Vous êtes sur la page 1sur 65

15.

1 SOLUTIONS 1039
CHAPTER FIFTEEN
Solutions for Section 15.1
Exercises
1. The point A is not a critical point and the contour lines look like parallel lines. The point B is a critical point and is a local
maximum; the point C is a saddle point.
2. To nd the critical points, we solve fx = 0 and fy = 0 for x and y. Solving
fx = 2x 2y = 0,
fy = 2x + 6y 8 = 0.
We see from the rst equation that x = y. Substituting this into the second equation shows that y = 2. The only critical
point is (2, 2).
We have
D = (fxx)(fyy) (fxy)
2
= (2)(6) (2)
2
= 8.
Since D > 0 and fxx = 2 > 0, the function f has a local minimum at the point (2, 2).
3. The partial derivatives are
fx = 6x 4 + 2y and fy = 2x 10y + 48.
Set fx = 0 and fy = 0 to nd the critical point, thus
2y 6x = 4 and 10y 2x = 48.
Now, solve these equations simultaneously to obtain x = 1 and y = 5.
Since fxx = 6, fyy = 10 and fxy = 2 for all (x, y), at (1, 5) the discriminant D = (6)(10) (2)
2
= 56 >
0 and fxx < 0. Thus f(x, y) has a local maximum value at (1, 5).
4. To nd the critical points, we solve fx = 0 and fy = 0 for x and y. Solving
fx = 3x
2
6x = 0
fy = 2y + 10 = 0
shows that x = 0 or x = 2 and y = 5. There are two critical points: (0, 5) and (2, 5).
We have
D = (fxx)(fyy) (fxy)
2
= (6x 6)(2) (0)
2
= 12x 12.
When x = 0, we have D = 12 < 0, so f has a saddle point at (0, 5). When x = 2, we have D = 12 > 0 and
fxx = 6 > 0, so f has a local minimum at (2, 5).
5. To nd the critical points, we solve fx = 0 and fy = 0 for x and y. Solving
fx = 3x
2
3 = 0
fy = 3y
2
3 = 0
shows that x = 1 and y = 1. There are four critical points: (1, 1), (1, 1), (1, 1) and (1, 1).
We have
D = (fxx)(fyy) (fxy)
2
= (6x)(6y) (0)
2
= 36xy.
At the points (1, 1) and (1, 1), we have D = 36 < 0, so f has a saddle point at (1, 1) and (1, 1). At (1, 1),
we have D = 36 > 0 and fxx = 6 > 0, so f has a local minimum at (1, 1). At (1, 1), we have D = 36 > 0 and
fxx = 6 < 0, so f has a local maximum at (1, 1).
1040 Chapter Fifteen /SOLUTIONS
6. To nd the critical points, we solve fx = 0 and fy = 0 for x and y. Solving
fx = 3x
2
6x = 0
fy = 3y
2
3 = 0
shows that x = 0 or x = 2 and y = 1 or y = 1. There are four critical points: (0, 1), (0, 1), (2, 1), and (2, 1).
We have
D = (fxx)(fyy) (fxy)
2
= (6x 6)(6y) (0)
2
= (6x 6)(6y).
At the point (0, 1), we have D > 0 and fxx < 0, so f has a local maximum.
At the point (0, 1), we have D < 0, so f has a saddle point.
At the point (2, 1), we have D < 0, so f has a saddle point.
At the point (2, 1), we have D > 0 and fxx > 0, so f has a local minimum.
7. To nd the critical points, we solve fx = 0 and fy = 0 for x and y. Solving
fx = 3x
2
3 = 0
fy = 3y
2
12y = 0
shows that x = 1 or x = 1 and y = 0 or y = 4. There are four critical points: (1, 0), (1, 0), (1, 4), and (1, 4).
We have
D = (fxx)(fyy) (fxy)
2
= (6x)(6y 12) (0)
2
= (6x)(6y 12).
At critical point (1, 0), we have D > 0 and fxx < 0, so f has a local maximum at (1, 0).
At critical point (1, 0), we have D < 0, so f has a saddle point at (1, 0).
At critical point (1, 4), we have D < 0, so f has a saddle point at (1, 4).
At critical point (1, 4), we have D > 0 and fxx > 0, so f has a local minimum at (1, 4).
8. Find the critical point(s) by setting
fx = (xy + 1) + (x +y) y = y
2
+ 2xy + 1 = 0,
fy = (xy + 1) + (x +y) x = x
2
+ 2xy + 1 = 0,
then we get x
2
= y
2
, and so x = y or x = y.
If x = y, then x
2
+2x
2
+1 = 0, that is, 3x
2
= 1, and there is no real solution. If x = y, then x
2
2x
2
+1 = 0,
which gives x
2
= 1. Solving it we get x = 1 or x = 1, then y = 1 or y = 1, respectively. Hence, (1, 1) and (1, 1)
are critical points.
Since
fxx(x, y) = 2y,
fxy(x, y) = 2y + 2x and
fyy(x, y) = 2x,
the discriminant is
D(x, y) = fxxfyy f
2
xy
= 2y 2x (2y + 2x)
2
= 4(x
2
+xy +y
2
).
thus
D(1, 1) = 4(1
2
+ 1 (1) + (1)
2
) = 4 < 0,
D(1, 1) = 4((1)
2
+ (1) 1 + 1
2
) = 4 < 0.
Therefore (1, 1) and (1, 1) are saddle points.
9. At a critical point, fx = 0, fy = 0.
fx = 8y (x +y)
3
= 0, we know8y = (x +y)
3
.
fy = 8x (x +y)
3
= 0, we know8x = (x +y)
3
.
Therefore we must have x = y. Since (x+y)
3
= (2y)
3
= 8y
3
, this tells us that 8y 8y
3
= 0. Solving gives y = 0, 1.
Thus the critical points are (0, 0), (1, 1), (1, 1).
15.1 SOLUTIONS 1041
fyy = fxx = 3(x +y)
2
, and fxy = 8 3(x +y)
2
.
The discriminant is
D(x, y) = fxxfyy f
2
xy
= 9(x +y)
4

_
64 48(x +y)
2
+ 9(x +y)
4
_
= 64 + 48(x +y)
2
.
D(0, 0) = 64 < 0, so (0, 0) is a saddle point.
D(1, 1) = 64 + 192 > 0 and fxx(1, 1) = 12 < 0, so (1, 1) is a local maximum.
D(1, 1) = 64 + 192 > 0 and fxx(1, 1) = 12 < 0, so (1, 1) is a local maximum.
10. To nd the critical points, we solve fx = 0 and fy = 0 for x and y. Solving
fx = 6 2x +y = 0,
fy = x 2y = 0.
We see from the second equation that x = 2y. Substituting this into the rst equation shows that y = 2. The only critical
point is (4, 2).
We have
D = (fxx)(fyy) (fxy)
2
= (2)(2) 1
2
= 3.
Since D > 0 and fxx = 2 < 0, the function f has a local maximum at (4, 2).
11. To nd the critical points, we solve fx = 0 and fy = 0 for x and y. Solving
fx = e
2x
2
+y
2
(4x) = 0,
fy = e
2x
2
+y
2
(2y) = 0,
shows that the only critical point is (0, 0).
We have
D = (fxx)(fyy) (fxy)
2
= e
2x
2
+y
2
(4 + (4x)
2
) e
2x
2
+y
2
(2 + (2y)
2
) (e
2x
2
+y
2
(4x 2y))
2
.
At (0, 0), we have D = 4 2 0
2
> 0 and fxx = 4 > 0, so the function has a local minimum at the point (0, 0).
12. At the origin f(0, 0) = 0. Since x
6
0 and y
6
0, the point (0, 0) is a local (and global) minimum. The second
derivative test does not tell you anything since D = 0.
13. At the origin g(0, 0) = 0. Since y
3
0 for y > 0 and y
3
< 0 for y < 0, the function g takes on both positive and negative
values near the origin, which must therefore be a saddle point. The second derivative test does not tell you anything since
D = 0.
14. At the origin h(0, 0) = 1. Since cos x and cos y are never above 1, the origin must be a local (and global) maximum. The
second derivative test
D = hxxhyy (hxy)
2
=
_
(cos xcos y)(cos xcos y) (sin xsin y)
2
_

x=0,y=0
=
_
cos
2
xcos
2
y sin
2
xsin
2
y
_

x=0,y=0
= 1 > 0
and hxx(0, 0) < 0, so (0, 0) is a local maximum.
15. At the origin, the second derivative test gives
D = kxxkyy (kxy)
2
=
_
(sin xsin y)(sin xsin y) (cos xcos y)
2
_

x=0,y=0
= sin
2
0 sin
2
0 cos
2
0 cos
2
0
= 1 < 0.
Thus k(0, 0) is a saddle point.
Problems
16. (a) P is a local maximum.
(b) Q is a saddle point.
(c) R is a local minimum.
(d) S is none of these.
1042 Chapter Fifteen /SOLUTIONS
17. Figure 15.1 shows the gradient vectors around P and Q pointing perpendicular to the contours and in the direction of
increasing values of the function.
y
x
0
0
0
0
1 2
3
4
5
6
1 1
2
3
4
5
6
6
1

5
6

4
5
6
6 Q R
P
S

-
.

.
.

-


Figure 15.1
18. Figure 15.2 shows the direction of f at the points where f is largest, since at those points the contours are closest
together.
y
x
0
0
0
0
1 2
3
4
5
6
1 1
2
3
4
5
6
1

5
6

4
5
6
6
Q

R
P
S
Figure 15.2
19. First, we identify the critical points. The partials are fx(x, y) = 3x
2
and fy(x, y) = 2ye
y
2
. These will vanish
simultaneously when x = 0 and y = 0, so our only critical point is (0, 0). The discriminant is
D = fxx(x, y)fyy(x, y) f
2
xy
(x, y) = (6x)(4y
2
e
y
2
2e
y
2
) 0 = 12xe
y
2
(2y
2
1).
Unfortunately, the discriminant is zero at the origin so the second derivative test can tell us nothing about our critical point.
We can, however, see that we are at a saddle point by looking at the behavior of f(x, y) along the line y = 0. Here we have
f(x, 0) = x
3
+ 1, so for positive x, we have f(x, 0) > 1 = f(0, 0) and for negative x, we have f(x, 0) < 1 = f(0, 0).
So f(x, y) has neither a maximum nor a minimum at (0, 0).
15.1 SOLUTIONS 1043
20. To nd critical points, set partial derivatives equal to zero:
Ex = sin x = 0 when x = 0, , 2,
Ey = y = 0 when y = 0.
The critical points are
(2, 0), (, 0), (0, 0), (, 0), (2, 0), (3, 0)
To classify, calculate D = ExxEyy (Exy)
2
= cos x.
At the points (0, 0), (2, 0), (4, 0), (6, 0),
D = (1) > 0 and Exx > 0 (SinceExx(0, 2k) = cos(2k) = 1).
Therefore (0, 0), (2, 0), (4, 0), (6, 0), are local minima.
At the points (, 0), (3, 0), (5, 0), (7, 0), , we have cos(2k + 1) = 1, so
D = (1) < 0.
Therefore (, 0), (3, 0), (5, 0), (7, 0), are saddle points.
21. To nd the critical points, we must solve the equations
f
x
= e
x
(1 cos y) = 0
f
y
= e
x
(sin y) = 0.
The rst equation has solution
y = 0, 2, 4, . . . .
The second equation has solution
y = 0, , 2, 3, . . . .
Since x can be anything, the lines
y = 0, 2, 4, . . .
are lines of critical points.
We calculate
D = (fxx)(fyy) (fxy)
2
= e
x
(1 cos y)e
x
cos y (e
x
sin y)
2
= e
2x
(cos y cos
2
y sin
2
y)
= e
2x
(cos y 1)
At any critical point on one of the lines y = 0, y = 2, y = 4, . . .,
D = e
2x
(1 1) = 0.
Thus, D tells us nothing. However, all along these critical lines, the value of the function, f, is zero. Since the function f
is never negative, the critical points are all both local and global minima.
22. At a critical point,
fx = cos xsin y = 0 so cos x = 0 or sin y = 0;
and
fy = sin xcos y = 0 so sin x = 0 or cos y = 0.
Case 1: Assume cos x = 0. This gives
x =
3
2
,

2
,

2
,
3
2
,
(This can be written more compactly as: x = k +/2, for k = 0, 1, 2, .)
If cos x = 0, then sin x = 1 = 0. Thus in order to have fy = 0 we need cos y = 0, giving
y =
3
2
,

2
,

2
,
3
2
,
1044 Chapter Fifteen /SOLUTIONS
(More compactly, y = l +/2, for l = 0, 1, 2, )
Case 2: Assume sin y = 0. This gives
y = 2, , 0, , 2,
(More compactly, y = l, for l = 0, 1, 2, )
If sin y = 0, then cos y = 1 = 0, so to get fy = 0 we need sin x = 0, giving
x = , 2, , 0, , 2,
(More compactly, x = k for k = 0, 0, 1, 2, )
Hence we get all the critical points of f(x, y). Those from Case 1 are as follows:

_

2
,

2
_
,
_

2
,

2
_
,
_

2
,
3
2
_


_

2
,

2
_
,
_

2
,

2
_
,
_

2
,
3
2
_


_
3
2
,

2
_
,
_
3
2
,

2
_
,
_
3
2
,
3
2
_

Those from Case 2 are as follows:
(, ), (, 0), (, ), (, 2)
(0, ), (0, 0), (0, ), (0, 2)
(, ), (, 0), (, ), (, 2)
More compactly these points can be written as,
(k, l), for k = 0, 1, 2, , l = 0, 1, 2,
and (k +

2
, l +

2
), for k = 0, 1, 2, , l = 0, 1, 2,
To classify the critical points, we nd the discriminant. We have
fxx = sin xsin y, fyy = sin xsin y, and fxy = cos xcos y.
Thus the discriminant is
D(x, y) = fxxfyy f
2
xy
= (sin xsin y)(sin xsin y) (cos xcos y)
2
= sin
2
xsin
2
y cos
2
xcos
2
y
= sin
2
y cos
2
x. (Use: sin
2
x = 1 cos
2
x and factor.)
At points of the form (k, l) where k = 0, 1, 2, ; l = 0, 1, 2, , we have
D(x, y) = 1 < 0 so (k, l) are saddle points.
At points of the form (k +

2
, l +

2
) where k = 0, 1, 2, ; l = 0, 1, 2,
D(k +

2
, l +

2
) = 1 > 0, so we have two cases:
If k and l are both even or k and l are both odd, then
fxx = sin xsin y = 1 < 0, so (k +

2
, l +

2
) are local maximum points.
If k is even but l is odd or k is odd but l is even, then
fxx = 1 > 0 so (k +

2
, l +

2
) are local minimum points.
23. At a local maximum value of f,
f
x
= 2x B = 0.
We are told that this is satised by x = 2. So 2(2) B = 0 and B = 4. In addition,
f
y
= 2y C = 0
and we know this holds for y = 1, so 2(1) C = 0, giving C = 2. We are also told that the value of f is 15 at the
point (2, 1), so
15 = f(2, 1) = A((2)
2
+ 4(2) + 1
2
2(1)) = A(5), so A = 10.
15.1 SOLUTIONS 1045
Now we check that these values of A, B, and C give f(x, y) a local maximum at the point (2, 1). Since
fxx(2, 1) = 2,
fyy(2, 1) = 2
and
fxy(2, 1) = 0,
we have that fxx(2, 1)fyy(2, 1) f
2
xy
(2, 1) = (2)(2) 0 > 0 and fxx(2, 1) < 0. Thus, f has a local
maximum value 15 at (2, 1).
24. (a) (1, 3) is a critical point. Since fxx > 0 and the discriminant
D = fxxfyy f
2
xy
= fxxfyy 0
2
= fxxfyy > 0,
the point (1, 3) is a minimum.
(b) See Figure 15.3.
1
3
x
y

0
1
4
1
6 3
2
6
4
1
2
0
Figure 15.3
1
1
x
y
0
0
0
0
1
3
5
7
9
1
3
5
7
9

7
Figure 15.4
25. (a) (a, b) is a critical point. Since the discriminant D = fxxfyy f
2
xy
= f
2
xy
< 0, (a, b) is a saddle point.
(b) See Figure 15.4.
26. Begin by constructing little pieces of the contour diagram around each of the points (1, 0), (3, 3), and (3, 3) where
some information about f is given. The general shape will be as in Figure 15.5, and the directions of increasing contour
values are indicated for each part. Then complete the diagram in any way. One possible solution is given in Figure 15.6.
Figure 15.5: Part of contour
diagram with arrows showing
direction of increasing
function values
0
1
2
0
1
2
3
4
Figure 15.6: Contour diagram
of f(x, y)
1046 Chapter Fifteen /SOLUTIONS
27. Since (2, 4) is a local minimum, the contours around (2, 4) are closed curves with increasing values as we go away from
the point (2, 4). We assume that the function values continue to increase as we move parallel to the y-axis to the point
(2, 1). Since (2, 1) is a saddle point, we draw the contours so that the values go down as we move up or down from this
point, and up as we move to the left or right. One possible contour diagram is shown in Figure 15.7.
1 1 2 3 4 5 6
1
1
2
3
4
5
6
0 1 2 3 4
1
1
2
3
4
x
y
Figure 15.7
f
=

1
f
=

1
f
=
0
f
=
0
f
=
1
f
=
1
(0, 0)
x
y
Figure 15.8
28. (a) Setting the partial derivatives equal to 0, we have
fx(x, y) = 2x(x
2
+y) + 2x(x
2
y) = 4x
3
= 0
fy(x, y) = (x
2
+y) + (x
2
y) = 2y = 0.
Thus, (0, 0) is the only critical point.
(b) Calculating D gives
D = (fxx)(fyy) (fxy)
2
= (12x
2
)(2) 0
2
= 24x
2
.
At x = 0, y = 0, we have
D(0, 0) = 0.
Thus the second derivative test tells us nothing about the nature of the critical point.
(c) Since f(0, 0) = 0, we sketch contours with values near 0. The contour f = 0 is given by
(x
2
y)(x
2
+y) = 0,
that is, the two parabolas
y = x
2
and y = x
2
We also sketch the contours f = 1 and f = 1. See Figure 15.8.
Since there are values of the function which are both positive (above f(0, 0)) and negative (below (f(0, 0)),
near the critical point (0, 0), the origin is neither a local maximum nor a local minimum; it is a saddle point.
29. We have fx = 2kx 4y and fy = 2y 4x, so fxx = 2k, fxy = 4, and fyy = 2. The discriminant is
D = (fxx)(fyy) (fxy)
2
= (2k)(2) (4)
2
= 4k 16.
Since D = 4k 16, we see that D < 0 when k < 4. The function has a saddle point at the point (0, 0) when k < 4.
When k > 4, we have D > 0 and fxx > 0, so the function has a local minimum at the point (0, 0). When k = 4,
the discriminant is zero, and we get no information about this critical point. By looking at the values of the function in
Table 15.1, it appears that f has a local minimum at the point (0, 0) when k = 4.
Table 15.1
y
x
0.1 0 0.1
0.1 0.01 0.01 0.09
0 0.04 0 0.04
0.1 0.09 0.01 0.01
(a) The function f(x, y) has a saddle point at (0, 0) if k < 4.
15.1 SOLUTIONS 1047
(b) There are no values of k for which this function has a local maximum at the point (0, 0).
(c) The function f(x, y) has a local minimum at (0, 0) if k 4.
30. The rst order partial derivatives are
fx(x, y) = 2kx 2y and fy(x, y) = 2ky 2x.
And the second order partial derivatives are
fxx(x, y) = 2k fxy(x, y) = 2 fyy(x, y) = 2k
Since fx(0, 0) = fy(0, 0) = 0, the point (0, 0) is a critical point. The discriminant is
D = (2k)(2k) 4 = 4(k
2
1).
For k = 2, the discriminant is positive, D = 12. When k = 2, fxx(0, 0) = 4 which is positive so we have a local
minimum at the origin. When k = 2, fxx(0, 0) = 4 so we have a local maximum at the origin. In the case k = 0,
D = 4 so the origin is a saddle point.
Lastly, when k = 1 the discriminant is zero, so the second derivative test can tell us nothing. Luckily, we can factor
f(x, y) when k = 1. When k = 1,
f(x, y) = x
2
2xy +y
2
= (x y)
2
.
This is always greater than or equal to zero. So f(0, 0) = 0 is a minimum and the surface is a trough-shaped parabolic
cylinder with its base along the line x = y.
When k = 1,
f(x, y) = x
2
2xy y
2
= (x +y)
2
.
This is always less than or equal to zero. So f(0, 0) = 0 is a maximum. The surface is a parabolic cylinder, with its top
ridge along the line x = y.
1
4
8
12
16
y
x
k = 2
1
1
5
5
10
10
20
20
30
30
k = 1
x
y
k = 0
1
1
4
4
8
8
12
12
16
16
4
4
8
8
12
12
16
16
1
1
x
y
1
1
5
5
10
10
20
20
30
30
k = 1
x
y
1
4
8
12
16
y
x
k = 2
Figure 15.9
1048 Chapter Fifteen /SOLUTIONS
31. The partial derivatives are
fx(x, y) = 3x
2
3y
2
and fy(x, y) = 6xy.
Now fx(x, y) will vanish if x = y and fy(x, y) will vanish if either x = 0 or y = 0. Since the partial derivatives
are dened everywhere, the only critical points are where fx(x, y) and fy(x, y) vanish simultaneously. (0, 0) is the only
critical point.
To nd the contour for f(x, y) = 0, we solve the equation x
3
3xy
2
= 0. This can be factored into
f(x, y) = x(x

