lA

mo
~8
1976
)upp.
\/lATH

Solutions Manual to Walter

Rudin's Principles of
Mathematical Analysis

Roger Cooke, University of Vermont

Su

Chapter 1

The Real and Complex

Number Systen1s

Exercise 1.1 If r is rational (r =f. 0) and x is irrational, prove that r +x and

rx are irrational.

Solution. If r and r + x were both rational, then x = r + x - r would also be

rational. Similarly if rx were rational, then x = 7 would also be rational.

Exercise 1.2 Prove that there is no rational number whose square is 12.

First Solution. Since v'f2 = 2.)3, we can inv~ke the previous problem and
prove that .J3 is irrational. If m and n are integers having no common factor
and such that m 2 ....:. 3n 2 , then m is divisible by 3 (since if m 2 is divisible by 3,
so ism). Let m = 3k. Then m 2 = 9k 2 , and we have 3k 2 = n 2 . It then follows
that n is also divisible by 3 contradicting the assumption that m and n have no
common factor.

Second Solution. Suppose m 2 = 12n 2 , where m and n have no common factor.

It follows that m must be even, and therefore n must be odd. Let m = 2r.
Then we have r 2 = 3n2 , so that r is also odd. Let r = 2s + 1 and n = 2t + 1.
Then
4s 2 + 4s + 1 = 3(4t2 + 4t + 1) = 12t2 + 12t + 3,
so that
4( s 2 +s - 3t 2 - 3t) = 2.

But this is absurd, since 2 cannot be a multiple of 4.

6 CHAPTER 1. THE REAL AND COMPLEX NUMBER SYSTEMS

Exercise 1.3 Prove Proposition 1.15, i.e., prove the following statements:
(a) If x =f. 0 and xy = xz, then y = z.
(b) If x =f. 0 and xy = x, then y = 1.
(c) If x =f. 0 and xy = 1, then y = 1/x.
(d) If x =f. 0, then 1/(1/x) = x.
Solution. (a) Suppose x ::J. 0 and xy = xz. By Axiom (M5) there exists an
element 1/x such that 1/x = 1. By (M3) and (M4) we have (1/x)(xy) =
((1/x)x)y = ly = y, and similarly (1/x)(xz) = z. Hence y = z.
(b) Apply (a) with z = 1.
(c) Apply (a) with z = ljx.
(d) Apply (a) with x replaced by 1/x, y = 1/(1/x), and z = x.

Exercise L4 Let E be a nonempty subset of an ordered set; suppose a is a

lower bound of E, and f3 is an upper bound of E. Prove that a< {3.
Solution. Since E is nonempty, there exists x E E. Then by definition of lower
and upper bounds we have a :::; x :::; {3, and hence by property ii in the definition
of an ordering, we have a< f3 unless a= x = {3.

Exercise 1.5 Let A be a nonempty set of real numbers which is bounded below.
Let -A be the set of all numbers -x, where x EA. Prove that

inf A=- sup( -A).

Solution: We need to prove that -sup( -A) is the greatest lower bound of A.
For brevity, let a= -sup( -A). We need to show that a:::; x for all x E A and
a ~ f3 if f3 is any lower bound of A.
Suppose x EA. Then, -x E -A, and, hence -x:::; sup( -A). It follows that
x ~ -sup( -A), i.e., a:::; x. Thus a is a lower bound of A.
Now let f3 be any lower bound of A. This means f3 :::; x for all x in A.
Hence -x:::; -{3 for all x E A, which says y:::; -{3 for ally E -A. This means
-{3 is an upper bound of -A. Hence -{3 ~ sup( -A) by definition of sup, i.e.,
f3:::; -sup( -A), and so- sup( -A) is the greatest lower bound of A.

Exercise 1.6 Fix b > 1.

(a) If m, n, p, q are integers, n > 0, q > 0, and r = mjn = pjq, prove that

Hence it makes sense to define br = (bm) l/n.

