Sie sind auf Seite 1von 3

Multivariable Fundamental Theorems of Calculus

Towards the end of this course, we have seen several theorems that look like the fundamental theorem of calculus, but involving crazy vector derivatives and integrals. Since we went through each one so quickly, it might be helpful to take a step back and line them up next to each other. Such is the purpose of this note. By the way, for a referesher on the various multivariable integrals, please refer to the handout posted online at
The Pattern.

httpXGGmthFhrvrdFeduGjyGintegrlsFpdf
F,
some type of derivative

We always have a scalar function or vector eld oriented spatial object

dF ,

and an

M.

Then the multivariable fundamental theorem of calculus says that

dF =
M
It turns out that not all types of derivative
pairs, and this is how we list them below.

F.
boundary(M )
work with all sorts of object

M:

they come in

The Gradient and Curves.

Let Let

be a scalar function in the plane or space, and take for

the gradient operator.

be an oriented curve, going from

to

Q;

we think of its boundary as the oriented

sum of points

Q + P . f
is a vector eld, so we can take its line integral; the function

Notice that the gradient of

itself can be evaluated at the points

and

Q.

The MFTC here is the same as the fundamental theorem of line integrals:

f dr = f (Q) f (P ).
C
It might be worth pointing out what happens when space. Then

is a curve inside

1-dimensional

is essentially an oriented interval

[a, b]

(where we don't necessarily assume

a < b),

and the MFTC here is just the old-fashioned fundamental theorem of calculus:

f (x) dx = f (b) f (a).


a

The Divergence and Volumes.

d the divergence operator. Let D be a S be its boundary: S is a closed surface, oriented outwards. Notice that the divergence of F is a scalar function, so we can take its volume integral over D ; the vector eld F has a ux integral over the surface S .
Let be a vector eld in space, and take for domain (a blob) in space, and let The MFTC here is the same as Gauss' theorem or the divergence theorem:

( F) dV =
D S

F da.

The Curl and Surfaces.

This example is a little more complicated than the others, because it falls into two parts.
In the plane. Let

be a vector eld in the plane, and take for

the curl operator.

Let

be a region (at blob) in the plane, and

counterclockwise (or, right-handedly) around Notice that the curl of vector eld

C R.

its boundary:

is a closed curve oriented

F is a

scalar function, so we can take its area integral over

R;

the

has a line integral over the curve

C. F dr.
C

The MFTC here is the same as Green's theorem:

( F) dA =
R
In space. Let

be a vector eld in space, and take for in space, and

the curl operator. Let

be

an oriented surface (oppy pancake?) oriented right-handedly around Notice that the curl of vector eld

its boundary:

is a closed curve

S. S;
the

is a vector eld, so we can take its ux integral over

has a line integral over the curve

C. F dr.
c

The MFTC here is the same as Stokes' theorem:

( F) da =
S
Relationships.

Let us sort these things out by dimensionpoints, curves, surfaces, volumesand see where everything else nds itself. Here is what happens inside space: points


FTLI

boundary

curves


Stokes

boundary

surfaces


Gauss

boundary

volumes

f (Pi )
scalar functions

F dr

F da

f dV

gradient

vector elds

curl

vector elds

divergence

scalar functions

The version inside the plane is a little simpler: points


FTLI

boundary

curves


Green

boundary

areas

f (Pi )
scalar functions

F dr

f dA

gradient

vector elds

curl

scalar functions

And the version inside the real line is the easiest: points


FTC

boundary

intervals

f (xi )
scalar functions

b a

f dx

derivative

scalar functions

We nish with some explanations, plus some assorted reections/questions: (1) On the bottom are things we can try to integrate. On the top are things we can try to intgrate them over. Between them is the correct integral of the bottom guy over the top guy. In particular, we use vector integrals precisely on vector elds, and scalar integrals (or point sums) on scalar functions. (2) After stating the MFTC at the beginning of this note, we paired things o: the gradient with curves, etc. Find each pairing in the diagram. In the center is the appropriate MFTC. Below it is the relevant derivative. Above it is the boundary from the relevant object. On the left and right are two types of integrals that are equal. It goes like this:

boundary(M ) F
boundary(M )

boundary

M dF
M

=
d

F
(3) We used the for the integral over

dF
This is because, yes,

boundary(M ).

boundary(M )

is always closed. (E.g., the boundary of a surface is a closed curve, and the boundary of a blob of volume is a closed surface.) The fact that

boundary(M )

has no boundary itself, so that

(4) It is a curious fact that both

boundary(M ) is closed means that boundary(boundary(M )) = . curl(grad f ) and div(curl F) are zero. (These formulas

appear in the section on curl and divergence in your text. Like we mentioned in class, this boils down to the fact that taking partial derivatives commute with one another:

fxy = fyx .)
correct

In particular, this is why gradients are irrotational (curl-free) and curls are

incompressible (divergence-free). Here is another way of saying all this: when using the

ds

in the diagram, we always have

d(dF ) = 0

or

0.

(5) The similarity between items (3) and (4) is not a coincidence: they are symmetric opposites, looking at each other across the mirror of the MFTC. This deep relationship is the beginning of a long and beautful story in modern mathematics called the de Rham theorem (which is not covered on your exam, thankfully). (6) Other things, however, are just lucky coincidences. In 1D, the concepts of a scalar function and a vector eld coalesceand the 1D diagram has no room for vector elds in the middle anyway! Also, the odd discrepancy that in 2D the curl is a scalar function, while in 3D the curl is a vector eld, seems to fall right into place. (7) Challenge question: What happens in 0D? How about 4D or 5D? If we count the numbers of components of the scalar/vector-valued functions in the diagrams, and turn the numbers upside down, we get the following arrangement, which happens to live inside Pascal's triangle. Is this just a coincidence?

1 1
Us:

1 2 1 3 1
Pascal:

1 1 1 1
. . .

1 2 1 3 6 4
. . . . . .

1 1 3

3 4
. . .

1 1

(8) Another: In 2D, the divergence operator is an embarassment of riches. There is a 2D divergence theorem (it's in your text). Can it be t into the diagrams?

Das könnte Ihnen auch gefallen