0% found this document useful (0 votes)
66 views366 pages

Concepts in Calculus III Multivariable Calculus Solutions Manual

Uploaded by

NABIL HUSSAIN
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
66 views366 pages

Concepts in Calculus III Multivariable Calculus Solutions Manual

Uploaded by

NABIL HUSSAIN
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 366

Concepts in Calculus III

Multivariable Calculus
Solutions Manual

K. DeMason and S. V. Shabanov

Department of Mathematics, University of Florida,


Gainesville, FL 32611 USA

For the Latest Edition of the Textbook: 2019


Fist Edition of the Textbook: University of Florida Press, Gainesville,
2012
ISBN 978–1–61610–162–6

c 2017 Sergei Shabanov, All Rights Reserved


Contents

Chapter 1. Vectors and the Space Geometry 1


1. Rectangular Coordinates in Space 3
2. Vectors in Space 2
3. The Dot Product 41
4. The Cross Product 60
5. The Triple Product 78
6. Planes in Space 96
7. Lines in Space 108
8. Euclidean spaces 124
9. Quadric Surfaces 132
Chapter 2. Vector Functions 155
10. Curves in Space and Vector Functions 155
11. Differentiation of Vector Functions 171
12. Integration of Vector Functions 183
13. Arc Length of a Curve 193
14. Curvature of a Space Curve 203
15. Applications to Mechanics and Geometry 218
Chapter 3. Differentiation of Multivariable Functions 239
16. Functions of Several Variables 239
17. Limits and Continuity 251
18. A General Strategy to Study Limits 265
19. Partial Derivatives 280
20. Higher-Order Partial Derivatives 286
21. Differentiability of Multivariable Functions 297
22. Chain Rules and Implicit Differentiation 315
23. The Differentials and Taylor Polynomials 333
24. Directional Derivative and the Gradient 358
25. Maximum and Minimum Values 375
26. Extreme Values on a Set 390
27. Lagrange Multipliers 405
Chapter 4. Multiple Integrals 429
28. Double Integrals 429
29. Properties of the Double Integral 443
1
2 CONTENTS

30. Iterated Integrals 452


31. Double Integrals over General Regions 460
32. Double Integrals in Polar Coordinates 476
33. Change of Variables in Double Integrals 490
34. Triple Integrals 509
35. Triple Integrals in Cylindrical and Spherical Coordinates 524
36. Change of Variables in Triple Integrals 539
37. Improper Multiple Integrals 552
38. Line Integrals 568
39. Surface Integrals 575
40. Moments of Inertia and Center of Mass 592
Chapter 5. Vector Calculus 607
41. Line Integral of a Vector Field 607
42. Fundamental Theorem for Line Integrals 619
43. Green’s Theorem 633
44. Flux of a Vector Field 645
45. Stokes’ Theorem 657
46. Gauss-Ostrogradsky (Divergence) Theorem 669
Acknowledgments 681
CHAPTER 1

Vectors and the Space Geometry

1. Rectangular Coordinates in Space


1–2. Find the distance between the specified points.
1. (1, 2, −3) and (−1, 0, −2)

Solution: Let d denote the distance between the given points.


Then, by the distance formula,
p √ √
d = (−1 − 1)2 + (0 − 2)2 + (−2 + 3)2 = 4 + 4 + 1 = 9 = 3

2. (−1, 3, −2) and (−1, 2, −1)

Solution: Let d denote the distance between the given points.


Then, by the distance formula,
p √ √
d = (−1 + 1)2 + (2 − 3)2 + (−1 + 2)2 = 0 + 1 + 2 = 2

3–4. Determine whether the given points lie in a straight line.


3. (8, 3, −3), (−1, 6, 3), and (2, 5, 1)

Solution: Given three points A, B, and C in Rn , we can


determine if they are colinear if

|AB| + |BC| = |AC|.

Therefore, we must compute the distances between each pair


of points and see if one is the sum of the other two. Let
A = (8, 3, −3), B = (−1, 6, 3), and C = (2, 5, 1). By the
distance formula,
p √ √
|AB| = (8 + 1)2 + (3 − 6)2 + (−3 − 3)2 = 81 + 9 + 36 = 126
p √ √
|BC| = (−1 − 2)2 + (6 − 5)2 + (3 − 1)2 = 9 + 1 + 4 = 14
p √ √
|AC| = (8 − 2)2 + (3 − 5)2 + (−3 − 1)2 = 36 + 4 + 16 = 56

3
4 1. VECTORS AND THE SPACE GEOMETRY

Note that |AB| = |BC| + |AC|, so they are colinear. To see


this, consider squaring the right hand side,
√ √ √ √
( 14 + 56)2 = 14 + 2 14 56 + 56

126 = 70 + 2 7 · 2 · 7 · 8
126 = 70 + 2(7)(4) = 70 + 56 = 126
4. (−1, 4, −2), (1, 2, 2), and (−1, 2, −1)

Solution: Given three points A, B, and C in Rn , we can


determine if they are colinear if
|AB| + |BC| = |AC|.
Therefore, we must compute the distances between each pair
of points and see if one is the sum of the other two. Let
A = (−1, 4, −2), B = (1, 2, 2), and C = (−1, 2, −1). By the
distance formula,
p √ √
|AB| = (−1 − 1)2 + (4 − 2)2 + (−2 − 2)2 = 4 + 4 + 16 = 24
p √ √
|BC| = (1 + 1)2 + (2 − 2)2 + (2 + 1)2 = 4 + 0 + 9 = 13
p √ √
|AC| = (−1 + 1)2 + (4 − 2)2 + (−2 + 1)2 = 0 + 4 + 1 = 5
Note that no one distance is the sum of the other two, hence
the points are not colinear.

5. Determine whether the points (1, 2, 3) and (3, 2, 1) lie in the plane
through O = (1, 1, 1) and perpendicular to the line through O and
B = (2, 2, 2). If not, find a point in this plane.

Solution: Let P1 = (1, 2, 3) and P2 = (3, 2, 1). Let P be the plane


that passes through a given point O and is perpendicular to the line
through O and B. An arbitrary point P ∈ R3 will be in P iff OP, OB,
and BP form a right triangle. Clearly, OP and OB must be perpen-
dicular, hence we must have
|OB|2 + |OP |2 = |BP |2
First checking P1 ,
p √
|OB| = (2 − 1)2 + (2 − 1)2 + (2 − 1)2 = 3,
p √
|BP1| = (2 − 1)2 + (2 − 2)2 + (2 − 3)2 = 2,
p √
|OP1 | = (1 − 1)2 + (1 − 2)2 + (1 − 3)2 = 5.
1. RECTANGULAR COORDINATES IN SPACE 5

Since 3 + 5 6= 2, P1 ∈
/ P. A similar argument shows that P2 ∈ / P. It
turns out that, in this problem, |OP1 | = |OP2 | and |BP1 | = |BP2 |.

To find a point in P, let P = (x, y, z). Then we must have that


3 + ((1 − x)2 + (1 − y)2 + (1 − z)2 ) = ((2 − x)2 + (2 − y)2 + (2 − z)2)
(2 − x)2 − (1 − x)2 + (2 − y)2 − (1 − y)2 + (2 − z)2 − (1 − z)2 − 3 = 0
(2 − x − (1 − x))(2 − x + 1 − x) + (2 − y − (1 − y))(2 − y + 1 − y)+
(2 − z − (1 − z))(2 − z + 1 − z) − 3 = 0
3 − 2x + 3 − 2y + 3 − 2z − 3 = 0
2x + 2y + 2z = 6
x+y+z =3

Thus any point P = (x, y, z) satisfying x + y + z = 3 will be in P. For


example, (3, 0, 0) ∈ P. In the future, you will learn an easier way to
arrive at this formula.

6. Find the distance from the point (1, 2, −3) to each of the coor-
dinate planes and to each of the coordinate axes.

Solution: Let P = (1, 2, −3). To find the distance between a point


and a plane, we must find a point P ∗ on the plane such that P P ∗ is
perpendicular to the plane. For the coordinate planes, this is easy –
simply project the point onto the plane. Let dxz represent the distance
between P and the projection of P onto the xz plane, P ∗ = (1, 0, −3).
Thus, by the distance formula,
p
dxz = (1 − 1)2 + (2 − 0)2 + (3 − (−3))2 = 2
Similarly, dyz and dxy are as follows
p
dyz = (1 − 0)2 + (2 − 2)2 + (3 − (−3))2 = 1
p
dxy = (1 − 1)2 + (2 − 2)2 + (3 − 0)2 = 3
To find the distance between P and a coordinate axis, we must find
our P ∗ and project it again onto the desired axis. For example, find
the distance between P and the x-axis, we must project P onto the
xz plane (P ∗ = (1, 0, −3)), and then project it again onto the x axis
(P ∗∗ = (1, 0, 0)). Thus, dx , the distance between P and the x-axis is
p √
dx = (1 − 1)2 + (2 − 0)2 + (3 − 0)2 = 13
6 1. VECTORS AND THE SPACE GEOMETRY

Similarly, dy and dz are as follows


p √
dy = (1 − 0)2 + (2 − 2)2 + (3 − 0)2 = 10
p √
dz = (1 − 0)2 + (2 − 0)2 + (3 − 3)2 = 5

7. Find the length of the medians of the triangle with vertices A =


(1, 2, 3), B = (−3, 2, −1), and C = (−1, −4, 1).

Solution: The coordinates of the midpoint of a straight line segment


with given endpoints are the half-sums of the corresponding coordi-
nates of the endpoints. Let Pa , Pb , and Pc be the midpoints of BC,
AC, and AB, respectively. Then
Pa = (−2, −1, 0) , Pb = (0, −1, 2) , Pc = (−1, 2, 1) .
By the distance formula,
p √ √
|APa | = (−2 − 1)2 + (−1 − 2)2 + (3 − 0)2 = 9 + 9 + 9 = 3 3 ,
p √ √
|BPb | = (0 + 3)2 + (−1 − 2)2 + (2 + 1)2 = 9 + 9 + 9 = 3 3 .
p √
|CPc | = (−1 + 1)2 + (2 + 4)2 + (1 − 1)2 = 0 + 62 + 0 = 6 .
8. Let the set S consist of points (t, 2t, 3t) where −∞ < t < ∞. Find
the point of S that is the closest to the point (3, 2, 1). Sketch the set S.

Solution: I will begin by answering the second part. Consider the


following: t(1, 2, 3). This is a restatement of the given set of points.
Interpreting (1, 2, 3) as a vector in space, we can see that t acts as a
scalar. Thus the set of points is a straight line parallel to the vector
h1, 2, 3i. To minimize the distance between our curve and the given
point P = (3, 2, 1), we must find a point Q on our curve such that the
function d = |P Q| is minimized. Let Q = (x, 2x, 3x). So,
p
d(x) = (3 − x)2 + (2 − 2x)2 + (1 − 3x)2
d(x)2 = (3 − x)2 + (2 − 2x)2 + (1 − 3x)2
2d(x)d0 (x) = −2(3 − x) − 4(2 − 2x) − 6(1 − 3x)
−3 + x − 4 + 4x − 3 + 9x
d0 (x) =
d(x)
14x − 10
d0 (x) =
d(x)
1. RECTANGULAR COORDINATES IN SPACE 7

10
Thus d(x) is minimized when x = 14 = 57 , as d(x) is never zero (the
curve does not intersect P ). So Q = ( 75 , 10
7 7
, 15 ). Our minimized distance
then is
r
5 10 15 4√
|P Q| = (3 − )2 + (2 − )2 + (1 − )2 ) = 21
7 7 7 7

9–18. Give a geometrical description of the given set S of points defined


algebraically and sketch the set:
9. S = {(x, y, z) | x2 + y 2 + z 2 − 2x + 4y − 6z = 0}

Solution: By completing the square, we obtain


(x − 1)2 − 1 + (y + 2)2 − 4 + (z − 3)2 − 9 = 0
⇒ (x − 1)2 + (y + 2)2 + (z − 3)2 = 14

Thus this is a sphere with radius R = 14 and center (1, −2, 3).

10. S = {(x, y, z) | x2 + y 2 + z 2 ≥ 4}

Solution: Take all of space, then remove the ball x2 + y 2 +


z 2 ≤ 4. Add back in the sphere x2 + y 2 + z 2 = 4. This is the
set (all of space with a spherical cavity).

11. S = {(x, y, z) | x2 + y 2 + z 2 ≤ 4, z > 0}

Solution: This is an intersection of two sets. The first is


a ball of radius 2 centered at the origin. The other is all of
space where z > 0. Thus the set is a hemisphere, minus a disk
of radius 2 in the xy plane.

12. S = {(x, y, z) | x2 + y 2 − 4y < 0, z > 0}

Solution: This is an intersection of two sets. Completing


the square on the first condition yields
x2 + (y − 2)2 − 4 = 0
⇒ x2 + (y − 2)2 = 4
This is a cylinder of radius 2 whose axis is parallel to the z
axis and passes through (0, 2, 0). The second set is all of space
with z > 0. So the set is the portion of the cylinder above the
xy plane.
8 1. VECTORS AND THE SPACE GEOMETRY

13. S = {(x, y, z) | 4 ≤ x2 + y 2 + z 2 ≤ 9}

Solution: This is a ball of radius 3 with a ball of radius


2 (except for its boundary) removed from it.

14. S = {(x, y, z) | x2 + y 2 ≥ 1, x2 + y 2 + z 2 ≤ 4}

Solution: This is an intersection of two sets. The first the


complement of a cylinder of radius 1 aligned along the z axis
(i.e., all of space with such a cylinder removed from it). The
second is a ball of radius 2. Thus this is a ball of radius 2 with
a cylindrical cavity of radius 1 aligned along the z axis (think
like an olive).

15. S = {(x, y, z) | x2 + y 2 + z 2 − 2z < 0, z > 1}

Solution: This is an intersection of two sets. The first, after


completing the square is

x2 + y 2 + (z − 1)2 − 1 < 0
⇒ x2 + y 2 + (z − 1)2 < 1

This is a ball of radius 1 centered at (0, 0, 1), without the


boundary. The second set is all of space with z > 1. Thus this
is a hemisphere, without the boundary.

16. S = {(x, y, z) | x2 + y 2 + z 2 − 2z = 0, z = 1}

Solution: This is an intersection of two sets. The first, after


completing the square is

x2 + y 2 + (z − 1)2 − 1 = 0
⇒ x2 + y 2 + (z − 1)2 = 1

This is a sphere of radius 1 centered at (0, 0, 1). The second


set is the plane z = 1. Thus this is a disk of radius 1 in the
plane z = 1, centered at (0, 0, 1).

17. S = {(x, y, z) | (x − a)(y − b)(z − c) = 0}

Solution: Solutions to the above yield x = a, y = a, or


z = a. These are three planes. Because any one can be true
1. RECTANGULAR COORDINATES IN SPACE 9

(notice the or), it is the union of these planes, not the inter-
section.

18. S = {(x, y, z) | |x| ≤ 1, |y| ≤ 2, |z| ≤ 3}

Solution: This is an intersection of three sets. The first,


|x| ≤ 1, is all of space between the planes x = −1 and x = 1.
The second is all of space between y = −2 and y = 2. The
third is all of space between z = −3 and z = 3. The inter-
section of these gives a solid box centered at the origin with
length 2 along the x-axis, 4 along the y-axis, and a height of
6 along the z-axis.

19–24. Sketch the given set of points and give its algebraic description.

19. A sphere whose diameter is the straight line segment AB, where
A = (1, 2, 3) and B = (3, 2, 1).

Solution: A sphere is determined by its radius and center. Since


AB is a diameter, the radius is
1 1p 1√ √
R = |AB| = (3 − 1)2 + (2 − 2)2 + (1 − 3)2 = 8= 2
2 2 2
The center P0 = (x0, y0 , z0) is the midpoint of the segment AB:
 
1 1 1
P0 = (1 + 3), (2 + 2), (3 + 1) = (2, 2, 2) .
2 2 2
Therefore an equation for the sphere has the standard form:
(x − x0 )2 + (y − y0 )2 + (z − z0)2 = R2
⇒ (x − 2)2 + (y − 2)2 + (z − 2)2 = 2 .
20. Three spheres centered at (1, 2, 3) that just barely touch the xy,
yz, and xz planes, respectively.

Solution: Because a sphere in space is uniquely determined by its


center and radius, we must find the radii of each sphere. Note that
if a sphere ”barely touches” a plane, then it is tangent to that plane.
We can construct a line AP0 , where P0 = (x0, y0 , z0) is the center of
the sphere and A is the point on the plane where the sphere touches
it. This line is perpendicular to the plane; hence |AP0 | is the shortest
distance between P0 and any point on the plane. Thus it is the radius
of our sphere.
10 1. VECTORS AND THE SPACE GEOMETRY

For the xy plane, we have that A = (1, 2, 0). It is easy to verify that
AP0 is perpendicular to the plane. Hence our radius R is
p
R = |AP0 | = (1 − 1)2 + (2 − 2)2 + (0 − 3)2 = 3
Recall that the standard form of a sphere is
(x − x0)2 + (y − y0 )2 + (z − z0)2 = R2
Thus our sphere is
(x − 1)2 + (y − 2)2 + (z − 3)2 = 9
For the yz plane, we have that B = (0, 2, 3). It is easy to verify
that BP0 is perpendicular to the plane. Hence our radius R is
p
R = |BP0| = (0 − 1)2 + (2 − 2)2 + (3 − 3)2 = 1
Recall that the standard form of a sphere is
(x − x0)2 + (y − y0 )2 + (z − z0)2 = R2
Thus our sphere is
(x − 1)2 + (y − 2)2 + (z − 3)2 = 1
For the xz plane, we have that C = (1, 0, 3). It is easy to verify
that CP0 is perpendicular to the plane. Hence our radius R is
p
R = |CP0 | = (1 − 1)2 + (0 − 2)2 + (3 − 3)2 = 2
Recall that the standard form of a sphere is
(x − x0)2 + (y − y0 )2 + (z − z0)2 = R2
Thus our sphere is
(x − 1)2 + (y − 2)2 + (z − 3)2 = 4
21. Three spheres centered at (1, −2, 3) that just barely touch the x,
y, and z coordinate axes, respectively.

Solution: Because a sphere in space is uniquely determined by its


center and radius, we must find the radii of each sphere. Note that if a
sphere ”barely touches” a line, then it is tangent to that line. We can
construct a line AP0 , where P0 = (x0, y0, z0 ) is the center of the sphere
and A is the point on the line where the sphere touches it. This line
is perpendicular to the given line; hence |AP0| is the shortest distance
between P0 and any point on the given line. Thus it is the radius of
our sphere.
1. RECTANGULAR COORDINATES IN SPACE 11

For the x axis, we have that A = (1, 0, 0). It is easy to verify that AP0
is perpendicular to the line. Hence our radius R is
p √ √
R = |AP0| = (1 − 1)2 + (0 + 2)2 + (0 − 3)2 = 4 + 9 = 13
Recall that the standard form of a sphere is
(x − x0)2 + (y − y0 )2 + (z − z0)2 = R2
Thus our sphere is
(x − 1)2 + (y + 2)2 + (z − 3)2 = 13
For the y axis, we have that B = (0, −2, 0). It is easy to verify that
BP0 is perpendicular to the line. Hence our radius R is
p √ √
R = |BP0| = (0 − 1)2 + (−2 + 2)2 + (0 − 3)2 = 1 + 9 = 10
Recall that the standard form of a sphere is
(x − x0)2 + (y − y0 )2 + (z − z0)2 = R2
Thus our sphere is
(x − 1)2 + (y + 2)2 + (z − 3)2 = 10
For the z axis, we have that C = (0, 0, 3). It is easy to verify that CP0
is perpendicular to the line. Hence our radius R is
p √ √
R = |CP0| = (0 − 1)2 + (0 + 2)2 + (3 − 3)2 = 1 + 4 = 5
Recall that the standard form of a sphere is
(x − x0)2 + (y − y0 )2 + (z − z0)2 = R2
Thus our sphere is
(x − 1)2 + (y + 2)2 + (z − 3)2 = 5
22. The largest solid cube that is contained in a ball of radius R cen-
tered at the origin. Solve the same problem if the ball is not centered
at the origin. Compare the cases when the boundaries of the solid are
included into the set or excluded from it.

Solution: For some problems, it helps to drop down into a lower


dimension first; That is, if the question were asking to find the largest
square contained in a circle of radius R. It is evident that the largest
square will have its corners touching the circle and its center at the
center of the circle. In other words, the distance from the center of the
circle to any corner of the square is R, and so the diagonal is 2R. Thus
the length of the sides of the square is d = √12 R.

We can now apply similar logic to the 3D case. Evidently, the largest
12 1. VECTORS AND THE SPACE GEOMETRY

solid cube will be that whose eight corners also touch points on the
sphere and whose center is also the center of the sphere. Now we must
find the length of each side of the cube. Suppose the side length is 2d.
Then we can find the distance between √ a corner and the center of a
2 2 2
nearby face by D = d + d , so D = 2d. Next, the distance between
this point and the center of the cube is d, so the distance between the
center of the cube and a corner is R2 = D2 + d2 = 3d2 , so d = √13 R,
and our side length is √23 R.

Let the center of the sphere be at (x0 , y0, z0). Then our solid cube
is given by
1 1 1
S = {(x, y, z) | |x − x0| ≤ √ R, |y − y0 | ≤ √ R, |z − z0 | ≤ √ R}
3 3 3
Note that I did not include the 2 for each √1 R. This is due to the
3
absolute value.

23. The solid region that is a ball of radius R that has a cylindrical hole
of radius R/2 whose axis is at a distance of R/2 from the center of the
ball. Choose a convenient coordinate system. Compare the cases when
the boundaries of the solid are included into the set or excluded from it.

Solution: Consider such a solid region in space. Then, up to a fi-


nite number of rigid transformations (distance and angle preserving),
this region is identical to one centered at the origin of a standard xyz
coordinate system, with the cylindrical hole aligned parallel to the z
(”up”) axis. The algebraic description for a ball (solid sphere) of radius
R centered at the origin is simply
x2 + y 2 + z 2 ≤ R2 .
Next, we want the axis of the cylinder to be a distance R/2 from
the origin. I will choose that the cylinder’s axis goes through the
point (0, R/2, 0). Realistically, however, it can go through any point
(R/2 cos θ, R/2 sin θ, 0) for any real θ. The general form of a solid
cylinder, whose axis passes through (0, R/2, 0, aligned with the z-axis,
is
x2 + (y − R/2)2 < (R/2)2
It is crucial that a < is used rather than ≤. This will be explained
later.
Since the cylinder is actually a hole, we must take the complement of
1. RECTANGULAR COORDINATES IN SPACE 13

this,
x2 + (y − R/2)2 ≥ R2 /4
If we had used < earlier, the ≥ would now be a >. This means that
the boundary between the solid and the hole would not be included.
The region is simply the intersection of these two solids,
S = {(x, y, z) ∈ R×R×R | x2 +y 2 +z 2 ≤ R2 ∩x2 +(y −R/2)2 ≥ R2 /4}
And without boundaries,
S = {(x, y, z) ∈ R×R×R | x2 +y 2 +z 2 < R2 ∩ x2 +(y −R/2)2 > R2 /4}
24. The part of a ball of radius R that lies between two parallel planes
each of which is at a distance of a < R from the center of the ball.
Choose a convenient coordinate system. Compare the cases when the
boundaries of the solid are included into the set or excluded from it.

Solution: Then, up to a finite number of rigid transformations (dis-


tance and angle preserving), this region is identical to one centered at
the origin of a standard xyz coordinate system, with the parallel planes
also parallel to the xy plane. The algebraic description for a ball (solid
sphere) of radius R centered at the origin is simply
x2 + y 2 + z 2 ≤ R2 .
Let a < R be a real number. The two planes in question can be seen
as one above the center and another below the center. Recall that a
plane parallel to the xy plane is simply the set of all points whose z
coordinate is a constant value. Thus, the two planes in question are
z=a
z = −a
Now we want to classify all the points between these two planes. This
will be all of the points below z = a and above z = −a. So all the
points between the two planes are all the points where −a ≤ z ≤ a,
where we have leq instead of < because we want the boundaries to be
included. Then the set of all points in the ball of radius R and between
the two planes will be the intersection of the two found sets,
S = {(x, y, z) ∈ R × R × R | x2 + y 2 + z 2 ≤ R2 ∩ −a ≤ z ≤ a
And without boundaries,
S = {(x, y, z) ∈ R × R × R | x2 + y 2 + z 2 < R2 ∩ −a < z < a
It may help to consider the 2D analogue first to have a clear visual idea
of the problem.
14 1. VECTORS AND THE SPACE GEOMETRY

25. Consider the points P such that the distance from P to the point
(−3, 6, 9) is twice the distance from P to the origin. Show that the set
of all such points is sphere, and find its center and radius.

Solution: Let P = (x, y, z) be an arbitrary point in space. The


distance between P and the given point P0 = (−3, 6, 9) is
p
|P P0 | = (x + 3)2 + (y − 6)2 + (z − 9)2
Next, the distance between P and the origin is
p
|P O| = x2 + y 2 + z 2
If the distance between P and P0 is twice the distance from P to the
origin, then we have
p p
(x + 3)2 + (y − 6)2 + (z − 9)2 = 2 x2 + y 2 + z 2
⇒ (x + 3)2 + (y − 6)2 + (z − 9)2 = 4(x2 + y 2 + z 2)
⇒ 3x2 − 6x − 9 + 3y 2 + 12y − 36 + 3z 2 + 18z − 81 = 0
⇒ 3(x2 − 2x − 3) + 3(y 2 + 4y − 12) + 3(z 2 + 6z − 27) = 0
⇒ 3(x2 − 2x + 1 − 4) + 3(y 2 + 4y + 4 − 16) + 3(z 2 + 6z + 9 − 36) = 0
⇒ 3(x2 − 2x + 1) − 12 + 3(y 2 + 4y + 4) − 48 + 3(z 2 + 6z + 9) − 108 = 0
12 + 48 + 108
⇒ (x − 1)2 + (y + 2)2 + (z + 3)2 =
3
2 2 2
⇒ (x − 1) + (y + 2) + (z + 3) = 56 .

Thus it is a sphere with center (1, −2, −3) and radius 56.

26. Find the volume of the solid whose boundaries are the spheres
x2 + y 2 + z 2 − 6z = 0 and x2 + y 2 − 2y + z 2 − 6z = −9.

Solution: By completing the square, the equations of the spheres


are written in the standard form:
x2 + y 2 + (z − 3)2 = 9 , x2 + (y − 1)2 + (z − 3)2 = 1 .
It follows then that the first sphere has radius R1 = 3 and is centered
at (0, 0, 3), while the second sphere lies inside the first one because it
has radius R2 = 1 and is centered at (0, 1, 3). So the solid is the ball
bounded by a sphere of radius R1 with a spherical cavity of radius R2 .
Therefore the volume of the solid is the difference of the volumes of the
ball and the cavity:
4π 3 4π 3 4π 3 104π
V = R1 − R2 = (3 − 13 ) = .
3 3 3 3
1. RECTANGULAR COORDINATES IN SPACE 15

27. Find the volume of the solid that is described by the inequalities
|x − 1| ≤ 2, |y − 2| ≤ 1, and |z + 1| ≤ 2. Sketch the solid.

Solution: The solid is a rectangular prism. Recall that if |a| < b,


then −b < a < b, and the set of points satisfying this is bounded be-
low by −b and above by b. Thus the solid is bounded by the planes
x − 1 = 2 ⇒ x = 3 and x − 1 = −2 ⇒ x = −1,
y − 2 = 1 ⇒ y = 3 and y − 2 = −1 ⇒ y = 1,
z + 1 = 2 ⇒ z = 1 and z + 1 = −2 ⇒ z = −3.
We can calculate the side lengths by finding the distance in between
each set of parallel planes. Hence the side length parallel to x is
3 − (−1) = 4, the side length parallel to y is 3 − 1 = 2, and the side
length parallel to z is 1 − (−3) = 4. Notice that we had |x − 1| ≤ 2,
and the side length along the x axis was 2 ∗ 2. This is no coincidence.
If |x − a| ≤ b, the side length along the x axis will be 2b. Thus the
volume is V = 4(2)(4) = 32.

28. The solid region is described by the inequalities |x − a| ≤ a,


|y − b| ≤ b, |z − c| ≤ c, and (y − b)2 + (z − c)2 ≥ R2 . If R ≤ min(b, c),
sketch the solid and find its volume.

Solution: First note that the solid described by the inequalities


|x − a| ≤ a, |y − b| ≤ b, and |z − c| ≤ c is a rectangular prism with side
lengths 2a, 2b, and 2c (see problem 27 for a more detailed explanation).
The rectangular prism is centered so that (a, b, c) is the center of it.

Next, we have a hole formed by (y − b)2 + (z − c)2 ≥ R2 . In 2D,


this imagine having a square with a circle perfectly embedded inside it.
Because we want the set of points greater than the radius, but bounded
by the square, we will have a circular hole. In 3D, the inequality de-
scribes a cylinder, because of its independence of x. Furthermore, this
tells us that the cylinder’s axis is parallel to the x axis, and that its axis
is along the line (t, b, c). That is, the axis of the cylinder cuts through
the center of the prism. The height of the cylinder is 2a, because it is
parallel to the x axis.

Thus the volume of the cylinder is Vc = πR2 (2a) = 2aπR2, where


R ≤ min(b, c) (Note: The min function is used to guarantee that the
cylinder lies perfectly inside the prism). The volume of the prism
is Vs = 2a(2b)(2c) = 8abc. Thus the volume of the solid is V =
8abc − 2aπR2 .
16 1. VECTORS AND THE SPACE GEOMETRY

29. Sketch the set of all points in the xy plane that are equidistant
from two given points A and B. Let A and B be (1, 2) and (−2, −1),
respectively. Give an algebraic description of the set.

Solution: Imagine a square, with two opposite (diagonal) corners


labeled A and B. It is obvious that the other two diagonal corners are
equidistant to A and B. Furthermore, the midpoint of the diagonal
is equidistant from A and B. This leads us to conclude that the set
of all points equidistant from A and B is the set of points on the line
through the other diagonal. Intuitively, this makes sense. This line is
perpendicular to the diagonal containing A and B and goes through
the midpoint. Call the midpoint O. Pick a point on the line, P . Then
triangle OAP is congruent to triangle OBP by SAS, as they share the
same side in common, and an angle in between (the right angle).
−1−2
The slope of the line connecting A and B is m = −2−1 = 1. So a
−1
line perpendicular to this will have slope M = m = −1. The midpoint
of AB is ( −2+1
2
, 2−1
2
) = ( −1 , 1 ). So the set of points equidistant from A
2 2
and B make the line y = −(x + 12 ) + 12 ) = −x. So the set of points is
S = {(x, y) ∈ R × R | y = −x}.

30. Sketch the set of all points in space that are equidistant from
two given points A and B. Let A and B be (1, 2, 3) and (−3, −2, −1),
respectively. Give an algebraic description of the set.

Solution: Extending the idea of number 29 to 3D, the solution


will be a plane through the midpoint of AB and perpendicular to
it. Let P = (x, y, z) ∈ P. Let O be the midpoint of AB, so O =
( 1−3
2
, 2−2
2
, 3−1
2
) = (−1, 0, 1). Because P is perpendicular to AB, trian-
gle OAP is a right triangle. Then |OA|2 + |OP |2 = |AP |2 . |OA|= ((1 +
1)2 + (2 − 0)2 + (3 − 1)2 ) = 12. So,
12 + ((x + 1)2 + (y − 0)2 + (z − 1)2 ) = (x − 1)2 + (y − 2)2 + (z − 3)2
(x − 1)2 − (x + 1)2 + (y − 2)2 − y 2 + (z − 3)2 − (z − 1)2 = 12
(−2)(2x) + (−2)(2y − 2) + (−2)(2z − 4) = 12
(x) + (y − 1) + (z − 2) = −3
x+y+z = 0
Where I applied that a2 − b2 = (a − b)(a + b) several times.
1. RECTANGULAR COORDINATES IN SPACE 17

So P = {(x, y, z) ∈ R × R × R | x + y + z = 0}

31. A point P = (x, y) belongs to the set S in the xy plane if


|P A| + |P B| = c where A = (a, 0), B = (−a, 0), and c > 2a. Show
that S is an ellipse.

Solution: One might remember that an ellipse can be generated by


two foci A, B as above. If not, then we have
p p
(x − a)2 + (y − 0)2 + (x − (−a))2 + (y − 0)2 =c
p p
(x − a)2 + y 2 + (x + a)2 + y 2 =c
p p
( (x − a)2 + y 2 + (x + a)2 + y 2)2 = c2
p p
(x − a)2 + y 2 + 2 (x − a)2 + y 2 (x + a)2 + y 2 + (x + a)2 + y 2 = c2
p p
(x − a)2 + (x + a)2 + 2y 2 + 2 (x − a)2 + y 2 (x + a)2 + y 2 = c2

We could continue by bringing all the radicals to one side and squaring
again, but I find that tedious. Instead, return to the original equation
and do the following:
p p
(x − a)2 + y 2 + (x + a)2 + y 2 = c
p p
(x − a)2 + y 2 = c − (x + a)2 + y 2
p
(x − a)2 + y 2 = (c − (x + a)2 + y 2)2
p
(x − a)2 + y 2 = c2 − 2c (x + a)2 + y 2 + (x + a)2 + y 2
p
(x − a)2 − (x + a)2 = c2 − 2c (x + a)2 + y 2
p
(x − a + x + a)(x − a − (x + a)) = c2 − 2c (x + a)2 + y 2
p
−4ax = c2 − 2c (x + a)2 + y 2
p
2c (x + a)2 + y 2 = c2 + 4ax
p c 2ax
(x + a)2 + y 2 = +
2 c

Where the identity a2 − b2 = (a + b)(a − b) has been used. We can


obtain a similar formula by substituting a for −a,

p c 2ax
(x − a)2 + y 2 = −
2 c
18 1. VECTORS AND THE SPACE GEOMETRY

Substituting these into the first equation we found yields


c 2ax c 2ax
(x − a)2 + (x + a)2 + 2y 2 + 2( + )( − ) = c2
2 c 2 c
c 2ax 2
(x − a)2 + (x + a)2 + 2y 2 + 2(( )2 − ( ) ) = c2
2 c
c2 8a2 x2
x2 − 2ax + a2 + x2 + 2ax + a2 + 2y 2 + ( − 2
) = c2
2 c
2 2
8a x c2
2x2 − + 2a 2
+ 2y 2
=
c2 2
2 2 2
(2c − 8a )x 2 c2
+ 2y = − 2a2
c2 2
4x2 4y 2
+ =1
c2 c2 − 4a2
This is an ellipse. Note that the coefficient of y 2 must be positive.
hence c2 − 4a2 > 0 ⇔ c2 > 4a2 ⇔ c > 2a or c < −2a Since c > 0, it
must be that c > 2a, hence the condition given in the problem.

32. Determine whether the points A = (1, 0, −1), B = (3, 1, 1), and
C = (2, 2, −3) are vertices of a right-angled triangle.

Solution: Suppose that the given points A, B, C are points of a right-


angled triangle. Then, by the pythagorean theorem, the square of the
largest side length must be equal to the sum of the squares of the other
perpendicular side lengths. So we must first find |AB|, |BC|, and |AC|
as follows
p √
|AB| = (1 − 3)2 + (0 − 1)2 + (−1 − 1)2 = 9 = 3
p √
|BC| = (3 − 2)2 + (1 − 2)2 + (1 + 3)2 = 18
p √
|AC| = (1 − 2)2 + (0 − 2)2 + (−1 + 3)2 = 9 = 3
It is evident that |AB|2 + |AC|2 = |BC|. Hence ABC is a right-angled
triangle.
2. VECTORS IN SPACE 19

2. Vectors in Space
1–5. Find the components and norms of each of the following vectors.
−→
1. AB where A = (1, 2, 3) and B = (−1, 5, 1).

Solution: Recall that


−→ −−→ −→
AB = OB − OA,
−→ −−→
where OA and OB are the position vectors of A and B, respectively.
So,
−→
AB = h−1, 5, 1i − h1, 2, 3i = h−1 − 1, 5 − 2, 1 − 3i = h−2, 3, −2i
Furthermore,
−→ p √
kABk = (−2)2 + 32 + (−2)2 = 17

−→
2. BA where A = (1, 2, 3) and B = (−1, 5, 1).

Solution: Recall that


−→ −→
BA = −AB.
Then, by 1,
−→
BA = −h−2, 3, −2i = h2, −3, 2i.
Since only the direction has changed,
−→ −→ √
kBAk = kABk = 17
−→
3. AC where C is the midpoint of the line segment AB with A =
(1, 2, 3) and B = (−1, 5, 1).

Solution: Since C is the midpoint of AB, we have that


−→ 1 −→ 1 3
AC = AB = h−2, 3, −2i = h−1, , −1i.
2 2 2
Furthermore, because C is the midpoint of AB, we know that
1
|AC| = |AB|
2
Thus
−→ 1 −→ 1√
kACk = kABk = 17
2 2
4. The position vector of a point P obtained from the point A =
(−1, 2, −1) by moving the latter along a straight line by a distance of
3 units in the direction of the vector u = h2, 2, 1i then by a distance of
20 1. VECTORS AND THE SPACE GEOMETRY

10 units in the direction of the vector w = h−3, 0, −4i.

Solution: The position vector of the point A is a = h−1, 2, −1i.


Let p be the position vector of the point P . Then by the rules of
vector algebra
p = a + tu + sw
where positive numbers t and s must be chosen so that the first dis-
placement vector tu has a length of 3 units and the second displacement
vector sw has a length of 10 units:

3 = ktuk = tkuk = t 22 + 22 + 1 = 3t ⇒ t = 1 ,

10 = kswk = skwk = s 32 + 0 + 42 = 5s ⇒ s = 2 .
Therefore
p = h−1, 2, −1i + h2, 2, 1i + 2h−3, 0, −4i = h−5, 4, −8i
⇒ P = (−5, 4, −8) .
5. The position vector of the vertex C of a triangle ABC in the first
quadrant of the xy plane if A is at the origin, B = (a, 0, 0), the angle
at the vertex B is 2π/3, and |BC| = 2a.

Solution: First recognize that since the angle at vertex B is 2π/3,


−−→ −→
the angle between OB and OC is actually π/3. This can be seen by
−→
first sketching triangle ABC and moving the tail of OC to the tail of
−−→ −−→
OB. Since OB is parallel to the x axis, we know that the angle between
−→
OC and the x axis is π/3. Next, recall that we can write the position
vector of a point P in the xy plane as
−→
OP = hp cos(θ), p sin(θ), 0i
where p is taken to be the distance between the origin and P and θ
is the smallest angle between OP and the x axis. This is simply the
polar representation of a point P in the xy plane. Thus,
−−→ √
BC = h2a cos(π/3), 2a sin(π/3), 0i = ha, 3a, 0i
−−→ −→
Note that this is the vector for BC! If this were OC, then we would
−→
have that |AC| = 2a, but we are given |BC| = 2a. To find OC , observe
that
−→ −−→ −−→ √ √ √
OC = OB + BC = ha, 0, 0i + ha, 3a, 0i = h2a, 3a, 0i = ah2, 3, 0i
Furthermore,
√ 2 √
q
−→
kOCk = a 22 + 3 = 7a
2. VECTORS IN SPACE 21

6. Are the points A = (−3, 1, 2), B = (1, 5, −2), C = (0, 3, −1), and
D = (−2, 3, 1) vertices of a parallelogram?

Solution: If the given points are indeed the sides of a parallelo-


gram, then we must have that two pairs of sides are equal in length
and parallel. That is, the we need to find two pairs of vectors between
two distinct points that are equal (up to a negative sign, which just
indicates direction). So,
−→
AB = h1, 5, −2i − h−3, 1, 2i = h4, 4, −4i
−→
AC = h0, 3, −1i − h−3, 1, 2i = h3, 2, −3i
−−→
AD = h−2, 3, 1i − h−3, 1, 2i = h1, 2, −1i
−−→
BC = h0, 3, −1i − h1, 5, −2i = h−1, −2, 1i
−−→
BD = h−2, 3, 1i − h1, 5, −2i = h−3, −2, 3i
−−→
CD = h−2, 3, 1i − h0, 3, −1i = h−2, 0, 2i

Observe that
−→ −−→
AC = −BD
−−→ −−→
AD = −BC
So ACBD is a parallelogram, with A and B opposite each other and
C and D opposite each other.

7. If A = (2, 0, 3), B = (−1, 2, 0), and C = (0, 3, 1), determine the


point D such that A, B, C, and D are vertices of a parallelogram with
sides AB, BC, CD, and DA.

Solution: From the description of the parallelogram it follows that


the side AB is parallel to CD and the side BC is parallel to DA. In
−−→ − −→
particular, AD = BC. Let a = h2, 0, 3i be the position vector of the
point A. Then, by the parallelogram rule, the position vector of the
point D is
−−→ −−→
a + AD = a + BC = h2, 0, 3i + h0 − (−1), 3 − 2, 1 − 0i = h3, 1, 4i .
So, D = (3, 1, 4).

8. A parallelogram has a vertex at A = (1, 2, 3) and two sides a =


h1, 0, −2i and b = h3, −2, 6i adjacent at A. Find the coordinates of
22 1. VECTORS AND THE SPACE GEOMETRY

the point of intersection of the diagonals of the parallelogram.

Solution: The point of intersection of the diagonals is the midpoint


of either of them. By the parallelogram rule, the diagonal extended
from A is a + b. Therefore the position vector of the midpoint relative
to the point A is 12 (a + b) so that the position vector of the midpoint
relative to the origin is
−→ 1 1
OA + (a + b) = h1, 2, 3i + h1 + 3, 0 − 2, −2 + 6i
2 2
= h1, 2, 3i + h2, −1, 2i = h3, 1, 5i
Thus, the point in question is (3, 1, 5).

9. Draw two vectors a and b that are neither parallel and nor perpen-
dicular. Sketch each of the following vectors: a + 2b, b − 2a, a − 12 b,
and 2a + 3b.
10. Draw three vectors a, b, and c in a plane, with none of them
parallel to either of the others. Sketch each of the following vectors:
a + (b − c), (a + b) − c, 2a − 3(b + c), and (2a − 3b) − 3c.
11. Let a = h2, −1, −2i and b = h−3, 0, 4i. Find unit vectors â and b̂.
Express 6â − 15b̂ in terms of a and b.

Solution: We first have that


p
kak = 22 + (−1)2 + (−2)2 = 3
p
kbk = (−3)2 + 42 = 5
So
1 1
â = a= a
kak 3
1 1
b̂ = b= b
kbk 5
And thus
6 15
6â − 15b̂ = a − b = 2a − 3b
3 5
12. Let a and b be vectors in the xy plane such that their sum c = a+b
makes the angle π/3 with a and has the length twice the length of a.
Find b if a is based at the origin, has its terminal point in the first
quadrant, makes an angle π/3 with the positive x-axis, and has length
a. There are two vectors b with these properties. Find both of them.
2. VECTORS IN SPACE 23

Solution: First we must find a. Since it has length a and makes


an angle of π/3 with the positive x-axis, we can write a in polar coor-
dinates as √
1 3
a = ha cos(π/3), a sin(π/3)i = ah , i
2 2
Now, there are two options for c which will lead to our two answers for
b, both of which are symmetric about a. The first is when c is ”above
a”. That is, the angle it makes with the positive x-axis is greater than
π/3. Since we know that the angle c makes with a is π/3, and the
angle a makes with the positive x-axis is π/3, we know that the angle
c makes with the positive x-axis in this case is 2π/3. The other option
is when c is ”below a”. In this case, the angle c will make with the
positive x-axis is 0. Let us split these into two cases.

Case 1: c makes angle 2π/3 with the positive x-axis.


In this case, we can write c in polar form as

c = 2ahcos(2π/3), sin(2π/3)i = ah−1, 3i,
where the 2a came from the fact that kc| = 2a.
By vector algebra,
√ √
√ 1 3 3 3
b = c − a = ah−1, 3i − ah , i = ah− , i
2 2 2 2

Case 2: c makes angle 0 with the positive x-axis.


In this case, we can write c in polar form as
c = 2ahcos(0), sin(0)i = 2ah1, 0i,
where the 2a came from the fact that kc| = 2a.
By vector algebra,
√ √
1 3 3 3
b = c − a = ah2, 0i − ah , i = ah , − i
2 2 2 2

13. Consider a triangle ABC. Let a be a vector from the vertex A to


the midpoint of the side BC, let b be a vector from B to the midpoint
of AC, and let c be a vector from C to the midpoint of AB. Use vector
algebra to find a + b + c, that is, do not resort to writing vectors in
component-form; just use properties of vector addition, subtraction,
and multiplication by scalars.
24 1. VECTORS AND THE SPACE GEOMETRY

Solution: First observe the following: If P and Q are vertices of


a triangle with midpoint M between them, then

1
m = (p + q).
2

Where m, p, q are taken to be the position vectors of M, P, Q respec-


tively.
Let D be the midpoint of BC, E the midpoint of AC, and F the
midpoint of AB. Then,

−−→ −−→ −→
a + b + c = DA + EB + F C
−→ −−→ −−→ − −→ −→ −→
= OA − OD + OB − OE + OC − OF
−→ −−→ −→ 1 −−→ −→ 1 −→ −→ 1 −→ −−→
= OA + OB + OC − (OB + OC) − (OA + OC) − (OA + OB)
2 2 2
−→ −−→ −→ 1 −→ 1 −→ 1 −−→ 1 −−→ 1 −→ 1 −→
= OA + OB + OC − OA − OA − OB − OB − OC − OC
2 2 2 2 2 2
=0

14. Let ûk , k = 1, 2, ..., n, be unit vectors in the plane such that the
smallest angle between the two vectors ûk and ûk+1 is 2π/n. What is
the sum vn = û1 + û2 +· · ·+ ûn for an even n? Sketch the sum for n = 1,
n = 3, and n = 5. Compare the norms kvn k for n = 1, 3, 5. Investi-
gate the limit of vn as n → ∞ by studying the limit of kvn k as n → ∞.

Solution: Consider a regular n-gon with its center located at the


origin of the xy plane, where each vertex is one unit away from the
origin. Then each ûk is the position vector of a vertex of the n-gon.
To see this, observe that we can split an entire angle (2π) into n equal
sized pieces, each of which has magnitude 2π/n (like cutting a pizza).
For the even n, it is obvious that one could rotate the n-gon by π and
obtain the original image. This implies that for every vertex, there
is another vertex exactly opposite. The sum of the position vectors
of these vertices will cancel (equal magnitude but opposite direction).
Thus vn = 0 for even n. The same will be true for odd n, with the
exception of when n = 1. (Sketches to come).
Referring back to the fact that we can construct each ûk as the vertex
2. VECTORS IN SPACE 25

of a regular n-gon as above, we have that:


v1 = hcos(0), sin(0)i = h1, 0i
v3 = hcos(0), sin(0)i + hcos(2π/3), sin(2π/3)i + hcos(4π/3), sin(4π/3)i
√ √
1 3 1 3
= h1, 0i + h− , i + h− , − i
2 2 2 2
=0
v5 = hcos(0), sin(0)i + hcos(2π/5), sin(2π/5)i + hcos(4π/5), sin(4π/5)i
+hcos(6π/5), sin(6π/5) + hcos(8π/5), sin(8π/5)i
=0
Therefore vn = 0 as n → ∞

15. Let ûk , k = 1, 2, ..., n, be unit vectors as defined in Exercise 14.


Let wk = ûk+1 − ûk for k = 1, 2, ..., n − 1 and wn = û1 − ûn . Find the
limit of kw1 k + kw2 k + · · · + kwn k as n → ∞. Hint: Use a geometrical
interpretation of the sum.

Solution: As with Exercise 14, we can consider a regular n-gon cen-


tered at the origin of the xy plane with vertices distance 1 from the
origin. Enumerate the vertices as 1, 2, ..., n. Consider vertices k and
k + 1. Recall that we discovered that the position vector for vertex k
is ûk . Convince yourself that wk = ûk+1 − ûk is the vector from vertex
k to vertex k + 1. Therefore the sum of all kwk k from 0 to n is simply
the perimeter of the n gon. As n → ∞, the n-gon approaches a circle.
Therefore kw1 k + kw2 k + · · · + kwn k = 2π as n → ∞.
The fact that the sum of the wk gives a length is important for the
formulation of a line integral in later chapters.

16. Suppose a wind is blowing at a speed of u mi/h in the direc-


tion that is 0 < α < 90◦ degrees west of the northerly direction. A
pilot is steering a plane in the direction that is 0 < β < 90◦ degrees
east of the northerly direction at an airspeed (speed in still air) of
v > u mi/h. The true course of the plane is the direction of the re-
sultant of the velocity vectors of the plane and the wind. The ground
speed is the magnitude of the resultant. Find the true course and the
ground speed of the plane. If α, u, and v are fixed, what is the direction
in which the pilot should steer the plane to make the true course north?

Solution: First we must define a coordinate system. Let the positive


y direction be north. Then the vector representing the wind speed is
26 1. VECTORS AND THE SPACE GEOMETRY

given by
u = hu cos(α + π/2), u sin(α + π/2)i
= uhcos(α) cos(π/2) − sin(α) sin(π/2), sin(α) cos(π/2) + cos(α) sin(π/2)i
= uh− sin(α), cos(α)i
where we had to shift the arguments of the cos and sin by π/2 because
north is aligned along the positive y-axis. To find v, we will apply the
same thing as follows:
v = hv cos(−β + π/2), u sin(−β + π/2)i
= vhsin(β), cos(β)i
where there is a negative sign in front of β in the first line because to
move east, we must rotate clockwise, which is in the negative direction.
The resultant vector, then, is
v + u = hv sin(β) − u sin(α), v cos(β) + u cos(α)i
and the magnitude of the resultant is
p
ku + vk = (v sin(β) − u sin(α))2 + (v cos(β) + u cos(α))2
p
= v 2 + u2 + 2uv(cos(α) cos(β) − sin(α) sin(β))
p
= v 2 + u2 + 2uv cos(α + β)

If the course is true north, then the resultant vector must be parallel
to the y-axis. So the x-component of the resultant must be 0. Thus,
v sin(β) − u sin(α) = 0
v sin(β) = u sin(α)
u
sin(β) = sin(α)
v
u
β = arcsin( sin(α))
v
Note that if u > v, it may be that β cannot be computed. Hence the
restriction that v > u in the exercise.

17. Use vector algebra (do not resort to writing vectors in component-
form) to show that the line segment joining the midpoints of two sides
of a triangle is parallel to the third side and half its length.

Solution: Let A, B, C be the vertices of a triangle. Let D, E be


2. VECTORS IN SPACE 27

the midpoints of AC and BC, respectively. We must then prove that


−−→ 1 −→
DE = AB.
2
First, by Exercise 13, we know that the position vector of the midpoint
M of a line with endpoints P and Q is
1
m = (p + q)
2
where m, p, q are taken to be the position vectors of M, P, Q respec-
tively. So, we have that
1
d = (a + c)
2
1
e = (a + b)
2
Then
−−→ − −→ −−→
DE = OE − OD
1 1
= (b + c) − (a + c)
2 2
1 1 1 1
= b+ c− a− c
2 2 2 2
1 1
= b− a
2 2
1 −−→ −→
= (OB − OA)
2
1 −→
= AB
2
Note: Working from
−−→ 1 −→
DE = AB
2
to
−−→ −−→
DE = DE
would be an incorrect proof; it is in the wrong direction.

18– 21. Describe geometrically the set of points whose position vectors
r satisfy the given conditions.
18. kr − ak = k and r lies in the xy plane, where a is a vector in
the xy plane and k > 0.

Solution: First, we can recognize a as the position vector


of some point A in the xy plane. Then, we can interpret
kr − ak = k as all the points a fixed distance (k) from A.
28 1. VECTORS AND THE SPACE GEOMETRY

This is simply a circle with center A and radius k.

19. kr − ak + kr − bk = k and r lies in the xy plane, where a and


b are vectors in the xy plane and k > ka − bk.

Solution: First, a and b represent the position vectors of


two points in the xy plane, A and B. Recall that an ellipse
can be generated by the set of all points such that the sum of
the distances to two points is a fixed value (see Section 1.1,
Exercise 31). This is exactly what is written above, so it is an
ellipse with foci A and B.

20. kr − ak = k, where a is a vector in space and k > 0.

Solution: This is the 3D analogue to Exercise 18. Thus


we have a sphere of raidus k whose center is at A, where A is
the point with position vector a.

21. kr − ak + kr − bk = k, where a and b are vectors in space and


k > ka − bk.

Solution: This is the 3D analogue to Exercise 19. Thus


we have an ellipsoid with foci A and B, whose position vectors
are a and b, respectively.

22. Let point-like massive objects be positioned at Pi , i = 1, 2, ..., n,


and let mi be the mass at Pi . The point P0 is called the center of
mass if

−−→ −−→ −−−→


m1 P0 P1 + m2P0 P2 + · · · + mn P0 Pn = 0

Express the position vector r0 of P0 in terms of the position vectors


ri of Pi . In particular, find the center of mass of three point masses,
m1 = m2 = m3 = m, located at the vertices of a triangle ABC for
A = (1, 2, 3), B = (−1, 0, 1), and C = (1, 1, −1).

Solution: First observe that for any 1 ≤ i ≤ n,

−−→
P0 Pi = ri − r
2. VECTORS IN SPACE 29

So we have that
−−→ −−→ −−−→
m1 P0 P1 + m2 P0 P2 + · · · + mn P0 Pn = 0
n
X −−→
mk P0 Pk = 0
k=1
n
X
mk (rk − r) = 0
k=1
n
X n
X
m k ri − mk r = 0
k=1 k=1
n
X n
X
mk r = m k ri
k=1 k=1
Pn
m k ri
r = Pk=1
n
k=1 mk

In the given problem, we are asked to find the center of mass of


three point masses, each with mass m, located at A = (1, 2, 3), B =
(−1, 0, 1), and C = (1, 1, −1). So,
Pn
m k ri
r = Pk=1
n
k=1 mk
m(h1, 2, 3i + h−1, 0, 1i + h1, 1 − 1i)
=
3m
1
= h1 − 1 + 1, 2 + 0 + 1, 3 + 1 − 1i
3
1
= h1, 3, 3i
3

23. Consider the graph y = f(x) of a differentiable function and the


line tangent to it at a point x = a. Express components of a vector
parallel to the line in terms of the derivative f 0 (a) and find a vector
perpendicular to the line. In particular, find such vectors for the graph
y = x2 at the point x = 1.

Solution: Recall that we can write the equation for the tangent line
to a differentiable function at x = a as

yt = f 0 (a)(x − a) + f(a)
30 1. VECTORS AND THE SPACE GEOMETRY

and for the normal line as


1
yn = − (x − a) + f(a)
f 0 (a)
A vector parallel to this line can simply be the difference between
the position vectors of two points on the line. I will use the points
A = (a, f(a)) and Bt = (0, yt (0)). So,
−→ −−→
vt = OA − OBt
= ha, f(a)i − h0, −af 0 (a) + f(a)i
= ha, af 0 (a)i || h1, f 0 (a)i
is a vector parallel to the tangent line, and using Bn = (0, yn (0)),
−→ −−→
vn = OA − OBn
a
= ha, f(a)i − h0, 0 + f(a)i
f (a)
a
= ha, − 0 i || hf 0 (a), −1i
f (a)

In particular when f(x) = x2 and x = 1,


vt = h1, f 0 (1))i = h1, 2i

vn = h−f 0 (1), 1)i = h−2, 1i

24. Let the vectors a, b, and c have fixed lengths a, b, and c, respec-
tively, while their direction may be changed. Is it always possible to
achieve a + b + c = 0? If not, formulate the most general condition
under which it is possible.

Solution: No it is not always possible. Consider a = h3, 0i, b =


c = h1, 0i. Notice that the best way to get to the zero vector is to have
b, c be antiparallel to a, as this will have the smallest magnitude of
the resultant vector. In this case, a − b − c = h3 − 1 − 1, 0i = h1, 0i,
which is not the zero vector.
What we are trying to achieve is to form a triangle from the vectors
a, b, c. In this case, we were unable to form a triangle. So the question
boils down to: Given three lines of lengths a, b, c, is there any way to
make a triangle from them? Recall the cosine rule
c2 = a2 + b2 − 2ab cos(C)
2. VECTORS IN SPACE 31

So we must have that


c2 − a 2 − b 2
C = arccos( )
2ab
For C to be defined,
c2 − a 2 − b 2
| | ≤1
2ab
|c2 − a2 − b2| ≤ |2ab|

25. Let the vectors a and b have fixed lengths, while their directions
may be changed. Put c± = ka ± bk. It is always possible to achieve
that c− > c+ , or c− = c+ , or c− < c+ ? If so, give examples of the
corresponding relative directions of a and b.

Solution: Yes it is possible. First, note that c− and c+ are real


numbers, hence there must be some way to relate them (via <, =, or
>). Consider a and b to be vectors with length 1. If a, b are parallel
(in the same direction), then the lengths will simply add and we will
have that c+ = 2 and c− = 0. If a, b are parallel (in the opposite
direction), then the lengths will subtract when the vectors are added,
so c+ = 0 and c− = 2. If a, b are perpendicular, then c+ and c− will
simply be the length of the hypotenuse of the triangle formed by a and
b and a and −b, respectively. These two triangles share a common an-
gle (perpendicular) between two sides of the same length, hence they
are congruent (SAS), so their hypotenuses are the same.

26. A point object travels in the xy plane, starting from an initial


point P0 . Its trajectory consists of straight line-segments, where the
nth segment starts at point Pn−1 and ends at point Pn (for n ≥ 1).
When the object reaches Pn , it makes a 90–degree counterclockwise
turn to proceed towards Pn+1 . (Thus each segment, other than the
first, is perpendicular to the preceding segment.) The length of the
first segment is a. For n ≥ 2, the nth segment is s times as long as the
(n − 1)st segment, where s is a fixed number between 0 and 1 (strictly).
If the object keeps moving forever,
(i) what is the farthest distance it ever gets from P0 , and
(ii) what is the distance between P0 and the limiting position P∞ that
the object approaches?
Hint: Investigate the components of the position vector of the object
in an appropriate coordinate system.
32 1. VECTORS AND THE SPACE GEOMETRY

Solution: First, suppose that the initial point P0 is at the origin


and its initial trajectory is along the positive x-axis. Notice that, since
0 < s < 1, we have for all n ≥ 1
|Pn−1 Pn | > |Pn Pn+1 |
(i) Consider the first 6 points P0 , P1 , ...P5. If you were to plot these out,
you would notice that they form a spiral. The point object, starting
from P0 , moves to P1 , then turns and moves to P2 . Its distance from
the origin is now greater than P0 P1 . Then it turns and moves to P3 (in
the opposite direction of the original direction), but since the lengths
are decreasing, it will not reach the y-axis, but will get closer to it.
Hence the distance to the origin has decreased. If you continue in this
fashion, you will see that the maximum distance √ is at P2 . We know
that P2 is at (a, sa), so the maximum distance is a 1 + s2 .

(ii) We must make two infinite series: one for the x-component and one
for the y component. Notice that the object alternates traveling along
the x and y directions. So one sum will be composed of odd powers of
s and one will be composed of even powers of s. We can see that since
P1 = (a, 0), the x-component sum will involve even powers of s. We
also know that each sum will be an alternating series. We can continue
finding the first couple coordinates as follows
P0 = (0, 0)
P1 = (a, 0)
P2 = (a, sa)
P3 = (a − s2 a, sa)
P4 = (a − s2 a, sa − s3a)
P5 = (a − s2 a + s4 a, sa − s3 a)
P6 = (a − s2 a + s4 a, sa − s3 a + s5 a)

I will only consider the points with even subscripts, to help simplify
the formulation of the series. This will not matter in the limiting case.
So we have that
n−1
X n−1
X
k 2k
P2n = (a (−1) s , a (−1)k s(2k+1) )
k=0 k=0
n−1
X n−1
X
2 k
P2n = (a (−s ) , as (−s2 )k )
k=0 k=0
2. VECTORS IN SPACE 33

And the distance from the origin is simply


v
u n−1 n−1
u X X
d(n) = (a
t 2 k 2
(−s ) ) + (sa (−s2 )k )2
k=0 k=0
n−1
X √
=a (−s2)k 1 + s2
k=0
1 − (−s2)n √
=a 1 + s2
1 − (−s2 )
1 − (−s2)n √
=a 1 + s2
1 + s2
by recalling that
n−1
X 1 − rn
ark = a
k=0
1−r
n
an if |r| < 1, then r → 0 as n → ∞. So,
a
lim d(n) = √
n→∞ 1 + s2
34 1. VECTORS AND THE SPACE GEOMETRY

3. The Dot Product


1–5. Find the dot product a · b for the given vectors a and b.
1. a = h1, 2, 3i and b = h−1, 2, 0i

Solution: Recall that the dot product between two vectors


u = hux , uy , uz i and v = hvx , vy , vz i is
u · v = ux v x + uy v y + uz v z
So
a · b = 1(−1) + 2(2) + 3(0) = 3
−→ −−→
2. a = AB and b = BC where A = (1, −2, 1), B = (2, −1, 3),
and C = (1, 1, 1).

Solution: First we have that


−→ −−→ −→
AB = OB − OA = h2, −1, 3i − h1, −2, 1i = h1, 1, 2i

−→ −→ −−→
BC = OC − OB = h1, 1, 1i − h2, −1, 3i = h−1, 2, −2i
So
a · b = 1(−1) + 1(2) + 2(−2) = −3
3. a = ê1 + 3ê2 − ê3 and b = 3ê1 − 2ê2 + ê3

Solution: Recall that eˆ1 , eˆ2, eˆ3 are the unit direction vec-
tors parallel to the x, y, z axes, respectively. So, a = h1, 3, −1i
and b = h3, −2, 1i, and
a · b = 1(3) + 3(−2) − 1(1) = −4
4. a = u1 +3u2 −u3 and b = 3u1 −2u2 +u3 where un , n = 1, 2, 3,
are orthogonal vectors and kun k = n.

Solution: Recall that if two vectors u and v are orthogo-


nal then u · v = 0 and u · u = kuk2. So
a · b = 1(3)ku1 k2 + 3(−2)ku2 k2 − 1(1)ku3 k2 = 3 − 6(4) − 9 = −30
5. a = 2c − 3d and b = c + 2d if c is a unit vector that makes
the angle π/3 with the vector d and kdk = 2

Solution: Recall that, given two vectors u and v, another


way to write the dot product of them is
u · v = cos(θ)|uk|vk
3. THE DOT PRODUCT 35

where θ is the smallest positive angle between u and v. So we


have that
c · d = cos(π/3)(2) = 1
since the angle between c and d is π/3, c is a unit vector so
kck = 1, and kdk = 2 as given. Thus,
a · b = 2c · c + 4c · d − 3d · c − 6d · d
= 2kck2 + c · d − 6kdk2
= 2 + 1 − 6(4) = −21
6. Let a be a nonzero vector. Show that 0 is the only vector that is
both parallel and perpendicular to a.

Solution: Let a be a nonzero vector. Suppose u is a vector that


is both parallel and perpendicular to a. Since u is parallel to a, there
exists a real number s such that u = sa. Since u is perpendicular to
a, cos(θ) = 0 ⇒ a · u = 0. But a · u = a · sa = s(a · a) = skak2 . So
skak2 = 0. Since a is a nonzero vector, kak =
6 0. Thus s = 0, and u = 0

7–9. Are the given vectors a and b orthogonal, parallel, or neither?


7. a = h5, 2i and b = h−4, −10i

Solution: To determine the relationship between a and b,


we must investigate the angle θ between them. If two vectors
are orthogonal, θ = π/2. If two vectors are parallel, θ = 0. If
neither, then θ 6= 0, π/2. We have that
a·b
θ = arccos( )
kakkbk
a·b = 5(−4) + 2(−10) = −40
√ √
kak = 52 + 22 = 29
p √
kbk = (−4)2 + (−10)2 = 2 29
−40 20
θ = arccos( √ √ ) = arccos(− )
29(2 29) 29
Therefore a and b are neither parallel nor orthogonal.

8. a = h1, −2, 1i and b = h0, 1, 2i

Solution: To determine the relationship between a and b,


we must investigate the angle θ between them. If two vectors
36 1. VECTORS AND THE SPACE GEOMETRY

are orthogonal, θ = π/2. If two vectors are parallel, θ = 0. If


neither, then θ 6= 0, π/2. We have that
a·b
θ = arccos( )
kakkbk
a · b = 1(0) − 2(1) + 1(2) = 0
0
θ = arccos( ) = π/2
kakkbk
Therefore a and b are orthogonal.

9. For what values of b are the vectors h−6 , b , 2i and hb , b2 , bi orthog-


onal?

Solution: The two given vectors will be orthogonal if their dot prod-
uct vanishes. Hence,
h−6, b, 2i · hb, b2, bi =0
−6(b) + b(b2) + 2(b) =0
b(b2 − 4) =0
b = ±2, 0

10. Use the dot product to find all unit vectors that are perpendicular
to the vectors h1, 1, −2i and h1, −2, 4i.

Solution: Suppose that u = ha, b, ci is perpendicular to both of the


given vectors. Then,
u · h1, 1, −2i = 0 ⇐⇒ a + b − 2c = 0
u · h1, −2, 4i = 0 ⇐⇒ a − 2b + 4c = 0
which gives a system of two linear equations in three unknowns. We
can solve for a, b as follows:
(
a + b = 2c
a − 2b = −4c
By subtracting the first from the second,
(
a + b = 2c
−3b = −6c ⇐⇒ b = 2c
3. THE DOT PRODUCT 37

And by substituting b = 2c into the first,


(
a=0
b = 2c
So u = h0, 2c, ci = ch0, 2, 1i. Now we must determine what c is so that
u is a unit vector. If u is a unit vector then kuk = 1, so

|c| 02 + 22 + 12 = 1

|c| 5 = 1
1
c = ±√
5
Therefore u = ± √15 h0, 2, 1i.

11. Find the angle at the vertex A of a triangle ABC for A = (1, 0, 1),
B = (1, 2, 3), and = C(0, 1, 1). Express the answer in radians.

Solution: By the geometrical properties of the dot product, the angle


θ between two vectors a and b is determined by the equation
a·b
cos θ =
kakkbk
−→ −→
Setting a = AB = h0, 2, 2i and b = AC = h−1, 1, 0i, it follows that
a·b = √
0 · (−1) + 2 · 1 + 2√· 0 = 2
2 1
kak = p02 + 22 + 22 = 2 2√ ⇒ cos θ = √ =
2( 2) 2 2
kbk = (−1)2 + 12 + 02 = 2
So, θ = π/3.

12. Find the cosines of the angles of a triangle ABC for A = (0, 1, 1),
B = (−2, 4, 3), and C = (1, 2, −1).

Solution: We can easily find the cosines of the angles by represent-


ing finding vectors parallel to the sides of the triangle, and using the
geometric interpretation of the dot product. So,
−→
AB = h−2, 4, 3i − h0, 1, 1i = h−2, 3, 2i
−−→
BC = h1, 2, −1i − h−2, 4, 3i = h3, −2, −4i
−→
AC = h1, 2, −1i − h0, 1, 1i = h1, 1, −2i
38 1. VECTORS AND THE SPACE GEOMETRY

and their magnitudes are


−→ p √
kABk = (−2)2 + 32 + 22 = 17

−→ p √
kBCk = 32 + (−2)2 + (−4)2 = 29
−→ p √
kACk = 12 + 12 + (−2)2 = 6

So the cosines of the angles of triangle ABC are as follows:


−→ −→
AB · AC −2(1) + 3(1) + 2(−2) 3
cos α = −→ −→ = √ √ = −√ √
kABkkACk 6 17 6 17
−→ − −→
BA · BC 2(3) − 3(−2) − 2(−4) 20
cos β = −→ −→ = √ √ =√ √
kBAkkABk 17 29 17 29
−→ −−→
CA · CB −1(−3) − 1(2) + 2(4) 9
cos γ = −→ − −→ = √ √ =√ √
kCAkkCBk 6 29 6 29

where α, β, γ are the angles at vertex A, B, and C respectively. Note


that the each vector in a dot product start at the vertex where the
angle is. This is to correctly find the angle they enclose. Otherwise
you might get the obtuse angle when it is actually acute, and vice verse.

13. Consider a triangle whose one side is a diameter of a circle and


the vertex opposite to this side is on the circle. Use vector algebra to
prove that any such triangle is right-angled. Hint: Consider position
vectors of the vertices relative to the center of the circle.

Solution: Consider a circle centered at the origin with radius R.


The triangle in question can be formed by three points as follows:
A = (R, 0), B = (−R, 0), and C = (R cos θ, R sin θ) for 0 ≤ θ < 2π.
−→ − −→
We wish to prove that AC ⊥ BC. To show this, we need only demon-
strate that their dot product vanishes. So,
−→ −→ −→
AC = OC − OA = hR cos θ − R, R sin θ

−→ −→ −→
BC = OC − OA = hR cos θ + R, R sin θ
−→ − −→
AC · BC = (R cos θ − R)(R cos θ + R) + (R sin θ)2
= (R cos θ)2 − R2 + (R sin θ)2
= R2 − R2 = 0
3. THE DOT PRODUCT 39

14. Let a = sû + v̂ and b = û + sv̂ where the angle between unit
vectors û and v̂ is π/3. Find the values of s for which the dot product
a · b is maximal, minimal, or zero if such values exist.

Solution: First let us compute û · v̂. By the geometric properties


of the dot product,
cos θ
û · v̂ =
kûkkv̂k
cos π/3 1
= =
(1)(1) 2
Note that since the dot product is commutative û · v̂ = v̂ · û, and that
û · û = kûk. Next we will compute a · b as follows
a · b = (sû + v̂) · (û + sv̂)
= sû · û + sû · sv̂ + v̂ · û + v̂ · sv̂
= skûk + (s2 + 1)û · v̂ + skv̂k
1
= s + (s2 + 1)( ) + s
2
1 2 1 1 3
= (s + 4s + 1) = (s2 + 4s + 4 − 3) = (s + 2)2 −
2 2 2 2

The function f(s) = 12 (s+2)2 − 23 is a parabola opening upwards. Hence


there is no maximum value. There is a single minimum occurring at
the vertex, when s = −2. The zeroes occur when f(s) = 0,
1 3
(s + 2)2 − =0
2 2
(s + 2)2 = 3

s+2 =± 3

s = −2 ± 3
15. Consider a cube whose edges have length a. Find the angle be-
tween its largest diagonal and any edge adjacent to the diagonal.

Solution: Consider a cube of side length a situated in the first octant,


with one corner at the origin and side lengths parallel to the x, y, z axes.
The largest diagonal (of four, actually, but since this is a cube they are
all equivalent) is the one connecting the corner at (0, 0, 0) with the cor-
ner at (a, a, a). We can find the angle between any edge adjacent to the
diagonal by considering the dot product between the vectors parallel to
these. Clearly a vector parallel to the largest diagonal is ha, a, ai. Any
40 1. VECTORS AND THE SPACE GEOMETRY

adjacent edge is simply an edge of the cube that meets the diagonal at
a corner. Notice that the angles must be the same, first because the
problem asks for ”the” angle, but also because you can rotate the cube
around its diagonal and still have a cube. So I will consider the angle
between the largest diagonal and the edge along the x-axis. A vector
parallel to this edge is ha, 0, 0i. So we have that

ha, 0, 0i · ha, a, ai
θ = arccos( )
kha, 0, 0ikkha, a, aik
a2
= arccos( √ )
(a)(a 3)
1
= arccos( √ )
3

16. Consider a parallelepiped with adjacent sides a = h1, −2, 2i,


b = h−2, −2, 1i, and c = h−1, −1, −1i (see the definition of a paral-
lelepiped in Study Problem ??). It has four vertex-to-opposite-vertex
diagonals. Express them in terms of a, b, and c and find the largest
one. Find the angle between the largest diagonal and the adjacent sides
of the parallelepiped.

Solution: There are four diagonals we must find. Recall that a par-
allelogram formed from two non-parallel vectors u and v has diagonal
u + v, and recall that the faces of a parallelpiped are parallelograms.
Notice that each diagonal can be formed as the sum/difference of a di-
agonal of a face and the other vector which does not make up the face.
It may be help to form a sketch. For example, consider a face formed
from the vectors a and b. The diagonal of this face is a + b, and the
other vector that does not make up this face is c. One of the diagonals
in question then is d1 = a + b + c, and another is d2 = a + b − c. Next
consider the face formed by the vectors a and c. Then the diagonal of
this face is a + c. So the third diagonal in question is d3 = a + c − b.
Finally, consider the face formed by the vectors b and c. Then the
diagonal of this face is b + c. So the fourth diagonal in question is
d4 = b + c − a. Clearly, the largest diagonal is d1 (subtracting any
vector will decrease the magnitude of the resultant).

Next we must find the angle between the largest diagonal, d1 = h−2, −5, 2i,
and the adjacent sides a, b, and c. First let us find the magnitudes of
3. THE DOT PRODUCT 41

each of these:
p √
kd1 k = (−2)2 + (−5)2 + 22 = 33
p
kak = 12 + (−2)2 + 22 = 3
p
kbk = (−2)2 + (−2)2 + 12 = 3
p √
kck = (−1)2 + (−1)2 + (−1)2 = 3
So the angles between them are:
d1 · a 12 4
α = arccos( ) = arccos( √ ) = arccos( √ )
kd1 kkak 3 33 33
d1 · b 16
β = arccos( ) = arccos( √ )
kd1 kkbk 3 33
d1 · c 5 5
γ = arccos( ) = arccos( √ √ ) = arccos( √ )
kd1 kkck 3 33 3 11
Where α, β, γ represent the angles between d1 and a, d1 and b, and
d1 and c, respectively.

17. Let a = h1, 2, 2i. For the vector b = h−2, 3, 1i, find the scalar
and vector projections of b onto a and construct the orthogonal de-
composition b = b⊥ + bk relative to a.

Solution: Recall that the scalar projection of b onto a is given by


b·a
b|| = .
kak
And the vector projection of b onto a is given by
b||
b|| = b|| â = a
kak
And so,
−2(1) + 3(2) + 1(2) 6
b|| = √ = √ =2
2 2
1 +2 +2 2 9
2
b|| = a
3
2
b⊥ = b − b|| = b − a
3
−→ −−→
18. Find the scalar and vector projections of AB onto BC if A =
(0, 0, 4), B = (0, 3, −2), and C = (3, 6, 2).
42 1. VECTORS AND THE SPACE GEOMETRY

−→ −−→
Solution: First we must compute AB and BC as follows:
−→ −−→ −→
AB = OB − OA = h0, 3, −2i − h0, 0, 4i = h0, 3, −6i
−−→ −→ −−→
BC = OC − OB = h3, 6, 2i − h0, 3, −2ii = h3, 3, 4i

And so
−→ − −→
AB · BC 0(3) + 3(3) − 6(4) 15
AB || = −
−→ = √ = −√
kBCk 32 + 32 + 42 34
−→ AB || −−→ 15
AB || = − −→ BC = − 34 h3, 3, 4i
kBCk
−→ −−→
are the scalar and vector projections of AB onto BC, respectively.

19. Find all vectors that have a given length a and make an angle
π/3 with the positive x axis and the angle π/4 with the positive z axis.

Solution: Recall that the direction of a vector in space is uniquely


determined by three angles, the so-called direction angles α, β, and γ.
These angles satisfy the following equation:
cos2 α + cos2 β + cos2 γ = 1
We are given two direction angles already. Set α = π/3 and γ = π/4,
then
cos2 π/3 + cos2 β + cos2 π/4 = 1
1 1
+ cos2 β + =1
4 2
1
cos2 β =
4
1
cos β = ±
2

Further, recall that a unit vector with these three direction angles is of
the form
û = hcos α, cos β, cos γi.
Since we require that the vector must be of length a, we simply have
that
1 1 1
u = aû = ah , ± , √ i
2 2 2
3. THE DOT PRODUCT 43

20. Find the components of all unit vectors û that make an angle φ with
the positive z axis. Hint: Put û = av̂ + bê3, where v̂ is a unit vector in
the xy plane. Find a, b, and all v̂ using the polar angle in the xy plane.

Solution: First recognize that û, av̂, and bê3 form a triangle with
hypotenuse 1 and side lengths a, b where the side with length a is op-
posite of φ. This means that tan φ = a/b. Thus a = b tan φ. Next we
have that û is a unit vector, so

a 2 + b2 = 1
a 2 + b2 = 1
b2 tan2 φ + b2 = 1
b2(tan2 φ + 1) = 1
b2(sec2 φ) = 1
b2 = cos2 φ
b = ± cos φ

As it turns out, later we will be able to discard the negative solution.


In any case, we can now express a as
a = ± cos φ tan φ = ± sin φ
Next we must find v̂. We can write it in polar form as
v̂ = hcos θ, sin θ, 0i
where θ is taken to be an angle measured counterclockwise from the
positive x axis in the xy plane (the standard polar angle). So we have
that
û = h± cos θ sin φ, ± sin θ sin φ, 0i + h0, 0, cos φi
= h± cos θ sin φ, ± sin θ sin φ, cos φi
Consider the negative solution for b,
û0 = h− cos θ0 sin φ, − sin θ0 sin φ, cos φi
Next, consider the transformation θ0 = θ + π. Then we will have
û0 = hcos θ sin φ, sin θ sin φ, cos φi
So we may discard the negative solution, and simply have
û = hcos θ sin φ, sin θ sin φ, cos φi
You will learn later that this is the conversion from spherical coordi-
nates to Cartesian coordinates.
44 1. VECTORS AND THE SPACE GEOMETRY

21. If c = kakb + kbka, where a and b are non zero vectors, show that
c bisects the angle between a and b. Hint: compare the angle between
c and a to the angle between c and b.

Solution: First we will compute the angle α between a and c. By


the geometric properties of the dot product, we have that
a·c
α = arccos
kakkck
kak(a · b) + kbkkak2
= arccos
kakkck
a · b + kbkkak
= arccos
kck

Similarly, we must compute the angle β between b and c


b·c
β = arccos
kbkkck
kbk2 kak + kbk(b · a)
= arccos
kbkkck
kbkkak + b · a
= arccos
kck

Evidently α = β, so c bisects the angle between a and b.

22. A rhombus is a parallelogram with sides of equal length. Prove


that the diagonals of a rhombus meet at right angles.

Solution: The adjacent sides of a rhombus can be formed from two


vectors a and b where kak = kbk. The two diagonals of the rhombus
are d1 = a + b and d2 = b − a. Convince yourself of this by sketching
a picture, if necessary. Next we have that
d1 · d2 = a · b − kak2 + kbk2 − b · a = 0 ⇒ d1 ⊥ d2
since kak = kbk.

23. Consider a parallelogram with adjacent sides of length a and b.


If d1 and d2 are the lengths of the diagonals, prove the parallelogram
law: d21 + d22 = 2(a2 + b2). Hint: Consider the vectors a and b that
are adjacent sides of the parallelogram and express the diagonals via a
3. THE DOT PRODUCT 45

and b. Use the dot product to evaluate d21 + d22 .

Solution: The adjacent sides of a parallelogram can be formed from


two vectors a and b, as stated in the hint. The two diagonals of the
parallelogram are d1 = a + b and d2 = b − a. Convince yourself of
this by sketching a picture, if necessary. Notice that
d1 · d1 = d21
On the other hand, we also have that
d1 · d1 = a2 + 2a · b + b2
So d21 = a2 + 2a · b + b2 .
Similarly for d2 we have that
d2 · d2 = d22
On the other hand, we also have that
d2 · d2 = a2 − 2a · b + b2
So d22 = a2 − 2a · b + b2 . Thus
d21 + d22 = a2 + a2 + 2a · b − 2a · b + b2 + b2 = 2(a2 + b2 )
24. Consider a right-angled triangle whose adjacent sides at the right
angle have lengths a and b. Let P be a point in space at a distance c
from all three vertices of the triangle (c ≥ a/2 and c ≥ b/2). Find the
angles between the line segments connecting P with the vertices of the
triangle. Hint: Consider vectors with the initial point P and terminal
points at the vertices of the triangle.

Solution: The solution I have is much easier than doing anything


with dot products. Notice that the solid formed by the vertices of
the triangle and P is a triangular pyramid. So each face is a trian-
gle. Consider one such triangle (that is not the base). Two√sides are
of length c and one is of length l, where l is either a, b, or a2 + b2 .
Call the angle between the sides of length c θ. Draw a line from P to
the side of length l, which will be a perpendicular bisector. We now
have two right triangles with side lengths l/2, c and another irrelevant
one, and an angle θ/2. Consider sin θ/2. The opposite side has length
l/2 and the hypoteneuse has length c. So sin θ/2 = l/2c. Therefore
θ/2 = arcsin l/2c, and θ = 2 arcsin l/2c. So the angles
√ in question are
α = 2 arcsin a/2c, β = 2 arcsin b/2c, γ = 2 arcsin a2 + b2 /2c

25. Show that the vectors u1 = h1, 1, 2i, u2 = h1, −1, 0i, and u3 =
h2, 2, −2i are mutually orthogonal. For a vector a = h4, 3, 4i find the
46 1. VECTORS AND THE SPACE GEOMETRY

scalar orthogonal projections of a onto ui , i = 1, 2, 3, and the numbers


si such that a = s1u1 + s2 u2 + s3 u3.

Solution: To show that uj are mutually orthogonal for j = 1, 2, 3,


we must show that their dot products vanish as follows:

u1 · u2 = 1(1) + 1(−1) + 2(0) = 1 − 1 = 0


u2 · u3 = 1(2 + −1(2) + 0(−2) = 2 − 2 = 0
u1 · u3 = 1(2) + 1(2) + 2(−2) = 2 + 2 − 4 = 0

So they are mutually orthogonal. Next consider a · u1 . This will be


s1ku1 k2 + 0 + 0. In general,

a · uj
sj =
kuj k2

for j = 1, 2, 3. So,

4(1) + 3(1) + 4(2) 4+3+8 5


s1 = √ 2 = =
12 + 12 + 22 6 2
4(1) + 3(−1) + 4(0) 4−3 1
s2 = p 2 = =
12 + (−1)2 + 02 2 2
4(2) + 3(2) + 4(−2) 8+6−8 1
s3 = p 2 = =
22 + 22 + (−2)2 12 2

Recall that the the scalar projection of y onto x is

x·y
y|| = ,
kxk

which we can rewrite as


x·y
y|| = kxk
kxk2

Let y = a and x = uj . Then the scalar projection of a onto uj is

uj · a
a|| = kuj k = sj kuj k
kuj k2
3. THE DOT PRODUCT 47

So the necessary scalar projections are



5√ 5 6
s1ku1 k = 6=
2 √2
1√ 2
s2ku2 k = 2=
2 2
1√ √
s2ku3 k = 12 = 3
2

26. For two nonzero vectors a and b find all vectors coplanar with a
and b that have the same vector projection onto a as the vector b.
Express these vectors in terms of a and b.

Solution: First recall that the vector projection of b onto a is uniquely


determined by
a·b
b|| =
kak
If a is a given vector, then all we can change in this equation is a · b.
So, suppose v is a vector with the same scalar projection onto a as b.
Then,
a·v a·b
= ⇒ a · v = a · b ⇒ a · (v − b) = 0
kak kak
So v − b is orthogonal to a. This means that v − b is parallel to b⊥ ,
so
v − b = tb⊥ ⇒ v = b + tb⊥
for all t ∈ R.

27. A point object traveled 3 meters from a point A in a particular


direction, then it changed the direction by 60◦ and traveled 4 meters,
and then it changed the direction again so that it was traveling at 60◦
with each of the previous two directions. If the last stretch was 2 me-
ters long, how far from A is the object?

Solution: Let A, B, C, D be the four points the object stops or


changes direction at. Then we have that
−→ −−→ −−→
kABk = 3, kBCk = 4, kCDk = 2
−−→
and we are being asked to find kADk. We are also given that the object
travels in one direction from A, then changes direction at B and travels
48 1. VECTORS AND THE SPACE GEOMETRY

60◦ from the original direction. So


−→ −−→ −→ −−→
AB · BC = kABkkBCk cos θ = 3(4)(cos 60◦ ) = 6.
The object then changes direction again at C and travels 60◦ from both
−→ −−→
of the previous directions (i.e., AB and BC). So
−→ −−→ −→ −−→
AB · CD = kABkkCDk cos θ = 3(2)(cos 60◦ ) = 3
−−→ −−→ −−→ −−→
BC · CD = kBCkkCDk cos θ = 4(2)(cos 60◦ ) = 4

Recall that kuk2 = u · u for a vector u. So,


−−→ −−→ −−→
kADk2 = AD · AD
−→ −−→ −−→ −→ −−→ −−→
= (AB + BC + CD)(AB + BC + CD)
−→ −−→ −−→ −→ −
−→ −→ −−→ −−→ −−→
= kABk2 + kBCk2 + kCDk2 + 2AB · BC + 2AB · CD + 2BC · CD
= 32 + 42 + 22 + 2(6 + 3 + 4)
= 55
−−→ √
kADk = 55

28. Two balls of the same mass m are connected by a piece of rope of
length h. Then the balls are attached to different points on a horizontal
ceiling by a piece of rope with the same length h so that the distance
L between the points is greater than h but less than 3h. Find the
equilibrium positions of the balls and the magnitude of tension forces
in the ropes.

