We concur with both referees that the article as written lacked sufficient clarity. Because we cannot fix this within the length constraints of PRL, we simply resubmit to PRB as a full length paper, which in any case we were preparing. Most of the referees' comments have been incorporated into this resubmission. Of particular importance is the inclusion of the reference to van Setten which we were unaware of and which is clearly highly relevant. We thank both referees for calling our attention to it, and have included some comparisons in the revised text. Below are our responses to those few points that we chose not to address in the revised manuscript. Referee A. The referee asks for absolute instead of relative energies. Since the VASP energies are already defined relative to a fairly arbitrary value (specifically EATOM=71.3643eV for the GGA PAW potentials) we feel there is limited use for the absolute values. Including them would obscure the very fine energy differences which are at the meV scale. Since the issue was raised, we are happy to give some data to the reviewer. Our energy for alpha, calculated with a 3x3x3 grid and Accurate precision (our standard) is E=-6.67720eV/atom. From this all other absolute energies can be calculated if desired. We vigorously disagree with the final paragraph regarding "broken symmetry". Coupling of neighboring cells in the bulk structure breaks the degeneracy that would be needed to achieve disorder at absolute zero. The structure is ordered in some fashion, most likely as a superlattice of the symmetry-broken beta structure, as is required by the third law of thermodynamics. Referee B. *Page 4, paragraph 5: Why is it noteworthy that a particular site reports an unusually large Debye-Waller factor? *We feel that this is explained sufficiently clearly in the sentences that follow. *Page 6: The authors mention variants stacked along the hexagonal axis -- would these be 1x1x3 supercells in their notation? *No it is actually a (1,-1,0)x(0,1,-1)x(1,1,1) supercell. *Page 6: Since the authors tried three unit cells stacked along the hexagonal axis, did they also try other directions, i.e., 3x1x1 supercells to get an idea of whether these lower the energy beyond the 2x1x1 result? *No, the only studies done with that number of atoms were in the hexagonal unit cell. There is certainly the possibility that a larger supercell would further optimize the structure and we mention this in the discussion section. *On page 7 the authors mention that "atomic vibrations contribute strongly to the free energy". This is true. But how does the contribution change between broken- symmetry state and the state with restored symmetry? The difference in free energy determines the transition temperature. *We do not know the answer at this time. Part of the problem is that the symmetry-restored state can only be represented with a very large supercell for which we cannot perform our calculations. We do include a comment in our revised version that the vibrational free energy differences among nearly- optimal symmetry-broken variants should be small compared with the free energy difference between alpha and beta.