Thin and dressed Polyakov loops from spectral sums of lattice differential operators
Erek Bilgici, Christian Hagen, Falk Bruckmann, Christof Gattringer
aa r X i v : . [ h e p - l a t ] O c t Thin and dressed Polyakov loops from spectralsums of lattice differential operators
Christian Hagen ∗ and Falk Bruckmann Institut für Theoretische Physik, Universität Regensburg, D-93040 Regensburg, Germany
Erek Bilgici ∗ and Christof Gattringer Institut für Physik, FB Theoretische Physik, Universität Graz, A-8010 Graz, AustriaE-mails: [email protected], [email protected],[email protected], [email protected]
We represent thin and dressed Polyakov loops as spectral sums of eigenvalues of differential op-erators on the lattice. For that purpose we calculate complete sets of eigenvalues of the staggeredDirac and the covariant Laplace operator for several temporal boundary conditions. The spec-tra from different boundary conditions can be combined such that they represent single (thin)Polyakov loops, or a collection of loops (dressed Polyakov loops). We analyze the role of theeigenvalues in the spectral sums below and above the critical temperature.
The XXV International Symposium on Lattice Field TheoryJuly 30 - August 4 2007Regensburg, Germany ∗ Based on presentations by Erek Bilgici and Christian Hagen. c (cid:13) Copyright owned by the author(s) under the terms of the Creative Commons Attribution-NonCommercial-ShareAlikeLicence. http://pos.sissa.it/ hin and dressed Polyakov loops
1. Motivation
The phenomenology of QCD is governed by two prominent features, confinement and sponta-neous breaking of chiral symmetry. As one increases the temperature T above some critical value T c , the theory becomes deconfined and chiral symmetry is restored. This suggests that there couldbe a relation between the two phenomena. Establishing or ruling out such a relation would be amajor insight into key mechanisms of QCD.The finite temperature transition of pure gauge theory, where the system changes from theconfined ( T < T c ) into the deconfined phase ( T > T c ) , can be understood as spontaneous breakingof the center symmetry [1], and the Polyakov loop is a suitable order parameter with h P i = T < T c and h P i 6 = T > T c (see Fig. 1). Concerning chiral symmetry breaking the orderparameter is given by the chiral condensate with h ¯ yy i 6 = T < T c and h ¯ yy i = T > T c .The chiral condensate in turn is related to the spectral density of the Dirac operator by the Banks-Casher formula [2]: h ¯ yy i = − pr ( ) , where r ( ) is the spectral density at the origin. Whenincreasing the temperature above the critical value, the spectral density at the origin vanishes andwith it the chiral condensate (see Fig. 2).In a series of recent papers, [3]-[6], the Polyakov loop, or other quantities that serve as orderparameters for the breaking of the center symmetry, were related to spectral sums of differen-tial operators on the lattice, in particular the Dirac and the covariant Laplace operators. In thisway also confinement is related to spectral quantities, as is chiral symmetry breaking through theBanks Casher formula. Since eigenvalues can be divided into IR and UV regions in a natural way,the spectral representations allow one to analyze whether confinement is dominated by IR or UVmodes. In this paper we review and extend our contributions to this enterprise.
