A convenient category of cubes
Sanjeevi Krishnan and Emily Rudman · 2025
Read A convenient category of cubes by Sanjeevi Krishnan and Emily Rudman (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-2503-13663.txt.
Introduction
Sanjeevi Krishnan and Emily Rudman study a category of cubes in a 2025 mathematics paper. Cubical sets are shapes built from squares and cubes, used in homotopy theory and type theory. The authors argue that a particular cube category, whose maps preserve intervals and order between finite Boolean lattices, is a good general purpose base for such sets.
They show the category is the largest of its kind that excludes reversals and diagonals. Its maps can be built from a small set of generators. A model structure on their cubical sets matches the classical one for simplicial sets, and is proper, so its fibrations can interpret dependent types in logic. The work aims to unify several cubical traditions.
Catalog record
| CREATOR | Sanjeevi Krishnan and Emily Rudman |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 112 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-13663
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A convenient category of cubes Sanjeevi Krishnan and Emily Rudman Abstract . We claim that the cube category whose morphisms are the interval-preserving monotone functions between finite Boolean lattices is a convenient general-purpose site for cubical sets. This category is the largest possible concrete Eilenberg-Zilber variant excluding the reversals and diagonals. The category admits as monoidal generators all functions between the ordinals [ 0 ] [0] and [ 1 ] [1] and all monotone surjections [ 1 ] n → [ 1 ] [1]^{n}\rightarrow[1] . Consequently, morphisms in the minimal symmetric monoidal variant of the cube category containing coconnections of one kind can be characterized as the interval-preserving semilattice homomorphisms between finite Boolean lattices. There exists a model structure on our variant of cubical sets that is at once Quillen equivalent to and left induced from the classical model structure on simplicial sets along triangulation. This model structure is proper and hence its fibrations interpret Martin-Lof dependent types. 1. Introduction Spaces are often combinatorially represented as simplicial or cubical sets. A cubical representation has some advantages over a simplicial representation. One is that homotopy invariants like cup products and triad homotopy groups are more naturally described in terms of cubes [ 7 ] than simplices [ 41 , 42 ] . Another is that cubical sets, unlike simplicial sets [ 28 , Remark A.3] , admit approximation theorems with strong claims of naturality (eg. [ 26 , 28 ] ). Another is that some fundamental axioms in synthetic simplicial homotopy theory become theorems in synthetic cubical homotopy theory [ 6 ] . Yet another is that non-positive curvature is easier to characterize in cubical [ 22 ] than in simplicial (cf. [ 25 ] ) settings. Just as simplicial sets are presheaves over the category \Delta of non-empty finite ordinals [0],[1],[2],\ldots , cubical sets are presheaves over a category \square of the finite Boolean lattices (1) [0],[1],[1]^{2},[1]^{3},\ldots The problem is that the definition of \square -morphisms is not standard [ 8 , 21 ] . The \Delta -morphisms can be both explicitly characterized as the monotone functions between non-empty finite ordinals [0],[1],[2],\ldots and implicitly characterized by a presentation in terms of cofaces and codegeneracies [ 33 ] . The original variant \Delta_{1}^{*} of \square in the literature, the free monoidal category with unit [0] generated by the category \Delta_{1} of ordinals [0] and [1] and all functions between them, has a simple presentation in terms of cofaces and codegeneracies [ 21 , Theorem 4.2] . But unlike \Delta -morphisms, \Delta_{1}^{*} -morphisms do not admit a short and explicit characterization. And formal properties of the simplex category \Delta convenient for combinatorial homotopy are lacking in the original cube category \Delta_{1}^{*} . Recent years have seen an explosion of variants for \square proposed to address different shortcomings of \Delta_{1}^{*} (eg. [ 1 , 8 , 21 , 24 , 32 ] ). The current situation in cubical theory resembles an earlier situation in topology. Various technical lemmas in homotopy theory, at their natural levels of generality, called for various fussy point-set axioms on a topological space. For some variants of \square , finite Cartesian products of cubical sets model finite Cartesian products of the spaces they represent up to