MATH EXPLORERS' CLUB Cornell Department of Mathematics 


 Appetizers
corner
Appetizers

Before jumping in to some of the more mystifying applications of induction, let’s take a look at how it works in some more straightforward (but by no means trivial) examples. Consider the following claim:

The sum of the first n squares is equal to (1/6)n(n + 1)(2n + 1).


Once again, this is easy enough to check by hand for the first few values of n:

12 = 1  and  (1/6)(1)(1 + 1)(2 x 1 + 1) = (1/6)(2)(3) = 1,
yes;
12 + 22 = 1 + 4 = 5  and  (1/6)(2)(2 + 1)(2 x 2 + 1) = (1/6)(2)(3)(5) = 5,
yes;
12 +22 +32 = 1+4+9 = 14  and  (1/6)(3)(3+1)(2 x 3+1) = (1/6)(3)(4)(7) = 14,

yes. So the claim holds up to the sum of the first 3 squares, but already things are starting to get cumbersome, and this method of checking by hand has no hope of yielding a proof for all such sums. We need to find a better way to go about things.

We might try the same method here as was used in proving that all squares of even numbers are divisible by four. Let n be any number, and consider the sum of the first n squares:

12 + 22 + 32 + 42 + ... + (n − 1)2 + n2 .

If we could somehow add up this sum and show that the result is (1/6)n(n + 1)(2n + 1) then we would be done. Problematically, there is no obvious way to do this addition (try it and see). We turn to induction.

 First comes the base case n = 1: we must prove that the sum of the first 1 squares is equal to (1/6)(1)(1 + 1)(2 x 1 + 1). This has already been done, above, and it was rather easy; it is often true that the base case of an inductive argument is easy or trivial.

Next comes the inductive step. We start by making our inductive hypothesis: assume that the claim is true for n. Our task is to show that it is also true for n + 1. In other words, we need to show that the sum of the first n + 1 squares is equal to

(1/6)(n + 1)[(n + 1) + 1][2(n + 1) + 1],                                 (1)

which is just the original formula with n replaced everywhere by (n + 1).

Our inductive hypothesis amounts to the following equation:

12 + 22 + ... + (n − 1)2 + n2 = (1/6)n(n + 1)(2n + 1).

Adding (n + 1)2 to both sides of the above equation yields

12 + 22 + ... + (n − 1)2 + n2 + (n + 1)2 = (1/6)n(n + 1)(2n + 1) + (n + 1)2.

Notice that the left hand side of this equation is exactly the sum of the first (n + 1) squares! If we could show that the right hand side of this equation is equal to the formula labeled by (1), above, then we will have completed the induction step. Although it is not immediately apparent, the two formulas in question are in fact equal. The right sequence of algebraic manipulations reveals this to be so:

(1/6)n(n + 1)(2n + 1) + (n + 1)2  = (n + 1)[(1/6)n(2n + 1) + (n + 1)]
                                              = (n + 1)[(1/3)n2 + (7/6)n + 1]
                                                  = (n + 1) [(1/6)n + (1/3)][2n + 3]
                                            = (1/6)(n + 1)(n + 2)(2n + 3).

Exercise 2

Verify the above equalities.

The reader is invited to verify these calculations in more detail. Whatever detail is omitted here is done so for the sake of clarity: there are only so many lines of algebra one can reasonably be expected to read through before going cross-eyed. The best path to understanding is to work out the details for your-self, using the above as a guide.

This completes the induction and therefore finishes the proof: we have verified that the sum of the first n squares is always equal to (1/6)n(n + 1)(2n + 1). It is worth noting, however, that I have left the origins of this formula a complete mystery. Induction proved quite useful in verifying that the given formula is the correct one, but how one might come to suspect that formula in the first place is another issue entirely. I refer the reader to [1], chapter 2, for a very pleasing exploration of this issue (the rest of the book is nice, too).

Next I want to give an example of induction used in a very different situation: to solve a puzzle which at first glance doesn’t seem to have much use for induction at all. The puzzle is a tiling puzzle. It asks:

Can a chessboard minus a rook’s square be tiled with triominos?


