0/1

P1 Calculus 2
Calculus of multiple variables

4 lectures, MT 2023/24 David Murray

Rev November 16, 2023 Updated by Vicente Grau (vicente.grau@eng.ox.ac.uk)

Course Page More information might be available on Canvas

Introduction

So far in your explorations of the dierential and integral calculus, you have considered

only scalar functions of one scalar variable  y = f (x ), and all that.

But you do not have to look hard to nd scalar quantities which depend on more than

one independent variable. The pressure in an oil eld will depend on latitude, longitude

and depth, and no doubt will change over time. The lectures in Calculus 2.I answer

the questions of how our notions of dierentiation and integration can be extended to

cover multiple variables.

Lecture Content

1. The partial derivative. First and higher partial derivatives. Total and partial dif-

ferentials, and their use in estimating errors. Testing for total (or perfect) dier-

entials. Integrating total dierentials to recover original function.

2. Relationships involving rst order partial derivatives. Function of a function. Com-

posite functions, the Chain Rule and the Chain Rule for Partials. Implicit Func-

tions.

3. Transformations from one set of variables to another. Transformations as old in

terms of new and new in terms of old. Jacobians. Transformations to Plane,

spherical and polar coordinates. What makes a good transformation? Functional

dependence. Shape.

4. Double, triple (and higher) integrals using repeated integration. Transformations,

the use of the Jacobian. Plane, spherical and polar coordinates.

0/2

Tutorial Sheets

The tutorial sheet associated with this course is

 1P1H: Calculus of Multiple Variables

Reading

 James, G. (2001) Modern Engineering Mathematics, Prentice-Hall, 3rd Ed.,

ISBN: 0-13-018319-9 (paperback).

 Stephenson, G. (1973) Mathematical Methods for Science Students, Longman

Scientic & Technical, 2nd Ed., ISBN: 0-582-44416-0 (paperback).

 Kreyszig, E. (1999) Advanced Engineering Mathematics, John Wiley & Sons,

8th Ed., ISBN: 0-471-15496-2 (paperback).

Course WWW Pages

Pdf copies of these notes (in colour), copies of the lecture slides, the tutorial sheets,

corrections, answers to FAQs etc, will be accessible from

More information might be available on Canvas

Only the notes and the tute sheets will appear on Canvas.
Lecture 1
Total and Partial Derivatives and Dierentials

1.1 Reminders: continuity and the derivative for one variable

Suppose we have a value x = a. We can dene a neighbourhood of x -values around

x = a by requiring jx aj <  or, equivalently, (x a)2 < 2. A function y = f (x ) is
said to be continuous at x = a if, for every positive number  (however small), one can

nd a neighbourhood jx aj <  in which jf (x ) f (a)j < . Another way of expressing

this is: limx !a f (x ) = f (a).

You will also recall that the derivative is dened as

f (x + x ) f (x )
 

x f (x ) = xlim :
d

d !0 x (1.1)

f (x ) is dierentiable if this limit exists, and exists independent of how x ! 0.

1.2 Moving to more than one variable

The big changes we will encounter all take place when going from one variable to any

number more than one. So for much of the time we can keep the page uncluttered by

dealing with functions of only n = 2 variables, as in f = f (x; y ), where x and y are

now independent variables.

Fig. 1.1 shows that functions of two variables are conveniently represented graphically

using the Cartesian axes Oxyz . The function representation is a surface, as opposed

to a plane curve for a one variable function. It is of course progressively harder to

represent functions of more than two variables.

1.2.1 Continuity for functions of several variables

A function f = f (x; y ), dened in some region R , is continuous at a point (x; y ) =

(a; b) in R if, for every positive number  (however small), it is possible to nd a positive
1
 1/2 LECTURE 1. TOTAL AND PARTIAL DERIVATIVES AND DIFFERENTIALS

3
20 1

10 0.5 2

f(x,y)
f(x,y)

f(x,y)
0 0
1
−10 −0.5

−20 −1 0
5 5 5
4 4 4
0 2 0 2 0 2
0 0 0
−2 −2 −2
y −5 −4 y −5 −4 y −5 −4
x x x

(a) (b) (c)

Figure 1.1: Surface plots of (a) (x +y )(x j

y ), (b) exp[ (x 2 +y 2 )=10] sin(2x ) cos(4y ), and (c) (x y ) 1=2 . j
(The Matlab code for such plots is given as an Appendix.)

 such that
jf (x; y ) f (a; b)j < ; (1.2)

for all points in the neighbourhood dened by

(x a )2 + ( y b )2 <  2 (1.3)

Or equivalently, as before, lim(x ;y )!(a;b) f (x; y )=f (a; b). Note that for functions of
more variables f (x1 ; x2 ; x3 ; :::) the neighbourhood would be dened by (x1 a ) 2 + ( x2
b)2 + (x3 c )2 + : : : < 2. The two-variable case is illustrated in Figure 1.2.

Figure 1.2: Neighbourhood for continuity for a function of 2 variables

1.2. MOVING TO MORE THAN ONE VARIABLE 1/3

1.2.2 The partial derivative

The extension of the idea of continuity to functions of several variables was direct.

Extending the notion of the derivative is not quite as simple, because the slope or

gradient of the f (x; y ) surface at (x; y ) depends on which direction you move o in.
The key idea is to derive the gradient in particular directions. For f (x; y ) the obvious

directions are those along the x axis and the y axis. And obviously if one wants to

move o from (x; y ) in the x direction, one has to keep y xed, and vice versa. This

leads to the following denitions.

The partial derivative of f (x; y ) wrt x , then wrt y

@f f (x + x; y ) f (x; y ) @f f (x; y + y ) f (x; y )
       
= lim ; = lim
@x x !0 x @y y !0 y
y x
(1.4)

The subscript y indicates that y is being kept constant, and then similarly for x
1.2.3 Geometrical interpretation of the partial derivative

Figure 1.3 shows the geometrical interpretation of the partial derivatives of a function

of two variables.

1.2.4 A possibly useful shorthand

Given that the list of variables is known, and the one being varied is explicit, writing

the the held constant subscripts after the derivative is unnecessary. The subscripts

are often omitted, but you may wish to leave them in until you are more familiar with

the subject. Another commonly used shorthand is

@f @f
   
fx = @x and fy = @y : (1.5)

1.2.5 More than two variables

If we are dealing with a function of more variables, we keep all but the one variable

constant. Eg for f (x1; x2; x3; :::) we have

The partial derivative again ...

@f f (x1; x2; [x3 + x3]; x4; : : :) f (x1; x2; x3; x4; : : :)

   
fx 3 =
@x3 = lim
x !0 x3
3
1/4 LECTURE 1. TOTAL AND PARTIAL DERIVATIVES AND DIFFERENTIALS

Figure 1.3: Interpreting partial derivatives as the slopes of slices through the function

1.3 Finding partial derivatives is easy

The denition indicates that actually doing partial dierentiation is exactly the same

as normal dierentiation with respect to one variable, while remembering to treat all

the other variables as constants.

| Examples.

Q: Find the rst partial derivatives of f (x; y ) = x 2y 3 2y 2 .

A: First assume y is a constant, and dierentiate wrt x ; and then vice versa:
fx = 2xy 3 and fy = 3x 2y 2 4y : (1.6)

Q: Find the 1st partial derivatives of f (x; y ) = e ( x 2 +y 2 )

sin(xy 2 )
A:

fx = e (x 2 +y 2 )
[ 2x sin(xy 2 ) + y 2 cos(xy 2 )] (1.7)

fy = e (x 2 +y 2 )
[ 2y sin(xy 2 ) + 2xy cos(xy 2 )] : (1.8)

Q: If f (x; y ) = ln(xy ), derive an expression for fx fy in terms of f.

1.4. HIGHER PARTIAL DERIVATIVES? 1/5

A:

f (x; y ) = ln(x ) + ln(y ) (1.9)

)fx = 1=x and fy = 1=y (1.10)

)fx fy = 1=xy = e f (x ;y )
: (1.11)

1.4 Higher partial derivatives?

Although you do not need to consider complicated relationships involving 2nd order

derivatives, you do need to know how to nd them. It is very obvious! fx and fy are

(well, probably are) perfectly good functions of (x; y ), so we can dierentiate both with
respect to either x or y.
| Example.

f (x; y ) = x 2y 3 2y 2 (1.12)

fx = 2xy 3 (1.13)

fy = 3 x 2 y 2 4y (1.14)

fx x = 2y 3 (1.15)

fy y = 6x 2 y 4 : (1.16)

But we should also consider

@ @ 2f
@y fx = fy x = @y@x : in this case; 6xy 2 (1.17)

and

@ @ 2f
@x fy = fx y = @x@y : in this case; 6xy 2 : (1.18)

Spooky! In this case fx y = fy x  is that always true?

1/6 LECTURE 1. TOTAL AND PARTIAL DERIVATIVES AND DIFFERENTIALS

| A meatier random example:

f (x; y ) = e ( x 2 +y 2 )
sin(xy 2 ) (1.19)

fx = e 2
( x +y ) 2
[ 2x sin(xy 2 ) + y 2 cos(xy 2 )] (1.20)

fy = e 2
( x +y ) 2
[ 2y sin(xy 2 ) + 2xy cos(xy 2 )] (1.21)

) fx y = e 2
( x +y ) 2
[ 4x 2 y cos(xy 2 ) + 2y cos(xy 2 ) 2y 3 x sin(xy 2 ) (1.22)

2y [ 2x sin(xy ) + y cos(xy )]]

2 2 2

= e ( x 2 +y 2 )
[sin(xy 2 )[ 2y 3 x + 4xy ] + cos(xy 2 )[ 4x 2 y + 2y 2y 3 ]]
and fy x = e ( x 2 +y 2 )
[ 2y 3 cos(xy 2 ) + 2y cos(xy 2 ) 2xy 3 sin(xy 2 ) (1.23)

2x [ 2y sin(xy 2 ) + 2xy cos(xy 2 )]]

= e ( x 2 +y 2 )
[sin(xy 2 )[ 2xy 3 + 4xy ] + cos(xy 2 )[ 2y 3 + 2y 4x 2 y ]]

So they are equal in this case too.

When both sides exist, and are continuous at the point of interest...

The dierential operators are equivalent:

@2 @2
=
@x@y @y@x (1.24)

This result has an interesting consequence for higher partial derivatives.

| Example.

Q: For a general function f show that

@ 3f @ 3f @ 3f
 4  5
=0
@ 2x@y @y@x 2 @x@y@x (1.25)

A: Using the ordering result

@ 3f @ @ 2f @ 3f @ 3f
@ 2x@y
= =
@x @x@y @x@y@x = @y@x@x ; (1.26)

and thus the equation is p:p p5  which is indeed zero.

