Gauge-invariant screening masses and static quark free energies in $N_f = 2+1$ QCD at non-zero baryon density

We discuss the extension of gauge-invariant electric and magnetic screening masses in the Quark-Gluon Plasma to the case of a finite baryon density, defining them in terms of a matrix of Polyakov loop correlators. We present lattice results for $N_f=2+1$ QCD with physical quark masses, obtained using the imaginary chemical potential approach, which indicate that the screening masses increase as a function of $\mu_B$. A separate analysis is carried out for the theoretically interesting case $\mu_B/T=3 i \pi$, where charge conjugation is not explicitly broken and the usual definition of the screening masses can be used for temperatures below the Roberge-Weiss transition. Finally, we investigate the dependence of the static quark free energy on the baryon chemical potential, showing that it is a decreasing function of $\mu_B$ which displays a peculiar behavior as the pseudocritical transition temperature at $\mu_B=0$ is approached.


I. INTRODUCTION
Static color charges are useful probes of the properties of strongly interacting matter. At low temperature, the potential between a heavy quark-antiquark pair, which can be derived from Wilson loop expectation values or from Polyakov loop correlators, can be used to investigate the confining properties of the medium and the spectrum of heavy quark bound states. At high temperature, static charge interactions permit instead to investigate screening effects in the Quark-Gluon Plasma (QGP), which are at the basis of interesting phenomenology, like the dissociation of heavy quark bound states [1] (see Ref. [2] for a recent review). Moreover, the asymptotic (large distance) behavior of Polyakov loop correlators gives access to the free energy of static color charges, which is a useful probe for confinement/deconfinement.
Interactions between heavy quarks have been widely studied in lattice QCD simulations by means of Polyakov loop correlators, in particular by projecting over color group representations (e.g., singlet or octet) after proper gauge fixing. Gauge invariant observables can also be studied and, using charge conjugation symmetry, it is possible to build gauge invariant operators that couple only to the chromo-magnetic or the chromo-electric sector [3,4]. In the high temperature phase correla-tors of these observables permit to define in a gaugeinvariant and nonperturbative way magnetic and electric screening masses (inverse of the screening lengths), which have been the subject of recent lattice QCD investigations [5,6] Screening effects in the QGP are expected to be influenced by external parameters which can change the properties of the thermal medium. In general an increase of color screening effects (i.e. an increase of the screening masses) is expected as the system is driven deeper into the deconfined region. An example is the introduction of a magnetic background field B, which is known to induce a decrease of the pseudocritical temperature T c [7] and to affect the confining properties of QCD [8,9], favoring the onset of deconfinement. Magnetic and electric screening masses in the presence of an external background field have been determined by lattice simulations in Ref. [10] and have indeed been shown to be increasing functions of B, in agreement with analytical studies of screening effects in the QGP [11][12][13].
The baryon chemical potential µ B is another parameter of obvious phenomenological relevance. Also in this case one expects an increase of screening effects as a function of µ B , since a finite baryon density favors the onset of deconfinement; this is confirmed by perturbative predictions [14] and by lattice QCD studies considering correlators projected over color representations after gauge fixing [15]. In this case, however, when considering gauge invariant screening masses one has to face the problem that charge conjugation symmetry is explicitly broken by the presence of the baryon chemical potential, so that a clear separation into electric and magnetic sectors cannot be performed any more [4].
One of the purposes of this study is to propose an extension of the gauge invariant definition of the screening masses to the case µ B = 0, which is based on the analysis of a full matrix of Polyakov loop correlators, i.e. including the mixed electric-magnetic correlator which turns out to be non-zero when µ B = 0. Such an extension is then implemented in numerical simulations of N f = 2 + 1 QCD with imaginary values of the baryon chemical potential, so that the behavior of the gauge invariant screening masses as a function of µ B is finally obtained by analytic continuation to real µ B . In this way we will show that they increase as a function of µ B . A special attention and a separate discussion will be devoted to those values of the imaginary chemical potential for which an exact charge conjugation symmetry can be recovered (Roberge-Weiss transition points).
