You are on page 1of 39

Chapter 5 – Function of Several Variables

Subject: Real Analysis Level: M.Sc.


Source: Syed Gul Shah (Chairman, Department of Mathematics, US Sargodha)
Collected & Composed by: Atiq ur Rehman (atiq@mathcity.tk), http://www.mathcity.tk

v Introduction
There is the basic difference between the calculus of functions of one variable and
the calculus of functions of two variables. But there is a slight difference between the
calculus of two variable and the calculus of functions of three, four or of many
variables. Therefore we shall emphasise mainly on the study of functions of two
variables.
v Function of two variables
If to each point ( x, y ) of a certain part of xy - plane, there is assigned a real number
z , then z is known to be a function of two variable x and y .
e.g. z = x 2 - y 2 , z = x 2 + y 2 , z = xy etc.
v Neighbourhood (nhood)
A neighbourhood of radius d of a point ( x0 , y0 ) of the xy - plane is the set of
points which lies inside a circle with centre at ( x0 , y0 ) and has radius d .
Nd ( x0 , y0 ) = ( x - x0 ) + ( y - y0 ) < d 2
2 2

Similarly, a nhood of a radius d of a point ( x0 , y0 , z0 ) of a space is a sphere with


centre at ( x0 , y0 , z0 ) and radius d .
Nd ( x0 , y0 , z0 ) = ( x - x0 ) + ( y - y0 ) + ( z - z0 ) < d 2
2 2 2

This definition can be extended to the definition of a nhood of a point of a space of


any dimension.
v Open Set
A set is known to be open set if each point ( x0 , y0 ) of the set has a nhood which
totally lies inside the set.
v Domain
A set D which is not empty and open is known to be a domain, if any two points
of the set can be joined by a broken line which lies completely with in D .
v Region
A domain D is known to be a region if some or all of the boundary points are
contained in D .
v Closed Region
A region is known to be closed if it contains all the boundary points.
e.g. i) x 2 + y 2 < 1 (Domain) ii) x y < 1 (Domain)
x + y = 1 (Boundary)
2 2
xy=2 (Boundary)
x2 + y2 £ 1 (Closed region) xy £ 1 (Closed Region)
v Limit & Continuity
Let z = f ( x, y ) be a function of two variables defined in a domain D . Suppose
there is a point ( x0 , y0 ) Î D or is a boundary point then
lim f ( x, y ) = c
x® x0
y ® y0

It means that given e > 0 $ a d > 0 such that


f ( x, y ) - c < e whenever ( x, y ) - ( x0 , y0 ) < d " ( x, y ) Î Nd ( x0 , y0 )
2 Chap 5 – Function of several variables

If limit of a function is equal to actual value of function then f is said to be


continuous at the point ( x0 , y0 )
lim f ( x, y ) = f ( x0 , y0 )
x ® x0
y ® y0

If f is continuous at every point of D , then f is said to be continuous on D .


v Theorem
Let f ( x, y ) & g ( x, y ) be defined in a domain D and suppose that
lim f ( x, y ) = u1 & lim g ( x, y ) = v1
x ® x0 x® x0
y ® y0 y ® y0

a) then (i) lim [ f ( x, y ) + g ( x, y )] = u1 + v1


x ® x0
y ® y0

(ii) lim [ f ( x, y ) × g ( x, y )] = u1v1


x ® x0
y ® y0

f ( x, y ) u1
(iii) lim =
x ® x0 g ( x, y ) v1
y ® y0

b) If f ( x, y ) & g ( x, y ) are defined in D , then


lim f ( x, y ) = f ( x0 , y0 ) & lim g ( x, y ) = g ( x0 , y0 )
x ® x0 x ® x0
y ® y0 y ® y0

i.e. f ( x, y ) , g ( x, y ) are continuous at ( x0 , y0 ) then so are the functions


f ( x, y )
f ( x, y ) + g ( x, y ) , f ( x, y ) g ( x, y ) and , provided g ( x, y ) ¹ 0 .
g ( x, y )
Proof
a) (i) Q lim f ( x, y ) = u1 , lim g ( x, y ) = v1
x ® x0 x® x0
y ® y0 y ® y0

e
\ given > 0 $ a d1 , d 2 > 0 such that
2
e
f ( x, y ) - u1 < " ( x, y ) Î Nd1 ( x0 , y0 )
2
e
& g ( x, y ) - v1 < " ( x, y ) Î Nd 2 ( x0 , y0 )
2
then [ f ( x, y ) + g ( x, y )] - [u1 + v1 ] = [ f ( x, y ) - u1 ] + [ g ( x, y ) - v1 ]
£ f ( x, y ) - u1 + g ( x, y ) - v1
e e
< + " ( x, y ) Î Nd ( x0 , y0 )
2 2
where d = min (d 1, d 2 )
Which show that
lim [ f ( x, y ) + g ( x, y )] = u1 + v1
x ® x0
y ® y0

(ii) f ( x, y ) × g ( x, y ) - u1v1 = f ( x, y ) × g ( x, y ) - u1 g ( x, y ) + u1g ( x, y ) - u1v1


= g ( x, y ) [ f ( x, y ) - u1 ] + u1 [ g ( x, y ) - v1 ]
£ g ( x, y ) [ f ( x, y ) - u1 ] + u1 [ g ( x, y ) - v1 ]
e e
< g ( x, y ) + u1 = e1 " ( x, y ) Î Nd ( x0 , y0 )
2 2
Chap 5 – Function of several variables 3

Þ lim f ( x, y ) × g ( x, y ) = u1v1
x ® x0
y ® y0

1 1
iii) We prove that lim =
x® x0 g ( x, y ) v1
y ® y0

1 1 v - g ( x, y )
- = 1
g ( x, y ) v1 v1 g ( x, y )
g ( x, y ) - v1 e
= < 2
v1 g ( x, y ) v1 g ( x, y )
e e
< 2 < 2
v1 g ( x, y ) - v1 + v1 v1 ( g ( x, y ) - v1 + v1 )
e
< 2 = e1 " ( x, y ) Î Nd 2 ( x0 , y0 )
v1 ( e +v
2 1 )
1 1
Þ lim =
x ® x0 g ( x, y ) v1
y ® y0

Q Þ lim f ( x, y ) = u1 & lim g ( x, y ) = v1


x ® x0 x® x0
y ® y0 y ® y0

By (ii) of theorem
1 f ( x, y ) u1
lim f ( x, y ) × = lim =
x ® x0 g ( x, y ) x® x0 g ( x, y ) v1
y ® y0 y ® y0

b) Since it is given that the limiting values are the same as the actual values of the
f ( x, y )
functions f ( x, y ) + g ( x, y ) , f ( x, y ) × g ( x, y ) and at the point ( x0 , y0 )
g ( x, y )
therefore these function are continuous on ( x0 , y0 ) .
Note
It is to be noted that there is a difference between lim f ( x, y ) and lim f ( x, y )
( x , y ) ®( a ,b ) x ®a
y ®b

i.e.
x ®a
y ®b
(
lim f ( x, y ) = lim lim f ( x, y )
y ®b x ®a
) or
x ®a
y ®b
x ®a (
lim f ( x, y ) = lim lim f ( x, y )
y ®b )
Obviously in the two cases limits are taken first w.r.t one variable and then w.r.t
other variable. These limits are called the repeated limits. Since these are taken along
the special path, therefore repeated limits are the special cases of limits.
lim f ( x, y ) exists if and only if limiting vales are not depend upon any path
( x , y ) ®( a ,b )

along which ( x, y ) ® (a, b) .


v Example
Consider f : ¡ 2 ® ¡ given by
ì x2 y2
ï , ( x, y ) ¹ (0,0)
f ( x, y ) = í x 4 + y 4
îï 0 , ( x, y ) = (0,0)
Now lim f ( x, y ) = lim f ( x, y ) = 0
x®0 y ®0
y ®0 x ®0
However along the straight line y = mx , we have
4 Chap 5 – Function of several variables

m4
lim f ( x, y ) =
( x , y ) ®(0,0) 1 + m4
which is different for different values of m . Hence lim f ( x, y ) does not exist.
( x , y ) ®(0,0)

v Example
Consider f : ¡ 2 ® ¡ given by
ì x 2 cos x - y 2 cos y
ï , ( x, y ) ¹ 0
f ( x, y ) = í x2 + y 2
ïî 0 , ( x, y ) = 0

then lim élim f ( x, y ) ù = lim cos x = 1


x ®0 ê
ë y ®0 úû x®0
and lim élim f ( x, y ) ù = lim(- cos y ) = -1
y ®0 ë x ®0 û y ®0
Þ lim f ( x, y ) does not exist.
x ®0
y ®0

v Example
Consider f : ¡ 2 ® ¡ given by
ì sin( x 2 + y 2 )
ï( x + y ) , ( x, y ) ¹ (0,0)
f ( x, y ) = í x2 + y2
ïî 0 , ( x, y ) = (0,0)
sin x
Use < 1 to get
x
f ( x, y ) - 0 £ x + y < x + y
e e
Thus f ( x, y ) - 0 < e whenever x < , y <
2 2
e
Take d = ,
2
It follows that for given e > 0 , we can find d > 0 such that
f ( x, y ) - f (0,0) < e whenever ( x - 0)2 + ( y - 0)2 < d
i.e. " ( x, y ) Î Nd (0,0)
Limit of the function at (0,0) is equal to actual value of function at (0,0) .
Hence f is continuous at (0,0) .
v Partial Derivative
Let z = f ( x, y ) be defined in a domain D of xy -plane and take ( x0 , y0 ) Î D ,
then f ( x, y0 ) is a function of x alone and its derivative may exist. If it exists then its
value at ( x0 , y0 ) is known to be the partial derivative of f ( x, y ) at ( x0 , y0 ) and is
¶f ¶z
denoted as or
¶x ( x0 , y0 ) ¶x ( x0 , y0 )
The other notations are z x , f x , f1 .
¶f f ( x + Dx, y0 ) - f ( x, y0 )
= lim
¶x ( x0 , y0 ) Dx®0 Dx
¶f
We can define in the same manner.
¶y
Chap 5 – Function of several variables 5

v Geometrical Interpretation
z = f ( x, y ) represents a surface in space. y = y0 is a plane. z = ( x, y0 ) is the curve
¶f
which arises when y = y0 cuts the surface z = f ( x, y ) . Thus denotes the
¶x ( x0 , y0 )
¶f
slope of tangent to the curve z = f ( x, y0 ) at x = x0 . Similarly denotes the
¶y ( x0 , y0 )
slope of the tangent to the curve z = f ( x0 , y ) at y = y0 .
If the point ( x0 , y0 ) varies, then f x & f y are themselves functions of x & y .
In the case of functions of more than three variables it is necessary to indicate the
variable held constant during the process of differentiation as a suffix to avoid the
confusion.
æ ¶z ö æ ¶z ö
For example, z = f ( x, y, u , v) , then partial derivatives are written as ç ÷ , ç ÷
è ¶u ø x è ¶y øv
and so on. We take an example: x = u + v , y = u - v
æ ¶x ö æ ¶x ö æ ¶y ö æ ¶y ö
ç ÷ = 1 , ç ÷ = 1 , ç ÷ = 1 , ç ÷ = -1
è ¶u øv è ¶v øu è ¶u øv è ¶v øu
Also x + y = 2u and x = 2u - y , then
æ ¶x ö æ ¶x ö
ç ÷ = 2 & ç y ÷ = -1 and so on.
è ¶u ø y è ¶ øu
v Total Differential
In the case of partial derivative we have considered increments Dx & Dy
separately.
Now take ( x, y ) & ( x + Dx, y + Dy ) two points in the domain of definition of z
then if ( z + Dz ) correspond to the point ( x + Dx, y + Dy ) we have
Dz = f ( x + Dx, y + Dy ) - f ( x, y )
If the increment Dz can be expressed as
Dz = aDx + bDy + e1Dx + e 2Dy
and e1 , e 2 ® 0 as Dx, Dy ® 0 , then aDx + bDy is known to be the total differential
of z denoted by dz , and we write
Dz = dz + e1Dx + e 2 Dy
In case when z is differentiable function dz gives very close approximation of Dz .
v Theorem
If z = f ( x, y ) has a total differential at a point ( x, y ) Î D , then
¶z ¶z
a= & b= .
¶x ¶y
Proof
We have
Dz = dz + e1Dx + e 2 Dy where e1 , e 2 ® 0 as Dx, Dy ® 0
Let us suppose that Dy = 0
then Dz = aDx + e1Dx
Taking the limit as Dx ® 0
¶z
=a
¶x
¶z
Similarly we can get =b.
¶y
6 Chap 5 – Function of several variables

v Theorem (Fundamental Lemma)


