UNIT 5 PARTIAL DIFFERENTIATIION

5.1 Introduction
Objectives
5.2 Functions of Several Variables
5.3 Limits and Continuity
5.4 Partial Derivatives
5.5 The Total Differential
5.6 The Chain Rule
5.6.1 Chain Rule for Functions of two Variables
5.6.2 Implicit Differentiation
5.7 Maxima and Minima
5.8 Jacobians
5.9 Surfaces
5.10 Summary
5.11 Solutions/Answers
5.1 INTRODUCTION
In Unit 1, you learnt about the concept of functions of one independant variable and some
operations concerning functions of one variable. In many physical situations, you feel the
absolute necessity to express quantities of interest in terms of two, three or more variables.
In such circumstances, you feel the need of extending the concept of functions of one
variable to functions of more than one variable.
In this unit, you will learn how to extend concepts you already learnt concerning functions
of one variable to functions of more than one variable. You will also learn some analogous
results when functions of one variable are extended to functions of more than one variable
in general and two or three variables in particular.
After reading this unit, you should be able to :
*
find partial derivatives of functions of two or three variables.
*
find the total differential of functions of two or three variables.
*
find directional derivative in any given direction.
*
find where a function of two variables is having relative maximum or relative
minimum.
*
to evaluate Jacobians of type
a ( x , y ) ln, a ( x , y , z )
a ( u , v ) a ( 14, v, w
5.2 FUNCTIONS OF SEVERAL VARIABLES
In this section, you will know how to extend the concept of function of one variable to a
f un~t i on of two, three or more variables.
5.21 Functions of Two Variables
R denotes the set of real numbers.
Suppose D is a collection of pairs of real numbers ( x , y ) (D is a subset of @ ). Then a real
valued fundion of two variables off is a rule that assigns to each point (x, y ) in D a unique
real number denoted by f ( x, y ).
The set D is called the domain'off.
The set ( f (x, y ) : ( x , y ) E D ), which is the set of values the function f takes on, is called
the range off:
We generally use the letter z to denote the values that a function of two variables takes.
Then we have
z = f ( x , y )
We call the symbol z the dependent variable off and f as the fuoction of two independent
variables: x and y.
Example 1 :
The volume of a right circular cone of base radius r and height h is given by
1
V = - l r ? h
3
Here we call Vas the dependent variable and r and h as the independent variables.
We now define a function of three variables as follows :
Suppose D be a collection of triple of real numbers ( x, y, z ). ( D is a subset of R ~ . )
Then a real valued function of three variables f is a rule that assigns to each point
( x, y, z ) in D a unique real number denoted by f ( x , y, z ).
The set D is called the domain off.
The set
( f ( x , y , z ) : ( x , y , z ) ED) ,
which is the set of values the function f takes on, is called the range off .
We often use the letter w to denote the values that a function of three variables
takes.
Then we have
w=f ( x , y , z )
Here we call the symbol w the dependent variable of function f and x, y, z the
independent variables off.
Example 2 :
The volume of a rectangular box of sides x, y, z is given by
v = x y z
Hare V is the dependent variable and x, y, z are the independent variables.
In general, we can define a function of n variables XI, xz, ...., x,, in a similar
manner.
In this unit we consider functions of two, three or more variables. These functions
are obviously single valued.
f 5.3 LIMITS AND CONTINUITY
In this section, you will learn about the concepts of limit and continuity of functions of
several variables. In this unit we confine most of our discussion of functions of several
variables to those of two variables. We can easily extend these ideas in a similar manner to
functions of more than two variables.
*
You have learnt the concepts of open and closed intervals in connection with functions of
a
one variable. Now, we extend these ideas in connection with functions of two variables.
Let ( .Q, yo ) be a point in D, which is a subset of 2.
Consider the equation
Since ( x, y ) and ( xo, yo ) are in R2, we get
~ ( x , Y ) - ( x a Y o ) l = ~ ( x - x a Y - Y o ) l = ~ ~ ~ - ~ o ~ ~ + ( Y- YO) ~ (2)
From (1) and (2), we get
Squaring both sides of (3), we get
2
( ~ - x o ) ~ + ( ~ - y o ) ~ 1. r . (4)
We note that (4) represent a circle with centre ( xo, yo ) and radius r.
We also see that the set of points whose coordinates ( x , y ) satisfy the inequality
is the set of points in R2 interior to and on the circle given by (4) as shown in Figure 5.1.
From the above discussion we formulate the following defmition
t
(i) The open disc D centred at ( a, yo ) with radius r is the subset of R2 given by
~ ( x , Y ~ : I ~ ~ ~ Y ~ - ~ ~ o ~ y o ~ l ~ ~ ~ .
(ii) The closed disc D centred at (xc, yo )with radius r is the subset of 2 given by
{( 4Y) : l ( x, Y) - ( xo, Yo) l s r).
(iii) The boundary of the open or closed disc defined in (i) or (ii) is the circle
{(x, Y): I(x, Y)- (XO,YO)I = r}.
(iv) A neighbourhood of a point (.Q , yo) in R2 is an open disc centred at
(xo, yo ).