3 y)(x +

3 y) = 0
whose roots are x = 0, x =

3 y and x =

3 y. Each of these roots describes a line through the origin; the three of
them divide the plane into six regions. Crossing any one of these lines will change the sign of only one of the three factors
of f(x, y), which will change the sign of f(x, y).
2
1 2 1
0
0
1
2
0
1 1
2
0
1
2
0
1 1
2

y = x

y = x

3
f > 0 f < 0
f < 0 f > 0
f > 0 f < 0
x
y
Figure 15.10
Solutions for Section 15.2
Exercises
1. Mississippi lies entirely within a region designated as 80s so we expect both the maximum and minimum daily high
temperatures within the state to be in the 80s. The southwestern-most corner of the state is close to a region designated as
90s, so we would expect the temperature here to be in the high 80s, say 87-88. The northern-most portion of the state is
located near the center of the 80s region. We might expect the high temperature there to be between 83-87.
Alabama also lies completely within a region designated as 80s so both the high and low daily high temperatures
within the state are in the 80s. The southeastern tip of the state is close to a 90s region so we would expect the temperature
here to be about 88-89 degrees. The northern-most part of the state is near the center of the 80s region so the temperature
there is 83-87 degrees.
Pennsylvania is also in the 80s region, but it is touched by the boundary line between the 80s and a 70s region. Thus
we expect the low daily high temperature to occur there and be about 80 degrees. The state is also touched by a boundary
line of a 90s region so the high will occur there and be 89-90 degrees.
New York is split by a boundary between an 80s and a 70s region, so the northern portion of the state is likely to be
about 74-76 while the southern portion is likely to be in the low 80s, maybe 81-84 or so.
California contains many different zones. The northern coastal areas will probably have the daily high as low as
65-68, although without another contour on that side, it is difcult to judge how quickly the temperature is dropping off
to the west. The tip of Southern California is in a 100s region, so there we expect the daily high to be 100-101.
Arizona will have a low daily high around 85-87 in the northwest corner and a high in the 100s, perhaps 102-107 in
its southern regions.
Massachusetts will probably have a high daily high around 81-84 and a low daily high of 70.
15.2 SOLUTIONS 1049
2. Let the line be in the form y = b +mx. When x equals 1, 0 and 1, then y equals b m, b, and b +m, respectively. The
sum of the squares of the vertical distances, which is what we want to minimize, is
f(m, b) = (2 (b m))
2
+ (1 b)
2
+ (1 (b +m))
2
.
To nd the critical points, we compute the partial derivatives with respect to m and b,
fm = 2(2 b +m) + 0 + 2(1 b m)(1)
= 4 2b + 2m2 + 2b + 2m
= 2 + 4m,
f
b
= 2(2 b +m)(1) + 2(1 b)(1) + 2(1 b m)(1)
= 4 + 2b 2m+ 2 + 2b 2 + 2b + 2m
= 4 + 6b.
Setting both partial derivatives equal to zero, we get a system of equations:
2 + 4m = 0,
4 + 6b = 0.
The solution is m = 1/2 and b = 2/3. One can check that it is a minimum. Hence, the regression line is y =
2
3

1
2
x.
3. The function f has no global maximum or global minimum.
4. The function g has a global minimum (it is 0) but no global maximum.
5. The function h has no global maximum or minimum.
6. Since f(x, y) 0 for all x, y and sincef(0, 0) = 0, the function has a global maximum (it is 0) and no global minimum.
7. Suppose x is xed. Then for large values of y the sign of f is determined by the highest power of y, namely y
3
. Thus,
f(x, y) as y
f(x, y) as y .
So f does not have a global maximum or minimum.
8. To maximize z = x
2
+ y
2
, it sufces to maximize x
2
and y
2
. We can maximize both of these at the same time by
taking the point (1, 1), where z = 2. It occurs on the boundary of the square. (Note: We also have maxima at the points
(1, 1), (1, 1) and (1, 1) which are on the boundary of the square.)
To minimize z = x
2
+y
2
, we choose the point (0, 0), where z = 0. It does not occur on the boundary of the square.
9. To maximize z = x
2
y
2
it sufces to minimize x
2
and y
2
. Thus, the maximum is at (0, 0), where z = 0. It does not
occur on the boundary of the square.
To minimize z = x
2
y
2
, it sufces to maximize x
2
and y
2
. Do this by taking the point (1, 1), (1, 1), (1, 1),
or (1, 1) where z = 2. These occur on the boundary of the square.
10. To maximize this function, it sufces to maximize x
2
and minimize y
2
. We can do this by choosing the point (1, 0), or
(1, 0) where z = 1. These occur on the boundary of the square.
To minimize z = x
2
y
2
, it sufces to maximize y
2
and minimize x
2
. We can do this by taking the point (0, 1), or
(0, 1) where z = 1. These occur on the boundary of the square.
11. The maximum value, which is slightly above 30, say 30.5, occurs approximately at the origin. The minimum value, which
is about 20.5, occurs at (2.5, 5).
12. The maxima occur at about (/2, 0) and (/2, 2). The minimum occurs at (/2, ). The maximum value is about 1,
the minimum value is about 1.
13. The maximum value, which is about 11, occurs at (5.1, 4.9). The minimum value, which is about 1, occurs at (1, 3.9).
Problems
14. To nd critical points of g we solve
gx(x, y) = Ax +By D = 0
gy(x, y) = Bx +Cy E = 0
1050 Chapter Fifteen /SOLUTIONS
which is equivalent to
Ax +By = D
Bx +Cy = E.
The second derivative test shows that the critical point gives a minimum for g because gxxgyy g
2
xy
= AC B
2
> 0
and gxx = A > 0.
15. The variables are a and b, so we set
S
a
= 2(a +b) + 8(4a +b 2) + 18(9a +b 4) = 0
S
b
= 2(a +b) + 2(4a +b 2) + 2(9a +b 4) = 0,
so, collecting terms and dividing by 4 and 2 respectively,
49a + 7b 22 = 0
14a + 3b 6 = 0.
Solving gives a = 24/49, b = 2/7.
Since there is only one critical point and S is unbounded as a, b , this critical point is the global minimum.
Therefore, the best tting parabola is
y =
24
49
x
2

2
7
.
16. We must nd b and m to minimize the function
g(b, m) =
_
1
0
(x
2
(b +mx))
2
dx
=
_
1
0
(x
4
2bx
2
+b
2
2mx
3
+ 2bmx +m
2
x
2
)dx
=
1
5

2
3
b +b
2

m
2
+bm+
m
2
3
.
The critical points of g are given by the equations
g
b
=
2
3
+ 2b +m = 0
g
m
=
1
2
+b +
2
3
m = 0.
The solution is b = 1/6 and m = 1. The linear least squares approximation of x
2
on the interval [0, 1] is L(x, y) =
1/6 +x.
17. We must nd b and m to minimize the function
g(b, m) =
_
1
0
(x
3
(b +mx))
2
dx
=
_
1
0
(x
6
2bx
3
+b
2
2mx
4
+ 2bmx +m
2
x
2
)dx
=
1
6

1
2
b +b
2

2m
5
+bm+
m
2
3
.
The critical points of g are given by the equations
g
b
=
1
2
+ 2b +m = 0
g
m
=
2
5
+b +
2
3
m = 0.
The solution is b = 1/5 and m = 9/10. The linear least squares approximation of x
3
on the interval [0, 1] is L(x, y) =
0.2 + 0.9x.
15.2 SOLUTIONS 1051
18. We calculate the partial derivatives and set them to zero.
(range)
t
= 10t 6h + 400 = 0
(range)
h
= 6t 6h + 300 = 0.
10t + 6h = 400
6t + 6h = 300
solving we obtain
4t = 100
so
t = 25
Solving for h, we obtain 6h = 150, yielding h = 25. Since the range is quadratic in h and t, the second derivative test
tells us this is a local and global maximum. So the optimal conditions are h = 25% humidity and t = 25

C.
19. (a) This tells us that an increase in the price of either product causes a decrease in the quantity demanded of both
products. An example of products with this relationship is tennis rackets and tennis balls. An increase in the price of
either product is likely to lead to a decrease in the quantity demanded of both products as they are used together. In
economics, it is rare for the quantity demanded of a product to increase if its price increases, so for q1, the coefcient
of p1 is negative as expected. The coefcient of p2 in the expression could be either negative or positive. In this case,
it is negative showing that the two products are complementary in use. If it were positive, however, it would indicate
that the two products are competitive in use, for example Coke and Pepsi.
(b) The revenue from the rst product would be q1p1 = 150p1 2p
2
1
p1p2, and the revenue from the second product
would be q2p2 = 200p2 p1p2 3p
2
2
. The total sales revenue of both products, R, would be
R(p1, p2) = 150p1 + 200p2 2p1p2 2p
2
1
3p
2
2
.
Note that R is a function of p1 and p2. To nd the critical points of R, set R = 0, i.e.,
R
p1
=
R
p2
= 0.
This gives
R
p1
= 150 2p2 4p1 = 0
and
R
p2
= 200 2p1 6p2 = 0
Solving simultaneously, we have p1 = 25 and p2 = 25. Therefore the point (25, 25) is a critical point for R. Further,

2
R
p
2
1
= 4,

2
R
p
2
2
= 6,

2
R
p1p2
= 2,
so the discriminant at this critical point is
D = (4)(6) (2)
2
= 20.
Since D > 0 and
2
R/p
2
1
< 0, this critical point is a local maximum. Since R is quadratic in p1 and p2, this is a
global maximum. Therefore the maximum possible revenue is
R = 150(25) + 200(25) 2(25)(25) 2(25)
2
3(25)
2
= (6)(25)
2
+ 8(25)
2
7(25)
2
= 4375.
This is obtained when p1 = p2 = 25. Note that at these prices, q1 = 75 units, and q2 = 100 units.
1052 Chapter Fifteen /SOLUTIONS
20. The total revenue is
R = pq = (60 0.04q)q = 60q 0.04q
2
,
and as q = q1 +q2, this gives
R = 60q1 + 60q2 0.04q
2
1
0.08q1q2 0.04q
2
2
.
Therefore, the prot is
P(q1, q2) = R C1 C2
= 13.7 + 60q1 + 60q2 0.07q
2
1
0.08q
2
2
0.08q1q2.
At a local maximum point, we have grad P =

0 :
P
q1
= 60 0.14q1 0.08q2 = 0,
P
q2
= 60 0.16q2 0.08q1 = 0.
Solving these equations, we nd that
q1 = 300 and q2 = 225.
To see whether or not we have found a local maximum, we compute the second-order partial derivatives:

2
P
q
2
1
= 0.14,

2
P
q
2
2
= 0.16,

2
P
q1q2
= 0.08.
Therefore,
D =

2
P
q
2
1

2
P
q
2
2


2
P
q1q2
= (0.14)(0.16) (0.08)
2
= 0.016,
and so we have found a local maximum point. The graph of P(q1, q2) has the shape of an upside down paraboloid since
P is quadratic in q1 and q2, hence (300, 225) is a global maximum point.
21. (a) The revenue R = p1q1 +p2q2. Prot = P = R C = p1q1 +p2q2 2q
2
1
2q
2
2
10.
P
q1
= p1 4q1 = 0 gives q1 =
p1
4
P
q2
= p2 4q2 = 0 gives q2 =
p2
4
Since

2
P
q
2
1
= 4,

2
P
q
2
2
= 4 and

2
P
q
1
q
2
= 0, at (p1/4, p2/4) we have that the discriminant, D = (4)(4) > 0
and

2
P
q
2
1
< 0, thus P has a local maximum value at (q1, q2) = (p1/4, p2/4). Since P is quadratic in q1 and q2, this
is a global maximum. So P =
p
2
1
4
+
p
2
2
4
2
p
2
1
16
2
p
2
2
16
10 =
p
2
1
8
+
p
2
2
8
10 is the maximum prot.
(b) The rate of change of the maximum prot as p1 increases is

p1
(max P) =
2p1
8
=
p1
4
.
22. We want to maximize the theaters prot, P, as a function of the two variables (prices) pc and pa. As always, P = RC,
where R is the revenue, R = qcpc + qapa, and C is the cost, which is of the form C = k(qc + qa) for some constant k.
Thus,
P(pc, pa) = qcpc +qapa k(qc +qa)
= rp
3
c
krp
4
c
+sp
1
a
ksp
2
a
To nd the critical points, solve
P
pc
= 3rp
4
c
+ 4krp
5
c
= 0
P
pa
= sp
2
a
+ 2ksp
3
a
= 0.
We get pc = 4k/3 and pa = 2k.
15.2 SOLUTIONS 1053
This critical point is a global maximum by the following useful, general argument. Suppose that F(x, y) = f(x) +
g(y), where f has a global maximum at x = b and g has a global maximum at y = d. Then for all x, y:
F(x, y) = f(x) +g(y) f(b) +g(d) = F(b, d),
so F has global maximum at x = b, y = d.
The prot function in this problem has the form
P(pc, pa) = f(pc) +g(pa),
and the usual single-variable calculus argument using f

and g

shows that pc = 4k/3 and pa = 2k are global maxima


for f and g, respectively. Thus the maximum prot occurs when pc = 4k/3 and pa = 2k. Thus,
pc
pa
=
4k/3
2k
=
2
3
.
23. Let the sides be x, y, z cm. Then the volume is given by V = xyz = 32.
The surface area S is given by
S = 2xy + 2xz + 2yz.
Substituting z = 32/(xy) gives
S = 2xy +
64
y
+
64
x
.
At a critical point,
S
x
= 2y
64
x
2
= 0
S
y
= 2x
64
y
2
= 0,
The symmetry of the equations (or by dividing the equations) tells us that x = y and
2x
64
x
2
= 0
x
3
= 32
x = 32
1/3
= 3.17 cm.
Thus the only critical point is x = y = (32)
1/3
cm and z = 32/
_
(32)
1/3
(32)
1/3
_
= (32)
1/3
cm. At the critical point
SxxSyy (Sxy)
2
=
128
x
3

128
y
3
2
2
=
(128)
2
x
3
y
3
4.
Since D > 0 and Sxx > 0 at this critical point, the critical point x = y = z = (32)
1/3
is a local minimum. Since
S as x, y , the local minimum is a global minimum.
24. Let the sides of the base be x and y cm. Let the height be z cm. Then the volume is given by xyz = 32 and the surface
area, S, is given by
S = xy + 2xz + 2yz.
Substituting z = 32/(xy) gives
S = xy +
64
y
+
64
x
.
At a critical point
S
x
= y
64
x
2
= 0
S
y
= x
64
y
2
= 0.
The symmetry of the equations tells us that x = y and
x
64
x
2
= 0
x
3
= 64
x = 4 cm.
1054 Chapter Fifteen /SOLUTIONS
Thus the only critical point is x = y = 4 cm and z = 32/(4 4) = 2 cm. At the critical point
D = SxxSyy (Sxy)
2
=
128
x
3

128
y
3
1
2
=
(128)
2
x
3
y
3
1.
Since D > 0 and Sxx > 0 at this critical point, the critical point x = y = 4, z = 2 is a local minimum. Since S as
x, y , the local minimum is a global minimum.
25. If the coordinates of the corner on the plane are (x, y, z), the volume of the box is V = xyz. Since z = 1 3x 2y on
the plane, the volume is given by
V = xy(1 3x 2y) = xy 3x
2
y 2xy
2
.
The domain is the triangular region 0 x
1
3
, 0 y (1 3x)/2. At a critical point,
V
x
= y 6xy 2y
2
= y(1 6x 2y) = 0
V
y
= x 3x
2
4xy = x(1 3x 4y) = 0,
One solution is x = y = 0. Another is x = 0, y =
1
2
; another is y = 0, x =
1
3
. Another is the solution of
1 6x 2y = 0
1 3x 4y = 0,
namely x =
1
9
, y =
1
6
.
If either x = 0 or y = 0, then V = 0, so these solutions do not give the maximum volume. Since
D = VxxVyy (Vxy)
2
= (6y)(4x) (1 6x 4y)
2
D
_
1
9
,
1
6
_
=
_
6
1
6
__
4
1
9
_

