Invariant Random Subgroups, Soficity, and Lück’s determinant conjecture
Aareyan Manzoor · 2025
Read Invariant Random Subgroups, Soficity, and Lück’s determinant conjecture by Aareyan Manzoor (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-2508-15154.txt.
Introduction
Aareyan Manzoor wrote a 2025 math paper extending a famous determinant conjecture from groups to random subgroups. Lück's conjecture bounds the determinants that arise from a group's operators, and it is tied to deep questions about which groups can be approximated by finite ones, a property called soficity. A random subgroup setting is a wider framework where non sofic objects are known to exist.
For every free group, Manzoor constructs a random subgroup that satisfies the determinant conjecture yet is provably not approximable in the sofic sense. That separation hints the conjecture may be strictly weaker than soficity, and could hold for all groups. The proof borrows tools from nonlocal games and recent complexity theory breakthroughs. It also gives a general recipe for locating exotic objects inside operator algebras.
Catalog record
| CREATOR | Aareyan Manzoor |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 34 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-15154
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Invariant Random Subgroups, Soficity, and Lück’s determinant conjecture Aareyan Manzoor Abstract We extend Lück’s determinant conjecture from groups to invariant random subgroups (IRS) of free groups, a framework generalizing groups where a non-sofic object is known to exist. For every free group, we prove the existence of an IRS satisfying the determinant conjecture that is not co-hyperlinear, and hence not co-sofic. This provides evidence that satisfying the determinant conjecture might be a weaker property than soficity for groups, and consequently the conjecture possibly holds for all groups. We use techniques from non-local games and 𝖬𝖨𝖯 ∗ = 𝖱𝖤 \mathsf{MIP}^{*}=\mathsf{RE} , showing more generally when the latter can be used to narrow down when a von Neumann algebra (or IRS) contains a non-Connes embeddable object. keywords Connes embedding, Aldous-Lyons, invariant random subgroup, Lück’s determinant conjecture Classification: 46L10 (primary), 20F69, 81P99 (secondary) 1 Introduction Let \Gamma be a discrete group, L(\Gamma) its group von Neumann algebra, and \tau the standard trace on L(\Gamma) . We say that \Gamma satisfies Lück’s determinant conjecture if (\tau\otimes\mathrm{tr}_{n})\big(\ln_{+}(A^{*}A)\big)\geq 0\quad\forall\,A\in M_{n}(\mathbb{Z}[\Gamma]),\ n\in\mathbb{N}, where \mathrm{tr}_{n} is the normalized trace on M_{n}(\mathbb{C}) , and \ln_{+}(t)=\begin{cases}0,&t=0,\\ \ln(t),&t>0.\end{cases} When exponentiated, this quantity is called the modified Fuglede–Kadison determinant of A . In other words, \ln_{+}(A^{*}A) is the logarithm of A^{*}A with zeros ignored; exponentiating recovers the product of its non-zero singular values. Conjecture 1.1 ( [ Lüc02 ] ) . Every discrete group satisfies Lück’s determinant conjecture. The determinant conjecture is deeply connected to both operator algebras and topology. For instance, if a finitely generated group satisfies it and has vanishing first \ell^{2} -Betti number, then L(\Gamma) is strongly 1 -bounded [ HJE23 , Shl21 ] , a rigidity property in free probability that rules out embeddings of free group factors. It is also essential for the definition of L^{2} -torsion [ Lüc09 ] , a powerful invariant in geometry and topology. The determinant conjecture is known to hold for all sofic groups [ ES05 ] . Whether every group is sofic remains an open problem, though it is widely believed that non-sofic groups exist. This makes it difficult to distinguish the determinant conjecture from soficity solely in the setting of groups. We can circumvent this problem by working in the framework of invariant random subgroups (IRSs) of the free group. IRSs provide a setting where non-sofic analogues are known to exist, allowing us to separate the determinant conjecture from soficity in a way not possible for groups themselves. An IRS is a conjugation-invariant Borel probability measure on the set of subgroups of a group. Dirac measures on normal subgroups are examples of IRSs, so this concept generalizes the notion of a normal subgroup. Since normal subgroups of free groups correspond to all finitely generated groups via quotients, IRSs of a free group can be seen as a natural generalization of groups themselves. The IRS analogue of soficity is called co-soficity . Recent