Vous êtes sur la page 1sur 14

Vector Calculus Applications

1. Introduction

The divergence and Stokes theorems (and their related results) supply fundamental tools
which can be used to derive equations which can be used to model a number of physical sit-
uations. Essentially, these theorems provide a mathematical language with which to express
physical laws such as conservation of mass, momentum and energy. The resulting equations
are some of the most fundamental and useful in engineering and applied science.
In the following sections the derivation of some of these equations will be outlined. The goal
is to show how vector calculus is used in applications. Generally speaking, the equations are
derived by rst using a conservation law in integral form, and then converting the integral
form to a dierential equation form using the divergence theorem, Stokes theorem, and
vector identities. The dierential equation forms tend to be easier to work with, particularly
if one is interested in solving such equations, either analytically or numerically.

2. The Heat Equation

Consider a solid material occupying a region of space V . The region has a boundary surface,
which we shall designate as S. Suppose the solid has a density and a heat capacity c.
j + z k is the
If the temperature of the solid at any point in V is T (r, t), where r = x + y
position vector (so that T depends upon x, y, z and t), then the total heat eergy contained
in the solid is 
cT dV .
V

Heat energy can only get in or out of the region V by owing across the boundary S. Lets
suppose that the heat ux vector is called q. This vector measures rate of energy ow past
a point per unit area. Thus, if we have a small element of surface area dS with outward
unit normal vector n , the rate of energy ow outward through this element of surface area is
q n
dS. Integrating over the entire surface, the total rate of energy ow (i.e., the ux) out
of the region is therefore 
 q n
dS .
S
This is shown schematically in Fig. 1.
Conservation of energy then requires that the rate at which the energy within the region
V and the rate at which energy crosses the boundary of region must balance, i.e.,
d  
cT dV +  q n
dS = 0 .
dt V S

W.
c L. Kath, 2004. E-mail: kath@northwestern.edu Version 1.0, 4 December 2004

1
Fall 2004 Vector calculus applications Calculus 215

z q

n
V S

Figure 1: Schematic diagram indicating the region V , the boundary surface S, the normal
to the surface n
, and the heat ux vector q.

The signs of the terms must appear in this manner, because if the ux of energy out of
the region is positive, then the total energy inside the region must be decreasing, hence its
derivative will be negative.

To simplify this, we rst take the time derivative inside the volume integral:
d  

cT dV = (cT ) dV .
dt V V t

We can do this because the region V over which the integral is being taken is independent of
time; V is the same region for all time. Thus, at any xed time, integration over V just gives
a single number, i.e., the energy inside V at that time. The energy will change with time
because T (and possibly and c) changes with time. Note that the time derivative must
be turned into a partial derivative when it is moved inside the volume integral, however,
because the functions being integrated depend upon both position and time. Similarly, we
apply the divergence theorem to the ux integral:
 
 q n
dS =  q dV .

S V

The result is  


 q dV ,
(cT ) dV +
V t V
or, equivalently,
  
 q dV = 0 .
(cT ) +
V t

2
Fall 2004 Vector calculus applications Calculus 215

It would be nice to say at this point that the quantity in brackets is zero; this would allow
us to get rid of the integral. This would be true, but the right argument is needed, because
an integral can be zero either because the integrand is zero or because it is equally positive
and negative, and when integrated there is perfect cancellation.

In the present case, however, there is one more fact: the integral must be zero for arbitrary
volumes V . This forces the integrand to be zero everywhere. The proof is by contradiction.
First, assume that the integrand is not zero everywhere. Pick a point at which it is positive,
and then choose a region V that is entirely within the region where its positive. Since we
are integrating a quantity that is entirely positive, the resulting integral must be zero. This
cant be true, however, since we know the integral must be zero. Therefore, we must have
 q = 0 .
(cT ) +
t
This is the partial dierential equation form of the fundamental principle of energy conser-
vation for heat transfer.

