An algorithm for the high-energy expansion of multi-loop diagrams to next-to-leading logarithmic accuracy

From MaRDI portal
(Redirected from Publication:875836)



Abstract: We present an algorithm to compute arbitrary multi-loop massive Feynman diagrams in the region where the typical energy scale sqrt{s} is much larger than the typical mass scale M, i.e. s>>M^2, while various different energy and mass parameters may be present. In this region we perform an asymptotic expansion and, using sector decomposition, we extract the leading contributions resulting from ultraviolet and mass singularities, which consist of large logarithms log(s/M^2) and 1/epsilon poles in D=4-2epsilon dimensions. To next-to-leading accuracy, at L loops all terms of the form alpha^L epsilon^{-k} log^j(s/M^2) with j+k=2L and j+k=2L-1 are taken into account. This algorithm permits, in particular, to compute higher-order next-to-leading logarithmic electroweak corrections for processes involving various kinematical invariants of the order of hundreds of GeV and masses M_W sim M_Z sim M_H sim M_t of the order of the electroweak scale, in the approximation where the masses of the light fermions are neglected.












This page was built for publication: An algorithm for the high-energy expansion of multi-loop diagrams to next-to-leading logarithmic accuracy

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q875836)