¡@

Double Integrals and Iterated Integrals

Download Maple Worksheet

First let's recall the basic facts about definite integrals of one-variable functions. If f(x)  is continuous for all x  on [a, b], the definite integral of f on [a, b] is given by the limit of Riemann sums:

int(f(x),x = a .. b) = limit(sum(f(c[i])*Delta*x,i = 1 .. n),n = infinity)

where Delta*x = (b-a)/n and c[i]is any point in  [a+(i-1)*Delta*x, a+i*Delta*x]  for all i = 1, ..., n.

If 0 <= f(x)  on [a, b], int(f(x),x = a .. b)  represents the area of the region bounded by the graph of   f, x = a, x = b  and the  x-axis.

[Maple Plot]

 

In a similar manner, we consider a function f of two variables continuously defined on a closed rectangle

[Maple OLE 2.0 Object]

and we first suppose that 0 <= f(x,y). The graph of f  is a surface with equation z = f(x,y). Let S  be the solid that lies above R  = [Maple OLE 2.0 Object]  and under the graph of f, that is,

[Maple OLE 2.0 Object]


What is the the volume of
S?

If we take f(x,y) = 2-x^2-y^2  on R = [0, 1] x [0, 1], here is the graph of S .

[Maple Plot]

Dividing the interval [a, b] into n subintervals [x[i-1], x[i]] of equal length Delta*x = (b-a)/n  and dividing the interval [c, d] into msubintervals [y[i-1], y[i]] of equal length Delta*y = (d-c)/m, We divide the rectangle R  into subrectangles

R[i*j] = [x[i-1], x[i]]x [y[j-1], y[j]]

of area Delta*A = Delta*x*Delta*y , where [Maple OLE 2.0 Object] , [Maple OLE 2.0 Object].  If we pick a sample point (x[i*j], y[i*j]) in each R[ij], then we can approximate S by "thin" rectangular columns.

The graph below shows the case where (x[i*j], y[i*j]) is taken to be such that f(x[i*j],y[i*j])  is the minimum value of f over R[i*j] :

[Maple Plot]

While the graph below shows the case where (x[i*j], y[i*j]) is taken to be such that f(x[i*j],y[i*j])  is the maximum value of f over R[i*j]  :

[Maple Plot]

Notice that the sum of the volume of all these "thin" rectangular columns equals to

[Maple OLE 2.0 Object]  -----  (1)

¡@

Note that the sum in (1) equals

 sum(sum(f(x[i*j],y[i*j])*Delta*y,j = 1 .. m)*Delta*x,i = 1 .. n) = sum(sum(f(x[i*j],y[i*j])*Delta*x,i = 1 .. n)*Delta*y,j = 1 .. m)

As nand m go larger, the sum in (1) with f(x[i*j],y[i*j]) to be the maximum value of f  over R[i*j] gets smaller and the sum in (1) with f(x[i*j],y[i*j]) to be the minimum value of f over R[i*j] gets larger, as shown in the animation below :

[Maple Plot]

 

[Maple Plot]

 

So we would expect the they eventually would reach the same value which should be the volume of S. In this case, no matter how we choose the sample point (x[i*j], y[i*j]), the sum in (1) would still approach to volume of S. The animation below shows the case that (x[i*j], y[i*j]) is taken to be  ((x[i-1]+x[i])/2, (y[j-1]+y[j])/2) :

[Maple Plot]

  3DVIEW

So the volume of S equals to

limit(sum(sum(f(x[i*j],y[i*j])*Delta*A,j = 1 .. m),i = 1 .. n),(m, n) = infinity) = limit(sum(sum(f(x[i*j],y[i*j])*Delta*A,i = 1 .. n),j = 1 .. m),(m, n) = infinity)

 

Definition   The double integral  of   f over the rectangle R is

[Maple OLE 2.0 Object]

if the limit exists. The sum

[Maple OLE 2.0 Object]  

is called a double Riemann sum .

When  f   is a nonnegative function, the double integral [Maple OLE 2.0 Object]  can be interpreted as the volume of the solid under the surface z = f(x,y)  and above the region R  in the xy -plane. If f   is not a nonnegative function, the double integral [Maple OLE 2.0 Object]  can be can be interpreted as the "signed volume" (that means we take the volume of the solid above xy-plane to be positive and the volume of the solid under xy-plane to be negative) of the solid between the surface z = f(x,y)  and in the xy -plane over the region R .

Properties of Double Integrals

1. [Maple OLE 2.0 Object] .

2. [Maple OLE 2.0 Object] , where c  is a constant.

3. If f(x,y) <= g(x,y)  for all ( x, y ) in R, then [Maple OLE 2.0 Object] .

How to compute the double integrals ?

Suppose that f  is continuous over R = [a, b] x [c, d] (so for each xin [a, b], g(y) = f(x,y) is continuous, for all y  in [c, d].). For each x  in [a, b], we can define

A(x) = int(f(x,y),y = c .. d)

Note that A(x) is continuous on [a, b].

