Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
15
S E VERAL VA RIABLES
15.1 Functions of Two or More Variables (ET Section 14.1)
Preliminary Questions
1. What is the difference between a horizontal trace and a level curve? How are they related?
SOLUTION A horizontal trace at height c consists of all points (x, y, c) such that f (x, y) = c. A level curve is the
curve f (x, y) = c in the x y-plane. The horizontal trace is in the z = c plane. The two curves are related in the sense that
the level curve is the projection of the horizontal trace on the x y-plane. The two curves have the same shape but they are
located in parallel planes.
2. Describe the trace of f (x, y) = x 2 sin(x 3 y) in the x z-plane.
SOLUTION The intersection of the graph of f (x, y) = x 2 sin(x 3 y) with the x z-plane is obtained by setting y = 0 in
the equation z = x 2 sin(x 3 y). We get the equation z = x 2 sin 0 = x 2 . This is the parabola z = x 2 in the x z-plane.
3. Is it possible for two different level curves of a function to intersect? Explain.
SOLUTION Two different level curves of f (x, y) are the curves in the x y-plane defined by equations f (x, y) = c1 and
f (x, y) = c2 for c1 = c2 . If the curves intersect at a point (x 0 , y0 ), then f (x 0 , y0 ) = c1 and f (x 0 , y0 ) = c2 , which
implies that c1 = c2 . Therefore, two different level curves of a function do not intersect.
4. Describe the contour map of f (x, y) = x with contour interval 1.
SOLUTION The level curves of the function f (x, y) = x are the vertical lines x = c. Therefore, the contour map of f
with contour interval 1 consists of vertical lines so that every two adjacent lines are distanced one unit from another.
5. How will the contour maps of f (x, y) = x and g(x, y) = 2x with contour interval 1 look different?
SOLUTION The level curves of f (x, y) = x are the vertical lines x = c, and the level curves of g(x, y) = 2x are the
vertical lines 2x = c or x = 2c . Therefore, the contour map of f (x, y) = x with contour interval 1 consists of vertical
lines with distance one unit between adjacent lines, whereas in the contour map of g(x, y) = 2x (with contour interval
1) the distance between two adjacent vertical lines is 12 .
Exercises
In Exercises 14, evaluate the function at the specified points.
f (2, 2) = 2 + 2 23 = 18
f (1, 4) = 1 + 4 (1)3 = 5
f 6, 12 = 6 + 12 63 = 114
y
2. g(x, y) = 2 , (1, 3), (3, 2)
x + y2
SOLUTION We substitute (x, y) = (1, 3) and (x, y) = (3, 2) in the function to obtain
3 3 2 2
g(1, 3) = 2 = ; g(3, 2) = =
1 +3 2 10 32 + (2)2 13
1
h(3, 8, 2) = 3 8 22 = 3 8 =6
4
1 1
h(3, 2, 6) = 3 (2) (6)2 = 6 =
36 6
604 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
5. f (x, y) = 4x 7y
SOLUTION The function is defined for all x and y, hence the domain is the entire x y-plane.
6. f (x, y) = 9 x 2
SOLUTION The function f (x, y) = 9 x 2 is defined if 9 x 2 0, that is, if x 2 9. In other words, 3 x 3.
This region is the region enclosed by the two vertical lines x = 3 and x = 3 (including the two lines themselves).
7. f (x, y) = ln(y 2x)
SOLUTION The function is defined if y 2x > 0 or y > 2x. This is the region in the x y-plane that is above the line
y = 2x.
y > 2x
1
8. h(x, t) =
x +t
SOLUTION The function is defined if x + t = 0, that is, x = t. The domain is the xt-plane with the line x = t
excluded.
y
x = t
9. G(x, t) = e1/(x+t)
SOLUTION The function is defined if x + t = 0, that is, x = t. The domain is the same as the domain of h(x, t) in
the previous exercise.
y
x = t
S E C T I O N 15.1 Functions of Two or More Variables (ET Section 14.1) 605
1
10. g(y, z) =
z + y2
SOLUTION The function is defined if z + y 2 = 0, that is, z = y 2 . The domain is the (y, z) plane with the parabola
z = y 2 excluded.
D = (y, z) : z = y 2
y
z = y 2
z + y 2 = 0
y
11. f (x, y) = sin
x
SOLUTION The function is defined for all x = 0. The domain is the x y-plane with the y-axis excluded.
D = {(x, y) : x = 0}
y
x = 0
12. H (r, s) = 3r 2 s 4
SOLUTION The function is defined for all r and s, hence the domain is the entire r s-plane.
13. F(I, R) = I R
SOLUTION The function is defined if I R 0. Therefore the domain is the first and the third quadrants of the I R-plane
including both axes.
y
IR 0
D = {(x, y) : 1 x + y 1}
y
1
x+
y=
1
x
0
x+
y=
1
x+
y=
1
1
15. g(r, t) = 2
r t
SOLUTION The function is defined if r 2 t = 0, that is, t = r 2 . The domain is the r t-plane with the parabola t = r 2
excluded.
D = (r, t) : t = r 2
y
t = r2
t = r 2
y y = 4x2
17. f (x, y, z) = x z + e y
SOLUTION The domain of f is the entire (x, y, z)-space. Since f takes all the real values, the range is the entire real
line.
18. P(r, s) = er/s
SOLUTION The domain is the r s-plane with the r -axis excluded, since the function is defined only if s = 0. That is,
D = {(r, s) : s = 0}. The range is all positive real numbers.
19. f (x, y, z) = 9 x 2 y 2 z 2
SOLUTION The function is defined if 9 x 2 y 2 z 2 0, i.e, x 2 + y 2 + z 2 9. Therefore, the domain is the set
{(x, y, z) : x 2 + y 2 + z 2 9}, which is the ball of radius 3 centered at the origin, including the sphere x 2 + y 2 + z 2 = 9.
To find the range, we examine the values taken by f . For in the range, the following equation has solutions x, y, z:
9 x 2 y2 z2 = (1)
S E C T I O N 15.1 Functions of Two or More Variables (ET Section 14.1) 607
9 x 2 y2 z2 = 2
x 2 + y2 + z2 = 9 2
{ R : 0 3} .
20. Match the functions (a)(f) with their graphs (A)(F) in Figure 21.
(a) f (x, y) = |x| + |y|
(b) f (x, y) = cos(x y)
1
(c) f (x, y) =
1 + 9x 2 + y 2
1
(e) f (x, y) =
1 + 9x 2 + 9y 2
z z
y y
x x
(A) (B)
z z
y
x
y
(C) x (D)
z z
y
y
x x
(E) (F)
FIGURE 21
SOLUTION
(a) |x| + |y|. The level curves are |x| + |y| = c, y = c |x|, or y = c + |x|. The graph (D) corresponds to the
function with these level curves.
(b) cos(x y). The vertical trace in the plane x = c is the curve z = cos(c y) in the plane x = c. These traces
correspond to the graph (C).
608 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
1 1
(c) (e) . The level curves of the two functions are:
1 + 9x 2 + y 2 1 + 9x 2 + 9y 2
1 1
=c =c
1 + 9x 2 + y 2 1 + 9x 2 + 9y 2
1 1
1 + 9x 2 + y 2 = 1 + 9x 2 + 9y 2 =
c c
1 1
9x 2 + y 2 = 1 9x 2 + 9y 2 = 1
c c
1 + c
x 2 + y2 =
9c
For suitable values of c, the level curves of the function in (c) are ellipses as in (E), and the level curves of the function
(e) are circles as in (A).
(d) cos(x 2 )e1/(x +y )
2 2
(f) cos(x 2 + y 2 )e1/(x +y ) . The value of |z| is decreasing to zero as x or y are decreasing, hence the possible graphs
2 2
In Exercises 2130, sketch the graph and describe the vertical and horizontal traces.
21. f (x, y) = 12 3x 4y
SOLUTION The graph of f (x, y) = 12 3x 4y is shown in the figure:
z
12
3
4 y
The vertical traces obtained by setting x = a or y = a are the lines z = (12 3a) 4y and z = 3x + (12 4a) in
the planes x = a and y = a, respectively.
z z
y y
x x
22. f (x, y) = x 2 + y
SOLUTION The graph is shown in the figure:
S E C T I O N 15.1 Functions of Two or More Variables (ET Section 14.1) 609
2
1 1
z
0
0.5
1
1
0 y
0.5
0 0.5
x 0.5
1 1
20
10
z 5
0
2.5
5 0 y
2.5
0 2.5
x 2.5
5 5
The vertical trace obtained by setting x = a is the line z = y + a 2 in the plane x = a, and the vertical trace obtained by
setting y = a is the parabola z = x 2 + a in the plane y = a.
30 30
20 20
z 10 5 z 10 5
0 0
2.5 2.5
5 5
0 y 0 y
2.5 2.5
0 2.5 0 2.5
x 2.5 x 2.5
5 5 5 5
23. f (x, y) = x 2 + 4y 2
SOLUTION The graph of the function is shown in the figure:
The horizontal trace at height c is the curve x 2 + 4y 2 = c, where c 0 (if c = 0, it is the origin). The horizontal traces
are ellipses for c > 0.
610 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
The vertical trace in the plane x = a is the parabola z = a 2 + 4y 2 , and the vertical trace in the plane y = a is the
parabola z = x 2 + 4a 2 .
z
z
y
y
x
x
24. f (x, y) = y 2
SOLUTION The graph of the function is shown in the figure:
15
10 4
z
5
2
0
2
0 y
1
0 2
x 1
2 4
The horizontal trace at height c is y 2 = c. For c > 0 the trace consists of the two lines y = c and y = c in the
plane z = c, and for c = 0 it is the line y = 0.
1
0.5
y
0
0.5
1
1
0.75
z 0.5
0.25
0
4
2
0
x 2
4
1
0.5
y
0
0.5
1
1
0.75
z 0.5
0.25
0
4
2
0
x 2
4
0.75
z 0.5
0.25
0
4
2
0
x 2
4
1
25. f (x, y) = 2
x + y2 + 1
SOLUTION The graph of the function is shown in the figure:
The vertical trace in the plane x = a is the following curve in the plane x = a:
1 1
z= 2 z=
a + y2 + 1 (1 + a 2 ) + y 2
z
1
The vertical trace in the plane y = a is the curve z = 2 in this plane.
x + a2 + 1
z
26. f (x, y) = 7
SOLUTION The graph of the function is the horizontal plane z = 7. The plane z = 7 is the only horizontal trace. The
vertical traces obtained by setting x = a or y = a are the lines z = 7 in these planes.
27. f (x, y) = x + |y|
SOLUTION The graph of f is shown in the figure:
The vertical traces obtained by setting x = a or y = a are z = a + |y| or z = x + |a| in the planes x = a and y = a,
respectively.
z z
y y
x x
28. f (x, y) = 9 x 2 y 2
SOLUTION The graph of the function is shown in the figure:
9
6 2
z
3
1
0
2 0 y
1
0 1
x 1
2 2
The horizontal trace at height c is 9 x 2 y 2 = c or x 2 + y 2 = 9 c. For c < 9 it is a circle of radius 9 c centered
at the origin, and for c = 9 it is the origin (0, 0).
2
y
0
2
0
x
2
The vertical traces in the planes x = a and y = a are the parabolas z = (9 a 2 ) y 2 and z = (9 a 2 ) x 2 in these
planes, respectively.
614 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
2 2
y y
0 0
2 2
5 5
z 0 z 0
5 5
2 2
0 0
x x
2 2
29. f (x, y) = 1 x 2 y2
SOLUTION The graph of f is shown in the figure below:
x
The vertical traces in the planes x = a and y = a (for |a| 1) are z = 1 a 2 y 2 and z = 1 a 2 x 2 in these
planes.
S E C T I O N 15.1 Functions of Two or More Variables (ET Section 14.1) 615
z z
y y
x x
30. f (x, y) = 1 + x 2 + y 2
SOLUTION The graph of the function is shown in the figure:
3
z
1
2
2 0
0 y
x 2 2
The vertical traces in the planes x = a and y = a are z = 1 a y and z = 1 a x in these planes, respectively.
0.5 0.5
0.25 0.25
y y
0 0
0.25 0.25
0.5 0.5
1.25
1.25
1
1
z 0.75
z 0.75
0.5
0.5
0.25
0.4 0.4
0.2 0.2
x 0 x 0
0.2 0.2
616 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
31. Sketch the contour maps of f (x, y) = x + y with contour intervals m = 1 and 2.
SOLUTION The level curves are x + y = c or y = c x. Using contour interval m = 1, we plot y = c x for various
values of c.
4
2
0
2
4
4 2 0 2 4
4
2
0
2
4
4 2 0 2 4
32. Sketch the contour map of f (x, y) = x 2 + y 2 with level curves c = 0, 4, 8, 12, 16.
SOLUTION The level curves are x 2 + y 2 = c for c 0. We sketch the level curves c = 0, 4, 8, 12, 16:
4
2
0
2
4
4 2 0 2 4
The function f (x, t) = t 1/2 ex /t , whose graph is shown in Figure 22, models the temperature along a
2
33.
metal bar after an intense burst of heat is applied at its center point.
(a) Sketch the graphs of the vertical traces at times t = 0.5, 1, 1.5, 2. What do these traces tell us about the way heat
diffuses through the bar?
(b) Sketch the vertical trace at x = 0. Describe how temperature varies in time at the center point.
(c) Sketch the vertical trace x = c for c = 0.2, 0.4. Describe how temperature varies in time at points near the
center.
Temperature T
4
3
0.40.2 2 Time t
0 0.2
0.4 1
x
Metal bar
SOLUTION
(a) The vertical traces at times t = 0.5, 1, 1.5, 2 are
z = 2e2x in the plane t = 0.5
2
z = ex in the plane t = 1
2
4 4
3 3
2 2
1 1
4 4
3 3
2 2
1 1
Trace at t = 1.5 Trace at t = 2
At each time the temperature decreases as we move away from the center point. Also, as t increases, the temperature at
each point in the bar (except at the middle) increases and then decreases (as can be seen in Figure 22).
(b) The vertical trace at x = 0 is z = t 1/2 = 1 . This trace is shown in the following figure:
t
y
x
t
As suggested by the graph, the temperature is decreasing with time at the center point, according to 1 .
t
(c) The vertical traces x = c for the given values of c are:
0.04
z = 1 e t in the planes x = 0.2 and x = 0.2
t
0.16
z = 1 e t in the planes x = 0.4 and x = 0.4.
t
We see that for small values of t the temperature increases quickly and then slowly decreases as t increases.
T
In Exercises 3443, draw a contour map of f (x, y) with an appropriate contour interval, showing at least six level
curves.
y
34. f (x, y) =
x
SOLUTION The level curves are xy = c or y = cx. We plot y = cx for c = 2, 1, 0, 1, 2, 3 using contour interval
m = 1:
618 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
0.5
0
0.2 0.4 0.6 0.8 1
0.5
c = e3 c = e2 c=e
0.1
0.05
0
0 0.05 0.1
y
36. f (x, y) = 2
x
y
SOLUTION The level curves are 2 = c or y = cx 2 . We use the contour interval m = 2 and plot y = cx 2 for c = 4,
x
2, 0, 2, 4, 6. For c = 0 these are parabolas.
1
0.5
0.5
1
1 0.5 0 0.5 1
37. f (x, y) = x y
c
SOLUTION The level curves are x y = c or y = x . These are hyperbolas in the x y-plane. We draw a contour map of
the function using contour interval m = 1 and c = 0, 1, 2, 3:
4
2
0
2
4
4 2 0 2 4
38. f (x, y) = x y
SOLUTION The level curves are the lines x y = c or y = x c. We plot a contour map using contour interval m = 2
and c = 4, 2, 0, 2, 4, 6:
4
2
0
2
4
4 2 0 2 4
S E C T I O N 15.1 Functions of Two or More Variables (ET Section 14.1) 619
39. f (x, y) = x + 2y 1
SOLUTION The level curves are the lines x + 2y 1 = c or y = x2 + c+1 2 . We draw a contour map using the
contour interval m = 4 and c = 9, 5, 1, 3, 7, 11. The corresponding level curves are:
x x x x
y= 4, y= 2, y= , y= + 2,
2 2 2 2
c=9 c=5 c=1 c=3
x x
y = + 4, y = +6
2 2
c=7 c=11
4
2
0
2
4
4 2 0 2 4
40. f (x, y) = x 2 y
SOLUTION The level curves are the parabolas y = x 2 + c. We draw a contour plot with contour interval m = 1, for
c = 0, 1, 2, 3, 4, 5:
6
5
4
3
2
1
3 2 1 0 1 2 3
41. f (x, y) = x 2
SOLUTION The level curves are x 2 = c. For c > 0 these are the two vertical lines x = c and x = c and for
c = 0 it is the y-axis. We draw a contour map using contour interval m = 4 and c = 0, 4, 8, 12, 16, 20:
4
2
0
2
4
4 2 0 2 4
42. f (x, y) = 3x 2 y 2
SOLUTION The level curves are the hyperbolas 3x 2 y 2 = c, c = 0, and for c = 0 it is the two lines y = 3x. We
plot a contour map with contour interval m = 2 using c = 4, 2, 0, 2, 4, 6:
2
2
2 1 0 1 2
10
43. f (x, y) =
1 + x 2 + y2
SOLUTION The level curves are:
10 10 10
=c 1 + x 2 + y2 = x 2 + y2 = 1
1 + x 2 + y2 c c
For 0 < c < 10 these are circles of radius 10c 1 centered at the origin. For c = 10 it is the point at the origin. We
plot a contour map with contour interval m = 1.5 for c = 10, 8.5, 7, 5.5, 4, 2.5.
620 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
C
470
B
450
i
ii iii
400 E
P iv
D
A
44. Match the contour maps (A) and (B) in Figure 23 with the two functions f (x, y) = x 2y and g(x, y) = 2x y.
c = 2
c=0
y y c=2
c = 2
2 c=0 2
1 c=2
x x
2 2 2 1 1 2
1
2
(A) (B)
FIGURE 23
x c
SOLUTION The level curves of the function f (x, y) = x 2y are the lines x 2y = c or y = 2 2 . The level
curves of g(x, y) = 2x y are the lines 2x y = c or y = 2x c. The slope of the lines in the contour map of g is
greater than the slope in the contour map of f . Therefore (A) is a contour map of f and (B) is a contour map of g.
45. Which linear function has the contour map shown in Figure 24 (with level curve c = 0 as indicated),
assuming that the contour interval is m = 6? What if m = 3?
2
1
x
6 3 3 6
1
2
c=0
FIGURE 24
f (x, y) = x + y + (1)
x + y + = c (2)
By the given information, the level curve for c = 0 is the line passing through the points (0, 1) and (3, 0). We find
the equation of this line:
0 (1) 1
y= (x + 3) y = (x + 3) x + 3y + 3 = 0
3 0 3
Setting c = 0 in (2) gives x + y + = 0. Hence,
= = = 3 , = 3
1 3 3
Substituting into (2) gives
x + 3 y + 3 = c
or
c
x + 3y + 3 = (3)
S E C T I O N 15.1 Functions of Two or More Variables (ET Section 14.1) 621
Case 1: m = 6. In this case, the closest line above c = 0 corresponds to c = 6. This line is parallel to x + 3y + 3 = 0
and passes through the origin, hence its equation is x + 3y = 0. Setting c = 6 in (3) gives
6 6
x + 3y + 3 = or x + 3y = 3
This line coincides with the line x + 3y = 0 only if 6 3 = 0, that is, = 2. Substituting = 2, = 3 = 6, and
= 3 = 6 in (1) gives
f (x, y) = 2x + 6y + 6
Case 2: m = 3. In this case the closest line above c = 0 corresponds to c = 3. Substituting c = 3 in (3) gives the level
curve x + 3y + 3 = 3 or x + 3y = 3 3. This line passes through the origin if 3 3 = 0, that is, = 1. Hence,
= 1, = 3 = 3, = 3 = 3. Substituting in (1) gives
f (x, y) = x + 3y + 3
In Exercises 4649, f (S, T ) denotes the density of seawater at salinity level S (parts per thousand) and temperature T
(degrees Celsius). Refer to the contour map of f (S, T ) in Figure 25.
Salinity (ppt)
31.5 32.0 32.5 33.0 33.5 34.0 34.5 35.0
25 25
30
1.02
35
20 1.02 0 20
Temperature T C
Temperature T C
4
1 2 45
. 0
A 1.02
15 50 15
1.02 55
1.02
260
10 1.0 10
2 65
1.0 B
0
5 27 5
1.0
C D
0 0
31.5 32.0 32.5 33.0 33.5 34.0 34.5 35.0
Salinity (ppt)
FIGURE 25 Contour map of seawater density f (S, T ) (kg/m3 ).
46. Calculate the average ROC of density with respect to temperature from C to A.
SOLUTION The segment C A spans 5 level curves and the contour interval is 0.0005. Since the density is decreasing in
the direction from C to A, the change in density is d = 0.0005 5 = 0.0025 kg/m3 . The temperature at A is 17 C
and at C is 2 C, so the difference in temperature from C to A is T = 17 2 = 15 C. Hence,
d 0.0025
Average ROC from C to A = = = 0.000167 kg/m3 C.
T 15
47. Calculate the average ROC of density with respect to salinity from C to D.
SOLUTION For fixed temperature, the segment C D spans one level curve and the level curve of D is to the right of the
level curve of C. Therefore, the change in density from C to D is d = 0.0005 kg/m3 . The salinity at D is greater than
the salinity at C and s = 0.8 ppt. Therefore,
d 0.0005
Average ROC from C to D = = = 0.000625 kg/m3 ppt.
s 0.8
48. At a fixed level of salinity, is seawater density an increasing or decreasing function of temperature?
SOLUTION The level of salinity is fixed on each vertical line. The vertical lines intersect level curves with decreasing
values in the direction of increasing temperature (which is the upward direction). Therefore, at a fixed level of salinity,
seawater density is a decreasing function of temperature.
49. Does water density appear to be more sensitive to a change in temperature at point A or point B?
SOLUTION The two adjacent level curves are closer to the level curve of A than the corresponding two adjacent level
curves are to the level curve of B. This suggests that water density is more sensitive to a change in temperature at A than
at B.
70
Average ROC from A to B = 0.047.
1500
The change in elevation from A to C is obtained by multiplying the number of level curves between A and C, which is 8,
by the contour interval 10 meters, giving 8 10 = 80 m. Using the scale, we approximate the distance AC by 3500 km.
Therefore,
80
Average ROC from A to C = 0.023.
3500
52. Estimate the average ROC from P to points i, ii, iii, and iv.
SOLUTION The points i, ii, and iii are on a level curve adjacent to the level curve of P, hence the change in elevation
is 10 meters. Using the scale we approximate the distances from P to the points i, ii, and iii:
From P to i: 400 m
From P to ii: 400 m
From P to iii: 640 m
Therefore,
10 = 0.025
Average ROC from P to i 400
10 = 0.025
Average ROC from P to ii 400
10 0.016
Average ROC from P to iii 640
Point iv is on the level curve of P, hence there is no change in elevation from P to iv, and the average ROC is zero:
0
Average ROC from P to iv = = 0.
Distance from P to iv
C
470
B
450
i
ii iii
400 E
P iv
D
A
SOLUTION Starting at D or E, we draw a path that everywhere along the way points on the steepest direction, that is,
moves as straight as possible from one level curve to the next. We obtain the following paths:
C
470
B
450
i
ii iii
400 E
P iv
D
A
Contour interval = 10 m
0 1 2 km
(b) At which of (A)(E) is temperature decreasing most rapidly in the northern direction?
(c) In which direction at (B) is temperature increasing most rapidly?
SurfaceTemperature (F) 12Z Sun Sep 18 2005
50 45 45
50
45
40 55
40 60
55
45 B C 60
40
50 55
50
65
D 60
A
40 70
65
55
70
60 E
70 75 75
75
80
80
WW2010 (http://ww2010.atmos.uiuc.edu/)
Atmospheric Sciences, University of Illinois at Urbana-Champaign
FIGURE 27
SOLUTION Ranked according to temperature, we have D (the lowest), A and B (equal temperature), C, and E (the
highest). San Francisco appears to be at about 52.
(a) Increasing to the east at D, B, and C.
(b) Most rapid decrease at point (E), where the contour lines are close together.
(c) Most rapid increase to the south.
r cos
f (r, ) = = cos .
r
z
x
y
The level curves are the curves cos = c in the r -plane, for |c| 1. For 1 < c < 1, c = 0, the level curves cos = c
are the two rays = cos1 c and = cos1 c.
z
y
624 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
For c = 0, the level curve cos = 0 is the y-axis; for c = 1 the level curve cos = 1 is the nonnegative x-axis.
z
x
y
(b) Show that the level curve is empty if |c| > 12 and is a hyperbola if |c| < 12 . What is the level curve for c = 12 ?
xy
SOLUTION The graph of f (x, y) = is shown below:
x 2 +y 2 +1
z
y
x 2 + y2
x 2 + y 2 2|x y| |x y|
2
Therefore, (1) gives
x 2 +y 2
xy |x y| 1 x 2 + y2 1 1
|c| = 2 = 2 = < 1=
x + y2 + 1 x 2 + y2 + 1 x 2 + y2 + 1 2 x 2 + y2 + 1 2 2
That is, |c| < 12 . It follows that for c 12 the level curve is empty.
y
Let t = y and s = x 2c . Then for |c| < 12 the equation of the level curve (2) can be rewritten as
1
1 t 2 s2 = 1
4c2
t2 2
2 s = 1
2c
14c2
This is the equation of a hyperbola in standard position in the ts-plane. The equation (2) for |c| < 12 is the equation of a
hyperbola obtained by a change of variables for a hyperbola in standard position, in the x y-plane. For c = 0.5 the level
curve is empty:
z
x
626 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
Since the limit exists, we may compute it by approaching (2, 3) along the vertical line x = 2. This gives
SOLUTION We show that the limit lim(x,y)(0,0) f (x, y) does not exist. Indeed, if the limit exists, it may be computed
by approaching (0, 0) along the x-axis or along the y-axis. We compute these two limits:
Since the limits are different, f (x, y) does not approach one limit as (x, y) (0, 0), hence the limit lim(x,y)(0,0) f (x, y)
does not exist.
Exercises
In Exercises 110, use continuity to evaluate the limit.
1. lim (x 2 + y)
(x,y)(1,2)
lim (x 2 + y) = 12 + 2 = 3
(x,y)(1,2)
x
2. lim
(x,y)( 94 , 29 ) y
S E C T I O N 15.2 Limits and Continuity in Several Variables (ET Section 14.2) 627
SOLUTION The function xy is continuous at the point 49 , 29 , hence we compute the limit by substitution:
4
x
lim = 29 = 2
(x,y) 4 2 y
9,9 9
3. lim (3x 2 y 2x y 3 )
(x,y)(3,1)
SOLUTION The function 3x 2 y 2x y 3 is continuous everywhere because it is a polynomial. We evaluate the limit by
substitution:
x 2 3y 2
4. lim
(x,y)(2,4) 4x + y
3y 2 2
SOLUTION The function x4x+y is a rational function and the denominator is not zero at the point (2, 4). Therefore
the function is continuous at (2, 4) and we may use substitution to compute the limit:
x 2 3y 2 (2)2 3 42
lim = = 11
(x,y)(2,4) 4x + y 4 (2) + 4
x2 + y
5. lim
(x,y)(1,2) x + y 2
The rational function x +y2 is continuous at (1, 2) since the denominator is not zero at this point. We
2
SOLUTION
x+y
compute the limit using substitution:
x2 + y 12 + 2 3
lim = =
(x,y)(1,2) x + y 2 1 + 22 5
y
7. lim arctan
(x,y)(4,4) x
SOLUTION Since the function arctan xy is continuous at the point (4, 4), we compute the limit by substitution:
y 4
lim arctan = arctan = arctan 1 =
(x,y)(4,4) x 4 4
e x y
2 2
8. lim
(x,y)(0,0)
e x y = e0 0 = e0 = 1
2 2 2 2
lim
(x,y)(0,0)
e x ey
2 2
9. lim
(x,y)(1,1) x+y
SOLUTION The function is the quotient of two continuous functions, and the denominator is not zero at the point (1, 1).
Therefore, the function is continuous at this point, and we may compute the limit by substitution:
e x ey e1 e1
2 2 2 2
e 1e 1
lim = = = (e e1 )
(x,y)(1,1) x+y 1+1 2 2
628 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
In Exercises 1118, evaluate the limit or determine that it does not exist.
x
11. lim
(x,y)(0,1) y
SOLUTION The function xy is continuous at the point (0, 1), hence we compute the limit by substitution:
x 0
lim = =0
(x,y)(0,1) y 1
x
12. lim
(x,y)(1,0) y
SOLUTION We examine the limit as (x, y) approaches (1, 0) along the line x = 1:
x 1
lim = lim
(x,y)(1,0) y y0 y
along x=1
Because this limit does not exist, the limit lim(x,y)(1,0) xy does not exist.
13. lim e x y ln(1 + x y)
(x,y)(1,1)
SOLUTION The function is continuous at (1, 1), hence we compute the limit using substitution:
x 2 + y2
14. lim
(x,y)(0,0) 1 + y 2
The function x +y2 is continuous everywhere since it is a rational function whose denominator is never
2 2
SOLUTION
1+y
zero. We evaluate the limit using substitution:
x 2 + y2 02 + 02
lim = =0
(x,y)(0,0) 1 + y 2 1 + 02
SOLUTION The function x 2 |y|3 is continuous everywhere. We compute the limit using substitution:
x y2
16. lim
(x,y)(1,2) |x|
2
SOLUTION The function x|x|
y
is the quotient of two continuous functions and the denominator is not zero at the point
(1, 2). Therefore, the function is continuous at (1, 2) and we may use substitution to compute the limit:
x y2 (1)(2)2
lim = = 4
(x,y)(1,2) |x| |1|
y2
17. lim
(x,y)(4,2) x2 4
SOLUTION The function is continuous at the point (4, 2), since it is the quotient of two continuous functions and the
denominator is not zero at (4, 2). We compute the limit by substitution:
y2 22 0
lim = = =0
(x,y)(4,2) x 4
2 4 4
2 12
S E C T I O N 15.2 Limits and Continuity in Several Variables (ET Section 14.2) 629
sin x
18. lim
(x,y)( ,0) sin y
SOLUTION We examine the limits as (x, y) approaches ( , 0) along the line x = and along the line y = x :
sin x sin 0
lim = lim = lim = lim 0 = 0
(x,y)( ,0) sin y y0 sin y y0 sin y y0
along x=
The two limits are different, therefore the given limit does not exist.
SOLUTION Using the Sum Law and the Constant Multiples Law we get
SOLUTION e f (x,y) is the composition of G(u) = eu and u = f (x, y). Since eu is continuous, we may evaluate the
limit as follows:
22. lim ln g(x, y) 2 f (x, y)
(x,y)(2,5)
SOLUTION The function is the composition G F, where G(u) = ln u and F(x, y) = g(x, y) 2 f (x, y). We first
use the Limit Laws to evaluate the limit:
lim F(x, y) = lim g(x, y) 2 f (x, y) = lim g(x, y) 2 lim f (x, y)
(x,y)(2,5) (x,y)(2,5) (x,y)(2,5) (x,y)(2,5)
=723=1
Since G(u) = ln u is continuous at the point u = 1, we may evaluate the limit as follows:
lim ln g(x, y) 2 f (x, y) = lim ln F(x, y) = ln lim F(x, y) = ln 1 = 0
(x,y)(2,5) (x,y)(2,5) (x,y)(2,5)
In Exercises 2332, evaluate the limit or determine that the limit does not exist. You may evaluate the limit of a product
function as a product of limits as in Example 3.
(sin x)(sin y)
23. lim
(x,y)(0,0) xy
SOLUTION We evaluate the limit as a product of limits:
(sin x)(sin y) sin x sin y
lim = lim lim = 1 1 = 1.
(x,y)(0,0) xy x0 x y0 y
e x y
2 2
24. lim
(x,y)(2,1)
630 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
Notice that since e x y is continuous everywhere, we may evaluate the limit by substitution:
2 2
e x y = e2 1 = e3 .
2 2 2 2
lim
(x,y)(2,1)
1
26. lim sin x cos
(x,y)(0,0) y
SOLUTION Note that cos 1y is always between 1 and 1. Thus, as (x, y) gets close to (0, 0) (or to any other number),
we have that sin x sin x cos 1y sin x, and since both sin x and sin x go to 0 as x goes to 0, we use the Squeeze
Theorem to conclude that
1
lim sin x cos = 0.
(x,y)(0,0) y
(2 + k)2 4
27. lim h4
(h,k)(2,0) k
SOLUTION We write the limit as a product of limits:
We show that this limit exists. Notice that the limit is the derivative of f (x) = x 2 at x = 2. That is,
(2 + k)2 4 f (2 + k) f (2)
lim = lim = f
(2) = 2x =4
k0 k k0 k x=2
Combining with (1), we conclude that the given limit is 4 4 = 16.
h(k 2 + 4)
28. lim
(h,k)(0,0) k
SOLUTION We compute the limit as (h, k) approaches (0, 0) along the lines h = 0 and h = k, respectively. This gives
h(k 2 + 4) 0(k 2 + 4)
lim = lim = lim 0 = 0
(h,k)(0,0) k k0 k k0
along h=0
h(k 2 + 4) k(k 2 + 4)
lim = lim = lim (k 2 + 4) = 02 + 4 = 4
(h,k)(0,0) k k0 k k0
along h=k
Since the two limits are different, the limit itself does not exist.
1
29. lim e1/x tan1
(x,y)(0,0) y
SOLUTION We show that the limit along the line y = x does not exist. We have,
1 1
lim e1/x tan1 = lim e1/x tan1 (1)
(x,y)(0,0) y x0 x
along y=x
We consider the two one-sided limits. As x 0+, e1/x is increasing without bound and lim tan1 1x = 2 . Therefore,
x0+
1
lim e1/x tan1 = +
x0+ x
S E C T I O N 15.2 Limits and Continuity in Several Variables (ET Section 14.2) 631
As x 0, we have
1 1
lim e1/x tan1 = lim e1/x lim tan1 =0 =0
x0 x x0 x0 x 2
Since the one-sided limits are not equal, the limit in (1) does not exist. Therefore, the given limit does exist.
ek 1
30. lim x2
(x,k)(1,0) k
First notice that the limit lim e k1 is the derivative of f (u) = eu at u = 0. That is,
k
SOLUTION
k0
ek 1 e0+k e0 d u
lim = lim = e = eu = e0 = 1
k0 k k0 k du u=0 u=0
Since the limit exists, we may use the Product Rule to obtain
2 ek 1 2 ek 1
lim x = lim x lim = 12 1 = 1
(x,k)(1,0) k x1 k0 k
31. lim x ln y
(x,y)(0,0)
SOLUTION We first compute the limit as (x, y) approaches the origin along the positive y-axis (x = 0, y > 0):
We now compute the limit as (x, y) approaches the origin along the curve y = e1/x , x > 0 (notice that y 0+ as
x 0+):
1
lim x ln y = lim x ln e1/x = lim x = lim (1) = 1
(x,y)(0,0) x0+ x0+ x x0+
along y=e1/x , x>0
Since the limits along the two paths are not equal, the limit itself does not exist.
1 1
32. lim
(x,y)(0,0) 2x y x y(x + 2)
SOLUTION We rewrite the function as follows:
1 1 1 1 1 1 x +22 x 1
= = = =
2x y x y(x + 2) y 2x x(x + 2) y 2x(x + 2) y 2x(x + 2) 2y(x + 2)
Hence we consider the limit
1
lim
(x,y)(0,0) 2y(x + 2)
We show that the limit along the y-axis does not exist. We have
1 1 1
lim = lim = lim
(x,y)(0,0) 2y(x + 2) y0 2y(0 + 2) y0 4y
1 does not exist, hence the given limit does not exist.
The limit lim 4y
y0
x 3 + y3
33. Let f (x, y) = 2 .
x + y2
(a) Show that
|x 3 | |x|(x 2 + y 2 ), |y 3 | |y|(x 2 + y 2 )
|x 3 + y 3 | |x 3 | + |y 3 |
| f (x, y)| =
x +y
2 2 x 2 + y2
We continue using the inequality in part (a):
Let > 0. Then if |x| < 2 and |y| < 2 , we have by (1)
| f (x, y) 0| |x| + |y| < + = (2)
2 2
Notice that if x 2 + y 2 < 4 , then x 2 < 4 and y 2 < 4 . Hence |x| < 2 and |y| < 2 , so (1) holds. In other words, using
2 2 2
D 2 to represent the punctured disc of radius /2 centered at the origin, we have
(x, y) D |x| <
2 2
and
|y| < | f (x, y) 0| <
2
We conclude by the limit definition that
lim f (x, y) = 0
(x,y)(0,0)
x = r cos , y = r sin
Then, (x, y) (0, 0) if and only if x 2 + y 2 0, that is, if and only if r 0+. We have
x 2 + y2 = r 2
x 3 + y 3 = r 3 cos3 + r 3 sin3 = r 3 (cos3 + sin3 )
Hence,
x 3 + y3 r 3 (cos3 + sin3 )
f (x, y) = 2 = = r (cos3 + sin3 ) (3)
x + y2 r2
As r approaches zero, 2r approaches zero, hence (4) and the Squeeze Theorem imply that
y2
34. Show that lim does not exist.
(x,y)(0,0) x 2 + y 2
2
SOLUTION We compute the limit of f (x, y) = 2y 2 as (x, y) approaches (0, 0) along the line y = mx. Substituting
x +y
y = mx we get
(mx)2 m2x 2 m2
f (x, mx) = = 2 =
x 2 + (mx)2 x (1 + m )
2 1 + m2
We see that f (x, y) is constant along every line y = mx through the origin, hence
m2
lim f (x, mx) =
x0 1 + m2
This limit depends on the value of m, so there is no single value L such that | f (x, y) L| is small for all (x, y) near
(0, 0). Hence the limit does not exist.
x
35. Show that lim does not exist. Hint: Consider the limits along the lines y = mx.
(x,y)(0,0) x + y
SOLUTION We compute the limit as (x, y) approaches the origin along the line y = mx, for a fixed value of m.
x , we get
Substituting y = mx in the function f (x, y) = x+y
x x
f (x, mx) = =
x + mx x(1 + m)
As (x, y) approaches (0, 0), (x, y) = (0, 0). Therefore x = 0 on the line y = mx. Thus,
x 1 1
lim f (x, y) = lim = lim =
(x,y)(0,0) x0 x(1 + m) x0 1 + m 1+m
along y=mx
We see that the limits along the lines y = mx are different, hence f (x, y) does not approach one limit as (x, y) (0, 0).
We conclude that the given limit does not exist.
x
36. Show that lim does not exist. Hint: Use polar coordinates.
(x,y)(0,0) x 2 + y 2
SOLUTION We use polar coordinates x = r cos , y = r sin . Then x 2 + y 2 = r 2 and (x, y) (0, 0) if and only if
r 0+. Substituting the polar coordinates and evaluating the limit gives
x r cos r cos
lim = lim = lim = lim cos
(x,y)(0,0) x 2 + y2 r 0+ r 2 r 0+ r r 0+
doesnot approach one limit, since the limit depends on the angle . For instance, approaching
We see that the function
along the lines y = x = 4 and y = 0( = 0) gives different limits. Hence the limit does not exist.
x a yb
37. Let a, b 0. Show that lim = 0 if a + b > 2 and the limit does not exist if a + b 2.
(x,y)(0,0) x 2 + y 2
SOLUTION We first show that the limit is zero if a + b > 2. We compute the limit using the polar coordinates
x = r cos , y = r sin . Then (x, y) (0, 0) if and only if x 2 + y 2 0, that is, if and only if r 0+. Therefore,
Since a + b > 2, lim r a+b2 = 0, therefore (2) and the Squeeze Theorem imply that
r 0+
x a yb
lim =0
(x,y)(0,0) x 2 + y 2
634 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
We now consider the case a + b < 2. We examine the limit as (x, y) approaches the origin along the line y = x. Along
this line, = 4 , therefore (1) gives
1 a 1 b
x a yb
= a+b2 cosa sinb = lim a+b2 =
r a+b2
lim lim r r lim
(x,y)(0,0) x 2 + y 2 r 0+ 4 4 r 0+ 2 2 r 0+ ( 2)a+b
Since a + b < 2, we have a + b 2 < 0 therefore lim r a+b2 does not exist. It follows that if a + b < 2, the given
r 0+
limit does not exist. Finally we examine the case a + b = 2. By (1) we get
x a yb
lim = lim (r 0 cosa sinb ) = lim cosa sinb = cosa sinb
(x,y)(0,0) x 2 + y 2 r 0+ r 0+
We see that the function does not approach one limit. For example, approaching the origin along the lines y = x (i.e.,
a+b
= 4 ) and y = 0 (i.e., = 0) gives two different limits cosa 4 sinb 4 = 22 and cosa 0 sinb 0 = 0. We conclude
that if a + b = 2, the limit does not exist.
x 2 + y2
38. Evaluate lim .
(x,y)(0,0) x 2 + y2 + 1 1
SOLUTION We rewrite the function by dividing and multiplying it by the conjugate of x 2 + y 2 + 1 1 and using
the identity (a b)(a + b) = a 2 b2 . This gives
2 2 (x 2 + y2) x 2 + y2 + 1 + 1 (x 2 + y2) x 2 + y2 + 1 + 1
x +y
= =
x 2 + y2 + 1 1 x 2 + y2 + 1 1 x 2 + y2 + 1 + 1 x 2 + y2 + 1 1
(x 2 + y 2 ) x 2 + y2 + 1 + 1
= = x 2 + y2 + 1 + 1
x 2 + y2
The resulting function is continuous, hence we may compute the limit by substitution. This gives
x 2 + y2
lim = lim x 2 + y2 + 1 + 1 = 02 + 02 + 1 + 1 = 2
(x,y)(0,0) x 2 + y2 + 1 1 (x,y)(0,0)
39. Figure 8 shows the contour maps of two functions with contour interval m = 2. Explain why the limit
lim f (x, y) does not exist. Does lim g(x, y) appear to exist in (B)? If so, what is its limit?
(x,y)P (x,y)P
8 4
12
8
P P
4
0
0
4
SOLUTION As (x, y) approaches arbitrarily close to P, the function f (x, y) takes the values 0, 4, and 8. Therefore
f (x, y) does not approach one limit as (x, y) P. Rather, the limit depends on the contour along which (x, y) is
approaching P. This implies that the limit lim(x,y)P f (x, y) does not exist. In (B) the limit lim(x,y)P g(x, y) appears
to exist. If it exists, it must be 6, which is the level curve of P.
y ln(1 + x)
ln f (x, y) = ln (1 + x) y/x = ln(1 + x) = y (1)
x x
Using LHopitals Rule we have
1
ln(1 + x) 1 1
lim = lim 1+x = lim = =1
x0 x x0 1 x0 1 + x 1+0
S E C T I O N 15.2 Limits and Continuity in Several Variables (ET Section 14.2) 635
Since this limit exists, we may use the Product Rule to compute the limit of (1):
ln(1 + x)
lim ln f (x, y) = lim y lim =21=2 (2)
(x,y)(0,2) y2 x0 x
ln u approaches 2 if and only if u is approaching e2 . Therefore, the limit in (2) implies that
lim f (x, y) = e2 .
(x,y)(0,2)
sin(x y)
41. The function f (x, y) = is defined for x y = 0.
xy
(a) Is it possible to extend the domain of f (x, y) to all of R2 so that the result is a continuous function?
(b) Use a computer algebra system to plot f (x, y). Does the result support your conclusion in (a)?
SOLUTION
(a) We define f (x, y) on the x- and y-axes by f (x, y) = 1 if x y = 0. We now show that f is continuous. f is continuous
at the points where x y = 0. We next show continuity at (x 0 , 0) (including x 0 = 0). For the points (0, y0 ), the proof is
similar and hence will be omitted. To prove continuity at P = (x 0 , 0) we have to show that
sin x y
lim f (x, y) = lim =1 (1)
(x,y)P (x,y)P xy
x y
The graph shows that, near the axes, the values of f (x, y) are approaching 1, as shown in part (a).
xy xy
42. Repeat Exercise 41 for the functions f (x, y) = and g(x, y) = 2 .
x+y x + y2
xy
SOLUTION For f (x, y) = :
x+y
xy
(a) The function f (x, y) = x+y is defined for y = x. We show that f cannot be extended continuously to all of R 2 ,
by showing that lim(x,y)(1,1) f (x, y) does not exist. We examine the limit along the line y = x + 2.
x(x + 2) x(x + 2)
lim f (x, y) = lim = lim (1)
(x,y)(1,1) x1 x + x + 2 x1 2(x + 1)
along y=x+2
It follows that the limit in (1) does not exist, hence also the limit lim f (x, y) does not exist. Similarly one can
(x,y)(1,1)
show that lim f (x, y) does not exist for x 0 = 0. We conclude that f cannot be extended continuously to all of
(x,y)(1,1)
R2 .
xy
(b) The following figure shows the graph of f (x, y) = x+y :
636 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
0 z
2
1 2
0
x 1
1 2
2 1 0
2 y
0.5
0.25
z 0
0.25
2
0.5 1
0
2 1 1 x
0 1 2
y 2
x2 y
43. The function f (x, y) = 4 provides an interesting example where the limit as (x, y) (0, 0) does
x + y2
not exist, even though the limit along every line y = mx exists and is zero (Figure 9).