A second aim of our study is the investigation of the dependence on µ B of the free energy F Q of a static color charge. This is related, after proper renormalization [6,[16][17][18][19], to the large distance behavior of Polyakov correlators hence, by cluster property, to the Polyakov loop expectation value. This quantity has been extensively studied in the past, at finite T and zero chemical potential, for its connection to the confining properties of the thermal medium. Here we consider the part of the free energy which is related to the introduction of the baryon chemical potential, ∆F Q (T, µ B ) ≡ F Q (T, µ B ) − F Q (T, 0) and can be obtained by studying the ratio of Polyakov loop expectation values. Also in this case we will present numerical results obtained for imaginary values of µ B and then exploit analytic continuation to extract the dependence of ∆F Q (T, µ B ) for small values of µ B /T .
Simulations have been performed on a line of constant physics of N f = 2 + 1 QCD with physical quark masses, discretized via stout improved staggered fermions. Most results have been obtained on a 32 3 × 8 lattice and for temperature above the pseudocritical temperature T c . In particular, the screening masses have been investigated for temperatures ranging from 217 to 300 MeV, while a wider range has been explored to study the behavior of ∆F Q (T, µ B ). The paper is organized as follows. In Section II we review the definition of the gauge invariant screening masses in terms of Polyakov loop correlators and discuss the extension to the case of non-zero chemical potential. We then present the main features of the phase diagram at imaginary chemical potential and we describe our numerical setup. In Section III we present our results for the screening masses and the quark free energies as a function of µ B . Finally, in Section IV, we draw our conclusions.

II. OBSERVABLES AND NUMERICAL METHODS
Our investigation is based on lattice QCD simulations. As it is well known, the introduction of a baryon chemical potential µ B makes the Euclidean path integral measure complex, so that standard Monte-Carlo simulations are not feasible. A possible solution to this problem, which has been widely explored in the literature , is to perform simulations at imaginary values of the chemical potential, µ B = iµ B,I , then exploiting analytic continuation, which is expected to be valid at least for sufficiently small values of µ B /T . We will consider a theory with N f = 2 + 1 flavors (up, down, strange) and physical quark masses. In order to move along a line of non-zero baryon chemical potential, with electric and strangeness chemical potentials set to zero, we have chosen degenerate quark chemical potentials µ f,I = µ I ≡ µ B,I /3. As usual, in the discretized theory, the chemical potentials are introduced in the Dirac operator in an exponentiated form attached to temporal gauge links [44], in order to avoid the insurgence of ultraviolet divergencies in the continuum limit. In this way, the introduction of a non-zero µ B can also be viewed as a rotation of temporal boundary conditions for quark fields by a factor exp(µ B /(3T )) = exp(iµ I /T ).

A. Gauge invariant screening masses at µB = 0
It is well known that a perturbative definition of the screening masses of QCD [45][46][47] gets into trouble because of non-perturbative contributions arising in the finite temperature gluon propagator at the next-to-leading order, which are due to magnetostatic gluons and stems from the non-Abelian nature of the gauge group [48][49][50][51]. Nevertheless, it has been shown that well defined screening masses can be accessed by studying the large distance behaviour of suitable correlators of gauge invariant quantities, such as the Polyakov loops projected onto the electric and magnetic sectors [3,4]. In the continuum the Polyakov loop is defined as where P is the path-ordering operator and N c is the number of colors; on the lattice it is easily built in terms of gauge links in the temporal direction. As shown in Ref. [4] (and retraced in some recent lattice works [5,6,10]), it is possible to separate the contributions to screening of the color-magnetic and color-electric gluons by defining suitable combinations of Polyakov loop correlators. This is achieved by performing a symmetry decomposition involving the Euclidean time reversal (R) and the charge conjugation (C) operators. As a result [4,6,10], one finds that the relevant role is played by the fluctuations of the real and imaginary parts of the Polyakov loop, so that the correlators take contributions, respectively, only from the colormagnetic (R-even and C-even) and the color-electric (R-odd and C-odd) sectors. In the same way, it can be shown [6] that the cross sectors (R-even and C-odd or its opposite) are trivial: that manifests in the vanishing of the mixed correlators between the real and imaginary part of the Polyakov loop. At large distances, these correlators are expected to behave as [3,4] where m M (T ) and m E (T ) are the color-magnetic and color-electric screening masses. The considerations above are valid when charge conjugation C is an exact symmetry of the theory. The introduction of a baryon chemical potential µ B (or, in general, of chemical potentials coupled to quark number operators) breaks C explicitly, so that a trivial extension of the definition of magnetic and electric screening masses to finite density QCD is not possible [4]. What happens is that the electric and the magnetic sectors are not separated any more, a fact that manifests through the appearance of a non-zero mixed electric-magnetic correlator, which is defined as follows (we drop the dependence on T and µ B for the sake of readability): A non-zero value of this correlator means that the real and the imaginary part of the Polyakov loop do not undergo independent fluctuations any more. The correlator is obviously symmetric under exchange of electric and magnetic components, i.e. TrReL(0)TrImL(r) = TrImL(0)TrReL(r) so that, when charge conjugation is explicitly broken, one can actually define a symmetric matrix of correlators The practical effect of this mixing is that the asymptotic, large distance behavior of all correlators will now be dominated by a single mass: that will be confirmed explicitly by our numerical data in the following. The information about the second gauge invariant mass is now hidden in non-leading corrections to the asymptotic behavior of the correlators, which are usually difficult to be detected directly. This situation is quite common in the numerical investigation of the spectrum of quantum field theories, where one usually considers a set of mixed correlators in the same channel and needs to derive the masses of the lightest independent physical states coupled to them. A possible solution is to diagonalize the correlator matrix, i.e. to solve the eigenvalue equation where C stands for the matrix and the diagonal correla-tors C 1 and C 2 are easily found to be It is reasonable to expect the long distance behavior of the diagonalized correlators to be similar to that found for the electric and magnetic ones at zero µ B , i.e.
