The function is continuously differentiable when all its first partial derivatives exist on and are continuous there. Equivalently, has a Fréchet derivative at every point and the map is continuous. For a scalar-valued function,
On the open domain, continuous first partial derivatives ensure that this linear expression is the total derivative.
For differentiable maps with compatible domains, the multivariable chain rule is
In Jacobian matrix notation this is matrix multiplication. In particular, if both maps are continuously differentiable, the composition is continuously differentiable. Componentwise,
Let . This is a linear map, hence smooth; composing it with the given function proves that is continuously differentiable by the chain rule. Write for the first partial derivatives, evaluated at . Then
Adding these identities gives
Geometrically, , so is constant along every line in the diagonal direction; its derivative in that direction must vanish.
An embedded -dimensional submanifold is locally a coordinate -plane: for every , there are open neighbourhoods , and a diffeomorphism with
For the present setting, and its inverse are continuously differentiable; in the smooth convention both are smooth.
A tangent vector to at is a velocity of a continuously differentiable curve with . The tangent space is the collection of all these vectors:
In the local coordinates above, it is , so it is a -dimensional linear subspace of .
The regular level set theorem says that for a map , if has rank on , then is an -dimensional submanifold and
Here and the given rank condition makes the zero set a two-dimensional submanifold. It is nonempty because . Part (c) gives
Consequently
There is also a direct curve proof: , so the entire line through in this direction lies in . Its velocity at zero is exactly the required tangent vector.

Articles by others on the same topic (0)

There are currently no matching articles.