Вы находитесь на странице: 1из 16

Questions and Answers about proof by induction

September 16, 2005 I will only hand out the rst four pages of these questions and answers in class. Print out the rest at home; you can nd it at:
http://www.math.lsu.edu/ verrill/teaching/discrete2020/

Contents
1 Questions about proof in general 2 Questions about proof by induction 3 Questions about Assumptions 4 Induction and Sums 5 Where do the formulas for sums come from? 6 Induction problems involving exponents 7 Strong induction and double induction 8 Other examples of proof by induction 9 I dont understand 10 Comments on group work 1 2 6 7 10 11 13 14 17 17

Questions about proof in general


1. Q: Can you give us a one or two class crash course on everything we need to know about proofs? A: This whole course should be giving you lots of things you need to know about proofs. We are starting o with some strategies for proving results; the rst is proof by induction. We will cover more strategies later. Everything you need to know is too much for one or two classes. A few points about theorems and proofs: The statement of a theorem generally has hypothesis and conclusion. You need to assume the hypothesis, and show how these lead to the conclusion. A proof is a sequence of true statements, one of which follows from the next via a logical argument. A statement is a sentence which can be either true or false, e.g., The sky is orange, I like turnips, and 47 > 90 are all statements (though they might not be true, or provable). We also call sentences like x is a perfect square and x > y statements, though whether or not they are true depends on what x and y actually are. But is 5 bigger than 3?, x + y , the yellow of the sunset, Baton Rouge and Hello are not statements. Logical arguments usually follow by using some kind of algebra, or geometry, or some other tools in some particular area of mathematics. The hypothesis is a statement H , maybe H (x), and the conclusion is also a statement. A typical theorem might have the form: Theorem: If H (x) holds, then C (x) holds 1

This theorem can also be written as H (x) C (x). For an example, we might have H (x): x is a real number and x > 1 C (x): x2 > x. In this case, the theorem is true, which means that whenever H (x) is true C (x) is also true. Some people nd it useful to write truth tables. Heres one with H (x) and C (x) as above. H (x) T T F F C (x) T F F F H (x) C (x) T F T T examples with these truth values x=7 no examples, since for this case H (x) C (x) is always true x=0 x = 5

Here T means true, and F means false. A truth table can help in understanding the logic of a proof, but a truth table is not a proof. Sometimes it might be necessary to use more complicated logical constructions, and well talk a bit about this when we cover another proof strategy, called proof by contradiction. We will talk more about proofs as the course progresses. Meanwhile, there are some useful web pages about proofs. Have a look at: http://www.math.ucsd.edu/ebender/proofs.html http://pass.maths.org.uk/issue7/features/proof1/index-gifd.html http://en.wikipedia.org/wiki/Mathematical induction http://www.math.csusb.edu/notes/logic/lognot/node1.html 2. Q: Is there a certain set of words to be used when writing proofs? Like a standard example you could show us, where we could just go back and ll in the blanks of a particular problem? A: There are many dierent kinds of proofs, so it would be possible to give a ll in the blanks for some kinds, but there are too many to give such a form for all proofs. However, for proof by induction for sums, we have given a ll in the blanks method in class. 3. Q: Do you always use words in proving or is there a shorter way to prove it by only numbers? A: Usually you have to use words and numbers. Quite often the words can be minimal, since you can assume the reader knows what proof by induction is, so you might just write, Proof by induction, base case: induction step, and QED. Writing clearly is very important, especially for anyone who wants to become a teacher. 4. Q: After doing base cases and nding patterns, I dont know how to use induction to prove a case. I dont know how to put answers in mathematical terms. A: I think this is a question of practice and experience. The more examples you read through, and the more exercises you try, the easier this will become. There are quite a lot of examples in this hand out.

