Kinematics and Dynamics, An Analytic Approach

Kinematics and dynamics, an analytic approach
Given ( ) a t find ( ) v t and ( ) r t

0 0 0
0
0
0 0
( )
( ) ( ) ( )
( ) ( ) ( ) ( ) ( )
(1) ( ) ( ) ( ) therefore given ( ) and the initial velocity ( )
one can find
t t t
t t t
t
t
dv t
a t dv t a t dt
dt
dv t a t dt v t v t a t dt
v t a t dt v t a t v t
= =
' ' ' ' ' = =
' ' = +
} } }
}
the velocity at all future times.

Similarly this can be done with the position
0 0 0
0
0
0 0
( )
( ) ( ) ( )
( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) therefore given ( ) and the initial position ( )
one can find t
t t t
t t t
t
t
dr t
v t dr t v t dt
dt
dr t v t dt r t r t v t dt
r t v t dt r t v t r t
= =
'' '' '' '' '' = =
'' '' = +
} } }
}
{ }
0 0
0 0
he velocity at all future times. However, we are not given ( )
but we can find it from expression (1) and substitute it in to this expression for ( ).
( ) ( ) ( ) ( )
(2
t t
t t
v t
r t
r t a t dt v t dt r t
''
' ' '' = + +
} }
{ }
0 0
0 0 0
0 0
) ( ) ( ) ( )( ) ( ) This is now an explicit solution
finding ( ) from a given function ( ) and two initial conditions ( ) and ( ). Let's
solve this for th
t t
t t
r t a t dt dt v t t t r t
r t a t v t r t
''
' ' '' = + +
} }
e simplest ( ) a constant. a t a =

{ }
{ }
0 0
0
0 0
0
0
0 0 0
0 0 0 0
0 0 0 0
2
0
Let ( ) then
( ) ( )( ) ( )
( ) ( ) ( )( ) ( )
( ) ( )( ) ( )
( )
( ) (
2
t t
t t
t
t
t t
t t
t
t
t
t
a t a
r t a dt dt v t t t r t
r t a t t dt v t t t r t
r t at dt at dt v t t t r t
a t
r t at t v t
''
=
' '' = + +
'' '' = + +
'' '' '' = + +
''
'' = +
} }
}
} }
( )
( )
0 0 0
2 2
2 0
0 0 0 0 0
2 2
0 0 0 0 0
2
0 0 0 0
)( ) ( )
( ) ( )( ) ( )
2 2
( ) 2 ( )( ) ( )
2
1
( ) ( )( ) ( ) This is the expression for the trajectory of a
2
partic
t t r t
at at
r t at t at v t t t r t
a
r t t t t t v t t t r t
r t a t t v t t t r t
+
= + + +
= + + +
= + +
le undergoing uniform acceleration. It must be emphasized that this is a
case and that equation (2) is the general solution.
Taking the derivative of this special case gives the velocity growin
SPECIAL
( )
0 0
g linearly in time:
( ) ( )
This discussion demonstrates that if the acceleration of a particle is known
as a function of time, as well as its initial position and velocity, than a unique
sol
v t a t t v t = +
ution for the trajectory of that particle can be found. Try some of the examples below.

a(t)
t
v(t)
t
r(t)
t
v(0)= 2
r(0)= 0
a(t)
t
v(t)
t
r(t)
t
v(0)= 0 r(0)= 0
a(t)
t
v(t)
t
r(t)
t
v(0)= 0
r(0)= 0
a(t)
t
v(t)
t
r(t)
t
v(0)= 0
r(0)= 0

As an exercise find the trajectory analytically for the example shown above,
( ) sin( ), (0) 0, r(0) 0 a t t v = = =

To turn the discussion to dynamics, we must consider the action of the forces on the object.
From high school you learned that F ma = . Lets be more precise and reformulate this for the
kinematical discussion that was given above. Better to write
1
( ) ( ) a t F t
m
=

. This says that
you need to sum all the forces (as vectors) on the object to obtain the net force, then dividing by
the mass gives the acceleration (also a vector). Having now obtained the acceleration the
trajectory can be determined as described above. Mathematically this is equivalent to writing
0 0
0 0 0
1
( ) ( ) ( )( ) ( )
t t
t t
r t F t dt dt v t t t r t
m
''

' ' '' = + +
`
)
} }
. Looks nasty, and it certainly can be for
complicated forces. That is why you only learned a couple special cases in high school. How
can we go further? We could wait until you have learned a lot more mathematics, but even that
would not help with many situations because there is just no way to analytically solve this
expression. One way forward, and this is what many researchers and scientist do, is use
computers to assist in the solution of these more complicated and interesting problems. This is
the approach that we will follow in this class.

Kinematics and dynamics, a computational approach
0 0
We begin the discussion of numerical approaches to mechanics with the definition for the velocity
( ) ( )
( ) lim lim or ( ) ( ) ( ) for small . This defin
t t
r r t t r t
v t r t t r t v t t t
t t
A A
A + A
= = + A ~ + A A
A A
ition
comes from mathematics, calculus specifically, and does not contain any real physics.
2
2
Where the physics comes in is through the definition of momentum. We know that
1
and inverting this and solving for the velocity gives so if we can find then
1
w
mv
p
v
c
p
m
v p
p
mc
=
| |
|
\ .
=
| |
|
\ .
2
e find the position using
( )
( ) ( ) ( ) ( ) This tells us how to find the
( )
1
position at the next increment in time as long as we know the current position and
p t
m
r t t r t v t t r t t
p t
mc
+ A ~ + A = + A
| |
|
\ .
the momentum. So the question becomes how to find the momentum. For this
we will use Newton's relation between the net force on an object and the time rate
of change of the momentum. Mathematically: rearranging this expression
or ( ) ( ) tells us that if we know the
present momentum and the net force on an object, than we can find the momentum
dp
F
dt
dp F dt p t dt p t F dt
=
| | | |
= + = +
| |
\ . \ .

at
a future time. Let's see how this works with a simple example, an object acted on by a constant
force.

F

Start with (0) 0 and (0) 0 Use position and momentum update to calculate
future positions ( ) ( ) and ( ) ( ) ( )
also let the nonrelativistic limit. Let's
p r
p t t p t F t r t t r t v t t
p
v
m

2
2
2
2
2
2
2
2
2
set up a chart for each time step. Note .
0 0 0 0
1 0
2 2 2
3
3 3 3
6
4 4 4
10
5 5 5
15
6 6 6
21
7 7 7
28
8 8 8
36
9 9 9
45
10 10 10
t n t
n t p r
t F t
F t
t F t
m
F t
t F t
m
F t
t F t
m
F t
t F t
m
F t
t F t
m
F t
t F t
m
F t
t F t
m
F t
t F t
m
F t
t F t
m

Graphically this looks like this:

where blue is from the data above and green is the analytic result. You can see that there is a
difference between the two sets of data. The numerical calculation does not exactly reproduce
the analytic result. Recognizing and comprehending the source and magnitude of this difference
is very important to your understanding of the solution to the problem. Below shows a graph of
the absolute error
_ _ _ abs error analytic result computational result (blue) and the fractional error
( _ _ )
_
_
analytic result computational result
frac error
analytic result
(red)
Examining the graph shows that the absolute error increases linearly with time, but the fractional
error decreases. It is often important for you to understand whether you care about either the
absolute or fractional error most. Sometimes you can determine this and other you cant.
0
50
100
150
200
250
0 2 4 6 8 10 12 14 16 18 20
analytic calculation
numerical calculation

There is an ambiguity that is worthy of a bit of discussion. In the numerical solution the two
key line are
???
(classical limit)
new old
new old
P P F t
P
r r t
m
| |
= + A
|
\ .
= + A

The question boils down to what momentum do you use for the second expression? This is
equivalent to asking in which order do you evaluate these expressions:

P P F t
P
r r t
m
| |
= + A
|
\ .
= + A
or
P
r r t
m
P P F t
= + A
| |
= + A
|
\ .

Do you update the momentum first and then the position using the new momentum or do you
update the position first using the old momentum and then the momentum using the new
position? Actually, there are a couple more permutations that you can think of. Sometimes these
seemingly minor differences in the algorithm really make a significant change to the accuracy of
your calculation. Topics such as these come under the name of numerical methods. In this
class we will be doing very little in the way of trying to get the best answer by using the best
algorithm. However, occasionally this might come up and you should be at least aware of the
issue.
0
20
40
60
80
100
120
0 2 4 6 8 10 12 14 16 18 20
absolute error
fractional error (%)
One simple way of improving the accuracy of almost any algorithm is to reduce the time
increment t A . Selection of the proper, or reasonable, t A for a problem often involves a trade-off
between the necessary accuracy and the length of time that you can afford to run the simulation.
It also involves understanding the natural time-scales of the physical situation and what
approximations you can make while still obtaining the information that you need to know.

Lets examine a real physical system. Consider an electron in a vacuum. This is a model for an
electron moving in a vacuum tube, such as an old TV tube (in the days before flat panel screens).
To provide a constant force we produce a uniform electric field in the region of space containing
the electron. A simple schematic would be:

The force on the electron is the electric charge (q) times the electric field, F qE = ; in the region
between the two plates the field is approximately constant, therefore so is the force. Consider a
simulation time of 10
-10
sec. What should the t A be? Let explore this computationally. Below is
a graph of position vs time with data from simulations with two different time steps, large 10
-12

sec and small 10
-14
sec.

There is clearly a difference between the two data sets. Which is correct? This problem is one
of the rare situations where we know the answer analytically,
2
1
( )
2
x
eE
x t t
m
| |
=
|
\ .
assuming the
electron starts at (0) 0, 0, 0 and (0) 0, 0, 0 r v =< > =< >. The parabolic analytic solution is
F
10 V 0 V
1 m
indistinguishable from the green dots. However we can look more precisely by plotting the
fractional error, defined above.

The simulation with the larger time step (blue) has an error greater than 3% for all of the time,
whereas using the smaller time step reduces the error well below 1% for almost all of the
simulation time. Since in this problem we know the exact answer we can calculate these errors,
generally this is not possible. What do you do then? The simplest thing is to try a time step,
then try a smaller one and see how much your answer changes. If it changes a lot (that means
more than you would like your error to be) try an even smaller time step and repeat this process
until you either have answers to the accuracy that you desire or you run out of computer power
and the simulations take too long. This is when you wish you had a faster computer and have
joined the legions of research scientists wanting more powerful computation machines.
from __future__ import division
from visual import *
from visual.graph import * # import graphing features

#Simulation of an electron acted on by a constant force.
#The simulation shows two different time steps and the relative errors.

time = 0 #start the clock at zero
small_dt = 1e-14 #small time increment
large_dt = 1e-12 #large time increment
simulation_time = 1e-10 #time between scattering events

c = 3e8 #speed of light m/s
e = 1.6e-19 #unit of charge
electron_mass = 9.11e-31 #mass of electron

electron_init_pos = vector(0,0,0) #call origin of system
electron_init_vel = vector(0,0,0) #initial velocity,
electron_init_mom = electron_mass*electron_init_vel #initial mommentum

E = vector(10,0,0) #Electric field in x direction, in Volts/meter

Electric_Force = -e*E #constant force applied by electric field

#graphics
position = gdisplay(title='Position',
xtitle='time (sec)', ytitle = 'distance (m)',
x=0, y=0, width=1000,
height=400,
background= color.black) #window for position

error = gdisplay(title='Percent Error',
xtitle='time (sec)', ytitle = 'factional error (%)',
x=0, y=400, width=1000,
ymax=10, height=400,
background= color.black) #window for error

pos_large_step = gdots(gdisplay=position, color=color.cyan) # a graphics curve for large #steps
pos_small_step = gdots(gdisplay=position, color=color.green) # a graphics curve for small #steps
analytic = gcurve(gdisplay=position, color=color.red) # a graphics curve for exact

delta_small_dot = gdots(gdisplay=error, color=color.green) # error for small steps
delta_large_dot = gdots(gdisplay=error, color=color.cyan) # error for large steps

#initialize all loop variables
electron_mom_large = electron_init_mom
electron_mom_small = electron_init_mom

electron_pos_large = electron_init_pos
electron_pos_small = electron_init_pos

time_small = time
time_large = time

#main loops (typically only one but this program used two different time steps

while (time_large< simulation_time): #one nanosec of acceleration

electron_mom_large = electron_mom_large + (Electric_Force)*large_dt #update electron momentum
electron_pos_large = electron_pos_large + (electron_mom_large/electron_mass)*large_dt #update electron position

analytic_pos = electron_init_pos + electron_init_vel*time + 0.5*(Electric_Force/electron_mass)*(time_large)**2

pos_large_step.plot(pos=(time_large, electron_pos_large.x )) # plot position

if(time_large > large_dt): #can't measure rel error at t = 0, delay one time step
delta_large = 100*abs((analytic_pos.x - electron_pos_large.x)/analytic_pos.x) #find error
delta_large_dot.plot(pos=(time_large, (delta_large))) #plot percent error

time_large = time_large + large_dt #increment time by large step

while (time_small< simulation_time): #one nanosec of acceleration

electron_mom_small=electron_mom_small + (Electric_Force)*small_dt #update electron momentum
electron_pos_small = electron_pos_small + (electron_mom_small/electron_mass)*small_dt #update electron position

analytic_pos = electron_init_pos + electron_init_vel*time + 0.5*(Electric_Force/electron_mass)*(time_small)**2

pos_small_step.plot(pos=(time_small, electron_pos_small.x )) # plot position

if(time_small > small_dt): #can't measure rel error at t = 0, delay one time step
delta_small = 100*abs((analytic_pos.x - electron_pos_small.x)/analytic_pos.x) #find error
delta_small_dot.plot(pos=(time_small, (delta_small))) #plot percent error

analytic.plot(pos=(time_small, analytic_pos.x )) #plot classical result

time_small = time_small + small_dt #increment time by small step

print electron_mom_small/(electron_mass*c) #check to see if velocity is much less than c, classical limit
print "average velocity", electron_pos_small/time_small
from random import random, uniform # import random number functions

#Simulation of an electron acted on by a constant force and suffering
#elastic and inelastic scattering. Drude model for electron conduction.

time = 0 # start the system clock at zero
elastic_time = 0 # time since last elastic collision
inelastic_time = 0 # time since last inelastic collision
count = 0 # count number of cycles in loop
dt = 5e-16 # time increment (sec)

simulation_time = 1e-10 #time of simulation
elastic_scattering = 1e-14 #time between elastic scattering events
inelastic_scattering = 1e-13 #time between inelastic scattering events

two_pi = 2*3.14159

electron_init_mom = electron_mass*electron_init_vel #initial momentum



#graphics
position = gdisplay(title = 'Position vs time',
xtitle = 'time', ytitle = 'x position',
x=0, y=0,
width=1000, height=400,
background= color.black) # window for position vs. time

pos_vs_time = gdots(gdisplay=position, color=color.cyan) # a graphics curve for large steps

electron_mom = electron_init_mom
electron_pos = electron_init_pos

#main loop

while (time < simulation_time): #length of time to simulate

electron_mom = electron_mom + (Electric_Force)*dt #update electron momentum
electron_pos = electron_pos + (electron_mom/electron_mass)*dt #update electron position

if count % 10 == 0: # only plot every n'th point
pos_vs_time.plot(pos=(time, electron_pos.x )) # plot x position vs time

if(elastic_time > elastic_scattering): #is it time for an elastic collision?
elastic_time = 0 #reset elastic time clock
#randomize electron momentum
electron_mom = rotate(electron_mom, angle = uniform(0,two_pi), axis = (0,0,1))
if(inelastic_time > inelastic_scattering): # is it time for an inelastic collision?
inelastic_time = 0 #resent inelastic time clock
electron_mom = vector(0,0,0) #reset electron momentum to zero
#print "inelastic scatter"

time = time + dt # increment system time by large step
elastic_time = elastic_time + dt # increment time since last elastic collision
inelastic_time = inelastic_time + dt # increment time since last inelastic collision
count = count + 1 # total number of cycles
#end main loop

print electron_mom/(electron_mass*c) #check to see if velocity is much less than c,
print "average velocity", electron_pos/time

Electron conduction in a metal
We have just solved approximately a problem, uniformly accelerating electron, that you could
have done exactly while in high school. The power of the computational techniques becomes
apparent in a slightly more complicated problem. Consider an electron moving under the
influence of the same field however now
it is in a metal. How can we change the
model to take into account the other
atoms? This schematic is out of scale in
almost all respects, but you know what it
means. (Scientists draw figures like this
all the time.)
Clearly the model must take into account the interactions of the electron with the atoms arranged
in a crystal. To do this properly requires quantum mechanics (QM) and draws on the wave
nature of electrons; however we can make a model that captures a lot of the physics without
using QM. Think about a racquetball bouncing around in a court. You often ignore the
complicated interactions of the ball as it bounces off the wall and instead concentrated on what
change occurred to the momentum of the ball from before to after the collision. The same
approach is a good one here. How can the interaction with the lattice of atoms change the
momentum of the electron? Two ways, it can alter the direction and/or the magnitude
(remember momentum is a vector). In fact, there are two general types of collision processes,
elastic where only the direction of the electron momentum is changed and inelastic where the
electron is slowed down or stopped. The time between elastic events is usually shorter than
inelastic. For metals at room temperature typical times are 10
-14
sec between elastic collisions
and 10
-13
between inelastic events. (These times vary with the perfection of the atomic crystal
lattice and the temperature.)
It is very easy to add these two processes to our simulation for a particle undergoing uniform
acceleration. You simply need to add two timers, one each for elastic and inelastic collisions.
Each cycle through the main loop, after the momentum and position are updated, check the
elastic and inelastic clocks to determine if such a process is due to occur. If so, either select a
random direction for the momentum, elastic event, or set the momentum to zero for an inelastic
event. (You would not have to decrease the momentum this much, but it is a fine
approximation.) Reset the timer to zero and continue on with the simulation.
What do you expect the results of such a simulation to show? First, consider the motion for
times comparable to the elastic time. The particle is gaining momentum and then bounces off an
atom. This should look like a set of piecewise continuous parabolas strung together. With the
elastic collision time set at 10
-14
sec and the simulation time at 10
-13
sec, a graph of the x position
vs. time is shown below. (
16
10 t

sec, you need the time step to be smaller than the collision
times)
F
10 V 0 V
1 m

As you can see, the graph shows the electron some times moving forward and sometime back.
However, the acceleration is always in the negative direction. Look at each sement of the curve,
between collisions the curvature is always downward. So, even though the momentum is
randomly rotated in the x-y plane at each elastic event, the electron has made a net motion in the
x direction.
Now consider the velocity vs time graph:

The x component of the velocity is always decreasing, except for the jumps due to scattering.
The red spots correspond to an inelastic event where the velocity was reduce to zero. During
these collisions the kinetic energy of the electron is lost and converted to heat, i.e vibration of the
atomic cores.
Moving to a longer simulation time, there are more inelastive collisions that reset the momentum
to zero. Below is a plot of the x position vs time for 5 x 10
-13
sec, or 4 inelastic collisions. You
can begin to see the electron drifting towards the negative x direction, but the trajectory is still
pretty irregular.