4

So the order of higher partials is unimportant  but a sensible ordering can save time!

| Example.

Q: Find

@3  5
(y + xy ) cosh(cosh(x 2 + 1=x )) + y 2 tx :

@t@y@x (1.27)
1.5. A STERN WARNING. PARTIAL DERIVATIVES ARE NOT FRACTION-LIKE. 1/7

A: Thoughtless Method. Grind away blindly dierentiating with respect to x then y

then t. This may take a fortnight because the functions of x and y are moderately

unpleasant. You will be cross when you remember that ...

A: Thoughtful Method. ...

1.5 A stern Warning. Partial derivatives are not fraction-like.

Although total derivatives have fraction-like qualities, this is NOT the case with partial

derivatives.

| Example of single variable ...

Q: Find d y=dx given y = u 1=3, u = v 3 and v = x 2.

A:

y y du dv 1
d

dx
=
d

du

dv
dx =
3
u 2=3
:3v 2:2x = 2x (1.28)

which you can check by nding y = x 2 explicitly.

| Example. But but but! Suppose we were asked:

Q: Given the perfect gas law pV = RT , determine

@p @V @T
   

@V @T @p : (1.29)

A: You would guess +1 of course. But WRONG!

If pV = RT then p = RT=V , V = RT=p and T = V p=R. Thus

@p @V @T RT R V
       
= = 1:
@V @T @p
2
V p R (1.30)

In fact, study of implicit functions would show that for any function f (x; y; z ) = 0,
@x @y @z
= 1:
@y @z @x
Partial derivatives, unlike total derivates, do not behave as fractions.

 Although we CAN write down expressions like d f = :::

 We CANNOT write down @f = : : :.
 If you ever think that you are dividing partials, THINK AGAIN!
1/8 LECTURE 1. TOTAL AND PARTIAL DERIVATIVES AND DIFFERENTIALS

1.6 Total and partial dierentials

First, note that a dierential is dierent from a derivative.

Dierentials are about the following. Suppose that we have a continuous function

f (x; y ) in some region, and both fx and fy are continuous in that region. How much

does the value of the function change as one moves innitesimal amounts d x and d y
in the x and y f
directions? The amount, d , is the total dierential  how do we

express it?

The total or perfect dierential of f (x; y )

@f @f
d f =
@x dx +
@y dy (1.31)

df x y
@f @f
The total dierential is the sum of the partial dierentials d and d .
@x @y

Figure 1.4: Total dierential as the sum of partial dierentials. Remember that dx and dy are innites-

imals.

Explanation Given small changes in x and y it is easy enough to write the change in

f as:

f = f (x + x; y + y ) f (x; y ) : (1.32)

By adding in two cancelling terms this can be rewritten as

d f = [f (x + dx; y + dy ) f (x; y + dy )] + [f (x; y + dy ) f (x; y )] : (1.33)

But recall that

@f f (x + x; y ) f (x; y ) @f f (x; y + y ) f (x; y )

   
= lim & = lim
@x x !0 x @y y !0 y (1.34)
1.6. TOTAL AND PARTIAL DIFFERENTIALS 1/9

so that

@f @f
(  )  
d f =
@x (y + y ) x + @y y : (1.35)
y x

@f
(y + y ) @f
y + y .



where means evaluated at But
@x y @x y

@f @f @ 2f
(y + y ) 
     

@x y @x y
+
@y@x y (1.36)

and

@f @ 2f @f
f  @x x + @y@x xy + @y y :
 
(1.37)

In the limit as x and y tend to dx and dy , the middle term becomes negligible.

1.6.1 When is an expression a total dierential?

Suppose we are given some expression p(x; y )dx +q (x; y )dy . In this, p(x; y ) and q (x; y )
could be quite properly be any functions of (x; y ). But can we determine whether the

expression is also the total dierential of some function f (x; y )?

Now, if it is,

d f = p (x; y )dx + q (x; y )dy : (1.38)

But then we must have that

@f @f
p(x; y ) = @x and q (x; y ) = @y (1.39)

and using fx y = fy x
@p @ 2f @q @ 2f
@y = @y@x =
@x =
@x@y (1.40)

Hence:

The test: p(x; y )dx + q (x; y )dy is a total dierential i

@p @q
@y = @x : (1.41)

| Example. Show that there is no function whose total dierential is (xy dx + 2x 2 dy ).

Set p = xy and q = 2x 2. Then
@p @q
=x and these are not equal:)No function
@y and
@x = 4x (1.42)
1/10 LECTURE 1. TOTAL AND PARTIAL DERIVATIVES AND DIFFERENTIALS

1.6.2 Recovering the function from its total dierential

Suppose we found p(x; y )dx + q (x; y )dy was a total dierential using the above test.
Could we recover the function f ? To recover f we must perform the reverse of partial

dierentiation.

As @f =@x = p(x; y ):
Z
f = p(x; y )dx + g (y ) + K1 (1.43)

where g is a function of y alone and K1 is a constant. You can see that we need
the g (y ) because when we dierentiate with respect to x it vanishes. As far as x is

concerned, g (y ) is a constant of integration.

Similarly,
Z
f = q (x; y )dy + h(x ) + K2 (1.44)

We now need to resolve the two expressions for f , and this is possible, up to a constant
K , as the following example shows.
| Example. Let f = xy + sin x sin y + 6y + 10  but pretend we do not know it.
3

Instead we are asked

Q: Determine whether

(y 3 + cos x sin y )dx + (3xy 2 + sin x cos y + 6)dy (1.45)

is a total dierential and, if so, of what function f?

A: Testing whether @p=@y = @q=@x , we nd that
@ 3
@y (y + cos x sin y ) = 3y 2 + cos x cos y (1.46)

@
(3xy 2 + sin x cos y + 6) = 3y 2 + cos x cos y
@x
They are the same, so it is a total dierential.

Integrating (y 3 + cos x sin y ) x and (3xy 2 + sin x cos y + 6) over y

over we nd:

f = y 3x + sin x sin y + g (y ) + K1 (1.47)

and

f = xy 3 + sin x sin y + 6y + h(x ) + K2 : (1.48)

1.7. SUMMARY 1/11

Comparing and resolving these expressions we have

f = y 3x + sin x sin y + 0 + g (y ) + K1
l l l l (1.49)
f = xy 3
+ sin x sin y + h (x ) + 6y + K2
So

f = xy 3 + sin x sin y + 6y + K1 : (1.50)

Thus we have indeed recovered the original function, up to a constant of integration.

We would need some extra piece of information to recover this  say the value of the

function at a particular point.

Given p(x; y )dx + q (x; y )dy

 Test whether: @p=@y = @q=@x . If good:

 p(x; y ) wrt x , remembering g (y ) is a const of integration

Integrate

 q (x; y ) wrt y , remembering h(x ) is a const of integration

Integrate

 Resolve the two expressions.

1.7 Summary

We have dened partial dierentation, and seen that the process of nding partial

derivatives is straightforward.

We have found that the ordering of dierentiation in higher partial derivatives is of no

consequence, other than that of speed.

We have dened the total or perfect dierential of a function of several variables, and

seen how to test for its validity, and then to recover the original function up to a

constant. The value of the total dierential in various operations will become clear

later.
1/12 LECTURE 1. TOTAL AND PARTIAL DERIVATIVES AND DIFFERENTIALS

1.8 A.1 [**] For interest: Plotting Functions of several variables

The Figure shows the surface plot and the level contour plots of the function f (x; y ) =
0:6x + 0:1x
4 2
y 2
2x + 2 y + 5 .
2 2

10

6
f(x,y)

0
1
2
0 1
0
−1
y −1 −2
x

The following is some simple Matlab code to give a surface plot of a function.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% surface plot script
%
% vectors of grid x and y positions
x = [ -2:0.1:2];
y = [ -1:0.1:1];
% turn these in array suitable for surface plotting
[X , Y ] = meshgrid (x , y );
% find the function values at each point in mesh
Xsq = X .* X ;
Ysq = Y .* Y ;
Z = 0.6* Xsq .* Xsq + 0.1* Xsq .* Ysq - 2.*( Xsq - Ysq ) +5;
% initial graphics
set ( gca , ' FontSize ' , 18);
colormap ( ' default ');
brighten (0);
% now plot surface
surfc (X ,Y , Z );
% label axes
xlabel ( 'x '); ylabel ( 'y '); zlabel ( ' f (x , y ) ');
% save as postscript ( vector graphics )
print ( ' - depsc ' , ' surfplot . eps ');
% save as png image
print ( ' - dpng ' , ' surfplot . png ');
Lecture 2
Partial derivatives and function character
We now consider a number of relationships which can be derived if the function of

several variables has a certain form.

There is a danger that the various cases will become an unconnected jumble. It seems

helpful always to ask the question

Of what, exactly, is f a function?

If you are considering a function of multiple variables, think in terms of partial deriva-

tives; if a function of just one variable, think in terms of total derivatives.

2.1 A function of a function. f = f (u ) where u = u (x ; y ).

Suppose f (x; y ) = xy sin(xy ). We could nd the partial derivatives in the usual way as

@f
 

@x = y sin(xy ) + xy 2 cos(xy ) (2.1)

@f
 

@y = x sin(xy ) + x 2y cos(xy ) : (2.2)

But you might notice that by writing u = xy , f = f (u ) = u sin u So, u = u (x; y ) is

a function of more than one variable, but f can be written as a function of the single

variable u. f =du exists.

Hence total d

When f = f (x; y ) and u = u (x; y ) is found such that f = f (u )

@f df @u @f df @u
       
= =
@x du @x and
@y du @y (2.3)

If you think this result is obvious because you can cancel du and @u you need to

re-read the stern warning section!

1
2/2 LECTURE 2. PARTIAL DERIVATIVES AND FUNCTION CHARACTER

2.1.1 Proof

The result is proved as follows. The total dierentials of f (x; y ) and u (x; y ) are:
@f @f @u @u
       
d f =
@x d x+ @y d y; and du =
@x dx +
@y dy : (2.4)

We are allowed to take ratios of total dierentials, so that

x+ y
 
@f @f
f


d d
@x @y
:
d
=
du
(2.5)
x+ y
 
@u @u


d d
@x @y

But x and y are independent, and hence so are dx and dy . If we take y to be xed, so

that dy = 0, we have

df dx @f @u
@f

  

du
@x
= @u
=
@x @x ; (2.6)
dx
@x

and if dx = 0

df @f @u
  

du
=
@y @y : (2.7)

The result quoted follows immediately.

| Example.

Q: Find (@f =@x ) and (@f =@y ) when f = tan 1 (y=x ).

