4
$\begingroup$

I was working on these two questions for a while, I think they are kind of related but I could not figure out any of them:

1)Let $\mu$ be a regular Borel measure on $[0,1]$ such that $\int_{0}^{1}{x^{2n}d\mu}=0$ for all $n=0,1,2,...$. Prove that $\mu=0$.

2) Describe all regular signed measures $\mu$ on $[-1,1]$ such that $\int_{-1}^{1}{x^{2n+1}d\mu}=0$ for all $n=0,1,2,...$.

My attempt: For the fist one, I tried using this fact that as a consequence of Stone-Weierstrass Theorem, every continuous function can be uniformly approximated by an even polynomial and then I tried to use Riesz-Markov Theorem, then I think I lost there

Recalling Riesz-Markov Theorem: Let $X$ be a locally compact Hausdorff space and $T$ a positive linear functional on $C_c(X)$, then there is a unique Radon measure (regular Borel measure) $\mu$ on $B(X)$, the Borel $\sigma$-algebra associated with the topology on $X$, for which $T(f)=\int_{X}{fd\mu}$ for all $f\in{C_c(X)}$.

Any hint or comment would be highly appreciated.

$\endgroup$
  • $\begingroup$ You can extend $F_\mu:C_c([0,1]) \rightarrow \mathbb{R}$ to $F^*:C([0,1]) \rightarrow \mathbb{R}$ from Hahn-Banach theorem. From assumption $F^* \equiv 0$ on polynomial subspace (dense) of $C([0,1])$ , then this would imply $F^*\equiv 0$, and therefore $F_\mu\equiv 0$ . $\endgroup$ – Xiao Apr 26 '16 at 19:50
  • $\begingroup$ @Xiao: but if $F_{\mu}$ is identically zero, does it mean $\mu=0$? $\endgroup$ – Mary Apr 26 '16 at 21:10
  • $\begingroup$ Yes. Easier to think about it backward, the zero measure $\mu\equiv 0$ gives you the zero continuous linear functional from integration, by uniqueness, the zero continuous linear functional has to correspond to the zero measure. $\endgroup$ – Xiao Apr 27 '16 at 0:22
  • $\begingroup$ I see what you mean. Thank you! $\endgroup$ – Mary Apr 27 '16 at 14:35
  • $\begingroup$ @Mary Out of interest, where do the problems come from? $\endgroup$ – Milen Ivanov Apr 28 '16 at 19:51
1
$\begingroup$

Part 1 has been answered in the comments. The classification I got for part 2 is:

$\mu$ satisfies the given condition if and only if $\mu$ is "even", that is, for every Borel set $E \subset [-1;1]$ $\mu(E) = \mu(-E)$

Proof: Assume $\mu(E) = \mu(-E)$ for every Borel $E$. Let $f_n $ be simple functions which increase to $x^{2n+1}$, as defined on the interval $[0;1]$ Extend $f_n$ to odd functions on $[-1;1]$ As $f_n$ are simple and $\mu(E) = \mu(-E)$ it follows $\int f_n d\mu = 0$ so by the Dominated convergence theorem $\int _{-1}^1 x^{2n+1} d\mu = 0$, as required. (we can even write $f_n$ explicitly but there is no need)

Now, assume $\int _{-1}^1 x^{2n+1} d\mu = 0$ for all $n$ First, we will show that every odd continuous function on $[-1;1]$ can be approximated uniformly by an odd polynomial. Let $f$ be an odd function and $p_n$ be a sequence of polynomials, converging uniformly to $f$, as given by the Stone-Weierstrass theorem. Then $\frac{p_n(x) - p_n(-x)}{2} \to \frac{f(x) - f(-x)}{2} = f(x)$, since $f$ is odd. \frac{p_n(x) - p_n(-x)}{2} is also an odd polynomial, which we wanted to show.

Note that every odd polynomial is of type $a_k x^{2k+1} + a_{k-1}x^{2k-1} + ... + a_1x$, so such a polynomial integrates to 0 with respect to$\mu$ As a corollary of the uniform convergence we established, for every odd function $f$, we have $\int f d\mu = 0$

Finally, fix a Borel set $E \subset [0;1]$ We want to show $\mu (E) = \mu(-E)$ and evidently there is no loss of generality in assuming that $E \subset [0;1]$ only. Find a compact set $K$ and an open set $U$ such that $K \subset E \subset U$ and $\mu(U-K) < \epsilon$. By Urysohn's lemma there exists a continuous function $f_\epsilon : [0;1] \to [0;1]$ which has value $1$ on $K$ and $0$ on $U^c$. Thus, we have approximated $\mathbb{1}_E $ by a continuous function and $| \int_0^1 \mathbb{1}_E - f_\epsilon d\mu| < \epsilon $ Extending all functions to odd functions on $[-1;1]$ gives that $ |\int_{-1}^1 \mathbb{1}_E - \mathbb{1}_{-E} - f_\epsilon d\mu| < 2\epsilon $ Letting $\epsilon \to 0$ and recalling that $f_\epsilon$ integrates to $0$ gives

$\int \mathbb{1}_E - \mathbb{1}_{-E} d\mu = 0$ or, in other words $\mu(E) = \mu(-E)$ which we wanted to prove.

$\endgroup$

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.