Molecular-dynamics simulations of direct reactive ion etching of silicon by fluorine and chlorine

Abstract

Listen

Translate article

We report results from molecular-dynamics simulations of F+ and Cl+ impact of silicon surfaces, at normal incidence and over a range of energies (10, 25, and 50 eV). The halogen content of the silicon layer increases with halogen fluence, and the simulations are continued until an apparent, quasisteady state in halogen coverage has occurred. Although in some cases the quantitative results differ, F+ and Cl+ are qualitatively similar in steady-state halogen coverage, depth of penetration, etch mechanisms, and etch yield dependence on ion energy. In both cases, a mixed halogenated silicon layer forms, with a substantial degree of surface roughness (∼1–2 nm for 25 or 50 eV ions). At 10 eV for both F+ and Cl+, the apparent steady-state coverage is about 2 equivalent monolayers and the depth of F+ (Cl+) penetration is about 15 Å. For 25 and 50 eV ions, the corresponding coverage (approximately independent of ion energy and type) is about 3 monolayers. The corresponding depth of penetration is about 35 Å. The silicon etch yield is a function of ion composition and energy: These values ranged from 0.25 at 10 eV to 0.45 at 50 eV for F+ and from 0.06 at 10 eV to 0.14 at 50 eV for Cl+. Simulations revealed that the dominant Si etch mechanism and the silicon etch product stoichiometry changed with ion energy. In addition to physical and chemical sputtering, an additional etch mechanism has been observed in the simulations. We term this mechanism direct abstractive etching (DAE). In DAE, an incoming F+ reacts with a surface SiFx, creating a volatile SiFx+1 that leaves the surface with nonthermal energies, similar to physically sputtered products. At 10 eV F+, DAE accounts for ∼80% of the etched species, but at higher energies this mechanism is less likely. DAE is observed for Cl+ impact as well. Comparison of simulation predictions to available experimental results shows at least qualitative agreement, suggesting the approximations made in the simulations are reasonable.