Model theory of class-sized logics

Jonathan Osinski and Trevor Wilson · 2026

Read Model theory of class-sized logics by Jonathan Osinski and Trevor Wilson (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-2604-21678.txt.

Introduction

Osinski and Wilson study fragments of class-sized logic, where statements may range over proper classes rather than only sets. In unrestricted class logics, key model-theoretic properties fail. The authors ban certain mixes of infinitary quantifiers and boolean connectives, and show the restricted setting behaves far better. Its compactness and Löwenheim-Skolem properties characterise large cardinals, including Weak Vopěnka's Principle, Ord is Woodin, and Shelah cardinals.

Catalog record

CREATORJonathan Osinski and Trevor Wilson
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE98695
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.21678

Browse the full library · About the Tower of Babel

Full text

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

Model theory of class-sized logics Jonathan Osinski and Trevor Wilson Abstract Abstract. We study compactness and Löwenheim-Skolem properties of fragments of the class-sized logic ℒ ∞ ​ ∞ \mathcal{L}_{\infty\infty} and of class-sized versions of second-order and sort logics. In these fragments, certain combinations of infinitary quantifiers and boolean connectives are banned. While model-theoretic properties fail for unrestricted class logics, this drastically changes in our more restricted setting. We show that model-theoretic properties of class logics characterise a wide array of large cardinals, and that some of them can even be obtained in ZFC . In particular, we give a characterisation of Weak Vopěnka’s Principle and Ord is Woodin by downwards Löwenheim-Skolem properties, and a characterisation of Shelah cardinals by a compactness property of class-sized logics. We further strengthen many known results about properties of set-sized logics by studying how they transfer to class-sized extensions. 1 Introduction In † † The results of this article also appear in the first author’s PhD thesis [ Osi24 , Chapter 5] . many discussions of model theory of strong logics \mathcal{L} , i.e., of extensions of first-order logic, it has become somewhat standard to reserve the term logic to set-sized languages, i.e., to languages for which, given any set of non-logical symbols \tau , the collection \mathcal{L}[\tau] of \mathcal{L} -sentences over \tau is again a set (cf., e.g., [ Mak85 , BDGM24 , Osi24 , Lüc25 ] as examples for this naming practice). As a consequence, some languages classically studied in abstract model theory are not considered to be a logic in this sense, e.g., the infinitary language \mathcal{L}_{\infty\infty} allowing for conjunctions and disjunctions over arbitrarily sized sets of sentences, and quantification over arbitrarily sized sets of variables. Let us call \mathcal{L} a class-sized logic or class logic when \mathcal{L}[\tau] , the collection of \mathcal{L} -sentences over \tau , is a proper class for some \tau . The reason that class-sized logics are often not considered is that many of the usual properties studied in abstract model theory are easily seen to fail for \mathcal{L}_{\infty\infty} (cf. Proposition 6 ), arguably the easiest example of a class-sized logic available. Hence, already \mathcal{L}_{\infty\infty} does not have much of an interesting model theory. In this article, we show that this drastically changes when considering class-sized logics which are not closed under negation. We study class-sized subsystems of \mathcal{L}_{\infty\infty} , as well as of stronger languages like class-sized versions of second-order and sort logics. We show that assertions about properties of many of these systems are equivalent to the existence of large cardinals, and that some of them even have model-theoretic properties obtainable in ZFC. In some cases, these results strengthen known theorems about properties of set-sized logics. In others, these results give completely new characterizations of large cardinals. As a motivation, let us mention some background about the relation of model theory of strong logics and large cardinals. How large cardinals are characterized by model-theoretic properties of strong logics is a classical realm of study since Tarski [ Tar62 ] asked the question how the Compactness Theorem for first-order logic carries over to infinitary