FoMP: Vectors, Tensors and Fields

Example problems with solutions

1 Given that the vector a has components (1, 2, 3) in a orthonormal, right-handed basis:–

(i) construct a unit vector â parallel to a ;

(ii) construct a vector b orthogonal to a, with no z-component (i.e. b3 = 0) ;
(iii) construct a unit vector b̂ parallel to b ;
(iv) construct a vector c orthogonal to both a and b by requiring that c·a = c·b = 0;
(v) construct a unit vector ĉ orthogonal to both â and b̂, such that â, b̂ and ĉ form
a right-handed triad ;
(vi) Verify that ĉ = â × b̂ .

(i) Construct â = a /a, where a is the magnitude

√ of a. Now a2 = a · a = 14, so
that in principle, we could take a = ±1/ 14 but we choose the positive root
so that â is parallel, rather than anti-parallel to a. The desired unit vector is
thus
1
â = √ (1, 2, 3)
14
(ii) The vector b is of the form (b1 , b2 , 0) and orthogonal to a so that b · a = 0,
giving b1 + 2b2 = 0. If we choose b2 = 1, say, then the desired vector is

b = (−2, 1, 0)

Clearly an arbitrary scalar multiple of this would also satisfy the orthogonality
requirement.
(iii) With the above choice for b, we have b2 = b · b = 5. Again we take the positive
square root to yield
1
b̂ = √ (−2, 1, 0)
5
(iv) Writing c = (c1 , c2 , c3 ) and demanding that c·a = 0 tells us that c1 +2c2 +3c3 =
0. Similarly, c · b = 0 implies that −2c1 + c2 = 0. Thus 5c1 = −3c3 and
5c2 = −6c3 . If we choose c1 = 3 we obtain

c = (3, 6, −5)
 √
(v) With this choice of c, we see that c2 = c · c = 70 so that c = ±1/ 70. The
choice of sign is dictated by the requirement that â, b̂, ĉ form a right-handed
triad, so we require that (â, b̂, ĉ) = 1:
1 2 3
√ √ √
14 14 14 1 2 3
−2 1 1
(â, b̂, ĉ) = ± √ √ 0 =± −2 1 0 = ±(−1)
5 5 70
3 6 −5 3 6 −5
√ √ √
70 70 70
1
Thus we must take the - sign, giving ĉ = √ (−3, −6, 5).
70
(vi) From the definition of the vector product

e1 e2 e3
1 2 3
√ √ √ 3 6 5
â × b̂ = 14 14 14 = −e1 √ − e2 √ + e3 √ = ĉ
−2 1 70 70 70
√ √ 0
5 5

2 (i) The velocity of a point particle, rotating with angular velocity vector ω through
the origin, is v = ω × r. The angular momentum vector of such a particle is
L = r × mv where m is the mass.
Show that Lω , the component of the angular momentum vector along the axis
of rotation, is given by
h i
Lω = mω r2 − |r · ω̂|2

What is the geometrical meaning of the term in the square bracket?

(ii) The kinetic energy of such a particle is given by K = m(v · v)/2.
Show that
mw2 h 2 i
K= r − |r · ω̂|2
2
The definition of the required component is Lω = L · ω̂
In full, the angular momentum is L = mr × (ω × r) = m ω r × (ω̂ × r).
Expanding the triple vector product using BAC-CAB, this gives L = mω [ω̂r2 − r(r · ω̂)].
Taking L · ω̂ now gives the required result.
The term in square brackets is r2 (1 − |r̂ · ω̂|2 ) = (r sin θ)2 , where θ is the angle between
r̂ and ω̂. R = r sin θ is the radius of the circle around which the particle moves. From
this 2D point of view, we would clearly expect Lω = mωR2 .
The definition of K involves v · v = (ω × r) · (ω × r). Using the cyclic property of the
triple scalar product, this becomes (r × [ω × r]) · ω.
 Expanding r × [ω × r] using BAC-CAB, we get ωr2 − r(ω · r).
Taking the dot product with ω, we then get v · v = ω 2 r2 − (r · ω)2 = ω 2 (r2 − (r · ω̂)2 .

