Further evidence that the role of hydrogen in stage 1 KHg-GIC's is to accept electrons comes from an examination of the effect of stoichiometry on the superconductivity of KHg alloys. In these compounds, Tc decreases monotonically with increasing potassium content, from 4.2 K in pure mercury to 0.9 K in KHg itself.[17,18] These data provide strong support for the idea that adding electrons to Hg decreases rather than increases the transition temperature. One can then think of the carbon layers in the potassium-mercury graphite intercalation compounds as a sink for some of the intercalant electrons, and use this idea to explain why Tc is higher in stage 1 KHg-GIC's than it is in the KHg alloy. The hypothesis that hydrogen increases the transition temperature by removing electrons from the intercalate bands is then fully consistent.
The density of states model described above could also be extended to treat stage 2 KHg-GIC's. Several different experiments show that the Fermi level in stage 2 is about 0.5 eV lower than in stage 1.[15,16,19] The finding that Tc = 1.9 K is higher in stage 2 than in stage 1 may indicate that the same type of density of states shift that occurs upon hydrogenation also occurs from stage 1 to stage 2. This conclusion is supported by EELS data which indicates that the density of intercalate states is higher in stage 2.[16] The EELS result is apparently in conflict with the finding from specific heat measurements that the total N(Ef) decreases substantially from stage 1 to stage 2.[9] The conflict stems from the fact that for the two results to be consistent, one must assign a contribution to N(Ef) from the graphite pi bands which is an order of magnitude larger than that expected for the graphite pi bands at this Fermi level.[19] The seeming discrepancy between the density of states model and the specific heat measurements can be resolved by noting that the stage 1 GIC used for the specific heat measurement was found to be not superconducting down to 0.8 K.[9] Therefore the upper limit on the transition temperature of the GIC used in the specific heat experiment is below the lowest Tc measured for any of the samples described in the present work, which implies that the electronic structure of that compound may well have been different from that of the GIC's described here. If this assertion be correct, then other experiments present a unified picture of the behavior of the density of states in these compounds.