If z = f ( x, y ) has a continuous first order partial derivative in D then z has total
¶z ¶z
differential dz = Dx + Dy at every point ( x, y ) Î D .
¶x ¶y
Proof
Take a point ( x, y ) as a fixed point in the domain D . Suppose x changes alone.
Then we have
Dz = f ( x + Dx, y ) - f ( x, y )
= f x ( x1 , y )Dx ( x < x1 < x + Dx ) ( It is by M. V. Theorem)
Q f x is continuous
\ e1 = f x ( x1, y ) - f x ( x, y ) ® 0 as Dx ® 0
Þ f ( x + Dx, y ) - f ( x, y ) = f x ( x, y )Dx + e1Dx ………. (i)
Now if both x, y changes, we obtain a change Dz in z as
Dz = f ( x + Dx, y + Dy ) - f ( x, y )
= [ f ( x + Dx, y ) - f ( x, y ) ] + [ f ( x + Dx, y + Dy ) - f ( x + Dx, y )]
that is we have expressed Dz as the sum of terms representing the effect of a
change in x alone and subsequent change in y alone.
Now f ( x + Dx, y + Dy ) - f ( x + Dx, y ) = f y ( x + Dx, y1 ) Dy ( y < y1 < y + Dy )
(It is by use of M.V. theorem)
Q f y is given to be continuous
\ e 2 = f y ( x + Dx, y1 ) - f y ( x, y ) ® 0 as Dx, Dy ® 0
Þ f ( x + Dx, y + Dy ) - f ( x + Dx, y ) = f y ( x, y )Dy + e 2 Dy ……….. (ii)
Using (i) & (ii), we have
Dz = f x ( x, y )Dx + f y ( x, y )Dy + e1Dx + e 2Dy where e1 , e 2 ® 0 as Dx, Dy ® 0
which shows that the total differential dz of z exist & is given by
dz = f x ( x, y )Dx + f y ( x, y )Dy ………….. (iii)

Note
(a) For reasons to be explained later; Dx & Dy can be replaced by dx & dy in (iii).
¶z ¶z
Thus we have dz = dx + dy
¶x ¶y
Which is the customary way of writing the differential. The preceding analysis
extends at once to functions of three or more variables. For example, if
¶w ¶w ¶w ¶w
w = f ( x, y , u , v) , then dw = dx + dy + du + dv .
¶x ¶y ¶u ¶v
(b) In the following discussion, the function and their Ist order partial derivatives
will be considered to be continuous in their respective domain of definition.
Example
If z = x 2 - y 2 , then dz = 2 xdx - 2 ydy .
Example:
xy y x xy
If w = , then dw = dx + dy - 2 dz
z x z z
Chap 5 – Function of several variables 7

PROBLEMS

¶z ¶z
1) Evaluate and if
¶x ¶y
x ¶z y 2 - x2 ¶z -2 xy
a) z= 2 Ans: = , =
x + y2 ¶x ( ) ¶x ( )
2 2
x2 + y 2 x2 + y 2
¶z ¶z
b) z = x sin xy Ans: = sin xy + xy cos xy , = x 2 cos xy
¶x ¶y
¶z 3 x 2 + y 2 - 2 xz ¶z e x +2 y - y
c) x 3 + xy 2 - x 2 z + z 3 - 2 = 0 Ans: = , =
¶x x 2 - 3z 2 ¶y e x+2 y - y 2
2) Evaluate the indicated partial derivatives:
æ ¶u ö æ ¶v ö
a) ç ÷ and ç ÷ if u = x 2 - y 2 , v = x + 2 y
è ¶x ø y è ¶y ø x
æ ¶x ö æ ¶y ö æ ¶x ö æ ¶y ö 1
b) ç ÷ and ç ÷ if u = x - 2 y , v = u + 2 y Ans: ç ÷ = 1 , ç ÷ =
è ¶u ø y è ¶v øu è ¶u ø y è ¶v øu 2
3) Find the differentials of the following functions
x ydx - xdy
a) z = Ans:
y y2
xdx + ydy
b) z = log x 2 + y 2 Ans:
x2 + y2
æ yö - ydx + ydy
c) z = tan -1 ç ÷ Ans:
èxø x2 + y 2
1 -( xdx + ydy + zdz )
d) u= Ans: 3
x2 + y2 + z 2 ( x2 + y 2 + z 2 ) 2
4) If z = x 2 + 2 xy , find Dz in terms of Dx , Dy for x = 1 , y = 1 .
Ans: Dz = 4Dx + 2Dy + Dx 2 + 2DxDy , dz = 4Dx + 2Dy , dz = 4Dx + 2Dy .

…………………………………..
8 Chap 5 – Function of several variables

v Derivative and Differential of functions of functions


In the following discussion, the function and their first order partial derivatives
will be considered to be continuous in their respective domain of definitions.
v Theorem (Chain Rule I)
Let z = f ( x, y ) , x = g (t ) & y = h(t ) be defined in a domain D , then
dz ¶z dx ¶z dy
= × + ×
dt ¶x dt ¶y dt
Proof
Q z = f ( x, y ) , x = g (t ) , y = h(t ) are defined in D , are continuous and have Ist
order partial derivatives.
\ By using the fundamental lemma we have
¶z ¶z
Dz = Dx + Dy + e1Dx + e 2 Dy ………… (i)
¶x ¶y
where e1 , e 2 ® 0 as Dx, Dy ® 0
Also Dx = g (t + Dt ) - g (t )
Dy = h(t + Dt ) - h(t )
Dividing (i) by Dt , we get
Dz ¶z Dx ¶z Dy Dx Dy
= × + × + e1 + e2
Dt ¶x Dt ¶y Dt Dt Dt
Take the limit as Dt ® 0 , we get
dz ¶z dx ¶z dy
= × + × as desired.
dt ¶x dt ¶y dt
v Theorem (Chain Rule II)
Let z = f ( x, y ) , x = g (u , v) , y = h(u , v ) be defined in a domain D and have
continuous first order partial derivative in D , then
¶z ¶z ¶x ¶z ¶y
= × + ×
¶u ¶x ¶u ¶y ¶u
¶z ¶z ¶x ¶z ¶y
and = × + ×
¶v ¶x ¶v ¶y ¶v
Proof
Q the functions are continuous having first order partial derivatives in D ,
therefore by the fundamental lemma, we have
¶z ¶z
Dz = Dx + Dy + e1Dx + e 2 Dy ………… (i)
¶x ¶y
where Dx = g (u + Du , v ) - g (u , v ) , Dy = h(u + Du , v) - h(u , v)
and e1 , e 2 ® 0 as Dx, Dy ® 0 i.e. Du ® 0
Dividing (i) by Du throughout to have
Dz ¶z Dx ¶z Dy Dx Dy
= × + × + e1 + e2
Du ¶x Du ¶y Du Du Du
Taking the limit as Du ® 0 i.e. Dx, Dy ® 0 , we have
¶z ¶z ¶x ¶z ¶y
= × + ×
¶u ¶x ¶u ¶y ¶u
Similarly if Dx = g (u , v + Dv) - g (u , v)
Dy = h(u , v + Dv) - h(u , v)
Then dividing (i) by Dv throughout, we obtain
Chap 5 – Function of several variables 9

Dz ¶z Dx ¶z Dy Dx Dy
= × + × + e1 + e2
Dv ¶x Dv ¶y Dv Dv Dv
Taking the limit as Dv ® 0 , we have
¶z ¶z ¶x ¶z ¶y
= × + ×
¶v ¶x ¶v ¶y ¶v
v Note
We have proved in chain rule I, that if z = f ( x, y ) , x = g (t ) , y = h(t ) , then
dz ¶z dx ¶z dy
= × + × ……….. (i)
dt ¶x dt ¶y dt
The three functions of t considered here: x = g (t ) , y = h(t ) , z = f ( g (t ), h(t ) )
dx dy dz
have differentials dx = Dt , dy = Dt , dz = Dt .
dt dt dt
From (i) we conclude that
dz ¶z æ dx ö ¶z æ dy ö
Dt = ç Dt ÷ + ç Dt ÷
dt ¶x è dt ø ¶y è dt ø
¶z ¶z
Þ dz = dx + dy ………… (ii)
¶x ¶y
¶x ¶x
Similarly, dx = Du + Dv
¶u ¶v
¶y ¶y
dy = Du + Dv
¶u ¶v
¶z ¶z
dz = Du + Dv
¶u ¶v
are the corresponding differentials when z = f ( x, y ) , x = g (u , v) , y = h(u , v )
æ ¶z ¶x ¶z ¶y ö æ ¶z ¶x ¶z ¶y ö
Þ dz = ç × + × ÷ Du + ç × + × ÷ Dv
è ¶x ¶u ¶y ¶u ø è ¶x ¶v ¶y ¶v ø
¶z æ ¶x ¶x ö ¶z æ ¶y ¶y ö
= ç Du + Dv ÷ + ç Du + Dv ÷
¶x è ¶u ¶v ø ¶y è ¶u ¶v ø
¶z ¶z
= dx + dy
¶x ¶y
which is again (ii)
The generalization of this permits to conclude that:
The differential formula
¶z ¶z ¶z
dz = dx + dy + dt + ......
¶x ¶y ¶t
which holds when z = f ( x, y , t ,....) and dx = Dx , dy = Dy , dt = Dt ,…..., remain the
true when x, y , t ,..... , and hence z , are all functions of other independent variables
and dx, dy, dt ,....., dz are the corresponding differentials.
As a consequence we can conclude:
Any equation in differentials which is correct for one choice of independent
variables remains true for any other choice. Another way of saying this is that any
equation in differentials treats all variables on an equal basis.
1 3
Thus, if dz = 2dx - 3dy at a given point, then dx = dz + dy is the
2 2
corresponding differentials of x in terms of y and z .
10 Chap 5 – Function of several variables

v Example
x2 - 1 2 xy dx - ( x 2 - 1)dy
If z = , then dz =
y y2
¶z 2 x ¶z 1 - x 2
Hence = , =
¶x y ¶y y2
v Example
If r 2 = x 2 + y 2 , then rdr = xdx + ydy
æ ¶r ö x æ ¶r ö y æ ¶x ö r
and ç ÷ = , ç ÷ = , ç ÷ = , etc.
è ¶x ø y r è ¶y ø x r è ¶r ø y x
v Example
æ yö
If z = tan -1 ç ÷ ( x ¹ 0) , then
èxø
1 æ yö xdy - ydx
dz = .d ç ÷ = 2
æ yö èxø
2
x + y2
1+ ç ÷
èxø
and hence
¶z y ¶z x
=- 2 , =
¶x x + y2 ¶y x 2 + y 2

