Time-resolved CVD of Group 13-Nitrides

Polla Rouf
Abstract & Cover

Group 13 nitrides (AlN, GaN and InN) and their alloys are semiconductor materials with a wide bandgap span covering from UV down to IR range. Their excellent electronic properties make them extremely attractive materials for light emitting diodes (LEDs) and different kind of transistor structures, especially high electron mobility transistors (HEMTs). These materials are routinely deposited by chemical vapor deposition (CVD) at high temperatures. The most sought-after material among the group 13 nitrides is InN due to its high electron mobility making it extremely useful in transistor structures. InN needs to be deposited at low temperatures as it decomposes at high temperatures. This does not only limit the deposition temperature for InN growth but also for all the other materials that will be deposited on top of InN. In this thesis the deposition of group 13 nitrides is investigated by low temperature atomic layer deposition (ALD) via both a thermal and plasma route. This was conducted by both process development and by improving the deposition chemistry by developing new precursors.  Carbon impurities is one of the greater challenges when using the standard aluminum precursor trimethylaluminum (TMA) in ALD due to the strong Al–C bonds in the molecule. An in-situ removal of carbon impurities was investigated by introducing a cleaning pulse, after the TMA pulse. The cleaning pulse consisted of an H2, N2 or Ar gas pulse perpendicular to the surface. The introduction of the cleaning pulse reduced the carbon impurity in the AlN film from 3 at% down to under 1 at%. This made it possible to deposit AlN at higher temperature to obtain better crystalline quality and on the same time reduce the impurity levels. Kinetic simulations showed that the cleaning pulse cleans the surface from desorbed methyl groups resulting in a suppressed reabsorption pathway.  To further reduce carbon impurities, the strong M–C bonded precursors was replaced with a M–N bonded one. The precursor used were tris(dimethylamido)gallium together with ammonia (NH3) plasma to deposit GaN. The precursor showed ALD behavior and the resulting GaN film possessed significantly lower carbon impurities compared to M-C bonded precursor at low deposition temperatures. This precursor could also produce epitaxial GaN directly on 4H-SiC without a need of a seed layer. To further investigate the precursor impact on deposition chemistry and ultimately the film quality, three indium precursors were evaluated, indium(III)guanidinate, indium(III)amidinate and indium(III)formamidinate. All three precursors have more or less the same structure, only difference being the size of the substituent on the endocyclic carbon position (-NMe2, -Me and -H respectively). Experimental results showed that smaller groups on the endocyclic carbon position improved the InN film quality in terms of crystallinity, morphology, stoichiometry and optical properties. Density functional theory (DFT) calculations showed that smaller moieties on the endocyclic position will lead to less surface and steric repulsion with the exocyclic position. As the size is decreased the exocyclic groups can fold up closer towards the endocyclic position leading to elongated metal-ligand bonds which will result in easier removal of the ligand for the upcoming NH3 plasma pulse.  From these results a new ligand was developed to further improve the deposition chemistry where the endocyclic carbon atom in the ligand backbone of the foramidinate ligand was replaced by a N atom to form a triazenide ligand (iPr–N–N=N–iPr). The triazenide ligand possess no moiety on the endocyclic position compared to the ligands used previously and hence should result in improved material quality if extrapolated from our previous study. The ligand was placed on indium and gallium forming In(III)triazenide and Ga(III)triazenide respectively. Both precursors showed excellent thermal properties making them good ALD precursors. Their use for depositing InN and GaN was investigated with NH3plasma. The resulting films showed excellent quality where no carbon could be detected for either InN nor GaN using XPS and ERDA. Both InN and GaN showed epitaxial growth behavior on 4H-SiC at deposition temperature of 350 °C, a factor of three lower deposition temperature compared to CVD. Interestingly, several linear growth regimes (ALD windows) upon changing the temperature were observed, two and three for InN and GaN respectively. This indicated that the precursors decomposed upon increasing the temperature to form smaller fragments which increased the growth rate but on the same time the smaller precursor fragments saturated the surface. This was further confirmed by DFT calculations.    The In(III)triazenide and Ga(III)triazenide was further used to deposit the ternary InGaN phase. A new method was developed where both precursors were mixed in the bubbler and co-sublimed into the reactor via a single pulse. The composition of the films could be tuned via bubbler temperature, deposition temperature and premixed ratio of the precursors in the bubbler. Near In0.5Ga0.5N could be obtained at low deposition temperatures confirmed by both XPS, ERDA and bandgap measurement. Deposition at 350 °C on 4H-SiC resulted in epitaxial In1-xGaxN without a need of a seed layer. 

Linköping University
(Linköping, Sweden)
External Link
Read Thesis
linkedin invite