Investigating the excitation function of HBT radii for Lévy-stable sources

Abstract

One of the main goals of today’s heavy-ion physics research is to explore the phase diagram of strongly interacting matter and search for signs of the possible critical endpoint on the QCD phase diagram. Femtoscopy is among the important tools used for this endeavor; there have been indications that combinations of femtoscopic radii parameters (referred to as HBT radii for identical boson pairs) can be related to the system’s emission duration. An apparent non-monotonic behavior in their excitation function thus might signal the location of the critical point. In this paper, we show that conclusions drawn from the results obtained with a Gaussian approximation for the pion source shape might be altered if one utilizes a more general Lévy-stable source description. We find that the characteristic size of the pion source function is strongly connected to the shape of the source and its possible power-law behavior. Taking this into account properly changes the observed behavior of the excitation function.

[first]organization=ELTE Eötvös Loránd University,addressline=Pázmány Péter sétány 1/A, city=Budapest, postcode=1117, country=Hungary

1 Introduction

Exploring the phase diagram of strongly interacting matter is a cornerstone of heavy-ion physics research [1]. The quest to pinpoint the critical endpoint on the Quantum Chromodynamics (QCD) phase diagram is ongoing, and femtoscopy serves as a pivotal tool in this endeavor [2]. It has been demonstrated that femtoscopic radii parameters (also called HBT radii, after R. Hanbury-Brown and R. Q. Twiss [3]) extracted from two-particle interferometry measurements are sensitive to the emission duration of heavy-ion collisions [4]. In the RHIC Beam Energy Scan program [2], a non-monotonic collision energy (and consequently baryochemical potential) dependence of these parameters was observed [5], which was suggested to be a sign of possible critical behavior [6]. These measurements assumed a Gaussian distribution for the shape of the two-particle emission distribution. Recent results from both experiment [7, 8, 9, 10, 11] and phenomenology [12, 13, 14, 15] showed that the shape of the pion pair-source follows a Lévy-stable distribution [16] and exhibits a power-law tail. While new measurements in this direction are ongoing, in this paper we describe a method to estimate the effect of the different source shapes on the previous results, based on already published data. In particular, we demonstrate that using a Lévy-stable assumption for the pion pair source can drastically change the trends of the radii parameters as a function of collision energy, and thus the previous conclusions drawn from the Gaussian measurements might need to be reiterated.

1.1 Basic definitions

Utilizing the smoothness approximation [17], where the members of the investigated particle pair have approximately identical momenta, the two-particle correlation function (as a function of their relative momentum $\boldsymbol{q}$ and average momentum $\boldsymbol{K}$) is defined as

$$C_{2}(\boldsymbol{q},\boldsymbol{K})=\int d^{3}rD(\boldsymbol{r},\boldsymbol{K})|\psi^{(2)}_{\boldsymbol{q}}(\boldsymbol{r})|^{2},$$ψぷさい start_POSTSUPERSCRIPT ( 2 ) end_POSTSUPERSCRIPT start_POSTSUBSCRIPT bold_italic_q end_POSTSUBSCRIPT ( bold_italic_r ) | start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT , (1)

where $D(\boldsymbol{r},\boldsymbol{K})$ is the pair source distribution (also called spatial correlation function or pair emission function), and $\psi^{(2)}_{\boldsymbol{q}}(\boldsymbol{r})$ψぷさい start_POSTSUPERSCRIPT ( 2 ) end_POSTSUPERSCRIPT start_POSTSUBSCRIPT bold_italic_q end_POSTSUBSCRIPT ( bold_italic_r ) is the symmetrized pair wave function [18]. Neglecting final-state interactions, the modulus square of the wave function is simply ${1+\cos{(\boldsymbol{q}\boldsymbol{r}})}$, thus the correlation function is in direct connection with the Fourier-transform of the spatial correlation function:

$$C_{2}^{(0)}(\boldsymbol{q},\boldsymbol{K})\approx 1+\widetilde{D}(\boldsymbol{q},\boldsymbol{K}),$$ (2)

where the $(0)$ superscript denotes the lack of final-state interactions, and $\widetilde{D}$ denotes the Fourier transform of $D$ in its first variable.