…………………………………..
Chap 5 – Function of several variables 11

v Implicit Function
If F ( x, y, z ) is a given function of x, y & z , then the equation F ( x, y , z ) = 0 is a
relation which may describe one or several functions z of x & y .
Thus if x 2 + y 2 + z 2 - 1 = 0 , then
z = 1 - x2 - y 2 or z = - 1 - x2 - y2
Where both functions being defined for x 2 + y 2 £ 1 . Either function is said to be
implicitly defined by the equation x 2 + y 2 + z 2 - 1 = 0 .
Similarly, an equation F ( x, y, z, w) = 0 may define one or more implicit functions
w of x, y , z . If two such equations are given;
F ( x, y , z , w) = 0 , G ( x, y , z , w) = 0 ,
It is in general possible (at least in theory) to reduce the equations by elimination to
the form
w = f ( x , y ) , z = g ( x, y )
i.e. to obtain two functions of two variables. In general, if m equations in n
unknown are given (m < n) , it is possible to solve for m of the variables in terms of
the remaining n - m variables; the number of dependent variables equals the number
of equations
v Example
If 3 x + 2 y + z + 2w = 0
2x + 3y - z - w = 0
then w = f ( x, y ) = -5 x - 5 y & z = g ( x, y ) = 7 x + 8 y
v Example
Suppose that the functions w = f ( x, y ) & z = g ( x, y ) are implicitly defined by
2 x 2 + y 2 + z 2 - zw = 0
x 2 + y 2 + 2 z 2 - 8 + zw = 0
Then taking the differentials, we obtain
4 xdx + 2 ydy + 2 zdz - wdz - zdw = 0 ……… (i)
wdz + zdw + 2 xdx + 2 ydy + 4 zdz = 0 ……… (ii)
Eliminate dw between (i ) and (ii ) to have
6 xdx + 4 ydy + 6 zdz = 0
x 2y
Þ dz = - dx - dy
z 3x
¶z x ¶z 2y
Þ =- , =-
¶x z ¶y 3z
Eliminating of dz from (i ) and (ii ) gives
6 x ( 2 z + w ) dx + 4 y ( z + w ) dy - 6 z 2 dw = 0
x ( 2 z + w) 2 y ( z + w)
Þ dw = dx + dy
z2 3z 2
¶w x (2 x + w) ¶w 2 y ( z + w)
= , =
¶x x 2
¶y z2
12 Chap 5 – Function of several variables

v Examples
Suppose that the functions w = f ( x, y ) & z = g ( x, y ) are implicitly define by
F ( x, y, z, w) = 0 and G ( x, y , z ) = 0 , then
Fx dx + Fy dy + Fz dz + Fwdw = 0
and Gx dx + Gy dy + Gz dz + Gwdw = 0
Þ Fz dz + Fwdw = - éë Fx dx + Fy dy ùû
and Gz dz + Gwdw = - éëGx dx + G y dy ùû
Then by crammer rule, we have
Fx dx + Fy dy Fw Fx Fw Fy Fw
-
Gx dx + G y dy Gw Gx Gw G y Gw
dz = =- dx - dy
Fz Fw Fz Fw Fz Fw
Gz Gw Gz Gw Gz Gw
Fx Fw Fy Fw
¶z G Gw ¶z G y Gw
Þ =- x , =-
¶x Fz Fw ¶y Fz Fw
Gz Gw Gz Gw
¶( F , G) ¶( F , G )
¶z ¶ ( x, w) ¶z ¶ ( y, w) ¶( F , G)
Þ =- , =- provided ¹0
¶x ¶( F , G) ¶y ¶( F , G ) ¶ ( z , w)
¶ ( z , w) ¶ ( z , w)
Similarly, we have
Fz Fx dx + Fy dy
Gz Gx dx + G y dy
dw = -
Fz Fw
Gz Gw
¶w ¶w
and we can find and in the same manner.
¶x ¶y
v Particular Cases
i) One equation in 2 unknowns i.e. F ( x, y ) = 0
Þ Fx dx + Fy dy = 0

Þ
dy
dx
F
=- x
Fy
(F y ¹ 0)

ii) One equation in 3 unknowns i.e. F ( x, y , z ) = 0


Fx dx + Fy dy + Fz dz = 0
¶z F ¶z F
Þ =- x , = - y ( Fz ¹ 0)
¶x Fz ¶y Fz
iii) 2 equations in 3 unknown
F ( x, y , z ) = 0 , G ( x , y , z ) = 0
¶( F , G ) ¶( F , G )
¶z ¶ ( y, x ) ¶z ¶ ( w, y )
=- , =-
¶x ¶( F , G ) ¶y ¶( F , G )
¶ ( y, z ) ¶ ( z , w)
………………………………
Chap 5 – Function of several variables 13

v Example
Find the partial derivatives w.r.t x & y , when
u + 2v - x 2 + y 2 = 0
2u - v - 2 xy = 0
Solution
Take the differentials
du + 2 dv - 2 x dx + 2 y dy = 0 ………… (i)
2 du - dv - 2 x dy - 2 y dx = 0 …………. (ii)
Eliminating dv between (i ) and (ii ) , we have
5du - (2 x + 4 y )dx + (2 y - 4 x )dy = 0
1 1
Þ du = ( 2 x + 4 y ) dx - ( 2 y - 4 x ) dy
5 5
¶u 1 ¶u 1
Þ = ( 2x + 4 y ) & = - (2 y - 4x )
¶x 5 ¶y 5
Eliminating du between (i ) and (ii ) , we get
5dv - ( 4 x - 2 y ) dx + ( 4 y + 2 x ) dy = 0
1 1
Þ dv = ( 4 x - 2 y ) dx - ( 4 y + 2 x ) dy
5 5
¶v 1 ¶v 1
Þ = (4x - 2 y ) & = - (4 y + 2x)
¶x 5 ¶y 5
v Question
Give that
2 x + y - 3 z - 2u = 0
& x + 2y + z + u = 0
æ ¶x ö æ ¶y ö æ ¶z ö æ ¶y ö
Find ç ÷ , ç ÷ , ç ÷ , ç ÷
è ¶y ø z è ¶x øu è ¶u ø x è ¶z ø x
Solution
Take the differentials
2 dx + dy - 3dz - 2du = 0 ………… (i)
dx + 2dy + dz + du = 0 ………...… (ii)
Eliminating du between (i) and (ii), we have
4 dx + 5dy - dz = 0 ………. (iii)
5 1
Þ dx = - dy + dz
4 4
æ ¶x ö 5
Þ ç ÷ =-
è ¶y ø z 4
From (iii), we have
5dy = dz - 4dx
1 4
Þ dy = dz - dx
5 5
æ ¶y ö 1
Þ ç ÷ =
è ¶z ø x 5
Eliminating dz between (i ) & (ii ) , we get
5dx + 7 dy + du = 0
14 Chap 5 – Function of several variables

5 1
Þ dy = - dx - du
7 7
æ ¶y ö 5
Þ ç ÷ =-
è ¶x øu 7
Now eliminating dy between (i ) & (ii ) , we get
-3dx - 5dz - 3du = 0
3 3
Þ dz = - dx - du
5 5
æ ¶z ö 3
Þ ç ÷ =-
è ¶u ø x 5
v Question
Given that
x 2 + y 2 + z 2 - u 2 + v 2 = 1 …………... (i)
x 2 - y 2 + z 2 + u 2 + 2v 2 = 2 ………… (ii)
a) Find du & dv in terms of dx, dy & dz at the point
x =1 , y =1 , z = 2 , u = 3 & v = 2 .
æ ¶u ö æ ¶v ö
b) Find ç ÷ , ç ÷ at the point given above.
è ¶x ø( y , z ) è ¶y ø( x , z )
c) Find approximately the values of u & v for x = 1 × 1 , y = 1.2 , z = 1.8
Solutions
Differential gives
2 xdx + 2 ydy + 2 zdz - 2udu + 2vdv = 0 ……… (iii)
2 xdx - 2 ydy + 2 zdz + 2udu + 2vdv = 0 ……… (iv)
a) Putting x = 1 , y = 1 , z = 2 , u = 3 & v = 2 in (iii) & (iv), we obtain
2 dx + 2dy + 4dz - 6du + 4dv = 0 ………… (v)
& 2 dx - 2dy + 4dz + 6du + 8dv = 0 ………… (vi)
Adding gives
12dv = - ( 4dx + 8dz )
1
Þ dv = - ( dx + 0 × dy + 2dz )
3
Similarly eliminating dv between (v ) and (vi ) , we get
1
du = ( dx + 3dy + 2dz )
9
1
b) Q du = ( dx + 3dy + 2dz )
9
æ ¶u ö 1
\ ç ÷ =
è ¶x ø y , z 9
1
& Q dv = - ( dx + 0 × dy + 2dz )
3
æ ¶v ö
\ ç ÷ =0
è ¶y ø x , z

…………………………………..
Chap 5 – Function of several variables 15

v Question
Find the transformation of x = r cosq , y = r sin q from rectangular to polar
coordinates. Verify the relations
a) dx = cosq dr - r sinq dq
dy = sin q dr + r cosq dq
b) dr = cosq dx + sin q dy
sin q cosq
dq = - dx + dy
r r
æ ¶x ö æ ¶x ö ¶ (r ,q ) 1
c) ç ÷ = cosq , ç ÷ = secq , =
è ¶r øq è ¶r ø y ¶ ( x, y ) r
Solutions
Given that x = r cosq & y = r sin q
a) Differential gives
dx = cosq dr - r sinq dq ………. (i)
dy = sin q dr + r cosq dq ………. (ii)
b) Multiplying (i) by cosq & (ii) by sinq and adding, we get
dr = cosq dx + sin q dy
Now multiply (i) by sinq & (ii) by cosq and subtract to obtain
sin q cosq
dq = - dx + dy
r r
c) Given x = r cosq
æ ¶x ö
Þ ç ÷ = cosq
è ¶r øq
We have already shown that dr = cosq dx + sin q dy
dr
Which can be written as dx = - tan q dy
cosq
æ ¶x ö
Þ ç ÷ = secq
è ¶r ø y
¶x ¶x
¶ ( x, y ) ¶r ¶q cosq - r sin q
= = = r cos 2 q + r sin 2 q = r
¶ (r ,q ) ¶y ¶y sin q r cosq
¶r ¶q
¶r ¶r
cosq sin q
¶ (r ,q ) ¶x ¶y 1 1 1
and = = sin q cosq = cos 2 q + sin 2 q =
¶ ( x, y ) ¶q ¶q - r r r
r r
¶x ¶y
v Question
Given that x 2 - y 2 cos uv + z 2 = 0
x 2 + y 2 - sin uv + 2 z 2 = 2
and xy - sin u cos v + z = 0
æ ¶x ö æ ¶x ö p
Find ç ÷ , ç ÷ at x = 1 , y = 1 , u = , v = 0 , z = 0
è ¶u øv è ¶v øu 2
Solution
Differential gives
2 x dx - 2 y cos uv dy + y 2 sin uv × udv + y 2 sin uv × vdu + 2 zdz = 0 ……. (i)
16 Chap 5 – Function of several variables