3 The two sets of basis vectors {ei } and {ei ′ } are both right-handed orthonormal triads
such that e1 ′ is in the direction of (e1 − e3 ) and e2 ′ is in the direction of (e1 + e2 + e3 ).
(i) Construct the correctly normalised basis vectors e1 ′ , e2 ′ , e3 ′ .
(ii) Write down the transformation matrix λ from the basis {ei } to the basis {ei ′ }.
(iii) Express the two vectors A = 2e1 + e3 and B = e1 − e2 − e3 in the basis {ei ′ }.
(iv) Verify that the scalar product of these two vectors is an invariant of the trans-
formation.
(v) Show that the components of the vector product, evaluated in the two bases,
are related as follows:

(A × B)i ′ = λij (A × B)j

(i) e1 ′ is a unit vector in the direction of e1 + e3 and so

1
e1 ′ = √ (e1 − e3 )
2
e2 ′ is a unit vector in the direction of e1 + e2 + e3 and so
1
e2 ′ = √ (e1 + e2 + e3 )
3
e3 ′ = e1 ′ × e2 ′ since the basis {ei ′ } is right-handed
1 1
e3 ′ = √ (e1 − e3 ) × (e1 + e2 + e3 ) = √ (e3 − e2 − e2 + e1 )
6 6
1
= √ (e1 − 2e2 + e3 )
6
(ii) The elements of λ are given by λij = ei ′ · ej which can be read off directly
from (i)
1 1
 
 √ 0 −√ 
 2 2 
 1 1 1
 
λ =  √ √ √ 


 3 3 3 

1 2 1 
√ −√ √
 
6 6 6
√
(iii) A′i = λij Aj so that, given A1 = 2, A3 = 1 and A2 = 0 we find that A′1 = 1/ 2,
√ √
A′2 = 3/ 3 and A′3 = 3/ 6, giving
1 3 3
A = √ e1 ′ + √ e2 ′ + √ e3 ′
2 3 6
 √ √ √
Similarly, Bi′ = λij Bj and B1 = 2/ 2, B2 = −1/ 3 and B3 = 2/ 6, yielding

2 1 2
B = √ e1 ′ − √ e2 ′ + √ e3 ′
2 3 6

(iv) The scalar product is A · B = Ai Bi = A′i Bi′ . To check this we calculate

Ai Bi = 2 1 + 0 − 1 1 = 1 and
s s
1 √ √ 1 3 2
Ai Bi = √ 2 − 3 √ +
′ ′
=1
2 3 2 3

(v) Assuming the formula (A × B)i = ǫijk Aj Bk

(A × B)1 = A2 B3 − A3 B2 = 1
(A × B)2 = A3 B1 − A1 B3 = 3
(A × B)3 = A1 B2 − A2 B1 = −2
√
(A × B)′1 = A′2 B3′ − A′3 B2′ = 3/√2
(A × B)′2 = A′3 B1′ − A′1 B3′ = 2/√3
(A × B)3 = A1 B2 − A2 B1 = −7/ 6
′ ′ ′ ′ ′

so we can check the relations between components in the two bases:

√
λ1j (A × B)j = 3/ 2 = (A × B)′1
√
λ2j (A × B)j = 2/ 3 = (A × B)′2
√
λ3j (A × B)j = −7/ 6 = (A × B)′3

4 Construct the elements λij of the transformation matrix and calculate its determinant
for (i) a reflection of basis in the e1 –e2 plane; (ii) a rotation of basis through an angle θ
about the e3 axis where a positive rotation is taken to be in r.h. screw direction; (iii) a
rotation of basis through an angle θ about the e1 axis where a positive rotation is taken to
be in the r.h. screw direction. Show that the matrices in (ii) and (iii) do not commute.
Finally, consider the case of small rotations and work to first order in angles, so that
cos θ → 1 and sin θ → θ, and show that the rotations now commute. Thus argue that
angular velocity can be represented by a vector, even though rotations cannot.
(i) Reflection in the e1 –e2 plane:
e_’2
e_ 3
  
e_’ e1 ′ = e1  1 0 0
1 
λ e2 =
′
e2 ⇒λ= 0 1

0 

e_
e3 = −e3 0 0 −1
′

2


e_1 det λ = −1
e_’
3

(ii) Rotation about e3 :

 e_ 2 e1 ′ · e1 = cos θ
e_ 2’
e1 ′ · e2 = sin θ
e1 ′ · e3 = 0
e_ ’
1
θ e2 ′ · e1 = − sin θ
θ e2 ′ · e2 = cos θ
e_1
O
e2 ′ · e3 = 0
where we have chosen a r.h. basis and assumed the r.h. screw rule. Thus
 
cos θ sin θ 0
 − sin θ cos θ
λ = λ1 =  0  and det λ = cos2 θ + sin2 θ = 1
0 0 1

(iii) Rotation about e1 : proceed similarly to get

 
1 0 0
λ = λ2 =  0 cos θ sin θ  and det λ = cos2 θ + sin2 θ = 1


0 − sin θ cos θ

We now need to compute the product of these matrices. It is quicker to use the shorthand
C = cos θ, S = sin θ:
S2
   