A: Set f = tan 1 (u )
u = y=x . Then with

df x2 @u @u
   

du
= 2
x + y2 ;
@x = y=x 2
;
@y = 1=x (2.8)

@f y @f x
) @x = x 2 + y 2 ; @y
   
= 2
x + y2 (2.9)
2.1. A FUNCTION OF A FUNCTION. F = F (U ) WHERE U = U (X; Y ). 2/3

| Example. This is a rather harder example.

Q: A quantity z (x; t ) is given by z = f (x ct ) where c is a constant. Show that z

satises the wave equation

@ 2z 1 @ 2z
   

@x 2
c2 =
@t 2 ;
no matter what f is exactly. (So z might be z = (x
ct )2 or z = 1= cosh(x ct ),
and so on, but we have to prove the result without knowing f .)

A: Write u = x ct , so that z = f (u ), and use the result just proved.

@z df @u df df
   
= = (1) =
@x du @x du du
(2.10)

But df =du is also just a function of u  let's say df =du = g (u ). So if we dierentiate

wrt x again

@z dg @u dg d f
2
 2   

@x 2 = du @x = du (1) = du 2 (2.11)

Now repeat the exercise, but dierentiating wrt t.

@z f @u f
   
( c ) = cg (u ) :
d d
= =
@t du @t du
(2.12)

So if we dierentiate wrt t again

@ 2z g @u g 2d f
2
 
c du @t c du ( c ) = c du 2 :
d d
= =
@t 2 (2.13)

The result follows straight away.

NB we have written nothing about the nature of function f, and you might care to

check the result for any arbitrary function.

2/4 LECTURE 2. PARTIAL DERIVATIVES AND FUNCTION CHARACTER

2.2 Composite functions and the Chain Rule

There are various cases of composite functions, and we deal with them in turn. This

section will also introduce the chain rule, which you will use over and over again.

2.2.1 What if f = f (x; y ) and x = x (t ) and y = y (t ) ?

Function f is said to be a composite function. Note that f is eectively a function of

t alone, so that total d f =dt exists.

There is a chain rule:

A Chain Rule: When f = f (x; y ) and x = x (t ) and y = y (t ):

f @f x @f y
   
:
d d d
= +
dt @x dt @y dt
(2.14)

Proof

To prove this, write the total dierential as

@f @f
   
d f =
@x dx +
@y d y (2.15)

We know that f, x and y are functions of t alone so the d f =dt , dx=dt , and dy=dt all

exist. So, dividing by d t (which we are allowed to do as it is a total dierential)

f @f dx @f y
   
d d
= +
dt @x dt @y dt
(2.16)

2.2.2 For a function of many variables ...

Suppose f is f (x1; x2; x3; : : : ; xn ) with xi = xi (t ). Again f must be a function of t alone,

so d f =dt exists.

The Chain Rule becomes

f @f x @f x n 
@f dxi
    
+ ::: =
t :
d d 1 d 2 X
= +
dt @x1 dt @x2 dt @xi (2.17)
d
i =1

| Example. Find d f =dt when

f = f (x; y; z ) = xy + yz + zx ; and x = t; y = e t ; z = cos t (2.18)

2.2. COMPOSITE FUNCTIONS AND THE CHAIN RULE 2/5

Thus

f @f dy @f dy @f dz
     
d
= + +
dt @y dt @y dt @z dt (2.19)

dx dy dz
= ( y + z ) + (x + z ) + (x + y )
dt dt dt
(2.20)

= e (1 t cos t sin t ) t sin t:

t
(2.21)

You could check the result by explicit substitution.

2.2.3 What if z = f (x; y ) and x = x (t1; t2; :::) and y = y (t1; t2; :::) ?

This is like the previous composite function, but now f is eectively a function of

several variables t1; t2; : : :. Suppose we x all but one of the ti . We would end up with

a composite function as above, but now instead of total derivatives d f =dt x= t

and d i d

we must have partial derivatives @f =@tj and @xi =@tj Thus

@f @f @x @f @y
       
= +
@t
 1
@x @t @y @t
   1    1
(2.22)

@f @f @x @f @y
= +
@t2 @x @t2 @y @t2 (2.23)

and so on.

We can now easily generalize this result to a function f (x1; x2; x3; :::; xn ) with xi =
xi (t1; t2; :::; tm ). After chomping your way through the indices you should nd that

The Chain Rule for Partials

@f @f @x1 @f @x2 @f @x3

       

@tj =
@x1 @tj +
@x2 @tj +
@x3 @tj + : : : : j = 1; : : : ; m: (2.24)

It turns out that the chain rule for partials is very commonly used in transforming from

one set of coordinates to another. Here is one example, but we shall return to it again.

| Example. x = r cos  and y = r sin  denes the transformation between Cartesian

and plane polar coordinates. Find @f =@r and @f =@ when f (x; y ) = x + y .
2 2

So, t1  r and t2  , and x1  x and x2  y .

2/6 LECTURE 2. PARTIAL DERIVATIVES AND FUNCTION CHARACTER

Using the chain rule for partials:

@f @f @x @f @y
= = 2x cos  + 2y sin 
@r @x @r + @y @r (2.25)

= 2r cos2  + 2r sin2  = 2r (2.26)

@f @f @x @f @y
= 2x ( r sin ) + 2y (r cos )
@ = @x @ + @y @ (2.27)

= 2r 2 cos  sin  + 2r 2 sin  cos  = 0 : (2.28)

(Because f= x 2 + y 2 = r 2 , this can be checked directly! As a bonus, note that

df @f
=
@r
dr @f
+ @ d = 2rdr , which is again consistent.)
2.2.4 What if f = f (x; y ) andy = y (x ) ?

Clearly f is a composite function of x alone.

So this is like t being x. (Eectively

x = x (x ), but one wouldn't bother to write this down.)

So,

f @f x @f y
   
d d d
= +
dt @x dt @y dt
(2.29)

morphs into

f @f x @f y
    
d d d
= +
dx @x dx @y dx
(2.30)

Ie,

f @f @f y
   
:
d d
= +
dx @x @y dx
(2.31)

Note that this expression contains both total and partial derivates of f wrt x. They

are dierent; they can co-exist!

| Example.

Q: Suppose z = xy + x=y and y = x.

p Find d z=dx .
A: Using the results just obtained,

z 1 1 1
y+y + x (1
y 2) 2 x
d 1=2
=
dx
(2.32)

1 1= 2
= x 1=2 + x 1=2 + x x

1=2
(2.33)
2
3 1
= x 1=2 + x 1=2 (2.34)
2 2
a result which can be checked directly in this case, using z = x 3=2 + x 1=2.
2.3. IMPLICIT FUNCTIONS 2/7

2.3 Implicit Functions

This result leads us on to discuss an important class of functions, implicit functions.

In calculus and analytic geometry it is common to see the equations of plane curves

written as

f (x; y ) = 0: (2.35)

p
For example the equation of a circle of radius 2 with centre at (0; 1) is

x 2 + (y 1)2 2=0 (2.36)

In this case it is easy to write an equivalent equation = y (x ) = 1  (2 x 2 )1=2 

y
but often it is the case that although one can write down f (x; y ) = 0 one cannot solve

for y = y (x ) explicitly. But if we were to solve for y numerically we might trace out a

curve y = y (x ) which was single valued and dierentiable. In other words, f (x; y ) = 0

may dene a y = y (x ) implicitly. Such a y = y (x ) is an implicit function.

It turns out that for many purposes we don't need an explicit expression for the function.

It is enough to characterize the function by its derivative.

So how do we compute the derivative d y=dx of an implicit function?

2.3.1 Finding the derivative of an implicitly dened function

At the end of Ÿ2.2.4 we saw that for f (x; y ) with y = y (x ),

f @f @f y
   
:
d d
= +
dx @x @y dx
(2.37)

Now suppose that we have f (x; y ) = 0. We have just argued that this gives us an

implicity = y (x ), so our previous result should hold. Indeed it does, but now we have
f = 0 always, so that df =dx = 0 always. Hence:
@f @f dy
0=
@x + @y dx : (2.38)

so that, (provided @f =@y is not zero at the point of interest),

dy @f =@x
dx
=
@f =@y : (2.39)
2/8 LECTURE 2. PARTIAL DERIVATIVES AND FUNCTION CHARACTER

| Example.

First consider an example where the implicit function can be derived explicitly. This

gives us a check on our results.

Q: Find d y=dx when f (x; y ) = x x 2y 3 = 0.

A: Use the result for implicit functions just derived:

y @f =@x 1 2xy 3
 
:
d
= =
dx @f =@y 3x 2 y 2
(2.40)

As a check, here we can nd y explicitly as y =x 1= 3

. Hence d y=dx = x 4=3
=3. Now

substitute for y in

y 1 2xy 3 1 2xx 1
1
     
x =3 :
d 4= 3
= = = =
dx 3x y 3x x 3x
(2.41)
2 2 2 2= 3 4=3

But remember! The whole point about implicit functions is that you do not need an

explicit y = y (x ) to get information about the derivatives. This is made clear in the

next example.

| Example.

Q: Find d y=dxf (x; y ) = ax 2 + 2hxy + by 2 + 2gx + 2f y + c = 0.

when

A: You may recognize f as a conic section  an ellipse or hyperbola, depending on

the constants. Now we can't nd y as a function of x , but using the result for implicit

functions we have

y 2ax + 2hy + 2g
:
d
=
dx 2by + 2hx + 2f
(2.42)
Lecture 3
Changing variables & Jacobians
3.1 Introduction

The slope of the f = f (x; y ) surface depends on the direction in which one moves from
the point (x; y ), and in Lecture 1 we learnt that the partial derivative must be dened

in terms of the change along a particular direction or axis.

For a function f (x; y ) we decided that the obvious directions to choose are along the
x and y axes, keeping y and x xed, respectively.
The question now is are these really the obvious directions?

For example, consider the function

f (x; y ) = e ( x 2 +y 2 )
cos(4(x 2 + y 2 ))
shown in Figure 3.1. Does it really make sense to impose a square  x -constant, y -
constant cake-rack mesh onto this function? Probably not! It would be in more

sympathy with the function to use a mesh with radial symmetry as in Figure 3.2.

To achieve this we must make an appropriate transformation to a new set of variables.

This raises the questions of

 How to choose the new variables and thus describe the transformation.
 How to describe the function in the new variables.
 How to nd the partial derivatives with respect to these new variables.

3.1.1 The general problem

The more general problem is illustrated in Figure 3.3. Subgure (a) shows a surface

f (x; y ) with lines of constant x and y underneath. The partial derivatives wrt x and y
are found by slicing or moving along these directions.

Figure 3.3(b) shows the exactly the same surface with lines of constant u and v un-

derneath. The partial derivatives wrt u and v are found by slicing along these new

directions.