(a) Show that the limit along any line y = mx exists and is equal to 0.
(b) Calculate f (x, y) at the points (101 , 102 ), (105 , 1010 ), (1020 , 1040 ). Do not use a calculator.
(c) Show that lim f (x, y) does not exist. Hint: Compute the limit along the parabola y = x 2 .
(x,y)(0,0)
SOLUTION
2
(a) Substituting y = mx in f (x, y) = 4x y 2 , we get
x +y
x 2 mx mx 3 mx
f (x, mx) = = = 2
x + (mx)
4 2 x 2 (x 2 + m 2 ) x + m2
We compute the limit as x 0 by substitution:
mx m0
lim f (x, mx) = lim = 2 =0
x0 x0 x 2 + m 2 0 + m2
(b) We compute f (x, y) at the given points:
(c) We compute the limit as (x, y) approaches the origin along the parabola y = x 2 (by part (b), the limit appears to be
1 ). We substitute y = x 2 in the function and compute the limit as x 0. This gives
2
x2 x2 x4 1 1
lim f (x, y) = lim f (x, x 2 ) = lim 2
= lim 4
= lim =
(x,y)0 x0 x0 x 4 + (x 2 ) x0 2x x0 2 2
along y=x 2
However, in part (a), we showed that the limit along the lines y = mx is zero. Therefore f (x, y) does not approach one
limit as (x, y) (0, 0), so the limit lim f (x, y) does not exist.
(x,y)(0,0)
44. Is the following function continuous?
x 2 + y2 if x 2 + y 2 < 1
f (x, y) =
1 if x 2 + y 2 1
x2 y
FIGURE 9 Graph of f (x, y) = .
x 4 + y2
SOLUTION f (x, y) is defined by a polynomial in the domain x 2 + y 2 < 1, hence f is continuous in this domain. In
the domain x 2 + y 2 > 1, f is a constant function, hence f is continuous in this domain also. Thus, we must examine
continuity at the points on the circle x 2 + y 2 = 1.
y
1
y2 + x2
x
0
Since lim f (r, ) = lim r 2 = 1 and lim f (r, ) = lim 1 = 1, we have lim f (r, ) = 1. Therefore f (r, ) is
r 1 r 1 r 1+ r 1+ r 1
continuous at r = 1, or f (x, y) is continuous on x 2 + y 2 = 1. We conclude that f is continuous everywhere on R 2 .
f x x y = f x yx = f yx x
Exercises
1. Use the limit definition of the partial derivative to verify the formulas
x y2 = y2, x y 2 = 2x y
x y
(x + h)y 2 x y 2 x y 2 + hy 2 x y 2 hy 2
x y 2 = lim = lim = lim = lim y 2 = y 2
x h0 h h0 h h0 h h0
2
2. Use the Product Rule to compute (x + y)(x + y 4 ).
y
SOLUTION Using the Product Rule we obtain
2
(x + y)(x + y 4 ) = (x 2 + y) (x + y 4 ) + (x + y 4 ) (x 2 + y)
y y y
= (x 2 + y) 4y 3 + (x + y 4 ) 1 = 4x 2 y 3 + 5y 4 + x
S E C T I O N 15.3 Partial Derivatives (ET Section 14.3) 639
y
3. Use the Quotient Rule to compute .
y x + y
SOLUTION Using the Quotient Rule we obtain
y (x + y) y (y) y y (x + y) (x + y) 1 y 1 x
= = =
y x + y (x + y)2 (x + y)2 (x + y)2
4. Use the Chain Rule to compute ln(u 2 + uv).
u
SOLUTION d ln = 1 d . Applying this with = u 2 + uv gives
By the Chain Rule du du
1 2 2u + v
ln(u 2 + uv) = 2 (u + uv) = 2
u u + uv u u + uv
f z (x, y, z) = (x yz) = x y
z
Substituting the given point we get
f z (2, 3, 1) = 2 3 = 6
d
sin(cx) = c cos(cx), sin(x y) = y cos(x y)
dx x
d
SOLUTION d x sin(cx) is the derivative of the single-variable function sin(cx), where c is a constant. x sin(x y) is the
partial derivative of the two-variable function sin(x y) with respect to x. While differentiating, the variable y is considered
constant, hence it resembles the first differentiation, and the results are the same where c is replaced by y.
7. The plane y = 1 intersects the surface z = x 4 + 6x y y 4 in a certain curve (Figure 7). Find the slope of the tangent
line to this curve at the point P = (1, 1, 6).
x
y
SOLUTION The slope of the tangent line to the curve z = z(x, 1) = x 4 + 6x 1, obtained by intersecting the surface
z = x 4 + 6x y y 4 with the plane y = 1, is the partial derivative zx (1, 1).
z 4
= (x + 6x y y 4 ) = 4x 3 + 6y
x x
z
m= (1, 1) = 4 13 + 6 1 = 10
x
f f
8. Determine if the partial derivatives and are positive or negative at the point P in Figure 7.
x y
SOLUTION The graph shows that f is increasing in the direction of growing x and f is decreasing in the direction of
growing y. Therefore, xf P > 0 and yf P < 0.
640 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
90 50
y 10
4
50
2
C
B 10 10
0
50
90
10
2 A 130
30
4
x
4 2 0 2 4
FIGURE 8 Contour map of f (x, y).
12. z = x 2 + y 2
SOLUTION We compute z x (x, y) by treating y as a constant, and we compute z y (x, y) by treating x as a constant:
2 2
(x + y 2 ) = 2x; (x + y 2 ) = 2y
x y
13. z = x 4 y 3
SOLUTION Treating y as a constant (to find z x ) and x as a constant (to find z y ) and using Rules for Differentiation, we
get,
4 3 4
(x y ) = y3 (x ) = y 3 4x 3 = 4x 3 y 3
x x
4 3
(x y ) = x 4 (y 3 ) = x 4 3y 2 = 3x 4 y 2
y y
14. z = x 4 y + x y 2
SOLUTION We obtain the following partial derivatives:
4
(x y + x y 2 ) = 4x 3 y + y 2
x
4
(x y + x y 2 ) = x 4 + x (2y 3 ) = x 4 2x y 3
y
15. V = r 2 h
SOLUTION We find rV and h
V:
V
= ( r 2 h) = h (r 2 ) = h 2r = 2 hr
r r r
V
= ( r 2 h) = r 2
h h
x
16. z =
y
SOLUTION Treating y as a constant we have
x 1 1 1
= (x) = 1 =
x y y x y y
We now find the derivative z y (x, y), treating x as a constant:
x 1 1 x
=x =x 2 = 2.
y y y y y y
x
17. z =
xy
SOLUTION We differentiate with respect to x, using the Quotient Rule. We get
x (x y) x (x) x x (x y) (x y) 1 x 1 y
= = =
x xy (x y)2 (x y)2 (x y)2
We now differentiate with respect to y, using the Chain Rule:
x 1 1 1 x
=x =x (x y) = x (1) =
y x y y x y (x y)2 y (x y)2 (x y)2
18. z = 9 x 2 y2
SOLUTION Differentiating with respect to x, treating y as a constant, and using the Chain Rule, we obtain
1 2x x
9 x 2 y2 = (9 x 2 y 2 ) = =
x 2 9 x 2 y2 x 2 9 x 2 y2 9 x 2 y2
We now differentiate with respect to y, treating x as a constant:
1 2y y
9 x 2 y2 = (9 x 2 y 2 ) = =
y 2 9 x y y
2 2 2 9x y
2 2 9 x 2 y2
642 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
x
19. z =
x 2 + y2
SOLUTION We compute zx using the Quotient Rule and the Chain Rule:
1 x 2 + y 2 x x x 2 + y 2 x 2 + y 2 x 2x2 2
z 2 x +y x 2 + y2 x 2 y2
= 2 = = =
x x 2 + y2 (x 2 + y 2 )
3/2
(x 2 + y 2 )
3/2
x 2 + y2
(sin x sin y) = sin y sin x = sin y cos x
x x
(sin x sin y) = sin x sin y = sin x cos y
y y
21. z = sin(u 2 v)
SOLUTION By the Chain Rule,
d d d d
sin = cos and sin = cos .
du du dv dv
Applying this with = u 2 v gives
sin(u 2 v) = cos(u 2 v) (u 2 v) = cos(u 2 v) 2uv = 2uv cos(u 2 v)
u u
sin(u 2 v) = cos(u 2 v) (u 2 v) = cos(u 2 v) u 2 = u 2 cos(u 2 v)
v v
x
22. z = tan
y
SOLUTION By the Chain Rule,
d 1 du d 1 du
tan u = and tan u = .
dx cos2 u d x dy cos2 u d y
(We could also say that the derivative of tan u is sec2 u, but of course sec2 u = 1/ cos2 u, so it really is the same thing.)
We apply this with u = xy to obtain
x 1 x 1 1 1
tan = = =
x y 2
cos y x x y 2
cos y x y ycos xy
2
x 1 x 1 x x
tan = = =
y y 2
cos y x y y 2
cos y x y 2
y cos2 xy
2
d 1 du d 1 du
tan1 u = and tan1 u =
dw 1 + u 2 dw dz 1 + u 2 dz
Using this rule with u = wz gives
dS 1 z
= tan1 (wz) = (wz) =
dw w 1 + (wz) w
2 1 + w2 z 2
dS 1 w
= tan1 (wz) = (wz) =
dz z 1 + (wz)2 z 1 + w2 z 2
S E C T I O N 15.3 Partial Derivatives (ET Section 14.3) 643
24. z = ln(x + y)
SOLUTION We use the Chain Rule to obtain
z 1 1 1
= (x + y) = 1=
x x + y x x+y x+y
z 1 1 1
= (x + y) = 1=
y x + y y x+y x+y
25. z = ln(x 2 + y 2 )
SOLUTION Using the Chain Rule we have
z 1 2 1 2x
= 2 (x + y 2 ) = 2 2x = 2
x x + y2 x x + y2 x + y2
z 1 2 1 2y
= 2 (x + y 2 ) = 2 2y = 2
y x + y 2 y x + y2 x + y2
26. W = er +s
SOLUTION We use the Chain Rule to compute r
W and W :
s
W
= er +s (r + s) = er +s 1 = er +s
r r
W
= er +s (r + s) = er +s 1 = er +s
s s
27. z = e x y
SOLUTION We use the Chain Rule, ddx eu = eu du d u u du
d x ; d y e = e d y with u = x y to obtain
xy
e = e x y (x y) = e x y y = ye x y
x x
xy
e = e x y (x y) = e x y x = xe x y
y y
28. R = ev /k
2
29. z = ex y
2 2
z 2
= ex y (x 2 y 2 ) = ex y (2x) = 2xex y
2 2 2 2 2
x x
z
= ex y (x 2 y 2 ) = ex y (2y) = 2yex y
2 2 2 2 2 2
y y
30. P = e y +z
2 2
y y 2 y2 + z2 y2 + z2
P 2 2z z
= e y +z y 2 + z 2 = e y +z = e y +z
2 2 2 2 2
z z 2 y +z
2 2 y + z2
2
644 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
31. z = y x
SOLUTION To find zy , we use the Power Rule for differentiation:
z
= x y x1
y
z
= y x ln y
x
33. z = sinh(x 2 y)
SOLUTION By the Chain Rule, ddx sinh u = cosh u ddux and ddy sinh u = cosh u du 2
d y . We use the Chain Rule with u = x y
to obtain
sinh(x 2 y) = cosh(x 2 y) (x 2 y) = 2x y cosh(x 2 y)
x x
sinh(x y) = cosh(x y) (x 2 y) = x 2 cosh(x 2 y)
2 2
y y
34. w = x y 2 z 3
SOLUTION The partial derivatives of w are
w
= y2 z3
x
w
= x z 3 (y 2 ) = x z 3 2y = 2x z 3 y
y y
w
= x y 2 (z 3 ) = x y 2 3z 2 = 3x y 2 z 2
z z
x
35. w =
y+z
SOLUTION We have
w x 1 1
= = (x) =
x x y+z y + z x y+z
To find w w
y and z , we use the Chain Rule:
w 1 1 1 x
=x =x (y + z) = x 1=
y y y + z (y + z) y
2 (y + z)2 (y + z)2
w 1 1 1 x
=x =x (y + z) = x 1=
z z y + z (y + z)2 z (y + z)2 (y + z)2
36. Q = r e
SOLUTION The partial derivatives are
Q
= (r e ) = e (r ) = e
r r r
Q
= (r e ) = r (e ) = r e
S E C T I O N 15.3 Partial Derivatives (ET Section 14.3) 645
x
37. w =
(x 2 + y 2 + z 2 )3/2
SOLUTION To find w
x , we use the Quotient Rule and the Chain Rule:
3/2 1/2
w 1 (x 2 + y 2 + z 2 ) x 32 (x 2 + y 2 + z 2 ) 2x (x 2 + y 2 + z 2 ) x 3x
= = (x 2 + y 2 + z 2 )1/2
x (x 2 + y 2 + z 2 )
3 (x 2 + y 2 + z 2 )3
x 2 + y 2 + z 2 3x 2 y 2 + z 2 2x 2
= 5/2
= 5/2
(x 2 + y 2 + z 2 ) (x 2 + y 2 + z 2 )
w 1 3/2
=x = x (x 2 + y 2 + z 2 )
y y (x 2 + y 2 + z 2 )3/2 y
3 5/2 3x y
=x (x 2 + y 2 + z 2 ) 2y =
5/2
2 x + y2 + z2
2
w 1 3/2
=x
= x (x 2 + y 2 + z 2 )
z z x 2 + y 2 + z 2 3/2 z
3 5/2 3x z
=x (x 2 + y 2 + z 2 ) 2z = 5/2
2 (x + y 2 + z 2 )
2
A
= cos(4 9t) (4 9t) = 4 cos(4 9t)
A
= cos(4 9t) (4 9t) = 9 cos(4 9t)
t t
er t
39. U =
r
SOLUTION We have
U ter t r er t 1 (1 + r t)er t
= =
r r2 r2
and also
U r er t
= = er t
t r
f x (x, y) = 6x y + 12x 2 y 2 7y 5
2
f y (x, y) = cos(x 2 y) (x y) = cos(x 2 y) (1) = cos(x 2 y)
y
Evaluating at (0, ) we obtain
1 u
gu (u, v) = (u ln(u + v)) = 1 ln(u + v) + u = ln(u + v) +
u u+v u+v
At the point (1, 2) we have
1 1
gu (1, 2) = ln(1 + 2) + = ln 3 + .
1+2 3
2 3
43. h(x, z) = e x zx z , h z (2, 1)
SOLUTION We first find the partial derivative h z (x, z) using the Chain Rule:
2 3 2 3 2 3
h z (x, z) = e x zx z (x z x 2 z 3 ) = e x zx z (x x 2 3z 2 ) = (x 3x 2 z 2 )e x zx z
z
At the point (2, 1) we get
2 3
44. h(x, z) = e x zx z , h z (1, 0)
SOLUTION In Exercise 43 we obtained the following partial:
2 3
h z (x, z) = (x 3x 2 z 2 )e x zx z
h z (1, 0) = (1 0)e00 = 1.
W W
45. Calculate and , where W = eE/kT .
E T
SOLUTION We use the Chain Rule
d u du d u du
e = eu and e = eu
dE dE dT dT
E , to obtain
with u = kT
W E 1 1 E/kT
= eE/kT = eE/kT
= e
E E kT kT kT
W E/kT E E/kT E 1 E/kT E 1 E E/kT
=e =e =e 2 = e
T T kT k T T k T kT 2
46. According to the ideal gas law, P V = n RT , where P, V , and T are the pressure, volume, and temperature (R and
P P
n are constants). View P as a function of V and T , and compute and .
T V
SOLUTION We differentiate the two sides of the equation P V = n RT with respect to V (treating T as a constant).
Using the Product Rule we obtain
P V P
PV = V +P =V + P; n RT = 0
V V V V V
Hence,
P
V +P =0
V
We substitute P = n RT P
V and solve for V . This gives
P n RT P n RT
V + =0 = 2
V V V V
We now differentiate P V = n RT with respect to T , treating V as a constant:
P
PV = V ; n RT = n R
T T T
S E C T I O N 15.3 Partial Derivatives (ET Section 14.3) 647
Hence,
P P nR
V = nR = .
T T V
V V
47. The volume of a right-circular cone of radius r and height h is V = 3 r 2 h. Calculate and .
r h
SOLUTION We obtain the following derivatives:
V 2 h 2 h 2 hr
= r h = r = 2r =
r r 3 3 r 3 3
V 2 2
= r h = r
h h 3 3
48. A right-circular cone has r = h = 12 cm. What leads to a greater increase in V , a 1-cm increase in r or 1-cm
increase in h? Argue using partial derivatives.
SOLUTION The partial derivatives rV and h
V can be estimated for small values of r and h by
Therefore an increase r = 1 cm in r leads to an increase of rV (12, 12) 1 in the volume, and an increase h = 1 cm
in h leads to an increase of h
V (12, 12) 1 in V . We compute these values, using the partials computed in Exercise 47.
This gives
V 2 hr 2 12 12
(12, 12) = = = 301.6
r 3 (12,12) 3
V
(12, 12) = 122 = 150.8
h 3
We conclude that an increase of 1 cm in r leads to a greater increase in V than an increase of 1 cm in h.
49. Use the contour map of f (x, y) in Figure 9 to explain why the following are true:
(a) f x and f y are both larger at P than at Q.
(b) f x (x, y) is decreasing as a function of y, that is, for any x, f x (x, b1 ) > f x (x, b2 ) if b1 < b2 .
y
20 16
14
Q
10
6
P 2
2
x
FIGURE 9 Contour interval 2.
SOLUTION
(a) Since the vertical and horizontal lines through P meet more level curves than vertical and horizontal lines through
Q, f is increasing more rapidly in the y and x direction at P than at Q. Therefore, f x and f y are both larger at P than
at Q.
(b) For any fixed value of x, a horizontal line through (x, b2 ) meets fewer level curves than a horizontal line through
(x, b1 ), for b1 < b2 . Thus, f x is decreasing as a function of y.
50. Over most of the earth, a magnetic compass does not point to true (geographic) north; instead, it points at some angle
east or west of true north. The angle D between magnetic north and true north is called the magnetic declination. Use
Figure 10 to determine which of the following statements is true. Note that because of the way longitude is measured,
the x-axis increases from right to left.
D D D D
(a) > (b) >0 (c) >0
y A y B x C y C
648 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
A
C
30N 10
5 0
SOLUTION
(a) To estimate D D
y A and y B , we move vertically from A and B to the points on the next level curve in the direction
of increasing y (upward). From A, we quickly come to a level curve corresponding to higher value of D; but from B,
moving vertically, there is hardly any change as we move along the curve. The statement is thus true.
(b) The derivative D D
x C is estimated by x . Since x varies in the horizontal direction, we move horizontally from C
to a point on the next level curve in the direction of increasing x (leftwards). Since the value of D on this level curve is
greater than on the level curve of C, D = 1. Also x > 0, hence
D D 1
= > 0.
x C x x
The statement is correct.
(c) Moving from C vertically upward (in the direction of increasing y), we come to a point on a level curve with a
smaller value of D. Therefore, D = 1 and y > 0, so we obtain
D D 1
= <0
y C y y
Hence, the statement is false.
(d) A is located on the level curve with value 15, and C is on the level curve with value 12. That is, at the point A we
have D = 15, and at C we have D = 12. Therefore, D is larger at A than at C by 15 (12) = 27 . The statement is
therefore false.
51. Seawater Density Refer to Example 6 and Figure 4.
(a) Estimate and at points B and C.
T S
(b) Which has a greater effect on density at B, a 1 increase in temperature or a 1-ppt increase in salinity?
(c) True or false: The density of warm seawater is more sensitive to a change in temperature than the density of cold
seawater? Explain.
SOLUTION
(a) We estimate T at B. Since T varies in the vertical direction, we move vertically from B to a point B
on the next
level curve, in the direction of increasing T (upward). The change in from B to B
is = 1.0235 1.0240 =
0.0005. The change in T from B to B
is T = 2.5 . Hence,
0.0005
= 2 104
T B T 2.5
B
1.0235
1.0240
B
We now compute T |C . Following the same procedure as before, we find that
0.0005
= 0.0002
T C T 2.5
We now estimate S at C. Since S varies in the horizontal direction, we move horizontally from C to a point C
on the
next level curve, in the direction of increasing S (to the right). The change in from C to C
is = 1.0255 1.0250 =
0.0005. The change in S from C to C
is S = 0.5. Hence,
0.0005
= = 0.001.
S C S 0.5
S E C T I O N 15.3 Partial Derivatives (ET Section 14.3) 649
Finally, we now estimate S at B, using the same technique. We get
0.0005
= 0.0008
S B S 0.6
(b) The change in the density as a result of a change T in temperature (where S remains unchanged), and as a
result of a change S in salinity (where T remains unchanged), can be estimated by
T ; S (1)
T C S C
We substitute T = 1 , S = 1 ppt, T C 0.0002, and S C 0.001 (obtained in part (a)) in (1), to obtain
0.0002 for the change in temperature
0.001 for the change in salinity.
We conclude that the effect caused by the change in salinity is greater than the effect caused by the change in temperature,
at the point C.
(c) At low temperatures the vertical distances between adjacent level curves are greater than at higher temperatures. This
means that the density of cold water is less sensitive to a change in the temperature than the density of warm water. Thus,
the statement is true.
52. Estimate the partial derivatives at P of the function whose contour map is shown in Figure 11.
y
21
18
15
4 12
9
6
P 3
2
x
0 2 4 6 8
FIGURE 11
SOLUTION The contour interval is m = 3. To estimate the partial derivative xf at P, we estimate the change x
between P and the point P
on the next level curve to the right, which is the length of the segment P P
:
x 2
y
21
18
15 12
4 P" 9
P' 6
P 3
2
x
0 2 4 6 8
To estimate the partial derivative yf at P, we estimate the change y between P and the point P
on the next level
curve vertically above P:
y 0.5
The change in f is f = 3 (since the level curve of P
is to the left of the level curve of P). Hence,
f f 3
= 6.
y P y 0.5
2 f 2 f
53. f (x, y) = 3x 2 y 6x y 4 , and
x 2 y 2
650 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
f f
= 6x y 6y 4 ; = 3x 2 6x 4y 3 = 3x 2 24x y 3
x y
2 f 2 f
= f x = 6y; = f y = 24x 3y 2 = 72x y 2 .
x2 x y 2 y
1 2x
fy = x ln y 2 = x ln y 2 = x 2 2y =
y y y y
1 2x
f yy (x, y) = f y = 2x = 2 .
y y y y
The derivative at (2, 3) is thus
2 2 4
f yy (2, 3) = = .
32 9
xy 2g
55. g(x, y) = ,
xy x y
SOLUTION By definition we have
2g g
= g yx =
xy x y
Differentiating g
y with respect to x, using the Quotient Rule, we obtain
2g g x2 2x(x y)2 x 2 2(x y) 2x y
= = = =
xy x y x (x y)2 (x y)4 (x y)3
g y (x, y) = x (yey ) = x 1 ey + y ey (1) = xey (1 y)
y
We now differentiate g y with respect to y to find g yy :
y
g yy (x, y) = xe (1 y) = x ey (1 y) = x ey (1 y) + ey (1) = xey (y 2)
y y
At the point (1, 0) we get
g yy (1, 0) = 1e0 (0 2) = 2
u
57. h(u, v) = , h uv (u, v)
u+v
S E C T I O N 15.3 Partial Derivatives (ET Section 14.3) 651
SOLUTION
By definition h uv = v h . We compute h using the Quotient Rule:
u u
h u 1 (u + v) u 1 v
= = =
u u u + v (u + v)2 (u + v)2
v 1 (u + v)2 v 2(u + v) uv
h uv (u, v) = = =
v (u + v) 2 (u + v)4 (u + v)3
h 3y 2
= 3
y x + y3
so thus
2h 3y 2 9x 2 y 2
= =
xy x x 3 + y3 (x 3 + y 3 )2
59. Estimate and at (S, T ) = (34, 2) and (35, 10).
T S
SOLUTION We estimate T at the given points using the values in Table 1 and the following approximation:
60. For fixed salinity S = 33, is concave up or concave down as a function of T ? Hint: Determine if the
2
quotients /T are increasing or decreasing. What can you conclude about the sign of ?
T 2
652 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
SOLUTION The function (33, T ) is concave up (concave down) if T (33, T ) is an increasing (decreasing) function
of T . We use Table 1 to estimate whether the function T (33, T ) is increasing or decreasing. We compute the following
values:
(33, 4) (33, 2) 26.23 26.38
(33, 2) = = 0.075
T 2 2
(33, 6) (33, 4) 26 26.23
(33, 4) = = 0.115
T 2 2
(33, 8) (33, 6) 25.73 26
(33, 6) = = 0.135
T 2 2
(33, 10) (33, 8) 25.42 25.73
(33, 8) = = 0.155
T 2 2
(33, 12) (33, 10) 25.07 25.42
(33, 10) = = 0.175
T 2 2
These values indicate that T (33, T ) is a decreasing function of T , which means that the second derivative is negative,
2
i.e., 2 (33, T ) < 0 and the graph of (33, T ) is concave down.
T
61. Compute f x yz for
z + z 1
f (x, y, z) = sin(yx) + tan
x x 1
f yz (x, y, z) = x cos(yx) = 0
z
Hence,
f yzx (x, y, z) = 0
We conclude that at the points where the partial derivatives are continuous,
62. Let
x2 + ey v
f (x, y, u, v) =
3y 2 + ln(2 + u 2 )
What is the fastest way to show that f uvx yvu (x, y, u, v) = 0 for all (x, y, u, v)?
SOLUTION We first differentiate with respect to v, obtaining
x2 ey
f v (x, y, u, v) = + v
v 3y 2 + ln(2 + u 2 ) v 3y 2 + ln(2 + u 2 )
ey ey
=0+ =
3y 2 + ln(2 + u 2 ) 3y 2 + ln(2 + u 2 )
We now differentiate f v with respect to x. Since f v does not depend on x, we have
f vx (x, y, u, v) = 0
Hence also,
f uvx yvu (x, y, u, v) = (0) = 0
u y v u
S E C T I O N 15.3 Partial Derivatives (ET Section 14.3) 653
fu = cos(u + v 2 ) = sin(u + v 2 ) (u + v 2 ) = sin(u + v 2 )
u u
f uu = sin(u + v 2 ) = cos(u + v 2 )
u
f uuv = cos(u + v 2 ) = sin(u + v 2 ) (u + v 2 ) = 2v sin(u + v 2 )
v v
64. g(x, y, z) = x 4 y 5 z 6 , gx x yz
SOLUTION For g(x, y, z) = x 4 y 5 z 6 , we have
gx = y 5 z 6 x 4 = y 5 z 6 4x 3 = 4x 3 y 5 z 6
x
gx x = 4y 5 z 6 x 3 = 4y 5 z 6 3x 2 = 12x 2 y 5 z 6
x
gx x y = 12x 2 z 6 (y 5 ) = 12x 2 z 6 5y 4 = 60x 2 y 4 z 6
y
6
gx x yz = 60x 2 y 4 z = 60x 2 y 4 6x 5 = 360x 2 y 4 z 5
z
65. F(r, s, t) = r (s 2 + t 2 ), Fr st
SOLUTION For F(r, s, t) = r (s 2 + t 2 ), we have
Fr = s 2 + t 2
Fr s = 2s
Fr st = 0
66. g(x, y, z) = x 4 y 5 z 3 , gx x yz
SOLUTION We have
gx = y 5 z 3 x 4 = y 5 z 3 4x 3 = 4x 3 y 5 z 3
x
5 3 3
gx x = 4y z x = 4y 5 z 3 3x 2 = 12x 2 y 5 z 3
x
gx x y = 12x 2 z 3 y 5 = 12x 2 z 3 5y 4 = 60x 2 y 4 z 3
y
3
gx x yz = 60x 2 y 4 (z ) = 60x 2 y 4 3z 2 = 180x 2 y 4 z 2
z
We now differentiate u x with respect to x, using the Product Rule and the Chain Rule:
1 3/2
x 2 /4t 1 3/2 x 2 /4t x 2 /4t 2x
ux x = t xe = t 1e +x e
2 x 2 4t
1 x 2 x 2 /4t 1 x2
= t 3/2 ex /4t = t 3/2 ex /4t 1
2 2
e
2 2t 2 2t
654 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
u
68. R(u, v, w) = , Ruvw
v+w
SOLUTION We differentiate R with respect to u:
u 1
Ru = =
u v+w v+w
We now differentiate Ru with respect to v, using the Chain Rule:
1 1
Ruv = =
v v + w (v + w)2
Finally we differentiate Ruv with respect to w:
2
Ruvw = (v + w)2 = 2(v + w)3 = .
w (v + w)3
69. g(x, y, z) = x 2 + y2 + z2, gx yz
SOLUTION Differentiating with respect to x, using the Chain Rule, we get
1 2 1 x
gx = x 2 + y2 + z2 = (x + y 2 + z 2 ) = 2x =
x 2 x + y + z x
2 2 2 2 x +y +z
2 2 2 x + y2 + z2
2
We now differentiate gx with respect to y, using the Chain Rule. This gives
2 2 2 1/2 1 3/2 x y
gx y = x (x + y + z ) =x (x 2 + y 2 + z 2 ) 2y =
y 2 (x + y 2 + z 2 )
2 3/2
ux = sech2 (x t) = 2 sech(x t) sech(x t) tanh(x t) (x t) = 2 sech2 (x t) tanh(x t)
x x
We now use the Product Rule and the Chain Rule to differentiate u x with respect to x:
u x x = 2 2 sech(x t) sech(x t) tanh(x t) tanh(x t) + sech2 (x t) sech2 (x t)
= 4 sech2 (x t) tanh2 (x t) 2 sech4 (x t) = 2 sech2 (x t) 2 tanh2 (x t) sech2 (x t)
f f
71. Find (by guessing) a function such that = 2x y and = x 2.
x y
SOLUTION The function f (x, y) = x 2 y satisfies yf = x 2 and xf = 2x y.
f f
72. Prove that there does not exist any function f (x, y) such that = x y and = x 2 . Hint: Show that f
x y
cannot satisfy Clairauts Theorem.
SOLUTION Suppose that there exists a function f (x, y) such that xf = x y and yf = x 2 . Hence,
f
fx y = = xy = x
y x y
S E C T I O N 15.3 Partial Derivatives (ET Section 14.3) 655
f 2
f yx = = x = 2x
x y x
The mixed partials f x y and f yx are continuous everywhere, but f x y = f yx for x = 0. This contradicts Clairauts
Theorem on Equality of Mixed Partials. We conclude that there does not exist any function f (x, y) with the given
partials.
73. Assume that f x y and f yx are continuous and that f yx x exists. Show that f x yx also exists and that f yx x = f x yx .
SOLUTION Since f x y and f yx are continuous, Clairauts Theorem implies that
f x y = f yx (1)
f yx x = fy = f yx = f x y = f x yx
x x x x
Therefore, f x yx also exists and f yx x = f x yx .
74. Show that u(x, t) = sin(nx) en t satisfies the heat equation (Example 10) for any constant n:
2
f 2 f
= 3
t x2
SOLUTION We compute u
t using the Chain Rule:
u 2
= sin(nx) en t = sin(nx)en t (n 2 t) = n 2 sin(nx)en t
2 2
t t t
We now find u x :
2
u x = en t sin(nx) = en t cos(nx) n = n cos(nx)en t
2 2
x
Differentiating u x with respect to x gives
2 2
u x x = nen t cos(nx) = nen t sin(nx) (nx) = nen t sin(nx) n = n 2 en t sin(nx)
2 2
x x
We see that u t = u x x , therefore u satisfies the heat equation.
75. Find all values of A and B such that f (x, t) = e Ax+Bt satisfies Eq. (3).
SOLUTION We compute the following partials, using the Chain Rule:
f
= (e Ax+Bt ) = e Ax+Bt ( Ax + Bt) = Be Ax+Bt
t t t
f Ax+Bt
= (e ) = e Ax+Bt ( Ax + Bt) = Ae Ax+Bt
x x x
2 f
= ( Ae Ax+Bt ) = A (e Ax+Bt ) = Ae Ax+Bt ( Ax + Bt) = A2 e Ax+Bt
x 2 x x x
Substituting these partials in the differential equation (3), we get
Be Ax+Bt = A2 e Ax+Bt
B = A2
We conclude that f (x, t) = e Ax+Bt satisfies equation (5) if and only if B = A2 , where A is arbitrary.
In Exercises 7679, the Laplace operator is defined by f = f x x + f yy . A function u(x, y) satisfying the Laplace
equation u = 0 is called harmonic.
ux = (x) = 1; u x x = (1) = 0
x x
uy = (x) = 0; u yy = (0) = 0
y y
Since u x x + u yy = 0, u is harmonic.
(b) We compute the partial derivatives of u(x, y) = e x cos y:
x
ux = e cos y = cos y e x = (cos y)e x
x x
x
uy = e cos y = e x cos y = e x sin y
y y
ux x = (cos y)e x = cos y e x = (cos y)e x
x x
x
u yy = e sin y = e x sin y = e x cos y
y y
Thus,
1 2x
ux = ln(x 2 + y 2 ) = 2 2x = 2
x x +y 2 x + y2
1 2y
uy = ln(x 2 + y 2 ) = 2 2y = 2
y x +y 2 x + y2
We now find u x x and u yy using the Quotient Rule:
2x 2(x 2 + y 2 ) 2x 2x 2(y 2 x 2 )
ux x = = =
x x 2 + y2 (x 2 + y 2 )
2
(x 2 + y 2 )
2
2y 2(x 2 + y 2 ) 2y 2y 2(x 2 y 2 )
u yy = = =
y x 2 + y 2 (x 2 + y 2 )
2
(x 2 + y 2 )
2
Thus,
2(y 2 x 2 ) 2(x 2 y 2 )
u x x + u yy = 2
+ 2
= 0.
(x 2 + y 2 ) (x 2 + y 2 )
SOLUTION We compute the first-order partials u x and u y and the second-order partials u x x and u yy of the given
polynomial u(x, y). This gives
u x = 3ax 2 + 2bx y + cy 2
u y = bx 2 + 2cx y + 3d y 2
u x x = 6ax + 2by
u yy = 2cx + 6d y
This equality holds for all x and y if and only if the coefficients of x and y are both zero. That is, 6a + 2c = 0 (so
c = 3a) and 2b + 6d = 0 (so b = 3d). We conclude that the harmonic polynomials in the given form are
u(x, y) = ax 3 3d x 2 y 3ax y 2 + d y 3
u u
78. Show that if u(x, y) is harmonic, then the partial derivatives and are harmonic.
x y
SOLUTION We assume that the second-order partials are continuous, hence the partial differentiation may be performed
in any order. By the given data, we have
u x x + u yy = 0 (1)
(u x )x x + (u x ) yy = 0 and (u y )x x + (u y ) yy = 0
0 = (u x x )x + (u yy )x = u x x x + u x yy = (u x )x x + (u x ) yy (2)
2
ux x = cos(ax)eby = a 2 cos(ax)eby = a 2 u
x2
2
u yy = cos(ax)eby = b2 cos(ax)eby = b2 u
y 2
4u t + u x x x + 12uu x = 0
u x = 2 sech2 (x t) tanh(x t)
u x x x = 16 sech4 (x t) tanh(x t) 8 sech2 (x t) tanh3 (x t)
Hence,
y(x 4 + 4x 2 y 2 y 4 )
f x (x, y) =
(x 2 + y 2 )2
x(x 4 4x 2 y 2 y 4 )
f y (x, y) =
(x 2 + y 2 )2
(b) Use the limit definition of the partial derivative to show that f x (0, 0) = f y (0, 0) = 0 and that f yx (0, 0) and f x y (0, 0)
both exist but are not equal.
(c) Use a computer algebra system to show that for (x, y) = (0, 0):
x 6 + 9x 4 y 2 9x 2 y 4 y 6
f x y (x, y) = f yx (x, y) =
(x 2 + y 2 )3
Show that f x y is not continuous at (0, 0). Hint: Show that lim f x y (h, 0) = lim f x y (0, h).
h0 h0
(d) Explain why the result of (b) does not contradict Clairauts Theorem.
SOLUTION
(a) We use a CAS to verify the following partials for (x, y) = (0, 0):
y(x 4 + 4x 2 y 2 y 4 )
f x (x, y) = 2
(x 2 + y 2 )
x(x 4 4x 2 y 2 y 4 )
f y (x, y) = 2
(x 2 + y 2 )
(b) Using the limit definition of the partial derivatives at the point (0, 0) we have
h 0 h 2 02 0
2 2
f (h, 0) f (0, 0) h +0 0
f x (0, 0) = lim = lim = lim =0
h0 h h0 h h0 h
0 k 02 k 2 0
2 2
f (0, k) f (0, 0) 0 +k 0
f y (0, 0) = lim = lim = lim = 0
k0 k k0 k k0 k
We now use the derivatives in part (a) and the limit definition of the partial derivatives to compute f yx (0, 0) and f x y (0, 0).
By the formulas in part (a), we have
h(h 4 0 0)
f y (0, 0) = 0, f y (h, 0) = 2
=h
(h 2 + 0)
k(0 + 0 k 4 )
f x (0, 0) = 0, f x (0, k) = 2
= k
(02 + k 2 )
Thus,
f y (h, 0) f y (0, 0) h0
f yx (0, 0) = f y = lim = lim = lim 1 = 1
x (0,0) h0 h h0 h h0
f x (0, k) f x (0, 0) k 0
f x y (0, 0) = f x = lim = lim = lim (1) = 1
y (0,0) k0 k k0 k k0
We see that the mixed partials at the point (0, 0) exist but are not equal.
S E C T I O N 15.4 Differentiability, Linear Approximation, and Tangent Planes (ET Section 14.4) 659
(c) We verify, using a CAS, that for (x, y) = (0, 0) the following derivatives hold:
x 6 + 9x 4 y 2 9x 2 y 4 y 6
f x y (x, y) = f yx (x, y) = 3
(x 2 + y 2 )
To show that f x y is not continuous at (0, 0), we show that the limit lim(x,y)(0,0) f x y (x, y) does not exist. We compute
the limit as (x, y) approaches the origin along the x-axis. Along this axis, y = 0; hence,
h 6 + 9h 4 0 9h 2 0 0
lim f x y (x, y) = lim f x y (h, 0) = lim 3
= lim 1 = 1
(x,y)(0,0) h0 h0 (0 + h 2 ) h0
along the x -axis
We compute the limit as (x, y) approaches the origin along the y-axis. Along this axis, x = 0, hence,
0 + 0 + 0 h6
lim f x y (x, y) = lim f x y (0, h) = lim 3
= lim (1) = 1
(x,y)(0,0) h0 h0 (0 + h 2 ) h0
along the y -axis
Since the limits are not equal f (x, y) does not approach one value as (x, y) (0, 0), hence the limit
lim f x y (x, y) does not exist, and f x y (x, y) is not continuous at the origin.
(x,y)(0,0)
(d) The result of part (b) does not contradict Clairauts Theorem since f x y is not continuous at the origin. The continuity
of the mixed derivative is essential in Clairauts Theorem.
15.4 Differentiability, Linear Approximation, and Tangent Planes (ET Section 14.4)
Preliminary Questions
1. How is the linearization of f (x, y) at (a, b) defined?
SOLUTION The linearization of f (x, y) at (a, b) is the linear function
That is,
or
Exercises
1. Let f (x, y) = x 2 y 3 .
(a) Find the linearization of f at (a, b) = (2, 1).
(b) Use the linear approximation to estimate f (2.01, 1.02) and f (1.97, 1.01), and compare your estimates with the
values obtained using a calculator.
SOLUTION
(a) We compute the value of the function and its partial derivatives at (a, b) = (2, 1):
f (x, y) = x 2 y 3 f (2, 1) = 4
f x (x, y) = 2x y 3 f x (2, 1) = 4
f y (x, y) = 3x 2 y 2 f y (2, 1) = 12
(b) For h = x 2 and k = y 1 we have the following form of the linear approximation at (a, b) = (2, 1):
2. Write the linear approximation to f (x, y) = x(1 + y)1 at (a, b) = (8, 1) in the form
Use it to estimate 7.98/2.02 and compare the estimate with the value obtained using a calculator.
SOLUTION We first compute the value of f (x, y) = x(1 + y)1 and its partial derivatives at (a, b) = (8, 1):
Hence,
1
f (8 + h, 1 + k) 4 + h 2k (1)
2
7.98 1
f (7.98, 1.02) = 4 + (0.02) 2(0.02) = 3.95
2.02 2
The actual value is
7.98
= 3.950495 . . .
2.02
3. Let f (x, y) = x 3 y 4 . Use Eq. (5) to estimate the change f = f (2.03, 0.9) f (2, 1).
SOLUTION We compute the function and its partial derivatives at (a, b) = (2, 1):
f (x, y) = x 3 y 4 f (2, 1) = 8
f x (x, y) = 3x 2 y 4 f x (2, 1) = 12
f y (x, y) = 4x 3 y 5 f y (2, 1) = 32
4. Assume that f (3, 2) = 2, f x (3, 2) = 1, and f y (3, 2) = 3. Use the linear approximation to estimate f (3.1, 1.8).
SOLUTION The linear approximation at the point (a, b) = (3, 2) is
That is, the value of f at the point (3.1, 1.8) can be estimated by f (3.1, 1.8) 1.3.
5. Use the linear approximation of f (x, y) = e x +y at (0, 0) to estimate f (0.01, 0.02). Compare with the value
2
We first must compute f and its partial derivative at the point (0, 0). Using the Chain Rule we obtain
f (x, y) = e x +y
2
f (0, 0) = e0 = 1
f x (x, y) = 2xe x +y
2
f x (0, 0) = 2 0 e0 = 0
f y (x, y) = e x +y
2
f y (0, 0) = e0 = 1
x2
6. Let f (x, y) = 2 . Use the linear approximation at an appropriate point (a, b) to estimate f (4.01, 0.98).
y +1
SOLUTION We use the linear approximation at the point (a, b) = (4, 1), which is the closest point with integer coor-
dinates. That is,
x2
f (x, y) = 2 f (4, 1) = 8
y +1
2x
f x (x, y) = 2 f x (4, 1) = 4
y +1
1 2y 2x 2 y
f y (x, y) = x 2 = x2 = f y (4, 1) = 8
y y + 1
2
(y 2 + 1)
2
(y 2 + 1)
2
4.012
f (4.01, 0.98) = = 8.202
0.982 + 1
7. Use Eq. (3) to find an equation of the tangent plane to the graph of f (x, y) = 2x 2 4x y 2 at the point (1, 2).
SOLUTION The equation of the tangent plane at the point (1, 2, 18) is
We compute the function and its partial derivatives at the point (1, 2):
f (x, y) = 2x 2 4x y 2 f (1, 2) = 18
f x (x, y) = 4x 4y 2 f x (1, 2) = 20
f y (x, y) = 8x y f y (1, 2) = 16
That is,
z = 34 20x + 16y
We compute the function and its partial derivatives at the point (2, 1). Using the chain Rule we obtain
z = e2 + e2 (x 2) 2e2 (y 1) = e2 + e2 x 2e2 y
z = e2 + e2 x 2e2 y
xy
9. Find the linear approximation to f (x, y, z) = at the point (2, 1, 2).
z
S E C T I O N 15.4 Differentiability, Linear Approximation, and Tangent Planes (ET Section 14.4) 663
10. Assume that f (1, 0, 0) = 3, f x (1, 0, 0) = 2, f y (1, 0, 0) = 4, and f z (1, 0, 0) = 2. Use the linear approximation
to estimate f (1.02, 0.01, 0.03).
SOLUTION The linear approximation at (1, 0, 0) is
We substitute h = 0.02, k = 0.01, l = 0.03 and the given values to obtain the following estimation:
That is,
In Exercises 1118, find an equation of the tangent plane at the given point.
f (x, y) = x 2 y + x y 3 f (2, 1) = 6
f x (x, y) = 2x y + y 3 f x (2, 1) = 5
f y (x, y) = x 2 + 3x y 2 f y (2, 1) = 10
We now substitute these values in (1) to obtain the following equation of the tangent plane:
That is,
z = 5x + 10y 14.
x
12. f (x, y) = , (4, 4)
y
SOLUTION The equation of the tangent plane at (4, 4) is
f (x, y) = x 2 + y 2 f (4, 1) = 17
f x (x, y) = 2x f x (4, 1) = 8
f y (x, y) = 2y 3 f y (4, 1) = 2
We substitute these values in (1) to obtain the following equation of the tangent plane:
z = 5 + 4(r 2) 5(s 1) = 4r 5s + 2.
z =0+1s +0t =s
That is, z = s.
