Since atomic level processing is driven by the chemistry of precursors at surfaces it is a natural candidate for atomic level modelling [link to atomistic modelling section of the website], using first principles density functional theory (DFT). DFT is able to model surfaces and surface chemistry to a reasonable level of accuracy, treating hundreds of atoms at a time. One can determine the thermodynamics of potential reactions and compute activation barriers for particular processes. These allow one to assess if the chemistry is feasible and what reaction by-products are possible. Furthermore, one can use activation barriers from DFT as input to kinetic Monte Carlo simulations to explore film growth on much larger and complex models, obtaining results that can be directly compared to experiment.
Our focus in modelling atomic level processing includes (1) plasma ALD of metals (2) ALD of nitrides as barrier and liner layers in transistors, (3) atomic layer etching of high-k materials and metals and (4) molecular layer deposition of hybrid organic-inorganic materials for low-k materials or encapsulation.
A recent example is our work in our SFI-NSF China partnership project on the modelling of the metal pulse in nitrogen plasma ALD of cobalt, which is important in nanodevices, e.g. as interconnect metal showing lower resistivity than copper at small dimensions. We first determined structures cobalt surfaces terminated in NH and NH2 groups which are proposed to be present after the nitrogen/hydrogen plasma pulse (similar to production of surface –OH species in oxygen plasma ALD) and predicted the coverage of NH/NH2 under typical ALD operating conditions. Figure 1(a) shows the atomic structure of bare Co (001) and (100) surfaces, while Figure 1(b) shows how the stability of different coverages of NH and NH2 changes with temperature (the grey boxes show the typical experimental temperature ranges for Co ALD) and Figure 1(c) shows the atomic structures of Co (001) and (100) with NH and NH2 terminations. https://pubs.acs.org/doi/abs/10.1021/acs.jpcc.9b06287
With these surface models we have investigated how the well-known Co(Cp)2 precursor interacts with the surface and the energy pathway to eliminate Cp ligands via hydrogen transfer from the surface to a Cp ring, releasing CpH. Figure 2(a) shows the Co(Cp)2 precursor interacting with the NHx-terminated Co surfaces to allow us predict the preferred orientation of the precursor; the preference depends on the surface structure and is driven by the geometry of the surface layer. Figure 2(b) shows the computed energy pathway for the elimination of Cp from the precursor, including the activation barrier for the first H transfer step. This barrier is rather moderate at 0.56 eV but the preference is for one Cp ligand to be eliminated, leaving CoCp on the surface. This analysis is extended to Ru ALD, following a similar approach. https://pubs.acs.org/doi/abs/10.1021/acs.jpcc.0c02976