Function of Two Variables Notes - 2
Function of Two Variables Notes - 2
Recall that the chain rule for a function of one variable says that for differentiable
function f and g ,
d
f ( g ( x ) ) = f ( g ( x ) ) g ( x ) .
dx
Now, we extend the chain rule to functions of two or more variables.
Theorem 1: If f ( x, y ) is a differentiable function of x and y , and both x and y are
differentiable functions of a single variable t , then f ( x ( t ) , y ( t ) ) is differentiable with
f
g ( t + t ) g ( t )
= lim
t 0
t 0
t
g ( t ) = lim
For
simplicity,
f = f
if
we
write
( ( x ( t + t ) , y ( t + t ) ) ) f ( ( x ( t ) , y ( t ) ) ) .
t
x = x ( t + t ) x ( t ) ,
y = y ( t + t ) y ( t )
and
( ( x ( t ) , y ( t ) ) ) = g ( t ) = lim ft .
t 0
f
f
x + y + 1x + 2 y
x
y
...(1)
x ( t + t ) x ( t ) dx
y ( t + t ) y ( t ) dy
x
y
= lim
=
and lim
= lim
=
. Also,
t 0 t
t 0
t 0 t
t 0
t
dt
t
dt
since x ( t ) and y ( t ) are differentiable, they are continuous as well, so that
Now, lim
lim x = lim x ( t + t ) x ( t ) = 0,
t 0
t 0
lim y = lim y ( t + t ) y ( t ) = 0.
t 0
t 0
Consequently,
( x, y ) ( 0,0 )
t 0
lim
t 0
f f dx f dy
=
+
t x dt y dt
d
f
dx f
dy
f ( x ( t ) , y ( t ) ) = ( x ( t ) , y ( t ) ) + ( x ( t ) , y ( t ) ) .
dt
x
dt y
dt
dw w dx w dy
=
+
.
dt x dt y dt
To remember it, we use the tree diagram as shown below:
z
z
= 2 xe y ,
= x 2e y , x ( t ) = 2t , y ( t ) = cos t .
x
y
By applying Theorem 1, we obtain
dz z dx z dy
=
+
dt x dt y dt
= 2 xe y ( 2t ) + x 2 e y ( cos t )
2
dw w dx w dy
=
+
dt x dt y dt
= 2 xy ( cos t ) + ( x 2 2 y ) et
= 2sin t ( et ) cos t + ( sin 2 t 2et ) et .
When t = 0 , it follows that dw dt = 2 .
Remarks: In Examples 1 and 2, we could have first substituted for x and y , then computed
measures the labour force (in thousands of workers). Suppose that when l = 2 and k = 6 , the
labour force is decreasing at the rate of 20 workers per year and capital is growing at the rate
of $400,000 per year. Determine the rate of change of production.
Solution: Let t denote the time in years and g ( t ) = P ( k ( t ) , l ( t ) ) . From the Chain rule, we
have
g (t ) =
P dk P dl P
P
+
=
k (t ) +
l (t ) .
k dt l dt k
l
P
P
( 6,2 ) 2.1935 and
( 6,2 ) 19.7411 .
k
l
Also, we have k ( t ) = 0.4 and l ( t ) = 0.02 . Thus, we obtain
g (t ) =
P
P
k (t ) +
l ( t ) 2.1935 ( 0.4 ) + 19.7411( 0.02 ) = 0.48258.
k
l
This indicates that the production is increasing at the rate of approximately one-half unit per
year.
Example 4: Two objects are travelling in elliptical paths given by the following parametric
equations.
x1 = 4cos t and y1 = 2sin t
...first object
...second object
The distance s between the two objects, as a function of four variables, is given by
s=
( x2 x1 )
+ ( y2 y1 ) ,
dx1
dy
dx
dy
= 4sin t = 0, 1 = 2 cos t = 2, 2 = 4 cos 2t = 4, 2 = 6sin 2t = 0.
dt
dt
dt
dt
Thus, using the appropriate Chain rule, we obtain that the distance is changing at the rate of
Also, at t = ,
Remarks: We can easily extend the Chain Rule given in Theorem 1 to the case of a function
f ( x, y ) , where both x and y are functions of two independent variables s and t ,
z z x z y
=
+
;
s x s y s
z z x z y
=
+
.
t x t y t
Solution: First, we find the partial derivatives of x and y with respect to s and t .
x
x
y 1 y
s
= 2 s,
= 2t ,
= ,
= 2 .
s
t
s t t
t
Hence, by the Chain rule given in Theorem 2, we get
w w x w y
1 6 s 2 + 2t 2
1 s
=
+
= ( 2 y )( 2 s ) + 2 x = 2 2 s + 2 ( s 2 + t 2 ) =
,
s x s y s
t
t
t t
2
3
w w x w y
s s
s 2 st 2 s
=
+
= ( 2 y )( 2t ) + 2 x 2 = 2 2t + 2 ( s 2 + t 2 ) 2 =
.
t x t y t
t2
t t
t
Remarks: The Chain Rule in Theorem 2 can be extended to any number of variables. For
example, if w is a function of n variables x1 , x2 ,..., xn , where each xi is a differentiable
function of the m variables t1 , t2 ,..., tm , then for w = f ( x1 , x2 ,..., xn ) , we have
w w x1 w x2
w xn
=
+
++
t1 x1 t1 x2 t1
xn t1
w w x1 w x2
w xn
=
+
++
t2 x1 t2 x2 t2
xn t2
w w x1 w x2
w xn
=
+
+ +
tm x1 tm x2 tm
xn tm
The tree diagram below gives the Chain rule for a function of three variables x, y , z , where
each of these intermediate variables are functions of two variables u and v .
dw dx = Fx ( x, y )( dx dx ) + Fy ( x, y )( dy dx ) .
Since w = F ( x, y ) = 0 , we have dw dx = 0 , so that Fx ( x, y )
dx dx = 1 , so that if Fy ( x, y ) 0 , we get
theorem of implicit differentiation:
dx
dy
+ Fy ( x, y ) = 0 . But
dx
dx
F ( x, y )
dy
= x
. We formally state this result as a
dx
Fy ( x, y )
F ( x, y )
dy
.
= x
dx
Fy ( x, y )
The above theorem can be extended to differentiable functions defined implicitly with any
number of variables.
Example 5: Find dy dx , given y 3 + y 2 5 y x 2 + 4 = 0.
Solution: Let F ( x, y ) = y 3 + y 2 5 y x 2 + 4 . Then we have
Fx ( x, y ) = 2 x, Fy ( x, y ) = 3 y 2 + 2 y 5 .
