9

The Structure of the Real Number System

"The integers God made; all the rest is the work of man. " L. Kronecker.

The invention of the calculus presented the mathematical world with a severe
difficulty. No one was in doubt of its usefulness, nor that it could produce many
correct results that could not be obtained by other methods, yet the foundations of
the calculus were unsatisfactory. There was some doubt as to the exact nature of a
limit, and proofs of various results left much to be desired: indeed, even from the
earliest times of the calculus, mathematicians had produced 'proofs' of results which
were visibly false. The calculus became a matter of some controversy, well
exemplified in Berkeley's book, The Analyst, or A Discourse Addressed to an infidel
Mathematician, published in 1734; this is still worth reading.

As is human nature, mathematicians, when faced with the apparently insoluble

problem of putting calculus on a rigorous foundation, tried to ignore the matter, and
until the end of the eighteenth century calculus was accepted largely as a matter of
faith. By that time, however, the issue had become pressing, and major progress was
made in the early nineteenth century, with the rigorous development of the idea of
limit and its consequences. The stumbling block at last became apparent; there was
no precise statement of the properties of the real number system. This was rectified
around 1870 and the modern subject of analysis assumed increasing importance
thereafter.

To an extent, we have followed a similar path. We introduced the idea of limit and
deduced its consequences, but we noticed that there were results we needed to
assume. For example, we did not prove the theorem that a bounded increasing
sequence must tend to some real limit, but deferred this issue. The important thing is
to appreciate that there is an issue here; no matter how plausible the statement may
seem, it does require justification unless it is to be accepted as a basic premise. If we
accept the property as true, for the moment, then it provides a distinction between the
set of real numbers and the set of rational numbers. To see this, notice that if s„ = 1
+ 1/1! + 1/2! + .. + 1/n! , then (s„) is increasing and bounded above by 3 but the
limit, the number usually called e , is not rational even though all the terms s aren

rational. (These assertions are the substance of Problems 8.11 and 8.20.)

To cope with this task we need to state clearly what we are assuming about Κ . In
fact the result we wish (that an increasing sequence of real numbers which is
bounded above tends to some real limit) cannot be proved from the assumptions we
have so far made, either explicitly or tacitly. These are what we rather glibly
described in Chapter 3 as "the normal properties of the number system" and the four
112 The Structure of the Real Number System [Ch. 9

order properties at the beginning of Chapter 5, which for convenience we shall re­
state formally here. These we shall accept as given; the formal word is axioms.

AXIOMS
Al For all real χ , y , and ζ , χ + (y + ζ) = (χ + y) + ζ and x(yz) = (xy)z •
Α2 For all real χ and y , x + y- y + x and xy = yx .
A3 There is a real number 0 such that, for all real χ , x + 0 = 0 + x = x.
A4 For all real numbers χ , there is a real number y for which
x+y=y+x=u.
A5 There is a real number 1 such that, for all real χ, x\ = 1 χ = χ .

A6 For all real χ Φ 0 , there is a real y for which xy - yx = 1 .

A7 For all real χ , y , and ζ , x(y + ζ) = (xy) + (xz).

A8 1 * 0.
A9 For all real χ , y , and ζ , exactly one of the three statements χ < y , χ = y ,
χ > y holds.

A10 For all real χ , y , and ζ , if x < y and y <z then χ < z •

Al 1 For all real χ, y , and ζ , if x < y then χ + z < y + z •

A12 For all real x ,y , and ζ , if x < y and z > 0 then xz < yz •
These axioms, however, are also true of the system of rational
numbers, so any result which can be deduced from them alone
must be true of the system of rational numbers and, indeed, of any
other system satisfying Al - A12. Since we know that a bounded
increasing sequence in the system of rational numbers need not
have a limit within that system, the property we wish is not true of
Q and, therefore, not deducible from Al - A12 alone.