2 x dx + 2 y dy - cos uv × udv - cos uv × vdu + 4 zdz = 0 ……… (ii)


& x dy + y dx - cos u × cos v du + sin u × sin v dv + dz = 0 ……... (iii)
At the given point, these equations reduce to
2 dx - 2dy = 0 …………..…. (iv)
p
2 dx + 2dy - dv = 0 ……… (v)
2
& dx + dy + dz = 0 …………... (vi)
Adding (iv) & (v), we have
p
4 dx - dv = 0
2
p æ ¶x ö æ ¶x ö p
Þ dx = dv + 0 × du Þ ç ÷ = 0 , ç ÷ =
8 è ¶u øv è ¶v øu 8
v Question
æ ¶u ö
Find ç ÷ if x 2 - y 2 + u 2 + 2v 2 = 1
è ¶x ø y
x 2 + y 2 - u 2 - v2 = 2
Solution
Taking the differentials, we have
2 x dx - 2 y dy + 2u du + 4v dv = 0
2 x dx + 2 y dy - 2u du - 2v dv = 0
Eliminating dv , we get
6 x dx + 2 y dy - 2u du = 0
3x y
Þ du = dx + dy
u u
æ ¶u ö 3x
Þ ç ÷ =
è ¶x ø y u
v Question
Given the transformation
x = u - 2v
y = 2u + v
a) Write the equations of the inverse transformation
b) Evaluate the Jacobian of the transformation and that of the inverse
transformation.
Solution
a) From the equations, we have
1 2
u= x+ y
5 5
2 1
v=- x+ y
5 5
which are the equations of the inverse transformation.
¶x ¶x
¶ ( x, y ) ¶u ¶v
b) Jacobian of the given transformation = =
¶ (u , v) ¶y ¶y
¶u ¶v
1 -2
= =5
2 1
Chap 5 – Function of several variables 17

¶u ¶u
¶ (u , v) ¶x ¶y
Jacobian of the inverse transformation = =
¶ ( x, y ) ¶v ¶v
¶x ¶y
1 2
5 5 1
= =
2 1 5
-
5 5
v Question
¶ ( x, y )
Given the transformation x = f (u , v) , y = g (u , v) with Jacobian J = , show
¶ (u , v)
that for the inverse transformation one has
¶u 1 ¶y ¶u 1 ¶x ¶v 1 ¶y ¶u 1 ¶x
= , =- , =- , =
¶x J ¶v ¶y J ¶v ¶x J ¶u ¶y J ¶u
Solution
The given equations are
f (u , v ) - x = 0 ………. (i)
g (u , v) - y = 0 ………. (ii)
Differentiating w.r.t. x , we get
¶u ¶v
fu + fv - 1 = 0
¶x ¶x
¶u ¶v
gu + gv -0=0
¶x ¶x
Solving these equations by Crammer’s rule, we have
-1 f v
¶u 0 gv gv 1 ¶y æ ¶y ö
=- = = çQ = gv ÷
¶x fu fv J J ¶v è ¶v ø
gu gv
fu -1
¶v g 0 g 1 ¶y
=- u =- u =-
¶x J J J ¶u
Differentiating (i) & (ii) w.r.t. y , we have
¶u ¶v
fu + fv -0=0
¶y ¶y
¶u ¶v
gu + gv -1 = 0
¶y ¶y
Solving these equations by Crammer’s rule, we get
0 fv
¶u -1 g v f 1 ¶x
=- =- v =-
¶y J J J ¶v
fu 0
¶v g -1 f 1 ¶x
=- u = u =
¶y J J J ¶u
18 Chap 5 – Function of several variables

v Question
Given the transformation
x = u 2 - v2
y = 2uv
a) Compute its Jacobian.
æ ¶u ö æ ¶v ö
b) Evaluate ç ÷ & ç ÷
è ¶x ø y è ¶x ø y
Solution
The given equations can be written as
u 2 - v 2 - x = 0 ……….. (i)
2uv - y = 0 ……...…… (ii)
Differentiating (i) & (ii) partially w.r.t. x , we have
¶u ¶v
2u - 2v - 1 = 0 …………. (iii)
¶x ¶x
¶u ¶v
2v + 2u - 0 = 0 …………. (iv)
¶x ¶x
¶x ¶x
¶ ( x, y ) ¶u ¶v 2u -2v
J= = = = 4(u 2 + v 2 )
¶ (u , v ) ¶y ¶y 2v 2u
¶u ¶v
Solving (iii) & (iv) by Crammer’s rule, we have
-1 -2v
æ ¶u ö 0 2u 2u u
ç ÷ =- = =
è ¶x ø y 4(u + v 2 ) 2(u + v 2 )
2 2
J
2u -1
æ ¶v ö 2v 0 -2v -v
ç ÷ =- = =
è ¶x ø y 4(u 2 + v 2 ) 2(u + v 2 )
2
J

Note
æ ¶u ö æ ¶v ö
ç ¶y ÷ & ç ¶y ÷ can be determined in the same manner.
è øx è øx
v Question
Prove that if F ( x, y , z ) = 0 , then
æ ¶z ö æ ¶x ö æ ¶y ö
ç ÷ × ç ÷ × ç ÷ = -1
è ¶x ø y è ¶y ø z è ¶z ø x
Solution
F ( x, y , z ) = 0
Þ Fx dx + Fy dy + Fz dz = 0
Fy Fz æ ¶x ö F
Þ dx = - dy - dz Þ ç ÷ =- y
Fx Fx è ¶y ø z Fx
Fx F æ ¶y ö F
& dy = - dx - z dz Þ ç ÷ =- z
Fy Fy è ¶z ø x Fy
Fx F æ ¶z ö F
dz = - dx - y dy Þ ç ÷ =- x
Fz Fz è ¶x ø y Fz
Chap 5 – Function of several variables 19

Hence
æ ¶z ö æ ¶x ö æ ¶y ö æ Fx ö æ Fy ö æ Fz ö
ç ÷ ×ç ÷ ×ç ÷ = ç - ÷×ç - ÷ × çç - ÷÷ = -1
è ¶x ø y è ¶y ø z è ¶z ø x è Fz ø è Fx ø è Fy ø
v Question
Prove that, if x = f (u , v) , y = g (u , v) , then
æ ¶x ö æ ¶u ö æ ¶y ö æ ¶v ö
ç ÷ ç ÷ =ç ÷ ç ÷
è ¶u øv è ¶x ø y è ¶v øu è ¶y ø x
æ ¶x ö æ ¶v ö æ ¶u ö æ ¶y ö
and ç ÷ ç ÷ =ç ÷ ç ÷
è ¶v øu è ¶x ø y è ¶y ø x è ¶u øv
æ ¶x ö æ ¶y ö
also that ç ÷ ç ÷ = 1
è ¶y øu è ¶x øu
Solution
Q f (u , v ) - x = 0
g (u , v) - y = 0
æ ¶u ö g æ ¶v ö gu
\ ç ÷ = v , ç ÷ =-
è ¶x ø y J è ¶x ø y J
æ ¶u ö fv æ ¶v ö f
ç ÷ =- , ç ÷ = u as already shown
è ¶y ø x J è ¶y ø x J
Taking differentials of the given equations, we have
f u du + f v dv - dx = 0
g u du + g v dv - dy = 0
Þ dx = f u du + f v dv ………… (i)
dy = g u du + g v dv ……..…. (ii)
æ ¶x ö æ ¶x ö
Þ ç ÷ = fu , ç ÷ = f v
è ¶u øv è ¶v øu
æ ¶y ö æ ¶y ö
ç ÷ = gu , ç ÷ = gv
è ¶u øv è ¶v øu
æ ¶x ö æ ¶u ö æ ¶y ö æ ¶v ö
Now ç ÷ ×ç ÷ = ç ÷ ×ç ÷
è ¶u øv è ¶x ø y è ¶v øu è ¶y ø x
g f
Þ f u × v = g v × u , which is true
J J
Similarly, we have the second relation.
Eliminating dv between (i ) & (ii ) , we get
( fu × gv - fv × gu ) du - gv dx + f v dy = 0
f g - fv gu f
Þ dx = u v × du + v dy
gv gv
g f g - f v gu
and dy = v dx - u v du
fv fv
æ ¶x ö f æ ¶y ö g
Þ ç ÷ = v & ç ÷ = v
è ¶y øu g v è ¶x øu f v
æ ¶x ö æ ¶y ö f g
Þ ç ÷ ×ç ÷ = v × v =1
è ¶y øu è ¶x øu g v f v
20 Chap 5 – Function of several variables

v Question
Given that x = f (u , v, w) , y = g (u , v, w) , z = h(u , v, w) with the Jacobian
¶ ( x, y , z )
J= , show that for the inverse transformation one has
¶ (u , v, w)
¶u 1 ¶ ( y, z ) ¶u 1 ¶ ( z, x ) ¶u 1 ¶ ( x, y )
i) = , = , =
¶x J ¶ (v, w) ¶y J ¶ (v, w) ¶z J ¶ (v, w)
¶v 1 ¶ ( y, z ) ¶v 1 ¶ ( z , x ) ¶v 1 ¶ ( x, y )
ii) = , = , =
¶x J ¶ ( w, u ) ¶y J ¶ ( w, u ) ¶z J ¶ ( w, u )
¶w 1 ¶ ( y, z ) ¶w 1 ¶ ( z , x) ¶w 1 ¶ ( x, y )
iii) = , = , =
¶x J ¶ (u, v) ¶y J ¶ (u , v ) ¶z J ¶ (u , v )
Solution
We have f (u , v, w) - x = 0
g (u , v, w) - y = 0
h (u , v, w) - z = 0
Differentiating w.r.t. to x , we get
¶u ¶v ¶w
fu + fv + f w -1 = 0
¶x ¶x ¶x
¶u ¶v ¶w
gu + gv + gw -0=0
¶x ¶x ¶x
¶u ¶v ¶w
hu + hv + hw -0=0
¶x ¶x ¶x
By Crammer’s rule, we have
-1 f v f w
0 gv g w gv g w
¶u 0 hv hw hv hw 1 ¶ ( g , h) 1 ¶( y, z)
=- = = =
¶x J J J ¶ ( v, w) J ¶ (v, w)
fu -1 f w
gu 0 gw gu gw
¶v h 0 hw h hw 1 ¶ ( g , h) 1 ¶( y, z)
=- u =- u = =
¶x J J J ¶ ( w, u ) J ¶ ( w, u )
fu f v -1
gu gv 0 gu g v
¶w h hv 0 h hv 1 ¶ ( g , h) 1 ¶ ( y, z )
=- u = u = =
¶x J J J ¶ (u , v) J ¶ (u , v)
We can find the other relations in the same way by differentiating given relation
w.r.t. y and w.r.t. z respectively.

