High-throughput computational screening for two-dimensional magnetic materials based on experimental databases of three-dimensional compounds

Abstract

We perform a computational screening for two-dimensional (2D) magnetic materials based on experimental bulk compounds present in the Inorganic Crystal Structure Database and Crystallography Open Database. A recently proposed geometric descriptor is used to extract materials that are exfoliable into 2D derivatives and we find 85 ferromagnetic and 61 antiferromagnetic materials for which we obtain magnetic exchange and anisotropy parameters using density functional theory. For the easy-axis ferromagnetic insulators we calculate the Curie temperature based on a fit to classical Monte Carlo simulations of anisotropic Heisenberg models. We find good agreement with the experimentally reported Curie temperatures of known 2D ferromagnets and identify 10 potentially exfoliable 2D ferromagnets that have not been reported previously. In addition, we find 18 easy-axis antiferromagnetic insulators with several compounds exhibiting very strong exchange coupling and magnetic anisotropy.

Dynamic transverse magnetic susceptibility in the projector augmented-wave method: Application to Fe, Ni, and Co

“…For these calculations, 12 empty shell bands per atom were used in the band summation of Eq. (77) and a 1000 eV plane wave cutoff was used in the plane wave representation of the Dyson equation (51). Furthermore, a constant frequency shift was applied in order to fulfill the Goldstone condition.…”

“…To this end, itinerant magnets seem to be the most challenging, as the Stoner spectrum is gapped for insulators and the magnons less sensitive towards k-point sampling and broadening. In addition, the Heisenberg model often provides a rather accurate description for insulators, with parameters that can be obtained directly from ground state DFT calculations [49][50][51]. Still, it would be of fundamental interest to compare the dispersion relations obtained from a first principles Heisenberg model with a direct computation from TDDFT.…”

“…( 77) is truncated to include a finite number of excited states, and a finite plane wave basis set is used to invert the Dyson equation in Eq. (51). The effect of these parameters is investigated in Secs.…”

“…To study the convergence of the low-frequency Stoner continuum further, it is worthwhile to remark that the macroscopic spectrum of Stoner excitations is much cheaper to compute than the full transverse magnetic excitation spectrum, as no extra plane wave components are needed, when the Dyson equation (51) does not have to be inverted. Thus, it would be of great value if the convergence of the lowfrequency Stoner continuum could be assessed from the Kohn-Sham spectral function itself.…”

“…This is bad news, of course, but there are several practical ways to circumvent these limitations. As an example, one can invert the Dyson equation (51) in another basis set than plane waves, a strategy previously shown to yield smaller gap errors than the ones reported here [24]. Additionally, different strategies have been developed to enforce Goldstone's theorem by introducing information about the exchange-correlation kernel into the Kohn-Sham susceptibility [26] or vice versa [24,25] and in that way achieve the consistency needed to guarantee a Goldstone mode.…”