2. Derivation of spectral sums
The thin Polyakov loop averaged over space is given by P = V (cid:229) ~ x L ( ~ x ) with L ( ~ x ) = Tr c " N t (cid:213) x = U ( ~ x , x ) , (2.1)where L ( ~ x ) is a straight line of links in temporal direction closed around the periodic boundary and Tr c denotes the color trace. Obviously, this quantity has a rather singular support. Due to this factit is known to have a poor continuum limit with large renormalization effects.To address this problem, below we will define a new observable P ( q ) which we refer to as the dressed Polyakov loop with winding number q . It is a collection of loops all with the same number q of windings around compactified time. So, for example, P ( ) is a sum of the thin Polyakov loop andall other loops winding once around the lattice in time direction, where the latter loops may havearbitrary complicated detours in spatial directions and are weighted proportional to their length.In the following two sections we derive the spectral sums for thin and dressed Polyakov loops.For this purpose it is convenient to summarize our conventions: The staggered Dirac operator is2 hin and dressed Polyakov loops -0.2 -0.1 0 0.1 0.2 0.3-0.3-0.2-0.100.10.20.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.3-0.2-0.100.10.20.3T < T c T > T c T » T c Figure 1:
Scatter plot for the thin Polyakov loop in the complex plane for an ensemble of quenched SU(3)configurations. As the temperature increases (left to right) the Polyakov loop acquires a non-vanishingexpectation value. c T > T c T » T c Figure 2:
Histograms of the spectral density of the staggered lattice Dirac operator as a function of themodulus of the eigenvalues. Below the critical temperature the spectral density at the origin is non-zero.Near T c the density at the origin decreases and vanishes above the critical temperature. given by D ( x , y ) = m d x , y + (cid:229) m = h m ( x ) (cid:2) U m ( x ) d x + ˆ m , y − U m ( x − ˆ m ) † d x − ˆ m , y (cid:3) , (2.2)where h m ( x ) = ( − ) x + x + ... + x m − is the staggered phase. The covariant Laplace operator on thelattice is defined as D ( x , y ) = ( + m ) d x , y − (cid:229) m = (cid:2) U m ( x ) d x + ˆ m , y + U m ( x − ˆ m ) † d x − ˆ m , y (cid:3) . (2.3)Below we will use the staggered Dirac operator for representing the thin Polyakov loop, while thedressed loops are constructed from the Laplace operator. We stress, however, that it is an easyexercise to generalize our formulas, such that both the thin and the dressed Polyakov loops can beobtained from both differential operators. The derivation can be extended further to arbitrary latticeoperators with only nearest neighbor interaction (such as Domain Wall Fermions). The advantageof these operators is that they are numerically cheap.3 hin and dressed Polyakov loops In order to connect the thin Polyakov loop to spectral sums we follow the derivation in [3]. Weconsider powers of the staggered Dirac operator (2.2). The n -th power of the operator then containspaths up to length n , dressed with a product of n links. For the special case where we consider theoperator to the power N t , where N t is the time extent of our lattice, we obtain Tr c D N t ( x , x ) = N t Tr c N t (cid:213) s = U ( ~ x , s ) − N t Tr c N t − (cid:213) s = U ( ~ x , N t − s ) † (2.4) + other loops, trivially closed = N t (cid:2) L ( ~ x ) − L ∗ ( ~ x ) (cid:3) + other loops, trivially closed . (2.5)From the set of all loops that contribute we have singled out the thin Polyakov loop and its complexconjugate. These two straight loops are the only ones that can close around compactified time. Allothers are ’trivially closed’, i.e., they do not wrap around the lattice in temporal direction.The key observation is that a change of the temporal boundary conditions U ( ~ x , N t ) → zU ( ~ x , N t ) , | z | = , (2.6)where z is a complex phase, only affects the Polyakov loops L −→ z L , (2.7)while the trivially closed loops remain unchanged. After such a change the expression (2.5) reads Tr c D N t z ( x , x ) = N t (cid:2) zL ( ~ x ) − z ∗ L ∗ ( ~ x ) (cid:3) + other loops , (2.8)where D z is the Dirac operator on a configuration where the temporal boundary conditions arechanged by a factor z . Averaging this expression over space and time we obtain Tr D N t z = V N t ( zP − z ∗ P ∗ + X ) , (2.9)where X is the sum of all trivially closed paths. We make use of this behavior to cancel the un-wanted trivial contributions and project onto the thin Polyakov loop by taking linear combinations(coefficients a i ) of this expression for three different boundary conditions z , z and z P ! = N t (cid:229) i = a i Tr ( D N t z i )= P ( z a + z a + z a ) | {z } ! = − P ∗ ( z ∗ a + z ∗ a + z ∗ a ) | {z } ! = + X ( a + a + a ) | {z } ! = . (2.10)This leads to a set of linear equations for the coefficients a , a and a z z z z ∗ z ∗ z ∗ a a a = . (2.11)4 hin and dressed Polyakov loops One can show that this equation has a unique solution as long as the boundary conditions z , z and z are different. For the particular choice z = z = z ∗ = e i p ≡ z the thin Polyakov loop isthen given by P = N t V " (cid:229) i (cid:16) l ( i ) (cid:17) N t + z ∗ (cid:229) i (cid:16) l ( i ) z (cid:17) N t + z (cid:229) i (cid:16) l ( i ) z ∗ (cid:17) N t , (2.12)where the l ( i ) z are the eigenvalues of the staggered Dirac operator for a given boundary condition z . Similar expressions can be obtained for other differential operators on the lattice. In order to obtain the spectral sum for the dressed Polyakov loops we now use the Laplace operatorand apply a different strategy than used for the thin loops. First, we rewrite the covariant Laplaceoperator on the lattice from Eq. (2.3) in the following way D = k [ − k H ] , H ( x , y ) = (cid:229) m = h U m ( x ) d x + ˆ m , y + U m ( x − ˆ m ) † d x − ˆ m , y i , (2.13)where the hopping matrix H collects all terms which connect nearest neighbors on the lattice andthe hopping parameter k is related to the bare mass m via k = ( + m ) − . The inverse Laplaceoperator D − ( x , x ) can be expressed as geometric series in terms of powers of H , D − ( x , x ) = ¥ (cid:229) j = k j + H j ( x , x ) . (2.14)Since the hopping matrix H contains only terms that connect nearest neighbors, the powers H j ( x , x ) in (2.14) correspond to closed loops of length j , again dressed with the corresponding products ofthe link variables. Thus we can organize D − in terms of loops and order these with respect to theirwinding around time. We obtain (after taking the trace)Tr D − = (cid:229) x Tr c D − ( x , x ) = (cid:229) n ∈ Z (cid:229) l ∈ L n k | l | + Tr c (cid:213) ( y , m ) ∈ l U m ( y ) , (2.15)where L n is the set of loops that wind exactly n -times around the time direction and | l | denotesthe length of a loop l . It is important that now the sum on the right hand side contains loopsof arbitrary length, with all possible numbers of windings. To disentangle the different windingnumbers we introduce U(1)-valued temporal boundary conditions for the Laplace operator whichare most conveniently implemented by the replacement U ( ~ x , N t ) −→ e i j U ( ~ x , N t ) , (2.16)which corresponds to a parameterization z = e i j in (2.6). Each loop acquires a phase factor e i j n ,corresponding to its winding number n :Tr (cid:16) D j (cid:17) − = (cid:229) n ∈ Z e i j n (cid:229) l ∈ L n k | l | + Tr c (cid:213) ( y , m ) ∈ l U m ( y ) . (2.17)5 hin and dressed Polyakov loops Integrating over j with a factor of e − i j q projects to winding number q and we end up with thedressed Polyakov loop P ( q ) (which we normalize with V = N N t ), P ( q ) ≡ V Z p − p d j p e − i j q Tr ( D j ) − = V (cid:229) l ∈ L q k | l | + Tr c (cid:213) ( y , m ) ∈ l U m ( y ) . (2.18)After the projection the loops still can be arbitrarily long but all have the same winding number q .The individual loops are weighted with a factor k | l | + , where | l | is the length of the loop. This leadsto an exponential suppression of long loops and the rate of the suppression is determined by thevalue of k . Nevertheless arbitrary long loops may contribute and the formula (2.18) is a sensibledefinition of dressed loops.The final step to make the connection to spectral properties of the staggered Dirac operator isto write the trace over D − as a spectral sum giving P ( q ) = V Z p − p d j p e − i j q V (cid:229) j = k − kl ( j ) j = V (cid:229) l ∈ L q k | l | + Tr c (cid:213) ( y , m ) ∈ l U m ( y ) , (2.19)where l ( j ) j is the j -th eigenvalue of the hopping matrix H (2.13) for boundary condition angle j .The dressed Polyakov loop transforms in the same way as the thin Polyakov loop. Thus the dressedPolyakov loops are proper order parameters for the breaking of the center symmetry. They arealso not as singular as the thin Polyakov loops and thus are expected to have a better continuumlimit with smaller renormalization effects. In an analogous way dressed Polyakov loops can beconstructed also with the eigenvalues of the Dirac operator instead of the Laplace spectrum.We stress that both equations (2.12) and (2.19) are exact results that hold for individual gaugeconfigurations. However, for a practical evaluation with numerically generated gauge configura-tions the two formulas have a different status. Equation (2.12) for the thin loops holds to ma-chine precision when the result of the spectral sum is compared to the direct evaluation of the thinPolyakov loop from the gluonic definition (2.1). The expression (2.19) for the dressed loop on theother hand contains an integral over a continuum of boundary conditions. That this integral canindeed be approximated numerically in a sensible way will be demonstrated in the next section. b L × N t a[fm] T[MeV]7 .
