Classical Mechanics Notes

copyright 2012, George Gollin
1
1
2

Physics 325

Classical Mechanics I

George Gollin
University of Illinois at Urbana-Champaign
Fall 2012

2
2


Classical Mechanics I
Physics 325

Welcome!

Here is an introduction and syllabus for the course.

Lectures: Tuesday, Thursday, 11:00 12:20 in Loomis 151.
Discussion sections: Monday, at various times, in Loomis 236.
Office hours: To be arranged.

Weekly problem sets are to be submitted in class immediately
before the beginning of the Thursday lecture. Ill return them, in
class the following Thursday. We wont be using the yellow
homework boxes that are on the second floor of Loomis.

Text: Classical Dynamics of Particles and Systems, 5th edition,
Stephen T. Thornton, J erry B. Marion. The fourth edition is fine
too, and MUCH less expensive. The third edition has a chapter on
chaos and nonlinear dynamics, and you can always borrow a later
3
3
Physics 325, fall 2012 Introduction University of Illinois
4
edition when its time to take a look at that material next semester
if youve found an even earlier edition. (The fourth and third
editions can be purchased through online merchants.)

My lecture notes are available at Notes-n-Quotes, 502 E. J ohn St.
in Champaign. You should buy your own copy and bring it to
class.

Take note:
Well use calculus all the time.
No calculators or computers are to be used as computational
aids, or as portals to access reference sources unless I
explicitly give the OK for a particular problem. You are not
to use them when working problem sets, problem session
exercises, or exams.
There is to be no laptop, cellphone, iPhone, iPad usage
during class. I do not object to snacks and beverages.
Ill hand out a short credit/no credit written exercise in most
classes. They are intended to give me some feedback about
what is (and isnt) making sense to you. I expect theyll take
about five minutes to complete.
I encourage you to work together on problem sets, but please
make sure everyone in your group participates in generating
4
4
5
the solutions and understands how they were obtained. You
are not to submit solutions that you have found on the web, in
books of worked problems, in archival copies of earlier
offerings of the course, or any other source of that nature.
That would be cheating: submitting work for a grade that is
not your own.
There will be two (evening) midterms, no time limit, and one
final exam. Exams count for approximately 75% of your
course grade.
Late homework will be penalized 50%. Homework more than
one week late will not be graded. I will penalize you for
failing to submit homeworks, approximately a half-letter
grade for every two missed assignments. Homework counts
for approximately 20% of your course grade.
Attendance at lectures is mandatory. You are to arrive on
time and stay until the end of class. If you are ill, have the
emergency dean send me the relevant documentation. If you
need to miss class for a job interview, athletic competition,
family event, or other legitimate reason let me know ahead of
time, and in writing. I will penalize you for unexcused
absences, probably a half-letter grade for every two missed
classes.
5
5
6
Come to the Monday problem sessions: your grade will
include points earned from your participation in these.
I will use email, sent to your UIUC email address, as an
important means of communication. Please be sure to check
this at least once per day. I dont think the physics
departments bulk mail system is currently able to send
messages to anything other than your illinois.edu address, so
I wont be able to get class-wide announcements to you
through other services such as gmail.
All of the material I distribute is copyrighted. That means
you are not to redistribute it without my permission. From
time to time I find that my courses material is offered for sale
on the web. That makes me cross, and generally triggers a
phone call to the Office of University Legal Counsel.
About academic integrity and plagiarism: you are responsible
for knowing what the Student Code requires of you. Much of
it is common senseyou are to conduct yourself in an honest
manner, not represent the work of others as your own, and
provide full attribution for sources you use in your written
work. Cheating and plagiarism are unacceptable; lifting even
a significant portion of a single sentence from an existing
work is an act of plagiarism. See
http://admin.illinois.edu/policy/code/.
6
6
7

Course staff
Lecturer:
George Gollin, Professor of Physics. Loomis 437d
Teaching assistants:
---, Graduate student.
Graders:

7
7
8
Lecture Notes Contents
1. Newtonian Mechanics, mostly in one dimension

Newtons Laws for motion in an inertial reference frame...........1-1
A comment about gravitational vs. inertial mass........................1-3
Invariance of Newtons Laws under Galilean transformations...1-5
Solving the equation of motion of a particle................................1-8
1. Force is zero........................................................................1-9
2. Force is constant................................................................1-10
3. Force is a function of time only........................................1-14
4. Force is a function of position only...................................1-17
5. Force is only a function of velocity...................................1-24
6. Force is a function of everything.......................................1-30
Nonrelativistic rocket.................................................................1-32

2. Dynamics in three dimensions

Conservative forces.....................................................................2-1
A reminder about the gradient operator ..................................2-1
Line integral around a loop as a test for a conservative force.2-5
Form of the equations in three dimensions.............................2-8
Interesting special cases........................................................2-10
Examples of forces derivable from potentials.......................2-15
Zero curl as a test for a conservative force............................2-20
Two examples in three dimensions.......................................2-28

3. Systems of particles

Definition of the center of mass...................................................3-1
Motion of the center of mass.......................................................3-2
Total angular momentum and torque...........................................3-4
Reduced mass coordinates for two-particle systems...................3-5

8
8
9
4. Oscillations

An overview................................................................................4-2
Force vs. displacement and period vs. amplitude........................4-4
Nature of the motion close to an equilibrium point.................4-4
Simple pendulum, two ways: energy method; force method..4-8
Behavior of the period...........................................................4-12
Successive approximation approach to the nonlinear problem....
......................................................................................... 4-13
Next order solution................................................................4-21
Size of the effects induced by nonlinearities.........................4-26
A reminder about initial conditions...........................................4-28
Phase diagrams..........................................................................4-29
Inhomogeneous elastic deformations: introduction to tensors...4-32
Strain (displacement) vs. stress for identical springs............4-35
Strain vs. stress for an inhomogeneous set of springs...........4-36
Transformation properties of the elastic stress tensor under
rotations............................................................................4-40
Rank-n tensor transformation properties...............................4-47
The outer product of two vectors is a tensor.........................4-49
Using complex exponentials to solve the oscillator equation....4-51
Damped oscillations...................................................................4-55
Case 1: underdamped oscillations.........................................4-57
Oscillator Q (Quality factor) .................................................4-59
Case 2: overdamped oscillator ..............................................4-63
Case 3: critically damped oscillator......................................4-64
Electrical analogues...................................................................4-69
Driven oscillations.....................................................................4-73
Particular solution to the inhomogeneous equation...............4-76
Characteristics of x
p
...............................................................4-85
Non-sinusoidal driving forces....................................................4-96
Fourier series...........................................................................4-100
Orthogonality relationships.................................................4-102
Square wave........................................................................4-108
Periodic impulses of short duration.....................................4-113
9
9
10
Impulse response and an introduction to Greens functions....4-119
Response of an oscillator to a single, short impulse............4-121
Response of an oscillator to a non-periodic driving force...4-126
Greens functions................................................................4-128
Greens function for the Coulomb potential........................4-130
Satisfying the initial conditions...........................................4-132
Some Greens function examples........................................4-134

5. Motion in rotating frames of reference

A qualitative explanation for the existence of Coriolis forces.....5-2
Pendulums and hurricanes.........................................................5-13
Quantitative description of motion in rotating frames...............5-16
Velocities and accelerations......................................................5-27
Effective force...........................................................................5-32
Angular acceleration term..........................................................5-34
Centrifugal term.........................................................................5-34
Coriolis term..............................................................................5-36
The earths oblateness................................................................5-38
Coriolis effect on a hockey puck...............................................5-39
Coriolis effect on a free-falling object.......................................5-41
Foucault pendulum analysis......................................................5-44

6. An introduction to fluid dynamics

Utility of conservation laws in fluid dynamics............................6-2
Partial derivatives and convective derivatives.............................6-4
Conservation of mass...................................................................6-8
Current density and the mass in a macroscopic volume............6-15
Density changes in a comoving frame...................................6-16
(Ir)rotational flow......................................................................6-18
Equations of motion for an ideal fluid.......................................6-22
Shear stress in a Newtonian fluid...........................................6-27
Conservation laws......................................................................6-33
Energy conservation..................................................................6-36
10
10
11
Stress tensor...............................................................................6-43
Including viscosity; the Navier-Stokes equations......................6-47
Water flowing through a long, cylindrical pipe.........................6-48
Summary....................................................................................6-53

7. Lagrangians and the calculus of variations

Introduction.................................................................................7-2
Definition of the Lagrangian.......................................................7-4
The connection between the Lagrangian and Newtons laws......7-5
Using the Lagrangian...................................................................7-8
Sliding mass, sliding wedge....................................................7-8
Nasty pendulum....................................................................7-14
Definition of the (classical) action.............................................7-20
Classical action for a free-falling mass......................................7-21
Hamiltons principle..................................................................7-23
Demonstration of the equivalence of Hamiltons principle
and the Euler-Lagrange equations.............................................7-24
Setting up a calculus of variations approach to the
problem.................................................................................7-25
Minimizing the action through proper choice of
trajectory...............................................................................7-29
The Euler-Lagrange equations and other minimization
problems...........................................................................7-35
A different minimization problem.............................................7-36
Brachistochrone.....................................................................7-37
A few comments........................................................................7-49
Generalized coordinates........................................................7-49
Feynmans use of the principle of least action......................7-51
More Lagrangian examples.......................................................7-53
Double Atwoods machine, done two ways..........................7-53
Brief recap.............................................................................7-61
Spherical pendulum...............................................................7-63
Conical pendulum.................................................................7-70
Small perturbations to the conical pendulum........................7-71
11
11
12
A comment on generalized momenta........................................7-78
The Hamiltonian as a Legendre transform of the Lagrangian...7-81
Lagrangian and Hamiltonian for a particle in a magnetic field.7-90
Poisson brackets........................................................................7-93

Math review

1. series approximations.............................................................m-2
Taylors Theorem...................................................................m-2
Binomial approximation.........................................................m-2
sin(x), for x in radians and x close to zero.............................m-4
cos(x), for x in radians and x close to zero.............................m-5
2. some geometry........................................................................m-6
right triangles.........................................................................m-6
triangles in general .................................................................m-7
circles.....................................................................................m-8
3. exponentials............................................................................m-9
4. differential equations............................................................m-10
5. Vectors..................................................................................m-13
Scalar product.......................................................................m-13
Cross product.......................................................................m-14
Differentiation of vectors.....................................................m-15
Integration of vectors...........................................................m-16
Line integrals........................................................................m-17
6. Vector transformation properties..........................................m-18
Under rotations.....................................................................m-18
Matrix multiplication and rotation matrices.........................m-20
How unit vectors transform under rotations.........................m-26
Successive rotations in three dimensions.............................m-27
Kronecker delta symbol .......................................................m-30
Levi-Civita symbol...............................................................m-31
Tensors.................................................................................m-32
7. Non-Cartesian coordinate systems........................................m-35
cylindrical coordinates.........................................................m-36
spherical coordinates............................................................m-37
12
12
13
8. partial derivatives..................................................................m-38
using partial derivatives to approximate changes in functions
........................................................................................m-40
gradient operator ..................................................................m-41
divergence operator..............................................................m-44
curl operator.........................................................................m-45
9. quadratic formula..................................................................m-46

Physics 325 course calendar
Week
Day Date Time Room Class activity Topic
1 Tue 8/28 11 am LLP 151 Lecture 1 Newtonian Mechanics, one dimension
1 Thu 8/30 11 am LLP 151 Lecture 2 Newtonian Mechanics, one dimension
2 Mon 9/3 Labor Day
2 Tue 9/4 11 am LLP 151 Lecture 3 Newtonian Mechanics, one dimension
2 Thu 9/6 12 am in class HW 1 due
2 Thu 9/6 11 am LLP 151 Lecture 4 Dynamics in three dimensions
3 Mon 9/10 3, 4, 5, 6, 7, 8 pm LLP 236 Problem session week 2 material
3 Tue 9/11 11 am LLP 151 Lecture 5 Dynamics in three dimensions
3 Thu 9/13 11 am LLP 151 Lecture 6 Dynamics in three dimensions
4 Tue 9/18 11 am LLP 151 Lecture 7 Systems of particles
4 Thu 9/20 11 am LLP 151 Lecture 8 Oscillations
5 Tue 9/25 11 am LLP 151 Lecture 9 Oscillations
5 Thu 9/27 11 am LLP 151 Lecture 10 Oscillations
6 Tue 10/2 11 am LLP 151 Lecture 11 Driven oscillations
6 Thu 10/4 11 am LLP 151 Lecture 12 Driven oscillations
7 Tue 10/9 11 am LLP 151 Lecture 13 Driven oscillations
7 Thu 10/11 11 am LLP 151 Lecture 14 Impulse response
7 Thu 10/11 7 pm TBA Exam 1 Material from weeks 1 - 6
8 Tue 10/16 11 am LLP 151 Lecture 15 Impulse response
8 Thu 10/18 11 am LLP 151 Lecture 16 Impulse response
13
13
14
9 Tue 10/23 11 am LLP 151 Lecture 17 Motion in rotating frames
9 Thu 10/25 11 am LLP 151 Lecture 18 Motion in rotating frames
10 Tue 10/30 11 am LLP 151 Lecture 19 Fluid dynamics
10 Thu 11/1 11 am LLP 151 Lecture 20 Fluid dynamics
11 Tue 11/6 11 am LLP 151 Lecture 21 Fluid dynamics
11 Thu 11/8 11 am LLP 151 Lecture 22 Fluid dynamics
12 Tue 11/13 11 am LLP 151 Lecture 23 Lagrangians and calculus of variations
12 Thu 11/15 11 am LLP 151 Lecture 24 Lagrangians and calculus of variations
12 Thu 11/15 7 pm TBA Exam 2 Material from weeks 7 - 12
Thanksgiving
13 Thu 11/29 11 am LLP 151 Lecture 26 Lagrangians and calculus of variations
14 Thu 12/6 11 am LLP 151 Lecture 28 Hamiltonian dynamics
15 Mon 12/10 3, 4, 5, 6, 7, 8 pm LP 236 Problem session week 14 material
15 Tue 12/11 11 am in class HW 14 due
15 Tue 12/11 11 am LLP 151 Lecture 29 Hamiltonian dynamics

Tue 12/18
8 am
TBA Final exam
Everything

14
14
15
15
15
16

16
16
Physics 325, fall 2012 Newtonian mechanics University of Illinois
Newtonian Mechanics, mostly in one dimension
Newtonian Mechanics, mostly in one dimension........................1-1
Newtons Laws for motion in an inertial reference frame........1-1
A comment about gravitational vs. inertial mass.....................1-3
Invariance of Newtons Laws under Galilean transformations 1-5
Solving the equation of motion of a particle.............................1-8
1. Force is zero.......................................................................1-9
2. Force is constant..............................................................1-10
3. Force is a function of time only.......................................1-14
4. Force is a function of position only.................................1-17
5. Force is only a function of velocity.................................1-24
6. Force is a function of everything.....................................1-30
Nonrelativistic rocket..............................................................1-32

Newtons Laws for motion in an inertial reference frame
1
st
law: No force uniform motion (constant vector velocity)
2
nd
law: Applied force = induced momentum change per unit time
3
rd
law: Bodies exert equal-and-opposite forces on each other.

Lets assume v
c so we can define p m =

.

1
st
law 0 when 0
dp
F
dt
= =

2
nd
law
dp dv
F m
dt dt
= = =
ma

.
copyright 2012 George Gollin PHY_325_lec_notes_01.doc

1-1
17
17

The 1
st
and 2
nd
laws are related: the 1
st
follows from the 2
nd
.

The 3
rd
law comes from the first two.

Heres how: consider an isolated system (with no external forces
being applied).
1
p
2
p
1
2
1
p
2
p
1
2

Total momentum is
1 2
P p p = +

.
Since no external force acts on the system, 0
dP
dt
=
.
Therefore,
( )
1 2
0
d p p
dt
+
=

so
1 2
dp dp
dt dt
=

.

1
2 on
dp
F
dt

=
1
and
2
1 2 on
dp
F
dt

=

2 1 1 on on
F F
2
=


1-2
18
18
This sounds odd, in a way: the gravitational force exerted on you
by the earth is equal, but opposite to the force exerted by you on
the earth.
Note that
2
2
dv d x
F ma m m
dt dt
= = =

only works when m is constant.

If m changes (for example, a rocket) then we have to use
F dp d =

t .

This differential equation lets us figure out how x
changes with
time when we know the nature of F
.

A comment about gravitational vs. inertial mass.
(Saywhut?)

Inertial mass is what determines how brisk/sluggish the response is
to a given force. Big mass small acceleration.

For example: charge Q, mass m, in an electric field E
:


1-3
19
19
F Q =

E so
F QE
a
m m
= =

.

A peculiar thing about gravity:

F m : g =

the charge seen by a gravitational field is the same
thing as the objects mass (which determines how briskly the
object accelerates in response to a given force).

As best as can be measured, the gravitational mass (the charge
seen by gravity) exactly equals the inertial mass.

Coincidence? Noits a hint about gravity and the interplay
between mass, space, and time. This led Einstein to guess that the
effects of uniform acceleration and a static gravitational field were
equivalent: you couldnt tell which was which without looking out
the window of your lab.


1-4
20
20
acceleration
acceleration
acceleration
acceleration
gravity gravity gravity gravity
acceleration
acceleration
acceleration
acceleration
acceleration acceleration acceleration
gravity gravity gravity gravity gravity gravity gravity gravity

Left: accelerating elevator being traversed by a light pulse.
Right: stationary elevator in uniform gravitational field.

Light bends in gravitational fields. Black holes, general
relativity

Invariance of Newtons Laws under Galilean transformations


1-5
21
21
O
x
y
O
x
y
O
x
y
O
x
y
O
x
y
O
x
y

The primed frame has velocity
0
v
seen from unprimed frame. Say

origin of (as viewed from O) at t =0. Also,
0
is at O

, .
x
, x x y y z z

An event (useful even in non-relativistic situations) at , x t
as seen
by observers in O will occur at
( )
0 0
x x x v t =

and t t . =

Two events,
( )
1 1
, x t
and
( )
2 2
, x t
, will have

2 1
x x x =

,
( ) ( )
2 0 0 2 1 0 0 1
x x x v t x x v t =

2 1
t t t = ,
2 1 2 1
t t t t t

= = = t


1-6
22
22
So that
0
, x x v t =

or
0
x x
v
t t

=

.

A velocity v seen from O will relate to v
seen from O like this:

0
v v v =

A change in velocity will correspond to
2
v v

1
v
( ) ( )
2 1 2 0 1 0
v v v v v v v = = =

v
.

Therefore

v v
t t

=

so that
2 2
2 2

md x md x
dt dt

Newtons Laws are left undisturbed by a Galilean transformation:

F ma ma = =

.


1-7
23
23
We see the response to a gravitational pull obeying F ma =

whether were at rest with respect to the earths surface or in
uniform motion.

We can choose the particular inertial frame that suits our
convenience (e.g., center-of-mass) without loss of generality.

Solving the equation of motion of a particle

Physics input:

1. (a differential equation) F ma =

2. identify all the forces acting on a particle and model their
dependences on , , , etc. x v t

Lets categorize some of the kinds of forces we might encounter.

1. 0 F = (easy: 0 F = means v

is constant)
2. (example: falling body, close to the earths
surface, neglecting air resistance.)
constant F =

1-8
24
24
3. F
a function of time only (uniform through all space)

4. F
a function of position only (constant in time)

5. F
a function of v
only (e.g., air resistance)

6. F
a function of everything

Lets look at each case in turn.

1. Force is zero

Zero force means is constant since there is no acceleration.
Therefore v t .
v
( )
0 v
( )
0
v =

Find
( )
x t
by integrating:
dx
vdt dt dx x C
dt
= = = +

0

where C is a constant of integration.

Because we have
0
v v =

0 0 0
vdt v dt v t C v t x = = + = +

since
( )
0
0 t
x t x
=
=

.

1-9
25
25

If youre troubled by the integration of a vector in vdt
, keep in
mind that this is just shorthand for three terms:
x y z
v y v z v v x = + +

so that

x y z x y
vdt v xdt v ydt v zdt xv t yv t z v t = + + = + +

z

when are constant. ,
,
v v v
z x y

Dealing with the integration constant C
is simple:

( )
0
x t v t = +

C so
( )
0
0 x C x =

.

2. Force is constant
An example of this is a free-falling body close to the earths
surface if we can neglect air resistance.

Solve for the motion by integrating twice:


1-10
26
26
2
2
so
d dx dx d x d dx
F m m Fdt m dt m d
dt dt dt dt dt dt

= = = =

or

1
dx
Ft m C
dt
= +

[Equation 1]

since the force is constant. Do some algebra:

( ) ( )
1 1
so that
dx dx
Ft C m Ft C dt m dt mdx
dt dt
= = =

or

2
1 2
2
Ft
C t mx C = +
. [Equation 2]

Recall that the force is constant so that a F m =

. Define
1 1
C C

m and
2 2
C C
to rewrite Equations 1 and 2 this

way:
Equation 1 ; Equation 2
1
at C v
+ =
2
1 2
1
2
at C t C x

+ + =

Therefore

1-11
27
27
( )
1 0
0 C v v

x and
( )
2 0
0 C x

so that
( )
2
0 0
1
2
x t at v t x = + +

y
.

Heres the classic constant-force problem: projectile motion.
We have
0 0 0
cos sin v v x v = +
and would like to solve for

where the projectile lands.

x
y
x
y

( )
2
0 0
1
2
x t at v t x = + +

Break this up into three equations:
( ) ( )
( ) ( )
( )
0 0
2 2
0 0
0
cos
1
sin ( 9.8 m/sec )
2
(motion is in the plane)
x t x v t
y t y v t gt g
z t z xy
= +
= + +
=

The projectile lands when
0
y y = at a time . 0 t >


1-12
28
28
For convenience: place the origin at the projectiles starting point
so that .
0 0 0
0 x y z = = =

We have
( ) ( )
2
0
1
sin
2
y t v t gt = +
so solve for the positive time that yields y =0.

Two ways come to mind: either use the quadratic formula or just
factor the polynomial. (Its easier to factor the polynomial in this
case.)

( )
2
0
1
sin 0
2
y v t gt = + =
0
1
sin 0
2
t v gt

+ =

Solutions: both and 0 t =
0 0
sin 2 0 (so 2 sin ) v gt t v g + = =
will satisfy the equation. Its the t >0 solution we want: recall that
. 9.8 g

After time
0
2 sin t v g = , the projectile has traveled this distance
in x:


1-13
29
29
2
0 0
0
2 sin 2 cos sin
cos
v v
x v
g g

= = .

I want you to be capable of using various trig identities to simplify
expressions. An example:
(
1
2
cos sin sin 2
)
= (see the Math
review section of the notes).

Proving this: use cos , sin
2 2
i i i
e e e e
i
i

+
= =

to write
2 2
2 2
1 1
cos sin
2 2 4
1 1
sin2
2 2 2
i i i i i
i i
e e e e e e
i i
e e
i
i

+ +
= =

= =

So the projectile range is
2
0
sin2 v
g
. Note the maximum at 45.

3. Force is a function of time only

If possible, integrate twice:
d dx
F m
dt dt
so we have

1-14
30
30

F d dx
dt dt v
m dt dt

= =

dx
dt
=

and then
( )
F dx
dt dt dt dx x t
m dt

= = =

.

Since is a function of time only, the integral on the left can be
attacked, even if we need to do it numerically, via computer.
F

An example: exponentially waning force:

( )
0

t
F t ma e x

=

We have: m
2
2
d x
m
dt
=
0
t
a e

and also
2 2
2 2

d y d z
m m
dt dt
0 = = .

Useful notation:
2
2
,
dx d x
x x
dt
dt
so we can write
( )
0

d
t
x a e
dt

= .

Multiply by dt and integrate:
( )
0
t
d x a e dt

=

or


1-15
31
31
0 0
t
x c a e

= where c
0
is a constant of integration. [Equation 1]

We have
0 0
dx
t
c a e
dt

= which can be integrated again:

( )
0 0
t
dx c a e dt

=

so that

( )
2
1 0 0
t
x t c c t a e

= + + [Equation 2]

DONT jump to conclusions about the integration constants:
( ) ( )
1 0
0 , 0 c x c v .

From Equation 1:
( )
0
0
0
x c a = so
0 0
c v a
0
= +

From Equation 2:
( )
2
1
0
0
x c a = + so
2
1 0 0
c x a =

As a result,
( ) ( )
/
0 0 0
t
x t v a a e

= +
and
( ) ( ) ( )
2 2
0 0 0 0 0
t /
x t x a v a t a e

= + + + .


1-16
32
32
4. Force is a function of position only

** For the time being, lets restrict ourselves to problems in one
dimension **

We have
( )
2
2
d x
m mx F
dt
= = x .

Theres a trick we can use to integrate this out:

( )
mx F x =
( )
mxx F x x = .

Integrate over time:
( )
mxxdt F x xdt =

.
dx
xdt dt dx
dt
= = so
( ) ( )
mxxdt F x xdt F x dx = =

.

Also: note that
2
1
2
d
mx mxx
dt

=

As a result,
2 2
1 1
2 2
d
mxxdt mx dt mx C
dt

= =

+

Finally, we have
( )
2
1
2
F x dx mx C = +
.

1-17
33
33
This is the work-energy theorem.

Define the kinetic energy
2
1
2
T mv so that the change in kinetic
energy is the same as the work done on the
particle.
( )
0
T T F x dx =

The work-energy theorem is is only guaranteed to work when F is
a function of position. Velocityor time-dependentforces are
more complicated. For velocity-dependent forces, work is
expended in other ways (heating something up) in addition to
changing the kinetic energy of the object.

I have no idea why the letter T is used for kinetic energy!

Lets define a special function V(x) (sometimes well call it U(x)
instead) so that
( )
( )
dV x
F x
dx
= .

Then
( )
( )
( ) ( ) ( ) ( )
B B
A A
dV x
F x dx dx V B V A T B T A
dx
= = = +


1-18
34
34
We have a conservation rule in the making: T V = going
from point A to point B so

0 T V + = .

Define the total energy E =T+V as the sum of the kinetic and
potential energies. In this case,
0 E A B =
so E is constant.

This gives us an equation in which the time doesnt appear
explicitly:
T V E + = or
( )
2
1
2
mv V x E + = .

Therefore
( )
( )
2 E V x
v x
m

=

and we can solve for velocity as a function of position.

Finishing the job:

1-19
35
35
( )
dx
v x
dt
= so
( )
dx
dt
v x
= and
( )
( )
1
2
2 E V x
dx
dt dx
v x m

= =

so that
( )
1
2
2 E V x
t d
m

= +

x C
x
.

Crank out the integral, then invert to write x as a function of time.

Heres an example: (a spring). F k =

Step 1: define V(x) so that
( )
( )
dV x
F x
dx
= .
Then:
( ) ( )
F x dx kxdx dV x = =

. As a result,
( ) ( )
2
1
0
2
V x kx V = + (V(0) is a constant of integration).

If we decide to place the zero of potential energy at x =0 then we
have
( )
2
1
2
V x kx = .

1-20
36
36

Step 2: determine E from the initial conditions on v and x since E is
constant:
2 2 2
0 0
1 1 1 1
2 2 2 2
E mv kx mv k = + = +
2
x

Step 3: solve for v:
( )
2
2
1
2
2
2
E kx
E kx
v x
m m

= = .

For convenience, define , 2 k m A E k . We have:
( )
2
v x A x =
2
after a bit of algebra.

As a result,
2
dx
2
A x
dt
= so
2 2
dx
dt
A x
=

.

I expect you to be able to deal with integrals of this sort. Use the
trig substitution sin x A = so that
( )
1
sin / x A

= (sin
-1
is the
arcsine function):

2 2
dx
dt
A x
=

2
cos cos
cos
1 sin
A d d
dt C
A

= = =

+

1-21
37
37

since
2
1 sin cos = and
( )
sin cos d A A d = .

As a result,
( )
1
const. sin / t x

+ = A .

Step 4: solve for x as a function of time.

( ) ( )
1
sin const. sin sin / t x

+ = =

A x A so that
( )
( )
sin x A t t = +
where Ive renamed the constant to be .

Both A and are constants of integration and must be determined
from initial conditions on two of: x(0), v(0), E.

Simple harmonic motion!!

Why is SHM so important? Heres why. Say V(x) (potential
energy) looks like so:


1-22
38
38
x
V(x)
x
V(x)

Local (stable) equilibrium points have
2
2
0 and 0
dV d V
dx
dx
= >
since the force felt by the particle is
dV
F
dx
= and theres no force

at an equilibrium point.

Taylor expand F(x) around an equilibrium point at x =x
0
:

( )
( ) ( ) ( )
0 0 0
0 0
1
!
n
n
n
dF d F
F x F x x x x
x x x x
n dx
dx
x = + + + +
= =

Keep in mind that
( )
0
0 F x = since were at an equilibrium point.
Remember too that
0
x x
dF
dx
=
is a constant since weve plugged in
0
x x = .

If
(
we have
)
0
1 x x
( ) ( )
0
F x k x x where
0
x x
dF
k
dx
=
.

1-23
39
39
The force is a restoring force, pointing opposite to the
displacement from x
0
thus the negative sign in front of k.

Most systems act like simple harmonic oscillators when displaced
a small distance from equilibrium.

5. Force is only a function of velocity
( )
dv
mx m F v
dt
= = so that
( )
m
dv dt
F v
=

.

Perhaps we can do the integral, then invert
( )
t v to find .
( )
v t

We wont get anywhere trying to define a potential energy function
since that had assumed that
( )
( )
dV x
F x
dx
= and F here depends
on velocity.

An example: linear damping force F m v = ( is a positive
constant so the force opposes the velocity, slowing the object).

m
dv
m
dt
= v so
dv
dt
v
= .

1-24
40
40

Integrate: or ln
dv
dt v t C
v

= = +

.

Use both sides as exponents:
lnv t C C
e e e e
t +
= = or
C t
v e e

=

Initial condition: so
( )
0
0 v t v = =
0
C
e v = :
0
t
v v e

= .

Integrate again :
0
t
dx
v v
dt
e

= = so or
0
t
dx v e dt

=

0 t
v
x e C
= + .

Initial condition:
( )
0
0 x x = so
0 0
0 0
or
v v
x C C x

= + = +
which yields
( ) ( )
0
0
1
t
v
x t e

= + x .

Note that
( )
0
0
lim
t
v
x t
= + x which is finite.

The object doesnt get all that far away from us; the bigger the
damping, the closer it stays.


1-25
41
41

What about cubic damping? The damping force is stronger for
large v than in the case of linear damping, but falls off more
quickly for small v. Lets see what happens

Lets take a look at
3
1
2
F mc = v (c is constant, not intended to be
the speed of light.)

m
1
2
dv
m
dt
=
3
cv so
3
2
dv
cdt
v
= .

Integrate to find
2
2
1 1
or ct C v
v c
= + =
t C +
.

At t =0,
2
0
1
v
C
=
so
2
0
1
C
v
= . Therefore,

2
2
0 0
2 2
2
0 0
0
1
so that
1 1
1
v v
v v
ct v v ct
v ct
= = =
+ +
+
.

Integrate again, since v dx dt = :


1-26
42
42
( )
1
2 2 0 0
0 0
2 2
0 0
or 1
1 1
v v dx
dx x dt v v ct dt
dt
v ct v ct
= = = =
+ +

+

This is an easy integral. Let
2
0
1 u v ct + . Note that
so we can write
2
0
du v cdt
( )
1
1
2 2
2
0 0 0
2
0
1 1
2 2
0 0
1
1 2
du
x v v ct dt v u
v c
u du u C
v c v c

+
= + =
= = +

or
2
0
0
2
1 x v ct C
v c
= + + .

Since
( )
0
0 x x = we can solve for C:
0
0
2
x C
v c
= + so
0
0
2
C x
v c
= . As a result, we have

2
0 0
0 0
2 2
1 x v ct x
v c v c
= + + .

Note that as x t .


1-27
43
43
Another example: falling body with linear air resistance. Lets do
this in one dimension: the object falls straight down.

dv
m m v
dt
= + mg (g is negative if up is positive)

Integrate:
dv
dt
v g
=
+

v and define u g so du dv =
in order to write

1 du
dt
u
=

or
1
ln const. u t
= +
or
( )
ln C g v t = +

so that
C t
g v e e

= .

At t =0 so we can write
C
0
g v e

=
( )
0
t
g v g v e

=
or

( )
0
0
t
t
g g v e
g g
v v

= =

+ .
Note that
g
v
regardless of the initial velocity as . t

Integrate once more to get
( )
x t (+x is up):

1-28
44
44
0
t
dx g g
v v e
dt

= = +

so
0
t
g g
dx v e dt

= +

Note that nearly everything in the integrand is constant. We find
0
1
t
g gt
x v e

= + +

C.

Use the fact that
( )
0
0 x x = to finish this off:
0 0 0 0
1 1
or
g g
x v C C x v

= + = +

so that
0 0
1 1
t
g gt
x v e x v
0
g

= + + +

.

After some algebra, we find
( )
0 0
1
1
t
g g
x x v e

t

= + +

.

It is instructive to look at x for large and small times.

Large t:
0 0
1 g g
x x v
t

+ +

(note the constant speed)


1-29
45
45
Small t: expand
t
e

around 0 t = :

( ) ( )
2 3
1 ...
2! 3!
t
t t
e t
= + + so
( ) ( )
2 3
1 ...
2! 3!
t
t t
e t
= +

( )
( ) ( )
( ) ( )
2 3
0 0
0
2 3
0 0 0 0
1
...
2! 3!
1 1
...
2 6
t
t t
g g
x t x v t
x v t g v t g v t

+ + +

= + + +

Note the linear increase in x for small times.

6. Force is a function of everything

In general, this is a mess that needs to be addressed with a
computer. Some special cases do lend themselves to analytic
techniques, however.

For example, if is a product of functions: F


1-30
46
46
a)
( ) ( ) (
, F v t f v g t =
)
( ) ( )

dv
f v g t m
dt
= so
( )
( )
mdv
g t dt
f v
=

and perhaps you can do the integrals.

b)
( ) ( )
dv
F f v h x m
dt
= =
Use a trick:
dv dv dx dv
v
dt dx dt dx
= = .
Then
( ) ( )

dv dv
m mv f v h
dt dx
= = x so that
( )
( )
mvdv
h x dx
f v
=

.

Try to solve for v as a function of x then use v(x) =dx/dt to write
( )
dx
dt
v x
=

and crank it out as best as you can.

c)
( ) ( )
F h x g t = get thee to a workstation!!

The general idea in all of this: Newtons laws and their
consequences (for example, conservation of momentum) will

1-31
47
47
allow us to write differential equations we can solve, even though
we might be forced to resort to numerical (computer) techniques.

Nonrelativistic rocket

Lets work another example, also in one dimension, that lends
itself to analytic (non-computer) solution. Its a non-relativistic
rocket problem, neglecting gravity.

The rocket burns fuel, shooting it out the back as exhaust. The
exhaust speed with respect to the rocket is u; the rockets speed
with respect to an inertial frame is
( )
v t . The tricky part is that the
rockets mass decreases as it burns fuel.

How we deal with this: consider the momentum of the rocket and
fuel that it carries at time t, then equate this to the momentum of
the (slightly lighter, slightly faster) rocket at t dt + added to the
momentum of the exhaust gases. A little algebra will let us
construct a differential equation from the pieces of our momentum
conservation equation.

Momentum at time t carried by the rocket and its onboard fuel:

1-32
48
48
( ) ( ) ( )
rocket
p t m t v t =

where m(t) is the combined mass of the rocket and the fuel it
contains at time t. Lets do our calculations in the inertial frame in
which we see the rocket moving at velocity v at time t.

At time t +dt the rocket is slightly lighter and moving slightly
faster. It carries momentum
( ) ( ) ( ) ( ) ( )
rocket
p t dt m t dt v t dt m t dm v t dv + = + + = + +

(Note that dm is negative.) The change in the rockets momentum
is
( ) ( ) ( ) ( )
( ) ( )
( ) ( )
rocket
p m t dm v t dv m t v
m t dv v t dm dmdv
m t dv v t dm
= + +

= + +
+

t

since the product of the two infinitesimals dmdv
can be neglected.

The change in the momentum of exhaust gas flying through space
is just the product of its mass and its velocity:
( ) ( )
exhaust
p v t u d = +

m

since the exhaust gas has mass dm and travels with velocity
. (Recall that u is with respect to the rocket.)
( )
v t u +


1-33
49
49
Because the rocket +fuel comprises a closed system, the net
momentum in the inertial frame we are using must remain
constant: so 0
rocket exhaust
p p + =

( ) ( ) ( ) (
( )
0
rocket exhaust
p p m t dv v t dm v t u dm
m t dv udm
+ = + + +

= =

)

or
( )
( )
so that
udm
m t dv udm dv
m t
= =

.
Solve for the velocity by integrating from the start of engine burn
(rocket mass is m
i
) to the end of engine burn (rocket mass is m
f
):
( )
( )
so ln
f f
f
i
i i
m m v v
m
f i
m
m m v v
udm
dv u m v v v
m t
= =
= =
= =

Recall that
( )
ln ln ln a b a b = so that
( ) ( )
ln ln
f i i
v u m m u m m = =

f
.
( is in the same direction as u
f i
v v

.)

You do better by increasing the exhaust velocity for a given
quantity of fuel than you do by increasing the fuel carrying
capacity of your rocket.


1-34
50
50
this page intentionally blank

1-35
51
51

1-36

52
52
Physics 325, fall 2012 Dynamics in three dimensions University of Illinois
Dynamics in three dimensions
Dynamics in three dimensions.....................................................2-1
Conservative forces..................................................................2-1
A reminder about the gradient operator.................................2-1
Line integral around a closed loop as a test for a conservative
force.......................................................................................2-5
Form of the equations in three dimensions............................2-8
Interesting special cases.......................................................2-10
Examples of forces derivable from potentials.....................2-15
Zero curl as a test for a conservative force..........................2-20
Two examples in three dimensions......................................2-28

Conservative forces
A reminder about the gradient operator
Recall the definition of the gradient operator in Cartesian
coordinates in three dimensions (see the Math Review, if
necessary):

d
x y z
x dy z

+ +

.

With this definition, we can calculate the change in a scalar
function when we shift from
(
, , u x y z
)
to x x +

(we are
copyright 2012 George Gollin PHY_325_lec_notes_02.doc 2-1

53
53
assuming that
is very small):

( )
( )
( )

x y
u u x u x
u u u
x y z
u u
x y z
z
u
x y z
u
= +

+ +

= + +

=

This is what led us to this particular definition for the gradient: we
can use it to calculate for any scalar function u. u

A note of caution: in non-Cartesian coordinates,
isnt so simple.

For example, you might expect
to be
r z
r z

+ +

in
cylindrical coordinates. That would be WRONG!

Heres why. Keep in mind that we want to use
this way:
( )
( )
( )
u u r u r u = + =

.

Imagine that
points in the
direction, as drawn below:


54
54
x
z
y
r
r
x =r cos
y =r sin
= r
x
z
y
r
rrr
x =r cos
y =r sin
= r
^
x =r cos
y =r sin
= r

Notice that the length of
depends on r since, for a particular

value of the angle , the length of
increases as r increases.

Allowing for
to point in an arbitrary direction, we have

( )

r z
r r

z = + +

rather than the incorrect expression

r z
r z
+ + .

