Can Dehn surgery yield three connected summands? (Q531914)

The Cabling Conjecture of \textit{F. González-Acuña} and \textit{H. Short} [Math. Proc. Camb. Philos. Soc. 99, 89--102 (1986; Zbl 0591.57002)] states that Dehn surgery on a knot in \(\mathbb S^3\) can produce a reducible 3-manifold only if the knot is a cable knot and the surgery slope is that of a cabling annulus. Since the case of a cable knot yields exactly two connected summands [see \textit{C. McA. Gordon} and \textit{J. Luecke}, Math. Proc. Camb. Philos. Soc. 102, No. 1, 97--101 (1987; Zbl 0655.57500)], a weaker conjecture may be stated, asserting that a manifold obtained by Dehn surgery (i.e. \(M^3=M(K, r)\), where \(K\) is a knot in \(\mathbb S^3\) and \(r \in \mathbb Q \cup \{\infty\}\)) cannot by expressed as a connected sum of three non-trivial manifolds. Note that results by \textit{N. Sayari} [Topology Appl. 87, No.~1, 73--78 (1998; Zbl 0926.57020)], \textit{L. G. Valdez Sánchez} [Topology Appl. 98, No.~1--3, 355--370 (1999; Zbl 0935.57024)] and \textit{J. Howie} [J. Pure Appl. Algebra 173, No.~2, 167--176 (2002; Zbl 1026.20019)] ensure that if \(M(K, r)\) has three connected summands, then two of these must be lens spaces and the third must be a \(\mathbb Z\)-homology sphere. Moreover, in the same hypothesis, \(r\) must be an integer: see [Gordon and Luecke, loc. cit.]. The present paper faces the problem of three connected summands by making use of standard techniques of intersection graphs by Scharlemann and Gordon-Luecke. The main result proves that, if \(K\) has bridge-number \(b\) and \(M(K, r)= M_1 \#M_2\#M_3,\) \(M_1\) and \(M_2\) being lens spaces and \(M_3\) being a homology sphere but not a homotopy sphere, then \[ |\pi_1(M_1)| + |\pi_1(M_2)| \leq b+1. \] As a consequence, the inequality \[ |r| = |\pi_1(M_1)| \cdot |\pi_1(M_2)| \leq \frac{b(b+2)}{4} \] is obtained, which is a sharpening of a similar inequality due to \textit{N. Sayari} [see J. Knot Theory Ramifications 18, No.~4, 493--504 (2009; Zbl 1188.57004)].