[Paper Review] A Logic of Injectivity

Jiřı́ Adámek, M. Hébert+1 more|ArXiv.org|Sep 16, 2007

Homotopy and Cohomology in Algebraic Topology12 references3 citations

TL;DR

This paper introduces a formal deduction system—Injectivity Logic—for reasoning about injectivity consequences in categories, proving that three core rules (cancellation, pushout, and transfinite composition) are both sound and complete in 'strongly locally ranked' categories. The key contribution is a completeness theorem for both infinitary and finitary versions of the logic, with applications to algebra, topology, and homotopy theory.

ABSTRACT

Injectivity of objects with respect to a set $\ch$ of morphisms is an important concept of algebra, model theory and homotopy theory. Here we study the logic of injectivity consequences of $\ch$, by which we understand morphisms $h$ such that injectivity with respect to $\ch$ implies injectivity with respect to $h$. We formulate three simple deduction rules for the injectivity logic and for its finitary version where \mor s between finitely ranked objects are considered only, and prove that they are sound in all categories, and complete in all "reasonable" categories.

Motivation & Objective

• . The paper aims to formalize the logical structure of injectivity consequences in category theory.
• It addresses the problem of determining which morphisms h are logical consequences of a set H of morphisms, in the sense that H-injectivity implies h-injectivity.
• The objective is to develop a sound and complete deduction system for injectivity logic in a broad class of categories.
• It extends classical equational logic and injectivity in modules, Kan complexes, and other structures to a general categorical framework.
• The paper seeks to establish completeness for both finitary and infinitary versions of the logic, with applications to model theory and homotopy theory.

Proposed method

• . The paper formulates three core deduction rules: cancellation (h₂·h₁ ⇒ h₁), pushout (h ⇒ h′ in a pushout square), and transfinite composition (hᵢ ⇒ h for λ-composite h).
• It defines 'strongly locally ranked categories' as a class of categories where the full Injectivity Logic is complete.
• For the finitary version, the system is restricted to morphisms between objects of finite rank, with rules for identity and composition.
• The completeness of the finitary logic is proven via a model-theoretic argument, using a category extension to construct countermodels.
• The proof of completeness for the infinitary logic relies on properties of locally presentable categories and injectivity classes.
• The paper uses category-theoretic tools such as colimits, pure subobjects, and injective envelopes to analyze the structure of injectivity classes.

Experimental results

Research questions

• RQ1 . What are the minimal set of deduction rules that capture all injectivity consequences in a category?
• RQ2 How can injectivity logic be formalized in a way that is both sound and complete for a broad class of categories?
• RQ3 Can the completeness of injectivity logic be established for finitary morphisms (between finite-rank objects)?
• RQ4 To what extent does the logic remain complete when restricted to subcategories of finite-rank objects?
• RQ5 How does this logic relate to classical equational logic and homotopical constructions like Kan complexes?

Key findings

• . The three deduction rules—cancellation, pushout, and transfinite composition—are sound in all categories and complete in all 'strongly locally ranked' categories.
• . The finitary Injectivity Logic, restricted to morphisms between objects of finite rank, is complete, with a corresponding compactness theorem: every finitary injectivity consequence is derivable from a finite subset of assumptions.
• . The completeness of the full Injectivity Logic in strongly locally ranked categories (e.g., varieties of algebras, topological spaces, locally presentable categories) is established via model-theoretic techniques.
• . The paper demonstrates that completeness fails when restricting the logic to the full subcategory of finite-rank objects, even if the deduction system operates within that subcategory.
• . A counterexample in the category of graphs shows that H |=ω (0 →1) does not imply H ⊢ω (0 →1), proving that completeness does not extend to subcategories of finite-rank objects.
• . The paper shows that each of the four finitary deduction rules (identity, cancellation, composition, pushout) is essential—omitting any one leads to incompleteness.