Assessing the dynamics of CO adsorption on Cu(110) using the vdW-DF2 functional and artificial neural networks

In this work, we revisit the dynamics of carbon monoxide molecular chemisorption on Cu(110) by using quasi-classical trajectory calculations. The molecule-surface interaction is described through an atomistic neural network approach based on Density Functional Theory calculations using a nonlocal exchange-correlation (XC) functional that includes the effect of long-range dispersion forces: vdW-DF2 [Lee et al. Phys. Rev. B, 82, 081101 (2010)]. With this approach, we significantly improve the agreement with experiments with respect to a similar previous study based on a semi-local XC functional. In particular, we obtain excellent agreement with molecular beam experimental data concerning the dependence of the initial sticking probability on surface temperature and impact energy at normal incidence. For off-normal incidence, our results also reproduce two trends observed experimentally: (i) the preferential sticking for molecules impinging parallel to the [1̄10] direction compared to [001] and (ii) the change from positive to negative scaling as the impact energy increases. Nevertheless, understanding the origin of some remaining quantitative discrepancies with experiments requires further investigations.