Sei sulla pagina 1di 4

MATH1116, James Borger

Mathematical Induction
Mathematical induction (induction, for short) is a method of proof based on recursion and, for this reason, especially well-suited to proving things about recursively dened mathematical objects. In these notes, I will state what Ill call basic induction (I just invented this terminology for the purpose of these notes), and Ill give some examples of applications. I will also discuss some variants that appear to be more sophisticated, but which can in fact be reduced to the basic form. Let N denote the {0, 1, 2, . . . }, the set of non-negative integers.

1. Basic induction Let P (n) be a statement that depends on a non-negative integer n. The principle of mathematical induction states that P (n) is true for all n N if the following hold: (1) P (0) is true (2) if n N and P (n) is true, then P (n + 1) is true. You can think of it like this: if youre able to stand on the rst rung of a ladder (rung number zero), and if youre also able to climb from one rung to the next (rung n to rung n + 1 for any n), then youre able to climb to any rung you want. Some terminology: (1) is often called the base case, and (2) is the induction step. The hypothesis that P (n) is true in (2) is often called the induction hypothesis. 1.1. Example: Prove 0 + 1 + + n = 1 2 n(n + 1) for all n N. Proof. For any n N, let P (n) denote the statement 0 + 1 + + n = 1 2 n(n + 1). First we 1 show the base case P (0) is true. But this is the statement 0 = 2 0 1, which is clear. Now we show the induction step. Suppose n N and that P (n) is true. In other words, we assume we know 1 0 + 1 + + n = n(n + 1). 2 We therefore have 0 + 1 + + (n + 1) = [0 + 1 + + n] + (n + 1) 1 = n(n + 1) + (n + 1) 2 1 = n + 1 (n + 1) 2 1 = (n + 1)(n + 2). 2 But this is precisely the statement P (n + 1). Thus whenever P (n) is true, P (n + 1) is true. The principal of mathematical induction then shows P (n) is true for all n N. Here is the same proof without training wheels.
1

Proof. The statement is clear for n = 0. Assuming it holds for some n N, we have 0 + 1 + + (n + 1) = [0 + 1 + + n] + (n + 1) 1 = n(n + 1) + (n + 1) 2 1 = (n + 1)(n + 2). 2 It therefore holds for n + 1. By induction, this completes the proof. 1.2. Exercise: Prove that 12 + 22 + + n2 = 1 6 n(n + 1)(2n + 1) for all n N. 2. Variants 2.1. Changing the starting point. We dont have to begin with n = 0. We can begin with n = 4 (or any number), but then we will end up proving P (n) only for n 4. For example, here is a proof that n! > 2n for integers n 4. Proof. For n = 4, the statement to be proved is 24 > 16, which is clearly true. Now suppose for induction that it holds for some n 4. Then we have (n + 1)! = (n + 1) n! > (n + 1) 2n >22
n

by the induction hypothesis (and because n + 1 > 0) because n + 1 > 2 (and 2n > 0)

= 2n+1 , which completes the proof, by induction. We can think of what we just did in the following way. For n N, let P (n) be the statement n! > 2n and let Q(n) be P (n + 4). Therefore proving P (n) for n 4 is the same as proving Q(n) for n 0. We could write out the proof as a basic induction proof (i.e. starting at n = 0) using Q(n), but really this is just the same as working with P (n) but starting at n = 4. 2.2. Looking back more than one step. Sometimes when trying to prove P (n + 1), its useful to use not just P (n) but also P (n 1) or even P (n 2) and so on. Its not necessary, but before I explain why, lets look at an example. Let F0 = 0, F1 = 1, and for n 2 (an integer as always) let Fn = Fn1 + Fn2 . This is called the Fibonacci sequence. The rst few terms are 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55. We will prove that Fn < 2n for all n N. I will give the proof rst, and after discuss how it diers from one by basic induction. Proof. The statement to be proved is clear for n = 0 and n = 1. Now, for any n 1, suppose we know Fr < 2r for all r such that 0 r n. Because n 1, both r = n and r = n 1 lie in this range. We therefore have Fn < 2n and Fn1 < 2n1 . Adding these together, we see Fn + Fn1 < 2n + 2n1 = 2n 1 + 1 2 < 2n 2 = 2n+1 .