Fx ( x, y , z ) = 6 xz 2 xy ,2
Fy ( x, y, z ) = 2 x 2 y + 3z ,
Fz ( x, y , z ) = 3x 2 + 6 z 2 + 3 y.
Hence, we obtain
Fx ( x, y , z )
z
2 xy 2 6 xz
=
=
x
Fz ( x, y , z ) 3 x 2 + 6 z 2 + 3 y
F ( x, y , z )
z
2 x 2 y 3z
= y
= 2
.
y
Fz ( x, y, z ) 3 x + 6 z 2 + 3 y
Directional Derivatives:
Recall that by Chain Rule, if f ( x, y ) is differentiable, then the rate at which f
changes with respect to t along a differentiable curve x = g ( t ) , y = h ( t ) is
df f dx f dy
=
+
.
dt x dt y dt
At any point P0 ( x0 , y0 ) = P0 ( g ( t0 ) , h ( t0 ) ) , this equation gives the instantaneous rate of
change of f with respect to increasing t and therefore depends, among other things, on the
direction of motion along the curve.
Example 1: Suppose the direction of motion is a straight line and s is the arc length
parameter along the line measured from P0 ( x0 , y0 ) in the direction of a given unit vector u .
df
at P0 is the instantaneous rate of change of f with respect to distance in its
ds
domain in the direction of u . By varying u , we find the rates at which f changes with
Then
Ty ( x0 , y0 ) . The question is how to find the direction of greatest temperature change, which
may be in a direction not parallel to either of the coordinate axes.
Example 3: While hiking in rugged terrain, we may think of the altitude on a hill side at the
point given by longitude x and latitude y as defining a function f ( x, y ) . If you face due east
(in the direction of the positive x-axis), then we know that the slope of the terrain is given by
the partial derivative f x . Similarly, facing due north, the slope of the terrain is given by
f y . However, how would one compute the slope in some other direction, say north-by-
northwest? Also, how one would find the direction of steepest ascent or descent?
To answer these questions, we now introduce the concept of directional derivative and
understand its geometrical interpretation.
Suppose that we want to find the instantaneous rate of change of f ( x, y ) at the point P ( a , b )
and in the direction given by the unit vector u = u1i + u2 j . Let Q ( x, y ) be any point on the
line through P ( a , b ) in the direction of u (see figure above). Notice that the vector PQ is
then parallel u . Since two vectors are parallel if and only if one is a scalar multiple of the
other, we have that PQ = h u for some scalar h, so that
( x a ) i + ( y b ) j = h u = ( hu1 ) i + ( hu2 ) j .
It then follows that x a = hu1 , y b = hu2 , so that x = a + hu1, y = b + hu2 . The point Q is
then described by ( a + hu1 , b + hu2 ) , as indicated in Figure above. The average rate of change
of z = f ( x, y ) along the line from P to Q is therefore
f ( a + hu1 , b + hu2 ) f ( a, b )
.
h
The instantaneous rate of change of f ( x, y ) at the point P ( a, b ) and in the direction of the
unit vector u is then found by taking the limit as h 0 . We give this limit a special name in
the following definition.
Du f ( a, b ) = lim
h0
f ( a + hu1 , b + hu2 ) f ( a, b )
h
Remarks:
a) Notice that this limit resembles the definition of partial derivative, except that in this
case, both variables may change.
b) Calculating directional derivatives by this definition is similar to finding the
derivative of a function of one variable by the limit process. However, a simpler
working formula for finding directional derivatives involving the partial derivatives
f x and f y is given in Theorem 1 below.
c) At a particular point P ( a, b ) , there are infinitely many directional derivatives to the
function f ( x, y ) , one for each direction radiating from P ( a, b ) . Two of these are the
partial derivatives f x ( a, b ) and f y ( a, b ) . To see this, note that if u = i (so u1 = 1 and
u2 = 0 ), then
Di f ( a, b) = lim
h 0
f ( a + h, b ) f ( a, b )
= f x ( a, b ) ,
h
and if u = j (so u1 = 0 and u2 = 1 ),
D j f ( a, b ) = lim
h 0
f ( a, b + h ) f ( a , b )
= f y ( a, b ) .
h
(1)
f ( a + hu1, b + hu2 ) f ( a , b )
h 0
h
F ( h ) F (0)
= lim
= F ( 0) .
h 0
h
Du f ( a , b ) = lim
dF f dx f dy
=
+
= f x ( x, y ) u1 + f y ( x, y ) u2 .
dh x dh y dh
This vertical plane intersects the surface to form a curve C . The slope of the surface at
( x0 , y0 , f ( x0 , y0 ) ) in the direction of u is defined as the slope of the curve C at
represented by the equations x = x0 + tu1, y = y0 + tu2 so that for any value of t , the point
Q ( x, y ) lies on the line L . The points on the surface corresponding to P and Q are
( x , y , f ( x , y ))
0
( x x0 )
+ ( y y0 ) =
( tu1 )
( u is a unit vector)
f ( x, y ) f ( x0 , y0 ) f ( x0 + tu1, y + tu2 ) f ( x0 , y0 )
=
.
t
t
Letting t 0 , we obtain the slope of the tangent line to the curve C at ( x0 , y0 ) . But this is,
by definition, the directional derivative of f at ( x0 , y0 ) in the direction of u :
lim
t 0
f ( x0 + tu1 , y + tu2 ) f ( x0 , y0 )
= Du f ( x0 , y0 ) .
t
Thus, geometrically, the directional derivative in a given direction at any point on the surface
gives the slope of the surface in that direction.
Remarks:
1) Note that since u is a unit vector, we may write u = cos i + sin j , where is the
angle that the vector u makes with the positive x axis. With this notation, any point
on L can be written as x = x0 + t cos , y = y0 + t sin .
2) In order to compute the directional derivative of f at the point ( a , b ) in the direction
of the vector v = v1i + v2 j , where v is not a unit vector, v must be normalized.
That is, consider the unit vector u = v v in the direction of v and then find Du f
using the formula (1) proved in Theorem 1.
Example 1: Find the directional derivative of f ( x, y ) = 3 2 x 2 + y 3 at the point P (1, 2 ) in
the direction of the unit vector u = (1 2 ) i 3 2 j .
(Ans: 2 6 3 )
and ( ii ) u in the direction from ( 2,1) to ( 4,0 ) .
3 2 i + (1 2 ) j ,
(Ans: 2 3 4 ; 16
5)
f ( x, y ) = f x ( x, y ) i + f y ( x, y ) j .
