Lecture 2

Multiple Products. Geometry using Vectors

2.1 Scalar triple product

Using mixtures of the pairwise scalar product and vector product, it is possible to derive
“triple products” between three vectors, and indeed n-products between n vectors.
This is the scalar product of a vector product and a third vector, i.e. a · (b × c). Using
the pseudo-determinant expression for the vector product, we see that the scalar triple
product can be represented as the true determinant

a1 a2 a3
a · (b × c) = b1 b2 b3 (2.1)
c1 c2 c3

You will recall that if you swap a pair of rows of a determinant, its sign changes; hence
if you swap two pairs, its sign stays the same.
a1 a2 a3 c1 c2 c3 c1 c2 c3
b1 b2 b3 1st SWAP b1 b2 b3 2nd SWAP a1 a2 a3
(2.2)
c1 c2 c3 a1 a2 a3 b1 b2 b3
+ − +

This says that

(i) a · (b × c) = b · (c × a) = c · (a × b) (Called cyclic permutation.)

(ii) a · (b × c) = −b · (a × c) and so on. (Called anti-cyclic permutation.)
(iii) The fact that a · (b × c) = (a × b) · c allows the scalar triple product to be written
as [a, b, c]. This notation is not very helpful, and we will try to avoid it below.

1
2/2 LECTURE 2. MULTIPLE PRODUCTS. GEOMETRY USING VECTORS

2.1.1 Geometrical interpretation of scalar triple product

The scalar triple product gives the volume of the parallelopiped whose sides are repre-
sented by the vectors a, b, and c.
We saw earlier that the vector product (a × b) has magnitude equal to the area of the
base, and direction perpendicular to the base. The component of c in this direction is
equal to the height of the parallelopiped shown in Figure 2.1(a).

n
c cos β b
c
c
β

b
a
a
Figure 2.1: (a) Scalar triple product equals volume of parallelopiped. (b) Coplanarity yields zero scalar
triple product.

2.1.2 Linearly dependent vectors

If the scalar triple product of three vectors is zero

a · (b × c) = 0 (2.3)

then the vectors are linearly dependent. That is, one can be expressed as a linear
combination of the others. For example,

a = λb + µc (2.4)

where λ and µ are scalar coefficients.

You can see this immediately in two ways:

• The determinant would have one row that was a linear combination of the others.
You’ll remember that by doing row operations, you could reach a row of zeros,
and so the determinant is zero.
• The parallelopiped would have zero volume if squashed flat. In this case all three
vectors lie in a plane, and so any one is a linear combination of the other two.
(Figure 2.1b.)
2.2. VECTOR TRIPLE PRODUCT 2/3

2.2 Vector triple product

This is defined as the vector product of a vector with a vector product, a × (b × c).
Now, the vector triple product a × (b × c) must be perpendicular to (b × c), which
in turn is perpendicular to both b and c. Thus a × (b × c) can have no component
perpendicular to b and c, and hence must be coplanar with them. It follows that the
vector triple product must be expressible as a linear combination of b and c:

a × (b × c) = λb + µc . (2.5)

The values of the coefficients can be obtained by multiplying out in component form.
Only the first component need be evaluated, the others then being obtained by sym-
metry. That is

(a × (b × c))1 = a2 (b × c)3 − a3 (b × c)2 (2.6)

= a2 (b1 c2 − b2 c1 ) + a3 (b1 c3 − b3 c1 )
= (a2 c2 + a3 c3 )b1 − (a2 b2 + a3 b3 )c1
= (a1 c1 + a2 c2 + a3 c3 )b1 − (a1 b1 + a2 b2 + a3 b3 )c1
= (a · c)b1 − (a · b)c1

The equivalents must be true for the 2nd and 3rd components, so we arrive at the
identity

a × (b × c) = (a · c)b − (a · b)c . (2.7)

a
bx c
In arbitrary direction

a x (bx c)

Figure 2.2: Vector triple product.

2/4 LECTURE 2. MULTIPLE PRODUCTS. GEOMETRY USING VECTORS

2.2.1 Projection using vector triple product

An example of the application of this formula is as follows. Suppose v is a vector and
we want its projection into the xy -plane. The component of v in the z direction is
v · k̂, so the projection we seek is v − (v · k̂)k̂. Writing k̂ ← a, v ← b, k̂ ← c,

a × (b × c) = (a · c)b − (a · b)c (2.8)

↓ ↓ ↓
k̂ × (v × k̂) = (k̂ · k̂)v − (k̂ · v)k̂ (2.9)
= v − (v · k̂)k̂

So, now we recognize that we have two equivalent expressions for the vector component
of a vectot v perpendicular to the unit direction k̂:
v − (v · k̂)k̂ = k̂ × (v × k̂) . (2.10)

(Hot stuff! But the expression v − (v · k̂)k̂ is much easier to understand, and cheaper
to compute!)

