Sharp lower bounds for a generalized Jensen inequality (Q583412)

Let k be a positive integer and \(0<d<1\). A polynomial \(P(z)=a_ 0+a_ 1+...+a_ Nz^ N\) is said to have concentration d at degree k if the following condition (*) holds: \(\sum^{k}_{0}| a_ j| \geq d\sum^{N}_{0}| a_ j|.\) Beauzamy and Enflo, who introduced this notion, demonstrated the existence of a constant \(C_{d,k}\) such that if P(z) has concentration d at degree k then \[ J(P):=\int^{2\pi}_{0}\log | P(e^{i\theta})| d\theta /2\pi - \log (\sum | a_ j|)\geq C_{d,k}. \] The authors first observe that the condition (*) on P(z) makes sense for any function \(f(z)=\sum^{\infty}_{0}a_ jz^ j\), analytic in \(| z| <1\), provided \(\sum^{N}_{0}| a_ j|\) is replaced by \(\sum^{\infty}_{0}| a_ j|\). They redefine \(C_{d,k}\) to be the infimum of J(f) over all functions bounded and analytic in \(| z| <1\) and satisfying (*). They then consider the problem of determining \(C_{d,k}\) and extremal functions satisfying \(J(f)=C(d,k)\). Their first theorem proves that \(C_{d,0}=\log d\), and gives a characterization of the functions f for which equality is achieved in \(J(f)=C_{d,0}\). As they observe, this is equivalent to Jensen's inequality (in its modern version involving outer functions). Their remaining results restrict f to be in the class \({\mathcal H}\) of Hurwitz functions, defined as follows: \(f\in {\mathcal H}\) if and only if it has a product expansion \[ f(z)=\prod_{j}(1-(z/z_ j)),\quad where\quad {\mathfrak R}z_ j<0, \] for all j and the set of \(z_ j\) is closed under complex conjugation. Their remaining results concern \(C^{{\mathcal H}}_{d,k}=\inf J(f)\), where the infimum is over all \(f\in {\mathcal H}\) satisfying (*). They conjecture that \(C_{d,k}=C^{{\mathcal H}}_{d,k}\). In Theorem 2, they determine an explicit value for \(C^{{\mathcal H}}_{d,k}\) and an extremal element of \({\mathcal H}\). Theorem 3 gives the asymptotic results \(\lim_{d\to 0+}(C^{{\mathcal H}}_{d,k}/\log d)=1\) and \(\lim_{k\to \infty}(C^{{\mathcal H}}_{d,k}/k)=-2 \log 2.\)