What? The first step to solving any problem is understanding the question, so we start with that. First, a ‘triomino’ is a two-dimensional figure made out of three equal squares glued together in the configuration pictured in Figure 1.
figure1
Figure 1: A triomino

A ‘chessboard’ is a more familiar object: it is an 8 x 8 square made by gluing together 64 smaller squares. Implicit in the question is the assumption that the squares that make up a triomino are the same size as the squares that make up the chessboard. A ‘rook’s square’ is any one of the four corner squares on the chessboard. Finally, the ‘tiling’ that the question asks for is an arrangement of triominos on the chessboard such that every square of the chessboard (except for one corner square) is covered by a square from a triomino. No two triominos are allowed to overlap, and every triomino must be positioned so that it lies entirely over the chessboard. Can this be done?

The answer is yes, and if you can get your hands on some triominos (or make some yourself), after a little experimentation you’ll be convinced of this. So we can ask a harder question:
 
Can every 2n x 2n board, minus a corner square, be tiled by triominos?


The chess board example corresponds to the case n = 3. We can proceed in the general case by induction in a rather surprising way. That induction is useful here is perhaps not so surprising, given the nature of the claim we wish to prove (i.e. we want to prove the claim for n = 1, 2, 3, . . .). But the geometrical aspect of this problem contrasts sharply with the previous example.

The base case for the induction is n = 1, so we must consider a 21 x 21 board with one of the corner pieces removed and determine whether this can be tiled by triominos. But of course it can be, since what remains of the board is precisely the same shape as a single triomino. Once again, the base case was easy.

Now the induction step. We assume the result is true for n and try to prove it true for n + 1. This means that we can assume that the 2n x 2n board, with one corner piece removed, can be tiled by triominos. See Figure 2.
figure2
Figure 2: A tiling of the 2n x 2 n board, minus the upper right corner square.

Note that the actual tiling pattern is not shown; at this point we need not be concerned with how the tiles are arranged, but merely that the board minus the corner is tiled in some way.

How can we use this tiling, our inductive hypothesis, to obtain the analogous result for a 2n+1 x 2n+1 board? Before reading ahead to the solution, try to play around with the possibilities for a time. The answer is not complicated or difficult to understand, but it requires the right flashes of insight to come up with it.

The first important insight is that the 2n+1 x 2n+1 board can be divided into exactly four 2n x 2n boards by cutting it vertically and horizontally down the center lines. Our inductive hypothesis tells us that we know how to tile such boards with triominos, minus a corner square. So let that be done.

The second insight is best provided in picture form: we glue the four quarters back together in the configuration shown in Figure 3. This leaves one corner square (the one in the upper right) untiled, along with three central squares. But look! The three central squares that are untiled have been glued back together in exactly the shape of a triomino! Therefore, with the addition of one extra triomino in that conspicuous space, we have managed to completely tile the 2n+1 x 2n+1 board, minus the square in the upper right corner. This completes the induction and finishes the proof.

Before moving on, it is worth noting a small corollary to the result we just proved. Since each triomino is composed of exactly three squares, any area that can be tiled with triominos must consist of exactly 3k squares, where k is the number of triominos used in the tiling.

Exercise 3

Show that the converse of this statement is false. That is, show that there are areas consisting 3k squares, for some natural number k, that cannot be tiled with triominos.

figure3
Figure 3: Gluing the four quarters back together.

The number of squares in a 2n x 2n board is 2n x 2n = 22n . So a tiling of this board minus a corner square is a tiling of an area consisting of exactly 22n − 1 squares. We can therefore deduce that 22n − 1 = 3kn for some kn (the subscript n is included to indicate that the number kn depends on n). This isn’t the most interesting result in the world, but it is not exactly obvious and we get it for free from our tiling proof.


Exercise 4

Prove this result directly. That is, prove by induction on n without reference to tilings that for all natural numbers n the number 22n − 1 is divisible by 3.

Exercise 5

Find an explicit formula for kn in terms of n. (Hint: use the tiling proof or your answer to the previous exercise to guess the formula and then verify it.)


Previous lesson Top Next lesson