00 6 × , × . ( ) .
60 6 × , × . ( ) .
91 6 × , × . ( ) .
40 12 × . ( ) .
06 12 × . ( ) .
40 12 × . ( ) Table 1:
Ensembles used in our simulations. For the 6 × × × hin and dressed Polyakov loops
3. Numerical analysis
For the numerical analysis of our spectral sums (2.12) and (2.19) we use quenched SU ( ) gaugeconfigurations generated with the Lüscher-Weisz action [7]. We work on lattices with differentsizes ranging from 6 × × T c (seeTable 1). For those we compute complete spectra of both the staggered Dirac operator and thehopping matrix H of the Laplace operator for up to 16 different boundary condition. The latticespacing was set [8] with the Sommer parameter r = . f m . All error bars we show are statisticalerrors determined with single elimination jackknife. For the thin Polyakov loops we restrict ourselves to the three boundary conditions z = z = z ∗ = e i p ≡ z which are used to derive the spectral sum (2.12). A different choice wouldchange our results only marginally.After computing all eigenvalues of the staggered Dirac operators we order them w.r.t. to theirabsolute value and organize them in bins. For each bin we calculate a number of observables.The first observable is the distribution of the eigenvalues, which we show in Fig. 3 for differentspatial and temporal extents of the lattice and for three different temperatures above, below andapproximately at the critical value T c . The l.h.s. set of those plots shows the distribution as afunction of the size of the eigenvalues in lattice units, while the r.h.s. plots are in MeV. One canclearly see that the spectral density at the origin vanishes and chiral symmetry is restored as oneincreases the temperature above the critical value. The density of eigenvalues reaches a maximumafter about three quarters of the eigenvalues and then quickly drops towards the UV cutoff.Fig. 3 is particularly important when one analyzes which part of the spectrum contributes tothe spectral sum (2.12) for the thin Polyakov loop. For the interpretation of such an analysis it isnecessary to take into account the global distribution of the eigenvalues, and this is exactly whatFig. 3 provides.In the spectral sum (2.12) the thin Polyakov loop emerges through a shift of the eigenvalues asthe boundary condition is changed. It is interesting to analyze how strongly different eigenvaluesof the spectrum are shifted when the boundary condition is changed. We quantify this question bystudying the averaged shift of the eigenvalues given by s ( l ) = ( | l − l z | + | l − l z ∗ | + | l z − l z ∗ | ) / . (3.1)Since above the phase transition the thin Polyakov loop obtains a finite value while in the confinedphase it is approximately zero, the shift of the eigenvalues should change as one increases thetemperature. This is exactly what we find in our results for the average shift s ( l ) presented inFig. 4). The individual data points are obtained by averaging over all eigenvalues in a bin and thenover all configurations in the ensemble.It is obvious, that over the whole range of the spectrum the average shift is almost zero belowthe critical temperature. For high temperatures, however, we find large shifts for low-lying eigen-values and little or no shift when the UV region is approached. The small shift at the UV end ofthe spectrum disappears when finer and larger lattices are used.7 hin and dressed Polyakov loops < D n / D | l | > T > T c T » T c T < T c x4 < D n / D | l | > < D n / D | l | > x4 < D n / D | l | > a | l | < D n / D | l | > x6 | l | [MeV] < D n / D | l | > Figure 3:
Distribution of the eigenvalues of the staggered Dirac operator for different lattices and differenttemperatures. The l.h.s. plots are in lattice units, while on the r.h.s. we use MeV.