We nate that m open disc does not contain any point on its boundary just as an open
interval does not contain its end points. A closed disc contains all points on its boundary just
as the dosed interval contaius all its boundary points.
Unit :
If the values of the function z = f ( x , y ) can be made as close as we like to a fixed
number L by taking the point ( x , y ) close to the point ( xo, yo ), but not equal to
( xa yo ), then we say that L is the limit of f as the point ( x , y ) approaches the
point(xo,~0).
We write this in symbols as
We read this as : the limit off as ( x , y ) approaches (xo, yo ) is L.
l rl h~si i udi st anceof ( x, y) ~m( xo, yo) mmel y) l ( r - xo12 + 0, - yo)'
is small in some sense, then we mean that ( 4 y ) is close to ( xo, yo ).
Since
and
we have :
which will be small if 6 is sufficiently small. Therefore when we calculate limits,
we may think either in terms of distance in the plane or in terms of differences in
wordhates as shown in Figure 5.2 below.
TbeopensquarcIx-xoI< b, ( y- yoJ<bliesinsi&theopendisc
\/(I - xO)l + (y - yo12 <
6, which also contains the open disc
Ncrw we have two equivalent definitions of limit as given below.
Tbe limit off ( x, y ) as ( x, y ) approaches (xo, ya ) is the numberL if
for any e > 0, there exists a 8 >Osuch that for all points either
(i) 0<4( x - xo12 + 0, - yo)2 < ti i mp~i egt hat l f ( x, y) - LI<E, O~
(ii) O<Ix-xoI<bandO<I y-yoI<8implythatIf(x,y)-L ( < E
t
show that lim f (xy) does not exist.
(ry) -. (0.0)
Solution :
Here the point ( x, y ) can approach the origin infinitely in many ways.
For example, if we approach the origin along x-axis, then y = 0. In this case,
-xC
lim f (xy) = lim - = -1
Civ) 440) (x;y)-.(QO) 2
On the other hand, if we approach the origin along y-axis, then x = 0 a n d k k a ,
case we get
Thus we get different limits depending on how we approach the or i gi n. s ns WI
have shown that for any open disc centmd at the origin, there are points at which
takes on the values + 1 and -1. Thereforef~&ot have the limit as
(x,~)-'(40).
Tbis example leads us to a general rule for nonexistence of the limit, which we
can state as follows.
If we get two or more different values for
as we approach ( xo, ya ) along different paths ( path is another name for a curve
joining (x,y ) to a given point ( xo, yo ) ), then
does not exist
; We state the following theorem without proof.
f (x,y) - Ll and 1im g (xy) = L2,
(*Y) -. ( wd (rv) + ( w d
then
(iv) lim [ K f ( x, y ) ] - KL1, K being any number, and
(xy)-txo,h)
r
A function f (x, y ) is said to be continuous at a point ( a, yo ) if
(i)
f is defined at ( a , yo )
(ii) lim f ( x, y ) exists, and
(+Y)-. (%, Yo)
(iii) lim f Cr , y) =f ( a, yo)
( x Y ) ' ~ . Y ~
Example 4 :
Determine the points at which f (x, y) is continuous if
Solution :
We approach (0,O) along the line y = m x, m being an arbitrary constant. Then
x mx d m
lim f (x, y) = lirrp 1 im
cry) - ~ " 3 ) (v) 440) x -
( ~ y ) +o.o) 2 (1 + m2) = G?
Here we get different limits for different values of m.
Therefore, lim f (x, y ) does not exist.
( r y) - ( O, O)
So f is not continuous at ( O,0 )
when( % b) *( 0, 0) , h+ b2 *Oand
ab
lim f ( x, y) = - =
a2 + b2
f ( a , b )
(+Y) -
Therefore f is continuous at all points ( U, b ) *(O, 0). So we conclude that f is
continuous at all points except the origin.
Here we make a note : if one or more of three conditions in the definition of
continuity off (x,y ) fails to hold, then f is discontinuous at the point under
consideration.
El
Test the continuity at the origin for the Function
Determine the points of continuity of the function
, if (x, Y ) (090)
5.4 PARTIAL DERIVATIVES
In this section , you will learn how to extend the concept of derivative of a function of one
variable to functions of more than one variable . For simplicity, we consider here the
functions of two variables only. Based on similar arguments, we can get the corresponding
results involving functions of three or more variables.
Take two points Po ( xo, yo ) and PI (XI, yl ) of D with value at Poand wl at PI
respectively as shown in Figure 5.3. Then the increment in w in going fiom rn at Po to wl
Aw = Wl-wo = f(x1,y1)-f(xo,yo).
(1)
corresponding to Ar = XI - xo and Ay = yl -m We keep Pb fixed and make PI approach
Y A
0 X
m- 53
some specific. smooth curve in the x y plane . We now s u h e that PI approaches
d w - Aw
= lim - = lim
f(xl*Yl)-f(xo, Yo) (2)
d s PI+PO A s PI-& ~ ( A X ) ~ + ( A ~ ) ~
exists, its value is called the directional derivative of w = f ( x, y ) at ( xo, yo ) in the
direction of L. Here we use the adjedive 'directional' becruse the answer in (2) depends on
the function at PI, the pomt Po and the direction in which PI approaches Po. We will take
up two special cases of the directional derivative. In Section 5.7, we discuss #e general case
of the directional derivative in any dhection.
