A Study of Group 13-Nitride Atomic Layer Deposition : Computational Chemistry Modelling of Atomistic Deposition Processes

Abstract: The crystalline solids aluminium nitride (AlN), gallium nitride (GaN) and indium nitride (InN), together with their alloys, are of huge interest in the semiconductor industry. Their bandgaps span an extensive range from 6.0 eV for AlN to 0.7 eV for InN, with GaN in between at a bandgap of 3.6 eV. Thus, with bandgap tuning from infrared (IR) to ultraviolet (UV) they are well suited for photoelectric applications such as light emitting diodes (LED). The higher bandgaps of AlN and GaN compared to that of silicon (1.1 eV) makes them suitable for high power applications while the high electron mobility of InN makes it attractive for high frequency transistors. Since aluminium, gallium, and indium belong to group 13, their nitrides are termed group 13-nitrides (13Ns).  The deposition techniques chemical vapor deposition (CVD) and atomic layer deposition (ALD) can be used to produce thin films upon a substrate through reactions by suitable precursor molecules in the gas phase or at the surface. These techniques have successfully deposited thin films of 13Ns using commercially available precursors, e.g., trimethyl aluminium (TMA), trimethyl gallium (TMG) and trimethyl indium (TMI) as metal precursor and ammonia (NH3) as nitrogen precursor. However, the chemistry between these precursors is not well developed, as evidenced by the large nonstoichiometric ratio between the metal and nitrogen precursors, in the order of 1:100-1:105. This is not sustainable for mass production of these materials, as significant amounts of precursor gas are wasted and must either be cleaned from the exhaust or be released into the atmosphere. In my thesis, the gas phase decomposition and the surface adsorption of these precursors and alternatives are investigated by computational approaches.  Gas phase decomposition of ammonia is investigated by kinetic modelling at relevant temperature and pressures. At these conditions, a very small fraction of the initial ammonia molecules can decompose within the expected residence time for the gases in the process. The conclusion is that the low reactivity of ammonia is intrinsic and is not due to decomposition into unreactive nitrogen and hydrogen gas. Methylamines as alternative nitrogen precursors are explored for CVD of GaN. Although these are more reactive in the gas phase, their lower surface reactivity compared to ammonia limits their use as a replacement for ammonia in 13N CVD. The origin of the surface reactivity of ammonia in thermal ALD of AlN and GaN, in comparison to the lack of reactivity on InN, is explored. Comparing GaN and InN surface chemistry, the surface adsorption process on InN is less favourable than on GaN as well as being many orders of magnitude slower, indicating that the lack of any reported thermal ALD process on InN arises from the low reactivity of ammonia towards the InN surface.  The resulting surface terminations after ammonia dosing determines how the metal precursors adsorb and react. A series of nitrogen rich surface terminations of the 13Ns is investigated by density functional theory (DFT) modelling and their stability and prevalence at different temperature and pressures are determined from statistical thermodynamics. At low temperatures the surfaces are terminated by hydrogen bonding amino groups while at high temperatures the surface is bare, with the transition temperature between the two structures decreasing from AlN to GaN to InN. TMA can adsorb onto the amino terminated surface and loses ligands by decomposing. Subsequent TMA molecules are found to decompose in two ways depending on how close it adsorbed to an already adsorbed and decomposed molecule.  A suitable alternative class of metal precursor for 13N ALD are molecules with nitrogen to metal bonds, such as formamidinates, amidinates, trisguanidinates, or triazenides. Ammonia will have an easier process to break the weaker metal nitrogen bond compared to a metal carbon bond. The gas phase decomposition of a trisguanidinate precursor is investigated but it is shown to be likely to decompose during volatilization, limiting its use as an InN ALD precursor.  My thesis consists of detailed atomistic simulations of the deposition of AlN, GaN and InN thin films. The simulations in cooperation with experimental work are used to elucidate the detailed atomistic mechanisms occurring during the process. It gives insight into the shortcomings of the current processes and precursors and can be used as a basis for how to improve them, rendering the 13N a suitable material in a sustainable large-scale production for a variety of semiconductor applications.