If one assumes a shape or functional form for the pair source distribution, one can calculate the shape of the correlation function as well. The dependence on $\boldsymbol{K}$ is then usually understood to be realized through the parameters of the source, such as its width. Subsequently, one can test the initial assumption via fits to experimentally measured momentum correlation functions. Recent investigations [7, 9, 10, 8] showed that assuming a so-called Lévy-stable distribution (a generalization of the Gaussian distribution) for the shape of the source might provide a statistically superior description to the measured two-pion correlation functions.

The three-dimensional symmetric Lévy-stable distribution is defined as [19]

$$\mathcal{L}(\alpha,\boldsymbol{R^{2}};\boldsymbol{x}) =\frac{1}{2\pi^{3}}{\int}d^{3}\boldsymbol{\omega}\exp{\left(i\boldsymbol{\omega}\boldsymbol{x}{-}\frac{1}{2}|\boldsymbol{\omega}^{T}\boldsymbol{R^{2}}\boldsymbol{\omega}|^{\alpha/2}\right)},$$αあるふぁ , bold_italic_R start_POSTSUPERSCRIPT bold_2 end_POSTSUPERSCRIPT ; bold_italic_x ) πぱい start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT end_ARG ∫ italic_d start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT bold_italic_ωおめが roman_exp ( italic_i bold_italic_ωおめが bold_italic_x - divide start_ARG 1 end_ARG start_ARG 2 end_ARG | bold_italic_ωおめが start_POSTSUPERSCRIPT italic_T end_POSTSUPERSCRIPT bold_italic_R start_POSTSUPERSCRIPT bold_2 end_POSTSUPERSCRIPT bold_italic_ωおめが | start_POSTSUPERSCRIPT italic_αあるふぁ / 2 end_POSTSUPERSCRIPT ) , (3)

$$\textnormal{where }\boldsymbol{R^{2}} =\begin{pmatrix}R_{o}^{2}&R_{os}^{2}&R_{ol}^{2}\\ R_{os}^{2}&R_{s}^{2}&R_{sl}^{2}\\ R_{ol}^{2}&R_{sl}^{2}&R_{l}^{2}\end{pmatrix}.$$ (4)

Here, $\boldsymbol{R^{2}}$ is a positive definite matrix containing the HBT-radii parameters (also called Lévy-scale parameters), and $\alpha$αあるふぁ is the Lévy exponent. In the above definition of $\boldsymbol{R^{2}}$ we already utilized the usual notation of the Bertsch-Pratt $out-side-long$ coordinate system [20, 21], where the $out$ direction represents the direction of the average transverse momentum of the particle pair, the $long$ direction is the beam direction, and the $side$ is perpendicular to both. In the case of an azimuthally integrated experimental analysis, it is usual to neglect the off-diagonal terms, and in the following we will also do this.

Utilizing such a distribution as the pair-source function, in the interaction-free case, the correlation function takes the following form:

$$C_{2}(q_{o},q_{s},q_{l})=1+\lambda\exp{\left(-|R_{o}^{2}q_{o}^{2}+R_{s}^{2}q_{s}^{2}+R_{l}^{2}q_{l}^{2}|^{\alpha/2}\right)}.$$λらむだ roman_exp ( - | italic_R start_POSTSUBSCRIPT italic_o end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_q start_POSTSUBSCRIPT italic_o end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT + italic_R start_POSTSUBSCRIPT italic_s end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_q start_POSTSUBSCRIPT italic_s end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT + italic_R start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT italic_q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT | start_POSTSUPERSCRIPT italic_αあるふぁ / 2 end_POSTSUPERSCRIPT ) . (5)

Here, we also introduced the $\lambda$λらむだ correlation strength or intercept parameter [22, 23], that corresponds to the value of the correlation at zero relative momentum (in an interaction-free case). In the case of $\alpha=2$αあるふぁ = 2, the shape of the pair-source and the correlation function is both a Gaussian. In the case of $\alpha<2$αあるふぁ < 2, the pair-source exhibits a power-law behavior, and the correlation function takes the form of a stretched-exponential function. Of course, in the case of an experimental analysis, one needs to properly account for the final-state Coulomb interaction. For Lévy-stable sources, details of such calculations were discussed in Refs. [24, 25, 26]; in the following, we use the results presented in Ref. [26].