1
3/2 LECTURE 3. CHANGING VARIABLES & JACOBIANS

0.5

f(x,y)
0

−0.5
5
4
0 2
0
−2
y −5 −4
x

Figure 3.1: The function exp (x 2 + y 2 ) cos(4(x 2 + y 2 )).

Figure 3.2: A radial dartboard mesh reects the symmetry rather better than the x; y constant cake

rack.

f (x , y ) F (u , v )

(a) (b)

Figure 3.3: (a) Surface f (x ; y ) above a set of x -constant and y -constant lines. (b) Surface F (u; v )
above a set of u -constant and v -constant curves but plotted in the x ; y frame. The surface has exactly
the same shape.
3.2. REWRITING THE FUNCTION AND FINDING THE DERIVATIVES 3/3

3.1.2 Choosing the new variables

There is no routine method for doing this, but usually the problem symmetry drops

very large hints as to the best transformation.

Later we will consider some of the most standard transformations, viz between Cartesian

and

 Plane Polar Coordinates (2D): for radial symmetry

 Spherical Polar Coordinates (3D): for spherical symmetry, and

 Cylindrical Polars Coordinates (3D): for cylindrical symmetry.

We also consider what would make a bad transformation.

3.2 Rewriting the function and nding the derivatives

First though we should consider the general problems of rewriting the function and

nding the partial derivatives with respect to arbitrary new variables (u; v ).
There are two main cases to consider:

1. The straightforward case. OLD variables in terms of NEW.

x = x (u; v ) and y = y (u; v ) : (3.1)

2. The more complicated case. NEW variables in terms of OLD.

u = u (x; y ) and v = v (x; y ) : (3.2)

3.2.1 CASES 1A and 1B: OLD variables in terms of NEW

We are transforming from (x; y ) to (u; v ) coordinates, and the transformation is given
as old in terms of new variables. That is:

x = x (u; v ) and y = y (u; v ) : (3.3)

Case 1A:

If we know what the function f (x; y ) is explicitly, we can replace x and y in the function
to get the new function as

f = f (x (u; v ); y (u; v )) = F (u; v ) : (3.4)

Then one can work out the partial derivatives with respect to u and v directly from the
new function F (u; v ).
3/4 LECTURE 3. CHANGING VARIABLES & JACOBIANS

| Example of Case 1A.

Let f (x; y ) = x=y , and let the transformation be x = u and y = u=v .

u @f @f
)f = F (u; v ) = (u=v )
= v and )
@u = 0;
@v = 1 : (3.5)

Case 1B: In this case we don't have an explicit form for f (x; y ), and, therefore, cannot
nd an explicit form for function F (u; v ).
But we can still write expressions for the partial derivatives using the Chain Rule for

partials from Lecture 2:

Case 1B: Use the Chain Rule for partials to obtain the derivatives

@f @f @x @f @y
       
= +
@u @x @u @y @u (3.6)

@f @f @x @f @y
       
= +
@v @x @v @y @v (3.7)

| Example of Case 1B.

We are told f = f (x; y ), and x = u and y = u=v .

Given this transformation, the best we can do is to write for any function f (x; y )
@f @f @f 1 @f @f 1
          
= (1) + = +
@u @x @y v @x @y v (3.8)

@f @f @f u @f u
       

@v =
@x (0) + @y v2 =
@y v 2 : (3.9)

To check this result, let's use (as in the example of Case 1A) f (x; y ) = x=y . Then

@f 1 x 1 v  u  v 2  v  v 
    
=
@u y (1) +  y2 v = u v u2 =
u u =0 (3.10)

@f 1 x  u  u 2v
 
=
@v y (0) + y2 v 2 = u 2v =1 (3.11)

This is exactly as found earlier at Equation 3.5.

3.2.2 A source of confusion introduced, and cleared up

Before continuing, it is worth checking that you are happy that the functions f () and
F () have dierent names. They work dierently on their parameters, so should be

named dierently.
3.2. REWRITING THE FUNCTION AND FINDING THE DERIVATIVES 3/5

But notice that the chain rule for partials seems to deliver an expression for @f =@u and
@f =@v .
But you might complain that f is a function of (x; y ) NOT (u; v ), and so we should
have written @F=@u and @F=@v . You are correct. So how are we getting away with it?

We are using f as a go-between value f = f (x; y ) and f = F (u; v ). This is correct

because at corresponding values of (x; y ) and (u; v ) the functions do return exactly

the same value. (Look back at Figure 3.3.)

z
Some books introduce another symbol ( , say) as the go-between. They write z =
f (x; y ) and z = F (u; v ) and then write
@z @z @x @z @y @z @z @x @z @y
= +
@u @x @u @y @u ; @v =
@x @v +
@y @v ; (3.12)

and so on  but this seems over-fussy.

3.2.3 CASES 2A and 2B: NEW variables in terms of OLD

We still wish to transform the function f = f (x; y ) to (u; v ) coordinates. But now the

transformation is given as new in terms of old variables, that is:

u = u (x; y ) and v = v (x; y ) : (3.13)

One's rst thought is that this should be the better way of dening a transformation,

but it turns out to be really inconvenient!

New-to-Old prevents us from nding either F (u; v ) or using the Chain Rule for Partials.
We cannot immediately work out @x=@u , etc.
We proceed in one of two ways.

Case 2A: Try to invert the transformation.

Use u = u (x; y ) and v = v (x; y ) as simultaneous equations from which to nd x =

x (u; v ) and y = y (u; v ).
If you can invert, the problem becomes a Case 1 problem.

| Example of Case 2A

Suppose you are told that

u=x and v = x=y : (3.14)

These are new in terms of old. But it is obvious that

x = u; and )v = u=y; ) y = u=v : (3.15)

3/6 LECTURE 3. CHANGING VARIABLES & JACOBIANS

We now have Old-in-terms-of-New, and can move to Case 1 above.

Case 2B: When the transformation is not invertible Write down the Chain Rule for

Partials the wrong way round! That is, we want @f =@u , so start with @f =@x .
@f @f @u @f @v
   
= +
@x @u @x  @v  @x (3.16)

@f @f @u @f @v
 
= +
@y @u @y @v @y (3.17)

Treat these as simultaneous equations in the unknown @f =@u and @f =@v and re-

arrange. I.e.:

@f @f @v @f @v @u @v @u @v
   

@u =
@x @y @y @x = @x @y @y @x (3.18)

@f @f @u @f @u @u @v @u @v
   

@v =
@y @x @x @y = @x @y @y @x (3.19)

So we can recover the partial derivatives with respect to the new variables without an

explicit expression for the function in the new variables.

3.2.4 SUMMARY

To summarize:

1: If you have the transformation as OLD variables in terms of NEW

either

1A: nd the function explicity and nd the partials from it directly, or

1B: nd just the partials using the Chain-rule-for-partials directly.

2: If you have the transformation as NEW variables in terms of OLD

either

2A: invert the given transformation to get OLD variables in terms of NEW, and

then go to CASE 1, or

2B: use the Chain-rule-for-partials for the OLD variables to give simultaneous

equations for the partial derivatives with respect to to the NEW variables.
3.2. REWRITING THE FUNCTION AND FINDING THE DERIVATIVES 3/7

| Example #1

Q: The elevation of a hill is described by

z = z (x; y ) = exp (x 2 + y 2 ) :

(3.20)

By deriving the function in plane polars, nd @z=@r and @z=@ and comment on their
values. Determine the value of r where the hill is steepest.

A: Plane polars are (r; ) where x = r cos , y = r sin .

This is old in terms of new, and we know the function, ) CASE 1A.

The hill is therefore

z = exp r2 :

(3.21)

So

@z @z
2r exp r2
= and
@ = 0 :

@r (3.22)

The change of height is all in the radial direction. @z=@ = 0 means that if you move
at constant r (that is, round the hill) you will not change height at all.

The gradient (Slope in the equation below) is a function of r alone, so we can nd the

total derivative

( 2r exp r2 ) = ( 2 + 4r 2 ) exp r2
d d
=
 
d r Slope
d r (3.23)

p
p
This is zero (the 2D equivalent of an inection point) when r = 1= 2 and Slope =
2 exp f 1=2g.
| Example #2.

Here is one that can be solved using all approaches.

Q: Find the partial derivatives @f =@u and @f =@v when

f = f (x; y ) = x 2 y2 and u = (x + y ) v = (x y ) : (3.24)

A: Transformation is new in terms of old. ) CASE 2.

Try CASE 2A: Can we invert the transformation? Yes! Adding then substracting

we nd

1 1
x = 2 (u + v ) y = 2 (u v ) : (3.25)
3/8 LECTURE 3. CHANGING VARIABLES & JACOBIANS

Now go to CASE 1. We know the function, ) CASE 1A.

1
f = F (u; v ) = x 2 y 2 = 4 (u + v )2 (u v )2 = uv


(3.26)

So the derivatives are

@f @f
=v and =u
@u @v (3.27)

That's the problem solved. But now notice that if CASE *A works, we can always nd

just the derivatives using CASE *B.

For example, having inverted the transformation to get

1 1
x = 2 (u + v ) y = 2 (u v ) :
we could use CASE 1B to nd just the derivatives:

@f @f @x @f @y 1 1
= = (2x ): + ( 2y ): = x y = v
@u @x @u + @y @u 2 2
(3.28)

@f @f @x @f @y 1 1
= = (2x ): + ( 2y ): =x +y =u
@v @x @v + @y @v 2 2
(3.29)

Now pretend that we did not (or could not) invert the transformation. We could still

nd the derivative using Case 2B.

Write the chain rule for partials the other way around:

@f @f @u @f @v
   
= +
@x @u @x  @v  @x (3.30)

@f @f @u @f @v
 
= +
@y @u @y @v @y (3.31)

and for simplicity ll in the values you know

@f @f
   
2x = (1) + (1)
@u @v (3.32)

@f @f
   
2y = (1) +
@u @v ( 1) (3.33)

Now use as simultaneous equations for @f =@u and @f =@v :

@f @f
)
   
2x 2y = 2 = (x y) = v ;
@u @u (3.34)
3.2. REWRITING THE FUNCTION AND FINDING THE DERIVATIVES 3/9

and subtracting gives, once again, the same results:

@f @f
)
   
2x + 2 y = 2 = (x + y ) = u :
@v @v (3.35)

| Example #3. Harder!

Q: A quantity V = V (x; y ) satises Laplace's equation

@ 2V @ 2V
@x 2 + @y 2 =0 : (3.36)

Express this equation in plane polar coordinates (r; )

x = r cos , y = r sin .
where

A: We want rst to write down expressions for @V=@x and @V=@y which will involve

@r=@x , @r=@y , @=@x , and @=@y . This is the wrong way round for the transformation.
So

 Either invert the transformation (Case 2A), or

 Write the Chain Rule for Partials the other way round (Case 2B) and use simul-

taneous equations

Both will work, but let's use 2B ...