We have already stated that R is a set with various arithmetical

properties and our picture of the real number system is that it
corresponds to the points of a line. We therefore wish a property
which would allow us to replace the following with something
3
more precise (see Fig. 9.1 ): The function given by y - χ - 2 has
a negative value at χ = 1 and a positive one at χ = 2, so, since the
curve has no discontinuities, it "must" cross the axis at some point.
This assertion that the curve "must" cross the axis is not so much a
proof as an act of defiance ("you tell me how else it could change
from being negative to being positive"). Fig. 9.1

The feature we shall make precise is the idea that there are no gaps in the set of real
numbers, as we think of there being no gaps in a line. (There certainly are gaps in
the set of rational numbers and lies in one of them.) Looking more closely, we
see that we can choose two related subsets of R , {JC e R-.x* -2 < 0} and
Ch. 9] The Structure of the Real Number System 113

3
{JC e R: JC - 2 > 0 } , with the property that every member of the first set is less than
every member of the second set. The existence of the number Λ/2 is equivalent to
the existence of a real number which marks the division between these two sets.
This corresponds to a picture of dividing a line into two: if we divide the line into
two parts, one lying entirely to the left of the other, then there must be a point which
marks the division, and this point must lie in one portion or the other (see Fig. 9.2).
This is what motivates the definition below. If A and Β are two sets, we define
their union , AuB , to be the set {JC: JC € A or JC e B] and their intersection , A n S ,
to be {JC: JC e A and JC e B) .

Η
Fig 9.2

Axiom of Continuity (Dedekind's Axiom)

The set R has the property that if R = A<uB where A and Β are both non-empty
and for all jce A, and all y e Β , JC < y , then either A has a greatest element or Β
has a smallest.

Notice that the conditions ensure that A and Β have no elements in common, and
that the axiom guarantees that real certain numbers exist.

Notation: Let a, b e R and a< b . We define the closed interval [a, b] to be the
set {jce R: a<x<b] and the open interval (a, b) to be {jce R:a<x<b}.
(a, b] and [a, b) have their obvious analogous meanings. For convenience we use a
similar notation for {JCe R: a < JC}, denoting this by (-OO, a] and {JCe IR : JC>a]
by (α, T»), with the corresponding interpretation of (a, OO) and [a, OO). As before , O
O
is used as a notation, not a number.

Theorem 9.1 Let R = AKJB where A and Β are both non-empty and for all
xe A , and all y e Β , JC < y . Then there is a real number a such that either
A - (-OO, a] and Β - (α, OO) or A = (-OO, a) and Β = [a, OO).

Proof: By the axiom of continuity, either A has a greatest element or Β has a

smallest.

First, suppose A has a greatest element, a . Then a e A and VJC e A , JC < a so A

contains only numbers belonging to (-OO, a]; we need to show A contains all such
numbers. Let y e R , y < a. Since y e R , y e A or y e B. But y e B=>a<y
(since a e A, by the condition given connecting A and B), so this is impossible.
Thus ye A so A = (-OO, a], since y was a typical number in (-«>, ] . Since a

B={xe R : JC ί A} = {JC e I R : j c i a } , Β = (α, OO). (That>5={jce Κ. : χ « Λ}

uses our knowledge that Au/? = R and AnZi = 0 . )

The case when β has a smallest element is similar. •

114 The Structure of the Real Number System [Ch. 9

In practice, the Axiom of Continuity is inconvenient to use, so we must develop a

more readily applicable tool.

Definitions A subset S of Κ is said to be bounded above if there is a number u

with the property that Vx e S , χ < u . Such a number u is called an upper bound
for 5 . Notice that if u is an upper bound for S so are u + 1, u + 2, u+j,.... so a
set which is bounded above will have many upper bounds.

Example 9.1 Show that the number u is not an upper bound for the set S if and
only if 3x e 5 such that x> u .

Solution: χ is not an upper bound for S if and only if it is false that V x e S, x< u ,
that is, if and only if 3x e S such that χ > u . In other words, u satisfies the
negation of the statement that it be an upper bound for 5 .

This is a useful property, well worth remembering, for it gives us a positive property
of numbers which are not upper bounds of a given set. By similar reasoning the
number / is not a lower bound of the set S if and only if 3x e S such that χ < I .•

Theorem 9.2 Let S be a non-empty subset of I R which is bounded above. Then

among all of the upper bounds of S there is a smallest upper bound, i.e. the set of
upper bounds for S has a smallest element.

Proof: Let Β be the set of all upper bounds for S ; we know Β Φ 0 . Let A be the
set of all real numbers which are not upper bounds for S . From the definitions it is
obvious that A u S = . Since S Φ0 , there is an element s e S, whence s - 1 is
not an upper bound for S and Α Φ 0 .

