The presence of charged carbon clusters of hundreds to thousands of atoms in the gas phase was predicted in the charged cluster model by Hwang et al. and confirmed experimentally by Homann et al. in the CVD diamond process. In the present investigation, we have studied the effect of these charged nanometer-size carbon particles on deposition and etching of CVD diamond. In the first part of experiment, a graphite sheet and a Mo sheet were placed side by side in a hot filament CVD chamber in order to experimentally confirm that etching of graphite can take place simultaneously with deposition of diamond. A gas mixture of 1% $CH_4$ - 99% $H_2$ was used. The result of the experiment clearly showed that a diamond film was deposited on the Mo substrate and the graphite sheet was etched simultaneously. A thermodynamic analysis of this experimental observation leads to thermodynamic paradox if we accept the conventional concept of crystal growth by atomic unit. The gas phase of the C-H system has retrograde solubility of carbon; the solubility is minimum at around 1500 K and increases toward the substrate because temperature decrease with approaching to the substrate. Considering this retrograde carbon solubility, the gas phase nucleation of the carbon clusters is thought to provide the driving force for etching at the substrate. Therefore, at the substrate, the driving force is for etching of both diamond and graphite. If the carbon clusters in the gas phase deposit as diamond film, the thermodynamic paradox can be avoided. This explanation appears to be the only possibility to avoid the thermodynamic paradox.
In the second part, diamond powder was placed on a Mo substrate and etched with $H_2$ in a hot filament chamber. This etching process made the surface of some diamond powder very porous. The pore distribution was non-uniform and the pore size is of tens of nanometer. The porous powder was observed only at very high substrate temperature, above around 1100℃. The crystalline phase of this porous part was analyzed by micro Raman spectroscopy. The wave number was $1322 ~ 1329 cm^{-1}$ for repeated measurements of different parts, different from that of cubic diamond, $1332 cm^{-1}$. This peak is rather close to the wave number for hexagonal diamond, which is reported to be $1315 ~ 1326 cm^{-1}$. Therefore, the porous part is thought to be hexagonal diamond. Hexagonal diamond is presumed to be a metastable phase between graphite and the cubic diamond. The appearance of hexagonal diamond seems to be an intermediate step in transition from cubic diamond to graphite. Possible explanation for this experimental observation is as follows. The direct transition from cubic diamond to graphite has higher activation energy than the two step transition from cubic diamond to hexagonal diamond and from hexagonal diamond to graphite. Graphite was not detected on the porous part of the specimen by the Raman spectroscopy. This result might be due to etching of graphite by hydrogen is spite of the formation of graphite from hexagonal diamond.
