Thermodynamics of alkali feldspar solid solutions with varying Al-Si order: atomistic simulations using a neural network potential

TL;DR

This study uses neural network potentials to simulate the thermodynamics of alkali feldspar solid solutions, revealing how disorder affects mixing properties and observing short-range ordering that could lead to exsolution phenomena.

Contribution

It introduces atomistic simulations with a neural network potential to analyze alkali feldspar thermodynamics across different disorder states.

Findings

Good agreement with literature on Gibbs energy, enthalpy, and entropy of mixing.

Disorder correlates with increased ideality in Na-K mixing.

Observation of short-range Na-K ordering as a precursor to exsolution lamellae.

Abstract

The thermodynamic mixing properties of alkali feldspar solid solutions between the Na and K end members were computed through atomistic simulations using a neural network potential. We performed combined molecular dynamics and Monte Carlo simulations in the semi-grand canonical ensemble at 800 {\deg}C and considered three quenched disorder states in the Al-Si-O framework ranging from fully ordered to fully disordered. The excess Gibbs energy of mixing, excess enthalpy of mixing and excess entropy of mixing are in good agreement with literature data. In particular, the notion that increasing disorder in the Al-Si-O framework correlates with increasing ideality of Na-K mixing is successfully predicted. Finally, a recently proposed short range ordering of Na and K in the alkali sublattice is observed, which may be considered as a precursor to exsolution lamellae, a characteristic phenomenon in alkali feldspar of intermediate composition leading to perthite formation during cooling.