(b) Prove that br+s = brbs if r and s are rational.
 7

(c) If x is real, define B (x) to be the set of all numbers bt, where t is rational
and t ::; x. Prove that
br = supB(r)
when r is rational. Hence it makes sense to define

bx =sup B(x)

for every real x.

(d) Prove that bx+y = bx bY for all real x and y.
Solution. (a) Let k = mq == np. Since there is only one positive real number c
such that cnq = bk (Theorem 1.21), if we prove that both (bm)lfn and (bP) 1 fq
have this property, it will follow that they are equal. The proof is then a routine
computation: ((bm)Ifntq = (bm)q = bmq = bk, and similarly for (bP) 1 fq.
(b) Let r = !!!:. w • Then r + s = mw+vn
n and s = ..!!.. nw ' and

by the laws of exponents for integer exponents. By the corollary to Theorem

1.21 we then have

where the last equality follows from part (a).

(c) It will simplify things later on if we amend the definition of B(x) slightly,
by defining it as {bt : t rational, t < x }. It is then slightly more difficult to
prove that br = sup B(r) if r is rational, but the technique of Problem 7 comes
to our rescue. Here is how: It is obvious that br is an upper bound of B(r).
We need to show that it is the least upper bound. The inequality b1 fn < t if
n > (b- 1)/(t- 1) is proved just as in Problem 7 below. It follows that if
0 < x < br, there exists an integer n with b1 fn < br jx, i.e., x < br-l/n E B(r).
Hence x is not an upper bound of B(r), and so br is the least upper bound.
(d) By definition bx+y = supB(x + y), where B(x + y) is the set of all
numbers bt with t rational and t < x + y. Now any rational number t that is
less than x + y can be written as r + s, where r and s are rational, r < x, and
s < y. To do this, let r be any rational number satisfying t - y < r < x, and
let s = t - r. Conversely any pair of rational numbers r, s with r < x, s < y
gives a rational sum t = r + s < x + y. Hence B(x + y) can be described as the
set of all numbers brbs with r·< x, s < y, and rands rational, i.e., B(x+y) is
the set of all products uv, where u E B(x) and v E B(y).
Since any such product is less than supB(x)supB(y), we see that the num-
ber M = sup B(x) sup B(y) is an upper bound for B(x + y). On the other
hand, suppose 0 < c < supB(x)supB(y). Then cj(supB(x)) < supB(y). Let
m = (1/2)(c/(supB(x)) + supB(y)). Then c/ supB(x) < m < supB(y), and
there exist u E B(x), v E B(y) such that cjm < u and m < v. Hence we have
8 ' CHAPTER 1. THE REAL AND COMPLEX NUMBER SYSTEMS

c = (cjm)m < uv E B(x + y), and soc is not an upper bound for B(x + y). It
follows that supB(x) supB(y) is the least upper bound of B(x + y), i.e.,

as required.

Exercise 1.7 Fix b > 1, y > 0, and prove that there is a unique real x such
that bx = y, by completing the following outline. (This xis called the logarithm
of y to the base b.)
(a) For any positive integer n, bn - 1 2: n(b - 1).
(b) Hence b-12: n(b1 fn -1).
(c) If t > 1 and n > (b- 1)/(t- 1), then b1 fn < t.
(d) If w is such that bw < y, then bw+(l/n) < y for sufficiently large n; to see
this apply part (c) with t = y · b-w.
(e) If bw > y, then bw-(l/n) > y for sufficiently large n.
(f) Let A be the set of all w such that bw < y, and show that x = sup A
satisfies bw = y.
(g) Prove that this x is unique.
Solution. (a) The inequality bn - 1 2: n(b- 1) is equality if n = 1. Then, by
induction bn+l -1 = bn+ 1 -b+ (b-1) = b(bn -1) + (b-1) 2: bn(b-1) + (b-1) =
(bn + 1)(b- 1) 2: (n + 1)(b- 1).
(b) Replace b by b1 fn in part (a).
(c) The inequality n > (b -1)/(t- 1) can be rewritten as n(t- 1) > (b -1),
and since b- 1 2: n(b 1 fn- 1), we have n(t- 1) > n(b 1 fn- 1), which implies
t > blfn.
(d) The application of part (c) with t = y · b-w > 1 is immediate.
(e) The application of part (c) with t = bw · (1 j y) yields the result, as in
part (d) above.
(f) There are only three possibilities for the number x =sup A: 1) bx < y; 2)
bx > y; 3) bx = y. The first assumption, by part (d), implies that x+ (1/n) E A
for large n, contradicting the assumption that xis an upper bound for A. The
second, by part (e), implies that x- (1/n) is an upper bound for A if n is large,
contradicting the assumption that x is the smallest upper bound. Hence the
only remaining possibility is that bx = y.
(g) Suppose z =/:. x, say z > x. Then bz = bx+(z-x) = bxbz-x > bx = y.
Hence x is unique. (It is easy to see that bw > 1 if w > 0, since there is a
positive rational number r = 7: with 0 < r < w, and br = (bm )1 /n. Then
bm > 1 since b > 1, and (bm)lfn > 1 since 1n = 1 < bm.)
 9

