Unraveling the effects of multi-phonon scattering mechanisms on lattice thermal conductivity of La2Zr2O7, La2Sr2AlO7 and LaPO4 prototype thermal barrier coatings from both perturbative and non-perturbative methods based on machine learning potential

The multi-phonon scattering mechanisms and lattice thermal conductivity of the three prototypes thermal barrier coatings including the La₂Zr₂O₇, La₂SrAl₂O₇ and LaPO₄ are studied using both the perturbative linear Boltzmann transport equation (LBTE) under the three- and four-phonon scattering schemes and the non-perturbative equilibrium molecular dynamics simulations coupled with Kubo-Green formula (EMD-KG) method based on properly trained and validated moment tensor potentials (MTPs). It is revealed that incorporating multi-phonon scatterings in LBTE method under either 3ph or 3ph + 4ph schemes could underestimate the lattice thermal conductivity of the thermal barrier coatings mainly due to the poor description of heat conduction for highly localized phonon modes (locons) in the temperature renormalized phonon spectrum within the perturbative approach. Meanwhile, the non-perturbative EMD-KG method not only could accurately reproduce the values of lattice thermal conductivity that are in good agreement with experiments in the temperature range between 300 K and 1500 K, but also correctly predicts the flat thermal conductivity at elevated temperatures for all three thermal barrier coatings. The phonon quasi-particle spectrum and quasi-particle lifetimes are also obtained, and results elucidate that the perturbative LBTE method predicts the phonon relaxation times significantly smaller than those of phonon quasi-particle lifetimes calculated from non-perturbative EMD method especially for the low-frequency phonon modes. Therefore, the EMD-KG method combined with highly accurate machine learning potential provides a numerically efficient and physically sound methodology to study phonon transport properties of thermal barrier coatings.