I'm trying to calculate coexistence curves and get a rough estimate for the critical point of simple fluids using the well-known Gibbs Ensemble Monte Carlo (GEMC) (c.f. Panagiotopoulos, Molecular Physics 61 1987). In doing so, my test has been the square-well potential (SW) since there exist numerous literature data already. It seems however, that GEMC is highly dependent on the total initial concentration. If the initial concentration is set to a known critical point, everything works fine. However, if it is set to low, the fluid phase separates but the coexistence densities for the liquid are underestimated. On the flip side, if it is initially set to high, the fluid fails to decompose due to a low probability of acceptance for particle transfer moves.
My question: How do you determine the initial concentrations for a GEMC simulation when the critical point is not known?