Math 21a Homework 19 Solutions Spring, 2014

1. Directional derivatives warm-up.

(a) (Stewart 11.6 #4 ) Find the directional derivative of the function f (x, y) = x2 y 3 − y 4 at the point (2, 1) in the
direction indicated by the angle θ = π/4.
Solution: When we evaluate the gradient ∇f = h2xy 3 , 3x2 y 2 − 4y 3 i at the point (2, 1), we get ∇f (2, 1) = h4, 8i.
The unit vector in the direction θ is u = hcos(θ), sin(θ)i = h √12 , √12 i. The directional derivative is then
√
 
1 1 12
Du f (2, 1) = ∇f (2, 1) · u = h4, 8i · √ , √ = √ = 6 2.
2 2 2

(b) (Stewart 11.6 #12 ) Find the directional derivative of the function f (x, y) = ln(x2 + y 2 ) at the point (2, 1) in the
direction v = h−1, 2i.
Solution: In this case, ∇f (x, y) = h x22x , 2y i. At point (2, 1), this is ∇f (2, 1) = h 54 , 25 i. The directional
+y 2 x2 +y 2
derivative is
h 4 , 2 i · h−1, 2i
 
v 4 2 h−1, 2i
Dv f (2, 1) = ∇f (2, 1) · = , · = 5 5 = 0.
|v| 5 5 |h−1, 2i| |h−1, 2i|
√
(c) (Stewart 11.6 #16 ) Find the directional derivative of the function f (x, y, z) = xyz at the point (3, 2, 6) in the
direction v = h−1, −2, 2i.
D p q p E
Solution: Here ∇f (x, y, z) = 21 · yz , 1 · xz
x 2
, 1 · xy
y 2 z
, so ∇f (3, 2, 6) = h1, 23 , 12 i. The directional derivative
in the direction of v is
 
v 3 1 h−1, −2, 2i
Dv f (3, 2, 6) = ∇f (3, 2, 6) · = 1, , · = −1.
|v| 2 2 |h−1, −2, 2i|

(d) (Stewart 11.6 #24 ) Find the maximum rate of change of f (x, y, z) = (x + y)/z at the point (1, 1, −1), and the
direction in which it occurs.
Solution: The maximum rate of change occurs in the direction of the gradient. The gradient is
 
1 1 x+y
∇f (x, y, z) = , ,− 2 , or at the given point ∇f (1, 1, −1) = h−1, −1, −2i.
z z z
The unit vector in the direction of the gradient is u = h− √16 , − √16 , − √26 i and the rate of change of f in this
direction is
√
 
1 1 2
Du f (1, 1, −1) = ∇f (1, 1, −1) · u = h−1, −1, −2i · − √ , − √ , − √ = 6.
6 6 6
Note that this is |∇f | at the point (1, 1, −1).

2. (Stewart 11.6 #32 ) Suppose you are climbing a hill whose shape is given by the equation z = 1000 − 0.005x2 − 0.01y 2 ,
where x, y and z are measured in meters, and you are standing at a point with coordinates (60, 40, 966). The positive
x-axis points east, and the positive y-axis points north.
(a) If you walk due south, will you start to ascend or descend? At what rate?
Solution: We compute the partial derivatives
∂z ∂z
= −0.01x and = −0.02y.
∂x ∂y

∂z ∂z ∂z 
If you walk due south, the directional derivative is D−j z = −j · h ∂x , ∂y
i = − ∂y  = 0.8. Since this is
(60,40,966)
positive, you will ascend at the rate 0.8 vertical meters per horizontal meter.

(b) If you walk northwest, will you start to ascend or descend? At what rate?
Solution: Northwest here means the direction u = h− √12 , √12 i. The directional derivative is
 
1 1 1   1
Du f = ∇f · − √ , √ = √ · − (−0.01) · 60 + (−0.02) · 40 = −0.2 · √ .
2 2 2 2
0.2
Since this is negative, we will descend at the rate √
2
m/sec ≈ 0.14 vertical meters per horizontal meter.
 (c) In which direction should you walk if you want to ascend as steeply as possible? What is the rate of ascent in that
direction? At what angle above the horizontal does the path in that direction begin?

