Yamaguchi et al.(1) examined the embrittlement of nickel (Ni) by progressively adding sulfur (S) atoms to a grain boundary (GB). From first-principles calculations, they concluded that S atoms tend to aggregate at the GB and that the repulsive S-S interactions induce boundary expansion, thus weakening Ni-Ni binding across the boundary. The agreement between their calculated critical S concentration and the measured data (2) suggests that the GB embrittlement of Ni is due to the aggregation of S segregants. Although we believe that the first-order calculations of Yamaguchi et al.(1) are reliable, we question the interpretations of the calculated binding energies and argue that the distribution of S near the GB remains uncertain.According to the Yamaguchi et al. calculations, the average binding energy for a S atom on site GB0/GB2 ( Fig. 1) with a 100% occupation is -4.75/-4.66 eV, and the binding energy drops to -4.23 eV when both GB0 and GB2 sites are fully occupied. With the calculated segregation energy DE seg , which is defined as the energy lowering when an impurity moves from inner bulk (binding energy 0 -2.96 eV/S) to the GB region, Yamaguchi et al. estimated the occupation number using McLean_s equation of equilibrium segregation (3), C GB 0 EC bulk exp(-DE seg /RT )^/ E1 þ C bulk exp(-DE seg /RT )^, with the impurity concentration in the bulk and the temperature as parameters. They noted that a binding energy of -4.23 eV is large enough for S atoms to segregate fully to the GB0 and GB2 sites, and went on to discuss the volume expansion effect of such GB0-GB2 S combinations.This conclusion is only valid if the segregation process starts and terminates instantly, and we know that segregation can take hours or days (2). As shown in (1), when GB0 (GB2) sites are occupied by S, the binding energy of GB2 (GB0) sites reduces greatly as a result of the repelling interaction between S atoms. For instance, if an S monolayer is formed at GB0 sites first, then the binding energy for a 1/4 monolayer of S at GB2 sites decreases from -4.67 to -3.45 eV. The occupation probability is only on the order of 1% under the experimental conditions in (2) (T 0 918 K and C bulk 0 25 atomic parts per million), according to McLean_s equation. This means that a high concentration of GB0-GB2 pairs is unlikely to appear at the Ni GB.To conduct a more comprehensive search for S-S pairs at the NiS5 (012) tilt GB, we calculated the binding energy of S in the form of both GB0-GBn (n 0 1, 3, 4, 5, 6) and GB2-GBm (m 0 1, 3, 4, 5, 6, -2, -3, -4, -5, -6) pairs at one monolayer concentration using the same code EVienna Ab initio Simulation Package (4)^and parameters reported in (1). Our calculations demonstrate that although the GBn (n 0 3, 4, 5, 6) site is still less stable than the GB0 site, the binding energy difference is greatly reduced when GB0 is occupied by S. For example, a GB0-GB3 pair is 0.05 eV less stable than a GB0-GB2 pair; in an isolated 1/4 monolayer, S at the GB3 site is 1.16 eV less stable than at the GB2 site (1). Interesti...