|Abstract:||The principal contribution of this Ph.D. research is to explore the chemical interactions of co-reactants with the film growth surface during low temperature chemical vapor deposition. Ammonia emerges as a particularly important additive because of: (i) the ability of catalytically active surfaces to dissociate NH3, which supplies H and N atoms to the growth reaction; (ii) the requirement for NHx species to drive precursor transamination pathways; and (iii) the ability of ammonia to bind with acidic OH groups on oxide surfaces, which completely extinguishes film nucleation and affords a new and effective method for area-selective film growth. In addition, we demonstrate the use of C alloying to improve the tribological properties of HfB2 hard coatings, as well as a new route to achieve atomic layer etching of metals.
The conformal deposition of highly conductive or superconductive transition metal nitride thin films is technologically important for a wide range of applications. For early transition metal nitrides, reaction of dialkylamido precursors and ammonia follow the transamination reaction pathway for the removal of ligands. However, insufficient ammonia decomposition on surface leads to ligand incorporation in film and a high electrical resistivity. In this work, I identify an important mechanism that affords growth of high quality VN films by CVD: the VN surface is catalytic for ammonia decomposition, hence, NHx groups are generated efficiently, which drives transamination to completion at low temperatures. I also extend this idea to low temperature growth of superconductive molybdenum carbonitrides from metal-carbonyl precursor and ammonia.
In the second part, I show how co-reactants can selectively passivate CVD reactions and result in area-selective deposition. Many nanoscale electronic devices are fabricated using a top-down approach involving blanket film deposition, patterning, and etching steps. However, as feature sizes shrink toward 10 nm, pattern registry becomes very difficult. A bottom-up solution is area selective deposition (ASD), in which a thin metallic film grows selectively on metal but not on dielectric surfaces. In CVD or ALD processes, selective growth occurs when film nucleation is inherently difficult on the oxide surface, or when a surface is rendered passive by chemical termination or by the deposition of a dense self-assembled monolayer.
However, the limit of selectivity is determined by the onset of nucleation on the intended non-growth surface, which may be associated with defects in the inherent chemical properties or in the passivation treatments. A robust ASD process must ensure that no nucleation occurs on the oxide over the total time needed to deposit the desired film thickness on the metal.
Here, I show that this can be accomplished by continuously injecting a neutral molecule inhibitor along with the precursor, which reduces the nucleation rate on oxide to values near zero. For the CVD of copper from the precursor Cu(hfac)VTMS, we demonstrate that the magnitude of the rate inhibition by VTMS is much larger on other surfaces and can be used to afford essentially perfect selectivity: there is rapid growth of Cu on RuO2 but essentially no growth on thermal SiO2 or porous, carbon-doped SiO2 (CDO). We explain the mechanism by which selectivity is achieved: on dielectric surfaces, VTMS associatively desorbs Cu(hfac) intermediates, which shuts down nucleation, whereas Cu growth proceeds on metal surfaces because the growth rate from the disproportionation reaction is kinetically faster than the desorption rate.
I also report a second approach to completely suppress film growth on oxide surfaces during CVD of MoCxNy, Fe, and Ru from metal carbonyl precursors of Mo(CO)6, Fe(CO)5, and Ru3(CO)12: we inject NH3, a strong Lewis base, along with the precursor. We show that the use of NH3 completely eliminates the nucleation of metal on oxide surfaces for the 1-2 hour duration of the experiments performed. At the same time, film nucleation and growth occur readily on metallic seed layers of Ti or VN despite the presence of NH3 in the feed gas, i.e., the approach affords perfect selectivity. We interpret our results in the context of catalysis studies, which report the average oxide surface charge (acidity) and the existence of inhibitor adsorption sites with various desorption energies.
I also report that in hard coating and tribology applications, alloying carbon in HfB2 films decreases the coefficient of sliding friction to 0.05, while maintaining high hardness and elastic modulus. Thin films of HfBxCy with carbon contents of 5–35 at.% are deposited at temperatures of 250–600 °C using a halogen free precursor, Hf(BH4)4, in combination with 3,3-Dimethyl-1-butene, (CH3)3CCH=CH2, (DMB), as the carbon source. Increasing carbon content decreases the hardness and elastic modulus of low-temperature deposited films; however, HfBxCy films have
higher H/E and H3/E2 ratios than for HfB2, which is predicted to improve the wear performance in tribological applications. The use of DMB retards the film growth rate and enhances the conformal coating of HfBxCy within deep trenches, including high aspect ratio structures.
In the last section, I show a general approach for thermal atomic layer etching of copper by means of sequential steps of oxidation and oxide removal by an acidic chelating agent. One important requirement is that the oxidation step must create a spatially uniform oxide thickness in the range of Å to nm that does not depend on underlying microstructural features such as crystallographic orientation, grain boundaries or extended defects. It must also convert the copper to an oxidation state suitable for reaction with the chelating agent. At 275 °C, we find that the oxidation of copper by molecular oxygen is very nearly self-limiting, and the resulting Cu2O overlayer reaches a final thickness of 0.3 nm. The Hhfac reacts with the copper oxide but not with the underlying copper to form Cu(hfac)2, and this step is fully self-limiting. This process is selective for the etching of metals, so that nearby SiO2 or SiNx are not removed.