@V @V @x @V @y @V @V
       
= + = cos  + sin 
@r @x @r @y @r @x @y (3.37)

@V @V @x @V @y @V @V
       
( r sin ) +
@ =
@x @ +
@y @ =
@x @y (r cos ) (3.38)

Sorting out the simultaneous equations (do check!) gives

@V @V sin  @V
= cos 
@x @r r @ (3.39)

@V @V cos  @V
= sin 
@y @r + r @ (3.40)

We can drop the V and consider

@
,
@
,
@x @r @
@
as operators

@ @ sin  @
= cos 
@x @r r @ (3.41)

@ @ cos  @
= sin  +
@y @r r @ (3.42)
3/10 LECTURE 3. CHANGING VARIABLES & JACOBIANS

So the operator

@2 @ sin 
@ @ sin  @
  
= cos  cos 
@x 2 @r r @ @r r @ (3.43)

@2 1 @ 1 @2
 
= cos  2 cos sin 
r 2 @ + r @r@
2
@r (3.44)

1 @ @2 sin  @
cos  @2
sin 
   

r sin  sin  @r + cos  @@r + r r @ r @2

which can be tidied up a little. Note how the operators move through to the right,

operating as they go.

Let's follow through a couple of terms. The following is straightforward:

@ @ @2
  
cos  cos  = cos 2
 @r 2
@r @r (3.45)

but this requires the product rule

@ sin  @ 1 @ 1@2
      
cos  = cos  sin  cos  sin 
@r r @ r2 @ r @r@ (3.46)

The tidied expression for y is (you should check!)

@2 @2 @ 1 @2 2
 
= sin  2
2
cos sin  2
@y 2 @r r @ r @@r (3.47)

1 @ @2 1 2 @
2
 
+ cos  cos  + sin 
r @r @@r + r 2 cos  @2 (3.48)

Adding them up:

@2 @2 @2 1 @ 1 @2
 
+ 2 2 :
@x 2 + @y 2 = @r 2 + r @r r @ (3.49)

So Laplace's equation in plane polar coordinates is

@ 2V 1 @V @ 2V
1
+ = 0:
@r 2 r @r + r 2 @2 (3.50)
3.3. JACOBIANS (JACOBIAN DETERMINANTS) 3/11

3.3 Jacobians (Jacobian determinants)

Recall the equations that appeared from solving the simultaneous equations in case 2B:

@f @f @v @f @v @u @v @u @v
   

@u =
@x @y @y @x = @x @y @y @x (3.51)

@f @f @u @f @u @u @v @u @v
   

@v =
@y @x @x @y = @x @y @y @x (3.52)

Notice that the denominator is the same in both expressions. It has special signicance,

and is called the Jacobian.

It is common to write the Jacobian as a determinant, but there is also another useful

notation.

@u @v @v @u @u @v
@ (u; v )
 
J= @x @y @x @y = @x
@u
@x
@v =
@ (x; y ) (3.53)
@y @y
As we shall see later, the Jacobian is especially useful when changing variables from

(x; y ) to (u; v ) in multiple integrals.

3.3.1 Multiple denitions of the Jacobian determinant?

In dierent books one sees dierent denitions of the determinant:

@u @u
@ (u; v ) @u @v
@ (u; v )
@ (x; y ) =
@x
@u
@x
@v while in others
@ (x; y ) =
@x
@v
@y
@v : (3.54)
@y @y @x @y
This is puzzling, until you realize that they are identical! However, perhaps the rst

way is to be preferred for reasons we see now.

3.3.2 Jacobians and their inverses

In examples given later you will see that

@ (x; y ) @ (u; v )

=1
@ (u; v ) @ (x; y ) (3.55)

We shall gain an intuitive understanding of why this must be so when we come to use

Jacobians in multiple integration. However it is quite easy to prove using the Jacobian

matrix. *NOTE: This is not on the syllabus.*

The Chain rule for partials says

@f @f @x @f @y @f @f @x @f @y
=
@u @x @u + @y @u @v =
@x @v +
@y @v (3.56)
3/12 LECTURE 3. CHANGING VARIABLES & JACOBIANS

or, switching the order of the products,

@f @x @f @y @f @f @x @f @y @f
=
@u @u @x + @u @y @v =
@v @x + @v @y (3.57)

We can treat these as operator equations: in steps ...

@ @x @ @y @ @ @x @ @y @
@u f =
@u @x f +
@u @y f @v f =
@v @x f +
@v @y f (3.58)

@ @x @ @y @ @ @x @ @y @
) @u f  @u @x f +
@u @y f @v f  @v @x f +
@v @y f (3.59)

In matrix notation, this operator is

@=@u @x=@u @y=@u @=@x @=@x

      
= =J
@=@v @x=@v @y=@v @=@y @=@y (3.60)
x ;y
u;v

The matrix J x ;y is a Jacobian matrix.

u;v

Now repeat with the Chain Rule around the other way. You will arrive at the following

@=@x @u=@x @v=@x @=@u @=@u

      
= =J
@=@y @u=@y @v=@y @=@v @=@v (3.61)
u;v
x ;y

But it follows that

@=@u @=@x @=@u

     
=J =J J
@=@v @=@y @=@v (3.62)
x ;y x ;y u;v
u;v u;v x ;y

so that

) J
1
J J =I J :
h i
x ;y
u;v
u;v
x ;y
x ;y
u;v
= u;v
x ;y
(3.63)

But a result from linear algebra tells us that, if A and B are square, jABj = jAjjBj
,

and hence jAjjA j = 11

. The result for Jacobian determinants follows immediately.

3.4 Three standard transformations

There are certain transformations which occur very frequently:

(1) Cartesian to plane polar coordinates;

(2) Cartesian to spherical polar coordinates; and

(3) Cartesian to cylindrical polar coordinates.

3.4. THREE STANDARD TRANSFORMATIONS 3/13

y (x,y) (r, ϕ )
r

Figure 3.4: Cartesian to plane polars. Many texts use  rather than .

3.4.1 Cartesian to plane polars

This is useful for problems with radial symmetry in the plane.

The transformation is

x = r cos ; y = r sin  (3.64)

) @x
@r = cos ;
@x
@ = r sin ;
@y
@r = sin ;
@y
@ = r cos  (3.65)

Writing the Chain Rule for Partials in operator form gives

@ @x @ @y @ @ @
= + = cos  + sin 
@r @r @x @r @y @x @y (3.66)

@ @x @ @y @ @ @
= r sin  + r cos 
@ = +
@ @x @ @y @x @y : (3.67)

Using these simultaneous equations gives

@ @ sin 
@ @ @ cos  @
= cos  = sin 
@x @r r @ and
@y @r +
r @ : (3.68)

(You could also nd these by inverting the transformation.)

The Jacobian
@x @x
@ (x; y ) @r @
= r cos2  + r sin2  = r :
@ (r; ) = @y
@r
@y
@
(3.69)

and
@r @r
@ (r; ) cos  sin  1
= (cos )( (sin )(
r :
@x @y
@ (x; y ) = @
@x
@
@y
r )
r )= (3.70)
3/14 LECTURE 3. CHANGING VARIABLES & JACOBIANS

3.4.2 3D Cartesian to cylindrical polars

Cylindrical polars add a z -axis, perpendicular to the plane, which is identical to the

Cartesian z -axis.
z r (x,y,z) (r, ϕ ,z)

ϕ
x

Figure 3.5: Cartesian x; y; z to cylindrical polars r; ; z

The transformation is

x = r cos ; y = r sin ; z = z: (3.71)

So

@ @ @
= cos  + sin 
@r @x @y (3.72)

@ @ @ @ @
@ = r sin  @x + r cos  @y and
@z =
@z (3.73)

and using these as simultaneous equations (or inverting the transformation):

@ @ sin  @
= cos 
@x @r r @ (3.74)

@ @ cos  @ @ @
= sin  + =
@y @r r @ and
@z @z (3.75)

The Jacobian is now a 33 determinant

@x @y @z
cos  sin  0
@ (x; y; z ) @r
@x
@r
@y
@r
@z
r sin  r cos  0 = r;
@ (r; ; z ) = @
@x
@
@y
@
@z
=
0 0 1
(3.76)

@z @z @z

and (do check!)

@r @ @z
@ (r; ; z ) @x @x @x 1
r :
@r @ @z
@ (x; y; z ) = @y
@r
@y
@
@y
@z
= (3.77)

@z @z @z

Most entries can be copied over from the plane polar case.
3.4. THREE STANDARD TRANSFORMATIONS 3/15

3.4.3 Cartesian to spherical polars

Polar angle (x,y,z) (r, θ, ϕ )

r
y

ϕ
x

Azimuthal angle

Figure 3.6: Cartesian to spherical polars

The transformation involves the radius r of a sphere r 2 = x 2 + y 2 + z 2, the polar angle

0 and the azimuthal angle 0    2

x = r sin  cos ; y = r sin  sin ; z = r cos : (3.78)

Hence

@ @ @ @
= sin  cos  + sin  sin  + cos 
@r @x @y @z (3.79)

@ @ @ @
@ = r cos  cos  @x + r cos  sin  @y r sin  @z (3.80)

@ @ @
@ = r sin  sin  @x + r sin  cos  @y (3.81)

and

@ @ 1 @ sin  @
= sin  cos  + cos  cos 
@x @r r @ r sin  @ (3.82)

@ @ 1 @ cos  @
= sin  sin  + cos  sin  +
@y @r r @ r sin  @ (3.83)

@ @ sin  @
= cos 
@z @r r @ (3.84)
3/16 LECTURE 3. CHANGING VARIABLES & JACOBIANS

The Jacobian

sin  cos  r cos  cos  r sin  sin 

@x @x @x
@ (x; y; z ) @r @ @
@y @y @y
= sin  cos  r cos  sin  r sin  cos  = r 2 sin ;
@ (r; ; ) = @r
@z
@
@z
@
@z
cos  r sin  0
@r @ @
(3.85)

and as an exercise you might show that

@r @r @r
@ (r; ; ) @x @y @z
1
@ (x; y; z ) =
@
@x
@
@y
@
@z =
r 2 sin  : (3.86)
@ @ @
@x @y @z

3.5 A good transformation

Cylindrical and spherical polars are particularly useful when solving problems involving

cylindrical and spherical regions. But what in general makes a good transformation?

If you were asked what is the nicest region shape to t into Cartesian coordinates Oxy
you would answer a rectangle.

Similarly, if you drew your Ouv coordinate system at right angles and asked the same

question, you would again answer a rectangle.