………………………………….
Chap 5 – Function of several variables 21

v Partial Derivative of Higher Order


¶z ¶z
Let a function z = f ( x, y ) be given. Then its two partial derivatives &
¶x ¶y
are themselves functions of x & y .
¶z ¶z
i.e. = f x ( x, y ) , = f y ( x, y )
¶x ¶y
Hence each can be differentiable w.r.t. x & y .
Thus, we obtain four partial derivatives
¶2 z ¶2z
= f xx ( x, y ) , = f xy ( x, y )
¶x 2 ¶x ¶y
¶2z ¶2z
= f yx ( x, y ) , = f yy ( x, y )
¶y ¶x ¶y 2
¶2 z ¶z ¶2 z
is the result of differentiating w.r.t. x , where is the result of
¶x 2 ¶x ¶y ¶x
¶z
differentiating w.r.t. y . If all the derivatives concerned are continuous in the
¶x
¶2z ¶2z
domain considered, then = i.e. order of differentiation is immaterial.
¶x ¶y ¶y ¶x
Third and higher order partial derivatives are defined in the same manner and
under appropriate assumptions of continuity the order of differentiation does not
matter.
v Laplacian of z
If z = f ( x, y ) , then the Laplacian of z is denoted by Ñ 2 z is the expression
¶2 z ¶2 z
Ñ z= 2 + 2
2

¶x ¶y
if w = f ( x, y , z ) , the Laplacian of w is the expression
¶ 2 w ¶ 2w ¶ 2w
Ñ w= 2 + 2 + 2
2

¶x ¶y ¶z
The symbol “ Ñ ” is a vector differential operator define as
¶ ¶ ˆ ¶ ˆ
Ñ = iˆ + j+ k
¶x ¶y ¶x
We then have symbolically
¶2 ¶2 ¶2
Ñ = Ñ×Ñ = 2 + 2 + 2
2

¶x ¶y ¶z
v Harmonic Function
If z = f ( x, y ) has continuous second order derivatives in a domain D and Ñ 2 z = 0
in D , then z is said to be Harmonic in D . The same term is used for the function of
three variables which has continuous 2nd derivatives in a domain D in space and
whose Laplacian is zero in D . The two equations for harmonic functions
¶2 z ¶2 z
Ñ z= 2 + 2 =0
2

¶x ¶y
¶ 2 w ¶ 2 w ¶ 2w
Ñ w= 2 + 2 + 2 =0
2

¶x ¶y ¶z
are known as the Laplace equations in two and three dimensions respectively.
22 Chap 5 – Function of several variables

v Bi-Harmonic Equations
Another important combination of derivatives occurs in the equation
¶4 z ¶4 z ¶4 z
+ 2 + =0
¶x 4 ¶x 2¶y 2 ¶y 4
which is known to be the Bi-harmonic equation. This combination can be
expressed in terms of Laplacian as
( )
Ñ2 Ñ2 z = Ñ4 z = 0
The solutions of Ñ 4 z = 0 are termed as Pri-harmonic functions.
v Higher Derivatives of Functions of Functions
(1) Let z = f ( x, y ) and x = g (t ) , y = h(t ) so that z can be expressed in terms of t
alone. Then
dz ¶z dx ¶z dy
= + ……….. (i)
dt ¶x dt ¶y dt
d 2 z d æ dz ö ¶z d 2 x dx d æ ¶z ö ¶z d 2 y dy d æ ¶z ö
= ç ÷= + ç ÷+ + ç ÷ ……… (ii)
dt 2 dt è dt ø ¶x dt 2 dt dt è ¶x ø ¶y dt 2 dt dt è ¶y ø
Using (i), we have
d æ ¶z ö ¶ 2 z dx ¶ 2 z dy
ç ÷= +
dt è ¶x ø ¶x 2 dt ¶y ¶x dt
d æ ¶z ö ¶ 2 z dx ¶ 2 z dy
& ç ÷= +
dt è ¶y ø ¶x¶z dt ¶y 2 dt
Putting these values in (ii ) , we have
2 2
d 2 z ¶z d 2 x ¶ 2 z æ dx ö ¶ 2 z dx dy ¶ 2 z æ dy ö ¶z d 2 y
= + ç ÷ +2 × + ç ÷ +
dt 2 ¶x dt 2 ¶x 2 è dt ø ¶x ¶y dt dt ¶y 2 è dt ø ¶y dt 2
(2) If z = f ( x, y ) and x = g (u , v) , y = h(u , v ) , then
¶z ¶z ¶x ¶z ¶y
= × + × ……….. (iii)
¶u ¶x ¶u ¶y ¶u
¶z ¶z ¶x ¶z ¶y
= × + × ……….. (iv)
¶v ¶x ¶v ¶y ¶v
¶ 2 z ¶z ¶ 2 x ¶ æ ¶z ö ¶x ¶z æ ¶ 2 y ö ¶y ¶ æ ¶z ö
= × + ç ÷× + ç ÷+ ×
¶u 2 ¶x ¶u 2 ¶u è ¶x ø ¶u ¶y è ¶u 2 ø ¶u ¶u çè ¶y ÷ø
…………. (iv)

Using (iii), we have


¶ æ ¶z ö ¶ 2 z ¶x ¶ 2 z ¶y
ç ÷= × + ×
¶u è ¶x ø ¶x 2 ¶u ¶y ¶x ¶u
¶ æ ¶z ö ¶ 2 z ¶x ¶ 2 z ¶y
= +
¶u èç ¶y ÷ø ¶x ¶y ¶u ¶y 2 ¶u
and

Putting these values in (iv ) , we get


2
¶ 2 z ¶z ¶ 2 x ¶ 2 z æ ¶x ö ¶ 2 z ¶x ¶y ¶ 2 z æ ¶y ö ¶z ¶ 2 y
= + ç ÷ + 2 × + ç ÷ + ×
¶u 2 ¶x ¶u 2 ¶x 2 è ¶u ø ¶x ¶y ¶u ¶u ¶y 2 è ¶u ø ¶y ¶u 2
¶2z ¶2 z
We can find the values of & in the same manner.
¶u ¶v ¶v 2

………………………………………..
Chap 5 – Function of several variables 23

v The Laplacian in Polar, Cylindrical and Spherical Co-ordinate


We consider first the two-dimensional Laplacian
¶ 2 w ¶ 2w
Ñ w= 2 + 2
2

¶x ¶y
and its expression in terms of polar co-ordinates r & q .
Thus we are given w = f ( x, y ) and x = r cosq , y = r sin q and we wish to express
Ñ 2 w in terms of r , q and derivatives of w with respect to r and q . The solution is
as follows. One has
¶w ¶w ¶r ¶w ¶q
= × + ×
¶x ¶r ¶x ¶q ¶x
¶w ¶w ¶r ¶w ¶q
= × + × by chain rule
¶y ¶r ¶y ¶q ¶y
¶r ¶q ¶r ¶q
To evaluate , , , , we use the equations
¶x ¶x ¶y ¶y
dx = cosq dr - r sin q dq
dy = sin q dr + r cosq dq
These can be solved for dr and dq by determinants or by elimination to give
dr = cosq dx + sin q dy
sin q cosq
dq = - dx + dy
r r
¶r ¶r ¶q sin q ¶q cosq
Hence = cosq , = sin q , =- and =
¶x ¶y ¶x r ¶y r
¶w ¶w
Putting these values above in expressions of & , we have
¶x ¶y
¶w ¶w sin q ¶w ü
= cosq -
¶x ¶r r ¶q ïï
¶w cosq ¶w ýï
.............. (i )
¶w
= sin q +
¶y ¶r r ¶q ïþ
These equations provide general rules for expressing derivatives w.r.t. x or y in
terms of derivatives w.r.t. r and q . By applying the first equation to the function
¶w
, one finds that
¶x
¶ 2 w ¶ æ ¶w ö ¶ æ ¶w ö sin q ¶ æ ¶w ö
= ç ÷ = cosq ç ÷- ç ÷
¶x 2
¶x è ¶x ø ¶r è ¶x ø r ¶q è ¶x ø
By (i) this can be written as follows:
¶2w ¶æ ¶w sin q ¶w ö sin q ¶ æ ¶w sin q ¶w ö
= cosq ç cosq - ÷- ç cosq - ÷
¶x 2
¶r è ¶r r ¶q ø r ¶q è ¶r r ¶q ø
The rule for differentiation of a product gives finally
¶2w ¶ 2 w 2sin q cosq ¶ 2 w sin 2 q ¶ 2 w
= cos q × 2 -
2
× + 2 × 2
¶x 2 ¶r r ¶r ¶q r ¶q
sin 2 q ¶w 2sin q cosq ¶w
+ × + × ……. (ii)
r ¶q r2 ¶q
In the same manner one finds
¶ 2 w ¶ æ ¶w ö ¶æ ¶w cosq ¶w ö cosq ¶ æ ¶w cosq ¶w ö
= ç ÷ = sin q ç sin q + ÷+ ç sin q + ÷
¶y 2
¶y è ¶y ø ¶r è ¶r r ¶q ø r ¶q è ¶r r ¶q ø
24 Chap 5 – Function of several variables

¶ 2 w 2sin q cosq ¶ 2 w cos 2 q ¶ 2 w


= sin q × 2 +
2
× + × 2
¶r r ¶r ¶q r2 ¶q
cos 2 q ¶w 2sin q cosq ¶w
+ × - × …… (iii)
r ¶r r2 ¶q
Adding (ii) & (iii), we conclude
¶ 2 w ¶ 2 w ¶ 2 w 1 ¶ 2 w 1 ¶w
Ñ w= 2 + 2 = 2 + 2
2
+ ………. (iv)
¶x ¶y ¶r r ¶q 2 r ¶r
This is the desired result.
Equation (iv) at once permits one to write the expression for the 3-demensional
Laplacian in cylindrical co-ordinates for the transformation of coordinates
x = r cosq , y = r sin q , z = z
involves only x & y . In the same way as above, we have
¶ 2 w ¶ 2w ¶ 2w
Ñ2w = + +
¶x 2 ¶y 2 ¶z 2
¶ 2 w 1 ¶ 2 w 1 ¶w ¶ 2 w
= 2 + 2 + +
¶r r ¶q 2 r ¶r ¶z 2
v Laplacian in Spherical Polar Coordinates
The transformation form rectangular to spherical polar coordinates is
x = r sin j cosq , y = r sin j sin q , z = r cos j
Writing r = r sin j , we have
x = r cosq , y = r sin q , z = z
Which can be considered as a transformation from rectangular to cylindrical
coordinates (r ,q , z )
We have
¶ 2 w 1 ¶ 2 w 1 ¶w ¶ 2 w
Ñ2w = 2 + 2 + + ……….. (i)
¶r r ¶q 2 r ¶r ¶z 2
where z = r cos j ü
ý ………. (ii)
r = r sin j þ
We have transformation from ( x, y ) to (r ,q ) as
¶ 2 w ¶ 2 w ¶ 2 w 1 ¶ 2 w 1 ¶w
+ = + +
¶x 2 ¶y 2 ¶r 2 r 2 ¶q 2 r ¶q
Now if we take transformation from ( z , r ) to ( r ,j ) , then
¶ 2 w ¶ 2 w ¶ 2 w 1 ¶ 2 w 1 ¶w
Þ + = + +
¶z 2 ¶r 2 ¶r 2 r 2 ¶j 2 r ¶j
¶w ¶w ¶r ¶w ¶j
Also = × + ×
¶r ¶r ¶r ¶j ¶r
r
Where r 2 = z 2 + r 2 , tan j =
z
¶r ¶r r r sin j
Þ 2r = 2r Þ = = = sin j
¶r ¶r r r
¶j 1 ¶j cos 2 j cos 2 j cos j
& sec 2 j × = Þ = = =
¶r z ¶r z r cos j r
¶w ¶w ¶w cos j
Þ = × sin j + × ………… (iv)
¶r ¶r ¶j r
Substituting (iii) & (iv) in (i), we have
Chap 5 – Function of several variables 25

