I'm working through a proof and the setting is that I have a convex, balanced, open subset $U$ of a topological vector space $V$.

The claim that I can't verify is made briefly: the Minkowski functional, $p_{C} := inf\{t\geq 0 : x\in tC\}$ is continuous. This is supposed to follow from the fact that $\{x\in V:p_{C}(x) < 1\}$ is open. But this I cannot verify.

Can anyone point out the logic link I am missing?

  • 1
    $\begingroup$ Your definition of Minkovski's functional is incorrect. $\endgroup$
    – Norbert
    Jan 26 '12 at 21:38

I think your definition of the functional is incorrect (for once, taking an infimum over a vector space is not well-defined, and it is clearly not a fuctional, as the image of $\inf\lbrace x\in V\mid x\in tC\rbrace$ is not in $\mathbb R$). I assume you meant $p_C(x):=\inf\lbrace t\in \mathbb R\mid x\in tC\rbrace$.

Let $B=\lbrace x\in V\mid p_C(x)<1\rbrace$, we aim to show that in the case where $C$ is open we have that $B=C$ is open.

The fact that $B\subseteq C$ is immediate: if $x\in B$ then $p_C(x)<1$ implying that for any $t\ge 1,\: x\in tC$ (since $C$ is convex). We get that $x\in 1C=C$.

To the converse direction, let $x\in C$ be arbitrary. Since $C$ is open we have an open neighbourhood $x\in U\subseteq C$. Seeing that the action of multiplyication by a scaler is continuous, the pre-image of $U$ is open in $\mathbb C\times V$, and containes the element $\langle 1,x\rangle$. Thus, there exists some $\delta>0$ and an open ball $B(1,\delta)$ such that $B(1,\delta)\cdot x\subseteq C$.

In particular, this implies that $(1+\delta/2)x\in C$, and so $p_C(x)\le\frac{1}{1+\delta/2}<1$ and hence $x\in B$.

Once you have that, showing that $p_C$ is continuous is indeed immediate, I'll leave it to you for now, not to spoil all the fun :-)

  • 1
    $\begingroup$ I have a quick question in regards to the definition of the Minkowski functional. Is the above definition the same as the definition where you take infimum over the set $\lbrace t>0 \mid t^{-1}x \in C \rbrace$, where $C$ is just open and convex? $\endgroup$
    – Bartuc
    Mar 26 '16 at 10:26

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.