Dual nanoscale roughness on plasma-etched Si surfaces: Role of etch inhibitors

The nanoroughness formation and evolution during fluorine-based plasma etching of Si surfaces is investigated both experimentally and theoretically. Dual nanoscale morphology, as well as, almost linear increase of both root mean square roughness and correlation length versus etching time is observed in the experiment. The effect of etch inhibitors from the plasma environment is modeled. Two kinds of etch inhibitors can explain and predict the nanoroughness formation and evolution. Key factors in the nanoroughness formation is the angular distribution of etch inhibitors versus that of ions and their sticking probability compared to that of reactive neutral species.

Figure 1

(Color online) AFM images $(5\times 5\mathrm{\mu }{\mathrm{m}}^2)$ of Si surfaces etched by ${\mathrm{SF}}_6$ plasma after etching for (a) $4\min$ ($z$ scale: $45\mathrm{nm}∕\mathrm{division}$, $\mathrm{rms}=9.1\mathrm{nm}$, etched depth $16000\mathrm{nm}$) and (c) $16\min$ ($z$ scale: $100\mathrm{nm}∕\mathrm{division}$, $\mathrm{rms}=51.1$, etched depth $64000\mathrm{nm}$). (b) The height distribution of the surface in (a). (d) The height distribution of the surface in (c). Note that the bumps have a low aspect ratio and they are not “grass.”

Figure 2

(Color online) rms roughness (squares) and correlation length ($\xi$, circles) of Si surfaces etched by ${\mathrm{SF}}_6$ plasma versus etching time. The values at $4\min$ and $16\min$ correspond to the surfaces of Fig. 1. Note that $\xi$ increases due to the increase of the width of the bumps.

Figure 3

Two mechanisms for the roughness increase during plasma etching of Si. (a) Mechanism with “hard” inhibitors: The ratio of ions to inhibitors is higher in the valleys than in the hills, due to more intense shadowing of inhibitors compared to the ions. Inhibitors have isotropic angular distribution while ions reach almost vertically the etched surface. (b) Mechanism with “soft” inhibitors: The ratio of reactive neutral species to inhibitors is higher in the valleys than in the hills, due to the lower sticking probability of reactive neutral species $(S_{\mathrm{RN}}<S_{\mathrm{SINH}})$. Both reactive neutral species and inhibitors have isotropic angular distribution.

Figure 4

(Color online) (a) Two-dimensional cross section of the AFM image of Fig. 1b. (b) Etched profile after simulation with the mechanism of “hard” inhibitors. (c) Another two-dimensional cross section of the AFM image of Fig. 1b. (d) Etched profile after simulation for the mechanism of “soft” inhibitors. (e) The height distribution of the etched profile for the mechanism of “hard” inhibitors. (f) The same as (e) for the mechanism of “soft” inhibitors. (g) The rms roughness and correlation length versus etching time for the mechanism of “hard” inhibitors. (h) The same as (g) for the mechanism of “soft” inhibitors. Please note that the time axes of Figs. 4g, 4h should not be compared since they depend on the type and the concentration of inhibitors.