C CS C S 0
λ1 λ2 =  −S

C2 SC  λ2 λ1 =  −SC

C2 S .
0 −S C S 2 −SC C

These two matrices differ. But if we take the small-angle limit and replace C by 1 and
S by θ and S 2 by 0, then things do commute:
 
1 θ 0
 −θ
λ1 λ2 = λ2 λ1 =  1 θ .
0 −θ 1
If we regard the rotation angle as θ = ω dt, this show that the rotations from two angular
velocities will combine linearly.

5 A certain type of anisotropically conducting crystal allows current to flow in one direction
only, along which there is a linear response of current to applied voltage.. Suppose that
this crystal is orientated so that current can flow only along the x axis:
(i) Show that this statement can be written as a tensor relation between the applied
electric field vector, E and the current density vector, J, and write down the components
of the conductivity tensor in this basis.
(ii) The system is now viewed using a system of coordinates rotated anticlockwise by an
angle θ about the z axis. Give the transformation matrix λ that relates vectors in the old
and new coordinate systems.
(iii) Derive the general transformation law that must be satisfied by the components of
a conductivity tensor in order to preserve the correct relation between electric field and
current density in the new coordinate system, and apply it to deduce the transformed
conductivity tensor in the present example.
(iv) Consider in particular rotations of π/2, π and 2π and comment on whether your
mathematical results make physical sense.
(i) In the x direction, we have the usual Ohm’s law: Jx = g Ex , where g is a conductivity
measure. Jx is independent of Ey and Ez and Jy = Jz = 0. This can be written in tensor
form Ji = Gij Ej , where all components of Gij are zero except G11 = g.
(ii) This is as usual from the notes:
 
cos θ sin θ 0
λ =  − sin θ cos θ 0 

.
0 0 1

(iii) Changing basis cannot alter the form of a physical law. Therefore, if our relation
reads Ji = Gij Ej , we must have

Ji′ = G′ij Ej′ where Ji′ = λij Jj and Ej′ = λjk Ek .

Rewriting the original relation, we have λ−1 −1 ′

ik Jk = Gij λjℓ Eℓ . Now multiply on the left by
′

λmi and use the fact that λmi λ−1

ik = δmk . this gives
 
′
Jm = λmi Gij λ−1
jℓ Eℓ′

and using the orthogonal property of the transformation matrix, we therefore identify
the transformed components of the tensor:

G′mℓ = λmi Gij (λT )jℓ ,

which in matrix notation is

G′ = λ G λT .

Carrying out the matrix multiplication,

     
cos θ sin θ 0 g 0 0 cos θ − sin θ 0
G′ =  − sin θ cos θ 0  0 0 0  sin θ cos θ 0 
  
    
0 0 1 0 0 0 0 0 1
cos2 θ
 
− sin θ cos θ 0
= g  − sin θ cos θ

sin2 θ 0 
.
0 0 0

(iv) Thus θ = π gives the same result as θ = 0 or 2π. This makes sense: current runs in
the x direction only, but can flow equally well in either direction. The case θ = π/2 has
the effect of swapping the x and y axes and indeed Gij in that case says just Jy = gEy .
 √
6 Three equal masses are placed at the origin, at r = 2ae1 and at r = ae1 + 3ae2 .
(i) Calculate the centre of mass position vector, R, and the position vectors, sα ,
of the particles relative to the centre of mass.
(ii) Construct the inertia tensor relative to the centre of mass.
(iii) Use the parallel axes theorem to construct the inertia tensor relative to the
origin

(i) The position vector of the centre of mass is

1 X α α
mα = 3m
X
R= m r where M =
M α α

Thus
√
!
1 h i a
R= m(2ae1 ) + m(ae1 + 3ae2 ) = ae1 + √ e2
3m 3
Position vectors relative to the centre of mass are sα = rα − R thus
! ! !
(1) a (2) a (3) 2a
s = −ae1 − √ e2 , s = ae1 − √ e2 , s = √ e2
3 3 3