Solution: The setup involves two balls and three ropes, where the
ropes and the ceiling make a regular trapezoid with the largest base
facing up. The free body diagram for the left ball involves three forces:
Gravity, the tension T1 from the rope connecting the ball to the ceiling,
and the tension T2 connecting the ball to the other ball. Note that Fg
points down, T1 to the up and left, and T2 horizontally to the right.
If the system is in equilibrium, then the vector sum of the forces must
vanish. Let θ denote the angle between the ceiling and T1 . Then we
must have
−T1 sin θ + T2 = 0
T1 cos θ − Fg = 0 ⇒ T1 cos θ = mg
where the first equation is the force balance of the x-components, the
second is the force balance of the y-components, and right and up are
3. THE DOT PRODUCT 49

taken to be the positive x and y directions, respectively.


Next, refer back to the trapezoidal setup. Imagine bringing the hori-
zontal rope of length h up onto the ceiling (keeping it horizontal and
in line with where it previously was), then connect the end points with
the balls. This forms a rectangle within the trapezoid and two right
triangles, with bases 12 (L − h). Look at the left triangle. The an-
gle between the base and the hypotenuse is θ, as defined before. The
hypotenuse is of length h, as stated in the problem (the rope that con-
nects
q the balls to the ceiling is length h), so the vertical side is of length
h2 + ( 21 (L − h))2 . Then cos θ = L−h
2h
. Referring back to our free body
diagram equations, we have that
mg 2mgh
T1 = =
cos θ L−h
q
2mg h2 + ( 21 (L − h)2
T2 = T1 sin θ = mg tan θ =
L−h
are the tensions in the ropes connecting the balls each other and to
the ceiling, respectively (the tensions in the right ball are the same as
those in the left due to symmetry).

29. A cart is pulled up a 20◦ slope a distance of 10 meters by a hori-


zontal force of 30 newtons. Determine the work.

Solution: Recall that the work done by a force F over a distance


d on an object is given by
−−→
W = F · P1 P2
where P1 is the initial point of the object and P2 is the final point, and
|P1 P2 | = d. So
−−→
W = F · P1 P2 = F d cos θ = 30(10) cos 20◦ = 300 cos 20◦ J
30. Two tug boats are pulling a barge against the river stream. One
tug is pulling with the force of magnitude 20 (in some units) and at
the angle 45◦ to the stream and the second with the force of magni-
tude 15 at the angle 30◦ so that the angle between the pulling ropes
is 45◦ + 30◦ = 75◦ . If the barge does not move in the direction of the
stream, what is the drag force exerted by the stream on the barge?
Does the barge move in the direction perpendicular to the stream?

Solution: First we must set up our problem. Suppose the stream


flows in the negative x direction, where positive x is defined to be right
50 1. VECTORS AND THE SPACE GEOMETRY

and positive y is defined to be up. Then we can write the force of the
two ropes and the drag force as follows:
T1 = hT1 cos θ1 , T1 sin θ1 i = h20 cos 45◦ , 20 sin 45◦ i
T2 = hT2 cos θ2 , T2 sin θ2 i = h15 cos(−30◦ ), 15 sin(−30◦ )i
D = h−D, 0i
where θ2 = −30◦ and not 30◦ because we must have that θ1 − θ2 = 75◦ .
Notice that D points in the negative x direction because that is how
the stream is trying to pull the barge. Next we must have that the
x component of the vector sum vanishes (because the barge does not
move in the direction of the stream), so
√ 15 √
20 cos 45◦ + 15 cos(−30◦ ) − D = 0 ⇒ D = 10 2 + 3
2
Furthermore since only the x-component vanishes, the barge moves in
the y direction, which is perpendicular to the direction of the stream
(negative x).

31. A ball of mass m is attached by three ropes of the same length a to


a horizontal ceiling so that the attachment points on the ceiling form
a triangle with sides of length a. Find the magnitude of the tension
force in the ropes.

Solution: Notice that the three ropes and the ceiling form a tri-
angular pyramid. Furthermore, the face is an equilateral triangle with
side lengths a. Each of T1, T2, and T3 have the same magnitude as
the problem is symmetric (rotating the setup by 120◦ does not change
anything). Consider one rope and its tension, say T1 . Construct a line
segment from the ball to the center of the triangular base and another
line from the center of the base to the attachment point of the rope.
Call the angle between the rope and the ceiling θ. Then T1 sin θ de-
scribes the vertical component of T1 . We must find the length of the
base of the triangle, b. We know its hypotenuse is length a and the
base connects the center of an equilateral triangle with sides√ of length
a to a vertex. Evidently, the height of the entire triangle is 23 a. Draw
three lines from each vertex to the opposite side. These lines should be
perpendicular to the opposite side and have length equal to the height.
Observe that we can split one height into two pieces, one of length b and
another of length c. Choose another line of length b, and this will form
a right-angled triangle with interior angles 60◦ and 30◦ . We have that
b = sec(30◦ ) 21 a = √a3 . So the height of the first triangle we were looking
3. THE DOT PRODUCT 51

√ q
a2

at (with base b and hypotenuse a) is h = a 2 − b2 = a 2 − 3
= a √23 .

h √2 .
Then sin θ = a
= 3
Since there are three tension vectors, we have
that
3T1 sin θ = Fg

mg 3
T1 = √
3 2
mg
=√
6
mg
And so T = T1 = T2 = T3 = √ .
6

32. Four dogs are at the vertices of a square. Each dog starts running
toward its neighbor on the right. The dogs run with the same speed
v. At every moment of time each dog keeps running in the direction of
its right neighbor (its velocity vector always points to the neighbor).
Eventually, the dogs meet in the center of the square. When will this
happen if the sides of the square have length a? What is the distance
traveled by each dog? Hint: Is there a particular direction from the
center of the square relative to which the velocity vector of a dog has
the same component at each moment of time?

Solution: First, recognize that the problem is symmetric. Rotat-


ing the setup by 90◦ at any time will result in essentially the same
thing prior to rotation. It is easy, then, to see that at any time, the co-
ordinates of the dogs form a square with the same center as the original
(at t = 0). This can be seen by doing the following: take a line segment
of length d and rotate it around a point by 90◦ so that one endpoint
of the new segment lands on an endpoint of the original segment, and
keep doing this, you will get a square.
With this in mind, the velocity vector of any dog is at an angle of
3π/4 with its position vector. This is because the velocity vector of
a dog will point towards the dog it is chasing (i.e., along the edge of
−−→
the square). Choose a dog. Let the position vector of the dog be OD.
Then the scalar projection of the dog’s velocity vector onto its position
vector is given by
−−→ −−→ √
v · OD vkODk cos 3π/4
v|| = −−→ = −−→ = v 2.
kODk kODk
where v is the speed of the dog, a constant value. This decomposes v
−−→ √
into two vectors. One is along OD with magnitude v 2. The other is
52 1. VECTORS AND THE SPACE GEOMETRY

perpendicular to this, which gives rise to an angular velocity, motion


in a circle. This is seen by considering that the tangent line to a point
−→
P on a circle is orthogonal to CP , from the center C of the circle.
Recall that the velocity vector is parallel to the tangent line at P of
a particle’s motion (dy/dx s(x) = v(x)). So, the true motion of the
dog is the superposition of a circular motion around the center of the
square and a motion along a straight line towards the center. Since the
circular motion does not affect the distance of the dog to the center ,
the time it takes for the dog to reach the center is the same as if the dog
moved
√ only in a straight line (”radial”). The radial distance traveled
is a 2, and so the time taken to reach the center is

a 2 a
T = √ =
v 2 v
So the distance traveled by the dog, given by its velocity multiplied by
the time taken to reach the center, is
a
D=v =a
v
Notice that it is as if the dog is chasing a stationary object.
4. THE CROSS PRODUCT 53

4. The Cross Product


1–7. Find the cross product a × b for the given vectors a and b.
1. a = h1, 2, 3i and b = h−1, 0, 1i

Solution: Recall that the definition of the cross product


u × v for two vectors u = hu1, u2 , u3i and v = hv1, v2, v3i is:
 
ê1 ê2 ê3      
u2 u3 u1 u3 u1 u2
u×v = det u1 u2 u3 = hdet
  , − det , det i
v2 v3 v1 v3 v1 v2
v1 v2 v3

Thus, by the definition of the cross product,


 
ê1 ê2 ê3      
23 1 3 1 2
a×b = det 1 2 3 = hdet
  , − det , det i = h2, −4, 2i
01 −1 1 −1 0
−1 0 1

2. a = h1, −1, 2i and b = h3, −2, 1i

Solution: Given the definition above, we can find the cross


product as
 
ê1 ê2 ê3      
−1 2 1 2 1 −1
a×b = det  1 −1 2  = hdet , − det , det i = h3, 5, 1i
−2 1 31 3 −2
3 −2 1

3. a = ê1 + 3ê2 − ê3 and b = 3ê1 − 2ê2 + ê3

Solution: Given the definition above, we can find the cross


product as
 
ê1 ê2 ê3      
3 −1 1 −1 1 3
a×b = det  1 3 −1 = hdet , − det , det i = h1, −4, −11i
−2 1 3 1 3 −2
3 −2 1

4. a = 2c − d, b = 3c + 4d where c × d = h1, 2, 3i.

Solution: Recall the following properties of the cross prod-


uct:
(1) For any vector u, we have that u × u = 0
(2) For any vectors u and v, we have that u × v = −v × u
(3) It is left and right distributive
(4) For any vectors su and v, where s ∈ R is a scalar, we have
54 1. VECTORS AND THE SPACE GEOMETRY

(2u) × v = 2(u × v). Thus,


a × b = (2c − d) × (3c + 4d)
= 2c × (3c + 4d) − d × (3c + 4d)
= 2c × 3c + 2c × 4d − d × 3c − d × 4d
= 8c × d + 3c × d
= 11h1, 2, 3i
5. a = u1 + 2u2 + 3u3 and b = u1 − u2 + u3 where the vectors
ui , i = 1, 2, 3, are mutually orthogonal, have the same length
3, and u1 × u2 = 3u3 . Express the answer as a linear combi-
nation of ui (not their cross products).

Solution: Recall that a × (b × c) = b(a · c) − c(a · b).


Therefore we have
u1 × 3u3 = u1 × (u1 × u2) = u1 (u1 · u2 ) − u2 (u1 · u1 )
= −u2 ku1k2 = −9u2
9
⇒ u1 × u3 = − u2 = −3u2
3
Similarly, u2 × u3 = 3u1

Using the properties of the cross product,


a × b = u1 × u1 + u1 × (−u2 ) + u1 × u3 + 2u2 × u1 + 2u2 × (−u2)
+2u2 × u3 + 3u3 × u1 + 3u3 × (−u2 ) + 3u3 × u3
= −u1 × u2 + u1 × u3 + 2u2 × u1 + 2u2 × u3 + 3u3 × u1 − 3u3 × u2
= −u1 × u2 − 2u1 × u2 + u1 × u3 − 3u1 × u3 + 2u2 × u3 + 3u2 × u3
= −3u1 × u2 − 2u1 × u3 + 5u2 × u3
= −3(3u3 ) − 2(−3u2 ) + 5(3u1 )
= 3(5u1 + 2u2 − 3u3 )
It turns out that this is the same as h1, 2, 3i × h1, −1, 1i mul-
tiplied by 3. This is no coincidence.

6. a has length 3 units, lies in the xy plane, and points from the
origin to the first quadrant at the angle π/3 to the x axis and
b has length 2 units and points from the origin in the direction
of the z axis.

Solution: From the given information, we can write a in


4. THE CROSS PRODUCT 55


polar form as a = 3hcos π/3, sin π/3, 0i = 23 h1, 3, 0i and we
can write b as b = 2h0, 0, 1i. So,
 √ 
 
ê1 √
ê2 ê3 √   

3 3 0 1 0 1 3
a×b = ∗2 det  1 3 0  = 3hdet , − det , det i = 3h 3, −1, 0i
2 0 1 0 1 0 0
0 0 1
7. a and b point from the origin to the second and first quad-
rant of the xy plane, respectively, so that a makes the angle
15◦ with the y axis and b makes the angle 75◦ with the x axis,
and kak = 2, kbk = 3.

Solution: If a makes an angle of 15◦ with the y axis and


is in the second quadrant, then it makes an angle of 105◦
with the positive x axis. So the angle between a and b is
105◦ − 75◦ = 30◦ . Recall that for vectors u and v we have that
ku × vk = kukkvk sin θ
But a and b both lie in the xy plane. So their cross product
must be parallel to the z axis, i.e.
a × b = h0, 0, ki = kh0, 0, 1i
for some scalar k. Clearly, ka × bk = k. On the other hand,
ka × bk = kakkbk sin θ = 2 ∗ 3 sin 30◦ = 3. So k = 3, and
a × b = 3h0, 0, 1i.

8. Let a = h3, 2, 1i, b = h−2, 1, −1i, and c = h1, 0, −1i. Find


a × (b × c), b × (c × a), and c × (a × b). Verify the Jacobi iden-
tity.

Solution: Recall the ”bac-cab” rule, as follows:


a × (b × c) = b(a · c) − c(a · b)
The other cross products can be calculated similarly, by permuting the
vectors in the formula above. If a × (b × c) → b × (c × a), then
we have moved each vector to the left once, and moved the front one
to the back. The effect of this is to permute the vectors in the first
term of the right-hand side to the right once and to permute the vec-
tors in the second term to the left once. I.e, b(a · c) → c(b · a) and
c(a · b) → a(b · c). If you have trouble remembering this, look at what
happens to the first vector in the LHS when you permute the LHS. In
this case a turns into b. That means in the RHS, wherever a appears,
b must now appear. Shift the vectors accordingly. Try to practice by
56 1. VECTORS AND THE SPACE GEOMETRY

determining what c × (a × b) should be before continuing.

We therefore have the following identities:


a × (b × c) = b(a · c) − c(a · b)
b × (c × a) = c(b · a) − a(b · c)
c × (a × b) = a(c · b) − b(c · a)
And so,
a × (b × c) = b(3(1) + 2(0) + 1(−1)) − c(3(−2) + 2(1) + 1(−1)) = 2b + 5c = h1, 2, −7i
b × (c × a) = c(−2(3) + 1(2) − 1(1)) − a(−2(1) + 1(0) − 1(−1)) = −5c + a = h−2, 2, 6i
c × (a × b) = a(1(−2) + 0(1) − 1(−1)) − b(1(3) + 0(2) − 1(1)) = −a − 2b = h1, −4, 1i
Thus,
a × (b × c) + b × (c × a) + c × (a × b) = 0
9. Let a be a unit vector orthogonal to b and c. If c = h1, 2, 2i, find
the length of the vector a × [(a + b) × (a + b + c)]

Solution: Recall that if u is a unit vector, then u·u = kuk2 = 12 = 1.


Furthermore, it u and v are two perpendicular vectors, then u · v = 0.
Thus, using the ”bac-cab” rule on a × [(a + b) × (a + b + c)] yields
a × [(a + b) × (a + b + c)] = (a + b)(a · (a + b + c)) − (a + b + c)(a · (a + b)
= (a + b)(a · a + a · b + a · c) − (a + b + c)(a · a + a · b)
= (a + b)(1) − (a + b + c)(1)
= −c
So p
ka × [(a + b) × (a + b + c)]k = (1)2 + (2)2 + (2)2 = 3
10. Given two nonparallel vectors a and b, show that the vectors a,
a × b and a × (a × b) are mutually orthogonal, and, hence, form an
orthogonal basis in space.

Solution: By the definition of the cross product, a is orthogonal


to a × b and a × (a × b). Thus we must only verify that a × b and
a × (a × b) are orthogonal. Certinaly this is also true, since for any
two nonparallel vectors u and v, u × v will be orthogonal to both u
and v. Since a and b are nonparallel, their cross product is nonzero
and orthogonal to both vectors. So a × b is not parallel to a (0 is the
only vector simultaneously parallel and orthogonal to any vector). So
a × (a × b) is orthogonal to a × b
4. THE CROSS PRODUCT 57

11. Suppose a lies in the xy plane, its initial point is at the origin,
and its terminal point is in first quadrant of the xy plane. Let b be
parallel to ê3. Use the right-hand rule to determine whether the angle
between a × b and the unit vectors parallel to the coordinate axes lies
in the interval (0, π/2) or (π/2, π) or equals π/2.

Solution: By the right hand rule, a × b will be in the xy plane


in the fourth quadrant. Let α, β, and γ be the angles between a × b
and ê1, ê2, and ê3. Clearly, α ∈ (0, π/2). Since a × b is in the fourth
quadrant, it is closer to the negative y-axis than the positive y-axis,
thus β ∈ (π/2, π). Since a × b is orthogonal to b, γ = π/2.

12. If vectors a, b, and c have the initial point at the origin and
lie, respectively, in the positive quadrants of the xy, yz, and xz planes,
determine the octants in which the pairwise cross products of these vec-
tors lie by specifying the signs of the components of the cross products.

Solution: To denote the sign of a component, I will simply write


the sign in the component. Thus, a vector with positive x and z com-
ponents, but a negative y component will be written as h+, −, +i.

First examine a × b. Note that a is h+, +, 0i and b is h0, +, +iBy


the right hand rule a × b will be h+, −, +i. This can be seen by con-
sidering the ”dot product” as follows:
a · (a × b) = h+, +, 0i · h+, −, +i = +(+) + +(−) + 0(+)
It is feasible to make this equal 0, with appropriate choices for the
components. But couldn’t we switch the + and - in the cross product
and also achieve this? To refute this, investigate the ”dot product”
with b as follows:
b · (a × b) = h0, +, +i · h+, −, +i = 0(+) + +(−) + +(+)
If the x and y component signs of the cross product are reversed, then
the ”dot product” with b will always be positive. It follows, in general,
that the component two of a, b, and c share must be positive/negative
in the cross product, with the other two components the opposite sign.
Thus a × c will be h+, −, −i and b × c will be h+, +, −i, by the right-
hand rule and the above discussion.

13. If a, b, and c are coplanar vectors, find (a × b) × (b × c).

Solution: If a, b, and c are coplanar, then each of a × b and b × c


58 1. VECTORS AND THE SPACE GEOMETRY

will be perpendicular to the plane a, b, and c span. Therefore a × b


is parallel to b × c, and their cross product will be 0.

14. Find a unit vector perpendicular to the vectors ê1 + ê2 − 2ê3
and ê1 − 2ê2 + 4ê3.

Solution: A unit vector perpendicular to the given vectors will sim-


ply be a unit vector parallel to their cross product. Let a = h1, 1, −2i
and b = h1, −2, 4i. Then
 
ê1 ê2 ê3      
1 −2 1 −2 1 1
a×b = det  1 1 −2 = hdet , − det , det i = h0, −6, −3i
−2 4 1 4 1 −2
1 −2 4
Let u = a × b. Then
p √
kuk = 02 + (−6)2 + (−3)2 = 3 5
Thus a unit vector perpendicular to the given vectors is
1 1 1
û = u = √ h0, −6, −3i = − √ h0, 2, 1i
kuk 3 5 5
15. Find the area of a triangle whose vertices lie on the different coor-
dinate axes at distances a, b, and c from the origin.

Solution: The vertices of the triangle are A = (a, 0, 0), B = (0, b, 0),
and C = (0, 0, c). Given the coordinates of the vertices, the area is
1 −→ −→
A∆ = kAB × ACk .
2
Therefore
−→
AB = h−a, b, 0i
−→
AC = h−a, 0, ci
−→ −→
AB × AC = hbc − 0, −(−ac − 0), 0 − (−ab)i = hbc, ac, abi
1√ 2 2
A∆ = b c + a 2 c2 + a 2 b 2 .
2
16. Find the area of a triangle ABC for A(1, 0, 1), B(1, 2, 3), and
C(0, 1, 1) and a nonzero vector perpendicular to the plane containing
the triangle.
−→ −→
Solution: Let a = AB = h0, 2, 2i and b = AC = h−1, 1, 0i. By
the geometrical properties of the cross product, the vector a × b is
orthogonal to both a and b and, hence, to the plane containing the
4. THE CROSS PRODUCT 59

triangle. Furthermore, ka × bk is the area of the parallelogram with


adjacent sides a and b and, hence, the area of the triangle is 12 ka × bk.
One has
 
ê1 ê2 ê3
a × b = det  0 2 2  = ê1(0 − 2) − ê2(0 + 2) + ê3(0 + 2)
−1 1 0
= h−2, −2, 2i = 2h−1, −1, 1i

ka × bk = 2kh−1, −1, 1ik = 2 3
√ √
Therefore the area of the triangle is 12 · 2 3 = 3 and the vector
h−1, −1, 1i is orthogonal to the plane in which the triangle lies.

17. Use the cross product to show that the area of the triangle whose
vertices are midpoints of the sides of a triangle with area A is A/4.
Hint: Define sides of the triangle of area A as vectors and express the
sides of the other triangle in question in terms of these vectors.

Solution: Without loss of generality, suppose one vertex of the tri-


angle is at the origin. Let the two other vertices be B and C. The area
of triangle OBC is given by
1 −−→ −→
A = kOB × OCk
2

−→ −→ −−→
because |OB × OCk gives the area of the parallelogram spanned by OB
−→
and OC, and the parallelogram is formed from two copies of OBC. Let
the midpoint of OB, OC, and BC be D, E, and F , respectively. Then
−−→ 1 −−→ − −→ −→ −→ −−→ −→ −−→ −−→
OD = 2 OB, OE = 12 OC , and OF = 12 (OB + OC) = OD + OE. Then,
−−→ −→ −→ −−→
two sides of the triangle are given by F D = − 12 OC and F E = − 21 OB.
Thus the area of the triangle is
1 −−→ −→ 1 1 −→ 1− −→ 1 1 −−→ −→ 1
A0 = kF D× F Ek = k− OC ×(− OB)k = ( kOB × OCk) = A
2 2 2 2 4 2 4
18. Consider a triangle whose vertices are midpoints of any three sides
of a parallelogram. If the area of the parallelogram is A, find the
area of the triangle. Hint: Define adjacent sides of the parallelogram
as vectors and express the sides of the triangle in terms of these vectors.

Solution: This problem can actually be easily done without cross


products. I will give both methods.
−→ −−→
First, consider the parallelogram spanned by two vectors OA and OB,
where A is on to the x-axis (all parallelograms are isomorphic to such
60 1. VECTORS AND THE SPACE GEOMETRY

a parallelogram, up to rigid motions). Then the area of this parallelo-


gram is
−→ −−→
A = kOA × OBk
Next, consider three midpoints D1 , D2 , D3 , where D1 is the midpoint
of OA, D2 the midpoint of OB, and D3 is the midpoint of the line
segment formed by connecting A to the last vertex.
−−→ −→ −−→ −−→ −−→
Cross product method: Then OD1 = 21 OA, OD2 = 12 OB, and OD3 =
−→ 1 −−→
OA + 2 OB. Two sides of the triangle formed from D1 ,D2, and D3 are
−−−→ −→ −−→ −−−→ −→
given by D3 D1 = − 21 (OA + OB) and D3 D1 = −OA. So the area is
given by
1 −−−→−−−→ 1 1 −→ −−→ −→ 1 −→ −
−→ 1
A0 = kD3 D1 D3 D2 k = k− (OA+OB)×(−OA) = kOA×OBk = A
2 2 2 4 4
−→ −→
Because OA × OA = 0.

Other method: Recognize that the line D2 D3 splits the parallelogram in


two. Thus there are two regions, one of area A/2 with the triangle, and
one of the same area without the triangle. Look at the triangles OD1 D2
and AD1 D3 . Let the height of the parallelogram be h and the base be
b. Then the triangles share the same base length and height, namely
h/2 and b/2. Thus their area is A∆ = 1/2(h/2b/2) = 1/2(hb/4). So
their combined area is hb/4, which is A/4. The only remaining area
is that of the triangle in question. Thus its area is A−A/2−A/4 = A/4.

19. Let A = (1, 2, 1) and B = (−1, 0, 2) be vertices of a parallelo-


gram. If the other two vertices are obtained by moving A and B along
straight lines by a distance of 3 units in the direction of the vector
a = h2, 1, −2i, find the area of the parallelogram.
−→
Solution: First, AB = h−2, −2, 1i Translating a point in the di-
rection of a by 3 units is done by adding 3â to the position vector of
the point. So first we must find 3â.
3
kak = 3 ⇒ a=a
kak
−−→
Let A0 be the translated point of A. Then AA0 = a. So the area of the
parallelogram, S, is
−−→ −→ −→
S = kAA0 × ABk = ka × ABk
4. THE CROSS PRODUCT 61

The cross product is computed as follows:


 
ê1 ê2 ê3      
−→ 1 −2 2 −2 2 1
a×AB = det  2 1 −2 = hdet , − det , det i = h−3, 2, −2i
−2 1 −2 1 −2 −2
−2 −2 1
Thus √
−→ p
S = ka × ABk = (−3)2 + 22 + (−2)2 = 17
20. Consider four points in space. Suppose that the coordinates of the
points are known. Describe a procedure based on the properties of the
cross product to determine whether the points are in one plane. In par-
ticular, are the points (1, 2, 3), (−1, 0, 1), (1, 3, −1), and (0, 1, 2) in one
plane?
21. Let the sides of a triangle have lengths a, b, and c and let the
angles at the vertices opposite to the sides a, b, and c be, respectively,
α, β, and γ. Prove that
sin α sin β sin γ
= = .
a b c
Hint: Define the sides as vectors and express the area of the triangle
in terms of the vectors adjacent at each vertex of the triangle.

Solution: Let A, B, C be the vertices of a triangle. Let α be the


angle at vertex A, β at B, and γ at C. Let |BC| = a, |AC| = b, and
|AB| = c. Then a is opposite α, b is opposite β, and c is opposite γ.
The area of the triangle can be computed in three ways as follows:
1 −→ −→ 1 −→ −→ 1
A∆ = kAB × ACk = kABkkACk sin α = (bc) sin α
2 2 2
1 −→ −−→ 1 −→ −−→ 1
= kBA × BCk = kBAkkBCk sin β = (ac) sin β
2 2 2
1 −→ −−→ 1 −→ − −→ 1
= kCA × CBk = kCAkkCBk sin γ = (ab) sin γ
2 2 2
We can equate these as follows
1 1 sin α sin β
(bc) sin α = (ac) sin β ⇒ =
2 2 a b
1 1 sin β sin γ
(ac) sin β = (ab) sin γ ⇒ =
2 2 b c
22. A polygon ABCD in a plane is a part of the plane bounded by
four straight line segments AB, BC, CD, and DA. Suppose that the
polygon is convex, that is, a straight line segment connecting any two
points in the polygon lies in the polygon. If the coordinates of the ver-
tices are specified, describe the procedure based on vector algebra to
62 1. VECTORS AND THE SPACE GEOMETRY

calculate the area of the polygon. In particular, let a convex polygon


be in the xy plane and put A = (0, 0), B = (x1, y1), C = (x2 , y2), and
D = (x3 , y3). Express the area in terms of xi and yi , i = 1, 2, 3.

Solution: The polygon ABCD can be divided into two parts along
the line AC. This forms two triangles, ABC and ACD. The area of the
polygon will be the sum of the area of each triangle. Since the polygon
is convex, each interior angle will be less than 180◦ , so ABC and ACD
are indeed triangles. Let the area of triangle ABC be A1 and that of
triangle ACD be A2. The total area, At , then is A1 + A2. So,
1 −→ −→
A1 = kAB × ACk
2
1 −→ − −→
A2 = kAC × ADk
2
At = A1 + A2
−→ −→ −−→
In particular, when A = (0, 0), AB, AC, and OD are the position
vectors of B, C, and D. Since each point lies in the xy plane, the
cross products above will only have a z component. Furthermore, their
magnitudes will be that z component. So,
 
−→ −→ x1 y1
AB × AC = h0, 0, det i = h0, 0, x1 y2 − x2 y1i
x2 y2
1 −→ −→ 1
A1 = kAB × ACk = |x1y2 − x2y1 |
2 2  
−→ −−→ x2 y2
AC × AD = h0, 0, det i = h0, 0, x2 y3 − x3 y2i
x3 y3
1 −→ − −→ 1
A2 = kAC × ADk = |x2y3 − x3y2 |
2 2
Thus the area of the polygon is
1
At = |x1y2 + x2y3 − x2y1 − x3y2|
2
This is known as the shoelace method.

23. Consider a parallelogram. Construct another parallelogram whose


adjacent sides are diagonals of the first parallelogram. Find the rela-
tion between the areas of the parallelograms.

Solution: Let a and b be vectors that form a parallelogram. The


diagonals of this parallelogram are given by d1 = a+ b and d2 = a− b.
4. THE CROSS PRODUCT 63

Then the area of the parallelogram spanned by these vectors is


A0 = kd1×d2 k = k(a+b)×(a−b)k = ka×a−a×b+b×a−b×bk = k2(b×a)k = 2A
where A is the area of the original parallelogram.

24. Given two nonparallel vectors a and b, show that any vector r
in space can be written as a linear combination r = xa + yb + za × b
and that the numbers x, y, and z are unique for every r. Express z in
terms of r, a and b. In particular, put a = h1, 1, 1i and b = h1, 1, 0i.
Find the coefficients x, y, and z for r = h1, 2, 3i. Hint: See Study
Problems ?? and ??.

Solution: This is actually a fundamental fact of linear algebra (see


direct sums of subspaces for more information). First, let P represent
the plane spanned by a and b, that is P = {sa + tb|s, t ∈ R}. Then
a × b is orthogonal to P. Let r be a vector in space. Then r can be
uniquely decomposed into two vectors r|| and r⊥ (Corollary 3.1). We
know that
r|| (a × b) · r
r|| = a×b= a×b
ka × bk ka × bk2
Next, we have that r⊥ = r − r||, where r⊥ ∈ P since a × b is orthogonal
to P. So r⊥ = xa + yb for some unique x, y ∈ R. This can easily
be shown as follows: Let r⊥ = hr1 , r2, r3 i , a = ha1 , a2, a3i, and b =
hb1 , b2, b3i. So if r⊥ = xa + yb, then we must have that

xa1 + yb1 = r1

xa2 + yb2 = r2

xa + yb = r
3 3 3

Where each equation was found by setting the components equal to


each other. Now, since a and b are not parallel, there exist some i, j
between 1 and 3 where ai 6= saj and bi 6= sbj for all s ∈ R. What
this means is that at least two of the equations above are linearly
independent. So we reduce our system of equations to the following
(
xai + ybi = ri
xaj + ybj = rj

for some i, j between 1 and 3. Now we can rewrite this as


    
x a i bi ri
=
y a j bj rj
64 1. VECTORS AND THE SPACE GEOMETRY

 
a i bi
Now recall that the determinant of A = is ai bj − bi aj . It ends
a j bj
 
bj −bi
up being that A−1 = A1 , so its existence is dependent on
−aj ai
the existence of det(A). Were it true that ai = saj and bi = sbj , then
det(A) = (saj )bj − (sbj )aj = 0. So it would be that A−1 wouldn’t
exist, but this is not true, so we are okay. What this means is that A−1
exists, and it is unique (evident since each of its entries are unique and
the determinant is too). Thus,
   
x r
= i A−1
y rj
and x, y are unique.
Now suppose two vectors r1 and r2 are such that their x, y are equal,
but r1 6= r2 . It remains to be shown that their z are unequal. Consider
the vector xa + yb. Draw a line perpendicular to P through the point
whose position vector is xa + yb. By construction, both r1 and r2 will
be position vectors for points on this line. Consider the quantity
(a × b) · r krk cos θ
z= =
ka × bk ka × bk
Thus, given a and b, z is determined by the magnitude of r and its
angle to a × b, parallel to the constructed line. Suppose kr1k = kr2 k.
There are exactly two points on the line a distance kr1k (so long as r1
and r2 do not lie in the plane, but if they do, there is only one point,
and so r1 = r2 , a contradiction). Thus, either r1 = r2 or r1 corresponds
to one of these points and r2 corresponds to the other. The former is a
contradiction so assume the latter. This, however, implies that θ1 , the
angle between r1 and a × b is unequal to θ2 (indeed, it should be that
cos θ1 = − cos θ2 ). Thus their z are of opposite sign and hence unequal.
Now suppose that θ1 = θ2 . Then it must be, for both vectors to lie on
the line, that they are equivalent, a contradiction. So z is unique.
Thus the first part of the question is proved, since the triplet (x, y, z)
is unique.

For the second part, we first calculate a × b where a = h1, 1, 1i and


b = h1, 1, 0i. So,
 
ê1 ê2 ê3      
11 11 11
a×b = det 1 1 1 = hdet
  , − det , det i = h−1, 1, 0i
10 10 11
1 1 0
4. THE CROSS PRODUCT 65

By the above, we have that


(a × b) · r h−1, 1, 0i · h1, 2, 3i 1
z= 2
=p 2 =
ka × bk (−1)2 + 12 + 02 2

And also,
r − r|| = xa + yy
1
h1, 2, 3i − h−1, 1, 0i = xh1, 1, 1i + yh1, 1, 0i
2
3 3
h , , 3i = hx + y, x + y, xi
2 2
It is clear then that x = 3 and x + y = 23 , so y = − 23 .

25. A tetrahedron is a solid with four vertices and four triangular


faces. Let v1, v2, v3, and v4 be vectors with lengths equal to the areas
of the faces and directions perpendicular to the faces and pointing out-
ward. Show that v1 + v2 + v3 + v4 = 0. Hint: Set up vectors being the
egdes of a tetrahedron. There are six edges. So all these vectors can
be expressed as linear combinations of particular three non-coplanar
vectors. Use the cross product to find the vectors vj , j = 1, 2, 3, 4, in
terms of the three non-coplanar vectors.

Solution: Let A, B, C and D be the coordinates of the vertices of


the tetrahedron, in the following way: With A oriented at the top,
looking down, vertices B, C, and D are coplanar and go counterclock-
wise (so B can be at the bottom right of the base, C at the top, and
−→ −→ − −→ − −→
D at the bottom left, with A still overhead). Then AB, AC, AD, BC,
−−→ −−→
BD, CD are the six edges of the tetrahedron. We have then that four
−→ −→
vectors perpendicular to the faces of the tetrahedron are AB × AC,
−→ −−→ −−→ −→ −−→ −−→
AC × AD, AD × AB, and BD × BC. The first three can be found
by orienting the tetrahedron such that vertex A points at the top, and
then looking from the top down. There will be three faces visible,
whose normal vectors correspond to the first three found. The order
in which the vectors are in the cross product can be found using the
right-hand rule, and that the normal vectors must point out. Because
the bottom face does not include vertex A, the last vector can be found
by choosing any two nonparallel vectors not including A, for example
−−→ −−→
BC and BD. The order in which the cross product is taken is based
on the right-hand rule.
−→ −→ −→ −−→ −−→ −→
Next, we set v1 = 12 AB × AC, v2 = 21 AC × AD, v3 = 12 AD × AB, and
66 1. VECTORS AND THE SPACE GEOMETRY

−−→ − −→
v4 = 12 BD × BC. Recall that the area of a triangle spanned by vertices
−→ −→
O, P , and Q is A∆ = 21 kOP × OQk. This explains the 12 in each of vj
for j = 1, 2, 3, 4, because kvj k needs to be equal to the area of each
triangle face. Next we must expand each of vj as follows, because in
their current form they are not useful.

1 −→ −→
v1 = AB × AC
2
1 −−→ −→ −→ −→
= (OB − OA) × (OC − OA)
2
1 −−→ −→ −→ −→ −→ −→
= (OB × (OC − OA) − OA × (OC − OA))
2
1 −−→ −→ −−→ −→ −→ −→
= (OB × OC − OB × OA − OA × OC )
2

Similarly, we conclude that

1 −→ −−→ −→ −→ −→ −−→
v2 = (OC × OD − OC × OA − OA × OD)
2
1 −−→ −−→ −−→ −→ −→ −−→
v3 = (OD × OB − OD × OA − OA × OB)
2
1 −−→ −→ −−→ −−→ −−→ −→
v4 = (OD × OC − OD × OB − OB × OC)
2

Computing v1 + v2 yields

1 −
−→ −→ −−→ −→ −→ −→ 1 −→ −−→ −→ −→ −→ −−→
v1 + v2 = (OB × OC − OB × OA − OA × OC) + (OC × OD − OC × OA − OA × OD)
2 2
1 −
−→ −→ −−→ −→ −→ −→ −→ −−→ −→ −→ −→ −−→
= (OB × OC − OB × OA − OA × OC + OC × OD − OC × OA − OA × OD)
2
1 −
−→ −→ −−→ −→ −→ −→ −→ −→ −→ −−→ −→ −−→
= (OB × OC − OB × OA − OA × OC − OC × OA + OC × OD − OA × OD)
2
1 −
−→ −→ −−→ −→ −→ −→ −→ −→ −→ −−→ −→ −−→
= (OB × OC − OB × OA − OA × OC + OA × OC + OC × OD − OA × OD)
2
1 −
−→ −→ −−→ −→ −→ −−→ −→ −−→
= (OB × OC − OB × OA + OC × OD − OA × OD)
2
4. THE CROSS PRODUCT 67

Computing v3 + v4 yields

1 −−→ −−→ −−→ −→ −→ − −→ 1 −−→ −→ −−→ −−→ −−→ −→


v3 + v4 = (OD × OB − OD × OA − OA × OB) + (OD × OC − OD × OB − OB × OC)
2 2
1 −−→ −−→ −−→ −→ −→ − −→ −−→ −→ −−→ − −→ −−→ −→
= (OD × OB − OD × OA − OA × OB + OD × OC − OD × OB − OB × OC)
2
1 −−→ −−→ −−→ − −→ −−→ −→ −→ −−→ −−→ −→ −−→ −→
= (OD × OB − OD × OB − OD × OA − OA × OB + OD × OC − OB × OC)
2
1 −−→ −→ −→ −−→ −−→ −→ − −→ −→
= (−OD × OA − OA × OB + OD × OC − OB × OC)
2
1 − −→ −→ − −→ −→ −→ −−→ −→ −−→
= (−OB × OC + OB × OA − OC × OD + OA × OD)
2

where in the last step, terms have been rearranged and the equality
u × v = −v × u has been used multiple times. Comparing v1 + v2 to
v3 +v4, it is now easy to see the cancellation. Hence v1 +v2 +v3 +v4 = 0

26. If a is a non-zero vector, a · b = a · c, and a × b = a × c,


does it follow that b = c?

Solution: By basic algebraic properties of the cross and dot prod-


ucts, the stated equations can be cast in the following form which is
then analyzed using the geometrical properties of the products:

a × c = a × b ⇒ a × (c − b) = 0 ⇒ a k c − b
a · c = a · b ⇒ a · (c − b) = 0 ⇒ a ⊥ c − b

But there exists only one vector that is perpendicular and parallel to
a non-zero vector a; it is the zero vector. Thus, c − b = 0 or c = b.

27. Given two non-parallel vectors a and b, construct three mutu-


ally orthogonal unit vectors ûi , i = 1, 2, 3, one of which is parallel to
a. Are such unit vectors unique? In particular, put a = h1, 2, 2i and
b = h1, 0, 2i and find ûi .

Solution: It is easy to see that a, a × b, and a × (a × b) are mutually


orthogonal. In fact, this was shown in problem 10. Therefore we set
1 1 1
û1 = kak a, û2 = ka×bk a × b, and û3 = ka×(a×b)k a × (a × b). These
vectors are not unique, as we could have chosen the negatives of ûi
i = 1, 2, 3.
68 1. VECTORS AND THE SPACE GEOMETRY

In particular, when a = h1, 2, 2i and b = h1, 0, 2i, we have


 
ê1 ê2 ê3      
2 2 1 2 12
a × b = det  1 2 2  = hdet , − det , det i = h4, 0, −2i
02 12 10
1 0 2
 
ê1 ê2 ê3      
2 2 1 2 12
a × (a × b) = det  1 2 2  = hdet , − det , det i = h−4, 10, −8i
0 −2 4 −2 40
4 0 −2
It follows that
1 1
û1 = √ h1, 2, 2i = h1, 2, 2i
2 2
1 +2 +2 2 3
1 1
û2 = p h4, 0, −2i = √ h2, 0, −1i
2 2
4 + 0 + (−2) 2 5
1 1
û3 = p h−4, 10, −8i = − √ h2, −5, 4i
(−4)2 + 102 + (−8)2 3 5
28. Find the area of a quadrilateral ACDB if A = (1, 0, −1), B =
−−→ −→ −→
(2, 1, 2), C = (0, 1, 2), and AD = 2AC − AB.
29. Find the area of a quadrilateral ABCD whose vertices are obtained
as follows. The vertex B is the result of moving A by a distance of 6
units along the vector u = h2, 1, −2i, C is obtained from B by moving
the latter by a distance of 5 units along the vector v = h−3, 0, 4i, and
−−→
CD = v − u.
30. Let ûi , i = 1, 2, 3, be an orthonormal basis in space with the prop-
erty that û3 = û1 × û2 . If a1 , a2, and a3 are the components of vector
a relative to this basis and b1, b2, and b3 are the components of b, show
that the components of the cross product a × b can also be computed
by the determinant rule given in Definition ?? where êi are replaced
by ûi . Hint: Use the “bac-cab” rule to find all pairwise cross products
of the basis vectors ûi .

Solution: We have that a = a1 û1 + a2û2 + a3û3 and b = b1û1 +


b2û2 + b3 û3 . Using the “bac-cab” rule and that û3 = û1 × û2 , we have
û1 × û3 = û1 × (û1 × û2 )
= û1 (û1 · û2) − û2 (û1 · û1 )
= −û2kû1 k2
= −û2
where I used the fact that û1 ·û2 = 0 since they are mutually orthogonal,
and that û1 · û1 = kû1k2 = 1, since it is a unit vector.
4. THE CROSS PRODUCT 69

Similarly, we have that û2 × û3 = û1 . Thus,


a × b = (a1û1 + a2 û2 + a3û3 ) × (b1û1 + b2û2 + b3 û3 )
= a1û1 × (b1 û1 + b2 û2 + b3û3 ) + a2 û2 × (b1 û1 + b2û2 + b3û3 )
+a3û3 × (b1û1 + b2 û2 + b3û3 )
= a1û1 × (b1 û1) + a1 û1 × (b2 û2 ) + a1û1 × (b3û3 ) + a2û2 × (b1 û1) + a2û2 × (b2û2 )
+a2û2 × (b3û3 ) + a3 û3 × (b1 û1 ) + a3û3 × (b2û2 ) + a3û3 × (b3 û3)
= a1b1 û1 × û1 + a1 b2û1 × û2 + a1b3 û1 × û3 + a2 b1û2 × û1 + a2b2 û2 × û2
+a2b3 û2 × û3 + a3 b1û3 × û1 + a3b2 û3 × û2 + a3 b3û3 × û3
= a1b2 û1 × û2 + a1 b3û1 × û3 + a2b1 û2 × û1 + a2 b3û2 × û3 + a3b1 û3 × û1 + a3 b2û3 × û2
= a1b2 û1 × û2 + a1 b3û1 × û3 − a2b1 û1 × û2 + a2b3 û2 × û3 − a3 b1û1 × û3 − a3b2û2 × û3
= a1b2 û3 + a1b3 (−û2) − a2b1 û3 + a2b3û1 − a3b1 (−û2) − a3 b2û1
= a1b2 û3 − a1b3û2 − a2b1 û3 + a2b3 û1 + a3 b1û2 − a3b2 û1
= (a2b3 − a3b2)û1 − (a1b3 − a3b1 )û2 + (a1 b2 − a2b1 )û3
     
a2 a3 a1 a3 a1 a2
= det u1 − det û2 + det û3
b2 b3 b1 b3 b1 b2
 
u1 û2 û3
= det  a1 a2 a3 
b1 b2 b3

31. Let the angle between the rigid rods in Example ?? be 0 < ϕ < π.
Find the equilibrium position of the system.

Solution: Recall that τ = r × F. It may be helpful to sketch the


problem, where the rods are pointing down and fixed at some common
point. Note that the position vector of the ball with mass m1 , r1 , is
coplanar with the gravitational force on that ball, F1 = m1 g (they
both lie in the plane that the rods lie in). This means that the torque
produced, τ 1 is orthogonal to this plane. The same goes for τ 2 , so they
are both orthogonal to the same plane. Then τ 1 and τ 2 are coplanar.
For the rod system to be in equilibrium, it must be that the sum of
its torques must be 0. Since the two vectors are coplanar, their mag-
nitudes simply add. By the right hand rule, the two torques are in
opposite directions. So, we have that τ1 − τ2 = 0 ⇒ τ1 = τ2 .
Recall that τ1 = L1 F1 sin φ1 where φ1 is the smallest angle between
r1 and F1 . Sketch these two vectors, noting that the gravitational
force points down. Create a horizontal line and define the angle θ1
to be the smallest angle between r1 and the horizontal. It should
be seen that θ1 + φ1 = π/2. Extend r1 enough to connect r2 to it,
70 1. VECTORS AND THE SPACE GEOMETRY

and extend the horizontal line, so that the horizontal line, r1 , and r2
form a triangle. It should be that the lower two angles of the triangle
are exactly θ1 and θ2 . As per the problem, we have ϕ + θ1 + θ2 =
π, so θ1 + θ2 = π − ϕ. Finally, using the trigonometric identity
sin(π/2 − θ) = cos(θ), we have that τ1 = L1 F1 sin(φ1 ) = L1 F1 cos(θ1 ).
Similarly, τ2 = L2 F2 cos(θ2 ) = L2 F2 cos(π − ϕ − θ1). Recall that
cos(A − B) = cos(A) cos(B) + sin(A) sin(B). So cos(π − ϕ − θ1 ) =
cos(π −ϕ) cos(θ1 )+sin(π −ϕ) sin(θ1 ) = − cos(ϕ) cos(θ1)+sin(ϕ) sin(θ1)

Returning to our equilibrium condition, τ1 = τ2 it follows that

τ1 = τ2
L1 m1g cos(θ1) = L2 m2g(− cos(ϕ) cos(θ1 ) + sin(ϕ) sin(θ1))
L1 m1
= − cos(ϕ) + sin(ϕ) tan(θ1)
L2 m2

32. Two rigid rods of the same length are rigidly attached to a ball
of mass m so that the angle between the rods is π/2. A ball of mass
2m is attached to one of the free ends of the system. The remaining
free end is used to hang the system. Find the angle between the rod
connecting the pivot point and the ball of mass m and the vertical axis
along which the gravitational force is acting. Assume that the masses
of the rods can be neglected as compared to m.

Solution: Construct a triangle OAB in the xy plane such that A


is in the third quadrant and B in the fourth, where |OA| = |AB| = L,
the angle OAB is π/2, and the smallest angle between OA and the
vertical is φ. This effectively models the situation, with the ball at
mass m at point A and the ball of mass 2m at point B. The rigid rods
are represented by the line segments OA and AB.
Recall the definition of torque as follows:

τ = r×F

Where r is the position vector to where the force F is being applied.


There are two torques in the problem, each produced by the gravita-
tional force acting on a ball, F1 = mg and F2 = 2mg. So r for each
−→ −−→
torque will simply be OA or OB. Thus we have
−→
kττ 1 k = τ1 = kOAkF1 k sin θ1 = L(mg) sin θ1

−→
kττ 2 k = τ2 = kOBkkF2 k sin θ2 = L(2mg) sin θ2
4. THE CROSS PRODUCT 71

−→
where θ1 and θ2 are defined to be the smallest angles between OA and
−−→
F1 and OB and F2 , respectively. Moreover, the directions of τ 1 and
τ 2 are opposite, by the right hand rule. Thus in equilibrium, it follows
that
τ1 − τ2 = 0 ⇔ τ1 = τ2
By definition, we have that θ1 = φ. This can be seen by moving the end
point of F1 to the origin. Since F1 is vertical and pointing downwards,
the smallest angle is φ, which we had defined earlier to be exactly this
smallest angle. Next, observe that OAB is a right triangle, where the
other two angles are both π/4. Translate F2 so its endpoint lands at
−−→
the origin. It follows that the smallest angle between OB and F2 is
θ2 = π/4 − φ. Recall that sin(A − B) = sin(A) cos(B) − sin(B) cos(A).
Therefore,

L(mg) sin φ = 2L(2mg) sin(π/4 − φ)

sin φ = 2 2 sin(π/4 − φ)
√ √
√ 2 2
= 2 2( cos φ − sin φ)
2 2
= 2(cos φ − sin φ)
1
= cot φ − 1
2
3
= cot φ
2
2
tan φ =
3
2
φ = arctan
3
33. Three rigid rods of the same length are rigidly joined by one end
so that the rods lie in a plane and the other end of each rod is free.
Let three balls of masses m1 , m2 , and m3 are attached to the free ends
of the rods. The system is hanged by the joining point and can rotate
freely about it. Assume that the masses of the rods can be neglected as
compared to the masses of the balls. Find the angles between the rods
at which the system remains in a horizontal plane under gravitational
forces acting vertically.

Solution: Without loss of generality, we can assume that the pivot


point is at the origin and one rigid rod is parallel to the negative y-axis.
Let this be the rod with mass m3 . This is convenient as it eliminates
the torque that would be produced by the ball attached to this rod.
72 1. VECTORS AND THE SPACE GEOMETRY

The torque vanishes since the position vector and the gravitational
force are both parallel to the negative y-axis. Notice that the solution
will be determined by only two angles, as the third can be found from
subtracting the sum of the two from 2π.
5. THE TRIPLE PRODUCT 73

5. The Triple Product


1–5. Find the triple products a · (b × c), b · (a × c), and c · (a × b) for
given vectors a, b, and c.
1. a = h1, −1, 2i, b = h2, 1, 2i, and c = h2, 1, 3i.

Solution: Recall that the definition of the triple product


a · (b × c) for three vectors a = ha1, a2, a3 i, b = hb1 , b2, b3 i and
c = hc1 , c2, c3 i is:
 
a1 a2 a3      
b 2 b 3 b 1 b 3 b 1 b 2
a·(b×c) = det  b1 b2 b3  = a1 det −a2 det +a3 det
c2 c3 c1 c3 c1 c2
c1 c2 c3
A fact of linear algebra is that swapping any two rows in a
matrix inverts the sign of its determinant. Therefore in the
triple product, switching any two vectors inverts the result.
Thus b · (a × c) = −a · (b × c). Moreover, switching any
two vectors twice will not change the sign. Switching any two
vectors twice can be viewed as permuting the vectors in the
triple product. Thus a · (b × c) = b · (c × a) = c · (a × b). So,
 
1 −1 2      
12 22 21
a·(b×c) = det 2 1 2 = 1 det
  +1 det +2 det = 1(1)+1(2)+2(0) = 3
13 23 21
2 1 3
Therefore the answers are 3, −3, 3 respectively.

2. a, b, and c are, respectively, position vectors of the points


A = (1, 2, 3), B = (1, −1, 1), and C = (2, 0, −1) relative to the
point O = (1, 1, 1).
−→ −−→ −→
Solution: First we must compute OA, OB, and OC as fol-
lows:
−→
a = OA = h1, 2, 3i − h1, 1, 1i = h0, 1, 2i
−−→
b = OB = h1, −1, 1i − h1, 1, 1i = h0, −2, 0i
−→
c = OC = h2, 0, −1i − h1, 1, 1i = h1, −1, −2i
Thus we have that
 
0 1 2      
−2 0 0 0 0 −2
a·(b×c) = det 0 −2 0  = 0 det −1 det +2 det =4
−1 −2 1 −2 1 −1
1 −1 −2
Therefore the answers are 4, −4, 4, respectively.
74 1. VECTORS AND THE SPACE GEOMETRY

3. a, b, and c are coplanar so that c = 2a − 3b.

Solution: Note that if a, b, and c are coplanar then b × c is


not only orthogonal to b and c, but also a, as they all lie in the
same plane and b×c is orthogonal to this plane. Therefore the
dot product between them vanishes, and the triple product is
0.

4. a = u1 + 2u2 , b = u1 − u2 + 2u3 , and c = u2 − 3u3 if


u1 · (u2 × u3) = 2

Solution: We have that


b × c = (u1 − u2 + 2u3 ) × (u2 − 3u3 )
= u1 × (u2 − 3u3 ) − u2 × (u2 − 3u3 ) + 2u3 × (u2 − 3u3 )
= u1 × u2 − 3u1 × u3 − u2 × u2 + 3u2 × u3 + 2u3 × u2 − 6u3 × u3
= u1 × u2 − 3u1 × u3 + 3u2 × u3 + 2u3 × u2
= u1 × u2 − 3u1 × u3 + u2 × u3
= u1 × (u2 − 3u3 ) + u2 × u3
And so,
a · (b × c) = (u1 + 2u2 ) · (u1 × (u2 − 3u3 ) + u2 × u3 )
= u1 · (u1 × (u2 − 3u3 ) + u2 × u3 ) + 2u2 · (u1 × (u2 − 3u3 ) + u2 × u3)
= u1 · (u1 × (u2 − 3u3 )) + u1 · (u2 × u3 ) + 2u2 · (u1 × (u2 − 3u3 )) + 2u2 · (u2 × u3)
= u1 · (u2 × u3 ) + 2u2 · (u1 × (u2 − 3u3 ))
= u1 · (u2 × u3 ) + 2u2 · (u1 × u2 ) − 6u2 · (u1 × u3 )
= u1 · (u2 × u3 ) − 6u2 · (u1 × u3 )
= u1 · (u2 × u3 ) + 6u1 · (u2 × u3 )
= 7u1 · (u2 × u3 )
Therefore the answers are 14, −14, 14, respectively.
5. a, b, and c are pairwise perpendicular and kak = 1, kbk = 2,
and kck = 3. Is the answer unique under the specified condi-
tions?

Solution: Recall that the absolute value of the triple product


gives the volume of the parallelpiped generated by the three
vectors. In this case, since they are all pairwise perpendicular,
the volume is simply |a · (b × c)| = kakkbkkck = 1(2)(3) = 6.
5. THE TRIPLE PRODUCT 75

Therefore a · (b × c) = ±6. The answer is not unique; it is


either 6, −6, 6 or −6, 6, −6, respectively.
6. Verify whether the vectors a = ê1 + 2ê2 − ê3, b = 2ê1 − ê2 + ê3 , and
c = 3ê1 + ê2 − 2ê3 are coplanar.

Solution: By the geometric properties of the triple product, three


vectors are coplanar iff their triple product vanishes. Thus we com-
pute the triple product as follows:
 
1 2 −1      
−1 1 2 1 2 −1
a·(b×c) = det 2 −1 1  = det −2 det −det = 1+14−5 = 10
1 −2 3 −2 3 1
3 1 −2
So they are not coplanar.

7. Consider the vectors a = h1, 2, 3i, b = h−1, 0, 1i and c = hs, 1, 2si


where s is a number.
(i) Find all values of s, if any, for which these vectors are coplanar.
(ii) If such s exists, find the area of the quadrilateral whose three ver-
tices have position vectors a, b, and c relative to the fourth vertex.
Hint: Determine which of the vectors a, b, and c is a diagonal of the
quadrilateral.

Solution: (i) By the geometric properties of the triple product, three


vectors are coplanar iff their triple product vanishes. Thus we must
compute the triple product as follows:
 
1 2 3      
0 1 −1 1 −1 0
a·(b×c) = det −1 0 1  = det −2 det +3 det = −1+6s−3 = 6s−4
1 2s s 2s s 1
s 1 2s
For the triple product to vanish then, it follows that s = 32 .
(ii) Evidently the diagonal will be the vector whose magnitude is the
largest. It is easy to see that this is a by comparing its components to
those of b and c. So the area of the quadrilateral will be
A = kb × ck
 
ê1 ê2 ê3
= k det −1 0 1  k
2
3
1 43
     
01 −1 1 −1 0
= khdet , − det 2 4 , det 2 ik
1 43 3 3 3
1

= kh−1, 2, −1ik = 6
76 1. VECTORS AND THE SPACE GEOMETRY

8. Determine whether the points A = (1, 2, 3), B = (1, 0, 1), C =


(−1, 1, 2), and D = (−2, 1, 0) are in one plane and, if not, find the
volume of the parallelepiped with adjacent edges AB, AC, and AD.
−→ −→ −−→
Solution: First compute AB, AC, and AD.
−→
AB = h1, 0, 1i − h1, 2, 3i = h0, −2, −2i
−→
AC = h−1, 1, 2i − h1, 2, 3i = h−2, −1, −1i
−−→
AD = h−2, 1, 0i − h1, 2, 3i = h−3, −1, −3i
By the geometric properties of the triple product, the triple product
gives the volume of the parallelpiped spanned by three vectors. Thus
we need only compute the triple product to answer the question.
 
0 −2 −2      
−1 −1 −2 −1 −2 −1
a · (b × c) = det −2 −1 −1 = 0 det
  + 2 det − 2 det
−1 −3 −3 −3 −3 −1
−3 −1 −3
= 2(3) − 2(−1) = 8
Thus the points are not in one plane, and the volume of the paral-
lelpiped is V = 8.
9. Find:
(i) all values of s at which the points A = (s, 0, s), B = (1, 0, 1),
C = (s, s, 1), and D = (0, 1, 0) are in the same plane;
(ii) all values of s at which the volume of the parallelepiped with adja-
cent edges AB, AC, and AD is 9 units.
−→ −→
Solution: Let a = AB = h1 − s, 0, 1 − si, b = AC = h0, s, 1 − si,
−−→
and c = AD = h−s, 1, −si. Then the vectors a, b, and c are coplanar
(and the points in question are in one plane) if and only if their triple
product vanishes. So the points in question are in one plane at the
values of s at which the triple vanishes (Part (i)). Furthermore, the
absolute value of the triple product is the volume of the parallelepiped
in question. The latter condition give an equation for s to be solved
(Part (ii)). One has
 
1−s 0 1−s
a · (b × c) = det  0 s 1 − s
−s 1 −s
= (1 − s)[−s2 − (1 − s)] + (1 − s)[0 + s2 ]
= (1 − s)[−s2 − (1 − s) + s2] = −(1 − s)2
5. THE TRIPLE PRODUCT 77

The triple product vanishes at s = 1 and, hence, A, B, C, and D are


in one plane when s = 1. The volume is

V = |a · (b × c)| = | − (1 − s)2 | = (s − 1)2 .

The condition V = 9 yields

(s − 1)2 = 9 ⇒ s − 1 = ±3 ⇒ s = 1 ± 3.

The volume of the parallelepiped is 9 units when s = 4 or s = −2.

10. Let a = h1, 2, 3i, b = h2, 1, 0i, and c = h3, 0, 1i. Find the vol-
ume of the parallelepiped with adjacent sides sa + b, c − tb, and a − pc
if the numbers s, t, and p satisfy the condition stp = 1.

Solution: By the geometric properties of the triple product, we know


its absolute value gives the volume of the parallelpiped spanned by the
three vectors. We first use the properties of the dot and cross product
to help simplify the problem:

V = |(sa + b) · ((c − tb) × (a − pc))| = |(sa + b) · (c × (a − pc) − tb × (a − pc))|


= |(sa + b) · (c × a − pc × c − tb × (a − pc))| = |(sa + b) · (c × a − tb × (a − pc))|
= |(sa · (c × a − tb × (a − pc)) + b · (c × a − tb × (a − pc))|
= |(sa · (c × a − t(b × a) + tp(b × c)) + b · (c × a) − tb · (b × (a − pc))|
= |sa · (c × a) − sta · (b × a) + stpa · (b × c)) + a · (b × c)|
= |(stp + 1)||a · (b × c)|

So we must calculate a · (b × c) now:


 
123      
10 20 21
a·(b×c) = det 2 1 0 = det
  −2 det +3 det = 1−2(2)+3(−3) = −12
01 31 30
301

Thus V = |stp + 1|| − 12| = 24.

11. Let the numbers u, v, and w be such that uvw = 1 and u3 +


v 3 + w3 = 1. Are the vectors a = uê1 + vê2 + wê3, b = vê1 + wê2 + uê3 ,
and c = wê1 + uê2 + vê3 coplanar? If not, what is the volume of the
parallelepiped with adjacent edges a, b, and c?
78 1. VECTORS AND THE SPACE GEOMETRY

Solution: We have that


 
u v w      
w u v u v w
a · (b × c) = det  v w u  = u det − v det + w det
u v wv w u
w u v
= u(vw − u2 ) − v(v 2 − uw) + w(uv − w2 ) = uvw − u3 − v 3 + uvw + uvw − w3
= 3uvw − (u3 + v 3 + w3 ) = 3 − 1 = 2

Thus the vectors are not coplanar

12. Prove that


 
a·c b·c
(a × b) · (c × d) = det .
a·d b·d

Hint: Put n = a × b. Use the invariance of the triple product under


cyclic permutations of vectors in it and the “bac-cab” rule (??).

Solution: Let n = a × b. Then, by the invariance of the triple


product under cyclic permutations of its vectors, we have

n · (c × d) = c · (d × n) = c · (d × (a × b))

Using the ”bac-cab” rule to evaluate d × (a × b) yields


 
a·c b·c
c·(d×(a×b)) = c·(a(d·b)−b(d·a)) = (c·a)(d·b)−(c·b)(d·a) = det
a·d b·d

This completes the proof.

13. Let P be a parallelepiped of volume V . Find the volumes of


all parallelepipeds whose adjacent edges are diagonals of the adjacent
faces of P .

Solution: Let a, b, and c be the adjacent sides of P . Then V =


|a · (b × c)|. The adjacent faces of P are parallelograms with adjacent
sides being the pairs of these vectors. Therefore the diagonals of the
faces are d1 = a + b, d2 = a + c, and d3 = b + c. Using the basic
algebraic properties of the dot, cross, and triple products, the volume
5. THE TRIPLE PRODUCT 79

in question is

V0 = |d1 · (d2 × d3 )|
= |d1 · (a × b + a × c + c × b + c × c)|
= |(a + b) · (a × b + a × c + c × b + 0)
= |a · (a × b) + a · (a × c) + a · (c × b) + b · (a × b) + b · (a × c) + b · (c × b)|
= |0 + 0 + a · (c × b) + 0 + b · (a × c) + 0|
= | − a · (b × c) − a · (b × c)|
= 2|a · (b × c)| = 2V .

14. Let P be a parallelepiped of volume V . Find the volumes of all par-


allelepipeds whose two adjacent edges are diagonals of two non-parallel
faces of P , while the third adjacent edge is a diagonal of P (the seg-
ment connecting two vertices of P that does not lie in a face of P ).

Solution: Let a, b, and c be the adjacent edges of P . First, the


diagonals of all the faces are:
Face 1,2: d1 = a + b Face 3,4: d2 = b + c Face 5,6: d3 = a + c.

Next, the diagonals of P are D1 = a + b + c, D2 = a − b + c,


D3 = a + b − c, and D4 = −a + b + c.

Let Vijk be the volume of the parallelpiped formed from diagonals


di , dj , and Dk . Notice that the only difference between each Dk is
the inclusion of a single minus sign and its location. To save time with
computation, I will generally call Dk = ska a + skb b + skc c, where each
s is either a + or a −, with a maximum of one −. Furthermore, notice
that each di can be written as sia a + sib b + sic c, where each s is either
a + or a 0, with a maximum of one 0. Thus, for the parallelpipeds
formed from di , dj , and Dk we have

Vijk = di · ((sja a + sjb b + sjc c) × (ska a + skb b + skc c))


= di · (sja a × (ska a + skb b + skc c) + sjb b × (ska a + skb b + skc c)
+sjc c × (ska a + skb b + skc c))
= di · (sja ska a × a + sja skb a × b + sja skc a × c + sjb ska b × a + sjb skb b × b + sjb skc b × c
+sjc ska c × a + sjc skb c × b + sjc skc c × c)
= di · (sja skb a × b + sja skc a × c + sjb ska b × a + sjb skc b × c + sjc ska c × a + sjc skb c × b)
80 1. VECTORS AND THE SPACE GEOMETRY

Vijk = (sia a + sib b + sic c) · ((sjaskb − sjb ska )a × b + (sja skc − sjc ska )a × c
+(sjb skc − sjc skb )b × c)
= sia a · ((sja skb − sjb ska )a × b + (sja skc − sjc ska )a × c + (sjb skc − sjc skb )b × c)
+sib b · ((sja skb − sjb ska )a × b + (sja skc − sjc ska )a × c + (sjb skc − sjc skb )b × c)
+sic c · ((sja skb − sjb ska )a × b + (sja skc − sjc ska )a × c + (sjb skc − sjc skb )b × c)
= sia (sjb skc − sjc skb )a · (b × c) + sib (sja skc − sjc ska )b · (a × c)
+sic ((sja skb − sjb ska )c · (a × b)
= sia (sjb skc − sjc skb )a · (b × c) − sib (sja skc − sjc ska )a · (b × c)
+sic ((sja skb − sjb ska )a · (b × c)
= (sia (sjb skc − sjc skb ) − sib (sja skc − sjc ska ) + sic ((sja skb − sjb ska ))a · (b × c)
   
sia sib sic sia sib sic
= det sja sjb sjc  a · (b × c) = det  sja sjb sjc  V
ska skb skc ska skb skc

Now it is simply a matter of doing all the computations, substituting


all the proper entries into the matrix. As an example, here is the
computation of V121
 
110      
1 1 0 1 01
V121 = det 0 1 1 V = (det − det + 0 det )V = (0 − (−1) + 0)V = V
11 11 11
111

All the possible volumes are as follows:

V121 = V V122 = 3V V123 = V V124 = V


V131 = V V132 = V V133 = V V134 = 3V
V231 = V V232 = V V233 = 3V V234 = V

15. Given two non-parallel vectors a and b, find the most general vec-
tor r that satisfies the conditions a · (r × b) = 0 and b · r = 0.

Solution: Using the invariance of the triple product under cyclic


permutations of its vectors, we have that a · (r × b) = r · (b × a) = 0. It
follows that r is perpendicular to both a × b and b. The first condition
implies that r lies in the plane spanned by a and b, and the latter
condition requires that r is orthogonal to b. By the first condition,
r = sa + tb for any real numbers s, t. Adding the second condition
5. THE TRIPLE PRODUCT 81

requires that s(a · b) + tkbk2 = 0 ⇔ t = − s(a·b)


kbk2
. Thus

s(a · b)
r = sa − b
kbk2

for any real number s is the most general vector r satisfying the con-
ditions.

16. Let a set S1 be the circle x2 + y 2 = 1 and let a set S2 be the


line through the points (0, 2) and (2, 0). Find the distance between the
sets S1 and S2 .

Solution: Recall that the distance D between two sets S1 and S2


is inf |A1A2|, where A1 ∈ S1 and A2 ∈ S2 . In this scenario, we must
find two points A1 and A2 such that the line segment A1A2 is perpen-
dicular to both sets. Since S1 is a circle, any straight line through its
center (here, the origin) that intersects it will be perpendicular to it.
2−0
The slope of the line passing through (0, 2) and (2, 0) is m = 0−2 = −1.
Thus a line perpendicular to it will have slope M = − m1 = 1. So a
straight line perpendicular to both sets must pass through the origin
and have slope √M = 1.√ There is only one such line, y = x, that passes
through S1 at ( 2/2, 2/2) and S2 at (1, 1). Thus the distance is

q √ √ q √ q√ √
D= (1 − 2/2)2 + (1 − 2/2) = 2(1 − 2/2) = ( 2 − 1)2 = 2−1
2 2

17. Consider a plane through three points A = (1, 2, 3), B = (2, 3, 1),
and C = (3, 1, 2). Find the distance between the plane and a point
P obtained from A by moving the latter 3 units along a straight line
segment parallel to the vector a = h−1, 2, 2i.

Solution: The distance can be interpreted as the distance between


two parallel planes. One plane goes through A, B, and C. The other
plane is parallel to this and passes through P . Recall that the distance
between two planes is

−→ −→ −→
|AP · (AB × AC)|
D= −→ −→
kAB × ACk
82 1. VECTORS AND THE SPACE GEOMETRY

−→ −→ −→
First we compute AB, AC, and AP as follows:
−→
AB = h2, 3, 1i − h1, 2, 3i = h1, 1, −2i
−→
AC = h3, 1, 2i − h1, 2, 3i = h2, −1, −1i
−→ 3 3
AP = a= √ a=a
kak 1+4+4
−→ −→
It follows that AB × AC is
 
ê1 ê2 ê3      
−→ −→ 1 −2 1 −2 1 1
AB×AC = det  1 1 −2 = hdet , − det , det i = h−3, −3, −3i
−1 −1 2 −1 2 −1
2 −1 −1
−→ −→
Thus kAB × ACk is
−→ −→ √
kAB × ACk = kh−3, −3, −3ik = k − 3h1, 1, 1ik = 3kh1, 1, 1ik = 3 3
−→ −→ −→
Furthermore we have that |AP · (AB × AC)| is
−→ −→ −→
|AP ·(AB×AC)| = |h−1, 2, 2i·h−3, −3, −3i| = |−1(−3)+2(−3)+2(−3)| = 9
Thus the distance between the plane and the point P is
9 3 √
D= √ = √ = 3
3 3 3
18. Consider two lines. The first line passes through the points (1, 2, 3)
and (2, −1, 1), while the other passes through the points (−1, 3, 1) and
(1, 1, 3). Find the distance between the lines.

Solution: First we must determine if the two lines are skew, parallel,
coincident, or incident. Set A = (1, 2, 3), B = (2, −1, 1), C = (−1, 3, 1),
and P = (1, 1, 3). Recall that two lines are skew iff
−→ −→ −→
AC · (AB × CP ) 6= 0

where A and B are distinct points on line L1 and C and D are distinct
−→ −→ −→
points on line L2 . We compute each of AC, AB, and CP as follows:
−→
AC = h−1, 3, 1i − h1, 2, 3i = h−2, 1, −2i
−→
AB = h2, −1, 1i − h1, 2, 3i = h1, −3, −2i
−→
CP = h−1, 3, 1i − h1, 1, 3i = h−2, 2, −2i
5. THE TRIPLE PRODUCT 83

−→ −→
It follows that AB × CP is
 
ê1 ê2 ê3      
−→ −→ −3 −2 1 −2 1 −3
AB × CP = det 1 −3 −2 = hdet
  , − det , det i
2 −2 −2 −2 −2 2
−2 2 −2
= h10, 6, −4i
−→ −→ −→
Thus AB · (AB × CP ) is
−→ −→ −→
AB·(AB×CP ) = h−2, 1, −2i·h10, 6, −4i = −2(10)+1(6)−2(−4) = −20+6+8 = −20+14 = −6
So the lines are skew. We therefore use the formula for the distance
between skew lines, which is as follows:
−→ −→ −→
|AB · (AB × CP )|
D= −→ −→
kAB × CP k
−→ −→
We must calculate kAB × CP k, which is
−→ −→ p √ √ √
kAB × CP k = 102 + 62 + (−4)2 = 100 + 36 + 16 = 152 = 2 38
It follows that the distance is
| − 6| 3
D= √ =√
2 38 38
19. Find the distance between the line through the points (1, 2, 3)
and (2, 1, 4) and the plane through the points (1, 1, 1), (3, 1, 2), and
(1, 2, −1). Hint: If the line is not parallel to the plane, then they in-
tersect and the distance is 0. So check first whether the line is parallel
to the plane. How can this be done?

Solution: Let r be a vector parallel to a line L. Observe that if


L is parallel to the plane P, for every O ∈ P there exists a P ∈ P such
−→
that OP = r. It is sufficient to check this for only one such O. This
can be interpreted in the following way: Take a point in the plane, and
try to make r from it. If the vector is not in the plane, then clearly it
simply passes through, and it is not parallel.

First we compute r using the points given in the line


r = h2, 1, 4i − h1, 2, 3i = h1, −1, 1i
Let O = (1, 1, 1), A = (3, 1, 2), and B = (1, 2, −1). Then
−→
OA = h3, 1, 2i − h1, 1, 1i = h2, 0, 1i
−−→
OB = h1, 2, −1i − h1, 1, 1i = h0, 1, −2i
84 1. VECTORS AND THE SPACE GEOMETRY

−→
Suppose there exists a P ∈ P such that OP = r. Then P is in the
−→ −−→ −→
plane iff OA, OB, and OP are coplanar. Thus we compute the triple
product
 
1 −1 1      
−→ −→ − −→ 0 1 2 1 20
OP · (OA × OB) = det 2 0
 1  = (1) det − (−1) det + (1) det
1 −2 0 −2 01
0 1 −2
= −1 + (−4) + 2 = −3

So the triple product does not vanish. It follows that the vectors are
−→
not coplanar, hence no P exists such that OP = r. Thus L is not
parallel to P. So the distance is 0.

20. Consider the line through the points (1, 2, 3) and (2, 1, 2). If a
second line passes through the points (1, 1, s) and (2, −1, 0), find all
values of s, if any, at which the distance between the lines is 3/2 units.

Solution: Let A = (1, 2, 3), B = (2, 1, 2), C = (1, 1, s), and P =


(2, −1, 0). Then
−→
AB = h2, 1, 2i − h1, 2, 3i = h1, −1, −1i
−→
CP = h2, −1, 0i − h1, 1, si = h1, −2, −si
−→
It is clear that there does not exist a real number s such that AB is
−→
parallel to CP . Thus the lines will always be skew. It follows that we
can use the formula for the distance between skew lines, which reads
−→ −→ −→
|AC · (AB × CP )|
D= −→ −→
kAB × CP k
for any distinct points A, B on one line and C, P on the other line.
−→
We first calculate AC, which is
−→
AC = h1, 1, si − h1, 2, 3i = h0, −1, s − 3i
−→ −→
Then AB × CP is
 
ê1 ê2 ê3      
−→ −→ −1 −1 1 −1 1 −1
AB × CP = det 1 −1 −1 = hdet
  , − det , det i
−2 −s 1 −s 1 −2
1 −2 −s
= hs − 2, s − 1, −1i
5. THE TRIPLE PRODUCT 85

It follows that
−→ −→ p √
kAB × CP k = (s − 2)2 + (s − 1)2 + (−1)2 = s2 − 4s + 4 + s2 − 2s + 1 + 1

= 2s2 − 6s + 6
−→ −→ −→
Moreover, |AC · (AB × CP )| is
−→ −→ −→
|AC · (AB × CP )| = |h0, −1, s − 3i · hs − 2, s − 1, −1i| = |0(s − 2) − (s − 1) + (s − 3)(−1)|
= |2s − 4| = 2|s − 2|
So the distance between the lines is
2|s − 2|
D= √
2s2 − 6s + 6
The problem asks for the values of s such that D = 3/2, so
2|s − 2| 3
√ =
2s2 − 6s + 6 2

4|s − 2| = 3 2s2 − 6s + 6
16(s − 2)2 = 9(2s2 − 6s + 6)
16s2 − 64s + 64 = 18s2 − 54s + 54
2s2 + 10s − 10 = =0
s2 + 5s − 5 = =0
To which applying the quadratic formula yields

−5 ± 3 5
s=
2
as the solutions.

21. Consider two parallel straight line segments in space. Formulate


an algorithm to compute the distance between them if the coordinates
of their end points are given. In particular, find the distance between
AB and CD if:
(i) A = (1, 1, 1), B = (4, 1, 5), C = (2, 3, 3), D = (5, 3, 7);
(ii) A = (1, 1, 1), B = (4, 1, 5), C = (3, 5, 5), D = (6, 5, 9)
Note that this distance does not generally coincide with the distance
between the parallel lines containing AB and CD. The segments may
even be in the same line at a nonzero distance.

Solution: The issue with simply applying the distance between par-
allel lines formula is that it assumes the lines extend infinitely, or
at least that there exist points L1 in L1 and L2 in L2 where L1 L2
86 1. VECTORS AND THE SPACE GEOMETRY

is perpendicular to both lines. However, consider the line segment


L1 = {(x, 2)|x ∈ [−2, −1]} (line segment of length 1 centered at −3/2)
and the line segment [0, 1]. Clearly no L1 , L2 exist to satisfy the re-
quired criteria. The distance then is simply the shortest distance be-
tween the endpoint of one line and the endpoint of the other line (in
this case, the distance between (−1, 2) and (0, 0)). It can be seen that
L1 and L2 exist when one line is ”above” another, with respect to the
plane the two line segments are in. First, everything will be conducted
where |AB| ≥ |CD|. Thus step 1 of the algorithm is to compute the
length of each line segment. Whichever is longer, call the endpoints
A and B, and whichever is shorter, call the endpoints C and D (note:
these may not match the names given in the problem. Ignore those
−→ −→ −−→
and rename them accordingly). Step 2 is to calculate AB, AC, AD,
−−→ −−→
BC, and BD, where the first vector is the vector parallel to the line
segments and the other vectors are the vectors parallel to the lines
passing through all the endpoint pairs, where one endpoint is from one
line and the other is from the other. We have the following cases for
what can occur:

Case 1:
−→ −→ − −→ −−→ −−→
AB || AC || AD || BC || BD
In this case, the two line segments lie on the same line. The distance
−→ −−→ −−→ −−→
is inf(S) where S = {kACk, kADk, kBCk, kBDk}.

Case 2:
There will be many sub-cases here. First, it is important to compute
all the following dot products:
−→ −→ −→ − −→ −→ − −→ −→ −−→
d1 = AB · AC, d2 = AB · AD, d3 = BA · BC, d4 = BA · BD
The sign of each of these dot products will determine which sub-case
we are in, as the sign determines whether the angle between two vec-
tors is acute (positive), right (zero), or obtuse (negative).

a).
Two dot products are negative.
In this case, segment CD is not ”above” AB. Thus, we cannot use
the distance formula, and the distance is inf(S), where S is defined as
above.
This case is the one used in the example above.

b).
5. THE TRIPLE PRODUCT 87

One dot product is negative.


In this case, one end point of segment CD is ”above” AB. Thus we
can form a perpendicular segment from that endpoint to a point on
AB. Thus the distance formula can be used.

c).
No dot products are negative.
In this case, both endpoints of segment CD are above AB. We can
do something similar to case 2b, and hence we can use the distance
formula.

Overall, we can use the distance formula when one or none of the
dot products are negative. Otherwise, compute inf(S).

(i): The length of the line segments are


p √
l1 = (4 − 1)2 + (1 − 1)2 + (5 − 1)2 = 9 + 16 = 5
p √
l2 = (5 − 2)2 + (3 − 3)2 + (7 − 3)2 = 9 + 16 = 5
Since the lengths are equal, it does not matter which segment we call
AB. Thus use the designations in the problem. Next, we compute the
required vectors
−→
AB = h4, 1, 5i − h1, 1, 1i = h3, 0, 4i
−→
AC = h2, 3, 3i − h1, 1, 1i = h1, 2, 2i
−−→
AD = h5, 3, 7i − h1, 1, 1i = h4, 2, 6i
−−→
BC = h2, 3, 3i − h4, 1, 5i = h−2, 2, −2i
−−→
BD = h5, 3, 7i − h4, 1, 5i = h1, 2, 2i
And finally compute the required dot products
−→ −→
d1 = AB · AC = h3, 0, 4i · h1, 2, 2i = 3(1) + 0(2) + 4(2) = 11
−→ −−→
d2 = AB · AD = h3, 0, 4i · h4, 2, 6i = 3(4) + 0(2) + 4(6) = 36
−→ −−→
d3 = BA · BC = −h3, 0, 4i · h−2, 2, −2i = −(3(−2) + 0(2) + 4(−2)) = 14
−→ −−→
d2 = BA · BD = −h3, 0, 4i · h1, 2, 2i = −(3(1) + 0(2) + 4(2)) = −11
Since only one dot product is negative, we can use the distance formula
as follows:
−→ −→
kAB × ACk
D= −→
kABk
88 1. VECTORS AND THE SPACE GEOMETRY

We have already calculated the denominator; it is |AB|. Thus we must


−→ −→
calculate kAB × ACk.
 
ê1 ê2 ê3      
−→ −→ 0 4 3 4 30
AB×AC = det  3 0 4  = hdet , − det , det i = h−8, −2, 6i
22 12 12
1 2 2
It follows that
−→ −→ p √ √
kAB × ACk = (−8)2 + (−2)2 + 62 = 104 = 2 26
Therefore the distance is

2 26
D=
5
(ii): The length of the line segments are
p √
l1 = (4 − 1)2 + (1 − 1)2 + (5 − 1)2 = 9 + 16 = 5
p √
l2 = (6 − 3)2 + (5 − 5)2 + (9 − 5)2 = 9 + 16 = 5
Since the lengths are equal, it does not matter which segment we call
AB. Thus use the designations in the problem. Next, we compute the
required vectors
−→
AB = h4, 1, 5i − h1, 1, 1i = h3, 0, 4i
−→
AC = h3, 5, 5i − h1, 1, 1i = h2, 4, 4i
−−→
AD = h6, 5, 9i − h1, 1, 1i = h5, 4, 8i
−−→
BC = h3, 5, 5i − h4, 1, 5i = h−1, 4, 0i
−−→
BD = h5, 3, 7i − h4, 1, 5i = h2, 4, 4i
And finally compute the required dot products
−→ −→
d1 = AB · AC = h3, 0, 4i · h2, 4, 4i = 3(2) + 0(4) + 4(4) = 22
−→ −−→
d2 = AB · AD = h3, 0, 4i · h5, 4, 8i = 3(5) + 0(4) + 4(8) = 47
−→ −−→
d3 = BA · BC = −h3, 0, 4i · h−1, 4, 0i = −(3(−1) + 0(4) + 4(0)) = 3
−→ −−→
d2 = BA · BD = −h3, 0, 4i · h2, 4, 4i = −(3(2) + 0(4) + 4(4)) = −22
Since only one dot product is negative, we can use the distance formula
as follows:
−→ −→
kAB × ACk
D= −→
kABk
5. THE TRIPLE PRODUCT 89

We have already calculated the denominator; it is |AB|. Thus we must


−→ −→
calculate kAB × ACk.
 
ê1 ê2 ê3      
−→ −→ 0 4 3 4 30
AB×AC = det  3 0 4  = hdet , − det , det i = h−16, −4, 12i
44 24 24
2 4 4
It follows that
−→ −→ p √ √
kAB × ACk = (−16)2 + (−4)2 + 122 = 416 = 4 26
Therefore the distance is √
4 26
D=
5
22-25. Consider the parallelepiped with adjacent edges AB, AC, and
AD where A = (3, 0, 1), B = (−1, 2, 5), C = (5, 1, −1), D = (0, 4, 2).
Find the specified distances.
22. The distances between the edge AB and all other edges parallel
to it.
23. The distances between the edge AC and all other edges parallel
to it;
24. The distances between the edge AD and all other edges parallel
to it;
25. The distances between all parallel planes containing the faces
of the parallelepiped.
26. The distances between all skew lines containing the edges of
the parallelepiped.
90 1. VECTORS AND THE SPACE GEOMETRY

6. Planes in Space
1. Find an equation of the plane through the origin and parallel to the
plane 2x − 2y + z = 4. What is the distance between the two planes?

Solution: Recall that the standard form of a plane is


n · r = n · r0
where n is a vector normal to the plane and r0 is the position vector of
a point in the plane. Thus for a plane to be parallel to the given one,
it must have the same normal vector. For the plane to pass through
the origin, we can choose r0 to be the position vector of the origin. So
n · r0 = 0, and
2x − 2y + z = 0
is a plane parallel to the given one passing through the origin.
Given two parallel planes
n · r = d1 , n · r = d2
The distance between the planes is defined to be
|d2 − d1 |
D=
knk
Applying this formula to the problem yields
|4 − 0| 4 4
D=p =√ =
22 + (−2)2 + 12 9 3
2. Do the planes 2x + y − z = 1 and 4x + 2y − 2z = 10 intersect?

Solution: An equivalent question is if the planes are parallel or co-


incide. The planes are parallel or coincide iff their normal vectors are
parallel. The normal vector of plane P1 defined by 2x+y−z = 1 is n1 =
h2, 1, −1i. The normal vector of plane P2 defined by 4x + 2y − 2z = 10
is n2 = h4, 2, −2i. Suppose there is a real number s where n1 = sn2 .
Then 2 = s(4), 1 = s(2), and −1 = s(−2) by equating the compo-
nents. Solving the first yields s = 1/2. Substituting into the other two
equations yields true statements, so the normal vectors are parallel. By
definition, intersecting planes have nonparallel normal vectors. So the
planes do not intersect.

3. Determine whether the planes 2x + y − z = 3 and x + y + z = 1 are


intersecting. If they are, find the angle between them.
6. PLANES IN SPACE 91

Solution: An equivalent question is if the planes are parallel or co-


incide. The planes are parallel or coincide iff their normal vectors are
parallel. The normal vector of plane P1 defined by 2x + y − z = 3 is
n1 = h2, 1, −1i. The normal vector of plane P2 defined by x + y + z = 1
is n2 = h1, 1, 1i. Suppose there is a real number s where n1 = sn2 .
Then 2 = s(1), 1 = s(1), and −1 = s(1) by equating the compo-
nents.Solving the first yields s = 2, but substituting into the second
yields 1 = 2, a contradiction. Thus the vectors are not parallel. So the
planes are not parallel nor coincidental, and must intersect.

