38 Vector algebra

The last characterization of the solution to our problem

° a−b ° shows that the solution set is
° °
the locus of all points having fixed distance ( 2 ) from the midpoint of the segment
AB. Indeed, if π is any plane containing AB, and C is any point of the circle of π
having AB as diameter, it is known by elementary geometry that the triangle ACB
is rectangle in C.

3 (right angle in B; solution with explicit mention of coordinates) With

−→ −→ −→
this approach, the vectors required to be orthogonal are BA and BC. Since BA =
−→
(−1, 3, 5) and BC = (x − 1, y + 4, z + 4), the following must hold
D E
−→ −→
0 = BA, BC = 1 − x + 3y + 12 + 5z + 20

that is,

x − 3y − 5z = 33

The solution set is the plane π through B and orthogonal to (1, −3, −5).
4 (right angle in B; same solution, with no mention of coordinates) Pro-
ceeding as in the previous point, with the notation of point 2, the condition to be
required is

0 = ha − b, x − bi

that is,

ha − b, xi = ha − b, bi

Thus the solution plane π is seen to be through B and orthogonal to the segment
connecting point B to point A.
Exercises 39

5 (right angle in B) It should be clear at this stage that the solution set in this
case is the plane π 0 through A and orthogonal to AB, of equation

hb − a, xi = hb − a, ai

It is also clear that π and π 0 are parallel.

12.8.10 n. 15 (p. 456)

I c1 − c2 + 2c3 = 0 I + II 3c1 + c3 = 0
II 2c1 + c2 − c3 = 0 I + 2II 5c1 + c2 = 0

c = (−1, 5, 3)

12.8.11 n. 16 (p. 456)

p = (3α, 4α) I 3α + 4β = 1
q = (4β, −3β) II 4α − 3β = 2

4II + 3I 25α = 11 1
(α, β) = (11, −2)
4I − 3II 25β = −2 25

1 1
p= (33, 44) q= (−8, 6)
25 25

12.8.12 n. 17 (p. 456)

The question is identical to the one already answered in exercise 7 of this section.
−→
Recalling solution 2 to that exercise, we have that p ≡ OP must be equal to αb =
−→
αOB, with

ha, bi 10
α= =
hb, bi 4

Thus
µ ¶
5 5 5 5
p = , , ,
2 2 2 2
µ ¶
3 1 1 3
q = − ,− , ,
2 2 2 2
40 Vector algebra

12.8.13 n. 19 (p. 456)

It has been quickly seen in class that
ka + bk2 = kak2 + kbk2 + 2 ha, bi
substituting −b to b, I obtain
ka − bk2 = kak2 + kbk2 − 2 ha, bi
By subtraction,
ka + bk2 − ka − bk2 = 4 ha, bi
as required. You should notice that the above identity has been already used at the
end of point 2 in the solution to exercise 14 of the present section.
Concerning the geometrical interpretation of the special case of the above
identity
ka + bk2 = ka − bk2 if and only if ha, bi = 0
it is enough to notice that orthogonality of a and b is equivalent to the property
for the parallelogram OACB (in the given order around the perimeter, that is, with
vertex C opposed to the vertex in the origin O) to be a rectangle; and that in such
a rectangle ka + bk and ka − bk measure the lengths of the two diagonals.
12.8.14 n. 20 (p. 456)

ka + bk2 + ka − bk2 = ha + b, a + bi + ha − b, a − bi
= ha, ai + 2 ha, bi + hb, bi + ha, ai − 2 ha, bi + hb, bi
= 2 kak2 + 2 kbk2
The geometric theorem expressed by the above identity can be stated as follows:
Theorem 2 In every parallelogram, the sum of the squares of the four sides equals
the sum of the squares of the diagonals.
12.8.15 n. 21 (p. 457)
Exercises 41

Let A, B, C, and D be the four vertices of the quadrilateral, starting from left
−→ −→
in clockwise order, and let M and N be the midpoint of the diagonals AC and DB.
In order to simplify the notation, let
−→ −→ −→ −→
u ≡ AB, v ≡ BC, w ≡ CD, z ≡ DA = − (u + v + w)

Then
−→ −→
c ≡ AC = u + v d ≡ DB = u + z = − (v + w)
−−→ −→ −−→ 1 1
MN = AN − AM = u − d − c
2 2
−−→
2MN = 2u+ (v + w) − (u + v) = w + u
° °
°−−→°2
4 °M N ° = kwk2 + kuk2 + 2 hw, ui
kzk2 = kuk2 + kvk2 + kwk2 + 2 hu, vi + 2 hu, wi + 2 hv, wi
kck2 = kuk2 + kvk2 + 2 hu, vi
kdk2 = kvk2 + kwk2 + 2 hv, wi

We are now in the position to prove the theorem.

¡ ¢ ¡ ¢
kuk2 + kvk2 + kwk2 + kzk2 − kck2 + kdk2
= kzk2 − kvk2 − 2 hu, vi − 2 hv, wi
= kuk2 + kwk2 + 2 hu, wi
° °
°−−→°2
= 4 °MN °

12.8.16 n. 22 (p. 457)

Orthogonality of xa + yb and 4ya − 9xb amounts to

0 = hxa + yb, 4ya − 9xbi

¡ ¢
= −9 ha, bi x2 + 4 ha, bi y 2 + 4 kak2 − 9 kbk2 xy

Since the above must hold for every couple (x, y), choosing x = 0 and y = 1 gives
ha, bi = 0; thus the condition becomes
¡ ¢
4 kak2 − 9 kbk2 xy = 0

and choosing now x = y = 1 gives

4 kak2 = 9 kbk2

Since kak is known to be equal to 6, it follows that kbk is equal to 4.

42 Vector algebra

Finally, since a and b have been shown to be orthogonal,

k2a + 3bk2 = k2ak2 + k3bk2 + 2 h2a, 3bi
= 22 · 62 + 32 · 42 + 2 · 2 · 3 · 0
= 25 · 32
and hence
√
k2a + 3bk = 12 2
12.8.17 n. 24 (p. 457)
This is once again the question raised in exercises 7 and 17, in the general context of
the linear space Rn (where n ∈ N is arbitrary). Since the coordinate-free version of
the solution procedure is completely independent from the number of coordinates of
the vectors involved, the full answer to the problem has already been seen to be
hb, ai
c = a
ha, ai
hb, ai
d = b− a
ha, ai
12.8.18 n. 25 (p. 457)
(a) For every x ∈ R,
ka + xbk2 = kak2 + x2 kbk2 + 2x ha, bi
= kak2 + x2 kbk2 if a ⊥ b
2
≥ kak if a ⊥ b
(b) Since the norm of any vector is nonnegative, the following biconditional is true:
¡ ¢
(∀x ∈ R, ka + xbk ≥ kak) ⇔ ∀x ∈ R, ka + xbk2 ≥ kak2
Moreover, by pure computation, the following biconditional is true:
∀x ∈ R, ka + xbk2 − kak2 ≥ 0 ⇔ ∀x ∈ R, x2 kbk2 + 2x ha, bi ≥ 0

If ha, bi ³is positive,´the trinomial x2 kbk2 + 2x ha, bi is negative for all x in the open
interval − 2ha,bi 2 , 0 ; if it is negative, the trinomial is negative in the open interval
³ ´ kbk
0, − 2ha,bi
kbk2
. It follows that the trinomial can be nonnegative for every x ∈ R only
if ha, bi is zero, that is, only if it reduces to the second degree term x2 kbk2 . In
conclusion, I have proved that the conditional
(∀x ∈ R, ka + xbk ≥ kak) ⇒ ha, bi = 0
is true.
Projections. Angle between vectors in n-space 43

12.9 Projections. Angle between vectors in n-space

12.10 The unit coordinate vectors

12.11 Exercises
12.11.1 n. 1 (p. 460)

ha, bi = 11 kbk2 = 9

The projection of a along b is

µ ¶
11 11 22 22
b= , ,
9 9 9 9

12.11.2 n. 2 (p. 460)

ha, bi = 10 kbk2 = 4

The projection of a along b is

µ ¶
10 5 5 5 5
b= , , ,
4 2 2 2 2

12.11.3 n. 3 (p. 460)

(a)

ci = ha, ii 6
cos a =
kak kik 7
cj = ha, ji = 3
cos a
kak kjk 7
ci = ha, ki = − 2
cos a
kak kkk 7

(b) There are just two vectors as required, the unit direction vector u of a, and its
opposite:
µ ¶
a 6 3 2
u = = , ,−
kak 7 7 7
µ ¶
a 6 3 2
− = − ,− ,
kak 7 7 7
44 Vector algebra

12.11.4 n. 5 (p. 460)

Let

A ≡ (2, −1, 1) B ≡ (1, −3, −5) C ≡ (3, −4, −4)

−→ −→ −→
p ≡ BC = (2, −1, 1) q ≡ CA = (−1, 3, 5) r ≡ AB = (−1, −2, −6)