Questions about proof by induction

5. Q: What exactly is proof by induction? We are learning how to solve, but what is the formal denition? A: Let P (n) be a statement involving a positive whole number. (See Question 1 for a denition of a statement; see the previous hand out on induction for three examples.) A proof by induction of the statement P (n) for all n n0 is a proof that (a) demonstrates that P (n0 ) is true (b) Shows how to prove P (n) P (n + 1) (this is called the base case) (this is called the induction step)

This is a proof of the result for all n. E.g., if you take any given value of n, e.g., n = 17, then, supposing that n0 = 3. By substituting n = 3 into (b), we have P (3) P (4) By substituting n = 4 into (b), we have P (4) P (5) By substituting n = 4 into (b), we have P (5) P (6) And so on, so we get P (3) P (4) P (5) P (6) P (7) P (8) P (9) P (10) P (11) P (12) P (13) P (14) P (15) P (16) P (17) So, because we showed that P (3) is true (step (a), the base case), this means that P (17) is true. 6. Q: How can you prove something by induction? Do you mean in general, or in particular? See Question 5 for a formal description. The idea of proof by induction can be written even more formally, as follows, where P (n) is a statement involving an integer n.

P (n0 ) P (n) P (n + 1)n n0 P (n)n n0

Here, means and, and means for all, and means implies. If this doesnt mean anything to you, just use the denition at the the beginning of this answer. There are examples of proof by induction on the next few pages. Note that proof by induction is not the same as inductive reasoning. Inductive reasoning, also called the inductive method or the Scientic Method takes many many examples, and uses these to make a guess about what a result might be. Inductive reasoning can be a very useful method of working out what you want to prove. But even if you had many examples where you had checked that 1 + 2 + + n = n(n + 1)/2, just showing me all the examples would not convince me that the result was always true. E.g., dened two functions, f and g on positive integers, by f (n) g (n) = = n the remainder when you divide n by 8

You can check that g (n) = f (n) for n = 1, 2, 3, 4, 5, 6 and 7, but this does not imply that g (8) = f (8). 7. Q: Can you give us a formal/Layman denition of the word induction? A: Do you want the formal denition, or the laymans one? For formal denitions, see Questions 5 and 6 above. There are many analogies used to describe what a proof by induction means. Here are some (the rst is very standard; I made up the others, hopefully they help too). Domino analogy If I have an innite line of dominoes, if I knock over the rst (base case), and if the dominoes are close enough so that when I knock down domino n it knocks over domino n + 1 (induction step), then Ill knock down all the dominoes.

If I knock down the rst one, everything falls It it necessary for the rst one to fall (base case)

dominoes stretching to innity same distance between each pair of dominoes Shorter than a domino It is necessary the dominoes are close enough together (induction step)

Steps analogy If I am trying to climb some stairs, as long as I can get to the rst step (base case), and if I know that I can always get from step n to step n + 1, then I know I can climb as far as I want. Ladder analogy If I can reach the rst rung of a ladder, and if the rungs are close enough so that once Im at rung n its possible to reach rung n + 1, then I know its possible to climb the whole ladder. 8. Q: If you know how to prove something starting with the nth case, do you still need to start with the base case? A: Yes, since if the base case is false, even if P (n) P (n + 1), you could still have everything false. E.g., heres an example. Im going to dene a new kind of triangle, call it the 1 2-Pascal triangle (this is a made up name); this is like Pascals triangle, but it starts o with a row with 1 and 2, instead of starting at 1. The rules for getting to the next row are the same. So, I dene b1,0 b1,1 b1,i bn,r Here are the rst few rows: 1 1 1 1 6 5 14 4 9 16 = = = = 1 2 0 for i = 0, 1 bn1,r1 + bn1,r for n > 1 1 3 5 7 9 2 2 2 2 2

It would be a good project to investigate some patterns in this tablewhat can you see? Anyway, one statement which is not true is P(n): The sum of the elements in row n of the 1-2-Pascal triangle is 2n For some notation, let f (n) = bn,0 + bn,1 + + bn,n =
i=0 n

bn,i

So P (n) says f (n) = 2n . You can check that from the denition, it follows that the sum of row n + 1 is twice as much as the sum of the elements in row n. (The details are one of the project questions.) So we have P (n) P (n + 1). So the induction step is true. But f (1) = 3, so the base case is false. So we cant prove this result by induction.