We f h t make PI approach Po along the line y = yo parallel to the x-axis. Seumdly we
make PI approach Po along the line x = xo parallel to the y-axis. In the first case where PI
approaches Po along the liney = yo, we get
We d l the resulting limit the partial derivative of w = f ( x, y ) with respect to x at
Po ( xo, yo ) . From (3) we note that this is just the ordinary derivative with respect to x of the
Eunction f (x ) = f ( A yo) obtained from f (x, y ) by holding the variable y as constant. This
also measures the instantaneous rate of change at Po ofthe function w = f ( x, y ) per unit
a w
change in r We use the notation - or to denote the partial derivative of w with respect
a x
to x. By deleting the subscript 0 every where in (3), we obtain the partial derivative of w at
( x, y ) and the same is given by
We also note that in order to calculate such a partial derivative fiom the equation for w, we
simply apply the rules of ordinary differentiation, treating y as constant.
Considering the second case and proceeding on similar lines, we get the partial derivatives
a w
of w = f (4 y ) with respect to y, denoted by f, or - as
a Y
We make the fdlowing observations :
(1) The definition allows us to calculate partial derivatives much in the same way as we
calculate the ordinary derivatives by allowing only one of the variables to vary at a
time. So we can use all those formulae learnt on differentiation of functions of one
variable for partial derivatives.
a w a w
) The partial derivatives - and -give us the rate of change of w with respect to
ax a y
each of the variables x and y with the other one held constant. We will discuss in
' Section 5.7 about how w changes when bothx and y change simultaneously.
a w a w
( 3) Even though we calculate -and -by holding one of the variables as constant,
a x a y
we note that each is a function of both the variables x and y.
Now we can easily obtain the partial derivatives of a sum, difference, product and quotient
of two functions much in the same way as that of ordinary derivatives in view of
obsewadon 1. We h n now easily extend the definition of partial derivatives to functions of
three or more variables.
Higher Order Pnrtiol Derivatives : The partial derivatives of a given function of two
variables rre also functions of two variables and we can some times differentiate them
further to get higher order partial derivatives.
Let w = f ( x, y ) and its partial derivatives be defined over a domain D. Then for ( 4 y ) E Dl
we define the second order partial derivatives as
a2w
-
a aw
a i - 5. - G ( a x ) .
If w = f ( 5 y ) and its first and sewnd order partial derivatives are defined over a domain D,
then for ( 5 y ) ED, we define third order partial derivatives as :
a2w a2w
We call the partial deriva'tives - - as mixed partial derivatives of second order
ax a y ' a y ax
and
a3 w a' w a3 w
--
a3 w
etc. as the mixed partial derivatives of
a x a y a x ' a x a # ' a y 2 a x g a y ax a y
third order.
On similar lines we can defme partial derivatives of orders higher than three.
Example 5 :
Obtain all the second order partial derivatives for w = 2 yZ -xyS
Solution :
Find ux, u,, u,, u,, 'iri and u, if u = log(ux + by), where a and b are arbitrary
constants.
E4
Obtain all the nine second order partial derivatives and verify k t all the three
pairs of mixed partial derivatives are equal if u = n)S - 2 + l?jcz.
We note that if w = f (x, y ) a d its partial detivatives are continuous, the order of
differentiation is immaterial, that is,
Example 6 :
Consider z = x cosy - y cos L
-
a z
= cosy + y s h x , - - - x s i n y - a x ,
a x a Y
$2 azz
We note here that - = - '
a x a y a y a x
5.5 THE TOTAL DIFFERENTIAL
We first prove the increment theorem for the function of two variables.
1
Let the function w = f (x, y ) be continuous and possess partial derivatives f , fY
I
throughouttheregioaR: (x-xo( <h,(y-yp(<kofthexyplane.htf,andf,be
continuous at (xa, yo ). Let
Aw f ( x o + Ax, + AY) - f ( x o , y o ) . (1)
Then
Aw - f , ( xoyo) AX + ~ ~ ( X O Y O ) Ay + el Ax + ezAy (2)
wbexe el and ez -, 0 when Ax and Ay 4 0.
Proof :
We note here that the region R is a redangle with centre at (xo, yo ) with sides 2h,
2k.Wechoost AxandAytobcsmallsothatthepoints(aa,~),
(xo+Ax,yo+Ay), (m+&,yo),(yhyo+Ay)alllieinR. Wealsoassumethat
f is continuous and has partial derivativesf, and& throughout the rectangle R.
We can use mean value theorem of a function of one variable as all conditions
required are now satisfied.
We note that the increment Awis the change in f fromA ( xo , YO ) to
B(xo+ Ax,yo+ Ay)inR.
We write
Aw = Awl + + Aw;!
where
Awl = f(xo+Ax,Yo)-f(xGYo) ( 5)
and
Here Awl is the change in w fromA to C and Awz is the change in w from C to B.