After understanding the global distribution of the eigenvalues and their shift under a changeof the boundary conditions, we can start to analyze the individual contributions to the spectral sum(2.12). This individual contribution is given by c ( l ) = N t V h ( l ) N t + z ∗ ( l z ) N t + z ( l z ∗ ) N t i . (3.2)Our results are shown in Fig. 5, where we plot the absolute value of the individual contri-butions after normalizing them with P , i.e., we divide by the value of the thin Polyakov loop. Itis obvious that mainly the UV modes contribute to the thin Polyakov loop. This observation isessentially independent of the temperature and the size of the lattice. This is a finding which is nota-priori obvious, given the fact that it is the IR modes that show the largest shifts in Fig. 4. On theother hand, the large power N t in the spectral sum (2.12) drastically enhances the contributions ofthe UV eigenvalues.Interesting are also the dips which form with increasing temperature at about 1 . a for N t = hin and dressed Polyakov loops a < s ( l ) > T > T c T » T c T < T c x4 < s ( l ) > [ M e V ] a < s ( l ) > x4 < s ( l ) > [ M e V ] a | l | a < s ( l ) > x6 | l | [MeV] < s ( l ) > [ M e V ] Figure 4:
Average shift of the eigenvalues calculated as defined in Eq. (3.1) as a function of l . Again weuse lattice units in the l.h.s. set of plots and MeV on the r.h.s. and at 0 . a and 1 . a for N t =
6. It can be shown that these dips are correlated with a change of thedirection of the shift of the eigenvalues for z -valued boundary conditions relative to the eigenvaluescomputed with periodic boundary conditions.The plot in Fig. 5 shows that the individual contributions are biggest at the UV end of thespectrum. On the other hand, Fig. 3 shows that for the largest eigenvalues the density shows asharp drop. It is an interesting question which effect, size of contribution versus density, wins out.This question can be addressed by considering only partial sums of (2.12). In Fig. 6, we showsuch partial sums as a function of the cutoff, i.e., as a function of the largest eigenvalue included in(2.12). Again we divide by the Polyakov loop P , such that when all eigenvalues are summed, i.e.,at the largest value of l in Fig. 6, the curve approaches 1.For the 12 × l . Thusthe behavior of the partial sums is non-monotonic. On the other hand, for small temperatures thePolyakov loop should vanish. Thus the overshooting is not problematic, since the accumulated9 hin and dressed Polyakov loops < | c ( l ) / P | > T > T c T » T c T < T c x4 < | c ( l ) / P | > < | c ( l ) / P | > x4 < | c ( l ) / P | > a | l | < | c ( l ) / P | > x6 | l | [MeV] < | c ( l ) / P | > Figure 5:
Contribution (3.2) of individual eigenvalues to the spectral sum for the thin Polyakov loop givenby Eq. (2.12). As before we use lattice units in the l.h.s. set of plots and MeV on the r.h.s. contributions stay approximately zero over the whole range of the spectrum. We also find that thethin Polyakov loop is recovered only after all UV modes are included in the spectral sum. Againwe see the appearance of dips at exactly the same values which we have already identified in Fig. 5.So far we have only considered absolute values of quantities. However, the Polyakov loop andalso the spectral sums are complex numbers and we can analyze their relative phase. In Fig. 7 weplot the angle