leading to the definition of two independent screening masses m 1 (T, µ B ) and m 2 (T, µ B ). Let us discuss the connection between this pair of states with those that can be determined at µ B = 0: if at µ B = 0 the excited state mass m * M in the magnetic sector is higher than m E , these masses satisfy the natural relations lim µB →0 m 1 (T, µ B ) = m M (T ) and lim µB →0 m 2 (T, µ B ) = m E (T ). In general it is not easy to extract the value of m * M with good precision, however we have verified that in all the cases explored in this paper our numerical data are consistent with m * M (µ B = 0) ≥ m E (µ B = 0). Let us conclude with some remark on the variational approach used in the context of hadron spectroscopy [52][53][54], which is different from the one described above. In that case, one solves a generalized eigenvalue problem where r 0 is a reference distance, which defines the socalled principal correlators λ 1/2 (r, r 0 ). We have verified that, in the explored cases, the two different approaches lead to consistent results. The results presented in the following will be based on the diagonal correlators C 1 (r) and C 2 (r).
B. The special case of the Roberge-Weiss point As we have already mentioned above, the introduction of an imaginary baryon chemical potential can be rephrased in terms of a modification of the temporal boundary conditions for all quark fields by a phase θ = µ I /T . That implies that an exact charge conjugation symmetry is recovered for special values of θ: θ = π is one example, however all values θ = (2k + 1)/(3π) with k integer are equivalent to each other after a global center transformation on gauge fields [55]. In the case θ = π the charge conjugation symmetry has the same form as for µ B = 0, and it can also be viewed as a switch from thermal antiperiodic to periodic boundary conditions for fermion fields; therefore we will take it as a reference in the following.
On the contrary, for T < T RW one can keep the standard definition of electric and magnetic sectors and ask what modifications they undergo with respect to the standard case at zero chemical potential. In particular, one expects a significant change in the behavior of the electric correlator, since the imaginary part of the Polyakov loop is the order parameter of the Roberge-Weiss transition at θ = π.
C. Dependence of the free energy on µB The free energy F Q of heavy quarks in the thermal medium can be inferred from the asymptotic behavior of the unsubtracted Polyakov loop correlator, i.e. the squared modulus of TrL , as follows: This definition is plagued by additive ultraviolet divergencies and needs renormalization [6,[16][17][18][19], which is usually performed by subtracting zero temperature contributions. However the introduction of chemical potentials in the discretized theory is not expected to introduce further divergencies, at least when this is done by exponentiating it in the temporal gauge links [44], since that amounts to just a change of the temporal boundary conditions for fermion fields, which has no effect at all on the theory in the zero temperature limit (at least for small enough chemical potentials).
For this reason, the contribution to the heavy quark free energy related to the introduction of a baryon chemical potential, where β is the inverse bare gauge coupling, is expected to be a renormalized quantity with a well defined continuum limit. That will be checked explicitly, based on our numerical results, in Section III B.

