Solid state interaction between thin metal films and SiO2 substrate.

Abstract

Thin films of Cu, Ir, Mo, Pt, Re, Y, Yb and Zr were vacuum deposited onto Si02 substrates of thermally oxidized Si< 100 > wafers and were annealed in a vacuum tube furnace. Samples were then analysed using Rutherford Backseattering Spectrometry (RBS) and X-Ray Diffraction (XRD) to determine whether solid state reactions took place and to identify the compound phases formed. It was found that Y, Yb and Zr reacted with the S1O2, while none of the other metals reacted even after heating for hours at high temperatures of up to 900° C. These results as well as all the other metal-S1O2 interaction studies that could be found in the literature were summarized and extensively analysed. In all the cases where the metal reacted with the S1O2 (Hf, Nb, Ta, Ti, V and Zr) with the exception of the rare-earth metals Y, Tb and Yb, it was found that a thin layer of metal silicide formed sandwiched between the S1O2 substrate and a top layer of metal oxide. From this configuration the following model for the diffusion process could be formulated. Metal atoms diffuse to the SiC^/silicide interface where interaction with and dissociation of SiC>2 takes place. The released oxygen atoms then diffuse through the silicide layer to form a metal oxide on top of the silicide. It could therefore be concluded that oxygen diffuses more readily through the silicide than silicon through the metal oxide. If the opposite was true the sample configuration after reaction would have been SiC^/MO^/MSiy. In the case of the rare-earth metals, solid-state reaction always took place, but with all the elements M, Si and O being present throughout the reaction region. In the case of Yttrium ternary phases were identified, whereas XRD showed that a mixture of silicides and oxides formed in the case of Ytterbium.

In this study an extensive table of heats of formation for the silicides and oxides was compiled, from which heats of reaction could be calculated. The theoretical thermodynamic predictions were found to be consistent with experimental observations. This study also showed that the results could be correlated with the mean electronegativity of the metal which offers a convenient empirical method of predicting whether a metal will react with Si02 or not. It is found that metals with an average electronegativity (average of Allred-Rochow, relative compactness and Pauling electronegativities) of less than 1.45 on the Pauling scale react with Si02- An even better correlation with electronegativity could be obtained when using the Miedema electronegativity parameter (<?*). Metals with c»* values of less than 4.45 V were found to react with Si02 and those with values equal to or greater than 4.45V were found not to react.

It has also been shown how ternary phase diagrams can be used to predict solid-state interaction between metal films and SiC^.