Predictive, High-throughput Computation of Spin-Crossover*

Spin-crossover (SCO) is a change in the total spin state of a magnetic ion (typically 3d) in a molecular complex driven by perturbation by another degree of freedom, e.g. temperature T.  A common experimental manifestation is rather abrupt change in bulk system magnetic susceptibility as a function of T.  At T_1/2 half the molecules are in the high-spin (HS) state, the other half in low-spin (LS).   Another important parameter is Γ, the width (in T) of the SCO hysteresis loop. 

    SCO is expected to be a key ingredient for quantum information systems based upon molecular magnetic materials. That requires predictive ways to screen myriad candidate molecules for SCO.  But they are big molecules in condensed phase at finite T.  Candidate systems may number in the thousands.  Dauntingly for theorists, the molecular electronic adiabatic energy difference ΔE_HL= E_HS-E_LS, that drives SCO is inaccessible spectroscopically.  It is inferred from T_1/2 and the spin and vibrational entropy differences. Thus, ΔE_HL is a molecular property found from bulk measurements. 

   High-cost, explicit wave-function quantum chemistry methods clearly are disqualified.  Tuning a density functional approximation (DFA) to parametrize experiment gives no predictive capacity beyond a small set of chemically similar complexes.  I will discuss our recent progress on this challenge.  First is a method for calculating Hubbard-U corrections to the simple PBE DFA with a half and half LS-HS ensemble self-adapted to each complex.  With it and careful calculation of entropic and enthalpic contributions, we predict T_1/2 values that are as good as those from much more expensive methods. Briefly I will discuss progress on calculating Γ parameters, though the method still is too costly for high throughput.

Thanks to A. Albavera-Mata and R. G. Hennig for a very fruitful collaboration.