Classical Mechanics Lectures

lecture name.
PNG
1. Introduction.pdf
2. Elementary Principles.pdf
3. Lagranges Equations.pdf
4. Hamiltons Principle.pdf
5. Central Force Problem.pdf
6. Central Force-Kepler Problem.pdf
7. Central Force-Scattering Problem.pdf
8. Rigid Body Motion.pdf
9. Rigid Body-Kinematics.pdf
10. Rigid Body-Equation of Motion.pdf
11. Rigid Body-Top and Precession.pdf
12. Oscillations.pdf
13. Forced Oscillations.pdf
14. Special Relativity.pdf
15. Relativity-Tensor,Force.pdf
16. Relativity-Relativistic Kinematics.pdf
17. Relativity-Lagrangian Formulation.pdf
18. Hamiltonian Equations.pdf
19. Hamiltonian Equations of Motion.pdf
20. Canonical Transformations.pdf
23. Continuous Systems and Fields.pdf
24. Relativistic Field Theory.pdf
Mechanics
Physics 151
Fall 2003
Masahiro Morii
Teaching Staff
!
Lecturer: Masahiro Morii

!
Section leaders: Srinivas Paruchuri and

Abdol-Reza Mansouri
!
Tuesday/Thursday 11:30 1:00. Jefferson 256
Two or three 1-hour sections per week

! Date/time to be announced
! Please fill out the student survey
Course assistant: Carol Davis

!
She will have all course materials (problem sets, etc.)
Getting the Best Out of Us

!
Speak up!
!
!
Ask questions at any time in the lectures and sections

Come to my office hours
! Hours will be posted on the door/on the web
Morii
Paruchuri
Mansouri
Davis
Office
Lyman 239
Pierce 402
MaxwellDworkin 316
Lyman 237
Phone
5-3279
5-0719
6-0318
6-1041
E-mail morii@physics srinivas@physics mansouri@deas davis@physics

!
Tell us when something in the class is not working for you
Prerequisite Courses
!
Physics 15a and 15b

!
Mathematics 21a and 21b (or equivalent)

!
Introductory Mechanics, Relativity and E&M

Multivariate Calculus, Linear Algebra and Diff. Equations
Take these prerequisites seriously

!
!
Without sufficient background, youll get lost quickly

If you still want to try, you need to get a written permission
from the Head Tutors (Howard Georgi, David Morin)
Textbook
!
Classical Mechanics, Goldstein, Poole and Safko

!
!
Required
Classic (literally) textbook. Originally published in 1950
! A must-read for serious physicists
3rd edition came out in 2001
nd edition still good (or better) Get it if you can
! 2
Will follow this textbook closely

!
Except for skipping a few advanced materials

! Its a 600-page book written for graduate students
Grading
!
Grades will be based on a weighted average of

!
!
Homework
40%
Mid-term exam 20%
! 1-hour exam. After 10 lectures
Final exam
40%
! Exam period. 3 hours
Homework
!
Problem sets are distributed on Thursdays

!
Typical format:
!
!
Reports are due at the next weeks Thursday lecture

23 problems that will be discussed at sections
34 problems you solve and turn in report
Work together in groups

!
!
Groups will be assigned according to the Survey

Each of you must turn in your own report, though
Is This Course For Me?

!
Yes if you have serious interest in Physics

!
Ill explain what we will study in the next 15 minutes

!
Definitely if you want to try a career in Physics

Plus a mini-lecture at the end
If you are not sure at the end of this lecture

!
!
Come to my office (Lyman 239) and ask questions

Ill be there everyday this week
Mechanics
Physics 151
What Is Classical Mechanics?
Mechanics
!
Mechanics concerns
!
!
Motion of objects " Velocity and acceleration

Cause of the motion " Force and energy
The objects move, but do not change their properties

!
!
Mechanics: A branch of physical

science that deals with energy and
forces and their effect on bodies
(Websters)
Idealized particles and rigid bodies

Mass and moment of inertia are all what matters
Newtons Three Laws of Motion

!
!
You remember them, right?

Principia (1687) pretty much wrapped it up
Classical vs. Modern

!
Modern in physics means 20th century

!
!
Classical Mechanics = pre-Quantum Mechanics

!
Quantum Mechanics
Relativity
We include special relativity as well as E&M
What happened between the 17th and 20th centuries?
Do We Care?
!
We know Relativity and QM are the right answers

!
Newtonian Mechanics is a human-scale approximation
Isnt that enough?

!
Why should we learn the theory that has been superseded?
(An advanced course in classical mechanics) introduces no new

physical concepts to the graduate student. It does not lead him
directly into current physics research. Nor does it aid him, to
any appreciable extent, in solving the practical mechanics
problems he encounters in the laboratory.
Goldstein, Preface to the First Edition
Why Classical Mechanics?

!
Three Good Reasons

!
Close connection to Modern Physics

! Mastering CM gives you clearer view of QM
Powerful and versatile mathematical tools
! Indispensable for advanced studies in physics
Reformulate familiar laws of physics using completely
different approaches
! Cleaner and more general formalism
! Its just cool
But first, lets go back to the 17th century
Newtonian Mechanics
!
In principle, Newtons Equation of Motion predicted

the motion of any object(s) from the force
!
All you need is a big, fast computer
In reality, life was not so easy

!
!
Intel was not founded until 1968

More fundamentally, the force may not be known
! It may depend on time, location, or even velocity
! E.g. 2 objects attracting each other by gravity
! Solving 3-body problem turns out to be impossible
" Quest for more powerful mathematics
Generalizing Equation of Motion

!
Newtonian Mechanics deals with the objects position

!
!
But there are infinite other ways to describe motion

!
E.g. a more natural way for a pendulum

x = L cos , y = L sin , z = 0, = (t )
Number of free variables may not be 3N

Lets call the new variables generalized coordinates
Goal: finding x = x(t), y = y(t), z = z(t)

3 coordinates for each object " 3N for N objects
What are the Equations of Motion for generalized

coordinates?
Lagrangian Formulation
!
Newtons Equation is about force F = ma

!
You start from F = F(x, t) for all particles

! 3N functions corresponding to 3N coordinates
Forget the force. Introduce something else

!
!
Lagrangian: L = L(q, q! )
Lagranges Equation
d L L
=0
!
dt q q
Coordinate q and its time derivative
Everything about this system is

embodied in a scalar function L
Lagrangian does not depend on a coordinate system

!
Switching to a different set of coordinates is a snap
Hamiltons Principle
!
Hamiltons Principle derives Lagranges Equation

from a simple rule:
2
The time integral of L is stationary for the path
Ldt = 0
1
taken by an actual physical system
Rather weird statement
Newtons Laws were found by induction

!
!
!
It is so because it agrees with many observations

Deriving them from a principle means knowing why it is so
Not quite that dramatic, but it does suggest deeper reason
! Eventually connected to Feynmans path integral
Besides, calculus of variations is a useful technique
Hamiltonian Formulation
!
Hamilton Equation
!
!
H
H
, p! =
q
p
Canonical variables ~ position and momentum

!
(p, q) are canonical variables

H is a function called Hamiltonian
q! =
Arent they related by p = mv?
Position and momentum as independent variables

!
Allows wider range of variable transformations than

Lagrangian formulation
! Formalism is clean, symmetric and cool
! Eerie similarity to what QM does with the uncertainty
principle
Success out of a Failure

!
Quest for tools that solve the 3-body problem failed

!
Byproducts (Lagrangians and Hamiltonians) turned

out to be the cornerstones of Quantum Mechanics
!
Unless you count invention of computers
Development of QM was guided by analogies to Lagrangian

and Hamiltonian formulations
! Pioneers of QM grew up with Classical Mechanics
Classical Mechanics is the missing link between

Newton and Schrdinger
!
It allows you to fully appreciate QM
What We Will Study

!
Lagranges Equations, Hamiltons Principle

!
!
!
!
Hamilton Equation, Canonical transformations

!
Central force problem

Rigid body motion
Oscillation
Extension to special relativity
Hamilton-Jacobi Equation
Advanced stuff
!
!
!
Classical chaos?
Perturbation theory?
Field theory?
Mechanics
Physics 151
Lecture 1
Elementary Principles
(Goldstein Chapter 1)
Goals for Today

!
Review basic principles of Newtonian Mechanics

!
Discuss motion of a single particle

!
!
Very quickly so that you dont fall asleep

Define standard notations and usages
Momenta, conservation laws, kinetic & potential energies
You (should) already know all this
Single Particle
!
Particle = object with insignificant size

!
!
!
It has mass m, and location r

!
Electron in a CRT
Baseball thrown by a pitcher
Earth orbiting the Sun
dr
= r!
Velocity v =
dt
Linear momentum p = mv = mr!
Newtons
2nd
law of motion
dp
F=
= p! = m!!
r
dt
Inertial System
!
The origin O of r is somewhat arbitrary

!
A choice of origin " a reference frame
Inertial system = a reference frame in which F = p!

holds
!
Newtons 2nd Law should be stated more precisely as
There exist reference frames in which the time derivative

of the linear momentum equals to the force
!
And there are infinite number of such frames
Inertial Systems
!
Consider two inertial systems A and B

!
!
A particle is at rA in A, rB in B
Origin of A is at rB rA in B
F = m!r!A = m!r!B !r!B !r!A = 0
r!B r! A = const
rA
OA
rB rA
rB
OB
Any two inertial systems are moving relative to each

other at a constant velocity
!
Equivalence of such systems was pointed out by Galileo

! Hence the name Galilean system
Angular Momentum
!
Define
Angular momentum L = r p
! Moment of force (= torque) N = r F
From F = p! one can deduce N = L!
!
The order
matters!
Too easy to write down here
Subtlety: the definitions depend on the origin O

!
!
Because r is defined from O

The equation N = L! holds for any origin
Momentum Conservation
!
Two conservation theorems follow

!
From F = p!
If the total force F is zero,

the linear momentum p is conserved
!
From N = L!
If the total torque N is zero,

the angular momentum L is conserved
This is really getting too easy
Work by External Force

!
Particle moves from point 1 to 2 under force F

!
Work W12 done by the force F is defined by

2
W12 = F ds
1
ds
F
mv 2
One can define the kinetic energy T
2
! Then derive W12 = T2 T1
Work done equals to the change in the kinetic energy
Conservative Force
!
If W12 is the same for any possible

path from 1 to 2, the force F is
conservative
!
W12 depends only on the end points, 1

not on the path
Equivalently, if you make a closed

loop, the total work is zero
F ds =
F ds + F ds = 0
2
2
1
Potential Energy
!
F is conservative F is expressed by F = V (r )
!
V is the potential energy

2
Work W12 is then expressed by W12 = 1 F ds = V1 V2

!
Which was equal to T2 T1
T1 + V1 = T2 + V2
If the force is conservative, the total energy T + V is conserved

Energy Conservation Theorem
Summary
!
Reviewed basic principles of Newtonian Mechanics

!
!
I hope everything looked familiar, if boring

!

It will get better from here
Next: multi-particle system & constraints
Mechanics
Physics 151
Lecture 2
Elementary Principles
Administravia
!
First Problem Set

!
If you havent filled the Survey, please do it

!
3 problems for the section

! Work on them before coming to your section!
3 for the report (due next week)
We need it for sectioning and study-grouping
Section time: Tue. 6 PM, 7 PM and Wed. 5 PM

!
!
If none of these slots works for you, let me know

Sectioning will be announced on Monday by email
! We will also assign you into study groups (~6 each)
What We Did Last Time

!
Reviewed basic principles of Newtonian Mechanics

!
!

Concentrated on the motion of a single particle
Goals for Today

!
Single " multi-particle system

!
Newtons 3rd Law
Introduce constraints
!
Force between particles

! Laws of action and reaction
Holonomic and nonholonomic constraints
Introduce Lagranges Equation
System of Particles
!
More than one particles? " Just add indices!

Fi = p! i
N i = L! i
Subtlety: F may be working between particles

!
Distinguish between internal and external forces

Fi = F ji + Fi( e )
j
Force acting
on particle i
Force from outside

Force from particle j
Now add up over i to see the overall picture
Sum of Particles
(e)
(e)
F
=
F
+
F
=
F
+
F
+
F
i ji i ( ji ij ) i
i
i, j
i j
i< j
This term vanishes if Fij = F ji

!
Weak law of action and reaction
(e)
F
=
F
i i
i
Forces two particle exert on each other are equal and opposite
C.f. the strong law of action and reaction
Forces two particle exert on each other are equal, opposite,
and along the line joining the particles
!
Sum of Particles
!
Now consider the equations of motion
F = F
(e)
i
d2
= p! i = 2
dt
i
Define center of mass
m r
i i
mr mr
R
=
M
m
i i
i i
!! = F ( e ) F ( e )
MR
i
i
Center of mass moves like a particle of mass M under total

external force F(e)
Total Linear Momentum

!
The sum of the linear momenta is
!
P = p i = mi r!i = MR
i
Taking the time derivative

!! = F ( e )
P! = MR
Newtons equation
of motion for the
center of mass
Conservation of total linear momentum
If the total external force F(e) is zero, the total linear momentum
P is conserved
Assumed weak law of action & reaction
Total Angular Momentum

!
The sum of the angular momenta is

L = L i = ri p i
i
Take time derivative and use p! i = Fi = F ji + Fi(e )

j
L! = ri F ji + ri Fi( e )
i, j
i j
Total external
torque
This term vanishes only if Fji satisfies the strong law of

action and reaction

!
Assuming strong law of action and reaction

L! = ri Fi( e ) = N i( e ) = N ( e )
i
" Conservation of total angular momentum

If the total external torque N(e) is zero, the total angular
momentum L is conserved
A multi-particle system (= extended object) can be treated

as if it were a single particle if the internal forces obey the
strong law of action and reaction
Laws of Action and Reaction

!
F12
E.g. Lorenz force felt by moving charges

Conservation of linear & angular momenta fails
Take into account the EM field

Particles exchange forces with the field
! The field itself has linear & angular momenta
" Conservation laws restored
!
F21 = 0
Gravity, electrostatic force
There are rare exceptions

!
+Q
Most forces we know obey strong law of

action and reaction
v2
+Q
v1
Conservation Laws
Weak law of action-reaction
Conservation of P
Strong law of action-reaction
Conservation of L
We will see (in 2 lectures) that P and L must be

conserved if the laws of physics are isotropic in space
!
!
No special origin
No special orientation
If we accept these symmetries as fundamental

principles, all forces must satisfy the action-reaction
laws " Proof of Newtons 3rd Law

!
Define particle is position from the center of mass

ri = ri R
!
!
!
Also define the velocities vi = r!i v = R
Calculate the total angular momentum
L = ri p i = (ri + R ) mi (vi + v )
i
L = R Mv + ri mi vi
i
Angular momentum of
motion concentrated at
the center of mass
Angular momentum
of motion around the
center of mass
Kinetic Energy
!
The work done by force W12 = 1 Fi ds i

2
Positions 1 and 2 are now configurations (sets of positions)
Use equations of motion to derive

1
W12 = T2 T1 where T = mi vi2
i 2
One can split T into two pieces

1
1
1
T = mi ( v + vi ) ( v + vi ) Mv 2 + mi vi2
2
i 2
i 2
Motion concentrated at
the center of mass
Motion around the

center of mass
Potential Energy
!
yi
zi
Assume conservative external force Fi( e ) = iVi

i
xi
F dsi = iVi dsi = Vi 1

2
(e)
i
Assume also conservative internal forces F ji = iVij

!
To satisfy strong law of action/reaction

Potential depends only
Vij = Vij (| ri r j |)
on the distance
i, j
i j
F ji dsi = iVij dsi

2
i, j
i j
Bit of work
2
1
Vij
1
2 i, j
i j
Energy Conservation
!
If all forces are conservative, one can define total

potential energy: V = V + 1 V
2 i, j
ij
i j
Then the total energy T + V is conserved
The second term is internal potential energy

!
!
It depends on the distances between all pairs of particles

Constant if particles relative configuration is fixed
" Rigid bodies
Constraints
!
(e)
!
!
m
r
=
F
=
F
+ F ji assumes
Equation of motion i i
i
i
j
that particles can
move anywhere in space
!
!
Not generally true

! In fact never true Free space is an idealization
Amusement-park ride constrained (hopefully) on a rail
Billiard balls on a pool table
How can we accommodate constraints in the equation

of motion?
!
Depends on the type of the constraint
Holonomic Constraints
!
Constraints may be expressed by

f (r1 , r2 , r3 , , t ) = 0
Particle on the x-y plane z = 0
Rigid body (ri r j ) 2 cij2 = 0
All other cases are called nonholonomic

!
!
!
A holonomic
constraint
It means we dont really want to mess with it

May be inequalities such as z 0
May depend on derivatives such as r!i
We will deal only with holonomic constraints
Independent Variables
!
A holonomic constraint reduces the number of

independent variables by 1
!
!
If z = 0, you are left with only x and y

You may be able to solve the constraint for one variable
f (r1 , r2 , r3 , , t ) = 0
x1 = g ( y1 , z1 , r2 , r3 , , t )
Then you can drop this variable

You may have to switch to a different set of variables
2
2
2
2
! For a particle on a sphere x + y + z = c
a good choice is (, )
!
New set of variables " Generalized coordinates
Generalized Coordinates
!
N particles have 3N degrees of freedom

!
!
Introducing k holonomic constraints reduces it to 3N k

Using generalized coordinates q1, q2,, q3N k
ri = ri (q1 , q2 , q3 , , t )
Transformation equations from (rl) to (ql)
x = c sin cos
Example: y = c sin sin

z = c cos
Transformation
from (x, y, z)
to (, )
Now What?
!
We know the equations of motion for (ri)

mi!r!i = Fi = Fi( e ) + F ji
j
We know how to include constraints by switching to

generalized coordinates
ri = ri (q1 , q2 , q3 , , t )
How can we transform the equation of motion to the

generalized coordinates?
Lagranges Equations
Why Constraints?
!
Constraint is an idealized classical concept

!
Nothing is perfectly constrained in QM Uncertainty
How useful is it to switch between coordinates?

More than it seems
Constraints and Force

!
A holonomic constraint is an infinitely strong force

!
Or an infinitely high potential wall
Reality is always smoother

!
V(r)
E.g. electron of a hydrogen atom

free
V(r)
Its still true that the

electron feels strong
radial (binding) force,
while it can move
freely around the
nucleus
constrained
Binding force
makes the r
direction special
Force and Symmetry

!
Without forces, all coordinate systems are equal

!
Forces break the symmetry

!
x-y-z system is the simplest

Some coordinate system works better than others
Generalized coordinates offer natural way of handing

systems with such forces
Constraints are extreme cases
!
We develop our technique with them

OK, back to the business
Lagranges Equations
d L
dt q! j
L
=0

q j
Recipe
Kinetic energy
L(q, q! , t ) T V
Potential energy
Lagrangian
!
Express L = T V in terms of generalized coordinates

{q j } , their time-derivatives {q! j }, and time t
!
!
The potential V = V(q, t) must exist

i.e. all forces must be conservative
Lets do a quick example to see how it works
d L
dt q! j
Ex: Particle on a Line

!
L
=0

q j
A particle moving on the x-axis x = x(t ), y = 0, z = 0

!
Kinetic and potential energies:

m
T = x! 2 V = V ( x)
2
Lagranges Eqn
L=
m 2
x! V ( x)
2
V
=0
mx!! +
x
V
Equivalent to Newtons Eqn given that Fx =
x
OK, it works
Summary
!
Discussed multi-particle systems

!
Introduced constraints
!
!
Internal and external forces

Holonomic and nonholonomic constraints
Generalized coordinates
Introduced Lagranges Equations

!
Next: Prove that Lagranges and Newtons Equations are

equivalent
Mechanics
Physics 151
Lecture 3
Lagranges Equations
Hamiltons Principle
(Chapter 2)

!
Discussed multi-particle systems

!
Introduced constraints
!
Internal and external forces

Generalized coordinates
Introduced Lagranges Equations

!
... and didnt do the derivation
" Lets pick it up and start from there
Todays Goals
!
Derive Lagranges Eqn from Newtons Eqn

!
!
Use DAlemberts principle

There will be a few assumptions
! Will make them clear as we go
Introduce Hamiltons Principle

!
!
!
Equivalent to Lagranges Equations

! Which in turn is equivalent to Newtons Equations
Does not depend on coordinates by construction
Derivation in the next lecture
Lagranges Equations
d L
dt q! j
L
=0

q j
Recipe
Kinetic energy
L(q, q! , t ) T V
Potential energy
Lagrangian
Express L = T V in terms of generalized coordinates

{q j } , their time-derivatives {q! j }, and time t
!
!
The potential V = V(q, t) must exist

i.e. all forces must be conservative
Virtual Displacement
!
Consider a system with constraints r1 = r1 (q1 , q2 ,..., qn , t )
r = r (q , q ,..., q , t )
2 2 1 2
! Ordinary coordinates ri (i = 1...N)
n
"
! Generalized coordinates qj (j = 1...n)
rN = rN (q1 , q2 ,..., qn , t )
! Imagine moving all the particles
slightly ri ri + ri q j q j + q j
!
Virtual displacement
Note that ri must satisfy the constraints
ri
ri =
qj
j q j
3N coordinates
not independent
n coordinates
independent
DAlemberts Principle
!
From Newtons Equation of Motion

Fi = p! i
Fi p! i = 0
Part of the force Fi must be due to constraints

Fi = Fi( a ) + fi
applied force
!
constraint force
Applied force is known Fi( a ) = Fi( a ) (r1 , r2 ,..., ri ,..., rN , t )
Constraint force fi (usually) does no work
Movement is perpendicular to the force fi ri = 0

! Exception: friction
!
Now multiply Fi( a ) + fi p! i = 0 by ri and sum over i
(F
(a)
i
p! i ) ri = 0
constraint force is out of the game.

You can forget (a)
!
!
Force of constraints dropped out because fi ri = 0

Called DAlemberts Principle (1743)
Now we switch from ri to qj
ri
1st term = Fi
q j = Q j q j
i
j q j
j
!
!
Unit of Qj not always [force]

Qj qj is always [work]
ri
Q j Fi
q j
i
Generalized force
2nd term = p! i ri = p! i
i
ri
d vi2 vi2
A bit of work can show !r!i

!
q j
dt q j 2 q j 2
d T
=
q!
dt
j
j
!
ri
ri
!!
q j = mi ri
qj
q j
q j
i, j
T

q j
q j
DAlemberts Principle becomes

d T
j dt q!
j
T
Q j q j = 0

q j
mvi2
T
2
i
Lagranges Equations
d T
j dt q!
j
!
T
Q j q j = 0

q j
Generalized coordinates qj are independent

d T
dt q! j
These are free
T
= Qj

q j
Almost there!
Assume forces are conservative Fi = iV

ri
ri
V
= iV
=
Q j Fi
q j
q j
q j
i
i
Throw this
back in
Lagranges Equations
d T
dt q! j
!
(T V )
=0

q j
Assume that V does not depend on q! j

Finally
d L
dt q! j
L
=0

q j
V
=0
q! j
L = T (q j , q! j , t ) V (q j , t )
Done!
Assumptions We Made
!
Constraints are holonomic

!
We always assume this
Constraint forces do no work

!
ri = ri ( q1 , q2 ,..., qn , t )
fi ri = 0
Forget frictions
Applied forces are conservative
Fi = iV
Lagranges Eqn. itself is OK if V depends explicitly on t

V
!
=0
! Potential V does not depend on q j
q! j
!
Will review the last assumption later
Example: Time-Dependent
!
Transformation functions may depend on t

!
!
Generalized coordinate system may move

E.g. coordinate system fixed to the Earth
An example
ri = ri ( q j , t )
spring constant K
natural length l
mass m on a rail
l+r
angular velocity
!
x = (l + r ) cos t
Transformation functions:
y = (l + r ) sin t
m
m
Kinetic energy T = { x! 2 + y! 2 } = {r! 2 + (l + r ) 2 2 }
2
2
K 2
Potential energy V = r
2
m 2
K 2
2 2
L = {r! + (l + r ) } r
2
2
d L L
2
!!
Lagranges Equation
mr
m
(l + r ) + Kr = 0
dt r! r
d L L
2
!!
mr
m
(l + r ) + Kr = 0
dt r! r
2
m
l
2
mr!! + ( K m ) r
=0
2
K m
!
!
K m 2
m
If K > m2, a harmonic oscillator with =

! Center of oscillation is shifted by
If K < m2, moves away exponentially
If K = m2, velocity is constant
! Centripetal force balances with the spring force
Note on Arbitrarity
!
Lagrangian is not unique for a given system

!
If a Lagrangian L describes a system

dF (q, t )
L = L +
works as well for any function F
dt
One can prove
d dF dF
dF F
F
!
=
0
=
q+
using
!
dt q dt q dt
dt q
t
Assumptions We Made
!
Constraints are holonomic

!
We always assume this
Constraint forces do no work

!
ri = ri ( q1 , q2 ,..., qn , t )
fi ri = 0
Forget frictions
Applied forces are conservative
Fi = iV
Lagranges Eqn. itself is OK if V depends explicitly on t

V
!
=0
! Potential V does not depend on q j
q! j
!
Lets review the last assumption
Velocity-Dependent Potential
V
V
= 0 so that
! We assumed Q j =
and
q! j
q j
This had to be 0
d T T
d (T V ) (T V )
= Qj
=0

q j
dt q! j q j
dt q! j
!
We could do the same if we had

U d U
Qj =
+
q j dt q! j
U = U (q j , q! j , t )
L = T (q j , q! j , t ) U (q j , q! j , t )
Generalized,
or velocitydependent
potential
EM Force on Particle
!
Lorentz force on a charged particle

F = q[E + ( v B)]
!
Velocity-dependent.
Cant find a usual
potential V
E and B fields are given by

A
E =
Physics 15b
B = A
t
Force is v-dependent " Need a v-dependent potential
U = q qA v works check
1 2
! Lagrangian is L = mv q + qA v
2
Monogenic System
!
If all forces in a system are derived from a generalized

potential,
U d U
its called a monogenic system
Qj =
+
!
!
dt q! j
A monogenic system is conservative only if U = U (q)

!
U is a function of q, q! , t
Lorentz force is monogenic
q j
U U
=
=0
Or
q!
t
Lagranges Equation works on a monogenic system
Hamiltons Principle
!
We derived Lagranges Eqn from Newtons Eqn using

a differential principle
!
DAlemberts principle uses infinitesimal displacements
Its possible to do it with an integral principle

Hamiltons Principle
Configuration Space
!
Generalized coordinates q1,...,qn fully describe the

systems configuration at any moment
configuration
Imagine an n-dimensional space
!
Each point in this space (q1,...,qn)

corresponds to one configuration of the system
Time evolution of the system " A curve in the
configuration space
real space
configuration space
space
Action Integral
!
!
A system is moving as q j = q j (t ) j = 1...n

Lagrangian is L(q, q! , t ) = L(q(t ), q! (t ), t )
integrate
!
!
t2
I = Ldt
t1
Action, or action integral
Action I depends on the entire path from t1 to t2

