2018-08-15: I'm still looking for an answer that does not rely on
$$\int{f\left[g(x)\right]g'(x)}\,\mathrm dx = F\left[g(x)\right]$$
I'm refreshing my old calculus skills, and the textbook (Kalkulus by Tom Lindstrøm, 3rd edition, a Norwegian book) asks me to find $\int{\arctan x}\,\mathrm dx$.
I start with integration by parts:
$$\int{\arctan x}\,\mathrm dx = \int{1\cdot\arctan x}\,\mathrm dx = x\cdot\arctan x - \int{x\cdot \frac{1}{x^2+1}}\,\mathrm dx$$
Next, it would be natural to use substitution, and I can do that to get the answer:
$$x\cdot\arctan x - \frac{\ln(x^2+1)}{2}+c$$
Which is correct, but it's clear from context that the book wants me to do it without using substituation. I feel like I've tried everything, looking at the book's own examples, but I must be missing some essential trick.
Here are some of the approaches I've tried:
1. Integration by parts, again
1.1. $u = x$ , $v' = \frac{1}{x^2+1}$
$$x\cdot\arctan x - \left(x\cdot\arctan x - \int{1\cdot\arctan x}\,\mathrm dx\right) = \int{\arctan x}\,\mathrm dx$$
Back where I started.
1.2. $u = \frac{1}{x^2+1}$ , $v' = x$
$$x\cdot\arctan x - \left(\frac{x^2}{2} \cdot \frac{1}{x^2+1} - \int{\frac{x^2}{2} \cdot \frac{-2x}{(x^2+1)^2}}\,\mathrm dx\right)$$ $$= x\cdot\arctan x - \frac{x^2}{2(x^2+1)} - \int{\frac{x^3}{(x^2+1)^2}}\,\mathrm dx$$
I've tried hacking away at this, but it doesn't look like it's getting any easier.
2. Adding and subtracting $x^2$ in the numerator
$$x\cdot\arctan x - \int{x\cdot \frac{(x^2+1)-x^2}{x^2+1}}\,\mathrm dx$$ $$= x\cdot\arctan x - \int{x\cdot \left(1 - \frac{x^2}{x^2+1}\right)}\,\mathrm dx$$ $$= x\cdot\arctan x - \int{x - x\cdot\frac{x^2}{x^2+1}}\,\mathrm dx$$ $$= x\cdot\arctan x - \frac{x^2}{2} + \int{x\cdot\frac{x^2}{x^2+1}}\,\mathrm dx$$
The book uses a similar trick (adding and subtracting 1 in the numerator) to solve $\int{x\cdot\arctan x}\,\mathrm dx$. In the process it finds that $\int{\frac{x^2}{x^2+1}}\,\mathrm dx = x - \arctan x$. Perhaps I need to use this result:
Partial integration with $u = x$ , $v´ = \frac{x^2}{x^2+1}$
$$= x\cdot\arctan x - \frac{x^2}{2} + \left(x\cdot(x-\arctan x) - \int{1\cdot (x-\arctan x)}\,\mathrm dx\right)$$ $$= x\cdot\arctan x - \frac{x^2}{2} + \left(x^2 - x\cdot\arctan x - \frac{x^2}{2} + \int{\arctan x}\,\mathrm dx\right)$$ $$= x\cdot\arctan x - \frac{x^2}{2} + x^2 - x\cdot\arctan x - \frac{x^2}{2} + \int{\arctan x}\,\mathrm dx$$ $$= \int{\arctan x}\,\mathrm dx$$
Back where I started, again. I've been trying for days, different tricks and manipulations. The book doesn't even list it as a particularly tricky question, so I'm feeling a bit dumb.
3. Searching for the answer
I've looked at some other questions, including
But they don't seem to answer my particular question.
I don't know if this is what you are looking for or not, but functions $f$ and their inverses $f^{-1}$ are pretty cool. If you know the derivative of $f$, you know the derivative of $f^{-1}$. And if you know the antiderivative of $f$, you know the antiderivative of $f^{-1}$ as well via this interesting formula on integration of inverse functions . I write it out here
$$ \int f^{-1}(y) \;dy = yf^{-1}(y) - F \;\circ \; f^{-1}(y) + C$$
So if you know that the antiderivative of $\tan{x}$ is $-\ln(\cos{x})$, then the result follows. This is a trick, so I'm not sure it's what you are looking for. Also, in order to find the antiderivative of $\tan{x}$, you use u-substitution which goes against our goal of integrating without u-substitution. Anyways, hopes this helps