æ ¶ 2 w ¶ 2 w ö 1 ¶ 2 w 1 ¶w
Ñ2w = ç 2 + 2 ÷ + 2 +
è ¶ z ¶r ø r ¶ q 2
r ¶r
¶ 2 w 1 ¶ 2 w 1 ¶w 1 ¶ 2w 1 æ ¶w ¶w cos j ö
= 2+ 2 + + 2 2 × 2 + ç sin j +
¶r r ¶j 2
r ¶r r sin j ¶q r sin j è ¶r ¶j r ÷ø
¶2w 1 ¶ 2w 1 ¶w 1 ¶ 2 w 1 ¶w cot j ¶w
= 2+ × 2+ × + 2 2 × 2+ × + 2 ×
¶r r2 ¶j r ¶r r sin j ¶q r ¶r r ¶j
¶2w 1 ¶ 2w 2 ¶w 1 ¶ 2 w cot j ¶w
= 2+ × 2+ × + 2 2 × 2+ 2 ×
¶r r2 ¶j r ¶r r sin j ¶q r ¶j

v Question
If u & v are functions of x & y defined by the equations
xy + uv = 1 , xu + yv = 1
¶ 2u
then find .
¶x 2
Solution
y dx + x dy + v du + u dv = 0 ……… (i)
u dx + v dy + x du + y dv = 0 ……... (ii)
Eliminating dv between (i) & (ii)
( )
y 2 - u 2 dx + ( xy - uv ) dy + ( vy - ux ) du = 0
u 2 - y2 uv - xy
Þ du = dx + dy
vy - ux vy - ux
¶u u 2 - y 2 u 2 - y 2
Þ = = ( using given eq. )
¶x vy - ux 1 - 2ux
¶u é ¶u ù
(1 - 2ux) × 2u × - (u 2 - y 2 ) ê(-2u ) - 2 x ú
¶u
2
¶x ë ¶x û
Þ =
¶x (1 - 2ux )
2 2

v Question
¶ 2w ¶ 2w
Find , when
¶x 2 ¶y 2
1
i) w =
x2 + y2
y
ii) w = tan -1
x
x2 - y 2
iii) w = e
v Question
Show that the following functions are harmonic in x & y
i) e x cos y
ii) x 3 - 3xy 2
iii) log x 2 + y 2

………………………………………
26 Chap 5 – Function of several variables

v Sufficient Condition for the Validity of Reversal in the Order


of Derivation
We now prove two theorems which lay sufficient conditions for the equality of f xy
and f yx .

v Schawarz’s Theorem
If (a, b) be a point of the domain of a function f ( x, y ) such that
i) f x ( x, y ) exists in a certain nhood of (a, b) .
ii) f xy ( x, y ) is continuous at (a, b) .
then f yx (a, b) exists and is equal to f xy (a, b) .
Proof
The given conditions imply that there exists a certain nhood of (a, b) at every point
( x, y ) of which f x ( x, y ) , f y ( x, y ) and f xy ( x, y ) exist. Let (a + h, b + k ) be any point
of this nhood. We write
f (h, k ) = f (a + h, b + k ) - f (a + h, b) - f (a, b + k ) + f (a, b)
g ( y ) = f (a + h, y ) - f (a, y )
so that f (h, k ) = g (b + k ) - g (b) ……….. (i)
Q f y exists in a nhood of ( a, b ) , the function g ( y ) in derivable in [b, b + k ] , and,
therefore, by applying the M.V. theorem to the expression on R.H.S of (i ) , we have
f (h, k ) = kg ¢(b + q k ) ( 0 < q < 1)
= k ( f y (a + h, b + q k ) - f y (a, b + q k ) ) ……….. (ii)
Again since f xy exists in a nhood of (a, b) , the function f y ( x, b + q k ) of x is
derivable w.r.t. x in interval (a, a + h) and, therefore, by applying the M.V. theorem
to the right of (ii ) , we have
f (h, k ) = hk f xy (a + q ¢h, b + q k ) ( 0 < q ¢ < 1)
1 æ f ( a + h, b + k ) - f ( a , b + k ) f ( a + h , b ) - f ( a , b ) ö
or ç - ÷ = f xy (a + q ¢h, b + q k )
kè h h ø
Since f x ( x, y ) exists in a nhood of (a, b) , this gives when h ® 0 ,
f x (a , b + k ) - f x ( a , b )
= lim f xy (a + q ¢h, b + q k )
k h®0

Let, now, k ® 0 . Since f xy ( x, y ) is continuous at (a, b) , we obtain


f yx (a, b) = lim lim f xy (a + q ¢h, b + q k ) = f xy (a, b)
k ®0 h ®0

v Young’s Theorem
If (a, b) be a point of the domain of definition of a function f ( x, y ) such that
f x ( x, y ) and f y ( x, y ) are both differentiable at (a, b) , then
f xy (a, b) = f yx (a, b)
Proof
The differentiability of f x and f y at (a, b) implies that they exist in a certain
nhood of (a, b) and that f xx , f yx , f xy , f yy exist at (a, b) .
Let (a + h, b + h) be a point of this nhood. We write
f (h, h) = f (a + h, b + h) - f (a + h, b) - f (a , b + h) + f (a, b)
g ( y ) = f (a + h, y ) - f (a, y )
so that f (h, h) = g (b + h) - g (b) ………. (i )
Chap 5 – Function of several variables 27

Since f y exists in a nhood of (a, b) , the function g ( y ) is derivable in (b, b + h) ,


and, therefore, by applying the M.V. theorem to the expression on the right of (i ) ,
we have
f (h, h) = h g ¢(b + q h) ( 0 < q < 1)
= h ( f y (a + h, b + q h) - f y (a, b + q h) ) ……….. (ii )
Since f y ( x, y ) is differentiable at (a, b) , we have, by definition,
f y (a + h, b + q h) - f y (a, b) = h f xy (a, b) + q h f yy (a, b)
+ h j1 (h, h) + q hy 1 (h, h) …… (iii)
and f y (a, b + q h) - f y (a, b) = q h f yy (a, b) + q hy 2 (h, h) ….….. (iv )
where j1 , y 1 , y 2 all ® 0 as h ® 0
From (ii ) , (iii ) and (iv ) , we obtain
f (h, h)
= f xy (a, b ) + f1 (h, h) + qy 1 (h, h) - q y 2 (h, h) …..…. (v )
h2
By a similar argument and on considering
g ( x ) = f ( x, b + k ) - f ( x, b )
We can show that
f (h, h)
= f yx (a, b) + y 3 (h, h) + q ¢j 2 (h, h) - q ¢j 3 (h, h) ……….. (vi )
h2
where j 2 , j 3 ,y 3 all ® 0 as h ® 0
Equating the right hand side of (v ) and (vi ) and making h ® 0 , we obtain
f xy (a, b) = f yx (a, b)

………………………………….
28 Chap 5 – Function of several variables

v Maxima and Minima for Functions of Two Variables


Let ( x0 . y0 ) be the point of the domain of a function f ( x, y ) , then f ( x0 , y0 ) said to
an extreme value of the function f ( x, y ) , if the expression
Df = f ( x0 + h, y0 + h) - f ( x0 , y0 )
preserves its sign for all h and k .
The extreme value of f ( x0 , y0 ) being called a maximum or a minimum value
according as this difference is positive or negative respectively.
Necessary Condition
The Necessary Condition for f ( x0 , y0 ) to be an extreme value of function f ( x, y )
is that f x ( x0 , y0 ) = 0 = f y ( x0 , y0 ) , provided that these partial derivatives exist.
It is to be noted that it is impossible to determine the nature of a critical point by
studying the function f ( x, y0 ) and f ( x0 , y ) .
e.g. Let f ( x, y ) = 1 + x 2 - y 2
then f (0, y ) = 1 - y 2 Þ f ¢(0, y ) = -2 y = 0 Þ (0,0) is a turning point.
Now f ¢¢(0, y ) = -2 Þ (0,0) is a point of maximum value.
But f ( x,0) = 1 + x 2
Þ f ¢( x,0) = 2 x = 0 Þ x = 0 Þ (0,0) is the critical point
Þ f ¢¢( x,0) = 2 > 0 Þ (0,0) is the maximum value
Hence we fail to decide the nature of the critical point in this way.
Sufficient Condition
Let z = f ( x, y ) be defined and have continuous 1st and 2nd order partial derivatives
in a domain D . Suppose ( x0 , y0 ) is a point of D for which f x and f y are both zero.
Let A = f xx ( x0 , y0 ) , B = f xy ( x0 , y0 ) , C = f yy ( x0 , y0 ) ,
then we have the following cases
i) B 2 - AC < 0 and A + C < 0 Þ relative maximum at ( x0 , y0 ) .
ii) B 2 - AC < 0 and A + C > 0 Þ relative minimum at ( x0 , y0 )
iii) B 2 - AC > 0 Þ saddle point at ( x0 , y0 )
iv) B 2 - AC = 0 Þ nature of the critical point is undetermined
Proof
By the application of M.V. theorem for function of two variables we have
Df = hf x ( x0 + q h, y0 + q k ) + kf y ( x0 + q h, y0 + q k ) (0 < q < 1)
= h [ f x ( x0 + q h, y0 + q k ) - f x ( x0 , y0 ) ] + k éë f y ( x0 + q h, y0 + q k ) - f y ( x0 , y0 ) ùû
(it is because f x ( x0 , y0 ) = f y ( x0 , y0 ) = 0 , a turning point)
= h éëq hf xx ( x0 , y0 ) + q kf yx ( x0 , y0 ) + e1q h + e 2q k ùû
+ k éëq hf xy ( x0 , y0 ) + q kf yy ( x0 , y0 ) + e 3q h + e 4q k ùû
where e1 , e 2 , e 3 & e 4 ® 0 as h, k ® 0
Df = h 2 f xx ( x0 , y0 ) + 2hk f xy ( x0 , y0 ) + k 2 f yy ( x0 , y0 ) +e1h 2 + (e 2 + e 3 )hk + e 4k 2
Þ Df = h 2 A + 2hkB + k 2C + e1h2 + (e 2 + e 3 )hk + e 4 k 2
The sign of Df depends upon the quadratic d 2 f = h2 A + 2hk B + k 2C
i & ii) Let B 2 - AC < 0 , ( A ¹ 0)

