(387c) A Theoretical Model Based on a Group Contribution Method to Describe Atomic Layer Deposition of Zirconia on Silicon

Zirconium is an industrially important ceramic material that has a multitude of applications in biomedical sciences and electronics. Thin films of zirconia, bearing a high dielectric constant, chemical stability and high wear resistance, can be used as insulators in the fabrication industries. Zirconia has applications in building novel photonic materials and they are also used as thermal barrier coatings. Atomic layer deposition (ALD) technique could be used to deposit an ultra-thin and highly conformal layer of ZrO₂ using metal precursors like Tris(dimethylamino)cyclopentadienyl zirconium (ZyALD) or Tetrakis(dimethylamido)zirconium(IV) (TDMAZ) on metal substrates. Metal precursors are commercially developed chemically complex compounds that contain many chemical species. During ALD, the metal precursor is introduced on the surface of the metal substrate and one or a few of the chemical components are adsorbed on the surface. The rest is purged out by passing an inert gas and the reaction is completed by a subsequent pulse of the oxidizer that serves as the second reactant. However, developing a successful ALD recipe requires time and effort as the reaction occurs at a specific window of temperature and pressure. Understanding how the different components of the metal precursor interact among themselves and with the substrate will help explain the surface chemistry of the materials and facilitate development of the ALD recipe.

A model characterizing the growth of the deposited metal oxide on the substrate with increasing number of ALD cycles was proposed in the study using adsorbate solid solution theory (ASST). Experimental data were obtained from atomic layer deposition of zirconia on silicon (100) substrates using a custom-made ALD system (Patent #10214817) and a commercial ALD system (Model ALD150-LE, Kurt J. Lesker Co.). Multiple ALD depositions, involving ZyALD or TDMAZ as the metal precursor and oxygen/ozone or ethanol as the oxidizer/co-reactant were carried out. The interaction of precursor components with the substrate was estimated according to the Universal Quasi-chemical Functional-group Activity Coefficient Group Contribution Method (UNIFAC GCM) applying ASST and a theoretical growth rate was calculated for each experimental precursor pulse using an optimization algorithm. The growth rate of the deposited material over varying precursor pulse time should reach a plateau after a certain number of pulses displaying saturation of the substrate surface with the metal precursor and the plot is a characteristic feature of any ALD reaction. The model predicts the theoretical growth rate based on the calculated interaction parameters and compares it with the experimental growth rate which may help to identify a potential ALD window or to provide suggestions to rectify the recipe. Additionally, with sufficient information, a data bank can be formed that can be used to predict the outcome of an ALD reaction. Also, it can provide insights for formation of a new precursor or modification of the existing ones.