In the previous section we explored the fundamental theorem of calculus and the link it provides between definite integrals and antiderivatives. Indeed, integrals with simple integrands are usually evaluated via this link. In this section we start to explore methods for integrating more complicated integrals. We have already seen — via Theorem 1.2.1 — that integrals interact very nicely with addition, subtraction and multiplication by constants:
for \(A,B\) constants. By combining this with the list of indefinite integrals in Theorem 1.3.17, we can compute integrals of linear combinations of simple functions. For example
\begin{align*}
\int_1^4\left(e^x - 2\sin x + 3x^2 \right)\dee{x} &= \int_1^4e^x\dee{x} -2\int_1^4 \sin x \dee{x} +3 \int_1^4x^2 \dee{x}\\
&= \left(e^x + (-2)\cdot(-\cos x) + 3\frac{x^3}{3} \right)\bigg|_1^4 &\text{and so on}
\end{align*}
Of course there are a great many functions that can be approached in this way, however there are some very simple examples that cannot.
In each case the integrands are not linear combinations of simpler functions; in order to compute them we need to understand how integrals (and antiderivatives) interact with compositions, products and quotients. We reached a very similar point in our differential calculus course where we understood the linearity of the derivative,
but had not yet seen the chain, product and quotient rules 1
If your memory of these rules is a little hazy then you really should go back and revise them before proceeding. You will definitely need a good grasp of the chain rule for what follows in this section.
. While we will develop tools to find the second and third integrals in later sections, we should really start with how to integrate compositions of functions.
It is important to state up front, that in general one cannot write down the integral of the composition of two functions — even if those functions are simple. This is not because the integral does not exist. Rather it is because the integral cannot be written down as a finite combination of the standard functions we know. A very good example of this, which we encountered in Example 1.3.4, is the composition of \(e^x\) and \(-x^2\text{.}\) Even though we know
With that caveat out of the way, we can introduce the substitution rule. The substitution rule is obtained by antidifferentiating the chain rule. In some sense it is the chain rule in reverse. For completeness, let us restate the chain rule:
In order to apply the substitution rule successfully we will have to write the integrand in the form \(f(u(x))\cdot u'(x)\text{.}\) To do this we need to make a good choice of the function \(u(x)\text{;}\) after that it is not hard to then find \(f(u)\) and \(u'(x)\text{.}\) Unfortunately there is no one strategy for choosing \(u(x)\text{.}\) This can make applying the substitution rule more art than science 2
Thankfully this does become easier with experience and we recommend that the reader read some examples and then practice a LOT.
. Here we suggest two possible strategies for picking \(u(x)\text{:}\)
Factor the integrand and choose one of the factors to be \(u'(x)\text{.}\) For this to work, you must be able to easily find the antiderivative of the chosen factor. The antiderivative will be \(u(x)\text{.}\)
Look for a factor in the integrand that is a function with an argument that is more complicated than just “\(x\)”. That factor will play the role of \(f\big(u(x)\big)\) Choose \(u(x)\) to be the complicated argument.
We want to massage this into the form of the integrand in the substitution rule — namely \(f(u(x))\cdot u'(x)\text{.}\) Our integrand can be written as the product of the two factors
and we start by determining (or guessing) which factor plays the role of \(u'(x)\text{.}\) We can choose \(u'(x)=9\sin^8(x)\) or \(u'(x)=\cos(x)\text{.}\)
If we choose \(u'(x)=9\sin^8(x)\text{,}\) then antidifferentiating this to find \(u(x)\) is really not very easy. So it is perhaps better to investigate the other choice before proceeding further with this one.
If we choose \(u'(x)=\cos(x)\text{,}\) then we know (Theorem 1.3.17) that \(u(x)=\sin(x)\text{.}\) This also works nicely because it makes the other factor simplify quite a bit \(9\sin^8(x) = 9u^8\text{.}\) This looks like the right way to go.
Note that \(9\sin^8(x) \cos(x)\) is a function of \(x\text{.}\) So our answer, which is the indefinite integral of \(9\sin^8(x) \cos(x)\text{,}\) must also be a function of \(x\text{.}\) This is why we have substituted \(u=\sin(x)\) in the last step of our solution — it makes our solution a function of \(x\text{.}\)
The second factor, \(\cos\big(x^3\big)\) is a function, namely \(\cos\text{,}\) with a complicated argument, namely \(x^3\text{.}\) So we try \(u(x)= x^3\text{.}\) Then \(u'(x) = 3x^2\text{,}\) which is the other factor in the integrand. So the integral becomes
If we choose \(u'(x)=e^x\) then \(u(x)=e^x\) and the other factor becomes \(\sin(u)\) — this looks promising. Notice that if we applied the other strategy of looking for a complicated argument then we would arrive at the same choice.
But what happened to the limits of integration? We can incorporate them now. We have just shown that the indefinite integral is \(-\cos(e^x)\text{,}\) so by the fundamental theorem of calculus
Theorem 1.4.2, the substitution rule for indefinite integrals, tells us that if \(F(u)\) is any antiderivative for \(f(u)\text{,}\) then \(F\big(u(x)\big)\) is an antiderivative for \(f\big(u(x)\big) u'(x)\text{.}\) So the fundamental theorem of calculus gives us
\begin{align*}
\int_a^b f\big(u(x)\big) u'(x)\,\dee{x} &= F\big(u(x)\big)\bigg|_{x=a}^{x=b}\\
&= F\big(u(b)\big) - F\big(u(a)\big)\\
&= \int_{u(a)}^{u(b)} f(u)\,\dee{u} & \text{since $F(u)$ is an antiderivative for $f(u)$}
\end{align*}
Notice that to get from the integral on the left hand side to the integral on the right hand side you
substitute 3
A good way to remember this last step is that we replace \(\diff{u}{x} \dee{x}\) by just \(\dee{u}\) — which looks like we cancelled out the \(\dee{x}\) terms: \(\frac{\dee{u}}{\cancel{\dee{x}}}\cancel{\dee{x}} = \dee{u}\text{.}\) While using “cancel the \(\dee{x}\)” is a good mnemonic (memory aid), you should not think of the derivative \(\diff{u}{x}\) as a fraction — you are not dividing \(\dee{u}\) by \(\dee{x}\text{.}\)
\(u(x)\rightarrow u\) and \(u'(x)\dee{x}\rightarrow \dee{u}\text{,}\)
set the lower limit for the \(u\) integral to the value of \(u\) (namely \(u(a)\)) that corresponds to the lower limit of the \(x\) integral (namely \(x=a\)), and
set the upper limit for the \(u\) integral to the value of \(u\) (namely \(u(b)\)) that corresponds to the upper limit of the \(x\) integral (namely \(x=b\)).
Also note that we now have two ways to evaluate definite integrals of the form \(\int_a^b f\big(u(x)\big)u'(x)\,\dee{x}\text{.}\)
We can find the indefinite integral \(\int f\big(u(x)\big) u'(x)\,\dee{x}\text{,}\) using Theorem 1.4.2, and then evaluate the result between \(x=a\) and \(x=b\text{.}\) This is what was done in Example 1.4.5.
Or we can apply Theorem 1.4.2. This entails finding the indefinite integral \(\int f(u)\,\dee{u}\) and evaluating the result between \(u=u(a)\) and \(u=u(b)\text{.}\) This is what we will do in the following example.
In this example the integrand is already neatly factored into two pieces. While we could deploy either of our two strategies, it is perhaps easier in this case to choose \(u(x)\) by looking for a complicated argument.
The second factor of the integrand is \(\sin\big(x^3+1\big)\text{,}\) which is the function \(\sin\) evaluated at \(x^3+1\text{.}\) So set \(u(x)=x^3+1\text{,}\) giving \(u'(x)=3x^2\) and \(f(u)=\sin(u)\)
The first factor of the integrand is \(x^2\) which is not quite \(u'(x)\text{,}\) however we can easily massage the integrand into the required form by multiplying and dividing by \(3\text{:}\)
We have already observed that one factor of the integrand is \(\sin\big(x^3+1\big)\text{,}\) which is \(\sin\) evaluated at \(x^3+1\text{.}\) Thus we try setting \(u(x)=x^3+1\text{.}\)
This makes \(u'(x)=3x^2\text{,}\) and we replace \(u(x)=x^3+1\rightarrow u\) and \(\dee{x}\rightarrow\frac{1}{u'(x)}\dee{u} = \frac{1}{3x^2}\dee{u}\text{:}\)
Starting with the first integral, we see that it is not too hard to spot the complicated argument. If we set \(u(x)=2x+1\) then the integrand is just \(\sqrt{u}\text{.}\)
We can evaluate the second integral in much the same way. Set \(u(x)=3x-2\) and replace \(\dee{x}\) by \(\frac{1}{u'(x)}\dee{u} = \frac{1}{3}\dee{u}\text{:}\)
This last example illustrates that substitution can be used to easily deal with arguments of the form \(ax+b\) (with \(a,b\) constants and \(a\ne 0\)), i.e. that are linear functions of \(x\text{,}\) and suggests the following theorem.
In this example we should set \(u=3x\text{,}\) and substitute \(\dee{x} \rightarrow \frac{1}{u'(x)}\dee{u} = \frac{1}{3}\dee{u}.\) When we do this we also have to convert the limits of the integral: \(u(0)=0\) and \(u(\pi/2)=3\pi/2\text{.}\) This gives
We can also do this example more directly using the above theorem. Since \(\sin(x)\) is an antiderivative of \(\cos(x)\text{,}\) Theorem 1.4.10 tells us that \(\frac{\sin(3x)}{3}\) is an antiderivative of \(\cos(3x)\text{.}\) Hence
The rest of this section is just more examples of the substitution rule. We recommend that you after reading these that you practice many examples by yourself under exam conditions.
Note that the lower limit of the \(u\)-integral, namely \(1\text{,}\) is larger than the upper limit, which is \(0\text{.}\) There is absolutely nothing wrong with that. We can simply evaluate the \(u\)-integral in the normal way. Since \(-\cos(u)\) is an antiderivative of \(\sin(u)\text{:}\)
We could do this one using Theorem 1.4.10, but its not too hard to do without. We can think of the integrand as the function “one over a cube” with the argument \(2x+1\text{.}\) So it makes sense to substitute \(u=2x+1\text{.}\) That is
The integrand can be rewritten as \(x \cdot \frac{1}{1+x^2}\text{.}\) This second factor suggests that we should try setting \(u=1+x^2\) — and so we interpret the second factor as the function “one over” evaluated at argument \(1+x^2\text{.}\)
Remember that we are using the notation “\(\log\)” for the natural logarithm, i.e. the logarithm with base \(e\text{.}\) You might also see it written as “\(\ln x\)”, or with the base made explicit as “\(\log_e x\)”.
The integrand is the product of \(\cos\) evaluated at the argument \(x^4+2\) times \(x^3\text{,}\) which aside from a factor of \(4\text{,}\) is the derivative of the argument \(x^4+2\text{.}\)
Before proceeding further, we should note that this is an indefinite integral so we don’t have to worry about the limits of integration. However we do need to make sure our answer is a function of \(x\) — we cannot leave it as a function of \(u\text{.}\)
Unlike all our previous examples, we have not cancelled out all of the \(x\)’s from the integrand. However before we do the integral with respect to \(u\text{,}\) the integrand must be expressed solely in terms of \(u\) — no \(x\)’s are allowed. (Look that integrand on the right hand side of Theorem 1.4.2.)
But all is not lost. We can rewrite the factor \(x^2\) in terms of the variable \(u\text{.}\) We know that \(u=1+x^2\text{,}\) so this means \(x^2 = u-1\text{.}\) Substituting this into our integral gives
At first glance there is nothing to manipulate here and so very little to go on. However we can rewrite \(\tan x\) as \(\frac{\sin x}{\cos x}\text{,}\) making the integral \(\int \frac{\sin x}{\cos x}\dee{x}\text{.}\) This gives us more to work with.
Now think of the integrand as being the product \(\frac{1}{\cos x}\cdot \sin x\text{.}\) This suggests that we set \(u=\cos x\) and that we interpret the first factor as the function “one over” evaluated at \(u=\cos x\text{.}\)
In all of the above substitution examples we expressed the new integration variable, \(u\text{,}\) as a function, \(u(x)\text{,}\) of the old integration variable \(x\text{.}\) It is also possible to express the old integration variable, \(x\text{,}\) as a function, \(x(u)\text{,}\) of the new integration variable \(u\text{.}\) We shall see examples of this in Section 1.9.
What is the integral that results when the substitution \(u= \sin x\) is applied to the integral \(\displaystyle \int_0^{\pi/2} f(\sin x)\,\dee{x}\text{?}\)
Let \(f(t)\) be any function for which \(\displaystyle\int_1^8 f(t)\,\dee{t}=1\text{.}\) Calculate the integral \(\displaystyle\int_1^2 x^2 f(x^3)\,\dee{x}\text{.}\)
Questions 18 through 22 can be solved by substitution, but it may not be obvious which substitution will work. In general, when evaluating integrals, it is not always immediately clear which methods are appropriate. If this happens to you, don’t despair, and definitely don’t give up! Just guess a method and try it. Even if it fails, you’ll probably learn something that you can use to make a better guess. 4