2.3 Other repeated products

Many combinations of vector and scalar products are possible, but we consider only
one more, namely the vector quadruple product (a × b) × (c × d). By regarding a × b as
a single vector, we see that this vector must be representable as a linear combination
of c and d. On the other hand, regarding c × d as a single vector, we see that it must
also be a linear combination of a and b. This provides a means of expressing one of
the vectors, say d, as linear combination of the other three, as follows:

(a × b) × (c × d) = [(a × b) · d]c − [(a × b) · c]d (2.11)

= [(c × d) · a]b − [(c × d) · b]a

Hence
[(a × b) · c] d = [(b × c) · d] a + [(c × a) · d] b + [(a × b) · d] c (2.12)
or
[(b × c) · d] a + [(c × a) · d] b + [(a × b) · d] c
d= = αa + βb + γc . (2.13)
[(a × b) · c]
This is much more complicated way of obtaining exactly the same result that we derived
earlier!
2.4. GEOMETRY USING VECTORS: LINES, PLANES 2/5

a
d

Figure 2.3: The projection of a (3-)vector onto a set of (3) basis vectors is unique. Ie in d = αa+βb+γc,
the set {α, β, γ} is unique.

2.4 Geometry using vectors: lines, planes

2.4.1 The equation of a line
The equation of the line passing through the point whose position vector is a and lying
in the direction of vector b is

r = a + λb (2.14)

where λ is a scalar parameter. If you make b a unit vector, r = a + λb̂ then λ will
represent metric length.
For a line defined by two points a1 and a2

r = a1 + λ(a2 − a1 ) (2.15)

or for the unit version

(a2 − a1 )
r = a1 + λ (2.16)
|a2 − a1 |

^
λb
Point r traces
a out line.
r

Figure 2.4: Equation of a line. With b̂ a unit vector, λ is in the length units established by the definition
of a.
2/6 LECTURE 2. MULTIPLE PRODUCTS. GEOMETRY USING VECTORS

2.4.2 The shortest distance from a point to a line

When working out the shortest distance from a point to a line it is usually sufficient to
assume that this distance is found along the perpendicular to the line. First here, let’s
prove that too.
Referring to Figure 2.5(a) the vector p from c to any point on the line is p = a+λb̂−c =
(a − c) + λb̂ which has length squared p 2 = (a − c)2 + λ2 + 2λ(a − c) · b̂ . Rather
than minimizing length, it is easier to minimize length-squared. The minumum is found
when d p 2 /dλ = 0, ie when
λ = −(a − c) · b̂ . (2.17)
So the minimum length vector is
p = (a − c) − ((a − c) · b̂)b̂. (2.18)
You might spot that is the component of (a − c) perpendicular to b̂ (as expected!).
Furthermore, using the result of Section 2.2.1,
p = b̂ × [(a − c) × b̂] . (2.19)
Because b̂ is a unit vector, and is orthogonal to [(a − c) × b̂], the modulus of the vector
can be written rather more simply as just
pmin = |(a − c) × b̂| . (2.20)

c
λb µd
λb

a P
r−c a
r Q

(a) (b)

Figure 2.5: (a) Shortest distance point to line. (b) Shortest distance, line to line.

2.4.3 The shortest distance between two straight lines

If the shortest distance between a point and a line is along the perpendicular, then the
shortest distance between the two straight lines r = a + λb̂ and r = c + µd̂ must be
found as the length of the vector which is mutually perpendicular to the lines.
2.4. GEOMETRY USING VECTORS: LINES, PLANES 2/7

The unit vector along the mutual perpendicular is

p̂ = (b̂ × d̂)/|b̂ × d̂| . (2.21)

(Yes, don’t forget that b̂ × d̂ is NOT a unit vector. b̂ and d̂ are not orthogonal, so
there is a sin θ lurking!)
The minimum length is therefore the component of a − c in this direction

pmin = (a − c) · (b̂ × d̂)/|b̂ × d̂| . (2.22)

♣ Example

Q Two long straight pipes are specified using Cartesian co-ordinates as follows:
Pipe A has diameter 0.8 and its axis passes through points (2, 5, 3) and (7, 10, 8).
Pipe B has diameter 1.0 and its axis passes through the points (0, 6, 3) and
(−12, 0, 9).
Determine whether the pipes need to be realigned to avoid intersection.