Carrying the simulation out for the same time as for the free electron (10
-10
sec) yields this
trajectory:

This is an amazing result! The system has nothing in it but uniform acceleration and collisions
and yet the result is an approximately constant velocity of the electron down the wire. Lets
compare the average velocity of the free electron and the electron in the metal for the same
simulation time of 10
-10
sec. Remember the force on the electrons in two cases was the same.
Looking back at the graph for the free electron it traveled ~8 x 10
-9
m in 10
-10
sec or an average
velocity in the negative x-direction of ~80 m/s. Whereas the electron in the metal only traveled
~1 x 10
-12
m in the same time for an average velocity of 0.01 m/s (1 cm/sec). This is a reduction
in average velocity over this short time period of 10
4
. If we looked at each electron after they
had traveled 1 meter the free electron would have a velocity of 1.8 x 10
6
m/s while the electron
in the metal would still have a velocity of 0.01 m/s.
What do you think happens if the voltage across the plates, and thus the electric field and force,
is doubled? If you guessed that the drift velocity doubles, you are correct! In the graph below
the applied voltage in the x direction is 10V (black), 20V (cyan) and 40V (red). Clearly the
slope, drift velocity
d
v , increases linearly with the applied force.

Now look at what we just said: velocity is proportional to force. This is certainly not Newtons
second law. Due to the collisions in the model, the behavior changes from acceleration being
proportional to force (which it is between collisions) to velocity being approximately
proportional to force, quite a remarkable and important result.
The result that the velocity of the electron is proportional to the applied voltage (or force) is a
very useful result and often goes by the name Ohms law. Most students have seen V IR
where V is the applied voltage, I is the current, and R the resistance. The current is the product
of the charge and the drift velocity,
d
I e v
. From our model, the drift velocity is constant in

time and proportional to the applied voltage. This then shows how Ohms law arises for metallic
conduction. With a little bit more analysis you can determine the resistance in terms of the
density of electrons and the elastic scattering time. It is interesting that from an electrical
measurement a parameter for the mechanical model can be determined.

As a final thought, we have not looked
at the other two components, y or z, of
the motion for electron in the metal.
The force due to the electric field is
strictly in the x direction, so the only
way that the electron acquires any
momentum in the other directions is by
collisions. The momentum is randomly
chosen and so on average the electron
should not have moved in the lateral
directions. However, each trajectory is
different and it is an interesting problem
to determine the statistics of how far the
electron would travel in y or z on its
random walk.


#Simulation of an electron acted on by a constant force







#graphics
position = gdisplay(title = 'Position vs time',
xtitle = 'time', ytitle ='x position',
x=0, y=0,
foreground=color.black,
background= color.white) # window for position vs. time

pos_vs_time = gdots(gdisplay=position, color=color.cyan) # a graphics curve for position vs. time

#electron_kinetic_energy = mag2(electron_mom)/(2*electron_mass)

#main loop

while (time < simulation_time): # length of time to simulate

electron_mom = electron_mom + (Electric_Force)*dt # update electron momentum
electron_pos = electron_pos + (electron_mom/electron_mass)*dt # update electron position



ave_vel = electron_pos/time

print ("time", time)
print ("distance traveled along field", electron_pos.x)
print ("average velocity", ave_vel)
print ("final velocity", electron_mom/electron_mass)


#Simulation of an electron acted on by a constant force and suffering
#elastic and inelastic scattering. Drude model for electron conduction.

elastic_time = 0 # time since last elastic collision
inelastic_time = 0 # time since last inelastic collision
inelastic_collision = 0 # if last suffered an inelastic collision

elastic_scattering = 1e-14 #time between elastic scattering events
inelastic_scattering = 1e-13 #time between inelastic scattering events

two_pi = 2*3.14159




#graphics
position = gdisplay(title = 'x Position vs time',
xtitle = 'time', ytitle ='x position',
x=0, y=0,

velocity_display = gdisplay(title = 'x velocity vs time',
xtitle = 'time', ytitle ='x velocity',
x=0, y=450,
background= color.white) # window for energy loss vs. time

pos_vs_time = gdots(gdisplay=position, color=color.cyan) # a graphics curve for position vs. time
velocity_vs_time = gdots(gdisplay=velocity_display, color = color.black) #a graphics curve for velocity vs. time

electron_kinetic_energy = mag2(electron_mom)/(2*electron_mass)

#main loop

while (time < simulation_time): # length of time to simulate

electron_mom = electron_mom + (Electric_Force)*dt # update electron momentum
electron_pos = electron_pos + (electron_mom/electron_mass)*dt # update electron position

if(inelastic_collision == 1):
velocity_vs_time.color = color.red
else:
velocity_vs_time.color = color.black
velocity_vs_time.plot(pos=(time, electron_mom.x/electron_mass)) # plot v_x vs. time

if(elastic_time > elastic_scattering): # is it time for an elastic collision?
elastic_time = 0 # reset elastic time clock
electron_mom = rotate(electron_mom, angle = uniform(0,two_pi), axis = (0,0,1)) #randomize electron momentum
inelastic_collision = 0

if(inelastic_time > inelastic_scattering): # is it time for an inelastic collision?
inelastic_time = 0 # resent inelastic time clock
electron_mom = vector(0,0,0) # reset electon momentum to zero
electron_kinetic_energy = electron_kinetic_energy + mag2(electron_mom)/(2*electron_mass) # calc KE lost and sum it
inelastic_collision = 1

elastic_time = elastic_time + dt # increment time since last elastic collision
inelastic_time = inelastic_time + dt # increment time since last inelastic collision

ave_vel = electron_pos/time
current = -e*electron_pos.x/time
voltage_drop = E.x*electron_pos.x

print ("time", time)
print ("distance traveled along field", electron_pos.x)
print ("average velocity", ave_vel)
print ("current", current)

Air resistance or drag is a very common type of friction experienced in many situations, a leaf
falling from a tree, riding your bicycle or a jet flying through the air. It is often impossible to
ignore the effects of air drag, even though many introductory physics courses do. In this lesson
you will learn that friction due to air (or other mediums) can be modeled reasonably easily and is
simply another force that can be included in your numerical simulations. (As an aside, the
analytic solution is very complicated or impossible to write in closed form.) We do not derive
an expression for the air drag, but instead we rely on experiment to give approximate or
empirical expressions. The most commonly used expression is
2
1
2
air
F CA v v where
is the density of the medium (air in this case), is the velocity, is the unit vector in the direction
of motion, is the frontal cross sectional area, and is the coefficient of drag.
v v
A C
The negative sign makes the resistive force oppose the motion. This expression is only
approximately correct, and not even for all speeds, but it is a good approximation over a large
enough range of speeds to be extremely useful. The coefficient C hides a lot of the details in the
problem. For instance it contains the shape of the object and perhaps if it is near a surface or out
in as open space. Take a car as an example, a sleek racing car will have a smaller coefficient of
drag than a boxy jeep. Often on semi-trucks you will see a curved piece of plastic above the cab,
right in front of the flat part of the trailer. This significantly reduces air drag, i.e. C , and
increases the fuel economy of the truck. In any case, C must be determined empirically, by
experiment. For many common shapes the coefficient of friction is tabulated and can be looked
up.
Lets examine a very common example where air resistance plays a major role. A skydiver
jumps out of an airplane and starts to accelerate towards the earth. As his velocity increases the
air begins to put an opposing force that reduces his downward acceleration. Eventually, the
upward force of the air equals the downward force of gravity and the skydiver ceases
accelerating and falls with a constant speed. This is called the terminal velocity. It is
interesting to note that terminal velocity for a skydiver depends on his position. If falling
headfirst in a streamlined fashion his cross sectional frontal area is just the head and shoulders,
however is he is falling spread-eagle then the frontal area is his entire body. These areas have a
ratio of ~20 and significantly reduce the terminal velocity for skydivers falling chest first.
Now we will make some estimates and do a simulation. The coefficient of drag is ~1.1, the
frontal area (spread-eagle) 0.75 m
2
, the density of air is ~1.2 kg/m
3
, the mass of a sky diver
might be around 80 kg. This allows us to determine the terminal speed by equating the
gravitational force and the air drag:
2
1
2
terminal
mg CA v or
2
terminal
mg
v
CA
Substituting in the
values from above give a terminal velocity (spread-eagle) of ~40 m/s or ~90 mi/h. If you go
headfirst the terminal velocity is closer to 180 mi/h. (This is why I dont skydive! I know too
much physics.) A more difficult question to answer is how far, or long, do you fall to achieve
90% of terminal velocity? This cannot be solved easily analytically, but numerical computation
will work very well. The forces on the object are:
2
1
2
net
F F mg CA v v
. How do we
code this? Here is the pseudocode for the main loop:
Main Loop
F_grav = m* g
F_air = -0.5*A*C*density*(momentum/mass)
2
*(momentum unit vector)
momentum = momentum +( F_grav + F_air)*dt
position = position + (Momentum/mass)*dt
time = time + dt
This looks exactly like the racquetball problem, but we have included F_air in the total force.
Before we include air drag lets review the results from our high school physics class for free fall
with no drag. In this case the x and y motion are completely decoupled, the velocity in y
changes by -9.8 m/s/s, and the velocity in x is constant. Below is a graph of such a trajectory,
note that the axis are y position vs x position, NOT time. The sky diver jumps out of a plane
flying at an altitude of 3000 m while traveling at 50 m/s in the x direction. Not including air
resistance the trajectory is shown below:

The dots are equally spaced in time and you can see a nice parabola where the even spacing in
the x-direction shows a constant speed in that direction, while the increasing vertical spacing
between points shows the acceleration due to gravity in the y direction. No doubt you spent a
considerable amount of time studying this in high school. While this is an excellent limiting
case, in actuality rarely can air resistance be ignored. Lets consider the same problem but
include air drag. The trajectory looks like this:

Comparing the two trajectories you can see quite a difference! The most obvious thing is that in
the second situation the skydiver comes down essentially vertically and the velocity in the x-
direction has gone to zero. This is due to the drag of the air gradually eliminating the forward
motion of the skydiver. Compare the x range for the two scenarios: without air drag the skydiver
lands ~1200 m down range (directly under the plane, assuming it continuums its original course),
whereas when air drag is included the range is reduced to ~185 m. This is a very significant
difference (reduction by 1/6
th
). Anyone who has played golf is aware that the golf ball comes
down from a long drive almost vertically. Or simply throw a crumpled piece of paper
horizontally and watch how it falls. Doesnt look like a parabola, does it? Ignoring air
resistance is a very large approximation. Is it reasonable? Where do you want your helpers
waiting for you after you land?
A second important point to notice is a bit more subtle. Look at the spacing of the dots as the
skydiver is in the final vertical descent. Are they increasing in spacing as they would be for
uniform acceleration due to gravity? No, they are not. In fact, they look evenly spaced.
To examine this more closely, we should plot the y-position, y-velocity and y-acceleration vs.
time. These three graphs are shown below:

The y-position vs time plot shows a small amount of downward curvature, and the velocity vs.
time plot similarly starts off with a slop of -9.8 m/s
2
, but after about 5 sec, when the speed is ~30
m/s, the velocity curve starts to bend over and level off. This is reflected in the y-position curve
becoming a linear relationship. The terminal velocity is ~40 m/s, just as we determined above
from equating the air drag and gravitational forces. It is interesting to see how rapidly the
skydiver obtains terminal velocity, the last plot of the acceleration vs. time shows that after just
~10 sec, downward acceleration is essentially zero and from then on his speed remains constant.
That is, of course, until he hits the Earth.
We have omitted one very important aspect of a sky jumpers flight, the parachute. A parachute
dramatically increases the frontal cross sectional area that is being resisted by the air. A typical
parachute might have an effective area of 40 m
2
, ~50 times larger than just the spread-eagle
jumper. Most jumpers want to have their chute deployed by ~300 m above the Earth. We can
add a couple lines of code to the simulation to allow a large increase in the area A and a smaller
increase in the drag coefficient C, from 1.1 to 1.5, when the skydiver is below 300 m in altitude.
Below is a plot of the trajectory.

This trajectory looks very similar to the one above, except the dots get much more closely spaced
starting at 300 m. The y-position and y-velocity vs. time plots are more revealing.

After falling for about 70 seconds, the skydiver is at a height of 300 m and the chute deploys.
Very rapidly the downward velocity is decreased from 40 m/s to a new terminal velocity of 4.6
m/s (~10 mi/hr). This is the impact speed for the skydiver unless some more fancy things are
done. Hitting the ground at 10 mi/hr is still pretty fast and broken bones are not uncommon.
It is interesting to note how large the positive acceleration is when the chute opens

At the start of the dive the y-acceleration is -9.8 m/s
2
and after about 10 seconds it is reduced to
zero. At 70 sec the y-acceleration jumps up momentarily to over 700 m/s
2
or ~70 gs. The
duration of this extremely high force is quite small, each point on this part of the curve is 0.01
sec, so after ~ 0.1 sec the acceleration is at a relatively more manageable 100 m/s
2
or ~10gs and
still rapidly falling. There must be one heck of a
da
jerk
dt
, when that chute opens.
Of course, we do not have a properly detailed model for the chute to open and this is an
artificially high value. Our chute opens in 0.01 sec, whereas a real chute would take much
longer. This would reduce the extremely high initial values of the acceleration. From a video
we could try to determine the area of the chute as a function of time and this would allow a better
model for the acceleration during chute deployment.
This last argument is an important one to understand from a big picture point of view. We
have introduced a very simple approximation for the chute opening, one time step it is not there
and in the next it is fully deployed. This is not good enough to answer detailed questions like,
what is the maximum acceleration that the skydiver experiences? However the model is just fine
for determining the speed at which the skydiver hits the ground. This is a good example of
knowing where you have a reasonable model, when it is accurate, and where it falls short. Pun
intended.


#Simulation of a skydiver jumping from a plane and popping a parachute at 300m
#above the ground

dt = .5 # time increment (sec)
count = 0 # count number of cycles through main loop

mass = 80 # mass of skydiver
g = vector(0,-9.8,0) # acceleration near Earth
rho = 1.22 # density of air kg/m**3
A = 0.75 # frontal area of skydiver, 0.75 m**2 spread eagle, 0.1 m**2 headfirst
C = 1.1 # Coefficient of drag

Force_grav = mass*g # constant force applied by gravity field

parachute = 1 # turns the opening of the parachute on (1) or off (0)

#graphics
position = gdisplay(title = ' Y Position vs time',
xtitle = 'time (s)', ytitle ='y position (m)',
x=0, y=0,

velocity = gdisplay(title = ' Y Velocity vs time',
xtitle = 'time (s)', ytitle ='velocity (m/s)',
x=0, y=450,
background= color.white) # window for velocity vs. time

trajectory = gdisplay(title = 'height vs x position',
xtitle = 'x position (m)', ytitle ='height (m)',
x=0, y=0,

acceleration = gdisplay(title = 'Y acceleration vs time',
xtitle = 'time (s)', ytitle ='Y acceleraton (m/s**2)',
x=0, y=0,

pos_vs_time = gdots(gdisplay=position, color=color.black) # a graphics curve for position vs. time
vel_vs_time = gdots(gdisplay=velocity, color=color.black) # a graphics curve for velocity vs. time
y_vs_x = gdots(gdisplay=trajectory, color=color.black) # a graphics curve for trajectory
acc_vs_time = gdots(gdisplay=acceleration, color=color.black) # a graphics curve for trajectory

initial_vel = vector(50,0,0) # 50 m/s in x is 120 mi/hi for a slow plane
initial_mom = mass*initial_vel
initial_pos = vector(0,3000,0)

position = initial_pos
momentum = initial_mom

#main loop
while (position.y > 0): # simulate while skydiver is above ground

if parachute and position.y < 300: #if have parachute, open at 300 meters high
A = 41 # parachute cross sectional area (m**2), 24' diameter
C = 1.5 # Coefficient of drag of chute
dt =.01 # need small dt to better handle the large force at chute opening

if mag(momentum) == 0: #if no velocity, air drag is 0
Force_air = vector(0,0,0)
else:
Force_air = -0.5*C*A*rho*mag2(momentum/mass)*norm(momentum) #air drag

momentum = momentum + (Force_grav + Force_air)*dt # update momentum
position = position + (momentum/mass)*dt # update position
accel = (Force_grav + Force_air)/mass # only used for plotting

pos_vs_time.plot(pos=(time, position.y )) # plot y position vs time
vel_vs_time.plot(pos=(time, momentum.y/mass )) # plot y velocity vs time
y_vs_x.plot(pos=(position.x, position.y)) # plot trajectory y vs. x
acc_vs_time.plot(pos=(time,accel.y)) # plot acceleration y vs. time

time = time + dt # increment system time
count = count + 1 # count number of cycles through main loop
print momentum.y/mass
Relativistic Dynamics
Many students are interested in how Einsteins theory of relativity affects our understanding of
mechanics. The answer is: profoundly. To thoroughly understand relativity, first special relativity and
then general, you must suspend your everyday physics intuition and rely on the beautiful mathematics that
describes the phenomena. Of course, there have now been countless experiments testing relativity to a
very high precision. So, although it seems to violate your low-speed sensibilities, there is no doubt that it
is a more complete theory than classical Newtonian mechanics.
Even though the behavior is strange and the mathematics at times complicated, the theory of special
relativity is based on only two very simple postulates:
1) The laws of physics must be the same for all observers, no matter what inertial reference frame, i.e.
frames in uniform translation.
2) The speed of light in a vacuum is the same for all observers, independent of the motion of the emitting
body and no matter what inertial reference frame of the observer.
This is the beauty of physics. Einstein started with just two assumptions and from that he deduced an
entirely new world view.

