Title: A constructive approach to Schaeffer's conjecture

Abstract: This article is devoted to the question what is the smallest $\mathcal{S}=\mathcal{S}(n)$ such that \[ \left|\det T\right|\left\Vert T^{-1}\right\Vert \leq \mathcal{S}\left\Vert T\right\Vert ^{n-1} \] holds for any induced matrix norm and any invertible complex $n\times n$ matrix. J. J. Schaeffer proved that $\mathcal{S}\leq\sqrt{en}$ and conjectured that $\mathcal{S}$ was bounded for any choice of invertible matrices. Schaeffer's conjecture was rebutted by Gluskin-Meyer-Pajor using a probabilistic approach: they determined an appropriate norm and proved the existence of a sequence of matrices $T=T(n)$ such that $\mathcal{S}$ grows unboundedly. This approach was refined by J. Bourgain boiling down lower estimates on $\mathcal{S}$ to Turan-type power sum inequalities. Combined with a number theoretic analysis of such inequalities H. Queffelec proved the currently strongest estimate $\mathcal{S}\geq\sqrt{n}(1-\mbox{O}(1/n))$. However, finding the optimal $\mathcal{S}$ and the construction of explicit sequences of matrices with growing $\mathcal{S}$ remain open tasks. In this article we provide a constructive approach to Schaeffer's conjecture. We derive new upper estimates on $\mathcal{S}$ and we construct an explicit class of matrices that reaches the asymptotic growth due to Schaeffer's theorem. Our framework naturally extends to provide sharp estimates on the resolvent $\left\Vert (\zeta-T)^{-1}\right\Vert$ when $\left|\zeta\right|\leq\left\Vert T\right\Vert$ and $\zeta$ does not intersect the spectrum of $T$.