9. Q: Is it always necessary to have g (f (n + 1)) in the right side of the inductive step, even if the problem is not a sum? A: I think you mean g (n + 1) here. No, this is not necessary. This method (described in class) is only for proving results where P (n) says something like f (1) + f (2) + + f (n) = g (n) 10. Q: Why is proof by induction better for solving some problems rather than using a dierent method? A: This depends on the problem. Some problems can be solved in many dierent ways, and induction, even if it can be used, is not always the best way. E.g., for proving that 1 + 2 + n =
n(n+1) 2

We can use Gaussian pairing (see page 75 of the text book; also done in class). My personal taste is that if possible, very often Id prefer a proof that does not use induction. E.g., I think the Gaussian pairing is much nicer than the induction proof of the same result. Usually induction does not tell us anything about how to get the result, where it comes from, or why its true. But this also depends on the situation. For some problems induction might be the best way. Question 37 gives three dierent proofs of one result. Some people collect dierent proofs of the same result. E.g., look at: http://www.cut-the-knot.org/pythagoras/index.shtml for 9 dierent proofs of the Pythagorean theorem. Which one you like best is a question of personal taste, like which kind of music do you prefer. 11. Q: Is it possible to prove a false statement true by induction? If so, what is the most likely aw? A: If induction is applied correctly, then you cant prove something false is true. However, if you missed out a step, you might have this happen. Probably the most likely aw would be to not do the base case, like in example 8. But sometimes you might think youve proved P (k ) P (k + 1), but there is some aw in the argument. Here is an example (this is well known; I dont know who rst thought of it): Theorem If you have set of horses, then they are all the same color. Proof The statement we want to prove for all n is: P (n): in a set of n horses, all horses have the same color Base case: P (1) is true, since there is only one horse. Inductive step: Suppose P (k ) is true. Suppose we have a set of k + 1 horses. Remove one horse. By induction, the remaining k horses are all the same color. Now put that horse back, and remove a dierent horse; the remaining horses are again all the same color, so the horse you added back has the same color as all the others. Hence all horses in the set of size k + 1 have the same color. Weve shown that P (k ) P (k + 1). Thus by induction all horses in any set of horses are the same color. What do you think is wrong? See http://www.personal.kent.edu/rmuhamma/Algorithms/MyAlgorithms/DeductInduct.htm for an explanation. 12. Q: Do we always have to show that the previous case is true when writing our proof? A: No, you always assume the previous case, i.e., if the statement is P (n), then you assume that P (n) is true, and show this would imply P (n + 1). But the only case you need to prove is the base case. E.g., Suppose A1 is a certain animal, and if n > 1, then An+1 is the mother of An . Suppose P (n) is the statement An is a rabbit. Then (assuming no genetic engineering or evolution, and that the world has existed for an innite length of time), we have P (n) P (n + 1). 5

So, if P (1) is true, then all P (n) are true. I.e., if A1 is a rabbit, all the ancestors of A1 are also rabbits. But if the base case is not true, then even though we still have P (n) P (n + 1), all P (n) are false. The statement If my dog was a rabbit, her mother would have been a rabbit, is true even though my dog is not actually a rabbit. 13. Q: The pattern of perfect squares by induction. A: This is not written as a question. Do you mean you want to know what happens when you add up the perfect squares? Or do you want a sum that adds up to perfect squares? 14. How can you prove a power series by induction? A: I dont understand this question. You can only prove statements by induction. A statement is a sentence that can be true of false. A power series is not a statement, it is a thing; it cant be true or false. Maybe you want to prove some result about a power series? See Question 1 for a denition of a statement. Give me a question, and Ill try and give an answer (or at least try). Give me a statement, and Ill try and work out if it is true or false. 15. Q: Is there a case where the base case f (1) is true, and f (n + 1) exists, but f (n) doesnt? A: Im not 100% sure if youve written this down properly Did you mean for a statement P (n), can P (1) and P (n) be true, but P (n + 1) false? If you meant this, then yes, there are lots of cases where the base case is true, but the other cases are false. E.g., let P (n) be the statement P (n) : 2n = n2 Suppose we thought the base case was n = 2. Then P (2) is true. But for n > 1 the result is false. P (3) is false.

