Optimization of Al-CVD process based on elementary reaction simulation and experimental verification: From the growth rate to the surface morphology Masakazu Sugiyama a, * , Tomohisa Iino b , Tohru Nakajima c , Takeshi Tanaka c , Yasuyuki Egashira d , Kohichi Yamashita c , Hiroshi Komiyama c , Yukihiro Shimogaki b a Department of Electronic Engineering, School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan b Department of Materials Engineering, School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan c Department of Chemical System Engineering, School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan d Department of Chemical Science and Engineering, Graduate School of Engineering Science, Osaka University, 1-3 Machikaneyama-cho, Toyonaka, Osaka 560, Japan Available online 10 August 2005 Abstract We propose a method to reduce the surface roughness of Al film in the chemical vapor deposition (CVD) using dimethyl-aluminum- hydride (DMAH) as the precursor. An elementary reaction simulation was executed not only to predict the deposition rate but also to predict the coverage of the film by surface adsorbates. It was assumed that high surface coverage is essential in order to deposit smooth films because the adsorbates protect the surface from oxidation which causes discontinuous growth of crystal grains. According to this principle, the condition, that realizes both high surface coverage and high deposition rate at the same time by using the elementary reaction simulation, was sought. A nozzle inlet was used instead of a conventional showerhead. This drastically improved the surface morphology, showing the effectiveness of this theoretical optimization procedure. D 2005 Elsevier B.V. All rights reserved. Keywords: CVD; Aluminium; Dimethyl-aluminum-hydride; Elementary reaction simulation; Surface morphology; Surface adsorbate; Nozzle reactor 1. Introduction There is an increasing demand for the simulation- assisted development of chemical vapour deposition (CVD) processes. This is due to the need for designing a new process for a novel material with less time and cost. A CVD process has a suitable chemistry for a particular material, that is, a new chemistry has to be designed for depositing a new material. Therefore, the simulation of a CVD process always requires an adequate reaction model. In most cases, the lack of a reaction model obstacles the development of CVD simulations. The reaction models of CVD systems are divided into two categories; the overall model and the elementary reaction model. An overall model is usually obtained based on well- defined experiments [1,2] and is very robust. The range of deposition conditions in which an overall model can be applied is limited to the range of experiments by which the model was obtained. Most of the overall models yield the deposition rate and the composition of deposited films. In contrast, an elementary reaction model includes the chemical reactions occurring in a CVD system ‘‘as is.’’ The number of reactions in an elementary reaction model is very large compared with the overall model because a reaction in an overall model is the combination of some elementary reactions. The merit of an elementary reaction model is that it can be applied to a wide range of deposition conditions. Furthermore, an elementary reaction model can potentially predict additional properties of a film other than the deposition rate and the composition. The crystallographic structure and the surface morphology of a film, for example, 0040-6090/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2005.07.058 * Corresponding author. Tel.: +81 3 5841 2336; fax: +81 3 5841 1183. E-mail address: sugiyama@ee.t.u-tokyo.ac.jp (M. Sugiyama). Thin Solid Films 498 (2006) 30 – 35 www.elsevier.com/locate/tsf