17. f (x, y) = e x ln y, (0, 1)
SOLUTION The equation of the tangent plane at (0, 1) is
f (x, y) = e x ln y f (0, 1) = 0
f x (x, y) = e x ln y f x (0, 1) = 0
ex
f y (x, y) = f y (0, 1) = 1
y
Substituting in (1) gives the following equation of the tangent plane:
z = 0 + 0x + 1(y 1) = y 1
That is, z = y 1.
18. f (x, y) = ln(x 2 + y 2 ), (1, 1)
SOLUTION The equation of the tangent plane at (1, 1) is
z = ln 2 + 1(x 1) + 1(y 1) = x + y + ln 2 2
That is, z = x + y 2 + ln 2.
19. Find the points on the graph of z = 3x 2 4y 2 at which the vector n = 3, 2, 2 is normal to the tangent plane.
SOLUTION The equation of the tangent plane at the point a, b, f (a, b) on the graph of z = f (x, y) is
or
f x (x, y) = 6x f x (a, b) = 6a
f y (x, y) = 8y f y (a, b) = 8b
Therefore, the vector v = 6a, 8b, 1 is normal to the tangent plane at (a, b). Since we want n = 3, 2, 2 to be
normal to the plane, the vectors v and n must be parallel. That is, the following must hold:
6a 8b 1
= =
3 2 2
20. Find the points on the graph of z = x 2 e y at which the tangent plane is parallel to 5x 2y + 12 z = 0.
SOLUTION The equation of the tangent plane at the point a, b, f (a, b) is
or
2aeb a 2 eb 1
= = 1 = 2
5 2
2
2aeb = 10
a 2 eb = 4
The first equation implies that aeb = 5. Substituting in the second equation gives
4
a(aeb ) = a 5 = 4 a =
5
Therefore, 45 eb = 5 or eb = 25 25
4 , yielding b = ln 4 . We compute the z-coordinate of the point:
4 2 ln 25 16 25
z= e 4 = = 4.
5 25 4
We conclude that the point on the graph at which the tangent plane is parallel to 5x 2x + 12 z = 0 is 45 , ln 25
4 ,4 .
21. Use a computer algebra system to plot the graph of f (x, y) = x 2 + 3x y + y together with the tangent plane
at (x, y) = (1, 1) on the same screen.
S E C T I O N 15.4 Differentiability, Linear Approximation, and Tangent Planes (ET Section 14.4) 667
1 5
0 1
3
22. Suppose that the plane tangent to the surface z = f (x, y) at (2, 3, 4) has equation z + 4x + 2y = 2. Estimate
f (2.1, 3.1).
SOLUTION The tangent plane z = 2 4x 2y is also a linear approximation for f near (2, 3), so we can thus
calculate the following:
Find an approximation to the equation of the tangent plane to the graph of f at (1, 2, 10).
SOLUTION The equation of the tangent plane at the point (1, 2) is
Since the values of the partial derivatives at (1, 2) are not given, we approximate them as follows:
f (1.1, 2) f (1, 2) f (1.1, 2.01) f (1, 2)
f x (1, 2) =3
0.1 0.1
f (1, 2.1) f (1, 2) f (1.04, 2.1) f (1, 2)
f y (1, 2) = 3
0.1 0.1
Substituting in (1) gives the following approximation to the equation of the tangent plane:
z = 10 + 3(x 1) 3(y 2)
Substituting these values and h = 0.1, k = 0.1 in (1) gives the following estimation:
9.1 3 1 3
+ 0.1 (0.1) 1.5271
3.9 2 12 16
The value obtained by a calculator is 9.1
3.9 1.5275. The error is 0.0004 and the percentage error is
0.0004 100
percentage error 0.0262%
1.5275
In Exercises 2530, use the linear approximation to estimate the value. Compare with the value given by a calculator.
We compute the value of the function and its partial derivatives at (2, 1):
f (x, y) = x 3 y 2 f (2, 1) = 8
f x (x, y) = 3x 2 y 2 f x (2, 1) = 12
f y (x, y) = 2x 3 y f y (2, 1) = 16
Substituting these values and h = 0.01, k = 0.02 in (1) gives the approximation
The value given by a calculator is 8.4487. The error is 0.0087 and the percentage error is
0.0087 100
Percentage error = 0.103%
8.4487
26. 4.1/7.9
SOLUTION The number 7.9 is a value of the function f (x, y) = x y 1 . We use the linear approximation at the point
4.1
(4, 8), which is
We compute the values of the function and its partial derivatives at (4, 8):
1
f (x, y) = x y 1 f (4, 8) =
2
1
f x (x, y) = y 1 f x (4, 8) =
8
1
f y (x, y) = x y 2 f y (4, 8) =
16
We substitute these values and h = 0.1, k = 0.1 in (1) to obtain the following approximation:
4.1 1 1 1 83
+ 0.1 (0.1) = = 0.51875
7.9 2 8 16 160
0.00024 100
Percentage error = 0.04625%
0.51899
27. 3.012 + 3.992
SOLUTION This is a value of the function f (x, y) = x 2 + y 2 . We use the linear approximation at the point (3, 4),
which is
Substituting these values and h = 0.01, k = 0.01 in (1) gives the following approximation:
3 4
3.012 + 3.992 5 + 0.01 + (0.01) = 4.998
5 5
The value given by a calculator is 3.012 + 3.992 4.99802. The error is 0.00002 and the percentage error is at most
0.00002 100
Percentage error = 0.0004002%
4.99802
0.982
28.
2.013 + 1
2
SOLUTION We use the linear approximation of the function f (x, y) = 3x at the point (1, 2), which is
y +1
We compute the values of f and its partial derivatives at (1, 2). We get:
x2 1
f (x, y) = 3 f (1, 2) =
y +1 9
2x 2
f x (x, y) = 3 f x (1, 2) =
y +1 9
1 3x 2 y 2 4
f y (x, y) = x 2 2
3y 2 = 2
f y (1, 2) =
(y 3 + 1) (y 3 + 1) 27
Substituting these values and h = 0.02, k = 0.01 in (1) gives the following approximation:
0.982 1 2 4
+ (0.02) 0.01 0.1052
2.013 + 1 9 9 27
2
The value given by a calculator is 0.98 0.1053. The error is 0.0001 and the percentage error is at most
3 2.01 +1
0.0001 100
Percentage error 0.095%
0.1053
29. (1.9)(2.02)(4.05)
SOLUTION We use the linear approximation of the function f (x, y, z) = x yz at the point (2, 2, 4), which is
We compute the values of the function and its partial derivatives at (2, 2, 4):
f (x, y, z) = x yz f (2, 2, 4) = 4
yz 1 yz
f x (x, y, z) = = f x (2, 2, 4) = 1
2 x yz 2 x
xz 1 xz
f y (x, y, z) = = f y (2, 2, 4) = 1
2 x yz 2 y
xy 1 xy 1
f z (x, y, z) = = f z (2, 2, 4) =
2 x yz 2 z 2
Substituting these values and h = 0.1, k = 0.02, l = 0.05 in (1) gives the following approximation:
1
(1.9)(2.02)(4.05) = 4 + 1 (0.1) + 1 0.02 + (0.05) = 3.945
2
670 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
8.01
30.
(1.99)(2.01)
SOLUTION We use the linear approximation of the function f (x, y, z) = xyz at the point (8, 2, 2), which is
We compute the values of the function and its partial derivatives at (8, 2, 2). This gives
x
f (x, y, z) = f (8, 2, 2) = 4
yz
1 1
f x (x, y, z) = f x (8, 2, 2) =
yz 2
1 1
f y (x, y, z) = x (yz)1/2 = x(yz)3/2 z = x y 3/2 z 1/2 f y (8, 2, 2) = 1
y 2 2
1 1
f z (x, y, z) = x (yz)1/2 = x(yz)3/2 y = x y 1/2 z 3/2 f z (8, 2, 2) = 1
z 2 2
Substituting these values and h = 0.01, k = 0.01, l = 0.01 in (1) gives the following approximation:
8.01 1
= 4 + 0.01 1 (0.01) 1 0.01 = 4.005
(1.99)(2.01) 2
The value given by a calculator is 4.00505. The error is 0.00005 and the percentage error is at most
0.00005 100
Percentage error 0.00125%
4.00505
31. Estimate f (2.1, 3.8) given that f (2, 4) = 5, f x (2, 4) = 0.3, and f y (2, 4) = 0.2.
SOLUTION We use the linear approximation of f at the point (2, 4), which is
Substituting the given values and h = 0.1, k = 0.2 we obtain the following approximation:
32. A boy has weight W = 34 kg and height H = 1.3. Use the linear approximation to estimate the change in I if
(W, H ) changes to (36, 1.32).
SOLUTION Let I = I (36, 1.32) I (34, 1.3) denote the change in I . Using the linear approximation of I at the
point (34, 1.3) we have
I I
I (34 + h, 1.3 + k) I (34, 1.3) (34, 1.3)h + (34, 1.3)k
W H
For h = 2, k = 0.02 we obtain
I I
I (34, 1.3) 2 + (34, 1.3) 0.02 (1)
W H
We compute the partial derivatives in (1):
I W 1 I
= = 2 (34, 1.3) = 0.5917
W W H 2 H W
I 2 2W I
=W H = W (2H 3 ) = (34, 1.3) = 30.9513
H H H3 H
Substituting the partial derivatives in (1) gives the following estimation of I :
33. Suppose that (W, H ) = (34, 1.3) and W increases to 35. Use the linear approximation to estimate the increase in H
required to keep I constant.
SOLUTION The linear approximation of I = W2 at the point (34, 1.3) is:
H
I I
I = I (34 + h, 1.3 + k) I (34, 1.3) (34, 1.3)h + (34, 1.3)k (1)
W H
In the earlier exercise, we found that
I I
(34, 1.3) = 0.5917, (34, 1.3) = 30.9513
W H
We substitute these derivatives and h = 1 in (1), equate the resulting expression to zero and solve for k. This gives:
I 0.5917 1 30.9513 k = 0
0.5917 = 30.9513k k = 0.0191
That is, for an increase in weight of 1 kg, the increase in height must be approximately 0.0191 meters (or 1.91 centimeters)
in order to keep I constant.
34. (a) Show that I 0 if H/W H/2W .
(b) Suppose that (W, H ) = (25, 1.1). What increase in H will leave I (approximately) constant if W is increased by
1 kg?
SOLUTION
(a) The linear approximation implies that
I I
I W + H
W H
Hence, I 0 if
I I
W + H = 0 (1)
W H
(c) The volume of a certain cylinder V is determined by measuring r and h. Which will lead to a greater error in V : a
1% error in r or a 1% error in h?
SOLUTION
(a) The linear approximation is
V Vr r + Vh h (1)
V 2 hr r + r 2 h
We divide by V = r 2 h to obtain
V 2 hr r r 2 h 2 hr r r 2 h 2r h
+ = + = +
V V V r h
2 r 2h r h
That is,
V 2r h
+
V r h
(b) The percentage increase in V is, by part (a),
V r h
100 2 100 + 100
V r h
V
100 = 2 2 + 2 = 6%
V
(c) The percentage error in V is
V r h
100 = 2 100 + 100
V r h
V
100 = 2 1 + 0 = 2%
V
A 1% in h implies that h
h 100 = 1. Assuming that there is no error in r , we get
V
100 = 0 + 1 = 1%
V
We conclude that a 1% error in r leads to a greater error in V than a 1% error in h.
36. Use the linear approximation to show that if I = x a y b , then
I x y
a +b
I x y
I I x x + I y y (1)
I ax a1 y b x + bx a y b1 y
S E C T I O N 15.4 Differentiability, Linear Approximation, and Tangent Planes (ET Section 14.4) 673
I ax a1 y b x bx a y b1 y ax a1 y b x bx a y b1 y x y
+ = a b
+ a b
=a +b
I I I x y x y x y
That is,
I x y
a +b .
I x y
37. The monthly payment for a home loan is given by a function f (P, r, N ), where P is the principal (the initial size
of the loan), r the interest rate, and N the length of the loan in months. Interest rates are expressed as a decimal: A 6%
interest rate is denoted by r = 0.06. If P = $100,000, r = 0.06, and N = 240 (a 20-year loan), then the monthly
payment is f (100,000, 0.06, 240) = 716.43. Furthermore, with these values, we have
f f f
= 0.0071, = 5,769, = 1.5467
P r N
Estimate:
(a) The change in monthly payment per $1,000 increase in loan principal.
(b) The change in monthly payment if the interest rate increases to r = 6.5% and r = 7%.
(c) The change in monthly payment if the length of the loan increases to 24 years.
SOLUTION
(a) The linear approximation to f (P, r, N ) is
f f f
f P + r + N
P r N
We are given that Pf = 0.0071, rf = 5769, Nf = 1.5467, and P = 1000. Assuming that r = 0 and N = 0,
we get
The change in monthly payment per thousand dollar increase in loan principal is $7.1.
(b) By the given data, we have
The interest rate 6.5% corresponds to r = 0.065, and the interest rate 7% corresponds to r = 0.07. In the first case
r = 0.065 0.06 = 0.005 and in the second case r = 0.07 0.06 = 0.01. Substituting in (1), assuming that
P = 0 and N = 0, gives
f 1.5467 48 = 74.2416
38. Automobile traffic passes a point P on a road of width w ft at an average rate of R vehicles per second. The arrival
of automobiles is irregular, however, and traffic engineers have found that the formula T = te Rt gives a good model for
the average waiting time T until there is a gap in traffic of at least t seconds. A pedestrian walking at a speed of 3.5 ft/s
(5.1 mph) requires t = w/3.5 s to cross the road. Therefore, the average time the pedestrian will have to wait before
crossing is f (w, R) = (w/3.5)ew R/3.5 s.
(a) What is the pedestrians average waiting time if w = 25 ft and R = 0.2 vehicles per second?
(b) Use the linear approximation to estimate the increase in waiting time if w is increased to 27 ft.
(c) Estimate the waiting time if the width is increased to 27 ft and R decreases to 0.18.
(d) What is the rate of increase in waiting time per 1-ft increase in width when w = 30 ft and R = 0.3 vehicles per
second?
SOLUTION
674 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
(a) We are given that the average time the pedestrian will have to wait for a t-second gap in traffic is
w w R/3.5
f (w, R) = e
3.5
Substituting the values w = 25 and R = 0.2, we obtain
25 (250.2)/3.5
f (25, 0.2) = e 29.8 seconds
3.5
(b) The linear approximation at (w, R) = (25, 0.2) is,
We compute the partial derivatives. Using the Product Rule and the Chain Rule we have
1 R ew R/3.5 wR
f w (w, R) = ew R/3.5 + wew R/3.5 = 1+
3.5 3.5 3.5 3.5
By the Chain Rule we get
w w R/3.5 w w 2
f R (w, R) = e = ew R/3.5
3.5 3.5 3.5
At the point (25, 0.2) we have
f = 2.9 2 = 5.8
That is, the waiting time is increased by approximately 1.54 seconds. Using part (a), the estimated waiting time is
39. The volume V of a cylinder is computed using the values 3.5 m for diameter and 6.2 m for height. Use the linear
approximation to estimate the maximum error in V if each of these values has a possible error of at most 5%.
SOLUTION We denote by d and h the diameter and height of the cylinder, respectively. By the Formula for the Volume
of a Cylinder we have
2
d
V = h = d2h
2 4
The linear approximation is
V V
V d + h (1)
d h
We compute the partial derivatives at (d, h) = (3.5, 6.2):
V V
(d, h) = h 2d = hd (3.5, 6.2) 34.086
d 4 2 d
V 2 V
(d, h) = d (3.5, 6.2) = 9.621
h 4 h
Substituting these derivatives in (1) gives
We are given that the errors in the measurements of d and h are at most 5%. Hence,
d
= 0.05 d = 0.175
3.5
h
= 0.05 h = 0.31
6.2
Substituting in (2) we obtain
f (h) = Mh + h(h) 7
h h(h + 1) + h 2 h
Mh + h E(h) = 1 h + h = = = F(h)
h+1 h+1 h+1
That is, F(x) is differentiable at x = 0, hence also f (x) is differentiable at x = 0.
(b) We have f (0) = 0, hence f is differentiable at x = 0 if (7) holds. We compute M:
3 1/2
f
(x) = x M = f
(0) = 0
2
We define the function E(h) by
f (h) h 3/2
E(h) = = = h 1/2
h h
Then lim E(h) = 0 and equality (7) holds, since
h0
(c) We show that f (x) = x 1/2 is not differentiable at x = 0, by showing that for any constant M, the function E(h)
defined by Eq. (7) does not approach zero as h 0. For f (h) = h 1/2 , Eq. (7) becomes
h = Mh + h E(h)
We solve for E(h) to obtain
h Mh = h E(h)
h Mh 1
E(h) = = M
h h
As h 0+ , E(h) is increasing without bound, therefore E(h) does not approach zero as h 0. We conclude that
f (x) = x is not differentiable at x = 0.
42. Assumptions Matter Recall that if f (x, y) is differentiable at (a, b), then the partial derivatives f x (a, b) and
f y (a, b) exist. This exercise shows that the mere existence of f x (a, b) and f y (a, b) does not imply f (x, y) is differen-
2x y(x + y)
tiable. Define g(x, y) = for (x, y) = 0 and g(0, 0) = 0.
x 2 + y2
(a) Prove that |g(x, y)| |x + y|. Hint: Use the inequality
(x y)2 0
2x y
to show that 2 1.
x + y2
(b) Use (a) to show that g(x, y) is continuous at (0, 0).
(c) Use the limit definitions to show that gx (0, 0) and g y (0, 0) exist and both are equal to zero.
(d) Show that if a function f (x, y) is locally linear at (0, 0) and f (0, 0) = 0, then f (h, h)/ h approaches 0 as h 0.
Show that g(x, y) does not have this property and hence is not differentiable at (0, 0).
(e) Explain why gx (x, y) and g y (x, y) cannot both be continuous at (0, 0). In fact, since g(x, y) is symmetric in x and
y, both partial derivatives are discontinuous at (0, 0).
(f) Show directly that lim gx (x, y) does not exist by verifying that gx (x, 0) = 0 for x = 0 and gx (0, y) = 2 for
(x,y)(0,0)
y = 0.
SOLUTION
(a) The inequality (|x| |y|)2 0 implies that
0 (|x| |y|)2 = x 2 2|x y| + y 2 x 2 + y 2 2|x y|
For (x, y) = (0, 0) we divide by x 2 + y 2 to obtain
2|x y| 2x y
1 or
x 2 + y2 x 2 + y2 1
(d) If f (x, y) were differentiable at (0, 0), then there would exist a function (h, k) satisfying lim (h, k) = 0
(h,k)(0,0)
such that
f (h, k) = L(h, k) + (h, k) h 2 + k 2
Therefore we get
f (h, k) = (h, k) h 2 + k 2 (1)
Therefore,
0 |R1 (h, k)| |(h, k)| and 0 |R2 (h, k)| |(h, k)|.
2h 2 (h + h) f (h, h)
f (h, h) = = 2h =2
h2 + h2 h
SOLUTION The gradient of f (x, y) is a vector with two components, hence the possible value of the gradient f =
f f
x , y is (b).
2. True or false: A differentiable function increases at the rate f P in the direction of f P ?
SOLUTION The statement is true. The value f P is the rate of increase of f in the direction f P .
3. Describe the two main geometric properties of the gradient f .
SOLUTION The gradient of f points in the direction of maximum rate of increase of f and is normal to the level curve
(or surface) of f .
f
4. Express the partial derivative as a directional derivative Du f for some unit vector u.
x
SOLUTION The partial derivative xf is the following limit:
f f (a + t, b) f (a, b)
(a, b) = lim
x t0 t
Considering the unit vector i = 1, 0, we find that
f f (a + t 1, b + t 0) f (a, b)
(a, b) = lim = Di f (a, b)
x t0 t
We see that the partial derivative xf can be viewed as a directional derivative in the direction of i.
5. You are standing at point where the temperature gradient vector is pointing in the northeast (NE) direction. In which
direction(s) should you walk to avoid a change in temperature?
(a) NE (b) NW (c) SE (d) SW
SOLUTION The rate of change of the temperature T at a point P in the direction of a unit vector u, is the directional
derivative Du T (P), which is given by the formula
Du T (P) = f P cos
To avoid a change in temperature, we must choose the direction u so that Du T (P) = 0, that is, cos = 0, so = 2 or
= 32 . Since the gradient at P is pointing NE, we should walk NW or SE to avoid a change in temperature. Thus, the
answer is (b) and (c).
N
NW NE
T(P)
W E
P
SE
6. What is the rate of change of f (x, y) at (0, 0) in the direction making an angle of 45 with the x-axis if f (0, 0) =
2, 4?
SOLUTION By the formula for directional derivatives, and using the unit vector 1/ 2, 1/ 2 , we get 2, 4
1/ 2, 1/ 2 = 6/ 2 = 3 2.
Exercises
1. Let f (x, y) = x y 2 and c(t) = ( 12 t 2 , t 3 ).
(a) Calculate f c
(t).
d
(b) Use the Chain Rule for Paths to evaluate f (c(t)) at t = 1 and t = 1.
dt
SOLUTION
Also,
1 2
3
c
(t) = t , t = t, 3t 2
2
f c
(t) = y 2 , 2x y t, 3t 2 = y 2 t + 6x yt 2 .
d 1
f (c(t)) = t 6 t + 2 t 2 3 t 3 t 2 = 4t 7
dt 2
At the point t = 1 and t = 1, we get
d d
( f (c(t))) = 4 17 = 4, ( f (c(t))) = 4 (1)7 = 4.
dt t=1 dt t=1
d f (c(t)) using the Chain Rule and the results of part (a). This gives
(b) We find dt
d f dx f dy
f (c(t)) = + = (ye x y ) 3t 2 + (xe x y ) 1
dt x dt y dt
dt
f (c(t)) = et +t
3 4
d
(a) Find f (c(t)) without making any calculations. Explain.
dt
(b) Verify your answer to (a) using the Chain Rule.
SOLUTION
(a) The level curves of f (x, y) are the circles x 2 + y 2 = c2 . Since c(t) is a parametrization of the unit circle, f has
d f (c(t)) = 0.
constant value 1 on c. That is, f (c(t)) = 1, which implies that dt
d f (c(t)) using the Chain Rule:
(b) We now find dt
d f dx f dy
f (c(t)) = + (1)
dt x dt y dt
We compute the derivatives involved in (1):
f 2 f 2
= x + y 2 = 2x, = x + y 2 = 2y
x x y y
dx d dy d
= (cos t) = sin t, = (sin t) = cos t
dt dt dt dt
Substituting the derivatives in (1) gives
d
f (c(t)) = 2x( sin t) + 2y cos t
dt
Finally, we substitute x = cos t and y = sin t to obtain
d
f (c(t)) = 2 cos t sin t + 2 sin t cos t = 0.
dt
4. Figure 16 shows the level curves of a function f (x, y) and a path c(t), traversed in the direction indicated. State
d
whether the derivative f (c(t)) is positive, negative, or zero at points AD.
dt
8
4 C 20
D
2 10 B
0
0
A 10
2 20
30
4
6
6 4 2 0 2 4 6 8
FIGURE 16
SOLUTION At points A and D, the path is (temporarily) tangent to one of the contour lines, which means that along
d f (c(t)) is zero. At point B, the path
the path c(t) the function f (x, y) is (temporarily) constant, and so the derivative dt
is moving from a higher contour (of 10) to a lower one (of 20), so the derivative is negative. At the point C, where
the path moves from the contour of 10 towards the contour of value 0, the derivative is positive.
5. f (x, y) = cos(x 2 + y)
SOLUTION We find the partial derivatives using the Chain Rule:
f
= sin x 2 + y x 2 + y = 2x sin x 2 + y
x x
f
= sin x 2 + y x 2 + y = sin x 2 + y
y y
The gradient vector is thus
f f
f = , = 2x sin x 2 + y , sin x 2 + y = sin x 2 + y 2x, 1
x y
x
6. g(x, y) = 2
x + y2
S E C T I O N 15.5 The Gradient and Directional Derivatives (ET Section 14.5) 681
SOLUTION We compute the partial derivatives. We first find gx using the Quotient Rule:
g 1 x 2 + y 2 x 2x y2 x 2
=
=
2
x x 2 + y2
2
x 2 + y2
We compute g
y using the Chain Rule:
g 1 1 2x y
=x =x
2 2y =
2
y y x 2 + y 2 x +y
2 2 x + y2
2
7. h(x, y, z) = x yz 3
SOLUTION We compute the partial derivatives of h(x, y, z) = x yz 3 , obtaining
h h h
= yz 3 , = x z 3 , = x y 3z 4 = 3x yz 4
x y z
The gradient vector is thus
h h h
h = , , = yz 3 , x z 3 , 3x yz 4 .
x y z
8. r (x, y, z, w) = x ze yw
SOLUTION We find the partial derivatives of r (x, y, z, w) = x ze yw :
r r r r
= ze yw , = x zwe yw , = xe yw , = x zye yw
x y z w
The gradient vector is thus
r r r r
r = , , , = ze yw , x zwe yw , xe yw , x zye yw = e yw z, x zw, x, x zy
x y z w
d
In Exercises 920, use the Chain Rule to calculate f (c(t)).
dt
9. f (x, y) = 3x 7y, c(t) = (cos t, sin t), t =0
SOLUTION By the Chain Rule for paths, we have
d
f (c(t)) = f c(t) c
(t) (1)
dt
We compute the gradient and the derivative c
(t):
f f
f = , = 3, 7 , c
(t) = sin t, cos t
x y
We determine these vectors at t = 0:
We now use the Chain Rule for paths to compute the following derivative:
d
f (c(t)) = f c(2) c
(2) = 3, 7 4, 12 = 72
dt t=2
d
f (c(t)) = f c(t) c
(t) (1)
dt
We compute the gradient and c
(t):
f f
f = , = 2x 3y, 3x
x y
c
(t) = sin t, cos t
12. f (x, y) = x 2 3x y, t = 2
c(t) = (cos t, sin t),
SOLUTION In the previous exercise we found that
d
f (c(t)) = f c(t) c
(t) (1)
dt
We compute the gradient and c
(t):
f f
f = , = y cos(x y), x cos(x y)
x y
S E C T I O N 15.5 The Gradient and Directional Derivatives (ET Section 14.5) 683
c
(t) = 2e2t , 3e3t
d
f (c(t)) = f c(t) c
(t) (1)
dt
We compute the gradient and c
(t):
f f
f = , = 2x y 3 , 3x 2 y 2
x y
c
(t) = et , 2e2t
c
(4) = 2 4, 2 4 4 = 8, 4
f c(4) = f (16,0) = 1 0, 16 = 1, 16
We now use the Chain Rule for Paths to compute the following derivative:
d
f (c(t)) = f c(4) c
(4) = 1, 16 8, 4 = 8 64 = 56
dt t=4
c(0) = (0, 0)
c
(0) = 0, 4
f c(0) = f (0,0) = e0 1, 0 = 1, 0
Using the Chain Rule for Paths we obtain the following derivative:
d
f (c(t)) = f c(0) c
(0) = 1, 0 0, 4 = 0
dt t=0
2 2 2
c = cos , = ,
4 4 4 2 16
2
2
c
= sin , = ,
4 4 4 2 2
16
f c
= f 2 2 = 2, 2
4 2 , 16
Using the Chain Rule for Paths we obtain the following derivative:
d 16 2 8
f (c(t))
= fc c
= 2, 2 , = 1 + 1.546
dt
t= 4 4 4 2 2
c(1) = (1, 1, 0)
c
(1) = 2, 3, 1
gc(1) = g(1,1,0) = e0 1, 1, 1 = 1, 1, 1
Using the Chain Rule for Paths we obtain the following derivative:
d
g (c(t)) = gc(1) c
(1) = 1, 1, 1 2, 3, 1 = 2 + 3 + 1 = 6
dt t=1
d
g (c(t)) = g c(t) c
(t) (1)
dt
We compute the gradient and c
(t):
g g g
g = , , = yz 1 , x z 1 , x yz 2
x y z
c
(t) = et , 1, 2t
c(1) = (e, 1, 1)
c
(1) = e, 1, 2
gc(1) = g(e,1,1) = 1, e, e
gc(1) = 1, 2, 3, 5
c
(1) = 2, 3, 1, 1
We now use the Chain Rule for Paths to obtain the following derivative:
d
g (c(t)) = gc(1) c
(1) = 1, 2, 3, 5 2, 3, 1, 1 = 2 + 6 + 3 + 5 = 16
dt t=1
In Exercises 2130, calculate the directional derivative in the direction of v at the given point. Remember to normalize
the direction vector or use Eq. (2).
v 1 1 1
u= = (i + j) = i + j
v 2 2 2
23. f (x, y) = x 2 y 3 , v = i + j, P = ( 16 , 3)
SOLUTION We normalize v to obtain a unit vector u in the direction of v:
v 1 1 1
u= = (i + j) = i + j
v 2 2 2
v 1
u= = 1, 1
v 2
We compute the gradient of f (x, y) = sin(x y) at the point P:
f f 1 1
f = , = cos(x y), cos(x y) f
, = cos , cos = ,
x y 2 6 3 3 2 2
The directional derivative in the direction v is thus
1 1 1
Du f (P) = f
, u = , 1, 1 = 0
2 6 2 2 2
v 1
u= = 1, 1
v 2
2
26. f (x, y) = e x yy , v = 12, 5, P = (2, 2)
S E C T I O N 15.5 The Gradient and Directional Derivatives (ET Section 14.5) 687
v 2i j + k 1
u= = = (2i j + k)
v 6
2
22 + (1) + 12
31. Find the directional derivative of f (x, y) = x 2 + 4y 2 at P = (3, 2) in the direction pointing to the origin.
SOLUTION The direction vector is v = P O= 3, 2. A unit vector u in the direction v is obtained by normalizing
v. That is,
v 3, 2 1
u= = = 3, 2
v 3 +2
2 2 13
32. Find the directional derivative of f (x, y, z) = x y + z 3 at P = (3, 2, 1) in the direction pointing to the origin.
SOLUTION The direction vector is v = P O = 3, 2, 1. We normalize v to obtain a unit vector u in the direction v:
v 3, 2, 1 1
u= = = 3, 2, 1
v 9+4+1 14
33. A bug located at (3, 9, 4) begins walking in a straight line toward (5, 7, 3). At what rate is the bugs temperature
changing if the temperature is T (x, y, z) = xe yz ? Units are in meters and degrees Celsius.
SOLUTION The bug is walking in a straight line from the point P = (3, 9, 4) towards Q = (5, 7, 3), hence the rate of
change in the temperature is the directional derivative in the direction of v = P Q. We first normalize v to obtain
v = P Q= 5 3, 7 9, 3 4 = 2, 2, 1
v 2, 2, 1 1
u= = = 2, 2, 1
v 4+4+1 3
The rate of change of the bugs temperature at the starting point P is the directional derivative
1 e5
Du f (P) = T (3,9,4) u = e5 1, 3, 3 2, 2, 1 = 49.47
3 3
The answer is 49.47 degrees Celsius per meter.
34. Suppose that f P = 2, 4, 4. Is f increasing or decreasing at P in the direction v = 2, 1, 3?
SOLUTION We compute the derivative of f at P with respect to v:
f
= 1 e x y + xe x y 2x = e x y 1 + 2x 2
2 2 2
x
f
= xe x y
2
y
The gradient vector is thus
f f
= e x y 1 + 2x 2 , xe x y = e x y 1 + 2x 2 , x
2 2 2
f = ,
x y
At the point P = (1, 1) we have
f P = e0 1 + 2, 1 = 3, 1 f P = 32 + (1)2 = 10
(b) The rate of change of f in the direction of the gradient vector is the length of the gradient, that is, f P = 10.
(c) Let ev be the unit vector making an angle of 45 with f P . The rate of change of f in the direction of ev is the
directional derivative of f in the direction ev , which is the following dot product:
1
Dev f (P) = f P ev = f P ev cos 45 = 10 1 = 5 2.236
2
36. Let f (x, y, z) = sin(x y + z) and P = (0, 1, ). Calculate Du f (P) where u is a unit vector making an angle
= 30 with f P .
SOLUTION The directional derivative Du f (P) is the following dot product:
Du f (P) = f P u
Since u is a unit vector making an angle = 30 with f P , we have by the properties of the dot product
3
Du f (P) = f P u cos 30 = f P (1)
2
We now must find the gradient at P and its length:
f f f
f = , , = y cos(x y + z), x cos(x y + z), cos(x y + z) = cos(x y + z) y, x, 1
x y z
f (0,1, ) = cos 1, 0, 1 = 1 1, 0, 1 = 1, 0, 1
Hence,
f (0,1, ) = 12 + 02 + (1)2 = 2
37. Let T (x, y) be the temperature at location (x, y). Assume that T = y 4, x + 2y. Let c(t) = (t 2 , t) be a path
in the plane. Find the values of t such that
d
T (c(t)) = 0
dt
SOLUTION By the Chain Rule for Paths we have
d
T (c(t)) = Tc(t) c
(t) (1)
dt
3t 2 6t = 3t (t 2) = 0 t1 = 0, t2 = 2
f x (x, y, z) = 2x
f y (x, y, z) = 2y f P = f (3,1,2) = 6, 2, 4
f z (x, y, z) = 2z
x2 y2
40. Find the two points on the ellipsoid + + z 2 = 1 where the tangent plane is normal to v = 1, 1, 2.
4 9
x2 y2
SOLUTION The gradient f P is normal to the level surface f (x, y, z) = + + z 2 = 1. To find where the
4 9
tangent plane is normal to v = 1, 1, 2, we simply find where the gradient is parallel to v = 1, 1, 2. This occurs
2y
at f P = 1, 1, 2, and since f P = 2x
4 , 9 , 2z, we conclude that the two points of interest are (2, 9/2, 1) and
(2, 9/2, 1).
In Exercises 4144, find an equation of the tangent plane to the surface at the given point.
f P x 2, y 2, z 1 = 0 (1)
f P = 2 2, 6 2, 8 1 = 4, 12, 8
4, 12, 8 x 2, y 2, z 1 = 0
4(x 2) + 12(y 2) + 8(z 1) = 0
x 2 + 3(y 2) + 2(z 1) = 0
or
x + 3y + 2z = 10
f P x 2, y 1, z 1 = 0 (1)
f P = 9, 10, 5
9, 10, 5 x 2, y 1, z 1 = 0
9(x 2) + 10(y 1) + 5(z 1) = 0
or
9x + 10y + 5z = 33
3
43. x 2 + z 2 e yx = 13, P = 2, 3,
e
SOLUTION We compute the gradient of f (x, y, z) = x 2 + z 2 e yx at the point P = 2, 3, 3 :
e
f f f
f = , , = 2x z 2 e yx , z 2 e yx , 2ze yx
x y z
At the point P = 2, 3, 3 we have
e
9 9 3
f P = 4 e, e, 2 e = 5, 9, 6 e
e e e
The equation of the tangent plane at P is
3
f P x 2, y 3, z = 0
e
That is,
3
5(x 2) + 9(y 3) + 6 e z =0
e
or
5x + 9y + 6 ez = 35
f P (x 3, y 1, z 6.1876) = 0 (1)
or
45. Verify what is clear from Figure 17: Every tangent plane to the cone x 2 + y 2 z 2 = 0 passes through the origin.
FIGURE 17 Graph of x 2 + y 2 z 2 = 0.
SOLUTION The equation of the tangent plane to the surface f (x, y, z) = x 2 + y 2 z 2 = 0 at the point P =
(x 0 , y0 , z 0 ) on the surface is
f P x x 0 , y y0 , z z 0 (1)
f P = 2x 0 , 2y0 , 2z 0
2x 0 , 2y0 , 2z 0 x x 0 , y y0 , z z 0 = 0
x 0 (x x 0 ) + y0 (y y0 ) z 0 (z z 0 ) = 0
x 0 x + y0 y z 0 z = x 02 + y02 z 02
Since P = (x 0 , y0 , z 0 ) is on the surface, we have x 02 + y02 z 02 = 0. The equation of the tangent plane is thus
x 0 x + y0 y z 0 z = 0
y
4
x
4 2 2 4
2
47. Find a function f (x, y, z) such that f is the constant vector 1, 3, 1.
SOLUTION The gradient of f (x, y, z) must satisfy the equality
f f f
f = , , = 1, 3, 1
x y z
Equating corresponding components gives
f
=1
x
f
=3
y
f
=1
z
One of the functions that satisfies these equalities is
f (x, y, z) = x + 3y + z
We differentiate f with respect to y and substitute in the second equation. This gives
f
= (yx + g(y)) = x + g
(y)
y y
Hence,
f (x, y) = yx + C
f = fx , f y = y2, x
f f P v 6
f f P v
Hence,
55. Find the unit vector n normal to the surface z 2 2x 4 y 4 = 16 at P = (2, 2, 8) that points in the direction of the
x y-plane.
SOLUTION The gradient vector f P is normal to the surface f (x, y, z) = z 2 2x 4 y 4 = 16 at P. We find this
vector:
f f f
f = , , = 8x 3 , 4y 3 , 2z f (2,2,8) = 8 23 , 4 23 , 2 8 = 64, 32, 16
x y z
We normalize to obtain a unit vector normal to the surface:
fP 64, 32, 16 64, 32, 16 1
= = = 4, 2, 1
f P 16 21 21
(64)2 + 322 + 162
56. Suppose, in the previous exercise, that a particle located at the point P = (2, 2, 8) travels toward the x y-plane in the
direction normal to the surface.
(a) Through which point Q on the x y-plane will the particle pass?
(b) Suppose the axes are calibrated in centimeters. Determine the path c(t) of the particle if it travels at a constant speed
of 8 cm/s. How long will it take the particle to reach Q?
SOLUTION
(a) The particle travels along the line through P = (2, 2, 8) in the direction (4, 2, 1). The vector parametrization of
this line is
We must find the point where this line intersects the x y-plane. At this point the z-component is zero. Hence,
8t =0 t =8
The particle will pass through the point Q = (34, 18, 0) on the x y-plane.
(b) If v is a direction vector of the line P Q, so that v = 8, the following parametrization of the line has constant speed
8:
c(t) = 2, 2, 8 + tv
(This has speed 8 because c
(t) = v = 8). In the previous exercise, we found the unit vector n = 1 4, 2, 1,
21
therefore we use the direction vector v = 8n = 8 4, 2, 1, obtaining the following parametrization of the line:
21
8 32 16 8t
c(t) = 2, 2, 8 + t 4, 2, 1 = 2 + t, 2 + t, 8
21 21 21 21
To find the time needed for the particle to reach Q if it travels along c(t), we first compute the distance P Q:
P Q = (34 2)2 + (18 2)2 + (0 8)2 = 1344 = 8 21
(a) We compute the partial derivatives of f (x, y) = tan1 xy . Using the Chain Rule we get
f 1 1 y
= 2 y = 2
x x x + y2
1+ y
f 1 x x
= 2 2 = 2
y x y x + y2
1+ y
(c) Note that f is not defined for y = 0. For x = 0, the level curve of f is the y-axis, and the gradient vector is 1y , 0,
which is perpendicular to the y-axis. For y = 0 and x = 0, the level curves of f are the curves where f (x, y) is constant.
That is,
x
tan1 =k
y
x
= tan k (for k = 0)
y
1
y= x
tan k
We conclude that the lines y = mx, m = 0, are level curves for f .
(d) By part (c), the level curve through P = (x 0 , y0 ) is the line y = xy00 x. This line has a direction vector 1, xy00 . The
gradient at P is, by part (a), f P = 2 1 2 y0 , x 0 . We verify that the two vectors are orthogonal:
x0 +y0
y y0 1 1 x 0 y0
1, 0 f P = 1, y0 , x 0 = 2 y0 =0
x0 x0 x 2 + y 2 x 0 + y02 x0
0 0
Since the dot products is zero, the two vectors are orthogonal as expected (Theorem 6).
58. Suppose that the intersection of two surfaces F(x, y, z) = 0 and G(x, y, z) = 0 is a curve C and let P be a
point on C. Explain why the vector v = FP G P is a direction vector for the tangent line to C at P.
S E C T I O N 15.5 The Gradient and Directional Derivatives (ET Section 14.5) 697
SOLUTION The gradient F P is orthogonal to all the curves in the level surface F(x, y, z) = 0 passing through P.
Similarly, G P is orthogonal to all the curves in the level surface G(x, y, z) = 0 passing through P. Therefore, both
FP and G P are orthogonal to the intersection curve C at P, hence the cross product FP G P is parallel to the
tangent line to C at P.
59. Let C be the curve of intersection of the two surfaces x 2 + y 2 + z 2 = 3 and (x 2)2 + (y 2)2 + z 2 = 3. Use the
result of Exercise 58 to find parametric equations of the tangent line to C at P = (1, 1, 1).
SOLUTION The parametric equations of the tangent line to C at P = (1, 1, 1) are
where v = a, b, c is a direction vector for the line. By Exercise 58 v may be chosen as the following cross product:
v = FP G P (2)
Fx (x, y, z) = 2x
Fy (x, y, z) = 2y F P = 2 1, 2 1, 2 1 = 2, 2, 2
Fz (x, y, z) = 2z
G x (x, y, z) = 2(x 2)
G y (x, y, z) = 2(y 2) G P = 2(1 2), 2(1 2), 2 1 = 2, 2, 2
G z (x, y, z) = 2z
Hence,
i j k
v = 2, 2, 2 2, 2, 2 = 2 2 2 = (4 + 4)i (4 + 4)j + (4 + 4)k = 8i 8j = 8, 8, 0
2 2 2
Therefore, v = a, b, c = 8, 8, 0, yielding a = 8, b = 8, c = 0. Substituting in (1) gives the following equations
of the tangent line: x = 1 + 8t, y = 1 8t, z = 1.
60. Let C be the curve of intersection of the two surfaces x 3 + 2x y + yz = 7 and 3x 2 yz = 1. Find the parametric
equations of the tangent line to C at P = (1, 2, 1).
SOLUTION The parametric equations of the tangent line to C at P = (1, 2, 1) are
where v = a, b, c is a direction vector for the line. By Exercise 58, v may be chosen as the cross product:
v = FP G P (2)
a = 4, b = 26, c = 25.
Substituting in (1) gives the following parametric equations of the tangent line:
( f g) = f g + g f
(b) By the limit definition of the directional derivative, and for u = u 1 , u 2 a unit vector, we have
1/3
f (tu 1 , tu 2 ) f (0, 0) t 2u1u2 0 u u
Du f (0, 0) = lim = lim = lim 11/32
t0 t t0 t t0 t
This limit does not exist unless u 1 = 0 or u 2 = 0. u 1 = 0 corresponds to the unit vector j, and u 2 = 0 corresponds to
the unit vector i.
S E C T I O N 15.5 The Gradient and Directional Derivatives (ET Section 14.5) 699
(c) If f was differentiable at (0, 0), then Du f (0, 0) would exist for any vector u. Therefore, using the result obtained in
part (b), f is not differentiable at (0, 0).
66. Use the definition of differentiability to show that if f (x, y) is differentiable at (0, 0) and
then
f (x, y)
lim =0 7
(x,y)(0,0) x 2 + y2
SOLUTION If f (x, y) is differentiable at (0, 0), then there exists a function (x, y) satisfying lim(x,y)(0,0) (x, y) =
0 such that
f (x, y) = L(x, y) + (x, y) x 2 + y 2 (1)
L(x, y) = f (0, 0) + f x (0, 0)x + f y (0, 0)y = f x (0, 0)x + f y (0, 0)y
Therefore,
f (x, y) f x (0, 0)x f y (0, 0)y
lim = lim (x, y) = 0
(x,y)(0,0) x 2 + y2 (x,y)(0,0)
67. This exercise shows that there exists a function which is not differentiable at (0, 0) even though all directional
x2 y
derivatives at (0, 0) exist. Define f (x, y) = 2 for (x, y) = 0 and f (0, 0) = 0.
x + y2
(a) Use the limit definition to show that Dv f (0, 0) exists for all vectors v. Show that f x (0, 0) = f y (0, 0) = 0.
(b) Prove that f is not differentiable at (0, 0) by showing that Eq. (7) does not hold.
