Lecture 23: Curvilinear Coordinates (RHB 8.

10)
It is often convenient to work with variables other than the Cartesian coordinates xi ( = x,
y, z). For example in Lecture 15 we met spherical polar and cylindrical polar coordinates.
These are two important examples of what are called curvilinear coordinates. In this
lecture we set up a formalism to deal with these rather general coordinate systems.
Actually we have effectively covered much of the materical in lectures 18 and 19 where we
studied surface and volume integrals. There we considered parameterisation of surfaces and
volumes. Here we do the same thing but think about how this realises non-Cartesian Co-
ordinate systems.

Suppose we change from the Cartesian coordinates (x1 , x2 , x3 ) to the curvilinear coordinates,
which we denote ui , each of which are functions of the xi :-

u1 = u1 (x1 , x2 , x3 )
u2 = u2 (x1 , x2 , x3 )
u3 = u3 (x1 , x2 , x3 )

The ui should be single-valued, except possibly at certain points, so the reverse transforma-
tion,
xi = xi (u1 , u2 , u3 )
can be made. A point may be referred to by its Cartesian coordinates xi , or by its curvilinear
coordinates ui . For example, in 2-D, we might have:-

Now consider coordinate surfaces defined by keeping one coordinate constant.

• The Cartesian coordinate surfaces ‘xi = constant’ are planes, with constant unit
normal vectors ei (or e1 , e2 and e3 ), intersecting at right angles.

• The surfaces ‘ui = constant’ do not, in general, have constant unit normal vectors,
nor in general do they intersect at right angles.

89
Example: Spherical polar co-ordinates
( )
z y
q  
−1 −1
r = x2 + y2 + z2 ; θ = cos √ 2 ; φ = tan .
x + y2 + z2 x

The surfaces of constant r, θ, and φ are:-

r = constant =⇒ spheres centred at the origin unit normal er
θ = constant =⇒ cones of semi-angle θ and axis along the z-axis unit normal eθ
φ = constant =⇒ planes passing through the z-axis unit normal eφ

These surfaces are not all planes, but they do intersect at right angles.
If the coordinate surfaces intersect at right angles (i.e. the unit normals intersect at right
angles), as in the example of spherical polars, the curvilinear coordinates are said to be
orthogonal.

23. 1. Orthogonal Curvilinear Coordinates

Unit Vectors and Scale Factors

Suppose the point P has position r = r(u1 , u2 , u3 ). If we change u1 by a small amount, du1 ,
then r moves to position (r + dr ), where
∂r
dr = du1 ≡ h1 e1 du1
∂u1
where we have defined the unit vector e1 and the scale factor h1 by
 
 ∂r  1 ∂r
h1 =  and e1 = .
 

 ∂u1  h1 ∂u1

• The vector e1 is a unit vector in the direction of increasing u1 .

• The scale factor h1 gives the magnitude of dr when we make the change u1 → u1 +du1 .
Thus for an infinitessimal change of u1
|dr| = h1 du1

90
Similarly, we can define hi and ei for i = 2 and 3.

• The unit vectors ei are not constant vectors. In general they are non-
Cartesian basis vectors, they depend on the position vector r, i.e. their
directions change as the ui are varied.

• If ei · ej = δij , then the ui are orthogonal curvilinear coordinates.

For Cartesian coordinates, the scale factors are unity and the unit vectors ei reduce to the
Cartesian basis vectors we have used throughout the course:
∂r
r = x e1 + y e2 + z e3 so that h1 e1 = = e1 , etc.
∂x

Example: spherical polars: u1 = r, u2 = θ and u3 = φ in that order:

r = r sin θ cos φ i + r sin θ sin φ j + r cos θ k,

where to avoid confusion in this section we use i, j, k for the Cartesian basis vectors. (By
this stage there should be no confusion with the suffices i, j, k). Therefore
∂r
= sin θ cos φ i + sin θ sin φ j + cos θ k =⇒ hr = 1
∂r
∂r
= r cos θ cos φ i + r cos θ sin φ j − r sin θ k =⇒ hθ = r
∂θ
∂r
= −r sin θ sin φ i + r sin θ cos φ j =⇒ hφ = r sin θ
∂φ
Thus
er = sin θ cos φ i + sin θ sin φ j + cos θ k = r / r
eθ = cos θ cos φ i + cos θ sin φ j − sin θ k
eφ = − sin φ i + cos φ j

These unit vectors are normal to the level surfaces described above (spheres, cones and
planes) and are clearly orthogonal: er · eθ = er · eφ = eθ · eφ = 0
and form a RH orthonormal basis: er × eθ = eφ , eθ × eφ = er , eφ × er = eθ .

Example: Cylindrical polars: u1 = ρ, u2 = φ and u3 = z in that order:

r = ρ cos φ i + ρ sin φ j + z k .