A Each pipe axis is defined using two points. The vector equation of the axis of pipe
A is
√
r = [2, 5, 3] + λ0 [5, 5, 5] = [2, 5, 3] + λ[1, 1, 1]/ 3 (2.23)

The equation of the axis of pipe B is

√
r = [0, 6, 3] + µ0 [12, 6, 6] = [0, 6, 3] + µ[−2, −1, 1]/ 6 (2.24)

The perpendicular to the two axes has direction

ı̂ ̂ k̂
[1, 1, 1] × [−2, −1, 1] = 1 1 1 = [2, −3, 1] = p (2.25)
−2 −1 1

The length of the mutual perpendicular is

[2, −3, 1] [2, −3, 1]

(a − c) · √ = [2, −1, 0] · √ = 1.87 . (2.26)
14 14
But the sum of the radii of the two pipes is 0.4 + 0.5 = 0.9. Hence the pipes do
not intersect.
2/8 LECTURE 2. MULTIPLE PRODUCTS. GEOMETRY USING VECTORS

2.4.4 The equation of a plane

There are a number of ways of specifying the equation of a plane.

1. If b and c are two non-parallel vectors (ie b × c 6= 0), then the equation of the
plane passing through the point a and parallel to the vectors b and c may be
written in the form
r = a + λb + µc (2.27)
where λ, µ are scalar parameters. Note that b and c are free vectors, so don’t
have to lie in the plane (Figure 2.6(a).)
2. Figure 2.6(b) shows the plane defined by three non-collinear points a, b and c in
the plane (note that the vectors b and c are position vectors, not free vectors as
in the previous case). The equation might be written as
r = a + λ(b − a) + µ(c − a) (2.28)

3. Figure 2.6(c) illustrates another description is in terms of the unit normal to the
plane n̂ and a point a in the plane
r · n̂ = a · n̂ . (2.29)

^
n
r r c

b r
a
c a a

NB that these are

O b parallel to the plane, not
necessarily in the plane
O O
(a) (b) (c)

Figure 2.6: (a) Plane defined using point and two lines. (b) Plane defined using three points. (c) Plane
defined using point and normal. Vector r is the position vector of a general point in the plane.

2.4.5 The shortest distance from a point to a plane

The shortest distance from a point d to the plane is along the perpendicular. Depending
on how the plane is defined, this can be written as
|(d − a) · (b × c)|
D = |(d − a) · n̂| or D= . (2.30)
|b × c|
2.5. ROTATION OF VECTORS, IN BRIEF 2/9

2.5 Rotation of vectors, in brief

In 3D, a rotation matrix R is a 3 × 3 matrix. The orthonormality condition R> = R−1
imposes 6 constraints on the matrix so that R has only 3 degrees of freedom. Its
determinant is unity, its trace is 2 cos θ + 1 where θ is the rotation angle and its
eigenvalues are (1, e±iθ ), where the eigenvalue 1 corresponds to the eigenvector which
is the rotation axis Ra = a.
The three degrees of freedom in one rotation matrix can be considered to lie in the
direction of the rotation axis (2 d.o.f) and the rotation angle about it (1 d.o.f.). How-
ever, often it is easier to compose the overall rotation as a sequence of 3 “elementary”
rotations about either fixed or successive principal axes (eg., x, y , z). There are 24
possibilites for choosing these combinations, subsets of which are called variously as
Euler, Cardan or Tait-Bryan angles. One well-known example from aircraft dynamics is
Yaw-Pitch-Roll where successive rotations are measured successive principal directions
of an aircraft’s body.
The other issue to consider is whether the rotation is leaving the “world” in place and
rotating the coordinates, or rotating the world within the same fixed coordinates.

y’ y y

(x y )
x’ 1 1

θ x θ x
(x y )
0 0
Figure 2.7: Left: The world is stationary, but the coordinate system is rotated. Right: The world is
rotated within fixed coordinates.