2 Scaling method

In the first published experimental heavy-ion analysis utilizing a Lévy-stable description for the pion pair source [7], a scaling variable was found, defined as

$$\widehat{R}=\frac{R}{\lambda\cdot(1+\alpha)},$$λらむだ ⋅ ( 1 + italic_αあるふぁ ) end_ARG , (6)

where $R$ is the one-dimensional Lévy scale parameter.

When fitting the same correlation function with different fixed $\alpha$αあるふぁ values, this parameter seemed remarkably stable. This implies that if we know the $\lambda$λらむだ and $R$ results for a given $\alpha$αあるふぁ assumption, we can deduce the value of these for another $\alpha$αあるふぁ assumption. In particular, if the Gaussian parameters $R_{\rm G}$ and $\lambda_{\rm G}$λらむだ start_POSTSUBSCRIPT roman_G end_POSTSUBSCRIPT (obtained with an assumption of $\alpha=2$αあるふぁ = 2) are known and the $\lambda_{{\rm free}\alpha}$λらむだ start_POSTSUBSCRIPT roman_free italic_αあるふぁ end_POSTSUBSCRIPT and $\alpha$αあるふぁ values are also known for a free-$\alpha$αあるふぁ fit, then the value of $R_{{\rm free}\alpha}$αあるふぁ end_POSTSUBSCRIPT can be estimated based on

$$\widehat{R}_{\rm G}\equiv\frac{R_{\rm G}}{\lambda_{\rm G}\cdot(1+2)}=\widehat{R}_{{\rm free}\;\alpha}\equiv\frac{R_{{\rm free}\;\alpha}}{\lambda_{{\rm free}\;\alpha}\cdot(1+\alpha)}$$λらむだ start_POSTSUBSCRIPT roman_G end_POSTSUBSCRIPT ⋅ ( 1 + 2 ) end_ARG = over^ start_ARG italic_R end_ARG start_POSTSUBSCRIPT roman_free italic_αあるふぁ end_POSTSUBSCRIPT ≡ divide start_ARG italic_R start_POSTSUBSCRIPT roman_free italic_αあるふぁ end_POSTSUBSCRIPT end_ARG start_ARG italic_λらむだ start_POSTSUBSCRIPT roman_free italic_αあるふぁ end_POSTSUBSCRIPT ⋅ ( 1 + italic_αあるふぁ ) end_ARG (7)

as

$$R_{{\rm free}\;\alpha}=R_{\rm G}\frac{\lambda_{{\rm free}\;\alpha}\cdot(1+\alpha)}{\lambda_{\rm G}\cdot(1+2)}.$$αあるふぁ end_POSTSUBSCRIPT = italic_R start_POSTSUBSCRIPT roman_G end_POSTSUBSCRIPT divide start_ARG italic_λらむだ start_POSTSUBSCRIPT roman_free italic_αあるふぁ end_POSTSUBSCRIPT ⋅ ( 1 + italic_αあるふぁ ) end_ARG start_ARG italic_λらむだ start_POSTSUBSCRIPT roman_G end_POSTSUBSCRIPT ⋅ ( 1 + 2 ) end_ARG . (8)