In Awl , we hold y = yo fixed and have an increment of a function of x that is
continuous and differentiable. Applying mean value theorem, we get
Awl = ~ ( X O + A X , ~ O ) - ~ ( . ~ ~ , Y O ) f x ( x i , ~ o ) ~ (7)
for some xl between ro and xo + Ax i.e. xo < XI < xo + br
Similarly applying mean value theorem for Awz after holding x + xo + Ax ftxed
and having an increment of a function y that is both continuous and differentiable,
we get
for some yl between yo and yo + A y i.e. Yo < yi < Yo + AY-
So we get
AW = f x ( x l , y ~ ) A x + f , ( x o + A x , y ~ ) A ~ (9)
forsomexl i nm <XI < x o + ~ x and f or someyl i n~o<yl <yo +AY.
Using the fact thatfx and f , are continuous at ( xo, yo ), we get
fx(xl , yo)-, fx(ro, ~o) (10)
and
f v ( ~ o + Ax , ~ l ) - , f , ( x o , ~ o ) (10')
as A x and Ay approach zero.
So we can now write
fx(x1,yo) ' fx(ro,Yo)+el
and
where ~1 and EZ both approach zero as Ax and Ay approach zero.
!
Making use of (1 1) and (1 1') in (9), we get
AW = f , ( xqyo) Ax+f , ( xqyo) Ay+~1A~+~2Ay
where^^ and ez+Oas Axand Ay-0.
Now we can extend (2) easily to functions of three variables as follows :
t
Let w = f (x, y, z ) be continuous and'possess partial derivativesf,, fv , fi at and in
some neighbourh60d o f a e point ( m, yo, zo ) whose partial derivatives are
continuous at.that point. Then we have
where EI, ~ 2 , ~3 + 0 as AX, Ay, A z + 0.
Now we define the differential of w = f ( x, y ) by the formula
dw = fxAx+fYAy.
We can write, in view of (3), that
In the increment Ax and A y are small, then the differential dw is a good
approximation for the change Aw in the value of w = f (x, y) as we move from the
point(x,y)to (x+Ax, y+Ay),since
Aw - d w
E ~ A X + EZAY - 0 as Ax, Ay +O.
As in the case of functions of one variable, writing Ax and A y as dx and dy
respectively, we get
It is of interest to note here that the function w = f ( x, y ) is differentiable in R
whenever (13) holds.
Similarly, for a function of three variables w = f (x, y, z ), we defme the
differential w by
Note :
aw a w aw
In (14), the separate terms - & , - dy and - dz are sometimes called
ax: @ az
partial differentials of w with respect to x, y and z respectively.
We can then call the sum of these partial differentials of w as the total differential
dw. So dw is called as the total difPerential of w, when w is a function of more
than one variable.
Example 7 :
Find the total differential of w = x sin y - y sin x.
Solution :
- a w=, y - y e o s x , - a w = x ms y - sinx
a x a y
To give you an idea about the use of differentials, we consider a rectangle with
sides x and y. The area of this rectangle is
If we increase the dimensions of the rectangle to x t Ax and y t Ay, the change in
area is
The differential estimate for this change in area is
r :
The error of our estimate, the difference between the actual change and the
estimated change, is the difference A A - d4 = A x Ay
Example 8 :
Use the differential to estimate
Solution :
We know
m = 5, Vian = 10.
What we need to find is an estimate for the increase of
f (x, y) = xH yY3
fromx=25, y = lOOOtox=27, y=1021.
gives
1 1 % -45
d f = - ~ - ~ ~ ~ d x + - x y dy.
2 3
Putting x = 25, y = 1000, dx = 2, dy = 21, we have
Hence
DiilerQ0l.l- E5
Find the total differential of w = 2 + 2 xy - 3y2.
Use differentials to fmd the approximate value of a m.
E7
Estimate by a differential the change in the volume of a right circular cylinder if
the height is increased from 12 to 12.1 cm and the radius is decreased from 6 to 5.8
cm.
5.6 THE CHAIN RULE
In this sedion we obtain the chain rules for finding partial derivatives of functions of two or
three variables. In a similar way we can get the chain rules of partial differentiation for
functions of more than three variables.
5.6.1 Chain Rule for Functions of Two Variables
If w = f (x, y) has continuous partial derivatives fx and f , and if x = x (t), y = y (t) are
differentiable functions oft, then the composite w = f (x (t), y (t)) is a differentiable
function of 1.
In this case we get,
Wetaketoasanyvalueoft.Thenxo=x(to)andyo=y(to).Ifwenowtake Ax, Ayas
changes that &mu in x and y when t is changed from to to to + A 1, we then get from (13) of
Section 5.5 that
where el, ep +O as Ax, Ay 4.
Dividing both sides of (2) by At , we get
We also note that, as Af + 0,
So we get
The equations we want to prove are simply the statements that (4) holds at every admissible
value oft, which infact it does.
