page 1 of Section 2.

CHAPTER 2 COORDINATE SYSTEMS

SECTION 2.0 REVIEW
the velocity vector
Suppose the position at time t of a particle moving in 2—space is given by the
parametric equations
x = x(t), y = y(t).

The position vector

Û Û Û
r(t) = x(t)i + y(t)j

is drawn with its tail at the origin and its head at the point (x(t), y(t)) on the
path (Fig 1).
The velocity vector
Û Û Û
v(t) = x'(t)i + y'(t)j

is drawn with its tail at the point (x(t),y(t)) on the path (Fig 1). It points in
the instantaneous direction of motion (in the direction of increasing t) so it's
Û
tangent to the curve. The instantaneous speed of the particle at time t is «v(t)«.
Á
the differentials ds and ds
Look at the path x=x(t), y=y(t).
Change t by dt so that a little piece of path is traced out. Let ds be the length
of the little piece (Fig 2). Then

(1) ds = «v« dt (because distance = speed ≈ time)

In other words,

ds =
√
dx 2 dy 2
( ) + ( )
dt dt
dt

I'd like to get an arrow that approximates the little piece of path (Fig 3). The
arrow should be tangent to the little piece, point in the direction of increasing t,
and have the same length as the little piece (i.e., have length ds). The arrow v(t)
itself has the right direction but its length is wrong. Its length is how far the
1
particle travels in one second. If the particle only travels say 10 of a second then
1
the approximating arrow should be 10 v(t). Here, the particle travels dt seconds. So,
in general:
If x = x(t), y = y(t) and t changes by dt (Fig 3) the
piece of path traced out can be approximated by the arrow
Û dx Û dy Û
(2) v(t) dt = ( i + j ) dt
dt dt
Á Á Á
This arrow is variously called dr, ds, d.

•
t=3
→ t=2•
v path
x=x(t) Û
t=1• y=y(t) v(t)dt
→ length
r ds • time t+dt •
time t+dt
• time t • time t
FIG 1 FIG 2 FIG 3
 page 2 of Section 2.0

graphing parametric equations

I'll get the graph of the parametric equations x = t, y = t2 (which I think of as
the position of a particle at time t).
Eliminate the parameter by inspection and get y = x2.
The graph is a parabola (Fig 1). The detailed path of the particle (including
direction and timing) is in Fig 2.

y-axis y-axis
t=
-3 t=3
• •
(-3,9) (3,9)

t= -
2 =2
(-2,4) • • t
(2,4)

t= -1 • • t=1
•t x-axis
x-axis =0

FIG 1 FIG 2

But it's not always that straightforward.

Look at the parametric equations x = t2, y = t4.

Again, y = x2. But this time x ≥ 0 (since x = t2) and the graph is only half a
parabola (Fig 3)
In particular, if the parametric equations describe the position of a particle at
time t then the particle moves down the right half of the parabola and then turns
around and goes back up again (Fig 4)
(In Fig 4, I drew the down and up paths as if they were different curves so you
could see the motion better. But it's all supposed to be happening on the one
parabola y = x2.)

NOTICE. Even though I eliminated the parameter to get y = x2, the graph of the
parametric equations is not the entire parabola y = x2. It's just the right half of
the parabola.
t=
-2 =2
• t
(4,16)
y-axis
y-axis

t=
-1
• t=1
(1,1)
x-axis t=0 •
x-axis

FIG 3 FIG 4
 page 3 of Section 2.0

example 1
Sketch the graph of the parametric equations x = sin t, y = sin2 t.
solution
Again y = x2 but this time -1 ≤ x ≤ 1 and 0 ≤ y ≤ 1 so the graph is only a portion
of the parabola (Fig 5)

If you think of the equations as the path of a particle then the particle moves
back and forth forever on a piece of the parabola (Fig 6). For instance, as t goes
from -π/2 to π/2 the particle moves along the parabola from point (-1,1) to point
(1,1); as t goes from π/2 t0 3π/2 the particle moves down the parabola from point
(1,1) back to (-1,1).
2
3π/ (1,1)
• t=
=
•
(-1,1) t
y-axis
/2 π/2
−π 2π
t= t= π
(-1,1) • • (1,1)
• t=
t=
x-axis 0

FIG 5 FIG 6

In general, suppose you want the graph of the parametric equations

x = x(t)
y = y(t)

and you eliminate the parameter to get a single equation in x and y.

The graph of the parametric equations might be only part of the graph of that single
equation.
You should look carefully at the parametric equations themselves to decide on the
extent of the parametric graph.

Eliminating the parameter does not necessarily tell the whole story.

example 2
Sketch the graph of the parametric equations
x = et
y = 3

solution
The graph lies on the line y = 3 (it takes no effort to "eliminate" the parameter).
But the graph of the parametric equations is only part of line y = 3, the part where
x > 0 (Fig 7) because et is always > 0.

Fig 8 shows the details.

The particle "starts" at point (0,3) when t = -™ and moves to the right forever.
00
00

y
−∞

y
-1

5
0
t=

t=

t=
t=

3 o • • •
3o

x ' x
1
FIG 7 FIG 8
 page 4 of Section 2.0

two famous sets of parametric equations

Let a and b be positive numbers.
Look at the parametric equations

x = a cos t
y = a sin t (the parameter is t)

Then
x2 + y2 = a2 cos2 t + a2 sin2 t
= a2 (cos2 t + sin2 t)
= a2

The graph of the parametric equations is the circle with plain equation
x2 + y2 = a2. (The particle travels counterclockwise and t is the usual angle œ.)

Look at the parametric equations

x = a cos t
y = b sin t (the parameter is t) (where a ≠ b)

Then
x
= cos t
a
y
= sin t
b
x 2 y 2
( ) + ( ) = cos2 t + sin2 t
a b
= 1

The graph of the parametric equations is the ellipse with plain equation
x2 y2
+ = 1 (Fig 9). (The particle travels counterclockwise and t is not the usual
a2 b2
angle œ except at the "corners".)
b

FIG 9
 page 1 of Section 2.1

SECTION 2.1 INTRODUCTION

defining a 2-dim coordinate system

Equations of the form

x = x(u,v), y = y(u,v)

can be thought of as relating a u,v coordinate system and an x,y Cartesian

coordinate system, both in the same plane.
A curve of the form u = u0 is called a v-curve; only v is changing on it.
A curve of the form v = v0 is called a u-curve; only u is changing on it.

example 1 (parabolic coordinates)

The parabolic coordinate system is defined by
1
(1) x = 2(u2 - v2), y = uv, v ≥ 0.
I want to draw some u—curves and v—curves.

The v-curve u=2

The curve has parametric equations
1
x = 2(4 - v2)
(2) y = 2v
v ≥ 0 (the parameter is v).

To help identify the curve, eliminate the parameter (one way to do it is to solve
the second equation for v and substitute in the first equation) to get the plain x,y
equation
1 1
x = (4 - y2),
2 4

which you should recognize as a parabola. But notice from (2) that v ≥ 0 and
y = 2v so y ≥ 0. The v—curve u=2 is just the upper half of the parabola (Fig 1).

The v-curve u = -2
The curve has parametric equations
1
x = 2(4 - v2)
(3) y = -2v
v ≥ 0

Eliminate the parameter to get the plain equation

1 1
x = 2(4 - 4 y2),

the same parabola as before. Notice from (2) that y ≤ 0 this time since v ≥ 0 so
the curve u = -2 is the lower half of the u=2 parabola.

The u-curve v=1

The curve has parametric equations
1
x = 2(u2 - 1)
y = u (the parameter is u).

It's a parabola with the plain equation

1
x = 2(y2 - 1)
 page 2 of Section 2.1

Fig 1 shows a piece of parabolic coordinate paper.

The parabolic coordinates u and v don't have geometric significance like Cartesian
coordinates x and y (over and up). In Fig 1, I plotted point P with coordinates u=2,
v=1 by finding the intersection of the curves u=2, v=1.
y-axis
u=2 v=2

v=3/2

u=1
v=1
P
•
v=1/2

u=0 v=0 x-axis

u=-1

u=-2

FIG 1 parabolic coordinate paper

polar coordinates
The polar coordinate system is defined by the equations

x = r cos œ, y = r sin œ, r ≥ 0

The coordinates r and œ have the physical significance shown in Fig 2.

r is the distance to the origin; r = √

x2+y2.

œ is the angle measured counterclockwise from the positive x—axis.

Fig 3 shows the œ—curve r=3 (a circle) and the r—curve œ=π/4 (a half—line).

θ=π/4
3
r=

• r
θ

•
FIG 2 FIG 3 FIG 4

œ versus Arctan[x,y] versus Arctan y/x

Look at point (-2, -2) (Fig 4). One person could say that it has angle 5π/4.
Another person could say it has angle -3π/4. Another could say that it has angle
13π/4 etc. The coordinate œ is not unique.
Fig 5 show some points with œ chosen between 0 and 2π.
Fig 6 shows the same points but œ chosen between -π and π.

