The theory of phase stability in the Ni-Au alloy system is a popular t
opic due to the large size mismatch between Ni and Au, which makes the
effects of atomic relaxation critical and also to the fact that Ni-Au
exhibits a phase separation tendency at low temperatures, but measure
ments at high-temperature show an ordering-type short-range order. We
have clarified the wide disparity which exists in the previously calcu
lated values of mixing energies and thermodynamic properties by comput
ing 'state-of-the-art' energetics (Cull-potential, fully-relaxed LDA t
otal energies) combined with 'state-of-the-art' statistics (k-space cl
uster expansion with Monte Carlo simulations) for the Ni-Au system. We
find: (i) LDA provides accurate mixing energies of disordered Ni1-x A
u-x alloys (Delta H-mix less than or equal to + 100 meV/atom) provided
that both atomic relaxation (a similar to 100 meV/atom effect) and sh
ort-range order (similar to 25 meV/atom) are taken into account proper
ly. (ii) Previous studies using empirical potentials or approximated L
DA methods often underestimate the formation energy of ordered compoun
ds and hence also underestimate the mixing energy of random alloys. (i
ii) Measured values of the total entropy of mixing combined with calcu
lated values of the configurational entropy demonstrate that the non-c
onfigurational entropy in Ni-Au is large and leads to a significant re
duction in miscibility gap temperature. (iv) The calculated short-rang
e order agrees well with measurements and both predict ordering in the
disordered phase. (v) Consequently, using inverse Monte Carlo to extr
act interaction energies from the measured/calculated short-range orde
r in Ni-Au would result in interactions which would produce ordering-t
ype mixing energies, in contradiction with both experimental measureme
nts and precise LDA calculations.