work has shown that every free group admits a non-co-sofic IRS [ BCV24 , Bow+24 ] . In [ Man25 ] , the methods were significantly simplified and extended to produce a non-co-hyperlinear IRS on each free group—one failing the natural IRS analogue of hyperlinearity. We can extend the determinant conjecture from groups to IRSs in a straightforward way, leading to the following central question: Problem 1.2 . Do all IRSs of a free group satisfy Lück’s determinant conjecture? While we cannot yet resolve this fully, our main result provides evidence towards a positive answer. Theorem 1.3 . Let \Gamma be a free group. There exists an IRS H on \Gamma that satisfies Lück’s determinant conjecture but is not co-hyperlinear. Such an H cannot be co-sofic, as co-sofic IRSs are co-hyperlinear. This separation suggests that the determinant conjecture may be strictly weaker than soficity, strengthening the case that it could hold for all groups, including potential non-sofic ones. Note that the measured determinant conjecture [ LSW10 ] implies a positive solution to Problem 1.2 , so our result also provides evidence for the measured determinant conjecture. Our proof of Theorem 1.3 adapts the approach of [ Man25 ] , which is based on non-local games . This framework, together with the computability result \mathsf{MIP}^{*}=\mathsf{RE} [ Ji+22 ] , was used to prove the existence of a non–Connes embeddable von Neumann algebra. Much of the recent progress in narrowing down possible counterexamples to Connes’ embedding problem has relied on this method. A well-known criticism, however, is that the argument is highly non-constructive: \mathsf{MIP}^{*}=\mathsf{RE} ensures existence but does not produce a useful explicit example. For example, one reason to search for a non-sofic group is that it might clarify when a group fails the determinant conjecture. If such a group were obtained purely via \mathsf{MIP}^{*}=\mathsf{RE} , we would have no way to check if that group actually satisfied the determinant conjecture. While this is a limitation, the method can still be used to narrow down such a group. Section 3.1 serves as a proof of concept. Suppose there exists a “nice” class of algebras (resp. IRSs) that: 1. admits nice upper bounds, and 2. contains all finite-dimensional tracial von Neumann algebras (resp. IRSs). Then \mathsf{MIP}^{*}=\mathsf{RE} implies that this class must contain a non–Connes embeddable algebra (resp. IRS). Similarly, by \mathsf{MIP}^{\mathsf{co}}=\mathsf{coRE} , if a “nice” class admits computable lower bounds, it cannot contain every tracial von Neumann algebra (resp. IRS). This means the class of all tracial von Neumann algebras (resp. IRS) has no “nice” approximation property that encompasses the whole class. The IRSs satisfying the determinant conjecture fall in the former category, which allows us to get the result. See Proposition 3.6 for the formal statement. In Section 2 we extend the determinant conjecture to IRSs. Section 3 develops the non-local game framework and proves Theorem 1.3 , while Section 4 discusses broader implications. Section 3.1 develops a NPA hierarchy for strategies coming from IRSs satisfying the determinant conjecture, which forms the technical core of the paper. This hierarchy not only yields Theorem 1.3 , but also provides a template for applying non-local game techniques to other approximation problems involving IRSs and traces/von Neumann algebras. Acknowledgements. I would like to thank Michael Brannan for discussions on these results and supervision . I would also like to thank Georges Skandalis and Balchi for sharing Balchi’s PhD thesis with me, which helped inspire the NPA heirarchy in this paper. 2 Invariant Random Subgroups In this section we introduce the basic terminology for invariant random subgroups (IRSs) and recall how they give rise to canonical traces and von Neumann algebras. For a more detailed discussion, see [ Man25 , Section 2] . Definition 2.1 . Let \Gamma be a discrete group. Denote by \mathfrak{sub}(\Gamma) the space of subgroups of \Gamma , topologized as a subset of \{0,1\}^{\Gamma} . A random subgroup of \Gamma is a Borel probability measure on \mathfrak{sub}(\Gamma) . If the measure is invariant under the conjugation action of \Gamma on \mathfrak{sub}(\Gamma) , we call it an invariant random subgroup (IRS). We write \operatorname{IRS}(\Gamma) for the convex set of IRSs on \Gamma . Example 2.2 . If N\trianglelefteq\Gamma is a normal subgroup, then the Dirac measure \delta_{N} is an IRS. Thus, IRSs generalize the notion of normal subgroups. In what follows