The inner structure of the dilogarithm in algebraic fields (Q761499)

The author has championed the dilogarithm function \(Li_ 2(z)=\sum z^ n/n^ 2\) over many years. The present paper is concerned with the mysterious relations that hold between the values of the dilogarithm at certain algebraic points. An example is Watson's identity \(Li_ 2(\alpha)-Li_ 2(\alpha^ 2)=\pi^ 2/42+\log^ 2\alpha,\) where \(\alpha =\sec (2\pi /7).\) The author has found some common patterns behind these identities and suggests that identities should appear in the form \(N^{-1} Li_ 2(x^ N)=\sum_{r| N}r^{-1} A_ r Li_ 2(x^ r)+A_ 0 \pi^ 2+A \log^ 2x,\) where the algebraic number x satisfies an equation of the shape \(x^{2A} (1-x^ N)=\prod_{r| N}(1-x^ r)^{A_ r}\) and the A's are rational. A measure of his success is that he is able to generate scores of new identities. On the other hand, this also deepens the mystery because there is no systematic procedure for proving the identities, and no analytic proof whatsoever for many of them. The evidence points to some underlying algebraic structure. For example, the identities might arise by specializing the functional equations for the dilogarithm. It appears that the known functional equations are not enough, though again this has not been proved. The dilogarithm appears tantalisingly in analysis starting from partition identities and hyperbolic geometry and some of the identities can be proved in this way. Results obtained in this way necessarily fit into the pattern described in the paper, but the underlying connections are unclear. The problem seems to require the application of Weil's dictum: ''When in doubt, find the group''.