Solution: The direction of steepest ascent is the unit vector in the direction of the gradient ∇z(60, 40) =
h−0.60, −0.80i. In our specific example, the gradient is already a unit vector. The rate of ascent in the direction
∇f
of the gradient is |∇f |
· ∇f = |∇f |. Since the gradient is a unit vector, this is 1.0 vertical meters per horizontal
meter. Thus the angle of ascent is 45 degrees.

3. A very important idea. In this problem, we’ll explore a very important property of the gradient of a function, ∇f .
We’ll consider the function f (x, y) = x2 /4 + y 2 , but the property you’ll discover is very general.
3
(a) The plot to the right shows the level sets z =
1, 2, 3, 4, 5, 6 for the function f (x, y) = x2 /4+y 2 .
For each of the following points P (a, b), find the 2
gradient ∇f (a, b) at P , and draw that vector on
the plot with its tail at P . (Note that each point 1
below is on one of the level curves drawn!)

(i) P (2, 0) (ii) P (0, 2) -4 -2 2 4

(iii) P (0, −1) (iv) P (−4, 0) -1

(v) P (2, 1) (vi) P (−2, 2)

-2
(vii) P (−4, −1) (viii) P (2, −2)
-3

(b) Given what you drew in part (a), what do you hypothesize is the relationship between the direction of the gradient
∇f (a, b) and the direction of the level curve of f through (a, b)?

Solution: We’ve drawn an expanded picture in Figure 1, showing the full lengths of the gradients. From this
picture, we see that the direction of the gradient is orthogonal to the direction of the level curve.

(c) Read: the second paragraph on page 798. Explain how the paragraph confirms your hypothesis from (b). Also
explain what Stewart means when he writes, “it is intuitively plausible because the values of f remain constant as
we move along the [level] curve”.

Solution: Stewart means that: Since f (x, y) is constant along these curves, then it makes sense that the largest
rate of change should come from the orthogonal component.

(d) Show that the level curves of the function f (x, y) = x2 /4 + y 2 and the level curves of the function g(x, y) = x4 /y
are orthogonal trajectories. (That is, show that for any point (a, b), the level curve of f through (a, b) and the
level curve of g through (a, b) meet orthogonally at (a, b).)

Solution: The level curves (tangent vectors) are orthogonal if and only if the gradients (normal vectors) are
orthogonal. These gradients are ∇f = hx/2, 2yi and ∇g = h4x3 /y, −x4 /y 2 i, so their dot product is
Dx E  4x3 −x4  2x4 2x4
∇f · ∇g = , 2y · , 2 = − = 0,
2 y y y y

as desired.
Figure 1: Sketch of gradients for Problem 3
4. Another method for finding tangent planes to surfaces.
(a) Read: The section Tangent Planes to Level Surfaces on page 796. The last paragraph on the page can be
summarized: If a surface S is the level surface f (x, y, z) = k of a function f , then the gradient ∇f (a, b, c) is the
normal vector of the tangent plane to S at the point (a, b, c).
(b) Using part (a), find an equation of the tangent plane to the surface x2 + xy 3 − z = 3 at the point (2, 2, 17).

Solution: The gradient of f is ∇f = h2x+y 3 , 3xy 2 , −1i. At the point (2, 2, 17), this is ∇f (2, 2, 17) = h12, 24, −1i.
Since the gradient is the normal vector to the tangent plane, this plane has the form 12x + 24y − z = C for some
C. Since the tangent plane passes through the point (2, 2, 17), we find C = 55. That is, the tangent plane at the
point (2, 2, 17) is 12x + 24y − z = 55.

