In my lecture notes it is stated that $x^a \log{x}$ is integrable on $(0,1)$ for all $a>-1$, but a proof is not given. It doesn't seem to satisfy any of the comparison tests we have learned. What is the best way to go about showing this? Thanks in advance.
3 Answers
In THIS ANSWER, I showed using only the limit definition of the exponential function and Bernoulli's inequality that the logarithm function satisfies the inequalities
$$\frac{x-1}{x}\le \log(x)\le x-1\tag 1$$
For $x\in (0,1]$, $\log(x)<0$ and we see from $(1)$ that
$$|\log(x)|\le \frac{1}{x}-1$$
Then, using $\log(x^b)=b\log(x)$ along with $(1)$, we find that for $b>0$
$$\left| x^a\log(x)\right|\le \frac{x^{a-b}-x^a}{b}\tag 2$$
Note that $(2)$ holds for all $b>0$. In particular, $(1)$ holds for $0<b<1+a$.
Inasmuch as $\frac{x^{a-b}-x^a}{b}$ is integrable for $0<b<1+a$, $x^a\log(x)$ is integrable for $a>-1$ as was to be shown!
We can proceed further and evaluate the integral in closed form. Note that we can write
$$\begin{align} \int_0^1 \log(x)x^a\,dx&=\frac d{da}\int_0^1 x^a\,dx\\\\ &=\frac d{da}\left(\frac{1}{a+1}\right)\\\\ &=-\frac{1}{(a+1)^2} \end{align}$$
In fact, we can prove convergence and evaluate the integral in one fell swoop. Note that integrating by parts with $u=\log(x)$ and $v=\frac{x^{a+1}}{a+1}$ reveals
$$\begin{align} \lim_{\epsilon\to0}\int_0^1 \log(x)x^a\,dx&=\lim_{\epsilon\to0}\left(\left.\left(\log(x)\frac{x^{a+1}}{a+1}\right)\right|_\epsilon^1-\int_\epsilon^1 \frac{x^a}{a+1}\right)\\\\ &=-\frac{1}{(a+1)^2} \end{align}$$
as was to be shown!

- 179,405
Let $0<\varepsilon<1$ and consider the integral $$I(\varepsilon):=\int_{\varepsilon}^1x^a\log x\,dx$$ A change of variable $\log x=u$ yields $$I(\varepsilon)=\int_{\log\varepsilon}^0e^{u(a+1)}u\,du$$ Using integration by parts \begin{align} I(\varepsilon)& =\frac{1}{a+1}e^{u(a+1)}u\Big|^0_{\log\varepsilon}-\frac{1}{a+1}\int^0_{\log\varepsilon}e^{u(a+1)}\,du\\ &=\frac{1}{a+1}e^{u(a+1)}u\Big|^0_{\log\varepsilon}-\frac{1}{(a+1)^2}e^{u(a+1)}\Big|^0_{\log\varepsilon}\\&=-\frac{\varepsilon^{a+1}}{a+1}\log\varepsilon-\Big[\frac{1}{(a+1)^2}-\frac{\varepsilon^{a+1}}{(a+1)^2}\Big]\\&=-\frac{\varepsilon^{a+1}}{a+1}\log\varepsilon+\frac{\varepsilon^{a+1}}{(a+1)^2}-\frac{1}{(a+1)^2} \end{align} The last expression is well defined whenever $a\neq -1$. In particular when $a>-1$ we take the limit $$\lim_{\varepsilon\to 0}I(\varepsilon)=\lim_{\varepsilon\to 0}\Big[-\frac{\varepsilon^{a+1}}{a+1}\log\varepsilon+\frac{\varepsilon^{a+1}}{(a+1)^2}-\frac{1}{(a+1)^2}\Big]=-\frac{1}{(a+1)^2}<\infty$$ The second term vanishes, while the first term by L'Hospital's rule vanishes as well. Note also that for $x\in(0,1)$ it holds $|\log x|=-\log x$ so the above estimate suffices for the claim.

- 6,277
-
Why change variables? Simply let $u=\log(x)$ and $v=\frac{x^{a+1}}{a+1}$ and integrate by parts. – Mark Viola Feb 21 '18 at 14:01
-
Nothing in particular, just a matter of preference. I like to work with exponential than logarithms. – Arian Feb 21 '18 at 17:28
Use that $|\log(x) | \leq c_\epsilon x^{-\epsilon}$ for $x\in (0,1)$ and a suitable (depending on $a$) $\epsilon >0$.

- 36