_
1 6
1
9
4
1
6
_
2
=
4
9

1
9
=
1
3
> 0,
and Vxx(
1
9
,
1
6
) = 1 < 0, the point x =
1
9
, y =
1
6
, is a local maximum at which V = (1/9)(1/6) 3(1/9)
2
(1/6)
2(1/9)(1/6)
2
= 1/162.
Since all points on the boundary of the domain give V = 0, the local maximum is a global maximum.
26.
w
l
h
Figure 15.11
Let w, h and l be width, height and length of the suitcase in cm. Then its volume V = lwh, and w + h + l 135.
To maximize the volume V , choose w +h +l = 135, and thus l = 135 w h,
V = wh(135 w h)
= 135wh w
2
h wh
2
Differentiating gives
Vw = 135h 2wh h
2
,
V
h
= 135w w
2
2wh.
Find the critical points by solving Vw = 0 and V
h
= 0:
Vw = 0 gives 135h h
2
= 2wh,
V
h
= 0 gives 135w w
2
= 2wh.
15.2 SOLUTIONS 1055
As hw = 0, we cancel h (and w respectively) in the above equations and get
135 h = 2w
135 w = 2h
Subtracting gives
w h = 2(w h)
hence w = h. Therefore, substituting into the equation Vw = 0
135h h
2
= 2h
2
and therefore
3h
2
= 135h.
Since h = 0, we have
h =
135
3
= 45.
So w = h = 45 cm. Thus, l = 135 w h = 45 cm. To check that this critical point is a maximum, we nd
Vww = 2h, V
hh
= 2w,
V
wh
= 135 2w 2h,
so
D = VwwV
hh
V
2
wh
= 4hw (135 2w 2h)
2
.
At w = h = 45, we have Vww = 2(45) < 0 and D = 4(45)
2
(135 90 90)
2
= 6075 > 0, hence V is maximum
at w = h = l = 45.
Therefore, the suitcase with maximum volume is a cube with dimensions width = height = length = 45 cm.
27.
w
l
h
Figure 15.12
The box is shown in Figure 15.12. Cost of four sides = (2hl + 2wh)(1)c/. Cost of two bottoms = (2wl)(2)c/. Thus
the total cost C (in cents) of the box is
C = 2(hl +wh) + 4wl.
But volume wlh = 512, so l = 512/(wh), thus
C =
1024
w
+ 2wh +
2048
h
.
To minimize C, nd the critical points of C by solving
C
h
= 2w
2048
h
2
= 0,
Cw = 2h
1024
w
2
= 0.
We get
2wh
2
= 2048
2hw
2
= 1024.
1056 Chapter Fifteen /SOLUTIONS
Since w, h = 0, we can divide the rst equation by the second giving
2wh
2
2hw
2
=
2048
1024
,
so
h
w
= 2,
thus
h = 2w.
Substituting this in C
h
= 0, we obtain h
3
= 2048, so h = 12.7 cm. Thus w = h/2 = 6.35 cm, and l = 512/(wh) =
6.35 cm. Now we check that these dimensions minimize the cost C. We nd that
D = C
hh
Cww C
2
hw
= (
4096
h
3
)(
2048
w
3
) 2
2
,
and at h = 12.7, w = 6.35, C
hh
> 0 and D = 16 4 > 0, thus C has a local minimum at h = 12.7 and w = 6.35.
Since C increases without bound as w, h 0 or , this local minimum must be a global minimum.
Therefore, the dimensions of the box that minimize the cost are w = 6.35 cm, l = 6.35 cm and h = 12.7 cm.
28. The square of the distance from the point (x, y, z) to the origin is
S = x
2
+y
2
+z
2
.
If the point is on the plane, z = 1 3x 2y, we have
S = x
2
+y
2
+ (1 3x 2y)
2
.
At the critical point
S
x
= 2x + 2(1 3x 2y)(3) = 2(10x + 6y 3) = 0
S
y
= 2y + 2(1 3x 2y)(2) = 2(6x + 5y 2) = 0.
Simplifying gives
10x + 6y = 3
6x + 5y = 2,
with solution x = 3/14, y = 1/7. At this point z = 1/14. We have
D = SxxSyy (Sxy)
2
= (20)(10) 12
2
= 56,
so D > 0 and Sxx > 0. Thus, the point x = 3/14, y = 1/7 is a local minimum. Since S as x, y , the local
minimum is a global minimum. Thus, x = 3/14, y = 1/7, z = 1/14 is the closest point to the origin on the plane.
29. We minimize the square of the distance from the point (x, y, z) to the origin:
S = x
2
+y
2
+z
2
.
Since z
2
= 9 xy 3x, we have
S = x
2
+y
2
+ 9 xy 3x.
At a critical point
S
x
= 2x y 3 = 0
S
y
= 2y x = 0,
so x = 2y, and
2(2y) y 3 = 0
giving y = 1, so x = 2 and z
2
= 9 2 1 3 2 = 1, so z = 1. We have
D = SxxSyy (Sxy)
2
= 2 2 (1)
2
= 4 1 > 0,
so, since D > 0 and Sxx > 0, the critical points are local minima. Since S as x, y , the local minima are
global minima.
If x = 2, y = 1, z = 1, we have S = 2
2
+ 1
2
+ 1
2
= 6, so the shortest distance to the origin is

6.
15.2 SOLUTIONS 1057
30. (a) Let t be the number of years since 1960 and let P(t) be the population in millions in the year 1960 + t. We assume
that P = Ce
at
, and therefore
ln P = at + ln C.
So, we plot ln P against t and nd the line of best t. Our data points are (0, ln 180), (10, ln 206), and (20, ln 226).
Applying the method of least squares to nd the best-tting line, we nd that
a =
ln 226 ln 180
20
0.0114,
ln C =
ln 206
3

ln 226
6
+
5 ln 180
6
5.20
Then, C = e
5.20
= 181.3 and so
P(t) = 181.3e
0.0114t
.
In 1990, we have t = 30 and the predicted population in millions is
P(30) = 181.3e
0.01141(30)
= 255.3.
(b) The difference between the actual and the predicted population is about 6 million or 2
1
2
%. Given that only three data
points were used to calculate a and c, this discrepancy is not surprising. Thus, the 1990 census data does not mean
that the assumption of exponential growth is unjustied.
(c) In 2010, we have t = 50 and P(50) = 320.7.
31. Let P(K, L) be the prot obtained using K units of capital and L units of labor. The cost of production is given by
C(K, L) = kK +L,
and the revenue function is given by
R(K, L) = pQ = pAK
a
L
b
.
Hence, the prot is
P = R C = pAK
a
L
b
(kK +L).
In order to nd local maxima of P, we calculate the partial derivatives and see where they are zero. We have:
P
K
= apAK
a1
L
b
k,
P
L
= bpAK
a
L
b1
.
The critical points of the function P(K, L) are solutions (K, L) of the simultaneous equations:
k
a
= pAK
a1
L
b
,

b
= pAK
a
L
b1
.
Multiplying the rst equation by K and the second by L, we get
kK
a
=
L
b
,
and so
K =
a
kb
L.
Substituting for K in the equation k/a = pAK
a1
L
b
, we get:
k
a
= pA
_
a
kb
_
a1
L
a1
L
b
.
We must therefore have
L
1ab
= pA
_
a
k
_
a
_

b
_
a1
.
Hence, if a +b = 1,
L =
_
pA
_
a
k
_
a
_

b
_
(a1)
_
1/(1ab)
,
1058 Chapter Fifteen /SOLUTIONS
and
K =
a
kb
L =
a
kb
_
pA
_
a
k
_
a
_

b
_
(a1)
_
1/(1ab)
.
To see if this is really a local maximum, we apply the second derivative test. We have:

2
P
K
2
= a(a 1)pAK
a2
L
b
,

2
P
L
2
= b(b 1)pAK
a
L
b2
,

2
P
KL
= abpAK
a1
L
b1
.
Hence,
D =

2
P
K
2

2
P
L
2

_

2
P
KL
_
2
= ab(a 1)(b 1)p
2
A
2
K
2a2
L
2b2
a
2
b
2
p
2
A
2
K
2a2
L
2b2
= ab((a 1)(b 1) ab)p
2
A
2
K
2a2
L
2b2
= ab(1 a b)p
2
A
2
K
2a2
L
2b2
.
Now a, b, p, A, K, and L are positive numbers. So, the sign of this last expression is determined by the sign of 1 a b.
(a) We assumed that a + b < 1, so D > 0, and as 0 < a < 1, then
2
P/K
2
< 0 and so we have a unique local
maximum. To verify that the local maximum is a global maximum, we focus on the cost. Let C = kK + L. Since
K 0 and L 0, K C/k and L C/. Therefore the prot satises:
P = pAK
a
L
b
(kK +L)
pA
_
C
k
_
a
_
C

_
b
C
= mC
a+b
C
where m = pA(1/k)
a
(1/)
b
. Since a +b < 1, the prot is negative for large costs C, say C C0 (C0 = m
1ab
will do). Therefore, in the KL-plane for K 0 and L 0, the prot is less than or equal to zero everywhere on or
above the line kK +L = Co. Thus the global maximum must occur inside the triangle bounded by this line and the
K and L axes. Since P 0 on the K and L axes as well, the global maximum must be in the interior of the triangle
at the unique local maximum we found.
In the case a + b < 1, we have decreasing returns to scale. That is, if the amount of capital and labor used is
multiplied by a constant > 0, we get less than times the production.
(b) Now suppose a +b 1. If we multiply K and L by for some > 0, then
Q(K, L) = A(K)
a
(L)
b
=
a+b
Q(K, L).
We also see that
C(K, L) = C(K, L).
So if a +b = 1, we have
P(K, L) = P(K, L).
Thus, if = 2, so we are doubling the inputs K and L, then the prot P is doubled and hence there can be no
maximum prot.
If a +b > 1, we have increasing returns to scale and there can again be no maximum prot: doubling the inputs
will more than double the prot. In this case, the prot increases without bound as K, L go toward innity.
32. We have
fx = 2x(y + 1)
3
= 0 only when x = 0 or y = 1
fy = 3x
2
(y + 1)
2
+ 2y = 0 never when y = 1 and only for y = 0 when x = 0
We conclude that fx = 0 and fy = 0 only when x = 0, y = 0, so f has only one critical point, namely (0, 0).
15.2 SOLUTIONS 1059
The second derivative test at (0, 0) gives
D = fxxfyy (fxy)
2
= 2(y + 1)
3
(6x
2
(y + 1) + 2) (6x(y + 1)
2
)
2
= 2(1)(2) 0 > 0 when x = 0, y = 0
Since fxx > 0 at (0, 0), this means f has a local minimum at (0, 0).
[Alternatively, if we expand (y + 1)
3
, then we can view f(x, y) as x
2
+y
2
+ (terms of degree 3 or greater in x and
y), which means that f behaves likes x
2
+y
2
near (0, 0).]
Although (0, 0) is a local minimum, it cannot be a global minimum since for xed x, say x = 1, the function f(x, y)
is a cubic polynomial in y and cubics take on arbitrarily large positive and negative values.
In the single-variable case, suppose a function f dened on the real line is differentiable and its derivative is con-
tinuous. Then if f has only one critical point, say x = 0, then if that critical point is a local minimum, it must also be a
global minimum. This is because f

cannot change sign without f

= 0 so we must have f

< 0 for x < 0 and f

> 0
for x > 0. Thus f is decreasing for all x < 0 and increasing for all x > 0, which makes x = 0 the global minimum for
f.
33. (a) We have f(2, 1) = 120.
(i) If x > 20 then f(x, y) > 10x > 200 > f(2, 1).
(ii) If y > 20 then f(x, y) > 20y > 400 > f(2, 1).
(iii) If x < 0.01 and y 20 then f(x, y) > 80/(xy) > 80/((0.01)(20)) = 400 > f(2, 1).
(iv) If y < 0.01 and x 20 then f(x, y) > 80/(xy) > 80/((20)(0.01)) = 400 > f(2, 1).
(b) The continuous function f must achieve a minimum at some point (x0, y0) in the closed and bounded region R

:
0.01 x 20, 0.01 y 20. Since (2, 1) is in R

, we must have f(x0, y0) f(2, 1). By part (a), f(x0, y0)
is less than all values of f in the part of R that is outside R

, so f(x0, y0) is a minimum for f on all of R. Since


(x0, y0) is not on the boundary of R, it must be a critical point of f.
(c) The only critical point of f in R is the point (2, 1), so by part (b) f has a global minimum there.
34. (a) The function f is continuous in the region R, but R is not closed and bounded so a special analysis is required.
Notice that f(x, y) tends to as (x, y) tends farther and farther from the origin or tends toward any point on
the x or y axis. This suggests that a minimum for f, if it exists, can not be too far from the origin or too close to the
axes. For example, if x > 10 then f(x, y) > 4x > 40, and if y > 10 then f(x, y) > 5y > 50. If 0 < x < 0.1 then
f(x, y) > 2/x > 20, and if 0 < y < 0.1 then f(x, y) > 3/y > 30.
Since f(1, 1) = 14, a global minimum for f if it exists must be in the smaller region R

: 0.1 x 10,
0.1 y 10. The region R

is closed and bounded and so f does have a minimum value at some point in R

, and
since that value is at most 14, it is also a global minimum for all of R.
(b) Since the region R has no boundary, the minimum value must occur at a critical point of f. At a critical point we
have
fx =
2
x
2
+ 4 = 0 fy =
3
y
2
+ 5 = 0.
The only critical point is (
_
1/2,
_
3/5) (0.7071, 0.7746), at which f achieves the minimum value
f(
_
1/2,
_
3/5) = 4

2 + 2

15 13.403.
35. (a) By the chain rule applied to M(a) = f(h(a), a) we have M

(a) = fx(h(a), a)h

(a) + fa(h(a), a)da/da =


fx(h(a), a)h

(a) + fa(h(a), a). Since x = h(a) is a critical point of g(x) = f(x, a), we have g

(h(a)) =
fx(h(a), a) = 0. Thus M

(a) = fa(h(a), a).


(b) Let n(a) = f(x, a) be a cross-section of f with x xed. We have n

(a) = fa(x, a). If x = h(a), then n(a) =


f(h(a), a) = M(a) so the point (a, M(a)) is on the graph of n. And n

(a) = fa(h(a), a) = M

(a) so n and M
have the same slope at the point (a, h(a)). Their graphs are tangent.
(c) The quadratic function g(x) = f(x, a) = (1/2)x
2
+ 2ax a
2
has a maximum where g

(x) = x + 2a = 0 or
x = 2a. Thus h(a) = 2a. The maximum value is M(a) = g(h(a)) = f(2a, a) = a
2
. The graph in Figure 15.13
shows that the graph of M(a) is the upper boundary, called the upper envelope, of the family of cross-sections of f
with x xed.
1060 Chapter Fifteen /SOLUTIONS
M(a)
f(2, a)
f(1, a)
f(0, a) f(1, a)
f(2, a)
f(3, a)
x
y
Figure 15.13
Solutions for Section 15.3
Exercises
1. Our objective function is f(x, y) = x +y and our equation of constraint is g(x, y) = x
2
+y
2
= 1. To optimize f(x, y)
with Lagrange multipliers, we solve f(x, y) = g(x, y) subject to g(x, y) = 1. The gradients of f and g are
f(x, y) =

i +

j ,
g(x, y) = 2x

i + 2y

j .
So the equation f = g becomes

i +

j = (2x

i + 2y

j )
Solving for gives
=
1
2x
=
1
2y
,
which tells us that x = y. Going back to our equation of constraint, we use the substitution x = y to solve for y:
g(y, y) = y
2
+y
2
= 1
2y
2
= 1
y
2
=
1
2
y =
_
1
2
=

2
2
.
Since x = y, our critical points are (

2
2
,

2
2
) and (

2
2
,

2
2
). Since the constraint is closed and bounded, maximum
and minimum values of f subject to the constraint exist. Evaluating f at the critical points we nd that the maximum
value is f(

2
2
,

2
2
) =

2 and the minimum value is f(

2
2
,

2
2
) =

2.
2. Our objective function is f(x, y) = 3x 2y and our equation of constraint is g(x, y) = x
2
+ 2y
2
= 44. Their gradients
are
f(x, y) = 3

i 2

j ,
g(x, y) = 2x

i + 4y

j .
15.3 SOLUTIONS 1061
So the equation f = g becomes 3

i 2

j = (2x

i + 4y

j ). Solving for gives us


=
3
2x
=
2
4y
,
which we can use to nd x in terms of y:
3
2x
=
2
4y
4x = 12y
x = 3y.
Using this relation in our equation of constraint, we can solve for y:
x
2
+ 2y
2
= 44
(3y)
2
+ 2y
2
= 44
9y
2
+ 2y
2
= 44
11y
2
= 44
y
2
= 4
y = 2.
Thus, the critical points are (6, 2) and (6, 2). Since the constraint is closed and bounded, maximum and minimum
values of f subject to the constraint exist. Evaluating f at the critical points, we nd that the maximum is f(6, 2) =
18 + 4 = 22 and the minimum value is f(6, 2) = 18 4 = 22.
3. Our objective function is f(x, y) = xy and our equation of constraint is g(x, y) = 4x
2
+y
2
= 8. Their gradients are
f(x, y) = y

i +x

j ,
g(x, y) = 8x

i + 2y

j .
So the equation f = g becomes y

i +x

j = (8x

i + 2y

j ). This gives
8x = y and 2y = x.
Multiplying, we get
8x
2
= 2y
2
.
If = 0, then x = y = 0, which does not satisfy the constraint equation. So = 0 and we get
2y
2
= 8x
2
y
2
= 4x
2
y = 2x.
To nd x, we substitute for y in our equation of constraint.
4x
2
+y
2
= 8
4x
2
+ 4x
2
= 8
x
2
= 1
x = 1
So our critical points are (1, 2), (1, 2), (1, 2) and (1, 2). Since the constraint is closed and bounded, maximum
and minimum values of f subject to the constraint exist. Evaluating f(x, y) at the critical points, we have
f(1, 2) = f(1, 2) = 2
f(1, 2) = f(1, 2) = 2.
Thus, the maximum value of f on g(x, y) = 8 is 2, and the minimum value is 2.
1062 Chapter Fifteen /SOLUTIONS
4. The objective function is f(x, y) = x
2
+y
2
and the equation of constraint is g(x, y) = x
4
+y
4
= 2. Their gradients are
f(x, y) = 2x

i + 2y

j ,
g(x, y) = 4x
3

i + 4y
3

j .
So the equation f = g becomes 2x

i + 2y

j = (4x
3

i + 4y
3

j ). This tells us that


2x = 4x
3
,
2y = 4y
3
.
Now if x = 0, the rst equation is true for any value of . In particular, we can choose which satises the second
equation. Similarly, y = 0 is solution.
Assuming both x = 0 and y = 0, we can divide to solve for and nd
=
2x
4x
3
=
2y
4y
3
1
2x
2
=
1
2y
2
y
2
= x
2
y = x.
Going back to our equation of constraint, we nd
g(0, y) = 0
4
+y
4
= 2, so y =
4

2
g(x, 0) = x
4
+ 0
4
= 2, so x =
4

2
g(x, x) = x
4
+ (x)
4
= 2, so x = 1.
Thus, the critical points are (0,
4

2), (
4

2, 0), (1, 1) and (1, 1). Since the constraint is closed and bounded,
maximum and minimum values of f subject to the constraint exist. Evaluating f at the critical points, we nd
f(1, 1) = f(1, 1) = f(1, 1) = f(1, 1) = 2,
f(0,
4

2) = f(0,
4

2) = f(
4

2, 0) = f(
4

2, 0) =

2.
Thus, the minimum value of f(x, y) on g(x, y) = 2 is

2 and the maximum value is 2.


