I just started a self-study of functional Analysis from "Functional analysis, Sobolev spaces and partial differential equations" by Haim Brezis. I struggle to understand the underlying assumptions for the proof of Hahn-Banach theorem on the extension of function. In brief, the theorem is written as follows

Let $p : E \rightarrow \mathbb{R} $ be a Minkowski function.

Let $G \subset E$, $g : G \rightarrow \mathbb{R} $ such that $g \leq p $ is a linear function.

Then there exists a linear function $f$ such that $f=g, \forall x \in G $ and $f \leq p, \forall x \in E$.

The definition of a Minkowski is

$1. p(\lambda x) = \lambda p(x) \forall \lambda > 0$,

$2. p(x+y) \leq p(x)+p(y)$.

My question is why we need the first strict condition for this theorem. Why any convex function does not work with the condition $p(\lambda x) \leq \lambda p(x), \forall \lambda \geq 1$.


1 Answer 1


First, for the case that $p(0)=0$ holds, one can show that your conditions on $p$ imply that $p$ is a Minkowski function.

But (perhaps) more important is that Hahn-Banach only compares the function $p$ to a linear function ($f$ or $g$). For comparing a function $p$ with a linear function in this context, it suffices to only consider the ``worst'' values of $p$ along each ray $\{t x : t>0\}$.

To formalize this, suppose $p$ is convex. We define $q$ via $q(x):=\inf \{t p(x/t) : t>0 \}$. Then one can show that $q$ is a Minkowski function and that $$ g \leq p \iff g \leq q $$ holds for all linear functions $g$ which are defined on a linear subspace (I will not give the detailed proof here for these claims).

From this one can conclude that the Hahn-Banach theorem also holds for convex functions $p$, and not just Minkowski functions.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.