The synergetic effects of process parameters on hydrogenated amorphous carbon (a-C:H) films properties were quantitatively analyzed in plasma-enhanced chemical vapor deposition. Predictive models, created from experimental datasets and machine learning, indicate a synergetic effect between the H2 ratio and ion energy impacting on the substrate. At a H2 ratio of 75 %, etch rates of 200 nm/min were predicted, regardless of radio frequency (RF) bias. These rates decreased to 120–140 nm/min at lower H2 ratios, primarily depending on RF bias. Quadrupole mass spectrometry and Raman scattering were used to investigate the underlying mechanisms. High H2 ratios led to a greater presence of low-mass hydrogen-rich molecules in the plasma. The C2H3 radical intensity at a 75 % H2 ratio was 15 times greater than that at 0 % at an RF bias of 50 W, and similar trends were observed for other low-mass neutrals. The synergetic effects of H2 ratio and RF bias decreased the film's hydrogen content. Machine learning and those diagnostics revealed that ion bombardment induces dehydrogenation of a-C:H, influenced by hydrogen-rich species. This study demonstrates that machine learning can uncover complex plasma processes and optimize material synthesis.
Read full abstract