The modulus of Section 1.2 measures a complex number
by a real one. Picture as the point of the plane: by
Pythagoras, is its distance from the
origin, and is the distance between the points and . In a triangle, one
side is never longer than the other two together. This
section proves the algebraic form of that fact, , from the definitions alone. The
picture motivates it; no step of the proof uses it.
Why the old
proof does not transfer
For an ordered field, Number Systems: From to , Chapter 4
proved by trapping (Number Systems: From to , Proposition 4.1.3
(e)). Each of and
lies between
its absolute
value, (part (c) there); adding the two traps
traps between
, and part (d) reads that as the
inequality.
That route needs : a comparison between itself and a number.
For a complex ,
“”
compares a complex number with a real one, and has been given no
order in which to make the comparison. Section 1.4 shows
that no order compatible with the arithmetic of exists (Proposition 1.4.3).
So the proof must compare real numbers
only.
Two kinds of real number are at hand. Moduli are
real, and so are real parts. The device that connects
them to the arithmetic of is the identity of
Theorem 1.2.2 (c).
A modulus is a square root, and square roots do not add
well; its square is a product, and products expand. So
we will compare squares and then come back.
Comparing squares. Coming back from
squares to the numbers needs one rule about the real
numbers. Let with and . Then and likewise with on both sides. The
strict form is Number Systems: From to , Proposition 2.3.5
(g). For the first equivalence: if , then by (g), and
if the
squares agree. Conversely, if and yet , then (g) gives
, which
contradicts by trichotomy (O1); so . The second
equivalence follows from the first, applied in both
directions, and trichotomy. In words: for
non-negative reals, squaring neither creates nor
destroys a comparison. Both hypotheses and matter: , and yet
.
The real
part is at most the modulus
The single estimate the proof needs is that a real
part never exceeds the modulus. It comes with two
companions, which say that the modulus and the “taxicab
size” are
comparable.
Lemma
1.3.1 (Real and Imaginary Parts against the
Modulus)
For every , In particular .
Here
is the absolute value of a real number. By Proposition 1.2.1
it is the same whether it is read as the absolute value
of the ordered field from Number
Systems: From
to ,
Chapter 4 (Number Systems: From to , Definition 4.1.1)
or as the modulus of a complex number that happens to be
real, so the lemma has only one reading.
Idea. Every quantity in the lemma
is a non-negative real, so compare squares. The square
of is , and the
squares of
and are
and , each at most the
sum; the square of is the sum plus the non-negative cross
term .
Proof. Write with . By Definition 1.2.1
and the defining property of the square root (Number Systems: From to , Proposition 6.1.4),
and
.
Since
(Number Systems: From to , Proposition 2.3.5
(e) if ;
otherwise),
adding gives
by Number Systems: From to , Proposition 2.3.5
(b). Also and (Number Systems: From to , Proposition 4.1.3
(a), (b)). Hence , with both and non-negative, and
comparing squares gives . The
same argument with the roles of and exchanged gives . Then
by Number Systems: From to , Proposition 4.1.3
(c), so .
For the last inequality, expand with Number Systems: From to , Proposition 2.3.2
(d): where the inequality holds because as a
product of non-negative reals, so ,
and Number Systems: From to , Proposition 2.3.5
(b) adds it. Both and are non-negative,
so comparing squares gives . This proves the lemma.
For
the three inequalities read , and . For they read , and : a real makes the first and
third equalities. The inequality is strict for , where it reads
; the
absolute value inside the first inequality is what makes
it sharp for real
of either sign.
Quick check. Find a non-zero with .
What do the three inequalities of Lemma 1.3.1 say
for it?
Solution. For example , with ,
and . The inequalities read , and . Any with works the same
way: , so the second and third inequalities
are equalities.
The
triangle inequality
We can now state the theorem. Its proof is the first
full run of the move this chapter is built around.
Theorem
1.3.2 (Triangle Inequality for Complex
Numbers)
For all ,
Try it first (ten minutes): show by expanding both sides with ,
and note which term you cannot yet bound.
Idea. Expanding the left side with
and conjugation of sums gives ; the right side is . Everything matches except the cross
term , and an attempt usually stops there:
it looks like a complex number, and complex numbers
cannot be compared with . The way
out is to notice that is the
conjugate of , so the cross term is with : a real number after
all. A real part is at most the modulus, and . Squares compare, so the moduli
do.
Before reading on: where is used, and why may we compare
squares?
