The melting curve between &egr; and liquid iron (100GPa<P<300GPa) has been derived by computing Gibbs free energies at high pressures and high temperatures from equations of state of the &agr;, &egr; and liquid phases. The most uncertainty lies in the equation of state (EOS) of the &egr; phase. By comparing the calculations to experimental data, the internal thermodynamic consistency of the three phases are examined. The best fits to the melting curves of Boehler [1993> and Williams et al. [1987> can be obtained with lower bulk moduli than determined by static compression. Using available equations of state of the iron phases, our calculations indicate that if sub-solidus iron is of the &egr;, phase, Boehler's melting curve is thermodynamically more consistent than Williams et al.'s. The problem is complicated by the possible existence of a new phase between the &egr; and the liquid fields. ¿ American Geophysical Union 1995