Canonical: https://knowlpedia.com/real-analysis/total-derivative-frechet-derivative/

# Total derivative (Fréchet derivative in ℝ^k)

The linear map Df(a) giving the best first-order approximation f(a+h)=f(a)+Df(a)h+o(‖h‖).

Let \(U\subseteq\mathbb{R}^k\) be [open](https://knowlpedia.com/topology/open-set/) and let \(f:U\to\mathbb{R}^m\). The function \(f\) is **(Fréchet) differentiable at \(a\in U\)** if there exists a [linear map](https://knowlpedia.com/linear-algebra/linear-map/) \(A:\mathbb{R}^k\to\mathbb{R}^m\) such that
\[
\lim_{h\to 0}\frac{\|f(a+h)-f(a)-A h\|_2}{\|h\|_2}=0.
\]

The map \(A\) is uniquely determined when it exists and is called the **(total) derivative** of \(f\) at \(a\), denoted \(Df(a)\).

This definition captures the best linear approximation of \(f\) near \(a\). In coordinates, \(Df(a)\) is represented by the [Jacobian matrix](https://knowlpedia.com/real-analysis/jacobian-matrix/) \(J_f(a)\) when \(f\) has continuous [partial derivatives](https://knowlpedia.com/real-analysis/partial-derivative/).

## Examples

- If \(f(x)=Ax+b\) is affine (with \(A\) an \(m\times k\) matrix), then \(Df(a)=A\) for all \(a\).
- If \(f:\mathbb{R}^2\to\mathbb{R}\), \(f(x,y)=x^2+y^2\), then \(Df(a)\) is the [linear map](https://knowlpedia.com/linear-algebra/linear-map/) \(h\mapsto 2\langle a,h\rangle\) (equivalently, gradient dot \(h\)).
- Existence of all [partial derivatives](https://knowlpedia.com/real-analysis/partial-derivative/) at \(a\) does not necessarily imply existence of \(Df(a)\) (Fréchet differentiability).
