Process optimization of copper MOCVD using modeling experimental design

Abstract An optimisation of copper CVD was carried out through the use of screening and modeling experimental designs. The copper precursor [Cu(hfac)tmvs] was delivered through a bubbler using hydrogen as carrier gas. Water vapour was used as reactant. Films were deposited on sputtered titanium nitride substrate. The influence of substrate temperature, carrier gas flow, water flow, water injection time and bubbler pressure were studied and led to experimental laws, which show the dependence of resistivity and deposition rate with any of these parameters. It was found that the optimum procedure was to inject water during a limited time at the beginning of the growth (typically 2 minutes, for an overall deposition time of 30 minutes). This improves the nucleation but avoids the oxidation of the film. Consequently, the resistivity is very low and the deposition rate is relatively high. For the optimum working point, in terms of resistivity (1.9 μΩ · cm after anneal), X-ray photoelectron spectroscopy (XPS) depth profiling indicates a pure copper phase. The adhesion on the TiN substrate was excellent according to the “Scotch tape” test.