The orthogonal subcategory problem for a class of morphisms in a category asks whether the full subcategory of objects orthogonal to is a reflective subcategory.
This problem is related to the problem of localization. Suppose is indeed a reflective subcategory; let be the reflector (the left adjoint to the inclusion ).
If belongs to , then is an isomorphism in .
By definition of orthogonality, for every object in , induces an isomorphism of hom-sets
Since , this means that for all in the map
is an isomorphism, so that is a natural isomorphism between representables. By the Yoneda lemma, this means is an isomorphism.
This lemma can be sharpened. First, given a category , there is a Galois connection between classes of morphisms and classes of objects , induced by the predicate that is a bijection for all and all . The induced monad on the collection of morphism classes (partially ordered by inclusion) may be called “saturation”, denoted . An easy argument due to Gabriel-Zisman is that if is finitely cocomplete, then satisfies the axioms for a calculus of fractions, so that the localization can be constructed. An easy argument then establishes the following sharpened lemma:
To be connected with things like small object argument, Bousfield localization, and others…
Peter Freyd, Max Kelly, Categories of continuous functors I, Jour. Pure Appl. Alg. 2 (1972), 169-191.