Iwata and Watanabe's model for the observed low-temperature specific heat of neutron-irradiated graphite ͓T. Iwata and M. Watanabe, Phys. Rev. B 81, 014105 ͑2010͔͒ assumes that self-interstitial atoms exist as clusters of nearly free C 2 molecules. We suggest that their hypothesis is not supported by other experiments and theory, including our own calculations. Not only is it inconsistent with the long-known kinetics of interstitial prismatic dislocation loop formation, density-functional theory shows that the di-interstitial is covalently bonded to the host crystal. In such calculations no prior assumptions are made about the nature of the bonding, covalent or otherwise. DOI: 10.1103/PhysRevB.82.056101 PACS number͑s͒: 65.40.Ba, 61.72.JϪ, 61.80.Hg, 71.20.Ϫb A recent study by Iwata and Watanabe 1 showed that neutron irradiation of graphite produces an enhancement of the material's low-temperature specific heat. They conclude from these measurements that the hindered rotation of interstitial C 2 molecules is responsible for this effect. These are elegant, precise experiments which provide important evidence about the nature of such materials. Nevertheless, we profoundly disagree with the analysis of results that they put forward because it contains unjustified assumptions which can be refuted in a number of different ways, particularly with regard to the C 2 entities invoked.First, recent calculations reported by us, based on densityfunctional theory ͑DFT͒, using large supercells ͑with four graphite layers and up to 290 atoms͒ conclusively demonstrate that the binding energy between pairs of interstitial atoms to yield C 2 is large ͑3 eV͒, and hence their formation must be irreversible at the temperatures of irradiation cited by Iwata and Watanabe ͑333 K͒.2 It is clear from the covalently bonded structures illustrated in Ref. 2 that any migration path for C 2 is unlikely to exist with as low an energy as suggested by Iwata and Watanabe or that any free rotation could occur. Isolated self-interstitial atoms also form covalent bonds with the host crystal, according to our calculations, and in agreement with others. Certainly, none of the structures obtained give rise to c : a aspect ratios or formation volumes comparable with those inferred in their paper.Our calculations were not based on the assumption of a model, either covalent or noncovalent. They are only conjugate-gradient geometry optimizations starting from various initial positions for two C atoms placed between perfect graphitic planes, as in Iwata and Watanabe's C 2 diagram in their Fig. 5. No special distortions or atomic displacements were applied before optimization. There is nothing to prevent the formation of freely floating C 2 units, if this is a valid outcome. The optimization algorithm cannot traverse any energy barrier it experiences; it only proceeds downhill.However, in every case the system spontaneously reorganized into fully covalently bonded structures, integrated with the host lattice, either in the same layer or between layers. Inde...