Open Book Publishers logo Open Access logo
  • button
  • button
  • button
GO TO...
Contents
Copyright
book cover
BUY THE BOOK

Problem 32:  Integration by substitution ( ) 1999 Paper II

Find dy dx if

y = ax + b cx + d. ()

By using changes of variable of the form (), or otherwise, show that

01 1 (x + 3)2 ln x + 1 x + 3dx = 1 6ln3 1 4ln2 1 12,

and evaluate the integrals

01 1 (x + 3)2 ln x2 + 3x + 2 (x + 3)2 dx  and   01 1 (x + 3)2 ln x + 1 x + 2dx.

Comments

You will find that the change of variable in each case is clearly signalled: it is really only the denominator of () that matters.

For the first integral, you do the obvious thing, but for the second and third integrals you have to be quite ingenious to get the argument of the logs in a suitable form. Once you have got the idea for the second integral, you should be able to see the connection with the third integral, but it would be hard to do the third integral without having done the second. That’s what I like about this question: one thing leads to another.

I have written below the version of this question that was proposed by the setter, because I thought you would be interested to see how a question evolves.

By changing to the variable y defined by

y = 2x 3 x + 1 ,

evaluate the integral

24 2x 3 (x + 1)3 ln 2x 3 x + 1 dx.

Evaluate the integral

9252z3 2 5z2 ln2 5z1 2 dz.

Note in particular the way that the ideas in the final draft are closely knit and better structured: the first change of variable is strongly signalled but not explicit and the following parts, though based on the same idea, require increasing ingenuity. The first draft required quite a jump to evaluate the second integral.

Note also that the final integral of the first draft has an unpleasant contrived appearance, whereas the the integrals of the final draft are rather pleasing: beauty matters to mathematicians.

Solution to problem 32

Differentiating gives dy dx = a(cx + d) c(ax + b) (cx + d)2 = ad bc (cx + d)2.

For the first integral, set y = x + 1 x + 3:

01 1 (x + 3)2 ln x + 1 x + 3dx = 1 21312 lnydy = 1 2ylny y1312,

which gives the required answer. (The integral of lny is a standard integral; it can be done by parts, first substituting z = lny, if you like.)

The second integral can be expressed as the sum of two integrals of the same form as the first integral, since

ln x2 + 3x + 2 (x + 3)2 = ln (x + 1)(x + 2) (x + 3)2 = ln x + 1 x + 3 + ln x + 2 x + 3.

We have already done the first of these integrals. Using the substitution y = x + 2 x + 3  in the second of these integrals gives

2334 lnydy = 17 12ln3 13 6 ln2 1 12.

For the final integral, note that

ln x + 1 x + 2 = ln x + 1 x + 3 ln x + 2 x + 3,

so the required integral is the difference of the two integrals that we summed in the previous part, i.e.

5 4ln3 + 23 12ln2.

Post-mortem

I hope you had time to try the original version of this question, given in the comments above. The last part is a nice puzzle.

For the first integral of the original version, the change of variable gives

1 51 3 1ylnydy = 1 90ln3 2 45.

I checked this using Wolfram Alpha. For the second integral, you have to guess the substitution. There are plenty of clues, but the most obvious place to start is the log. The argument of the log can be written as

2z 5 z .

It is now a bit of a leap in the dark (too much of a leap, I thought), but if we decide to convert this exactly into the previous integral, we should take z = (x + 1)2. And, magic (or contrived STEP question), it works, giving the first integral almost exactly (note especially the the limits transform as they should). The only difference is a factor of 2 so the answer is twice the previous answer.

A change of variable of the form () is called a linear fractional or Möbius transformation. It is of great importance in the theory of the geometry of the complex plane.