The generalized Lax conjecture is true for topological reasons related to compactness, convexity and determinantal deformations of increasing products of pointwise approximating linear forms

Alejandro González Nevado · 2026

Read The generalized Lax conjecture is true for topological reasons related to compactness, convexity and determinantal deformations of increasing products of pointwise approximating linear forms by Alejandro González Nevado (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-12267.txt.

Introduction

Alejandro González Nevado claims a proof of the generalized Lax conjecture by topological means. The conjecture links rigidly convex sets to determinants of symmetric linear pencils. RZ polynomials are real polynomials with set zero traits tied to those sets. His route uses compactness, convexity, and determinantal deformations. He builds determinants of large pencils as products of approximating linear forms grow. A proof would close a long open problem in convex algebraic geometry. The work sets out that topological case in full.

Catalog record

CREATORAlejandro González Nevado
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-12267

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

The generalized Lax conjecture is true for topological reasons related to compactness, convexity and determinantal deformations of increasing products of pointwise approximating linear forms Alejandro González Nevado Abstract We develop a topological approach to prove the generalized Lax conjecture using the fact that determinants of sufficiently big symmetric linear pencils are able to express the rigidly convex sets of RZ polynomials of any degree d d . Monicity of the representation is assessed through a topological argument that allows us to perturbate a sufficiently close linear approximation into a suitable nice determinantal multiple of the initial RZ polynomial with the same rigidly convex set. The perturbation can be smoothly performed. This fact is what will allow us to determine that the multiple obtained respects the initial rigidly convex sets. This argument provides thus a full proof of the generalized Lax conjecture. However, an effective proof providing the representation in nice terms seems far from reachable at this moment. 1 Introduction The generalized Lax conjecture is one of the most important open problems at the intersection of mathematical optimization and real algebraic geometry. It deals with the representation of sets delimited at the same time by an algebraic (monic symmetric determinantality) and a geometric property (keeping rigidly convex set). These sets are defined by a border given by the set of real zeros around the origin of a polynomial having only real zeros across each real line through the origin. The exact definition is as follows. Definition 1 (RZ polynomials and RCSs) . Let p\in\mathbb{R}[\mathbf{x}]=\mathbb{R}[x_{1},\dots,x_{n}] be a polynomial with real coefficients. We say that p is real zero if, for each a\in\mathbb{R}^{n} , we have that the univariate polynomial p(ta) has all its roots real. Let V\subseteq\mathbb{R}^{n} be the set of real zeros of p . We call the set \rcs(p) given by the Euclidean closure of the connected component of \mathbf{0} in \mathbb{R}^{n}\smallsetminus V the rigidly convex set of p . A classic example of such polynomials are determinants of MSLMP (monic symmetric linear matrix polynomials). It is know that, in the case of two variables, these are actually all the polynomials verifying that property. Theorem 2 (Helton-Vinnikov) . [ 5 , Theorem 2.2] Let p\in\mathbb{R}[x_{1},x_{2}] be a RZ polynomial of degree d with p(\mathbf{0})=1 . Then there exist real symmetric matrices A_{1} and A_{2} of size d such that p=\det(I_{d}+x_{1}A_{1}+x_{2}A_{2}). And the entries of these matrices can be generically explicitly given by continuous functions of the coefficients. The generalized Lax conjecture asks also loosely if something like this happens for more variables. The immediate and naive generalization of such result is not true by dimensionality reasons. It is already well known that, for dimensionality reasons, there are more RZ polynomials than there are determinantal polynomials when n>2 . Theorem 3 (Dimensionality obstruction) . [ 8 , Theorem] The space of RZ polynomials in n variables of degree d contains an open inside the space of polynomials in the same number of variables and degree \mathbb{R}[\mathbf{x}]_{\leq d} . In fact, every such RZ polynomial p\in\mathbb{R}[\mathbf{x}]_{\leq d} can be expressed as the limit of smooth real zero polynomials of the same degree and number of variables. The dimension of the space \mathbb{R}[\mathbf{x}]_{\leq d} is, for n>2 and d>1 , bigger than the amount of (independent) entries in any possible MSLMP representation of these polynomials. The cited theorem of Nuij says even more about the topology of this space. We copy the theorem adapting it to the environment of RZ polynomials directly. Theorem 4 (Simplicity as topological space) . The topological space of RZ polynomials verifies the following properties. 