The surface chemistry of the atomic layer deposition of ruthenium on aluminum and tantalum oxide surfaces

Abstract

The surface chemistry of Ru atomic layer deposition (ALD) processes based on the use of tris(2,2,6,6-tetramethyl-3,5-heptanedionato)ruthenium(III) (Ru(tmhd)₃) and either molecular oxygen or atomic hydrogen on aluminum oxide films was characterized by a combination of surface-sensitive techniques. The thermal decomposition of the Ru metalorganic precursor was determined, by using a combination of reflection-absorption infrared spectroscopy (RAIRS), temperature programmed desorption (TPD), and X-ray photoelectron spectroscopy (XPS), to start below 400 K and to take place in a stepwise fashion over a wide range of temperatures. Gas-phase products from this chemistry include 2,2,6,6-tetramethyl-3,5-heptanedione (the protonated ligand, Htmhd; in a TPD peak at 520 K), isobutene (540 K; indicating the fragmentation of the organic ligands), and other products from isomerization and/or aldol condensation (650 and 730 K). This chemistry is accompanied by the reduction of the Ru³⁺ ions in two stages, involving the loss of some of their ligands and their direct bonding to the substrate first (between 500 and 600 K) and a full reduction to a metallic state later on (600–700 K). ALD cycles using either molecular oxygen or atomic hydrogen resulted in the slow build-up of Ru on the surface, but the co-deposition of carbon could not be avoided, at least in the initial cycles, while the alumina surface was still exposed. With O₂, the Ru atoms alternate between partially-oxidized (after the O₂ exposures) and zero-valent (after the Ru(tmhd)₃ doses) states, and some Ru loss in the form of the volatile RuO₄ oxide was seen after the second half of the ALD cycles; neither the Ru oxidation state alternation nor the elimination of some Ru from the surface were observed when using H·. The deposited Ru was determined, by combining results from angle-resolved XPS (ARXPS) and low-energy ion scattering (LEIS) experiments, to grow as 3D nanoparticles rather than as a layer-by-layer 2D film, presumably because the Ru precursor preferentially adsorbs (and decomposes more cleanly) on the metal surface. A discussion is provided of the implications of these results for the design of ALD processes.