The angle between two planes is defined as


|n1 · n2|
cos θ =
kn1 kkn2k
Thus,

h2, 1, −1i · h1, 1, 1i 2+1−1 2
θ = arccos p √ = arccos √ √ = arccos
22 + 12 + (−1)2 12 + 12 + 12 6 3 3
4–6. Consider a parallelepiped with one vertex at the origin O at
which the adjacent sides are the vectors a = h1, 2, 3i, b = h2, 1, 1i, and
c = h−1, 0, 1i. Let OP be its diagonal extended from the vertex O.
Find equations of the following planes.
4. The planes that contain the faces of the parallelepiped.

Solution: There are 6 faces to find, 3 through the origin


−→
and 3 through the point P . First computing OP gives
−→
OP = a + b + c = h1, 2, 3i + h2, 1, 1i + h−1, 0, 1i = h2, 3, 5i
Next we must find the normal vector of each plane. These
normal vectors will be a × b, b × c, and a × c.
 
ê1 ê2 ê3      
2 3 1 3 12
n1 = a × b = det  1 2 3  = hdet , − det , det i
11 21 21
2 1 1
= h(2 − 3), −(1 − (6)), (1 − (4)i = h−1, 5, −3i
 
ê1 ê2 ê3      
11 2 1 2 1
n2 = b × c = det 2 1 1 = hdet
  , − det , det i
01 −1 1 −1 0
−1 0 1
= h1, −(2 − (−1)), (0 − (−1)i = h1, −3, 1i
92 1. VECTORS AND THE SPACE GEOMETRY

 
ê1 ê2 ê3      
2 3 1 3 1 2
n3 = a × c = det  1 2 3  = hdet , − det , det i
01 −1 1 −1 0
−1 0 1
= h2, −(1 − (−3)), (0 − (−2))i = h2, −4, 2i

Recall that the standard equation of a plane is

n · r = n · r0

where n is a normal vector to the plane, r is a general position


vector, and r0 is the position vector of a point in the plane.
The left hand side for each plane can now be computed using
the dot product, with r = hx, y, zi

n1 · r = −x + 5y − 3z
n2 · r = x − 3y + z
n3 · r = 2x − 4y + 2z

The right hand side can be calculated using the dot product,
−→
with r0 as either h0, 0, 0i or OP . The former dot product is
trivially 0 for each plane. The latter is
−→
n1 · OP = −1(2) + 5(3) − 3(5) = −2
−→
n2 · OP = 1(2) − 3(3) + (5) = −2
−→
n3 · OP = 2(2) − 4(3) + 2(5) = 2

Thus the three planes passing through the origin are

x − 5y + 3z = 0 x − 3y + z = 0 x − 2y + z = 0
−→
and the three planes passing through OP are

x − 5y + 3z = 2 x − 3y + z = −2 x − 2y + z = 1

5. The planes that contain the diagonal OP and the diagonal of


each of three its faces adjacent at P .

Solution: The diagonals at the faces adjacent at P are


d1 = −(a + b), d2 = −(b + c), and d3 = −(a + c). If a
plane contains two vectors, then the normal vector must be
parallel to the cross product of those vectors. Thus we can
−→ −→ −→
let n1 = OP × d1 , n2 = OP × d2 , and n3 = OP × d3 . We
6. PLANES IN SPACE 93

−→ −→ −→
calculate OP × a, OP × b, and OP × c separately below
 
ê1 ê2 ê3      
−→ 3 5 2 5 23
OP × a = det  2 3 5  = hdet , − det , det i = h−1, −1, 1i
23 13 12
1 2 3
 
ê1 ê2 ê3      
−→ 3 5 2 5 23
OP × b = det  2 3 5  = hdet , − det , det i = h−2, 8, −4i
11 21 21
2 1 1
 
ê1 ê2 ê3      
−→ 35 2 5 2 3
OP × c = det 2 3 5 = hdet
  , − det , det i = h3, −7, 3i
01 −1 1 −1 0
−1 0 1
Then, we have that
−→
n1 = OP × d1 = −(h−1, −1, 1i + h−2, 8, −4i) = −h−3, 7, −3i = h3, −7, 3i
−→
n2 = OP × d2 = −(h−2, 8, −4i + h3, −7, 3i) = −h1, 1, −1i = h−1, −1, 1i
−→
n3 = OP × d3 = −(h−1, −1, 1i + h3, −7, 3i) = −h2, −8, 4i = h= 2, 8, −4i
These normal vectors should seem very familiar. This is no
−→ −→
coincidence, because OP = a + b + c So, for example, −OP ×
−→ −→ −→ −→
(a+ b) = −OP × (a+ b+ c − c) = −OP × (OP − c) = OP × c.
Notice that each ni for 1 ≤ i ≤ 3 is the cross product of some
−→
nonzero vector and OP . Thus the dot product of each normal
−→
vector with OP will be 0. It follows that the equation for each
plane is
3x − 7y + 3z = 0 x + y − z = 0 x − 4y + 2z = 0
6. The planes that contain parallel diagonals in the opposite faces
of the parallelepiped.

Solution: The following argument may be hard to visualize.


There is a set of symmetry operations that will map parallel
diagonals on opposite faces to each other, while keeping the
diagonal fixed. Specifically, take any two parallel faces. Con-
structing a line segment from the center of these faces gives
an axis. Rotate about this axis. Construct a plane where the
rotation axis is orthogonal to it. Reflecting the parallelpiped
over this plane provides the above set of operations. This im-
plies that the planes that go through the origin are exactly
those in problem 5. However, there are still three more to
−→
compute. The normal vectors will be n1 = OP × (a − b),
94 1. VECTORS AND THE SPACE GEOMETRY

−→ −→
n2 = OP × (b − c), and n3 = OP × (a − c). So,
−→
n1 = OP × d1 = h−1, −1, 1i − h−2, 8, −4i = h1, −9, 5i
−→
n2 = OP × d2 = h−2, 8, −4i − h3, −7, 3i = h−5, 15, −7i
−→
n3 = OP × d3 = h−1, −1, 1i − h3, −7, 3i = h−4, 6, −2i
NEEDS TO BE COMPLETED

7. Find an equation of the plane with x intercept a 6= 0, y intercept


b 6= 0, and z intercept c 6= 0. What is the distance between the origin
and the plane? Find the angles between the plane and the coordinate
planes.

Solution: If the plane P has intercepts (a, 0, 0), (0, b, 0), and (0, 0, c)
(a, b, c nonzero) then the plane must pass through these points. The
general form of a plane is
nx x + ny y + nz z = d
for real numbers nx , ny , nz , d. Thus, nx a = d, ny b = d, nz c = d by
substituting each intercept into the equation for the plane. So, nx =
d/a, ny = d/b, and nz = d/c, which are valid since a, b, c are nonzero.
It follows that
x y z
(d/a)x + (d/b)y + (d/c)z = d ⇔ + + =1
a b c
by substituting nx , ny , nz into the equation for the plane.

The distance between the plane and the origin can be answered by
finding the distance between the plane and a parallel plane passing
through the origin, x/a + y/b + z/c = 0. Applying the distance formula
yields
|1 − 0| abc
D= q =√
( 1 )2 + ( 1 )2 + ( 1 )2 b 2 c2 + a 2 c2 + a 2 b 2
a b c

Recall that the angle between two planes whose normals are n1 and
n2 , respectively, is
θ = arccos |nˆ1 · nˆ2 |
We first find nˆ1 to be
1 1 1 1 1
nˆ1 = q h , , i= √ hbc, ac, abi
( a1 )2 + ( 1b )2 + ( 1c )2 a b c b c + a 2 c2 + a 2 b 2
2 2
6. PLANES IN SPACE 95

Next, nˆ2 will be one of h1, 0, 0i, h0, 1, 0i, or h0, 0, 1i, for the angle
between the plane and the yz, xz, or xy plane, respectively. Thus we
have
1 bc D
θyz = arccos | √ hbc, ac, abi · h1, 0, 0i| = | √ |=
2 2 2 2
b c +a c +a b 2 2 b c + a 2 c2 + a 2 b 2
2 2 |a|
1 ac D
θxz = arccos | √ hbc, ac, abi · h0, 1, 0i| = | √ |=
2 2 2 2
b c +a c +a b 2 2 b c + a 2 c2 + a 2 b 2
2 2 |b|
1 ab D
θxy = arccos | √ hbc, ac, abi · h0, 0, 1i| = | √ |=
2 2 2 2
b c +a c +a b 2 2 b c + a 2 c2 + a 2 b 2
2 2 |c|
8. Show that the points A = (1, 1, 1), B = (1, 2, 3), C = (2, 0, −1) and
D = (3, 1, 0) are not in a plane and therefore vertices of a tetrahedron.
Any two of the four faces of the tetrahedron are intersecting along one
of its six edges. Find the angles of intersection of the face BCD with
the other three faces.

Solution: To answer the first part, we can find an equation for the
plane containing face BCD (which we will need anyway later), and
show that point A does not satisfy the equation. The normal to the

−→ −−→ −−→ −−→
plane can be n1 = BC × CD. Thus we must find BC and BD as
follows:
−−→ −→ − −→
BC = OC − OB = h2, 0, −1i − h1, 2, 3i = h1, −2, −4i
−−→ −−→ −−→
BD = OD − OB = h3, 1, 0i − h1, 2, 3i = h2, −1, −3i
It follows that
 
ê1 ê2 ê3      
−2 −4 1 −4 1 −2
n1 = det  1 −2 −4 = hdet , − det , det i = h2, −5, 3i
−1 −3 2 −3 2 −1
2 −1 −3
−−→
Next, point B must be a part of the plane. So we can let d1 = n1 · OB.
Thus,
d1 = h2, −5, 3i · h1, 2, 3i = 2(1) − 5(2) + 3(3) = 1
Thus an equation for the plane in question is
2x − 5y + 3z = 1
The reader can verify that points C and D indeed satisfy this equation.
Substitution of point A into the left hand side gives
2(1) − 5(1) + 3(1) = 5 − 5 = 0 6= 1
Thus A is not a point in the plane, and the four points are not coplanar.
96 1. VECTORS AND THE SPACE GEOMETRY

The other three faces can be found with the following normal vectors
−→ − −→ −→ −−→ −→ −−→
n2 = BA × BC, n3 = BA × BD, n4 = CA × CD. These vectors were
chosen simply to reduce the amount of computations. The necessary
vectors are
−→ −→ −−→
BA = OA − OB = h1, 1, 1i − h1, 2, 3i = h0, −1, −2i
−→ −→ −→
CA = OA − OC = h1, 1, 1i − h2, 0, −1i = h−1, 1, 2i
−−→ −−→ −→
CD = OD − OC = h3, 1, 0i − h2, 0, −1i = h1, 1, 1i
The necessary normal vectors are
 
ê1 ê2 ê3      
−1 −2 0 −2 0 −1
n2 = det 0 −1 −2 = hdet
  , − det , det i = h0, −2, 1i
−2 −4 1 −4 1 −2
1 −2 −4
 
ê1 ê2 ê3      
−1 −2 0 −2 0 −1
n3 = det  0 −1 −2 = hdet , − det , det i = h1, −4, 2i
−1 −3 2 −3 2 −1
2 −1 −3
 
ê1 ê2 ê3      
1 2 −1 2 −1 1
n2 = det −1 1 2  = hdet , − det , det i = h−1, 3, −2i
11 1 1 1 1
1 1 1
−−→
The right hand side for each plane can be calculated as d2 = n2 · OB,
−−→ −→
d3 = n3 · OB, and d4 = n4 · OC. The points were chosen to be the ones
that ”repeated” in the cross product vectors. The constants are, then,
d2 = h0, −2, 1i · h1, 2, 3i = 0(1) − 2(2) + 1(3) = −1
d3 = h1, −4, 2i · h1, 2, 3i = 1(1) − 4(2) + 2(3) = −1
d4 = h−1, 3, −2i · h2, 0, −1i = −1(2) + 3(0) − 2(−1) = 0
Thus, an equation for each plane (containing face XY Z) is:
BCD : 2x − 5y + 3z = 1
ABC : −2y + z = −1
ABD : x − 4y + 2z = −1
ACD : −x + 3y − 2z = 0
Let α, β, and γ denote the angle between plane BCD and plane ABC,
ABD, or ACD, respectively. The angle between two planes, whose
normal vectors are n1 and n2 is:
|n1 · n2 |
θ = arccos
kn1kkn2 k
6. PLANES IN SPACE 97

Then,
|h2, −5, 3i · h0, −2, 1i| |2(0) − 5(−2) + 3(1)| 13
α = arccos p p = √ √ =√ √
22 2 2 2
+ (−5) + 3 0 + (−2) + 1 2 2 38 5 38 5
|h2, −5, 3i · h1, −4, 2i| |2(1) − 5(−4) + 3(2)| 28
β = arccos p p = √ √ =√ √
22 + (−5)2 + 32 12 + (−4)2 + 22 38 21 38 21
|h2, −5, 3i · h−1, 3, −2i| |2(−1) − 5(3) + 3(−2)| 23
γ = arccos p p = √ √ =√ √
22 + (−5)2 + 32 (−1)2 + 32 + (−2)2 38 21 38 14

9. Find equations of all planes that are perpendicular to the line


through (1, −1, 1) and (3, 0, −1) and that are at the distance 2 from
the point (1, 2, 3).

Solution: Since the planes in question are perpendicular to the line


through two given points, a normal vector of all such planes may be
chosen as the vector whose initial and terminal points are the given
points, n = h3 − 1, 0 − (−1), −1 − 1i = h2, 1, −2i, so that knk = 3. All
planes perpendicular to the line are described by equations
2x + y − 2z = d
where d is real. So the question is reduced to determining the values
of d at which the distance D from the point with the position vector
r1 = h1, 2, 3i to the plane is D = 2. If r0 is the position vector of a
particular point in the plane, then d = n · r0 . The position vector of
the point (1, 2, 3) relative to a particular point in the plane is r1 − r0 .
The distance in question is the absolute value of the scalar projection
of this vector onto n:
|n · (r1 − r0 )| |n · r1 − d| |2 + 2 − 6 − d| |d + 2|
D= = = = ,
knk knk 3 3
D = 2 ⇔ |d + 2| = 6 ⇔ d + 2 = ±6 ⇔ d = −2 ± 6
So equations of the planes in question are 2x + y − 2z = −8 and
2x + y − 2z = 4.

10. Find an equation for the set of points that are equidistant from
the points (1, 2, 3) and (−1, 2, 1). Give a geometrical description of the
set.

Solution: Because we’re in the section on planes, this is obviously


a plane. However, an intuition for why this is a plane can be made
by dropping down into the second dimension. Consider two distinct
98 1. VECTORS AND THE SPACE GEOMETRY

points. Connect the two with a line segment, then draw a normal line
through the midpoint of this segment. This line is the set of all points
equidistant to the two given points. This can be seen by choosing a
point, and making line segments from it to the two given ones. Two
triangles will be made, which are clearly congruent. Extending this
idea into the third dimension transforms the line into a plane. The
algebra for it is as follows:
−→
Let A = (1, 2, 3), B = (−1, 2, 1), and n = AB. Then the plane through
−→
A whose normal is n, PA , is given by n · r = n · OA. Similarly, the
−−→
plane through B whose normal is n, PB , is given by n · r = n · OB.
−→ −−→
Consider the plane P given by n · r = n · (OA + OB)/2. Let P be a
point in this plane. Then the distance between A and P is given by
−→ −−→
D = kAP k. Consider the point M in P such that AM is normal to P.
−−→
Then kAMk is the distance between PA and P, which is
−→ − −→ −→ −→ −−→ −→ −→ −→
−−→ |d2 − d1 | |n · (OA + OB)/2 − n · OA| |n · (−OA + OB)| |AB · AB| kABk
kAMk = = −→ = −→ = −→ =
knk kABk 2kABk 2kABk 2
−−→ −→ −−→
A nearly identical argument shows that kBMk = kABk/2 = kAMk.
Thus M is equidistant to both A and B, and AM, BM are normal
to the plane. This implies that AM and BM are orthogonal to MP .
−−→
Consider the distance D0 = kBP k. Triangles AMP and BMP are
congruent by SAS (|AM| = |BM|, |MP | = |MP |, and a π/2 angle
enclosed between these sides), thus |AP | = |BP |, and D0 = D. So P
is the set of points equidistant to A and B.
−−→ −→
In this problem, we have that n = OB − OA = h−1, 2, 1i − h1, 2, 3i =
−→ −−→
h−2, 0, −2i and 1/2(OA+OB) = 1/2(h1, 2, 3i+h−1, 2, 1i) = 1/2h0, 4, 4i =
h0, 2, 2i. The equation for the set of points that are equidistant from
the given ones is the plane
h−2, 0, −2i·r = h−2, 0, −2i·h0, 2, 2i ⇔ −2x−2z = −2(0)+0(2)−2(2) ⇔ x+z = 2

11. Find an equation of the plane that is perpendicular to the plane


x + y + z = 1 and contains the line through the points (1, 2, 3) and
(−1, 1, 0).

Solution: A plane is determined by a particular point P0 in the


plane and a normal n (a non-zero vector perpendicular to the plane).
6. PLANES IN SPACE 99

Since the plane contains the segment AB, one can take
P0 = A = (1, 2, 3) .
The vector n1 = h1, 1, 1i is a normal of the given plane. Since the
plane in question is perpendicular to the given plane, their normals
must be perpendicular. Furthermore, the normal n is perpendicular
−→
to the vector AB = h−2, −1, −3i because the segment AB lies in the
plane in question. By the geometrical properties of the cross product,
−→ −→
the vector n1 × AB is perpendicular to both n1 and AB and can be
taken as a normal to the plane in question:
 
n ⊥ n1
 ê 1 ê2 ê3
−→
−→ ⇒ n = n1 × AB = det  1 1 1  = h−2 , 1 , 1i
n ⊥ AB −2 −1 −3
Therefore an equation of the plane in question reads:
−2(x − 1) + (y − 2) + (z − 3) = 0 ⇒ −2x + y + z = 3 .
12. To which of the planes x + y + z = 1 and x + 2y − z = 2 is the
point (1, 2, 3) the closest?

Solution: The point A = (1, 2, 3) is not in the plane x + y + z = 1


because its coordinates do not satisfy the equation of the plane, but A
lies in the second plane because 1 + 2 · 2 − 3 = 2. Thus, A is the closest
to the second plane.
Alternative solution: The plane x + y + z = 1 has a normal
n1 = h1, 1, 1i and passes through the point P1 = (1, 0, 0). Therefore
the distance from the point A = (1, 2, 3) to this plane is
−−→
|n1 · P1 A| |h1, 1, 1i · h0, 2, 3i| |2 + 3| 5
D1 = = = √ =√ .
kn1k kh1, 1, 1ik 3 3
Similarly, the plane x + 2y − z = 2 has a normal n2 = h1, 2, −1i and
passes through the point P2 = (2, 0, 0). Therefore the distance from
the point A = (1, 2, 3) to this plane is
−−→
|n2 · P2 A| |h1, 2, −1i · h−1, 2, 3i| | − 1 + 4 − 3|
D2 = = = √ = 0 < D1 .
kn2k kh1, 2, −1ik 6
Thus, the point A is closer to the second plane.

13–15. Give a geometrical description of each of the following fam-


ilies of planes where c is a numerical parameter.
100 1. VECTORS AND THE SPACE GEOMETRY

13. x + y + z = c .

Solution: This is the collection of all planes whose normal


vector is h1, 1, 1i.

14. x + y + cz = 1 .

Solution: This is the collection of all planes whose normal


vector is h1, 1, ci. Each plane intersects the coordinate axes
at (1, 0, 0), (0, 1, 0), and (0, 0, 1/c). Furthermore, each plane
contains the line x + y = 1, seen by substituting x + y = 1 and
z = 0 into the equation.

15. x sin c + y cos c + z = 1 .

Solution: This is the collection of all planes whose normal


vector is hsin c, cos c, 1i, passing through the z axis at (0, 0, 1).
This normal vector can be seen as a sum of a unit vector in
the xy plane and ê3. Each plane is tangent to the unit circle
in the xy plane. Therefore the planes can be seen as tak-
ing one plane (e,g, y + z = 1) and rotating it about the z
axis. The space enclosed by the union of all these planes and
the disk {(x, y, z)|x2 + y 2 = 1, z = 0} is in the shape of a cone.

16. Find values of c for which the plane x + y + cz = 1 is closest to the


point P = (1, 2, 1) and farthest from P .

Solution: An equivalent problem is to find the constants c that give


a maximum or minimum distance between the plane x + y + cz = 1
and a parallel plane through the point P . The distance between two
parallel planes n · r = d1 and n · r = d2 is
|d2 − d1 |
D=
knk
We have that n = h1, 1, ci. For one plane, set d1 = 1. For the plane
−→
passing through P , let d2 = n · OP = h1, 1, ci · h1, 2, 1i = 1(1) + 1(2) +
c(1) = 3 + c. Thus the distance between them is
|3 + c − 1| |c + 2|
D=√ =√
12 + 12 + c2 2 + c2
Clearly, D is minimal when c = −2. This is the case when the plane
x + y + cz = 1 is coincidental with the plane passing through P (they
are the same plane). Finding when D is maximal could be done by
6. PLANES IN SPACE 101

differentiating D with respect to c, finding the zeroes, and so on. But


that is tedious, so instead let’s do a geometric investigation. Notice
that in all cases, when z = 0, we have that x + y = 1. Thus, no matter
what c is, the plane will always intersect the line x + y = 1 in the xy
plane (see problem 14). This implies that varying c rotates the plane
along an axis produced by the line x + y = 1 in the xy plane. Start
with the plane passing through P . The plane can be divided into two
”parts” by the line x + y = 1; one part with z > 0 and one with z ≤ 0.
In this state, one part goes through P and the other one is maximally
far away. Rotating the plane causes the part that had intersected with
P to move away from P , and the part that was farthest from P to move
towards it. Eventually, you will rotate the plane until both parts are
equidistant; this is when there exists a point Q on the line x+y = 1 such
−→
that P Q is normal to the plane. Continuing to rotate the plane will
bring one of the parts closer to P again, thus the previous state yields
a maximum distance between the plane and P . Thus, let Q be a point
on the line x+y = 1 in the xy plane. So Q satisfies (a, 1−a, 0) for some
−→
real a. Next, we require P Q to be parallel to the normal of the plane.
−→ −→ −→
Thus, P Q = OQ − OP = sh1, 1, ci ⇔ ha − 1, −1 − a, −1i = sh1, 1, ci
for some real s. Setting the first two components equal to each other
yields a − 1 = s and −1 − a = s. So a − 1 = −1 − a ⇔ a = 0. Thus
s = −1. Equating the third components gives −1 = sc, so c = 1.

17. Consider three planes with normals n1 , n2 , and n3 such that each
pair of the planes is intersecting. Under what condition on the normals
are the three lines of intersection parallel or even coincide?

Solution: Consider planes P1 and P2, whose normals are n1 and


n2 , respectively. Then line L12 , formed from the intersection of planes
P1 and P2, must be parallel to a vector that is perpendicular to both
n1 and n2 . Let P and Q be two distinct points on the line. Then P and
−→ −→
Q also lie in P1, so P Q is parallel to the plane. Thus P Q is perpen-
dicular to n1 . The same argument applies for plane 2. It follows that
−→
P Q is parallel to n1 × n2 . Suppose that the three lines of intersection
formed from the three planes pairwise intersecting are parallel. Then
n1 × n2 || n1 × n3 || n2 × n3 . Note that since n1 × n2 and n2 × n3 are
parallel, their cross product is 0. Hence,

(n1 × n2 ) × (n2 × n3 ) = 0
102 1. VECTORS AND THE SPACE GEOMETRY

By the BAC-CAB rule,


(n1 ×n2)×(n2 ×n3) = n2 ((n1 ×n2)·(n3 ))−n3((n1 ×n2 )·(n2)) = n2 ((n1×n2 )·(n3 ))
It easily follows that
n2 ((n1 × n2 ) · (n3 )) = 0
Because n2 is not 0 (as required in the definition of a plane), it must
be that (n1 × n2) · (n3 ) = 0. Thus the normals must be coplanar.

18. Find equations of all the planes that are perpendicular to the
plane x + y + z = 1, have the angle π/3 with the plane x + y = 1, and
pass through the point (1, 1, 1).

Solution: Let n = hnx , ny , nz i be the normal vector of any such


plane. By the first condition, the normal vectors of the two planes
must be orthogonal. So, nx + ny + nz = 0. By the second condition,
the normal vectors must have an angle of π/3. So,
√ q
|hnx , ny , nz i · h1, 1, 0i| 2
cos π/3 = p √ ⇔ |nx +ny | = n2x + n2y + n2z
2 2 2
nx + ny + nz 1 + 1 2 2 2
However, we can do the following
√ q
2
|nx + ny | = n2x + n2y + n2z
2
√ q
2|nx + ny | = n2x + n2y + n2z
2(nx + ny )2 = n2x + n2y + n2z
2(nx + ny )2 = n2x + 2nx ny − 2nx ny + n2y + n2z
2(nx + ny )2 = (nx + ny )2 + n2z − 2nx ny
(nx + ny )2 − n2z = −2nx ny
(nx + ny − nz )(nx + ny + nz ) = −2nx ny
0 = −2nx ny
where the result from the first condition has been used in the second
to last line. It follows that at least one of nx or ny (or both) must be
zero. If both nx and ny are zero, then by the original deduction from
the second condition, it must be that nz = 0. This, however, would
result in the normal vector being the zero vector. Thus only one of nx
or ny must be zero.
6. PLANES IN SPACE 103

Any such plane will be described by an equation of the form


nx x + ny y + nz z = d
for some real number d. By the third condition (1, 1, 1) is a point on
the plane and therefore must satisfy this equation. So nx + ny + nz =
d ⇔ d = 0, by the first condition.

Suppose first that nx = 0. Returning to the first condition, we know


that nx + ny + nz = 0. Since nx = 0, we have ny = −nz . In this case,
the normal vector is h0, −nz , nz i = nz h0, −1, 1i for some real nz . Thus
an equation for the plane would be
nz (−y + z) = 0 ⇔ y = z
Suppose now that ny = 0. A nearly exact argument shows that nx =
−nz . In this case, the normal vector is h−nz , 0, nz i = nz h−1, 0, 1i for
some real nz . Thus an equation for the plane would be
nz (−x + z) = 0 ⇔ x = z

19. Let a = h1, 2, 3i and b = h1, 0, −1i. Find an equation of the plane
that contains the point (1, 2, −1), the vector a and a vector orthogonal
to both a and b.

Solution: A vector orthogonal to both a and b is a × b. We can


therefore let n = a × (a × b). We can compute n using the BAC-CAB
rule as follows:
n = a × (a × b) = a(a · b) − b(a · a) = a(a · b) − bkak2
In this problem, with a = h1, 2, 3i and b = h1, 0, −1i, we have that
a · b = h1, 2, 3i · h1, 0, −1i = 1(1) + 2(0) + 3(−1) = −2
√ √
kak = 12 + 22 + 32 = 14
n = −2h1, 2, 3i − 14h1, 0, −1i = h−16, −4, 8i = −4h4, 1, −2i
The general form of a plane is
n·r=d
for a general position vector r. Thus if P = (1, 2, −1) is a point on the
plane, its position vector must satisfy the above equation. It follows
that
−→
d = n· OP = −4h4, 1, −2i·h1, 2, −1i = −4(4(1)+1(2)−2(−1)) = −4(8)
104 1. VECTORS AND THE SPACE GEOMETRY

Thus an equation for the plane is


−4(4x + y − 2z) = −4(8) ⇔ 4x + y − 2z = 8
20. Consider the plane P through three points A = (1, 1, 1), B =
(2, 0, 1) and C = (−1, 3, 2). Find all the planes that contain the seg-
ment AB and have the angle π/3 with the plane P. Hint: see Study
Problem ??.

Solution: To find the angle between two planes, we must first find
−→ −→
their normal vectors. The normal of P is parallel to AB × AC, where
−→ −→
AB = h2, 0, 1i − h1, 1, 1i = h1, −1, 0i and AC = h−1, 3, 2i − h1, 1, 1i =
−→
h−2, 2, 1i = −2AB + h0, 0, 1i. Thus the cross product reduces to
−→ −→ −→
AB × (−2AB + h0, 0, 1i) = AB × h0, 0, 1i. We have
 
ê1 ê2 ê3      
−→ −1 0 10 1 −1
AB×h0, 0, 1i = det 1 −1 0 = hdet
  , − det , det i = h−1, −1, 0i
0 1 01 0 0
0 0 1
So we may set the normal of P to be n1 = h1, 1, 0i.
Next, let the normal of the plane in question be given by n = hnx , ny , nz i
where n is a normal vector. Then the angle between the two planes is
|n1 · n| |nx + ny |
cos θ = = √
kn1 kknk 2
It follows that since θ = π/3 that

2
|nx + ny | =
2
−→
Next, we require AB to be in the plane. Therefore AB must be or-
thogonal to its normal. So,
−→
AB · n = 0 ⇔ nx − ny = 0
Squaring each obtained equation gives
1
n2x + n2y + 2nx ny =
2
n2x + n2y − 2nx ny = 0
Adding the two, then multiplying by 1/2 gives
1
n2x + n2y =
4
6. PLANES IN SPACE 105

Since n is a unit vector, it must be that



q
2 2 2
1 2 3
nx + ny + nz = 1 ⇔ ( ) + nz = 1 ⇔ nz = ±
4 2
To find nx and ny , we must have two different cases.

Case 1: nx + ny = 1
We also have that nx − ny = 0. Adding the two then multiplying by
1/2 gives nx = 12 . So ny = 21 . The possible normal vectors then are

1 1 3
n11 = h , , i
2 2 2

1 1 3
n12 = h , , − i
2 2 2

Case 2: nx + ny = −1
We also have that nx − ny = 0. Adding the two then multiplying by
1/2 gives nx = − 21 . So ny = − 21 . The possible normal vectors then are

1 1 3
n21 = h− , − , i
2 2 2

1 1 3
n22 = h− , − , − i
2 2 2
However, notice that n21 = −n12 and n22 = −n11. Therefore the planes
generated in Case 1 are identical to those in Case 2.

The planes then are


−→ −→
n11 · r = n11 · OA, n12 · r = n12 · OA

21. Find an equation of the plane that contains the line through (1, 2, 3)
and (2, 1, 1) and cuts the sphere x2 + y 2 + z 2 − 2x + 4y − 6z = 0 into
two hemispheres.

Solution: In order for a plane to cut a sphere into two hemisphere,


the plane must contain the center of the sphere. By completing the
squares, the given equation is reduced to the standard form
(x − 1)2 − 1 + (y + 2)2 − 4 + (z − 3)2 − 9 = 0
⇒ (x − 1)2 + (y + 2)2 + (z − 3)2 = 14
106 1. VECTORS AND THE SPACE GEOMETRY

So, the center of the sphere is C = (1, −2, 3). Since the plane in
question contains there known points, a normal n of the plane is pro-
portional to the cross product
−→ −→
AB × AC = h1, −1, −2i × h0, −4, 0i = h−8, 0, −4i = −4h2, 0, 1i
⇒ n = h2, 0, 1i

Therefore, taking A as a particular point of the plane, the equation


reads
2(x − 1) + (z − 3) = 0 or 2x + z = 5 .
22. Find all planes perpendicular to n = h1, 1, 1i whose intersection
with the ball x2 + y 2 + z 2 ≤ R2 is a disk of area πR2 /4.

Solution: For the intersection to be a disk of area πR2/4, it must


be that the radius of the disk is r = R/2. Construct a line L paral-
lel to n = h1, 1, 1i that goes through the center of the sphere. There
are two points on the line that, when a plane perpendicular to the
line passes through the point, a disk of radius r = R/2 is formed.
Next, consider the triangle connecting one such point, the center of
the sphere, and a point where the disk of radius r = R/2 is tangent
to the sphere. It will have one leg of length R/2 with a hypotenuse
of length p
R. By the pythagoren
p theorem, √ the length of the other leg
2 2 2
must be R − (R/2) = 3R /4 = 3R/2. This is the distance
between√the center and √one of the two points. The length of n is
knk = 12 + 12 + 12 = 3. Therefore we can find the two points by
traveling parallel and antiparallel to R/2n from the origin. This gives
P = (R/2, R/2, R/2) and Q = (−R/2, −R/2, −R/2) as the two points.
An equation for each plane is as follows

x + y + z = h1, 1, 1i · h±R/2, ±R/2, ±R/2i = ±3R/2

23. Find an equation of the plane that is tangent to the sphere


x2 + y 2 + z 2 − 2x − 4y − 6z + 11 = 0 at the point (2, 1, 2). Hint:
What is the angle between a line tangent to a circle at a point P and
the segment OP where O is the center of the circle? Extend this ob-
servation to a plane tangent to a sphere to determine a normal of the
tangent plane.
−→
Solution: The hint shows that OP is perpendicular to a line tan-
gent to the circle at P . Therefore given a point P on a sphere, a plane
−→
tangent to the sphere at P will have a normal vector of OP .
6. PLANES IN SPACE 107

By completing the square, the given equation for the sphere becomes
(x − 1)2 − 1 + (y − 2)2 − 4 + (z − 3)2 − 9 + 11 = 0
⇒ (x − 1)2 + (y − 2)2 + (z − 3)2 = 3
The center of the circle then is O = (1, 2, 3). Therefore,
−→
OP = h2, 1, 2i − h1, 2, 3i = h1, −1, −1i
Thus an equation for a plane tangent to the given sphere at (2, 1, 2) is
x − y − z = h1, −1, −1i · h2, 1, 2i = 1(2) − 1(1) − 1(2) = −1
24. Find the family of planes through the point (0, 0, a), a > R, that
are tangent to the sphere x2 + y 2 + z 2 = R2 . Hint: Compare this family
with the family of planes in Exercise 15.

Solution: First notice that the family of planes can be generated


by taking one plane that passes through (0, 0, a) (for a > R) that is
tangent to the sphere and rotating it about the z axis. Choose a plane.
Suppose it is tangent to the point P on the sphere. Rotating this plane
around the z axis corresponds to translating P along a circular path.
This circular path lies in another plane containing P , parallel to the
xy plane.

We are interested in finding the radius of this circular path, r. Because


the path is parallel to the xy plane, it must be that there exists a point
on the sphere whose y coordinate is 0. Call this point S = (r, 0, sz ).
Construct a line from (0, 0, a) to S. By construction, this line lies in
one of the planes in the family, so it must be that this line is tangent
to the circular path, and thus the sphere, at S. Let the line be given
by z = −cx + a in the xz plane. The line intersects the sphere at the
points whose x coordinate satisfy
x2 + 02 + (−cx + a)2 = R2 ⇔ (1 + c2 )x2 + (−2ca)x + (a2 − R2 ) = 0
For the line to be tangent, we only want one point. Thus there must be
only one x coordinate, and only one solution to the above quadratic.
This is achieved when the discriminant vanishes,

−(−2ca)2 − 4(1 + c2 )(a2 − R2 ) = 0 ⇔ c = a2 − R2 /R
It follows that

a a2 −R2
√ √
−(−2ca) R aR a2 − R2 R a2 − R2
r= = 2 2 = =
2(1 + c2) 1 + a R−R
2
R2 + a2 − R2 a
108 1. VECTORS AND THE SPACE GEOMETRY

We can also find sz as follows:



2 2 R a2 − R2 2 R4
r +0 + =R ⇔s2z 2
s2z
= R −(2
) = 2
a a
2
So sz = R /a, where the positive root has been chosen since a > R > 0.
The constant sz can be interpreted as the height of the circular path
above the xy plane.
−→
We know that the normal of the plane must be parallel to OP . We can
−→
easily find OP using our construction. First, recognize that it is the
sum of two vectors, sz ê3 and the vector connecting the center of the
circular path to P (this latter vector is also horizontal). Notice that
a vector from the center of the circular path to a point on the path
will be of the form rhcos b, sin b, 0i for some b. The quantity b revolves
the vector around the circular path, achieving the revolution necessary.
This vector is very similar in form to that of exercise 15. Thus,
√ √
−→ R a2 − R2 R a2 − R2 R2
OP = rhcos b, sin b, 0i+sz ê3 = h cos b, sin b, i
a a a
Finally, the plane must pass through the point (0, 0, a). Therefore
−→
d = OP · h0, 0, ai = R2 . It follows that an equation for a distinct plane
in the family is
√ √
R a2 − R2 R a2 − R2 R2
( cos b)x + ( sin b)y + z = R2
a a a
which can be reduced to
√ √
( a2 − R2 cos b)x + ( a2 − R2 sin b)y + Rz = aR

25. Consider a sphere of radius R centered at the origin and two points
P1 and P2 whose position vectors are r1 and r2 . Suppose that kr1k > R
and kr2k > R (the points are outside the sphere). Find the equation
n·r = d of the plane through P1 and P2 whose distance from the sphere
is maximal. What is the distance? Hint: Show first that a normal of
the plane can always be written in the form n = r1 + c(r2 − r1 ). Then
find a condition to determine the constant c.
7. LINES IN SPACE 109

7. Lines in Space
1–7. Find vector, parametric, and symmetric equations of the specified
line.

1. The line containing the segment AB where A = (1, 2, 3) and


B = (−1, 2, 4).

Solution: The vector, parametric, and symmetric equations of a line


parallel to a vector v = hv1, v2, v3 i through a point P = (x0 , y0, z0) are
r = r0 + tv − ∞ < t < ∞
x = x0 + tv1, y = y0 + tv2, z = z0 + tv3, −∞ < t < ∞
x − x0 y − y0 z − z0
= =
v1 v2 v3
Note that the last one, the symmetric equation of a line, only applies
if v 6= 0. If one component is zero, say v1 = 0, then the symmetric
equation becomes:
y − y0 z − z0
x = x0 , =
v2 v3
Therefore we need only find v and P . v is a vector parallel to the line,
−→
so we may simply choose v = AB = h−1, 2, 4i − h1, 2, 3i = h−2, 0, 1i.
P is any point on the line, so we may choose P = A = (1, 2, 3). Thus
we have
r = h1, 2, 3i + th−2, 0, 1i, −∞ < t < ∞
x = 1 − 2t, y = 2, z = 3 + t, −∞ < t < ∞
x−1
= z − 3, y = 2
−2
2. The lines containing the diagonals of the parallelogram whose adja-
cent sides at the vertex (1, 0, −1) are a = h1, 2, 3i and b = h−1, 2, 1i.

Solution: Let the vertices of the parallelogram be OABC, where


−→ −−→
O = (1, 0, −1), a = OA, and b = OB. Then the final vertex C is given
by a + b. It follows that a line parallel to diagonal OC is parallel to
a + b = h1, 2, 3i + h−1, 2, 1i = h0, 4, 4i. Thus for the first line we have
that a vector parallel to it is v1 = h0, 1, 1i. We must now choose any
point on the line, say P1 = O = (1, 0, −1). So the vector, parametric,
and symmetric equations of the line through diagonal OC are given by
r = h1, 0, −1i + th0, 1, 1i, −∞ < t < ∞
x = 1, y = t, z = −1 + t, −∞ < t < ∞
110 1. VECTORS AND THE SPACE GEOMETRY

x = 1, y = z + 1
A vector parallel to diagonal AB is given by a − b = h1, 2, 3i −
h−1, 2, 1i = h2, 0, 2i. Thus a vector parallel to this line is v2 = h1, 0, 1i.
We must now find a point on the line. We can choose P2 = A =
(1, 0, −1) + (1, 2, 3) = (2, 2, 2). So the vector, parametric, and symmet-
ric equations of the line through diagonal AB are given by
r = h2, 2, 2i + th1, 0, 1i, −∞ < t < ∞

x = 2 − t, y = 2, z = 2 − t, −∞ < t < ∞
x−2 z−2
= , y=2 ⇔ x=z y=2
−1 −1
3. The line through the vertex A of a triangle ABC and perpendic-
ular to the sides AB and AC if A = (1, 0, −1), B = (−1, 1, 2), and
C = (2, −1, −2)

Solution: A line is determined by a particular point P0 in it and


a non-zero vector v parallel to the line. Since the line passes through
A,
P0 = A = (1, 0, −1) = (x0, y0, z0 ) .
−→ −→
The vector v must be perpendicular to the vectors AB and AC and
hence, by the geometrical properties of the cross product, be parallel
to their cross product:
 
−→ )
ê1 ê2 ê3
v ⊥ AB = h−2, 1, 3i −→ −→
−→ ⇒ v = AB×AC = det −2 1 3  = h2, 1, 1i .
v ⊥ AC = h1, −1, −1i 1 −1 −1
The symmetric and parametric equations of the line read, respectively,
x − x0 y − y0 z − z0 x−1
= = ⇒ =y =z+1
v1 v v3 2
2 
 x = x0 + v 1 t  x = 1 + 2t
y = y0 + v 2 t ⇒ y=t
z = z0 + v 3 t z = −1 + t
 

4. The line through the vertex C of a triangle ABC and parallel to the
edge AB if A = (1, 0, −1), B = (−1, 1, 2), and C = (2, −1, −2)
−→
Solution: The line in question must be parallel to AB. We have
−→
that AB = h−1, 1, 2i − h1, 0, −1i = h= h−2, 1, 3i. Thus we can
set v = h−2, 1, 3i. Since the line must pass through C, we can let
7. LINES IN SPACE 111

P = C = (2, −1, −2). Thus the vector, parametric, and symmetric


equations of a line through C parallel to side AB are given by
r = h2, −1, −2i + th−2, 1, 3i, −∞ < t < ∞
x = 2 − 2t, y = −1 + t, z = −2 + 3t, −∞ < t < ∞
x−2 z+2
=y+1 =
−2 3
5. A line through the origin that makes an angle of 60◦ with the x and
y axes. Is such a line unique? Explain.

Solution: First, since such a line must pass through the origin, we
can have P = (0, 0, 0) for all such lines. Next, such a line is not unique.
Let v = hv1 , v2, v3i be a unit vector parallel to the line. The angle
between two lines is defined to be
cos(θ) = |v̂1 · v̂2 |
Note that the x and y axes can be represented as lines parallel to ê1
and ê2, respectively, through the origin. So we must have that
1
|v · ê1 | = |v1| = cos 60◦ =
2
1
|v · ê2 | = |v2| = cos 60◦ =
2
So we have that v = h±1/2, ±1/2, v3 i. Since we required v to be a
unit vector, it must be that
q
v12 + v22 + v32 = 1
1/4 + 1/4 + v32 = 1
v32 = 1/2

v3 = ± 2/2

Therefore we have that u = h±1, ±1, ± 2i is a vector parallel to such
a line (note that u is simply 2v). Thus the vector, parametric, and
symmetric equations for such a line are given by

r = h0, 0, 0i + th±1, ±1, ± 2i, −∞ < t < ∞

x = ±t, y = ±t, z = ± 2t, −∞ < t < ∞
z
±x = ±y = ± √
2
3
Note: There are 2 = 8 combinations for the plus and minus, but since
−v is parallel to v, there are only four distinct lines.
112 1. VECTORS AND THE SPACE GEOMETRY

6. The line through the vertex A = (1, 2, 3) of a parallelogram with


adjacent sides at A being a = h1, 2, 2i and b = h−2, 1, −2i that bisects
the angle of the parallelogram at A. Hint: See Exercise 21 in Section
??.

Solution: By Exercise 21 in Section ??, we know that c = kbka +


kakb bisects the angle between vectors a and b. Thus we have
p √
c = (−2)2 + 12 + (−2)2 h1, 2, 2i+ 12 + 22 + 22 h−2, 1, −2i = 3h−1, 3, 0i
The line in question must be parallel to this vector. Thus we may
choose v = h−1, 3, 0i. Since the line must pass through A = (1, 2, 3),
we may choose P = A. Thus the line through vertex A that bisects
the angle of the parallelogram at A is given by
r = h1, 2, 3i + th−1, 3, 0i, −∞ < t < ∞
x = 1 − t, y = 2 + 3t, z = 3, −∞ < t < ∞
y−2
1−x = , z=3
3
7. The line parallel to the vector h1, −2, 0i that contains a diameter of
the sphere x2 + y 2 + z 2 − 2x + 4y − 6z = 0.

Solution: By completing the square, the given equation for the sphere
becomes
(x − 1)2 − 1 + (y + 2)2 − 4 + (z − 3)2 − 9 = 0
⇒ (x − 1)2 + (y + 2)2 + (z − 3)2 = 14
Any line that contains a diameter of the sphere must include its center.
Thus any such line passes through (1, 2, 3). We may choose this point
as P . We are given the vector v that the line is parallel to. Thus we
have that the vector, parametric, and symmetric equations for such a
line are
r = h1, 2, 3i + th1, −2, 0i, −∞ < t < ∞
x = 1 + t, y = 2 − 2t, z = 3, −∞ < t < ∞
2−y
x−1 = , z=3
2
8. Show that the line through P1 = (1, 2, −1) and parallel to v1 =
h1, −1, 3i coincides with the line through P2 = (0, 3, −4) and parallel
to v2 = h−2, 2, −6i as points sets in space.

Solution: Since v1 = −2v2 , the two lines are at least parallel. They
coincide now iff they share the given points P1 and P2 . If they share
7. LINES IN SPACE 113

−−→
P1 and P2 , then P1 P2 must be parallel to v1 or v2 (it doesn’t matter
which since they are parallel). Thus we have
−−→
P1 P2 = h0, 3, −4i − h1, 2, −1i = h−1, 1, −3i = −v1
−−→
So P1 P2 is parallel to v1. Thus the lines coincide.

9. Do the lines x − 1 = 2y = 3z and r = r0 + tv, where r0 = h7, 3, 2i


and v = h6, 3, 2i, coincide as point sets in space?

Solution: Converting the second line into its symmetric equation


yields
x−7 y−3 z−2
= = ⇔ x−7 = 2y −6 = 3z −6 ⇔ x−1 = 2y = 3z
6 3 2
Since the lines share the same symmetric equation, they must be the
same line. Hence they coincide.

10. Find parametric equations of the line through the point (1, 2, 3)
and perpendicular to the plane x + y + 2z = 1. Find the point of
intersection of the line and the plane.

Solution: If the line is perpendicular to the given plane, then it


must be parallel to the plane’s normal vector, n = h1, 1, 2i. We may
choose P = (1, 2, 3). Thus the parametric equations of the line are
given by
x = 1 + t, y = 2 + t, z = 3 + 2t, −∞ < t < ∞
We can find the value of t when the line intersects with the plane as
follows
4
(1 + t) + (2 + t) + 2(3 + 2t) = 1 ⇔ 6t + 9 = 1 ⇔ t = −
3
So the point of intersection is
4 4 8 1 2 1
(1 − , 2 − , 3 − ) = (− , , )
3 3 3 3 3 3
11. Find parametric and symmetric equations of the line of intersec-
tion of the planes x + y + z = 1 and 2x − 2y + z = 1.

Solution: Since the line of intersection lies in both planes, a vec-


tor parallel to the line must be orthogonal to both n1 = h1, 1, 1i and
114 1. VECTORS AND THE SPACE GEOMETRY

n2 = h2, −2, 1i. Thus we must compute n1 × n2.


 
ê1 ê2 ê3      
1 1 1 1 1 1
n1 ×n2 = det  1 1 1  = hdet , − det , det i = h3, 1, −4i
−2 1 21 2 −2
2 −2 1
Thus we may set v = h3, 1, −4i. Now we must find a point of inter-
section. We are free to choose one coordinate at will, so long as the
corresponding component of v does not vanish (otherwise the paramet-
ric equation for this component is simply a constant). For example, we
may choose x0 = 0. By inspection, we can let z0 = 1 and it easily
follows that y0 = 0. So we may let P = (0, 0, 1). The parametric and
symmetric equations of the line of intersection then are
x = 3t, y = t, z = 1 − 4t, −∞ < t < ∞
x 1−z
=y=
3 4
12–13. Determine whether the given lines are are parallel, skew, or
intersecting. If they intersect, find the point of intersection and the
angle between the lines.
12. The line through the points (1, 2, 3) and (2, −1, 1) and the line
through the points (0, 1, 3) and (1, 0, 2)

Solution: Let L1 and L2 be two lines, parallel to v1 and


v2, respectively, and passing through P1 and P2 , respectively.
−−→
Let r12 = P1 P2 , where P1 is a point on L1 and P2 is a point
on L2 . Two lines can be skew, intersecting, parallel, or co-
incidental. To determine which, it is useful to first compute
r12 · (v1 × v2 ). Let P1 = (1, 2, 3) and P2 = (0, 1, 3). First,
v1 = h1, 2, 3i − h2, −1, 1i = h−1, 3, 2i
v2 = h1, 0, 2i − h0, 1, 3i = h1, −1, −1i
r12 = h0, 1, 3i − h1, 2, 3i = h−1, −1, 0i
The triple product then is
 
−1 −1 0      
3 2 −1 2 −1 3
r12·(v1×v2) = det −1 3 2  = − det +det +0 det =0
−1 −1 1 −1 1 −1
1 −1 −1
So the three vectors are coplanar, and therefore the lines are
not skew. Notice further that v1 and v2 are not parallel, so
the lines are neither parallel nor coincidental. Thus they are
7. LINES IN SPACE 115

intersecting. The parametric equations for each line are as


follows
x = 1 − t, y = 2 + 3t, z = 3 + 2t, −∞ < t < ∞
x = s, y = 1 − s, z = 3 − s, −∞ < t < ∞
Notice that two different parameters have been used. This
is because the lines may intersect at different ”times” (the
parameters may attain different values at the point of inter-
section). Equating each coordinate yields
 
 1−t = s  t − 1 = −s
2 + 3t = 1 − s ⇒ 2 + 3t = 1 − s
 3 + 2t = 3 − s  3 + 2t = 3 − s

We can then add 1 to each side of the first equation and sub-
stitute it into the second, which gives
2 + 3t = t ⇔ t = −1
Therefore the point of intersection is at
(1 − (−1), 2 + 3(−1), 3 + 2(−1)) ⇔ (2, −1, 1)
We can verify this is correct by solving for s. Using the first
equation, s = 1 − (−1) = 2. Substituting this into the para-
metric equations for y and z yield y = 1 − (2) = −1 and
z = 3 − (2) = 1, which are correct.

Next we must find the angle between the two lines. The angle
is defined to be
|v1 · v2|
cos θ =
kv1kkv2k
Thus the angle between the two lines is given by
|h−1, 3, 2i · h1, −1, −1i| | − 1(1) + 3(−1) + 2(−1)| 6
cos θ = p p = √ √ =√ √
(−1)2 2 2 2
+ 3 + 2 1 + (−1) + (−1)2 2 14 3 14 3
So θ = arccos √146√3 .

13. The lines x = 1 + 2t, y = 3t, z = 2 − t and x + 1 = y − 4 =


(z − 1)/3.

Solution: Using the intuition developed in Exercise 12, we


must find r12 · (v1 × v2 ). Let P1 = (1, 3, 2), a point on the first
116 1. VECTORS AND THE SPACE GEOMETRY

line (this is found by looking at the constant terms in the para-


metric equations). Let P2 = (−1, 4, 1), a point on the second
line (this is found by recalling the definition of the symmetric
equation of a line). We find v1 = h2, 3, −1i and v2 = h1, 1, 3i.
We calculate r12 as r12 = h−1, 4, 1i − h1, 3, 2i = h−2, 1, −1i
Then the triple product is
 
−2 1 −1      
3 −1 2 −1 23
r12·(v1×v2) = det 2 3 −1 = −2 det
  −det −det = −26
1 3 1 3 11
1 1 3
So the vectors are not coplanar. Since v1 and v2 are not par-
allel, it must be that the vectors are skew, and thus do not
intersect.

−→ −→ −−→
14. Let AB = h1, 2, 2i, AC = h2, −1, −2i, and AD = h0, 3, 4i be the
adjacent sides of a parallelepiped. Show that the diagonal of the paral-
lelepiped extended from the vertex A intersects the diagonal extended
from the vertex D and find the angle between the diagonals.

Solution: First, the diagonal extending from vertex A is parallel


−→ −→ −−→ −→ −→ −−→
to AB + AC + AD. So we may let v1 = AB + AC + AD. Second,
−→ −→ −−→
the diagonal extending from vertex D is parallel to AB + AC − AD.
−→ −→ −−→
So we may let v2 = AB + AC − AD. We may let P1 = A as a point
on the first line (the diagonal through A) and P2 = D as a point on
−−→ −−→
the second line (the diagonal through D). Let r12 = P1 P2 = AD. For
the diagonals to be intersecting, the triple product r12 · (v1 × v2) must
vanish. Computing the triple product yields:
−→ −→ −−→ −→ −→ −−→
r12 · (v1 × v2) = r12 · ((AB + AC + AD) × (AB + AC − AD))
−→ −→ −→ −→ −→ −→ −−→ −→ −→ −
−→
= r12 · ((AB + AC) × (AB + AC) − (AB + AC) × AD + (AB + AC) × AD
−−→ −−→
+AD × AD)
= r12 · 0 = 0
Next we must check if v1 and v2 are parallel. Their dot product is
−→ −→ −−→ −→ −→ −−→
v1 · v2 = (AB + AC + AD) · (AB + AC − AD)
−→ −→ −→ −→ −→ −→ −−→ −→ −→ − −→
= (AB + AC) · (AB + AC) − (AB + AC) · AD + (AB + AC) · AD
−−→ −−→
+AD · AD
−→ −→ −−→
= kAB + ACk2 + kADk2
7. LINES IN SPACE 117

Suppose that v1 and v2 are parallel. Then


|v1 · v2| = kv1kkv2k
−→ −→ −−→ −→ −→ −−→
= kAB + AB + ADkkAB + AB − ADk
−→ −→ −−→ −→ −→ −−→
≤ (kAB + ACk + kADk)(kAB + ACk − kADk)
−→ −→ −−→ −→ −→ −−→ −→ −→ −−→
≤ kAB + ACk2 + kADkkAB + ACk − kADkkAB + ACk − kADk2
−→ −→ −−→
≤ kAB + ACk2 − kADk2
where the Cauchy-Schwarz inequality has been used. Comparing this
to what we previously obtained yields
−→ −→ −−→ −→ −→ −−→ −−→
|kAB + ACk2 + kADk2 | ≤ kAB + ACk2 − kADk2 ⇔ 2kADk2 ≤ 0
−−→ −−→
Since kADk ≥ 0 by definition, it must be that kADk = 0. But this
is not true. Hence v1 and v2 are not parallel. Therefore they must
intersect.

To find the angle between v1 and v2, we must first compute them
as follows
−→ −→ − −→
v1 = AB + AC + AD = h1, 2, 2i + h2, −1, −2i + h0, 3, 4i = h3, 4, 4i
−→ −→ −−→
v2 = AB + AC − AD = h1, 2, 2i + h2, −1, −2i − h0, 3, 4i = h3, −2, −4i
The angle them is
|h3, 4, 4i · h3, −2, −4i| 15
θ = arccos( √ p ) = arccos √ √
32 + 42 + 42 32 + (−2)2 + (−4)2 41 29

15. Are the four lines containing the diagonals of a parallelepiped


intersecting at a point? Prove your answer. If they are intersecting,
find the position vector of the point of intersection relative to a vertex
at which the adjacent sides of the parallelepiped are a, b, and c.
−→ −→
Solution: Let A, B, C, D be points such that AB = b, AC = c, and
−−→
AD = c. Notice that in Exercise 14, the choice of using the diagonal
extended from point D was more or less arbitrary. We could have cho-
−→ −→ −−→
sen point B, and the vector would have been v3 = −AB + AC + AD,
or we could have chosen point C, and the vector would have been
−→ −→ − −→
v3 = AB − AC + AD. Note that all that changes is where the minus
sign is. If we chose point B, we could rename it to something like D0
and rename D to B 0. The argument then looks nearly identical to the
one presented in Exercise 15, with the exception of changing B to B 0
118 1. VECTORS AND THE SPACE GEOMETRY

and D to D0 . We can do the same thing if we had chosen point C.


Thus all the diagonal extending from each of B, C, and D intersect
the one extending from A.

Let us continue working with the diagonal extending from point D.


The line that contains this diagonal is given by the vector equation
−−→ −→ −→ − −→ −−→
r = OD + t(AB + AC − AD) = OD + t(a + b − c), −∞ < t < ∞
and the vector equation for the line containing the diagonal extended
from point A is given by
−→ −→ −→ −−→ −→
r = OA + s(AB + AC + AD) = OA + s(a + b + c), −∞ < s < ∞
note that a different parameter has to be used.

We know these lines intersect, so we can equate them


−−→ −→
OD + t(a + b − c) = OA + s(a + b + c)
−−→
AD + t(a + b − c) − s(a + b + c) = 0
c + (t − s)a + (t − s)b + (−t − s)c = 0
(t − s)a + (t − s)b + (1 − t − s)c = 0
The only way this can be true for any arbitrary noncoplanar, nonpar-
allel, nonzero vectors a, b, c must be if t − s = 0 and 1 − t − s = 0. So
t = s, and substituting this into the second equation yields
1
1 − s − s = 0 ⇔ 1 = 2s ⇔ s =
2
We can substitute this into the second vector equation to give
−→ 1
r = OA + (a + b + c)
2
Relative to A, this point of intersection is at 12 (a + b + c).
The same answer is produced by using the first equation, but some
tedious vector manipulations are involved.

The same argument as in the beginning paragraph can be used to


show this is where all the diagonals intersect.

16. Find vector and parametric equations of the straight line segment
from the point (1, 2, 3) to the point (−1, 1, 2).

Solution: Let A = (1, 2, 3) and B = (−1, 1, 2). The line segment from
−→
A to B must be parallel to AB = h−1, 1, 2i − h1, 2, 3i = h−2, −1, −1i.
7. LINES IN SPACE 119

Thus we may choose it to be v = h−2, −1, −1i. Since we want the line
segment to emanate from A, we should choose it as our representative
point. Thus the vector and parametric equations are:
r = h1, 2, 3i + th−2, −1, −1i
x = 1 − 2t, y = 2 − t, z = 3 − t
You may have noticed that I did not include an interval for the param-
eter t. This is because we must find it ourselves. We want t to range
−→ −−→
so that r ranges from OA to OB. Thus we have:
h1, 2, 3i = h1, 2, 3i + th−2, −1, −1i ⇔ t = 0
h−1, 1, 2i = h1, 2, 3i+th−2, −1, −2i ⇔ h−2, −1, −1i = th−2, −1, −2i ⇔ t = 1
The vector and parametric equations then are
r = h1, 2, 3i + th−2, −1, −1i, 0 ≤ t ≤ 1

x = 1 − 2t, y = 2 − t, z = 3 − t, 0 ≤ t ≤ 1
17. Let r1 and r2 be position vectors of two points in space. Find a
vector equation of the straight line segment from r1 to r2.

Solution: Clearly the line must be parallel to the vector from r1


to r2 . This vector is easily found to be r2 − r1 (use the parallelogram
law to visualize if necessary). We should choose the point whose posi-
tion vector is r1 as our point, because this is where the line starts from.
Thus we have
r = r1 + t(r2 − r1)
You may have noticed I did not include an interval for t. Every para-
metric equation must include an interval that defines the possible values
of t. Here, we must find it ourselves. We need r to range from r1 to
r2, so we have
r1 = r1 + t(r2 − r1) ⇔ t = 0
r2 = r1 + t(r2 − r1) ⇔ (1 − t)r2 = (1 − t)r1 ⇔ t = 1
Thus we have 0 ≤ t ≤ 1. So the vector equation is
r = r1 + t(r2 − r1 ), 0 ≤ t ≤ 1
18. Find the distance from the point (1, 2, 3) to the line 2x = y + 1,
z = 3.

Solution: For an arbitrary line, with v a vector parallel to it, P0


120 1. VECTORS AND THE SPACE GEOMETRY

a point on the line, and P1 an arbitrary point, the distance between


the line and P1 is given by
−−→
kv × P0 P1 k
D=
kvk
We find from the symmetric equation of the line that it is parallel to
v = h1, 2, 0i and passes through (0, −1, 3). Let P0 = (0, −1, 3) and
−−→
P1 = (1, 2, 3). Then P0 P1 = h1, 2, 3i − h0, −1, 3i = h1, 3, 0i. So,
 
ê1 ê2 ê3      
−−→ 2 0 1 0 12
v×P0P1 = det  1 2 0  = hdet , − det , det i = h0, 0, 1i
30 10 13
1 3 0
−−→
Then kv × P0 P1 k = 1. Furthermore,
√ √
kvk = 12 + 22 + 02 = 5
It follows that the distance is
1
D=√
5
19. Consider the plane x + y − z = 0 and a point P = (1, 1, 2) in it.
Find parametric equations of the lines through the origin that lie in the
plane and are at a distance of 1 unit from P . Hint: A vector parallel
to these lines can be taken in the form v = h1, c, 1 + ci where c is to be
determined. Explain why!

Solution: Let v be a vector parallel to the line. Then v is orthogonal


to the normal of the plane, n = h1, 1, −1i. Recall that the distance
between a line and a point P1 is given by
−−→
kv × P0 P1 k
D=
kvk
where v is parallel to the line and P0 is a point on the line. Since the line
in question passes through the origin, we may choose P0 = O = (0, 0, 0).
Note that P1 = (1, 1, 2) is a point on the plane, and the plane passes
−−→ −−→
through the origin. Then OP1 is in the plane. So OP1 is orthogonal
−−→
to the normal as well. Then v × P0 P1 is parallel to the normal of the
−−→
plane. Let v × P0 P1 = sn, where v = hvx , vy , vz i. Then,
 
ê1 ê2 ê3      
−−→ vy vz vx vz vx vy
v × P0 P1 = det vx vy vz = hdet
  , − det , det i
1 2 1 2 1 1
1 1 2
= h2vy − vz , vz − 2vx , vx − vy i
7. LINES IN SPACE 121

Notice that adding the first two equations gives a multiple of the third
equation. This means there are too many degrees of freedom. We are
free to choose one of the coordinates, say vx = 1. Then the second and
third equations become vz = s + 2 and vy = 1 + s. The transformation
c = s + 1 gives the vector in the hint. So we have v = h1, c, 1 + ci for
−−→
some constant c. Note that v × P0 P1 = (c − 1)n. So
−−→ p √
kv × P0 P1 k = |(c − 1)|knk = |c − 1| 12 + 12 + (−1)2 = |c − 1| 3
Next, we have
p √
kvk = 12 + c2 + (1 + c)2 2c2 + 2c + 2
By the distance formula, we have
−−→ √ √
kv × P0 P1 k
1= ⇔ 2c2 + 2c + 2 = |c − 1| 3
kvk
Then,
√ √
2c2 + 2c + 2 = |c − 1| 3
2c2 + 2c + 2 = 3(c − 1)2
2c2 + 2c + 2 = 3c2 − 6c + 3
c2 − 8c + 1 = 0
Applying the quadratic formula to this gives
√ √

p
−(−8) ± (−8)2 − 4(1)(1) 8 ± 64 − 4 8 ± 2 15
c= = = = 4± 15
2(1) 2 2
√ √
So v = h1, 4 ± 15, 5 ± 15i. Then the parametric equations of the
lines are
√ √
x = t, y = (4 ± 15)t, z = (5 ± 15)t, −∞ < t < ∞
20. Find the parallel lines intersecting the line x = 2 + t, y = 1 + t,
z = 2 + 2t at a right angle and parallel to the plane x + 2y − 2z = 1
that are at a distance of 1 unit from the plane. Hint: Find values of t
at which the distance from a point in the given line to the plane is 1.
This determines the points of intersection of the lines in question with
the given line.

Solution: First using the hint, we should find the two points on the
line that are a distance of 1 unit from the plane. The distance between
a point on the line and the plane is the length of the perpendicular
122 1. VECTORS AND THE SPACE GEOMETRY

segment from the point to the plane. Note that this segment is perpen-
dicular to the plane, and is thus parallel to its normal, n = h1, 2, −2i.
We find when the line intersects the plane as follows:
(2 + t) + 2(1 + t) − 2(2 + 2t) = 1 ⇔ t = −1
Let P = (1, 0, 0), the point of intersection. Let Q be a point on the
−→
line a distance of 1 unit from the plane. Then P Q = h2 + t, 1 + t, 2 +
2ti − h1, 0, 0i = h1 + t, 1 + t, 2 + 2ti = (1 + t)h1, 1, 2i. The absolute
−→
value of the scalar projection of P Q onto n gives the distance of Q to
the plane. Recall the scalar projection of b onto a is
a·b
b|| =
kak
So we have that the scalar projection s is
(1 + t)h1, 1, 2i · h1, 2, −2i (1 + t)(1(1) + 1(2) + 2(−2) (1 + t)(−1)
s= p = √ =
12 + 22 + (−2)2 9 3
By the above analysis, we require |s| = 1, so
|1 + t|
= 1 ⇔ 1 + t = ±3 ⇔ t = −4, 2
3
We can find the coordinates of Q as follows:
−→ −→ −→ −→ −→ −→
P Q = OQ − OP ⇔ OQ = P Q + OP
So the possible coordinates are:
−−→
OQ1 = h1 − 4, 1 − 4, 2 + 2(−4)i + h1, 0, 0i = h−2, −3, −6i
−−→
OQ2 = h1 + 2, 1 + 2, 2 + 2(2)i + h1, 0, 0i = h4, 3, 6i
Any such line in question is parallel to the plane. Thus they are or-
thogonal to the plane’s normal vector. The problem requires the lines
to also be perpendicular to the given line, so they must be orthogonal
to v = h1, 1, 2i. This means the vector parallel to any line in question
must be orthogonal to both n and v. In other words it is parallel to
n × v. The cross product is
 
ê1 ê2 ê3      
2 −2 1 −2 12
n×v = det  1 2 −2 = hdet , − det , det i = h6, −4, −1i
1 2 1 2 11
1 1 2
Let v1 = h6, −4, −1i. Then the two lines in question are given by
−−→
r = OQ1 + tv1, −∞ < t < ∞
−−→
r = OQ2 + tv1, −∞ < t < ∞
7. LINES IN SPACE 123

21. Find parametric equations of the line that is parallel to v =


h2, −1, 2i and goes through the center of the sphere x2 + y 2 + z 2 =
2x + 6z − 6. Restrict the range of the parameter to describe the part
of the line that is inside the sphere.

Solution: To determine the center of the sphere (and, hence, a par-


ticular point of the line in question), the equation of the sphere should
be transformed to the standard form by completing the squares:
x2 + y 2 + z 2 − 2x − 6z = −6
⇒ (x − 1)2 − 1 + y 2 + (z − 3)2 − 9 = −6
⇒ (x − 1)2 + y 2 + (z − 3)2 = 4
So the sphere has radius 2 and its center is at the point (1, 0, 3). Hence
the parametric equations of the line are
x = 1 + 2t , y = −t , z = 3 + 2t
The points of intersection of the line and the sphere are determined
by the values of t at which the point (1 + 2t, −t, 3 + 2t) satisfies the
equation of the sphere:
2
(2t)2 + (−t)2 + (2t)2 = 4 ⇔ 9t2 = 4 ⇔ t=±
3
Since the center of the sphere corresponds to t = 0, the line segment
inside the sphere is described by restricting the range of t to the inter-
val: − 23 ≤ t ≤ 32 .

22. Let the line L1 pass through the point A(1, 1, 0) parallel to the vec-
tor v = h1, −1, 2i and let the line L2 pass through the point B(2, 0, 2)
parallel to the vector w = h−1, 1, 2i. Show that the lines are intersect-
ing. Find the point C of intersection and parametric equations of the
line L3 through C that is perpendicular to L1 and L2 .

Solution: The parametric equations for lines L1 and L2 are


x = 1 + t, y = 1 − t, z = 2t, −∞ < t < ∞
x = 2 − s, y = s, z = 2 + 2s, −∞ < s < ∞
If the lines are intersecting, there will be a unique pair of real num-
bers (t0 , s0) such that substituting the respective parameter into the
parametric equations yield identical points. Thus we attempt to find a
solution by setting the parametric equation for each coordinate equal
to each other. The following three equations are obtained: 1+t = 2−s,
124 1. VECTORS AND THE SPACE GEOMETRY

1 − t = s, 2t = 2 + 2s ⇔ t = 1 + s. Substituting the third equation


into the first yields
1 + (1 + s) = 2 − s ⇔ 2 + s = 2 − s ⇔ s = 0
Thus t0 = 1 and s0 = 0. Using the parametric equations for L1 , the
point of intersection is
(1 + 1, 1 − 1, 2(1)) = (2, 0, 2)
This is actually point B, so line L2 trivially passes through it.

A line L3 that passes through B and is perpendicular to L1 and L2


must be parallel to a vector that is orthogonal to both v and w. Thus
it is parallel to their cross product
 
ê1 ê2 ê3      
−1 2 1 2 1 −1
v×w = det 1 −1 2 = hdet
  , − det , det i = h−4, −4, 0i
1 2 −1 2 −1 1
−1 1 2
Thus we may choose u = h1, 1, 0i = − 41 (v × w). The parametric
equations of L3 then are
x = 2 + t, y = t, z = 2, −∞ < t < ∞

23. Find parametric equations of the line through (1, 2, 5) that is per-
pendicular to the line x − 1 = 1 − y = z and intersects this line.

Solution: Parametric equations of the given line are x = 1 + t,


y = 1 − t, z = t. This line is parallel to u = h1, −1, 1i. The point
Pt = (1 + t, 1 − t, t) is a generic point of the given line. Any line
through A and Pt intersects the given line, but not at the right angle.
However, there should exists a particular value of t at which the vector
−−→
APt = ht, −1 − t, t − 5i is perpendicular to the given line or u and,
−−→
hence, the line parallel to such APt is the line in question:
−−→ −−→ 4
APt ⊥ u ⇒ APt · u = 0 ⇒ t + (1 + t) + t − 5 = 0 ⇒ t =
3
Thus, parametric equations of the line in question L are:
−−→ 1
L k v = APt 4 = h4, −7, −11i
t= 3 3
⇒ x = 1 + 4t , y = 2 − 7t , z = 5 − 11t .
Note that a non-zero scaling factor in v may always be changed.
Alternative solutions: A vector parallel to the line in question
7. LINES IN SPACE 125

−−→
may also be found in other ways. Let w = AP0 = h0, −1, −5i (the
vector from A to a particular point of the given line). Then
v = u × (u × w)
Indeed, by the geometrical properties of the cross product, u × w is
perpendicular to the plane containing A and the given line. Therefore,
v is perpendicular to both u × w and u and, hence, parallel to the
line in question. One can also make an orthogonal decomposition of
w relative to u: w = wk + w⊥ where wk is the vector projection of w
onto u. Then w⊥ is parallel to the line in question:
w·u
v = w⊥ = w − wk = w − u
kuk2
The two above solutions are related to one another through the ”bac-
cab” rule applied to the double cross cross product in the former solu-
tion.

24. Find the distance between the lines x = y = z and x + 1 =


y/2 = z/3.

Solution: Let L1 be defined by the symmetric equation x = y = z


and L2 be defined by the symmetric equation x + 1 = y/2 = z/3.
Then L1 is parallel to v1 = h1, 1, 1i and L2 is parallel to v2 = h1, 2, 3i.
Clearly, v1 and v2 are not parallel. It is also fairly obvious that the lines
do not intersect, for if they did, then we would have x + 1 = x/2 = x/3.
Solving x/2 = x/3 gives x = 0. But 0 + 1 6= 0, so there is no solution.
Thus the lines are skew. Recall the distance between skew lines is given
by
−→ −→ −→
|AC · (AB × CP )|
D= −→ −→
kAB × CP k
where A, B are distinct points on L1 and C, P are distinct points on
−→ −→
L2 . We may choose A, B and C, P such that AB = v1 and CP = v2 .
Furthermore, we may also choose A = (0, 0, 0) and C = (−1, 0, 0).
−→ −→
Thus AC = OC = h−1, 0, 0i. We must then compute v1 × v2 as
follows:
 
ê1 ê2 ê3      
1 1 1 1 11
v1×v2 = det  1 1 1  = hdet , − det , det i = h1, −2, 1i
23 13 12
1 2 3
Then we have
p √
kv1 × v2 k = 12 + (−2)2 + 12 = 6
126 1. VECTORS AND THE SPACE GEOMETRY

The triple product then is


−→
OC · (v1 × v2 ) = h−1, 0, 0i · h1, −2, 1i = −1
It follows that the distance is
| − 1| 1
D= √ =√
6 6
25. A small meteor moves with the speed v along a straight line par-
allel to a unit vector û. If the meteor passed the point r0 , find the
condition on û so that the meteor hits an asteroid of the shape of a
sphere of radius R centered at the point r1. Determine the position
vector of the impact point.

Solution: Let L be the line r = r0 +tv û, 0 ≤ t < ∞. Notice the inclu-
sion of v, otherwise the units do not make sense (t is time). Let P be the
plane containing L and the center of the sphere. We need only consider
what is happening within this plane. Since the plane passes through
the center of the sphere, the portion of the sphere within the plane is a
circle of radius R. Suppose L is just tangent with the sphere. Then L is
also just tangent with this circle. Let P be the point of tangency. Then
the triangle with vertices P , where the meteor started from, and the
center of the circle is right. Note that the hypotenuse of this triangle is
parallel to r1 − r0 , and one leg is parallel to û. Let θ be the angle be-
tween r1 − r0 and û. Then sin θ = R/kr1 − r0 k ⇔ kr1 − r0 k sin θ = R.
This is the ”maximal” case. If the left hand side exceeds R, then the
line is no longer intersecting.
Suppose now that L passes through the center of the sphere. Clearly
L intersects the sphere. Then û is parallel to r1 − r0. So sin θ = 0.

Combining these two facts, we know it must be that


0 ≤ kr1−r0k sin θ ≤ R ⇔ 0 ≤ kr1−r0 kkûk sin θ ≤ R ⇔ k(r1 −r0)×ûk ≤ R
Where the fact that kûk = 1 and k(r1 − r0 ) × ûk = kr1 − r0 kkûk sin θ
have been used.

To find the position vector of the point of intersection, we must first


find the smallest positive value of t such that L|t is a point on the
sphere. Recall that L is given by
r = r0 + tv û, 0 ≤ t < ∞
and the sphere is given by
kr − r1 k = R
7. LINES IN SPACE 127

Substituting the equation for the line into that of the sphere yields
kr0 + tv û − r1 k = R
ktv û + r0 − r1k2 = R2
t2v 2kûk2 + 2tv û · (r0 − r1) + kr0 − r1k2 = R2
v 2t2 + 2tv û · (r0 − r1) + kr0 − r1k2 − R2 = 0
This is a second degree polynomial in t, so we can solve for t using the
quadratic formula.
p
−(2v û · (r0 − r1 )) ± (2v û · (r0 − r1))2 − 4(v 2)(kr0 − r1 k2 − R2 )
t=
2(v 2)
p
−2v û · (r0 − r1 ) ± 2v (kûkkr0 − r1 k cos θ)2 − (kr0 − r1k2 − R2 )
=
p 2v 2
−û · (r0 − r1 ) ± kr0 − r1 k2(cos θ)2 − kr0 − r1k2 + R2
=
p v
−û · (r0 − r1 ) ± R − kr0 − r1 k2(1 − (cos θ)2 )
2
=
p v
−û · (r0 − r1 ) ± R − kr0 − r1 k2(sin θ)2
2
=
p v
−û · (r0 − r1 ) ± R2 − kû × (r0 − r1 )k2
=
v
Since we require t ≥ 0, it must be that
p
R2 − kû × (r0 − r1 )k2 − û · (r0 − r1)
t=
v
Substitution of this into the equation for the line gives the desired po-
sition vector.

26. A projectile is fired in the direction v = h1, 2, 3i from the point


(1, 1, 1). Let the target be a disk of radius R centered at (2, 3, 6) in the
plane 2x − 3y + 4z = 19. If the trajectory of the projectile is a straight
line, determine whether it hits a target in two cases R = 2 and R = 3.

27. Consider a triangle ABC where A = (1, 1, 1), B = (3, 1, −1),


C = (1, 3, 1). Find the area of a polygon DP QB where the vertices D
and Q are the midpoints of CB and AB, respectively, and the vertex
P is the intersection of the segments CQ and AD.
128 1. VECTORS AND THE SPACE GEOMETRY

8. Euclidean Spaces.
9. QUADRIC SURFACES 129

9. Quadric Surfaces
1–10. Use traces to sketch and identify each of the following surfaces:

1. y 2 = x2 + 9z 2 ;

Solution: First letting z = 0, we have that y 2 = x2 ⇔


y = ±x. These are two orthogonal, straight lines through the
origin.
Next, letting y = 0 we have that x2 + 9z 2 = 0. This is similar
to a circle or ellipse of radius 0, and is simply a point at the
origin.
Finally, letting x = 0 we have that y 2 = 9z 2 ⇔ y = ±3z.
These are also two straight lines through the origin.

This is an elliptic double cone aligned along the y axis.

2. y = x2 − z 2 ;

Solution: First letting z = 0, we have that y = x2. This


is a parabola, pointing up, whose vertex is the origin.
Next, letting y = 0 we have that x2 − z 2 = 0 ⇔ x = ±z.
These are two orthogonal, straight lines through the origin.
Finally, letting x = 0 we have that y = −z 2 . This is a
parabola, pointing down, whose vertex is the origin.

This is a saddle (hyperbolic paraboloid) aligned along the y


axis.

3. 4x2 + 2y 2 + z 2 = 4;

Solution: First letting z = 0, we have that 4x2 + 2y 2 =


4 ⇔ x2 + y 2 /2 = 1. This is an ellipse, centered
√ at the origin,
with a major axis along the y axis of length 2 and a minor
axis along the x axis of length 1.
Next, letting y = 0 we have that 4x2 +z 2 = 4 ⇔ x2 +z 2 /4 = 1.
This is an ellipse, centered at the origin, with a major axis
along the z axis of length 2 and a minor axis along the x axis
of length 1.
Finally, letting x = 0 we have that 2y 2 + z 2 = 4 ⇔ y 2/2 +
z 2 /4 = 1. This is an ellipse, centered at the origin, with a
major axis along the z axis of length 2 and a minor axis along
130 1. VECTORS AND THE SPACE GEOMETRY


the y axis of length 2.

This is an ellipsoid with a = 1, b = 2, and c = 2.

4. x2 − y 2 + z 2 = −1;
2 2
√ First letting z = 0, we have that x − y = −1 ⇔
Solution:
y = ± 1 + x2 . This is a hyperbola opening up and down
along the y axis. For sufficiently large x, the hyperbola begins
to look like y = ±x
Next, letting y = 0 we have that x2 + z 2 = −1. There are no
solutions to this.
Finally,
√ letting x = 0 we have that −y 2 + z 2 = −1 ⇔ y =
± 1 + z 2. This is a hyperbola opening up and down along
the y axis. For sufficiently large z, the hyperbola begins to
look like y = ±z

This is a hyperboloid of two sheets, aligned along the y axis.

5. y 2 + 4z 2 = 16;

Solution: First letting z = 0, we have that y 2 = 16 ⇔


y = ±4. These are lines parallel to the x axis in the xy plane.
Next, letting y = 0 we have that 4z 2 = 16 ⇔ z = ±4. These
are lines parallel to to the x axis in the xz plane.
Finally, letting it does not matter what we let x equal, since
the equation is independent of it. Therefore the surface will
be aligned along the x axis and will be identical for any cross
section x = k. Every cross section will look like y 2 + 4z 2 =
16 ⇔ y 2/16 + z 2 /4 = 1, an ellipse centered at the origin with
major axis along the y axis of length 4 and minor axis along
the z axis of length 2.

This is an elliptic cylinder, aligned along the x axis.

6. x2 − y 2 + z 2 = 1;
2 2
p First letting z = 0, we have that x − y = 1 ⇔
Solution:
x = ± 1 + y 2. This is a hyperbola opening up and down
along the x axis. For sufficiently large y, the hyperbola begins
to look like x = ±y
9. QUADRIC SURFACES 131

Next, letting y = 0 we have that x2 + z 2 = 1. This is a circle


of radius 1 centered at the origin.
Finally,
p letting x = 0 we have that −y 2 + z 2 = 1 ⇔ z =
± 1 + y 2. This is a hyperbola opening up and down along
the z axis. For sufficiently large y, the hyperbola begins to
look like z = ±y

This is a hyperboloid of one sheet, aligned along the y axis.


(One way to try and visualize this is to connect the traces.
There should be a circle and two hyperboloids tangent to the
circle, where you can rotate one by π/2 to get the other. Fill-
ing in the rotations gives the full surface).

7. x2 + 4y 2 − 9z 2 + 1 = 0;

Solution: First letting z = 0, we have that x2 + 4y 2 + 1 =


0 ⇔ x2 + 4y 2 = −1. This has no solutions.
Next,√letting y = 0 we have that x2 − 9z 2 + 1 = 0 ⇔ z =
±1/3 1 + x2. This is a hyperbola opening up and down along
the z axis. For sufficiently large x, the hyperbola begins to look
like z = ±1/3x
Finally, letting
p x = 0 we have that 4y 2 − 9z 2 + 1 = 0 ⇔
z = ±2/3 1 + y 2. This is a hyperbola opening up and down
along the z axis. For sufficiently large y, the hyperbola begins
to look like z = ±2/3y

This is a hyperboloid of two sheets, aligned along the z axis.

8. x2 + z = 0;

Solution: First letting z = 0, we have that x2 = 0 ⇔ x = 0.


This is the y axis.
Next, it does not matter what y we choose, for the surface is
independent of y. Therefore at every cross section y = k for
real k, the surface will be identical. At every such cross sec-
tion, the surface looks like x2 + z = 0 ⇔ z = −x2, a parabola
opening down with vertex at the origin.
Finally, letting x = 0 we have that z = 0. This the y axis once
again.
132 1. VECTORS AND THE SPACE GEOMETRY

This is parabolic cylinder, parallel to the y axis.

9. x2 + 9y 2 + z = 0;

Solution: First letting z = 0, we have that x2 + 9y 2 =


0 ⇔ x = 0. This is simply the origin.
Next, letting y = 0 we have that x2 + z = 0 ⇔ z = −x2, a
parabola opening down with vertex at the origin.
Finally, letting x = 0 we have that 9y 2 + z = 0 ⇔ z = −9y 2 ,
a parabola opening down with vertex at the origin.

This is an elliptic paraboloid opening downwards parallel to


the z axis.

10. y 2 − 4z 2 = 16.

Solution: First letting z = 0, we have that y 2 = 16 ⇔


y = ±4. These are lines parallel to the x axis in the xy plane.
Next, letting y = 0 we have that −4z 2 = 16. There are no
solutions to this.
Finally, letting it does not matter what we let x equal, since
the equation is independent of it. Therefore the surface will
be aligned along the x axis and will be identical for any cross
section x = k for real k. Every
p cross section will look like
2 2
y − 4z = 16 ⇔ y = ±4 1 + z 2 /4, a hyperbola opening
up and down along y. For sufficiently large z, the hyperbola
begins to look like y = ±2x

This is a hyperbolic cylinder, aligned along the x axis.

11–15. Reduce each of the following equations to one of the standard


forms, classify the surface, and sketch it:
11. x2 + y 2 + 4z 2 − 2x + 4y = 0;

Solution: By completing the square, the given equation be-


comes
(x − 1)2 − 1 + (y + 2)2 − 4 + 4z 2 = 0
⇒ (x − 1)2 + (y + 2)2 + 4z 2 = 5
(x − 1)2 (y + 2)2 z2
⇒ √ 2 + √ 2 + √5 2 = 1
5 5 ( 2)
9. QUADRIC SURFACES 133


This√is an ellipsoid centered at (1, −2, 0) with a = b = 5 and
c = 5/2.

12. x2 − y 2 + z 2 + 2x − 2y + 4z + 2 = 0;

Solution: By completing the square, the given equation be-


comes
(x + 1)2 − 1 − (y + 1)2 + 1 + (z + 2)2 − 4 = −2
⇒ (x + 1)2 − (y + 1)2 + (z + 2)2 = 2
This is a hyperboloid of one sheet aligned along the y axis
centered at (−1, −1, −2).

13. x2 + 4y 2 − 6x + z = 0;

Solution: By completing the square, the given equation be-


comes
(x − 3)2 − 9 + 4y 2 + z = 0
⇒ (z − 9) = −((x − 3)2 + 4y 2)
This is an elliptic paraboloid concave down with vertex at
(3, 0, 9).

14. y 2 − 4z 2 + 2y − 16z = 0;

Solution: By completing the square, the given equation be-


comes
(y + 1)2 − 1 − 4(z + 2)2 + 16 = 0
⇒ −(y + 1)2 + 4(z + 2)2 = 15
This is a hyperbolic cylinder parallel to the x axis, whose axis
passes through (0, −1, −2).

15. x2 − y 2 + z 2 − 2x + 2y = 0 .

Solution: By completing the square, the given equation be-


comes
(x − 1)2 − 1 − (y − 1)2 + 1 + z 2 = 0
⇒ (y − 1)2 = (x − 1)2 + z 2
134 1. VECTORS AND THE SPACE GEOMETRY

This is an elliptic double cone whose axis is parallel to the y


axis and with vertex at (1, 1, 0).

16–20. Use rotations in the appropriate coordinate plane to reduce


each of the following equations to one of the standard forms and classify
the surface:
16. 6xy + x2 + y 2 = 1;

Solution: Let x = x0 cos φ−y 0 sin φ and y = y 0 cos φ+x0 sin φ.