Since
( )
( )
( )
u r u r u + =

is, in effect, forcing us into a

particular definition of , we are required to have
u r z
r r z
u

= + +


55
55
so that dotting
( )
u
with
to calculate
( )
u

will actually give
us u rather than something else.

Heres a summary of in different coordinate systems. u

Cartesian coordinates x, y, z : u x y z u
x y z

= + +

Cylindrical coordinates r, , z :
u r z
r r z
u

= + +

a
x
z
y
r
x =r cos
y =r sin
z
a
x
z
y
r
z
aa
x
z
y
r
x =r cos
y =r sin


56
56
Spherical coordinates r, , :

sin
d
u r
r r r

u

= + +

sin cos x r =
sin sin y r =
cos z r =
a
x
z
y
r
sin cos x r =
sin sin y r =
cos z r =
aa
x
z
y
r

Line integral around a closed loop as a test for a conservative
force
Imagine we map out
( )
U r
along a path from A to B like so:


57
57
A
B
ds
A
B
ds

Along each short segment of path ds
along the route, U will

change by
( )
dU U ds =

. If we add up (actually integrate:
keep in mind that an integral is really just a big sum) the changes
in U over each of the small path segment, well get the net change
in U between A & B:
( ) ( ) ( ) ( )
B B
B
i
A
A A
U B U A dU dU U ds U B U A = = =

.

If the net force acting on our 1-particle system can be set equal to
the gradient of a scalar function of position so that F = U

then
we can write
( ) ( )
B B
A A
U B U A U ds F ds = =

B
A
F ds

so that the work
done by the force
equals the reduction in potential

energy, and the gain in kinetic energy.


58
58
U
Note that the path taken doesnt enter into the result: as long as
we automatically find F =

B
A
F ds U =

.

All we need to know to calculate U is the location of the starting
and ending points A and B of the path to calculate.

Forces that have this property are called conservative forces: if A
and B are the same point (true for any round trip), no net energy is
expended during the trip.

Not all conceivable forces are conservative. Friction is an obvious
example. But also: force fields like the ones sketched below are
non-conservative since you can find loops about which there is a
net energy gain or loss when you go around once.

and and

59
59

Form of the equations in three dimensions

F ma =

is shorthand for 3 equations:

( )
, , , , , ,
x
mx F x y z x y z t =
( )
, , , , , ,
y
my F x y z x y z t =
( )
, , , , , ,
z
mz F x y z x y z t =

If we can write

( )
( ) ( ) ( )
, , , , , , , ,
x y z
F x x t x F x x t yF y y t zF z z t = + +

Then the equations of motion for the , , x y z directions are
decoupled and becomes three (often simpler) equations: F ma =

( )
( )
( )
, ,
, ,
, ,
x
y
z
mx F x x t
my F y y t
mz F z z t
=
=
=

OFTEN we deal with forces that do not separate in

60
60
Cartesian coordinates. A familiar example: 1/r
2
central force
(gravity, for example).

In spherical coordinates,
1 2
2
Gmm
F
r
r
(easy: it separates).
In Cartesian coordinates,
2 2 2
r x y z
2
= + +
and
[ ]
sin cos sin cos r x y z = + +
x
z
y
r
^
x
z
y
r
^
r
^

so everything gets messy. Plugging in yields

[ ]
1 2
2 2 2
cos sin sin sin cos
Gmm
F x y
x y z
z
= +
+ +
+
and since


61
61
2 2 2 2
2 2 2 2 2 2
sin , cos , sin
x y x y
2
x y z x y x y

+
= =
+ + + +
=

we can conclude that

( ) ( ) ( )
1 2 1 2 1 2
3 2 3 2 3 2
2 2 2 2 2 2 2 2 2

x y z
Gmm x Gmm y Gmm z
F F F
x y z x y z x y z

= = =
+ + + + + +

Our equations relating force to the particles coordinates do not
separate. Looks nasty!

Interesting special cases

In general, there isnt an analytic solution to the equations since
they can be quite complicated: the component of force in any
direction might be a function of many variables ( , , , , , , x y z x y z t ).

But there are a number of interesting special cases that yield
simpler equations of motion which can (sometimes) be solved.
Here are some examples:

62
62

1. Say s
is a constant vector (for example, x ). If F
is always
perpendicular to s
, then the component of momentum parallel to

s
will be constant.

Heres the proof:

( )
d p s
dp
F s s
dt dt

= =

since s
is constant and can be brought

inside the derivative.

Because is perpendicular to F
well always have so

that it will also be the case that
0 F s =

( )
0
d p s
F s
dt
= =

: the value of
never changes since its time derivative is zero. p s

The length of the component of p
along a direction parallel to s
is
s p s
p
s s
. In order for p s

to remain unchanged, the
component of parallel to p
must remain constant.

2. Another special case: define the torque exerted by a force to be
r F =

where r is measured with respect to a fixed point in an

63
63
inertial reference frame. Define angular momentum to be
. Then, if L r p =

0, L =

is constant and the object moves in a

plane perpendicular to . (Note that when F
is a central force so
that , well automatically have r F F

r 0 =
.)

Heres the proof:

( )
( )
d d d dp
L r p r
dt dt

= =

p r
dt
dt

0 =
dr
p v mv
dt
=

.

Therefore,
dL dp
r r F
dt dt

= =

.

So if 0 =
, then is constant. Also, since L
( )
p r p

,
must be in a plane perpendicular to
p
.

You might not be comfortable with this part of the first step:
( )
d d dp
r p r r
dt dt dt

= +

p

. If so, consider writing out the

64
64
expression in terms of components and grinding it out:
( ) ( ) ( )

y x z y x z
r p xp yp z yp zp x zp xp y = + +

so
( ) ( ) ( ) ( )

y x z y x z
y
x
y x
z
z y
x z
x z
d d
r p xp yp z yp zp x zp xp y
dt dt
dp
dp dx dy
p x p y z
dt dt dt dt
dp dy dz dpy
p y p z x
dt dt dt dt
dp dp dz dx
p z p x y
dt dt dt dt
= + +

= + + +

+ +

+ +

+
so that

65
65
( )
[ ]
y
x
x y y x
z
y z z y
x z
z x x z
x y y x y z z y
z x x z
y
x
dp
dp d
r p v p x v p y z
dt dt dt
dp dpy
v p y v p z x
dt dt
dp dp
v p z v p x y
dt dt
v p v p z v p v p x
v p v p y
dp
dp
x y
dt dt

= + + +

+ +

+ +

= + + +

+
z
x z
dp dpy
z y z x
dt dt
dp dp
z x y
dt dt

+ +

or
( )
d dp dr
r p v p r p r
dt dt dt dt
= + = +
dp

.

Whew.

3. Last special case well discuss, no proof to be offered

If is only a function of position, and if F
B
A
F ds

is independent
of the path taken between A and B, then one can always define a

66
66
potential such that
( )
U r
( ) ( )
U r F r =

B B
A A
dT
. (We discussed this
earlier, but for one dimensional systems.) The great utility of U is
that the sum of potential and kinetic energies is constant in such a
system.
Recall the work-energy theorem: F ds =

where we have
defined kinetic energy in the usual way as
2
2 T m = v
U
. Since
we can write
( ) )
U r r
(
F =

( )
B B
A A
T
U B
=

T =
( )
T d F ds
U A
= =
=

B B
A A
U ds dU
U
=
=

Since we can define the total energy E T U + to find
that E is conserved in systems of this sort: regardless of the choice
of A, B, and the path connecting them it is always true that
E T U + U = + 0 U = = . The total energy E is constant
everywhere along the path linking A & B.

Examples of forces derivable from potentials

67
67

Here are some examples of forces derivable from potentials.

1. Uniform gravitational field:

z F mg =
.

Note that so
( )
mgz mgz =
const. U mgz = + is a solution.

We could have written the gradient in Cartesian coordinates and
integrated, if we wanted to brute-force it:

F mgz U x y z U
x y z

= = = + +

so

0
U U
x y

= =

and U can only be a function of z.

As a result,
U dU
z d
=
z
and we can integrate:

68
68

dU
F mgz z
dz
= =
or
dU
mg
dz
= so that mgdz dU =

.

As before, we find that const. U mgz = +

Since the sum of kinetic and potential energies is constant, we have
2
1
constant
2
mv mgz E + = =
while the object falls.

2. Inverse square law:
2
kr
F
r
x
z
y
x
z
y
rr

Note that it is most natural to use spherical coordinates, since
always points inwards, towards the origin. Lets figure out what F
(
, , U r
)
is. (Clearly U is independent of , .)


69
69
( )
( )
3 2 3
2 2 2 2
x x y y z z
kr kr
k
r r
x y z
+ +

= =
+ +

What function u has
3
kr
r

as its gradient? Lets take a look: use the
form of in spherical coordinates (see the Math Review).

( ) ( )
, , , ,
sin
U r U r F r
r r d r

= + + =

.

Since always points towards the origin and doesnt have a F
or component, 0
U U

= =

. As a result, U is only a function
of r. Therefore,
( )
2
U r
kr
U r
r r
= =

Since U only depends on r,
U dU
r dr
and we can integrate:

( )
2
dU r
kr
r
dr r
= so
( )
2
dU r
k
dr r
= or
2
k
dU dr
r
=

.


70
70
We find
( )
constant
k
U r
r
= + .

3. A last example: the Yukawa force, important in
understanding nuclear potentials. The force is a central force: it
depends only on r, not on or .

0
2
0
1
r r
k r
F e
r r
r

= +

.

Note that the force becomes the same as the Coulomb force when
. For the strong nuclear force, (a fermi!)
0
r r
13
0
10 cm r

The force is radial so

0
2
0
1
r r
U k r
F U r e
r r r
= = = +

.

Integrate this, just like you would for a total derivative:


71
71
0 0
2 2
0 0
1 1
r r r r
U dU k r k r
e U e
r dr r r r r

dr

= = + = +

We can take all sorts of brute-force approaches to this integral, but
take note of the following useful fact:
0 0 0 0
2 2
0 0
1
r r r r r r r r
d k k k k r
e e e e
dr r r rr r r

= + = +

We can use this to rewrite the integrand so that

0 0
2
0
1
r r r r r r
k r d k k
U e dr e dr e
r r dr r r

= + = =

0
.

For nuclei, k ~1 GeV/fermi ~1.6 10
5
J oules/meter.

Zero curl as a test for a conservative force
In the examples above we derived position-dependent potentials
from conservative forces. The fact that we could find potentials
that depend only on position guaranteed the conservative
nature of the forces.
( )
U x

72
72

The path independence of the potential energy is equivalent to the
fact that the net change in potential around any closed loop is zero:
path independence means that both staying put and also going on a
round trip will result in since the starting/ending points are
the same.
0 U =

U
Theres a way to investigate whether or not a force is conservative
without having to integrate it to solve F =

for the potential
U.

Lets work with a microscopic version of the no-work-around-a-
closed-loop idea to develop another test for a conservative force.

Imagine that a force field acts on a particle as we move it
counterclockwise around a tiny rectangular loop, as shown in the
diagram.

73
73
z
y
x
Rectangular loop is
centered at (x, y)
y
x
1
2
3
4
x-x/2 x+x/2
y-y/2
y+y/2
z
y
x
Rectangular loop is
centered at (x, y)
y y
x x
11
22
33
44
x-x/2 x+x/2
y-y/2
y+y/2

The midpoints of the four sides of the loop are at positions
( , 2, 0) x y y , ( 2, , 0) x x y + , ( , 2, 0) x y y + , and
( 2, , 0) x x y .

The work done by the external force as we push the object around
the loop is .
loop
F ds

If the loop is small enough we can approximate the integral: the
force that acts as we move the mass along each side can be well
represented as constant, with the value it assumes at the midpoint
of each side.

The force acting on our particle as it moves along side 1 is

74
74
approximately
1
( , 2) F F x y y =

, the value halfway between the
ends of side 1.

The forces acting as the particle moves along the other sides are
approximately
2
( 2, ) F F x x y = +

,
3
( , 2) F F x y y = +

, and
4
( 2, F F x x y =

).

The line integral around the loop can be approximated as a
sum over the four sides:
F ds
4
1
i i
i
loop
F ds F s
=

with
1
s x x =
,
2
s y y =
,
3
s x x =
, and
4
s y y =
.

We have:

75
75
4
1
( , 2) ( 2, )
( , 2) ( 2, )
( , 2) ( 2, )
( , 2) ( 2, )
( , 2) ( , 2)
( 2, ) (
i i
i
x y
x y
x x
y y
F s F x y y x x F x x y y y
F x y y x x F x x y y y
F x y y x F x x y y
F x y y x F x x y y
F x y y F x y y
x y
y
F x x y F
=
= + +
+
= + +
+
+
=

+
+
2, ) x x y
x y
x

where I have grouped the terms and written the force in terms of its
x, y components:
x y
F F x F = +
y ( and
x y
F x F F y F = =

).

Recall the definition of a partial derivative:

( ) ( ) ( )
( ) ( )
, , ,
0
, 2 ,
0
x x x
x x
F x y F x y y F x y
lim
y y
y
F x y y F x y y
lim
y
y
+

=
2

We can use this to rewrite

76
76
4
1
y
x
i i
i
F
F
F ds F s x y x y
dy dx
=
= +
.

If is conservative, F
0 F dx =
and well have

0
y
x
F
F
x y
dx dy

=

.

Define the curl of a vector function F
to be

y x z y x z
F F F z F F x F F
x y y z z x

= + +

y .

With this we can write
( )
0 F ds F z x y =
.

Our conclusion is that conservative forces have zero curl.

Note that theres a compact way to write the definition of the curl
using notation described in the Math review. Recall the definition
of the Levi-Civita symbol
ijk
:


77
77
( )
( )
1 if is an even permutation of 123 123, 312, or 231
if is an odd permutation of 123 213, 132, or 321
if any two (or more) of , , are equal
1
0
ijk
ijk
ijk
i j k
+
=

Its not hard to see that

3 3 3
1 1 1 , ,

also written as
j i k
ijk i j ijk i j
i j k i j k
a b a b k a b k
= = =
= = =

=

and that
, ,
j
ijk
i j k
i
F
F k
x

where
1
x x ,
2
x y , and
3
x z .

Note that we can express F ds
as a sum of line integrals around

subloops:


78
78

z
y
x
y
x
1 2
3 4

Along each small interior side we always have the edges of two
small loops touching, with opposite directions of travel. Since F
is
the same, but s
in the opposite direction along interior loops

adjacent edges, the contributions to a grand sum will cancel. As a
result,
small loops
big loop loop

i
i
F ds F ds =

.

Lets define an area patch dA
with an associated direction using

the right-hand rule:


79
79
z
y
x
dy
dx
dA dxdy z =
z
y
x
dy
dx
z
y
x
dy dy
dx dx
dA dxdy z =

By doing this we can rewrite the sum over small loops as an
integral:
( ) ( )
tinyloop
F ds F zdxdy F dA =

so
( )
big loop
F
tinyloops
i looparea
F ds F ds dA = =

This is Stokes Theorem, named for the Irish mathematical
physicist George Gabriel Stokes (1819-1903); it is true for any
vector field.

Two examples in three dimensions
1. Projectile in a uniform gravitational field with linear air
resistance.


80
80
v Linear air resistance: F mgz m =

.

Fortunately, this equation separates:

x x
F m v m x = =
y y
F m v m y = =
( )
z z
F m v mg m g z = = +

As a result,
( )
, ,
y
x z
x y z
dv
dv dv
v v g
dt dt dt
v = = = + .

Easy to solve:
ln
x
x
x
dv
dt v t c
v
= = +

so
c
x
v e e
t
= .

At t =0,
0 x x
v v = so
0
c
x
e v = . The y solution works the same way:
0 y y
v v
t
e

= .

Now for z: let
z
u v g + . Note that
z
du dt dv dt = .

81
81
We have
( )
z
z
dv du
g v
dt dt
u = + = so
c t
u e e

= .

At t =0
0 z
u v g = + so
( )
0
t
z
u v g e

= + .

As a result,

( )
0
t
z z
v u g v g e g
= = + .

To solve for
( )
x t
, integrate and use

( )
0
0 x x =

. The final result is

( )
0 0
1
t
gt gz e
x t x z v

= + +

Is this sensible? Lets check.

In the limit 0
,
x y
v v
(no air resistance), this should look like a freely
falling body
( )
constant, etc. etc.

Use
( )
2
1
2!
t
t
e t
= + + to rewrite


82
82
( )
2
2
1

2 2
t
t
e t
t
= + .

As a result,
( )
( )
2
2
0 0 0
1
2
2
t
x t x v t gt z v

= + +

Yep, looks ok. In the limit 0 we have
( )
2
0 0
1
2
x t x v t gt + z

.

2. Motion of a charged particle in a constant magnetic field.

The equations dont separate, but theres a handy trick youll find
useful in other circumstances.

Lets say and B B =
z 0 E =
.

The Lorentz force is
1 2
F k qE k qv B = +

. The specific values for

depend on the units you choose for charge, length, etc. In
the commonly use SI system of units (Tesla, meters, Coulombs,
J oules, and so forth)
1
and k
2
k
1 2
=1 and 1 k k = .


83
83
We have
( )
2 2 2

y x
F k qv B k qv zB k qB v x v y = = =

.

The equation F ma =

becomes

2
or
x y
k qB k qB
a v x
m m
= =
2
y [Equation 1]
2
or y
y x
k qB k qB
a v
m m
= =
2
x [Equation 2]
0
z
a =

Looks nasty: the equations are coupled, since the x component of
acceleration depends on the y component of velocity!

Heres the trick: define , u x iy w x iy + so that
,
2 2
u w u w
x y
i
+
= =

Then add Equation 1 to i times Equation 2:
( ) ( ) ( )
2
2 2
2
y
y
k qB
x y
m
k qB k qB
x i ix y i
m m
k qB
i i x
m
+ + = + = +

x iy

84
84
or
2
k qB
u i
m
= u .

Now subtract i times Equation 2 from Equation 1 and do a small
amount of algebra to find

2
k qB
w i w
m
=

Integrate the u equation once:
2 2
so
ik qB ik qB du du
u u
dt m u m
= = =

dt
or

( )
2
ln
ik qB
u t
m
= + c .

Similarly, integrate the equation: w
2 2
so
ik qB ik qB dw dw
w w
dt m w m
= = =

dt
or

( )
2
ln
ik qB
w t
m
= + c .


85
85

A
To save effort better, Im going to be clever about redefining the
integration constant : c ln c i + . I will require A to be a real
number.

Since
( )
2 2
ln ln
ik qB ik qB
u t c t i
m m

= + = + A

we find (after using the left and right sides of the last equation as
exponents)
2
2 2
ln
ln
k qB
ik qB ik qB
i t
t i A t i
A m
m m
u e e e Ae

+
+

= = = .

Recall that cos sin
i
e

i
= so that

2
2 2
cos sin
k qB
i t
m
k qB k qB
u Ae A t i t
m m

+

= = +

+

and that u x . iy = +

The real part of u is x while the imaginary part of u is . As a
result,
y

86
86

2
cos
k qB
x A t
m

= +

and
2
sin
k qB
y A t
m

= +

.

Integrating once more gives
2
2
sin
k qB Am
x t
k qB m

= +

and
2
2
cos
k qB Am
y t
k qB m

= +

.

Note that
2
2 2
2
Am
x y
k qB

+ =

independent of time.

Therefore,
2
2 2
2
Am
x y
k qB

+ =

or
2 2
2
0 0
k qB
A x y
m
= +

87
87

2-36

so we can write
2 2
2 2
0 0
2
2 2
2
0 0
sin
sin
k qB k qB m
x x y t
m k qB m
k qB
x y t
m

= +

= + +

and, similarly,
2 2
2
0 0
cos
k qB
y x y t
m

= + +

.

We can solve for the remaining unknown :
0 0
sin cos tan x y = =
so
( )
1
0 0
tan x y

= .

Note that the motion (projected onto the x-y plane) is a circle. If we
allow for the a non-zero (but constant) velocity in the z direction
then the particle will move in a helix.

88
88
Physics 325, fall 2012 Systems of particles University of Illinois
Systems of particles
Systems of particles.....................................................................3-1
Definition of the center of mass................................................3-1
Motion of the center of mass....................................................3-2
Total angular momentum and torque........................................3-4
Reduced mass coordinates for two-particle systems................3-5

So far weve only been dealing with a single particle moving under
the influence of an externally applied force. A system of particles
exhibits more complicated behavior: external forces can act on the
system, and forces between particles in the system can also affect
things.

Definition of the center of mass
There are some useful definitions we can employ to simplify the
description of whats going on.

Define the position of the center-of-mass of the system to be
1
R m r
M

where M m
and the sum runs over all particles in the system.


89
89
Lets define f
to be the force exerted on particle by particle

. We know f f
from Newtons laws.

Motion of the center of mass
Useful fact: if external forces , , , ... F F F

act on particles , ,
, the response of the center-of-mass position will be the same
as if there were a single particle of mass
i
i
M m =
at the point R

being acted on by a force
i
F F

. The particles dont even have
to be bound together to make this true.

Proof: the
th
particle feels a net force which is a sum of the inter-
particle forces f
acting on it added to the external-to-the-system

force F
. As a result,
m r F f

= +

.
We could have included = in the sum, since a particle does not
exert a force on itself: 0 f f

= =

.

Summing over all particles:

90
90
,
m r F f F f

= + = +

.

The double-sum on the right is zero because of Newtons law
concerning equal-and-opposite forces. Lets list terms explicitly to
see this:

12 13 14
,
21 23 24
31 32 34
+
+
f f f f
f f f
f f f

= + + +
+ + +
+ + +

Note that , etc. etc. For each team
12 21 13 31
0, 0 f f f f + = + =

ij
f
in
the sum theres a corresponding term
ji
f
that appears and cancels

the
ij
f
term: . 0
ij ji
f f + =

As a result, we can rewrite the equation
,
m r F f

= +

as . m r F F

=

Also, since
2
2
d
m r m r
dt

=

and
1
R m r
M

we can

91
91
rewrite once more:
m r F MR F

= =
. (QED!)

This works for things like exploding projectiles, where the external
force acting on individual pieces can vary significantly from piece
to piece. It works even if some pieces go on to orbit while others
fall to earth.

Total angular momentum and torque
Another useful system-of-particles-quantity is the total angular
momentum: pick an origin, at rest in an inertial frame. Then

L r p

.

Define the net torque as r F

.

It turns out that
dL
dt
=

, so if 0 =
the total angular momentum of

the system is constant. Here, F
are the externally applied forces;

inter-particle forces dont enter into the picture. (look in Marion,

92
92
or elsewhere, for a proof of this if youre interested.)

Reduced mass coordinates for two-particle systems
Lets consider a system of two particles which interact through a
central force so that
1-on-2
f
is parallel to
2
r
1
r

and its magnitude
depends only on
2 1
r

r
. Naturally,
1- 2-on-1 on-2
f f =

.

Lets further restrict ourselves to the case where no external forces
act on the particles.

We have
( )
1 1 2 on 1 2 1
ma f f r r

=

.
Note that were just using this as a way to define a force
( )
2 1
f r r

in terms of .
2 on 1
f

We also have
( )
2 2 1 on 2 2 1
m a f f r r

= = +

from Newton.


93
93
With a small amount of algebra it is easy to show that

( )
( )
( )
1 2 2 1 2 1 1 2
2 1
1 2 1 2 1 2
1 2
2 1
1 2 1 2
2 1
mm a mm a mm
a a
m m m m m m
m m
f r r
m m m m
f r r
=
+ + +

= +

+ +

=

so that
( ) (
1 2
2 1 2 1
1 2
mm
a a f r r
m m
)
=
+

.

Defining the reduced mass to be
1 2
1 2
mm
m m

+
we can rewrite the
previous equation as
( ) ( )
2 1 2 1
a a f r r =

1
.

If we define we will have turned our two-particle
systems equations of motion into the same equation wed use for a
single particle of mass
2
r r r

:

( )
r f r =

.

94
94

3-7
2

We can always extract from
1
, r r

r
and the center-of-mass

position : R

1 2 2
1 2 1
1 2 1 2 1 2
2 1 1
2
1 2
m m m
R r r r R
m m m m m m
r r r m
r R
m m

= + =

+ + +

=
= +
+

r
r

This sort of technique will let us deal with situations like orbits of
binary star systems.

95
95


96
96
Physics 325, fall 2012 Oscillations University of Illinois

Oscillations
Oscillations..................................................................................4-1
An overview.............................................................................4-2
Force vs. displacement and period vs. amplitude.....................4-4
Nature of the motion close to an equilibrium point...............4-4
Simple pendulum, done two ways: energy method; force
method...................................................................................4-8
Behavior of the period.........................................................4-12
Successive approximation approach to the nonlinear problem...
............................................................................................ 4-13
Next order solution..............................................................4-21
Size of the effects induced by nonlinearities.......................4-26
A reminder about initial conditions........................................4-28
Phase diagrams.......................................................................4-29
Inhomogeneous elastic deformations: an introduction to tensors
................................................................................................4-32
Strain (displacement) vs. stress for identical springs...........4-35
Strain vs. stress for an inhomogeneous set of springs.........4-36
rotations...............................................................................4-40
Rank-n tensor transformation properties.............................4-47
The outer product of two vectors is a tensor........................4-49
Using complex exponentials to solve the oscillator equation.4-51
Damped oscillations................................................................4-55
Case 1: underdamped oscillations........................................4-57
Oscillator Q (Quality factor)................................................4-59
Case 2: overdamped oscillator.............................................4-63
Case 3: critically damped oscillator.....................................4-64
Electrical analogues................................................................4-69
Driven oscillations..................................................................4-73
Particular solution to the inhomogeneous equation.............4-76

97
97

x
p
.............................................................4-85
Non-sinusoidal driving forces.................................................4-96
Fourier series........................................................................4-100
Orthogonality relationships...............................................4-102
Square wave.......................................................................4-108
Periodic impulses of short duration...................................4-113
Impulse response and an introduction to Greens functions.4-119
Response of an oscillator to a single, short impulse..........4-121
Response of an oscillator to a non-periodic driving force.4-126
Greens functions...............................................................4-128
Greens function for the Coulomb potential ......................4-130
Satisfying the initial conditions.........................................4-132
Some Greens function examples......................................4-134

Well spend about two and a half weeks on oscillations. There are
many useful techniques that come into play here, including the use
of orthogonal functions to generate series solutions to the
equations of motion of linear systems. Please see sections in
Thornton and Marion for material not covered in the lecture notes.

An overview
Weve discussed one-dimensional simple harmonic oscillation
briefly already. We found that a force of the form F k = gave

98
98

rise to
( ) ( )
sin x t A t = + with k m = . The two constants
and A came about from integrating the differential equation
twice to solve for mx = kx
( )
x t
and
.

A We can determine both given two appropriate pieces of
information such as the initial position and velocity
0
,
0
x v or the
initial position and kinetic energy
0
,
0
x T or the initial potential
energy and total energy U or
0
, E

Recall that the oscillators potential energy is
( )
2
2 U x kx = .

What more is there to this oscillation business than just
( )
x t
( )
sin A t = + ?

Lots.

We can investigate what happens if
an oscillator is damped so v F kx = ;
a (damped) oscillator is driven by a sinusoidal driving force
with frequency
d
that is not necessarily the same as k m;

99
99

the oscillator isnt linear so
n
F kx v x = + ;
the oscillator is driven by an arbitrary driving force
( )
F t
rather than a sinusoidal force;
the oscillator can move in three dimensions.

To address these issues well make use of some mathematical tools
with which you may be unfamiliar:

Fourier analysis and linear algebra;
Greens function techniques;
tensors.

Dont be alarmed! You already know about Greens functions,
even though you didnt hear them being called by that name.
Tensors arent such a big deal, either.

Force vs. displacement and period vs. amplitude
Nature of the motion close to an equilibrium point

100
100


4-5
Imagine that a mass moves in a one-dimensional potential with a
stable equilibrium at some point which well refer to as x
0
.

A stable equilibrium corresponds to a local minimum in so
( )
U x
0
0
x x
dU
dx
=
= and
0
2
2
0
x x
d U
dx
=
> .

( )
U x might look like so:
x
U(x)
local minima

There are three local minima in the figure; for sufficiently small
amplitudes each can serve as an equilibrium point for oscillations,
though only one is a global minimum.

We can Taylor-expand
( )
U x about a minimum at x
0
:

101
101

( ) ( ) ( )
( ) ( ) ( )
( ) ( )
0 0
0
2
2
0
0 0
2
3
3
0
3
2!
3!
x x x x
x x
dU x x x d U x
U x U x x x
dx dx
x x d U x
dx
= =
=
= + +
+ +

Because is a minimum,
(
0
U x
)
( )
0
0
x x
dU x
dx
=
= and
( )
0
0
x x =
>
2
2
d U x
dx
.

Note that F dU d = x
x
so it doesnt matter whether we define
or any other (constant) value U
0
: the derivative zaps the
constant when we calculate the force so the force will always be
zero at a minimum in the potential energy.
( )
0
0 U x =

Since the force corresponds to a quadratic potential F k =
2
2 kx
we have (from our Taylor expansion for F )

( )
( ) ( )
( )
0
2
2
2
0
0
2
1
2! 2
x x
x x d U x
U x k x x
dx
=
= + = +

102
102


so the spring constant is
( )
0
2
2
x x
d U x
k
dx
=
= .

If wed instead expanded
( )
F x , wed have gotten
( ) ( ) ( )
( ) ( ) ( )
0 0
2
2
0
0 0
2
2!
x x x x
dF x x x d F x
F x F x x x
dx dx
= =
= + + +

Keep in mind that
( )
0
0 F x = since x
0
is an equilibrium point:

( ) ( )
( ) ( ) ( )
( )
( ) ( )
0 0
0
2
2
0
0
2
2
2
0
0
2
2!
2!
x x x x
x x
dF x x x d F x
F x x x
dx dx
x x d F x
k x x
dx
= =
=
= + +
= + +

If the mass moves in a potential thats described to us, we can
locate a minimum, Taylor-expand the potential around the
minimum, and immediately read off the oscillation frequency by
identifying

103
103

( )
0
2
2
x x
d U x
k
dx
=
= and k m = .

Simple pendulum, done two ways: energy method; force method
Heres an example that we can do using potential energy.

cos
mgsin

The mass swings on an inextensible massless rod of length .
When the mass is at an angle its height above the lowest point in
its swing is cos . At the distance along its arc that it has
travelled away from its equilibrium point is s = .

Lets write the potential energy in terms of s.
mg
mg
mgcos

104
104


[ ] ( )
cos 1 cos U mg mg = =

Recall
2 4
cos 1
2! 4!

= + so that

( )
2 4
2 4 2 2 4 4
2
2 2 4 4 2 4
2 2
1 cos 1 1
2! 4!
2! 4! 2 24
.
2 24 2 24
U mg mg
mg
mg
mg mg s s

= = +

= + = +

= + = +

Therefore
( )
2 4
3
1
.
2 24
mg mg
U s s s =

Since the potential is quadratic in s, the pendulum executes simple
harmonic motion. Further, because

( )
0
2
2
x x
d U x
k
dx
=
= and k m =

105
105

we can read off the spring constant and angular frequency
immediately: k mg = and g = .

If you prefer, we could have worked the problem using force
instead of energy, keeping in mind that the force component
perpendicular to the arc is cancelled by the tension in the rod.

We have
( )
3
3 3
3
3
sin
3!

3!
.
3!
F mg mg
s s
mg
mg mg
s s F s

= = +

= +

= + =

The kinetic energy is
[ ]
2
2
2 2
2
1 1
2 2
1
2
1
.
2
d
mv m
dt
m
ms

=

=
=

106
106

The pendulums acceleration along its arc is s =

.

Note that theres also a component of acceleration perpendicular to
the arc: acceleration is
dv
dt
, and since both the direction of v and

the magnitude
change, a
is neither completely tangent to the

arc nor perpendicular to it. (Recall that motion in a circular arc
with constant speed is associated with an inwards centripetal
acceleration that is perpendicular to the velocity.)

Putting this all together, we have
( )
3
.
3!
s mg s
F s mg ms

= + =

If 1 s we can ignore the
( )
3
s term to write

mg
ms s =

so that
( ) ( )
sin s t A t = + . As before,
g
=

. and A have
to be determined from initial conditions.

107
107


Behavior of the period
Note that the period of the motion is determined only by the ratio
g and not by the amplitude of the motion. The period is the
same, independent of how large the swings become, as long as s
stays small.

Whats happening is this: the further out the pendulum goes, the
stronger the return force becomes. When F kx = , the increased
strength of the return force exactly makes up for the increased
distance to be covered to return to equilibrium.

Wed expect that the period for the pendulum should increase
when the amplitude is large, since the return force grows less
quickly then linearly:
( )
3
2
6
mg s
F s s

= +

108
108


4-13
s
|F(s)|
perfectly linear
pendulum

Successive approximation approach to the nonlinear problem
We can actually get a handle on this via the method of successive
approximations. This is a very powerful technique, useful in many
circumstances (e.g., quantum mechanics) where you cant solve
the problem at hand exactly, but can write a solution to a similar
problem.

This is discussed in various textbooks; heres my own version of it.

Lets say we can express
( )
F x as a series in powers of x, where
all terms after the part are small for all values of x accessible
to our oscillator (since its energy is fixed:
kx
( )
2 3
2 3
F x kx a x a x = + + +.

109
109


As long as
2
2
a x kx ,
3
3
a x kx everywhere that the
oscillator can reach, we expect the motion to be very similar to that
of the linear-force-law oscillator.

If the force grows with x less quickly than kx , as is the case for
our pendulum, were
3
~ 6 F s s
2
, we expect the oscillation
period to increase with increasing amplitude, since the restoring
force grows weaker (compared to the linear case) then wed need
to keep to period (frequency) constant, independent of amplitude.

Lets try modeling the motion as a sine with shifted frequency:
( ) ( )
sin x t A t = . Lets plug this into our F ma = equation and see
what happens. (Im going to assume all terms except and
are zero to make the problem tractable.)
kx
3
3
a x

If I assume that
( ) ( )
sin x t A t = then, differentiating twice and
multiplying by m yields
( )
2
sin t mx mA = .

The restoring force is
( )
2 3
2 3 3
F x kx a x a x kx a x = + + + +
3

110
110

in this example. Plug in
( ) ( )
sin x t A t = and see if it is possible to
make the equation work by choosing properly.

3
3
sin kA t a A
3 3
3
sin kx a x t We have + =
mx kx =
+ so now lets see
if we can make this work in the equation .
3
3
a x +

Since our assumption for the form of x gave us
2
sin mx mA t = ,
the equation turns into
3
3
a x
sin t kA t
mx kx = +

2 3
3
sin sin mA a A t
3
+

where Ive replaced the = with a because I dont know if it
can be made to be true. Rework the non-equation a little:

2 2
3
sin k m a A
2
t
Nope, it doesnt work: the left side is constant, while the right side
wiggles. Rats.

But how bad is the disagreement?

111
111


4-16
a
3
A
2
t

If we pick
2 2
3
2 k m a A = things wont be as far off as theyd
be if we left
2
as
2
0
k m .

Lets try this: since
2
0
m k = , lets use
( )
2 2 2
0 3
2 m a = A or
2 2
2 2 2
3 3
0 0
1
2 2
a A a A
m k

= =

.
For small , we can use a binomial approximation to do the
square root:
3
a
2
3
0
1
4
a A
k

.

Notice the coupling between the best frequency and the
amplitude: the larger A, the lower the frequency (and the longer the
period).

As a result,

112
112

2
3
0
sin 1
4
a A
x A t
k

=

ought to be a better description of the motion than
( )
0
sin x A t = .

We need to pay attention to what we mean by better: is the
best thing to do to minimize the amount of average disagreement
between
( )
2
k m and
2
3
sin a A t ? Depending on what we really
mean by best we might be led to different choices about how to
make our approximations.

We might have gone about things in a slightly different fashion.

For example, lets try dealing with the
3
sin t in the expression
2 3
3
sin sin sin mA t kA t a A t
3
+
by using a trig identity which you can prove using complex
exponentials:
( )
3
sin 3sin sin3 4 t t = t
3
.

With this,
2 3
3
sin sin sin mA t kA t a A t + turns into
( )
3 3
2
3 3
3
sin sin sin3
4 4
a A a A
kA mA t t t

113
113

or
2 2
2 2
3 3
0
3
sin sin3
4 4
a A a A
t t
m m

where Ive used
2
0
k m .

In this approach we want to make the sin t term look as much
like the sin3 t term as possible. How do we decide what is best?

One possibility is to make the average of the square of the
difference between the two terms be as small as possible. (This
isnt the only possibility; its not clear at this point what the best
one is.) Rewriting slightly,
2
2 2
3
0
2
3
3 4
sin sin3
4
a A m
t t
a A m

Doing this: lets see what value of B makes
( )
2
sin sin3 B t t
as small as possible. The large angle brackets
( )
mean to take
the average of whats inside the brackets. Well use this as a guide
to our choice of this time around: well want to make
2
2 2
3
0
2
3
3 4
4
a A m
B
a A m

=

114
114

assume the right value by choosing properly.

Lets calculate the average of the function
( )
2
sin sin3 B t t
over one period of sin t :

( ) ( ) ( )
2
2 2
0
2
2 2 2
0
1
sin sin3 sin sin3
2
1
sin sin 3 2 sin sin3
2
t
t
B t t B t t d t
B B d

=
=
=
= +

where I have made the substitution t . Work this out more:

2 2
2 2 2 2
0 0
sin , sin 3 B d B d

= =

, and

( )
2
2
0 0
sin2 sin4
sin sin3 0
4 8
d
= so

( )
2 2
2 1
sin sin3
2 2
B B
B t t

+ +
= = .

The minimum value occurs for B =0, of course.

115
115


Since we had defined
2
2 2
3
0
2
3
3 4
4
a A m
B
a A m

we should use
2 2
2 2 2
3 3
0 0
2 2
2
3 3
0 0
2
0
3 3
or
4 4
3 3
1
4 4
a A a A
m m
a A a A
k k

= =
= =

since
2
0
k m = .

We can use a binomial approximation to do the square root:
2
3
0
3
1
8
a A
k

.

Note that we got a somewhat different answer than the one a few
pages back!

Whats happening is that we havent really defined what best
means, or at least not why one choice is better than another.

116
116


Next order solution
Our first order solution was sin x A t = with
2
3
0
3
1
4
a A
k

=

.

Keep in mind that were looking for an approximate solution to the
equation of motion . When we plug our first
order trial solution for x into this equation we find
3
3
0 mx kx a x + =

2 3
3
sin sin sin 0 m A t kA t a A t +
3

since our solution isnt exact.

Recall that and use this to rewrite the above expression:
2
0
m =
( )
2 2 3 3
0 3
sin sin 0 mA t a A t . Using that trig identity to
replace the last term yields

117
117

( )
( )
2 2 3 3
0 3
3 3
2 2 3 3
0
2 3
2 2
3 3
0
sin sin
3 3
sin sin sin3
4 4
3 3
sin sin3
4 4
0.
mA t a A t
a A a A
mA t t t
a A a A
mA t t
m

= +

= +

We had picked so that
2
2 2
3
0
3
0
4
a A
m
=
)
. As a result, only
the
(
sin 3 t term is there to spoil things for us.

We can try an improved (next order) solution by guessing that a
modification to our first order correction might work: rather than
just adjusting the oscillation frequency, perhaps we can add a small
amount of
(
sin 3 t
)
into our trial solution.