Exercise 1.8 Prove that no order can be defined in the complex field that turns
it into an ordered field. Hint: -1 is a square.
Solution. By Part (a) of Proposition 1.18, either i or -i must be positive. Hence
-1 = i 2 = ( -i) 2 must be positive. But then 1 = ( -1) 2, must also be positive,
and this contradicts Part (a) of Proposition 1.18, since 1 and -1 cannot both
be positive.

Exercise 1.9 Suppose z = a+ bi, w = c + di. Define z < w if a < c, and

also if a = c but b < d. Prove that this turns the set of all complex numbers
into an ordered set. (This type of order relation is called a dictionary order, or
lexicographic order, for obvious reasons.) Does this ordered set have the least
upper bound property?
Solution. We need to show that either z < w or z = w, or w < z. Now sin~e
the real numbers are ordered, we have a < c or a = c, or c < a. In the first
case z < w; in the third case w < z. Now consider the second case. We must
have b < d or b = d or d < b. In the first of these cases z < w, in the third case
w < z, and in the second case z = w.
We also need to show that if z <wand w < u, then z < u. Let u = e + fi.
Since z < w, we have either a < cor a= c and b < d. Since w < u we have
either c < for c = f and d <g. Hence there are four possible cases:
Case 1: a< c and c <f. Then a< f and so z < u, as required.
Case 2: a < c and c = f and d < g. Again a < f, and z < u.
Case 3: a = c and b < d and c < f. Once again. a < f and so z < u.
Case 4: a= c and b < d and c = j, and d <g. Then a= f and b < g, and
so z < u.

Ex~rcise 1.10 Suppose z =a+ bi, w = u + iv, and

a= ('wl2+ ur/2, b = cwl2- ur/2.

Prove that z 2 = w if v 2: 0 and that (z) 2 = w if v :::; 0. Conclude that every

complex number (with one exception) has two complex square roots.
Solution.

Now
a2 _ b2 = !w! + u _ !w! - u = u
2 2 '
and, since (xy) 112 . x 112y 112,
2
_ (!w!+u!w!-u)l/2_ (!w! -u2)1/2
2ab - 2 - 2 .
2 2 4
10 CHAPTER 1. THE REAL AND COMPLEX NUMBER SYSTEMS

Hence
2ab = 2( (~)
2
r
12

Now (x 2)112= x if x 2: 0 and (x 2)112= -x if x ~ 0. We conclude that 2ab = v

if v 2: 0 and 2ab = -v if v ~ 0. Hence z 2 = w if v 2: 0. Replacing b by -b, we
find that (z) 2 = w if v ~ 0.
Hence every non-zero complex number has (at least) two complex square
roots.

Exercise 1.11 If z is a complex number, prove that there exists an r > 0 and a
complex number w with lwl = 1 such that z = rw. Are wand r always uniquely
determined by z?
Solution. If z =' 0, we take r = 0, w = 1. (In this case w is not unique.)
Otherwise we taker= !zl and w = z/lzl, and these choices are unique, since if
z = rw, we must haver= r!wl = !rw! = !z!, zjr.

Exercise 1.12 If z1 , ... , Zn are complex, prove that

Solution. The case n = 2 is Part (e) of Theorem 1.33. We can then apply this
result and induction on n to get

!(zl + Z2 + · · · + Zn-l) + Znl

< !zl + Z2 + · · · + Zn-ll + lzn I
< lz1l + lz2l + · · · + lzn-11 + lznl·

Exercise 1.13 If x, yare complex, prove that

j!xl-iYII ~ !x- Yl·

Solution. Since x = x- y + y, the triangle inequality gives

lxl ~ !x- Yl + IYI,

so that lxl- !YI ~ !x- Yl· Similarly IY!-!xl ~ !x- Yl· Since !xl- IYI is a real
number we have either j!x!-IYII = !xi-IYI or j!xi-IY!I = IYI-!xl. In either
case, we have shown that jlxl - IYII ~ lx- Yl·
 11

Exercise 1.14 If z is a complex number such that \z\ = 1, that is, such that
zz = 1, compute

Solution. \1 + z\ 2 = (1 + z)(1 + z) = 1 + z + z + zz = 2 + z + z. Similarly

\1- z\ 2 = (1- z)(1- z) = 1- z- z + zz = 2- z- z. Hence

\1 + z\ 2 + \1- z\ 2 = 4.

Exercise 1.15 Under what conditions does equality hold in the Schwarz in-
equality?
Solution. The proof of Theorem 1.35 shows that equality can hold if B = 0 or
if Baj- Cb.i = 0 for all j, i.e., the numbers aj are proportional to the numbers
bj. (In terms of linear algebra this means the vectors a = (a1, a2, ... , an) and
b = (bll b2, ... , bn) in complex n-dimensional space are linearly dependent. Con-
versely, if these vectors are linearly independent, then strict inequality holds.)

Exercise 1.16 Suppose k 2: 3, x, y E Rk, \x- y\ = d > 0, and r > 0. Prove:

(a) If 2r > d, there are infinitely many z E Rk such that

\z - x\ = \z - y\ = r.
(b) If 2r = d, there is exactly one such z.
(c) If 2r < d, there is no such z.
How must these statements be modified if k is 2 or 1?
Solution. (a) Let w be any vector satisfying the following two equations:

w·(x-y)

\w\2

From linear algebra it is known that all but one of the components of a solution
w of the first equation can be arbitrary. The remaining component is then
uniquely determined. Also, if w i~ any non-zero solution of the first equation,
there is a unique positive number t such that tw satisfies both equations. (For
example, if x 1 =/:. y1 , the first equation is satisfied whenever

Zl =
Z2(X2- Y2) + ... + Zk(Xk - Yk)
.
Yl- Xl

If (z 1 , z2, ... , zk) satisfies this equation, so does (tz 1 , tz2, ... , tzk) for any real
number t.) Since at least two of these components can vary independently, we
can find a solution with these components having any prescribed ratio. This
12 CHAPTER 1. THE REAL AND COMPLEX NUMBER SYSTEMS

ratio does not change when we multiply by the positive number t to obtain
a solution of both equations. Since there are infinitely many ratios, there are
infinitely many distinct solutions. For each such solution w the vector z -
~x + ~y + w is a solution of the required equation. For

ly;x+wl2
y-xl2 x-y 2
l -2- +2w·-2-+lwl
·d2 d2
- +0+r 2 - -
4 4
r2
'
and a similar relation holds for iz - y! 2 .
(b) The proof of the triangle inequality shows that equality can hold in this
inequality only if it holds in the Schwarz inequality, i.e., one of the two vectors
is a scalar multiple of the other. Further examination of the proof shows that
the scalar must be nonnegative. Now the conditions of this part of the problem
show that
!x- Yl = d = !x- zi + iz - Yl·
Hence it follows that there is a nonnegative scalar t such that

x- z = t(z- y).
However, the hypothesis also shows immediately that t = 1, and so z is uniquely
determined as
x+y
Z= -2-.

