Let $\theta \in \ell^2(\mathbb{N})$ satisfy $$ \sum_{j=1}^n \theta_j^2 a_j^2 \leq Q $$ for some $Q > 0$ where $$ a_j = \begin{cases} j^\beta & \text{$j$ even}\\ (j-1)^\beta & \text{$j$ odd} \end{cases} $$ for some $\beta > 1/2$. Let $\{\varphi_j\}_{j=1}^\infty$ denote the trigonometric basis of $[0,1]$, i.e. $\varphi_1(x) = 1$, $\varphi_{2k}(x) = \sqrt{2} \cos(2\pi k x)$ and $\varphi_{2k+1}(x) = \sqrt{2} \sin(2\pi k x)$ for $k \in \mathbb{N}$. Show that $f = \sum_{j=1}^\infty \theta_j \varphi_j$ is continuous.

I know that $\varphi_j$ is Lipschitz with constant $\pi\lfloor j+1 \rfloor$ but I don't think that helps very much. Any ideas?


Note first that the hypothesis is just a silly way of writing $$\sum_{j=1}^\infty\theta_j^2j^{2\beta}<\infty.$$

Proof by Looking it Up

The hypothesis says precisely that $f$ is in the Sobolev space often denoted $H^\beta$; since $\beta>1/2$ the Sobolev Embedding Theorem shows that $f$ is continuous.

Actual Proof

$$\sum|\theta_j|=\sum|\theta_j|j^\beta j^{-\beta}\le\left(\sum\theta_j^2j^{2\beta}\right)^{1/2}\left(\sum j^{-2\beta}\right)^{1/2}<\infty,$$since $2\beta>1$. So the series converges uniformly, hence $f$ is continuous.

Proof of a Stronger Result

Say $x<y$ and $y-x$ is small. Choose an integer $N$ with $1/2\le N(y-x)\le 2$. Now $$f(y)-f(x)=\left(\sum_{j=1}^N+\sum_{j=N+1}^\infty\right)\theta_j(\phi_j(y)-\phi_j(x)):=I+II.$$Now the Lipschitz constant for $\phi_j$ shows that $$I\le c(y-x)\sum_{j=1}^N\theta_jj^\beta j^{1-\beta}\le c(y-x)\left(\sum_{j=1}^N\theta_j^2j^{2\beta}\right)^{1/2}\left(\sum_{j=1}^Nj^{2-2\beta}\right)^{1/2}\le c(y-x)N^{\frac32-\beta}\le c(y-x)^{\beta-\frac12}.$$To estimate $II$ you just use the fact that $|\phi_j|\le1$: $$II\le 2\sum_{j=N+1}^\infty\theta_j j^\beta j^{-\beta}\le c\left(\sum_{j=N+1}^\infty j^{-2\beta}\right)^{1/2}\le cN^{\frac12-\beta}\le c(y-x)^{\beta-\frac12}.$$Since $\beta-\frac12>0$ this shows that $f\in Lip_{\beta-\frac12}$.


It's worthwhile spending some time understanding this argument, because the outline "use a trivial uniform estimate for the high-order terms and the Lipschitz condition on the low-order terms" comes up a lot in this sort of thing.

| cite | improve this answer | |

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.