But the left-hand side is Fn+1 , by denition, and therefore the statement to be proved holds for n + 1. There are two dierences with basic induction:

(1) the induction hypothesis is that the statement to be proved holds, not just for n, but for all smaller values as well, and (2) there is more than one base case (both n = 0 and n = 1). The need for the rst dierence, as you can see by studying the proof, is because the Fibonacci sequence is dened in terms of the two pervious terms in the sequence. The need for second dierence is a consequence of having the rst. If we did not treat n = 1 by hand, we would have to allow n = 0 in the induction step (where were proving the statement for n + 1). But then the number r = n 1 = 1 would not satisfy the requirement 0 r n. Notice the similarity with the denition of the Fibonacci sequence. Because the recurrence relation Fn = Fn1 + Fn2 involved two previous terms, it was necessary to dene to initial terms (F0 and F1 ) to get started. This proof can be set in the formalism of a basic-induction proof in the following way. Let P (n) denote the statement Fn < 2n . This is what we ultimately want to prove for all n. But now let Q(n) denote the statement for all r such that 0 r n, we have Fr < 2r . Note that Q(n) is a formally stronger statementif Q(n) holds then taking r = n, we see P (n) must hold. But the proof above is really a basic induction proof using Q(n) rather than P (n). Observe that verifying the base case Q(1) is (by the denition of Q) the same as verifying both P (0) and P (1). And the induction step shows how Q(n) implies Q(n + 1). I could have written the proof above in the strict form of a basic-induction proof, but it would have had a somewhat heavy style, and so, like most authors, I didnt. 3. Inventors Paradox Another point that the proof of Fn < 2n brings up is what George P olya has called the Inventors Paradox: The more ambitious plan may have more chances of success. It is usually true that it is harder to prove a formally stronger statement. If Q implies P , then any proof of Q immediately gives a proof of P , but not the other way around. But in this example, we werent able to prove P (n) for all n by itself; we paradoxically had to prove the stronger statement Q(n). The reason for this, and the reason why the Inventors Paradox is not a paradox, lies at the heart of what induction is. In proving Q(n + 1) by induction, we assume Q(n). And while it is true that Q(n + 1) is a stronger statement than P (n + 1), it is also true that Q(n) is a stronger statement than P (n). In other words, even though we are trying to prove something harder, were also giving ourselves more equipment to do it with, and so it might actually be easier to do. And so the Inventors Paradox might also be called the Borrowers Tautology: its easier to make more money if you have more money than less if you have less. Exercise: Prove that for any n N, 13 + 23 + + n3 is the square of an integer. (Hint: Remember the Inventors Paradox!) Exercise: Consider the following two-player game. There are several piles of stones in a row. In turns, each player takes some number of stones from one pile and places them in any piles to the left it. (Both players are facing the same way and so agree on which way is left.) The rst player who is unable to move loses. Prove that no matter how they play, the game will end after a nite number of steps. 4. Appendix: the logical status of induction You may want a proof of the validity of induction, but I dont actually know what its logical status is. I think I read once that it is one of the basic building blocks that logicians use to construct N itself, and so I refer you to a course or book in logic. Like most logical issues, it is really of no consequence to the working mathematician. Here is another point of view on justifying it; if youre not interested, please skip it. Suppose (1) and (2) from the rst section hold but P (n) isnt always true. Then let m be the smallest

number such that P (m) is false, and let r = m 1. Because P (0) is true (by assumption #1), we see that r N and therefore that P (r) makes sense. Because r is smaller than m and because m is the smallest counter-example, we know P (r) is true. But then by assumption #2, P (r + 1) is true. Because m = r + 1, we see P (m) is true and hence have a contradiction. Therefore the original assumption that P (n) is not always true must be false, and so P (n) is always true. This may look like an improvement, but all I did was show that induction is a consequence of the so-called well-orderedness principle (that every non-empty subset of non-negative integers has a smallest element) if we allow ourselves to use proof by contradiction. But logicians tend not to like to base things on proofs by contradiction, and its also not clear to me that the well-orderedness principle should be thought of as more fundamental than induction. Whatever the case, if youre interested in these things, you should consult a logician, or maybe an expert on topos theory.

Potrebbero piacerti anche