Then
√ √
b = h−q, ri 35 35 41
cos A =√ √ =
k−qk krk 35 41 41
√ √
b = hp, −ri = √ 6√ = 6 41
cos B
kpk k−rk 6 41 41
b = h−p, qi = √ √
cos C
0
=0
k−pk kqk 6 35
There is some funny occurrence in this exercise, which makes a wrong solution
apparently correct (or almost correct), if one looks only at numerical results. The
b
angles in A, B, C, as implicitly argued above, are more precisely described as B AC,
b ACB;
C BA, b that is, as the three angles of triangle ABC. If some confusion is made
between points and vectors, and/or angles, one may be led into operate directly with
−→ −→ −→
the coordinates of points A, B, C in place of the coordinates of vectors BC, AC, AB,
−→ −→
respectively. This amounts to work, as a matter of fact, with the vectors OA, OB,
−→
OC, therefore computing
D E
−→ −→
OB, OC
° °° ° b an angle of triangle OBC
°−→° °−→° = cos B OC
°OB °° OC °
D E
−→ −→
OC, OA
° °° ° b an angle of triangle OCA
°−→° °−→° = cos C OA
°OC ° °OA°
D E
−→ −→
OA, OB
° °° ° b an angle of triangle OAB
°−→° °−→° = cos AOB
°OA° °OB °

b
instead of cos B AC, b
cos C BA, b
cos ACB, respectively. Up to this point, there is
nothing funny in doing that; it’s only a mistake, and a fairly bad one, being of
conceptual type. The funny thing is in the numerical data of the exercise: it so
happens that

1. points A, B, C are coplanar with the origin O; more than that,

−→ −→ −→ −→
2. OA = BC and OB = AC; more than that,
Exercises 45

b is a right angle.
3. AOB

Point 1 already singles out a somewhat special situation, but point 2 makes
OACB a parallelogram, and point 3 makes it even a rectangle.

−→ −→ −→
OA red, OB blue, OC green, AB violet

It turns out, therefore, that

° ° ° ° ° ° ° ° ° ° ° °
°−→° °−→° °−→° °−→° °−→° °−→°
°OA° = °BC ° = kpk °OB ° = °AC ° = k−qk = kqk °OC ° = °AB ° = krk

b = C BA
C OA b b = B AC
B OC b b = ACB
AOB b

and such a special circumstance leads a wrong solution to yield the right numbers.
12.11.5 n. 6 (p. 460)
Since

ka + c ± bk2 = ka + ck2 + kbk2 ± 2 ha + c, bi

from

ka + c + bk = ka + c − bk

it is possible to deduce

ha + c, bi = 0

This is certainly true, as a particular case, if c = −a, which immediately implies

cc = π
a
c = 7π
c = π − ab
bc
8
46 Vector algebra

Moreover, even if a + c = 0 is not assumed, the same conclusion holds. Indeed, from

hc, bi = − ha, bi

and

kck = kak

it is easy to check that

c= hb, ci ha, ci c
cos bc =− = − cos ab
kbk kck kbk kak

and hence that bcc =π±a cc = 78 π, 98 π (the second value being superfluous).
12.11.6 n. 8 (p. 460)
We have
r
√ n (n + 1) (2n + 1) n (n + 1)
kak = n kbn k = ha, bn i =
v 6 2
u 2 r
n(n+1) u n (n+1)
2
3 n+1
cos [a bn = √ q 2 = t n2 (n+1)(2n+1)
4
=
2 2n + 1
n n(n+1)(2n+1)
6 6
√
3 π
lim cos [ a bn = lim [ a bn =
n→+∞ 2 n→+∞ 3
12.11.7 n. 10 (p. 461)
(a)

ha, bi = cos ϑ sin ϑ − cos ϑ sin ϑ = 0

kak2 = kbk2 = cos2 ϑ + sin2 ϑ = 1

(b) The system

µ ¶µ ¶ µ ¶
cos ϑ sin ϑ x x
=
− sin ϑ cos ϑ y y

that is,
µ ¶µ ¶ µ ¶
cos ϑ − 1 sin ϑ x 0
=
− sin ϑ cos ϑ − 1 y 0

has the trivial solution as its unique solution if

¯ ¯
¯ cos ϑ − 1 sin ϑ ¯
¯ ¯ 6= 0
¯ − sin ϑ cos ϑ − 1 ¯
Exercises 47

The computation gives

1 + cos2 ϑ + sin2 ϑ − 2 cos ϑ 6= 0

2 (1 − cos ϑ) 6 = 0
cos ϑ 6 = 1
ϑ ∈
/ {2kπ}k∈Z

Thus if

ϑ ∈ {(−2kπ, (2k + 1) π)}k∈Z

the only vector satisfying the required condition is (0, 0). On the other hand, if

ϑ ∈ {2kπ}k∈Z

the coefficient matrix of the above system is the identity matrix, and every vector in
R2 satisfies the required condition.
12.11.8 n. 11 (p. 461)
Let OABC be a rhombus, which I have taken (without loss of generality) with one
vertex in the origin O and with vertex B opposed to O. Let
−→ −→ −→
a ≡ OA (red) b ≡ OB (green) c ≡ OC (blue)

Since OABC is a parallelogram, the oriented segments AB (blue, dashed) and OB

are congruent, and the same is true for CB (red, dashed) and OA. Thus
−→ −→
AB = b CB = a

From elementary geometry (see exercise 12 of section 4) the intersection point M of

the two diagonals OB (green) and AC (violet) is the midpoint of both.
The assumption that OABC is a rhombus is expressed by the equality

kak = kck ((rhombus))

48 Vector algebra

The statement to be proved is orthogonality between the diagonals

D E
−→ −→
OB, AC = 0

that is,

ha + c, c − ai = 0

or

kck2 − kak2 + ha, ci − hc, ai = 0

and hence (by commutativity)

kck2 = kak2

The last equality is an obvious consequence of ((rhombus)). As a matter of fact, since

norms are nonnegative real numbers, the converse is true, too. Thus a parallelogram
has orthogonal diagonals if and only if it is a rhombus.
12.11.9 n. 13 (p. 461)
The equality to be proved is straightforward. The “law of cosines” is often called

Theorem 3 (Carnot) In every triangle, the square of each side is the sum of the
squares of the other two sides, minus their double product multiplied by the cosine of
the angle they form.

The equality in exam can be readily interpreted according to the theorem’s

−→ −→
statement, since in every parallelogram ABCD with a = AB and b = AC the
−→
diagonal vector CB is equal to a − b, so that the triangle ABC has side vectors a,
b, and a − b. The theorem specializes to Pythagoras’ theorem when a ⊥ b.
12.11.10 n. 17 (p. 461)
(a) That the function
X
Rn → R, a 7→ |ai |
i∈n

is positive can be seen by the same arguments used for the ordinary norm; nonnega-
tivity is obvious, and strict positivity still relies on the fact that a sum of concordant
numbers can only be zero if all the addends are zero. Homogeneity is clear, too:

∀a ∈ Rn , ∀α ∈ R,
X X X
kαak = |αai | = |α| |ai | = |α| |ai | = |α| kak
i∈n i∈n i∈n
Exercises 49

Finally, the triangle inequality is much simpler to prove for the present norm (some-
times referred to as “the taxi-cab norm”) than for the euclidean norm:

∀a ∈ Rn , ∀b ∈ Rn ,
X X X X
ka + bk = |ai + bi | ≤ |ai | + |bi | = |ai | + |bi |
i∈n i∈n i∈n i∈n
= kak + kbk

(b) The subset of R2 to be described is

© ª
S ≡ (x, y) ∈ R2 : |x| + |y| = 1
= S++ ∪ S−+ ∪ S−− ∪ S+−

where
© ª
S++ ≡ (x, y) ∈ R2+ : x + y = 1 (red)
S−+ ≡ {(x,
© y) ∈ R−2 × R+ : −x + ªy = 1} (green)
S−− ≡ (x, y) ∈ R− : −x − y = 1 (violet)
S+− ≡ {(x, y) ∈ R+ × R− : x − y = 1} (blue)
Once the lines whose equations appear in the definitions of the four sets above are
drawn, it is apparent that S is a square, with sides parallel to the quadrant bisectrices

-2 -1 0 1 2

-1

-2

(c) The function

X
f : Rn → R, a 7→ |ai |
i∈n

is nonnegative, but not positive (e.g., for n = 2, f (x, −x) = 0 ∀x ∈ R). It is

homogeneous:

∀a ∈ Rn , ∀α ∈ R,
¯ ¯ ¯ ¯ ¯ ¯
¯X ¯ ¯ X ¯ ¯X ¯
¯ ¯ ¯ ¯ ¯ ¯
kαak = ¯ αai ¯ = ¯α ai ¯ = |α| ¯ ai ¯ = |α| kak
¯ ¯ ¯ ¯ ¯ ¯
i∈n i∈n i∈n
50 Vector algebra

Finally, f is subadditive, too (this is another way of saying that the triangle inequality
holds). Indeed,

∀a ∈ Rn , ∀b ∈ Rn ,
¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
¯X ¯ ¯X X ¯¯ ¯¯X ¯¯ ¯¯X ¯¯
¯ ¯ ¯
ka + bk = ¯ (ai + bi )¯ = ¯ ai + bi ¯ ≤ ¯ ai ¯ + ¯ bi ¯
¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
i∈n i∈n i∈n i∈n i∈n
= kak + kbk

12.12 The linear span of a finite set of vectors

12.13 Linear independence

12.14 Bases

12.15 Exercises
12.15.1 n. 1 (p. 467)

x (i − j) + y (i + j) = (x + y, y − x)

¡1 1
¢
(a) x + y = 1 and y − x = 0 yield (x, y) = ,
2 2
.
¡ ¢
(b) x + y = 0 and y − x = 1 yield (x, y) = − 12 , 12 .

