4

I have tried out many trigonometric substitution like $x=\sin^{\frac{2}{5}}z$. But it did not work.

Arun
  • 816
Aman
  • 119
  • 1
    According to wolframalpha the solution is really ugly: http://www.wolframalpha.com/input/?i=%5Cint+x%2F(x%5E5-7)+dx – Cornman Oct 06 '17 at 03:07
  • https://www.wolframalpha.com/input/?i=integral+x%2F(x%5E5-7) something disgusting –  Oct 06 '17 at 03:07
  • Is this a problem from your text book? – imranfat Oct 06 '17 at 03:07
  • 8
    The way you would do it is decomposition into partial fractions. In this case, it would involve factorising $x^5 - 7$, the roots of which are the complex fifth roots of $7$. So, you'd get a factor of $x - 5^{1/7}$, as well as two nasty irreducible quadratics. Decomposing into partial fractions will yield some horrible constants, producing the aforementioned disgusting result. – Theo Bendit Oct 06 '17 at 03:19
  • Yep partial fraction decomposition is the way to go here – Joel Oct 06 '17 at 03:41

2 Answers2

4

Substitute $x=\sqrt[5]{7}u$, and you have $$ \begin{align} \int\frac{\sqrt[5]{7}u}{7u^5-7}\,\sqrt[5]{7}du &=\frac{\sqrt[5]{49}}{7}\int\frac{u}{u^5-1}\,du\\ &=\frac{\sqrt[5]{49}}{7}\int\left(\frac{A}{u-1}+\frac{Bu+C}{u^2-2hu+1}+\frac{Du+E}{u^2-2ku+1}\right)\,du\\ \end{align} $$

where $h=\cos(2\pi/5)=\frac{\sqrt{5}-1}{4}$ and $k=\cos(4\pi/5)=\frac{-\sqrt{5}-1}{4}$. The values of $A,B,C,D,E$ can be found using linear algebra, after recombining the three terms and comparing the coefficients of powers of $u$ in the numerator to that of $\frac{u}{u^5-1}$.

Continuing: $$ \begin{align} &=\frac{\sqrt[5]{49}}{7}\int\left(\frac{A}{u-1}+\frac{Bu+C}{(u-h)^2+1-h^2}+\frac{Du+E}{(u-k)^2+1-k^2}\right)\,du\\ &=\frac{\sqrt[5]{49}}{7}\int\left(\frac{A}{u-1}+\frac{B(u-h)+Bh+C}{(u-h)^2+1-h^2}+\frac{D(u-k)+Dk+E}{(u-k)^2+1-k^2}\right)\,du\\ &=\frac{\sqrt[5]{49}}{7}\left(A\ln\lvert u-1\rvert+\frac{B}{2}\ln\left((u-h)^2+1-h^2\right)+\frac{Bh+C}{\sqrt{1-h^2}}\arctan\left(\frac{u-h}{\sqrt{1-h^2}}\right)\right.\\ &\phantom{{}={}}\left.+\frac{D}{2}\ln\left((u-k)^2+1-k^2\right)+\frac{Dh+E}{\sqrt{1-k^2}}\arctan\left(\frac{u-k}{\sqrt{1-k^2}}\right)\right)+\text{constant}\\ \end{align} $$ Note $1-h^2=\sin^2(2\pi/5)$ and $1-k^2=\sin^2(4\pi/5)$, which may help some of you needed to carry this through to an explicit end.

The "constant" may take two different values on either side of $u=1$ where the discontinuity is.

The last step would be to back-substitute from $u$ to $x$. Then the "constant" might change values at $x=\sqrt[5]{7}$.

2'5 9'2
  • 54,717
  • OUCH. This is the kind of integration the Cambridge mathematics faculty used to put on the calculus section of the Tripos exam to screw with the egomaniac students.But it's important to be able to crank out a computation like this and it's easy to let your skills lapse after years of doing pure math proofs and letting computers do the work. – Mathemagician1234 Oct 06 '17 at 04:40
1

To make the problem more general, consider $$I_n=\int\frac x {x^n-A}\,dx=A^{\frac 2n-1}\int\frac t {t^n-1}\,dt=A^{\frac 2n-1}\int\frac t {\prod_{i=1}^n(t-r_i)}\,dt$$ where $r_i$ are the roots of unity. Using partial fraction decomposition, you will end with $$I_n=A^{\frac 2n-1}\sum _{i=1}^n\int \frac {\alpha_i}{t-r_i}\,dt=A^{\frac 2n-1}\sum _{i=1}^n{\alpha_i}\log({t-r_i})$$ For sure, since the ${\alpha_i},r_i$ terms will be complex numbers, you will need to recombine them by pairs if you want to get rid of all complex terms (coefficients and logarithms).

There is also a shorter notation $$J_n=\int\frac t {t^n-1}\,dt=-\frac{1}{2} t^2 \, _2F_1\left(1,\frac{2}{n};1+\frac{2}{n};t^n\right)$$ where appears the Gaussian or ordinary hypergeometric function.