In this Letter we report a simulation study in which we compare the solid-liquid interfacial free energy of NaCl at coexistence, with the value that follows from the height of the homogeneous nucleation barrier. We find that the two estimates differ
by more than 100%. Similar, although smaller discrepancies are found for crystals of hard-sphere colloids and of Lennard-Jones (argon) particles. We consider a variety of possible causes for this discrepancy and are forced to conclude that it is due to a finite-size effect that cannot be corrected for by any simple thermodynamic procedure. Importantly, we find that the surface free energies that follow from real nucleation experiments should be subject to large finite size effects. Taking this in to account, we obtain quantitative agreement between the simulation data and the surface free energy of NaCl that follows from nucleation experiments. Our finding suggests that most published solid-liquid surface free energies derived from nucleation experiments will have to be revised.
We report a numerical simulation of the rate of crystal nucleation of sodium chloride from its melt at moderate supercooling. In this regime nucleation is too slow to be studied with brute-force Molecular Dynamics simulations. The melting temperature
of (Tosi-Fumi) NaCl is $sim 1060$K. We studied crystal nucleation at $T$=800K and 825K. We observe that the critical nucleus formed during the nucleation process has the crystal structure of bulk NaCl. Interestingly, the critical nucleus is clearly faceted: the nuclei have a cubical shape. We have computed the crystal-nucleation rate using two completely different approaches, one based on an estimate of the rate of diffusive crossing of the nucleation barrier, the other based on the Forward Flux Sampling and Transition Interface Sampling (FFS-TIS) methods. We find that the two methods yield the same result to within an order of magnitude. However, when we compare the extrapolated simulation data with the only available experimental results for NaCl nucleation, we observe a discrepancy of nearly 5 orders of magnitude. We discuss the possible causes for this discrepancy.
