eNote 26 1

eNote 26

Vector Fields

In eNote 10 vectors in the 2D plane and in 3D space were introduced and studied. In eNote 20
we considered the gradients of functions f ( x, y) of two variables. A gradient vector field of a
function of two variables is – as the name amply hints – an example of a plane vector field. In
this eNote we will begin to study vector fields in general, both in the ( x, y) plane and in
3-dimensional ( x, y, z) space. We will clarify what it means to flow with a given vector field
and compute where you then arrive at in the space or in the plane in this way after a given
period of time. In order to find these so-called flow curves we need to be able to solve (suitably
simple) systems of first-order differential equations. Thus eNote 16 together with the two
eNotes above become background material for the present eNote. We will also begin to
investigate what happens to larger systems of points or particles when they individually flow
with the vector field.

Updated: 31-01-2023, shsp.

26.1 Vector Fields

A vector field V in 3D space is given by 3 smooth functions V1 ( x, y, z) , V2 ( x, y, z) and

V3 ( x, y, z) that are all functions of the three variables x, y, and z like this:

V( x, y, z) = (V1 ( x, y, z) , V2 ( x, y, z) , V3 ( x, y, z) ) for ( x, y, z) ∈ R3 . (26-1)

eNote 26 26.1 VECTOR FIELDS 2

A vector field V( x, y, z) is drawn and usually stated in ( x, y, z) space by – at a

suitable number of chosen points ( xi , yi , zi ) – plotting the vector as an arrow
with the base point at ( xi , yi , zi ) and the end point at ( xi + V1 ( xi , yi , zi ), yi +
V2 ( xi , yi , zi ), zi + V3 ( xi , yi , zi )). There can be good reasons to state the vector
field in other ways. E.g. if there is great variation in the length of the vectors
in a given vector field, then it can be advantageous to use the thickness of the
arrows as indication of the vector length.

A vector field in the plane is similarly given by the two coordinate functions V1 ( x, y)
and V2 ( x, y) that are both smooth functions of the two variables x and y:

V( x, y) = (V1 ( x, y) , V2 ( x, y) ) for ( x, y) ∈ R2 . (26-2)

The gradient vector field of a smooth function f ( x, y) of two variables (introduced and
studied in eNote 20) is an example of a vector field in the plane, see Figure 26.1.

Figure 26.1: A graph surface of a function of two variables and the corresponding gradient
vector field in the plane together with some of the level curves. The gradient vector field is
everywhere perpendicular to the level curves, see eNote 20.

The gradient vector field of functions f ( x, y, z) of three variables is defined similarly to

that of two variables:
eNote 26 26.1 VECTOR FIELDS 3

Definition 26.1 Gradient Field in ( x, y, z) Space

Let f ( x, y, z) denote a smooth function of three variables in R3 . Then the gradient
vector field of f ( x, y, z) is defined in the following way by the use of the first three
partial derivatives of f ( x, y, z):
 
∇ f ( x, y, z) = f x′ ( x, y, z) , f y′ ( x, y, z) , f z′ ( x, y, z) , ( x, y, z) ∈ R3 . (26-3)

Example 26.2 A Gradient Field in ( x, y, z) Space

We let f ( x, y, z) denote the quadratic polynomial

f ( x, y, z) = 2 · x2 + 2 · y2 + 2 · z2 − 2 · x · z − 2 · x − 4 · y − 2 · z + 3 . (26-4)

The gradient vector field of f ( x, y, z) is then:

∇ f ( x, y, z) = (4 · x − 2 · z − 2 , 4 · y − 4 , −2 · x + 4 · z − 2) . (26-5)

See Figure 26.2. We refer to Example 24.5 in eNote 24 about the construction of the ellipsoidal
level surface K0 ( f ) of f ( x, y, z) shown. The level surface and the computed gradient vector
field are indicated in Figure 26.2. The gradient vector field is seen to be perpendicular to the
level surface.

The question is now, do all smooth vector fields in the plane and all smooth
vector fields in 3D space stem from a function in the way that each individu-
ally is the gradient vector field of some function of two and three variables,
respectively? But it is not that simple!
eNote 26 26.1 VECTOR FIELDS 4

Figure 26.2: The level surface (in an open version) of a function (quadratic polynomial) of three
variables and some corresponding gradient vectors from the gradient vector field in ( x, y, z)
space.

