2

While teaching about the general technique to integrate irrational functions$\left(\text{like} \int\frac{\mathrm dx}{L_1\sqrt L_2}, \int\frac{\mathrm dx}{L\sqrt Q}, \int\frac{\mathrm dx}{Q\sqrt L}\text{ and }\int\frac{\mathrm dx}{Q_1\sqrt Q_2}\right)$, my teacher taught that integrals of the form $\int\frac{\mathrm dx}{Q_1\sqrt Q_2}$ where $Q_1$ and $Q_2$ are quadratic functions in $x$ can be integrated by the substitution $x=\frac1{t}$. This works well in examples where the coefficient of $x$ in both the quadratics is zero as I can take the resulting quadratic in the square root as $u$, but when it's not, I am not able to proceed with this substitution as it seems to complicate the integral. Can someone please explain how to do so?

$\big{(}$I am aware of hyperbolic substitutions to eliminate $\sqrt Q_2$, but here I am specifically interested in the substitution $x=\frac1{t}$.$\big{)}$

For example,

$$\mathcal I=\int\frac{\mathrm dx}{(x^2+x+1)\sqrt{x^2+2x+3}}$$

Putting $x=\frac1{t}$, $$\mathcal I=-\text{sgn}x\int\frac{t\mathrm dt}{(t^2+t+1)\sqrt{3t^2+2t+1}}$$ $$=-\frac{\text{sgn}x}{\sqrt3}\int\frac{t\mathrm dt}{(t^2+t+1)\sqrt{\left(t+\frac13\right)^2+\frac29}}$$

Then, I tried to rewrite the numerator as $t=\frac12\left(2t+\frac23\right)-\frac13$ and take $t^2+\frac{2t}3+\frac13=u^2$ for the first integral:

$$\mathcal I=-\frac{\text{sgn}x}{\sqrt3}\int\frac{\mathrm du}{u^2+\frac13\sqrt{u^2-\frac29}+\frac59}-\frac{\text{sgn}x}{3\sqrt3}\int\frac{\mathrm dt}{(t^2+t+1)\sqrt{t^2+\frac{2t}3+\frac13}}$$

For the first integral, I thought of rationalizing the denominator, but it seems to complicate it even more. And ironically, the second integral is itself of the very form I wish to compute.

Integreek
  • 8,530
  • 2
    +1 Good question. Are you interested in an answer showing the integration of $\displaystyle \int \frac{1}{(x^2+ax+b)\sqrt{x^2+cx+e}} \mathrm dx$ with first substitution $x = \dfrac{1}{t}$? If you are, I can provide an insight. Let me know. – M. A. Nov 06 '24 at 11:10
  • @Bml Yes, I am interested in the general answer. – Integreek Nov 06 '24 at 11:11

1 Answers1

1

Not really a complete answer, but it may help:

When the coefficients of $x$ and $x^2$ in $Q_{1}$ and $Q_{2}$ have the same (non zero) ratio, you can complete the square, and find a suitable substitution.

For example,

$Q_{1}\sqrt{Q_{2}} \ := \ (2x^2+8x+9)\sqrt{x^2+4x+7} \ = \ \ (2(x+2)^2+1)\sqrt{(x+2)^2+3} \ $ and then the substitution $u=x+2 \ $ reduces the integrand to your required form.

If that doesn't work, we can try factoring $Q_{1}$ to get an integrand of the form $ \frac{1}{L_{1} L_{2}\sqrt{Q_2}}$ Where $L_{1}$ and $L_{2}$ are linear functions of $x$. From here, we proceed with partial fractions to get an integrand of the form $ \frac{a}{L_{1} \sqrt{Q_{2}}} + \frac{b}{L_{2} \sqrt{Q_{2}}} $ which can be solved using the substitution $L=\frac{1}{t}$.

In your particular example, this may not work either since $Q_{1}$ has complex roots, so you'll probably have to resort to trigonometric/hyperbolic substitutions.

Sorry if this wasn't particularly helpful.

  • So when $Q_1$ has complex roots, is it not possible to proceed using $x=\frac1{t}$? And don’t worry, your answer is quite useful :) – Integreek Nov 06 '24 at 10:25
  • well, you'll can't proceed if you wish to avoid complex numbers. you may try writing the $I$ as $ \frac{1}{\sqrt{3}i} \int \frac{dx}{(x + \frac{1 - \sqrt{3}i}{2})(\sqrt{x^2+2x+3})} - \frac{1}{\sqrt{3}i} \int \frac{dx}{(x + \frac{1 + \sqrt{3}i}{2})(\sqrt{x^2+2x+3})}$ and proceed with $x=\frac{1}{t}$ if you really want to use this substitution – Edmaand Vlad Phiyik Nov 06 '24 at 10:32
  • after the substitution, you'll get two integrals of the for $\frac{1}{\sqrt{Q}}$ but I don't know if we can solve those, given the complex coefficients. – Edmaand Vlad Phiyik Nov 06 '24 at 10:47
  • Well, in the general case, wouldn't substituting $x=\frac1{t}$ in the form $\int\frac{\mathrm dx}{L\sqrt Q}$ again yield an integral of the same form? Rather, $L=\frac1{t}$ would be efficient, but I want to ask if it is possible to proceed with $x=\frac1{t}$ by manipulating the integrand in any way? – Integreek Nov 06 '24 at 10:50
  • right, I'm sorry I means $L=\frac{1}{t}$ Edited answer.

    regarding $x=\frac{1}{t}$ I do not think it's possible, but I'm not sure of it.

    – Edmaand Vlad Phiyik Nov 06 '24 at 10:51
  • If we substitute $L=\frac1{t}$ in the case of complex numbers, it would be quite tedious to find square roots of complex numbers to complete the square and then at the end, use the logarithmic definitions for inverse trig and hyperbolic functions, is there any shorter way? – Integreek Nov 06 '24 at 10:53
  • none that I know of. which is why one may prefer hyperbolic substitution in the first place. but I am way in over my head here, I'm analyzing as I go. Perhaps someone more learned can provide an answer. sorry. – Edmaand Vlad Phiyik Nov 06 '24 at 10:57