Þ d2 f =
1 2
A
(
h A + 2hk AB + k 2 AC )
Chap 5 – Function of several variables 29

=
A
(
1 2 2
)
h A + 2hk AB + k 2 B 2 + (k 2 AC - k 2 B 2 )

=
1
A
( )
(hA + kB )2 + k 2 ( AC - B 2 )

Since (hA + kB )2 is positive and AC - B 2 (supposed) is +ive, therefore the sign of


d 2 f depends upon the sign of A .
Þ Df > 0 if A > 0 & Df < 0 if A < 0
Again, since B 2 - AC < 0 Þ B 2 < AC Þ AC > 0
Þ A and C are either both +ive or both –ive.
If A > 0 , C > 0 then A + C > 0 and if A < 0 , C < 0 then A + C < 0 .
Hence we have the following result
a) Df > 0 when A + C > 0 Þ ( x0 , y0 ) is a point of minimum value.
b) Df < 0 when A + C < 0 Þ ( x0 , y0 ) is a point of maximum value.
iii) Let B 2 - AC > 0 , then
d2 f =
1
A
( )
(hA + kB) 2 + k 2 ( AC - B 2 )

=
1
A
( )
(hA + kB )2 - k 2 ( B 2 - AC )
which may be +ive or –ive for certain value of h & k , therefore ( x0 , y0 ) is a
saddle point.
iv) Let B 2 - AC = 0 , A ¹ 0
1
Þ d 2 f = ( hA + kB )
2

A
which may vanish for certain values of h and k , implies that nature of the point
remain undetermined.
v Question
Test for maxima and minima
z = 1 - x2 - y2
Solution
¶z
= -2 x = 0 Þ x = 0
¶x
¶z
= -2 y = 0 Þ y = 0
¶y
Þ (0,0) is the only critical point.
¶2z ¶2z ¶2z
A = 2 = -2 , B = = 0 , C = 2 = -2
¶x ¶x ¶y ¶y
B - AC = 0 - 4 = -4 < 0 and A + C = -2 - 2 = -4 < 0
2

Þ the function has maximum value at (0,0) .


v Question
Test for maxima and minima
z = x 3 - 3xy 2
Solution
¶z
= 3x 2 - 3 y 2 = 0 Þ x = - y & x = y
¶y
¶z
= -6 xy = 0 Þ xy = 0
¶y
30 Chap 5 – Function of several variables

Þ (0,0) is the critical point.


¶2 z
A = 2 = 6 x = 0 at (0,0)
¶x
¶2 z
B= = -6 y = 0 at (0,0)
¶x¶y
¶2z
C = 2 = -6 x = 0 at (0,0)
¶y
B - 4 AC = 0 also A + C = 0
2

Therefore we need further consideration for the nature of point


Dz = z (0 + h,0 + k ) - z (0,0)
= z (h, k ) - z (0,0)
= h 3 - 2hk 2
For h = k
Dz = h 3 - 3h3 = -2h3
Þ Dz > 0 if h < 0 & Dz < 0 if h > 0
Hence (0,0) is a saddle point.
v Question
Examine the function
z = f ( x, y ) = x 2 y 2
Solution
f x = 0 Þ 2 xy 2 = 0
f y = 0 Þ 2 yx 2 = 0
implies that (0,0) is the critical point
A = f xx = 2 y 2 = 0 at (0,0)
B = f xy = -4 xy = 0 at (0,0)
C = f yy = 2 x 2 = 0 at (0,0)
Since B 2 - 4 AC = 0 and also A + C = 0
Therefore we need further consideration for the nature of point.
Df = f (h, k ) - f (0,0)
= h 2k 2
Df > 0 for all h & k
Hence (0,0) is the point where function has minimum value.

……………………………………..
Chap 5 – Function of several variables 31

v Lagrange’s Multiplier
(Maxima & Minima for Function with Side Condition)
A problem of considerable importance for application is that of maximizing and
minimizing of function (optimization) of several variables where the variables are
related by one or more equations, which are turned as side condition. e.g. the
problem of finding the radius of largest sphere inscribable in the ellipsoid
x 2 + 2 y 2 + 3z 2 = 6 is equivalent to minimizing the function w = x 2 + y 2 + z 2 with the
side condition x 2 + 2 y 2 + z 2 = 6 .
To handle such problem, we can, if possible, eliminate some of the variables by
using the side conditions and reduce the problem to an ordinary maximum and
minimum problem such as that consider previously.
This procedure is not always feasible and following procedure often is more
convenient which treat the variable in more symmetrical manner, so that various
simplifications may be possible.
Consider the problem of finding the extreme values of the function f ( x1, x2 ,...., xn )
when the variable are restricted by a certain number of side conditions say
g1 ( x1, x2 ,...., xn ) = 0
g 2 ( x1, x2 ,...., xn ) = 0
……………………
…………………....
……………………
g m ( x1, x2 ,...., xn ) = 0
We then form the linear combination
j ( x1 ,..., xn ) = f ( x1 ,..., xn ) + l1g1 ( x1 ,..., xn ) + l2 g 2 ( x2 ,..., xn ) + ....... + lm g m ( x1,..., xn )
where l1 , l2 ,...., lm are m constants.
We then differentiate j w.r.t. each coordinate and consider the following system
of n + m equations.
Drj ( x1 , x2 ,...., xn ) = 0 , r = 1,2,...., n
g k ( x1, x2 ,...., xn ) = 0 , k = 1, 2,...., m
Lagrange discovered that if the point ( x1, x2 ,...., xn ) is a solution of the extreme
problem then it will also satisfy the system of n + m equation.
In practise, we attempt to solve this system for n + m unknowns, which are
l1 , l2 ,...., lm & x1, x2 ,...., xn
The point so obtain must then be tested to determine whether they yield a
maximum, a minimum or neither.
The numbers l1 , l2 ,...., lm , which are introduced only to help to solve the system
for x1, x2 ,...., xn are known as Lagrange’s multiplier. One multiplier is introduced for
each side condition.
v Question
Find the critical points of w = xyz , subject to condition x 2 + y 2 + z 2 = 1 .
Solution
We form the function
j = xyz + l ( x 2 + y 2 + z 2 - 1)
then
¶j
= yz + 2l x = 0
¶x
¶j
= xz + 2l y = 0
¶y
32 Chap 5 – Function of several variables

¶j
= xy + 2l z = 0
¶z
& x2 + y 2 + z 2 -1 = 0
Multiplying the first three equations by x, y & z respectively, adding and using
the fourth equation, we find
3x y z
l=-
2
using this relation we find that ( 0,0, ±1) , ( 0, ±1,0 ) , ( ±1,0,0 ) and
æ 1 1 1 ö
ç± ,± ,± ÷ are the critical points.
è 3 3 3ø
v Question
Find the critical points of w = xyz , where x 2 + y 2 = 1 & x - z = 0 . Also test for
maxima and minima.
Solution
( )
Consider F = xyz + l1 x 2 + y 2 + 1 + l2 ( x - z )
For the critical points, we have
Fx = yz + 2l1 x + l2 = 0 ……….. (i)
Fy = xz + 2l1 y = 0 ………...….. (ii)
Fz = xy - l2 = 0 …………...….. (iii)
and x 2 + y 2 = 1 ………...….…. (iv)
x - z = 0 …………………. (v)
xz
From (iii ) , l2 = xy & from (ii ) l1 = -
2y
Use these values in equation (i ) to have
x2 z
yz - + xy = 0
y
Þ y 2 z - x 2 z + xy 2 = 0
Q x = z from (v )
\ y 2 x - x 3 + xy 2 = 0 Þ 2 xy 2 - x 3 = 0
But y 2 = 1 - x 2 , from (iv )

( )
\ 2 x 1 - x 2 - x3 = 0 Þ 2 x - 3 x3 = 0 Þ x=0 , ±
2
3
æ 2 1 2ö æ 2 1 2ö
This implies the critical points are ç ± , ,± ÷ ç
, ± , - , ± ÷,
è 3 3 3 ø è 3 3 3 ø
( 0,1,0 ) , ( 0, -1,0 )
A = Fxx = 2l1
B = Fxy = z
C = Fyy = 2l1
B 2 - AC = z 2 - 4l12
x2 z 2
= z -4 2 =
2 (
z2 y 2 - x2 )
4y y2
æ 2 1 2ö
(i) At ç ± , ,± ÷ , we have
è 3 3 3 ø
Chap 5 – Function of several variables 33

2æ1 2ö
ç - ÷
3è3 3ø
B - AC =
2
<0
1
3
æ 2 ö
& A = Fxx = 2l1 = - = - ç 3 ÷ < 0
xz
ç 1 ÷
y ç ÷
è 3ø
æ 2 1 2ö
Þ function has maximum value at ç ± , , ± ÷
è 3 3 3 ø
Similarly, we can show that F is also maximum at (0, -1,0) and is minimum at
remaining points. (Check yourself)
v Question
Find the point of the curve
x 2 - xy + y 2 - z 2 = 1 , x 2 + y 2 = 1
which is nearest to the origin.
Solution
Let a point on a given curve be ( x, y, z )
Implies that we are to minimize the function
f = d 2 = x2 + y 2 + z 2
subject to the conditions
x 2 - xy + y 2 - z 2 = 1
x2 + y2 = 1
Consider
( ) ( )
F = x 2 + y 2 + z 2 + l1 x 2 - xy + y 2 - z 2 - 1 + l2 x 2 + y 2 - 1
For the critical points
Fx = 2 x (1 + l1 + l2 ) - l1 y = 0 ……….. (i )
Fy = 2 y (1 + l1 + l2 ) - l1x = 0 ……….. (ii )
Fz = 2 z (1 - l1 ) = 0 ………… (iii )
x 2 - xy + y 2 - z 2 = 1 ………… (iv )
x 2 + y 2 = 1 ………….. (v )
From equation (iii ) , we have
z = 0 and l1 = 1
Put z = 0 in equation (iv ) , gives
x 2 - xy + y 2 - 1 = 0
Þ xy = x 2 + y 2 - 1
Þ xy = 0 by (v )
Þ x = 0 or y = 0 or both are zero.
z = 0 , x = 0 in (v ) gives, y 2 = 1 Þ y = ±1
Þ ( 0, ±1,0 ) are the critical points.
z = 0 , y = 0 Þ x = ±1 Þ ( ±1,0,0 ) are the critical points.
We can not take x = 0 , y = 0 at the same time, because it gives ( 0,0,0 ) which is
origin itself as a critical point.
Q d 2 = 1 at all these four points.
\ these are the required point at which function is nearest to origin.
34 Chap 5 – Function of several variables

v Question
Find the point on the curve
x2 + y 2 + z 2 = 1
which is farthest from the point (1,2,3)
Solution
We are the maximize the function
f = ( x - 1) + ( y - 2 ) + ( z - 3 )
2 2 2

subject to the condition


x2 + y 2 + z 2 = 1
Let
2 2 2
( )
F = ( x - 1) + ( y - 2 ) + ( z - 3) + l x 2 + y 2 + z 2 - 1
For the critical points, we have
x - 1 + l x = 0 …………. (i )
y - 2 + l y = 0 …..…..… (ii )
z - 3 + l z = 0 ………… (iii )
& x 2 + y 2 + z 2 = 1 ……….. (iv )
1 2 1
Þ x= , y= , z=
1+ l 1+ l 3+l
Putting in (iv )
2
æ 1 ö
÷ (1 + 4 + 9 ) = 1 Þ (1 + l ) = 14 Þ l = -1 ± 14
2
ç
è1+ l ø
1 2 3
Þ x= , y= , z=
± 14 ± 14 ± 14
Þ critical points are
æ 1 2 3 ö
ç± ,± ,± ÷
è 14 14 14 ø
Its clear that the required point which is farthest from the point (1,2,3) is
æ 1 2 3 ö
ç- ,- ,- ÷
è 14 14 14 ø