SOLUTION
(a) Let v = 0 be the vector v = v1 , v2 . By the definition of the derivative Dv f (0, 0), we have
(tv1 ) tv2
2
f (tv1 , tv2 ) f (0, 0) 0
(tv1 )2 +(tv2 )2
Dv f (0, 0) = lim = lim
t0 t t0 t
t 3 v 2 v2 v12 v2 v12 v2
= lim 1 = lim = (1)
t0 t 3 v 2 + v 2 t0 v 2 + v 2 v12 + v22
1 2 1 2
12 0
f x (0, 0) = 2 =0
1 + 02
Substituting v1 = 0, v2 = 1 in (1) gives
02 1
f y (0, 0) = 2 =0
0 + 12
Also f (0, 0) = 0, therefore for (x, y) = (0, 0) we have
x y 2
f (x, y) f x (0, 0)x f y (0, 0)y 0x 0y x2 y
x 2 +y 2
lim = lim = lim 3
(x,y)(0,0) x 2 + y2 (x,y)(0,0) x 2 + y2 (x,y)(0,0) 2
(x + y 2 ) 2
700 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
We compute the limit along the line y = 3x:
3 3
x2 y x 2 3x 3x 3x 3
lim = lim
3/2 = lim
3/2 = lim = = 0
(x,y)(0,0) 3/2 3 8
(x 2 + y 2 ) x0 2 x0 4x 2 x0 8x
along y= 3x x +
2 3x
Since this limit is not zero, f does not satisfy Eq. (7), hence f is not differentiable at (0, 0).
68. Prove that if f (x, y) is differentiable and f (x,y) = 0 for all (x, y), then f is constant.
SOLUTION Since f = f x , f y = 0, 0 for all (x, y), we have
Let Q 0 = (x 0 , y0 ) be a fixed point and let P = (x 1 , y1 ) be any other point. Let c(t) = x(t), y(t) be a parametric
equation of the line joining Q 0 and P, with P = c(t1 ) and Q 0 = c(t0 ). We define the following function:
F(t) is defined for all t, since f (x, y) is defined for all (x, y). By the Chain Rule we have
dx dy
F
(t) = f x (x(t), y(t)) + f y (x(t), y(t))
dt dt
Combining with (1) we get F
(t) = 0 for all t. We conclude that F(t) = const. That is, f is constant on the line c(t). In
particular, f (P) = f (Q 0 ). Since P is any point, it follows that f (x, y) is a constant function.
69. Prove the following Quotient Rule where f, g are differentiable:
f g f f g
=
g g2
f f g f g
f f f f g x f gx g y f y g z f z
= , , = , ,
g x g y g z g g2 g2 g2
f f f
g x g y g z f g
f gx
g
y f z g f f f f g g g
= , , , , = 2 , , 2 , ,
g2 g2 g2 g2 g2 g2 g x y z g x y z
g f g f f g
= 2 f 2 g =
g g g2
In Exercises 7072, a path c(t) = (x(t), y(t)) follows the gradient of a function f (x, y) if the tangent vector c
(t) points
in the direction of f for all t. In other words, c
(t) = k(t) f c(t) for some positive function k(t). Note that in this case,
c(t) crosses each level curve of f (x, y) at a right angle.
y
(t) fy
70. Show that if the path c(t) = (x(t), y(t)) follows the gradient of f (x, y), then
= .
x (t) fx
SOLUTION Since c(t) follows the gradient of f (x, y), we have
c
(t) = k(t) f c(t) = k(t) f x (c(t)) , f y (c(t))
Hence,
y
(t) fy
=
x
(t) fx
S E C T I O N 15.5 The Gradient and Directional Derivatives (ET Section 14.5) 701
71. Find a path of the form c(t) = (t, g(t)) passing through (1, 2) that follows the gradient of f (x, y) = 2x 2 + 8y 2
(Figure 18). Hint: Use Separation of Variables.
y
2
1
x
2 1 1 2
1
2
SOLUTION By the previous exercise, if c(t) = (x(t), y(t)) follows the gradient of f , then
dy y
(t) fy
=
= (1)
dx x (t) fx
We find the partial derivatives of f :
2 2
fy = 2x + 8y 2 = 16y, fx = 2x + 8y 2 = 4x
y x
Substituting in (1) we get
dy 16y 4y
= =
dx 4x x
We solve the differential equation using separation of variables. We obtain
dy dx
=4
y x
dy dx
=4
y x
ln y = 4 ln x + c = ln x 4 + c
or
y = eln x +c = ec x 4
4
y = kx 4
Since we want the path to pass through (1, 2), there must be a solution t for the equation
t, kt 4 = (1, 2)
or
t =1
k 14 = 2 k=2
kt 4 = 2
72. Find the curve y = g(x) passing through (0, 1) that crosses each level curve of f (x, y) = y sin x at a right
angle. If you have a computer algebra system, graph y = g(x) together with the level curves of f .
SOLUTION Using f x = y cos x, f y = sin x, and y(0) = 1, we get
dy tan x
= y(0) = 1
dx y
We solve the differential equation using separation of variables:
y d y = tan x d x
y d y = tan x d x
1 2
y = ln | cos x| + k
2
y 2 = 2 ln | cos x| + k = ln cos2 x + k
y = ln cos2 x + k
Since y(0) = 1 > 0, the appropriate sign is the positive sign. That is,
y = ln cos2 x + k (1)
Hence,
k=1
The following figure shows the graph of the curve (3) together with some level curves of f .
y
y sin x = c
c = 0.15
y = 1-ln (cos2x)
x
0
SOLUTION The composite function f (u, v) is obtained by replacing u and v in the formula for f (u, v) by the corre-
sponding functions u = r s and v = r + s. This gives
f u(r, s), v(r, s) = u(r, s)ev(r,s) = r ser +s
u(1, 1) = 1 1 = 1; v(1, 1) = 1 + 1 = 2
f
4. According to the Chain Rule, is equal to (choose correct answer):
r
f x f x
(a) +
x r x s
f x f y
(b) +
x r y r
f r f s
(c) +
r x s x
SOLUTION For a function f (x, y) where x = x(r, s) and y = y(r, s), the Chain Rule states that the partial derivative
f
r is as given in (b). That is,
f x f y
+
x r y r
5. Suppose that x, y, z are functions of the independent variables u, v, w. Given a function f (x, y, z), which of the
f
following terms appear in the Chain Rule expression for ?
w
f x f w
(a) (b)
v v w x
f z f v
(c) (d)
z w v w
f
SOLUTION By the Chain Rule, the derivative w is
f f x f y f z
= + +
w x w y w z w
Therefore (c) is the only correct answer.
x f
6. With notation as in the previous question, does appear in the Chain Rule expression for ?
v u
SOLUTION The Chain Rule expression for uf is
f f x f y f z
= + +
u x u y u z u
The derivative vx does not appear in differentiating f with respect to the independent variable u.
Exercises
1. Let f (x, y, z) = x 2 y 3 + z 4 and x = s 2 , y = st 2 , and z = s 2 t.
f f f
(a) Calculate the primary derivatives , , .
x y z
x y z
(b) Calculate , , .
s s s
704 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
f
(c) Compute using the Chain Rule:
s
f f x f y f z
= + +
s x s y s z s
Express the answer in terms of the independent variables s, t.
SOLUTION
To express the answer in terms of the independent variables s, t we substitute x = s 2 , y = st 2 , z = s 2 t. This gives
f 3 2 2 3
= 4s 2 (st 2 ) s + 3(s 2 ) (st 2 ) t 2 + 8(s 2 t) st = 4s 6 t 6 + 3s 6 t 6 + 8s 7 t 4 = 7s 6 t 6 + 8s 7 t 4 .
s
x y
= 2v, = 1.
v v
Substituting these derivatives and the primary derivatives computed in part (a) in the Chain Rule (1) gives
f
= cos(y) 2v x sin(y) (1) = 2v cos(y) + x sin(y)
v
(c) We substitute u = 2, v = 1 in x = u 2 + v 2 and y = u v, and determine (x, y) for (u, v) = (2, 1). This gives
x = 22 + 12 = 5, y = 2 1 = 1.
In Exercises 310, use the Chain Rule to calculate the partial derivatives. Express the answer in terms of the independent
variables.
f f
3. , : f (x, y, z) = x y + z 2 , x = s 2 , y = 2r s, z = r 2
s r
S E C T I O N 15.6 The Chain Rule (ET Section 14.6) 705
Step 3. Express the answer in terms of r and s. We substitute x = s 2 , y = 2r s, and z = r 2 in sf and rf in step 2, to
obtain
f
= 2r s 2s + s 2 2r = 4r s 2 + 2r s 2 = 6r s 2 .
s
f
= 2s 2 s + 4r 2 r = 2s 3 + 4r 3 .
r
f f
4. , : f (x, y, z) = x y + z 2 , x = r + s 2t, y = 3r t, z = s 2
r t
SOLUTION We use the following steps:
Step 1. Compute the primary derivatives. The primary derivatives of f (x, y, z) = x y + z 2 are
f f f
= y, = x, = 2z
x y z
Step 2. Apply the Chain Rule. By the Chain Rule,
f f x f y f z x y z
= + + =y +x + 2z (1)
r x r y r z r r r r
f f x f y f z x y z
= + + =y +x + 2z (2)
t x t y t z t t t t
We compute the partial derivatives of x, y with respect to r and t:
x y z
= 1, = 3t, =0
r r r
x y z
= 2, = 3r, =0
t t t
Substituting in (1) and (2), we get
f
= y + 3t x + 2z 0 = y + 3xt
r
f
= y (2) + x 3r + 2z 0 = 2y + 3xr
t
Step 3. Express the answer in terms of r and t. We substitute x = r + s 2t, y = 3r t, and z = s 2 in rf and tf
obtained in step 2. This gives
f
= 3r t + 3(r + s 2t)t = 3r t + 3r t + 3st 6t 2 = 6r t + 3st 6t 2
r
706 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
f
= 2 3r t + 3(r + s 2t)r = 6r t + 3r 2 + 3sr 6tr = 12r t + 3r s + 3r 2
t
g g
5. , : g(x, y) = cos(x 2 y 2 ), x = 2u 3v, y = 5u + 8v
u v
SOLUTION We use the following steps:
Step 1. Compute the primary derivatives. The primary derivatives of g(x, y) = cos(x 2 y 2 ) are
g g
= 2x sin(x 2 y 2 ), = 2y sin(x 2 y 2 )
x y
Step 2. Apply the Chain Rule. By the Chain Rule,
g g x g y x y
= + = 2x sin(x 2 y 2 ) + 2y sin(x 2 y 2 )
u x u y u u u
g g x g y x y
= + = 2x sin(x 2 y 2 ) + 2y sin(x 2 y 2 )
v x v y v v v
We find the partial derivatives of x and y:
x y
= 2, = 5
u u
x y
= 3, =8
v v
We substitute these derivatives to obtain
g
= 4x sin(x 2 y 2 ) 10y sin(x 2 y 2 ) = (4x + 10y) sin(x 2 y 2 ) (1)
u
g
= 6x sin(x 2 y 2 ) + 16y sin(x 2 y 2 ) = (6x + 16y) sin(x 2 y 2 ) (2)
v
Step 3. Express the answer in terms of u and v. We substitute x = 2u 3v, y = 5u + 8v in (1) and (2), to obtain
g
= (8u 12v 50u + 80v) sin (2u 3v)2 (5u + 8v)2
u
= (42u 68v) sin 68uv 21u 2 55v 2
g
= (12u 18v 80u + 128v) sin (2u 3v)2 (5u + 8v)2
v
= (110v 68u) sin 68uv 21u 2 55v 2
h x
6. : h(x, y) = , x = t1 t2 , y = t12 t2
t2 y
SOLUTION We use the following steps:
Step 1. Compute the primary derivatives. The primary derivatives of h(x, y) = xy are
h 1 h x
= , = 2
x y y y
Step 2. Apply the Chain Rule. By the Chain Rule,
h h x h y 1 x x y
= + = 2
t2 x t2 y t2 y t2 y t2
We compute the partial derivatives of x and y with respect to t2 :
x y
= t1 , = t12
t2 t2
Hence,
h t x
= 1 2 t12
t2 y y
S E C T I O N 15.6 The Chain Rule (ET Section 14.6) 707
h t t1 t2 t12 1 1
= 21 = =0
t2 t1 t2 2
(t1 t2 )
2 t t
12 t 1 t2
h = 0 (as
Remark: Notice that h x(t1 , t2 ), y(t1 , t2 ) = h(t1 , t2 ) = t12t2 = t11 . h(t1 , t2 ) is independent of t2 , hence t
t1 t2 2
obtained in our computations).
F F
7. , : F(u, v) = eu+v , u = x 2 , v = x y
x y
SOLUTION We use the following steps:
Step 1. Compute the primary derivatives. The primary derivatives of F(u, v) = eu+v are
f f
= eu+v , = eu+v
u v
Step 2. Apply the Chain Rule. By the Chain Rule,
F F u F v u v u v
= + = eu+v + eu+v = eu+v +
x u x v x x x x x
F F u F v u v u v
= + = eu+v + eu+v = eu+v +
y u y v y y y y y
We compute the partial derivatives of u and v with respect to x and y:
u v
= 2x, =y
x x
u v
= 0, =x
y y
We substitute to obtain
F
= (2x + y)eu+v (1)
x
F
= xeu+v (2)
y
Step 3. Express the answer in terms of x and y. We substitute u = x 2 , v = x y in (1) and (2), obtaining
F F
= (2x + y)e x +x y , = xe x +x y .
2 2
x y
f
8. : f (x, y) = x 2 + y 2 , x = eu+v , y = u + v
u
SOLUTION We use the following steps:
Step 1. Compute the primary derivatives. The primary derivatives of f (x, y) = x 2 + y 2 are
f f
= 2x, = 2y
x y
Step 2. Apply the Chain Rule. By the Chain Rule,
f f x f y x y
= + = 2x + 2y
u x u y u u u
y
We compute ux and u :
x y
= eu+v , =1
u u
Hence,
f
= 2xeu+v + 2y (1)
u
Step 3. Express the answer in terms of u and v. We substitute x = eu+v and y = u + v in (1) to obtain
f
= 2eu+v eu+v + 2(u + v) = 2 e2(u+v) + u + v
u
708 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
f f
9. , : f (r, ) = r sin2 , x = r cos , y = r sin
x y
SOLUTION We use the following steps:
Step 1. Compute the primary derivatives. The primary derivatives of f (r, ) = r sin2 are
f f
= sin2 , = r 2 sin cos = r sin 2
r
Step 2. Apply the Chain Rule. By the Chain Rule,
f f r f r
= + = sin2 + r sin 2 (1)
x r x x x x
f f r f r
= + = sin2 + r sin 2 (2)
y r y y y y
We compute the partial derivatives of r and with respect to x and y. Since r 2 = x 2 + y 2 , we have
r r x
2r = 2x =
x x r
r r y
2r = 2y =
y y r
1 y y cos2
= 2 =
cos2 x x x x2
1 1 cos2
= =
cos2 y x y x
Substituting these derivatives in (1) and (2) gives
f 2 x y cos2 x sin2 yr sin 2 cos2
= sin + r sin 2 = (3)
x r x 2 r x2
y y
sin = =
r x + y2
2
2x y
sin 2 = 2 sin cos = 2
x + y2
We substitute in (3) and (4) to obtain
f x y2 y x 2 + y2 2x y x2
= 2
x x 2 + y2 x + y2 x2 x 2 + y2 x 2 + y2
x y2 2y 2 x x y 2
= 3/2
3/2
= 3/2
(x 2 + y 2 ) (x 2 + y 2 ) (x 2 + y 2 )
f y y2 x 2 + y2 2x y x2
= 2 + 2 2
y x +y
2 2 x +y 2 x x +y 2 x + y2
y3 2x 2 y y(y 2 + 2x 2 )
= 3/2
+ 3/2
= 3/2
(x 2 + y 2 ) (x 2 + y 2 ) (x 2 + y 2 )
f f
10. , : f (x, y, z) = x y z 2 , x = r cos , y = cos2 , z = r
r
SOLUTION We use the following steps:
S E C T I O N 15.6 The Chain Rule (ET Section 14.6) 709
Step 1. Compute the primary derivatives. The primary derivatives of f (x, y, z) = x y z 2 are
f f f
= y, = x, = 2z
x y z
Step 2. Apply the Chain Rule. By the Chain Rule,
f f x f y f z x y z
= + + =y +x 2z
x y z
f f x f y f z x y z
= + + =y +x 2z
r x r y r z r r r r
We compute the partial derivatives of x, y, and z with respect to and r :
x y z
= r sin , = 2 cos sin = sin 2 , =0
x y z
= cos , = 0, =1
r r r
Thus,
f
= yr sin x sin 2 (1)
f
= y cos 2z (2)
r
Step 3. Express the answer in terms of and r . We substitute x = r cos , y = cos2 , and z = r in (1), (2) to obtain
f 1 3
= r cos2 sin r cos sin 2 = r cos sin 2 r cos sin 2 = cos sin 2
2 2
f
= cos2 cos 2r = cos3 2r
r
In Exercises 1116, use the Chain Rule to evaluate the partial derivative at the point specified.
f f
11. and at (u, v) = (1, 1), where f (x, y, z) = x 3 + yz 2 , x = u 2 + v, y = u + v 2 , z = uv.
u v
SOLUTION The primary derivatives of f (x, y, z) = x 3 + yz 2 are
f f f
= 3x 2 , = z2, = 2yz
x y z
By the Chain Rule we have
f f x f y f z x y z
= + + = 3x 2 + z2 + 2yz (1)
u x u y u z u u u u
f f x f y f z x y z
= + + = 3x 2 + z2 + 2yz (2)
v x v y v z v v v v
We compute the partial derivatives of x, y, and z with respect to u and v:
x y z
= 2u, = 1, =v
u u u
x y z
= 1, = 2v, =u
v v v
Substituting in (1) and (2) we get
f
= 6x 2 u + z 2 + 2yzv (3)
u
f
= 3x 2 + 2vz 2 + 2yzu (4)
v
We determine (x, y, z) for (u, v) = (1, 1):
Finally, we substitute (x, y, z) = (0, 0, 1) and (u, v) = (1, 1) in (3), (4) to obtain the following derivatives:
f
= 6 02 (1) + 12 + 2 0 1 (1) = 1
u (u,v)=(1,1)
f
= 3 02 + 2 (1) 12 + 2 0 1 (1) = 2
v (u,v)=(1,1)
f
12. at (r, s) = (1, 0), where f (x, y) = ln(x y), x = 3r + 2s, y = 5r + 3s.
s
SOLUTION The primary derivatives of f (x, y) = ln(x y) are
f y 1 f x 1
= = , = =
x xy x y xy y
By the Chain Rule we have
f f x f y 1 x 1 y
= + = + (1)
s x s y s x s y s
We compute sx and sy :
x y
= 2, =3
s s
Substituting in (1) we get
f 2 3
= + (2)
s x y
We now must determine (x, y) for (s, r ) = (1, 0):
x = 3 0 + 2 1 = 2, y =50+31=3
g 1
13. at (r, ) = (2 2, 4 ), where g(x, y) = , x = r sin , y = r cos .
x + y2
SOLUTION We compute the primary derivatives of g(x, y) = 1 :
x+y 2
g 1 g 2y
= , =
x (x + y 2 )
2 y (x + y 2 )
2
x y
= r cos , = r sin
Hence,
g r
= (cos 2y sin ) (1)
(x + y 2 )
2
At the point (r, ) = 2 2, 4 , we have x = 2 2 sin 4 = 2 and y = 2 2 cos 4 = 2. Substituting (r, ) = 2 2, 4
and (x, y) = (2, 2) in (1) gives the following derivative:
g 2 2 2 1 4 1
= cos 4 sin = = .
(r, )= 2 2,
(2 + 2 )
2 2 4 4 18 2 2 6
4
S E C T I O N 15.6 The Chain Rule (ET Section 14.6) 711
g
14. at s = 4, where g(x, y) = x 2 y 2 , x = s 2 + 1, y = 1 2s.
s
SOLUTION We find the primary derivatives of g(x, y) = x 2 y 2 :
g g
= 2x, = 2y
x y
Applying the Chain Rule gives
g g d x g d y dx dy
= + = 2x 2y (1)
s x ds y ds ds ds
dx dy
= 2s, = 2
ds ds
Substituting in (1) we obtain
g
= 4xs + 4y (2)
s
We now determine (x, y) for s = 4:
x = 42 + 1 = 17, y = 1 2 4 = 7
g
15. at (u, v) = (0, 1), where g(x, y) = x 2 y 2 , x = eu cos v, y = eu sin v.
u
SOLUTION The primary derivatives of g(x, y) = x 2 y 2 are
g g
= 2x, = 2y
x y
By the Chain Rule we have
g g x g y x y
= + = 2x 2y (1)
u x u y u u u
y
We find ux and u :
x y
= eu cos v, = eu sin v
u u
Substituting in (1) gives
g
= 2xeu cos v 2yeu sin v = 2eu (x cos v y sin v) (2)
u
We determine (x, y) for (u, v) = (0, 1):
Finally, we substitute (u, v) = (0, 1) and (x, y) = (cos 1, sin 1) in (2) and use the identity cos2 sin2 = cos 2 , to
obtain the following derivative:
g
= 2e0 cos2 1 sin2 1 = 2 cos 2 1 = 2 cos 2
u (u,v)=(0,1)
h
16. at (q, r ) = (3, 2), where h(u, v) = uev , u = q 3 , v = qr 2 .
q
SOLUTION We first find the primary derivatives of h(u, v) = uev :
h h
= ev , = uev
u v
712 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
u v
= 3q 2 , = r2
q q
Substituting in (1) gives
h
= ev 3q 2 + ur 2 (2)
q
We now determine (u, v) for (q, r ) = (3, 2):
u = 33 = 27, v = 3 22 = 12
17. The temperature at a point (x, y) is T (x, y) = 20 + 0.1(x 2 x y) (degrees Celsius). A particle moves clockwise
along the unit circle at unit speed (1 cm/s). How fast is the particles temperature changing at time t = ?
SOLUTION The particle moves along the unit circle at unit speed, hence its trajectory is parametrized by the arc length
parametrization of the unit circle. That is,
T T
= 0.2x 0.1y, = 0.1x
x y
By the Chain Rule we have
dT T dx T dy dx dy
= + = (0.2x 0.1y) 0.1x (1)
dt x dt y dt dt dt
dx dy
= sin t, = cos t
dt dt
Substituting in (1) gives
dT
= (0.2x 0.1y) sin t 0.1x cos t (2)
dt
We determine (x, y) for t = :
x = cos = 1, y = sin = 0
1
u2 = u r2 + 2 u 2 7
r
SOLUTION By the Chain Rule we have
u = u x x + u y y (1)
u r = u x xr + u y yr (2)
S E C T I O N 15.6 The Chain Rule (ET Section 14.6) 713
Since x = r cos and y = r sin , the partial derivatives of x and y with respect to r and are
x = r sin , y = r cos
xr = cos , yr = sin
We now solve these equations for u x and u y in terms of u and u r . Multiplying (3) by ( sin ) and (4) by r cos and
adding the resulting equations gives
19. Let u(r, ) = r 2 cos2 . Use Eq. (7) to compute u2 . Then compute u2 directly by observing that u(x, y) =
x 2 and compare.
SOLUTION By Eq. (7) we have
1
u2 = u r2 + 2 u 2
r
We compute the partial derivatives of u(r, ) = r 2 cos2 :
That is,
We now compute u2 directly. We first express u(r, ) as a function of x and y. Since x = r cos , we have
u(x, y) = x 2
f f x f y f f f f
= + = 1+ 1= +
s x s y s x y x y
f f x f y f f f f
= + = 1+ (1) =
t x t y t x y x y
f f f f f f
21. Express the derivatives , , in terms of , , , where (, , ) are spherical coordinates.
x y z
SOLUTION The spherical coordinates are
22. Suppose that z is defined implicitly as a function of x and y by the equation F(x, y, z) = x z 2 + y 2 z + x y 1 = 0.
(a) Calculate Fx , Fy , Fz .
z z
(b) Use Eq. (6) to calculate and .
x y
S E C T I O N 15.6 The Chain Rule (ET Section 14.6) 715
SOLUTION
(a) The partial derivatives of F are
Fx = z 2 + y, Fy = 2yz + x, Fz = 2x z + y 2
z Fx z2 + y
= =
x Fz 2x z + y 2
z Fy 2yz + x
= =
y Fz 2x z + y 2
z z
23. Calculate and at the points (3, 2, 1) and (3, 2, 1), where z is defined implicitly by the equation z 4 + z 2 x 2
x y
y 8 = 0.
SOLUTION For F(x, y, z) = z 4 + z 2 x 2 y 8 = 0, we use the following equalities, (Eq. (6)):
z Fx z Fy
= , = (1)
x Fz y Fz
The partial derivatives of F are
Fx = 2z 2 x, Fy = 1, Fz = 4z 3 + 2zx 2
z 2z 2 x zx
= 3 = 2
x 4z + 2zx 2 2z + x 2
z 1
= 3
y 4z + 2zx 2
At the point (3, 2, 1), we have
z 13 3 z 1 1
= = , = =
x (3,2,1) 2 12 + 32 11 y (3,2,1) 4 13 + 2 1 32 22
z Fx
= (1)
x Fz
We compute the partial derivatives of F:
Fx = 2x y + z 2 , Fz = y 2 + 2x z
z 2x y + z 2
=
x 2x z + y 2
w
25. , x 2 w + w 3 + wz 2 + 3yz = 0
z
716 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
F(x, w, z) = x 2 w + w 3 + wz 2 + 3yz
Fw = x 2 + 3w 2 + z 2 , Fz = 2wz + 3y
z
26. , e x y + sin(x z) + y = 0
y
SOLUTION We use Eq. (6):
z Fy
= (1)
y Fz
The partial derivatives of F(x, y, z) = e x y + sin(x z) + y are
Fy = xe x y + 1, Fz = x cos(x z)
r t
27. and , r 2 = te s/r
t r
SOLUTION We use the formulas obtained by implicit differentiation of F(r, s, t) = r 2 tes/r (Eq. (6)):
r Ft t Fr
= , = (1)
t Fr r Ft
The partial derivatives of F are
s st
Fr = 2r tes/r 2 = 2r + 2 es/r
r r
Ft = es/r
r es/r r 2 es/r
= =
t st
2r + 2 es/r 2r 3 + stes/r
r
w 1 1
28. , + 2 = 1 at (x, y, w) = (1, 1, 1)
y w2 + x 2 w + y2
SOLUTION Using the formula obtained by implicit differentiation (Eq. (6)), we have
w Fy
= (1)
y Fw
2y 2w 2w
Fy = 2
, Fw = 2
2
(w 2 + y 2 ) (w 2 + x 2 ) (w 2 + y 2 )
We substitute in (1) to obtain
2y
2 2
w (w2 +y 2 )
2 y(w 2 + x 2 ) y(w 2 + x 2 )
= 2w
= =
2
y 2
2w
2
2
w(w 2 + y 2 ) + w(w 2 + x 2 )
2 2
w (w 2 + y 2 ) + (w 2 + x 2 )
(w2 +x 2 ) (w2 +y 2 )
S E C T I O N 15.6 The Chain Rule (ET Section 14.6) 717
U T
29. and , (T U V )2 ln(W U V ) = 1 at (T, U, V , W ) = (1, 1, 2, 4)
T U
SOLUTION Using the formulas obtained by implicit differentiation (Eq. (6)) we have,
U FT T FU
= , = (1)
T FU U FT
FT = 2U (T U V ) ln(W U V )
V V (T U V )
FU = 2T (T U V ) ln(W U V ) + (T U V )2 = (T U V ) 2T ln(W U V )
W UV W UV
At the point (T, U, V , W ) = (1, 1, 2, 4) we have
FT = 2(1 2) ln(4 2) = 2 ln 2
2(1 2)
FU = (1 2) 2 ln(4 2) = (2 ln 2 1) = 1 2 ln 2
42
Substituting in (1) we obtain
U 2 ln 2 T 1 + 2 ln 2
= , = .
T (1,1,2,4) 1 + 2 ln 2 U (1,1,2,4) 2 ln 2
30. The pressure P, volume V , and temperature T of a van der Waals gas with n molecules (n constant) are related by
the equation
an 2
P + 2 (V nb) = n RT
V
P V
where a, b, and R are constant. Calculate and .
T P
SOLUTION Let F be the following function:
an 2
F(P, V , T ) = P+ (V nb) n RT
V2
By Eq. (6),
F F
P V
= TF , = FP (1)
T P
P V
f = F
(r )er 8
r
where er = and r = x, y, z.
r
718 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
Hence,
x y z F
(r ) r
f = F
(r ) , F
(r ) , F
(r ) = x, y, z = F
(r ) = F
(r )er
r r r r r
32. Let f (x, y, z) = ex y z = er , with r as in Exercise 31. Compute f directly and using Eq. (8).
2 2 2 2
f = F (r )er
r
f = 2r er = 2er r
2 2
r
Both answers agree, as expected.
1
33. Use Eq. (8) to compute and (ln r ).
r
1 1
SOLUTION To compute r using Eq. (8), we let F(r ) = r :
1
F
(r ) = 2
r
We obtain
1 1 r 1
= F
(r )er = 2 = 3r
r r r r
1 r r
(ln r ) = F
(r )er = = 2
r r r
34. Show that if f (x) is differentiable and c = 0 is a constant, then u(x, t) = f (x ct) satisfies the so-called advection
equation
u u
+c =0
t x
SOLUTION For s = x ct, we have u(x, t) = f (s). We use the Chain Rule to compute u u
t and x :
u s
= f
(s) = f
(s) (c) = c f
(s) (1)
t t
u s
= f
(s) = f
(s) 1 = f
(s) (2)
x x
S E C T I O N 15.6 The Chain Rule (ET Section 14.6) 719
35. Jessica and Matthew are running toward the point P along the straight paths that make a fixed angle of (Figure
3). Suppose that Matthew runs with velocity va m/s and Jessica with velocity vb m/s. Let f (x, y) be the distance from
Matthew to Jessica when Matthew
is x meters from P and Jessica is y meters from P.
(a) Show that f (x, y) = x + y 2 2x y cos .
2
(b) Assume that = /3. Use the Chain Rule to determine the rate at which the distance between Matthew and Jessica
is changing when x = 30, y = 20, va = 4 m/s, and vb = 3 m/s.
va
x vb
y
B
FIGURE 3
SOLUTION
(a) This is a simple application of the Law of Cosines. Connect points A and B in the diagram to form a line segment
will call f . Then, the Law of Cosines says that f = x + y 2x y cos . By taking square roots, we find that
that we 2 2 2
f = x 2 + y 2 2x y cos .
(b) Using the chain rule,
df f dx f dy
= +
dt x dt y dt
so we get
df (x y cos )d x/dt (y x cos )d y/dt
= +
dt x + y 2x y cos
2 2 x 2 + y 2 2x y cos
and using x = 30, y = 20, and d x/dt = 4, d y/dt = 3, we get
df 180 170 cos
=
dt 1300 1200 cos
(a) Let F(a, b, c, ) = a 2 + b2 2ab cos c2 . We use the formulas obtained by implicit differentiation (Eq. (6)) to
write
F F F
= aF , = bF , = cF (1)
a b c
2c c
= =
c 2ab sin ab sin
(b) The linear approximation for is
a + b + c = 1+ 1+ 2 (2)
a b c a b c
We find the partial derivatives for a = 10, b = 16, c = 22. We first find using the relation c2 = a 2 + b2 2ab cos .
This gives
In Exercises 3740, a function f (x, y, z) is called homogeneous of degree n if f ( x, y, z) = n f (x, y, z) for all
R.
37. Show that the following functions are homogeneous and determine their degree.
(a) f (x, y, z) = x 2 y + x yz (b) f (x, y, z) = 3x + 2y 8z
xy
(c) f (x, y, z) = ln (d) f (x, y, z) = z 4
z2
SOLUTION
f ( z) = ( z)4 = 4 z 4 = 4 f (z)
38. Prove that if f (x, y, z) is homogeneous of degree n, then f x (x, y, z) is homogeneous of degree n 1. Hint: Either
use the limit definition or apply the Chain Rule to f ( x, y, z).
S E C T I O N 15.6 The Chain Rule (ET Section 14.6) 721
SOLUTION We are given that f ( x, y, z) = n f (x, y, z) for all , and we must show that f x ( x, y, z) =
n1 f x (x, y, z). We use the limit definition of f x . Since for all
= 0, h 0 if and only if h 0, we get
f ( x + h, y, z) f ( x, y, z) f ( (x + h), y, z) f ( x, y, z)
f x ( x, y, z) = lim = lim
h0 h h0 h
n f (x + h, y, z) n f (x, y, z) n1 f (x + h, y, z) n1 f (x, y, z)
= lim = lim
h0 h h0 h
f (x + h, y, z) f (x, y, z)
= n1 lim = n1 f x (x, y, z)
h0 h
Alternatively, we prove this property using the Chain Rule. We use the Chain Rule to differentiate the following equality
with respect to x:
f ( x, y, z) = n f (x, y, z)
We get
( x) ( y) ( z)
f x ( x, y, z) + f y ( x, y, z) + f z ( x, y, z) = n f x (x, y, z)
x x x
f x ( x, y, z) = n f x (x, y, z) or f x ( x, y, z) = n1 f x (x, y, z)
Hint: Let F(t) = f (t x, t y, t z) and calculate F
(1) using the Chain Rule.
SOLUTION We use the Chain Rule to differentiate the function F(t) = f (t x, t y, t z) with respect to t. This gives
f (t x) f (t y) f (t z) f f f
F
(t) = + + =x +y +z (1)
x t y t z t x y z
On the other hand, since f is homogeneous of degree n, we have
F(t) = f (t x, t y, t z) = t n f (x, y, z)
f f f
x +y +z = nf
x y z
Hence,
f f f 1 1 2
x +y +z =x +y +z = 0 = 0 f (x, y, z)
x y z x y z
41. Suppose that x = g(t, s), y = h(t, s). Show that f tt is equal to
2
x 2 x y y 2x 2 y
fx x + 2 fx y + f yy + fx 2 + f y 2 10
t t t t t t
SOLUTION We are given that x = g(t, s), y = h(t, s). We must compute f tt for a function f (x, y). We first compute
f t using the Chain Rule:
x y
ft = f x + fy
t t
To find f tt we differentiate the two sides with respect to t using the Product Rule. This gives
x 2x y 2 y
f tt = ( fx ) + fx 2 + ( f y ) + fy 2 (1)
t t t t t t
By the Chain Rule,
x y
( fx ) = fx x + fx y
t t t
x y
( f y ) = f yx + f yy
t t t
Substituting in (1) we obtain
x y x 2x x y y 2 y
f tt = + fx y
fx x + f x 2 + f yx + f yy + fy 2
t t t t t t t t
2
2
2
x y x x x y y 2 y
= fx x + fx y + f x 2 + f yx + f yy + fy 2
t t t t t t t t
If f x y and f yx are continuous, Clairauts Theorem implies that f x y = f yx . Hence,
2
x 2 x y y 2x 2 y
f tt = f x x + 2 fx y + f yy + fx 2 + f y 2
t t t t t t
42. Let r = x 12 + + x n2 and let g(r ) be a function of r . Prove the formulas
g xi 2g xi2 r 2 xi2
= gr , = grr + gr
xi r xi2 r2 r3
g r 2xi x
= g
(r ) = gr = gr i
xi xi 2 x 12 + + x n2 r
S E C T I O N 15.6 The Chain Rule (ET Section 14.6) 723
We differentiate g
x with respect to xi . Using the Product Rule we get
i
2g xi xi
= (gr ) + gr (1)
xi 2 xi r xi r
d r 2xi xi
(gr ) = (gr ) = grr = grr (2)
xi dr xi 2 2
2 x1 + + xn r
We compute x xri using the Quotient Rule and the Chain Rule:
i
r
x i 1 r x i xi r xi xri r 2 xi2
= = = (3)
xi r r2 r2 r3
Substituting (2) and (3) in (1), we obtain
2g xi xi r 2 xi2 x2 r 2 xi2
= grr + gr = i2 grr + gr
xi 2 r r r 3 r r3
2g 2g n1
+ + = grr + gr
x 12 x n2 r
2g xi2 r 2 xi2
= grr + gr
xi 2 r2 r3
Hence,
2g 2g x 12 r 2 x 12 x n2 r 2 x n2
+ + = grr + gr + + grr + gr
xi 2 xn 2 r2 r3 r 2 r3
x 2 + + x n2 1 2
= 1 grr + gr (r x 2 ) + + (r 2 x 2 )
1 n
r2 r3
r2 1
= 2 grr + 3 gr nr 2 (x 12 + + x n2 )
r r
1 r2 n1
= grr + 3 gr (nr 2 r 2 ) = grr + 3 gr (n 1) = grr + gr
r r r
In Exercises 4448, the Laplace operator is defined by f = f x x + f yy . A function f (x, y) satisfying the Laplace
equation f = 0 is called harmonic. A function f (x, y) is called radial if f (x, y) = g(r ), where r = x 2 + y 2 .
x y x y
= r sin , = r cos , = cos , = sin ,
r r
2x 2 y 2x 2 y
= r cos , = r sin , = 2 =0
2 2 r 2 r
By Eq. (10) we have
2
x 2 y x y 2x 2 y
f = f x x + f yy + 2 fx y + fx 2 + f y 2
= f x x r 2 sin2 + f yy r 2 cos2 2r 2 sin cos f x y (r cos ) f x (r sin ) f y (1)
724 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
and
2
x 2 y x y 2x 2 y
frr = f x x + f yy + 2 fx y + fx 2 + f y 2
r r r r r r
= f x x cos2 + f yy sin2 + (2 cos sin ) f x y (2)
x y
fr = f x + fy = f x (cos ) + f y (sin ) (3)
r r
We now compute the right-hand side of the equality we need to prove. Using (1), (2), and (3), we obtain
1 1
frr + 2 f + fr = f x x cos2 + f yy sin2 + (2 cos sin ) f x y + f x x sin2
r r
cos sin cos sin
+ f yy cos2 (2 sin cos ) f x y fx f y + fx + fy
r r r r
= f x x cos2 + sin2 + f yy sin2 + cos2
= f x x + f yy = f
f = f x x + f yy
y
47. Verify that f (x, y) = tan1 is harmonic using both the rectangular and polar expressions for f .
x
SOLUTION
(a) Using the rectangular expression for f :
f = f x x + f yy
We compute the partial derivatives of f (x, y) = tan1 xy . Using the Chain Rule we get
1 y y
fx =
y 2 = 2
1+ x x 2 x + y2
1 1 x
fy =
y 2 x = 2
1+ x + y2
x
y 2x y
fx x =
2 2x =
2
x 2 + y2 x 2 + y2
x 2x y
f yy =
2 2y =
2
x 2 + y2 x 2 + y2
Hence,
2x y 2x y
f x x + f yy = 2
2
=0
(x 2 + y 2 ) (x 2 + y 2 )
(b) Using the polar expression for f ,
1 1
f = frr + 2 f + fr (1)
r r
sin = tan , we have f (x, y) = tan1
y = tan1 ( ) = . We compute the partial derivatives:
Since xy = rr cos x
fr = 0, f = 1, frr = 0, f = 0.
f f f f 2 f 1
r =1 +r = + r 2 = fr + r frr = r frr + fr
r r r r r r r r
Hence,
1 f
frr + fr = r 1 r (1)
r r r
Now, suppose that f (x, y) is a radial harmonic function. Since f is radial, f (x, y) = g(r ), therefore f = 0. Substi-
tuting in the polar expressions for f gives
1 1 1
f = frr + 2 f + fr = frr + fr = 0
r r r
Combining with (1), we get
f f
r 1 r = 0 or r =0
r r r r
726 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
yielding
f C
r =C fr =
r r
We now integrate the two sides to obtain
C
fr dr = dr or f (r ) = C ln r + b.
r
F(x, y, z) = x 2 + y 2 z 2 12x 8z 4 = 0
(a) Use the quadratic formula to solve for z as a function of x and y. This gives two formulas, depending on the choice
of a sign.
(b) Which formula defines the portion of the surface satisfying z 4? Which formula defines the portion satisfying
z 4?
z
(c) Calculate using the formula z = f (x, y) (for both choices of sign) and again via implicit differentiation. Verify
x
that the two answers agree.
z = 4
x
SOLUTION
(a) We rewrite F(x, y, z) = 0 as a quadratic equation in the variable z:
z 2 + 8z + 4 + 12x x 2 y 2 = 0
Hence,
8 4 x 2 + y 2 12x + 12
z 1,2 = = 4 x 2 + y 2 12x + 12
2
We obtain two functions:
z = 4 + x 2 + y 2 12x + 12, z = 4 x 2 + y 2 12x + 12
(b) The formula with the positive root defines the portion of the surface satisfying z 4, and the formula with the
negative root defines the portion satisfying z 4.
(c) Differentiating z = 4 + x 2 + y 2 12x + 12 with respect to x, using the Chain Rule, gives
z 2x 12 x 6
= = (1)
x 2 x 2 + y 2 12x + 12 x 2 + y 2 12x + 12
z Fx
= (2)
x Fz
S E C T I O N 15.6 The Chain Rule (ET Section 14.6) 727
Fx = 2x 12, Fz = 2z 8
Hence,
z x y Fx Fy Fz
= = 1
x y z Fz Fx Fy
x = y z, y = x z, z = x y
Hence,
z x y
= 1, = 1, = 1
x y z
Eq. (12) holds since
z x y
= (1) (1) (1) = 1
x y z
(c) If P V T = 0, then
T T
T = PV, P= , V =
V P
Hence,
T V T P 1
= P, = 2, =
V P P T V
We have
T V P T 1 T 1
=P 2 =
V P T P V P V
728 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
f x = 2x, f x x = 2, f y = 2y, f yy = 2, fx y = 0
gx = 2x, gx x = 2, g y = 2y, g yy = 2, gx y = 0
Therefore, f x = f y = 0 at (0, 0) and gx = g y = 0 at (0, 0). That is, the two functions have one critical point, which
is the origin. Since the discriminant of f is D = 4 > 0, f x x > 0, and the discriminant of g is D = 4 < 0, f has
a local minimum (which is also a global minimum) at the origin, whereas g has a saddle point there. Moreover, since
lim g(0, y) = and lim g(x, 0) = , g does not have global extrema on the plane. Similarly, f does not have a
y x
global maximum but does have a global minimum, which is f (0, 0) = 0.
2. Identify the points indicated in the contour maps as local minima, maxima, saddle points, or neither (Figure 14).
3
1
3
1 1 3
0 1 3 1 0
3 10 10
1 6 6
3
2 2
1 0
FIGURE 14
SOLUTION If f (P) is a local minimum or maximum, then the nearby level curves are closed curves encircling P. In
Figure (C), f increases in all directions emanating from P and decreases in all directions emanating from Q. Hence, f
has a local minimum at P and local maximum at Q.
P Q
10 10
6 6
2 2
In Figure (A), the level curves through the point R consist of two intersecting lines that divide the neighborhood near R
into four regions. f is decreasing in some directions and increasing in other directions. Therefore, R is a saddle point.
3
1
3
1
R 0 1
3
1
3
Figure (A)
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 729
S 1
3 1 0 3
Figure (B)
3. Let f (x, y) be a continuous function on a domain D in R2 . Determine which of the following statements are true:
(a) If D is closed and bounded, then f takes on a maximum value on D.
(b) If D is neither closed nor bounded, then f does not take on a maximum value of D.
(c) f (x, y) need not have a maximum value on D = {(x, y) : 0 x, y 1}.
(d) A continuous function takes on neither a minimum nor a maximum value on the open quadrant {(x, y) : x > 0,
y > 0}.
SOLUTION
(a) This statement is true. It follows by the Theorem on Existence of Global Extrema.
(b) The statement is false. Consider the constant function f (x, y) = 2 in the following domain:
y
x
1
Obviously f is continuous and D is neither closed nor bounded. However, f takes on a maximum value (which is 2) on
D.
(c) The domain D = {(x, y) : 0 x, y 1} is the following rectangle:
y
x
1
D = {(x, y) : 0 x, y 1}
D is closed and bounded, hence f takes on a maximum value on D. Thus the statement is false.
(d) The statement is false. The constant function f (x, y) = c takes on minimum and maximum values on the open
quadrant.
Exercises
1. Let P = (a, b) be a critical point of f (x, y) = x 2 + y 4 4x y.
(a) First use f x (, x, y) = 0 to show that a = 2b. Then use f y (x, y) = 0 to show that P = (0, 0), (2 2, 2), or
(2 2, 2).
(b) Referring to Figure 15, determine the local minima and saddle points of f (x, y) and find the absolute minimum
value of f (x, y).
730 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
FIGURE 15
SOLUTION
(a) We find the partial derivatives:
f x (x, y) = x 2 + y 4 4x y = 2x 4y
x
f y (x, y) = x 2 + y 4 4x y = 4y 3 4x
y
Since P = (a, b) is a critical point, f x (a, b) = 0. That is,
2a 4b = 0 a = 2b
4b3 4a = 0 a = b3
We obtain the following equations for the critical points (a, b):
a = 2b
a = b3
2b = b3
b1 = 0
b3 2b = b(b2 2) = 0 b2 = 2
b3 = 2
Since a = 2b, we have a1 = 0, a2 = 2 2, a3 = 2 2. The critical points are thus
P1 = (0, 0), P2 = 2 2, 2 , P3 = 2 2, 2
(b) Referring to Figure 14, we see that P1 = (0, 0) is a saddle point and P2 = 2 2, 2 , P3 = 2 2, 2 are
local minima. The absolute minimum value of f is 4.
