Precise Control of Nanoscale Cu Etching via Gas-Phase Oxidation and Chemical Complexation

Abstract

We present a cyclic process for selective and anisotropic atomic layer etching of copper: an oxygen plasma modulates the depth and directionality of the oxidized layer, while formic acid vapor selectively removes the copper oxide scale from the metallic copper. Via density functional theory, with finite temperature and pressure free energy corrections, we evaluate the feasibility of formation of gas-phase Cu­(II) and Cu­(I) complexes with formate, water, formic acid, and combinations thereof as ligands. These complexes result from the neutralization reaction between copper oxide (CuO and Cu2O) and formic acid, with and without water. We identified and evaluated the formation free energies of formato, formic acid, aquahydroxo, and aquaformato complexes of Cu­(II) and Cu­(I). Under relevant experimental pressures, we find the water-free dimeric tetra­(μ-formato)­dicopper­(II) “paddlewheel” complex (Cu2(HCOO)4) to be the most favorable etching product, with its formation reaching equilibrium conditions from CuO. The most likely precursor for the dimer is the diformatodi­(formic acid)­copper­(II) monomer, which favorably dimerizes under the same water-lean condition at which the dimer persists. Stabilization of gas-phase Cu (oxide) derivatives thus can be achieved through complexation, enabling gas-phase etching of Cu. This work provides complementary experimental and theoretical studies that illuminate the nature of highly controlled etching with formic acid of nanoscopic CuO(s) layers covering Cu nanoarchitectures, which is relevant for the fabrication of next-generation integrated circuits.

Plasma Oxidation of Copper: Molecular Dynamics Study with Neural Network Potentials

“…In their experiments on Cu (100), only Cu 2 O was observed initially, and only prolonged exposure leads to a “sandwich” structure of CuO, Cu 2 O, and Cu. This is consistent with ex situ XPS results under conditions targeted by our model showing the existence of CuO . The “sandwich” structure is not observed in the current MD simulation, likely due to the much lower thickness that we were able to reach in the MD simulation: while a concentration gradient is observed in our MD structures, a Cu 2 O-like environment is not observed via the BOO parameters.…”

“…This is consistent with ex situ XPS results under conditions targeted by our model showing the existence of CuO. 5 The "sandwich" structure is not observed in the current MD simulation, likely due to the much lower thickness that we were able to reach in the MD simulation: while a concentration gradient is observed in our MD structures, a Cu 2 O-like environment is not observed via the BOO parameters. The molecular dynamics technique also does not provide electronic structure information, preventing a direct comparison with the auger electron spectroscopy (AES) and near-edge X-ray absorption fine structure (NEXAFS) characterizations.…”

“…Comparison with Figure 6 suggests that the resulting oxide after long exposure to oxygen plasma will have the structure and composition of CuO. This result is consistent with an experimental X-ray photoelectron spectroscopy study, 5 although in that case the Cu 2 O is not ruled out. The q ̅ 6 result suggests no Cu 2 O structure is present (Figure S6).…”

“…The three velocities components are initialized similarly to the x , y -velocities of the ions. Such ratios are typical given the experimental conditions . Apart from the velocity, the charge is not explicitly modeled since most of the ions are neutralized by Auger electrons before reaching the surface .…”

“…Such ratios are typical given the experimental conditions. 5 Apart from the velocity, the charge is not explicitly modeled since most of the ions are neutralized by Auger electrons before reaching the surface. 16 All the ions are deposited essentially vertically downward: a collision-free presheath is assumed.…”