Choice of coordinates qj does not matter
! Action is invariant under coordinate transformation
Hamiltons Principle
The action integral of a physical system is stationary
for the actual path
!
This is equivalent to Lagranges Equations

!
We will prove this
We will also define stationary
Three equivalent formulations

!
!
!
Newtons Eqn depends explicitly on x-y-z coordinates

Lagranges Eqn is same for any generalized coordinates
Hamiltons Principle refers to no coordinates
! Everything is in the action integral
Hamiltons Principle is more fundamental
probably...
Stationary
!
Consider two paths that are close to each other

Difference is infinitesimal
t2
Stationary means that the

difference of the action integrals is
zero to the 1st order of q(t)
t2
t2
t1
t1
I = L(q + q, q! + q! , t )dt L(q, q! , t )dt = 0
Almost same as saying minimum

!
q(t )
Similar to first derivative = 0
configuration space
It could as well be maximum
q (t ) + q (t )
t1
q(t1 ) = q(t2 ) = 0
Infinitesimal Path Difference

!
Whats q(t)?
!
!
!
configuration space
t2
Its arbitrary sort of

It has to be zero at t1 and t2
Its well-behaving
Dont worry
q(t )
too much
Continuous, non-singular,
continuous 1st and 2nd derivatives
!
q (t ) + q (t )
t1
Have to shrink it to zero

!
Trick: write it as q(t ) = (t )

! is a parameter, which well make " 0
! (t) is an arbitrary well-behaving function (t1 ) = (t2 ) = 0
Hamilton " Lagrange

!
To derive Lagranges Eqns from Hamiltons Principle

t2
t2
t1
t1

!
t2
Define I ( ) t L(q (t ) + (t ), q! (t ) + ! (t ), t )dt

1
d
=0
I is then lim [ I ( ) I (0)]
I
We must show that
= 0 leads to Lagranges Eqns
=0
A bit of work. Will do it on Thursday
Summary
!
Derived Lagranges Eqn from Newtons Eqn

!
Assumptions we made:
!
!
!
Using DAlemberts Principle Differential approach

Constraints are holonomic " Generalized coordinates
Forces of constraints do no work " No frictions
Other forces are monogenic " Generalized potential
Introduced Hamiltons Principle

!
!
!
U d U
Qj =
+
q j dt q! j
Integral approach
Defined the action integral and stationary
Derivation in the next lecture
Mechanics
Physics 151
Lecture 4
Hamiltons Principle
(Chapter 2)
Administravia
!
Problem Set #1 due

!
Problem Set #2 is here

!
Solutions will be posted on the web after this lecture

Due next Thursday
Next lecture (Tuesday) will be given by Srinivas and

Abdol-Reza
!
I will be attending a workshop at Stanford

!
Derived Lagranges Eqn from Newtons Eqn

!
Using DAlemberts Principle = differential approach
Lagranges Equations work if

!
!
!
Constraints are holonomic " Generalized coordinates

Forces of constraints do no work " No frictions
Other forces are monogenic " Generalized potential
U d U
Qj =
+
q j dt q! j
Todays Goals
!
Discuss Hamiltons Principle

!
!
!
Derive Lagranges Eqn from Hamiltons Principle

Calculus of variation
Looks unfamiliar, but not so difficult
Discuss conservation laws again

!
!
!
Using Lagrangian formalism

Linear, angular momenta
Connection between symmetry, invariance of the
Lagrangian, and conservation of generalized momentum
Configuration Space
!
Generalized coordinates q1,...,qn fully describe the

systems configuration at any moment
configuration
Imagine an n-dimensional space
!
Each point in this space (q1,...,qn)

corresponds to one configuration of the system
Time evolution of the system " A curve in the
configuration space
real space
configuration space
space
Action Integral
!
!
A system is moving as q j = q j (t ) j = 1...n

Lagrangian is L(q, q! , t ) = L(q(t ), q! (t ), t ) = L(t )
integrate
!
!
t2
I = Ldt
t1
Action, or action integral
Action I depends on the entire path from t1 to t2

Choice of coordinates qj does not matter
! Action is invariant under coordinate transformation
Hamiltons Principle
The action integral of a physical system is stationary
for the actual path
!
This is equivalent to Lagranges Equations

!
We will prove this
We will also define stationary
Three equivalent formulations

!
!
!
Newtons Eqn depends explicitly on x-y-z coordinates

Lagranges Eqn is same for any generalized coordinates
Hamiltons Principle refers to no coordinates
! Everything is in the action integral
Hamiltons Principle is more fundamental
probably...
Stationary
!
Consider two paths that are close to each other

Difference is infinitesimal
t2
Stationary means that the

difference of the action integrals is
zero to the 1st order of q(t)
t2
t2
t1
t1
Almost same as saying minimum

!
q(t )
Similar to first derivative = 0
configuration space
It could as well be maximum
q (t ) + q (t )
t1
q(t1 ) = q (t2 ) = 0
Infinitesimal Path Difference

!
Whats q(t)?
!
!
!
configuration space
t2
Its arbitrary sort of

It has to be zero at t1 and t2
Its well-behaving
Dont worry
q(t )
too much
continuous 1st and 2nd derivatives
!
q (t ) + q (t )
t1
Have to shrink it to zero

!
Trick: write it as q(t ) = (t )

! is a parameter, which well make " 0
! (t) is an arbitrary well-behaving function (t1 ) = (t2 ) = 0
Hamilton " Lagrange

!
Consider 1 generalized coordinate q

!
!
Add q(t) to q(t), then make q(t) " 0

Do this by q(t ) = (t )
configuration space
t2
q (t , ) = q(t ) + (t )
!
!
is a parameter " 0
(t) is an arbitrary well-behaving
function
(t1 ) = (t2 ) = 0
!
q(t )
continuous ' and ''
t2
Lets define I ( ) t L(q(t , ), q! (t , ), t )dt

1
q (t ) + q (t )
t1
NB: this also
depends on (t)
Calculus of Variations
!
t2
Lets define I ( ) = t L(q(t , ), q! (t , ), t )dt

1
If the action is stationary

t2 L dq
dI ( )
L dq!
dt
=
+
t
1
d
q d q! d
Some work!
t2
t1
L d L dq
dt

q dt q! d
NB: this also

depends on (t)
dI ( )
=0
d =0
for any (t)

q (t , ) = q(t ) + (t )
= (t )
Arbitrary function
Lagranges Equation
Fundamental lemma
x2
x1
M ( x) ( x)dx = 0 for any ( x)

!
M ( x) = 0 for x1 < x < x2
We got
t2
t1
L d L
(t )dt = 0

q dt q!
L d L
=0
q dt q!
Done!
Notation of Variation
!
For shorthand, we use for infinitesimal variation

!
I.e. -derivative at = 0
t2
d
dI
I
L(q(t , ), q! (t , ), t )dt d
d =
t
d 1
d =0
dq
d = (t )d
d =0
q
!
Hamiltons Principle can be written as

I =
t2
t1
L d L
qdt = 0

q dt q!
Going Multi-Coordinates
!
Trivial to expand q " (q1, q2, , qn)

!
See Goldstein Section 2.3
I =
t2
t1
L d L
i q dt q! qi dt = 0
i
i
= 0 for each i
!
Assumption: q1, q2, are arbitrary and independent

! Not true for x-y-z coordinates if there are constraints
! True for generalized coordinates if the system is
holonomic
Hamiltons Principle
t2
I = L(q, q! , t )dt = 0
t1
Action I describes the entire motion of the system

!
Action I does not depend on the choice of the

coordinates
!
It is sufficient to derive the equations of motion
Lagrange formalism is coordinate invariant
Adding dF/dt to L would add F(t2) F(t1) to I

!
It wouldnt affect I Variations are 0 at t1 and t2
dF (q, t )
Arbitrarity of L is obvious L = L +
dt
Calculus of Variation
!
Technique has wider applications

!
x2
In general for J = f ( y ( x), y( x), x)dx

x1
J =0
dy
y
dx
f d f

=0
y dx y
Examples in Goldstein Section 2.2
Most famous: the brachistochrone problem

Fastest path via gravity
Conservation Laws
!
Weve seen (in Lectures 1&2) conservation of linear,

angular momenta and energy in Newtonian mechanics
!
!
How do they work with Lagranges equations?

Should better be the same
Well find a few differences and assumptions

!
They are, in fact, limitations we ignored so far
!
Lets consider a simple system

mi ( x! + y! + z! )
L = T V =
V ( xi , yi , zi , t )
2
i
2
i
L
= mi x!i = pix
x!i
!
2
i
2
i
Momentum
Potential does
not depend on
velocity
L
V
=
= Fix
xi
xi
Force
Momentum pix conserved if V does not depend on xi
Now try to generalize from here
Generalized Momentum
L
! Lets call p j
the generalized momentum
q! j
!
Also known as canonical or conjugate momentum
Equals to usual momentum for simple x-y-z coordinates

dp j L
=0
! Lagranges equation becomes
dt q j
!
pj is conserved if L does not depend explicitly on qj
Such qj is called cyclic (or ignorable)
Generalized momentum associated

with a cyclic coordinate is conserved
Linear momentum
conservation is a
special case
Generalized Momentum
!
L
pj
q! j
Generalized momentum may not look like linear

momentum
!
Dimension may vary, if qj is not a space coordinate

! pjqj always has the dimension of action (= work time)
Form may vary if V depends on velocity
! Example: a particle in EM field
1
L = mv 2 q + qA v
px = mx! + qAx
2
Extra term due to velocitydependent potential
Symmetry
!
Linear momentum p = (px, py, pz) is conjugate of

(x, y, z) coordinates
!
Conserved if Lagrangian does not depend explicitly on

position
I.e. if Lagrangian is invariant under space translation
( x , y , z ) ( x + x , y + y , z + z )
Such a system is called symmetric under space translation
Symmetry of a system = Invariance of Lagrangian

" Conservation of conjugate momentum
!
Lets study an example of angular momentum
Angular Momentum
!
ri = ri (q1 ,..., qn , t )
Consider a multi-particle system

!
Suppose q1 turns the whole system around

! Example: in ri = ( xi , yi , zi ) = ( ri cos , ri sin , zi )
n
Assume V does not depend on !
r
i
Conjugate momentum is
L T
p
=
!
!
bit of
work
= n Li = n L
d
zi
ri ( )
Axis of rotation
ri ( + d )
Total angular
momentum
Bit of Work
ri = ri ( , q2 ,..., qn , t )
ri ! n ri
r
+
q!k + i
r!i =
t
k = 2 qk
mi
T = r!i r!i
2
i
!
!
r!i ri
=
!

r!i
ri
T
= mi r!i
= mi r!i
!
!
i
i
dri is perpendicular to both n and ri

Size of dri is ri sin d ri
= n ri
= mi r!i (n ri )
ri
= mi n (ri r!i ) = n Li
i
dri
Angular Momentum
!
Angular momentum is conserved if the system is

symmetric under rotation
!
How does this relate to the total torque N?

Generalized
force
!
L
Q
This must be zero

if is cyclic
T cannot depend on Rotating doesnt change vi2

ri
L
V
=
= Fi
= Fi ( n ri ) = n ri Fi
i
i
i
Total torque is zero along the axis of symmetry
Torque
Conservation Laws
!
Following statements are equivalent:

!
!
!
!
System is symmetric wrt a generalized coordinate

The coordinate is cyclic (does not appear in Lagrangian)
The conjugate generalized momentum is conserved
The associated generalized force is zero
Symmetry
Coordinate
Momentum
Force
Spatial translation
Distance along an axis
Linear
Force
Rotation
Angle around an axis
Angular
Torque
Summary
!
Derived Lagranges Eqn from Hamiltons Principle

!
Discussed conservation laws

!
!
Calculus of variation
L
Generalized (conjugate) momentum p j
q! j
Symmetry of the system
" Invariance of the Lagrangian
" Conservation of momentum
We are almost done with the basic concepts

!
!
Finish up next Tuesday with energy conservation

Some applications are in order " Central force problem
Mechanics
Physics 151
Lecture 5
Central Force Problem
(Chapter 3)
Introduced Hamiltons Principle
Discussed conservation laws
Action integral is stationary for the actual path

Derived Lagranges Equations
Used calculus of variation
Generalized (conjugate) momentum
Symmetry Invariance Momentum conservation
We are almost done with the basic concepts
One more thing to cover
Goals for Today
Energy conservation
Central force problem First application
Motion of a particle under a central force

Simplify the problem using angular momentum conservation
Discuss qualitative behavior of the solution
Define energy function

Subtle difference from the Newtonian version
Use energy conservation

Distinguish bounded/unbounded orbits
Actual solution Thursday
Energy Conservation
Consider time derivative of Lagrangian

dL(q, q , t )
L dq j
L dq j L
=
+
+
j dt
dt
t
j q j dt
j q
L d L
Using Lagranges equation
=
q
q
dt
j
j
one can derive
L
L
d
L +
=0
q j
t
dt j
q j
Define this as energy function h(q, q , t )
Conserved if Lagrangian does not depend explicitly on t
Energy Function?
h(q, q , t ) q j
j
Does energy function represent the total energy?
L
L
q j
Lets try an easy example first
Single particle moving along x axis

mx 2
L=
V ( x)
2
How general is this?
h = mx 2 L
mx 2
=
+ V ( x) = T + V
2
Total
energy
Energy Function
h(q, q , t ) q j
j
L
L
q j
Suppose L can be written as

L(q, q , t ) = L0 (q, t ) + L1 (q, q , t ) + L2 (q, q , t )
True in most cases

of interest
Derivatives satisfy
L0
=0
q j
L1
j q j q = L1
j
2nd order in q
1st order in q
L2
j q j q = 2 L2
j
L
h(q, q , t ) q j
L = L2 L0
q j
j
Eulers
theorem
Energy Function
h(q, q , t ) = L2 L0
L = T V
Energy function equals to the total energy T + V if

T = L2 and V = L0
1st condition is satisfied if transformation from ri to qj is

time-independent
2nd condition holds if the potential is velocity-independent
No frictions Friction would dissipate energy
Lets look into the 1st condition
Kinetic Energy
mi 2
T = ri
2
i
ri = ri (q1 ,..., qn )
mi 2
mi

i 2 ri = i 2
ri ri
mi ri ri

q j qk = q j qk
2 q j qk
j , k q j qk
j ,k
i
dri
ri
q j
=
Using the chain rule
dt
j q j
2nd order homogeneous
Time-independent
No q
This wouldnt work if ri = ri (q1 ,..., qn , t ) because

dri
ri
ri

=
qj +
t
dt
j q j
Energy Conservation
Energy function equals to the total energy if
Constraints are time-independent

Kinetic energy T is 2nd order homogeneous function of
the velocities
Potential V is velocity-independent
Energy function is conserved if
L
L
h(q, q , t ) q j
q j
j
Lagrangian does not depend explicitly on time
These are restatement of the energy conservation

theorem in a more general framework
Conditions are clearly defined
Central Force Problem
Consider a particle under a central force
Force F parallel to r
Assume F is conservative F = V (r )
V is function of |r| if F is central
Such systems are quite common
Planet around the Sun

Satellite around the Earth
Electron around a nucleus
These examples assume the body at the center is heavy
and does not move
Two-Body Problem
Consider two particles without external force
r1 and r2 relative to center of mass
m1
r1
Lagrangian is
2 2 m r 2
(m1 + m2 )R
L=
+ i i V (r )
2
2
i =1
Motion of CoM
Motion of particles
around CoM
m2
r1 =
r
(m1 + m2 )
m1
r2 =
r
(m1 + m2 )
CoM
r m2
2
Potential is function of
|r| = |r2 r1|
Strong law of action and reaction
mi ri2 1 m1m2
2

r
=
2
2 (m1 + m2 )
i =1
2
Two-Body Central Force

2 1 mm
(m1 + m2 )R
1 2
L=
+
r 2 V (r )
2
2 (m1 + m2 )
m1
r
R is cyclic
CoM moves at a constant velocity

Move O to CoM and forget about it
L=
CoM
m2
1 m1m2
r 2 V (r )
2 (m1 + m2 )
Relative motion of two particles is identical to the motion of

one particle in a central-force potential
1
1
1
m1m2
=
+
Reduced mass =
or
m1 m2
(m1 + m2 )
Hydrogen and Positronium
Positronium is a bound state of a positron

and an electron
Similar to hydrogen except m(p) >> m(e+)

Potential V(r) is identical
Turn them into central force problem
me me
me
positronium =
=
(me + me ) 2
m p me
hydrogen =
me
( m p + me )
Spectrum of positronium identical to

hydrogen with me me/2
e
e+
e
p
q2
V (r ) =
r
Spherical Symmetry
Central-force system is spherically symmetric
Angular momentum is conserved L = r p = const
It can be rotated around any axis through the origin

2 ) V (r ) doesnt depend on the
Lagrangian L = T (r
direction
Direction of L is fixed
r L by definition r is always in a plane
Choose polar coordinates
Polar axis = direction of L

r = r ( r , , ) = r (r , )
Azimuth
Zenith = 1/2
L
O
More Formally
Lagrangian in polar coordinates r = r (r , , )

m 2 2 2 2 2 2
L = T V = ( r + r sin + r ) V (r )
2
is cyclic, but is not

d L L
2
2 ) = 0

=
mr
(
sin
cos
dt
We can choose the polar axis so that the initial condition is
= 2 , = 0
2nd term vanishes
= 0
Now is constant. We can forget about it
Angular Momentum
m 2 2 2
L = T V = (r + r ) V (r )
2
is cyclic. Conjugate momentum p conserves
L
Magnitude of
p = = mr 2 = const l
angular momentum
Alternatively
dr
dA 1 2
Areal velocity
= r = const
dt 2
Keplers 2nd law
True for any central force
dA
Radial Motion
m 2 2 2
L = T V = (r + r ) V (r )
2
d
V ( r )
2

(mr ) mr +
=0
Lagranges equation for r
r
dt
Derivative of V is the force
V ( r )
mr = mr 2 + f (r )
f (r ) =
r
Centrifugal force
Central force
Using the angular momentum l
l2
mr =
+ f (r )
3
mr
l = mr 2
We know how to integrate this.

But we also know what well
get by integrating this
Energy Conservation
m 2 2 2
m 2 1 l2
E = T + V = (r + r ) + V (r ) = r +
+ V (r ) = const
2
2
2
2 mr
r =
2
l2
E V (r )
m
2mr 2
1st order differential

equation of r
One can solve this (in principle) by

t
r
dr
t = dt =
= t (r )
0
r0
l2
2
E V (r )
m
2mr 2
Then invert t(r) r(t)

l

Then calculate (t) by integrating =
mr 2
NB: This never

goes negative
Done! (?)
Degrees of Freedom
A particle has 3 degrees of freedom
Each conservation law reduces one differentiation
By saying time-derivative equals zero
We used L and E 4 conserved quantities
Eqn of motion is 2nd order differential 6 constants
Left with 2 constants of integration = r0 and 0
We dont have to use conservation laws
Its just easier than solving all of Lagranges equations
Qualitative Behavior
Integrating the radial motion

2
l2
r =
E V (r )
isnt always easy

m
2mr 2
More often impossible
You can still tell general behavior by looking at

l2
Quasi potential including
V (r ) V ( r ) +
the centrifugal force
2mr 2
Energy E is conserved, and E V must be positive
mr 2
mr 2
E > V (r )
E=
+ V (r )
= E V (r ) > 0
2
2
Plot V(r) and see how it intersects with E
Inverse-Square Force
Consider an attractive 1/r2 force
k
k
f (r ) = 2
V (r ) =
r
r
Gravity or electrostatic force
k
l2
V (r ) = +
r 2mr 2
1/r2 force dominates at large r
Centrifugal force dominates at
small r
A dip forms in the middle
l2
2mr 2
r
V (r )
k
Unbounded Motion
Take V similar to 1/r2 case
Only general features are relevant
E = E1 r > rmin E1 = V (rmin )
V (r )
E1
1 2
mr
2
Particle can go infinitely far
E2
r
Arrive from r =
E3
Turning point
E =V
r = 0
Go toward r =
A 1/r2 force would

make a hyperbola
Bounded Motion
E = E2 rmin < r < rmax
Particle is confined between two

circles
Goes back and
forth between
two radii
V (r )
E1
E2
1 2
mr
2
r
E3
Orbit may or may not be

closed. (This one isnt)
A 1/r2 force would

make an ellipse
Circular Motion
V (r )
E = E3 r = r0 (fixed)
Only one radius is allowed
E1
Stays on a circle
E = V (r0 )
r = 0
r = const = r0
E2
r
E3
r0
Classification into unbounded, bounded and circular motion

depends on the general shape of V
Not on the details (1/r2 or otherwise)
Another Example
a
V = 3
r
3a
f = 4
r
a
l2
V = 3 +
2mr 2
r
Attractive r-4 force
V has a bump
Particle with energy E may be
either bounded or unbounded,
depending on the initial r
l2
2mr 2
E
r
Stable Circular Orbit
Circular orbit occurs at the bottom of a dip of V

mr 2
= E V = 0
2
dV
mr =
=0
dr
r = const
Top of a bump works in theory,

but it is unstable
Initial condition must be exactly

r = 0 and r = r0
r
stable
unstable
d 2V
Stable circular orbit requires
>0
2
dr
r0
Power Law Force

dV
l2
= f (r0 ) 3 = 0
dr r = r0
mr0
df
dr
<
r = r0
l2
V ( r ) V ( r ) +
2mr 2
d 2V
df
=
2
dr r = r
dr
0
r = r0
3l 2
+ 4 >0
mr0
3 f (r0 )
r0
Suppose the force has a form f = kr n
k > 0 for attractive force

Condition for stable circular orbit is
knr0n 1 < 3kr0n 1
n > 3
Power-law forces with n > 3 can make stable circular orbit
Summary
Started discussing Central Force Problems
Reduced 2-body problem into central force problem
Problem is reduced to one equation
Used angular momentum conservation
l2
+ f (r )
mr =
3
mr
l2
Qualitative behavior depends on V ( r ) V ( r ) +
2mr 2
Unbounded, bounded, and circular orbits
Condition for stable circular orbits
Next step: Can we actually solve for the orbit?
Mechanics
Physics 151
Lecture 6
Kepler Problem
(Chapter 3)
Discussed energy conservation
Started discussing Central Force Problems
Defined energy function h Conserved if L t = 0

Conditions for h = E
Reduced 2-body problem into central force problem
Problem is reduced to one equation mr = l 2 + f (r )

3
Used angular momentum conservation

Energy conservation gives
m 2 1 l2
E = r +
+ V (r ) = const
2
2
2 mr
Now we must solve this
mr
Goals for Today
Analyze qualitative behavior of central-force problem
Solutions: bounded or unbounded

Determined by the shape of the potential
Solve the Kepler problem
Get the shape of the orbit

As if we dont know yet
Derive Keplers 3rd Law
Period of rotation is proportional to the 3/2 power of the
major axis
Integrating the radial motion

2
l2
r =
E V (r )
isnt always easy

m
2mr 2
More often impossible
You can still tell general behavior by looking at

l2
Quasi potential including
V (r ) V ( r ) +
the centrifugal force
2mr 2
Energy E is conserved, and E V must be positive
mr 2
mr 2
E > V (r )
E=
+ V (r )
= E V (r ) > 0
2
2
Plot V(r) and see how it intersects with E
Inverse-Square Force
Consider an attractive 1/r2 force
k
k
f (r ) = 2
V (r ) =
r
r
Gravity or electrostatic force
k
l2
V (r ) = +
r 2mr 2
1/r2 force dominates at large r
Centrifugal force dominates at
small r
A dip forms in the middle
l2
2mr 2
r
V (r )
k
Unbounded Motion
Take V similar to 1/r2 case
Only general features are relevant
E = E1 r > rmin E1 = V (rmin )
V (r )
E1
1 2
mr
2
Particle can go infinitely far
E2
r
Arrive from r =
E3
Turning point
E =V
r = 0
Go toward r =
A 1/r2 force would

make a hyperbola
Bounded Motion
E = E2 rmin < r < rmax
Particle is confined between two

circles
Goes back and
forth between
two radii
V (r )
E1
E2
1 2
mr
2
r
E3
Orbit may or may not be

closed. (This one isnt)
A 1/r2 force would

make an ellipse
Circular Motion
V (r )
E = E3 r = r0 (fixed)
Only one radius is allowed
E1
Stays on a circle
E = V (r0 )
r = 0
r = const = r0
E2
r
E3
r0
Classification into unbounded, bounded and circular motion

depends on the general shape of V
Not on the details (1/r2 or otherwise)
Another Example
a
V = 3
r
3a
f = 4
r
a
l2
V = 3 +
2mr 2
r
Attractive r-4 force
V has a bump
Particle with energy E may be
either bounded or unbounded,
depending on the initial r
l2
2mr 2
E
r
Stable Circular Orbit
Circular orbit occurs at the bottom of a dip of V

mr 2
= E V = 0
2
dV
mr =
=0
dr
r = const
Top of a bump works in theory,

but it is unstable
Initial condition must be exactly

r = 0 and r = r0
r
stable
unstable
d 2V
Stable circular orbit requires
>0
2
dr
r0
Orbit Equation
We have been trying to solve r = r(t) and = (t)

We are now interested in the shape of the orbit r = r ( )
Switch from dt to d
d
l d
2
l = mr
=
dt mr 2 d
const
l d l dr l 2
dV
=0
2
3 +
2
r d mr d mr
dr
Switch from r to u 1 r
l2
dV
mr 3 +
=0
mr
dr
du
d 1
1 dr
=
= 2
d d r
r d
d
2 d
= u
dr
du
Orbit Equation
Switching variables
Solving this equation gives the shape of the orbit
d 2u
m dV ( u1 )
+u+ 2
=0
2
d
l du
Not that its easy (How could it be?)