Lets try
( ) ( ) ( )
sin sin 3 x t A t B t = + , using the same and A
as before, but now solving for B.

Plus this version of x into the left side of our equation of motion
that reads and recall
3
3
0 mx kx a x + =
2
0
k m = :

118
118

( )
( )
3 2
3
2
3
3
sin sin
9 sin3 sin3
sin sin3
mx kx a x mA t kA t
mB t kB t
a A t B t

+ = +
+
+

( )
2 3
sin3 3 sin 3 mB g t a B t
3

( )
3
3
sin sin3 a A t B t +

( )
( )
( )
2 2 3 3
0 3
2 2 3 3
0 3
2 2 2 2
3 3
sin sin
9 sin3 sin 3
3 sin sin3 3 sin sin 3
mA t a A t
mB t a B t
a A t B t a AB t t

=
+

( )
( )
2 3
2 2
3 3
0
2 3
2 2 3 3
0
3
3
sin sin 3
4 4
3
9 sin3 sin 9
4 4
3 sin sin3 sin5 sin7
4 2 2 4 4
a A a A
mA t t
m
a B a B
mB t t
m
A B A A B B
a AB t t t t

= +

+ +

+ + +

This looks really messy. Group terms and rewrite: well want to
try to kill off each of the different frequency components
separately, if thats possible.

119
119

( )
2
3 2 2
3
3 0 3
2 3 2
2 2
3 3 3
0
2 3
3 3
3
3
= 3
4 4 2
3 3
9 s
4 4 2
3
3 sin5 sin7 sin9 .
4 4 4
a A A B
mx kx a x mA a AB t
m
a B a A a A B
mB t
m
a AB a B A B
a AB t t t

sin
in 3

+ +

+ +

+ + +

Keep in mind that we expect that
0
is very small (as is
2 2
0
),
is very small, and B is very small. If we only keep terms which are
first order in small stuff, we can write:
3
a

( ) ( )
2
3 2 2 3
3 0
3
2 2 3
0
3
sin
4
9 sin
4
a A
mx kx a x mA t
m
a A
mB t

3

+

+ +

since Im dropping terms proportional to , etc.
3
a B
2
3
a B

Wed like to be as close to zero as possible. If we
select
3
3
mx kx a x +
2
2 2
3
0
3
4
a A
m
= well kill off the
( )
sin t term just like
before.

120
120

If we choose B so that
( )
2 2 3
0 3
9 mB a A 4 0 + = well kill off the
sin3 t term too. (Note that we are not free to choose since we used
that to tune up the sint term.)

We want B to make
( )
3
2 2 3
0
9 0
4
a A
mB + = :
( )
3 3
3 3
2 2 2
2 2
3 0
0 0
3
3
2 2
0 3
27a 4 9
4 9
4
a
.
32 27
a A a A
B
A m
m
m
A
m a A

= =

=

Weve only been keeping terms that are 1
st
order in small
parameters like and B, so we cant expect our calculation of B to
be any more accurate than 1
st
order in . We can drop the
term in the denominator since theres another, larger term, in the
denominator and the numerator is already 1
st
order in . As a
result,
3
a
3
a
2
3
27a A
3
a
3
3
2
0
a
B
32
A
m

and our improved description of the motion is

121
121


( ) ( ) ( )
3
3
2
0
sin sin 3
32
a A
x t A t t
m

= +
with
( )
2 2
0 3
3 4 a A m = .

Size of the effects induced by nonlinearities
Ever hear the term total harmonic distortion (THD) used to
describe your receivers amplifier? What it means is this: when the
signal going in is a pure sine wave (eg, sin A t ), the signal going out
to your speakers contains harmonics of the input signal (harmonics are
signal components with frequencies that are integral multiples of the
fundamental frequency, eg, 3). The amount of power in the
harmonics, compared to the amount of power in the fundamental, tells
you how much distortion your amplifiers generating.

If our slightly non-linear oscillator were a speaker cone/driver/. . .
mechanism, wed find that it would oscillate in a way that generated
overtones (e.g. the 3 component) when we fed it with a pure sine

122
122

wave intended to create only an
( )
sin A t pressure-vs.-time signal.
The ratio of power into the overtones to power into the fundamental
would go like the squares of the amplitudes of the frequency
components, or

2
3
3
32
l
a A
=
2
2 2
0 3
2
0
Power into 3 harmonic
Power into fundamenta 32
m a A
A m

Louder (bigger A) means more distortion. Note that the ratio of the
distortion power to the signal power increase like A
4
or like the square
of the input power, which is proportional to A
2
.

Well discuss this more when we deal with forced oscillators.

How big are these effect for the simple pendulum? Earlier, we found
that
( )
3
3!
s mg s
F s ms mg

= = +

so that
3
0
3!
s mg s
ms mg

+

.

123
123


Match terms with :
3
3
0 mx kx a x +
mg
k
,
3
3
mg
a
, s x .
0
k g
m
=

A is the maximum value of s so that
max
A = or
max
A = .

2 2
2 2
max 3
0 3
0
3
3 3
8
8 8
mg
a A
mg
k
max

= = =
where
max
is in radians.

2
2 2
3
3 m
2
0
32
32 32
mg
A
a A B
mg
A m
ax
= = =
.

A reminder about initial conditions
F ma =

is a second order differential equation. In effect, when we
solve it, were integrating twice to remove all the derivatives.
Each time you perform an indefinite integral, you get a constant of

124
124


4-29
integration. As a result, to define completely
( )
x t , you need two
pieces of information which tip you off to
( )
0 and
(
0
)
x v
( )
0 ,
. Any
pair of initial conditions from which you can extract
( )
0 x v
will do fine. Besides the obvious
( ) ( ) { }
0 , x v 0 other possible
combinations include
( ) { } ( ) { }
, 0 , 0 , , E U t = x E and so forth. You
can use E =constant =
( ) (
0 U t
)
2 1
2
mv 0 + = , for example.

Phase diagrams
Once
( )
0 x and
(
0
)
x are specified,
( )
x t
( )
is completely determined
for our simple harmonic oscillator.

( )
x t vs. : We can make plots of x t
x
.
A

x
A

125
125


With some thought you can see that the direction of travel is
clockwise in our phase-plane loop.

The maximum value of
( )
sin x A t = + is A; the maximum of
( )
cos is x A t A = + .

Recall that

( ) ( )
2 2 2 2 2 2 2
1 1 1 1
cos sin
2 2 2 2
E mx kx mA t kA t = + = + + +

and that
2
k m = so that

( ) ( )
2 2 2 2 2 2 2 2
1 1
cos sin .
2 2
E mA t mA t mA
1
2
= + + + =

Note also that
( ) ( )
2 2
2 2
2 2 2
sin cos 1.
x x
t t
A A

+ = + + +
=

Because
2 2
2 2
1
x y
a b
+ = is the equation of an ellipse with

126
126


4-31
semi-major axes a and b, our oscillators trajectory in the x x
phase plane is an ellipse.

The variables , , , , , x x y y z z span whats called a six-dimensional
phase space. The shapes of orbits in the phase space are
determined by the x, y, z dependences of the potential. If wed
used a potential with hard walls rather than one that was
quadratic

U

wed have a potential in which our particle moved without
experiencing a force (since F dU dx = ) until it reached x A = .
At the walls, the particle would bounce, reversing its velocity. This
would yield the phase trajectory shown in the following diagram.

x
A -A

127
127


4-32
x
A

Note that the rate at which the object moves around the
trajectory isnt constant along the path: since a bounce off the
walls of the potential happens very quickly, the time spent along
the x A = legs of the path is small.

This phase space business plays a role in more advanced
formulations of mechanics, quantum mechanics, optics, accelerator
physics, and other subjects. We will work with it next semester
when we do a unit on nonlinear dynamics and chaos.

Inhomogeneous elastic deformations: an introduction to tensors
Lets work up a description of a mass attached to four springs that
can move in the x-y plane by sliding on a frictionless table so that
only its x, y components are of interest. Heres a diagram:

128
128


4-33

2
3 1
4

Lets refer to the equilibrium point as 0, 0 x y = = and
displacements from equilibrium as
.

If we move the object up and to the right, well compress springs 1
and 2 while stretching springs 3 and 4. Say each spring has initial
length . If the objects new position is
( )
,
x y
=
, the springs
will look like this:

2
3 1
4

129
129


Change in the length of spring 2:

( ) ( ) ( ) ( )
( ) ( )
( )
2 2
2 2
1
2 2
2
1
1 2
1
y x y x
y y x
y

+ = +

= + +

where Ive used a binomial approximation and only kept terms to
first order in
.

As a result, the change in length is 1
y
y
.

Similarly, the change in the length of: spring 3 is
x
+
spring 4 is
y
+

130
130

spring 1 is
x
.

Strain (displacement) vs. stress for identical springs
If all springs have identical spring constant k, we can push the
mass m to the position
by applying the right amount of force to

stretch or compress each spring by the appropriate amount.
Working to first order in (notice that springs 1 and 3 remain
nearly parallel to y while 2 and 4 remain nearly parallel to x ),
well find
2 2 or 2
x y
F k x k y F k = + =

.

In the case where all the springs have the same spring constant we
find that the displacement
is proportional to, and in the same

direction as, the applied force F
.

We can write a matrix equation to relate F
with
, though it
hardly seems necessary:

131
131


4-36
2 F k =

2 0
0 2
k
F
k

=

.

Lets define . (Thats a lower case Greek sigma).
2 0
0 2
k
k

Strain vs. stress for an inhomogeneous set of springs
Now imagine what would happen if springs 2 and 4 were very stiff
(had large spring constants k
2
) compared to springs 1 and 3 (with
smaller spring constants k
1
). Heres a diagram:

2
k
2
k
2
k
1
k
1
3 1
4

We will work only to first order in the ratios of displacements-to-
spring lengths.

132
132

As before, if we push in the x direction, well effect a
displacement satisfying

1
2k F =
x so that
1
2
F
x
k
=

where . Note that only springs 1 and 3 stretch, to first
order.
F F =
x
y

Pushing instead along will give y
2
2k F =
so
2
2
F
y
k
=
.

The displacements will be unequal for these two directions of
applied force. This isnt at all surprising since springs 2 and 4 are
stiffer than 1 and 3.

But note also that the displacement is still parallel to the applied
force as long as the force is parallel to one of the two principal
axes of the collection of springs.

What will happen if we push the mass at an angle with respect to
the x axis so that cos sin F F x F y = +
?

133
133


What happens is quite simple: the mass will move to a position
where the stretching/compression of the springs counterbalances
the applied force. If we damp its motion, itll eventually come to
rest at this new position.

Lets say that the resulting displacement is
x y
x y = +
. In order
to have the springs forces cancelling our applied force, well need
to have
1
2 cos
x
k x F x = and 2 y
2

y
k sin F y = or
1
cos

2
x
F
k
= and
2
sin

2
y
F
k
= .

Since k
2
is bigger than k
1
,
y
is smaller than
x
.

The mass does not shift in a direction which is parallel to the
applied force unless . Though the force is applied at an
angle , the mass is displaced at an angle with
1
k k =
2

1
2 1 2
sin cos
tan tan .
2 2
y x
k F F
k k k

= = =

134
134

Since and F
are not parallel, it is not possible to write the

relationship between and as F F k =

for k a single numerical
constant. Instead, we can take our two equations

1
2
2
2
x x
y y
F k
F k
=
=

and rewrite them as the matrix equation

1
2
2 0
0 2
x x
y y
F
k
F k

or
F =

where
1
2
2 0
0 2k
k
is sometimes called the elastic stress tensor.

Note that the response to the force is still linear: pushing twice as
hard will effect a displacement that is twice as great. Its just that
the displacement is usually not going to be parallel to the applied
force.

135
135


4-40

rotations
How would the relationship between applied force and
displacement look to someone using coordinate axes that are
rotated by an angle with respect to ours?

Keep in mind that and F
are vectors and, as shown in the

diagram, oriented at angles and with respect to the x axis. The
angle is a positive quantity for the axis rotation shown in the
figure.

x'
x
y
y'
F

The coordinates of forces and displacement, according to observers

136
136

using the , x y axes will be

cos sin
sin cos
x x
y
y
F F
F
F

and
cos sin
sin cos
x x
y
y

.

The matrix with the sines and cosines is a rotation matrix in two
dimensions, of course. In this particular context we use it to
describe how observers in a different frame of reference will
describe the same vector that we are observing.

Lets give it a name: define
( )
cos sin
sin cos
R

.

The inverse of
( )
R
is
( ) ( )
1
R R
=

as you can easily see.
Making use of the fact that
( )
sin sin = and
( )
cos cos = + we have:

137
137


( ) ( )
( ) ( )
( ) ( )
( ) ( )
( ) ( )
2 2
2 2
cos sin cos sin
sin cos sin cos
cos sin sin cos sin cos
sin cos sin cos sin cos
1 0
0 1
R R
I

=

+ +

=

+ +

=

Note that the transpose of is the same as the inverse of .
Recall that the transpose of a matrix is the new matrix obtained by
exchanging the rows and columns of the original matrix. An
element in the i
th
row, j
th
column (R
ij
) in the original matrix appears
in the j
th
row, i
th
column of the transposed matrix:
R
( )
T
ij
ji
R =
R .
We know that and F RF =

R =

where F
and
are the
descriptions of the very same force and displacement vectors that
we see, but written according to their components along the x and
y axes used by observers in the other coordinate system. This has
to be the case since force and displacement are vector quantities
and automatically transform this way under rotations.

We found some pages ago that the relationship between force

138
138

and displacement is linear, and that
F =

where
1
2
2 0
0 2k
k
.

We certainly expect observers in the primed frame to agree with us
that the force-displacement relation is linear, so that doubling the
force always doubles the displacement. We also expect that they
will measure the same angle
( )
between F
and
F
that we
do. It is natural to expect that they ought conclude that =

.
But what are , F
, and
?

F
and
RF

are easy, since both are vectors: we already know that
and F =
R =

. But what about
?

Lets get to it this way: we will start with the known-to-be-true
equation F =

and evolve it in hopes of ending up with a new

equation from which we can read off the identity of
. Its
tempting to think that
might be R

. But its not!

Watch this: start with
F =

139
139

and then multiply both sides of the equation on the left by R

RF R =

.
Since we have F RF =

F R =

.

But since matrix multiplication is not commutative, we cannot
rewrite R

as
( )
R R = =

.

We are free, however, to insert an identity matrix anywhere that it
is convenient in order to write
RF F R I = =

I
.
Because
( )
we can rewrite the previous equation
as
1 T
R R R R
= =

T
F R I R R R = =

.

Since matrix multiplication is associative
( ) ( ) ( )
A BC AB C =

we
can then rewrite the equation as F

140
140

( ) ( )
T T
F R R R R R = =

.

We conclude that the connection between applied force and
resulting displacement, according to observers using the , x y
axes, will necessarily be

1
where
T
F R R R R
= =

.

Not surprisingly, in the other coordinate system the equation shows
that a fractional increase in applied form is matched by the same
fractional increase in displacement.

If we abbreviate
1
2
0
,
0
cos , c and sin s we can

calculate
without getting writers cramp:

141
141

( )
( )
1
2
1 1
2 2
2 2
1 2 2 1
2 2
2 1 1 2
0
0
T
c s c s
R R
s c s
c s c s
s c s c
c s cs
cs s c

= =

+
=

+

Note that a displacement along x does not result from a force
along x . Using F =

with
as expressed above we have

( ) ( )
2 2
1 2 2 1

x x
F c s x cs y

= + +

when
x
x

=
.

Only for
1 2
= is always parallel to F
.

This all works the same way in three dimensions: with springs
along the k, y, and z axes with spring constants k
1
, k
2
, k
3
wed have

142
142

1
2
3
2 0 0
0 2 0
0 0 2
k
F k
k

=

.

If we concoct the 33 rotation matrix that gets us from x, y, z to
, , x y z we will still have

( )
, = , and
T
F RF R F R R = =

so that
T
R R =

.

Any object which transforms under rotations in the same way as

is a tensor of rank two. (A vector is a tensor of rank one, a scalar a
tensor of rank zero.)

Rank-n tensor transformation properties
It isnt immediately clear how to extend the transformation
properties of rank-1 tensors (vectors) and rank-2 tensors to higher
rank: we have A RA =

and . What should we

do to transform a rank-3 tensor? ? ?
1 T
T RTR RTR
= =

T
RRWR

RWR

T T
R

143
143


The pattern is obvious if we write out the matrix multiplications in
terms of sums of products of components, making use of the fact
that
( )
T
ji
ij
R R =

:

3
1
i
p pi i
i
A R A
=
=
=

( )
3 3 3 3
1 1 1 1
3 3
1 1
3 3
1 1 ,
j j i i
T
pq pj ji pj ji qi
iq
j i j i
j i
pj qi ji
j i
j i
pj qi ji pj qi ji
j i i j
T R T R R T
R R T
R R T R R T
= = = =
= = = =
= =
= =
= =
= =

= =

=

= =

Notice the pattern of indices, as shown below.

,
pq pj qi ji
i j
T R R
= T

144
144

H
3
z

Extension to rank-n is straightforward:

, , , , ,
abcdef ai bj ck dl em fn ijklmn
i j k l m n
H R R R R R R =

etc.

The outer product of two vectors is a tensor
Imagine we have two vectors
1 2
a a x a y a = + +
and
. The outer product of
1 2 3
b b x b y b = + +
z b and a

is a 33 object
defined like so:

1 1 1 2 1 3
2 1 2 2 2 3
3 1 3 2 3 3
a b a b a b
a b a b a b a b
a b a b a b

=

If observers using the primed coordinate system form the outer
product from their own versions of the same vectors (e.g.,
Loomis Illini Union), well find that the outer product
transforms like a rank-two tensor. In equations, T a
a b

and
automatically forces . T a
T
T RTR =

145
145


The antisymmetric combination of two outer products also
transforms like a tensor: P a b b a

. Note that the

diagonal elements of P
are zero, since

ij ji
P P = . If you know P
12
,
P
13
, P
23
you can . . .

( ) ( )
( ) (
( ) ( )
2 1 1 2 3 1 1 3
2 1 1 2 3 2 2 3
3 1 1 3 3 2 2 3
0
0
0
a b a b a b a b
)
P a b a b a b a b
a b a b a b a b

=

predict the values for the other components of P
.

This is actually familiar to you. The cross product of two vectors
produces the three linearly independent components of the
antisymmetric combination of two outer products:

( ) ( ) ( )

x y y x y z z y z x x z
L r p r p r p z r p r p x r p r p y = + +

.

These are the (1,2), (2,3), and (3,1) components of r p p r

.
So in a sense is not really a vector, but more properly some
kind of rank-two tensor. This is a fine point we wont need to
L

146
146

3
worry about.

For the time being, well concentrate on one-dimensional
oscillators (or isotropic ones:
1 2
k k k = = ) so that displacement and
applied force are always parallel.

Using complex exponentials to solve the oscillator equation
Back to our simple harmonic oscillator. Wed like to develop the
tools to handle damping and driving forces.

Lets rewrite F =ma (working in one dimension for now) as
0
k
x x
m
+ = .
If we guess that a solution might be
( )
bt
x t ae = and plug in, well
require
2
0
bt bt
k
b ae ae
m
+ = for x to be a valid solution:
2
b k = m
or b i k = m. Note that this solution has reduced our differential
equation to an algebraic one! Very handy trick, that

147
147

Lets define
0
k m = . Note that we have no information about
a yet.

Also note that any function of the form
( )
0 0
i t i t
x t Ae Be

= +
will satisfy 0
k
x x
m
+ = .

Since
( )
( )
0 0
0
i t i t
dx
v t i Ae Be
dt

= =
we can use initial conditions for x(0), v(0) to extract A, B:

( )
0 x A B = + so
( ) ( )
0
Re Re A B x + = while
( ) ( )
Im Im 0 A B + = .

( ) ( )
0
0 v i A B = so
( ) ( )
Re Re 0 A B =
0 0
while
( )
Im Im
( ) ( )
A B v = .

(Keep in mind that Re(1 2) 1 i + = while Im(1 2) 2 i + = , not 2i.)

We have Re Re A B = and
( ) ( )
Im Im A B =
* B A
so it must be the case
that , or, equivalently, * B = A = .

148
148


As a result,
( )
0 0
*
i t i t
x t Ae A e

= +
with
0
2ReA x = and
0 0
2ImA v = .

Also,
0 0
0
2 2
x iv
A
= and
0 0
0
*
2 2
x iv
A
= + .

Note that
( )
x t is always real:
0
0 0
cos sin
i t
e t i
t = + and
0
0
co sin
i t
e t i
0
t s
= so that

( )
0 0 0 0 0
0 0
0 0
0 0 0 0 0
0 0
0 0
0
0 0 0
0
cos sin
cos sin
2 2 2 2
cos sin
cos sin
2 2 2 2
cos sin
0
0
x x i v t v
x t t i t
t
x x i v t v
t i t
v
x t t
t

= + +
+ + +
= +

The same is true for
( )
0 0 0 0
: cos sin v v t v t x t
0
= .

The potential energy associated with the force F kx = is

149
149

( )
2
1
2
U F x dx kx = =
where we take U to be zero when x is zero.

The kinetic energy is
2
1
2
T m = v , of course.

Plugging in for x, v gives
( ) ( ) ( )
2
2 2 2
0 0 0
0 0 0 0 0
0 0
1
cos sin 2 cos sin
2
v x v
U t k x t t t t

= + +

( ) ( )
2
2 2 2
0 0 0 0 0 0 0 0 0 0
1
cos sin 2 cos sin
2
T t m v t x t x v t t

= +

The total energy is

( ) ( )
2
2 2 2 2 2 2 0
0 0 0 0 0
2
0
0 0
0 0 0 0 0
0
1 1 1 1
cos sin
2 2 2 2
cos sin .
E U t T t
v
kx mv t k mx t
kx v
m x v t t
0

= +

= + + +

+

Because
0
k k k m = = mk and
0
m m k m m = = k , the
last term is zero.
Replacing
2
0
by k m in the coefficient of the
2
0
sin t term lets

150
150

us write

2 2 2 2 2 2
0 0 0 0 0
2 2
0 0
1 1 1 1
cos sin
2 2 2 2
1 1
2 2
E kx mv t mv kx
kx mv
0
t

= + + +

= +

since
2 2
cos sin 1 + = .

Not surprisingly, total energy for our oscillator is constant.

Damped oscillations
If theres a velocity-dependent drag on our oscillator well have
F kx ax =
where a is real, and positive.

Lets rewrite this as
2
0
2 0 x x x + + =

where
0
k m and 2 a m .

151
151


As before, try a solution of the form
( )
Bt
x t Ae = . If we really do
have a solution, well find that

2 2
0
2 0
Bt Bt Bt
AB e A e A Be + + = .

Again, we obtain no information about A.

We have ; use the quadratic formula to write
the two solutions for B:
2
0
2 B B + + =
2
0

2 2
0 2 2
0
2 4 4
2
B

= = .

Recall 0 = corresponds to no damping.

We have three qualitatively different cases to consider:
1)
2 2
0
0 <
2)
2 2
0
0 >
3) .
2 2
0
0 =

152
152

Lets do #1 first.

Case 1: underdamped oscillations
2 2
0
0 < so
2 2 2 2
0 0
i =

Define
2
0 d
2
= to write
2 2
0 d
B i = =
so that
( )
( )
d d
i t i t t
x t A e Ae e

= = .
As was true with the undamped case, we can write
( )
d d
i t i t t t
x t Ae e Ce e

= +
and determine A and C from boundary conditions on x and v.

Since
( )
x t is real, at we can conclude A and C must be real
so that Im
0 t =
Im A C = .

Since
( )
0 v is real, we have
( ) ( ) ( )
0
d d
v i A i C = + + with

153
153

the obvious requirement that Re Re A C = . (Otherwise,
( )
0 v
would contain an imaginary part from
d d
i A i C + .)

Combining Im Im A C = and Re Re A C = allows us to conclude
that * A C = and, of course, * A C = .

Since
( )

i t t
*
d d
i t t
x t Ae e
A e e

= +
0 2Re
we find that
( )
x A = , or
( )
Re 0 2 A x = .

In addition,
( ) ( ) ( )
d d
i A*
d d
i t t
v t i Ae e
i t t
e e

= + + so
that
( ) ( )( ) ( )

d
A i A 0 * v A * A + = + .

Recall that * A A + is the same thing as A A 2ReA * , while is the
same thing as 2 so that Im i A
( ) ( )
0 0 v x 2 Im
d
A =
and
( ) ( )
d
0 0 x v
+

Im
2
A

= .

As a result,

154
154

( ) ( ) ( )
0 0
2 2
d
x v x
A i

+
=

0
.

Lets expand the complex exponentials now.

( ) ( )
cos sin , etc. etc.
d
i t
d d
e t i t
= + so that

( ) ( ) ( )
( ) ( )
( ) ( )
0 0 0
0 0 0
0 0
0
cos sin
2 2
cos sin
2 2
cos sin
t
d d
d
t
d d
d
t t
d d
d
x v x
x t i e t i
x v x
i e t i
v x
x e t e t

+
= + t
t

+
+ +

+
= +

The motion grows smaller with time due to the
t
e

factors, but
oscillates with a (slightly) shifted frequency
2 2
0 d
= .

This is called underdamping.

Oscillator Q (Quality factor)

155
155


A comment about energy loss in the underdamped oscillator: we
can define a quality factor Q which tells us how well the
oscillator holds onto its energy. There are lots of nearly-
equivalent ways to define Q; one commonly used definition is
meant to look like the rate of energy loss per radian of advance in
the arguments of the oscillators sines and cosines. In effect, you
look to see what fraction of the energy is lost in one cycle, divide
this into the stored energy, and perhaps throw in a factor of 2 to
convert cycles into radians.

Lets say we watch the oscillator from time t to time t+dt.

At t, its energy is
( ) ( )
2 2 1 1
2 2
k x t m v t +

.

At t+dt its energy is
( ) ( )
2 2 1 1
2 2
k x t dt m v t dt + + +

.

156
156

( ) ( ) ( ) ( )
[ ] [ ]
2 2 2
2 2
1 1
2 2
1 1
2 2
.
t t
E k x t dt x t m v t dt v t
d x d v
k dt mdt
dt dt
kxv mva dt kx ma vdt

= + + +

+
= + = +
2
v

We know 2 F ma kx m = = so 2 ma kx mv + = .

This lets us rewrite
2
2 4 E mv dt Tdt = = for
2
1
2
T m v .

Not surprisingly, is time dependent. Times during the
oscillators cycle when it has larger velocity (and therefore smaller
potential energy) are the times during which the damping force
acts most strongly.
E

Since
1
2
T E , the average fractional energy loss in time dt is
4
2
2
E Tdt
dt
E T

= .

The time required for the cosines (or sines) argument to change

157
157

by one radian is given by 1
d
t = , so 1
d
t = .

In time dt the argument advances by d radians for
d
d dt = .
Consequently, we can write

2
2 .
d
E
dt d
E
=

We define the quality factor Q so that E E d Q = for a time
interval dt which corresponds to an advance of d radians in the
oscillators cycle. We have, then
2
d
Q

=

Big Q: small energy loss per cycle, so the oscillator is weakly
damped.

Small Q: the oscillator loses energy quickly (thus low quality).

People are generally interested in building oscillators with as small
a rate of energy loss as possible, thus the definition of Q so that big
Q (which is desirable) corresponds to small damping.

158
158


Case 2: overdamped oscillator
For the square root in
2 2
0
0 >
2
0
B
2
= is real. We
have
( )
( ) ( )
2 2 2 2
0 0
t t
x t Ae Be
+
= +
and
( )
( )
( )
( )
( )
2 2 2 2
0 0
2 2 2 2
0 0
t t
v t Ae Be

+
= + +

( )
0 x A B = + , so Im Im A B = to keep x real.
( ) ( ) ( ) ( )
2 2 2 2
0 0 0
0 v A B A B x A = + + = + B

We already know Im Im A B = . To keep the second term in our
expression real, well also need Im
( )
0 v Im A B = + , so that
Im 0, Im 0 A B = = .

A little algebra reveals
0 0
2 2
0

v x
A B

+
=
so that

159
159


( )
0 0
0
2 2
0
2
2
v x
x
A

+
= +
and
( )
0 0
0
2 2
0

2
2
v x
x
B

+
=
.

than once through x =0, and in fact does not oscillate at all: if

Note that an overdamped oscillator (this case) never passes more
2 2 2 2
0 0
0
t t
t t
Ae e Be e

+ =
we have

2 2
0
2
e
t
B
A

=
at the on t

e time t tha
( )
0 x t = . No other values of time can give
after this.
ow lets consider the critically damped oscillator.
lution
( )
0 x t =

Case 3: critically damped oscillator
N

We try a so
( )
Bt
x t Ae = and find
2 2
0
B = .
However,
2 2
0
= , so it seems like theres only a single linearly

160
160

e second order differential equation
ther than two solutions.
What I mean b
independent solution to th
ra

y linearly independent is this: two functions
( ) ( )
and f t g
)
t
t ar
= to be satisfied for all t is for a = b = 0. (a, b are
onstants.)
Examples of linearly in endent functions:
e linearly independent if the only way for
( ) (
af t bg + 0
c

dep
( ) ( )
2
, , cos , sin t t t t
d.e.
pendent solutions. For the
under/over damped oscillator,

A basic property of differential equations is that an n
th
order
must always have n linearly inde
( )
2 2
0
~ x t Ae

and
(
t

+
)
2 2
0
~
t
Be

x t

re two linearly independent solutions. a

When
2 2
0
0 = , these two solutions collapse to the single
function
t
e

. As a result, theres another solution (which we
avent found yet) lurking somewhere. h

161
161

for an under d
over a second solution to our critically damped
scillator equation.
The under damped oscillator e n had two solut ns:
Investigate this: consider the difference between the two solutions
amped oscillator. Well look at what happens when
2 2
0
0 to unc
o

quatio io
( )
1
d
i t t
x t e e

= and
( )
2
d
i t t
x t e e

= .
Because they are both solution
2
0
2 0 x x x + + = s to we have
and
2
1 1 0 1
2 0 x x x + + =
2
2 1 0 2
2 0 x x x + + = .
w ns after multiplying the first by
and the second by to write
0.
Now define . Obviously,

Lets combine the t o equatio
1
a
2
a
2 2
1 1 2 2 1 1 2 2 0 1 1 0 2 2
2 2 a x a x a x a x a x a x + + + + + =

( )
1 1 2 2
u t a x a x = +
1 1 2 2
u a x a x = + and
1 1
a x = +
2 2
a x . u

We can rewrite that
1 1 2 2
... a x a x + equ
2
0
u u ation as 2 0 u + + = : as
long as x
1
and x
2
are solutions, then
1 1 2 2
u a x a x
= + is automatically

162
162

This is an important property of linear differential
quations.)
With the specific values
a solution. (
e

1 2
1 1
,
2 2
d d
a a
i i
= = we have

( )
( )
sin
2
d d
i t i t t
d t
d d
t
e e e
u t e
i

= =

.
nds to the case
underdamped oscillator, in the limit that

Our critically damped oscillator correspo of an
2 2
0
0 (or
d
0 , by definition of
d
).
the limit that
( ) ( )
1 2
0, , .
t t
d
x t e x t e

In However,

( )
( )
0 0
sin
lim lim
d d
d t t
d
t
u t e te

= =

giving us
( )
sin
d
d
d d
t
t
t

= since
( )
sin as 0 .

163
163

his is our second solution.
Crank through the use of initial conditions now. Lets say

T

( )
t t
x t Ae Bte

= + .
we have At 0 t =
( )
0 x A = so
0
A x = .

t
( )
t t
v t Ae Be Bte

= +
So
x
( )
0
0 v B A B = =
or
0 0
B v x = +
yielding
( ) ( )
0 0 0
t
x t x v x t e

= + +

Critical damping corresponds to the oscillator set up which allows
the fastest return (without overshoot) to equilibrium. This is good
r shock absorbers, and so forth. fo

164
164


4-69

Electrical analogues
Any system which obeys an equation of the form 0 x kx ax + + =
will exhibit

oscillatory behavior: x doesnt have to represent
osition.
Heres an electrical analogue:
l to the
amount of change on the capacitor, among other things:

p

L
C
I

The sum of the voltage drops around the circuit must be zero.
Recall that the voltage across the capacitor is proportiona
0
.
Q
V Ed d
A
= =

165
165


4-70
ductors respond to changes in current since changes in the
magnetic field inside the inductor will induce voltages around the
loops in the inductor that op se th hanges in current.

he voltage across the inductor is
++++
_ _ _ _

+Q
Q
In
po e c

T , LdI dt negative since the

The net voltage drop around the loop is zero so the voltage across
e capacitor and inductor must be the same:
voltage tends to induce a current to counter the change in current
experienced by the inductor.
th
Q dI
L
C dt
= or 0.
Q dI
L
C dt
+ =

The current flowing in the loop must be constant everywhere.
Keep in mind that the positive charge arriving on one capacitor
L I

166
166


4-71
citor
rge. (Theres just as much positive charge on
ne plate as negative charge on the other.) As a result, the current
equals the rate at which positive charge flows off one of the
capacitors plates:
plate expels an equal amount from the other plate so the capa
contains no net cha
o

.
dQ
I
dt
=
We have:
2
2
0
Q dI Q d dQ Q d Q
L L L
C dt C dt dt C dt

+ = + = + =

or
1
0 Q Q +
LC
=
.
As a result,
( ) ( )
t
0
sin Q Q = + with 1 LC = .

We could install a resistor in series:

The voltage across the resistor is V =IR so our sum-of-voltages-
I

L
R
C

167
167

und-the-loop equation becomes aro
2
2
0
Q LdI Q Ld Q dQ
IR R
C dt C dt dt
+ + = + + =
r o
1
0
R
Q Q Q
L LC
+ +

ith some pages back:
= .

2
0
2 0 x x x + + = Compare this w
2
R
L
and
2
0
1
LC

so well have
( )
2
1
2 2
~
R R
t
L L LC
Q t e

.

By using initial conditions on charge and current, we can select the
exact form of the solution, solving for A and B.

( )
( )
( ) ( ) ( ) ( ) ( ) ( )
2
2 1 R L LC t
Ae Be

2
2 1
2
R L LC t
Rt L
Q t e

= +

168
168

Driven oscillations
e can apply a driving force to our oscillator so that W

( )
F mx kx ax f t = = +

( )
f t for an arbitrary driving force.

ets start by considering the effect of a sinusoidal driving force: L

( ) ( )
cos
0
f t F t = .

Note that well want to be able to choose an arbitrary frequency so
at (often) th
0
. For convenience, Ill define t = 0 to be when
sitive. (Well only deal with an
under damped oscillator for the time being.)
the applied force is greatest, and po

Our equation of motion is
( )
0
cos
F a k
x x x t
m m m
+ + =

This is a second order, inhomogeneous differential equation. By

169
169

homogeneous I mean that whats left over (on the right side) in
after we collect all the
( )
2 2
, , x t dx dt d x dt terms is different
from zero. (In our case, its
( ) ( )
cos F m t
0
.)

Since the equation is second order, well have to do something that
these
hen a particular solution to the inhomogeneous equation. The
ons of the homogenous
solutions and the particular solution will be what we want.
Heres the homogeneous equation:
amounts to integrating it twice to solve it: there will be two
(initially undetermined) constants of integration that arise.

A basic property of n
th
-order differential equations is that there will
always exist n linearly independent solutions. Well generate
by finding two independent solutions to the homogeneous equation
and t
sum of appropriate linear combinati

0 x x x
m m
a k
+ + = .
Lets refer
( )
1
x t and
( )
2
x t to its linearly independent solutions as .

solution x
p
to the inho
equation
If we can find a particular mogeneous

170
170

( )
0
cos
F a k
p p p
x x x t
m m m
+ + =
then anything of the form
( )
1 1 2 2 p
x t A x A x x = + + will also satisfy
the inhomogeneous equation. This is easy to see: write the three
equations

( )
1 1 1
2 2 2
0
a k
x x x
0
0
cos
p p p
a k
x x x
m m
m m
F a k
x x x t
m m m

+ + =
+ + =

and add them up, after multiplying the first two by the constants
A
1
, A
2
:

+ + =
( ) ( )
( ) ( )
2
1 1 2 2
0
cos .
p p
p
a d
x A x A x x
dt m dt
F
1 1 2 2
2
d
A x A x + +
1 1 2 2
k
A x A x x t
m m

+ + +
+ + + =

( )
As a result any function in the form
1 1 2 2 p
x t A x A x x = + + will
work fine since it satisfies the equation
( )
0
cos mx ax kx F t + + = .

171
171

y choosing two different sets of
{ }
1 2
, A A B values we can concoct
two different linearly our inhomogeneous
quation.

on to
independent solutions to
e

Particular soluti the inhomogeneous equation
We know how to generate
1
x and
2
x see previous material some
ages earlier in the notes. e he functions which damp p (Th se are t
away to zero as t .) If we can find/invent/stumble over a
particular solution
p
x , were done.

Heres a reasonable guess: the oscillator will oscillate with the
ame period as the driving force, though we may need to wait for
y-damped homogeneous equation solutions
s
the exponentiall
1
x and
2
x to die away.
Lets try

( )
i t i t
p
x t D e D e
+
+
= + and rewrite
( )
0
cos
F
t
m
in terms
of complex exponentials. Our equation for the particular solution
is

172
172

( )
0
2
i t i t
p p p
F a k
x x x + + e e
m m m

= +

or, inserting our trial solution,

( )
2
2
0
.
2
i t i t i t
i t i t i t i t i t
a k
D e i D e D e
m m
F a k
D e i D e D e e e
m

m m

+ + +
+ + +

+ +
+ = +

Grind away for a bit:

2 2
0 0

2
i t i t
F F
e e
m
+
2
i t i t
ia D kD ia D kD
D e D e
m m m m
m

+ +
+

+ + +

= +

Keep in mind that
i t
e
i t
e

is linearly independent of
+
: the only
way we can have 0
i t i t
Ae Be

+ = for all t is to have A=B=0.

We can rewrite the above equation slightly as

173
173


2 0
2
0
2

2
i t
i t
F ia D kD
D e
m m m
F ia D kD
D e
m m m
+ +
+

+ + +

+

=

which tells us that

2
0
+
2
F ia D kD
D
m m m
0
+ +
+
+ =

and also that

2
0
0.
2
F ia D kD
D
m m m
+ =
We know
2
, a, m, k, and F
0
. As a result, we can immediately
solve for . Recall and
+
D D
0
k m so we can rewrite the
above equations as

2 2
0
0
2
F iaw
D
m m

+

+ + =

and
2 2
0
0
2
F ia
D
m m

+ =

.

As a result,

( )
0
2 2
0

2 2
F
D
m ia
+
=
+
and
( )
0
2 2
0

2 2
F
D
m ia
=

174
174


or (using a common denominator)

( )
( )
( )
2 2
0 0
2
2 2 2 2 2
0
2
F m ia
D
m a

+

=
+

( )
( )
( )
2 2
0 0
2
2 2 2 2 2
0
2
F m ia
D
m a

+

=
+

Note that we have no freedom to adjust anything: D
+
and are
completely determined. Also note that
D
*
D D
+
= .

When
2 2
0
a , D
+
and D
-
are nearly real. When
2 2
0
a , D
+
and D
-
are nearly imaginary. (Keep in mind
that a is the coefficient of the damping term, not the acceleration.
Small a means small damping.)

