Sie sind auf Seite 1von 9

# 3.

## Dierentiation of vector elds

Why are there two notions of dierentiation of vector elds (div and curl)? What do they represent physically? Do they obey the rules of calculus I know from 1D? Which second derivatives make sense?

## 3.1 The two kinds of vector dierential

There are two key notions of dierentiation of a vector eld v (x, y, z ). They are the so-called divergence div v and curl curl v (sometimes called rot v ). Consider vector elds in 2D (think of the wind map on the weather forecast) Loosely speaking: Divergence represents the extent to which a particle being carried by the vector eld (the wind) is being pulled apart.
1

y 0.5
1 0.5 0.5 0.5 1

Curl represents the amount to which a particle being carried by the vector eld is being rotated

24

y 0.5
1 0.5 0 0.5 0.5 1

We will return to precise denitions of pulled apart and rotated in Chapter 6, after we have dealt with integration. Calculating div v and curl v . Divergence: scalar!

## div v = v = , , (v1, v2, v3) x y z = Curl: vector! i curl v = v =

v1 + v2 + v3 x y z j k
y z

v1 v2 v3 v3 v2 v1 v3 v2 v1 = i+ j+ k y z z x x y

Note that the direction of curl v is the axis about which the vector eld is being rotated (in an anti-clockwise sense). Worked example 3.1 Calculate div v and curl v for the vector elds 1. v = 4xy i + yz j + xk
25

2. v = xi + y j + z k (=position vector r)

## 3.2 Applications of divergence and curl

1. Rigid body mechanics The velocity of a rigid body B rotating about a xed axis can be described by a vector , where = | | is the angular rotation speed and the direction of is the axis of (anti-clockwise) rotation. The velocity eld at distance r from the axis can be shown to be given by v = r Then we can show that div v = 0, curl v = 2 (2.1)

that is, curl of the velocity eld has the direction of the axis of rotation, and magnitude that is twice the angular speed of rotation Worked example 3.2 Taking = k, show explicitly the formulae (2.1) 2. Fluid mechanics Let v be the velocity of a uid ow. Consider the following situations Linear simple shear ow between two plates in relative motion V y v = , 0, 0 d
x

## V curl v = (0, 0, ), d div v = 0

This is a rotational (curl v = 0), incompressible (div v = 0) ow. The rotation is perpendicular to the (x, y )-plane.
26

Vortex
x

x y , v = 2 x + y 2 x2 + y 2 curl v = 0,

div v = 0. This is an irrotational, incompressible ow. 3. Incompressibility Denition div v =0 A vector eld is said to be incompressible if

This has a natural interpretation in uid mechanics, where the equation of continuity states that the uid density (r ) (a scalar eld) and the uid velocity v (r) are linked by the continuity equation + div (v ) = 0. t So that if the uid has constant density (e.g. water, to good approximation, but not air) we have = 0 =const., and hence div (0v ) = 0 div (v ) = 0 div (v ) = 0.

Incompressible vector elds are also called solenoidal 4. Irrotational ow, conservative forces Denition A uid ow that whose velocity eld v is curl free, curl v = 0 is called irrotational. Denition A force eld F that satises curl F =0 is said to be conservative. More generally it can be shown that curl v = 0 if and only if v = grad , for some scalar eld
27

is called the scalar potential of a conservative vector eld. Proof. First, we show that v = grad curl v = 0 This is by direct calculation: v= i+ j+ k x y z i so, curl v =
x x

j k
y y

z z

2 2 2 2 2 2 i + j + k = zy yz xz zx yx xy = (0i + 0j + 0k) = 0.

Second, we show that curl v = 0 v = for some scalar eld (r). Now curl v = 0 v3 v2 = 0 y z v1 v3 = 0 z x v2 v1 = 0 x y
Now, taking x of (2.2) and Setting these equal gives y

2 v3 xy .