5. The objective function is f(x, y) = x
2
+ y
2
and the constraint equation is g(x, y) = 4x 2y = 15, so grad f =
(2x)

i + (2y)

j and grad g = 4

i 2

j . Setting grad f = grad g gives


2x = 4,
2y = 2.
From the rst equation we have = x/2, and from the second equation we have = y. Setting these equal gives
y = 0.5x.
Substituting this into the constraint equation 4x 2y = 15 gives x = 3. The only critical point is (3, 1.5).
We have f(3, 1.5) = (3)
2
+ (1.5)
2
= 11.25. One way to determine if this point gives a maximum or minimum
value or neither for the given constraint is to examine the contour diagram of f with the constraint sketched in, Fig-
ure 15.14. It appears that moving away from the point P = (3, 1.5) in either direction along the constraint increases the
value of f, so (3, 1.5) is a point of minimum value.
15.3 SOLUTIONS 1063
1 2 3 4
5 10 15 20
3
2
1
1
Constraint: 4x 2y = 15
P = (3, 1.5)
x
y
Figure 15.14
2 2
2
2
(

5/2, 1/2) (

5/2, 1/2)
0
0.75
1.5
2.25
x
2
y
2
= 1
x
y
Figure 15.15: Graph of x
2
y
2
= 1
6. Our objective function is f(x, y) = x
2
+y and our equation of constraint is g(x, y) = x
2
y
2
= 1. Their gradients are
f(x, y) = 2x

i +

j ,
g(x, y) = 2x

i 2y

j .
Thus f = g gives
2x = 2x
1 = 2y
But x cannot be zero, since the constraint equation, y
2
= 1, would then have no real solution for y. So the equation
f = g becomes
=
2x
2x
=
1
2y
1 =
1
2y
2y = 1
y =
1
2
.
Substituting this into our equation of constraint we nd
g(x,
1
2
) = x
2

1
2
_
2
= 1
x
2
=
5
4
x =

5
2
.
So the critical points are (

5
2
,
1
2
) and (

5
2
,
1
2
). Evaluating f at these points we nd f(

5
2
,
1
2
) = f(

5
2
,
1
2
) =
5
4

1
2
=
3
4
. This is the minimum value for f(x, y) constrained to g(x, y) = 1. To see this, note that for x
2
= y
2
+ 1,
f(x, y) = y
2
+ 1 + y = (y + 1/2)
2
+ 3/4 3/4. Alternatively, see Figure 15.15. To see that f has no maximum on
g(x, y) = 1, note that f as x and y on the part of the graph of g(x, y) = 1 in quadrant I.
7. The objective function is f(x, y) = x
2
xy +y
2
and the equation of constraint is g(x, y) = x
2
y
2
= 1. The gradients
of f and g are
f(x, y) = (2x y)

i + (x + 2y)

j ,
g(x, y) = 2x

i 2y

j .
Therefore the equation f(x, y) = g(x, y) gives
2x y = 2x
x + 2y = 2y
x
2
y
2
= 1.
1064 Chapter Fifteen /SOLUTIONS
Let us suppose that = 0. Then 2x = y and 2y = x give x = y = 0. But (0, 0) is not a solution of the third equation,
so we conclude that = 0. Now lets multiply the rst two equations
2y(2x y) = 2x(x + 2y).
As = 0, we can cancel it in the equation above and after doing the algebra we get
x
2
4xy +y
2
= 0
which gives x = (2 +

3)y or x = (2

3)y.
If x = (2 +

3)y, the third equation gives


(2 +

3)
2
y
2
y
2
= 1
so y 0.278 and x 1.038. These give the critical points (1.038, 0.278), (1.038, 0.278).
If x = (2

3)y, from the third equation we get


(2

3)
2
y
2
y
2
= 1.
But (2

3)
2
1 0.928 < 0 so the equation has no solution. Evaluating f gives
f(1.038, 0.278) = f(1.038, 0.278) 0.866
Since y on the constraint, rewriting f as
f(x, y) =
_
x
y
2
_
2
+
3
4
y
2
shows that f has no maximum on the constraint. The minimum value of f is 0.866. See Figure 15.16.
2
2
(1.038, 0.278)
(1.038, 0.278)
0.2
0.866
2
x
x
2
y
2
= 1
y
Figure 15.16
8. The objective function is f(x, y, z) = x + 3y + 5z and the equation of constraint is g(x, y, z) = x
2
+ y
2
+ z
2
= 1.
Their gradients are
f(x, y, z) =

i + 3

j + 5

k ,
g(x, y, z) = 2x

i + 2y

j + 2z

k .
So the equation f = g becomes

i + 3

j + 5

k = (2x

i + 2y

j + 2z

k ). Solving for we nd
=
1
2x
=
3
2y
=
5
2z
.
Which provides us with the equations
2y = 6x
10x = 2z.
15.3 SOLUTIONS 1065
Solving the rst equation for y gives us y = 3x. Solving the second equation for z gives us z = 5x. Substituting these
into the equation of constraint, we can nd x:
x
2
+ (3x)
2
+ (5x)
2
= 1
x
2
+ 9x
2
+ 25x
2
= 1
35x
2
= 1
x
2
=
1
35
x =
_
1
35
=

35
35
.
Since y = 3x and z = 5x, the critical points are at (

35
35
, 3

35
35
,

35
7
). Since the constraint is closed and bounded, max-
imum and minimum values of f subject to the constraint exist. Evaluating f at the critical points, we nd the maximum
is f(

35
35
, 3

35
35
,

35
7
) =

35
35
35
=

35, and the minimum value is f(

35
35
, 3

35
35
,

35
7
) =

35.
9. Our objective function is f(x, y, z) = x
2
2y +2z
2
and our equation of constraint is g(x, y, z) = x
2
+y
2
+z
2
1 = 0.
To optimize f(x, y, z) with Lagrange multipliers, we solve f(x, y, z) = g(x, y, z) subject to g(x, y, z) = 0. The
gradients of f and g are
f(x, y, z) = 2x

i 2

j + 4z

k ,
g(x, y) = 2x

i + 2y

j + 2z

k .
We get,
x = x
1 = y
2z = z
x
2
+y
2
+z
2
= 1.
From the rst equation we get x = 0 or = 1.
If x = 0 we have
1 = y
2z = z
y
2
+z
2
= 1.
From the second equation z = 0 or = 2. So if z = 0, we have y = 1 and we get the solutions (0, 1, 0),(0, 1, 0). If
z = 0 then = 2 and y =
1
2
. So z
2
=
3
4
which gives the solutions (0,
1
2
,

3
2
), (0,
1
2
,

3
2
).
If x = 0, then = 1, so y = 1, which implies, from the equation x
2
+y
2
+z
2
= 1, that x = 0, which contradicts
the assumption.
Since the constraint is closed and bounded, maximum and minimum values of f subject to the constraint exist. There-
fore, evaluating f at the critical points, we get f(0, 1, 0) = 2, f(0, 1, 0) = 2 and f(0,
1
2
,

3
2
) = f(0,
1
2
,

3
2
) =
4. So the maximum value of f is 4 and the minimum is 2.
10. Our objective function is f(x, y, z) = 2x + y + 4z and our equation of constraint is g(x, y, z) = x
2
+ y + z
2
= 16.
Their gradients are
f(x, y, z) = 2

i + 1

j + 4

k ,
g(x, y, z) = 2x

i + 1

j + 2z

k .
So the equation f = g becomes 2

i + 1

j + 4

k = (2x

i + 1

j + 2z

k ). Solving for we nd
=
2
2x
=
1
1
=
4
2z
=
1
x
= 1 =
2
z
.
Which tells us that x = 1 and z = 2. Going back to our equation of constraint, we can solve for y.
g(1, y, 2) = 16
1
2
+y + 2
2
= 16
y = 11.
1066 Chapter Fifteen /SOLUTIONS
So our one critical point is at (1, 11, 2). The value of f at this point is f(1, 11, 2) = 2 + 11 + 8 = 21. This is the
maximum value of f(x, y, z) on g(x, y, z) = 16. To see this, note that for y = 16 x
2
z
2
,
f(x, y, z) = 2x + 16 x
2
z
2
+ 4z = 21 (x 1)
2
(z 2)
2
21.
As y , the point (

16 y, y, 0) is on the constraint and f(

16 y, y, 0) , so there is no minimum
value for f(x, y, z) on g(x, y, z) = 16.
11. The region x
2
+y
2
4 is the shaded disk of radius 2 centered at the origin (including the circle x
2
+y
2
= 4) shown in
Figure 15.17.
We will rst nd the local maxima and minima in the interior of the disk. So we need to nd the extrema of
f(x, y) = x
2
+ 2y
2
in the region x
2
+y
2
< 4.
For this we compute the critical points:
fx = 2x = 0
fy = 4y = 0
So the critical point is (0, 0). As fxx(0, 0) = 2, fyy(0, 0) = 4 and fxy(0, 0) = 0 we have
D = fxx(0, 0) fyy(0, 0) (fxy(0, 0))
2
= 8 > 0 and fxx(0, 0) = 2 > 0.
Therefore (0, 0) is a minimum point and f(0, 0) = 0.
Now lets nd the local extrema of f on the boundary of the disk, hence this time we have to solve a constraint
problem. We want the extrema of f(x, y) = x
2
+ 2y
2
subject to g(x, y) = x
2
+ y
2
4 = 0. We use Lagrange
multipliers:
grad f = grad g and x
2
+y
2
= 4,
which give
2x = 2x
4y = 2y
x
2
+y
2
= 4.
From the rst equation we have x = 0 or = 1. If x = 0, from the last equation y
2
= 4 and therefore (0, 2) and
(0, 2) are solutions.
If x = 0 then = 1 and from the second equation y = 0. Substituting this into the third equation we get x
2
= 4 so
(2, 0) and (2, 0) are the other two solutions.
The region x
2
+ y
2
4 is closed and bounded, so maximum and minimum values of f in the the region exist.
Therefore, as f(0, 2) = f(0, 2) = 8 and f(2, 0) = f(2, 0) = 4, (0, 2) and (0, 2) are global maxima and (0, 0) is
the global minimum on the whole region. The maximum value of f is 8 and the minimum value of f is 0.
4 4
4
4
1
4
8
16
x
y
Figure 15.17

2
1
3
5
1
3
5
x
y
Figure 15.18
15.3 SOLUTIONS 1067
12. The region x
2
+y
2
2 is the shaded disk of radius

2 centered at the origin (including the circle x


2
+y
2
= 2) as shown
in Figure 15.18.
We rst nd the local maxima and minima of f in the interior of our disk. So we need to nd the extrema of
f(x, y) = x + 3y, in the region x
2
+y
2
< 2.
As
fx = 1
fy = 3
f does not have critical points. Now lets nd the local extrema of f on the boundary of the disk. We want to nd the
extrema of f(x, y) = x + 3y subject to the constraint g(x, y) = x
2
+y
2
2 = 0. We use Lagrange multipliers
grad f = grad g and x
2
+y
2
= 2,
which give
1 = 2x
3 = 2y
x
2
+y
2
= 2.
As cannot be zero, we solve for x and y in the rst two equations and get x =
1
2
and y =
3
2
. Plugging into the third
equation gives
8
2
= 10
so =

5
2
and we get the solutions (
1

5
,
3

5
) and (
1

5
,
3

5
). Evaluating f at these points gives
f(
1

5
,
3

5
) = 2

5 and
f(
1

5
,
3

5
) = 2

5
The region x
2
+ y
2
2 is closed and bounded, so maximum and minimum values of f in the region exist. Therefore
(
1

5
,
3

5
) is a global maximum of f and (
1

5
,
3

5
) is a global minimum of f on the whole region x
2
+y
2
2.
13. The domain x
2
+2y
2
1 is the shaded interior of the ellipse x
2
+2y
2
= 1 including the boundary, shown in Figure 15.19.
1 1

2
0.1
0.3
0.5
0.1
0.3
0.5
x
y
Figure 15.19
First we want to nd the local maxima and minima of f in the interior of the ellipse. So we need to nd the extrema
of
f(x, y) = xy, in the region x
2
+ 2y
2
< 1.
For this we compute the critical points:
fx = y = 0 and fy = x = 0.
So there is one critical point, (0, 0). As fxx(0, 0) = 0, fyy(0, 0) = 0 and fxy(0, 0) = 1 we have
D = fxx(0, 0) fyy(0, 0) (fxy(0, 0))
2
= 1 < 0
1068 Chapter Fifteen /SOLUTIONS
so (0, 0) is a saddle and f does not have local extrema in the interior of the ellipse.
Now lets nd the local extrema of f on the boundary, hence this time well have a constraint problem. We want the
extrema of f(x, y) = xy subject to g(x, y) = x
2
+ 2y
2
1 = 0. We use Lagrange multipliers:
grad f = grad g and x
2
+ 2y
2
= 1
which give
y = 2x
x = 4y
x
2
+ 2y
2
= 1
From the rst two equations we get
xy = 8
2
xy.
So x = 0 or y = 0 or 8
2
= 1.
If x = 0, from the last equation 2y
2
= 1 so y =

2
2
and we get the solutions (0,

2
2
) and (0,

2
2
).
If y = 0, from the last equation we get x
2
= 1 and so the solutions are (1, 0) and (1, 0).
If x = 0 and y = 0 then 8
2
= 1, hence =
1
2

2
. For =
1
2

2
x =

2y
and plugging into the third equation gives 4y
2
= 1 so we get the solutions (

2
2
,
1
2
) and (

2
2
,
1
2
).
For =
1
2

2
we get
x =

2y
and plugging into the third equation gives 4y
2
= 1, and the solutions (

2
2
,
1
2
) and (

2
2
,
1
2
). So nally we have the
solutions: (1, 0), (1, 0), (

2
2
,
1
2
), (

2
2
,
1
2
), (

2
2
,
1
2
), (

2
2
,
1
2
).
Evaluating f at these points gives:
f(0,

2
2
) = f(0,

2
2
) = f(1, 0) = f(1, 0) = 0
f(

2
2
,
1
2
) = f(

2
2
,
1
2
) =

2
4
f(

2
2
,
1
2
) = f(

2
2
,
1
2
) =

2
4
.
The region x
2
+ 2y
2
1 is closed and bounded, so the maximum and minimum values of f in the region exist. Hence
the maximum value of f is

2
4
and the minimum value of f is

2
4
.
14. The region x
2
+y
2
1 is the shaded disk of radius 1 centered at the origin (including the circle x
2
+y
2
= 1) shown in
Figure 15.20.
Lets rst compute the critical points of f in the interior of the disk. We have
fx = 3x
2
= 0
fy = 2y = 0,
whose solution is x = y = 0. So the only one critical point is (0, 0). As fxx(0, 0) = 0, fyy(0, 0) = 2 and fxy(0, 0) =
0,
D = fxx(0, 0) fyy(0, 0) (fxy(0, 0))
2
= 0
which does not tell us anything about the nature of the critical point (0, 0).
But, if we choose x,y very small in absolute value and such that x
3
> y
2
, then f(x, y) > 0. If we choose x,y very
small in absolute value and such that x
3
< y
2
, then f(x, y) < 0. As f(0, 0) = 0, we conclude that (0, 0) is a saddle
point.
We can get the same conclusion looking at the level curves of f around (0, 0), as shown in Figure 15.21.
So, f does not have extrema in the interior of the disk.
Now, lets nd the local extrema of f on the circle x
2
+ y
2
= 1. So we want the extrema of f(x, y) = x
3
y
2
subject to the constraint g(x, y) = x
2
+y
2
1 = 0. Using Lagrange multipliers we get
grad f = grad g and x
2
+y
2
= 1,
15.3 SOLUTIONS 1069
which gives
3x
2
= 2x
2y = 2y
x
2
+y
2
= 1.
From the second equation y = 0 or = 1.
If y = 0, from the third equation we get x
2
= 1, which gives the solutions (1, 0), (1, 0).
If y = 0 then = 1 and from the rst equation we get 3x
2
= 2x, hence x = 0 or x =
2
3
. If x = 0, from the
third equation we get y
2
= 1, so the solutions (0, 1),(0, 1). If x =
2
3
, from the third equation we get y
2
=
5
9
, so the
solutions (
2
3
,

5
3
), (
2
3
,

5
3
).
Evaluating f at these points we get
f(1, 0) = 1, f(1, 0) = f(0, 1) = f(0, 1) = 1
and
f
_

2
3
,

5
3
_
= f
_

2
3
,

5
3
_
=
23
27
.
The region x
2
+y
2
1 is closed and bounded, so maximum and minimum values of f in the region exist. Therefore
the maximum value of f is 1 and the minimum value is 1.
1
1
x
y
Figure 15.20
1 1
1
1
0 0.3 0.7 0.1
0.3
0.7
x
y
Figure 15.21: Level curves of f
15. The region x +y 1 is the shaded half plane (including the line x +y = 1) shown in Figure 15.22.
1 2 3
1
2
3
1
3
5
1
3
5
x
y
Figure 15.22
Lets look for the critical points of f in the interior of the region. As
fx = 3x
2
fy = 1
1070 Chapter Fifteen /SOLUTIONS
there are no critical points inside the shaded region. Now lets nd the extrema of f on the boundary of our region. We
want the extrema of f(x, y) = x
3
+y subject to the constraint g(x, y) = x +y 1 = 0. We use Lagrange multipliers
grad f = grad g and x +y = 1,
which give
3x
2
=
1 =
x +y = 1.
From the rst two equations we get 3x
2
= 1, so the solutions are
(
1

3
, 1
1

3
) and (
1

3
, 1 +
1

3
).
Evaluating f at these points we get
f(
1

3
, 1
1

3
) = 1
2
3

3
f(
1

3
, 1 +
1

3
) = 1 +
2
3

3
.
From the contour diagram in Figure 15.22, we see that (
1

3
, 1
1

3
) is a local minimum and (
1

3
, 1 +
1

3
) is a local
maximum of f on x +y = 1. Are they global extrema as well?
If we take x very big and y = 1 x then f(x, y) = x
3
+y = x
3
x + 1 which can be made as big as we want (if
we choose x big enough). So there will be no global maximum.
Similarly, taking x negative with big absolute value and y = 1 x, f(x, y) = x
3
+y = x
3
x +1 can be made as
small as we want (if we choose x small enough). So there is no global minimum. This can also be seen from Figure 15.22.
Problems
16. We know that a maximum or minimum value of f subject to the constraint equation g(x, y) = c occurs where grad f
is parallel to grad g, or at the endpoints of the constraint. The vectors grad f and grad g are parallel where the graph of
g(x, y) = c is tangent to the contours of f, which occurs at approximately x = 6 and y = 6. At the point (6, 6), we
have f = 400. The graph of g(x, y) = c crosses the contours f = 300, f = 200, f = 100 but does not cross any
contours with f-values greater than 400. We see that the maximum of f subject to the constraint is 400 at the point (6, 6).
It appears that f takes on its minimum value (less than 100) at one of the endpoints, which are approximately (10.5, 0)
and (0, 13.5).
17. (a) The contour for z = 1 is the line 1 = 2x + y, or y = 2x + 1. The contour for z = 3 is the line 3 = 2x + y, or
y = 2x + 3. The contours are all lines with slope 2. See Figure 15.23.
1 2 3 4 5 1 2 3 4 5
1
2
3
4
5
1
2
3
4
5
7
5
3
1
1
3
5
7
x
y
Figure 15.23
1 2 3 4 5 1 2 3 4 5
1
2
3
4
5
1
2
3
4
5
7
5
3
1
1
3
5
7
x
y
Figure 15.24
15.3 SOLUTIONS 1071
(b) The graph of x
2
+y
2
= 5 is a circle of radius

5 = 2.236 centered at the origin. See Figure 15.24.


