You are on page 1of 3

Appendix 1

The Real Numbers


For the practical purposes of computation, nite decimals, which are only a small part of the set of rational numbers, are more than adequate; since computers get along with only a small part of the set of nite decimals. But in trying to discover useful results, restriction to even the rationals would prove extremely clumsy because there would be no numbers representing most square roots, and no number representing the ratio perimeter/diameter of a circle. With only the rationals we would be hobbled in our search for new results. The basic idea behind the extension of our system of numbers to include these reals, is to use pairs of rationals to zero in on any point on a line. For example, to narrow down on 2 , we use the innite decimal 1.414213562373095.... by which we mean the sequence of intervals I 1 = ( 1, 2 ) I 2 = ( 1.4, 1.5 ) which pins down
2

2 to the numbers between 1 and 2


2

(because 1 < 2 and 2 < 2 ) which pins down


2

2 to the numbers between 1.4 and 1.5


2

(because 1.4 < 2 and 2 < 1.5 ) I 3 = ( 1.41, 1.42 ) , etc. The basic properties of these intervals are Each interval is part of the previous one. As n grows large, the length of the n-th intervals gets close to 0. (In the decimal case, the length of the n-th n1 interval above is 1 10 ; in the binary case used in n1 computers, the length would be 1 2 .)

A sequence of intervals with the properties above is called a nested sequence of intervals. Each nested sequence determines a single point on a line, and for us, a real number. One problem: a single point can be specied by lots of different nested sequences like 1.999... and 2.000... both determine the same point (the real number 2). So, we need a means of deciding when two nested sequences

1999 by CRC Press LLC

(which you can think of as innite decimals) stand for the same number. (We had this same type of problem for rationals, where two different fractions could represent the same rational number, e.g., 1/2 and 3/6.) Its simple. Two nested sequences represent the same number if each interval from one of the sequences overlaps with all intervals from the other sequence. Denote two such sequences by R and S. If one of the intervals in R lies completely to the left of any of the intervals of S, then R < S , as illustrated in Figure A1-1.

S
Figure A1-1

If the intervals from one of the sequences always overlap with the intervals from the other sequence, then the two nested sequences must zero down on the same point, and therefore stand for the same real number. The old rules for inequality still hold i.e., if R < S and S < T , then R < T , etc.

We need to know how to add and multiply real numbers. For the positive reals (those representing points to the right of 0, if the n-th interval S n of S is written S n = ( S nL, S nR ) , that of T as T n = ( T nL, T nR ) , then the nth interval of the sum S + T is ( S nL + T nL, S nR + T nR ) and of the product S T is ( S nL T nL, S nR T nR ) . A real R represents a rational r if r is in every interval of R. Its known that nested sequences can represent numbers such as 2 , which arent rational.1 If we now form nested sequences of reals, its easy to
1. See Modern University Calculus, by Bell, Blum, Lewis and Rosenblatt, [1]

1999 by CRC Press LLC

see that we dont create any additional numbers because given any nested sequence of reals, we can create a nested sequence of rational intervals, equal to it. Just make the corresponding rational intervals a trie bigger, as illustrated in Figure A1-2 (the intervals with rational end-points indicated as darker).

Figure A1-2

You might want to think about just how you would accomplish this. (Hint: Each nested sequence of reals has each of its end-points determined by a nested sequence of rational intervals. As the rational end-point associated with one of the left real end-points, use one of the left rationals used to dene this left real end-point going far down enough in this sequence, so that this rational is getting very close to the real left end when you get far down the real sequence.) Because repeating this zeroing in process which created the reals from the rationals doesnt enlarge the reals, we say that the reals are complete . Completeness will be an important property of other mathematical structures which arise from a similar extension (e.g., the set of Lebesgue integrals, and the set of generalized functions). In many situations we will solve a problem using a process of bisection, to settle down on a number with some property. This could have been done to nd 2 , as follows: 1< 2<2 1 < 2 < 1.5 1.25 < 2 < 1.5 . . and then showing that the real number we constructed has the property we wanted in this illustration, calling this number R, that R 2 = 2. The real numbers have other properties which are important for deriving essential results, which we introduce in the main body of the book as needed.

1999 by CRC Press LLC

You might also like