## of (2.3) we get two expressions for

2 v2 2 v1 3 = := zx zy xyz
28

(2.5)

for some scalar function (x, y, z ). This can only be true if + b1y + c1 z + d1 and v2 = + b2 x + c 2 z + d 2 . v1 = x y For some functions b1(x), c1 (x), d1(x) and b2(y ), c2(y ), d2(y ) (where = dierentiate). But, by redening to be b1y c1 z d1x b2xy c2 yz d2y, we can without loss of generality choose b1,2 = c1,2 = d1,2 = 0. Similarly, from (2.3) and (2.4) we nd 2 v2 3 2 v3 = = xy xz xyz so that, without loss of generality , v2 = v3 = z y v = (v 1 , v 2 , v 3 ) = , , = 2 x y z Worked example 3.3 Find scalar functions whose gradients are (i) (2xy + z 3)i + x2j + 3xz 2 k. (ii) 2xi + 4y j + 8z k.

## 3.3 Rules of vector dierentiation

We have dened three kinds of derivative involving the operator . grad = , div = v , curl = v The good news is that you can apply all the usual formulae for dierentiation with d/dx replaced by provided you are careful. This is because grad and curl are vectors, whereas div is a scalar. Also div and curl apply to vector elds, whereas grad applies to scalar elds. Let u(r) and v (r) be vector elds, f (r ) and g (r) be scalar elds and and be constants:
29

1. dierentiation is linear grad (f + g ) = f + g = grad f + grad g div ( u + v ) = u + v = div u + div v curl ( u + v ) = u + v = curl u + curl v 2. product rules: multiplication by scalars grad (f g ) = (f g ) = f g + g f = f grad g + g grad f div (f v ) = (f v ) = f v + f v + = f div v + grad f v curl (f v ) = (f v ) = f v + f v = f curl v + grad f v 3. a vector product rule div (u v ) = v curl u curl v u Sketch proof of 3: From triple scalar product

(u v ) = ( u) v and (u v ) = ( v ) u. But in order for the dierential operator to see both u and v , we must add these two forms, i.e. (u v ) = ( u) v ( v ) u which gives the required result. 2 Alternatively can prove each of the above using co-ordinates (not the operator) - see example sheet. Worked example 3.4 Let v = (3xyz 2 , 2xy 3, x2yz ), = 3x2 yz

Find (i) div v , (ii) v grad , and hence (iii) div (v ) at the point (1, 1, 1).
30

## 3.4 Second derivatives

Having dened grad , div and curl in terms of applying to a vector or scalar eld. What do we get if we apply twice? The following make no sense: grad grad div div v grad curl v curl div v = = = = ( v ) ( v ) ( v )

The next two are identically zero curl grad = () 0, div curl v = ( v ) 0. (2.6) (2.7)

Proof. (2.6) was proved before when dening conservative vector elds (curl v = 0 v = ). So consider (2.7). Worked example 3.5 Show that div curl v 0 aside: In fact, it can be shown that any incompressible (solenoidal) vector eld u (i.e. one for which div u = 0) can be written in the form u = curl v . Such a v is called a vector potential. So we are left with only three 2nd derivatives that make sense div grad := 2 (=Laplacian del-squared) div grad = = i, j , k i, j , k x y z x y z 2 2 2 = + + scalar! x2 y 2 z 2 grad div v = ( v )
=

vector!
2 2

v2 v3 v1 i + ... + + 2 x xy xz
31

curl curl v = ( v )

vector!

( v ) = ( v ) ( )v 2 2 2 2 v3 v2 v1 v1 i + ... + = xy y 2 z 2 xz note = 2v := (2v 1, 2v 2, 2v 3) in Cartesian co-ordinates. Worked example 3.6 Verify that ( v ) = ( v ) ( )v where v = 2xz 2i yz j + 3xz 3k. The Laplacian 2 is an important dierential operator in its own right. It can be shown that any vector eld that is both irrotational and incompressible satises v = where 2 = 0. (2.8) (see example sheet). The equation (2.8) is called Laplaces equation and is one of the most fundamental Partial Dierential Equations in engineering science (see later in EMa II). It occurs in electrostatics, incompressible uid dynamics (water, not air), and in stress analysis.

## 3.5 Summary (Chapters 2 and 3)

i+ j+ k x y z ) (directional derivative D f = a v1 v2 v3 div v = v = + + x y z grad = =

i j k curl v = v =
x y z

v1 v2 v3 v = 0 irrotational (=conservative), v = 0 incompressible Only certain 2nd derivatives make sense, the most important of which is 2, the Laplacian.
32