Questions about Assumptions

16. Q: Do you ever assume things when doing induction? A: See question 18 below. 17. Q: What can we assume the person knows? A: See question 18 below. 18. Q. What is safe to assume (axioms, theorems, etc.) when using induction? A. Axioms are something that are always true and can be assumed. E.g., in Euclidean geometry, it is an axiom that for any two points in the plane, there is a straight line between these two points. You dont need to prove this. But you should clearly state what axioms are used. If you use a theorem, you should clearly state what you are assuming. You need to state what the hypothesis of the theorem is, what the result is, and make sure the theorem applies to the problem you are working on. In most cases, for any theorem you use, you should also be able to prove it, so ideally you should also give the proof of the theorem, unless the question says you may assume a certain result. E.g., in most cases, you can assume an am = an+m , without even mentioning that you are using a theorem, since this is a fairly basic result, which everyone knows. But if a question asks you to prove this, then you should give the details. And in any case, you should know how to prove this result, if asked. (See Question 30 for a proof). 19. Q: Is the key to proof by induction using the characteristics and denitions of a function? A: This depends on what you are trying to prove, and whether there is a function or not. But yes, in general, whatever objects your proof is about, using the denition of that object is often the key (whether the object is a polygon with n-sides, or a function, or something else). Denitions are important in all proofs, not just proof by induction. The denitions are even more basic than your axioms. They are your starting point on which you build. 6

Induction and Sums


2

20. Q: How can we prove that 1+3=


i=1

(2i 1)

by induction? A: This is not really an induction problem, since the left hand of this statement just means (2 1 1) + (2 2 1) and we can see this is the same as 1 + 3 just by algebra. The induction version of this problem would be to show that
n

(2i 1) = n2
i=1

This is saying that the sum of the rst n odd numbers is n2 Ill leave this for an exercise, since this is one of the projects. To do this by induction we can just work out what f (i) and g (n) are, and plug into the method given in class. But here is an alternative picture proof: The areas of each region is written in it 7 5 5 3 3 3 1 1 1 1 To get to the next square in each case, we just add the next odd number. E.g., to get to 52 , wed have to add a region of area 9 around the edge of the box. To make this into a proof still needs a bit more work, mainly that (n + 1)2 n2 = 2n + 1 This shows that the dierence between any two successive squares is an odd number. Ill leave it as an exercise to ll in the details. 21. Q: Find a formula for 11 + 22 + + nn and prove by induction. A: This looks hard! The rst few terms are: n 1 2 3 4 5 6 7 8 9 10 nth term of sum 1 4 27 256 3125 46656 823543 16777216 387420489 10000000000 sum 1 5 32 288 3413 50069 873612 17650828 405071317 10405071317

I have no idea what or whether there is a simple formula for this. Maybe there isnt! I looked up in sloanes integer sequences data base, at

http://www.research.att.com/njas/sequences/index.html which does not give any formula, (though it does say something about the ratio of the successive sums). You could investigate this sequence further as a project. 22. Is the nth term given? I dont know what this question means. The nth term of what? If youre talking about proving a formula like m(m + 1)(2m + 1) 1 + 22 + + m 2 = 6 then the nth term on the left side would be n2 , though this is not explicitly stated; youre supposed to look at the pattern, and work that out. 23. What is the sum of all negative numbers between 0 and 1000? Prove by induction. A: This is a wonderful trick question I might use this on students in the future. There are no negative numbers between 0 and 1000, and so the sum of them is 0! However, Im not sure this is what you meant? Maybe you want to nd the sum of the positive numbers between 0 and 1000? The answer is: 1 + 2 + 1000 = 1000 1001 = 500500 2

To prove this by induction, well prove the more general result P (n) : 1 + 2 + n = and then plug in n = 1000. Im not going to write out the proof by induction here, since we have done this in class. You can nd lots of examples of formulas for sums at: http://mathworld.wolfram.com/PowerSum.html try proving as many from there as you can. 24. Q: Prove by induction 1 + 2 + 3 + + (n2 + 2)2 n(n + 1)(2n + 4) = 3 4 n (n + 1) , 2