This will be handy: lets figure out what the complex conjugate of
i t
De
looks like.

( ) (
Re Im D D i = +
)
D and
( ) ( )
cos sin
i t
e t i
t = + so

175
175

( ) ( )
( ) ( )
Re cos Im sin
Re sin Im cos .
i t
De D t D t
i D t D t

+
=

+ +

similarly,

( ) ( )
( ) ( )
*
Re cos Im sin
Re sin Im cos .
i t
D e D t D t
i D t D t
=

+

Notice the sign-switch on the imaginary part of
* i t
D e

relative to
that of .
i t
De
+
As a result,
* i t
D e

is the complex conjugate of
i t
De
+
, and vice versa:

( )
*
* i t i t
De D e

= .

A few pages back we had
( )
i t i t
p
x t D e D e

+
= +

Lets rename We found . D
+
D
*
D D
+
= , so we have
and can write
*
D D

( ) ( )
*
*
.
i t i t i t i t
p
x t De D e De De

= + = +

176
176


Something added to its complex conjugate gives a real quantity,
just as wed expect for
( )
p
x t .

Lets rewrite
( )
p
x t in terms of real functions now.

( ) ( ) ( )
( ) ( )
( ) ( )
( ) ( )
( ) ( ) (
*
Re cos Re sin
Im cos Im sin
Re cos Re sin
Im cos Im sin
2 Re cos 2 Im sin .
i t i t
p
De De D t i D t
i D t D t
D t i D t
i D t D t
D t D t x

+ = +
+
+

= =
)
t

From a few pages ago,

( ) ( )
( )
( )
2 2
0 0
2
2 2 2 2
0
2Re 2Re
F m
D D
m a

2

+

= =
+

and
( )
( )
0
2
2 2 2 2
0
2Im
F a
D
m a
2
=
+

so that

177
177

( )
( )
( )
2 2
0
0 2
2 2 2 2 2
0
cos sin
p
F
x t m
m a
t a t

= +

+
.

The response of the oscillator is out of phase with the
( )
cos t
driving force: take note of the sin a t term in
( )
p
x t in addition
to the cosine term.

Useful trig identity:
( )
cos cos cos sin sin t t t = + .

With some algebraic cleverness, we can apply this to
( )
p
x t .

Lets look at the coefficients of cos and sin t t in the
( )
p
x t
expression:

( )
( )
( )
( )
( )
2 2
0
0
2 2
2 2 2 2 2 2 2 2 2 2
0 0
2
2 2 2 2 2
0
cos
sin
p
m
F
x t t
m a m a
a
t
m a

=

+ +
+

+

178
178

a Ive split the
( )
2
2 2 2 2 2
0
m
denominator on the previous

page into a pair of square roots.

Notice that the sum of the squares of the terms in front of the
cos t and sin t is 1:
( )
( ) ( )
( )
( ) ( )
( )
( )
2 2
2 2
0
2 2
2 2 2 2 2 2 2 2 2 2
0 0
2
2 2 2
2 2
0
2 2
2 2 2 2 2 2 2 2 2 2
0 0
2
2 2 2 2 2
0
2
2 2 2 2 2
0
1.
m
a
m a m a
m
a
m a m a
m a
m a

+

+ +

= +
+ +
+
= =
+

This is just like
2 2
sin cos 1 + = ! Lets define so that

( )
( )
2 2
0
2
2 2 2 2
0
cos
m
m a

=
+

and
( )
2
2 2 2 2
0
sin
a
m a
2

=
+

179
179

and
( )
( )
2 2
0
tan .
a
m

This way we can rewrite the equation for
( )
p
x t as
( )
( )
[ ]
( )
( )
0
2
2 2 2 2 2
0
0
2
2 2 2 2 2
0
cos cos sin sin
cos

p
F
x t t
m a
F
t
m a
t

= +
+
=
+
t

where I made use of the trig identity
( )
cos cos cos sin sin t t = + .

Reverting to the definition of 2 a m = we can rewrite
( )
p
x t as

( )
( )
( )
( )
( )
0
0
2 2
2 2 2 2 2 2 2 2 2 2
0 0
cos cos
4 4
p
F t t
F
x t
m
m m

= =
+ +

where
( )
2 2
0
2
tan
.

180
180


So thats our solution! If we plug
( )
p
x t into the equation of
motion some pages back, well find it really does work as a
solution.

p

There are a number of features to notice about
( )
p
x t .
Amplitude vs. especially near
0
.
Phase shift between driving force and response
( )
p
x t
Effect of damping
( )
on frequency response
Very high, very low frequency limits, etc. etc.

Lets discuss these qualitatively, first, without resorting to
equations.

Amplitude vs. driving frequency
Ever push a child on a playground swing? To get the largest
amplitude oscillations, you push the swing once per cycle, at the

181
181

time the swing reaches its maximum angle with respect to the
bottom of its arc.

Youre not driving the swing with a sinusoidal force, but
qualitatively it makes sense: if you push the swing at its natural
frequency, youre always adding energy. If you push it at a
different frequency sometimes your push adds energy but
sometimes your push acts instead to slow the swing down.

We can be more quantitative by graphing the amplitude vs. :
below is a plot of the coefficient of
( )
cos t in our x
p

expression:

182
182


Since

( )
( )
( )
0
2
2 2 2
0
cos
4
p
t
F
x t
m 2

=
+

183
183

the maximum amplitude occurs when the
( )
2
2 2 2
0
4
2
+
2

denominator is a minimum. Note that this happens when the term
inside the square root is a minimum: this saves a little of the work
in the differentiation.

The frequency (call it
max
) ought to be close to
0
.

Differentiate
( )
2
2 2 2
0
4 + to find where it is a minimum:
( )
( )( )
2
2 2 2 2
0
2 2 2
0
4
2 2
d
d

8

+

= + .

( )( )
2 2 2 2 2
0 max max max
2 2 8 + 0 =
so
( )
2 2
0 max
2
2
=
or
2 2
max 0
2
2
=
and
2 2
max 0
2 = .

The frequency which corresponds to the maximum

184
184

amplitude is shifted slightly from the natural frequency
0

because of the damping.

The maximum amplitude is
( ) ( ) ( )
( ) ( )
0 0
2 2
2 2 2 2 2 2 2 2 2
0 max max 0 0 0
0
2 2 2 2 2
0
0
2 2
0
4 2 4
4 4 2
.
2
F m F m
F m
F m
2
2

=
+ + +
=
+
=

The oscillator stores energy pumped into it by the driving force,
only losing it through the effects of the (non-conservative)
damping force. Note that the amplitude blows up as 0 :
without a mechanism for energy loss, the oscillator goes bonkers.

Recall that we defined
2 2 2
0 0 d
2
= for the under-
damped oscillator. We can rewrite our maximum amplitude as
0
2
d
F
m
using this.

185
185

The frequency
1/2
at which the amplitude drops to half its
maximum is given by
( )
2
2 2 2 2
0 1/2 1/2
2 2
0
1
4
1
1
2
2

+
=

or
( )
( )
2 2 2
0
2
2 2 2 2
0 1/2 1/2
4
1
4
4

=
+

Solve for
2
1/2
to find
2 2
1/2 max
2 3 .
d
=

If is small (light damping, high Q),
( )
2
1/2 max max
2 3 1 3
d
=
since
max 0 d
for light damping.

As a result, the amplitude vs. frequency curve has a full-width-at-
half-maximum of
max
FWHM 2 3 and a fractional width of

186
186

max
FWHM
2 3 .

Smaller damping gives a narrower resonance peak.

Phase shift between driving force and response x
p

Now for the phase . From some pages earlier we had
( )
( )
2 2
0
tan
a
m
.
Our driving force was
( )
0
cos F t and the particular solution
p
x
was proportional to cos
( )
t .

is a phase lag: when our force goes through its positive
maximum the position wont go through its positive max until
0 or t t = = .

If we drive the oscillator at low frequency
0
, well have
( )
2
0
tan 1
a
m

so

187
187

2
0
a
m
.

In the low frequency case, the oscillator keeps up with the force:
you push it in the positive direction, and it tends to move so that
most of your force just goes towards keeping the spring stretched.

In the high frequency case,
0
and
( )
2
tan .
a a
m m

=

Since
( )
tan sin cos = , well either need sin to be small and
negative (with cos positive) or sin to be small and positive
with cos negative. The latter turns out to be what happens: we
end up with close to 180 ( radians). In this case, as the
oscillator passes through its positive maximum, the force passes
through its negative maximum.

Near resonance,
0
and tan 1 .

Large
( )
tan corresponds to phase angles near 2 for which
( )
0 cos .

188
188


2
= corresponds to pushing hardest (in the positive direction)

when the oscillator goes through equilibrium in the positive
direction, i.e. when its velocity is a (positive) maximum. Note that
this phase gives the largest value of F v

, the power expended by
the driving force.

Heres a plot of vs. driving frequency :

189
189


Keep in mind that weve just been talking about the particular
solution x
p
. The general solution includes contributions from the
homogeneous equations solutions x
1
(t) and x
2
(t). Since x
1
and x
1

damp away to zero, theyre sometimes referred to as transients.

You need initial conditions on x
0
and v
0
(or equivalent information

190
190

on kinetic and potential energy [even though we dont really have
P.E. for a damped system!!] to determine how much of the
solutions x
1
and x
2
to include.

We can concoct an electrical analogue of the driven oscillator by
taking that circuit some pages back and adding a signal generator,
which maintains a voltage
( ) ( )
0
cos v t v t = between its output
terminals:
I

The sum of voltages around the circuit must obey this equation:
0
cos
Q dI
IR L v
C dt
t + + =
or
0
1
cos
v RQ
Q Q
L LC L
t + + =
.

L
R
C
~
0
cos v t

191
191

t
Non-sinusoidal driving forces
A forced oscillator, driven by a sinusoidal driving force, can be
described by the equation

0
cos F mx kx ax F = = +

After a lot of algebraic grunge, we found we could write
( )
x t as

( ) ( )
( )
( )
0
1 1 2 2
2
2 2 2
0
cos
4
t
F
A x t A x t
m 2

+ +
+

After the transients
1
x ,
2
x (which satisfy the homogeneous
equation, corresponding to
0
0 F = ) die out our oscillator moves
with a sinusoidal displacement vs. time, though with a phase shift
relative to the driving force. This is useful for speaker design, for
example: as long as the restoring force on our speaker cone is a
(damped) spring force, it will always move in a way which
produces a sinusoidal displacement when we drive it with a
sinusoidal force.

If our system is nonlinear, the motion is much more complicated:

192
192

the natural frequency varies with amplitude. As our oscillators
motion changes in amplitude, corresponding to gains/losses of
energy associated with the effects of the driving force, the rate at
which it gains/loses energy can vary dramatically.

We wont deal with the driven nonlinear oscillator at this time.

What happens if we drive our (linear) oscillator with a non-
sinusoidal driving force?

Start with an easy case: a two-frequency driving force.

Let
( ) ( ) ( )
1 1 2
be cos cos F t F t F t
2
+ .

The equation of motion is

( ) ( )
1 1 2
cos cos mx kx ax F t F t
2
+ + = + .

This is linear: the left side does not contain terms which are
products of two or more xs or derivatives. There are no terms like
( )
2
2
or or . xx x x

193
193

We know how to solve the case when
2 1
0 and/or 0. F F = =

Lets say
( )
1
x t
mx
is the solution (including particular and transient
pieces) to
1
cos kx ax F t
1
+ + = and
( )
2
x t is the solution to
2 2
cos F t mx kx ax + + = .

(Note that
( )
1
x t and
( )
2
x t no longer stand for solutions to the
homogeneous equation.)

We have:

( )
2
1 0 1 1 1 1
cos
a
x x x F
m
t + + =
(
2
2 0 2 2 2 2
cos
a
)
x x x F
m
t + + = since
2
0
k m = .

Add these two equations to find

( )
( )
( )
2
1 2 1 2 2
0 1 2 1 1 2 2
2
cos cos .
d x x d x x
a
x x F t
dt m dt
F t
+ +
+ + + = +

In short,
1 2
x x x = + is a solution to our inhomogeneous equation

194
194

( )
2
0
a
x x x F
m
+ + = t when
( )
1 1 2
cos cos F t F t F t
2
= + .

Note that this would also work for a three-component force.

In general, if we have the solutions to all the necessary equations
( )
( )
( )
2
1 0 1 1 1
2
2 0 2 2 2
2
0 n n n n
a
x x x F
m
a
t
x x x F
m
a
t
x x x F
m
+ + =
+ + =
+ + =

t

then
( ) ( ) ( ) ( )
1 2 n
x t x t x t x t + + + will automatically satisfy
( )
2
0
a
x x x F
m
+ + = t
as long as
( ) ( ) ( ) ( )
1 2 n
F t F t F t F t = + + + .

The trick, when dealing with a non-sinusoidal function (like a
square wave, or
1 1 2 2
sin sin , or a t a t +
( )
F t
) is to come up with a
way of writing as a sum of (co)sinusoidal forces, since we
know how to solve the equations of motion for a sinusoidal driving
force.

195
195


Fourier series
Fouriers Theorem (crudely stated): any periodic function
can be represented as a sum of sines and cosines with frequencies
that are integer multiples of
( )
F t
2 T = where T is the period of
.
( )
F t

For the time being, well restrict ourselves to periodic functions in
order to explore this. How do we cook up a technique to represent
as a sum of sines and cosines?
( )
F t

Lets assume that has period T, and well work with
representing it by studying it during the interval
( )
F t
2 2 T t T + .
The frequency associated with period T is 2 T = , so we can
express the time range as t + .

Well want to build up
( )
F t as a sum of terms like this:

196
196

( ) ( ) ( )
0
cos sin .
n n
n
F t a n t b n t
=
= +

Keep in mind that has n periods in the interval
(
cos nwt
)
2 2 T t T + , as does
( )
n t sin .
( )
0
cos 0 a t is just a constant
since ;
( )
cos 0 1 =
(
0
sin 0 b
)
t is zero and can be neglected.

In general, well know the form of
( )
F t in any particular problem.
(For example, perhaps its a square wave). Wed like to work out
the values of all the coefficients. ,
n
a b
n

What were going to do is to develop a sort of mathematical filter
that we can bump up against
( )
F t , thatll kill off everything
except one term in the sum while passing that term undisturbed.

Schematically:

197
197


4-102
( ) ( ) ( )
0
cos sin
n n
n
F t a n t b n t
=
= +

Orthogonality relationships
We need to do some preliminary work in order to develop our
filter. Lets evaluate
( ) ( )
2
2
cos cos
T
nm
T
I n t m t dt
+

.
Here, 0, 0, and 0 n m n m +

Use complex exponentials for the cosines to rewrite this:

k
n k
n = k
a
k

198
198

( ) ( ) ( ) ( )
( ) ( )
2
2
2
2
2 2
2 2
2 2
1
4
1 1
cos cos .
2 2
T
inwt inwt imwt imwt
nm
T
T
i n m wt i n m wt i n m wt i n m wt
T
T T
T T
e e e e
I d
e e e e d
n m t dt n m t dt
+

+
+ +
+ +

+ +
=

= + + +

= + +

t
t

Look at the two integrals separately:
( )
( )
( ) ( )
( )
( )
( )
( )
( )
[ ]
2 2
2 2
2
2
1 1
cos cos
2 2
1
sin
2
1
sin
2
1
0 0
2
0.
T T
T T
T
T
n m t dt n m t d n m t
n m
n m t
n m
n m t
n m
n m

+ +

+
+ = + +

+
= +

+
= +

+
=
+
=

For the case that the second integral works out the same
way:
n m

199
199

( )
( )
( ) ( )
( )
( )
2 2
2 2
1 1
cos cos
2 2
1
sin
2
0.
T T
T T
n m t dt n m t d n m t
n m
n m t
n m

+ +

+
=

When n = m we have this for the second integral:

( )
2 2
2 2
1 1
cos 1 .
2 2
T T
T T
T
n m t dt dt
+ +

2
= =

As a result,
( ) ( ) ( )
2
2
cos cos 0 .
2
T
nm nm
T
T
I n t m t dt n m
+
= =
+

We say that the functions
( )
cos n t and
( )
cos m t are orthogonal
when we use an integral over the specified interval as our tool for
performing an inner product between two functions.

For the case we have 0 m n = =
( ) ( )
2
00
2
cos 0 cos 0 .
T
T
I t t dt T
+
= =

200
200

Keep in mind that both of the cosines in I
mn
have an integer
number of periods occurring during the time interval
2 2 T t T + .

It is easy to show, in similar fashion, that

( ) ( ) ( )
2
2
sin sin 0, 0, 0
2
T
nm
T
T
n t m t dt n m n m
+
= +

as well as
( ) ( ) ( )
2
2
cos sin 0 all , .
T
T
n t m t dt n m
+

The orthogonality of functions in the set
( ) ( ) { }
cos ,sin n t m t
( )
sin n t

provides us with a useful tool to figure out all the coefficients in
the expansion
( ) ( )
0
cos
n n
n
F t a n t b
=
= +
.

We can extract the value for a particular a
k
by evaluating the
integral
( ) ( )
2
2
2
cos .
T
k
T
I k t F t dt a
T

+
=

= (Keep in mind that we
know the functional form for F(t).)

201
201

Heres why this works. First consider the case k >0.

( ) ( )
( ) ( ) ( )
2
2
2
0
2
2
cos
2
cos cos sin
T
T
T
n n
n
T
k t F t dt
T
k t a n t b n t dt
T
+

=
= +

( ) ( )
( ) ( )
[ ]
2
0
2
2
0
2
0 0
2
cos cos
2
cos sin
2
0
2
.
T
n
n
T
T
n
n
T
n kn
n n
k
a k t n t dt
T
b k t n t d
T
T
a
T
a

=

+
=

= =

=

+

= +

=

t

For k =0 we have instead

202
202


( ) ( )
( ) ( ) ( )
2
2
2
0
2
2
0 0
2
1
cos 0
1
cos 0 cos sin
1
.
T
T
T
n n
n
T
T
T
F t dt
T
a n t b n t
T
a dt a
T

+
+

=
= + dt

= =

The act of multiplying by
( )
2cos k t
T
and then integrating kills off

all but one of the terms in the sums on the right side. As long as
we are able to perform the integral
( )
( )
2
2
2cos
T
T
k t
F t dt
T
, we can
extract the a
k
values one-by-one.

Similarly, we can multiply by
( )
2sin k t
T
and integrate to extract

b
k
.

Lets work some examples. (Keep in mind that 2 T .)

203
203


4-108
Square wave
We have
( )
0
0
3
0 ,
2 2
3
, 2
2 2
T T
F t T t
F t
T T
F t T t
+ < <
=

< <
,
, T

so that looks like this:
( )
F t

+F
0
t
-F
0

T

Lets build up as a sum of cosines and sines. The
fundamental frequency is
( )
F t
2 T = . We want to evaluate all the
Fourier coefficients and in the expansion
k
a
k
b
( ) ( ) ( )
0
cos sin
n n
n
F t a n t b n t
=
= +
.

Extract (k > 0):
k
a

204
204

( )
( )
( ) ( )
( ) ( ) ( ) ( )
( ) ( )
2
2
2 0
0 0
2 0
2 0
0
2 0
0 2
0
2 0
2cos
2cos 2cos
2
cos cos
2
sin sin .
T
k
T
T
T
T
T
T
T
k t
a F t dt
T
k t k t
F dt F dt
T T
F
k t d k t k t d k t
k T
F
k t k t
k T
=
= + +

= +

= +

Since 2 T = we have

( ) ( )
0
0
0
0
2
sin sin
2
2sin 0.
k
F
a k t k
k T
F
k
k T
t

= +

= =

Its also easy to show that

( )
2
0
2
0.
T
T
F t
a d
T
= =
t

205
205

Now extract :
k
b

( )
( )
( ) ( )
( ) ( ) ( ) ( )
( ) ( )
2
2
2 0
0 0
2 0
0
0
0
0
0
0
2sin
2sin 2sin
2
sin sin
2
cos cos
T
k
T
T
T
k t
b F t dt
T
k t k t
F dt F dt
T T
F
k t d k t k t d k t
k T
F
k t k t
k T

=
= + +

= +

=

so (since 2 T = )

( ) ( )
( ) ( )
0
0
0
2
1 cos cos 1
2
1 1 1 1
4
is odd
0 is even.
k
k k
F
b k
k
F
k
F
k
k
k
k
=

= +

+

0 0
1 2 3
4 4
0 0
3
F F
b b b
4
b

= = = =

0 0
5 6 7
4 4
0 0
5 7
i
F F
b b b b

= = = =

206
206


and our square wave is

( ) ( ) ( )
0
odd
4
sin 2 .
k
F
F t k t T
k

= =

Lets use this to write a solution to the equation of motion for our
damped oscillator when we drive it with a square wave:

( )
( )
2
0
0
odd
4 1
2 s
k
F t
F
in x x x k
m m k
t
+ + = =

.
The particular solution for one frequency component (lets forget
about transients for now) will satisfy the equation

( )
2 0
0
4 1
2 s
k k k
F
in x x x k
m k
t
+ + = .

From earlier work (but driving the oscillator with a sine instead of
a cosine) we know

( )
( )
( )
( )
( )
0
2
2 2
2 2
0
sin
4
4
k
k
k t
F
x t
m k
k k

=
+

207
207


where
( )
( )
2
2
0
2
tan
k
k
k

.

As a result,
( ) ( )
odd
k
k
x t x =

t will satisfy

( )
( )
2
0
0
odd
4 1
2 sin
k
F t
F
x x x k t
m k m

+ + = =
.

Notice that the amplitudes of the various sines get smaller and
smaller with increasing k once were past the resonant frequency
0
. Two reasons:
1. Individual b
k
have a k in the denominator so the large-k
Fourier components of the force contribute smaller and
smaller amplitudes;
2. The line shape of
( )
k
x t (the oscillators response to the
frequency component k) causes the amplitude to drop with
increasing k, roughly like
2 2
k .

208
208


Periodic impulses of short duration
Another example: lets try a driving force
( )
F t that looks like this:

Ive indicated the period T. To make life simple, Im going to pick
as my example the following force:
( ) ( )
0
0 2
1
cos 10 1 10 10
2
0 1
T t T
F t F t T T t T
T t T
10
0 2

= +
+ +
+

with the shape of repeating, as indicated in the diagram
above.
( )
F t

You might expect that only a
( )
cos 10 t term will contribute to the
Fourier series. This isnt correct, however: you need a lot of other
T
-T/10 +T/10

209
209

terms to keep zero except during a narrow time window.
( )
F t

J ust a reminder:
( ) ( ) ( )
( )
( )
( ) ( )
( )
( )
0
2
2
2
2
2
2
2co
2sin
n
a
F t
T
0
2
cos sin
s
0
n n
T
T
T
n
T
T
n
T
F t n t b n t
T
a dt
n t
a F t dt n
T
n t
b F t dt
T
= + =

=
= >
=
;

Do first:
0
a

( )
( )
2 10
0 0
2 10
11
cos 10 1 .
2
T T
T T
F t
a dt F t T
T T

+ +

= = + dt

Note the change in integration limits since
( )
F t is zero during the
period that includes t =0 except for the indicated interval.

210
210

10
10
10
0
0
10
10 10
0 0
0
11 10 10
cos 1 sin
2 2 10
2
.
2 10 10
T
T
T
T
T T
F t T t
F dt
T T T T
F F T
a
T

+
+
+

t

+ = +

= = =

Now evaluate :
n
a

( )
( )
( )
( )
( ) ( ) ( )
2 10
0
2 10
10 10
0
10 10
2cos 2cos
cos 10 1
2
cos cos 5 cos .
T T
n
T T
T T
T T
n t n t
F
a F t dt t T
T T
F
n t t dt n t dt
T

+ +

+ +

= = dt +

= +

Again, note the change in integration limits. Because the limits are
no longer 2 T , neither integral inside the square brackets will be
zero, generally. Instead, we get this, doing the integral in two
pieces:

211
211

( ) ( )
( ) ( )
{ }
( )
( )
( )
( )
( )
( )
( )
( )
( )
10
0
10
10
0
10
10 10
10 10
0
10
10
0 0
cos cos 5
cos 5 cos 5
2
sin 5 sin 5
5
2 5 2 5
sin 5
5
2 5 10
sin 5
sin 5 10
5
T
T
T
T
T T
T T
T
T
F
n t t dt
T
F
n t n t dt
T
n t n t
n
n n
F
T
n t
T
n
n
n t
n T
F F
T n T

+ +

+
= + +

+
+
=

+

+ =
+

= +
+
( )
10
5
2 5
10 5.
n
n
T n

Since 2 T = this first integral evaluates to

( ) ( )
( )
( )
( )
( )
10
0
10
0
0
cos cos 5
sin 5 5 sin 5 5
5
2 5 5
10 5.
T
T
F
n t t dt
T
n n
F
n
n n
F n

+

+

= +

To finish the calculation of , we need to evaluate
n
a

212
212

( )
( )
10
10
0 0 0
10
10
sin
cos sin .
5
T
T
T
T
n t
F F F n
n t dt
T T n n

+
+

= =

Adding the two integrals yields

( )
( )
( )
( )
( )
( )
0
sin 5 5 sin 5 5
2sin 5
5
2 5 5
n
n n
n
F
a n
n n n

+

= + +

+

Also,
( )
0 0
5
2sin
10 2 5 10
F F F
a
0
= + = .

We already found that
0
0
10
F
a = .

Now for .
n
b
( )
F t is an even function:
( ) ( )
F t F t = + but
( )
sin n t is odd:
( ) (
sin n t n t
)
sin = . As a result,
( ) ( )
n t sin F t is an odd
function. Recall that the integral of an odd function over an
interval symmetric about 0 t = is zero.
Consequently, . 0
n
b =

213
213


As with the square wave driving force, we know the particular
solution for one frequency component will satisfy
( ) ( ) ( ) ( )
2
0
1
2 c
k k k k
os x t x t x t a k
m
t + + =
and that
( )
( )
( )
( )
( )
2
2 2
2 2
0
cos
4
k
k
k
k t
a
x t
m
k k

=
+

. . . and that
( )
( )
2
2
0
2
tan
k
k
k

.

Also,
( ) ( )
all
k
k
x t x =
t will satisfy

( )
2
0
2 x x x F t + + = m.

Here are plots of the coefficients for different terms in the Fourier
expansions for the square wave (only the sine terms are plotted)
and the force weve just analyzed (only the cosine terms appear).

214
214


Note the greater importance of higher frequency terms in the
second case. In general, the more structure there is at time scales
which are short compared to the period, the greater the importance
of the high frequency components.

Impulse response and an introduction to Greens functions
We ignored transients during our discussion of Fourier series.
Lets take them into account now. In fact, well work out a way to
describe an oscillator which is driven by an arbitrary (non-
periodic) force.

215
215

Well go about it this way. If we know the solution x
i
(t) to the
equation
( )
2
0
a
x x x F
i i i
m
+ + = t
for i =1 then we can define , 2, 3, n

( ) ( ) ( ) ( )
1
and
n
i i
i
x t x t F t F
=
= =
t

and know that
( )
x t will automatically satisfy
( )
2
0
a
x x x F
m
+ + = t .

This was handy for solutions to equations in which the driving
force was periodic so that we could break
( )
F t up into its Fourier
series representation as a sum of sines and cosines.

But Fourier analysis may seem a little more abstract than youd
like, and it is of limited use when the driving force is not periodic.

216
216

Response of an oscillator to a single, short impulse
Heres another approach: lets figure out the response of the
oscillator to an impulse, which is an applied force of short
duration. After we do this we can then describe an arbitrary force
as a sum of impulses.

Heres a sketch:

217
217


The smooth curve is approximately equal to the sum of the
individual impulses. With more impulses of shorter time duration,
our approximation would be more accurate.
( )
F t

As with Fourier series, if we can work out a solution to the motion
of an oscillator subjected to the impulse

( )
0

0
i
i i i
i
t t
F t F t t t t
t t t
<
= < +
> +

with appropriate
( ) ( )
0 , 0 ,
i i
x v
( )
i
then well be able to sum all the x
i

to have
( )
i
x t x t =
.

The first step is to figure out what a damped oscillator does when
we hit it with a single pulse. Thornton and Marion work this up in
a fashion thats considerably more complicated than necessary.

Heres a better way, based on a sort of successive approximations
technique.

218
218


Consider the response of a free particle of mass m, initially at rest
at the origin, that is hit with a constant force F
0
for a very short
time t. The particle accelerates briefly, reaching a speed
0
v F t m = as it glides through the position
( )(
2
0
1
2
)
x F m t =
just as the impulse ends.

A damped oscillator wont behave exactly the same way: as it
speeds up, the drag force will oppose the applied impulse. And as
it moves away from equilibrium, the spring will also begin to
oppose the applied force. But how important are these effects?

The average speed of the mass during the interval in which the
impulse is applied is close to
0
2 v F t 2m = so the average drag
force is approximately
( )
0
2 2 m v F t = .

The force exerted by the spring at the end of the impulse is
( )(
2
0
2
k
k x F m t =
)
; the average force during the impulse is, of
course, smaller.

219
219

If we correct the strength of the impulse to account for drag and
the opposition from the spring, we can say that the net force F
net
on
the mass is between F
0
and
( )( ) ( )
2 2
0 0 0 0
1
2 2
k k
F F t F m t F t t
m

=

.
As a result, the fractional error F in the force if we completely
neglect both damping and spring forces is

( )
( )
2
0 0
2
0
1
2
2
k
F F t t
F k
m
t t
F F

=
m
+ .

In short, as long as we use a sufficiently brief time interval, the
fractional error we incur using only the free particle force is
negligible.

This lets us treat the behavior of an oscillator initially at rest at the
origin which is struck by an impulse as if it were nothing more
than an exercise in non-zero initial conditions. The oscillator
behaves the same as one that is at the origin, but with velocity
0 0
v F t = m, at t =0.

220
220

For an underdamped oscillator with x
0
=0 we found
( ) ( ) ( )
0 0
sin sin
t t
d d
d d
v F t
x t e t e
m

t

= =

where
2 2
0 d
= .

For the overdamped case we have (again, for x
0
=0)

( )
( ) ( )
2 2 2 2
0 0
0
2 2
0
2
t t
F t
x t e e
m

+

=

For the critically damped oscillator (with x
0
=0) we have

( ) ( )
0 0
t t
x t v te F t m te

= = .

Naturally, for all three cases
( )
0 x t = when 0 t < .

If we had delivered the impulse at time
0
t t = instead of our
solutions would look similar to those listed above except for the
replacement everywhere. Further, wed also have
when . For example, the underdamped oscillator
0 t =
0
t t t
0
t t <
( )
0 x t =

221
221

would move with
( )
( )
( ) ( )
0
0
0
0 0
0
.
sin
t t
d
d
t t
x t
F t
e t t
m

t t
<

Response of an oscillator to a non-periodic driving force
We can approximate a general force F(t) as a series of pulses
delivered at times , each with strength
k
t k = t
( )
k
F t , as shown in
the diagram a few pages back. Lets refer to the solution to the
oscillators motion for a single impulse delivered at t
k
as
( )
k
x t .

As long as the oscillator is at rest at the origin before F(t) turns on,
well have the oscillators position as a function of time, in
response to the full force F(t), as
( ) ( )
k
k
x t x t =
.

Keep in mind the distinction between t
k
, the time that the k
th

impulse is delivered, and t, namely the time on the clock on the
wall in our laboratory.

222
222

t
)
Rewriting our earlier equation by substituting and
, as well as rearranging terms slightly, we have
0
t k
(
0
F F k t
( ) ( )
( )
( ) ( )
0
sin
t k t k
d
d
t k
x t F k t
e t k t t
m

<
t
t k t

Our full solution involves a sum over all k such that kt is in the
past (kt <t):
( )
( )
( )
( ) ( )
sin
t k t
d
k
d
F k t
x t e t k
m

t t
.

In the limit that this becomes an integral, with the
integration variable replacing kt and
0 t
t dt replacing t:

( )
( )
( )
( ) ( )
sin
t t
t t
d
d t
F t
x t e t
m
=

=
t dt
.

Anthropomorphizing somewhat: as far as the integral is concerned,
t behaves like a constant. Only t interacts with the integrals
calculus machinery.

223
223

Greens functions
The form of our solution is this (neglecting for the moment any
non-zero initial conditions):

( ) ( )
( )
( ) ( )
sin
t t
t t
d
d t
e t t
x t F t
m

=
=
dt

=

.

Since t inside the integral, our solution is causal: it shows no
influence from things that have not yet happened.
t

The quantity in square brackets inside the integral is called a
Greens function, named for the English engineer George Green.

For an underdamped oscillator we define
( )
( )
( ) ( )
sin
,
t t
d
d
e t
G t t
m

t
= ,
which allows us to write
( ) ( ) ( )
,
t t
t
x t F t G t t dt
=
=
=

.

The Greens function serves as a helper function that allows us

224
224

to sum the cumulative effects of the driving force over the
oscillators past. This way we can predict its present behavior if we
know its history.

Theres a simpler way to think about the oscillators Greens
function: it tells us how the oscillator behaves at time t in response
to an (earlier) unit-strength impulse at time . t (By unit strength I
mean .)
0
1 F t =

This is easy to see: a unit impulse at t gives rise to motion
satisfying

( )
( )
( ) ( )
( )
( ) ( )
0
sin
sin
t t
d t t
d
d d
e t
F t
x t e t t
m m
= =

t

we have
( ) ( )
, x t G t t = .

Note that tells us how to propagate the effects of an
impulse at t forward in time to t so we can add it to the effects of
other impulses, also propagated forward to time t.
(
, G t t
)

225
225


Greens function for the Coulomb potential
This notion that you can use a helper function to figure out how
to combine the effects of multiple sources/forces/causes is familiar
from electrostatics.

For example, the potential at a point r
caused by a charge q
i

at
position is
i
r
( )
i
q
r
r r
=

.
If there are multiple charges wed superpose their effects to write
( )
1
n
i
i
i
q
r
r r
=
=

.

If instead theres a charge density
( )
r
so that the charge

contained in a small volume dv is
( )
r dv dq =

, wed write the
Coulomb potential as

( )
( )
( )
all
all
space
space
1
r dv
r r
r r r r
dv

226
226


The helper function
1
r r

is the electrostatic Coulomb Greens
function:
( )
1
,
c
G r r
r r
=

.

Note that is a function of position, not time, and that the
potential at caused by a unit charge at r
c
G
r
is
( ) ( )
1
,
c
r G
r r
r r = =

t
.

Electrodynamics is a considerably richer subject than electrostatics
because complications arise due to the finite speed of propagation
of information. Even so, a Greens function approach to
calculating the Coulomb potential has intuitive appeal. Heres how
it works

Allowing for the movement of charges that are the sources for the
potential we need to include time dependence in the charge
density:
( ) ( )
, r r

. Note that

227
227

( ) ( )
all
space
1
, , r t r t dv
r r

since the potential cannot depend on the instantaneous distribution
of charge. (Only after enough time has passed for the observer at r

to become aware of the charge density at r
will the potential

reflect whats happening at . r
)

Instead, we have
( )
all
space
1
, ,
r r
r t r t dv
c r r

=

since the time delay associated with information arriving at from
is
r
r r c

. In electrodynamics courses youll hear t r r c

referred to as the retarded time.

Satisfying the initial conditions
Keep in mind that were working on generating a solution to the
inhomogeneous equation
( )
2
0
2 x x x F + + = t and that I had
assumed that both x and x were zero at t =0.

228
228

0
Recall that I had defined x
1
and x
2
to be solutions to the
homogeneous equation
2
1,2 1,2 0 1,2
2 x x x + + = . We are always
free to add arbitrary amounts of the solutions x
1
and x
2
into x(t) in
order to satisfy non-zero initial conditions on position and velocity.
Any function of the form
( ) ( ) ( )
1 1 2 2
x t A x t A x t + + will
automatically satisfy the inhomogeneous equation as long as x(t) is
also a solution.

Weve been assuming that the oscillator was at rest before we
applied a non-zero force. That way, when we represent the force
as a series of impulses
( ) ( ) ( )
1 2
F t F t F t = + + and calculate
( )
k
x t
( )
for each impulse, well be able to sum them so that
( )
k
k
x t x
t = . If we didnt impose the condition that each

( )
k
x t
was zero, with zero first derivative at t =0, then x would blow up
on us at t =0 if we sum a large number of successive impulses.

So: what do we do if we want
( )
0 x 0 and/or
( )
0 v 0 ? Its easy.
All we have to do is find a solution to the homogeneous equation
which has the desired
( ) ( )
nd 0 0 a x v , and then add it to our
Greens functionderived
( )
x t . Thatll do it. Take note: if we

229
229

specify initial conditions at 0 t = , we must use 0 t = as the lower
limit in our Greens function integration. (Because the solutions to
the homogeneous equation are exponentially damped, they will die
away completely if we impose them instead at t = . As a result,
if we know the complete history of our oscillator, rather than
starting it at time zero, we can neglect the transient solutions.)

Assuming the oscillator is under damped, the results from a few
lectures ago are all that we need:
( ) ( ) ( )
( )
, t t
( )
0
0 0
0
cos .
t t
t
d
x t F t G dt
v x
sin
t t
d d
x e t

+ + e t
=
=

=
+

Some Greens function examples
Since
( )
( )
( ( ) )
sin
,
t t
d
d
e t
G t t
m
t
=
and

230
230

( ) ( ) ( )
,
t t
t
x t F t G t t dt
=
=
=

or
( ) ( ) ( )
( ) ( )
0
0 0
0
,
cos sin
t t
t
t t
d d
d
x t F t G t t dt
v x
x e t e

=
=

=

+
+ +

problems in which the applied force can be expressed in terms of
polynomials and/or (possibly complex) exponentials are often
amenable to solution.

Lets do an easy one:
( )
0
0 0
0

t t
F t
F t t
=

>
with
0 0
0, 0 x v = = .

We have
( ) ( )
( )
( ) ( )
( )
( ) ( )
0
0
0
0 0
sin
0
sin
t t
t t
d
d t
t t
t t
d
d t t
e t t
x t F t dt
m
t t
e t t
F d
m

=
=

=
=

=
t t t
<

231
231

Note the change in lower limit.

Lets plunge ahead with the integral: rewrite
( ) ( )
( ) ( )
sin
2
d d
i t t i t t
d
e e
t t
i

=
so
( )
( )( ) ( )( )
0
0
2
d d
t t
i t t i t t
d t t
F
x t e e
im

dt
=
+
=

=

.

Change variables: u t t

( )
( ) ( )
( )
( )
( )
( )
( )
( )
( )( )
( )
( )
( )( )
( )
0
0 0
0
0
0
0
0 0
0
0
2 2
2
1 1
2
1
2
1
d d
d d
d
d
u
i u i u
d u t t
u u
i u i u
d d d
u t t u t t
i t t
d
i t t
d d
d
F
x t e e du
im
F
e e
im i i
i e
F
im
i e

=
+
=
= =

= =

+

=

= +
+

+

= +

+
+

+

so

232
232

( )
( )
( ) ( ) ( )
( )
( ) ( ) ( )
( )
( )
( )
( ) ( )
( )
( ) ( )
( )
0 0 0
0 0 0
0 0
0
0 0
0
0
2 2
0
2 2
0
2 2
2
2
2
2
1
d d
d d
d d
d d
t t i t t i t t
d d
t t i t t i t t
d d
i t t i t t
t t
d d
i t t i t t
d d
t t
d
i i e e e
F
x t
im
e e e
e e
e
F
m
e e
e
i
F
m

+

=

+

+

=

+

=
+
( )
( ) ( ) ( ) ( )
0
0 0
cos sin .
t t
d d
d
e t t t

t

+

Recall that
2 2
0 d
2
so
2 2 2
0 d
k m + = = . As a result,

( )
( )
( ) ( ) ( ) ( )
0 0
0 0
1 cos sin
t t
d d
d
F
x t e t t t t
k

= +

for . Thats our solution!
0
t t

Lets check that
( ) ( )
0 0
0, 0 x t v t = = :
( ) ( ) ( )
0 0
0
1 cos 0 sin 0
d
F
x t e
k

= + =

0.

