English

Discriminant and root separation of integral polynomials

Number Theory 2015-01-29 v2 Probability

Abstract

Consider a random polynomial GQ(x)=ξQ,nxn+ξQ,n1xn1+...+ξQ,0 G_Q(x)=\xi_{Q,n}x^n+\xi_{Q,n-1}x^{n-1}+...+\xi_{Q,0} with independent coefficients uniformly distributed on 2Q+12Q+1 integer points {Q,...,Q}\{-Q, ..., Q\}. Denote by D(GQ)D(G_Q) the discriminant of GQG_Q. We show that there exists a constant CnC_n, depending on nn only such that for all Q2Q\ge 2 the distribution of D(GQ)D(G_Q) can be approximated as follows supabP(aD(GQ)Q2n2b)abφn(x)dxCnlogQ, \sup_{-\infty\leq a\leq b\leq\infty}|\mathbb{P}(a\leq \frac{D(G_Q)}{Q^{2n-2}}\leq b)-\int_a^b\varphi_n(x)\, dx|\leq\frac{C_n}{\log Q}, where φn\varphi_n denotes the distribution function of the discriminant of a random polynomial of degree nn with independent coefficients which are uniformly distributed on [1,1][-1,1]. Let Δ(GQ)\Delta(G_Q) denote the minimal distance between the complex roots of GQG_Q. As an application we show that for any ε>0\varepsilon>0 there exists a constant δn>0\delta_n>0 such that Δ(GQ)\Delta(G_Q) is stochastically bounded from below/above for all sufficiently large QQ in the following sense P(δn<Δ(GQ)<1δn)>1ε. \mathbb{P}(\delta_n<\Delta(G_Q)<\frac1{\delta_n})>1-\varepsilon .

Keywords

Cite

@article{arxiv.1407.6388,
  title  = {Discriminant and root separation of integral polynomials},
  author = {Friedrich Götze and Dmitry Zaporozhets},
  journal= {arXiv preprint arXiv:1407.6388},
  year   = {2015}
}
R2 v1 2026-06-22T05:11:36.400Z