Example 26.3 A Vector Field that Is Not a Gradient Vector Field

Let V( x, y) denote the very simple vector field in the plane V( x, y) = (−y, x ), where
( x, y) ∈ R2 . Then no function f ( x, y) exists that satisfies ∇ f ( x, y) = V( x, y).

Viz. if we (until we reach a contradiction) assume that such a function with this property
exists:
∇ f ( x, y) = V( x, y) , such that
 
f x′ ( x, y), f y′ ( x, y) = (−y, x ) , then we get that

f x′ ( x, y) = −y
(26-6)
f y′ ( x, y) = x , and thereby

′′
f xy ( x, y) = −1
′′
f yx ( x, y) = 1 ,

and this is not compatible with the fact that for all smooth functions

′′ ′′
f xy ( x, y) = f yx ( x, y) , for all ( x, y) ∈ R2 . (26-7)
eNote 26 26.1 VECTOR FIELDS 5

This shows that a function whose gradient vector field is the given vector field does not
exist.

The gradient vector fields are thus only examples of vector fields – but a very
large and very important collection of examples of vector fields.

A vector field V( x, y) in the ( x, y) plane can easily be extended to a vector

field W( x, y, z) in ( x, y, z) space by simply displacing all the vectors from the
plane in the direction of the z-axis and in addition putting W3 ( x, y, z) = 0 for
all ( x, y, z) ∈ R3 :

The quite general plane vector field V( x, y) = (V1 ( x, y), V2 ( x, y)) thus has the
following spatial extension:

W( x, y, z) = (V1 ( x, y), V2 ( x, y), 0) i.e.

W1 ( x, y, z) = V1 ( x, y) , (26-8)
W2 ( x, y, z) = V2 ( x, y) ,
W3 ( x, y, z) = 0 .

See Figure 26.3 that hints at the extensions of the three different vector fields
V( x, y) = (1, 0), V( x, y) = ( x, y), and V( x, y) = (−y, x ), that is, W( x, y, z) =
(1, 0, 0), W( x, y, z) = ( x, y, 0), and W( x, y, z) = (−y, x, 0), respectively.

Figure 26.3: Three plane vector fields are here extended to spatial vector fields.
eNote 26 26.1 VECTOR FIELDS 6

Figure 26.4: The three spatial vector fields from Figure 26.3 are here modified to have W3 = 1/2
in place of W3 = 0 .

Some vector fields are particularly simple. In particular this applies to those vector
fields where all three coordinate functions are polynomials of at the most the first degree
in the spatial variables ( x, y, z), i.e.

V( x, y, z) = ( a11 x + a12 y + a13 z + b1 ,

a21 x + a22 y + a23 z + b2 , (26-9)
a31 x + a32 y + a33 z + b3 ) .

In this case the vector field can be written in short form by use of the matrix, A , that
has the elements aij and the vector, b, that has the coordinates bi :

Definition 26.4 Vector Field of the First Degree

A vector field of the first degree is a vector field V( x, y, z) that can be written in the
following form by the use of a constant matrix A and a constant vector b :
⊤
V⊤ = (V( x, y, z))⊤ = A · + b⊤

x y z , (26-10)

where ⊤ means transposition of the respective matrices, such that

       
V1 ( x, y, z) a11 a12 a13 x b1
 V2 ( x, y, z)  =  a21 a22 a23  ·  y  +  b2  . (26-11)
V3 ( x, y, z) a31 a32 a33 z b3
eNote 26 26.1 VECTOR FIELDS 7

Example 26.5 Constant Vector Field

A constant vector field can e.g. model a constant wind locally close to the ( x, y) plane (the
face of the earth):
V( x, y, z) = b , (26-12)
where b is a constant vector, e.g. b = (0, 7, 0) – if the wind blows with 7km/h in the direction
of the y-axis.

Example 26.6 Rotating Vector Field

An example of a so-called rotating vector field is given by

V( x, y, z) = (−y, x, 0) . (26-13)

See Figure 26.3 in the middle.

Definition 26.7
The trace of a square n × n matrix A with the elements ai j is the sum of the n diagonal
elements of the matrix:
i =n
trace( A) = ∑ ai i . (26-14)
i =1

Exercise 26.8

Find A and b (as in Definition 26.4) for the vector field in example 26.6. What is the trace of
A in this case? Can A be diagonalized (diagonalization is described in eNote 14)?