Will do this for inverse-square force later
One more useful knowledge can be extracted without

solving the equation
Symmetry of Orbit
d 2u
m dV ( u1 )
+u+ 2
=0
2
d
l du
Equation is even, or symmetric, in
Replacing with does not change the equation

Solution u() must be symmetric if the initial condition is
Choosing = 0 at t = 0, makes
du
du
du
OK
if
u (0) u (0)
(0)
(0)
(0) = 0
OK
d
d
d
Orbit is symmetric at angles where du/d = 0
Symmetry of Orbit
Orbit is symmetric about every

turning point = apsidals
Orbit is invariant under reflection
about apsidal vectors
Thats why I didnt care too much about

the sign of r
Solve the orbit between a pair of apsidal
points Entire orbit is known
Now its time to solve the equation
du
=0
d
Solving Orbit Equation
d 2u
m dV ( u1 )
+u+ 2
=0
2
d
l du
Integrating the diff eqn will give energy conservation
One can use energy conservation to save effort
mr 2
l2
E=
+
+ V (r )
2
2
2mr
Switch variables
2
l2
r =
V (r )
E
2
m
2mr
l du
r =
m d
2mV ( u1 )
du
2mE
2
=
u
2
d
l
l2
Integrate this
Inverse Square Force

k
f = 2
r
k
V =
r
du
2mE 2mku
2
u
=
+
d
l2
l2
du
2 mE
l2
2 mku
l2
= d
Look it up in a math text book and find
dx
2
+ 2 x
+ x + x 2 = arccos 2 4
Just substitute , and

Or
Working It Out Yourself

d =
du
2 mE
l2
2
+ 2 mku
u
2
l
=
2 mE
l2
m2 k 2
l4
du
m2 k 2
l4
mk
l2
du
mk
l2
2 mE
l2
u
+
sin
=
d =
sin
cos = cos( ) =
2 mE
l2
m2 k 2
l4
du =
mk
l2
2 mE
l2
Define as cos
2 mE
l2
m2 k 2
l4
sin d
u
+
m2 k 2
l4
Solve this for u = 1/r
Solution
1 mk
2 El 2
u = = 2 1 + 1 +
cos( )
2
r l
mk
This matches the general equation of a conic

1
= C (1 + e cos( ) )
One focus is at the origin
r
e is eccentricity
e>1
E>0
hyperbola
e=1
E=0
parabola
e<1
E<0
ellipse
e=0
mk 2
E= 2
2l
circle
Matches the
qualitative
classification
of the orbits
Energy and Eccentricity
E = 0 separates unbounded and

bounded orbits
l2
2mr 2
Borderline = Parabola
Circular orbit requires

k
l2
V (r0 ) = +
=E
2
r0 2mr0
dV
k
l2
= 2 3 =0
dr r0 r0 mr0
mk 2
E= 2
2l
Hyperbola
Parabola
V (r )
Ellipse
Circle
Unbound Orbits
e > 1 hyperbola
1
= C (1 + e cos( ) )
r
is the turning point (perihelion)

cos( ) > 1/e limits
e = 1 parabola
Bound Orbits
1
= C (1 + e cos( ) )
r
mk
C= 2
Ends of the major axis are 1/ r = C (1 e )
l
Length of the major axis
2 El 2
e = 1+
1 1
1
k
2
+
=
a=
mk
2 C (1 + e) C (1 e)
2E
Major axis is given by the total energy E
2b
2a
Minor axis is
2
l
b = a 1 e2 =
2mE
Rotation Period
k
a=
2E
l2
b=
2mE
Area of the orbit
l 2k 2
A = ab =
8mE 3
We know that the areal velocity is constant
dA 1 2
l
Period of
= r=
rotation
dt 2
2m
Express in terms of a
m 3/ 2
= 2
a
k
mk 2
dA
=A
= 3
dt
2E
Period of rotation is proportional

to the 3/2 power of the major axis
Keplers Third Law of Planetary Motion
Keplers Third Law
Keplers third law is not exact
m 3/ 2
= 2
a
k
The reason: reduced mass

k is given by the gravity
Mm
k
f = G 2 = 2
k = GMm
r
r
Period of rotation becomes
= 2
3/ 2
1 1
+
M m
1
a 3/ 2
= 2
G ( M + m)
Coefficient is same for all planets only if M >> m
planet
Sun
Time Dependence
So far we dealt with the shape of the orbit: r = r()
Why arent we doing it?
We dont have the full solutions r = r(t) and = (t)

Its awfully complicated
Not that bad to get t = t() See Goldstein Section 3.8
Inverting to = (t) impossible
Physicists spent centuries calculating approximate
solutions
Already got physically interesting features of the solution
Leave it to the computers
Summary
Studied qualitative behavior of the orbits
Derived orbit equation from the eqn of radial motion
Bounded or unbounded Shape of
l2
V ( r ) V (r ) +
2mr 2
r (or u = 1/r) as a function of
Analyzed the Kepler Problem
1 mk
2 El 2
cos( )
= 2 1 + 1 +
2
r l
mk
Solved the orbit

Conic depending on E
k
For elliptic orbit, major axis depends only on E a =
2E
Keplers third law of planetary motion
Mechanics
Physics 151
Lecture 7
Scattering Problem
(Chapter 3)
Discussed Central Force Problems
Analyzed qualitative behaviors
Problem is reduced to one equation
l2
mr =
+ f (r )
3
mr
2
l
Unbounded, bounded, and circular orbits
V (r ) = V (r ) +
2mr 2
Condition for stable circular orbits
Defined orbit equation and solved it for the Kepler

problem
1 mk
2 El 2
Conic orbits depending on E
cos( )
= 2 1 + 1 +
2
r l
mk
Goals for Today
Introduce the scattering problem
Define scattering cross section
What happens when a particle collides

How often a particle gets scattered in a given direction
How to calculate it from the potential
Examples
1/r2 force Rutherford scattering

Rainbow scattering
Scattering Problem
Consider unbound motion under central force
Particle comes from infinity goes to infinity
Assume f(r) 0 as r
Orbit approaches straight lines at large r
Straight section A
on
i
t
c
a
r
e
Int
How are sections A and B related?
B
n
o
ti
c
e
ts
h
g
ai
r
t
S
Why Scattering Problem?
Physical observations are scattering phenomena
Experiments on microscopic objects
Photons scattered by an object Seeing

Electrons scattered by an object Electron microscope
Electron-nucleus scattering to probe nuclear structure
Neutrino-electron scattering to measure neutrino energy
Classical description fails with such targets
Still a good approximation in many cases

Classical framework of describing scattering used in QM as
well and its more intuitive to understand
Lousy Shooter Model
Imagine shooting bullets at a small target
Suppose you have very poor aim
Bullets spread uniformly

Number of bullets / area / time = intensity I
Number of hits will be proportional to the target size
Hit frequency
(bullets/sec)
N hits = I
Target cross section (m2)

Intensity (bullets/m2/sec)
Spherical Target
Imagine the target is a solid sphere
We want to know which direction the bullets ricochet

Number of bullets ricocheting into solid
angle d around a direction is
N = I ( ) d
Differential cross
section (m2/str)
Concentrate on the scattering angle
Target is round = has rotational symmetry

()d = () sin d d
Number of bullets between and + d is
2
N = d I () sin d = I ()2 sin d

0
Differential Cross Section
Classical mechanics is deterministic
Scattering angle is determined by the impact parameter s

Probability of scattering between
d
and + d is proportional
to the area of this ring
( s )
2 sds = ()2 sin d
s ds
( ) =
sin d
Absolute value taken because
ds/d might be negative
ds
Total Cross Section

s ds
() =
sin d
Lets check if this matches our idea of the total cross

section
Integrating over the entire solid angle
T = ()d = 2 sin ()d

4
= 2 sds = a 2
0
Total area of
the target
Central Force Scattering
Now consider scattering by general central force
s ds
How does relate to s?
( ) =
sin d
We need to know the shape of the orbit
at large r
d 2u
m dV (1/ u )
+u+ 2
=0
Look at the orbit equation
2
d
l
du
Angular momentum l is related to s by
l = r p 0 = rp0 sin = sp0
If we assume V(r) 0 as r
p02
E =T =
2m
l = sp0 = s 2mE
p0
Central Force Scattering

d 2u
1 dV (1/ u )
+u+ 2
=0
2
d
2s E
du
Orbit equation in terms of

the impact parameter s
and the energy E
One can in principle solve this equation and get

u = u ( , s, E )
r at =
u (, s, E ) = 0
Solve
Then we can calculate

s ds
(, E ) =
sin d
s = s ( , E )
Lets look at the orbit we already

know
Inverse-square force
Inverse Square Force
Consider a repulsive 1/r2 force
k
k
Ex: electrostatic force between
f (r ) = 2 V (r ) =
two like-sign charged particles
r
r
Equation and solution same as Kepler problem
Just flip the sign of k
1
mk
2 El 2
cos( )
= 2 1 + 1 +
2
r
l
mk
Radius > 0
Eccentricity
2 El 2
= 1+
>1
2
mk
Hyperbola
Hyperbolic Orbit
l = s 2mE
1
mk
2 El 2
= 2 1 + 1 +
cos( )
2
r
l
mk
2 El 2
= 1+
>1
2
mk
Solution is a hyperbola
>1E>0
1
1/r > 0 cos( ) <

Scattering angle is
= 2
A bit of
work
cos = 1/
2 Es
2
cot = 1 =
2
k
s=
cot
2E
2
Weve got what we need!
Differential Cross Section
s=
cot
2E
2
Differential cross section is

2
1 k
d
s ds
() =
=
cot
cot
sin d sin 2 E
2 d
2
2
1 k
1
=
4 2 E sin 4
Scattering of particles with charges Ze and Ze k = ZZ e 2

2
1 ZZ e
1
() =
4 2 E sin 4
2
Rutherford scattering:
particle (Z = 2) scattered
by atomic nuclei with Z
Existence of nuclei in atoms
Rutherford Scattering
Before Rutherfords discovery
Electron was known to exist in matter

Positive charge must exist in atoms, but the distribution was
unknown
2 2
1 ZZ e
1
Measurement of () showed () =
4 2 E sin 4 2
Positive charge of +Ze is in one particle
2 2
Z Z e
1
e.g. Z particles of +e each would give
4 2 E sin 4 2
Discovery of atomic nuclei
Total Cross Section
Integrating Rutherford cross section gives

T = ()d =
4
ZZ e 1 d (sin 2 )
= 2
=
0
3
sin 2
2E
Because electrostatic force is long-range
No matter how large is the impact parameter s, the particle
still gets slightly deflected
Reality: electrostatic field is shielded by the electrons around
the nucleus Finite cross section
2
1 ZZ e
1
2 sin
4 2 E sin 4
2
Rainbow Scattering
Equation for () assumes that s() is single-valued
True for Rutherford scattering, but not always
If s() is not monotonous

si dsi
( ) =
i sin d
s ds
( ) =
sin d
Sum up for
possible ss
At maximum = m
d
=0
ds
( ) =
Called rainbow scattering
s2
s1
( )
From Physics 15c
Rainbow
Youve probably heard of how rainbows are made
But the scattering angle depends on

where the light enters the drop
If you add up all possible positions,
rainbow will be washed out
They lied
Real rainbow is made by

1
the light that reflected
s
internally
2
2
2
2
Total deflection is
= 21 4 2 +
Rainbow
s
= 21 4 2 + sin 1 =
sin 1 = n sin 2
R
has a minimum around 137.5
From Physics 15c
n = 1.33
Illuminate a water droplet

with uniform light
min = 2.40
What is the distribution of
light intensity in ?
A bit difficult problem
Covered in Physics 143a and 151
s ds
This goes to infinity
The answer: I ()
at the turning point there
sin d
s R
From Physics 15c
Rainbow
Minimum of Sharp peak of intensity I()
I ()
min
r < rmin
smin
r > rmin
s R
Reflection observed only at

min This depends on n,
which depends slightly on
This is really how rainbow
is created
min
min
Attractive Force
Repulsive force can only scatter by 0 < <

Attractive force can do more
If the potential and the energy are just right,

particle can make multiple turns before emerging
Called spiraling or orbiting
V (r )
Orbiting region:
E V is small
r varies slowly
E
r
Summary
Discussed scattering problem
Defined and calculated cross sections
Foundation for all experimental particle physics

Differential cross section
Rutherford scattering
Done with central force problems

Next: Rigid Bodies
N hits = I
s ds
( ) =
sin d
Mechanics
Physics 151
Lecture 8
Rigid Body Motion
(Chapter 4)

!
Discussed scattering problem

!
Defined and calculated cross sections

!
!
Foundation for all experimental physics

Differential cross section and impact parameter
Rutherford scattering
( ) =
Translated into laboratory system

!
!
N hits = I
Angular translation + Jacobian

Shape of () changes
s ds
sin d
Goals For Today

!
Start discussing rigid-body motion

!
Concentrate on representing the rotation

!
!
!
Multi-particle system with fixed shape

Which generalized coordinates should we use?
Define Euler angles
Define infinitesimal rotation
! Will use this for angular velocities, etc
Todays lecture is largely mathematical

!
Assume knowledge of linear algebra
Rigid Body
!
Multi-particle system with fixed distances

!
Constraints: rij = ri r j = const for all i, j
How should we define generalized coordinates?

!
!
How many independent coordinates are there?

If you start from 3N and subtract the number of constraints
N ( N 1) 7 N N 2
3N
=
2
2
0 for N 7
Not all the constraints are independent

!
Right answer: 3 translation and 3 rotation = 6

Todays theme
2-D Rotation
!
2-dimensional rotation is specified by a 22 matrix

x cos sin x
=

y
sin
cos

y
i i i j x
=

j
i
j
j
y
!
Try the same thing with 3-d rotation

z
y
z
y
x
y
j
i
x
x
i
3D Rotation
!
Vector r is represented in x-y-z and x-y-z as

r = xi + yj + zk = xi + yj + z k
!
Using angles ij between two axes

x = r i = xi i + yj i + zk i
= cos 11 x + cos 12 y + cos 13 z
z
z 13
y = cos 21 x + cos 22 y + cos 23 z

z = cos 31 x + cos 32 y + cos 33 z
x cos 11 cos 12
or y = cos 21 cos 22
z cos 31 cos 32
cos 13 x
cos 23 y
cos 33 z
11
12
x
3D Rotation
!
Simplify formulae by renaming

( x, y, z ) ( x1 , x2 , x3 ) ( x, y, z ) ( x1, x2 , x3 )
!
Rotation is now expressed by

xi = cos ij x j = aij x j = aij x j
j
Einstein convention:
Implicit summation over repeated index
!
We got 9 parameters aij to describe a 3-d rotation

!
Only 3 are independent
Constraints of Rotation
!
Rotation cannot change the length of any vector

!
Exactly the constraints we need for rigid body motion
r 2 = xi xi = xixi
!
Using the transformation matrix

xi = aij x j
xixi = aij x j aik xk
therefore
1 ( j = k )
aij aik = jk
0 ( j k )
Matrix A = [aij] is orthogonal

! =1
AA
Transpose of A
6 conditions
reduces free
parameters
from 9 to 3
Orthogonal Matrix
!
a11
A = a21
a31
a12
a22
a32
a13
a23
a33
Goldstein Section 4.3 covers algebra of matrices
You must have learned this already

aij aik = jk
!
! Orthogonal matrix A satisfies AA = 1
! Consider the determinants
Transposed matrix
2
!
!
AA = A A = A = 1
A = 1
!
!
!
|A| = +1 " proper matrix

|A| = 1 " improper matrix
Space Inversion
!
Space inversion is represented by

1 0 0
r = r = Sr 0 1 0 r
0 0 1
!
!
S = 1
S is orthogonal Doesnt change distances

But it cannot be a rotation
! Coordinate axes invert to become left-handed
! Orthogonal matrices with |A| = 1 does this
Rigid body rotation is represented by proper

orthogonal matrices
Rotation Matrix
!
A operating on r can be interpreted as

!
Rotating r around an axis by an angle

! Positive angle = clockwise rotation
Rotating the coordinate axes around the same axis by the
same angle in the opposite direction
! Positive angle = counter clockwise rotation
Both interpretations are useful

!
We are more interested in the latter for now
How do we write A with 3 parameters?

!
r = Ar
There are many ways
Euler Angles
!
Transform x-y-z to x-y-z in 3 steps

( x, y , z )
Rotate CCW by around z axis
( , , )
( , , )
Rotate CCW by around axis
( x, y, z)
Rotate CCW by around axis
Dx
CDx
Ax = BCDx
z y
Euler Angles
cos
D = sin
0
sin
cos
0
0
0
1
0 C = 0 cos
0 sin
1
cos cos cos sin sin

A = sin cos cos sin cos
sin sin
!
0
cos
sin B = sin
0
cos
sin
cos
cos sin + cos cos sin

sin sin + cos cos cos
sin cos
Definition of Euler angles is somewhat arbitrary

!
!
May rotate around different axes in different order

Many conventions exist Watch out!
0
0
1
sin sin
cos sin
cos
Rigid Body Motion

!
Motion of a rigid body can be described by:

!
!
!
Define x-y-z axes (body axes) attached to the rigid body

! Same direction as x-y-z (space axes) at t = 0
! Origin fixed at one point of the rigid body (e.g. CoM)
Use R(t) to describe the motion of the origin
Use A(t) to describe the rotation of the x-y-z axes
! Use Euler angles (t), (t), (t)
! A(0) = 1 " (0) = (0) = (0) = 0
6 independent coordinates (x, y, z, , , )
Eulers Theorem
The general displacement of a rigid body with
one point fixed is a rotation about some axis
!
In other words
!
!
Arbitrary 3-d rotation equals to one rotation around an axis

Any 3-d rotation leaves one vector unchanged
For any rotation matrix A

!
!
There exists a vector r that satisfies Ar = r

A has an eigenvalue of 1
Eigenvector with
eigenvalue 1
Eulers Theorem
!
If a matrix A satisfies Ar = r
( A 1)r = 0
A 1 = 0 or r = 0 or A - 1 = 0
!
Since A 1 = A
! = 1 A
!
( A 1) A
! = 1 A
!
A 1 A
A 1 = 1 A
!
For odd-dimensioned matrices M = M
A 1 = A 1 = 0
Q.E.D.
Rotation Vector?
!
Eulers theorem provides another way of describing

3-d rotation
!
!
Direction of axis (2 parameters) and angle of rotation (1)

It sounds a bit like angular momentum
Critical difference: commutativity

!
Angular momentum is a vector

! Two angular momenta can be added in any order
Rotation is not a vector
! Two rotations add up differently depending on which
rotation is made first
Infinitesimal Rotation
!
Small (infinitesimal) rotations are commutative

!
!
They can be represented by vectors

We also need them to describe how a rigid body changes
orientation with time
Infinitesimal rotation must be close to non-rotation

xi = xi + ij x j or x = (1 + )x
!
ij " 1
Two successive infinitesimal rotations make

(1 + 1 )(1 + 2 ) = 1 + 1 + 2 + 1 2
2nd order of vanishes
= 1 + 1 + 2
! Obviously commutative
!
Inverse of an infinitesimal rotation is

(1 + ) 1 = 1
!
!
Using A 1 = A
! =
!
(1 + )(1 ) = 1 + = 1
1 + ! = 1
is antisymmetric
We can write as
0
= d 3
d 2
d 3
0
d 1
d 2
d 1
0
d = (d 1 , d 2 , d 3 )
behaves almost like a
vector
Well see
!
A vector r is rotated by (1 + ) as r = (1 + )r
d 3 d 2 x1
0
dr r r = r = d 3
d 1 x2 = r d
0
d 2 d 1
0 x3
! Eulers theorem says this equals to a
n
d
rotation by an infinitesimal angle d
around an axis n
dr = r nd
d = nd
dr
Axial Vector
!
d behaves pretty much like a vector

!
d rotates the same way as r with coordinate rotations
Space inversion S reveals difference

!
!
Ordinary vector flips r = Sr = r

d doesnt
(dr ) = r d
d = d
= dr = r d = r d
!
Such a vector is called an axial vector

!
Examples: angular momentum, magnetic field
Parity
!
Parity operator P represents space inversion

P
( x, y, z ) ( x, y, z )
Quantity
Parity
Eigenvalue
Scalar
PS = S
+1
Pseudoscalar PS* = S*
Vector
PV = V
Axial vector
PV* = V*
+1
V V = V*
V V* = V
S *V = V*
V V* = S *
V* V* = S
S *V* = V
etc.
Summary
!
Discussed 3-dimensional rotation

!
Looked for ways to describe 3-d rotation

!
!
Euler angles one of the many possibilities

Eulers theorem
Defined infinitesimal rotation d

!
!
Preparation for rigid body motion

! Movement in 3-d + Rotation in 3-d = 6 coordinates
Commutative (unlike finite rotation)

Behaves as an axial vector (like angular momentum)
Ready to go back to physics
Mechanics
Physics 151
Lecture 9
Rigid Body Motion
(Chapter 4, 5)
Discussed 3-dimensional rotation
Looked for ways to describe 3-d rotation
Preparation for rigid body motion

Movement in 3-d + Rotation in 3-d = 6 coordinates
Euler angles one of the many possibilities
Eulers theorem
Defined infinitesimal rotation d
d = nd
dr = r d
Commutative (unlike finite rotation)

Behaves as an axial vector (like angular momentum)
Goals For Today
Time derivatives in rotating coordinate system
Try to write down Lagrangian for rigid body
To calculate velocities and accelerations

Coriolis effect
Express angular velocity using Euler angles
Separate rotation from movement of CoM
Define inertia tensor
Will almost get to the equation of motion
Body Coordinates
Consider a rotating rigid body
Define body coordinates (x, y, z)
Between t and t + dt, the body

coordinates rotate by d = nd
The direction is CCW because we

are talking about the coordinate axes
Observed in the space coordinates,
any point r on the body moves by
dr = d n r = d r
Sign opposite from last lecture
because the rotation is CCW
y
x
d
dr
General Vectors
Now consider a general vector G
How does it move in space/body coordinates?

i.e. whats the time derivative dG/dt ?
Movement dG differs in space and body coordinates

because of the rotation of the latter
( dG )space = ( dG )body + ( dG )rot
Difference is due to rotation
If G is fixed to the body
( dG )body = 0
and
( dG )space = d G
( dG )rot = d G
Generally true
Angular Velocity
For any vector G ( dG )space = ( dG )body + d G

dG
dG
=
+ G
dt space dt body
dt = d = d n
= instantaneous angular velocity

Direction = n = instantaneous axis of rotation
Magnitude = d/dt = instantaneous rate of rotation
Since this works for any vector, we can say

d d
= +
dt s dt r
space coordinates
rotating coordinates
Coriolis Effect
d d
= +
dt s dt r
Imagine a particle observed in a rotating system
e.g. watching an objects motion on Earth

Velocity: v s = v r + r
dv s dv s
Acceleration: a s =
=
+ vs
dt s dt r
= a r + 2 v r + ( r )
Newtons equation works in the space (inertial) system,
i.e. F = ma s
ma r = Feff = F 2m( v r ) m ( r )
Object appears to move according to this force
Coriolis Effect
ma r = Feff = F 2m( v r ) m ( r )
Last term is centrifugal force

vr
Middle term is Coriolis effect
Occurs when object is moving in the
rotating frame
v r
Moving objects on Earth appears to deflect toward right in
the northern hemisphere
Hurricane wind pattern
L
Foucault pendulum
Coriolis Effect
Free-falling particle
Drop a ball and ignore air resistance

Coriolis effect is 2m( v )
Pointing east (out of screen) x axis
Ignoring the centrifugal force
mx = 2m vz sin
mz = mg
x=
x=
gt 3 sin
3
sin
3
gt 2
,z=
2
(2 z )3
g
For sin = 1 and z = 100 m x = 2.2 cm
Euler Angles
Use angular velocity to calculate particles velocities

Use Euler angles to describe the rotation of rigid bodies
How are they connected?
Infinitesimal rotations can be added like vectors

= n + n + n
z
nz
y
x
z y
z
y
n z
Euler Angles
Lets express in x-y-z
Doing this in x-y-z equally easy (or difficult)

Must express nz, n, nz in x-y-z
nz Anz, n Bn (nz is OK)
z
nz
y
x
z y
z
y
n z
Euler Angles
From the last lecture
cos
D = sin
0
sin
cos
0
0
0
1
0 C = 0 cos
0 sin
1
cos cos cos sin sin

A = sin cos cos sin cos
sin sin
0
cos
sin B = sin
0
cos
cos sin + cos cos sin

sin sin + cos cos cos
sin cos
sin
cos
0
0
0
1
sin sin
cos sin
cos
0 sin sin
1 cos
0
n z = A 0 = cos sin n = B 0 = sin n z = 0
1 cos
0 0
1
Euler Angles
sin sin + cos

= n z + n + n z = cos sin sin
cos +
Remember: this is in the x-y-z system
Now we know how to express velocities in terms of

time-derivatives of Euler angles
We can write down the Lagrangian
Kinetic Energy
Kinetic energy of multi-particle system is

1
1
2
T = Mv + mi vi2
2
2
Motion of CoM
Remember Einstein convention

Motion around CoM
If we define the body axis from the center of mass

1
T = M ( x 2 + y 2 + z 2 ) + T (,, )
2
T depends only on the angular velocity

Must be a 2nd order homogeneous function
Potential Energy
Potential energy can often be separated as well

V = V1 ( x, y, z ) + V2 ( , , )
Uniform gravity g
V1 = g r
Uniform magnetic field B and magnetic dipole moment M
V2 = M B
Lagrangian can then be written as

L = Lt ( x, y, z , x , y , z ) + Lr ( , , , ,, )
Translational
Rotational
It is often possible to separate the translational and rotational motions by

taking the center of mass as the origin of the body coordinate axes
Rotational Motion
We concentrate on the rotational part
Translational part same as a single particle Easy
Consider total angular momentum L = mi ri vi
vi is given by the rotation as v i = ri
L = mi ri ( ri )
mi (ri 2 xi2 )
mi xi yi
mi xi zi
2
2
2
mi (ri yi )
= mi ri ri (ri ) = mi yi xi
mi yi zi
2
2
mi zi xi
m
z
y
m
(
r
z
i i i
i i
i )
Inertia tensor I
Inertia Tensor
Diagonal components are familiar

moment of inertia
I xx = mi (ri x ) = mi ri sin
2
2
i
What are the off-diagonal components?
Iyx produces Ly when the object is turned around x axis

Imagine turning something like:
Unbalanced one has non-zero
off-diagonal components,
which represents wobbliness
of rotation
Balanced
Unbalanced
Inertia Tensor
Using ( xi , yi , zi ) ( xi1 , xi 2 , xi 3 )
mi (ri 2 xi2 )
mi xi yi
mi xi zi
2
2
mi (ri yi )
mi yi zi
I = mi yi xi
2
2
mi zi xi
m
z
y
m
(
r
z
i i i
i i
i )
I jk = mi ( ri 2 jk xij xik )
We can also deal with continuous mass distribution (r)

I jk = (r ) ( r 2 jk x j xk )dr
Kinetic Energy
1
Kinetic energy due to rotation is T = mi v i v i
2
Using v i = ri
1
mi v i ( ri ) = mi (ri v i ) = L
2
2
2
I
Using L = I
T=
2
Defining n as the unit vector in the direction of = n
T=
n In 2 1 2
= I where I = n In = mi ri 2 (ri n) 2
T=
2
2
Moment of inertia about axis n
NB: n moves with time

I = I(t) must change accordingly with time
Shifting Origin
Origin of body axes does not have to be at the CoM
Its convenient Separates translational/rotational motion
If it isnt, I can be easily translated

from origin
ri = R + ri
from CoM
I = mi (ri n) 2 = mi [(R + ri) n]2

= M (R n) 2 + mi (ri n) 2 + 2mi (R n) (ri n)
I of CoM
I from CoM
Summary
Found the velocity due to rotation
Connected with the Euler angles