A: First we have to work out what the general term is. It says that the last term is going on:
(n2 +2)2 . 3

Lets call this f (n), and make a table of some values, to see whats

n 1 2 3 4

(n2 +2)2 3 (1+2)2 3 (22 +2)2 3 (32 +2)2 3 (42 +2)2 3

= = = =

32 3 (4+2)2 3 (9+2)2 3 (16+2)2 3

=3 = =
62 3

= 36/3

= 18 = 40 1 3 = 108

121 3 182 3

Well, this is a bit funny, since the person wrote that the rst three terms were 1, 2, 3, which does not seem to be the case. So, lets just assume the question should have been: 3 + 18 + (n2 + 2)2 n(n + 1)(2n + 4) 121 + + = 3 3 4 8

We better check this is the right result! What do we get when we add up the rst n terms? Lets write g (n) = n(n + 1)(n + 2) n(n + 1)(2n + 4) = 4 2

Lets try adding up the f (i) and comparing with the g (n): n 1 2 3 4
(n2 +2)2 3

sum of rst n terms 3 19 59 1 3 167 1 3

g (n) 3 12 30 60

3 18 40 1 3 108

This doesnt seem to work at all, except when n = 1! I wonder what the question was supposed to be? I will prove a dierent result instead of the one in the question: Theorem For all positive integers n, we have 1 3 3 + 9 + 18 + 30 + 45 + 63 + 84 + + n(n + 1) = n(n + 1)(n + 2) 2 2 Proof Note that the sum on the left side can also be written as s(n), where
n

s(n) =
i=1

3 i(i + 1). 2

1 Base case: We have s(1) = 3 2 1 2 = 3, and g (1) = 2 1(1 + 1)(1 + 2) = 3, so the statement is true for n = 1. Induction step: Assume the result for n = k . So we assume that

3 1 s(k ) = 3 + 9 + 18 + 30 + 45 + 63 + 84 + + k (k + 1) = k (k + 1)(k + 2) = g (k ). 2 2 Now we have 3 3 s(k + 1) = 3 + 9 + 18 + 30 + 45 + 63 + 84 + + k (k + 1) + (k + 1)(k + 2) 2 2 By substitution from (1), we have 3 3 s(k + 1) = s(k ) + (k + 1)(k + 2) = g (k ) + (k + 1)(k + 2). 2 2 Some algebra gives: 3 g (k ) + (k + 1)(k + 2) 2 = = We also have 1 3 k (k + 1)(k + 2) + (k + 1)(k + 2) 2 2 (k + 1)(k + 2) (k + 1)(k + 2) 1 k+ 3 = (k + 1)(k + 2) k + 3 2 2 2

(1)

(2)

(3)

1 1 (k + 1)(k + 1 + 1)(k + 1 + 2) = (k + 1)(k + 2)(k + 3) 2 2 Putting (2) (3), and (4) together gives us: g (k + 1) = s(k + 1) = g (k + 1)

(4)

25. Q: More examples would help me understand induction. A: Are there enough above? Let me know if you still need more! Would you prefer these on a handout, on the board, both? Have you looked at the examples in the text book, and these on web sites? (a) http://bigcheese.math.sc.edu/sumner/numbertheory/induction/Induction.html (b) http://www.cc.gatech.edu/people/home/idris/AlgorithmsProject/ProofMethods/Induction/UnderstandingInduction.html

Where do the formulas for sums come from?

