Discussion

The most obvious conclusion to be extracted from the numbers in Table is that the bismuth in the CsBi-GIC's is acting as an acceptor of electrons, similar to behavior of the heavy metal in other trilayer ternary GIC's. This conclusion is drawn due to the lower E_F and f_C of the bismuth GIC's compared to the binary Cs-GIC's, and also due to the fairly high E_2g2 Raman frequencies displayed by the CsBi-GIC's.[270] According to Refs. [37] and [194], a higher Raman frequency for donor compounds is indicative of a smaller in-plane carbon-carbon distance and thereby a smaller f_C. Therefore, all evidence points to less cesium charge per carbon atom being resident in the graphite layers than in C₈Cs and possibly even C₂₄Cs.

The picture then is of positively charged cesium layers and negatively charged bismuth and carbon layers. Since it is generally accepted that the superconducting MHg-GIC's and MTl-GIC's also have layers of alternating charge[110], this insight would appear to be of little help in understanding the apparent lack of superconductivity in the CsBi-GIC's. It is hard to see why back transfer of charge from the carbon layers would reduce T_c in the MBi compounds when it seems to raise T_c in the MTl- and MHg-GIC's.

A closer look at Table shows that, roughly speaking, GIC's with | f_C | less than about 0.042 (= 1/24) are not found to be superconducting. An accumulation of experimental evidence[89,62,66,199] supports the conclusion that in the stage 2 alkali metal GIC's that the alkali metal is fully ionized, and in fact it seems reasonable to suppose that in any ternary compound where | f_C | ;SPM_lt; 0.04 the alkali metal atom will be fully ionized. The thinking behind this is that the graphitic pi bands have such a high affinity for the alkali-metal s-electron that the only reason that f_C can fall very low is for the heavy metal acceptor to outcompete the pi-band for the s-charge. Therefore a low f_C in a ternary compound implies that f_M = +1. The data accumulated in Table therefore suggests that superconductivity in ternary GIC's is suppressed by an empty s band, a conclusion in keeping with the Al Jishi model[4] of superconductivity in binary graphite intercalation compounds. This model suggests that s-band occupancy is necessary for superconductivity in GIC's.

The next natural question to arise is that of why physically the s-band occupation should be imperative for superconductivity. More intuition on this question can be developed by examining quantities closely linked to s-band occupancy, namely the c-axis resistivity and the resistivity anisotropy. Table , taken directly from the work of McRae and Marêché[166], shows that the compounds identified above as having an empty s-band also have a high rho_c and a high resistivity anisotropy == rho_c/rho_a. High values for the anisotropy and rho_c indicate an almost two-dimensional band structure, often with hopping conduction along ^c.[227,229]

  
Table: Two tables prepared by McRae and Marêché[166] which list the c-axis resistivity and resistivity anisotropy of many GIC's, both donors and acceptors. A== rho_c/rho_a. Correct sources for these numbers are given in Ref. [166].

The chain of reasoning developed in this section seems to imply that two-dimensionality of band structure tends to suppress superconductivity in GIC's. This is a conclusion which has been previously reached by leading superconductivity experts[163,102] when they turned their attention to the problem of superconductivity in GIC's. The reasons why two-dimensional superconductivity might be expected in the closely related transition metal dichalcogenides intercalation compounds and not in the GIC's is discussed in Section .

The apparent lack of superconductivity in the MBi-GIC's is a great disappointment. Many experimentalists wanted to take advantage of the new materials' increased air stability and reported higher T_c's to extend the scope of experiments that could reasonably be performed on superconducting GIC's. Despite the letdown, attempts to answer the question of why the MBi-GIC's are not superconducting have led to some insight about the properties of the MHg-GIC's, which are the topic of the rest of this thesis. Finally, it should be mentioned that the irreproducibility of transition temperatures in other intercalation compounds leaves one with the hope that superconductivity may yet be confirmed in the MBi-GIC's.