LAUSR

In the recent paper by Teys [1], an atomic model for the Si(331) reconstructed surface (hereby referred to as T-model) was proposed on the basis of highresolution scanning tunneling microscopy (STM) images. While detailing the virtues against previous and abandoned models, the author avoids any reference to the rather distinct 8P-model advocated few weeks earlier by Zhachuk and Teys [2], casting doubts to his own work. Formulated that way, findings from Ref. [1] leave readers of JETP Letters with a partial and confusing view of the problem, and above all, leaves the observations open to ambiguous interpretation. The 8P-model is also based on STM measurements, and unlike the Tmodel, passed through the scrutiny of first-principles calculations. According to the authors, the 8P reconstruction consistently described the STM imagery and showed a remarkable low surface formation energy. Since the T-model is solely based on STM data, the above ambiguity can only be dissipated if we compare 8P and T structures on an equal foot. This means testing the T-model in terms of surface energy and STM simulations from first-principles. Using the same procedure as in Ref. [2], we found that the T-model is actually unstable. After atomic relaxation, a Si-Si bond in the surface trimer breaks, leading to a strong rearrangement of the surface atoms. Not surprisingly, the resulting simulated constant-current STM image is incompatible with the experimental analogues shown in Figs. 3(a) and 3(b) of Ref. [1]. The surface energy of the structure attained after relaxing the T-model is 8 meV/Å higher than the energy of the 8P structure, much higher than the typical error bar (below 1 meV/Å) which allows us to discriminate surface stability orderings. Combining these figures with the upper limit for the Si(331) surface energy [2], we conclude that according to the T-model, the Si(331) surface should be unstable against decomposition into Si(111) and Si(110) facets, in obvious contradiction with the observations.