Note that the gradient f ( x, y ) is a vector in the plane and not a vector in space. Also,
think of as an operator which produces a vector in the plane. We read f as del f .
Example 4: The gradient of f ( x, y ) = x 2 y + y 3 is
f ( x , y ) =
x y + y 3 ) i + ( x 2 y + y 3 ) j = ( 2 xy ) i + ( x 2 + 3 y 2 ) j .
(
x
y
(Ans: 6i + 4 j )
Du f ( a, b ) = f ( a , b ) u .
Proof: Since f ( x, y ) = f x ( x, y ) i + f y ( x, y ) j for all x and y , and u = u1i + u2 j , we have
f ( a, b ) u = f x ( a , b ) i + f y ( a , b ) j u1i + u2 j
= f x ( a , b ) u1 + f y ( a , b ) u2
= Du f ( a , b ) ,
b) Linearity Rule
c) Product Rule
( fg ) = f g + g f .
d) Quotient Rule
f g f f g
=
, ( g 0) .
g2
g
e) Power Rule
f n = nf n1f .
a) Constant Rule
( af + bg ) i + ( af + bg ) j = ( af x + bg x ) i + ( af y + bg y ) j
x
y
) (
= a f x i + f y j + b g x i + g y j = af + bg .
( fg ) =
( fg ) i + ( g ) j = ( fg x + f x g ) i + ( fg y + f y g ) j
x
y
) (
= f g x i + g y j + g f xi + f y j = f g + g f .
(d) Using the quotient rule of partial (ordinary) derivatives, we obtain for g 0 ,
f f
f
= i +
g x g y g
g f x i + f y j f
=
g2
gf x fg x gf y fg y
j =
i +
g2
g2
g xi + g y j
g f f g
=
.
g2
f n ) i + ( f n ) j = nf n 1 f xi + nf n 1 f y j = nf n 1 f x i + f y j = nf n 1f .
(
x
y
1) If f ( a , b ) = 0, then Du f ( a , b ) = 0 for any direction u .
2) If f ( a , b ) 0 , then
( a, b) is f ( a, b ) . The
maximum rate of change of f at ( a , b ) is f ( a , b ) .
b) The direction of minimum increase of f at ( a , b ) is f ( a , b ) . The
minimum rate of change of f at ( a , b ) is f ( a , b ) .
The rate of change of f at ( a , b ) is 0 in the directions orthogonal to f ( a , b ) .
The gradient f ( a , b ) is orthogonal to the level curve f ( x, y ) = c at the point ( a , b ) ,
where c = f ( a , b ) .
a) The direction of maximum increase of f at
3)
4)
Proof: 1) Given that f ( a , b ) = 0 , we have Du f ( a, b ) = f ( a, b ) u = 0 u = 0 for any
direction u.
2) Let f ( a , b ) 0 . Then
(a) Using Theorem 2, we have
Du f ( a, b ) = f ( a, b ) u = f ( a, b ) u cos = f ( a , b ) cos .
where is the angle between the gradient vector at ( a, b ) and the direction vector u .
Now, f ( a , b ) cos has its maximum value when cos assumes its largest value,
that is, when = 0 , so that cos = 1 . Thus, the direction of maximum increase of f
at ( a , b ) is attained when the angle between f ( a, b ) and u is 0. In other words,
f ( a, b ) is in the same direction as u . Hence, the direction of maximum increase of
f ( a, b )
, and the largest value of Du f ( a, b ) is f ( a , b ) .
f at ( a , b ) is u =
f ( a, b )
(b) As in part (a), the minimum value of Du f ( a, b ) occurs when cos = 1 and
= . This value occurs when u points towards f ( a , b ) , and in this direction
Du f = f ( a , b ) ( 1) = f ( a, b ) .
3) Since Du f ( a, b ) = f ( a , b ) u and f ( a , b ) u = 0 if and only if f ( a , b ) is
orthogonal to u , we get that Du f ( a , b ) = 0 in the directions orthogonal to f ( a , b ) .
4) Let the vector equation of the level curve be r ( t ) = x ( t ) i + y ( t ) j , where t is a
x dt y dt
dt
This implies that f is normal to velocity vector r ( t ) for every point on the level
curve. But r ( t ) is tangent to the level curve at every point, so that f is normal to
the level curve at every point. In particular, f ( a , b ) is orthogonal to the level curve
at the point ( a, b ) .
Example 8: In what direction is the function defined by f ( x, y ) = xe 2 y x increasing most
rapidly at the point P ( 2,1) , and what is the maximum rate of increase? In what direction is
f = f xi + f y j = e2 y x + xe2 y x ( 1) i + xe2 y x ( 2 ) j
= e2 y x (1 x ) i + 2 xj .
Now, at the point P ( 2,1) , we have f ( 2,1) = i + 4 j . Thus, the most rapid rate of increase
is f ( 2,1) = 17 and it occurs in the direction of i + 4 j . The most rapid rate of decrease
therefore occurs in the direction f ( 2,1) = i 4 j.
Example 9: Find the direction of maximum increase on the contour plot of the surface given
by f ( x, y ) = 3x x 3 3xy 2 from the point P ( 0.6, 0.7) and sketch the path of steepest
ascent.
so that f ( 0.6, 0.7) =< 0.45,2.52 > . The unit vector in this direction is then
u =< 0.176,0.984 > , which is the direction of maximum increase. Note that the path of
steepest ascent is a curve that remains perpendicular to each level curve through which it
passes. Finding the path of steepest ascent is a difficult one. The unit vector u does not point
towards the maximum (the peak of the surface) at (1,0 ) . However, a plausible path of
steepest ascent is shown in the figure below:
Remarks: The directional derivative and gradient concepts can easily be extended to
functions of three or more variables. The basic properties of the gradient and maximal
property is valid for function or three or more variables. For example, for a function
f
f f
j + k , and the
f ( x, y , z ) of three variables, the gradient f is defined by f = i +
x
y
z
directional derivative Du f of f ( x, y , z ) at the point P0 ( x0 , y0 , z0 ) in the direction of the unit
vector u is given by
Du f ( x0 , y0 , zo ) = f ( x0 , y0 , zo ) u .
Example10: Let f ( x , y , z ) = xy sin ( xz ) . Find f (1, 2, ) and then compute the directional
derivative of f at (1, 2, ) in the direction of the vector v = 2i + 3 j 5k .