The computer program Mathematica chooses to measure the polar coordinate œ of point
(x,y) in the -π to π range and uses the notation ArcTan[x,y] to stand for that
unique value of œ (this ArcTan is a function of two variables).
 page 3 of Section 2.1

In[3]:=
ArcTan[-2, -2]
Out[3]=
-3 Pi
-----
4

−π/2
π/2

π/2
3π • −π
π/ 2

3π /4 / 4 /4 • 4
• •π
/4
π/
4 • • • π/
• •
π
−π• •0 0 •
π• •0
•0
2π • • • •-
• • / 4 • π/4 • π

−π
/4 • 7π π
5π −3

−π/2

−π/2
/4

/4
/4

π/2
3π/2

one set of œ values values of ArcTan[x,y] values of ArcTan y/x

(another set of œ values )
FIG 5 FIG 6 FIG 7

ArcTan[y/x] is another story (this ArcTan is a function of one variable).

Too often, students use the notation Arctan y/x for the polar coordinate œ of point
(x,y). Not right. Don't do it. Here's what Arctan y/x actually means.

Arctan y/x is the angle between -π/2 and π/2 whose tangent is y/x

Fig 7 shows values of ArcTan y/x at various points.

For instance, for the point (-2, -2), a value of œ is 5π/4. Another is -3π/4 etc.
You won't get any of these using ArcTan y/x:

y/x = -2/-2 = 1.
There are many angles whose tangent is 1, including π/4, -3π/4, -7π/4 etc.
ArcTan picks the one between -π/2 and π/2, namely π/4 which is not an angle
for the original point at all.
In[23]:=
ArcTan[-2/-2]
Out[23]=
Pi
---
4
warning ArcTan[x,y] is one of the œ values of the point (x,y), the one
between -π and π.
ArcTan[y/x] may not even go with point (x,y) at all.
It will if the point is in quadrants I or IV but not otherwise.
So, in general, do not write that œ = arctan y/x

3-dim u,v,w coordinate systems

Equations of the form

x = x(u,v,w), y = y(u,v,w), z = z(u,v,w)

can be thought of as relating a u,v,w coordinate system, with the usual x,y,z system.
The graph of u = u0 is a surface, called a coordinate surface.
Similarly, other coordinate surfaces are v = v0 and w = w0.
A u—curve is a curve of the form v=v0, w=w0 (only u changes on it).
A v—curve is a curve of the form u=u0, w=w0 (only v changes on it).
A w—curve is a curve of the form u=u0, v=v0 (only w changes on it).
 page 4 of Section 2.1

cylindrical coordinates
The equations

x = r cos œ, y = r sin œ, z = z, r ≥ 0

(i.e., z together with polar coordinates) define cylindrical coordinates.

Fig 8 shows the physical significance of the coordinates:

r is the distance to the z—axis; r = 
√
x2+y2.
œ is the angle around from the x,z plane.
z measures up and down from the x,y plane as usual.

Fig 9 shows some coordinate surfaces and coordinate curves:

The coord surface r=3 is a cylinder.
The coord surface œ=π/4 is a half plane.
The coord surface z=4 is a plane.
The œ—curve r=3, z=4 is a circle.
The r—curve œ=π/4, z=4 is a half—line.
The z—curve r=3, œ=π/4 is a line.

z-axis

r=3
θ=π/4
z=4
θ -curve
r-cu
rve
z-curve
z-axis

r
θ y-axis
•
y-axis
r z x-axis
θ
x-axis
cylindrical coords r,œ,z

FIG 8 FIG 9
 page 5 of Section 2.1

spherical coordinates
The spherical coord system is defined by the equations

x = ® sin ƒ cos œ
(4) y = ® sin ƒ sin œ
z = ® cos ƒ
0 ≤ ƒ ≤ π, ® ≥ 0

Fig 10 shows spherical (along with cylindrical and Cartesian) coordinates:

® is the distance to the origin; ® = 

√
x2+y2+z2.
œ is the angle around from the x,z plane (same as the cylindrical coord œ).
ƒ is the angle down from the positive z—axis.

If you remember Fig 10 you can figure out the various relations among the three sets of coordinates.

z-axis

curve
ρ-
φ=π/6

z-axis θ=π/4
θ -curve

θ r

e
φ -curv
•P ρ=3
z y-axis
ρ
φ z

θ y-axis
x
r
x-axis
y
x-axis
cylindrical coords
spherical coords ®, ƒ, œ curves
Cartesian coords

FIG 10 FIG 11

Fig 11 shows the coordinate surfaces and coordinate curves:

The coordinate surface ®=3 is a sphere.
The coordinate surface ƒ=π/6 is a cone.
The coordinate surface œ=π/4 is a half—plane.
The ®—curve œ=π/4, ƒ=π/6 is the ray of intersection of a half—plane and a cone.
The ƒ—curve ®=3, œ=π/4 is the great semicircle of intersection of a sphere and a
half—plane.
The œ—curve ®=3, ƒ=π/6 is the circle of intersection of a sphere and a cone.

Remember Figs 10 and 11

 page 6 of Section 2.1

Fig 12 gives a cartographer's view of spherical coordinates. If ®=®o then the point
lies on a sphere centered at the origin with radius ®o, an "earth".
On the earth, ƒ measures "down" from the north pole N. On the great circle NASBN, ƒ
goes from 0o at N, to 90o at A, to 180o at S, then back to 90o at B and finally back
to 0o at N. Each parallels of latitude (including the equator) is a circle on which
ƒ and ® are fixed (a œ—curve).
On the earth, œ measures "around" from the prime meridian in the x,z plane. Each
meridian of longitude is a great semicircle on which œ and ® are fixed (a ƒ—curve).

North pole
N•

ρ=ρο
para
llel
of latitude
e

meridian
longitud

y
equator
an of

•
A
di
eri m

S•
South pole
FIG 12

warning
In a u,v system the curve v=2 is not a v—curve. It's a u—curve (because u is varying
on it).
In a u,v,w system in 3—space, v=2 is not a v—curve. It's not a curve at all. It's a
surface. A v—curve is a curve of the form u=u0, w=w0 on which only v varies.
u,v,z cylindrical coordinates
Any 2—dim u,v coordinate system can be turned into a 3—dim u,v,z in a natural way
as follows.
Suppose x = x(u,v), y = y(u,v) define the zippo coord system in 2—space with the
u—curves and v—curves shown in Fig 13. Then the equations

x = x(u,v), y = y(u,v), z = z

define the zippo cylindrical coord system. Fig 14 show some coordinate surfaces, called
cylinders (I couldn't fit them all into one picture).
To draw the picture of the surface u=4 in the lefthand diagram in Fig 14, I copied
the curve u=4 from Fig 9 and pasted it above the original curve and drew some
connecting vertical lines.
Similarly, the surface v=7 in the center diagram in Fig 14 is gotten by sliding the
v=7 curve from Fig 9 up and down in the z direction. The surfaces z=3 and z=5 in the
righthand diagram in Fig 10 are planes parallel to the x,y plane.
The u—curve v=7, z=5 in Fig 15 is like the curve v=7 from Fig 3 but raised up 5;
it's the intersection of the surface v=7 with the plane z=6. The z—curve u=4,v=7 in
Fig 15 is the vertical line of intersection of the surface u=4 and the surface v=7.
 page 7 of Section 2.1

y-axis

x-axis

u=5

u=4

v=8 v=7

The 2—dim zippo coord system

FIG 13

z
z v=7
z=5
z
y
y z=3
u=4
x y
x
x

Some coordinate surfaces in the 3—dim zippo cylindrical coordinate system

FIG 14

z—curve
z
u=4 z
ur ve z
z=5
u—c

7
v=
z=5 v—curve
y
y v=7
y x
x
u=4
x

A u—curve, v—curve, z—curve in the zippo cylindrical coord system

FIG 15

PROBLEMS FOR SECTION 2.1

1. Let x = 2u + 3v, y = u - v define a u,v coordinate system.

(a) Identify the u—curve that goes through the point x = 0, y = 5 (find parametric
equations for it and then a plain equation).
(b) Identify the u—curve v = 1 (find parametric equations for it and then a plain
equation).
(c) Identify the v—curve u = 0.
(d) Identify the v—curve u = u0 and the u—curve v = v0 in general and then sketch a
piece of u,v coordinate paper.
 page 8 of Section 2.1

2. Let x = er cos œ, y = er sin œ define a new coordinate system.

(a) Identify the œ-curve r=3.
(b) Identify the r—curve œ = π/7.
(c) Sketch a piece of coordinate paper with enough curves on it to make the pattern
clear.