Thus ∂r ∂r ∂r
= cos φ i + sin φ j ; = −ρ sin φ i + ρ cos φ j ; = k
∂ρ ∂φ ∂z
and
hρ = 1 ; hφ = ρ ; hz = 1
Therefore
eρ = cos φ i + sin φ j ; eφ = − sin φ i + cos φ j ; ez = k

91
These unit vectors are normal to the level surfaces described by cylinders about the z-axis
(ρ = constant), planes through the z-axis (φ = constant), planes perpendicular to the z axis
(z = constant) and are clearly orthogonal.

Remark: An example of a curvilinear coordinate system which is not orthogonal is provided

by the system of elliptical cylindrical coordinates (see tutuorial 9.4).
r = a ρ cos θ i + b ρ sin θ j + z k (a 6= b)

In the following we shall only consider orthogonal systems

Arc Length
The arc length ds is the length of the infinitesimal vector dr :-

(ds)2 = dr · dr .

In Cartesian coordinates
(ds)2 = (dx)2 + (dy)2 + (dz)2 .
In curvilinear coordinates, if we change all three coordinates ui by infinitesimal amounts
dui , then we have

∂r ∂r ∂r
dr = du1 + du2 + du3
∂u1 ∂u2 ∂u3
= h1 du1 e1 + h2 du2 e2 + h3 du3 e3

For the case of orthogonal curvilinears, because the basis vectors are orthonormal we have

(ds)2 = h21 du21 + h22 du22 + h23 du23

For spherical polars, we showed that

hr = 1, hθ = r, and hφ = r sin θ

therefore
(ds)2 = (dr)2 + r2 (dθ)2 + r2 sin2 θ (dφ)2

92
Lecture 23 continued:

23. 2. Elements of Area and Volume

Basically we just repeat using scale factors what we did in lectures 18 and 19.
Vector Area
If u1 → u1 + du1 , then r → r + dr 1 , where dr 1 = h1 e1 du1 , and if u2 → u2 + du2 , then
r → r + dr 2 , where dr 2 = h2 e2 du2 . (Curvature greatly exaggerated in figure!)

On the surface of constant u1 the vector area bounded by dr2 and dr 3 is given by

dS 1 = (dr 2 ) × (dr3 ) = (h2 du2 e2 ) × (h3 du3 e3 ) = h2 h3 du2 du3 e1 ,

since e2 × e3 = e1 for orthogonal systems.

Thus dS1 is a vector pointing in the direction of the normal to the surfaces ‘u1 =constant’,
its magnitude being the area of the small parallelogram with edges dr 2 and dr 3 . Similarly,
one can define dS 2 and dS 3 .
For the case of spherical polars, if we vary θ and φ, keeping r fixed, then
   
dS r = hθ dθ eθ × hφ dφ eφ = hθ hφ dθ dφ er = r2 sin θ dθ dφ er .

Similarly for dS θ and dS φ .

Volume
The volume contained in the parallelepiped with edges dr1 , dr2 and dr3 , is

dV = dr1 · dr2 × dr3

= (h1 du1 e1 ) · (h2 du2 e2 ) × (h3 du3 e3 )
= h1 h2 h3 du1 du2 du3

because e1 · e2 × e3 = 1.
For spherical polars, we have

dV = hr hθ hφ dr dθ dφ = r2 sin θ dr dθ dφ

93
Components of a Vector Field in Curvilinear Coordinates
A vector field A(r) can be expressed in terms of curvilinear components Ai , defined as:–
X
A(r) = Ai (u1 , u2 , u3 ) ei
i

where ei is the ith basis vector for the curvilinear coordinate system.
For orthogonal curvilinear coordinates, the component Ai is obtained by taking the scalar
product of A with the ith (curvilinear) basis vector ei

Ai = ei · A(r)

NB Ai must be expressed in terms of ui (not x, y, z) when working in the ui basis.

Example If A = i in Cartesian coordinates, then in spherical polars,Ar = A·er = sin θ cos φ,

etc.
A(r, θ, φ) = sin θ cos φ er + cos θ cos φ eθ − sin φ eφ .

Example If A = r then in cylindrical polars Aρ = A · eρ = ρ cos2 φ + ρ sin2 φ = ρ etc, and

A(ρ, φ, z) = r = ρ eρ + z ez

23. 3. Grad, Div, Curl, and the Laplacian in Orthogonal Curvilinears

We defined the vector operators grad, div, curl firstly in Cartesian coordinates, then most
generally through integral definitions without regard to a coordinate system. Here we com-
plete the picture by providing the definitions in any orthogonal curvilinear coordinate system.
Gradient
In section (2) we defined the gradient in terms of the change in a scalar field f when we let
r → r + dr
δf = ∇ f (r) · dr (1)
Now consider what happens when we write f in terms of orthogonal curvilinear coordinates
f = f (u1 , u2 , u3 ). As before, we denote the curvilinear basis vectors by e1 , e2 , and e3 .
Let u1 → u1 + du1 , u2 → u2 + du2 , and u3 → u3 + du3 .
By Taylor’s theorem, we have
∂f ∂f ∂f
δf = du1 + du2 + du3
∂u1 ∂u2 ∂u3
We have already shown that
dr = h1 du1 e1 + h2 du2 e2 + h3 du3 e3 .