Solution: By definition, we have f = f xi + f y j + f z k . We now compute the partial
derivatives at (1, 2, ) :
f y = ( xy sin ( xz ) ) = x sin ( xz )
f y (1, 2, ) = 0;
y
f z = ( xy sin ( xz ) ) = xy cos ( xz ) x
f z (1, 2, ) = 2.
z
fx =
2i + 3 j 5k
( 2 )
+ ( 3) + ( 5)
1
2i + 3 j 5k ,
38
1
1
Du f (1, 2, ) = f (1, 2, ) u = 2 i + 2k
2i + 3 j 5k =
( 4 10) 3.66.
38
38
( x ( t ) , y (t ) , z (t ))
d
f ( x ( t ) , y ( t ) , z ( t ) ) = 0 .
dt
Applying the Chain rule on left hand side, we get that for all t I ,
f x ( x (t ) , y (t ) , z (t ) )
dx
dy
dz
+ f y ( x (t ) , y (t ) , z ( t ) ) + fz ( x (t ) , y (t ) , z ( t ) ) = 0
dt
dt
dt
or
dR
f ( x ( t ) , y ( t ) , z ( t ) )
= 0.
dt
=0
f ( a, b, c )
dt ( a ,b,c )
...(1) .
dR
0,
But since the curve is smooth, the velocity vector to the curve C at P , namely,
dt ( a ,b ,c )
and it is given that f ( a, b, c ) 0 . Therefore, from (1), we must have that the vector
f ( a, b, c ) is normal to the vector dR dt
. That is, f ( a, b, c ) is orthogonal to the level
( a ,b ,c )
surface of f through P ( a, b, c ) .
Remarks: The preceding theorem implies that the gradient vector at a point P ( a, b, c ) on the
level surface is normal to the tangent vector T = dR dt on each curve C on the surface that
passes through P . Thus, all these tangent vectors lie in a single plane through P with normal
vector N = f ( a, b, c ) . This plane is the tangent plane to the surface at P . Note that since
P is an arbitrary point on the level surface, there is a unique tangent plane at every point on
the level surface of f where f 0 .
Example 11: Find a vector that is normal to the level surface x 2 + 2 xy yz + 3z 2 = 7 at the
point P (1,1, 1) .
Solution: By the preceding theorem, the gradient vector f (1,1, 1) is normal to the level
f = f xi + f y j + f z k = ( 2 x + 2 y ) i + ( 2 x z ) j + ( 6 z y ) k .
At the point (1,1, 1) , we have, therefore, f (1,1, 1) = 4i + 3 j 7k as the required normal.
Example 12: Sketch the level curve f ( x, y ) = c corresponding to c = 1 for the function
Solution: The level curve for c = 1 is the hyperbola given by x 2 y 2 = 1 (trace it). The
gradient vector is normal to the level curve, by Theorem 3 part (4). We have
f = f xi + f y j = 2 xi 2 yj .
quadrant of the xy plane. Suppose this square is heated in such a way that the temperature
at the point P ( x, y ) is given by T ( x, y ) = x 2 + y 2 . In what direction will heat flow from the
point P ( 3, 4 ) ?
Solution: The level curves of T are called isothermal curves. From physics it is known that
the flow of heat is perpendicular to the isothermal curves, and points in the direction of
decreasing temperature. Thus, if H ( x, y ) denotes the heat flow at a point ( x, y ) in the
H ( x, y ) = k T ( x, y )
...(2) .
H ( 3, 4 ) = k 6i + 8 j .
Because the thermal conductivity k is positive, the heat flows from P ( 3, 4 ) in the direction
of the unit vector u given by
u=
6i + 8 j
( 6 )
+ ( 8 )
( 6) i + ( 8)
100
j = 3 i + 4 j.
100
5 5
Example 14: Find equations for the tangent plane and the normal line at the point P (1, 1, 2 )
F ( x, y, z ) = x 2 y + y 2 z + z 2 x .
The gradient vector F is normal to the level surface S at P (1, 1, 2 ) . We have
F ( x, y , z ) = ( 2 xy + z 2 ) i + ( x 2 + 2 yz ) j + ( y 2 + 2 xz ) k .
Thus, a normal vector at P (1, 1, 2 ) is N = F (1, 1, 2 ) = 2i 3 j + 5k . Hence, if Q ( x, y , z )
is any point on the plane, then equation for the tangent plane passing through P (1, 1, 2 ) and
with normal vector N is given by QP N = 0 , or equivalently
2 ( x 1) 3 ( y + 1) + 5 ( z 2 ) = 0
or
2 x 3 y + 5 z = 15.
The normal line to the surface at P (1, 1, 2 ) with direction numbers < 2, 3,5 > is
x = 1 + 2t ,
y = 1 3t,
z = 2 + 5t ,
t .
Example 15: Find the equation for the tangent plane and the normal line to the cone
z2 = x2 + y 2
Fx ( x, y , z ) = 2 x, Fy ( x, y, z ) = 2 y , Fz ( x, y , z ) = 2 z.
Therefore, the direction numbers of the normal to the cone at ( 3, 4,5 ) are
y = 4 + 8t , z = 5 10t , t .
(1) We call f (a, b) the absolute maximum of f on the region R if f (a, b) f (x, y) for
all (x, y) R.
(2) We call f (a, b) is called the absolute minimum of f on R if f (a, b) f (x, y) for all
(x, y) R.
(3) In either case (1) or (2), f (a, b) is called an absolute extremum of f.
Relative Extrema
Local Maximum
Local Minimum
f x ( x0 , y0 ) = 0 and f y ( x0 , y0 ) = 0 .
2)
f ( x0 , y0 ) = f x ( x0 , y0 ) i + f y ( x0 , y0 ) j = 0i + 0 j .
Therefore, every directional derivative of f at ( x0 , y0 ) must be 0. This implies that the graph
of the function f has a horizontal tangent plane at the point ( x0 , y0 , f ( x0 , y0 ) ) . Since some
critical points yields saddle points, ( x0 , y0 ) is a possible location for relative extremum. See
figures below:
f x ( x0 , y0 ) = f y ( x0 , y0 ) = 0 .
Remarks: The preceding theorem implies that there is a horizontal plane at each extreme
point where the first partial derivatives exist. However, the theorem does not say that
whenever a horizontal tangent plane occurs at a point P , there must be an extremum there.
All that we can deduce is that such a point P is a possible location for a relative extremum.
