WorksPhysics and Astronomy
Canonical partition function and equation of state in the heavy quark region of finite-density lattice gauge theory
Abstract
We investigate the phase structure and equations of state in lattice gauge theory at finite temperature and finite density, focusing on the canonical partition function obtained by decomposing the grand partition function with respect to particle number. We demonstrate that when quark masses are heavy, the complex phase of the quark determinant, which causes the sign problem, can be controlled by decomposing the grand partition function according to particle number. In this study, we propose an algorithm to calculate the canonical partition function for heavy-quark, high-density effective theory and test it for the first time using one-flavor gauge theory, a case where the sign problem arises. When the grand partition function of is decomposed into canonical partition functions, the sign problem is serious in odd-particle sectors, while it is mild in even-particle sectors. The odd-particle sectors also expose a fundamental issue: center symmetry makes the canonical partition function of vanish unless the particle number is an integer multiple of . Although appropriate in the confinement phase, this restriction must be treated carefully in the deconfinement phase. We resolve the issue by introducing an infinitesimal external field, calculate the canonical partition functions by Monte Carlo simulation, reconstruct the grand partition function, and determine thermodynamic observables over a wide range of temperature and density. Furthermore, we find that the chemical-potential dependence of the equation of state at low temperatures exhibits behavior characteristic of the hadron resonance gas model, and that no softening of the QCD matter occurs at finite density.
E-mail: [email protected]
© Copyright owned by the author(s) under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND 4.0). All rights for text and data mining, AI training, and similar technologies for commercial purposes, are reserved. ISSN 1824-8039. Published by SISSA Medialab.
Fugacity expansion and center symmetry
In lattice QCD at finite density, the fermion determinant is a complex number, which prevents the direct application of Monte Carlo methods. This makes numerical studies at finite density difficult. We attempt to calculate the equation of state at finite density by performing the fugacity expansion [1]. The grand partition function is given by the path integral over the gauge fields and is related to the canonical partition function at fixed particle number through the following fugacity expansion:
where is the fugacity factor. When the quark determinant is expanded in terms of the hopping parameter , the resulting terms can be expressed as a sum of products of Wilson loops. At finite temperature, the temporal direction is compactified with periodic boundary conditions. Classifying the terms according to their net winding number around the temporal direction, we obtain the following fugacity expansion:
where we denote the coefficient of as .
The Polyakov loop is the order parameter of center symmetry: in the symmetric phase and in the spontaneously broken phase. The reason acts as an order parameter is that it winds around the time direction. Like , , which has winding number , transforms nontrivially under the center symmetry. Under the center transformation, the Polyakov loop and fixed-number determinant transform as
for gauge theory. Consequently, like , must vanish unless is a multiple of , since and . Because is the sum of Boltzmann weights for a fixed particle number , this implies that states in which the particle number is not a multiple of cannot exist. Furthermore, the complex phase is expected to be one of the origins of the sign problem.
On the other hand, in the deconfinement phase, states in which the particle number is not a multiple of should be realized. In general, to determine whether a quantity that has a vanishing expectation value because of a symmetry acquires a nonzero value through spontaneous symmetry breaking, one must introduce an infinitesimal external field to break the symmetry, and then take the limit where the volume goes to infinity and the external field goes to zero. Therefore, for such quantities, the expectation value is determined solely by the response to an external field. By setting to zero those components that vanish due to the complex phase arising from the center transformation, one may resolve the sign problem associated with that phase. Furthermore, the probability of sampling states whose particle number is not a multiple of , a key indicator of the deconfinement phase transition, can be calculated by imposing an external field. It is anticipated that this approach will enable the calculation of the partition function and the equation of state. In this study, to verify the validity of this method, we calculate the partition function and the equation of state for a simplified version of finite-density lattice gauge theory: specifically, one-flavor, heavy-dense lattice gauge theory, in which the sign problem arises.
Canonical partition function in SU(2) gauge theory
From Eq. (2), we express the canonical partition function as using the partition function and expectation value of quenched QCD, and . For one-flavor , the center transformation gives and . Thus, the ensemble average cancels exactly for odd , whereas even sectors do not suffer this sign cancellation. To break the center symmetry explicitly, we add an external source . We define
For small , can be interpreted as the determinant of an additional heavy fermion. The lattice size is . Averaging over each configuration and its center transform gives
This averaging procedure makes quantities that must vanish by symmetry exactly zero, even when their direct Monte Carlo calculations have large statistical errors. In particular, vanishes for odd in the limit . For small , the ratios required to reconstruct the grand partition function are independent of :
Heavy-dense effective theory and numerical method
To explicitly calculate the canonical partition function, we discuss an effective theory for heavy-quark, high-density gauge theory. The Wilson fermion matrix is
We use the heavy-quark, high-density limit , , with fixed. Neglecting spatial and backward temporal hopping terms gives heavy-dense QCD [3, 4, 5, 6]. The determinant is
where , . Using in Eq. (7),
The coefficients are obtained as follows. Assign an index to each spatial lattice site, and let denote the coefficient obtained after multiplying by the factors from point 1 up to point . Using , can be obtained recursively as follows:
For ,
Here, for all and the initial values are for . These recursive calculations generate all coefficients on every gauge configuration. After the last point, . In particular, and . The ratio of to is therefore
The canonical functions are obtained by summing these logarithmic ratios from for even and from for odd .
