Abstract
Spin crossover materials are bi-stable systems with potential applications as molecular scale electronic switches, actuators, thermometers, barometers and displays. However, calculating the enthalpy difference, DH, between the high spin (HS) and low spin (LS) states has been plagued with difficulties. For example, many common density functional theory (DFT) methods fail to even predict the correct sign of DH, which determines the low temperature state. Here, we study a collection of Fe(II) and Fe(III) materials, where DH has been measured, and which has previously been used to benchmark density functionals. The best performing hybrid functional, TPSSh, achieves a mean absolute error compared to experiment of 11 kJ/mol for this set of materials. However, hybrid functionals scale badly in the solid state; therefore, local functionals are preferable for studying crystalline materials, where the most interesting SCO phenomena occur. We show that both the Liechtenstein and Dudarev DFT+U methods are a little more accurate than TPSSh. The Dudarev method yields a mean absolute error of 8 kJ/mol for Ueff=1.6 eV. However, the MAE for both TPSSh and DFT+U are dominated by a single material - if this is excluded from the set then DFT+U achieves chemical accuracy. Thus, DFT+U is an attractive option for calculating the properties of spin crossover crystals, as its accuracy is comparable to that of meta-hybrid functionals, but at a much lower computational cost.