Example 26.9 Explosion and Implosion Vector Fields

An example of what we could call an explosion vector field is given by the following coordinate
functions (see why in Figure 26.6):

V( x, y, z) = ( x, y, z) . (26-15)
eNote 26 26.2 FLOW CURVES OF A VECTOR FIELD 8

Similarly the following is an example of a vector field that we can call an implosion vector field
(see why in Figure 26.7):
V( x, y, z) = (− x, −y, −z) . (26-16)

Exercise 26.10

Find A and b for the vector fields in Example 26.9. What is the trace of A for the two vector
fields? Can A be diagonalized?

We will now argue the (dynamic) names that we have given the vector fields in the
above examples 26.6 and 26.9. To do so we will move together with – or flow along –
the vector field in a very precise way that we shall now define.

26.2 Flow Curves of a Vector Field

Let us first repeat that if we are given a curve with a parametric representation

Kr : r(t) = ( x (t), y(t), z(t)) ∈ R3 , t ∈ [ a, b] , (26-17)

then this curve has for every value of the parameter t a tangent vector, viz.

r′ (t) = ( x ′ (t), y′ (t), z′ (t)) . (26-18)

If we consider the parameter t ∈ [ a, b] as a time parameter for the motion (of a particle) in
( x, y, z) space that is given by r(t) then r′ (t) is the velocity of the particle at time t.

If we construct sufficiently many curves (a curve through every point in the given
space), each curve intersecting neither itself nor any other curve, then we in this way
get a vector field in the space.

The obvious inverse question is now: Given an initial point p = ( x0 , y0 , z0 ) and given
a vector field V( x, y, z) in ( x, y, z) space, then does a parametrized curve r(t) through
p (with r(0) = ( x0 , y0 , z0 ) ) exist, such that the tangent vector field of the curve all the
way along the curve precisely is the vector field V( x, y, z) along the curve? If this is the case
then we shall call the curve r(t) an integral curve or a flow curve of the vector field. These
names are partly due to the fact that the curve can be found by integration (solution to
eNote 26 26.2 FLOW CURVES OF A VECTOR FIELD 9

a system of differential equations) and partly that motion along the curve is similar to
flying or floating with the given vector field, i.e. with a speed and a direction given by
the vector field at every point of the motion, since the requirements to the motion r(t)
are expressed by:

Definition 26.11 Flow Curves, Integral Curves

Let V( x, y, z) denote a smooth vector field in ( x, y, z) space. A parametrized curve

Kr : r(t) = ( x (t), y(t), z(t)) , t ∈ [ a, b] , (26-19)

is called a flow curve or an integral curve of the vector field V( x, y, z) if r(t) fulfills the
flow curve equation:

V(r(t)) = r′ (t) for all t ∈ [ a, b] , (26-20)

which is equivalent to the following system of first-order differential equations

x ′ (t)
   
V1 ( x (t), y(t), z(t))
 y′ (t)  = (V( x (t), y(t), z(t)))⊤ =  V2 ( x (t), y(t), z(t))  . (26-21)
z′ (t) V3 ( x (t), y(t), z(t))

If V( x, y, z) is given and if we have been given an initial point p = r( a) for

a flow curve then the task is of course the typical one, to find the solution to
the system of differential equations with this initial condition, i.e. to find the
coordinate functions x (t), y(t), and z(t) so that p = ( x ( a), y( a), z( a)).

In other words: If we are given a vector field in 3D space then the task is to start
a particle (a small ball) moving along the vector field such that the velocity
vector of the ball at every instant is given by the value of the vector field in
the point where the ball is positioned at that point in time. And it is of course
interesting to be able to decide where the ball is after a long period of time. And
it is interesting to find out how a multitude of balls (particles that to begin with
are close to each other) develop in time – is the multitude of balls more dense
or more thin, squeezed together or stretched out?

The following existence and uniqueness theorem applies and will be the foundation for
our first examples and the first consideration about the natural questions concerning
eNote 26 26.2 FLOW CURVES OF A VECTOR FIELD 10

flow curves and their behaviour.

