2 removes the leading lifetime τ\tau, window-width ww, and common net-baryon-density dependences on fugacity, leaving a smoothly varying prefactor multiplying a response that increases with ξeq\xi_{\mathrm{eq}}.

This quantitative cancellation relies on the use of narrow rapidity windows. For wider acceptances, thermal smearing and longitudinal inhomogeneity increasingly mix contributions from sources associated with different spacetime rapidities and, consequently, different local thermodynamic conditions, thereby reducing the normalized response, as examined in Figure 2. From this perspective, the restriction to narrow rapidity windows is not merely a technical simplification, but is essential for maintaining a transparent connection between the measured cumulants and the local thermodynamic conditions on the freeze-out hypersurface. As shown in our previous studies [17], when the rapidity window is sufficiently narrow, the measured cumulants are dominated by sources with nearby spacetime rapidities, such that thermal smearing between source rapidity and momentum rapidity gives only a subleading correction. The resulting net-baryon fluctuations are therefore governed primarily by the local values of (T,μB)(T,\mu_B) associated with that portion of the freeze-out hypersurface. This locality is particularly useful near midrapidity, where an approximately boost-invariant interval may persist even at intermediate collision energies [13]. The cross-rapidity cumulants considered here are consequently best suited to relatively narrow, adjacent but nonoverlapping rapidity windows near midrapidity, which preserve an approximately local and homogeneous thermodynamic interpretation while retaining sensitivity to correlations generated by long-wavelength critical modes. Such a separation of scales is important because the critical correlation length can be large compared with microscopic thermal scales, while still remaining small compared with the macroscopic size of the fireball [38].

Finally, for the numerical implementation shown in Figure 2, we need the sNN\sqrt{s_{NN}} dependences of the finite spacetime-rapidity extent of the fireball, denoted by (−yf0,yf0)(-y_{f0},y_{f0}), the freeze-out temperature TT, the baryon chemical potential μB\mu_B, and the local equilibrium correlation length ξeq\xi_{\mathrm{eq}}. The homogeneous freeze-out conditions are parametrized as functions of sNN\sqrt{s_{NN}},

T(sNN)=0.1584 GeV1+exp⁡(2.60−ln⁡sNN0.45),μB(sNN)=1.3075 GeV1+0.288sNN.(57)\begin{aligned} T\left(\sqrt{s_{NN}}\right)&=\frac{0.1584\ \mathrm{GeV}}{1+\exp\left(2.60-\frac{\ln\sqrt{s_{NN}}}{0.45}\right)},\\ \mu_B\left(\sqrt{s_{NN}}\right)&=\frac{1.3075\ \mathrm{GeV}}{1+0.288\sqrt{s_{NN}}}. \tag*{(57)} \end{aligned}

The values of yfoy_{\mathrm{fo}} are obtained by interpolating several values extracted from the hydrodynamic simulations in Ref. [13], as shown in Figure F1. We then obtain ξeq\xi_{\mathrm{eq}} from the universal three-dimensional Ising model and then map the Ising

Figure F1. Interpolated spacetime-rapidity extent (−yfo,yfo)(-y_{\mathrm{fo}}, y_{\mathrm{fo}}) of the freeze-out hypersurface as a function of sNN\sqrt{s_{NN}}. The solid blue curve shows the interpolation, while the red markers indicate the values extracted from the hydrodynamic simulations of Ref. [13].

variables onto the QCD (T,μB)(T,\mu_B) plane. Denoting the Ising scaling variables by rr and hh, we write

ξeqξ0=ΞIsing(r,h),(58)\frac{\xi_{\mathrm{eq}}}{\xi_0}=\Xi_{\mathrm{Ising}}(r,h), \tag*{(58)}

where ξ0\xi_0 is the noncritical baseline correlation length. In the numerical illustration we take ξ0=1 fm\xi_0=1\ \mathrm{fm}, and define the mapped critical region by

ξeqξ0>1.(59)\frac{\xi_{\mathrm{eq}}}{\xi_0}>1. \tag*{(59)}

The affine map is centered at the assumed CEP, (μc,Tc)(\mu_c,T_c), and is chosen as

h=T−TcΔT,r=μc−μBΔμB.(60)h=\frac{T-T_c}{\Delta T},\qquad r=\frac{\mu_c-\mu_B}{\Delta\mu_B}. \tag*{(60)}

Equivalently,

T=Tc+h ΔT,μB=μc−r ΔμB.(61)T=T_c+h\ \Delta T,\qquad\mu_B=\mu_c-r\ \Delta\mu_B. \tag*{(61)}