Then
x2 = (x0 cos φ − y 0 sin φ)2
= x02 cos φ2 − 2x0 y 0 cos φ sin φ + y 02 sin φ2
1 1
= (1 + cos(2φ))x02 + (1 − cos(2φ))y 02 − sin(2φ)x0 y 0
2 2

y 2 = (y 0 cos φ + y 0 sin φ)2


= y 02 cos φ2 + 2x0 y 0 cos φ sin φ + x02 sin φ2
1 1
= (1 − cos(2φ))x02 + (1 + cos(2φ))y 02 + sin(2φ)x0y 0
2 2

xy = (x0 cos φ − y 0 sin φ)(y 0 cos φ + x0 sin φ)


= x0y 0(cos φ2 − sin φ2 ) + (x02 − y 02) cos φ sin φ
1
= sin(2φ)(x02 − y 02) + cos(2φ)x0 y 0
2
We are interested in finding an angle φ such that the mixed
term x0y 0 is ”annihilated”. Thus we need only consider the
contributions of the mixed terms for now. We have then that
6(cos(2φ)) + (− sin(2φ)) + (sin(2φ)) = 0 ⇔ cos(2φ) = 0
We may then choose φ = π/4. So,
1 2π 1 2π 2π
x2 = (1 + cos( ))x02 + (1 − cos( ))y 02 − sin( )x0 y 0
2 4 2 4 4
1 1
= x02 + y 02 − x0y 0
2 2
1 2π 1 2π 2π
y2 = (1 − cos( ))x02 + (1 + cos( ))y 02 + sin( )x0y 0
2 4 2 4 4
1 02 1 02 0 0
= x + y +xy
2 2
9. QUADRIC SURFACES 135

1 2π 2π
xy = sin( )(x02 − y 02) + cos( )x0y 0
2 4 4
1 02 02
= (x − y )
2
Substitution of these into the given equation yields
1 1 1 1 1
6( (x02 − y 02)) + ( x02 + y 02 − x0y 0) + ( x02 + y 02 + x0y 0 ) = 1
2 2 2 2 2
3x − 3y + x02 + y 02 = 1
02 02

4x02 − 2y 02 = 1
So this is a hyperbolic cylinder.

17. 3y 2 + 3z 2 − 2yz = 1;

Solution: We can play the same game as in Exercise 16,


except in the yz plane. Thus we let y = y 0 cos φ − z 0 sin φ and
z = z 0 cos φ + y 0 sin φ. Then
1 1
y 2 = (1 + cos(2φ))y 02 + (1 − cos(2φ))z 02 − sin(2φ)y 0z 0
2 2
1 1
z 2 = (1 − cos(2φ))y 02 + (1 + cos(2φ))z 02 + sin(2φ)z 0 z 0
2 2
1
yz = sin(2φ)(y 02 − z 02) + cos(2φ)y 0z 0
2
We are interested in finding an angle φ such that the mixed
term is annihilated. So we need only consider the mixed term
contribution of each of the above equations. Thus,
3(− sin(2φ)) + 3(sin(2φ)) − 2(cos(2φ)) = 0 ⇔ cos(2φ) = 0
So we may let φ = π/4. Thus we have
1 2π 1 2π 2π
y2 = (1 + cos( ))y 02 + (1 − cos( ))z 02 − sin( )y 0 z 0
2 4 2 4 4
1 02 1 02
= y + y − y 0z 0
2 2

1 2π 1 2π 2π
z2 = (1 − cos( ))y 02 + (1 + cos( ))z 02 + sin( )y 0z 0
2 4 2 4 4
1 02 1 02 0 0
= y + y +yz
2 2
136 1. VECTORS AND THE SPACE GEOMETRY

1 2π 2π
yz = sin( )(y 02 − z 02) + cos( )y 0z 0
2 4 4
1 02 02
= (y − z )
2
Substitution of these into the given equation yields
1 1 1 1 1
3( y 02 + y 02 − y 0z 0 ) + 3( y 02 + y 02 + y 0z 0 ) − 2( (y 02 − z 02)) = 1
2 2 2 2 2
3 02 3 02 3 02 3 02
y + y + y + y − y z + y z − y 02 + z 02 = 1
0 0 0 0
2 2 2 2
2y 02 + 4z 02 = 1
This is an elliptical cylinder centered at the origin with axis
parallel to the x axis.

18. x − yz = 0;

Solution: As with Exercise 17, we will be rotating in the


yz plane to attempt to annihilate the mixed yz term. Thus
we let y = y 0 cos φ − z 0 sin φ and z = z 0 cos φ + y 0 sin φ. Then
1
yz = sin(2φ)(y 02 − z 02) + cos(2φ)y 0z 0
2
We are only interested in the mixed contributions of this equa-
tion. So, −(cos(2φ)) = 0 and we may let φ = π/4. Then we
have
1 2π 2π
yz = sin( )(y 02 − z 02) + cos( )y 0z 0
2 4 4
1 02
= (y − z 02 )
2
Substitution of this into the given equation yields
1 1
x − y 02 + z 02 = 0 ⇔ 2x = y 02 − z 02
2 2
This is a hyperbolic paraboloid.

19. xy − z 2 = 0;

Solution: As in Exercise 16, we will be rotating in the


xy plane, attempting to annihilate the mixed xy term. Let
9. QUADRIC SURFACES 137

x = x0 cos φ − y 0 sin φ and y = y 0 cos φ + x0 sin φ. Then


1
xy sin(2φ)(x02 − y 02) + cos(2φ)x0y 0
2
Since we are interested in annihilating the mixed term, we
need only consider the mixed contributions from the above
equation. Thus, cos(2φ) = 0, and we may let φ = π/4. So,
1
xy = (x02 − y 02)
2
Substituting this into the given equation yields
1 02
(x − y 02) − z 2 = 0 ⇔ x02 − y 02 − 2z 2 = 0 ⇔ x02 = y 02 + 2z 2
2
This is an elliptic double cone parallel to the x axis. However,
prior to rotation it would have been parallel to the line y = x
in the xy plane.

20. 2xz + 2x2 − y 2 = 0.

Solution: We will need to rotate in the xz plane in order to


annihilate the mixed xz term. Thus, let x = x0 cos φ − z 0 sin φ
and z = z 0 cos φ + x0 sin φ. Then
1 1
x2 = (1 + cos(2φ))x02 + (1 − cos(2φ))z 02 − sin(2φ)x0z 0
2 2
1
xz = sin(2φ)(x02 − z 02) + cos(2φ)x0 z 0
2
We need only use the mixed contribution of these to find a
value of φ to kill the mixed term. Thus we have
2 cos(2φ) + 2(− sin(2φ)) = 0 ⇔ tan(2φ) = 1
and we may let 2φ = π/4 ⇔ φ = π/8. So
1 2π 1 2π 2π
x2 = (1 + cos( ))x02 + (1 − cos( ))z 02 − sin( )x0 z 0
2 √ 8 √ 2 √ 8 8
2 + 2 02 2 − 2 02 2 0 0
= x + z − xz
4 4 2
1 2π 2π
xy = sin( )(x02 − z 02) + cos( )x0z 0
2√ 8 √ 8
2 02 2 0 0
= (x − z 02) + xz
4 2
138 1. VECTORS AND THE SPACE GEOMETRY

Substituting these into the given equation yields


√ √ √ √ √
2 02 02 2 0 0 2 + 2 02 2 − 2 02 2 0 0
2( (x − z ) + x z ) + 2( x + z − x z ) − y2 =0
4 √ 2 4 √ 4 √ 2
2 02 √ 2 + 2 02 2 − 2 02 √ 0 0
(x − z 02) + 2x0 z 0 + x + z − 2x z − y 2 =0
2 2 √ 2 √
2 + 2 2 02 2 − 2 2 02
x + z − y2 =0
2 √ 2 √
(1 + 2)x02 + (1 − 2)z 02 = y2
This is an elliptic double cone.
21. Find an equation for the surface obtained by rotating the line
y = 2x about the y axis. Classify the surface.

Solution: First, this is an elliptic double cone. This can be seen by


looking at cross sections through y = k, which are circles, and through
the xy and yz planes, which are intersecting lines. When z = 0, it
follows that we should have y = ±2x ⇔ y 2 = 4x2 . Furthermore, when
x = 0 we should have y = ±2z ⇔ y 2 = 4z 2 (this can be seen by
rotating y = 2x until it lies in the yz plane). Combining these with
the general formula for an elliptic double cone yields
1
y 2 = 4x2 + 4z 2 ⇔ y 2 = x2 + z 2
4
22. Find an equation for the surface obtained by rotating the curve
y = 1 + z 2 about the y axis. Classify the surface.

Solution: This is an elliptic paraboloid. The trace when x = 0


should be y = 1 + z 2 . The trace when z = 0 should look identical, so
we should have y = 1 + x2. Combining these with the general form of
an elliptic paraboloid yields
y = 1 + x2 + z 2 ⇔ (y − 1) = x2 + z 2

23. Find equations for the family of surfaces obtained by rotating


the curves x2 − 4y 2 = k about the y axis where k is real. Classify the
surfaces.

Solution: We must separate this into three cases: k = 0, k < 0,


and k > 0.
Case 1: k = 0
9. QUADRIC SURFACES 139

The traces in the xy plane look like x2 − 4y 2 = 0, which are inter-


secting lines. This is indicative of an elliptic double cone. In the yz
plane, we obtain the same traces, just replacing x with z. So we have
z 2 − 4y 2 = 0. Combining these two with the general form of an elliptic
double cone yields

x2 + z 2 − 4y 2 = 0 ⇔ 4y 2 = x2 + z 2

Case 2: k < 0
The traces in the xy plane look like x2 − 4y 2 = k. These are hyper-
boloids opening up towards y. So, rotation of these around the y axis
produces the same traces in the yz plane, with x replaced by z. So we
have z 2 − 4y 2 = k.Combining the two equations, we have

x2 + z 2 − 4y 2 = k ⇔ −x2 + 4y 2 − z 2 = −k

In the form on the right, it is clear we have a hyperboloid of two sheets


(note that −k > 0).

Case 3: k > 0
The traces in the xy plane look like x2 − 4y 2 = k. These are hyper-
boloids opening up towards x. So, rotation of these around the y axis
produces the same traces in the yz plane, with x replaced by z. So we
have z 2 − 4y 2 = k.Combining the two equations, we have

x2 + z 2 − 4y 2 = k

In the form on the right, it is clear we have a hyperboloid of one sheet.

24. Find an equation for the surface consisting of all points that are
equidistant from the point (1, 1, 1) and the plane z = 2.

Solution: Some may be familiar that a parabola is generated by


a point (a focus) and a line (a directrix), where the parabola is the set
of points equidistant from the focus and the directrix. The 3D ana-
logue of this is described in the problem. Thus we will have an elliptic
paraboloid.
First, notice that the distance between a point P = (x0, y0, z0 ) and a
plane z = k is simply |z0 − k|. This is because Q = (x0 , y0, k) is a
point on the plane, and P Q is orthogonal to the plane. So the distance
between P and the plane is simply |P Q|, which is trivially |z0 − k|.
140 1. VECTORS AND THE SPACE GEOMETRY

Thus we have
p
(x − 1)2 + (y − 1)2 + (z − 1)2 = |z − 2|
(x − 1)2 + (y − 1)2 + (z − 1)2 = (z − 2)2
(x − 1)2 + (y − 1)2 + (z − 1)2 − (z − 2)2 = 0
(x − 1)2 + (y − 1)2 + (z − 1 − (z − 2))(z − 1 + z − 2) = 0
(x − 1)2 + (y − 1)2 + (2z − 3) = 0
(2z − 3) = −((x − 1)2 + (y − 1)2 )
which is an elliptic paraboloid, as hypothesized, that is concave down
with vertex at (1, 1, 32 )
p
25. Sketch the solid region bounded by the surface z = x2 + y 2
2 2 2
from below and by x + y + z − 2z = 0 from above.
p
Solution: The surface z = x2 + y 2 is the top part of an elliptic
double cone. The surface x2 + y 2 + z 2 − 2z = 0 ⇔ x2 + y 2 + (z − 1)2 = 1
is a sphere of radius
p 1 centered at (0, 0, 1). Incidentally, when z = 1,
the surface z = x2 + y 2 is a circle of radius 1. Thus the double cone
cuts through the sphere at all points along this circle. Since the circle
is parallel to the xy plane and contains a diameter of the sphere, you
can imagine the top portion as just the upper hemisphere. The solid
now looks like an ice-cream cone.

26. Sketch the solid region bounded by the surfaces y = 2 − x2 − z 2 ,


y = x2 + z 2 − 2, and x2 + z 2 = 1.

Solution: The surface y = 2 − x2 − z 2 ⇔ (y − 2) = −(x2 + z 2 ) is an


elliptic paraboloid aligned along the y axis, concave down, with vertex
at (0, 2, 0). The surface y = x2 + z 2 − 2 ⇔ (y + 2) = x2 + z 2 is an
elliptic paraboloid aligned along the y axis, concave up, with vertex at
(0, −2, 0). The surface x2 + z 2 = 1 is a cylinder or radius 1 parallel
to the y axis, whose axis is the y axis. The solid region can first be
viewed as an ”egg” shape inside the two paraboloids. The upper part
of the egg looks like y = 2 − x2 above the y axis, and the lower looks
like y = x2 − 2 below the y axis. Then, shave away everything lying
outside the cylinder x2 + z 2 = 1. The solid looks like that bounded by
y = 2 − x2 , y = x2 − 2, x = ±1 revolved about the y axis.

27. Sketch the solid region bounded by the surfaces x2 + y 2 = R2


and x2 + z 2 = R2
9. QUADRIC SURFACES 141

Solution: The surface x2 + y 2 = R2 is a cylinder of radius R aligned


along the z axis. The surface x2 + z 2 = R2 is a cylinder of radius R
aligned along the y axis. The solid generated is fairly hard to visual-
ize. First, we can set the two equations equal to each other and obtain
y = ±z. This means that the intersection of the two solids lies in the
planes y = ±z. Construct these planes, with only one cylinder in mind.
Focus only on one of the four regions of this cross such that the axis of
the cylinder does not go through it (i.e., there is no hole), then rotate
it by π/2, π, and 3π/2 around the axis of the other cylinder. For more
information see Steinmetz Solid.

28. Find an equation for the surface consisting of all points P for
which the distance from P to the y axis is twice the distance from P
to the zx plane. Identify the surface.

Solution: First note that the xz plane is given by y = 0. We


have established in Exercise 24 that the distance between a point
P = (x0, y0, z0) and a plane y = k is simply |y0 − k|. The projec-
tion of P onto the y axis is given by P 0 = (0, y0, 0). So the distance
between P and the y axis is simply |P P 0 |, since P P 0 is orthogonal to
the y axis. Combining these we have
p
(x − 0)2 + (y − y)2 + (z − 0)2 = 2|y − 0|
√ 2
x2 + z 2 = 22 |y|2
4y 2 = x2 + z 2
This is an elliptic double cone.

29. Show that if the point (a, b, c) lies on the hyperbolic paraboloid z =
y 2 −x2, then the lines through (a, b, c) and parallel to v = h1, 1, 2(b−a)i
and u = h1, −1, −2(a + b)i both lie entirely on this paraboloid. Deduce
from this result that the hyperbolic paraboloid can be generated by the
motion of a straight line. Show that hyperboloids of one sheet, cones,
and cylinders can also be obtained by the motion of a straight line.
Remark. The fact that hyperboloids of one sheet are generated by the
motion of a straight line is used to produce gear transmissions. The
cogs of the gears are the generating lines of the hyperboloids.
30. Find an equation for the cylinder of radius R whose axis goes
through the origin and is parallel to a vector v.
142 1. VECTORS AND THE SPACE GEOMETRY

Solution: Let r be a point on the cylinder. Connect this point to the


axis of the cylinder. A right triangle with the point, the origin, and the
point on the axis is now formed. The leg along the axis of the cylinder
is given by sv for some real s. Then the are of the triangle is given by
1 s
kr × svk = kr × vk
2 2
On the other hand, the area of the triangle is also given by
1 s
Rksvk = Rkvk
2 2
Combining these we have
s s
kr × vk = Rkvk ⇔ skr × vk = sRkvk
2 2
Suppose s 6= 0, then we have
kr × vk = Rkvk
If s = 0, then r is orthogonal to v and krk = R, so the above equation
still holds.

31. Show that the curve of intersection of the surfaces x2 − 2y 2 +


3z 2 − 2x + y − z = 1 and 2x2 − 4y 2 + 6z 2 + x − y + 2z = 4 lies in a
plane.

Solution: Subtracting twice the first equation from the second yields
2x2 −2x2−4y 2 +4y 2+6z 2 −6z 2 +x+4x−y−2y+2z+2z = 4−2 ⇔ 5x−3y+4z = 2
This is clearly a plane.

32. The projection of a point set S onto the xy plane is obtained by


setting the z coordinates of all points of S to zero. The projections of S
onto the other two coordinate axes are defined similarly. What are the
curves that bound the projections of the ellipsoid x2 + y 2 + z 2 − xy = 1
onto the coordinate planes?

Solution: RETURN TO THIS, IT’S WRONG

The surface given is an ellipsoid (imagine just removing the mixed


xy term for some insight). An important distinction is that we cannot
immediately just substitute x = 0, y = 0, or z = 0 to get the necessary
projections. The curves must bind the projection of any trace onto
the respective coordinate plane. Imagine a sphere or radius 1 centered
at (0, 0, 1). The sphere is just tangent to the point (0, 0, 0), which is
9. QUADRIC SURFACES 143

the only point on the surface where z = 0. But the projection of the
sphere onto the xy plane should be a circle of radius 1, the largest
circle produced by any trace z = k for real k. So we must ensure that
the largest curve is already in the appropriate coordinate plane before
setting x = 0, y = 0, or z = 0. For ellipsoids, this will be the case when
the center is at the origin. The ellipsoid given is indeed one centered
at the origin, albeit rotated. Thus we are free to continue simply by
substituting x = 0, y = 0, or z = 0.

Case 1: Projection onto the xy plane


Here we set z = 0. So we have
x2 + y 2 − xy = 1
Let x = x0 cos φ − y 0 sin φ and y = y 0 cos φ + x0 sin φ. Then
1 1
x2 = (1 + cos(2φ))x02 + (1 − cos(2φ))y 02 − sin(2φ)x0y 0
2 2
1 1
y 2 = (1 − cos(2φ))x02 + (1 + cos(2φ))y 02 + sin(2φ)x0 y 0
2 2
1
xy = sin(2φ)(x02 − y 02) + cos(2φ)x0 y 0
2
We are interested in finding an angle φ such that the mixed term x0y 0
is ”annihilated”. Thus we need only consider the contributions of the
mixed terms for now. We have then that
− sin(2φ) + sin(2φ) − cos(2φ) = 0 ⇔ cos(2φ) = 0
We may then choose φ = π/4. So,
1 2π 1 2π 2π
x2 = (1 + cos( ))x02 + (1 − cos( ))y 02 − sin( )x0 y 0
2 4 2 4 4
1 02 1 02
= x + y − x0 y 0
2 2
1 2π 1 2π 2π
y2 = (1 − cos( ))x02 + (1 + cos( ))y 02 + sin( )x0y 0
2 4 2 4 4
1 1
= x02 + y 02 + x0y 0
2 2
1 2π 2π
xy = sin( )(x02 − y 02) + cos( )x0y 0
2 4 4
1 02 02
= (x − y )
2
144 1. VECTORS AND THE SPACE GEOMETRY

Substitution of these into the above equation yields


1 02 1 02 1 1 1
x + y − x0 y 0 + x02 + y 02 + x0y 0 + (x02 − y 02) = 1
2 2 2 2 2
1 02 1 02 1 02 1 02 1 02 1 02
x + x + x + y + y − y − x0 y 0 + x0 y 0 = 1
2 2 2 2 2 2
3 02 1 02
x + y =1
2 2
p √
This is an ellipse with a = 2/3 and b = 2. However, to obtain the
curve pre-rotation, we must rotate this by π/4 clockwise. The ellipse
then is aligned with major axis along y = x.

Case 2: Projection onto the xz plane


Here we set y = 0. So we have
x2 + z 2 = 1
This is a circle of radius 1.

Case 2: Projection onto the yz plane


Here we set x = 0. So we have
y2 + z2 = 1
This is a circle of radius 1.
CHAPTER 2

Vector Functions

1. Curves in Space and Vector Functions


1–5. Find the domain of each of the following vector functions:
1. r(t) = ht, t2 , eti;

Solution: Recall that a vector function r(t) may be writ-


ten as r(t) = hx(t), y(t), z(t)i for functions x(t), y(t), z(t).
The domain of r is the intersection of the domains of each
component. Here, x(t) = t, y(t) = t2, z(t) = et. The domains
are as follows:
x(t) = t : −∞ < t < ∞
y(t) = t2 : −∞ < t < ∞
z(t) = et : −∞ < t < ∞
since polynomials and exponential functions are defined on the
entire real number line. Thus, the domain of r(t) is the inter-
section of the three domains, and is −∞ < t < ∞.

2. r(t) = h t, t2 , et i;

Solution: We √ find the domain in a similar manner as above.


Here, x(t) = t, y(t) = t , z(t) = et. The latter two are the
2

same as in Exercise√1, so we need only find the domain of x(t),


which is wherever t is defined. Clearly, this is when 0 ≤ t.
Thus the domain of r(t) is 0 ≤ t.

3. r(t) = h 9 − t2 , ln t, cos ti;

Solution: We √ find the domain in a similar manner as above.


Here, x(t) = 9 − t2, y(t) = ln t, z(t) = cos t. The domains
are

x(t) = 9 − t2 : 0 ≤ 9 − t2 ⇔ −3 ≤ t ≤ 3
y(t) = ln t : 0 < t
z(t) = cos t : −∞ < t < ∞
145
146 2. VECTOR FUNCTIONS

Thus the domain of r(t) is 0 < t ≤ 3.

4. r(t) = hln(9 − t2 ), ln |t|, (1 + t)/(2 + t)i;

Solution: We find the domain in a similar manner as above.


Here, x(t) = ln(9 − t2 ), y(t) = ln |t|, z(t) = (1 + t)/(2 + t).
The domains are
x(t) = ln(9 − t2) : 0 < 9 − t2 ⇔ −3 < t < 3
y(t) = ln |t| : t 6= 0
z(t) = (1 + t)/(2 + t) : t =6 −2
Thus the domain of r(t) is (−3, −2) ∪ (−2, 0) ∪ (0, 3)
√ √
5. r(t) = h t − 1, ln t, 1 − ti.

Solution: We √ find the domain in a similar


√ manner as above.
Here, x(t) = t − 1, y(t) = ln t, z(t) = 1 − t. The domains
are

x(t) = t − 1 : 0 ≤ t − 1 ⇔ 1 ≤ t
y(t) = ln t : 0 < t

z(t) = 1 − t : 0 ≤ 1 − t ⇔ t ≤ 1
Thus the domain of r(t) is t = 1.

6–15. Find each of the following limits or show that it does not exist:

6. limt→1 h t, 2 − t − t2 , 1/(t2 − 2)i;

Solution: Recall that a vector function r(t) may be written


as r(t) = hx(t), y(t), z(t)i for functions x(t), y(t), z(t). The
limit of r as t approaches some value can be found by find-
ing the limits of the each component. If any one of the limits
does not exist, the limit of r(t) does not either. Note that
the components are simply functions of one variable,√ hence
normal limit laws can be used. Here, x(t) = t, y(t) =
2 − t − t2, z(t) = 1/(t2 − 2). The limits are as follows:

lim x(t) = 1 = 1
t→1
lim y(t) = 2 − 1 − 12 = 0
t→1
1
lim z(t) = = −1
t→1 12 − 2
1. CURVES IN SPACE AND VECTOR FUNCTIONS 147

If a function is continuous at the limit point, then you may


simply substitute. Polynomials are continuous everywhere,
and rational functions are continuous everywhere such that
the denominator is nonzero. Thus,

limh t, 2 − t − t2, 1/(t2 − 2)i = h1, 0, −1i
t→1

7. limt→1 h t, 2 − t − t2 , 1/(t2 − 1)i;

Solution: We find the limit in a similar manner as above.


Note that z(t) = 1/(t2 − 1) is not defined at t = 1, and the
limit tends to infinity. Thus the limit does not exist.

8. limt→0 het , sin t, t/(1 − t)i;

Solution: We find the limit in a similar manner as above.


Here, x(t) = et, y(t) = sin t, z(t) = t/(1 − t). The limits are
lim x(t) = e0 = 1
t→0
lim y(t) = sin 0 = 0
t→0
0
lim z(t) = =0
t→0 1−0
Thus,
limhet , sin t, t/(1 − t)i = h1, 0, 0i
t→0

9. limt→∞ he−t , 1/t2 , 4i;

Solution: We find the limit in a similar manner as above.


Here, x(t) = e−t, y(t) = 1/t2 , z(t) = 4. The limits are
lim x(t) = 0
t→∞
lim y(t) = 0
t→∞
lim z(t) = 4
t→∞

Thus,
lim he−t , 1/t2 , 4i = h0, 0, 4i
t→∞
√ √
10. limt→∞ he−t , (1 − t2 )/t2 , 3 t/( t + t)i;

Solution: We find the limit in a similar manner as above.


148 2. VECTOR FUNCTIONS

√ √
Here, x(t) = e−t, y(t) = (1 − t2 )/t2, z(t) = 3
t/( t + t). The
limits are
lim x(t) = 0
t→∞
lim y(t) = lim (1/t2 − 1) = −1
t→∞ t→∞
1
lim z(t) = lim ( √
6

3 2
)=0
t→∞ t→∞ t+ t
Thus,
√ √
lim he−t , (1 − t2 )/t2,
3
t/( t + t)i = h0, −1, 0i
t→∞

11. limt→−∞ h2, t2 , 1/ ti;
3

Solution: We find the limit in a similar manner as above.


Note that y(t) = t2, and the limit tends to infinity. Thus the
limit does not exist.

12. limt→0+ h(e2t − 1)/t, ( 1 + t − 1)/t, t ln ti;

Solution: We find the limit in √ a similar manner as above.


2t
Here, x(t) = (e − 1)/t, y(t) = ( 1 + t − 1)/t, z(t) = t ln t.
The limits are
P∞ n
P∞ n ∞
n=0 (2t) /(n)! − 1 n=1 (2t) /(n)!
X
lim x(t) = lim = lim = lim 2n tn−1 /(n)! = 2
t→0+ t→0+ t t→0+ t t→0+
n=1

1/2(1 + t)−1/2 1 1
lim y(t) = lim = √ =
t→0+ t→0+ 1 2 1+0 2
ln t 1/t
lim z(t) = lim = lim = −(0) = 0
t→0+ t→0+ 1/t t→0+ −1/t2

In calculating the limit of x(t), you need only find the constant
term in the Taylor expansion. You could also use L’Hopitals
rule. I did this for y(t), z(t). Thus,

lim h(e2t − 1)/t, ( 1 + t − 1)/t, t ln ti = h2, 1/2, 0i
t→0+

13. limt→0 hsin2 (2t)/t2 , t2 + 2, (cos t − 1)/t2 i;

Solution: We find the limit in a similar manner as above.


Here, x(t) = sin2 (2t)/t2, y(t) = t2 + 2, z(t) = (cos t − 1)/t2 .
1. CURVES IN SPACE AND VECTOR FUNCTIONS 149

The limits are


4 cos(2t) sin(2t) sin(4t) 4 cos(4t)
lim x(t) = lim = lim = lim =4
t→0 t→0 2t t→0 t t→0 1
lim y(t) = 02 + 2 = 2
t→0
P∞ n 2n
P∞ n 2n ∞
n=0 (−1) t /(2n)! − 1 n=1 (−1) t /(2n)!
X
lim z(t) = lim = lim = lim (−1)n t2n−2 /(2n)!
t→0 t→0 t2 t→0 t2 t→0
n=1
= (−1)1 /(2)! = −1/2
where L’Hopitals rule has been used in the first two limits, and
a Taylor series in the third. Thus,
limhsin2(2t)/t2 , t2 + 2, (cos t − 1)/t2 i = h4, 2, −1/2i
t→0

14. limt→0 h(e2t − 1)/t, t cos t, 1 + ti;

Solution: We find the limit in a similar manner √ as above.


Here, x(t) = (e2t − 1)/t, y(t) = t cos t, z(t) = 1 + t. The
limits are
lim x(t) = 2
t→0
lim y(t) = 0 cos 0 = 0
t→0

lim z(t) = 1 + 0 = 1
t→0

Thus,

limh(e2t − 1)/t, t cos t, 1 + ti = h2, 0, 1i
t→0

15. limt→∞ he2t / cosh2 t, t2012 e−t , e−2t sinh2 ti.

Solution: We find the limit in a similar manner as above.


Here, x(t) = e2t/ cosh2 t, y(t) = t2012e−t, z(t) = e−2t sinh2 t.
First, cosh(t) = (et + e−t )/2 and sinh(t) = (et − e−t )/2. From
these it can be shown that d/dt cosh(t) = sinh(t) and d/dt sinh(t) =
cosh(t). The limits are
4e2t 4e2t 4
lim x(t) = lim t −t 2
= lim 2t −2t
= lim −2t
=4
t→∞ t→∞ (e + e ) t→∞ e + 2 + e t→∞ 1 + 2e + e−4t
t2012 2012t2011 2012 ∗ 2011t2010 2012!
lim y(t) = lim t = lim t
= lim t
= ... = lim =0
t→∞ t→∞ e t→∞ e t→∞ e t→∞ et
(et − e−t )2 e2t − 2 + e−2t
lim z(t) = lim 2t
= lim 2t
= lim (1/4 − 1/2e−2t + 1/4e−4t ) = 1/4
t→∞ t→∞ 4e t→∞ 4e t→∞
150 2. VECTOR FUNCTIONS

where repeated L’Hopitals rule has been used in finding the


limit of y(t). Thus,
lim he2t/ cosh 2 t, t2012e−t, e−2t sinh2 ti = h4, 0, 1/4i
t→∞

16–22. Sketch each of the following curves and identify the direction
in which the curve is traced out as the parameter t increases:
16. r(t) = ht, cos(3t), sin(3t)i;

Solution: For these problems, the general strategy will be


to split r(t) into a sum of two vectors, r1(t) and r2 (t), one
with zero as one component and the other with zero in the
other two components. For example, here it would be wise to
let r1 (t) = ht, 0, 0i and r2 (t) = h0, cos(3t), sin(3t)i. It is clear
then that projected in the yz plane, r(t) traces a circle. The
action of r1 (t) is to stretch out this circle along the x axis – to
make a helix of radius 1. To determine the direction, simply
choose some sample points. In the yz plane, it should be evi-
dent that the curve is traced counterclockwise. As t increases,
the curve extends in the positive x direction.

17. r(t) = h2 sin(5t), 4, 3 cos(5t)i;

Solution: Let r1 (t) = h2 sin(5t), 0, 3 cos(5t)i. This traces


an ellipse with major axis of length 6 along the z axis and
minor axis of length 4 along the x axis. Let bfr2 (t) = h0, 4, 0i.
This has the action of shifting r1 (t) to the plane y = 4 rather
than y = 0. In the xz plane, it traverses clockwise.

18. r(t) = h2t sin t, 3t cos t, ti;

Solution: Let r1(t) = h2t sin t, 3t cos t, 0i. This curve re-
sembles that of the above, notably it seems to be t times the
previous r1(t). The multiplication of t has the effect of causing
the ellipse to spiral out, as |r1(t)| = t. Letting r2 (t) = h0, 0, ti
reveals that the curve traced is helical in nature. In fact, this
curve lies on the surface of the double cone x2 /4 + y 2 /9 = z 2 .
It has a clockwise orientation viewed from above.

19. r(t) = hsin t, cos t, ln ti;

Solution: Let r1 (t) = hsin t, cos t, 0i. This is clearly a circle


1. CURVES IN SPACE AND VECTOR FUNCTIONS 151

oriented clockwise. Let r2(t) = h0, 0, ln ti. Since ln t is always


increasing and nonnegative, this is a helical shape above the
x axis of radius 1. Notably, it is simply ln t wrapped around
the cylinder x2 + y 2 = 1.

20. r(t) = ht, 1 − t, (t − 1)2 i;

Solution: Let r1 (t) = ht, 1 − t, 0i = h0, 1, 0i + th1, −1, 0.


This is a line through (0, 1, 0) parallel to (1, −1, 0), lying in
the xy plane. Let r2 (t) = h0, 0, (t − 1)2 i. The effect of this is
to raise up the line by (t − 1)2 . The curve is a parabola with
vertex at (1, 0, 0) rotated so that it is in the plane orthogonal
to z = 0 containing r1(t).

21. r(t) = ht2 , t, sin2 (πt)i;

Solution: Let r1 (t) = ht2 , t, 0i. This is a parabola with


vertex at (0, 0) opening towards the positive x axis, lying in
the xy plane. Let r2 (t) = h0, 0, sin2 (πt)i. The effect of this is
to wrap sin2 (πt) so that it sits on top of the parabola.

22. r(t) = hsin t, sin t, 2 cos ti.

Solution: Let r1 (t) = h0, sin t, 2 cos ti. Then r1 (t) √ traces
an ellipse in the yz plane with major axis of length 2 along
the z axis, oriented clockwise. Let r2 (t) = hsin t, 0, 0i. What is
the effect of adding r2 (t) to r1 (t)? Notice that both have the
same period, 2π. Therefore the shape of r1 (t) will be more or
less maintained, but maybe skewed – it will not form a helix.
The effect of r2(t) is to move the points of the ellipse along
the x axis by some amount (according to sin t). Starting with
t = 0 and traversing to t = π, the ellipse is pushed in the pos-
itive x direction. It reaches a maximum at t = π/2, where the
ellipse is pushed by a full one unit. Starting now with t = π
and traversing to t = 2π, the ellipse is pushed in the negative
x direction, eventually reaching the start again. Overall, this
effect just skews the ellipse. You can imagine that the ellipse,
first in the yz plane, is rotated by 45◦ from the y axis toward
the x axis. Alternatively, because the x and y coordinates are
equal, the curve must lie in the plane y = x. In fact, it can
be shown that this is the intersection of x2 + y 2 + z 2 = 2 with
152 2. VECTOR FUNCTIONS

y = x as follows:
Let x(t) = y(t) = sin t, then

sin2 t+sin2 t+z 2 = 2 ⇔ z 2 = 2(1−sin2 t) = 2 cos2 t ⇔ z(t) = ± 2 cos t
where the sign signifies the starting point and the orientation.
23. Two objects are said to collide if they are at the same position at
the same time. Two trajectories are said to intersect if they have com-
mon points. Let t be the physical time. Let two objects travel along
the space curves r1(t) = ht, t2 , t3i and r2 (t) = h1 + 2t, 1 + 6t, 1 + 14ti.
Do the objects collide? Do their trajectories intersect? If so, find the
collision and intersection points.

Solution: We first answer if their trajectories intersect. To do this,


use different parameters t, s. If it happens that t = s, then the objects
collide. Thus we are looking for a (t0, s0) such that r1(t0 ) = r2 (s0 ) and
determining if t0 = s0 or not. So we have
t = 1 + 2s, t2 = 1 + 6s, t3 = 1 + 14s
Squaring both sides of the first gives an equation that we may substitute
into the second, as follows:
(1+2s)2 = 1+6s ⇔ 1+4s+4s2 = 1+6s ⇔ 4s2 −2s = 0 ⇔ s = 0, 1/2
Substituting these into the first equation gives (1, 0) and (2, 1/2) as
possible solutions. Both of these are consistent in the third equation.
Therefore there are two points of intersection but the objects do not
collide. The two points of intersection are:
r1(0) = h1, 1, 1i, r2 (2) = h2, 4, 8i
24. Find a simple parameterization of the curve of intersection of the
surfaces x2 + y 2 /4 + z 2 /9 = 1, y ≥ 0 and z = x2 . Sketch the curve.
Let x(t) = t, a standard choice because it is easy to work with. Then
z(t) = t2. Substitution of these into the first equation yields

t2+y(t)2/4+t4 /9 = 1 ⇔ 36t2 +9y(t)2+4t4 = 36 ⇔ y(t) = ±2/3 9 − 9t2 − t4
This parameterization is simple because x(t) and z(t) are monotoni-
cally increasing.

25–32. Find two vector functions that traverse a given curve C in


the opposite directions if C is the curve of intersection of two surfaces:
25. y = x2 and z = 1;

Solution: Simply let x(t) = t. Then y(t) = x(t)2 = t2 .


1. CURVES IN SPACE AND VECTOR FUNCTIONS 153

So r(t) = ht, t2, 1i describes the intersection. To traverse C in


the opposite direction, simply replace t with −t. This yields
r(t) = h−t, t2, 1i.

26. x = sin y and z = x;

Solution: Let y(t) = t. Then z(t) = x(t) = sin y(t) = sin t.


So the intersection is given by r(t) = hsin t, t, sin ti, and in the
opposite direction is given by r(t) = h− sin t, −t, − sin ti.

27. x2 + y 2 = 9 and z = xy;

Solution: Let x(t) = 3 cos t, since when an x2 and√y 2 appears


it is natural to use sine and cosine. Then y(t) = ±3 1 − cos2 t =
±3 sin t. Choose y(t) = 3 sin t to avoid having a negative sign.
Then z(t) = x(t)y(t) = 9 cos t sin t = 9/2 sin(2t). So the in-
tersection is given by r(t) = h3 cos t, 3 sin t, 9/2 sin(2t)i in one
direction and r(t) = h3 cos t, −3 sin t, −9/2 sin(2t)i in the op-
posite direction.

28. x2 + y 2 = z 2 and x + y + z = 1;

Solution: I am going to employ a little trick first. I will


define x(t) = z(t) cos t and y(t) = z(t) sin t, despite the fact
that we do not know what z(t) is. These satisfy the first equa-
tion naturally. To satisfy the second equation, it must be that
1
z(t)(cos t + sin t) + z(t) = 1 ⇔ z(t) =
1 + cos t + sin t
So r(t) = hcos t/(1 + cos t + sin t), sin t/(1 + cos t + sin t), 1/(1 +
cos t+sin t)i gives the curve of intersection in one direction and
r(t) = hcos t/(1 + cos t − sin t), − sin t/(1 + cos t − sin t), 1/(1 +
cos t − sin t)i in the opposite. I believe there are actually two
curves of intersection, but I am not sure how to get the other.

29. z = x2 + y 2 and y = x2 ;

Solution: Let x(t) = t. Then y(t) = t2 and z(t) = x(t)2 +


y(t)2 = t2 + t4 . So r(t) = ht, t, t2 + t4i gives the curve of inter-
section in one direction and r(t) = h−t, t2, t2 + t4i in the other.
154 2. VECTOR FUNCTIONS

30. x2 /4 + y 2 /9 = 1 and x + y + z = 1;

Solution: Let x(t) = 2 cos t and y(t) = 3 sin t. These satisfy


the first equation. To satisfy the second equation it must be
that z(t) = 1 − 2 cos t − 3 sin t. So r(t) = h2 cos t, 3 sin t, 1 −
2 cos t − 3 sin ti gives the curve of intersection in one direction
and r(t) = h2 cos t, −3 sin t, 1 − 2 cos t + 3 sin ti in the other.

31. x2 /2 + y 2 /2 + z 2 /9 = 1 and x − y = 0;

Solution: By the second equation we have x(t) = y(t).


Substitution of this into the first yields x(t)2 + z(t)2 /9 = 1.
So let x(t) = y(t) = cos t. Then z(t) = 3 sin t. So r(t) =
hcos t, cos t, 3 sin ti gives the curve of intersection in one direc-
tion and r(t) = hcos t, cos t, −3 sin ti in the opposite.

32. x2 + y 2 − 2x = 0 and z = x2 + y 2 .

Solution: Note that substitution of the second equation into


the first gives that z(t) = 2x(t). Completing the square in the
first yields
(x − 1)2 + y 2 = 1
So let x(t) = 1+cos t and y(t) = sin t. Then z(t) = 2(1+cos t).
Thus r(t) = h1 + cos t, sin t, 2(1 + cos t)i gives the curve of
intersection in one direction andr(t) = h1 + cos t, − sin t, 2(1 +
cos t)i in the opposite.
33. Specify the parts of the curve r(t) = hsin t, cos t, 4 sin2 ti that lie
above the plane z = 1 by restricting the range of the parameter t.

Solution: To solve this we need only find the t such that 4 sin2 t > 1.
So
sin2 t > 1/4 ⇔ sin t < −1/2, sin t > 1/2
So we are looking for the t where −1 ≤ sin t < −1/2 and 1/2 < sin t ≤
1. Solving these gives 7π/6 < t < 11π/6 and π/6 < t < 5π/6. Inclu-
sion of the endpoints allows for r(t) to intersect z = 1.

34. Find the values of the parameters a and b at which the curve
r(t) = h1 − at2, b − t, t3i passes through the point (1, 2, 8).

Solution: Equating the third coordinates yields t = 2. Substitut-


ing this gives r(2) = h1 − 4a, b − 2, 8i = h1, 2, 8i. Equating the first
1. CURVES IN SPACE AND VECTOR FUNCTIONS 155

coordinates yields 1 − 4a = 1 ⇔ a = 0 and equating the second


coordinates yields b − 2 = 2 ⇔ b = 4.

35–39. Find the values of a, b, and c, if any, for which each of the
following vector functions is continuous: r(0) = ha, b, ci and, for t 6= 0,
35. r(t) = ht, cos2 t, 1 + t + t2 i;

Solution: Recall that a vector function r(t) is continuous


at t = t0 iff limt→t0 = r(t0). So, to make each of these function
continuous, we must define r(0) = limt→0 r(t), if it exists. In
this problem, the limit exists since polynomials are continuous
and sine and cosine are continuous. So a = limt→0 x(t) = 0,
b = limt→0 y(t) = 1, and c = limt→0 z(t) = 1 and r(0) =
h0, 1, 1i.

36. r(t) = ht, cos2 t, 1 + t2i;

Solution: As with above, we must define r(0) = limt→0 r(t),


if it exists. In this problem, the limit exists since polynomials

are continuous, sine and cosine are continuous, and 1 + t2
is continuous in its domain. So a = limt→0 x(t) = 0, b =
limt→0 y(t) = 1, and c = limt→0 z(t) = 1 and r(0) = h0, 1, 1i.

37. r(t) = ht, cos2 t, ln |t|i;

Solution: As with above, we must define r(0) = limt→0 r(t),


if it exists. In this problem, the limit does not exist since ln |t|
is not continuous at t = 0.

38. r(t) = hsin(2t)/t, sinh(3t)/t, t ln |t|i;

Solution: As with above, we must define r(0) = limt→0 r(t),


if it exists. To determine if the limit exist, we examine the
limit of each component individually
P∞ n 2n+1 2n+1 ∞
n=0 (−1) 2 t /(2n + 1)! X
a = lim x(t) = lim t→0 = lim (−1)n 22n+1 t2n/(2n + 1)! = 2
t→0 t t→0
n=0
e3t − e−3t 3e3t + 3e−3t
b = lim y(t) = lim = lim =3
t→0 t→0 2t t→0 2
ln |t| 1/t
c = lim z(t) = lim = lim = lim(−t) = 0
t→0 t→0 1/t t→0 −1/t2 t→0
156 2. VECTOR FUNCTIONS

Thus r(0) = h2, 3, 0i.

39. r(t) = ht cot(2t), t1/3 ln |t|, t2 + 2i;

Solution: As with above, we must define r(0) = limt→0 r(t),


if it exists. To determine if the limit exist, we examine the
limit of each component individually
t cos(2t) cos(2t) − 2t sin(2t) 1
a = lim x(t) = lim = lim =
t→0 t→0 sin(2t) t→0 2 cos(2t) 2
ln |t| 1/t
b = lim y(t) = lim −1/3 = lim = lim(−3t1/3) = 0
t→0 t→0 t t→0 −1/3t−4/3 t→0

where L’Hopitals rule has been used. Polynomials are contin-


uous so we need not find the limit. Thus r(0) = h1/2, 0, 2i.
40. Suppose that the limits limt→a v(t) and limt→a u(t) exist. Prove
the basic laws of limits for the following vector functions:
(1.1) lim(v(t) + u(t)) = lim v(t) + lim u(t),
t→a t→a t→a
(1.2) lim(sv(t)) = s lim v(t),
t→a t→a
(1.3) lim v(t) · u(t) = lim v(t) · lim u(t),
t→a t→a t→a
(1.4) lim v(t) × u(t) = lim v(t) × lim u(t).
t→a t→a t→a

Solution:
Throughout the proofs, let vi (t), vj (t), and vk (t) be the components
of v(t). Moreover, let limt→a vi (t) = vi(a) and limt→a vj (t) = vj (a).
Similarly, let ui (t), uj (t), and uk (t) be the components of u(t). More-
over, let limt→a ui (t) = ui (a) and limt→a uj (t) = uj (a). Note that the
components of these vector functions are functions of a single variable,
so basic limit laws apply.

Proof of (1.1):
Look at one component of each, say vi (t) and ui (t). Then
lim(vi (t) + ui (t)) = lim vi(t) + lim ui (t) = vi (a) + ui (a)
t→a t→a t→a

Hence the limit exists for each component. By Theorem 10.1 the proof
is complete.

Proof of (1.2):
1. CURVES IN SPACE AND VECTOR FUNCTIONS 157

Look at one component of v(t), say vi (t). Then


lim(svi(t)) = s lim vi (t) = svi(a)
t→a t→a

Hence the limit exists for each component. By Theorem 10.1 the proof
is complete.

Proof of (1.3):
Look at one component of each, say vi (t) and ui (t). Then
lim(vi (t)ui(t)) = (lim vi(t))(lim ui (t)) = vi(a)ui (a)
t→a t→a t→a

Hence the limit exists for each component. By Theorem 10.1 the proof
is complete.

Proof of (1.4):
Look at one component of the cross product, say the ith component.
Then
lim(vj (t)uk (t) − vk (t)ui (t)) = lim(vj (t)uk (t)) − lim(vk (t)uj (t))
t→a t→a t→a
= (lim vj (t))(lim uk (t)) − (lim vk (t))(lim uj (t))
t→a t→a t→a t→a
= vj (a)uk (a) − vk (a)uj (a)
Hence the limit exists for each component. By Theorem 10.1 the proof
is complete.

41. Prove the last limit law in Exercise 40 directly from Definition
10.2, i.e., without using Theorem 10.1. Hint: see Study Problem 10.7.

Solution: The proof is more or less identical to the one given in


Study Problem 10.7, just replace all dots with crosses. The only no-
ticeable is to prove a Cauchy-Schwarz inequality for the cross product,
the proof of which is below:
|a × b| = kakkbk sin θ ≤ kakkbk
owing to the fact that −1 ≤ sin θ ≤ 1.

42–47. Let

v(t) = h(e2t − 1)/t, ( 1 + t − 1)/t, t ln |t|i,
u(t) = hsin2 (2t)/t2 , t2 + 2, (cos t − 1)/t2 i,
w(t) = ht2/3, 2/(1 − t), 1 + t − t2 + t3i.
Use the basic laws of limits established in Exercise 40 to find:
158 2. VECTOR FUNCTIONS

42. limt→0 (2v(t) − u(t) + w(t));

Solution: It would first be wise to find the individual lim-


its limt→0 v(t), limt→0 u(t), and limt→0 w(t). We first find
limt→0 v(t):
e2t − 1
lim x(t) = lim = lim 2e2t = 2
t→0 t→0
√ t t→0

1+t−1 1 1
lim y(t) = lim = lim √ =
t→0 t→0 t t→0 2 1+t 2
ln |t| 1/t
lim z(t) = lim = lim = lim(−t) = 0
t→0 t→0 1/t t→0 −1/t2 t→0

Thus limt→0 v(t) = h2, 1/2, 0i.

We next find limt→0 u(t):


sin2 (2t) 4 sin(2t) cos(2t) 2 sin(4t)
lim x(t) = lim 2
= lim = lim = lim 4 cos(4t) = 4
t→0 t→0 t t→0 2t t→0 2t t→0
2
lim y(t) = lim(t + 2) = 2
t→0 t→0
cos t − 1 − sin t − cos t 1
lim z(t) = lim = lim = lim = −
t→0 t→0 t2 t→0 2t t→0 2 2
Thus limt→0 u(t) = h4, 2, −1/2i.

Finally we find limt→0 w(t):


lim x(t) = t2/3 = 0
t→0
2
lim y(t) = lim =2
t→0 t→0 1 − t

lim z(t) = lim(1 + t − t2 + t3) = 1


t→0 t→0

Thus limt→0 w(t) = h0, 2, 1i.

With these in hand we may now proceed. The limit in question


reduces to
lim(2v(t) − u(t) + w(t)) = 2 lim v(t) − lim u(t) + lim w(t)
t→0 t→0 t→0 t→0
= 2h2, 1/2, 0i − h4, 2, −1/2i + h0, 2, 1i
= h0, 1, 3/2i
43. limt→0 (v(t) · u(t));
1. CURVES IN SPACE AND VECTOR FUNCTIONS 159

Solution: From Exercise 40 the limit can be evaluated as


limt→0 v(t) · limt→0 u(t). Thus,
lim(v(t) · u(t)) = h2, 1/2, 0i · h4, 2, −1/2i = 8 + 1 = 9
t→0

44. limt→0 (v(t) × u(t));

Solution: From Exercise 40 the limit can be evaluated as


limt→0 v(t) × limt→0 u(t). Thus,
lim(v(t) × u(t)) = h2, 1/2, 0i · h4, 2, −1/2i = h−1/4, 1, 2i
t→0

45. limt→0 [w(t) · (v(t) × u(t))];

Solution: From Exercise 40 the limit can be evaluated as


lim[w(t) · (v(t) × u(t))] = lim w(t) · lim(v(t) × u(t)).
t→0 t→0 t→0

Thus,
lim[w(t) · (v(t) × u(t))] = h0, 2, 1i · h−1/4, 1, 2i = 2 + 2 = 4
t→0

46. limt→0 [w(t) × (v(t) × u(t))];

Solution: From Exercise 40 the limit can be evaluated as


lim[w(t) × (v(t) × u(t))] = lim w(t) × lim(v(t) × u(t)).
t→0 t→0 t→0

Thus,
lim[w(t)·(v(t)×u(t))] = h0, 2, 1i×h−1/4, 1, 2i = h4−1, −(−1/4), 0−2(−1/4)i = h3, 1/4, 1/2i
t→0

47. limt→0 [w(t) × (v(t) × u(t)) + v(t) × (u(t) × w(t)) + u(t) ×


(w(t) × v(t))].

Solution: By the Jacobi identity, [w(t) × (v(t) × u(t)) +


v(t) × (u(t) × w(t)) + u(t) × (w(t) × v(t))] = 0. So
lim[w(t) × (v(t) ×u(t)) +v(t) ×(u(t) ×w(t)) +u(t) ×(w(t) ×v(t))] = 0
t→0
.
47. Suppose that the vector functions v(t) × u(t) where v(t) and u(t)
are non-vanishing. Does this imply that the both vector functions v(t)
and u(t) are continuous? Support your arguments by examples.

Solution: Consider v(t) = ht, 1, 0i and u(t) = h−1, ln |t|, 0i, which
are both non-vanishing. Then v(t) × u(t) = h0, 0, t ln |t| + 1i. It has
already been shown that this vector function is continuous. Yet u(t) is
160 2. VECTOR FUNCTIONS

not continuous at t = 0.

48. Suppose that the vector functions v(t) × u(t) and v(t) 6= 0 are
continuous. Does this imply that the vector function u(t) is continu-
ous? Support your arguments by examples.

Solution: See the above example.


2. DIFFERENTIATION OF VECTOR FUNCTIONS 161

2. Differentiation of Vector Functions


1–6. Find the derivatives and differentials of each of the following
vector functions:
1. r(t) = h1, 1 + t, 1 + t3 i;

Solution: To find the derivative r0(t) simply differentiate


each component as follows:

r0(t) = h0, 1, 3t2 i

2. r(t) = hcos t, sin2 t, t2 i;

Solution: As above, to find the derivative r0 (t) simply dif-


ferentiate each component as follows:

r0 (t) = h− sin t, 2 sin t cos t, 2ti = h− sin t, sin(2t), 2ti

3. r(t) = hln t, e2t, te−t i;

Solution: As above, to find the derivative r0 (t) simply dif-


ferentiate each component as follows:
1 1
r0 (t) = h , 2e2t, e−t − te−t i = h , 2e2t, e−t (1 − t)i
t t
√ √
4. r(t) = h 3 t − 2, t2 − 4, ti;

Solution: As above, to find the derivative r0 (t) simply dif-


ferentiate each component as follows:
1 t
r0(t) = h , √ , 1i
3(t − 2)2/3 t2 − 4

5. r(t) = a + bt2 − cet;

Solution: To find the derivative r0 (t), we need only differen-


tiate the coefficients of each vector. The reason being that the
vectors are not functions of t themselves. So,

r0 (t) = 2tb − etc

6. r(t) = ta × (b − cet).
162 2. VECTOR FUNCTIONS

Solution: As above, we need only differentiate the coeffi-


cients. However, we must remember the product rule for dif-
ferentiating cross products:
d
[v(t) × u(t)] = v0(t) × u(t) + v(t) × u0 (t)
dt
Thus r0(t) is
r0(t) = (ta)0 ×(b−cet)+(ta)×(b−cet)0 = a×(b−cet)+(ta)×(−cet) = a×b−et(1+t)a×c
7. Sketch the curve traversed by the vector function r(t) = h2, t1, t2 +1i.
Indicate the direction in which the curve is traversed by r(t) with in-
creasing t. Sketch the position vectors r(0), r(1), r(2) and the vec-
tors r0(0), r0 (1), r0 (2). Repeat the procedure for the vector function
R(t) = r(−t)h2, −t1, t2 + 1i for t = −2, −1, 0.

Solution: Consider r1(t) = h0, t − 1, t2 + 1. This traces a curve


exclusively in the yz plane. We can actually algebraically determine
the shape of the curve. Solving for t in y gives t = y + 1. Substitution
of this into z gives z = (y + 1)2 + 1, a parabola facing up with vertex
at (−1, 1). Let r2 (t) = h2, 0, 0i. The effect of adding this to r1 (t) is to
simply translate the curve into the plane x = 2. The curves traced by
r(t) and r1 (t) are shown below:

It is easy to see that with increasing t, r(t) increases in the positive y


direction (due to the increasing nature of t − 1). The position vectors
r(0), r(1), r(2) are given by
r(0) = h2, −1, 1i
r(1) = h2, 0, 2i
r(2) = h2, 1, 5i
I will sketch these on the plot of r1 (t) for simplicity. r(0), r(1), r(2)
are in red/orange, yellow, and purple, respectively
2. DIFFERENTIATION OF VECTOR FUNCTIONS 163

Next we must find r0 (0), r0(1), r0 (2). It should be clear that r0(t) =
h0, 1, 2ti. Thus we have

r0 (0) = h0, 1, 0i
r0 (1) = h0, 1, 2i
r0 (2) = h0, 1, 4i

Again, these will be sketched on the plot of r1(t). To the left are just
the three tangent vectors, to the right are both tangent and position
vectors.

R(t) simply traces the same curve but in the opposite direction. The
final plot is produced below, with t = 0 corresponding to red/orange,
t = 1 with yellow, and t = 2 with purple as before (note: R(0) and
R(1) are parallel, so their position vectors overlap).
164 2. VECTOR FUNCTIONS

8–12. Determine if the curve traced out by each of the following vector
functions is smooth for a specified interval of the parameter. If the
curve is not smooth at a particular point, graph it near that point.
8. r(t) = ht, t2 , t3 i, 0 ≤ t ≤ 1;

Solution: A curve is smooth on an interval if its unit tan-


gent vector is continuous on that interval. Recall that the unit
tangent vector is given by
r0 (t)
T̂(t) =
kr0 (t)k
so long as r0(t) exists and does not vanish. First we find r0 (t)
as
r0(t) = h1, 2t, 3t2 i
Note that r0 (t) never vanishes since the first component is al-
ways 1. So kr0 (t)k is

kr0(t)k = 1 + 4t2 + 9t4
Then T̂(t) is found to be
1
T̂(t) = √ h1, 2t, 3t2 i
1 + 4t2 + 9t4
which is continuous everywhere, hence r0 (t) is smooth on the
interval.

9. r(t) = ht2 , t3 , 2i, −1 ≤ t ≤ 1;

Solution: As above, we must find T̂. First we find r0 (t)


as
r0(t) = h2t, 3t2 , 0i
2. DIFFERENTIATION OF VECTOR FUNCTIONS 165

Note that r0 (t) vanishes when t = 0, which is in the given


interval. So the unit tangent vector does not exist and r(t)
is not smooth there. However, we must still determine if the
unit tangent vector, when it exists, is continuous. So kr0 (t)k
is √ √
kr0 (t)k = 4t2 + 9t4 = |t| 4 + 9t2
Then T̂(t) is found to be
1
T̂(t) = √ h2t, 3t2 , 0i
|t| 4 + 9t 2

Note that for t > 0 we have


1 1
T̂(t) = √ h2t, 3t2 , 0i = √ h2, 3t, 0i
t 4 + 9t2 4 + 9t2
which is continuous, and for t < 0 we have
1 1
T̂(t) = √ h2t, 3t2 , 0i = − √ h2, 3t, 0i
−t 4 + 9t2 4 + 9t2
which is also continuous. It becomes evident from this that
T̂(t) is discontinuous at t = 0 as the left and right limits as
t → 0 are not equal. Thus r(t) is only not smooth at t = 0. A
sketch is shown below

10. r(t) = ht1/3 , t, t3 i, −1 ≤ t ≤ 1;

Solution: As above, we must find T̂. First we find r0 (t)


as
1
r0(t) = h t−2/3, 1, 3t2 i
3
0
Note that r (t) never vanishes since the third component is
always 1, but it does not exist when t = 0 due to the first
component. So r(t) is possibly not smooth at t = 0. However,
166 2. VECTOR FUNCTIONS

we must still determine if it is not smooth anywhere else in


the interval. So kr0(t)k is
r
0 1 −4/3
kr (t)k = t + 1 + 9t4
9
Then T̂(t) is found to be
1 1 3t2/3 1
T̂(t) = q h t−2/3, 1, 3t2 i = √ h t−2/3, 1, 3t2i
1 −4/3
9
t + 1 + 9t4 3 1 + 9t4/3 + 81t16/3 3
1
= √ h1, 3t2/3, 9t8/3i
1+ 9t4/3 + 81t16/3
It ends up being that T̂(t) is not smooth at t = 0, illustrated
below

11. r(t) = ht5 , t3 , t4 i, −1 ≤ t ≤ 1;

Solution: As above, we must find T̂. First we find r0 (t)


as
r0 (t) = h5t4 , 3t2 , 4t3i
Note that r0(t) exists everywhere but vanishes at t = 0. So
r(t) is possibly not smooth at t = 0. However, we must still
determine if it is not smooth anywhere else in the interval. So
kr0(t)k is
√ √ √
kr0(t)k = 25t8 + 9t4 + 16t6 = |t2| 25t4 + 9 + 16t2 = t2 9 + 16t2 + 25t4
Then T̂(t) is found to be
1 1
T̂(t) = √ h5t4 , 3t2 , 4t3i = √ h5t2 , 3, 4ti
t2 2
9 + 16t + 25t 4 9 + 16t2 + 25t4
2. DIFFERENTIATION OF VECTOR FUNCTIONS 167

It ends up being that limt→0 T̂(t) exists and is h0, 1, 0i. So r(t)
is smooth everywhere in the interval.

12. r(t) = hsin3 t, 1, t2 i, −π/2 ≤ t ≤ π/2.

Solution: As above, we must find T̂. First we find r0 (t)


as
r0(t) = h3 sin2 t cos t, 0, 2ti
Note that r0(t) exists everywhere but vanishes at t = 0. So
r(t) is possibly not smooth at t = 0. However, we must still
determine if it is not smooth anywhere else in the interval. So
kr0(t)k is
p
kr0(t)k = 9 sin4 t cos2 t + 4t2
Then T̂(t) is found to be
1
T̂(t) = √ 4
h3 sin2 t cos t, 0, 2ti
2
9 sin t cos t + 4t 2

Note that
2t
lim T3(t) = lim √ =1
t→0+ t→0+ 9 sin4 t cos2 t + 4t2
whereas
2t
lim T3(t) = lim √ = −1
t→0− t→0− 9 sin4 t cos2 t + 4t2
and thus the limit does not exist, so r(t) is not smooth at
t = 0. The curve is shown below

13. Determine whether a cardioid described by the polar graph r =


1 − a cos θ, 0 ≤ θ ≤ 2π, is smooth, where |a| ≤ 1. Sketch the cardioid
for a = 0, a = ±1/2, and a = ±1. Hint: to find parametric equations
of the cardioid, use the relations between the polar and rectangular
168 2. VECTOR FUNCTIONS

coordinates and θ as a parameter.

Solution: First I will write t instead of θ; this is purely a cosmetic


change. Recall that the conversion from polar to rectangular coordi-
nates is x(t) = r cos t, y(t) = r sin t. So r(t) = h(1 − a cos t) cos t, (1 −
a cos t) sin ti describes the cardioid. From this we must find the unit
tangent vector. We have that

r0(t) = h−(1 − a cos t) sin t + (a sin t) cos t, (1 − a cos t) cos t + (a sin t) sin ti
= h− sin t + 2a sin t cos t, cos t − a cos2 t + a sin2 ti
= h− sin t + a sin(2t), cos t − a cos(2t)i
0
p
kr (t)k = (− sin t + a sin(2t))2 + (cos t − a cos(2t))2
q
= sin2 t − 2a sin t sin(2t) + a2 sin2 (2t) + cos2 t − 2a cos t cos(2t) + a2 cos2 (2t)
q
= (sin2 t + cos2 t) + (a2 sin2(2t) + a2 cos2 (2t)) − 2a(sin t sin(2t) + cos t cos(2t))

= 1 + a2 − 2a cos t

It follows then that r(t) never vanishes so long as (−2 cos t)2 −4 < 0 ⇔
4 cos2 t < 4. This inequality is broken t = 0, π, 2π. However, when
t = π, 1 + a2 − 2a cos t = 1 + a2 + 2a = (1 + a)2 , which is nonnegative so
long as a 6= −1. When t = 0, 2π, 1+a2 −2a cos t = 1+a2 −2a = (1−a)2 ,
which is nonnegative so long as a 6= 1. For any other combination of
a, t, kr0(t)k is nonzero. Then the unit tangent vector can be defined as

1
T̂(t) = √ h(1 − a cos t) cos t, (1 − a cos t) sin ti
1 + a2 − 2a cos t

Again, when a = −1, r(t) is not smooth at t = π and when a = 1,


r)(t) is not smooth at t = 0, 2π. The graphs are depicted below:
2. DIFFERENTIATION OF VECTOR FUNCTIONS 169

From left to right, top to bottom: a = 1, −1, 0, 1/2, −1/2.

14–15. Find the parametric equations of the tangent line to each of


the following curves at a specified point:
14. r(t) = ht2 − t, t3 /3, 2ti, P0 = (6, 9, 6);

Solution: We begin by finding the value of t such that


−−→
r(t) = OP0 . Thus we have t3 /3 = 9 ⇔ t3 = 27 ⇔ t = 3.
Next we find r0(t) by differentiating each component,
r0 (t) = h2t − 1, t2 , 2i
Then r0 (3) = h5, 9, 2i, which is parallel to the tangent line
at P0 . Recall that in general a line parallel to a vector v =
hvi , vj , vk i passing through a P0 = (x0, y0 , z0) is given by
−−→
R(t) = OP0 + tv,
with parametric equations given by x = x0 + tvi, y = y0 + tvj ,
and z = z0 +tvk . Thus the parametric equations of the tangent
line are
x = 6 + 5t, y = 9 + 9t, z = 6 + 2t
170 2. VECTOR FUNCTIONS


15. r(t) = hln t, 2 t, t2 i, P0 = (0, 2, 1).

Solution: As above, we begin by finding the value of t such


−−→ √
that r(t) = OP0 . Thus we have 2 t = 2 ⇔ = 1. Next we
find r0(t).

r0(t) = h1/t, 1/ t, 2ti
Then r0(1) = h1, 1, 2i. So the parametric equations of a line
tangent to P0 are
x = t, y = 2 + t, z = 1 + 2t
16–17. Find the unit tangent vector to the curve traversed by the
specified vector function at the given point P0 :
16. r(t) = h2t + 1, 2 arctan t, e−t i, P0 = (1, 0, 1);
−−→
Solution: First we find the value of t such that r(t) = OP0 .
Then e−t = 1 ⇔ t = 0. Next we find r0 (t) by differentiating
each component as follows:
r0(t) = h2, 2/(1 + t2 ), −e−ti
√ √
Then r0 (0) = h2, 2, −1i. So kr0(0)k = 22 + 22 + 1 = 9 = 3
So the unit tangent vector is
r0(0) 1
T̂(0) = 0
= h2, 2, −1i
kr (0)k 3

17. r(t) = hcos(ωt), cos(3ωt), sin(ωt)i, P0 = (1/2, −1, 3/2),
where ω is a positive constant.

Solution: As above we find the value of t such that r(t) =


−−→
OP0 . Then cos(ωt) = 1/2, so t = π/(3ω) or√t = 5π/(3ω). Of
these, only t = π/(3ω) satisfies sin(ωt) = 3/2. Of course,
adding multiples of 2π/ω to t will also produce valid t, but
these may be neglected; the effect is to essentially add 0. Next
we find r0 (t) by differentiating each component as follows:
r0 (t) = h−ω sin(ωt), −3ω sin(3ωt), ω cos(ωt)i
0

Then
p r (π/(3ω)) = h−ω 3/2, 0, ω/2i. So kr0(π/(3ω))k =
ω 2 (3/4) + 0 + ω 2 /4 = ω So the unit tangent vector is
r0 (π/(3ω)) 1 √ √
T̂(π/(3ω)) = = h−ω 3/2, 0, ω/2i = h− 3/2, 0, 1/2i
kr0 (π/(3ω))k ω
2. DIFFERENTIATION OF VECTOR FUNCTIONS 171

18. Find r0 (t) · r00 (t) and r0 (t) × r00 (t) if r(t) = ht, t2 − 1, t3 + 2i.

Solution: First we compute r0 (t) and r00(t) as follows:


r0(t) = h1, 2t, 3t2 i
r00(t) = h0, 2, 6ti
Then r0 (t) · r00(t) is
r0(t) · r00(t) = h1, 2t, 3t2 i · h0, 2, 6ti = 4t + 18t3
And r0(t) × r00(t) is
 
ê1 ê2 ê3
2t 3t2 1 3t2
     
0 00 1 2t
r (t)×r (t) = det 1 2t 3t
 2
= hdet , − det , det i = h6t2 , −6t, 2i
2 6t 0 6t 0 2
0 2 6t

19. Is there a point on the curve r(t) = ht2 − t, t3 /3, 2ti at which the
tangent line is parallel to the vector v = h−5/2, 2, 1i? If so, find the
point.

Solution: We must first compute r0 (t) as follows


r0 (t) = h2t − 1, t2 , 2i
If there is a point on the curve r)(t) such that the tangent line is parallel
to v, then there must exist an s and t such that v = sr0(t). So we have
h−5/2, 2, 1i = h2ts − s, t2s, 2si
Equating the third components yields 2s = 1 ⇔ s = 1/2. Equating
the second components and substituting s = 1/2 yields 2 = t2/2 ⇔
t = ±2. With t = −2 we have 2(−2)/2 − 1/2 = −5/2 and with t = 2
we have 2(2)/2 − 1/2 = 3/2. Thus there is a point, when t = −2,
r(−2) = h2, −8/3, −4i.

20. Let r(t) = het , 2 cos t, sin(2t)i. Use the best linear approxima-
tion L(t) near t = 0 to estimate r(0.2). Use a calculator to assess the
accuracy kr(0.2) − L(0.2)k of the estimate. Repeat the procedure for
r(0.7) and r(1.2). Compare the errors in all three cases.

Solution: Recall that the best linear approximation, L(t), near a


point t = t0 is defined to be
L(t) = r(t0) + (t − t0 )r0(t0)
So we must first calculate r0 (t), which is
r0(t) = het , −2 sin t, 2 cos(2t)i
172 2. VECTOR FUNCTIONS

Then we have that r(0) = h1, 2, 0i and r0(0) = h1, 0, 2i. So then L(t)
around t = 0 is
L(t) = h1, 2, 0i + th1, 0, 2i = h1 + t, 2, 2ti.
We have that r(0.2) ≈ L(0.2), so
r(0.2) ≈ L(0.2) = h1.2, 2, 0.4i
Using a calculator to compute kr(0.2) − L(0.2)k gives
kr(0.2) − L(0.2)k = 0.0464695041096
For t = 0.7 and t = 1.2 we have
r(0.7) ≈ L(0.7) = h1.7, 2, 1.4i, r(1.2) ≈ L(1.2) = h2.2, 2, 2.4i
And so
kr(0.7) − L(0.7)k = 0.701063103676, kr(1.2) − L(1.2)k = 2.419718929
Clearly the error increases as t moves away from 0. This makes sense
as the approximation is accurate for small neighborhoods of t0.

21. Find the point of intersection of the plane y + z = 3 and the


curve r(t) = hln t, t2, 2ti. Find the angle between the normal of the
plane and the tangent line to the curve at the point of intersection.

Solution: To find the point of intersection simply substitute y(t),


z(t) into the equation of the plane and solve for t:
t2 + 2t = 3 ⇔ t2 + 2t − 3 = 0 ⇔ (t + 3)(t − 1) = 0 ⇔ t = −3, 1
Since ln(−3) is not real, the only point of intersection is when t = 1, at
r(1) = h0, 1, 2i. The normal of the plane is n = h0, 1, 1i. The tangent
line to the curve at the point of intersection is parallel to r0(1). First
we must compute r0 (t), which is
r0(t) = h1/t, 2t, 2i
Then r0 (1) = h1, 2, 2i. So the angle between the tangent line and the
normal of the plane is simply the angle between r0(1) and n, which is
r0 (1) · n 2+2 4
θ = arccos( ) = arccos( √ √ ) = arccos( √ )
kr0(1)kknk 1+4+4 1+1 3 2
22. Does the curve r(t) = h2t2 , 2t, 2−t2i intersect the plane x+y+z =
−3? If not, find a point on the curve that is closest to the plane. What
is the distance between the curve and the plane. Hint: Express the
distance between a point on the curve and the plane as a function of t,
then solve the extreme value problem.
2. DIFFERENTIATION OF VECTOR FUNCTIONS 173

Solution: To test if r(t) intersects the plane, substitute x(t), y(t),


and z(t) into the equation of the plane and attempt to solve for t as
follows:
2t2 + 2t + 2 − t2 = −3 ⇔ t2 + 2t + 3 = 0
The discriminant of this quadratic (in t) is 4 − 4(3)(1) < 0, so there
are no real solutions. Thus there is no point of intersection.
Recall that the distance between a point, whose position vector is r1 ,
and a plane is
|n · r1 − d|
D=
knk
Here we are interested in the closest distance between the curve and
the plane, so we let r1 = r(t), and D becomes a function of t. We may
then minimize D by computing its derivative and finding the critical
points as follows
|n · r(t) − d| 0 |(n)0 · (r(t) − r0)) + (n) · (r(t) − r0))0 | |n · r0 (t)|
D0 (t) = ( ) = =
knk knk knk
D has a critical value when n · r0(t) = 0. We first compute r0 (t) as
r0 (t) = h4t, 2, −2ti
Then to satisfy the condition n · r0 (t) = 0 we must have
4t + 2 − 2t = 0 ⇔ t = −1
So n · r(t) = 2t2 + 2t + 2 − t2 = t2 + 2t + 2 = (t + 1)2 + 1. Thus
n · r(−1) = 1. It follows that the minimum distance is
|1 − (−3)| 4
D(−1) = √ =√
1+1+1 3
An alternative solution can be found below.

Recall that the distance between a point P on a curve and a plane


is the distance of QP , where QP is orthogonal to the plane and Q
is on the plane. The question can now be reformulated as to mini-
−→
mize the distance QP . We want QP to be parallel to n = h1, 1, 1i, so
h2t2 − x0, 2t − y0, 2 − t2 − z0i = sh1, 1, 1i for some s, where (x0 , y0, z0) is
a point on the plane. Then z0 = −3−x0 −y0 . Furthermore, 2t−y0 = s,
so 2t2 −x0 = s = 2t−y0 ⇔ 2t2 −2t−x0 +y0 = 0. Finally 2−t2 +3+x0 +
y0 = 2t2 − x0 ⇔ 5 − 3t2 + 2x0 + y0 = 0. Adding two times the first to
the second gives t2 −4t+5+3y0 = 0 ⇔ y0 = −1/3(t2 −4t+5). Solving
for s gives s = 2t − y0 = 2t + 1/3(t2 − 4t + 5) = t2 /3 + 2/3t + 5/3. Note
174 2. VECTOR FUNCTIONS

−→
then that |QP | = kQP k = ksh1, 1, 1ik, and so we need only minimize
s. We have that
s = 1/3(t2 + 2t + 5)
and so
s0 (t) = 1/3(2t + 2)
and s attains a minimum when t = −1. The minimum value of s then
is s = 1/3(1 − 2 + 5) = 4/3. Finally, the minimum distance is
√ √
|QP | = |s|kh1, 1, 1ik = 4/3 3 = 4/ 3
23. Find the point of intersection of two curves r1 (t) = ht, 1− t, 3+ t2 i
and r2 (s) = h3 − s, s − 2, s2 i. If the angle at which two curves inter-
sect is defined as the angle between their tangent lines at the point of
intersection, find the angle at which the above two curves intersect.

Solution: To find a point of intersection we must find an s, t such


that r1 (t) = r2 (s). So we have
ht, 1 − t, 3 + t2i = h3 − s, s − 2, s2 i
Equating the first components gives t = 3 − s. Equating the third
components and substituting this gives 3 + 9 − 6s + s2 = s2 ⇔ s = 2.
Thus t = 1. It is easily verified that r1 (1) = h1, 0, 4i and r2 (2) =
h1, 0, 4i, which are indeed equal. The angle between the tangent lines
is equal to the angle between the vectors parallel to the tangent lines,
of which we may choose r01 (1) and r02 (2). We have that
r01 (t) = h1, −1, 2ti
r02 (s) = h−1, 1, 2si
So r01(1) = h1, −1, 2i and r02(2) = h−1, 1, 4i. Therefore the angle is
r01(1) · r02(2) −1 − 1 + 8 6 1
θ = arccos( 0 0
) = arccos( √ √ ) = arccos( √ √ ) = arccos( √ )
kr1(1)kkr2 (2)k 1 + 1 + 4 1 + 1 + 16 6 18 3
24. State the condition under which the tangent lines to the curve
r(t) at two distinct points r(t1) and r(t2) are intersecting, or skew, or
parallel. Let r(t) = h2 sin(πt), cos(πt), sin(πt)i, t1 = 0, and t2 = 1/2.
Determine whether the tangent lines at these points are intersecting
and, if so, find the point of intersection.

Solution: Recall the conditions for which two lines are intersect-
ing, skew, or parallel:
Let L1 be a line through P1 and parallel to v1 and L2 a line through
2. DIFFERENTIATION OF VECTOR FUNCTIONS 175

−−→
P2 parallel to v2 . Set r12 = P1 P2 Then
1. L1 and L2 are intersecting iff
r12 · (v1 × v2) = 0, v1 × v2 6= 0
2. L1 and L2 are skew iff
r12 · (v1 × v2) 6= 0
3. L1 and L2 are parallel iff
v1 × v2 = 0, r12 × v1 6= 0
Simply let v1 = r0 (t1), v1 = r0 (t2), and r12 = r(t2) − r(t1) (the order
does not matter) to formulate the condition under which the tangent
lines are intersecting, skew, or parallel.
With r(t) = h2 sin(πt), cos(πt), sin(πt)i, t1 = 0, and t2 = 1/2, first
find r0 (t1) and r0 (t2) as follows:
r0 (t) = h2π cos(πt), −π sin(πt), π cos(πt)i
Then r0(0) = h2π, 0, πi and r0 (1/2) = h0, −π, 0i. The cross product
r0(0) × r0(1/2) can easily be found to be hπ 2 , 0, −2π 2i =
6 0. So the lines
are either intersecting or skew, and we must calculate r(1/2) − r(0) to
determine which. Evidently r(1/2) = h2, 0, 1i and r(0) = h0, 1, 0i, so
r(1/2) − r(0) = h2, −1, 1i. The triple product (r(1/2) − r(0)) · (r0 (0) ×
r0(1/2)) can then be calculated as
(r(1/2)−r(0))·(r0 (0)×r0 (1/2)) = h2, −1, 1i·hπ 2 , 0, −2π 2i = 2π 2 −2π 2 = 0
Therefore the lines are skew.

25. Suppose a smooth curve r(t) does not intersect a plane through a
point P0 and orthogonal to a vector n. Assume that, among the points
on the curve, there is one that is closest to the plane. What is the
angle between n and a tangent vector to the curve at the point that is
the closest to the plane?

Solution: Let r0 be the position vector for P0 . Recall that the dis-
tance between a point, whose position vector is r1, and a plane is
|n · (r1 − r0)|
D=
knk
Here we are interested in the closest distance between the curve and
the plane, so we let r1 = r(t), and D becomes a function of t. We may
176 2. VECTOR FUNCTIONS

then minimize D by computing its derivative and finding the critical


points as follows
|n · (r(t) − r0)| 0 |(n)0 · (r(t) − r0 )) + (n) · (r(t) − r0 ))0| |n · r0(t)|
D0 (t) = ( ) = =
knk knk knk
D has a critical value when n · r0(t) = 0, i.e. when n and r0 (t) are
orthogonal, so the angle is π/2.

26. Suppose r(t) is twice differentiable. Show that (r(t) × r0 (t))0 =


r(t) × r00(t).

Solution: First recall the differentiation rule for cross products:


(v(t) × u(t))0 = v0(t) × u(t) + v(t) × u0 (t)
The proof follows easily:
(r(t)×r0(t))0 = (r(t))0×(r0 (t))+(r(t))×(r0(t))0 = r0 (t)×r0(t)+r(t)×r00(t) = r(t)×r00(t)
owing to the fact that v × v = 0 for any vector v.

27. Suppose that r(t) is differentiable three times. Show that [r(t) ·
(r0(t) × r00(t))]0 = r(t) · (r0(t) × r000(t)).

Solution: Recall the differentiation rule for dot products:


(v(t) · u(t))0 = v0(t) · u(t) + v(t) · u0 (t)
Then we have
[r(t) · (r0(t) × r00(t))]0 = (r(t))0 · (r0(t) × r00(t)) + (r(t)) · (r0(t) × r00(t))0
= r0(t) · (r0(t) × r00(t)) + r(t) · (r0(t) × r00(t))0 = r(t) · (r0(t) × r00(t))0
owing to the fact that v · (v × u) = 0 for vectors v, u. Next recall the
differentiation rule for cross products:
(v(t) × u(t))0 = v0(t) × u(t) + v(t) × u0 (t)
The proof now follows:
r(t) · (r0 (t) × r00(t))0 = r(t) · [(r0(t))0 × (r00(t)) + (r0(t)) × (r00(t))0] = r(t) · [r00(t) × r00(t) + r0 (t) ×
= r(t) · (r0(t) × r000(t))
owing to the fact that v × v = 0 for any vector v.

28. Let r(t) be a differentiable vector function. Show that (kr(t)k)0 =


r(t) · r0 (t)/kr(t)k.
2. DIFFERENTIATION OF VECTOR FUNCTIONS 177

p
Solution: Recall that kr(t)k2 = r(t) · r(t) ⇔ kr(t)k = r(t) · r(t).
Using the differentiation rule for dot products we obtain:

r0 (t)r(t) + r(t)r0(t) r(t) · r0 (t)


(kr(t)k)0 = p =
2 r(t) · r(t) kr(t)k

Clearly this is only defined when kr(t)k =


6 0.

29. A space warship can fire a laser cannon forward along the tan-
gent line to its trajectory. If the trajectory is traversed by the vector
function r(t) = ht, t, t2 + 4i in the direction of increasing t and the tar-
get is the sphere x2 +y 2 +z 2 = 1 find the part of the trajectory in which
the laser cannon can hit the target. Hint: If a line L is tangent to the
trajectory at t = t0, then the target is hit when the distance between
L and the origin is less or equal 1. State this geometrical condition as
an algebraic condition on t0. To solve this algebraic condition, show
that the trajectory is a parabola in the plane y = x. So, find points on
the parabola at which its tangent is at a distance less or equal 1 to the
origin.

Solution: A tangent line to the trajectory at t = t0 is given by

r1(t) = r(t0)+(t−t0)r0 (t0) = ht0, t0 , t20+4i+(t−t0)h1, 1, 2t0 i = ht, t, 2t0t−t20+4i

The target will be hit if the distance between r1 (t) and the origin is less
than or equal to 1 for some t. We can then only focus on the extreme
case when the closest distance is less than or equal to 1.
To find the distance between the origin and a line, we need to find a
−→
point P on the line such that OP is orthogonal to a vector parallel to
the line. The distance |OP | will be the closest distance between the
tangent line and the origin. Clearly a vector parallel to the tangent
line at t = t0 is r0 (t0) = h1, 1, 2t0 . So we require that

r1 (t) · h1, 1, 2t0 i = 0

for some t. Then

t + t + (2t0 t − t20 + 4)(2t0 ) = 0


(4t20 + 2)t − 2t30 + 8t0 = 0
t3 − 4t0
t= 02
2t0 + 1
178 2. VECTOR FUNCTIONS

The closest distance then is


s
t3 − 4t0 2 t3 − 4t0 2 t3 − 4t0
D = ( 02 ) +( 02 ) + (2t0 ( 0 2 ) − t20 + 4)2
2t0 + 1 2t0 + 1 2t0 + 1
1
q
= 2 2t20 (t20 − 4)2 + (2t40 − 8t20 + (−t20 + 4)(2t20 + 1))2
2t0 + 1
1
q
= 2 2t20 (t20 − 4)2 + (2t40 − 8t20 − 2t40 + 8t20 − t20 + 4)2
2t0 + 1
1
q
= 2 2t20 (t20 − 4)2 + (−t20 + 4)2
2t0 + 1
t20 − 4
q
= 2 2t20 + 1
2t0 + 1
t2 − 4
= p0
2t20 + 1
We then require that D ≤ 1, so
t2 − 4
1 ≥ p0 2
2t0 + 1
2t20 + 1 ≥ (t20 − 4)2
2t20 + 1 ≥ t40 − 8t20 + 16
0 ≥ t40 − 10t20 + 15
This is quadratic in t20 . We may then use the quadratic equation to get
the roots

√ √
p
10 ± 102 − 4(1)(15)
q
2 10 ± 100 − 60
t0 = = = 5± 10, t0 = ± 5 ± 10
2 2

Finally we must determine when t40 − 10t20 + 15 is less than 0. To do this


note that the polynomial is even, so it symmetric. Substitutingp t√ =0
gives 15, which
p is√positive, so in the strip between t0 = − 5 − 10
and
p t0 = 5 − 10 it p is positive. This means thatpbetween t0 =
√ √ √
− 5p + 10 and t0 = − 5 − 10, and between t0 = 5 − 10 and

t0 = 5 + 10, it is negative. Thus for these values of t0, the dis-
tance is less than 1. Because the laser shoots forwards in pthe direction

of increasing t, we only consider the t between t 0 = − 5 + 10 and
p √
t0 = − 5 − 10.