3. Suppose the restriction v ≥ 0 is removed from the parabolic coordinate system in

example 1.
(a) How does the v—curve u=2 change.
(b) Draw a piece of the new coordinate paper.
(c) What is there about the new coordinate paper that led to the decision to
make v ≥ 0.

4. Sketch in spherical coords

(a) the coordinate surface ƒ = 120o
(b) the œ—curve ƒ = 120o, ® = 4
(c) the coordinate surface œ = 80o
(d) the ƒ—curve ® = 2, œ = -20o

5. (a) Sketch the r—curve œ = -π/2, z = 5 in cylindrical coordinates.

(b) Sketch the œ—curve r = 4, z = 1 in cylindrical coordinates.
(c) sketch the ®—curve œ = -π/2, ƒ = 30o in spherical coordinates.

6. Parabolic cylindrical coordinates are defined by

1
x = 2(u2 - v2), y = uv, z = z, v ≥ 0 (parabolic coords plus z).
Sketch
(a) the coordinate surface v=1
(b) the coordinate surface u=2
(c) the coordinate surface z=3
(d) the u—curve, v—curve, and z—curve through the point P where u=2,v=1,z=3
 page 1 of Section 2.2

SECTION 2.2 BASIS VECTORS

Û Û
the vectors eu and ev
In a Cartesian coordinate system, vectors are expressed in terms of the basis
Û Û
vectors i and j. Here are the basis vectors used at the point u=u0, v=v0 in a
coordinate system defined by x = x(u,v), y= y(u,v) (Fig 1):

Û
eu is a unit vector tangent to the u—curve v=v0 and of the two tangent directions
it points toward increasing u.
Û
ev is a unit vector tangent to the v—curve u=u0 and of the two tangent directions
it points toward increasing v.

Fig 1 shows eu and ev at the point u=1, v=3 in a hypothetical u,v coord system.

eu

v=3
e
v

u=2
v=7 u=1

FIG 1

Here's how to find eu and ev algebraically. The u—curve v=v0 has parametric
equations
x = x(u,v0), y = y(u,v0) (the parameter is u)

If you think of u as "time" then the velocity vector is

ºx Û ºy Û
velu = i + j
ºu ºu

Then velu is tangent to the curve and points in the direction of increasing u so

Û ºx Û ºy Û
eu = (velu)unit = ( i + j )unit
ºu ºu
(1) Û ºx Û ºy Û
ev = (velv)unit = ( i + j )unit
ºv ºv

Similarly for a u,v,w coordinate systems in 3—space,

Û ºx Û ºy Û ºz Û
eu = (velu)unit = ( i + j + k )unit
ºu ºu ºu
Û ºx Û ºy Û ºz Û
(2) ev = (velv)unit = ( i + j + k )unit
ºv ºv ºv

Û ºx Û ºy Û ºz Û
ew = (velw)unit = ( i + j + k )unit
ºw ºw ºw
 page 2 of Section 2.2

basis vectors in polar, cylindrical and spherical coordinates

In polar coordinates (Fig 2), er points away from the origin (along an r—curve)
and eœ points "around" counterclockwise.
Û Û Û Û
Note that unlike i and j, the basis vectors er and eœ vary from point to point.
In cylindrical coordinates (Fig 3), er points away from the z—axis, tangent to an
r—curve; eœ points "around" counterclockwise as seen from above, tangent to a œ—
curve; ez = k, tangent to a z—curve.
The cylinder in Fig 3 is r = constant. At points on that cylinder,
er is perp to the cylinder
ez lies on the cylinder
eœ is tangent to the cylinder.

z-curve
e
θ e
r
e z
r ez

e
θ
θ-curv er
e e
θ r-c
urv
e
θ y e
er x
FIG 2 polar coords FIG 3 cylindrical coords

In spherical coordinates (Fig 4), e® points away from the origin, tangent to a ®—
curve; eƒ points "down" tangent to a circle of longitude, a ƒ—curve; eœ points
"around" counterclockwise as seen from above (as in polar and cylindrical coords),
tangent to a œ—curve..
In Fig 4, the sphere is ® = constant. At points on the sphere,

e® is perp to the sphere

eœ is tangent to the sphere
eƒ is tangent to the sphere

In Fig 4, the first cone is ƒ = ƒo where 0 < ƒo< π/2; the second cone is ƒ = ƒo
where π/2 < ƒo < π. For each cone, at any point,

e® lies on the cone

eœ is tan to the cone
eƒ is perp to the cone
 page 3 of Section 2.2

z eρ y
eρ
x
e e
• θ • θ •
e
y θ
eφ eφ eρ
eφ
x y
x
FIG 4 spherical coords

Û Û Û
Now I want to find the components (w.r.t. i,j, k) of these basis vectors. It's
easier for er, eœ, ez and e® because it's easy to describe their directions (out,
around, up, out).

(3a) erin polar and cylindrical = (xi + yj)unit

(this is away from the origin in 2-space
and away from the z—axis in 3—space)
x y
= i + j

√ 2
x +y2 
√ 2
x +y2

r cos œ r sin œ
= i + j
r r

= cos œ i + sin œ j

(3b) eœ in polar, cylindrical and spherical

= er = (-yi + xj)unit = -sin œ i + cos œ j
left turn
(3c) ez = k (by inspection)

(3d) e® = (xi + yj + zk)unit

x y z
= i + j + k

√ x2+y2+z2 
√ x2+y2+z2 
√ x2+y2+z2
Û Û Û
= sin ƒ cos œ i + sin ƒ sin œ j + cos ƒ k

To find eƒ, I'll use (2):

ºx ºy ºz
eƒ = ( i + j + k )unit
ºƒ ºƒ ºƒ

= (® cos ƒ cos œ i + ® cos ƒ sin œ j - ® sin ƒ k)unit

Normalizing factor = √

®2 cos2ƒ cos2œ + ®2 cos2ƒ sin2œ + ®2 sin2ƒ

= ® √
cos2ƒ (cos2œ + sin2œ) + sin2ƒ
1

= ®
√
cos2 ƒ + sin2 ƒ

= ®
 page 4 of Section 2.2

So

(3e) eƒ = cos ƒ cos œ i + cos ƒ sin œ j - sin ƒ k

These are all on the reference page (but an exam could ask you to explain where
they came from).

example 1 (away from the origin and around the origin)

Û Û
Let F(x,y) = x i + y j(Fig 5). The F vector attached to point (x,y) points away
from the origin and has length 
√
x2 + y2 = r (the polar coordinate r). So

Û Û Û
F = x i + y j = r er (Fig 5)
Û Û
Let G(x,y) = -y i + x j (Fig 6). Then G is a left turn from F and has the same length
as F, so
Û Û Û
G = -y i + x j = r eœ (Fig 6)

Let H(x,y,z) = xi + yj + zk. The H vector attached to point (x,y,z) points away from
the origin and has length √
x2 + y2 + z2 = ®. So
Û
H = xi + yj + zk = ® e®

FIG 5 xi + yj = rer FIG 6 -yi + xj = reœ

famous electric force fields (do not memorize formulas)

Let F be the electric force field (force per unit charge) in 3-space due to a unit charge at the origin.
The vectors point away from the origin and at a point which is distance d from the
origin the vector has length 1/d2 (Fig 7). In Cartesian coords,
1 1 Û Û Û
F(x,y,z) = (x,y,z)unit = (xi + yj + zk)
2 2
x + y + z 2 3/2
(x2 + y2 + z2)
and in spherical coords,
1
F(®,ƒ,œ) = e
®2 ®

Let F be the electric force field in 3-space due to a line of unit charge density along the z-axis . The
vectors point away from the z—axis and at a point which is distance d from the
z—axis the vector has length 1/d (Fig 8). The distance from (x,y,z) to the z—axis is

√ x2+y2 so in Cartesian coordinates,
1 x Û y Û
F(x,y,z) = (x,y,0)unit = i + j
2 2 2 2

√ x2 + y2 x +y x +y
 page 5 of Section 2.2

and in cylindrical coordinates,

1
F(r,œ,z) = e .
r r

Let F be the electric force field in 2-space due to a unit charge at the origin. The vectors point
away from the origin and at a point which is distance d from the origin the vector
has length 1/d2 (Fig 9) In Cartesian coordinates,

1 x Û y Û
F(x,y) = (x,y)unit = i + j
(x2 + y2) (x2+y2)3/2 (x2+y2)3/2

and in polar coords,

1
F(r,œ) =
e
r2 r
(The field in Fig 9 is a cross section of the field in Fig 7; i.e., the electric
field in 2-space due to a unit charge at the origin is part of the electric field in
3—space due to a unit charge at the origin.)
z
z

x
y
x
3—dim electric field electric field due 2—dim electric field
due to a point charge to a line of charge due to a point charge
FIG 7 FIG 8 FIG 9

example 2
Let
Û Û Û
G(x,y,z) = x
√
x2 + y2 + z2 i + y
√
x2 + y2 + z2 j + z
√
x2 + y2 + z2 k