So given some arbitrary region shape in the xy plane, the best sort of transformation
is one which maps it onto a rectangle in the uv plane, as illustrated in Figure 3.7(a).

A specic example is given in Figure 3.7(b) for plane polars. The quadrant's extent is

awkard to describe in Cartesians, but easy in describe in plane polars.

3.5.1 Will any transformation work? When is one bad?

Another question is will any transformation work? The answer is no. If we start with

two (or n) independent variables (x; y ), we must ensure that the new two (or n) are
also independent.

The test for independence is rather simple.

If u (x; y )and v (x; y ) are functionally INDEPENDENT then

@ (u; v )
@ (x; y ) 6= 0: (3.87)

Recall that the Jacobian is the denominator in a relationship relating partial derivatives

with respect to dierent sets of variables. Trouble is expected if this is zero.

3.5. A GOOD TRANSFORMATION 3/17

y v

x u

y ϕ

π/2

x r

a a

Figure 3.7: Useful transformations tend to take an awkward region shape in Ox y into a rectangle in

Ouv .

Proof: For reading only

The proof is straightforward. u (x; y ) and v (x; y ) are functionally dependent, then
If

u = u (v ), which in turn means that a function z = z (u; v ) = 0 exists. Dierentiating

z with respect to x and y gives:
@z @u @z @v
@u @x + @v @x = 0 (3.88)

@z @u @z @v
@u @y + @v @y = 0 (3.89)

For consistency between these two equations we must have ux = uy and vx = vy ,

where is some constant. Thus

@ (u; v ) u v
= x x = (ux vy vx uy ) = ( uy vy vy uy ) = 0 :
@ (x; y ) uy v y (3.90)
3/18 LECTURE 3. CHANGING VARIABLES & JACOBIANS

| Example

Q: Test whether there is a functional dependence between the variables (u; v ) if

u = x2 + y + 1 and v = x 4 + 2x 2y + y 2 x 2 y: (3.91)

A:

@ (u; v )
= ( ux v y v x uy )
@ (x; y ) (3.92)

= (2x )(2x 2 + 2y 1) (4x 3 + 4xy 2x )(1)

= (4x 3 + 4xy 2x ) (4x 3 + 4xy 2x )
= 0

So, yes the variables are not independent. With a little work you'll spot that the

dependence is v =u 2
3u + 2 .
Lecture 4
Multiple Integration

4.1 The integral for a single variable

Recall that in the Calculus I lectures we considered a function f (x ) dened over some
bounded region R of x , and divided up R into n subregions, where xi denoted the width

of the i th subregion. Letting fi be the associated function value, the single integral was

the limit
Z n

f (x )dx = fi xi :
X
lim (4.1)
n!1 i =1
xi !0
Region R

f(x i )

δx
i
x

x
i

Figure 4.1: The integral for one variable

4.2 Double Integral

The extension to two variables is straightforward. Consider a function f (x; y ) which

is dened in some bounded region R of the (x; y ) plane. Let R be divided up into n

subregions, where Ai denotes the area of the i th subregion. Let fi be the function

1
4/2 LECTURE 4. MULTIPLE INTEGRATION

value associated with the i th subregion. If the sum exists and is nite as n!1 and
Ai !0, then its limit is the double integral:
ZZ n

f (x; y )dA = fi Ai :

X
lim (4.2)
n!1 i =1
Ai !0
We assume here that it is irrelevant how the region is subdivided and the the xi ; yi
chosen. This is always true for a continuous function f (x; y ). Later we indicate how

to extend integration to discontinuous functions.

Function
surface

fi
y

dA i
Region R
x in x,y plane

Figure 4.2: The double integral

| Example

Q: Consider a thin plate, whose surface density (mass per unit area) is a function

(x; y ). R of the plate?

What is the mass of region

A: The innitesimal element of mass in area dA at (x; y ) must be

M = (x; y )dA )
ZZ ZZ
d Total M= d M= (x; y )dA : (4.3)
R R

4.2.1 Tiling the region

To progress further we have to (i) dene the region of integration, and (ii) dene d A.
You can think of placing the d A's as generating a tiling across the region.
In Cartesian coordinates, the xy plane is divided into a cake rack of x -constant, y -

constant lines. The tiles are rectangles, so that in the limit

d A = dx dy : (4.4)
4.2. DOUBLE INTEGRAL 4/3

Tile dxdy in general position

y

Region in
the x,y plane

x
Figure 4.3: Placing a tile in general position

| Example /ctd The integral for the mass is written

ZZ
Mass = (x; y )dx dy: (4.5)
R

You start by placing a tile in general position  at an (x; y ) value that is not special

in any way  as shown in Figure 4.3.

4.2.2 Summing tiles into strips

Suppose we rst summed the d d x y tiles into a strip parallel to the y -axis. For each

strip this involves holding x constant and summing up the length along y from one

boundary intersection to the other.

Sum over tiles

in a strip

y (x)
2
Region in
the x,y plane

y 1(x)

x x+dx x
Figure 4.4: Summing tiles into a strip parallel to the y -axis.
4/4 LECTURE 4. MULTIPLE INTEGRATION

Note that we are assuming that the x =constant line intersects the boundary just twice.
More on this later.

| Example /ctd The mass of the thin strip is

"Z #
y =y2 (x )
d M= (x; y )dy d x (4.6)
y =y1 (x )

4.2.3 Summing strips into the region

Then we want to sum the strip values up from the smallest to the largest values of x,
Sum over strips
in the region

Region in
the x,y plane

x1 x2
Figure 4.5: Summing strips into the complete region

| Example /ctd The total mass of the region is

"Z #
Z x2 y =y2 (x )
Mass of region = (x; y )dy dx (4.7)
x =x1 y =y1 (x )

Note that, because the inner limits are function of x, the length of the strip varies

automatically as x changes.

4.2.4 Evaluating the integral

This is done in two stages, as a repeated integral.

 (x; y )dy x is held constant, simply

hR i
y =y2 (x )
First, consider the integral . Because
y =y1 (x )
treat it no dierently from any other constant. (Recall that we already used this

method in Lecture 1 when integrating total or perfect dierentials.) We will then

4.3. A STANDARD RECIPE FOR SUCCESS ... 4/5

be left with some new function of (x; y ) which is evaluated between the limits

y = y1(x ) and y = y2(x ). This yields some function of x alone, g (x ) say.

 Then integrate the function g (x ) over x and evaluate it between the limits x1 and
x2 .
Thus the double integral is broken down into two single integrals.

| Example

Q: Derive the area of the shaded triangular region of the xy plane by integration.

y
y 2 (x)
1

y 1 (x)
x
1 2

A: Obviously the result should be 1.

y =1
2 2
1
ZZ Z Z  Z  
A= x y=
d d d y dx = 1
2
x d x (4.8)
R x =0 y =x =2 x =0

1 22 1

= x x =2
4 0 4
4= 1 (4.9)

4.2.5 The order of integration does not matter

You can just as well sum tiles into strips along the x -direction, then sum strips in the
y -direction. You must take care to redene the region.

1 x =2y 1
ZZ Z Z  Z
A= x y=
d d d x dy = 2 y dy (4.10)
R y =0 x =0 y =0

= y 2 10 =

1 (4.11)

4.3 A standard recipe for success ...

The recipe for success is:

4/6 LECTURE 4. MULTIPLE INTEGRATION

x 1 (y) 1 x 2 (y) 2

1. Sketch the region of integration.

2. Draw the tile in general position
3. Think about limits as you build the tile in a strip, then the strip into the region.
4. Think about the  quantity associated with the tile.
5. Then write down the integral, and nally
6. Solve it.

| Example

Q: The mass per unit area of the thin triangular plate varies as (x; y ) = k xy , where
k is constant. Find the plate's mass.

A:

Step 1 and 2: Sketch region and place tile

y
y 2 (x)
1

y 1 (x)
x
1 2

Step 3: Think about limits

Tile into strip: y goes from x=2 to 1. Strip into region: x goes from 0 to 2.

Step 4: Delta quantity. This is the physics

d M = (x; y )dA = k xy dx dy : (4.12)

Step 5: Write down integral

x =2 y =1
Z Z 
M=k x y dy dx : (4.13)
x =0 y =x = 2
4.4. COMPLICATED REGIONS 4/7

Step 6: Do it

x =2 y =1
Z Z 
M = k x y dy dx (4.14)
x =0 y =x =2

x2 k x3
Z x =2 x =2
1
  Z  
= k x2 1 4
d x =
2
x 4
d x
x =0 x =0

k x2 x4 2
k k

= = [2 1] =
2 2 16 0
2 2

4.4 Complicated regions

Earlier on it was noted that for a continuous function it does not matter how you

subdivide the region. There are several types of complicated regions where care must

be taken.

(a) Suppose we wish to nd the mass of a plate covering a region R comprised of

two sub-regions R1 having a density function 1(x; y ) and R2 having a density function
2(x; y ). Now it does matter how you subdivide R. No little element A should straddle
the boundary of R1 and R2 . The obvious solution is to perform two separate double

integrations and add up the results:

dA =  dA + 2dA :
RR RR RR
R R 1
1 R2

Note that the functions 1 and 2 can be discontinuous at the boundary.

R1

σ1 σ2

R2
R1 R2

Needs two Needs two One region Needs two

regions regions regions
(a) (b) (c)

Figure 4.6: Three region problems discussed above

4/8 LECTURE 4. MULTIPLE INTEGRATION

(b). Any closed smooth contour has two points where the tangent is parallel to the x
axis and another two where it is parallel to the y axes, which coincide with the extremal
values of x and y respectively. Problems are caused when there are more tangent points

which are parallel to either the x or y axes, but which do not coincide with the extremal
values. In this case, there will be some lines of constant x and/or constant y which

cut the boundary more than twice (but note that the number of crossings is always

even). But integrals only have two limits! The solution is to subdivide the region R

into subregions not troubled by this shape problem, integrate over the subregions and

then sum the results.

(c). Sometimes the boundary is not dened by a single function. In the example

shown, summing tiles into strips in the y direction would be straightforward, as the

two boundaries provide the lower and upper limits. However, summing tiles along the

x -direction would require splitting the region into two.

4.5 Change of variables in double integrals

The denition of the double integral involved a subdivision of the region into arbitrarily

shapes tiles d Ai . Then, for a function f (x; y ), it was noted that the obvious tiling is
d A = dx d y .
However, there is nothing particularly special about chopping the region of integration

into little rectangles. Indeed, as we discussed in Lecture 3, often the shape and sym-

metry of a problem suggests a dierent tiling or mesh, introduced by a transformation

to a new set of variables.

Suppose we specify a transformation to u; v coordinates as

x = x (u; v ) ; y = y (u; v ) : (4.15)

Let us keep the function in xy space and consider the new tiling.

