A characterization of Heegaard diagrams for the 3-sphere (Q5893774)

The excellent survey-research article by \textit{F. Waldhausen} [Algebr. Geom. Topol., Stanford/Calif. 1976, Proc. Symp. Pure Math. 32, Pt. 2, 313--322 (1978; Zbl 0397.57007)] serves as background for this review. \textit{H. Zieschang} [Transl., II. Ser., Am. Math. Soc. 92, 127--137 (1970; Zbl 0206.25605); translated from Mat. Sb., Nov. Ser. 66(108), 230--239 (1965; Zbl 0201.25403)] proved that if the algebraic length of a Heegaard diagram \(H=(F_ g;\partial v,\partial w)\) can be reduced by replacing the system of meridians \(w\) with some \(w'\), then the geometric length can also be reduced by replacing \(w\) with some \(\hat w\) obtained algorithmically. The main result of the paper is Theorem 2: If \(H\) represents \(S^3\), then there exists a system \(w'\) such that the cyclically reduced group presentation arising from \((F_g;\partial v,\partial w')\) has generators \(\{x_1,\ldots,x_g\}\) and relators \(\{x_1,\ldots,x_g\}\). Note that if this is true, Zieschang's theorem algorithmically yields a system \(\hat w\) such that \((F_g;\partial v,\partial \hat w)\) is standard, thus solving the old problem of the algorithmical recognition of \(S^3\). Unfortunately Theorem 2 is false. In fact, according to \textit{J. H. C. Whitehead} [Proc. Lond. Math. Soc., II. Ser. 41, 48--56 (1936; Zbl 0013.24801)] any diagram \((F;\partial v,\partial w)\) of \(S^3\) obtained from the standard one \((F;\partial v,\partial \hat w)\) by replacing \(\hat w\) with some \(w\) must have a cut-point (see also the paper by the reviewer [``Two questions on Heegaard diagrams of \(S^3\)'', Proc. Am. Math. Soc. 102, No. 2, 421--425 (1988; Zbl 0656.57002)]). However, \textit{O. Ya. Viro} and \textit{V. L. Kobel'skiĭ} [Usp. Mat. Nauk 32, No. 5(197), 175--176 (1977; Zbl 0367.57001)] found a Heegaard diagram of \(S^3\), \((F_4;\partial v,\partial w)\), such that neither \((F_4;\partial v,\partial w)\) nor the dual \((F_4;\partial w,\partial v)\) have a cut-point. Both diagrams are counterexamples of Theorem 2 in the paper.