Write G in spherical coords.

solution G = ®(xi + yj + zk) = ®…®e® = ®2 e®

example 3
Parabolic coordinates (example 1, Section 2.1) are defined by
1
x = 2(u2 - v2), y = uv, v ≥ 0
By (1), the basis vectors are
Û Û Û u Û v Û
eu = (ui + vj)unit = i + j

√ u2 + v2 
√ u2 + v2
(4)
Û Û Û -v Û u Û
ev = (-vi + uj)unit = i + j

√ u2 + v2 
√ u2 + v2

At point u = 2, v = 1 for instance,

Û 2 Û 1 Û Û 1 Û 2 Û
(5) eu = i + j and ev = - i + j (Fig 10)
5
√ 5
√ √ 5 5
√
 page 6 of Section 2.2

y-axis
v=2
u=2
ev
eu
u=1 v=1
eu x-axis
v=0
u=0
u=-1 eu
v=1
ev ev
u=-2
v=2
FIG 10

warning
1. Remember that eu and ev are unit vectors.
ºx Û ºy Û
After finding i + j you have to normalize it to get eu.
ºu ºu
2. If a point has coordinates u=2, v=1 don't call the point (2,1). This notation is
reserved for the point x=2, y=1 in the usual x,y system. You can use ((2,1)) or
(2,1)u,v but you must distinguish it from the usual (2,1).
If a vector p is 3eu + 7ev, don't wrote p = (3,7) because that notation is reserved
for 3i + 7j.

orthogonal coordinate systems

A 2—dim coordinate system is called orthogonal if eu and ev are perpendicular at
each point, i.e., if eu…ev = 0, i.e., if the u—curves and v—curves intersect
perpendicularly.
Similarly a 3—dim coordinate system is orthogonal if eu, ev, ew are pairwise
perpendicular at each point, i.e., if the u—curves, v—curves, w—curves intersect
perpendicularly, i.e., if the coordinate surfaces intersect perpendicularly.

Polar coordinates, cylindrical coordinates and spherical coordinates are orthogonal

coordinate systems (I can tell from the geometry -- see problem 14 for an algebraic
argument). So is the less famous parabolic coord system (see problem 13).

To test for orthogonality, you can test velu, velv and velw instead of the unit
vectors eu, ev, ew since they have the same directions respectively.

warning
The definition of eu and ev says nothing about perpendicularity. They are not
necessarily perp to each other or perp to u—curves and v—curves. It may happen, but
not necessarily.

example 4
Let x = 2u + 3v2, y = u - v.
Is the u,v coordinate system orthogonal.

solution
velu = 2i + j, velv = 6vi - j.
velu…velv = 12v - 1. This is not 0 at every point (it is only 0 at points where
v = 1/12). So the coord system is not orthogonal.
 page 7 of Section 2.2

physical significance of the scalars a and b when p = aeu + bev in an orthogonal u,v coord system
Û
First of all, remember that the phrase "component of p in direction eu" has a
precise meaning. It means "signed projection onto the eu line" (see Section 1.0)
Û
Look at Fig 11, where p = 3eu - 2ev.
The scalar 3 is the component of p in the direction of eu (i.e., signed projection)
(see (4) in Section 1.0), and the scalar -2 is the component of p in the direction
of ev. This worked because eu and ev in Fig 11 are orthogonal.
It doesn't work when eu and ev are not orthog (which is one of the many reasons why
we usually prefer to use orthog coord systems). In Fig 12, p = 3eu + 2ev. But the
component of p in direction eu is 5, not 3, and the component of p in direction ev
is 4, not 2. If p is a force vector and you want to know "how many pounds of force
push you in the eu direction", the answer is 5 pounds, not 3 pounds.
So what would you call the scalars 3 and 2 in Fig 12? You don't have to call them
anything. Or call them the first coord and second coord of p in the u,v coord
system. Or even the first component and second component of p. But don't call them
"the components of p in the eu and ev directions" because they are not signed
projections.

5
p
3

eu
3
p
eu e
ev v
2

2 4
FIG 11 FIG 12

how to convert a vector from Cartesian coords to u,v coordinates

Û
As just decided, in an orthogonal coordinate system, when p is written as
aeu + bev, the scalars a and b are the signed projections of p onto eu and ev.
By the formulas for signed projections ((4) in Section 1.0),

(6) p = (p…eu) eu + (p…ev) ev (Fig 13)

The formula in (6) doesn't work if the coordinate system is not orthogonal.
p

ev
(p • ev) ev eu

(p • eu )eu

FIG 13
 page 8 of Section 2.2

example 5
Let q = 3i + 2j.
Switch to the parabolic coordinate system (an orthogonal system).

(a) Write q in terms of the basis vectors at point u=2, v=1

(b) Write q in terms of the basis vectors at point u = 3, v=1.

solution
(a) I found the basis vectors in (5): At the point where u=2, v=1,

Û 2 Û 1 Û Û 1 Û 2 Û
eu = i + j, ev = - i + j
√ 5 5
√ √ 5 5
√
By (6),
8 1
q = (q…eu)eu + (q…ev)ev = e + e
5 u
√ 5 v
√
(b) The basis vectors in general are in (4). The basis vectors at point u = 3, v=1
are
3 1 -1 3
eu = i + j, ev = i + j.
10
√ 10
√ 10
√ 10
√
By (6),

11 3
q = e + e
10 u
√ 10 v
√

PROBLEMS FOR SECTION 2.2

1. Sketch eu and ev at point A and point B in the diagrams.

v=2
v=3
• v=2
A B• u=5
v=3
u=8

u=5
u=1

2. Sketch in cylindrical coordinates

(a) the coordinate surface r=3
(b) the coordinate surface œ = 2π/3
(c) the coordinate surface z = 1
(d) the r—curve, œ—curve and z—curve through the point r=3, œ = 2π/3, z=1 and the
basis vectors at the point.

3. Sketch in spherical coordinates

(a) the coordinate surface ®=2
(b) the coordinate surface ƒ = 2π/3
(c) the coordinate surface œ=2π/3
(d) the ƒ—curve, ®—curve, œ—curve through the point ƒ=π/6, œ = -π/3, ® = 1 and the
basis vectors at the point.

4. I found er and eœ (polar coords) in (3), mostly with geometry. Use the formula
in (1) to find them again and check that they are perpendicular.
 page 9 of Section 2.2

5. Sketch the vector field cos œ er in 2—space (polar coords).

6. Sketch the vector field veu in parabolic coordinates.

7. Plot these vector fields in 3—space. Draw enough arrows to make the pattern clear
and also write a brief description.
A picture in 3—space usually looks clearer if there are "props" in the picture,
like circles, spheres, cones, planes etc.
(a) reœ (cylindrical coords)
(b) ®eœ (spherical coords)
(c) zer (cylindrical coords)
Û
(d) n r k (cylindrical coords)
(e) sin ƒ e® (spherical coords)

8. Convert these vector fields to polar or spherical coordinates.

-y Û x Û x Û y Û Û Û Û
(a) i + j (b) - i - j (c) xi + yj + zk
2
x + y 2 2
x + y 2 2
x + y 2 2
x + y 2

9.(a) Find k…eƒ (a) algebraically and (b) geometrically.

10. Parabolic cylindrical coordinates (parabolic coords plus z) are defined by

1
x = 2(u2 - v2), y = uv, z = w, v ≥ 0

Sketch the basis vectors at the point P where u=2, v=1, z=3.

11. Verify that the parabolic coord system is orthogonal.

12. Let x = x(u,v), y = y(u,v) define a u,v coordinate system.

Find whichever of the following are possible with just this info:
ev…ev, eu…ev, «eu«

13. I thought it was clear from geometry that the spherical coordinate basis vectors
were orthogonal. Use (3b), (3(d), (3e) to check it algebraically.
Û
14. Let F = xi. Convert F to the parabolic coordinate system.

15. The diagram shows eu, ev and p.

(a) Estimate a and b so that p = aeu + bev.
(b) Estimate the component of p in the direction of eu (i.e., signed projection)
and the component of p in the direction of ev.
p
eu
ev
 page 10 of Section 2.2

16. Let x = u2 + uv, y = u - v.

(a) Is the coordinate system orthogonal
(b) Find eu and ev.

Û
17. Let F = k. Convert F to spherical coordinates.

18. The formulas in (1) express the basis vectors eu and ev in terms of i and j. Now
write eu and ev in terms of eu and ev.

