Problem 200
Medium Numerical and Functional Series

Using the Weierstrass test, show that the series sum

Using the Weierstrass test, show that the series $\displaystyle\sum_{n=2}^{\infty}\ln\!\left(1+\frac{x^2}{n\ln^2n}\right)$ converges uniformly on $[-a,a]$ for any fixed $a>0$.

Explanation of Method

Using $\ln(1+t)\le t$ for $t\ge0$, each term is bounded by $\dfrac{x^2}{n\ln^2n}\le\dfrac{a^2}{n\ln^2n}$ on $[-a,a]$. The majorant series $\sum\dfrac{a^2}{n\ln^2n}$ converges by the Cauchy integral test, since $\displaystyle\int_2^\infty\frac{du}{u\ln^2u}$ converges (substitute $v=\ln u$). The M-test then finishes the job.

  1. Bound $\ln(1+t)\le t$ to get a term bound $\dfrac{a^2}{n\ln^2n}$ on $[-a,a]$.
  2. Test convergence of $\sum\dfrac1{n\ln^2n}$ via the integral test.
  3. Apply the Weierstrass M-test to conclude uniform convergence.

Step 1 · Uniform bound

For $t\ge0$, $\ln(1+t)\le t$ (concavity of $\ln$, tangent line at $t=0$). With $t=\dfrac{x^2}{n\ln^2n}\ge0$: $$ \ln\!\left(1+\frac{x^2}{n\ln^2n}\right) \le \frac{x^2}{n\ln^2n}. $$ On $[-a,a]$, $x^2\le a^2$, so each term is bounded by $M_n=\dfrac{a^2}{n\ln^2n}$, independent of $x$.

Step 2 · Convergence of the majorant

Consider $f(u)=\dfrac1{u\ln^2u}$ for $u\ge2$: positive, continuous, decreasing. By the Cauchy integral test, $\sum_{n=2}^{\infty}f(n)$ converges iff $\displaystyle\int_2^{\infty}\frac{du}{u\ln^2u}$ converges. Substituting $v=\ln u$, $dv=du/u$: $$ \int_2^{\infty}\frac{du}{u\ln^2u} = \int_{\ln2}^{\infty}\frac{dv}{v^2} = \left[-\frac1v\right]_{\ln2}^{\infty} = \frac1{\ln2} < \infty. $$ So $\sum_{n=2}^{\infty}1/(n\ln^2n)$ converges, and hence $\sum M_n=a^2\sum1/(n\ln^2n)<\infty$.

Step 3 · Weierstrass M-test

With $M_n=\dfrac{a^2}{n\ln^2n}$, both M-test conditions hold, so the series converges uniformly and absolutely on $[-a,a]$ for every fixed $a>0$. Since $a$ is arbitrary, this gives uniform convergence on every bounded interval, hence (local uniform convergence) a continuous sum on all of $\mathbb R$.


Summary Table

QuantityValue
Majorant$M_n=a^2/(n\ln^2n)$
$\sum1/(n\ln^2n)$converges (integral test, $=1/\ln2$ for the integral bound)

Additional Notes

Result

Converges uniformly and absolutely on $[-a,a]$ for every fixed $a>0$:

$$ \sum_{n=2}^{\infty}\ln\!\left(1+\frac{x^2}{n\ln^2n}\right) \text{ converges uniformly on } [-a,a],\ \forall a>0. $$