We study SU (2) lattice gauge theory at T > 0 in a finite box with fixed holonomy value at the spatial boundary. We search for (approximate) classical solutions of the lattice field equations and find in particular the dissociated calorons recently discussed by van Baal and collaborators.The quark confinement has not yet found a satisfactory explanation. Several models are under consideration. The dual superconductor scenario views confinement as a dual Meissner effect due to the condensation of Abelian monopoles. An alternative promising approach is based on the center-vortex dominance picture. On the other hand there is the semiclassical approach based on instanton solutions. It provides successful phenomenology for many phenomena in hadron physics. Unfortunately, instanton gas or liquid models fail to explain confinement. The question arises, whether other extended classical objects -e.g. monopoles or dyons -could be suited to describe confinement within a semi-classical approach.We consider SU (2) lattice gauge theory at finite temperature with periodic boundary conditions characterized additionally by a non-trivial holonomy P(x) at the spatial boundary. The Polyakov line at the boundary is then the trace of the holonomy L(x) = caloron solutions appear once a non-trivial holonomy P(x) at |x| → ∞ is admitted. These solutions differ from the 't Hooft periodic instantons employed for the standard semi-classical approach at finite temperatures [3][4][5]. The latter solutions have trivial holonomy i.e. P(x) → 1 for |x| → ∞.The most interesting feature of the new calorons is the fact that monopole constituents of an instanton can become explicit as degrees of freedom [7,8]. They carry magnetic charge (in fact, they are BPS monopoles [6]) and 1/N color units of topological charge. Being part of classical solutions of the Euclidean field equations, one can hope that the instanton constituents can play an independent role in the semiclassical analysis of T = 0 Yang-Mills theory (and of full QCD).Here we present an exploratory study where we have searched for characteristic differences between the two phases as far as semiclassical background fields are concerned. The latter become visible in the result of cooling.We have fixed during the simulation and under cooling the boundary time-like link variables in order to keep a certain value of P(x) = P ∞ everywhere on the spatial surface of the system while conserving periodicity. In this case, the influence of the respective phase, that we want to describe, is twofold : (i) the cooling starts from genuine thermal Monte Carlo gauge field configurations,