Theorem 26.12 Existence and Uniqueness

Let V( x, y, z) be a vector field of the first degree, given by a coefficient matrix A and
a vector b as in Definition 26.4. Let ( x0 , y0 , z0 ) denote an arbitrary point in ( x, y, z)
space. Then exactly one curve r(t) exists that fulfills the two conditions:

r(0) = ( x0 , y0 , z0 ) and
(26-22)
′
r (t) = V( x (t), y(t), z(t)) for all t ∈ [−∞, ∞] .

The last equation (26-22) is equivalent to the following system of differential equa-
tions with a constant coefficient matrix A:
 ′ 
x (t)
 y′ (t)  = (V( x (t), y(t), z(t)))⊤
z′ (t)
⊤
+ b⊤

= A · x (t) y(t) z(t) (26-23)
     
a11 a12 a13 x (t) b1
= a21 a22 a23 · y(t) + b2  .
    
a31 a32 a33 z(t) b3

If we are given a vector field of the first degree, then we can therefore ”start” a point,
a particle, at an arbitrary position in the space and let it ”flow” with the vector field
such that the particle is situated on a uniquely determined flow curve to every time
thereafter.

Two flow curves cannot intersect each other, because if they did there could
not be a unique flow curve through the point of intersection.
eNote 26 26.2 FLOW CURVES OF A VECTOR FIELD 11

The theorem can be extended to vector fields that are not necessarily of the first
degree, but then it is no longer certain that all the time-parameter intervals
for the flow curves will be of the double infinity interval R =] − ∞, ∞[ . The
integral curves of a vector field of the first degree can be found and shown
with Maple and are exemplified in the figures 26.5, 26.6, and 26.7.

If the vector field is not of the first degree there is as previously stated no guar-
antee that flow curves can be determined explicitly (not even using Maple),
but in certain cases numerical tools can anyway be applied with success inside
”windows” where the solutions exist and are well-defined.

The argument, the proof, for Theorem 26.12 is known from the study of systems of linear
coupled differential equations, see eNote 17. Let us shortly repeat the considerations
needed in order to find the flow curves of some of the simplest vector fields.

Example 26.13 Flow Curves of a Constant Vector Field

The constant vector field V( x, y, z) = (0, 7, 0) has flow curves ( x (t), y(t), z(t)) that fulfill the
two conditions: The initial condition ( x (0), y(0), z(0)) = ( x0 , y0 , z0 ) and the 3 differential
equations for x (t), y(t), and z(t) following from the velocity vector condition

r′ (t) = ( x ′ (t), y′ (t), z′ (t)) = V( x (t), y(t), z(t)) = (0, 7, 0) . (26-24)

The task is to find the three coordinate functions x (t), y(t), and z(t) such that

x ′ (t) = 0
y′ (t) = 7 (26-25)
′
z (t) = 0 .

The 3 differential equations in this case are not coupled and they are solved directly with the
given initial conditions with the following result: x (t) = x0 , y(t) = y0 + 7 t, and z(t) = z0 . I.e.
the flow curves are (not surprisingly) all the straight lines parallel to the y-axis, parametrized
such that all have the speed 7.

Example 26.14 Flow Curves of a Rotating Vector Field

The example with the rotating vector field V( x, y, z) = (−y, x, 1) has corresponding flow
curves that now have to fulfill the conditions: ( x (0), y(0), z(0)) = ( x0 , y0 , z0 ) together with
the differential equations

r′ (t) = ( x ′ (t), y′ (t), z′ (t)) = (−y(t), x (t), 1) . (26-26)

eNote 26 26.2 FLOW CURVES OF A VECTOR FIELD 12

Therefore the task is here to find the three coordinate functions x (t), y(t), and z(t) such that

x ′ (t) = −y(t)
y′ (t) = x (t) (26-27)
′
z (t) = 1 .

The differential equations for x (t) and y(t) are coupled linear differential equations with
constant coefficients and are solved precisely as in eNote 17. Note that the system matrix
has already been found in Exercise 26.8. The result is x (t) = x0 cos(t) − y0 sin(t) , y(t) =
x0 sin(t) + y0 cos(t) , and z(t) = z0 + t. These flow curves can be found and inspected with
Maple. It is also apparent from this that it is quite reasonable to call the vector field a rotating
vector field. See Figure 26.5.