For a rotation of θ in the rh sense about the z−axis, the first and second are
   
cos θ sin θ 0 cos θ − sin θ 0
x0 =  − sin θ cos θ 0  x and x1 =  sin θ cos θ 0  x0 (2.31)
0 0 1 0 0 1

Which sin term has the minus sign is easily determine using the trial point — [1, 0, 0] in
our case. For the rotating axes case, the fixed point must have coordinates [x 0 , y 0 , z 0 ] =
[cos θ, − sin θ, 0] placing the minus sign as shown.
2/10 LECTURE 2. MULTIPLE PRODUCTS. GEOMETRY USING VECTORS

2.6 Angular velocity/acceleration and moments, in brief

A rotation can represented by a vector whose direction is along the axis of rotation
in the sense of a r-h screw, and whose magnitude is proportional to the size of the
rotation (Fig. 2.8). The same idea can be extended to the derivatives, that is, angular
velocity ω and angular acceleration ω̇.
Angular accelerations arise because of a moment (or torque) on a body. In mechanics,
the moment of a force F about a point Q is defined to have magnitude M = F d, where
d is the perpendicular distance between Q and the line of action L of F.

M ω
ω
in right−hand screw
sense
dθ δ t
dt
α r r
v
F
(a) (b)

Figure 2.8: (a) The angular velocity vector ω is along the axis of rotation and has magnitude equal to
the rate of rotation. (b)

The vector equation for moment is

M=r×F (2.32)

where r is the vector from Q to any point on the line of action L of force F. The
resulting angular acceleration vector is in the same direction as the moment vector.
The instantaneous velocity of any point P on a rigid body undergoing pure rotation can
be defined by a vector product as follows. The angular velocity vector ω has magnitude
equal to the angular speed of rotation of the body and with direction the same as that
of the r-h screw. If r is the vector OP , where the origin O can be taken to be any point
on the axis of rotation, then the velocity v of P due to the rotation is given, in both
magnitude and direction, by the vector product v = ω × r.
2.7. SOLUTIONS TO GENERAL VECTOR EQUATIONS 2/11

2.7 Solutions to general vector equations

Not all equations for a variable vector x have a unique point solution, and often the
locus is not a simple recognizable curve or region.
Examples of such equations are

x = x × a+b (2.33)
x·a = γ . (2.34)

A standard recipe for proceeding in such cases is as follows:

1. Decide upon two non-parallel vectors appearing in the vector relationship. These
might be a, b and their vector product (a × b).
If there is only one vector a, say, then b can be freely chosen, with the proviso
that it is non-parallel to a.
2. Generate a×b.
3. Express the unknown vector x as a linear combination of these vectors
x = λa + µb + νa × b
where λ, µ, ν are scalars to be found.
4. Substitute the above expression for x into the vector relationship to determine the
constraints on λ, µ and ν for the relationship to be satisfied.

♣ Example #1
Q: Find the vectors x that satisfy the equation x = x × a + b.
A: Steps (1 & 2): Use a and b and their vector product a × b as basis vectors.
Step (3): x = λa + µb + νa × b.
Step (4): Bung this expression for x into the equation!
 
λa + µb + νa × b = λa + µb + νa × b × a + b (2.35)
= 0 + µ(b × a) + ν(a × b) × a + b (2.36)

But using the expression for the vector triple product

(a × b) × a = a × (b × a) (2.37)
2
= (a · a)b − (a · b)a = a b − (a · b)a (2.38)
2/12 LECTURE 2. MULTIPLE PRODUCTS. GEOMETRY USING VECTORS

Hence

λa + µb + νa × b = −ν(a · b)a + (νa2 + 1)b − µ(a × b) (2.39)

We have learned that any vector has a unique expression in terms of a basis set, so
that the coefficients of a, b and a × b on either side of the equation much be equal.

⇒λ = −ν(a · b) µ = νa2 + 1 ν = −µ (2.40)

so that
1 1 a·b
µ= ν=− λ= . (2.41)
1 + a2 1 + a2 1 + a2
So finally the solution in this case is a single point:
1
x= ((a · b)a + b − (a × b)) (2.42)
1 + a2

♣ Example #2
Q: The second example x · a = K is in 2A1A. Notice that the geometry is rather
obvious in this case! x must lie on the plane x · â = K/a. A plane with unit normal
â and perpendicular distance K/a from the origin. However, that does not answer the
question.
A: Using the recipe you choose b arbitrarily, and will find
K − λa2
 
x = λa + b + ν(a×b) (2.43)
a·b
where λ and ν are free. This is correct, but why does it look so complicated? It is
because b has been chosen arbitrarily and is one of the basis vectors.
As we can see upfront that this must be a plane, a cunning selection would be to choose
the second vector to be perpendicular to both a and a × b. You still have to choose b
arbitrarily, but using our knowledge, we can write without further thought
K
x= a + µ(a × (a × b)) + ν(a × b) . µ, ν are free (2.44)
a2
Can you see why?

Revised December 1, 2015