D f between the truncated sums and the resulting thin Polyakov loop as a functionof the cutoff used in the truncated sum. We show data for a 12 × × T > T c . We find that on the former lattice the accumulated contributions start with the wrongsign and recover the correct sign at the end. On the other lattice, although starting and ending withthe correct sign, there is an intermediate region in which the truncated sum has the wrong sign.A comparison with lattices of different spatial size but the same temporal extent reveals, that thisbehavior depends only on the temporal extent.A possible explanation for this phenomenon is the fact that the eigenvalues l of the staggeredDirac operator are purely imaginary. Thus l n is a real positive number, while l n + is real but10 hin and dressed Polyakov loops < | S c ( l ) / P | > T > T c T = T c T < T c x4 < | S c ( l ) / P | > < | S c ( l ) / P | > x4 < | S c ( l ) / P | > a | l | < | S c ( l ) / P | > x6 | l | [MeV] < | S c ( l ) / P | > Figure 6:
Partial sums of the spectral formula (2.12), plotted as a function of the cutoff. negative. If we consider only the first term in the truncated sum we find: c ( l ( ) ) = N t V (cid:20)(cid:16) l ( ) (cid:17) N t + z ∗ (cid:16) l ( ) z (cid:17) N t + z (cid:16) l ( ) z ∗ (cid:17) N t (cid:21) = N t V [ a + z ∗ ˜ a z + z ˜ a z ∗ ] ˜ a z ∗ ≈ ˜ a z ≡ ˜ a = N t V [ a + ( z ∗ + z ) ˜ a ]= N t V [ a − ˜ a ] ( < N t = n as a , ˜ a > > N t = n + a , ˜ a < , (3.3)where we assume that the boundary conditions with phases z and z ∗ lead to approximately the samelowest eigenvalue ˜ a , which is larger in size than the eigenvalue at periodic boundary conditions[9, 5]. Interestingly, this last relation between the relative sizes of the eigenvalues for non-trivial andperiodic boundary conditions seems to change in certain regions in the spectrum of the staggeredDirac operator, thus creating the observed sign change in the truncated sums. The points wherethese sign changes occur depend only on the time extend of the lattice.11 hin and dressed Polyakov loops < | Df | > wrong right < | Df | > x4 wrong right a | l | < | Df | > right wrong | l | [MeV] < | Df | > x6 wrongright Figure 7:
Phase shift of the truncated sums relative to the thin Polyakov loop for T > T c . The upper plotsare for a 12 × × For the thin Polyakov loops the numerical analysis of the last section demonstrates that thespectral sums for the thin loops are predominantly built up from the UV modes. However, as wehave already speculated above, part of the UV dominance might come from the singular natureof the support of the thin loop. Thus it is interesting to perform the same spectral analysis forthe dressed Polyakov loop. Such an analysis is planned for a future publication. Here we focuson the feasibility of computing the integral over the boundary angle in (2.19) numerically andestablish that the dressed Polyakov loop is indeed an order parameter for the breaking of the centersymmetry.For the dressed Polyakov loops we compute complete spectra of H for 16 values of the bound-ary angle j ∈ [ , p ) . The integrand in the spectral sum (2.19) is given by (omitting the normaliza-tion and the phase exp ( − i j q ) ) S ( j ) = V V (cid:229) j = k − kl ( j ) j . (3.4)Integrating S ( j ) over j with the phase exp ( − i j q ) projects to the loops with winding number q .The important technical question is how smooth a function of j this integrand is.Fig. 8 shows the integrand S ( j ) averaged over 20 configurations above T c with real Polyakovloops (l.h.s. plots) and 20 configurations with complex Polyakov loops (r.h.s.). The individualcurves are for different values of the mass parameter m in the vicinity of the critical mass m c wherethe Laplace operator ceases to be invertible. In the bottom plots a constant was subtracted.12 hin and dressed Polyakov loops Figure 8:
