I've just proved the fact that every linear function on a finite dimensional normed vector space is uniformly continuous. Because, let $T:U\to V$ be a linear function on $U$ with basis : $(u_1,u_2,\dots,u_n)$ and suppose that $\|\cdot\|_u$, $\|\cdot\|_v$ be the respective norms associated with $U$ and $V$. Setting $M= \max(\|Tu_1\|_v,\dots,\|Tu_n\|_v)$, we have $$\|Tu\|_v=\|a_1Tu_1 + a_2Tu_2 +\dots+ a_nTu_n\|_v \le |a_1|\cdot\|Tu_1\|_v+\dots+|a_n|\cdot\|Tu_n\|_v \le M\|u\|_1,$$ where $\|u\|_1=|a_1|+\dots+|a_n|$. Since all norms on $U$ are equivalent, we have $\|Tu\|_v\le C\|u\|_u$ for some $C>0$. Thus, $\|Tx-Ty\| \le C\|x-y\|$ for all $x$, $y$ in $U$, thereby satisfying Lipschitz' condition.
My questions are:

  1. Is it true that a linear function from one vector space to another is always continuous? (finite dimension not assumed)
  2. If so, or else, is a continuous linear map always uniformly continuous?
  • 1
    $\begingroup$ For the first question, you could have a look here: Discontinuous linear functional $\endgroup$ Commented May 19, 2012 at 6:00
  • $\begingroup$ BTW Your post was edited using TeX markup - for better readability. I think learning basics of TeX is not that difficult, see e.g. this meta thread. BTW I'd think that $\|\cdot\|_U$ would be slightly better notation than $\|\cdot\|_u$, but I tried to stick with your original version as much as possible, when editing. $\endgroup$ Commented May 19, 2012 at 6:04

2 Answers 2


You don't need the Axiom of Choice.

Take the space of polynomials in one variable, with the norm $||p|| = \sup_{t\in [0,1]}|p(t)|$. Then take the sequence of polynomials $p_n$ defined by $p_n(t) = (\frac{x}{2})^n$. Clearly $p_n \rightarrow 0$ in this norm, but the linear functional $\phi(p) = p(4)$ is unbounded (in fact, $||p_n|| = \frac{1}{2^n}, \phi(p_n) = 2^n)$.

  • $\begingroup$ You are right there. I was too quick. $\endgroup$
    – abatkai
    Commented May 19, 2012 at 10:01

(1): no. There is no direct construction, but assuming the Axiom of Choice, it is possible to construct such a map. You can easily construct non-continuous maps which are not everywhere defined, but only densely: differential operators are typical canonical examples.

(2): yes. A continuous map is continuous at zero, hence bounded, hence Lipschitz as can be seen from your proof.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .