The method of integration by substitution comes from the rule for differentiation.
The method of integration by parts comes from the rule for differentiation.
Hint+
Read back over Sections 1.4 and 1.7
of the CLP-2 text. When these methods are introduced, they are justified using the
corresponding differentiation rules.
Answer+
chain; product
Full solution+
Integration by substitution is just using the chain rule, backwards:
Suppose you want to evaluate an integral using integration by parts. You choose part of your integrand to be u, and part to be dv. The part chosen as u will be: (differentiated, antidifferentiated). The part chosen as dv will be: (differentiated, antidifferentiated).
Hint+
Remember our rule: ∫udv=uv−∫vdu. So, we take u and use it to make du, and we take dv and use it to make v.
Answer+
The part chosen as u will be differentiated. The part chosen as dv will be antidifferentiated.
Full solution+
Remember our rule: ∫udv=uv−∫vdu. So, we take u and use it to make du–that is, we differentiate it. We take dv and use it to make v–that is, we antidifferentiate it.
Let f(x) and g(x) be differentiable functions. Using the quotient rule for differentiation, give an equivalent expression to ∫g(x)f′(x)dx.
Hint+
According to the quotient rule,
dxd{g(x)f(x)}=g2(x)g(x)f′(x)−f(x)g′(x).
Antidifferentiate both sides of the equation, then solve for the expression in the question.
Answer+
∫g(x)f′(x)dx=g(x)f(x)+∫g2(x)f(x)g′(x)dx
Full solution+
We'll use the same ideas that led to the method of integration by parts. (You can review this in your text, or see the solution to Question 1 in this section.)
According to the quotient rule,
Suppose we want to use integration by parts to evaluate ∫u(x)⋅v′(x)dx for some differentiable functions u and v. We need to find an antiderivative of v′(x), but there are infinitely many choices. Show that every antiderivative of v′(x) gives an equivalent final answer.
Hint+
Remember all the antiderivatives differ only by a constant, so you can write them all as v(x)+C for some C.
Answer+
All the antiderivatives differ only by a constant, so we can write them all as v(x)+C for some C. Then, using the formula for integration by parts,
In order to figure out the first product (and the second integrand), you need to know the antiderivative of f(x)–but that's exactly what you're trying to figure out!
Full solution+
Suppose we choose dv=f(x)dx, u=1. Then v=∫f(x)dx, and du=dx. So, our integral becomes:
In order to figure out the first product (and the second integrand), you need to know the antiderivative of f(x)–but that's exactly what you're trying to figure out! So, using integration by parts has not eased your pain.
We note here that in certain cases, such as ∫logxdx (Example 1.7.8 in the CLP-2 text), it is useful to choose dv=1dx. This is similar to, but crucially different from, the do-nothing method in this problem.
2Stage 2Procedural
Practising the skill itself, until applying it is automatic.
You'll probably want to use integration by parts. (It's the title of the section, after all). You'll break the integrand into two parts, integrate one, and differentiate the other. Would you rather integrate logx, or differentiate it?
Answer+
2x2logx−4x2+C
Full solution+
For integration by parts, we want to break the integrand into two pieces, multiplied together. There is an obvious choice for how to do this: one piece is x, and the other is logx. Remember that one piece will be integrated, while the other is differentiated. The question is, which choice will be more helpful. After some practice, you'll get the hang of making the choice. For now, we'll present both choices–but when you're writing a solution, you don't have to write this part down. It's enough to present your choice, and then a successful computation is justification enough.
In Example 1.7.8 of the CLP-2 text, we found the antiderivative of logarithm, but it wasn't trivial. We might reasonably want to avoid using this complicated antiderivative. Indeed, Option 2 (differentiating logarithm, antidifferentiating x) looks promising–when we multiply the blue equations, we get something easily integrable– so let's not even bother going deeper into Option 1.
That is, we perform integration by parts with u=logx and dv=xdx,
so that du=xdx and v=2x2.
Our integrand is the product of two functions, and there's no clear substitution. So, we might reasonably want to try integration by parts. Again, we have two obvious pieces: logx, and x−7. We'll consider our options for assigning these to u and dv:
Again, we remember that logarithm has some antiderivative we found in Example 1.7.8 of the CLP-2 text, but it was something complicated. Luckily, we don't need to bother with it: when we multiply the red equations in Option 1, we get a perfectly workable integral.
We perform integration by parts with u=logx and dv=x−7dx,
so that du=xdx and v=−6x−6.
CLP-2 text shows you how to find the antiderivative. Then the Fundamental Theorem of Calculus Part 2 gives you the definite integral.
Answer+
π
Full solution+
To integrate by parts, we need to decide what to use as u, and what to use as dv. The salient parts of this integrand are x and sinx, so we only need to decide which is u and which dv. Again, this process will soon become familiar, but to help you along we show you both options below.
Compare to Question 8. Try to do this one all the way through without peeking at another solution!
Answer+
2π−1
Full solution+
When we have two functions multiplied like this, and there's no obvious substitution, our minds turn to integration by parts. We hope that our integral will be improved by differentiating one part and antidifferentiating the other. Let's see what our choices are:
At first glance, multiplying the red functions and multiplying the blue functions give largely equivalent integrands to what we started with–none of them with obvious antiderivatives. In previous questions, we were able to choose u=x, and then du=dx, so the “x" in the integrand effectively went away. Here, we see that choosing u=x3 will lead to du=3x2dx, which has a lower power. If we repeatedly perform integration by parts, choosing u to be the power of x each time, then after a few iterations it should go away, because the third derivative of x3 is a constant.
So, we start with Option 2: u=x3, dv=exdx, du=3x2dx, and v=ex.
Now, we take u=x2 and dv=exdx, so du=2xdx and v=ex. We're only using integration by parts on the actual integral–the rest of the function stays the way it is.
Remark: In order to be technically correct in our antidifferentiation, we should add the +C as soon as we do the first integration by parts. However, when we are using integration by parts, we usually end up evaluating an integral at the end, and we add the +C at that point. Since the +C comes up eventually, it is common practice to not clutter our calculations with it until the end.
Similarly to Question 10, look for a way to use integration by parts a few times to simplify the integrand until it is antidifferentiatable.
Answer+
2x2log3x−43x2log2x+43x2logx−83x2+C
Full solution+
Since our integrand is two functions multiplied together, and there isn't an obvious substitution, let's try integration by parts. Here are our salient options.
This calls for some strategizing. Using the template of Example 1.7.8
in the CLP-2 text, we could probably figure out the antiderivative of log3x. Option 1 is tempting, because our x-term goes away. So, there might be a benefit there, but on the other hand, the antiderivative of log3x is probably pretty complicated.
Now let's consider Option 2. When we multiply the blue functions together, we get something similar to our original integrand, but the power of logarithm is smaller. If we were to iterate this method (using integration by parts a few times, always choosing u to be the part with a logarithm) then eventually we would end up differentiating logarithm. This seems like a safer plan: let's do Option 2.
We use integration by parts with u=log3x, dv=xdx, du=x3log2xdx, and v=21x2.
Once again, technically there is a +C in the work after the first integration by parts, but we follow convention by conveniently suppressing it until the final integration.
Neither option gives us something immediately integrable, but Option 1 replaces our x2 term with a lower power of x. If we repeatedly apply integration by parts, we can reduce this power to zero. So, we start by choosing u=x2 and dv=sinxdx, so du=2xdx and v=−cosx.
Using integration by parts again, we want to be differentiating (not antidifferentiating) x, so we choose u=x, dv=cosxdx, and then du=dx (x went away!), v=sinx.
If you let u=logt in the integration by parts, then du works quite nicely with the rest of the integrand.
Answer+
(t3−25t2+6t)logt−31t3+45t2−6t+C
Full solution+
This problem is similar to Questions 6 and 7: integrating a polynomial multiplied by a logarithm. Just as in these questions, if we use integration by parts with u=logt, then du=t1dt, and our new integrand will consist of powers of t–which are easy to antidifferentiate.
So, we use u=logt, dv=3t2−5t+6, du=t1dt, and v=t3−25t2+6t.
Those square roots are a little disconcerting– get rid of them with a substitution.
Answer+
es(2s−4s+4)+C
Full solution+
Before we jump to integration by parts, we notice that the square roots lend themselves to substitution. Let's take w=s. Then dw=2s1ds, so 2wdw=ds.
∫sesds=∫w⋅ew⋅2wdw=2∫w2ewdw
Now we have nearly the situation of Question 10. We can repeatedly use integration by parts, with u as the power of w, to get rid of the polynomial part. We'll start with u=w2, dv=ewdw, du=2wdw, and v=ew.
=2[uvw2ew−∫vdu2wewdw]=2w2ew−4∫wewdw
We use integration by parts again, this time with u=w, dv=ewdw, du=dw, and v=ew.
This can be solved using the same ideas as Example 1.7.8 in the CLP-2 text.
Answer+
xlog2x−2xlogx+2x+C
Full solution+
Let's use integration by parts. What are our parts? We have a few options.
Following Example 1.7.8 in the CLP-2 text,
we choose u=log2x and dv=dx, so that du=x2logxdx and v=x.
∫log2xdx=uvxlog2x−∫vdu2logxdx
Here we can either use the antiderivative of logarithm from memory, or re-derive it. We do the latter, using integration by parts with u=logx, dv=2dx, du=x1dx, and v=2x.
Our integrand is two functions multiplied together: logx and logx. So, we will use integration by parts with u=logx, dv=logx, du=x1dx, and (using the antiderivative of logarithm, found in Example 1.7.8
in the CLP-2 text) v=xlogx−x.
Not every integral should be evaluated using integration by parts.
Answer+
ex2+1+C
Full solution+
This is your friendly reminder that to a person with a hammer, everything looks like a nail. The integral in the problem is a classic example of an integral to solve using substitution. We have an “inside function," x2+1, whose derivative shows up multiplied to the rest of the integrand. We take u=x2+1, then du=2xdx, so
You know, or can easily look up, the derivative of arccosine.
You can use a similar trick as the book did when antidifferentiating other inverse trigonometric functions in Example 1.7.9 of the CLP-2 text.
Answer+
yarccosy−1−y2+C
Full solution+
In Example 1.7.9 of the CLP-2 text, we saw that integration by parts was useful when the integrand has a derivative that works nicely when multiplied by x. We use the same idea here.
Let u=arccosy and dv=dy,
so that v=y and du=−1−y2dy.
The integrand (2y)2+14y2 is a rational function.
So the remaining integral can be evaluated using the method of partial
fractions, starting with long division. But it is easier to just notice
that
4y2+14y2=4y2+14y2+1−4y2+11.
We therefore have:
We've got an integrand that consists of two functions multiplied together, and no obvious substitution. So, we think about integration by parts. Let's consider our options. Note in Example 1.7.9 of the CLP-2 text, we found that the antiderivative of arctangent is
xarctanx−21log(1+x2)+C.
What if we had tried the other option? That is, u=x2, du=2xdx, dv=arctanx, and v=xarctanx−21log(1+x2). It's not always the case that both options work, but sometimes they do. (They are almost never of equal difficulty.) This solution takes advantage of two previously hard-won results: the antiderivatives of logarithm and arctangent.
This example is similar to Example 1.7.10 in the CLP-2 text. The functions ex/2 and cos(2x) both do not substantially alter when we differentiate or antidifferentiate them. If we use integration by parts twice, we'll end up with an expression that includes our original integral. Then we can just solve for the original integral in the equation, without actually integrating.
Answer+
172ex/2cos(2x)+178ex/2sin(2x)+C
Full solution+
This example is similar to Example 1.7.10 in the CLP-2 text. The functions ex/2 and cos(2x) both do not substantially alter when we differentiate or antidifferentiate them. If we use integration by parts twice, we'll end up with an expression that includes our original integral. Then we can just solve for the original integral in the equation, without actually antidifferentiating.
Let's use u=cos(2x) and dv=ex/2dx, so du=−2sin(2x)dx and v=2ex/2.
Remark: remember that C is a stand-in for “we can add any real constant". Since C can be any number in (−∞,∞), also 17C can be any number in (−∞,∞). So, rather than write 17C in the last line, we re-named 17C to C.
This looks a bit like a substitution problem, because we have an “inside function."
It might help to review Example 1.7.11 in the CLP-2 text.
Answer+
2x[sin(logx)−cos(logx)]+C
Full solution+
This question looks like a substitution, since we have an “inside function." So, let's see where that leads: let u=logx. Then du=x1dx. We don't see this right away in our function, but we can bring it into the function by multiplying and dividing by x, and noting from our substitution that eu=x.
∫sin(logx)dx=∫xxsin(logx)dx=∫eusinudu
Using the result of Example 1.7.11 in the CLP-2 text:
It's not clear how to antidifferentiate the integrand, but we can certainly differentiate it. So, keeping in mind the method of Example 1.7.11 in the CLP-2 text, we take u=sin(logx) and dv=dx, so du=x1cos(logx)dx and v=x.
∫sin(logx)dx=uvxsin(logx)−∫vducos(logx)dx
Continuing on, we again use integration by parts, with u=cos(logx), dv=dx, du=−x1sin(logx)dx, and v=x.
Remark: remember that C is a stand-in for “we can add any real constant". Since C can be any number in (−∞,∞), also 2C can be any number in (−∞,∞). So, rather than write 2C in the last line, we re-named 2C to C.
This is similar to the integral ∫xexdx, which we saw in Example 1.7.1 of the CLP-2 text. Let's write 2=elog2 to take advantage of the easy integrability of ex.
=∫x⋅exlog2dx
We use integration by parts with u=x, dv=exlog2dx; du=dx, v=log21exlog2. (Remember log2 is a constant. If you'd prefer, you can do a substitution with s=xlog2 first, to have a simpler exponent of e.)
It's not obvious where to start, but in general it's nice to have the arguments of our trig functions the same. So, we use the identity sin(2x)=2sinxcosx.
∫ecosxsin(2x)dx=2∫ecosxcosxsinxdx
Now we can use the substitution w=cosx, dw=−sinxdx.
=−2∫wewdw
From here the integral should look more familiar. We can use integration by parts with u=w, dv=ewdw, du=dw, and v=ew.
We've got an integrand that consists of several functions multiplied together, and no
obvious substitution. So, we think about integration by parts. We know an antiderivative for
(1−x)21, because we know dxd1−x1=(1−x)21.
So let's try dv=(1−x)2dx and u=xe−x. Then v=1−x1
and du=(1−x)e−xdx. So, by integration by parts,
You'll want to do an integration by parts for (a)–check the end result to get a guess as to what your parts should be. A trig identity and some amount of algebraic manipulation will be necessary to get the final form.
Answer+
(a)
We integrate by parts with u=sinn−1x and dv=sinxdx,
so that du=(n−1)sinn−2xcosx and v=−cosx.
Moving the last term on the right hand side to the left hand side gives
n∫sinnxdx=−sinn−1xcosx+(n−1)∫sinn−2xdx
Dividing across by n gives the desired reduction formula.
(b) 25635π≈0.4295
Full solution+
(a)
The “parts" in the integrand are powers of sine. Looking at the right hand side of the reduction formula, we see that it looks a little like the derivative of sinn−1x, although not exactly. So, let's
integrate by parts with u=sinn−1x and dv=sinxdx,
so that du=(n−1)sinn−2xcosx and v=−cosx.
Let R be the region between the curves T(x)=xe3x and B(x)=x(1+2x) on the interval 0≤x≤3. (It is true that T(x)≥B(x) for all 0≤x≤3.) Compute the volume of the solid formed by rotating R about the x-axis.
Hint+
Your integral can be broken into two integrals, which yield to two different integration methods.
Answer+
π(3617e18−4373)
Full solution+
For a fixed value of x, if we rotate about the x-axis, we form a washer of
inner radius B(x) and outer radius T(x) and hence of area π[T(x)2−B(x)2].
We integrate this function from x=0 to x=3 to find the total volume V:
Let f(0)=1, f(2)=3 and f′(2)=4.
Calculate
∫04f′′(x)dx.
Hint+
Think, first, about how to get rid of the square root in the argument of f′′,
and, second, how to convert f′′ into f′. Note that you are told that f′(2)=4
and f(0)=1, f(2)=3.
Answer+
12
Full solution+
To get rid of the square root in the argument of f′′, we make the change of variables (also called “substitution")
x=t2,dx=2tdt.
∫04f′′(x)dx=2∫02tf′′(t)dt
Then, to convert f′′ into f′, we integrate by parts with
u=t, dv=f′′(t)dt, v=f′(t).
As we saw in Section 1.1 of the CLP-2 text, there are many different ways to interpret a limit as a Riemann sum. In the absence of instructions that restrain our choices, we go with the most convenient interpretations.
With that in mind, we choose:
that our Riemann sum is a right Riemann sum (because we see i, not i−1 or i−21)
Δx=n2 (because it is multiplied by the rest of the integrand, and also shows up multiplied by i),
then xi=a+iΔx=n2i−1, which leads us to a=−1 and
f(x)=xex.
Finally, since Δx=nb−a=n2 and a=−1, we have b=1.
So, the limit is equal to the definite integral
n→∞limi=1∑nn2(n2i−1)en2i−1=∫−11xexdx
which we evaluate using integration by parts with u=x, dv=exdx, du=dx, and v=ex.