(c) x + y = 3 and y − x = −5 yield (x, y) = (4, −1).

(d) x + y = 7 and y − x = 5 yield (x, y) = (1, 6).

12.15.2 n. 3 (p. 467)

I 2x + y = 2 I + III 3x = 9
II −x + 2y = −11 II + III y = −4
III x − y = 7 III (check) 3 + 4 = 7

The solution is (x, y) = (3, −4).

Exercises 51

12.15.3 n. 5 (p. 467)

(a) If there exists some α ∈ R ∼ {0} such that∗ a = αb, then the linear combination
1a − αb is nontrivial and it generates the null vector; a and b are linearly dependent.
(b) The argument is best formulated by counterposition, by proving that if a and b
are linearly dependent, then they are parallel. Let a and b nontrivially generate the
null vector: αa + βb = 0 (according to the definition, at least one between α and β
is nonzero; but in the present case they are both nonzero, since a and b have been
assumed both different from the null vector; indeed, αa + βb = 0 with α = 0 and
β 6= 0 implies b = 0, and αa + βb = 0 with β = 0 and α 6= 0 implies a = 0). Thus
αa = −βb, and hence a = − αβ b (or b = − αβ a, if you prefer).
12.15.4 n. 6 (p. 467)
Linear independency of the vectors (a, b) and (c, d) has been seen in the lectures
(proof of Steinitz’ theorem, part 1) to be equivalent to non existence of nontrivial
solutions to the system
ax + cy = 0
bx + dy = 0
The system is linear and homogeneous, and has unknowns and equations in equal
number (hence part 2 of the proof of Steinitz’ theorem does not apply). I argue by
the principle of counterposition. If a nontrivial solution (x, y) exists, both (a, c) and
(b, d) must be proportional to (y, −x) (see exercise 10, section 8), and hence to each
other. Then, from (a, c) = h (b, d) it is immediate to derive ad − bc = 0. The converse
is immediate, too
12.15.5 n. 7 (p. 467)
By the previous exercise, it is enough to require
(1 + t)2 − (1 − t)2 =
6 0
4t = 6 0
t = 6 0
12.15.6 n. 8 (p. 467)
(a) The linear combination
1i + 1j + 1k + (−1) (i + j + k)
is nontrivial and spans the null vector.
(b) Since, for every (α, β, γ) ∈ R3 ,
αi + βj + γk = (α, β, γ)
∗
Even if parallelism is defined more broadly − see the footnote in exercise 9 of section 4 − α
cannot be zero in the present case, because both a and b have been assumed different from the null
vector.
52 Vector algebra

it is clear that

∀ (α, β, γ) ∈ R3 , (αi + βj + γk = 0) ⇒ α = β = γ = 0

so that the triple (i, j, k) is linearly independent.

(c) Similarly, for every (α, β, γ) ∈ R3 ,

αi + βj + γ (i + j + k) = (α + γ, β + γ, γ)

and hence from

αi + βj + γ (i + j + k) = 0

it follows

α+γ =0 β+γ =0 γ =0

that is,

α=β=γ=0

showing that the triple (i, j, i + j + k) is linearly independent.

(d) The last argument can be repeated almost verbatim for triples (i, i + j + k, k)
and (i + j + k, j, k), taking into account that

αi + β (i + j + k) + γk = (α + β, β, β + γ)
α (i + j + k) + βj + γk = (α, α + β, α + γ)

12.15.7 n. 10 (p. 467)

(a) Again from the proof of Steinitz’ theorem, part 1, consider the system

I x+y+z =0
II y + z = 0
III 3z = 0

It is immediate to derive that its unique solution is the trivial one, and hence that
the given triple is linearly independent.
(b) We need to consider the following two systems:

I x+y+z =0 I x+y+z =0
II y + z = 1 II y + z = 0
III 3z = 0 III 3z = 1

It
¡ is 1again ¢ immediate that the unique solutions to the systems are (−1, 1, 0) and
0, − 3 , 13 respectively.
Exercises 53

(c) The system to study is the following:

I x+y+z =2 III z = 53
II y + z = −3 (↑) ,→ II y = − 14
3
III 3z = 5 (↑) ,→ I x = 5

(d) For an arbitrary triple (a, b, c), the system

I x+y+z = a
II y + z = b
III 3z = c
¡ ¢
has the (unique) solution a − b, b − 3c , 3c . Thus the given triple spans R3 , and it is
linearly independent (as seen at a).
12.15.8 n. 12 (p. 467)
(a) Let xa + yb + zc = 0, that is,

x+z = 0
x+y+z = 0
x+y = 0
y = 0

then y, x, and z are in turn seen to be equal to 0, from the fourth, third, and first
equation in that order. Thus (a, b, c) is a linearly independent triple.
(b) Any nontrivial linear combination d of the given vectors makes (a, b, c, d) a
linearly dependent quadruple. For example, d ≡ a + b + c = (2, 3, 2, 1)
(c) Let e ≡ (−1, −1, 1, 3), and suppose xa + yb + zc + te = 0, that is,

x+z−t = 0
x+y+z−t = 0
x+y+t = 0
y + 3t = 0

Then, subtracting the first equation from the second, gives y = 0, and hence t, x,
and z are in turn seen to be equal to 0, from the fourth, third, and first equation in
that order. Thus (a, b, c, e) is linearly independent.
(d) The coordinates of x with respect to {a, b, c, e}, which has just been seen to be
a basis of R4 , are given as the solution (s, t, u, v) to the following system:

s+u−v = 1
s+t+u−v = 2
s+t+v = 3
t + 3v = 4
54 Vector algebra

It is more direct and

¡ orderly
¢ to work just with the table formed by the system
(extended) matrix A x , and to perform elementary row operations and column
exchanges.
 
s t u v x
 
 
 1 0 1 −1 1 
 
 1 1 1 −1 2 
 
 1 1 0 1 3 
0 1 0 3 4
a1 ↔ a3 , 2 a0 ← 2 a0 − 1 a0
 
u t s v x
 
 
 1 0 1 −1 1 
 
 0 1 0 0 1 
 
 0 1 1 1 3 
0 1 0 3 4
a2 ↔ a3 , 2 a0 ↔ 3 a0
 
u s t v x
 
 
 1 1 0 −1 1 
 
 0 1 1 1 3 
 
 0 0 1 0 1 
0 0 1 3 4
0
4a ← 4 a0 − 3 a0
 
u s t v x
 
 
 1 1 0 −1 1 
 
 0 1 1 1 3 
 
 0 0 1 0 1 
0 0 0 3 3
Thus the system has been given the following form:
u+s−v = 1
s+t+v = 3
t = 1
3v = 3
which is easily solved from the bottom to the top: (s, t, u, v) = (1, 1, 1, 1).
Exercises 55

12.15.9 n. 13 (p. 467)

(a)
¡√ ¢ ¡ √ ¢ ¡ √ ¢
α 3, 1, √ 0 + β 1, 3, 1 + γ 0, 1, 3 = (0, 0, 0)
I 3α√+ β = 0
⇔ II α + √3β + γ = 0
III β + 3γ = 0√
I√ 3α + β = 0
⇔ 3II − III − I 2β =√0
III β + 3γ = 0
⇔ (α, β, γ) = (0, 0, 0)

and the three given vectors are linearly independent.

(b)
¡√ ¢ ¡ √ ¢ ¡ √ ¢
α 2, 1,
√ 0 + β 1, 2, 1 + γ 0, 1, 2 = (0, 0, 0)
I 2α√+ β = 0
⇔ II α + √2β + γ = 0
III β + 2γ = 0
√
This time the sum of equations I and III (multiplied by 2) is the same as twice
equation II, and a linear dependence in¡√ the equation
√ ¢ system suggests that it may
well have nontrivial solutions. Indeed, 2, −2, 2 is such a solution. Thus the
three given triple of vectors is linearly dependent.
(c)

α (t, 1, 0) + β (1, t, 1) + γ (0, 1, t) = (0, 0, 0)

I tα + β = 0
⇔ II α + tβ + γ = 0
III β + tγ = 0

It is clear that t = 0 makes the triple linearly dependent (the first and third vector
coincide in this case). Let us suppose, then, t 6= 0. From I and III, as already
noticed, I deduce that α = γ. Equations II and III then become

tβ + 2γ = 0
β + tγ = 0

2
a 2 by 2 homogeneous system with determinant©√ of coefficient
√ ª matrix equal to t − 2.
Such a system has nontrivial solutions for t ∈ 2, − 2 . In conclusion, the given
© √ √ ª
triple is linearly dependent for t ∈ 0, 2, − 2 .
56 Vector algebra

12.15.10 n. 14 (p. 468)

Call as usual u, v, w, and z the four vectors given in each case, in the order.
(a) It is clear that v = u + w, so that v can be dropped. Moreover, every linear
combination of u and w has the form (x, y, x, y), and cannot be equal to z. Thus
(u, w, z) is a maximal linearly independent triple.
(b) Notice that
1 1
(u + z) = e(1) (u − v) = e(2)
2 2
1 1
(v − w) = e(3) (w − z) = e(4)
2 2
Since (u, v, w, z) spans the four canonical vectors, it is a basis of R4 . Thus (u, v, w, z)
is maximal linearly independent.
(c) Similarly,

u − v = e(1) v − w = e(2)
w − z = e(3) z = e(4)

and (u, v, w, z) is maximal linearly independent.