Figure 26.5: The rotating vector field from Example 26.14, one ”flow curve” for an individual
particle and the system of flow curves, passing through a cube (the nethermost cube flows along
the flow curves with the vector field until the time π).

Exercise 26.15

Let V( x, y, z) = (−y, x, 0) and use Maple to find flow curves and the motion of points in the
same cube as in Figure 26.5 when the time interval of the flow is T = [0, 2 π ] . Next compare
with ’the effect’ of the vector fields W( x, y, z) = (−y, − x, 0) W( x, y, z) = (−y, 2 x, 0) on
the points of the cube for the same time interval. Explain the difference between the three
’effects’ of the three different vector fields on the cube.
eNote 26 26.2 FLOW CURVES OF A VECTOR FIELD 13

Example 26.16 Flow Curves of an Explosion and Implosion Vector Field

The explosion vector field

V( x, y, z) = ( x, y, z) (26-28)
has flow curves that satisfy an initial condition

( x (0), y(0), z(0)) = ( x0 , y0 , z0 ) (26-29)

and the differential equations

r′ (t) = ( x ′ (t), y′ (t), z′ (t)) = ( x (t), y(t), z(t)) . (26-30)

We find the three coordinate functions x (t), y(t), and z(t) such that

x ′ (t) = x (t)
y′ (t) = y(t) (26-31)
z′ (t) = z(t) .

The differential equations for x (t), y(t), and z(t) are here uncoupled linear differential equa-
tions that can easily be solved, one at a time. The result is

x (t) = x0 exp(t) , y(t) = y0 exp(t) , and z(t) = z0 exp(t) . (26-32)

Note that if ( x (0), y(0), z(0)) = (0, 0, 0) then ( x (t), y(t), z(t)) = (0, 0, 0) for all t ∈ [−∞, ∞] .
Therefore the flow curve ’through’ the point (0, 0, 0) is not a proper curve but consists only
of the point itself. Note also that all other flow curves run arbitrarily close to the point
(0, 0, 0) for t → −∞, since exp(t) → 0 for t → −∞ , but they do not run through the point.
Therefore if we follow the flow curves in Figure 26.6 back in time from t = 0 through
negative values we will see an exponentially decreasing implosion of the cube. If we on
the contrary follow the flow curves forward in time from t = 0 through larger and larger
positive values for t we will see an exponentially increasing explosion of the cube. The flow
curves can again be found and inspected using Maple. See Figure 26.6.

The implosion vector field is given by

V( x, y, z) = (− x, −y, −z) (26-33)

with ”time-reversed” solution (as related to the explosion vector field)

x (t) = x0 exp(−t) , y(t) = y0 exp(−t) , og z(t) = z0 exp(−t) . (26-34)

See Figure 26.7.

eNote 26 26.3 THE DIVERGENCE OF A VECTOR FIELD 14

Figure 26.6: The explosion vector field from example 26.16 together with the integral curves
passing through a cube. The cube is shown as a solid to the time t = 1 and as ”open” to the time
t = 0. Compare with Figure 26.7.

Exercise 26.17

Let V denote the vector field V( x, y, z) = (− x, −2y, −3z) . Find and show a suitable number
of the flow curves of the vector field through the ball that has its centre at (1, 0, 0) and a radius
of 41 .

26.3 The Divergence of a Vector Field

For the purpose of geometric analysis of vector fields and their flow curve properties we
will here introduce two tools, two concepts, for the local description of general smooth
vector fields. The description is local because both concepts are expressed by the partial
derivatives of the coordinate functions of the vector field.
eNote 26 26.3 THE DIVERGENCE OF A VECTOR FIELD 15

Figure 26.7: The implosion vector field from Example 26.16 together with the integral curves
passing through a cube. The cube is shown as a solid to the time t = 1 and as ”open” to the time
t = 0. Compare with Figure 26.6.

Definition 26.18
Let V( x, y, z) = (V1 ( x, y, z), V2 ( x, y, z), V3 ( x, y, z) ) be a vector field in ( x, y, z)
space. The divergence of V at the point ( x0 , y0 , z0 ) is defined like this:

∂V1 ∂V ∂V
Div(V)( x0 , y0 , z0 ) = ( x0 , y0 , z0 ) + 2 ( x0 , y0 , z0 ) + 3 ( x0 , y0 , z0 ) . (26-35)
∂x ∂y ∂z

If V( x, y) = (V1 ( x, y), V2 ( x, y) ) is a plane vector field we define quite similarly:

∂V1 ∂V
Div(V)( x0 , y0 ) = ( x0 , y0 ) + 2 ( x0 , y0 ) . (26-36)
∂x ∂y

The divergence of a smooth vector field in R3 is a smooth function in R3 .