………………………………………
Chap 5 – Function of several variables 35

v Directional Derivative
i) Let f : V ® ¡ , where V Ì ¡ n , is nhood of a Î ¡ n . Then the directional
derivative Db f at a in the direction of b Î ¡ n , is defined by the limit, if it exists,

Db f (a ) = lim
(
f a + hb - f ( a ) )
hh ®0

ii) The directional derivative of f ( x1, x2 ,..., xi ,..., xn ) at a = ( a1 , a2 ,..., ai ,..., an ) in


the direction of the unit vector ( 0,0,...,1,0,0,...,0 ) is called partial derivative of f at
a w.r.t. the ith component xi and is denoted by
¶f ( a )
Di f ( a ) or or f xi (a)
¶x
f ( a1, a2 ,..., ai + h,..., an ) - f ( a1, a2 ,..., ai ,..., an )
where Di f ( a ) = lim
h®0 h
v Example
Let f ( x, y ) = x 2 + y 2 + x + y , then f has a directional derivative in every direction
and at every point in ¡ 2 .
Since, if b = ( a, b ) Î ¡ 2 , we have
( x + ha )
+ ( y + hb ) + ( x + ha ) + ( y + hb ) - x 2 - y 2 - x - y
2 2

Db f ( x, y ) = lim
h®0 h
h ®0
(
= lim 2ax + 2by + ha + hb + a + b
2 2
)
= 2 ( ax + by ) + a + b
v Exercise

ï
(
ì xy x 2 - y 2 ) ; x4 + y 4 ¹ 0
Let f ( x, y ) = í x 4 + y 4
ï ; ( x, y ) ¹ (0,0)
î 0
Note that if b = ( a, b ) Î ¡ 2 ,
( 0 + ah )( 0 + bh ) éë( 0 + ah ) - ( 0 + bh ) ù
2 2

Db f (0,0) = lim û
h é( 0 + ah ) + ( 0 + bh ) ù
h ®0 4 4
ë û

= lim
(
ab a 2 - b 2 )
h ®0 h(a 4
+ b4 )
This limit obviously exists only if b = (1,0 ) or ( 0,1) . Hence the directional
derivatives of f at ( 0,0 ) that exists are the partial derivatives f x and f y given by
fx = 0 , fy = 0.

v Example
Let
ì xy 2
ï ; ( x, y ) = ( 0,0 )
f ( x, y ) = í x 4 + y 4
ï 0 ; ( x, y ) ¹ ( 0,0 )
î
It is discontinuous at ( 0,0 ) . To see it, note that
1
lim f ( x, y ) is zero along y = 0 and is along y 2 = x .
( x , y ) ®(0,0) 2
36 Chap 5 – Function of several variables

However, if b = ( a, b ) , then
( 0 + ah )( 0 + bh )
2

f b ( 0,0 ) = lim
h ( 0 + ah ) + ( 0 + bh ) ù
h ®0 é 2 4
ë û
ah × b 2 h 2 ab 2
= lim = lim
h ®0 h é a 2 h 2 + b 4 h 4 ù h ®0 a 2 + h 2b 4
ë û
ïìb a , a ¹ 0
2


ïî 0 , a = 0
Hence the directional derivative of f at ( 0,0 ) exists in every direction.
v Question
Let z = f ( x, y ) , x = u 2 - v 2 , y = 2uv . Then show that
ìïæ ¶z ö2 æ ¶z ö2 üï
2
æ ¶z ö æ ¶z ö
2
1
ç ÷ +ç ÷ = íç ÷ + ç ÷ ý
è ¶x ø è ¶y ø 4 u 2 + v 2 ( ) îïè ¶u ø è ¶v ø þï
Solution
We have
¶x ¶x ¶y ¶y
= 2u , = -2v , = 2v , = 2v
¶u ¶v ¶u ¶v
Also
¶u ¶v ¶u ¶v
1 = 2u - 2v - 2v
, 0 = 2u
¶x ¶x ¶y ¶y
¶u ¶v ¶u ¶v
and 0 = 2v + 2u , 1 = 2v + 2u
¶x ¶x ¶y ¶y
¶u ¶v ¶u ¶v
Solving these four equations for , , & , we get
¶x ¶x ¶y ¶y
¶u u ¶v -v
= , =
(
¶x 2 u 2 + v 2 )
¶x 2 u 2 + v 2 ( )
¶u v ¶v u
= , =
(
¶y 2 u + v 2
2
) (
¶y 2 u + v 2
2
)
And
¶z ¶z ¶u ¶z ¶v
= × + ×
¶x ¶u ¶x ¶v ¶x
1 é ¶z ¶z ù
= u× -v× ú
( 2 ê
2 u + v ë ¶u
2
)
¶v û
¶z ¶z ¶u ¶z ¶v
& = × + ×
¶y ¶u ¶y ¶v ¶y
1 é ¶z ¶z ù
= v × + u ×
(
2 u 2 + v 2 êë ¶u ) ¶v úû
Hence
ïìæ ¶z ö æ ¶z ö ïü
2
æ ¶z ö æ ¶z ö
2 2 2
1
ç ÷ +ç ÷ = íç ÷ + ç ÷ ý
è ¶x ø è ¶y ø 4 u 2 + v 2 ( ) ïîè ¶u ø è ¶v ø ïþ
…………………………………………………..
Chap 5 – Function of several variables 37

v Question
Let f : ¡ 2 ® ¡ be given by
ì xy
ï ; ( x, y ) ¹ ( 0,0 )
f ( x, y ) = í x 2 + y 2
ï 0 ; ( x, y ) = ( 0,0 )
î
Show that f x , f y exist at ( 0,0 ) but f is discontinuous at ( 0,0 ) .
Solution

f b ( 0,0 ) = lim
( ah )( bh ) where b = ( a, b )
h ( ah ) + ( bh ) ù
h®0 é 2 2
ë û
ab
= lim
(
h ®0 h a + b 2
2
)
Which exists only when b = (1,0 ) or ( 0,1) .
Þ f x & f y exist at ( 0,0 )
Now
xy
lim f ( x, y ) = lim
( x , y ) ®(0,0) ( x , y ) ®(0,0) x 2 + y 2

Let y = mx , then
xy mx 2
lim = lim 2
( x , y ) ®(0,0) x 2 + y 2 x ®0 x + m 2 x 2

m
= lim
x ®0 1 + m 2

Which is different for different m .


Þ f ( x, y ) is discontinuous at ( 0,0 ) .
v Question
Let f : ¡ 2 ® ¡ be given by
ì x2 y
ï ; ( x, y ) ¹ ( 0,0 )
f ( x, y ) = í x 4 + y 2
ï 0 ; ( x, y ) = ( 0,0 )
î
Show that f x , f y exist at ( 0,0 ) but f is discontinuous at ( 0,0 ) .
Solution

f b ( 0,0 ) = lim
( a h ) ( bh )
2 2

, b = ( a, b )
h®0 h éë a 4 h 4 + b 2h 2 ùû
a 2b
= lim 4 2
h ®0 a h + b 2

ì a2
ï , b¹0
=í b
ï 0 , b=0
î
1
Now lim f ( x, y ) is zero along x = 0 and is along y = x 2
( x , y ) ®(0,0) 2
Þ it is discontinuous at (0,0) .
…………………………………..
38 Chap 5 – Function of several variables

v Question
Find the greatest volume of the box contained in the ellipsoid 3x 2 + 2 y 2 + z 2 = 18 ,
when each of its edges is parallel to one of the coordinate axes.
Solution
V = volume of the box = (12 x )( 2 y )( 2 z ) = 8xyz
We need to find maximum of V subject to 3x 2 + 2 y 2 + z 2 - 18 = 0
( )
Consider j ( x, y, z ) = 8 xyz + l 3 x 2 + 2 y 2 + z 2 - 18 = 0
Then
j x = 8 yz + 6l x = 0
j y = 8 xz + 4l y = 0
j z = 8 xy + 2l z = 0
Þ 4 xyz + 3l x 2 = 0
2 xyz + l y 2 = 0
4 xyz + l z 2 = 0
(
Þ l 3x 2 - 2 y 2 = 0 )
l (3x 2
)
- z2 = 0
2 y2 z2
Þ x = 2
=
3 3
Substituting these values in
3x 2 + 2 y 2 + z 2 - 18 = 0
We get
3 x 2 + 3 x 2 + 3 x 2 = 18 Þ 9 x 2 = 18
Þ x = 2 , y = 3 and z = 6
Which gives
f ( x, y , z ) = 8 xyz = 48

……………………………………..
Chap 5 – Function of several variables 39

v Definition
If f : ¡ n ® ¡ , a Î ¡ n then
n
¶f ( a ) ¶f ( a ) ¶f ( a ) ¶f ( a )
Ñf ( a ) = å = + + .... +
k =1 ¶xk ¶x1 ¶x2 ¶xn
v Definition
Let f : G ® ¡ , G is an open set in ¡ n .
i) f is said to have a local maximum at a Î G , if there is a nhood Ve ( a ) such that
f ( x ) £ f ( a ) " x ÎVe .
ii) f is said to have a local minimum at a Î G , if there is a nhood Ve ( a ) such that
f ( x ) ³ f ( a ) " x ÎVe .
v Theorem
Let f : G ® ¡ , G is an open set in ¡ n . If f has a local extremum at a Î G , then
Ñf ( a ) = 0 .
Proof
¶(a)
It is clear that Ñf ( a ) = 0 iff = 0 , i = 1,2,3,...., n
¶xi
Write f ( xi + t ) = f ( x1 , x2 ,...., xi + t ,...., xn ) = f ( x )
If f has a local maximum at a , then
f ( ai + t ) - f ( ai )
£ 0 if t > 0
t
f ( ai + t ) - f ( ai )
Þ lim £ 0 if t > 0
t ®0 t
¶f ( a )
So that £ 0
¶xi
Similarly,
f ( ai + t ) - f ( ai )
lim ³ 0 if t < 0
t ®0 t
¶f ( a )
So that ³0
¶xi
¶f ( a )
Hence = 0 , i = 1,2,3,...., n
¶xi
Þ Ñf ( a ) = 0
Note
There are situations when Ñf ( a ) = 0 but f has no local maximum or minimum at
a . If so and if the sign of f ( x ) - f ( a ) depends upon the direction of x and a , f is
said to have a saddle point at a .

={ END }=