1
00:00:00,499 --> 00:00:03,260
PROFESSOR: So let's look at
a last example of applying

2
00:00:03,260 --> 00:00:07,290
the well ordering principle,
this time to something that we

3
00:00:07,290 --> 00:00:11,410
actually care about-- a theorem
that really does require some--

4
00:00:11,410 --> 00:00:14,780
The theorem is the
following famous formula

5
00:00:14,780 --> 00:00:17,730
for the sum of a geometric
sum-- for a geometric series

6
00:00:17,730 --> 00:00:18,980
or a geometric sum.

7
00:00:18,980 --> 00:00:22,540
So the numbers on the left,
the powers of r starting at 1,

8
00:00:22,540 --> 00:00:26,090
which is r to the 0 followed
by r, which is r to the 1,

9
00:00:26,090 --> 00:00:29,600
followed by r squared, up
through the nth powerful of r.

10
00:00:29,600 --> 00:00:31,940
You add all those numbers
up and it turns out

11
00:00:31,940 --> 00:00:34,890
that there's a nice,
simple, fixed formula that

12
00:00:34,890 --> 00:00:37,700
doesn't have those
three dots in it that

13
00:00:37,700 --> 00:00:41,130
tells you exactly what the value
of that sum is in the formula.

14
00:00:41,130 --> 00:00:45,230
As you can read is r to the n
plus 1 minus 1 is the numerator

15
00:00:45,230 --> 00:00:48,630
and r minus 1 is
the denominator.

16
00:00:48,630 --> 00:00:51,610
And the claim is
that this identity

17
00:00:51,610 --> 00:00:55,650
holds for all
non-negative integers n

18
00:00:55,650 --> 00:00:58,820
and for all real numbers r
that aren't 1 because I don't

19
00:00:58,820 --> 00:01:02,610
want the denominator to be 0.

20
00:01:02,610 --> 00:01:04,110
So how are we going
to prove this?

21
00:01:04,110 --> 00:01:06,240
Well, I'm going to
prove it by using

22
00:01:06,240 --> 00:01:08,850
the well ordering
principle, and let's suppose

23
00:01:08,850 --> 00:01:13,190
that this identity didn't hold
for some non-negative integer

24
00:01:13,190 --> 00:01:14,340
n.

25
00:01:14,340 --> 00:01:17,420
So we'll apply the
well ordering principle

26
00:01:17,420 --> 00:01:20,660
and we'll let m be
the smallest number

27
00:01:20,660 --> 00:01:26,470
n where this equality fails--
it becomes an inequality.

28
00:01:26,470 --> 00:01:28,920
Now, what I know
about m immediately

29
00:01:28,920 --> 00:01:33,340
is that this equality, if
you look at it, when n is 0

30
00:01:33,340 --> 00:01:35,860
the left-hand side comes down.

31
00:01:35,860 --> 00:01:38,620
It degenerates to
just r to the 0, or 1.

32
00:01:38,620 --> 00:01:40,520
The right-hand side,
if you check it,

33
00:01:40,520 --> 00:01:44,670
is r minus 1 over r
minus 1, which is also 1.

34
00:01:44,670 --> 00:01:49,220
So equality holds
when n is 0, and that

35
00:01:49,220 --> 00:01:52,870
means that the least m for
which equality doesn't hold

36
00:01:52,870 --> 00:01:54,880
has to be positive.

37
00:01:54,880 --> 00:01:58,220
So, what we know
about the least number

38
00:01:58,220 --> 00:02:01,760
where this equality fails
is that it's positive.

39
00:02:01,760 --> 00:02:05,160
And that means in particular
since it's the least one

40
00:02:05,160 --> 00:02:08,550
[? word ?] fails, if you
go down one to m minus 1,

41
00:02:08,550 --> 00:02:10,110
the equality holds.

42
00:02:10,110 --> 00:02:15,660
So we can assume that the sum
of the first m powers of r,

43
00:02:15,660 --> 00:02:18,920
starting at 0 and ending
at r to the m minus 1,

44
00:02:18,920 --> 00:02:24,530
is equal to the formula where
you plug in m minus 1 for n

45
00:02:24,530 --> 00:02:26,610
and you get that formula
on the right, which

46
00:02:26,610 --> 00:02:29,010
I'm not going to read to you.

47
00:02:29,010 --> 00:02:30,710
Well, we can simplify
it a little bit.

48
00:02:30,710 --> 00:02:34,980
If you look at the exponent,
r to the m minus 1 plus 1

49
00:02:34,980 --> 00:02:38,230
is after all just r to the m.

50
00:02:38,230 --> 00:02:40,570
So repeating what I've
got is that the sum

51
00:02:40,570 --> 00:02:45,150
of those first powers of r
up to m minus 1 we can assume

52
00:02:45,150 --> 00:02:50,090
is equal to the formula r to the
m minus 1 divided by r minus 1

53
00:02:50,090 --> 00:02:55,260
because m failed and this
was the number one less where

54
00:02:55,260 --> 00:02:57,290
it had to succeed.

55
00:02:57,290 --> 00:02:59,540
So now we take the
obvious strategy.

56
00:02:59,540 --> 00:03:01,530
What I'm interested
in is properties

57
00:03:01,530 --> 00:03:05,159
of the sum of the
powers up to r to the m.