12.15.11 n. 15 (p. 468)
(a) Since the triple (a, b, c) is linearly independent,

α (a + b) + β (b + c) + γ (a + c) = 0
⇔ (α + γ) a + (α + β) b + (β + γ) c = 0
I α+γ =0
⇔ II α+β =0
III β + γ = 0
I + II − III 2α = 0
⇔ −I + II + III 2β = 0
I − II + III 2γ = 0

and the triple (a + b, b + c, a + c) is linearly independent, too

(b) On the contrary, choosing as nontrivial coefficient triple (α, β, γ) ≡ (1, −1, 1),

(a − b) − (b + c) + (a + c) = 0

it is seen that the triple (a − b, b + c, a + c) is linearly dependent.

12.15.12 n. 17 (p. 468)
Let a ≡ (0, 1, 1) and b ≡ (1, 1, 1); I look for two possible alternative choices of a
vector c ≡ (x, y, z) such that the triple (a, b, c) is linearly independent. Since

αa + βb + γc = (β + γx, α + β + γy, α + β + γz)

Exercises 57

my choice of x, y, and z must be such to make the following conditional statement

true for each (α, β, γ) ∈ R3 :
 
I β + γx = 0   α=0
II α + β + γy = 0 ⇒ β=0
 
III α + β + γz = 0 γ=0

Subtracting equation III from equation II, I obtain

γ (y − z) = 0

Thus any choice of c with y 6= z makes γ = 0 a consequence of II-III; in such a case,

I yields β = 0 (independently of the value assigned to x), and then either II or III
yields α = 0. Conversely, if y = z, equations II and III are the same, and system
I-III has infinitely many nontrivial solutions

α = −γy − β
β = −γx
γ free

provided either x or y (hence z) is different from zero.

As an example, possible choices for c are (0, 0, 1), (0, 1, 0), (1, 0, 1), (1, 1, 0).
12.15.13 n. 18 (p. 468)
The first example of basis containing the two given vectors is in point (c) of exercise
14 in this section, since the vectors u and v there coincide with the present ones.
Keeping the same notation, a second example is (u, v, w + z, w − z).
12.15.14 n. 19 (p. 468)
(a) It is enough to prove that each element of T belongs to lin S, since each element
of lin T is a linear combination in T , and S, as any subspace of a vector space, is
closed with respect to formation of linear combinations. Let u, v, and w be the three
elements of S (in the given order), and let a and b be the two elements of T (still in
the given order); it is then readily checked that

a=u−w b = 2w

(b) The converse inclusion holds as well. Indeed,

1 1
v =a−b w= b u=v+w = a− b
2 2
Thus

lin S = lin T
58 Vector algebra

Similarly, if c and d are the two elements of U ,

c=u+v d = u + 2v
which proves that lin U ⊆ lin S. Inverting the above formulas,
u = 2c − d v =d−c
It remains to be established whether or not w is an element of lin U . Notice that
αc + βd = (α + β, 2α + 3β, 3α + 5β)
It follows that w is an element of lin U if and only if there exists (α, β) ∈ R2 such
that

 I α+β =1
II 2α + 3β = 0

III 3α + 5β = −1
The above system has the unique solution (3, −2), which proves that lin S ⊆ lin U .
In conclusion,
lin S = lin T = lin U
12.15.15 n. 20 (p. 468)
(a) The claim has already been proved in the last exercise, since from A ⊆ B I can
infer A ⊆ lin B, and hence lin A ⊆ lin B.
(b) By the last result, from A ∩ B ⊆ A and A ∩ B ⊆ B I infer
lin A ∩ B ⊆ lin A lin A ∩ B ⊆ lin B
which yields
lin A ∩ B ⊆ lin A ∩ lin B
(c) It is enough to define
A ≡ {a, b} B ≡ {a + b}
where the couple (a, b) is linearly independent. Indeed,
B ⊆ lin A
and hence, by part (a) of last exercise,
lin B ⊆ lin lin A = lin A
lin A ∩ lin B = lin B
On the other hand,
A∩B =∅ lin A ∩ B = {0}
The vector space Vn (C) of n-tuples of complex numbers 59

12.16 The vector space Vn (C) of n-tuples of complex numbers

12.17 Exercises
60 Vector algebra
 Chapter 13
APPLICATIONS OF VECTOR ALGEBRA TO ANALYTIC
GEOMETRY

13.1 Introduction

13.2 Lines in n-space

13.3 Some simple properties of straight lines

13.4 Lines and vector-valued functions

13.5 Exercises
13.5.1 n. 1 (p. 477)
−→
A direction vector for the line L is P Q = (4, 0), showing that L is horizontal. Thus
a point belongs to L if and only if its second coordinate is equal to 1. Among the
given points, (b), (d), and (e) belong to L.
13.5.2 n. 2 (p. 477)
−→
A direction vector for the line L is v ≡ 12 P Q = (−2, 1). The parametric equations
for L are

x = 2 − 2t
y = −1 + t

If t = 1 I get point (a) (the origin). Points (b), (d) and (e) have the second coordinate
equal to 1, which requires t = 2. This gives x = −2, showing that of the three points
only (e) belongs to L. Finally, point (c) does not belong to L, because y = 2 requires
t = 3, which yields x = −4 6= 1.
62 Applications of vector algebra to analytic geometry

13.5.3 n. 3 (p. 477)

The parametric equations for L are

x = −3 + h
y = 1 − 2h
z = 1 + 3h

The following points belong to L:

(c) (h = 1) (d) (h = −1) (e) (h = 5)

13.5.4 n. 4 (p. 477)

The parametric equations for L are

x = −3 + 4k
y = 1+k
z = 1 + 6h

The following points belong to L:

µ ¶ µ ¶
1 1
(b) (h = −1) (e) h= (f ) h=
2 3
−→
A direction vector for the line L is P Q = (4, 0), showing that L is horizontal.
Thus a point belongs to L if and only if its second coordinate is equal to 1. Among
the given points, (b), (d), and (e) belong to L.
13.5.5 n. 5 (p. 477)
I solve each case in a different a way.
(a)
−→ −→
P Q = (2, 0, −2) QR = (−1, −2, 2)

The two vectors are not parallel, hence the three points do not belong to the same
line.
(b) Testing affine dependence,

I 2h + 2k + 3l = 0 I − 2II + III 2l = 0
II −2h + 3k + l = 0 I + II 5k + 4l = 0
III −6h + 4k + l = 0 2I − III 10h + 5l = 0

the only combination which is equal to the null vector is the trivial one. The three
points do not belong to the same line.
Exercises 63

(c) The line through P and R has equations

x = 2 + 3k
y = 1 − 2k
z = 1
Trying to solve for k with the coordinates of Q, I get k = − 43 from the first equation
and k = −1 from the second; Q does not belong to L (P, R).
13.5.6 n. 6 (p. 477)
The question is easy, but it must be answered by following some orderly ¡path, ¢ in
order to achieve some economy of thought and of computations (there are 82 = 28
different oriented segments joining two of the eight given points. First, all the eight
points have their third coordinate equal to 1, and hence they belong to the plane
π of equation z = 1. We can concentrate only on the first two coordinates, and,
as a matter of fact, we can have a very good hint on the situation by drawing a
twodimensional picture, to be considered as a picture of the π

15

10

-15 -10 -5 0 5 10 15

-5

-10

-15

−→
Since the first two components of AB are (4, −2), I check that among the
twodimensional projections of the oriented segments connecting A with points D to
H
−→ −→ −→
pxy AD = (−4, 2) pxy AE = (−1, 1) pxy AF = (−6, 3)
−→ −→
pxy AG = (−15, 8) pxy AH = (12, −7)
64 Applications of vector algebra to analytic geometry

−→
only the first and third are parallel to pxy AB. Thus all elements of the set P 1 ≡
{A, B, C, D, F } belong to the same (black) line LABC , and hence no two of them can
both belong to a different line. Therefore, it only remains to be checked whether or
not the lines through the couples of points (E, G) (red), (G, H) (blue), (H, E) (green)
coincide, and whether or not any elements of P 1 belong to them. Direction vectors
for these three lines are
1 −→ 1 −→ −−→
pxy EG = (−7, 4) pxy GH = (9, −5) pxy HE = (−13, 7)
2 3
and as normal vectors for them I may take
nEG ≡ (4, 7) nGH ≡ (5, 9) nHE ≡ (7, 13)
By requiring point E to belong to the first and last, and point H to the second, I end
up with their equations as follows
LEG : 4x + 7y = 11 LGH : 5x + 9y = 16 LHE : 7x + 13y = 20
All these three lines are definitely not parallel to LABC , hence they intersect it exactly
at one point. It is seen by direct inspection that, within the set P 1 , C belongs to
LEG , F belongs to LGH , and neither A nor B, nor D belong to LHE .
Thus there are three (maximal) sets of at least three collinear points, namely
P 1 ≡ {A, B, C, D, F } P 2 ≡ {C, E, G} P 3 ≡ {F, G, H}
13.5.7 n. 7 (p. 477)
The coordinates of the intersection point are determined by the following equation
system
I 1 + h = 2 + 3k
II 1 + 2h = 1 + 8k
III 1 + 3h = 13k
Subtracting the third equation from the sum of the first two, I get k = 1; substituting
this value in any of the three equations, I get h = 4. The three equations are
consistent, and the two lines intersect at the point of coordinates (5, 9, 13).
13.5.8 n. 8 (p. 477)
(a) The coordinates of the intersection point of L (P ; a) and L (Q; b) are determined
by the vector equation
P + ha = Q + kb (13.1)
which gives
P − Q = kb − ha
that is,
−→
P Q ∈ span {a, b}
Exercises 65

