The chain rule run backwards. Recognising when it applies is most of the skill; executing it is bookkeeping.
By the end of this chapter you can
Recognise an integrand that is a chain rule derivative
Carry out a u-substitution, including changing the limits
Choose u when several candidates present themselves
Say why a substitution sometimes fails
Every differentiation rule runs backwards into an integration technique. The
chain rule runs backwards into substitution, which is the one technique AB
needs and the foundation of every technique BC adds.
The chain rule says
dxdF(u(x))=F′(u(x))⋅u′(x)
Read right to left: anything of the form f(u)⋅u′ integrates to F(u).
The whole method is recognising that shape.
The shape to look for
An integrand that contains a function and its own derivative, up to a
constant factor.
∫2xcos(x2)dx
The inside is x2; its derivative is 2x; and 2x is sitting right there.
Substitute u=x2, so du=2xdx, and the integral becomes
∫cosudu=sinu=sin(x2).
∫2xcos(x2)dx=sin(x2)
dxdsin(x2)=2xcos(x2)
checked by the build
Both directions checked. The second line is the honest test of any integration:
differentiate your answer and see whether you get the integrand back. That
takes seconds and catches nearly every mistake, which makes skipping it
inexcusable on an exam.
Working the bookkeeping
∫xex2dx
The derivative of x2 is 2x and only x is present — off by a constant,
which is always fixable.
Let u=x2, so du=2xdx, hence xdx=21du:
∫xex2dx=21∫eudu=21eu=21ex2
∫xex2dx=2ex2
∫sin(x)cos(x)dx=2sin2(x)
∫(2x+1)ex2+xdx=ex2+x
checked by the build
Definite integrals: change the limits
With definite integrals there are two routes, and one is reliably safer.
Substitute the limits too. When u=x2 and x runs from 0 to 2, then
u runs from 0 to 4:
∫02xex2dx=21∫04eudu=21(e4−1)
Or convert back to x first, then use the original limits. Both are
correct; the first is less error-prone, because forgetting to convert back is a
standard way to evaluate the antiderivative at the wrong numbers.
∫eudu=eu
dxd2ex2=xex2
checked by the build
Choosing u
When more than one candidate presents itself, these heuristics hold up:
The inside of a composition. In cos(x2), take u=x2.
Whatever is under a root, inside a denominator, or in an exponent.
The thing whose derivative is also present, which is the real criterion —
the others are shortcuts to it.
A quick test before committing: compute du and see whether the rest of the
integrand is exactly du up to a constant. If a stray x survives, that choice
of u is wrong. Trying the other candidate takes fifteen seconds.
∫xlog(x)1dx=log(log(x))
∫tan(x)dx=−log(cos(x))
∫x2+1xdx=2log(x2+1)
checked by the build
Those three are all the same move. In the first, u=lnx gives
du=xdx, which is exactly what remains. In the second,
tanx=cosxsinx and u=cosx gives du=−sinxdx.
In the third, u=x2+1 gives du=2xdx, off by the usual constant.
Once the pattern is visible, these stop being separate problems.
{
"question": "Which substitution evaluates ∫ x·√(x² + 1) dx?",
"options": [
{
"text": "u = x² + 1, because du = 2x dx and the x is already present",
"correct": true,
"why": "The integrand contains the inside function and its derivative up to the constant 2. It becomes ½∫√u du = ⅓(x²+1)^{3/2}, and differentiating that returns x√(x²+1)."
},
{
"text": "u = x, since that is the variable of integration",
"correct": false,
"why": "That substitution changes nothing — du = dx and the integral is exactly as it was. A substitution has to simplify the composition to be worth making."
},
{
"text": "u = √(x² + 1), because it is the outermost function",
"correct": false,
"why": "Workable but needlessly painful: du involves the square root in the denominator and has to be untangled. Take the inside of the composition, not the outside."
},
{
"text": "No substitution works; this needs integration by parts",
"correct": false,
"why": "Parts is for products of unrelated functions. Here the two factors are related — one is essentially the derivative of what is inside the other — which is exactly the substitution signature."
}
]
}