We extend (4) to functions of three variables as follows ;
If w= f ( &y, z) hascontinuouspartial derivativesandx=x(t),y=y(t),z=z(t )are
differentiable functions oft, then the composite f ( x ( t ), y (t ), z ( t ) ) is a differentiable
function of t. In this case we get
If w = f (x, y ) and if x = x ( r, s ), y = y ( r, s ), are any functions, then the composite
f ( x ( r, s ), y ( r, s ) ),when defined, is a function of r and s. Ifx, y and f have continuous
partial derivatives, then the partial derivatives of w with respect to r and s exist and in this
case we get
.
aw " a x aw*
- r - + - (6)
ar ax ar g ar
and
aw a ~ a x + a w a y
- - - - (9
as ax as g as
We can extend (6) and (7) to functions of three variables as follows :
I f w=f (x,y,z)andifx - x( r , s ) , y = y( r , s) , z = z(r,s)areanyfunctions,thenthe
composite function w = f (x (r, s), y (r, s), z (r, s)), when defined, is a function of r and s. If
x, y, z and f have continuous partial derivatives, then the partial derivatives of w with respect
to r and s exist and we get
a~ a w a x + a w a y + a ~ a ~
-
I -
ar ax ar g ar az ar
(8)
aw " a x aw* a wa z
- I--+-
+ - -
a~ ax a~ g a~ a~ a~
(9)
To show what variables are assumed independent in computing a partial derivative, we
employ the following notations.
8w
" means -with x and y independent.
( ax
x and z independent.
with y, x and t independent.
Example 9 :
Ifw = ~ + y - r + s h t a n d x + y - t,Bad(r) and(ii) -
( ax) yJ (z)4z
Solution :
(i) Witb x, y, z independent, we have
t = x+y, w = 2+y-z+si n(x+y).
Therefore
(ii) Witb x, t, z independent, we have
y - t-X, W - g+( t - x) +s i nt .
Thus
Solution :
Therefore we have
Similarly we have
If z be a function of x and y and x' = e ' + e -", y = e ' - e ", prove that
~f v = f , prove that
(2,
X
(Hint:TakeE = - q = :, then v = f(5, q)
z '
El l
Iff ( x, y, z )
0 , prove that
5.6,2 Implicit Differentiation
You would recall that the method of implicit differentiation was used in Unit 2 to find the
derivative of a function defined implicitly by an equation. In this section you will see how
the derivative of an implicitly defmed function can be obtained through the use of partial
derivatives.
Let
where
Y = Q , ( x )
By the chain rule ( with x = x, y = + (x)), we get
We have assumed that y is a differentiable function of x.
Example 11 :
If Q, is a differentiable function such that y = Q, ( x ) satisfies the equation
2 + yZ + sinxy = 0 ,
fmd *
dr '
Soluthon :
Let
Then
Hence
Find &, given .hat y = 4 ( r ) satisfies each equation.
a5
(i) x s e c y + y s e c x - 6 = 0
(ii) 21ny- $ +2 -4.
PuttlDaPe!renti.t&
5.7 MAXIMA AND MINIMA
In this section you will learn about the maxima and minima of functions of two variables.
We give some definitions to start with.
A function of two variables f ( x , y ) is said to have an absolute or global
maximum at Crayo) in a region R if
f ( x, y) s f(xo,yo)forall(x,y)inR.
An absolute or global minimum of f in R occurs at ( xo, yo ) if
( f ( x, y) 2 f(xo,yo) forall 4 x 9 ~
in R.
Definition 2 :
A relative or local maximum off occurs at ( xo, yo) if a circle D about
(m,yo)eXistswithf (x,y) sf (xo,yo)forall(x,y)intheinteri~~ofD.
A relative or local minimum off occurs at ( ~ a , yo) if a circle D about
(~~,yo)existswithf (x,y) zf(xo,yo)forafl(x,y)intheinteriorofD.
We refer to both maxima and minima as extrpma, and relative maximum
and relative minimum as relative e m m a
A function f of two variables is said to have a crltical point at ( xo, ya )in
the domain off if either
(i)
both partial derivatives of fare zero at (xo, yo )
or
(ii) at least one of tpe partial derivatives fails to exist at ( xo, yo ).
Suppose that a relative maximum value of the function z = f ( x, y ) occurs at an
interior point ( a, b ) of the domain of the function and that tY and both exist
1 ax ay
at ( a, b ) as shown in the figure 55.
Then we observe the following facts :
(1) x = a is an interior point of the domain of the curve CI : z = f ( x, b ) in which
theplaney= bcutsthesurfacez= f(x, y).
(2) The function z = f (x, b ) bas a relative maximum at x = a.
(3) The value of the derivative of z = f ( x , b ) at x = a is therefore zero.
Since this derivative is precisely ( q( 4b) , we condude that
On the lines of similar argument with the function z = f ( a, y ), we get
Thuswehavefx(a, b) =O =f , ( a, b).
So we have a necessary condition for f to have an extreme value at an interior point
(a, 6) :
Definition 4 :
The point where there is no relative maximum or no relative minimum
but fx and f, are both zero, is called a saddle point
We apply the second derivative test to determine whether a function
z - f ( x, y ) has a relative maximum or minimum value at a point P( a, b ) where
fx and f, both vanish.
