1
Partial Dierentiation
Let z = f (x, y) be a real valued function of two independent variables, x and y.
We dene the partial derivative of f with respect to (wrt) x as:
[
f (x + x, y) f (x, y)
f
= lim
x0
x
x
where in this limit y is assumed to have a xed value.
Similarly, we dene the partial derivative of f wrt y as
[
f
f (x, y + y) f (x, y)
= lim
y y0
y
where, this time, x is assumed to have a xed value.
Example 1.1
Find f /x and f /y for f (x, y) = x3 + y 2 .
Solution 1.1
f
f
= 3x2 ,
= 2y
x
y
Example 1.2
Find f /x and f /y for f (x, y) = x tan1 y + x3 y + y 3/2 .
Solution 1.2
f
= tan1 y + 3x2 y
x
f
x
3
=
+ x3 + y 1/2
2
y
1+y
2
Note: All the standard rules of ordinary dierentiation (eg product rule, quotient rule etc) apply
unchanged.
Example 1.3
If f (x, y) = ey ln(x + y) then nd f /y.
Solution 1.3
f
ey
y
= e ln(x + y) +
y
x+y
r
y
x
Figure 1:
Example 1.4
Consider the transformation to plane-polar coordinates: From Fig. 1, we have
x = r cos ,
y = r sin
r 2 = x2 + y 2 ,
tan =
y
x
We could, for example, write x as a function of (r, ), (r, y) or (, y):
x = r cos , x =
r2 y 2 , x = y cot
From these we can compute the following partial derivatives in turn:
x
x
r
x
r
= cos ,
= 2
= cot
= = sec ,
2
r
r
x
y
r y
Note: When computing any partial derivative, all the independent variables in the expression
being dierentiated - except the one in the dierentiation - are assumed to be held constant!!
Question: Why do we get two dierent results for x/r above? Well, in the rst case is
being held constant, and in the second case, y is being held constant.
1.1
Higher Order Derivatives
We mean, for example,
f
x
2f
x2
f
2f
=
y y
y 2
(
)
f
2f
=
x y
xy
(
)
f
2f
=
y x
yx
Example 1.5
f (x, y) = x3 + y 2
f
2f
= 3x2 ,
= 6x
x
x2
f
2f
= 2y,
=2
y
y 2
2f
=
(2y) = 0
xy
x
2f
( 2)
=
3x = 0
yx
y
Example 1.6
f (x, y) = x tan1 y + x3 y + y 3/2
f
f
x
3
= tan1 y + 3x2 y,
=
+ x3 + y 1/2
2
x
y
1+y
2
2
2
f
f
2xy
3
= 6xy,
=
+ y 1/2
2
2
2
2
x
y
(1 + y )
4
2
2
f
1
f
1
=
+ 3x2 ,
=
+ 3x2
2
yx
1+y
xy
1 + y2
Note: For all well-behaved functions, then
2f
2f
=
xy
yx
can be assumed true - all functions considered on this course are well-behaved.
Example 1.7
Given f (x, y) = x2 tan1 (y/x) nd fxy at (1, 1).
Solution 1.7
Note that fxy is short notation for 2 f /xy.
f
y
Hence
( )
1
y
= x
1 + (y/x)2 y x
x4 1
x3
= 2
=
x + y2 x
x2 + y 2
2
2f
=
xy
x
f
y
3x2
2x4
x2 + y 2 (x2 + y 2 )2
Thus, at (1, 1), we nd
2f
3 2
= =1
xy
2 4
1.2
Notes
A common notation is dened as follows:
f
f
2f
2f
fx
, fy =
, fxx =
,
f
=
, etc
xy
x
y
x2
xy
Partial dierentiation can be extended to functions of more than two variables in a natural way.
Example 1.8
Suppose f (x, y, z) = x3 + x2 z 4 + xy 5 , then
f
f
f
= 3x2 + 2xz 4 + y 5 ,
= 5xy 4 ,
= 4x2 z 3
x
y
z
1.3
The Ordinary Chain Rule
The ordinary chain rule can be stated as follows: If y = f (x) where x = g(t) then
dy
dy dx
=
dt
dx dt
1.4
The Generalized Chain Rule
The ordinary chain rule extends in an obvious way to functions of two or more variables.