We will not be able to study the theory of relativity in any detail in this class. However, we can still use
one of the results to greatly improve our study of high speed particles.
Special relativity does not alter Newtons second law:
dP
F
dt
=
. This means that the momentum

update statement
new old
P P F t
| |
~ + A
|
\ .

is correct to use. However, the kinematical statement
new old
P
r r t
m
~ + A

is no longer adequate when the particle has a large speed. This statement comes from
the differential equation
dr
v
dt
, which is a kinematical definition and is certainly correct. What is no

longer correct is that v
is simply equal to
P
m
.Instead, the correct relativistic expression is

2
1
P
m
v
P
mc
=
| |
+
|
\ .
. This come from solving for velocity using the correct relativistic expression for the
momentum P mv

. In the low speed limit v c where the term
2
1
P
mc
| |
|
\ .
, it can be neglected
and the expression for the velocity reduces to the familiar P mv ~

. However, in the case of very large
momentum
2
1
P
mc
| |
|
\ .
, the expression reduces to v cv ~
. This is the case for an ultrarelativistic

particle. No matter how large you make the particles momentum, its speed will approach c from below,
but never be equal to . c

Returning to the iterative update equations, the correct expression for finding the new position is
2
1
new old
P
m
r r t
P
mc
~ + A
| |
+
|
\ .
. Other than this change in the position update, the algorithm for calculating
the trajectory of a particle remains the same.
Lets see how this works with an example. Consider an electron in an old television tube (not a flat
panel). An electric potential difference is used to put a force on the particle. In a typical television this
voltage is ~50kV and the distance between the front and rear electrodes is ~1 m. The force on the
electron is given by the expression
V
F q
D
= where q is the charge on the e
, V is the voltage between the

electrodes and Dis the separation.

The two graphs below show the position and velocity vs. time for an election accelerated through 50kV in
the television tube. The red curve shows the trajectory computed nonrelativistically (with Newtonian
mechanics), while the black trajectory used the relativistic expressions. You can see that the calculated
positions differ by several cm over the 1 m length.

Classical
Relativistic
Equally interesting is the velocity of the particle. By the end of the trajectory the electron is traveling at
40% of the speed of light. So inside your old TV set you have quite a few relativistic particles. Note that
the relativistically calculated velocity is lower than that calculated classically. In the correct relativistic
treatment the force acting on the particle increases the momentum, but does not increase the velocity as
much as it would have been classically. This is seen by the black relativistic velocity falling below the
red classical calculation.

These effects become more apparent as the energy of the particle increases. If the electron is in a particle
accelerator (a very modest one) and subject to a 5 MV potential then the relativistic effects completely
dominate the behavior of the particle. The same two graphs are shown below. In the first, position vs.
time, the difference between the correct relativistic calculation and the now very inappropriate classical
trajectory is at least 500%, too large an error for this classical calculation to be good for anything.
Classical
Relativistic

Perhaps even more dramatic are the differences in calculated velocities. The relativistic calculation has
the speed of the electron approaching the speed of light. However the classical calculation has the
electron speed at 10 times the speed of light!

Classical
Relativistic
Classical
Relativistic
Gamma
Einsteins theory of relativity supersedes Newtonian mechanics. However, that does not mean
that all mechanics problems need to be solved relativistically. Newtons second law still holds:
dP
F
dt
=
, however the definition of the momentum for a particle is changed to

2
1
mv
P
v
c
=
| |
|
\ .

, where c

is the speed of light in a vacuum. This new factor
2
1
1
v
c
| |
|
\ .
occurs so frequently in
the theory of special relativity that it is given its own symbol, gamma
2
1
( )
1
v
v
c

| |
|
\ .
. The
first thing to do when you come across a new function is to examine its behavior. For speeds
low in comparison to the speed of light,
8
3 x 10 m/s c ~ , we can neglect the / v c term and then
0
lim 1
v
c
= . This just returns us to the Newtonian definition for momentum, P mv = . At the

other extreme, as the speed of the particle approaches c , diverges to infinity and clearly
Newtonian mechanics is not valid. So at what intermediate speeds does one need to begin
worrying about relativistic effects? Of course, the answer depends on how accurately you need
to know the answer. But first lets graph ( / ) v c .

So for small velocities, gamma is indeed close to 1.
0
x
0
x x
x
0
( ) f x
0 0
( ) ( ) f x f x x
0
( ) f x x

To do better than this we can use a Taylor series expansion to more closely approximate gamma.
Recall that a function can be expanded around any point by an infinite series containing
derivatives of the function evaluated at the point that is being expanded around. Mathematically
this is expressed as:
2 3
1 1 1
( ) ( ) ( ) ( ) ( ) ... ( ) ...
2 6 !
n n
f x f x f x f x f x f x
n

where the behavior of f in the neighbor of x is to be approximated and delta defines the range
of the approximation.

Lets take a look at this polynomial approximation term by term. In the figure below the red line
is a sketch of the function to be approximated, the point of interest,
0
x x , is shown in red and
the point that is being expanded about,
0
x , is black. The zeroth order approximation to the
function is
0 0
( ) ( ) f x x f x . This just says that near the point of interest the function can be
approximated as a constant, shown in blue. The error made by this approximation is the vertical
distance between the correct value, red dot and the approximation blue dot. This might be a good
enough approximation for
some situations, but we do
better.

The next term in the series
is:
0
0
( )
x
df
f x x x f
dx
.
This is the slope of the
function multiplied by the
distance away for the
selected point. With the
addition of this term in the
approximation, now
0 0 0
( ) ( ) ( ) f x x f x f x x

, this approximates the
function as a line with
non-zero slope, show in
green. The value for this
approximation is the green dot and the error is the difference between the red dot and the green.
The error is reduced and clearly the green line is a better approximation to the function than is
the blue one.

Lets examine a couple common functions. First find an approximation for exp( ) x where x is
near 3. The expansion for the function is:
2 3
0 0 0
3
1 1
exp( ) exp(3) exp(3)( ) exp(3)( ) exp(3)( )
2 6
x
x x x x x x x

Below are graphs of the zeroth (orange), first (green), second (red) and third (blue) order
approximation around x = 3. The true function is shown in black, but is so well fit by the third
order curve that you cannot see it.

The error decreases as the order of the approximation is increased, same colors as above.

A similar expansion is shown below for sin(x) around x = 1. The second order approximation is
shown in red. Typically in physics the second order expansion is used. The reason for this will
become apparent later in the semester.

Returning to the problem at hand, the understanding of
2
1
( )
1
v
v
c

| |
|
\ .
, in this case we are
concerned about low speeds, or
0
0 x = , and
2
v
c
o
| |
=
|
\ .
, therefore we need to find an
approximation for
1
( )
1
f o
o
=

Using the expression above
2 3
1 1 1
( ) (0) (0) (0) (0) ... (0)
2 6 !
n n
f f f f f f
n
o o o o o ' '' ''' = + + + + +
Lets look at the expression term by term. The first term says that ( ) (0) f f o ~ and since
(0) 1 f = this is just the simplest approximation 1 ~ . This Newtonian approximation will often
be good enough, but how much better can we do by adding the next term in the expansion?
Look at ( ) 1 (0) f f o o ' = + , to evaluate this we need the derivative
( ) 1
1
df d
d d
o
o o
o
| |
=
|
\ .
(Note
that
2
v
c
o
| |
=
|
\ .
so the radical is of (1 o ) not (
2
1 o )). Performing the derivative gives
( )
3
2
( ) 1 1 1
2
1 1
df d
d d
o
o o
o o
| |
| |
|
= =
|
|
\ .
\ .
and we need this evaluated at 0 o = and so
0
( ) 1
2
df
d
o
o
o
=
= . Finally
2
1 1
( ) 1 or ( ) 1
2 2
v
f v
c
o o
| |
= + ~ +
|
\ .
this is the next level of
approximation for gamma. I leave it to you to show that the third level of approximation is
2 4
1 3
( ) 1
2 8
v v
v
c c
| | | |
~ + +
| |
\ . \ .
. Lets plot the Newtonian gamma (red), the first (orange) and second
(green) approximations and gamma (black) on the same graph.

You can see that the first and second approximations bend upward and match the true function
for higher speeds better than 1 ~ . If we plot the difference between the approximations and
gamma we can more closely examine the error. Using the same color scheme the graph is
below.

The red curve is the difference between the Newtonian approximation and gamma. The
horizontal line is at 1% error. You can see that up to 4% of the speed of light the error is very
small. Four percent of the speed of light is ~1 x 10
7
m/s, extremely fast for macroscopic objects.
The graph below better shows the error for such low speeds. The 1 approximation is in red.
The error is small, but look how much better the Taylor series expansion does.

This is why for most everyday objects Newtonian mechanics is sufficient. However, as the
speed increases toward c the error starts to grow rapidly. The first order approximation (orange)
reduces this error considerable out until ~0.2c and the second order terms extend the
approximations out to ~0.4c.
Lets look at one last point. Remember that we used the relativistic expression
2
( 1) KE mc = .
What does this reduce to when we use the low velocity expression for ? Clearly, 1 ~ is not
useful, as it just say that the 0 KE ~ . To do better we can use
2
1
( ) 1
2
v
v
c
| |
~ +
|
\ .
this means the
kinetic energy is
2
2 2
1 1
(1 1)
2 2
v
KE mc mv
c
| |
= + =
|
\ .
. This is just the familiar classical expression
for kinetic energy. It is very reassuring to be able to show that the Newtonian kinetic energy is
just the low velocity limit of the relativistic expression.

# Plots gamma, several approximations to gamma and their absolute errors



gamma_function = gdisplay( title = 'gamma vs. speed/c',
xtitle = 'v/c',
ytitle = 'gamma',
background= color.white) #set display #1

error = gdisplay( x=0, y=400, #plot the next display down from the top
title = 'error vs. speed/c',
xtitle = 'v/c', ymax =.05,
ytitle = 'gamma - approx_gamma (%)',

gamma_func = gcurve(gdisplay = gamma_function, color = color.black) # curve for gamma
one = gcurve(gdisplay = gamma_function, color = color.red) # Newtonian approx gamma = 1
approx_func_2 = gcurve(gdisplay = gamma_function, color = color.orange) # curve for gamma = 1 + (v/c)**2
approx_func_4 = gcurve(gdisplay = gamma_function, color = color.green) # curve for gamma = 1 + (v/c)**2 + 0.375(v/c)**3
approx_func_0 = gcurve(gdisplay = error, color = color.red) # curve for Newtonian i.e. gamma =
delta_gamma = gcurve(gdisplay = error, color = color.orange) # gamma - (1+(v/c)**2) in percent
delta_gamma_2 = gcurve(gdisplay = error, color = color.green) # gamma - (1+(v/c)**2 + 0.375(v/c)**3) in percent
one_percent = gcurve(gdisplay = error, color = color.black) # draw line at 1% for error

v = 0 #scaled speed in terms of c i.e. v/c
dv = .005 #graphing increment

while (v < 1):
#for v in arange (0, 0.999, dv): # Could use for loop instead of while

gamma = 1/sqrt(1-v**2)
approx_gamma = 1 + 0.5*(v**2)
approx_gamma_4 = 1 + 0.5*(v**2)+ 3/8*(v**4)

gamma_func.plot (pos=(v, gamma )) # plot real gamma
one.plot (pos=(v, 1 )) # plot gamma = 1
approx_func_2.plot (pos=(v,approx_gamma)) # plot gamma = 1 + .5(v/c)**2
approx_func_4.plot (pos=(v,approx_gamma_4)) # plot gamma = 1 + .5(v/c)**2 + .375(v/c)**2

if v > .2: # limit the range of the second plot to .2c
pass # do nothing, not interested in this range for second graph
else:
delta_gamma.plot (pos = (v, 100*(gamma - approx_gamma)/gamma)) #difference in percent
approx_func_0.plot (pos = (v, 100*(gamma - 1)/gamma)) #differece between classical and special relativity
one_percent.plot (pos = (v, 1)) #1% error line
delta_gamma_2.plot (pos = (v, 100*(gamma - approx_gamma_4)/gamma))
v = v + dv #advance to next point on graphs

Force and dynamics with a spring, numerical approach
It may strike you as strange that the first force we will discuss will be that of a spring. It is not
one of the four Universal forces and we dont use springs every day. Or do we? The basic
motion of a spring is to oscillate. Many, many mechanical systems oscillate. However, they
might not appear to be springs. Take a tree branch bouncing up and down in the wind, is it a
spring? How about the vibrations between two atoms in a molecule or a cork bobbing up and
down on the surface of water, are these spring systems? From a physics point of view, all of
these can be modeled by the same equations as those that describe a spring. This is why
examining spring systems are so important. Later in the class we will make this mathematically
precise, but for now just remember that the study of springs is far, far more general and
important than a coil of metal with a mass hanging from it.
The force law for an ideal spring is
0
( ) F k L L k L . This says that the force that a spring
applies is proportional to the amount that the spring is stretched from its equilibrium length. The
constant of proportionality is negative, meaning that the force is directed in the opposite
direction from the extensive or compressive displacement L . Lets look at the simplest spring
system:

In this figure the mass is displaced to the right and the spring exerts a restoring force to the left.
Writing Newtons equation for motion gives
0
( )
dP
k x L
dt
, combining with
( )
( )
dx t
v t
dt
and
nonrelativistically ( ) ( ) p t mv t gives a coupled set of first order differential equations.
We have turned the physical problem of a spring-mass system, through Newtons equations of
motion, into a mathematical problem, solving a coupled set of differential equations. Now we
need to solve the equations to learn about the behavior of the mass-spring system. There are a
number of ways to solve differential equations; in this section we will solve the problem
numerically.
Since we know the force law, the equations can be integrated simply. The algorithm looks like
this:

0
L
0
L
L
L

F

Initialize needed variables
Initialize graphics
Main loop
Calculate Forces on object(s)
Force_total equals sum of forces on object(s)
Calculate new momenta by adding impulse
mom = mom + (F
total
)*(dt)
Calculate new position(s) using momentum
pos = pos + (mom/mass)*(dt)
time = time + dt
display results

Analyze results, often with aid of graphics

The code that we wrote together in class looked like this; the physics is highlighted in yellow:


#Simulation of a spring mass system, no gravity

mass = 1 # mass of object
k = 100 # spring constant
L_0 = vector(0,0,0) # relaxed position of the spring

#graphics
x=0, y=0,

x=0, y=450,

x=0, y=0,

acc_vs_time = gdots(gdisplay=acceleration, color=color.black) # a graphics curve for acceleration vs. time

initial_vel = vector(0,0,0)


#main loop
while (time < 10): # simulate time < 10

F_s = -k*(position - L_0) # force produced by spring, a vector
momentum = momentum + (F_s)*dt # note using total force in momentum update
position = position + momentum/mass * dt # position update

#graphics
accel = F_s/mass # calculate acceleration


If we want to include the force of gravity there are only a couple of changes, highlighted in red:

#Simulation of a spring mass system with gravity

mass = 1 # mass of object
k = 100 # spring constant
L_0 = vector(0,0,0) # relaxed position of the spring
g = vector(0,-9.8,0) # acceleration near Earth

#graphics
x=0, y=0,

x=0, y=450,

x=450, y=450,

acc_vs_time = gdots(gdisplay=acceleration, color=color.black) # a graphics curve for acceleration vs. time

initial_vel = vector(0,0,0)


#main loop
while (time < 10): # simulate time < 100

Force_grav = mass*g # constant force applied by gravity field, a vector
F_s = -k*(position - L_0) # force produced by spring, a vector
F_net = F_s + Force_grav
momentum = momentum + F_net*dt # note using total force in momentum update
position = position + momentum/mass * dt # position update

accel = F_net/mass # calculate acceleration


That is all that needs to be changed!
Using the initial conditions of dropping the mass from the equilibrium position of the spring
yields this trajectory.

The mass oscillates about the position where the spring would be extended with the mass
hanging at rest,
mg
x
k
for the conditions used m = 1, k = 100, yields 1 x .
Graphing position, velocity and acceleration vs time shows the phase relationship between the
three. Note that the amplitudes are arbitrary.

The velocity (cyan) leads the position by
2
and the acceleration (red) is out of phase by .

So we are able to integrate Newtons law for a spring and produce oscillatory motion. Note that
no where have we used the sine or cosine functions; the oscillations emerge naturally from the
differential equations.