D. Simulation details
We adopted a rooted staggered fermion discretization of N f = 2 + 1 QCD. The partition function is where M f st and µ f,I are respectively the fermion matrix and the imaginary chemical potential for the quark flavor f , while is the tree level improved Symanzik gauge action [80,81] and W n×m i;µν stands for the trace of the n × m rectangular parallel transport starting from site i and spanning the µ-ν directions. The rooted staggered Dirac matrix is written in terms of two times stout smeared links U (2) i;ν with isotropic smearing parameter ρ = 0.15 [82]. Bare parameters entering the action have been chosen so as to move on a line of constant physics [83][84][85], with degenerate light quark masses, m u = m d = m l , a physical pion mass m π ≃ 135 MeV, and a physical strange-to-light quark mass ratio ratio m s /m l = 28.15.
Most of our Monte-Carlo simulations have been performed for a fixed value of N t , precisely on a 32 3 × 8 lattice. Screening masses have been measured for four values of the bare parameter β (corresponding to four different lattice spacings a and temperatures T = 1/(aN t )) and various imaginary chemical potentials µ I . The temperature range has been chosen in order to stay well above T RW , which for N t = 8 is T RW ≃ 200 MeV [67], so that the full range of imaginary chemical potentials, up to the first Roberge-Weiss transition line at µ I /T = π/3, is available for analytic continuation. A summary of the simulation parameters is reported in Table I.
No attempt to estimate the magnitude of discretization effects is performed regarding the screening masses, since we have investigated just one lattice spacing for each temperature; however we stress that results obtained at µ B = 0 [6] show that, in the same range of temperatures explored in our study, discretization errors for N t = 8 lattices are not large and within the statistical accuracy of the results that we are going to show.
Polyakov loop correlators have been measured, for each run, on a set of about 5 × 10 3 configurations separated by five molecular dynamics trajectories. As a noisereduction technique we have applied to temporal links 3 steps of APE smearing with parameter α = 0.5. Correlators have been extracted for generic orientations (i.e. not just along the lattice axes) and distances using a Fast Fourier Transform algorithm, then averaging correlators corresponding to equal distances and different orientations. A blocked jackknife resampling technique has been used in order to correctly estimate statistical errors.
In order to extract the screening masses, a fit procedure based on the model in Eq. (3) has been performed, estimating the statistical errors on the mass parameters I: Values of the bare coupling β we used for the determination of the screening masses, together with the corresponding lattice spacings a and temperatures T for our simulations on a 32 3 ×8 lattice. In each case we report also the explored values of the imaginary quark chemical potential. The lattice spacing determination has a systematic uncertainty of the order of 2-3% [84,85]. by resampling techniques. The errors reported in the following will also include systematic errors estimated by changing the range of fitted data points.
As for the determination of the static quark free energy ∆F Q (T, µ B ), given the better statistical accuracy which is reachable in this case even with limited statistics, we have explored a larger range of temperatures and also, in some cases, different values of N t , in order to check for finite cutoff corrections. All numerical simulations have been performed using an RHMC algorithm running on GPUs [86,87].

A. Screening masses
We start by discussing results obtained for Polyakov loop correlators and the gauge invariant screening masses. In Fig. 1 we report some of the correlators obtained for lowest explored temperature, T ≃ 217 MeV. At zero chemical potential the magnetic and electric correlators, C M + and C E − , show the standard behavior already clarified in previous studies [5,6,10]: the magnetic correlator is larger than the electric one, by around one order of magnitude, and decreases more slowly as a function of the distance, in agreement with the expected hierarchy m M < m E ; best fits according to Eq. (3) are reported together with the data points. The mixed electricmagnetic correlator is zero within statistical errors, as expected, and is not reported in the figure.
In Fig. 1 we report also correlators obtained at the same T for µ I > 0, in particular for µ I /T = π/3. In this case the situation is quite different. The mixed correlator C X turns out to be different from zero and, as a result of this mixing, the long-range behavior of the electric correlator is modified: indeed, apart from an overall factor, the magnetic and the electric correlators show the same behavior at large distances, which seems to be governed by the same long distance correlations length. This phenomenon is more clearly visible in Fig. 2, where the ratio C E − /C M + is reported as a function of r: while at µ I = 0 the ratio becomes compatible with zero at large  distances, at µ I = 0 it approaches a constant non-zero value, which indicates that the two correlators fall off in the same way and that their large distance behavior is dominated by one single mass. This mixing is eliminated when one considers the two diagonalized correlators C 1 and C 2 defined in Eq. (7), which permits to obtain two well distinct screening masses m 1 and m 2 . An example of such correlators is shown in Fig. 1, together with a best fit according to Eq. (8). The whole set of diagonalized correlators at T ≃ 217 MeV is shown in Fig. 3, for all the explored values of the imaginary chemical potential.