Example 1: Discuss the nature of the critical ( 0,0) for the quadric surfaces
a) z = x 2 + y 2 .
b) z = 1 x 2 y 2 .
b) g x ( x, y ) = 2 x, g y ( x, y ) = 2 y , so again
( 0, 0)
for all x and y with a relative maximum z = 1 occurring when ( x, y ) = ( 0,0) , that is
the point ( 0,0) is the point of relative maximum.
( 0,0) ,
and hence
hyperbolic paraboloid h ( x, y ) = y 2 x 2 .
f ( x, y ) = 2 x 2 + y 2 + 8 x 6 y + 20 .
13
b) g ( x , y ) = 1 ( x 2 + y 2 )
(a)
(b)
( 0, 0)
We begin by finding the critical points of f and g . The critical points of f and g are
found using their partial derivatives. The partial derivatives of f and g are
f x ( x, y ) = 4 x + 8,
g x ( x, y ) =
f y ( x, y ) = 2 y 6;
2 x
3( x + y
2
2 23
, g y ( x, y ) =
2 y
3( x + y 2 )
2
23
( 2,3)
( x, y ) ( 2,3) ,
we
conclude that a relative minimum of f occurs at ( 2,3) , and the value of the
relative minimum is f ( 2,3) = 3 .
(b) Both partial derivatives g x and g y exist for all x and y except for ( 0,0 ) . Since the
critical points are those points for which one of the partials must not exist or both
must be 0, we have that the critical points are the solutions of the equations
g x ( x, y ) = 0 = g y ( x, y ) . Solving these equations, we get that ( 0,0 ) is the only critical
point of g . Since g ( 0,0 ) = 1 and
13
g ( x, y ) = 1 ( x 2 + y 2 ) < 1 ( x, y ) 2 ,
we find that g has a relative maximum at ( 0,0 ) .
Second Partials Test:
In the preceding examples, it was relatively easy to find the relative extrema of given
functions or determine when a critical point of the given function is a saddle point. The
arguments were algebraic. However, for more complicated functions, it is better to rely on
analytical means presented in following Second partials test:
Theorem 3: Let f have continuous second order partial derivatives on an open region
f x ( x0 , y0 ) = 0 and f y ( x0 , y0 ) = 0 .
To test for relative extrema of f at ( x0 , y0 ) , consider the quantity, called discriminant of f ,
defined by the equation
D ( x0 , y0 ) = f xx ( x0 , y0 ) f y ( x0 , y0 ) f xy2 ( x0 , y0 ) .
( x0 , y0 ) if D ( x0 , y0 ) > 0 and
equivalently, D ( x0 , y0 ) > 0 and f yy ( x0 , y0 ) < 0 ).
(2) A relative minimum occurs at ( x0 , y0 ) if D ( x0 , y0 ) > 0 and
equivalently, D ( x0 , y0 ) > 0 and f yy ( x0 , y0 ) > 0 ).
(3) A saddle point occurs at ( x0 , y0 ) if D ( x0 , y0 ) < 0 .
(4) The test is inconclusive if D ( x0 , y0 ) = 0 .
(1) A relative maximum occurs at
f xx ( x0 , y0 ) < 0 (or
f xx ( x0 , y0 ) > 0 (or
Solution: (a) We first find the critical points of f . Because the partial derivatives
f x ( x, y ) = 3x 2 + 4 y and f y ( x, y ) = 4 x 4 y
both exist for all x and y , the only critical points are those for which both f x ( x, y ) and
To test these points for relative extrema, we now compute the second partials:
f xx ( x, y ) = 6 x, f yy ( x, y ) = 4, f xy ( x, y ) = 4.
For the critical point ( 0,0 ) , we have f ( 0,0) = 1 , and the discriminant of f
2
g xx ( x, y ) = 2 y 2 , g yy ( x, y ) = 2 x 2 , g xy ( x, y ) = 4 xy .
If either x = 0 or y = 0 , then
2
D ( x , y ) = f xx ( x , y ) f yy ( x , y ) f xy ( x , y ) = 4 x 2 y 2 16 x 2 y 2 = 12 x 2 y 2 = 0 .
So, the second partials test fails. However, because f ( x, y ) = 0 for every point along
x axis or y axis, and f ( x, y ) = x 2 y 2 > 0 for all other points, we conclude that
each of these critical points is yields an absolute minimum. (See figure below)
Example 5: Find the absolute extrema of the function f ( x, y ) = sin ( xy ) on the closed
region given by 0 x , 0 y 1.
f x ( x, y ) = y cos ( xy ) , f y ( x, y ) = x cos ( xy ) ,
we obtain that the critical points in the domain of f are ( 0,0 ) and each point lying in the
domain and on the hyperbola given by xy = 2 . Since every point on the hyperbola
xy = 2 yields the value
f ( x, y ) = sin ( 2 ) = 1 ,
which we know is the absolute maximum of f , we get that all the critical points lying on the
hyperbola xy = 2 are points of absolute maximum.
The critical point ( 0,0) yields an absolute minimum of f because for every point in the
domain 0 x , 0 y 1, we have 0 xy which implies that
0 sin ( xy ) 1 0 f ( x, y ) 1 .
To locate other absolute extrema, we consider the points of the boundary of the domain:
x = 0, x = , y = 0, y = 1 . Since sin ( xy ) = 0 for all the points on the x axis, y axis and at
the point ( ,1) , we get that each of these points yields an absolute minimum of f . (See
figure below)
Example 6: Find all critical points on the graph of the following functions and classify each
point as a relative extremum or a saddle point:
(1) f ( x, y ) = 8 x 3 24 xy + y 3 .
(2) g ( x , y ) = x 2 y 4 .
(Try on your own and draw the rough sketch of the surface in each case.)
Example 7: Find the absolute extrema of the function h ( x , y ) = e x
x2 + y 2 1 .
y2
y2
and h y ( x, y ) = 2 ye x
y2
exist for
all the points ( x, y ) , and both are zero only when x = 0 and y = 0 , it follows that ( 0,0 ) is
the only critical point of h and it is inside the unit disk x 2 + y 2 1 . Also, h ( 0,0 ) = e0 = 1 .
But the discriminant of h at
( 0,0)
x 2 1 x 2
= e2 x
Since F ( x ) = 4 xe2 x
( 0, 1)
for 1 x 1.
( 1,0 ) ,
[ 1,1] ,
the
Thus, by computing the values of h for all the above critical points:
h ( 0,1) = e 1 , h ( 0, 1) = e 1, h (1, 0 ) = e, h ( 1,0 ) = e ,
we find that the absolute maximum value of h on the given unit disk is e , which occurs at
(1, 0 )
and ( 1,0 ) ; the absolute minimum value is e1 , which occurs at ( 0,1) and ( 0, 1) .
