8
$\begingroup$

Let $E=(0,1]$. For every $\alpha,\beta\in\mathbb{R}$, let $f(x)=x^{\alpha}\sin{x^\beta}$. For what values of $\alpha,\beta$ is $f\in L^1(E)$?

I think I know the answer: when $\alpha>-1$ or $\alpha+\beta>-1$. I don't know how to prove that $f$ is not integrable when neither of these inequalities hold.

First suppose that $\alpha>-1$. Since $\left|\sin{x^\beta}\right|\leq 1$, we have that $\left|x^{\alpha}\sin{x^\beta}\right|\leq x^{\alpha}$. By the monotonicity of Lebesgue integration, the monotone convergence theorem, and the fact that the Lebesgue and Riemann integral of continuous functions over closed intervals are equal, we can see: $$ \int_E\left|f(x)\right|\,d\lambda\leq\int_E\left|x^\alpha\right|\,d\lambda=\int_Ex^\alpha\,d\lambda=\lim_{n\to\infty}\int_{\frac{1}{n}}^1x^\alpha\,d\lambda=\lim_{n\to\infty}\int_{\frac{1}{n}}^1x^\alpha\,dx=\lim_{n\to\infty}\frac{1^{\alpha+1}-n^{-\alpha-1}}{\alpha+1}=\frac{1}{\alpha+1}$$

Now suppose that $\alpha+\beta>-1$. So $\underset{n\to\infty}{\lim}n^{-\alpha-\beta-1}=0$. Note that $\left|\sin{u}\right|\leq u$ for all $u\geq 0$. Thus, for $x>0$ we have that $\left|\sin{x^\beta}\right|\leq x^\beta$ and so $\left| f(x)\right|\leq x^{\alpha+\beta}$ for $x\in E$. Thus, we have: $$\int_E\left|f\right|\,d\lambda\leq\int_E\left|x^{\alpha+\beta}\right|\,d\lambda=\int_Ex^{\alpha+\beta}=\lim_{n\to\infty}\int_{\frac{1}{n}}^1x^{\alpha+\beta}\,d\lambda=\lim_{n\to\infty}\int_{\frac{1}{n}}^1x^{\alpha+\beta}\,dx=\lim_{n\to\infty}\frac{1^{\alpha+\beta+1}-n^{-\alpha-\beta-1}}{\alpha+\beta+1}=\frac{1}{\alpha+\beta+1}.$$

That's all I have so far. If anyone can help me show that $f$ is not integrable when $\alpha\leq -1$ and $\alpha+\beta\leq -1$, I would be grateful. Or perhaps I am wrong and these are not the right constraints.

$\endgroup$

1 Answer 1

8
$\begingroup$

Suppose that $\alpha\leq -1$ and $\alpha+\beta\leq -1$. We will first consider the case when $\beta=0$. Then $f(x)=x^{\alpha}\sin{x^{\beta}}=x^\alpha \sin{(1)}$ for all $x\in (0,1]$. Thus, $f\in L^1(E)$ if and only if $x^{\alpha}\in L^1(E)$. Our previous argument shows us that for $\alpha\neq -1$, $\int_Ex^\alpha\,d\lambda=\lim\limits_{\epsilon\to 0}\frac{1-\epsilon^{\alpha+1}}{\alpha+1}$. This limit exists if and only if $\alpha>-1$. From the Monotone Convergence Theorem (MCT) we have $$\int_Ex^{-1}\,d\lambda=\lim\limits_{\epsilon\to 0}\int_{\epsilon}^1x^{-1}\,dx=\lim_{\epsilon\to 0}\left(\ln{(1)}-\ln{(\epsilon)}\right)=\infty.$$ Hence, if $\alpha\leq -1$, $\alpha+\beta\leq -1$, and $\beta=0$, $f\notin L^1(E)$.

Now consider the case when $\beta\neq0$. Let $u=x^{\beta}$. So $du=\beta x^{\beta-1}dx$ which means that $dx=\frac{1}{\beta}x^{1-\beta}du=\frac{1}{\beta}u^{(1/\beta)-1}du$. Further, $x^\alpha=u^{\alpha/\beta}$. So we have that $x^\alpha\sin{x^{\beta}}dx=\frac{1}{\beta}u^{((\alpha+1)/\beta)-1}\sin{u}\,du$. Let $\gamma=\frac{\alpha+1}{\beta}-1$.

If $\beta>0$, we have that:

$$ \alpha+\beta\leq -1\Rightarrow\frac{\alpha}{\beta}+1\leq-\frac{1}{\beta}\Rightarrow\frac{\alpha+1}{\beta}+1\leq 0\Rightarrow\frac{\alpha+1}{\beta}-1=\gamma\leq-2. $$

