Analysis of hydroxyl group controlled atomic layer deposition of hafnium dioxide from hafnium tetrachloride and water

Abstract

Atomic layer deposition (ALD) has recently gained interest because of its suitability for the fabrication of conformal material layers with thicknesses in the nanometer range. Although the principles of ALD were realized 30 to 40 years ago, the description of many physicochemical processes that occur during ALD growth is still under development. “Substrate-inhibited (SI)” ALD growth is one phenomenon not yet well understood. In SI-ALD, the growth-per-cycle (GPC) increases in the beginning of the growth, goes through a maximum, and levels off to a constant value. The origin of SI growth is investigated in this work with two recent models of ALD: Model A of Puurunen [Chem. Vap. Deposition 9, 249 (2003)] and Model B of Alam and Green [J. Appl. Phys. 94, 3403 (2003)]. The hafnium tetrachloride/water ALD process, of interest for gate dielectric applications, is taken to represent typical SI growth. The possible reaction chemistry is evaluated with two models: Model C of Ylilammi [Thin Solid Films, 279, 124 (1996)] and Model A. Model A seemingly allows higher amounts of species adsorbed at saturation than Model C. The ligand exchange reaction of hafnium tetrachloride with one surface hydroxyl (OH) group is chosen as the chemical basis of the modeling. Models A and B are, despite their apparent dissimilarity, found to treat the GPC identically when the same chemical reactions are assumed. According to Model A, the maximum in the GPC of hafnium dioxide ALD originates from a maximum in the surface concentration of OH groups, whereas according to Model B, the maximum is caused by a sudden decrease in the fraction of OH groups reacted with hafnium tetrachloride. The physical picture obtained with Model A is in better agreement with other investigations. Analysis of Model B reveals that OH surface concentrations produced by the model are systematically too high and that the numerical solution of Model B is based on an assumption not valid for the hafnium oxide ALD process. In addition, Model B is constructed assuming that ALD is a continuous process. A theoretical example of random deposition as a growth mode in ALD compared to continuous deposition illustrates that the noncontinuous, discrete nature of ALD affects the resulting mathematic equations.