Example 8: Find the point on the plane x + 2 y + z = 5 that is closest to the point P ( 0, 3, 4 ) .
Solution: Let Q ( x, y , z ) be any point on the plane x + 2 y + z = 5 . Then z = 5 x 2 y and
2
f ( x, y ) = x 2 + ( y 3) + (1 x 2 y ) .
D ( 5 6 ,4 3) = f xx ( 5 6, 4 3) f yy ( 5 6, 4 3) f xy ( 5 6, 4 3) = 4 (10 ) 16 > 0 ,
and f xx ( 5 6,4 3) = 4 > 0 , so a relative minimum occurs at
( 5 6, 4 3) .
relative minimum must also be an absolute minimum because there must be exactly one point
on the plane that is closest to the given point. The corresponding z value is
25
5
5 4
5
4
=
.
d = + 3 + 1 + 2 =
6
6
6 3
6
3
Example 9: A rectangular box is resting on the xy plane with one vertex at the origin. The
opposite vertex lies on the plane 6 x + 4 y + 3z = 24 . Find the maximum volume of the box.
See figure below.
Solution: Let x, y , z represent the length, width and height of the box, respectively. Because
one vertex of the box lies at the origin and the vertex opposite to this vertex lies on the plane
6 x + 4 y + 3z = 24 , the point ( x, y, z ) must satisfy the equation of the this plane. Thus, we get
z=
1
( 24 6 x 4 y ) .
3
Thus, we can write the volume V = xyz of the box as a function of two variables as follows:
1
1
V ( x, y ) = xy ( 24 6 x 4 y ) = ( 24 xy 6 x 2 y 4 xy 2 ) ,
3
3
where the domain of V is the triangle in the xy plane bounded by the lines x = 0, y = 0
and 3 x + 2 y 12 = 0 . By setting the first partial derivatives equal to 0,
y
1
24 y 12 xy 4 y 2 ) = ( 24 12 x 4 y ) = 0
(
3
3
1
x
V y ( x, y ) = ( 24 x 6 x 2 8 xy ) = ( 24 6 x 8 y ) = 0
3
3
Vx ( x , y ) =
we obtain that the critical points are ( 0,0 ) and ( 4 3, 2 ) . At ( 0,0 ) , the volume is 0 so that
Because
V yy =
8 x
,
3
Vxy =
1
( 24 12 x 8 y ) .
3
and
Vxx ( 4 3,2 ) = 8 < 0 , we conclude that the maximum volume of the box occurs at
( x, y ) = ( 4 3,2 )
2
1 4
4
4 2 64
V ( 4 3, 2 ) = 24 2 6 2 4 ( 2 ) =
cu units.
3 3
3
3
9
Note that the volume is 0 at the boundary points of the triangular domain of V.
Example 10: An electronics manufacturer determines that the profit (in dollars) obtained by
producing x units of a DVD player and y units of a DVD recorder is approximated by the
By setting these partial derivatives equal to 0, and after simplifying, we obtain the following
system of equations:
2 x + y = 8000
x + 2 y = 10,000.
Solving this system produces x = 2000 and y = 4000 . The second partial derivatives of P at
( x, y ) = ( 2000,4000)
are
Pxx ( 2000,4000 ) = 0.002
Pyy ( 2000, 4000 ) = 0.002
Pxy ( 2000, 4000 ) = 0.001.
Pxx ( 2000, 4000 ) Pyy ( 2000,4000 ) Pxy ( 2000, 4000 ) = ( 0.002 ) ( 0.001) > 0 .
we conclude that the production level of x = 2000 units and y = 4000 units yields a maximum
profit. The maximum profit is P ( 2000, 4000 ) = 18,000 , that is $18,000.
Remarks: Note that in the preceding example, it was assumed that the manufacturing plant
is able to produce the required number of units to yield a maximum profit. In actual practice,
the production would be bounded by physical constraints. We will next study such
constrained optimization problems and learn how deal with them by a very ingenious
technique called the method of Lagrange Multiplier.
g ( x, y ) = 0 .
In figure (a), each point of intersection of g ( x, y ) = 0 with a level curve is a candidate for a
solution, since these points lie on the constraint curve. Among the seven such intersections
shown in the figure, the maximum value of f ( x, y ) occurs at the intersection ( x0 , y0 ) where
f ( x, y ) has a value of 400. Note that at ( x0 , y0 ) , the constraint curve and the level curve just
touch and thus have a common tangent line at this point. Since f ( x0 , y0 ) is normal to the
level curve f ( x, y ) = 400 at ( x0 , y0 ) , and since g ( x0 , y0 ) is normal to the constraint curve
f ( x0 , y0 ) = g ( x0 , y0 )
(*)
for some scalar . The same condition holds at points on the constraint curve where f ( x, y )
has a minimum. For example, if the level curves are as shown in the figure (b), then the
minimum value of f ( x, y ) occurs where the constraint curve just touches a level curve.
Thus, to find the maximum or minimum of f ( x, y ) subject to the constraint g ( x, y ) = 0 , we
look for points at which (*) equation holdsthis is the method of Lagrange multipliers. The
scalar is called a Lagrange multiplier.
To see how this method works, let us look at the following problem. Suppose we want
to find the greatest and the smallest values that the function
f ( x, y ) = xy
takes on the ellipse
x2 y2
+
= 1.
8
2
That is, to find the extreme values of f ( x, y ) = xy subject to the constraint
g ( x, y ) =
x2 y 2
+
1 = 0 .
8
2
The farther the hyperbolas lie from the origin, the larger the absolute value of f . We want
to find the extreme values of f ( x, y ) given that the point ( x, y ) is in the domain of f and
also lies on the ellipse x 2 + 4 y 2 = 8 . Which hyperbolas intersecting the ellipse lie farthest
from the origin? The hyperbolas that just graze the ellipse, the ones that are tangent to it, are
farthest. To find the appropriate hyperbola, we use the fact that two curves are tangent at a
point if and only if their gradient vectors are parallel. This means that f ( x, y ) must be a
scalar multiple of g ( x, y ) at the point of tangency, so that
f ( x , y ) = g ( x , y )
yi + xj = xi + yj .
4
we
obtain
( 2 y )
+ 4 y 2 = 8 4 y 2 + 4 y 2 = 8 y = 1.
Therefore,
the
function
( 2,1) , ( 2, 1) .
