Origin of the (110) orientation of Y2O3 and CeO2 epitaxial films grown on (100) silicon

A perplexing issue in the growth of epitaxial oxide films on (100) silicon is the observed (110) orientation of yttria (Y2O3) and ceria (CeO2) despite the (100) orientation having a lower lattice mismatch. As expected, yttria-stabilized zirconia (YSZ) grows with the (100) orientation, yet it has a worse lattice match than both Y2O3(100) and CeO2(100) with silicon (100). The orientations observed would be expected if an epitaxial metal silicide layer forms initially during growth, before the oxidizing ambient is introduced. Calculation of the ensuing lattice mismatch between the oxide and silicide layer and multiplicity (sigma(oxide)) of the near coincident-site lattice for the oxide lattice shows that the (110) orientation is better lattice-matched than the (100) orientation for both Y2O3 (+2.3% x -2.4%, sigma(oxide) = 2, vs. -2.4% x +3.5%, sigma(oxide) = 4) and CeO2 (+1.8% x -4.0%, sigma(oxide) = 5, vs. 1.9% x 1.9%, sigma(oxide) = 5) and that for YSZ, (100) is better lattice matched than (110) (-0.8% x -1.6%, sigma(oxide) = 1, vs. -1.1% x +4.8%, sigma(oxide) = 3) In each case, the in-plane orientation yielding the lowest mismatch with the silicide layer is consistent with the in-plane orientation observed between the oxide film and silicon substrate. Furthermore, the commonly observed rotational twinning in the oxide film can be accounted for by the expected orthogonal domain multipositioning in both the silicide and oxide layers. In CeO2, multipositioning I allows two equally matched sets of orthogonal domains. One set consists of the two experimentally observed orientations (related by a 90 degrees rotation). The other set is rotated 37 degrees from the commonly observed orientations. Only the set with orientations aligned to the surface steps of Si(100) is observed, indicating the likely influence of graphoepitaxy in selecting between the two degenerate sets of orientations. When grown on Pt(100), Y2O3 grows with predominantly the (100) orientation and no (110) orientation, suggesting that without silicide formation, Y2O3 will grow with the expected well matched (100) orientation even when polar interfaces are involved.