Proof. Let . By Theorem 1.2.2 (c),
then (a), then Proposition 1.1.2
(c) together with (a) and (b) (distributivity,
commutativity, associativity), By Theorem 1.2.2 (b)
and (a), . So with
, the two middle terms are , by Theorem 1.2.2 (c).
Using (c) once more for and , Every term here is real, and sums and products
of real numbers are the same in as in (Proposition 1.1.1),
so the inequalities below take place in the ordered
field . By Lemma 1.3.1, , and by Theorem 1.2.2 (d),
.
Adding this inequality to itself ( by
distributivity) and then adding (Number Systems: From to , Proposition 2.3.5
(b) twice), and expanding the square with Number Systems: From to , Proposition 2.3.2
(d), Both and are non-negative real numbers, so
comparing squares gives . This proves the triangle
inequality.
The proof answers its prompt in two lines. is used once, to bound the cross term
by .
Comparing squares is legitimate because the two
quantities being compared, and , are
both non-negative; without that, a smaller square would
say nothing about the numbers.
This route has a name in this chapter: square
the modulus. The modulus is a square root,
and square roots are hard to handle; its square is
a product, so square, expand with conjugates, bound or
compare the real numbers you get, and take the
non-negative square root back. Section 1.2 used the
same move once already, for
(Theorem 1.2.2
(d)). The linear algebra volumes take the same route,
square, expand, bound the cross term, for the length of
a vector in an inner product space.
Warning. The square of the modulus
is , not
. Expanding
as
gives nonsense: for , ,
while . The cross term of the correct
expansion is , and only that form is real.
Quick check. Compute and , and check them against the
triangle inequality.
Solution. and , so the right side is . The sum is , and . Indeed . In the picture, the two points and are three corners of a
parallelogram with the origin as the fourth, and the
diagonal is
shorter than the two sides .
Three
variants
Three consequences are used often enough to be stated
once: a lower bound for the modulus of a difference, the
inequality for any finite number of terms, and a lower
bound for the modulus of a sum.
Corollary
1.3.3 (Variants of the Triangle
Inequality)
Let .
- (a) (Reverse
triangle inequality.) .
- (b) For every with and all ,
.
- (c) .
In (a), the outer bars are the absolute value of the
real number . In words: (a) says that two points at
distance from each other have distances from the
origin that differ by at most that much; (b) says that a
path of steps is
at least as long as the straight line from start to end;
(c) says that adding can lower the modulus
by at most .
Idea. For (a), split as and apply the
theorem; that bounds from above, and swapping and bounds it from below.
For (b), add one term at a time. For (c), apply (a) to
and .
Proof.
(a) Since , Theorem 1.3.2
gives , so . Exchanging and gives . Moreover , so by Theorem 1.2.2 (d),
, where
by Proposition 1.2.1.
Let . We have shown and
,
that is, : if , then
by Number Systems: From to , Proposition 2.3.5
(a) and Number Systems: From to , Proposition 2.3.2
(a), and if , then . By Number Systems: From to , Proposition 4.1.3
(d), .
(b) By induction
on (Logic and Proof, Axiom 4.2.1,
through Number Systems: From to , Corollary 1.7.3
(b), applied to the statement for , ), with the sums
defined by appending the last term, as for real sums (Sets and Functions, Definition 2.3.1).
For , both
sides are . Suppose the
inequality holds for , and let . Then by Theorem 1.3.2,
then the induction hypothesis with Number Systems: From to , Proposition 2.3.5
(b).
(c) Apply (a) to
and . Since (Number Systems: From to , Proposition 2.3.2
(a)) and as in (a), it reads
. By Number Systems: From to , Proposition 4.1.3
(c), . This proves
the corollary.
Part (b) with is the theorem itself, so the corollary
contains it. Algebraic Preliminaries, Chapter 4
uses (b) and (c) to show that a polynomial expression of
degree at least
is large in modulus when is large, a step
towards the Fundamental Theorem of Algebra.
Quick check. Take and . Compute both
sides of (a) and of (c).
Solution. and , so .
For (a): , and because
(comparing squares). For (c): too, and again . Both
bounds are far from sharp here; (a) is an equality when
, for
instance.
Exercises
A. Check your
understanding
Exercise
(A1)
These questions are about the proof of Theorem 1.3.2.
- (a) Where in the
proof is used, and for which ?
- (b) Why may the
proof pass from back to ?
- (c) Run the proof on
and : compute the cross
term ,
then both sides of the squared inequality.
- (d) State the idea
of the proof in two sentences.
- (e) True or false:
for all . Justify your answer.
Solution.
(a) Once, to
bound the cross term: with , .