We assume that f and its first and second order partial derivatives are continuous in
some region R about P ( a, b ) as shown in the figure 5.6. We take
S ( a + h, b + k ) to be a point close enough to P so that P S lies in R. We take the
parametric equation of P S to be
We study the values of F ( x, y ) along P S by considering the function
F ( t ) = f ( a +h t , b +k t )
We know that F is a differentiable function of t since the first partial derivatives
are continuous and x = a + th, y = b + tk are differentiable functions of t. Using
\
1
the chain rule, we get '
Further, we know that F ' is continuous on the closed interval 0 s t s 1 since
t
2 and ' are continuous throughout R. Once again, making use of chain rule, we
I aY
i get
F " (t) - h
aF ( t ) aF1 ( t )
ax
+ k
ay
Now, F ' ( t ) satisfy the conditions of Taylor's theorem on the closed interval
[ 0 , l 1, namely, F and F ' are continuous in ( 0 , l ) and F ' is differentiable in open
interval ( 0, l ) . So we have
forsomec i nO<c < 1.
F (1) = f (a + h, b + k)
F ( O ) - f ( a , b )
F' ( 0) - y ; ( a , b ) +y , ( a , b )
Substituting these values in (4), we get
f (a + h, b + k)
f (a, b) + hfx (a, b) + k fAa, b)
Supposingfx = 0, f, = 0 at ( a, b ), we wish to determine whether f ( x, y ) has a
maximum or minimum at ( a, b ). We write ( 5 ) as
Since a minimum or maximum valire of f at ( a, b ) depends on the sign of
f ( a +h , b + k ) - f (a, b),weareledtoconsiderthesignof
or, considering c very small,
Q(0) = h2f= (a, b) + 2 h kfv (a, b) + 2fw (a, b) (7)
Multiplying both sides of (7) by fn(a, b), we get
fx Q (0) - (hf, + kfv)2 + (firfw - fv 2, 2 (8)
The sign of Q(0) can be determined from (7). Consequently we get the following
+ri for the behaviour off ( x, y ) at ( a, b ).
1
(1) 1ffa<0orf, <0rndfaf, -fv2 > 0 at(a,b),thenQ(O)<Oforallsmall
non z m values of h and k and f has a relative maximum value at ( a, b ).
(2) ~f f , r ~or f , >~andf , f , - f v2 > O a ( 4 b),then Q (O)>Oforall
small non zero values of h and k and f has a relative maximum value at ( a, b ).
1
(3) If f, f,- f$ < 0 at (a, b ), then it can be shown that there are
combinations of arbitrary small non zero values of h and k for which
Q ( 0 ) > 0 and Q ( 0 ) < 0 .Thus, arbitrarily close to the point !
Po ( 4 b, f (a,b))onthesurfacez = f (&y)therearepointsabovePa
and also points below Po. The function f therefore has a saddle point at
(4 b)-
(4) Finally, iff= fw - f 2v = Q we can draw no conclusion about the sign of
Q ( 0 ) and we requh some other test to settle this case.
Now we give a brief summary of maximum and minimum tests as follows :
If z = f (4 y ) is continuous, then extreme values of f may occur at
(i) boundary points of the domain of f
(ii) interior points where f r = fy = 0
(iii) points wherefr orfy fails to exist.
Further, iff has continuous first and second order partial derivatives on some open
disccontaining(a, b)andif f r( a, b ) = f y(4 b ) = 9 then
(i) f, < 0 or f, < 0 and f,f, -fg > 0 at ( a, b ) implies that f has a local
maximum at ( a, b ).
(ii) f, > 0 or f, > 0 and f,f, -f& > 0 at ( a, b ) implies that f has local
minimum at (a, b ).
(iii) f, f, - fg < 0 at ( a, b ) implies that the point ( 4 b ) is a saddle point of
f ;
(iv) fa fyv - f$ = 0 at ( a, b ) implies that the test is inconclusive as this case
requires some further investigations. ,- -
Find the maximum and minimum values of
Tbe likely points where f (5 y ) has maximum or minimum, arc
(O,O), (0, * 0, (* LO).
2
fa - 4 - 1 2 2 , fv - 0, f, = - 4 + 12y.
The results are summarized in the following table.
I
Find the maximum and minimum values of xy + 27
(i.4
Point f x x f v v - f x ~ fx x
- 16
-
32 4
( 0 , - 1 ) 32 4
( 1 9 0 ) 32 -8
( - 1 , 0 ) 32 -8
Investigate the maximum and minimum values of
2 y 2 - 5 x 2 - 8 x y - 5 y 2
5.8 JACOBIANS
In this section, you will learn about Jawbians. Jawbian's play an important role in
evaluation of multiple integrals when the variables of an integral are changed by a suitable
transformation.