The gauge invariant screening masses that we have obtained are reported as a function of µ I /(πT ) and for various temperatures in Tab. II and plotted in Fig. 4. Actually, we report the quantity m 1/2 /T , since this dimensionless ratio is known to be almost independent of T at zero chemical potential. Reported error bars include also systematic errors related to the choice of fitting range.
The screening masses are expected to be even functions of µ B . Therefore, at the leading order in a Taylor expansion, we can write the following general ansatz: where µ B = 3iµ I . In our fit we have discarded the values obtained for µ I /(πT ) = 1/3, the reason being that for this value of the chemical potential the screening masses are expected to vanish with a non-analytical behavior as the temperature approaches the second order Roberge-Weiss endpoint from above [67].
As a matter of fact, it is possible to fit all data using a unique value for the quadratic coefficients, i.e. setting b 1/2 (T ) = b, where b is a T independent constant, obtaining b = 1.37 (36) with χ 2 = 15.24/23. A positive value b indicates that screening masses increase as a function of µ B : this is in qualitative agreement with the fact that the introduction of a finite baryon density tends to drive the system farther away from the confined phase. The fact that b is independent of T is in agreement with leading order perturbative estimates of the Debye screening mass [14] giving b = 1.5N f /(6 + N f ) = 0.5 for N f = 3, which is not far from our estimate. The fact that a common value of b describes both masses implies that the ratio m 1 /m 2 is independent of µ B : notice however that our data do not put a stringent constraint on this and there is room for different behaviors within the present errors.  Let us turn to a discussion of the results obtained for the dependence of the heavy quark free energy on the baryon chemical potential, ∆F Q (µ B , T ), defined in Eq. (11). To start with, we show in Fig. 5 the results obtained at T ≃ 251 MeV for two different values of the temporal extension, N t = 6, 8, and from both smeared and unsmeared Polyakov loops. The fact that all determinations agree within errors is a convincing test that this quantity is well defined and does not need renormalization, as expected. Moreover, data suggest that O(a 2 ) corrections are not significant, within present errors, already for N t = 6. It is not the aim of this study to provide a continuum extrapolation for ∆F Q (µ B , T ), therefore for the other temperatures we just provide results obtained for N t = 8, which are shown for some of the explored temperatures in Fig. 6. The free energy is expected to be an even and analytic function of the chemical potential, therefore fitting numerical data with a truncated Taylor expansion in (µ B /T ) 2 is the natural choice. Within the statistical accuracy of our data we are able to reliably estimate just the first term in the expansion, moreover we have noticed that the quality of the fit improves if one tries to fit directly the behavior of the squared ratio of Polyakov loops, i.e. the exponential of the free energy, according to Such a functional dependence, with higher order terms neglected, describes reasonably well data with µ I /T < π/3 at all temperatures: the results of our best fits are reported in Table III. The analytic continuation of ∆F Q (µ B , T ) to real chemical potentials according to Eq. (16) is shown in Fig. 7. The free energy of the static quark is a decreasing function of µ B , in agreement with the fact that a finite baryon chemical potential enhances deconfinement. Also the fact that the coefficient χ Q,µ 2 B becomes larger as T decreases can be understood qualitatively. Close to the pseudocritical temperature, the chemical potential acts as a transition driving parameter: if the Polyakov loop were an exact order parameter for deconfinement, its dependence on µ B would become sharper and eventually diverge at the transition. Of course this is not the case: the Polyakov   loop is not an order parameter and there is no real transition in QCD at the physical point; however it is reasonable to expect χ Q,µ 2 B to become larger and larger as T c is approached from above.
This behavior is clearly visible in Fig. 8, where we plot the inverse coefficient, 1/χ Q,µ 2 B as a function of T . It is striking to notice that, looking at the high temperature region, one would be tempted to predict a vanishing of 1/χ Q,µ 2 B in a region of temperatures around 150 MeV, i.e. roughly coinciding with T c . Of course the behavior is then smoothed out as T c is approached more closely. We cannot assert if this hints at a more strict connection between ∆F Q (µ B , T ) and the confinement/deconfinement transition; however it indicates that this is a quantity which is surely worth further investigation.