Force and dynamics with a spring, analytic approach
It may strike you as strange that the first force we will discuss will be that of a spring. It is not
one of the four Universal forces and we dont use springs every day. Or do we? The basic
motion of a spring is to oscillate. Many, many mechanical systems oscillate. However, they
might not appear to be springs. Take a tree branch bouncing up and down in the wind, is it a
spring? How about the vibrations between two atoms in a molecule or a cork bobbing up and
down on the surface of water, are these spring systems? From a physics point of view, all of
these can be modeled by the same equations as those that describe a spring. This is why
examining spring systems is so important. Later in the class we will make this mathematically
precise, but for now just remember that the study of springs is far, far more general and
important than a coil of metal with a mass hanging from it.
The force law for an ideal spring is
0
( ) F k L L k L . This says that the force that a spring
applies is proportional to the amount that the spring is stretched from its equilibrium length. The
constant of proportionality is negative, meaning that the force is directed in the opposite
direction from the extensive or compressive displacement L . Lets look at the simplest spring
system:

In this figure the mass is displaced to the right and the spring exerts a restoring force to the left.
Writing Newtons equation for motion in the lateral direction gives
0
( )
x
dP
k x L
dt
,
combining with
( )
( )
dx t
v t
dt
and nonrelativistically ( ) ( ) p t mv t gives
2
0 2
( )
( )
d x t k
x L
dt m
.
This is a linear second-order inhomogeneous differential equation with constant coefficients (that
is a mouth-full), with a harmonic solution.

We have turned the physical problem of a spring, through Newtons equations of motion, into a
mathematical problem: solving a differential equation. Now we need to solve the equation to
learn about the behavior of the mass-spring system. There are a number of ways to solve a
differential equation; one is simply guessing the solution. This is perfectly acceptable, but often
difficult. The equation before us,
2
0 2
( )
( )
d x t k
x L
dt m
, is quite simple. We can guess one
solution right away,
0
( ) x t L . This is a correct solution, however pretty boring. The mass is
0
L
0
L
L
L

F
simply sitting at the equilibrium position of the spring and not moving. To find more interesting
solutions we need to consider the equation that removes that particular solution:
2
2
( )
( )
d x t k
x t
dt m
= . This equation says that after taking two derivatives of the function you get
the function back multiplied by a constant. This leads us to consider a function that when
differentiated remains proportional to itself: namely the exponential ( )
t
x t Ae
o
= . Rewriting the
equation as
2
2
( )
( ) 0
d x t k
x t
dt m
+ = and substituting in the exponential gives
2
0
k
m
o + = . This is
called the secular equation. The solutions are
k
m
o = . Now, this is an interesting situation
because both k and m are positive, but lets proceed. So we have two independent solutions for
the problem
1 1
( )
k
i t
m
x t Ae = and
2 2
( )
k
i t
m
x t A e
= . The general solution is the sum of these, plus

the particular solution
0
( ) x t L = . Specifically the general solution is
1 2 0
( )
k k
i t i t
m m
x t Ae A e L
= + + .
This equation has two adjustable constants
1
A and
2
A . These can be determined by using the
initial conditions.

So, we have analytically solved the mathematical problem. The only issue is that the solution is
complex and certainly not the simple oscillatory motion that we expected. Lets start by
renaming the quantity
k
m
e . e has the units of inverse time and is called the angular
frequency. We need to understand what the complex exponential
i t
e
e
means. The best way to
do this is expand the function in a Taylor series.
2 3 4 5 6 7 8
2 3 4 5 6 7 8
1 ...
2! 3! 4! 5! 6! 7! 8!
i t
e i t t i t t i t t i t t
e
e e e e e e e
e = + + + + +
This infinite series does not look much better. It is still complex and contain an infinite number
of terms. Lets simplify a bit by collecting real and imaginary terms:
2 4 6 8 3 5 7
2 4 6 8 3 5 7
1 ... ...
2! 4! 6! 8! 3! 5! 7!
i t
e t t t t i t t t t
e
e e e e e e e
e
| | | |
= + + + + + +
| |
\ . \ .
or
( ) ( )
2 2 1
0 0
2 ! (2 1)!
n n
i t
n n
t t
e i
n n
e
e e
+

= =
= +
+

Again, this might not seem to improve the situation until you realize that those two summations
are the Taylor series expansions for the cosine and sine function. Go ahead and try it yourself,
expand the cosine and sine function and find that
( )
2
2 4 6 8
2 4 6 8
0
cos( ) 1 ...
2! 4! 6! 8! 2 !
n
n
t
t t t t t
n
e
e e e e
e

=
= + + + =

and
( )
( )
2 1
3 5 7
3 5 7
0
sin( ) ...
3! 5! 7! 2 1 !
n
n
t
t t t t t
n
e
e e e
e e
+
=
= + + =
+
.
This simplifies things a bit and produces one of the most amazing equations in mathematics,
namely Eulers relation: cos( ) sin( )
i t
e t i t
e
e e = + . Now this is progress: we have produced the
common oscillatory functions that we expected. However, it is still a complex expression. That
can be fixed. Look at the similar expression for the other solution cos( ) sin( )
i t
e t i t
e
e e
= , here
we have used the fact that cosine is an even function, and sine is odd. By adding these two
solutions together we get cos( )
2
i t i t
e e
t
e e
e

+
= . This is still a solution to the original problem.
Similarly subtraction yields sin( )
2
i t i t
e e
t
i
e e
e

= . Having now done all of this we have

reformed our original complex solution
1 2 0
( )
k k
i t i t
m m
x t Ae A e L
= + + into a new form

0
( ) sin( ) cos( )
k k
x t A t B t L
m m
= + + that contain only real functions ( A and B could still be
complex if we wanted, typically we dont).

Often the solution is written in a second form that contains a phase angle. This change is simply
a trigonometric identity.
0 0
( ) sin cos or ( ) sin
k k k
x t A t B t L x t C t L
m m m

| | | | | |
= + + = + +
| | |
| | |
\ . \ . \ .

These two expressions for ( ) x t are equivalent in that there are two free parameters in each , A B
in the first and , C in the second and the function is periodic with frequency
1
2
k
f
m t
= . The
two free parameters can be determined from two additional conditions such as the initial position
and initial velocity.

There is a relationship between the parameters in each expression so that you can go from one to
the other as suits your needs:
2 2 1
and tan , alternately cos and sin
B
C A B A C B C
A

| |
= + = = =
|
\ .
. This can be
remembered from basic trigonometry, see the figure.

A

For now we will use
0
( ) cos
k
x t C t L
m

| |
= + +
|
|
\ .

The constant
0
L is a reflection of where the origin is placed. If it is placed at the equilibrium
location of the end of the spring,
0
L will be zero. This is a common way to set up your
coordinate system. The amplitude of oscillation is given by C , and the range of y varies from
C to C + . The term inside the parenthesizes,
k
t
m
+ , is called the phase. There is a time
independent term , called the phase shift,initial phase, or phase constant, that determines where
in the cycle the motion starts at time zero. In other words, it determines whether the object starts
at the equilibrium position and has a nonzero momentum or at a point of maximum displacement
with zero momentum, or anywhere in between these positions. To determine C and you must
be given two pieces of information about the system, such as from what position did it start and
how fast was it going initially. This will allow you to uniquely determine these constants.
So what about the time dependent part of the phase,
k
t
m
? This term determines the frequency
of oscillation. Every time this part of the phase increases by 2t the motion has completed one
A

B
C
more full cycle. That means that the period T for an oscillation is given by 2
k
T
m
or
2
m
T
k
. If that is the time per cycle, the number of cycles per second (Hz), the frequency f
, can be found from
1 1
2
k
f
T m
. A closely related quantity, the angular frequency ,
gives the oscillation rate in radians per second and is defined by 2
k
f
m
.
Much of the behavior of the oscillator can be understood from this figure:

where ( ) t t . Picture a vector rotating with a constant rate, i.e. the second hand on a
clock, only this rotates counter-clockwise. The position of the oscillator is given by the x-
component of the vector, i.e. where the dashed line on the figure intersects the x-axis. The
vector rotates at a rate radians per second and the position of the oscillator ranges from C to
C . This is called harmonic motion. On a position vs. time plot the motion looks like a sine
curve:
C
C

( ) t

( ) x t

The parameters that were used are
3
1, 1,
4
C

The program at the end of this section has an animation with a rotating vector and the position of
the y-coordinate tracing out oscillations of harmonic motion.

# Plots sine curve and rotating vector representation for harmonic motion


sine_function = gdisplay( title = 'harmonic position vs. time',
xtitle = 'time',
ytitle = 'position',
x=0, y=400, width=1000,

sine_func = gdots(gdisplay = sine_function, color = color.black) # curve for sine

dt = .1 #graphing increment
C = 1 # C is amplitude of oscillation
w = 1 # angular frequency
phi = 3*3.14/4 # phase shift, set initial conditions

#graphics
vector = arrow(pos=(0,0,0), axis=(C*cos(phi),C*sin(phi),0), shaftwidth=.01)
position = sphere(pos = (C*cos(phi),0,0), radius = .05, color = color.red)

for t in arange (0, 20, dt): # time points from 0 to 10 interval dt
rate(10)
sine = C*cos(w*t + phi) # harmonic function
sine_func.plot (pos=(t, sine )) # plot harmonic function
vector.axis = (C*cos(w*t + phi),C*sin(w*t + phi),0) # rotating arrow
position.pos = (C*cos(w*t + phi),0,0) # x position of arrow
Until now weve had quite a bit of luxury in making our step size smaller to get more accuracy. If were
calculating a gravitation field though, the numbers can get large very quickly. Think of it this way. For
each point in space we need to calculate the field. So we will need an index to run over x, y, and z to
cover a grid in space where we want to calculate the field. Now for each point in space we will need to
calculate the vector sum where we run over x, y;, and z over the mass. Lets rewrite our integral so it
goes over a grid of cubes.
g(r) = -
0p [r

r - r
JxJyJz
(r -r)
2
v

This is way of dividing up space using Cartesian coordinates is often useful and makes numerical
implementation easier. It is, however, not the only way of dividing up a volume into small parts.
Sometimes its better to divide it up into concentric spherical shells, and then subdivide each shell up
using an a pair of angles like longitude and latitude. This method uses spherical coordinates. Or
sometimes its best to divide space up into concentric cylindrical shells. Each cylinder can be divided up
into one angular slice and one linear slice along the axis of the cylinder. Naturally enough, this is the
method of using cylindrical coordinates. There are several (but a finite number) of other, more
complicated ways of dividing up space, but the three mentioned here are most common.

http://www.math.byu.edu/~math302/content/learningmod/coordinates/
Calculating the volume of those elements (in the limit of small size) is easiest for rectangular coordinates.
Let implement this numerically. As an example, suppose we have a slab 1x4x9 m
3
. We learn from 2001:
A Space Odyssey that these dimensions are needed for teleportation.
Here is vpython code to calculate the field from this object:

from visual.graph import *

vscale = 1.0*10**7 # scales g vectors size

# size of cube: see 2001 a space odessey
x_box = 1.0
y_box = 4.0
z_box = 9.0

# primed variables step in monolith This will make a 10x10x10 grid
dx_p = 0.1
dy_p = 0.4
dz_p = 0.9

density = 1000.0 # This is waters density in kg/m^3 (water)
G = 6.67*10**(-11) # Newtons universal G

dx=dy=dz = 4 # sets the density of points where field is calcualted

xyz_limit = 20 # calculate field in cube of dimesion xyz_limit^3

# define monolith
shape = box(pos=(0,0,0), size=(x_box,y_box,z_box), color=color.green,opacity=0.5)

# step over space to calculate
x = -xyz_limit/2
while x < xyz_limit/2:
x = x + dx
y = -xyz_limit/2
while y < xyz_limit/2:
y = y + dy
z = -xyz_limit/2
while z < xyz_limit/2:
z = z + dz

# calculate g at this position g = sum(over primed) dg
g = vector(0,0,0) # initalize g

x_p = -x_box/2
while x_p < x_box/2:
x_p = x_p + dx_p
y_p = -y_box/2
while y_p < y_box/2:
y_p = y_p + dy_p
z_p = -z_box/2
while z_p < z_box/2:
z_p = z_p + dz_p

# physics here
r_i = vector(x,y,z)-vector(x_p,y_p,z_p)
dg = -G*density*(dx_p*dy_p*dz_p)*norm(r_i)/mag2(r_i)
g = g + dg

# make arrow
arrow(pos=vector(x,y,z), axis=vscale*g, color=color.yellow)

The first 3 while statements step over a 5x5x5 grid (20/4) of points where the field is calculated. The
second set of 3 while statements step over the primed coordinates a 10x10x10 grid containing the
monolith. Notice how this is not a fast calculation even though our grid is coarse. The three physics
statements execute:
r
i
= r - r
= (x, y, z) - (x
, y
, z
)

Ag = -
0pJx'Jy'Jz'
r
i
2
r
i

g = Ag
i

Try running it and see what you get.
Rotate the picture and think about the field. Do the arrows point toward the center of mass? Only at
some points. Generally not. Notice the symmetry: if you make a reflection about the xy, xz, or yz plane
you get the same answer. If you zoom out it goes looks like the field from a point mass: The arrows
point toward the center and the magnitude decreases like 1/r
2
.

Let also work out an analytical example. Lets say you have a long thin rod, mass M and length L
and you want to know the field (or force) on a mass along the bisector. A picture is below:

We already know how to calculate the field from a point mass. A good starting point is to ask how to
divide up the mass into small masses. A clue that youre choosing a good way to divide things up is that
you can name the limits pretty easily. Here we can divide the rod into segments. Pick a radom segment
somewhere in the mass and calculate the small gravitational field due to the small mass.
Jg
= -
0pr Jm
r
2

The mass dm is related to the length Jm = [
M
L
Jx. We can say g = ]Jg
, but we cant write g = ]Jg.

You must integrate the way you add, and you cant add magnitudes, but you can add components. From
the symmetry theres only a net y-component. g
= ]Jg
= ]-|Jg
|cos0. Putting this all together

gives:
g
= _ -
0H
I
cos0Jx
r
2
L2
-L2

Anything that changes with x must be expressed in terms of x. This gives:
g
= _ -
0H
I
yJx
(x
2
+ y
2
)
32
L2
-L2

Look it up. (Or try substitution.) The answer is:

g
= -
0H
y(I2)
2
+ y
2

In the limit where youre far from the rod you recover Newtons law of gravity for a point mass.
Keplerian Orbits
If two otherwise isolated particles interact through a
2
1
r
force law their trajectories can be
reduced to conic sections. This is called Keplers problem after Johannes Kepler who studied
planetary motion. If one of the particles is so massive that it can be approximated as fixed, then
the other particles motion will be a conic section. The trajectories can be hyperbolic, parabolic,
and elliptical. Linear and circular cases are special examples of these orbits. These trajectories
can be classified as either open or closed by whether the particle visits infinity and therefore
never returns (because the force of interaction has vanished). On an open orbit, parabolic and
hyperbolic, the particles reach the point of closest approach only once and then separate forever
after, i.e. one traversal; whereas for a closed orbit, elliptical, the motion is periodic repeating
over and over identically.
It is interesting to note that there are only two mathematical types of forces that produce such
motion that exactly repeats over-and-over. They are the linear Hookes law and the inverse
square force law. As such, these are often considered the two most important problems in all of
classical mechanics. All other forms of interaction force produce trajectories that either dont
trace out the same path (though dont separate infinitely) or just move apart forever.
Lets take a look at a comet entering our solar system (assume mass comet << mass

sun). At
some position r it has a momentum p . The force that is exerted on the comet by the sun is
, , 2
,
C S C S
C S
Mm
F G r
r
. It is very useful to decompose this force into components parallel ( ) p and
perpendicular ( ) n to the comets momentum (shown in red):

F
on C by S
F
parallel
F
perpendicular
S

C

p
n
p

The F
parallel
( ) F causes a change in the speed of the Comet, while F
perpendicular
( ) F
causes a
change in direction of the momentum. The osculatory or kissing circle gives the radius of
curvature at the location C. It is the circle that best approximates the curvature of the
trajectory at point C.
The comet momentum is defined as p p p , taking the derivative with respect to time using
the product rule gives,

d p
dp dp
p p
dt dt dt

Therefore equating the parallel and perpendicular components of Newtons equation yields,
linear motion) and (curvature)

d p
dp
F p F p
dt dt

So just what is
dp
dt
, the change in time of the unit vector of momentum?
Here is a diagram:

p p

speed direction
Osculatory or kissing circle

R
R

i
p

f
p

v t

Consider the comet at two times, initial i and final f where
f i
t t t . The comet moves along
the trajectory, shown in heavy black, in this interval by an amount v t (red) and the unit vectors
for the momentum change by p =
f
p -
i
p . Using the fact that the momentum unit vector does
not change in magnitude (it cant since it is a unit vector) then if we find a circle that
approximates the local trajectory we can construct similar triangles as shown below.