logics of the form \mathcal{L}_{\kappa\kappa} for an uncountable regular cardinal \kappa . The answer led to the introduction of weakly and strongly compact cardinals. Until the 1980s, many prominent large cardinal axioms were found to be equivalent to statements about model-theoretic properties of strong logics (cf., e.g., [ Mag71 , Ben78 , Mak85 ] ). Recently, this type of research had a resurgence, and many more connections of large cardinals to compactness properties of logics, generalizing the Compactness Theorem, as well as to (upward and downward) Löwenheim-Skolem properties, generalizing the Löwenheim-Skolem Theorems, were discovered (cf. [ MV11 , BV16 , Bon20 , HM22 , BDGM24 , HLM24 , Osi24 , GKV25 , GO25 , Lüc25 , BO25 , OP26 ] ). The article is structured as follows. In Section 2 we recall some basic definitions and background from abstract model theory, introduce the class-sized logics we will work with, and collect definitions and background on large cardinal notions we will consider. In Section 3 we study how switching from the set-sized versions of first-order logic, infinitary logics, second-order logics, and sort logics, to certain class extensions impacts their respective compactness and upward Löwenheim-Skolem properties. We show that in some cases, this switch comes with a jump in large cardinal strength of the considered property, while in others, the strength stays the same. In Section 4 , we study downwards Löwenheim-Skolem properties of class extensions of sort logic. In this vein, we obtain the first known characterization of \Pi_{n} -strong cardinals , Weak Vopěnka’s Principle , and Ord is Woodin by Löwenheim-Skolem properties. The connection between \Pi_{n} -strong cardinals, Weak Vopěnka’s Principle, and Ord is Woodin was recently uncovered by work of Wilson [ Wil22b ] and Bagaria and Wilson [ BW23 ] . That the usual Vopěnka’s Principle has several model-theoretic characterizations is well-known. For instance, it is equivalent to axiom schemas stating that all logics have certain compactness properties and downwards Löwenheim-Skolem properties, respectively. Whether the same is true for Weak Vopěnka’s Principle has been open. We give a positive answer by showing that Weak Vopěnka’s Principle is equivalent to the simultaneous existence of certain Löwenheim-Skolem numbers for a large collection of class logics. Independently, also Boney and Osinski [ BO25 ] , and, Holy, Lücke, and Müller [ HLM24 ] , showed that Weak Vopěnka’s Principle can be characterized by model-theoretic properties of strong logics. Both approaches use compactness principles though. By our characterization by Löwenheim-Skolem properties, we thus strengthen the model-theoretic analogy between Vopěnka’s Principle and Weak Vopěnka’s Principle in also carrying over to the latter types of properties. Finally, in Section 5 we provide the (to our best knowledge) first known characterization of Shelah cardinals by properties of strong logics, using a property akin to weak compactness of a class logic. 2 Preliminaries 2.1 Logics Let us start by making precise what we understand by a logic , as an abstract object of study. The definition is the standard one taken from [ BF85 , Chapter II, Definition 1.1.1] . Note that we do not follow the naming practice described in the beginning, in which a logic has to have set-many sentences over set-sized vocabularies. We will consider many-sorted vocabularies, as one of the logics we will consider is sort logic , whose semantics is naturally expressed in this context. As in standard definitions in first-order model theory, a (many-sorted) vocabulary \tau is a set consisting of finitely many sort symbols and an arbitrary amount of finitary relation, function, and constant symbols. We denote the set of sort symbols in \tau by \text{s}(\tau) . We demand \text{s}(\tau)\neq\emptyset . A \tau -structure is a tuple consisting of non-empty sets A_{s} for every sort symbol s\in\tau and interpretations for the relation, function, and constant symbols. We write A for \bigcup\{A_{s}\colon s\in\text{s}(\tau)\} . If \tau and \sigma are vocabularies, a bijection f:\tau\rightarrow\sigma is called a renaming if it respects arities and restricts to bijections of the respective subsets of sort, relation, function, and constant symbols in \tau and \sigma . If \mathcal{A} is a \tau -structure and f:\tau\rightarrow\sigma a renaming, then f

Read the complete text