1

$$\int\ln(x+\sqrt{x^2-1})dx$$ To evaluate the integral we can apply integral by parts (using $u=\ln(x+\sqrt{x^2-1})$ and $dv=dx$). but my question is: is there other approach to evaluate the integral faster? I think there might be something because we can write it as:

$$\ln\left(x+\sqrt{x^2-1}\times\dfrac{x-\sqrt{x^2-1}}{x-\sqrt{x^2-1}}\right)=\ln\left(\cfrac1{x-\sqrt{x^2-1}}\right)=-\ln(x-\sqrt{x^2-1})$$ But I don't know how to use this.

Etemon
  • 6,437

3 Answers3

4

Use $x=\cosh t$

$$\int \sinh t \ln(\cosh t + \sinh t) dt = \int t \sinh t dt = t\cosh t - \sinh t$$

then substitute back.

Ninad Munshi
  • 34,407
  • +1..but IMHO I will put the constant :-) – Sebastiano Jan 21 '21 at 22:19
  • This is nice, recognizing the integrand as $;\text{arsinh}, x;$ ...but I don't think it is faster or simpler than integration by parts. – DonAntonio Jan 21 '21 at 22:20
  • 1
    @Sebastiano haha my bad. I've been in the habit of problem writing for an integration bee, no $+C$s there. – Ninad Munshi Jan 21 '21 at 22:20
  • @DonAntonio one thing I did not explain clearly was how I did the last step. I "guessed" the antiderivative needed a $t\cosh t$ term, and then asked what did I need to subtract to get the integrand I want after differentiating. It is a less systematic, but faster, equivalent method to by parts. – Ninad Munshi Jan 21 '21 at 22:22
  • 1
    @NinadMunshi Perhaps it is a matter of opinion, but I can't see how is this faster than by parts. Simply substituting back is already longer... – DonAntonio Jan 21 '21 at 22:24
4

Note that this is an inverse function and hence IBP is standard (see below). As soon as you recognize $$ \ln(x+\sqrt{x^2-1})=\mbox{arcosh}(x) $$ you can simplify the IBP computations, since no computation is really needed:

The following formula holds for any continuous bijective function: $$ \int f^{-1}(x) dx= x f^{-1}(x)- \int x (f^{-1})'(x) dx = x f^{-1}(x)- \int f( f^{-1}(x)) (f^{-1})'(x) dx \\ = x f^{-1}(x)- \int f(u) du $$

[Side note: there is a picture proof of this formula in the Book "Proofs without Words". The proof is basically just observing that $$ \int_{a}^x f(x) dx + \int_{f(a)}^{f(x)} f^{-1}(u) du $$ graphically the signed between the rectangle with corners $(0,0), (0,f(x)), (x,0), (x,f(x))$ and the rectangle with corners $(0,0), (0,f(a)), (a,0), (a,f(a))$ The picture proof can be seen in this answer. ]

Now set $$f(x)= \cosh(x) \\ f^{-1}(x)=\mbox{arcosh}(x)$$ and integrate $\int \cosh(u) du$.

If you accept the above formula as a "trick", then that's it.

Note Whenever when you know the integral of a bijection $f(x)$, the above formula gives you the integral of $f^{-1}$.

Since the formula is just IBP, it explains why $\int \ln(x) dx, \int \arcsin(x) ds$.... are always easy to calculate by parts.

N. S.
  • 132,525
3

Note $$\int\ln\left(x+\sqrt{x^2-1}\right)dx=\int \cosh^{-1}xdx \overset{IBP}=x \cosh^{-1}x -\sqrt{x^2-1}$$

Quanto
  • 97,352
  • I remember the $\mathrm{settcos, h}$ if I not have done a mistake. +1 – Sebastiano Jan 21 '21 at 22:34
  • 1
    @Sebastiano - what is settcos. – Quanto Jan 22 '21 at 14:04
  • I add the link in Italian language :-( sigh...but you can see the operator sett cos h and sett sin h. I think that the operator is defined only in Italian language: https://www.youmath.it/lezioni/analisi-matematica/integrali/612-integrali-con-sostituzioni-iperboliche.html – Sebastiano Jan 22 '21 at 21:04