(c) The circle representing the constraint equation in Figure 15.24 appears to be tangent to the contour close to z = 5 at
the point (2, 1), and this is the contour with the highest z-value that the circle intersects. The circle is tangent to the
contour z = 5 approximately at the point (2, 1), and this is the contour with the lowest z-value that the circle
intersects. Therefore, subject to the constraint x
2
+ y
2
= 5, the function f has a maximum value of about 5 at the
point (2, 1) and a minimum value of about 5 at the point (2, 1).
Since the radius vector, 2

i +

j , at the point (2, 1) is perpendicular to the line 2x + y = 5, the maximum is


exactly 5 and occurs at (2, 1). Similarly, the minimum is exactly 5 and occurs at (2, 1).
(d) The objective function is f(x, y) = 2x + y and the constraint equation is g(x, y) = x
2
+ y
2
= 5, and so grad f =
2

i +

j and grad g = (2x)

i + (2y)

j . Setting grad f = grad g gives


2 = (2x),
1 = (2y).
On the constraint, x = 0 and y = 0. Thus, from the rst equation, we have = 1/x, and from the second equation
we have = 1/(2y). Setting these equal gives
x = 2y.
Substituting this into the constraint equation x
2
+y
2
= 5 gives (2y)
2
+y
2
= 5 so y = 1 and y = 1. Since x = 2y,
the maximum or minimum values occur at (2, 1) or (2, 1). Since f(2, 1) = 5 and f(2, 1) = 5, the function
f(x, y) = 2x+y subject to the constraint x
2
+y
2
= 5 has a maximum value of 5 at the point (2, 1) and a minimum
value of 5 at the point (2, 1). This conrms algebraically what we observed graphically in part (c).
18. We want to minimize
C = f(q1, q2) = 2q
2
1
+q1q2 +q
2
2
+ 500
subject to the constraint q1 +q2 = 200 or g(q1, q2) = q1 +q2 200 = 0.
Since f = (4q1 +q2)

i + (2q2 +q1)

j and g =

i +

j , f = g gives
4q1 +q2 =
2q2 +q1 = .
Solving we get
4q1 +q2 = 2q2 +q1
so
3q1 = q2.
We want
q1 +q2 = 200
q1 + 3q1 = 4q1 = 200.
Therefore
q1 = 50 units, q2 = 150 units.
19.
r
`

Figure 15.25
Let V be the volume and S be the surface area of the container. Then
V = r
2
h and S = 2rh + 2r
2
1072 Chapter Fifteen /SOLUTIONS
where h is the height and r is the radius as shown in Figure 15.25. We have V = 100 cm
3
as our constraint. Since
S = (2h + 4r)

i + 2r

j = ((2h + 4r)

i + 2r

j )
and V = 2rh

i +r
2

j = (2rh

i +r
2

j ),
at the optimum
S = V, we have
((2h + 4r)

i + 2r

j ) = (2rh

i +r
2

j ),
that is 2h + 4r = 2rh and 2r = r
2
, hence =
2
r
.
We assume r = 0 or else we have a very awkward cylinder. Then, plug = 2/r into the rst equation to obtain:
2h + 4r = 2
_
2
r
_
rh
2h + 4r = 4h
h = 2r.
Finally, solve for r and h using the constraint:
V = r
2
h = 100
r
2
(2r) = 100
r
3
=
50

r =
3
_
50

.
Solving for h, we obtain h = 2r = 2
3
_
50

.
20. Constraint is G = P1x +P2y K = 0.
Since Q = G, we have
cax
a1
y
b
= P1 and cbx
a
y
b1
= P2.
Dividing the two equations yields
cax
a1
y
b
cbx
a
y
b1
=
P1
P2
, or simplifying,
ay
bx
=
P1
P2
. Hence, y =
bP1
aP2
x.
Substitute into the constraint to obtain P1x +P2
bP1
aP2
x = P1
_
a +b
a
_
x = K, giving
x =
aK
(a +b)P1
and y =
bK
(a +b)P2
.
We now check that this is indeed the maximization point. Since x, y 0, possible maximization points are (0,
K
P2
),
(
K
P1
, 0), and (
aK
(a +b)P1
,
bK
(a +b)P2
). Since Q = 0 for the rst two points and Q is positive for the last point, it follows
that (
aK
(a +b)P1
,
bK
(a +b)P2
) gives the maximal value.
21. (a) The company wishes to maximize P(x, y) given the constraint C(x, y) = 50, 000. The objective function is P(x, y)
and the constraint equation is C(x, y) = 50, 000. The Lagrange multiplier is approximately equal to the change in
P(x, y) given a one unit increase in the budget constraint. In other words, if we increase the budget by $1, we can
produce about more units of the good.
(b) The company wishes to minimize C(x, y) given the constraint equation P(x, y) = 2000. The objective function is
C(x, y) and the constraint equation is P(x, y) = 2000. The Lagrange multiplier is approximately equal to the
change in C(x, y) given a one unit increase in the production constraint. In other words, it costs about dollars to
produce one more unit of the good.
15.3 SOLUTIONS 1073
22. The company wants to maximize f(x, y) = 500x
0.6
y
0.3
given the constraint g(x, y) = 10x + 25y = 2000. Setting
grad f = grad g gives
500(0.6x
0.4
)y
0.3
= 10,
500x
0.6
(0.3y
0.7
) = 25.
From the rst equation we have = 30y
0.3
/x
0.4
, and from the second equation we have = 6x
0.6
/y
0.7
. Setting these
equal gives
y = 0.2x.
Substituting this into the constraint equation 10x + 25y = 2000 gives x = 133.33. Since y = 0.2x, the maximum value
occurs at x = 133.33 and y = 26.67.
(a) The company should purchase 133.33 units of chemical X and 26.67 units of chemical Y. With these purchases, the
company will be able to produce f(133.33, 26.67) = 25, 219 units of chemical Z.
(b) When x = 133.33 and y = 26.67, we see that = 11.348. If $1 is added to the budget, the company will be able to
produce about 11.348 additional units of chemical Z.
23. (a) Let c be the cost of producing the product. Then c = 10W + 20K = 3000. At optimum production,
q = c.
q =
_
9
2
W

1
4
K
1
4
_

i +
_
3
2
W
3
4
K

3
4
_

j , and c = 10

i + 20

j . Equating we get
9
2
W

1
4
K
1
4
= 10, and
3
2
W
3
4
K

3
4
= 20.
Dividing yields K =
1
6
W, so substituting into c gives
10W + 20
_
1
6
W
_
=
40
3
W = 3000.
Thus W = 225 and K = 37.5. Substituting both answers to nd gives
=
9
2
(225)

1
4
(37.5)
1
4
10
= 0.2875.
We also nd the optimum quantity produced, q = 6(225)
3
4
(37.5)
1
4
= 862.57.
(b) At the optimum values found above, marginal productivity of labor is given by
q
W

(225,37.5)
=
9
2
W

1
4
K
1
4

(225,37.5)
= 2.875,
and marginal productivity of capital is given by
q
K

(225,37.5)
=
3
2
W
3
4
K

3
4

(225,37.5)
= 5.750.
The ratio of marginal productivity of labor to that of capital is
q
W
q
K
=
1
2
=
10
20
=
cost of a unit of L
cost of a unit of K
.
(c) When the budget is increased by one dollar, we substitute the relation K1 =
1
6
W1 into 10W1 +20K1 = 3001 which
gives 10W1 + 20(
1
6
W1) =
40
3
W1 = 3001. Solving yields W1 = 225.075 and K1 = 37.513, so q1 = 862.86 =
q + 0.29. Thus production has increased by 0.29 , the Lagrange multiplier.
24. (a) The problem is to maximize
V = 1000D
0.6
N
0.3
subject to the budget constraint in dollars
40000D + 10000N 600000
or (in thousand dollars)
40D + 10N 600
1074 Chapter Fifteen /SOLUTIONS
(b) Let B = 40D + 10N = 600 (thousand dollars) be the budget constraint. At the optimum
V = B,
so
V
D
=
B
D
= 40
V
N
=
B
N
= 10.
Thus
V
D
V
N
= 4.
Therefore, at the optimum point, the rate of increase in the number of visits to the number of doctors is four times the
corresponding rate for nurses. This factor of four is the same as the ratio of the salaries.
(c) Differentiating and setting V = B yields
600D
0.4
N
0.3
= 40
300D
0.6
N
0.7
= 10
Thus, we get
600D
0.4
N
0.3
40
= =
300D
0.6
N
0.7
10
So
N = 2D.
To solve for D and N, substitute in the budget constraint:
600 40D 10N = 0
600 40D 10 (2D) = 0
So D = 10 and N = 20.
=
600(10
0.4
)(20
0.3
)
40
14.67
Thus the clinic should hire 10 doctors and 20 nurses. With that staff, the clinic can provide
V = 1000(10
0.6
)(20
0.3
) 9,779 visits per year.
(d) From part c), the Lagrange multiplier is = 14.67. At the optimum, the Lagrange multiplier tells us that about 14.67
extra visits can be generated through an increase of $1,000 in the budget. (If we had written out the constraint in
dollars instead of thousands of dollars, the Lagrange multiplier would tell us the number of extra visits per dollar.)
(e) The marginal cost, MC, is the cost of an additional visit. Thus, at the optimum point, we need the reciprocal of the
Lagrange multiplier:
MC =
1


1
14.67
0.068 (thousand dollars)
i.e. at the optimum point, an extra visit costs the clinic 0.068 thousand dollars, or $68.
This production function exhibits declining returns to scale (e.g. doubling both inputs less than doubles output,
because the two exponents add up to less than one). This means that for large V , increasing V will require increasing
D and N by more than when V is small. Thus the cost of an additional visit is greater for large V than for small. In
other words, the marginal cost will rise with the number of visits.
25. (a) The solution to Problem 23 gives = 0.29. We recalculate with a budget of $4000.
The condition that grad q = grad(budget) in Problem 23 gives
9
2
W
1/4
K
1/4
= (10) and
3
2
W
3/4
K
3/4
= (20),
so K =
1
6
W. Substituting into the budget constraint after replacing the budget of $3000 by $4000 gives
10W + 20(
1
6
W) =
40
3
W = 4000.
Thus, W = 300 and K = 50 and q = 1150.098.
Multiplying the rst equation by W and the second by K and adding gives
W(
9
2
W
1/4
K
1/4
) +K(
3
2
W
3/4
K
3/4
) = W(10) +K(20).
15.3 SOLUTIONS 1075
So
_
9
2
+
3
2
_
W
3/4
K
1/4
= (10W + 20K)
6W
3/4
K
1/4
= (4000)
Thus,
=
6W
3/4
K
1/4
4000
=
1150.098
4000
= 0.29
Thus, the value of remains unchanged.
(b) The solution to Problem 24 shows that = 14.67. We solve the problem again with a budget of $700,000.
The condition that grad V = grad B in Problem 24 gives
600D
0.4
N
0.3
= 40
300D
0.6
N
0.7
= 10
Thus, N = 2D. Substituting in the budget constraint after replacing the budget of 600 by 700 (the budget in measured
in thousands of dollars) gives
40D + 10(2D) = 700
so D = 11.667 and N = 23.337 and V = 11234.705. As in part a), we multiply the rst equation by D and the
second by N and add:
D(600D
0.4
N
0.3
) +N(300D
0.6
N
0.7
) = D(40) +N(10),
so
(600 + 300)D
0.6
N
0.3
= (400 + 10N)
900D
0.6
N
0.3
= (700)
Since V = 1000D
0.6
N
0.3
= 11234.705, we have
=
900D
0.6
N
0.3
700
=
9
7
(
V
1000
) = 14.44.
Thus, the value of has changed with the budget.
(c) We are interested in the marginal increase of production with budget (that is, the value of ) and whether it is affected
by the budget.
Suppose $B is the budget. In part (a) we found
=
6W
3/4
K
1/4
B
and in part (b) we found
=
900D
0.6
N
0.3
B
.
In part (a), both W and K are proportional to B. Thus, W = c1B and K = c2B, so
=
6(c1B)
3/4
(c2B)
1/4
B
=
6c
3/4
1
C
1/4
2
B
3/4
B
1/4
B
= 6c
3/4
1
c
1/4
2
.
So we see is independent of B.
In part (b), both D and N are proportional to B, so D = c3B and N = c4B. Thus,
=
900(c3B)
0.6
(c4B)
0.3
B
=
900c
0.6
3
C
0.3
4
B
0.6
B
0.3
B
= 900c
0.6
3
c
0.3
4
1
B
0.1
.
1076 Chapter Fifteen /SOLUTIONS
So we see is not independent of B.
The crucial difference is that the exponents in Problem 23 add to 1, that is 3/4+1/4 = 1, whereas the exponents
in Problem 24 do not add to 1, since 0.6 + 0.3 = 0.9.
Thus, the condition that must be satised by the Cobb-Douglas production function
Q = cK
a
L
b
to ensure that the value of is not affected by production is that
a +b = 1.
This is called constant returns to scale.
26. (a) The curves are shown in Figure 15.26.
20 40 60 80 100
500
1000
1500
I
II
III
(50, 500)
s = 1000 10l
l
s
Figure 15.26
(b) The income equals $10/hour times the number of hours of work:
s = 10(100 l) = 1000 10l.
(c) The graph of this constraint is the straight line in Figure 15.26.
(d) For any given salary, curve III allows for the most leisure time, curve I the least. Similarly, for any amount of leisure
time, curve III also has the greatest salary, and curve I the least. Thus, any point on curve III is preferable to any point
on curve II, which is preferable to any point on curve I. We prefer to be on the outermost curve that our constraint
allows. We want to choose the point on s = 1000 10l which is on the most preferable curve. Since all the curves
are concave up, this occurs at the point where s = 1000 10l is tangent to curve II. So we choose l = 50, s = 500,
and work 50 hours a week.
27. The maximum of f(x, y) = ax
2
+ bxy + cy
2
subject to the constraint g(x, y) = 1 where g(x, y) = x
2
+ y
2
occurs
where grad f = grad g. Since grad f = (2ax +by)

i + (bx + 2cy)

j and grad g = 2x

i + 2y

j we have
2ax +by = 2x
bx + 2cy = 2y
x
2
+y
2
= 1
Adding x times the rst equation to y times the second gives x(2ax +by) +y(bx + 2cy) = (2x
2
+ 2y
2
). Dividing by
2 and using the constraint equation gives f(x, y) = ax
2
+ bxy + cy
2
= (x
2
+ y
2
) = . This equation holds for all
solutions (x, y, ) of the three equations, including the solution that corresponds to the maximum value of f subject to
the constraint. Thus the maximum value is f(x, y) = .
28. (a) Let f(x1, x2, x3) =

3
i=1
xi
2
= x1
2
+ x2
2
+ x3
2
and g(x1, x2, x3) =

3
i=1
xi = 1. Then grad f = grad g
gives
2x1 = and 2x2 = and 2x3 = .
so
x1 = x2 = x3 =

2
.
15.3 SOLUTIONS 1077
Then x1 +x2 +x3 = 1 gives
3

2
= 1 so =
2
3
so x1 = x2 = x3 =
1
3
.
These values of x1, x2, x3 give the minimum (rather than maximum) because the value of f increases without bound
as x2, x2, x3 .
(b) A similar argument shows that

n
i=1
xi has its minimum value subject to

n
i=1
xi = 1 when
x1 = x2 = = xn =
1
n
.
29. (a) The gradient vectors, f, point inward around a local maximum. See the two points marked A in Figure 15.27.
(b) Some of the gradient vectors around a saddle are pointing inward toward the point; some are pointing outward away
from the point. See the point marked B in Figure 15.27.
(c) The critical points on g = 1 are at points where f is perpendicular to the curve g = 1. There are four of them, all
marked with a dot in Figure 15.27. Imagine the level surfaces of f sketched in everywhere perpendicular to f; the
maximum value of f is at the point marked C in Figure 15.27
(d) Again imagine level curves of f. The minimum value of f is at the point marked D.
A A
C
B
D
Figure 15.27
(e) At C, the maximum on g = 1, the vector g points outward (because it points toward g = 2), while f points
inward. The Lagrange multiplier, , is dened so that f = g, so must be negative.
30. We want to minimize the function h(x, y) subject to the constraint that
g(x, y) = x
2
+y
2
= 1,000
2
= 1,000,000.
Using the method of Lagrange multipliers, we obtain the following system of equations:
hx =
10x + 4y
10,000
= 2x,
hy =
4x + 4y
10,000
= 2y,
x
2
+y
2
= 1,000,000.
Multiplying the rst equation by y and the second by x we get
y(10x + 4y)
10,000
=
x(4x + 4y)
10,000
.
1078 Chapter Fifteen /SOLUTIONS
Hence:
2y
2
+ 3xy 2x
2
= (2y x)(y + 2x) = 0,
and so the climber either moves along the line x = 2y or y = 2x.
We must now choose one of these lines and the direction along that line which will lead to the point of minimum
height on the circle. To do this we nd the points of intersection of these lines with the circle x
2
+ y
2
= 1,000,000,
compute the corresponding heights, and then select the minimum point.
If x = 2y, the third equation gives
5y
2
= 1,000
2
,
so that y = 1,000/

5 447.21 and x = 894.43. The corresponding height is h(894.43, 447.21) = 2400 m.


If y = 2x, we nd that x = 447.21 and y = 894.43. The corresponding height is h(447.21, 894.43) =
2900 m. Therefore, she should travel along the line x = 2y, in either of the two possible directions.
31. The objective function is
f(x, y, z) =
_
(x a)
2
+ (y b)
2
+ (z c)
2
,
and the constraint is
g(x, y, z) = Ax +By +Cz +D = 0.
Partial derivatives of f and g are
fx =
1
2
2 (x a)
f(x, y, z)
=
x a
f(x, y, z)
,
fy =
1
2
2 (y b)
f(x, y, z)
=
y b
f(x, y, z)
,
fz =
1
2
2 (z c)
f(x, y, z)
=
z c
f(x, y, z)
,
gx = A, gy = B, and gz = C.
Using Lagrange multipliers, we need to solve the equations
grad f = grad g
where grad f = fx

i +fy

j +fz

k and grad g = gx

i +gy

j +gz

k . This gives a system of equations:


x a
f(x, y, z)
= A
y b
f(x, y, z)
= B
z c
f(x, y, z)
= C
Ax +By +Cz +D = 0.
Now
xa
A
=
yb
B
=
zc
C
= f(x, y, z) gives
x =
A
B
(y b) +a,
z =
C
B
(y b) +c,
Substitute into the constraint,
A
_
A
B
(y b) +a
_
+By +C
_
C
B
(y b) +c
_
+D = 0,
_
A
2
B
+B +
C
2
B
_
y =
A
2
B
b Aa +
C
2
B
b Cc D.
Hence
15.3 SOLUTIONS 1079
y =
(A
2
+C
2
)b B(Aa +Cc +D)
A
2
+B
2
+C
2
,
y b =
B(Aa +Bb +Cc +D)
A
2
+B
2
+C
2
x a =
A
B
(y b)
=
A(Aa +Bb +Cc +D)
A
2
+B
2
+C
2
z c =
C
B
(y b)
=
C(Aa +Bb +Cc +D)
A
2
+B
2
+C
2
Thus the minimum f(x, y, z) is
f(x, y, z) =
_
(x a)
2
+ (y b)
2
+ (z c)
2
=
_
_
A(Aa +Bb +Cc +D)
A
2
+B
2
+C
2
_
2
+
_
B(Aa +Bb +Cc +D)
A
2
+B
2
+C
2
_
2
+
_
C(Aa +Bb +Cc +D)
A
2
+B
2
+C
2
_
2

1/2
=
|Aa +Bb +Cc +D|

A
2
+B
2
+C
2
.
The geometric meaning is nding the shortest distance from a point (a, b, c) to the plane Ax +By +Cz +D = 0.
32. (a) The objective function is the energy loss, i
2
1
R1 +i
2
2
R2, and the constraint is i1 +i2 = I, where I is a constant. The
Lagrangian function is
L(i1, i2, ) = i
2
1
R1 +i
2
2
R2 (i1 +i2 I).
We look for solutions to the system of equations we get from grad L =