19. The formulas in (3a) and (3b) express er and eœ in terms of i and j:
er = cos œ i + sin œ j
eœ = -sin œ i + cos œ j

Now write i and j in terms of er and eœ.

 page 1 of Section 2.3

SECTION 2.3 DOTS, NORMS AND CROSS PRODUCTS IN ORTHOGONAL COORDINATE SYSTEMS

norms and dot products in an orthogonal coord system

Û Û Û
Start with an orthogonal coordinate system with basis vectors eu, ev, ew. Norms and
dots are done "as usual". In particular, if
Û Û Û Û
a = au eu + av ev + aw ew
Û Û Û Û
b = bu eu + bv ev + bw ew
then

(1) «a« =
Û Û
√

2 2
au + av + aw
2

(2) a…b = aubu + avbv + awbw

geometric proof of (1) in 2-space

Û
Since the coord system is orthogonal, au is the signed projection of a onto eu
Û
and similarly av is the signed projection of a onto ev (Fig 1). So, by the
Û 2 2
Pythagorean theorem, «a«2 = au + av.
algebraic proof of (2) in 2-space
Û Û Û Û
a…b = (au eu + av ev )…(bu eu + bv ev)
= aubu eu…eu + aubv eu…ev + avbu eu…ev + avbv ev…ev by dot properties

But eu…ev = 0 since the system is orthogonal. And eu…eu = 1, ev…ev = 1 since eu and
ev are unit vectors (see property (g) of dot products in ±1.0). So a…b = aubu+ avbv.
→
a
→
a

eu
a ev
u
a
v
FIG 1

example 1
Let p = 2i + 8j.
(a) Express p in terms of er and eœ at point r=2, œ=π/3, in polar coordinates.
(b) Find «p«, using p = 2i + 8j
(c) Find «p« again, but this time use your p in terms of er and eœ from part (a).
solution (a) From ±2.2 (and the reference page),
er = cos œ i + sin œ j, eœ = -sin œ i + cos œ j
1 1
At point r=2, œ=π/3, er = 2 i + 2  √ 3 j, eœ = - 12 √ 3 i + 12 j.
Since the polar coord system is orthogonal (this wouldn't work otherwise), you can
use (6) in the preceding section to write p in terms of er and eœ:
p = (p…er)er + (p…eœ)eœ = (1 + 4√ 3)er + (-√ 3 + 4)eœ

(b) «p« = 
√
4 + 64 = √68.
 page 2 of Section 2.3

(c) Since er and eœ are orthogonal (this wouldn't work otherwise),

«p« = √

(1+4√ 3)2 + (-√ 3+4)2

= √
1 + 48 + 8√ 3 + 3 - 8√ 3 + 16

= √68 (same as in part (b), as it should be)

righthanded sets of basis vectors

A set of basis vectors eu, ev, ew, in that order, is called righthanded if your thumb
points more or less like (makes an acute angle with) ew when the fingers of your
right hand curl like eu turning into ev (Fig 2); if eu, ev, ew are orthogonal as well
as righthanded then your thumb would point exactly like ew.
For example, Fig 3 shows several sets of axes for which the vectors i, j, k, in that
order are righthanded.
For example, in Fig 4 (where the underlying i,j,k coord system is righthanded), the
basis vectors e®, eƒ, eœ, in that order, are righthanded; if your fingers curl like
e® turning into eƒ your thumb will point like eœ. Also in Fig 4, eœ, e®, eƒ are
righthanded and so are eƒ, eœ, e®.

ev
ew ev

ew eu eu
eu
ev
ew

righthanded eu, ev, ew lefthanded eu, ev, ew

FIG 2

y-axis

z-axis j
k z-axis
i k y-axis
j x-axis j
y-axis k i
x-axis i z-axis x-axis

FIG 3 righthanded i, j, k

eρ

eθ
•
y
x eφ

FIG 4
 page 3 of Section 2.3

cross products using orthogonal righthanded basis vectors

Û Û
a ≈ b is defined as follows, independent of any coordinate system:

length «a ≈ b« = «a« «b« sin œ where œ is the angle between a and b.

direction a ≈ b is perp to both a and b and furthermore, of the two perp directions,
it points like your thumb when the fingers of your righthand curl like a
turning into b.

In a previous calculus course you were shown that if you use righthandeded i, j, k
as in Fig 3 then the cross product can be computed like this:
If
a = 2i + 3j + 4k
b = i - j + 5k
then

 i j k 
(3) a ≈ b =  2 3 4  = 19i - 6j - 5k

 1 -1 5 

Suppose you continue with any of the coordinate systems in Fig 3 but rewrite a and
b like this (same vectors but write the components in a different order):

a = 4k + 2i + 3j
b = 5k + i - j

When you compute

 k i j 
 4 2 3 
 
 5 1 -1 

you get -5k + 19i - 6j, same as before; i.e., you still get a ≈ b. Notice that k,i,j
in Fig 3 are still righthanded.

But suppose you rewrite a and b like this:

a = 3j + 2i + 4k
b = - j + i + 5k

When you compute

 j i k 
 3 2 4 
 
 -1 1 5 

it turns out to be 6j - 19i + 5k. But this is the -(a ≈ b). Notice that j, i, k in
Fig 3 is a lefthanded triple

Here is what mathematicians say to straighten this all out (proofs omitted).

In an orthogonal coordinate system where the basis vectors eu, ev, ew are
righthanded, to get the cross product (with the length and direction described
above) use this rule:
 page 4 of Section 2.3

Û Û Û Û Û Û Û Û
If a = au eu + av ev + aw ew and b = bu eu + bv ev + bw ew then

 eu ev ew 
(4) a ≈ b =  au av aw 
 bu bv bw 
If eu, ev, ew are lefthanded and you use the determinant in (4) you will end up with
-(a ≈ b).

example 2

I'll use the spherical coordinate system in Fig 4. Suppose

a = 2e®
b = 3eƒ + 4eœ

Choose a righthanded order, like e®, eƒ, eœ. Then

 e® eƒ eœ 
 
a ≈ b = 
 2 0 0


 
 0 3 4 

2 0 0 2 0 0 2 0 0
= eρ e + eθ
4 φ
0 3 4 - 0 3 0 3 4
= -8eƒ + 6eœ

Since eœ, e®, eƒ is also a righthanded triple you can find a ≈ b using

 eœ e® eƒ 
 
 
 0 2 0 
 
 4 0 3 
But e®, eœ, eƒ are lefthanded so do not use

 e® eœ eƒ 
 
 
 2 0 0 
 
 0 4 3 
(or if you do, remember that this is -(a ≈ b) ).

example 3
Let
a = 2eu + 3ev - ew
b = 5eu + 2ev + 4ew
in a u,v,w coordinate system. If the system is orthogonal then

a…b = 10 + 6 - 4 = 12
«a« = 
√14
 page 5 of Section 2.3

And if eu, ev, ew are righthanded as well as orthogonal then

 3 -1   2 -1   2 3 
a ≈ b =  2  eu -  5  ev +  5  ew
4 4 2

= 14eu - 13ev - 11ew

PROBLEMS FOR SECTION 2.3

1. If a = 2e® - 3eƒ + eœ and b = 4e® + 2eƒ - 5eœ (spherical coordinates) find a…b
and a ≈ b.

2. Let p = i + 4j, q = 5i - j.
(a) Express p and q in terms of eu and ev at point u=2, v=3 in parabolic
coordinates.
(b) Find p…q and «p«, using the Cartesian coordinates of p and q.

(c) Find p…q and «p« again, but this time using the representation of p and q in
terms of eu and ev from part (b).

Û
3. Let a = e® (spherical coordinates)
Û
b = eœ (spherical coordinates)

Û Û
(a) Find a ≈ b without any algebra at all, just by looking at Fig 4 and thinking
about the length and direction of the cross product.
Û Û
(b) Find a ≈ b again using their coordinates in the spherical coordinate system.

(c) In Cartesian coordinates,

a = (xi + yj + zk)unit (away from the origin)

b = (-yi + xj)unit
Û Û Û Û
Find a ≈ b again using the Cartesian coordinates of a and b and keep going until
you see that it agrees with (a) and (b).

4. Suppose that at a point, eu= (2i+k)unit, ev = (2i+j+3k)unit, ew = (i+4j+k)unit

Are eu, ev, ew (in that order) a righthanded triple?

3 Û 4 Û
5. The vector 5 i + 5 j is a unit vector and points sort of northeast.
3 4
Look at 5 eu + 5 ev in a u,v coord system.
(a) Is it a unit vector.
(b) Does it point sort of northeast.
 page 1 of Section 2.4

SECTION 2.4 SCALE FACTORS

the scale factors hu and hv

Start with a u,v coordinate system given by

x = x(u,v), y = y(u,v)

Let dsu be the arc length traced out when u starts at u0 and changes by du and v
stays fixed at v0 (Fig 1). The little piece of u—curve traced out has parametric
equations

x = x(u,v0), y = y(u,v0), u0 ≤ u ≤ u0 + du (the parameter is u)

If you think of the parameter u as time then there is a velocity vector

ºx ºy
velu = i + j.
ºu ºu

By (1) in ±2.0,

dsu = «velu« du (distance = speed ≈ time).

