Multiple Proton Diffusion and Film Densification in Atomic Layer Deposition Modeled by Density Functional Theory

To investigate the atomic layer deposition (ALD) reactions for growth of HfO2 from Hf(NMe2)(4) (TDMAHf) and H2O, a density functional theory (DFT) slab model was employed. We inspected all energy steps, from the early stage of adsorption of each ALD precursor to the densification of multiple atoms into bulk-like HfO2 layers. The activation energy calculations show that repeated proton diffusion from the surface to amide ligands and rotation of the protonated amine is more energetically accessible than the simple elimination of the amine in the initial stage. We therefore propose that multiple protons diffuse to the amide ligands of the Hf precursor before desorption of protonated ligands takes place. Loss of a proton from surface oxygen frees it up for bonding to Hf of the precursor. Protonation of ligands, and especially desorption of ligands, frees up Hf for bonding to surface oxygen. These effects are termed "densification", as they bring Hf-O packing closer to the bulk scenario. Densification is associated with substantial release of energy. During the metal pulse, saturation of the surface by remaining fragments HEX causes adsorption of further metal precursor to stop. The presence of these fragments prevents further chemisorption of HfX4, since this requires the creation of a strong dative bond between Hf and O. Next, during the H2O pulse, Hf exchanges its remaining ligands with OH groups. The exchange occurs due to the decomposition of adsorbed H2O molecules in clusters of HOC. Decomposition of H2O when adsorbed onto a (Hf(NMe2))(x) (x >= 2) cluster (e.g., dimers) also increases the coordination of Hf and O. Simultaneously, low-coordinated oxygen atoms appear at the surface, which are reactive sites for the next metal pulse. With saturation of the surface by OH groups, H2O molecules begin to appear. This detailed description of ALD chemistry allows us to make qualitative predictions about how the process depends on temperature. The data can also be inputted into kinetic simulations for a quantitative view of the complex film growth process.