(c) If z were to satisfy this condition, the triangle inequality would be vio-
lated, i.e., we would have

lx- Yl = d > 2r = lx- zi + !z- Yl·

When k = 2, there are precisely 2 solutions in case (a). When k = 1, there
are no solutions in case (a). The conclusions in cases (b) and (c) do not require
modification.

Exercise 1.17 Prove that

if x E Rk and y E Rk. Interpret this geometrically as a statement about

parallelograms.
Solution. The proof is a routine computation, using the relation
 13

If x and y are the sides of a parallelogram, then x + y and x - y are its

diagonals. Hence this result says that the sum of the squares on the diagonals
of a parallelogram equals the sum of the squares on the sides.

Exercise 1.18 If k 2:: 2 and x E Rk, prove that there exists y E Rk such that
y =I= 0 but x · y = 0. Is this also true if k = 1?
Solution. If x has any components equal to 0, then y can be taken to have
the corresponding components. equal to 1 and all others equal to 0. If all the
components of x are nonzero, y can be taken as ( -x 2 , x 1 , 0, ... , 0). This is, of
course, not true when k = 1, since the product of two nonzero real numbers is
nonzero.

Exercise 1.19 Suppose a E Rk, bE Rk. Find c E Rk and r > 0 such that

lx - ai = 2jx- hi

if and only if lx- cl = r. (Solution: 3c = 4b- a, 3r = 2lb- a!.)

Solution. Since the solution is given to us, all we have to do is verify it, i.e., we
need to show that the equation

lx-al = 2lx- bl

is equivalent to lx- cl = r, which says

4 1 I -lb-al
2
Ix- -b+
3 -a=
3 3 .
If we square both sides of both equations, we an equivalent pair of equations,
the first of which reduces to

3lxl 2 + 2a · x- 8b · x- lal 2 + 4lbl 2 = 0,

and the second of which reduces to this equation divided by 3. Hence these
equations are indeed equivalent.

Exercise 1.20 With reference to the Appendix, suppose that property (III)
were omitted from the definition of a cut. Keep the same definitions of order
and addition. Show that the resulting ordered set has the least-upper-bound
property, that addition satisfies axioms (A1) to (A4) (with a slightly different
zero element!) but that (A5) fails.
Solution. We are now defining a cut to be a proper subset of the rational
numbers that contains, along with each of its elements, all smaller rational
14 CHAPTER 1. THE REAL AND COMPLEX NUMBER SYSTEMS

numbers. Order is defined by containment. Now given a set A of cuts having an

upper bound {3, let a be the union of all the cuts in A. Obviously a is properly
contained in {3, and so is a proper subset of the rationals. It also obviously
satisfies the property that if p E a and q < p, then q E a; hence a is a cut. It is
further obvious that a contains each elements of A, and so is an upper bound
for A. It remains to prove that there is no smaller upper bound.
To that end, suppose, "/ < a, then a contains an element x not in "Y· By
definition of a, x must belong to some cut b in A. But then "Y < b, and so "Y is
not an upper bound for A. Thus a is the least upper bound.
The proof given in the text goes over without any change to show that (Al),
(A2), and (A3) hold. As for (A4) let 0 = {r : r ~ 0}. We claim 0 +a = a.
The proof is easy. First, we obviously have 0 +a~ a. For r + s ~ s if r ~ 0.
Hence r + s E a if s E a. Conversely a ~ 0 +a, since each s in a can be written
as 0 + s.
Unfortunately, if 0' = { r : r < 0}, there is no element a such that a + 0' =
0. For a + 0' has no largest element. If x = r + s E a + 0', where r E a and
s E 0', there is an element t E 0' with t > s, and so r+t E a+O' and r+t > s.
Since 0 has a largest element (namely 0), these two sets cannot be equal.