Prove that $\log\left(1+\frac{1}{x^2}\right)$ is $L^1$ on $(0,\infty)$.

99 Views Asked by At

Let $f(x) = \log(1+\frac{1}{x^2})$. I am trying to show $f \in L^1(\mathbb{R}^+)$ where $\mathbb{R}^+ = (0,\infty)$.

What I have done so far

Since $f$ is strictly positive, I just have to show that $\int_{\mathbb{R}^+}f < \infty$. I already proved that this is true on $(\varepsilon,\infty)$ for any $\varepsilon>0$: since $\log(1+1/x^2) \leq 1/x^2$ for all $x > 0$, we have $$ \int_{(\varepsilon,\infty)}\log\left(1+\frac{1}{x^2}\right)dx \leq \int_{(\varepsilon,\infty)}\frac{1}{x^2}dx = \frac{1}{\varepsilon}, $$ so $\int_{(\varepsilon,\infty)}f < \infty$, where I used the Riemann integrability of $1/x^2$.

But here I run into trouble where I can't figure out how to bring the lower bound to zero, as $1/\varepsilon$ blows up as $\varepsilon \to 0$.

I attempted to use the fact that for any $\alpha >0$, $$ \lim_{x \to \infty}\frac{\log(1+x)}{x^{\alpha}} = 0, $$ i.e. there must exist an $M_{\alpha}>0$ such that $x>M_{\alpha}$ implies that $\log(1+x)\leq x^{\alpha}$. Letting $z = 1/x^2$, we see that for $x<1/\sqrt{M_{\alpha}}$ we have $$ \log\left(1+\frac{1}{x^2}\right)\leq \frac{1}{x^{2\alpha}}. $$ Then we set $\alpha = 1/3$ or something like that, so for $x < 1/\sqrt{M_{1/3}} = \varepsilon$ we have $f(x)\leq 1/x^{2/3}$. I'm thinking then to use some sort of Monotone Convergence Theorem argument, i.e. let $f_n(x) = f(x)\chi_{[1/n,\varepsilon]}$, making sure to start at large enough $n$ such that $1/n < \varepsilon$, then $$ \int_{(0,\varepsilon)}f_n(x)dx = \int_{(0,\varepsilon)}\log\left(1 + \frac{1}{x^2}\right)\chi_{[1/n,\varepsilon]}dx \leq \int_{1/n}^{\varepsilon}\frac{1}{x^{2/3}}dx = 3\left(\sqrt[3]{\varepsilon} - \frac{1}{\sqrt[3]{n}}\right), $$ where I again used Riemann integrability, which implies Lebesgue integrability. Sending $n \to \infty$ and using monotone convergence theorem we have that $\int_{(0,\varepsilon)}f(x)dx \leq 3 \sqrt[3]{\varepsilon}< \infty$.

My main question

I have shown then that $f$ is Lebesgue integrable on $(0,\varepsilon)$ for sufficiently small $\varepsilon$, and also that it is Lebesgue integrable on $(\varepsilon,\infty)$ for any $\varepsilon>0$. Is it possible to "glue" these results together to get that $f\in L^1(\mathbb{R}^+)$? This solution seems kind of sloppy and disconnected to me, but it's the closest I could come to a solution and I'd really appreciate any help someone could provide.

Please let me know if something is unclear, thank you!

2

There are 2 best solutions below

1
On BEST ANSWER

The Lebesgue integral is additive for all nonnegative measurable functions, even those not necessarily integrable, so for any $ ε > 0$,
$$ ∫_{ℝ_+} f \ \mathrm dλ= ∫_{ℝ_+} f ( 1_{(0,ε)} + 1_{[ε,∞)}) \ \text dλ = ∫_{(0,ε)} f \ \text dλ + ∫_{[ε,∞)} f \ \text dλ, $$ which justifies the gluing procedure.

1
On

By integration by parts we have

$$\int \log\left(1+\frac{1}{x^2}\right)dx = x\log\left(1+\frac{1}{x^2}\right)+\int\frac{2dx}{1+x^2} = x\log\left(1+\frac{1}{x^2}\right)+2\tan^{-1}x+C$$

which is a direct antiderivative. And since $\log\left(1+\frac{1}{x^2}\right) > 0$ for all $x\in(0,\infty)$ means $f = |f|$, we have that

$$||f||_{L^1(\Bbb{R}^+)} = \Biggr[x\log\left(1+\frac{1}{x^2}\right)+2\tan^{-1}x\Biggr]_0^\infty = \pi$$