Quark confinement from dyons Dmitri Diakonov Petersburg Nuclear Physics Institute QUARKS-08, May Gatchina

Confinement criteria in a pure glue theory (no dynamical quarks): 1) Average Polyakov line 2) Linear rising potential energy of static quark and antiquark 3) Area law for the average Wilson loop 4) Mass gap: no massless states, only massive glueballs

We shall consider quantum Yang-Mills theory at nonzero T, as we shall be interested not only in confinement at small T but also in the deconfinement phase transition at T>Tc. Quarks are switched off. According to Feynman, the partition function is given by a path integral over all connections periodic in imaginary time, with period 1/T : A helpful way to estimate integrals is by the saddle point method. Dyons are saddle points, i.e. field configurations satisfying the non-linear Maxwell equation:

Bogomolnyi-Prasad-Sommerfield monopoles or dyons are self-dual configurations of the Yang-Mills field, whose asymptotic electric and magnetic fields are Coulomb- like, and the eigenvalues of the Polyakov line are non-trivial. For the SU(N) gauge group there are N kinds of elementary dyons: Inside the dyons cores, whose size is, the field is large and, generally, time-dependent, the non-linearity is essential. Far away the field is weak and static. holonomy

In the saddle point method, one has to compute small-oscillation determinants about classical solutions. The small-oscillation determinant about a single dyon is infrared-divergent (because of the Coulomb asymptotics at infinity) isolated dyons are unacceptable, they are not saddle points. One has to take neutral clusters of N kinds of dyons. The corresponding exact solutions are known as Kraan-van Baal-Lee-Lu (KvBLL) calorons or instantons with non-trivial holonomy (1998). The KvBLL instantons generalize standard instantons to the case when the Polyakov loop (the holonomy) is nontrivial, The analytical solution shows what happens when dyons come close to each other:

Action density as function of time of three dyons of the SU(3) group. At large dyon separations, we have three static dyons. When dyons merge, they become a standard time-dependent instanton. In all cases the full action is the same. The small-oscillation determinant about KvBLL instantons is finite; computed exactly by D.D., Gromov, Petrov, Slizovskiy (2004) as function of separations between N dyons the phases of the Polyakov line temperature T, the renormalized scale parameter

The 1-loop statistical weight (or probablity) of an instanton with non-trivial holonomy: fugacity The expression for the metric of the moduli space G is exact, valid for all separations between dyons. If holonomy is trivial, or T -> 0, the measure reduces to that of the standard instanton, written in terms of center, size and orientations [D.D. and Gromov (2005)].

It has N degenerate minima when all are equal (mod 1) i.e. when the Polyakov loop belongs to one of the N elements of the group center: In perturbation theory, deviation from these values are forbidden as exp(- const.V). For confinement, one needs Tr L = 0, which is achieved at the maximum of the perturbative energy! The perturbative potential energy (it is present even in the absence of dyons ) as function of the Polyakov loop phases :

SU(2) SU(3) Perturbative potential energy has N minima corresponding to trivial holonomy: scale as T^4 However, the non-perturbative free energy of the ensemble of dyons has the minimum at Tr L = 0 ! At low T it wins confinement ! At T<Tc the dyon-induced free energy prevails and forces the system to pick the confining holonomy

To see it, one has to calculate the partition function of the grand canonical ensemble of an arbitrary number of dyons of N kinds and arbitrary s, and then minimize the free energy in s (and also compute the essential correlation functions). number of dyons of kind m 3d coordinate of the i-th dyon of kind m fugacity, function of T, moduli space metric, function of dyon separations

G is the moduli space metric tensor whose dimension equals the full # of dyons: Properties: 1)the metric is hyper-Kaehler (a very non-trivial requirement) 2) same-kind dyons repulse each other, whereas different-kind attract e.o. 3) if dyons happen to organize into well separated neutral clusters with N dyons in each (= instantons), then det G is factorized 4) identical dyons are symmetric under permutations: they should not know what instanton they belong to! This is an unusual statistical physics based not on the Boltzmann exp(-U/T) but on the measure det G; it can be written as exp(Tr Log G), but then there will be many-body forces!

It turns out that this statistical ensemble is equivalent to the exactly solvable 3d Quantum Field Theory! Two tricks: 1) «fermionization» [Berezin] 2) «bosonization» [Polyakov] Here the «charges» Q are Grassmann fields but they can be easily integrated out [ Diakonov and Petrov (2007) ]

The partition function of the dyon ensemble can be presented identically as a QFT with 2N boson fields v_m, w_m, and 2N anticommuting (ghost) fields: boson and ghost determinants cancel. Classical calculation is exact! periodic N-particleToda potential

1 st result, 1 st criterion of confinement: The minimum of the free energy is at equidistant values of corresponding to the zero average value of the Polyakov line! Indeed, the dyon-induced potential energy as function of, has the minimum at i.e. at equidistant, which implies Tr L = 0 !

The correlation function of two Polyakov lines defines the potential energy between two static quarks: 2nd result, 2 nd criterion of confinement: Тhe potential energy of static quark and antiquark is linearly rising with separation, with a calculable slope, or string tension. The string tension has a finite limit at small T. It is stable in the number of colours Nc, as it should be.

3d result, 3d criterion Along the surface spanning the loop there is a large (dual) field, the string, leading to the area behaviour of the average Wilson loop ! At low T the magnetic string tension coincides with the electric one, as it should be: The Lorentz symmetry is restored, despite the 3d formulation. Moreover, in SU(N) there are N different string tensions, classified by the N-ality of the representation, in which the Wilson loop is considered. We find for the rank-k antisymmetric tensor representation. The string tension in the adjoint representation (k=0) is asymptotically zero. the results for the two string tensions are the same although they are computed in two very different ways

4 th result, thermodynamics of the deconfinement phase transition: In the confinement phase, the free energy is dyon-inducedperturbative energy at maximum Stefan-Boltzmann gluons are cancelled from the free energy, as it should be in the confining phase! The 1 st order confinement-deconfinement phase transition is expected at (At N=2 the free energy depends only on one variable, and the phase transition is explicitly 2 nd order)

Critical temperature T_c in units of the string tension for various numbers N_c : [lattice data: Lucini, Teper and Wenger (2003)] Another important quantity characterizing the non-perturbative vacuum – the topological susceptibility :

Summary 1)The statistical weight of gluon field configurations in the form of N kinds of dyons has been computed exactly to 1-loop 2) Statistical physics of the ensemble of interacting dyons is governed by an exactly solvable 3d QFT 3)The ensemble of dyons self-organizes in such a way that all criteria of confinement are fulfilled Non-trivial holonomy allows the existence of dyons, dyons request the holonomy to be maximally non-trivial ! 4) All quantities computed are in good agreement with lattice data 5) A simple picture of a semi-classical vacuum based on dyons works surprisingly well!