(b) Both and are
non-negative reals, and for non-negative reals if and only if
(comparing squares, from Number Systems: From to , Proposition 2.3.5
(g) and trichotomy).
(c) , so , with real part : the cross term is
. Then and , and . So .
(d) Square both
sides and expand with :
everything matches except the cross term, which is .
A real part is at most the modulus, and , so the squares compare, and so
do the non-negative moduli.
(e) False. For
and , , but
.
B. Practice
Exercise
(B1: The equality case)
Let . Prove that if and only if is a
non-negative real number, that is, and .
Hint: trace the proof of the triangle inequality
and ask where an inequality could be strict.
Solution. Let with . By the proof of Theorem 1.3.2,
the second because
(Theorem 1.2.2
(d)). Both moduli are non-negative, so (comparing
squares) holds if and only if the
squares are equal, that is, if and only if , that is,
.
It remains to show that holds exactly when and . If , then , and squaring
gives , so and (Number Systems: From to , Proposition 2.3.5
(e): a non-zero square is positive). Conversely, if
and , then ,
since and
(uniqueness in Number Systems: From to , Proposition 6.1.4).
Hence if and only if is a
non-negative real number.
For instance , : , and indeed . In the picture: equality holds exactly
when one of is
or they point in
the same direction from the origin. The linear algebra
volumes prove the version for terms.
Exercise
(B2: A distance that does not change)
Let
with .
Prove that for every . The first two
steps are done: both by expanding with Theorem 1.2.2.
Name the move, then complete it.
Solution. The move is
square the modulus: compare the
squares, which are algebra, and take the non-negative
square roots back.
Substitute into both lines. The first becomes , and so does the second, since . So . Both and are
non-negative, and non-negative reals with equal squares
are equal (comparing squares). Hence .
(For the record, the given lines. Conjugation
respects differences: by Theorem 1.2.2 (b),
Proposition 1.2.1
and Number Systems: From to , Proposition 2.3.2
(c), so with (a), ; and by Proposition 1.2.1.
Then , and , so .)
Exercise
(B3)
Prove that for every , and, for each
of the two inequalities, find a non-zero for which it is an
equality.
C. Going deeper
Exercise
(C1: Exit check: the triangle inequality)
State and prove the triangle inequality for complex
numbers. Your proof must include a proof that for every .
Solution. Statement. For
all ,
.
The estimate. Let with . Then , since . Both and are non-negative,
so by comparing squares (Number Systems: From to , Proposition 2.3.5
(g) with trichotomy). Hence (Number Systems: From to , Proposition 4.1.3
(c)).
The inequality. Let . By Theorem 1.2.2
(a)–(c), By the estimate and Theorem 1.2.2 (d),
. Hence . Both and are
non-negative, so comparing squares gives .
Marking scheme. (10 points) - 2:
proves , with a reason for passing from
squares to moduli. - 3: expands with
conjugates and identifies the cross term as ,
a real number. - 2: bounds the cross term by , citing
. - 1: concludes . - 2: takes square roots, justified
by the non-negativity of both sides. A second route,
with coordinates , : expanding and isolating the cross term
earns
the 3 points of the second line; proving , for instance from , which follows from , earns
the 2 points of the third line if proved, not if quoted;
the last two lines are marked as above. The 2 points for
are still owed on that route, since the
problem asks for it. Fatal errors (0 for the step):
writing or any inequality between non-real
complex numbers; dropping the conjugate, as in .
Exercise
(C2: Inside the unit disc)
Let
with
and .
Prove that . Here is the start of a
proof: square both sides and expand. Name the move,
then complete it.
Solution. The move is
square the modulus. Both sides are
non-negative, so it is enough to show (comparing squares, strict form:
Number Systems: From to , Proposition 2.3.5
(g)). Expanding each as , exactly as in
Exercise (B2: A distance that does not change),
The non-real terms cancel in the difference:
Since , comparing squares gives ,
so ; likewise . A product of two positive
reals is positive ((O4)). So the difference is positive,
, and .
In particular . Since (a modulus is unchanged by
conjugation, Theorem 1.2.2
(d)), the inequality says that for all in the open unit disc: the map sends the open unit disc into
itself, a fact complex analysis builds on.
Marking scheme. (10 points) - 2:
reduces to the squared inequality, with the
non-negativity of both sides as the reason. - 3: expands
both squares with conjugates. - 3: factors the
difference as . - 2: shows both
factors positive, from ,
and concludes strictly. Fatal errors (0 for the step):
expanding as ; comparing with a real
number.