If x = g ( u, v ) and y = h ( u, v ) be differentiable then the Jacobian of x and y with
respect to u and v, denoted by is given by
a ( 4 v)'
a x a x
--
~ ( x , Y ) = a~ a v
a (u, V ) Q
au a v
If x = g ( u , v, w) , y = h (4 v, w), t = j ( u, v, w ) aredifferentiable,thenwedefme
the Jacobian of the transformation Erom a region U in u v w -space to a region W in
a x ax a x
-
a u a v Z
~ ( x , Y , z ) = Q 2 a
a (u, v, W) a u a v a w
a t a t a z
---
a~ a~ a~
max / min / saddle point
saddle point
minimum value - 1
minimum value - 1
maximum value 1
maximum value 1
In a similar manner we can defme Jacobians of order higher than three.
Note :
-is sometimes denoted by J
a (u, V,W)
We now prove some important properties of Jacobians. We prove these properties
for Jacobians of second order only. You can get similar such properties for
Jacobians of order higher than 2.
Thearem 3 :
then
JJ' = 1.
Consider u=+( x, y) and v=Ip(x,y).Wecansolvex,yintermsof u and v
andget x = +i ( y v)and y=Ipi (u, v).
We also get
and
av av
dv = - dr+- dy.
ax ay
So we get that
au
-
au
1 , - =
au
0, - = 1 and -&
au av av av ' O.
Now we have
av ax av 9 a~ ax + av ay
+ -
ax: au ay au ax av ay av
= I - -
Theorem 4 :
If u and v are functions ofp and q and p and q are functions of x and y, then
h o f :
,
Since u and v are functions ofp and q, we get
We also get
and
Now consider
by interchanging rows and columns of the second determinant.
So we get now
If the functions u and v of two independent variables x and y be such that
f (4 v) = 0, then
a0 = 0.
a (x,Y)
Proof :
Sincef ( u , v)=O,wehave
and
Eliminating a from the above two equations
a ~ ' a v
we get
by interchanging rows and columns of the determinant
Therefore,
Example 13 :
I f x =r ws O, y =rsi nO, z=z,find
a (x, Y, 2)-
a (r, 0, Z)
Solution :
a , , 1 1 1 h e 8 o
a (r, 0, Z) a r ae a z I
find d.
find a (XI , XZ, 1 3 )
a ( XI , ~ 2 , ~ 3 ) a Cvl, m PI'
find
a (x9 YJ) and a (u, V, W)
a (u, V, W) a (x, Y, 2) '
5.9 SURFACES
In this section, you will leam'about some basic ideas about surfaces in general.
We fhst give its definition.
1
/
A swface is tbe locus of a point whose cartesian coordinates x, y, z are fune
tions of two independent parameters u and v.
-
So we get
to be the parametric equations of a surface. When we represent a surface in this way, we call
this as the Gaussian form of representing a surface; we also call the two parameten u and v
as the c ~i l i ni a r coordinates of a current point on a surface.
For any point ( x , y , z ) on the surface, we can uniquely determine the values of u and v and
we refer to this as ( u , v ).
If we eliminate parameters u and v in the parametric equations of a surface, then the relation
so obtained is of the form
and this also represents a surface.
When a surface is represented in this fom, we call this as the constraint equation of a
surface. Consider
Eliminating u, v in (I), we get
2 - # = z .
Next take
Again on elimination of u and v in (2). we get
Finally consider
x=u&hv, y=us i nhv, z =~
on elimination of u and v m (3), we get
2-3 = z.
.
We see that the dBerent parametric equations (1) , (2) and (3) give rise to the
samesurface2-9 = z
Thmfme we note that the parametric equations of a given surface are not
necessarily unique. Further, we obsewe that the constraint equation 3 -y2 - z of ,
the surface represents the whole of i t The parametric equations of the surface
represent a portion of i t
Fur example, the rametric equations (3) of a surface represent only the part of
t hesOmce*2-9 = z for which z a 0 as u is real.
Therefore we note here that the parametric equation and the constraint equation to
a surface are not always equivalent.
If we write the equation to a surface in the form
z - f(x,y)
We call this as ~o n g e ' s form of the equation to a surface.
In Monge's form of equation to A surface, we can regard x and y themselves as
parameters
Tangent Plane and Nomal Line at a Point bn a Sotface.
Let
F(x,y,z) = 0
represent a surfice. We take P (5 y, z) and Q (x + Ax, y + by, z + Az) as two points
close to each other on the surface represented by equations (4).
Let arc PQ be As and chord PQbe k.
We know that
From our knowledge of coordinate geometry of three dimension, we know that the
direction satics of PQ are
Ax A J &
L e., 2 &,
We write the above as
We al s onot e t hr t c s ds +O, Q+P, ~t e ~t oa tangent 1inePT;
Now noting that the coordinates of my point on arc ?Q arc functions of s only, we
get the direction cosines of PTto be
& 4 dz. -
ds' ds' ds
Now differentiating both sides of the equation (4) witb respect to s, we get
i
1
Again using our knowledge of coordinate geometry of three dimensions, we get
h (6) that line with direaim mines $,%,$ is pexpendiculn to the line
a~ a~ aF
having direction ratios as
Since we can take different curves joining Q to P , we get a number of tangent
a~ a~ aF
lines at P and the line with direction ratios - - -will be perpendicular to all
ax' g' a
such trngent lines at P. Thus we set that all tangent line at P lie in a plant rhrougb
P perpendicular to the line with direction ratios
So we get the equation of the tangent plane to surfrce given by the equation (4) at
point P to be
Where ( X, Y, Z ) are the cwrent coordinates of my point on the tangent plane.