Let f and g have continuous first order partial derivatives such that f has an extremum at
a point ( a, b) on the smooth constraint curve g ( x, y ) = c . If g ( a, b ) 0 , then there is a
real number such that
f ( a , b ) = g ( a , b ) .
Proof: Let us represent the smooth curve given g ( x, y ) = c by the vector-valued function
r ( t ) = x ( t ) i + y ( t ) j,
r ( t ) 0,
h ( t0 ) = f ( x ( t0 ) , y ( t0 ) ) = f ( a , b )
is an extreme value of h . This implies that h ( t0 ) = 0 . But by Chain rule, we have
h ( t0 ) = f x ( a, b) x ( t0 ) + f y ( a, b ) y ( t0 ) = f ( a, b ) r ( t0 ) = 0 .
Thus, we obtain that f ( a, b ) is orthogonal to r ( t0 ) . Moreover, g ( a, b ) is orthogonal to
r ( t0 ) . Consequently, the gradients f ( a, b ) and g ( a, b ) are parallel. Hence, there exists a
scalar such that f ( a, b ) = g ( a, b ) . This completes the proof.
2. Evaluate f at each solution point obtained in the first step. The largest value yields
the maximum of f subject to the constraint g ( x, y ) = c , and the smallest value
yields the minimum of f subject to the constraint g ( x, y ) = c .
Example 11: Find the maximum value of f ( x, y ) = 4 xy where x > 0 and y > 0 , subject to
the constraint x 2 9 + y 2 16 = 1 .
Solution: For the rectangle to be inscribed inside the given ellipse, we must have one of its
vertices in the first quadrant. Let ( x, y ) be the vertex of the rectangle in the first quadrant, so
that x > 0 and y > 0 (see figure above). Note that the other three vertices of the rectangle are
then determined uniquely as ( x, y ) , ( x, y ) , ( x, y ) . Now, because the rectangle has sides
of lengths 2 x and 2 y , its area is given by (the objective function) f ( x, y ) = 4 xy . We want
to find x and y such that f ( x, y ) is a maximum. However, the choice of ( x, y ) is restricted
to the first-quadrant points that lie on the ellipse (the constraint function)
x2 y2
+
= 1.
9 16
We now use the method of Lagrange multiplier. Let g ( x, y ) = x 2 9 + y 2 16 = 1 . By equating
f ( x, y ) = 4 yi + 4 xj and g ( x, y ) = ( 2 x 9 ) i + ( 2 y 16 ) j ,
we obtain the following system of equations.
2
4 y = x,
9
1
4 x = y,
8
x2 y 2
+
= 1 (constraint equation).
9 16
From the first equation, we obtain = 18 ( y x ) and substitution into the second equation
produces 4 x = (1 8)(18 y x ) y x 2 = ( 9 16) y 2 . Substituting this value of x 2 into the third
equation gives us y 2 = 8 , so that y = 2 2 . But y > 0 , so we take the positive value
y = 2 2. This gives x 2 = 9 2 or x = 3
will occur at 3
3
f
,2 2 = 4
2 2 = 24.
2
Geometry of the solution: We consider the constraint equation to be the fixed level curve
x2 y 2
+ . The level curves of f , f 1 ( k ) , k a constant,
9 16
represent a family of hyperbolas f ( x, y ) = 4 xy = k (see figure below).
g ( x, y ) = 1 , where g ( x, y ) =
In this family, the level curves that meet the given constraint correspond to the hyperbolas
that intersect the ellipse. Moreover, to maximize f ( x, y ) , we wanted to find the hyperbola
that just barely satisfies the constraint. The level curve that does this is the one that is tangent
to the ellipse.
Example 12: The Cobb-Douglas production function for a software manufacturer is given by
f ( x, y ) = 100 x 3 4 y1 4 ,
where x represents the units of labour (at $150 per unit) and y represents the units of capital
(at $250 per unit). The total cost of labour and capital is limited to $50,000. Find the
maximum production level for this manufacturer.
Solution: The limit on the cost of labour and capital produces the constraint equation
f ( x, y ) = 75 x1 4 y1 4i + 25 x3 4 y 3 4 j .
By Lagranges theorem, there exists such that f ( x, y ) = g ( x, y ) . This gives us the
following system of equations
75 x 1 4 y1 4 = 150 , 25 x 3 4 y 3 4 = 250 and 150 x + 250 y = 50,000.
34
14
(50)
Remarks:
Economists call the Lagrange multiplier obtained in a production function the marginal
productivity of money. For instance, in the preceding example, the marginal productivity of
money at x = 250 and y = 50 is
1 4
14
1
( 250) ( 50)
= x 1 4 y 1 4 =
2
2
0.334
which means that for each additional dollar spent on production, an additional 0.334 unit of
the product can be produced.
g ( x, y ) = x 2 + y 2 1 .
First, we will find the constrained relative extrema. For this purpose, we need the gradients
f ( x, y ) = yi + xj and g ( x, y ) = 2 xi + 2 yj . Now, g ( x, y ) = 0 if and only if x = 0 and
y = 0 , so that g ( x, y ) 0 for any point on the circle x 2 + y 2 = 1 . Thus, at a constrained
relative extremum, we must have
f ( x, y ) = g ( x, y ) yi + xj = 2 xi + 2 yj .
This gives us the pair of equations y = 2 x, x = 2 y . It follows from these equations that if
y
x
and =
.
2x
2y
2 , y = 1
1 1 1 1 1 1 1 1
the points
,
,
,
,
,
,
,
.
2 2 2 2 2 2 2 2
1 1 1 1
Thus, the absolute maximum of f ( x, y ) = xy is 1 2 at the points
,
,
,
.
2 2 2 2
1 1 1 1
Also, note that f has an absolute minimum 1 2 at the points
,
,
,
. In
2 2 2 2
the figure above, some level curves xy = c , and the constraint curve x 2 + y 2 = 1 are shown in
the vicinity of maxima of f .
Example 14: Find the maximum and minimum values of the function f ( x, y ) = 3x + 4 y on
the circle x 2 + y 2 = 1 .
f ( x, y ) = 3x + 4 y, g ( x, y ) = x 2 + y 2 1 ,
and look for the values of x, y and that satisfies the equations
f ( x, y ) = g ( x, y ) 3i + 4 j = 2 xi + 2 yj ;
x 2 + y 2 1 = 0.
g ( x, y ) = 0
The gradient equation implies that 0 , and gives that x = 3 2 , y = 2 . These equations
tell us that both x and y have the same sign. The constraint equation gives us that
2
5
3 2
+ 1 = 0 = .