(ii) The inertia tensor relative to the centre of mass is

mα (sα· sα ) δij − sα α
X n o
Iij (G) = i sj
α
We can compute as follows:

mα (sα α 2 2
X
1 s1 ) = ma (1 + 1 + 0) = 2ma
α
mα (sα α 2 2
X
2 s2 ) = ma (1/3 + 1/3 + 4/3) = 2ma
α
mα (sα α
X
3 s3 ) = 0
α
mα (sα α mα (sα α mα (sα α
X X X
1 s2 ) = 2 s3 ) = 3 s1 ) = 0
α α α
X
α α α
m (s · s ) = ma (4/3 + 4/3 + 4/3) = 4ma2
2

α
Thus
     
1 0 0 1 0 0 1 0 0
2 2 2
I(G) = 4ma  0 1 0  − 2ma  0 1 0  = 2ma  0 1 0 
 

0 0 1 0 0 0 0 0 2

An equally acceptable method is to calculate the inertia tensors for the indi-
vidual masses and add them together, as in lectures.
 (iii) The Parallel Axes Theorem states that
n o
Iij (O) − Iij (G) = M (R · R) δij − Ri Rj
√
Now M = 3m, R · R = 4a2 /3 and R1 = a , R2 = a/ 3 and R3 = 0. Thus
     √ 
1 0 0 2 1 0 0 √ 3 3 0
4a 
I(O) = 2ma2  0 1 0  + 3m ×
  2
 0 1 0  − ma  3 1 0 

3
0 0 2 0 0 1 0 0 0
 √ 
√3 − 3 0 
= ma2  − 3 5 0 
0 0 8

7 With reference to a given Cartesian basis a second-rank, symmetric tensor T has com-
ponents
 
4 1 1
 1 4 −1  .


1 −1 2
Calculate its eigenvalues and eigenvectors, and verify that the latter are orthogonal.
Construct a transformation matrix λ whose rows are the normalised eigenvectors of the
tensor T . Verify that λ λT = δ (where δ denotes the identity matrix) and that λ T λT is
diagonal, with its diagonal components the eigenvalues of T .
The characteristic equation is

4−t 1 1
1 4 − t −1 = (4 − t)[(4 − t)(2 − t) − 1] − [(2 − t) + 1] − [−1 − 4 + t] = 0
1 −1 2 − t

Thus
(4 − t)[5 − 6t + t2 ] = 0 ⇒ (1 − t)(4 − t)(5 − t) = 0
Thus the eigenvalues are 1, 4 and 5.
For t = t(1) = 1 we denote the corresponding eigenvector by n(1) and the equations for
the components of n(1) are (dropping the label (1))
    
3 1 1 n1 0
 1   n2  =  0 
3 −1  ⇒ n1 = −n2 , n3 = 2n2
    

1 −1 1 n3 0
 
−1
1 
Thus a normalised eigenvector is: n̂(1) =√  1   For t = t(2) = 4
6 2
    
0 1 1 n1 0
 1

0 −1   n2  =  0 
  
 ⇒ n1 = n3 , n2 = −n3
1 −1 −2 n3 0
  
1
1
Thus a normalised eigenvector is: n̂(2) =√   −1 

3 1
For t = t(3) = 5
    
−1 1 1 n1 0
 1 −1 −1   n2  =  0  ⇒ n1 = n 2 , n3 = 0
    

1 −1 −3 n3 0
 
1
1
Thus a normalised eigenvector is: n̂(3) =√  1 

2 0
Clearly
1
n̂(1) · n̂(2) = √ (−1 − 1 + 2) = 0
18
1
n̂(1) · n̂(3) = √ (−1 + 1 + 0) = 0
12
1
n̂(2) · n̂(3) = √ (1 − 1 + 0) = 0
6