1. The space of smooth RZ polynomials is open. 2. Every RZ polynomial is the limit of smooth RZ polynomials. 3. The space of normalized ( p(\mathbf{0})=1 ) RZ polynomials is connected and simply connected. We cannot have a determinantal representation because of the dimensionality obstruction presented in Theorem 3 . However, the requirement of a determinantal representation can be softened in order to try to produce a positive result. Several softenings have been shown to have exceptions and therefore are not true in general. Other softenings are well known to be true but the conditions are so weak that the original rigidly convex set could be destroyed along the way. For this reason, the generalized Lax conjecture (GLC) sits in the middle of these. The conjecture that we are going to prove here is the next softening of having a MSLMP determinantal representation. Conjecture 5 (Generalized Lax conjecture) . [ 1 , Conjecture 1.5] Let p\in\mathbb{R}[\mathbf{x}] be a RZ polynomial with p(\mathbf{0})=1 . Then there exist another RZ polynomial q\in\mathbb{R}[\mathbf{x}] with \rcs(p)\subseteq\rcs(q) and symmetric matrices A_{1},\dots,A_{n} such that qp=\det(I+x_{1}A_{1}+\cdots+x_{n}A_{n}) . As we can see, the algebraic part is softened as much as possible while the geometric requirement of keeping the original rigidly convex set untouched is required. In this way, we keep a LMI representation of \rcs(p) , which is another way of phrasing the GLC. Conjecture 6 (GLC) . Every rigidly convex set is spectrahedral, i.e., admits a representation as a MSLMI of the form \rcs(p)=\{\mathbf{x}\in\mathbb{R}^{n}\mid I+x_{1}A_{1}+\cdots+x_{n}A_{n}\mbox{\ is PSD}\}. This conjecture has been proved in several instances of polynomials having further structure. See [ 1 ] , for some examples. This includes cases for some well-known polynomials coming from combinatorics. However, it seems difficult to obtain general algebraic effective results for polynomials not associated to combinatorial objects. This is something one could expect. The form of the entries of the matrices in the LMI could be given in terms of very wild algebraic functions of the coefficients and there is no reason believe that we know much about these functions. Thus, this general algebraic effectivity seems like a too strong requirement for three reasons. We detail here which are these reasons and what makes us think that following general effectivity is a too complicated task to pursue upfront. 1.1 Size explosion It is well-known that the size of the matrices involved in the possible monic symmetric determinantal representation of an arbitrary RZ polynomial grows very fast . This implies that any effective algebraic description would have to deal with too much information and therefore being extremely difficult to obtain. In particular, the accelerated growth we mention comes in the form of a theorem that bounds our problem from below. This means that the situation could in general be even much worse. Remark 7 (Compact RCS) . For the next theorem (and in general in the future unless we say the opposite), we will suppose wlog that our RZ polynomials have compact RCSs. Theorem 8 (Uncontrollable growth of size of the LMP) . [ 11 , Theorem 1.1] There exists an absolute constant \kappa>0 such that for all sufficiently large n,d , there exists a RZ polynomial p\in\mathbb{R}[\mathbf{x}] of degree d in n variables whose rigidly convex set \rcs(p) does not admit an \eta -approximate MSLMI representation of size B\leq(n/d)^{\kappa d} for \eta=1/n^{4nd}. As approximations will be central for our work here, we also remind the reader of what kind of approximations the result above talks about. Definition 9 (Approximation) . Let (X,d) be a metric space and A\subseteq X bounded. For x\in X we define the radius map r(x,A)=\inf\{d(x,a)\mid a\in A\}

Read the complete text