26. Q: Where do you get formulas to prove? A: This often comes from trial and error and experimentation. Induction will not give you the formula, that has to come from somewhere else. See the next example. 27. Q: Can you use induction to nd the formula for sums, or do you only prove a known formula using induction? A: Generally, induction does not give you what the formula is. You have to try and guess that. The guess will only be a guess until you actually prove it, and induction is one way to prove your guess was correct. 28. Q: I dont understand how to prove by induction when only one formula is given. It makes sense when 1 + x2 + x3 + + n2 = formula but not when it says solve by induction, and only one formula is given. A: This is a good point, though Im not sure what example in particular you are talking about... Maybe you meant 1 + x2 + x3 + + xn = formula? Anyway, youre right, induction will never tell you what the formula is. For this, you have to try several cases, and collect some data. Then in the data, you should hopefully see a pattern. Then if your pattern is correct, you should be able to prove the result by induction. E.g., We have

1 = 1 1+3 = 4 1+3+5 = 9 1 + 3 + 5 + 7 = 16 You probably see a pattern here the sums are always squares. So you might guess that 1 + 3 + + (2n 1) = n2 Then you can prove this by induction. Finding the formula in the rst place is good practice, and helps you understand better what the formula means. 29. Q: What is the point of induction if you are already given a proved formula. A: This is a good question! (compare with Question 28 where someone complained about not being given a proved formula). There are lots of puzzles and problems for which noone yet knows the solutions. For these, if you could guess a solution, you would then have to prove it was true, and induction might be one way. Perhaps someone in this class will solve such a problem in the future. Here is an example of a simple problem that no-one yet knows a formula for:

10

How many ways are there to fold a map? That is, given an n m grid, with crease lines dividing the paper into 1 1 squares, how many ways are there to fold the map so that the resulting folded map has a 1 1 area? Heres a fun special case of this: Take some letter paper, and make creases as in the following picture. Can you fold along these in such a way that when folded, each numbered sheet lies directly on top of the sheet with the previous number?
1 4 8 5 2 3 7 6

For lots and lots of interesting sequences, take a look at Sloanes integer sequences data base. http://www.research.att.com/njas/sequences/index.html You can nd whats known about map folding there.

Induction problems involving exponents

30. Q: What is another way to write (1)n+1 ? 1n (1) = 1n A: Im not 100% sure what you mean. The answer, is of course 1. To prove this by induction, you have the correct induction step, that 1n+1 = 1 1n . If 1n = 1, then this implies that 1n+1 = 1. You just have to check the base case, 11 = 1, which is easy. Here is a similar problem: Denition: For a real number a and a positive integer n, we dene a1 a
k+1

= =

a, ak a for all positive integers k

(5) (6)

Note that this is called a recursive denition, which means that one case depends on the previous case, i.e., the denition of an+1 depends on the denition of an . Theorem: For positive integers n and m, we have am+n = am an Proof: Do this by induction on n, i.e., we x m, and prove for any xed m, the result holds for all n. I.e., for a xed m, the statement P (n) is: P (n) : am+n = am an Base case: am+1 = am a This is true by the denition, part (6), plugging in k = m. P (n) says: am+n = am an We assume this is true. We want to show that this implies P (n + 1), which says: am+n+1 = am an+1 We have am+n+1 = = = am+n a by plugging k = m + n into (6) am an a by applying P (n) am an+1 by applying (6) with k = n

This shows that P (n) P (n + 1) So by induction we conclude the result holds for all m, n. 11