The spectral sum S ( j ) for the Laplace operator as a function of j . The l.h.s. plots are froman average over 20 configurations with real Polyakov loop, the r.h.s. plots are for 20 configurations with aPolyakov loop with phase ∼ exp ( i p / ) . In the top plots the full sum S ( j ) is shown, while in the bottomplots the constant c = R d j S L ( j ) was subtracted. The plots show a clear cosine behavior for the ensemble with real Polyakov loops (l.h.s.) and ashifted cosine for the complex ensemble (r.h.s.). For low-lying modes a similar dependence on thephase at the boundary was observed in [10]. It is obvious that dividing the interval [ , p ) into 16sub-intervals should be sufficient for a reliable numerical estimate of the j -integral in the case of q =
1, where the phase factor is simply cos j − i sin j . This is the most interesting case of dressedPolyakov loops with a single winding.Fig. 8 establishes that a numerical integration over j is feasible. We implement this integrationfor the case of q = P ( ) transformsunder center transformations in the same way as the thin Polyakov loop P defined in (2.1). Thus itshould be an order parameter for breaking of center symmetry, but one even expects that it showsa similar behavior as shown in Fig. 1 for the thin loop. In order to test this expectation we use 20configurations below T c and 20 above T c . The latter all have a Polyakov loop which is essentiallyreal, i.e., configurations in the right leg of the distribution shown in the r . h . s . plot of Fig. 1. Onthese two sets of configurations we compare the thin loop with the dressed loop at two differentvalues of k . The corresponding results in the complex plane are shown in Fig. 9.The plot shows clearly that for T < T c both, the thin and the dressed Polyakov loops clusteraround the origin, while above T c they develop a non-vanishing value near the positive real axis.Even the pattern in the distribution of the data points is similar for thin and dressed loops.13 hin and dressed Polyakov loops Figure 9:
Scatter plots of thin (l.h.s. plot) and unscaled dressed Polyakov loops (center and l.h.s. plots)below (blue symbols) and above T c (red symbols) for 20 configurations on a 12 × The analysis of Fig. 9 shows that the dressed Polyakov loop (2.19) is a suitable order parameterfor the breaking of center symmetry. It is less singular and is expected to have a proper continuumlimit. We are currently analyzing the volume and temperature dependence of the dressed loops andwill then return to studying which part of the spectrum predominantly contributes to the spectralsum for the dressed loops (2.19).
Acknowledgments
We thank Pierre van Baal, Tamas Kovacs, Kurt Langfeld, Wolfgang Söldner and AndreasWipf for stimulating discussions. Erek Bilgici has been supported by Fonds zur Förderung derwissenschaftlichen Forschung (FWF DK W1203-N08) and by NAWI-Graz. Christian Hagen hasbeen supported by BMBF.
References [1] L. D. McLerran and B. Svetitsky, Phys. Rev. D , 450 (1981).[2] T. Banks and A. Casher, Nucl. Phys. B , 103 (1980).[3] C. Gattringer, Phys. Rev. Lett. , 032003 (2006) [arXiv:hep-lat/0605018].[4] F. Bruckmann, C. Gattringer and C. Hagen, Phys. Lett. B , 56 (2007) [arXiv:hep-lat/0612020].[5] F. Synatschke, A. Wipf and C. Wozar, Phys. Rev. D , 114003 (2007) [arXiv:hep-lat/0703018].[6] W. Söldner, PoS(LAT2007) 222.[7] M. Lüscher and P. Weisz, Commun. Math. Phys. , 59 (1985); Err.: , 433 (1985); G. Curci,P. Menotti, and G. Paffuti, Phys. Lett. B , 205 (1983); Err.: B , 516 (1984).[8] C. Gattringer, R. Hoffmann, and S. Schaefer, Phys. Rev. D , 094503 (2002).[9] T. Kovacs, private communication.[10] F. Bruckmann and E. M. Ilgenfritz, Phys. Rev. D , 30 (2003) [arXiv:hep-lat/0212029]., 30 (2003) [arXiv:hep-lat/0212029].