Use the Second Derivative Test to determine the local minimum, maximum, and saddle points. Match f (x, y) and g(x, y)
with their graphs in Figure 16.
z z
y
x
y
x
(A) (B)
FIGURE 16
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 731
SOLUTION
Step 1. Find the critical points. We set the first partial derivatives equal to zero and solve:
f x = 2x + 6 = 0 x = 3
f y = 4y 4 y=1
fx x = 2
f yy = 4
fx y = 0
gx x = 2
g yy = 0
gx y = 12
The graph in Figure 16(A) has a saddle point, therefore it is the graph of g(x, y). The graph in Figure 16(B) corresponds
to f (x, y), since it has a local minimum.
3. Find the critical points of
f (x, y) = 8y 4 + x 2 + x y 3y 2 y 3
Use the contour map in Figure 17 to determine their nature (minimum, maximum, saddle point).
1
0.1
0.2
0.5 0.3
0 0.1 0
0.2
0.3
0.5
1
1 0.5 0 0.5 1
FIGURE 17 Contour map of f (x, y) = 8y 4 + x 2 + x y 3y 2 y 3 .
SOLUTION The critical points are the solutions of f x = 0 and f y = 0. That is,
f x (x, y) = 2x + y = 0
f y (x, y) = 32y 3 + x 6y 3y 2 = 0
The first equation gives y = 2x. We substitute in the second equation and solve for x. This gives
4. Let f (x, y) = y 2 x yx 2 + x y.
(a) Show that the critical points (x, y) satisfy the equations
y(y 2x + 1) = 0, x(2y x + 1) = 0
(b) Show that there are three critical points with either x = 0 or y = 0 (or both) and one critical point with x, y = 0.
(c) Use the second derivative to determine the nature of the critical points.
SOLUTION
(a) The critical points are the solutions of the two equations f x (x, y) = 0 and f y (x, y) = 0. That is,
(b) We find the critical points by solving the equations obtained in part (a):
y(y 2x + 1) = 0 (1)
x(2y x + 1) = 0 (2)
Equation (1) implies that y = 0 or y = 2x 1. Substituting y = 0 in (2) and solving for x gives
x(x + 1) = 0 x = 0 or x =1
We obtain the solutions (0, 0) and (1, 0). We now substitute y = 2x 1 in (2) and solve for x. We get
x(4x 2 x + 1) = 0
1
x(3x 1) = 0 x = 0 or x=
3
We compute the y-coordinate, using y = 2x 1:
y = 2 0 1 = 1
1 1
y =2 1=
3 3
We obtain the solutions (0, 1) and 13 , 13 . To summarize, the critical points are (0, 0), (1, 0), (0, 1), and 13 , 13 .
Three of the critical points have at least one zero coordinate, and one has two nonzero coordinates.
(c) We compute the second-order partial derivatives:
2
f x x (x, y) = (y 2yx + y) = 2y
x
f yy (x, y) = (2yx x 2 + x) = 2x
y
2
f x y (x, y) = (y 2yx + y) = 2y 2x + 1
y
The discriminant is
We now apply the Second Derivative Test. We first compute the discriminants at the critical points:
D(0, 0) = 1 < 0
D(1, 0) = 1 < 0
D(0, 1) = 1 < 0
2
1 1 1 1 2 2 1
D , = 4 + 1 = > 0,
3 3 3 3 3 3 3
1 1 1 2
fx x , = 2 = >0
3 3 3 3
The Second Derivative Test implies that the points (0, 0), (1, 0), and (0, 1) are saddle points, and f 1, 1 is a local
3 3
minimum.
In Exercises 520, find the critical points of the function. Then use the Second Derivative Test to determine whether they
are local minima or maxima (or state that the test fails).
5. f (x, y) = x 2 + y 2 x y + x
SOLUTION
Step 1. Find the critical points. We set the first-order partial derivatives of f (x, y) = x 2 + y 2 x y + x equal to zero
and solve:
f x (x, y) = 2x y + 1 = 0 (1)
f y (x, y) = 2y x = 0 (2)
Equation (2) implies that x = 2y. Substituting in (1) and solving for y gives
1
2 2y y + 1 = 0 3y = 1 y=
3
The corresponding value of x is x = 2 13 = 23 . The critical point is 23 , 13 .
Step 2. Compute the Discriminant. We find the second-order partials:
The discriminant is
6. f (x, y) = x 3 x y + y 3
SOLUTION
Step 1. Find the critical points. We set the first-order partial derivatives of f (x, y) = x 3 x y + y 3 equal to zero and
solve:
f x (x, y) = 3x 2 y = 0 (1)
f y (x, y) = x + 3y 2 = 0 (2)
Equation (1) implies that y = 3x 2 . Substituting in equation (2) and solving for x gives
2
x + 3(3x 2 ) = 0
1
x + 27x 4 = x(1 + 27x 3 ) = 0 x = 0, x=
3
2
The y-coordinates are y = 3 02 = 0 and y = 3 13 = 13 . The critical points are thus (0, 0) and 13 , 13 .
734 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
The discriminant is
D(0, 0) = 1 < 0
1 1 1 1 1 1 1
D , = 36 1 = 3 > 0, f x x , =6 =2>0
3 3 3 3 3 3 3
Thus, (0, 0) is a saddle point, whereas f 13 , 13 is a local minimum.
7. f (x, y) = x 3 y + 12x 2 8y
SOLUTION
Step 1. Find the critical points. We set the first-order partial derivatives of f (x, y) = x 3 y + 12x 2 8y equal to zero
and solve:
Equation (2) implies that x = 2. We substitute in equation (1) and solve for y to obtain
6(2y + 8) = 0 or y = 4
f x x (x, y) = 6x y + 24, f yy = 0, f x y = 3x 2
D(x, y) = f x x f yy f x2y = 9x 4
D(2, 4) = 9 24 < 0
Step 1. Find the critical points. We set the first-order partial derivatives of f (x, y) = x 3 + 2x y 2y 2 10x equal to
zero and solve:
f x (x, y) = 3x 2 + 2y 10 = 0 (1)
f y (x, y) = 2x 4y = 0 (2)
Equation (2) implies that x = 2y. We substitute in (1) and solve for y. This gives
3 (2y)2 + 2y 10 = 0
12y 2 + 2y 10 = 0
6y 2 + y 5 = 0
1 1 4 6 (5) 1 11 5
y1,2 = = y1 = 1 and y2 =
12 12 6
We find the x-coordinates using x = 2y:
5 5
x 1 = 2 (1) = 2, x2 = 2 =
6 3
The critical points are thus (2, 1) and 53 , 56 .
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 735
The discriminant is
9. f (x, y) = 4x 3x 3 2x y 2
SOLUTION
Step 1. Find the critical points. We set the first-order derivatives of f (x, y) = 4x 3x 3 2x y 2 equal to zero and solve:
f x (x, y) = 4 9x 2 2y 2 = 0 (1)
f y (x, y) = 4x y = 0 (2)
10. f (x, y) = x 3 + y 4 6x 2y 2
SOLUTION
Step 1. Find the critical points. We set the first-order derivatives of f (x, y) = x 3 + y 4 6x 2y 2 equal to zero and
solve:
The discriminant is
11. f (x, y) = x 4 + y 4 4x y
SOLUTION
Step 1. Find the critical points. We set the first-order derivatives of f (x, y) = x 4 + y 4 4x y equal to zero and solve:
Equation (1) implies that y = x 3 . Substituting in (2) and solving for x, we obtain
3
(x 3 ) x = x 9 x = x(x 8 1) = 0 x = 0, x = 1, x = 1
y = 03 = 0, y = 13 = 1, y = (1)3 = 1
D(0, 0) = 16 < 0
D(1, 1) = 144 16 = 128 > 0, f x x (1, 1) = 12 > 0
D(1, 1) = 144 16 = 128 > 0, f x x (1, 1) = 12 > 0
We conclude that (0, 0) is a saddle point, whereas f (1, 1) and f (1, 1) are local minima.
SOLUTION
Step 1. Find the critical points. We set the first partials of f (x, y) = e x y +4y equal to zero and solve:
2 2
Since e x y +4y = 0, the first equation gives x = 0 and the second equation gives 2y + 4 = 0 or y = 2. We obtain
2 2
2xe x y +4y = 2xe x y +4y (2y + 4) = 4x(2 y)e x y +4y
2 2 2 2 2 2
f x y (x, y) =
y
The discriminant is
2 2
D(x, y) = f x x f yy f x2y = 4e2 x y +4y 2y 2 8y + 7 16x 2 (2 y)2 e2 x y +4y
2 2
1 + 2x 2
SOLUTION
Step 1. Find the critical points. We compute the partial derivatives of f (x, y) = x yex y , using the Product Rule
2 2
f y (x, y, z) = x 1 ex y + yex y (2y) = xex y 1 2y 2
2 2 2 2 2 2
We set the partial derivatives equal to zero and solve to find the critical points. This gives
yex y 1 2x 2 = 0
2 2
xex y 1 2y 2 = 0
2 2
xex = 0
2
y = 0: x =0
1 1 1
1
e 2 y 1 2y 2 = 0 1 2y 2 = 0 y =
2
x= :
2 2 2
1 1 1 y 2 1
x = : e 2 1 2y 2 = 0 1 2y 2 = 0 y =
2 2 2
738 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
x 2 y 2
1 2y 2 = x ex y (2y) 1 2y 2 + ex y (4y)
2 2 2 2
f yy (x, y) = x e
y
= 2yxex y 3 2y 2
2 2
yex y = 1 2x 2 1 ex y + yex y (2y)
2 2 2 2 2 2
f x y (x, y) = f x = 1 2x 2
y y
= ex y 1 2x 2 1 2y 2
2 2
The discriminant is
D(x, y) = f x x f yy f x2y
Step 3. Apply the Second Derivative Test. We construct the following table:
ln y + 0 = 0 ln y = 0 y = 1.
We obtain the critical point (0, 1). f x and f y do not exist at the points where x + y = 0, but these points are not in the
domain of f , hence they are not critical points. The critical point is thus (0, 1).
Step 2. Compute the discriminant. We find the second-order derivatives:
x 1 1 (x + y) x 1 1 y x + 2y
fx x = ln(x + y) + = + = + =
x x+y x+y (x + y)2 x+y (x + y)2 (x + y)2
x x
f yy = =
y x + y (x + y)2
x 1 (x + y) x 1 y
f x y = f yx = = =
x x + y (x + y)2 (x + y)2
The discriminant is
x(x + 2y) y2
D(x, y) = f x x f yy f x2y =
(x + y)4 (x + y)4
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 739
12
D(0, 1) = 0 = 1 < 0
(0 + 1)4
Therefore, (0, 1) is a saddle point.
15. f (x, y) = e x xe y
SOLUTION
Step 1. Find the critical points. We set the first-order derivatives of f (x, y) = e x xe y equal to zero and solve:
f x (x, y) = e x e y = 0
f y (x, y) = xe y = 0
Since e y = 0, the second equation gives x = 0. Substituting in the first equation, we get
e0 e y = 1 e y = 0 ey = 1 y=0
D(0, 0) = 0 e0 = 1 < 0
We set the partial derivatives equal to zero and solve to find the critical points:
2
e yx 1 2x 2 6x y = 0
2
e yx (3 + x + 3y) = 0
2
Since e yx = 0, we obtain the following equations:
1 2x 2 6x y = 0
3 + x + 3y = 0
The second equation gives x = 3(1 + y). We substitute for x in the first equation and solve for y:
D(x, y) = f x x f yy f x2y
Step 3. Apply the Second Derivative Test. We obtain the following table:
Critical
Point
fx x f yy fx y D Type
16 , 17
18 2.4 1.13 0.38 2.57 D > 0, f x x > 0, local minimum
17. f (x, y) = ln x + 2 ln y x 4y
SOLUTION
Step 1. Find the critical points. We set the first-order partials of f (x, y) = ln x + 2 ln y x 4y equal to zero and
solve:
1 2
f x (x, y) = 1 = 0, f y (x, y) = 4=0
x y
The first equation gives x = 1, and the second equation gives y = 12 . We obtain the critical point 1, 12 . Notice that f x
and f y do not exist if x = 0 ory = 0, respectively, but these are not critical points since they are not in the domain of f .
The critical point is thus 1, 12 .
Step 2. Compute the discriminant. We find the second-order partials:
1 2
f x x (x, y) = 2 , f yy (x, y) = 2 , f x y (x, y) = 0
x y
The discriminant is
2
D(x, y) = f x x f yy f x2y = 2 2
x y
Step 3. Apply the Second Derivative Test. We have
1 2 1 1
D 1, = 2 = 8 > 0, f x x 1, = 2 = 1 < 0
2 2 1
12 12
We conclude that f 1, 12 is a local maximum.
SOLUTION
Step 1. Find the critical points. We set the first-order derivatives equal to zero and solve:
Since ex y = 0, the first equation gives x = 0 or x 2 + y 2 = 1. We substitute x = 0 in the second equation and solve
2 2
for y:
2yey (1 y 2 ) = 0
2
Since ey = 0, the solutions are y = 0 or y = 1. The corresponding points are (0, 0), (0, 1), (0, 1). The solution
2
x 2 + y 2 = 1 also satisfies the second equation. We conclude that there are infinitely many critical points, namely, the
points on the unit circle x 2 + y 2 = 1 and its center (0, 0).
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 741
= 2ex y 2x 2 + 2x 4 + 2x 2 y 2 + 1 3x 2 y 2
2 2
= 2ex y 2x 4 5x 2 + 2x 2 y 2 y 2 + 1
2 2
Also,
x 2 y 2
f x y (x, y) = e 2x 2x 3 2x y 2
y
= 2yex y 2x 2x 3 2x y 2 + ex y (4x y)
2 2 2 2
= 4x yex y 1 + x 2 + y 2 1
2 2
= 4x yex y 2 + x 2 + y 2
2 2
The discriminant is
D(x, y) = f x x f yy f x2y
= 4e2(x +y ) 2x 4 5x 2 + 2x 2 y 2 y 2 + 1 2y 4 5y 2 + 2x 2 y 2 x 2 + 1
2 2
2
16x 2 y 2 e2 x +y (2 + x 2 + y 2 )
2 2
Step 3. Apply the Second Derivative Test. We first check the critical point (0, 0):
Therefore, f (0, 0) is a local minimum. We check the points of the unit circle x 2 + y 2 = 1. We rewrite D(x, y) as
follows:
2 2 " #" #
D(x, y) = 4e2 x +y 2x 2 x 2 + y 2 x 2 + y 2 4x 2 + 1 2y 2 x 2 + y 2 x 2 + y 2 4y 2 + 1
2
16x 2 y 2 e2 x +y
2 2
x 2 + y2 2
= 16e2 x 2 y 2 16e2 x 2 y 2 = 0
Therefore, the Second Derivative Test is inconclusive. However, for the given function we can define t = x 2 + y 2 to
obtain the function g(t) = tet . The critical point of g(t) is
SOLUTION
Step 1. Find the critical points. We set the partial derivatives of f (x, y) = x y 2 ln(x + y) equal to zero and solve.
1 1
f x (x, y) = 1 = 0, f y (x, y) = 2y =0
x+y x+y
1 = 1. Substituting in the second equation gives
The first equation implies that x+y
1
2y 1 = 0 2y = 1 y=
2
1 1 3
1 =0 x =1 x=
x 12 2 2
We obtain the critical point 32 , 12 . Notice that although f x and f y do not exist where x + y = 0, these are not critical
points since f is not defined at these points.
Step 2. Compute the discriminant. We compute the second-order partial derivatives:
1 1
f x x (x, y) = 1 =
x x+y (x + y)2
1 1
f yy (x, y) = 2y = 2 +
y x+y (x + y)2
1 1
f x y (x, y) = 1 =
y x+y (x + y)2
The discriminant is
1 1 1 2
D(x, y) = f x x f yy f x2y = 2 + =
(x + y)2 (x + y)2 (x + y)4 (x + y)2
Step 3. Apply the Second Derivative Test. We have
3 1 2
D , = 2 = 2 < 0
2 2 3 1
2 2
We conclude that 32 , 12 is a saddle point.
Find the critical points. We set the partial derivatives of f (x, y) = (x y)e x y equal to zero and solve:
2 2
SOLUTION
f x (x, y) = e x y + (x y)e x y 2x = e x y 2x 2 2x y + 1 = 0
2 2 2 2 2 2
f y (x, y) = e x y + (x y)e x y (2y) = e x y 2y 2 2x y 1 = 0
2 2 2 2 2 2
2x 2 2x y + 1 = 0
2y 2 2x y 1 = 0
We add and subtract the two equations to obtain the following equations:
2 x 2 + y 2 4x y = 0
2 x 2 y2 + 2 = 0
The first equation can be rewritten as x 2 2x y + y 2 = 0 or (x y)2 = 0, yielding x = y. Substituting in the second
equation gives 2 = 0, we conclude that the two equations have no solutions, that is, there are no critical points (notice
that f x and f y exist everywhere). Since local minima and local maxima can occur only at critical points, it follows that
f (x, y) = (x y)e x y does not have local minima or local maxima.
2 2
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 743
21. Show that f (x, y) = x 2 + y 2 has one critical point P and that f is nondifferentiable at P. Show that f (P) is an
absolute minimum value.
SOLUTION Since f (x, y) = x 2 + y 2 0 and f (0, 0) = 0, f (0, 0) is an absolute minimum value. To find the
critical point of f we first find the first derivatives:
2x x
f x (x, y) = x 2 + y2 = =
x 2 x 2 + y2 x 2 + y2
2y y
f y (x, y) = x 2 + y2 = =
y 2 x +y
2 2 x + y2
2
Since f x and f y do not exist at (0, 0) and the equations f x (x, y) = 0 and f y (x, y) = 0 have no solutions, the only
critical point is P = (0, 0), a point where f is non-differentiable.
22. Let f (x, y) = (x 2 + y 2 1)2 .
(a) Find the critical points of f (x, y).
(b) Use a computer algebra system to graph f (x, y) and identify the critical points on the graph.
(c) Find all points where f takes on a minimum value. Does the Second Derivative Test apply?
SOLUTION
(a) We set the partial derivatives equal to zero and solve. This gives
f x (x, y) = 2 x 2 + y 2 1 2x = 0 x x 2 + y2 1 = 0
f y (x, y) = 2 x 2 + y 2 1 2y = 0 y x 2 + y2 1 = 0
The solutions of the first equation are x = 0 or x 2 + y 2 = 1. The second equation implies that y = 0 or x 2 + y 2 = 1.
We obtain the following critical points: (0, 0); and the points on the unit circle x 2 + y 2 = 1.
2
(b) The following figure shows the graph of f (x, y) = x 2 + y 2 1 and the critical points (namely, the origin and
the unit circle) on this graph.
z
x
2
(c) Since f (x, y) = x 2 + y 2 1 0, and f (x, y) = 0 for (x, y) on the unit circle x 2 + y 2 = 1, f takes on a
minimum value on the unit circle. We try to use the Second Derivative Test. The second-order partials are:
3
fx x = 4x + 4x y 2 4x = 12x 2 + 4y 2 4
x
2
f yy = 4x y + 4y 3 4y = 4x 2 + 12y 2 4
y
3
fx y = 4x + 4y 2 x 4x = 8yx
y
The discriminant is
D(x, y) = f x x f yy f x2y = 16 3x 2 + y 2 1 3y 2 + x 2 1 64y 2 x 2
Hence f (0, 0) is a local maximum. To compute D(x, y) at the point on the unit circle, we first rewrite D(x, y) as follows:
D(x, y) = 16 2x 2 + x 2 + y 2 1 2y 2 + x 2 + y 2 1 64y 2 x 2
Hence,
D(x, y) = 16 2x 2 + 1 1 2y 2 + 1 1 64y 2 x 2 = 64x 2 y 2 64y 2 x 2 = 0
x 2 +y 2 =1
We see that the Second Derivative Test is inconclusive for the critical points on the unit circle.
744 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
23. Use a computer algebra system to find numerical approximations to the critical points of
2 2
f (x, y) = (1 x + x 2 )e y + (1 y + y 2 )e x
x
y
2 2
FIGURE 18 Plot of the function f (x, y) = (1 x + x 2 )e y + (1 y + y 2 )e x .
SOLUTION The critical points are the solutions of f x (x, y) = 0 and f y (x, y) = 0. We compute the partial derivatives:
2
2
f x (x, y) = (1 + 2x)e y + 1 y + y 2 e x 2x
2 2
f y (x, y) = 1 x + x 2 e y 2y + (1 + 2y)e x
Hence, the critical points are the solutions of the following equations:
2
2
(2x 1)e y + 2x 1 y + y 2 e x = 0
2
2
(2y 1)e x + 2y 1 x + x 2 e y = 0
Using a CAS we obtain the following solution: x = y = 0.27788, which from the figure is a local minimum.
24. Use the contour map in Figure 19 to determine whether the critical points A, B, C, D are local minima, maxima, or
saddle points.
3
3
2
2 1
1 0
1 1
A B 2
3
0
0
1
1 D C
2
3
3 2 1 0 1 2 3
FIGURE 19
SOLUTION The nearby level curves at A and C are closed curves encircling A and C. As we move towards A the
function increases in all directions, while moving towards C the function decreases in all directions. We conclude that
the function has a local maximum at A and a local minimum at C. The level curves through B and D consist of two
curves intersecting at these points respectively. These curves divide the neighborhoods near B and D into four regions.
In some of the regions the function is increasing and in others it is decreasing as we move towards B or D. This implies
that B and D are saddle points.
25. Which of the following domains are closed and which are bounded?
(a) {(x, y) R2 : x 2 + y 2 1}
(b) {(x, y) R2 : x 2 + y 2 < 1}
(c) {(x, y) R2 : x 0}
(d) {(x, y) R2 : x > 0, y > 0}
(e) {(x, y) R2 : 1 x 4, 5 y 10}
(f) {(x, y) R2 : x > 0, x 2 + y 2 10}
SOLUTION
(a) {(x, y) R2 : x 2 + y 2 1}: This domain is bounded since it is contained, for instance, in the disk x 2 + y 2 < 2.
The domain is also closed since it contains all of its boundary points, which are the points on the unit circle x 2 + y 2 = 1.
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 745
(b) {(x, y) R2 : x 2 + y 2 < 1}: The domain is contained in the disk x 2 + y 2 < 1, hence it is bounded. It is not closed
since its boundary x 2 + y 2 = 1 is not contained in the domain.
(c) {(x, y) R2 : x 0}:
y
This domain is not contained in any disk, hence it is not bounded. However, the domain contains its boundary x = 0 (the
y-axis), hence it is closed.
(d) {(x, y) R2 : x > 0, y > 0}:
y
The domain is not contained in any disk, hence it is not bounded. The boundary is the positive x and y axes, and it is not
contained in the domain, therefore the domain is not closed.
(e) {(x, y) R2 : 1 x 4, 5 y 10}:
y
D (1, 10) C (4, 10)
10
7.5
A (1, 5) B (4, 5)
x
1 4
This domain is contained in the disk x 2 + y 2 112 , hence it is bounded. Moreover, the domain contains its boundary,
which consists of the segments AB, BC, C D, AD shown in the figure, therefore the domain is closed.
(f) {(x, y) R2 : x > 0, x 2 + y 2 10}:
y
This domain is bounded since it is contained in the disk x 2 + y 2 10. It is not closed since the part {(0, y) R2 : |y|
In Exercises 2629, determine the global extreme values of the function on the given set without using calculus.
26. f (x, y) = x + y, 0 x, y 1
SOLUTION The sum x + y is maximum when x = 1 and y = 1, and it is minimum when x = 0 and y = 0. Therefore,
the global maximum of f on the given set is f (1, 1) = 1 + 1 = 2 and the global minimum is f (0, 0) = 0 + 0 = 0.
746 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
27. f (x, y) = 2x y, 0 x 1, 0 y 3
SOLUTION f is maximum when x is maximum and y is minimum, that is x = 1 and y = 0. f is minimum when
x is minimum and y is maximum, that is, x = 0, y = 3. Therefore, the global maximum of f in the set is f (1, 0) =
2 1 0 = 2 and the global minimum is f (0, 3) = 2 0 3 = 3.
28. f (x, y) = (x 2 + y 2 + 1)1 , 0 x 3, 0 y 5
SOLUTION f (x, y) = 2 1 2 is maximum when x 2 and y 2 are minimum, that is, when x = y = 0. f is minimum
x +y +1
when x 2 and y 2 are maximum, that is, when x = 3 and y = 5. Therefore, the global maximum of f on the given set is
1 1 1 .
f (0, 0) = (02 + 02 + 1) = 1, and the global minimum is f (3, 5) = (32 + 52 + 1) = 35
29. f (x, y) = ex y ,
2 2
x 2 + y2 1
The function f (x, y) = e(x +y ) = is maximum when e x +y is minimum, that is, when x 2 + y 2
2 2 1 2 2
SOLUTION 2 2
e x +y
is minimum. The minimum value of x 2 + y 2 on the given set is zero, obtained at x = 0 and y = 0. We conclude that the
maximum value of f on the given set is
f (0, 0) = e0 0 = e0 = 1
2 2
f is minimum when x 2 + y 2 is maximum, that is, when x 2 + y 2 = 1. Thus, the minimum value of f on the given disk
is obtained on the boundary of the disk, and it is e1 = 1e .
30. Assumptions Matter Show that f (x, y) = x + y has no global minimum or maximum on the domain 0 < x, y <
1. Does this contradict Theorem 3?
SOLUTION The largest and smallest values of f on the closed square 0 x, y 1 are f (1, 1) = 2 and f (1, 1) =
2. However, on the open square 0 < x, y < 1, f can never attain these maximum and minimum values, since the
boundary (and in particular the points (1, 1) and (1, 1)) are not included in the domain. This does not contradict
Theorem 3 since the domain is open.
31. Let D = {(x, y) : x > 0, y > 0}. Show that D is not closed. Find a continuous function that does not have a global
minimum value on D.
SOLUTION The boundary of D consists of two rays: the positive x and y axes.
y
x
x > 0, y > 0
1 is continuous on D (since
These rays are not contained in D, therefore D is not closed. The function f (x, y) = x+y
x + y = 0 in D). However, f does not have a global minimum value on D, since lim x+x 1 = . Thus, when
x0+
x approaches zero along the ray {(x, x) : x > 0}, the values of f are decreasing without bound.
32. The goal is to find the extreme values of f (x, y) = x 2 2x y + 2y on the square 0 x, y 2 (Figure 20).
(a) Evaluate f at the critical points that lie in the square.
(b) On the bottom edge of the square, y = 0 and f (x, 0) = x 2 . Find the extreme values of f on the bottom edge.
(c) Find the extreme values of f on the remaining edges of the square.
(d) Find the largest and smallest values of f among the values computed in (a), (b), and (c).
y f(x, 2) = x 4x + 4
2
Edge y = 2
Edge x = 0 Edge x = 2
f(0, y) = 2y f(2, y) = 4 2y
x
2
Edge y = 0
f(x, 0) = x 2
FIGURE 20 The function f (x, y) = x 2 2x y + 2y on the four boundary segments of the square 0 x 2.
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 747
SOLUTION
(a) We find the critical points that lie in the interior of the square, by solving f x = 0, f y = 0 for 0 < x, y < 2. This
gives
f x = 2x 2y = 0, f y = 2x + 2 = 0
The second equation gives x = 1. Substituting in the first equation and solving for y gives
2 1 2y = 0 2y = 2 y=1
x
Edge y = 0 2
(c) On the right edge, x = 2 and 0 y 2, and f takes the values f (2, y) = 4 4y + 2y = 4 2y. The global
maximum is f (2, 0) = 4 2 0 = 4 and the global minimum is f (2, 2) = 4 2 2 = 0. On the top edge, y = 2
and 0 x 2, and f takes the values f (x, 2) = x 2 2x 2 + 2 2 = x 2 4x + 4 = (x 2)2 . The minimum
value is f (2, 2) = (2 2)2 = 0 and the maximum value is f (0, 2) = (0 2)2 = 4. Finally, on the left edge x = 0,
0 y 2, and f takes the values f (0, y) = 2y. The minimum value is f (0, 0) = 2 0 = 0 and the maximum value is
f (0, 2) = 2 2 = 4.
(d) The values computed in parts (a), (b), and (c) are
We examine the minimum and maximum values among these values, and determine the global minimum and the global
maximum of f on the square. We obtain the following conclusions:
The global minimum is f (2, 2) = f (0, 0) = 0.
The global maximum is f (2, 0) = f (0, 2) = 4.
(a) This is best done by converting to polar coordinates. We see that f (r, ) = (r cos + r sin ) ln r 2 , or in other
words, f (r, ) = (cos + sin )2r ln r . We take derivatives and find that
f f
= (cos + sin )(2 ln r + 2), = ( sin + cos )(2r ln r )
r
Since a critical point makes both equations equal to zero, and since 2 ln r + 2 and 2r ln r cant both be zero at the same
r , we get that either cos + sin = 0 or sin + cos = 0, which means either x = y orx = y. These
solutions
leadto r = 1 and r = 1/e, respectively,
giving us our four critical points of (1/e 2, 1/e 2), (1/e 2, 1/e 2),
(1/ 2, 1/ 2), and (1/ 2, 1/ 2).
(b) Looking at a graph, we see that (1/ 2, 1/ 2) and (1/ 2, 1/ 2) are saddle points, while (1/e 2, 1/e 2) is a
minimum and (1/e 2, 1/e 2) is a maximum.
z
1
0.5 1.5
1.5
0 y
1
x
748 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
2 1 y
1 2
x 2
In Exercises 3541, determine the global extreme values of the function on the given domain.
f x (x, y) = 3x 2 = 0, f y (x, y) = 2
The two equations have no solutions, hence there are no critical points.
Step 2. Check the boundary. The extreme values occur either at the critical points or at a point on the boundary of the
domain. Since there are no critical points, the extreme values occur at boundary points. We consider each edge of the
square 0 x, y 1 separately.
The segment O A: On this segment y = 0, 0 x 1, and f takes the values f (x, 0) = x 3 . The minimum value is
f (0, 0) = 0 and the maximum value is f (1, 0) = 1.
y
C (0, 1) B (1, 1)
x
D (0, 0) A (1, 0)
The segment AB: On this segment x = 1, 0 y 1, and f takes the values f (1, y) = 1 2y. The minimum value
is f (1, 1) = 1 2 1 = 1 and the maximum value is f (1, 0) = 1 2 0 = 1.
The segment BC: On this segment y = 1, 0 x 1, and f takes the values f (x, 1) = x 3 2. The minimum value
is f (0, 1) = 03 2 = 2 and the maximum value is f (1, 1) = 13 2 = 1.
The segment OC: On this segment x = 0, 0 y 1, and f takes the values f (0, y) = 2y. The minimum value is
f (0, 1) = 2 1 = 2 and the maximum value is f (0, 0) = 2 0 = 0.
Step 3. Conclusions. The values obtained in the previous steps are
The smallest value is f (0, 1) = 2 and it is the global minimum of f on the square. The global maximum is the largest
value f (1, 0) = 1.
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 749
f y (x, y) = 8yex y + 4y 2 x 2 ex y (2y) = 2yex y 4 + 4y 2 x 2 = 0
2 2 2 2 2 2
Since ex y = 0, the first equation gives x = 0 or x 2 = 1 + 4y 2 . Substituting x = 0 in the second equation gives
2 2
2yey 4 + 4y 2 = 0.
2
Since ey = 0, we get
2
y 1 + y 2 = y(y 1)(y + 1) = 0 y = 0, y = 1, y = 1
We obtain the three points (0, 0), (0, 1), (0, 1). We now substitute x 2 = 1 + 4y 2 in the second equation and solve for
y:
2ye15y 4 + 4y 2 1 4y 2 = 0
2
2ye15y (5) = 0
2
y=0
x 2 = 1 + 4 02 = 1 x = 1
We obtain the solutions (1, 0) and (1, 0). We conclude that the critical points are
(0, 0), (0, 1), (0, 1), (1, 0), and (1, 0).
All of these points are in the interior x 2 + y 2 < 2 of the given disk.
Step 2. Check the boundary. The boundary is the circle x 2 + y 2 = 2. On this set y 2 = 2 x 2 , hence the function
f (x, y) takes the values
f (x, y) = g(x) = 4 2 x 2 x 2 e2 = 5x 2 + 8 e2
x 2 +y 2 =2
2 2 2
is, g(x)= 5e x + 8e . We determine the interval of x. Since x + y = 2, we have 0 x 2 or
That 2 2 2
2 x 2.
y
8e2
2 2
x
1.265 1.265
750 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
We thus must find the extreme values of g(x) = 5e2 x 2 + 8e2 on the interval 2 x 2. With the aid of the
2
graph of g(x), we conclude that the maximum value is g(0) = 8e and the minimum value is
2
g 2 =g 2 = 5e2 2 + 8e2 = 10e2 + 8e2 = 2e2 0.271
We conclude that the points on the boundary with largest and smallest values of f are
f 0, 2 = 8e2 1.083, f 2, 0 = 2e2 0.271
Step 3. Conclusions. The extreme values either occur at the critical points or at the points on the boundary, found in step
2. We compare the values of f at these points:
f (0, 0) = 0
f (0, 1) = 4e1 1.472
f (0, 1) = 4e1 1.472
f (1, 0) = e1 0.368
f (1, 0) = e1 0.368
f 0, 2 1.083
f 2, 0 0.271
We conclude that the global minimum is f (1, 0) = f (1, 0) = 0.368 and the global maximum is f (0, 1) =
f (0, 1) = 1.472.
39. f (x, y) = x 3 + x 2 y + 2y 2 , x, y 0, x + y 1
SOLUTION We use the following steps.
Step 1. Examine the critical points. We find the critical points of f (x, y) = x 3 + x 2 y + 2y 2 in the interior of the domain
(the standard region in the figure).
y
B (0, 1)
x
0 A (1, 0)
The first equation gives x = 0 or y = 32 x. Substituting x = 0 in the second equation gives 4y = 0 or y = 0. We obtain
the critical point (0, 0). We now substitute y = 32 x in the second equation and solve for x:
3
x 2 + 4 x = x 2 6x = x(x 6) = 0 x = 0, x = 6
2
We get the critical points (0, 0) and (6, 9). None of the critical points (0, 0) and (6, 9) is in the interior of the domain.
Step 2. Check the boundary. The boundary consists of the three segments O A, O B, and AB shown in the figure. We
consider each part of the boundary separately.
The segment O A: On this segment y = 0, 0 x 1, and f (x, y) = f (x, 0) = x 3 . The minimum value is
f (0, 0) = 03 = 0 and the maximum value is f (1, 0) = 13 = 1.
The segment O B: On this segment x = 0, 0 y 1, and f (x, y) = f (0, y) = 2y 2 . The minimum value is
f (0, 0) = 2 02 = 0 and the maximum value is f (0, 1) = 2 12 = 2.
The segment AB: On this segment y = 1 x, 0 x 1, and
f (x, y) = x 3 + x 2 (1 x) + 2(1 x)2 = x 3 + x 2 x 3 + 2 1 2x + x 2 = 3x 2 4x + 2
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 751
We find the extreme values of g(x) = 3x 2 4x + 2 in the interval 0 x 1. With the aid of the graph of g(x), and
with setting the derivative g
equal to 0, we find that the minimum value is
2
2 2 1 2 2 2
g = f , =3 4 +2=
3 3 3 3 3 3
and the maximum value is
g(0) = f (0, 1) = 3 02 4 0 + 2 = 2
y
g(x) = 3x 2 4x + 2
x
0 2 1
3
Step 3. Conclusions. We compare the values of f (x, y) at the points obtained in step (2), and determine the global
extrema of f (x, y). This gives
2 1 2
f (0, 0) = 0, f (1, 0) = 1, f (0, 1) = 2, f , =
3 3 3
We conclude that the global minimum of f in the given domain is f (0, 0) = 0 and the global maximum is f (0, 1) = 2.
40. f (x, y) = x 3 + y 3 3x y, 0 x, y 1
SOLUTION We use the following steps.
Step 1. Examine the critical points in the interior of the domain. We set the partial derivatives equal to zero and solve:
f x (x, y) = 3x 2 3y = 0
f y (x, y) = 3y 2 3x = 0
The first equation gives y = x 2 . We substitute in the second equation and solve for x:
2
3 x 2 3x = 0
3x 4 3x = 3x x 3 1 = 0 x = 0, y = 02 = 0
or x = 1, y = 12 = 1
The critical points (0, 0) and (1, 1) are not in the interior of the domain.
Step 2. Find the extreme values on the boundary. We consider each part of the boundary separately.
y
D(0, 1) C(1, 1)
x
A(0, 0) B(1, 0)
The edge AB: On this edge, y = 0, 0 x 1, and f (x, 0) = x 3 . The maximum value is obtained at x = 1 and the
minimum value is obtained at x = 0. The corresponding extreme points are (1, 0) and (0, 0).
this edge x = 1, 0 y 1, and f (1, y) = y 3y + 1. The critical points are
The edge BC: On 3
d y 3 3y + 1 = 3y 2 3 = 0, that is, y = 1. The point in the given domain is y = 1. The candidates
dy
for extreme values are thus y = 1 and y = 0, giving the points (1, 1) and (1, 0).
The edge DC: On this edge y = 1, 0 x 1, and f (x, 1) = x 3 3x + 1. Replacing the values of x and y in the
previous solutions we get the points (1, 1) and (0, 1).
The edge AD: On this edge x = 0, 0 y 1, and f (0, y) = y 3 . Replacing the values of x and y obtained for the
edge AB, we get (0, 1) and (0, 0).
752 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
By Theorem 3, the extreme values occur either at a critical point in the interior of the square or at a point on the boundary
of the square. Since there are no critical points in the interior of the square, the candidates for extreme values are the
following points:
f (0, 0) = 03 + 03 3 0 = 0
f (1, 0) = 13 + 03 3 1 0 = 1
f (1, 1) = 13 + 13 3 1 1 = 1
f (0, 1) = 03 + 13 3 0 1 = 1
We conclude that in the given domain, the global maximum is f (1, 0) = f (0, 1) = 1 and the global minimum is
f (1, 1) = 1.
41. f (x, y) = x 3 y 5 , the set bounded by x = 0, y = 0, and y = 1 x. Hint: Use a computer algebra system
to find the minimum along the boundary curve y = 1 x, which is parametrized by (t, 1 t) for 0 t 1.
SOLUTION
B (0, 1)
y=1 x
x
0 A (1, 0)
f x (x, y) = 3x 2 y 5 = 0, f y (x, y) = 5y 4 x 3 = 0
The solutions are all the points with at least one zero coordinate. These points are not in the interior of the region, hence
there are no critical points in the interior of the region.
Step 2. Check the boundary. We consider each part of the boundary separately.
The segment O A: On this segment y = 0 and f (x, 0) = 0, hence f has the constant value 0 on this part of the
boundary.
The segment O B: On this segment x = 0, hence f (0, y) = 03 y 5 = 0. f attains the constant value 0 on this part of
the boundary.
5
The curve y = 1 x, 0 x 1: On this curve we have f (x, y) = g(x) = x 3 1 x .
5
We find the points on the interval 0 x 1 where g(x) = x 3 1 x has maximum and minimum values. Differen-
tiating g gives
5
4 1
5 5
4
g
(x) = 3x 2 1 x + x 3 5 1 x = 3x 2 1 x x 2.5 1 x
2 x 2
4 5
4 11
= x2 1 x 33 x x = x2 1 x 3 x
2 2
The points where g has extreme values in the interval 0 x 1 are x = 0, x = 1, and x = 121 36 . We find the
y-coordinates of these points from y = 1 x:
x = 0: y = 1 0 = 1
x = 1: y = 1 1 = 0
36 36 5
x= : y =1 =
121 121 11
We conclude
that the global extrema of f on the curve y = 1 x, 0 x 1 are obtained at the points (0, 1), (1, 0),
36 , 5 .
and 121 11
Step 3. Conclusions. We examine the values of f (x, y) = x 3 y 5 at the points obtained in the previous parts. The candi-
dates for global extrema are f = 0, the values of f on the segments O A and O B:
f (0, 1) = f (1, 0) = 0
36 5 36 3 5 5
f , = = 0.0005
121 11 121 11
We conclude that the minimum value of f (x, y) in the given domain is 0 and the maximum value is f 36 5
121 , 11
0.0005.
42. Show that the rectangular box (including a top and bottom) with fixed volume V with the smallest possible surface
area is a cube (Figure 21).
SOLUTION
Step 1. Find a function to be maximized. The surface area of the box with sides lengths x, y, z is
S = 2(x z + yz + x y) (1)
We express the surface area in terms of x and y alone using the equation V = x yz for the volume of the box. This
equation implies that z = xVy , hence by (1) we get
V V V V 2V 2V
S = S(x, y) = 2 x +y + xy = 2 + + xy = + + 2x y
xy xy y x y x
That is,
2V 2V
S= + + 2x y
y x
Step 2. Determine the domain. The variables x and y express lengths, therefore, they must be nonnegative. Also, S is
not defined if x = 0 or y = 0, therefore the domain is
We must find the minimum value of S on D. Because this domain is neither closed nor bounded, we have no guarantee
that an absolute minimum exists. However, it can be proved (see later Justifications) that S has a minimum value on D,
hence it must occur at a critical point in D.
754 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
Differentiating S = 2V 2V
y + x + 2x y and equating the partial derivatives to zero, we get
2V 2V
Sx (x, y) = 2 + 2y = 0, S y (x, y) = 2 + 2x = 0
x y
V V
3
y = 2 = 2/3 = V
3 V
V
The critical point is 3
V, 3
V . We find the value of z, using z = xVy :
V V
3
z=
3
3
= 2/3 = V
V V V
But how can we show that this critical point is a minumum? We provide two justifications.
Justification 1: Using the second derivative test, we have Sx x = 4V /x 3 , so Sx x ( 3 V ) = 4; S yy = 4V /y 3 , so
S yy ( 3 V ) = 4; and Sx y = 2. Thus, D = 4 4 22 = 12 > 0, and since Sx x > 0, we do indeed have a minimum surface
area. This makes sense, because when x or y go to 0 or to , then S (which is 2V /x + 2V /y + 2x y) clearly goes to .
Justification 2: We show that the function S(X, Y ) = 2V 2V
y + x + 2x y has a minimum value in the domain D =
{(x, y) : x > 0, y > 0}.
y
x
D
S(2, 2) = 2V + 8 = a0 > 8
Since lim Vx = , it follows by (1) that there exists 0 < r1 < 1 such that, for all 0 < x < r1 and for all values of y,
x0+
S(x, y) > a0
Since lim Vy = , it follows by (2) that there exists 0 < r2 < 1 such that, for all 0 < y < r1 and for all values of x,
y0+
S(x, y) > a0
R = {(x, y) : x r1 , y r2 , x y a0 }
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 755
xy = a 0
1 R
r2
x
0 r1 1
R is closed and bounded and S(x, y) is continuous in R, therefore S has a minimum value in R.
We now show that this minimum is also the minimum value of S in D. First notice that, by the above considerations,
S(x, y) > a0 for all (x, y) outside R. At the point (2, 2), S(2, 2) = a0 , and this point is in R, since 2 r1 , 2 r2 (recall
that 0 < r1 , r2 < 1) and 2 2 = 4 < 8 < a0 . Therefore, the minimum value of S(x, y) in R is also the minimum value
of S in D. We thus proved that S attains a minimum value on D.
43. Consider a rectangular box B with a bottom and sides but no top such that B has minimal surface area
among all boxes with fixed volume V .
(a) Do you think B is a cube as in the solution to Exercise 42? If not, how would you make its shape differ from a cube?
(b) Find the dimensions of B and compare with your response to (a).
SOLUTION
(a) Each of the variables x and y is the length of a side of three faces (for example, x is the length of the front, back, and
bottom sides), whereas z is the length of a side of four faces.
y
z
S = 2x z + 2yz + x y (1)
y
z
To express the surface in terms of x and y only, we use the formula for the volume of the box, V = x yz, giving z = xVy .
We substitute in (1) to obtain
V V 2V 2V
S = 2x + 2y + xy = + + xy
xy xy y x
That is,
2V 2V
S= + + x y.
y x
Step 2. Determine the domain. The variables x, y denote lengths, hence they must be nonnegative. Moreover, S is not
defined for x = 0 or y = 0. Since there are no other limitations on the variables, the domain is
We must find the minimum value of S on D. Because this domain is neither closed nor bounded, we are not sure that a
minimum value exists. However, it can be proved (in like manner as in Exercise 42) that S does have a minimum value
on D. This value occurs at a critical point in D, hence we set the partial derivatives equal to zero and solve. This gives
2V
Sx (x, y) = 2 + y = 0
x
756 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
2V
S y (x, y) = 2 + x = 0
y
The first equation gives y = 2V2 . Substituting in the second equation yields
x
2V x4 x3
x =x = x 1 =0
4V 2 2V 2V
x4
The solutions are x = 0 and x = (2V )1/3 . The solution x = 0 is not included in D, so the only solution is x = (2V )1/3 .