Note that the divergence of a plane vector field is the same as the divergence
of the spatial extension of the field.
eNote 26 26.3 THE DIVERGENCE OF A VECTOR FIELD 16

Example 26.19 Simple Divergences

Every constant vector field V( x, y, z) = b has the divergence Div(V) = 0 .

The explosion vector field V( x, y, z) = ( x, y, z) has the constant divergence Div(V) = 3 .

The implosion vector field V( x, y, z) = (− x, −y, −z) has the divergence Div(V) = −3 .

The rotating vector field V( x, y, z) = (−y, x, 0) also has a constant divergence Div(V) = 0 .

Exercise 26.20

Let V( x, y, z) = ( x + sin(y), z + cos(y), x + y − z). Determine Div(V) at every point in the

( x, y, z) space.

Exercise 26.21

Let V( x, y, z) be a vector field of the first degree with the matrix representation as in Equation
(26-10). Show that the divergence of V( x, y, z) is constant and equal to the trace of A.

Exercise 26.22

Let V( x, y, z) = ∇h( x, y, z) be the gradient vector field for a given function h( x, y, z) . Show
that the divergence of V( x, y, z) is

∂2 h ∂2 h ∂2 h
Div(∇h( x, y, z)) = + 2+ 2 . (26-37)
∂ x2 ∂y ∂z

In the applications of vector analysis the divergence of gradient vector fields of given
functions, Div(∇h( x, y, z)) is very often used and therefore is given its own name:
eNote 26 26.4 THE CURL OF A VECTOR FIELD 17

Definition 26.23
Let h( x, y, z) denote a smooth function in R3 . Then we write:

∆h( x, y, z) = Div(∇h( x, y, z))

∂2 h ∂2 h ∂2 h (26-38)
= + + 2 .
∂ x 2 ∂ y2 ∂z

The function ∆h( x, y, z) is called the Laplacian of the function h( x, y, z).

Example 26.24 The Laplacian

The Laplacian of some elementary functions of three variables:

Function ∇ f ( x, y, z) ∆ f ( x, y, z)
f ( x, y, z) = a · x + b · y + c · z ( a, b, c) 0
2 2
f ( x, y, z) = x + y + z 2 (2x, 2y, 2z) 6 (26-39)
f ( x, y, z) = y · sin( x ) (y · cos( x ), sin( x ), 0) −y · sin( x )
f ( x, y, z) = ex · cos(z) (ex · cos(z), 0 , −ex · sin(z)) 0

The Laplacian of a smooth function f ( x, y, z) is the trace of the 3 × 3-Hessian

matrix of f (see eNote 22):

∆ f ( x, y, z) = trace( H f ( x, y, z)) . (26-40)

26.4 The Curl of a Vector Field

The other quite central concept, a tool for the analysis of vector fields, is the following:
eNote 26 26.4 THE CURL OF A VECTOR FIELD 18

Definition 26.25 The Curl of a Vector Field

Let V( x, y, z) = (V1 ( x, y, z), V2 ( x, y, z), V3 ( x, y, z) ) be a vector field in ( x, y, z)
space. The curl of V at the point ( x0 , y0 , z0 ) is defined as the following vector:

∂V3 ∂V2
Rot(V)( x0 , y0 , z0 ) = ( ( x0 , y0 , z0 ) − ( x0 , y0 , z0 ) ,
∂y ∂z
∂V1 ∂V3
( x0 , y0 , z0 ) − ( x0 , y0 , z0 ) , (26-41)
∂z ∂x
∂V2 ∂V1
( x0 , y0 , z0 ) − ( x0 , y0 , z0 ) ) .
∂x ∂y

The curl of a smooth vector field in R3 is in itself a smooth vector field in R3 .

Example 26.26 The Curl of Simple Vector Fields

The explosion vector field V( x, y, z) = ( x, y, z) has constant curl Curl(V) = 0 .