Thus the correlation length assigned to a point in the QCD phase diagram is

ξeq(μB,T)ξ0=ΞIsing(μc−μBΔμB,T−TcΔT).(62)\frac{\xi_{\mathrm{eq}}(\mu_B,T)}{\xi_0}=\Xi_{\mathrm{Ising}}\left(\frac{\mu_c-\mu_B}{\Delta\mu_B},\frac{T-T_c}{\Delta T}\right). \tag*{(62)}

For the phase diagram shown in the Letter, we use ΔT=0.02 GeV\Delta T=0.02\ \mathrm{GeV}, ΔμB=0.10 GeV\Delta\mu_B=0.10\ \mathrm{GeV}, and cap the finite-size critical enhancement by taking ΞIsing(0,0)=10\Xi_{\mathrm{Ising}}(0,0)=10. Outside the mapped critical region, the correlation length is reset to its baseline value,

ξeq(μB,T)ξ0={ΞIsing(μc−μBΔμB,T−TcΔT),(μB,T) inside the mapped critical region,1,(μB,T) outside the mapped critical region.(63)\frac{\xi_{\mathrm{eq}}(\mu_B,T)}{\xi_0}= \begin{cases} \Xi_{\mathrm{Ising}}\left(\dfrac{\mu_c-\mu_B}{\Delta\mu_B},\dfrac{T-T_c}{\Delta T}\right), & (\mu_B,T)\ \text{inside the mapped critical region},\\ 1, & (\mu_B,T)\ \text{outside the mapped critical region}. \end{cases} \tag*{(63)}

This construction is used only as an equilibrium proof of principle. It fixes the input ξeq\xi_{\mathrm{eq}} that controls the magnitude and position of the nonmonotonic response in Figure 2 of the main text.

With the above ingredients, one can also define and extract an effective correlation length by replacing ξeq\xi_{\mathrm{eq}} with ξeff\xi_{\mathrm{eff}} in Eq. (10). This gives the algebraic equation

C1,1αB1B2=πGp2mp28RK(T)T(ξeff2R2M1(2Rξeff)+ξeffR),(64)\frac{C_{1,1}^{\alpha}}{B_{1}B_{2}} = \frac{\pi G_{p}^{2}m_{p}^{2}}{8R}\frac{\mathcal{K}(T)}{T}\left(\frac{\xi_{\mathrm{eff}}^{2}}{R^{2}}\mathsf{M}_{1}\left(\frac{2R}{\xi_{\mathrm{eff}}}\right)+\frac{\xi_{\mathrm{eff}}}{R}\right), \tag*{(64)}

which can be solved for ξeff\xi_{\mathrm{eff}}. This effective correlation length incorporates the thermal-smearing effects contained in the left-hand side. Using the same rapidity-bin selections as in Figure 2, this extraction is shown in Figure F2. For

Image not extracted

Figure F2. The effective correlation length extracted from Eq. (9) in three rapidity bins, w=0.01w = 0.01 (dashed, red), w=0.10w = 0.10 (dotted, green), and w=0.50w = 0.50 (dot-dashed, orange). The benchmark ξeq/ξ0\xi_{\mathrm{eq}}/\xi_0 is shown as the thin blue solid line.

For very narrow rapidity windows, such as w=0.01w = 0.01, the extracted ξeff\xi_{\mathrm{eff}} closely follows the input equilibrium correlation length ξeq\xi_{\mathrm{eq}}. This is consistent with the expectation that thermal-smearing effects are negligible in sufficiently narrow rapidity windows. For wider windows, thermal smearing causes each momentum-rapidity bin to receive particles from a broader range of source rapidities, diluting the normalized critical response and lowering the extracted effective correlation length.