To demonstrate the stability of this scaling variable, we take three example correlation functions from Refs. [7, 9] (from their respective HEPdata entries). Using the correlation function calculation method described in Ref. [26] and the software package given in Ref. [27], we fit these example correlation functions with different fixed $\alpha$αあるふぁ values, with the functional form

$$N\left[1-\lambda+\lambda(1+e^{-(qR)^{\alpha}})K_{C}(q;R,\alpha)\right](1+\epsilon q),$$λらむだ + italic_λらむだ ( 1 + italic_e start_POSTSUPERSCRIPT - ( italic_q italic_R ) start_POSTSUPERSCRIPT italic_αあるふぁ end_POSTSUPERSCRIPT end_POSTSUPERSCRIPT ) italic_K start_POSTSUBSCRIPT italic_C end_POSTSUBSCRIPT ( italic_q ; italic_R , italic_αあるふぁ ) ] ( 1 + italic_ϵ italic_q ) , (9)

where $K_{C}$ is the Coulomb correction as calculated in Ref. [26], $N$ is a normalization parameter and $\epsilon$ responsible for accounting for any residual, long-range background from non-femtoscopic effects (such as energy and momentum conservation, flow, or minijets). From the resulting fit parameters, we calculate the corresponding $\widehat{R}_{{\rm fix}\;\alpha}$αあるふぁ end_POSTSUBSCRIPT values. All fit parameters and fit quality indicators are shown in the Appendix. We find that, indeed, in the investigated range, the scaling variable $\widehat{R}$ remains approximately constant as a function of the fixed $\alpha$αあるふぁ value, as demonstrated in Figure 1.

Figure 1: Figure 2 of Ref. [9] and Figs. 3a-3b of Ref. [7]. See detailed parameters in Appendix.

While an experimental confirmation is not yet available, it is natural to assume that this scaling relation also holds for the three-dimensional radii parameters, at least approximately, as a smaller $\alpha$αあるふぁ leads to a more “peaked” correlation function, hence to describe the same data one requires a larger $R$ and a larger $\lambda$λらむだ value. We also assume that the $\alpha$αあるふぁ parameter measured under the assumption of a spherically symmetric source is identical with (or close to) the $\alpha$αあるふぁ obtained from a three-dimensional measurement (which does not assume a spherically symmetric source). Thus we can estimate how the radii would change from the fixed $\alpha=2$αあるふぁ = 2 Gaussian case to the free-$\alpha$αあるふぁ case using published data, as described in the next section.

3 Results and discussion

Measurements based on a Gaussian source assumption were performed in Ref. [5], where also the collision energy dependence of the source parameters (radii and $\lambda$λらむだ) was investigated at $K_{T}\approx 0.22$ GeV$/c^{2}$, where $K_{T}$ is the transverse component of $\boldsymbol{K}$. In this paper, a non-monotonic collision energy dependence of $\sqrt{R_{\rm o,G}^{2}-R_{\rm s,G}^{2}}$ was found, as mentioned above.

Figure 2: Scaled Lévy parameters $R_{\rm o,s,l}$, compared to the original Gaussian values from Ref. [5]

Based on STAR preliminary results presented at the Zimányi School 2023 and CPOD 2024 conferences, one can make a rough approximation of the collision energy dependence of the $\lambda_{{\rm free}\;\alpha}$λらむだ start_POSTSUBSCRIPT roman_free italic_αあるふぁ end_POSTSUBSCRIPT and $\alpha$αあるふぁ parameters (also at $K_{T}\approx 0.22$ GeV$/c^{2}$) in the form of

$$A_{\rm par}\left(\frac{\sqrt{s_{NN}}}{E_{0}}\right)^{B_{\rm par}},$$ (10)

with

$$A_{\alpha}=1.85,B_{\alpha}=-0.06,\;\;A_{\lambda}=0.6,B_{\lambda}=0.06,$$αあるふぁ end_POSTSUBSCRIPT = 1.85 , italic_B start_POSTSUBSCRIPT italic_αあるふぁ end_POSTSUBSCRIPT = - 0.06 , italic_A start_POSTSUBSCRIPT italic_λらむだ end_POSTSUBSCRIPT = 0.6 , italic_B start_POSTSUBSCRIPT italic_λらむだ end_POSTSUBSCRIPT = 0.06 , (11)