The tiling is no longer one of lines of constant x separated by x and constant y

separated by y , but rather one of lines of constant u separated by u and constant
v separated by v . Typically, these will be curves in the xy plane. (Indeed (u; v ) are
commonly called curvilinear coordinates.)

Two key things to note are that (i) these tiles are NOT rectangular, and (ii) they

do NOT have area d u v

d .
4.5. CHANGE OF VARIABLES IN DOUBLE INTEGRALS 4/9

To transform the integral we require three things:

1. The function value in (u; v ) coordinates.

2. The tile area
3. The region dened in (u; v ) space.

1. A function value for each (u; v ). This is obtained by substituting for x and y
using x = x (u; v ) and y = y (u; v ). Thus F (u; v ) = f (x (u; v ); y (u; v )).
2. The area of the tile. This is best found using vector calculus, which we cover

more fully later, but the argument is not dicult to follow.

Let r(x; y ) = x^ + y^ be the position vector of points in the x; y -plane, with  ^; ^
the usual unit vectors. But as x = x (u; v ) and y = y (u; v ), we may write d r, an
innitesimal change in position, in two ways:

@r @r
d r = dx^ + dy^
  and d r = @u du + @v dv : (4.16)

The tile in the new coordinates is not a rectangle but a parallelogram and its area

can be found from the vector product of its sides, which we have just seen are

@r @r
@u du and
@v dv : (4.17)

We can re-express these vector sides using

@x @x @y @y
dx = @u du + @v dv and dy = @u du + @v dv ; (4.18)

inserting these into Eq 4.16, and then equating the coecients of du and dv . (We

can equate coecients because du and dv are independent.) This gives the two

y y

x x

Figure 4.7: A region (left) showing lines of constant x and y giving rise to a dx dy tiling and (right)

showing lines of constant u and v giving rise to a du dv tiling. These tiles are NOT rectangles, and do

NOT have area du dv .

4/10 LECTURE 4. MULTIPLE INTEGRATION

vector equations

@r @x @y @r @x @y
@u =
@u^ + @u^
  and
@v =
@v ^ + @v ^ :
  (4.19)

Thus the elemental area we want is

@r @r

 
dA = mod
@u du @v dv
^ ^
^ k
@x @y @x @y
 
= mod @x
@u
@y
@u
0 du dv =
@u @v @v @u u v :
d d
@x @y
@v @v
0

But the j:::j term is nothing other than the modulus of the Jacobian, so that

@ (x; y )
 
d A = dx dy = mod
@ (u; v ) u v:
d d (4.20)

3. The last issue we have to consider is the region of integration. We have to express

the regionR0 in uv space using the limits of integration of u and v to exactly cover
the physical region R previously expressed in terms of x and y limits.

This requires some care, but is best discussed by example.

| Example #1

Q: Consider a semicircle of radius a centred at the origin of xy space which lies in the

x > 0 half space.

i) Work out its area using integration in Cartesian coordinates.

ii) Do the same by transforming to plane polars.

(The answer should be a2=2 of course!)

A: (i) Place the d d x y x; p
y . Integrate tiles into strips palong the
tile in general position

y -direction, so that that limits are ylower (x ) = a2 x 2 and yhigher (x ) = + a2 x 2.

Then strips are integrated over x from x = 0 to x = a.

So the integral is
p
Z x =a Z y =+ a2 x 2 Z x =a
A= y x= a 2 x 2 dx :
p
p d d 2 (4.21)
x =0 y= a2 x2 x =0
4.5. CHANGE OF VARIABLES IN DOUBLE INTEGRALS 4/11

(a) (b)

Figure 4.8: Semicircle region in (a) Cartesians and (b) Plane Polars

This is now a standard 1-variable integral. We need a trig substitution, eg x = a sin p

and d x = a cos pdp
x =a =2
a2 =2
a2 
Z Z Z
a x dx = 2a p dp = (1 + cos 2p )dp = :
p
2 2 2 2
cos 2
(4.22)
x =0 p =0 2 p =0 2
A: (ii) The transformation is x = r cos ; y = r sin .
 The function is f (x; y ) = 1, so that F (r; ) = 1.
 Find the MODULUS of the Jacobian.

@x @x @y @y
= cos ; = r sin  = sin ;
@r @ @r @ = r cos  (4.23)

so

@ (x; y )
= r cos2  ( r sin2 ) = r :
@ (r; ) (4.24)

Now r > 0, so jr j = r , and dA = r dr d.

 Find region. This is 0  r  a and =2    =2. Note that the new region
is a rectangle in r;  space. As we have noted before this is the sort of convenient

shape we should hope for when make a good transformation.

So the new integral is entirely separable

a +=2 +=2 a
1
Z Z Z Z
A= rdrd = d rdr = 2 a2: (4.25)
r =0 = =2 = =2 r =0
4/12 LECTURE 4. MULTIPLE INTEGRATION

| Example #2

If you were not convinced about the merits of transformation by the previous example,

try this.

Q: The number of dopant atoms per unit area in a at semicircular semiconducting

wafer is
2
(x + y )
2 3=2
. Determine the average dopant level per unit area.

A: We must work out the total number of atoms and divide by the area. In Cartesians

the total number of atoms is

Z a Z +pa 2 x2
N= p (x 2 + y 2 )3=2 dy dx (4.26)
x =0 y= a2 x2

which is a mess.

Transforming ...

 Find function in terms of r

: f (x; y ) = (x 2 + y 2)3=2 = r 3.
and

 Modulus of the Jacobian, as before, is r , and dA = r dr d.

 Semicircle in xy space transforms to rectangle 0  r  a, =2    =2.

So

=2 a
a5
Z Z
N= r 3 r dr d =  5 : (4.27)
= =2 r =0

Now divide by the area, A = a2=2, and the average number per unit area is T=A =
2a3 =5.

4.6 Triple integrals

The method we have discussed for double integrals is, straightforwardly extensible to

triple and higher integrals, and is best explored now through examples. However, here

are the formalities.

Denition: The triple integral is dened as

ZZZ n

f (x; y; z )dV fi Vi

X
= lim
n !1
(4.28)
R i =1

Volume element: In Cartesians, we can write d V = dx dy dz and evaluate as a repeated

integral.
4.6. TRIPLE INTEGRALS 4/13

Region of Integration: Dening the region of integration tends to be harder in 3D,

but the plan is the same. You place a volume element (a cuboid in Cartesians) in

general position (x; y; z ), integrate the element rst into a rod, then integrate the

rod into a lamina or plate, and nally integrate the plate into the complete volume.

Changing variables: One can also change variables, and again the pattern is exactly

the same. Suppose we wish and we wish to change to variables u; v; w where

x = x (u; v; w ) y = y (u; v; w ) z = z (u; v; w ) (4.29)

Transforming the integral involves

1. Finding the new function by substitution

F (u; v; w ) = f (x (u; v; w ); y (u; v; w ); z (u; v; w )) : (4.30)

2. Finding the modulus of the Jacobian

@x @y @z
@ (x; y; z ) @u @u @u @ (x; y; z )
) V
 

@ (u; v; w ) du dv dw
@y
mod = mod @x @z
=
@ (u; v; w ) @v
@x
@v
@y
@v
@z
d

@w @w @w
(4.31)

3. Dening the new region R0

4. And then integrating ...

@ (x; y; z )
ZZZ
I= F (u; v; w ) @ (u; v; w ) du dv dw: (4.32)
R0

| Example #1

Q: x 2 + y 2 + z 2  a which occupies the

Find the mass of that part of the solid sphere

octant x  0, y  0, z  0, and which has volume density  = (x; y; z ) = xyz .

A: This is an obvious candidate for a transformation, but let's set up the integral rst

in Cartesian coordinates.

M = (x; y; z )dV ) M =
ZZZ
d xyz dx dy dz : (4.33)
R

Place the volume element in general position (x; y; z ), integrate into a rod over x , with
y and z constant, then integrate over y into a plate, then nally over z.
4/14 LECTURE 4. MULTIPLE INTEGRATION

z
Volume element
dxdydz

ymax

x
x max

x , variable x changes from xmin = 0 to xmax = a2 y 2 z 2.

p
Integrating into a rod along

Note that xmax is not 1. The extreme position for the volume element is on the surface
of the sphere, so xmax + y + z = a .
2 2 2 2

Integrating the rod along y , y changes from 0 to ymax , where (x = 0) + ymax + z = a ,

p2 2
2 2 2 2

so ymax = a z.
The plate is integrated over z going from 0 to a.

Thus the mass is

Z a Z pa 2 z2 Z pa 2 y2 z2
M = z y x dx dy dz (4.34)

pa
z =0 y =0 x =0
a Z pa 2 z2
1
Z
2 y2 z2
= z y x 2
y z

d d (4.35)
2 z =0 y =0
0

a Z pa 2 z2
1
Z
=
2
z y ( a 2 y 2 z 2 ) dy dz (4.36)
z =0 y =0
a Z pa 2 z2
1
Z
=
2
z y ( a 2 z 2 ) y 3 dy dz (4.37)
z =0 y =0
p
1 a
y 2
y 4 a2 z 2
Z 
=
2
z 2
(a 2 z 2) 4
d z (4.38)
z =0 y =0
(4.39)
4.7. THE ROLE OF MOD[JACOBIAN] 4/15

1 a 1 2 1 2
Z  
M = z
2 z =0 2
(a z )
2 2
4
(a z 2 ) 2 dz (4.40)

1 a
Z
=
8 z =0
z (a 2 z 2 )2 d z (4.41)

1 a
Z
=
8 z =0
za 4
2z 3 a2 + z 5 dz (4.42)

1 1 2 4 1 4 2 1 6a

=
8 2
z a 2z a + 6z (4.43)
z =0
1 6
=
48
a (4.44)

We will return to perform this integral in spherical polar coordinates after reviewing the

standard transformations introduced in Lecture 3.

4.7 The role of mod[Jacobian]

Looking at the double and triple integrals

@ (x; y ) @ (x; y; z )
ZZ ZZZ
I= F (u; v ) @ (u; v ) du dv and I= F (u; v; w ) @ (u; v; w ) du dv dw
R0
R0

(4.45)

will remind you that inserting the modulus of the Jacobian was the method of turning

u v
d d into a proper area, and d d d u v w into a proper volume. That is in our curvilinear

coordinate system

@ (x; y ) @ (x; y; z )
d A = @ (u; v ) du dv and d V =
@ (u; v; w ) du dv dw : (4.46)

In plane, cylindrical and spherical polars, although the lines of constant r; , etc, are
not straight, they always meet each other at right angles, and it turns out then to be

quite easy to visualise the area and volume elements. Being familiar with them will help

you dene regions of integration after transformations.

