0
$\begingroup$

Define $$\mathcal{A}:=\Big\{p_n(x) = a_0+\sum^n_{k=1}(a_k\cos(kx)+b_k\sin(kx)): 1\leq k\leq n, \,{\rm and}\,\, a_0,a_k,b_k\in \mathbb{R}\Big\}$$ a) Prove that $\mathcal{A}$ is not dense in $C([0,2\pi])$ but $\mathcal{A}$ is dense in $$\mathcal{G}:=\{g\in C([0,2\pi]:g(0) = g(2\pi)\}$$ b) Prove that $\mathcal{A}$ is dense in $L^1([0,2\pi])$

I have recently studied the Stone-Weierstrass theorem and suspect that I can use it for this problem. I think that every $p_n(x)$ separates points because when sine does not separate points, cosine separates, and when cosine does not separate sine separates. So I believe $\mathcal{A}$ satisfies the conditions for Stone-Weierstrass theorem. But I have failed to prove that $\mathcal{A}$ not dense in $[0,2\pi]$ and moreover why is $\mathcal{A}$ dense in $\mathcal{G}$.

$\endgroup$
  • 2
    $\begingroup$ Take $f(x) = x$. Note that $f(0) = 0$, $f(2 \pi) = 2 \pi$, whereas any function $g \in {\cal A}$ satisfies $g(0)=g(2 \pi) = a_0$. $\endgroup$ – copper.hat Feb 9 '17 at 6:31
  • $\begingroup$ There is a little-known theorem of Fejer, proved by elementary Fourier series methods: For $f\in C[0,2\pi]$ with $f(0)=f(2\pi)$, let $F_n(x)$ be the partial sum of the Fourier series for$ f$ (up to the terms in $\cos nx$ & $\sin nx)$. Then $\;(1/n)\sum_{j=1}^n F_j(x)\;$ converges uniformly to $f(x)$ on $[0,2\pi]$ as $n\to \infty.$ $\endgroup$ – DanielWainfleet Feb 9 '17 at 14:36
2
$\begingroup$

Let $f \in C([0,2\pi])$ such that $f(0) \ne f(2 \pi)$ (for example $f(x)=x$.)

Suppose that $\mathcal{A}$ is dense in $C([0,2\pi])$. Then there is a sequence $(p_n)$ in $\mathcal{A}$ which converges uniformly to $f$ on $[0,2\pi]$.

Then we get from $p_n(0)=p_n( 2 \pi)$, with $n \to \infty$, that $f(0) = f(2 \pi)$, a contradiction.

$\endgroup$
1
$\begingroup$

The example in the comment and other answer shows why ${\cal A}$ is not dense in $C[0,2 \pi]$. The key missing feature is that the subalgebra does not separate points. There is no $g \in {\cal A}$ that separates $0,2 \pi$.

For ${\cal G}$, we can consider the functions $C(\mathbb{T})$ instead, and note that $\mathbb{T}$ is compact in the usual metric and ${\cal A}$ does separate points of $\mathbb{T}$ and contains the function $x \mapsto 1$. Hence Stone Weierstraß applies and hence ${\cal A}$ is dense in ${\cal G}$.

To prove that ${\cal A}$ is dense in $L^1[0,1]$, note that ${\cal A}$ is dense in indicator functions of intervals, hence of open sets and hence of measurable sets. Since simple functions are dense in $L^1[0,1]$, we obtain the desired result.

$\endgroup$

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.