0 :
L
i1
= 2i1R1 = 0
L
i2
= 2i2R2 = 0
L

= (i1 +i2 I) = 0.
Combining
L
i1

L
i2
= 2(i1R1 i2R2) = 0 with
L

= 0 gives the two equation system


i1R1 i2R2 = 0
i1 +i2 = I.
Substituting i2 = I i1 into the rst equation leads to
i1 =
R2
R1 +R2
I
i2 =
R1
R1 +R2
I.
(b) Ohms Law states that across a resistor
Voltage = Current Resistance.
Since /2 = i1 R1 = i2 R2, the Lagrange multiplier equals twice the voltage across the resistors.
1080 Chapter Fifteen /SOLUTIONS
33. (a) We draw the level curves (parallel straight lines) of f(x, y) = ax + by + c. We can see that the level lines with the
maximum and minimum f-values which intersect with the disk are the level lines that are tangent to the boundary of
the disk. Therefore, the maximum and minimum occur at the boundary of the disk. See Figure 15.28.
f =max
f =min
f increases
`
Figure 15.28
f =max
f =min
f increases

Figure 15.29
f =max
f =min
f increases
`
Figure 15.30
(b) Similar to part (a), we see the level lines with the largest and smallest f-values which intersect with the rectangle must
pass the corner of the rectangle. So the maximum and minimum occur at the corners of rectangle. See Figure 15.29.
When the level curves are parallel to a pair of the sides, then the points on the sides are all maximum or minimum, as
shown below in Figure 15.30.
(c) The graph of f is a plane. The part of the graph lying above a disk R is either a at disk, in which case every point is
a maximum, or is a tilted ellipse, in which case you can see that the maximum will be on the edge. Similarly, the part
lying above a rectangle is either a rectangle or a tilted parallelogram, in which case the maximum will be at a corner.
34. The point P is the solution to the constraint optimization problem of maximizing the square of the distance function.
D = x
2
+y
2
+x
2
subject to the constraint
g(x, y, z) = f(x, y) z = 0.
(We take the square of the distance between the point (x, y, z) and the origin, which is
Distance =
_
(x 0)
2
+ (y 0)
2
+ (z 0)
2
=
_
x
2
+y
2
+z
2
,
because it makes the calculations easier.) Therefore, at point P, we have D = g, so D is parallel to g.
We know that g is perpendicular to the surface g(x, y, z) = 0; that is, perpendicular to the surface z = f(x, y).
Also
D = 2x

i + 2y

j + 2z

k .
15.3 SOLUTIONS 1081
At point P, whose position vector is p = a

i +b

j +c

k , we have
D = 2(a

i +b

j +c

k ) = 2 p .
Thus, p is parallel to D and therefore p is also perpendicular to the surface.
35. (a) The objective function f(x, y) = px +qy gives the cost to buy x units of input 1 at unit price p and y units of input
2 at unit price q.
The constraint g(x, y) = u tells us that we are only considering the cost of inputs x and y that can be used to
produce quantity u of the product.
Thus the number C(p, q, u) gives the minimum cost to the company of producing quantity u if the inputs it
needs have unit prices p and q.
(b) The Lagrangian function is
L(x, y, ) = px +qy (xy u).
We look for solutions to the system of equations we get from grad L =

0 :
L
x
= p y = 0
L
y
= q x = 0
L

= (xy u) = 0.
We see that = p/y = q/x so y = px/q. Substituting for y in the constraint xy = u leads to x =
_
qu/p,
y =
_
pu/q and =
_
pq/u. The minimum cost is thus
C(p, q, u) = p
_
qu
p
+q
_
pu
q
= 2

pqu.
36. (a) The objective function U(x, y) gives the utility to the consumer of x units of item 1 and y units of item 2.
Since px + qy gives the cost to buy x units of item 1 at unit price p and y units of item 2 at unit price q, the
constraint px + qy = I tells us that we are only considering the utility of inputs x and y that can be purchased with
budget I.
Thus the number V (p, q, I) gives the maximum utility the consumer can get with a budget of I if the two items
have unit prices p and q.
The indirect utility function tells how much utility the consumer can buy, depending on his budget and the prices
of the two items.
(b) The value of the Lagrange multiplier is the rate of change of the maximum utility V the consumer can get with
his budget as the budget increases. This means that for small changes I in the budget, smart buying will result in a
change V I in the utility to the consumer of his purchases.
(c) The Lagrangian function is
L(x, y, ) = xy (px +qy I).
We look for solutions to the system of equations we get from grad L =

0 :
L
x
= y p = 0
L
y
= x q = 0
L

= (px +qy I) = 0.
We see that = y/p = x/q so y = px/q. Substituting for y in the constraint px + qy = I leads to x = I/(2p),
y = I/(2q) and = I/(2pq). The maximum utility is thus
V (p, q, I) = U(
I
2p
,
I
2q
) =
I
2p

I
2q
=
I
2
4pq
.
The marginal utility of money is
(p, q, I) = =
I
2pq
.
1082 Chapter Fifteen /SOLUTIONS
37. (a) The critical points of h(x, y) occur where
hx(x, y) = 2x 2 = 0
hy(x, y) = 2y 4 = 0.
The only critical point is (x, y) = (, 2) and it gives a minimum value for h(x, y). That minimum value is m() =
h(, 2) =
2
+ (2)
2
(2 + 4(2) 15) = 5
2
+ 15.
(b) The maximum value of m() = 5
2
+ 15 occurs at a critical point, where m

() = 10 + 15 = 0. At this
point, = 1.5 and m() = 5 1.5
2
+ 15 1.5 = 11.25.
(c) We want to minimize f(x, y) = x
2
+ y
2
subject to the constraint g(x, y) = 15, where g(x, y) = 2x + 4y. The
Lagrangian function is L(x, y, ) = x
2
+y
2
(2x + 4y 15) so we solve the system of equations
L
x
= 2x 2 = 0
L
y
= 2y 4 = 0
L

= (2x + 4y 15) = 0.
The rst two equations give x = and y = 2. Substitution into the third equation gives 2 + 4(2) 15 = 0
or = 1.5. Thus x = 1.5 and y = 3. The minimum value of f(x, y) subject to the constraint is f(1.5, 3) =
1.5
3
+ 3
2
= 11.25.
(d) The two question have the same answer.
38. You should try to anticipate your opponents choice. After you choose a value , your opponent will use calculus to
nd the point (x, y) that maximizes the function f(x, y) = 10 x
2
y
2
2x (2x + 2y). At that point, we have
fx = 2x 2 2 = 0 and fy = 2y 2 = 0, so your opponent will choose x = 1 and y = . This gives a
value L(1, , ) = 10(1)
2
()
2
2(1)(2(1)+2()) = 11+2+2
2
which you
want to make as small as possible. You should choose to minimize the function h() = 11 + 2 + 2
2
. You choose
so that h

() = 2 + 4 = 0, or = 1/2. Your opponent then chooses (x, y) = (1 , ) = (1/2, 1/2), giving


a nal score of L(1/2, 1/2, 1/2) = 10.5. No choice of that you can make can force the value of L below 10.5. But
your choice of = 1/2 makes it impossible for your opponent to force the value of L above 10.5.
Solutions for Chapter 15 Review
Exercises
1. The critical points of f are obtained by solving fx = fy = 0, that is
fx(x, y) = 2y
2
2x = 0 and fy(x, y) = 4xy 4y = 0,
so
2(y
2
x) = 0 and 4y(x 1) = 0
The second equation gives either y = 0 or x = 1. If y = 0 then x = 0 by the rst equation, so (0, 0) is a critical point. If
x = 1 then y
2
= 1 from which y = 1 or y = 1, so two further critical points are (1, 1), and (1, 1).
Since
D = fxxfyy (fxy)
2
= (2)(4x 4) (4y)
2
= 8 8x 16y
2
,
we have
D(0, 0) = 8 > 0, D(1, 1) = D(1, 1) = 16 < 0,
and fxx = 2 < 0. Thus, (0, 0) is a local maximum; (1, 1) and (1, 1) are saddle points.
2. At a critical point
fx(x, y) = 2xy 2y = 0
fy(x, y) = x
2
+ 4y 2x = 0.
SOLUTIONS to Review Problems for Chapter Fifteen 1083
From the rst equation, 2y(x 1) = 0, so either y = 0 or x = 1. If y = 0, then x
2
2x = 0, so x = 0 or x = 2. Thus
(0, 0) and (2, 0) are critical points. If x = 1, then 1
2
+ 4y 2 = 0, so y = 1/4. Thus (1, 1/4) is a critical point. Now
D = fxxfyy (fxy)
2
= 2y 4 (2x 2)
2
= 8y 4(x 1)
2
,
so
D(0, 0) = 4, D(2, 0) = 4, D(1,
1
4
) = 2
so (0, 0) and (2, 0) are saddle points. Since fyy = 4 > 0, we see that (1, 1/4) is a local minimum.
3. At a critical point
fx(x, y) = 6x
2
6xy + 12x = 0
fy(x, y) = 3x
2
12y = 0
From the second equation, we conclude that 3(x
2
+ 4y) = 0, so y =
1
4
x
2
. Substituting for y in the rst equation
gives
6x
2
6x
_

1
4
x
2
_
+ 12x = 0
or
x
2
+
1
4
x
3
+ 2x =
x
4
(4x +x
2
+ 8) = 0.
Thus x = 0 or x
2
+ 4x + 8 = 0. The quadratic has no real solutions, so the only one critical point is (0, 0).
At (0, 0), we have
D(0, 0) = fxxfyy (fxy)
2
= (12)(12) 0
2
= 144 < 0,
so (0, 0) is a saddle point.
4. The partial derivatives are
fx = cos x + cos (x +y).
fy = cos y + cos (x +y).
Setting fx = 0 and fy = 0 gives
cos x = cos y
For 0 < x < and 0 < y < , cos x = cos y only if x = y. Then, setting fx = fy = 0:
cos x + cos 2x = 0,
cos x + 2 cos
2
x 1 = 0,
(2 cos x 1)(cos x + 1) = 0.
So cos x = 1/2 or cos x = 1, that is x = /3 or x = . For the given domain 0 < x < , 0 < y < , we only
consider the solution when x = /3 then y = x = /3. Therefore, the critical point is (

3
,

3
).
Since
fxx(x, y) = sin x sin (x +y) fxx(

3
,

3
) = sin

3
sin
2
3
=

3
fxy(x, y) = sin (x +y) fxy(

3
,

3
) = sin
2
3
=

3
2
fyy(x, y) = sin y sin (x +y) fyy(

3
,

3
) = sin

3
sin
2
3
=

3
the discriminant is
D(x, y) = fxxfyy f
2
xy
= (

3)(

3) (

3
2
)
2
=
9
4
> 0.
Since fxx(

3
,

3
) =

3 < 0, (

3
,

3
) is a local maximum.
5. We nd critical points:
fx(x, y) = 12 6x = 0
fy(x, y) = 6 2y = 0
so (2, 3) is the only critical point. At this point
D = fxxfyy (fxy)
2
= (6)(2) = 12 > 0,
and fxx < 0, so (2, 3) is a local maximum. Since this is a quadratic, the local maximum is a global maximum.
Alternatively, we complete the square, giving
f(x, y) = 10 3(x
2
4x) (y
2
6y) = 31 3(x 2)
2
(y 3)
2
.
This expression for f shows that its maximum value (which is 31) occurs where x = 2, y = 3.
1084 Chapter Fifteen /SOLUTIONS
6. The partial derivatives are fx = 2x 3y, fy = 3y
2
3x. For critical points, solve fx = 0 and fy = 0 simultaneously.
From 2x 3y = 0 we get x =
3
2
y. Substituting it into 3y
2
3x = 0, we have that
3y
2
3(
3
2
y) = 3y
2

9
2
y = y(3y
9
2
) = 0.
So y = 0 or 3y
9
2
= 0, that is, y = 0 or y = 3/2. Therefore the critical points are (0, 0) and (
9
4
,
3
2
).
The contour diagram for f in Figure 15.31 (drawn by a computer), shows that (0, 0) is a saddle point and that (
9
4
,
3
2
) is a
local minimum.
4 3 2 1 1 2 3 4
4
3
2
1
1
2
3
4
110
90
70
50
30
10
10
30
50
70
90
10
30
0
0
0
x
y
(
9
4
,
3
2
)
Figure 15.31: Contour map of f(x, y) = x
2
+y
3
3xy
We can also see that (0, 0) is a saddle point and (
9
4
,
3
2
) is a local minimum analytically. Since fxx = 2, fyy =
6y, fxy = 3, the discriminant is
D(x, y) = fxxfyy f
2
xy
= 12y (3)
2
= 12y 9.
D(0, 0) = 9 < 0, so (0, 0) is a saddle point.
D(
9
4
,
3
2
) = 9 > 0 and fxx = 2 > 0, we know that (
9
4
,
3
2
) is a local minimum. The point (
9
4
,
3
2
) is not a global minimum
since f(
9
4
,
3
2
) = 1.6875, whereas f(0, 2) = 8.
7. The partial derivatives are
fx = y +
1
x
, fy = x + 2y.
For critical points, solve fx = 0 and fy = 0 simultaneously. From fy = x + 2y = 0 we get that x = 2y. Substituting
into fx = 0, we have
y +
1
x
= y
1
2y
=
1
y
(y
2

1
2
) = 0
Since
1
y
= 0, y
2

1
2
= 0, therefore
y =
1

2
=

2
2
,
and x =

2. So the critical points are


_

2,

2
2
_
and
_

2,

2
2
_
. But x must be greater than 0, so
_

2,

2
2
_
is
not in the domain.
The contour diagram for f in Figure 15.32 (drawn by computer), shows that
_

2,

2
2
_
is a saddle point of f(x, y).
SOLUTIONS to Review Problems for Chapter Fifteen 1085
4 5 6 7 8
4
3
2
1
1
2
3
4
x
y
0

10
`
10

f(

2,

2
2
) =
21
2
+ ln

2
20
0
10
20
30
Figure 15.32: Contour map of f(x, y) = xy + ln x +y
2
10
We can also see that
_

2,

2
2
_
is a saddle point analytically.
Since fxx =
1
x
2
, fyy = 2, fxy = 1, the discriminant is:
D(x, y) = fxxfyy f
2
xy
=
2
x
2
1.
D
_

2,

2
2
_
= 2 < 0, so
_

2,

2
2
_
is a saddle point.
8. Note that the x-axis and the y-axis are not in the domain of f. Since x = 0 and y = 0, by setting fx = 0 and fy = 0 we
get
fx = 1
1
x
2
= 0 when x = 1
fy = 1
4
y
2
= 0 when y = 2
So the critical points are (1, 2), (1, 2), (1, 2), (1, 2). Since fxx = 2/x
3
and fyy = 8/y
3
and fxy = 0, the
discriminant is
D(x, y) = fxxfyy f
2
xy
=
_
2
x
3
_
_
8
y
3
_
0
2
=
16
(xy)
3
.
Since D < 0 at the points (1, 2) and (1, 2), these points are saddle points. Since D > 0 at (1, 2) and (1, 2) and
fxx(1, 2) > 0 and fxx(1, 2) < 0, the point (1, 2) is a local minimum and the point (1, 2) is a local maximum.
No global maximum or minimum, since f(x, y) increases without bound if x and y increase in the rst quadrant; f(x, y)
decreases without bound if x and y decrease in the third quadrant.
9. The objective function is f(x, y) = 3x4y and the constraint equation is g(x, y) = x
2
+y
2
= 5, so grad f = 3

i 4

j
and grad g = (2x)

i + (2y)

j . Setting grad f = grad g gives


3 = (2x),
4 = (2y).
From the rst equation we have = 3/(2x), and from the second equation we have = 2/y. Setting these equal gives
x =
3
4
y.
Substituting this into the constraint equation x
2
+ y
2
= 5 gives y
2
= 16/5, so y = 4/

5 and y = 4/

5. Since
x =
3
4
y, there are two points where a maximum or a minimum might occur:
(3/

5, 4/

5) and (3/

5, 4/

5).
Since the constraint is closed and bounded, maximum and minimum values of f subject to the constraint exist. Since
f(3/

5, 4/

5) = 5

5 and f(3/

5, 4/

5) = 5

5, we see that f has a minimum value at (3/

5, 4/

5) and
a maximum value at (3/

5, 4/

5).
1086 Chapter Fifteen /SOLUTIONS
10. The objective function is f(x, y) = x
2
+ 2y
2
and the constraint equation is g(x, y) = 3x + 5y = 200, so grad f =
(2x)

i + (4y)

j and grad g = 3

i + 5

j . Setting grad f = grad g gives


2x = 3,
4y = 5.
From the rst equation, we have = 2x/3, and from the second equation we have = 4y/5. Setting these equal gives
x = 1.2y.
Substituting this into the constraint equation 3x + 5y = 200 gives y = 23.256. Since x = 1.2y, we have x = 27.907. A
maximum or minimum value of f can occur only at (27.907, 23.256).
We have f(27.907, 23.256) = 1860.484. From Figure 15.33, we see that the point (27.907, 23.256) is a minimum
value of f subject to the given constraint.
10 20 30 40 50
10
20
30
40
50
1000
1860
3000
4000
(27.9, 23.6)
x
3x + 5y = 200
y
Figure 15.33
11. The objective function is f(x, y) = 2xy and the constraint equation is g(x, y) = 5x + 4y = 100, so grad f =
(2y)

i + (2x)

j and grad g = 5

i + 4

j . Setting grad f = grad g gives


2y = 5,
2x = 4.
From the rst equation we have = 2y/5, and from the second equation we have = x/2. Setting these equal gives
y = 1.25x.
Substituting this into the constraint equation 5x +4y = 100 gives x = 10 and y = 12.5. A maximum or minimum value
for f subject to the constraint can occur only at (10, 12.5).
We have f(10, 12.5) = 250. From Figure 15.34, we see that the point (10, 12.5) gives a maximum.
10 20
10
20
30
(10, 12.5)
100
250
400
x
5x + 4y = 100
y
Figure 15.34
SOLUTIONS to Review Problems for Chapter Fifteen 1087
12. We will use the Lagrange multipliers with:
Objective function: f(x, y) = 3x
2
2y
2
+ 20xy
Constraint: g(x, y) = x +y 100
We rst nd
f = (6x + 20y)

i + (4y + 20x)

j
g =

i +

j .
To optimize f, we must solve the equations
f = g
(6x + 20y)

i + (4y + 20x)

j = (

i +

j ) =

i +

j
We have a vector equation, so we equate the coordinates:
6x + 20y =
20x 4y = .
So 6x + 20y = 20x 4y
24y = 26x
y =
13
12
x
Substituting into the constraint equation x +y = 100, we obtain:
x +
13
12
x = 100
25
12
x = 100
x = 48.
Consequently, y = 52, and f(48, 52) = 37,600. The point (48, 52) leads to the extreme value of f(x, y), given that
x + y = 100. Note that f has no minimum on the line x + y = 100 since f(x, 100 x) = 3x
2
2(100 x)
2
+
20x(100 x) = 25x
2
+ 2400x 20000 which goes to as x goes to . Therefore, the point (48, 52) gives the
maximum value for f on the line x +y = 100.
13. Our objective function is f(x, y, z) = x
2
y
2
2z and our equation of constraint is g(x, y, z) = x
2
+ y
2
z = 0.
To optimize f(x, y, z) with Lagrange multipliers, we solve f(x, y, z) = g(x, y, z) subject to g(x, y, z) = 0. The
gradients of f and g are
f(x, y, z) = 2x