It is probably easier to define a new coordinate system in the plane


x = y and work with only two coordinates rather than three.
2. DIFFERENTIATION OF VECTOR FUNCTIONS 179

30–32. A plane normal to a curve at a point P0 is the plane through


P0 whose normal is tangent to the curve at P0 . For each of the follow-
ing curves find a suitable parameterization, the tangent line, and the
normal plane at a specified point:
30. y = x, z = x2 , P0 = (1, 1, 1);

Solution: First we parameterize the intersection by letting


x(t) = t. Then y(t) = t and z(t) = x(t)2 = t2. Next we must
−−→
find the value of t such that r(t) = hx(t), y(t), z(t)i is OP0 .
Clearly this corresponds to t = 1. We then find r0 (t) as
r0(t) = h1, 1, 2ti
and at t = 1 we have r0(1) = h1, 1, 2i. So the tangent line at
P0 is
r1 (t) = h1, 1, 1i + th1, 1, 2i = h1 + t, 1 + t, 2t + 1i
Finally, the normal plane is given by
r0 (1) · hx, y, zi = r0 (1) · h1, 1, 1i ⇔ x + y + 2z = 1 + 1 + 2 = 4
31. x2 + z 2 = 10, y 2 + z 2 = 10, P0 = (1, 1, 3);

Solution: As above, √ we first parameterize


√ the intersection
by
√ letting x(t) = 10 cos t. Then z(t) = 10 sin t and y(t) =
10 cos t (since y = x, as indicated by P0 ). Next we must find
−−→
the value of t such that r(t) =√hx(t), y(t), z(t)i is OP0 . Clearly
this corresponds to arccos(1/ 10) (in the first quadrant since
both sine and cosine must be positive). We then find r0 (t) as
√ √ √
r0(t) = h− 10 sin t, − 10 sin t, 10 cos ti
√ √
and at t = arccos(1/ 10) we have r0 (arccos(1/ 10)) = h−3, −3, 1i.
So the tangent line at P0 is
r1(t) = h1, 1, 3i + th−3, −3, 1i = h1 − 3t, 1 − 3t, 3 + ti
(you could have also used h3, 3, −1i instead; this is what pro-
fessor Shabanov does).
Finally, the normal plane is given by
√ √
r0(arccos(1/ 10))·hx, y, zi = r0(arccos(1/ 10))·h1, 1, 3i ⇔ −3x−3y+z = −3−3+3 = −3
32. x2 + y 2 + z 2 = 6, x + y + z = 0, P0 = (1, −2, 1);

Solution: I will use the following trick to parameterize the


180 2. VECTOR FUNCTIONS

intersection: Note that the first equation describes a sphere


with center at the origin and the second equation describes a
plane through the origin. Therefore the intersection
√ will be a
circle with
√ center at the origin with radius r = 6 (that the
radius is 6 is a consequence of its center being the same as
that of the sphere). So it may be parameterized as
√ √
r(t) = 6 cos tu + 6 sin tv
for some orthogonal unit vectors u and v, both of which will
be orthogonal to n = h1, 1, 1i. √
We are free to choose one of the vectors, so let u = 1/ 2h1, −1, 0i.
Clearly this is orthogonal to n. We √ may √ then√ find v to be
a vector√ parallel to u × n = h− 2, − 2, 2 2i. Then let
v = 1/ 6h−1, −1, 2i. It follows that r(t) is
√ √ √
r(t) = 3 cos th1, −1, 0i−sin th1, 1, −2i = h 3 cos t−sin t, − 3 cos t−sin t, 2 sin ti
It is left to the reader to verify that this is a valid parameteri-
zation by substituting x(t), y(t), and z(t) into the given equa-
−−→
tions. Next we must find the value of t such that r(t) = OP0 .
By equating the third components we have 2 sin t = 1 ⇔
t = π/6, 5π/6 (additions of multiples √ of 2π are ignored). Only
t = π/6 gives a positive value for 3 cos t − sin t. We then find
r0 (t) as
√ √
r0(t) = h− 3 sin t − cos t, 3 sin t − cos t, 2 cos t, i
0
√ √ √
and
√ at√ t = π/6 we
√ have
√ r (π/6) = h− 3/2 − 3/2, 3/2 −
3/2, 3i = h− 3, 0, 3i. So the tangent line at P0 is
√ √ √ √
r1 (t) = h1, −2, 1i + th− 3, 0, 3i = h1 − 3, −2, 1 + 3i
(you could have also used h1, 0, −1i instead; this is what pro-
fessor Shabanov does).
Finally, the normal plane is given by
r0 (π/6) · hx, y, zi = r0 (π/6) · h1, −2, 1i ⇔ x − z = 0
33. Show that tangent lines to a circular helix have a constant angle
with the axis of the helix.

Solution: Without loss of generality we may assume the axis of the


helix is the z axis (this is because any other helix may be obtained
from such a helix via rigid motions). The general form of a helix is
r(t) = hr cos(ωt), r sin(ωt), ti
2. DIFFERENTIATION OF VECTOR FUNCTIONS 181

where r gives the radius of the helix and ω describes how fast the
helix winds. Then r0(t) = h−rω sin(ωt), rω cos(ωt), 1i. Fix a t = t0 .
Then the tangent line to the helix at t = t0 is parallel to r0 (t0) =
h−rω sin(ωt0), rω cos(ωt0), 1i. The axis of the helix is parallel to h0, 0, 1i.
Therefore the angle between the tangent line and the axis of the helix
is
r0(t0) · h0, 0, 1i 1 1
θ = arccos( 0 ) = arccos( p ) = arccos( ),
kr (t0 )kkh0, 0, 1ik 2 2 2 2 2 2
r ω sin (ωt0 ) + r ω cos (ωt0 ) |rω|
a constant.

33. Consider a line through the origin. Any such line sweeps a circular
cone when rotated about the z axis and, for this reason, is called a gen-
erating line of a cone. Prove that the curve r(t) = het cos t, et sin t, eti
intersects all generating lines of the cone x2 + y 2 = z 2 at the same
angle. Hint: Show that parametric equations of a line in the cone
are x = s cos θ, y = s sin θ, and z = s Define the points of intersec-
tion of the line and the curve and find the angle at which they intersect.

Solution: First we verify that a generating line is given by x = s cos θ,


y = s sin θ, and z = s. Note that the variable is s and θ is a constant.
Then x2 + y 2 = s2 cos2 θ + s2 sin2 θ = s2 = z 2 , so they satisfy the
equation of the cone for all s and for all θ. Next consider r(t). As
t → −∞, r(t) approaches h0, 0, 0i. So one point of intersection, where
the curve traced by r(t) emanates from, is the origin. Next we let
r(t) = hs cos θ, s sin θ, si for some θ. Equating the third component
gives et = s. Equating the first and second components and substitut-
ing et = s gives cos t = cos θ and sin t = sin θ. This implies that t = θ
(with only one equality, this is not necessarily true), because both sin t
and cos t must have the same sign as sin θ and cos θ, respectively. So
s = eθ , and the other point of intersection is (eθ cos θ, eθ sin θ, eθ ). A
line tangent to the curve at the point of intersection is parallel to r0 (θ).
First, r0 (t) = h−et sin t + et cos t, et cos t + et sin t, eti. So
r0 (θ) = eθ hcos θ − sin θ, cos θ + sin θ, 1i
and kr0(θ)k is
p √
kr0(θ)k = eθ cos2 θ − 2 sin θ cos θ + sin2 θ + cos2 θ + 2 sin θ cos θ + sin2 θ + 1 = eθ 3
A vector parallel to the generating line is v = hcos θ, sin θ, 1i. Note
that r0 (θ) = eθ (h− sin θ, cos θ, 0i + v). So
r0(θ) · v = eθ (− sin θ cos θ + cos θ sin θ) + eθ kvk = eθ kvk
182 2. VECTOR FUNCTIONS

Finally the angle between r0 (θ) and v is


r0 (θ) · v eθ kvk 1
φ = arccos( 0
) = arccos( √ ) = arccos( √ )
kr (θ)kkvk θ
e 3kvk 3
3. INTEGRATION OF VECTOR FUNCTIONS 183

3. Integration of Vector Functions


1–7. Find the indefinite and definite integrals over specified intervals
for each of the following functions:
1. r(t) = h1, 2t, 3t2 i, 0 ≤ t ≤ 2;

Solution: We first begin by finding the indefinite integral. R


Recall
R thatR if r(t) =R hx(t), y(t), z(t)i is integrable, then r(t)dt =
h x(t)dt, y(t)dt, z(t)dt. If X(t), Y (t), Z(t) R are such that
X 0 (t) = x(t), Y 0 (t) = y(t), Z 0(t) = z(t) then r(t)dt = hX(t), Y (t), Z(t)i+
c for some constant vector c. So we first find the antiderivative
of each component as follows:
Z Z
x(t)dt = 1dt = t + c1
Z Z
y(t)dt = 2tdt = t2 + c2
Z Z
z(t)dt = 3t2dt = t3 + c3

for constants c1 , c2 , c3. Let c = hc1 , c2, c3 i. Then


Z
r(t)dt = ht, t2, t3i + c

Recall that the definite integral on the interval [a, b] is given


by
Z b Z b Z b Z b
r(t)dt = h x(t)dt, y(t)dt, z(t)dti = hX(b)−X(a), Y (b)−Y (a), Z(b)−Z(a)i
a a a a
by the fundamental theorem of calculus. Thus we have
Z 2
r(t)dt = h2 − 0, 22 − 02 , 23 − 03 i = h2, 4, 8i
0
2. r(t) = hsin t, t3 , cos ti, −π ≤ t ≤ π;

Solution: As above, we first find the antiderivative of each


component as follows:
Z Z
x(t)dt = sin tdt = − cos t + c1
Z Z
y(t)dt = t3dt = t4/4 + c2
Z Z
z(t)dt = cos tdt = sin t + c3
184 2. VECTOR FUNCTIONS

for constants c1 , c2 , c3. Let c = hc1 , c2, c3 i. Then


Z
r(t)dt = h− cos t, t4/4, sin ti + c

And the definite integral is


Z π
r(t)dt = h− cos π+cos(−π), π 4/4−(−π)4/4, sin π−sin(−π)i = h0, 0, 0i
−π
√ √
3. r(t) = ht2 , t 1 − t, ti, 0 ≤ t ≤ 1;

Solution: As above, we first find the antiderivative of each


component as follows:
Z Z
x(t)dt = t2 dt = t3 /3 + c1
√ √
Z Z Z
y(t)dt = t 1 − tdt = (u − 1) udu = 2/5u5/2 − 2/3u3/2 + c2

= 2/5(1 − t)5/2 − 2/3(1 − t)3/2 + c2



Z Z
z(t)dt = tdt = 2/3t3/2 + c3

for constants c1 , c2 , c3. Let c = hc1 , c2, c3 i. Then


Z
r(t)dt = ht3 /3, 2/5(1 − t)5/2 − 2/3(1 − t)3/2, 2/3t3/2i + c

And the definite integral is


Z 1
r(t)dt = h 1 3/3 − 03 /3, (2/5(1 − 1)5/2 − 2/3(1 − 1)3/2) − (2/5(1 − 0)5/2 − 2/3(1 − 0)3/2),
0
2 /3(1)3/2 − 2/3(0)3/2 i = h1/3, −2/5 + 2/3, 2/3i = h1/3, 4/15, 2/3i
4. r(t) = ht ln t, t2 , e2ti, 0 ≤ t ≤ 1;

Solution: As above, we first find the antiderivative of each


component as follows:
Z Z Z
x(t)dt = t ln tdt = t /2 ln t − t2 /2(1/t)dt = t2 /2 ln t − t2 /4 + c1
2

Z Z
y(t)dt = t2 dt = t3 /3 + c2
Z Z
z(t)dt = e2tdt = 1/2e2t + c3
3. INTEGRATION OF VECTOR FUNCTIONS 185

for constants c1 , c2 , c3. Let c = hc1 , c2, c3 i. Then


Z
r(t)dt = ht2/2 ln t − t2/4, t3 /3, 1/2e2t i + c

And the definite integral is


Z 1
r(t)dt = h(12 /2 ln 1 − 12 /4) − lim(t2 /2 ln t − t2 /4), 13 /3 − 03 /3, 1/2e2 − 1/2e0 i
0 t→0
2
= h−1/4, 1/3, 1/2(e − 1)i
5. r(t) = h2 sin t cos t, 3 sin t cos2 t, 3 sin2 t cos ti, 0 ≤ t ≤ π/2;

Solution: As above, we first find the antiderivative of each


component as follows:
Z Z
x(t)dt = 2 sin t cos tdt = sin2 t + c1
Z Z
y(t)dt = 3 sin t cos2 tdt = − cos3 t + c2
Z Z
z(t)dt = 3 sin2 t cos tdt = sin3 t + c3

for constants c1 , c2 , c3. Let c = hc1 , c2, c3 i. Then


Z
r(t)dt = hsin2 t, − cos3 t, sin3 ti + c

And the definite integral is


Z π/2
r(t)dt = hsin2 π/2−sin2 0, − cos3 π/2+cos3 0, sin3 π/2−sin3 0i = h1, 1, 1i
0

6. r(t) = a + cos tb, 0 ≤ t ≤ π, a and b are constant vectors;

Solution: Since a and b are constant vectors, their com-


ponents are constants. Thus
R they mayR be treated simply as
constants would, that is f(t)adt = a f(t)dt for any (inte-
grable) function f. Thus,
Z Z Z
r(t)dt = adt + cos tbdt = ta + sin tb + c

and the definite integral is


Z π
r(t)dt = πa + sin πb − ((0)a + sin 0b) = πa
0
186 2. VECTOR FUNCTIONS

7. r(t) = a × (u0 (t) + b), 0 ≤ t ≤ 1 if u0 (t) is continuous and


u(0) = a and u(1) = a − b.

Solution: As above, a and b may be treated as constants,


so the indefinite integral is
Z
r(t)dt = a × (u(t) + tb) + c

and the indefinite integral is


Z 1
r(t)dt = a×(u(1)+(1)b)−a×(u(0)+(0)b) = a×(a−b+b)−a×a = 0
0

8–11. Find r(t) if the derivatives r0 (t) and r(t0 ) are given:
8. r0 (t) = h1, 2t, 3t2 i, r(0) = h1, 2, 3i;

Solution: These types of problems are essentially the same


as those from Calculus I, except three bundled into one. Recall
that if r0 (t) = hx0 (t), y 0(t), z 0(t)i is integrable then the definite
integral is given by r(t) = hx(t), y(t), z(t)i + c. The goal is
to solve for c by substituting the given value of r(t0 ). So we
begin by computing the indefinite integral,
Z Z Z Z
r(t)dt = h 1dt, 2tdt, 3t2dti = ht, t2 , t3i + c

We have that r(0) = h1, 2, 3i, thus


h0, 0, 0i + c = h1, 2, 3i ⇒ c = h1, 2, 3i
Hence
r(t) = ht + 1, t2 + 2, t3 + 3i

9. r0 (t) = ht − 1, t2 , ti, r(1) = h1, 0, 1i;

Solution: As above we begin by computing the indefinite


integral,

Z Z Z Z
2
r(t)dt = h (t − 1)dt, t dt, tdti = ht2/2 − t, t3/3, 2/3t3/2 i + c

We have that r(1) = h1, 0, 1i, thus


h−1/2, 1/3, 2/3i+c = h1, 0, 1i ⇒ c = h1, 0, 1i−h−1/2, 1/3, 2/3i = h3/2, −1/3, 1/3i
Hence
r(t) = ht2 /2 − t + 3/2, t3 /3 − 1/3, 2/3t3/2 + 1/3i
3. INTEGRATION OF VECTOR FUNCTIONS 187

10. r0 (t) = hsin(2t), 2 cos t, sin2 ti, r(π) = h1, 2, 3i;

Solution: As above we begin by computing the indefinite


integral. It is wise to use the identity sin2 t = 1/2(1 − cos(2t)),
Z Z Z Z
r(t)dt = h sin(2t)dt, 2 cos tdt, 1/2(1 − cos(2t))dti
= h− cos(2t)/2, 2 sin t, t/2 − sin(2t)/4i + c
We have that r(π) = h1, 2, 3i, thus
h−1/2, 0, π/2i+c = h1, 2, 3i ⇒ c = h1, 2, 3i−h−1/2, 0, π/2i = h3/2, 2, 3−π/2i
Hence
r(t) = h− cos(2t)/2 + 3/2, 2 sin t + 2, t/2 − sin(2t)/4 + 3 − π/2i
11. r0 (t) = h2t, et , 4t3 i, r(0) = h1, 3, 0i.

Solution: As above we begin by computing the indefinite


integral,
Z Z Z Z
r(t)dt = h 2tdt, e dt, 4t3 dti = ht2 , et, t4i + c
t

We have that r(0) = h1, 3, 0i, thus


h0, 1, 0i + c = h1, 3, 0i ⇒ c = h1, 3, 0i − h0, 1, 0i = h1, 2, 0i
Hence
r(t) = ht2 + 1, et + 2, t4i
12–14. Find the solution r(t) of each of the following initial value
problems:
12. r00 (t) = h0, 2, 6ti, r(0) = h1, 2, 3i, r0 (0) = h1, 0, −1i;

Solution: These types of problems are also essentially the


same as those from Calculus I. Recall that if r00(t) = hx00 (t), y 00(t), z 00(t)i
Ris twice integrable then the definite integral is given by r(t) =
(hx0 (t), y 0(t), z 0(t)i + c1 )dt = hx(t), y(t), z(t)i + tc1 + c2 for
constant vectors c1 and c2 . The goal is to solve for c1 first
by substituting the condition for r0(t) after integrating once.
Then, after substituting this, integrate once more and pro-
ceed as in Exercises 8-11. So we begin by computing the first
indefinite integral,
Z Z Z Z
r (t)dt = h 0dt, 2dt, 6tdti = h0, 2t, 3t2 i + c1
00
188 2. VECTOR FUNCTIONS

We have that r0 (0) = h1, 0, −1i, thus


h0, 0, 0i + c1 = h1, 0, −1i ⇒ c1 = h1, 0, −1i
Hence
r0 (t) = h1, 2t, 3t2 − 1i
Now we may compute r(t) by integrating the above
Z Z Z Z
r (t)dt = h 1dt, 2tdt, (3t2 − 1)dti = ht, t2, t3 − ti + c2
0

We have that r(0) = h1, 2, 3i, thus


h0, 0, 0i + c2 = h1, 2, 3i ⇒ c2 = h1, 2, 3i
Finally,
r(t) = ht + 1, t2 + 2, t3 − t + 3i
13. r00 (t) = ht1/3 , t1/2 , 6ti, r(1) = h1, 0, −1i, r0 (0) = h1, 2, 0i;

Solution: As above we begin by computing the first indefi-


nite integral,
Z Z Z Z
r (t)dt = h t dt, t dt, 6tdti = h3/4t4/3 , 2/3t3/2 , 3t2 i + c1
00 1/3 1/2

We have that r0 (0) = h1, 2, 0i, thus


h0, 0, 0i + c1 = h1, 2, 0i ⇒ c1 = h1, 2, 0i
Hence
r0(t) = h3/4t4/3 + 1, 2/3t3/2 + 2, 3t2 i
Now we may compute r(t) by integrating the above
Z Z Z Z
r (t)dt = h (3/4t +1)dt, (2/3t +2)dt, 3t2 dti = h9/28t7/3 +t, 4/15t5/2+2t, t3i+c2
0 4/3 3/2

We have that r(1) = h1, 0, −1i, thus


h37/28, 34/15, 1i+c2 = h1, 0, −1i ⇒ c2 = h1, 0, −1i−h37/28, 34/15, 1i = h−9/28, −34/15, −2i
Finally,
r(t) = h9/28t7/3 + t − 9/28, 4/5t5/2 + 2t − 34/15, t3 − 2i
14. r00 (t) = h− sin t, cos t, 1/ti, r(π) = h1, −1, 0i, r0 (π) = h−1, 0, 2i.

Solution: As above we begin by computing the first indefi-


nite integral,
Z Z Z Z
00
r (t)dt = h − sin tdt, cos tdt, 1/tdti = hcos t, sin t, ln |t|i + c1
3. INTEGRATION OF VECTOR FUNCTIONS 189

We have that r0 (π) = h−1, 0, 2i, thus


h−1, 0, ln πi+c1 = h−1, 0, 2i ⇒ c1 = h−1, 0, 2i−h−1, 0, ln πi = h0, 0, 2−ln πi
Hence
r0 (t) = hcos t, sin t, ln |t| + 2 − ln πi
Now we may compute r(t) by integrating the above. Note that
Z Z
ln |t|dt = t ln |t| − t(1/t)dt = t ln |t| − t + C

and so,
Z Z Z
0
r (t)dt = h cos tdt, sin tdt, ln |t|+2−ln πi = hsin t, − cos t, t ln |t|+(1−ln π)ti+c2

We have that r(π) = h1, −1, 0i, thus


h0, 1, πi + c2 = h1, −1, 0i ⇒ c2 = h1, −1, 0i − h0, 1, πi = h1, −2, −πi
Finally,
r(t) = hsin t + 1, − cos t − 2, t ln |t| + (1 − ln π)t − πi
15. Solve the equation r00 (t) = a where a is a constant vector if r(0) = b
and r(t0) = c for some t = t0 6= 0.

Solution: Since a is a constant vector, r00(t) is twice integrable, thus


Z
r (t) = r00(t)dt = ta + c1
0

for some constant vector c1, and


Z
r(t) = r0(t)dt = t2 /2a + tc1 + c2

Our goal is to solve for c1 and c2 using the given conditions. Substitu-
tion of the conditions yields
02 /2a + 0c1 + c2 = b, t20 /2a + t0c1 + c2 = c
Solving the first easily gives c2 = b. Substituting this into the second
yields
t20/2a + t0c1 + b = c ⇔ c1 = 1/t0 (c − b) − t0/2a
Thus
r(t) = t2 /2a + t(1/t0(c − b) − t0 /2a) + b
16. Find the most general vector function whose nth derivative van-
ishes, r(n) (t) = 0, in an interval.
190 2. VECTOR FUNCTIONS

Solution: Suppose that r(n) (t) = 0 for some positive integer n. Then
r(n) (t) is integrable and
Z
(n−1)
r (t) = r(n)(t)dt = n!cn

for some constant vector n!cn !. Now, r(n−1) (t) is integrable, and
Z
(n−2)
r (t) = r(n−1) (t)dt = tn!/1!cn + (n − 1)!cn−1

for some constant vector (n − 1)!cn−1 . Then r(n−2) (t) is integrable, and
Z
(n−3)
r (t) = r(n−2) (t)dt = t2n!/2!cn + t(n − 1)!/1!cn−1 + (n − 2)!cn−2

for some constant vector (n − 2)!cn−2 . Continuing in this fashion, we


reach
r(t) = tn−1 cn + tn−2 cn−1 + ... + tc2 + c1
for constant vectors ck , k = 1, 2, ..., n.

17. Show that a continuously differentiable vector function r(t) satis-


fying the equation r00(t) × r0(t) = 0, where r0 (t) is never zero, traverses
a straight line (or a part of it).

Solution: Suppose r(t) = hx(t), y(t), z(t)i is a continuously differ-


entiable vector function such that r00(t) × r0 (t) = 0. Then
r00(t)×r0(t) = hy 00(t)z 0(t)−y 0(t)z 00(t), z 00(t)x0(t)−x00(t)z 0(t), x00(t)y 0(t)−y 00(t)x0(t)i = 0
We may now employ the following trick:
y 00(t)z 0(t) − y 0(t)z 00(t) z 00 (t)x0(t) − x00(t)z 0(t) x00(t)y 0(t) − y 00(t)x0(t)
r00(t) × r0 (t) = h , , i
z 0(t)2 x0(t)2 y 0(t)2
y 0(t) z 0 (t) x0(t)
= h( 0 )0, ( 0 )0 , ( 0 )0 i = 0
z (t) x (t) y (t)
(there is a slight caveat that it is not guaranteed that all three of x(t),
y(t), and z(t) are nonzero. If this is the case, however, the numerator
would have been zero to begin with and no division is necessary. In
fact, it one component is like this, necessarily another one must be as
well. Then you need only work with the remaining one component,
keeping the others as 0).
Now, integrating component by component yields
y 0(t) z 0(t) x0(t)
= c 1 , = c 2 , = c3
z 0 (t) x0 (t) y 0(t)
3. INTEGRATION OF VECTOR FUNCTIONS 191

for nonzero (if one is zero, then the caveat above would have applied)
constants c1, c2 , c3 . Solving for z 0 (t) and y 0(t) in the second two equa-
tions gives
1
z 0 (t) = c2 x0(t), y 0(t) = x0(t)
c3
and so we have
1 1
r0(t) = hx0 (t), x0(t), c2x0 (t)i = x0 (t)h1, , c2i
c3 c3
and hence
1
r(t) = c + x(t)h1, , c2 i
c3
which is a line. Compare this to the general form of a line below
L(t) = r0 + f(t)v
18. If a particle was initially at point (1, 2, 1) and had velocity v =
h0, 1, −1i, find the position vector of the particle after it has been mov-
ing with acceleration a(t) = h1, 0, ti for 2 units of time.

Solution: The given information can be restated in the following


initial conditions r(0) = h1, 2, 1i, v(0) = r0 (0) = h0, 1, −1i, and a(t) =
r00(t) = h1, 0, ti. This is essentially like Exercises 12-14. We begin by
computing r0 (t) as follows
Z Z Z Z
r (t) = r (t)dt = h 1dt, 0dt, tdti = ht, 0, t2 /2i + c1
0 00

We are given that r0 (0) = h0, 1, −1i, so


h0, 0, 0i + c1 = h0, 1, −1i ⇒ c1 = h0, 1, −1i
Hence,
r0 (t) = ht, 1, t2/2 − 1i
Integrating this, we get
Z Z Z
r(t) = h tdt, 1dt, (t2/2 − 1)dti = ht2 /2, t, t3 /6 − ti + c2

We are given that r(0) = h1, 2, 1i, so


h0, 0, 0i + c2 = h1, 2, 1i ⇒ c2 = h1, 2, 1i
Finally
r(t) = ht2 /2 + 1, t + 2, t3 /6 − t + 1i
Then r(2) is
r(2) = h22 /2 + 1, 2 + 2, 23 /6 − 2 + 1i = h3, 4, 1/3i
192 2. VECTOR FUNCTIONS

19. A particle of unit mass moves under a constant force F. If a parti-


cle was initially at the point r0 and passed through the point r1 after
2 units of time, find the initial velocity of the particle. What was the
velocity of the particle when it passed through r1 ?

Solution: Recall Newton’s second law F = ma. We are given that


m = 1, so F = a, a constant vector. Now, compare this problem to
Exercise 15 with b = r0 , c = r1 and t0 = 2. By Exercise 15 we have
that
r(t) = t2/2F + t(1/2(r1 − r0 )) + r0
Then v(t) = r0(t) is
v(t) = tF + 1/2(r1 − r0 )
and so v(0) is
v(0) = 1/2(r1 − r0 )
and v(2) is
v(2) = 2F + 1/2(r1 − r0 )
20. A particle of mass of 1 kg was initially at rest. Then during 2
seconds a constant force of magnitude of 3 N was applied to the parti-
cle in the direction of h1, 2, 2i. How far is the particle from its initial
position in 4 seconds?

Solution: First, the conditions may be restated √ as m = 1,√ v(0) =


0 2
r (0) = 0, and F = h1, 2, 2i. Note that kFk = 1 + 4 + 4 = 9 = 3,
so its magnitude is indeed three N (otherwise we would have to make
the direction vector a unit vector and multiply it by the desired mag-
nitude). Note that I am not including units because they will all work
out nicely; everything is in the appropriate SI unit. Newton’s second
law dictates that F = ma ⇔ a = h1, 2, 2i. So,
Z
v(t) = a(t)dt = ht, 2t, 2ti + c1

Since v(0) = 0 (initially at rest), we have that c1 = 0. Then the


distance traveled (over the first two seconds only!) is given by the
magnitude of
Z 2
d = r(2) − r(0) = v(t)dt = h2, 4, 4i
0

Next, we must find the particle’s position at t = 4. From t = 2 to


t = 4, there is no force and hence a = 0. Thus the particle will move
3. INTEGRATION OF VECTOR FUNCTIONS 193

with a constant velocity given by v(2) = h2, 4, 4i. Then we have that
Z 4
r(4) − r(2) = v(t)dt = h4, 8, 8i
2
So we have that
r(4)−r(2)+r(2)−r(0) = r(4)−r(0) = h4, 8, 8i+h2, 4, 4i = h6, 12, 12i = 6h1, 2, 2i
And thus the distance from the particle’s initial position is
kr(4) − r(0)k = 18
21. The velocity of a particle is v(t) = h2t, 5, 2t − 16i. Find its position
r(t) when the speed of the particle is minimal if r(0) = 0.

Solution: First we have that speed is kv(t)k. Recall from Exercise


28 of section 11 that
v(t) · v0(t)
(kv(t)k)0 =
kv(t)k
Thus speed is minimal when v(t) · v0(t) = 0. It is clear that v0(t) =
h2, 0, 2i, so
v(t) · v0(t) = h2t, 5, 2t − 16i · h2, 0, 2i = 4t + 4t − 32 = 8t − 32
Thus speed is minimal when 8t − 32 = 0 ⇔ t = 4. The particle’s
position at t = 4 is given by
Z 4 Z 4 Z 4 Z 4
r(4) = r(4)−r(0) = v(t) = h 2tdt, h 5dt, (2t−16)dti = h16, 20, −48i
0 0 0 0

owing to the fact that r(0) = 0.

22. A projectile is fired from the ground at an initial speed of 400


m/s and at an angle of elevation of 30◦ . Find the range of the projec-
tile, the maximum height reached, and the speed at impact.

Solution: We are given that v(0) = 400hcos(30◦ ), sin(30◦ )i = 200h 3, 1i.
Since gravity in action, we have that a(t) = h0, −gi, where up is taken
to be the positive direction and g is the acceleration due to gravity (on
Earth, about 9.81 m/s). So
Z
v(t) = a(t)dt = h0, −gti + c1

We are given v(0), so


h0, 0i + c1 = v0
194 2. VECTOR FUNCTIONS

thus √
v(t) = h0, −gti + v0 = h200 3, 200 − gti
The projectile is at its maximum height when the vertical component
of the velocity vector is 0 (imagine the case of simply throwing a ball
up). So the maximum height is when
200
−gt + 200 = 0 ⇔ t =
g
The height at this time is
Z 200
g 200 g 200 2 40000 1 40000 20000
h= 200−gtdt = 200( )− ( ) = − = ≈ 2038.73598369
0 g 2 g g 2 g g
Noting that the projectile’s path is a parabola, the range is given by
twice the horizontal distance from the initial position to where the
maximum height occurs. Thus we have that
Z 200 √
g √ √ 200 80000 3
R=2 200 3dt = 400 3( )= ≈ 14124.7772279
0 g g
Also due to the symmetry of the parabola, we have that impact will
occur at twice the time it takes to reach the maximum height. So the
speed at the impact is

r
400 400 2 p 2
kv( )k = 2002 (3) + (200 − g( )) = 200 (3) + 2002 = 200 4 = 400
g g
23. A ball of mass m is thrown southward into the air at an initial
speed of v0 at an angle of θ to the ground. An east wind applies a
steady force of magnitude F to the ball in a westerly direction. Find
the trajectory of the ball. Where does the ball land and at what speed?
Find the deviation of the impact point from the impact point A when
no wind is present. Is there any way to correct the direction and the
initial speed in which the ball is thrown so that the ball still hits A?
Is it possible to achieve the goal only by adjusting the direction, while
keeping the initial speed fixed?

Solution: Take the positive x direction to mean east, the posi-


tive y direction to be north, and z the vertical height. Without loss
of generality we may assume that the initial point is at the origin.
The initial condition tells us that v(0) = v0h0, cos θ, sin θi (the an-
gle is with respect to the ground!). There is a easterly wind (force)
of magnitude F , represented by hF, 0, 0i. There is also an the force
of gravity, given by h0, 0, −gmi. The net force then is hF, 0, −gmi
3. INTEGRATION OF VECTOR FUNCTIONS 195

Using Newton’s second law, we have that a = hF/m, 0, −gi. Inte-


grating this gives v(t) = hF/mt, 0, −gti + c. The initial velocity is
v(0) = v0h0, cos θ, sin θi. So
v(t) = hF/mt, v0 cos θ, v0 sin θ − gti
Then the trajectory of the ball is
Z
r(t) = v(t)dt = hF/(2m)t2, v0 cos θt, v0 sin θt − g/2t2 i

because we specified that r(0) = 0.


The ball lands when the vertical component is 0. So
v0 sin θt − g/2t2 = 0 ⇔ t(v0 sin θ − g/2t) = 0 ⇔ t = 0, t = 2v0 sin θ/g
We ignore the solution t = 0 because this corresponds to the initial
position. The speed then is
q
kv(2v0 sin θ/g)k = (F/m2v0 sin θ/g)2 + v02 cos2 θ + (v0 sin θ − g(2v0 sin θ/g))2
p
= v0 (2F/m sin θ/g)2 + 1
The ball lands at
2v0 sin θ F 2v0 sin θ 2 2F v02 sin2 θ
x = x( )= ( ) =
g 2m g mg 2
2v0 sin θ 2v0 sin θ 2v 2 cos θ sin θ v 2 sin(2θ)
y = y( ) = v0 cos θ( )= 0 = 0
g g g g
2v0 sin θ
z = z( )=0
g
If there is no wind present, then we may simply set F = 0. Thus the
trajectory and velocity become
r1(t) = h0, v0 cos θt, v0 sin θt − g/2t2 i
v(t) = h0, v0 cos θ, v0 sin θ − gti
The impact still occurs at the same time; thus the y and z coordinates
remain the same. The deviation occurs in x, where the x coordinate is
now 0.
It is possible to correct the direction and the initial speed in which
the ball is thrown so that the ball still hits A (the impact point when
no wind is present). To do this, change the initial speed so that it is
v(0) = hu, v0 cos θ, v0 sin θi. Then the trajectory is
r2(t) = hut + F/(2m)t2 , v0 cos θt, v0 sin θt − g/2t2 i
196 2. VECTOR FUNCTIONS

We wish to have that at t = 2v0 sin θ/g, x(t) = 0. So we require that


u(2v0 sin θ/g) + 2F v02 sin2 θ/(mg 2 ) = 0 ⇔ u = −F v0 sin θ/m
this will, however, change the initial speed.

24. A rocket burns its on-board fuel while moving through space. Let
v(t) and m(t) be the velocity and mass of the rocket at time t. It can
be shown that the force exerted by the rocket jet engines is m0(t)vg ,
where vg is the velocity of the exhaust gases relative to the rocket.
Show that v(t) = v(0) − ln(m(0)/m(t))vg . The rocket is to accelerate
in a straight line from rest to twice the speed of its own exhaust gases.
What fraction of its initial mass would the rocket have to burn as fuel?

Solution: There is no gravity, so Fnet = m0(t)vg , the force exerted


by the rocket jet engines. On the other hand, we have by Newton’s
second law that Fnet = m(t)a(t) = m(t)v0(t). So,
m(t)v0(t) = m0(t)vg
Let vi (t) be a component of v(t) and vgi the corresponding component
of vg . Then
m0(t)
m(t)vi0(t) = m0(t)vgi ⇔ vi0 (t) = vgi
m(t)
Integrating this gives
Z t
vi(t)−vi(0) = vi0 (u)du = ln(m(t))vgi−ln(m(0))vgi = ln(m(t)/m(0))vgi = − ln(m(0)/m(t))vgi
0
where u is a dummy variable. Rearranging this gives
vi (t) = vi (0) − ln(m(0)/m(t))vgi
Since this is true for each component, we have
v(t) = v(0) − ln(m(0)/m(t))vg
If the rocket accelerates from rest to twice the speed of its exhaust
gases, then v(0) = 0 and v(t0) = −2vg (negative because the rocket
moves in the opposite direction as its exhaust gases) for some t0. Then,
−2vg = − ln(m(0)/m(t0))vg ⇔ −2 = − ln(m(0)/m(t0 )) ⇔ m(t0)/m(0) = e−2
25. The acceleration of a projectile is a(t) = h0, 2, 6ti. The projectile
is shot from (0, 0, 0) with an initial velocity v(0) = h1, −2, −10i. It is
supposed to destroy a target located at (2, 0, −12). The target can be
destroyed if the projectiles speed is at least 3.1 at impact. Will the
target be destroyed?
3. INTEGRATION OF VECTOR FUNCTIONS 197

Solution: We must first determine r(t) from the given information.


Integrating a(t) gives
v(t) = h0, 2t, 3t2 i + v(0) = h1, 2t − 2, 3t2 − 10i
Integrating this gives
r(t) = ht, t2 − 2t, t3 − 10ti + r(0) = ht, t2 − 2t, t3 − 10ti
because r(0) = 0. Next we must find when the projectile hits the
target. This is evidently when t = 2 by equating the first component.
Checking the other two coordinates ensures that the projectile does
indeed hit the target. Next we find the speed at this time
p √
kv(2)k = 12 + (2(2) − 2)2 + (3(2)2 − 10)2 = 1 + 22 + 22 = 3
and the target is not destroyed.
198 2. VECTOR FUNCTIONS

4. Arc Length of a Curve


1–6. Find the arc length of each of the following parametric curves:
1. r(t) = h3 cos t, 2t, 3 sin ti, −2 ≤ t ≤ 2;

Solution: First recall that the arc length of a curve C param-


eterized by a simple, continuously differentiable vector func-
tion r(t) on a ≤ t ≤ b is
Z b
L= kr0(t)kdt
a

So we must first calculate r0 (t) as follows


r0 (t) = h−3 sin t, 2, 3 cos ti
and so
0
p √
kr (t)k = 9 sin2 t + 4 + 9 cos2 t = 13
where the fundamental trigonometric identity sin2 t+cos2 t = 1
has been used. Then the arc length is
Z 2√ √ √
L= 13dt = 13(2 − (−2)) = 4 13
−2

2. r(t) = h2t, t3 /3, t2 i, 0 ≤ t ≤ 1;

Solution: As above, we must first calculate r0 (t) as follows


r0 (t) = h2, t2 , 2ti
and so
√ p
kr0(t)k = 4 + t4 + 4t2 = (t2 + 2)2 = t2 + 2
Finally the arc length is
Z 1
13 03 1 7
L= (t2 + 2)dt = + 2(1) − ( − 0) = + 2 =
0 3 3 3 3
3. r(t) = h3t2 , 4t3/2 , 3ti, 0 ≤ t ≤ 2;

Solution: As above, we must first calculate r0 (t) as follows


r0 (t) = h6t, 6t1/2, 3i
and so
√ √ p
kr0(t)k = 36t2 + 36t + 9 = 3 4t2 + 4t + 1 = 3 (2t + 1)2 = 3(2t + 1)
4. ARC LENGTH OF A CURVE 199

Finally the arc length is


Z 2
L= 3(2t + 1)dt = 3(22 + 2 − (02 + 0)) = 18
0

t
4. r(t) = he , 2t, e−ti, −1 ≤ t ≤ 1;

Solution: As above, we must first calculate r0 (t) as follows



r0 (t) = het , 2, −e−ti
and so
√ p
kr0(t)k = e2t + 2 + e−2t = (et + e−t )2 = et + e−t
Finally the arc length is
Z 1
1 1 2
L= (et + e−t )dt = (e − ) − ( − e) = 2e −
−1 e e e
5. r(t) = hcosh t, sinh t, ti, 0 ≤ t ≤ 1;

Solution: As above, we must first calculate r0 (t) as follows


r0 (t) = hsinh t, cosh t, 1i
and so
r
p p e2t + e−2t 1 √
kr0(t)k = sinh2 t + cosh2 t + 1 = cosh(2t) + 1 = + 1 = √ e2t + e−2t + 2
2 2
t −t
e +e
= √
2
where the identity cosh2 t + sinh2 t = cosh(2t) and the result
from Exercise 4 have been used. Finally the arc length is
Z 1 t
e + e−t e − 1/e 1 − 1 e − 1/e
L= √ dt = √ − √ = √
0 2 2 2 2
2
6. r(t) = hcos t − t sin t, sin t + t cos t, t i, 0 ≤ t ≤ 2π; Hint: Find
the decomposition r(t) = v(t) + tw(t) + t2ê3, where v, w, and
ê3 are mutually orthogonal, and v0(t) = w(t), w0 (t) = −v(t).
Use the Pythagorean theorem to calculate kr0(t)k.

Solution: The decomposition in the hint is easily found to


be
r(t) = hcos t, sin t, 0i + th− sin t, cos t, 0i + t2ê3
with v(t) = hcos t, sin t, 0i and w(t) = h− sin t, cos t, 0i. It is
easily verified that v(t) · w(t) = 0, v(t) · ê3 = 0, and w(t) · ê3 =
200 2. VECTOR FUNCTIONS

0 are met (hence they are mutually orthogonal), as well as


v0(t) = w(t), w0 (t) = −v(t). By the product rule we have
r0(t) = v0(t)+w(t)+tw0(t)+2tê3 = w(t)+w(t)+t(−v(t))+2tê3 = 2w(t)−tv(t)+2tê3
where I have also used the conditions v0(t) = w(t), w0 (t) =
−v(t). Because w(t), v(t), and ê3 are mutually orthogonal,
we may use the Pythagorean theorem to calculate kr0 (t)k as
follows
p √
kr0(t)k = 22 + (−t)2 + (2t)2 = 5t2 + 4
since w(t), v(t), and ê3 are unit vectors. Then the arc length
is
Z 2π √ √ √ √
2
L= 5t2 + 4dt = √ ln( 5π 2 + 1 + 5π) + 2π 5π 2 + 1
0 5
The easiest way to compute this integral is √ to first find the
antiderivative using the substitution t = 2/ 5 tan θ, then ap-
plying the fundamental theorem of calculus. The steps are
shown below:
Z √ r
2 4 4
Z Z
2
5t2 + 4dt = 2
√ sec θ 5( ) tan θ + 4dθ = √ sec3 θdθ
5 5 5
The indefinite integral sec3 θdθ may be computed as follows
R
Z Z
sec θdθ = sec θ(tan2 θ + 1)dθ
3

Z Z
= (sec θ tan θ) tan θdθ + sec θdθ
Z Z
3
= sec θ tan θ − sec θdθ + sec θdθ
Z Z
3
2 sec θdθ = sec θ tan θ + sec θdθ

R
The indefinite integral sec θdθ may be computed as follows
sec θ + tan θ
Z Z
sec θdθ = sec θ( )dθ
sec θ + tan θ
du
Z
=
u
= ln |u| + C
= ln | sec θ + tan θ| + C
4. ARC LENGTH OF A CURVE 201

where the substitution u = sec θ + tan θ has been used. Then


we finally have that
Z
2 sec3 θdθ = sec θ tan θ + ln | sec θ + tan θ| + C

and so
4 2 2
Z Z
√ sec3 θdθ = √ (2 sec3 θdθ) = √ (sec θ tan θ+ln | sec θ+tan θ|)+C
5 5 5

Referring back to √ the original substitution
√ t = 2/ 5 tan θ, we
p
2
have√that tan θ = 5/2t and sec θ = tan θ + 1 = 5/4t2 + 1 =
1/2 5t2 + 4. And so the integral is
Z √ √ √
2 1 √ 5 1 √ 5
I(t) = 5t2 + 4dt = √ (ln | 5t2 + 4 + t| + ( 5t2 + 4)( t)) + C
5 2 2 2 2
2 √ √ 1√ 2
= √ ln | 5t2 + 4 + 5t| + 5t + 4t + C 0
5 2

since −2/ 5 ln(2) is just a constant, we may absorb it into C.
By the fundamental theorem of calculus, the definite integral
is just I(2π) − I(0). It is left to the reader to verify that the
answer given above is indeed I(2π) − I(0).
7. Find the arc length of the curve r(t) = he−t cos t, e−t sin t, e−ti,
0 ≤ t ≤ ∞. Hint: Put r(t) = e−t u(t), differentiate, show that u(t)
is orthogonal to u0 (t), and use the Pythagorean theorem to calculate
kr0(t)k.

Solution: First note that, by the hint, r(t) = e−t hcos t, sin t, 1i and
so u(t) = hcos t, sin t, 1i. Then u0 (t) = h− sin t, cos t, 0i and
u(t) · u0 (t) = cos t(− sin t) + sin t cos t + 0 = 0
and so u(t) and u0 (t) are orthogonal. Thus
r0 (t) = (e−t u(t))0 = −e−t u(t) + e−t u0(t)
is a linear combination of two mutually orthogonal vectors. Therefore
we may use the Pythagorean theorem to compute kr0 (t)k as follows
q √ √
0
kr (t)k = (−e−t( 2))2 + (e−t )2 = 3e−t
√ √
where
√ the factor of 2 comes from the fact that ku(t)k = cos2 t + sin2 t + 1 =
2. Then the arc length is
Z ∞√ √ √ √
L= 3e−t dt = lim (− 3e−t) + 3 = 3
0 t→∞
202 2. VECTOR FUNCTIONS

8. Find the arc length of the portion of the helix r(t) = hcos t, sin t, ti
that lies inside the sphere x2 + y 2 + z 2 = 2

Solution: We must first find the points of intersection of the he-


lix with the sphere to determine the interval of t such that r(t) lies
inside the sphere. So,
cos2 t + sin2 t + t2 = 2 ⇔ t2 = 1 ⇔ t = −1, 1
Since r(0) = 0, which lies inside the sphere, for all t with −1 ≤ t ≤ 1,
r(t) lies inside the sphere. We then have that
r0(t) = h− sin t, cos t, 1i
and so p √
kr0 (t)k = sin2 t + cos2 t + 1 = 2
Thus the arc length is
Z 1 √ √ √
L= 2dt = 2(1 − (−1)) = 2 2
−1

9. Find the arc length of the portion of the curve r(t) = h2t, 3t2 , 3t3 i
that lies between the planes z = 3 and z = 24.

Solution: First we must find the points of intersection of the curve


with the two planes to determine all t such that r(t) lies in between
them. We have then that
3t3 = 3, 3t3 = 24 ⇔ t = 1, t = 2
Since 3t3 is increasing it follows that for all t such that 1 ≤ t ≤ 2, r(t)
lies in between the two planes. We then have that
r0(t) = h2, 6t, 9t2 i
and so
p √ p
kr0(t)k = 22 + (6t)2 + (9t2 )2 = 81t4 + 36t2 + 4 = (9t2 + 2)2 = 9t2 +2
Then the arc length is
Z 2
L= (9t2 + 2)dt = (3(2)3 + 2(2)) − (3(1)3 + 2(1)) = 28 − 5 = 23
1

10. Find the arc length of the portion of the curve r(t) = hln t, t2 , 2ti
that lies between the points of intersection of the curve with the plane
y − 2z + 3 = 0.
4. ARC LENGTH OF A CURVE 203

Solution: We must first find the points of intersection with the given
plane. So,
t2 − 2(2t) + 3 = 0 ⇔ t2 − 4t + 3 = 0 ⇔ (t − 1)(t − 3) = 0 ⇔ t = 1, 3
We then have that
1
r0(t) = h , 2t, 2i
t
and so
r r r
1 1 1 1
kr0 (t)k = ( )2 + (2t)2 + 22 = 4t2 + 4 + 2 = (2t + )2 = 2t +
t t t t
Thus the arc length is
Z 3
1
L= (2t + )dt = (32 + ln 3) − (12 + ln 1) = 8 + ln 3
1 t
11. Let C be the curve of intersection of the surfaces z 2 = 2y and
3x = yz Find the length of Cfrom the origin to the point (36, 18, 6).

Solution: First we must find a simple, continuously differentiable


parameterization of C. To do this, begin by letting z(t) = t. Then
y(t) = 1/2z(t)2 = 1/2t2 . Finally, x(t) = 1/3y(t)z(t) = 1/6t3 . We then
have that r(t) = h1/6t3 , 1/2t2 , ti is a parameterization of C. Since t is
injective (one-to-one), r(t) is automatically a simple parameterization,
and it is obviously differentiable since its components are polynomials.
We must now find the value of t such that r(t) = h36, 18, 6i. Evidently
this is when t = 6. Clearly, r(0) = 0, so our interval is 0 ≤ t ≤ 6. Then
we have that
r0(t) = h1/2t2 , t, 1i
and so
r
t4 1√ 4 1p 2 1
kr0(t)k = + t2 + 1 = t + 4t2 + 4 = (t + 2)2 = (t2 + 2)
4 2 2 2
Then the arc length is
Z 6
1 2 1 63 03 1
L= (t + 2)dt = ( + 2(6) − ( + 2(0))) = (72 + 12) = 42
0 2 2 3 3 2
12–15. For each of the following curves defined by given equations with
a parameter a > 0, find suitable parametric equations and evaluate
the arc length between a given point A and and a generic point B =
(x0, y0 , z0):
204 2. VECTOR FUNCTIONS

12. y = a arcsin(x/a), z = (a/4) ln[(a − x)/(a + x)], A = (0, 0, 0);

Solution: We begin by finding a suitable parameterization.


As ugly as the above equations may seem, it is actually best
(as far as I can tell) to use them directly (it is not obvious,
but it can be shown that z = −a/2 tanh−1 (2x/a); this does not
help, however). So we let x(t) = t, then y(t) = a arcsin(t/a)
and z(t) = (a/4)(ln(a − t) − ln(t + a)). Then A corresponds
to t = 0 and B corresponds to t = x0 . Next, r0(t) is

0 1 a 1 1 a a2 1
r (t) = h1, p , ( − )i = h1, √ , i
1 − (t/a)2 4 t − a t + a a2 − t2 2 t2 − a2
and so
s
a2 a4
kr0 (t)k = 1− +
t2 − a2 4(t2 − a2 )2
1 p
= 2 2
4(t2 − a2 )2 − 4a2 (t2 − a2) + a4
2(t − a )
1 p
= (2(t2 − a2 ) − a2)2
2(t2 − a2)
2t2 − 3a2 a
= 2 2
= 1 − 2 2
= 1 − z 0(t)
2(t − a ) 2(t − a )
where partial fraction decomposition has been used. Then the
arc length is
x0
a a − x0
Z
L= (1 − z 0(t))dt = (x0 − z(x0)) − (0 − z(0)) = x0 − ln[ ]
0 4 a + x0
13. (x − y)2 = a(x + y), x2 − y 2 = 9z 2 /8, A = (0, 0, 0); Hint: Use
new variables u = x + y and v = x − y to find parametric
equations;

Solution: As above, we begin by finding a suitable parame-


terization. Using the hint, we have that
u+v u−v
x= , y=
2 2
substituting u, v into the given equations produces
9z 2
v 2 = au, uv =
8
4. ARC LENGTH OF A CURVE 205

Multiply both sides of the second equation by a and substi-


tuting this into the second gives

8 3
2 2 2 3/2
z = v ⇔ z= √ v
9a 3 a

where we have chosen the positive square root.√ Finally,


√ 3/2 let
2
v(t) = t. Then u(t) = t /a and z(t) = (2 2)/(3 a)t .
Substituting these into the equations for x(t) and y(t) gives
x(t) = t2 /(2a) + t/2 and y(t) = t2/(2a) − t/2. Then
√ A cor-
responds to t = 0 and B corresponds to t0 = [(3 az0)]2/3/2.
Next, r0(t) is

2
r (t) = ht/a + 1/2, t/a − 1/2, √ t1/2i
0
a

and so
r r √
0 t2 t 1 t2 t 1 2t 2t2 2t 1 2t 1
kr (t)k = ( 2 + + ) + ( 2 − + ) + = + + = +√
a a 4 a a 4 a a2 a 4 a 2

Then the arc length is


t0
√ √
2t 1 2 2 t0
Z
L= ( + √ )dt = t0 + √
0 a 2 2a 2

14. x2 + y 2 = az, y = x tan(z/a), A = (0, 0, 0); Hint: Use polar


coordinates in the xy-plane to find parametric equations;

Solution: As above, we begin by finding a suitable


p parame-
terization. Using
p the hint, we have that x(t) = az(t)) cos(z(t)/a)
and y(t) = az(t) sin(z(t)/a). It is easy to verify that these
satisfy the given equations. So we may let z(t) = at. Thus
√ √
r(t) = ha t cos t, a t sin t, ati

Then A corresponds to t = 0 and B corresponds to t0 = z0 /a.


Next, r0(t) is

cos t √ sin t √ a √
r0(t) = ah √ − t sin t, √ + t cos t, 1i = √ hcos t−2t sin t, sin t+2t cos t, 2 ti
2 t 2 t 2 t
206 2. VECTOR FUNCTIONS

and so
a p
kr0(t)k = √ (cos t − 2t sin t)2 + (sin t + 2t cos t)2 + 4t
2 t
a p
= √ cos2 t − 4t sin t cos t + 4t2 sin2 t + sin2 t + 4t sin t cos t + 4t cos2 t + 4t
2 t
a √ a √ a
= √ 1 + 4t2 + 4t = √ (2t + 1) = a t + √
2 t 2 t 2 t
Then the arc length is

Z t0
a 2a 3/2 √ √ 2
L= (a t + √ )dt = t0 + a t0 = a t0(1 + t0)
0 2 t 3 3
p
15. x2 + y 2 + z 2 = a2 , x2 + y 2 cosh(arctan(y/x)) = a, A =
(a, 0, 0); Hint: Represent the second equation as a polar graph.

Solution: As above, we begin by finding a suitable param-


eterization. Using the hint, we let x = r cos t and y = r sin t.
Then
a
r cosh(t) = a ⇔ r =
cosh t
So we should have that x(t) = a/ cosh(t) cos t and y(t) =
a/ cosh(t) sin t. Thus

a2 a2 a2 (cosh2 t − 1 a2 sinh2 t
+z 2
= a 2
⇔ z 2
= a 2
− = = = a2 tanh2 t
cosh2 t cosh2 t cosh2 t cosh2 t
where the fundamental hyperbolic trigonometric identity cosh2 t−
sinh2 t = 1 has been used. Thus z(t) = a tanh t (taking the
positive root). So,
a cos t a sin t a
r(t) = h , , a tanh ti = hcos t, sin t, sinh ti
cosh t cosh t cosh t
Then A corresponds to t = 0 and B corresponds to t0 =
tanh−1 (z0/a). Next, r0 (t) is
−a sinh t a
r0(t) = 2 r(t) + h− sin t, cos t, cosh ti
cosh t cosh t
a
= h− sinh t cos t − sin t cosh t, − sinh t sin t + cosh t cos t, − sinh2 t + cosh2 ti
cosh2 t
a
= h−(cos t sinh t + sin t cosh t), cos t cosh t − sin t sinh t, 1i
cosh2 t
4. ARC LENGTH OF A CURVE 207

where the fundamental hyperbolic trigonometric identity cosh2 t−


sinh2 t = 1 has been used. And so,
a p
kr0(t)k = (−(cos t sinh t + sin t cosh t))2 + (cosh t cos t − sinh t sin t)2 + 1
cosh2 t p
a
= 2 sinh2 t + cosh2 t + 1
cosh t
a p
= cosh(2t) + 1
cosh2 t
a √
= √ e2t + e−2t + 2
2 cosh2 t
a t −t 2a et + e−t 2a
= √ 2
(e + e ) = √ 2
( )= √
2 cosh t 2 cosh t 2 2 cosh t
Then the arc length is
Z t0
2a 2a 2a z0
L= √ dt = √ (arctan(sinh t0 )) = √ arcsin( )
0 2 cosh t 2 2 a
as it can be shown that arctan(sinh(t)) = arcsin(tanh(t)) To
compute the integral, first compute the indefinite integral. The
steps are shown below
1 cosh t
Z Z
dt = 2 dt
cosh t cosh t
cosh t
Z
= dt
sin2 t + 1
1
Z
= 2
dt
u +1
= arctan(u) + C
= arctan(sinh t) + C
where the identity cosh2 t − sinh2 t = 1 and the substitution
u = sinh t have been used.
16–20. Reparameterize each of the following curves with respect to
the arc length measured from the point where t = 0 in the direction of
increasing t:
16. r(t) = ht, 1 − 2t, 5 + 3ti;

Solution: Our goal for these problems is to find


Z t
s(t) = kr0 (u)kdu,
0
208 2. VECTOR FUNCTIONS

then find the inverse of this, t(s), and substitute it for t in


r(t). This will give us a new vector function R(s) = r(t(s))
parameterized with respect to arc length from t = 0. We begin
by calculating r0 (t)
r0 (t) = h1, −2, 3i
and so √ √
kr0(t)k = 1 + 4 + 9 = 14
Thus we have that
Z t√ √
s(t) = 14du = 14t
0
and so t(s) = √s . Therefore the natural parameterization is
14
s 2s 3s
R(s) = h √ , 1 − √ , 5 + √ i
14 14 14
2 2
17. r(t) = 2t/(t + 1)ê1 + (2/(t + 1) − 1)ê3 ;

Solution: As above, we begin by calculating r0(t)


2(t2 + 1) − 2t(2t) 2 2
r0 (t) = h 2 2
, 0, 2 2
i= 2 h−t2 + 1, 0, −2ti
(t + 1) (t + 1) (t + 1)2
and so
2 p 2 √ 2
kr0(t)k = 2 2
(−t2 + 1)2 + (2t)2 = 2 2
t4 + 2t2 + 1 = 2
(t + 1) (t + 1) t +1
Thus we have that
t
2
Z
s(t) = du = 2 arctan t
0 u2 +1
and so t(s) = tan(s/2). Therefore the natural parameteriza-
tion is
2 tan(s/2) 2
R(s) = h 2 , 0, 2 − 1i
tan (s/2) + 1 tan (s/2) + 1
2 tan(s/2) 2
=h 2
, 0, 2
− 1i
sec (s/2) sec (s/2)
= h2 tan(s/2) cos2 (s/2), 0, 2 cos 2(s/2) − 1i
= h2 sin(s/2) cos(s/2), 0, 2 cos 2 (s/2) − 1i
= hsin s, 0, cos si
where the identity 2 cos2 θ − 1 = cos(2θ) has been used.
4. ARC LENGTH OF A CURVE 209

18. r(t) = hcosh t, sinh t, ti;

Solution: As above, we begin by calculating r0(t)


r0 (t) = hsinh t, cosh t, 1i
and so
p
2 2
p √
0
kr (t)k = sinh t + cosh t + 1 = 2 cosh2 t = 2 cosh t
where the identity cosh2 t − sinh2 t = 1 has been used. Thus
we have that
Z t√ √
s(t) = 2 cosh udu = 2 sinh t
0
−1

and
p so t(s) = sinh (s/ 2). Note that cosh(sinh−1 y) =
1 + y 2 . Therefore the natural parameterization is
r
s2 s s
R(s) = h 1 + , √ , sinh−1 √ i
2 2 2
19. x = a(t − sin t), y = a(1 − cos t), z = 0, a > 0, and 0 ≤ t ≤ 2π;

Solution: As above, we begin by calculating r0(t)


r0(t) = ha(1 − cos t), a sin t, 0i
and so
p √ √ √ t
kr0(t)k = a 1 − 2 cos t + cos2 t + sin2 t = a 2 − 2 cos t = 2a 1 − cos t = 2a sin
2
2
where the identity sin θ = (1 − cos θ)/2 has been used. No
absolute value is necessary since sin t/2 is positive on 0 ≤ t ≤
2π. Thus we have that
Z t
u t
s(t) = 2a sin du = 4a − 4a cos
0 2 2
and so t(s) = 2 arccos(1 − s/(4a)). Therefore the natural pa-
rameterization is
s s s
R(s) = ah2 arccos(1 − ) − sin(2 arccos(1 − )), 1 − cos(2 arccos(1 − )), 0i
4a 4a 4a
s s s
= ah2 arccos(1 − ) − 2 sin(arccos(1 − )) cos(arccos(1 − )),
4a 4a 4a
s 2
1 − 2 cos(arccos(1 − )) + 1, 0i
r4a
s s s s
= ah2 arccos(1 − ) − 2 1 − (1 − )2 (1 − ), 2 − 2(1 − )2 , 0i
4a 4a 4a 4a
210 2. VECTOR FUNCTIONS

20. r(t) = ethcos t, sin t, 1i.

Solution: As above, we begin by calculating r0 (t). Let


u(t) = hcos t, sin t, 1i, then
r0 (t) = etu(t) + etu0 (t)
where u0 (t) = h− sin t, cos t, 0i. Note that u(t) · u0 (t) = 0, so
they are orthogonal. Thus kr0(t)k may be computed using the
Pythagorean theorem, as follows
q √ √
0
kr (t)k = (et 2)2 + e2t = 3et

where
√ the factor of √ 2 comes from the fact that ku(t)k =
2 2
cos t + sin +1 = 2. Thus we have that
√ u √ √
Z t
s(t) = 3e du = 3et − 3
0

and so t(s) = ln(s/ 3 + 1). Therefore the natural parameter-
ization is
s s s s
R(s) = ( √ + 1)hcos(ln( √ + 1)), sin(ln( √ + 1)), ln( √ + 1)i
3 3 3 3
21. A particle travels from the point (R, 0, 0) into the positive quad-
rant along a helix of radius R that rises h units of length per turn. If
the z axis is the symmetry axis of the helix, find the position vector of
the particle after it travels the distance 4πR.

Solution: We must first parameterize the helix. Each turn occurs


with every 2π (if the arguments of sine and cosine are just t). So we
want that z(2π + (k + 1)) − z(2π + k) = h for all k. In particular,
z(2π) − z(0) = z(2π) = h. So z(t) = ht/(2π). Then x(t) = R cos t and
y(t) = R sin t. Indeed, if
ht
r(t) = hR cos t, R sin t, i

then the particle will start at (R, 0, 0) and travel into the positive quad-
rant, since both sine and cosine will be positive (for 0 ≤ t ≤ π/2). We
are looking for the T such that
Z T
4πR = kr0(t)kdt
0
so we must first find r0 (t). This is
h
r0 (t) = h−R sin t, R cos t, i

4. ARC LENGTH OF A CURVE 211

and so
r r
h h2
kr0(t)k = R2 sin2 t + R2 cos2 t + ( )2 = R2 +
2π 4π 2
Thus we have that
Z Tr r
h2 h2 4πR 8π 2R
R2 + 2 dt = R2 + 2 T = 4πR ⇔ T = q =√
0 4π 4π h2
R2 + 4π 4π 2 R + h2
2

The position of the particle at this time is


8π 2R 8π 2R 4πRh
r(T ) = hR cos( √ ), R sin( √ ), √ i
4π 2R + h2 4π 2R + h2 4π 2R + h2
22. A particle travels along a curve traversed by the vector function
√ = hu, cosh u, sinh ui from the point (0, 1, 0) with a constant speed
r(u)
2 m/s so that its x coordinate increases. Find the position of the
particle in one second.

Solution: Recall that kr0(t)k can be interpreted as the speed of a


particle with trajectory given by r0(t). So we require that kr0 (t)k = 2.
If D is the distance traveled over the interval a ≤ t ≤ b then
Z b
D= kr0(t)kdt
a

and so if a = 0 and b = 1, then D = 2. Thus we are looking for a
point on the curve r(u)
√ = hu, cosh u, sinh ui such that x is positive and
kr(u) − h0, 1, 0ik = 2. Note that r(0) = h0, 1, 0i. So it is sufficient to
find a value of u, u0 , such that
√ Z u0
L= 2= kr0(u)kdu
0
We have that
r0(u) = h1, sinh u, cosh ui
and so p √
kr0 (u)k = 1 + sinh2 u + cosh2 u = 2 cosh u
(this result was proven in Exercise 18). Thus,
Z u0 √ √ √
2 cosh udu = 2 sinh u0 = 2 ⇔ sinh u0 = 1
0

Recall that cosh(sinh−1 (y)) = 1 + y 2 and sinh−1 (y) = ln(y+ 1 + y 2),


p p

so √ √
r(u0 ) = hln(1 + 2), 2, 1i
212 2. VECTOR FUNCTIONS

where x is indeed positive, as required.

23. Let C be a smooth closed curve whose arclength is L. Let r(t), a ≤


t ≤ b, be a simple, continuously differentiable parameterization of C.
Prove that there is a number a ≤ t∗ ≤ b such that kr(t∗)k = L/(b − a).
Hint: Recall the integral mean value theorem.

Solution: The integral mean value theorem states that if f is contin-


uous on [a, b] then there exists a c such that
Z b
1
f(c) = f(x)dx
b−a a
In particular, take f(t) = kr0 (t)k. Since r0(t) is continuously differen-
tiable, f(t) is continuous. Hence there exists a t∗ such that
Z b
∗ 0 ∗ 1 L
f(t ) = kr (t )k = kr0 (t)kdt =
b−a a b−a
24. A particle travels in space a distance D in time T . Show that
there is a moment of time 0 ≤ t ≤ T at which the speed of the particle
coincides with the average speed D/T .

Solution: ThisR is an application of Exercise 23, with a = 0, b = T ,


T
and taking D = 0 kr0 (t)k, where kr0(t)k is the speed of a particle with
trajectory r(t). nd the target is not destroyed.
5. CURVATURE OF A SPACE CURVE 213

5. Curvature of a Space Curve


1–10. Find the curvature of each of the following parameterized curves
as a function of the parameter t, and find the radius of curvature at
the indicated point P :
1. r(t) = ht, 1 − t, t2 + 1i, P = (1, 0, 2);

Solution: Recall the definition of curvature of a parame-


terized curve,
kr0 (t) × r00(t)k
κ(t) =
kr0(t)k3
where r(t) must be twice differentiable and r0(t) 6= 0. We then
begin by calculating r0(t) as follows
r0 (t) = h1, −1, 2ti
and so
√ √ √
kr0(t)k = 1 + 1 + 4t2 = 2 1 + 2t2
and r00(t) is
r00 (t) = h0, 02i
thus,
r0 (t) × r00(t) = h−2, −2, 0i
so the curvature is

2 2 1
κ(t) = √ =
2 2(1 + 2t ) 2 3/2 (1 + 2t2 )3/2
Recall that the radius of curvature, ρ(t), is given by
1
ρ(t) =
κ(t)
We must first find the parameter of t such that r(t) = h1, 0, 2i.
Evidently this is when t = 1. Thus,
1
ρ(1) = = (1 + 2(1)2 )3/2 = 33/2
κ(1)
2. r(t) = ht2 , t, 1i, P = (4, 2, 1);

Solution: As above, we begin by calculating r0(t) as follows


r0(t) = h2t, 1, 0i
and so √
kr0 (t)k = 1 + 4t2
214 2. VECTOR FUNCTIONS

and r00(t) is
r00(t) = h2, 0, 0i
thus,
r0 (t) × r00(t) = h0, 0, 2i
so the curvature is
2
κ(t) =
(1 + 4t2 )3/2
To find the radius of curvature at P , we must first find the
parameter of t such that r(t) = h4, 2, 1i. Evidently this is
when t = 2. Thus,
1 (1 + 4(2)2 )3/2 173/2
ρ(2) = = =
κ(2) 2 2
3. y = sin(x/2), P = (π, 1);

Solution: Recall the curvature of a graph y = f(x) is given


by
|f 00(x)|
κ(x) =
(1 + f 0 (x)2 )3/2
Here, we have f(x) = sin(x/2), so f 0 (x) = 1/2 cos(x/2) and
f 00 (x) = −1/4 sin(x/2). Thus the curvature is
| − 1/4 sin(x/2)| 1 | sin(x/2)|
κ(x) = 2 3/2
=
(1 + (1/2 cos(x/2)) ) 4 (1 + 1/4 cos(x/2)2 )3/2
The given point P corresponds to x = π. Thus,
1 4(1 + 1/4 cos(π/2)2 )3/2
ρ(π) = = =4
κ(π) | sin(π/2)|
4. r(t) = h4t3/2 , −t2 , ti, P = (4, −1, 1);

Solution: As above, we begin by calculating r0(t) as follows


r0(t) = h6t1/2, −2t, 1i
and so √
kr0 (t)k = 36t + 4t2 + 1
and r00(t) is
r00 (t) = h3t−1/2, −2, 0i
thus,
r0(t) × r00(t) = h−2, −3t−1/2, −6t1/2i
5. CURVATURE OF A SPACE CURVE 215

so the curvature is
p
36t + 4 + 9/t
κ(t) =
(4t2 + 36t + 1)3/2
To find the radius of curvature at P , we must first find the
parameter of t such that r(t) = h4, −1, 1i. Evidently this is
when t = 1. Thus,
1 (4(1)2 + 36(1) + 1)3/2 413/2
ρ(1) = = p =
κ(1) 36(1) + 4 + 9/(1) 7
5. x = 1 + t2 , y = 2 + t3 , P = (2, 1);

Solution: Recall the curvature of a planar curve r(t) =


hx(t), y(t), 0i is
|x0y 00 − x00y 0|
κ(t) =
[(x0)2 + (y 0)2 ]3/2
We have that x0 (t) = 2t, y 0(t) = 3t2 and x00(t) = 2, y 00(t) = 6t.
Thus the curvature is
|2t(6t) − 2(3t2 )| 6t2 6
κ(t) = 2 2 2 3/2
= 3 2 3/2
=
[(2t) + (3t ) ] |t| [4 + 9t ] |t|[4 + 9t2 ]3/2
To find the radius of curvature at P , we must find the value of
t such that x(t) = 2, y(t) = 1. Evidently this is when t = −1.
Thus,
1 | − 1|[4 + 9(−1)2 ]3/2 133/2
ρ(−1) = = =
κ(−1) 6 6
6. x = et cos t, y = 0, z = et sin t, P = (1, 0, 0);

Solution: As with Exercise 5, we may use the formula for


the curvature of a planar curve; just replace y with z. We
have that x0(t) = et cos t − et sin t, z 0 (t) = et sin t + et cos t and
x00(t) = −2et sin t, z 00(t) = 2et cos t. Thus the curvature is
|(et cos t − et sin t)(2et cos t) − (et sin t + et cos t)(−2et sin t)| 2e2t 1
κ(t) = t t 2 t t 2 3/2
= t 3 3/2
= √
[(e cos t − e sin t) + (e sin t + e cos t) ] |e | [2] et 2
To find the radius of curvature at P , we must find the value of
t such that x(t) = 1, y(t) = 0, z(t) = 0. Evidently this is when
t = 0. Thus,
1 √ √
ρ(0) = = e(0) 2 = 2
κ(0)
216 2. VECTOR FUNCTIONS


7. r(t) = hln t, t, t2 i, P = (0, 1, 1);

Solution: As above, we begin by calculating r0(t) as follows


1 1
r0(t) = h , t−1/2, 2ti
t 2
and so r
0 1 1
kr (t)k = 2
+ + 4t2
t 4t
and r00(t) is
1 1
r00(t) = h− 2 , − t−3/2, 2i
t 4
thus,
3 4 1
r0 (t) × r00(t) = h t−1/2, , t−5/2 i
2 t 4
so the curvature is
p
9/(4t) + 16/t2 + 1/(16t5 )
κ(t) =
(1/t2 + 1/(4t) + 4t2)3/2
To find the radius of curvature at P , we must first find the
parameter of t such that r(t) = h0, 1, 1i. Evidently this is
when t = 1. Thus,
1 (1/(1)2 + 1/(4(1)) + 4(1)2 )3/2 (21/4)3/2 213/2
ρ(1) = = p = p = √
κ(1) 9/(4(1)) + 16/(1)2 + 1/(16(1)5 ) 293/16 2 293
8. r(t) = hcosh t, sinh t, 2 + ti, P = (1, 0, 2);

Solution: As above, we begin by calculating r0(t) as follows


r0 (t) = hsinh t, cosh t, 1i
and so
p √
kr0(t)k = sinh2 t + cosh2 t + 1 = 2 cosh t
and r00(t) is
r00(t) = hcosh t, sinh t, 0i
thus,
r0 (t) × r00(t) = h− sinh t, cosh t, −1i
so the curvature is

2 cosh t 1
κ(t) = √ =
( 2 cosh t) 3 2 cosh2 t
5. CURVATURE OF A SPACE CURVE 217

To find the radius of curvature at P , we must first find the


parameter of t such that r(t) = h1, 0, 2i. Evidently this is
when t = 1. Thus,
1 e + 1/e 2 (e + 1/e)2
ρ(1) = = 2 cosh2 t = 2 cosh(1)2 = 2( ) =
κ(1) 2 2

9. r(t) = het , 2t, e−ti, P = (1, 0, 1);

Solution: As above, we begin by calculating r0(t) as follows



r0 (t) = het , 2, −e−ti
and so
√ p
kr0(t)k = e2t + 2 + e−2t = (et + e−t )2 = et + e−t
and r00(t) is
r00(t) = het , 0, e−t i
thus,
√ √ √ √
r0(t) × r00(t) = h 2e−t , −2, − 2eti = 2he−t , − 2, −eti
so the curvature is
√ t √
2(e + e−t ) 2
κ(t) = =
(et + e−t )3 (et + e−t )2

(this happens to also be 2/(4 cosh 2 t)). To find the radius of
curvature at P , we must first find the parameter of t such that
r(t) = h1, 0, 1i. Evidently this is when t = 0. Thus,
1 (1 + 1)2 4
ρ(0) = = √ = √ = 23/2
κ(0) 2 2
10. r(t) = hsin t − t cos t, t, cos t + t sin ti, P = (0, 1, 1).

Solution: As above, we begin by calculating r0(t) as follows


r0 (t) = ht sin t, 1, t cos ti
and so
p √
kr0(t)k = t2 sin2 t + 1 + t2 cos2 t = 1 + t2
and r00(t) is
r00(t) = hsin t + t cos t, 0, cos t − t sin ti
thus,
r0(t) × r00(t) = hcos t − t sin t, t2, − sin t − t cos ti
218 2. VECTOR FUNCTIONS

so the curvature is

1 + t2 + t4
κ(t) =
(1 + t2)3/2
To find the radius of curvature at P , we must first find the
parameter of t such that r(t) = h0, 1, 1i. Evidently this is
when t = 1. Thus,
1 (1 + 1)3/2 23/2
ρ(1) = =√ = √
κ(1) 1+1+1 3
11. Find the curvature of r(t) = ht, t2 /2, t3 /3i at the point of its inter-
section with the surface z = 2xy + 2/3.

Solution: First we find the point of intersection by substituting


x(t), y(t), and z(t) into the equation of the surface, as follows
t3 t2 2
= 2(t)( ) + ⇔ t3 = 3t3 + 2 ⇔ t = −1
3 2 3
To find the curvature, we begin by calculating r0 (t)
r0 (t) = h1, t, t2i
and so √
kr0(t)k = 1 + t2 + t4
and r00(t) is
r00(t) = h0, 1, 2ti
thus,
r0 (t) × r00 (t) = ht2, −2t, 1i
Then the curvature is

t4 + 4t2 + 1
κ(t) =
(t4 + t2 + 1)3/2
and in particular κ(1) is
p √ √
1 + 4(1) + 1 6 2
κ(1) = = √ =
(1 + 1 + 1)3/2 3 3 3
12. Find the maximal and minimal curvatures of the graph y = cos(ax)
and the points at which they occur. Sketch the graph for a = 1 and
mark the points of the maximal and minimal curvature, local maxima
and minima of cos x, and the inflection points.

Solution: I will ignore the case when a = 0; this is a straight line


5. CURVATURE OF A SPACE CURVE 219

and has no curvature. We first find the formula for the curvature (of a
graph) as

|f 00(x)|
κ(x) =
[1 + f 0 (x)2 ]3/2

Then,

f 00 (x)f 000(x)/(|f 00(x)|)[1 + f 0 (x)2]3/2 − 3|f 00 (x)|f 0 (x)f 00(x)[1 + f 0 (x)2]1/2


κ0 (x) =
[1 + f 0 (x)2]3
f 00 (x)(f 000(x)(1 + f 0 (x)2) − 3f 0 (x)f 00 (x)2)
=
|f 00 (x)|[1 + f 0 (x)2]5/2

so long as |f 00(x)| =
6 0. The maximum and minimum curvature occur
at the critical points of κ0 (x), i.e. when κ0 (x) = 0 or does not exist.
This is when either f 00 (x)(f 000(x)(1 + f 0 (x)2 ) − 3f 0 (x)f 00 (x)2) = 0 or
|f 00(x)| = 0, since [1 + f 0 (x)2]5/2 can never be 0. However, |f 00(x)| = 0
exactly when f 00 (x) = 0, which is included in the first condition; so we
need only find x such that f 00 (x)(f 000(x)(1+ f 0(x)2) − 3f 0 (x)f 00(x)2 ) = 0.
Here we have f(x) = cos(ax). So f 0 (x) = −a sin(ax), f 00 (x) = −a2 cos(ax),
and f 000(x) = a3 sin(ax). Then we are looking for solutions to

−a2 cos(ax)(a3 sin(ax) + a5 sin3 (ax) + 3a5 sin(ax) cos2 (ax)) = 0


cos(ax) sin(ax)(1 + a2 sin2 (ax) + 3a2 cos2 (ax)) = 0
cos(ax) sin(ax)(1 + a2 + 2a2 cos2 (ax)) = 0

Evidently the solutions to cos(ax) = 0 and sin(ax) = 0 are x = (2k +


1)π/(2a) and x = kπ/a for integers k. However, there are no real
solutions to 1 + a2 + 2a2 cos2(ax) = 0, since the sum of three positive
numbers is never 0. So our only critical points are at x = kπ/(2a) and
x = kπ/a for integers k. Substituting these into the formula for κ gives

(2k + 1)kπ kπ a2
κ( ) = 0, κ( ) = 2 2 3/2
= a2
2a a (1 + a (0) )

so the minimum curvature occur at x = (2k + 1)kπ/(2a) and the max-


imum at x = kπ/a. These are shown on the first plot below, with the
points in red those with maximum curvature, and those in blue with
minimum. The second plot shows the min and max (again in blue and
red, respectively), and the points of inflection (green) of cos x.
220 2. VECTOR FUNCTIONS

13. Find the maximal and minimal curvatures of the graph y = 1/x.

Solution: Let f(x) = 1/x, so f 0 (x) = −1/x2 , f 00(x) = 2/x3 , and


f 000(x) = −6/x4 . Using the formula for κ0 (x) established in Exercise
12, we have
2/x3 (−6/x4 (1 + 1/x4 ) − 3(−1/x2 )(4/x6 ))
κ0(x) =
|2/x3 |(1 + 1/x4 )5/2
3|x3 |(−2/x4 (1 + 1/x4 ) + 4/x8 )
=
x3 (1 + 1/x4 )5/2
3|x3 |(−2/x4 + 2/x8 )
=
x3 (1 + 1/x4 )5/2
−6|x3|(x4 − 1)
=
x3 ∗ x8(1 + 1/x4 )5/2
−6|x3 |(x4 − 1)
=
x3 (1 + 1/x4 )1/2(x4 + 1)2
−6|x3 |(x4 − 1)
=
x(x4 + 1)5/2
It ends up being that the critical points are when x = 0, ±1 (it is easy
to check that limx→0 κ0 (x) = 0). We also have that
|2/x3 |
κ(x) =
(1 + 4/x4 )3/2

it ends up being that limx→0 κ(x) = 0 and κ(±1) = 2/2, so the max-
imal curvature occurs at x = ±1 and the minimal curvature at x = 0.
One could also argue that minimal curvature ”occurs” ax x approaches
±∞.
5. CURVATURE OF A SPACE CURVE 221

14. Use a geometrical interpretation of the curvature to guess the


point on the graphs y = ax2 and y = ax4 where the maximal curvature
occurs. Then verify your guess by calculations.

Solution: First, y = ax2 has a vertex. This intuitively seems to


be where maximal curvature is (especially considering maximal cur-
vature to be minimal radius of curvature). So our guess is at x = 0
for y = ax2. On the other hand, y = ax4 is fairly flat at x = 0 and
starts to curve up on both sides. So the maximal curvature will prob-
ably occur at two points fairly close to x = 0. To verify this, first let
f(x) = ax2. Then f 0 (x) = 2ax, f 00 (x) = 2a (note that |f 00(x)| = f 00 (x)),
and f 000(x) = 0. Then by Exercise 12, κ0 (x) is
−3(2ax)(2a)2 −24a3 x
κ0 (x) = =
(1 + 4a2x2 )5/2 (1 + 4a2x2 )5/2
Since the denominator is never 0, the only critical point of κ0 (x) is at
x = 0. Because κ0 (−1) > 0 and κ0(1) < 0, a sign chart reveals that this
is indeed a maximum, as desired.
Next let f(x) = ax4. Then f 0 (x) = 4ax3, f 00 (x) = 12ax2 (note that
|f 00(x)| = f 00 (x)), and f 000(x) = 24ax. Then by Exercise 12 we have
24ax(1 + 16ax6 ) − 3(4ax3 )(144a2 x4 ) 24ax(1 − 56a2 x6)
κ0 (x) = =
(1 + 16ax6 )5/2 (1 + 16a2 x6)5/2
One critical point then is at x = 0 whereas two more are at x =
±1/(56a2 )1 /6. A sign chart reveals that indeed, at x = 0 curvature is
minimal whereas at x = ±1/(56a2 )1/6 curvature is maximal.

15–17. Let f(x) be twice continuously differentiable function and κ(x)


be the curvature of the graph y = f(x).
15. Does κ attain a local maximum value at every local minimum
and maximum of f? If not, state an additional condition on
f under which the answer to this question is affirmative.

Solution: By Exercise 12, we have that


0 f 00 (x)(f 000(x)(1 + f 0 (x)2) − 3f 0 (x)f 00(x)2 )
κ (x) =
|f 00(x)|[1 + f 0 (x)2]5/2
so long as f 00 (x) 6= 0. Suppose f 0 (x0) = 0 (i.e., there is a local
min/max at x = x0). Then
f 00(x0 )f 000(x0)
κ0(x0 ) =
|f 00 (x0)|[1 + f 0 (x0)2 ]5/2
222 2. VECTOR FUNCTIONS

So κ0(x0 ) may not be 0. It will, however, always be 0, if in


addition, f 000(x0) = 0.

16. Prove that κ = 0 at inflection points of the graph.

Solution: By definition,

|f 00(x)|
κ(x) =
(1 + f 0 (x)2 )3/2

Clearly if f 00 (x) = 0, then κ(x) = 0. So κ = 0 at inflection


points.

17. Show by an example that the converse of the statement in Ex-


ercise 16 is not true, i.e., that the curvature vanishes at x = x0
does not imply that the point (x0, f(x0 )) is an inflection point.

Solution: Consider f(x) = 1/4x4 . Then f 0 (x) = x3 and


f 00 (x) = 3x2 . Note that f 00 (x) ≥ 0 for all x. A necessary
condition for a point x = x0 to be an inflection point of f is
that f 00(x0 ) = 0; however, this is not a sufficient condition.
One also needs that a change in the sign occurs, i.e., that f 00 is
positive or negative just before x = x0 and negative or positive
just after. So while

3x2
κ(x) =
(1 + x6)3/2

vanishes at x = 0, the fact that there is no change in sign


(f 00 (x) is either 0 or positive) ensures that x = 0 is not an
inflection point.
18. Let f be twice differentiable at x0 . Let T2 (x) be its Taylor poly-
nomial of the second degree about x = x0. Compare the curvatures of
the graphs y = f(x) and y = T2(x) at x = x0.

Solution: Recall that the second degree Taylor polynomial of f about


x = x0 is given by
1
T2(x) = f(x0 ) + f 0 (x0)(x − x0) + f 00(x0 )(x − x0 )2
2
5. CURVATURE OF A SPACE CURVE 223

The curvature for y = f(x) is easy to compute, and can be verified to


be
|f 00(x0 )|
κ(x0) =
(1 + f 0 (x0 )2)3/2
We have that T20 (x) = f 0 (x0) + f 00(x0 )(x − x0 ) and T200(x) = f 00 (x0), so
T20 (x0 ) = f 0 (x0) and T200(x0 ) = f 00 (x0). Since the curvature is only de-
pendent on the first and second derivatives, their curvatures are equal.

19–21. Find the curvature of each of the following polar graphs:


19. r = 1 + cos θ;

Solution: Let r = f(θ) be a polar graph. Study Problem


14.2 details that the curvature of a polar graph is
|2f 0 (θ)2 − f(θ)f 00 (θ) + f(θ)2 |
κ(θ) =
[f 0(θ)2 + f(θ)2 ]3/2
Here, f(θ) = 1 + cos θ, so f 0 (θ) = − sin θ and f 00(θ) = − cos θ.
Then the curvature is
|2(− sin θ)2 − (1 + cos θ)(− cos θ) + (1 + cos θ)2 | 3 + 3 cos θ 3
κ(θ) = 2 2 2
= 3/2
= √
[(sin θ) + (1 + cos θ) ] [2 + 2 cos θ] 2 2(1 + cos θ)1/2
20. r = eθ ;

Solution: We follow the same steps as above. First, let


f(θ) = eθ , so f 0 (θ) = eθ and f 00 (θ) = eθ . Then the curvature
is
|2e2θ − e2θ + e2θ | 2e2θ 1
κ(θ) = = √ = √ ,
(e2θ + e2θ )3/2 2 2e3θ 2eθ
which was actually found in Exercise 6.
21. r = | sin(3θ)|.