More can be done if one has an explicit result for the heat ux q in terms of the temper-
ature. In many cases, observations show that the heat ux is given by Ficks law,
 ,
q = k T

i.e., the heat transfer is proportional to the negative of the temperature gradient. Thus, heat
tends to ow from hotter regions to colder regions. While this is true in most cases, it is not
always true; in particular, if phase transitions are possible it can be violated.

If we use Ficks law in the equation for conservation of heat energy, it becomes
 
(cT ) = 
 k T .
t
Finally, if , c and k are all constant, this equation simplies to the heat equation

T
= 2 T , (1)
t

where = k/c is the thermal diusivity.

A special case of this equation results if the object is in thermal equilibrium, i.e., the
temperature doesnt change with time:

2 T = 0 . (2)

This is known as Laplaces equation.

3
Fall 2004 Vector calculus applications Calculus 215

V S

Figure 2: Schematic diagram indicating the region V , the boundary surface S, the normal
to the surface n
, the uid velocity vector eld v , and the particle paths (dashed lines).

3. Fluid mechanics

3.1. Conservation of mass

Conservation laws are also the basis for the equations of uid mechanics. Let be the uid
density and v be the uid velocity. Then the mass in a region V is given by

dV ,
V

and the rate at which mass leaves the region through its boundary S is

 v n
dS ;
S

see Fig. 2 for a diagram of the situation. Here v is the mass ux vector; this has units of
kg/m3 m/sec = kg/m2 sec, i.e., mass per unit area per second.

The rate of change of the mass in V must be balanced by the rate at which mass leaves
through S,
d  
dV +  v n
dS = 0 .
dt V S
Pulling the time derivative inside the V integral, and converting the surface integral into a
volume integral using the divergence theorem, we have
  

+ (v ) dV = 0 .
V t

4
Fall 2004 Vector calculus applications Calculus 215

As before, because the region V is arbitrary, we must have the terms between the brackets
be identically zero, which gives


+ (v ) = 0 . (3)
t

This is the partial dierential equation form of conservation of mass.


 (v ) =
Equation 3 can be simplied somewhat by using the vector identity  v +
 v = v
 +
 v , which gives

 +
 v = 0 .
+ v (4)
t

The combination /t + v  occurs so much in uid mechanics that it is given a special


notation,
 D .
+ v
t Dt
This is called the material or substantial derivative, and physically it represents taking the
time derivative following the uid, i.e., taking the time derivative of something transported
along by the uid. One can see this by considering (x, t) not at a xed position x, but
instead (x(t), t), where dx/dt = v ; because the position x moves at the uid velocity v , this
is the density of a specic uid element. Taking the time derivative using the chain rule,
d  dx + = + dx
 = + v
 = D .
(x, t) =
dt dt t t dt t Dt
In any event, with this notation we have

D  v = 0 .
+ (5)
Dt

This has a nice physical interpretation: if  v > 0, so that the uid velocity is locally
 v < 0, so
expanding, then D/Dt < 0, i.e., the local density is decreasing. Similarly, if
the uid velocity is locally compressing, then D/Dt > 0 and the density is increasing.

We can also consider the case of an incompressible uid : if is constant, then D/Dt = 0
and therefore

 v = 0 .
(6)

Water is an example of a uid that is almost incompressible.

5
Fall 2004 Vector calculus applications Calculus 215

3.2. Conservation of momentum

The next main conservation law is momentum. Actually, of course, the law is that the time
rate of change of momentum is equal to force, d(momentum)/dt = force. To apply it to a
uid, we need to identify all of the sources of momentum and all of the forces involved.

First, the momentum density (per unit volume) of the uid is v . Therefore, the momen-
tum in a volume V of uid is 
v dV .
V
Of course, what we really need is the time rate of change of this momentum,
d 
v dV .
dt V

Another way that the momentum in V can change is if the uid transports momentum out
of the region across the boundary S. The ux of momentum out of V is then

 v (v n
) dS .
S

This is similar to the previous case of conservation of mass, where we had mass density times
the normal component of the velocity, giving mass ux. In this case, we have momentum
density times the normal component of the velocity, giving momentum ux.

