Szczarba's twisting cochain and the Eilenberg-Zilber maps (Q820783)

Suppose we are given a fibre bundle \(F\hookrightarrow E\to B\) over a path connected base, (equally well thought of as a simplicial fibre bundle). \textit{E. H. Brown jun.} [Ann. Math. (2) 69, 223--246 (1959; Zbl 0199.58201)] showed how the homology of \(E\) was related to that of a `twisted tensor product', \(C(F)\otimes_tC(B)\) of the chain complexes of the base and the fibre, and in which \(t\) was a `twisting cochain'. In the same year, Barratt, Gugenheim and Moore [\textit{M. Barratt} et al., Am. J. Math. 81, 639--657 (1959; Zbl 0127.39002)] described the structure of a simplicial \(G\)-fibre bundle as a `twisted cartesian product', \(E\cong F\times_\tau B\), where \(\tau\) is a `twisting function' \(\tau:B_{>0}\to G\). The two results were clearly linked and shortly afterwards \textit{R. H. Szczarba} [Trans. Am. Math. Soc. 100, 197--216 (1961; Zbl 0108.35901)], gave an explicit formula for \(t\) in terms of \(\tau\), whilst \textit{W. Shih} [Publ. Math., Inst. Hautes Étud. Sci. 13, 93--176 (1962; Zbl 0105.16903)], gave an algorithm for computing \(t\) using the maps that are used in the Eilenberg-Zilber theorem, which handles the case in which \(E=F\times B\), and thus gave how in that case, \(C(E)\) and \(C(F)\otimes C(B)\) are related. His method uses what is now known as homological perturbation theory, and thus considers the result as a perturbation or `twisted' form of the Eilenberg-Zilber theorem. His method starts with the usual Eilenberg-Zilber contraction, \((AW, \nabla, H)\), with \(AW\) the Alexander-Whitney map, \(\nabla\) the shuffle map, and \(H\) being the usual contracting homotopy. The twisting cochain derived from Shih's method is not the same as the one by Szczarba's, but there was computational evidence that the two twisting cochains were closely related. The main result of this paper is a detailed statement, and proof, of the exact relationship between them. This is that Szczarba's twisting cochain is the same as the result of applying Shih's perturbation algorithm to \((AW, \nabla, \tilde{H})\), where \(\tilde{H}\) is obtained from \(H\) by reversing simplices and transposing factors. The detailed proof produces, along the way, several new identities and new full proofs of some known ones, and investigates the way of splitting \(C(X\times Y\times Z)\) using the classical Eilenberg-Zilber mechanisms.