13.5.9 n. 9 (p. 477)

X (t) = (1 + t, 2 − 2t, 3 + 2t)

(a)
d (t) ≡ kQ − X (t)k2 = (2 − t)2 + (1 + 2t)2 + (−2 − 2t)2
= 9t2 + 8t + 9
(b) The graph of ¡the function ¢ t 7→ d (t) ≡ 9t2 + 8t + 9 is a parabola with the
4 65
point (t0 , d (t0 )) = −√ ,
9 9
as vertex. The minimum squared distance is 65
9
, and the
65
minimum distance is 3 .
(c)
µ ¶ µ ¶
5 26 19 22 1 10
X (t0 ) = , , Q − X (t0 ) = , ,−
9 9 9 9 9 9

22 − 2 + 20
hQ − X (t0 ) , Ai = =0
9
that is, the point on L of minimum distance from Q is the orthogonal projection of
Q on L.
13.5.10 n. 10 (p. 477)
(a) Let A ≡ (α, β, γ), P ≡ (λ, µ, ν) and Q ≡ (%, σ, τ ). The points of the line L
−→
through P with direction vector OA are represented in parametric form (the generic
point of L is denoted X (t)). Then
X (t) ≡ P + At = (λ + αt, µ + βt, ν + γt)
f (t) ≡ kQ − X (t)k2 = kQ − P − Atk2
= kQk2 + kP k2 + kAk2 t − 2 hQ, P i + 2 hP − Q, Ai t
= at2 + bt + c
where
a ≡ kAk2 = α2 + β 2 + γ 2
b
≡ hP − Q, Ai = α (λ − %) + β (µ − σ) + γ (ν − τ )
2
c ≡ kQk2 + kP k2 − 2 hQ, P i = kP − Qk2 = (λ − %)2 + (µ − σ)2 + (ν − τ )2
The above quadratic polynomial has a second degree term with a positive coefficient;
its graph is a parabola with vertical axis and vertex in the point of coordinates
µ ¶ Ã 2 2 2
!
b ∆ hQ − P, Ai kAk kP − Qk − hP − Q, Ai
− ,− = ,
2a 4a kAk2 kAk2
66 Applications of vector algebra to analytic geometry

Thus the minimum value of this polynomial is achieved at

hQ − P, Ai α (% − λ) + β (σ − β) + γ (τ − ν)
t0 ≡ 2 =
kAk α2 + β 2 + γ 2
and it is equal to
kAk2 kP − Qk2 − kAk2 kP − Qk2 cos2 ϑ
2 = kP − Qk2 sin2 ϑ
kAk
where
−→
\ −→
ϑ ≡ OA, QP
is the angle formed by direction vector of the line and the vector carrying the point
Q not on L to the point P on L.
(b)
hQ − P, Ai
Q − X (t0 ) = Q − (P + At0 ) = (Q − P ) − A
kAk2
hQ − P, Ai
hQ − X (t0 ) , Ai = hQ − P, Ai − hA, Ai = 0
kAk2
13.5.11 n. 11 (p. 477)
The vector equation for the coordinates of an intersection point of L (P ; a) and
L (Q; a) is
P + ha = Q + ka (13.2)
It gives
P − Q = (h − k) a
and there are two cases: either
−→
PQ ∈
/ span {a}
and equation (13.2) has no solution (the two lines are parallel), or
−→
P Q ∈ span {a}
that is,
∃c ∈ R, Q − P = ca
and equation (13.2) is satisfied by all couples (h, k) such that k − h = c (the two lines
intersect in infinitely many points, i.e., they coincide). The two expressions P + ha
and Q + ka are seen to provide alternative parametrizations of the same line. For
each given point of L, the transformation h 7→ k (h) ≡ h + c identifies the parameter
change which allows to shift from the first parametrization to the second.
Planes in euclidean n-spaces 67

13.5.12 n. 12 (p. 477)

13.6 Planes in euclidean n-spaces

13.7 Planes and vector-valued functions

13.8 Exercises
13.8.1 n. 2 (p. 482)
We have:
−→ −→
P Q = (2, 2, 3) P R = (2, −2, −1)

so that the parametric equations of the plane are

x = 1 + 2s + 2t (13.3)
y = 1 + 2s − 2t
z = −1 + 3s − t
¡ ¢
(a) Equating (x, y, z) to 2, 2, 12 in (13.3) we get

2s + 2t = 1
2s − 2t = 1
3
3s − t =
2
¡ ¢
yielding s = 12 and t = 0, so that the point with coordinates 2, 2, 12 belongs to the
plane. ¡ ¢
(b) Similarly, equating (x, y, z) to 4, 0, 32 in (13.3) we get

2s + 2t = 3
2s − 2t = −1
1
3s − t =
2
¡ ¢
yielding s = 12 and t = 1, so that the point with coordinates 4, 0, 32 belongs to the
plane.
(c) Again, proceeding in the same way with the triple (−3, 1, −1), we get

2s + 2t = −4
2s − 2t = 0
3s − t = −2
68 Applications of vector algebra to analytic geometry

yielding s = t = −1, so that the point with coordinates (−3, 1, −1) belongs to the
plane.
(d) The point of coordinates (3, 1, 5) does not belong to the plane, because the system

2s + 2t = 2
2s − 2t = 0
3s − t = 4

is inconsistent (from the first two equations we get s = t = 1, contradicting the third
equation).
(e) Finally, the point of coordinates (0, 0, 2) does not belong to the plane, because
the system

2s + 2t = −1
2s − 2t = −1
3s − t = 1

is inconsistent (the first two equations yield s = − 12 and t = 0, contradicting the

third equation).
13.8.2 n. 3 (p. 482)
(a)

x = 1+t
y = 2+s+t
z = 1 + 4t

(b)

u = (1, 2, 1) − (0, 1, 0) = (1, 1, 1)

v = (1, 1, 4) − (0, 1, 0) = (1, 0, 4)

x = s+t
y = 1+s
z = s + 4t
Exercises 69

13.8.3 n. 4 (p. 482)

(a) Solving for s and t with the coordinates of the first point, I get

I s − 2t + 1 = 0 I + II − III t + 3 = 0
II s + 4t + 2 = 0 2III − I − II 2s − 3 = 0
3
III 2s + t = 0 check on I 2
− 6 + 1 6= 0

and (0, 0, 0) does not belong to the plane M . The second point is directly seen to
belong to M from the parametric equations

(1, 2, 0) + s (1, 1, 2) + t (−2, 4, 1) (13.4)

Finally, with the third point I get

I + II − III t+3=2
I s − 2t + 1 = 2 2III − I − II 2s − 3 = −5
II s + 4t + 2 = −3 check on I −1 + 2 + 1 = 2
III 2s + t = −3 check on II −1 − 4 + 2 = −3
check on III −2 − 1 = −3

and (2, −3, −3) belongs to M .

(b) The answer has already been given by writing equation (13.4):

P ≡ (1, 2, 0) a = (1, 1, 2) b = (−2, 4, 1)

13.8.4 n. 5 (p. 482)

This exercise is a replica of material already presented in class and inserted in the
notes.
(a) If p + q + r = 1, then

pP + qQ + rR = P − (1 − p) P + qQ + rR
= P + (q + r) P + qQ + rR
= P + q (Q − P ) + r (R − P )

and the coordinates of pP + qQ + rR satisfy the parametric equations of the plane

M through P , Q, and R..
(b) If S is a point of M, there exist real numbers q and r such that

S = P + q (Q − P ) + r (R − P )

Then, defining p ≡ 1 − (q + r),

S = (1 − q − r) P + qQ + rR
= pP + qQ + rR
70 Applications of vector algebra to analytic geometry

13.8.5 n. 6 (p. 482)

I use three different methods for the three cases.
(a) The parametric equations for the first plane π 1 are

I x = 2 + 3h − k
II y = 3 + 2h − 2k
III z = 1 + h − 3k

Eliminating the parameter h (I − II − III) I get

4k = x − y − z + 2

Eliminating the parameter k (I + II − III) I get

4h = x + y − z − 4

Substituting in III (or, better, in 4 · III),

4z = 4 + x + y − z − 4 − 3x + 3y + 3z − 6

I obtain a cartesian equation for π 1

x − 2y + z + 3 = 0

(b) I consider the four coefficients (a, b, c, d) of the cartesian equation of π 2 as un-
known, and I require that the three given points belong to π 2 ; I obtain the system