The factor «velu« is called the arc length magnification factor and is denoted by hu.
So all in all,

(1) dsu = hu du

where

(2) hu = «velu« =
√
ºx 2 ºy 2
( ) + ( )
ºu ºu

You can think of hu as speed if you think of u as time. The actually units on hu are
meters/u—unit.

Similarly if u is fixed at u0 and v changes by dv then arc length dsv is traced out
on the v—curve u=u0 where

dsv = hv dv

hv = «velv« =
√
ºx 2 ºy 2
( ) + ( )
ºv ºv

• u=u +du
0
u
ds

h curve v=v
n gt 0
r s=rα
l•e α
u=u0

FIG 1 FIG 2

review of a geometry formula

Let s be the arc length cut off by a central angle of å radians (Fig 2).
Then s = rå.
 page 2 of Section 2.4

scale factors in polar coordinates

hr = 1
(3) hœ = r

You can derive these scale factors algebraically using (1) (see problem 1) but you
can also see them geometrically as follows:
Fig 3 shows r changing by dr. dsr is dr so hr = 1.
Fig 4 shows œ changing by dœ. By the rule in Fig 2, the arc length traced out is
dsœ = r dœ. So hœ = r. This means that for the same dœ (Fig 5) you get larger dsœ
when r is larger: If you start at point A in Fig 5 and change œ by dœ you trace out
arc length AB. But if you start at point C, where r is larger, and change œ by the
same dœ you trace out the larger arc length CD.
lar
D ger
ds
dsr • B
sm
al θ
l
ds

er
θ C

ds θ
• dr
A
r dθ r dθ
θ

FIG 3 FIG 4 FIG 5

scale factors in cylindrical coordinates

hr = 1
hœ = r
(4) hz = 1

scale factors in a u,v,z coordinate system (see the zippo cylindrical coordinate system in Section 2.1)
hz is always 1 because when you change z by dz,you get arclength dz.

example 1 (scale factors in parabolic coordinates)

In the parabolic system,
1
x = 2(u2 - v2), y = uv, v ≥ 0.
So
velu = ui + vj
velv = -vi + uj

hu = «velu« = √
u2 + v2 (by (1))

hv = «velv« = 
√
u2 + v2
If u=3 and v=2 then hv = √13. A particle starting at point u=3, v=2 and moving so
that u stays fixed while v goes up by dv will move distance dsv = 
√13 dv (Fig 6).
u=3 v=2+dv
ds v
v=2

FIG 6
 page 3 of Section 2.4

scale factors in spherical coordinates

h® = 1
hƒ = ®
(5) hœ = ® sin ƒ

Here's the geometric reason for these scale factors.

Fig 7 shows a ®—curve where ƒ and œ are fixed and ® changes by d® sweeping out arc
length ds®. By inspection, ds® = d® so h® = 1
Fig 8 shows a ƒ—curve where ® and œ are fixed and ƒ changes by dƒ sweeping out arc
length dsƒ. By the formula in Fig 2, dsƒ = ® dƒ so hƒ = ®.
Fig 9 shows a œ—curve where ® and ƒ are fixed and œ changes by dœ, sweeping out arc
length dsœ. Angle A in triangle ABC is 90o so AB = ® sin ƒ. Then, by the formula in
Fig 2, dsœ = AB dœ = ® sin ƒ dœ. So hœ = ® sin ƒ.

ve
dsφ θ-cur
φ
dφ A
dsρ • •
ρ dθ
dρ θ
• •
ve

B
ρ φ ρ
cur

dsθ
C•
φ-

FIG 7 FIG 8 FIG 9

connection between eu and hu

The normalizing factor that turns the tangent vector velu into the unit vector eu
is «velu« which happens to be hu. So

Û ºx Û ºy 1 ºx Û ºy Û
(6) eu = ( i + j)unit = ( i + j )
ºu ºu hu ºu ºu

For example, in spherical coordinates,

Û 1 ºx Û ºy Û ºz Û
eƒ = ( i + j + k )
hƒ ºƒ ºƒ ºƒ
1 Û Û Û
= (® cos ƒ cos œ i + ® cos ƒ sin œ j - ® sin ƒ k )
®
Û Û Û
(7) = cos ƒ cos œ i + cos ƒ sin œ j - sin ƒ k

PROBLEMS FOR SECTION 2.4

1. Find the scale factors in polar coordinates algebraically.

2. Find the spherical coordinate scale factors algebraically.

 page 4 of Section 2.4

3. Define a u,v coordinate system with the equations x = u2 v3, y = 3u + 2v.

A particle moves from point u = 2, v = 1 so that v stays fixed and u changes by du.
How far does it go.

4. The diagrams show some u,v coordinate paper.

(a) Which of points A,B,C,D has the largest hv.
(b) Which of points P,Q,R,S has the largest hv.
u=2
A u=2
• v=1 P
B • v=2 • v=1
Q
• v=100
C • R
v=3
• v=1000
S
D •
v=10000
• v=4
(a) (b)

5. Look at the diagram below.

(a) Is hv larger at A or at B.
(b) Which is larger at B, hu or hv.

B
•
A u=10
•
u=20
v=3
v=2

6. Let x = u2 + v, y = u + v. Which creates more arc length, fixing u and changing v

by a little bit or fixing v and changing u by the same little bit.

7. Find the scale factors in parabolic cylindrical coordinates.

 page 1 of Section 2.5

SECTION 2.5 CALCULUS IN ORTHOGONAL COORDS

repeated from Section 2.2

Û
(1) In an orthogonal coordinate system, when p = aeu + bev, the scalars a and b
are the signed projections of p onto eu and ev.

gradient
Suppose temperature f is expressed as a function of u and v in a u,v orthogonal
coordinate system. I want to express ∞f in terms of eu and ev; i.e., I want to find
a and b so that
∞f = aeu + bev.

Since the u,v system is orthogonal,

a = signed projection of ∞f onto eu (by (1) above)

But by (A) in Section 1.2, the signed projection of ∞f in a direction is the rate
of change of temp as you walk in that direction. So

a = degrees/meter—in—the—u—direction

Now let's see how to compute it.

ºf/ºu is degrees per u—unit (i.e., rate of change of f with respect to u)

hu (i.e., «velu«) is meters per u—unit.

Here is how a, ºf/ºu and hu are related

ºf
a times hu =
ºu
degrees meters degrees
meter u-unit u-unit

So

1 ºf
a =
hu ºu

Similarly for b. So

1 ºf 1 ºf
(2) ∞f = e + e
hu º u u hv º v v

And similarly, in 3—space,

1 ºf 1 ºf 1 ºf
∞f = e + e + e
hu º u u hv º v v hw º w w
 page 2 of Section 2.5

example 1
If f(r,œ) = r2 + cos œ in polar coordinates then
1 1 1
∞f = 2r er - sin œ eœ = 2rer - sin œ eœ
hr hœ r
1
At point r = 2, œ = π/6 for instance, ∞f = 4er - 4 eœ.

gradients in some special cases

Suppose the temperature at a point in 2—space depends only on r where r is the
usual polar coord, and not on œ. Or suppose the temperature at a point in 3—space
depends only on r where r is the usual cylindrical coordinate (distance to the
z—axis). Then

d(temp)
(3) ∞ temp = er
dr
Similarly, if the temperature at a point in 3—space depends only on the spherical
coordinate ®, and not on ƒ or œ, then

d(temp)
(4) ∞ temp = e®
d®

proof of (3) in 2-space

1 ºtemp 1 ºtemp
∞temp = er + eœ
hr ºr h œ ºœ

d(temp)
But ºtemp/ºœ = 0 here, and hr = 1, so ∞temp turns into just er
dr

example 2

If the temperature in 2—space is 1/r then

1
∞temp = - e
r2 r

Fig 1 shows some level sets and gradients.

FIG 1
antigradients of some special fields
The results in (3) and (4) can be read backwards to get antigradients for certain
vector fields:
1 1
If F = r3 er (in polar or cylindrical coordinates) then F = ∞ 4 r4 ; i.e., 4 r4 is
an antigradient of F.
If F = 2® e® (in spherical coords) then F = ∞®2, i.e., ®2 is an antigrad of F.

Û Û Û
If F = r—stuff er in polar or cyl coords then F = ∞(antideriv of r—stuff).

