In my studies of convexity I have recently come across the following:

Let $ V=\mathbb{R}^d ; d \geq 1 $ be a Euclidean space and $ S \neq \emptyset $ be a non empty set of the vector space $ V $, where we define its dual as usual $ S^* = \{y\in V | \forall x \in S ; \langle y,x \rangle \geq 0 \} $. Now I have already proved that $ S^* $ is always a closed convex cone, now I wish to prove the following: $ S^* = (\bar{S})^* = {(conv(S))}^*= {(cone(S))}^* $ where "conv" is the usual convex hull and "cone" is the minimal convex cone containing $S$ and $ \bar{S} $ is the closure of $S$.

I have proved that the dual set is always a closed convex cone, is there a way to prove these equalities I have written, I seem to be stuck on this. All help is appreciated.


Assume that $S$ has nonzero vector

Note that for $x\neq 0\in S$ $H_x:=\{ y|y\cdot x\geq 0\}$ is closed half space That is $ S^\ast$ is intersection of these closed half space Hence $S^\ast$ is closed convex In further closed half space is cone so that $S^\ast$ is cone

(1) $S^\ast = (\overline{S})^\ast$ : From this argument we have that $$ S\subset \overline{S}\Rightarrow S^\ast \supseteq (\overline{S})^\ast $$

If $x_n\in S$ and $x_n\rightarrow x$ let $$ v\in \bigcap_{n} H_{x_n} $$

so that $$ v\cdot x_n\geq 0 \Rightarrow v\cdot x\geq 0 $$

That is $v\in H_x$ so that $\bigcap_{n} H_{x_n} \subset H_x $

(2) $S^\ast =({\rm conv}\ S)^\ast$ : $x_i\in S$ then $v:=\sum_{i=1}^n c_ix_i\in {\rm conv}\ S$ where $c_i\geq 0,\ \sum_i c_i=1$ Assume that $$w\in \bigcap_i H_{x_i}$$ Then $$ w\cdot v =\sum_i c_ix_i\cdot w \geq 0\Rightarrow w\in H_v$$

(3) $({\rm conv}\ S)^\ast =({\rm cone}\ S)^\ast$ : For $v\in {\rm cone}\ S$, we have $$ w\in {\rm conv}\ S,\ tw=v,\ t>0 $$

Hence $$ H_v=H_w$$


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.