From the similar triangles,

or
1
p v t p v
R t R

. So that describes the change in

magnitude of
dp
dt
, but in what direction? In order for a vector to change its direction and not its
length, the change must take place in a perpendicular direction. In this case it is in the inward
radial direction of the osculatory, n . The limit as t goes to zero shows that
v
dp
n
dt R
.
Substituting into the expressions for the change in the momentum
and
d p
dp
F p F p
dt dt
, we can now use the nonrelativistic momentum to find,

2
linear motion) and = m = m

d v v
dp v
F m p F p v n n
dt dt R R
or
2
v
F m n
R
(curvature), where R is the radius of the osculatory circle and n the radially inward
unit vector directed normal to the trajectory or equivalently to the center of the osculatory.
R
R
v t

i
p

f
p

p

The radius R of the osculatory

is the local radius of curvature of the trajectory. If the orbital
equation is known then the radius of curvature can be found from (you will learn this in Calc III):
3
2 2
2
2
1
dy
dx
R
d y
dx
(
| |
+
(
|
\ .
(

= or
3
2 2 2
2 2
2 2
dx dy
dt dt
R
dx d y dy d x
dt dt dt dt
| |
| | | |
+
|
| |
|
\ . \ .
\ .
=
| | | |
| | | |
| | | |
\ . \ .
\ . \ .
, where ( ) x x t = and ( ) y y t =

F
on C by S
F
parallel
F
perpendicular
S

C

Osculatory or kissing circle

Radius (R) of curvature at C

p
n
p

So now we have the condition for motion in a closed elliptical orbit,
2
v
F m n
R
. This is a very
important expression to understand. In order for the object to be traveling in a closed elliptical
orbit the component of the net force on the object perpendicular to the trajectory must be equal to
2
v
m
R
.
There is an important special case that comes up often, that of a circular orbit. This is a limiting
case of an ellipse with zero eccentricity with both foci at the center of the circle. In this case the
trajectory and the osculatory are one in the same. The local curvature of a circle is always its
radius, so R is the radius of the trajectory.

F
on C by S
= F
perpendicular

F
parallel
S

C

Trajectory and Osculatory are the
same for circular motion

Radius (R) is the radius of curvature at C

p
n
p

Since
d v
F m p
dt
, the speed of the comet is constant. The speed is determined by
2
v
F F m n
R
, the radially inward force sets a relationship that must be followed by the
comet to move in a circular orbit. Of course, we know what the force due to gravity is between
these masses,
, 2
C S
Mm
F G n
R
equating this with the force needed to keep a circular orbit yields,
2
2

v Mm GM
m n G n v
R R R
. Thus the speed of an object in circular motion due to
gravitational attraction is set by the ratio of the mass of the stationary center to the radius of the
orbit. Note this is only true for circular orbits.

Changing topics and leaving the comet, lets consider a constant mass rocket that is launched
into space off the moon but does not escape and falls back to the surface. The moon has no
atmosphere, so we dont need to worry about air resistance. What type of trajectory should the
rocket have? Our high school physics prepared you well to say a parabolic trajectory, no air
resistance right? Not necessarily. If this is a closed orbit it must be an elliptical trajectory. Look
at the diagram below,

We know from the shell theorem that the gravitational force will act on the rocket as if all the
mass of the moon is concentrated at its center, figure on the left. This is clearly an elliptical orbit
for all the reasons that we have discussed above. The shape of the ellipse is determined by the
initial momentum of the rocket and the mass and radius of the moon.
Of course, the moon is not really a point particle; it only acts that way for masses above its
surface. When the rocket comes around it crashes back into the surface, as in the right hand
figure.
So the trajectory can be elliptical. Then why might you think that it should be parabolic?
Because in high school you learned that the force of gravity was uniform, giving a parabolic
path, and did not consider trajectories where the height was a significant fraction of the size of
the massive object. In fact, as the rocket moves away from the moon the gravitational force
decreases. If you make the approximation that gravity is a constant force F mg then your
solution is parabolic, but this is a poor approximation for long or high trajectories.

So lets put some of this theory to work. You are traveling on in a spaceship to mars through the
asteroid belt. You need to stop to make some repairs and spot a convenient asteroid, R2D2.
From your great distance you can only measure its rough diameter 121400. The density of this
type of rock is approximately 5000 kg/m
3
. Before landing you need to get in a circular orbit with
a radius of 120 km. What is the speed that you need to have in this circular orbit?

As you draw closer to R2D2 you realize that your spherical approximation is very crude; with
better resolution you determine that R2D2 actually looks like this:

How is the irregular shape of the asteroid going to affect your orbit?

You want to simulate what your orbit will be around R2D2. How will you go about determining
the gravitational attraction that you will feel at each point? Write down the algorithm
(procedure) to calculate this.
Conservation of momentum
So far we have been dealing with one particle systems that are experiencing external forces.
Examples include the mass on a spring oscillating, skydiver falling or moon in orbit. In each of
these cases there was a force external to the system that was responsible for the change in
momentum of the object.
In this class we are going to shift from one particle systems experiencing an external force to a
multiparticle system that is isolated from external forces. A good example of this is a binary
star system. Such star systems are actually much more common than isolated stars such as our
Sun. In any case, there are two objects interacting through a force, gravity. What can be said in
general about the system?
Lets start with a free-body diagram:

There are two forces in the system: the force that is placed on m by M, F
mM
and the force that is
placed on M by m, F
Mm
. These are equal and opposite forces, as can seen from the gravitational
force law, or more generally from Newtons third law. Given these forces the momentum of
each mass must be changing.
m
mM
dp
F
dt
and
M
Mm
dp
F
dt
. If the forces are known then it is
easy to find the motion of the bodies. However, sometimes the interaction force is not known, or
so complicated, that the equations of motion cannot be determined directly. In these situations it
is extremely useful to consider the total momentum of the system,
total m M
P p p . Looking at
the time rate of change of the total momentum gives:

0
total m M
mM Mm
dP dp dp
F F
dt dt dt
. This
is a very important result: the total momentum is constant. This is true because the system is
isolated and all the forces come as Newtons third law reaction pairs. When you sum up the
changes in the individual momenta they cancel. This derivation can trivially be extended to any
number of particles.
The law of conservation of linear momentum states that for an isolated system the total
momentum is a constant in time. Stated mathematically the total momentum defined as
m
M
F
Mm
F
mM
total i i
i i
P p mv = =

satisfies 0
total
dP
dt
= . This is true for any isolated system, with any
number of particles, containing any forces. It is this generality that makes the law of
Conservation of Linear Momentum so powerful.

The thrust of a rocket is an excellent example of how to use the law of conservation of
momentum for an isolated system. Consider an isolated (deep space) rocket of mass M, at time t,
moving with velocity v, as shown in the figure on the left. An instant of time t A later the
situation looks like figure on the right.

Since the system is isolated we know that 0
total
dP
dt
= this means that for a short time interval t A
that we can write 0
t t t
P P P
t t
+A
A
= =
A A
where
t t
P
+A
is the momentum for the system on the right
and
t
P is the momentum for the system on the left. Examining the figures shows that
( )( )
t t
P M M v v Mu
+A
= A +A +A and
t
P Mv = . Using conversation of momentum gives
| | | | ( )( )
0
t t t
M M v v Mu Mv P P Mv M v Mv M v Mu Mv
t t t
+A
A + A + A + A A A A + A
= = =
A A A
.
Taking the lim
t A
, dropping the term M v A A that is proportional to
2
( ) t A , and rearranging leads to
( )
dv dM
M u v
dt dt
= where
M dM
t dt
A
=
A

because the change in mass with respect to time is negative. The right-hand term only depends
on the rockets present properties, the left-hand allows us to predict where the rocket will be in
the future. Lets examine the right-hand side more closely.
dM
dt
is the negative of the mass of
the gas per second being exhausted out the back of the rocket. This is controlled by design of the
rocket. What about the term ( ) u v ? Looking at the figures above, this is just the velocity of
the exhaust gas relative to the rocket. If riding on the rocket you see gas moving backwards
Time t
Time t t +A
M
v
M-M
v+v
M
u
t
P
t t
P
+A

away from you with this velocity; it is a vector that points to the left in the figure above. You
should draw it in.
We will define
rel
v u v , this is the exhaust velocity of the gas and is also a designed property
of the rocket, via nozzle size etc. With this definition the equation becomes,
rel
dv dM
M v
dt dt
The two terms on the right are properties of the rocket and the product has
the dimensions of a force. This product is called the rocket thrust. The left hand side is the
mass times the acceleration of the rocket. This equation is often called the first rocket equation.
Lets take your homework problem, in it the thrust is 100 N and assuming a linear burn rate, or
loss of mass,
0.3
0.06
5sec
dM kg
kg s
dt
. From this the exhaust velocity of the gas can be
determined from
thrust
rel
F
v
dM
dt
= 1667 m/s. Typically, a rocket engineer designs an engine
to have an initial mass, burn rate, exhaust velocity and burn time. With these parameters she can
impart a force over a time (an impulse) to give the rocket the desired momentum.

A little more can be done to analyze the rocket equation,
rel
dv dM
M v
dt dt
. Multiply each side by
dt , rearranging the mass terms so they are all on the right-hand side yields:
final final
initial initial
t t
rel
t t
dM
dv v
M

and then integrate to produce the second rocket equation.
( ) ( ) ln ln
f
i
final initial rel rel
i f
M
M
v t v t v v
M M
This allows you to find the velocity change of a
rocket only knowing the exhaust velocity and the initial and final masses. The time of the burn
does not enter into the equation. Therefore small engines that burn a long time are equivalent to
large engines that only burn briefly. This result is true for rockets not experiencing gravitational
forces. For a rocket taking off from Earth this result is not correct. In that situation you want to
get rid of as much mass as possible, as quickly as possible. You will explore this a bit in part 2d
of your homework.

Center of momentum frame
Newtons laws of motion hold in any inertial frame of reference. However, it is often simplest to
examine an isolated system of particles from the frame of reference where the total momentum is
zero. Since the system is isolated, the total momentum is constant and in this Center of
Momentum frame remains zero as the particles interact with one another.
Consider the simplest isolated system containing a single particle. This particle will have a
constant momentum in any inertial frame, however if you select the frame that moves along with
the particle, the particle will have momentum equal to zero. Cant get simpler than that.
Things are considerably more interesting when the isolated system contains two interacting
particles. There are two different situations, a bound or an unbound system. Lets first consider
an unbound system. Here the particles are going to eventually end up well separated and no
longer interact with one another. If they also start far apart then you can consider this as a
collision. A collision starts with the particles isolated and having constant individual momenta,
then the particles interact and momentum is transferred, finally the particles separate and have
constant individual momenta. A common situation is where one of the particles is initially
stationary in this laboratory frame of reference The incoming particle collides with a fixed
target. Such a collision typically has a trajectory as shown below:
This is a collision between a proton and
a He nucleous. The He nucleous is
initially at rest and the proton has an
initial velocity of 0.05c. As the proton
approaches the target, its trajectory is
deflected. At the same time the
nucleous begins to accellerate away.
The black line shows the trajectory of
the center of mass (CM). The total
momentum of the isolated system is
constant and therefore the CM moves
to the right with a constant velocity
throughout the entire time. The
velocity of the CM is given by
total
CM
total
P
V
m
. It must be emphasized
that the CM moves before, during and
after the colision. The diagram of the
momentum before and after the collision is shown below.

Now we want to change our frame of
reference. Imagine moving along with the
CM at
CM
V . In this frame the He nucleus
moves towards the CM from the right, while
the proton moves to the left. Of course, the
CM (small black dot in figure) remains
stationary in this frame, the so called Center
of Momentum frame of reference. From this
frame of reference the same collision as above
is shown on the right.

The diagrams of the momenta before and after the collision are shown above left. The total
momentum in this frame of reference is zero and the collision can be viewed as a rotation of the
individual momenta. This is an elastic collision so the magnitudes of the momentum vectors are
unchanged. If energy had been lost in the collision the momenta would be equal and opposite,
but reduced in magnitude from the initial value.
One approach to solving this scattering problem is to start in the Lab frame, transform to the
Center of Momentum (CoM) frame, solve the interaction for changes in momenta of the two
particles and then transform back to the Lab frame to produce the desired result. This procedure
is done because it is so much simpler to solve the scattering in the CoM frame, it is a rotation of
the two momenta, with a possible change in magnitude for if the collision is inelastic. So the
question then becomes how does one transform to the CoM frame and back?
For an isolated system the total momentum
1 1 2 2
P mv m v is a constant. This is true in any
frame of reference. To transform between the Lab frame and the CoM from you need to know
initial proton
P
final proton
P
final He
P
final proton
P
final He
P
initial proton
P
initial He
P
the velocity of the CM. This can be found by taking the derivitive with respect to time of the
CM position,
1 1 2 2
1 2
CM
mr m r
R
m m
. Differentiating yeilds
1 1 2 2
1 2 1 2
CM
mv m v P
V
m m m m

. To find
the velocities of particles one and two in the CoM frame, we subtract the CM velocity from the
particles Lab velocity,
1, 1, CM Lab CM
v v V and similarly for particle two. Substituting in the
expression for
CM
V results in

2
1, 1, 2,
1 2
CM Lab Lab
m
v v v
m m

and

1
2, 2, 1,
1 2
CM Lab Lab
m
v v v
m m

.
From these we can see that the momenta of the particles in the CoM frame are

1 2
1, 1, 2,
1 2
CM Lab Lab
mm
p v v
m m

and

1 2
2, 2, 1,
1 2
CM Lab Lab
mm
p v v
m m

. Clearly these are opposite

vectors and so the total momentum in the CoM frame is indeed zero.

If you need to move from the CoM frame to the Lab reference, then you just invert these
transformations,
1, 1, Lab CM CM
v v V and
2, 2, Lab CM CM
v v V .

In the previous discussion we were dealing with two particles acting through a central force, a
force that depends on the particle separation and act along the line between them. Newtons
equations for this system are:
1
( )
dp
F r
dt
and
2
( )
dp
F r
dt
. Using the nonreletivistic
expression of momentum, we have
2
1
1 2
( )
d r
m F r
dt
and
2
2
2 2
( )
d r
m F r
dt
. Multipling each
equation by the other mass and subtracting them yeilds:
2 2
1 2
1 2 1 2 1 2 2 2
( ) ( )
d r d r
mm mm m m F r
dt dt
.
Simplifing provides a very interesting and important equation:

2
1 2
1 2
2
1 2
( )
d r r
mm
F r
m m dt
. Defining
1 2
r r r and
1 2
1 2
mm
m m

and noting that r r yeilds,

2
2
( )
d r
F r
dt
. This is the equation for a single partile with a mass acted on by the same
force as in the original two particle problem. The two particle problem has been reduced to a
one particle problem with an effective mass, called the reduced mass at position
1 2
r r r . So
what has happened to the other particle? There is still a second particle in the problem,
however its motion does not depend on the force ( ) F r . How can that be? The second particle is
located at the CM and acts with a mass given by
1 2
M m m , the total mass. This particle can
indeed be in motion, however for an isolated system it moves with a constant velocity
CM
V as
described previously, independent of F .
This simplification is very important, we have taken a two-body problem with both
particles being acted upon by a force and transformed it into two separate one-body
problems with only one of these being acted on by a force. This is an enormous reduction
in complexity. We can solve the one-particle problem of the reduced mass particle,
combine that with the constant velocity of the CM and produce the solution for the two
interacting particles.
The mathematical procedure follows this line. For an isolated system the total momentum
1 1 2 2
P mv m v is a constant. The other quantity of interest is the relative position,
1 2
r r r ,
between the two particles. Taking the derivitive of the relative position gives
1 2
v v v . We
can now solve for
1
v and
2
v in terms of the total momentum P and the relative velocity v . This
system of two equations and two unknowns yeilds,
2
1
1 2 1 2
m P
v v
m m m m

and
1
2
1 2 1 2
m P
v v
m m m m

. It convienent to express this in terms
of the reduced mass
1 2
1 2
mm
m m

and the velocty of the CM,

1 1 2 1
1 2 1 2
CM
mv m v P
V
m m m m

.
Rewriting the expressions for the individual particle velocities in terms of the reduced mass and
velocity of the CM gives:
1
1
CM
v V v
m
and
2
2
CM
v V v
m
or in terms of momenta
1 1 CM
p mV v and
2 2 CM
p mV v . The right hand side of these equations use the effective single particle
problem and CM, while the left hand side is the velocity of the two real particles. Therefore
these equations describe how to go from the motion of the effective one partiticle problem to that
of the original two particles.
The dynamics of the effective single particle problem are governed by ( )
dv
F r
dt
. From this
you can find ( ) v t ,and combined with the motion of the CM, the velocities of the real particles
can be found.
A few trajectories will help illustrate the concepts. Consider the binary star system in your
homework. You are in a nonaceelerating rocket (lab frame of reference
CM
v V ) observing the
motion of the stars. You would see something like this.

The stars trace out quite complicated
orbits. Now (having taken Physics
160) you know that things are simpler
when considered from the CoM
frame and so you change course and
match your velocity with that of
white point on the diagram. This is
the CM and it is tracing out a straight
path with constant velocity.
Remember that this is just a point in
space, there is actually no mass
located there.

From your new
reference frame the
orbits of the two stars
look quite different.
First, they appear to
be elipses with
periodic orbits. It
was difficult before to
identify an orbital
period because the
stars did not return to
the same point.
However, now the
system is clearly
periodic. Note that
the CM is a stationary
spot as it should be in
the CoM frame of reference.
If we look at the trajectory of
1 2
r r r = , shown left in white, we
see that it also traces out an elipse.
This is not an accident, but a
reflection of the fact that this
problem can be reduced to an
effective single particle problem
and this reduced mass particle will
trace out an orbit equal to that of
1 2
r r r = .

Consider the single particle problem ( )
dv
F r
dt
= . The interaction force is between the reduced
mass particle and the CM. The force is given by
( )
1 2
1 2 1 2
2 2 2
1 2
( )
m m
mm mm M
F r G r G r G r
r m m r r
+ | |
= = =
|
+
\ .
, the same force of interaction as
between the real particles with masses
1
m
and
2
m . The initial position and
momentum of the reduced-mass particle
can be determined from the initial
positions and momenta of the original
particles. Specifically,
, ,1 ,2 init init init
r r r
=
and
2 ,1 1 ,2
,
1 2
init init
init init
m p m p
p v
m m

= =
+
.
With these and the dynamical equation the
trajectory of the orbit can be determined
and is shown left. This orange orbit is the
same as the white one shown in the
previous figure. In this figure the white
dot is the location of the fixed CM and is
at the one of the foci of the ellipse. This is not the case in the previous figure. In that figure the
CM is stationary, but does not rest at a focus of either of the red orbital ellipses.
The simplification of the two-body problem to a single particle with an effective mass and the
constant motion of the center-of-mass is an important concept in mechanics. We will see this
effective mass reduction return very soon for oscillating systems.

