If is a diagram and is a limit of it in , then we may naïvely say that this limit is preserved by a functor if is a limit of the composite diagram . However, it is not enough to state this at the level of objects; we also need to impose some coherence conditions, preserving the entire universal cone. Furthermore, we can use a trick involving the Yoneda embedding to get a meaningful condition even if has no limit in at all.
Let be a functor, a diagram, and be a cone over . Suppose is a limit for . The functor preserves this limit if is a limit of .
Here, is a whiskering.
Dually, preserves a colimit of if preserves it as a limit of .
For instance:
Let be the empty category, so that a limit of the unique functor is a terminal object . Then preserves this terminal object if and only if is a terminal object of .
Let be the discrete category , so that picks out two objects and of and the limit of is a product of and . Note that this product comes equipped with product projections and . Then preserves this product if and only if is a product of and and furthermore the product projections are and .
If preserves all limits/colimits for a specified diagram , we say that preserves limits of .
If preserves all limits or colimits of a given type (i.e. over a given category ), we simply say that preserves that sort of limit (e.g. preserves products, preserves equalizers, etc.).
A functor that preserves all small limits in that exist is called a continuous functor.
Usually this term is only used when has all small limits, i.e. is a complete category.
If there exists at least one limit for that is preserved by , then every limit for is preserved by .
Any other limit is uniquely isomorphic to , and preserves this isomorphism (as all functors do). This implies is a limit for .
Thus, for a specified diagram , preservation of limits either holds vacuously (if has no limit in ), or holds for all limit cones over .
Suppose is a limit for and is a limit for . Since is a cone over , there is a canonical comparision map with for each . Then preserves if and only if this comparision map is an isomorphism.
preserves if and only if is a limit for . Since is already a limit for , that is if and only if is isomorphic to in the category of cones over . And since is terminal in this category, that’s if and only if the comparison map, between the unique morphism in the cone category, is an isomorphism.
Let be a family of functors, and a diagram. Then every preserves limits of if and only if their product preserves limits of .
We exploit limits in product categories are computed componentwise. If every preserves a limit of , then the collectively determine a limit , so that preserves this. Conversely, if preserves the limit, then limits being computed componentwise implies each preserves the limit.
For the following properties, we have a diagram, and functors.
If preserves limits of and preserves limits of , we have that preserves limits of .
Take a limit for . Then is a limit for since preserves this limit, and then is a limit for since preserves this limit, as required.
If preserves limits of and reflects limits of , we have that preserves limits of .
Take a limit for . We want to show is a limit for . Since reflects limits of , it suffices to show is a limit for . But this follows since preserves limits of .
If preserves limits of and lifts limits for , we have that preserves limits of .
Take a limit for . We want to show is a limit for . Since lifts limits for , there is some limit cone over with in the category of cones over . Thus, in the category of cones over . But the former is a limit for since preserves limits of , implying the latter is too.
The propositions above taken together imply the following:
the Yoneda embedding preserves limits (see there)
Let be any infinite set, and consider the constant functor sending every object of to , and every morphism to the identity.
Then this functor preserves binary products objectwise, since . However, it does not preserve the product of and in the sense above, since the projection maps get sent to identities , which do not form a limit cone in .
This illustrates that it is not enough to merely show that is a limit of at the level of objects - we also need that the projections get mapped to the corresponding projections.
We can also view this through the language of representable functors. Preserving limits “objectwise” means represents the presheaf on , so that there is a natural isomorphism . However, for to preserve the limit we don’t just care that represents this functor, but how it does - in other words, we need the corresponding universal element in to be applied to the projections in . Given the natural isomorphism, then, we can check preserves the limit by following the identity and checking it maps to the correct element.
Analogously, an enriched functor between enriched categories may preserve weighted limits. Are there any tricky points that we should mention?
Sometimes we want to say that a functor preserves a limit that does not actually exist in . For instance, a finitely continuous functor is usually defined as one that preserves all finite limits. If is a finitely complete category, then this is fine; such a functor is called left exact. But what if does not have all finite limits?
If and are locally small, then we can use the Yoneda lemma to turn the question into one involving categories that do have the required limits (and in fact have all limits), the presheaf categories and . (For colimits, use and ; for -enriched categories, use and , which will work if is complete.)
The left Kan extension of the composite along the Yoneda embedding (which always exists) is a functor from to , which may be written as (alluding to the bimodule nature of profunctors). A diagram becomes a diagram in , where it has a limit. If preserves this limit, then we say that preserves the hypothetical limit of .
Since the Yoneda embedding preserves and reflects all limits, if has a limit in , then this condition is equivalent to the condition that preserve it in the ordinary sense, but in general it is stronger than requiring that preserve the limit only if it exists in .
Finishing the motivating example, a flat functor may be defined as one that preserves all finite limits, whether or not they exist.
Saunders MacLane, §V.4 of: Categories for the Working Mathematician, Graduate Texts in Mathematics 5 Springer (1971, second ed. 1997) [doi:10.1007/978-1-4757-4721-8]
Francis Borceux, §2.4 in: Handbook of Categorical Algebra Vol. 1: Basic Category Theory [doi:10.1017/CBO9780511525858]
Emily Riehl, §3.3 in: Category Theory in Context, Dover Publications (2017) [pdf, book website]
Last revised on February 10, 2026 at 12:13:48. See the history of this page for a list of all contributions to it.