Û Û
If F = ®—stuff e® in spherical coords then F = ∞(antideriv of ®—stuff).
 page 3 of Section 2.5

divergence

If the u,v,w coordinate system is orthogonal and

F(u,v,w) = p(u,v,w) eu + q(u,v,w) ev + r(u,v,w) ew

(5a) then
1  º(hv hw p) º(hu hw q) º(hu hv r) 
div F =  + + 
hu hv hw  ºu ºv ºw 

º(hv hw p)
Note that to compute , multiply p by hvhw first and then differentiate the
ºu
product w.r.t. u (probably with the derivative product rule).The factors hv and hw
usually can't be pulled out of the derivative because they are usually not
constants.

If a 2—dim u,v coord system is orthogonal and

(5b) F = p(u,v) eu + q(u,v) ev then

1  º(hv p) º(hu q) 
div F =  + 
hu hv  ºu ºv 

why (5a) works

If F is a mass flux density in 3—space then div F dV is the net mass/sec leaving a
little box with volume dV. So

net mass/sec out

div F =
volume dV

I'll use this to find div F in the u,v,w coordinate system. I'm going to use the box
in Fig 2 swept out by starting at point A with coordinates u, v, w and changing u by
du, v by dv and w by dw.
To find the mass/sec out of the box, I'll find the flux out of all six faces.

Flux out the left face

= comp of F in -u direction ≈ area of left face
= -p evaluated at a point on the left face, say A ≈ area of left face (•)
= -p at A ≈ dsv dsw at left face (•)
= -p at A ≈ hv hw dv dw at left face
= (-phvhw at point A) dv dw
= -phvhw(u, v, w) dv dw

Similarly,

flux out the right face

= comp of F in the u direction ≈ area of right face
= (phvhw at point B) dv dw
= phuhv(u+du, v, w) dv dw
So
 page 4 of Section 2.5

net mass/sec out of the left and right faces

= phuhv(u+du,v,w) dv dw - phvhw(u,v,w) dv dw

=
[phuhv(u+du,v,w) - phvhw(u,v,w)] dv dw (factor)

phvhw(u + du,v,w) - phvhw (u,v,w)

= du dv dw (multiply and divide by du)
du
º(phvhw)
= du dv dw
ºu

D
volume dV
y dw

C
dv
h ds
e w b

by

ds v
engt

v
chang

th
ge

ng
an
arcl

e
ch

cl
ar
change u by du B
coords A arclength dsu
u, v, w
FIG 2

Similarly for the mass/sec out the top and bottom and the mass/sec out the front and
back. So

net mass/sec out of box

 º(hv hw p) º(hu hw q) º(hu hv r) 

(6) =  + +  du dv dw
ºu ºv ºw

And
dV = dsu dsv dsw (•)

(7) = huhvhw du dv dw
so
net mass/sec out
div F =
volume dV
(6)
=
(7)
1  º(hv hw p) º(hu hw q) º(hu hv r) 
=  + +  QED.
hu hv hw  ºu ºv ºw 

(•) These steps wouldn't work in a

non-orthogonal coordinate system
 page 5 of Section 2.5

why (5b) works

The vector field F(u,v) = p(u,v) eu + q(u,v) ev in 2-space in a coordinate system
where
x = x(u,v)
y = y(u,v)

can be considered to be

F(u,v,z) = p(u,v) eu + q(u,v) ev + 0 k

in the u,v,z cylindrical coordinate system where

x = x(u,v)
y = y(u,v)
z = z

Then (5a) becomes

1  º(hv hz p) º(hu hz q) º(hu hv 0) 

div F =  + + 
hu hv hz  ºu ºv ºz 

º(huhv0)
But hz = 1 in a u,v,z cylindrical coordinate system and is 0 so (5a)
ºz
collapses to (5b).

example 3
Let F(r,œ) = r sin œ er + r eœ in polar coords. Then hr = 1, hœ = r,
1 º(r2 sin œ) ºr
div F = ( + )
r ºr ºœ
1
= (2r sin œ + 0)
r
= 2 sin œ

warning
º(r sin œ) ºr
In example 2, div F is not + .
ºr ºœ
Don't forget to stick scale factors in all the right places,

example 4
Let f(®,ƒ,œ) = ®2 sin œ (spherical coords). Find Lapl f.

solution

Lapl f = div ∞f so first take gradient and then take divergence.

h® = 1, hƒ = ®, hœ = ® sin ƒ

1 ºf 1 ºf
∞f = e + e
h® º ® ® hœ º œ œ
1
= 2® sin œ e® + ®2 cos œ eœ
® sin œ
cos œ
= 2® sin œ e® + ® eœ
sin œ
 page 6 of Section 2.5

 º(®
cos œ
…®)

1  º(2® sin œ… ®2 sin ƒ) sin ƒ 
Lapl f = div ∞f =  + 
h®hƒhœ º® ºœ

1  ®2 sin œ 
=
2  6®2 sin œ sin ƒ -
sin ƒ 
® sin ƒ

sin œ
= 6 sin œ -
sin2 ƒ

curl

Suppose the u,v,w coord system is orthogonal and eu, ev, ew in that order are
righthanded. Let
Û Û Û Û
F(u,v,w) = p(u,v,w) eu + q(u,v,w) ev + r(u,v,w) ew
Then
(9)  1
e
hvhw u
1
huhw v
e
1
huhv w
e 
Û  
curl F =
 º
ºu
º
ºv
º
ºw 
 phu qhv rhw


The determinant in (9) is fake but it helps you remember the formula. It works only
if you expand the determinant across row 1. For instance, start like this

 1
e
hvhw u
1
e
huhw v
1
e
huhv w 
 
 ∂ ∂ ∂

 
∂u ∂v ∂w

phu qhv rhw

 º(rh ) º(qhv) 
1 w
to get e times  -  .
huhv u ºv ºw

Here's the whole curl formula written out.

 º(rh ) º(qhv) 
w
curl F =  -  1 eu
ºv ºw hvhw

 º(rh ) º(phu) 
w
-  -  1 ev
ºu ºv huhw

 º(qh ) º(phu) 
v 1
+  -  e
ºu ºv huuv w
 page 7 of Section 2.5

example 5
Û Û Û
Let F(®,ƒ,œ) = cos œ e® + ® eƒ + 2eœ in spherical coordinates.
Find curl F.

solution
e®, eƒ, eœ in that order are righthanded, and h® = 1, hƒ = ®, hœ = ® sin ƒ so


1
e
hƒ hœ ®
1
e
h® hœ ƒ
1
e
h® hƒ œ 
 
curl F =
 º
º®
º
ºƒ
º
ºœ 
 h® cos œ hƒ ® 2hœ


 1
e
1
e
1
e 
 ® sin ƒ ®
2 ® sin ƒ ƒ ® œ

=  º º º 
 º® ºƒ ºœ

 cos œ ®2 2® sin ƒ 
1  º(2® sin ƒ) º®2 
= e®  - 
2
® sin ƒ  ºƒ ºœ

-
1
® sin ƒ
eƒ
[ º(2® sin ƒ)
º®
-
º cos œ
ºœ ]
1  º®2 º cos œ 
+ e
® œ  º® - ºƒ 

1 1 1
= (2® cos ƒ - 0) e® - (2 sin ƒ + sin œ) eƒ + (2® - 0) e
2
® sin ƒ ® sin ƒ ® œ

=
2 cos ƒ
® sin ƒ
e® -
[2® + ®sinsinœƒ ]eƒ + 2eœ

warning
When you use the curl formula in (9), you must use a righthanded set of basis
vectors. For spherical coordinates that means e®, eƒ, eœ in that order (or any cyclic
permutation thereof) in row 1. But not e®, eœ, eƒ which is a lefthanded triple.

why (9) works

Let
F(u,v,w) = p(u,v,w) eu + q(u,v,w) ev + r(u,v,w) ew

I want to express curl F in terms of eu, ev and ew; i.e., I want to find a, b and c
so that
curl F = aeu + bev + cew.

I'll work on c. Since the u,v,w system is orthogonal, by (1) above, c is the signed
projection of curl F onto ew, i.e., the component of curl F in direction ew.
 page 8 of Section 2.5

But by (4) in ±1.5,

(comp of curl F in direction ew) dS

= righthanded circ on a little loop around ew, with area dS

So
righthanded circ on a little loop around ew
(10) c =
dS

To find c, first I need a loop around ew.

