Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

This question is in relation to the space $C(I)$, $I = [a, b]$. Define $P_n =\{ a_0+\dots+a_nx^n \mid a_i \in \mathbb{R}\}$ (any or all $a_i$ could be zero); clearly $P_n \subset C(I)$. The norm I'm using is $\lVert f\rVert_I = \sup_I |f(x)|$. Is $P_n$ closed under $\lVert\cdot\rVert_I$?

I am almost sure the answer is "yes", but I can't seem to prove it. My first instinct was to biject $P_n$ to $\mathbb{R}^{n+1}$, using coefficients as coordinates, and prove that sequences of degree-$n$ polynomials converge to degree-$n$ polynomials, but I can't prove that the metric $\lVert\cdot\rVert_I$ is equivalent to the standard metric on $\mathbb{R}^{n+1}$. Next, intuitively given a function $f \in C(I), \notin P_n$ I should be able to find some $\epsilon > 0$ such that there are no low-degree polynomials "nearby", then that function was not a limit point of $P_n$ so $P_n$ is closed. Again I have no idea how to prove this. What should I do?

share|cite|improve this question
A general result is that a finite-dimensional subspace of a normed space is closed. Maybe here a quite simple argument can be given. – Davide Giraudo Nov 12 '12 at 23:04
@ChrisEagle fixed. – Julien Clancy Nov 13 '12 at 0:27
up vote 8 down vote accepted

Generally, any finite dimensional subspace of a topological vector space is closed. However, in this case one can use properties of polynomials to show this directly.

Choose distinct points $x_0,...,x_n \in I$. Let $V$ be the $(n+1)\times (n+1)$ Vandermonde matrix $V=\pmatrix{1 &x_0 & \cdots & x_0^n\\ \vdots & \vdots & & \vdots \\1 &x_n & \cdots & x_n^n }$. A tedious calculation shows that $\det V \neq 0$ iff the $x_k$ are distinct. If $p$ is a polynomial of the form $p(x) = a_0+a_1x+...+a_nx^n$, then we see that (taking the coefficients as a element $a \in \mathbb{R}^{n+1}$) $\pmatrix{p(x_0) \\ \vdots \\ p(x_n)} = V a$, or equivalently $a = V^{-1} \pmatrix{p(x_0) \\ \vdots \\ p(x_n)}$.

Now define the operator $N: C(I) \to C(I)$ as follows: Let $f \in C(I)$, and define $a(f) = V^{-1} \pmatrix{f(x_0) \\ \vdots \\ f(x_n)}$, and let $N(f)(x) = f(x) - \sum_{k=0}^n [a(f)]_k x^k$. Noting that pointwise evaluation is continuous with respect to the $\sup$ norm, we see that $N$ is continuous, and furthermore, $N(f) = 0$ iff $f$ is a polynomial of degree $n$ or less. Hence $P_n = N^{-1}\{0\}$, the inverse image of a closed set, hence $P_n$ is closed.

share|cite|improve this answer
Very clever. Coincidentally we calculated the Vandermonde determinant recently in class. – Julien Clancy Nov 13 '12 at 0:02

It suffices to show that any sequence $(p_k)_{k\in\mathbb{N}}\subset P_n$ that converges with respect to $\|\cdot\|_I$ converges to a polynomial $p(x)\in P_n$. So, let $(p_k)$ be a sequence; for each $k$,


Claim: $p_k(x)$ converges (with respect to your norm) to the polynomial $p(x)=a_0+\ldots+a_nx^n$ where


I'll let you flush out the details.

share|cite|improve this answer
In the details you flush out, there is the fact that the sequences $\{a_i^{(k)}\}_i$ are convergent. – Davide Giraudo Nov 12 '12 at 23:05
@DavideGiraudo This is where I got stuck on my first attempt. Can you give me a hint as to how to do this? – Julien Clancy Nov 13 '12 at 0:00

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.