Overlap problem When is large, an overlap problem arises. Since approximately , configurations near contribute almost nothing to the Monte Carlo statistical sum. We therefore perform simulations incorporating an external Polyakov loop field,
and use the reweighting method, . When estimating the optimal source using a mean-field approximation for the case , we obtain . With this value, the condition is satisfied, ensuring that configurations for each are sampled with equal probability in a Monte Carlo simulation. Then, the statistical error is minimized. The left panel of Figure 1 shows the relative statistical error in as a function of at and on a lattice. The green arrows represent the optimal values of obtained by the mean-field approximation. For each there is an optimal with a small statistical error. We perform simulations for several values and calculate over a wide range of particle numbers by combining measurable windows for each .
Figure 1. Left: Relative statistical error in versus particle number . Middle: Discrete derivative of the canonical partition function at (blue) and (red). Right: An enlarged view of the low-density region in the middle figure.
Canonical partition functions and grand-canonical reconstruction
The simulations are performed on a lattice. The deconfinement transition is at ; we compare results at (confinement) and those at (deconfinement). The middle panel of Figure 1 shows the discrete derivative, . The blue curve represents the results for , and the red curve represents the results for . The maximum number of particles is , which occurs when the lattice is completely filled with particles. In the simulations, 10,000 configurations are used for each point; calculations are performed while varying the parameter , and the results with the smallest errors are then combined to cover the entire density range. As seen in this figure, the heavy-dense effective theory possesses high-density/low-density symmetry [6]. The right panel is an enlarged view of the region where is small. Circle symbols are the results for even values of , while square symbols are the results for odd values of . This figure shows that no significant difference is observed between the results for even and those for odd .
Separating even and odd sectors, the normalized grand partition function can be written as
Since and , the grand partition function can be reconstructed using the data for . We note that the odd-particle contribution is proportional to the order parameter . Thus the odd sectors do not contribute in the confinement phase because , which expresses the absence of a single-fermion contribution, while they survive in the deconfinement phase.
Equation of state and order parameters
Pressure and particle-number density are defined by and . The dimensionless density and the fermionic contribution to the pressure are
where is the pressure in the quenched theory. More explicitly, differentiating the grand partition function gives
We plot the pressure and the particle-number density in the left and right panels of Figure 2, respectively. The blue curve shows the results at calculated on a lattice, while the red curve shows the results at . The horizontal axis is . As shown in Figure 2, the density rises rapidly near and approaches fermion saturation at sufficiently large . The change in density in the confinement phase () is more rapid than that in the deconfinement phase (). When , the density saturates and remains constant, while the pressure increases linearly with because .
Figure 2. Fermionic contribution to the pressure (left) and particle-number density (right) as a function of at and on the lattice.
In the confinement phase, in Eq. (13) and only even contribute. In gauge theory, baryons are composed of two quarks; the absence of any contribution from states with an odd number of quarks is consistent with color confinement. In the deconfinement phase, on the other hand, takes a nonzero value, and thus contributions from odd also appear. This is consistent with the fact that single-quark excitations become allowed in the deconfined phase. The inset in the right panel of Figure 2 enlarges the low-density region for , , , and . In this region, the density increases exponentially from a value consistent with zero as the chemical potential is increased, even at the larger values of .
Speed of sound The speed of sound is an important quantity when discussing the softness of the matter. We use the following nonrelativistic definition:
Here, is the effective mass of the particles. The speed of sound characterizes the stiffness of the matter: a larger corresponds to a stiffer equation of state. This quantity can be calculated from Figure 2 by determining the slope in a plot where is on the vertical axis and is on the horizontal axis, with held constant for the partial differentiation. The results obtained on a lattice are shown in the left panel of Figure 3 as the dimensionless quantity . Over most of the density range, this quantity is nearly constant, but it diverges near fermion saturation. The inset is an enlarged view of the low-density region; in the confinement phase, the value is approximately over a wide range of , while in the deconfinement phase, it is approximately unity3. blue, light blue, green, and red symbols correspond to the results for , 2.2, 2.3, and 2.5, respectively. We also performed simulations using , 6, 8, and 12 at , which is the deconfinement transition point for [7]. These correspond to , 2/3, 1/2, and 1/3, respectively. We use 100,000 configurations for each simulation. The results are shown in Figure 3 (right). In the low-temperature phase the temperature dependence of the speed of sound is small.
Figure 3. Left: Speed of sound versus on the lattice at (blue), 2.2 (light blue), 2.3 (green), and 2.5 (red). Right: The results on the lattices with (red), 6 (green), 8 (light blue), and 12 (blue). The horizontal axis is .
The result in the confined phase can be understood in terms of the hadron resonance gas model. In the hadron resonance gas model, the pressure is expressed as a sum over meson and baryon states; its dependence on the chemical potential, after subtracting a constant, is proportional to [8]. Here, is the number of colors. In the heavy-dense effective theory used in this study, the effects of antibaryons are neglected; consequently, the chemical-potential-dependent part of the pressure is proportional to . In that case, the particle-number density, , is proportional to . Figure 4 shows the logarithm of the density as a function of . The left panel shows the results for the confinement phase at . We fit the data and determine the slope. The fitted slope is 2.0119, in good agreement with the hadron resonance gas prediction . On the other hand, the right panel shows the results for the deconfinement phase at , where the slope is 1.0173. This implies that, at high temperatures, particles consisting of a single quark are present. Using as the parameter, the speed of sound in the confined phase is
Figure 4. versus at (left) and 2.5 (right). The red line shows the fitting result.
These results are consistent with the results at low temperatures shown in Figure 3. Thus, the speed of sound does not decrease with increasing density, and no softening of the matter is observed.
Polyakov loop Next, we discuss the Polyakov loop, an indicator of confinement. The Polyakov loop at finite density is reconstructed from the coefficients as follows:
where and . We calculate the Polyakov loop using the data and plot the results in Figure 5 (left). When is small, takes a finite value above the phase transition point, , and is zero below it. At high densities, however, the value becomes finite for all . Even in the region , where the Polyakov loop takes on a finite value, the density remains zero as long as the chemical potential is below the threshold. It can be seen that and exhibit significantly different behaviors.
Figure 5. Polyakov loop (left) and chiral condensate (right) as functions of at , , , and .
Chiral condensate Finally, we plot the chiral condensate at (blue), (light blue), (green), and (red) in Figure 5 (right). In the heavy-dense effective theory, is given by
where is the particle number per spatial lattice site in lattice units. Thus the condensate decreases from as increases. This behavior is qualitatively similar to the recent tensor-renormalization-group result at and [9]. However, because the fermions are heavy in this study, the condensate does not vanish. Since the maximum value of is at saturation, remains finite and approaches .
Conclusions
Focusing on the canonical partition function, we investigated the phase structure and the equation of state of lattice gauge theory at finite temperature and finite density. We proposed an algorithm for calculating the canonical partition function in the heavy-dense effective theory and tested it first in one-flavor gauge theory. Due to the center symmetry, the signs in the odd-particle sector alternate, leading to a severe sign problem, whereas no such cancellations occur in the even sector. The sign problem was resolved by introducing an infinitesimal external field. Furthermore, by adjusting this external field, the overlap problem was also resolved. The reconstructed grand partition function yields the density, pressure, Polyakov loop, sound speed, and chiral condensate across the low- to high-density region. In the low-temperature confined phase, the particle-number density is consistent with the hadron resonance gas model. No softening of the equation of state is observed before the system reaches fermion saturation. Extending this method to gauge theory, where three center sectors and a genuinely complex phase must be treated, is the next important step.
References
- [1]S. Ejiri, “Canonical partition function and finite density phase transition in lattice QCD,” Phys. Rev. D 78, 074507 (2008).DOI
- [2]S. Ejiri, “Canonical partition function and center symmetry breaking in finite density lattice gauge theories,” Phys. Rev. D 106, 114505 (2022).DOI
- [3]T. C. Blum, J. E. Hetrick and D. Toussaint, “High density QCD with static quarks,” Phys. Rev. Lett. 76, 1019-1022 (1996).pubmed.ncbi.nlm.nih.gov/10061613
- [4]G. Aarts and I. O. Stamatescu, “Stochastic quantization at finite chemical potential,” JHEP 09, 018 (2008).DOI
- [5]G. Aarts, F. Attanasio, B. Jäger and D. Sexty, “The QCD phase diagram in the limit of heavy quarks using complex Langevin dynamics,” JHEP 09, 087 (2016).DOI
- [6]S. Ejiri and M. Koiida, “Phase structure of heavy dense lattice QCD and the three-state Potts model,” Phys. Rev. D 113, 074515 (2026).DOI
- [7]S. Ejiri, Y. Iwasaki and K. Kanaya, “Nonperturbative determination of anisotropy coefficients in lattice gauge theories,” Phys. Rev. D 58, 094505 (1998).
- [8]S. Ejiri, F. Karsch and K. Redlich, “Hadronic fluctuations at the QCD phase transition,” Phys. Lett. B 633, 275-282 (2006).DOI
- [9]Y. Sugimoto, S. Akiyama and Y. Kuramashi, “Phase structure of (3+1)-dimensional dense two-color QCD at T=0 in the strong coupling limit with the tensor renormalization group,” Phys. Rev. D 113, 034503 (2026).