Thus, if z = f (x, y) where x = g(t) and y = h(t) then
dz
z dx z dy
=
+
dt
x dt
y dt
Example 1.9
Thus, if z = sin(xy 2 ) where x = et and y = t3 then
z dx z dy
dz
=
+
dt
x dt
y dt
2
2 t
= y cos(xy )e + 2xy cos(xy 2 )3t2
= t6 cos(t6 et )et + 6et t5 cos(t6 et )
= (t + 6)t5 et cos(t6 et )
4
1.5
A Special Case of the Generalized Chain Rule
For the general case, the problem is: if z = f (x, y) where x g(t) and y h(t) then nd dz/dt,
and the solution is:
z dx z dy
dz
=
+
dt
x dt
y dt
Now suppose that x = t: then the problem becomes: if z = f (x, y) where y h(x), then nd
dz/dx, and the solution is:
dz
z z dy
=
+
dx
x y dx
Example 1.10
If z = x2 y + 1/y and y = ln x then nd dz/dx.
Solution
(
dz
1
= 2xy + x2 2
dx
y
1.6
1
1
= 2xy + x 2
x
xy
Implicit Dierentiation
A further application of the above is to nd dy/dx when y is dened as an implicit function of
x: That is, nd dy/dx when f (x, y) = 0. In this case, we write z f (x, y) so that, rstly
dz
z z dy
=
+
dx
x y dx
and then we note that, since z = 0 then we must have dz/dx = 0 also so that the above gives
directly:
(
) (
)
dy
f
f
=
/
dx
x
y
Example 1.11
Given f (x, y) x2 + 3xy + 4y 2 = 0 nd dy/dx.
Solution
dy
= (2x + 3y)/(3x + 8y).
dx
1.7
Small Error Estimates
The total derivative formula can be used to estimate errors in calculations: Clearly, if z z(x, y)
then
z
z
x +
y
(1)
z
x
y
where x and y are small changes (which can be errors) in x and y, and z is the corresponding
change (or error) in z.
Example 1.12
The value of z is to be calculated from
z=
xy
for x = 4 and y = 9 and where the corresponding errors in x and y are 0.1 and 0.2 respectively.
What is the corresponding maximum modulus for the error in z?
Solution
We have
z
1
=
x
2
y
3
= ,
x
4
z
1
=
y
2
x
1
=
y
3
Consequently, (1) gives
3
1
x + y
4
3
3
1
|z | | x + y|
4
3
1
3
1
3
|x | + |y | = (0.1) + (0.2) 0.14
|z |
4
3
4
3
z
That is, |z | 0.14.
Notes:
We have used |a + b| |a| + |b|.
The three-variable extension is
f =
f
f
f
x +
y +
z
x
y
z
The chain-rule is mainly used when we change form one independent variable to another as in coordinate transormation:
Example 1.13
If f f (x, y) and x = re and y = re then show that
f
f
f
= r
+
x
r
f
f
f
2y
= r
y
r
2x
Solution
f
r
f x f y
f
f
+
= e
+ e
x r
y r
x
y
f
r
= x
f
f
+y
x
y
(2)
Similarly
f
f x f y
f
f
+
= re
re
x
y
x
y
= x
f
f
y
x
y
(3)
Adding and subtracting (2) and (3) gives, respectively:
f
f
f
= r
+
x
r
f
f
f
2y
= r
y
r
2x
Alternative Solution
We could have rearranged x = re and y = re as
x
y
( )
= e2
x
1
= ln
2
y
xy = r2 r = (xy)1/2
and then used the relations
f
f r f
=
+
x
r x x
etc.
Example 1.14
By putting u = xy and v = x2 /y show that the partial dierential equation
x
z
z
+ 2y
= 3x3 cos(xy)
x
y
can be written in the form
z
= v cos u.
u
Hence, nd the general solution of the equation.
7
Solution
We have
z u z v
z
=
+
x
u x v x
z 2x z
= y
+
u
y v
z
z 2x2 z
z
z
x
= xy
+
=u
+ 2v
x
u
y v
u
v
Similarly,
z
z u z v
=
+
y
u y v y
z x2 z
= x
u y 2 v
z
z 2x2 z
z
z
2y
= 2xy
= 2u
2v
y
u
y v
u
v
Adding these together gives, nally,
x
z
z
z
+ 2y
= 3u
= 3x3 cos(xy)
x
y
u
z
3u
= 3uv cos u
u
z
= v cos u.
u
This pde can now be solved as follows: Integrate wrt to u:
z = v sin u + f (v)
( )
x2
x2
z(x, y) =
sin(xy) + f
y
y
