Ultracategories as colax algebras for a pseudo-monad on 𝖢𝖠𝖳 \mathsf{CAT}

Ali Hamad · 2025

Read Ultracategories as colax algebras for a pseudo-monad on 𝖢𝖠𝖳 \mathsf{CAT} by Ali Hamad (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-2502-20597.txt.

Introduction

Ali Hamad wrote a 2025 mathematics paper on ultracategories, a structure from category theory built to hold limit operations of infinite size. A conjecture by Shulman held that these structures, as defined by Lurie, are a kind of loose algebra over a monad-like structure on the universe of categories. Hamad proves that claim. Under this view, the two flavours of functors between ultracategories also fall out as the natural loose and tight maps between such algebras.

Catalog record

CREATORAli Hamad
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE129 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-20597

Browse the full library · About the Tower of Babel

Full text

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

Ultracategories as colax algebras for a pseudo-monad on 𝖢𝖠𝖳 \mathsf{CAT} Ali Hamad Abstract We show a result inspired by a conjecture by Shulman claiming that ultracategories as defined by Lurie are normal colax algebras for a certain pseudo-monad on the category of categories 𝖢𝖠𝖳 \mathsf{CAT} . Such definition allows us to regard left and right ultrafunctors as instances of lax/colax algebra morphisms. 1 Introduction The ultraproduct construction plays a fundamental role in model theory. Makkai introduced the concept of ultracategories and ultrafunctors [ 14 , 15 ] , and used it to show the strong conceptual completeness theorem, which states that for any small pretopos \mathcal{T} , there is an equivalence of categories between \mathcal{T} , and the category of ultrafunctors from \mathrm{Mod}(\mathcal{T}) to \mathsf{Set} . Here, ultrafunctors are functors that preserve the ultraproduct up to coherence. Early work exploring these notions further includes Marmolejo’s 1995 PhD thesis [ 16 ] , which gave a 2 -monadic treatment of Makkai’s ultracategories, and Zawadowski [ 23 ] , although this early work is not directly related to the results of the present paper. Lurie [ 13 ] built on Makkai’s work, introducing a much simplified notion of ultracategories, and various notions of functors between these, namely ultrafunctors, right ultrafunctors, and most importantly left ultrafunctors, and gave a different proof of Makkai’s conceptual completeness (although Lurie and Makkai’s definitions of ultracategories and ultrafunctors differ, they agree in the statement of Makkai’s conceptual completeness). Towards this, he showed another important result, which generalises Makkai’s conceptual completeness, and Deligne’s completeness (the statement that any coherent topos has enough points), and states that for any coherent topos there is an equivalence between \mathcal{T} and left ultrafunctors from \mathrm{Mod}(\mathcal{T}) to \mathsf{Set} . Left ultrafunctors are a weakening of the definition of ultrafunctors, and unlike ultrafunctors which seem to impose a very restrictive condition on functors (preservation of ultraproduct), are much more abundant in mathematical structures (for example, the \mathbf{GNS} construction in operator algebras is a left ultrafunctor [ 9 , 8.4] ). Rewriting ultracategories in a monadic definition was an early question. In this direction we mention Clementino and Tholen [ 6 ] , who introduce a notion of ultracategories as an instance of a generalised multicategory, in particular a (\beta\text{-}\mathsf{Set}) -multicategory where \beta is the ultrafilter monad, however their notion looked more like generalised ultracategories (which we will talk about in the next paragraph) than ordinary ultracategories. In a blog comment [ 20 ] , Shulman postulated that ultracategories are normal colax algebras for a certain pseudo-monad on the category of categories. We prove a corrected version of this conjecture, where we had to make some modifications to the definition suggested by Shulman, to obtain the intended result. We give the definition of the pseudo-monad T in section 3.2 , and we think of this monad as sending a category to the category of “formal” ultraproducts on this category, while the algebra structure in a category recovers the actual ultraproducts (think of any free algebraic structure monad for an oversimplified comparison). In general our approach is also similar to the unrelativised pseudo-monad of the relative ultrafilter monad [ 21 ] , this