Triples in matroid circuits (Q1086243)

This paper considers the problem of characterizing those matroids for which some triple of elements is not contained in a circuit. This problem was solved for graphic and cographic matroids by \textit{M. Watkins} and \textit{D. Mesner} [Can. J. Math. 19, 1319-1328 (1967; Zbl 0205.285)] and \textit{Chakravarti} and \textit{Robertson} [Ann. Discrete Math. 8, 247 (1980)], respectively. In this paper, the author generalizes these two results by solving the problem for all binary matroids. For non-binary matroids, the problem is still open. The author's first step is to show that, in the binary case, it suffices to solve the problem when the matroid M is 3-connected and internally 4-connected. The latter condition forbids the existence of a partition of E(M) into subsets X and Y, each having at least four elements, such that \(r(X)+r(Y)-r(M)\leq 2\). The main result of the paper is that if e, f and g are distinct elements in a 3- connected, internally 4-connected binary matroid M, then M has no circuit containing \(\{\) e,f,g\(\}\) if and only if \(\{\) e,f,g\(\}\) is a cocircuit, or M is isomorphic to the cycle matroid of a graph in which e,f and g are edges sharing a common endpoint.