To apply the Axiom of Continuity we need to show that every element of A is less
than every element of Β . Let x e A and y e B. Since χ is not an upper bound
for S , there is an element, say s , of 5 such that χ < s . Since y is an upper bound
for S , s <y. Thus χ < y . Because χ and y were typical elements of A and Β
respectively., we have shown that VJC e A and Vy e Β χ < y .

By the Axiom of Continuity, either A has a largest element or Β has a smallest. In

fact, A cannot have a largest element. Let a e A . Then a is not an upper bound
for 5 so there is an element of S , say s , such that a < s . Also a < (a + s)/2 < s
so (a + s)l2 is not an upper bound for 5 , (because it is less than the element s of
5), and therefore (a + s)/2 e A . This shows that a is not the greatest element of A.
Since a was a typical element of A , A can have no greatest element. By the
Axiom of Continuity Β has a smallest element. This is what we require. •

Comments: 1. This result does tell us something significant. The set (0,1) has no
smallest element even though all its members are greater than 0 , so not all bounded
sets have a smallest element. The Theorem tells us, however, that those sets which
are the set of all upper bounds of a non-empty set are special in that they must have a
smallest element.
Ch. 9] The Structure of the Real Number System 115

2. We do not claim that S has a largest element, but if it does, this element is the
smallest upper bound. (For the largest element of a set is an upper bound and no
smaller number can be an upper bound by Example 9.1.)

Example 9.2 The smallest upper bound for [0,1] is 1, and the smallest upper
bound for (0,1) is also 1. Notice that 1 belongs to one of the sets but not the other.
The smallest upper bound of the set ( y , -|, j,...} is 1 .

Solution: In all three cases 1 is an upper bound for the set. Because 1 e [0, 1 ] no
smaller number can be an upper bound, so in the first case we see that 1 is the least
upper bound. In the second case if we choose a number χ < 1 then there is an
element of (0, 1) which is greater than χ (for example 1/2 if χ < 1/2, or (x + l)/2
if χ > 1/2), so again we see that no number smaller than 1 can be an upper bound
for (0, 1 ) . In the last case again if we choose χ < 1 then there is an element of the
set, in this case a number of the form n/(n + 1) = 1 - l/(n + 1) for large enough η
which is greater than χ , showing again that χ is not an upper bound for the set.
(This uses the Archimedean property and will be dealt with more fully in the next
few pages.) •

Definition Let 5 be a non-empty set which is bounded above. We call the smallest
upper bound for S the supremum of S and denote it by sup S.

Lemma 9.3 Let S be a subset of I R . Then s = sup S if and only if

(i) VJC e S, χ < s , and (ii) Vs' < 5 3x e S such that s ' < χ .

Proof: Statement (i) says that * is an upper bound for S and (ii) that no smaller
number is an upper bound. A moment's thought will show that a number possesses
both properties if and only if it is the smallest upper bound for S . •

The two conditions of Lemma 6.3 offer a practical method for establishing that a
particular number is the supremum of a given set. Notice that there are two
conditions to be checked and that this complication is intrinsic. The supremum need
not be a member of the set concerned; matters would be greatly simplified if this
were so, for then every set with an upper bound would have a greatest element, but it
just is not true. Let us use these ideas to prove one of our deferred results.

Theorem 9.4 Let (a„) be an increasing sequence of real numbers which is bounded
above. Then there is a real number a such that a —> a as η —» °°,
n

Proof: We know that [a : n e N ]

n is bounded above; it is non-empty since it
contains a\. Let a = sup {a : n e Ν ) .
n

Let ε > 0 . Since a - ε < a , by Lemma 9.3 there is an element of {a : n e Ν ] which

n

is greater than a - ε ; let this element be a . . Then, since (a„) is increasing we see
N

that Vn > Ν α-ε<α Ν < a < a , the last inequality being true since a = sup [a : η
n n

e Ν } . Therefore, Vn > Ν \a - a\ < ε . Since ε > 0 was not further specified, this
n

holds for all ε > 0 and we have shown that a -> a as η —» °° . •

n
116 The Structure of the Real Number System [Ch. 9

Notice that Lemma 9.3 allows us to deduce the existence of at least one element of
the set {α„: n e Ν} greater than a - ε ; we do not know whether or not a - ε itself
belongs to the set.