2a + 3b + c + d = 0
−2a − b − 3c + d = 0
4a + 3b − c + d = 0

which I solve by elimination

 
2 3 1 1
 −2 −1 −3 1 
4 3 −1 1
 
µ 0 1 0 0
¶ 2 3 1 1
2a ← ( 2a + 1a )
2  0 1 −1 1 
00 0
3a ← 3a − 2 1a
0 −3 −3 −1
 
¡ ¢ 2 3 1 1
3a
0
← 12 ( 3 a0 + 3 1 a0 )  0 1 −1 1 
0 0 −3 1
Exercises 71

From the last equation (which reads −3c + d = 0) I assign values 1 and 3 to c and d,
respectively; from the second (which reads b − c + d = 0) I get b = −2, and from the
first (which is unchanged) I get a = 1. The cartesian equation of π 2 is
x − 2y + z + 3 = 0
(notice that π 1 and π 2 coincide)
(c) This method requires the vector product (called cross product by Apostol ), which
is presented in the subsequent section; however, I have thought it better to show its
application here already. The plane π3 and the given plane being parallel, they have
the same normal direction. Such a normal is
n = (2, 0, −2) × (1, 1, 1) = (2, −4, 2)
Thus π 3 coincides with π2 , since it has the same normal and has a common point
with it. At any rate (just to show how the method applies in general), a cartesian
equation for π 3 is
x − 2y + z + d = 0
and the coefficient d is determined by the requirement that π 3 contains (2, 3, 1)
2−6+1+d =0
yielding
x − 2y + z + 3 = 0
13.8.6 n. 7 (p. 482)
(a) Only the first two points belong to the given plane.
(b) I assign the values −1 and 1 to y and z in the cartesian equation of the plane M ,
in order to obtain the third coordinate of a point of M , and I obtain P = (1, −1, 1) (I
may have as well taken one of the first two points of part a). Any two vectors which
are orthogonal to the normal n = (3, −5, 1) and form a linearly independent couple
can be chosen as direction vectors for M . Thus I assign values arbitrarily to d2 and
d3 in the orthogonality condition
3d1 − 5d2 + d3 = 0
say, (0, 3) and (3, 0), and I get
dI = (−1, 0, 3) dII = (5, 3, 0)
The parametric equations for M are
x = 1 − s + 5t
y = −1 + 3t
z = 1 + 3s
72 Applications of vector algebra to analytic geometry

13.8.7 n. 8 (p. 482)

The question is formulated in a slightly insidious way (perhaps Tom did it on pur-
pose...), because the parametric equations of the two planes M and M 0 are written
using the same names for the two parameters. This may easily lead to an incorrect
attempt two solve the problem by setting up a system of three equations in the two
unknown s and t, which is likely to have no solutions, thereby suggesting the wrong
answer that M and M 0 are parallel. On the contrary, the equation system for the
coordinates of points in M ∩ M 0 has four unknowns:
I 1 + 2s − t = 2 + h + 3k
II 1 − s = 3 + 2h + 2k (13.5)
III 1 + 3s + 2t = 1 + 3h + k
I take all the variables on the lefthand side, and the constant terms on the righthand
one, and I proceed by elimination:
s t h k | const.
2 −1 −1 −3 | 1
−1 0 −2 −2 | 2
3 2 −3 −1 | 0

 0
 s t h k | const.
1r ← 1 r0 + 2 2 r0
 3 r0 ← 3 r0 + 3 2 r0  −1 0 −2 −2 | 2
0 0 0 −1 −5 −7 | 5
1r ↔ 2r
0 2 −9 −7 | 6

µ ¶ s t h k | const.
0 0 0
3r ← 3r + 2 2r −1 0 −2 −2 | 2
0 0
1r ↔ 2r 0 −1 −5 −7 | 5
0 0 −19 −21 | 16
A handy solution to the last equation (which reads −19h − 21k = 16) is obtained
by assigning values 8 and −8 to h and k, respectively. This is already enough to
get the first point from the parametric equation of M 0 , which is apparent (though
incomplete) from the righthand side of (13.5). Thus Q = (−14, 3, 17). However, just
to check on the computations, I proceed to finish up the elimination, which leads
to t = 11 from the second row, and to s = −2 from the first. Substituting in the
lefthand side of (13.5), which is a trace of the parametric equations of M , I do get
indeed (−14, 3, 17). Another handy solution to¡the equation¢ corresponding¡to the last ¢
2 2
row of the final elimination table is¡ (h, k) =
¢ − 5
, − 5
. This gives R = 25 , 75 , − 35 ,
and the check by means of (s, t) = − 25 , − 15 is all right.
13.8.8 n. 9 (p. 482)
(a) A normal vector for M is
n = (1, 2, 3) × (3, 2, 1) = (−4, 8, −4)
Exercises 73

whereas the coefficient vector in the equation of M 0 is (1, −2, 1). Since the latter is
parallel to the former, and the coordinates of the point P ∈ M do not satisfy the
equation of M 0 , the two planes are parallel.
(b) A cartesian equation for M is

x − 2y + z + d = 0

From substitution of the coordinates of P , the coefficient d is seen to be equal to 3.

The coordinates of the points of the intersection line L ≡ M ∩ M 00 satisfy the system
½
x − 2y + z + 3 = 0
x + 2y + z = 0

By sum and subtraction, the line L can be represented by the simpler system
½
2x + 2z = −3
4y = 3

as the intersection of two different planes π and π 0 , with π parallel to the y-axis, and
π 0 parallel
¡ 3 3to ¢the xz-plane.
¡ Coordinate
¢ of points on L are now easy to produce, e.g.,
3 3
Q = − 2 , 4 , 0 and R = 0, 4 , − 2
13.8.9 n. 10 (p. 483)
The parametric equations for the line L are

x = 1 + 2r
y = 1−r
z = 1 + 3r

and the parametric equations for the plane M are

x = 1 + 2s
y = 1+s+t
z = −2 + 3s + t

The coordinates of a point of intersection between L and M must satisfy the system
of equations

2r − 2s = 0
r+s+t = 0
3r − 3s − t = −3

The first equation yields r = s, from which the third and second equation give t = 3,
3
¡r = s5= −72¢. Thus L ∩ M consists of a single point, namely, the point of coordinates
−2, 2 . − 2 .
74 Applications of vector algebra to analytic geometry

13.8.10 n. 11 (p. 483)

The parametric equations for L are

x = 1 + 2t
y = 1−t
z = 1 + 3t

and a direction vector for L is v = (2, −1, 3)

(a) L is parallel to the given¡ plane¢ (which I am going to denote π a ) if v is a linear
combination of (2, 1, 3) and 34 , 1, 1 . Thus I consider the system

3
2x + y = 2
4
x + y = −1
3x + y = 3

where subtraction of the second equation from the third gives 2x = 4, hence x = 2;
this yields y = −3 in the last two equations, contradicting the first.
Alternatively, computing the determinant of the matrix having v, a, b as
columns
¯ ¯ ¯ ¯
¯ 2 2 3 ¯ ¯ 4 0 −5 ¯ ¯ ¯
¯ 4 ¯ ¯ 4 ¯ ¯ 5 ¯
¯ −1 1 1 ¯ = ¯ −1 1 1 ¯ = ¯ 4 − 4 ¯ = − 1
¯ ¯ ¯ ¯ ¯ 6 −2 ¯ 2
¯ 3 3 1 ¯ ¯ 6 0 −2 ¯

shows that {v, a, b} is a linearly independent set.

With either reasoning, it is seen that L is not parallel to πa .
−→
(b) By computing two independent directions for the plane π b , e.g., P Q = (3, 5, 2) −
−→
(1, 1, −2) and P R = (2, 4, −1) − (1, 1, −2), I am reduced to the previous case. The
system to be studied now is

2x + y = 2
4x + 3y = −1
4x + y = 3

and the three equations are again inconsistent, because subtraction of the third equa-
tion from the second gives y = −2, whereas subtraction of the first from the third
yields x = 12 , these two values contradicting all equations. L is not parallel to π b .
(c) Since a normal vector to π c has for components the coefficients of the unknowns
in the equation of π c , it suffices to check orthogonality between v and (1, 2, 3):

h(2, −1, 3) , (1, 2, 3)i = 9 6= 0

L is not parallel to π c .
Exercises 75

13.8.11 n. 12 (p. 483)

Let R be any point of the given plane π, other than P or Q. A point S then belongs
to M if and only if there exists real numbers q and r such that

S = P + q (Q − P ) + r (R − P ) (13.6)

If S belongs to the line through P and Q, there exists a real number p such that

S = P + p (Q − P )

Then, by defining

q≡p r≡0

it is immediately seen that condition (13.6) is satisfied.

