AVS 59th Annual International Symposium and Exhibition
    Thin Film Tuesday Sessions
       Session TF-TuM

Invited Paper TF-TuM3
Growth Simulations for Atomic Layer Deposition: Adsorption, Elimination and Densification Reactions

Tuesday, October 30, 2012, 8:40 am, Room 11

Session: ALD Reactions and Film Properties
Presenter: S. Elliott, Tyndall National Institute, Ireland
Authors: S. Elliott, Tyndall National Institute, Ireland
M. Shirazi, Tyndall National Institute, Ireland
Correspondent: Click to Email
The central idea in chemical vapour deposition (including atomic layer deposition, ALD) is that the thermodynamic tendency of atoms to aggregate and bond into a solid film can be delayed by surrounding the atoms with ligands and transporting the molecular complex as a vapour. Deposition thus involves adsorption of the molecule, removal of ligands and a concomitant increase in atomic coordination number as the solid is formed. The focus is often on adsorption and ligand removal, while the change in coordination during growth, which has been termed ‘densification’ [1], has often been neglected.

We apply first principles density functional theory (DFT) to the Hf(NR2)4+H2O system for the ALD of HfO2, which reveals how important densification can be in explaining the characteristics of oxide ALD. We consider R=Me but expect that similar reactions occur for larger R. Transfer of H from the surface to the ligand is strongly affected by rotation of the ligand around the Hf-N bond, which in turn depends on the crowding associated with proximity of Hf to the surface. Dissociation of HNR2 is facilitated if multiple ligands on an adsorbed Hf centre are protonated, contrary to the usual assumption of dissociation one by one. Once sufficient protonated ligands have desorbed, Hf is freed up to bond to more surface O (densification), with substantial release of energy. Thus this example illustrates the importance of densification reactions in transforming molecular precursors into solid films.

Next, we use the DFT activation energies for this reaction mechanism as inputs to a Kinetic Monte Carlo (KMC) model to explicitly model film growth over multiple ALD cycles. KMC allows a large set of inter-dependent events (such as growth reactions) to be combined into a sequence over a variety of timescales. It is therefore suitable for the coarse-graining in time that is necessary in order to simulate ALD cycles [2]. We have modified the KMC modules of the SPPARKS code [3] for oxide growth by ALD. We include 162 possible reactions (each with DFT-derived activation energies) at 8000 reaction sites under typical values of temperature and pressure. The results show which reactions predominate, how layer-by-layer growth takes place and how roughness evolves in time.

We gratefully acknowledge funding from Science Foundation Ireland under the Strategic Research Cluster ‘FORME’ (www.tyndall.ie/forme) and coding support and computing time at the SFI/HEA funded ‘Irish Centre for High End Computing’.

[1] S. Olivier et al., Chem. Mater., 2008, 20, 1555–1560.

[2] A. Esteve et al., J.Chem. Theory Comput., 2008, 4, 1915-1927.

[3] A. Slepoy et al., J. Chem. Phys., 2008, 128, 205101.