So far so good.

233
233


Differentiate to find the velocity for :
0
t t
( )
( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
( )
( ) ( )
( )
( ) ( )
0
0
0
0 0
0
0 0
2 2
0
0
0
0
cos sin
sin cos
sin
sin .
t t
d d
d
d d d
t t
d
d
d
t t
d
d
t t t t
F e
v t
k
t t t t
F e
t t
k
F e
t t
m

+

=

+

+
=

=

We have
( ) ( )
0
0
0
sin 0 0.
d
F e
v t
m
= =

Since we wanted
( ) ( )
0 0
0, 0 x t v t =
( )
0 0
,
= , were done. If instead wed
wanted
( )
0 0
x t x v t v = = as our initial conditions wed need to
add to an extra piece to write

( )
( )
( )
( ) ( ) ( ) ( )
( )
( ) ( )
( )
( ) ( )
0
0 0
0
0 0
2 2
0 0
0 0
1 cos sin
cos sin .
t t
d d
d
d
t t t t
d d
d
F
x t e t t t t
m
v x
0
x e t t e t

= +

+

+
+ +

t

234
234


Heres another example, where I will assume that the underdamped
oscillators position and velocity are zero at t =0:

( ) ( )
0 0
0
0 0
sin 0 .
0
d
t
F t F t t t
t t
= <
>

0 =

Since we wont have to do anything more than
integrate the Greens function expression:
( ) ( )
0 0, 0 x v =

( ) ( )
( )
( ) ( )
( )
( )
( ) ( )
( )
0
min ,
0
0
sin
sin
sin .
t t
t t
d
d t
t t t t t
d
d
d t
e t t
x t F t dt
m
e t t
F t
m

=
=
=
=

=

dt =

Keep in mind that is zero when
( )
F t
0
t t > , so well have to
watch how we handle to upper limit when
0
t t > .

Replace
( ) ( )
sin sin
d
t t t
d
by complex exponentials:

235
235

( ) ( )
( ) ( )
( ) ( ) 2 2
sin sin
2 2
1
4
d d d d
d d d d
i t t i t t i t i t
d d
i t t i t t i t i t
e e e e
t t t
i i
e e e e

=

= + +

so that
( )
( ) ( )
( )
( )
( )
0
0
min ,
2 2
0
0
min ,
2 2
0
0
2
0
4
4
2cos
4
d d d d
d d d d d d
d d
t t t
t
i t t i t t i t i t t
d t
t t t
t
i t i t i t i t i t i t t t t
d t
t
i i t t
d
d
F e
x t e e e e e dt
m
F e
e e e e e e e e e e dt
m
F e
t e e e
m
+
=
=

= + +

= + +

= +
( ) ( )
( )
0
min ,
2
0
d d
t t t
t i t i t
t
e e dt

=
+
=

t

or
( ) ( )
( ) ( )
( )
( )
( )
( )
( )
( )
( )
0
0 0
0
min ,
2 2
0
0
min , min ,
2
0 0
0
min ,
2
0
2cos
4
2cos
2
4
2
d d d d
d
d
d
d
t t t
t
i t i t i t i t t
d
d t
t t t t t t
i t
i t d t
t
d
t t
t t t
i t
d
i t
d
t
F e
x t t e e e e
m
t
e
e e
i
F e
m
e
e
i
+
=
= =

= =
=
+
=

e dt = + +

( )
( )
( )
( ) ( )
( )
( ) ( )
( )
0 0
0
min , 2 min ,
0
2 min ,
2cos
1 1
2
4
1
2
d
d
d
d
i t
t t i t t d
t
d
i t
i t t d
d
t
e
e e
i
F e
m
e
e
i

=

+

+

236
236

Grind away to find:
( )
( )
( )
( )
( )
( )
( ) ( )
( )
( )
( )
0
0
0
0
2 2
2 2
0 0
0
0
2 2
2 2
2 1 1
cos sin
4
4
cos
1
2
cos 2 2 sin 2
2 4
cos 2 sin
2 4
t t
d
d d
d
d
t d
d d d d d t t
d
d
d d d
d
e e
F
t t
m
t
e
x t
t t t t
F
e e
m
t t

t t
t t
+
+
+

=

+

+ >

Its messy, but not complicated. Note: no guarantees that my
integrals are correct on this one!!

Lets stop dealing with harmonic oscillators at this point. Well
return to them a bit next semester in Physics 326.

237
237


238
238


239
239


4-144
240
240
Physics 325, fall 2012 University of Illinois
Motion in rotating frames of reference

Motion in rotating frames of reference........................................5-1
A qualitative explanation for the existence of Coriolis forces....
...............................................................................................5-2
Pendulums and hurricanes...................................................5-13
Quantitative description of motion in rotating frames.........5-16
Velocities and accelerations.................................................5-27
Effective force.....................................................................5-32
angular acceleration term.....................................................5-34
centrifugal term....................................................................5-34
Coriolis term........................................................................5-36
The earths oblateness..........................................................5-38
Coriolis effect on a hockey puck.........................................5-39
Coriolis effect on a free-falling object.................................5-41
Foucault pendulum analysis................................................5-44

There are a lot of cross products that arise when dealing with
motions in rotating frames of reference. Theyre there mostly
because we choose to use as an origin a point that keeps shifting
from one inertial frame to another.

Since an object not experiencing a net force moves with constant
(vector) velocity, when viewed from an inertial frame, this same
object, viewed from a non-inertial frame will have its apparent
velocity shifting as time passes.

5-1
241
241

A qualitative explanation for the existence of Coriolis forces
The surface of the earth is an example of a rotating reference
frame: were spinning around (and also moving in a ellipse around
the sun), so the apparent velocity of distant objects (like stars) will
change as the day and year evolve.

There are kinematic consequences associated with viewing
something from a noninertial frame.

Many of them arise because of the following facts:

1. An object moving in a circle of radius r with constant speed
v, when viewed from an inertial frame, must experience a net
force
2
mv r which is inwards. If the net force on it is
different from this, it wont move in a circle, or wont move
with constant speed.


5-2
242
242

2. Angular momentum is conserved unless some external force
creates a torque which acts on the object.

Lets see how these facts influence/affect what happens if we glue
a coordinate system to the earths surface at some latitude .

F
net
R

Earth
=2 radians/day

Distance from object (person ?) to the earths axis is cos R . The
object travels in a circle of radius cos R with 2 =
radians/day. The objects speed is
( )
cos R .

Net force on the object, if it is at rest on the earths surface, is
( )
2 2
cos cos mv R m R = .


5-3
243
243

R is the radius of the earth, of course.

The net force is directed to the left in the diagram. This force is a
sum of gravity, the normal force from the ground pushing on the
object, and frictional force (which keeps the object from sliding
towards the equator).

Heres a diagram of the forces:
friction (parallel
to surface)
normal force (earths
surface doesnt compress)
gravity (towards the
center of the earth)

Another diagram, doing the vector addition:

5-4
244
244

x
gravity
normal force
friction
m
2
Rcos net force to the left

Im assuming that the earth is a perfect sphere. The normal force
becomes whatever is necessary to keep the earths surface from
compressing. This is also true, of course, if we allow for the
oblateness (ellipsoidal shape) of the earth.

What happens if you go for a walk?

The earth turns west-to-east. Define a coordinate system whose
origin is fixed to a point on the surface of the earth, which has
pointing south, and pointing east. up, z y

5-5
245
245

If you begin to walk east, youre increasing your effective , since
youre moving in the same direction as the earths rotation. If you
walk with constant speed, the forces acting on you have to readjust
themselves to increase the (net) inwards centripetal force acting on
you:

1: gravity doesnt change;
4: net (centripetal) force has to increase since your is greater;
2: friction has to increase, but stays parallel to earths surface;
1

3

2

4

1

3

2

4

north pole
z

y
x

5-6
246
246

3: normal force has to decrease, but stays to earths surface.

Without friction, youd slide south, with greater acceleration than
if you werent walking.

Lets say youre standing still, holding a plumb bob. The only
forces acting on the plumb bob are gravity and the tension in the
string; these two forces have to sum to
2
m r since the plumb bob
is turning through a circle of radius r once per day. Here,
cos .
earth
r R =

A diagram:

mg towards the earths center

m
2
r net force
tension along string


5-7
247
247

r, if you prefer:

ets rotate the picture so that straight-towards-the-center-of-the-
Note that the plumb bob hangs a bit to the south of straight down.
O

m
2
r net force


L
earth is down in the sketch.

north

east
m
2
r net force


5-8
248
248

you walk to the east while holding the plumb bob If increases
. (youre going in the same direction the earths surface is moving)
Once the plumb bob steadies down, it will also be moving in a
circle with the increased . As a result, the required centripeta
force in
2
m r
l
must now be greater than it was before. The
gravitatio orce doesnt change, so only the string tension
change to provide this additional inwards force. What has to
happen is this:

nal f can
The line of the string has the plumb bob hanging further to the
alking to the east appears to produce a southerly force which

south than before.

W
acts on the plumb bob (and on you, which is counteracted by the
increase in north-directed friction). In reality, the direction of the
north

east
m
2
r net force
mg
north

east
new m
2
r
mg
new tension


5-9
249
249

he opposite would happen if you walked west:
string tension is changing in order to provide the additional force
that is directed towards the earths rotation axis.

T would
decrease,
2
m r would decrease, so the orientation of the s
would be closer to straight up-and-down. You would see the
plumb bob shift towards the north.

tring
or both the east and west strolls, F would be changing while
you do all this in the southern hemisphere, the directions of the
f
r
would be staying constant.

If
apparent forces are opposite to those in the northern hemisphere.
The line of the string puts the plumb bob somewhat to the north o
where it would be in the absence of the earths rotation.


5-10
250
250

north pole
x
z
y

m
2
r net force



m
2
r net force


Walking east makes the plumb bob shift to the north (so the string
can exert more of a southerly pull, etc. etc.

Lets go back to thinking about motion in the northern hemisphere.


5-11
251
251

Imagine you walk south now. In this case, stays constant (it still
takes 24 hours to go around once) but r is increasing. As long as
youre walking, youre continually increasing r, since cos r R = .

Angular momentum is
2
mvr m r = .

Unless you apply a torque to the plumb bob, for the plumb
bobs rotation about the earths axis will decrease as you walk
south since angular momentum is conserved. What happens is
this: the plumb bob lags behind during your rotation around the
earths axis, dragging behind you to the west. You have to tow it
along to force it to keep up as you move south, and the earth spins
west-to-east. As a result, walking south makes it appear as if a
westerly force was acting on the plumb bob, which you have to
counteract by pulling on the string.

If you walk north, r decreases so the plumb bob wants to turn
around the earths axis with increased . It tries to sail off to the
east, so it looks (to you) as if walking north caused an easterly
force to act on the plumb bob.


5-12
252
252

Direction you walk Apparent force on bob
north east
south west
east south

west north
N

E

The plumb bob always seems to feel a force which pushes it to
your right when you go for a walk while holding the string of the
plumb bob.

Pendulums and hurricanes
A pendulum which is initially swinging in, say, an east-west plane
will feel this coriolis force pushing it to the south during its west-
to-east swing and to the north during its east-to-west swing.

East

West


5-13
253
253

The plane of the pendulum will rotate. This is what makes a
Foucault pendulums arc turn in a circle. (Note that all the
Foucault pendulums youve seen in museums have a small amount
of driving applied to maintain the swing. Theyll come to a stop
after a while without this.)

The rate of rotation will depend on how the apparent force behaves
as a function of your velocity on the earths surface. For example,
at the equator, walking north or south doesnt change the radius of
the circle through which your plumb bob turns each day, to first
order. Therefore, the bobs angular momentum doesnt change, so
theres no apparent force which acts on it. Walking east or west,
the tension in the string decreases or increases, but since the bob
hangs straight down, theres no change in its orientation.

If you suspend a Foucault pendulum above the north or south pole,
all that happens is that the earth rotates underneath the pendulum,
which keeps the plane in which it swings fixed in an inertial frame.
Seen from an observer standing at a fixed point on the earths
surface near the pole, the pendulums plane rotates one full turn
every 24 hours.


5-14
254
254

As a result, as you move from the equator to the poles, the rate of
rotation increases from no rotation at all to one full circle every 24
hours. (Note that after 12 hours the east-to-west swing has become
a west-to-east swing so the plane of the motion looks like it did at
.) 0 t =

Hurricanes and other large storms are also influenced by coriolis
forces: air rushing inwards toward a low pressure region is given a
counter-clockwise rotation in the northern hemisphere and
clockwise in the southern hemisphere.

H

L

H

H


5-15
255
255

Quantitative description of motion in rotating frames
We need to be more quantitative now. Since well be interested in
how various vectors change when viewed from a rotating
coordinate system (recall that velocity is dr dt
), its time to figure

out how to describe a vector which changes when viewed from an
inertial frame but from the perspective of observers in a rotating
frame.

Let frame O be an inertial frame, let O' be the rotating frame.
Lets assume that the origins of O and O' sit at the same point in
space as O' twirls around. Lets also assume that the rotation axis
of frame O' stays fixed in space, when viewed from O. We dont
really need to impose these restrictions, but itll make it easier to
visualize what happens.

One last restriction: lets orient the axes for O and O' so that the z
axis of O (the non-rotating frame) is along the frame O' rotation
axis, with the axis of O' along its rotation axis. Lets also
assume that the x and x' axes are aligned when
z
0 t = .


5-16
256
256

Here is the power of using vectors and other objects that transform
sensibly under rotations. Our conclusions, as long as we write them
using equations expressed as relationships among vectors, wont
depend on our chosen orientations for the O and O' axes, any more
than our conclusion that your plumb bob is pushed to your right
when you walk depended on the choice of east/west as
perpendicular to the earths rotation axis.

Heres how O' looks from the perspective of O at 0 t = and
. Ive drawn a vector t d = t A
on the plot which is stationary in the

inertial frame O.

'
axis at 0 x t =
'
A
z axis unchanged
axis at x t dt =
axis at 0 y t =
axis at y t dt =
'


5-17
257
257

O' is twirling so that its angular velocity
is along z + as viewed
fromO.

To observers in O', the vector A
seems to be twirling around so

that its projection on the x y plane moves in a direction that is
from the axis to the y x axis.

At time the 0 t = , , x y z components of A
seen from O' are

sin cos , sin sin , cos x A y A z A = = =

where A A

. Keep in mind that sin A is the length of the
component of A
projected onto the x y plane.

At time t , the axes of O' have rotated through the angle d = t so
that the vectors , , x y z components are now

( ) ( )
sin cos , sin sin , cos x A y A z A = = = .


5-18
258
258

For small we can use derivatives to help evaluate the changes
in , , x y z :

( )
sin cos cos
cos
sin
sin sin ,
x A
d
A
d
A

=

( )
sin sin sin
sin cos ,
y A
A

=

=

0. z =

We have, of course, A x x y y z z = + +
.
When viewed from the inertial frame O, the coordinate system O'
is twirling () with z =
. We can come up with a vector that

corresponds to a rotation angle for frame O' by defining
dt =

. In our case, z =
but we could certainly do this

with other orientations of the O' rotation axis if we wanted to.

5-19
259
259

Note that A

is a vector that points in the same direction as the
change in A
, from the perspective of observers in the rotating

frame: A

is perpendicular to the plane defined by A
and
.
Further, note that A

is proportional to both an A d

.

This seems like a good candidate to describe the change in A

since sin sin sin cos A A x A y =

.
(Since A
and A
are the same at t =0 Im going to leave off the

primes some of the time in the following equations)
x
A
z
y
'

'

5-20
260
260

x
A
z
y
'

Lets calculate A

now to see if it actually is what we need:
( ) ( )
( )

sin sin sin cos
x y y x y z z y
z x x z
y z x

A A A z A A
A A y
A x A y
x
A x A y
A

= +
+
=
=
=

after plugging in.
This is just what we want: A

turns out to be the same as A
.
Since dt =

, we can rewrite the expression for the change in
A
, in the limit that , as 0 t


5-21
261
261

( ) ( )
dA A dt A dt = =

.

What would happen if A
wasnt a constant vector when viewed

from the inertial frame O ? (For example, A
might be the position

vector of an object that is moving.) We expect that the observers in
the rotating frame will see A
changing both due to its intrinsic

behavior in the inertial frame O and also the rotation of frame O'.

Lets work this it in the following way: define a pair of vectors
1
and
2
A A

.
1
A
corresponds to the vector A
at time 0 t = , while
2
A

corresponds to A
at time t dt = . Since A
is the same as
1
A
at the
start of the interval and then the same as
2
A
at the end of the

interval, the difference between
1
A
at 0 t = and
2
A
at , from
the perspective of the rotating frame O', will tell us what we want
to know.
t = dt

Work it out now.

In the inertial (non-rotating) frame O we have

5-22
262
262

2 1
A A dA =

with
( ) (
1 1
0
)
A t A t = = =

dt as well as
( ) ( )
2 2
0 A t A t = = = dt

.

In the rotating frame O' we have

( ) ( )
( )
1 1 1 1
0 dA A dt A A dt = =

and
( )
2 2
dA A dt =

.

Note the primes on the vectors when viewed from the rotating
frame.

( ) ( )
( ) ( )
2 2
2 2
0
0 0
A t dt A t dA
. A t A t
= = = +
= = =

dt

As a result,

( ) ( ) ( ) ( ) (
( ) ( ) ( )
2 1 2 2 1
2 1 2
0 0 0
0 0
A t dt A t A t A t dt A t
)
0
0 . A t A t A t
= = = = = =

= = = =

dt


5-23
263
263

At , the 0 t = , , x y z axes of O' coincide with the x, y, z axes of O.
As a result, a vector in O' at 0 t = has the same components as the
vector viewed from O at t 0 = . In equations:

( ) ( )
( ) ( )
1 1
2 2
0 0
0 0
A t A t
A t A t
= = =
= = =

so that
( ) ( ) ( ) ( )
2 1 2 1
0 0 0 A t A t A t A t

= = = = =

0
d
.

Lets call .
2 1 fixe
frame
A A dA

As a result, we can rewrite the equation for
( ) (
2 1
0 A t dt A t
)
= =

on the last page as

( ) ( ) ( )
2 1 fixed 2
frame
0 0 A t dt A t dA A t d

= = = =

t

In the limit that
2
0, dt A A
1

of course. We can write

rotating fixed
frame frame
dA dA A dt =


5-24
264
264

where Ive defined
( ) ( )
rotating 2 1
frame
0 dA A t dt A t = =

.

As a result, we have

rotating fixed
frame frame
dA dA
A
dt dt

where
is the rotating frames angular velocity as seen by

observers in the fixed (non-rotating) frame.

This is true for any vector, as long as it has the same magnitude
and direction in both frames at the instant 0 t = . Also note that we
can use either or A A

in the A

piece of dA dt
since the two

vectors coincide, having the same magnitude and direction at 0 t = .

We can rearrange terms to write

fixed rotating rotating
frame frame frame
.
dA dA
A
dt dt

= +


5-25
265
265

What happens if
changes? We can use the equation we just

developed to describe this (imagine that
is nothing more than a

vector oriented along the z direction):

fixed rotating
frame frame
d d
dt dt

= +

But 0 =

so d dt
is the same as d dt
.

Keep in mind that the equation
frame frame frame
dA dA
A
dt dt

= +

is
true regardless of the orientation of the frame O' rotation axis. In
addition, it doesnt matter whether we use components of A
as
seen in O or A
as seen in O': at 0 t = , their components are the

same. But we are forced to require that and A A

have the same
magnitude and direction in the two frames at the instant t . 0 =


5-26
266
266

Velocities and accelerations
Now lets take a look at how velocities behave. Replace , A A

by
a position vector and use the machinery weve just developed.

fixed rotating
frame frame
dr dr
r
dt dt

= +

.
Lets define
fixed
dr
v
dt
and
rotating
frame
dr
v
dt
. Note that an object

fixed to one point on the surface of the (rotating) earth will have
. 0 v =

We have
fixed rotating
frame frame
dr dr
r v v
dt dt
r
= + = +

.

Note: if the origin of O' had been moving with velocity V as seen
by observers in O, wed have
v V v r = + +

instead.

How about accelerations?

Always keep in mind that our equation

5-27
267
267

frame frame frame
dA dA
A
dt dt

= +

only works if the vectors and A A

have the same magnitude and
direction in the two frames at the instant 0 t = . So we cant just
replace A
by v and
by v
in hopes that
fixed
frame
dv
dt
will be the
same as
rota
frame
ting
v
rotating
frame
dv
dt
+
. It doesnt work since and can

be very different from each other as the O and O' axes align
momentarily. For example, something that stays at a fixed point on
the earths surface will have a non-zero velocity when viewed from
an inertial frame.
v

If we are working in the inertial frame O, well need to ask
observers in the rotating frame O' to tell us the components of the
vector
rotating
frame
dr
v
dt
they measure at 0 t = and then at t in

order to allow us to take the derivative:
dt =
( ) ( )
0
fixed
lim
t
v t t v t
dv
dt t

+
=
.


5-28
268
268

Well also do the same thing with r
, request that observers in O',

radio us the components they measure at at 0 t = and then at t d
so that we can calculate
t =
fixed
dr dt
.

Note the use if the subscript fixed since this is what we calculate
in the inertial frame based on numbers radioed to us by observers
in the other frame. This is a strange sort of equation: are
measured by observers in O', not in O, but the derivative is actually
taken by observers in O.
, v r

If corresponds to an object moving due east at constant speed
along the earths surface, well be told (by observers in O') that v
v

does not change so well conclude that
fixed
0 dv dt =
.

Do the differentiation of the equation v v r = +

to find:
[ ]
fixed
fixed
fixed fixed
.
d v r
dv
dt dt
dv dr
r
dt dt

+
=

= + +

Note the presence and absence of primes.

5-29
269
269

What we really want to be able to calculate is
rotating
frame
dv
dt
and
rotating
frame
dr
dt
in terms of positions, velocities, and accelerations in the

inertial frame taken in combination with the angular velocity (and
angular acceleration, if necessary) of the rotating frame.

Use the equation
frame frame frame
dA dA
A
dt dt

= +

from a few
pages ago to write

frame frame frame
dv dv
v
dt dt

= +

and also

fixed rotating
frame frame
rotating
frame
dr dr
r
dt dt

= +


5-30
270
270

Plug these into our earlier equation

fixed fixed fixed
frame frame frame
dv dv dr
r
dt dt dt

= + +

to write

fixed rotating rotating rotating rotating
frame frame frame frame frame
.
dv dv dr
v r r
dt dt dt

= + + + +

Well often assume that 0 =
.

Note that
rotating
dr
v
dt
.

Combine everything to write (dropping redundant rotating fixed
labels):

( )
2
dv dv
v r
dt dt

r = + + +


5-31
271
271

a The actual, true, genuine acceleration (as in F m = ) is a dv d =

t .

The effective acceleration, from the point of view of an observer
sitting still in the rotating frame is
eff
a dv dt =

.

Effective force
Multiply everything by m and rearrange terms to write

( )
effective
2 F F m r m r m v =

.

What does this all mean? Imagine we have a projectile thats
traveling with (apparent) speed v
as seen by observers stuck to the

surface of the rotating earth. Neglecting air resistance, the only
force that really acts on the projectile is the gravitational force
(here is towards the center of the earth). This term is
what comes just to the right of the equal sign, and is represented as
. The other terms arise from our insistence on using an earth-
based coordinate system that twirls around.
mgr
F
r


5-32
272
272

Lets be rather quantitative about our projectile: define a
coordinate system n O' to be used by people on the ground which
has its origin at the earths center, z up, x parallel to due south,
and parallel to due east. The coordinate axes rotate with the
earth.
y

north pole
z
x
y
you are here
east

Lets say the non-rotating coordinate system O also has its origin
at the earths center and that its axes coincide with the rotating
frames axes just as the projectile passes through the point
0 x x = = , , 0 y y = = z z = =radius of the earth.

In the fixed frame,
mdv
mgz
dt
=
so
dv
gz F
dt
=
.

5-33
273
273

5-34

What about the other terms in the equation

( )
effective
2 F F m r m r m v =

we just derived?

angular acceleration term
For our projectile problem, 0 =
so m r

is zero.

centrifugal term
The next term is
(
m
)
r

. When the projectile is close to the
earths surface for r the earths radius. The angular velocity
is
r rz
( )
cos z x
sin =
so cos r r y = +

, yielding

( ) ( )
2
cos sin cos r r x z = +

.


274
274

5-35
)
r

z
x
y
( )
m r

r

The direction of
(
m

is perpendicular to the earths
rotation axis. Its referred to as a centrifugal force because the
earth is turning, therell be an apparent force (not a real force)
which throws things radially outwards from the center of the circle
through which they turn. Note that the centrifugal term is
independent of , the projectiles velocity as seen by observers
using the rotating coordinate system.
v

Centrifugal force isnt a real force because it only comes about
due to the fact that your (rotating) coordinate system accelerates
away from the object, not because the object experiences a force,
when observed from an inertial frame.


275
275

5-36
Coriolis term
2m v

: this is the force which arises because a moving
object might be increasing (or decreasing) the diameter of the
circle through which it turns each day, or increasing (or
decreasing) its angular velocity. This is the Coriolis force
described earlier in the notes.

Note that 2m v

is perpendicular to both
and v
. It can
contain a component parallel to z
; if the Coriolis force isnt strong

enough to overcome gravity, something sliding around on the
ground will only exhibit accelerations associated with the
component of v

which is perpendicular to z
(the ground
wont compress; normal force cancels the z
component of the
Coriolis force.)

In the northern hemisphere, the Coriolis force pushes you to the
right. In the southern hemisphere it pushes you to the left:


276
276

5-37

2m v

v
2m v

v

To summarize, weve defined
eff
F
to be the product of mass and

the apparent rate-of-change of velocity seen from the perspective
of observers in the rotating frame. The true force F
is mass times
acceleration viewed from an inertial frame. We found

( )
2
eff
F F m r m r m v =

:

)
r
(
m

is a centrifugal term
2m v

is a Coriolis term.


277
277

5-38
The earths oblateness
I have a comment on centrifugal forces, then well do a couple of
examples.

Ive been assuming all along that the earth is a sphere. Actually,
its not. Every see a pizza chef toss/spin dough? You spin a lump
of dough and it becomes more oblate: it flattens out, getting larger
in the rotation axis. The earth does this too, though its a small
effect: the earth flattens out due to centrifugal (
( )
m r

)
effects until the surface is everywhere perpendicular to the sum of
the gravitational pull and centrifugal forces.

apparent net force
centrifugal force

mg towards the
earths center
round
oblate

A frozen lake will have its surface oriented to perpendicular to the
apparent net force, cancelling centrifugal effects.


278
278

5-39
oriolis effect on a hockey puck
e a hockey puck due north on a
C
Heres an example problem: slid
frictionless lake with an initial velocity thats 1 m/sec. The lake is
at 45 north latitude. What happens? Gravity, centrifugal, and
normal forces from the lake surface acting on the hockey puck sum
to zero so we have 2
eff
F m v =

.

as cos sin v v x v y +
where Lets describe is the angle

between due north and v
, and Im using a coordin te system as

drawn a few pages back, where
a
x is south, y is east, and z is up.
Im going to ignore the extremel small angl between the axis
y e z
(points away from the earths center) and the perpendicular to the
ices surface (points a tad to the north of z ).
Assume the puck doesnt go far, so the la ud tit e is ~constant.
( )
sin cos z x =

( ) [ ]
2 2 in cos sin sin cos sin m v m v y x z s =

.

279
279

5-40
The cos sin z term just changes the apparent weight of the
puck on the smooth ice and has no effect on the motion. As a
result, the effective force causing the Coriolis acceleration is

( )
2 sin cos sin
eff
F m v y x = + +

Note that
( )
2
2 sin cos sin cos sin 0
eff
v F m v = =

so .
eff
F v

Also note that
2
2 2 2 2
4 sin
eff
F m v =
, independent of since
2 2
cos sin 1 + = .

Consequently, the effective force causes the puck to move in a
circle.

Since
2
eff
mv
F
r
=

for the circle of radius r made by the puck,
2
=
eff
mv m
r
F
=
2
2
v
m
v
=
2 sin sin v

.

280
280

5-41

We have , 1 m/sec v =
5
7.3 10 rad/sec

for the earths rotation,
and sin so the pucks circular path on the ice has radius 1 2 =
( )( )
4
10 m=10 km
5
1
2 7.3 10
r
.

Coriolis effect on a free-falling object
Lets do another example. What is the net horizontal deflection on
a free-falling rock near the earths surface due to Coriolis effects?

For the rock,
( ) ( )
2
eff
F mgz m x xr m x v =

.

I am tired of writing primes on everything, so lets switch to using
unprimed variables to refer to quantities taken with respect to our
rotating coordinate system.

We have
( )
sin cos z x =
.
As a result, the centrifugal term in the effective force is
( ) ( )
2
cos sin cos m r m r x z = +


281
281

v
allowing us to write
( )
2 2 2
cos cos sin 2
eff
F m r mg z m r x m = +

.

,
z x z
v v v v
y
so neglect all contributions to the Coriolis
force from the tiny ,
x y
v v components which arise due to the
Coriolis forces associated with the much larger . We obtain
z
v

( ) ( )
2 2 2 co
z z
m v m v z m v s y =

.

Acceleration associated with
eff
F
is /
eff eff
a F = m
. Ill drop the

subscript for the time being and write this as /
eff
a F m =

.

We have, collecting the terms,

2
2 2
cos sin
2 cos
cos .
x
y z
z
a x r
a y v
a z g r

= =
= =
= = +

The earths rotation rate is
5
2 rad day 7.29 10 rad sec

= .
This yields
2
5.31 10
9

. The earths radius is

5-42
282
282

6
6.37 10 m r so
2 2 2 2
cos 3.39 10 cos r

( )
2 2
3.39 10 cos z g
. As a result, we
obtain
g = +
g t
( ) ( )
.
Because of this,
z
v = which yields
2
1
2
z t z 0 . g t
2 cos 2 cos
z
y v

The y equation is
g t = = +
( )
,
which is easy to integrate, yielding
3
cos
3
g
y t t

.

The x equation is
2
cos sin x r = +
(constant acceleration!) so
( ) ( )
2 2 2
1 1
cos sin sin 2
2 4
x r t r
2
t = + =

Dropped from a height h, our rock will land when 2 t h g so


5-43
283
283

( )
( )
2
3
3
2
3
5
2
sin 2
1.73 10 sin 2
2
cos 2
2.2 10 cos
3
r h
x h
g
g h
y h
g

where were working in meters.

For 100 meters, h = 17.3 cm, x = 1.55 cm y = at 45 N.

Foucault pendulum analysis
Lets do one last example. Imagine we build a Foucault pendulum
that hangs from a very long rod of length , with 0 x y = = its
equilibrium , x y position.


5-44
284
284

5-45
v

mg
x
z
y
T

The net effective force on the pendulums mass is
2
eff
F mg z T m = +

where Ive picked to include the small effect associated with
centrifugal forces, and am ignoring the small difference between
g
z
and the direction of the sum of the gravitational and centrifugal
forces. T is the tension in the rod holding the pendulum mass, of
course.

If , 0,0 x y = is the equilibrium position for the pendulum, then

restoring
y x
F Ty =

Tx
for small amplitude swings.


285
285

5-46
The earths rotation frequency is cos sin x z = +
so

cos sin sin
y x
v v z v y
y
v x +

I am assuming the pendulums support is so long compared to the
maximum displacement from equilibrium that I can ignore the z
component of the pendulums velocity, which will be proportional
the
( )
2
1 cos 2 for the angle between the pendulums
support and the z axis.

The accelerations in and x y are, consequently,
2 sin
2 sin
x y
y x
xT
a v
m
yT
a v
m

= +

Coupled equations! We can deal with this by adding
x y
a ia + :

( ) ( )
2 sin
x y x
T
a ia x iy v iv
m i
y

+ = + + +

Let . We can rewrite the last equation as q x iy +

286
286

5-47
2 sin or 2 sin
T T
q q i q q i q q
m m

= + + =

0
since 1i i = .

Try a solution of the form to find
Bt
q Ae =

2
2 sin
T
B i B
m
0 + + =

for
Bt
Ae to be a solution. We can deal with this using the
quadratic formula:

2 2
4
2 sin 4 sin
2
T
i
m
B

=

or
2 2
sin sin
T
B i i
m
= +
.

As a result, we obtain
2 2 2 2
sin sin
sin
T T
i t i t
i t
m m
q e Ae Be

+ + +

= +

.


287
287

5-48
Now impose initial conditions. Say the pendulum is swinging in
the x z plane
( )
. Since 0 at 0 y t = =
( )
, 0 q x iy q = + must be
real since the initial swing is only in the x z plane. The only way
to satisfy this is to have A B = .

Rewrite q:

sin 2 2
2 cos sin
i t
T
q x iy Ae t
m

= + = +

since 2cos
iB iB
e e
+ = .

At , so all the motions in the x direction, since q
is entirely real. By the time
sin
0, 1
i t
t e

= =
sin
2
t

= we find that q has
become entirely imaginary, so the plane of swinging has rotated
into the y z plane. Consequently,
2 sin
t

corresponds to
the time it takes for the plane of the Foucault pendulums swing to
rotate through 90 degrees.


288
288

5-49
One full rotation of the plane of swing through 360
o
requires an
amount of time thats
2
sin

.

Since refers to the earths rotation frequency,
2
1 day
= .

We find that the period for pendulum planes rotation is
1 day
sin
.


289
289

5-50

290
290

5-51

291
291


5-52

292
292
An introduction to fluid dynamics
An introduction to fluid dynamics...............................................6-1
Utility of conservation laws in fluid dynamics.........................6-2
Partial derivatives and convective derivatives..........................6-4
Conservation of mass................................................................6-8
Current density and the mass in a macroscopic volume.........6-15
Density changes in a comoving frame................................6-16
(Ir)rotational flow...................................................................6-18
Equations of motion for an ideal fluid....................................6-22
Shear stress in a Newtonian fluid........................................6-27
Conservation laws...................................................................6-33
Energy conservation...............................................................6-36
Stress tensor............................................................................6-43
Including viscosity; the Navier-Stokes equations...................6-47
Water flowing through a long, cylindrical pipe......................6-48
Summary.................................................................................6-53

Im going to follow the development of the subject as presented in
Mechanics, 3
rd
edition, K. R. Symon, Addison-Wesley Publishing,
1971. See chapter 8, sections 6-9.

Another good reference is Lectures in Elementary Fluid Dynamics:
Physics, Mathematics and Applications, J .M. McDonough,
Departments of Mechanical Engineering and Mathematics,
University of Kentucky, Lexington, KY (2009):
http://www.engr.uky.edu/~acfd/me330-lctrs.pdf .
copyright 2012 George Gollin 6-1 PHY_325_lec_notes_06.doc

293
293

Utility of conservation laws in fluid dynamics
Many (most ??) of the useful equations in fluid dynamics come
about because of various conservation laws. Well only deal with
non-relativistic fluids, so well always have one set of equations
which comes about because mass is conserved; if the density in a
fluid at some point in space increases/decreases, it must be
associated with a net inflow/outflow of stuff from that point in
space.

Schematically:
in out
dV dV
dm decreases dm dV =

Other conservation laws which might be useful:


294
294
Perhaps momentum conservation? Wed want to work out a
way to say something like this: momentum inside the
volume element dV can only change when (more/less)
momentum enters the box than leaves it.

Maybe conservation of energy? This ones more
complicated since if you compress some collection of
particles in the fluid, you do work on them, increasing their
potential energy. Also, if the fluid flows uphill (against the
earths gravitational field), its potential energy changes.

An added piece of complication comes about because there are two
kinds of derivatives that well be interested in. We might want to
know how, for example, the pressure at a fixed point changes with
time. But we also might want to know about the rate of change of
the pressure at a point which moves along with the fluid. For
example, a chunk of air moving over an airfoil will show
condensation fog if it is humid and the pressure drops suddenly,
cooling the air.


295
295
Partial derivatives and convective derivatives
This is a natural place to consider the difference between partial
and total derivatives.

Laminar (non-turbulent, layered) air flow over a wing might
look like this:
B A

At the point A, fixed to remain in front of the airfoil, the pressure
will be constant in time. The same is true at point B. Since were
talking about holding x, y, z constant, if the pressure is a function
of x, y, z, t (lets call it P(x, y, z, t) ), we can write

point A
0
P
t
and
point B
0
P
t
.

Since P is a function of four variables, we need to specify (by
taking a partial derivative) that all but t are being held fixed in our
description of the behavior of the pressure at points A and B.


296
296
A group of air molecules in a small volume element that flows
from point A to point B will experience a changing pressure since
the local air pressure goes up as the air gets crowded in front of
the air foils leading edge then drops as the air moves over the
upper surface of the wing. We would like to be able to describe
this kind of thing too. Lets investigate.

From the chain rule, the total derivative of P is

dP P P dx P dy P dz
dt t x dt y dt z dt

= + + +

.

Note the partials of P, but the total derivatives for x, y, z with
respect to time. What were doing, in effect, is declaring how we
want to move around in space by saying well cruise around with
the same velocity as the average fluid velocity of a particular tiny
fluid packet. Well then figure out how P changes with time in the
packet which moves with (changing) velocity

dx dy dz
v x y
dt dt dt
= + +
z .

Recall:
P P P
x y z
x y z

+ + =

P
(see the math review)


297
297

As a result, we can write
dP P
v P
dt t
= +
.

This is handy: it lets us say how the pressure changes if we move
around with velocity v inside the fluid volume. Sometimes
(usually??) well choose the velocity to be the same as the local
fluid flow velocity. Symbolically, we can write

.
d
v
dt t
= +

This quantity is nothing more than the total (time) derivative
operator for a function that depends on x, y, z, and t. It is
sometimes referred to as the convective derivative, Stokes
derivative, Lagrangian derivative, or any one of a number of
other names.

More on this: imagine were in a submarine, measuring the water
temperature T(x,y,z,t) as we motor along. If we come to a stop and
measure how the temperature changes with time, well be

298
298
determining the partial derivative of T: T t since were holding
x, y, z fixed by halting the submarine.

If we dont stop our submarine, the rate of change of the
temperature will depend on how much the temperature varies from
place to place (and how quickly were moving around), as well as
any built-in time dependence, for example, from the sun heating
the ocean during the day.

Taking both the built-in and position-related effects into account,
well measure
( ) ( )
( )
( )
, , , , , ,
, , ,
dT x y z t T x y z t
v T x y z t
dt t
= +

with v the velocity of our submarine.

We might want our submarine to cruise along with the same
velocity as the water around us (in that case, well get to observe
the behavior of the temperature of the same water molecules all
day long), but we arent required to do this for the above equation
to hold.