From what we leunt in coordinate geometry of mime dimensions,we get the
equation of a plane through a point ( xo ya zo ) having a line witb direction ratios a,
b , c asnormal tobe
We also get the equations of a normal to the surfrce at P which is a line through P
perpendicular to the tangent plane at P as
X-x Y-y z-z
I - - -
Fx Fy Ft
Example 14 :
,-
3
Find the equations of the tangent plane and the normal to the s mc e xyz - a at
(xlylzl)whercaisaconstant.
So we get the equation of the tangent plane at any point ( XI, yl, n ) as
i.e.,xyln+xlyzl+xlylz- 3xlylzl
The equation of normal at ( xl YI a ) becomes
x1(x-x1)- y1(y-y1)- s ( z- 21) .
Find the equations of the tangent plane a d the nonnal to the s d c e
3 6 3 + 9 9 + 4 z = 72 at(O.2.3).
5.10 SUMMARY
(1) f ( 4 y ) is said to tend to the limit L as (x, y) + (xa yo), written as lim - L, if for any
k Y ) ~ f ( l s Y )
e > 0, there exists a 8 .' O such that either
!
(2) f ( x, y ) is said to be 'continuous at a point (xa yo) if
i
( i ) fisdefinedat(xo,yo)
( ii ) lim f (x. y) exists
C ~ ; Y ) - . ~ Y . )
a w
(3) The partid derivative of w = f (x, y) with respect to x , denoted by f, or - , is
a x
a w
x + A X, Y) - f (x, y)
- - - l h f (
a x AX
9
Ax-0
provided the limit exists.
(4) The second order partid derivatives of w - f (x, y) are fn, fm f* fp wheae
(5) The differential of w - f (x, y) is defined by the formula d w - f, dr +h dy.
(6) If w - f (x, y, z) and i f x = x (r, s), y = y (r, s), z - z (r, s), then
(7) Iff (x, y) = 0 and if y is a function of x ,
then
(8) Iff ( x, y ) has continuous first and second order partial derivatives on some open disc
containing ( a , b ) and iff, (a, b) = 0 = f , (a, b), then
i
(i) f, < 0 and f , f, - f ' , > 0 at (a, b) =+ f has a local maximum at ( a, b ).
1
(ii) f, > 0 and f,h - f$ > 0 at ( a, b ) *f has a local minimum at ( a, b ).
I
(iii) f, f , - f$ < 0 at (a, 6) * the point ( a, b ) is a saddle point off , i.e., ( a, b ) is
neither a maximum nor a minimum.
(iv) f, f, - f$ = 0 at (a, 6) the test is in conclusive; this requires further
investigation.
9 . I f x =g ( u , v , w ) , y = h ( u , v , w) , z = j(u,v,w)aredifferentiable,thentbe
Jacobian from a region U in uvwspace to a region Win x y z-space, denoted by
ao is defined as
a (4 v, w)'
(10) The equation of the tangent plane to tbe surface f ( x, y, z ) = 0 at the point
( 5 y , z ) i s
(1 1) The equation of the nonnal line to the surface f ( x, y, t ) = 0 at the point ( x, y, z ) are
X- x Y- y Z-z
I -
b fY fi '
5.11 SOLUTIONS / ANSWERS
El
f ( x, y ) is discontinuous at the origin (0 ,0).
f (x, y ) is continuous at all points except at the origin (0 ,0). .
I
2
E3 a b a
Ux = - uy = -
a + by' ax + by'
2
u V = - - * and
(af b y ) ' = -(ux+by)'
3
u*= -2Ozx +2p, u,,. = 6xy, u, = 0,
2
u, = 3y2+2xz, uu, = -5x4+2xy, uy, = 2, Up = 3y +az, u a = -5x4+2xy
and u, = 2.
V = nr 2h dV = V, d r + V, = 2zrhdr+m2dh
For r = 12, h = 6, dr = 0.1 and dh = - 0.2,
dV = 2nx12x6x0. 1- nx12x12x0. 2
.I 4 . 8 ~ - 28. 8~ = -203.
(i) Letf(x, y) - xsecy+ysecx-6
.'. fx = secy+ysecx tanx
f , = xsecy tany+secx
Hence 9 I -fi -secy+ysecx tanx
dr f, secx+xsecytany
(ii) Leff(x, y) = x 3 ~n y - y 3 +x 2 - 4
dy f x=3x2yl ny+2r y
Hence - = - -
& f , ~ ~ - 3 ~ ~
The value of the function attains a minimum at x = y = 3 and the minimum value
is 27.
El4
Function has maximum value 0 at (0,O ) and minimum value - 81 at ( d, D ).
El5
:. Equation of tangent plane at (0,2,3) is
o(x-xI)+~~(~-~I)+~(z-zI) - 0
i . e. , 9( y- yl ) +( r- z1) = 0
Equation of normal at (0,2,3) is