Contact Forces So far we have not dealt with contact forces, i.e. normal, tension, etc. Consider
this very simple situation to illustrate the key issues. Two blocks in contact are in deep space,
i.e. gravity plays no role; a force is applied as shown in the figure.

As in all force problems the first thing to do is draw free body diagrams for each mass.

From these we can write Newtons equations of motion for each mass:
1 2
1,2 2,1

A
dp dp
F F F
dt dt
, where of course
1,2 2,1
F F as a Newtons third reaction pair.
Using the nonrelativistic momentum these equations become:
2,1 2,1
1 2
1 2

A
F F F
dv dv
dt m dt m
. Since the two blocks must move together,

1 2

dv dv
dt dt
and
therefore
2,1 2,1
1 2
A
F F F
m m
. Solving this for

2,1
F yields the result that we are looking for:
2
2,1
1 2
A
m
F F
m m

This result should strike you as quite strange as the contact force is proportional to the applied
force. How does the right side of block 1 know what is happening at the left side? If the
applied force is increased, then the contact force increases mysteriously. This is certainly
different from the four Universal Laws. This force does not depend on mass or charge or any
physical property of block 1 or 2, very strange indeed. If your use of the Normal force has not
bothered you before, it should now.
To explain this mystery, consider a slightly different situation. Place a massless spring between
the two blocks.

A
F
2
1
2,1
F
A
F
1,2
F
A
F
2
1
Now the free body diagrams look like this:

Where the subscript S indicates that the spring is applying the force. We know from Hooks law
that
1, 2, S S
F F k l , with k the spring constant and l the distance the spring is compressed.
Using the result for the contact force from above,
2 2
1 2 1 2

A
A
m m F
k l F l
m m m m k

.
This gives a result that is not mysterious, i.e. we can understand it. The applied force
compresses the spring by an amount l and the spring applies the equal and opposite forces on
the two blocks. The only problem with this picture is that we do not have a spring between the
two blocks. Or do we? What spring could exist between the blocks? The atoms and molecules
themselves!
Atoms are typically arranged in a regular way, with space between them. If a
force is applied the spacing changes, just as with a spring. To a good first
approximation the change in spacing is proportional to the applied force, also
just like a spring.
Having the picture in your mind of atoms connected by springs is a very nice model for many
phenomena concerning solids. Take a look at the cover of Matter & Interaction textbook. This
is a nice rendition of the model.
So just what is the spring stiffness between atoms, i.e in the atomic bonds? Lets use copper as
an example. One mole of Cu is 6.02x10
23
atoms and has a mass of 64 grams and a density of
8.94 g/cm
3
. First we need to find the volume of a mole of Cu atoms,
3
mole Cu 3
64 g
7.159 cm
8.94 g/cm
V . From this, finding the volume of a single Cu atom gives
24
11.89 x 10
atom
V

cm
3
. The distance between atoms can be approximated by viewing this
volume as a cube and finding the length of an edge,
3
atom atom
L V = 22.8 x 10
-9
cm or 2.28 .
This is a typical spacing between many types of atoms in most all crystals. So from this we
know the equilibrium length,
0
L , of the bonds between the Cu atoms.
But, what is the spring constant k in Newtons/meter? It turns out that this is a standard property
to measure for a solid. The quantity that is measured is called the Youngs modulus and it
relates the change in length of a solid when it experiences a compressive or tensile force. The
definition is
stress
strain
F
A
Y
L
L

where
force
stress
cross-sectional area
or the pressure and the
A
F
1,S
F
2,S
F
strain
L
L
is the dimensionless fractional change in the length along the force. Using Hooks
law
F
L
k
, Youngs modulus can be rewritten as
Lk AY
Y k
A L
. So lets
consider the little cubic atom
The length of a side is
0
L (= 2.28 ), so
0
A
L
L
and thus
0
k L Y . Youngs modulus can be
found from experiment and is tabulated for many materials. For Cu it ranges from 110 to128 x
10
9
N/m
2
. (The unit of pressure is a Pascale (Pa) or N/m
2
and the natural scale for elemental
Youngs modulii is a Giga-Pascale or GPa).
Evaluating the spring constant for a Cu atom,
10 11
2 Cu atom
(2.3 10 )(1.2 10 ) 28
N N
k x m x
m m
.
This is a weak spring, 28 N, or 6 lbs, can extent the spring by one meter.

So why is Cu not very compressible, equivalently why is Y so large? If you push on a block, it
does not seems to compress much. But, yet the atomic bonds are soft. Consider a large block of
Cu containing many atoms.

How much will the block extend when forces are applied? We know how much an individual
atom will deform, now how about a collection of atoms? There are two important cases to
understand: Springs in series and springs in parallel.

F F
L
F

F
L
L
F

F

2L
2 L
F

F

2
L

Looking at the springs in series, the applied force F acts on each spring. This can be confusing
at first. Think about one of the two end springs. The external force on it is F . Since the spring
is in equilibrium it must have zero net force on it. Therefore the neighboring spring must be
placing a force of F on the terminal spring. But this means now, by Newtons third law, that
the next spring is experiencing a force F . And so the process continues. Each spring in the
series experiences a force F and extends an amount L . So the total extension of the
combined length of strings in series extends by N L , with N being the number of springs. So
with an applied force of F generating an extension of N L ,
individual
N in series
k F
k
N L N

Springs in series form a combined softer spring, i.e. it extends more for the same applied force.

Conversely, springs in parallel form a stiffer combined spring. In this case the force is split
among the N springs and the force each spring experiences is
F
N
so the extension of each is
L
N
.
This is also the extension of the combined springs in parallel. Therefore the effective spring
constant is
N in parallel individual
F
k Nk
L
N
The spring constant for a parallel combination of

springs is increased by the number of individual component springs.

These parallel and series combinations basically explain why a block of atoms is so stiff. The
number of springs in the length of the block grows linearly, and they make the block softer than
an individual atom. However, the number of atoms in the cross section of the block grows
quadratically, and they make the block stiffer. Consider a 1 cm sized block; the number of atoms
in series is ~10
8
, whereas the number in parallel is ~10
16
. This vast difference in the number of
springs in parallel that stiffen the material vs. those in series that soften, is the reason that Cu
blocks are not very compressible.
So for the entire block
Block
F
k
L
. Since
Block
k grows linearly with A and decreases as
1
L
, to
take the geometry out of the problem, consider
0
atom
Block
F
k L
A
k Y
L
A L
L

. So Youngs modulus
is a material property that does not depend on the geometry of the object, but is determined by
the atomic bond strength and the bond length.

Work
Work is the change in energy of a system (neglecting heat transfer). To examine what could
cause work, lets look at the dimensions of energy:
| | | |
2
2
L ML
E M L F L
T T
| | | |
= = =
| |
\ . \ .
so
dimensionally energy is equal to a force times a length. Consider a constant force F on a block.
The block generically could be displaced in three
ways. (This displacement could be due to other
forces or an initial momentum.) What is the work
done by F on the block? Work is a scalar and so
the function must take two vectors, F and d , and
produce a scalar. It turns out that the correct way to
do this is to look at parallel components of F and
d . Work is defined as
x y z
W F dx F dy F dz = + + or
W F d = . This is correct only for a constant force. Looking at the diagram above, what is the
work done during a displacement
R
d ? It is simply the product
R
Fd . The force is doing positive
work on the block. What about for a displacement
L
d ? In this case the force and displacement
are antiparallel and the work is
L
Fd . The force reduces the energy of the block, i.e. it would
slow the block down and reduce the kinetic energy. How about for a displacement that is
perpendicular to the force? Here the work done is zero. The force does not change the energy of
the block as it moves along this perpendicular path.
There is an important special case that has this perpendicular relationship between the force and
displacement. Consider your circular orbit and the arrows that you drew. The force (or
acceleration) was radially inward while the velocity (instantaneous displacement d v t A = A ) is
tangent to the circle and therefore perpendicular to the force. This means the work done by this
force is zero.
Another case were the work is zero is when there is no displacement of the object. An
interesting case is that of jumping off the floor. What is the work done by the floor on your
body? Clearly the floor places a force on your body, but yet the floor does not move so the work
done by the floor is zero. This may seem strange and you should ask where you get your energy
to jump up when the floor does no work. We will return to this interesting question in a couple
of weeks.
The last case with constant forces to consider is a displacement at an
arbitrary angle to the force. Using the definition for the dot product
cos W F d Fd u = = .
F
R
d
L
d
U
d
F
d
u
Now the general case for non-constant forces along an arbitrary trajectory needs to be examined.

The red arrows show the force that a particle experiences along a trajectory shown in black. If
the trajectory is broken up into many small segments as shown in the figure below, then these
segments will be approximately straight and the work done is cos
i i i i
i i
W F r F r u = A = A

where
i
u is the angle between the force and the small displacement at location i .

Of course, we want to take the limit of this so that r dr A and the sum becomes an integral, a
line integral. Taking this limit, the general expression for the work done by a non-constant force
along an arbitrary trajectory is
f f f
i i i
f x y z
x y z
i x y z
W F dr F dx F dy F dz = + +
} } } }
. You will learn how
to evaluate this integral in your class on calculus of several variables. Right now you could
evaluate this using numerical techniques.
Lets find the work done by a variable force from a spring, where
0
( ) F k x x x = . The
displacement is along the x-axis. Here is the situation:

0
x
i
x
f
x
dr F
The work done by the spring on the block as it moves from
i
x to
f
x is given by:
2 2 2
0 0 0
1 1
( ) ( ) ( )
2 2
f
f
f
i
i
i
x
x
x
f i f i
x
x
x
W k x x x xdr kx kx x k x x kx x x
. You need to be
careful when doing this calculation to make sure that F dr has the correct sign. Quite often it is
convenient to place the origin of the coordinate system at the point of equilibrium for the spring,
i.e.
0
0 x . With this choice the work simplifies to
2 2
1
( )
2
f i
W k x x . Note that in this case
the force that the spring does on the block is negative. It is removing kinetic energy from the
block. Eventually, the block will come to rest and begin to move back towards
0
x ; during this
part of the cycle the spring will do positive work on the block. If a complete cycle is considered
the spring will perform zero net work and the block will return to the original point with exactly
the same kinetic energy as it had the cycle before.

Do example problem

Potential Energy
Up till now we have been considering single particle systems. Once there are multiple particles
in the system there can be interactions between the particles. Consider this scenario with two
particles that are isolated from any other surroundings:

Initially the particles are at rest and some time later their kinetic energy has increased. Where
did the energy come from? The particles are isolated so no work can be done by external forces.
Clearly the energy came from the particles interacting with one another. The energy is called
potential energy it depends on the relative positions of the two particles. As the kinetic energy
increases, the potential energy decreases such that the total energy is constant. Mathematically
this is expressed,
t
E U KE = + where U is the potential energy of the interaction. The general
case is represented below:

Consider particle 1. The work done on this particle by the applied forces is
1 1, 1, 1 1,int 1,
or E
i ext ext
i ext
E F F dr W W
| | | | | |
A = + A = +
| | |
\ . \ . \ .

. Now do this for all the particles and
sum
1 2 3 1,int 2,int 3,int 1, 2, 3, int i ext ext ext ext
i
E E E E W W W W W W W W A = A + A + A = + + + + + = +
therefore
1 2 3 int ext
E E E W W A +A +A = . Defining
int
U W A , this becomes
0 KE =
0 KE >
1
3,ext
F
3
2
2,ext
F
1,ext
F
1,2
F
3,1
F
2,3
F
3,2
F
1,3
F
2,1
F
1 2 3 t ext
E E E E U W . So the change in the total energy of the system is equal to
the work done by external forces. A total energy can be defined now as
t i
i
E E U
writing
this out explicitly
2 2 2
1 2 3 1 2 3 t
E mc m c m c K K K U . For an isolated system 0
ext
W
and so 0
t
E or
t
E is constant. This is the very important statement of conservation of energy
for an isolated system. It is one of the three conservation laws that you will learn this term.

A change in potential energy U has been introduced as the negative of the work done by
internal forces, specifically
f
i
U W F dr
. Not all forces can have their work done

expressed as a change in potential energy. In order for a force to have a potential energy
function a couple of things must be satisfied. The first is path independence, see the figure
below.

In order for a force to have a potential function the change in potential energy going from point i
to point f must be the same whether path Aor path B is taken. Mathematically this means
, f i
A
U F dr
must be equal to
, f i
B
U F dr
. In other words the work done by the force

must not depend on the path taken. The only way for this to be true is if
1,2 1,2
U U ,
independent of the path taken. So for an object that takes a complete round trip, the work done
by the force must be zero,
,
0
i i
U F dr
. If this path independence holds, then the force

is called conservative. Some examples of each type of force:
Conservative:
Universal law of Gravitation
1 2
2
G
mm
F G r
r

Coulombs Law
1 2
2
C
q q
F k r
r
.
A
B
f
i
Hookes law
0
( ) F k r r =
Uniform gravity F mgr =
Nonconservative:
Air drag
2
1
2
F ACv v =
Fluid drag F vv =

If mechanical energy is changed to heat energy during the motion, then the force will not be
conservative.

The explicit change in potential energy function for conservative forces can be determined from
f
i
U F dr A =
}
. We know that this will be path independent and so the simplest path to
evaluate the integral is usually choosen. Consider the case for a uniform gravitational field
where F mgy = . If an object is moved from one height to another there will be a change in
potential energy given by
f f
g f i
i i
U mgy dr mgdy mgy mgy A = = =
} }
. Note that since the
path does not matter,
f i f i
U U U mgy mgy A = = . Comparing terms a potential energy
function for a uniform gravitational field can be identified as ( ) U y mgy C = + . The constant C
is arbitrary and can be adjusted to suit the situation. This is equivalent to choosing
0
y in
0
( ) ( ) U y mg y y = , which is a choice of the level where 0 U = Its selection will not affect the
results of any calculation because the only physically meaningful quantity is a change in
potential energy U A .

The same procedure can be done for the Universal law of Gravitation, where
1 2
2
f
G
i
mm
U G r dr
r
A =
}
. Consider a radial path moving from
i
r to
f
r , then this reduces to
1 2 1 2 2
1 1
f
f
i
i
r
r
G
r
r
U Gmm dr Gmm
r r
| |
A = =
|
\ .
}
. Evaluating yields
1 2 1 2
G f i
f i
Gmm Gmm
U U U
r r
| |
| |
A = =
|
|
|
\ .
\ .
and the identification of the potential energy due to
gravity to be
1 2
( )
G
Gmm
U r C
r
| |
= +
|
\ .
. It is customary to set the gravitational potential energy at
infinity to be zero. This choice sets the arbitrary constant 0 C = and the potential energy due to
gravity becomes
1 2
( )
G
Gmm
U r
r
= . It may concern you that this function is always negative.

When I was a student this always bothered me. But remember that only changes in potential
energy are physically relevant. The changes in potential energy from this function can be either
positive or negative. This will be explored in more detail in the next lecture.
The potential energy for the Coulomb interaction is of exactly the same form as for gravity:
1 2
( )
C
kq q
U r
r
| |
=
|
\ .
. This can be either positive or negative because the charge interactions can be
either repulsive or attractive.

The last potential energy function that we will find is that for a spring or Hookes law. What is
the change in potential energy of the spring when the block moves from
i
x to
f
x ?

The calculation proceeds very similarly to the one we did for finding the work:
2 2 2
0 0 0 0
1 1
( ) ( ) ( ) ( )
2 2
f
f f
f
i
i i
i
x
x x
x
f i f i
x
x x
x
U k x x x dr k x x dx kx kx x k x x kx x x A = = = =
} }

This expression is not in the form
G f i
U U U A = . To put it in the proper form group the initial
and final terms together, and complete the square by adding and subtracting
2
0
2
x
k . Then
2 2 2 2 2 2
0 0 0 0 0
1 1 1
( ) ( ) ( 2 ) ( 2 )
2 2 2
f i f i f f i i
U k x x kx x x k x kx x x k x kx x x A = = + +
2 2
0 0
1 1
( ) ( )
2 2
f i
U k x x k x x A =
0
x
i
x
f
x
dr F
This is now in the proper form
G f i
U U U , so that the potential energy can be identified as
2
0
1
( ) ( )
2
S
U x k x x C
If the origin of the coordinate system is chosen such that
0
0 x , then it can be written as
2
1
( )
2
s
U x kx , which is probably familiar from a previous physics class.

So in summary, the potential energy for these conservative forces is

Conservative forces and their potential energies:
Law of Gravitation
1 2
2
G
mm
F G r
r

1 2
( )
G
mm
U r G
r

Coulombs Law
1 2
2
C
q q
F k r
r

1 2
( )
C
q q
U r k
r

Hookes law
0
( )
S
F k x x x
2
0
1
( ) ( )
2
S
U x k x x
Uniform gravity
g
F mgy
0
( ) ( )
g
U y mg y y

It is important to note that although the forces are vector expressions, the potential energy is a
scalar, only depending on the location of the particle. Again only changes in potential energy are
physically meaningful. Equivalently, each potential energy must be referenced to a point. For
the
1
r
potentials this is typically infinity where conventionally the potential energy is set to zero.
It is instructive to consider the potential energy of a mass near the surface of a planet. The
gravitational potential energy is ( )
G
mM
U r R G
R h

where h is the height above the surface

of the planet and is much smaller than R . This can be rewritten as
( )
1
G
mM mM
U r R G G
h R h
R
R
= =
+ | |
+
|
\ .
. Letting 1
h
x
R
= this becomes
( )
1
1
G
GmM
U
R x
=
+
.
So what is the approximation for
1
( )
1
f x
x
=
+
for small x ? Using a Taylor series expansion,
2
0
1
( ) (0) (0) ... = 1+ ... 1
(1 )
f x f f x x x
x
' ~ + + + ~
+
. So the potential energy
2
(1 )
G
GmM h GmM GmM
U h
R R R R
~ = + . The first term is a constant and so it is irrelevant,
the second term is linear in the height above the planet and the mass of the object. This is what
we are looking for. Rewriting yields ( ) ( )
G g
U h U h mgh ~ = where
2
GM
g
R
= . This is the same
identification of the local acceleration due to gravity g with the near-surface expression from the
Universal law of gravitation.

