Changes of variable that turn an integrand with surds into a rational function.
A rationalising substitution changes the variable so that an integrand with
surds or exponentials becomes rational, after which the partial-fraction
methods apply. A typical
choice is a root u that clears all rational powers of one base, or an
exponential substitution such as u=ex/2. Not every algebraic integrand
can be rationalised by a single substitution, so check the transformed
integrand before committing to the method.
For an integrand whose roots are rational powers of a common expression (or
whose exponential exponents are rationally related so that u=ecx makes
them rational powers), choose a new variable that makes the integrand
rational. If the result is improper, divide; then decompose into partial
fractions and integrate.
Replacing the radical is not enough. After rewriting both the integrand and
dx, every occurrence of the new variable must have an integer exponent;
only then do polynomial division and partial fractions apply. Forgetting to
rewrite dx in terms of du gives the wrong powers of u.
When several rational powers of x appear, set x=uN where N is a common
multiple of their denominators, so every power becomes an integer power of
u. Work on an interval where the original real powers and the chosen inverse
are defined; when N is even, x=uN covers only x≥0.
The trigonometric and hyperbolic table
clears a2±x2 and x2−a2 by a named identity. Use a
rationalising substitution for a surd that is not one of those three shapes.
For a fixed denominator after cancellation and factorisation, every proper rational function f has a unique partial-fraction expansion with terms of the form
(x−a)kA,(x2+bx+c)kBx+C,
where the quadratic x2+bx+c is irreducible over the real numbers (i.e. b2−4c<0).