What the formula says
∫ f(g(x)) g′(x) dx = F(g(x)) + C
Substitution is for integrands that contain an inner function g(x) and, as a factor beside it, its
derivative g′(x). F is an antiderivative of the outer function f. The theorem says you can then
integrate the outer function as if the inner one were a single variable. It is a theorem, and it is
the chain rule run backwards: the derivative of F(g(x)) is f(g(x)) · g′(x).
It is usually written with a new variable. Let u = g(x). Then du = g′(x) dx, and the integral becomes ∫ f(u) du.
Worked examples
cos(x²) · 2x. The inner function is x², and its derivative 2x sits beside it. With u = x² and du = 2x dx:
- ∫ cos(x²) · 2x dx = ∫ cos u du = sin u + C = sin(x²) + C.
2x(x² + 1)⁵. With u = x² + 1 and du = 2x dx:
- ∫ 2x(x² + 1)⁵ dx = ∫ u⁵ du = u⁶/6 + C = (x² + 1)⁶/6 + C.
x/(x² + 1). With u = x² + 1, du = 2x dx, but the integrand only has x dx, which is du/2. A constant factor can be moved out:
- ∫ x/(x² + 1) dx = ½ ∫ 1/u du = ½ ln|u| + C = ½ ln(x² + 1) + C. The absolute value is not needed, because x² + 1 is always positive.
Common mistakes
- Changing the variable but keeping dx. ∫ cos u dx is not an integral you can work out. The whole expression, g′(x) dx included, is replaced by du.
- Keeping the old limits. In a definite integral the limits become g(a) and g(b) when you work in u. The alternative is to put x back first and then use the old limits.
- Substituting when the derivative is missing. A constant factor can be adjusted, as in the last example. A missing factor of x cannot: ∫ cos(x²) dx cannot be solved by substitution.
Common questions
How do you see what to set u to?
Look for something that sits inside another function, a bracket, an exponent or a denominator, and whose
derivative appears as a factor elsewhere in the integrand. That is u.
What happens to the limits under substitution?
They go through g. If the integral runs from a to b in x, it runs from g(a) to g(b) in u.
∫₀¹ 2x(x² + 1)⁵ dx becomes ∫₁² u⁵ du = (64 − 1)/6 = 10.5.
How is this different from integration by parts?
Substitution is the chain rule backwards and suits a function sitting inside another.
Integration by parts is the product rule backwards and suits two independent
factors multiplied together, such as x · eˣ.