The elementary theory of the 2-category of small categories
Calum Hughes, Adrian Miranda
TL;DR
The paper develops ET2CSC, an elementary theory describing the $2$-category of small categories by axiomatising internal categories, functors, and natural transformations within a setting modeling ETCS. It extends Bourke's internal-category framework to capture the two-dimensional structure, introducing notions such as $2$-well-pointedness, full-subobject classifiers, and a categorified axiom of choice, and showing how generating families and orthogonal factorisations transfer to $ extbf{Cat}( extbf{E})$. The central result is a biequivalence between the 2-category of ETCS models and the 2-category of ET2CSC models, realized via the nerve/discretisation adjunctions and the correspondence between properties in $ extbf{E}$ and in $ extbf{Cat}( extbf{E})$. The work also characterises when $ extbf{Cat}( extbf{E})$ models ET2CSC in terms of ETCS-models for $ extbf{E}$ and establishes a Morita biequivalence between ETCS and ET2CSC, with an eye toward future extensions such as the axiom of replacement and higher categorical generalisations. This provides a purely 2-categorical, first-order framing of the foundational structure of categories and their morphisms, paralleling ETCS for sets.
Abstract
We give an elementary description of $2$-categories $\mathbf{Cat}\left(\mathcal{E}\right)$ of internal categories, functors and natural transformations, where $\mathcal{E}$ is a category modelling Lawvere's elementary theory of the category of sets (ETCS). This extends Bourke's characterisation of $2$-categories $\mathbf{Cat}\left(\mathcal{E}\right)$ where $\mathcal{E}$ has pullbacks to take account for the extra properties in ETCS, and Lawvere's characterisation of the (one dimensional) category of small categories to take account of the two-dimensional structure. Important two-dimensional concepts which we introduce include $2$-well-pointedness, full-subobject classifiers, and the categorified axiom of choice. Along the way, we show how generating families (resp. orthogonal factorisation systems) on $\mathcal{E}$ give rise to generating families (resp. orthogonal factorisation systems) on $\mathbf{Cat}\left(\mathcal{E}\right)_{1}$, results which we believe are of independent interest.
