The uniqueness of Lyapunov rank among symmetric cones
Michael Orlitzky, Giovanni Barbarino · 2025
Read The uniqueness of Lyapunov rank among symmetric cones by Michael Orlitzky, Giovanni Barbarino (2025) 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-2503-06143.txt.
Introduction
Michael Orlitzky and Giovanni Barbarino answer, in a 2025 math paper, a sorting question about a class of cones. A cone's Lyapunov rank is the size of the symmetry algebra tied to it. That rank does not change when a cone is reshaped by a linear map. Yet different cones can share the same rank, so rank alone usually cannot name a cone.
The authors narrow the field to symmetric cones, a well behaved family used in optimization. They ask whether any of those cones can be picked out uniquely by rank alone. For cones that cannot be split into simpler parts, they give a complete answer. For the general case, they make partial progress. The result clarifies what a single number can and cannot tell about shape, in a setting where such invariants guide theory.
Catalog record
| CREATOR | Michael Orlitzky, Giovanni Barbarino |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 60 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-06143
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The uniqueness of Lyapunov rank among symmetric cones Michael Orlitzky, Giovanni Barbarino Abstract The Lyapunov rank of a cone is the dimension of the Lie algebra of its automorphism group. It is invariant under linear isomorphism and in general not unique—two or more non-isomorphic cones can share the same Lyapunov rank. It is therefore not possible in general to identify cones using Lyapunov rank. But suppose we look only among symmetric cones. Are there any that can be uniquely identified (up to isomorphism) by their Lyapunov ranks? We provide a complete answer for irreducible cones and make some progress in the general case. Keywords: symmetric cone, Lyapunov rank, Euclidean Jordan algebra, isomorphism problem MSC2020: 15B48, 17C20, 17B45, 90C25 1 Introduction The motivation for Lyapunov rank comes from linear programming. The familiar complementary slackness condition \left\langle{x},{s}\right\rangle=0 for x,s\in\mathbb{R}^{n}_{+} can be separated into n equations \left\{{x_{i}s_{i}=0}\ \middle|\ {i=1,2,\ldots,n}\right\} using the nonnegativity of the components, and these n equations should be easier to solve simultaneously than the one equation we started with. Complementary slackness is a necessary condition for optimality, so being able to solve that system can lead us to candidate solutions. Rudolf, Noyan, Papp, and Alizadeh [ 7 ] realized that this scenario generalizes to a cone program over a proper cone K and its dual K^{*} . In a cone program, complementary slackness is still a necessary condition for optimality, and the the condition \left\langle{x},{s}\right\rangle=0 for x\in K and s\in K^{*} can always be separated into a system of equations via “bilinear complementarity relations.” The number of equations, denoted \beta\left({{K}}\right) , depends only on the cone K and is called the bilinearity rank (by them) or the Lyapunov rank (by us) of K . If you plan to search for candidate solutions by decomposing the complementary slackness condition, Lyapynov rank is a measure of how cooperative the cone will be. If the cone K lives in an n -dimensional space, then the set of all candidate pairs \left\{{\left({x},{s}\right)}\ \middle|\ {x\in K,\,s\in K^{*},\,\left\langle{x},{s}\right\rangle=0}\right\} forms an n -dimensional manifold [ 7 ] . We might therefore hope to obtain n or more equations ( \beta\left({{K}}\right)\geq n ) to be solved simultaneously. Rudolf, Noyan, Papp, and Alizadeh call such a cone bilinear , and show that several important cones are not bilinear. A few years later, Gowda and Tao were able to connect these ideas to Lyapunov-like operators [ 3 ] . The Lyapunov-like operators on K form the Lie algebra of its automorphism group, a vector space whose dimension is \beta\left({{K}}\right) . Linearly independent Lyapunov-like operators correspond to “extra” equations, and vice-versa. For this reason the name Lyapunov rank was chosen. Around the same time, Seeger and Sossa coined the term Loewnerian cone for a cone that is isomorphic to the real symmetric PSD cone, \mathcal{S}^{n}_{+} . As demonstrated by Seeger and Sossa, Loewnerian cones have important applications to complementarity problems [ 8 ] . But how do we know if a cone is