the differential of a map

Suppose a smooth map F carries one manifold into another — think of a deformation that bends a flat sheet into a curved shape, or a function from a surface into space. Zoom in very close to a single point and any smooth deformation looks like a simple linear stretch-and-rotate: that local linear picture is the differential. It is the multivariable, coordinate-free version of the ordinary derivative, capturing how F distorts tiny vectors near a point.

Precisely, if F maps manifold M to manifold N and p is a point of M, the differential of F at p, written dF_p, is a linear map from the tangent space T_p M to the tangent space T_(F(p)) N. It acts on velocity vectors by the chain rule: take a curve through p with velocity v; push the curve forward through F to get a curve through F(p); the velocity of that image curve is dF_p(v). In coordinates dF_p is exactly the Jacobian matrix of partial derivatives of F — but the manifold viewpoint frees the idea from any particular coordinates, so the same object makes sense between any two manifolds.

The differential is the workhorse that turns nonlinear geometry into linear algebra one point at a time. Its rank tells you whether F is locally an embedding, a submersion, or folds things over; the inverse and implicit function theorems are stated through it. A frequent slip is to confuse F itself (which moves points) with dF_p (which moves tangent vectors and is always linear). And note the direction: vectors are pushed forward along F, whereas covectors and functions are pulled back — opposite arrows, a distinction that becomes central with the cotangent space.