Normal two-dimensional hypersurface triple points and the Horikawa type resolution (Q1312821)

The aim of the first part of the article is to construct an analog for triple coverings of Horikawa's resolution for double coverings. Let \(W\) be a nonsingular complex analytic surface and \(\pi : \overline L \to W\) the \(\mathbb{P}^ 1\)-bundle associated to a line bundle \(L\) on \(W\). Let \(S\) be an irreducible reduced divisor on \(\overline L\) linearly equivalent to \(3T\), where \(T = {\mathcal O}_{\overline L} (1)\). Then we get a triple covering \(\varphi : S \to W\), we call \((S,W,L)\) a ``triple section surface'', and we want to construct a reduction process of singularities of the surface \(S\). There are two classes of singular points \(Q\) on \(S\): the ``inner double points'', for which \(\# \varphi^{ - 1} (\varphi (Q)) = 2\), and the ``target singularities'', for which \(\# \varphi^{ - 1} (\varphi (Q)) = 3\), and in this case we call \(P = \varphi (Q)\) a ``target point'' of \(W\). If \(\tau_ 1 : W_ 1 \to W\) is the blow-up of a target point \(P\) of \(W\), we can construct a new triple section surface \((S_ 1, W_ 1, L_ 1)\), where the line bundle \(L_ 1\) depends on \(L\) and on \(P\), such that we have a birational morphism \(\overline \tau_ 1 : \overline S_ 1 \to S\). We call this a ``triplet blow-up'' at \(P\), and we get the following result: Theorem: Let \((S,W,L)\) be a triple section surface with \(S\) normal and \[ (S,W,L) \gets (S_ 1, W_ 1, L_ 1) \gets \cdots \gets (S_ r, W_ r, L_ r) \gets \cdots \] be the reduction process by successive triplet blow-ups. Then this process terminates in finite steps, namely there exists \(r\) such that \((S_ r, W_ r, L_ r)\) has no target point. \(\tau : (S_ r, W_ r, L_ r) \to (S,W,L)\) is called the ``canonical reduction'' and then the author studies the singularities of the (not necessarily normal) surface \(S_ r\) and the exceptional divisor of the resolution of \(S_ r\). He shows that the singularities are only relative cusps, relative nodes and isolated inner double points, all of multiplicity at most 2. In the second part the author solves Durfee's conjecture for two- dimensional hypersurface singularities of multiplicity 3. If \((V,p)\) is a such singularity, there exists a triple section surface \((S,W,L)\) and a target singularity \(P\) on \(S\) such that \((V,p)\) and \((S,P)\) are analytically equivalent. We can use the results of the first part of the article and we get: Theorem: Let \((V,p)\) be a normal two-dimensional hypersurface singularity of multiplicity 3. Then we have: \(\mu (V,p) \geq 6p_ g (V,p) + 2\), where \(\mu (V,p)\) is the Milnor number and \(p_ g (V,P)\) the geometric genus of the singularity. -- Especially the signature of the Milnor fiber of \((V,p)\) is negative. Moreover, the equality holds if and only if \((V,p)\) is a simple elliptic singularity of type \(\widetilde E_ 6\) in the sense of Saito.