31. Q: How can you make 1m+1 disappear? Take natural log. A: Im not 100% sure what you mean. Perhaps this is a fragment of a bigger statement or proof? 32. Q: Prove using induction that 5n is divisible by 5 for all n Z A: This is not true for all n Z, e.g., if n = 1, we get 51 = 1 5 , and usually we only talk about divisibility for integers. Well assume that n 1. You dont really need induction to prove this result, since 5n = 5 5n1 . For a number x to be divisible by 5 means that there is some integer y such that x = 5y So, here we just take y = 5n1 . Do you really want to force me to prove this by induction? Heres another way to prove it: Check the result for the base case n = 1. This gives 51 = 5, which is divisible by 5. Now assume that 5n is divisible by 5. So we assume we can write 5n = 5x for some x. Now 5n+1 = 5 5n = 5 (5x) Since (5x) is divisible by 5, any integer multiple of 5x is also divisible by 5x. I.e., k 5x is divisible by 5 for all integers k . In particular, this is true with k = 5. So we have that 5n+1 = 5(5x) is divisible by 5. I hope this is clear - sometimes proving something too easy by induction might make things unnecessarily complicated. 33. Q: How do you show that 7n 1 is divisible by 6 for all positive integers n? A: Since this is the same as a homework question, Ill give a similar, but slightly dierent example. Theorem For all positive integers m, we have that if m is odd, then 3m + 4m is divisible by 7. Proof We prove this by induction. If m is odd, then for some n, we have m = 2n + 1, so the The statement we want to prove for all n is P (n) : n N 32n+1 + 42n+1 is divisible by 7 Base case: P (0) says that 3 + 4 is divisble by 7, which is easy to see is true. Induction step. Suppose for some value of n, say n = k , we know that P (k ) is true. This means that there is some number qk (it depends on k what this is) such that 32k+1 + 42k+1 = 7qk Now for n = k + 1, the statement P (k + 1) will be true if we can show that 7 divides 3 2(k+1)+1 + 42(k+1)+1 , so we will try and write this in terms of 32k+1 + 42k+1 . We have 32(k+1)+1 + 42(k+1)+1 = 32k+3 + 42k+3 = = = = = = So, assuming P (k ), weve shown that P (k + 1) is true. By induction P (n) is true for all n N Note, in this question, N means the set of non-negative whole numbers. 12 9 32k+1 + 16 42k+1 9 32k+1 + (9 + 7) 42k+1 9 32k+1 + 9 42k+1 + 7 42k+1 9(32k+1 + 42k+1 ) + 7 42k+1 9 7 qk + 7 42k+1 7 9qk + 42k+1 QED.

Strong induction and double induction

34. Q: What is double induction? Could you give an example of double induction? A: I use this term to describe proofs by induction where there are two variables, and you need an induction that works for both. The statement has the form P (n, m). This is a common kind of problem. This use of the term double induction is also used in a good book on proofs, How to prove it by Daniel Velleman. Some examples with two variables are Question 35 and Question 30 below. 35. Q: When can you use strong induction? A: If you want to prove a statement P (n), sometimes its not enough to assume just P (n 1). Here is a problem where you can apply this method:
Example: Whats the least number of breaks needed to break a n m bar of chocolate into individual 1 1 pieces? E.g., for 4 5, we cen do it in 19 steps as below. Can you do it in any less? How many?

3 breaks

4 4 = 16 breaks

Theorem Given an n m rectangle with n, m positive integers, and lines dividing the rectangle into 1 1 squares, then however the rectangle is cut into individual squares, it requires exactly nm 1 cuts. Proof Base case: for m = n = 1 no cuts are needed, and mn 1 = 0. You cant have less than 0 cuts, so this is the least number necessary. So the statement of the theorem is true when n = m = 1. Induction step: Given an n m rectangle. Assume the result is true for k1 k2 rectangles whenever k1 k2 < nm. How can we break the n m bar? There must always be a rst break, which splits the rectangle into two parts. We have two possible cases, depending on which direction the rst break is:
m m

r n n s r s

Each case is the same, so Ill just discuss the rst. We now have made one cut, and have rectangles of sizes n r and n s, with r + s = m. Now by the inductive hypothesis, we need exactly nr 1 and ns 1 cuts to break these smaller rectangles up. So, the total number of breaks must be: 1 + (nr 1) + (ns 1) = nr + ns 1 = n(r + s) 1 = nm 1 So, by assuming the result for smaller rectangles, weve also proved it for the n m rectangle. So by induction the result is true for all integers sided rectangles. Notice that here we had to use strong induction, since the rst break might be anywhere, we dont know, so we have to assume the result for all smaller rectangles, not just those of size n (m 1) for example. There is a java applet to describe this problem, at the bottom of the page at: http://www.cut-the-knot.com/ctk/memes.shtml click on lines to make the rectangle split along that line. This site gives more explanation at: http://www.cut-the-knot.com/ctk/chocolate.shtml Another example of strong induction is given in the last proof in Question 37, though in that case proof by usual induction is also sucient, as the rst example shows. 13