Lagrangian translational and rotational parts
Often possible if body axes are defined from the CoM
I = mi ( ri 2 jk xij xik )
Defined the inertia tensor
Used it to find Coriolis effect
d d
= +
dt s dt r
Calculated angular momentum

and kinetic energy
L = I T =
Next: Equation of motion (finally!)
nIn 2 1 2
= I
2
2
Mechanics
Physics 151
Lecture 10
Rigid Body Motion
(Chapter 5)

!
Found the velocity due to rotation

!
!
!
Connected with the Euler angles

Lagrangian " translational and rotational parts
!
Used it to find Coriolis effect
d d
= +
dt s dt r
Often possible if body axes are defined from the CoM
Defined the inertia tensor

!
Calculated angular momentum

and kinetic energy
I 1 2
L = I T =
= I
2
2
Diagonalizing Inertia Tensor

mi (ri 2 xi2 )
mi xi yi
mi xi zi
2
2
mi yi zi = mi ( ri 2 jk xij xik )
mi (ri yi )
I = mi yi xi
2
2
mi zi xi
m
z
y
m
(
r
z
i i i
i i
i )
Inertia tensor I can be made diagonal

I1
! = 0
I D = RIR
0
I2
0
0
0
I 3
R is a rotation matrix
Ii > 0
Kinematical properties of a rigid body are fully described by its

mass, principal axes, and moments of inertia
Principal Axes
!
Consider a rigid body with body axes x-y-z

!
!
Inertia tensor I is (in general) not diagonal

!
But it can be made diagonal by I D = RIR
Rotate x-y-z by R " New body axes x-y-z

!
In x-y-z coordinates
= R L = RL
= RI
!
= RIRR
! =1
RR
= I D
How do I
find them?
One can choose a set of

body axes that make the
inertia tensor diagonal
" Principal Axes
Finding Principal Axes

!
0
0
I 3
ni is an eigenvector of I with eigenvalue Ii
To find the principal axes and principal moments:

!
!
!
!
0
I2
0
Consider unit vectors n1, n2, n3 along principal axes

Ini = I i ni
I1
! = 0
I D = RIR
Express I in in any body coordinates

Solve eigenvalue equation
= I1 , I 2 , I 3
I = 0
(I )r = 0
Eigenvectors point the principal axes
Use them to re-define the body coordinates to simplify I
You can often find the principal axes by just looking at

the object
Rotational Equation of Motion

!
Concentrate on the rotational motion

!
Newtonian equation of motion gives

dL
dL
N
=
+ L = N
dt s
dt b
space axes
d d
= +
dt s dt b
body axes
Take the principal axes as the body axes

I1
L = I = 0
0
0
I2
0
0 1 I11
0 2 = I 2 2
I 3 3 I 33
Eulers Equation of Motion

I11 1 I11 N1
d
+ I = N
I
2 2
2 2 2 2
dt
I 33 3 I 33 N 3
I1"1 23 ( I 2 I 3 ) = N1
I 2" 2 31 ( I 3 I1 ) = N 2
I 3" 3 1 2 ( I1 I 2 ) = N 3
Eulers equation of motion for rigid body

with one point fixed
!
Special cases:
! 2 = 3 = 0
I1"1 = N1
I 2 = I3
I1"1 = N1
Torque-Free Motion
!
!
No linear force " Conservation of linear momentum

No torque " Conservation of angular momentum
Try N = 0 in Eulers equation of motion

I1"1 23 ( I 2 I 3 ) = 0
I 2" 2 31 ( I 3 I1 ) = 0
I 3" 3 1 2 ( I1 I 2 ) = 0
Integrating these equation will give us

energy and angular momentum
conservation
We will do something more intuitive (hopefully)

!
Geometrical trick by L. Poinsot
Inertia Ellipsoid
!
For any direction n, I = n In = ni I ij n j

!
Moment of inertia
about axis n
If we express n using principal axes x-y-z

I = I i ni2 = I1n12 + I 2 n22 + I 3 n32
! Consider a vector =
n
I
1 = I i i2 = I1 12 + I 2 22 + I 3 32
Inertia ellipsoid
Inertia Ellipsoid
n
=
I
I3
1 = I1 12 + I 2 22 + I 3 32
Inertia ellipsoid represents

the moment of inertia of a
rigid body in all directions
1
1
I1
I2
Usefulness of this
definition will become
apparent soon
n
=
I
Inertia Ellipsoid
!
Inertia along axis n is I = n In

F () I = I1 12 + I 2 22 + I 3 32 = 1
F is a function (like potential) defined in the space

!
!
F() = 1 " Inertia ellipsoid

Normal of the ellipsoid
given by the gradient
F = 2I
Using = n
2
F =
L
T
I =
2T
Surface of the
inertia ellipsoid is
perpendicular to L
F=1
Invariable Plane
2
F =
L
T
!
L is conserved!
Surface of inertia ellipsoid at is

perpendicular to the angular momentum L
As moves, inertia ellipsoid must rotate to satisfy this

condition continuously
Consider the projection of on L

L L
2T
=
=
L
L
L 2T
!
constant
The ellipsoid is touching

a fixed plane perpendicular to L
Invariable plane
2T
L
Invariable Plane
!
Inertia ellipsoid touches the invariable plane

!
Distance between the center and the plane is
2T L
Determined by the initial conditions

!
Touching point = instantaneous axis of rotation

!
Tip of the vector is momentarily at rest in space

d

= =
=0
dt
2T
i.e. its not sliding, but
rolling without slipping

on the invariable plane
Invariable Plane
!
Inertia ellipsoid rolls on the invariable plane

!
!
Touching point draws curves

!
!
Rotation of ellipsoid gives the rotation of the body

Direction of gives the direction of in space
Curve drawn on the ellipsoid = polhode
Curve drawn on the invariable plane
= herpolhode
Lets examine a few

simple cases
Simple Cases
!
Inertia ellipsoid is a sphere (I1 = I2 = I3)

!
!
is constant and parallel to L

Stable rotation
Simple Cases
!
Initial axis is close to one of the principal axes

!
!
!
Assume I1 > I2 > I3

Stable rotation around I1 and I3
Not so obvious around I2
! If 1 = 3 = 0, 2 is constant
I1"1 23 ( I 2 I 3 ) = 0
I 2" 2 31 ( I 3 I1 ) = 0
I 3" 3 1 2 ( I1 I 2 ) = 0
!
Small deviation leads to instability
I3
I1
Simple Cases
!
Since I1 > I2 > I3, distance 1 I 2 allows a polhode that

wraps around the inertia ellipsoid
Rotation around a principal axis is stable except for the

one with the intermediate moment of inertia
Simple Cases
!
Inertia ellipsoid is symmetric around one axis

!
!
!
!
I1 = I2 I3
draws a cone (space cone) on the invariable plane
draws a cone (body cone) in the inertia ellipsoid
Body cone rolls on the space cone
Precession
!
I1 = I2 turns the Eulers equation of motion to

I1"1 = 23 ( I1 I 3 )
I1" 2 = 31 ( I 3 I1 )
I 3" 3 = 0
3 is constant
Consider it as a given initial condition
"1 = 2 , " 2 = 1
I 3 I1
3
I1
1 = A cos t , 2 = A sin t
!
precesses around the I3 axis

!
Draws the body cone
Rotation Under Torque

!
I1"1 23 ( I 2 I 3 ) = N1
I 2" 2 31 ( I 3 I1 ) = N 2
We introduce torque
Things get messy
I 3" 3 1 2 ( I1 I 2 ) = N 3
Consider a spinning top

!
Define Euler angles
Lagrangian
!
Assume I1 = I2 I3
!
!
1
1
2
2
Kinetic energy given by T = I1 (1 + 2 ) + I 332
2
2
Use Euler angles
" sin sin + " cos
= " cos sin " sin

" cos + "
I3 "
I1 " 2 "2 2
T = ( + sin ) + ( cos + " ) 2
2
2
Lagrangian
!
Potential energy is given by the height of the CoM

V = Mgl cos
Lagrangian is
I3 "
I1 " 2 "2 2
L = ( + sin ) + ( cos + " ) 2 Mgl cos
2
2
!
!
Finally we are in real business!

How we solve this?
! Note and are cyclic
!
Can define conjugate momenta that conserve
To be continued
Summary
!
Discussed rotational motion of rigid bodies

!
Analyzed torque-free rotation

!
!
!
Eulers equation of motion

Introduced the inertia ellipsoid
It rolls on the invariant plane
Dealt with simple cases
I3
y
x
1
1
Started discussing heavy top

!
Found Lagrangian " Will solve this next time

I3 "
I1 " 2 "2 2
L = ( + sin ) + ( cos + " ) 2 Mgl cos
2
2
I1
I2
Mechanics
Physics 151
Lecture 11
Rigid Body Motion
(Chapter 5)
Administravia
Please fill out the midterm evaluation form
Critical feedback for me

to evaluate how well (or badly) Im teaching
to adjust the level of the course according to your needs
to receive your suggestions for improvements
Be critical, and be specific
I cant fix them if you dont tell me whats wrong
Its anonymous and confidential
Thank you!
Discussed rotational motion of rigid bodies
Analyzed torque-free rotation
Eulers equation of motion

Introduced the inertia ellipsoid
It rolls on the invariant plane
Dealt with simple cases
I3
y
x
1
1
I1
Started discussing heavy top
Found Lagrangian Analyze it today

L=
I
I1 2 2 2
( + sin ) + 3 ( cos + ) 2 Mgl cos
2
2
I2
Heavy Top
Top is spinning on a fixed point
Lagrangian is
I3
I1 2 2 2
L = ( + sin ) + ( cos + ) 2 Mgl cos
2
2
and are cyclic
Symmetry
p and p are conserved
Conserved Momenta
I3
I1 2 2 2
L = ( + sin ) + ( cos + ) 2 Mgl cos
2
2
L
p =
= I 3 ( cos + ) = I 33 = const. I1a

L
p = = I1 sin 2 + I 3 cos ( cos + ) = const. I1b
Solve them for and

b a cos

=
sin 2
=
I1a
b a cos
cos
I3
sin 2
We need (t) to get (t) and (t)
Got rid of 2 degrees of freedom
Energy Conservation
I3
I1 2 2 2
E = ( + sin ) + ( cos + ) 2 + Mgl cos
2
2
2
Middle term is 12 I 3 3
conserved
I 332 I1 2 I1 (b a cos ) 2
E = E
=
+
+ Mgl cos
2
2
2
2
sin
Weve got a 1-dim equation of motion of
It looks like a particle of mass I1
under a potential
I1 b a cos
V ( ) =
+ Mgl cos
2 sin
2
b a cos

=
sin 2
conserved
conserved
1-D Equation of Motion
Simplify the equation of motion by defining

2 E I 332
I1
2Mgl
and
I1
b a cos
2

= +
+ cos
sin
2
EoM becomes
Switch variable from to u = cos
u 2 = (1 u 2 )( u ) (b au ) 2
Integrate
u (t )
du
t=
u (0)
(1 u 2 )( u ) (b au ) 2
EoM
Elliptic
integral
Try to extract qualitative behavior
Same way as we did with central force problem
Consider the RHS of the last equation

u 2 = f (u ) (1 u 2 )( u ) (b au ) 2
= u 3 ( + a 2 )u 2 + (2ab )u + ( b 2 )
Physical range is f (u ) = u 2 0 and 1 u 1
f(u) is a cubic function of u with
f (1) = (b au ) 2 0
2 Mgl
I1
>0
These conditions constrain

the shape of f(u)
u = cos
Shape of f(u)
f(u) = 0 has 3 roots 1 u1 u2 1 u3
Solution for u 2 = f (u ) is
bounded inside u1 u u2
f (u )
oscillates between
arccos(u1) and arccos(u2)
and determined by
b a cos

=
sin 2
I1a
b a cos

=
cos
sin 2
I3
u1
1
u2
1 u
3
Nutation
b a cos b au

=
Consider the sign of =
2
sin
1 u2
changes sign at u = u = b / a
u < u1 or u > u2
is monotonous
u1 < u < u2
switches direction
locus
Initial Condition
Suppose the figure axis is initially at rest
Spin the top, then release it quietly

f (ut =0 ) = 0
ut =0 = u1 or u2
t =0 = 0
b aut =0 = 0
ut =0 = u
= 0
t =0
t=0
Initially, the figure axis falls

It then picks up precession in
How does it know which way to go?
Origin of Precession
Angular momentum conservation

L
= I 33
p =

L
p = = I1 sin 2 + I 33 cos
3 is constant
As the figure axis falls, 3s contribution to p decreases

must start precessing to make up for it
Direction of precession is same as that of spin
Uniform Precession
Can we make a top precess without bobbing?
i.e. = 0, = const
We need to have a double root for f(u) = 0
b a cos

=
sin 2
f (u0 ) = (1 u02 )( u0 ) (b au0 ) 2 = 0
f (u0 ) = 2u0 ( u0 ) (1 u02 ) + 2a (b au0 ) = 0

Combine
I1a I 33
2Mgl
I1
= a 2u0
Mgl = ( I 33 I1 cos 0 )
f (u )
u0
1
1 u
3
Uniform Precession
Mgl = ( I 33 I1 cos 0 )

For any given value of 3 and cos0, you must give exactly
the right push in to achieve uniform precession
Quadratic equation 2 solutions
Same top can do fast or slow precession
For the solutions to exist I 3232 > 4 MglI1 cos 0
2
MglI1 cos 0
3 >
I3
Uniform precession is achieved only by a fast top
Magnetic Dipole Moment
Consider a rigid body made of charged particles
Mass mi, charge qi, position ri, velocity vi
If there is uniform magnetic field B
Each particle feels force Fi = qi v i B

No sum over i!
If CoM is at rest and qi/mi = const
q
No net force
F = qi v i B = mi v i B = 0
m
How about the torque?
q
= ri Fi = qi ri ( v i B) = mi ri ( v i B)
m
Using v i = ri
q
q
N = mi ri ( v i B) = mi ( ri )(ri B) B
m
m
Explicit calculation using polar coordinates
ri
sin
( ri )(ri B) = ri 2 B sin cos (sin cos sin + cos cos )
0
Take time average Assume rotation is fast

q
q
2
N=
LB
mi ( ri sin ) B =
2m
2m
N=
q
LB
2m
Magnetic dipole M in B feels the torque N = M B
Fast spinning charged rigid body has a magnetic

moment M = L
q
=
2m
gyromagnetic ratio
dL
= LB
Equation of motion
dt
This makes L to precess around B
Angular velocity of precession is

q
precess = B =
B
2m
Larmor frequency
Elementary Particles
Particles such as electrons or protons have
Spin, or intrinsic angular momentum, s

Magnetic moment
q
Dirac equation for a spin-1/2 particle predict = s
m
Differs from classical charged object by factor 2
classical object
1
gq
Particle physicists say =
s g=
Dirac particle
2m
2
g = 2 for electron and muon Dirac particles
g = 2.8 for proton, 1.9 for neutron composite particles
Anomalous Magnetic Moment
of electron and muon known very accurately

g electron = 2.002319304374 0.000000000008
g muon = 2.002331832 0.0000000012
Not pure Dirac particles, but surrounded by thin cloud of

virtual particles due to quantum fluctuation
Measurement uses spin precession
Store particles with known spin orientation in B field

Measure spin direction after time t
precess
gq
=
B
2m
Need to know B very accurately
Muon g2 Experiment
BNL E-821 muon storage ring
g muon = 2.0023318404
0.0000000030
Summary
Analyzed the motion of a heavy top
Magnetic dipole moment of spinning charged object
Reduced into 1-dimensional problem of

Qualitative behavior Precession + nutation
Initial condition vs. behavior
M = L, where = q/2m is the gyromagnetic ratio
L precesses in magnetic field by = B
of elementary particles contains interesting physics
Done with rigid bodies
Next: Oscillation
Mechanics
Physics 151
Lecture 12
Oscillations
(Chapter 6)
Analyzed the motion of a heavy top
Reduced into 1-dimensional problem of

Qualitative behavior Precession + nutation
Initial condition vs. behavior
Magnetic dipole moment of spinning charged object
M = L, where = q/2m is the gyromagnetic ratio

L precesses in magnetic field by = B
of elementary particles contains interesting physics
Oscillations
Oscillations are everywhere
Youve learned it in 15c or 16
Small motion near equilibrium (perturbation)

Appears in many QM problems
Rather easy
x = 2 x
x = A exp(i t )
Well do more formal treatment of multi-dimensional

oscillation
Did this (less formally) in 15c for coupled oscillators

Georgis book does this You may already know
Problem Definition
Consider a system with n degrees of freedom
Generalized coordinates {q1, , qn}
Generalized force at the equilibrium

V
Qi =
=0
qi 0
Potential V is at
an extremum
V must be minimum at a stable equilibrium
Taylor expansion of V using qi = q0i + i
constant
V
1 2V
V = V0 +
i +
2 qi q j
qi 0
zero
i j +
0
1
Viji j
2
Constant
symmetric
matrix
Problem Definition
Kinetic energy is a 2nd-order homogeneous function of

velocities
1
1
T=
mij (q1 , , qn )qi q j =
miji j
2
2
This requires that the transformation functions do not
explicitly depend on time, i.e.
qi = qi ( x1 , , xN , t )
mij generally depend on {qi} Taylor expansion

mij
mij = ( mij ) +
k +
0
qk 0
Tij
1
T Tiji j
2
Constant
symmetric
matrix
Lagrangian
For small deviation {i} from equilibrium
1
1
1
1
L = T V = Tiji j Viji j = T V
2
2
2
2
The equations of motion are
Tij j + Vij j = 0
(Textbook Eq 6.8 wrong)
This looks similar to mx + kx = 0
Difficulty: Tij and Vij have off-diagonal components

If T and V were diagonal
Tij j + Vij j Tiii + Viii = 0
no sum over i
Can we find a new set of coordinates that diagonalizes T and V?
Solving Lagranges Equations
Assume that solution will be i = Cai eit

Tij j + Vij j = 0
2Tij a j + Vij a j = 0 or Va Ta = 0
Its a slightly odd eigenvalue equation
=2
Solution can be found by V T = 0

n-th order polynomial of Expect n solutions for
must be real and = 2 > 0
Can be proven by
a bit of work
Reality of Eigenvalues
Start from Va = Ta
Take adjoint (complex conjugate + transpose)

a V = *aT
Multiplying by a or a gives a Va = aTa = *aTa
( * )aTa = 0
Writing a as + i ( and are real)
aTa = ( i)T( + i) = T + T + i (T T )
Since T = 12 T is positive for any real

aTa = T + T > 0
zero
* = 0, i.e. is real
Reality of Eigenvectors
Now that we know is real

Va = Ta
V + iV = T + i T
and both satisfy the same eigenvalue equation
Suppose that the eigenvalues are not degenerate

V T = 0
= 1 , 2 , , n
Will worry about the degenerate case later

Each eigenvalue corresponds to 1 eigenvector
and are proportional to each other
a = + i =
All different
a complex number
a can be made real by absorbing in C of i = Cai eit
Positive Definiteness
We now have an all-real equation aVa = aTa

aVa
=
aTa
Positive because V =
V 0 for any real

if V is minimum at the equilibrium
1
2
Already shown to be positive
= 2 is positive definite
We now have a guarantee that each solution of the

eigenvalue equation gives an oscillating solution
= Cae i t with a definite frequency = 2
Normalization
Eigenvector satisfying Va = Ta has arbitrary scale
a Ca can absorb such scale as well as imaginary phase

We fix the normalization by declaring
aTa = 1
Just the sign () remains ambiguous
aVa
This turns =
aTa
= aVa
Principal Axis Transformation
There are n eigenvectors Call them aj

Va j = j Ta j
j = 1, 2,..., n
Take transpose
no sum over j or k
a k V = k a k T
a k Ta j = jk
( j k )a k Ta j = 0
Assuming j k for j k
a k Va j = j jk
If we stack aj to make A = [a1 a 2
an ]
ATA = 1
1
AVA = =
0
T and V are diagonalized

by the principal axis
transformation
Normal Coordinates
1
1
1
1
L
=
T
=
T
V
Lagrangian was
ij i j
ij i j
2
2
2
2
Once we have A, we can switch to new coordinates
A 1
Normal coordinates
A-1 does exist because ATA = 1
A 0
Lagrangian becomes
1
1
1
1
1
1
L = ATA AVA = = k k k k k
2
2
2
2
2
2
Solutions are obvious

k = k k
k = Ck e ik t
= k
2
k
No cross terms
Normal coordinates are independent simple harmonic oscillators
Initial Conditions
The coefficients Ck is fixed by the initial conditions
Suppose at t = 0 = (0) = (0)
(0) = A(0)
j (0) = a jk Re Ck
(0) = A(0)
j (0) = a jk Re(i k Ck ) = a jk k Im Ck
Using ATA = 1
Re Ck = alk Tlj j (0)
Im Ck =
k = Ck e i t
alk Tlj j (0)
no sum over k
We need an example now
Remember
to take the
real part!
Linear Triatomic Molecule
Consider a molecule like CO2
Consider only motion along the axis
m 2
T = (1 + 22 + 32 )
2
m 0 0
T = 0 m 0
0 0 m
k
k
2
V = (2 1 ) + (3 2 ) 2
2
2
k
V = k
0
k
2k
k
0
k
k
We want to solve eigenvalue equation (V 2 T)a = 0

k 2m
k
0
V 2T = k
2k 2 m
k
=0
0
k
k 2m
2 (k 2 m)(3k 2 m) = 0
Solutions are 1 = 0
k
2 =
m
3 =
3k
m
Is this OK?
1
1
a1 =
1
3m
1
1
1
a2 =
0
2m
1
1
1
a3 =
2
6m
1
First solution is linear movement of the molecule

1 = 0
1
1
a1 =
1
3m
1
This is not an oscillation

Consider it as an oscillation with infinitely long period
Although V is minimum at the equilibrium, it does not
increase when the whole molecule is shifted
Position of the CoM is a cyclic coordinate
Total momentum is conserved
Two normal oscillation modes exist

1
1
k
a2 =
0
2 =
m
2m
1
3k
3 =
m
1
1
a3 =
2
6m
1
CoM does not move Orthogonal to the first solution
Putting together a1, a2 and a3

2
3
1
1
A=
2
0
2
6m
2 3 1
Normal coordinates are
1 =
m
3
(1 + 2 + 3 ) 2 =
2
m
1
A =
3
6
1
m
2
(1 3 ) 3 =
1 2
k
2
2
L = ( 1 + 2 + 3 )
( 22 + 3 32 )
2
2m
1 is cyclic as we expect
2
0
2
m
6
3
1
(1 22 + 3 )
Degenerate Solutions
We assumed j k for j k
What if the eigenvalue equation has multiple roots?

Can we still achieve ATA = 1 ?
Quick answer: Dont worry
Multiple root corresponds to multiple eigenvectors

V T = ( ) m f ( ) = 0
(V T)a j = 0 ( j = 1, , m)
= is an m-fold root
m eigenvectors
Any linear combination cjaj is also an eigenvector

It is always possible to find a set of m orthogonal vectors
Recipe
For an m-fold root and m eigenvectors
First normalize a1 with a1Ta1 = 1

Next turn a2 into a2 = a 2 (a1Ta 2 )a1
This satisfies a1Ta2 = 0
Normalize a2 with a2 Ta2 = 1
Next turn a3 into a3 = a3 (a1Ta3 )a1 (a2 Ta3 )a2
This satisfies a1Ta3 = 0 and a2 Ta3 = 0
Continue
It is now guaranteed that we can make ATA = 1
Summary
Studied oscillation
Equation of motion Eigenvalue problem Va = Ta
Showed that oscillating solutions exist i = Cai e i t

Eigenvalues 2 are positive definite
Provided that V is minimum at the equilibrium
Principal axis transformation diagonalizes T and V
Discussed general features of multi-dimensional oscillators
Normal coordinates behave as independent oscillators
Next: forced oscillation
Mechanics
Physics 151
Lecture 13
Oscillations
(Chapter 6)
Studied oscillation
Equation of motion Eigenvalue problem Va = 2Ta
Discussed general features of multi-dimensional oscillators

Showed that oscillating solutions exist i = Cai e i t
Eigenvalues 2 are positive definite
Provided that V is minimum at the equilibrium
Principal axis transformation diagonalizes T and V
Normal coordinates behave as independent oscillators
Forced Oscillation
Force Fj is applied to coordinate q j = q0 j + j
Switch to normal coordinates {i}
Linear force if qj is a Cartesian coordinate

New generalized force is Qi = Fj
j
i
j = Aji i

= Fj Aji or Q = AF
Without force, i was satisfying i + i2 i = 0
Now it must satisfy i + i2 i = Qi

No cross-terms between
coordinates
Sinusoidal Force
Suppose Fj is sinusoidal with time Fj (t ) = Re ( Fj eit )
Pressure from sound wave

Microwave radiation on molecule in food
Any periodic force will do Fourier series
Generalized force on a normal coordinate is

Qi (t ) = Qi e i t = Fj Aji e i t
Equation of motion is
i + i2 i = Qi eit
Thats easy to solve
Complex
number
General Solution
i + i2 i = Qi e i t
General solution for inhomogeneous equation is
1 particular solution + any solution for the homogeneous equation

Qi
ii t
i t
+ Ci e
In this case i = Bi e
with Bi = 2
i 2
Forced
solution
Force-free
solution
Ci is arbitrary Fixed by the initial conditions

Amplitude Bi of forced oscillation depends on i2 2
The closer to the natural frequency, the bigger
An example will help here
Triatomic Molecule
Take the triatomic molecule from last week
Suppose the atoms are charged

Molecule is polarized
Plane EM waves are arriving
i t
E = E0 e
at the atoms
Forces are
q
F = 2q E0 e i t
q
Switch to the normal coordinate
Use A from last week
2q
q
F = 2q E0 e i t
q
Triatomic Molecule
2
1
A=
2
6m
2
0
2 = A
3 1
3
2q
Generalized force for is

0

Q = AF = 0 qE0 e i t
6/ m
Only 3 gets forced oscillation

Direction of the forces are just right for this mode
Triatomic Molecule
Send the EM waves at an angle
Fj have different phases now
qe i ( t + dk sin )
i t
F = 2qe
E0 cos
qe i ( t dk sin )
A bit of
work
2
1
A=
2
6m
2
2
0
3 1
3
d
q
32m (cos 1)
2
Q = i m sin qE0 cos e i t
2
3m (cos + 2)
dk sin
2q
2
=
k
Triatomic Molecule
32m (cos 1)
2
Q = i m sin qE0 cos e i t
2
3m (cos + 2)
Oscillation of 2 is maximized by
F = 2 qE0 cos e i t
dk sin =
2
+i
1st and 3rd atoms are driven in
the opposite directions
Generally, a normal mode is excited

when the external forces push the
particles in the direction the mode
wants to move them
2
=
k
Microwave Ovens
Microwave ovens use microwaves
Microwave generator (magnetron) was invented during