If the sun isnt shining (so theres nothing heating the water) and
the ocean current flows steadily, without change (so we dont

299
299
suddenly have cold water from the bottom blasting past us), well
expect the temperature at a fixed point in the ocean to remain
unchanged. In that case, 0 T t = so that well only sense a
change in the temperature if we change our position:
( )
( )
, , ,
, , ,
dT x y z t
v T x y z t
dt
=
.

Conservation of mass
Lets work up a differential equation that expresses the idea that
mass is neither created nor destroyed in our fluid.

We will investigate the inflow/outflow of mass in a small volume
at the point x, y, z. (Im only going to draw it in two
dimensions, to simplify the picture.)
dV dxdydz =


300
300

Ive drawn the fluid flow so that the fluids velocity into the box
from the bottom and left sides is greater than the velocity out of the
box through the top and right sides. As a result, we expect fluid to
build up in the box, so the density should increase.

We can take as the average fluid velocity for the fluid that enters
the left side of the box the exact fluid velocity at the center of the
left side:
, ,
2 2
left
y z
v v x y z

+ +

.
The volume of fluid that flows into the box from the left side
during time dt is shown as a shaded region in the following
diagram.

x
y
y
x

301
301
x
y
left
v xdt

The volume of the region of fluid that flows into the left side of the
box is approximately

, ,
2 2
left
y z
v xdt y z v x y z xdt y z

+ +

since the thickness of it in the x direction is v dt x
.

The volume that flows in from the bottom is
, ,
2 2
x z
v x y z ydt x z

+ +

.

The volume that flows out the top is

302
302
, ,
2 2
x z
v x y y z ydt x z

+ + +

.

The volume that exits the right side is
, ,
2 2
y z
v x x y z xdt y z

+ + +

.

The mass which flows in from the left is the product of the local
density and the in-flowing volume:

, , , , , ,
2 2 2 2
left
y z y z
dm x y z t v x y z t xdt y z

+ + + +

Note that Im approximating density and velocity for the left side
using the value of and v

at the center of the left face of the
volume element.

Adding up the in-flowing and out-flowing mass for all six faces of
the volume element (and omitting explicit time dependence, to
save myself the effort of writing , t everywhere) gives


303
303
, , , ,
2 2 2 2
, , , ,
2 2 2 2
, , , ,
2 2 2 2
, , , ,
2 2 2 2
x
x
y
y
y z y z
dm x y z v x y z dt y z
y z y z
x x y z v x x y z dt y z
x z x z
x y z v x y z dt x z
x z x z
x y y z v x y y z

= + + + +

+ + + + + +

+ + + + +

+ + + + + +

, , , ,
2 2 2 2
, , , ,
2 2 2 2
z
z
dt x z
x y x y
x z v x z dt x y
x y x y
x z z v x z z dt x

+ + + + +

+ + + + + +

y

Note that the difference of the first two terms can be rewritten:

( ) ( )
, , , ,
2 2 2 2
, , , ,
2 2 2 2
, , , , , ,
.
x
x
x
y z y z
x y z v x y z dt y z
y z y z
x x y z v x x y z dt y z
x y z t v x y z t
dtdxdydz
x

+ + + +

+ + + + + +

=

We can write similar expressions for the other two pairs of terms to
conclude

304
304
[ ] [ ]
( )
.
y
x z
v
v v
dm dtdxdydz
x y z
dtdxdydz v

= + +

=

Recall that the mass inside the small box with volume
is dV dxdydz =
( )
, , , m x y z t dxdydz. =

Since were evaluating things for a stationary volume element at
the point x, y, z with side lengths dx, dy, dz we can say that the
total time derivative of the mass inside the volume is

( )
( )
fixed
, , , , ,
, , ,
, , ,
x y z dx dy dz
d x y z t dxdydz
x y z t
dm
dxdydz
dt dt t

= =

because thats what we mean by taking a partial derivative: we
hold all the variables except one fixed.

But we also know that
( )
dm dtdxdydz v =

so it must be true
that
( )
dm
v dxdydz
dt
=

.


305
305
As a result, we can write

( )
( )
, , , x y z t
dm
dxdydz v dxdydz
dt t
= =

so that
( )
( )
, , , x y z t
v
t

or
( )
( )
, , ,
0 (fixed , , )
x y z t
v x
t
y z
+ =

This equation expresses the fact that mass is conserved in our fluid,
even though density and velocity of flow can change with time.
Keep in mind that I worked this up using a volume element dV that
was fixed in space: with x, y, z held constant, our time derivative is
a partial derivative t , not a total derivative d dt .


306
306
Current density and the mass in a macroscopic volume
We can describe the change in mass contained in a macroscopic
volume V by integrating the above expression:
( )
, , ,
V
M x y z t dV =

so
( )
( )
( )
, , ,
, , ,
V
V V
dM d
x y z t dV
dt dt
x y z t
dV v dV
t
= =

Are you familiar with the divergence theorem? If so, youll recall
that, for any vector field
( )
, , , J x y z t
,
( ) ( )

, , , , , ,
closed surface volume enclosed
J x y z t dA J x y z t dV =

.

By identifying we can rewrite v

J
( )
V
v dV

as
and conclude that
A
v dA
( )
V A
v dV
dM
v dA
dt
= =

.

The quantity v
is the mass current density flowing in the fluid.


307
307
The integral over a closed surface of the current density flowing
through the surface tells us the rate at which mass is entering (or
leaving) the volume.

Density changes in a comoving frame
If we wanted to discuss the rate of change of density of the fluid as
we move along with it, we could use the expression for the
convective derivative from some pages back:
d
v
dt t
= +
.

As long as we plug in the velocity v
that corresponds to the

(moving) point in the fluid were observing, well learn something
useful.

We want to calculate an expression for the total derivative d dt ,
so lets use the connection between convective and partial
derivatives, above:
d
v
dt t

= +
. [1]

308
308
Weve already figured out how the partial time derivative of
(when were holding x, y, z fixed) behaves:
( )
v
t

. [2]
It comes from conservation of mass: if the density is decreasing at
a particular point, there has to be an outflow of matter from that
point. (The outflow is what the divergence is telling us about.)

Use the result in [2] to replace the first term to the right of the =
in [1]:
( )
(
d
v v
dt
)
= +

[3]
or
( )
( )
0
d
v v
dt
+ =

.

Now,
( ) ( ) ( ) ( )
( )
x y z
y
x z
x y z
v v v v
x y z
dv
dv dv d d
v v
dx dx dy dy dz dz
v v

= + +

= + + + + +

= +

d
v


309
309
so I can rewrite
( )
d
v v
dt
= +

(this was equation [3])
as

( )
d
v v v
dt
v
= +
=

or
0 (moving along with fluid)
d
v
dt
+ =

. [4]

Thats what we wanted: as we cruise along with the fluid, well see
the density changing in accord with this equation. The physical
meaning is straightforward: if we see atoms in our fluid streaming
out from a point (so that the divergence of the velocity is non-
zero), well expect to see the fluids density at that point changing
with time.

(Ir)rotational flow

Remember about curl? Look over material in the lecture notes on
conservative forces for a refresher, if necessary.

310
310

Lets say the velocity in the fluid near a vortex looks like this:

To have something concrete to work with, lets say the velocity at
a point
( )
, r is
( )
v v r =
where 0 r = is the center of the vortex.

In Cartesian coordinates, we could write

( )
( )
2 2
cos sin v v x y y x = +
.

If you look up the form for curl in cylindrical coordinates youll
find that its

1 1

z r z r
A A A A A A A
A r z
r z z r r r

= + + +

r
.

As a result, since , we have
( )
v v r
=


311
311
( )
( )
v r
v v r
r r

= +

z

because . 0
r z
v v = =

A circular flow pattern like this has non-zero curl.

If the layers are all turning with the same angular velocity, therell
be no shearing between adjacent layers. This means that one
layer in the flow does not slide past another, and there will be no
energy loss associated with viscous drag that one layer will exert
on another.

If all layers turn with the same angular velocity , well have
( )
v r r = which will yield
( )
( )
2
v r
v v r
r r

= + = + =

,
independent of r.

If depends on r, therell be shear dislocations of adjacent
layers relative to each other, which will lead to energy dissipation
in a viscous fluid.
v


312
312

If , the flow is irrotational: a small pinwheel placed in
the fluid wont spin.
0 v =

We can be a little more formal than using the pinwheel analogy by
referring to Stokes theorem:

( )
loop area
v ds v dA =

.

If the velocity seems to go around in circles so that 0
loop
v ds
,
well automatically have non-zero curl inside the loop.

Itll also be true that fluid flow with a velocity gradientfor
example, having v
x
increase with ycan correspond to rotational
flow, even if all molecules in the fluid are traveling in the x
direction.


313
313
Equations of motion for an ideal fluid
Lets assume for the moment that were working with a fluid that
has zero viscosity, so that one layer of fluid sliding past another
layer of fluid does not exert a force on the other layer, attempting
to drag it along. In the language of fluid mechanics, there is no
shear stress exerted by one layer moving across another layer.

Shear stress is defined as the force per unit area that one layer
exerts on another layer. Ideal fluids do not support shear
stresses.

The net force acting on a small volume element can come from
two sources: a body force (for example, gravity) and a pressure
gradient that makes the force on one side of the volume element
different from the force on the opposite side of the volume
element.

Lets say the volume is dV dxdydz = for a small rectangular solid
in the fluid. The areas of the six faces are dxdy, dydz, and so forth.


314
314

The force associated with pressure on the face centered at
, ,
2 2
dy dz
x y z

+ +

is , ,
2 2
dy dz
P x y z dydz
+ +
since force is the

product of pressure and area.

The net force associated with pressure in the x direction is
, , , ,
2 2 2 2
.
dy dz dy dz
P x y z dydz P x dx y z dydz
P
dxdydz
x

+ + + + +

since the pressure on the right-side face creates a push to the left.

x
y
, ,
2 2
dy dz
x y z

+ +

z
, ,
2 2
dy dz
x dx y z

+ + +


315
315
Consequently, we can write the net force, associated with changes
in pressure across the volume element, as

( )
y z
.
gradient
P P P
F x dxdydz
x y z
P dV

=

=

If the gravitational force acting on dV is to be included, well have
an additional piece to incorporate which is
( )
mg dV g =

.

The equation of motion for the volume element dV is just F ma =

,
or
( )
F dV dv =

dt so
( )
( )
( )
+
dv
dV P dV dV g
dt
=

Note that v refers to the velocity of our tiny volume of fluid, since
were referring to how this velocity changes when there are
unequal pressures on opposite sides of the small volume element.

We can rewrite the last equation as
dv
P g
dt
+ =

, or


316
316
dv P
g
dt
+ =
.
Keep in mind that the pressure P appears in this equation because
gradients in pressure will cause unequal forces to be exerted on
opposite sides of our tiny volume, causing it to accelerate.

The equation lets us relate density, the pressure gradient, and the
acceleration of a packet of fluid as it moves along. Its just telling
us that force is mass times acceleration, nothing more. If the fluid
is viscous, therell be additional terms associated with viscous
forces acting on our small packet of fluid.

We can write a new version of this describing what happens at a
fixed point (as opposed to what happens to a particular group of
molecules in the tiny volume that flows from place to place) by
using our convective derivative since thatll let us get to . v t

Use
d
v
dt t
= +
to rewrite the previous equation as

.
v P
v v g
t

+ + =


317
317
Now we have an equation with a partial derivative. In the case that
the flow is steady, so that the behavior of the (moving) fluid at one
point in space doesnt change, well have . v v P g + =

Heres what I mean by the curious term v v

(curious because we
appear to be taking the gradient of a vector, instead of a scalar). It
is perhaps more clearly expressed as
( )
v v

:

( )
.
x y z
x x x
x y z
y y y
x y z
z z z
x y z
v v v v v v
x y z
v v v
v v v x
x y z
v v v
v v v y
x y z
v v v
v v v z
x y z

= + +

= + +

+ + +

+ + +

The equation
v P
v v
t

+ + =

v
is referred to as Eulers
equation of motion for a moving fluid subject to a gravitational
force. It is nonlinear due to the presence of the v

term.


318
318

Shear stress in a Newtonian fluid
Imagine you drag a sheet of plywood of area A across the floor,
using a spring balance to measure the force F necessary to
overcome friction. The force per unit area required to move the
plywood is F/A, of course; we refer to this as the stress caused by
the interaction between the plywood and the floor.

Imagine instead that we have a layer of fluid with non-zero
viscosity between a (rough) plate of area A and the rough bottom
of a large tank, and drag the plate with velocity
plate
v v x =
across
the top of the fluid, as shown in the figure.

fluid-filled tank with rough bottom
rough plate, area A,
floating on surface of fluid
d is depth of fluid
v
x
y
z
d is depth of fluid
v
d is depth of fluid
v
x
y
z
x
y
z

If the bottom of the tank and the surface of the plate that is in
contact with the fluid are both sufficiently rough, the layer of fluid

319
319
in contact with the surfaces should move with the same velocity as
the surfaces. The fluid at y =d immediately below the plate will
move to the right with velocity
plate
v x while the fluid at the bottom
of the tank will remain at rest.

In a Newtonian fluid, the fluid velocity below the moving plate
increases linearly from 0 to
plate
v x as y increases from 0 to d. We
can write
( )
plate
, ,
.
x
v v x y z
y d

A Newtonian fluid is viscous (otherwise thered be no shear forces
to make the fluid move parallel to the plate), and will exert a stress
(force per unit area) on the dragged plate that is proportional to
how fast were pulling it, among other things.

(Liquid) water behaves like a Newtonian fluid. Examples of non-
Newtonian fluids include Silly Putty and chilled caramel ice cream
toppings: stress them hard enough and theyll behave like solids.
(Silly Putty will shatter, for example, when the applied force
comes from a hammer.)


320
320
The thicker the layer of fluid, the less the force needed to drive the
plate with constant speed in opposition to the viscous drag exerted
by the fluid.

If we can assume that the fluid at the bottom of the tank has zero
velocity, while the fluid in contact with the bottom of the plate has
the same velocity as the plate, the required force will be
proportional to 1/d for a Newtonian fluid. As a result, the shear
stress on the plate (and on the bottom of the tank) has magnitude
plate
.
x
v
v F
A d y

= =

The constant is an intrinsic property of the fluid, and is called the
fluids viscosity.

Its not going to be true in general that the fluid velocity increases
linearly with distance transverse to the direction of flow: thats
only the case for the example of a plate being dragged above a
rough surface.

Imagine weve set up a steady flow in a fluid: theres no explicit
time dependence to the fluid velocity so we can write it as
Lets look at how viscous forces are exerted on one
small volume element by adjacent volume elements. Lets make
(
, , . v x y z
)

321
321
them all the same size, and assume that the general direction of
flow is along x, and that the flow speed increases (but not
necessarily linearly) with y.

Viscous drag from the bottom volume element creates shear stress
that tries to slow the middle element down; shear stress from the
upper volume element will try to speed the middle element up. The
shear stresses will also distort the volume element; well look at
whats going on at the instant when the collection of molecules in
the middle volume happens to form a rectangular solid.

x
y
z
dy
dz
dx
( )
, ,
x
v v x y dy z x =
( )
, ,
x
v v x y z x =
( )
, ,
x
v v x y dy z x = +
x
y
z
x
y
z
dy
dz
dx
( )
, ,
x
v v x y dy z x =
( )
, ,
x
v v x y z x =
( )
, ,
x
v v x y dy z x = +

The net (shear) force on the middle element will be


322
322
upper lower
surface surface
2 2
2 2
.
x x x
x
x x
dv dv dv d
F dxdz dydxdz
dy dy dy dy
d v d v
dxdydz dV
dy dy

=

= =

If we allow for a similar velocity gradient in v
x
in the z direction,
well have an additional contribution to the force on our middle
volume element. As a result, we expect the force from viscosity on
our volume element to be
2 2
2 2
.
x x
x
d v d v
F d
dy dz

= +

V

There are also viscous effects that act along the direction of motion
of the fluid: if you've ever poured honey into a cup of tea, you've
seen this. Even though gravity is pulling honey off your spoon, the
viscosity of the honey keeps it from going into freefall.

x
y
z
( )
, ,
x
v v x y z dz x =
( )
, ,
x
v v x y z x =
( )
, ,
x
v v x y z dz x = +
x
y
z
x
y
z
( )
, ,
x
v v x y z dz x =
( )
, ,
x
v v x y z x =
( )
, ,
x
v v x y z dz x = +


323
323

The volume element to the left exerts a pull in the negative x
direction, the volume to the right a pull in the positive x direction.
In a Newtonian fluid, the pull along the direction of motion
associated with viscosity (this is distinct from a pull associated
with a pressure gradient) will also be proportional to the derivative
of the velocity:
.
x x
F v
A x

I wont derive it, but the proportionality constant in a Newtonian
fluid is the same viscosity constant as for shear forces.

The net force on the volume from the non-zero
x
v x is
right left
surface surface
2 2
2 2
.
x x x
x
x x
dv dv dv d
F dydz dxdydz
dx dx dx dx
d v d v
dxdydz dV
dx dx

=

= =

As a result, the full expression for the viscous force is nicely
symmetric:
( )
2 2 2
2
2 2 2
.
x x x
x x
d v d v d v
F dV
dx dy dz

= + + =

dV v

324
324
Including the effects of possible flow velocities along y and z gives

( )
2 2 2
.
x y z
F dV x v y v z v dV = + + =

2
v
Conservation laws
We worked up the equations some pages back to express the fact
that mass is conserved in our fluid.

Lets try for a conservation-of-momentum equation now. Start
with this equation that I had derived for a non-viscous fluid:

.
dv P
g
dt
+ =

This describes the relationship between velocity, pressure, density,
and the gravitational acceleration for a point moving along with the
fluid, and not at a fixed x, y, z. Thats because we have a total
derivative dv dt
, not a partial derivative, v t
, as had been used

in Eulers equation of motion.

For a particular set of molecules in the fluid (we tag them, and
keep track of them), the volume they occupy will increase when

325
325
the density decreases, and decrease when the density increases
The mass of this particular small set o molecules i
occupy volume dV and have density
.
f s dm; they
. We have dm dV = ;
since were keeping track of what these particular molecules are
doing as they flow along, dm will be constant as long as we dont
ose track of any of them. As a result, dV is constant. lo

Multiply our equation
dv P
g
dt
+ =
by dV to write

( )
dv
dV P dV gdV
dt
+ =

.
Because dV is constant, we can put it inside the time
derivative:
( )
( )
.
d
dt
vdV g P dV =

Note that vdV

s wer
is the momentum carried by the particular set of
olecule e watching. Keep in mind that both m and
If we integrate over a finite volume, we can write

dV
change as the fluid flows.


326
326
( )
volume volume
dt

d
vdV g P dV

=

.
I mentioned the divergence theorem some pages ago:

( ) ( )
, , , , , ,
closed surface volume enclosed
J x y z t dA J x y z t dV =

.
Theres a generalized version of this that says:

volume surface
PdV PdA

=

.

As a result, we can rewrite the equation
( )
volume volume
d
vdV g P dV
dt

=

as
volume volume surface
dt

. vdV gdV PdA
=

d


327
327
surface
PdA

is just the net force acting on a macroscopic volume of
fluid associated with changes in pressure over the surface of the
fluid.
This equation serves as a statement about momentum conservation

Energy conservation
ets develop an energy conservation equation, again assuming the
viscosity

in our fluid.

L
is zero. Start with : F ma =

( )
( )
( ) ( )
dV g P dV g P dV
dt
d dV v
= =

. [1]

( )
dV v

is the momentum of our volume of fluid dV
( )
dV g

is the gravitational force acting on dV
( )
P dV
is the force acting on dV due to a pressure gradient


328
328
Recall that
( )
2
d dv dv dv
v v v
dt dt dt dt

= + =

v v

so that

2
1
2 dt dt

dv d
v

=

.

Also recall that
v
dV for our parcel of fluid is constant: the larger

the density, the smaller the volume it occupies as it moves along.
Take the dot product of the first equation ([1]) in this ection with
to write

s
v
( )
( )
d dV v
v g P dV v
dt

or
2
1
2
d
v dV v gdV v PdV
dt

=

.

What weve done is taken F ma =

and
dotted it with the velocity;

call that is the pow e applied forces are feeding into the
e rate of change of the kinetic energy of our parcel of fluid
equals the power pumped into it by gravity and the driving force
associated with the pressure gradient.

F v

er th re
system. Not surprisingly, the equation we have now tells us that
th

329
329
Lets try to replace the second term on the right side. Note that

( ) ( )
( )
( )
d dP d
PdV dV P dV
dt dt dt
P d

dV v P dV P dV
t dt
= +
= + +

(thats just our convective derivative coming into play).
The term
( )
d
dV
dt
in
( )
d
P dV
dt
refers to the rate-of-change of the
olume occupied by the particular ensemble of molecules were
ng as they cruise along in the fluid flow.

Since
v
watchi
, dV dxdydz =
( ) ( )
( ) ( ) ( )
.
d d
dV dxdydz
dt dt
d d d
dx dydz dy dxdz dz dxdy
dt dt dt

Now, dx is the t ickness of the volume element in the
=

= + +

h x
direction. The rate at which it changes is proportional to e th
difference in the x component of velocity of the right and left sides
of the volume element. Since

330
330
x
right left
v
v v

dx
x
=

e have w
( )
x
dv d
dx = dx
dt x
.
e have similar results for dy and dz.

As a result,

W

( )
( )
dydz v dV
dt x y z
=

y
x z
dv
dv dv d
dV dx

= + +

.

can use this to rewrite the equation We
( )
( )
( )
d P d
PdV dV v P dV P dV
dt t dt
= + +

as
( )
(
d P
) ( )
PdV dV v
= + P dV P v dV
dt t
+

.

e have, after rearranging terms, W
( )
( )
( )
P
V dV P v dV
dt t
d
v P dV Pd

= + +

.


331
331
Use this to rewrite the equation
2
1
2
d
v dV v gdV v PdV
dt

=

from several pages back as

( )
( )
2
1
2
d d P
v dV v gdV PdV dV P v dV
dt dt t

= + +

or
( )
( )
2
1
.
2
d P
v dV PdV v g dV dV PdV v
dt

t

+ = + +

[1]

Lets define the potential energy per unit mass associated with the
esence of gravity as pr
( )
U r
so that
( )
g U = r

.
ith this definition,
( )
v g v U r =

W .

Now (convective derivatives to the rescue!),

dU U U dx U dy y dz
dt t x dt y dt z dt
U
v U
t

= + + +
= +


332
332
So
dU U
v U
dt t
and also, from before, v U v g =

.
U is time-independent when x, y, z are held fixed (neither the
earths mass nor Newtons gravitational onstant are changing!)
o
c
0
U
t
. As a result, s
dU
v g =
dt

and we can rewrite our previous equation ([1]) as

( )
2
1 d P
v dV UdV PdV dV P v dV
2 dt t

+ + = +

.

we have steady-state flow, the pressure at any particular x, y, z If
point is constant, so 0
P
t
.

Lets define the potential energy of the volume element with mass
dm associated with compression of the volume element by
l
y per unit mass, and its a Roman w, not a Greek omega.)

pressure from the surrounding fluid to be wdm. (w is the potentia
energ

333
333
With some work, we can show for our steady-state flowing fluid
that
( )
( )
d
wdm P v dV
dt
=

.
Consequently, we can rewrite our earlier equation as

2
1
0
2
d
v dV UdV PdV wdV
dt

+ + + =

since dm dV = .

Using constant dV dm = = , we can rewrite the above as
[ ]
2
1
0
2
d P
dm v U w
dt

+ + + =

or
2
1 P
+ + + = constant v U w
2
.

For an incompressible fluid (e.g., water), w is onstant so we have

c
2
1
constant
2

P
v U + + . =

334
334
This is called Bernoullis equation; its why airplanes fly. As v
increases, P decreases. Thus lift is generated by airfoils as the
reater speed across the curved, upper surface is accompanied by
If the fluid flow does not go uphill or downhill so that U is
constant, Bernoullis equation simplifies further:
g
decreased pressure.

2
1
2
P
+ =
Keep in mind how we got to this point: we were investigating how
a small volume of fluid accelerated due to the existence of a
pressure gradient. Modification of those equations to tell us how
the kinetic energy of the small volume of fluid changed thanks to
the power applied by the pressure gradient to ou
constant. v

r volume gave us
ernoullis equation. It says, in effect, that the pressure will drop
as the fluid expends e

e talked about the elastic tensor during the oscillation unit. (See
material beginning around p. 4-35.)

B
nergy speeding itself up.

Stress tensor
W

335
335
When I described viscous forces that produce shear strains, I
concluded that F
x
contained contributions from the second
derivatives of v
x
.

The stress tensor is something we can use to streamline some of
our notation in fluid dynamics. Im going to follow the general
approach in J .M. McDonoughs paper, cited at the beginning of
this section of my lecture notes.

Lets define something called the viscous stress tensor this way:
xx xy xz
yx yy yz
zx zy zz

=

with
2 , 2 , 2
y
x z
xx yy zz
v
v v
x y z

= = =

and
, ,
.
y y
x z
xy yx yz zy
x z
zx xz
v v
v v
y x z y
v v
x z

= = + = = +

= = +


336
336
Im not going to prove this, but it is not hard to show that

transforms like a tensor under rotations, just as was true for the
elastic tensor.

Heres what happens when we take the divergence of .
This will
look a little strange to you since youre used to seeing the
divergence of a vector field, which produces a scalar. Here, were
taking the divergence of a rank-2 tensor to produce a vector.

.
xx xy xz
yx yy yz
zx zy zz
yx xy yy zy
xx zx
yz
xz zz
x y z
x y
x y z x y z
z
x y z

=

= + + + + +

+ + +

Take a closer look at the x component.
2
2 2 2 2
2 2
2
2 2 2 2 2
2 2 2 2
2
.
yx y
2
xx zx x x z
y
x x x x z
v
v v v
x
v
x y z x y y x z x z
v
v v v v v
x y z x y x z x

+ + = + +

= + + + +


337
337

We can rewrite the second term:
( )
2
2 2
2
.
y y
x x z z
v
v v
v v v v
x y x z x x x y z x

+ + = + + =

Recall that conservation of mass gave us
( )
( )
, , ,
.
x y z t
v
t

For an incompressible fluid like water, the density is constant so
zeroing out this term. 0, v =

As a result, for an incompressible fluid, the x component of

is
( )
2 2 2
2 2 2
x x x
v v v
x
x y z

= + +

.

When I worked out the viscous forces on a volume element some
pages ago, I concluded that

2 2 2
2 2 2
.
x x x
x
d v d v d v
F d
dx dy dz

= + +

V


338
338
So the divergence of the viscous stress tensor is proportional to the
force acting on our volume element! We can express the vector
version of the force (not just the x component) that acts on our
volume element this way:
( )
viscous
. F d = V

I had already found that
( )
2
F dV v =

so we have
( )
( )
2
viscous
. F dV d = =

V v

Including viscosity; the Navier-Stokes equations
I developed a description the rate of change of momentum for a
particular group of fluid molecules some pages ago for a non-
viscous fluid. Lets include viscous forces now.

( )
( )
( ) ( )
d dV v
dV g P dV dV
dt
.
For an incompressible fluid (water!) both and dV are constant for
our small volume element. In this case,
.
dv
g P
dt
=

Here, v
is the momentum density (momentum per unit volume)

carried by the fluid; g

is the gravitational force per unit volume;

339
339

P
is the force per unit volume due to a pressure gradient;

is the force per unit volume associated with viscosity.
Using our convective derivative we can rewrite the last equation as

.
v
v v P g
t

+ = +

This is called the Navier-Stokes Equation. Sometimes the
continuity equation
( )
( )
, , ,
0
x y z t
v
t
+ =

is included as
another of the Navier-Stokes equations.

Water flowing through a long, cylindrical pipe
This is a good example of the power of the Navier-Stokes
equations. Im going to follow the development of the subject in
J .M. McDonoughs lecture notes, referenced at the beginning of
this unit.

Something I forgot to mention: to good accuracy, the layer of a
viscous fluid that is in contact with a surface is always moving at
exactly the same velocity as the surface. This is called th
condition.
e no-slip
340
340

We need to rewrite the Navier-Stokes equations in cylindrical
oordinates. Thats not a big deal; looking it up (rather than
quation:
c
grinding it out myself) yields this:

Continuity (conservation of mass) e
( ) ( ) ( ) 1 1
0
r z
rv v v
z
t r r r

+ + + =

Equation of motion for r component of momentum:

( )
2
1 1
r r r r
r z
t r r r z

rr
r rz
r
v v v v v v
v v
r
P
F
r r r r r z

+ + + =

+ + +

Equation of motion for z component of momentum:
( ) 1 1
z z z z
r z
t r r z

rz
z zz
z
v v v v v
v v
r
P
F
z r r r z

+ + + =

+ + +

For a Newtonian incompressible fluid the momentum equations
duce to the following: re

341
341

Equation of motion for r component of momentum:
( )
2
2 2
2 2 2 2
1 1 2
r r r r
r z
t r r r z

r
r r
r
v v v v v v
v v
rv
v v v P
F
r r r r r r z

+ + + =

+ + + +

Equation of motion for z component of momentum:
2 2
2 2 2
1 1
z z z z
r z
t r r z

z z z
z
v v v v v
v v
v v v P
r F
z r r r r z

+ + + =

+ + + +

Due to the cylindrical symmetry of the pipe there cannot be a
ependence to anything. If there are no gravitational effects, F
r
and
the fluid far enough from the entrance of the pipe so
at weve arrived at a place where the flow is steady, v
r
must be

d
F
z
are zero.

If we look at
th
zero since the fluid is not passing out of the walls of the pipe.
Naturally, v
is also zero. Further, the density has no explicit time

dependence. As a result, the continuity equation becomes

342
342
( ) ( ) ( ) ( ) 1 1
0
r z z
z
rv v v v
v
t r r r z z z

so that v
z
can only depend on r, but not on z.

he r momentum equation becomes
+ + + = = =

0 P r = T so that the pressure
: P(z). is independent of r, and can only depend on z

The z momentum equation becomes

( ) ( ) 1
0
z
P z v r
r

= +
z r r r

or
( ) ( )
.
z
P z v
r
z r r r
=

r
te that the left side is only a function of z, while the right side is
only a function of r. This can only be true if both sides are
constant. There is a pressure drop in the pipe, of course, so lets
No
identify the constant as P/L for a total pipe length of L. Note that
P is negative: its final pressure minus initial pressure.

Solve for v
z
now, replacing our partial derivatives with total
derivatives to be able to do the integrals:

343
343
( ) ( )
z z
v r dv r
P P
r d r rdr

= = ;
r r r L dr L

Integrate:
( )
( )
( )
2
1
1
2
2
z
z
z
dv r
P
d r rdr
dr L
dv r
P
r r C
dr L
C P
d v r r dr
L r
= +

= +

or
( ) ( )
2
1 2
ln .
4
z
P
v r r C r C
L
= + +

Two boundary conditions on v
z
: it must be finite at r =0, and it
must be zero at the walls of the pipe, where r =R. The first
quires C
1
=0; the second that
2
2
4 . C PR L = re As a result,
( )
2
2
1 .
4
z
P
r
v r R
L R

=

We are neglecting messy things
2
like turbulence!

The average flow velocity in the pipe is this:

344
344
2 2
2
1
R
P
r
2 2
0 0
2
3 2 4
2 2
0
0
1
4
.
2 2 2 4 8
z
R
R
v R rd dr
R L R
P P P R
r r r
r dr
L R L R L

= = =

he volume of fluid passing through the pipe is just the average

=
T
flow velocity multiplied by the cross sectional area of the pipe:

4
flow rate .
8
P R
L
=
Summary
for fluid dynamics, made with
s
onservation of mass (Continuity Equation)
These are the master equations
certain simplifying assumptions such as constant, homogeneou
viscosity that is independent of flow velocity, pressure,
temperature, and so forth.

C
( )
( )
, , ,
0 (fixed , , )
x y z t
v x y z
t
+ =


345
345

Conservation of momentum (Navier-Stokes Equation)
(Newtonian incompressible fluid)
v
v v P g
t

+ = +

Conservation of Energy (Bernoullis Equation)
2
1
constant (incompressible, non-viscous fluid)
2
P
v U
+ + =

The Navier-Stokes equations are nonlinear, and capable of
producing the chaotic behaviorturbulencethat is observed in
many fluid systems.

There are loads of concepts in fluid dynamics that I have
omittedMach number, Reynolds number, turbulence, supersonic
flow, and so forth.

346
346
Lagrangians and the calculus of variations

Lagrangians and the calculus of variations..................................7-1
Introduction..............................................................................7-2
Definition of the Lagrangian....................................................7-4
The connection between the Lagrangian and Newtons laws...7-5
Using the Lagrangian................................................................7-8
Sliding mass, sliding wedge..................................................7-8
Nasty pendulum...................................................................7-13
Definition of the (classical) action..........................................7-20
Classical action for a free-falling mass...................................7-21
Hamiltons principle...............................................................7-23
Demonstration of the equivalence of Hamiltons principle
and the Euler-Lagrange equations..........................................7-24
Setting up a calculus of variations approach to the
problem................................................................................7-25
Minimizing the action through proper choice of
trajectory..............................................................................7-29
The Euler-Lagrange equations and other minimization
problems..............................................................................7-35
A different minimization problem..........................................7-36
Brachistochrone...................................................................7-37
A few comments.....................................................................7-49
Generalized coordinates.......................................................7-49
Feynmans use of the principle of least action.....................7-51
More Lagrangian examples....................................................7-53
Double Atwoods machine, done two ways........................7-53
Brief recap...........................................................................7-61
Spherical pendulum.............................................................7-63
Conical pendulum................................................................7-70
Small perturbations to the conical pendulum.......................7-71

347
347

A comment on generalized momenta.....................................7-78
The Hamiltonian as a Legendre transform of the
Lagrangian..............................................................................7-81
Lagrangian and Hamiltonian for a particle in a magnetic
field.........................................................................................7-90
Poisson brackets.....................................................................7-93

Introduction
For the moment lets restrict ourselves to the study of systems in
which all the relevant forces can be derived from a potential (and
are therefore conservative) and also lack explicit time dependence.
The force acting on a particle may change, of course, as the
particle moves through a region of space in which the force is not
spatially constant.

Earlier wed sometimes solve for a particles motion by integrating
Newtons (differential) equation F mx =
. And sometimes wed

resort to various tricks, such as the use of conservation of the sum
of kinetic and potential energies to write an equation that linked
position and velocity:
( )
0
x E
2
2 mv U + =
.

Perhaps the approach in which we work directly with the forces is
the most intuitively clear: we have all understood how objects

348
348

move in response to applied forces since we were toddlers. But the
constraints we encounteran object might be constrained to roll
without slipping on a curved surface, for examplecan render it
fiendishly difficult to express the various forces in a form that is
amenable to calculation.

Another example for which the motion in response to an applied
force can be complicated is the behavior of a small object in the
gravitational field of a massive object like the sun. The suns
gravitation exerts an inwards pull along , r but conservation of
angular momentum couples the behavior of r with changes in .

It would be very convenient to develop a technique for generating
the equations relating second derivatives of coordinates with other
kinematic quantities (
x
F mx = is an example of this) without
plunging into a geometrical morass.

Well work on the Euler-Lagrange equations: these will allow us to
generate the desired equations of motion with ease.


349
349

Definition of the Lagrangian
We define a new quantity, the Lagrangian, this way:
L T U = .
The Lagrangian will usually be a function of both position and
velocity so that we can write it as
( )
1 2 1 2
, ,... , , ,...
n n
L q q q q q q where
are the coordinates of the various particles whizzing
around in whatever coordinate system we are using (for example
1 2
, ,...
n
q q q
( )
, , r in spherical coordinates) and are their first
(total) time derivatives:
1 2
, ,...
n
q q q
.
k
dt
k
q q d =

At first glance the Lagrangian is an odd thing. Unlike the total
energy, it changes with time. This is obvious: imagine the time
dependence of L for an object dropped from rest in a constant
gravitational field. Initially T is zero while U is positive. But as it
falls U decreases while T becomes positive. Consequently, L
changes. For our falling mass it is a simple calculation to
determine the value of L as a function of time.


350
350

The connection between the Lagrangian and Newtons laws
For the moment, lets work in Cartesian coordinates. Also keep in
mind that Im still restricting our considerations to conservative
forces. In this case we have a Lagrangian
( )
1 2 1 2
, ,... , , ,...
n n
L q q q q q q
which can be rewritten
( )
,... .
n n
x
1 2 1 2
, ,... , , L x x x x x

In Cartesian coordinates the kinetic energy only depends explicitly
on the components of the velocity
1 2
, ,...
n
x x x . (Note that this is not
true in, for example, cylindrical coordinates where the

component of velocity is r
.)

Lets take a look at the partial derivatives of the Lagrangian.

You should keep in mind the difference between a partial
derivative and a total derivative. When taking a partial derivative,
all but one of the variables
1 2 1 2
, ,... , , ,...
n n
x x x x x x is held fixed. Even
though a free-falling objects position and velocity are changing as

351
351

it falls, we will still have 0 T y = since
2
2 T mv = . It is the total
derivative dT dy that is different from zero.

Since L depends explicitly on
1 2
, ,...
n
x x x and
1 2
, ,...
n
x x x we have
( )
( )

i i
i i
i i
T U
L U
i
x x x
U x F x
F mx

= =

= =
= =

because the kinetic energy does not depend explicitly on position.

I am restricting our discussion to systems in which the forces are
derivable from potentials, and are therefore independent of
velocity. Because of this,
( )
( )
2 2 2
1 2 3 1
2
i i i
i
i
T U
L T
x x x
m x x x
mx
x

= =

+ +
= =

Note that
(from before)
i
i i
d L L
mx
dt x x

= =

so that

352
352

0.
i i
d L L
dt x x

=

This is called the Euler-Lagrange equation. It also happens to be
true when we work in non-Cartesian coordinates, although Im not
going to demonstrate this in class.

I still havent discussed why I bothered to define L as T U rather
than choosing to work with something else, such as : K T U + if I
had done that wed have a sign flip in the Euler-Lagrange
equations, but it would still be useful for us as an equation-of-
motion generator, involving derivatives of K instead of L.

Theres a good reason for working with L instead of something
else (for example T +U) because of the connection between the
Lagrangian and the classical action associated with the systems
motions. More on that in a while.


353
353

Using the Lagrangian
Sliding mass, sliding wedge
Lets do an example.

Do you recall the annoying problem from a Physics 325 problem
session, in which a block slid down an inclined plane that was free
to move on the surface of an ice sheet? This problem lends itself
nicely to Lagrangian techniques, and is considerably simpler when
done this way.

Heres the diagram.

m
1
x
1
x
2
m
2
g
m
1
x
1
x
2
m
2
g


354
354

We have two masses, both of which can move in the plane of the
paper. At first glance, it might look like wed need four
coordinates to characterize the entire system: x, y for each mass.

However, the wedge never leaves the surface of the frictionless
table, so its y position is always zero. In addition, the other mass is
always on the inclined surface of the wedge (until it slides onto the
table). If we knew the x position of the wedge, and how far down
the wedge the second mass has traveled, wed be able to draw a
unique picture of the system: wed have all the information we
needed to describe the system completely. So lets use x
1
, x
2
as
illustrated in the diagram. Note that a point with x
2
=0 moves as
the wedge moves.

Since L =T U, we need to express the total kinetic energy and
total potential energy in terms of x
1
, x
2
, and their time derivatives.