Determination of Force from Potential Energy
We have learned how to find a potential energy function (if it exists) from a force. But, if given
a potential energy ( ) U r how do you determine the force ( ) F r ? Again you will learn more
about the mathematics of this is in your Calc III class, but since the potential energy is the
negative integral of the force, you might expect that the force would be determined by taking the
negative derivative of the potential. This is correct, except for the force being a vector quantity
and the potential energy a scalar.
What is needed is a 3-dimensional derivative. This exists and is called the gradient, it is denoted
by the symbol , ,
x y z
| | c c c
V
|
c c c
\ .
. The force acting on a particle at position r with a potential
energy ( ) U r is ( ) ( ) F r U r = V , which means that components of force are given by partial
derivatives of U . For example,
( )
( )
x
U r
F r
x
c
=
c
. This operation is not difficult to perform.
As an example consider the general 2-dimensional spring potential energy given by
( )
2 2
1
( )
2
x y
U r k x k y = + . The force is given by
( )
( ) ( ) ( )
2 2 2 2 2 2
2 2
1 1
( ) , ,
2 2
x y x y x y
x y
k x k y k x k y k x k y
F r k x k y
x y z
| |
c + c + c +
| |
| = V + =
|
|
c c c
\ .
\ .
. The
symbol
x
c
c
is called the partial derivative. It means take the derivative with respect to x while
considering y and z as constants, so
( )
2 2
2
x y
x
k x k y
k x
x
c +
=
c
,
( )
2 2
2
x y
y
k x k y
k y
y
c +
=
c
and
( )
2 2
0
x y
k x k y
z
c +
=
c
. Therefore, the force field generated by the potential
( )
2 2
1
( )
2
x y
U r k x k y = +
is ( ) ( , , 0)
x y
F r k x k y = .

Lets consider a second example, the potential energy due to gravity is
1 2
( )
G
m m
U r G
r
= , how
do we find the force associated with this? First we will do it the long way. The radius is
2 2 2
r x y z = + + , so
1 2
2 2 2
( , , )
Gm m
F x y z
x y z
| |
| = V
|
+ +
\ .
must be computed. Explicitly:
1 2
2 2 2 2 2 2 2 2 2
1 1 1
( , , ) , , F x y z Gm m
x y z
x y z x y z x y z
| | | | | | | |
c c c
| | | | =
| | | | c c c
+ + + + + +
\ . \ . \ . \ .
.
This is not as bad as it looks! Take it one component at a time:
( )
3
2 2 2
2 2 2
2
1 x
x
x y z
x y z
| |
c
| =
|
c
+ +
\ .
+ +
. The other components are exactly the same so the force
becomes
( ) ( ) ( )
1 2 3 3 3
2 2 2 2 2 2 2 2 2
2 2 2
( , , ) , ,
x y z
F x y z Gm m
x y z x y z x y z
| |
|
=
|
|
+ + + + + +
\ .
. This may not
look familiar, but it is correct. Simplifying we have
( )
( )
1 2
1 2 2 2
2 2 2
2
( , , )
( , , )
Gm m x y z
F x y z
x y z
x y z
| |
|
=
|
+ +
|
+ +
\ .

and using the definition of r this becomes
1 2
2
( )
Gm m
F r r
r
= , as it has to be.

I mentioned that this was the long way to solve this problem. The situation that we considered,
the gravitational potential energy from a single mass has a tremendous amount of symmetry,
namely, the spherical symmetry. Obviously, you would like to work in a coordinate system that
has this symmetry, and the Cartesian (x,y,z) coordinate system does not do this. The proper
coordinate system is the spherical system ( , , ) r u , defined in the figure below. This
representation exploits the fact that there is only a radial dependence in the problem and thus the
potential energy and force only depend on r .

The coordinate transformation is:
2 2 2
1
1
2 2 2
tan
cos
r x y z
y
x
z
x y z
u
= + +
| |
=
|
\ .
| |
| =
|
+ +
\ .
and the inverse
cos sin
sin sin
cos
x r
y r
z r
u
u
=
=
=

Now the important point for this discussion is that the gradient in spherical coordinates takes a
very convenient form for our problem:
1 1

sin
spherical
r
r r r
u
u
c c c
V = + +
c c c
. This is far more natural to use than
y z
Cartesian
x
x y z
c c c
V = + +
c c c
. Explicitly,
1 2
2
1 1

( ) ( )
sin
spherical
Gm m
F r U r r
r r r r
u
u
| | c c c
= V = + +
|
c c c
\ .
. Now since there is no
and u dependence in the problem, last two of these terms vanish! You are left with
1 2
( )
Gm m
F r r
r r
c | |
=
|
c
\ .
. This is a simple derivative of one variable ( r does not change with ) r
and it is directly found that
1 2
2
( )
Gmm
F r r
r
. This is much simpler than using the Cartesian
coordinate system.
As you go on farther in physics and mathematics, the discussion of symmetry will emerge as one
of the most important aspects. This is not only due to the computational advantages shown here;
but there are very deep connections between the symmetries of a problem and the conserved
quantities for that physical situation.

Potential energy diagrams
Consider an arbitrary potential energy shown schematically below.

There are a number of important qualitative features of the behavior of the system that can be
determined by just knowing this curve. The first thing to notice is that since the kinetic energy
of a particle must be greater than or equal to zero, the total energy of the system must be greater
than the potential energy. This means that the coordinates of a particle (position, Energy) can
never be in the hashed region. This is called the forbidden region of the diagram.

When dealing with an isolated mechanical system the total energy will be constant represented
by a horizontal line.

Energy
Position
Energy
Position
Energy
Position

The particle can only be at the regions shown in red. In the graph above there are two allowed
regions, the particle must be located somewhere in the red. Once we know where the particle is,
then you can determine the force on it and its kinetic energy.

The particle is represented by the red point. Its position is
0
r and the kinetic energy is the
difference between the total energy (red line) and the potential energy. The force on the particle
is the negative of the slope
x
dU
F
dx
. So you can determine the speed and the acceleration of
the particle, but not the direction of motion. The particle as shown above will oscillate back and
forth stopping at the turning points,
l
tp and
r
tp . This is a bound trajectory. If the particle were
in the right side region it would be in an unbound orbit, visiting the single turning point at most
once.
One last thing that you can immediately determine are the points of equilibrium. An equilibrium
point is a point where the force on the particle equals zero. In the situation above there are five
equilibrium points. Further, there are two types of equilibrium points, stable and unstable. A
particle perturbed from a point of stable equilibrium will experience a force acting to return it to
the equilibrium point. This leads to oscillatory motion around the stable equilibrium point.
The converse of this is if the particle is disturbed and experiences a force that continues to
accelerate its motion away from the point of equilibrium. Then this is an unstable equilibrium.
Stable equilibria occur at local minima of the potential energy, while unstable equilibria occur at
local maxima.
It is clear that a lot of information about the possible motions can be gleaned by simply
examining the potential energy diagram.
To continue the discussion, consider the potential energy for two particles interacting
gravitationally,
1 2
( )
G
mm
U r G
r
. The graph of this is shown below.
0
r
KE
dU
dx
r
tp
l
tp
Energy
Position
Bound
Unbound

The first thing to note is that the potential energy, as defined, is always negative, but since only
changes in potential energy are physically meaningful this is fine. Consider the particles
changing their separation from 10 to 2. Then we can read off the graph
(2) (10) ( 5) ( 1) 4 U U U A = ~ = , so the potential energy decreased. If on the other hand,
the particle separation went from 2 to 10 then (10) (2) ( 1) ( 5) 4 U U U A = ~ = and the
potential energy of the particles has increased. So both increases and decreases in potential
energy of the system can be realized.
Now consider the special case of circular motion of a light particle around a stationary massive
one. Let the radius of the orbit be r . What are the total energy, potential energy and kinetic
energy for this system? Since the motion will be much less than the speed of light we can ignore
the rest energy of the particles, then the total energy will be
1 2
mm
E KE U KE G
r
| |
= + = +
|
\ .
.
For a circular orbit we can find the kinetic energy through
2
2
1 1
1 1

2 2
G G
v
F m mv rF
r
= = .
Therefore the kinetic energy equals
2 1 2 1 2
1 2
1 1 1
2 2 2
mm mm
KE mv rG G
r r
= = = . Putting this together
for the total energy yields:
1 2 1 2 1 2
1 1
2 2
mm mm mm
E G G G
r r r
= = . The total energy is less than zero! How does this
appear on the potential energy diagram?

On this diagram the orbit of
1
m about
2
m is represented as a point with the coordinates
1 2
0 0
( , ) ( , )
2
mm
r E r G
r
. From the graph we can tell that the orbit has a fixed radius
0
r and a
constant kinetic and potential energies. The potential energy is
1 2
0
( )
mm
U r G
r
and the kinetic
energy is the difference between the total energy and the potential energy,
1 2
0
( )
2
mm
KE E U r G
r
. Note that twice the kinetic energy is equal to the negative of the
potential energy, 2KE U , this is a specific case of a very general theorem, the virial theorem
in mechanics.

In the above special case of circular motion we were able to solve for all the energies easily.
This is not often the case, and it is in these circumstances that a potential energy diagram is most
helpful. Consider an elliptical orbit, for this case the diagram looks like this:

0
( ) U r
0
r
KE
E

Again the total energy E is a constant, as the system is isolated. However this time the particle
separation varies from the perihelion (
p
r , closest approach) to the aphelion (
a
r , farthest point).
Along with this comes a variation in both the kinetic energy and the potential energy such that
their sum is always constant.

What does a rocket leaving the surface of a planet look like on such a diagram?

( )
a
U r
p
r
p
KE
E
a
r
( )
p
U r
a
KE
0
( )
i
E U r
0
r
escape
KE
0
f
E

Before blast off the kinetic energy of the rocket is zero. To escape the gravitational attraction of
the planet the rocket must move from the surface to infinity. The amount of energy needed to
move the rocket to infinity and arrive with zero velocity is called the escape energy. This
amount of energy is shown on the graph as the brace, it is defined as ( )
escape surface
KE U r . If
you want to escape the planet and have a nonzero velocity, say to go to a farther galaxy, then the
rocket must be able to supply more energy than this.
For an object of mass mto escape from the earth the kinetic energy needed is

Earth
escape Earth
earth
mM
KE G
r
. More typically people look at the escape velocity defined as
2

1
2
Earth
escape escape Earth
earth
mM
mv KE G
r
and so
2
Earth
escape
earth
GM
v
r
. Note that this velocity does not
depend on the mass of the object.
There are a couple general things that can be seen from the potential energy diagrams. If the
total energy of the system is less than ( ) 0 U then the mass cannot escape and this is called a
bound system. If the total energy of the system is zero the mass can reach infinity with zero
velocity. Lastly, if the total energy is greater than zero the system is unbound and eventually the
two particles will separate.
Moving to unbound orbits, there are two general ways to have them. The first we have already
considered, that is the KE is too large to be bound. This is represented below.

( ) U r
r
KE
E
( ) U r

The particle is above to travel to infinity. (There is a detail here that is being over looked and
that is conservation of angular momentum. In three dimensions the singularity at zero particle
separation is removed by including the effects of angular momentum.)
The other way to have an unbound orbit is to have a repulsive potential as will occur between
two charges of like sign. Such a situation is shown below.

This particle is either moving toward or away from a repulsive potential. The closest that it can
get is
tp
r , at this point the particle has zero kinetic energy and reverses its motion and accelerates
away from the turning point, never to return.
Before leaving gravitational potential energy we need to examine the case when the size of the
mass cannot be neglected. For distances less than the radius of one of the masses we know that
the force is linear in distance from the center. This gives a potential energy that is parabolic with
distance. The expression for the potential energy inside of a mass with radius R is
2
3
1 3
( )
2 2
G
GMm GMm
U r R r
R R
. This function smoothly matches with the potential energy
( )
G
GmM
U r R
r
outside of the mass. In the graph below the radius of the mass is 2 and the
red dot shows the value of the potential energy at that position.
E
KE
U
tp
r

The potential energy function can be approximated near the surface by a linear relationship. This
line is shown in red in the next plot. The equation for this line is,
2
( ) 2
G linear
GmM GmM
U r R r
R R
= + . This can be rewritten as
2 2
( ) ( 2 ) ( )
G linear
GmM GmM GmM
U r R R r r R
R R R
= + = . Now call the height above the large
mass h r R = and this becomes
2
( )
G linear
GM GmM
U h m h
R R
| |
=
|
\ .
. Remember, the local
acceleration due to gravity is
2
GM
g
R
= , then this becomes

( )
G linear
U h mgh C = + . That should
look familiar. The constant C was
GmM
R
in the equation above, but since only differences in
potential energy are physically meaningful, this constant can be adjusted set the zero of the
potential.

A graphical construction of this is shown below. The red line is a plot of the linear
approximation. The slope is
2
U mg h GmM
mg
r h R
A A
= = =
A A
. Of course, the range of validity is
much smaller than shown on the graph.

Turning now to a different situation, consider the potential energy of a harmonic oscillator where
the equilibrium point is
0
3 x .

KE
tp
r
tp
r
E
0
r
h
mg h

The total energy of the system is 35 and the two turning points are -1.1 and 7.1. The particle will
oscillate between these two extremes with the energy being transferred back and forth between
kinetic and potential. The energies are schematically shown below.

The kinetic energy is red, the potential energy is black and the total energy is shown in blue. The
kinetic and potential energies oscillate at twice the frequency of motion and so make two cycles
while the particle makes one oscillation in the potential well.
Lets look at a general potential shown below:

one period of oscillation T
0
r
E
tp
r
tp
r

The equilibrium point is
0
3 r and the total energy of the system is guaranteed to be greater than
-5. With the total energy as shown (-2.9) the turning points are at 2.5 and 4.1. Although the
motion of this particle is oscillatory, this potential is clearly not symmetric or harmonic and the
oscillations would not be sinusoidal at this amplitude. With a lower energy the oscillations
would more closely resemble those of a harmonic oscillator. This will be a topic for a future
discussion.

Small oscillations
The theory of small oscillations is an extremely important topic in mechanics. Consider a system
that has a potential energy diagram as below:

There are three points of stable equilibrium, A, B, and C. The general coordinate x can be
thought of as a position or angle. The force on a particle at position
0
x in this potential is given
by
0
x
dU
F
dx
By definition a particle located at any of the stable equilibrium points
experiences no force, and the slope of the potential energy is zero at these positions.

What happens when the particle is disturbed from one of this equilibrium points? If the particle
is pushed to the right, it experiences a restoring force to the left. The behavior should remind
you of a spring. The potential energy for a spring is given by the parabola
2
1
( )
2
spring
U x kx .
This parabolic potential does not look globally much like that pictured above. However, locally
parabolas approximate the potential near the stable equilibria.

A
B
C
c
C
U
x
A
B
C
c
C
U
x
If the best parabolas can be determined then the potential at the equilibrium point can be
approximated by that of a spring.
The question reduces to finding an approximate mathematical description of the potential energy
near equilibrium points. To do this, recall that a Taylor series expansion is such a local
approximation to a function. Specifically near a point
0
x :
2 3
0 0 0 0 0 0
1 1 1
( ) ( ) ( ) ( ) ( ) ... ( )
2 6 !
n n
f x f x f x f x f x f x
n

Consider the potential energy U near an equilibrium point
0
x . Expanding to second order
2
0 0 0 0
1
( ) ( ) ( ) ( )
2
U x U x U x U x . Consider each term,
0
( ) U x is the constant value of
the potential energy at the local minimum. Since only differences in potential energy are
physically significant this constant term is not important. The next term is
0
( ) U x and is equal
to zero. This is because at the equilibrium point the first derivative of the potential energy
vanishes. The last term
2
0
1
( )
2
U x is very interesting and important. It is the equation of a
parabola in , the displacement from the equilibrium point. This is just what we needed. Near
an equilibrium point the potential energy is a constant plus a quadratic dependence in the
displacement,

2
0 0 0
1
( ) ( ) ( )
2
U x U x U x . This is the equation for the potential energy of a
spring with a spring constant, or stiffness, equal to
0
( ) U x .

Look at the figure of the potential energy and points, A, B, and C. Although all of these are
stable equilibria, the curvature, or equivalently
0
( ) U x , varies with each. Point C has the
highest curvature, followed by A and then B, equivalently ( ) ( ) ( )
C A B
U x U x U x . Again,
this curvature plays the role of the spring stiffness k .