13.8.12 n. 13 (p. 483)
Since every point of L belongs to M , M contains the points having coordinates equal
to (1, 2, 3), (1, 2, 3) + t (1, 1, 1) for each t ∈ R, and (2, 3, 5). Choosing, e.g., t = 1, the
parametric equations for M are

x = 1+r+s
y = 2+r+s
z = 3 + r + 2s

It is possible to eliminate the two parameters at once, by subtracting the first equaiton
from the second, obtaining

x−y+1 =0

13.8.13 n. 14 (p. 483)

Let d be a direction vector for the line L, and let Q ≡ (xQ , yQ , zQ ) be a point of L.
A plane containing L and the point P ≡ (xP , yP , zP ) is the plane π through Q with
−→
direction vectors d and QP

x = xQ + hd1 + k (xP − xQ )
y = yQ + hd2 + k (yP − yQ )
z = zQ + hd3 + k (zP − zQ )

L belongs to π because the parametric equation of π reduces to that of L when k is

assigned value 0, and P belongs to π because the coordinates of P are obtained from
the parametric equation of π when h is assigned value 0 and k is assigned value 1.
Since any two distinct points on L and P determine a unique plane, π is the only
plane containing L and P .
76 Applications of vector algebra to analytic geometry

13.9 The cross product

13.10 The cross product expressed as a determinant

13.11 Exercises
13.11.1 n. 1 (p. 487)
(a) A × B = −2i + 3j − k.
(b) B × C = 4i − 5j + 3k.
(c) C × A = 4i − 4j + 2k.
(d) A × (C × A) = 8i + 10j + 4k.
(e) (A × B) × C = 8i + 3j − 7k.
(f) A × (B × C) = 10i + 11j + 5k.
(g) (A × C) × B = −2i − 8j − 12k.
(h) (A + B) × (A − C) = 2i − 2j.
(i) (A × B) × (A × C) = −2i + 4k.

13.11.2 n. 2 (p. 487)

A×B
(a) kA×Bk = − √426 i + √326 j + √1 k
26
A×B
or − kA×Bk = √4 i
26
− √3 j
26
− √1 k.
26

A×B 41 √ 18 j √7 k A×B √ 41 i √ 18 j √ 7 k.
(b) kA×Bk
= − √2054 i− 2054
+ 2054
or − kA×Bk = 2054
+ 2054
− 2054

A×B
(c) kA×Bk
= − √16 i − √2 j
6
− √1 k
6
A×B
or − kA×Bk = √1 i
6
+ √2 j
6
+ √1 k.
6

13.11.3 n. 3 (p. 487)

(a)
−→ −→
AB × AC = (2, −2, −3) × (3, 2, −2) = (10, −5, 10)
1°°−→ −→° 15
°
area ABC = °AB × AC ° =
2 2
(b)
−→ −→
AB × AC = (3, −6, 3) × (3, −1, 0) = (3, 9, 15)
√
1° °
°−→ −→° 3 35
area ABC = °AB × AC ° =
2 2
Exercises 77

(c)
−→ −→
AB × AC = (0, 1, 1) × (1, 0, 1) = (1, 1, −1)
1° ° √3
°−→ −→°
area ABC = °AB × AC ° =
2 2
13.11.4 n. 4 (p. 487)
−→ −→
CA × AB = (−i + 2j − 3k) × (2j + k) = 8i − j − 2k
13.11.5 n. 5 (p. 487)
Let a ≡ (l, m, n) and b ≡ (p, q, r) for the sake of notational simplicity. Then

ka × bk2 = (mr − nq)2 + (np − lr)2 + (lq − mp)2

= m2 r2 + n2 q 2 − 2mnrq + n2 p2 + l2 r2
−2lnpr + l2 q 2 + m2 p2 − 2lmpq
¡ ¢¡ ¢
kak2 kbk2 = l2 + m2 + n2 p2 + q2 + r2
= l2 p2 + l2 q 2 + l2 r2 + m2 p2 + m2 q 2
+m2 r2 + n2 p2 + n2 q 2 + n2 r2
¡ ¢
(ka × bk = kak kbk) ⇔ ka × bk2 = kak2 kbk2
⇔ (lp + mq + nr)2 = 0
⇔ ha, bi = 0
13.11.6 n. 6 (p. 487)
(a)
ha, b + ci = ha, b × ai = 0
because b × a is orthogonal to a.
(b)
hb, ci = hb, (b × a) − bi = hb, (b × a)i + hb, −bi = − kbk2
because b × a is orthogonal to b. Thus
iπ i
hb, ci < 0 c = − kbk < 0
cos bc c
bc ∈ ,π
kck 2
c = −π is impossible, because
. Moreover, bc

kck2 = kb × ak2 + kbk2 − 2 kb × ak kbk cos (b\

× a) b
2 2 2
= kb × ak + kbk > kbk
since (a, b) is linearly independent and kb × ak > 0. √
(c) By the formula above, kck2 = 22 + 12 , and kck = 5.
78 Applications of vector algebra to analytic geometry

13.11.7 n. 7 (p. 488)

(a) Since kak = kbk = 1 and ha, bi = 0, by Lagrange’s identity (theorem 13.12.f)

ka × bk2 = kak2 kbk2 − ha, bi2 = 1

so that a × b is a unit vector as well. The three vectors a, b, a × b are mutually

orthogonal either by assumption or by the properties of the vector product (theorem
13.12.d-e).
(b) By Lagrange’s identity again,

kck2 = ka × bk2 kak2 − ha × b, ai2 = 1

(c) And again,

k(a × b) × bk2 = ka × bk2 kbk2 − ha × b, bi2 = 1

k(a × b) × ak2 = kck2 = 1

Since the direction which is orthogonal to (a × b) and to b is spanned by a, (a × b)×b

is either equal to a or to −a. Similarly, (a × b) × a is either equal to b or to −b.
Both the righthand rule and the lefthand rule yield now

(a × b) × a = b (a × b) × b = −a

(d)
 
(a3 b1 − a1 b3 ) a3 − (a1 b2 − a2 b1 ) a2
(a × b) × a =  (a1 b2 − a2 b1 ) a1 − (a2 b3 − a3 b2 ) a3 
(a2 b3 − a3 b2 ) a2 − (a3 b1 − a1 b3 ) a1
 2 
(a2 + a23 ) b1 − a1 (a3 b3 + a2 b2 )
=  (a21 + a23 ) b2 − a2 (a1 b1 + a3 b3 ) 
(a21 + a22 ) b3 − a3 (a1 b1 + a2 b2 )

Since a and b are orthogonal,

 
(a22 + a23 ) b1 + a1 (a1 b1 )
(a × b) × a =  (a21 + a23 ) b2 + a2 (a2 b2 ) 
(a21 + a22 ) b3 + a3 (a3 b3 )

Since a is a unit vector,

 
b1
(a × b) × a =  b2 
b3

The proof that (a × b) × b = −a is identical.

Exercises 79

13.11.8 n. 8 (p. 488)

(a) From a × b = 0, either there exists some h ∈ R such that a = hb, or there exists
some k ∈ R such that b = ka. Then either ha, bi = h kbk2 or ha, bi = k kak2 , that
is, either h kbk = 0 or k kak = 0. In the first case, either h = 0 (which means a = 0),
or kbk = 0 (which is equivalent to b = 0). In the second case, either k = 0 (which
means b = 0), or kak = 0 (which is equivalent to a = 0). Geometrically, suppose
that a 6= 0. Then both the projection ha,bi
kak2
a of b along a and the projecting vector
(of length kb×ak
kak
) are null, which can only happen if b = 0.
(b) From the previous point, the hypotheses a 6= 0, a × (b − c) = 0, and ha, b − ci =
0 imply that b − c = 0.
13.11.9 n. 9 (p. 488)
c and observe that a and c are orthogonal. From a×b = kak kbk |sin ϑ|,
(a) Let ϑ ≡ ab,
in order to satisfy the condition
a×b= c
the vector b must be orthogonal to c, and its norm must depend on ϑ according to
the relation
kck
kbk = (13.7)
kak |sin ϑ|
kck
Thus kak ≤ kbk < +∞. In particular, b can be taken orthogonal to a as well, in
a×c c×a
which case it has to be equal to kak 2 or to
kak2
. Thus two solutions to the problem
are
µ ¶
7 8 11
± ,− ,−
9 9 9
(b) Let (p, q, r) be the coordinates of b; the conditions a × b = c and ha, bi = 1 are:
I −2q − r = 3
II 2p − 2r = 4
III p + 2q = −1
IV 2p − q + 2r = 1
Standard manipulations yield
2II + 2IV + III 9p = 9
(↑) ,→ II 2 − 2r = 4
(↑) ,→ I −2q + 1 = 3
check on IV 2+1−2 =1
check on III 1−2 =1
that is, the unique solution is
b = (1, −1, −1)
The solution to this exercise given by Apostol at page 645 is wrong.
80 Applications of vector algebra to analytic geometry

13.11.10 n. 10 (p. 488)

Replacing b with c in the first result of exercise7,

(a × c) × a = c

Therefore, by skew-simmetry of the vector product, it is seen that the c × a is a

solution to the equation

a×x=c (13.8)

However, c × a is orthogonal to a, and hence it does not meet the additional require-
ment

ha, xi = 1 (13.9)

We are bound to look for other solutions to (13.8). If x and z both solve (13.8),

a × (x − z) = a × x − a × z = c − c = 0

which implies that x − z is parallel to a. Thus the set of all solutions to equation
(13.8) is
© ª
x ∈ R3 : ∃α ∈ R, x = a × c + αa

From condition (13.9),

ha, a × c + αai = 1

that is,
1
α=−
kak2
Thus the unique solution to the problem is
a
b ≡ a × c−
kak2
13.11.11 n. 11 (p. 488)
(a) Let
−→
u ≡ AB = (−2, 1, 0)
−→
v ≡ BC = (3, −2, 1)
−→
w ≡ CA = (−1, 1, −1)