We also need the forces acting on V . First of all, we have the net pressure force, which is

 p n
dS ,
S

since the pressure acts normally to each element of the surface (with an inward force when
the pressure is positive, hence the minus sign). Similarly, if we have any body forces, such as
gravity, we have to include them. Its traditional to use f as the body force per unit mass,
so that 
f dV
V
is the total body force.

Putting everything together, we have conservation of momentum:

d    
v dV +  v (v n
) dS =  p n
dS + f dV . (7)
dt V S S V

Note that other forces have been neglected, such as forces due to viscosity, where a bit of
uid exerts a drag force on nearby uid elements.

6
Fall 2004 Vector calculus applications Calculus 215

The next step should be to convert the surface integrals to volume integrals using the
divergence theorem. Unfortunately, the form of the surface integrals is not right for the
divergence theorem, because the integrands of the surface integrals are vectors, not scalars.
Thus, we need to somehow modify the divergence theorem.

Lets look rst at the force due to pressure. Here, we are missing a vector. We can x
this by taking the dot product of the integral (neglecting the sign for the moment) with
a constant vector a:
  
dS =  p a n
a  p n dS =  (p a) n
dS .
S S S

Now we can apply the divergence theorem:


 
 (p a) n
dS =  (p a) dV

S V

Now we use the vector identity  (p a) = p


 a + p
 a. This is equal to p
 a, since
 a = 0, because a is a constant vector. Thus,

    


a  p n
dS =  (p a) n
dS =  (p a) dV =
 a) dV = a
p  dV .
p
S S V V V

Thus, we have    


a  p n
dS  dV = 0 .
p
S V

Because a is an arbitrary vector, this means that the terms in brackets must be zero. There-
fore
 
 pn
dS =  dV .
p (8)
S V

This is an extension of the divergence theorem.

Next, consider the integral for the ux of momentum,



 v (v n
) dS .
S

We cant apply the divergence theorem because of the extra v . Its also not so easy to do
something like what we did for the pressure force, so instead lets expand the extra v into

components, v = 3n=1 in vn :

3 
 v (v n
) dS = in  vn (v n
) dS .
S n=1 S

7
Fall 2004 Vector calculus applications Calculus 215

Now we can apply the divergence theorem to the integral,


3 
3 
in  vn (v n
) dS = in  (vnv ) dV

n=1 S n=1 V

Applying vector identities, we have


 (vnv ) =
 [vn (v )] = v  (v ) = v v
 n (v ) + vn  n + vn
 (v )

The result of the divergence theorem then simplies to


3  
in v v  (v ) dV =
 n + vn  v + v
v   (v ) dV
n=1 V V

Collecting all of the converted terms in Eq. (7) and dropping the integrals because the
region is arbitrary, we then have
 v + v  + f .
 (v ) = p
(v ) + v 
t
This can be further simplied since
v  (v ) + v ,
(v ) = v + = v
t t t t
where for this last part we have used conservation of mass, Eq. (5), to eliminate /t.
Cancelling the terms, this then gives
 
v  + f ,
 v = p
+ v  (9)
t

or, alternatively

Dv  + f .
= p (10)
dt

This is the equation for conservation of momentum, the uid-mechanical equivalent of New-
tons law ma = F . Together with the equation for conservation of mass, they are called the
Euler equations. Note that to complete the equations, the force f and an equation for the
pressure p need to be specied. Sometimes the pressure can be given as a function of the
density; this is called the equation of state. A fully complete set of equations, however, also
has to include one additional equation, that for conservation of energy.

8
Fall 2004 Vector calculus applications Calculus 215

3.3. Irrotational ows and Bernoullis equation

First, use the last vector identity in the Appendix with F = v and G
 = v to write

 v = 1 (
v   v v ) v (
 v ) . (11)
2

Lets also write 


 = v; here
 is called the uid vorticity, and measures the local rotation
rate of the ow. Then, conservation of momentum, Eq. (9), becomes

v 1  
p
+ (v v ) v
 = + f . (12)
t 2
Now, take the curl of this equation. We need another vector identity,
 (v

 ) = (  v )
 ) v (  v v 
 
 + .