WW-II for radars
Use for food was discovered accidentally Radarange
Frequency (~2.5 GHz) must be close to resonance
Water, sugar and fat has vibration modes with natural

frequencies around there
Wavelength much longer than the size of the molecule
All atoms feel same E field Direction doesnt matter
Molecules must be polarized
Dissipative Force
i = Bi e
+ Ci e
ii t
Qi
Bi = 2
i 2
Forced solution goes when = i
i t
Because there is no friction/loss/dissipation
no sum over i
Introduce velocity-dependent frictional force Fi = Fii
Non-conservative force cannot be expressed by potential

Add this to Lagrange's equations by
d L L
= Qi = Fij q j
dt qi qi
In general not diagonal
Dissipation Function
Force Fij q j dissipates energy
Rate of work done against it is given by Fij qi q j
Rayleighs dissipation function is defined by

F=
1
Fij qi q j
2
F
Fi =
qi
2F = rate of energy dissipation
Lagranges equation is modified as

d L L F
=0
+
dt qi qi qi
This is invariant under

coordinate transformation
Coordinate Transformation
Switch coordinates from {qi} to {si}
qi = qi ( s1 ,..., sn , t )
Lagranges equation is invariant

d L L
= Fi
dt qi qi
d L L
= Qi
dt si si
We know that the forces transform as Qi = Fj
F
If Fi =
qi
q j
si
F q j
F q j
F
Qi =
=
=
q j si
q j si
si
Dissipation Function
d L L F
Switch
+
=
0
coordinates

dt qi qi qi
Modified Lagranges equation with dissipation

function is invariant under coordinate transformation
Handy when moving to normal coordinates
Its still a kluge
d L L F
+
=0
dt si si si
No fundamental principle behind it

Unavoidable Dissipative forces are not fundamental
We can use it to solve our problem at hand
Damped Oscillator
With dissipation, equation of motion becomes

i + Fij j + i2 i = 0
no sum over i
Fij is generally not diagonal
If it is diagonal, i.e., Fij = Fi ij
i = Ci e i t
2 i Fi + i2 = 0
2
F
Fi
2
i
= i
i
4
2
Oscillation
no sum over i
Solution is an oscillation with

exponentially decreasing amplitude
F
exp i
2
True even if Fij

is non-diagonal
Triatomic Molecule
Suppose there is friction Fi = f i
f is same for all the atoms

1
f
Fij diagonal

F = f ii =
at this stage
2
2
2
1
A=
2
6m
2
2
0
3 1
3
F1
Switch to normal coordinates

= A
f
F = AA
2
In this problem, T =
m 0 0
m
0
0
0 0 m
F2
F3
Fij not necessarily

diagonal, but
= m1
f
f
f

F = AA =
ATA =
2
2m
2m
Fij happens to be diagonal
Equation of Motion
f
With the dissipation function F =

2m
2
EoM is i + i + i i = 0 no sum over i
Normal-mode solutions are i = Ceit

2 i + i2 = 0
i
= i2
4
2
All solutions decay with exp ( 2 t )

Energy (amplitude)2 decays with exp ( t )
Rate of energy loss
f
Fij =
m
Damped-n-Forced Oscillation
Turn on the EM waves

Q1 = Q2 = 0
2q
6
qE0 e i t
m
Q3 =
Equation of motion for 3 is
3 + 3 + 32 3 =
Solve it by
3 = Cei t
6
qE0 e i t
m
6
( i + )C =
qE0
m
2
2
3
qE0
6
C=
m 2 i + 32
We know how this

looks like
In Case You Forgot
C=
qE0
6
m 2 i + 32
Phase
Amplitude
0
arg(C )
C
Q=3
Q = 10
Q = 10
Q=3
2m qE0
3 k
There is resonance at = 3
3
Near-resonance behavior depends on Q =
Quality
factor
Summary
Discussed forced oscillation
Qi
i2 2
Qi depends on how well the forces match the motion pattern
of the normal mode
Solution i = Bi e i t + Ci e iit with Bi =
Introduced dissipation function F

Generalized force given by Q = AF
Modifies Lagranges equation
Next: Special Relativity
d L L F
+
=0
dt qi qi qi
Now thats something I really like
Mechanics
Physics 151
Lecture 14
Special Relativity
(Chapter 7)
Special Relativity
You know it quite well already
We skip all the historical discussions
Learned in Physics 15a/16

E.g. about non-existence of absolute time
Start from two principles and build a clean formalism
Laws of physics are the same in all inertial frames

Speed of light in vacuum is the same in all inertial frames
Maxwells equations are always correct
Frames and Events
Consider two inertial frames S and S
An event is specified by the position and the time

(t , x, y, z ) in S (t , x, y, z ) in S
Consider the distance between two events 1 and 2
We can always move the origins so that event 1 is

(t , x, y, z ) = (t , x, y, z ) = (0, 0, 0, 0)
Physics does not depend on absolute position or time
We denote event 2 with (t , x, y, z ) and (t , x, y, z )
Spacetime Distance
Light is emitted at event 1 and detected at event 2
Speed of light, measured in S and S, must be the same
c=
x2 + y2 + z 2
=
t
x2 + y 2 + z 2
t
c 2t 2 ( x 2 + y 2 + z 2 ) = 0
c 2 t 2 ( x 2 + y 2 + z 2 ) = 0
Define distance in spacetime (s)2 = c 2t 2 ( x 2 + y 2 + z 2 )

We assume that the spacetime distance s between any
two events is the same in all inertial frames
Light Cone
(s ) 2 = c 2t 2 ( x 2 + y 2 + z 2 ) = 0 represents light
If an object slower than light travels from event 1 to 2,
t must be larger (s)2 > 0
If (s)2 < 0, there isnt enough time for even light to reach
Interval between two events can fall into 3 regions
(s ) 2 > 0
timelike Reachable by ordinary objects
(s ) 2 < 0
spacelike
(s ) 2 = 0
lightlike Reachable by light
Reachable only by tachyons (v > c)
Light Cone
(s)2 = 0 draws a cone in the

4-dimensional space
Light cone divides spacetime

into past, future and elsewhere
t
(s ) 2 > 0
future
elsewhere
Division is frame-independent
Protects causality
Past is past, future is future,
no matter which frame you are in
(s ) 2 < 0
past
Time Dilation
Consider an object moving at a velocity v in S

(dx, dy, dz ) = (vx , v y , vz )dt
Define S so that this object is constantly at its origin

Consider small movement from event 1 to event 2
(s ) 2 = (ct ) 2 in S
= (ct ) 2 (x 2 + y 2 + z 2 ) in S
Make it infinitesimal
(cdt ) 2 = (cdt ) 2 (dx 2 + dy 2 + dz 2 ) = (c 2 v 2 )dt 2
dt d = 1 vc2 dt = 1 2 dt < dt
2
Proper time
Moving clock
appears to run
slower
4-Vectors
Express an event (t, x, y, z) as a

4-dimensional vector
c is there to fix the unit

Length of this vector in ordinary
4-d space would be
x = xx = c 2t 2 + x 2 + y 2 + z 2
ct
x ct

=
x=
y r

z
Totally useless for us
Its more useful to define length as

1 0 0 0
0 1 0 0
2
x = c 2t 2 ( x 2 + y 2 + z 2 )
x = xx = x
0 0 1 0
Metric tensor g
0 0 0 1
Minkowski Space
x = x x = xgx = c 2t 2 ( x 2 + y 2 + z 2 )
2
4-d space in which the length is defined as above is

called the Minkowski space
Metric tensor g expresses how length is defined

Metric tensor for ordinary space = unit matrix 1
We now say
Length of any 4-vector in Minkowski space is the same
in all inertial frames
Lorentz Transformation
So far we have not specified how S and S are related

The Question Of The Day is
What is the general form of the transformation

L
(ct , x, y, z )
(ct , x, y, z )
that keeps the spacetime distance s2 invariant?
We know the name: Lorentz transformation
Lets figure out its characteristics
What is the general form of the transformation

that conserves the length in the Minkowski space?
Linearity
Physics does not depend on absolute position or time
Shifting the origin of S must simply shift the origin of S

x x+a
x x + a i.e. L(x + a) = L(x) + a
Try x = 0 L(a) = L(0) + a
L(x + a) = L(x) + L(a) L(0)
If we define L(x) L(x) L(0)
L(x + a) = L(x) + L(a)
True for any x and a
L is linear
L(x) must be a linear function of x + fixed offset
Poincar Transformation
Knowing that the transformation is linear

ct
ct
x Any linear
x
x = x = transformation
y is written as
y

z

z
x = Lx + a
Translation
of origin
4x4 matrix
This is called Poincar transformation

a moves the origin No physical consequences
We ignore the translation and consider L only
This is called homogeneous Lorentz transformation

Let me write this as HLT
Now we have x = Lx
L is a 4x4 real matrix
What are the constraints on L?
L must conserve the length of any 4-vector

x = xgx
2
x = xgx = xLgLx
2
In terms of components
L ji g jk Lkl = gil
16 real equations
g is symmetric Equation is symmetric

6 of 16 equations are duplicates
There are 10 constraints
L has 6 degrees of freedom
LgL = g
Rotation
Remember how rotation matrices were defined?
Conserve the length of any 3-vector in Cartesian space

In fact any rotation A in 3-space satisfies the condition
x = c 2t 2 ( x 2 + y 2 + z 2 ) = const
2
Unaffected
Conserved
3-d rotation is a subset of HLT

Not the most exciting part of it
Rotation has 3 degrees of freedom (Euler angles)
There must be 3 more in HLT
Lorentz Boost
Consider two inertial frames S and S
Origin of S must be moving at a constant velocity in S

x = vx t y = v y t z = vz t
( x, y, z ) = 0
Rotate S and S so that x and x are parallel to v
Plus, y and y, z and z are parallel to each other

OK to do this because rotation is a subset of HLT
ct L00 L01 0 0 ct
x L
x
y and z
0
0
L
11
= 10

are out of
y 0
0 1 0 y
the game

0 0 1 z
z 0
L00
L
L = 10
0
Lorentz Boost
L must satisfy LgL = g
L01 0 0
L11 0 0
0 1 0
0 0 1
2
2
L200 L10
= 1 L201 L11
= 1 L00 L01 L10 L11 = 0
Origin of S is x = 0
0 = L10 ct + L11 x
This must satisfy x = vt
L00 L01
Solve!
=
L
10 L11
Looks familiar except
L10
v
=
c
L11
What are these

signs?
1
1 2
Sign Ambiguities
There are 4-fold sign ambiguities

L00
L
10
L01
=
L11
1
1 2
Sign of ct is arbitrary
Sign of x is arbitrary
Only 1 represents Lorentz transformation
Think about the low-velocity limit

0 1 0
1
0
1
L00
L
10
L01
=
L11
This must be unit matrix
Space-Time Reversal
What are the other solutions?
L00
L
10
L01
=
L11
L00
L
10
L01
=
L11

flips x axis: Space inversion

flips t : Time reversal
They reverse the orientation of time and/or space
Laws of physics are happy with such transformation

Not continuous part of the Lorentz transformation
We arent interested for the moment

L=
0
General Boost
Weve got L for boost in x
Split 3-vector r into two parts

(r )
Parallel Perpendicular
r =
2
r = r r
Parallel component transforms like x above

to v
0
0
We can rotate S and S to get L for boost in any direction
We can also use a bit of vector algebra
to v
ct = ct r = ct ( r )
r = ct + r + r = ct + r +
(r )( 1)
0 0
0 0
1 0
0 1
Proper Lorentz Transformation
Writing down explicitly for general
Proper Lorentz
transformation
x
y
z

x y
x2
xz
( 1) 2
x 1 + ( 1) 2 ( 1) 2
L=
yx
y2
yz
y ( 1) 2
1 + ( 1) 2 ( 1) 2
2
z y
z x
z

(
1)
(
1)
1
(
1)
2
2
2
=

1 + ( 1) 2
3 degrees of freedom (x, y, z)

General form of Lorentz
transformation without rotation
Lorentz T at a Glance
Proper Lorentz Transformation
Homogeneous Lorentz Transformation LA
PLT
3 degrees of freedom in
Orientation of the axes are unchanged
L=

1 + ( 1) 2
A is rotation. 3 + 3 = 6 degrees of freedom

Origin is unchanged
Poincar transformation x = LAx + a
a shifts the origin. 6 + 4 = 10 degrees of freedom

Completely general form of transformation between inertial
systems that satisfy special relativity
Properties of L
L 1 when 0
L-1 is given by
x
y
z

x y
x2
xz
1
(
1)
(
1)
(
1)
x
2
2
2
L=
yx
y2
yz
y ( 1) 2
1 + ( 1) 2 ( 1) 2
z
y

( 1) z 2 x
( 1) 2
1 + ( 1) z2
z
1 + ( 1) 1 + ( 1) = 1
2
2
Check
this!
Diagonally symmetric
Remember: rotation matrix A is anti-symmetric

LA (for HLT) is neither symmetric nor anti-symmetric
Addition of Velocities
Make two PLTs L and L in series
If and are parallel, we can define x as their direction

LL =

(1 + ) ( + )
=
+
+
(
)
(1
)
+
Same as one PLT with =
1 +
|| < 1 for any and as expected
Only ct and x
components
shown
Addition of
velocities
What if we add two velocities that are not parallel?
For example, is in x, is in y
0

1
0
LL = 0
0
0
0

Obviously LL LL
0
0
1
Only ct, x, y
components
shown
Ugly
asymmetric
matrix
PLTs are not commutative
In general, LL is asymmetric
On the other hand, two HLTs do add up to a HLT
Two PLTs do not add up to a PLT

HLTs are defined as linear transformations that conserve
length of 4-vectors and do not move the origin
Two successive HLTs automatically satisfy this
Therefore two PLTs must make a HLT
i.e., LL must be written as LA

i.e., PLT + PLT = PLT + a rotation
Goldstein section 7.3 makes a big fuss about this
Lorentz Group
Homogeneous Lorentz transformations make a closed

group (Lorentz group)
Rotations make a group
Product of two members is always a member

Which is a subgroup of the Lorentz group
Proof Eulers theorem
Proper Lorentz transformations do not make a group
Summary
Discussed Lorentz transformation
Derived general form of homogeneous Lorentz

transformation
Linear transformation of 4-vectors that conserve the length

in Minkowski space
Product of rotation and proper Lorentz transformation

Found explicit matrix expression of PLT
HLTs form a group, PLTs dont
Now we know how the spacetime transforms
Next: physical quantities in the spacetime
Mechanics
Physics 151
Lecture 15
Special Relativity
(Chapter 7)
Defined Lorentz transformation

in Minkowski space

transformation

HLT HLT = HLT
PLT PLT = HLT
Fun With Paradox
In general, two PLTs dont add up to a PLT
Rotation becomes involved
Example: two objects are moving in parallel

rotation!
stationary
observer
Seen from
an observer
moving
upwards
Can you see where the rotation come from?
moving
observer
Fun With Paradox
How do you know two objects passed the line

simultaneously?
By sending light and receiving reflection
l
c
l
t
=
+
Reflections come back at
l
c
Took same time same distance
l
Came back simultaneously reflected
at the same time
What happens if the observer was moving?
Light is sent at t =
Fun With Paradox
For a moving observer

A
Light is sent at the same moment
Reflection from A comes back

earlier than from B
A must have passed earlier!

B
Definition of simultaneous
depends on the observer
Causing this effect
rotation!
Discussed Lorentz transformation

transformation

in Minkowski space

We got the stage = spacetime
Lets get the actors = physical quantities
4-Vectors
We write 4-vectors as x
Greek index = 03
x 0 = ct x1 = x x 2 = y x 3 = z
It seems confusing, but youll get used to it
Lets follow a particle traveling in 4-space
Trajectory is given by x 0 ( ), x1 ( ), x 2 ( ), x 3 ( )
is a parameter that varies monotonously along the curve
Proper time is a convenient possibility for
At any point on the curve, we can define a tangent 4-vector
dx
u =
d
Call it 4-velocity
4-Velocity
S is observers frame. S is the particles rest frame

Particles 3-velocity in S is v
Roman index = 13
x 0 = ct = c
x i = i c = v i
We define 4-velocity
x 0
u =
=c
Length is
0
x i
u =
= vi
u 0 u 0 u i u i = 2 (c 2 v 2 ) = c 2
Space part is proportional

to the 3-velocity
Lorentz invariant
4-velocity is the relativistic extension of v

Its timelike, by the way
4-Momentum
Multiplying 4-velocity with mass gives 4-momentum

p 0 = mu 0 = m c
p i = mu i = m v i
Space part is natural extension of 3-momentum p = m v

Time part is energy E / c = p 0 = m c
This needs to be confirmed after introducing force
Lorentz invariance is obvious
p 0 p0 pi pi = m2c 2
E 2 = m2c 4 + p 2c 2
Kinetic energy is defined as

T = E Ev =0 = m c + p c mc
2 4
2 2
Becomes p2/2m
at low-speed limit
Lorentz Tensor
Proper Lorentz transformation turns a 4-vector into

another 4-vector
Consider it a linear function of 4-vector
Express it as x = L x
Upper index = 4-vector
Lower index = function that accepts 4-vector
You can define a whole bunch of quantities using this

convention Call them tensors
X
X X X X X
Well find their physical meanings as we go
Tensor Product
Tensor product of two 4-vectors is defined by

T = u v
Write this as
T = u v
T is a tensor of rank 2
You can repeat this to define tensors of rank n
Lorentz transformation of T can be easily found

T = u v = L u L v = L L T
Use as many Lorentz tensor as necessary to convert all

indices
This (irresponsible-sounding) rule works
for Lorentz transformation of any tensor
Scalar Product
We define the scalar product of two 4-vectors

u v = u 0 v 0 u i v i = u g v
Two lower indices

Metric tensor
because it takes two 4-vectors and
returns a scalar
How does it transform?
1 0 0 0
0 1 0 0
g=
0 0 1 0
0
0
0
1
u v = u g v = u L g L v = u g v = u v
Scalar product is Lorentz invariant, as expected
Metric Tensor
A coordinate system in general has basis vectors

u = u e = u 0e0 + u1e1 + u 2e 2 + u 3e3
Scalar product u v can be written as
u v = u e v e
g = e e
Metric tensor defines the scalar products of the basis vectors

Lengths of and angles between the basis vectors
Thats what metric means
General Metric Tensor
Metric g in Minkowski space is diagonal
Coordinate axes are always orthogonal
Formalism of tensors allows more flexibility
Useful in curved space coordinates

General relativity makes full use of this
Lets not get into it for now
1-Form
Lets look at a scalar product in a different way

scalar
u v = u g v
4-vector
function that turns a 4-vector into a scalar
We call u = u g as the 1-form of u

x0 = x 0 = ct
xi = x i
Difference between 4-vector and 1-form seems small
Why do we make such distinction?

non-Minkowski metric can make them really different
There are physical quantities that are naturally 1-form
Gradient
Consider a scalar function f ( x )
A particle goes along a curve x = x ( )

Rate of increase of f ( x ( )) is given by
f ( x ) dx f ( x )
gradient of f
=
v
x d
x
Gradient operates on velocity to make a scalar 1-form
Gradient operator is defined by =

x
Also known as d
Lower index shows
But Ill avoid this notation
its a 1-form
4-Vector and 1-Form
4-vector can be turned into its 1-form by u = u g
Obviously you can do the reverse

u = g u where g g =
g looks identical to g in Minkowski space
This gives us Lorentz transformation for 1-form

u = g u = g L u = g L g u = L u
Works just the same as 4-vector, as it should
Rank of Tensors
A general tensor has n upper and p lower indices
()
n
p
T is rank
()
2
1
Call it a tensor of rank
It takes p 4-vectors and n 1-forms and return a scalar
T a b c = scalar
4-vector and its 1-form are interchangeable using g
We can turn a tensor into equivalent tensors with different

rank, as long as n + p is conserved
Example:
T a b c = T ( g a )b c
T = T g
We can find Lorentz transformation for any tensor
Transform all indices using Lorentz tensor

Example:
T a b = scalar Transform this to get
T a b = T L a L b = T a b
T = L L T
We now know all the rules for
Lorentz transformation
Moving indices up and down
They arent even all that complicated
Force
Newtons laws must be correct if the velocity is zero
dp
in the rest frame of the object
F = p =
dt
Momentum transforms as a 4-vector
Time dilation changes the time derivative
dp
Natural extension would be
= K
d
K must be a 4-vector
is proper time. Connected with t by dt = d
How do we find the 4-force K ?
Electromagnetic Force
We assume Maxwells equations are always correct
Predicts constant speed of light c

i.e. relativistically correct
We want to rewrite it in a covariant form
EM force on a charged particle can be derived from

This should better
the generalized potential U = e( A v)
be a scalar
4-velocity u was (u 0 , u) = ( c, v )
Define 4-potential as A = ( A0 , A) = ( c , A)
Looks
promising
Scalar product is
A u =
c A v = ( A v )
U = eA u
U
=
e
(
A
u ) is a scalar if A = ( c , A) is a 4-vector
What we are looking for if the laws of EM are covariant
U d U
+
Force in 3-d is given by Fi =
xi dt xi
Extend this to 4-d
Careful to make it a real 4-vector
(eA u ) d
K =
d
x
(eA u )
A
dA
u
= e
d
u
x
Minkowski force for a charged particle in EM field
A
dA
K = e
u
x
d

A0
Ai
A
A
u0 +
ui
u0
ui
= e
x
x0
xi
Rewrite using E and B

i
0
i
A
A
A
i
E =
= c
x
x
t
0
i
A j Ai u j
( v B ) = ( v ( A) ) = +
xi x j
i
A bit of
work
K =
0
ev E
i
i
K = e E + ( v B)
i
dp
i
i
i
Space part
= K = e E ( v B)
d
dp
Agrees with
= F = e E ( v B )
dt
dp 0
i i
0
= K = ev E
Time part
d
c
dp 0 W
=
dt
c
Integrate
Rate of work done

by the EM force
E
p =
c
0
Energy!
Confirmation
promised earlier
Faraday Tensor
A
dA
K = e
u
x d

u eF u
Tensor F is called Faraday, or EM field tensor
A A
= e
x x
0
E
A A
x
=
x x E y
Ez
Ex
0
cBz
cBy
E and cB form a tensor of rank 2

cf. /c and A make a 4-vector
Ey
cBz
0
cBx
Ez
cBy
cBx
Other Forces
What happens with forces other than EM?
There is no general method for making forces covariant

You must deal with each force, case by case
There are 4 (known) fundamental forces in nature
EM, gravity, weak and strong

Covariant form has been found for weak and strong
Need quantum field theory to do this
Gravity cannot be made covariant
Need general relativity
Summary
Defined covariant form of physical quantities
4-vectors: velocity, momentum

Tensors: metric, Lorentz
1-forms: gradient
Found how to Lorentz transform them

Covariant form of Newtons equation with EM force
EM potential 4-vector, EM force tensor
dp
Equation of motion
= K
d
Mechanics
Physics 151
Lecture 16
Special Relativity
(Chapter 7)
Defined covariant form of physical quantities
Collectively called tensors

Scalars, 4-vectors, 1-forms, rank-2 tensors,
Found how to Lorentz transform them
Use Lorentz tensor & metric tensor
Covariant form of Newtons equation with EM force

dp
= K
Equation of motion
d
EM potential 4-vector (/c, A)
EM field Faraday tensor I gave you a wrong one
Faraday Tensor
A A
K = e
x
x
u
e
F
u
Faraday Tensor
F is derivatives of the EM potential E and B fields
A0 A1
Ax
Ex =
= c
x t
x1 x0
Az Ay
A3 A2
Bx =
=
+
y
z
x2 x3
Ex c E y c Ez c
0
E c
B
B
x
z
y
=
Ey c
Bx
0
Bz
0
Bx
Ez c By
What
WhatIIgave
gaveyou
you
(and
(andthe
thetextbook)
textbook)
was
waswrong
wrongby
byaa
factor
factorcc
Multi-Particle System
Consider a system with particles s = 1, 2,
Total momentum P = ps
s
dps
Equation of motion for each particle
= K s
d s
Different time for each particle!
dps
dP
EoM for the total momentum is
= s
= s K s
dt
d s
s
s
Not a very clean equation
Trouble ahead
dp
dP
= s s = s Ks
dt
d s
s
s
Imagine a 2-particle system with no external force
But there are internal forces between the particles

Law of action and reaction K12 = K 21
dP1+ 2
= 1 K 21 + 2 K12
dt
This is zero only if
1 = 2
To conserve the total momentum in all frames,

the particles interacting with each other must have the
same velocity
Is this a weird restriction, or what?
Local Interaction
Suppose the particles interact only when in contact
Forces exchanged only when they collide

They share the rest frame momentarily Same
Total momentum of a multi-particle system is conserved
if the interactions between the particles are local
Furthermore, Law of action/reaction K12 = K 21

cannot work over a distance instantaneously
There must be delays Conservation laws cannot hold
Relativity and non-local interactions dont mix
Particle Collisions
Interactions between particles must be local
Force exchanged when they collide

Free motion between collisions
Consider the collision as a black box

?
We dont know what happens in the box (not classical)

Motion outside the box is easy Relativistic Kinematics
How much can we learn without opening the box?
Center-of-Momentum Frame
Local interactions conserve total 4-momentum
i.e., total energy and total 3-momentum are conserved

n
n
n
E
p = ps = , p
E = Es p = p s
c
s=1
s =1
s =1
We know how to Lorentz transform it
as usual
p = psi = L psi = L p
si
si
Define the center-of-momentum frame in which p = 0
Its the frame in which the total 3-momentum is zero

Or, the center of mass is at rest
Often called center-of-mass frame as well
CoM Energy and Boost

E
E
p = ps = , p CoM frame
p = , 0
c
c
s =1
There are two particularly useful quantities
CoM energy E
Lorentz invariance p p = p p
E is the smallest possible E
E 2 E 2
2
=
p
c2
c2
Boost of CoM frame relative to the lab. frame
Lorentz transformation
E E
,
p = L p =
c
c
pc
=
E
E
E
Two-Particle Collision
Consider collision of particle 1 on particle 2 at rest

p1 = ( E1 c , p1 )
p2 = (m2 c, 0)
Fixed-target collision
Total 4-momentum is p = ( E1 c + m2 c, p1 )
Total CoM energy
E 2 = p p c 2 = (m12 + m22 )c 4 + 2 E1m2 c 2
Boost of CoM frame is
p1
m1 1 v1
=
=
E1 c + m2 c m1 1c + m2 c
E grows slowly with E1

Approaches v1/c for large E1
Creation Threshold
Suppose we are trying to create a new particle
In the best scenario, particles 1 and 2 merge to create a new

heavy particle 3
Total 4-momentum would be simply p = p1 + p2 = p3
Total CoM energy is
CoM energy must
2
2
2 2
2
match the mass of
E c = p3 p3 = m3 c
E = m3c
How much energy E1 do we have to

give particle 1?
E 2 = (m12 + m22 )c 4 + 2 E1m2 c 2 = m32 c 4
For large m3, E1 grows with m32
the new particle
(m32 m12 m22 )c 2

E1 =
2m2
Fixed-Target vs. Collider
Consider hitting a proton with a proton to make a

Higgs particle, which is X times heavier than a proton
(m32 m12 m22 )c 2 X 2
E1 =
mpc2
2m2
2
For X > 100, wed need a >5000 GeV accelerator
Particle colliders are more energy-efficient

p1 = ( E1 , p1 )
p2 = ( E2 , p 2 )
Laboratory is CoM p = ( E1 + E2 , 0)
Just need E1 + E2 = m3c 2 = Xm p c 2
p1 = p 2
50 GeV + 50 GeV
Elastic Scattering
Particle 1 hits particle 2 and get elastically scattered

y
p1
x
p3
p3
CoM
p4
p1
p2
p4
Cross section is calculated in CoM frame

By treating it as a central-force problem
Experiment is done in the laboratory frame
We need to learn how to translate between the CoM

and the laboratory frames
Elastic Scattering
First, whats the boost?
Total momentum is p = p1 + p2 = ( E1 c + m2 c, p1 )
p
p1 c
= 0 =
p
E1 + m2 c 2
Lets get as well
p p = ( E1 c + m2 c) 2 p1 = m12 c 2 + m22 c 2 + 2 E1m2
2
p0
p p
E1 + m2 c 2
m12 c 4 + m22 c 4 + 2 E1m2 c 2
Elastic Scattering
Now we can boost p1 to CoM

p1 = ( E1 c , p1 , 0, 0)
p1 = ( E1 c , p1, 0, 0)
E1 = ( E1 p1c)
p1 = ( p1 E1 c)
p3 is given by rotating p1 by
p3 = ( E1 c , p1 cos , p1 sin , 0)
Boost this back to get p3
E3 = ( E1 + p1c cos ) p31 = ( p1 cos + E1 c)
p32 = p1 sin
Scattering angle in lab frame is
p
sin
sin
tan =
=
=
p
(cos + E1 ( p1c)) (cos + 1)
2
3
1
3
Velocity of
1 in CoM
Elastic Scattering
What happens to the kinetic energy?

