Breakdown of phonon band theory in MgO

Abstract

We present a series of detailed images of the distribution of kinetic energy among frequencies and wave vectors in the bulk of an MgO crystal as it is heated slowly until it melts. These spectra, which are Fourier transforms of mass-weighted velocity-velocity correlation functions calculated from accurate molecular dynamics (MD) simulations, provide a valuable perspective on the growth of thermal disorder in ionic crystals. We use them to explain why the most striking and rapidly progressing departures from a band structure occur among longitudinal optical (LO) modes, which would be the least active modes at low temperature (T) if phonons did not interact. The degradation of the LO band begins, at low T, as an anomalously large broadening of modes near the center of the Brillouin zone (BZ), which gradually spreads towards the BZ boundary. The LO band all but vanishes before the crystal melts, and transverse optical (TO) modes' spectral peaks become so broad that the TO branches no longer appear band-like. Acoustic bands remain relatively well defined until melting of the crystal manifests in the spectra as their sudden disappearance. We argue that, even at high T, the long wavelength acoustic (LWA) phonons of an ionic crystal can remain partially immune to disorder generated by its LO phonons; whereas, even at low T, its LO phonons can be strongly affected by LWA phonons. This is because LO displacements average out in much less than the period of an LWA phonon; whereas during each period of an LO phonon, an LWA phonon appears as a quasistatic perturbation of the crystal, which warps the LO mode's intrinsic electric field. LO phonons are highly sensitive to acoustic warping of their intrinsic fields because their frequencies depend strongly on them: They cause the large frequency difference between LO and TO bands known as LO-TO splitting. We calculate vibrational spectra from MD trajectories using a method that we show to be classically exact and therefore applicable, with equal validity, to any solid or liquid in any thermal or nonthermal state. By demonstrating its power and generality, we show that it has become possible to go far beyond the reach of perturbation theories and mean-field theories in the study of vibrations in materials.