Of course,   = 0. We then have for the curl of the equation (remember the curl of a
gradient is zero)


 =
+ v   v (
   v )  p +
  f .
t

This can also be written


D  v (
 v ) 1   f .
 +
 
=  + 2 p
Dt


If  f = 0,
 = 0, which will happen if is a constant or if p is a function of , and if
p

which will happen if f is a conservative force eld, then
D  v (
 v )
 
= 
Dt
The interesting thing about this equation (called the vorticity equation), is that if
 is zero
initially, then it shows that D /Dt = 0, so that the vorticity wont change. Thus,  will
remain zero for all time (or until some source of vorticity is introduced into the uid).
 v = 0. Such a uid ow is said to be irrotational.
 = 0, this means that 
If the vorticity

In this case, we know that there is a scalar potential such that v = . The advantage of
this is that it reduces three unknowns (the components of v ) to one (the potential ).

In the case when the density is constant, conservation of mass becomes  v = 0.


As described earlier, this ow is said to be incompressible. If, in addition, the ow is also

9
Fall 2004 Vector calculus applications Calculus 215


irrotational, so that v = , then the equation of conservation of mass becomes Laplaces
equation,
2 = 0 .
This equation is a lot simpler than the original version of conservation of mass.

Another interesting result is that happens to conservation of momentum in this case.


 in Eq. (12), we have
Substituting v =
 
 1    2  p  = 0,
() + || + + U
t 2

where we have assumed f = U


 . This can then be written
 
 + 1 ||
 2 + p + U = 0.
t 2

For the gradient of a function to be zero it must be constant. Therefore,

1  2 p
+ || + + U = constant . (13)
t 2

This is Bernoullis equation. It can be thought of as an equation similar to conservation of



energy; note that in the case with no body force (U = 0), where the uid speed (i.e., ||)
is small, the pressure will be large, and where the uid speed is large, the pressure will be
small.

4. Maxwells equations

The equations describing the behavior of electric and magnetic elds can also be derived
in a manner similar to that done for uids. In each case, the starting point is a principle
based upon observation. Note that fundamentally, conservation of mass, momentum and
energy are principles that are based upon observation. The situation is much the same for
electromagnetics; the principles are just a little dierent.

Historically, one of the rst results to be obtained was that for the force between charged
objects. In particular, the force between two metal spheres with charges Q1 and Q2 was
observed to be
Q1 Q2 r
F = ,
4 r2
where r is the distance between the centers of the two spheres, r is a unit vector pointing
from Q1 to Q2 when we want the force on Q2 due to the charge on Q1 , and is a constant
(called the permittivity, or the dielectric constant).

10
Fall 2004 Vector calculus applications Calculus 215

If the second object is replaced with a test charge q, then the force on the test charge due
to the charge Q1 is
F = q E,
 where E  = Q1 r .
4 r2
Here E is the electric eld produced by the charge Q1 ; in this case r is a unit vector that
points away from its position.

In the case where more than one charge is present, or there is a distribution of charge with
density (r, t), experimentally one can verify Gausss law:

 E
 n
dS = Q , (14)
S

i.e., the total ux of the electric eld out through the surface S is proportional to the total
charge Q. Furthermore, if one has a distribution of charge with density (r, t), then the
above becomes
 
 E
 n
dS = dV . (15)
S V

Applying the divergence theorem to the ux integral, and noting that the volume V is
arbitrary, this becomes
 
 E
 = . (16)

The vector D  = E is called the electric displacement, or the electric ux density (E


 is still
the electric eld). This is the rst of Maxwells equations.
 except there are no free magnetic
A similar result holds for the magnetic ux density, B,
charges. This means that we have

B
 n
dS = 0 , (17)
S

or equivalently,

 B
 = 0. (18)

This is the second of Maxwells equations.