E3 = 2 (1 2 cos ) E1 (1 cos ) p1c
Makes sense
At = 0 E3 = 2 (1 2 ) E1 = E1
With a little bit of work
Sign wrong in textbook
T3
2 (1 + E1 2)
2
Kinetic
E
=
T
m
c
(1 cos ) 1 1 1
= 1
2
(1 + ) + 2 E1
T1
energies
= m1 m2
Worst case is =
(T3 ) min
(1 ) 2
=
(1 + ) 2 + 2 E1
T1
Elastic Scattering
(T3 ) min
(1 ) 2
=
(1 + ) 2 + 2 E1
T1
Non-relativistic limit
(T3 ) min (1 ) 2
If m1 << m2, i.e., the target is heavy,
=
almost no energy is lost in the collision
(1 + ) 2
T1
Ultra-relativistic limit
(T3 ) min (1 ) 2 (m2 m1 ) 2 c 2
=
=
(T3)min is independent of T1
2 E1
2m2T1
T1
As T1 increases, the energy loss becomes very large
Particle Decays
Some particles are unstable and decay after a while

p = pi
i
Mass of the mother is known from the 4-momenta of the

daughters by calculating the total CoM energy
mc 2 = E = p p
Also called the invariant mass of the system
This is how particle physicists find particles that do not live

long enough to be directly seen
Example: Discovery of J/ (November 1974)
Particle Decays
A day at the BABAR experiment at SLAC
Collide e+ and e to generate a few 100,000 (4S) particles

each of which decays into two B0 mesons
some of which decays into a J/ and a KS
J/ decays into e+ and e, or + and
KS decays into + and
Measure 3-momenta of the stable particles
Masses known Calculate 4-momenta
Rebuild the decay chain backwards and calculate invariant
masses of them all Do they match the expected masses?
J/ mass
Combine e+e or + and to see if they make a J/
B0 mass
Events / 2.5 MeV/c
Combine J/ and KS to make B0

180
160
BABAR
140
120
100
80
Found 440 signals

out of 30 million
(4S) generated
60
40
20
0
5200
5210
5220
5230
5240 5250 5260 5270 5280 5290 5300

2
Beam-Energy Substituted Mass (MeV/c
)
Summary
Discussed multi-particle system
Relativistic kinematics for particle physics
Only local forces are amiable with relativity

CoM frame, CoM energy, invariant mass and boost
Particle creation and collider experiment
Elastic scattering
Particle decays
Next lecture: Lagrangian formalism
Mechanics
Physics 151
Lecture 17
Special Relativity
(Chapter 7)
Worked on relativistic kinematics
Essential tool for experimental physics

Basic techniques are easy:
Define all 4 vectors
Calculate c-o-m energy and boost
Go about with business
Examples:
Particle creation
Elastic scattering
Particle decays
Todays Goals
Relativistic Lagrangian formulation
Two different approaches: practical and truly relativistic

Neither is perfect Will cover both
Will do a few easy examples in the process
Proper Approach
Set up a covariant form of Hamiltons principle

Keep everything in clean tensor forms
Practical Approach
Build a Lagrangian that reproduces 3-force in a frame

May or may not be correct in other frames
Works OK pretty often, but no guarantee
Practical Formalism
For a single particle of mass m

L = mc 2 1 2 V (x)
= reduced velocity
Lets check if this works

1 2
L
mc i
i
=
mc
=
=
p
vi
i
1 2
3-d equation
of motion
Space component
good. But no time
component
d L L
V
i
i
i

=
+
=
=0
p
p
F
i
i
i
dt v x
x
Looks OK for the 3-d part Try to push this path
Generalized Potential
Expand the definition to allow v-dependent potential
Consider the EM force
L = mc 2 1 2 U (x, v) = mc 2 1 2 q + qA v
We know that U gives us

U d U
i + i = qE i + q( v B)i
x dt v
Did this before

Still works fine
Only difference is the definition of the momentum

Pi =
Canonical momentum
L
i
i
=
p
+
qA
vi
Same thing happened

without relativity
Classical 3-momentum
No big deal
Energy Function
Energy function h is defined by
L = mc 2 1 2 V (x)
conservative
L
mc i vi
mc 2
2
2
h = x i L =
+ mc 1 + V =
+V
2
2
x
1
1
i
This is total energy

Its conserved if V is time-independent
Proved this before No changes by going relativistic
Simple Example
Particle accelerating under constant force
Electron in an electric field

L = mc 2 1 2 + eEx
E
e
=0
=V
Lagranges equation
d L L d mc
eE = 0
=

dt v x dt 1 2
dt 1 2
eE
=
mc
Integrate twice, assuming x = 0, v = 0 at t = 0
eE
mc
1+ ( t)
eE
mc
mc 2
x=
eE
eE
1 + ( mc
t )2 1
Simple Example
mc 2
x=
eE
x
eE
1 + ( mc
t )2 1
nonrelativistic
Relativistic solution is a
hyperbola
mc 2
x = ct
eE
Approaches v = c
Non-relativistic solution
(parabola) accelerates faster
Simple Example
=
eE
mc
1+ ( t)
eE
mc
mc 2
x=
eE
eE
1 + ( mc
t )2 1
eE
eE 2
t x=
t
2m
m
Low-velocity limit v =
t limit 1 x ct
Look at it in terms of energy

mc 2
x=
( 1)
eE
All as expected
eEx = mc 2 ( 1)
LHS = V ( x) RHS = p 0 c mc 2 = T
Energy
conservation
Relativistic Oscillator
Consider a 1-dim. harmonic oscillator

L = mc 2 1 2 V
V = 12 kx 2
Lets use energy conservation this time

E=
mc 2
1
+ V = const
2 4
m
c
2 = 1
>0
2
(E V )
Solution exists only when

E V > mc 2
Oscillation between two
points expected
Whats the
frequency?
V ( x)
E
E mc 2
Semi-Relativistic Oscillator
Integrate for of the cycle
1 dx
m2c 4
=
= 1
c dt
( E V )2
Oscillation period
b
c 1 ( EmVc )2
2 4
dx
b is given by E = mc 2 + 12 kb 2
Nasty
integral
E V
k
2
2
2
2
=
+
b
+
b
x
1
(
)
1
(
)
2
2
mc
2mc
Approximate for V << mc2

E V
= 1+
2
mc
1
1 (1 + )
1
2 3
1 + 34
2
Semi-Relativistic Oscillator
4 b 1 + 34 (b 2 x 2 )
2
=
dx =
2
2
0
c
c 2
2 (b x )
Period is longer than non-relativistic oscillator
3kb 2
3 Vmax
=
=
2
0 16mc 8 mc 2
m
3kb 2
3 2
1 +
1 + b = 2
2
k 16mc
8
Wrong sign in textbook /
Relativistic solution slower than the non-relativistic one

Difference depends on the amplitude of oscillation
Limitations of Practical Approach

2
2
L
=
mc
1
V (x) gives correct relativistic answers
for many practical problems

It is an ad-hoc technique
Not Lorentz covariant by construction

Time is treated separately from space
Lorentz transformation of Lagrangian is not given
Must redefine L in each inertial frame
Truly relativistic theory should respect relativity from

the principle all the way up
Lets see how well it works
Practical Approach
Build a Lagrangian that reproduces 3-force in a frame

May or may not be correct in other frames
Works OK pretty often, but no guarantee
Proper Approach
Set up a covariant form of Hamiltons principle

Keep everything in clean tensor forms
but it quickly runs into difficulties even for a single particle. For a system
of more than one particle, it breaks down almost from the start. No satisfactory
formulation for an interacting multiparticle system exists in classical relativistic
mechanics except for some few special cases
Goldstein, p. 313
Truly Relativistic Formalism
Hamiltons principle
We want the action integral to be Lorentz scalar

Integration should not be by t, but by a Lorentz-invariant
variable Proper time could be a good choice?
Lagrangian L must then be a Lorentz scalar
Lagranges equation should look like

L d L

x
d u
I = Ldt = 0
=0
Symmetric for time and

space components
Solution is not unique. None of them perfect
Lets look at one Goldstein Section 7.10 for more
Free Lagrangian
We try a force-free Lagrangian = 12 mu u
Looks like the non-relativistic kinetic energy

Lorentz scalar
d d (mu )
=0
Lagranges equation would be
=
d u
d
Conservation of 4-momentum
Time component is conservation of energy
Energy function doesnt give total energy, though
1
1 2
h=u
= mu u = mc
u
2
2
Conserved,
but not energy
EM Force
We know only one force in 4-vector form EM
Potential was given by qu A
Lagrangian can be ( x , u ) = 12 mu u + qu A
Lagranges equations
d
d
A
d

( mu + qA ) qu = 0
=
d
x
x
A dA
d (mu )
K
= qu
d
x
d
This looks promising
4-force found
last week
Limitations of Purist Approach
We dont know 4-force for anything but EM
Most real-world problems cannot be solved this way
What to do with multi-particle system

I = Ld
L d L

d u
x
=0
Proper time of what?
Lagrangian formalism allows coordinate transformation

Each coordinate does not correspond to a single particle
Problem will be solved only when we give up the

particle picture
Summary
Constructed Lagrangian formulation
Practical approach provides useful tools

Relativistic solutions can be
L = mc 2 1 2 V (x)
found for many systems
Not really relativistic at heart
Purist approach can be built only for limited cases
E.g. single particle in EM field
= 12 mu u + qu A
Done with special relativity
Next: Hamiltonian formalism
Mechanics
Physics 151
Lecture 18
Hamiltonian Equations of Motion
(Chapter 8)
Whats Ahead
We are starting Hamiltonian formalism
May not cover Hamilton-Jacobi theory
Hamiltonian equation Today and 11/26

Canonical transformation 12/3, 12/5, 12/10
Close link to non-relativistic QM
Cute but not very relevant
What shall we do in the last 2 lectures?
Classical chaos?
Perturbation theory?
Classical field theory?
Send me e-mail if you have preference!
Hamiltonian Formalism
Newtonian Lagrangian Hamiltonian
Describe same physics and produce same results

Difference is in the viewpoints
Symmetries and invariance more apparent
Flexibility of coordinate transformation
Hamiltonian formalism linked to the development of
Hamilton-Jacobi theory
Classical perturbation theory
Quantum mechanics
Statistical mechanics
Lagrange Hamilton
Lagranges equations for n coordinates

d L L
= 0 i = 1, , n
dt qi qi
2nd-order differential
equation of n variables
n equations 2n initial conditions qi (t = 0) qi (t = 0)
Can we do with 1st-order differential equations?
Yes, but youll need 2n equations

We keep qi and replace qi with something similar
L(q j , q j , t )
We take the conjugate momenta pi
qi
Configuration Space
We considered (q1 , , qn ) as a point in an n-dim. space
Called configuration space

Motion of the system
A curve in the config space
qi = qi (t )
When we take variations,

we consider qi and qi as
independent variables
i.e., we have 2n independent variables in n-dim. space

Isnt it more natural to consider the motion in 2n-dim space?
Phase Space
Consider coordinates and momenta as independent
State of the system is given by (q1 , , qn , p1 , , pn )

Consider it a point in the 2n-dimensional phase space
We are switching the

(qi , qi , t ) (qi , pi , t )
A bit of mathematical trick

is needed to do this
qi = qi (t )
pi = pi (t )
Legendre Transformation
Start from a function of two variables f ( x, y )
Total derivative is
f
f
df = dx + dy udx + vdy
x
y
Define g f ux and consider its total derivative
dg = df d (ux) = udx + vdy udx xdu = vdy xdu
i.e. g is a function of u and y
g
If f = L and ( x, y ) = ( q, q )
g
= x
=v
u
y
L(q, q ) g ( p, q ) = L pq
This is what
we need
Hamiltonian
Opposite sign from

Legendre transf.
Define Hamiltonian: H (q, p, t ) = qi pi L(q, q, t )
Total derivative is
L
L
L
dH = pi dqi + qi dpi
dqi
dqi
dt
qi
qi
t
L
Lagranges equations say
= pi
qi
L
dH = qi dpi pi dqi
dt
t
This must be equivalent to
H
H
H
dH =
dpi +
dqi +
dt
pi
qi
t
Putting them
together gives
Hamiltons Equations
H
H
H
L
=
= qi
= pi and
We find
t
t
pi
qi
2n equations replacing the n Lagranges equations
1st-order differential instead of 2nd-order

Symmetry between q and p is apparent
There is nothing new We just rearranged equations
First equation links momentum to velocity

This relation is given in Newtonian formalism
Second equation is equivalent to Newtons/Lagranges
equations of motion
Quick Example
Particle under Hookes law force F = kx

L=
m 2 k 2
x x
2
2
p=
L
= mx
x
m 2 k 2
H = xp L = x + x
2
2
p2 k 2
=
+ x
2m 2
Hamiltons equations
H
H p
= kx
p=
x=
=
x
p m
Replace x with
p
m
Usual harmonic
oscillator
Energy Function
Definition of Hamiltonian is identical to the energy

function h(q, q, t ) = qi L L(q, q, t )
qi
Difference is subtle: H is a function of (q, p, t)
This equals to the total energy if
Lagrangian is L = L0 (q, t ) + L1 (q, t )qi + L2 (q, t )q j qk

Constraints are time-independent
This makes T = L2 ( q, t ) q j qk
See Lecture 4, or
Forces are conservative
Goldstein Section 2.7
This makes V = L0 ( q )
Hamiltonian and Total Energy
If the conditions make h to be total energy, we can

skip calculating L and go directly to H
For the particle under Hookes law force

p2 k 2
H = E = T +V =
+ x
2m 2
This works often, but not always
when the coordinate system is time-dependent

e.g., rotating (non-inertial) coordinate system
when the potential is velocity-dependent
e.g., particle in an EM field
Lets look at this
Particle in EM Field
For a particle in an EM field

m 2
L = xi q + qAi xi
2
pi = mxi + qAi
We cant jump on H = E
because of the last term, but
mxi2
H = (mxi + qAi ) xi L =
+ q
This is in fact E
2
Wed be done if we were calculating h
For H, we must rewrite it using pi = mxi + qAi

( pi qAi ) 2
H ( xi , pi ) =
+ q
2m
Particle in EM Field
Hamiltons equations are

H pi qAi
xi =
=
pi
m
( pi qAi ) 2
H ( xi , pi ) =
+ q
2m
p j qAj Aj
H
pi =
=q
q
xi
xi
xi
m
Are they equivalent to the usual Lorentz force?

Check this by eliminating pi
Aj
d
(mxi + qAi ) = qxi

q
xi
xi
dt
A bit of work
d
(mvi ) = qEi + q ( v B)i
dt
Conservation of Hamiltonian
Consider time-derivative of Hamiltonian

dH (q, p, t ) H
H
H
q+
p+
=
q
p
t
dt
H
= pq + qp +
t
Hamiltonian is
conserved if it does not
depend explicitly on t
H may or may not be total energy

If it is, this means energy conservation
Even if it isnt, H is still a constant of motion
Cyclic Coordinates
A cyclic coordinate does not appear in L
By construction, it does not appear in H either

H ( q , p, t ) = qi pi L( q , q, t )
Hamiltons equation says
H
Conjugate momentum of a
=0
p=
cyclic coordinate is conserved
q
Exactly the same as in the Lagrangian formalism
Cyclic Example
Central force problem in 2 dimensions

m 2 2 2
L = (r + r ) V (r )
2
pr = mr p = mr 2
1 2 p2
H=
pr + 2 + V (r )
r
2m
1 2 l2
=
pr + 2 + V (r )
r
2m
is cyclic
p = const = l
Hamiltons equations
pr
l2
V (r )
r=
pr = 3
m
mr
r
Cyclic variable drops off by itself
Going Relativistic
Practical approach
Purist approach
Find a Hamiltonian that works

Does it represent the total energy?
Construct covariant Hamiltonian formalism
For one particle in an EM field
Dont expect miracles
Fundamental difficulties remain the same
Practical Approach
Start from the relativistic Lagrangian that works

L = mc 2 1 2 V (x)
mvi
L
=
pi =
vi
1 2
H =h=
Did this last time
p 2 c 2 + m 2 c 4 + V ( x)
It does equal to the total energy

Hamiltons equations
H
V
pi c 2
pi
H
pi =
=
= Fi
=
=
xi =
xi
xi
pi
p 2 c 2 + m2 c 4 m
Practical Approach w/ EM Field
Consider a particle in an EM field

L = mc 2 1 2 q (x) + q ( v A)
Hamiltonian is still total energy
H = m c 2 + q
Can be easily checked
= m 2 2 v 2 c 2 + m2 c 4 + q
Difference is in the momentum pi = m vi + qAi

H = (p qA) 2 c 2 + m 2 c 4 + q
Not the usual linear momentum!
Practical Approach w/ EM Field

H = (p qA) 2 c 2 + m 2 c 4 + q
Consider H q
( H q ) 2 (p qA)2 c 2 = m2 c 4
constant
It means that ( H q , pc qAc) is a 4-vector,

and so is ( H , pc)
Similar to 4-momentum (E/c, p) of
a relativistic particle
Remember p here is not the linear momentum!
This particular Hamiltonian + canonical momentum

transforms as a 4-vector
True only for well-defined 4-potential such as EM field
Purist Approach
Covariant Lagrangian for a free particle = 12 mu u
= mu
p =
u
H=
We know that p0 is E/c
We also know that x0 is ct
p p
2m
Energy is the conjugate momentum of time
Generally true for any covariant Lagrangian
You know the corresponding relationship in QM
Purist Approach
Value of Hamiltonian is
p p mc 2
=
H=
This is constant!
2m
2
What is important is Hs dependence on p
Hamiltons equations
dx
H p
=
=
d p
m
dp
H
=
=0
d
x
Time components are

d (ct ) E
d ( E c)
=
=c
=0
d
mc
d
4-momentum
conservation
Energy definition
and conservation
Purist Approach w/ EM Field
With EM field, Lagrangian becomes

( x , u ) = 12 mu u + qu A
H=
mu u
2
p = mu + qA
( p qA )( p qA )
2m
Hamiltons equations are

( p qA ) A
H
dx H p qA dp
=
=
=
=
d
x
m
x
d p
m
A bit of work can turn them into
A A
du
= q
m
u = K
4-force
x
d
x

EM Field and Hamiltonian
In Hamiltonian formalism, EM field always modify

the canonical momentum as p A = p0 + qA
With EM field
Without EM field
A handy rule:
Hamiltonian with EM field is given by replacing p
in the field-free Hamiltonian with p qA
Often used in relativistic QM to introduce EM interaction
Summary
Constructed Hamiltonian formalism
Equivalent to Lagrangian formalism

Simpler, but twice as many, equations
Hamiltonian is conserved (unless explicitly t-dependent)
Equals to total energy (unless it isnt) (duh)
Cyclic coordinates drops out quite easily
A few new insights from relativistic Hamiltonians
Conjugate of time = energy

p qA rule for introducing EM interaction
Mechanics
Physics 151
Lecture 19
Hamiltonian Equations of Motion
(Chapter 8)
Constructed Hamiltonian formalism

H (q, p, t ) = qi pi L(q, q, t )
H
H
H
L
=
= qi
= pi
pi
t
t
qi
Equivalent to Lagrangian formalism

Simpler, but twice as many, equations
Hamiltonian is conserved (unless explicitly t-dependent)
Equals to total energy (unless it isnt)
Hamiltons Principle
Lagranges equations were derived from

t2
I Ldt = 0
t1
Hamiltons Principle
How does it work in Hamiltonian formalism?

First, lets remember how this worked before
Consider the path q = q(t) and define variation by

q (t , ) q (t ) + (t )
t2
I ( ) f (q (t , ), q (t , ), t )dt
t1
dI ( )
I
d
d =0
Hamiltons Principle
q (t , ) q (t ) + (t )
t2
I ( ) f (q (t , ), q (t , ), t )dt
t1
dI ( )
I
d
d =0
A bit of work shows

t2 f
f
d f
dI ( )
= t1 (t )dt + (t )
d =0
q
t1
dq dt q
t2
So I = 0 is equivalent to
f d f
Lagranges equations
=0
dq dt q
= 0 at t1, t2
Hamiltons Principle
Rewrite action integral using Hamiltonian

t2
t2
t1
t1
I Ldt =
( pi qi H (q, p, t ) ) dt
Now denotes variation in the phase space

i.e. both qi(t) and pi(t) are varied independently
qi (t , ) qi (t ) + i (t ) pi (t , ) pi (t ) + i (t )
Calculation of I goes exactly the same way
Just consider 2n variables qi and pi instead of qi only

f (q, q, p, p, t ) = pi qi H (q, p, t )
This can be omitted
Hamiltons Principle
I = 0 is equivalent to
f (q, q, p, t ) = pi qi H (q, p, t )
f
d f
f
d f

= 0 and
=0
dqi dt qi
dpi dt pi
H
H
qi
=0
+ pi = 0
dpi
dqi
Well, that was easy

One subtlety remains the end points
Hamiltons
equations
End Point Constraints
Variation in Lagrangian formalism required

q (t1 ) = q(t2 ) = 0
The end points are fixed in the config space
Derivation of Hamiltons equations requires

More restrictive
q (t1 ) = q (t2 ) = p (t1 ) = p(t2 ) = 0
Or does it?
t2
t2
What we really need is (t ) + (t ) = 0

q
p
t1
t1
But f does not depend
on p !
This is 0
This doesnt
f (q, q, p, t ) = pi qi H (q, p, t )
So we need only q(t1 ) = q(t2 ) = 0
have to be 0
as before
We may still want to require

q (t1 ) = q (t2 ) = p(t1 ) = p(t2 ) = 0
Keeps q and p symmetric

Adds flexibility to the definition of the action integral
You can add time-derivative of any function F(q, p, t)
t2
dF (q, p, t )
I pi qi H (q, p, t ) +
dt
t1
dt
Difference would be
t2 dF ( q , p , t )
t2
dt
=
F
(
q
(
t
),
p
(
t
),
t
)
[
]t1
=0
t1 dt

dF (q, p, t )
I pi qi H (q, p, t ) +
dt
t1
dt
dF
= qi pi qi pi
Choose for example F = qi pi
dt
t2
t2
t1
( pi qi H (q, p, t ) ) dt
This is no longer the Lagrangian
Action integral can be defined without referring to L
This allows a larger set of coordinate transformation

canonical transformation for Hamiltonian formalism
Will do this next time, but lets take a sneak peek
Canonical Transformation
Lagrangian Hamiltonian formalism meant moving

from the configuration space to the phase space
(q1 , , qn )
(q1 , , qn , p1 , , pn )
Lagrangian formalism is invariant under coordinate

transformations such as Qi = Qi (q1 , , qn , t )
Is Hamiltonian formalism invariant under similar but more
general transformations? Qi = Qi (q1 , , qn , p1 , , pn , t )
Pi = Pi (q1 , , qn , p1 , , pn , t )
Qi = Qi (q1 , , qn , p1 , , pn , t ) Pi = Pi (q1 , , qn , p1 , , pn , t )
Well, no. Its much too general, but
There is a subset of such transformations canonical

transformations that work
We will find the rules
Hamiltonian formalism is more forgiving
Goal: Find the transformation that makes the problem easiest

to solve
We may make coordinates cyclic, as we did by using
polar coordinates in the central-force problem
Principle of Least Action
Principle of Least Action is a confusing term
Action changed its meaning historically

I would rather call Hamiltons principle as
the principle of least action
Least is not strictly true Extreme would be the right word
Lets follow Goldsteins usage for today
Principle of least action is expressed as

t2
pi qi dt = 0
t1
-variation
action
What are they?
-Variation
In -variation, end points are fixed in space and time
This requires a different kind of variation
What if we allowed time to vary?