i 2y

j 2

k ,
g(x, y, z) = 2x

i + 2y

k .
We get
2x = 2x
2y = 2y
2 =
x
2
+y
2
= z.
The third equation gives = 2 and from the rst x = 0, from the second y = 0 and from the fourth z = 0. So the only
solution is (0, 0, 0), and f(0, 0, 0) = 0.
To see what kind of extreme point is (0, 0, 0), let (a, b, c) be a point which satises the constraint, i.e. a
2
+ b
2
= c.
Then f(a, b, c) = a
2
b
2
2c = a
2
3b
2
0. The conclusion is that 0 is the maximum value of f and that there is
no minimum.
14. We rst nd the critical points in the disk
z = (8x y)

i + (8y x)

j
Setting z = 0 gives 8x y = 0 and 8y x = 0. The only solution is x = y = 0. So (0, 0) is the only critical point in
the disk.
1088 Chapter Fifteen /SOLUTIONS
Next we nd the extremal values on the boundary using Lagrange multipliers. We have objective function z =
4x
2
xy + 4y
2
and constraint G = x
2
+y
2
2 = 0.
z = (8x y)

i + (8y x)

j
G = 2x

i + 2y

j
z = G gives
8x y = 2x
8y x = 2y
If = 0 we get
8x y = 0
8y x = 0
with only solutions x = y = 0, which does not satisfy the constraint: x
2
+y
2
2 = 0. Therefore = 0 and we get:
2y(8x y) = 2x(8y x)
and
y(8x y) = x(8y x).
So x
2
= y
2
, x = y.
Substitute into G = 0, we get 2x
2
2 = 0 so x = 1. The extremal points on the boundary are therefore
(1, 1), (1, 1), (1, 1), (1, 1). The region x
2
+y
2
2 is closed and bounded, so minimum values of f in the region
exist. We check the values of z at these points :
z(1, 1) = 7, z(1, 1) = 7, z(1, 1) = 9, z(1, 1) = 9, z(0, 0) = 0
Thus (1, 1) and (1, 1) give the maxima over the closed disk and (0, 0) gives the minimum.
15. The region x
2
y is the shaded region in Figure 15.35 which includes the parabola y = x
2
.
10
30
50
10
30
50
70 70
x
y
Figure 15.35
We rst want to nd the local maxima and minima of f in the interior of our region. So we need to nd the extrema
of
f(x, y) = x
2
y
2
, in the region x
2
> y.
For this we compute the critical points:
fx = 2x = 0
fy = 2y = 0.
As (0, 0) does not belong to the region x
2
> y, we have no critical points. Now lets nd the local extrema of f on
the boundary of our region, hence this time we have to solve a constraint problem. We want to nd the extrema of
f(x, y) = x
2
y
2
subject to g(x, y) = x
2
y = 0. We use Lagrange multipliers:
grad f = gradg and x
2
= y.
SOLUTIONS to Review Problems for Chapter Fifteen 1089
This gives
2x = 2x
2y =
x
2
= y.
From the rst equation we get x = 0 or = 1.
If x = 0, from the third equation we get y = 0, so one solution is (0, 0). If x = 0, then = 1 and from the second
equation we get y =
1
2
. This gives x
2
=
1
2
so the solutions (
1

2
,
1
2
) and (
1

2
,
1
2
).
So f(0, 0) = 0 and f(
1

2
,
1
2
) = f(
1

2
,
1
2
) =
1
4
. From Figure 15.35 showing the level curves of f and the region
x
2
y, we see that (0, 0) is a local minimum of f on x
2
= y, but not a global minimum and that (
1

2
,
1
2
) and (
1

2
,
1
2
)
are global maxima of f on x
2
= y but not global maxima of f on the whole region x
2
y.
So there are no global extrema of f in the region x
2
y.
16. Let the line be in the form y = b + mx. Then, when x equals 0, 1, and 2, y equals b, b + m, and b + 2m respectively.
The sum of the squares of the vertical distances, which is what we want to minimize, is
f(m, b) = (4 b)
2
+ (3 (b +m))
2
+ (1 (b + 2m))
2
To nd critical points, set each partial derivative equal to zero.
fm = 0 + 2(3 (b +m))(1) + 2(1 (b + 2m))(2)
= 6b + 10m10
f
b
= 2(4 b)(1) + 2(3 (b +m))(1) + 2(1 (b + 2m))(1)
= 6b + 6m16
Setting both partial derivatives equal to zero and dividing by 2, we get a system of equations:
3b + 5m = 5
3b + 3m = 8
with solutions m =
3
2
and b =
25
6
. Thus, the line is y =
25
6

3
2
x.
Problems
17. Since fxx < 0 and D = fxxfyy f
2
xy
> 0, the point (1, 3) is a maximum. See Figure 15.36.
1
3
x
y

1
4
1
6

3
2
6
4

1
2
0
Figure 15.36
18. We rst express the revenue R in terms of the prices p1 and p2:
R(p1, p2) = p1q1 +p2q2
= p1(517 3.5p1 + 0.8p2) +p2(770 4.4p2 + 1.4p1)
= 517p1 3.5p
2
1
+ 770p2 4.4p
2
2
+ 2.2p1p2.
1090 Chapter Fifteen /SOLUTIONS
At a local maximum we have grad R = 0, and so:
R
p1
= 517 7p1 + 2.2p2 = 0,
R
p2
= 770 8.8p2 + 2.2p1 = 0.
Solving these equations, we nd that
p1 = 110 and p2 = 115.
To see whether or not we have a found a local maximum, we compute the second-order partial derivatives:

2
R
p
2
1
= 7,

2
R
p
2
2
= 8.8,

2
R
p1p2
= 2.2.
Therefore,
D =

2
R
p
2
1

2
R
p
2
2


2
R
p1p2
= (7)(8.8) (2.2)
2
= 56.76,
and so we have found a local maximum point. The graph of P(p1, p2) has the shape of an upside down paraboloid. Since
P is quadratic in q1 and q2, (110, 115) is a global maximum point.
19. Using Lagrange multipliers, let G = 2000 5x 10y = 0 be the constraint.
P =
_
1 +
2xy
2
2 10
8
_

i +
_
2 +
2yx
2
2 10
8
_

j =
_
1 +
xy
2
10
8
_

i +
_
2 +
yx
2
10
8
_

j .
G = 5

i 10

j .
Now, P = G, so
1 +
xy
2
10
8
= 5 and 2 +
yx
2
10
8
= 10.
Thus
2 +
2xy
2
10
8
= 2 +
yx
2
10
8
.
Solving, we get 2y = x or x = 0 or y = 0.
400
200
x
y
Constraint
G = 0
Figure 15.37
From G = 0 we have: when x = 0, y = 200, when y = 0, x = 400, and when x = 2y, x = 200, y = 100. So
(0,200), (400,0) and (200,100) are the critical points and they include the end points.
Substitute into P: P(0, 200) = 400, P(400, 0) = 400, P(200, 100) = 402 so the organization should buy 200 sacks of
rice and 100 sacks of beans.
20. We want to minimize cost C = 100L + 200K subject to Q = 900L
1/2
K
2/3
= 36000. Using Lagrange multipliers, we
get
Q =
_
450L
1/2
K
2/3
_

i +
_
600L
1/2
K
1/3
_

j .
C = 100

i + 200

j
C = Q gives
100 = 450L
1/2
K
2/3
and 200 = 600L
1/2
K
1/3
.
Since = 0 this gives
450L
1/2
K
2/3
= 300L
1/2
K
1/3
.
SOLUTIONS to Review Problems for Chapter Fifteen 1091
Solving, we get L = (3/2)K. Substituting into Q = 36,000 gives
900
_
3
2
K
_
1/2
K
2/3
= 36,000.
Solving yields K =
_
40
_
2
3
_
1/2
_
6/7
19.85, so L
3
2
(19.85) = 29.78. We can thus calculate cost using K = 20
and L = 30 which gives C = $7, 000.
21. We wish to minimize the objective function
C(x, y, z) = 20x + 10y + 5z
subject to the budget constraint
Q(x, y, z) = 20x
1/2
y
1/4
z
2/5
= 1, 200.
Therefore, we solve the equations grad C = grad Q and Q = 1, 200:
20 = 10x
1/2
y
1/4
z
2/5
or = 2x
1/2
y
1/4
z
2/5
,
10 = 5x
1/2
y
3/4
z
2/5
, or = 2x
1/2
y
3/4
z
2/5
,
5 = 8x
1/2
y
1/4
z
3/5
, or = 0.625x
1/2
y
1/4
z
3/5
,
20x
1/2
y
1/4
z
2/5
= 1, 200.
The rst and second equations imply that
x = y,
while the second and third equations imply that
3.2y = z.
Substituting for x and z in the constraint equation gives
20y
1/2
y
1/4
(3.2y)
2/5
= 1200
y 23.47,
and so
x 23.47 and z 75.1.
22. We want to maximize f(x, y) = 80x
0.75
y
0.25
subject to the budget constraint g(x, y) = 6x + 4y = 8000. Setting
grad f = grad g gives us
80(0.75x
0.25
)y
0.25
= 6,
80x
0.75
(0.25y
0.75
) = 4.
At the maximum, x, y = 0. From the rst equation we have = 10y
0.25
/x
0.25
, and from the second equation we have
= 5x
0.75
/y
0.75
. Setting these equal gives
y = 0.5x.
Substituting this into the constraint equation 6x+4y = 8000, we see that x = 1000. Since y = 0.5x, we obtain y = 500.
That the point (1000, 500) gives the maximum production is suggested by Figure 15.38, since the values of f decrease as
we move along the constraint away from (1000, 500). To produce the maximum quantity, the company should use 1000
units of labor and 500 units of capital.
1092 Chapter Fifteen /SOLUTIONS
500 1000 1500
500
1000
1500
2000
Budget constraint: 6x + 4y = 8000
P = (1000, 500)
50000
f = 60000
f = 70000
f = 80000
f = 90000
x
y
Figure 15.38
23. (a) We want to minimize C subject to g = x +y = 39. Solving C = g gives
10x + 2y =
2x + 6y =
so y = 2x. Solving with x +y = 39 gives x = 13, y = 26, = 182. Therefore C = $4349.
(b) Since = 182, increasing production by 1 will cause costs to increase by approximately $182. (because =
C
g
= rate of change of C with g). Similarly, decreasing production by 1 will save approximately $182.
24. (a) To be producing the maximum quantity Q under the cost constraint given, the rm should be using K and L values
given by
Q
K
= 0.6aK
0.4
L
0.4
= 20
Q
L
= 0.4aK
0.6
L
0.6
= 10
20K + 10L = 150.
Hence
0.6aK
0.4
L
0.4
0.4aK
0.6
L
0.6
= 1.5
L
K
=
20
10
= 2, so L =
4
3
K. Substituting in 20K + 10L = 150, we obtain
20K + 10
_
4
3
_
K = 150. Then K =
9
2
and L = 6, so capital should be reduced by
1
2
unit, and labor should be
increased by 1 unit.
(b)
New production
Old production
=
a4.5
0.6
6
0.4
a5
0.6
5
0.4
1.01, so tell the board of directors, Reducing the quantity of capital by 1/2 unit
and increasing the quantity of labor by 1 unit will increase production by 1% while holding costs to $150.
25. Cost of production, C, is given by C = p1W +p2K = b. At the optimal point, q = C.
Since q =
_
c(1 a)W
a
K
a
_

i +
_
caW
1a
K
a1
_

j and C = p1

i +p2

j , we get
c(1 a)W
a
K
a
= p1 and caW
1a
K
a1
= p2.
Now, marginal productivity of labor is given by
q
W
= c(1 a)W
a
K
a
and marginal productivity of capital is given by
q
K
= caW
1a
K
a1
, so their ratio is given by
q
W
q
K
=
c(1 a)W
a
K
a
caW
1a
K
a1
=
p1
p2
=
p1
p2
which is the ratio of the cost of one unit of labor to the cost of one unit of capital.
SOLUTIONS to Review Problems for Chapter Fifteen 1093
26. (a) Points A, B, C, D, E; that is, where a level curve of f and the constraint curve are parallel.
(b) Point F since the value of f is greatest at this point.
(c) Point D has the greatest f value of the points A, B, C, D, E.
9
1
0
1
1
1
2
1
3

g = c
A
B
C
D
E
F
27. (a) By the method of Lagrange multipliers, the point (2, 1) is a candidate when the gradient for f at (2, 1) is a multiple of
the gradient of the constraint function at (2, 1). The constraint function is g(x, y) = x
2
+y
2
, so grad g = 2x

i +2y

j .
We have grad g(2, 1) = 4

i + 2

j . This is not a multiple of grad f(2, 1) = 3

i + 4

j , so (2, 1) is not a candidate.


(b) The constraint function is g(x, y) = (x5)
2
+(y+3)
2
, so grad g = 2(x5)

i +2(y+3)

j . We have gradg(2, 1) =
6

i + 8

j . This is a multiple of grad f(2, 1) = 3

i + 4

j , so (2, 1) is a candidate.
The contours near (2, 1) are parallel straight lines with increasing f-values as we move in the direction of
3

i + 4

j (approximately toward the northwest). The center of the constraint circle is at (5, 3), approximately
southeast of the point (2, 1). Thus the point (2, 1) is a candidate for a maximum.
In general, if the constraint is a circle and grad f points outside the circle, then the point is a candidate for a
maximum.
(c) The constraint function is g(x, y) = (x + 1)
2
+ (y 5)
2
. Thus grad g(2, 1) = 6

i 8

j , which is again a multiple


of grad f, so (2, 1) is a candidate.
This time the center of the constraint circle, (1, 5) is approximately northwest of (2, 1), the same general
direction in which gradf is pointing. This means the point (2, 1) is a candidate for a minimum.
In general, if the constraint is a circle and grad f points inside the circle, then the point is a candidate for a
minimum.
28. (a) (i) Suppose N = kA
p
. Then the rule of thumb tells us that if A is multiplied by 10, the value of N doubles. Thus
2N = k(10A)
p
= k10
p
A
p
.
Thus, dividing by N = kA
p
, we have
2 = 10
p
so taking logs to base 10 we have
p = log 2 = 0.3010.
(where log 2 means log
10
2). Thus,
N = kA
0.3010
.
(ii) Taking natural logs gives
ln N = ln(kA
p
)
ln N = ln k +p ln A
ln N ln k + 0.301 ln A
Thus, ln N is a linear function of ln A.
1094 Chapter Fifteen /SOLUTIONS
(b) Table 15.2 contains the natural logarithms of the data:
Table 15.2 ln N and ln A
Island ln A ln N
Redonda 1.1 1.6
Saba 3.0 2.2
Montserrat 2.3 2.7
Puerto Rico 9.1 4.3
Jamaica 9.3 4.2
Hispaniola 11.2 4.8
Cuba 11.6 4.8
Using a least squares t we nd the line:
ln N = 1.20 + 0.32 ln A
This yields the power function:
N = e
1.20
A
0.32
= 3.32A
0.32
Since 0.32 is pretty close to log 2 0.301, the answer does agree with the biological rule.
29. (a) The objective function is the complementary energy,
f
2
1
2k1
+
f
2
2
2k2
, and the constraint is f1 +f2 = mg. The Lagrangian
function is
L(f1, f2, ) =
f
2
1
2k1
+
f
2
2
2k2
(f1 +f2 mg).
We look for solutions to the system of equations we get from grad L =

0 :
L
f1
=
f1
k1
= 0
L
f2
=
f2
k2
= 0
L

= (f1 +f2 mg) = 0.


Combining
L
f1

L
f2
=
f1
k1

f2
k2
= 0 with
L

= 0 gives the two equation system


f1
k1

f2
k2
= 0
f1 +f2 = mg.
Substituting f2 = mg f1 into the rst equation leads to
f1 =
k1
k1 +k2
mg
f2 =
k2
k1 +k2
mg.
(b) Hookes Law states that for a spring
Force of spring = Spring constant Distance stretched or compressed from equilibrium.
Since f1 = k1 and f2 = k2 , the Lagrange multiplier equals the distance the mass stretches the top spring
and compresses the lower spring.
SOLUTIONS to Review Problems for Chapter Fifteen 1095
30. The distance to the origin d =
_
x
2
+y
2
+z
2
. Minimizing d
2
minimizes d, so let d
2
= f(x, y, z) = x
2
+y
2
+z
2
. We
minimize f with constraint z = xy + 1, so we let g(x, y, z) = xy + 1 z.
Lagrange:

f =

g gives
2x = y
2y = x
2z =
z = xy + 1
_

_
which gives
x = 0
y = 0
z = 1
= 2.
The point is (0, 0, 1). This must be a minimum since there are points on z = xy + 1 innitely far from the origin.
(There is no maximum.)
31. Since patient 1 has a visit every x1 weeks, this patient has 1/x1 visits per week. Similarly, patient 2 has 1/x2 visits per
week. Thus, the constraint is
g(x1, x2) =
1
x1
+
1
x2
= m
To minimize
f(x1, x2) =
v1
v1 +v2

x1
2
+
v2
v1 +v2

x2
2
subject to g(x1, x2), we solve the equations
grad f = grad g
g(x1, x2) = m.
This gives us the equations
f
x1
=
v1
v1 +v2

1
2
=
_

1
x
2
1
_
=
g
x1
f
x2
=
v2
v1 +v2

1
2
=
_

1
x
2
2
_
=
g
x2
1
x1
+
1
x2
= m.
Dividing the rst equation by the second gives
v1
v2
=
x
2
2
x
2
1
.
As v1, v2, x1, x2, m are strictly positive we have
x2
x1
=
_
v1
v2
_1
2
.
Substituting for x2 in the constraint occasion gives
1
x1
+
_
v2
v1
_1
2

1
x1
= m
solving for x1 gives
1
x1
_
1 + (
v1
v2
)
1
2
_
= m
x1 =
(v1)
1
2
+ (v2)
1
2
m (v1)
1
2
,
and similarly
x2 =
(v1)
1
2
+ (v2)
1
2
m (v2)
1
2
.
32. We want to optimize
f(x1, x2) =
v1
v1 +v2

x1
2
+
v2
v1 +v2

x2
2
subject to
g(x1, x2) =
1
x1
+
1
x2
= m.
1096 Chapter Fifteen /SOLUTIONS
At the optimum point, x1, x2, and the Lagrange multiplier must satisfy the equations
v1
v1 +v2

1
2
=

x
2
1
v2
v1 +v2

1
2
=

x
2
2
1
x1
+
1
x2
= m.
Solving the rst and second equations for 1/x1 and 1/x2, respectively, gives
1
x
2
1
=
1
2

v1
(v1 +v2)
1
x
2
2
=
1
2

v2
(v1 +v2)
substituting into the constraint gives (note that < 0):
_

1
2

v1
(v1 +v2)
_
1/2
+
_
1
2

v2
(v1 +v2)
_
1/2
=
_

1
2
_
1/2

(v1)
1/2
+ (v2)
1/2
(v1 +v2)
1/2
= m.
So

1
2

v1 +v2 + 2(v1v2)
1/2
v1 +v2
= m
2
.
and thus
=
1
2m
2
_
1 +
2(v1v2)
1/2
v1 +v2
_
.
The units of are weeks
2
(since the units of m are 1/weeks). The Lagrange multiplier measures df/dm, which
represents the rate of change in the expected delay in tumor detection as the available number of visits per week increases.
The negative sign represents the fact that as the number of visits per week increases, the delay decreases.
33. We want to minimize the total cost, C, of the cable. The distance AB is