Loewnerian? Sossa, Gowda, Tao, and one of the authors would eventually meet. The problem naturally arose: can Lyapunov rank be used to show that a given cone is Loewnerian? If the Lyapunov rank of K differs from that of \mathcal{S}^{n}_{+} , then K and \mathcal{S}^{n}_{+} are not isomorphic. The question therefore amounts to whether or not there exist any non-Loewnerian K having \beta\left({{K}}\right)=\beta\left({{\mathcal{S}^{n}_{+}}}\right) . We are able to answer the question for \mathcal{S}^{n}_{+} with relative ease, but the problem, inspires us to investigate further. First we ask whether or not there are any irreducible symmetric (self-dual and homogeneous) cones whose Lyapunov ranks are unique among other, non-isomorphic symmetric cones. This question is a bit harder to resolve, and the answer is unexpected. Afterwards, we move on to symmetric cones that are not necessarily irreducible. This is related to the isomorphism problem for symmetric cones and can have real applications: there do exist generic symmetric-cone optimization algorithms, but practically speaking, it is much better to know if your symmetric cone is really collection of PSD and/or second-order cones in disguise. 2 Background The main objects of interest are symmetric cones in finite dimensions. All such cones arise as the cone of squares in some Euclidean Jordan algebra [ 2 ] . Symmetric cones are self-dual, and in particular proper : closed, convex, pointed, and solid. A convex cone is reducible if it is the direct sum of two nontrivial convex cones, and irreducible if not. Every proper cone can be expressed as a direct sum of nontrivial irreducible cones, and this sum is unique up to the order of its factors (Hauser and Güler [ 4 ] , Theorem 4.3). This general result holds even in the absence of an inner product. When we have an inner product, and when our cone is symmetric, we can say much more: the decomposition will be orthogonal, the factors will be symmetric, and all potential factors have been classified up to isomorphism. Chapter V of Faraut and Korányi classifies the simple Euclidean Jordan algebras up to isomorphism [ 2 ] . The classification of irreducible symmetric cones then follows from the correspondence between symmetric cones and Euclidean Jordan algebras. Starting from a symmetric cone, one obtains a Euclidean Jordan algebra (Faraut and Korányi, Chapter III) in which the given cone is the cone of squares. That algebra is the orthogonal direct sum of simple algebras by Proposition III.4.4. In each of those simple factors, the cone of squares is a symmetric cone, and it must be irreducible to avoid contradictions. The resulting decomposition of a symmetric cone is Faraut and Korányi’s Proposition III.4.5, which we now combine with the classification of simple algebras. We use the term Jordan isomorphism for an invertible Jordan-algebra homomorphism, and Jordan-isomorphic when two sets are identified by one Theorem 1 . Every symmetric cone is the unique (up to order) orthogonal direct sum of nontrivial irreducible symmetric cones, and every nontrivial irreducible symmetric cone is Jordan-isomorphic to a member of one of the five distinct families, 1. The n -dimensional Lorentz cones, \mathcal{L}^{{n}}_{+} , for n\notin\left\{{0,2}\right\} , 2. The n\times n real symmetric positive semidefinite cones, \mathcal{H}^{n}_{+}\left({\mathbb{R}{}}\right) , for n\geq 3 , 3. The n\times n complex Hermitian positive semidefinite cones \mathcal{H}^{n}_{+}\left({\mathbb{C}{}}\right) , for n\geq 3 , 4. The n\times n quaternion Hermitian positive semidefinite cones 1 1 1 These cones of squares were identified as positive-semidefinite cones only recently [ 6 ] . , \mathcal{H}^{n}_{+}\left({\mathbb{H}}\right) , for n\geq 3 , 5. The cone of squares in the Albert algebra, \mathcal{H}^{3}_{+}\left({\mathbb{O}}\right) . As a result, every symmetric cone is Jordan-isomorphic to a unique (up to order) orthogonal direct sum whose factors come from the list above. The trivial cone in the trivial space, obtained as an empty sum, is also symmetric but is spiritually a member of the family \mathcal{L}^{{n}}_{+} with n=0 . If the “up to order” is bothersome, one can imagine a total order on the set of irreducible factors listed in Theorem 1 . We then obtain true uniqueness: every symmetric cone is Jordan-isomorphic to exactly one orthogonal direct sum whose factors come from the list and are sorted. The restrictions on n in Theorem 1 ensure that the families are distinct up to