Solution: We follow the same steps as above. First, let


f(θ) = | sin(3θ)|. It is important to remember the derivative
of the absolute value of a function does not exist when the
function is 0, but does exist elsewhere. It can be explicitly
written as g 0(x)g(x)/|g(x)| if f(x) = |g(x)|. So,
3 cos(3θ) sin(3θ) 3 sin(6θ)
f 0 (θ) = =
| sin(3θ)| 2| sin(3θ)|
224 2. VECTOR FUNCTIONS

and
36 cos(6θ)| sin(3θ)| − 6 sin(6θ)( 2|3 sin(6θ)
sin(3θ)|
)
00
f (θ) =
(2| sin(3θ)|)2
36 cos(6θ)| sin(3θ)|2 − 3 sin(6θ)(3 sin(6θ))
=
| sin(3θ)|(2| sin(3θ)|)2
36(cos(6θ) sin(3θ)2 − sin2(3θ) cos2 (3θ))
=
4 sin2 (3θ)| sin(3θ)|
9(cos2(3θ) − sin2(3θ) − cos2 (3θ))
=
| sin(3θ)|
2
−9 sin (3θ)
=
| sin(3θ)|
Then the curvature is
2
|2( 3 cos(3θ) sin(3θ) 2
| sin(3θ)|
) − (| sin(3θ)|)( −9| sin(3θ)|
sin (3θ)
) + (| sin(3θ)|)2 |
κ(θ) =
(( 3 cos(3θ) sin(3θ) 2
| sin(3θ)|
) + (| sin(3θ)|)2)3/2
2 2 (3θ))
|2( (9 cos sin
(3θ) sin
2 (3θ) ) + 9 sin2(3θ) + sin2(3θ)|
= 2 2 (3θ))
(( (9 cos sin
(3θ) sin
2 (3θ) ) + sin2(3θ))3/2
|2((9 cos2 (3θ)) + 10 sin2 (3θ)|
=
(9 cos2 (3θ) + sin2 (3θ))3/2
|18 cos2 (3θ) + 10 sin2 (3θ)|
=
(9 cos2 (3θ) + sin2 (3θ))3/2
8 cos2 (3θ) + 10
=
(8 cos2(3θ) + 1)3/2
22–26. Find the equation of the osculating circle for each of the fol-
lowing planar curves at a specified point P :
22. y = x2 , P = (0, 0);

Solution: First recall the definition of the osculating circle

R(t) = r0 + ρ0 (1 − cos t)N̂0 + ρ0 sin tT̂0, 0 ≤ t < 2π


Thus to find the osculating circle at a point P , we must find ρ0 ,
N̂0 , and T̂0 . Calculation of the latter two is quite tedious for
space curves; however, for planar curves (like those given), it
is relatively easy. To begin with, first calculate the curvature,
5. CURVATURE OF A SPACE CURVE 225

which is
2
κ(x) =
(1 + 4x2 )3/2
so the radius of curvature at P , ρ0 , is
1 (1 + 4(0)2 )3/2 1
ρ0 = ρ(0) = = =
κ(0) 2 2
Next we must find the unit normal and tangent vectors. Let
the curve be parameterized by r(t) = ht,pf(t)i. Then the unit
tangent vector is given by T̂(t) = 1/( 1 + f 0 (t)2)h1, f 0 (t)i.
Now we need to find the unit normal vector, which is always
0
perpendicular to the unit tangent vector. Note
p that h−f (t),0 1i
is always perpendicular to T̂; so N̂ = ±1/( 1 + f 0 (t)2)h−f (t), 1i
(the appropriate sign must be chosen based on the concavity).
At t = 0, the unit normal and tangent vectors are N̂0 = h0, 1i
(positive chosen since f is concave up at t = 0) and T̂0 = h1, 0i.
Therefore the osculating circle is given by
1 1 1
R(t) = (1− cos t)h0, 1i+ sin th1, 0i = hsin t, 1− cos ti, 0 ≤ t < 2π
2 2 2
equivalently, this can be written as
1 1
x2 + (y − )2 =
2 4
(an easy way to find the above is to use the general form
(x − x0)2 + (y − y0 )2 = ρ20 , where x0 is the constant in the x
component of R(t) and y0 is the constant in the y component
of R(t)).

23. y = x − x2 /4, P = (2, 1);

Solution: As above, we begin by calculating the curvature,


which is
1/2
κ(x) =
(1 + (1 − x/2)2 )3/2
so the radius of curvature at P , ρ0 , is
1
ρ0 = ρ(2) = = 2(1 + (1 − 2/2)2 )3/2 = 2
κ(2)
Next we must find the unit normal and tangent vectors. At
t = 2, the unit normal and tangent vectors are N̂0 = h0, −1i
226 2. VECTOR FUNCTIONS

(negative chosen since f is concave down at t = 2) and T̂0 =


h1, 0i. Therefore the osculating circle is given by
R(t) = h2, 1i+2(1−cos t)h0, −1i+2 sin th1, 0i = h2+2 sin t, −1+2 cos ti, 0 ≤ t < 2π
equivalently, this can be written as
(x − 2)2 + (y + 1)2 = 4
24. y = 1/x, P = (1, 1);

Solution: As above, we begin by calculating the curvature,


which is
2/x3
κ(x) =
(1 + (−1/x2 )2 )3/2
so the radius of curvature at P , ρ0 , is

1 (1)3 (1 + (−1/(1)2 )2)3/2 2 2 √
ρ0 = ρ(1) = = = = 2
κ(1) 2 2
Next we must find the unit normal and tangent vectors.√At t =
1, the unit normal and tangent vectors are N̂0 = 1/ 2h1, 1i
(positive
√ chosen since f is concave up at t = 1) and T̂0 =
1/ 2h1, −1i. Therefore the osculating circle is given by
√ 1 √ 1
R(t) = h1, 1i+ 2(1−cos t)( √ )h1, 1i+ 2 sin t( √ )h1, −1i = h2+cos t+sin t, 2+cos t−sin ti
2 2
for 0 ≤ t < 2π. Equivalently, this can be written as
(x − 2)2 + (y − 2)2 = 2
25. x = a(t−sin t), y = a(1−cos t) (a cycloid) P = (a(π/2−1), a);

Solution: As above, we begin by calculating the curvature.


So x0 (t) = a(1−cos t), y 0 (t) = a sin t and x00(t) = a sin t, y 00(t) =
a cos t. Then the curvature is is
|a(1 − cos t)(a cos t) − (a sin t)(a sin t)| a2 | cos t − 1| 1
κ(x) = 2 = 3 3/2
= √
(a2(1 − cos t)2 + a2 sin t)3/2 a (2 − 2 cos t) 2a 2(1 − cos t)3/2
so the radius of curvature at P (which corresponds to t = π/2),
ρ0 , is
1 √ √
ρ0 = ρ(π/2) = = 2a 2(1 − cos π/2)3/2 = 2a 2
κ(π/2)
Next we must find the unit normal and tangent vectors. Note
0
that
√ r (π/2) = ha, ai. So the unit tangent vector is T̂0 =
1/ 2h1, 1i. The unit normal vector must be perpendicular
5. CURVATURE OF A SPACE CURVE 227


to this, so we have two options: ±1/ 2h−1, 1i. Choose the
negative version since the curve is concave down at t = π/2.
Therefore the osculating circle is given by
√ 1 √ 1
R(t) = ha(π/2 − 1), ai + 2a 2(1 − cos t)( √ )h1, −1i + 2a 2 sin t( √ )h1, 1i
2 2

=h + a − 2a cos t + 2a sin t, −a + 2a cos t + 2a sin ti, 0 ≤ t < 2π
2
equivalently, this can be written as

(x − ( + a))2 + (y + a)2 = 8a2
2
26. x = cos t, y = sin(2t), P = (1, 0) and P = (−1, 0).

Solution: I will first use P = (1, 0). As above, we begin by


calculating the curvature. So x0 (t) = − sin t, y 0(t) = 2 cos(2t)
and x00(t) = − cos t, y 00(t) = −4 sin(2t). Then the curvature is
|(− sin t)(−4 sin(2t)) − (− cos t)(2 cos(2t))| |4 sin t(sin(2t)) + 2 cos t(cos(2t))|
κ(x) = =
2 2
((− sin t) + (2 cos(2t)) ) 3/2 (sin2 t + 4 cos2 (2t))3/2
so the radius of curvature at P (which corresponds to t = 0),
ρ0 , is
1 (sin2 0 + 4 cos2 (0))3/2 43/2
ρ0 = ρ(0) = = = =4
κ(0) |4 sin(0)(sin(0)) + 2 cos(0)(cos(0))| 2
Next we must find the unit normal and tangent vectors. Note
that r0 (0) = h0, 2i. So the unit tangent vector is T̂0 = h0, 1i.
The unit normal vector must be perpendicular to this, so we
have two options: ±h1, 0i. Choose the negative version since
the curve bends inward to the origin at t = 0. Therefore the
osculating circle is given by
R(t) = h1, 0i+4(1−cos t)h−1, 0i+4 sin th0, 1i = h−3+4 cos t, 4 sin ti, 0 ≤ t < 2π
equivalently, this can be written as
(x + 3)2 + y 2 = 16
Now let P = (−1, 0), which corresponds to t = π Then the
radius of curvature, ρ0 , at P is
1 (sin2 (π) + 4 cos2(2π))3/2 43/2
ρ0 = ρ(π) = = = =4
κ(π) |4 sin(π)(sin(2π)) + 2 cos(π)(cos(2π))| 2
Next we must find the unit normal and tangent vectors. Note
that r0(π) = h0, 2i. So the unit tangent vector is T̂0 = h0, 1i.
The unit normal vector must be perpendicular to this, so we
228 2. VECTOR FUNCTIONS

have two options: ±h1, 0i. Choose the positive version since
the curve bends inward to the origin at t = 0. Therefore the
osculating circle is given by
R(t) = h−1, 0i+4(1−cos t)h1, 0i+4 sin th0, 1i = h3−4 cos t, 4 sin ti, 0 ≤ t < 2π
equivalently, this can be written as
(x − 3)2 + y 2 = 16
27. Find the maximal and minimal curvature of the ellipse x2 /a2 +
y 2/b2 = 1, a > b and the points where they occur. Give the equations
of the osculating circles at these points.

Solution: Assume a, b are positive. First, the ellipse may be param-


eterized as r(t) = ha cos t, b sin ti. Then x0(t) = −a sin t, y 0(t) = b cos t
and x00(t) = −a cos t, y 00(t) = −b sin t. Then the curvature at t is given
by
| − a sin t(−b sin t) − b cos t(−a cos t)| ab
κ(t) = 2 2 2 2 3/2
= 2 2
(a sin t + b cos t) (a sin t + b2 cos2 t)3/2
Differentiating this gives
−3ab(2a2 sin t cos t − 2b2 cos t sin t) −3(a2 − b2 )ab sin(2t)
κ0(t) = =
2(a2 sin2 t + b2 cos2 t)5/2 2(a2 sin2 t + b2 cos2 t)5/2
The critical points occur when 2t = 0, π, 2π, ..., when t = kπ/2 for
integers k. Evidently the sign of κ0 (t) depends only on sin(2t) and is
in fact the same sign as sin(2t) (by assumption, a > b, so a2 − b2 > 0).
It can be inferred from a sign chart that the maximal κ occur at t =
0, π, 2π, ... and the minimal κ occur at t = π/2, 3π/2, 5π/2, .... These
correspond to the points (a, 0) and (−a, 0) for maximal curvature and
(0, b) and (0, −b) for minimal. The curvature at each of these points is
ab ab a
κ(0) = κ(π) = = 3 = 2
(a2 sin2 (0)
2 2
+ b cos (0)) 3/2 b b
π 3π ab ab b
κ( ) = κ( ) = 2 2 π 2 2 π 3/2 = 3 = 2
2 2 (a sin ( 2 ) + b cos ( 2 )) a a

Let P = (a, 0). The unit normal vector is h−1, 0i (negative since the
ellipse bends in towards the origin). The center is ρ0 = b2/a units
from P in the direction of the unit normal vector; i.e., at (a − b2/a, 0).
Therefore the osculating circle is given by
b4
(x − a + b2/a)2 + y 2 =
a2
5. CURVATURE OF A SPACE CURVE 229

Let P = (−a, 0). The unit normal vector is h1, 0i (positive since the
ellipse bends in towards the origin). The center is ρ0 = b2/a units from
P in the direction of the unit normal vector; i.e., at (−a + b2/a, 0).
Therefore the osculating circle is given by
b4
(x + a − b2/a)2 + y 2 =
a2
Let P = (0, b). The unit normal vector is h0, −1i (negative since the
ellipse bends in towards the origin). The center is ρ0 = a2/b units
from P in the direction of the unit normal vector; i.e., at (0, b − a2 /b).
Therefore the osculating circle is given by

2 2 a4
2
x + (y − b + a /b) = 2
b
Let P = (0, −b). The unit normal vector is h0, 1i (positive since the
ellipse bends in towards the origin). The center is ρ0 = a2/b units from
P in the direction of the unit normal vector; i.e., at (0, −b + a2 /b).
Therefore the osculating circle is given by
a4
x2 + (y + b − a2/b)2 =
b2
28. Let r(t) = ht3 , t2 , 0i This curve is not smooth and has a cusp at
t = 0. Find the curvature for t 6= 0 and investigate its limit as t → 0

Solution: The curve is planar, so let x(t) = t3 and y(t) = t2. Then
x0(t) = 3t2 , y 0 (t) = 2t and x00(t) = 6t, y 00(t) = 2. Then the curvature
of the curve is given by
|3t2(2) − 6t(2t)| 6t2
κ(t) = =
((3t2)2 + (2t)2 )3/2 |t|3(9t2 + 4)3/2
for t 6= 0. We then have
6t2 6
lim κ(t) = lim 3 2 3/2
= lim = +∞
t→0+ t→0+ t (9t + 4) t→0+ t(9t + 4)3/2
2

6t2 6
lim κ(t) = lim 3 2 3/2
= lim − = +∞
t→0− t→0− −t (9t + 4) t→0− t(9t + 4)3/2
2

29. Show that the cardioid r = 1 + cos θ is not smooth at the origin.
Investigate the curvature of the cardioid as the origin is approached
along the cardioid.
230 2. VECTOR FUNCTIONS

Solution: By Exercise 19, the curvature of the given cardiod is


3
κ(θ) = √
2 2(1 + cos θ)1/2
The limit as θ approaches π is +∞. 30–31. Find an equation of the
osculating plane for each of the following curves at a specified point:
30. r(t) = h4t3/2 , −t2 , ti, P = (4, −1, 1);

Solution: Recall that the osculating circle, a planar curve,


lies completely in the osculating plane. The osculating circle
is also formed from a linear combination of N̂0 and T̂0. The
latter is parallel to r0 (t) while the former is a linear combi-
nation of r0 (t) and r00(t) (apply product rule). Therefore the
osculating plane contains the vectors r0 (t) and r00(t). Therefore
an equation of the osculating plane through a point P0 on a
curve C traced by a twice differentiable function r(t), where
the position vector of P0 is r(t0) = hx0 , y0, z0i, is
n1 (x − x0 ) + n2(y − y0) + n3 (z − z0) = 0
where n = r0(t0) × r00(t0 ) = hn1 , n2, n3 i. So we have that
r0(t) = h6t1/2, −2t, 1i
and
r00(t) = h3t−1/2 , −2, 0i
The point P = (4, −1, 1) has position vector r(t0 ) for t0 = 1.
Thus,
r0 (1) = h6, −2, 1i, r00(1) = h3, −2, 0i
So,
r0(1) × r00(1) = h2, 3, −6i
and an equation of the osculating plane is
2(x − 4) + 3(y + 1) − 6(z − 1) = 0 ⇔ 2x + 3y − 6z = −1

31. r(t) = hln t, t, t2 i, P = (0, 1, 1);

Solution: As above, we begin by computing r0 (t) and r00(t).


So we have that
1 1
r0(t) = h , t−1/2, 2ti
t 2
5. CURVATURE OF A SPACE CURVE 231

and
1 1
r00(t) = h− 2 , − t−3/2, 2i
t 4
The point P = (0, 1, 1) has position vector r(t0) for t0 = 1.
Thus,
1 1
r0 (1) = h1, , 2i, r00(1) = h−1, − , 2i
2 4
So,
3 1
r0 (1) × r00(1) = h , −4, i
2 4
and an equation of the osculating plane is
3 1
(x) − 4(y − 1) + (z − 1) = 0 ⇔ 6z − 16y + z = −15
2 4
32. Find an equation for the osculating and normal planes for the
curve r(t) = hln t, 2t, t2i at the point P0 of its intersection with the
plane y − z = 1. A plane is normal to a curve at a point if the tangent
to the curve at that point is normal to the plane.

Solution: First we find the point of intersection by substituting y(t)


and z(t) into the equation of the plane:
2t − t2 = 1 ⇔ t2 − 2t + 1 = 0 ⇔ (t − 1)2 = 0 ⇔ t = 1
And so t0 = 1, and the point of intersection is P = (0, 2, 1). Next we
compute r0(t) and r00(t)
1
r0(t) = h , 2, 2ti
t
and
1
r00(t) = h− 2 , 0, 2i
t
then at t0 = 1, l
r0 (1) = h1, 2, 2i, r00(1) = h−1, 0, 2i
therefore the normal plane is given by
(x − 0) + 2(y − 2) + 2(z − 1) = 0 ⇔ x + 2y + 2z = 6
To find the osculating plane, we must find r0(1) × r00(1), shown below
r0 (1) × r00(1) = h4, −4, 2i,
which is parallel to h2, −2, 1i. So the osculating plane is given by
2(x − 0) − 2(y − 2) + (z − 1) = 0 ⇔ 2x − 2y + z = −3
33. Is there a point on the curve r(t) = ht, t2 , t3 i where the osculating
plane is parallel to the plane 3x − 3y + z = 1?
232 2. VECTOR FUNCTIONS

Solution: This question is essentially asking if there exists a t, s


where r0(t) × r00(t) = sh3, −3, 1i. So we must calculate the first and
second derivatives, shown below:
r0(t) = h1, 2t, 3t2 i
and
r00(t) = h0, 2, 6ti
so r0(t) × r00(t) is
r0(t) × r00(t) = h6t2 , −6t, 2i = 2h3t2 , −3t, 1i
Evidently with t = 1 and s = 2, it is clear that r0 (1) × r00(1) is parallel
to h3, −3, 1i. The point is (1, 1, 1).

34. Prove that the trajectory of a particle has a constant curvature


if the particle moves so that the magnitudes of its velocity and accel-
eration vectors are constant.

Solution: Let the trajectory of a particle be given by r0(t). Then


the velocity and acceleration vectors are given by v(t) = r0 (t) and
a(t) = r00(t). Suppose the magnitudes of these vectors are constant;
then kr0 (t)k = v and kr00(t)k = a for constants v, a. But the curvature
is given by
kr0(t) × r00(t)k kr0(t)kkr00(t)k sin θ a sin θ
κ(t) = = =
kr0 (t)k3 kr0(t)k3 v2
All that is left is to show that sin θ is constant. To do this, consider
r0(t) · r0(t) = kr0(t)k2 = v 2. Differentiating both sides yields
r00(t) · r0 (t) + r0 (t) · r00(t) = 0 ⇔ r0(t) · r00(t) = 0
hence θ = π/2, and sin θ = 1, a constant.

Alternatively, one might consider the natural parameterization of the


trajectory, r(s), and simply note that κ(s) = kr00 (s)k, a constant, so κ
is constant.

35. Consider a graph y = f(x) such that f 00 (x0 ) 6= 0. At a point


(x0, y0 ) on the curve, where y0 = f(x0 ), find the equation of the oscu-
lating circle in the form (x − a)2 + (y − b)2 = R2 . Hint: Show first that
the vector h1, f 0 (x0 )i is tangent to the graph and a vector orthogonal
to it is h−f 0 (x0 ), 1i. Then consider two cases f 00 (x0) > 0 and f 00 (x0) < 0
5. CURVATURE OF A SPACE CURVE 233

Solution: Refer to Exercise 22 for the first part pof the hint. 0In
1/ 1 + f 0 (t)2h1, f (t)i
essence, the unit tangent vector is given by T̂(t) = p
and the unit normal vector is given by N̂(t) = ±1/ 1 + f 0 (t)2h−f 0 (t), 1i,
where plus is taken if the graph is concave up and negative is taken if
the graph is concave down. We now break the problem into cases.

Case 1: f 00 (x0) > 0. p


In this case, we take N̂ = 1/ 1 + f 0 (t)2h−f 0 (t), 1i. Since the graph
is concave up, the osculating circle’s center must lie above the curve,
because we want the circle to ”sit” in the graph. The center of the
osculating circle is found by traveling ρ(x0 ) units in the direction of N̂
from (x0, y0 ). In other words, the position vector of the center of the
circle is
f 0 (x0)ρ(x0 ) ρ(x0)
hx0, y0 i + ρ(x0 )N̂ = hx0 − p , y0 + p i
0
1 + f (x0 ) 2 1 + f 0 (x0)2
Recall that the radius of curvature is given by
1 [1 + f 0 (x)2 ]3/2
ρ(x) = =
κ(x) |f 00(x)|
at then ρ(x0) is
[1 + f 0 (x0)2 ]3/2 [1 + f 0 (x0)2 ]3/2
ρ(x0) = =
|f 00 (x0)| f 00 (x0)
since f 00 (x0) > 0. Then a = x0 − f 0 (x0)(1 + f 0 (x0 )2)/f 00 (x0 ) b = y0 +
(1 + f 0 (x0 )2)/f 00 (x0 ), and R2 = ρ(x0)2 = [1 + f 0 (x0)2 ]3/f 00 (x0)2 . Thus
the osculating circle is given by
f 0 (x0)(1 + f 0 (x0)2 ) 2 1 + f 0 (x0)2 2 (1 + f 0 (x0 )2 )3
(x−x0 + ) +(y −y 0 − ) =
f 00 (x0) f 00 (x0) f 00 (x0)2
p
Case 2: f 00 (x0) < 0 In this case, we take N̂ = −1/ 1 + f 0 (t)2 h−f 0 (t), 1i.
Note that the radius of curvature will still be the same, but will have
a negative sign in front (from evaluating the absolute value portion).
The combination of this negative sign with that of the unit normal
vector make a positive, which brings us exactly back to Case 1. The
equation is the same.

36. Let a smooth curve r = r(t) be planar and lie in the xy plane.
At a point (x0, y0) on the curve, find the equation of the osculating
circle in the form (x − a)2 + (y − b)2 = R2 . Hint: Use the result of
Study Problem 14.4 to express the constants a, b, and R via x0 , y0 ,
and the curvature at (x0 , y0).
234 2. VECTOR FUNCTIONS

Solution: First, it is obvious that R = ρ0 = 1/κ0 , where κ0 is the


curvature at (x0, y0). Let r(t) = hx(t), y(t),p0i, where r(t0) = hx0 , y0, 0i.
Then r0(t) = hx0 (t), y 0(t), 0i and T̂(t) = 1/ x0(t)2 + y 0(t)2hx0 (t), y 0(t), 0i.
The unit normal vector must be orthogonal to this. So we may choose

1
N̂(t) = p h−y 0 (t), x0(t), 0i
x0 (t)2 + y 0(t)2

The center of the osculating circle is found by traveling ρ0 units in the


direction of N̂0 = N̂(t0). Recall the radius of curvature can be found
using

1 (x0(t0 )2 + y 0(t0)2 )3/2


ρ0 = = 0
κ0 x (t0)y 00(t0 ) − x00(t0 )y 0(t0)

then the position vector of the center of the osculating circle is given
by

x0 (t0)2 + y 0(t0)2
hx0 , y0i + 0 00 00 0
h−y 0(t0), x0(t0 )i
x (t0)y (t0) − x (t0)y (t0 )

So,

0 x0 (t0)2 + y 0(t0)2 0 x0 (t0)2 + y 0(t0)2


a = x0−y (t0 )( 0 ), b = y0+x (t0 )( 0 )
x (t0)y 00(t0) − x00(t0)y 0 (t0) x (t0)y 00(t0) − x00(t0)y 0 (t0)

36. Find parametric equations of the osculating circle to the curve


r(t) = h4t3/2, −t2, ti at the point P = (4, −1, 1) by using the method
of Study Problem 14.4.

Solution: First, the point P corresponds to t = 1. We must find


5. CURVATURE OF A SPACE CURVE 235

T̂0 = T̂(1) and N̂0 = N̂(1). Then,


r0(t) = h6t1/2, −2t, 1i

kr0(t)k = 36t + 4t2 + 1
1 1
T̂(t) = 0
r0(t) = √ h6t1/2 , −2t, 1i
kr (t)k 36t + 4t2 + 1
1
⇒ T̂0 = T̂(1) = √ h6, −2, 1i
41
18 + 4t 1
T0(t) = − 2 3/2
h6t 1/2
, −2t, 1i + √ h3t−1/2 , −2, 0i
(36t + 4t + 1) 2
36t + 4t + 1
22 41 1
T0(1) = − √ h6, −2, 1i + √ h3, −2, 0i = √ h−9, −38, −22i
41 41 41 41 41 41

0 1 p
2 2 2
2009 7
⇒ kT (1)k = √ (−9) + (−38) + (−22) = √ =
41 41 41 41 41
1 1
N̂0 = 0
T0 (1) = √ h−9, −38, −22i
kT (1)k 7 41
0
√ √
1 kr (1)k 41 41 41
ρ0 = = = =
κ(1) kT0(1)k 7/41 7
The general equation of the osculating circle is
R(t) = r0 + ρ0(1 − cos t)N̂0 + ρ0 sin tT̂0, 0 ≤ t < 2π
where r0 = h4, −1, 1i. Let R(t) = hX(t), Y (t), Z(t)i. Then
√ √
41 41 9 41 41 6 173 369 246
X(t) = 4 + (1 − cos t)(− √ ) + sin t( √ ) = − + cos t + sin t
7 7 41 7 41 49 49 7
√ √
41 41 38 41 41 −2 1607 1558 82
Y (t) = −1 + (1 − cos t)(− √ ) + sin t( √ ) = − + cos t − sin t
7 7 41 7 41 49 49 7
√ √
41 41 22 41 41 1 853 902 41
Z(t) = 1 + (1 − cos t)(− √ ) + sin t( √ ) = − + cos t + sin t
7 7 41 7 41 49 49 7
236 2. VECTOR FUNCTIONS

6. Applications to Mechanics and Geometry


1–7. For each of the following trajectories of a particle, find the velocity,
speed, and the normal and tangential acceleration as functions of time,
and their values at a specified point P :
1. r(t) = ht, 1 − t, t2 + 1i, P = (1, 0, 2);

Solution: First we find the velocity and acceleration vec-


tors
v(t) = r0(t) = h1, −1, 2ti
a(t) = r00(t) = h0, 0, 2i
then the speed is
√ √
v(t) = kv(t)k = 1 + 1 + 4t2 = 2 + 4t2 .
Recall that the tangential and normal acceleration are given
by
v(t) · a(t)
aT (t) =
v(t)
kv(t) × a(t)k
aN (t) =
v(t)
Thus we have,
4t
aT (t) = √
2 + 4t2

kh−2, −2, 0ik 2 2 2
aN (t) = √ =√ =√
1 + 1 + 4t2 2 + 4t2 1 + 2t2
The point P = (1, 0, 2) corresponds to t = 1. The values of
the specified functions at this t are
v(1) = = h1, −1, 2i
a(1) = = h0, 0, 2i

v(1) = = 6

2 2
aT (1) = √
3
2
aN (1) = √
3
2. r(t) = ht2 , t, 1i, P = (4, 2, 1);
6. APPLICATIONS TO MECHANICS AND GEOMETRY 237

Solution: As above, first we find the velocity and accel-


eration vectors
v(t) = r0 (t) = h2t, 1, 0i
a(t) = r00(t) = h2, 0, 0i
then the speed is

v(t) = kv(t)k = 1 + 4t2 ,
and the tangential and normal acceleration are
v(t) · a(t) 4t
aT (t) = =√
v(t) 1 + 4t2
kv(t) × a(t)k kh0, 0, −2ik 2
aN (t) = = √ =√
v(t) 1 + 4t2 1 + 4t2
The point P = (4, 2, 1) corresponds to t = 2. The values of
the specified functions at this t are
v(2) = h4, 1, 0i
a(2) = h2, 0, 0i

v(2) = 17
8
aT (2) = √
17
2
aN (2) = √
17
3. r(t) = h4t3/2 , −t2 , ti, P = (4, −1, 1);

Solution: As above, first we find the velocity and accel-


eration vectors
v(t) = r0(t) = h6t1/2 , −2t, 1i
a(t) = r00(t) = h3t−1/2 , −2, 0i
then the speed is

v(t) = kv(t)k = 1 + 36t + 4t2,
and the tangential and normal acceleration are
v(t) · a(t) 18 + 4t
aT (t) = =√
v(t) 1 + 36t + 4t2
p
kv(t) × a(t)k kh2, 3t−1/2 , −6t1/2ik 4 + 9/t + 36t
aN (t) = = √ = √
v(t) 1 + 36t + 4t 2 1 + 36t + 4t2
238 2. VECTOR FUNCTIONS

The point P = (4, −1, 1) corresponds to t = 1. The values of


the specified functions at this t are
v(1) = h6, −2, 1i
a(1) = h3, −2, 0i

v(1) = 41
22
aT (1) = √
41
7
aN (1) = √
41

4. r(t) = hln t, t, t2 i, P = (0, 1, 1);

Solution: As above, first we find the velocity and accel-


eration vectors
1 1
v(t) = r0(t) = h , 1/2 , 2ti
t 2t
1 1
a(t) = r00(t) = h− 2 , − 3/2 , 2i
t 4t
then the speed is
p
v(t) = kv(t)k = 1/t2 + 1/(4t) + 4t2,
and the tangential and normal acceleration are
v(t) · a(t) −1/t3 − 1/(8t2 ) + 4t
aT (t) = =p
v(t) 1/t2 + 1/(4t) + 4t2
p
kv(t) × a(t)k kh3/(2t1/2 ), −4/t, 1/(4t5/2)ik 9/(4t) + 16/t2 + 1/(16t5 )
aN (t) = = p = p
v(t) 1/t2 + 1/(4t) + 4t2 1/t2 + 1/(4t) + 4t2
The point P = (0, 1, 1) corresponds to t = 1. The values of
the specified functions at this t are
1
v(1) = h1, , 2i
2
1
a(1) = h−1, − , 2i
4
p 1√
v(1) = 1 + 1/4 + 4 = 21
2
23
aT (1) = √
4 21
6. APPLICATIONS TO MECHANICS AND GEOMETRY 239


293
aN (1) = √
2 21
5. r(t) = hcosh t, sinh t, 2 + ti, P = (1, 0, 2);

Solution: As above, first we find the velocity and accel-


eration vectors

v(t) = r0(t) = hsinh t, cosh t, 1i


a(t) = r00(t) = hcosh t, sinh t, 0i

then the speed is


p √
v(t) = kv(t)k = sinh2 t + cosh2 t + 1 = 2 cosh t,

and the tangential and normal acceleration are


v(t) · a(t) 2 sinh t cosh t √
aT (t) = = √ = 2 sinh t
v(t) 2 cosh t

kv(t) × a(t)k kh− sinh t, cosh t, −1ik 2 cosh t
aN (t) = = √ =√ =1
v(t) 2 cosh t 2 cosh t
The point P = (1, 0, 2) corresponds to t = 0. The values of
the specified functions at this t are

v(0) = h0, 1, 1i
a(0) = h1, 0, 0i

v(0) = 2
aT (0) = 0
aN (0) = 1

6. r(t) = het , 2t, e−ti, P = (1, 0, 1);

Solution: As above, first we find the velocity and accel-


eration vectors

v(t) = r0 (t) = het, 2, −e−t i
a(t) = r00(t) = het , 0, e−t i

then the speed is



v(t) = kv(t)k = e2t + 2 + e−2t = et + e−t ,
240 2. VECTOR FUNCTIONS

and the tangential and normal acceleration are


v(t) · a(t) e2t − e−2t
aT (t) = = t = et − e−t
v(t) e + e−t
√ √ √ t
kv(t) × a(t)k kh 2e−t , −2, − 2et ik 2(e + e−t ) √
aN (t) = = = = 2
v(t) et + e−t et + e−t
The point P = (1, 0, 1) corresponds to t = 0. The values of
the specified functions at this t are

v(0) = h1, 2, −1i
a(0) = h1, 0, 1i
v(0) = 2
aT (0) = 0

aN (0) = 2
7. r(t) = hsin t − t cos t, t2 , cos t + t sin ti, P = (0, 0, 1).

Solution: As above, first we find the velocity and accel-


eration vectors
v(t) = r0 (t) = ht sin t, 1, t cos ti
a(t) = r00(t) = hsin t + t cos t, 0, cos t − t sin ti
then the speed is

v(t) = kv(t)k = 1 + t2 ,
and the tangential and normal acceleration are
v(t) · a(t) t
aT (t) = =√
v(t) 1 + t2

kv(t) × a(t)k khcos t − t sin t, t2, − sin t − t cos tik 1 + t2 + t4
aN (t) = = √ = √
v(t) 1 + t2 1 + t2
The point P = (0, 0, 1) corresponds to t = 0. The values of
the specified functions at this t are
v(0) = h0, 1, 0i
a(0) = h0, 0, 1i
v(0) = 1
aT (0) = 0
aN (0) = 1
6. APPLICATIONS TO MECHANICS AND GEOMETRY 241

8. Find the normal and tangential accelerations of a particle with the


position vector r(t) = ht2 + 1, t, t2 − 1i when the particle is at the least
distance from the origin.

Solution: Note that the distance from the particle’s position at a


moment t to the origin is given by kr(t)k. The find when the distance
is at a minimum, differentiate this. By Exercise 28 in section 11, we
have
r(t) · r0 (t) r(t) · v(t)
kr(t)k0 = =
kr(t)k kr(t)k
so we are looking for t such that r(t) · v(t) = 0. We have that
v(t) = h2t, 1, 2ti
a(t) = h2, 0, 2i
Thus
r(t) · v(t) = 2t3 + 2t + t + 2t3 − 2t = 4t3 − t
which is evidently 0 if t = 0 or t = ±1/2. The distance to the origin is

kr(t)k = 2t4 + t2 + 2
Evidently kr(1/2)k = kr(−1/2)k > kr(0)k. Thus the particle is at the
least distance from the origin at t = 0.
Next, the velocity and acceleration at t = 0 are
v(0) = h0, 1, 0i
a(0) = h2, 0, 2i
So the speed at t = 0 is v(0) = 1. Then the tangential and normal
accelerations are
0
aT (0) = = 0
1
kh2, 0, −2ik √
aN (0) = =2 2
1

9. Find the tangential and normal accelerations of a particle with


the position vector r(t) = hR sin(ωt − φ0), −R cos(ωt − φ0), v0ti where
R, ω, φ0 , and v0 are constants (see Study Problem 15.4).

Solution: We have that the velocity and acceleration vectors are


v(t) = hωR cos(ωt − φ0), ωR sin(ωt − φ0 ), v0i
a(t) = h−ω 2 R sin(ωt − φ0), ω 2 R cos(ωt − φ0), 0i
242 2. VECTOR FUNCTIONS

therefore the speed is


q
v(t) = ω 2 R2 + v02
and the tangential and normal acceleration are
−ω 3R2 sin(ωt − φ0 ) cos(ωt − φ0 ) + ω 3 R2 sin(ωt − φ0) cos(ωt − φ0)
aT (t) = p =0
ω 2 R2 + v02
p
kh−v0ω 2 R cos(ωt − φ0), −v0ω 2 R sin(ωt − φ0 ), ω 3 R2 ik ω 2 R v02 + ω 2 R2
aN (t) = p = p = ω2 R
ω 2 R2 + v02 ω 2 R2 + v02
This is essentially the general form of the converse of Study Problem
15.4

10. The shape of a winding road can be approximated by the graph


y = L cos(x/L), where the coordinates are in miles and L = 1 mile.
The condition of the road is such that if the normal acceleration of a
car on it exceeds 10 m/s2 , the car may skid off the road. Recommend
a speed limit for this portion of the road.

Solution: First convert L to meters, so that y = f(x) = aL cos(x/(aL)),


where a = 1609.34 m/mile. Recall that
v2
aN (x) = κ(x)v 2 =
ρ
Here, f 0 (x) = sin(x/(aL)) and f 00(x) = 1/(aL) cos(x/(aL)). Then
| cos(x/(aL))|
κ(x) =
aL(1 + sin(x/(aL))2 )3/2
The maximal curvature occurs at x = 0, since cos is at a maximum
there and sin is at a minimum. This curvature is
1
κ(0) =
aL
Let am be the maximum normal acceleration allowed. Then a reason-
able speed limit is to bound it using the maximum curvature. Thus,
set
v2
amax
N = a N (0) = κ(0)v 2
=
aL
we bound this above by am , and so
v2 p
< am ⇒ v < am (aL)
aL
6. APPLICATIONS TO MECHANICS AND GEOMETRY 243

note that this will have correct units, since am is in m/s2 and aL is in
m. So a suitable speed limit is, in m/s
p p
am (aL) = 10(1609) ≈ 126.86
Conversion of this to km/h or mph yields 456.7 km/h or 283.7 mph

11. A particle moves along the curve y = x2 + x3 in the direction


of increasing x. If the acceleration of the particle at the point (1, 2) is
a = h−3, −1i, find its normal and tangential accelerations.

Solution: Let f(x) = x2 + x3 . Then r(t) = ht, f(t) paremeterizes


the curve. Thus a tangent vector to the point (1, 2) (corresponding to
t = 1) is
r0 (1) = h1, f 0 (1)i = h1, 2(1) + 3(1)2 i = h1, 5i
√ √
Then the speed is kr0 (1)k = 1 + 25 = 26. The tangential and
normal accelerations are
h1, 5i · h−3, −1i 8
aT = √ = −√
26 26
kh1, 5, 0i · h−3, −1, 0ik 14
aN = √ =√
26 26
(note that v/kvk is just the unit tangent vector, so we may choose any
tangent vector to be v; in particular, we have chosen v = r0 (1)).

12. Suppose that a particle moves so that its tangential acceleration


aT is constant, while the normal acceleration aN remains 0. What is
the trajectory of the particle?

Solution: Recall that aT and aN are related by the equations


aT = v 0 , aN = κv 2
Suppose aT = a, where a is a constant. Then v 0 = a ⇒ v = at + v0 for
a constant v0. Suppose now that aN = 0. Then either κ = 0 or v 2 = 0,
or both. Suppose that v = 0. Then r = 0. Suppose now that v 6= 0.
Then κ = 0. Recall that straight lines have 0 curvature.

13. Suppose that a particle moves so that its tangential acceleration


aT remains 0, while the normal acceleration aN is constant. What is
the trajectory of the particle? Hint: Investigate the curvature of the
trajectory
244 2. VECTOR FUNCTIONS

Solution: As in Exercise 12, we will need the following equations


aT = v 0 , aN = κv 2
Suppose aT = 0. Then v 0 = 0 ⇒ v = v0, where v0 is a constant. Sup-
pose now that aN is constant. Then κ = an /v02 is a constant. Recall
that circles have constant curvature, hence this is a circle.

14. A race car moves with a constant speed v0 along an elliptic track
x2/a2 + y 2/b2 = 1, a > b. Find the maximal and minimal values of the
magnitude of its acceleration and the points where they occur.

Solution: By Exercise 27 of Section 14, maximal curvature occurs


at (±a, 0) and minimal curvature occurs at (0, ±b), where κmax = a/b2
and κmin = b/a2 . Recall that
aT = v 0 , aN = κv 2
Since v = v0, a constant, aT = 0. Moreover, since
a(t) = aT T̂ + aN N̂
where T̂ and N̂ are orthogonal, the total acceleration may be computed
using the Pythagorean theorem as
q
a = a2T + a2N
(positive since magnitude). Hence,
q
a = 0 + a2N = |aN | = |κv02|
Thus the maximal acceleration is av02/b2 and the minimal acceleration
is bv02/a2.

15. Does there exist a curve with zero curvature and a non-zero tor-
sion? Explain the answer.

Solution: Recall the Frenet-Serret equations


T̂0(s) = κ(s)N̂(s), N̂0 (s) = −κ(s)T̂(s)+τ (s)B̂(s), B̂0 (s) = −τ (s)N̂(s)
Suppose κ(s) = 0 and τ (s) = τ0 . Then,
T̂0 (s) = 0, N̂0(s) = τ0B̂(s), B̂0 (s) = −τ0 N̂(s)
Let r(s) be the natural parameterization of the curve, then r0 (s) =
T̂(s), and so r00(s) = T̂0(s) = 0. Integration of this gives r(s) = r0 +T̂0s
where r0 and T̂0 are constant vectors with T̂(s) = T̂0 , so a curve does
6. APPLICATIONS TO MECHANICS AND GEOMETRY 245

exist and it is a straight line. Take the derivative of both sides of the
second equation and substitute B̂0 (s) into the third. This yields

1/τ0 N̂00(s) = −τ0 N̂(s)

Let x(t) be one component of N̂(s). Then we have


1 00
x(t) = − x (t)
τ02
Suppose x(t) = A cos(τ0 s) + B sin(τ0 t). Then,
1
A cos(τ0 s)+B sin(τ0 s) = − (−Aτ02 cos(τ0 s)−Bτ02 sin(τ0 s)) = A cos(τ0 s)+B sin(τ0 s)
τ02
which is true for any choice of A, B. Then,

N̂(s) = cos(τ0 s)u + sin(τ0 s)v


for constant vectors u, v. Substituting s = 0 yields

N̂0 = N̂(0) = u
Taking the derivative of this gives

N̂0 (s) = −τ0 sin(τ0 s)N̂0 + τ0 cos(τ0 s)v = τ0 B̂(s)


And so,
B̂(s) = − sin(τ0 s)N̂0 + cos(τ0 s)v
Substitution of s = 0 yields

B̂0 = B̂(s) = v
and so
N̂(s) = cos(τ0 s)N̂0 + sin(τ0 s)B̂0
B̂(s) = cos(τ0 s)B̂0 − sin(τ0 s)N̂0
which describes a rotation (note the resemblance to the rotation ma-
trix!) So, as s increases, N̂(s) and B̂(s) rotate around the line, hence a
nonzero torsion. 16–20. For each of the following curves, find the unit
tangent, normal, and binormal vectors and the torsion at a specified
point P :
16. r(t) = ht, 1 − t, t2 + 1i, P = (1, 0, 2);
246 2. VECTOR FUNCTIONS

Solution: First recall the equations for the unit tangent,


normal, and binormal vectors, and torsion:

r0(t0)
T̂(t0 ) =
kr0(t0)k
r0(t0) × r00(t0 )
B̂(t0 ) = 0
kr (t0) × r00(t0 )k
N̂(t0 ) = B̂(t0) × T̂(t0)
r000(t0 ) · (r0(t0) × r00(t0 ))
τ (t0) =
kr0 (t0) × r00(t0)k2

So we must first find r0 (t0), r00(t0), and r000(t0). The given point
corresponds to t0 = 1. So,

r0(t) = h1, −1, 2ti ⇒ r0(1) = h1, −1, 2i


r00(t) = h0, 0, 2i ⇒ r00 (1) = h0, 0, 2i
r000(t) = h0, 0, 0i ⇒ r000(1) = 0

Thus we have

h1, −1, 2i 1
T̂(1) = = √ h1, −1, 2i
kh1, −1, 2ik 6
h−2, −2, 0i 1
B̂(1) = = √ h−1, −1, 0i
kh−2, −2, 0ik 2
1 1 1
N̂(1) = √ √ h−1, −1, 0i × h1, −1, 2i = √ h−2, 2, 2i = √ h−1, 1, 1i
2 6 2 3 3
0 · (h−1, −1, 0i)
τ (1) = √ 2 =0
2

17. r(t) = ht3 , t2 , 1i, P = (8, 4, 1);

Solution: As above, we must first find r0(t0), r00(t0 ), and


r000(t0). The given point corresponds to t0 = 2. So,

r0 (t) = h3t2 , 2t, 0i ⇒ r0 (2) = h12, 4, 0i


r00(t) = h6t, 2, 0i ⇒ r00(2) = h12, 2, 0i
r000(t) = h6, 0, 0i ⇒ r000(2) = h6, 0, 0i
6. APPLICATIONS TO MECHANICS AND GEOMETRY 247

Thus we have
h12, 4, 0i 1
T̂(2) = = √ h3, 1, 0i
kh12, 4, 0ik 10
h0, 0, −24i
B̂(2) = = h0, 0, −1i
kh0, 0, −24ik
1 1
N̂(2) = √ h0, 0, −1i × h3, 1, 0i = √ h1, −3, 0i
10 10
h6, 0, 0i · (h0, 0, −1i)
τ (2) = =0
12
18. r(t) = h4t3/2 , −t2 , ti, P = (4, −1, 1);

Solution: As above, we must first find r0(t0), r00(t0 ), and


r000(t0). The given point corresponds to t0 = 1. So,

r0 (t) = h6t1/2 , −2t, 1i ⇒ r0(1) = h6, −2, 1i


r00 (t) = h3t−1/2 , −2, 0i ⇒ r00(1) = h3, −2, 0i
r000(t) = h−3/2t−3/2 , 0, 0i ⇒ r000(1) = h−3/2, 0, 0i
Thus we have
h6, −2, 1i 1
T̂(1) = = √ h6, −2, 1i
kh6, −2, 1ik 41
h2, 3, −6i 1
B̂(1) = = h2, 3, −6i
kh2, 3, −6ik 7
1 1
N̂(1) = √ h2, 3, −6i × h6, −2, 1i = √ h−9, −38, 22i
7 41 7 41
h−3/2, 0, 0i · (h2, 3, −6i) 3
τ (1) = 2
=−
7 49
√ 2
19. r(t) = hln t, 2 t, t i, P = (0, 2, 1);

Solution: As above, we must first find r0(t0), r00(t0 ), and


r000(t0). The given point corresponds to t0 = 1. So,
1
r0 (t) = h , t−1/2, 2ti ⇒ r0 (1) = h1, 1, 2i
t
1 1 1
r00(t) = h− 2 , − t−3/2, 2i ⇒ r00(1) = h−1, − , 2i
t 2 2
000 2 3 −5/2 000 3
r (t) = h 3 , t , 0i ⇒ r (1) = h2, , 0i
t 4 4
248 2. VECTOR FUNCTIONS

Thus we have
h1, 1, 2i 1
T̂(1) = = √ h1, 1, 2i
kh1, 1, 2ik 6
h3, −4, 1/2i 2 1
B̂(1) = =√ h3, −4, 1/2i = √ h6, −8, 1i
kh3, −4, 1/2ik 101 101
1 1
N̂(1) = √ √ h6, −8, 1i × h1, 1, 2i = √ h−17, −11, 14i
101 6 606
h2, 34 , 0i · (h3, −4, 1i) 4(6 − 3) 12
τ (1) = √ 2 = =
101 /22 101 101
It is important to use the non-simplified form of the cross prod-
uct, because its magnitude must be squared.

20. r(t) = hcosh t, sinh t, 2 + ti, P = (1, 0, 2).

Solution: As above, we must first find r0(t0), r00(t0 ), and


r000(t0). The given point corresponds to t0 = 0. So,
r0(t) = hsinh t, cosh t, 1i ⇒ r0 (0) = h0, 1, 1i
r00(t) = hcosh t, sinh t, 0i ⇒ r00(0) = h1, 0, 0i
r000(t) = hsinh t, cosh t, 0i ⇒ r000(0) = h0, 1, 0i
Thus we have
h0, 1, 1i 1
T̂(0) = = √ h0, 1, 1i
kh0, 1, 1ik 2
h0, 1, −1i 1
B̂(0) = = √ h0, 1, −1i
kh0, 1, −1ik 2
1 1
N̂(0) = √ √ h0, 1, −1i × h0, 1, 1i = h2, 0, 0i = h1, 0, 0i
2 2 2
h0, 1, 0i · (h0, 1, −1i) 1
τ (0) = √ 2 =
2 2
21. Let r(t) = hcos t + t sin t, sin t − t cos t, t2 i. Find the speed, the tan-
gential and normal accelerations, the curvature and torsion, and the
unit tangent vector, normal, and binormal as functions of time t. Hint:
To simplify calculations, find the decomposition r(t) = v(t) − tw(t) +
t2ê3 where v, w, and ê3 are mutually orthogonal unit vectors such that
v0(t) = w(t), w0(t) = −v(t). Use the properties of the cross products
of mutually orthogonal unit vectors.
6. APPLICATIONS TO MECHANICS AND GEOMETRY 249

Solution: Let
r(t) = hcos t+t sin t, sin t−t cos t, t2i = hcos t, sin t, 0i−th− sin t, cos t, 0i+t2h0, 0, 1i
Then v(t) = hcos t, sin t, 0i and w(t) = h− sin t, cos t, 0i. It should be
evident that these satisfy the conditions in the hint. So,
r0(t) = v0(t) − w(t) − tw0 (t) + 2tê3 = w(t) − w(t) + tv(t) + 2tê3 = tv(t) + 2tê3
r00(t) = v(t) + tv0(t) + 2ê3 = v(t) + tw(t) + 2ê3
r000(t) = v0(t) + w(t) + tw0(t) = 2w(t) − tv(t)
Since v(t) and ê3 are orthogonal, the speed may be calculated using
the Pythagorean theorem, shown below
√ √
v(t) = |t| 1 + 4 = 5|t|
The dot product r0(t) · r00(t) is
r0(t) · r00(t) = t(v(t) + 2ê3 ) · (v(t) + tw(t) + 2ê3 ) = t[v(t) · v(t) + 4ê3 · ê3]
= t[kv(t)k2 + 4kê3 k2 ] = 5t
owing to the orthogonality of v(t), w(t), and ê3, and the fact that each
of those vectors is a unit vector.
The cross product r0(t) × r00(t) is
r0 (t) × r00(t) = t(v(t) + 2ê3 ) × (v(t) + tw(t) + 2ê3 )
= t[tv(t) × w(t) + 2v(t) × ê3 + 2ê3 × v(t) + 2tê3 × w(t)]
= t2[v(t) × w(t) + 2ê3 × w(t)]
= t2[ê3 − 2v(t)]
(the cross products can be easily calculated by imagining v(t) as ê1
and w(t) as ê2). The magnitude of the cross product is easily found
using the Pythagorean theorem, shown below:
√ √
kr0(t) × r00(t)k = t2 1 + 4 = 5t2
The cross product (r0(t) × r00(t)) × (r0 (t)) is
(r0 (t) × r00(t)) × (r0(t)) = t2[ê3 − 2v(t)] × (t)(v(t) + 2ê3 )
= t3[ê3 × v(t) − 4v(t) × ê3]
= t3[5ê3 × v(t)] = 5t3 w(t)
Finally, the triple product r000(t) · (r0 (t) × r00(t)) is
r000(t) · (r0(t) × r00(t)) = (2w(t) − tv(t)) · (t2[ê3 − 2v(t)])
= t2 [−2(−t)v(t) · v(t)] = 2t3
250 2. VECTOR FUNCTIONS

Then we have the tangential acceleration, normal acceleration, unit


tangent vector, unit binormal vector, unit normal vector, and torsion
are:

r0(t) · r00(t) 5t 5t
aT (t) = 0
=√ =
kr (t)k 5|t| |t|

r0(t) × r00(t) 5t2 t2
aN (t) = = √ = = |t|
kr0 (t)k 5|t| |t|
r0(t) 1 |t|
T̂(t) = 0 = √ [t(v(t) + 2ê3)] = √ (v(t) + 2ê3)
kr (t)k 5|t| t 5
0 00
r (t) × r (t) 1 1
B̂(t) = 0 00
= √ [t2(ê3 − 2v(t))] = √ (ê3 − 2v(t))
kr (t) × r (t)k 5t2 5
[r (t) × r00(t)] × r0(t)
0
5t3 w(t) |t|
N̂(t) = B̂(t) × T̂(t) = 0 = √ √ = w(t)
kr (t) × r00(t)kkr0(t)k 5t2 5|t| t
r000(t) · (r0(t) × r00(t)) 2t3 2t3 2
τ (t) = 0 00 2
= √ = 4
=
kr (t) × r (t)k ( 5t2)2 5t 5t

22. Let C be the curve of intersection of an ellipsoid x2 /a2 + y 2 /b2 +


z 2/c2 = 1 with the plane 2x − 2y + z = 0. Find the torsion and the
binormal B̂ along C.

Solution: Planar curves have a torsion of 0, so τ = 0. Let r(t)


parameterize the curve. Recall that the osculating plane at a point
with position vector r(t0) has a normal vector of r0(t0) × r00(t0). But
B̂(t0) = r0(t0) × r00(t0 )/kr0(t0 ) × r00(t0)k. So
h2, −2, 1i 1
B̂ = = h2, −2, 1i
kh2, −2, 1ik 3
CHAPTER 3

Differentiation of Multivariable Functions

1. Functions of Several Variables


1–13. Find and sketch the domain of each of the following functions:
1. f(x, y) = x/y.

Solution: Recall that the domain of a function is, in essence,


the set of n−tuples such that f is able to assign a value to each.
Therefore, a strategy for finding the domain is to take the set
of all n−tuples (i.e., Rn ) and remove from this those n−tuples
that f cannot assign a value to (so-called ”problem points”).
The easiest way to do this is to break f into the sum and
product of other simpler functions, find the domain of each of
these, and take the intersection of them. The intersection will
be the domain of f. For example, here the only issue is when
y = 0, because 1/y is not defined there. So the domain of f is
D = {(x, y) ∈ R2 | y 6= 0}
a sketch is

on the unit square, where blue represents points in the domain


of f and white those that are not.

2. f(x, y) = x/(x2 + y 2 ).

251
252 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

Solution: As above, we begin by finding the ”problem points”.


Here, the only issue is when x2 + y 2 = 0, because 1/(x2 + y 2)
is not defined there. So the domain of f is
D = {(x, y) ∈ R2 | x 6= 0 and y 6= 0}
a sketch is

on the unit square, where blue represents points in the domain


of f and white those that are not.

3. f(x, y) = x/(y 2 − 4x2 ).

Solution: As above, we begin by finding the ”problem points”.


Here, the only issue is when y 2 − 4x2 = 0 ⇔ y = ±2x, be-
cause 1/(y 2 − 4x2 ) is not defined there. So the domain of f
is
D = {(x, y) ∈ R2 |y 6= ±2x}
a sketch is

where blue represents points in the domain of f and white


those that are not.
1. FUNCTIONS OF SEVERAL VARIABLES 253

4. f(x, y) = ln(9 − x2 − (y/2)2 ).

Solution: As above, we begin by finding the ”problem points”.


Here, the only issue is when 9 − x2 − (y/2)2 ≤ 0 ⇔ 9 ≤
x2 + y 2/4 ⇔ 1 ≥ x/9 + y 2/36, because ln(9 − x2 − (y/2)2 ) is
not defined there. So the domain of f is

D = {(x, y) ∈ R2 |x/9 + y 2/36 < 1}

a sketch is

where blue represents points in the domain of f and white


those that are not (note: the border of the ellipse is not in-
cluded!).

p
5. f(x, y) = 1 − (x/2)2 − (y/3)2 .

Solution: As above, we begin by finding the ”problem points”.


is when 1 − (x/2)2 − (y/3)2 < 0 ⇔ 1 <
Here, the only issue p
x2/4+y 2 /9, because 1 − (x/2)2 − (y/3)2 is not defined there.
So the domain of f is

D = {(x, y) ∈ R2 |x/4 + y 2 /9 ≤ 1}

a sketch is
254 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

where blue represents points in the domain of f and white


those that are not (note: the border of the ellipse is included!).

p
6. f(x, y) = 4 − x2 − y 2 + 2x ln y.

Solution: As above, we begin by finding the ”problem points”.


Here, we have two simple
p functions to consider; the square
root and the logarithm. 4 − x2 − y 2 is not defined whenever
4 − x2 − y 2 < 0 ⇔ 4 < x2 + y 2 . ln y is not defined whenever
y ≤ 0. So the domain of f is

D = {(x, y) ∈ R2 |x2 + y 2 ≤ 4, y > 0}

a sketch is

where blue represents points in the domain of f and white


those that are not (note: the bottom border of the circle is
not included, but the rest is!).
1. FUNCTIONS OF SEVERAL VARIABLES 255

p
7. f(x, y) = 4 − x2 − y 2 + x ln y 2.

Solution: As above, we begin by finding the ”problem points”.


Here, we have two simple
p functions to consider; the square
root and the logarithm. 4 − x2 − y 2 is not defined whenever
4 − x2 − y 2 < 0 ⇔ 4 < x2 + y 2. ln y 2 is not defined whenever
y 2 ≤ 0 ⇔ y = 0. So the domain of f is

D = {(x, y) ∈ R2 |x2 + y 2 ≤ 4, y 6= 0}

a sketch is

where blue represents points in the domain of f and white


those that are not (note: entire border is included, except
when y = 0!).

p
8. f(x, y) = 4 − x2 − y 2 + ln(x2 + y 2 − 1).

Solution: As above, we begin by finding the ”problem points”.


Here, we have two simplep functions to consider; the square
root and the logarithm. 4 − x2 − y 2 is not defined whenever
4 − x2 − y 2 < 0 ⇔ 4 < x2 + y 2 . ln(x2 + y 2 − 1) is not defined
whenever x2 + y 2 − 1 ≤ 0 ⇔ x2 + y 2 ≤ 1. So the domain of
f is

D = {(x, y) ∈ R2 |x2 + y 2 ≤ 4, x2 + y 2 > 1}

a sketch is
256 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

where blue represents points in the domain of f and white


those that are not (note: entire border of the outer circle is
included, but not that of the inner circle!).

9. f(x, y, z) = x/(yz).

Solution: As above, we begin by finding the ”problem points”.


Here, we now work with R3 instead of R2 , so all our inequali-
ties will be in space. The only issue is when yz = 0 ⇔ y =
0, z = 0 because 1/(yz) is not defined there. So the domain of
f is
D = {(x, y, z) ∈ R3 | y 6= 0, z 6= 0}

a sketch is

where blue represents points in the domain of f and the white


grid planes are not included.
1. FUNCTIONS OF SEVERAL VARIABLES 257

10. f(x, y, z) = x/(x − y 2 − z 2 ).

Solution: As above, we begin by finding the ”problem points”.


The only issue is when x − y 2 − z 2 = 0 ⇔ x = y 2 + z 2 because
1/(x − y 2 − z 2) is not defined there. So the domain of f is

D = {(x, y, z) ∈ R3 | x 6= y 2 + z 2 }

a sketch is

where blue represents points in the domain of f and the white


grid paraboloid is not included.

11. f(x, y, z) = ln(1 − z + x2 + y 2 ).

Solution: As above, we begin by finding the ”problem points”.


The only issue is when 1 − z + x2 + y 2 ≤ 0 ⇔ x2 + y 2 ≤ z − 1
because ln(1 − z + x2 + y 2) is not defined there. So the domain
of f is

D = {(x, y, z) ∈ R3 | z − 1 < x2 + y 2}

a sketch is
258 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

where blue represents points in the domain of f. The region


above and including the paraboloid is not included.

p
12. f(x, y, z) = x2 − y 2 − z 2 + ln(1 − x2 − y 2 − z 2 ).

Solution: As above, we begin by finding the ”problem points”.


There are two simple functions, the square root and the log-
arithm. For the square root, the problem points are when
x2 − y 2 − z 2 < 0 ⇔ x2 < y 2 + z 2 . For the logarithm, the
problem points are when 1−x2 −y 2 −z 2 ≤ 0 ⇔ 1 ≤ x2 +y 2 +z 2 .
So the domain of f is

D = {(x, y, z) ∈ R3 | x2 + y 2 + z 2 < 1, x2 ≥ y 2 + z 2 }

a sketch is

The domain is the region between the double cone and the
unit sphere, not including the boundary of the unit sphere.
1. FUNCTIONS OF SEVERAL VARIABLES 259

13. f(t, r) = (t2 − krk2 )−1 , r = hx1 , x2 , ..., xni.

Solution: As above, we begin by finding the ”problem points”.


This is when t = ±krk. So the domain of f is
D = {(t, x1, ..., xn) ∈ Rn+1 | t 6= ±krk}
There is no good way to sketch this for general n.

14–19. For each of the following functions, sketch a contour map and
use it to sketch the graph:
14. f(x, y) = x2 + 4y 2 ;

Solution: For each of these problems, we let f(x, y) = k,


where k is in the range of f. This is simply a level set of
f, and a contour plot is a collection of level sets. To sketch
the graph, we interpret k as z; thus, take the contour plot,
and ”raise” or ”lower” a each contour by its k. Here, we have
x2 + 4y 2 = k, where k is a nonnegative real number (since
the sum of two positive numbers is always positive). This is
clearly an ellipse, given by x2/k + y 2/(k/4) = 1 for nonzero k;
when k = 0, the contour is simply the origin. Thus we have

the contours closer to blue correspond to lower k values and


those in yellow correspond to greater k values. A sketch of the
graph is produced to the right. This is an elliptic paraboloid.

15. f(x, y) = xy;

Solution: As above, we begin by forming the contour plot


using multiple level sets. Here, we have xy = k, where k
260 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

is a real number. This is a hyperbola for nonzero k given by


y = k/x. For positive k, the level sets are hyperbolas in the 1st
and 3rd quadrant; for negative k, in the 2nd and 4th. When
k = 0, the contour is xy = 0 ⇔ x = 0, y = 0. Thus we have

the contours closer to blue correspond to lower k values (teal


is k = 0) and those in yellow correspond to greater k values.
A sketch of the graph is produced to the right. Note that this
is a saddle rotated by 45 degrees, aligned along y = x.

16. f(x, y) = x2 − y 2 ;

Solution: As above, we begin by forming the contour plot


using multiple level sets. Here, we have x2 − y 2 = k, where k
is a real
√ number. This is a hyperbola for nonzero k given by
y = ± x2 − k. For positive k, the level sets are hyperbolas in
the opening along the x-axis; for negative k, along the y-axis.
When k = 0, the contour is x2 − y 2 = 0 ⇔ y = ±x. Thus we
have

the contours closer to blue correspond to lower k values (teal


1. FUNCTIONS OF SEVERAL VARIABLES 261

is k = 0) and those in yellow correspond to greater k values.


A sketch of the graph is produced to the right. Note that this
is a saddle in its canonical form.

p
17. f(x, y) = x2 + 9y 2 ;

Solution: As above, we begin by forming p the contour plot


using multiple level sets. Here, we have x2 + 9y 2 = k,
where k is a nonnegative real number. This is an ellipse
2 2 2 2
given
p by x /k + y /(k /9) = 1. When k = 0, the contour
is x2 + 9y 2 = 0 which consists of only the origin. Thus we
have

the contours closer to blue correspond to lower k values and


those in yellow correspond to greater k values. A sketch of the
graph is produced to the right. This is a cone.

18. f(x, y) = sin x;

Solution: As above, we begin by forming the contour plot


using multiple level sets. Here, we have sin x = k, where k sat-
isfies −1 ≤ k ≤ 1. The contour is a collection of parallel lines
given by x = arcsin k, repeating as necessary. Note that there
is no dependent on y, so the surface will look the same for any
y; you can imagine the surface as being formed by taking the
planar curve sin x and placing a copy for each y. Thus we have
262 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

the contours closer to blue correspond to lower k values and


those in yellow correspond to greater k values. A sketch of the
graph is produced to the right.

19. f(x, y) = y 2 + (1 − cos x).

Solution: As above, we begin by forming the contour plot


using multiple level sets. Here, we have y 2 + (1 − cos x) = k,
where k is a real number. Thus we have

the contours closer to blue correspond to lower k values and


those in yellow correspond to greater k values. A sketch of the
graph is produced to the right.
20–25. Describe and sketch the level sets of each of the following
functions:
20. f(x, y, z) = x + 2y + 3z;

Solution: The level sets for each of these problem will be


level surfaces. Thus we must describe and sketch the surfaces.
To do this, we set f(x, y, z) = k for k in the range of f. Here we
1. FUNCTIONS OF SEVERAL VARIABLES 263

have x + 2y + 3z = k for real k. These are parallel planes with


normal vector n = h1, 2, 3i. The level sets for k = −10, 0, 10
are produced below:

21. f(x, y, z) = x2 + 4y 2 + 9z 2 ;

Solution: As above, we set f(x, y, z) = k for k in the


range of f. Here we have x2 + 4y 2 + 9z 2 = k 2 for nonneg-
ative real k. Using k 2 rather than k is merely an aesthetic
choice. When k = 0, we have x2 + 4y 2 + 9z 2 = 0, the only
solution of which is the origin (since the sum of three nonneg-
ative numbers, not all zero, is nonzero). For positive k, we
have (x/k)2 + (y/(k/2))2 + (z/(k/3))2 = 1. This is an ellipsoid
with axes along the x, y, and z axis of length 2k, k, and 2/3k.
The level sets for k = 0, 1, 2 are produced below:

22. f(x, y, z) = z + x2 + y 2 ;

Solution: As above, we set f(x, y, z) = k for k in the range


of f. Here we have z + x2 + y 2 = k for real k. Note that this
can be rearranged as (z − k) = −(x2 + y 2 ); clearly, an elliptic
264 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

paraboloid shifted concave down shifted k units in the positive


z direction. The level sets for k = 0, 1, 2 are produced below:

23. f(x, y, z) = x2 + y 2 − z 2 ;

Solution: As above, we set f(x, y, z) = k for k in the range


of f. Here we have x2 + y 2 − z 2 = k for real k. In the case
when k = 0, we have z 2 = x2 + y 2, a cone. When k is nonzero,
we have x2 /k + y 2/k − z 2 /k = 1. The surface depends on the
sign of k; if k is positive, this is a hyperboloid of one sheets
otherwise it is a hyperholoid of two sheets. The level sets for
k = −1/2, 0, 1/2 (from inner to outer) are produced below:

24. f(x, y, z) = ln(x2 + y 2 − z 2 );

Solution: As above, we set f(x, y, z) = k for k in the range


of f. Here we have ln(x2 + y 2 − z 2) = k for real k. However,
we can manipulate this to x2 + y 2 − z 2 = ek = K for some pos-
itive K (since ek is never negative). Thus we are in a special
case of Exercise 23 where we only have hyperboloids of one
1. FUNCTIONS OF SEVERAL VARIABLES 265

sheet. The level sets for K = e−2 , (e2 + e−2 )/2, e2 (from inner
to outer) are produced below:

25. f(x, y, z) = ln(z 2 − x2 − y 2 ).

Solution: As above, we set f(x, y, z) = k for k in the range


of f. Here we have ln(z 2 − x2 − y 2 ) = k for real k. However,
we can manipulate this to z 2 − x2 − y 2 = ek = K for some pos-
itive K (since ek is never negative). Thus we are in a special
case of Exercise 23 where we only have hyperboloids of two
sheets. The level sets for K = e−2 , (e2 + e−2 )/2, e2 (from outer
to inner) are produced below:

26–32. Sketch the level sets of each of the following functions. Here
min(a, b) and max(a, b) denote the smallest number and the largest
number of a and b, respectively, and min(a, a) = max(a, a) = a.
26. f(x, y) = |x| + y;

Solution: Here, the level sets are given by |x| + y = k ⇔


y = k − |x| for real k. This is simply a vertical translation of
266 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

y = −|x| by k units. The level sets are shown as a contour


plot below:

27. f(x, y) = |x| + |y| − |x + y|;

Solution: Here, the level sets are given by |x|+|y|−|x+y| =


k for nonnegative k. k must be nonnegative because, due to
the triangle inequality, we have
|x + y| ≤ |x| + |y| ⇔ −|x + y| ≥ −|x| − |y| ⇔ |x| + |y| − |x + y| ≥ 0
Suppose k = 0. Then |x| + |y| − |x + y| = 0. Let y = 0, then
|x| − |x| = 0, which is true for all x. A similar argument holds
for x = 0. Thus the lines x = 0, y = 0 are the level set for
k = 0. Next, let k > 0. It follows that x = k/2, y = −k/2 is a
solution, since
|k/2| + | − k/2| − |k/2 − k/2| = k/2 + k/2 − 0 = k
The same is true for x = −k/2, y = k/2. Suppose x = k/2.
Then
|k/2| + |y| − |y + k/2| = k ⇔ |y| − |y + k/2| = k/2
Suppose now that y > −k/2. Then y + k/2 > 0, and so
|y+k/2| = y+k/2. However, y+k/2 > y, so |y|−|y+k/2| < 0.
But k is positive, so this cannot be. Clearly, the issue is with
the equality |y + k/2| = y + k/2. Suppose now that y ≤ −k/2.
Then |y + k/2| = −(y + k/2). We still have that y + k/2 > y,
and so −y > −(y + k/2). Since −y = |y| (both quantities are
positive, y is negative), we have |y| − |y + k/2| > 0. In this
case, we actually have direct equality. Thus the line x = k/2
for y ≤ −k/2 is part of the contour. Additionally, we have
y = −k/2 for x ≥ k/2, x = −k/2 for y ≥ k/2, and y = k/2 for
x ≤ −k/2. The level sets are shown as a contour plot below:
1. FUNCTIONS OF SEVERAL VARIABLES 267

28. f(x, y) = min(x, y);

Solution: Here, the level sets are given by min(x, y) = k


for real k. First, recall that min(x, x) = x, so clearly (k, k) is
a solution. Fix x = k. Then for any y ≥ k, min(x, y) = x = k.
Next, fix y = k. Then for any x ≥ k, min(x, y) = y = k. So
the level set consists of the two lines x = k for y ≥ k and y = k
for x ≥ k. The level sets are shown as a contour plot below:

29. f(x, y) = max(|x|, |y|);

Solution: Here, the level sets are given by max(|x|, |y|) = k


for real k. First, recall that max(x, x) = x, so clearly (k, k)
and (−k, −k) are solutions. Fix x = ±k. Then for any
−k ≤ y ≤ k, max(|x|, |y|) = |x| = k. Next, fix y = ±k.
Then for any −k ≤ x ≤ k, max(|x|, |y|) = y = k. So the level
set consists of the four lines x = ± for y ≥ k and y = ±k for
x ≥ k. Incidentally, these make squares of side length 2k. The
level sets are shown as a contour plot below:
268 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

30. f(x, y) = sign(sin x sin y); here sign(a) is the sign function, it
has the values 1 and 1 for positive and negative a, respectively.

Solution: Here, the level sets are given by sign(sin x sin y) =


k for k = −1, 0, 1. Recall that sin(x) = 0 for x = nπ where
n is an integer. Thus sign(sin x sin y) partitions R2 with grid
lines at x = nπ and y = nπ. Focus on the plot with 0 ≤
x ≤ 2π and 0 ≤ y ≤ 2π. This region is subdivided into four
squares (by the lines x = π and y = π). In the lower left
(0 < x < π and 0 < y < π), sin x and sin y are both positive,
so sign(sin x sin y) = 1. Next, in the lower right (π < x < 2π
and 0 < y < π) square, sin x is negative whereas sin y is pos-
itive, so sign(sin x sin y) = −1. In the upper left (0 < x < π
and π < y < 2π) square, sin x is positive whereas sin y is neg-
ative, so sign(sin x sin y) = −1. Finally, in the upper right
(π < x < 2π and π < y < 2π) square, sin y is negative and
sin y is negative, so sign(sin x sin y) = 1. Due to the periodic
nature of sin, this plot may be tessellated to cover all of R2 .
The level sets are shown as a contour plot below:
1. FUNCTIONS OF SEVERAL VARIABLES 269

Unfortunately, MATLAB would not color the space appropri-


ately. All white spaces should either be shaded yellow/blue,
depending on the border of the square. Teal corresponds to 0.

31. f(x, y) = (x + y)2 + z 2 ;

Solution: Here, the level sets are given by (x + y)2 + z 2 = k


for nonnegative k. Let x = x0 cos(φ) − y 0 sin(φ) and y =
y 0 cos(φ) + x0 sin(φ) [Recall Chapter 1]. Then
√ π √ π
x+y = x0(cos(φ)+sin(φ))+y 0(cos(φ)−sin(φ)) = 2x0 cos(φ− )− 2y 0 sin(φ− ),
4 4
so we have that
π π π π
(x+y)2 = 2(x02 cos(φ− )2 −2x0y 0 cos(φ− ) sin(φ− )+y 02 sin(φ− )2)
4 4 4 4
We wish to remove the mixed term, so let φ = π/4. Then the
level sets are given by
2x02 + z 2 = k
which are evidently elliptic cylinders that have been rotated
by π/4 (aligned along y = −x, because there are no solutions
there!). The level sets are shown with the surfaces (k = 1, 4, 9,
from inner to outer) below:

32. f(x, y) = arctan(2ay/(x2 + y 2 − a2 )), a > 0.

Solution: Here, the level sets are given by arctan(2ay/(x2 +


y 2 − a2)) = k for real k satisfying −π/2 < k < π/2 (the
range follows from the natural restriction of arctangent). This
can be rearranged to give 2ay = tan(k)(x2 + y 2 − a2). After
270 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

completing the square, one yields


x2 + (y − a/ tan(k))2 = a2 + (a/ tan(k))2
p
Evidently these are circles with radius a2 + (a/ tan(k))2 and
center (0, a/ tan(k)). The level sets are shown with the sur-
faces below:

33–36. Explain how the graph z = g(x, y) can be obtained from the
graph of f(x, y) if
33. g(x, y) = k + f(x, y), where k is a constant;

Solution: Recall that z − k for a constant k simply de-


scribes a shift by k units in the positive z direction. So
g(x, y) = k + f(x, y) is a translation of f(x, y) by k units.

34. g(x, y) = mf(x, y), where m is a nonzero constant;

Solution: First, suppose m > 0. In this case g(x, y) =


mf(x, y) is obtained where f(x, y) is shrunk (0 < m < 1) or
stretched (1 < m). Next, suppose m < 0. Then g(x, y) =
mf(x, y) is obtained by first reflecting f over the xy plane,
then stretching/shrinking f.

35. g(x, y) = f(x − a, y − b), where a and b are constants;

Solution: As with Exercise 33, g(x, y) = f(x − a, y − b)


is obtained by translating f by a units along the x axis and b
units along the y axis.

36. g(x, y) = f(px, qy), where p and q are nonzero constants.


1. FUNCTIONS OF SEVERAL VARIABLES 271

Solution: As with Exercise 34, the same shrinking/stretching


effect occurs along each axis depending on the sign of p, q and
whether |p|, |q| > 1.

37–39. Given a function f(x, y), sketch the graphs of the function
g(x, y) defined in Exercises 33-36. Analyze carefully various cases for
values of the constants (for example, m > 0, m < 0, p > 1, 0 < p < 1,
and p = −1, etc.)

37. f(x, y) = x2 + y 2 ;

Solution: Let f(x, y) = x2 + y 2. Note that f(y, x) = f(x, y),


so there is no reason to investigate changes in q. Further ob-
serve that f is even with respect to both x and y, so f(px, qy) =
f(−px, −qy), and thus these cases will not be investigated.
The plot to the left includes z = f(x, y) and z = ±4f(x, y).
The plot to the right includes z = f(2x, y), z = f(x, y), and
z = f(x/2, y) (from inner to outer).

38. f(x, y) = xy;

Solution: Let f(x, y) = xy. Note that f(y, x) = f(x, y),


so there is no reason to investigate changes in q. Further
observe that f(px, y) = pf(x, y), thus these cases will not
be investigated. The plot to the left includes z = f(x, y)
and z = 2f(x, y) = f(2x, y). The plot to the right includes
272 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

z = f(x, y) and z = −2f(x, y) = f(−2x, y)

39. f(x, y) = (a2 − x2 − y 2 )2 .

Solution: Let f(x, y) = (a2 − x2 − y 2)2 . Note that f(y, x) =


f(x, y), so there is no reason to investigate changes in q. Fur-
ther observe that f is even with respect to both x and y, so
f(px, qy) = f(−px, −qy), and thus these cases will not be
investigated. The plot to the left includes z = f(x, y) and
z = ±2f(x, y). The plot to the right (note: which is scaled dif-
ferently) includes z = f(x, y), z = f(2x, y), and z = f(x/2, y)

p
40. Find f(u) if f(x/y) = x2 + y 2/x, x > 0.

Solution: The goal is to express all instances of x and y in f(x/y) in


terms of x/y or y/x. Then we may substitute u = x/y and 1/u = y/x.
This is just algebra.
p r r
x x2 + y 2 y2 y
f( ) = = 1 + 2 = 1 + ( )2
y x x x
p √
so with u = x/y, f(u) = 1 + 1/u2 = u2 + 1/u.
1. FUNCTIONS OF SEVERAL VARIABLES 273

41. Find f(x, y) if f(x + y, y/x) = y 2 − x2

Solution: I will solve this by making the substitutions u = x + y


and v = y/x. Note that we can solve for x, y as follows. We have
that y = vx. Substitution of this into the former equation gives
u = x + vx = (1 + v)x, so x = u/(1 + v). Substituting this back
gives y = v(u/(1 + v)) = uv/(1 + v). Then we have
uv 2 u 2 u2 2 u2 (v − 1)
f(u, v) = ( ) −( ) = (v − 1) =
1+v 1+v (1 + v)2 1+v
A simple cosmetic change provides f(x, y) as f(x, y) = x2 (y−1)/(y+1).

Alternative Solution: One can also simply perform the follow-


ing algebraic manipulations
y−x y/x − 1
f(x + y, y/x) = (y + x)(y − x) = (y + x)2 = (x + y)2 ,
y+x y/x + 1
but I find this to be far less intuitive.
√ √
42. Let z = y + f( x − 1). Find the functions z and f if z = x when
y = 1.

Solution: Suppose z = x when y = 1. Let y = 1. Then


√ √
x = 1 + f( x − 1) ⇔ f( x − 1) = x − 1
and f(x) = (x + 1)2 − 1. Direct substitution verifies this:
√ √ √ 2
f( x − 1) = ( x − 1 + 1)2 − 1 = x − 1 = x − 1
√ √ √
Also, z = y + f( x − 1) = y + x − 1.

43. Graph the function F (t) = f(cos t, sin t) where f(x, y) = 1 if


y ≥ x and f(x, y) = 0 if y < x. Give a geometrical interpretation of
the graph of F as an intersection of two surfaces.

Solution: First let us determine what the surface z = f(x, y) looks


like. Consider the plane y = x. Everything ”above” this plane (with
y ≥ x) will be at z = 1, and everything below will be at z = 0.
Note that cos t = sin t for t = π/4 + nπ (n integer). On the interval
[0, π/4), cos t > sin t ⇔ x > y, and so F (t) = 0. On the interval
[π/4, 5π/4], sin t ≥ cos t ⇔ y ≥ x and F (t) = 1. On the interval
(5π/4, 2π], cos t > sin t ⇔ x > y, and so F (t) = 0. Thus F (t) is
1 if t is in the interval [π/4 + 2nπ, 5π/4 + 2nπ] and 0 if t is in the
274 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

interval (5π/4 + 2nπ, 9π/4 + 2nπ). F is the intersection of the surface


z = f(x, y) and the cylinder x2 + y 2 = 1.

44–47. Let f(u) be a continuous function for all real u. Investigate


the relation between the shape of the graph of f and the shape of the
following surfaces:
44. z = f(y − ax);

Solution: Let f be a continuous function. Fix a real a. Then


y − ax = d represents a particular plane for every d. Since d is
a constant, if y − ax = d, then f(y − ax) = f(d), a constant.
Thus f(y − ax) attains a constant value throughout the plane
y − ax = d. The effect of this is the following: Consider any
plane perpendicular to y = ax. Then a ”copy” of f(u) will be
in this plane. Since this is true for every perpendicular plane,
a cylinder is formed. The effect of a can be found by referring
to Exercise 36.
p
45. z = f( x2 + y 2 );
p
Solution: Let f be a continuous
p function. Suppose x2 + y 2 =
r for a constant r. Then f( x2 + y 2) = f(r), a constant. But
we also have x2 + y 2 = r2 , a circle. So f is constant on circular
regions (not a disc!). Imagine
p taking a circle x2 + y 2 = r2
in the xy plane. Then f( x2 + y 2 is produced by lifting this
circle f(r) units. The effect is that z is f rotated around the z
axis. However, it is only the surface with z ≥ f(0), since both
x2 + y 2 are positive, r can never be negative.
p
46. z = f(− x2 + y 2 );

Solution: Let f be a continuous function. This is in some


sense the opposite of Exercise 45. The reason being that only
z ≤ f(0) are allowed now. More precisely, it is f(−u) rotated
about the z axis for u ≥ 0. So, for functions that are neither
even nor odd (e.g., eu ), the functions in Exercises 45 and 46
will be completely different!

47. z = f(x/y);

Solution: Let f be a continuous function. Let x(t) = r cos t


1. FUNCTIONS OF SEVERAL VARIABLES 275

for 0 ≤ t < 2π and y(t) = r sin t for 0 ≤ t < 2π. Then


hx(t), y(t)i defines a circle of radius r in the xy plane. Also note
that f(x(t)/y(t)) = f(cot(t)). So for every circle x2 + y 2 = r2 ,
f looks identical.
276 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

2. Limits and Continuity


1–5. Use Definition 17.3 of the limit to verify each of the following
limits (i.e., given  > 0, find a neighborhood of the limit point with the
properties specified in the definition):
x3 − 4y 2 x + 5y 3
1. lim = 0;
r→0 x2 + y 2
Solution: Let  > 0. The limit point here is 0. The dis-
tance between r and the limit point is
p
R = kr − r0k = krk = x2 + y 2
From this equality, we immediately receive the following two
inequalities:
|x| ≤ R, |y| ≤ R
where equality is reached when one of x or y (or both) is 0.
We will also need the triangle inequality,
|x1 + x2 + ... + xn | ≤ |x1| + |x2| + ... + |xn |

To verify that the limit exists, we need to show that for every
 > 0, if kr − r0 k < δ [equivalent to R < δ!] then |f(r) − 0| =
|f(r)| < . Thus we need to evaluate the latter quantity in
some way, as follows.
x3 − 4y 2 x + 5y 3 |x|3 + 4y 2|x| + 5|y|3 R3 + 4R3 + 5R3
| | ≤ ≤ = 10R
x2 + y 2 R2 R2
Note that x2 + y 2 = R2 . Take close note, in particular, to the
sign change of the 4y 2x quantity due to the triangle inequality.
To get from the second to the third quantity, we employed the
two derived inequalities |x|, |y| ≤ R.

Now, we require that |f(r)| < . So if 10R < , then |f(r)| < .
Given  > 0, we may choose 10δ =  ⇔ δ = /10 to meet the
requirements R < δ and 10R < . However, we may actually
choose any δ ≤ /10; all this does is restrict our choice of R
further.

x3 − 4y 2 x + 5y 3
2. lim = 0;
r→0 3x2 + 4y 2
Solution: Let  > 0. The limit point here is 0. As above,
2. LIMITS AND CONTINUITY 277

we need to evaluate |f(r)| as follows.

x3 − 4y 2 x + 5y 3 |x|3 + 4y 2|x| + 5|y|3 R3 + 4R3 + 5R3 10R


| 2 2
| ≤ 2 2
≤ 2
=
3x + 4y 3x + 3y 3R 3

Note that 3x2 + 4y 2 ≥ 3x2 + 3y 2 = 3R2 ⇔ 1/(3R2 ) ≤


1/(3x2 + 4y 2 ). You want all your inequalities to be ≤, other-
wise you might run into trouble.

Now, we require that |f(r)| < . So if 10R/3 < , then |f(r)| <
. Given  > 0, we may choose 10δ/3 =  ⇔ δ = 3/10 to
meet the requirements R < δ and 10R/3 < . However, we
may actually choose any δ ≤ 3/10; all this does is restrict our
choice of R further.

x3 − 4y 4 + 5y 3x2
3. lim = 0;
r→0 3x2 + 4y 2
Solution: Let  > 0. The limit point here is 0. As above,
we need to evaluate |f(r)| as follows.

x3 − 4y 4 + 5y 3x2 |x|3 + 4y 4 + 5|y|3x2 R3 + 4R4 + 5R5 R3 + 4R3 + 5R3 10R


| 2 2
| ≤ 2 2
≤ 2
≤ 2
=
3x + 4y 3x + 4y 3R 3R 3

for sufficiently small R. This is because for R < 1, we have


R5 , R4 < R3 . So what about when R > 1? We will get to that
later.

Now, we require that |f(r)| < . So if 10R/3 < , then |f(r)| <
. Given  > 0, we may choose 10δ/3 =  ⇔ δ = 3/10 to
meet the requirements R < δ and 10R/3 < . However, we
may actually choose any δ ≤ 3/10. Remember about the case
when R > 1?. We can avoid this by requiring that δ is the
smallest number between 1 and 3/10. This is because if 3/10
is greater than 1, then we will still have that R < δ, and so R
may possibly be greater than 1. However, by restricting δ in
these cases, we can ensure that R is always less than 1. So,
δ ≤ δ1, where δ1 is the smallest number between 1 and 3/10.

x3 − 4y 2 x + 5y 3
4. lim = 0;
r→0 3x2 + 4y 2 + y 4

Solution: Let  > 0. The limit point here is 0. As above,


278 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

we need to evaluate |f(r)| as follows.


x3 − 4y 2 x + 5y 3 10R3 10R/3 10R/3 10R
| 2 2 4
| ≤ 2 4
≤ 2
≤ =
3x + 4y + y 3R + R (1 + R /3) 1 3
Note that 1 + R2 /3 ≥ 1 ⇔ 1/(1 + R2 /3) ≤ 1 for all R, so we
do not need to do anything like in Exercise 3.

Now, we require that |f(r)| < . So if 10R/3 < , then |f(r)| <
. Given  > 0, we may choose 10δ/3 =  ⇔ δ = 3/10 to
meet the requirements R < δ and 10R/3 < . However, we
may actually choose any δ ≤ 3/10.

3x3 + 4y 4 − 5y 5
5. lim = 0.
r→0 x2 + y 2 + z 2
Solution: Let  > 0. The limit point here is 0. As above,
we need to evaluate |f(r)| as follows. Note, we have z in this
limit; treat it the same as any other variable!
3x3 + 4y 4 − 5y 5 3R3 + 4R4 + 5R5 3R3 + 4R3 + 5R3 12R3
| | ≤ ≤ ≤ = 12R
x2 + y 2 + z 2 R2 R2 R2
for sufficiently small R (note: we’ll have to do something like
in Exercise 3 again!).

Now, we require that |f(r)| < . So if 12R < , then |f(r)| < .
Given  > 0, we may choose 12δ =  ⇔ δ = /12 to meet the
requirements R < δ and 12R < . However, we may actually
choose any δ ≤ δ1 where δ1 is the smallest number between
/12 and 1.