58
00:03:05,159 --> 00:03:06,700
Now, the left-hand
side is the powers

59
00:03:06,700 --> 00:03:08,910
up to r to the m
minus 1, so there's

60
00:03:08,910 --> 00:03:12,360
an obvious strategy for turning
the left-hand side into what

61
00:03:12,360 --> 00:03:13,230
I'm interested in.

62
00:03:13,230 --> 00:03:17,750
Namely, let's add r to
the m to both sides.

63
00:03:17,750 --> 00:03:20,240
So the left-hand side
becomes just the sum

64
00:03:20,240 --> 00:03:23,000
that I want and the
right-hand side becomes

65
00:03:23,000 --> 00:03:25,560
this messy thing, r to the
m minus 1 over r minus 1

66
00:03:25,560 --> 00:03:27,460
plus r to the m.

67
00:03:27,460 --> 00:03:29,980
Well, let's just
simplify a little bit.

68
00:03:29,980 --> 00:03:35,140
Let's put r to the m over
the denominator, r minus 1,

69
00:03:35,140 --> 00:03:38,710
which I do by multiplying
it by r minus 1.

70
00:03:38,710 --> 00:03:41,360
And then it comes
out to be r to the m

71
00:03:41,360 --> 00:03:44,150
plus 1 minus r to
the m over r minus 1.

72
00:03:44,150 --> 00:03:47,670
And I collect terms
and look what I got.

73
00:03:47,670 --> 00:03:51,380
I've got the formula r
to the m plus 1 minus 1

74
00:03:51,380 --> 00:03:55,070
over r minus 1, which means
that the identity that I was

75
00:03:55,070 --> 00:03:58,010
originally claiming,
in fact, holds

76
00:03:58,010 --> 00:04:02,835
at m contradicting the assertion
that it didn't hold at m.

77
00:04:02,835 --> 00:04:04,710
In other words, we've
reached a contradiction

78
00:04:04,710 --> 00:04:08,730
assuming there was a least
place were equality fails,

79
00:04:08,730 --> 00:04:11,380
that means there's
no counter example

80
00:04:11,380 --> 00:04:14,840
and the equality holds for
all non-negative integers n.

81
00:04:18,190 --> 00:04:21,010
So here's the
general organization

82
00:04:21,010 --> 00:04:23,260
of a well ordering proof
which we've been using.

83
00:04:23,260 --> 00:04:25,540
Let's just summarize it
into a kind of template

84
00:04:25,540 --> 00:04:26,800
for proving things.

85
00:04:26,800 --> 00:04:31,100
So what you have in mind is that
there's some property, P of n,

86
00:04:31,100 --> 00:04:32,176
of non-negative integers.

87
00:04:32,176 --> 00:04:34,050
And what you'd like to
prove is that it holds

88
00:04:34,050 --> 00:04:36,370
for every non-negative integer.

89
00:04:36,370 --> 00:04:41,400
So for all n in non-negative
integers, P of n holds.

90
00:04:41,400 --> 00:04:43,410
And we're going to
try to prove this

91
00:04:43,410 --> 00:04:45,030
by the well ordering
principle, which

92
00:04:45,030 --> 00:04:49,940
means that we're going to define
the set of numbers for which P

93
00:04:49,940 --> 00:04:51,159
doesn't hold.

94
00:04:51,159 --> 00:04:52,700
That is, the set of
counter examples,

95
00:04:52,700 --> 00:04:57,070
and call that C. So C is the
set of non-negative integers

96
00:04:57,070 --> 00:04:59,880
for which not P of n holds.

97
00:04:59,880 --> 00:05:03,350
Now, by the well
ordering principle

98
00:05:03,350 --> 00:05:05,290
there's got to be
a minimum element,

99
00:05:05,290 --> 00:05:11,600
call it m, that's in C.
And at this point the job--

100
00:05:11,600 --> 00:05:14,310
by assuming that m is the
smallest counter example,

101
00:05:14,310 --> 00:05:17,660
we have to reach a
contradiction somehow.

102
00:05:17,660 --> 00:05:23,320
Now, up to this second
bullet it's the template,

103
00:05:23,320 --> 00:05:26,200
but the third bullet is where
the real math starts and there

104
00:05:26,200 --> 00:05:27,730
isn't any template anymore.

105
00:05:27,730 --> 00:05:29,690
How you reach a
contradiction is by reasoning

106
00:05:29,690 --> 00:05:34,170
about properties of P of n,
and there's no simple recipe.

107
00:05:34,170 --> 00:05:36,760
But the usual organization
of the contradiction

108
00:05:36,760 --> 00:05:38,550
is one of two kinds.

109
00:05:38,550 --> 00:05:41,810
You find a counter example
that's smaller than m--

110
00:05:41,810 --> 00:05:44,006
you find a C that's in the
set of counter examples

111
00:05:44,006 --> 00:05:46,505
and C is less than m that would
be a contradiction because m

112
00:05:46,505 --> 00:05:50,080
is the smallest thing at C.
Or you reach a contradiction

113
00:05:50,080 --> 00:05:53,440
by proving that P does
hold for m, which means

114
00:05:53,440 --> 00:05:55,060
it's not a counter example.

115
00:05:55,060 --> 00:05:58,060
And those are the
two standard ways

116
00:05:58,060 --> 00:06:02,241
to organize a well
ordering proof.