Lets define U so that it is zero when x
2
=0. This way,
2 2
sin . U m gx =


355
355

Now for kinetic energy. Note that it is not
2
1 1 2 2
2 2 m x m x +
2
. The
speed of the wedge is just
1
x . However, even if x
2
is constant, m
2

can be in motion if x
1
is changing. We need to realize that m
2
has a
vertical component of velocity
2
sin x but a horizontal velocity
component
1 2
cos x x + .

Using this, we find that the kinetic energies are
2
1 1
2 m x for mass 1
and
( ) ( )
2
2 2 1 2
sin cos 2 m x x x
+ +
for mass 2. As a result, the

total kinetic energy for the system is
( )
2 2
1 2 1 2 2 2 1 2
1 1
cos
2 2
T m m x m x m x x = + + +
where I have used the fact that
2 2
sin cos 1. + =

The Lagrangian is

( )
2 2
1 2 1 2 2 2 1 2 2
1 1
cos sin .
2 2
L T U
m m x m x m x x m g x
2

=
= + + + +

The Euler-Lagrange equations are


356
356

1 1 2 2
0 and 0
d L L d L L
dt x x dt x x

= =

so we can plug in our Lagrangian and do the differentiations.

We have
( )
( )
1 2 1 2 2
1
1 2 1 2 2
cos
cos
d L d
m m x m x
dt x dt
m m x m x
= + +

= + +

and
1
0
L
x

so the first Euler-Lagrange equation becomes

( )
1 2 1 2 2
1 1
cos 0.
d L L d
m m x m x
dt x x dt

= + + =

Note that this equation tells us that the quantity in square brackets
is constant. If you look at it a little bit youll see that the equation
is informing us that the net horizontal momentum of the system is
constant.


357
357

We have

( )
( )
1 2 1 2 2
1 1
1 2 1 2 2
cos
cos 0
d L L d
m m x m x
dt x x dt
m m x m x

= + +

= + + =

so that
1 2 1
2
2
.
cos
m m x
x
m

+
=

The second Euler-Lagrange equation is

( )
2 2 1 2
2 2
cos sin 0.
d L L d
m x x m g
dt x x dt

= +

=

Consequently,
2 1
cos sin . x x g + =


358
358

(Keep in mind that is fixed: its just the angle of inclination of
the wedge.) Use the result from the first Euler-Lagrange equation
to rewrite the last equation as

1 2 1
1
2
cos sin
cos
m m x
x g
m

+
+ =

or
2
1 2
1
2
cos sin cos
m m
x g
m

+
=

so that
2 2
1
2 2
2 1 2 1 2
sin cos sin cos
.
cos sin
m g m g
x
m m m m m

= =
+

Constant acceleration! (Not at all surprising, actually.) You can
plug in to the earlier equations to find
2
x .

Nasty pendulum
Here is another example. Lets say we have the nasty pendulum
system shown in the following diagram. The first mass is free to

359
359

move along the x axis while the second mass hangs from a
massless rigid rod and can pivot about the point of attachment to
first mass, but is constrained to move in the x-y plane.
x
x
y
m
1
m
2
g
r
xx
x
y
m
1
m
2
gg
r

How many generalized coordinates do we need to describe the
system?

Since m
1
moves (without friction) along the x axis we only need
one coordinate to say where it is. And since m
2
swings from a rigid
rod whose motions are confined to the x-y plane, specifying the
angle ought to finish things off. So lets try working the problem
using x and .

For this system potential energy is only a function of how high m
2

has risen during its swing. For convenience, use y =0 as the place

360
360

where U =0. (Since well always be taking derivatives of the
Lagrangian, any constants we add to U will disappear later.)

We have
2
cos . U m gr =

Now work on the kinetic energy. The speed of m
1
is just x . The
speed of m
2
depends on how fast it is moving with respect to m
1
in
combination with how fast m
1
is moving.

If m
1
is stationary, m
2
s speed is just r
, so its velocity is
cos sin r x r y +

. If m
1
is moving, we have to add in its x
velocity to determine m
2
s net velocity:
( ) ( )
2
cos sin v x r x r y = + +

Square this and use in the calculation of system kinetic energy:
( )
2 2
1 2
1 1
2 cos
2 2
T m x m x r x r
2 2
= + + +

.


361
361

(Note that Ive used
2 2
cos sin 1. + = )

At this point, it is tempting to restrict things to small angles so
sin and cos 1. But note that you have to be a bit careful
doing this. If there were to be a difference of two almost-constant
(large) terms that appeared, a nave approach might not be
sufficiently accurate.

Also, were going to take some derivatives, which will turn cosines
into sines. Lets hold off on the cos 1, etc. substitutions for the
time being.

The Lagrangian is

( )
2 2 2
1 2 2 2 2
1 1
cos cos
2 2
L T U m m x m r x m r m gr = = + + + +

Its partial derivatives are


362
362

( )
2
2 2
2 2
1 2 2
cos
sin sin
cos
0.
L
m r x m r
L
m r x m gr
L
m m x m r
x
L
x

= +
= + +

The Euler-Lagrange equations are

0
d L L
dt

=

so that
[ ]
[ ]
2
2 2 2
2 2
2
2 2 2
sin cos
sin sin 0
cos sin 0
m r x m r x m r
m r x m gr
m r x m r m gr

+ +
+ + =
+ + =

as well as

0
d L L
dt x x

=

so

363
363

( ) [ ] [ ]
2
1 2 2 2
sin cos 0 m m x m r m r + + =

.

At this point I will impose the assumption 1 so that sin
and cos 1.

The first Euler-Lagrange equation becomes

2
2 2 2
2
2 2 2
cos sin 0
0
m r x m r m gr
m rx m r m gr

+ + =
+ + =

or

0. x r g + + =

The second equation becomes

( ) [ ] [ ]
( )
2
1 2 2 2
2
1 2 2 2
sin cos 0
0
m m x m r m r
m m x m r m r

+ + =
+ + =

Notice that the middle term is a nasty non-linear term.

Fortunately, its small ( 1 ) and a term like
is just

364
364

( ) ( )
0
lim
dt
t dt t
dt

+
.

As a result, the middle term is small to third order in . Lets
neglect it in comparison with the other two. Doing this we end up
with
( )
1 2 2
0 m m x m r + + =
.

Use this to eliminate x from the equation 0 x r g + + =
by
replacing x with
( )
2 1 2
m m m r
+ :
2
1 2
0
m
r r g
m m
+ +
+

=
or
1
1 2
0.
mr
g
m m
+ =
+

This is a harmonic oscillator, with angular frequency

( )
1 2
1
m m g
mr
+
= .

365
365

dt

We can then solve for x, having figured out .

Definition of the (classical) action
Theres a good reason for working with L instead of something
else (for example T +U) because of the connection between the
Lagrangian and the classical action associated with the systems
motions.

We could, if suddenly seized by an urge to do a time integral,
calculate the value of
( )
.
stop
start
t
t
S L t

The quantity S is called the action (sometimes the classical
action) that characterizes the particles motion between the times
start
t and
stop
t . Why is this useful? Youll see in a while


366
366

Classical action for a free-falling mass
Lets calculate S for an object released at rest with initial height h
in a uniform gravitational field. Its simple:
2
2 y h gt = and
so v g = t
( )
2
2
2 2
1 1
2 2
.
L T U
m gt mg h gt
mg t mgh
=

=

=

The beginning and ending time of its fall are 0
start
t = and
2
stop
t h = g so that
2
2 2
0
3 3
2
0.471 .
3
h g
S mg t mgh dt
m gh m gh
=

=

No big deal.


367
367

But what would happen if we didnt know that
2
2 y h gt =
(perhaps Newtons law F mx =
start
t
hadnt yet been discovered) and
calculated S based on a crazy guess at the functional form of y(t)?
We might still know that 0 = and 2
stop
t h = g from
measuring the time to fall from a height h so lets use those as the
end points of the integral.

Lets try this for our guess: assume the object falls with constant
velocity so that
0
2
stop
v v h t gh = = = and
0
. y h v t = We have
( )
2
0 0
1
2
3
.
4 2 4
L T U mv mg h v t
mgh gh mgh gh
mgh mg t mg t
= =
= + = +
2

Now calculate S:
2
0
3 3
3
2 4
3
2 2 2
0.354 .
8
h g
gh mgh
S mg t d
h h
mgh
g g
m
gh m gh

=

=

=
t


368
368

Not surprisingly, S is different.

Hamiltons principle
Note that the action for the correct trajectory is less than the action
for the crazy guess. (By less than I mean that
3
0.471m gh is
farther from positive infinity than
3
0.354 . m gh ) This will be true
in general: the correct description for the trajectory of a particle
moving under the influence of conservative forces will always
yield an extremum value for S. (Ill show why this is true in a little
while.) This is called Hamiltons principle.

Its nearly always the case that S is minimized, rather than
maximized, so Hamiltons principle is sometimes called the
principle of least action. Note that all trial trajectories we might
use in calculating S must begin at the same point in space at time
t
start
and also end at the same point at time t
stop
as illustrated in the
following diagram.

369
369

x
y
z
( )
stop
x t
( )
start
x t
x
y
z
( )
stop
x t
( )
start
x t

Demonstration of the equivalence of Hamiltons principle and
the Euler-Lagrange equations
The proof of the principle of least action involves something you
might not have seen before called the calculus of variations. With
it well be able to show that the following three statements are
equivalent:

1. Objects move so that F mx =
.

370
370

2. Objects move so that 0
k k
d L L
dt q q

=

where the Lagrangian
L is a function of the generalized coordinates
and their first (total) time derivatives .
n
q q q
1 2
, ,...
n
q q q
1 2
, ,...
3. Objects move so that
( )
dt is an extremum.
stop
start
t
t
S L t =

Setting up a calculus of variations approach to the problem
Weve already looked at the connection between equation 1 (one
of Newtons laws) and equation 2 (the Euler-Lagrange equation).
Now lets see why the Euler-Lagrange equations are equivalent to
the principle of least action.

Ill start with the principle of least action and show how it yields
the Euler-Lagrange equations. The calculus-of-variations approach
requires you to keep in mind the difference between a partial
derivative and a total derivative.


371
371

)
If we compare the action S for each of the trajectories in the
previous figure well find that there is one for which the action is a
minimum. (Note that all paths must begin at the same initial point
(
start
x t
and end at the same final point

( )
stop
x t
.) Lets refer to the

particular path that minimizes S as
( )
best
x t
, as sketched in the next

figure.

x
( )
stop
x t
( )
start
x t
( )
best
x t
y

If we calculate the action along a somewhat different (and
therefore unphysical) path well find a larger value for S. Lets
describe this other path as
( ) ( ) ( )
.
best
x t x t t = +

Ive illustrated what I mean in the following diagram. ( is a
lower-case Greek letter eta. The upper-case eta will look
familiar: its .)

372
372

( )
stop
x t
( )
start
x t
( )
best
x t
( ) ( ) ( )
best
x t x t t = +

y
x

To make it easier to see what were doing, let me switch to
schematic representations that show only one spatial dimension (as
the ordinate) and a time axis (as the abscissa). We have, then,
curves that look as shown in the next figure. Note that is
negative, except at the starting and stopping times
start
t and .
stop
t


373
373

x
( )
stop
x t
( )
start
x t
( )
best
x t
( ) ( ) ( )
best
x t x t t = +
start
t
stop
t
( )
t
t

Lets introduce a dial we can use to vary the function
( )
x t
smoothly from
( ) ( )
best
x t t + into
( )
best
x t . Do it this way:

( ) ( ) ( )
, .
best
x t x t t +

We can choose any we want; once we pick its value we leave it
alone until were ready to study x with a different value. (By this I
mean that acts like a constant whenever we take a time
derivative of x. Naturally,
( ) ( )
,0
best
x t x t = .

374
374

The total time derivative of
( )
, x t is
( )
, dx t dt or
( )
, x t .

We can write the Lagrangian for the path
( )
, x t in terms of x and
x , of course:
( ) ( ) ( )
, , , L x t x t .

Minimizing the action through proper choice of trajectory
The statement that the action is minimized by the trajectory that
has set to zero can be written this way:
0
0
dS
d

=
= .
Lets write this in terms of the integral of the Lagrangian and do
some work.

( ) ( ) ( )
( ) ( ) ( )
, , ,
, , ,
.
stop
start
stop
start
t
t
t
t
dS d
L x t x t dt
d d
dL x t x t
dt
d

=
=


375
375

Recall how to express the total derivative of L in terms of its
partials:

( ) ( ) ( )
, , , dL x t x t
L dx L dx
d x d

x d

= +

so that
.
stop
start
t
t
dS L dx L dx
dt
d x d x d

= +

Now, since
( ) ( ) ( )
,
best
, x t x t t = + the derivatives with respect
to are easy to evaluate:
( ) ( )
( )
( ) ( )
( )
.
best
best
d x t t
dx
t
d d
d x t t
dx
t
d d

+

= =
+

= =

Consequently,

376
376

.
stop
start
stop
start
stop
start
t
t
t
t
t
t
dS L dx L dx
dt
d x d x d
L L
dt
x x
L L d
dt
x x dt

= +

= +

= +

Note that
.
stop stop
start start
t t
t t
L d L
dt d
x dt x

=

Recall how to do an integration by parts. Since
dfg dg df
f g
du du du
= + or
[ ]
d fg fdg gdf = +

we can rewrite an integral making use of this:

[ ]
b b b
b
a
a a a
fg d fg fdg gdf = = +

or

377
377

.
b b
b
a
a a
fdg fg gdf =

Employ this now:
.
stop stop stop
start start start
t t t
t t t
L L
d d
L
x x x

=

Since is zero at the starting and stopping times, we have
0.
stop
start
t
t
L
x

Use this to rewrite our earlier integral:

378
378

stop
start
stop stop
start start
stop stop
start start
stop stop
start start
t
t
t t
t t
t t
t t
t t
t t
dS L L d
dt
d x x dt
L L
dt d
x x
L L
dt d
x x
L d
dt dt
x dt
L d L
x dt x

= +

=

L
x

=

=

.
stop
start
t
t
dt

Our starting point, that the correct trajectory minimizes the action,
is equivalent to the statement that
0
0. dS d
=
= Recall that
( ) (
,0
best
)
x t x = t and
( ) ( )
,0
best
x t x t = .

Consequently,
0
0
stop
start
t
best best t
dS L d L
dt
d x dt
=
x

= =

with
best
x the actual, physical trajectory.

379
379

I havent placed any restrictions on the function
( )
t other than its
zero value at
start
t and
stop
t : it could look like any of the curves
shown in the following diagram, for example.

x
start
t
stop
t
t

The only way for the integral
stop
start
t
best best t
L d L
dt
x dt x

to be zero,
regardless of the shape of , is for the integrand to be zero: it is
necessarily the case that the correct trajectory minimizing the
action S forces us to have
0
best best
L d L
x dt x

=

.

This, except for the trivial sign reversal of the two terms on the left
side, is the Euler-Lagrange equation.


380
380

)
n
We already know that the Euler-Lagrange equations are true;
theyre just a sophisticated way of rewriting the equation .
I could have run the steps on the last few pages backwards in order
to show that the truth of the Euler-Lagrange equations forces
Hamiltons principle to be true.
F ma =

For a Lagrangian that depends on more generalized coordinates:
we have a series of n equations, all of the
form
(
1 2 1 2
, ,... , , ,...
n
L q q q q q q
0
k k
d L L
dt q q

=

.

The validity of Hamiltons principle (which Ive been calling the
principle of least action) is equivalent to the Euler-Lagrange
equations holding true for a system.

The Euler-Lagrange equations and other minimization problems
We can state this more generally for functions that have nothing to
do with kinetic and potential energies since the math to show it is

381
381

)
n
exactly the same. Imagine we are given a function
and choose to express the values taken on
by the variables as functions of some other parameter s. (I am
defining
(
1 2 1 2
, ,... , , ,...
n
F q q q q q q
.
k k
q dq ds ) We might be interested in determining the
behavior of the variables ,
k
q q
k
so that the integral of F over some
range of s is minimized. In equations, a quantity G is to be
minimized where
( ) ( ) ( ) ( ) ( ) ( ) ( )
2
1
1 2 1 2
, ,... , , ,... .
s
n n
s
G F q s q s q s q s q s q s ds =

Once we succeed in effecting the minimization of G, we will
always find that the coordinates
( )
k
q s and their derivatives
( )
k
q s
automatically satisfy the Euler-Lagrange equations for each of the
: ,
k k
q q 0
k k
d F F
ds q q

=

. (Recall that .
k k
q dq ds )

A different minimization problem
Lets take a look at a different minimization problem that lends
itself well to a calculus of variations approach.

382
382

Keep in mind that my development of the equivalence of the
principle of least action and the validity of the Euler-Lagrange
equations didnt actually rely on how we defined the Lagrangian as
T U. As long as we know that its always the case that
0
k k
d L L
dt q q

=

for a function then well also be confident
that
(
1 2 1 2
, ,... , , ,...
n
L q q q q q q
)
n
( ) ( ) ( ) ( ) ( ) ( ) ( )
1 2 1 2
, ,... , , ,...
B
n n
A
S L q t q t q t q t q t q t dt =

is an extremum along the actual trajectory defined by the
coordinates
( ) ( ) ( )
1 2
, ,...
n
q t q t q t
( )
,...
n
q t
and their derivatives
.
( ) ( )
1 2
, q t q t

Brachistochrone
Heres our problem. A point mass slides along a stiff wire under
the influence of gravity after being released from rest at the point

383
383

) (
,
A A
x y . The wire passes through the point
( )
,
B B
x y . How should
the wire be shaped in order to minimize the transit time from point
A to point B?

An outline of the steps to be followed:

1. Concoct an integral which yields the time to move from A to B if
the shape of the path is specified (in other words, if the details of
the function y(x) are known).

2. Discuss what condition, applied to the integrand, will force the
integral to have the minimum possible value.

Heres a figure. Note that I will use +x to correspond to the down
direction in this problem.


384
384

As long as its safe for us to ignore friction we can use
conservation of energy to relate the speed of the particle to the
height through which it has fallen:
2
1
( )
2
A
mg x x mv =
so
( )
2 ( ).
A
v x g x x =

In the enlarged region in the figure the bead moves through
horizontal and vertical displacements of dy and dx , or a total
distance of
Note the unusual
definition of
positive x as down.
x ( )
,
A A
x y
( )
,
B B
x y
dx
Sliding bead
dy
g
y

385
385

( ) ( )
2 2
ds dx dy = +
.

If the enlarged region is centered at the point
( )
, x y the bead will
be moving with speed
( )
2 ( )
A
v x g x x = and will take time
to traverse the region. / dt ds v

For convenience, lets define the elevation of point A to be zero:
. We have 0
A
x =
( )
( ) ( )
( )
2 2
2
2
1
.
2
ds
dt
v x
dx dy
gx
dy dx
dx
gx
+
=
+
=

The time to travel from A to B is just the integral of this:

( )
2
1
.
2
B B
A A
dy dx
T dt d
gx
+
= =

x


386
386

Now we have an expression for the transit time T in terms of an
integral of a function of x and y. We are interested in figuring out
what shape for the wire will minimize T. Stated differently, wed
like to determine the function
( )
y x that minimizes the transit time.
Once weve determined
( )
y x we can bend our wire into this
shape.

How to go about this? Use the Euler-Lagrange equation that
corresponds to the minimization of the integral that yields T!

A few pages back I discussed minimizing a function G defined as
( ) ( ) ( ) ( ) ( ) ( ) ( )
2
1
1 2 1 2
, ,... , , ,...
s
n n
s
G F q s q s q s q s q s q s ds =

by choosing the proper form for the functions
( )
k
q s and their
derivatives .
( )
k
q s

The coordinates and their derivatives
( )
k
q s
( )
k
q s
,
k k
q
will satisfy the
Euler-Lagrange equations for each of the : q 0
k k
d F F
ds
q q
=

where .
k
ds
k
q q d

387
387

Compare our expression for T with our expression for G. The
integration variable is x instead of s, and we have
( )
y x instead of
We also have
( )
1
. q s dy dx (or
( )
y x if you prefer) instead of
. I suppose we could also associate the x in the denominator
inside the square root with a second generalized coordinate that is
just the identity function:
( )
1
q s
( )
2
x q s s . And the integrand is
( ) ( )
2
1
2
+ y x
gx
rather than
( )
1 2
, ,... q
1 2
, ,...
n n
q q q , F q q .

The Euler-Lagrange equation for our T minimization is

( ) ( ) ( ) ( )
2 2
1 1
2 2
0.
y x y x
gx gx d
dx y y

+ +

=

Now take a look at the terms on the left side of the equation. The
partial derivative to the right of the negative sign must be zero:

388
388

( ) ( )
2
1
2
0
y x
gx
y
+

since the square root does not depend explicitly on y.

The other term is

( )
( )
( )
( )
2
1
2 2
2
2
1
1 2 1
2 2
1
.
2 1
y
y gx d d
dx y dx gx gx
y
d
dx
y
x
g y

+

=

=

+

As a result, our Euler-Lagrange equation is

( )
( )
2
2
1
0
2 1
y
d
dx
x
g y

=

+

or

389
389

( )
( )
2
2
1
constant.
2 1
y
x
g y
=

+

It looks messy, no? But lets just square both sides, since the left
side is a constant. We have
( )
( )
2
2
1
constant.
2 1
y
x
g y
=

+

Lets define the constant so that we can rewrite this equation in
terms of another constant a:
( )
( )
2
2
.
2
1
y
x
a
y
=

+

This looks nasty. How might we integrate it to solve for y?

Take note of the denominator. This suggests that we
might make some headway with a trig substitution: recall that
( )
2
1 y +
( )
2 2 2
1 tan sec 1 cos + = = . So lets try this: tan . y


390
390

Doing this, we have
( )
( )
( )
( )
( )
2 2
2
2
2
tan tan
sec
1 tan
sin .
2
x
a

=

+

= =

As a result, we can also write
( ) ( )
2 2
cos 1 sin = so that
( )
( )
( )
2
2
2
sin
tan
cos
1
2 2
.
2
x x
a a
x
a x
=

=

=

Take the square root:
tan .
2
x
a x
=

Recall our original substitution tan : y now rewrite the
previous equation as
2
2
2
dy x x
y
dx a x
ax x
= = =


391
391

and integrate:
2
2
x
y d
ax x
=
x
)
.

This integral isnt so bad: use another trig substitution, replacing x
with
(
1 cos a . We have
( )
( ) ( )
( )
( )
( ) ( )
( )
( )
( )
( )
( )
( )
( )
[ ]
2
2 2
2
2
2
2
1 cos
1 cos
2 1 cos 1 cos
1 cos
1 cos
2 1 cos 1 cos
1 cos
sin
2 2cos 1 cos 2cos
1 cos
sin
1 cos
1 cos
sin
sin
1 cos
sin constant.
a
y d
a a
a
d
a
a
d
a
d
a
d
a d
a
=
+
=
=
= +

As a result, we have

392
392

( )
[ ]
1 cos
sin constant.
x a
y a

=
= +

If we define
a
x x x = and
a
y y y = we can rewrite these
equations as
( )
[ ]
1 cos
sin .
x a
y a

=
=

When is zero we have
( ) ( )
, ,
A A
. x y x y =

We need to determine a value for a somehow, and then dial up the
value of until we arrive at
( ) ( )
, ,
B B
. x y x y = In principle we have
enough information to solve the equations: there are two equations
in two unknowns, and we know the values for x and . But its
a nasty problem so it is easier to determine a value for and a
using some sort of numerical approximation technique.
y

At the point
( )
,
B B
x y we have
( )
( )
( )
( )
1 cos 1 cos
sin sin
B B
B B b
a
x
y a
B

= =


393
393

where
B
is the value of at that point. From this equation we can
use numerical techniques to determine an approximate (but
extremely accurate) value for
B
. After that we can solve the
equation
( )
1 cos
B
x a = for a:
( )
1 cos
B
a x = .

The equations
( )
1 cos x a = and
[ ]
sin y a = describe a
cycloid: a point on the rim of a wheel of radius a that rolls without
slipping traces out a cycloid. The describes the angle through
which the wheel has turned, as shown in the figure below for a
wheel of diameter 4. (In the figure, the wheel rolls without slipping
on the underside of the horizontal axis.)

0
1
2
3
4
0 2 4 6 8 10 12 14 16 18

cycloid


394
394

A few comments
Generalized coordinates
The coordinates q
1
, q
2
, used in the Lagrangian dont have to be
Cartesian coordinates x, y, We can, for example, use spherical
coordinates r, , , or cylindrical coordinates r, , z, or most
anything else that we find convenient. Since potential and kinetic
energy are scalars, L wont change under coordinate
transformations. You just have to make sure you express kinetic
and potential energy properly in the appropriate coordinate system.

Generalized, non-Cartesian coordinates are especially convenient
for systems with holonomic constraints. Holonomic constraints are
constraints which can be expressed as equations (not inequalities)
connecting the coordinates (not the velocities). One example is a
mass sliding on a bent wire (with known shape) under the
influence of gravity.


395
395

Another example is a pendulum hanging from a massless, rigid rod
fixed at the origin. If we work in Cartesian coordinates well have
the constraint that the x, y, z coordinates of the mass must satisfy

x
2
+y
2
+z
2
=a
2
.

z
g
y
a
x
m

A more sensible coordinate system would be spherical coordinates
r, , since r is constant, and will not appear in the equations of
motion except as a constant.

Heres another example: two particles that are connected by a rigid
rod of length d. We have, naively, 6 coordinates to work with: x
1
,
x
2
, y
1
, y
2
, z
1
, z
2
where the subscript labels which particle were
working with.

396
396

( )
1 1 1
, , x y z
( )
2 2 2
, , x y z
d

The constraint, obviously, is (x
1
x
2
)
2
+(y
1
y
2
)
2
+(z
1
z
2
)
2
=d
2
.
All sorts of possible generalized coordinates come to mind. An
especially convenient one would probably be the center-of-mass
position along with the angles , that give the direction of the
line connecting the first mass to the second mass.

Feynmans use of the principle of least action
A quick appeal to the small amount of quantum mechanics you had
in Physics 213/214.


397
397

Notice that the action S has units of energy time, the same as
Plancks constant. (Recall two ways of writing the uncertainty
principle: 2 E t
(
2 h
)
= and also 2 p x .)

In classical mechanics, the only possible trajectory for a particle is
the trajectory that minimizes the action.

In quantum mechanics, trajectories for which S isnt a minimum,
but only vary from S
min
by ~h are also possible ways for the
system to be seen to behave.

Richard Feynman did some striking, original work reformulating
quantum mechanics to take advantage of this not-quite-least-action
feature, and cooked up the theory of quantum electrodynamics, a
marvelously accurate, successful model of the interactions of
electrons, muons, etc. with electromagnetic fields. (He did this in
the late 40s.) This got him his Nobel prize.


398
398

More Lagrangian examples
Double Atwoods machine, done two ways
Lets do a double pulley example, first using non-Lagrangian
methods, then with reference to the Lagrangian. The pulleys and
strings are massless, and none of the strings stretch.

y
1

Lets say the lengths of the two strings are L
1
and L
2
. Recall that
the tension in a (massless, inextensible) string must be the same all
along its length. (If the tensions were to be different on the left and
right sides of a pulley the [massless] pulley would experience
infinite angular acceleration until the tensions became equal.)
y
2
T
1
T
1
m
1
T
2
T
2
m
2
m
3
y =0

y
3

399
399

Say the upper pulley is at y =0. Lets use y
1
to label the height of
m
1
. My sign convention is that y
1
is negative when m
1
is lower than
the upper pulley. Since the first string has total length L
1
, it must
be true that L
1
equals the sum of the upper-pulley-to-m
1
distance
and the upper-pulley-to-lower-pulley distance. As a result, the
height of the lower pulley is
( )
1 1
L y + .

The height of mass m
2
is y
2
with respect to the lower pulley. Since
the lower pulley is at
( )
1 1
y L y = + , the height of m
2
with respect
to the origin is
( )
2 2 1
y L y + = +
1
2 2
y
y y .

The height of mass m
3
is y
3
with respect to the lower pulley where
or . Since the lower pulley is at
, the height of m
3
with respect to the origin is
3 2
y y L =
(
1
y L = +
3 2
y L =
)
( )
1
y
( ) ( )
2 1 1
y L y + +
3 3
y y y + =
1 1 2
L y L + = .

A summary so far:

400
400

1 2
L L
( )
1 1
2 2 1 1
3 2 1
1
2
position
position
position
string 1 tension
string 2 tension
m y
m y y L
m y y
T
T
=
=
= +
=
=

Lets write down the forces acting on each of the masses and the
lower pulley now.
1) Net force on m
1
is
2
1
1 1 1
2
d y
T m g m
dt
=
2) Net force on m
2
is
( )
2
2 1 1
2 2 2
2
d y y L
T m g m
dt

=
3) Net force on m
3
is
( )
2
2 1 1 2
2 3 3
2
d y y L L
T m g m
dt

=
4) Net force on the lower pulley is zero (its massless) so T
1
=2T
2

We have four equations for the forces, and four unknowns: T
1
, T
2
,
y
1
, and y
2
.

Crank away:

1)
1 1 1 1
m y T m g =

401
401

2)
2 2 2 1 2 2
m y m y T m g =
3)
3 2 3 1 2 3
m y m y T m g =
4) T
1
=2T
2

J ust to put it all down on paper explicitly, heres the algebra
needed to slog through the four equations.

(a) Use equation 4) to replace T
2
with T
1
/2 everywhere.

1)
1 1 1 1
m y T m g =
2)
2 2 2 1 1 2
2 m y m y T m g =
3)
3 2 3 1 1 3
2 m y m y T m g =

(b) Use equation 1) to replace by
1
y
1 1
T m g in equations 2) and
3).

2)
2 2 2 1 1 2 1 2
2 m y m T m m g T m g = +
3)
3 2 3 1 1 3 1 3
2 m y m T m m g T m g = + +


402
402

(c) Note that both equations have on the left side. Divide both
equations by the appropriate mass (either m
2
or m
3
) and equate the
right sides.
2
y

( )
( )
2 1 1 1 3
2 1 1 1 2
2
2
y T m g T m
y T m g T m g
= + +
= +
g

so
( ) ( )
1 1 1 2 1 1 1 3
2 2 T m g T m g T m g T m g + = + + .

(d) Solve for T
1

1
1 2 3
2 1 1
4
2 2
T g
m m m

+ + =

so that

1 2 3
1
2 3 1 3 1 2
8
4
mm m g
T
m m mm mm
=
+ +
.


403
403

(e) Now that we have T
1
, solve for T
2
(its T
1
/2), then plug in to
calculate and . From before,
1
y
2
y
1 1 1 1
m y T m g = or
1 1 1
m y T g = so that
2 3
1
2 3 1 3 1 2
8
4
m m g
y g
m m mm mm
=
+ +
.
Were finished. That was a lot of work!

Now lets do the problem using Lagrangian techniques. We can
still use the following facts and definitions:

( )
1 1
2 2 1 1
3 2 1
1
2
position
position
position
string 1 tension
string 2 tension
m y
m y y L
m y y
T
T
1 2
L L
=
=
= +
=
=

Call y =0 the zero-point of potential energy. We have

( ) ( ) ( )
1 2 1 1 2 2 1 1 3 2 1 1 2
2 2 2
1 1 2 2 3 3
,
1 1 1
2 2 2
U y y m gy m g y y L m g y y L L
T mv m v m v
= + + + +
= + +


404
404

1
y
with
( )
1 1
2 2 1
3 2
v y
v y y
v y
=
=
= +

so
( ) ( )
2 2 2 2 2
1 1 2 2 1 1 2 3 2 1 1 2
1 1 1
2 2
2 2 2
T m y m y y y y m y y y y = + + + + + .

The Lagrangian is L =T U of course, and the Euler-Lagrange
equations are
0
i i
d L L
dt y y

=

for i =1, 2 as usual.

Do the i =1 equation first:

( ) (
1 1 2 1 2 2 3 1 3 2
1
1 2 3 1 3 2 2
L
m y m y m y m y m y
y
m m m y m m y
= + + +
= + + +

)

while
( )
1 2 3
1
L
m m m g
y
.

405
405

As a result, the Euler-Lagrange equation can be rewritten

( ) ( ) ( )
1 2 3 1 3 2 2 1 2 3
m m m y m m y m m m g + + + = . (equation 1)

Now work with the i =2 equation:

2 2 2 1 3 2 3 1
2
L
m y m y m y m y
y
= + +

and
( )
2 3
2
L
m g m g
y

so that

( ) ( ) ( )
3 2 1 2 3 2 3 2
m m y m m y m m g + + = . (equation 2)

We have produced two force equations directly, instead of the four
we had a few pages ago. Note that we havent needed to use the
strings tensions in our calculation.

406
406

)

Now solve for by multiplying equation 2 by
1
y
( ( )
3 2 2
m
3
m m m + then subtracting it from equation 1:

( )
( )
( )
( )
( )
2 2
3 2 3 2
1 2 3 1 1 2 3
2 3 2 3
m m m m
m m m y m m m g
m m m m

+ + = +

+ +

or
1 2 1 3 2 3
1
1 2 1 3 2 3
4
4
mm mm m m
y g
mm mm m m
+
=
+ +

You can see that this is equivalent to the answer I obtained without
using the Lagrangian.

Brief recap
To recap, Lagrangian techniques are especially handy when it is
difficult to write down the equations describing the net forces

407
407

acting on parts of a system. The block sliding down a moveable
inclined plane is a good example of this.

We need to do the following to make use of Lagrangians:

1) Concoct a suitable set of generalized coordinates which have no
constraints coupling them.
2) Calculate the potential and kinetic energies in terms of the
generalized coordinates.
3) Use the Euler-Lagrange equations to extract the differential
equations for the accelerations of the generalized coordinates.

Not all mechanics problems lend themselves well to Lagrangian
techniques. We need forces to be derivable from potentials:
frictional forces dont work in Lagrangians. In addition, constraints
governing velocities (something rolling without slipping) are
sometimes difficult to handle.

Lagranges equations work fine for problems that were easily
solved using older methods earlier in the course. For example, an

408
408

x undamped harmonic oscillator experiences a force leading
to the potential
F k =
2
2 U kx = .

We could have solved for its motion using the Lagrangian :
( )
2 2
1 1
2 2
0
L T U mx kx
d L L d
mx kx
dt x x dt
= =

= +

=

or
. mx kx =

Spherical pendulum
Lets do yet again another example, but not in Cartesian
coordinates.

Consider a spherical pendulum something free to move on the
surface of a sphere, subject only to gravity (acting in the z

409
409

2
direction) and constraint forces that keep
2 2 2
x y z + + = for
the length of the rod holding the pendulum mass.

Lets put the support point at (x, y, z) =(0, 0, 0) as shown in the
figure below. Also, lets use standard spherical coordinates to
describe the mass position:
( )
, , r =
os ,
. Note that the Cartesian
equivalents are sin c x sin y = sin = and cos . z =
sin cos x r =
sin sin y r =
cos z r =
a
x
z
y
r
sin cos x r =
sin sin y r =
cos z r =
aa
x
z
y
r

Use and as the generalized coordinates, since the constraint
reduces to
2 2 2
x y z + + =
2
r = in spherical coordinates, with no
constraints on or .

410
410

Now figure out the potential and kinetic energies so we can
evaluate the Lagrangian.

How fast is the mass moving? You have to be careful here.
( )
1
2 2 2 2
v x y z = + +

cos cos sin sin
cos sin sin cos
sin
x
y
z

=
= +
=

so
2 2 2 2 2 2
2 2
2 2 2
1
2 2 2
cos cos sin sin
2cos sin cos sin cos sin
sin cos 2cos sin cos sin
sin
v
2

= +
+
+ +

or
1
2 2 2 2 2 2 2
1
2 2 2 2
cos sin sin
sin .
v

= + +

= +


411
411

As a result,
( )
2 2 2 2 2
1 1
sin
2 2
cos
T mv m
U mgz mg
2

= = +
= =

so the Lagrangian is
( )
2 2 2 2 2
1
sin cos
2
L m mg = +

and the Euler-Lagrange equations are

( )
( )
2 2 2 2 2
2 2 2 2 2
1
sin cos 0
2
1
sin cos 0.
2
d
m mg
dt
d
m mg
dt

+

+

=

So

( )
2 2 2
sin cos sin 0
d
m m mg
dt

=

or
2 2 2
sin cos sin m m mg = +
.

412
412

The other Euler-Lagrange equation gives us
2 2
sin
0
d m
dt

=

.

Note that the absence of explicit dependence in L makes
2 2
sin m
a constant of motion.

Its fairly straight forward to see that
2 2
sin m is the component
of angular momentum parallel to the z axis, corresponding to
rotations in the x-y plane:
( ) ( ) ( )

y x z y x z
L r p xp yp z yp zp x zp xp y = = + +

so
( )
2 2 2 2 2
2 2
sin cos cos sin sin cos

sin sin cos cos sin sin
sin cos sin sin
sin .
z
L r p z
m
m
m
m

=
= +

= +

=


413
413

In general, if a particular coordinate q
i
does not appear explicitly in
the Lagrangian, the fact that 0
i
L
q
guarantees that there will be

some conserved quantity, given by
i
L
q
since Eulers equation tells

us that

( 0)
i i
d L L
dt q q
=

0. =

In the spherical pendulum case, the absence of from the potential
and kinetic energies yields conservation of one component of
angular momentum. This is sensible the torque associated with
gravity is r , which must be perpendicular to mg

g
. As a result,
the component of along the direction of L
must be constant.

Back to the spherical pendulums equations of motion.

We find (after dividing the results of the first Euler-Lagrangian
equation by )
2
m

414
414

2
sin
sin cos
g
=
.

Note that a pendulum with very little energy will tend to have
close to 180 which flips the sign of the sine (heh heh).

We found before that
2 2
sin m
is constant. Lets define the

constant
2
sin S

.

Replace
2
in the equation before this to write

2
3
sin cos
0
sin
g S
=

It might be tempting to try out some sort of small-angle
approximation, but keep in mind that can be very different from
zero.


415
415

Here are two simple cases that are easy to describe. The first is
0 =
: the pendulum swings in a planar arc, just like the simple

pendulum from earlier in the course. When 1 the motion is
simple harmonic.

Conical pendulum
The second case is the conical pendulum: initial angles and
velocities are chosen so that is constant, as is
.

For the conical pendulum 0, 0 = =

so the equation
2
sin
sin cos
g
=

becomes

416
416

2
cos
g

with both and
2
constant: we could call these steady-state

values
0
and
2
0
if we want. Also note that it will always be true
that 2 > so the cosine is negative.

Small perturbations to the conical pendulum
What is the motion like if
isnt quite constant, so isnt quite

constant? Lets look at our equation for the acceleration in (
),
above. Keep in mind that S
2
is constant.

Lets write to include a constant
0
and a small time-dependent
part, and see if we can make sense of the behavior of the time
dependent piece:
( ) ( )
0
t t + .


417
417

We expect
( )
0
t for a small perturbation away from the
conical pendulum situation, and will use binomial expansions
liberally.

Lets also write that S =S
0
+ where <<S
0
. Note that is
constant since S is constant.

Our equation for
on the last page becomes

( )
( )
( )
( )
2
0 0
0
3
0
sin cos
0
sin
g S

+ +
+
+
= .