C. Magnetic and electric screening masses close to the Roberge-Weiss endpoint
As we have stressed above, values of the imaginary chemical potential which at high T lead to the Roberge-Weiss transition play a special role. Consider in particular the case µ I /T = π, which corresponds to a simple shift of fermionic boundary conditions from antiperiodic to periodic ones: charge conjugation symmetry is not explicitly broken in this case. For T > T RW it gets spontaneously broken, so that a mixing between the electric and the magnetic sectors appears anyway, however for T < T RW it is not and the standard definition of electric and magnetic correlators, with the associated screening  Fig. 9. The vertical gray band indicates the location of TRW for Nt = 8 [67]: notice that in this case we report just the statistical error stemming from the determination of the critical coupling in Ref. [67], and not the systematic one stemming from the determination of the overall physical scale, which is common to all data in this limited range of temperatures.
masses, is well posed, so that it is interesting to compare their behavior with the case of zero chemical potential. In Fig. 9 we show the behavior of such correlators for three values of T . A striking difference with respect to the µ I /T = 0 case is clearly visible: the hierarchy is inverted, with the electric correlators being larger than the magnetic ones by more than one order of magnitude and growing significantly as T approaches T RW , which for the 40 3 × 8 lattice that we have used for these measurements is T RW (N t = 8) ∼ 200 MeV [67].
The reason is easy to understand: the electric correlator is the two-point function of the imaginary part of the Polyakov loop (see Eq. (2)), which is also an order parameter for the Roberge-Weiss transition at µ I /T = π. Therefore, being directly connected to the order parameter of the transition, it is expected to undergo the most critical modifications, due to large distance fluctuations, as one approaches the critical point.
To better clarify this point, we show the values of the electric screening masses at µ I /T = π (defined as in Eq. (3)) as a function of T in Fig. 10. The electric mass decreases rapidly as T approaches T RW , seemingly approaching zero at the transition, as expected. Notice however that, within the present statistical errors and limited number of data points, we are not able to check whether we are already close enough to T RW to reveal the correct critical behavior predicted for the correlation length at the transition, which for this case is in the 3D Ising universality class [67].

IV. CONCLUSIONS
We have investigated the behavior of Polyakov loops and of Polyakov loop correlators as a function of the baryon chemical potential in the deconfined phase of QCD with N f = 2 + 1 flavors and physical quark masses.
We have discussed the extension of the concept of gauge invariant screening masses at finite baryon density, where charge conjugation symmetry is explicitly broken and a mixing between the electric and the magnetic sector appears, leading to the inapplicability of the standard definition of color-electric and color-magnetic masses. Such an extension can be given in terms of diagonalized (or alternatively principal) correlators derived from a 2 × 2 matrix of correlators involving the real and the imaginary part of the Polyakov loop. In this way one can obtain a consistent definition of two different gauge invariant screening masses which characterize the thermal medium in the presence of a finite baryon density.
In order to obtain a determination of such screening masses, we have considered the theory discretized on N t = 8 lattices by means of stout smeared rooted staggered fermions and a tree level Symanzik improved gauge action. Numerical simulations have been performed for imaginary values of the baryon chemical potential, then exploiting analytic continuation. Both screening masses show an increasing behavior as the baryon chemical potential is switched on, and the slope of the relative increase (see Eq. (15)) seems to be independent of T and of the same order of magnitude as that predicted for the Debye screening mass at the lowest order of perturbation theory. Future investigations should extend present results to different values of N t in order to provide a continuum extrapolation for the gauge invariant screening masses at finite baryon density.
We have also shown that for µ I /T = π and T < T RW , where charge conjugation symmetry is exact and the standard definitions of electric and magnetic screening masses still holds, the hierarchy of screening lengths is inverted, with the electric mass being the lowest one and approaching zero as T → T RW . This has been interpreted in terms of the direct coupling existing between the electric correlator and the order parameter for the Roberge-Weiss transition.
Finally, we have investigated the dependence of the static quark free energy on the chemical potential, which is defined by the large distance behavior of the Polyakov loop correlators. In particular, we have considered the free energy variation due to the introduction of the chemical potential, ∆F Q (µ B , T ), which is related to the ratio of Polyakov loops and is not expected to undergo additive renormalization: this has been explicitly verified by comparing data obtained for different values of N t and different amounts of smearing on Polyakov loops. The static free energy decreases as a function of the baryon chemical potential, as expected on general grounds, with a slope which increases significantly as the temperature approaches the pseudocritical temperature T c from above. That could hint at a more strict connection between ∆F Q (µ B , T ) and the confinement/deconfinement transition: future investigations should consider a more systematic study of this quantity and its extrapolation to the continuum limit.