Measurements and calculations of Au-Si surface properties

The Au-Si system has recently attracted much attention because droplets of Au act as catalysts for the growth of Si nanowires. In order to understand the mechanisms of this growth and to model it, the knowledge of the surface properties of Au-Si alloys is crucial. These properties are also useful to study joining processes involving conductive metals on semiconductors. The literature on the Au-Si surface properties is quite poor. The surface energy of Au-Si alloys is difficult to measure because Si has high tendency to oxidize. We have measured it as a function of temperature by large sessile drop for three Au-rich alloys. The samples have been pre-alloyed by melting together Au and Si in the desired proportions. We have paid particular attention to obtain homogeneous samples, which is difficult because of the high melting temperature of Si. In order to avoid oxidation, the surface energy measurements were performed under a H2 rich atmosphere and with a Zr getter positioned very close to the sample. After the sessile drop experiments, the samples were analyzed by scanning electron microscopy. Nanowires curved and several micrometers long were observed in samples processed without the Zr getter. The measurements have been compared with calculations carried out using two different models. One of them is based on Butler's description of a surface as a layer of different composition from the bulk, where the components have the same activity of the bulk. This model is developed in the frame of statistical mechanical theory in conjunction with the Quasi-Lattice Theory (QLT). It allows a quantitative prediction of the transport properties (chemical diffusion, viscosity and electrical resistivity) and of the surface energy. The other model is based on statistical thermodynamics and the interactions in alloy melts are considered in terms of a subregular solution. The model can calculate the surface energy of a selected alloy and its surface composition gradient. The two models give similar results, but, in some cases, the calculations differ from the measured data. On increasing the Si content, the surface energy of the alloys decreases faster in the experiments than in the calculations. Agreements and discrepancies between measurements and computations are discussed in terms of O adsorption, structure and thermodynamics of the liquid. We conclude suggesting the values of surface energy which should be adopted.