We make a physical system to move from one state to
another, but allow it to take as much time as it wants
q(t, ) does not have to be q(t, 0) at t = t1, t2
It may take extra time, e.g. q(t2 + t2, ) = q(t2, 0)
How can we define such variation?
-Variation
Vary the path in the config space qi (t , ) qi (t ) + i (t )
This time, we do not fix the end points

i.e. i (t1 ) 0, i (t2 ) 0
Whats more, we vary the range of integration
t1 t1 + t1 , t2 t2 + t2
We define the -variation by

t2
t2 +t2
t1
t1 +t1
Ldt =
t2
L( )dt L(0)dt
t1
Less restrictive form

than what we need
-Variation
In the configuration space
t2
End points of integration

move by
qi (t1 ) = qi (t1 + t1 , ) qi (t1 )
1st-order approximation
qi (t1 ) = qi (t1 )t1 + qi (t1 )
Same for t2
We will later impose
qi (t1 ) = qi (t2 ) = 0
q (t )
t1
t2 + t2
q (t ) + q (t )
t1 + t1
These points will
be the same
-Variation
t2
t2 +t2
t1
t1 +t1
Ldt =
t2
L( )dt L(0)dt
t1
Considering only 1st-order terms of t and (t)

t2
t2
t2
t1
t1
t1
Ldt = L(t2 )t2 L(t1 )t1 + L( )dt L(0)dt
Last 2 terms are -variation with free end points

t2
t2
t1
t1
Ldt =
L d L
L
qi

qi dt +
dqi dt qi
qi
t1
t2
Assume Lagranges equation
Ldt = [ Lt + pi qi ]t
t2
t1
t2
1
Ldt = [ pi qi ]t
t2
t1
t2
1
-Variation
Ldt = [ Lt + pi qi ]t
t2
t2
t1
Using
qi (t1 ) = qi (t1 )t1 + qi (t1 )
t2
t2 + t2
qi (t2 ) = qi (t2 )t2 + qi (t2 )

q (t )
Ldt = [ Lt pi qi t + pi qi ]t
t2
t2
t1
q (t ) + q (t )
= [ pi qi H t ]t
t2
1
This is a very general form

Now we examine a more restrictive case
t1
t1 + t1
Restricted Variation
We impose three conditions
Initial and final states are fixed

qi = 0 at t = t1, t2
H is conserved H does not explicitly depend on t
Varied paths are restricted so that H is constant
Last two are basically energy conservation
-variation is now
Ldt = [ pi qi H t ]t = H (t2 t1 )
t2
t1
t2
1

t2
Ldt = H (t2 t1 )
t1
From the definition of Hamiltonian

t2
t2
t2
t1
t1
t1
Ldt = ( pi qi H )dt = pi qi dt H (t2 t1 )
Comparing the two equations

t2
pi qi dt = 0
t1
A conservative system takes a path that minimizes the integral
pi qi dt
pi qi dt ?
Lets consider a simple example
Now, what is
A particle under conservative potential L =

pi qi = mv 2 = 2T
m 2
v V ( x)
2
t2
Principle of least action is equivalent to Tdt = 0

t1
The particle tries to move from point 1 to point 2 with

minimum kinetic energy time
As slowly as possible, yet spending as little time as
possible
Principle of Ideal Commuting?
For a free particle, kinetic energy T is constant

t2
Tdt = T (t2 t1 ) = 0
Principle of Least Time
t1
This also means that a free particle takes the shortest path =
straight line between point 1 and point 2
Similar to Fermats principle in optics
Light travels the fastest path between two points
Summary
Hamiltons Principle in the Hamiltonian formalism
Derivation was simple I
t2
t1
Additional end-point constraints

q (t1 ) = q (t2 ) = p(t1 ) = p(t2 ) = 0
( pi qi H (q, p, t ) ) dt = 0
Not strictly needed, but adds flexibility to the definition

of the action integral
t2
pi qi dt = 0
t1
-derivative allows change of time

t2
For simple systems, equivalent to Tdt = 0

t1
Mechanics
Physics 151
Lecture 20
Canonical Transformations
(Chapter 9)
Hamiltons Principle in the Hamiltonian formalism
Derivation was simple I
t2
t1
( pi qi H (q, p, t ) ) dt = 0
Additional end-point constraints

q (t1 ) = q(t2 ) = p(t1 ) = p(t2 ) = 0
Not strictly needed, but adds flexibility to the definition
of the action integral
This connects to: Canonical Transformations
t2
Principle of Least Action t pi qi dt = 0

1
Got into
this a bit
Goal: To find transformations

Qi = Qi (q1 , , qn , p1 , , pn , t ) Pi = Pi (q1 , , qn , p1 , , pn , t )
that satisfy Hamiltons equation of motion

H
qi =
dpi
H
pi =
dqi
K
Pi =
dQi
K
Qi =
dPi
K is the transformed Hamiltonian K = K (Q, P, t )
Hamiltons principle requires
t2
t1
( pi qi H (q, p, t ) ) dt = 0
and
t2
t1
( PQ K (Q, P, t ) ) dt = 0
i
General Transformation
t2
t1
and
t1
( PQ K (Q, P, t ) ) dt = 0
i
Two types of transformations are possible
( pi qi H (q, p, t ) ) dt = 0
t2
PQ
i i K = ( pi qi H )
Scale transformation
dF
PQ
= pi qi H
Canonical transformation
i i K +
dt
Both satisfy Hamiltons principle
Combined, we find
dF
PQ
= ( pi qi H )
i i K +
dt
Extended Canonical
transformation
Scale Transformation
We can always change the scale of (or unit we use to

measure) coordinates and momenta
Pi = pi Qi = qi
To satisfy Hamiltons principle, we can define

K ( P, Q, t ) = H ( p, q, t )
PQ
i i K = ( pi qi H )
Scale transformation
This is trivial
We now concentrate on Canonical transformations
dF
= pi qi H
dt
Hamiltons principle
PQ
i i K +
t2
t1
( PQ K ) dt =
i
t2
t1
dF
t2
pi qi H
dt = [ F ]t1 = 0
dt
Satisfied if p = q = P = Q = 0 at t1 and t2
F can be any function of pi, qi, Pi, Qi and t
It defines a canonical transformation

Call it the generating function of the transformation
or generator
Simple Example [1]
dF
PQ
= pi qi H
i i K +
dt
Try a generating function: F = qi Pi Qi Pi
Canonical transformation generated by F is

dF
PQ
= K + (qi Qi ) Pi + Pq
i i K +
i i = pi qi H
dt
Qi = qi Pi = pi
Identity transformation
K=H
OK, that was too simple

Lets push this one step further
Simple Example [2]
dF
PQ
= pi qi H
i i K +
dt
Lets try this one: F = fi (q1 , , qn , t ) Pi Qi Pi

fi
are arbitrary functions of q1qn and t

fi
fi
dF
PQ
= K + ( f i Qi ) Pi + Pi
qj +
Pi = pi qi H
i i K +
dt
q j
t
Qi = fi (q1 , , qn , t )
pi =
f j
qi
Pj
All point transformations of

generalized coordinates are covered
Must invert these n equations to get Pi
fi
K=H+
Pi
t
We can do all what we could do before
Arbitrarity
Generating function F a canonical transformation
Opposite mapping is not unique

There are many possible Fs for each transformation
e.g. add an arbitrary function of time g(t) to F
dF
dF dg (t )
Does not affect
PQ
K
+
PQ
K
+
+
i i
i i
the action integral
dt
dt
dt
dg (t )
KK+
dt
Just modifies the Hamiltonian

without affecting physics
F is arbitrary up to any function of time only
So is the Hamiltonian
Finding the Generator
PQ
i i K +
dF
= pi qi H
dt
Lets look for a generating function
Suppose K (Q, P, t ) = H (q, p, t ) for simplicity
dF
= pi qi PQ
i i
dt
Easiest way to satisfy this would be
F
F
= pi
= Pi
F = F ( q, Q )
qi
Qi
Trivial example: F (q, Q) = qi Qi

pi = Qi Pi = qi
In the Hamiltonian formalism,

you can freely swap the
coordinates and the momenta
Type-1 Generator
PQ
i i K +
dF
= pi qi H
dt
F = F (q, Q) is not very general

It does not allow t-dependent transformation
Fix this by extending to F = F1 ( q, Q, t )
Call it Type-1
pi =
F1 (q, Q, t )
qi
Pi =
F1 (q, Q, t )
Qi
This affects the Hamiltonian

F1
F1
dF F1
=
qi +
Qi +
= pi qi PQ
i i +K H
dt qi
Qi
t
F1
K=H+
t
Harmonic Oscillator
Consider a 1-dimensional harmonic oscillator

p 2 kq 2
1
k
2
2 2 2
2
H ( q, p ) =
+
=
p +m q )

(
2m 2
2m
m
Sum of squares Can we make them sine and cosine?
f ( P)
Suppose p = f ( P ) cos Q q =
sin Q
m
2
f
(
P
)
{
}
Q is cyclic P is constant
K=H =
2m
Trick is to find f(P) so that the transformation is canonical

How?
Harmonic Oscillator
Lets try a Type-1 generator

F1 (q, Q, t )
F1
p=
q
F1
P=
Q
Express p as a function of q and Q

f ( P)
sin Q
p = f ( P) cos Q q =
m
Integrate with q
p = m q cot Q
m q2 q2
cot Q
F1 =
2
F1
m q 2
P=
=
Q 2sin 2 Q
We are getting somewhere
Harmonic Oscillator
F
p = 1 = m q cot Q
q
We need to turn H(q, p) into K(Q, P)

Solve the above equations for q and p
2P
q=
sin Q
m
F1
m q 2
P=
=
Q 2sin 2 Q
p = 2 Pm cos Q
Now work out the Hamiltonian

1
K=H =
p 2 + m 2 2 q 2 ) = P
(
2m
Things dont get much simpler than this
Harmonic Oscillator
K = P = E
Solving the problem is trivial
P = const =
K
= Q = t +
Q=
P
p = 2 Pm cos Q = 2mE cos( t + )

Finally
2P
2E
q=
sin Q =
sin( t + )
2
m
m
Phase Space
Oscillator moves in the p-q and P-Q phase spaces

p
2mE
2E
m 2
One cycle draws the same area
2 E
in both spaces
The area swept by a cyclic system in the phase space is

invariant
Will come back to this in Lecture 23
Other Types of Generators
Type-1 generator F = F1 (q, Q, t ) is still not so general
We need generating functions of different set of

Just try to find a generator for Qi = qi Pi = pi
In fact, we may have 4 basic types of them

F1 (q, Q, t ) F2 (q, P, t ) F3 ( p, Q, t ) F4 ( p, P, t )
We can derive them using the now-familiar rule
i.e. we can add any dF/dt inside the action integral
Type-2 Generator
In the last lecture, I used F = qi pi to convert
t2
t1
( pi qi H (q, p, t ) ) dt = 0
t2
t1
( pi qi H (q, p, t ) ) dt = 0
Switch the definition of canonical transformations

dF
dF
PQ
= pi qi H
Pi Qi K +
= pi qi H
i i K +
dt
dt
dF
= pi qi + Qi Pi + K H
dt
To satisfy this
F2
= pi
F = F2 (q, P, t )
qi
F2
= Qi
Pi
F2
K=H+
t
Type-2 Generator
If we go back to the original definition of generating

dF
= pi qi H
dt
F2
F2
F2
=
p
=
=
+
Q
K
H
F = F2 (q, P, t ) Qi Pi
i
i
Pi
qi
t
function PQ
i i K +
Trivial case: F2 = qi Pi
pi = Pi Qi = qi
Identity transformation
We push the same idea to define the other 2 types
Four Basic Generators

Generator
Derivatives
Trivial Case
Qi = pi
F1
F1
F1 (q, Q, t )
F1 = qi Qi
pi =
Pi =
Pi = qi
Qi
qi
Qi = qi
F2
F2
F2 (q, P, t ) Qi Pi
F2 = qi Pi
pi =
Qi =
Pi = pi
Pi
qi
Qi = qi
F3
F3
F3 ( p, Q, t ) + qi pi
F3 = pi Qi
qi =
Pi =
Pi = pi
pi
Qi
Qi = pi
F4
F4
F4 ( p, P, t ) + qi pi Qi Pi qi =
F4 = pi Pi
Qi =
Pi = qi
pi
Pi
The 4 types of generators are almost equivalent
It may look as if F1 is special, but it isnt
dF1
PQ
i i K +
dt
dF2
PQ
i i K +
dt
dF3
PQ
i i K +
dt
dF4
PQ
i i K +
dt
= pi qi H
= pi qi H
= pi qi H
= pi qi H
There is no reason to consider

any of these 4 definitions to
be more fundamental than the
others
We arbitrarily chose the first
form (which happens to be the
Lagrangian form) to write the
generating functions in the
table
Some canonical transformations cannot be generated

by all 4 types
This does not present a fundamental problem
e.g. identity transf. is generated only by F2 or F3

One can always swap coordinate and momentum
Qi = pi Pi = qi
One can always change sign by scale transformation
Qi = qi Pi = pi
These transformations make the 4 types practically

equivalent
One More Example

p2
1
H
=
+
1-dim system with
2 2q 2
Try P = pq
Lets use Type-2
F2
F2
=p
F = F2 (q, P, t )
=Q
q
P
P
p=
q
Step 1: Express p with q and P

Step 2: Integrate with q to get
F2 = P log q
assuming q > 0
Step 3: Differentiate to get Q = log q

q = eQ
Now we have a canonical transformation
V=
1
2q 2
One More Example

P
= Pe Q
q
Now rewrite the Hamiltonian
p2
1
P 2 + 1 2Q
+ 2 =
H=
e =E
2 2q
2
F2 = P log q q = eQ p =
constant
Equation of motion: P = ( P 2 + 1)e 2Q = 2 E

P = 2 Et + C
q=e =
Q
P2 + 1
C2 +1
2
= 2 Et + 2Ct +
2E
2E
Summary
Canonical transformations
They are all practically equivalent
Used it to simplify a harmonic oscillator
dF
= pi qi H
dt
Hamiltonian formalism is
invariant under canonical + scale transformations
Generating functions define canonical transformations
Four basic types of generating functions
PQ
i i K +
Invariance of phase space area
Mechanics
Physics 151
Lecture 21
(Chapter 9)
Canonical transformations
Hamiltonian formalism is
invariant under canonical + scale transformations
Generating functions define canonical transformations
Four basic types of generating functions
K + dF = p q H
PQ
i i
i i
dt
They are all practically equivalent
Used it to simplify a harmonic oscillator
Invariance of phase space

Generator
Derivatives
Trivial Case
Qi = pi
F1
F1
F1 (q, Q, t )
F1 = qi Qi
pi =
Pi =
Pi = qi
Qi
qi
Qi = qi
F2
F2
F2 (q, P, t ) Qi Pi
F2 = qi Pi
pi =
Qi =
Pi = pi
Pi
qi
Qi = qi
F3
F3
F3 ( p, Q, t ) + qi pi
F3 = pi Qi
qi =
Pi =
Pi = pi
Qi
pi
Qi = pi
F4
F4
F4 ( p, P, t ) + qi pi Qi Pi qi =
F4 = pi Pi
Qi =
Pi = qi
Pi
pi
Goals for Today
Dig deeper into Canonical Transformations
Infinitesimal Canonical Transformation

Very small changes in q and p
Define generator G for an ICT
Direct Conditions for Canonical Transformation
Necessary-and-sufficient conditions for any CT
Poisson Bracket
Invariant of any Canonical Transformation
Connect to Infinitesimal Canonical Transformation
Infinitesimal CT
Consider a CT in which q, p are changed by small

(infinitesimal) amounts
Qi = qi + qi
Pi = pi + pi
Infinitesimal Canonical
Transformation (ICT)
ICT is close to identity transf.

Generating function should be F2 (q, P, t ) = qi Pi + G (q, P, t )
Identity CT generator
Look at the
generator table
Since is
infinitesimal
Small
F2
F2
G
G
pi =
Qi =
= Pi +
= qi +
qi
qi
Pi
Pi
G
G
G
G
pi =

qi =
qi
Qi
Pi
pi
Generator of ICT
An ICT is generated by F2 (q, P, t ) = qi Pi + G (q, P, t )

Qi = qi +
G
Pi
Pi = pi
G
qi
G is called (inaccurately) the generator of the ICT

Since the CT is infinitesimal, G may be expressed in terms
of q or Q, p or P, interchangeably
G
G
For example: G = G (q, p, t ) Qi = qi +
Pi = pi
qi
pi
Hamiltonian
Consider G = H (q, p, t )
H
H
= qi pi =
qi =
= p i
pi
qi
What does look like? Infinitesimal time t

qi = qi t
pi = p i t
Hamiltonian is the generator of infinitesimal time

transformation
In QM, you learn that Hamiltonian is the operator that

represents advance of time
Direct Conditions
Consider a restricted Canonical Transformation
Generator has no t dependence

F
=0
K (Q, P) = H (q, p )
t
Hamiltonian
is unchanged
Q and P depends only on q and p

Qi = Qi (q, p ) Pi = Pi (q, p)
Qi
Qi
Qi H Qi H

Qi =
qj +
pj =
q j
p j
q j p j p j q j
Pi
Pi
Pi H Pi H

Pi =
q j +
p j =
q j
p j
q j p j p j q j
Hamiltons
equations
Direct Conditions
On the other hand, Hamiltons eqns say
Qi H Qi H
H H q j H p j

Qi =
Qi =
=
+
q j p j p j q j
Pi q j Pi p j Pi
H
H q j H p j

Pi =
=
Qi
q j Qi p j Qi
Direct
Conditions
for a Canonical
Transformation
Pi H Pi H

Pi =
q j p j p j q j
Qi
p j
q
P
j q , p i Q , P
Qi
q j
p
P
i Q , P
j q, p
Pi
q j
Pi
p j
p j
=
Qi Q , P
q , p
q j
=
q , p Qi Q , P
Direct Conditions
Qi
q j
Pi Q , P
p j q , p
Pi
q j
Pi
p j
p j
=
i Q , P
q , p
q j
=
q , p i Q , P
Direct Conditions are necessary and sufficient for a

time-independent transformation to be canonical
Qi
p j
q j q , p Pi Q , P
You can use them to test a CT
In fact, this applies to all Canonical Transformations
But the proof on the last slide doesnt work
G
G
qi =
Pi
pi
Infinitesimal CT
pi =
Does an ICT satisfy the DCs?
Qi (qi + qi )
2G
=
= ij +
q j
q j
Pi q j
p j
Qi (qi + qi )
2G
=
=
p j
p j
Pi p j
q j
Pi ( pi + pi )
2G
=
=
q j
q j
Qi q j
p j
Pi ( pi + pi )
2G
=
= ij
p j
p j
Qi p j
Pi
Pi
=
=
Qi
q j
Qi
Yes!
=
=
( Pj p j )
Pi
(Q j q j )
Pi
( Pj p j )
Qi
(Q j q j )
Qi
G
G

qi
Qi
2G
= ij +
Pi q j
2G
=
Pi p j
2G
=
Qi q j
2G
= ij
Qi p j
Successive CTs
Two successive CTs make a CT

dF1

PQ
= pi qi H
i i K +
dt
dF2
K

= PQ
Yi X i M +
i i
dt
d ( F1 + F2 )

= pi qi K
Yi X i M +
dt
True for unrestricted CTs
Direct Conditions can also be chained, e.g.,

Qi
p j
q j q , p Pi Q , P
X i
Q j
X i
p j
q
Y
j q , p i X ,Y
Pj
=
Q , P Yi X ,Y
Easy to prove
Unrestricted CT
Now we consider a general, time-dependent CT

Qi = Qi (q, p, t )
Pi = Pi (q, p, t )
Lets do it in two steps

q, p
Q(q, p, t0 ), P(q, p, t0 )
Time-independent CT
F
K=H+
t
Q(q, p, t ), P(q, p, t )
Time-only CT
Fixed time
First step is t-independent Satisfies the DCs
We must show that the second step satisfies the DCs
Unrestricted CT
Concentrate on a time-only CT Q(t0 ), P(t0 )
Q(t ), P(t )
Break t t0 into pieces of infinitesimal time dt

Q(t0 ), P(t0 )
Q(t0 + dt ), P(t0 + dt )
Q(t ), P(t )
Each step is an ICT Satisfies Direct Conditions

Integrating gives us what we needed
All Canonical Transformations satisfies the

Direct Conditions, and vice versa
The proof worked because a time-only CT is a continuous

transformation, parameterized by t
Poisson Bracket
For u and v expressed in terms of q and p

u v u v
Poisson Bracket
[u, v ]q , p
qi pi pi qi
This weird construction has many useful features

If you know QM, this is analogous to the commutator
1
1
[u, v ] (uv vu ) for two operators u and v
i=
i=
Lets start with a few basic rules
Poisson Bracket Identities
For quantities u, v, w and

constants a, b
[u, u ] = 0
[u, v ]q , p
u v u v
qi pi pi qi
[u , v] = [v, u ]
[au + bv, w] = a[u , w] + b[v, w]
[uv, w] = [u , w]v + u[v, w]
All easy to prove

Jacobis Identity
[u ,[v, w]] + [v,[ w, u ]] + [ w,[u, v]] = 0

This one is worth trying.
See Goldstein if you are lost
Fundamental Poisson Brackets
Consider PBs of q and p themselves

q j qk q j qk
[q j , qk ] =
=0
qi pi pi qi
[ p j , pk ] = 0
q j pk q j pk
= jk
[q j , pk ] =
qi pi pi qi
[ p j , qk ] = jk
Called the Fundamental Poisson Brackets
Now we consider a Canonical Transformation

q, p Q, P
What happens to the Fundamental PB?
Fundamental PB and CT
[Q j , Qk ]q , p
Q j Qk Q j Qk
Q j qi Q j pi
Q j
=
=
=0
qi pi pi qi
qi Pk pi Pk
Pk
[ Pj , Pk ]q , p
Pj Pk Pj Pk Pj qi Pj pi
Pj
=
=
+
=
=0
qi pi pi qi qi Qk pi Qk Qk
[Q j , Pk ]q , p
Q j Pk Q j Pk Q j qi Q j pi Q j
=
=
+
=
= jk
qi pi pi qi qi Qk pi Qk Qk
[ Pj , Qk ]q , p = [Qk , Pj ] = jk
Used Direct Conditions here
Fundamental Poisson Brackets are invariant under CT
Poisson Bracket and CT
What happens to a Poisson Bracket under CT?
For a time-independent CT
[u, v ]Q, P
u v u v
Qi Pi Pi Qi
u q j u p j v qk v pk u q j u p j v qk v pk
=
+
+
q P + p P q Q + p Q
q
Q
p
Q
q
P
p
P
j
i
k
i
k
i
j
i
k
i
k
i
j i
j i
=
=
u v
q j qk
u v
q j pk
[q j , qk ]Q , P +
jk
= [u , v ]q , p
u v
p j qk
u v
q j pk
[q j , pk ]Q , P +
u v
p j qk
[ p j , qk ]Q , P +
u v
p j pk
[ p j , pk ]Q , P
jk
Poisson Brackets are invariant under CT
Invariance of Poisson Bracket
Poisson Brackets are canonical invariants
True for any Canonical Transformations

Goldstein shows this using simplectic approach
We dont have to specify q, p in each PB
[u, v ]q , p
[u , v ]
good enough
ICT and Poisson Bracket
Infinitesimal CT can be expressed neatly with a PB
G
For a generator G, Qi = qi +
pi
On the other hand
G
Pi = pi
qi
qi G qi G
G
=
= qi
[qi , G ] =
q p p q
pi
j
j
j j
pi G pi G
G
=
= pi
[ pi , G ] =
q p p q
qi
j
j
j j
We can generalize further
ICT and Poisson Bracket
For an arbitrary function u(q,p,t), the ICT does

ICT
u
u + u = u +
u
u
u
qi +
pi + t
qi
pi
t
u G u G u
=u+
+ t
qi pi pi qi t
u
= u + [u, G ] + t
t
u
That is u = [u , G ] + t
t
Infinitesimal Time Transf.
Hamiltonian generates infinitesimal time transf.
Applying the Poisson Bracket rule

du
u
u
u = t[u, H ] + t
= [u , H ] +
t
t
dt
Have you seen this in QM?
u
=0
If u is a constant of motion, [u , H ] +
t
That is,
u
[H , u] =
t
u is a constant of motion
Infinitesimal Time Transf.
If u does not depend explicitly on time,

du
u
= [u , H ] +
= [u , H ]
t
dt
Try this on q and p

q H qi H H
qi = [qi , H ] = i
=
q j p j p j q j pi
pi H pi H
H
p i = [ pi , H ] =
=
q j p j p j q j
qi
Hamiltons
equations!
Summary
Direct Conditions
Necessary and sufficient

for Canonical Transf.
Infinitesimal CT
Poisson Bracket
Qi
q j
p j
=
Q ,P
q, p
Qi
p j
q j
=
Q ,P
q, p
P
i
q j
p j
=
q, p
Q ,P
Pi
p j
q j
=
Q ,P
q, p
[u , v ]
u v u v
qi pi pi qi
Canonical invariant
Fundamental PB [qi , q j ] = [ pi , p j ] = 0 [qi , p j ] = [ pi , q j ] = ij
u
ICT expressed by u = [u , G ] + t
t
Infinitesimal time transf. generated by Hamiltonian
Mechanics
Physics 151
Lecture 22
(Chapter 9)
Direct Conditions
Necessary and sufficient

for Canonical Transf.
Infinitesimal CT
Poisson Bracket
Qi
q j
p j
=
Q ,P
q, p
Qi
p j
q j
=
Q ,P
q, p
P
i
q j
p j
=
q, p
Q ,P
Pi
p j
q j
=
Q ,P
q, p
[u , v ]
u v u v
qi pi pi qi
Canonical invariant
Fundamental PB [qi , q j ] = [ pi , p j ] = 0 [qi , p j ] = [ pi , q j ] = ij
u
ICT expressed by u = [u , G ] + t
t
Infinitesimal time transf. generated by Hamiltonian
Hamiltons equations
Two Points of View
Canonical Transformation allows one system to be

described by multiple sets of coordinates/momenta
Same physical system is expressed in different phase spaces
p
P
This is the static view The system itself is unaffected

Is there a dynamic view?
Dynamic View of CT
A system evolves with time q(t0 ), p(t0 )
q (t ), p (t )
At any moment, q and p satisfy Hamiltons equations
The time-evolution must be a Canonical Transformation!
p
This movement
is a CT
Static View = Coordinate system is changing

Dynamic View = Physical system is moving
Infinitesimal Time CT
q (t + dt ), p (t + dt )
We know that the generator = Hamiltonian
u
du = dt[u , H ] + dt
q = [q, H ] p = [ p, H ]
t
Infinitesimal CT q(t ), p(t )
Hamiltonian
Hamiltonian isis the
the generator
generator of
of the
the systems
systems
motion
motion with
with time
time
Integrating it with time should give us the finite CT that

turns the initial conditions q(t0), p(t0) into the configuration
q(t), p(t) of the system at arbitrary time
Thats a new definition of solving the problem
Static vs. Dynamic
Two ways of looking at the same thing
System is moving in a fixed phase space

Hamiltons equations Integrate to get q(t), p(t)
System is fixed and the phase space is transforming
ICT given by the PB Integrate to get CT for finite t
Equations are identical
Youll find yourself integrating exactly the same equations

Did we gain anything?
Conservation
u
t
Consider an ICT generated by G u = [u , G ] +
t
Suppose G is conserved and
[G, H ] = 0
has no explicit t-dependence
How is H (without t-dependence) changed by the ICT?
H
H = [H , G] +
t = 0
t
If an ICT does not affect

Hamiltonian, its generator
is conserved
A transformation that does not affect H

Symmetry of the system
Generator of the transformation is conserved
Simplest example:
What is the ICT generated by momentum pi?

