Knowing
the local structure of metal ions in molten salt melts
is essential for understanding the chemistry related to corrosion
and solvation processes for various applications such as molten salt
reactors, solar thermal power systems, and molten salt-enabled materials
processing. However, modeling the dynamic local structure of metals
in salt melts is difficult because classical mechanics does not reproduce
the correct local atomic networking. The computational cost of carrying
out multiple first-principles dynamics calculations to ensure that
the compositional space is well sampled can be prohibitively large.
In order to address this issue, the current study explores the use
of the evolutionary algorithm Universal Structure Predictor: Evolutionary
Xtallography (USPEX) to predict coordination numbers of strontium
and zirconium in binary and ternary chloride and fluoride melts. Temperature-dependent
coordination number distributions for the metal atoms were computed
using a Boltzmann distribution. The calculated average coordination
numbers were found to be consistent with observations from extended
X-ray absorption fine structure (EXAFS) experiments and the expected
temperature trends. Furthermore, the most stable predicted crystal
structures compare well with EXAFS values, validating our approach
for predicting local structures in salt melts.