M399_27b 342..345

342 NATURE | VOL 399 | 27 MAY 1999 | www.nature.com A micro-Raman spectrometer was used for the further characterization of the deposited diamond. A laser beam of 1-mm spot size was focused onto the crystallites. Figure 3 shows the resulting Raman spectrum: it exhibits a first-order diamond peak centred at 1,332.8 cm. The high intensity of the diamond peak, its small full-width at half-maximum (3.4 cm) and the small contribution of graphite phase confirm the high quality of the diamond crystal film. The Raman analysis was also carried out on deposits surrounding the diamond crystals: these spectra showed graphitic-like bands and no signal from diamond. The diamond growth observed here, that is, in a hydrogen-free and oxidizing environment, differs from the results obtained in the earlier studies, including those using an oxygen-containing molecule in the source gas. The present findings clearly establish that not only hydrogen, but also oxygen is able to abstract hydrogen from the growing diamond surface to form a reactive site. Our findings also establish that it is possible to grow diamond under hydrogen-free conditions, where the etching of carbon (with sp, sp and sp configuration) by oxygen clearly competes with vapour deposition. That is, macroscopic diamond deposition is controlled by the relative rate of these two processes, with growth occurring at temperatures where the oxidation of sp and sp forms of carbon is faster than that of the sp configuration present in diamond. In order to examine the effect of substrate surface roughness on the growth of diamond, films were also deposited on an as-supplied sapphire (0001) substrate. This mirror-polished substrate had an irregular roughness of the order of 1 nm. The SEM image of a film deposited onto the as-received substrate is shown in Fig. 4. It demonstrates that reproducible growth of (111)-oriented diamond crystals with hexagonal shape can be attained on this substrate using the oxygen process but without achieving the in-plane arrangement of the hexagons on the surface obtained with the ultrasmooth substrates (compare Fig. 1). The dependence of crystal orientation on sapphire surface roughness indicates that the in-plane arrangement of the heteroepitaxial diamond crystals is governed by steps and/or terraces present on the surface. In conventional heteroepitaxial nucleation of diamond in which hydrogen is involved, oxygen is extracted from the oxide surface resulting in degradation of crystal structure of the substrate. By contrast, the present oxygen process preserves the original crystal structure of the oxide substrate surface and we therefore expect an atomically flat heteroepitaxial interface between diamond and sapphire. On the other hand, using XRD and high-energy electron diffraction, we have found that a thin layer of (0001)-oriented graphite is epitaxially grown on the ultrasmooth sapphire (0001) surface during the initial growth stage. The (0001) plane of sapphire has a lattice mismatch of 8.4% with the (0001) basal plane of graphite and 6.3% with the (111) diamond plane, implying that the initial formation of the strained graphite layer might be closely associated with the diamond nucleation and growth process. To further investigate this question, it will be necessary to first study the structure of the heteroepitaxial interface to determine whether diamond nucleates directly on sapphire or on the graphite layer formed on sapphire. M