*AD- A28O 964 Final Report Surface Processes in CVD Diamond Growth Contract N00014-90-J-1386 Prof. David G. Goodwin r"'J 191-,4 Division of Engineering and Applied Science C .EL California Institute of Technology fl ELFCTE Pasadena, CA 91125 LJ UN 29199411i U Executive Summary The work performed under this contract consisted of basic research with the objective of improving our understanding of the fundamental science of diamond chemical vapor d&eposi-. tion. The major areas of work were 1. numerical modeling of CVD diamond growth 2. theoretical studies of growth mechanisms 3. in-situ diagnostics of gas-phase species during hot-filament diamond growth 4. ab initio calculations of surface migration on diamond 5. theoretical studies of transport issues and scaling law development 6. measurements of the open site fraction on a diamond surface during H exposure. Each of these areas is described briefly below. Complete details are given in the published papers, which are attached. Modeling The numerical modeling work was the first numerical model which incorporated both de- 00 tailed gas-phase chemistry and transport. We first examined hot-filament reactors, and q (cid:127) outlined the basic features of the gas-phase chemistry in this environment [1]. In a second O ~ paper, we showed that catalytic H recombination on the substrate perturbed this environ- ment significantly, and we predicted large gradients in H concentration near the substrate I' (subsequently measured by others) [2]. We next modeled higher-pressure convective dia- I N mond growth methods (arcjets and combustion flames) [3]. By a detailed modeling of the g(cid:127) reacting boundary layers in these systems, we showed for the first time that the same growth mechanism which predicted growth rates well in hot-filament systems (the CH3 mechanism of Dr. Steve Harris at GM) also worked well for higher-pressure, high-growth rate methods. This provided support for CH as the principal growth species, and was the first indication 3 that the chemistry was the same in both high-rate and low-rate reactors. 94 6 29 102 " cc: Clinton Werner/ONR,Pasadena,CA Growth Mechanisms Although our modeling work showed that the Harris mechanism was very successful in pre- dicting growth rates, it was based on a (100) (lxl) diamond surface. Subsequently, STM images showed that the (100) growth surface was (2x1) reconstructed [4]. In collaborative work with Steve Harris, we developed a modified mechanism for growth by CH addition to 3 the (100) (2xl) surface [5]. This mechanism involved C11 insertion into dimer bonds using 3 a previously-lproposed mechanism by Garrison and co-workers [6], and the original Harris mechanism to bridge between dimer rows. We carried out a detailed thermochemical analy- sis, using the molecular mechanics potential MM3. We found that bridging between dimer rows was the rate-limiting portion of the mechanism, which offered some insight into the success of the original Harris mechanism. Diagnostics Our diagnostics work focused on in-situ detection of CH using resonance-enhanced multi- 3 photon ionization (REMPI) during hot-filament growth [7]. The objective of this work was to obtain data on CH to test numerical models of diamond CVD. In these experiments, we 3 also measured stable species concentrations at the substrate using sampling mass spectrom- etry. We studied the temperature dependence of the CH concentration at the substrate, 3 and also measured spatial profiles of CH for several substrate temperatures. We found a 3 previously-unexpected fall-off in CH near the substrate at low substrate temperatures. We 3 have argued that this can be explained as due to the reaction CH + H -- CH occurring 3 4 in the cool layer near the substrate, with perhaps additional contributions due to surface destruction of CH . Modeling work is currently in progress t. simulate these experiments. 3 Surface Migration Ab initio calculations were carried out to investigate the mobility of H, F, and Cl bonded to the diamond surface. Although surface mobility is crucial for the growth of most crystals, diamond growth mechanisms to date have neglected this possibility, due to the large bond strengths of C-H, C-F, or C-Cl. However, even a few diffusive hops of these species on the surface may be important for the growth mechanism, for example by relieving strain. We calculated activation barriers for these species using a small cluster which modeled the diamond (110) surface. We found that H migration to an empty neighboring site has an activation barrier of approximately 2.25 eV, and the value for Cl is only slightly higher. On the other hand, the activation energy for F (3.4 eV) is high enough to rule out F migration. Using transition state theory, we estimated hopping times, which were found to be on the 01 order of 10' s for H and Cl. It is interesting to note that this time is approximately equal to the radical site lifetime at typical diamond growth temperatures. We therefore conclude that a limited amount of diffusive hopping can occur during diamond growth, and may in fact be a crucial component of the growth mechanism. -W-02A_6 Availability Codes Avail andIor Dist Special 2 Scaling Law Development Recognizing that current diamond CVD processes must be scaled to larger areas and higher growth rates, a theoretical study was carried out to investigate the factors governing scaling for several different CVD methods [8, 91. This work consisted both of the formulation of a simplified diamond growth mechamism, and a detailed study of atomic hydrogen transport to the diamond surface. For convective flows (arcjets, RF torches. combustion torches) optimal conditions were identified to maximize the H concentration at the substrate, and therefore the achievable growth rate. Open site fraction measurements One of the critical parameters of tile chemistry of diamond CVD is the fraction of surface sites which are not terminated by hydrogen during H exposure. It may be shown by simple kinetic arguments that this fraction f* is independent of the hydrogen flux, and is fully determined by the ratio of the rate constant for H abstraction to tile rate constant for sticking on open sites (8]. Values assumed in tile literature for this number range from a percent or less, up to 40%. A measurement of this number would give useful information on the rate constant ratio, not obtainable in other ways. We are currently setting up an experiment to measure this quantity, using the system shown in the figure. Elastic recoil spectrometry is being used to measure the H coverage on a single-crystal diamond surface exposed to an atomic hydrogen source created using a microwave discharge. We should have results on the open site fraction within a few months. This work is continuing under other financial support. References [1] D. G. Goodwin and G. G. Gavillet, J. Appi. Phys. 68, 6393 (1990). [2] D. G. Goodwin and G. G. Gavillet, Proc. Ond Intl. Conf. New Diamond Sci. Tech., edited by R. Messier, J. T. Glass, J. E. Butler, and R. Roy (Materials Research Society, Pittsburgh, PA, 1991), pp. 335-340. [31 D. G. Goodwin, Appi. Phys. Lett. 59, 277 (1991). [4] T. Tsuno, T. Ihnai, Y. Nishibayashi, K. Hamada, and N. Fujimori, Jpn. J. Appl. Phys. 30, 1063 (1991). (51 S. J. Harris and D. G. Goodwin, J. Phys. Chem. 97, 23 (1993). (61 B. J. Garrison, E. J. Dawnkaski, D. Srivastava, and D. W. Brenner, Science 255, 835 (1992). [7] E. J. Corat and D. G. Goodwin, J. Appl. Phys. 74, 2021 (1993). [8] D. G. Goodwin, .J. Appl. Phys. 74, 6888 (1993). 3 (jD. C'. Goodwin.I .J rppl. Pliys. 74. 689-5 (1993). SaMPle Manipulator Sample heater ThermacuO 4e current measurement To mass Specfrteron t Gats ToT uor 2um F-i g -c t e e 4N