In Fig 3, I started at point A with coordinates u,v,w and drew the righthanded
orthogonal basis vectors eu, ev, ew. For my small closed curve around ew, I'll choose
the "rectangle" ABCD: Side AB came from changing u by du while v stayed fixed. Side
AD came from changing v by dv while u stayed fixed.
Let dS be the surface area inside the curve.
The righthanded direction on the curve w.r.t. ew is A to B to C to D to A.
The circ along the B—to—C segment for instance is the tangential component of F on
the segment, namely q, times length BC. Similarly for the circs along the other
three sides, so

righthanded circ = p on AB ≈ length AB

+ q on BC ≈ length BC
+ -p on CD ≈ length CD
+ -q on DA ≈ length DA
dv D
v by >
ew nge ds v C
h
ev cha en gt
cl area dS
ar
e
u
>
A•
B
change u by du
coords arclength dsu
u, v, w
FIG 3

Length AB = dsu = (hu on AB) du, and similarly for the other sides. So

righthanded circ = (phu on AB, say at point A) du

+ (qhv on BC, say at point B) dv
- (phu on CD, say at point D) du
- (qhv on DA, say at point A) dv
Rearrange to get

qhv(u+du,v,w) - qhv(u,v,w)
righthanded circ = du dv
du

phu(u,v+dv,w) - phu(u,v,w)
- du dv
dv

 º(qh ) º(phu)
v
(11) =  -  du dv
ºu ºv
 page 9 of Section 2.5

Since the coord system is orthogonal,

dS = dsu dsv
(12) = hudu hvdv.

Then, by (10),
 º(qh ) º(phu)
(11) 1 v
c = =  -  .
(12) hu hv ºu ºv

Similarly for a and b. The three formulas can be combined into the single rule
in (9). QED

APPENDIX LAPLACIAN IN POLAR, CYLINDRICAL AND SPHERICAL COORDINATES

These are just special cases of the general formulas for div and grad but you might
want to quote them directly in another course.

polar coordinates

º2f 1 ºf 1 º2f
Lapl[ f(r,œ) ] = + +
ºr2 r ºr r2 ºœ 2

cylindrical coordinates

º2f 1 ºf 1 º2f 1 º2f

Lapl [f(r,œ,z)] = + + +
ºr2 r ºr r2 ºœ 2 r2 ºz2

spherical coordinates

º2f 2 ºf 1 º2f cos ƒ ºf 1 º2 f

Lapl[ f(®,ƒ,œ)] = + + + +
º® 2 ® º® ®2 ºƒ 2 ® sin ƒ º ƒ
2 ® sin2ƒ
2 ºœ 2

PROBLEMS FOR SECTION 2.5

1. Find the gradients of the following scalar fields in spherical coordinates.

(a) sin ƒ (b) ® sin ƒ

2. Find the gradient of an arbitrary f(r,œ) in polar coordinates.

3. Find ∞temp using the most appropriate coordinate system. In each case, sketch the
temp level sets, the gradient field and describe them in words.
(a) Temperature at a point in 2—space is the cube of the distance from the point to
the origin.
(b) Temp at a point in 3—space is 1/(distance to z—axis).

4. The arrows of a vector field in 2—space point away from the origin and the length
of the arrow at a point is 1/(distance to origin).
Find an antigradient for F in the most appropriate coordinate system.
 page 10 of Section 2.5

Û Û Û Û Û
5. Let r = xi + yj + zk. Find ∞(«r«3)
(a) by sticking to the x,y,z coord system
(b) by switching to a new coord system and check that the two answers agree

6. Let F = uev in parabolic coordinates. Find div F.

7. Define an r,œ coordinate system with the equations x = er cos œ, y = er sin œ.
(a) Show that the coordinate system is orthogonal.
Û Û
(b) If F = rer + cos œ eœ find div F.
(c) Find the gradient of r3.
(d) Find Lapl of r3

1
8. Find Lapl . The answer should be in the same coordinate system as the
x2+y2+z2
original problem but along the way you might want to use another coord system.

9. Let v(r,œ) be an arbitrary scalar field in polar coords. Find Lapl v in polar
coordinates.

Û Û
10. Find div and curl of 2zer + 3zeœ (cylindrical coords).

1 1
11. Find an antigradient, div and curl of (a) e (b) e
®2 ® r2 r

12. Let F = eu in parabolic coordinates. Find curl F.

13. Let F = ®eœ (spherical coordinates).

Find curl F.

x y z
14. Let F = i + j + k
x2+y2+z2 x2+y2+z2 x2+y2+z2

(a) Find div F (using Cartesian coordinates)

(b) Find F using a non—Cartesian coordinate system.
And check that your answers to (a) and (b) agree.
(c) Find curl F.

15. Let F = cos œ eœ (polar coordinates). Find curl F.

 page 1 of review problems for Chapter 2

REVIEW PROBLEMS FOR CHAPTER 2

1. Find Lapl(® sin œ) (spherical coords)

1
2. In the parabolic coord system, x = 2(u2 - v2), y = uv, v ≥ 0.
Suppose the temperature at a point is u2 + 2v. If you walk from point u = 1, v = 2
in the direction of arrow 2eu - 3ev, what rate of change of temperature per foot do
you feel instantaneously.
Where exactly will you use the fact that the parabolic coord system is orthogonal.

3. Fig 1 in Section 2.1 showed parabolic coordinate paper. Where are the u—axis and
v—axis ?????
Û Û
4. If F(x,y) = 2i + 3j then F is a uniform field (all arrows have the same length
and same direction). Suppose F(u,v) = 2eu - 3ev in some u,v coordinate system. Is F
uniform.

x y
5. Let F(x,y) = i + j.
2
x + y2 x + y2
2

(a) Write F in polar coordinates.

(b) Find div F twice, using the x,y system and again using polar coordinates.
(c) Find curl F.

6. (a) Sketch the vector field eƒ (spherical coordinates).

(b) Find curl eƒ.

7. Let x = 2r cos œ, y = 4r sin œ, r ≥ 0 define a coordinate system.

(a) Identify all the œ—curves and sketch some.
(b) Identify the r—curve œ = π/4 and add it to the picture.
(c) Is the coord system is orthogonal
(d) Find the basis vectors at point r=2, œ = π/4 and sketch them in the picture.

(e) Find hr at point r=2, œ = π/4 and explain what it means physically.

8. Let F = 2i + 4j + 6k. Express F in spherical coordinates at the point where

® = 117, œ = π/2, ƒ = π/4.

Û Û y3 -xy2
9. Let F = pi + qj where p = and q =
(x2 + y2)2 (x2 + y2)2

Write F in polar coordinates.

 page 1 of appendix to Chapters 1,2

APPENDIX TO CHAPTERS 1 AND 2

using Mathematica to plot level sets

Here's how to get the -2, -1, 0, 1, 2 level sets of x2 + y - 2 (Fig 1 in ±1.1 but I
added the labels separately in a drawing program).

ContourPlot[x^2 + y - 2,{x,-4,4},{y,-4,4}, ContourShading->False,

Frame->False, Axes->True, PlotPoints->25, Contours->{0,-1,2,1,-2},
AxesOrigin->{0,0},Ticks->{{-4,4},{-4,4}},AspectRatio->Automatic];

2 level x2+y-2=2
1 level
0 level

-1 level

-2 level x2+y-2= -2

using Mathematica to plot vector fields

First load the PlotField package.

<< Graphics`PlotField`
Û Û Û
Here's how to plot the vector field F(x,y)= (y+3)i + yj for -2 ≤ x ≤ 2, -3 ≤ y ≤ 3
so that the arrows along a row occur at intervals of .7 and the arrows along a
column also occur at intervals of .7 (Fig 4 in ±1.1). If you leave out the two .7's
in the instruction, you get the default spacing which would give you more arrows
than I got, spaced closer together.

PlotVectorField[{y+3,y},{x,-2,2,.7},{y,-3,3,.7},Axes->True,Ticks->None];
 page 2 of appendix to Chapters 1,2

using Mathematica to find div, curl, gradient in various coordinate systems

First, load the package with all the vector operations in it.

<<Calculus`VectorAnalysis`

Û Û
Here's how to find the divergence of x2 y2 i + (2x + 3y)j. The program will only
take the divergence of a 3—dim vector field so use third component 0.
And you have to tell Mathematica which coordinate system to use.

SetCoordinates[Cartesian[x, y, z]]

Div[{x^2 y^2,2x + 3y, 0}]

2
3 + 2 x y

Curl[{x^2 y^2, x y z, 2x + 3y + 4z}]

2
{3 - x y, -2, -2 x y + y z}

Grad[x^2 y^3 z + 3z]

3 2 2 2 3
{2 x y z, 3 x y z, 3 + x y }

Instead of specifying a default coord system, or if you want to override your

default system, you can specify each time what coord system to use.

Div[{u^2 v^2,2u + 3v, 0}, Cartesian[u,v,w]]

2
3 + 2 u v

Div[{rho^3,0,0}, Spherical[rho,phi,theta]]

2
5 rho

Laplacian[v[r, theta], Cylindrical[r,theta,z]]//Together

(0,2) (1,0) 2 (2,0)

v [r, theta] + r v [r, theta] + r v [r, theta]
-----------------------------------------------------------
2
r

The notation v(2,0) means the second derivative of v w.r.t. its first variable,
º2v
i.e.,
ºr2