This leaves us in a position to tackle the detailed structure of R . We base all our
deductions on the assumptions we have made about R , which are that R satisfies
the arithmetical and order axioms A1-A12 stated above and that R obeys the axiom
of continuity. By the arithmetical properties, we can prove that R contains all
rational numbers (e.g. 2/3=(l + 1)/(1 + I + 1)), whose construction uses the
arithmetical properties). To show that R contains irrational numbers we need the
axiom of continuity, which we use indirectly. Define a sequence (a„) by a\ = 2, and
Vne Ν a n+] = \[an

Then, following earlier work in Example 7.7, we see that (a„) is decreasing and
bounded below, so there is a real number a such that a„ —> a as η —» oo . (This
2
uses Theorem 9.4 above, which we have now proved.) As before, a = 2 so a is
not rational. This guarantees the existence of a real number whose square is 2. That
is, our assumptions are enough to prove formally, and without appealing to anything
other than these assumptions, that irrational numbers exist.

The rational numbers are distributed throughout R and are thoroughly intermingled
with the irrational numbers. Before showing this we prove the Archimedean
property, which we previously had to assume.

Theorem 9.5 The set of natural numbers is not bounded above. In other words, for
every real number χ, there is a natural number η for which η > χ .

Remarks: This result may seem obvious. It certainly is obvious that Ν has no
greatest element, though that on its own does not preclude the existence of an upper
bound. Theorem 9.5 can be viewed as showing that R contains no "infinitely
large" numbers and is a form of confirmation that there are no further additional
properties which we shall have to assume. Though it need not concern us here, there
do exist systems satisfying A1-A12 for which the set corresponding to Ν is
bounded above.

Proof: Suppose M were bounded above. Then, since Ν * 0 , there must be a real
number which is the supremum of N; call it s . Then, by the properties of the
supremum, we see that since s - j < s , there is a number n e Ν with s - \ <η .

Then η + \ > s + j > s , which is a contradiction since η + 1 e Ν , but

η + 1 > sup Ν . This contradiction shows Ν cannot be bounded above.

The second part of the statement is just the observation that if x e R then χ is not an
upper bound for Ν . •

Corollary If x e R and χ > 0 then there is a natural number η with \ln<x .

Ch. 9] The Structure of the Real Number System 117

Proof: Since χ > 0 , 1/JC > 0 so by Theorem 9.5 l i e Ν such that η > \lx . For
this η , 1/n < χ. •

This result shows us that if χ is positive, no matter how small, then if we add χ to
itself sufficiently often, the result is greater than 1 : this just notices that 3n e Ν such
that nx > 1 and nx = χ + χ + ... + x. This could be interpreted as showing that there
are no 'infinitesimally small' elements of I R .

Lemma 9.6 If χ and y are real numbers and χ < y , there is a rational number q
with χ < q < y . (Loosely, between every two reals there is a rational number.)

Proof: S ince y — χ > 0 there is N e Ν with UN < y — χ. We now show that for
some integer m , m/N is between χ and y .

Since Nx is real, there is an m e Ν for which m\ > Nx so that χ < m /N. Since
{ t

-Nx e IR , there is an m e Ν for which m > -Nx , hence χ > -m^N. The set
0 0

A = {m/N: m e Ζ , -mo < m< m\) is a finite set which contains an element greater
than χ and another less than χ . If m IN is the smallest member of A which is
2

greater than χ then (m - 1 )IN < χ . (This number exists since every finite set of real
2

numbers has a minimum.) Let q = m /N so q e Q , χ <q and q-\IN <x . Thus

2

q < χ + UN < χ + (y - x) whence χ < q < y. •

Obviously, we can repeat the process just carried out to produce a rational number
«ι with χ < Οι < q < y and repeat again to produce as many rationals as we wish
between χ and y . There are, therefore, infinitely many rationals between χ and y,
since if this were not so there would be some maximum number of rationals which
could be found between χ and y .

Corollary 1 If JC e I R then JC = sup [q e Q : q < JC).

Proof: Let A = {q e Q : q < χ ) . We show the result by using Lemma 9.3.

Obviously, Vy e A , y < χ. Also if y' < JC then by Lemma 9.6 there is a q e Q
with y ' < q < χ , so 3q e A such that y' <q . Therefore, y' is not an upper bound for
A , and by Lemma 9.3 JC = sup A . •

Corollary 2 Every real number is the limit of a sequence of rational numbers.