The last two of Maxwells equations dont follow from the divergence theorem, but rather
from Stokes theorem. First of all, we have Amperes law, which says that

11
Fall 2004 Vector calculus applications Calculus 215

  
B
dr = J n
dS , (19)
C S

where is the magnetic permeability, J is the current density, C is an arbitrary closed


contour and S is an open surface that has C as its boundary. The basic form of this law
was discovered from experiments. Basically, this is the mathematical statement of the well-
 = B/
known fact that currents generate magnetic elds. It is traditional to dene H  to be
the magnetic eld intensity.

To eliminate the integrals in the above, we apply Stokes theorem:


 
 dr =
H  H

n  dS .
C S

We therefore have   
 H

n  J dS = 0 .
S
Since the contour C is arbitrary, this means the surface S must be arbitrary. By the same
argument used for the divergence theorem, the only way we can get zero is for the terms in
the brackets to be zero, i.e.,
 H
 = J .

Here, the current density J is generated by movement of charges. There is also another type
of current that must be added; in reality, an extra term must be added:


 = J + E .
 H (20)
t

This is the dierential equation form of Amperes law, and is the third of Maxwells equations.
The extra term is called the displacement current. It is an eective current that results from
the time-variation of the electric eld. One more thing: in the case of a simple resistive
material, J = E,
 where is the conductivity of the material.

Note that if one takes the divergence of the above equation, one gets

 J +   J + .
 =
0= ( E)
t t
This is an equation stating that charge is conserved.

Finally, we have Faradays law:


 
 dr =
E  n
B dS . (21)
C t S

12
Fall 2004 Vector calculus applications Calculus 215

The line integral of E is called the electromotive force. Essentially, it measures the increase
or decrease in the electric potential (i.e., voltage) when going around the contour, since E is
the gradient of the potential. Faradays law says that the voltage drop around the contour is
proportional to the time rate of change of magnetic ux passing through a surface spanning
the contour. If we now apply Stokes theorem to this situation, we nd

 = B .
 E (22)
t
This is the dierential equation form of Faradays law, and Maxwells fourth equation.

In summary, we have the full set of Maxwells equations:


 D
 = ,  = B ,
 E

t
(23)

 B
 = 0,  = J + E ,
 H

t

 = E
where D  and B
 = H,
 and, in the case of a resistive material, J = E.


As a last application, lets consider Maxwells equations in free space, with no free charges
( = 0 and J = 0). In addition, and are constant. Then Maxwells equations become

 E
 = 0,  = B ,
 E

t

 B
 = 0,  = E .
 B

t
 We do this by taking the curl of Faradays law:
Lets eliminate the magnetic eld B.
  2
 (
 =
 E)  =1 E,
 B
t c2 t2
where we used Amperes law for  B,
 and have written = 1/c2 . The constant c turns
 (
out to be the speed of light. We can also use a vector identity to simplify  E),
 i.e.,

 (
 E)
 = (
  E) .
 2 E

Thus, we get
1 2E
 =
2 E ,
c2 t2
 E
because  = 0. This is known as the wave equation.

13
Fall 2004 Vector calculus applications Calculus 215

To understand the behavior of the wave equations solutions, consider the very special
 = E1 (x, t)i. Then the equation becomes
case when E

2 E1 1 2 E1
= .
x2 c2 t2
It is relatively easy to verify that a solution of this equation is

E1 (x, t) = f (xct) ,

where f () is an arbitrary function. The solution is a wave propagating in the positive x


direction with velocity c. To see this, let xct = and note that if f () has its maximum
at = 0, then the position of this maximum will be at xct = 0, or x = ct. Thus, the
maximum will move in the positive x direction with speed c.

A. Appendix: the main theorems and formulas

Divergence Theorem:  


 F n
dS =  F dV

S V

Stokes Theorem:  
F dr =  F dS

n
C S

Vector Operators & Identities:


   
 F
 F +
=  F ,  F
 F +
=  F ,
   
div curl F =  F
 = 0, 
curl grad =  = 0,
   
 F
 
=  F 2 F ,
     
 F G
 = G  F F
  G
 ,
         
 F G
 =  G
 F
 F G
+ G

 F F
 G
,
         
 F G
  G
= F  +G  F + F
  G
+ G

 F .

14

Vous aimerez peut-être aussi