No label defined (Q4533410)

The class \({\mathcal L}\)-\({\mathcal P}\) consists of those entire functions that are uniform limits, on compact subsets of \(\mathbb C\), of polynomials with only real zeros. The operators \(L_n(\varphi)\), \(n\geq 0\), are defined for any \(\varphi \in {\mathcal L}\)-\({\mathcal P}\) by the equation \(|\varphi (x+iy)|^2=\sum _{n=0}^{\infty} L_n(\varphi(x))y^{2n}\), \((x,y\in \mathbb R)\). In the first part of the paper one finds a simple recursion for \(L_n\) and the closed form \(L_n(\varphi (x))=\sum _{j=0}^{2n} (-1)^{j+n} {2n\choose j}/(2n)! \varphi ^{(j)}(x)\varphi ^{(2n-j)}(x)\). It is also shown that a real entire function \(\varphi\) (with appropriate restrictions on the order and type of the function) is in the class \({\mathcal L}\)-\({\mathcal P}\) if and only if \(L_n(\varphi(x))\geq 0\) for all \(n=0,1,2,\ldots\) and for all \(x\in \mathbb R\). This result extends the well-known Laguerre inequalities \(\varphi^{(k)}(x)^2-\varphi^{(k-1)}(x)\varphi^{(k+1)}(x)\geq 0\), valid for \(\varphi \in {\mathcal L}\)-\({\mathcal P}\), positive integer \(k\) and real \(x\). The authors proceed to consider iterated Laguerre inequalities expressed in terms of operators \({\mathcal T}_k^{(1)}(\varphi (x)) :=\varphi^{(k)}(x)^2-\varphi^{(k-1)}(x)\varphi^{(k+1)}(x)\), \(k\geq 1\), \({\mathcal T}_k^{(n)}(\varphi (x)):= {\mathcal T}_{k-1}^{(n-1)}(\varphi (x)) {\mathcal T}_{k+1}^{(n-1)}(\varphi (x))\) if \(k\geq n\geq 2\). It is known [\textit{Th. Craven} and \textit{G. Csordas}, Pac. J. Math. 136, 241-259 (1989; Zbl 0699.30007)] that \({\mathcal T}_k^{(2)}(\varphi (x))\geq 0\) for all \(x\geq 0\), \(k\geq 2\), and \(\varphi \in {\mathcal L}\)-\({\mathcal P}^+\) (i.e. \(\varphi \in {\mathcal L}\)-\({\mathcal P}\) and all its Taylor coefficients are nonnegative). A short proof invoking extended Laguerre inequalities is presented here. This proof leads to new necesary conditions for entire functions to belong to the class \({\mathcal L}\)-\({\mathcal P}^+\). In the second part of the paper one considers iterated forms of the classical Turán inequalities. Though the computations suggest that all the multiplier sequences satisfy higher iterated Turán inequalities, there is available a proof only for rapidly decreasing sequences. The main result of this part asserts that the third iterated Turán inequalities are valid for \(\varphi \in {\mathcal L}\)-\({\mathcal P}^+\) with a double zero at the origin. Apparently, this last hypothesis is only needed to render the algebraic manipulations feasible on present day computers.