homotopy equivalence [ 32 ] . For some variants of \square , a cubical homotopy theory can be defined that models datatypes in certain higher order programming languages [ 1 , 10 ] . For some variants of \square , cubical sets can be used to model homotopy theories of spaces with extra geometry, such as a metric [ 19 , 28 ] , uniformity [ 28 ] , or directionality [ 19 , 27 ] . For some variants of \square , the natural tensor product on cubical sets is symmetric and thereby convenient for various applications [ 20 , 24 ] . Nowdays, a topological space in homotopy theory is typically redefined to just be a (weak) Hausdorff k-space so as to avoid having to make different point-set assumptions for different applications. In this same spirit, we propose a novel variant \boxplus for \square whose objects are the finite Boolean lattices ( 1 ) and whose morphisms are the interval-preserving monotone functions as a general-purpose choice for \square . This variant occupies a middle ground. On one hand, the \boxplus -morphisms include but are not generated by cofaces, codegeneracies, coordinate permutations, and codegeneracy-like maps called coconnections . On the other hand, the \boxplus -morphisms exclude all reversals and diagonals . The exclusion of reversals allows cubical sets to admit edge-orientations and thus admit directed topological realizations (e.g. [ 27 ] ). The exclusion of diagonals allows \square -objects to model topological cubes with their \ell_{p} metrics and hence allows cubical sets to admit \ell_{p} -realizations (e.g. [ 28 ] ), not just for p=\infty but for all 1\leqslant p\leqslant\infty ; \ell_{2} -realizations in particular generalize uniquely geodesic CAT(0) cubical complexes. The category \boxplus is the largest reasonable variant of \square excluding the reversals and diagonals: Proposition 3.21 . Consider a subcategory {\small\bigcirc} of \mathrm{\bf Set} such that the following all hold: (1) {\small\bigcirc} excludes the reversal [1]\rightarrow[1] and diagonal [1]\rightarrow[1]^{2} (2) {\small\bigcirc} contains \Delta_{1}^{*} as a wide subcategory (3) every {\small\bigcirc} -morphism factors into a composite of a surjective {\small\bigcirc} -morphism followed by an injective {\small\bigcirc} -morphism. Then {\small\bigcirc} is a subcategory of \boxplus . The choice {\small\bigcirc}=\boxplus satisfies all three conditions above. We can identify generators for \boxplus as a symmetric monoidal category. Unlike most variants of \square in the literature, \boxplus does not admit a finite monoidal presentation. Also unlike most variants of \square in the literature, \boxplus -morphisms do not generally preserve any obvious algebraic structure on finite Boolean lattices. Nonetheless, \boxplus -morphisms admit the following decomposition. Birkhoff Duality allows us to reinterpret the endomorphism operad of [1] in \boxplus as what we might call the distributive lattice operad , the operad of free and finite distributive lattices, where the operadic action substitutes generators from such a lattice with terms from other such lattices (c.f. [ 31 , Exercise 2.2.11] ). The symmetric monoidal category \boxplus is almost generated by this operad but for the presence of the terminal function [1]\rightarrow[0] , by the following splitting result. Theorem 3.23 . The category \boxplus is generated as a symmetric monoidal category by … (1) the unique function \sigma:[1]\twoheadrightarrow[0] (2) the two functions of the form [0]\rightarrow[1] (3) all monotone surjections [1]^{n}\twoheadrightarrow[1] For each \boxplus -morphism \phi:[1]^{m}\rightarrow[1]^{n} , there exists a unique n -tuple (m_{1},m_{2},\ldots,m_{n}) of natural numbers with m_{1}+m_{2}+\cdots+m_{n}\leqslant m and unique coset in \Sigma_{m}/(\Sigma_{m_{1}}\times\cdots\times\Sigma_{m_{n}}) such that for each representative g\in\Sigma_{m} of that coset, there exist unique monotone functions \phi_{i}:[1]^{m_{i}}\rightarrow[1] for 1\leqslant i\leqslant n , each non-constant in each of its coordinates, with \phi g=\phi_{1}\otimes\cdots\otimes\phi_{n}\otimes\sigma^{\otimes(m-m_{1}-\cdots-m_{n})} . This result has direct applications. The first is a computational implementation of cubical sets when \square=\boxplus . The theorem suggests a representation of \boxplus -morphisms in terms of natural numbers, formal expressions involving lattice operations, and permutations. The action of the distributive lattice operad and a solution to the word problem for free distributive lattices [ 39 ] can be used to give an algorithm for composition in \boxplus in terms of those