References

  1. [1]A. Bzdak, S. Esumi, V. Koch, J. Liao, M. Stephanov, and N. Xu, Mapping the Phases of Quantum Chromodynamics with Beam Energy Scan, Phys. Rept. 853, 1 (2020), arXiv:1906.00936 [nucl-th].DOI
  2. [2]A. Sorensen et al., Dense nuclear matter equation of state from heavy-ion collisions, Prog. Part. Nucl. Phys. 134, 104080 (2024), arXiv:2301.13253 [nucl-th].DOI
  3. [3]L. Du, A. Sorensen, and M. Stephanov, The QCD phase diagram and Beam Energy Scan physics: A theory overview, Int. J. Mod. Phys. E 33, 2430008 (2024), arXiv:2402.10183 [nucl-th].DOI
  4. [4]M. Arslandok et al., Hot QCD White Paper (2023), arXiv:2303.17254 [nucl-ex].arxiv.org/abs/2303.17254
  5. [5]Y. Aoki, G. Endrodi, Z. Fodor, S. D. Katz, and K. K. Szabo, The Order of the quantum chromodynamics transition predicted by the standard model of particle physics, Nature 443, 675 (2006), arXiv:hep-lat/0610114.
  6. [6]M. A. Stephanov, K. Rajagopal, and E. V. Shuryak, Signatures of the tricritical point in QCD, Phys. Rev. Lett. 81, 4816 (1998), arXiv:hep-ph/9806219.DOI
  7. [7]M. A. Stephanov, K. Rajagopal, and E. V. Shuryak, Event-by-event fluctuations in heavy ion collisions and the QCD critical point, Phys. Rev. D 60, 114028 (1999), arXiv:hep-ph/9903292.DOI
  8. [8]M. A. Stephanov, Non-Gaussian fluctuations near the QCD critical point, Phys. Rev. Lett. 102, 032301 (2009), arXiv:0809.3450 [hep-ph].DOI
  9. [9]C. Athanasiou, K. Rajagopal, and M. Stephanov, Using Higher Moments of Fluctuations and their Ratios in the Search for the QCD Critical Point, Phys. Rev. D 82, 074008 (2010), arXiv:1006.4636 [hep-ph].DOI
  10. [10]M. A. Stephanov, On the sign of kurtosis near the QCD critical point, Phys. Rev. Lett. 107, 052301 (2011), arXiv:1104.1627 [hep-ph]. Phys. Rev. C 110, 054910 (2024), arXiv:2310.06636 [nucl-th].DOI
  11. [11]B. Aboona et al. (STAR), Beam Energy Dependence of Fifth and Sixth-Order Net-Proton Number Fluctuations in Au+Au Collisions at RHIC, Phys. Rev. Lett. 130, 082301 (2023), [Erratum: Phys. Rev. Lett. 134, 139901 (2025)], arXiv:2207.09837 [nucl-ex].arxiv.org/abs/2207.09837
  12. [12]B. E. Aboona et al. (STAR), Precision Measurement of Net-Proton-Number Fluctuations in Au+Au Collisions at RHIC, Phys. Rev. Lett. 135, 142301 (2025), arXiv:2504.00817 [nucl-ex].
  13. [13]L. Du, H. Gao, S. Jeon, and C. Gale, Rapidity scan with multistage hydrodynamic and statistical thermal models, Phys. Rev. C 109, 014907 (2024), arXiv:2302.13852 [nucl-th].DOI
  14. [14]L. Du, C. Shen, S. Jeon, and C. Gale, Probing initial baryon stopping and equation of state with rapidity-dependent directed flow of identified particles, Phys. Rev. C 108, L041901 (2023), arXiv:2211.16408 [nucl-th].DOI
  15. [15]P. Braun-Munzinger, B. Friman, K. Redlich, A. Rustamov, and J. Stachel, Relativistic nuclear collisions: Establishing a non-critical baseline for fluctuation measurements, Nucl. Phys. A 1008, 122141 (2021), arXiv:2007.02463 [nucl-th].
  16. [16]V. Vovchenko, V. Koch, and C. Shen, Proton number cumulants and correlation functions in Au-Au collisions at sNN=7.7–200 GeV from hydrodynamics, Phys. Rev. C 105, 014904 (2022), arXiv:2107.00163 [hep-ph].DOI
  17. [17]J. Li, L. Du, and S. Shi, Rapidity scan approach for net-baryon cumulants with a statistical thermal model, Phys. Rev. C 109, 034906 (2024), arXiv:2311.11374 [nucl-th].DOI
  18. [18]R.-z. Zhao, S. Yin, S. Wu, L. Du, X. Luo, and W.-j. Fu, Critical net-proton number fluctuations with hydrodynamics (2026), arXiv:2607.26912 [nucl-th].arxiv.org/abs/2607.26912
  19. [19]A. Bzdak and V. Koch, Acceptance corrections to net baryon and net charge cumulants, Phys. Rev. C 86, 044904 (2012), arXiv:1206.4286 [nucl-th].DOI
  20. [20]A. Bzdak and V. Koch, Local Efficiency Corrections to Higher Order Cumulants, Phys. Rev. C 91, 027901 (2015), arXiv:1312.4574 [nucl-th].DOI
  21. [21]V. Skokov, B. Friman, and K. Redlich, Volume Fluctuations and Higher Order Cumulants of the Net Baryon Number, Phys. Rev. C 88, 034911 (2013), arXiv:1205.4756 [hep-ph].DOI
  22. [22]X. Luo, J. Xu, B. Mohanty, and N. Xu, Volume fluctuation and auto-correlation effects in the moment analysis of net-proton multiplicity distributions in heavy-ion collisions, J. Phys. G 40, 105104 (2013), arXiv:1302.2332 [nucl-ex].DOI
  23. [23]O. Savchuk, Net-proton fluctuations influenced by baryon stopping and quark deconfinement, Phys. Rev. C 111, 024913 (2025), arXiv:2407.17670 [hep-ph].arxiv.org/abs/2407.17670
  24. [24]M. Nahrgang, M. Bluhm, P. Alba, R. Bellwied, and C. Ratti, Impact of resonance regeneration and decay on the net-proton fluctuations in a hadron resonance gas, Eur. Phys. J. C 75, 573 (2015), arXiv:1402.1238 [hep-ph].DOI
  25. [25]M. Bluhm, M. Nahrgang, S. A. Bass, and T. Schaefer, Impact of resonance decays on critical point signals in net-proton fluctuations, Eur. Phys. J. C 77, 210 (2017), arXiv:1612.03889 [nucl-th].DOI
  26. [26]J. Hammelmann, M. Bluhm, M. Nahrgang, and H. Elfner, Fate of critical fluctuations in an interacting hadronic medium using maximum-entropy distributions,
  27. [27]Q. Lin, S. Yin, J. Li, H. Elfner, F. Rennecke, L.-G. Pang, and J. M. Pawlowski, Hadronic rescattering effects on net-proton cumulants from functional renormalization group calculations (2026), arXiv:2608.28299 [nucl-th].arxiv.org/abs/2608.28299
  28. [28]D. R. Brillinger, Time Series: Data Analysis and Theory, 2nd ed., Classics in Applied Mathematics, Vol. 36 (Society for Industrial and Applied Mathematics, Philadelphia, 2001).
  29. [29]Y.-K. Kong, Y. Kim, and M. Harada, Nuclear Matter and Finite Nuclei: Recent Studies Based on Parity Doublet Model, Symmetry 16, 1238 (2024), arXiv:2405.06956 [nucl-th].arxiv.org/abs/2405.06956
  30. [30]M. Kitazawa and M. Asakawa, Relation between baryon number fluctuations and experimentally observed proton number fluctuations in relativistic heavy ion collisions, Phys. Rev. C 86, 024904 (2012), [Erratum: Phys.Rev.C 86, 069902 (2012)], arXiv:1205.3292 [nucl-th].DOI
  31. [31]A. Baricz and T. K. Pogány, Functional inequalities for modified struve functions ii, Mathematical Inequalities & Applications 1387 (2014).
  32. [32]W.-j. Fu, J. M. Pawlowski, and F. Rennecke, QCD phase structure at finite temperature and density, Phys. Rev. D 101, 054032 (2020), arXiv:1909.02991 [hep-ph].DOI
  33. [33]J. Brewer, S. Mukherjee, K. Rajagopal, and Y. Yin, Searching for the QCD critical point via the rapidity dependence of cumulants, Phys. Rev. C 98, 061901 (2018), arXiv:1804.10215 [hep-ph].DOI
  34. [34]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 (2018), arXiv:1710.09425 [nucl-th].DOI
  35. [35]L. Du, Bulk medium properties of heavy-ion collisions from the beam energy scan with a multistage hydrodynamic model, Phys. Rev. C 110, 014904 (2024), arXiv:2401.00596 [hep-ph].DOI
  36. [36]H. De Vries, C. W. De Jager, and C. De Vries, Nuclear charge and magnetization density distribution parameters from elastic electron scattering, Atom. Data Nucl. Data Tabl. 36, 495 (1987).
  37. [37]M. Stephanov and Y. Yin, Hydrodynamics with parametric slowing down and fluctuations near the critical point, Phys. Rev. D 98, 036006 (2018), arXiv:1712.10305 [nucl-th].DOI
  38. [38]S. Mukherjee, R. Venugopalan, and Y. Yin, Real time evolution of non-Gaussian cumulants in the QCD critical regime, Phys. Rev. C 92, 034912 (2015), arXiv:1506.00645 [hep-ph].DOI
