Limit spectral distribution for non-degenerate hypersurface singularities (Q2148431)

Take a (complex) analytic hypersurface germ \(f^{-1}(0)\subset(\mathbb{C}^{n+1},o)\), with an isolated singularity. The spectrum of the singularity is a strong numerical (topological) invariant. It is often encoded via the spectral polynomial, \(\chi_f(t):=\frac{1}{\mu}\sum T^{\alpha_i}\), with \(\alpha_1,\dots,\alpha_\mu\in \mathbb{Q}\cap (0,n+1)\). Thinking of this as a distribution of numbers on the interval \((0,n+1)\), K. Saito studied the asymptotic distribution of the spectrum. He observed that for many germs the asymptotic distribution is determined by the dimension \((n+1)\) only, and does not depend on \(f\). In this case the continuous (asymptotic) distribution is denoted by \(\mathcal{F}(N_{n+1})\). The main result of the article is: \textbf{Theorem 1.} Given a Newton diagram \(\Gamma\) take its scaled version, \(\omega\Gamma\), by the scaling factor \(\omega\). Take any Newton-non-degenerate function germ \(f\), whose Newton diagram is \(\Gamma\). Then \(lim_{\omega\to \infty}\chi_{f_\omega}=\mathcal{F}(N_{n+1})\). Furthermore, K. Saito has defined the function \(\Phi_f:[0,1]\to\mathbb{R}\) as the difference of the continuous and the spectral distributions, \(r\to \int^r_0 N_{n+1}(s)-\frac{1}{\mu}\sum \delta(s-\alpha_i)ds\). A number \(0<r<\frac{n+1}{2}\) is called a dominating value if \(\Phi_f(r)>0\). \textbf{Proposition 5.} Take a plane curve singularity, \(f^{-1}(0)\subset(\mathbb{C}^2,o)\), with one Puiseux pair \((p,q)\). Then: (a) \(\Phi_f(\frac{1}{p}+\frac{1}{q})>0\) unless \(p=2\) and \(q\in \{3,5\}\). And \(lim_{p\to \infty}\Phi_f(\frac{1}{p}+\frac{1}{q})=0\). (b) \(\Phi_f(1-\frac{1}{pq})<0\) with \(lim_{p\to \infty}\Phi_f(1-\frac{1}{pq})=0\). \textbf{Theorem 6.} Take an irreducible plane curve singularity with value semigroup \(\{\beta_0,\beta_1,\dots\}\), such that \(\{\beta_0,\beta_1\}\) is not \(\{2,3\}\) or \(\{2,5\}\). Then \(\Phi_f(\frac{1}{\beta_0}+\frac{1}{\beta_1})>0\). In other words, \((\frac{1}{\beta_0}+\frac{1}{\beta_1})^2>\frac{2}{\mu}\). Moreover, \(lim_{n_g\to \infty}\Phi_f(\frac{1}{\beta_0}+\frac{1}{\beta_1})=0\).