Each side of the triangle ABC can be one of the two diagonals of the parallelogram
to be determined.
Exercises 81

−→ −→ −→
If one of the diagonals is BC, the other is AD, where AD = AB + AC = u − w. In
this case

D = A + u − w = “B + C − A” = (0, 0, 2)

−→ −→ −→
If one of the diagonals is CA, the other is BE, where BE = BC + BA = v − u. In
this case

E = B + v − u = “A + C − B” = (4, −2, 2)

−→ −→ −→
If one of the diagonals is AB, the other is CF , where CF = CA + CB = −v + w.
In this case

F = A − v + w = “A + B − C” = (−2, 2, 0)

(b)

µ¯ ¯ ¯ ¯ ¯ ¯¶
¯ 1 0 ¯ ¯ ¯ ¯ −2 1 ¯
u×w = ¯ ¯ , − ¯ −2 0 ¯,¯ ¯
¯ 1 −1 ¯ ¯ −1 −1 ¯ ¯ −1 1 ¯ = (−1, −2, −1)
√
√ 6
ku × wk = 6= area (ABC) =
2
82 Applications of vector algebra to analytic geometry

13.11.12 n. 12 (p. 488)

b + c = 2 (a × b) − 2b
ha, b + ci = −2 ha, bi = −4
c = ha, bi = 1
cos ab
kak kbk 2
2 2 2
ka × bk = kak kbk sin2 ab c = 12
kck2 = 4 ka × bk2 − 12 ha × b, bi + 9 kbk2 = 192
√
kck = 8 3
hb, ci = 2 hb, a × bi − 3 kbk2 = 48
√
c = hb, ci 3
cos bc =
kbk kck 2
13.11.13 n. 13 (p. 488)
(a) It is true that, if the couple (a, b) is linearly independent, then the triple

(a + b, a − b, a × b)

is linearly independent, too. Indeed, in the first place a and b are nonnull, since every
n-tuple having the null vector among its components is linearly dependent. Secondly,
(a × b) is nonnull, too, because (a, b) is linearly independent. Third, (a × b) is
orthogonal to both a + b and a − b (as well as to any vector in lin {a, b}), because
it is orthogonal to a and b; in particular, (a × b) ∈ / lin {a.b}, since the only vector
which is orthogonal to itself is the null vector. Fourth, suppose that

x (a + b) + y (a − b) + z (a × b) = 0 (13.10)

Then z cannot be different from zero, since otherwise

x y
(a × b) = − (a + b) − (a − b)
z z
x+y y−x
= − a+ b
z z
would belong to lin {a, b}. Thus z = 0, and (13.10) becomes

x (a + b) + y (a − b) = 0

which is equivalent to

(x + y) a+ (x − y) b = 0

By linear independence of (a, b),

x+y =0 and x−y =0

Exercises 83

and hence x = y = 0.
(b) It is true that, if the couple (a, b) is linearly independent, then the triple

(a + b, a + a × b, b + a × b)

is linearly independent, too. Indeed, let (x, y, z) be such that

x (a + b) + y (a + a × b) + z (b + a × b) = 0

which is equivalent to

(x + y) a + (x + z) b + (y + z) a × b = 0

Since, arguing as in the fourth part of the previous point, y + z must be null, and
this in turn implies that both x + y and x + z are also null. The system

x+y = 0
x+z = 0
y+z = 0

has the trivial solution as the only solution.

(c) It is true that, if the couple (a, b) is linearly independent, then the triple

(a, b, (a + b) × (a − b))

is linearly independent, too. Indeed,

(a + b) × (a − b) = a × a − a × b + b × a − b × b
= −2a × b

and the triple

(a, b, a × b)

is linearly independent when the couple (a, b) is so.

13.11.14 n. 14 (p. 488)
−→ −→
(a) The cross
³ ´ product AB × AC equals the null vector if and only if the couple
−→ −→ −→
AB, AC is linearly dependent. In such a case, if AC is null, then A and C coin-
cide and it is clear that the line through them and B contains all the three points.
−→
Otherwise, if AC is nonnull, then in the nontrivial null combination
−→ −→
xAB + y AC = 0

x must be nonnull, yielding

−→ y −→
AB = − AC
x
84 Applications of vector algebra to analytic geometry

that is,
y −→
B = A − AC
x
which means that B belongs to the line through A and C.
(b) By the previous point, the set
n o
−→ −→
P : AP × BP = 0

is the set of all points P such that A, B, and P belong to the same line, that is, the
set of all points P belonging to the line through A and B.
13.11.15 n. 15 (p. 488)
(a) From the assumption (p × b) + p = a,
hb, p × bi + hb, pi = hb, ai
hb, pi = 0
that is, p is orthogonal to b. This gives
kp × bk = kpk kbk = kpk (13.11)
and hence
1 = kak2 = kp × bk2 + kpk2 = 2 kpk2
√
that is, kpk = 22 .
(b) Since p, b, and p×b are pairwise orthogonal, (p, b, p × b) is linearly independent,
and {p, b, p × b} is a basis of R3 .
(c) Since p is orthogonal both to p × b (definition of vector product) and to b (point
a), there exists some h ∈ R such that
(p × b) × b = hp
Thus, taking into account (13.11),
¯ π¯
¯ ¯
|h| kpk = kp × bk kbk ¯sin ¯ = kpk
2
and h ∈ {1, −1}. Since the triples (p, b, p × b) and (p × b, p, b) define the same
orientation, and (p × b, b, p) defines the opposite one, h = −1.
(d) Still from the assumption (p × b) + p = a,
hp, p × bi + hp, pi = hp, ai
1
hp, ai = kpk2 =
2
p × (p × b) + p × p = p × a
1
kp × ak = kpk2 kbk =
2
The scalar triple product 85

Thus
1 1
p= a+ q (13.12)
2 2
where q = 2p − a is a vector in the plane π generated by p and a, which is orthogonal
to a. Since a normal vector to π is b, there exists some k ∈ R such that

q = k (b × a)

Now

kqk2 = 4 kpk2 + kak2 − 4 kpk kak cos p

ca
√ √
2 2
= 2+1−4 =1
2 2
kqk = 1 |k| = 1

If on the plane orthogonal to b the mapping u 7→ b × u rotates counterclockwise, the

vectors a = p + b × p, b × p, and b × a form angles of π4 , π2 , and 3π
4
, respectively,
with p. On the other hand, the decomposition of p obtained in (13.12) requires that
the angle formed with p by q is − π4 , so that q is discordant with b × a. It follows
that k = −1. I have finally obtained

1 1
p = a − (b × a)
2 2

13.12 The scalar triple product

13.13 Cramer’s rule for solving systems of three linear equations

13.14 Exercises

13.15 Normal vectors to planes

13.16 Linear cartesian equations for planes

86 Applications of vector algebra to analytic geometry

13.17 Exercises
13.17.1 n. 1 (p. 496)
(a) Out of the inifnitely many vectors satisfying the requirement, a distinguished one
is

n ≡ (2i + 3j − 4k) × (j + k) = 7i − 2j + 2k

(b)

hn, (x, y, z)i = 0 or 7x − 2y + 2z = 0

(c)

hn, (x, y, z)i = hn, (1, 2, 3)i or 7x − 2y + 2z = 9

13.17.2 n. 2 (p. 496)

(a)
µ ¶
n 1 2 2
= , ,−
knk 3 3 3

(b) The three intersection points are:

µ ¶ µ ¶
7 7
X-axis : (−7, 0, 0) Y -axis : 0, − , 0 Z-axis : 0, 0, .
2 2

(c) The distance from the origin is 73 .

(d) Intersecting with π the line through the origin which is directed by the normal
to π


 x=h

y = 2h
 z = −2h


x + 2y − 2z + 7 = 0

yields

h + 4h + 4h + 7 = 0
7
h = −
µ9 ¶
7 14 14
(x, y, z) = − ,− ,
9 9 9
Exercises 87

13.17.3 n. 3 (p. 496)

A cartesian equation for the plane which passes through the point P ≡ (1, 2, −3) and
is parallel to the plane of equation
3x − y + 2z = 4
is
3 (x − 1) − (y − 2) + 2 (z + 3) = 0
or
3x − y + 2z + 5 = 0
The distance between the two planes is
√
|4 − (−5)| 9 14
√ =
9+1+4 14
13.17.4 n. 4 (p. 496)
π1 : x + 2y − 2z = 5
π2 : 3x − 6y + 3z = 2
π3 : 2x + y + 2z = −1
π4 : x − 2y + z = 7
(a) π 2 and π 4 are parallel because n2 = 3n4 ; π 1 and π3 are orthogonal because
hn1 , n3 i = 0.
(b) The straight line through the origin having direction vector n4 has equations
x = t
y = −2t
z = t
and intersects π 2 and π 4 at points C and D, respectively, which can be determined
by the equations
3tC − 6 (−2tC ) + 3tC = 2
tD − 2 (−2tD ) + tD = 7
−→ ¡ ¢
Thus tC = 19 , tD = 76 , CD = 76 − 19 n4 , and
° ° 19 √
°−→° 19
°CD° = kn4 k = 6
18 18
Alternatively, rewriting the equation of π 4 with normal vector n2
3x − 6y + 3z = 21

|d2 − d4 | 19
dist (π, π 0 ) = =√
kn2 k 54