Quadratically hyponormal recursively generated weighted shifts need not be positively quadratically hyponormal (Q2385095)

A bounded linear operator \(T\) on a complex Hilbert space is hyponormal if the commutator \([T^*,T]:=T^*T-TT^*\) is positive, and quadratically hyponormal if \(T + s T^2\) is hyponormal for all complex numbers \(s\). For a bounded sequence \(\alpha = (\alpha_n)_{n\geq0}\) of positive numbers, the corresponding unilateral weighted shift \(W_\alpha\) is the operator defined on the usual Hilbert space \(l^2(\mathbb{Z}^+)\) of all square summable sequences by \(W_\alpha e_n:= \alpha_ne_{n+1}\), \(n\geq0\), where \((e_n)_{n\geq0}\) is the canonical orthonormal basis of \(l^2(\mathbb{Z}^+)\). For a complex number \(s\) and a positive integer \(n\), let \[ d_n(s):= \det\left(P_n[(W_\alpha + s W_\alpha^2)^{*}, W_\alpha + s W_\alpha^2]P_n\right), \] where \(P_n\) is the orthogonal projection onto the span of \(\{e_0, e_1, \dots, e_n \}\). Actually, each \(d_n(.)\) is a polynomial in \(t:=| s| ^2\) with Maclaurin expansion \(d_n(t):=\sum_{i=0}^{n+1} c(n,i) t^i.\) The shift \(W_\alpha\) is said to be positively quadratically hyponormal if \(c(n,i)\geq 0\) for all \(n,i \geq 0\) with \(0 \leq i \leq n+1\), and \(c(n,n+1) > 0\) for all \(n \geq 0\). Clearly, if \(W_\alpha\) is positively quadratically hyponormal, then \(W_\alpha\) is quadratically hyponormal. But the converse is not true in general, as was shown in [Proc.\ Am.\ Math.\ Soc.\ 128, No.\,12, 3639--3646 (2000; Zbl 0958.47009)]. The authors consider \(m+2\) positive weights \(0<\alpha_0<\alpha_1<\dots<\alpha_m<\alpha_{m+1}\), \(m\geq2\), and produce a recursively generated weighted sequence, denoted by \[ \alpha:=\alpha_0,\dots,\alpha_{m-2},(\alpha_{m-1},\alpha_m,\alpha_{m+1})^{\wedge}, \] having \(\alpha_0,\dots,\alpha_{m+1}\) as the initial segment of weights. They first establish anew condition of quadratic hyponormality of the corresponding recursively generated weighted shift \(W_\alpha\). Next, they characterize the difference between quadratic hyponormality and positive quadratic hyponormality for such shift \(W_\alpha\), and show that \(W_\alpha\) is positively quadratically hyponormal if and only if it is quadratically hyponormal and satisfies a finite number of conditions depending on \(m\). Furthermore, they show that these conditions are automatically satisfied provided that \(m=2\), and recapture a result due to \textit{I.\,B.\thinspace Jung} and \textit{S.\,S.\thinspace Park} [Integral Equations Oper.\ Theory 36, No.\,4, 480--498 (2000; Zbl 0962.47014)] which states that for \(\alpha:=\alpha_0,(\alpha_1,\alpha_2,\alpha_3)^{\wedge}\) (i.e., \(m=2\)), the shift \(W_\alpha\) is positively quadratically hyponormal if and only if it is quadratically hyponormal. They finally give an example to show that for \(m\geq 3\), a quadratically hyponormal recursively generated weighted shift \(W_\alpha\) need not be positively quadratically hyponormal.