A widely used formalism in quantum statistical physics is the formalism of Green's functions [1,2,3], where the techniques of second quantization and field theory are used from the beginning; the notion of exchange operators of indistinguishable particles is of course contained, but in a completely implicit way. In the formalism of Ursell operators [4,5,6], the starting point is first quantization with numbered particles, so that the role of exchange cycles becomes completely explicit: these cycles appear clearly in all diagrams and, for instance, they are the only source of diagrams for the ideal quantum gas. This reduces the distance between the formalism and, for instance, numerical calculations such as the PIMC method (Path Integral Monte Carlo), where particles are also numbered and the exchange cycles are explicitly sampled by random choices. Moreover, it becomes possible to assume that the particles obey Boltzmann statistics, just by "switching off the cycles", a task that would be difficult in the Green's function formalism. Needless to say, this does not mean that Green's functions are, in general, less powerful than the Ursell formalism! The opposite is actually closer to reality: for instance, Green's functions handle timedependent problems easily, while this is not the case in the Ursell formalism. But it remains true that, if one is interested in a detailed discussion of the effects of quantum statistics, it becomes more straightforward to resort to the Ursell formalism.In this article, we consider dense systems, for which it is not necessarily possible to limit oneself to first order density effects. In contrast to the situation in a dilute gas, a given particle may interact frequently with several others at the same time, and even liquefaction may take place. Therefore, we will no longer ignore all Ursell operators beyond U 2 , as was done in most of previous work in this formalism; operators U 3 , U 4 , etc. now become important. One may actually wonder what the role of these higher order operators is in general, and why exactly it is possible to ignore their role in a dilute system, as was done in [6] for instance. We will see that part of their contribution (what we will call their tree reducible part) groups naturally with U 2 through a chain of binary interactions, and builds an exponential of the mean-field energy. In other words, instead of making the problem more complicated, this contribution of the higher order operators builds exactly the exponential of U 2 that is needed to reconstruct a simple and natural expression of the mean-field. Nevertheless, this mean-field is expressed in terms of the matrix elements of U 2 , instead of the usual matrix elements of the potential itself; in a sense, what we obtain is the exponential of an exponential, since U 2 itself contains exponentials of the Hamiltonian and corresponds physically to the local change of the Boltzmann equilibrium. As a consequence, the logarithms that appeared in [6], and had to be expanded to first order in density, are actu...