Is translation of a closed set by a compact set is closed

functional-analysislinear algebranormed-spaces

Let $E$ be a normed vector space over $\mathbb R$, $A, B \subset E$ disjoint nonempty convex subsets, $A$ closed and $B$ compact.

In the proof of Theorem 1.7 of Brezis's Functional Analysis, Sobolev Spaces and Partial Differential Equations, he says, without proof, $C := A – B = \{ x – y \mid x \in A, y \in B \}$ is also convex and closed.

I can show that $C$ is convex. Although I am sure that $C$ is closed by drawing the situation, I do not give any proof for it.

Since $C = \bigcup_{x \in A} (x – B) = \bigcup_{y \in B} (A – y)$, it suffices to pick up finitely many $x_i \in A$ (or finitely many $y_j \in B$) so that $C = \bigcup_{i} (x_i – B)$ (or $= \bigcup_{j} (A – y_j)$), so that $C$ is now the union of finitely many closed sets.

I think I have to construct an open cover of the compact set $B$ in order to pick up finitely many open sets, which in turn are used to find $x_i$ or $y_j$. But I do not come up with such an open cover, nor other methods to proving the statement.

How to show that $C$ is actually closed?

Best Answer

Probably the easiest way to prove this is using sequences. Suppose $(c_n)$ is a sequence in $C$ converging to some point $x\in E$. For each $n$, we can write $c_n=a_n-b_n$ where $a_n\in A$ and $b_n\in B$. Since $B$ is compact, we can pass to a subsequence and assume that $(b_n)$ converges to some $b\in B$. But now since $c_n\to x$ and $b_n\to b$, $a_n=b_n+c_n$ must converge to $b+x$. Since $A$ is closed, this means $b+x\in A$. Thus $x=(b+x)-b\in C$, as desired.

Related Question