6–8. Use the squeeze principle to prove the following limits and find a
neighborhood of the limit point in which the deviation of the function
from the limit value does not exceed a small given number  (Hint:
| sin u| ≤ |u|):

6. lim y sin(x/ y) = 0
r→0

Solution: Recall the squeeze principle is if there exists a


function h of one variable such that
|f(r) − c| ≤ h(R) → 0 as kr − r0 k = R → 0+
then the limit as r → r0 of f(r) is c. Here, we have
p
R = kr − 0k = x2 + y 2
2. LIMITS AND CONTINUITY 279

Using the hint,


√ √ √ √
|y sin(x/ y) − 0| ≤ |y|| sin(x/ y)| ≤ |y||x/ y| ≤ |x| y ≤ R3/2
Evidently, R3/2 goes to 0 as R goes to 0. Hence the limit
exists. To find the neighborhood, simply use the second to
last inequality as follows:
√ √ √ √ √
|y sin(x/ y)| ≤ |x| y ⇒ −|x| y ≤ y sin(x/ y) ≤ |x| y
A delta such that δ ≤ 2/3 may be chosen.

7. lim[1 − cos(y/x)]x2 = 0
r→0

Solution: As above, we have


p
R = kr − 0k = x2 + y 2
The hint cannot exactly be used here, so we must develop our
own. Recall the Taylor series for cos x about x = 0. Then the
Taylor expansion of 1 − cos x about x = 0 is given by
∞ ∞ ∞
X (−1)n x2n X (−1)n x2n X (−1)n x2n x2 x4
1−cos x = 1− = 1−(1+ )=− = − +...
n=0
(2n)! n=1
(2n)! n=1
(2n)! 2 4!

and so |1 − cos x| ≤ x2 /2. Using this,


|[1 − cos(y/x)]x2 − 0| ≤ |1 − cos(y/x)|x2 ≤ |y 2/(2x2 )|x2 ≤ y 2/2 ≤ R2 /2
Evidently, R2 /2 goes to 0 as R goes to 0. Hence the limit
exists. To find the neighborhood, simply use the second to
last inequality as follows:
|[1 − cos(y/x)]x2| ≤ y 2/2 ⇒ −y 2/2 ≤ [1 − cos(y/x)]x2 ≤ y 2/2

A delta such that δ ≤ 2 may be chosen.

cos(xy) sin(4x y)
8. lim √ =0
r→0 y
Solution: As above, we have
p
R = kr − 0k = x2 + y 2
The hint can be used, but what about cos(xy)? Recall that
| cos(x)| ≤ 1. Using this,
√ √ √
cos(xy) sin(4x y) | cos(xy)|| sin(4x y)| 4|x| y
| √ −0| ≤ √ ≤ √ ≤ 4|x| ≤ 4R
y y y
280 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

Evidently, 4R goes to 0 as R goes to 0. Hence the limit ex-


ists. To find the neighborhood, simply use the second to last
inequality as follows:
√ √
cos(xy) sin(4x y) cos(xy) sin(4x y)
| √ | ≤ 4|x| ⇒ −4x ≤ √ ≤ 4x
y y
A delta such that δ ≤ /4 may be chosen.

9. Suppose that limr→r0 = 2 and r0 is in the domain of f. If nothing


else is known about the function, what can be said about the value
f(r0 )? If, in addition, f is known to be continuous at r0 , what can be
said about the value f(r0 )?

Solution: If nothing else is know, the value f(r0) is unknown. This


in part why the definition of continuity exists. Recall that a function
f is continuous at r0 in its domain if
lim = f(r0 )
r→r0

thus if (and only if) f is continuous, we may say that f(r0 ) = 2.

10–19. Find the points of discontinuity of each of the following func-


tions:
10. f(x, y) = xy/(x2 + y 2 );

Solution: f(x, y) is the quotient of two continuous functions,


so it is continuous everywhere the denominator is nonzero. The
denominator is zero only at the origin. Therefore f is discon-
tinuous at the origin. It is of interest to investigate
p the limit
at the origin. As a sketch, notice that if R = x2 + y 2 ,
xy R2
| | ≤ =1
x2 + y 2 R2
and hence by the squeeze principle the limit does not exist
(or, at least, the limit is not zero). However, the above rea-
soning for now is sufficient to show that the limit does not
exist. More extensive techniques will be developed later to
conclusively prove this. So, f is not continuously extendable
to the origin.

11. f(x, y, z) = xyz/(x2 + y 2 + z 2 );


2. LIMITS AND CONTINUITY 281

Solution: f(x, y) is the quotient of two continuous functions,


so it is continuous everywhere the denominator is nonzero. The
denominator is zero only at the origin. Therefore f is discon-
tinuous at the origin. It is of interest to investigate the limit
at the origin. One may be inclined to conclude, based on Ex-
ercise 10, that the limit immediately
p does not exist. However,
2
this is incorrect. If R = x + y , 2

xyz R3
| | ≤ =R
x2 + y 2 + z 2 R2
and hence by the squeeze principle the limit does exist and is
equal to 0. Hence, f is continuously extendable to the origin.

12. f(x, y) = sin( xy);

Solution: f(x, y) is the composition of two continuous func-


tions and thus is continuous everywhere on its domain.

13. f(x, y) = cos( xyz)/(x2 y 2 + 1);

Solution: Observe first that cos( xyz) is the composition of
two continuous functions and hence is continuous. Then f is
the ratio of two continuous functions and is discontinuous only
where the denominator vanishes. But, x2 y 2 + 1 is never 0 (it is
1 plus something nonnegative and hence always greater than
or equal to 1). So f is continuous everywhere in its domain.

14. f(x, y) = (x2 + y 2 ) ln(x2 + y 2 );

Solution: f is the product of two functions and is hence con-


tinuous wherever they are continuous. x2 + y 2 is a polynomial
and hence continuous everywhere. ln(x2 + y 2) is continuous
everywhere such that x2 + y 2 > 0, so everywhere but the ori-
gin. Thus f is discontinuous at the origin.

15. f(x, y) = 1 if either x or y is rational and f(x, y) = 0 elsewhere;

Solution: f is discontinuous everywhere. For every disc


around a point r0 , the disc will contain rational and irrational
points.
282 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

16. f(x, y) = (x2 − y 2 )/(x − y) if x 6= y and f(x, x) = 2x;

Solution: We may rewrite f as f(x, y) = (x + y) if x 6= y


and f(x, x) = 2x. Here, it is obvious that the limit as (x, y)
approaches (x, x) of f(x, y) is x + x = 2x, and hence f is con-
tinuous.

17. f(x, y) = (x2 − y 2 )/(x − y) if x 6= y and f(x, x) = x;

Solution: The continuous extension of a function is unique


and described in Exercise 16. Hence f is discontinuous ev-
erywhere along the line y = x except at the origin (because
2(0) = (0), and the extensions agree).

18. f(x, y, z) = 1/[sin(x) sin(z − y)];

Solution: f is the quotient of two functions and hence is


discontinuous everywhere the denominator is 0. Evidently,
this occurs when x = nπ for integers n and z − y = kπ for
integers k.

19. f(x, y) = sin(1/(xy)).

Solution: f is the composition of two functions. Note that


sin u is continuous everywhere, so we need only find when
1/(xy) is discontinuous. Evidently, this is whenever x = 0
or y = 0.

20–22. Each of the following functions has the value at the origin
f(0, 0) = c. Determine whether there is a particular value of c at
which the function is continuous at the origin if for (x, y) 6= (0, 0):
20. f(x, y) = sin(1/(x2 + y 2 ));
p
Solution: Let R = x2 + y 2. Then

1 1
|f(x, y) − c| ≤ + |c| ≤ 2 + |c|
x2 +y 2 R

which does not exist as R goes to 0, hence the limit does not
exist.
2. LIMITS AND CONTINUITY 283

21. f(x, y) = (x2 + y 2 )ν sin(1/(x2 + y 2 )), ν > 0;


p
Solution: Let R = x2 + y 2. Then

R2ν
|f(x, y) − 0| ≤ = R2ν−2
R2
which goes to 0 as R goes to 0 since ν > 0. So we can contin-
uously extend f to (0, 0) by letting c = 0.

22. f(x, y) = xn y m sin(1/(x2 + y 2 )), n ≥ 0, m ≥ 0, and n + m > 0.


p
Solution: Let R = x2 + y 2. Then

R2(n+m)
|f(x, y) − 0| ≤ 2
= R2n+2m−2
R
which goes to 0 as R goes to 0 since ν > 0. So we can contin-
uously extend f to (0, 0) by letting c = 0.

23–27. Use the properties of continuous functions to find the following


limits
(1 + x + yz 2 )1/3
23. lim ;
r→0 2 + 3x − 4y + 5z 2

Solution: Note that both (1 + x + yz 2 )1/3 and (2 + 3x −


4y + 5z 2 ) are continuous at r0 = 0. Thus we have that

(1 + x + yz 2)1/3 limr→0 (1 + x + yz 2)1/3 11/3 1


lim 2
= 2
= =
r→0 2 + 3x − 4y + 5z limr→0 [2 + 3x − 4y + 5z ] 2 2

24. lim sin(x y);
r→0

Solution: Note that both sin(x y) is the composition of two
continuous functions and hence continuous at r0 = 0. Thus
we have that

lim sin(x y) = sin(0) = 0
r→0

sin(x y)
25. lim ;
r→0 cos(x2 y)

Solution: Note that both sin(x y) and cos(x2 y) are contin-
uous at r0 = 0 and cos(0) = 1, so the quotient is continuous
284 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

at the limit point. Thus we have that



sin(x y) 0
lim = =0
r→0 cos(x2 y) 1
26. lim[exyz − 2 cos(yz) + 3 sin(xy)];
r→0

Solution: Note that all of exyz , −2 cos(yz), and 3 sin(xy)


are continuous at (0, 0, 0), so their sum is continuous there.
Thus we have that
lim[exyz − 2 cos(yz) + 3 sin(xy)] = e0 − 2(1) + 3(0) = −1
r→0

27. lim ln(1 + x2 + y 2 z 2 ).


r→0

Solution: Note that 1+x2 +y 2z 2 is continuous at (0, 0, 0) and


is equal to 1 at that point, so the composition is continuous
there. Thus we have that
lim ln(1 + x2 + y 2z 2 ) = ln(1 + 0 + 0) = 0
r→0
28. Use Theorem 17.1 and the properties of limits of numerical se-
quences to prove Theorem 17.2.

Solution: Proof: Suppose limr→r0 f(r) = p and limr→r0 g(r) = q.


Let pn and qn be sequences such that pn = f(rn ) and qn = f(rn ) for
some sequence of points rn in D that converges to r0 with rn 6= r0 .
Then by Theorem 17.1 we have
lim f(r) = p ⇔ lim f(rn ) = p ⇔ lim pn = p,
r→r0 n→∞ n→∞

lim g(r) = q ⇔ lim g(rn ) = q ⇔ lim qn = q


r→r0 n→∞ n→∞
Recall the properties of numerical sequences, produced below:
lim [pn + qn ] = p + q
n→∞
lim cpn = cp
n→∞
lim pn qn = pq
n→∞
pn p
= , q 6= 0
lim
n→∞ qn q
The proof immediately follows.

29. Use Theorem 17.1 to prove Theorem 17.6.

Solution: Proof: Since h is continuous, limr→r0 h(r) = h(r0 ). Let


2. LIMITS AND CONTINUITY 285

rn be a sequence that converges to r0 but rn 6= r0. Then by The-


orem 17.1, limn→∞ h(rn ) = h(r0). Consider the limit limr→r0 f(r).
Again by Theorem 17.1, we have limr→r0 f(r) = limn→∞ f(rn ). But
this is exactly limn→∞ g(h(rn )) since f(r) = g(h(r)). Since g is con-
tinuous, we have that limu→u0 g(u) = g(u0 ). Hence limn→∞ g(h(rn )) =
g(limn→∞ h(rn )) = g(h(r0 )). Therefore f is continuous.
286 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

3. A General Strategy to Study Limits


1–3. Prove the following statements:
1. Let f(x, y) = (x − y)/(x + y). Then
lim(lim f(x, y)) = 1, lim(lim f(x, y)) = −1
x→0 y→0 y→0 x→0
but the limit of f(x, y) as (x, y) → (0, 0) does not exist.

Solution: The repeated limits are easily computed as fol-


lows
x−0
lim(lim f(x, y)) = lim( ) = lim (1) = 1
x→0 y→0 x→0 x + 0 x→0
0−y
lim(lim f(x, y)) = lim( ) = lim(−1) = −1
y→0 x→0 y→0 0 + y y→0

By Corollary 18.1, if the repeated limits are unequal or at


least one does not exist, then the multivariable limit does not
exist.

2. Let f(x, y) = x2 y 2 /(x2 y 2 + (x − y)2 ). Then


lim(lim f(x, y)) = lim(lim f(x, y)) = 0
x→0 y→0 y→0 x→0
but the limit of f(x, y) as (x, y) → (0, 0) does not exist.

Solution: The repeated limits are easily computed as fol-


lows
0
lim(lim f(x, y)) = lim ( ) = lim(0) = 0
x→0 y→0 x→0 0 + (x − 0)2 x→0

0
lim(lim f(x, y)) = lim( ) = lim(0) = 0
y→0 x→0 y→0 0 + (x − 0)2 y→0

Note that both limits can be found since the denominator is


nonzero when x = 0, y 6= 0 or y = 0 and x 6= 0. By Corollary
18.1, if the repeated limits are unequal or at least one does
not exist, then the multivariable limit does not exist. Note
that it does not make any statement about the case when the
limits exist and are equal. That is to say, this is not an iff
statement; it gives a criteria for sufficiency, but not necessity.
Thus it may or may not be that the multivariable limit exists,
and we must investigate it independently. Let x(t) = t and
y(t) = t. Then the limit becomes
t4 t4
lim = lim = lim(1) = 1
t→0 t4 + (t − t)2 t→0 t4 t→0
3. A GENERAL STRATEGY TO STUDY LIMITS 287

Hence the limit is path dependent (another path was investi-


gated in the repeated limits), and so the multivariable limit
does not exist.

3. Let f(x, y) = (x + y) sin(1/x) sin(1/y). Then


lim(lim f(x, y)) and lim(lim f(x, y))
x→0 y→0 y→0 x→0
do not exist, but the limit of f(x, y) exists and equals 0 as
(x, y) → (0, 0). Does the result contradict to Theorem 18.2?
Explain.

Solution: Evidently, the repeated limits do not exist because


of the sin(1/x) and sin(1/y) terms, which undefined along the
lines x = 0 and y = 0. However, this means that those lines
are not in the domain of f (but are limit points); therefore,
this is not a contradiction of Corollary 18.1. The multivari-
able limit is computed as follows:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: It is not wise to evaluate the limit along a curve. This
is because sin(1/x(t)) and sin(1/y(t)) will be difficult to work
with. p
Step 4: Let R = x2 + y 2. Then

|f(x, y) − 0| ≤ |x + y|| sin(1/x)|| sin(1/y)| ≤ |x + y| ≤ |x| + |y| ≤ 2R

where the inequality | sin u| ≤ 1 has been used. Since 2R goes


to 0 as R goes to 0, the limit exists and is 0.
4–17. Find each of the following limits or show that it does not exist:
cos(xy + z)
4. lim ;
r→0 x4 + y 2 z 2 + 4

Solution:
Step 1: The continuity argument applies since the denomina-
tor is nonzero at the origin; so the quotient is continuous at
the origin since it is the quotient of two continuous functions.
Hence,

cos(xy + z) cos(0) 1
lim 4 2 2
= =
r→0 x + y z + 4 0+0+4 4
288 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

sin(xy) − xy
5. lim ;
r→0 (xy)3
Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: We are able to transform this into a one-variable
limit with the substitution u = xy. With this, u → 0 as
(x, y) → (0, 0). Hence the limit is
sin(xy) − xy sin u − u cos u − 1 − sin u − cos u 1
lim 3
= lim 3
= lim 2
= lim = lim =−
r→0 (xy) u→0 u u→0 3u u→0 6u u→0 6 6
where l’Hospital’s rule has been used three times.
p
xy 2 + 1 − 1
6. lim ;
r→0 xy 2
Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: We are able to transform this into a one-variable
limit with the substitution u = xy 2. With this, u → 0 as
(x, y) → (0, 0). Hence the limit is
p √ p √
xy 2 + 1 − 1 u+1−1 xy 2 + 1 − 1 1/(2 u + 1) 1
lim 2
= lim = 2
= lim =
r→0 xy u→0 u xy u→0 1 2
sin(xy 3)
7. lim ;
r→0 x2
Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: Consider the repeated limit
sin(xy 3) y 3 cos(xy 3)
lim(lim ) = lim (lim ) = lim(±∞) = ±∞
y→0 x→0 x2 y→0 x→0 x2 y→0

where y is treated as a constant in evaluating the inner limit.


Hence the limit does not exist.

x3 + y 5
8. lim ;
r→0 x2 + 2y 2
3. A GENERAL STRATEGY TO STUDY LIMITS 289

Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: Let x(t) = t and y(t) = st for some constant s. Then
the limit becomes
x3 + y 5 t 3 + s5 t 5 t + s5 t 3
lim 2 = lim = =0
r→0 x + 2y 2 t→0 t2 + 2s2 t2 1 + 2s2
since 1 + 2s2 is never 0 when s is real.

ekrk − 1 − krk
9. lim ;
r→0 krk2
Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: Let u = krk. Then the limit becomes
ekrk − 1 − krk eu − 1 − u eu − 1 eu 1
lim 2
= lim 2
= lim = lim =
r→0 krk u→0 u u→0 2u u→0 2 2
where l’Hospital’s rule has been used twice.

x2 + sin2 y
10. lim ;
r→0 x2 + 2y 2

Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: Let x(t) = t and y(t) = st for some constant s. Then
the limit becomes
x + sin2 y
2
t2 + sin2 (st) 2t + 2s sin(st) cos(st) 2 + 2s2 cos(2st) 1 + s2
lim 2 = lim = lim = lim =
r→0 x + 2y 2 t→0 t2 + 2s2 t2 t→0 2t + 4s2 t t→0 2 + 4s2 1 + 2s2
Therefore the limit depends on the path (depends on s), and
does not exist.

xy 2 + x sin(xy)
11. lim ;
r→0 x2 + 2y 2
Solution:
290 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

Step 1: The continuity argument does not apply because f is


not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: Let x(t) = t and y(t) = st for some constant s. Then
the limit becomes
xy 2 + x sin(xy) t(s2 t2) + t sin(t(st)) s2t2 + sin(st2 )
lim = lim = lim
r→0 x2 + 2y 2 t→0 t2 + 2s2 t2 t→0 t + 2s2 t
2ts2 + 2st cos(st2)
= lim =0
t→0 1 + 2s2
since 1 + 2s2 is never 0 for real s. However, this does not prove
the limit exists. All it does is give us an educated guess on
what the limit should be. Recall that the limit exists if all
curves have the above limit, but we did not test all curves,
only straight lines. So we must continue to conclusively prove
whether the limit exists or not, using our guess that the limit
should be 0. p
Step 4: Let R = x2 + y 2. Then
|x|y 2 + |x|| sin(xy)| R(R2 ) + R|xy| R3 + R3
|f(x, y) − 0| ≤ ≤ ≤ = 2R
x2 + y 2 R2 R2
which evidently goes to 0 as R goes to 0. Hence the limit exists
and is 0.

ln(x + ey )
12. lim p ;
(x,y)→(1,0) x2 + y 2
Solution:
Step 1: The continuity argument applies since both the nu-
merator and denominator are continuous at the limit point.
So the limit is
ln(x + ey ) ln(1 + e0 ) ln(2)
lim p = √ = √ = ln 2
(x,y)→(1,0) x2 + y 2 12 + 02 1
2 y2
13. lim(x2 + y 2 )x ;
r→0

Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
3. A GENERAL STRATEGY TO STUDY LIMITS 291

Step 3: Let x(t) = s cos t and y(t) = s sin t for some constant
s 6= 0. Then the limit becomes
2 y2 4 cos2 t sin2 t 4 sin2 (2t)/2
lim (x2 + y 2)x = lim(s2)s = lim(s)s =1
r→0 t→0 t→0

since s 6= 0, we have that s0 = 1. So we guess that the limit is


1. p
Step 4: Let R = x2 + y 2. UNSURE HOW TO PROCEED.
2 2 2 2
|f(x, y) − 1| = |(x2 + y 2)x y − 1x y | = |((x2 + y 2 )xy − 1xy )((x2 + y 2 )xy + 1xy )|
= |((x2 + y 2)xy − 1)((x2 + y 2)xy + 1)|
1 xy
14. lim tan( );
r→0 xy 1 + xy
Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: Let u = xy. Then the limit becomes
1 xy 1 u 1 1
lim tan( ) = lim tan( ) = lim sec2 (1− )[ 2
] = sec2 (1−1)(1) = 1
r→0 xy 1 + xy u→0 u 1+u u→0 1 + u (1 + u)
sin(x2 − y 2 ) 2
15. lim ln( );
r→0 x2 − y 2
Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: Let u = x2 − y 2 . Then the limit becomes
sin(x2 − y 2 ) 2 sin u 2
lim ln( 2 2
) = lim ln( ) = ln(1)2 = 0
r→0 x −y u→0 u
Note that sin u/u can be continuously extended to u = 0. It
is easy to verify that the value here should be 1. Then, the
limit above is a composition of continuous functions, and we
may simply substitute.

xy + 1 − 1
16. lim √ ;
r→0 y x
Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
292 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

one-variable limit. √
Step 3: Let x(t) = t and y(t) = s t. Then the limit becomes
√ √ √ √ √
xy + 1 − 1 st3/2 + 1 − 1 (3/2s t)/(2 st3/2 + 1) 3 t
lim √ = lim = lim = lim √ =0
r→0 y x t→0 st t→0 s t→0 4 st3/2 + 1

Hence we guess that


p the limit is 0 if it exists.
Step 4: Let R = x2 + y 2. Then

R2 + 1 − 1
|f(x, y) − 0| ≤
R3/2
which goes to 0 as R goes to 0. Therefore the limit exists and
is 0.

xb y a
17. lim , 0 < b < a, x, y > 0;
r→0 xa + y b

Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: Let x(t) = t and y(t) = st where s is a constant. Then
the limit becomes
b a
xy sa ta+b sa t a sa (a)...(b + 1)tb
lim a = lim = lim = lim =0
r→0 x + y b t→0 ta + sb tb t→0 ta−b + sb t→0 (a − b)!
Note that this limit is ”top heavy” since a > a − b > 0. The
above limit can be proved by using L’hopital’s rule until the
denominator is a constant; this establishes the second to last
equality. Thus our
p guess for the limit is 0 if it exists.
Step 4: Let R = x2 + y 2. Then we have
|xb ||y a| Ra+b
|f(x, y) − 0| ≤ b
≤ a
= Rb
|y | R
[Recall that |xa + y b| ≥ |xa + y a| for sufficiently small R. So,
1/|xa + y b | ≤ 1/|xa + y a|]. Evidently this goes to 0 as R goes
to 0 since b > 0. Therefore the limit exists and is 0.
18. Let
|x| + |y| − |x + y|
f(x, y) = 2 2 k
if x2 + y 2 6= 0, and f(0, 0) = c
(x + y )
Find all values of constants c and k > 0 at which the function is con-
tinuous at the origin.
3. A GENERAL STRATEGY TO STUDY LIMITS 293

Solution: Consider the limit along the path y = x. Evidently, the


limit will be 0 since the numerator evaluates to 0 directly. Now consider
the limit along the path y = −x. Then the limit becomes
|x| + | − x| − |x − x| 2|x|
lim 2 2 k
= lim k 2k
x→0 (x + (−x) ) x→0 2 x

We can break this limit into the left and right limits as follows
2x 1
lim k 2k = lim k−1 2k−1
x→0+ 2 x x→0+ 2 x
2x 1
lim − k 2k = lim − k−1 2k−1
x→0− 2 x x→0+ 2 x
The limit will not exist if we have something in the form 1/xK for
positive K. Hence, we require that 2k − 1√≤ 0 ⇔ √k ≤ 1/2. In the
case when k = 1/2, the limits evaluate to 2 and − 2, so the overall
limit does not exist. Hence k < 1/2; in these cases, both limits are 0,
and so the overall limit is 0. Note that this matches what we expect,
given that the path y = x showed that the plimit, if it exists, should be
0. Going immediately to step 4, let R = x2 + y 2. Then,
|x| + |y| + |x + y| |x| + |y| + |x| + |y| 4R 4
|f(x, y)−0| ≤ 2 2 k
≤ k
≤ 2k = 2k−1
(x + y ) R R R
Note that this will only go to 0 if 2k − 1 < 0, as expected. So choosing
c = 0 when 0 < k < 1/2 makes f continuous at the origin.

19. Let f(x, y) = x2 y/(x4 + y 2 ) if x2 + y 2 6= 0 and f(0, 0) = 0. Show


that f is continuous along any straight line through the origin, i.e.,
F (t) = f(x(t), y(t)) is continuous for all t where x(t) = t cos θ, y(t) =
t sin θ for any fixed θ, nevertheless f is not continuous at (0, 0). Hint:
Investigate the limits of f along power curves through the origin.

Solution: Let x(t) = t cos θ and y(t) = t sin θ for a fixed θ. Then
the limit becomes
t2 cos2 θ(t sin θ) t cos2 θ sin θ
lim 4 = lim =0
t→0 t cos4 θ + t2 sin2 θ t→0 t2 cos4 θ + sin2 θ

so long a θ 6= 0, π, 2π, ...; the above limit is found by simply substituting


t = 0 since both numerator and denominator are continuous and the
denominator is nonzero. For the θ given above, we have that x(t) = ±t
and y(t) = 0. So, f(x(t), y(t)) = t2(0)/(t4 + 0) = 0, and hence the limit
is 0 in these cases as well. So f is continuous along these curves.
Next, consider the limit along the curve x(t) = t, y(t) = at2 for some
294 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

constant a. I choose y to be a degree 2 polynomial so that in the


denominator I will have a polynomial that is easily factorizable. Then
the limit becomes
t2(at2) at4 a
lim 4 2 4
= lim 4 2
=
t→0 t + a t t→0 t (1 + a ) 1 + a2
which evidently depends on a, and hence the limit does not exist.

20. Let f(x, y) be continuous in a rectangle a < x < b, c < y < d.


Let g(x) be continuous on the interval (a, b) and takes values in (c, d).
Prove that the function F (x) = f(x, g(x)) is continuous on (a, b).

Solution: Proof: Let a < p < b. Consider the limit


lim F (x) = f(p, lim g(x)) = f(p, g(p) = F (p)
x→p x→p

So F is continuous. The first equality is reached due to continuity of


f; the second due to continuity of g.
2 −y)
21. Investigate the limits of the function f(x, y) = x2 e−(x along
the rays x(t) = cos(θ)t, y(t) = sin(θ)t, as t → ∞ for all 0 ≤ θ ≤ 2π.
Are the values of the function arbitrary small for all krk > δ if δ is
large enough? Does the limit limr→∞ f(x, y) exist?

Solution: Let x(t) = t cos θ, y(t) = t sin θ for a fixed θ. Then the
limit becomes
t2 cos2 θ 2t cos2 θ
lim t2 cos2 θ−t sin θ = lim 2 2
t→∞ e t→∞ (2t cos2 θ − sin θ)et cos θ−t sin θ

2 cos2 θ
= lim 2 2 2 2 =0
t→∞ (2 cos2 θ)et cos θ−t sin θ + (2t cos2 θ − sin θ)2 et cos θ−t sin θ

The second and third questions are actually equivalent. ”Arbitrarily


small” refers to  and whether |f(x, y)−c| <  for krk > δ. So we answer
the third question, which is easier to approach (the second question is
step 4). Consider x(t) = t and y(t) = t2 . Then the limit becomes
2 −t2 )
lim t2 e−(t = lim t2e0 = lim t2 = ∞
t→∞ t→∞ t→∞

and so the limit does not exist

22–31. Find the limit or show that it does not exist:


sin(x2 + y 2 + z 2 )
22. lim ;
r→0 x4 + y 4 + z 4
3. A GENERAL STRATEGY TO STUDY LIMITS 295

Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: As a guess, consider x(t) = t, y(t) = t, and z(t) = t.
Then the limit becomes

sin(x2 + y 2 + z 2 ) sin(3t2 ) 6t cos(3t2 ) cos(3t2)


lim = lim = lim = lim =∞
r→0 x4 + y 4 + z 4 t→0 3t4 t→0 12t3 t→0 2t2

This does not mean that the limit does not exist! Just that if
it exists, then it is ∞. We must now move to step 4.
Step 4: This step will be slightly modified given that the limit
is not a finite number. A limit is said to be ∞ if for every M >
0 there exists a δ > 0 such that if 0 < kr−r0 k < δ then f(r) >
M. The notation M is common when attempting to found an
upper/lower bound for a function. Here, we are saying that
for every attempted upper bound M, one can always find an
r such that f(r) exceedsp that bound; hence, f is unbounded.
So, suppose that R = x2 + y 2 + z 2 . Then,

| sin(x2 + y 2 + z 2 )| R2 1
M < |f(x, y, z)| ≤ 4 4 4
≤ 4
=
|x| + |y| + |z| 3R 3R2

2

And so we choose
√ R such that M < 1/(3R ) ⇔ R < 1/ 3M .
Then if δ ≤ 1/ 3M , the values of f for 0 < krk < δ will be
larger than any preassigned positive number M.

x2 + 2y 2 + 3z 2
23. lim ;
r→0 x4 + y 2 z 4
Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: As a guess, consider x(t) = t, y(t) = t, and z(t) = t.
Then the limit becomes

x2 + 2y 2 + 3z 2 6t2 6
lim 4 2 4
= lim 4 6
= lim 2 =∞
r→0 x +y z t→0 t + t t→0 t + t4
296 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

p
Step 4: Suppose that R = x2 + y 2 + z 2 . Then,
|x|2 + 2|y|2 + 3|z|2 R2 + 2R2 + 3R2 6R2 3
M < |f(x, y, z)| ≤ 4 2 4
≤ 4 6
≤ 6
= 4
|x| + |y| |z| R +R 2R R
owing to the fact that for sufficiently smallpR. And so we
choose R such that M < 3/(R4 ) p ⇔ R < 4 3/M . Then if
δ ≤ δ1 where δ1 is the smallest of 4 3/M and 1, the values of
f for 0 < krk < δ will be larger than any preassigned positive
number M.

ln(x2y 2z 2 )
24. lim ;
r→∞ x2 + y 2 + z 2

Hint : Consider the limits along the curves


2
x = y = z = t and x = e−t , y = z = t
Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: First let x(t) = y(t) = z(t) = t. Then the limit
becomes
ln(x2y 2 z 2) ln(t6) 2 ln t 2(1/t)
lim 2 = lim = lim = lim =0
r→∞ x + y 2 + z 2 t→∞ 3t2 t→∞ t2 t→∞ 2t
2
Now let x(t) = e−t and y(t) = z(t) = t. Then the limit
becomes
2
ln(x2y 2 z 2) ln(e−2t t4) −2t2 + 4 ln(t) −4t + 4/t
lim 2 2 2
= lim −2t 2 2
= lim −2t 2 2
= lim 2
r→∞ x + y + z t→∞ e + 2t t→∞ e + 2t t→∞ −4te−2t + 4t

−t2 + 1 −2t
= lim 2 = lim 2
t→∞ −t e2 −2t +t 2 t→∞ −2te −2t + 4t3e−2t2 + 2t
−1
= lim 2 2 = −1
t→∞ −e−2t + 23 e−2t + 1

therefore the limit does not exist because it is path dependent.


2 2 2
e3x +2y +z
25. lim ;
r→∞ (x2 + 2y 2 + 3z 2 )2012

Solution:
Step 1: The continuity argument does not apply because f is
3. A GENERAL STRATEGY TO STUDY LIMITS 297

not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: As a guess, let x(t) = y(t) = z(t) = t. Then the limit
becomes
2 2 2 2
e3x +2y +z e6t eu eu
lim 2 = lim = lim = ... = lim =∞
r→∞ (x + 2y 2 + 3z 2 )2012 t→∞ (6t2 )2012 u→∞ u2012 u→∞ 2012!

where the substitution u = 6t2 has been used, and l’Hospital’s


rule has been used many times.
Step 4: Consider the function g(x, y, z) = (1 + 3x2 + 2y 2 +
z 2 )/(x2 + 2y 2 + 3z 2 ). Owing to the inequality 1 + u ≤ eu ,
2 2 2
we have that g ≤ f(x, y, z) = e3x +2y +z /(x2 + 2y 2 + 3z 2 ).
It suffices to show thatp limr→∞ g(r) = ∞, due to the squeeze
theorem. Let R = x2 + y 2 + z 2. For limits at infinity we
need that g(r) > M for every M > 0 if krk > δ. Then,
z
26. lim ;
r→0 x2 + y2 + z2
Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: Take x(t) = y(t)t, z(t) = −t. Then the limit becomes
z −t −1
lim = lim 2 = lim = −∞
r→0 x2 + y 2 + z 2 t→0 3t t→0 3t

Now take x(t) = y(t)t, z(t) = −t2. Then the limit becomes
z −t2 −1 1
lim 2 2 2
= lim 2
= lim =−
r→0 x + y + z t→0 3t t→0 3 3
so the limit does not exist because it is path dependent
z
27. lim if z < 0;
r→0 x2 + y2 + z2
Solution: Both paths taken in Exercise 26 satisfy the con-
dition z < 0; Hence the limit still does not exist.

x2 + y 2
28. lim ;
r→∞ x2 + y 4

Solution:
298 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

Step 1: The continuity argument does not apply because f is


not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: Let x(t) = y(t) = t. Then the limit becomes
x2 + y 2 2t2 1
lim 2 4
= lim 2 4
= lim =0
r→∞ x + y t→∞ t + t t→∞ 1 + t2

Next, take x(t) = t and y(t) = t. Then the limit becomes
x2 + y 2 t2 + t 1+t
lim 2 4
= lim 2
= lim =1
r→∞ x + y t→∞ 2t t→∞ t
hence the limit does not exist.
πx
29. lim sin( );
r→∞ 2x + y
Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: Let x(t) = y(t) = t. Then the limit becomes

πx πt π 3
lim sin( ) = lim sin( ) = lim sin( ) =
r→∞ 2x + y t→∞ 2t + t t→∞ 3 2

Now let x(t) = t, y(t) = 2t. Then the limit becomes



πx πt π 2
lim sin( ) = lim sin( ) = lim sin( ) =
r→∞ 2x + y t→∞ 2t + 2t t→∞ 4 2
so the limit does not exist.

30. lim (x2 + y 2 )e−|x+y| ;


r→∞

Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: Let x(t) = y(t) = t. Then the limit becomes
2t2 4t 1
lim (x2 +y 2)e−|x+y| = lim 2t2e−2|t| = lim = lim = lim =0
r→∞ t→∞ t→∞ e−2t t→∞ −2e−2t t→∞ e−2t
3. A GENERAL STRATEGY TO STUDY LIMITS 299

we are allowed to drop the absolute value because t will even-


tually be positive (as it heads to positive infinity). Next, let
x(t) = t and y(t) = −t. Then the limit becomes
lim (x2 + y 2)e−|x+y| = lim 2t2 e0 = ∞,
r→∞ t→∞

hence the limit does not exist.


xy 2
31. lim ( )x ;
r→∞ x2 +y 2

Solution:
Step 1: The continuity argument does not apply because f is
not defined at r0 .
Step 2: No substitution is possible to transform the limit to a
one-variable limit.
Step 3: Let x(t) = t, y(t) = t. Then the limit becomes
xy x2 t 2 t2 1 2
lim ( 2 2
) = lim ( 2
) = lim ( )t = 0
r→∞ x +y r→∞ 2t r→∞ 2

Next let x(t) = 0 and y(t) = t. Then the limit becomes


xt x2
lim (lim( ) )
t→∞ x→0 x2 + t2
With respect to the inner limit, t is a constant. Thus we may
do the following:
xt x2
L = lim( )
x→0 x2 + t2
xt ln x + ln t − ln(x2 + t2 ) 1/x − 2x/(x2 + t2)
ln L = lim x2 ln( 2 ) = lim = lim
x→0 x + t2 x→0 1/x2 x→0 −2/x3
x2 − 2x4 /(x2 + t2)
= lim =0
x→0 −2
L = e0 = 1
Then the repeated limit is
xt x2
lim (lim( ) ) = lim (1) = 1
t→∞ x→0 x2 + t2 t→∞

Thus the limit is path dependent and does not exist.


32. Find the repeated limits

lim(lim logx (x + y)) and lim(lim logx (x + y))


x→1 y→0 y→0 x→1
300 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

What can be said about the corresponding two-variable limit?

Solution: First recall the change of base formula


ln b
loga b =
ln a
(proof: set loga b = k then b = a , and ln b = ln ak = k ln a. So
k

k = ln b/ ln a). The repeated limits are computed below:


ln(x + y) ln(x + 0)
lim(lim ) = lim ( ) = lim (1) = 1
x→1 y→0 ln x x→1 ln x x→1

ln(x + y) ln(1 + y)
lim(lim ) = lim( )=∞
y→0 x→1 ln x y→0 ln 1
So the multivariable limit does not exist.
4. PARTIAL DERIVATIVES 301

4. Partial Derivatives
1–7. Find the specified partial derivatives of each of the following
functions:
1. f(x, y) = (x − y)/(x + y), fx0 (1, 2), fy0 (1, 2);

Solution: Recall that the partial derivative of a multivari-


able function f with respect to one of its variables, say x,
means to treat all other variables as constants and take the
derivative with respect to x. So the partial derivatives are

(1)(x + y) − (1)(x − y) 2y
fx0 = 2
=
(x + y) (x + y)2
(−1)(x + y) − (1)(x − y) −2x
fy0 = 2
=
(x + y) (x + y)2

Note the sign change of (1) to (−1) in computing u0 in the quo-


tient rule; This is because the variable we are differentiating
with respect to changes from x to y. The partial derivatives
at (1, 2) are

2(2) 4
fx0 (1, 2) = 2
=
(1 + 2) 9
−2(1) 2
fy0 (1, 2) = 2
=−
(1 + 2) 9

One can also calculate the partial derivatives (at a specific


point only!) as follows:

x−2 0 x + 2 − (x − 2) 4
fx0 (1, 2) = f 0 (x, 2)|x=1 = ( ) |x=1 = 2
|x=1 =
x+2 (x + 2) 9

where we have substituted y = 2 and taken the derivative with


respect to x as usual. I will not take this approach so as to
familiarize you with taking partial derivatives in general, but
remember that the above approach is a suitable alternative for
these problems.

2. f(x, y, z) = (xy + z)/(z + y), fx0 (1, 2, 3), fy0 (1, 2, 3), fz0 (1, 2, 3);

Solution: As above, we begin by computing the partial


302 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

derivatives, shown below


y
fx0 = =
z+y
x(z + y) − (1)(xy + z) xz − z
fy0 = =
(z + y)2 (z + y)2
(1)(z + y) − (1)(xy + z) y − xy + z
fz0 = 2
=
(z + y) (z + y)2
(No quotient rule is necessary for fx0 because the denominator
does not contain x! Treat it like a constant!). The partial
derivatives at (1, 2, 3) are
2 2
fx0 (1, 2, 3) = =
3+2 5
(1)(3) − 3
fy0 (1, 2, 3) = =0
(3 + 2)2
2 − 1(2)
fz0 (1, 2, 3) = − =0
(3 + 2)2
3. f(r) = (x1 + 2x2 + ... + mxm)/(1 + krk2 ), fx0 i (0), i = 1, 2, ..., m;

Solution: As above, we begin by computing the partial


derivatives, shown below
i(1 + krk2) − (x1 + 2x2 + ... + mxm )(2i2xi )
fx0 i =
(1 + krk2)2
Recall that krk2 = (x1 )2 +(2x2)2 +...+(mxm)2, hence (krk2 )0 =
2i2 xi . Alternatively, recall that krk2 = r · r, hence (krk2 )0 =
r0 · r + r · r0 = 2r · r0. But r = hx1, 2x2 , ..., ixi, ..., mxmi, so
r0 = h0, 0, ..., i, ..., 0i, where the ith component is 1 since all
other components do not contain xi . The partial derivatives
at 0 are
i(1 + 02 ) − (0)(2i2 (0))
fx0 i (0) = =i
(1 + 02 )2
4. f(x, y, z) = x sin(yz), fx0 (1, 2, π/2), fy0 (1, 2, π/2), fz0 (1, 2, π/2);

Solution: As above, we begin by computing the partial


derivatives, shown below
fx0 = sin(yz)
fy0 = xz cos(yz)
fz0 = xy cos(yz)
4. PARTIAL DERIVATIVES 303

The partial derivatives at (1, 2, π/2) are


fx0 (1, 2, π/2) = sin(2(π/2) = 0)
fy0 (1, 2, π/2) = (1)(π/2) cos(2(π/2)) = −π/2
fz0 (1, 2, π/2) = (1)(2) cos(2(π/2)) = −2
p
5. f(x, y) = x + (y − 1) arcsin( x/y), fx0 (1, 1), fy0 (1, 1);

Solution: As above, we begin by computing the partial


derivatives, shown below
p
0 (1/2 1/(xy)) y−1
fx = 1 + (y − 1) q = 1+ √ √
p
1 − ( x/y)2 2 x y−x
√ √
0
p −1/2 xy −3/2 p x
fy = arcsin( x/y) + (y − 1) q = arcsin( x/y) − (y − 1) √
p 2y y − x
1 − ( x/y)2
The partial derivatives at (1, 1) are
fx0 (1, 1) = 1
fy0 (1, 1) = π/2
You will need to use a limit to compute both partial deriva-
tives.

6. f(x, y) = (x3 + y 3 )1/3 , fx0 (0, 0), fy0 (0, 0);

Solution: As above, we begin by computing the partial


derivatives, shown below
x2
fx0 =
(x3 + y 3)2/3
y2
fy0 = 3
(x + y 3)2/3
The partial derivatives at (0, 0) are
fx0 (0, 0) = 1
fy0 (0, 0) = 1
An intuition for this is that the degree of the numerator for
each is 2, and the ”degree” of the denominator is 3∗ (2/3) = 2,
so they are ”equal”. Hence the limit is just the ratio of the
coefficients: 1.
304 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

p
7. f(x, y) = |xy|, fx0 (0, 0), fy0 (0, 0).

Solution: As above, we begin by√ computing p the partial


derivatives. First, recall that |x| = x . So |x| = (x2 )1/4.
2

Therefore we have
p p p
0 (1/4)2x |y| x |y| x |y|
fx = = √ = p
(x2 )3/4 2( x2 )3/2 2( |x|)3/2
p
y |x|
fy0 = p
2( |y|)3/2

The partial derivatives at (0, 0) are


fx0 (0, 0) = 0
fy0 (0, 0) = 0
The squeeze principle can be used here; a sketch is produced
below √
0 R( R) R3/4
|fx − 0| ≤ =
2R3/4 2
which goes to 0 as R goes to 0.

8–23. Find the partial derivatives of each of the following functions:


8. f(x, y) = (x + y 2 )n ;

Solution: All these problems are just computing a bunch


of derivatives. The key point is when calculating one par-
tial derivative, you must treat all other variables as constants.
Therefore we have
fx0 = n(x + y 2)n−1
fy0 = n(x + y 2)n−1 (2y) = 2ny(x + y 2)n−1
9. f(x, y) = xy ;

Solution: We have
fx0 = yxy−1
fy0 = ln x(xy )
The latter partial derivative is computed as follows:
ln(f(x, y)) = y ln x ⇔ fy0 /f(x, y) = ln x ⇔ fy0 = f(x, y) ln x
4. PARTIAL DERIVATIVES 305

where x is treated as a constant.


2
10. f(x, y) = xe(x+2y) ;

Solution: We have
2 2 2
fx0 = e(x+2y) + x(2(x + 2y))e(x+2y) = (1 + 2x2 + 4xy)e(x+2y)
2 2
fy0 = (2(x + 2y)(2))xe(x+2y) = 4x(x + 2y)e(x+2y)
where the (2) in the unsimplified form of fy0 came from chain
rule with the x + 2y term.

11. f(x, y) = sin(xy) cos(x2 + y 2 );

Solution: We have
fx0 = y cos(xy) cos(x2 + y 2 ) + sin(xy)(− sin(x2 + y 2)(2x))
= y cos(xy) cos(x2 + y 2 ) − 2x sin(xy) sin(x2 + y 2)
fy0 = x cos(xy) cos(x2 + y 2 ) + sin(xy)(− sin(x2 + y 2 )(2y))
= x cos(xy) cos(x2 + y 2 ) − 2y sin(xy) sin(x2 + y 2)
12. f(x, y, z) = ln(x + y 2 + z 3 );

Solution: We have
1
fx0 =
x + y2 + z3
2y
fy0 =
x + y2 + z3
3z 2
fz0 =
x + y2 + z3
13. f(x, y, z) = xy 2 cos(z 2 x);

Solution: We have
fx0 = y 2 cos(z 2x) + xy 2(− sin(z 2x)(z 2 )) = y 2 cos(z 2 x) − xy 2z 2 sin(z 2 x)
fy0 = x(2y) cos(z 2 x) = 2xy cos(z 2 x)
fz0 = xy 2(− sin(z 2x)(2zx)) = −2x2 y 2z sin(z 2 x)
14. f(r) = (a1 x1 + a2 x2 + ... + am xm )n = (a · r)n ;

Solution: We have
fx0 i = ai n(a1x1 + a2 x2 + ... + am xm )n−1 = nai (a · r)n−1
306 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

for i = 1, 2, ..., m
15. f(x, y) = arctan(y/x);

Solution: We have
−y/x2 y
fx0 = = −
1 + (y/x)2 x2 + y 2
1/x x
fy0 = 2
= 2
1 + (y/x) x + y2
p
16. f(x, y) = arcsin(x/ x2 + y 2 );

Solution: We have
p p
0 1 x2 + y 2 − x/(2 x2 + y 2 )(2x)
fx = q ( )
p
2 2 2 x2 + y 2
1 − (x/ x + y )
p p p
x2 + y 2 x2 + y 2 − x2 /( x2 + y 2 )
= p ( )
x2 + y 2
p
x2 + y 2 1 − x2 /(x2 + y 2)
1 x2 + y 2 − x2 y2 |y|
= p ( ) = =
x2 + y 2 − x2 x2 + y 2 |y|(x2 + y 2) x2 + y 2
1
fy0 = q p (x(2y)(−1/2)(x2 + y 2)−3/2 )
1 − (x/ x2 + y 2)2
xy
=− p p
(x + y ) x + y 2 1 − x2/(x2 + y 2)
2 2 2

xy xy
=− =
|y|(x2 + y 2 )
p
(x2 + y 2) x2 + y 2 − x2
z
17. f(x, y, z) = xy ;

Solution: We have
z
fx0 = y z xy −1
z
fy0 = ln(x)zy z−1 xy
z
fz0 = ln(x) ln(y)y z xy
where the zy z−1 term in fy0 came from the chain rule in the
exponent, and fz0 is calculated as follows:
ln(ln f(x, y, z)) = ln(y z ln(x)) = z ln y+ln ln x ⇔ (fz0 /f(x, y, z))/ ln(f(x, y, z)) = ln y
18. f(x, y, z) = xy/z ;
4. PARTIAL DERIVATIVES 307

Solution: We have

fx0 = y/zxy/z−1
fy0 = ln(x)/zxy/z
fz0 = − ln(x)y/z 2 xy/z

19. f(x, y) = tan(x2 /y);

Solution: We have

fx0 = 2x/y sec2(x2 /y)


fy0 = −x2/y 2 sec2 (x2 /y)

20. f(x, y, z) = sin(x sin(y sin z));

Solution: We have

fx0 = sin(y sin z) cos(x sin(y sin z))


fy0 = x sin(z) cos(y sin z) cos(x sin(y sin z))
fz0 = xy cos(z) cos(y sin(z)) cos(x sin(y sin z))

This is really just excessive chain rule. It may be helpful to


define g(u) = sin u and write f as g(xg(yg(z))).

21. f(x, y) = (x + y 2 )/(x2 + y);

Solution: We have
(1)(x2 + y) − (x + y 2)(2x) y − 2xy 2 − x2
fx0 = =
(x2 + y)2 (x2 + y)2
(2y)(x2 + y) − (x + y 2)(1) 2yx2 + y 2 − x
fy0 = =
(x2 + y)2 (x2 + y)2
22. f(x, y, z) = a · (b × r), where a and b are constant vectors;

Solution: We have

fu0 = a0u · (b × r) + a · (b × r)0u = a · (b0u × r + b × r0u ) = a · (b × r0u )

since a0u = b0u = 0 (because they are constant). Moreover,


r0x = h1, 0, 0i, r0y = h0, 1, 0i, and r0z = h0, 0, 1i. So, b × r0x =
h0, b3 , −b2i, b × r0y = h−b3, 0, b1 i, and b × r0z = hb2 , −b1, 0i
308 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

Thus we have
fx0 = a2b3 − a3 b2 = (a × b)x
fy0 = −a1b3 + a3 b1 = (a × b)y
fz0 = a1b2 − a2 b1 = (a × b)z
23. f(x, y, z) = ka × rk, where a is a constant vector.

Solution: It is more useful to consider f(x, y, z)2. Note that


(f(x, y, z)2)0u = 2f(x, y, z)fu0 . So,
f(x, y, z)2 = ka×rk2 = (a×r)·(a×r) = (a·a)(r·r)−(a·r)(r·a) = kak2 krk2 −(a·r)2
where the last identity comes from the so-called ”quadruple
product” (See Chapter 1 Section 5 Exercise 12 for more infor-
mation). Hence,
(f(x, y, z)2)0u = kak2 (2u) − 2(a0u · r + a · r0u)(a · r) = 2u(a · a) − 2au(a · r)
We may then conclude that
1
2f(x, y, z)fu0 = 2u(a·a)−2au (a·r) ⇔ fu0 = (u(a·a)−au (a·r))
ka × rk
where a = hax , ay , az i.
24–28. For each of the following functions, determine whether the func-
tion f increases or decreases at a specified point P0 when one variable
increases, while the others are fixed:
24. f(x, y) = xy/(x + y), P0 = (1, 2);

Solution: To determine whether the function f increases


or decreases at a point P0 while one variable increases and
all others are fixed, we must consider the partial derivative of
each variable. First, keeping y constant, we have
y(x + y) − (1)(xy) y2
fx0 (x, y) = =
(x + y)2 (x + y)2
At P0 , we have
22 4
fx0 (1, 2) = 2
= >0
(1 + 2) 9
so f is increasing.
Next, keeping x constant, we have
x(x + y) − (1)(xy) x2
fy0 (x, y) = =
(x + y)2 (x + y)2
4. PARTIAL DERIVATIVES 309

At P0 , we have
12 1
fy0 (1, 2) = 2
= >0
(1 + 2) 9
so f is increasing.

25. f(x, y) = (x2 − 2y 2 )1/3 , P0 = (1, 1);

Solution: As above, first keep y constant. We have


2x
fx0 (x, y) =
3(x − 2y 2 )2/3
2

At P0 , we have
2(1) 2 2
fx0 (1, 1) = = = >0
3((1)2 2
− 2(1) ) 2/3 2
3((−1) ) 1/3 3
so f is increasing.
Next, keeping x constant, we have
4y
fy0 (x, y) = −
3(x2 − 2y 2 )2/3
At P0 , we have
4(1) 4
fy0 (1, 1) = − = − <0
3((1)2 − 2(1)2 )2/3 3
so f is decreasing.

26. f(x, y) = x2 sin(xy), P0 = (−1, π);

Solution: As above, first keep y constant. We have


fx0 (x, y) = 2x sin(xy) + x2 y cos(xy)
At P0 , we have
fx0 (−1, π) = 2(−1) sin(−π) + (−1)2 π cos(−π) = −π < 0
so f is decreasing.
Next, keeping x constant, we have
fy0 (x, y) = x3 cos(xy)
At P0 , we have
fy0 (−1, π) = (−1)3 cos(−π) = 1 > 0
so f is increasing.
310 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

27. f(x, y, z) = zx/(x + y 2 ), P0 = (1, 1, 1);

Solution: As above, first keep y, z constant. We have


z(x + y 2 ) − zx(1) zy 2
fx0 (x, y, z) = =
(x + y 2)2 (x + y 2)2
At P0 , we have
(1)(1)2 1
fx0 (1, 1, 1) = = >0
(1 + (1)2 )2 4
so f is increasing.
Next, keeping x, z constant, we have
zx(−2y))
fy0 (x, y, z) =
(x + y 2 )2
At P0 , we have
2(1)(1)(1) 2 1
fy0 (1, 1, 1) = − 2 2
=− =− <0
(1 + (1) ) 4 2
so f is decreasing.
Finally, keeping x, y constant, we have
x
fz0 (x, y, z) =
x + y2
At P0 , we have
1 1
fz0 (1, 1, 1) = 2
= >0
1 + (1) 2
so f is increasing.
p
28. f(x, y, z) = (x + yz)/ x + 2y + z 2 , P0 = (1, 2, 2).

Solution: As above, first keep y, z constant. We have


p p
x + 2y + z 2 − (x + yz)/(2 x + 2y + z 2 )
fx0 (x, y, z) =
x + 2y + z 2
At P0 , we have
p p
0 1 + 2(2) + (2)2 − (1 + (2)(2))/(2 1 + 2(2) + (2)2 ) 3 − 5/(2(3)) 13
fx (1, 2, 2) = 2
= = >0
1 + 2(2) + (2) 9 54
so f is increasing.
Next, keeping x, z constant, we have
p p
z x + 2y + z 2 − (x + yz)(2)/(2 x + 2y + z 2 )
fy0 (x, y, z) =
x + 2y + z 2
4. PARTIAL DERIVATIVES 311

At P0 , we have
(2)3 − 5/3 13
fy0 (1, 2, 2) = = >0
9 27
so f is increasing.
Finally, keeping x, y constant, we have
p p
0 y x + 2y + z 2 − (x + yz)(2z)/(2 x + 2y + z 2)
fz (x, y, z) =
x + 2y + z 2
At P0 , we have
2(3) − 5(2)/3 8
fz0 (1, 2, 2) = = >0
9 27
so f is increasing.
312 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

5. Higher-Order Partial Derivatives


1–7. Find all second partial derivatives of each of the following func-
tions and verify Clairauts theorem:
1. f(x, y) = arctan(xy);

Solution: The first partial derivatives are calculated as fol-


lows:
y
fx0 =
1 + x2 y 2
x
fy0 =
1 + x2 y 2

Then the second partial derivatives are

00 −2xy 3
fxx =
(1 + x2y 2 )2
00 −2yx3
fyy =
(1 + x2y 2 )2
00 (1)(1 + x2 y 2) − y(2x2y) 1 − x2 y 2
fxy = (fx0 )0y = =
(1 + x2 y 2)2 (1 + x2y 2 )2
00 (1)(1 + x2 y 2) − x(2xy 2) 1 − x2 y 2
fyx = (fy0 )0x = =
(1 + x2 y 2)2 (1 + x2 y 2)2

00 00
Clairaut’s theorem asserts that if fxy and fyx are continuous on
00 00
a region D then fxy = fyx on D. Evidently, the two are equal
here, since rational functions are continuous everywhere the
denominator is nonzero, and here the denominator is always
greater than or equal to 1. Thus Clairaut’s theorem is verified.

2. f(x, y, z) = x sin(zy 2 );

Solution: The first partial derivatives are calculated as fol-


lows:

fx0 = sin(zy 2)
fy0 = 2xyz cos(zy 2)
fz0 = xy 2 cos(zy 2)
5. HIGHER-ORDER PARTIAL DERIVATIVES 313

Then the second partial derivatives are


00
fxx =0
fyy = 2xz cos(zy 2 ) + 2xyz(2yz)(− sin(zy 2)) = 2xz cos(zy 2) − 4xy 2z 2 sin(zy 2)
00

00
fzz = −xy 4 sin(zy 2)
00
fxy = (fx0 )0y = 2zy cos(zy 2 )
00
fyx = (fy0 )0x = 2yz cos(zy 2 )
00
fxz = (fx0 )0z = y 2 cos(zy 2)
00
fzx = (fz0 )0x = y 2 cos(zy 2)
00
fzy = (fz0 )0y = 2xy cos(zy 2) + xy 2(2yz)(− sin(zy 2)) = 2xy cos(zy 2) − 2xy 3z sin(zy 2)
00
fyz = (fy0 )0z = 2xy cos(zy 2) − 2xy 3 z sin(zy 2)
Clearly all the mixed partial derivatives are compositions of
continuous functions and hence continuous. Therefore Clairaut’s
theorem is verified.

3. f(x, y, z) = x3 + zy + z 2 ;

Solution: The first partial derivatives are calculated as fol-


lows:
fx0 = 3x2
fy0 = z
fz0 = y + 2z
Then the second partial derivatives are
00
fxx = 6x
00
fyy = 0
00
fzz =2
00
fxy = 0
00
fyx =0
00
fxz =0
00
fzx = 0
00
fzy =1
00
fyz =1
Clearly all the mixed partial derivatives are continuous. There-
fore Clairaut’s theorem is verified.
314 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

4. f(x, y, z) = (x + y)/(x + 2z);

Solution: The first partial derivatives are calculated as fol-


lows:

(x + 2z) − (x + y) 2z − y
fx0 = 2
=
(x + 2z) (x + 2z)2
1
fy0 =
x + 2z
2(x + y)
fz0 = −
(x + 2z)2

Then the second partial derivatives are

00 2(y − 2z)
fxx =
(x + 2z)3
00
fyy =0
00 8(x + y)
fzz =
(x + 2z)3
00 1
fxy = −
(x + 2z)2
00 1
fyx = −
(x + 2z)2
00 2(x + 2z)2 − (2z − y)(2)(x + 2z)(2) 2[x + 2z − 2(2z − y)] 2[x + 2y − 2z]
fxz = = =
(x + 2z)4 (x + 2z)3 (x + 2z)3
00 (1)(x + 2z)2 − (x + y)(2)(x + 2z) x + 2z − 2x − 2y 2[x + 2y − 2z]
fzx = −2( 4
) = −2( 3
)=
(x + 2z) (x + 2z) (x + 2z)3
00 2
fzy = −
(x + 2z)2
00 −1(2) 2
fyz = 2
=−
(x + 2z) (x + 2z)2

And Clairaut’s theorem is verified.


p
5. f(x, y) = arccos( x/y);

Solution: Recall that d/du arccos u = −1/ 1 − u2 (this can
be established from the relation arccos u = − arcsin u + π/2.
5. HIGHER-ORDER PARTIAL DERIVATIVES 315

So the first partial derivatives are calculated as follows:


1 √ 1
fx0 = − q (1/(2 xy)) = − p
2 xy − x2
p
1 − ( x/y)2
1 √ x
fy0 = − q (− x/(2y 3/2)) = p
2y xy − x2
p
1 − ( x/y)2
Then the second partial derivatives are
00 1 y − 2x y − 2x
fxx =− 2 3/2
(−1/2) =
2 (xy − x ) 4(xy − x2 )3/2
00 x
fyy = [(y)−2 (−1)(xy − x2)−1/2 + (y)−1(xy − x2)−3/2(−1/2)(x)]
2
x 1 x
= − p ( + )
2y xy − x2 y 2(xy − x2)
00 1 x x
fxy = − (−1/2) =
2 (xy − x2)3/2 4(xy − x2 )3/2
p p
00 1 xy − x2 − x(1/ xy − x2 )(1/2)(y − 2x)
fyx = ( )
2y xy − x2
2xy − 2x2 − xy + 2x2 2(xy − x2 ) − x(y − 2x) x
= = 2 3/2
= 2 3/2
=
4y(xy − x ) 4y(xy − x ) 4(xy − x2)3/2
And Clairaut’s theorem is verified.

6. f(x, y) = xy ;

Solution: The first partial derivatives are calculated as fol-


lows:
fx0 = yxy−1
fy0 = ln(x)xy
Then the second partial derivatives are
00
fxx = y(y − 1)xy−1
00
fyy = ln(x)2 xy
00
fxy = xy−1 + y ln(x)xy−1 = (1 + y ln x)xy−1
00 1
fyx = (xy ) + ln(x)(y)xy−1 = (1 + y ln x)xy−1
x
And Clairaut’s theorem is verified.
316 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

7. f(x1 , x2 , ..., xm) = (a · r)(b · r), where r = hx1 , x2 , ..., xmi and
a and b are constant vectors.

Solution: The first partial derivatives are calculated as fol-


lows:
fx0 i = (a · r)0xi (b · r) + (a · r)(b · r)0xi
= (a0xi · r + a · r0xi )(b · r) + (a · r)(b0xi · r + b · r0xi )
= (a · r0xi )(b · r) + (a · r)(b · r0xi ) = ai(b · r) + bi (a · r)
for i = 1, 2, ..., m. Then the second partial derivatives are
fx00i xj = ai (b · r)0xj + bi (a · r)0xj
= ai (b0xj · r + b · r0xj ) + bi(a0xj · r + a · r0xj )
= ai (b · r0xj ) + bi (a · r0xj ) = ai bj + aj bi
And Clairaut’s theorem is verified since addition is commuta-
tive.

8–12. Show without explicit calculations of higher-order partial deriva-


tives why the hypotheses of Clairauts theorem are satisfied for the fol-
lowing functions:
8. f(x, y, z) = sin(x2 + y − z) cos(xy);

Solution: The only condition necessary for Clairaut’s the-


orem is continuity of the mixed partial derivatives. This can
often be established by seeing if the function in question is a
composition, product, quotient, etc. of continuous functions
with continuous derivatives. Here, f is the product of sine and
cosine, which have continuous derivatives of all orders. By the
chain rule, we must also consider the derivatives of their argu-
ments, which are polynomials and hence continuously differ-
entiable. Thus Clairaut’s theorem will hold.

9. f(x, y) = sin(x + y 2 )/(x2 + y 2 ), x2 + y 2 6= 0;

Solution: Here, f is the quotient of sine and a polyno-


mial. By the chain rule, we must consider the derivative of
the argument of sine, which is a polynomial and hence contin-
uously differentiable. Thus the derivatives are discontinuous
only whenever x2 + y 2 = 0, but these points are not considered
5. HIGHER-ORDER PARTIAL DERIVATIVES 317

in the problem. Thus Clairaut’s theorem will hold.


2 yz
10. f(x, y, z) = ex (y 2 + zx4 );

Solution: Here, f is the product of an exponential function


and a polynomial, which are both continuously differentiable.
Moreover, the argument of the exponential function is a poly-
nomial, which is continuously differentiable. Thus Clairaut’s
theorem will hold.

11. f(x, y) = ln(1 + x2 + y 4 )/(x2 − y 2 ), x2 6= y 2 ;

Solution: Here, f is the quotient of a logarithm and a poly-


nomial. The derivative of the logarithm will not exist whenever
its argument is 0, but this can never occur since the sum of
three positive numbers is positive. Upon differentiating once,
a 1+ x2 + y 4 term will appear in the denominator, but this will
not affect continuity since it is never 0. The denominator only
vanishes when x2 = y 2 , but these points are not considered in
the problem. Thus Clairaut’s theorem will hold.

12. f(x, y, z) = (x + yz 2 − xz 5 )/(1 + x2 y 2 z 4 ).

Solution: Here, f is the quotient of two polynomials and


hence its derivatives are discontinuous only when the denom-
inator vanishes. But, the denominator is the sum of two pos-
itive numbers so it is never 0. Thus Clairaut’s theorem will
hold.

13–20. Find the indicated partial derivatives of each of the following


functions:
13. f(x, y) = xn + xy + y m , fxxy
000 000
, fxyx 000
, fyyx 000
, fxyy ; here n and m are
positive integers;

Solution: Note that f is a polynomial and hence has con-


tinuous derivatives of all orders. Therefore Clairaut’s theorem
will hold, in particular for the mixed third derivatives. Thus
000 000 000 000
fxxy = fxyx and fyyx = fxyy . Then we have
000 000
fxyx = fxxy = ((fx0 )0x )0y = ((nxn−1 + y)0x)0y = (n(n − 1)xn−2 )0y = 0
000 000
fxyy = fyyx = ((fy0 )0y )0x = ((x + ny n−1 )0y )0x = (n(n − 1)y n−2 )0x = 0
318 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

14. f(x, y, z) = x cos(yx) + z 3 , fxyz


000 000
, fxxz 000
, fyyz ;

Solution: We have
000
fxyz = ((fx0 )0y )0z = ((cos(yx) − xy sin(yx))0y )0z = 0
000
fxxz = ((fx0 )0y )0z = ((cos(yx) − xy sin(yx))0x)0z = 0
000
fxxz = ((fx0 )0y )0z = (−x2 sin(yx))0y )0z = 0

000
In the calculation of fxyz , you need only show that fx0 contains
no z term, hence the mixed partials will not either and are
constant with respect to z.
(4) (4) (4)
15. f(x, y, z) = sin(xy)ez , ∂f 5 /∂z 5 , fxyzz , fzyxz , fzxzy ;

Solution: First, f is the product of sine and the expo-


nential function, which have continuous derivatives of all or-
ders. Moreover, the argument of the sine part is a polynomial
and also has continuous derivatives of all orders. Therefore
(4) (4) (4)
Clairaut’s theorem will hold and fxyzz = fzyxz = fzxzy . Next,
note that f(x, y, z) can be written as f(x, y, z) = g(x, y)h(z);
(4)
this will make our calculations easier. For example, fxyzz =
00 00
gxy hzz , since h is constant with respect to x, y and g is constant
with respect to z. Therefore we have

∂f 5
5
= g(x, y)h(5)
zzzzz = sin(xy)e
z
∂z
(4) (4) (4) 00 00
fzyxz = fzxzy = fxyzz = gxy hzz = (y cos(xy))0y ez = (cos(xy) − xy sin(xy))ez

(4)
16. f(x, y, z, t) = sin(x + 2y + 3z − 4t), fabcd where abcd denotes
all permutations of xyzt;

Solution: Observe that f is the composition of sine and


a polynomial and hence will have continuous derivatives of all
order. Therefore Clairaut’s theorem will hold, and differenti-
ation order does not matter (that is, permuting xyzt in the
mixed fourth derivative will not change anything). Therefore
we have
(4) 4
fabcd = ftzyx = (((−4 cos(x + 2y + 3z − 4t))0z )0y )0x = ((4(3) sin(x + 2y + 3z − 4t))0y )0x
= (4(3)(2) cos(x + 2y + 3z − 4t))0x = −4! sin(x + 2y + 3z − 4t)
5. HIGHER-ORDER PARTIAL DERIVATIVES 319

4
where I have chosen ftzyx arbitrarily to compute the fourth
mixed partial.
(4)
17. f(x, y) = exy (y 2 +x), fabcd where abcd denotes all permutations
of xxxy;

Solution: Note that f is the product of an exponential func-


tion whose argument is a polynomial and a polynomial; there-
fore its partial derivatives will be continuous and Clairaut’s
(4)
theorem will hold. Thus fabcd will be the same for any permu-
tation of xxxy. So we have that
(4) (4)
fabcd = fxxxy = (((yexy (y 2 + x) + exy )0x )0x )0y = (((yexy (y 2 + x + 1/y))0x )0x )0y
= ((y 2 exy (y 2 + x + 2/y))0x )0y = (y 3 exy (y 2 + x + 3/y))0y = (y 2exy (y 3 + xy + 3))0y
= 2yexy (y 3 + xy + 3) + xy 2exy (y 3 + xy + 3) + y 2exy (3y 2 + x)
= [2(y 3 + xy + 3) + xy(y 3 + xy + 3) + y(3y 2 + x)][yexy ]
= [2y 3 + 2xy + 6 + xy 4 + x2y 2 + 3xy + 3y 3 + xy][yexy]
= [x2y 2 + xy 4 + 6xy + 5y 3 + 6][yexy ]
(3)
18. f(x, y, z) = arctan((x + y + z − xyz)/(1 − xy − yz − xz)), fabc
where abc denotes all permutations of xyz;

Solution: Note that f will satisfy Clairaut’s theorem since


the derivative of arctangent gives a rational function with non-
vanishing denominator). Therefore we have
(3) 1 (1 − yz)(1 − xy − yz − xz) − (−y − z)(x + y + z − xyz) 0 0
fabc = ((( (x+y+z−xyz)2
)( ))y )z
1+ (1 − xy − yz − xz)2
(1−xy−yz−xz)2
(1 − yz)(1 − xy − yz − xz) + (y + z)(x + y + z − xyz) 0 0
= (( )y )z
(1 − xy − yz − xz)2 + (x + y + z − xyz)2
1 + y 2z 2 + y 2 + z 2 1
= (( )0y )0z = (( )0 )0 = 0
2 2 2 2
(1 + y z + y + z )(1 + x ) 2 1 + x2 y z
There is a lot of tedious algebra here that I will skip.
(4)
19. f(x, y, z, t) = ln((x − y)2 + (z − t)2 )−1/2 , fabcd where abcd de-
notes all permutations of xyzt;

Solution: Note that f will satisfy Clairaut’s theorem ex-


cept when x = y and z = t. Apart from that case, we have
(4)
that fabcd is independent of the permutation. To simplify the
320 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

problem, I will write f(x, y, z, t) = (ln(g(x, y, z, t)))−1/2 where


g(x, y, z, t) = (x − y)2 + (z − t)2. Hence

(4) (4) 1 g0
fabcd = fxyzt = (((− ln(g)−3/2 x )0y )0z )0t
2 g
0 0
3 g g
−5/2 x y 1 −3/2 xy
g 00 g − gx0 gy0 0 0
= (( ln(g) − ln(g) [ ])z )t
4 g2 2 g2

20. f(x, y) = ex sin y, ∂ n+m f/(∂ n x∂ m y)(0, 0), where n and m are
positive integers.

Solution: Observe that f will satisfy Clairaut’s theorem and


f may be written as f(x, y) = g(x)h(y) where g(x) = ex and
h(y) = sin y. Hence ∂ n f/∂ n x = g (n) (x)h(y) and ∂ m f/∂ m y =
g(x)h(m) (y). Therefore

∂ n+m f/(∂ n x∂ my) = g (n) (x)h(m) (y) = exh(m) (y)

we must now break this down into three cases:


Case 1: m is one greater than a multiple of 4 (that is, m
mod 4 ≡ 1). Then h(m) (y) = cos y, and h(m)(0) = cos 0 = 1
Case 2: m is three greater than a multiple of 4 (that is, m
mod 4 ≡ 3). Then h(m) (y) = − cos y, and h(m) (0) = − cos 0 =
−1
Case 3: m is even. Then h(m) (y) = ± sin y, and h(m)(0) =
± sin 0 = 0
Therefore,

1
 m mod 4 ≡ 1
n+m n m
∂ f/(∂ x∂ y)(0, 0) = −1 m mod 4 ≡ 3

0 m even

21–25. Given partial derivatives, find the function or show that it does
not exist:
21. fx0 = 3x2 y, fy0 = x3 + 3y 2 ;

Solution: We first test to see if the function may exist by


seeing if its mixed partial derivatives are equivalent; this is
known as the integrability condition. So we have
00
fxy = (3x2 y)0y = 3x2 , fyx
00
= (x3 + 3y 2 )0x = 3x2
5. HIGHER-ORDER PARTIAL DERIVATIVES 321

00 00
Hence fxy = fyx and a function f does exist. Thus we have
that Z
f(x, y) = fx0 dx = x3y + g(y)

where g(y) is used since it is a constant with respect to x (it


is thus easily verified that (x3y + g(y))0x = 3x2y + 0 = fx0 , as
expected). However, we must also have that
x3 + 3y 2 = fy0 = (x3y + g(y))0y = x3 + g 0 (y)
where we have calculated fy0 in two different ways. Thus we
have
x3 + 3y 2 = x3 + g 0 (y) ⇔ g 0 (y) = 3y 2 ⇔ g(y) = y 3 + C
Note the cancellation of the x3 term; this is necessary since g
must be a function of y only. Had the integrability conditions
not been met, the terms containing x may not vanish and thus
we would have a contradiction. Therefore f is
f(x, y) = x3y + y 3 + C
for arbitrary C. Note that one may have also started by inte-
grating fy0 . The choice of fx0 was arbitrary.

22. fx0 = yz + 3x2 , fy0 = xz + 4y, fz0 = xy + 1;

Solution: We first test to see if the function may exist by


seeing if its mixed partial derivatives are equivalent. So we
have
00
fxy = (yz + 3x2 )0y = z, fyx
00
= (xz + 4y)0x = z
00
fxz = (yz + 3x2 )0z = y, fzx00
= (xy + 1)0x = y
00
fyz = (xz + 4y)0z = x, fzy00
= (xy + 1)0y = x
So the mixed partial derivatives are equal and a function f
does exist. Thus we have that
Z
f(x, y, z) = fx0 dx = xyz + x3 + g(y, z)

where g(y, z) is used since it is a constant with respect to x.


However, we must also have that
xz + 4y = fy0 = (xyz + x3 + g(y, z))0y = xz + gy0 (y, z)
322 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

where we have calculated fy0 in two different ways. Thus we


have
xz + 4y = xz + gy0 (y, z) ⇔ gy0 (y, z) = 4y ⇔ g(y, z) = 2y 2 + h(z)
We must also have
xy + 1 = fz0 = (xyz + x3 + 2y 2 + h(z))0z = xy + h0(z)
Thus we have
xy + 1 = xy + h0 (z) ⇔ h0 (z) = 1 ⇔ h(z) = z + C
Therefore f is
f(x, y, z) = xyz + x3 + 2y 2 + z + C
for arbitrary C.

23. fx0 k = kxk , k = 1, 2, ..., m;

Solution: We first test to see if the function may exist by


seeing if its mixed partial derivatives are equivalent. So we
have
fx00i xj = (ixi)0xj = 0, fx00j xi = (jxj )0xi = 0
for i, j = 1, 2, ..., m with i 6= j. So the mixed partial derivatives
are equal and a function f does exist. Thus we have that
1
Z
f(x1 , x2, ..., xm) = fx0 1 dx1 = x21 + g(x2, x3 , ..., xm)
2
However, we must also have that
1
2x2 = fx0 2 = ( x21 + g(x2, x3 , ..., xm))0x2 = gx0 2 (x2 , x3, ..., xm)
2
Thus we have
2
2x2 = gx0 2 (x2, x3 , ..., xm) ⇔ gx0 2 (x2, x3, ..., xm)) = 2x2 ⇔ g(x2 , x3, ..., xm) = x22 +h(x3, ..., xm)
2
Continuing this process, we will have that f is
m
X k
f(x1 , x2, ..., xm) = x2k + C
k=1
2

for arbitrary C.
5. HIGHER-ORDER PARTIAL DERIVATIVES 323

24. fx0 = xy + z, fy0 = x2 /2, fz0 = x + y;

Solution: We first test to see if the function may exist by


seeing if its mixed partial derivatives are equivalent. So we
have
00
fxy = (xy + z)0y = 0, fyx
00
= (x2 /2)0x = x
which are not equivalent, so f does not exist.

25. fx0 = sin(xy) + xy cos(xy), fy0 = x2 cos(xy) + 1.

Solution: We first test to see if the function may exist by


seeing if its mixed partial derivatives are equivalent; this is
known as the integrability condition. So we have
00
fxy = (sin(xy) + xy cos(xy))0y = x cos(xy) + x cos(xy) − x2 y sin(xy) = 2x cos(xy) − x2y sin(xy)
00
fyx = (x2 cos(xy) + 1)0x = 2x cos(xy) − x2y sin(xy)
00 00
Hence fxy = fyx and a function f does exist. Thus we have
that
Z
f(x, y) = (x2 cos(xy) + 1)dy = x sin(xy) + y + g(x)

where g(x) is used since it is a constant with respect to y.


However, we must also have that
sin(xy)+xy cos(xy) = fx0 = (x sin(xy)+y+g(x))0x = sin(xy)+xy cos(xy)+g 0(x)
where we have calculated fx0 in two different ways. Thus we
have
sin(xy)+xy cos(xy) = sin(xy)+xy cos(xy)+g 0(x) ⇔ g 0 (x) = 0 ⇔ g(y) = C
Therefore f is
f(x, y) = x sin(xy) + y + C
for arbitrary C.

26–32. Verify that a given function is a solution of the indicated dif-


ferential equation:
26. f(t, x) = A sin(ct − x) + B cos(ct + x), c−2 ftt00 − fxx
00
= 0 if A,
B, and c are constants;
324 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

Solution: I will first calculate the necessary partial deriva-


tives as follows:
ftt00 = (Ac cos(ct − x) − Bc sin(ct + x))0t = −Ac2 sin(ct − x) − Bc2 cos(ct + x) = −c2 f
00
fxx = (−A cos(ct − x) − B sin(ct + x))0x = −A sin(ct − x) − B cos(ct + x) = −f
Then we have that
c−2 ftt00 − fxx
00
= c−2 (−c2f) − (−f) = f − f = 0
27. f(x, y, t) = g(ct − ax − by) + h(ct + ax + by), ftt00 = c2 (fxx
00 00
+ fyy )
if a, b, and c are constants such that a2 + b2 = 1, and g and h
are twice differentiable functions;

Solution: I will first calculate the necessary partial deriva-


tives as follows. Note that gtt00 /u00tt = gxx
00
/u00xx = gyy
00
/u00yy where
u = ct − ax − by; A similar argument holds for h (with
v = ct + ax + by). This is because we may rewrite g as g(u),
00
and hence gkk = g 00 (u)u00kk by the chain rule and the fact that
u0k is independent of k for k = x, y, t.
ftt00 = (cg 0 (u) + ch0 (v))0t = c2 g 00 (u) + c2 h00(v) = c2 (g 00(u) + h00(v))
00
fxx = (−ag 0 (u) + ah0 (v))0x = a2g 00(u) + a2h00 (v) = a2(g 00(u) + h00(v))
00
fyy = (−bg 0 (u) + bh0(v))0y = b2g 00(u) + b2 h00(v) = b2(g 00 (u) + h00 (v))
Then we have that
c2 (fxx
00 00
+fyy ) = c2 ([a2 +b2]g 00(u)+[a2 +b2]h00(v)) = c2(g 00 (u)+h00(v)) = ftt00
where we have applied the assumption that a2 + b2 = 1.

28. f(x, y) = ln(x2 + y 2 ), fxx


00 00
+ fyy = 0;

Solution: I will first calculate the necessary partial deriva-


tives as follows:
00 2x 0 2(x2 + y 2 ) − 2x(2x) 2y 2 − 2x2
fxx =( ) = =
x2 + y 2 x (x2 + y 2)2 (x2 + y 2)2
00 2y 0 2(x2 + y 2) − 2y(2y) 2x2 − 2y 2
fyy =( 2 ) = =
x + y2 y (x2 + y 2 )2 (x2 + y 2)2
Then we have that
2y 2 − 2x2 2x2 − 2y 2 2x2 − 2x2 + 2y 2 − 2y 2
+ = =0
(x2 + y 2)2 (x2 + y 2 )2 (x2 + y 2)2
5. HIGHER-ORDER PARTIAL DERIVATIVES 325

29. f(x, y) = ln(ex + ey ), fx0 + fy0 = 1 and fxx


00 00 00 2
fyy − (fxy ) = 0;

Solution: I will first calculate the necessary partial deriva-


tives as follows:
ex
fx0 =
ex + ey
00 ex 0 ex (ex + ey ) − ex(ex ) ex ey
fxx = ( x ) = =
e + ey x (ex + ey )2 (ex + ey )2
ey
fy0 =
ex + ey
00 ey 0 ey (ex + ey ) − ey (ey ) ex ey
fyy = ( x ) y = =
e + ey (ex + ey )2 (ex + ey )2
00 ex 0 (0)(ex + ey ) − ex (ey ) ex ey
fxy = ( x ) = = −
e + ey y (ex + ey )2 (ex + ey )2

Then we have that


ex ey ex + ey
+ = = 1,
ex + ey ex + ey ex + ey

ex ey ex ey ex ey 2 e2xe2y − e2xe2y
( x )( ) − (− x ) = =0
(e + ey )2 (ex + ey )2 (e + ey )2 (ex + ey )2
30. f(r) = exp(a · r), where a · a = 1 and r ∈ Rm , fx001 x1 + fx002 x2 +
... + fx00m xm = f;

Solution: I will first calculate the necessary partial deriva-


tives as follows:

fx00i xi = (exp(a · r)[a0xi · r + a · r0xi ])0xi = ai(exp(a · r))0xi = a2i exp(a · r)

for i = 1, 2, ..., m. Note that exp(x) = ex. Then we have that


m
X
a21 exp(a·r)+a22 exp(a·r)+...+a2m exp(a·r) = exp(a·r) a2k = [exp(a·r)](a·a) = f
k=1

where I have used the assumption that a · a = 1.

31. f(r) = krk2−m where r ∈ Rm , fx001 x1 + fx002 x2 + ... + fx00m xm = 0


if krk =
6 0;
326 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

Solution: I will first calculate the


√ necessary partial deriva-
tives as follows. Note that krk = r · r. Thus f = (r · r)1−m/2

fx00i xi = ((1 − m/2)(r · r)−m/2[r0xi · r + r · r0xi ])0xi = ((2 − m)xi (r · r)−m/2)0xi


(2 − m)[r · r − mx2i ]
= (2 − m)[(r · r)−m/2 + xi (−m/2)(r · r)−m/2−1 (2xi )] =
(r · r)m/2+1

for i = 1, 2, ..., m. Then we have that

(2 − m)[r · r − mx21] (2 − m)[r · r − mx22] (2 − m)[r · r − mx2m ]


+ + ... +
(r · r)m/2+1 (r · r)m/2+1 (r · r)m/2+1
m m m
2−m X 2 2−m X X
= m/2+1
[r · r − mxi ] = m/2+1
[r · r (1) − m x2i ]
(r · r) (r · r)
k=1 k=1 k=1
2−m
= [mr · r − mr · r] = 0
(r · r)m/2+1

00 00 00
32. f(x, y, z) = sin(kkrk)/krk, fxx +fyy +fzz +k 2f = 0 (Helmholtz
equation), where k is a constant.

Solution: First define g(x, y, z) = krk. Then f = sin(kg)/g.


Note that gu0 = u/g and gu00 = (g−ugu0 )/g 2 = (g 2 −u2)/g 3 where
u = x, y, z. The necessary partial derivatives are as follows:

(kgu0 cos(kg))g − (sin(kg))gu0 ku cos(kg)g − u sin(kg) ku cos(kg) u sin(kg)


fu0 = 2
= 3
= −
g g g2 g3
00 [k cos(kg) − k 2 ugu0 sin(kg)]g 2 − [ku cos(kg)](2ggu0 )
fuu =
g4
[sin(kg) + kugu cos(kg)]g 3 − [u sin(kg)](3g 2 gu0 )
0

g6
[kg cos(kg) − k 2 u2 sin(kg)] − [2ku2 cos(kg)](1/g)
=
g3
g 3 sin(kg) + ku2 g 2 cos(kg) − 3u2 g sin(kg)

g6
k cos(kg) k 2u2 sin(kg) 2ku2 cos(kg) sin(kg) ku2 cos(kg) 3u2 sin(kg)
= − − − − +
g2 g3 g4 g3 g4 g5
k cos(kg) −(k 2u2 + 1) sin(kg) 3ku2 cos(kg) 3u2 sin(kg)
= + − +
g2 g3 g4 g5
5. HIGHER-ORDER PARTIAL DERIVATIVES 327

Then we have that


k cos(kg) −(k 2 (x2 + y 2 + z 2 ) + 3) sin(kg) 3k(x2 + y 2 + z 2) cos(kg)
3( )+ −
g2 g3 g4
3(x2 + y 2 + z 2) sin(kg)
+ 5
+ k2f
g
3k cos(kg) −(k 2 g 2 + 3) sin(kg) 3kg 2 cos(kg) 3g 2 sin(kg)
= + − + + k2f
g2 g3 g4 g5
3k cos(kg) 3k cos(kg) k 2 sin(kg) 3 sin(kg) 3 sin(kg)
= − − − + + k 2 f = −k 2 f + k 2 f = 0
g2 g2 g3 g3 g3
where the three copies of k cos(kg)/g 2 and − sin(kg)/g 3 came
00 00 00
from each of fxx , fyy , and fzz .

33. Find a relation between constants a, b and c such that the function
u(x, y, t) = sin(ax+by +ct) satisfies the wave equation u00tt −u00xx −u00yy =
0. Give a geometrical description of such a relation, e.g., by setting val-
ues of c on a vertical axis, and the values of a and b on two horizontal
axes

Solution: We start by calculating the necessary partial derivatives


u00tt = (c cos(ax + by + ct))0t = −c2 sin(ax + by + ct) = −c2u
u00xx = (a cos(ax + by + ct))0x = −a2 sin(ax + by + ct) = −a2u
u00yy = (b cos(ax + by + ct))0y = −b2 sin(ax + by + ct) = −b2u
We then require that u satisfies the wave equation, so
−c2u − (−a2u) − (−b2 u) = 0 ⇔ (−c2 + a2 + b2 )(u) = 0
The only way for this to be guaranteed to be equal to 0 for all u is if
−c2 + a2 + b2 = 0 ⇔ c2 = a2 + b2, which is a double cone.
000
34. Let f(x, y, z) = u(t) where t = xyz. Show that fxyz = F (t)
and find F (t) in terms of u(t).