pseudo-monad resembles our T except that morphisms are not quotiented by an equivalence relation, the colax algebras for this pseudo-monad are what the authors call “weak ultracategories”. Rosolini and Garner independently investigated the same monad of this paper where it was called the ultracompletion pseudo-monad [ 18 ] , with the direction of their research going into understanding Makkai’s ultracategories, and drawing comparison to Lurie’s. Indeed, T restricted to the one-object category \mathbf{1} recovers exactly Blass’s classical category of ultrafilters [ 3 ] . Other work on categorification of ultraproducts include Garners construction of a process \mathcal{UF} , which he used among other things to give a categorical description of ultraproducts of sets [ 8 ] , and also to give a categorical description of the ultraproduct of ultrafilters (called in his case the directed sum of ultrafilters), we will later see that the construction \mathcal{UF} can be used to understand, for a category A the category TA , but without any functoriality or monad structure. Moreover, we should mention that a categorical description of ultraproducts, in this case for all models of coherent logic, was given by Di Liberti [ 7 ] . One motivating reason behind introducing this monad will not be clear until our further work which introduces generalised ultracategory in the sense of the author’s own work [ 10 ] , also called virtual ultracategories by Saadia [ 19 ] and ultraconvergence spaces by Tarantino, Marquès and van Gool [ 22 ] , these are relational counterparts to ultracategories in which we don’t have an ultraproduct but we have the data of formal “representables” at ultraproducts i.e we have for each object A and family of objects (B_{i})_{i\in I} a generalised \mathrm{Hom} -set, which we denote by \mathrm{Hom}(A,\int_{I}B_{i}d\omega) . It turns out that every pseudo-algebra for the pseudo monad T (which we will now call a strong ultracategory) is in fact a generalised ultracategory, so generalised ultracategories are a generalisation of T -pseudo-algebras (strong ultracategories). In Saadia’s formulation, the definition is simplified by having a Hom functor X\times TX\to\mathsf{Set} , where T is the present monad restricted to the discrete category on X . In a recent paper, Aristote and Tarantino [ 1 ] characterised generalised ultracategories/virtual ultracategories/ ultraconvergence spaces as normalised algebras for an extension of the ultracompletion pseudo-monad T (in their cases called \beta ) to the category of categories with profunctors and natural transformations \mathsf{Prof} , and strong ultracategories (i.e T pseudo-algebras) are the representables among those. 1.1 Outline of methodology The main result that we want to show is the following: Theorem . Let \mathrm{Ult}_{L} (resp. \mathrm{Ult}_{R} , resp. \mathrm{Ult} ) denote the bicategory of ultracategories with left ultrafunctors (resp. right ultrafunctors, resp. ultrafunctors) as 1 -cells, and natural transformations (in the appropriate structure-respecting sense) as 2 -cells, then 1. There is an equivalence of bicategories between \mathrm{Ult}_{L} (resp. \mathrm{Ult}_{R} , resp. \mathrm{Ult} ), and the bicategory of normal colax algebras for the pseudo-monad T as we defined it in 3.2 , with lax morphisms as 1 -morphisms (resp. colax morphisms, resp. pseudo-morphisms) and with the respective 2 -morphisms of those. 2. This equivalence is realised concretely by a natural bijection between ultrastructures on a category A and normal colax T -algebra structures on A , and given two ultracategories A and B by a bijection between left ultrafunctors, (resp. right ultrafunctors) (resp. ultrafunctors) and left/lax morphisms (resp. right/colax morphisms, resp. pseudo-morphisms) of normal colax algebras, moreover given two such functors, there is a bijection between natural transformations of left ultrafunctors (resp. right ultrafunctors) (resp. ultrafunctors) between these functors, and the corresponding 2\text{-} morphisms between the T -algebra morphism structures on these functors. The first few sections give some necessary definitions: in section 2 , we give a comprehensive definition of the 2 -categories of ultracategories in question: ultracategories with left (right) ultrafunctors and natural transformations of left (right) ultrafunctors. Section 3 provides a definition of various 2 -categories of colax algebras over a 2 -monad. After that, we give the necessary definitions needed to

Read the complete text