The Complexity of Multivariate Matching Polynomials

We study various versions of the univariate and multivariate matching and rook polynomials. We show that there is most general multivariate matching polynomial, which is, up the some simple substitutions and multiplication with a prefactor, the original multivariate matching polynomial introduced by C. Heilmann and E. Lieb. We follow here a line of investigation which was very successfully pursued over the years by, among others, W. Tutte, B. Bollobas and O. Riordan, and A. Sokal in studying the chromatic and the Tutte polynomial. We show here that evaluating these polynomials over the reals is ]P-hard for all points in R but possibly for an exception set which is semi-algebraic and of dimension strictly less than k. This result is analoguous to the characterization due to F. Jaeger, D. Vertigan and D. Welsh (1990) of the points where the Tutte polynomial is hard to evaluate. Our proof, however, builds mainly on the work by M. Dyer and C. Greenhill (2000).