Imaginary projections of polynomials (Q1994130): Difference between revisions

For a multivariate polynomial $f\in\mathbb{C}[\mathbf{z}]=\mathbb{C}[z_1,\dots,z_n],$ let $\mathcal{V}(f)$ be its variety. The article initiates a study of the \textit{imaginary projection} of $f$, $\mathcal{I}(f):= \{\mathrm{Im} (\mathbf z): \mathbf{z} \in \mathcal{V}(f)\}= \{\mathbf{y}\in \mathbb{R}^n: f(\mathbf{x}+i\mathbf{y}) =0\}$ noting various properties sets $\mathcal{I}(f)$ have and have not in common with amoebas $\mathcal {A}(f)=\{(\log |z_1|,\dots,\log |z_n|): \mathbf{z}\in \mathcal{V}(f)\cap (\mathbb{C}^*)^n\},$ which since their introduction in 1994 by Gelfand, Kapranov and Zelevisky [\textit{I. M. Gelfand} et al., Discriminants, resultants, and multidimensional determinants. Boston, MA: Birkhäuser (1994; Zbl 0827.14036)] have generated lots of interest. Another motivation for the study might well have been the impressive number of old conjectures recently solved in which \textit{stable polynomials} $f,$ i.e. those for which $\mathcal{I}(f)\cap \mathbb{R}_{>0}^n=\emptyset$ holds, played a role; see e.g: \textit{L. Gurvits} [Electron. J. Comb. 15, No. 1, Research Paper R66, 25 p. (2008; Zbl 1182.15008)]; \textit{J. Borcea} and \textit{P. Brändén} [Duke Math. J. 143, No. 2, 205--223 (2008; Zbl 1151.15013)]; \textit{A. W. Marcus} et al. [Ann. Math. (2) 182, No. 1, 327--350 (2015; Zbl 1332.46056)]. \par Section 2 gives a summary of stability results of Borcea and Brändén who generalized, for example, the classical Hermite-Biehler theorem for univariate polynomials to the multivariate case, and showed the stability of an arbitrary polynomial in $\mathbb{C}[\mathbf{z}]$ as equivalent to the stability of its polarization. Section 3 investigates the imaginary projection of affine linear polynomials, i.e. those of the form $f(\mathbf{z})=a_0+\sum_{i=1}^n a_j z_j.$ In this case $\mathcal I(f)=\mathbb{R}^n$ except possibly when $\mathbf{a}=(a_1,\dots,a_n)$ equals $e^{i\varphi } \mathbf{b}$ for some $\mathbf{b}\in\mathbb{R}^n,$ $\varphi\in [0,2\pi[.$ \par Section 4 investigates the complements in $\mathbb{R}^n$ of the closure of imaginary projections, $\overline{\mathcal I(f)}^c.$ In analogy to $\mathcal A(f)^c$, it is shown that $\overline{\mathcal I(f)}^c$ consists of finitely many convex components. The proof of this is short but uses Bochner's tube theorem [\textit{S. Bochner}, Ann. Math. (2) 39, 14--19 (1938; Zbl 0018.15303; JFM 64.0321.02)] as well as the fact that $\mathcal I(f)^c$ is a semialgebraic set so that bounds on Betti numbers associated to the names Oleinik, Petrowskij, Milnor can be applied. Using suitable hyperplane arrangements one can show via section 3 that for any integer $t>0$ an $f\in \mathbb{C}[\mathbf{z}]$ with exactly $t$ bounded components in $\mathcal I(f)^c$ can be constructed. A homogeneous polynomial in $\mathbb{R}[\mathbf{z}]$ is \textit{hyperbolic}, see \textit{L. Gårding} [J. Math. Mech. 8, 957--965 (1959; Zbl 0090.01603)], with respect to a point $\mathbf{x} \in \mathbb{R}^n$ and a direction $\mathbf{e} \in \mathbb{R}^n$ if $f(\mathbf{e})\neq 0$ and the polynomial $\mathbb{R}\ni t\mapsto f(\mathbf{x}+t\mathbf{e})$ has only real zeros. \par Theorem 4.6 says for homogeneous $f\in \mathbb{R}[\mathbf{z}]$ that $\mathcal I (f)\neq \mathbb{R}^n$ if and only if $f$ is hyperbolic with respect to some direction. Section 5 presents a complete classification of the imaginary projections of quadratic polynomials. For example, if $f=\sum_{j=1}^p z_j^2 -\sum_{j=p+1}^r z_j^2 +1,$ $3\leq r\leq n$ and $p=r-1,$ then $\mathcal I (f)=\{\mathbf{y}\in \mathbb{R}^n: \sum_{j=1}^{r-1} y_j^2> y_r^2 \}\cup \{\mathbf{0}\}.$ As a consequence of lemmas leading to the classification result, a necessary and sufficient criterion for the existence of an unbounded component in $\overline{\mathcal{I}(f)}^c,$ $f\in \mathbb{C}[\mathbf{z}]$ is found. \par Section 6 determines the set of limit directions of the imaginary projections of $f,$ namely $\mathcal I_\infty (f)=\lim_{r\rightarrow \infty} \left(\frac{1}{r}\mathcal{I}(f) \cap \mathbb{S}^{n-1}\right).$ Contrary to the analogously defined set of limit directions $\mathcal A_\infty (f)$ for amoebas, $\mathcal I_\infty (f)$ is usually not a spherical polyhedral complex. Consider a nonconstant $f\in\mathbb{C}[\mathbf{z}]$ and let its homogenization $f_h\in \mathcal C[z_0,\mathbf{z}]$ have a zero $\mathbf{p}_h =(0:p_1:\cdots :p_n)=(0:\mathbf{p}),$ then every point in the intersection of $\mathcal H=\{\lambda \mathrm{Re}(\mathbf{p})+ \mu \mathrm{Im}(\mathbf{p}):\lambda,\mu \in \mathbb{R} \}$ with $\mathbb S^{n-1}$ is a limit direction. \par The paper has colored pictures of amoebae and imaginary projections and closes with open questions. For example no results on the number of components of the complement of an imaginary projection $\mathcal I(f)$ as a function of the Newton polytope of $f$ seem to be known.