Suppose that $\beta>0$. Note that $\lim\limits_{x\to 0} x^\beta =0$ and $\lim\limits_{x\to 1}x^\beta=1$. Thus, we do not need to change the limits of integration from the improper Riemann integral $\int_0^1 x^\alpha\left|\sin{x^\beta}\right|\,dx$ to the improper Riemann integral $\int_0^1\frac{1}{\beta}u^\gamma\left|\sin{u}\right|\,du$. Since $u$ is a dummy variable, we have that $\int_0^1\frac{1}{\beta}u^\gamma\left|\sin{u}\right|\,du=\int_0^1\frac{1}{\beta}x^\gamma\left|\sin{x}\right|\,dx$. Since $x^\gamma\sin{x}=\left|x^\gamma\sin{x}\right|$ on $(0,1]$, we have from the MCT that \begin{align*} \int_Ex^\alpha\left|\sin{x^\beta}\right|\,d\lambda &=\lim_{\epsilon\to 0}\int_\epsilon^1x^\alpha\left|\sin{x^{\beta}}\right|\,d\lambda=\lim_{\epsilon\to 0}\int_\epsilon^1x^\alpha\left|\sin{x^{\beta}}\right|\,dx=\lim_{\epsilon\to 0}\int_\epsilon^1\frac{1}{\beta}x^\gamma\left|\sin{x}\right|\,dx\\ &=\lim_{\epsilon\to 0}\int_\epsilon^1\frac{1}{\beta}x^\gamma\left|\sin{x}\right|\,d\lambda=\int_E\frac{1}{\beta}x^{\gamma}\left|\sin{x}\right|\,d\lambda=\int_E\frac{1}{\beta}x^\gamma\sin{x}\,d\lambda. \end{align*} Note that $x\sin{(1)}\leq\sin{x}=\left|\sin{x}\right|$ on $(0,1]$. Thus: \begin{align*} \int_Ex^\alpha\left|\sin{x^\beta}\right|\,d\lambda &=\int_0^1\frac{1}{\beta}x^\gamma\sin{x}\,dx\geq\frac{\sin{(1)}}{\beta}\cdot\int_0^1x^{\gamma+1}\,dx\\ &=\begin{cases}\dfrac{\sin{(1)}}{\beta}\cdot\left(\lim\limits_{y\to 0}\ln{1}-\ln{y}\right)&\text{if }\gamma+1=-1\\\dfrac{\sin{(1)}}{\beta}\cdot\left(\lim\limits_{y\to 0}\dfrac{1-y^{\gamma+2}}{\gamma+2}\right)&\text{if }\gamma+1<-1\end{cases}=\infty. \end{align*} Therefore, if $\beta>0$, $\alpha\leq -1$, and $\alpha+\beta\leq -1$, $f\notin L^1(E)$.

If $\beta<0$, we have the following: $$ \alpha\leq -1\Rightarrow \alpha+1\leq 0\Rightarrow\frac{\alpha+1}{\beta}\geq 0 \Rightarrow \frac{\alpha+1}{\beta}-1=\gamma\geq -1.$$ Unlike before, when we perform the $u$-substitution, we must flip the limits of integration. So $\int_0^1x^\alpha\sin{x^\beta}\,dx=\int_{\infty}^1\frac{1}{\beta}x^\gamma\sin{x}\,dx=-\frac{1}{\beta}\int_1^\infty x^\gamma\sin{x}\,dx$. A similar argument as before gives us that $\int_0^1x^\alpha\left|\sin{x^\beta}\right|\,d\lambda=-\int_1^{\infty}\frac{1}{\beta}x^\gamma\left|\sin{x}\right|\,d\lambda$ We can see that: \begin{align*} \int_1^\infty x^\gamma \left|\sin{x}\right|\,d\lambda &\geq\int_{\pi}^\infty x^\gamma\left|\sin{x}\right|\,d\lambda=\lim_{n\to\infty}\int_\pi^{n\pi+\pi}x^{\gamma}\left|\sin{x}\right|\,d\lambda=\lim_{n\to\infty}\int_\pi^{n\pi+\pi}x^{\gamma}\left|\sin{x}\right|\,dx\\ &=\lim_{n\to\infty}\sum_{k=1}^n\int_{k\pi}^{k\pi+\pi}x^{\gamma}\left|\sin{x}\right|\,dx=\lim_{n\to\infty}\sum_{k=1}^n\int_{0}^{\pi}(x+k\pi)^{\gamma}\left|\sin{x+k\pi}\right|\,dx\\ &=\lim_{n\to\infty}\sum_{k=1}^n\int_{0}^{\pi}(x+k\pi)^{\gamma}\left|\sin{x}\right|\,dx=\lim_{n\to\infty}\sum_{k=1}^n\int_{0}^{\pi}(x+k\pi)^{\gamma}\sin{x}\,dx\\ &\geq\begin{cases}\lim\limits_{n\to\infty}\sum\limits_{k=1}^n\int_{0}^{\pi}(\pi+k\pi)^{\gamma}\sin{x}\,dx&\text{if }\gamma\geq 0\\\lim\limits_{n\to\infty}\sum\limits_{k=1}^n\int_{0}^{\pi}(k\pi)^{\gamma}\sin{x}\,dx&\text{if }\gamma< 0\end{cases}\\ &=\begin{cases}\pi^\gamma\lim\limits_{n\to\infty}\sum\limits_{k=1}^n(k+1)^{\gamma}\int_{0}^{\pi}\sin{x}\,dx&\text{if }\gamma\geq 0\\\pi^\gamma\lim\limits_{n\to\infty}\sum\limits_{k=1}^nk^{\gamma}\int_{0}^{\pi}\sin{x}\,dx&\text{if }\gamma< 0\end{cases}\\ &=\begin{cases}2\pi^\gamma\lim\limits_{n\to\infty}\sum\limits_{k=1}^n(k+1)^{\gamma}&\text{if }\gamma\geq 0\\2\pi^\gamma\lim\limits_{n\to\infty}\sum\limits_{k=1}^nk^{\gamma}&\text{if }\gamma< 0\end{cases}=\infty. \end{align*} Thus, $-\frac{1}{\beta}x^\gamma\sin{x}\notin L^1([1,\infty))$. By our previous argument, we have that $x^\alpha\sin{x^\beta}\notin L^1(E)$.

Thus, $f\in L^1(E)$ if and only if either $\alpha>-1$ or $\alpha+\beta>-1$.

$\endgroup$

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.