2
2
Thus,
x = 3 5 and y = 4 5 ,
( x, y ) = ( 3 5, 4 5)
and
f ( x, y ) = 3 x + 4 y
has
extreme
values
at
respectively, therefore
3
4 25
3
4 25
3 + 4 =
= 5 and 3 + 4 =
= 5.
5
5
5 5
5
5
T ( x, y ) = x 2 + 2 x + y 2
for points
( x, y )
Solution: The plate corresponds to the shaded region as shown in the figure below.
We first look for critical points of T ( x, y ) inside the region R . We have the gradient of T
T ( x, y ) = ( 2 x + 2 ) i + ( 2 y ) j .
At the critical point, we must have T ( x, y ) = 0 , that is 2 x + 2 = 0, 2 y = 0 which implies
x = 1, y = 0 . Thus, T is maximum or minimum inside the region R at the point ( 1,0) and
at this point T ( 1, 0) = 1 .
Now, we look for extrema of T ( x, y ) on the boundary of the region R , that is the ellipse
x 2 + 4 y 2 = 24 using the method of Lagrange multiplier. The constraint function is
g ( x, y ) = x 2 + 4 y 2 24 = 0 .
By Lagranges theorem, any extrema on the ellipse must satisfy the gradient equation
T ( x, y ) = g ( x, y ) .
and
2 y = 8 y .
constraint equation implies that y = 50 3 . Now, we compare all the values of T at all of
these points, namely, one interior critical point and the candidates for boundary extrema:
T ( 1,0 ) = 1,
( 24,0) = 24 + 2 24 33.8,
T ( 24,0 ) = 24 2 24 14.2,
T
4 50 14
T ,
4.7,
=
3
3
3
4
50 14
T ,
4.7.
=
3 3
3
Hence, the minimum value of T is 1 at the point ( 1,0) and the maximum value of T is
24 + 2 24 at the point
24,0 .
Solution: This problem is similar to Example 15. We divide the constraint into two cases.
a) For points on the circle x 2 + y 2 = 10 , use Lagrange multipliers to find that the
maximum value of f ( x, y ) is 24. This value occurs at ( 1,3) and at ( 1, 3) . In a
similar way, we can determine that the minimum value of f is approximately 6.675,
and this value occurs at
10, 0 .
b) For points inside the circle, use the techniques of finding the relative extrema to
conclude that the function has a relative minimum of 2 at the point (1, 0) .
By combining these two results, we conclude that f has a maximum of 24 at ( 1,3) and a
minimum of 2 at (1, 0) , as shown in the figure below.
f ( x , y , z ) = 2 x 2 + y 2 + 3z 2
2 x 3 y 4 z = 49 .
g ( x, y, z ) = 2i 3 j 4k and f ( x, y , z ) = 4 xi + 2 yj + 6 zk ,
so that Lagranges theorem gives us that f = g , we obtain the following system of
equations
4 x = 2 :
f x ( x, y , z ) = g x ( x, y , z )
2 y = 3 :
f y ( x , y , z ) = g y ( x, y , z )
6 z = 4 :
f z ( x, y, z ) = g z ( x, y , z )
2 x 3 y 4 z 49 = 0 :
(the constraint)
f ( 3, 9, 4 ) = 2 ( 3) + ( 9 ) + 3 ( 4 ) = 147.
Note that from the original function and the constraint, f has no maximum. So the optimum
value determined above is a minimum of f .
( x, y, z ) ,
the
point must lie on both surfaces defined by the constraints. Consequently, in order for there to
be a solution, we must assume that the two surfaces intersect. We further assume that g
and h are nonzero and are not parallel, so that the two surfaces intersect in a curve C and
are not tangent to one another. If f has an extremum at a point ( x0 , y0 , z0 ) on a curve C, then
f ( x0 , y0 , z0 ) must be normal to the curve. Notice that since C lies on both constraint
surfaces, g ( x0 , y0 , z0 ) and h ( x0 , y0 , z0 ) are both orthogonal to C at
implies that f ( x0 , y0 , z0 ) must lie in the plane determined by
( x0 , y0 , z0 ) .
g ( x0 , y0 , z0 )
This
and
f ( x, y, z ) = g ( x, y, z ) + h ( x, y, z ) .
The method of Lagrange multipliers for the case of two constraints then consists of finding
the point ( x, y , z ) and the Lagrange multipliers and (for a total of five unknowns)
satisfying the five equations defined by:
f x ( x , y , z ) = g x ( x , y , z ) + hx ( x , y , z )
f y ( x, y, z ) = g y ( x, y , z ) + h y ( x, y , z )
f z ( x, y , z ) = g z ( x, y, z ) + hz ( x, y , z )
g ( x, y , z ) = 0
h ( x, y, z ) = 0.
Solution: The intersection of the given plane and the paraboloid is shown below.
Finding the point on the ellipse that is closest to the origin is same as finding the point
( x, y, z )
( x 0) + ( y 0) + ( z 0 ) = x 2 + y 2 + z 2
f ( x, y, z ) = x 2 + y 2 + z 2 subject to the constraints
or
g ( x, y, z ) = x + y + z 12 = 0 and h ( x, y, z ) = x 2 + y 2 z = 0 .
At any extremum, we must have, by the method of Lagrange multipliers, that
f ( x, y, z ) = g ( x, y, z ) + h ( x, y , z )
) (
2 xi + 2 yj + 2 zk = i + j + k + 2 xi + 2 yj k .
or
Together with the constraint equations, we now have the system of five equations
2 x = + 2 x, 2 y = + 2 y , 2 z = , x + y + z 12 = 0, x 2 + y 2 z = 0
(*) .
From the first two equations, we have = 2 x 2 x and = 2 y 2 y . Equating these two
expressions of , we obtain that
2 x 2 x = 2 y 2 y
( x y )(1 ) = 0 .
(*)
0 = x + y + z 12 = x + x + 2 x 2 12 = 2 x 2 + 2 x 12 = 2 ( x + 3)( x 2 ) .
Thus, we get that either x = 3 or x = 2 . Since y = x and z = 2 x 2 , we have the points of
extremum of f as ( 3, 3,18 ) and ( 2, 2,8 ) . Finally, since
Example 20: The plane x + y + z = 1 cuts the cylinder x 2 + y 2 = 1 in an ellipse. Find the
points on the ellipse that lie closest to and farthest from the origin.
(Ans.: The closest points are (1, 0, 0 ) and ( 0,1,0 ) . Farthest point is 2 2, 2 2,1 + 2 )