Bond order potential for FeSi

Abstract

A new parameter set has been derived for FeSi using the Albe-Erhart-type bond order potential (BOP) and the PONTIFIX code for fitting the parameters on a large training set of various polymorphs. Ab initio calculations are also carried out to study the relative stability of various polymorphs and to use the obtained data in the training set. The original BOP formalism was unable to account for the correct energetic relationship between the B20 ($\epsilon$-FeSi) and B2 (CsCl) phases and notoriusly slightly favors incorrectly the B2 polymorph. In order to correct this improper behavior the BOP potential has been extended by a Columbic term (BOP+C) in order to account for the partial ionic character of FeSi. Using this potential we are able to account for the correct phase order between the most stable B20 and B2 (CsCl) polymorphs when the net atomic charges are properly set. Although this brings in a new somewhat uncertain parameter (the net charges) one can adjust properly the BOP+C potential for specific problems. To demonstrate this we study under high pressure the B2 phase which becomes more stable vs. B20 as it is found experimentally and expected to be taken place in the Earth mantle. The obtained BOP has also been tested for the metallic and semiconducting disilicides ($\alpha$-FeSi$_2$ and $\beta$-FeSi$_2$) and for the Si/$\beta$-FeSi$_2$ heterostructure. The obtained BOP, as many other BOP, overestimates the melting point ($T_m$) of the B20 phase by $\sim 1000$ K if the parameters in the radial part of the potential were obtained according to the Pauling relation (regular way).