(c) Your classmate, Stergios, is a little bit confused about gradients and normal vectors to surfaces. He is trying to
find the tangent plane to the graph of the function f (x, y) = x2 + xy 3 − 3 at the point (2, 2, 17). He explains to
you his reasoning:
(i) To find the equation of a plane, all I need is a point on the plane and the direction of its normal vector
(ii) I’m given the point (2, 2, 17), which is the point of tangency. So that part’s easy.
(iii) I also remember from part (a) that ∇f is the normal vector of the surface. Since f (x, y) = x2 + xy 3 ,
∇f = h2x + y 3 , 3xy 2 i and at x = 2, y = 2, z = 17 the normal vector to the graph of f is n = ∇f = h12, 24i
(iv) Now I have a point and a normal vector for the plane, the rest is easy: the equation is

n · hx − x0 , y − y0 , z − z0 i = 0 which gives the equation: h12, 24i · hx − 2, y − 2, z − 17i = 0.

Write a sentence to convince Stergios that his answer can’t possibly be right.
Hint: Something makes that “equation” nonsense. Explain what is wrong with the third step of his solution
that leads him to this nonsensical equation.

Solution: Dear Stergios, as you can very well see, the dot product h12, 24i · hx − 2, y − 2, z − 17i does not make
sense. The normal vector n you give has two components, but a vector in 3-space has three components. To find
the normal vector to this surface z = f (x, y) = x2 + xy 3 − 3 using a gradient, we need to write it as a level set. If
we write it as x2 + xy 3 − z = 3, then the surface we’re interested in is the level set of F (x, y, z) = x2 + xy 3 − z.
(This was the function we used to find the gradient in part (b).)

5. (Loosely based on Stewart 11.6 #58.) The plane y + z = 2 intersects the funky cylinder x2 + y 4 = 17 (no, this is not
standard terminology) in a curve C. You can see a picture of this funky cylinder and the plane by using:

funky = ContourPlot3D[ x^2 + y^4 == 17, {x,-5,5},{y,-5,5},{z,-5,5}]

plane = ContourPlot3D[ y+z == 2, {x,-5,5},{y,-5,5},{z,-5,5}]
Show[funky, plane]

(a) Find a parametric equation of the tangent line to C at the point (4, 1, 1).
Hint: Don’t try to find a parametrization of C. Instead, use the ideas from Problem 4.
(b) Explain carefully how you found the direction vector of the tangent line in part (a), and how you know that it’s
the right direction.

Solution: The tangent line to C is the intersection of two tangent planes at the point (x, y, z) = (4, 1, 1): the tangent
plane to the plane y + z = 2 (which is the plane itself!) and the tangent plane to x2 + y 4 = 17. Let’s find the equation
of this second tangent plane. The normal vector is the gradient ∇(x2 + y 4 ) = h2x, 4y 3 , 0i. Evaluated at point (4, 1, 1),
it is h8, 4, 0i. Thus, the tangent plane has the equation 8x + 4y = 40, where 40 was found by requiring that the plane
actually pass through the tangent point (4, 1, 1).

How can we find the line that is the intersection of these two tangent planes? One simple method is to notice that the
vector u in the direction of this line lies in both tangent planes, so u is perpendicular to the normals of both tangent
planes. So one choice for u is the cross product of these normals:
 
 i j k
 
u = h0, 1, 1i × h8, 4, 0i = 0 1 1  = h−4, 8, −8i.
8 4 0 
Thus the tangent line is through the point (4, 1, 1) and in the direction h−4, 8, −8i. One version of this line is r(t) =
h4, 1, 1i + th−4, 8, −8i.

Another approach to finding the intersection of y + z = 2 and 8x + 4y = 40 is more direct. We can express z and x
in terms of y: z = 2 − y from the first equation and x = 5 − y/2 from the second. Thus hx, y, zi = h5 − y/2, y, 2 − yi.
Setting t = y/2, we get the parametrization r(t) = h5 − t, 2t, 2 − 2ti or r(t) = h5, 0, 2i + th−1, 2, −2i. Note that this is
the same line as before, but a different parametrization.