  39. [39]S. Mukherjee, R. Venugopalan, and Y. Yin, Universal off-equilibrium scaling of critical cumulants in the QCD phase diagram, Phys. Rev. Lett. 117, 222301 (2016), arXiv:1605.09341 [hep-ph].DOI
  40. [40]L. Du, U. Heinz, K. Rajagopal, and Y. Yin, Fluctuation dynamics near the QCD critical point, Phys. Rev. C 102, 054911 (2020), arXiv:2004.02719 [nucl-th].DOI
  41. [41]M. Bluhm et al., Dynamics of critical fluctuations: Theory – phenomenology – heavy-ion collisions, Nucl. Phys. A 1003, 122016 (2020), arXiv:2001.08831 [nucl-th].arxiv.org/abs/2001.08831
  42. [42]X. An, G. Başar, M. Stephanov, and H.-U. Yee, Evolution of Non-Gaussian Hydrodynamic Fluctuations, Phys. Rev. Lett. 127, 072301 (2021), arXiv:2009.10742 [hep-th].DOI
  43. [43]M. Pradeep, K. Rajagopal, M. Stephanov, and Y. Yin, Freezing out fluctuations in Hydro+ near the QCD critical point, Phys. Rev. D 106, 036017 (2022), arXiv:2204.00639 [hep-ph].
  44. [44]X. An et al., The BEST framework for the search for the QCD critical point and the chiral magnetic effect, Nucl. Phys. A 1017, 122343 (2022), arXiv:2108.13867 [nucl-th].DOI
  45. [45]M. Nahrgang, S. Leupold, C. Herold, and M. Bleichner, Nonequilibrium chiral fluid dynamics including dissipation and noise, Phys. Rev. C 84, 024912 (2011), arXiv:1105.0622 [nucl-th].DOI
  46. [46]M. Nahrgang, M. Bluhm, T. Schaefer, and S. A. Bass, Diffusive dynamics of critical fluctuations near the QCD critical point, Phys. Rev. D 99, 116015 (2019), arXiv:1804.05728 [nucl-th].
  47. [47]L. Du, X. An, and S. Wu, Baryon transport and the QCD critical point, Phys. Rev. C 104, 064904 (2021), arXiv:2107.02302 [hep-ph].DOI
  48. [48]N. Abasi, X. An, and S. Wu, Cumulant dynamics in finite-memory diffusion (2026), arXiv:2606.02446 [hep-th].
  49. [49]B. Ling and M. A. Stephanov, Acceptance dependence of fluctuation measures near the QCD critical point, Phys. Rev. C 93, 034915 (2016), arXiv:1512.09125 [nucl-th].DOI
  50. [50]A. Bzdak and V. Koch, Rapidity dependence of proton cumulants and correlation functions, Phys. Rev. C 96, 054905 (2017), arXiv:1707.02640 [nucl-th].
  51. [51]J. Adamczyk-Musch et al. (HADES), Proton-number fluctuations in √sNN =2.4 GeV Au + Au collisions studied with the High-Acceptance DiElectron Spectrometer (HADES), Phys. Rev. C 102, 024914 (2020), arXiv:2002.08701 [nucl-ex].
  52. [52]M. Abdallah et al. (STAR), Cumulants and correlation functions of net-proton, proton, and antiproton multiplicity distributions in Au+Au collisions at energies available at the BNL Relativistic Heavy Ion Collider, Phys. Rev. C 104, 024902 (2021), [Erratum: Phys.Rev.C 111, 029902 (2025)], arXiv:2101.12413 [nucl-ex].DOI
  53. [53]M. Gadzicki and P. Seyboth, Search for Critical Behaviour of Strongly Interacting Matter at the CERN Super Proton Synchrotron, Acta Phys. Polon. B 47, 1201 (2016), arXiv:1506.08141 [nucl-ex].
  54. [54]P. Senger (CBM), Astrophysics in the Laboratory—The CBM Experiment at FAIR, Particles 3, 320 (2020), arXiv:2004.11214 [nucl-ex].DOI
  55. [55]D. Hu, X. Wang, M. Shao, Y. Zhou, S. Ye, L. Zhao, Y. Sun, J. Lu, and H. Xu, Design and performance testing of a T0 detector for the CSR External-target Experiment, Nucl. Instrum. Meth. A 1057, 168773 (2023), arXiv:2304.02944 [physics.ins-det].arxiv.org/abs/2304.02944
  56. [56]V. Kolesnikov (MPD), Prospects for Heavy-Ion Physics with the MPD Detector at NICA, Universe 4, 145 (2018).DOI

Paper details

vlnc:W_9r2v8gae
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 11

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