Other examples of proof by induction

36. Q: Prove by induction that the sequence of numbers in the rst diagonal of Pascals triangle is 1, 2, 3 , I.e., an,1 = n. 37. Q. How can you prove a triangle has 180 using induction? (If you can at all) A. Make sure to be precise in your questions. I think you mean how can you prove that the sum of the internal angles in a triangle is 180 ? This is not really a question to prove by induction, since there is no varying number n. This statement would be proved by using methods of geometry. There are several dierent ways to prove this. One method is based on the following picture: Add a line parallel to one side, through opposite vertex

There are many web sites with more details, e.g.,: http://mathworld.wolfram.com/Triangle.html http://www.mathsrevision.net/gcse/pages.php?page=17 Also, you will probably see a complete proof of this result in a geometry course, such as MATH4005. There might be a way to prove this by induction, but I cant think of one, because I cant easily see where the n would be. To make this into an induction question, you need to make a statement about an innite sequence of things. Here is one possibility: Theorem: The sum of the internal angles in an n-gon add up to 180(n 2) Proof: The base case is n = 3, which is a result about triangles. This result is true for triangles by the argument given above. Note that to prove the base case you cant use induction; this is part of the induction, and might need various methods. Mostly the base case is fairly simple, but this is not always true. How to go from n = k to n = k + 1: Assume the result is true for all k -gons. This is the inductive hypothesis. Give me any k -gon I can split a triangle o it:

10-gon split into a 9-gon and a triangle

By induction, the result is true for the k 1-gon, and we know its true for the triangle. By induction the k 1 gon has internal angles summing to 180(k 3). The angles of the triangle sum to 180 So the internal angles of the k -gon sum to 180(k 3) + 180 = 180(k 2) So the result follows by induction. 14

We can use strong induction to prove the same result: Theorem: The sum of the internal angles in an n-gon add up to 180(n 2) Proof: The base case is n = 3, which is a result about triangles. This result is true for triangles by the argument given above. Note that to prove the base case you dont use induction; this is part of the induction, and might need various methods. Mostly the base case is fairly simple, but this is not always true. How can we prove the statement for n = k if we know it for smaller values of n? Assume the result is true for all r-gons for r < k This is the inductive hypothesis. Give me any k -gon. Break it into two smaller polygons, e.g.:

10-gon split into a 5-gon and a 7-gon

Its always possible to do this if the polygon has more than 3 sides. The smaller polygons have k1 sides and k2 sides. Since k1 < k and k2 < k , the result is true for these polygons. So they have internal angles adding up to 180(k1 2) and 180(k2 2). So, the polygon with k sides has sum of internal angles 180(k1 2) + 180(k2 2) = 180(k1 + k2 4) Now notice that k1 + k2 = k + 2, which is because there is an extra edge on the two smaller polygons which was not an edge of the initial polygon. So, by substitution, the sum of internal angles of the initial polygon is 180(k 2). So, weve proved by induction that the result always holds. Heres a completely dierent proof, without using induction at all: Theorem: The sum of the internal angles in an n-gon add up to 180(n 2) Proof: You can divide the n-gon up into n 2 triangles:
10-gon split into 8 triangles

Assume the result for triangles. Then the total sum of the internal angles is sum of angles in a triangle number of triangles = 180(n 2). 38. Q: Can you prove the minimum number of steps to move n disks for the tower of Hanoi from one peg to another? A: Yes, but since these notes are getting rather long, Im going to refer you to some web pages: http://www.cl.cam.ac.uk/Teaching/1998/InroAlgs/notes98/node3.html http://www.cut-the-knot.com/recurrence/hanoi.shtml

15

I dont understand

39. It needs more explaining, still unsure how its done. 40. Im still very confused on the steps of induction. I hope some of the above examples help. Please read through everything on these sheets carefully, and let me know where you get stuck. Try and locate the step you dont understand.

10

Comments on group work

The comments about group work: For 1. Groups 2. I like groups 3. I like working in groups because we put our heads together against 1. More examples on board 2. Need more work on board 3. Too many groups 4. I think were doing too many groups conclusion: Perhaps this means I should do more work on the board! I will give at least one class next week with no group work.

16

Вам также может понравиться