4.8 Standard transformations revisited

Here we specialize the general theory to look at the shapes of the area and volume

elements generated by the standard transformations.

1. 2D Cartesian to plane polars. r which is always positive so that

The Jacobian is

its modulus is r . Thus the element of area is dA = r dr d. This is readily seen from

Fig. 4.9(a).
4/16 LECTURE 4. MULTIPLE INTEGRATION

2. 3D Cartesian to cylindrical polars. The Jacobian is r,

r > 0 so is the
and, as

modulus of the Jacobian. Thus the volume element of area is dV = r dr ddz . Again

this is evident from the diagram Fig. 4.9(b). Note also that the volume element is

dimensionally correct.

3. 3D Cartesian to spherical polars.

The Jacobian is r 2 sin  (see Lecture 3). Now theta ranges from 0 to  , so the Jacobian
is always positive. Thus the element of area is d V = r sin  dr d d.
2

r
One side of the volume element is d , the second is r d , and the third is r sin d.
Note that the third side is generated by swing the arm of length r sin  through the

change in azimuthal angle d . 

See Fig. 4.9(c).

| Example: Mass of octant, now in Spherical polars

Recall the problem to nd the mass of the positive octant of a sphere with density xyz .
We convert the integral to spherical polars as follows:

M = fxy z g x y z
RRR
d d d (4.47)
R
@ (x; y; z )
= f(r sin  cos )(r sin  sin )(r cos )g mod
@ (r; ; ) dr dd (4.48)
RRR
R

= r (sin2  cos )(sin  cos ) r 2 sin dr dd

RRR  3
(4.49)
R
= r (sin3  cos )(sin  cos )dr dd
RRR 5
(4.50)
R
Z a Z =2 Z =2
= r dr
5
sin  cos d
3
sin  cos d (4.51)
r =0  =0 =0

1 6a 1 4 =2 1 2 =2
  
= r
6 r =0 4
sin 
2
sin  (4.52)
 =0 =0
1 6
=
48
a (4.53)

Notice that the region of integration is now dened by limits of integration which are

all constants  which means that the integration is separable into the product of three

single-variable integrations.
4.8. STANDARD TRANSFORMATIONS REVISITED 4/17

y d A =r d r d ϕ

dϕ dr
r ϕ

x
(a)

r dϕ

dz
z
r
dr y

ϕ
dϕ
x
(b)

z
r sin θ d ϕ
θ dr
dθ
r sin
θ

r dθ

dϕ y
x ϕ
(c)

Figure 4.9: Elements of (a) Area in plane polars: dA = r dr d. (b) Volume in cylindrical polars.

dV = r dr ddz . (c) Volume in spherical polars. dV = r 2 sin  dr; d d

4/18 LECTURE 4. MULTIPLE INTEGRATION

| Example

a and, as shown in the diagram, is lled to a

A spherical settling tank has radius
0
Q:

depth 3a=2 with a uid suspension whose volume density is

a (3a=2 z ) where z is
measured upwards from the bottom of the tank. Calculate the mass of uid suspension

in the tank.

z In general posn
at r, ϕ ,z

z=3a/2
r max
a
a r max

z-a
a

z=0

Figure 4.10:

A: The spherical tank pulls you towards spherical polars, but the dependence of the

density on z alone should be enough to make you realize that cylindrical polars are

easier!

So

0 0
(3a=2 z )r dr ddz ) M =
ZZZ
dM = (3a=2 z )r dr ddz :
a a R
(4.54)

Now think about the region and the limits this places on the integrals.

 z =3a=2 =2 r =rmax

Z Z Z
M = a0 (3a=2 z ) r dr d dz (4.55)
z =0 =0 r =0

The only limit requiring thought is rmax. From the inset diagram it is obvious that

rmax
2
= a2 (z a)2 = (2az z 2) : (4.56)
4.9. THE DEFINITIONS OF SEVERAL INTEGRAL PROPERTIES 4/19

p
0 z =3a=2 =2 r = 2az z 2
Z Z Z
M =
a (3a=2 z) d  r dr d z
z =0 =0 r =0

0 z =3a=2 1  2 p2az z
Z
2
2
=
a z =0 (3a=2 z ) 2 r 0 dz

0 z =3a=2
Z
= a (3a=2 z )(2az z 2 )dz
z =0
0 a
Z z =3a=2 
7

= a z 3 2 z 2 + 3 a 2 z dz
z =0

0 1 4 7a 3 3a2 2 z =3a=2

=  a 4z 6
z + 2z
 z=0
9a 1 9a 7a 3a 3a 2
  2 
= 0 4 4 4 6 2
+
2
9a 3 9 7 3 9a 3 5 3 45
     
= 0 4 16 4 + 2 = 0 4 16 = 0a 64

It is encouraging that this is (i) positive, and (ii) dimensionally correct.

4.9 The denitions of several integral properties

It is important to be able to set up integrals from rst principles. You should focus rst

on dening the d(Quantity)  the total quantity is then simply the integral over the

region.

4.9.1 In 2D ...

Consider a lamina of uniform thickness and mass per unit area (x; y ) occupying the
region R in the
RR xy plane. Then
(1) Area is
RRR dx dy .
(2) Mass is
R
(x; y )dx dy .
y -axis is
(RRx; y ) x 2dx dy .
RR
(3) Moment of inertia about the
R
(4) Polar moment of inertia about the origin is
R
(x; y ) (x 2 + y 2)dx dy .
x -coordinate is x dx dRRy= dx dy .
RR RR
(5) Centroid

x -coordinate is (x; y ) x dx dy= (x; y ) dx dy .

RR
(6) Centre of mass
4/20 LECTURE 4. MULTIPLE INTEGRATION

4.9.2 In 3D ... and in more detail ...

V = dx dy dz ) V
x y z =
RRR
(1) Volume: d d d d
R

(2) Mass: Given a volume density function (x; y; z ),

) M=
ZZZ
d M = (x; y; z ) dx dy dz (x; y; z ) dx dy dz (4.57)

x, y z. x = (1=V ) x dx dy dz
RRR
(3) Centroid: Average value of or For example
R

(4) Centre of Mass: The x-coordinate is

1
ZZZ
xCoM = M x(x; y; z ) dx dy dz (4.58)
R

(5) Moment of inertia about an axis:

The value depends on the axis of rotation. Supposing this to the be the z -axis, the

moment of inertia is

) Jz =
ZZZ
d z J =r 2
d M = (x 2
+ y ) dM
2
(x 2 + y 2 )(x; y; z )dx dy dz : (4.59)
R

About x , the value is

ZZZ
Jx = (y 2 + z 2 )(x; y; z )dx dy dz : (4.60)
R

| Example

Q: Derive the moment of inertia Jx about the x-axis of the solid cylinder x 2 + y 2  a2
bounded by z = 0 and z = b. Assume uniform volume density .

A: The moment of inertia of a little element d d d x y z about the x -axis is

d x J = (y 2 + z 2 )dx dy dz (4.61)

So

Jx =  ( y 2 + z 2 ) dx dy dz :
RRR
(4.62)
R

The geometry suggests transforming to cylindrical polar coordinates. There are again

three tasks in making the transformation: (i) rewrite the function, (ii) derive the Mod
4.9. THE DEFINITIONS OF SEVERAL INTEGRAL PROPERTIES 4/21

a
b

Figure 4.11:

Jacobian, (iii) rewrite the region. (i) The transformation is given by x = r cos ,
y = r sin , z = z . Thus the function (y 2 + z 2 ) is (r 2 sin 2
 + z 2 ). (ii) The Jacobian is

cos  r sin  0
sin  r cos  0 =r
0 0 1

As r > 0 always, jr j = r . (iii) The region is 0r  a 0    2

, and 0z b .

So

Z a Z 2 Z b
Jx =  rdr d (r 2 sin2  + z 2 )dz (4.63)
r =0 =0 z =0
Z a Z 2
=  rdr (br 2 sin2  + b3 =3)d (4.64)
r =0 =0

Now
2 2
1
Z Z
sin d =
2
(1 cos 2)d = 
0 2 0

so that

a
b3
Z
Jx =  rdr (r b + 3 2) 2
(4.65)
r =0
a b b32a2
 4 
=  4
+
6
(4.66)

a2b
=  12
(3a2 + 4b2 ): (4.67)
4/22 LECTURE 4. MULTIPLE INTEGRATION

4.10 Volumes from double integrals

Sometimes volumes appear to arise from double integration. Here are two examples.

But don't be deceived  in each case you have performed the third integral behind

the scenes because it was trivial.

4.10.1 A volume of revolution: a double integral

Suppose a closed curve C enclosing a region R lying in the positive quadrant of the

xy plane is rotated about the x -axis. It sweeps out a volume known as a volume of

revolution.

Figure 4.12: Volume of revolution

To calculate the volume of the v.o.r. rst calculate the element of volume created by

rotation of an area element dA about the x -axis. This is obviously:

dV = 2ydA : (4.68)

The total volume of revolution is thus:

V = 2ydA = 2ydxdy :
RR RR
(4.69)
R R

But the y coordinate of the centroid of the region R is given as

ydxdy ydxdy
RR RR
y = RRR = R
A
dxdy
(4.70)
R

where A is the area of the region. In other words we arrive at the intuitive result that

V = 2Ay (4.71)

4.10.2 A volume under surface: a double integral

Suppose you have a surface z = f (x; y ), and want to know the volume enclosed by a
region R on the x = y = 0 plane as it rises up to hit the surface. The volume formed
4.10. VOLUMES FROM DOUBLE INTEGRALS 4/23

by an element of area dA = dxdy placed in general position (x; y ) on the x =y =0

plane and the surface is dV =f (x; y )dA, and so the total volume is simply
ZZ
V = f (x; y ) dxdy: (4.72)
R

| Example

Q: Find the volume bounded by the cylinder x 2 + y 2 = 4 and the planes y + z = 4;

z= 1.

Figure 4.13: Volume under a surface

A: This example has two surfaces, but it is easy enough to combine the volumes by

thinking about limits. The element dxdy is at (x; y ). The upper surface is a z = (4 y )
and the lower at z = 1, so the elemental dV = (4 y + 1)dxdy . Then integrate
over the circle by changing to plane polars ...

ZZ Z 2 Z 2
V = (5 y ) d x dy = (5 r sin )r dr d (4.73)
Circle =0 r =0
2
r2 r =2 2
Z  Z
= 5r sin  d  = [10 2 sin ] d (4.74)
=0 2 r =0 =0

= [10 + 2 cos j20 = 20 : (4.75)

(It is easy to check this example. Because it is a plane slicing through the cylinder, we

could take the average z

z = 4, for the top surface.
value, So the total volume should

be (4 + 1)   2 which is indeed 20 . Hurrah.)

2