The quaternionic Gauss-Lucas theorem (Q1756460)

This paper gives a quaternionic version of the well-known Gauss-Lucas theorem valid for any quaternionic polynomial. \par The ring $\mathbb{H}[X]$ of quaternionic polynomials is defined by fixing the position of the coefficients w.r.t. the indeterminate $X$ and by imposing commutativity of $X$ with the coefficients when two polynomials are multiplied together. \par The classical theorem states that if $p$ is a complex polynomial, then the zero set of its derivative $p'$ is contained in the convex hull $\mathcal{K}(p)$ of the zero set of $p$. \par Given any quaternionic polynomial $P=\sum_{k=0}^d X^k a_k$, consider the polynomial $P^c=\sum_{k=0}^d X^k \bar{a_k}$ and define the normal polynomial as $N(P)=P\cdot P^c$. Zeros of quaternionic polynomials can be isolated points or isolated 2-spheres centered at real points. If a quaternionic polynomial $P$ has a non-real isolated zero $\alpha+I\beta$, then its normal polynomial $N(P)$ vanishes identically on the 2-sphere $\alpha+\mathbb{S}\beta$, where $\mathbb{S}$ denotes the 2-sphere of imaginary units in $\mathbb{H}$. \par If complex numbers are replaced by quaternions then a number of properties are no longer true due mainly to the lack of commutativity. In particular the classical statement of the Gauss-Lucas theorem fails to be true in the quaternionic setting. An initial version valid only for quaternionic polynomial of degree $\leq 2$ was given in [\textit{F. Vlacci}, in: Hypercomplex analysis and applications. Selected papers presented at the session `Clifford and quaternionic analysis' of the 7th ISAAC conference, London, UK, July 2009. Basel: Birkhäuser. 275--282 (2011; Zbl 1221.30121)]. In more details, the authors define a quaternionic polynomial to be a Gauss-Lucas polynomial if the zero set $V(P')$ of $P'$ is contained in the convex hull of zeros $\mathcal{K}(N(P))$ of $N(P)$. First of all the authors show that any degree 2 quaternionic polynomial is a Gauss-Lucas polynomial (as stated in the paper by Vlacci [loc. cit.]). Afterwards, they show a large class of quaternionic polynomials that are not Gauss-Lucas polynomials. \par After these results they construct a suitable set to replace $\mathcal{K}(N(P))$ as follows: for any imaginary unit $I\in\mathbb{S}$, let $P^I$ be the polynomial obtained by projecting all coefficients of $P$ onto $\mathbb{C}_I:=\mathrm{span}_\mathbb{R}(1,I)$. Then $P^I$ can be seen as a complex polynomial. Let $\mathcal{K}_{\mathbb{C}_I}(P^I)$ be the convex hull in $\mathbb{C}_I$ of the zero set of $P^I$ and define the Gauss-Lucas snail of $P$ as \[\mathfrak{sn}(P):=\bigcup_{I\in\mathbb{S}}\mathcal{K}_{\mathbb{C}_I}(P^I).\] Then the quaternionic Gauss-Lucas version proved in the paper is the following inclusion: for every $P\in\mathbb{H}[X]$ of degree $d\geq 2$, \[V(P')\subset \mathfrak{sn}(P).\] \par After the main theorem, the authors prove some topological properties of the Gauss-Lucas snail and then they use these results to prove two estimates on the size of the zero set of $P$. \par The article is very well written in an exhaustive form but nonetheless various explanatory examples are given.