We find the value of y using y = 2V2 :
x
2V
y= = (2V )1/3
(2V )2/3
We conclude that the critical point, which is the point where the minimum value of S in D occurs, is (2V )1/3 , (2V )1/3 .
We find the corresponding value of z using z = xVy . We get
V V V 1/3 V 1/3
z= = 2/3 2/3 = 2/3 =
1/3
(2V ) (2V ) 1/3 2 V 2 4
We conclude that the sizes of the box with minimum surface area are
width: x = (2V )1/3 ;
length: y = (2V )1/3 ;
1/3
height: z = V4 .
$
n $
n $
n
m x 2j + b xj = xj yj
j=1 j=1 j=1
(xn , yn)
y = mx + b
(x2, y2)
(x1, y1)
(xj , yj)
x
FIGURE 22 The linear least-squares fit minimizes the sum of the squares of the vertical distances from the data points
to the line.
%n
2
SOLUTION We first find the critical points of E(m, b) = j=1 y j mx j b . Setting the partial derivatives equal
to zero, we get
$
n
$
n
E m (m, b) = 2 y j mx j b x j = 2 x j y j mx j b = 0
j=1 j=1
$
n
$
n
E b (m, b) = 2 y j mx j b (1) = 2 y j mx j b
j=1 j=1
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 757
$
n
= 2 y j mx j nb = 0
j=1
$
n $
n
yj m x j bn = 0
j=1 j=1
or
$
n $
n $
n
m x 2j + b xj = xj yj (1)
j=1 j=1 j=1
$
n $
n
m x j + bn = yj (2)
j=1 j=1
By Theorem 3 the minimum value of E(m, b) (if it exists) occurs at a critical point, which is the solution of equations
(1) and (2). It can be shown (see justification) that E(m, b) has a minimum value, hence E is minimized by the solution
of (1) and (2).
%
2
Justification: We show that E(m, b) = nj=1 y j mx j b has a minimum value. Let (m 0 , b0 ) be any point and
E(m 0 , b0 ) = E 0 . Since E(m, b) is increasing without bound as |m| and |b| , there exists a number R > 0
such that
The domain D = {(m, b) : |m| R and |b| R} is closed and bounded and E(m, b) is continuous on D, hence E has a
minimum value E M on D. The point (m 0 , b0 ) is in D (since E(m, b) > E 0 for all points (m, b) that are not in D), hence
E M E(m 0 , b0 ) = E 0 (4)
It follows by (1) and (2) that E M is the minimum value of E(m, b) on the entire mb-plane.
45. The power (in microwatts) of a laser is measured as a function of current (in milliamps). Find the linear least-squares
fit (Exercise 44) for the data points.
SOLUTION By Exercise 44, the coefficients of the linear least-square fit f (x) = mx + b are determined by the follow-
ing equations:
$
n $
n
m x j + bn = yj
j=1 j=1
$
n $
n $
n
m x 2j + b xj = xj yj (1)
j=1 j=1 j=1
$
6
x j = 1 + 1.1 + 1.2 + 1.3 + 1.4 + 1.5 = 7.5
j=1
$
6
y j = 0.52 + 0.56 + 0.82 + 0.78 + 1.23 + 1.50 = 5.41
j=1
758 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
$
6
x 2j = 12 + 1.12 + 1.22 + 1.32 + 1.42 + 1.52 = 9.55
j=1
$
6
x j y j = 1 0.52 + 1.1 0.56 + 1.2 0.82 + 1.3 0.78 + 1.4 1.23 + 1.5 1.50 = 7.106
j=1
7.5m + 6b = 5.41
9.55m + 7.5b = 7.106 (2)
We multiply the first equation by 9.55 and the second by (7.5), then add the resulting equations. This gives
We now substitute b = 1.5519 in the first equation in (2) and solve for m:
7.5m + 6 (1.5519)=5.41
m = 1.9629
7.5m=14.7214
46. Prove that if , 1 are numbers such that 1 + 1 = 1, then the following inequality holds for all x, y 0:
1 1
x + x xy
For example,
1 3 2 3/2 2 5/2 3 5/3
x + y x y, x + y x y, etc.
3 3 5 5
1 1
f (x, y) = x + y x y
We must show that f (x, y) 0 for x 0, y 0. Notice that f (0, 0) = 0, hence we must show that 0 is the minimum
value of f in the domain x 0, y 0. For R > 1, we denote the square D R = {(x, y) : 0 x R, 0 y R}. f
is continuous on the closed and bounded set D R , therefore f has a minimum value on D. We show that the minimum
value is zero.
Step 1. Examine the critical points in the interior of D R . We set the partial derivatives equal to zero and solve:
1 1
f x (x, y) = x + y x y = x 1 y = 0
x
1 1
f y (x, y) = x + y x y = y 1 x = 0
y
The first equation gives y = x 1 and the second gives x = y 1 . Therefore, the critical points (a, b) satisfy b = a 1
and a = b 1 . We compute the value of f at the critical points, in the interior of the square:
1 1 a b
f (a, b) = a + b ab = a 1 + b 1 ab
ab ba 1 1
= + ab = ab + ab = ab 1 ab = 0
Step 2. Find the minimum on the boundary of D R . We consider each part of the boundary separately.
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 759
A = (0, R) B = (R, R)
DR
x
0 C = (R, 0)
We now compute the values of f (x, y) = 1 x + 1 y x y at the points on the boundary of D R that we found above.
These points are:
(0, 0), (R 1/( 1) , R), (0, R), (R, R), (R, R 1/( 1) ), (R, 0)
Recalling that 1 + 1 = 1, hence = 1 , and = 1 , we get
f (0, 0) = 0
1 1
f R 1/( 1) , R = R 1 + R R 1/( 1) R
1 1
R 1 R 1 +1
1
= R 1 + 1
= R 1 R 1 = 0
1
f (0, R) = R >0
1 1 1+ 1 1 1
f R, R 1/( 1) = R + R 1 R 1 = 1 R 1 + R 1 R 1 = 0
1
f (R, 0) = R >0
The minimum value of f on the boundary of D R is therefore 0. Since the value at the critical points inside D R is also
zero, 0 is the minimum value of f on D R . Now, since R is an arbitrary positive number, we conclude that 0 is the absolute
minimum of f in the entire quadrant x 0, y 0.
* *
Remark: We show that 1 R + 1 R R 2 . Since 1 R = 0R x 1 d x and 1 R = 0R x 1 d x, the inequality
can be rewritten as
R R
x 1 d x + x 1 d x R 2 (1)
0 0
Notice that the functions x 1 and x 1 are inverse functions, since the equalities = 1 and = 1 imply that
1=
and 1 = . Hence,
1
(x 1 ) = x ( 1)( 1) = x = x
760 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
1
(x 1 ) = x ( 1)( 1) = x = x
By properties of inverse functions, their graphs are symmetric with respect to the line y = x. Therefore the area between
one graph and the x-axis is equal to the area between the other graph and the y-axis, over the intervals 0 x R
*
and 0 y R, respectively. In particular, the integral 0R x 1 d x can be thought of as the area between the graph
of y = x 1 and the y-axis, from y = 0 to y = R. So, recalling that the area of the squares is R 2 , and assuming that
2 , we have the following inequality (see figure, and note that the first integral gives the area of the lower-right
part of the square as well as that extra spike at the top of the square, while the second integral gives the area of the rest of
the square, as per our discussion above):
R R
R2 x 1 d x + x 1 d x
0 0
y
y = xa 1
R
(1, 1)
1
y = xb 1
x
1 R
47. The following problem was proposed by Fermat as a challenge to the Italian scientist Evangelista Torricelli
(16081647), a student of Galileo and inventor of the barometer. Given three points A = (a1 , a2 ), B = (b1 , b2 ), and
C = (c1 , c2 ) in the plane, find the point P = (x, y) that minimizes the sum of the distances
f (x, y) = A P + B P + C P
(a) Write out f (x, y) as a function of x and y, and show that f (x, y) is differentiable except at the points A, B, C.
(b) Define the unit vectors
AP BP CP
e = , f = , g =
A P B P C P
Show that the condition f = 0 is equivalent to
e+f+g=0 3
Prove that Eq. (3) holds if and only if the mutual angles between the unit vectors are all 120 .
(c) Define the Fermat point to be the point P such that angles between the segments A P, B P, C P are all 120 . Conclude
that the Fermat point solves the minimization problem (Figure 23).
(d) Show that the Fermat point does not exist if one of the angles in ABC is 120 . Where does the minimum occur
in this case? Hint: The minimum must occur at a point where f (x, y) is not differentiable.
B
A
f
e
P
B
g
140
C
A
C
(A) P is the Fermat point (B) Fermat point does not exist.
(the angles between e,
f, and g are all 120).
FIGURE 23
SOLUTION
(a)
P(x, y)
A(a 1 , a 2 )
B(b 1 , b 2 )
C(c 1 , c 2 )
S E C T I O N 15.7 Optimization in Several Variables (ET Section 14.7) 761
For all (x, y) other then (a1 , a2 ), (b1 , b2 ), (c1 , c2 ) the partial derivatives are continuous, therefore the Criterion for
Differentiability implies that f is differentiable at all points other than A, B, and C.
(b)
P e
A
f
B
Therefore, the condition f = 0 is equivalent to e + f + g = 0. We now show that Eq. (3) holds if and only if the mutual
angles between the unit vectors are all 120 . We place the axes so that the positive x-axis is in the direction of e.
y
g
a f
q x
e
762 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
Let and be the angles that f and g make with e, respectively. Hence,
Substituting in e + f + g = 0 we have
or
cos + cos + 1 = 0
sin + sin = 0
= 180 + , = 360
We substitute each solution in the first equation and solve for . This gives
= 180 + = 360
cos(180 + ) + cos + 1 = 0 cos(360 ) + cos + 1 = 0
cos + cos + 1 = 0 cos + cos + 1 = 0
1=0 2 cos = 1
cos = 12
= 120 = 240
= 360 = 240 = 360 = 120
or
e = 1, 0 , f = cos 120 , sin 120 , g = cos 240 , sin 240
f
120
x
120 e
120
g
or
y
B
A C
f
g e 120 g
120 P 120
e P
x 120 120 A 120
120 e
120
120
f f g
B C
120 120
P
120
B C
Suppose that the Fermat point P exists. The values of f at the critical points are
f ( A) = AB + AC
f (B) = AB + BC
f (C) = AC + BC
f (P) = A P + B P + PC
We show that f (P) < f ( A). Similarly it can be shown that also f (P) < f (B) and f (P) < f (C). By the Cosine
Theorem for the triangles AB P and AC P we have
2 2 2 2
AB = A P + B P 2 A P B P cos 120 = A P + B P + A P B P
2 2 2 2
AC = A P + C P 2 A P PC cos 120 = A P + C P + A P PC
Hence
2 2 2 2
f ( A) = AB + AC = AP + B P + AP B P + A P + C P + A P PC
A P + B P + PC = f (P)
The last inequality can be verified by squaring and transfering sides. Its best to use a computer to help with the algebra;
its a daunting task to do by hand.
(d) We show that if one of the angles of ABC is 120 , then the Fermat point does not exist. Notice that the Fermat
point (if it exists) must fall inside the triangle ABC.
A
C
B 120
120
120
1 2
120 120
P
1 120 1
B C
Therefore,
A = A1 + A2 = 60 B1 + 60 C1 = 120 (B1 + C1 ) < 120
764 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
We thus showed that if the Fermat point exists, then A < 120 . Similarly, one shows also that B and C must be
smaller than 120 . We conclude that if one of the angles in ABC is equal or greater than 120 , then the Fermat point
does not exist. In that case, the minimum value of f (x, y) occurs at a point where f x or f y do not exist, that is, at one of
the points A, B, or C.
2. Figure 8 shows a constraint g(x, y) = 0 and the level curves of a function f . In each case, determine whether f has
a local minimum, local maximum, or neither at the labeled point.
4
3
2
f 1 f
4
3
A 2 B
1
g(x, y) = 0 g(x, y) = 0
FIGURE 8
SOLUTION The level curve f (x, y) = 2 is tangent to the constraint curve at the point A. A close level curve that
intersects the constraint curve is f (x, y) = 1, hence we may assume that f has a local maximum 2 under the constraint
at A. The level curve f (x, y) = 3 is tangent to the constraint curve. However, in approaching B under the constraint,
from one side f is increasing and from the other side f is decreasing. Therefore, f (B) is neither local minimum nor
local maximum of f under the constraint.
3. On the contour map in Figure 9:
(a) Identify the points where f = g for some scalar .
(b) Identify the minimum and maximum values of f (x, y) subject to g(x, y) = 0.
y
4 2 2 6
Graph of
g (x, y) = 0
6 2 2 6
SOLUTION
(a) The gradient g is orthogonal to the constraint curve g(x, y) = 0, and f is orthogonal to the level curves of f .
These two vectors are parallel at the points where the level curve of f is tangent to the constraint curve. These are the
points A, B, C, D, E in the figure:
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 765
6 2 2 6
fA, gA g(x, y) = 0
A B
C
E D
6 2 2 6
(b) The minimum and maximum occur where the level curve of f is tangent to the constraint curve. The level curves
tangent to the constraint curve are
Therefore the global minimum of f under the constraint is 4 and the global maximum is 6.
Exercises
1. Use Lagrange multipliers to find the extreme values of the function f (x, y) = 2x + 4y subject to the constraint
g(x, y) = x 2 + y 2 5 = 0.
(a) Show that the Lagrange equation f = g gives x = 1 and y = 2.
(b) Show that these equations imply = 0 and y = 2x.
(c) Use the constraint equation to determine the possible critical points (x, y).
(d) Evaluate f (x, y) at the critical points and determine the minimum and maximum values.
SOLUTION
(a) The Lagrange equations are determined by the equality f = g. We find them:
f = f x , f y = 2, 4 , g = gx , g y = 2x, 2y
Hence,
or
(2x) = 2 x = 1
(2y) = 4 y = 2
(b) The Lagrange equations in part (a) imply that = 0. The first equation implies that x = 1 and the second equation
gives y = 2 . Therefore y = 2x.
(c) We substitute y = 2x in the constraint equation x 2 + y 2 5 = 0 and solve for x and y. This gives
x 2 + (2x)2 5 = 0
5x 2 = 5
x2 = 1 x 1 = 1, x2 = 1
Extreme values can also occur at the points where g = 2x, 2y = 0, 0. However, (0, 0) is not on the constraint.
(d) We evaluate f (x, y) = 2x + 4y at the critical points, obtaining
Since f is continuous and the graph of g = 0 is closed and bounded, global minimum and maximum points exist. So
according to Theorem 1, we conclude that the maximum of f (x, y) on the constraint is 10 and the minimum is 10.
2. Find the extreme values of f (x, y) = x 2 + 2y 2 subject to the constraint g(x, y) = 4x 6y = 25.
(a) Show that the Lagrange equations yield 2x = 4 , 4y = 6 .
(b) Show that if x = 0 or y = 0, then = 0 and the Lagrange equations give x = y = 0. Since (0, 0) does not satisfy
the constraint, you may assume that x and y are nonzero.
766 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
6
4
2
24 36 g(x, y) = 0
0 6 12
2 P
4
8 6 4 2 0 2 4 6 8
FIGURE 10 Level curves of f (x, y) = x 2 + 2y 2 and constraint g(x, y) = 4x 6y 25 = 0.
SOLUTION
(a) The gradients f and g are
f = g
2x, 4y = 4, 6
or
2x = 4
4y = 6
(b) If x = 0, the first equation gives 0 = 4 or = 0. Substituting in the second equation gives 4y = 0 or y = 0.
Similarly, if y = 0, the second equation implies that = 0, hence by the first equation also x = 0. That is, if x = 0, then
y = 0 and if y = 0 also x = 0. The point (0, 0) does not satisfy the equation of the constraint, hence we may assume
that x = 0 and y = 0.
(c) The first equation in part (a) gives = x2 . Substituting in the second equation we get
x 3
4y = 6 = 3x y= x
2 4
(d) We substitute y = 34 x in the constraint 4x 6y = 25 and solve for x and y. This gives
3
4x 6 x = 25
4
9
4x + x = 25
2
50 3 50 75
17x = 50 x= , y= =
17 4 17 34
We conclude that there is a unique critical point, which is 50 75
17 , 34 .
(e) We now refer to Figure 10. As (x, y) moves away from P along the line g(x, y) = 0, the values of f (x, y) increase,
hence P corresponds to a minimum value of f .
3. Apply the method of Lagrange multipliers to the function f (x, y) = (x 2 + 1)y subject to the constraint x 2 + y 2 = 5.
Hint: First show that y = 0; then treat the cases x = 0 and x = 0 separately.
SOLUTION We first write out the Lagrange Equations. We have f = 2x y, x 2 + 1 and g = 2x, 2y. Hence, the
Lagrange Condition for g = 0 is
f = g
2x y, x 2 + 1 = 2x, 2y
2x y = (2x) 2x(y ) = 0
(1)
x 2 + 1 = (2y) x 2 + 1 = 2 y
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 767
The second equation implies that y = 0, since there is no real value of x such that x 2 + 1 = 0. Likewise, = 0. The
solutions of the first equation are x = 0 and y = .
Case 1: x = 0. Substituting x = 0 in the second equation gives 2 y = 1, or y = 21 . We substitute x = 0, y = 21
(recall that = 0) in the constraint to obtain
1 1 1 1
02 + =5 4 2 = = =
4 2 5 20 2 5
The corresponding values of y are
1 1
y= = 5 and y= = 5
1
2 1
2 5
2
2 5
We obtain the critical points:
0, 5 and 0, 5
Case 2: x = 0. Then the first equation in (1) implies y = . Substituting in the second equation gives
x 2 + 1 = 2 2 x 2 = 2 2 1
2 2 1 + 2 = 5
3 2 = 6
2 = 2 = 2
The solution (x, y) are thus
= 2: y= 2, x = 2 2 1 = 3
= 2: y = 2, x = 2 2 1 = 3
In Exercises 413, find the minimum and maximum values of the function subject to the given constraint.
4. f (x, y) = 2x + 3y, x 2 + y2 = 4
SOLUTION We find the extreme values of f (x, y) = 2x + 3y under the constraint g(x, y) = x 2 + y 2 4 = 0.
Step 1. Write the Lagrange Equations. We have f = 2, 3 and g = 2x, 2y, hence the Lagrange Condition is
f = g
2, 3 = 2x, 2y
2 = (2x)
3 = (2y)
768 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
Step 2. Solve for x and y using the constraint. The two equations imply that x = 0 and y = 0, hence
1 3
= and =
x 2y
The two expressions for must be equal, so we obtain
1 3 3
= y= x
x 2y 2
f = g
2x, 2y = 2, 3
2x = 2
2y = 3
Step 2. Solve for in terms of x and y. Notice that if x = 0, then the first equation gives = 0, therefore by the second
equation also y = 0. The point (0, 0) does not satisfy the constraint. Similarly, if y = 0 also x = 0. We therefore may
assume that x = 0 and y = 0 and obtain by the two equations:
2
=x and = y.
3
Step 3. Solve for x and y using the constraint. Equating the two expressions for gives
2 3
x= y y= x
3 2
We substitute y = 32 x in the constraint 2x + 3y = 6 and solve for x and y:
3
2x + 3 x =6
2
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 769
12 3 12 18
13x = 12 x= , y= =
13 2 13 13
We obtain the critical point 12 , 18 .
13 13
Step 4. Calculate f at the critical point. We evaluate f (x, y) = x 2 + y 2 at the critical point:
2
12 18 12 2 18 468
f , = + = 2.77
13 13 13 13 169
Rewriting the constraint as y = 23 x + 2, we see that as |x| + then so does |y|, and hence x 2 + y 2 is increasing
without bound on the constraint as |x| . We conclude that the value 468/169 is the minimum value of f under the
constraint, rather than the maximum value.
6. f (x, y) = 4x 2 + 9y 2 , xy = 4
SOLUTION We find the extreme values of f (x, y) = 4x 2 + 9y 2 under the constraint g(x, y) = x y 4 = 0.
Step 1. Write out the Lagrange Equations. The gradient vectors are f = 8x, 18y and g = y, x, hence the
Lagrange condition is
f = g
8x, 18y = y, x
or
8x = y
18y = x
Step 2. Solve for in terms of x and y. We may assume that x = 0 and y = 0, since the points with x = 0 or y = 0 do
not satisfy the constraint. The two equations give
8x 18y
= and =
y x
Step 3. Solve for x and y using the constraint. We equate the two expressions for to obtain
8x 18y 2
= 8x 2 = 18y 2 y= x
y x 3
The constraint x y = 4 implies that x and y have the same sign, hence y = 23 x. We substitute y = 23 x in the constraint
and solve for x and y:
2
x x =4 x2 = 6 x1 = 6, x2 = 6
3
2 2 2 2
y1 = 6=2 , y2 = 6 = 2
3 3 3 3
The critical points are thus
2 2
6, 2 , 6, 2
3 3
Extreme values can also occur at the point where g = y, x = 0, 0, . However, the point (0, 0) is not on the
constraint.
Step 4. Calculate f at the critical points. We evaluate f (x, y) = 4x 2 + 9y 2 at the critical points:
2 2
f 6, 2 = 4 6 + 9 4 = 48
3 3
2 2
f 6, 2 = 4 6 + 9 4 = 48
3 3
On the constraint, y = 4x , thus f (x, y) = f x, 4x = h(x) = 4x 2 + 144 . Since lim h(x) = limx h(x) = , h
x2 x
has a global minimum of 48 (but no maximum!) on (, ).
770 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
7. f (x, y) = x y, 4x 2 + 9y 2 = 32
SOLUTION We find the extreme values of f (x, y) = x y under the constraint g(x, y) = 4x 2 + 9y 2 32 = 0.
Step 1. Write out the Lagrange Equation. The gradient vectors are f = y, x and g = 8x, 18y, hence the La-
grange Condition is
f = g
y, x = 8x, 18y
y = (8x)
x = (18y)
Step 2. Solve for in terms of x and y. If x = 0, then the Lagrange equations also imply that y = 0 and vice versa.
Since the point (0, 0) does not satisfy the equation of the constraint, we may assume that x = 0 and y = 0. The two
equations give
y x
= and =
8x 18y
Step 3. Solve for x and y using the constraint. We equate the two expressions for to obtain
y x 2
= 18y 2 = 8x 2 y= x
8x 18y 3
We now substitute y = 23 x in the equation of the constraint and solve for x and y:
2 2
4x 2 + 9 x = 32
3
4x 2
4x 2 + 9 = 32
9
8x 2 = 32 x = 2, x =2
2 4 2 4 2 4 2 4
y= (2) = , y = (2) = , y= 2= , y = 2=
3 3 3 3 3 3 3 3
We obtain the following critical points:
4 4 4 4
2, , 2, , 2, , 2,
3 3 3 3
Extreme values can also occur at the point where g = 8x, 18y = 0, 0, that is, at the point (0, 0). However, the point
does not lie on the constraint.
Step 4. Calculate f at the critical points. We evaluate f (x, y) = x y at the critical points:
4 4 8
f 2, = f 2, =
3 3 3
4 4 8
f 2, = f 2, =
3 3 3
Since f is continuous and the constraint is a closed and bounded set in R 2 (an ellipse), f attains global extrema on the
constraint. We conclude that 83 is the maximum value and 83 is the minimum value.
8. f (x, y) = x 2 y + x + y, xy = 4
SOLUTION Under the constraint x y = 4, then f (x, y) = x(x y) + x + y = 4x + x + 4x . Therefore, as x 0+,
f (x, y) + on the constraint, and as x 0, f (x, y) . Therefore there are no minimum and maximum
values of f (x, y) under the constraint.
9. f (x, y) = x 2 + y 2 , x 4 + y4 = 1
SOLUTION We find the extreme values of f (x, y) = x 2 + y 2 under the constraint g(x, y) = x 4 + y 4 1 = 0.
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 771
Step 1. Write out the Lagrange Equations. We have f = 2x, 2y and g = 4x 3 , 4y 3 , hence the Lagrange Condition
f = g gives
2x, 2y = 4x 3 , 4y 3
or
2x = 4x 3 x = 2 x 3
(1)
2y = 4y 3 y = 2 y 3
Step 2. Solve for in terms of x and y. We first assume that x = 0 and y = 0. Then the Lagrange equations give
1 1
= and =
2x 2 2y 2
Step 3. Solve for x and y using the constraint. Equating the two expressions for gives
1 1
= 2 y2 = x 2 y = x
2x 2 2y
We now substitute y = x in the equation of the constraint x 4 + y 4 = 1 and solve for x and y:
x 4 + (x)4 = 1
2x 4 = 1
1 1 1
x4 = x = 1/4 , x = 1/4
2 2 2
The corresponding values of y are obtained by the relation y = x. The critical points are thus
1 1 1 1 1 1 1 1
, , , , , , , (2)
21/4 21/4 21/4 21/4 21/4 21/4 21/4 21/4
We examine the case x = 0 or y = 0. Notice that the point (0, 0) does not satisfy the equation of the constraint, hence
either x = 0 or y = 0 can hold, but not both at the same time.
Case 1: x = 0. Substituting x = 0 in the constraint x 4 + y 4 = 1 gives y = 1. We thus obtain the critical points
Case 2: y = 0. We may interchange x and y in the discussion in case 1, and obtain the critical points:
Combining (2), (3), and (4) we conclude that the critical points are
1 1 1 1 1 1
A1 = , , A 2 = , , A 3 = , ,
21/4 21/4 21/4 21/4 21/4 21/4
1 1
A4 = 1/4 , 1/4 , A5 = (0, 1), A6 = (0, 1), A7 = (1, 0), A8 = (1, 0)
2 2
The point where g = 4x 3 , 4y 3 = 0, 0, that is, (0, 0), does not lie on the constraint.
Step 4. Compute f at the critical points. We evaluate f (x, y) = x 2 + y 2 at the critical points:
2 2
1 1 2
f ( A1 ) = f ( A2 ) = f ( A3 ) = f ( A4 ) = + = 1/2 = 2
21/4 21/4 2
f ( A5 ) = f ( A6 ) = f ( A7 ) = f ( A8 ) = 1
or
2x y 4 = (2x) x y4 = x
(1)
4y 3 x 2 = (4y) x 2 y3 = y
Step 2. Solve for in terms of x and y. Notice that if x = 0 or y = 0, then f (x, y) = x 2 y 4 has the value 0, which is
the minimum value (since f (x, y) 0). We thus assume that x = 0 and y = 0. The Lagrange equations (1) give
x y4 x 2 y3
= = y4, = = x 2 y2
x y
Step 3. Solve for x and y using the constraint. Equating the two expressions for gives
y4 = x 2 y2 y2 = x 2 y = x
Substituting y = x in the equation of the constraint x 2 + 2y 2 = 6 and solving for x and y gives
x 2 + 2x 2 = 6
3x 2 = 6
x2 = 2 x= 2, x = 2
The corresponding value of y is obtained by the relation y = x. We obtain the following points:
2, 2 , 2, 2 , 2, 2 , 2, 2
Extreme values can occur also at the point where g = 2x, 4y = 0, 0, that is, (0, 0). However, this point does not lie
on the constraint.
Step 4. Computing f at the critical points. We evaluate f (x, y) = x 2 y 4 at the critical points:
2 4 6
f 2, 2 = f 2, 2 = f 2, 2 = f 2, 2 = 2 2 = 2 =8
Recall that there are critical points with x = 0 or y = 0 at which the value of f is zero. Since f has global extrema on
the ellipse x 2 + 2y 2 = 6, we conclude that the minimum value of f on the constraint is 0 and the maximum value is 8.
11. f (x, y, z) = 3x + 2y + 4z, x 2 + 2y 2 + 6z 2 = 1
SOLUTION We find the extreme values of f (x, y, z) = 3x + 2y + 4z under the constraint g(x, y, z) = x 2 + 2y 2 +
6z 2 1 = 0.
Step 1. Write out the Lagrange Equations. The gradient vectors are f = 3, 2, 4 and g = 2x, 4y, 12z, therefore
the Lagrange Condition f = g is:
Substituting x = 92 z and y = 32 z in the equation of the constraint x 2 + 2y 2 + 6z 2 = 1 and solving for z we get
2 2
9 3
z + 2 z + 6z 2 = 1
2 2
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 773
123 2 2 2
z =1 z1 = , z2 =
4 123 123
9 2 9 3 2 3 2
x1 = = , y1 = = , z1 =
2 123 123 2 123 123 123
9 2 9 3 2 3 2
x2 = = , y2 = = , z2 =
2 123 123 2 123 123 123
We obtain the following critical points:
9 3 2 9 3 2
p1 = , , and p2 = , ,
123 123 123 123 123 123
Critical points are also the points on the constraint where g = 0. However, g = 2x, 4y, 12z = 0, 0, 0 only at the
origin, and this point does not lie on the constraint.
Step 4. Computing f at the critical points. We evaluate f (x, y, z) = 3x + 2y + 4z at the critical points:
27 6 8 41 41
f ( p1 ) = + + = = 3.7
123 123 123 123 3
27 6 8 41 41
f ( p2 ) = = = 3.7
123 123 123 123 3
Since f is continuous and the constraint is closed and bounded in R 3 , f has global extrema under the constraint. We
conclude that the minimum value of f under the constraint is about 3.7 and the maximum value is about 3.7.
12. f (x, y, z) = x 2 y z, x 2 y2 + z = 0
SOLUTION We show that the function f (x, y, z) = x 2 y z does not have minimum and maximum values subject
to the constraint x 2 y 2 + z = 0. Notice that the curve (x, x, 0) lies on the constraint, since it satisfies the equation of
the constraint. On this curve we have
f (x, y, z) = f (x, x, 0) = x 2 x 0 = x 2 x
Since lim (x 2 x) = , f does not have a maximum value subject to the constraint. Observe that the curve
x
0, z, z also lies on the constraint, and we have
f (x, y, z) = f 0, z, z = 02 z z = z + z
Since lim z + z = , f does not attain a minimum value on the constraint either.
z
13. f (x, y, z) = x y + 3x z + 2yz, 5x + 9y + z = 10
SOLUTION We show that f (x, y, z) = x y + 3x z + 2yz does not have minimum and maximum values subject to the
constraint g(x, y, z) = 5x + 9y + z 10 = 0. First notice that the curve c1 : (x, x, 10 14x) lies on the surface of the
constraint since it satisfies the equation of the constraint. On c1 we have,
The limit lim (3x 2 + 10x) = implies that f does not have a maximum value subject to the constraint.
x
1
5
0 3
0 1
1
1 3
2 5
2 1 0 1 2 3
FIGURE 11 Contour map of f (x, y) = x 3 + x y + y 3 and graph of g(x, y) = x 3 x y + y 3 = 1.
SOLUTION
(a) The gradients of f are f = 3x 2 + y, x + 3y 2 and g = 3x 2 y, x + 3y 2 , hence the Lagrange Condition
f = g is
3x 2 + y, x + 3y 2 = 3x 2 y, x + 3y 2
or
3x 2 + y = (3x 2 y)
(1)
x + 3y 2 = (x + 3y 2 )
Notice that if 3x 2 y = 0, the first equation implies that also 3x 2 + y = 0, hence y = 0 and x = 0. Since the point
(0, 0) does not satisfy the equation of the constraint, we may assume that 3x 2 y = 0. Similarly, if x + 3y 2 = 0, the
second equation implies that also x + 3y 2 = 0, therefore x = y = 0. We thus may also assume that x + 3y 2 = 0.
Using these assumptions, we have by (1):
3x 2 + y x + 3y 2
= , =
3x 2 y x + 3y 2
Equating the two expressions for we get
3x 2 + y x + 3y 2
=
3x 2 y x + 3y 2
3x 2 + y x + 3y 2 = x + 3y 2 3x 2 y
3x 3 + 9x 2 y 2 yx + 3y 3 = 3x 3 x y + 9x 2 y 2 3y 3
x 3 = y3 x=y
y3 y2 + y3 = 1
2y 3 y 2 1 = 0
We notice that y = 1 is a root of 2y 3 y 2 1, hence this polynomial is divisible by y 1. Long division yields
(y 1)(2y 2 + y + 1) = 0
Since 2y 2 + y + 1 > 0 for all y (the discriminant is negative), the only solution is y = 1. Then, x = y = 1 and the only
critical point is (1, 1).
(b) Figure 11 suggests that the values of f (x, y) are increasing as (x, y) approaches the critical point (1, 1) along the
constraint. Therefore, f has a local maximum at P, subject to the constraint.
(c) Figure 11 shows the behavior of f and g only in the range 3 x 3, so we cannot know whether P is a global
maximum, but it is reasonable to guess that it is.
15. Use Lagrange multipliers to find the point (a, b) on the graph of y = e x , where the value ab is as small as possible.
SOLUTION We must find the point where f (x, y) = x y has a minimum value subject to the constraint g(x, y) =
e x y = 0.
Step 1. Write out the Lagrange Equations. Since f = y, x and g = e x , 1 , the Lagrange Condition f = g
is
y, x = e x , 1
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 775
y = ex
x =
Step 2. Solve for in terms of x and y. The Lagrange equations imply that
= yex and = x
Step 3. Solve for x and y using the constraint. We equate the two expressions for to obtain
yex = x y = xe x
e x (xe x ) = 0
e x (1 + x) = 0
Since e x = 0 for all x, we have x = 1. The corresponding value of y is determined by the relation y = xe x . That is,
y = (1)e1 = e1
(1, e1 )
Step 4. Calculate f at the critical point. We evaluate f (x, y) = x y at the critical point.
f (1, e1 ) = (1) e1 = e1
We conclude (see Remark) that the minimum value of x y on the graph of y = e x is e1 , and it is obtained for x = 1
and y = e1 .
Remark: Since the constraint is not bounded, we need to justify the existence of a minimum value. The values
f (x, y) = x y on the constraint y = e x are f (x, e x ) = h(x) = xe x . Since h(x) > 0 for x > 0, the minimum
value (if it exists) occurs at a point x < 0. Since
x 1
lim xe x = lim = lim = lim e x = 0,
x x ex x ex x
then for x < some negative number R, we have | f (x) 0| < 0.1, say. Thus, on the bounded region R x 0, f
has a minimum value of e1 0.37, and this is thus a global minimum (for all x).
16. Find the rectangular box of maximum volume if the sum of the lengths of the edges is 300 cm.
SOLUTION We denote by x, y, and z the dimensions of the rectangular box.
y
x
Then the volume of the box is x yz. We must find the values of x, y and z that maximize the volume f (x, y, z) = x yz,
subject to the constraint g(x, y, z) = x + y + z = 300, x 0, y 0, z 0. (One could also argue that the sums of the
lengths of the edges is 4x + 4y + 4z = 300, but that would give a different answer, of course. Instead, we will choose to
interpret the problem with the constraint x + y + z = 300).
Step 1. Write out the Lagrange Equations. The Lagrange Condition is
f = g
yz, x z, x y = 1, 1, 1
yz =
xz =
xy =
Step 2. Solve for in terms of x, y, and z. The Lagrange equations already give in terms of x, y, and z. Equating the
expressions for we get yz = x z = x y.
776 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
yz = x z z(x y) = 0
xy = xz x(z y) = 0
If x = 0, y = 0, or z = 0, the volume has the minimum value 0. We thus may assume that x = 0, y = 0, and z = 0.
The first equation implies that x = y and the second equation gives z = y. We now substitute x = y and z = y in the
constraint x + y + z = 300 and solve for y:
y + y + y = 300
3y = 300 y = 100
Therefore, x = 100 and z = 100. The critical point is (100, 100, 100).
Step 4. Conclusions. The value of f (x, y, z) = x yz at the critical point is
The constraint x + y + z = 300, x 0, y 0, z 0 is the part of the plane x + y + z = 300 that lies in the first octant.
This is a bounded and closed set in R 3 . Since f is continuous on this set, f has global extreme values on this set. The
minimum value is zero (obtained if one of the variables is zero), hence the value 106 is the maximum value. We conclude
that the box with maximum value is a cube of edge 100 cm.
z
300
300 300 y
17. The surface area of a right-circular cone of radius r and height h is S = r r 2 + h 2 , and its volume is V = 13 r 2 h.
(a) Determine the ratio h/r for the cone with given surface area S and maximal volume V .
(b) What is the ratio h/r for a cone with given volume V and minimal surface area S?
(c) Does a cone with given volume V and maximal surface area exist?
SOLUTION
(a) Let S0 denote a given surface area. We must find the ratio hr for which the function V (r, h) = 13 r 2 h has maximum
value under the constraint S(r, h) = r r 2 + h 2 = r 4 + h 2 r 2 = S0 .
Step 1. Write out the Lagrange Equation. We have
2r h r 2 2r 3 + h 2 r hr 2
V = , and S = ,
3 3 r 4 + h2r 2 r 4 + h 2r 2
2r h 2r 3 + h 2 r 2h 2r 2 + h 2
= =
3 r 4 + h 2r 2 3 r 4 + h 2r 2
r2 hr 2 1 h
= =
3 r 4 + h 2r 2 3 r 4 + h2r 2
Step 2. Solve for in terms of r and h. These equations yield two expressions for that must be equal:
2h r 4 + h 2 r 2 1 4
= = r + h2r 2
3 2r + h
2 2 3h
Step 3. Solve for r and h using the constraint. We have
2h r 4 + h 2 r 2 1 4
= r + h 2r 2
3 2r + h
2 2 3h
1 1
2h =
2r 2 + h 2 h
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 777
h
2h 2 = 2r 2 + h 2 h 2 = 2r 2 = 2
r
We substitute h 2 = 2r 2 in the constraint r r 2 + h 2 = S0 and solve for r . This gives
r r 2 + 2r 2 = S0
r 3r 2 = S0
S 2S
3 r 2 = S0 r2 = 0 ,
h 2 = 2r 2 = 0
3 3
Extreme values can occur also at points on the constraint where S = 2r 4 +h 2r 2 , 4hr 2 2 = 0, 0, that is, at
2 2 2
r +h r r +h r
(r, h) = (0, h), h = 0. However, since the radius of the cone is positive (r > 0), these points are irrelevant. We conclude
that for the cone with surface area S0 and maximum volume, the following holds:
h 2S0 S
= 2, h = , r = 0
r 3 3
1 2
V (r, h) = r h = V0
3
Using the gradients computed in part (a), the Lagrange Condition S = V gives the following equations:
2r 3 + h 2 r 2r h 2r 2 + h 2 2h
= =
r +h r
4 2 2 3 r +h r
4 2 2 3
hr 2 r2 h
= =
r 4 + h2r 2 3 r 4 + h2r 2 3
1 2r 2 + h 2 h
= =
3 2h r 4 + h 2 r 2 r 4 + h 2r 2
2r 2 + h 2
=h
2h
2r 2 + h 2 = 2h 2
h
2r 2 = h 2 = 2
r
We conclude that the ratio hr for a cone with a given volume and minimal surface area is
h
= 2
r
(c) The constant V = 1 gives 13 r 2 h = 1 or h = 3 2 . As r , we have h 0, therefore
r
lim S(r, h) = lim r r 2 + h 2 =
r r
h0 h0
That is, S does not have maximum value on the constraint, hence there is no cone of volume 1 and maximal surface area.
18. In Example 1, we found the maximum of f (x, y) = x + 2y on the ellipse 3x 2 + 4y 2 = 3. Solvethis problem again
3
without using Lagrange multipliers. First, show that the ellipse is parametrized by x = cos t, y = sin t. Then find
2
3
the maximum value of f cos t, sin t using single-variable calculus. Which method do you find easier?
2
778 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
SOLUTION We want to find the maximum of f (x, y) = x + 2y on the ellipse 3x 2 + 4y 2 = 3 without using Lagrange
multipliers. We rewrite the equation of the ellipse in the form
2
y
x + =1
2
3
2
x2 y2
2
+ 2 =1
a b
(x, y) (x, y)
(x, y) (x, y)
x2 y2
FIGURE 12 Rectangle inscribed in the ellipse + = 1.
a2 b2
SOLUTION Since (x, y) is in the first quadrant, x > 0 and y > 0. The area of the rectangle is 2x 2y = 4x y. The
vertices lie on the ellipse, hence their coordinates (x, y) must satisfy the equation of the ellipse. Therefore, we must
find the maximum value of the function f (x, y) = 4x y under the constraint
x2 y2
g(x, y) = 2 + 2 = 1, x > 0, y > 0.
a b
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 779
Step 1. Write out the Lagrange Equations. The gradient vectors are f = 4y, 4x and g = 2x2 , 2y2 , hence the
a b
Lagrange Condition f = g gives
2x 2y
4y, 4x = 2 , 2
a b
or
2x x
4y = 2y = 2
a2 a
2y y
4x = 2x = 2
b2 b
Step 2. Solve for in terms of x and y. The Lagrange equations give the following two expressions for :
2ya 2 2xb2
= , =
x y
Equating the two equations we get
2ya 2 2xb2
=
x y
Step 3. Solve for x and y using the constraint. We solve the equation in step 2 for y in terms of x:
2ya 2 2xb2
=
x y
2y 2 a 2 = 2x 2 b2
x 2 b2 b
y2 = y= x
a2 a
2 2
We now substitute y = ab x in the equation of the constraint x 2 + y2 = 1 and solve for x:
a b
2
b
x2 ax
+ =1
a2 b2
x2 x2
+ =1
a2 a2
2x 2
=1
a2
a2 a
x2 = x=
2 2
b a b
y= =
a 2 2
We obtain the critical point a , b . Extreme values can also occur at points on the constraint where g = 2x2 , 2y2 =
2 2 a b
0, 0. However, the point (0, 0) is not on the constraint. We conclude
that if f (x, y) = 4x y has a maximum value on the
2 2
ellipse x 2 + y2 = 1 with x > 0, y > 0, then it occurs at the point a , b and the maximum value is
a b 2 2
a b a b
f , = 4 = 2ab
2 2 2 2
We now justify why the maximum value exists. We consider the problem of finding the extreme values of f (x, y) = 4x y
2 2
on the quarter ellipse x 2 + y2 = 1 in the first quadrant. Since the constraint curve is bounded and f (x, y) is continuous,
a b
f has a minimum and maximum values on the ellipse. The minimum volume occurs at the end points:
x = 0, y = b 4x y = 0 or x = a, y=0 4x y = 0
So the critical point a , b must be a maximum.
2 2
780 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
20. Show that the point (x 0 , y0 ) closest to the origin on the line ax + by = c has coordinates
ac bc
x0 = 2 , y0 = 2
a + b2 a + b2
SOLUTION We need to minimize the distance d(x, y) = x 2 + y 2 subject to the constraint g(x, y) = ax + by = c.
Notice that the distance d(x, y) is at a minimum at the same points where the square of the distance d 2 (x, y) is at a
minimum (since the function u 2 is increasing for u 0). Therefore, we may find the minimum of f (x, y) = x 2 + y 2
subject to the constraint ax + by = c.
Step 1. Write out the Lagrange Equations. The gradient vectors are f = 2x, 2y and g = a, b, hence the Lagrange
Condition f = g is
or
2x = a
2y = b
We now substitute y = bx
a in the equation of the constraint ax + by = c and solve for x:
b
ax + b x =c
a
b2
a+ x =c
a
a 2 + b2 ac
x =c x= 2
a a + b2
b ac bc
y= = 2
a a 2 + b2 a + b2
The critical point is thus
ac bc
x0 = 2 , y0 = 2 (1)
a + b2 a + b2
Step 4. Conclusions. It is clear geometrically that the problem has a minimum value and it does not have a maximum
value. Therefore the minimum occurs at the critical point. We conclude that the point closest to the origin on the line
ax + by = c is given by (1). To show that the vector x 0 , y0 is perpendicular to the line, we write the line in vector form
as x x 0 , y y0 a, b = 0. Thus, a, b is perpendicular to the line. Since x 0 , y0 = 2 c 2 a, b, then x 0 , y0 is
a +b
parallel to a, b, and thus also perpendicular to the line.
21. Find the maximum value of f (x, y) = x a y b for x, y 0 on the unit circle, where a, b > 0 are constants.
SOLUTION We must find the maximum value of f (x, y) = x a y b (a, b > 0) subject to the constraint g(x, y) =
x 2 + y 2 = 1.
Step 1. Write out the Lagrange Equations. We have f = ax a1 y b , bx a y b1 and g = 2x, 2y. Therefore the
Lagrange Condition f = g is
ax a1 y b , bx a y b1 = 2x, 2y
or
ax a1 y b = 2 x
(1)
bx a y b1 = 2 y
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 781
Step 2. Solve for in terms of x and y. If x = 0 or y = 0, f has the minimum value 0. We thus may assume that x > 0
and y > 0. The equations (1) imply that
ax a2 y b bx a y b2
= , =
2 2
Step 3. Solve for x and y using the constraint. Equating the two expressions for and solving for y in terms of x gives
ax a2 y b bx a y b2
=
2 2
ax a2 y b = bx a y b2
ay 2 = bx 2
b 2 b
y2 = x y= x
a a
We now substitute y = b 2 2
a x in the constraint x + y = 1 and solve for x > 0. We obtain
b 2
x2 + x =1
a
(a + b)x 2 = a
a a
x2 = x=
a+b a+b
We find y using the relation y = b
a x:
b a ab b
y= = =
a a+b a(a + b) a+b
Extreme points can also occur where g = 0, that is, 2x, 2y = 0, 0 or (x, y) = (0, 0). However, this point is not on
the constraint.