q j = [q j , pi ] = ij p j = [ p j , pi ] = 0
Thats a shift in qi by spatial translation
If Hamiltonian is unchanged by such shift, then [ H , pi ] = 0
Momentum pi is conserved
This is not restricted to linear momentum
Hamiltonian is
unchanged by
a shift of a
coordinate q
The generator
of the ICT is
the conjugate
momentum p
[ H , p] = 0
p is
conserved
Angular Momentum
Lets consider a specific case: Angular momentum
Pick x-y-z system with z being the axis of rotation

n particles positions given by ( xi , yi , zi )
Rotate all particles CCW around z axis by d
xi = xi yi d
yi = yi + xi d
( x, y) d
Momenta are rotated as well
pix = pix piy d piy = piy + pix d
( x, y )
Generator is G = xi piy yi pix
G
= yi d
d [ xi , G ] = d
pix
G
= piy d
d [ pix , G ] = d
xi
etc.
Angular Momentum
The generator G = xi piy yi pix is obviously Lz = (ri pi ) z
i.e. the z-component of the total momentum

Generator for rotation about an axis given by a unit vector n
should be
G = L n
We now know generators of 3 important ICTs
Hamiltonian generates displacement in time

Linear momentum generates displacement in space
Angular momentum generates rotation in space
Integrating ICT
I said we can integrate ICT to get finite CT
How do we integrate u = [u , G ] ?
First, lets rewrite it as du = d [u, G ]
We want the solution u() as a function

of , with the initial condition u(0) = u0
Taylor expand u() from = 0
du
= [u , G ]
d
2 d 2u
3 d 3u
du
u ( ) = u0 +
+
+
+"
2
3
d 0 2! d 0 3! d 0
This is [u,G]0
What can I do with these?
Integrating ICT
du
d
= [u , G ] is true for any u, we can say
= [, G ]
Since
d
d
Now apply this operator repeatedly
d ju
d 2u
d
=
[u , G ] = [[u , G ], G ]
= ["[[u , G ], G ]," , G ]
2
j
d
d
d
Going back to the Taylor expansion,
2 d 2u
3 d 3u
du
u ( ) = u0 +
+
+
+"
2
3
d 0 2! d 0 3! d 0
= u0 + [u, G ]0 +
[[u , G ], G ]0 +
[[[u , G ], G ], G ]0 + "
2!
3!
Now we have a formal solution But does it work?
Rotation CT
Lets integrate the ICT for rotation around z
Let me forget the particle index i G = xp y ypx

Parameter is in this case
Lets see how x changes with
x( ) = x0 + [ x, G ]0 +
[[ x, G ], G ]0 +
[[[ x, G ], G ], G ]0 + "
2!
3!
Evaluate the Poisson Brackets
[ x, G ] = y [[ x, G ], G ] = x [[[ x, G ], G ], G ] = y
[[[[ x, G ], G ], G ], G ] = x
Where does this lead us?
Repeats after this
Rotation CT
x( ) = x0 + [ x, G ]0 +
= x0 y0
2!
[[ x, G ], G ]0 +
x0 +
y0 +
3
3!
[[[ x, G ], G ], G ]0 + "
x0 "
3!
4!
2 4
3 5
= x0 1 + " y0 + "
2! 4!
3! 5!
= x0 cos y0 sin
2!
Similarly
y ( ) = y0 + [ y, G ]0 +
2
2!
= y0 cos + x0 sin
[[ y, G ], G ]0 +
3
3!
[[[ y, G ], G ], G ]0 + "
Free Fall
An object is falling under gravity
p2
Hamiltonian is H =
+ mgz
2m
Integrate the time ICT
t2
t3
z (t ) = z0 + t[ z , H ]0 + [[ z , H ], H ]0 + [[[ z , H ], H ], H ]0 + "
2!
3!
p
[[ z , H ], H ] = g [[[ z , H ], H ], H ] = 0
[ z, H ] =
m
z (t ) = z0 +
p0
g
t t2
m
2
Its easier than it looked
ICT for rotation is generated by G = L n
Weve studied infinitesimal rotation in Lecture 8

Infinitesimal rotation of d about n moves a vector r as
dr = nd r
Compare the two expressions
dr = d [r, L n] = nd r
[r, L n] = n r
Equation [r, L n] = n r holds for any r that rotates

together with the system
Several useful rules can be derived from this
Scalar Products
[r, L n] = n r
Consider a scalar product a b of two vectors
Try to rotate it [a b, L n] = a [b, L n] + b [a, L n]

= a (n b) + b (n a)
= a (n b) + a (b n)
=0
Obvious: scalar product doesnt change by rotation
Also obvious: length of any vector is conserved
Angular Momentum
Try with L itself
[ L, L n ] = n L
x-y-z components are

[ Lx , Lx ] = 0
[ Lx , Ly ] = Lz
[ Lx , Lz ] = Ly
[ Ly , Lx ] = Lz [ Ly , Ly ] = 0
[ Ly , Lz ] = Lx
[ Lz , Lx ] = Ly
These relationships are well-known in QM

They tell us two rather interesting things
[ Lz , Ly ] = Lx [ Lz , Lz ] = 0
[ Li , L j ] = ijk Lk
Angular Momentum
Imagine two conserved quantities A and B

[ A, H ] = [ B, H ] = 0
How does [A,B] change with time?

[[ A, B], H ] = [[ B, H ], A] [[ H , A], B] = 0
Jacobis identity
Poisson bracket of two conserved quantities is conserved
Now consider [ Li , L j ] = ijk Lk
If 2 components of L are conserved, the 3rd component must

Total vector L is conserved
Angular Momentum
Remember the Fundamental Poisson Brackets?

[qi , q j ] = [ pi , p j ] = 0
[qi , p j ] = [ pi , q j ] = ij
PB of two canonical momenta is 0
Now we know [ Li , L j ] = ijk Lk

Poisson brackets between Lx, Ly, Lz are non-zero
Only 1 of the 3 components of the angular momentum

can be a canonical momentum
On the other hand, [ L2 , Li ] = 0 , so |L| may be a canonical

momentum
QM: You may measure |L| and, e.g., Lz simultaneously, but
not Lx and Ly, etc.
Phase Volume
Static view: CT moves a point in one phase space to a point

in another phase space
Dynamic view: CT moves a point in one phase space to
another point in the same space
If you consider a set of points, CT moves a volume to

anther volume, e.g.
P
dp
How does the area change?
dq
q
Phase Volume
Easy to calculate the Jacobian for 1-dimension
Q q Q p
dQdP = M dqdp where M =
Q P P Q
M =
= [Q, P] = 1
dQdP = dqdp
q p q p
i.e., volume in 1-dim. phase space is invariant
This is true for n-dimensions
Goldstein proves it using simplectic approach
Volume in Phase Space is a Canonical Invariant
Harmonic Oscillator
Weve seen it in the oscillator example (Lecture 21)

p
2mE
2E
m 2
One cycle draws the same area

Thats static view
2 E
in both spaces
Dynamic View
Consider many particles moving independently
e.g., ideal gas molecules in a box

They obey the same EoM independently
Can be represented by multiple points
in one phase space
They move with time CT
p
Time
Ideal Gas Dynamics
Imagine ideal gas in a cylinder with movable piston
Each molecule has its own position

and momentum They fill up a
certain volume in the phase space
What happens when we compress it?
Extra momenta
Gas gets hotter!
Compress
slowly
Liouvilles Theorem
The phase volume occupied by a group of particles

(ensemble in stat. mech.) is conserved
Thus the density in phase space remains constant with time

Known as Liouvilles theorem
Theoretical basis of the 2nd law of thermodynamics
This holds true when there are large enough number of

particles so that the distribution may be considered
continuous
More about this in Physics 181
Summary
Introduced dynamic view of Canonical Transf.
Hamiltonian is the generator of the motion with time

Symmetry of the system Hamiltonian unaffected by the
generator Generator is conserved
How to integrate infinitesimal transformations

u ( ) = u0 + [u, G ]0 +
[[[u , G ], G ], G ]0 + "
3!
Discussed infinitesimal rotation [r, L n] = n r
Angular momentum QM
2!
[[u, G ], G ]0 +
Invariance of the phase volume
Liouvilles theorem Stat. Mech.
Mechanics
Physics 151
Lecture 23
Continuous Systems and Fields
(Chapter 13)
Where Are We Now?
Weve finished all the essentials
Last two lectures: Classical Field Theory
Final will cover Lectures 1 through 22

Start with wave equations, similar to Physics 15c
Do it with Lagrangian, and maybe with Hamiltonian
Go into relativistic field theory
Not enough time to discuss everything
Lets see how much we can do

And take it easy!
Longitudinal Mechanical Waves
An infinite elastic rod is vibrating longitudinally

dx
x
Model this with a chain of masses and springs Did this in 15c
x
m k
i-th masss position is i relative to equilibrium
1
T = mi2
i 2
1
V = k (i +1 i ) 2
i 2
Lets build the

Lagrangian
Lagrangian
1
2
2
Lagrangian is L = mi k (i +1 i )
i 2
Rearrange
a little
m/x is the linear density (mass/unit length)

kx is the elastic modulus K (force/fractional elongation)
1 m 2
i +1 i
= i k x
x
x
i 2
x
Think about Hookes law Its not Youngs modulus

L
F = kL = K
How much the spring is stretched
L
relative to its natural length
and K remain constant as we shrink x 0
Continuous Limit
1 2
i +1 i
Now we have L = i K
2
x
Re-label i with the equilibrium position x i ( x)
Shrink!
1
( x + x) ( x)
2
L = ( x) K
x
x

i 2
1
d
x 0
2
K
dx
2
dx
Lagrangian per unit length
Lagrangian Density
We can write the Lagrangian as

1 d
d
L =
K
2 dt
dx
dx Ldx
L is the Lagrangian density in 1-dimension

2
We may generally extend this to 3-dimensions
2
2
1 d
d
L = L dxdydz where L =
Y

2 dt
dx
is the volume density /A (A is the rods cross section)
Y is Youngs modulus K/A
Lagranges Equations
2
1
i +1 i
2
First, start from L = i K
x
x
i 2
Do the usual
Lagranges equations
d L L
K i +1 i K i i 1
=
+
x = 0
dt i i
x x x x
Shrink
x
d 2
K 2 = 0
dx
Thats wave equation with velocity v =
We want to get this from the continuous Lagrangian
Lagranges Equations
L d L
In the discrete case, we had

=0
i dt i
i became (x)
L
d L
Simple analogy gives

=0
( x) dt ( x)
But this doesnt work
We must go back to Hamiltons Principle

2
I = Ldt =
1
L dxdt = 0
for each i
Hamiltons Principle
1 d
d
Our Lagrangian density is L =
K
2 dt
dx
Lets get general
d d
L may depend on L = L , dx , dt , x, t
2
We need the path of and its variation

( x, t ; ) = ( x, t ;0) + ( x, t )
Nominal path
Variation
Set variation to zero at the boundaries

( x, t1 ) = ( x, t2 ) = ( x1 , t ) = ( x2 , t ) = 0
Initial
Will make 0
Final
OK, lets work
Edges
Dont really
matter for the
infinite rod
Hamiltons Principle
dI
d
=
d d
=
t2
t1
t2
x2
t1
x1
x2
x1
L , ddx , ddt , x, t dxdt
L d L d ddx L d ddt
+ d
+ d
dxdt
d dx d dt d
L d L d L d
dxdt
=
d d
t1 x1
dx dx dt dt d
Hamiltons Principle gives
L d L d L
t2 x2
dI
d d ( x, t )dxdt = 0
= t1 x1
d =0
dx dx dt dt
t2
x2
= 0!
Lagranges Equation
Lagranges equation for the 1-dim problem is

d L d L L
=0
d + d
dt dt dx dx
2
2
1 d
d
Lets try it with L =
K

2 dt
dx
d d d d
d 2
d 2
K
= 2 K 2 =0
dt dt dx dx
dt
dx
Yes, the right

wave equation
3-D Version
Easy to guess how it should look like in 3-dim.
L = L , ddx , ddy , ddz , ddt , x, y, z , t
I =
t2
t1
x2
y2
z2
x1
y1
z1
d L
d
dt dt
L , ddx , ddy , ddz , ddt , x, y, z , t dxdydzdt
d L d L d L L
=0
+ d + d + d
dx dx dy dy dz dz
Symmetric between time and space

Hope for relativistic formalism
Will look into this in the next lecture
Multi-Component Field
I defined as the displacement along x axis
General 3-dim. vibration may

be in any direction
= ( x , y , z )
We are now dealing with 3

functions of space and time
x , ddxx , ddyx , ddzx , ddtx ,
d y d y d y d y
y , dx , dy , dz , dt ,
L = L
d z d z d z d z
z , dx , dy , dz , dt ,
x, y , z , t
This is getting really tedious
Shorthand Notation
Lets use indices (0,1,2,3) instead of t, x, y, z
Similar to what we did in relativity

di d 2i
We need quantities like i
dx dx dx
Lets get lazy
d
d 2
d
,
,
and ,
etc.
dx
dx
dx dx
We can write, e.g.

L = L ( , , , x )
d
dx
L
=0

Conservation Laws
Lets try what we did with the energy function
Consider the total derivative of the Lagrangian density

d L L
L
L
, +
, +
=
L ( ,, , x )
,
x
dx
d
Using Lagranges equations:
dx
dL
d L
=
dx dx ,
d
=
dx
L
=0

L
L
, +
, +
,
x
L
L
, +
,
x
This is
d,
dx
Stress-Energy Tensor
d
We got
dx
L
, L
L
=
x
Stress-energy tensor
NB: T is not a tensor in the relativistic sense
Suppose L does not depend explicitly on x
For = 1, 2, 3, that means no external force

For = 0, that means no source/sink of energy
dT
dx
=0
What does this conservation

condition mean?
Free
field
Divergence of S-E Tensor
The condition
dT
dT 0
dT
dx
dTi
= 0 has a form of divergence

dT 0
+ T = 0
dx
dt
dxi
dt
Integrate over a fixed volume V and use Gausss Law
d
What escapes
=
T
=
S
T
dV
dV
d
0

from the surface
dt
Total T0
in the volume
This vector represents

the flow
Now we need to know what T0 and T are
Energy Density
L
First consider T00 =
L

, L
,
Looks just like the energy function, doesnt it?

Think about the 1-dim. elastic rod example
2
2
2
2
1 d
1 d
d
d
L =
T00 =
+K

K

2 dt
2 dt
dx
dx
T0 should be the energy flow

L
d
T01 = d = K

dx
dx
is it?
Kinetic
energy
Potential
energy
Energy Current Density
Consider a small piece
dx
Its stretched by
d
( x + dx) ( x) =
dx
dx
( x) ( x + dx)
This gives the Hookes law force
d
F = K
dx
The work done by this piece to the next piece is
d
d
F = K
equals to T01 = K
dx
dx
Momentum Density
T
L d
First consider Ti 0 =
dxi
Again with the 1-dim. elastic rod example
1 d
d
L =
K
dx
2 dt
, L
,
This isnt so obvious
T10 =
d d
dt dx
Momentum Density
How much mass is there between x and x + dx?
d d
T10 =
dt dx
dx to the zeroth order
dx
To the first order

d
1 dx
dx
( x) ( x + dx)
Its velocity is ,
d
d
dx
so the momentum is 1
dx dt
d d
= T10
Density of excess momentum is
dt dx
T10 may be considered as the momentum density
Stress-Energy Tensor
We can interpret the stress-energy tensor T as
T00 = energy density

T0i = energy current density
Ti0 = momentum density
Tij = momentum current density
The divergence condition
dT
dx
= 0 represents
conservation of energy and momentum
Summary
Built Lagrangian formalism for continuous system
Lagrangian L = L dxdydz
Lagranges equation
Derived simple wave equation
d
dx
L
=0

Energy and momentum conservation given by the

dT
L
energy-stress tensor
T
, L
,
dx
Conservation laws
take the form of (time derivative) = (flux into volume)
=0
Mechanics
Physics 151
Lecture 24
Continuous Systems and Fields
(Chapter 13)
Built Lagrangian formalism for continuous system
Lagrangian L = Ldxdydz
Lagranges equation
Derived simple wave equation
d
dx
L
=0

Energy and momentum conservation given by the

dT
L
energy-stress tensor
T
, L
=0
,
dx
Conservation laws
take the form of (time derivative) = (flux into volume)
Ran out of time here See Goldstein 13.3 if interested
Todays lecture doesnt use it
For a discrete system, we define conjugate momenta

L
pi =
qi
Then
H = pi qi L
d
For a continuous system, L = L( , , dx
, t , xi )dxdydz
i
L
Momentum should be (t , xi ) =
Hamiltonian H = Hdxdydz where H = L
Lets see how this works
The Rod Again
Consider the 1-dim elastic rod again

x
2
2
1 d
d
Lagrangian density L =
K

2 dt
dx
L
=
=
1 d
d
H = L =
+K
2 dt
dx
Wait!
What am I going to
do with this term?
K d
=
+
2 2 dx
2
Hamiltonian formalism treats time as special
Natural structure of classical field theory is symmetric

between time and space
d L L
Because of the way momentum is defined = L
At least in Lagrangian formalism

Hamiltonian is not so useful
as in the case of discrete systems
=0

dx ,
Quantum field theory is built primarily on Lagrangian
c.f. Non-relativistic QM is almost all Hamiltonian
Fourier Transformation
Consider an elastic rod with finite length L

( x, t )
L
At a given moment t, we can Fourier transform (x,t)
nx
Assuming
( x, t ) = qn (t ) sin
= qn (t ) sin kn x
(0) = (L) = 0
L
n=0
n =0
Or, using the complex form,
( x, t ) = qn (t )e
n =0
nx
L
= qn (t )eikn x
n =0
qn(t) is a complex function
Re() assumed
Fourier Transformation
What happens to the Lagrangian?
( x, t ) = qn (t ) sin kn x
n =0
2
2
1 d
d
L =
K

2 dt
dx
= qn sin kn x qm sin km x K qn kn cos kn x qm km cos km x

2 n
m
n
m
Integrate with x and use
L
sin kn x sin km xdx = nm etc.
2
qn2
kn2 qn2 L 2 Kkn2 2
L
qn
Ldx = K
= qn
2 n 2
2 2 n 2
2
n
What does this look like?
Harmonic Oscillators
The Lagrangian represents an infinite array of

independent harmonic oscillators
L 2 Kkn2 2
qn
qn
2 n 2
2
Angular frequencies are n =
Kkn2
= vkn
Wave velocity
Wavenumber
Vibration of continuous system can be decomposed

into a set of discrete oscillators
True for any linear system

Lagrangian density must be 2nd order homogeneous
function of the fields derivatives
Small oscillation around equilibrium always OK
Phonons
Harmonic oscillator in QM has discrete energy levels
(x,t) is a superposition of sine waves with different k
Possible values of E are E = (m + 12 ) (m = 0,1, 2, )

What does this mean for the continuous system?
Each mode is a harmonic oscillator
n = vkn
En = (mn + 12 ) n
( x, t ) = qn (t ) sin kn x
n =0
Vibration energy comes in small-but-finite pieces of n

As if its a bunch of particles
Vibration can be seen as particles
Called phonons in the case of mechanical vibration
Other Examples?
Linear fields Harmonic oscillators Particles
We know an excellent example: Electromagnetic field

Corresponding particle = photon
Photoelectric effect tells us E =
Is it possible that all particles are quantized field?
For a particle of mass m, E = m 2 c 4 + p 2 c 2

Make correspondence with a harmonic oscillator
2 4
Must satisfy
m
c
2
2 2
2 2
2 4
2 2
=m c + p c
= 2 +k c
this dispersion
relation
But first of all, the field must satisfy relativity
Relativistic Field Theory
We had difficulty with relativity and multi-particles
Each particles EoM looked like
When combined, we didnt know

whose time to use
Proper time of particle s
With field like (x,t), time is just another parameter
dps
= Ks
d s
Action integral and Lagranges equations d L L

=0

look symmetric for time and space
dx ,
Can we just call x0 = ct and call it done? I = Ldxdydzdt
Almost
Lagrangian Density
Everything depends on the action integral
It must be Lorentz invariant All the equations will follow

Write it as I = L dx 0 dx1dx 2 dx 3
The volume element dx0dx1dx2dx3 is Lorentz invariant
Because det(L) = 1 for any Lorentz tensor
Lagrangian density L must be a Lorentz scalar
You must construct L using covariant quantities
Your field may be scalar () or 4-vector () or tensor

You combine them so that the product is a scalar
Field Equation
We derived Lagranges equation from Hamiltons

principle for continuous field in the last lecture
Derivation is unchanged Same equations hold
d L L
0
1
2
3
=0
I = L dx dx dx dx = 0

dx ,
d L
d L d L
Note
d = d
=
0
dx ,0 d (ct ) d ( ct ) dt dt
Ready to look at an easy example
Didnt
change
this term
Scalar Field
The simplest field is a scalar field
d
dx
Lagrangian density may be a function of L( , , , x )

For free field, L has no explicit dependence on x
Only a few scalar quantities can be formed
Try L = ,, 02 2
d
dx
L
d
= ( 2, ) + 2 02 = 0

dx
d 2
2
+
0 = 0
dx dx
What kind of field is this?

Known as
Klein-Gordon equation
Klein-Gordon Equation
Lets do Fourier in space volume V

= qk eik r
k
1
ik r
where qk = e
dV
V
k takes all the values

that satisfy the
boundary condition
Klein-Gordon equation is then

2
d 2
d
1
ik r
2
2
2
2
2
+ 0 = 2 2 + 0 = 2 qk + k qk + 0 qk e = 0
dx dx
c dt
k c
1
For each mode k, 2 qk + k 2 qk + 02 qk = 0
c
Dispersion relation is k2 = c 2 (k 2 + 02 )
Harmonic
oscillator!
Corresponds to a particle with a finite mass m = 0

c
The Field What is It?
L = ,, 02 2 gives particles with mass m = 0

c
OK, but what is the field itself?
Vibration of elastic material Phonons
Vibration of electromagnetic field Photons
The field doesnt have to be physical
It exists only in the sense that quantized excitation of are

physical (particles)
Mystical ether anybody?
QM calls it wave function,
whose (amplitude)2 is interpreted as the probability of a
particle being there
Still an indirect definition of existence
Vector Field
Field can be more complicated than scalar
How about a 4-vector, for example?

Such field represents particles with spins
4-vector field Particles with spin = 1
Electromagnetic field is an obvious example
Corresponding particle is photon, with spin 1

Recall A = ( c , A) is a 4-vector
Connection
Ex c
0
E c
with E and B
0
F
A A
x
=
=
x x E y c
E z c
Ey c
Bz
Bz
By
Bx
Ez c
Bx
0
By
Electromagnetic Field
EM field interacts with charge

Maxwells
equations
E =
0
In terms of F,
F 0
E
=
=
c
c 0
x
1 E
B 2
= 0 j
c t
F i
1 E i
i
i
B
+
=
j
(
)
0
x
c 2 t
Defining 4-current as j = ( c, j) ,
dF
j
0
dx
Lets pick a unit

in which 0 = 1
dF
+
j
=0
dx
Whats the Lagrangian?
Electromagnetic Field
To build L, we can use A, F and j
Easy to find L that works: L =

Field equation is
d
dx
A ,
F F
4
+ j A
= A , A ,
L
d F F
A
dx
2 A ,
d F + F
=
j
dx
2
dF
What we
= j =0
wanted
dx
Free EM Field
L=
F F
4
+ j A
Does it satisfy the usual wave equation?
dF
For free field (j = 0), the field equation reduces to
=0
dx
A
d A A A
=0
dx x x x x x x
This doesnt give you the usual plane waves etc.

Problem: Given E and B, A is not uniquely defined
Extra condition to fix this ambiguity
Impose Lorentz gauge condition
A
=0
2 A
1 d 2 A
2
=
A =0
2
2
x x
c dt
EM waves
with v = c
Gauge Conditions
We may add a gradient of any function to A
A =A +
x
2
2
A A
=F +
= F
x
x
x x x x
A is not fully specified without a gauge condition

Youve probably seen Coulomb gauge in Physics 15b
A = 0
This is not Lorentz invariant
Natural relativistic extension is the Lorentz gauge A = 0
All gauge conditions give you same physics
Some are easier than the others to solve
Relativistic Field Theory
Classical field theory can be made relativistic
Not very difficult although I omitted many subtleties
Lagrangian density L must be a Lorentz scalar
Built using covariant fields and currents

This limits the possible forms of L
Guided physicists toward correct picture of Nature
Quantization of the field produces particles
Fourier transformation Harmonic oscillators

Quantum field theory has enjoyed great success in
describing elementary particles and their interactions
Summary
Weve come a long way
Covered all the essentials in Goldstein

Lagrangian, conservation laws, special relativity,
Hamiltonian, canonical transformations
Central force, rigid body, oscillation
Also talked about a lot of frivolous but intriguing topics
I dont expect you to keep everything in your brain
Hopefully, it will come back and help you when you see it in
the more advanced courses of physics
At least youll know which book to look up

Classical Mechanics Lectures

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 Lectures

Caricato da

Copyright:

Formati disponibili

lecture name.

Lecturer: Masahiro Morii

Section leaders: Srinivas Paruchuri and

Tuesday/Thursday 11:30 1:00. Jefferson 256

Two or three 1-hour sections per week

Course assistant: Carol Davis

She will have all course materials (problem sets, etc.)

Getting the Best Out of Us

Ask questions at any time in the lectures and sections

E-mail morii@physics srinivas@physics mansouri@deas davis@physics

Tell us when something in the class is not working for you

Physics 15a and 15b

Mathematics 21a and 21b (or equivalent)

Introductory Mechanics, Relativity and E&M

Take these prerequisites seriously

Without sufficient background, youll get lost quickly

Classical Mechanics, Goldstein, Poole and Safko

Will follow this textbook closely

Except for skipping a few advanced materials

Grades will be based on a weighted average of

Problem sets are distributed on Thursdays

Reports are due at the next weeks Thursday lecture

Work together in groups

Groups will be assigned according to the Survey

Is This Course For Me?

Yes if you have serious interest in Physics

Ill explain what we will study in the next 15 minutes

Definitely if you want to try a career in Physics

If you are not sure at the end of this lecture

Come to my office (Lyman 239) and ask questions

What Is Classical Mechanics?

Motion of objects " Velocity and acceleration

The objects move, but do not change their properties

Mechanics: A branch of physical

Idealized particles and rigid bodies

Newtons Three Laws of Motion

You remember them, right?

Classical vs. Modern

Modern in physics means 20th century

Classical Mechanics = pre-Quantum Mechanics

What happened between the 17th and 20th centuries?

We know Relativity and QM are the right answers

Newtonian Mechanics is a human-scale approximation

Isnt that enough?

Why should we learn the theory that has been superseded?

(An advanced course in classical mechanics) introduces no new

Why Classical Mechanics?

Three Good Reasons

Close connection to Modern Physics

But first, lets go back to the 17th century

In principle, Newtons Equation of Motion predicted

All you need is a big, fast computer

In reality, life was not so easy

Intel was not founded until 1968

" Quest for more powerful mathematics

Generalizing Equation of Motion

Newtonian Mechanics deals with the objects position

But there are infinite other ways to describe motion

E.g. a more natural way for a pendulum

Number of free variables may not be 3N

Goal: finding x = x(t), y = y(t), z = z(t)

What are the Equations of Motion for generalized

Newtons Equation is about force F = ma

You start from F = F(x, t) for all particles