Let $V$ be a real vector with finite dimension, let's say $\operatorname{dim}_\mathbb{R}(V) = n$. The dual space of $V$ is defined by $V^* = \{ f : V \to \mathbb{R} \text{ linear} \}$.

I was told that one can define a topology on $V^*$, $\operatorname{GL}(V)$ and $\operatorname{GL}(V^*)$ but I have no idea how it is defined. Could someone tell me how exactly the open resp. closed set over this topology look like? Thanks a lot!


All topologies induced by a norm over finite dimensional spaces (of the same dimension) are equivalent to the standard topology on $\mathbb R^n$ so it is very much up to you how to choose your topology, noting that $\dim V^*=\dim V$.

If you'd like, a dual vector in $V^*$ can be identified with some vector $v=(v_1, \dots v_n) \in V$ via the dual basis construction (not canonical.)

One can then define the usual open ball topology here induced by the norm $\|v\|:=\sqrt{v_1^2+ \dots+v_n^2}$.

A slightly more sophisticated approach, more amenable to generalization is the idea of the "continuous dual," which consists of functions that are continuous $f:V \to k$ (for finite dimensional spaces, this is every linear function.) Here, the topology would be induced by the norm $$\sup_{v \in V} \frac{\|f(v)\|}{\|v\|}$$ where $\| \cdot\|$ is any norm on $V$.

All in all: under the usual identification of a dual vector space with the original space, closed sets look exactly how you would expect them to.

  • 1
    $\begingroup$ What you say in your first sentence is not correct. For example, the topology induced by the discrete metric on $\mathbb R$ is surely not equivalent to the standard topology. You mean all topologies under which vector addition and scalar multiplication is continuous. $\endgroup$ – amsmath Oct 29 '17 at 19:17
  • $\begingroup$ Excuse me. Yes, you are correct $\endgroup$ – Andres Mejia Oct 29 '17 at 19:23

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.