Step 4. Conclusions. We compute f (x, y) = x a y b at the critical point:
a/2 b/2
a b a b a a/2 bb/2 a a bb
f , = = (a+b)/2
=
a+b a+b a+b a+b (a + b) (a + b)a+b
The function f (x, y) = x a y b is continuous on the set x 2 + y 2 = 1, x 0, y 0, which is a closed and bounded set
in R 2 , hence f has minimum and maximum values on the set. The minimum value is 0 (obtained at (0, 1) and (1, 0)),
hence the critical point that we found corresponds to the maximum value. We conclude that the maximum value of x a y b
on x 2 + y 2 = 1, x > 0, y > 0 is
a a bb
.
(a + b)a+b
22. Find the maximum value of f (x, y) = x a y b for x, y 0 on the line x + y = 1, where a, b > 0 are constants.
SOLUTION
a b
Step 1. Write the Lagrange Equations. We must find the maximum value of f (x, y) = x y under the constraints
g(x, y) = x + y 1, x > 0, y > 0. The gradient vectors are f = ax a1 b a
y , bx y b1 and g = 1, 1, hence the
Lagrange Condition f = g is
ax a1 y b , bx a y b1 = 1, 1
ax a1 y b =
ax a1 y b = bx a y b1
bx a y b1 =
782 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
Step 2. Solve for x and y using the constraint. We solve the equation in step 1 for y in terms of x. This gives
ax a1 y b = bx a y b1
b
ay = bx y= x
a
b
x+ x =1
a
a
(a + b)x = a x=
a+b
b a b
y= =
a a+b a+b
The critical point is thus
a b
, (1)
a+b a+b
Now, since f is continuous on the segment x + y = 1, x 0, y 0, which is a closed and bounded set in R 2 , then f
has minimum and maximum values on this segment. The minimum value is 0 (obtained at (0, 1) and (1, 0)), therefore
the critical point (1) corresponds to the maximum value. We conclude that the maximum value of x a y b on x + y = 1,
x > 0, y > 0 is
a a bb
(a + b)a+b
23. Find the maximum value of f (x, y, z) = x a y b z c for x, y, z 0 on the unit sphere, where a, b, c > 0 are constants.
SOLUTION We must find the maximum value of f (x, y, z) = x a y b z c subject to the constraint g(x, y, z) = x 2 + y 2 +
z 2 1 = 0, x 0, y 0, z 0.
a
Step 1. Write the Lagrange Equations. The gradient vectors are f = ax a1 y b z c , by b1 x z c , cz c1 x a y b and g =
2x, 2y, 2z, hence the Lagrange Condition f = g gives the following equations:
ax a1 y b z c = (2x)
by b1 x a z c = (2y) (1)
cz c1 x a y b = (2z)
Step 2. Solve for in terms of x, y, and z. If x = 0, y = 0, or z = 0, f attains the minimum value 0, therefore we may
assume that x = 0, y = 0, and z = 0. The Lagrange equations (1) give
ax a2 y b z c by b2 x a z c cz c2 x a y b
= , = , =
2 2 2
Step 3. Solve for x, y, and z using the constraint. Equating the expressions for , we obtain the following equations:
ax a2 y b z c = by b2 x a z c
(2)
ax a2 y b z c = cz c2 x a y b
We solve for x and y in terms of z. We first divide the first equation by the second equation to obtain
by b2 x a z c b z2
1= =
cz c2 x a y b c y2
(3)
b b
y2 = z2 y= z
c c
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 783
ax a2 y b z c a z2
1= =
cz c2 x a y b c x2
(4)
a a
x 2 = z2 x= z
c c
We now substitute x and y from (3) and (4) in the constraint x 2 + y 2 + z 2 = 1 and solve for z. This gives
2
2
a b
z + z + z2 = 1
c c
a b
+ + 1 z2 = 1
c c
a+b+c 2 c
z =1 z=
c a+b+c
We find x and y using (4) and (3):
a c ac a
x= = =
c a+b+c c(a + b + c) a+b+c
b c bc b
y= = =
c a+b+c c(a + b + c) a+b+c
We examine the point where g = 2x, 2y, 2z = 0, 0, 0, that is, (0, 0, 0): This point does not lie on the constraint,
hence it is not a critical point.
Step 4. Conclusions. We compute f (x, y, z) = x a y b z c at the critical point:
a
b
c
a b c a a bb cc
f (P) = =
a+b+c a+b+c a+b+c (a + b + c)a+b+c
24. Show that the Lagrange equations for f (x, y, z) = x 2 y + zy 2 subject to the constraint g(x, y) = x + yz = 4 have
no solution. What can you conclude about the minimum and maximum values of f subject to g = 0?
SOLUTION We find the Lagrange equations. For f (x, y, z) = x 2 y + zy 2 and g(x, y, z) = x + yz 4, we have f =
2x y, x 2 + 2yz, y 2 and g = 1, z, y, hence the Lagrange Condition f = g gives the following equations:
2x y = (1)
x 2 + 2yz = z (2)
y2 = y (3)
We try to solve the Lagrange equations. Equations (1) and (3) imply that
y 2 = 2x y y
784 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
1
(2x 1)y 2 = 0 y=0 or x=
2
If y = 0, then by (1) also = 0, hence by (2) we get x 2 + 2 0 z = 0 z, that is, x = 0. Substituting x = 0, y = 0
in the equation of the constraint gives 0 + 0 z 4 = 4 = 0. Therefore y = 0 cannot hold. If x = 12 , then (1) gives
y = . Substituting in (2) we get 14 + 2 z = z or z = 14 . Again, substituting x = 12 , y = , and z = 41 in the
equation of the constraint shows that this equation is not satisfied:
1 1 1 1 3
+ 4 = 4 = 3 = 0
2 4 2 4 4
We conclude that the Lagrange equations have no solution, therefore f does not have local minimum or maximum, nor
global minimum or maximum subject to the constraint x + yz 4 = 0.
25. Let f (x, y, z) = y + z x 2 .
(a) Find the solutions to the Lagrange equations for f subject to the constraint g(x, y, z) = x 2 y 2 + z 3 = 0. Hint:
Show that at a critical point, , y, z must be nonzero and x = 0.
(b) Show that f has no minimum or maximum subject to the constraint. Hint: Consider the values of f at the points
(0, y 3 , y 2 ), which satisfy the constraint.
(c) Does (b) contradict Theorem 1?
SOLUTION
(a) The gradient vectors are f = 2x, 1, 1 and g = 2x, 2y, 3z 2 , hence the Lagrange Condition f = g
yields the following equations:
2x = (2x) 2x = 2 x (1)
1 = (2y) 1 = 2 y (2)
1 = 3z 2 1 = 3 z 2 (3)
We solve the Lagrange equations. Equations (2) and (3) imply that y = 0, z = 0, = 0, and that
1 1
= and =
2y 3z 2
Equating and solving for y in terms of z gives
1 1 3
= 2 y = z2
2y 3z 2
Equation (1) gives
(2 + 2)x = 0 = 1 or x =0
By equation (3), must be positive, hence the solution = 1 is not appropriate. We check the solution x = 0.
Substituting x = 0 and y = 32 z 2 in the equation of the constraint x 2 y 2 + z 3 = 0 gives
2
3
02 z 2 + z 3 = 0
2
9
z4 + z3 = 0
4
Since z = 0, we can divide by z 3 to obtain
9 4
z+1=0 z=
4 9
2 3
(b) The points (0, y 3 , y 2 ) satisfy the constraint since 02 (y 3 ) + (y 2 ) = y 6 + y 6 = 0. We examine the values of
f (x, y, z) = y + z x 2 at these points:
f (0, y 3 , y 2 ) = y 3 + y 2 02 = y 3 + y 2
Since lim (y 3 + y 2 ) = and lim (y 3 + y 2 ) = , f has no absolute minimum or absolute maximum subject to
y y
the constraint.
(c) Theorem 1 states that the solutions of the Lagrange equations are the only points where local extrema may occur
subject to the constraint. As shown above, our function has no global extrema, but it does have local extrema. Its not
particularly hard to show that (0, 0, 0) is a local minimum and that (0, 8/27, 4/9) is a local minimum. There is no
contradiction.
26. Use Lagrange multipliers to find the point P = (x 0 , y0 , z 0 ) on the plane ax + by + cz = d closest to the origin.
Then calculate the distance from P to O.
SOLUTION We want to minimize the distance P = x 2 + y 2 + z 2 subject to ax + by + cz = d. Since the square
2 2
function u is increasing for u 0, the square P attains its minimum at the same point where the distance P attains
its minimum. Thus, we may minimize the function f (x, y, z) = x 2 + y 2 + z 2 subject to the constraint g(x, y, z) =
ax + by + cz = d.
Step 1. Write out the Lagrange Equations. We have f = 2x, 2y, 2z and g = a, b, c, hence the Lagrange Condi-
tion f = g gives the following equations:
2x = a
2y = b
2z = c
Assume for now that a = 0, b = 0, c = 0.
Step 2. Solve for in terms of x, y, and z. The Lagrange Equations imply that
2x 2y 2z
= , = , =
a b c
Step 3. Solve for x, y, and z using the constraint. Equating the expressions for give the following equations:
2x 2z a
= x= z
a c c
(1)
2y 2z b
= y= z
b c c
We now substitute x = ac z and y = bc z in the equation of the constraint ax + by + cz = d and solve for z. This gives
a
b
a z +b z + cz = d
c c
a2 b2
z+ z + cz = d
c c
a 2 + b2 + c2 z = dc
27. Let Q be the point on an ellipse closest to a given point P outside the ellipse. It was known to the Greek
mathematician Apollonius (third century BCE) that P Q is perpendicular to the tangent at Q (Figure 13). Explain in words
why this conclusion is a consequence of the method of Lagrange multipliers. Hint: The circles centered at P are level
curves of the function to be minimized.
FIGURE 13
SOLUTION Let P = (x 0 , y0 ). The distance d between the point P and a point Q = (x, y) on the ellipse is minimum
where the square d 2 is minimum (since the square function u 2 is increasing for u 0). Therefore, we want to minimize
the function
f (x, y, z) = (x x 0 )2 + (y y0 )2 + (z z 0 )2
x2 y2
g(x, y) = 2 + 2 = 1
a b
The method of Lagrange indicates that the solution Q is the point on the ellipse where f = g, that is, the point
on the ellipse where the gradients f and g are parallel. Since the gradient is orthogonal to the level curves of the
function, g is orthogonal to the ellipse g(x, y) = 1, and f is orthogonal to the level curve of f passing through Q.
But this level curve is a circle through Q centered at P, hence the parallel vectors g and f are orthogonal to the
ellipse and to the circle centered at P respectively. We conclude that the point Q is the point at which the tangent to the
ellipse is also the tangent to the circle through Q centered at P. That is, the tangent to the ellipse at Q is perpendicular
to the radius P Q of the circle.
28. Antonio has $5.00 to spend on a lunch consisting of hamburgers ($1.50 each) and French fries ($1.00 per order).
Antonios satisfaction from eating x 1 hamburgers and x 2 orders of French fries is measured by a function U (x 1 , x 2 ) =
x 1 x 2 . How much of each type of food should he purchase to maximize his satisfaction (assume that fractional amounts
of each food can be purchased)?
SOLUTION Antonio has $5.00 to spend on the lunch, hence the total cost 1.5x 1 + x 2 must satisfy
1.5x 1 + x 2 = 5
We thus want to maximize the function U (x 1 , x 2 ) = x 1 x 2 subject to the constraint g(x, y) = 1.5x 1 + x 2 = 5 with
x 1 > 0, x 2 > 0.
x2 x1
Step 1. Write out the Lagrange Equations. The gradient vectors are U = 12 x , x and g = 1.5, 1, hence the
1 2
Lagrange Condition U = g gives the following equations:
1 x2 x2
= 1.5 = 9 2
2 x1 x1
1 x1 x1
= = 4 2
2 x2 x2
Step 2. Solve for x 1 and x 2 using the constraint. The two equations in step 1 give
x2 x
2 = = 1
9x 1 4x 2
Therefore,
4x 22 = 9x 12
9 2 3
x 22 = x x2 = x1
4 1 2
We now substitute x 2 = 32 x 1 in the constraint 1.5x 1 + x 2 = 5 and solve for x 1 . We get
3
1.5x 1 + x1 = 5
2
5
3x 1 = 5 x1 =
3
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 787
3 5 5
x2 = =
2 3 2
We obtain the critical point:
5 5
,
3 2
Step 3. Conclusions. We conclude that Antonio should have 53 hamburgers and 52 orders of fries, to maximize his sat-
isfaction. Notice that U (x 1 , x 2 ) = x 1 x 2 is continuous on the set 1.5x 1 + x 2 = 5, x 1 0, x 2 0, which is closed
and bounded in R 2 (it is a triangle in the first quadrant). f has minimum and maximum values on this set. The minimum
value 0 is obtained for x 1 = 0 or x 2 = 0, hence the critical point that we found corresponds to the maximum value.
29. Find the maximum value of f (x, y, z) = x y + x z + yz 4x yz subject to the constraints x + y + z = 1 and
x, y, z 0.
SOLUTION
Step 1. Write out the Lagrange Equations. We have f = y + z 4yz, x + z 4x z, x + y 4x y and g =
1, 1, 1, hence the Lagrange Condition f = g yields the following equations:
y + z 4yz =
x + z 4x z =
x + y 4x y =
Step 2. Solve for x, y, and z using the constraint. The Lagrange equations imply that
x + z 4x z = y + z 4yz x 4x z = y 4yz
(1)
x + y 4x y = y + z 4yz x 4x y = z 4yz
x y + 4yz 4x z = 0
x y 4z(x y) = 0 (2)
1
(x y)(1 4z) = 0 x=y or z=
4
The second equation in (1) gives:
x z + 4yz 4x y = 0
x z 4y(x z) = 0 (3)
1
(x z)(1 4y) = 0 x=z or y=
4
We examine the possible solutions.
1. x = y, x = z. Substituting x = y = z in the equation of the constraint x + y + z = 1 gives 3z = 1 or z = 13 . We
obtain the solution
1 1 1
, ,
3 3 3
1 1 1
+ +z =1 z=
4 4 2
We obtain the solution
1 1 1
, ,
4 4 2
1 1 1
+y+ =1 y=
4 4 2
788 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
x y z
30. A plane with equation + + = 1 (a, b, c > 0) together with the positive coordinate planes forms a tetrahedron
a b c
of volume V = 16 abc (Figure 14). Find the plane that minimizes V if the plane is constrained to pass through the point
P = (1, 1, 1).
C = (0, 0, c)
B = (0, b, 0)
y
A = (a, 0, 0)
x
FIGURE 14
SOLUTION The plane is constrained to pass through the point P = (1, 1, 1), hence this point must satisfy the equation
of the plane. That is,
1 1 1
+ + =1
a b c
We thus must minimize the function V (a, b, c) = 16 abc subject to the constraint g(a, b, c) = a1 + b1 + 1c = 1, a > 0,
b > 0, c > 0.
Step 1. Write out the Lagrange Equations. We have V = 16 bc, 16 ac, 16 ab and g = 12 , 12 , 12 , hence the
a b c
Lagrange Condition V = g yields the following equations:
1 1
bc = 2
6 a
1 1
ac = 2
6 b
1 1
ab = 2
6 c
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 789
Step 2. Solve for in terms of a, b, and c. The Lagrange equations imply that
Since a, b, c are positive numbers, we conclude that a = b and c = b. We now substitute a = b and c = b in the equation
of the constraint a1 + b1 + 1c = 1 and solve for b. This gives
1 1 1
+ + =1
b b b
3
=1 b=3
b
Therefore also a = b = 3 and c = b = 3. We obtain the critical point (3, 3, 3).
Step 4. Conclusions. If V has a minimum value subject to the constraint then it occurs at the point (3, 3, 3). That is, the
plane that minimizes V is
x y z
+ + = 1 or x +y+z =3
3 3 3
Remark: Since the constraint is not bounded, we need to justify the existence of a minimum value of V = 16 abc
under the constraint a1 + b1 + 1c = 1. First notice that since a, b, c are nonnegative and the sum of their reciprocals is 1,
none of them can tend to zero. In fact, none of a, b, c can be less than 1. if a , b , or c , then
Therefore,
V . This means that we can find a cube that includes the point 13 , 13 , 13 such that, on the part of the constraint
that is outside the cube, it holds that V > V 13 , 13 , 13 = 162 1 . On the part of the constraint inside the cube, V has a
minimum value m, since it is a closed and bounded set. Clearly m is the minimum of V on the whole constraint.
31. With the same set-up as in the previous problem, find the plane that minimizes V if the plane is constrained to pass
through a point P = ( , , ) with , , > 0.
SOLUTION The plane ax + by + cz = 1 must pass through the point P( , , ), hence
+ + =1
a b c
We thus must minimize the function V (a, b, c) = 16 abc subject to the constant g(a, b, c) = a + b + c = 1, a > 0,
b > 0, c > 0.
Step 1. Write out the Lagrange Equations. We have V = 16 bc, 16 ac, 16 ab and g = 2 , 2 , 2 , hence the
a b c
Lagrange Condition V = g yields the following equations:
1 a 2 bc
bc = 2 =
6 a 6
1 b2 ac
ac = 2 =
6 b 6
1 c2 ab
ab = 2 =
6 c 6
Step 2. Solve for a, b, c using the constraint. The Lagrange equations imply the following equations:
a 2 bc c2 ab
=
b2 ac c2 ab
=
We simplify the two equations to obtain
abc( a c) = 0
abc( b c) = 0
790 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
c + + c = 1
c
+ + =1
c c c
3
=1 c = 3
c
We find a and b using (1):
a= 3 = 3 , b= 3 = 3
We obtain the solution
P = (3 , 3 , 3 )
Step 3. Conclusions. Since V has a minimum value subject to the constraint, it occurs at the critical point. We substitute
a = 3 , b = 3 , and c = 3 in the equation of the plane ax + by + cz = 1 to obtain the following plane, which minimizes
V:
x y z x y z
+ + = 1 or + + =3
3 3 3
32. In a contest, a runner starting at A must touch a point P along a river and then run to B in the shortest time
possible (Figure 15). The runner should choose the point P minimizing the total length of the path.
(a) Define a function f (x, y) = A P + P B, where P = (x, y). Rephrase the runners problem as a constrained opti-
mization problem, assuming that the river is given by an equation g(x, y) = 0.
(b) Explain why the level curves of f (x, y) are ellipses.
(c) Use Lagrange multipliers to justify the following statement: The ellipse through the point P minimizing the length
of the path is tangent to the river.
(d) Identify the point on the river in Figure 15 for which the length is minimal.
River
A B
FIGURE 15
SOLUTION
(a) Let A and B be the points A = (a, b) and B = (c, d).
P = (x, y)
A B g(x, y) = 0
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 791
We must minimize the function f subject to the constraint g(x, y) = 0 (since the point P = (x, y) must satisfy the
equation of the river).
(b) The level curves of f (x, y) are f (x, y) = k for positive constants k. That is,
(x a)2 + (y b)2 + (x c)2 + (y d)2 = k
The level curve consists of all the points P = (x, y) such that the sum of the distances to the two fixed points A = (a, b)
and B = (c, d) is constant k > 0. Therefore the level curves are ellipses with foci at A and B.
(c) The point P that minimizes the length of the path must satisfy the Lagrange Condition f P = g P . That is, the
gradients of f and g are parallel vectors. Since the gradient at P is orthogonal to the level curve of the function passing
through P, the level curve of f through P (which is the ellipse through P) is tangent to the level curve of g through P,
that is, it is tangent to the river.
(d) The path-minimizing point P is the point such that the ellipse through P is tangent to the river. This point is shown
in the figure below.
River
P
B
g(x, y) = 0
33. Let L be the minimum length of a ladder that can reach over a fence of height h to a wall located a distance b behind
the wall.
(a) Use Lagrange multipliers to show that L = (h 2/3 + b2/3 )3/2 (Figure 16). Hint: Show that the problem amounts to
minimizing f (x, y) = (x + b)2 + (y + h)2 subject to y/b = h/x or x y = bh.
(b) Show that the value of L is also equal to the radius of the circle with center (b, h) that is tangent to the graph of
x y = bh.
y
xy = bh
y L
L x
Wall
h Ladder (b, h)
b x
Fence
FIGURE 16
SOLUTION
(a) We denote by x and y the lengths shown in the figure, and express the length l of the ladder in terms of x and y.
A
y
D
E
b
o C x B
792 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
Since the function u 2 is increasing for u 0, l and l 2 have their minimum values at the same point. Therefore, we may
minimize the function f (x, y) = l 2 (x, y), which is
We now identify the constraint on the variables x and y. (Notice that h, b are constants while x and y are free to change).
Using proportional lengths in the similar triangles AE D and DC B, we have
AE ED
=
DC CB
That is,
y b
= x y = bh
h x
We thus must minimize f (x, y) = (x + b)2 + (y + h)2 subject to the constraint g(x, y) = x y = bh, x > 0, y > 0.
Step 1. Write out the Lagrange Equations. We have f = 2(x + b), 2(y + h) and g = y, x, hence the Lagrange
Condition f = g gives the following equations:
2(x + b) = y
2(y + h) = x
Step 2. Solve for in terms of x and y. The equation of the constraint implies that y = 0 and x = 0. Therefore, the
Lagrange equations yield
2(x + b) 2(y + h)
= , =
y x
Step 3. Solve for x and y using the constraint. Equating the two expressions for gives
2(x + b) 2(y + h)
=
y x
We simplify:
x(x + b) = y(y + h)
x 2 + xb = y 2 + yh
Extreme values may also occur at the point on the constraint where g = 0. However, g = y, x = 0, 0 only at the
point (0, 0), which is not on the constraint.
Step 4. Conclusions. Notice that on the constraint y = bh bh
x or x = y , as x 0+ then y , and as x , then
y 0+. Also, as y 0+, x and as y , x 0+. In either case, f (x, y) is increasing without bound. Using
this property and the theorem on the existence of extreme values for a continuous function on a closed and bounded set
(for a certain part of the constraint), one can show that f has a minimum value on the constraint. This minimum value
occurs at the point (2). We substitute this point in (1) to obtain the following minimum length L:
2 2
1/3 1/3
L= (b2 h) + h + (bh 2 ) +b
2/3 1/3 2/3 1/3
= (b2 h) + 2h(b2 h) + h 2 + (bh 2 ) + 2b(bh 2 ) + b2
4 4 4 4
= b 3 h 2/3 + 2h 3 b2/3 + h 2 + b2/3 h 3 + 2b 3 h 2/3 + b2
4 4
= 3b 3 h 2/3 + 3h 3 b2/3 + h 2 + b2
3
2
2
3
= h 2/3 + 3 h 2/3 b2/3 + 3h 2/3 b2/3 + b2/3
(b) The Lagrange Condition states that the gradient vectors f P and g P are parallel (where P is the minimizing
point). The gradient f P is orthogonal to the level curve of f passing through P, which is a circle through P centered
at (b, h). g P is orthogonal to the level curve of g passing through P, which is the curve of the constraint x y = bh.
We conclude that the circle and the curve x y = bh, both being perpendicular to parallel vectors, are tangent at P. The
radius of the circle is the minimum value L, of f (x, y).
34. Find the minimum and maximum of f (x, y, z) = y + 2z subject to two constraints, 2x + z = 4 and x 2 + y 2 = 1.
SOLUTION The constraint equations are:
g(x, y) = 2x + z 4 = 0, h(x, y) = x 2 + y 2 1 = 0
We now write out the Lagrange Equations. We have, f = 0, 1, 2, g = 2, 0, 1, and h = 2x, 2y, 0, so the
Lagrange Condition is
f = g + h
0, 1, 2 = 2, 0, 1 + 2x, 2y, 0 = 2 + 2 x, 2 y,
From the third coordinate we get that = 2, which then gives us the following from the first two coordinates:
0 = 4 + 2 x
1 = 2 y
From the second equation, we see that neither nor y can be zero, so we can write = 1/2y and substitute it into the
first equation, resulting in 0 = 4 + 2(1/2y)x = 4 + x/y, or in other words, x = 4y. Plugging this into the second
constraint, we find that 16y 2 + y 2 = 1, so y = 1/ 17. Thus, our two points of interest are
4 1 8 4 1 8
, ,4 + and , ,4
17 17 17 17 17 17
The function f at the first point is 17/ 17, and at the second point is 17/ 17, so these must be our maximum and
minimum values, respectively.
35. Find the minimum and maximum of f (x, y, z) = x 2 + y 2 + z 2 subject to two constraints, x + 2y + z = 3 and
x y = 4.
SOLUTION The constraint equations are
g(x, y, z) = x + 2y + z 3 = 0, h(x, y) = x y 4 = 0
Step 1. Write out the Lagrange Equations. We have f = 2x, 2y, 2z, g = 1, 2, 1, and h = 1, 1, 0, hence
the Lagrange Condition is
f = g + h
794 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
2x = +
2y = 2
2z =
Step 2. Solve for and . The first equation gives = 2x . Combining with the third equation we get
2z = 2x (1)
The second equation gives = 2 2y, combining with the third equation we get = 4z 2y. Substituting in (1) we
obtain
2z = 2x (4z 2y) = 2x 4z + 2y
x+y
6z = 2x + 2y z= (2)
3
Step 3. Solve for x, y, and z using the constraints. The constraints give x and y as functions of z:
xy=4 y = x 4
3x z
x + 2y + z = 3 y=
2
Combining the two equations we get
3x z
x 4=
2
2x 8 = 3 x z
11 z
3x = 11 z x=
3
We find y using y = x 4:
11 z 1 z
y= 4=
3 3
We substitute x and y in (2) and solve for z:
11z + 1z
3 3 11 z 1 z 10 2z
z= = =
3 9 9
9z = 10 2z
10
11z = 10 z=
11
We find x and y:
1 z 1 10
11 = 21 = 7
y= =
3 3 33 11
11 z 11 10
11 = 111 = 37
x= =
3 3 33 11
We obtain the solution
37 7 10
P= , ,
11 11 11
Step 4. Calculate the critical values. We compute f (x, y, z) = z 2 + y 2 + z 2 at the critical point:
2
37 2 7 2 10 1518 138
f (P) = + + = = 12.545
11 11 11 121 11
As x tends to infinity, so also does f (x, y, z) tend to . Therefore f has no maximum value and the given critical point
P must produce a minimum. We conclude that the minimum value of f subject to the two constraints is f (P) = 138 11
12.545.
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 795
f (x, y) = Ax + By + C, g(x, y) = Dx + E y + F
If f has a maximum value at a point P subject to g, then at this point f P g P . Since the gradient is normal to the
level curve of the function passing through P, the tangents to the level curves of f and g at P coincide. In our case, the
level curves of f (and of g) consist of parallel lines, hence since their tangents coincide, then these parallel contour lines
coincide. That is, the contour line f (x, y) = K is also the contour line g(x, y) = L for some K , L, or in other words,
Ax + By + C = K , Dx + E y + F = L
Therefore,
D = a A, E = a B, F L = a(C K )
g(x, y) = Dx + E y + F = a Ax + a By + aC a K + L
= a( Ax + By + C) + L a K = a f (x, y) + L a K
By Theorem 1 we conclude that if g is not in the form (1), f does not have a maximum subject to g(x, y) = 0.
37. Assumptions Matter Consider the problem of minimizing f (x, y) = x subject to g(x, y) = (x 1)3
2
y = 0.
(a) Show, without using calculus, that the minimum occurs at P = (1, 0).
(b) Show that the Lagrange condition f P = g P is not satisfied for any value of .
(c) Does this contradict Theorem 1?
SOLUTION
(a) The equation of the constraint can be rewritten as
(x 1)3 = y 2 or x = y 2/3 + 1
Therefore, at the points under the constraint, x 1, hence f (x, y) 1. Also at the point P = (1, 0) we have f (1, 0) = 1,
hence f (1, 0) = 1 is the minimum value of f under the constraint.
(b) We have f = 1, 0 and g = 3(x 1)2 , 2y , hence the Lagrange Condition f = g yields the following
equations:
1 = 3(x 1)2
0 = 2 y
The first equation implies that = 0 and x 1 = 1 . The second equation gives y = 0. Substituting in the equation
3
of the constraint gives
1 3 1
(x 1)3 y 2 = 02 = = 0
3 (3 )3/2
We conclude that the Lagrange Condition is not satisfied by any point under the constraint.
(c) Theorem 1 is not violated since at the point P = (1, 0), g = 0, whereas the Theorem is valid for points where
g P = 0.
38. Marginal Utility Goods 1 and 2 are available at prices (in dollars) of p1 per unit of good 1 and p2 per unit of
good 2. A utility function U (x 1 , x 2 ) is a function representing the utility or benefit of consuming x j units of good j.
The marginal utility of the jth good is U/ x j , the rate of increase in utility per unit increase in the jth good. Prove
the following law of economics: Given a budget of L dollars, utility is maximized at the consumption level (a, b) where
the ratio of marginal utility is equal to the ratio of prices:
Marginal utility of good 1 Ux1 (a, b) p
= = 1
Marginal utility of good 2 Ux2 (a, b) p2
796 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
That is, U (x 1 , x 2 ) is maximized at the consumption level (a, b), where the following holds:
marginal utility of good 1 Ux1 (a, b) p
= = 1
marginal utility of good 2 Ux2 (a, b) p2
Notice that the constraint is a segment in the x 1 x 2 -plane (if p1 > 0 and p2 > 0), which is a closed and bounded set in
this plane. Hence, if U is continuous, it assumes extreme values on this segment.
x2
x1
p1 x1 + p2 x2 = L
x1 , x2 0
p1 c c
x2 = =
p2 2 p1 2 p2
x 2 , x 1 = p1 , p2 (1)
or
x 2 = p1 x x
= 2 = 1
x 1 = p2 p1 p2
In part (a) we showed that at the maximizing point x 1 = 2 cp , therefore the value of is
1
x c
= 1 =
p2 2 p1 p2
(c) We compute dU
dc using the Chain Rule:
dU U U
= x 1 (c) + x 2 (c) = x 2 x 1 (c) + x 1 x 2 (c) = x 2 , x 1 x 1 (c), x 2 (c)
dc x1 x2
Substituting in (1) we get
dU
= p1 , p2 x 1 (c), x 2 (c) = p1 x 1 (c) + p2 x 2 (c) (2)
dc
We now use the Chain Rule to differentiate the equation of the constraint p1 x 1 + p2 x 2 = c with respect to c:
p1 x 1 (c) + p2 x 2 (c) = 1
Hint: Use the Chain Rule to differentiate the equation g(x(c), y(c)) = c with respect to c.
(b) Show that
d
f (x(c), y(c)) =
dc
Hint: Use the Chain Rule and the Lagrange Condition f P = g P .
(c) Conclude that is the rate of increase in f per unit increase in the budget level c.
SOLUTION
(a) We differentiate the equation g (x(c), y(c)) = c with respect to c, using the Chain Rule. This gives
g g
x (c) + y (c) = 1
x y
We rewrite this equality using the dot product and the definition of the gradient:
g g
, x (c), y (c) = 1
x y
g (x(c), y(c)) x (c), y (c) = 1
(b) We now differentiate f (x(c), y(c)) with respect to c, using the Chain Rule. We obtain
d f f f f
f (x(c), y(c)) = x (c) + y (c) = , x (c), y (c) = f x (c), y (c)
dc x y x y
We use the Lagrange Condition f = g and the result in part (a) to write
d
f (x(c), y(c)) = g x (c), y (c) = 1 =
dc
798 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
(c) The equality obtained in part (b) implies that is the rate of change in the maximum value of f (x, y), subject to the
constraint g(x, y) = c, with respect to c.
41. Let B > 0. Show that the maximum of f (x 1 , . . . , x n ) = x 1 x 2 x n subject to the constraints x 1 + + x n = B
and x j 0 for j = 1, . . . , n occurs for x 1 = = x n = B/n. Use this to conclude that
a + + an
(a1 a2 an )1/n 1
n
for all positive numbers a1 , . . . , an .
SOLUTION We first notice that the constraints x 1 + + x n = B and x j 0 for j = 1, . . . , n define a closed and
bounded set in the nth dimensional space, hence f (continuous, as a polynomial) has extreme values on this set. The
minimum value zero occurs where one of the coordinates is zero (for example, for n = 2 the constraint x 1 + x 2 = B,
x 1 0, x 2 0 is a triangle in the first quadrant). We need to maximize the function f (x 1 , . . . , x n ) = x 1 x 2 x n
subject to the constraints g (x 1 , . . . , x n ) = x 1 + + x n B = 0, x j 0, j = 1, . . . , n.
Step 1. Write out the Lagrange Equations. The gradient vectors are
f = x 2 x 3 x n , x 1 x 3 x n , . . . , x 1 x 2 x n1
g = 1, 1, . . . , 1
x2 x3 xn =
x1 x3 xn =
x 1 x 2 x n1 =
Step 2. Solving for x 1 , x 2 , . . . , x n using the constraint. The Lagrange equations imply the following equations:
x 2 x 3 x n = x 1 x 2 x n1
x 1 x 3 x n = x 1 x 2 x n1
x 1 x 2 x 4 x n = x 1 x 2 x n1
..
.
x 1 x 2 x n2 x n = x 1 x 2 x n1
We may assume that x j = 0 for j = 1, . . . , n, since if one of the coordinates is zero, f has the minimum value zero.
We divide each equation by its right-hand side to obtain
xn
=1
x1
x1 = xn
xn
=1
x2 x2 = xn
xn x3 = xn
=1
x3 ..
.. .
.
xn x n1 = x n
=1
x n1
x + xn + + xn = B
n
n
B
nx n = B xn =
n
Hence x 1 = = x n = Bn .
Step 3. Conclusions. The maximum value of f (x 1 , . . . , x n ) = x 1 x 2 x n on the constraint x 1 + + x n = B, x j 0,
j = 1, . . . , n occurs at the point at which all coordinates are equal to Bn . The value of f at this point is
B B B B n
f , ,... , =
n n n n
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 799
It follows that for any point (x 1 , . . . , x n ) on the constraint, that is, for any point satisfying x 1 + + x n = B with x j
positive, the following holds:
n
B
f (x 1 , . . . , x n )
n
That is,
x1 + + xn n
x1 xn
n
or
x + + xn
(x 1 x n )1/n 1 .
n
42. Let B > 0. Show that the maximum of f (x 1 , . . . , x n ) = x 1 + + x n subject to x 12 + + x n2 = B 2 is n B.
Conclude that
|a1 | + + |an | n(a12 + + an2 )1/2
1 = (2x 1 )
1 = (2x 2 )
..
.
1 = (2x n )
Step 2. Solve for in terms of x 1 , . . . , x n . The Lagrange equations imply that x j = 0 for j = 1, . . . , n. Therefore we
may divide by x j to obtain
1
=
2x 1
1
=
2x 2
..
.
1
=
2x n
Step 3. Solving for x 1 , . . . , x n using the constraint. Equating the expressions for gives the following equations:
1 1
=
2x 1 2x n x1 = xn
1 1
= x2 = xn
2x 2 2x n ..
.. .
.
1 1 x n1 = x n
=
2x n1 2x n
Substituting x 1 , . . . , x n1 in terms of x n in the equation of the constraint x 12 + + x n2 = B 2 and solving for x n , gives
x 2 + x n2 + + x n2 = B 2
n
n
nx n2 = B 2
800 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
B2 B
x n2 = |x n | =
n n
We conclude that |x 1 | = |x 2 | = = |x n | = B . Since x j = x n for all j, the maximum value occurs when x n is
n
positive, and the minimum value corresponds to the negative value of x n . We conclude that the maximizing point is
B
x1 = x2 = = xn =
n
Notice that the point where g = 2x 1 , 2x 2 , . . . , 2x n = 0 is the point at the origin, and this point does not lie on the
constraint.
Step 4. Conclusions. The maximum value of f (x 1 , . . . , x n ) = x 1 + + x n under the constraint is
B B B
f ,... , = n = nB
n n n
This means that for any point under the constraint, that is, for any (x 1 , . . . , x n ) such that x 12 + + x n2 = B 2 , we have
f (x 1 , . . . , x n ) n B
That is,
2
x1 + + xn n x 1 + + x n2 (1)
Notice that if (x 1 , . . . , x n ) is under the constraint, then (|x|1 , . . . , |x|n ) is also under the constraint, and the right-hand
side in (1) has the same value at these two points. Therefore, we also have
2 1/2
|x 1 | + + |x n | n x 1 + + x n2 .
S(x 1 , x 2 , x 3 ) = x 1 ln x 1 + x 2 ln x 2 + x 3 ln x 3
x1 + x2 + x3 = N , E 1 x1 + E 2 x2 + E 3 x3 = E
g (x 1 , x 2 , x 3 ) = x 1 + x 2 + x 3 N = 0
h (x 1 , x 2 , x 3 ) = E 1 x 1 + E 2 x 2 + E 3 x 3 E = 0
1 + ln x 1 = + E 1
1 + ln x 2 = + E 2
1 + ln x 3 = + E 3
ln x 1 ln x 3 = (E 1 E 3 )
ln x 2 ln x 3 = (E 2 E 3 )
S E C T I O N 15.8 Lagrange Multipliers: Optimizing with a Constraint (ET Section 14.8) 801
or
x
ln 1 = (E 1 E 3 )
x3 x 1 = x 3 e (E 1 E 3 )
(1)
x
ln 2 = (E 2 E 3 ) x 2 = x 3 e (E 2 E 3 )
x3
Substituting x 1 and x 2 in the equation of the constraint g(x 1 , x 2 , x 3 ) = 0 and solving for x 3 gives
x 3 e (E 1 E 3 ) + x 3 e (E 2 E 3 ) + x 3 = N
We multiply by e E 3 :
x 3 e E 1 + e E 2 + e E 3 = N e E 3
N e E 3
x3 = E
e 1 + e E 2 + e E 3
Substituting in (1) we get
N e E 3 N e E 1
x1 = E
e (E 1 E 3 ) = E
e 1 +e 2 +e 3
E E e 1 + e E 2 + e E 3
N e E 3 (E 2 E 3 ) = N e E 2
x2 = E e
e 1 + e E 2 + e E 3 e E 1 + e E 2 + e E 3
E E E3
Letting A = e 1 +e N 2 +e , we obtain
x 1 = A1 e E 1 , x 2 = A1 e E 2 , x 3 = A1 e E 3
S = x 1 ln x 1 + + x n ln x n
x1 + + xn = N , E 1 x1 + + E n xn = E
A = N 1 (e E 1 + + e E n )
This result is used in physics to determine the distribution of velocities of gas molecules at temperature T ; xi is the
number of molecules with kinetic energy E i ; = (kT )1 , where k is Boltzmanns constant. The quantity S is called
the entropy.
SOLUTION The constraints equations are
g (x 1 , . . . , x n ) = x 1 + + x n N
h (x 1 , . . . , x n ) = E 1 x 1 + + E n x n E
1 + ln x 1 = + E 1
1 + ln x 2 = + E 2
..
.
802 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
1 + ln x n = + E n
ln xi ln x n = (E i E n ) , i = 1, . . . , n 1
or
xi
ln = (E i E n ) xi = x n e (Ei E n ) , i = 1, . . . , n 1 (1)
xn
Substituting x 1 , . . . , x n1 in the equation of the constraint x 1 + + x n1 + x n = N and solving for x n , we get
x n e (E 1 E n ) + x n e (E 2 E n ) + + x n e ( E n1 E n ) + x n = N
x n e E 1 + x n e E 2 + + x n e E n1 + x n e E n = N e E n
x n e E 1 + e E 2 + + e E n = N e E n
N e E n
xn = E
e 1 + e E 2 + + e E n
N e E n (E i E n ) = N e E i
xi = E e
e 1 + e E 2 + + e E n e E 1 + e E 2 + + e E n
E1 E2 En
Letting A = e +e N++e , we obtain
xi = A1 e ei , i = 1, . . . , n 1
The value of is determined by the second constraint h(x 1 , . . . , x n ) = 0, although it would be very difficult to calculate.
Chapter Review Exercises 803
x 2 y2 0 x 2 y2 |x| |y|
x + 3 = 0 x = 3
x
3
xy0 yx
zyx
yz 0 yz
804 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
D = {(x, y, z)|z y x}
y = 4x 2
Since the logarithm function takes on all real values, the range of f is all real values.
3. Sketch the graph f (x, y) = x 2 y + 1 and describe its vertical and horizontal traces.
SOLUTION The graph is shown in the following figure.
The trace obtained by setting x = c is the line z = c2 y + 1 or z = (c2 + 1) y in the plane x = c. The trace obtained
by setting y = c is the parabola z = x 2 c + 1 in the plane y = c. The trace obtained by setting z = c is the parabola
y = x 2 + 1 c in the plane z = c.
4. Use a graphing utility to draw the graph of the function cos(x 2 + y 2 )e1x y in the domains [1, 1]
[1, 1], [2, 2] [2, 2], and [3, 3] [3, 3], and explain its behavior.
SOLUTION The graphs of the function f (x, y) = cos(x 2 + y 2 )e1x y in the given domains are shown in the following
figures.
z z
y y
x x
The graph in the domain [1, 1] [1, 1] shows a saddle point and two local maxima. In the domain [2, 2] [2, 2]
we see two additional local minima and two maxima and in the last graph two additional maxima and two additional min-
ima appear. We can see that when |x y| 0, cos(x 2 + y 2 ) is the dominant part of the function, and as x y grows, e1x y
gains more effect. When x y , the function oscillates between and , while for x y +, f (x, y) 0.
Chapter Review Exercises 805
y
y x
x x y y
x
(a) (b) (c) (d)
FIGURE 1
SOLUTION The function f = x 2 + y matches picture (b), as can be seen by taking the x = 0 slice. The function
f = x 2 + 4y 2 matches picture (c), as can be seen by taking z = c slices (giving ellipses). Since sin(4x y)ex y is
2 2
symmetric with respect to x and y, and so also is picture (d), we match sin(4x y)ex y with (d). That leaves the third
2 2
C
650
v
400
B
A D
SOLUTION
(a) From A to C: The segment AC spans 6 level curves and the contour interval is m = 50 m, so the change of altitude
is 6 50 = 300 m. From the horizontal scale of contour map we see that the horizontal distance from A to C is 2 km or
2000 m. Therefore,
altitude 300
Average ROC from A to C = = = 0.15
horizontal distance 2000
From A to D: A and D lie on the same level curve, hence there is no change in altitude from A to D. Therefore,
0
Average ROC from A to D = = 0.
horizontal distance
(b) We first estimate the gradient at A. We get
#
f ## f 0
= =0 #
x #A x x #
# f ## 0, 0.22
f # # f 50 A
0.25
y # A y 200
We estimate v, by v 49 , 1 0.44, 1, hence the cosine of the angle between v and the gradient at A is
Hence,
(Another method is to note that in the direction of v, we cross four contour lines in about 1000 meters; thus, the change
of f in that direction is about 4 50/10000 = 0.2.)
806 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
(a) The level curves of f (x, y) = e4xy are the curves e4xy = c in the x y-plane, where c > 0. Taking ln from both
sides we get 4x y = ln c. Therefore, the level curves are the parallel lines of slope 4, 4x y = ln c, c > 0, in the
x y-plane.
(b) The level curves of f (x, y) = ln(4x y) are the curves ln(4x y) = c in the x y-plane. We rewrite it as 4x y = ec
to obtain the parallel lines of slope 4, with negative y-intercepts.
(c) The level curves of f (x, y) = 3x 2 4y 2 are the hyperbolas 3x 2 4y 2 = c in the x y plane.
(d) The level curves of f (x, y) = x + y 2 are the curves x + y 2 = c or x = c y 2 in the x y-plane. These are parabolas
whose axis is the x-axis.
8. Match each function (a)(c) with its contour graph (i)(iii) in Figure 3:
(a) f (x, y) = x y
(b) f (x, y) = e x y
(c) f (x, y) = sin(x y)
3 3 3
2 2 2
1 1 1
0 0 0
1 1 1
2 2 2
3 3 3
3 2 1 0 1 2 3 3 2 1 0 1 2 3 3 2 1 0 1 2 3
In Exercises 914, evaluate the limit or state that it does not exist.
9. lim (x y + y 2 )
(x,y)(1,3)
SOLUTION The function f (x, y) = x y + y 2 is continuous everywhere because it is a polynomial, therefore we evaluate
the limit using substitution:
lim x y + y 2 = 1 (3) + (3)2 = 6
(x,y)(1,3)
lim ln(3x 3) =
x1+
Therefore f takes on arbitrary small values at the intersection of every disk around the point (1, 3) with the domain of
the function. This shows that lim ln(3x + y) does not exist.
