In mathematics, a submanifold of a manifold M is a subset S which itself has the structure of a manifold, and for which the inclusion map S → M satisfies certain properties. There are different types of submanifolds depending on exactly which properties are required. Different authors often have different definitions.
Contents 
In the following we assume all manifolds are differentiable manifolds of class C^{r} for a fixed r ≥ 1, and all morphisms are differentiable of class C^{r}.
An immersed submanifold of a manifold M is the image S of an immersion map i: N → M; in general this image will not be a submanifold as a subset, and an immersion map need not even be onetoone – it can have selfintersections – so the term is used loosely.
More narrowly, one can require that the map i: N → M be an inclusion (onetoone), in which we call it an injective immersion, and define an immersed submanifold to be the image subset S together with a topology and differential structure such that S is a manifold and the inclusion i is a diffeomorphism: this is just the topology on N, which in general will not agree with the subset topology: in general the subset S is not a submanifold of M, in the subset topology.
Given any injective immersion f : N → M the image of N in M can be uniquely given the structure of an immersed submanifold so that f : N → f(N) is a diffeomorphism. It follows that immersed submanifolds are precisely the images of injective immersions.
The submanifold topology on an immersed submanifold need not be the relative topology inherited from M. In general, it will be finer than the subspace topology (i.e. have more open sets).
Immersed submanifolds occur in the theory of Lie groups where Lie subgroups are naturally immersed submanifolds.
An embedded submanifold (also called a regular submanifold), is an immersed submanifold for which the inclusion map is a topological embedding. That is, the submanifold topology on S is the same as the subspace topology.
Given any embedding f : N → M of a manifold N in M the image f(N) naturally has the structure of an embedded submanifold. That is, embedded submanifolds are precisely the images of embeddings.
There is an intrinsic definition of an embedded submanifold which is often useful. Let M be an ndimensional manifold, and let k be an integer such that 0 ≤ k ≤ n. A kdimensional embedded submanifold of M is a subspace S ⊂ M such that for every point p ∈ S there exists a chart (U ⊂ M, φ : U → R^{n}) containing p such that φ(S ∩ U) is the intersection of a kdimensional plane with φ(U). The pairs (S ∩ U, φ_{S ∩ U}) form an atlas for the differential structure on S.
There are some other variations of submanifolds used in the literature. Sharpe (1997) defines a type of submanifold which lies somewhere between an embedded submanifold and an immersed submanifold.
Given any immersed submanifold S of M, the tangent space to a point p in S can naturally be thought of as a linear subspace of the tangent space to p in M. This follows from the fact that the inclusion map is an immersion and provides an injection
Suppose S is an immersed submanifold of M. If the inclusion map i : S → M is closed then S is actually an embedded submanifold of M. Conversely, if S is an embedded submanifold which is also a closed subset then the inclusion map is closed. The inclusion map i : S → M is closed if and only if it is a proper map (i.e. inverse images of compact sets are compact). If i is closed then S is called a closed embedded submanifold of M. Closed embedded submanifolds form the nicest class of submanifolds.
Manifolds are often defined as embedded submanifolds of Euclidean space R^{n}, so this forms a very important special case. By the Whitney embedding theorem any secondcountable smooth nmanifold can be smoothly embedded in R^{2n}.
