Operator systems and positive extensions over discrete groups
Evgenios T.A. Kakariadis, Malte Leimbach, Ivan G. Todorov, Walter D. van Suijlekom · 2026
Read Operator systems and positive extensions over discrete groups by Evgenios T.A. Kakariadis, Malte Leimbach, Ivan G. Todorov, Walter D. van Suijlekom (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-2603-29958.txt.
Introduction
Kakariadis and co-authors recast a classic extension problem through operator systems. The question asks if positive definite functions on part of a group extend to the whole group. Since Rudin's work, it has tied to sums of squares in group algebras. They show the extension property matches a quotient map on the Fourier-Stieltjes algebra. The factorization side matches a complete order embedding into the group algebra. Examples connect the frame to nuclearity and C-star envelopes.
Catalog record
| CREATOR | Evgenios T.A. Kakariadis, Malte Leimbach, Ivan G. Todorov, Walter D. van Suijlekom |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 153654 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.29958
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Operator systems and positive extensions over discrete groups Evgenios T.A. Kakariadis, Malte Leimbach, Ivan G. Todorov, Walter D. van Suijlekom Abstract . The extension problem asks whether positive semi-definite functions on a symmetric unital subset of a discrete group can be extended to positive semi-definite functions on the whole group. It has been known at least since the work of Rudin in the 1960s that this is closely related to the problem of finding sums of squares factorisations of positive elements in the group C*-algebra. We give an operator system perspective at these two problems explaining their equivalence: the extension property is characterised by a certain quotient map on the Fourier–Stieltjes algebra, and the factorisation property by a certain complete order embedding into the group C*-algebra. These properties are linked to the duality of the operator systems which have recently emerged from spectral and Fourier truncations in noncommutative geometry. We exemplify how one can relate certain extension problems to operator system techniques such as nuclearity and the C*-envelope. 1. Introduction The study of positivity of functions via sums of squares of other functions has its roots in the work of Hilbert [ 27 ] , with a corresponding factorisation question formulated by him and becoming widely known as Hilbert’s 17th problem [ 28 ] . In the context of classical Fourier analysis, affirmative first results and counterexamples for positive semi-definite functions defined on subsets of abelian groups were obtained by Rudin [ 49 ] . The study of the sums-of-squares problem revealed deep and intricate connections with harmonic analysis, first in the case of abelian groups [ 25 , 4 , 10 , 49 ] and then in the non-abelian case; see e.g. [ 5 , 6 ] . Despite the negative general results, the failure of the characterisation of positivity as an expression of sums of squares led to fruitful semi-definite optimisation hierarchies. In the commutative case, these were first studied by Lasserre [ 35 ] and have had lasting impact in optimisation theory, while in the context of quantum information theory, sums-of-squares approximations for the problem of determining the quantum commuting value of a non-local game have been proposed with great success by Navascués, Pironio and Acín [ 42 ] . At the heart of these optimisation hierarchies lies a constrained resolution approach, where the test set (usually convex) for a given optimisation problem is represented as an increasing union (or the closure thereof) of subsets that lead to computationally amenable estimations of the cost function. A constrained resolution approach to noncommutative geometry has independently generated recent interest; here, the \mathrm{C}^{*} -algebras in the spectral triple formalism are replaced by operator systems, as a means to describe physical systems with incomplete information [ 13 ] . Specifically, in the group setting, one considers the state space of spectral truncations, which leads to the problem of extending partially positive semi-definite functions to globally positive semi-definite ones. The goal of this paper is to present a unifying approach to the extension and factorisation problems in general (not necessarily abelian) discrete groups, using the formalism of operator systems, motivated by the main examples appearing in the context of noncommutative geometry [ 13 ] . To be more precise, in the latter context, operator systems arise in two main (and complementary) ways: on the one hand as compressions PAP of a \mathrm{C}^{*} -algebra A\subseteq\mathcal{B}(H) by a projection P\in\mathcal{B}(H) (e.g. a