(i)
Now suppose that λij = nj then

1 1 2
 
−√ √ √
6 6 6
 
 
1 1 1
 
√ −√ √
 
λ= 

 3 3 3 


1 1 
√ √ 0
 
2 2
1 1 2 1 1 1
  
−√ √ √   −√ √ √   

 6 6 6
 6 3 2  1 0 0
1 1 1 1 1 1 
    
T
√ −√ √ √ − √ √
   
λλ = = 0 1 0 
     
 3 3 3

 6 3 2   

1 1 
2 1 
0 0 1
√ √ 0 √ √ 0
  
2 2 6 3
1 4 5
 
 −√ √ √  

4 1 1

 6 3 2  1 0 0

 1 4 5 
λ T λT = λ  1 4 −1  λT = λ  √ −√ √  = 0 4 0 
    

1 −1 2
 6 3 2 

0 0 5
2 4

√ √ 0
 
6 3
Thus by choosing an orthonormal basis corresponding to the normalised eigenvectors of
the tensor, we find that the tensor is diagonal. This procedure of transforming to the
basis provided by the normalised eigenvectors is referred to as diagonalisation.
8 Calculate the gradient ∇f for f (r) = x2 +3y 2 +2z 2 . At the point (2, 3, 1) calculate (a) the
gradient to the level surface; (b) the unit normal; (c) the equation of the tangent plane;
(d) the directional derivative in the direction of the vector 2e1 − e2 ; (e) the maximum
value of f subject to the condition r2 = 1.
For the given scalar field
!
∂ ∂ ∂
∇f = e1 + e2 + e3 (x2 + 3y 2 + 2z 2 ) = 2xe1 + 6ye2 + 4ze3
∂x ∂y ∂z

(a) From this expression for ∇f we find

∇f = 4e1 + 18e2 + 4e3 at the point (2, 3, 1)

(b) To find the unit normal we divide ∇f by its magnitude:

∇f 1
n̂ = = √ (2e1 + 9e2 + 2e3 )
|∇f | 89

(c) The equation of the tangent plane is the equation of a plane with normal n̂ and
containing the point r0 . Thus

r · n̂ = r0 · n̂ ⇒ 2x + 9y + 2z = 33

(d) The directional derivative along a is â · ∇f , where a = 2e1 − e2 . Thus

1 10 √
â = √ (2e1 − e2 ) ⇒ â · ∇f = − √ = −2 5
5 5
where we have used the result of part (a) for ∇f at (2, 3, 1).
(e) If the constraint function is g(r) = r2 = 1, then we introduce a Lagrange multiplier λ
and look for ∇(f +λg) = 0 (note it would have been messier if we had taken g(r) = |r| =
1). f +λg = (1+λ)x2 +(3+λ)y 2 +(2+λ)z 2 , so ∇(f +λg) = 2[(1+λ)x, (3+λ)y, (2+λ)z].
This could vanish at the origin (but only for the case g = 0); otherwise we need λ =
−1, −3 or −2. These stationary points correspond to y = z = 0, x = z = 0, x = y = 0
respectively, with the remaining coordinate being of unit magnitude in each case (using
g = 1). Considering each of these in turn, the maximum value of f comes when y 2 = 1,
so the conditional maxima are at y = ±1, f = 3.

9 (i) Evaluate the line integral Z

F · dr
C

with F = (y, −x, 0), from the point (a, 0, 0) to the point (a, 0, 2πb) along
a) a circular helix, parameterized by

r = (a cos λ, a sin λ, bλ)

 .
b) a straight line, parameterized by

r = (a, 0, bλ) .

(ii) Repeat for the field F = r. and comment on the dependence of the integral on
path.

(i) a) The end points of the path correspond to λ = 0 and λ = 2π respectively.

With the given parameterization, we have x = a cos λ and y = a sin λ on the
helix so that

F = (a sin λ, −a cos λ, 0)
r = (a cos λ, a sin λ, bλ)
dr = (−a sin λ, a cos λ, b)dλ

Thus
F · dr = (−a2 sin2 λ − a2 cos2 λ) dλ = −a2 dλ
and we can write the line integral as
Z Z 2π
2
F · dr = −a dλ = −2πa2
C 0

b) The straight line between the point (a, 0, 0) to the point (a, 0, 2πb) is pa-
rameterized as
r = (a, 0, bλ) where λ : 0 → 2π
Thus dr = (0, 0, b)dλ, F = (y, −x, 0) = (0, −a, 0) and hence F · dr = 0.
(ii) a) This time F = r and so

F = (a cos λ, a sin λ, bλ)

dr = (−a sin λ, a cos λ, b)dλ

so that
F · dr = (−a2 cos λ sin λ + a2 cos λ sin λ + b2 λ) dλ
giving #2π
λ2
"
Z Z 2π
2 2
F · dr = b λdλ = b = 2π 2 b2
C 0 2 0

b) For the straight line path with F = r

r = (a, 0, bλ)
dr = (0, 0, b)dλ
F = (a, 0, bλ)
so that Z Z 2π
F · dr = b2 λdλ = 2π 2 b2
C 0
as before.
We see that the line integral in part (i) is path dependent but in part (ii) is
path independent. In part (i) the field has non-zero curl:

e1 e2 e3
∂ ∂ ∂
∇×F = ∂x ∂y ∂z
= 2e3
−y x 0

whereas in part (ii) the field F = r is irrotational :

∇×r =0