Question about calculating $\sum_{n=0}^{\infty}\frac{(-1)^n}{3n+1}$

calculusintegrationsequences-and-series

Evaulate:
$$\sum_{n=0}^{\infty}\frac{(-1)^n}{3n+1}$$
Answer:

To evaluate $\sum_{n=0}^{\infty}\frac{(-1)^n}{3n+1}$ we use maclaurin series of $f(x)=\frac1{1+x^3}$:

$$\frac1{1+x^3}=\sum_{n=0}^{\infty}(-1)^nx^{3n}$$

and after taking $\int_{0}^{1}$ of both sides, the series will be equal to $\sum_{n=0}^{\infty}\frac{(-1)^n}{3n+1}$ which we were looking for and left hand side is:

$$\int_{0}^{1}\frac{1}{x^3+1}dx=[\frac13\ln(1+x)-\frac16\ln(x^2-x+1)+\frac{\sqrt3}{3}tan^{-1}\left(\frac{2x-1}{\sqrt3}\right)]\big|_0^1$$

$$=\frac13\ln2+\frac{\sqrt3\pi}{9} $$

My question:

In the answer above we evaluated $\int_{0}^{1}\frac1{x^3+1}dx$ because $\int_{0}^{1}[\sum_{n=0}^{\infty}(-1)^nx^{3n}]dx=\sum_{n=0}^{\infty}\frac{(-1)^n}{3n+1}$.

But if we consider:

$$g(x)=\sum_{n=0}^{\infty}(-1)^nx^{3n}$$

$$G(x)=\int\sum_{n=0}^{\infty}(-1)^nx^{3n}dx$$

Then $$G(1)=\sum_{n=0}^{\infty}\frac{(-1)^n}{3n+1}$$

And we can see $G(1)$ is also equal to series we are trying to calculate. Hence for $f(x)=\frac1{1+x^3}$ because $f(x)=g(x)$, If we consider $F(x)$ as antiderivative of $f(x)$. Then $\sum_{n=0}^{\infty}\frac{(-1)^n}{3n+1}=G(1)=F(1)=\frac13\ln2+\frac{\sqrt3\pi}{\color{red}{18}}$. Which is different from what we get in the answer. So what is the problem of this method?

Best Answer

The short answer is that the antiderivative you selected doesn't vanish at zero, whereas $\sum_{n=0}^\infty (-1)^n \frac{x^{3n+1}}{3n+1}$ does. So when you evaluated the antiderivative at the endpoints in the first computation, you picked up a contribution at zero that you have lost in your second computation.

To do the second computation correctly, you would need to include a "$+C$" in your antiderivative and then match it with the series by plugging in a value of $x$ where you know the value of the series. The obvious choice for this value of $x$ is $0$. But now effectively you have selected the antiderivative to be $\int_0^x g(y) dy$ in the first place, so why go through these extra steps rather than just evaluating $\int_0^1 g(y) dy$ immediately?

In general, various identities involving integrals in analysis do not hold in their simple form when you start talking about indefinite integrals. Accordingly, I usually suggest using definite integrals (possibly with variable limits) whenever it is at all convenient to do so.