spectral projection for a Dirac operator on H ) and, on the other hand, as certain operator subsystems of A . In the case of the \mathrm{C}^{*} -algebra A=\mathrm{C}(\mathbb{T}) of continuous complex-valued functions on the circle, the operator systems of interest are the Toeplitz system (2) \displaystyle\mathrm{C}(\mathbb{T})^{(N)}:=\{(T_{ij})_{i,j=0}^{N-1}\in\mathrm{M}_{N}(\mathbb{C})\mid T_{ij}=T_{(i+1)(j+1)}\text{ for all }i,j=0,\dots,N-1\} (the addition is mod N ), consisting of the N\times N Toeplitz matrices, and the Fourier system (3) \displaystyle\mathrm{C}(\mathbb{T})_{(N)}:=\{f\in\mathbb{T}\mid\widehat{f}_{n}=0\text{ when }n\geq N\}, that is, the operator subsystem of A\cong\mathrm{C}^{*}(\mathbb{Z}) spanned by finitely supported sequences (\dots,0,a_{-N+1},\dots,a_{0},\dots,a_{N-1},0,\dots) . Using the matrix-valued Fejér–Riesz lemma it was shown that the Toeplitz and the Fourier system are unitally completely order isomorphic to each other’s operator system duals [ 13 , 20 ] . As we will explain shortly, it seems appropriate to consider this duality of operator systems as a property of the group \mathbb{Z} and the subset \{1,\dots,N\}\subseteq\mathbb{Z} ; it is then natural to ask, for which discrete groups \Gamma and finite subsets \Sigma\subseteq\Gamma analogous results hold. To this end, for a discrete group \Gamma and a finite subset \Sigma\subseteq\Gamma , we denote by \mathrm{C}^{*}(\Gamma)^{(\Sigma)} the Toeplitz system , i.e. the operator system of Toeplitz matrices (4) \displaystyle(T_{st^{-1}})_{s,t\in\Sigma}\in\mathrm{M}_{\Sigma}\cong\mathcal{B}(\ell^{2}(\Sigma)), and by \mathrm{C}^{*}(\Gamma)_{(\Sigma\Sigma^{-1})} the Fourier system , i.e. the operator subsystem of the group C*-algebra \mathrm{C}^{*}(\Gamma) spanned by \delta_{st^{-1}} , s,t\in\Sigma , where \delta_{s} , s\in\Gamma , are the canonical unitary generators of \mathrm{C}^{*}(\Gamma) . The Toeplitz system \mathrm{C}^{*}(\Gamma)^{(\Sigma)} arises both as the compression of the (reduced) group \mathrm{C}^{*} -algebra by the orthogonal projection P:\ell^{2}(\Gamma)\rightarrow\ell^{2}(\Sigma) , and as the image of the restriction map \rho:\mathrm{B}(\Gamma)\rightarrow\mathrm{M}_{\Sigma} of functions in the Fourier–Stieltjes algebra \mathrm{B}(\Gamma) to the set \Sigma\Sigma^{-1} . We mainly adopt the latter point of view. Recalling that the dual of the group \mathrm{C}^{*} -algebra \mathrm{C}^{*}(\Gamma) is the Fourier–Stieltjes algebra \mathrm{B}(\Gamma) , it is natural to expect that the Toeplitz and Fourier systems are dual to each other if and only if the restriction map \rho is a quotient map. We make this precise in terms of the involved matrix order structures by showing that the Toeplitz and the Fourier system are dual to each other if and only if every positive semi-definite Toeplitz matrix (T_{st^{-1}})_{s,t\in\Sigma} with matrix entries extends to a positive semi-definite function on \Gamma , Theorem 4.22 . The duality of Toeplitz matrices and sums of squares was already exploited by Rudin [ 49 ] in relating the extension problem for partially defined positive semi-definite functions to a Fejér–Riesz type factorisation property (see [ 52 , 51 , 7 ] for extensive treatments, and [ 5 , Theorem 7.1] for an example of the explicit use of operator system duality which is similar to our Theorem 4.22 ). We capture the sums-of-squares concept by introducing the sums-of-squares operator system \mathrm{SOS}(\Sigma) and we show that the Toeplitz system \mathrm{C}^{*}(\Gamma)^{(\Sigma)} is its operator system dual. We infer that the Toeplitz system \mathrm{C}^{*}(\Gamma)^{(\Sigma)} and the Fourier system \mathrm{C}^{*}(\Gamma)_{(\Sigma\Sigma^{-1})} are each other’s operator system duals if and only if the Fourier system is completely order isomorphic to the sums-of-squares system (see Theorem 4.22 ). Note that if the group \Gamma is amenable, its group \mathrm{C}^{*} -algebra \mathrm{C}^{*}(\Gamma)\cong\mathrm{C}^{*}_{\mathrm{r}}(\Gamma) is a coamenable compact quantum group in the sense of Woronowicz [ 56 ] (with comultiplication given by \mathrm{C}^{*}_{\mathrm{r}}(\Gamma)\ni\lambda(s)\mapsto\lambda(s)\otimes\lambda(s)\in\mathrm{C}^{*}_{\mathrm{r}}(\Gamma)\otimes\mathrm{C}^{*}_{\mathrm{r}}(\Gamma) ), and the Toeplitz and Fourier systems are the operator systems arising respectively as Peter–Weyl [ 36 ] and Fourier truncations [ 47 ] . In this way, we provide the connection between these two types of truncations. As a byproduct we obtain that the operator system of Toeplitz matrices is not (min,max)-nuclear, thereby simplifying the corresponding argument in [ 20 ] (see also [ 21 ] ). While the non-nuclearity of the Toeplitz system is a special instance of a general characterisation of nuclear finite-dimensional operator systems [ 30 ] , it