The implosion vector field V( x, y, z) = (− x, −y, −z) has (not surprisingly) also constant curl
Curl(V) = 0 .

The rotating vector field (that rotates counter-clockwise) V( x, y, z) = (−y, x, 0) has constant
curl that of course is different from 0 : Curl(V) = (0, 0, 2) .

The rotating vector field (that rotates clockwise) V( x, y, z) = (y, − x, 0) also has constant curl
that of course is the opposite of the counter-clockwise rotation: Curl(V) = (0, 0, −2) .

Exercise 26.27

Let V( x, y, z) = ( x + sin(y), z + cos(y), x + y − z). Determine Curl(V) at every point in the

space.
eNote 26 26.5 A BRIDGE BETWEEN DIVERGENCE AND CURL 19

Exercise 26.28

Let V( x, y, z) be a vector field of the first degree with the matrix representation as in equation
(26-10). Show that the curl of V( x, y, z) is a constant vector and express the vector by the
elements in A.

26.5 A Bridge between Divergence and Curl

We mention here some relations between divergence, curl and gradient vector fields:

Theorem 26.29 Divergence Versus Curl

Let h( x, y, z) denote a smooth function in ( x, y, z) space. Then

Curl(∇h) = 0 . (26-42)

Let V( x, y, z) and W( x, y, z) denote two vector fields in R3 . Then the following

identity applies

Div(V × W) = Curl(V) · W − V · Curl(W) . (26-43)

Therefore we have in particular: If W is a gradient vector field of a function h( x, y, z)

in R3 , i.e. in short form W = ∇h , then

Div(V × ∇h) = Curl(V) · ∇h . (26-44)

Exercise 26.30

Show by direct computation that the two equations (26-42) and (26-43) both are satisfied.
eNote 26 26.6 FLOWS OF CURVES AND SURFACES 20

From the quite simple considerations and examples that we have been through
in this eNote it is reasonable to expect that the divergence is a measure of how
much a given collection of particles are spread or squished when they flow
with the vector field. This is exactly the content of Gauss’ theorem which is so
important for the application of these concepts that it is given its own eNote26.

Similarly we must expect that the total rcurl of a collection of particles flowing
with the vector field can be expressed by use of the rotation vector field for
the given vector field. This is exactly the content of Stokes’ Theorem, which
therefore also – for the same reason – has its own eNote 27.

26.6 Flows of Curves and Surfaces

As already hinted with the figures 26.5, 26.6 and 26.7 we can let any geometrically well-
defined set, surface, or curve flow with a given vector field – in such a way that ev-
ery point on the object follows, within the vector field, the unique flow curve passing
through the point. The idea is to understand the geometry of the vector field by observ-
ing how it moves and deforms geometric objects. See also figures 26.8 and 26.9.

Figure 26.8: A line segment flows with the flow curves of the vector field V( x, y, z) =
(−y, x, 0.3).
eNote 26 26.6 FLOWS OF CURVES AND SURFACES 21

Figure 26.9: A square flows with the flow curves of the vector field V( x, y, z) = (−y +
( x/9), −z + (y/9), − x + (z/9)).

26.6.1 The Flow of Level Curves and Level Surfaces

The gradient vector fields of the functions of two and three variables also deform curve
and surfaces via their respective flow curves. One could now be led to believe that level
curves and level surfaces probably flow over into other level curves and level surfaces
by the gradient vector flow. Yet it is not that simple – but almost.

By closer consideration one will realize that it cannot be the case that the gra-
dient vector field in general should make level sets flow into level sets. If e.g.
two neighboring level curves in Figure 26.1 are close to each other then the gra-
dient vector is correspondingly large and vice versa if two neighboring level
curves lie further apart then the gradient vector is correspondingly smaller. I.e.
where the gradient vectors are large we observe slower flow and where they
are small we see faster flow for the level curves to flow into each other.
eNote 26 26.6 FLOWS OF CURVES AND SURFACES 22

Theorem 26.31 Level Set Flow

Let f ( x, y, z) denote a smooth function of three variables with a proper gradient
vector field ∇ f ( x, y, z) ̸= 0. Let V( x, y, z) be the square normed gradient vector
field:
∇ f ( x, y, z)
V( x, y, z) = . (26-45)
| ∇ f ( x, y, z) |2
If we let every point p on the level surface Kc ( f ) flow for the time t0 with the flow
curve of V( x, y, z) that starts at p, then the entire level surface will flow into the
level surface Kc+t0 ( f ).