Proof: Let a e I R . For each n e Ν , there is a rational number a with n

α-]/η <a„ < a , by the Lemma. Then Vn e Ν \a - a\ < 1/n so that a„ —> a as
n

η —» oo . •

In the foregoing work we have introduced and used the idea of the supremum of a
set. This is an important tool, so we need to obtain some skill in using it. Notice
that the definition of the supremum as the smallest upper bound is somewhat
complicated, and the equivalent statement in Lemma 9.3 has two parts. It is usual to
have to establish these two properties of the supremum separately.
118 The Structure of the Real Number System [Ch. 9

Example 9.3 Let A be a non-empty subset of R which is bounded above, and let
Β = {2x: χ e A) = [y: y/2 e A) . Then sup Β = 2sup A .

Solution: Let a = sup A . We need to show that 2a has the two properties required
by Lemma 9.3.

Let y e Β . Then y - 2x for some χ e A , so y/2 e A . Thus y/2 < a (as

a = sup A), and y < 2a . Therefore Vy e Β , y < 2α .

Let b' <2a . Then b'12 <a . Since α = sup Λ , from the second part of Lemma 9.3,
3x e A such that 072 < JC and so b' <2x . Thus 3y e β (y = 2JC before) such that
b'<y. This establishes the condition V/V<2a 3y e β such that b'<y.

We have established both conditions of Lemma 9.3, so 2a= sup Β . •

The two-part nature of the proof above is typical.

Definitions A real number / is said to be a lower bound for the set A if VJC e A,
x> I. A set is bounded below if it possesses a lower bound. The real number m is
called the infimum of the set A and denoted by inf A if (and only if) it is the
greatest lower bound of A. A set is said to be bounded if it is both bounded above
and bounded below.

To save repeating our work to produce the properties of the infimum, we notice the
following:

Lemma 9.7 Let Λ be a non-empty set of real numbers which is bounded above and
let Β = { JC: -JC e A} . Then Β is bounded below and inf Β = -sup A .

Proof: Let a = sup A . Then jce β :=> - x e A = > - x < a = > j c > - a . Thus - a is a
lower bound for Β . In passing, notice that this shows Β is bounded below.

Suppose that b is a lower bound for β ; we need to relate this to A. VJC e A ,

—x e Β so -JC > b and JC < —b . Thus VJC e A , JC < -b . Therefore -b is an upper
bound for A , hence a < -b since α = sup/t is the smallest upper bound for A.
Hence b<-a . W e have shown that if b is a lower bound for Β , b < -a , so since
-a is a lower bound for B, it is the greatest lower bound, that is, inf B. •

Lemma 9.8 Every non-empty set of real numbers which is bounded below has a
(real) infimum. If A is a subset of R , then t=\nfA ifandonlyif

(i) V x e A x>t and (ii) Vf'> t 3xe A such that f ' > x .

Proof: Let A be non-empty and bounded below. Let C = {JC: -JC e A ). Then C is
non-empty and bounded above and thus C has a supremum. (If / is a lower bound
for A , - / i s an upper bound for C , as is easily checked.) By Lemma 9.7, then,
inf A - - s u p C so inf A exists. The proof that the criteria are correct can either be
obtained by relating them to -t - sup C or by analogy with the proof of Lemma 6.3.
•

Lemma 9.9 If Λ is a bounded non-empty subset of R , inf A < sup/t.

Ch. 9] The Structure of the Real Number System 119

Proof: Since Α Φ 0 , we can choose χ e A. Then inf A < χ and χ < sup A so
inM<supA. •

Example 9.4 Let A and Β be two non-empty sets which are bounded above and
let C = [x + y: x e A and y e B] . Show that sup C = sup A + sup Β .

Solution: Let a = sup A and 6 = sup Β . Then if z e C there are JC e Λ and y e β

such that z = jc + y < a + ft (since JC < α and y < b). Thus Vz e C z<a + b .

Now let c' < a + /?. Then c' - b < a , so, since α = sup A ,3xe A with c ' - b < χ .
For this value of JC , c' - χ < b = sup Β , so 3y e Β with c' - JC < y . Thus letting
ζ = χ + y , we have z e C with c' < ζ . This gives the second property of sup C , so
a + b- sup C (by Lemma 9.3). •

Example 9.5 Suppose that A and Β are two non-empty sets in Κ with the
property that VJC e A and Vy e Β χ < y . Show that sup A < inf Β .

Solution: Let JC e A . Then Vy e Z? JC < y so that JC is a lower bound for β .