and $E_{0}=1$ GeV is introduced to remove the unit of $\sqrt{s_{NN}}$. The Gaussian measurements for $R_{G(o,s,l)}$ together with the corresponding $\lambda_{\rm G}$λらむだ start_POSTSUBSCRIPT roman_G end_POSTSUBSCRIPT values are available from Ref. [5]. Using these inputs, we calculated the corresponding $R_{{\rm free}\;\alpha(o,s,l)}$αあるふぁ ( italic_o , italic_s , italic_l ) end_POSTSUBSCRIPT values, as shown in Figure 2. The $\sqrt{R_{\rm o}^{2}-R_{\rm s}^{2}}$ difference is shown in Figure 3.

Figure 3: Scaled Lévy parameter difference $\sqrt{R_{\rm o}^{2}-R_{\rm s}^{2}}$, compared to the original Gaussian values from Ref. [5]

In line with the expected scaling behavior, the smaller $\alpha$αあるふぁ values increase the radii for the Lévy scaling predictions, and with this, the $out$ and $side$ difference also increases. However, it is interesting to note that the non-monotonic energy dependence is weakened, and the $out-side$ difference appears to be constant above $\sqrt{s_{NN}}=19$ GeV. These observations demonstrate the need for the three-dimensional HBT measurements utilizing Lévy-stable source distributions.

4 Summary and conclusions

We demonstrated based on three example correlation functions that the variable combination called $\widehat{R}$ is indeed a scaling variable, i.e., it does not strongly depend on the assumed $\alpha$αあるふぁ value. We used this scaling relation to infer HBT radii in the out-side-long system with a free $\alpha$αあるふぁ value from results with a Gaussian assumption ($\alpha=2$αあるふぁ = 2). Finally, we investigated the out-side difference as a proxy for emission duration. Our results indicate that the trend of this difference is strongly affected by the assumption for the shape, and unlike the case of a Gaussian assumption, there may be no non-monotonic behavior of the out-side difference as a function of collision energy. This underlines the importance of performing experimental measurements with an unrestricted $\alpha$αあるふぁ parameter.

Appendix A

Figs. 4, 5, and 6 show all fit parameters ($R$, $\lambda$λらむだ, $N$, $\epsilon$) and fit quality indicators ($\chi^{2}$χかい start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT, NDF, confidence level, fit status and covariance matrix status), along with the scaling variable $\widehat{R}$, for many cases of fixed $\alpha$αあるふぁ values from 0.8 to 2.0. Here NDF means the number of degrees of freedom (number of data points minus the number of free fit parameters), confidence level means the p-value of the fit (calculated from $\chi^{2}$χかい start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT and NDF), fit status and covariance matrix status come from the Minuit optimization library [28], and their best values are 0 and 3, respectively (indicating a fully converged fit and fully positive definite covariance matrix). These fit quality incicators are important as they demonstrate the good quality of the fits. Without them, one cannot be sure that the best fit parameters indeed do represent the data. These plots show that (especially for larger $\alpha$αあるふぁ, above $\sim$1.4) the scaling variable $\widehat{R}$ is indeed independent of the assumed fix $\alpha$αあるふぁ value, hence can be used to infer $R$ at a given $\alpha$αあるふぁ from $R$ at a different $\alpha$αあるふぁ.

Figure 4: Fit parameters and quality from various fixed-$\alpha$αあるふぁ fits to correlation function data of Figure 3a from Ref. [7]. The plots show parameters $R$ and $\lambda$λらむだ, fit normalization parameter $N$ and residual slope $\epsilon$, scaling parameter $\widehat{R}$, confidence level, $\chi^{2}$χかい start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT and NDF, as well as the “fit status” and “covariance matrix status” parameters.

Figure 5: Fit parameters and quality from various fixed-$\alpha$αあるふぁ fits to correlation function data of Figure 3b from Ref. [7], similarly to Fig. 4.

Figure 6: Fit parameters and quality from various fixed-$\alpha$αあるふぁ fits to correlation function data of Figure 2 from Ref. [9], similarly to Fig. 4.