Recall the Taylor expansions we can use on the sines and cosines:
( ) ( )
0
0 0 0
0 0
sin
sin sin ...
sin cos ...
d
d

=
+ = + +
= + +

since
0
=.

Also,

418
418

( ) ( )
0
0 0 0
0 0
cos
cos cos ...
cos sin ...
d
d

=
+ = + +
= +

so
( )
( )
( )
( )
0
0 0
3 3
0
0 0
0 0
0
3
0 0
0 0
3
0 0
cos
cos sin
sin
sin cos
cos sin
cos
1 3
sin sin
cos sin cos
1 3
sin cos sin

0
0
+
+

to first order in .

Also:
( )
2
2
0 0
2 S S
0
S + +
to first order in .

Recall that we found
( )
( )
( )
( )
2
0 0
0
3
0
sin cos
0
sin
g S

+ +
+
+
= .
Rewrite this equation, grouping terms according to how many
powers of , they contain:

419
419

2 0
0 0
3
0
2 0 0
0 0 0
3 3
0 0
2 2
cos
sin
sin
cos cos 3cos
cos 2 1
sin sin sin
terms in , , , and higher orders
0
g
S
g
S S
0
0

+ + + +

+
=

The constant term in the above equation is zero. A few pages back
I showed you that
2
3
sin cos
0
sin
g S
= .

As a result, keeping things only to first order in , we have:

2
0 0 0
0 0
3 3
0 0
cos 3cos 2 cos
cos 1
sin sin sin
S g
S
0
0

+ + + =

.

Looks horrible, right? DO NOT DESPAIR everything is constant
except .

420
420

Define the constants v and
0
this way:

2 2
0 0
0 0
3
0 0
2 0 0
0
3
0
cos 3cos
cos 1
sin sin
2 cos
sin
g
S
S

+ +

in order to write
2 2
0
v v + = .

We know how to solve this: its a harmonic oscillator equation
with a constant driving force which shifts the average value of
away from zero. The solution is

( )
0
sin A vt = + +

The oscillation frequency is

2
0 0 0
0
3
0 0
cos 3cos
cos 1
sin sin
S g

= + +


421
421

while the offset from zero is
0 0
0
3
0
2 cos
sin
S
v

= .

If we started the problem by specifying how different the z-
component of angular momentum was from the conical pendulum
case, wed know all we needed to know to solve the problem, as
long as we knew initial values for
( )
0
t = + and
, or some
equivalent set of parameters.

To restate this, once angular momentum is specified, S is known.
Now, for any S value the motion might correspond to a conical
pendulum, or to a wobbly pendulum. We need to know how much
the motion wobbles in order to describe whats going on.

Lets say at t =0 we know that the pendulum is at a particular
value of , moving with a particular . From this, we can calculate
S. If we also know
then wed have enough additional

information to specify the motion completely.

422
422

We have two independent parameters which describe the
perturbation from constant , constant
motion: (t) and .

Since S =S
0
+, (which is a constant) refers to a kick that would
increase the angular momentum of a pendulum that was initially
moving with constant ,
.

(t) describes the variation in , and could be excited without
changing S by bopping the pendulum in the
direction.
direction of kick


423
423

In either case, once we disturb the initially constant ,
motion,
we see oscillations in , and also in
, since
2
sin S

is
constant.

The Lagrangian is often useful for cooking up quantitative
descriptions of small oscillations away from some kind of
equilibrium. The general approach is similar to what Ive just
worked through.

A comment on generalized momenta
If our Lagrangian is a function of generalized coordinates q
i
and
their derivatives , well have an Euler-Lagrange equation for
each coordinate:
i
q
0
i i
d L L
dt q q

=

.


424
424

If L does not depend explicitly on some particular q
k
, well have a
conserved quantity:

0
k
d L
dt q
so
k
L
q
is constant.

For Cartesian coordinates,
i
i
L
mx
x

since all potentials are velocity independent and
( )
2 2 2
1 2 3
1
2
T m x x x = + +
for
1 2 3
, , x x x y x z .

As a result, if the potential is independent of x
i
so the i
th
component
of force
i
i
u
F
x
is zero, we expect
i
x to be constant. The
momentum component along the Cartesian coordinate i is
conserved.

No surprises here.

425
425

We found, in spherical coordinates, that the absence from L of any
explicit -dependence yielded conservation of the z component of
angular momentum.

We define the generalized momentum p
i
associated with
generalized coordinate q
i
this way:

i
i
L
p
q

It is obvious that p
i
is conserved whenever 0
i
L
q
.

Take note of the fact that the generalized momenta are not always
quantities with the units of mass velocity. For example, in
cylindrical coordinates we have a Lagrangian
( ) ( )
( )
2 2 2 2
1
, , , , , , ,
2
L r z r z m r r z U r z = + +

with the corresponding Euler-Lagrange equations

426
426

0 0
d L L d L L d L L
dt r r dt dt z z
0

= = =

and generalized momenta
2
r z
L L
p mr p mr p
r z

L
mz

= = =

.

Note that the component of generalized momentum p
is the
angular momentum, not the linear momentum.

One of the important properties of the Lagrangian is that it yields
the Euler-Lagrange equations, which serve as powerful tool for
generating differential equations relating the coordinates and their
derivatives: through solution of these equations we are able to
determine the trajectory
( )
q t
described by the coordinates of the

system.

The Hamiltonian as a Legendre transform of the Lagrangian
Would it be useful to rewrite the Lagrangian in terms of
rather than ? In cylindrical coordinates L would become
,
i i
p q
,
i
q q
i

427
427

( ) ( )
2
2 2
2
1
, , , , , , ,
2
r z r z
p
L r z p p p p p U r z
m r

= + +

but then the Euler-Lagrange equations would be a mess: wed need
to use the chain rule to reexpress the L

derivative in terms of
the momenta, for example. And for more exotic orthogonal
coordinate systems (confocal paraboloidal coordinates, for
example) it might be more trouble than its worth.

Have you learned about Legendre transforms in a math course? We
use a Legendre transform to assist in creating a new function (in
terms of new variables) from our original function (which
depended on a set of coordinates and their time derivatives ).
k
q
k
q

In one dimension it works this way, for an arbitrary function
:
( )
, F q q
( ) ( ) ( )
, , G q p qp F q q q p = ,
where
F
p
q
.
G is the Legendre transform of F. Note that we need to reexpress
in terms of q and p inside the function F. q

428
428

When F is a function of several variables we have
( )
( ) ( ) ( )
1 2 1 2
1 2 1 1 2 1 2 2 1 2 1 2
, ,..., , ,...
, ,..., , ,..., , ,... , , ,..., , ,... ,...
k k
k
G q q p p q p
F q q q q q p p q q q p p

=

.

The Legendre transform of the Lagrangian is called the
Hamiltonian:
( )
1 2 1 2
, ,... , ,...
k k
k
H q q p p q p L

=

.

For example, in Cartesian coordinates we have
( ) ( )
( ) ( )
2 2 2
1 2 3
2 2 2
1 2 3
1
, ,
2
1
, ,
2
L T U m x x x U x y z
p p p U x y z
m
= = + +
= + +

so that

429
429

( )
( ) ( )
( ) ( )
( ) ( )
2 2 2
1 2 3
2 2 2
1 2 3
2 2 2
1 2 3
, , , , ,
1
, ,
2
1
, ,
2
1
, ,
2
.
x y z k k
k
k
k
k
H x y z p p p x p
p p p U x y z
m
p
p
m
p p p U x y z
m
p p p U x y z
m
T U

=

+ +

=

+ +

= + + +
= +

The Hamiltonian is (nearly always) the systems total energy.

Of what use is the Hamiltonian? We already know that total energy
is conserved in many systems of interest. Is there anything else?
Well, yes: the Hamiltonian is also the time evolution operator for a
system. Once we know the functional form of the Hamiltonian we
can use it to calculate the time derivatives of the coordinates and
momenta in a straightforward fashion. Let me prove this for you.


430
430

Lets imagine that we figure out all the p
i
s using
i
i
L
p
q
so we
know the functional form for all the generalized momenta p
i
in
terms of all the coordinates and their derivatives :
1 2 1 2
, ,..., , ,... q q q q
( )
1 2 1 2
, ,..., , ,...
i i
p p q q q q .

Imagine that somehow we invert this so that we have written the
'
i
s q in terms of all the : ,
k k
q p
( )
1 2 1 2
, ,..., , ,...
i i
q q q q p p .

(Dont worry how wed actually do this.)

Once weve done this we can rewrite the Lagrangian in terms of
the and actually generate an explicit form for the
Hamiltonian:
,
k
q p
k


431
431

( )
( ) ( ) ( )
( )
1 2 1 2
1 2 1 1 2 1 2 2 1 2 1 2
1 2 1 2
, ,... , ,...
, ,... , ,..., , ,... , , ,..., , ,... ,...
, ,..., , ,... .
k k
k
k k
k
H q q p p q p
L q q q q q p p q q q p p
q p L q q p p

=

If we were to do this, wed be able to figure out the variation in H
if we were to change q
k
by
k
q and p
k
by
k
p since wed be able to
take the various partial derivatives to calculate H .

We would find
k k
k
k k
H H
H p q
p q

= +

just from the way the chain rule works.

Lets see if we can learn anything new from this by referring back
to our definition of the Hamiltonian:
( ) ( )
1 2 1 2 1 2 1 2
, ,..., , ,... , ,..., , ,... .
k k
k
H p q q q p p L q q p p

=


432
432

Doing the standard calculus stuff, lets calculate H when we
vary by ,
k
q p
k k
q and
k
p using our definition of H.

( )
( )
k k k k k k
k k
k k
k k k k k k
k
k k
L L
H p q p q q q
q q
L L
p q p q q q
q q

= + +

= +

.

This looks complicated, since we need to figure out how much
changes when we change since we really are working with
: is no longer an independent variable. If we
were to tough it out, wed write
k
q
,
k
q p
k
) (
1 1
, ... , ...
k
p q q
k
q

( )
( ) ( )
1 1
1 1 1 1
, ... , ...
, ... , ... , ... , ...
.
k k
m
k k
k
k k
p q
p q p q
q p
q p
q
q q

= +

But recall that
k
k
L
p
q
so we can rewrite our expression for H:


433
433

( )
( )
( )
.
k k k k k k
k
k k
k k k k k k k
k
k
k k k
k
k
L L
H p q p q q q
q q
L
p q p q p q q
q
L
p q q
q

= +

= +

Recall from the Euler-Lagrange equations that

k k
k k
L d L d
p p
q dt q dt

= = =

.

As a result, we can write H :
( ) [ ]
1 2 1 2
, ,..., , ,... .
k k k k
k
H q q p p q p p q =

Compare this with the basic, must-be-true calculus fact that

( )
1 2 1 2
, ,..., , ,... .
k k
k
k k
H H
H q q p p p q
p q

= +


434
434

For these two different expressions for H to be equal it must be
true that
k k
k k
H H
q p
p q

= =

.

These are called Hamiltons canonical equations of motion. Once
we calculate
( )
1 2 1 2
, ,... , ,...
k k
k
H q q p p q p L

=

we can take derivatives of H to extract the rate-of-change of q
k
and
p
k
.

This is what H is good for: if it does not contain q
k
explicitly, then
p
k
is constant. If H doesnt contain p
k
, then q
k
is constant.

It turns out that H is important in the formulation of mechanics
which was used as a basis for the invention of non-relativistic
quantum mechanics.

Do keep in mind that the generalized momentum is sometimes
quite different from mv. For example, in calculations involving
particles moving in (static) magnetic fields, youll learn in

435
435

g electrodynamics that it is convenient to work with somethin
called the vector potential A
which is related to the magnetic f

through the relation
ield
B A =

. In kinematic problems of this sort
the generalized momentum that appears in the Hamiltonian isnt
mv
, but rather mv qA c +

.

Lagrangian and Hamilto
c
nian for a particle in a magnetic field
You know that the force exerted on a particle in an electromagneti
field is
( )
. F q E v B = +

Because the force is velocity-dependent, its not going to work if
we try to define a scalar potential U such that . F U =

But there
is something we can do along these lines, since the magnetic force
cannot change the energy of a particle.

re you familiar with the vector potential generated by moving
to
A
charges? In principle its not a particularly complicated function
calculate, at least for a point charge: for non relativistic motion, its
nothing more than the scalar potential caused by the charge,

436
436

equations, now: for a point charge q at position
multiplied by the charges vector velocity, and divided by the
square of the speed of light.

r
In moving with
velocity , v
the scalar and vector potentials V and A
are
( )
2
0
1 c q q
0
4 4
V r
r r r r
= =

and
( ) ( )
0
2
4
v q
A r V r
c r
v
r
= =

.

always true that and
A
E V
t

B A =

It is so the force on
a charge is
.
A
F q V v A
t

= +

We can define a generalized potential energy that is velocity
dependent this way:
( )
U q V A v =

.
With this, we can define the Lagrangian f a particle in an
electromagnetic field like so:
or

437
437

. L T U T qV qA v = = +

Recall that the Hamiltonian is defined this way:
( )
, ,... , ,...
1 2 1 2 k k
k
H q q p p q p L

=

.
The generalized momenta are ,
k
k k
L L
p
q v

=

so the presence of
ian is going to a
onto the usual momentum of mass times velocity:
the term qA v in the Lagrang dd an extra piece

2
1
2
mv qA v

+

k k k
k
p mv qA
v
= = +

.
Since
( )
2
2 2
1
,
2
k k
k
H q p L v p L
v mv qA L mv qA v L
mv qA v T qV qA v mv qV

= =

= + = +

= + + = +


438
438

the Hamiltonian is just the total energy of the particle. And since
the relationship between the canonical momentum and the velocity
is we have , p mv qA = +

, mv p qA =

or
( )
2
1
.
2
H p qA qV
m
= +

In developing quantum mechanics, this provides a natural way to
incorporate magnetic fields into the Hamiltonian, by replacing the
classical kinetic energy term
(
2
1
2
p qA
m

)
with
2
1
.
2
qA
m i

Poisson brackets
There is another piece of (advanced) classical mechanics that
transfers naturally into quantum mechanics. In a course like
Physics 486 or 487 youll learn about commutators like
[ ]
, p x px xp =

439
439

where p and x are quantum mechanical operators rather than
kinematic variables: p
i x
while x is just x. Note that

[ ] [ ]
, , p x x p = .

Note that
[ ] ( )
,
0.
p x x x
i x i x
x x
i x i i x i

=

= + =

We say that p and x do not commute.

In classical mechanics, the Poisson bracket is defined as follows,
for functions and
( )
1 2 1 2
, ,... , ,... f q q p p
( )
1 2 1 2
, ,... , ,... g q q p p of the
coordinates and canonical momenta:
[ ]
, .
k
k k k k
f g f g
f g
q p p q

=


440
440

We can evaluate the time derivatives of the coordinates and
generalized momenta using the Hamiltonian: we found previously
that
k
k
H
q
p
and
k
k
H
p
q
.

It is not hard to show that the following are true:

[ ]
, 0 ,
,
, ,
, .
j k j k
j k jk
0
j j j
q q p p
q p
q H q p H p
df f
f H
dt t

j
= =

=

= =

= +

In addition, any quantity that commutes with the Hamiltonian
and does not explicitly depend on time (so that
[ ]
, 0 f H = and
0 f t = ) is a constant of motion: 0. df dt =

Thats it! Good luck!


441
441


442
442
Physics 325, fall 2012 Math review University of Illinois
Math review
Math review................................................................................m-1
1. series approximations..........................................................m-2
Taylors Theorem.................................................................m-2
Binomial approximation.......................................................m-2
sin(x), for x in radians and x close to zero............................m-4
cos(x), for x in radians and x close to zero...........................m-5
2. some geometry.....................................................................m-6
right triangles........................................................................m-6
triangles in general................................................................m-7
circles....................................................................................m-8
3. exponentials.........................................................................m-9
4. differential equations.........................................................m-10
5. Vectors...............................................................................m-13
Scalar product.....................................................................m-13
Cross product......................................................................m-14
Differentiation of vectors....................................................m-15
Integration of vectors..........................................................m-16
Line integrals......................................................................m-17
6. Vector transformation properties.......................................m-18
Under rotations...................................................................m-18
Matrix multiplication and rotation matrices.......................m-20
How unit vectors transform under rotations.......................m-26
Successive rotations in three dimensions............................m-27
Kronecker delta symbol ......................................................m-30
Levi-Civita symbol .............................................................m-31
Tensors................................................................................m-32
7. Non-cartesian coordinate systems.....................................m-35
cylindrical coordinates........................................................m-36
spherical coordinates..........................................................m-37
8. partial derivatives...............................................................m-38

m-1
copyright 2012 George Gollin PHY_325_math_review.doc

443
443
............................................................................................m-40
gradient operator.................................................................m-41
divergence operator ............................................................m-44
curl operator........................................................................m-45
9. quadratic formula...............................................................m-46

1. series approximations

Taylors Theorem

( )
( ) ( ) ( )
0 0 0
0 0
1
!
n
n
n
df d f
f x f x x x x
x x x x
n dx
dx
x = + + + +
= =

This can be handy. (Recall: n! =123 [n-1]n)

Binomial approximation

Work out an approximation for
( )
1
n
x + when x is small:

m-2

444
444
Use
0
0 x = .

( )
1
n
d
x
dx
+ =
( )
1
1
n
n x

+

( )
2
2
1
n
d
x
dx
+ =
( )
1
1
n
d
n x
dx
+ =
( )( )
2
1 1
n
n n x

+

so

( ) ( ) ( ) ( )
( )( ) ( )
1
0 0
2 2
0
1 1 1 0
1
1 1 0 ...
2!
n n n
x x
n
x
x x n x x
n n x x
= =
=

+ = + + +

+ +

+

=
( )
2
1
1
2
nx
n n
x + +

+

when 1 x << we end up with
( )
1 1
n
. x nx + +

This works when n is a fraction, too.

m-3

445
445

sin(x), for x in radians and x close to zero

0
0 x =

2
2
sin( ) sin( )
cos( ) sin( )
d x d x
x x
dx dx
= =

2
2
cos( ) cos( )
sin( ) cos( )
d x d x
x x
dx dx
= =

2 2
0 0
2
3 3
0
3
2 3
3
sin( ) sin( )
sin( ) sin(0)
2!
sin( )
...
3!
sin(0) cos(0) sin(0) cos(0) ...
2! 3!
0 0 ... for 1
3!
x x
x
d x x d x
x x
dx dx
x d x
dx
x x
x
x
x x x
= =
=
= + +
+
= + +
= + + +
+

m-4

446
446

cos(x), for x in radians and x close to zero

again,
0
0 x =
2 2
0 0
2
3 3
0
3
2 2
cos( ) cos( )
cos( ) cos(0)
2!
cos( )
...
3!
1 0 0 ... 1 for 1
2! 2!
x x
x
d x x d x
x x
dx dx
x d x
dx
x x
x
= =
=
= + +
+
= + + +
+

m-5

447
447
2. some geometry

right triangles

c
b
a
c
b
a

2 2
a b c + =
2
2
3
n

Integers n
1
, n
2
, n
3
for which
2 2
1 2
n n + = are called
Pythagorean triples.

Examples: 3, 4, 5; 5, 12, 13; 7, 24, 25; 8, 15, 17; 9, 40,
41;

Trigonometric functions for right triangles:

m-6

448
448
sin / cos = / tan = /
csc / sec = / cot =a/
b c a c b a
c b c a b

=
=

Useful trig identities:

2 2 2 2 2
sin cos 1 1 cot csc 1 tan =sec
2
+ = + + =

( )
( )
sin sin cos cos sin
cos cos cos sin sin

+ = +
+ =

( ) ( )
( ) ( )
( ) ( )
1 1
2 2
1 1
2 2
1 1
2 2
sin sin cos cos
cos cos cos cos
sin cos sin sin

= +
= + +
= + +

triangles in general

m-7
b
c
b
a

c
b
c
b
a

c
a


449
449

m-8
2
c
2 2
2 cos
c
a b ab + = (law of cosines)

( ) ( ) ( )
sin sin sin
a b
a b
c
c

= = (law of sines)

circles
r
s
r
tangent line
r
s
r
tangent line

1. tangent and radius are perpendicular
2.
( )
in radians s r =

450
450

3. exponentials

e
x
takes off like a rocket for large x.

so =
x n x
x x
n
de d e
e e
dx dx
= .

Taylors theorem:
2
1
2! !
n
x
e
x x
x
n
= + + + + +

This will be handy: for 1 i (so that ),
2
1 i =

( )
( ) ( ) ( )
2 3 4
1
2! 3! 4!
ix ix ix
ix
e ix = + + + + +

2 3 4
1 +
2! 3! 4!
ix
x x x
i = +

+

m-9

451
451

2 4 3
1
2! 4! 3!
x x x
i x

= + + + +

cos sin x i x = + .

4. differential equations
It is common in physics to describe how quantities change in
response to external circumstances. Because of this, calculus is the
natural language for describing the physical world. Many of our
statements about how things work are phrased as differential
equations.

An example: consider a box full of radioactive atoms. The more
atoms there are in the box, the more decays there should be per
unit time.

In addition, the greater the decay rate (the shorter the half-life) for
this species of atom, the more decays therell be per unit time.

m-10

452
452
N(t) number of surviving atoms inside the box at time t.
1 is the decay rate where is the mean life of an atom.
The number of decays per unit time is equal in magnitude, opposite
in sign, to the change per unit time of the surviving population in
the box.

Therefore, the statement . . .

The number of decays per unit time equals the number of
surviving atoms in the box times the decay rate for that species of
atom

is equivalent to the differential equation

( )
( )

dN t
N t
dt
= or
( )
( )
dN t
N t
dt
= .
How can we solve this differential equation for N(t)?
Note that
( ) dN t
dt
is proportional to N(t) since is a constant.
From a few pages back, after replacing x with t:
so
t at
t a
de de
e a
dt dt
= =
t
e

m-11

453
453
Note that
at
de
dt
is proportional to e
at
since a is a constant.
Compare:
( )
( )
dN t
N t
dt
= and
at
at
de
ae
dt
= .
It looks like and ( )
at
N t e a works: is a
solution to the differential equation. But
( )
t
N t e
=
2
t
e ( ) N t

= is also a
solution!

What to do? We have a first order differential equation, so we can
determine the unique solution with the help of one initial
condition. (A second order differential equation would require two
initial conditions.)

If we know N(0) we can use that: say there are N
0
atoms in the box
at . 0 t =

The only version of N(t) that solves the differential equation and
satisfies the initial condition is
( )
0
=
t
N t N e
.

m-12

454
454
5. Vectors
Vectors have a length and a direction.

b
a
bb
aa

Unit vectors in cartesian coordinates:

, y, z or , , x i j k

Addition, subtraction, you know about.

Scalar product
cos a b a b =

.

Component of a along b is
cos =
( )
a b b

.

Also, if = a
a x a y a z
z x y
+ +
= b
b b x y z
z x y
b + +

m-13

455
455
then

=
x x y y z
a b a b a b a b +
z
+

Cross product
= sin a b a b

Direction is given by right hand rule. For the vectors as drawn
above:

a b

(out of the paper)

b a

(into the paper)

( ) ( ) ( )

x y y x y z z y z x x z
a b a b a b z a b a b x a b a b y = + +
.

th_review.doc

m-14
copyright 2012 George Gollin PHY_325_ma
456
456
Differentiation of vectors

Go back to the basic definition of a derivative:

( )
( ) ( )
t 0
lim
a t t a t
d
a t
t dt

+

( ) u t
( ) u t t +
x
y
( ) ( ) u t t u t +

particle trajectory
( ) u t
( ) u t t +
x
y
( ) ( ) u t t u t +

particle trajectory

( ) ( )

d
v t x t
dt
=

is an example where this is useful.

m-15

457
457
Integration of vectors
Same idea: an integral is just a big sum.

( ) ( )
{ }
2
1
2
0
1
=lim
i
i
t t
i
t t
t
t
a t dt a t t
t
=
=

For example,
( ) ( )
{ }
2
1
2
0
1
=lim
i
i
t t
i
t t
t
t
v t dt v t t
t
=
=

Recall that (for constant velocity) velocity time =distance.

When is small, t
( )
v t t
i

displacement between and

.

i
t

i
+ t t

Therefore, is the net displacement between and .
( )
2
1

t
v t dt
t

1
t
2
t

m-16

458
458
Line integrals
A
B
S
A
B
S

Travel from point A to point B along the path labeled S.

Say theres an applied force acting, which might be a function of
position along the path S:
( )
F s

.

Recall that the work done by the force through a small interval s

is as long as is almost constant over the interval . F s

Net work done is
( )
all intervals
i
i
F s s

i .

In the limit that the intervals become infinitesimal,

Net work = .
B
A
F ds

m-17

459
459

6. Vector transformation properties

Under rotations

a
a
a
x
a
y
a
y
a
x
rotate vector by
aa
a a
a
x
a
y
a
y
a
x
rotate vector by

let a

a . cos sin a a a a
x y
= =

a = a since lengths of vectors are invariant under rotations.

( )
( )
cos sin
x y
a a a a = + = +

useful trig identities:

m-18

460
460
( )
( )
cos cos cos sin sin
sin sin cos cos sin

+ =
+ = +

use this to rewrite . . .

cos cos sin sin
cos sin
x
x y
a a a
a a

=
=

sin cos + cos sin
cos + sin
y
y x
a a a
a a

=
=

Note that the scalar product of two vectors is invariant under
rotations. In two dimensions:

x x y
a b a b a b = +
y

while

m-19

461
461

m-20
)
)
( )(
( )(
( )
2 2
2 2
cos sin cos sin
cos sin cos sin
cos sin
cos sin
cos sin
x x y y
x y x y
y x y x
x x y y x x y y
x y y x y x x y
x x y y
x x y y
a b a b a b
a a b b
a a b b
a b a b a b a b
a b a b a b a b
a b a b
a b a b
a b

= + =
=
+ + +
= + + +

+ + +

= + +

= +
=

Its also true in three dimensions, of course.

Matrix multiplication and rotation matrices

( ) ( )
( ) (

au bw av bx
a b u v
c d w x cu dw cv dx

+ +
=
+ +
)

also:

462
462

( )
( )

ax by
a b x
c d y
cx dy

+
=
+

We can write as a
a
x
a
y

, a
similarly to get . . .

cos sin

sin cos
a
a
x
x
a
a
y
y

Note the utility of this: successive rotations can be represented as
products of matrices

m-21

463
463
cos sin

sin cos
cos sin
sin cos
cos sin
sin cos
cos sin cos
sin cos
x
y
a a
x x
a
a
y
y
a
a

=
=

sin
sin cos
a
x
a
y

since matrix multiplication is associative.

This works fine in 3 dimensions:

a
a
x
x
a R a
y
a
a
z
z

where is 3 3 R
rotation matrix.

m-22

464
464
Lets define the rotation operators
( ) ( ) (
, , and
x y z
R R R
)

to
perform rotations of an object around the x axis by an angle , the
y axis by , and the z by .

Heres how a rotation around z will look:

z

th_review.doc

m-23

I am defining my angles so that a positive rotation moves a vector
initially along the positive x towards the positive y axis.

x

r'
r

y

copyright 2012 George Gollin PHY_325_ma
465
465
Note that
2 2
sin
x y
r r r + = and cos
z
r r = where r r

. We
have sin cos
x
r r = and sin sin
y
r r = .

Also note that .
z z
r r =

The form for
( )
z
R
that rotates r
into r
in the above diagram is

( )
cos sin 0
sin cos 0
0 0
z
R
1

since it only mixes the vectors x and y components.

Note that the unit vectors

1 0
0 1
0 0
x y z

= = =

0
0
1

are rotated into these vectors:

m-24

466
466

m-25
1
cos sin 0
sin cos 0
0 0
z z z
x R x y R y z R z z

= = = =

.

The rotation changes the components of each unit vector into the
values of the corresponding column of the rotation matrix.

A rotation around x, which mixes a vectors y and z components, is
represented by the following rotation operator. (Again, a positive
rotation moves a vector initially along the positive y towards the
positive z axis.)

( )
1 0 0
0 cos sin
0 sin cos
x
R

=

.

Finally, a rotation around y is generated by the following operator,
where a positive rotation is defined to move a vector initially along
+z towards the positive x axis:


467
467
:

( )
cos 0 sin
0 1 0
sin 0 cos
y
R

.

How unit vectors transform under rotations

(
1 0 0 x =
)
. In the rotated frame it becomes

11 12 13 11
21 22 23 21
31 31 32 33

1
0
0

R R R R
R R R R
R R R R

=

(Note that 1
st
index is row, 2
nd
is column).

So: the 1
st
column of is the same as the representation of R
x in
the new frame after rotation.

m-26

468
468
Similarly, the 2
nd
column of is the same as the representation of R
y and the 3
rd
is . z

0, 0, z 0, x y y z x = = =
R
regardless of which coordinate system

we use (scalar product is invariant under rotations) so each column
of is perpendicular to every other column.

11 12 21 22 31 32
12 13 22 23 32 33
13 11 23 21 33 31
0, 0, z 0 so...
0
0
z 0
x y y z x
x y R R R R R R
y z R R R R R R
x R R R R R R
= = =
= + + =
= + + =
= + + =

Transformations of this sort are called orthogonal transformations:
quantities that are orthogonal before remain orthogonal after.

Successive rotations in three dimensions

Successive rotations (where the two rotation axes might be
different) can be described like so:

m-27

469
469

2 1
a R R a

=

.

In general, rotations do not commute:

We can see this algebraically. For example,

m-28

470
470

( ) ( ) ( ) ( )
cos sin 0 1 0 0
sin cos 0 0 cos sin
0 0 1 0 sin cos
cos sin cos sin sin
sin cos cos cos sin
0 sin cos
z x z x
R R A R R A
A
A

=

=

but

( ) ( ) ( ) ( )
1 0 0 cos sin 0
0 cos sin sin cos 0
0 sin cos 0 0 1
cos sin 0
cos sin cos cos sin
sin sin sin cos cos
x z x z
R R A R R A
A
A

=

=

and

m-29

471
471
cos sin cos sin sin cos sin 0
sin cos cos cos sin cos sin cos cos sin
0 sin cos sin sin sin cos cos

Kronecker delta symbol
1
0
ij
i j
i j

The Kronecker delta symbol
ij
gives us another way of writing
dot products: a b a b
i j ij
ij
=

Also, since

1, 1, z z 1
0, 0, z 0
x x y y
x y y z x
= = =
= = =

we have

m-30

472
472
3
1
ij ki kj
k
R R
=
=
and
3
1
ij ik jk
k
R R
=
=
.

Levi-Civita symbol

( )
( )
1 if is an even permutation of 123 123, 312, or 231
if is an odd permutation of 123 213, 132, or 321
if any two (or more) of , , are equal
1
0
ijk
ijk
ijk
i j k
+
=

Its not hard to see that

3 3 3
1 1 1 , ,

also written as
j i k
ijk i j ijk i j
i j k i j k
a b a b k a b k
= = =
= = =

=

.

m-31

473
473
When working with these symbols, you can always write things
out explicitly, replacing the indices i, j, and k with numbers and
writing out each term in the sums. Its messy, but clear.

One can even show that
jm im
ijk lmk il jl
k
=
should you
be so moved!

Tensors

Lets say we constructed a 3 3 object out of vectors like so:
Start with ; define T so that , a b

a b a b a b
z x x x y x
T a b a b a b
z y x y y y
a b a b a b
z z z z x y

=

In terms of components, T a b
ij i j
= . (This is an outer product.)

m-32

474
474

m-33
b
If we switched to a (rotated) coordinate system, determined the
coordinates of in this system, wed calculate T using the
rule T .
, a b

b
j

a
ij i
=

How is T related to T ?

We know: and a Ra b R = =

.

In terms of components, a R
i
ik k
k
a =

and b R
j
jl l
l
b =

This comes from the definition of matrix multiplication.

As a result,

T a b R a R b
ij i j
ik k jl l
k l

= =

.

For a particular choice of each of , , , , i j k l , , , R R a b
ik il k l
are
just numbers so we can change the order of our sums:


475
475
.
ik k jl l
k l
T a b R a R b
ij i j
ik k jl l
k l
R a R b R R a b
ik k jl l ik jl k l
l k k l
R a R b

= = =

= =

If we want, we can group things:

T R T
ij
ik kl jl
l k

=

R

This is the il component of the product
th
T R

.

The transpose of is defined this way: . (Useful: the
transpose of a rotation matrix is also its inverse.)
R
T
R R
ij ji
=

This lets us write

( )
( )
T
T RT R
il
ij
lj
l

=

.

m-34

476
476
From the definition of matrix multiplication, the sum is simply the
component of the product of the matrices
th
ij and
T
RT R

.

As a result,

{ }

T T
T R R R T = =

T R

T
.

I find I prefer to write it in terms of components:

T R R
ij
ik jl kl
kl
=

.

Something that transforms this way is called a tensor (of rank 2).

Well see later how tensors can be useful.

7. Non-cartesian coordinate systems

Sometimes these will be convenient: dont be put off by their
unfamiliarity!

m-35

477
477

cylindrical coordinates
z
a
x
z
y
r
x =r cos
y =r sin
z
a
x
z
y
r
z
aa
x
z
y
r
x =r cos
y =r sin
, , r z

m-36

478
478
spherical coordinates

sin cos x r =
sin sin y r =
cos z r =
a
x
z
y
r
sin cos x r =
sin sin y r =
cos z r =
aa
x
z
y
r
, , r

The unit vectors changes which way they point as x
moves around
in space if we are working in a non-Cartesian coordinate system.
This complicates the taking of derivatives.

Note the unfortunate change in the meaning of the angle when
we switch from cylindrical to spherical coordinates.

m-37

479
479
8. partial derivatives

Well work with functions of several variables. Sometimes well
want to know how the function changes if we change one variable
while keeping all the others fixed.

For example, imagine we define a function h(x,y) that represents
the height above sea level in Champaign as a function of latitude
(x) and longitude (y).

Heres a graph of the function
( )
2 2
5 3
2 2
0.3
x y
h x y e

= .

m-38

480
480
The definition of a partial derivative with respect to y is this:

( ) ( ) ( )
, ,
,
0
h x y y h x y
h x y
lim
y y
y
+

Note that x is held constant.

This amounts to measuring the slope as (in this case) you move
parallel to the y axis.

Its easy to take partials: you treat all the variables except the
selected one as if they were constants.

For example, if
( )
2 3
, , 6
2
f x y z x xy z xz = + + then

( )
2
3 2
2
6

0 18 0
x
f xy z xz
y y y
xy z
= + +

= + +

If you look in a small region around a particular x,y point

m-39

481
481
any smooth function will look like a plane:


How much does h change if we go from x,y to + , y+ x x y ?

( ) ( )
( ) ( ) { } ( ) ( )
{ }
= , ,
= , , , ,
h h x x y y h x y
h x x y y h x y y h x y y h x y
+ +
+ + + + +

m-40

482
482
Ive just subtracted, then added the same thing.

For small x, y we have

( )
( )
( )
,
, ,
h x y y
h x x y y h x
x
x y y
+
+ +
+

from the definition of a partial derivative.

Also, as long as the partial derivative doesnt change violently with
position,

( ) ( )
,
,

h x y y
h x y
x x
+

.

As a result,
h h
h x y
x y

+

.

gradient operator

m-41

483
483
The gradient operator (in two dimensions) is defined this way:

x y
x y

+

We use it like so:

( ) ( ) ( )
, , ,
h
h x y x x y y x y
x y

= +

If we move a small distance away from the point
( )
, x y along the
direction x y
x y
= +
we find

h h
h
x y
x y

= +

.

With our definition for , we can rewrite this as

( )
h h =

.

In three (Cartesian) dimensions

m-42

484
484
( ) ( )
, , , , f x y z f
x y z
x y z x y z

+ +

.

Useful fact: the direction of h
is the direction in which

changes most rapidly.
( )
, h x y

In cylindrical coordinates , , r z we have

( ) ( )
, , , , f r z r z f r z
r r z

+ +

.

In spherical coordinates , , r we have

( ) ( )
, , , ,
sin
f r r f r
r r r

+ +

.

m-43

485
485
divergence operator

The divergence operator acts on vector fields and is defined as
specified below.

Cartesian coordinates:

( ) ( ) ( ) ( )
( )
( ) ( ) ( )

, , , , , , , ,
, , , , , ,
, ,
x y z
y x z
A A A y A z
A A A
A
x y
x y z x y z x x y z x y z
x y z x y z x y z
x y z
+ +

+ +

z

Cylindrical coordinates:

( ) ( ) ( ) ( )
( )
( )
, , , , , , , ,
1 1
, ,
r z
r
z
A r z A r z r A r z A r z
rA
A A
A r z
r r r z
+ +

+ +

Spherical coordinates:

( ) ( ) ( ) ( )
( )
( )
( )
2
2
, , , , , , , ,
sin
1 1 1
, ,
sin sin
r z
r
A r A r r A r A r
r A
A A
A r
r r r r

+ +

+ +

m-44

486
486

curl operator

The curl operator acts on vector fields. Its form in Cartesian
coordinates is this:

( )
( ) ( )
( ) ( )
( ) ( )

, , , ,
, ,
, , , ,
, , , ,
y x
y z
x z
A A
A
x y
A A
y z
A A
z x
x y z x y z
x y z z
x y z x y z
x
x y z x y z
y

+
+

The forms in cylindrical and spherical coordinates are more
complicated in appearance. See a math reference for them.

m-45

487
487
9. quadratic formula

Very handy: if then
2
0 ax bx c + + =
2
4
2
b b a
x
a

=
c

m-46

488
488

Classical Mechanics Notes

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Classical Mechanics Notes

Caricato da

Copyright:

Formati disponibili

copyright 2012, George Gollin

seen from unprimed frame. Say

seen from O like this:

copyright 2012 George Gollin PHY_325_lec_notes_01.doc

a function of time only (uniform through all space)

a function of position only (constant in time)

only (e.g., air resistance)

to rewrite Equations 1 and 2 this

and would like to solve for

. Note the maximum at 45.

= and theres no force

regardless of the initial velocity as . t

(note the constant speed)

direction, as drawn below:

depends on r since, for a particular

to point in an arbitrary direction, we have

is, in effect, forcing us into a

along a path from A to B like so:

along the route, U will

equals the reduction in potential

is a constant vector (for example, x ). If F

, then the component of momentum parallel to

is constant and can be brought

well always have so

along a direction parallel to s

must remain constant.

copyright 2012 George Gollin PHY_325_lec_notes_02.doc 2-11

is constant and the object moves in a

, then is constant. Also, since L

const. U mgz = + is a solution.

and we can integrate:

copyright 2012 George Gollin PHY_325_lec_notes_02.doc 2-20

and well have

as a sum of line integrals around

in the opposite direction along interior loops

with an associated direction using

, integrate and use

. The specific values for

and the sum runs over all particles in the system.

to be the force exerted on particle by particle

from Newtons laws.

acting on it added to the external-to-the-system

that appears and cancels

the total angular momentum of

are the externally applied forces;

and the center-of-mass

, and since both the direction of v and

is neither completely tangent to the

by applying the right amount of force to

is proportional to, and in the same

are not parallel, it is not possible to write the

is sometimes called the elastic stress tensor.

are vectors and, as shown in the

. But what about

and evolve it in hopes of ending up with a new

cos , c and sin s we can

without getting writers cramp:

as expressed above we have

and . What should we

. Note that the

are zero, since

where we take U to be zero when x is zero.

denominator on the previous

= corresponds to pushing hardest (in the positive direction)