Solution: Applying the chain rule we have


000
fxyz = ((u0 (t)t0x)0y )0z = ((u0(t)yz)0y )0z = (u00(t)t0y yz + u0(t)z)0z = (u00(t)xyz 2 + u0(t)z)0z
= u000(t)t0z xyz 2 + u00(t)xy(2z) + u00(t)t0z z + u0(t) = u000(t)x2y 2z 2 + 3u00(t)xyz + u0 (t)
= t2 u000(t) + 3tu00(t) + u0 (t)
So F (t) = t2u000(t) + 3tu00(t) + u0 (t).
328 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

35–36. Find (fx0 )2 + (fy0 )2 + (fz0 )2 and fxx


00 00
+ fyy 00
+ fzz for each of the
following functions:
35. f = x3 + y 3 + z 3 − 3xyz;

Solution: The necessary partial derivatives are

fx0 = 3x2 − 3yz ⇒ (fx0 )2 = 9x4 + 9y 2 z 2 − 18x2 yz


00
fxx = 6x
fy0 = 3y 2 − 3xz ⇒ (fy0 )2 = 9y 4 + 9x2z 2 − 18xy 2 z
00
fyy = 6y
fz0 = 3z 2 − 3xy ⇒ (fz0 )2 = 9z 4 + 9x2 y 2 − 18xyz 2
00
fzz = 6z

Therefore we have that

(fx0 )2+(fy0 )2 +(fz0 )2 = 9(x4 +y 4+z 4)+9(y 2 z 2 +x2z 2 +x2y 2)−18xyz(x+y+z)

and
00 00 00
fxx + fyy + fzz = 6(x + y + z)

36. f = (x2 + y 2 + z 2 )−1/2 .

Solution: Note that this is actually just a special case of


Exercise 31 with m = 3. So the sum of the second partial
derivatives will be 0. Thus we need only calculate the first
partial derivatives

fu0 = (2 − 3)u(x2 + y 2 + z 2)−3/2 = −u(x2 + y 2 + z 2)−3/2

And so,
1
(−x(x2+y 2+z 2)−3/2 )2+(−y(x2+y 2+z 2 )−3/2)2 +(−z(x2+y 2+z 2 )−3/2)2 =
x2 + y 2 + z 2

37–38. Let the action of K on a function f be defined by Kf =


xfx0 + yfy0 . Find Kf, K 2f = K(Kf), and K 3 f = K(K 2 f) if
37. f = x/(x2 + y 2 );

Solution: In order to calculate K 3f, we probably need to


calculate many partial derivatives and, really, who wants to
5. HIGHER-ORDER PARTIAL DERIVATIVES 329

do that? So first we will find Kf and see if anything interest-


ing happens.
x2 + y 2 − x(2x) y 2 − x2
fx0 = =
(x2 + y 2)2 (x2 + y 2 )2
−x(2y) 2xy
fy0 = 2 2 2
=− 2
(x + y ) (x + y 2)2
Thus Kf is
y 2 − x2 2xy x3 + xy 2 x
Kf = x( 2 2 2
) + y(− 2 2 2
) = − 2 2 2
=− 2
(x + y ) (x + y ) (x + y ) x + y2
So we have that Kf = −f. Therefore K 2f = K(Kf) =
K(−f) = −Kf = f and K 3 f = K(K 2 f) = K(f) = −f.

38. f = 1/2 ln(x2 + y 2 )

Solution: As above, we first calculate Kf.


x
fx0 = 2
x + y2
y
fy0 = 2
x + y2
Thus Kf is
x y
Kf = x( ) + y( 2 )=1
x2 +y 2 x + y2
So we have that Kf = 1. Therefore K 2 f = K(Kf) = K(1) =
0 and K 3 f = K(K 2 f) = K(0) = 0.

39. Let f(x, y) = xy 2 /(x2 + y 2 ) if (x, y) 6= (0, 0) and f(0, 0) = 0. Do


00 00
fxy (0, 0) and fyx (0, 0) exist?

Solution: To tackle this problem, find the first partial derivatives


and then use the identity gx0 = d/dxg(x, 0)|x=0 with g = fy0 to find
00
fyx (0, 0) and vice verse. So we have

y 2 (x2 + y 2 ) − xy 2(2x) y 4 − x2 y 2
fx0 = = 2
(x2 + y 2)2 (x + y 2 )2
2xy(x2 + y 2) − xy 2(2y) 2x3 y
fy0 = =
(x2 + y 2)2 (x2 + y 2)2
330 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

00
Then fyx (0, 0) is

00 d 0 d
fyx (0, 0) = (fy (x, 0)) = (0) = 0
dx dx
00
and fxy (0, 0) is

00 d 0 d y4 d y4
fxy (0, 0) = (fx (0, y)) = ( )=
dy dy (0 + y 2)2 dy y 4
which is actually not differentiable at y = 0 since fx0 (0, y) = 1 if y 6= 0.

40–43. If f = f(x, y) and g = g(x, y, z), find the most general so-
lution to each of the following equations:
00
40. fxx = 0;
00
Solution: Suppose a function f satisfies fxx = 0. Then
Z Z
fx0 = fxx00
dx = g(y), f(x, y) = fx0 dx = xg(y) + h(y)

for functions g(y), h(y).


00
41. fxy = 0;
00
Solution: Suppose a function f satisfies fxy = 0. Then
Z Z
fx = fxy dy = g (x), f(x, y) = fx0 dx = g(x) + h(y)
0 00 0

for once differentiable functions g(x), h(y); note that g 0 (x) is a


constant with respect to y and h(y) is a constant with respect
to x, hence both act like ”+C” in their respective integrals.

42. ∂ n f/∂y n = 0;

Solution: Suppose a function f satisfies ∂ n f/∂y n = 0. Then


Z Z
n−1 n−1 n n
∂ f/∂y = ∂ f/∂y dy = 0 dy = gn (x)
Z Z
n−2 n−2 n−1 n−1
∂ f/∂y = ∂ f/∂y dy = gn (x) dy = ygn (x) + gn−1 (x)

Eventually we reach f(x, y) = y n−1 gn (x) + y n−2 gn−1 (x) + ... +


g1 (x) for functions g1 (x), g2(x), ..., gn(x) (which are treated like
5. HIGHER-ORDER PARTIAL DERIVATIVES 331

constants with respect to y). Also note that we may incorpo-


rate any constant coefficient (like the factorials) into the func-
tions themselves. Note that the highest power of y must be
n − 1; if it were n, then the nth partial derivative would just
be the coefficient of the y n term.
000
43. gxyz = 0.
000
Solution: Suppose a function g satisfies gxyz = 0. Then
Z Z
00 000 00
gxy = gxyz dz = 0 dz = fxy (x, y)

And so
Z
gx0 = 00
fxy (x, y)dy = fx0 (x, y) + p0x (x, z)

since p0x (x, z) is constant with respect to y. Finally,


Z
g(x, y, z) = (fx0 (x, y) + p0x (x, z))dx = f(x, y) + p(x, z) + h(y, z)

for functions f(x, y), p(x, z), and h(y, z).

44–46. Find f(x, y) that satisfies the given conditions:


44. fy0 = x2 + 2y, f(x, x2 ) = 1;

Solution: First we have that


Z
f(x, y) = fy0 dy = x2 y + y 2 + g(x)

Using the second condition, we have


x2(x2 ) + (x2)2 + g(x) = 1 ⇒ g(x) = 1 − 2x4
Hence g(x) = 1 − 2x2 and
f(x, y) = x2y + y 2 + 1 − 2x2 = x2y + y 2 − 2x4 + 1
00
45. fyy = 4, f(x, 0) = 2, fy0 (x, 0) = x;

Solution: We have
Z
0 00
fy = fyy dy = 4y + g(x)

Using the third condition,


fy0 (x, 0) = 4(0) + g(x) = x ⇒ g(x) = x
332 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

and so fy0 = 4y + x. It follows that


Z
f(x, y) = fy0 dy = 2y 2 + xy + h(x)

With the second condition,


f(x, 0) = 2(0)2 + x(0) + h(x) = 2 ⇒ h(x) = 2
and so
f(x, y) = 2y 2 + xy + 2
00
46. fxy = x + y, f(x, 0) = x, f(0, y) = y 2.

Solution: With the first condition,


Z Z
fx0 = 00
fxy 2 0
dy = xy+1/2y +g (x) ⇒ f(x, y) = fx0 dx = 1/2x2 y+1/2y 2 x+g(x)+h(y)

for functions g(x), h(y). By the second and third conditions,


f(x, 0) = 1/2x2 (0) + 1/2(0)2 x + g(x) + h(0) = x ⇒ g(x) + h(0) = x
f(0, y) = 1/2(0)2 y + 1/2y 2 (0) + g(0) + h(y) = y 2 ⇒ g(0) + h(y) = y 2
Now, using the first equation we have g(0) + h(0) = 0. Adding
both equations gives g(x) + g(0) + h(0) + h(y) = x + y 2 ⇒
g(x) + 0 + h(y) = x + y 2 . Thus,
xy
f(x, y) = (x + y) + x + y 2
2
6. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS 333

6. Differentiation of Multivariable Functions


1. Let g(x, y) = 4 − x2 − y 2 . Sketch the graph z = g(x, y) for the
disk x2 + y 2 ≤ 4. Find gx0 (1, 1) and gy0 (1, 1). Consider the plane
z = g(1, 1) + gx0 (1, 1)(x − 1) + gy0 (1, 1)(y − 1) = L(x, y). Is this plane a
good linear approximation to the function near (1, 1)? Is the function
differentiable at (1, 1)?

Solution: Sketching the graph is easy enough; it is an elliptic pa-


raboloid concave down from 0 ≤ z ≤ 4 (owing to the fact that when
x2 + y 2 = 4, z = 4 − 4 = 0).

The partial derivatives of g are


gx0 (x, y) = −2x ⇒ gx0 (1, 1) = −2
gy0 (x, y) = −2y ⇒ gy0 (1, 1) = −2
Let L(x, y) = 4 − 12 − 12 − 2(x − 1) − 2(y − 1) = −2x − 2y + 6. To
see if L(x, y) is a good linear approximation to g near (1, 1), we must
investigate the limit
f(r) − L(r)
L = lim
r→r0 kr − r0 k

where r0 = h1, 1i here. So we have


4 − x2 − y 2 − (−2x − 2y + 6) −(x2 − 2x + 1) − (y 2 − 2y + 1)
L = lim p = lim p
r→r0 (x − 1)2 + (y − 1)2 r→r0 (x − 1)2 + (y − 1)2
−(x − 1)2 − (y − 1)2 p
= lim p = lim − (x − 1)2 + (y − 1)2 = 0
r→r0 (x − 1)2 + (y − 1)2 r→r0
Since the limit is 0, a good linear approximation exists. Hence g is
differentiable at (1, 1). This can also be concluded by noting that the
partial derivatives of g are continuous at (1, 1).
334 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

2. Let f(x, y) = 4 − x2 − y 2 if (x, y) is not in D = {(x, y)| x <


y < 2x − 1, x > 1} and f(x, y) = 0 if (x, y) is in D. Sketch D and
the graph z = f(x, y) for the disk x2 + y 2 ≤ 4. Use the definition
of partial derivatives to find fx0 (1, 1) and fy0 (1, 1). Consider the plane
z = f(1, 1) + fx0 (1, 1)(x − 1) + fy0 (1, 1)(y − 1) = L(x, y). Is this plane a
good linear approximation to the function near (1, 1)? Is the function
differentiable at (1, 1)? Hint: Investigate the difference f(x, y)−L(x, y)
along straight lines through the point (1, 1).

Solution: D is the set of (x, y) between the lines y = x and y = 2x−1


for x > 1 (note that the two lines intersect at x = 1). A sketch is shown
below, along with an approximate sketch of f (the one produced has a
slice cut out extending to (0, 0), but this is not actually the case).

The point (1, 1) is not in D (since x > 1 for all points in D), there-
fore the partial derivatives are the same as those in Exercise 1. So
L(x, y) = −2x − 2y + 6. The plane is not a good linear approximation
to f near (1, 1). Indeed, consider L(1 + ∆x, 1 + ∆y) = −2 − ∆x − 2 −
∆y + 6 = 2 − ∆x − ∆y for ∆x, ∆y > 0, whereas f(1 + ∆x, 1 + ∆y) = 0.
Consider a straight line path through (1, 1) given by x(t) = t + 1 and
y(t) = at + 1 where 1 < a < 2. This defines a line completely in D for
t > 0. Then f(x(t), y(t)) = 0 for t > 0, and
f(x(t), y(t)) − L(x(t), y(t)) 0 − (−2(t + 1) − 2(at + 1) + 6)
lim p = lim √
t→0+ (t + 1 − 1)2 + (at + 1 − 1)2 t→0+ |t| 1 + a2
2(1 + a)t − 2
= lim √
t→0+ |t| 1 + a2

does not exist, thus f is not differentiable at (1, 1).

3. Let f(x, y) = xy 2 if (x, y) 6= (0, 0) and f(0, 0) = 1 and let g(x, y) =


6. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS 335

p p
|x| + |y|. Are the functions f and g differentiable at (0, 0)?

Solution: Suppose f is differentiable at (0, 0). Then it is continu-


ous at (0, 0). However, it is obvious that f is not continuous there (if f
was, then f(0, 0) = 0, rather than 1). Therefore f is not differentiable.
Consider gx0 . By the chain rule, we have that
p
0 1 0 1 |x| |x|
gx = p (|x|) = p ( ) =
2 |x| 2 |x| x 2x

where we have p used the fact that |x|0 = |x|/x. Because g(x, y) =
g(y, x), gy0 = |y|/(2y). Then L(x, y) is
p p
|x| |y| 1 p p 1
L(x, y) = g(0, 0)+ (x−0)+ (y−0) = ( |x|+ |y|) = g(x, y)
2x 2y 2 2
Then,
p p
g(x, y) − L(x, y) 1 |x| + |y|
p = p
x2 + y 2 2 x2 + y 2
By the squeeze principle,

g(x, y) − L(x, y) 1 R1/2 + R1/2 1


| p − 0| ≤ = 1/2
2
x +y 2 2 R R

which does not tend to 0; hence the limit is not 0 (if it even exists), and
therefore g is not differentiable. This is reaffirmed by the observation
that gx0 is not continuous at (0, 0).

4. Let f(x, y) = xy 2 /(x2 + y 2 ) if x2 + y 2 6= 0 and f(0, 0) = 0. Show


that f is continuous and has bounded partial derivatives fx0 and fy0 , but
that f is not differentiable at (0, 0). Investigate continuity of partial
derivatives near (0, 0).

Solution: To show that f is continuous, we need to show that as


(x, y) → (0, 0), f(x, y) → 0. By the squeeze principle,

|x||y|2 R3
|f(x, y) − 0| = 2 ≤ 2 =R
x + y2 R
which goes to 0 as R goes to 0. Hence the limit is indeed 0, and f is
continuous.
336 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

The partial derivatives are


y 2 (x2 + y 2 ) − xy 2(2x) y 2(y 2 − x2 )
fx0 = =
(x2 + y 2)2 (x2 + y 2)2
2xy(x2 + y 2) − xy 2(2y) 2x3 y
fy0 = =
(x2 + y 2)2 (x2 + y 2)2
Clearly the numerator of fx0 is maximized when x2 is minimized (be-
cause of the y 2 − x2 factor). x2 is minimized when x = 0. Setting
x = 0, we have that fx0 (0, y) = 1. Thus fx0 is bounded above by 1. This
is shown with the below inequality
y 2(y 2 − x2 )
≤ 1 ⇒ y 4 − x2y 2 ≤ x4 + 2x2 y 2 + y 4 ⇒ 0 ≤ x4 + 3x2 y 2
(x2 + y 2 )2
which holds for all (x, y), since the sum of two positive numbers is
positive. In fact, 1 is the supremum of fx0 over R2 . √
It can also be shown that fy0 is bounded, where the supremum is 3 3/8.
As a hint, let x = r cos θ and y = r sin θ. Then
r3 cos θ(r sin θ)
fy0 (x, y) = 2 2
= cos3 θ sin θ
(r )
Owing to the fact that cos θ, sin θ < 1, fy0 (x, y) < 1.
Next we will investigate the continuity of fx0 at (0, 0). First let x = 0.
Then, as shown, fx0 (0, y) = 1. However, let y = 0. Then fx0 (x, 0) = 0.
Therefore fx0 is not continuous.
Finally, investigate the continuity of fy0 at (0, 0). Let x = t and y = at.
Then
2t3(at) 2a
fy0 (t, at) = 2 2 2 2
=
(t + a t ) (1 + a2)2
so the limit will certainly depend on a, and thus fy0 is not continuous
at (0, 0). Therefore f is not differentiable there either.
p
5. Show that the function f(x, y) = |xy| is continuous at (0, 0)
and has the partial derivatives fx0 (0, 0) and fy0 (0, 0), but that f is not
differentiable at (0, 0). Investigate continuity of the partial derivatives
near (0, 0).

Solution: f(x, y) is the composition of continuous functions, so f


is continuous.
By definition we have
d d
fx0 (0, 0) = f(x, 0)|x=0 = (0)|x=0 = 0
dx dx
6. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS 337

Similarly, fy0 (0, 0) = 0. So L(x, y) = 0. Therefore we have


p
f(x, y) − L(x, y) |xy|
L = lim p = lim p
r→0 x2 + y 2 r→0 x2 + y 2
Consider the path x(t) = t, y(t) = a2t. Then
p
t(a2t) |a||t| |a|
L = lim p = lim √ = lim √
t→0 t2 + (a2t)2 t→0 |t| 1 + a4 t→0 1 + a4
so the limit depends on a and does not exist. Thus f is not differen-
tiable.
The partial derivatives are
p p p
0 |y| 0 |y| |x| |xy| f
fx = p (|x|) = p ( ) = =
2 |x| 2 |x| x 2x 2x
p p p
|x| |x| |y| |xy| f
fy0 = p (|y|)0 = p ( ) = =
2 |y| 2 |y| y 2y 2y
which are not continuous at (0, 0). As an example, consider fx0 along
the path x = t2, y = 1. Then

t2 |t|
lim = lim
t→0 2t t→0 2t

which is 1/2 from the right, but −1/2 from the left. Hence the limit
does not exist. Similarly, fy0 is not continuous.

6. Let f(x, y) = x3 /(x2 + y 2 ) if (x, y) 6= (0, 0) and f(0, 0) = 0. Show


that f is continuous, has partial derivatives at (0, 0), but that f is not
differentiable at (0, 0).

Solution: If f is continuous, then the limit as (x, y) goes to (0, 0)


of f must be 0. By the squeeze principle,
|x|3 R3
|f(x, y) − 0| = ≤ =R
x2 + y 2 R2
which goes to 0 as R goes to 0. Hence the limit exists and is indeed 0;
therefore f is continuous. The partial derivatives of f are calculated
as follows:
d d 2
fx0 (0, 0) = f(x, 0)|x=0 = x |x=0 = 0
dx dx
d d
fy0 (0, 0) = f(0, y)|y=0 = (0)|y=0 = 0
dy dy
338 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

Hence L(x, y) = 0. Then we have

f(x, y) − L(x, y) x3
L = lim = lim 2
r→0 (x + y 2 )3/2
p
r→0 x2 + y 2
Let x = t and y = at. Then,
t3 1
L = lim = lim
t→0 (t2 + a2 t2 )3/2 t→0 (1 + a2 )3/2

which depends on a. Therefore the limit does not exist and f is not
differentiable.
7–13. Find the set on which each of the following functions is differen-
tiable:

7. f(x, y) = y x;

Solution: First recall the necessary and sufficient conditions


for differentiability of f at a point r0. It is necessary that f is
continuous at r0 and that the partial derivatives exist at r0. It
is sufficient to show that the partial derivatives are continuous
at r0.

First, f is continuous everywhere on D = {(x, y)| 0 < x <


∞, −∞ < y < ∞} and√clearly has continuous partial deriva-
tives there (both y and x are continuous, differentiable func-
tions on D). So f is differentiable on D.

8. f(x, y) = xy/(2x + y);

Solution: f is the quotient of two polynomials. Hence, it


is continuous everywhere the denominator does not vanish,
and has continuous partial derivatives at the same location.
The denominator vanishes when y = −2x. Hence f is differ-
entiable on D = {(x, y)| y 6= −2x}.

9. f(x, y, z) = sin(xy + z)ezy ;

Solution: f is the product of two continuous functions and


hence continuous everywhere. Moreover, those two functions
are continuously differentiable everywhere. Therefore f is dif-
ferentiable everywhere.
6. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS 339

p
10. f(x, y, z) = x2 + y 2 − z 2 ;

Solution: f is the composition of u and a polynomial.
So f is continuous when x2 + y 2 − z 2 ≥ 0. By the chain rule,
the partial derivatives exist and are continuous at the same
location. So f is differentiable when x2 + y 2 > z 2 .

11. f(r) = ln(1 − krk), where r = hx1 , x2 , ..., xmi;

Solution: f is continuous when 1 − krk > 0 ⇒ krk < 1. f


has continuous partial derivatives when 1 − krk =
6 0, which is
included in the above condition. Hence f is differentiable in
the ball krk < 1.
p
12. f(x, y) = 3 x3 + y 3 ;

Solution: f is continuous everywhere, but its partial deriva-


tives only exist when x3 + y 3 6= 0 (and are continuous else-
where). Therefore f is differentiable everywhere except the
origin.
2
13. f(r) = e−1/krk if r 6= 0 and f(r) = 0, where r = hx, yi. Hint:
Show fx0 (r) = fy0 (r) = 0 using the definition of partial deriva-
tives. Calculate the partial derivatives at r 6= 0. Investigate
continuity of partial derivatives.

Solution: When krk = 6 0, f is continuous and has continuous


partial derivatives; therefore it is differentiable. The partial
derivatives at (0, 0) are
d d −1/x
fx0 = f(x, 0)|x=0 = e = e−1/x /x2
dx dx
d d −1/y
fy0 = f(0, y)|y=0 = e = e−1/y /y 2
dy dy
It can be shown that the limit of these as x or y goes to 0 is
0. Hence L(x, y) = 0. But,
2
f(x, y) − L(x, y) e−1/u
L = lim = lim =0
r→0 krk u→0 u
So f is differentiable at the origin as well.
340 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

14–19. The line through a point P0 of a surface perpendicular to the


tangent plane at P0 is called the normal line. Show that each of the
following surfaces has a tangent plane at the given point P0 , and find an
equation of the tangent plane and symmetric equations of the normal
line to each of the following surfaces at the specified point:
14. z = x2 + 3y − y 3 x, P0 = (1, 2, −1);

Solution: f(x, y) = x2 + 3y − y 3x is differentiable at P0


because it is a polynomial and polynomials have continuous
partial derivatives. So a best linear approximation L(x, y) to
f at P0 exists, which is exactly given by the tangent plane.
So a tangent plane at P0 of z = f(x, y) exists. The partial
derivatives are
d 2
fx0 (1, 2) = (x + 6 − 8x)|x=1 = (2x − 8)|x=1 = −6
dx
d
fy0 (1, 2) = (1 + 3y − y 3)|y=2 = (3 − 3y 2)|y=2 = −9
dy
Therefore L(x, y) is given by
L(x, y) = −1 − 6(x − 1) − 9(y − 2) = 23 − 6x − 9y ⇒ 6x + 9y + z = 23
where a normal vector to L(x, y) is n = h−fx0 (1, 2), −fy0 (1, 2), 1i =
h6, 9, 1i. So a line parallel to n through P0 is given by
x−1 y−2
r(t) = h1, 2, −1i + th6, 9, 1i ⇒ = =z+1
6 9
p
15. z = x3 y, P0 = (1, 4, 2);
p
Solution: f(x, y) = x3 y is differentiable at P0 because
it has continuous partial derivatives there (they are only dis-
continuous at x, y = 0). So a tangent plane at P0 of z = f(x, y)
exists. The partial derivatives are
d
fx0 (1, 4) = (2x3/2 )|x=1 = (3x1/2)|x=1 = 3
dx
d √ √ 1
fy0 (1, 4) = ( y)|y=4 = (1/(2 y))|y=4 =
dy 4
Therefore L(x, y) is given by
1
L(x, y) = 2 + 3(x − 1) + (y − 4) ⇒ 12x + y − 4z = 8
4
6. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS 341

where a normal vector to L(x, y) is n = hfx0 (1, 4), fy0 (1, 4), −1i =
h3, 1/4, −1i. So a line parallel to n through P0 is given by
x−1
r(t) = h1, 4, 2i + th3, 1/4, −1i ⇒ = 4(y − 4) = −(z − 2)
3
16. z = y ln(x2 − 3y), P0 = (2, 1, 0);

Solution: f(x, y) = y ln(x2 − 3y) is differentiable at P0 be-


cause it has continuous partial derivatives there (they are only
discontinuous or do not exist for x2 − 3y ≤ 0, but 4 − 3 > 0).
So a tangent plane at P0 of z = f(x, y) exists. The partial
derivatives are
d 2x
fx0 (2, 1) = (ln(x2 − 3))|x=2 = ( 2 )|x=2 = 4
dx x −3
d 3y
fy0 (2, 1) = (y ln(4 − 3y))|y=1 = (ln(4 − 3y) − )|y=1 = −3
dy 4 − 3y
Therefore L(x, y) is given by
L(x, y) = 0 + 4(x − 2) − 3(y − 1) ⇒ 4x − 3y − z = 5
where a normal vector to L(x, y) is n = hfx0 (2, 1), fy0 (2, 1), −1i =
h4, −3, −1i. So a line parallel to n through P0 is given by
x−2 y−1
r(t) = h2, 1, 0i + th4, −3, −1i ⇒ =− = −z
4 3
17. y = arctan(xz 2 ), P0 = (1, π/4, −1);

Solution: f(x, z) = arctan(xz 2) is differentiable at P0 be-


cause it has continuous partial derivatives there (they only do
not exist when 1 + xz 2 = 0). So a tangent plane at P0 of
y = f(x, z) exists. The partial derivatives are
d 1 1
fx0 (1, −1) = (arctan x)|x=1 = ( 2
)|x=1 =
dx 1+x 2
d 2z
fz0 (1, −1) = (arctan(z 2))|z=−1 = ( )|z=−1 = −1
dz 1 + z4
Therefore L(x, z) is given by
π 1 1 π
y = L(x, z) = + (x − 1) − (z + 1) ⇒ (x − 1) − (y − ) − (z + 1) = 0
4 2 2 4
where a normal vector to L(x, z) is n = hfx (1, −1), −1, fz0 (1, −1)i =
0

h1/2, −1, −1i. So a line parallel to n through P0 is given by


r(t) = h1, π/4, −1i+th1/2, −1, −1i ⇒ 2(x−1) = −(y−π/4) = −(z+1)
342 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

18. x = z cos(y − z), P0 = (1, 1, 1);

Solution: f(y, z) = z cos(y − z) is differentiable at P0 be-


cause it has continuous partial derivatives there. So a tangent
plane at P0 of x = f(y, z) exists. The partial derivatives are
d
fy0 (1, 1) = (cos(y − 1))|y=1 = (− sin(y − 1))|y=1 = 0
dy
d
fz0 (1, 1) = (z cos(1 − z))|z=1 = (cos(1 − z) + z sin(1 − z))|z=1 = 1
dz
Therefore L(y, z) is given by
x = L(y, z) = 1 + 0(y − 1) + (z − 1) ⇒ x − z = 0
where a normal vector to L(y, z) is n = h−1, fy0 (1, 1), fz0 (1, 1)i =
h−1, 0, 1i. So a line parallel to n through P0 is given by
r(t) = h1, 1, 1i + th−1, 0, 1i ⇒ −(x − 1) = z − 1, y = 1
19. z − y + ln z − ln x = 0, P0 = (1, 1, 1).

Solution: f(x, z) = z + ln z − ln x is differentiable at P0


because it has continuous partial derivatives there (they only
do not exist or are discontinuous when x, z ≤ 0). So a tangent
plane at P0 of y = f(x, z) exists. The partial derivatives are
d 1
fx0 (1, 1) = (1 − ln x)|x=1 = (− )|x=1 = −1
dx x
d 1
fz0 (1, 1) = (z + ln z)|z=1 = (1 + )|z=1 = 2
dz z
Therefore L(x, z) is given by
y = L(x, z) = 1 − (x − 1) + 2(z − 1) ⇒ x + y − 2z = 0
where a normal vector to L(x, z) is n = h−fx0 (1, 1), 1, −fz0 (1, 1)i =
h1, 1, −2i. So a line parallel to n through P0 is given by
z −1
r(t) = h1, 1, 1i + th1, 1, −2i ⇒ x − 1 = y − 1 = −
2
20–22. Show that each of the following functions is differentiable at
the given point r0, and find its linearization at r0 :
2y+3
20. f(x, y) = 4x+1 , r0 = 0;

Solution: f is differentiable because it is a rational func-


tion, and rational functions are differentiable except when the
denominator vanishes; i.e., when x = −1/4. Since x = 0 at the
6. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS 343

given point, it is differentiable there. The partial derivatives


are
(2y + 3)(−4) 3(−4)
fx0 (0, 0) = | (x,y)=(0,0) = = −12
(4x + 1)2 (4(0) + 1)2
2
fy0 (0, 0) = |(x,y)=(0,0) = 2
4x + 1
The linearization then is
2(0) + 3
L(x, y) = − 12(x − 0) + 2(y − 0) = 3 − 12x + 2y
4(0) + 1
p
21. f(x, y, z) = z 1/3 x + cos2 (y), r0 = h0, 0, 1i;

Solution: f is differentiable because it is a product of dif-


ferentiable functions. z 1/3pis clearly differentiable so long as
z 6= 0, which is the case. x + cos2 (y) is differentiable (using
the product and chain rule) so long as x + cos2(y) 6= 0, which
is also not the case at r0 . The partial derivatives are
d p d √ 1 1
fx0 (0, 0, 1) = ((1) x + cos2 (0))|x=0 = ( x + 1)|x=0 = √ |x=0 =
dx dx 2 x+1 2
d p d
fy0 (0, 0, 1) = ((1) 0 + cos2 (y))|y=0 = | cos(y)|y=0 = 0
dy dy
d 1/3p d 1/3 1 1
fz0 (0, 0, 1) = (z 0 + cos2 (0))|z=1 = (z )|z=1 = 2/3 |z=1 =
dz dy 3z 3
(note that | cos(y)| = cos(y) near y = 0). The linearization
then is
1 1 2 x z
L(x, y) = f(0, 0, 1) + (x − 0) + 0(y − 0) + (z − 1) = + +
2 3 3 2 3
22. f(r) = sin(n·r), r0 = ha1 , a2 , ..., ami, where r = hx1 , x2 , ..., xmi
and n 6= 0 is a constant vector orthogonal to r0 .

Solution: f is differentiable because it is the composition


of differentiable functions. sin u is clearly differentiable and
n · r is differentiable since it is a polynomial in m variables.
The partial derivatives are
fx0 i (r) = cos(n · r)(n0xi · r + n · r0xi ) = cos(n · r)(ni )
where ni is the ith component of n (for 1 ≤ i ≤ m). Now, if
r0 is orthogonal to n, then n · r0 = 0. So,
fx0 i (r0 ) = cos(n · r0)ni = ni
344 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

The linearization then is


L(x1, x2 , ..., xm) = sin(n·r0 )+n1 (x1−a1)+n2 (x2−a2)+...+nm(xm −am ) = n·(r−r0) = n·r
23–26. Use the linearization to approximate the following numbers.
Then use a calculator to find the numbers. Compare the results.
p
23. 20 − 7x2 − y 2, where (x, y) = (1.08, 1.95);
p
Solution: Let f(x, y) = 20 − 7x2 − y 2. We will find the
linearization at (1, 2). Then the partial derivatives are
d√ 7x 7 7
fx0 (1, 2) = 16 − 7x2 |x=1 = − √ |x=1 = − √ = −
dx 16 − 7x2 9 3
d p y 2
fy0 (1, 2) = 13 − y 2|y=2 = − p |y=2 = −
dy 13 − y 2 3
Then the linearization is
7 2
L(x, y) = 3 − (x − 1) − (y − 2)
3 3
Then L(1.08, 1.95) is a good approximation for f(1.08, 1.95).
So,
7 2 8.54
L(1.08, 1.95) = 3 − (.08) − (−.05) = ≈ 2.84666666667
3 3 3
whereas f(1.08, 1.95) ≈ 2.83420182768, which is fairly close to
the approximation.

24. xy 2 z 3 , where (x, y, z) = (1.002, 2.003, 3.004);

Solution: Let f(x, y, z) = xy 2z 3 . We will find the lineariza-


tion at (1, 2, 3). Then the partial derivatives are
d
fx0 (1, 2, 3) = (2)2 (3)3 x|x=1 = 108
dx
d
fy0 (1, 2, 3) = (1)(3)3 y 2|y=2 = 27(2y)|y=2 = 108
dy
d
fz0 (1, 2, 3) = (1)(2)2 z 3 |z=3 = 4(3z 2 )|z=3 = 108
dz
Then the linearization is
L(x, y, z) = 108 + 108(x − 1) + 108(y − 2) + 108(z − 3)
Then L(1.002, 2.003, 3.004) is a good approximation for f(1.002, 2.003, 3.004).
So,
L(1.002, 2.003, 3.004) = 108(1+.002+.003+.004) = 108(1.009) = 108.972
6. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS 345

whereas f(1.002, 2.003, 3.004) ≈ 108.975634194, which is fairly


close to the approximation.
q p
2 3
25. (1.03) / 0.98 4 (1.05)2 ;

Solution: Let f(x, y, z) = x2 /(y 1/3z 1/6); observe that f(1.03, .98, 1.05)
is exactly the quantity we wish to approximate. We will find
the linearization at (1, 1, 1). Then the partial derivatives are
d 2 1/3 1/6
fx0 (1, 1, 1) = x /(1 (1 ))|x=1 = 2x|x = 1 = 2
dx
d 1 1
fy0 (1, 1, 1) = (1)2 /(y 1/3(11/6))|y=1 = − 4/3 |y=1 = −
dy 3y 3
d 1 1
fz0 (1, 1, 1) = (1)2 /(11/3 (z 1/6))|z=1 = − 7/6 |z=1 = −
dz 6z 6
Then the linearization is
1 1
L(x, y, z) = 1 + 2(x − 1) − (y − 1) − (z − 1)
3 6
Then L(1.03, .98, 1.05) is a good approximation for f(1.03, .98, 1.05).
So,
1 1 6.35
L(1.03, .98, 1.05) = 1+2(.03)− (−.02)− (.05) = ≈ 1.05833333333
3 6 5
whereas f(1.03, .98, 1.05) ≈ 1.05941847196, which is fairly close
to the approximation.

26. (0.97)1.05 .

Solution: Let f(x, y) = xy . We will find the linearization at


(1, 1). Then the partial derivatives are
d
fx0 (1, 1) = x|x=1 = 1
dx
d y
fy0 (1, 1) = 1 |y=1 = 0
dy
Then the linearization is
L(x, y) = 1 + (x − 1) + 0(y − 1) = 1 + (x − 1)
Then L(0.97, 1.05) is a good approximation for f(0.97, 1.05).
So,
L(0.97, 1.05) = 1 + (−.03) = .97
346 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

whereas f(0.97, 1.05) ≈ 0.968523852779, which is fairly close


to the approximation.

27–29. Use one iteration of Newtons method initiated at the given


point P0 to approximate the solution of each of the following system of
nonlinear equations. Assume that the solution exists near P0 . Explain
why Newtons method is applicable in each case. Check whether the
approximate solution is better than the initial guess P0 .
27. x2 + xy + 2y 2 = 4.2, x2 − x3 y + y 3 = 0.9, P0 = (1, 1);

Solution: Let f(x, y) = x2 + xy + 2y 2 and g(x, y) = x2 −


x3y + y 3. Then f(1, 1) = 4, g(1, 1) = 1. The functions are
differentiable, and in particular
fx0 (1, 1) = 2x + y|(x,y)=(1,1) = 3, fy0 (1, 1) = x + 4y|(x,y)=(1,1) = 5
gx0 (1, 1) = 2x − 3x2 y|(x,y)=(1,1) = −1, gy0 (1, 1) = −x3 + 3y 2 |(x,y)=(1,1) = 2
Put x = 1 + ∆x and y = 1 + ∆y. We may linearize f and g at
P0 and obtain a system of equations where ∆x and ∆y may
be solved for:
(
3∆x + 5∆y = 4.2 − f(1, 1) = .2 ⇒ ∆x = .9/11 = 9/110
−∆x + 2∆y = 0.9 − g(1, 1) = −.1 ⇒ ∆y = −.1/11 = −1/110
∆y can easily be found by adding three times the second equa-
tion to the first. Find ∆x by substituting for ∆y in the second
equation. Finally,
f(1 + ∆x, 1 + ∆y) ≈ 4.20611570248
g(1 + ∆x, 1 + ∆y) ≈ 0.888729997951
which are closer to 4.2 and 0.9, respectively, than the initial
guess.

28. ln(1 + y + 2x) + x2 + y = 3.2, x3 + y 2 = 2.7, P0 = (−1, 2);

Solution: Let f(x, y) = ln(1 + y + 2x) + x2 + y and g(x, y) =


x3 + y 2. Then f(−1, 2) = 3, g(−1, 2) = 3. The functions are
differentiable, and in particular
2 1
fx0 (−1, 2) = + 2x|(−1,2) = 0, fy0 (−1, 2) = + 1|(−1,2) = 2
1 + y + 2x 1 + y + 2x
gx0 (−1, 2) = 3x2 |(−1,2) = 3, gy0 (−1, 2) = 2y|(−1,2) = 4
6. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS 347

Put x = −1 + ∆x and y = 2 + ∆y. We may linearize f and


g at P0 and obtain a system of equations where ∆x and ∆y
may be solved for:
(
2∆y = 3.2 − f(−1, 2) = .2 ⇒ ∆y = .1
3∆x + 4∆y = 2.7 − g(−1, 2) = −.3 ⇒ ∆x = −.7/3
∆y can easily be found by adding three times the second equa-
tion to the first. Find ∆x by substituting for ∆y in the second
equation. Finally,
f(−1 + ∆x, 2 + ∆y) ≈ 3.16435270862
g(−1 + ∆x, 2 + ∆y) ≈ 2.53396296296
which are closer to 3.2 and 2.7, respectively, than the initial
guess.

29. y sin(y − x) + x = 1.2, x cos(y 2 − x) + y = 0.8, P0 = (1, 1).

Solution: Let f(x, y) = y sin(y − x) + x and g(x, y) =


x cos(y 2 − x) + y. Then f(1, 1) = 1, g(1, 1) = 2. The functions
are differentiable, and in particular
fx0 (1, 1) = [−y cos(y − x) + 1]|(1,1) = 0, fy0 (1, 1) = [sin(y − x) + y cos(y − x)]|(1,1) = 1
gx0 (1, 1) = [cos(y 2 − x) + x sin(y 2 − x)]|(1,1) = 1, gy0 (1, 1) = [−2xy sin(y 2 − x) + 1]|(1,1) = 1
Put x = 1 + ∆x and y = 1 + ∆y. We may linearize f and g at
P0 and obtain a system of equations where ∆x and ∆y may
be solved for:
(
∆y = 1.2 − f(1, 1) = .2 ⇒ ∆y = .2
∆x + ∆y = 0.8 − g(1, 1) = −1.2 ⇒ ∆x = −1.4
∆y can easily be found by adding three times the second equa-
tion to the first. Find ∆x by substituting for ∆y in the second
equation. Finally,
f(1 + ∆x, 1 + ∆y) ≈ 1.59982944122
g(1 + ∆x, 1 + ∆y) ≈ 1.30638555024
of which the former is farther away from the initial guess,
whereas the latter is closer.
348 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

30–31. The existence of partial derivatives at a point is not sufficient


for continuity of the function at that point (Example 21.2). Prove the
following assertions:
30. Suppose that a function f(x, y) is continuous with respect to
x at each fixed y and has a bounded partial derivative fy0 (x, y),
i.e., |fy0 (x, y)| ≤ M for some M > 0 and all (x, y). Then f is
continuous.

Solution: Proof: That f is continuous with respect to x


at each fixed y implies that f is defined for all (x, y). Indeed,
fix a y = y0. Then limx→x0 f(x, y0 ) = f(x0 , y0) since f is con-
tinuous with respect to x for fixed y. Hence f must exist at
each (x0, y0).
We may linearize f at a point P0 (x0 , y0) as
f(x, y) ≈ L(x, y) = f(x0 , y0) + fx0 (x0 , y0)(x − x0) + f 0 y(x0, y0 )(y − y0)
Then,
|f(x, y)−f(x0, y0)| ≤ |fx0 (x0, y0 )(x−x0)+fy0 (x0 , y0)| ≤ |fx0 (x0, y0 )||x−x0|+|M|(y−y0)|
31. Let f(x, y) be defined on open set D and have bounded partial
derivatives, |fx0 (x, y)| ≤ M1 and |fy0 (x, y)| ≤ M2 for all (x, y)
in D and some positive M1 and M2 . Then f is continuous on D.

Solution:
7. CHAIN RULES AND IMPLICIT DIFFERENTIATION 349

7. Chain Rules and Implicit Differentiation


1. Let f(x, y) = 4 − x2 − y 2 if (x, y) is not in D = {(x, y)| x < y <
2x − 1, x > 1} and f(x, y) = 0 if (x, y) is in D. Sketch D and the
graph z = f(x, y) for the disk x2 + y 2 ≤ 4. Use the definition of
partial derivatives to find fx0 (1, 1) and fy0 (1, 1). Consider a smooth sim-
ple curve through the point (1, 1) with parametric equations x = x(t),
y = y(t) so that x(0) = y(0) = 1 and x0(0) and y 0(0) do not vanish. Put
F (t) = f(x(t), y(t)). Is it true that F 0(0) = fx0 (1, 1)x0 (0) + fy0 (1, 1)y 0 (0)
for any smooth simple curve through (1, 1)? If not, give an example of
a curve for which the chain rule holds and an example of a curve for
which it does not hold.

Solution: For a sketch of D and f(x, y) in x2 + y 2 ≤ 4, see Exercise


2 of the previous section. The partial derivatives are
fx0 = −2x ⇒ fx0 (1, 1) = −2
fy0 = −2y ⇒ fy0 (1, 1) = −2
Let r(t) = hx(t), y(t) be a smooth curve satisfying the above conditions.
Then F (t) is
F (t) = f(x(t), y(t)) = 4 − x(t)2 − y(t)2
and so, by the chain rule for one variable
F 0(t) = 4 − x(t)(2x0(t)) − y(t)((2y 0(t)) = 4 − 2x(t)x0(t) − 2y(t)y 0(t)
Since x(0) = y(0) = 1 and fx0 (1, 1) = fy0 (1, 1) = −2 we have
F 0(0) = −2x0(0) − 2y 0 (0) = fx0 (1, 1)x0 (0) + fy0 (1, 1)y 0 (0)
Thus it is true that F 0(0) = fx0 (1, 1)x0 (0) + fy0 (1, 1)y 0 (0) for any smooth
simple curve through (1, 1).
p
2. Use the chain rule to find dz/dt if z = 1 + x2 + 2y 2 and x = 2t3 ,
y = ln t.

Solution: By the chain rule,


dz ∂z dx ∂z dy x 2y 1 12t5 + 2 ln t/t
= + =p (6t2)+ p ( )=p
dt ∂x dt ∂y dt 1 + x2 + 2y 2 1 + x2 + 2y 2 t 1 + 4t6 + 2(ln t)2
3. Use the chain
√ rule to find ∂z/∂s and ∂z/∂t if z = e−x sin(xy) and
x = ts, y = s2 + t 2 .
350 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

Solution: By the chain rule, we have


∂z ∂z ∂x ∂z ∂y s
= + = (−e−x sin(xy) + ye−x cos(xy))(t) + (xe−x cos(xy))( √ )
∂s ∂x ∂s ∂y ∂s s + t2
2

∂z ∂z ∂x ∂z ∂y t
= + = (−e−x sin(xy) + ye−x cos(xy))(s) + (xe−x cos(xy))( √ )
∂t ∂x ∂t ∂y ∂t s + t2
2

4–5. Use the chain rule to write the partial derivatives of each of
the following functions F with respect to the new variables:
4. F = f(x, y), x = x(u, v, w), y = y(u, v, w);

Solution: We will need to calculate the derivative of F with


respect to u, v, and w. By the chain rule,
∂f ∂x ∂f ∂y
Fu0 = + = fx0 x0u + fy0 yu0
∂x ∂u ∂y ∂u
∂f ∂x ∂f ∂y
Fv0 = + = fx0 x0v + fy0 yv0
∂x ∂v ∂y ∂v
∂f ∂x ∂f ∂y
Fw0 = + = fx0 x0w + fy0 yw0
∂x ∂w ∂y ∂w
5. F = f(x, y, z, t), x = x(u, v), y = y(u, v), z = z(w, s), t =
t(w, s).

Solution: We will need to calculate the derivative of F with


respect to u, v, w, and s. By the chain rule,
∂f ∂x ∂f ∂y ∂f ∂z ∂f ∂t
Fu0 = + + + = fx0 x0u + fy0 yu0
∂x ∂u ∂y ∂u ∂z ∂u ∂t ∂u
∂f ∂x ∂f ∂y ∂f ∂z ∂f ∂t
Fv0 = + + + = fx0 x0v + fy0 yv0
∂x ∂v ∂y ∂v ∂z ∂v ∂t ∂v
∂f ∂x ∂f ∂y ∂f ∂z ∂f ∂t
Fw0 = + + + = fz0 zw0 + ft0 t0w
∂x ∂w ∂y ∂w ∂z ∂w ∂t ∂w
∂f ∂x ∂f ∂y ∂f ∂z ∂f ∂t
Fs0 = + + + = fz0 zs0 + ft0 t0s
∂x ∂s ∂y ∂s ∂z ∂s ∂t ∂s
owing to the fact that zu0 = t0u = 0, zv0 = t0v = 0 , x0w = yw0 = 0,
and x0s = ys0 = 0
6. Find the rates of change ∂z/∂u, ∂z/∂v, ∂z/∂w when (u, v, w) =
(2, 1, 1) if z = x2 + yx + y 3 and x = uv 2 + w3, y = u + v ln w.
7. CHAIN RULES AND IMPLICIT DIFFERENTIATION 351

Solution: First note that x(2, 1, 1) = 3 and y(2, 1, 1) = 2. Then


we calculate the individual partial derivatives,

zx0 = 2x + y ⇒ zx0 (3, 2) = 8


zy0 = x + 3y 2 ⇒ zy0 (3, 2) = 15
x0u v 2 ⇒ x0u (2, 1, 1) = 1
=
x0v 2vu ⇒ x0v (2, 1, 1) = 4
=
x0w 3w2 ⇒ x0w (2, 1, 1) = 3
=
yu0 1 ⇒ yu0 (2, 1, 1) = 1
=
yv0 ln w ⇒ yv0 (2, 1, 1) = 0
=
v
yw0 = ⇒ yw0 (2, 1, 1) = 1
w

Then we have

zu0 (3, 2) = zx0 (3, 2)x0u (2, 1, 1) + zy0 (3, 2)yu0 (2, 1, 1) = 8(1) + 15(1) = 23
zv0 (3, 2) = zx0 (3, 2)x0v (2, 1, 1) + zy0 (3, 2)yv0 (2, 1, 1) = 8(4) + 15(0) = 32
zw0 (3, 2) = zx0 (3, 2)x0w (2, 1, 1) + zy0 (3, 2)yw0 (2, 1, 1) = 8(3) + 15(1) = 39

7. For a differentiable function f(x, y, z) find the rates of change


fu0 , fz0 , fw0 when (x, y, z) = (0, 2, −1) if x = 2/u − v + w, y = vuw,
z = w3 , and fx0 (0, 2, −1) = 2, fy0 (0, 2, −1) = 1, fz0 (0, 2, −1) = −1.

Solution: First we have that if x = 0, y = 2, and z = −1, then


w = −1 (from z = w3 ), u = 4, and v = −1/2. The partial derivatives
are

2 1 1
x0u = − 2
⇒ x0u (4, − , −1) = −
u 2 8
1
x0v = −1 ⇒ x0v (4, − , −1) = −1
2
1
x0w = 1 ⇒ x0w (4, − , −1) = 1
2
1 1
yu0 = vw ⇒ yu0 (4, − , −1) =
2 2
1
yv0 = uw ⇒ yv0 (4, − , −1) = −4
2
1
yw0 = 0
vu ⇒ yw (4, − , −1) = −2
2
352 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

1
zu0 = 0 ⇒ zu0 (4, − , −1) = 0
2
0 0 1
zv = 0 ⇒ zv (4, − , −1) = 0
2
0 2 0 1
zw = 3w ⇒ zw (4, − , −1) = 3
2
Then the rates of change are
1 1 1
fu0 (0, 2, −1) = fx0 (0, 2, −1)x0u (4, − , −1) + fy0 (0, 2, −1)yu0 (4, − , −1) + fz0 (0, 2, −1)zu0 (4, − , −1)
2 2 2
1 1 1
= (2)(− ) + (1)( ) + (−1)(0) =
8 2 4
1 1 1
fv0 (0, 2, −1) = fx0 (0, 2, −1)x0v (4, − , −1) + fy0 (0, 2, −1)yv0 (4, − , −1) + fz0 (0, 2, −1)zv0 (4, − , −1)
2 2 2
= (2)(−1) + (1)(−4) + (−1)(0) = −6
1 1 1
fw0 (0, 2, −1) = fx0 (0, 2, −1)x0w (4, − , −1) + fy0 (0, 2, −1)yw0 (4, − , −1) + fz0 (0, 2, −1)zw0 (4, − , −1)
2 2 2
= (2)(1) + (1)(−2) + (−1)(3) = −3

8. If z(u, v) = f(x, y) where x = eu cos v and y = eu sin v, show that


00 00
zxx + zyy = e−2u (zuu
00 00
+ zvv )

Solution: The derivation in the book for the chain rule for higher
derivatives only does so if x, y are with functions of one variable. Here,
we will derive it for two, although it is essentially the same. First we
calculate zu0 from the normal chain rule:

zu0 = fx0 x0u + fy0 yu0

Now we apply the chain rule and product rule to zu0


00
zuu = (fx0 x0u )0u + (fy0 yu0 )0u = (fx0 )0u x0u + fx0 x00uu + (fy0 )0u yu0 + fy0 yuu
00

00 0 00 0
= (fxx xu + fxy yu )x0u + fx0 x00uu + (fyx
00 0
xu + fyy yu )yu0 + fy0 yuu
00 0 00

00
= fxx (x0u )2 + fyy
00
(yu0 )2 + fx0 x00uu + fy0 yuu
00 00 0 0
+ 2fxy xu yu

Here we have that

x0u = eu cos v
x00uu = eu cos v
x0v = −eu sin v
x00vv = −eu cos v
7. CHAIN RULES AND IMPLICIT DIFFERENTIATION 353

yu0 = eu sin v
00
yuu = eu sin v
yv0 = eu cos v
00
yvv = −eu sin v
00 00
Then zuu and zvv are
00 00
zuu = fxx (eu cos v)2 + fyy
00
(eu sin v)2 + fx0 (eu cos v) + fy0 (eu sin v) + 2fxy
00
(eu cos v)(eu sin v)
00
= fxx (e2u cos2 v) + fyy
00
(e2u sin2 v) + fx0 (eu cos v) + fy0 (eu sin v) + 2fxy
00
(e2u cos v sin v)
00 00
zvv = fxx (−eu sin v)2 + fyy
00
(eu cos v)2 + fx0 (−eu cos v) + fy0 (−eu sin v) + 2fxy
00
(−eu sin v)(eu cos v)
00
= fxx (e2u sin2 v) + fyy
00
(e2u cos2 v) − fx0 (eu cos v) − fy0 (eu sin v) + 2fxy
00
(−e2u cos v sin v)
Note the cancellation in fx0 , fy0 , and fxy
00
when the two are added. Thus,
00 00 00
zuu +zvv = fxx (e2u cos2 v+e2u sin2 v)+fuu
00
(e2u sin2 v+e2u cos2 v) = e2u(fxx
00 00
+fyy ) = e2u (zxx
00 00
+zyy )
00 00
We do not need to apply chain rule when calculating zxx or zyy .

9. If z(u, v) = f(x, y) where x = u2 + v 2 and y = 2uv, find all the


second order partial derivatives of z(u, v).

Solution: Half of the work for this problem was derived in Exercise
00 00
8. To find zuv = zvu , we have
00
zuv = (zu0 )0v = (fx0 x0u )0v + (fy0 yu0 )0v = (fx0 )0v x0u + fx0 x00uv + (fy0 )0v yu0 + fy0 yuv
00

00 0 00 0
= (fxx xv + fxy xv )x0u + fx0 x00uv + (fyx
00 0 00 0
xv + fyy yv )yu0 + fy0 yuv
00

00 0 0 00 0 0 00
= fxx xu xv + fyy yu yv + fxy (x0u yv0 + x0v yu0 ) + fx0 x00uv + fy0 yuv
00

The partial derivatives of x, y are


x0u = 2u, yu0 = 2v
x0v = 2v, yv0 = 2u
x00uu = 00
2 yuu =0
x00vv = 00
2 yvv = 0
x00uv = 00
0 yuv =2
00 00 00
Then zuu , zvv , and zuv are
00 00
zuu = fxx (2u)2 + fyy
00
(2v)2 + fx0 (2) + fy0 (0) + 2fxy
00
(2u)(2v) = 2fx0 + 8uvfxy
00
+ 4u2 fxx
00
+ 4v 2 fyy
00

00 00
zvv = fxx (2v)2 + fyy
00
(2u)2 + fx0 (2) + fy0 (0) + 2fxy
00
(2v)(2u) = 2fx0 + 8uvfxy
00
+ 4v 2fxx
00
+ 4u2 fyy
00

00 00 00 00
zuv = fxx (2u)(2v) + fyy (2v)(2u) + fxy ((2u)(2u) + (2v)(2v)) + fx0 (0) + fy0 (2)
= 2fy0 + 4uv(fxx
00 00
+ fyy ) + 4(u2 + v 2 )fxy
00
= 2fy0 + 2y(fxx
00 00
+ fyy 00
) + 4xfxy
354 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

10. If z(u, v) = f(x, y) where x = u + v and y = u − v, show that


(zx0 )2 + (zy0 )2 = zu0 zv0

Solution: By the chain rule,


zu0 = zx0 x0u + zy0 yu0 = zx0 (1) + zy0 (1) = zx0 + zy0
zv0 = zx0 x0v + zy0 yv0 = zx0 (1) + zy0 (−1) = zx0 − zy0
Then,
zu0 zv0 = (zx0 + zy0 )(zx0 − zy0 ) = (zx0 )2 + (zy0 )2
11–16. Find the first and second partial derivatives of the function
g in terms of the first and second partial derivatives of the function f:
11. g(x, y, z) = f(x2 + y 2 + z 2 );

Solution: Let u = x2 + y 2 + z 2. Then g(x, y, z) = f(u).


These problems will be a little strange, as we are used to hav-
ing f be a function of x, y, z where x, y, z are functions of u, v,
etc. Here it is backwards; keep this in mind! By the chain rule
the first derivatives are
gx0 = f 0 (u)u0x = 2xf 0 (u)
gy0 = f 0 (u)u0y = 2yf 0 (u)
gz0 = f 0 (u)u0z = 2zf 0 (u)
Using the chain and product rules, the second derivatives are
00
gxx = 2f 0 (u) + 2x(f 00 (u)u0x) = 2f 0 (u) + 4x2 f 00(u)
00
gyy = 2f 0 (u) + 2y(f 00 (u)u0y ) = 2f 0 (u) + 4y 2f 00 (u)
00
gzz = 2f 0 (u) + 2z(f 00 (u)u0z ) = 2f 0 (u) + 4z 2 f 00 (u)
00 00
gxy = gyx = (0)f 0 (u) + 2x(f 00(u)u0y ) = 4xyf 00(u)
00 00
gxz = gzx = (0)f 0 (u) + 2x(f 00 (u)u0z ) = 4xzf 00(u)
00 00
gzy = gyz = (0)f 0 (u) + 2z(f 00 (u)u0y ) = 4yzf 00 (u)
12. g(x, y) = f(x, x/y);

Solution: Let u = x and v = x/y. Then g(x, y) = f(u, v).


By the chain rule, the first partial derivatives are
1
gx0 = fu0 u0x + fv0 vx0 = fu0 + fv0
y
x x
gy0 = fu0 u0y + fv0 vy0 = fu0 (0) − 2 fv0 = − 2 fv0
y y
7. CHAIN RULES AND IMPLICIT DIFFERENTIATION 355

Using the product and chain rules, the second partial deriva-
tives are
00 00 0 00 0 1 00 0 00 0 00 2 00 1 00
gxx = (fuu ux + fuv vx) + (fvu ux + fvv vx ) = fuu + fuv + 2 fvv
y y y
00 2x x 00 0 2x x x 00 2x x2 00
gyy = 3 fv0 − 2 (fvu 00 0
uy + fvv vy ) = 3 fv0 − 2 (− 2 fvv ) = 3 fv0 + 4 fuv
y y y y y y y
00 00 1 x 1 x 1 1 x 00 x 00
gxy = gyx = − 2 fv0 − 2 (fvu
00 0 00 0
ux + fvv vx) = − 2 fv0 − 2 (fuv 00
+ fvv00
) = − 2 fv0 − 2 fuv − 3 fvv
y y y y y y y y
13. g(x, y, z) = f(x, xy, xyz);

Solution: Let u = x, v = xy, and w = xyz. Then g(x, y, z) =


f(u, v, w). By the chain rule, the first partial derivatives are

gx0 = fu0 u0x + fv0 vx0 + fw0 wx0 = fu0 + yfv0 + yzfw0
gy0 = fu0 u0y + fv0 vy0 + fw0 wy0 = xfv0 + xzfw0
gz0 = fu0 u0z + fv0 vz0 + fw0 wz0 = xyfw0

Using the product and chain rules, the second partial deriva-
tives are
00 00 0 00 0 00
gxx = (fuu ux + fuv vx + fuw wx0 ) + y(fvu
00 0
ux + fvv 00 0
vx + fvw00
wx0 ) + yz(fwu 00
u0x + fwv
00 0 00
vx + fww wx0 )
00
= fuu + y 2fvv
00
+ y 2 z 2fww
00
+ 2yfuv00
+ 2yzfuw 00
+ 2y 2 zfvw00

00 00 0 00 0 00
gyy = x(fvu uy + fvv vy + fvw wy0 ) + xz(fwu00
u0y + fwv 00 0
vy + fww 00
wy0 ) = x2 fvv
00
+ x2z 2 fww
00
+ 2x2 zfvw00

00 00
gzz = xy(fwu u0z + fwv
00 0 00
vz + fww wz0 ) = x2y 2 fww
00

00 00
gxy = gyx = fv0 + x(fvu00 0
ux + fvv00 0
vx + fvw00
wx0 ) + zfw0 + xz(fwu
00
u0x + fwv
00 0 00
vx + fww wx0 )
= fv0 + xyfvv
00
+ zfw0 + xyz 2fww00 00
+ xfvu + xzfwu00
+ 2xyzfvw00

00 00
gxz = gzx = yfw0 + xy(fwu00
u0x + fwv
00 0
vx + fww00
wx0 ) = yfw0 + xyfwu00
+ xy 2fwv
00
+ xy 2zfww
00

00 00
gyz = gzy = xfw0 + xy(fwu00
u0y + fwv
00 0 00
vy + fww wy0 ) = xfw0 + x2 yzfww00
+ x2yfwv00

14. g(x, y) = f(x/y, y/x);

Solution: Let u = x/y and v = y/x. Then g(x, y) = f(u, v).


By the chain rule, the first partial derivatives are
1 0 y
gx0 = fu0 u0x + fv0 vx0 = fu − 2 fv0
y x
x 1
gy0 = fu0 u0y + fv0 vy0 = − 2 fu0 + fv0
y x
356 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

Using the product and chain rules, the second partial deriva-
tives are

00 1 00 0 00 0 2y y 00 0
gxx = (fuu ux + fuv vx ) + 3 fv0 − 2 (fvu 00 0
ux + fvv vx )
y x x
1 00 y 2 00 2 00 2y
= 2
fuu + 4
fvv − 2 fuv + 3 fv0
y x x x
00 2x 0 x 00 0 00 0 1 00 0 00 0
gyy = 3
fu − 2 (fuu uy + fuv vy ) + (fvu uy + fvv vy )
y y x
x2 00 1 00 2 00 2x
= 4
fuu + 2 fvv − 2 fuv + 3 fu0
y x y y
00 00 1 1 00 0 1 y 00 0
gxy = gyx = − 2 fu0 + (fuu uy + fuv 00 0
vy ) − 2 fv0 − 2 (fvu 00 0
uy + fvv vy )
y y x x
00 x 00 y 00 2 00 1 1
= gyx = − 3 fuu − 2 fvv + fuv − 2 fu0 − 2 fv0
y x xy y x

15. g(x, y, z) = f(x + y + z, x2 + y 2 + z 2 );

Solution: Let u = x + y + z and v = x2 + y 2 + z 2 . Then


g(x, y, z) = f(u, v). By the chain rule, the first partial deriva-
tives are

gx0 = fu0 u0x + fv0 vx0 = fu0 + 2xfv0


gy0 = fu0 u0y + fv0 vy0 = fu0 + 2yfv0
gz0 = fu0 u0z + fv0 vz0 = fy0 + 2zfv0

Using the product and chain rules, the second partial deriva-
tives are
00 00 0 00 0
gxx = fuu ux + fuv vx + 2fv0 + 2x(fvu
00 0 00 0
ux + fvv vx) = fuu00
+ 4x2 fvv00 00
+ 4xfuv + 2fv0
00 00 0 00 0
gyy = fuu uy + fuv 00 0
vy + 2fv0 + 2y(fvu 00 0
uy + fvv 00
vy ) = fuu + 4y 2 fvv
00 00
+ 4yfuv + 2fv0
00 00 0 00 0
gzz = fuu uz + fuv vz + 2fv0 + 2z(fvu
00 0 00 0
uz + fvv 00
vz ) = fuu + 4z 2 fvv
00 00
+ 4zfuv + 2fv0
00 00 00 0 00 0 00 0 00 0 00 00 00
gxy = gyx = fuu uy + fuv vy + 2x(fvu uy + fvv vy ) = fuu + 4xyfvv + (2x + 2y)fuv
00 00 00 0 00 0 00 0 00 0 00 00 00
gxz = gzx = fuu uz + fuv vz + 2x(fvu uz + fvv vz ) = fuu + 4xzfvv + (2x + 2z)fuv
00 00 00 0 00 0 00 0 00 0 00 00 00
gyz = gzy = fuu uz + fuv vz + 2y(fvu uz + fvv vz ) = fuu + 4yzfvv + (2y + 2z)fuv

16. g(x, y) = f(x + y, xy).

Solution: Let u = x + y and v = xy. Then g(x, y) = f(u, v).


7. CHAIN RULES AND IMPLICIT DIFFERENTIATION 357

By the chain rule, the first partial derivatives are


gx0 = fu0 u0x + fv0 vx0 = fu0 + yfv0
gy0 = fu0 u0y + fv0 vy0 = fu0 + xfv0
Using the product and chain rules, the second partial deriva-
tives are
00 00 0 00 0 00 0 00 0 00
gxx = fuu ux + fuv vx + y(fvu ux + fvv vx ) = fuu + y 2fvv
00 00
+ 2yfuv
00 00 0 00 0 00 0 00 0 00
gyy = fuu uy + fuv vy + y(fvu uy + fvv vy ) = fuu + x2fvv
00 00
+ 2xfuv
00 00 00 0 00 0 00 0 00 0 00 00 00
gxy = gyx = fuu uy + fuv vy + y(fvu uy + fvv vy ) = fuu + xyfvv + (x + y)fuv

00 00 00
17. Find gxx + gyy + gzz if g(x, y, z) = f(x + y + z, x2 + y 2 + z 2 ).

Solution: See Exercise 15 for a derivation of the necessary partial


derivatives. Then
00 00 00 00
gxx +gyy +gzz = 3fuu +4(x2+y 2 +z 2)fvv
00 00
+4(x+y+z)fuv +6fv0 = 3fuu
00 00
+4vfvv 00
+4ufuv +6fv0

18. Let x = r cos θ and y = r sin θ. Show that

00 00 1 ∂ ∂f 1 ∂ 2f
fxx + fyy = (r ) + 2 2
r ∂r ∂r r ∂ θ

Solution: Let z = f(x, y). As shown in Exercise 8,


00 00 00
zuu = fuu = fxx (x0u )2 + fyy
00
(yu0 )2 + fx0 x00uu + fy0 yuu
00 00 0 0
+ 2fxy xu yu
for u = r, θ. Therefore,
00 00
frr = fxx (cos θ)2 + fyy
00
(sin θ)2 + 2 cos θ sin θfxy
00

00 00
fθθ = fxx (r2 sin2 θ) + fyy
00
(r2 cos2 θ) − r cos θfx0 − r sin θfy0 − 2r2 cos θ sin θfxy
00

Note that
r2 frr
00 00
+ fθθ = r2 fxx
00
+ r2 fyy
00
− r(cos θfx0 + sin θfy0 )
Since fr0 = fx0 x0r + fy0 yr0 = cos θfx0 + sin θfy0 ,
r2 frr
00 00
+ fθθ = r2 (fxx
00 00
+ fyy ) − r(fr0 )
Rearranging this gives
00 00 1 00 1 00 1 1 00 1 ∂ ∂f 1 ∂ 2f
fxx +fyy = (rfrr +fr0 )+ 2 fθθ = (rfr0 )0r + 2 fθθ = (r )+ 2 2
r r r r r ∂r ∂r r ∂ θ
which was to be shown.
358 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

19. Let x = ρ sin φ cos θ, y = ρ sin φ sin θ, and z = ρ cos φ. The


variables (ρ, φ, θ) are called spherical coordinates. Show that
00 00 00 1 ∂ 2 ∂f 1 ∂ ∂f 1 ∂ 2f
fxx + fyy + fzz = (ρ ) + (sin φ ) +
ρ2 ∂ρ ∂ρ ρ2 sin φ ∂φ ∂φ ρ2 sin2 φ ∂θ2

Solution: Let f be a function of x, y, z. We will use the same rule


found in Exercise 8, though slightly modified (as f is a function of
three variables, not two):
00 00
fuu = fxx (x0u )2 +fyy
00
(yu0 )2+fzz
00
(zu0 )2 +fx0 x00uu +fy0 yuu
00
+fz0 zuu
00 00 0 0
+2fxy 00 0 0
xu yu +2fxz 00 0 0
xu zu +2fyz yu z u
for u = ρ, θ, φ. The above is easily verified. Therefore,
00 00
fρρ = fxx (sin2 φ cos2 θ) + fyy
00
(sin2 φ sin2 θ) + fzz
00
(cos2 φ) + fxy
00
(2 sin2 φ cos θ sin θ)
00 00
+fxz (2 sin φ cos φ cos θ) + fyz (2 sin φ cos φ sin θ)
00 00
fφφ = fxx (ρ2 cos2 φ cos2 θ) + fyy
00
(ρ2 cos2 φ sin2 θ) + fzz
00
(ρ2 sin2 φ)
+fx0 (−ρ sin φ cos θ) + fy0 (−ρ sin φ sin θ) + fz0 (−ρ cos φ)
00
+fxy (2ρ2 cos2 φ cos θ sin θ) + fxz
00
(−2ρ2 cos φ sin φ cos θ) + fyz
00
(−2ρ2 cos φ sin φ sin θ)
00 00
fθθ = fxx (ρ2 sin2 φ sin2 θ) + fyy
00
(ρ2 sin2 φ cos2 θ) + fx0 (−ρ sin φ cos θ) + fy0 (−ρ sin φ sin θ)
00
+fxy (−2ρ2 sin2 φ sin θ cos θ)
Consider the quantity ρ2 sin2 φfρρ00
+ sin2 φfφφ
00 00
+ fθθ . The coefficients
here are not obvious, but can be found by multiplying the equation
we’re trying to prove by ρ2 sin2 φ. The unmixed second partials in the
sum become
00
fxx : (ρ2 sin2 φ)(sin2 φ cos2 θ)fxx
00
+ (sin2 φ)(ρ2 cos2 φ cos2 θ)fxx
00
+ (ρ2 sin2 φ sin2 θ)fxx
00

= (ρ2 sin2 φ)(sin2 φ cos2 θ + cos2 φ cos2 θ + sin2 θ)fxx00


= (ρ2 sin2 φ)(cos2 θ + sin2 θ)fxx
00

= (ρ2 sin2 φ)fxx


00

00
fyy : (ρ2 sin2 φ)(sin2 φ sin2 θ)fyy
00
+ (sin2 φ)(ρ2 cos2 φ sin2 θ)fyy
00
+ (ρ2 sin2 φ cos2 θ)fyy
00

= (ρ2 sin2 φ)(sin2 φ sin2 θ + cos2 φ sin2 θ + cos2 θ)fyy


00
= (ρ2 sin2 φ)fyy
00

00
fzz : (ρ2 sin2 φ)(cos2 φ)fzz
00
+ (sin2 φ)(ρ2 sin2 φ)fzz
00
= (ρ2 sin2 φ)(cos2 φ + sin2 φ)fzz
00

= (ρ2 sin2 φ)fzz


00

The mixed second partials become


00
fxy : (ρ2 sin2 φ)(2 sin2 φ cos θ sin θ)fxy
00
+ (sin2 φ)(2ρ2 cos2 φ cos θ sin θ)fxy
00

+(−2ρ2 sin2 φ sin θ cos θ)fxy


0
= (2ρ2 sin2 φ cos θ sin θ)(sin2 φ + cos2 φ − 1)fxy
0
=0
00
fxz : (ρ2 sin2 φ)(2 cos φ sin φ cos θ)fxz
00
+ (sin2 φ)(−2ρ2 cos φ sin φ cos θ)fxz
00
=0
00 2 2 00 2 2 00
fyz : (ρ sin φ)(2 sin φ cos φ sin θ)fyz + (sin φ)(−2ρ cos φ sin φ sin θ)fyz = 0
7. CHAIN RULES AND IMPLICIT DIFFERENTIATION 359

Finally, the first partial derivatives become


fx0 : (sin2 φ)(−ρ sin φ cos θ)fx0 + (−ρ sin φ cos θ)fx0 = −ρ sin φ cos θ(sin2 φ + 1)fx0
= −ρ sin φ cos θ(2 sin2 φ + cos2 φ)fx0
fy0 : (sin2 φ)(−ρ sin φ sin θ)fy0 + (−ρ sin φ sin θ)fy0 = −ρ sin φ sin θ(sin2 φ + 1)fy0
= −ρ sin φ sin θ(2 sin2 φ + cos2 φ)fy0
fz0 : (sin2 φ)(−ρ cos φ)fz0 = −ρ cos φ sin2 φfz0
Noting that
fρ0 = fx0 x0ρ + fy0 yρ0 + fz0 zρ0 = fx0 (sin φ cos θ) + fy0 (sin φ sin θ) + fz0 (cos φ),
fφ0 = fx0 x0φ +fy0 yφ0 +fz0 zφ0 = fx0 (ρ cos φ cos θ)+fy0 (ρ cos φ sin θ)+fz0 (−ρ sin φ)
Consider the quantity −2ρ sin2 φfρ0 − sin φ cos φfφ0 . The first partials in
this quantity have coefficients
fx0 : −ρ(2 sin3 φ cos θ + sin φ cos2 φ cos θ)fx0 = −ρ sin φ cos θ(2 sin2 φ + cos2 φ)fx0
fy0 : −ρ(2 sin3 φ sin θ + sin φ cos2 φ sin θ)fy0 = −ρ sin φ sin θ(2 sin2 φ + cos2 φ)fy0
fz0 : −ρ(2 sin2 φ cos φ − sin2 φ cos φ)fz0 = −ρ cos φ sin2 φfz0
Which are exactly the coefficients derived above in the sum ρ2 sin2 φfρρ
00
+
2 00 00 2 0
sin φfφφ + fθθ . Hence the first partials will become −2ρ sin φfρ −
sin φ cos φfφ0 . Putting this all together,
ρ2 sin2 φfρρ
00
+sin2 φfφφ
00 00
+fθθ = ρ2 sin2 φ(fxx
00 00
+fyy 00
+fzz )−2ρ sin2 φfρ0 −sin φ cos φfφ0
Rearranging this give
ρ2 sin2 φ(fxx
00 00
+ fyy 00
+ fzz ) = 2ρ sin2 φfρ0 + ρ2 sin2 φfρρ
00
+ sin φ cos φfφ0 + sin2 φfφφ
00 00
+ fθθ
= sin2 φ[2ρfρ0 + ρ2 φfρρ
00
] + sin φ[cos φfφ0 + sin φfφφ
00 00
] + fθθ
00 00 00 sin2 φ 2 0 0 sin φ 0 0 1
fxx + fyy + fzz = 2 [ρ fρ ]ρ + 2 2 [sin φfφ ]φ + 2 f 00
2
ρ sin φ ρ sin φ ρ sin2 φ θθ
1 ∂ ∂f 1 ∂ ∂f 1 ∂ 2f
= 2 [ρ2 ] + 2 [sin φ ] + 2 2
ρ ∂ρ ∂ρ ρ sin φ ∂φ ∂φ ρ sin φ ∂θ2
which was to be shown.

20. Prove that if a function f(x, y) satisfies the Laplace equation


00 00
fxx + fyy = 0, then the function g(x, y) = f(x/(x2 + y 2 ), y/(x2 +
y )), x + y 2 > 0, also satisfies the Laplace equation.
2 2

Solution: Let u = x/(x2 + y 2) and v = y/(x2 + y 2). Then g(x, y) =


f(u, v). By the chain rule
00 00
gxx = fuu (u0x)2 + fvv
00
(yx0 )2 + fu0 u00xx + fv0 vxx
00 00 0 0
+ 2fuv ux v x
360 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

00 00 y 2 − x2 2 00 −2xy 2 2
0 2x(x − 3y )
2 2
0 2y(3x − y )
2
gxx = fuu ( ) + f vv ( ) + f u ( ) + f v ( )
(x2 + y 2)2 (x2 + y 2)2 (x2 + y 2)3 (x2 + y 2 )3
00 y 2 − x2 −2xy
+2fuv ( 2 )( 2 )
(x + y ) (x + y 2)2
2 2

4 2 2 4
00 y − 2x y + x 00 4x2 y 2 2
0 2x(x − 3y )
2 2
0 2y(3x − y )
2
= fuu ( ) + fvv ( 2 ) + fu ( ) + fv ( )
(x2 + y 2)4 (x + y 2)4 (x2 + y 2)3 (x2 + y 2 )3
2 2
00 −4xy(y − x )
+fuv ( )
(x2 + y 2)4
00 00 −2xy 2 00 x2 − y 2 2 2
0 2x(3y − x )
2 2
0 2y(y − 3x )
2
gyy = fuu ( 2 ) + f vv ( ) + f u ( ) + f v ( )
(x + y 2)2 (x2 + y 2)2 (x2 + y 2)3 (x2 + y 2 )3
00 −2xy x2 − y 2
+2fuv ( 2 )( )
(x + y 2)2 (x2 + y 2)2
00 4x2 y 2 4
00 y − 2x y + x
2 2 4 2
0 −2x(x − 3y )
2 2
0 −2y(3x − y )
2
= fuu ( 2 ) + f vv ( ) + fu ( ) + fv ( )
(x + y 2)4 (x2 + y 2)4 (x2 + y 2 )3 (x2 + y 2 )3
2 2
00 4xy(y − x )
+fuv ( )
(x2 + y 2)4
So,
00 00 00 y 4 + 2x2 y 2 + x2 4 2 2
00 y + 2x y + x
2
1
gxx +gyy = fuu ( )+fvv ( )=( 2 )(f 00 +f 00 ) = 0
2
(x + y ) 2 4 2 2
(x + y ) 4 (x + y 2)2 uu vv
if x2 + y 2 6= 0, which is assumed. The above is established since
00 00
fuu + fvv = 0.

21. Prove that if a function f(x, t) satisfies the diffusion equation


ft0 = a2 fxx
00
, a > 0, then the function
1 2 2 x x
g(x, t) = √ e−x /(4a t) f( 2 , − 4 ), t > 0
a t a t at
also satisfies the diffusion equation.

Solution: Let √ u(x, t) = x/(a2t) and v(x, t) = −x/(a4t). Define


2 2
h(x, t) = 1/(a t)e−x /(4a t). Then g(x, t) = h(x, t)f(u, v). By the
chain and product rule,
gt0 = h0t f + hft0 = h0t f + h(fu0 u0t + fv0 vt0 )
00
gxx = (h0x f + hfx0 )0x = (h0x f + h(fu0 u0x + fv0 vx0 ))0x
= h00xx f + h0x fx0 + h0x (fu0 u0x + fv0 vx0 ) + h(fu0 u0x + fv0 vx0 )0x
= h00xx f + 2h0x (fu0 u0x + fv0 vx0 ) + h(fuu 00
(u0x )2 + fvv
00
(vx0 )2 + 2fuv
00 0 0
vx ux + fu0 u00xx + fv0 vxx
00
)
7. CHAIN RULES AND IMPLICIT DIFFERENTIATION 361

The necessary partial derivatives are

1 −x2 /(4a2t) 1 x2 −x2 /(4a2t) 1 x2


h0t = − e + √ ( )e = − h + h
2at3/2 a t 4a2 t2 2t 4a2 t2
1 2 2 x
h0x = √ (−2x/(4a2 t))e−x /(4a t) = − 2 h
a t 2a t
1 x 1 x2 1
h00xx = − 2 h − 2 h0x = − 2 h + 4 2 h = 2 h0t
2a t 2a t 2a t 4a t a
x 1 x 1
u0t = − 2 2 = − u, vt0 = 4 2 = − v
at t at t
1 1
u0x = = u, uxx = 0 00
a2 t x
1 1
vx0 = − 4 = v, vxx 00
=0
at x
Then,
h xh 1
gt0 = h0t f − (fu0 u + fv0 v) = h0t f − 2 2 (fu0 − fv0 2 )
t at a
0 2 2
00 1 0 2hx 0 0 00 u 00 v 2uv 00
gxx = 2 ht f + (fu u + fv v) + h(fuu ( 2 ) + fvv ( 2 ) + 2 fuv )
a x x x x
1 −xh 1 h 00 1 2 00
= 2 h0t f + 4 2 (fu0 − fv0 2 ) + 4 2 (fuu 00
+ fvv ( 4 ) − 2 fuv )
a at a at a a
23. Show that the function g(x, y) = xn f(y/x2 ), where f is a differen-
tiable function, satisfies the equation xgx0 + 2ygy0 = ng.

Solution: Let u = y/x2. Then g(x, y) = xn f(u). By the chain


rule,

gx0 = nxn−1 f(u) + xn f 0 (u)u0x = nxn−1 f(u) + xn f 0 (u)(−2y/x3) = nxn−1 f(u) − 2xn−3 yf 0(u)
gy0 = xn f 0 (u)u0y = xn f 0 (u)(1/x2 ) = xn−2 f 0 (u)

Therefore,

xgx0 + 2ygy0 = nxn f(u) − 2xn−2 yf 0(u) + 2yxn−2 f 0 (u) = nxn f(u) = ng

24. Show that the function g(x, y, z) = xn f(y/xa, z/xb ), where f is a


differentiable function, satisfies the equation xgx0 + aygy0 + bzgz0 = ng.

Solution: Let u = y/xa and v = z/xb . Then g(x, y, z) = xn f(u, v).


362 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

By the chain rule,


gx0 = nxn−1 f(u, v) + +xn (fu0 u0x + fv0 vx0 ) = nxn−1 f(u, v) + xn (−ayfu0 /xa−1 − bzfv0 /xb−1 )
gy0 = xn (fu0 u0y + fv0 vy0 ) = xn (fu0 /xa)
gz0 = xn (fu0 u0z + fv0 vz0 ) = xn (fv0 /xb )
Therefore,
xgx0 +aygy0 +bzgz0 = nxn f(u, v)−axnyfu0 /xa −bxnzfv0 /xb +ayxn(fu0 /xa )+bzxn(fv0 /xb ) = ng
Observe that Exercise 23 is a special case of this with a = 2 and z = 0.

25–27. Assume in each case below that the given equation determines
z implicitly as a function of x and y. Find the first partial derivatives
of this function z(x, y):
25. x + 2y + 3z = ez ;

Solution: First, we must check that the hypothesis of Theo-


rem 22.2 are satisfied. We let F (x, y, z) = x+2y +3z −ez . F is
the sum of continuous functions, so it is continuous. Moreover,
it is the sum of differentiable functions, so it is differentiable.
In particular, Fz0 = 3 − ez , which vanishes only at z = ln 3.
Applying Theorem 22.2, we have
Fx0 1
zx0 = − 0
= z
Fz e −3
0
Fy 2
zy0 = − 0 = z
Fz e −3
except for z = ln 3.

26. x − z = arctan(yz);

Solution: Let F (x, y, z) = x − z − arctan(yz). F is a contin-


uous and differentiable function. Fz0 = −1 − 1+yy2 z2 , which is
continuous everywhere. Fz0 = 0 when y 2z 2 + y + 1 = 0. Thus
we have
Fx0 1 1 + y2z2
zx0 = − = − =
Fz0 −1 − 1+yy2 z2 y 2z 2 + y + 1
Fx0 − 1+yz2 z2 z
zy0 =− 0 =− y = 2 2
Fz −1 − 1+y2 z2 y z +y+1
7. CHAIN RULES AND IMPLICIT DIFFERENTIATION 363

For all x, y, z such that y 2z 2 + y + 1 6= 0

27. x/z = ln(z/y) + 1.

Solution: Let F (x, y, z) = x/z − ln(z/y) − 1. F is con-


tinuous whenever z 6= 0 and when z/y > 0. Fz0 is given by
Fz0 = − zx2 − 1z , which is continuous when z 6= 0. Additionally,
Fz0 = 0 when x + z = 0, so we must exclude these points.
Therefore,
1/z z
zx0 = − =
− zx2 − 1
z
x+z
1/y z2
zy0 = − =
− zx2 − 1
z
xy + yz
For z/y > 0, x + z 6= 0.
28–30. Assume in each case below that the given equation determines
z implicitly as a function of x and y. Find the second partial derivatives
of this function z(x, y):
28. x + 2y + 3z = ln z;

Solution: First, we will find the first partial derivatives us-


ing the implicit function theorem. Then, we will calculate the
second partial derivatives using the chain rule.
Let F (x, y, z) = x + 2y + 3z − ln z. Then F is continuous
when z > 0, and Fz0 = 3 − 1z , which is continuous when z 6= 0.
Moreover, Fz0 = 0 when 3z = 1. So, we have
1 z 1 1/3
zx0 = − 1 = = +
3− z
3z − 1 3 3z − 1
2 2z 2 2/3
zy0 = − 1 = = +
3− z
3z − 1 3 3z − 1

The second partial derivatives then are


00 −(1/3)3zx0 z
zxx = 2
=−
(3z − 1) (3z − 1)3
00
−(2/3)3zy0 4z
zyy = = −
(3z − 1)2 (3z − 1)3
00
−(1/3)3zy0 2z
zxy = 2
=−
(3z − 1) (3z − 1)3
364 3. DIFFERENTIATION OF MULTIVARIABLE FUNCTIONS

for z 6= 1/3.

29. x + z = arctan(yz);

Solution: Let F (x, y, z) = x + z − arctan(yz). Following


a similar procedure in Exercise 28, we will use the implicit
function theorem to find the first partials. The functions F
and Fz0 = 1 − y/(1 + y 2 z 2) are continuous for all x, y, z. Note
that Fz0 is zero when z 2 = (y − 1)/y 2 . By the implicit function
theorem,
1 1 + y 2z 2 y
zx0 = − 2 2
= 2 2
= −1 +
1 − y/(1 + y z ) y−y z −1 y − y z2 − 1
2

−z/(1 + y 2z 2 ) z
zy0 = − 2 2
=
1 − y/(1 + y z ) 1 + y z2 − y
2

Then the second partials are


00 y(−2y 2zzx0 ) −2y 3 z(1 + y 2z 2 )
zxx = =
(y − y 2z 2 − 1)2 (y − y 2z 2 − 1)3
00
zy0 (1 + y 2z 2 − y) − z(2yz 2 + y 2(2zzy0 ) − 1)
zyy =
(y − y 2z 2 − 1)2
z(1 + y 2z 2 − y) − z(2yz 2 + 2y 2 zzy0 − 1)(1 + y 2z 2 − y)
=
(y − y 2z 2 − 1)3
z(1 + y 2z 2 − y) − z(2yz 2 − 1)(1 + y 2z 2 − y) − z(2y 2 zzy0 )(1 + y 2 z 2 − y)
=
(y − y 2 z 2 − 1)3
z(2 − 2yz 2)(1 + y 2 z 2 − y) − 2y 2z 3
=
(y − y 2 z 2 − 1)3
00
(y − y 2z 2 − 1) − y(1 − 2yz 2 − 2y 2 zzy0 )
zxy =
(y − y 2 z 2 − 1)2
y − y 2z 2 − 1 − y + 2y 2z 2 + 2y 3 zzy0
=
(y − y 2 z 2 − 1)2
−1 + y 2z 2 + 2y 3zzy0
=
(y − y 2z 2 − 1)2

30. xz = eyz .

Solution:

You might also like