(x,y)(1,3)
x y + x y2
11. lim
(x,y)(0,0) x 2 + y 2
Chapter Review Exercises 807
SOLUTION We evaluate the limits as (x, y) approaches the origin along the lines y = x and y = 2x:
x y + x y2 x x + x x2 x2 + x3 1+x 1
lim = lim = lim = lim =
(x,y)(0,0) x + y
2 2 x0 x +x
2 2 x0 2x 2 x0 2 2
along y=x
x y + x y2 x 2x + x (2x)2 2x 2 + 4x 3 2 + 4x 2
lim = lim = lim = lim =
x(0,0) x + y
2 2 x0 x + (2x)
2 2 x0 5x 2 x0 5 5
along y=2x
Since the two limits are different, f (x, y) does not approach one limit as (x, y) (0, 0), therefore the limit does not
exist.
x 3 y2 + x 2 y3
12. lim
(x,y)(0,0) x 4 + y4
SOLUTION We use polar coordinates x = r cos , y = r sin . Then (x, y) (0, 0) if and only if r = x 2 + y2
0+. Therefore,
The minimum value of the function s = 2t 4 2t 2 + 1 is 12 . Therefore, since | cos | 1 and | sin | 1, we find that
# # # #
# cos2 sin2 (cos + sin ) # # cos2 sin2 (cos + sin ) #
# # # #
# ## # 2| cos + sin | 4
# 2 cos4 2 cos2 + 1 # # 1 #
2
Hence,
# #
# cos2 sin2 (cos + sin ) #
# #
0 #r # 4r
# 2 cos4 2 cos2 + 1 #
We now use the Squeeze Theorem to conclude that the limit as r 0+ is zero, hence also the given limit is zero.
13. lim (2x + y)ex+y
(x,y)(1,3)
SOLUTION The function f (x, y) = (2x + y)ex+y is continuous, hence we evaluate the limit using substitution:
(e x 1)(e y 1)
14. lim
(x,y)(0,2) x
SOLUTION We have
x
e 1 ey 1 ex 1
ex 1
lim = lim lim e y 1 = e2 1 lim (1)
(x,y)(0,2) x x0 x y2 x0 x
By LHopitals Rules,
d (e x 1)
ex 1 ex
lim = lim d x d = lim =1 (2)
d x (x)
x0 x x0 x0 1
808 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
(e x 1)(e y 1)
lim = (e2 1) 1 = e2 1.
(x,y)(0,2) x
15. Let
(x y)
p
(x, y) = (0, 0)
f (x, y) = x 4 + y 4
0 (x, y) = (0, 0)
Use polar coordinates to show that f (x, y) is continuous at all (x, y) if p > 2, but discontinuous at (0, 0) if p 2.
SOLUTION We show using the polar coordinates x = r cos , y = r sin , that the limit of f (x, y) as (x, y) (0, 0) is
zero for p > 2. This will prove that f is continuous at the origin. Since f is a rational function with nonzero denominator
for (x, y) = (0, 0), f is continuous there. We have
Since p 2 > 0, lim 2r 2( p2) = 0, hence by the Squeeze Theorem the limit in (1) is also zero. We conclude that f
r 0+
is continuous for p > 2.
We now show that for p < 2 the limit of f (x, y) as (x, y) (0, 0) does not exist. We compute the limit as (x, y)
approaches the origin along the line y = x.
p
(x 2 ) x2p x 2( p2)
lim f (x, y) = lim = lim = lim =
(x,y)(0,0) x0 x + x
4 4 x0 2x 4 x0 2
along y=x
x 2 y2
Therefore the limit of f (x, y) as (x, y) (0, 0) does not exist for p < 2. We now show that the limit lim
(x,y)(0,0) x +y
4 4
does not exist for p = 2 as well. We compute the limits along the line y = 0 and y = x:
x 2 y2 x 2 02 0
lim = lim = lim 4 = 0
(x,y)(0,0) x 4 + y 4 x0 x 4 + 04 x0 x
along y=0
x 2 y2 x2 x2 x4 1
lim = lim = lim =
(x,y)(0,0) x + y
4 4 x0 x + x
4 4 x0 2x 4 2
along y=x
Since the limits along two paths are different, f (x, y) does not approach one limit as (x, y) (0, 0). We thus showed
that if p 2, the limit lim f (x, y) does not exist, and f is not continuous at the origin for p 2.
(x,y)(0,0)
16. Calculate f x (1, 3) and f y (1, 3) for f (x, y) = 7x + y 2 .
SOLUTION To calculate f x (x, y) we treat y as a constant and use the Chain Rule. This gives
1 7
f x (x, y) = 7x + y 2 = 7x + y 2 =
x 2 7x + y 2 x 2 7x + y 2
Chapter Review Exercises 809
17. f (x, y) = 2x + y 2
SOLUTION To find f x we treat y as a constant, and to find f y we treat x as a constant. We get
2
fx = 2x + y 2 = (2x) + y =2+0=2
x x x
2
fy = 2x + y 2 = (2x) + y = 0 + 2y = 2y
y y y
18. f (x, y) = 4x y 3
SOLUTION We compute f x , treating y as a constant:
fx = (4x y 3 ) = 4y 3 (x) = 4y 3 1 = 4y 3
x x
We compute f y treating x as a constant:
fy = (4x y 3 ) = 4x (y 3 ) = 4x 3y 2 = 12x y 2 .
y y
2
fx = ln x 2 + x y 2 = 2
1
x 2 + x y2 = 2
1
2x + y 2 = 2x + y
x x + x y2 x x + x y2 x 2 + x y2
1 1 2x y
fy = ln x 2 + x y 2 = 2 x 2 + x y2 = 2 (2x y) = 2
y x + x y 2 y x + x y2 x + x y2
22. Fix c > 0. Show that for any constants , , the function u(t, x) = sin( ct + ) sin( x) satisfies the wave equation
2u 2u
= c2 2
t 2 x
SOLUTION We compute the partial derivatives u t and u x using the Chain Rule:
ut = sin( ct + ) sin( x) = sin( x) sin( ct + ) = sin( x) cos( ct + ) c
t t
ux = sin( ct + ) sin( x) = sin( ct + ) sin( x) = sin( ct + ) cos( x)
x x
We find u tt and u x x , differentiating u t and u x with respect to t and x respectively, we get
u tt = c sin( x) cos( ct + ) = 2 c2 sin( x) sin( ct + )
t
u x x = sin( ct + ) cos( x) = 2 sin( ct + ) sin( x)
x
We see that u tt = c2 u x x .
23. Find an equation of the tangent to the graph of f (x, y) = x y 2 x y + 3x 3 y at P = (1, 3).
SOLUTION The tangent plane has the equation
f x (x, y) = y 2 y + 9x 2 y f x (1, 3) = 32 3 + 9 12 3 = 33
f y (x, y) = 2x y x + 3x 3 f y (1, 3) = 2 1 3 1 + 3 13 = 8
Also, f (1, 3) = 1 32 1 3 + 3 13 3 = 15. Substituting these values in (1), we obtain the following equation:
z = 15 + 33(x 1) + 8(y 3)
or
z = 33x + 8y 42
24. Suppose that f (4, 4) = 3 and f x (4, 4) = f y (4, 4) = 1. Use the linear approximation to estimate f (4.1, 4) and
f (3.88, 4.03).
SOLUTION The linear approximation is
We use the linear approximation at the point (4, 4). Therefore, estimating f (3.88, 4.03),
h = 3.88 4 = 0.12
k = 4.03 4 = 0.03
f (3.88, 4.03) f (4, 4) + f x (4, 4) (0.12) + f y (4, 4) 0.03
f (3.88, 4.03) 3 1 (0.12) 1 0.03 = 3.09
h = 4.1 4 = 0.1
k =44=0
f (4.1, 4) f (4, 4) + f x (4, 4)(0.1) + f y (4, 4) 0
f (4.1, 4) 3 1 (0.1) 1 0 = 2.9
We use the linear approximation at the point (7, 5, 70), hence h = 7.1 7 = 0.1, k = 4.9 5 = 0.1, and l =
70.1 70 = 0.1. We get
f (7.1, 4.9, 70.1) f (7, 5, 70) + 0.1 f x (7, 5, 70) 0.1 f y (7, 5, 70) + 0.1 f z (7, 5, 70) (1)
Since z = 2x y 3 is the equation of the tangent plane at (2, 4), the following holds:
f x (2, 4) = 2
f y (2, 4) = 1 (1)
The tangency point (2, 4, f (2, 4)) lies on the tangent plane, hence
f (2, 4) = 2 2 4 3 = 3
(b) We now find linear approximation to f (2.2, 3.9) using the linear approximation at (a, b) = (2, 4). Since h =
2.2 2 = 0.2 and k = 3.9 4 = 0.1, we obtain
c 1.5
27. Jason earns S(h, c) = 20h 1 + dollars per month at a used car lot, where h is the number of hours worked
100
and c is the number of cars sold. He has already worked 160 hours and sold 69 cars. Right now Jason wants to go home
but wonders how much more he might earn if he stays another 10 minutes with a customer who is considering buying a
car. Use the linear approximation to estimate how much extra money Jason will earn if he sells his 70th car during these
10 minutes.
SOLUTION We estimate the money earned in staying for 16 hour more and selling one more car, using the linear
approximation
By the given information, a = 160, b = 69, h = 16 , and c = 1. We compute the partial derivative of the function:
c 1.5
S(h, c) = 20h 1 +
100
c 1.5
Sh (h, c) = 20 1 + Sh (160, 69) = 43.94
100
c 0.5 1 c 0.5
Sc (h, c) = 20h 1.5 1 + = 0.3h 1 + Sc (160, 69) = 62.4
100 100 100
Substituting the values in (1), we get the following approximation:
1 1
S = Sh (160, 69) + Sc (160, 69) 1 = 43.94 + 62.4 $69.72
6 6
We see that John will make approximately $69.72 more if he sells his 70th car during 10 min.
d
In Exercises 2831, compute f (c(t)) at the given value of t.
dt
28. f (x, y) = x + e y , c(t) = (3t 1, t 2 ) at t = 2
SOLUTION By the Chain Rule for Paths we have
d
f (c(t)) = f c (t) (1)
dt
We evaluate the gradient f and c (t):
d
f (c(t)) = f c(t) c (t) (1)
dt
We compute the gradient of f :
f f f
f = , , = z, 2y, x
x y z
d
f (c(t)) = 1 t, 2t 3 , t 1, 3t 2 , 1 = 1 t + 3t 2 2t 3 t = 6t 5 2t + 1
dt
d
f (c(t)) = f c(t) c (t) (1)
dt
Chapter Review Exercises 813
d t
c (t) = e , ln t = et , t 1 c (1) = e, 1 (3)
dt
Substituting (2) and (3) in (1) gives
#
d #
f (c(t)) ## = f c(1) c (1) = 1, 3e e3e e, 1 = e + 3e e3e = 4e e3e
dt t=1
Therefore,
#
d # 3 1 3 1 3 1
f (c(t)) ##
= fc c = , , = + =1
dt t= 3 3 3 2 2 2 2 4 4
v 2i + j 2 1
u= = = i+ j
v 22 + 12 5 5
Du f (3, 1) = f (3,1) u
Hence,
2 1 54 108 54
Du f (3, 1) = 27, 108 , = =
5 5 5 5 5
Du f (1, 1, 1) = f (1,1,1) u
Hence,
2 1 2 2 2 4
Du f (1, 1, 1) = 0, 2, 1 , , =0+ + =
3 3 3 3 3 3
2 2
34. f (x, y) = e x +y ,
2 2
P= , , v = 3, 4
2 2
SOLUTION We normalize v to obtain a vector u in the direction of v:
3, 4 3 4
u= = ,
5 5
32 + (4)2
Du f (P) = f P u (1)
Hence,
2 2
2
+ 22 2 2
f P = 2e 2
, = e 2 1, 1
2 2
1 1
u= 0, 1, 1 = 0, 1, 1
02 + 12 + 12 2
By the Theorem on Evaluating Directional Derivatives,
Dv f (P) = f P u (1)
f P = 0, 0, 1 .
36. Find the unit vector e at P = (0, 0, 1) pointing in the direction along which f (x, y, z) = x z + ex +y increases
2
most rapidly.
SOLUTION The gradient vector f P points in the direction of maximum rate of increase of f . Therefore we need to
find a unit vector in the direction of f P . We first find the gradient of f (x, y, z) = x z + ex +y :
2
f f f
= z 2xex +y , ex +y , x
2 2
f = , ,
x y z
At the point P = (0, 0, 1) we have
f P = 1, 1, 0 .
We normalize f P to obtain the unit vector e at P pointing in the direction of maximum increase of f :
fP 1 1
e= = , ,0 .
f P 2 2
37. Find an equation of the tangent plane at P = (0, 3, 1) to the surface with equation
ze x + e z+1 = x y + y 3
SOLUTION The surface is defined implicitly by the equation
F(x, y, z) = ze x + e z+1 x y y + 3
The tangent plane to the surface at the point (0, 3, 1) has the following equation:
4x (y 3) + 2(z + 1) = 0
4x y + 2z + 5 = 0
2z = 4x + y 5 z = 2x + 0.5y 2.5
38. Let n = 0 be an integer and r an arbitrary constant. Show that the tangent plane to the surface x n + y n + z n = r at
P = (a, b, c) has equation
a n1 x + bn1 y + cn1 z = r
SOLUTION The tangent plane to the surface, defined implicitly by F(x, y, z) = r at a point (a, b, c) on the surface,
has the following equation:
The given surface is defined by the function F(x, y, z) = x n + y n + z n . We find the partial derivative of F at a point
P = (a, b, c) on the surface:
Fx (x, y, z) = nx n1 Fx (a, b, c) = na n1
na n1 (x a) + nbn1 (y b) + ncn1 (z c) = 0
a n1 x a n + bn1 y bn + cn1 z cn = 0
a n1 x + bn1 y + cn1 z = a n + bn + cn (2)
816 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
The point P = (a, b, c) lies on the surface, hence it satisfies the equation of the surface. That is,
a n + bn + cn = r
a n1 x + bn1 y + cn1 z = r
f f
39. Let f (x, y) = (x y)e x . Use the Chain Rule to calculate and , where x = u v and y = u + v.
u v
SOLUTION First we calculate the Primary Derivatives:
f f
= e x (x y) + e x = e x (x y + 1), = e x
x y
y
Since ux = 1, u = 1, vx = 1, and vy = 1, the Chain Rule gives
f f x f y
= + = e x (x y + 1) 1 e x 1 = e x (x y + 1 1) = e x (x y)
u x u y u
f f x f y
= + = e x (x y + 1) (1) e x 1 = e x (y x 2)
v x v y v
We now substitute x = u v and y = u + v to express the partial derivatives in terms of u and v. We get
f
= euv (u v u v) = 2veuv
u
f
= euv (u + v u + v 2) = 2euv (v 1)
v
f f
40. Let f (x, y) = x 2 y + y 2 z. Use the Chain Rule to calculate and , where
s t
x = s + t, y = st, z = 2s t
SOLUTION We compute the Primary Derivatives:
f f f
= 2x y, = x 2 + 2yz, = y2
x y z
Since sx = 1, sy = t, z x y z
s = 2, t = 1, t = s, and t = 1, the Chain Rule gives
f f x f y f z
= + + = 2x y 1 + x 2 + 2yz t + y 2 2
s x s y s z s
= 2x y + x 2 + 2yz t + 2y 2
f f x f y f z
= + + = 2x y 1 + x 2 + 2yz s + y 2 (1)
t x t y t z t
= 2x y + x 2 + 2yz s y 2
We now substitute x = s + t, y = st, and z = 2s t to express the answer in terms of the independent variables s, t.
We get
f
= 2(s + t)st + (s + t)2 + 2st (2s t) t + 2s 2 t 2
s
= 2s 2 t + 2st 2 + s 2 + 2st + t 2 + 4s 2 t 2st 2 t + 2s 2 t 2
= 3s 2 t + 4st 2 + t 3 2st 3 + 6s 2 t 2
f
= 2(s + t)st + (s + t)2 + 2st (2s t) s s 2 t 2
t
= 2s 2 t + 2st 2 + s 2 + 2st + t 2 + 4s 2 t 2st 2 s s 2 t 2
= 4s 2 t + 3st 2 + s 3 + 4s 3 t 3s 2 t 2
f f f f f
41. Express the partial derivatives and of a function f (x, y, z) in terms of , , and , where (r, , z) are
r x y z
cylindrical coordinates.
Chapter Review Exercises 817
f f
42. Let f (x, y, z) = (x 2 + y 2 )ex z and P = (1, 0, 2). Use the result of Exercise 41 to calculate and at the
r
point P.
SOLUTION In Exercise 41 we showed that
f f f
= cos + sin
r x y
f f f
= r sin + r cos (1)
x y
We compute the partial derivatives of f (x, y, z) = x 2 + y 2 ex z .
f
= 2xex z + x 2 + y 2 (z)ex z = 2x x 2 z y 2 z ex z
x
f
= 2yex z
y
At the point P = (1, 0, 2), we have
f f
(1, 0, 2) = 0, (1, 0, 2) = 0
x y
Also, for this point, r = 1, = 0. Substituting in (1) gives
f
= 0 cos 0 + 0 sin 0 = 0
r
f
= 0 1 sin 0 + 0 1 cos 0 = 0
z
2
P = (1, 0, 2)
r=1 y
=0
x
818 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
SOLUTION We are given the function f (x, y), where x = u 3 v 3 and y = v 3 u 3 . Using the Chain Rule we have
the following derivatives:
g f x f y
= +
u x u y u
g f x f y
= + (1)
v x v y v
We compute the following partial derivatives:
x y
= 3u 2 , = 3u 2
u u
x y
= 3v 2 , = 3v 2
v v
Substituting in (1) we obtain
g f f f f
= 3u 2 + 3u 2 = 3u 2
u x y x y
g f f f f
= 3v 2 + 3v 2 = 3v 2
v x y x y
Therefore,
g g f f f f
v2 + u2 = 3u 2 v 2 3u 2 v 2 =0
u v x y x y
44. Let f (x, y) = g(u), where u = x 2 + y 2 and g(u) is a differentiable function in one variable. Prove that
2 2
f 2 f g
+ = 4u
x y u
SOLUTION We use the Chain Rule and the partial derivatives ux = 2x, u
y = 2y, to differentiate the equation
f (x, y, z) = g(u) with respect to x and to y. We get
f u
= g (u) = g (u) 2x
x x
f u
= g (u) = g (u) 2y
y y
Therefore,
2
f 2 f
2
2
+ = g (u) 2x + g (u) 2y = 4x 2 g (u)2 + 4y 2 g (u)2
x y
= 4 x 2 + y 2 g (u)2 = 4ug (u)2
z
45. Calculate , where xe z + ze y = x + y.
x
SOLUTION The function F(x, y, z) = xe z + ze y x y = 0 defines z implicitly as a function of x and y. Using
implicit differentiation, the partial derivative of z with respect to x is
z Fx
= (1)
x Fz
Chapter Review Exercises 819
Fx = e z 1
Fz = xe z + e y
(a) To find the critical points of the function f (x, y) = x 4 2x 2 + y 2 6y we set the partial derivatives equal to zero
and solve. This gives
f x (x, y) = 4x 3 4x = 4x x 2 1 = 0
x = 0, x = 1, x = 1, y = 3
f y (x, y) = 2y 6 = 2(y 3) = 0
The critical points are (0, 3), (1, 3), (1, 3). We now apply the Second Derivative Test to examine the critical points. We
compute the second-order partials:
f x x (x, y) = 12x 2 4, f yy = 2, fx y = 0
The discriminant is
D = f x x f yy f x2y = 2 12x 2 4 = 8 3x 2 1
This function has a minimum when x 2 1 = 0 and y 3 = 0, that is, x = 1 and y = 3. Therefore, the minimum
value is 10 obtained at the points (1, 3) and (1, 3).
In Exercises 4750, find the critical points of the function and analyze them using the Second Derivative Test.
47. f (x, y) = x 2 + 2y 2 4x y + 6x
SOLUTION To find the critical points of the function f (x, y) = x 2 + 2y 2 4x y + 6x, we set the partial derivatives
equal to zero and solve. This gives
f x (x, y) = 2x 4y + 6 = 0
f y (x, y) = 4y 4x = 0
2x 4x + 6 = 0
2x = 6 x=3
There is one critical point P = (3, 3). We now apply the Second Derivative Test to examine the critical point. We first
find the second-order partials:
48. f (x, y) = x 3 + 2y 3 x y
SOLUTION We set the partial derivatives of f (x, y) = x 3 + 2y 3 x y equal to zero and solve to find the critical points.
We get
f x (x, y) = 3x 2 y = 0
f y (x, y) = 6y 2 x = 0
54x 3 1 = 0 x 1 = 0, x 2 = 0.26
y1 = 0, y2 = 3 0.262 = 0.2
There are two critical points, (0, 0) and (0.26, 0.2). We next use the Second Derivative Test to examine the critical points.
We compute the second-order partials at these points:
We conclude that (0, 0) is a saddle point, whereas at (0.26, 0.2) the function has a local minimum.
49. f (x, y) = e x+y xe2y
SOLUTION We find the critical point by setting the partial derivatives of f (x, y) = e x+y xe2y equal to zero and
solve. This gives
The first equation gives e x+y = e2y and the second equation gives e x+y = 2xe2y . Equating the two expressions,
dividing by the nonzero function e2y , and solving for x, we obtain
1
e2y = 2xe2y 1 = 2x x=
2
We now substitute x = 12 in the first equation and solve for y, to obtain
1 1
e 2 +y e2y = 0
1
e 2 +y = e2y
1
+ y = 2y y=
2 2
There is one critical point, 12 , 12 . We examine the critical point using the Second Derivative Test. We compute the
second derivatives at this point:
1 1 1 1
f x x (x, y) = e x+y f x x , = e2+2 = e
2 2
1 1 1 1 1 1
f yy (x, y) = e x+y 4xe 2y f yy , = e 2 + 2 4 e2 2 = e
2 2 2
1 1 1 1 1
f x y (x, y) = e x+y 2e2y f x y , = e 2 + 2 2e2 2 = e
2 2
Therefore the discriminant at the critical point is
1 1
D , = f x x f yy f x2y = e (e) (e)2 = 2e2 < 0
2 2
Chapter Review Exercises 821
We conclude that 12 , 12 is a saddle point.
1
50. f (x, y) = sin(x + y) (x + y 2 )
2
SOLUTION We find the critical points by setting the partial derivatives of f (x, y) = sin(x + y) 0.5 x + y 2 equal
to zero and solve. We get
1
f x (x, y) = cos(x + y) =0
2
f y (x, y) = cos(x + y) y = 0
By the second equation y = cos(x + y). Substituting in the first equation gives y 12 = 0 or y = 12 . We set y = 12 in
the first equation and solve for x, to obtain
1 1
cos x + =0
2 2
1 1
cos x + =
2 2
The general solution is
1 1
x+ = + 2 k x = + 2 k
2 3 2 3
The critical points are thus
1 1 1 1
Pk = + + 2 k, , Q k = + 2 k,
2 3 2 2 3 2
We examine the critical points using the Second Derivative Test. We first compute the second-order partials at the critical
points:
3
f x x (x, y) = sin(x + y) f x x (Pk ) = sin + 2 k =
3 2
3
f x x (Q k ) = sin + 2 k =
3 2
3
f yy (x, y) = sin(x + y) 1 f yy (Pk ) = 1
2
3
f yy (Q k ) = 1
2
3
f x y (x, y) = sin(x + y) f x y (Pk ) =
2
3
f x y (Q k ) =
2
We compute the discriminant D = f x x f yy f x2y at the critical points:
2
3 3 3 3 3
D (Pk ) = 1 = > 0, f x x (Pk ) = <0
2 2 2 2 2
2
3 3 3 3
D (Q k ) = 1 = <0
2 2 2 2
We conclude that Q k = 12 3 + 2 k, 12 are saddle points, and at the points Pk = 12 + 3 + 2 k, 12 the
function has local maxima.
51. Prove that f (x, y) = (x + 2y)e x y has no critical points.
SOLUTION We find the critical points by setting the partial derivatives of f (x, y) = (x + 2y)e x y equal to zero and
solving. We get
f x (x, y) = e x y + (x + 2y)ye x y = e x y 1 + x y + 2y 2 = 0
822 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
f y (x, y) = 2e x y + (x + 2y)xe x y = e x y 2 + x 2 + 2x y = 0
We divide the two equations by the nonzero expression e x y to obtain the following equations:
1 + x y + 2y 2 = 0
2 + 2x y + x 2 = 0
The first equation implies that x y = 1 2y 2 . Substituting in the second equation gives
2 + 2 1 2y 2 + x 2 = 0
2 2 4y 2 + x 2 = 0
x 2 = 4y 2 x = 2y or x = 2y
x = 2y x = 2y
1 + 2y 2 + 2y 2 = 0 1 2y 2 + 2y 2 = 0
1 + 4y 2 = 0 1=0
y 2 = 14
In both cases there is no solution. We conclude that there are no solutions for f x = 0 and f y = 0, that is, there are no
critical points.
52. Find the global extrema of f (x, y) = x 3 x y y 2 + y on the square [0, 1] [0, 1].
SOLUTION
Step 1. Examine the critical points. We set the partial derivatives of f (x, y) = x 3 x y y 2 + y equal to zero and
solve to find the critical points in the interior of the square.
f x (x, y) = 3x 2 y = 0
f y (x, y) = x 2y + 1 = 0
The first equation gives y = 3x 2 . We substitute in the second equation and solve for x.
x 2 3x 2 + 1 = 0
6x 2 + x 1 = 0
1 1 + 24 1 5 1 1
x 1,2 = = x1 = , x2 =
12 12 2 3
The corresponding y-coordinates are determined by y = 3x 2 . That is,
2
1 2 3 1 1
y1 = 3 = , y2 = 3 =
2 4 3 3
Therefore, the critical points are
1 3 1 1
, , ,
2 4 3 3
Step 2. Find the global extrema on the boundary.
y
B = (1, 1)
C = (0, 1)
x
O A = (1, 0)
x
1
f(1, y) = 1 y2
The maximum value in the interval 0 y 1 occurs at y = 0, and the minimum value occurs at y = 1. The
corresponding points on the boundary of the square are (1, 0) and (1, 1).
The segment BC: On this segment y = 1, 0 x 1, hence f (x, 1) = x 3 x 1 + 1 = x 3 x.
y
x
1
f(x, 1) = x 3 x
Using calculus of one variable and referring to the graph of f (x, 1), we see that the maximum value occurs at x = 0
and x = 1 and the minimum value occurs at x = 1 . The corresponding points on the segment BC are
3
1
,1 , (0, 1), and (1, 1)
3
x
1 1
2
f(0, y) = y2 + y
The maximum value occurs at y = 12 and the minimum value occurs at y = 0 and y = 1. The corresponding points
on the segment OC are
1
0, , (0, 0), (0, 1)
2
Step 3. Conclusions. Since the global extrema occur either at critical points in the interior of the region or on the bound-
ary of the region, the candidates for global extrema are the following points:
1 3 1 1 1 1
, , , , (0, 0), (1, 0), (1, 1), (0, 1), 0, , ,1
2 4 3 3 2 3
f (0, 1) = 12 + 1 = 0
2
1 1 1 1
f 0, = + =
2 2 2 4
1 1 3 1 2 3
f ,1 = 1+1= = 0.38
3 3 3 9
We conclude that the maximum value of f on the square is f (1, 0) = 1 and the minimum value is f 1 , 1 = 0.38.
3
y
A y=4 B
y = x2
x
2 0 2
Step 1. Finding the critical points. We find the critical points in the interior of the domain by setting the partial derivatives
equal to zero and solving. We get
f x = 2y 1 = 0
1 1
f y = 2x 1 = 0 x= , y=
2 2
2
The critical point is 12 , 12 . (It lies in the interior of the domain since 12 < 4 and 12 > 12 ).
Step 2. Finding the global extrema on the boundary. We consider the two parts of the boundary separately.
The parabola y = x 2 , 2 x 2:
y
x
2 0 1 2
f (x, x 2 ) = 2x 3 x 2 x
On this curve, f (x, x 2 ) = 2 x x 2 x x 2 = 2x 3 x 2 x. Using calculus in one variable or the graph of the
function, we see that the minimum of f (x, x 2 ) on the interval occurs at x = 2 and the maximum at x = 2. The
corresponding points are (2, 4) and (2, 4).
The segment AB: On this segment y = 4, 2 x 2, hence f (x, 4) = 2 x 4 x 4 = 7x 4. The maximum
value occurs at x = 2 and the minimum value at x = 2. The corresponding points on the segment AB are (2, 4)
and (2, 4)
Step 3. Conclusions. Since the global extrema occur either at critical points in the interior of the domain or on the
boundary of the domain, the candidates for global extrema are the following points:
1 1
, , (2, 4), (2, 4)
2 2
We compute the values of f = 2x y x y at these points:
1 1 1 1 1 1 1
f , =2 =
2 2 2 2 2 2 2
f (2, 4) = 2 (2) 4 + 2 4 = 18
f (2, 4) = 2 2 4 2 4 = 10
We conclude that the global maximum is f (2, 4) = 10 and the global minimum is f (2, 4) = 18.
Chapter Review Exercises 825
yz, x z, x y = 2, 1, 4
or
yz = 2 , xz = , x y = 4
Step 2. Solve for in terms of x, y, and z. The Lagrange equations imply that
yz xy
= , = x z, =
2 4
Step 3. Solve for x, y, and z using the constraint. Equating the expressions for gives the following equations:
yz
= xz z(2x y) = 0
2
xy x(4z y) = 0
= xz
4
The first equation implies that z = 0 or y = 2x. The second equation implies that x = 0 or y = 4z. We examine all
possible solutions.
(1) z = 0 and x = 0: Then substituting in the constraint 2x + y + 4z = 1 gives 2 0 + y + 4 0 = 1 or y = 1. We
obtain the point (0, 1, 0).
(2) z = 0 and y = 4z: Then y = 4 0 = 0. Substituting
z = 0 and y = 0 in the constraint 2x + y + 4z = 1 gives
1 1
2x + 0 + 4 0 = 1 or x = 2 . We obtain the point 2 , 0, 0 .
(3) y = 2x and x = 0: Then y = 2 0 = 0. Substituting x = y = 0 in the constraint 2x + y + 4z = 1 gives
2 0 + 0 + 4z = 1 or z = 14 . The corresponding point is 0, 0, 14 .
(4) y = 2x, y = 4z: Then x = 2y and z = 4y . We substitute in the constraint 2x + y + 4z = 1 and solve for y:
y y
2 +y+4 =1
2 4
1
3y = 1 y=
3
Hence, x = 2y = 16 , z = 4y = 12
1 . We obtain the point 1 , 1 , 1 .
6 3 12
Step 4. Conclusions. We evaluate f (x, y, z) = x yz at the critical points:
f (0, 1, 0) = 0 1 0 = 0
1 1
f , 0, 0 = 0 0 = 0
2 2
1 1 1 1 1 1 1
f , , = =
6 3 12 6 3 12 216
1 1
f 0, 0, =00 =0
4 4
We conclude that the local maximum of f subject to the constraint is
1 1 1 1
f , , = .
6 3 12 216
Notice
that f does not have a global maximum on the plane 2x + y + 4z = 1 since, for all t, the point
t , 1 + 6t 2 , t 2 is on the plane and we have
2
lim f t 2 , 1 + 6t 2 , t 2 = lim t 4 1 + 6t 2 =
t t
55. Use Lagrange multipliers to find the minimum and maximum value of f (x, y) = 3x 2y on the circle x 2 + y 2 = 4.
SOLUTION
826 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
Step 1. Write out the Lagrange Equations. The constraint curve is g(x, y) = x 2 + y 2 4 = 0, hence g = 2x, 2y
and f = 3, 2. The Lagrange Condition f = g is thus 3, 2 = 2x, 2y. That is,
3 = 2x
2 = 2y
Note that = 0.
Step 2. Solve for x and y using the constraint. The Lagrange equations gives
3
3 = 2x x=
2
(1)
2 = 2y 1
y=
We substitute x and y in the equation of the constraint and solve for . We get
3 2 1 2
+ =4
2
9 1
+ 2 =4
4 2
1 13 13 13
=4 = or =
2 4 4 4
Substituting in (1), we obtain the points
6 4
x= , y =
13 13
6 4
x = , y=
13 13
The critical points are thus
6 4
P1 = ,
13 13
6 4
P2 = ,
13 13
Step 3. Calculate the value at the critical points. We find the value of f (x, y) = 3x 2y at the critical points:
6 4 26
f (P1 ) = 3 2 =
13 13 13
6 4 26
f (P2 ) = 3 2 =
13 13 13
y, x = 5, 1
y, x = 5 ,
y = 5 y
= , = x
x = 5
Chapter Review Exercises 827
Step 2. Solve for x and y using the constraint. Equating the two expressions for gives
y
= x y = 5x
5
We substitute y = 5x in the equation of the constraint 5x y = 4 and solve for x. This gives
5x (5x) = 4 2
x=
10x = 4 5
The y-coordinate is y = 5 25 = 2. We obtain the critical point 25 , 2 .
Step 3. Calculate the value at the critical point. The value of f (x, y) = x y at the critical point is
2 2 4
f , 2 = (2) = (1)
5 5 5
This value is the minimum value of f subject to the constraint:
y
5x y = 4
Note that since f (x, y) = x y is positive in the first and third quadrant, the minimum value of f subject to the constraints
part in the fourth quadrant is also the minimum value subject to the entire constraint. The part of the constraint in the
fourth quadrant is a closed and bounded segment, hence the minimum value of f on this segment exists, and is given in
(1).
We now find the minimum value of f (x, y) = x y subject to the constraint 5x y = 4 using the second way. On the
constraint 5x y = 4, we have y = 5x 4. We substitute in the function f (x, y) = x y and then find the minimum of
the resulting one-variable function. We get
We now find the minimum value of g(x) = 5x 2 4x in the interval < x < . We find the critical points:
2
g (x) = 10x 4 = 0 x=
5
The limits
lim g(x) = lim 5x 2 4x = and lim g(x) = lim 5x 2 4x =
x x x x
imply that g has a minimum value for < x < , and it occurs at the critical point. Therefore, the minimum
value of g occurs at x = 25 . The corresponding y-coordinate is y = 5 25 4 = 2, therefore the minimum value of
f (x, y) = x y is
2 2 4
f , 2 = (2) =
5 5 5
57. Find the minimum and maximum values of f (x, y) = x 2 y on the ellipse 4x 2 + 9y 2 = 36.
SOLUTION We must find the minimum and maximum values of f (x, y) = x 2 y subject to the constraint g(x, y) =
4x 2 + 9y 2 36 = 0.
Step 1. Write out the Lagrange Equations. The gradient vectors are f = 2x y, x 2 and g = 8x, 18y, hence the
Lagrange Condition f = g gives
2x y, x 2 = 8x, 18y = 8 x, 18 y
2x y = 8 x
x 2 = 18 y
828 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
Step 2. Solve for in terms of x and y. If x = 0, the equation of the constraint implies that y = 2. The points (0, 2)
and (0, 2) satisfy the Lagrange Equations for = 0. If x = 0, the second Lagrange Equation implies that y = 0.
Therefore the Lagrange Equations give
y
2x y = 8 x =
4
x2
x 2 = 18 y =
18y
Step 3. Solve for x and y using the constraint. We equate the two expressions for to obtain
y x2
=
4 18y
18y 2 = 4x 2
We now substitute 4x 2 = 18y 2 in the equation of the constraint 4x 2 + 9y 2 = 36 and solve for y. This gives
18y 2 + 9y 2 = 36 36 2 2
y2 = y1 = , y2 =
27y 2 = 36 27 3 3
2
We find the x-coordinates using x 2 = 9y2 :
9y 2
x2 =
2
9 4
x2 = = 6 x1 = 6, x2 = 6
2 3
We obtain the following critical points:
2
P1 = (0, 2), P2 = (0, 2), P3 =
6,
3
2 2 2
P4 = 6, , P5 = 6, , P6 = 6,
3 3 3
f (P1 ) = 02 2 = 0
f (P2 ) = 02 (2) = 0
2 12
f (P3 ) = f (P5 ) = 6 =
3 3
2 12
f (P4 ) = f (P5 ) = 6 =
3 3
Since the min and max of f occur on the ellipse, it must occur at critical points. Thus, we conclude that the maximum
12 and
and minimum of f subject to the constraint are 12 respectively.
3 3
58. Find the point in the first quadrant on the curve y = x + x 1 closest to the origin.
SOLUTION We need to minimize the distance d = x 2 + y 2 subject to the constraint g(x, y) = x + 1x y = 0.
Since the function u is increasing for u 0, the distance d is minimal where the square d 2 is minimal. Therefore, we
2
minimize the function f (x, y) = d 2 = x 2 + y 2 subject to the constraint.
Step 1. Write out the Lagrange Equations. The gradient vectors are f = 2x, 2y and g = 1 12 , 1 , hence the
x
Lagrange Condition f = g gives
1 1
2x, 2y = 1 2 , 1 = 1 2 ,
x x
The Lagrange Equations are
1
2x = 1 2
x
2y =
Chapter Review Exercises 829
Step 2. Solve for in terms of x and y. The second Lagrange equation gives = 2y, and the first equation gives
x2 1 2x 3
2x = = 2
x2 x 1
Step 3. Solve for x and y using the constraint. Equating the two expressions for , we get
2x 3 x3
2y = 2 y=
x 1 1 x2
We now substitute y as a function of x in the equation of the constraint and solve for x. This gives
x3 1 x2 + 1
= x + =
1 x2 x x
x4 = 1 x2 1 + x2 = 1 x4
2x 4 = 1 x = 21/4 , x = 21/4
23/4 21/4 1
= 21/4 21/2 + 1 = 21/4 + 21/4 = 2 +
4
y= 1/2
= 1/2
12 2 1 4
2
We obtain the critical point:
1 4 1
P=
4
, 2+
4
2 2
Step 4. Conclusion.
y
It is clear from the graph of y = x + 1x that the critical point is a minimum. Therefore, the point P is the closest to the
origin on the curve y = x + 1x in the first quadrant.
59. Find the extreme values of f (x, y, z) = x + 2y + 3z subject to the two constraints x + y + z = 1 and x 2 + y 2 +
z 2 = 1.
SOLUTION We must find the extreme values of f (x, y, z) = x + 2y + 3z subject to the constraints g(x, y, z) =
x + y + z 1 = 0 and h(x, y, z) = x 2 + y 2 + z 2 1 = 0.
Step 1. Write out the Lagrange Equations. We have f =< 1, 2, 3 >, g =< 1, 1, 1 >, h =< 2x, 2y, 2z >, hence
the Lagrange condition f = g + h gives
or
1 = + 2 x
2 = + 2 y
3 = + 2 z
1 = + 2 x =1 2 x
2 = + 2 y =2 2 y
3 = + 2 z =3 2 z
830 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
h
Chapter Review Exercises 831
f = 2 r h + 2 r 2
or
2 h + 2r, r = 2hr, r 2
2(h + 2r ) = 2hr h + 2r = hr
2r = r 2 2r = r 2
Step 2. Solve for in terms of r and h. The equation of the constraint implies that r = 0 and h = 0 (we assume that
V > 0). Therefore, the Lagrange equations give
h + 2r 1 2 2
= = + , =
hr r h r
Step 3. Solve for r and h using the constraint. Equating the two expressions for gives
1 2 2
+ =
r h r
2 1
= h = 2r
h r
r 2 2r = V
V 1/3
2 r 3 = V r=
2
We find h using the relation h = 2r :
1/3
V
h=2
2
1/3 1/3
The critical point is h = 2 2V , r = 2V .
Step 4. Conclusions. On the constraint r 2 h = V we have h = V 2 and r = V
h , hence
r
V V 2V
f r, 2 = 2 r 2 + 2 r 2 = + 2 r 2
r r r
V V V 2V
f , h = 2 h + 2 = 2 V h +
h h h h
61. Find the dimensions of the box of maximum volume with its sides parallel to the coordinate planes that can be
inscribed in the ellipsoid (Figure 4)
x 2 y 2 z 2
+ + =1
a b c
y
x
FIGURE 4
SOLUTION We denote the vertices of the box by (x, y, z), where x 0, y 0, z 0. The volume of the box is
V (x, y, z) = 8x yz
The vertices of the box must satisfy the equation of the ellipsoid, hence,
x2 y2 z2
g(x, y, z) = 2 + 2 + 2 1 = 0, x 0, y 0, z 0.
a b c
We need to maximize V due to the constraint: g(x, y, z) = 0, x 0, y 0, z 0.
Step 1. Write out the Lagrange Equations. We have V = 8 yz, x z, x y and g = 2x2 , 2y2 , 2z2 , hence the Lagrange
a b c
Condition V = g gives the following equations:
2x
yz = 2
a
2y
xz = 2
b
2z
xy = 2
c
Step 2. Solve for in terms of x, y, and z. If x = 0, y = 0, or z = 0, the volume of the box has the minimum value
zero. We thus may assume that x = 0, y = 0, and z = 0. The Lagrange equations give
a 2 yz b2 x z c2 x y
= , = , =
2x 2y 2z
Step 3. Solve for x, y, and z using the constraint. Equating the three expressions for yields the following equations:
a 2 yz c2 x y
= y c2 x 2 a 2 z 2 = 0
2 x 2 z
b2 x z c2 x y x c2 y 2 b2 z 2 = 0
=
2 y 2 z
Since x > 0 and y > 0, these equations imply that
az
c2 x 2 a 2 z 2 = 0 x=
c
(1)
c2 y 2 b2 z 2 = 0 bz
y=
c
We now substitute x and y in the equation of the constraint and solve for z. This gives
az 2 bz 2
c c z2
+ + 2 =1
a2 b 2 c
z2 z2 z2
2
+ 2 + 2 =1
c c c
3z 2 c
=1 z=
c2 3
Chapter Review Exercises 833
f (x 1 , . . . , x n ) = x 12 12 + + x n2 n2
1
' n
subject to x 1 + + x n = 1 is c, where c = 2
j
.
j=1
212 x 1 =
222 x 2 =
..
.
2n2 x n =
Step 2. Solve for x 1 , . . . , x n using the constraint. The Lagrange equations imply the following equations:
n2
212 x 1 = 2n2 x n x1 = xn
12
n2
222 x 2 = 2n2 x n x2 = xn
22 (1)
.. ..
. .
2 x 2 n2
2n1 n1 = 2n x n x n1 = xn
n1
2
We substitute these values in the equation of the constraint x 1 + + x n = 1 and solve for x n . This gives
n2 n2 n2
xn + xn + + xn + xn = 1
12 22 n1
2
1 1 1 1
n2 + + + + 2 xn = 1
12 22 n1
2 n
'
n
2
n
2 j xn = 1
j=1
834 C H A P T E R 15 D I F F E R E N T I AT I O N I N S E V E R A L VA R I A B L E S (ET CHAPTER 14)
(
n 2 1
Denoting c = j=1 j , we get
c
xn =
n2
Using (1) we get
n2 c c
x1 = =
12 n2 12
n2 c c
x2 = =
22 n2 22
..
.
n2 c c
x n1 = =
2
n1 n2 2
n1
63. A bead hangs on a string of length whose ends are fixed by thumbtacks located at points (0, 0) and (a, b) on a
bulletin board (Figure 5). The bead rests in the position that minimizes its height y. Use Lagrange multipliers to show
that the two sides of the string make equal angles with the horizontal, that is, show that 1 = 2 .
As an aside, note that the locus of the bead when pulled taut is an ellipse with foci at O and P, and the tangent line at
the lowest point is horizontal. Therefore, this exercise provides another proof of the reflective property of the ellipse (a
light ray emanating from one focus and bouncing off the ellipse is reflected to the other focus).
P = (a, b)
x
O
1 2
FIGURE 5
y
E (a, b)
2 2 b
A B C D
x
1
1 2 y
F
Chapter Review Exercises 835
Hence,
l = l1 + l2 (1)
For our second constraint, we note that tan 1 = AB/(y) and tan 2 = B D/(b y). Since AB + B D = a, we have
Note that 1 and 2 cant both be zero. If 1 = 0, then we get 0 = 2 (y sec2 2 ), which doesnt happen. A similar
argument holds for 2 . Thus, we can assume that both 1 and 2 are nonzero. Looking at the second coordinate of the
above equation, we get 1 (y sec 1 tan 1 ) = 2 (y sec2 1 ), which means 1 /2 = sec 1 / tan 1 = 1/ sin 1 . A
similar analysis of the third coordinate gives 1 /2 = sec 2 / tan 2 = 1/ sin 2 . We conclude that sin 1 = sin 2 ,
and hence 1 = 2 , as desired.