50
2
+x
2
meters and the distance BC is
_
30
2
+y
2
, and the cost for each segment is cost/meter length. Thus
C = 500
_
50
2
+x
2
+ 2000
_
30
2
+y
2
+ 300z,
where x, y, z must satisfy the constraint
g(x, y, z) = x +y +z = 100.
At the optimum point, C = g, so
C
x
=
500x

50
2
+x
2
= =
g
x
C
y
=
2000y
_
30
2
+y
2
= =
g
y
C
z
= 300 = =
g
z
.
Thus, using = 300, we have
2000y
_
30
2
+y
2
= 300
20y = 3
_
30
2
+y
2
400y
2
= 9(900 +y
2
)
391y
2
= 8100
y =
90

391
= 4.6 meters.
SOLUTIONS to Review Problems for Chapter Fifteen 1097
Similarly
500x

50
2
+x
2
= 300
5x = 3
_
50
2
+x
2
25x
2
= 9(2500 +x
2
)
16x
2
= 9 2500
x =
3 50
4
= 37.5 meters.
Thus z = 100 37.5 4.6 = 57.9 meters. The values x = 37.5 m, y = 4.6 m, and z = 57.9 m give the minimum cost.
How do we know these values of x, y, z give a minimum? Since the variables are constrained to lie in the rst octant
on the plane x + y + z = 100, we nd the minimum cost on the boundary, which consists of three line segments on the
coordinate planes, and compare with the cost at the critical point.
If x = 0, then z = 100 y and the minimum of C(x, y, z) = 500 50 + 2000
_
30
2
+y
2
+ 300(100 y) for
0 y 100 is C(0, 0, 100) = $115,000.
If y = 0, then z = 100 x and the minimum of C(x, y, z) = 500

50
2
+x
2
+ 2000 30 + 300(100 x) for
0 x 100 is C(37.5, 0, 62.5) = $110,000.
If z = 0, then y = 100 x and the minimum of C(x, y, z) = 500

50
2
+x
2
+ 2000
_
30
2
+ (100 x)
2
for
0 x 100 is C(100, 0, 0) = $115,902.
The cost at the critical point, C(37.5, 4.6, 57.9) = $109,321, is lower than the minimum $110, 000 on the boundary.
Thus, the minimum cost subject to the constraint is C(37.5, 4.6, 57.9) = $109,321.
34. The wetted perimeter of the trapezoid is given by the sum of the lengths of the three walls, so
p = w +
2d
sin
We want to minimize p subject to the constraint that the area is xed at 50 m
2
. A trapezoid of height h and with parallel
sides of lengths b1 and b2 has
A = Area = h
(b1 +b2)
2
.
In this case, d corresponds to h and b1 corresponds to w. The b2 term corresponds to the width of the exposed surface of
the canal. We nd that b2 = w+(2d)/(tan ). Substituting into our original equation for the area along with the fact that
the area is xed at 50 m
2
, we arrive at the formula:
Area =
d
2
_
w +w +
2d
tan
_
= d
_
w +
d
tan
_
= 50
We now solve the constraint equation for one of the variables; we will choose w to give
w =
50
d

d
tan
.
Substituting into the expression for p gives
p = w +
2d
sin
=
50
d

d
tan
+
2d
sin
.
We now take partial derivatives:
p
d
=
50
d
2

1
tan
+
2
sin
p

=
d
tan
2


1
cos
2


2d
sin
2

cos
From p/ = 0, we get
d cos
2

sin
2


1
cos
2

=
2d
sin
2

cos .
Since sin = 0 and cos = 0, canceling gives
1 = 2 cos
so
cos =
1
2
.
Since 0 < <

2
, we get =

3
.
1098 Chapter Fifteen /SOLUTIONS
Substituting into the equation p/d = 0 and solving for d gives:
50
d
2

1

3
+
2

3/2
= 0
which leads to
d =
_
50

3
5.37m.
Then
w =
50
d

d
tan

50
5.37

5.37

3
6.21 m.
When = /3, w 6.21 m and d 5.37 m, we have p 18.61 m.
Since there is only one critical point, and since p increases without limit as d or shrink to zero, the critical point
must give the global minimum for p.
35. (a)
A
B

d
a
b

1
Medium 1
Medium 2
A

R
B

Figure 15.39
See Figure 15.39. The time to travel from A to B is given by
T(1, 2) =
AR
v1
+
RB
v2
=
a
v1 cos 1
+
b
v2 cos 2
.
(b) The distance d = A

= A

R +RB

. Hence
d = a tan 1 +b tan 2.
(c) We imagine the following extreme case: the light ray rst travels through medium 1 to a point R on the boundary
far to the left of A

, then through medium 2 toward B. The distance traveled this way is very large, hence the travel
time is large as well. Similarly, if R is far to the right of B

, the travel time will be large. Therefore values of 1 near


/2 or /2 increase the time, T.
(d) The constrained optimization problem is: minimize T(1, 2) subject to g(1, 2) = a tan 1 + b tan 2 = d.
According to the method of Lagrange multipliers, the minimum point should be among those satisfying grad T =
grad g as well as the constraint.We have
grad T(1, 2) =
a
v1
sin 1
cos
2
1

i +
b
v2
sin 2
cos
2
2

j
and
grad g(1, 2) = a
1
cos
2
1

i +b
1
cos
2
2

j .
The condition grad T = grad g becomes
a
v1
sin 1
cos
2
1
= a
1
cos
2
1
and
b
v2
sin 2
cos
2
2
= b
1
cos
2
2
.
Eliminating we are left with
sin 1
v1
=
sin 2
v2
or
sin 1
sin 2
=
v1
v2
,
which is Snells law. The argument in part (c) shows that the critical point corresponding to 1, 2 satisfying Snells
law is indeed a minimum.
SOLUTIONS to Review Problems for Chapter Fifteen 1099
CAS Challenge Problems
36. (a) The partial derivatives of f are
f
x
=

a +x +
_
1 +

a +x
_
y
2

a +x

a +x +y
_
1 +

a +x +y
_
2
f
y
=
1 2 a 2 x +

a +x y
2

a +x +y
_
1 +

a +x +y
_
2
Solving f/x = 0, f/y = 0, we get x = 1/4 a, y = 1. The discriminant at this point is D = 16/625.
Thus, by the second derivative test, the point is a saddle point.
(b) The y coordinate of the critical points stays the same and the x coordinate is a units to the left of its position when
a = 0. The type is always a saddle point. This is because f is obtained from

x +y
1 +y +

x
by substituting x + a for
x, so that the graph is shifted a units in the negative x-direction but its shape remains the same.
37. (a) We have grad f = 2x

i +

j and grad g = (2x + 2y)

i + (2x + 2y)

j . So the equations to be solved in the method


of Lagrange multipliers are
2x = (2x + 2y)
1 = (2x + 2y)
x
2
+ 2xy +y
2
9 = 0
Solving these with a CAS, we get two solutions:
x = 1/2, y = 7/2, = 1/6, or x = 1/2, y = 5/2, = 1/6
Student A reasons that since f(1/2, 7/2) = 13/4 and f(1/2, 5/2) = 11/4 , the (global) maximum and
minimum values are 11/4 and 13/4, respectively. Student B graphs the constraint curve g = 0 and a contour diagram
of f. The constraint curve turns out to be two straight lines, since the constraint x
2
+ 2xy +y
2
9 = 0, which can
be rewritten as (x + y)
2
= 9, or x + y = 3. The value of f goes to innity on each of these straight lines. On the
line y = x + 3, f(x, y) = x
2
+ y = x
2
x + 3, and on the line y = x 3, f(x, y) = x
2
+ y = x
2
x 3.
Thus Student B is correct. The points Student A found are actually local maximum and local minimum values, not
global. Since the constraint is not bounded, there is no guarantee that there is a local maximum or minimum. See
Figure 15.40.
4 2 2 4
4
2
2
4
g
g
x
y
Figure 15.40: Contours of f and two straight
lines giving constraint g = 0
38. (a) We have grad f = 3

i + 2

j and grad g = (4x 4y)

i + (4x + 10y)

j , so the Lagrange multiplier equations are


3 = (4x 4y)
2 = (4x + 10y)
2x
2
4xy + 5y
2
= 20
1100 Chapter Fifteen /SOLUTIONS
Solving these with a CAS we get = 0.4005, x = 3.9532, y = 2.0806 and = 0.4005, x = 3.9532, y =
2.0806. We have f(3.9532, 2.0806) = 11.0208, and f(3, 9532, 2.0806) = 21.0208. The constraint equation
is 2x
2
4xy +5y
2
= 20, or, completing the square, 2(x y)
2
+3y
2
= 20. This has the shape of a skewed ellipse,
so the constraint curve is bounded, and therefore the local maximum is a global maximum. Thus the maximum value
is 21.0208.
(b) The maximum value on g = 20.5 is 21.0208 + 0.5(0.4005) = 21.2211. The maximum value on g = 20.2 is
21.0208 + 0.2(0.4005) = 21.1008.
(c) We use the same commands in the CAS from part (a), with 20 replaced by 20.5 and 20.2, and get the maximum
values 21.2198 for g = 20.5 and 21.1007 for g = 20.2. These agree with the approximations we found in part (b) to
2 decimal places.
CHECK YOUR UNDERSTANDING
1. True. By denition, a critical point is either where the gradient of f is zero or does not exist.
2. False. The point P0 could be a saddle point of f.
3. False. The point P0 could be a saddle point of f.
4. True. If P0 were not a critical point of f, then grad f(P0) would point in the direction of maximum increase of f, which
contradicts the fact that P0 is a local maximum or minimum.
5. True. The graph of this function is a cone that opens upward with its vertex at the origin.
6. False. The graph of this function is a saddle shape, with a saddle point at the origin. The function increases in the

i
direction and decreases in the

j direction.
7. True. Adding 5 to the function shifts the graph 5 units vertically, which leaves the (x, y) coordinates of the local extrema
intact.
8. True. Multiplying by 1 turns the graph of f upside down, so local maxima become local minima and vice-versa.
9. False. For example, the linear function f(x, y) = x +y has no local extrema at all.
10. False. The statement is only true for points sufciently close to P0.
11. False. Local maxima are only high points for f when compared to nearby values; the global maximum is the largest of
any values of f over its entire domain.
12. True. For unconstrained optimization, global extrema occur at one (or more) of the local extrema.
13. False. For example, the linear function f(x, y) = x + y has neither a global minimum or global maximum on all of
2-space.
14. True. The region is the unit disk without its boundary (the unit circle), and the distance between any two points in this
region is less than 2it does not stretch off to innity in any direction.
15. False. The region is the unit disk without its boundary (the unit circle), so it is not closed (in fact, it is open).
16. True. The global minimum is 0, which occurs at the origin. This is clear since the function f(x, y) = x
2
+ y
2
is greater
than or equal to zero everywhere, and is only zero at the origin.
17. False. On the given region the function f is always less than one. By picking points closer and closer to the circle
x
2
+ y
2
= 1 we can make f larger and larger (although never larger than one). There is no point in the open disk that
gives f its largest value.
18. False. While f can only have (at most) one largest value, it may attain this value at more than one point. For example, the
function f(x, y) = sin(x +y) has a global maximum of 1 at both (/2, 0) and (0, /2).
19. True. The region is both closed and bounded, guaranteeing both a global maximum and minimum.
20. True. The global minimum could occur on the boundary of the region.
21. True. The point (a, b) must lie on the constraint g(x, y) = c, so g(a, b) = c.
22. False. The point (a, b) is not necessarily a critical point of f, since it is a constrained extremum.
23. True. The constraint is the same as x = y, so along the constraint f = 2x, which grows without bound as x .
24. False. The condition grad f = grad g yields the two equations 1 = 2x and 2 = 2y. Substituting x = 2 in the rst
equation gives = 1/4, while setting y = 1 in the second gives = 1, so the point (2, 1) is not a local extremum
of f constrained to x + 2y = 0.
25. False. Since grad f and grad g point in opposite directions, they are parallel. Therefore (a, b) could be a local maximum
or local minimum of f constrained to g = c. However the information given is not enough to determine that it is a
PROJECTS FOR CHAPTER FIFTEEN 1101
minimum. If the contours of g near (a, b) increase in the opposite direction as the contours of f, then at a point with
grad f(a, b) = grad g(a, b) we have 0, but this can be a local maximum or minimum.
For example, f(x, y) = 4 x
2
y
2
has a local maximum at (1, 1) on the constraint g(x, y) = x + y = 2. Yet at
this point, grad f = 2

i 2

j and grad g =

i +

j , so grad f and grad g point in opposite directions.


26. False. A maximum for f subject to a constraint need not be a critical point of f
27. False. The condition for the Lagrange multiplier is grad f(a, b) = grad g(a, b).
28. False. Just as a critical point need not be a maximum or minimum for unconstrained optimization, a point satisfying the
Lagrange condition need not be a maximum or minimum for a constrained optimization.
29. True. Since f(a, b) = M, we must satisfy the Lagrange conditions that fx(a, b) = g(a, b) and fy(a, b) = gy(a, b),
for some . Thus fx(a, b)/fy(a, b) = gx(a, b)/gy(a, b).
30. True. Since f(a, b) = m, the point (a, b) must satisfy the Lagrange condition that fx(a, b) = g(a, b), for some . In
particular, if gx(a, b) = 0, then fx(a, b) = 0.
31. False. Whether increasing c will increase M depends on the sign of at a point (a, b) where f(a, b) = M
32. True. The value of at a maximum point gives the proportional change in M for a change in c.
33. False. The value of at a minimum point gives the proportional change in m for a change in c. If > 0 and the change
in c is positive, the change in m will also be positive.
PROJECTS FOR CHAPTER FIFTEEN
1. (a) If p = e
x
where x then p 0 with p > 0 and
lim
p0
p>0
(p ln p) = lim
x
(xe
x
) = 0,
since the exponential decreases faster than any power of x. Alternatively, use lHopitals rule:
lim
p0
p>0
(p ln p) = lim
p0
p>0
ln p
1/p
= lim
p0
p>0
1/p
1/p
2
= 0
(b) We apply the method of Lagrange multipliers to nd the critical points of S(p
1
, , p
30
). The constraint
function is g(p
1
, , p
30
) = p
1
+ + p
30
. We have
S
p
j
=

p
j
_

30

i=1
p
i
ln p
i
ln 2
_
=
1
ln2
(ln p
j
+ 1),
therefore
grad S =
1
ln 2
30

j=1
(ln p
j
+ 1)

k
j
where

k
1
, ,

k
30
are the unit vectors corresponding 30 independent directions of the p
j
-axes. Also,
grad g =
30

j=1

k
j
so the condition grad S = grad g becomes

1
ln 2
(ln p
j
+ 1) = , for i = 1, , 30.
Thus,
ln p
j
= ln 2 1
1102 Chapter Fifteen /SOLUTIONS
and, in particular, all the p
j
s must be equal. Since the p
j
s have to satisfy the constraint g(p
1
, , p
30
) = 1,
we see that p
j
=
1
30
and that the point (
1
30
,
1
30
, ,
1
30
) is the only critical point of S. We have
S
_
1
30
,
1
30
, ,
1
30
_
= 30
1
30
(ln 30)
ln 2
=
ln 30
ln 2
.
We will not prove that this is indeed the maximum value of S (this requires a higher-dimensional analogue
of the second derivative test). Since in part (c) we show that the minimum value of S is 0, the critical point
we have found here is not a global minimum; the maximum of S has to be attained somewhere and it is
reasonable to believe that it is attained at the unique critical point. The maximum entropy corresponds to
maximum uncertainty in the outcome of the competition.
(c) We already know that S 0. However, S can be zero: For example, if p
1
= 1 and p
2
= = p
30
= 0,
we have S(1, 0, , 0) = 0. Therefore the minimum value of S is 0. Now we determine all the values of
p
i
s for which S(p
1
, , p
30
) = 0. The condition
S =
30

i=1
p
i
ln p
i
2
= 0
together with the restrictions ln p
i
0 shows that, for S to vanish, each individual term in the above
sum has to vanish. This means p
i
ln p
i
= 0 for all i = 1, , 30, that is, p
i
= 0 or p
i
= 1 for i = 1, , 30.
Since

30
i=1
p
i
= 1, only one of the p
i
s is 1 whereas the other 29 are 0. This corresponds to the case where
one of the teams is certain to win, that is, there is no uncertainty. The result can be interpreted by saying
that zero entropy implies zero uncertainty.
2.
(xi, b +mxi)
(xi, yi) (x1, b +mx1)
(x1, y1)
y = b +mx
Figure 15.41
(a) Points which are directly above or below each other share the same x coordinate, therefore, the point on
the least squares line which is directly above or below the point in question will have x coordinate x
i
and
from the formula for the least squares line, it will have y coordinate b + mx
i
. (See Figure 14.1.)
(b) The general distance formula in two dimensions is d =
_
(x
2
x
1
)
2
+ (y
2
y
1
)
2
, so d
2
= (x
2
x
1
)
2
+
(y
2
y
1
)
2
. Since the x coordinates are identical for the two points in question, the rst term in the square
root is zero. This yields d
2
= (y
i
(b + mx
i
))
2
.
(c) In both cases we use the chain rule and our knowledge of summations to show the relationship.
f
b
=

b
_
n

i=1
(y
i
(b + mx
i
))
2
_
=
n

i=1

b
(y
i
(b + mx
i
))
2
=
n

i=1
2(y
i
(b + mx
i
))

b
(y
i
(b + mx
i
))
=
n

i=1
2(y
i
(b + mx
i
)) (1)
= 2
n

i=1
(y
i
(b + mx
i
))
PROJECTS FOR CHAPTER FIFTEEN 1103
f
m
=

m
_
n

i=1
(y
i
(b + mx
i
))
2
_
=
n

i=1

m
(y
i
(b + mx
i
))
2
=
n

i=1
2(y
i
(b + mx
i
))

m
(y
i
(b + mx
i
))
=
n

i=1
2(y
i
(b + mx
i
)) (x
i
)
= 2
n

i=1
(y
i
(b + mx
i
)) x
i
(d) We can separate
f
b
into three sums as shown:
f
b
= 2
_
n

i=1
y
i
b
n

i=1
1 m
n

i=1
x
i
_
Similarly we can separate
f
m
after multiplying through by x
i
:
f
m
= 2
_
n

i=1
y
i
x
i
b
n

i=1
x
i
m
n

i=1
x
i
2
_
Setting
f
b
and
f
m
equal to zero we have:
bn + m
n

i=1
x
i
=
n

i=1
y
i
b
n

i=1
x
i
+ m
n

i=1
x
2
i
=
n

i=1
x
i
y
i
(e) To solve this pair of linear equations, we multiply the rst equation by

n
i=1
x
2
i
, multiply the second one
by

n
i=1
x
i
, and subtract; we get
bn
n

i=1
x
2
i
b(
n

i=1
x
i
)
2
=
n

i=1
y
i
n

i=1
x
2
i

n

i=1
x
i
y
i
n

i=1
x
i
,
So,
b =
_
n

i=1
x
2
i
n

i=1
y
i

i=1
x
i
n

i=1
x
i
y
i
_
/
_
_
n
n

i=1
x
2
i

_
n

i=1
x
i
_
2
_
_
Similarly,
m =
_
n
n

i=1
x
i
y
i

i=1
x
i
n

i=1
y
i
_
/
_
_
n
n

i=1
x
2
i

_
n

i=1
x
i
_
2
_
_
(f) Applying the formulas to the given data, we have b =
1
3
, m = 1 which gives y = (1/3) + x, in
agreement with the example.

Vous aimerez peut-être aussi