94
Using the orthogonality of the basis vectors, ei · ej = δij , we can write
!
∂f ∂f ∂f
δf = e1 + e2 + e · (e1 du1 + e2 du2 + e3 du3 )
∂u1 ∂u2 ∂u3 3
!
1 ∂f 1 ∂f 1 ∂f
= e1 + e2 + e · (h1 e1 du1 + h2 e2 du2 + h3 e3 du3 )
h1 ∂u1 h2 ∂u2 h3 ∂u3 3
!
1 ∂f 1 ∂f 1 ∂f
= e1 + e2 + e · dr
h1 ∂u1 h2 ∂u2 h3 ∂u3 3
Comparing this result with equation (1) above, we obtain the following expression for ∇ f
in orthogonal curvilinears

1 ∂f 1 ∂f 1 ∂f
∇f = e1 + e2 + e
h1 ∂u1 h2 ∂u2 h3 ∂u3 3
3
X 1 ∂f
= e
i=1 hi ∂ui i

For spherical polars, we obtain

∂f 1 ∂f 1 ∂f
∇ f (r, θ, φ) = er + eθ + eφ .
∂r r ∂θ r sin θ ∂φ
For cylindrical polars, we obtain
∂f 1 ∂f ∂f
∇ f (ρ, φ, z) = eρ + eφ + ez .
∂ρ ρ ∂φ ∂z

Divergence
In orthogonal curvilinear coordinates

( )
1 ∂ ∂ ∂
divA = (A1 h2 h3 ) + (A2 h3 h1 ) + (A3 h1 h2 )
h1 h2 h3 ∂u1 ∂u2 ∂u3

This expression can be obtained by using the integral definition of divA, or alternatively
using vector operator identities (see BK 4.13, RHB 8.10).

For Cartesian coordinates, we have hi = 1, and we regain the usual expression for ∇ · A in
Cartesians.
For spherical polars we have
( )
1 ∂  2  ∂   ∂  
divA(r, θ, φ) = 2 r sin θAr + r sin θAθ + rAφ
r sin θ ∂r ∂θ ∂φ
( )
1 ∂  2  1 ∂   ∂  
= 2 r Ar + sin θAθ + A .
r ∂r r sin θ ∂θ ∂φ φ

95
where Ar , Aθ , and Aφ are the components of the vector field A in the basis (er , eθ , eφ ).
For cylindrical polars we have
( )
1 ∂   ∂ ∂
divA(ρ, φ, z) = ρAρ + Aφ + (ρAz )
ρ ∂ρ ∂φ ∂z
1 ∂   1 ∂ ∂
= ρAρ + Aφ + Az .
ρ ∂ρ ρ ∂φ ∂z
where Aρ , Aφ , and Az are the components of the vector field A in the basis (eρ , eφ , ez ).
Curl
In orthogonal curvilinear co-ordinates, curl is most conveniently written as
 
 h1 e1 h2 e2 h3 e3
 

 
1  ∂ ∂ ∂ 
∇×A = 



h1 h2 h3  ∂u1
 ∂u2 ∂u3 

 h1 A1 h2 A2 h3 A3
 


e.g. the first component is given by

( )
1 ∂ ∂
e1 · ∇ × A = (A3 h3 ) − (A2 h2 )
h2 h3 ∂u2 ∂u3
and the components of ∇ × A in the e2 and e3 directions may be obtained by cyclic permu-
tations of the suffices.
The above formula can be demonstrated by using the line integral definition of curl, as used
to proof Stokes’ theorem, (see tutorial) or by vector operator identities (BK. 4.13 or RHB
8.10).
For spherical polars we have
 
er r eθ r sin θ eφ
 
 
 
1  ∂ ∂ ∂ 
∇×A = 2 



r sin θ  ∂r ∂θ ∂φ 

Ar rAθ r sin θ Aφ
 
 

Laplacian of a Scalar Field

The Laplacian operator acting on a scalar field is defined by ∇2 f = ∇ · (∇ f ), giving:–

( ! ! !)
2 1 ∂ h2 h3 ∂f ∂ h3 h1 ∂f ∂ h1 h2 ∂f
∇f = + +
h1 h2 h3 ∂u1 h1 ∂u1 ∂u2 h2 ∂u2 ∂u3 h3 ∂u3

In spherical polars, we have

( ! ! !)
2 1 ∂ ∂f ∂ ∂f ∂ 1 ∂f
∇ f (r, θ, φ) = 2 r2 sin θ + sin θ +
r sin θ ∂r ∂r ∂θ ∂θ ∂φ sin θ ∂φ
∂2f
! ( ! )
1 ∂ ∂f 1 ∂ ∂f
= 2 r2 + 2 2 sin θ sin θ + .
r ∂r ∂r r sin θ ∂θ ∂θ ∂φ2

96