Canonical: https://knowlpedia.com/real-analysis/differentiability-criterion/

# Differentiability criterion

Characterization of differentiability via a best linear approximation.

**Differentiability criterion:** Let \(U\subseteq\mathbb R^k\) be an [open set](https://knowlpedia.com/topology/open-set/), let \(f:U\to\mathbb R^m\), and fix \(a\in U\). The following are equivalent:

1. \(f\) is [differentiable at](https://knowlpedia.com/real-analysis/differentiable-map/) \(a\) (in the [Fréchet](https://knowlpedia.com/real-analysis/frechet-derivative/) sense).
2. There exists a [linear map](https://knowlpedia.com/linear-algebra/linear-map/) \(A:\mathbb R^k\to\mathbb R^m\) such that
   \[
   \lim_{h\to 0}\frac{\|f(a+h)-f(a)-Ah\|}{\|h\|}=0.
   \]

In this case, \(A\) is unique and is denoted \(Df(a)\).

## Remarks

This formulation says that differentiability is exactly the existence of a first-order linear approximation with an error term that is \(o(\|h\|)\). In the one-dimensional case \(k=m=1\), it is equivalent to the existence of the usual [derivative](https://knowlpedia.com/real-analysis/derivative/) at \(a\).
