The entropic cost to tie a knot

We estimate by Monte Carlo simulations the configurational entropy of N-step polygons in the cubic lattice with fixed knot type. By collecting rich statistics of configurations with very large values of N we are able to analyse the asymptotic behaviour of the partition function of the problem for different knot types. Our results confirm that, in the large N limit, each prime knot is localized in a small region of the polygon, regardless of the possible presence of other knots. Each prime knot component may slide along the unknotted region contributing to the overall configurational entropy with a term proportional to ln N. Furthermore, we discover that the mere existence of a knot requires a well defined entropic cost that scales exponentially with its minimal length. In the case of polygons with composite knots it turns out that the partition function can be simply factorized in terms that depend only on prime components, with an additional combinatorial factor that takes into account the statistical property that by interchanging two identical prime knot components in the polygon the corresponding set of overall configurations remains unaltered. Finally, the above results allow one to conjecture a sequence of inequalities for the connective constants of polygons whose topology varies within a given family of composite knot types.