A similar result applies to gradient vector fields of smooth functions f ( x, y) of two

variables and their corresponding level curves in the plane.

Proof

We only have to show that if we start (at time t = 0) at a point p where f ( p) = c then we
end up with the flow curve r(t) of the vector field V( x, y, z) after the time t0 at a point r(t0 )
where f ( x, y, z) has the value f (r(t0 )) = c + t0 .

We use the chain rule for function value incrementation along a flow curve, see eNote 19:
d
f (r(t)) = ∇ f (r(t)) · r′ (t) , (26-46)
dt
and since r(t) is a flow curve of V( x, y, z) we know that r′ (t) = V(r(t)), which substituted
into (26-46) gives:
d
f (r(t)) = ∇ f (r(t)) · V(r(t))
dt
∇ f (r(t))
= ∇ f (r(t)) ·
| ∇ f (r(t)) |2 (26-47)
∇ f (r(t)) · ∇ f (r(t))
=
| ∇ f (r(t)) |2
=1 .
From this we get the result we wanted directly:
Z t0
d
f (r(t0 )) = c + f (r(t)) dt
0 dt
Z t0
(26-48)
= c+ 1 dt
0
= c + t0 .
eNote 26 26.6 FLOWS OF CURVES AND SURFACES 23

■
eNote 26 26.7 SUMMARY 24

26.7 Summary

We have in this eNote established the first concepts and methods for the analysis of
vector fields in the plane and 3D space.

• Some but not all vector fields are gradient vector fields of functions f ( x, y, z) (here
of three variables):
 
∇ f ( x, y, z) = f x′ ( x, y, z) , f y′ ( x, y, z) , f z′ ( x, y, z) , ( x, y, z) ∈ R3 . (26-49)

• Every vector field V( x, y, z) of the first degree can be written and stated by use of
a system matrix A and a constant vector b:
       
V1 ( x, y, z) a11 a12 a13 x b1
 V2 ( x, y, z)  =  a21 a22 a23  ·  y  +  b2  . (26-50)
V3 ( x, y, z) a31 a32 a33 z b3

• To understand the geometry of a given vector field it is important to be able to

determine the flow curves of the vector field – i.e. the parametrized curves r(t) =
( x (t), y(t), z(t)), t ∈ [ a, b], which at all curve-points have the given vector field as
the tangent vector. If the vector field is given by V( x, y, z) = (V1 ( x, y, z), V2 ( x, y, z), V3 ( x, y, z))
then the flow curve equation is:
 ′   
x (t) V1 ( x (t), y(t), z(t))
 y′ (t)  = (V( x (t), y(t), z(t)))⊤ =  V2 ( x (t), y(t), z(t))  . (26-51)
′
z (t) V3 ( x (t), y(t), z(t))

• The divergence of a vector field V( x, y, z) we have defined as the function that at

an arbitrary point ( x0 , y0 , z0 ) has the value:
∂V1 ∂V ∂V
Div(V)( x0 , y0 , z0 ) = ( x0 , y0 , z0 ) + 2 ( x0 , y0 , z0 ) + 3 ( x0 , y0 , z0 ) , (26-52)
∂x ∂y ∂z
and we have indicated through very simple examples that the divergence is a local
measure for how much the vector field spreads or squishes a given set of particles
flowing with the vector field, that is, follows the flow curves of the vector field.
• The curl of a vector field we have defined as the following vector field
∂V3 ∂V2
Curl(V)( x0 , y0 , z0 ) = ( ( x0 , y0 , z0 ) − ( x0 , y0 , z0 ) ,
∂y ∂z
∂V1 ∂V3
( x0 , y0 , z0 ) − ( x0 , y0 , z0 ) , (26-53)
∂z ∂x
∂V2 ∂V1
( x0 , y0 , z0 ) − ( x0 , y0 , z0 ) ) ,
∂x ∂y
eNote 26 26.7 SUMMARY 25

and we have indicated through very simple examples that the curl is a local mea-
sure for how much the vector field rotates a given set of particles that flow with
the vector field, i.e. follows the flow curves of the vector field.