Therefore j c < i n f f i . (inf β is the greatest lower bound.)

The conclusion of the above paragraph was reached only on knowing that JC e A so
it must hold for all such JC , hence VJC e A χ < inf B. Thus inf Β is an upper bound
for A, so sup A < inf Β . •

Example 9.6 Let Λ be a non-empty, bounded subset of R . Then

sup [\x\ : χ e A} = max(sup>4, -inf Λ) = max(|sup A\, |inf A\).

Solution: Let χ e A. Then inf Λ < JC < sup A, so Example 5.19 shows that
|x| < max(|sup A\, |inf A\) , proving that max(|sup A\, |inf A\) is an upper bound for
our set. Now we can choose sequences ( y j and (z„) in A with y„ —» inf A and
z„-» sup A as η - » °° . Therefore, |y„| - » |inf A\ and |z„| - 4 |sup A\ , so one of these
provides a sequence in sup {p:| : JC e A) which tends to max(|sup/\|, |inf A\) showing
that no smaller number is an upper bound. The comment following Example 5.19
shows us that max(sup A, -inf A) - max(|sup A\, |inf A\). •

Experience will show that in most examples where a supremum or infimum is to be

shown equal to a given number, the proof is in two steps. Where an inequality
involving sup or inf is concerned, one step may suffice and the two parts of
Lemma 9.3 can be viewed as establishing two opposite inequalities (e.g. if a is an
upper bound for A then sup A < a).
Notice that the supremum of a set need not belong to the set (consider (0,1)); life
would be much simpler if this were true! Also notice that if a' < sup A 3x 6 A such
that a ' < j c . This guarantees the existence of some JC e A with a'<x but not that
a' itself is in A. (Consider A = {-J.f . 7 . · · · } ·)

We give closed intervals their own name because they have useful properties, one of
which is that the closed interval [a, b] contains both its supremum and infimum. An
 120 The Structure of the Real Number System [Ch. 9

open interval contains neither its supremum nor its infimum. In the course of the
next chapter we shall encounter several situations where the distinction is crucial.

There is one common type of set which is simpler in this respect. Subsets of Ν ,
other than the empty set, necessarily have a smallest element so in this case the set
contains its infimum. This was proved in Lemma 4.1.

Problems
1. The following sets are bounded above; identify the set of upper bounds in
each case: (0, 1), [x: Mx > 2 ) , {x e Q: χ < 4 } , (x e Q: χ < 4 } , ,

1 } { }
2 4'4 -" ' 2 ' 4 ' 8 ' 1 6 ' - ' { : m , η, e Ν, 0<m<n and η is a
power of 2 ).

2. Let Λ be a non-empty subset of R such that Vx e Α , χ < 2. Prove that

sup A < 2. Give two examples of sets A satisfying this property, one with
sup A = 2, one with sup A < 2 .

3. Find the supremum of each of the sets in Ql and check that the properties
demanded by Lemma 9.3 are satisfied.

4. Recall that if r is an irrational number and g is a non-zero rational number,

rq is irrational. Use this to show that if χ and y are real numbers and x<y
then there is an irrational number s with χ < s < y . (Hint: consider xN2 and
v/V2.)

5. Show that for each n e Ν there is a rational number a„ satisfying

l
42— -<a n <V2, and deduce that there is a sequence of rational numbers
whose limit is V*2 .

6* Modify the construction in Q5 to produce an increasing sequence (b ) n of

rational numbers whose limit is V2.

7. Let A be a non-empty subset of R which is bounded above, and define Β

and C by β = [x e R : x - 1 e A ) and C= (x e R: (x + l ) / 2 e A ) . Prove
that sup Β = 1 + sup A , sup C = 2sup A - 1.

8. Let A be a non-empty set of positive numbers, which is bounded above, and

set Β = [χ € R: l/x e A } . Prove that Β is bounded below and infi? =
1/supA .

9. Suppose that A and Β are non-empty bounded subsets of R. Prove that sup
(A^JB)= max(sup A , sup β ) , inf {x + y: x e A and y € β ) = inf A + inf β

and sup [x - y: x e A and y e β ) = sup A - inf β .

10. Let A and Β be non-empty, bounded sets of positive numbers and define C
by C = (xy: x e A and y e β ) . Prove that C is bounded and that
sup C = sup A sup Β , inf C - inf A inf B.