If we now integrate the function A with respect to x from x = a to x = b, we get

int(A(x),x = a .. b) = int([int(f(x,y),y = c .. d)],x = a .. b)

The integral above is called an iterated integral . Usually, the brackets are omitted. Thus,

int(int(f(x,y),y = c .. d),x = a .. b) = int([int(f(x,y),y = c .. d)],x = a .. b)

Similarly, the iterated integral

int(int(f(x,y),x = a .. b),y = c .. d) = int([int(f(x,y),x = a .. b)],y = c .. d)

Let A[1](x) = int(f(x,y),y = c .. d) and A[2](y) = int(f(x,y),x = a .. b).

Recall that if f  is continuous on R , the volume of S  can be approximate by

sum(sum(f(x[i],y[j])*Delta*x,i = 1 .. n)*Delta*y,j = 1 .. m).

[Maple Plot]

 

Fixed y[j], sum(f(x[i],y[j])*Delta*x,i = 1 .. n)*Delta*y, the volume of each thin slice, is approximately equal to

A[2](y[j])*Delta*y = [int(f(x,y[j]),x = a .. b)]*Delta*y.

[Maple Plot]

So sum(sum(f(x[i],y[j])*Delta*x,i = 1 .. n)*Delta*y,j = 1 .. m)  is approximately equal to

sum([int(f(x,y[j]),x = a .. b)]*Delta*y,j = 1 .. m) = sum(A[2](y[j])*Delta*y,j = 1 .. m),

as j  goes to infinity,  the right hand side of the above equation approaches

int(A[2](y),y = c .. d) = int([int(f(x,y),x = a .. b)],y = c .. d) .

So we have

[Maple OLE 2.0 Object]  = int(int(f(x,y),x = a .. b),y = c .. d)

Similarly,  the volume of S  can be approximate by sum(sum(f(x[i*j],y[i*j])*Delta*y,j = 1 .. m)*Delta*x,i = 1 .. n) .

[Maple Plot]

 

For fixed x[i] , the volume of each thin slice is approximately equal to

  A[1](x[i])*Delta*x = [int(f(x[i],y),y = c .. d)]*Delta*x .

[Maple Plot]

So sum(sum(f(x[i*j],y[i*j])*Delta*y,j = 1 .. m)*Delta*x,i = 1 .. n)  is approximately equal to

  sum([int(f(x[i],y),y = c .. d)]*Delta*x,i = 1 .. n) = sum(A[1](x[i])*Delta*x,i = 1 .. n) ,

a s i  goes to infinity, the right hand side of the above equation approaches

int(A[1](x),x = a .. b) = int([int(f(x,y),y = c .. d)],x = a .. b)

So we get

[Maple OLE 2.0 Object]  = int(int(f(x,y),y = c .. d),x = a .. b)

¡@

¡@

Fubini's Theorem

If   f  is continuous on the rectangle R = [a, b] x  [c, d], then

[Maple OLE 2.0 Object] = int(int(f(x,y),x = a .. b),y = c .. d)  = int(int(f(x,y),y = c .. d),x = a .. b)

More generally, this is true if we assume that f  is bounded on R , f   is discontinuous only on a finite number of smooth curves, and the iterated integrals exist.

¡@

¡@

Example:   Evaluate the double integral [Maple OLE 2.0 Object] , where R  = [0, 2] x  [1, 2].

[Maple Plot]
3DVIEW

Note that the surface y = x-3*y^2  lies below the xy -plane, so the double integral [Maple OLE 2.0 Object] should be negative.

¡@

Solution : Fubini's theorem gives

[Maple OLE 2.0 Object]  = int(int(x-3*y^2,y = 1 .. 2),x = 0 .. 2) = int(x-7,x = 0 .. 2)  =   -12

¡@

Example:  Evaluate the double integral [Maple OLE 2.0 Object] , where R  = [1, 2] x  [0, pi].

[Maple Plot]
3DVIEW

Solution : We integrate with respect to x first,

[Maple OLE 2.0 Object] = int(int(y*sin(x*y),x = 1 .. 2),y = 0 .. Pi) = int(-cos(2*y)+cos(y),y = 0 .. Pi)  = 0

¡@

Remarks :

1. If we integrate with respect to y  first, the computation will be much more complicated.

2. [Maple OLE 2.0 Object] = 0 says that the volume of the solid under the surface z = y*sin(x*y) , above the xy-plane over R equals to of the solid above the surface z = y*sin(x*y) , below the xy-plane over R are equal.  

[Maple Plot]
3DVIEW

Example:  Find the volume of the solid S that is bounded by z = x^2*y, the planes x = 2  and y = 2, and the three coordinate planes.

[Maple Plot]
3DVIEW

Solution : Note that is x^2*y  positive on [ 0, 2 ] x  [ 0, 2 ], so the volume equals to

int(int(x^2*y,x = 0 .. 2),y = 0 .. 2) = int([y*int(x^2,x = 0 .. 2)],y = 0 .. 2)  = [int(x^2,x = 0 .. 2)]*[int(y,y = 0 .. 2)] = 16/3

Note that in the special case where f(x,y) = g(x)*h(y)  and R = [a, b]x [c, d], the Fubini's theorem gives

[Maple OLE 2.0 Object]   =    int(int(g(x)*h(y),x = a .. b),y = c .. d)  

                         =   int([int(g(x)*h(y),x = a .. b)],y = c .. d)

                        =   int(h(y)*[int(g(x),x = a .. b)],y = c .. d)

                           =   [int(g(x),x = a .. b)]*[int(h(y),y = c .. d)]


Download Maple Worksheet