The angular frequency of oscillations for a spring is given by
k
m
. Therefore the angular
frequency of oscillations about a stable equilibrium point is
0
( ) U x
m

. Such harmonic
behavior is a general feature of small amplitude oscillations about any stable equilibrium
point for an arbitrary potential energy. As the amplitude of oscillations increases, the parabolic
approximation to the potential energy becomes less accurate and the behavior becomes
anharmonic (periodic, but not sinusoidal).
Lets consider a specific example, let ( ) U r r
r
where 0 r and , 0 . Dimensionally

[ ] E L and [ ]
E
L
. Always sketch the potential:

The red curves show the two parts of the potential; there is a linear attractive piece and a
1
r
repulsive part, their sum generates the total potential. First the equilibrium point must be found.
This is done by setting
0
( ) 0 U r and solving for
0
r . The derivative is given by
2
( ) U r
r
and setting equal to zero and solving yields

0
r

. Note that dimensionally

this combination yields a length. Evaluating the potential at
0
( ) U r gives the minimum value of
the potential
0
( ) 2 U r

. We can now label these on the potential energy

curve.

In order to find the frequency of small oscillations, we need to find the curvature from the second
derivative of the potential energy,
3
2
( ) U r
r
o
'' = evaluating at the equilibrium point
3
0 3 3
0
2
2 2
( ) 2 U r
r
o o |
o
o
|
'' = = =
| |
|
\ .
. Once again you
should check the units to make sure that your answer can be correct, the dimensions should be
2
E
L
. We are now ready to look at the best approximate harmonic (parabolic) potential,
( )
2
0 0 0 0
1
( ) ( ~ ) ( ) ( )
2
H
U r U r r U r U r r r '' ~ = + , for this above case
2
3
0
( ) ( ~ ) 2
H
U r U r r r
| o
o|
o |
| |
~ = +
|
|
\ .
. The plots are done with 1 o = and 1 | = and so a
plot of ( )
2
( ) 2 1
H
U r r = + will show the best harmonic approximation near the equilibrium
point.
0
r
0
( ) U r

The red curve is the harmonic approximation to the actual potential. When the total energy is
just slightly above the minimum, the potential energy is well approximated by the harmonic
potential. This is actually a common occurrence in nature and so this small amplitude oscillation
approximation is very important.
The last thing to find is the frequency of these small oscillations. Using the harmonic
approximation the spring constant k is equal to the curvature evaluated at the equilibrium point,
0
( ) U r '' , therefore the angular frequency can be found from
0
0
( ) U r k
m m
e
''
= = or
3 1
3
2 4
0 1
2
2 2
m
m
| |
e
o
o
| |
= =
|
\ .
. The frequency is
1
3
4
0
0
1
2 2
f
m
e |
t o t
| |
= =
|
\ .
.

This procedure may be conducted near any point of stable equilibrium for any potential energy
profile, independent of the nature of the underlying interactions. This provides the true context
for the importance of the harmonic oscillator problem: near stable equilibra all systems behave as
if they were simple harmonic oscillators.
0
r
0
( ) U r
Molecular Dynamics
Physics Nobel Prize winner Richard Feynman once said that "If we were to name the most powerful
assumption of all, which leads one on and on in an attempt to understand life, it is that all things are made
of atoms, and that everything that living things do can be understood in terms of the jigglings and
wigglings of atoms."
Modern physics, chemistry and biology utilize numerical simulations to explore the behavior of
molecules and solids. Molecules used to treat disease, proteins, DNA, all of these are studied with
numerical simulations. To understand atoms at the deepest level you need to use quantum mechanics.
This was a subject that was developed starting the 1920s and has been tremendously successful in
explaining atomic behavior. However, the theory is complicated and difficult to apply to systems that are
larger than 100 atoms.
So what is done is that quantum mechanical calculations are used to develop approximations or
experiments are performed to make measurements. The atoms are reduced to point masses and the
approximate potential energies are used to find the forces by taking the negative gradient. So the problem
is reduced to finding the best potential energies. This is simple in principle, but in practice, the
calculations can be very complicated with 1000s of different types of potential energies. However, the
subject is still simple in theory.
Consider the molecule below. There are two generic types of energies, the first due to interactions
between bonded atoms (those with lines between them) and the second between non-bonded atoms. The
total energy of the molecule can be written
as
total bonded unbonded
E E E . The bonding
energy occurs between nearest neighbor
atoms. This is why it is so much easier to
deal with this energy, because there are a
very limited number of such interactions.
There are various types of bonded
interactions
bonded bond angle dihedral
E E E E . We will
discuss them one at a time.
bond
E is dependent on the distance between
the centers of two atoms. The simplest case
is a diatomic molecule, such as O
2
. A
commonly used approximation for the
energy of this bond is the Morse potential.
A mathematical form is
0
( ) 2
( ) (1 )
r r
Morse M M
U r E e E

. In this equation
0
r sets the position of the equilibrium separation,
M
E the depth of the potential well, and the curvature at the minimum i.e. the stiffness of the bond and
frequency of vibrations. Below is a graph of this potential with two different values of , but the same
0
r and
M
E

Taking the negative derivative of the function gives the force which is used to compute the motion. The
problem with this is that the Morse potential is time consuming to compute numerically. Computers are
relatively slow at determining the exponetiation of a value. Since this computation will be performed
literally billions of times, we need to find a faster method. This is where the theory of small oscillations
come into play.
We can find the curvature of the harmonic approximation be taking two derivatives of the Morse potential
0
2
2
2
2
m
r
d U
k E
dr
. Then it is easily to write the best parabolic approximation to the potential as

2 2
0
( ) ( ) 1
harmonic m
U r E r r . This is shown below in blue, while the Morse potential is in red.
0.75 black
1.00 red

To determine if this approximation is good enough, we need to know the amplitude of the vibrations. If
they are small enough then the harmonic potential will be satisfactory. The way to determine this is to
compare the energy of vibration to the thermal energy at the temperature of the simulation. I will just
state the answer below. Those of you that have had some chemistry may recognize the thermal energy
given by
B
k T . A typical soft molecular spring constant 50 k = N/m was used. At room temperature
( )
2
19
1
(0.025 )(1.6 10 / )
2
B
k r k T eV x J eV
A = ~ so
11
2
10
B
k T
r m
k

A ~ ~ or 0.1. This is a very
small amplitude 0.1 r A = . The small circle shown around the potential minimum gives the scale of the
oscillations in the bond length. So using the harmonic approximation
( )
2 2
0
( ) ( ) 1
harmonic m
U r E r r o =
will be quite accurate and will speed up the calculations many-fold. The force from this potential energy
is
( )
2 2 2
0 0
( ) 1 2 ( )
m m
F U r E r r E r r
r
o o
c
= V = =
c
. This is a spring with a stiffness
bond
k
given by
2
2
m
E o , in agreement with the potential energy above.

Turning from vibrations of the bond length, now we need to consider changes in bond angle. The origin
of this force can only be determined by a complicated quantum mechanical calculation. Scientists dont
typically even try to find such a realistic force, they just jump right to the harmonic approximation given
by, ( )
2
0
1
( )
2
angle
U k
u
u u u = . The effective spring constant is determined by experiment.
One very common molecule that has important bond angle oscillations is water, H
2
O. The equilibrium
bond angle is 104.45 . The figure below is a ball and stick representation of a water molecule

Morse potential
(red)
Harmonic
potential
(blue)
0
r

Experimental measurements have determined that the scissor motion of these two bonds occurs at 20GHz
or 20 10
9
cycles/sec. From this we can estimate the spring constant that should be used,
( )
2
27 9 2 5
0 0
1
2 1.7 10 ((2 )(20 10 )) ~ 5 10 /
2
k
f k m f N m
m
t t
t

~ ~ = . This is much
weaker than the vibrational bond and will be excited thermally much more readily. To be used in the
potential energy ( )
2
0
1
( )
2
angle
U k
u
u u u = This must be changed to an angular spring constant, the radius
is ~1 so
1
x
u
A
A = and
15 2
5 10 J/rad
bond
k x

~ .

The last type of motion for bonded atoms is the rotation of a group of atoms along a bond. Again it is
often too complicated to study exactly so an empirical form for the potential energy is used,
{ }
1
( ) 1 cos( )
2
dihedral
U k n
= + . This is really completely an approximate form that only takes into

account the symmetry of the problem. The factor n determines the number of relative minima for a
rotation of 2t . If the problem has 2-fold symmetry than 2 n = . In the situation shown below there is,
by symmetry, only one minimum in the dihedral potential energy. In the pictured configuration, the
energy is either at a maximum or a minimum. Without knowing more about the atoms we cannot tell
which.

Up till now we have been dealing with energy associated with configuration of bonded atoms. There are
other important interactions that occur between non-bonded atoms. Computationally these are much
more difficult to deal with because there are so many more interacting pairs to consider. However in
principle it is no more difficult to understand these interactions than the bonded.
Consider the interaction between
the red and blue atoms in the
molecule shown on the left. The
two atoms are not bonded, yet they
can interact. The origin of this
interaction must be electrostatic in
nature. They are too far apart to
have their electrons interact
quantum mechanically. The
electrostatic interactions are broken
down into two generic types.
The first interaction type is easy to
understand. There can be a net
charge in the region around the red
atom and the blue atom. This
typically arises when these atoms
make their bonds and some of the
electrons are partially transferred to
or from the bonding partners. You
have probably learned about
ionic bonds. This is a type of bond where one atom, the most electronegative, has more electrons
spending more time near it than its partner. Therefore this atom acquires a negative charge and the
bonding partner a positive charge.
So imagine in the molecule above that the red atom upon bonding acquired a negative charge and
similarly so did the blue atom. Remember these acquired charges occurred with the red and blue atom
neighbors and not directly between these two atoms. However now there is an electrostatic interaction, a
repulsion, that occurs between the red and blue atoms. The form of the potential energy is familiar to
you,
1 2
( )
electrostatic
q q k
U r
r
. This is just the Coulomb interaction with a new term in the denominator.
For like-charges the energy increases with decreasing separation and yields a repulsive force, while for
opposite-charges the potential energy is lowered as the atoms come nearer. The
1
term is an
approximation that takes into account the possibility of intervening material between the red and blue
atoms that could screen or reduce the strength of the interaction. In a vacuum 1 , whereas if there is
water around the molecule the electrostatic interaction is can be reduced by as much as 80.

The last interaction that we will consider is the most complex. It arises due to the motion of the
arrangement of charge in an atom. Electrons are in rapid motion about a nucleus. Therefore the atom
always has a fluctuating charge distribution. Even though the atom has no net charge, i.e. it is neutral,
some of the time more electrons are on the left then the right and vice versa. This is a dynamic process
with the location of the high density of electrons continuously moving. However, even though the charge
location averages out over time, this movement can induce an important interaction between atoms.
Here is a simplified picture of the situation. Consider two identical neutral atoms, there has been no
bonding and no charge transfer and so the Morse and electrostatic interaction as described about would be
zero. Now take an instant in time and schematically draw the charge distributions of the two atoms,
electrons in blue, protons in red. Lets assume that the charge distributions are symmetric at this time.

Then, as must occur, a fluctuation in the charge of the left atom develops:

There is now more negative charge (the electrons) of the right hand side of that atom. The left-hand
nucleus cannot accelerate fast enough to follow the electrons because it is too massive. These charge
fluctuations occur on a time scale of 10
-15
sec (femtoseconds). Now the electrons on the right atom
experience a repulsion and so they respond by moving away.

This causes a net attraction between the atoms, due to the attraction of transient positive and negative
regions. This is called a fluctuating dipole-dipole interaction. The potential energy associated with it
varies with distance as
6
r
|
. The interaction is attractive.
This attraction brings the atoms closer together until their electron clouds begin to overlap

At this separation distance the atoms experience a very strong repulsion. This effect is quantum
mechanical and often approximated by a potential energy
12
( ) U r
r
o
. It should noted that this form is
used only for calculational convenience and that any other large power or exponential rise would be just
as appropriate.

Putting these two terms together gives us the Lennard-Jones potential, sometimes called the 6-12
potential, ( )
12 6
U r
r r
o |
= . The constants can be determined from the equilibrium separation and the
binding energy of the molecule. You will have a chance to do this in a homework problem.

The Lennard-Jones potential is a very important energy term, especially in biological applications. It is
reasonable to say that all of the lock and key arrangements that occur in biology for proteins arise due to a
large number of these relatively weak Lennard-Jones interactions.
Putting all the term together the potential energy for the molecular is:
( ) { }
2
2 2 1 2
0 0 12 6

1 1
( ) ( ) 1 cos( )
2 2
m
bonds angles dihedral non bonded
angles atom pairs
q q k
U r E r r k k n
r r r
u
o |
o u u
c
= + + + + +

To find the force on a particle the negative gradient with respect to the proper coordinate is found.
Obviously, there can be a lot of forces on a lot of particles, but once the forces are found, the procedure to
update the momentum and position is just the same as you have been using.
There are a couple of results that researchers are looking for from molecular mechanics calculations.
First they might want to understand the vibrational properties of the molecule. This is experimentally
measured by infrared spectroscopy. We will learn a bit more about that in the discussion on resonance in
the next section.
The holy grail of these types of calculation is to provide a list of the atomic numbers of the constituent
atoms and have the output from the computer predict the resultant molecule and its structure. Researchers
are a long way from being able to do this, but over just the past ten years tremendous strides have been
taken in understanding the dynamics of large biological proteins.
More information about molecular mechanics and some simulation movies can be found at these URLs:
http://www2.ess.ucla.edu/~schauble/molecular_vibrations.htm
http://www.ch.embnet.org/MD_tutorial/pages/MD.Part1.html
http://lammps.sandia.gov/movies/crack.mpg
http://lammps.sandia.gov/movies.html
http://lammps.sandia.gov/movies/nanowire2.mov
http://lammps.sandia.gov/movies/friction.mpg
http://lammps.sandia.gov/movies/micelle.mpg
Driven Oscillations and Resonance
In the last section we learned about damping and how a damping force can remove energy from
the system, but what about the opposite effect? Can a force do positive work and feed energy
into the system? A constant force will disturb the system and move it away from equilibrium. If
there is damping the system will settle down to a new stable location where the sum of the forces
equals zero.
A more interesting case is when the applied force is itself oscillatory. This is a very common
physical situation. One of the first times you experienced this was when you jumped on a swing
and someone pushed you. The swing is a mechanical system that oscillates and if the person
pushing the swing times his pushes (forces) properly, energy is fed into the system and the
amplitude of oscillation grows. If, however, the pusher does not use the correct frequency, either
no work is done or worse negative work is performed and you eventually stop swinging. This is
the general problem of a driven oscillator.
Newtons equation for a one-dimensional spring-mass system is
x
dp
kx
dt
. Using the
nonrelativistic expression for momentum, this can be rewritten as
2
2
0
d x
m kx
dt
. To have a
driven system an additional external force must be applied. We want a force that can have a
variable frequency;
0
sin( )
drive
F F t is a simple example. Adding this to dynamical equation
yields,
2
0 2
sin( )
d x
m kx F t
dt
. It is standard to divide through by the mass, which gives
2
0
2
sin( )
F d x k
x t
dt m m
. An identification can be made,
2
0
k
m
, the natural frequency of free
oscillations. So the equation for a driven oscillator with no damping becomes
2
2 0
0 2
sin( )
F d x
x t
dt m
. This is a linear second-order inhomogeneous differential equation
with constant coefficients. It is one of the simplest differential equations to solve and you will
learn how to do this is Calc IV. I will just quote the solution,
0
0 2 2
0
( ) sin( ) sin( )
F
m
x t A t t

. This solution has two parts. The first we have seen

before:
0
sin( ) A t . This has two unknown constants , A that are determined from the
initial conditions. This is called the homogeneous solution because is solves the differential
equation with no driving term:
2
2
0 2
0
d x
x
dt
. The other part
0
2 2
0
sin( )
F
m
t

is called the
inhomogeneous or particular solution and depends on the amplitude and frequency of the
driving force.
There is a very important feature to this inhomogeneous part of the solution: as the drive
frequency approaches the natural frequency
0
the amplitude grows and diverges to infinity.
This phenomenon is called resonance: as a function of frequency the amplitude of response
of the system grows and peaks at a frequency call the resonant frequency. At the resonant
frequency the driving force only does positive work and steadily increases the energy and
amplitude of oscillations of the system. If there is no damping, as in the above mathematical
equation, the amplitude, and thus energy, grows without bound. Physically this is never the case
as some loss mechanism or nonlinearity enters the system and keeps the oscillations from
diverging.
It is very instructive to now examine the numerical solutions to a resonant system. A 0.1 kg
mass is connected to a fixed wall by a 0.3 m long spring with a force constant of 3 N/m. A
sinusoidal force is placed on the mass. (You can think of this as similar to magnet being moved
to and from attracting the mass with a time varying force.) The drive force on the mass is given
by
0
( ) sin( ) F t F t x , gravity is ignored. The amplitude of the driving force is
0
0.015 N F and
the angular frequency of the driving force, , is variable. There is no damping in the system.

The natural time scale in this problem is the period of oscillation. Using the parameters of the
problem this time is 2 1.15 sec
m
T
k
. It is more convenient to use the natural angular
frequency
0
5.48
k
m

radians/sec. In the graphs below the angular frequency of the driving
force is scaled by this natural frequency. Therefore resonance occurs when
0
1
. Each
graph shows the trajectory of the mass (only the x position varies with time) in black. Also
shown are graphs of the natural oscillation (
0
sin( ) t in cyan) and the driving force ( sin( ) t in
red); the amplitudes of these traces are arbitrary, what is important is the phase relationship
between the two.
This first graph shows the oscillation of the mass when the drive frequency is one half of the
natural frequency. Clearly not much is going on. The position of the mass, found by
numerically integrating momentum update, is close to its equilibrium and the amplitude of
oscillation is small. There is very little relationship to the free oscillation (red) and drive
frequency (cyan). This means that there is a rapid oscillation between the drive force doing
positive and negative work and the system does not build up a large amount of energy.

As the frequency of the driving force is increased to 0.8 of the natural frequency, some
interesting features in the response of the mass are observed. Oscillations in the amplitude are
clearly present. The period of this modulation is approximately five cycles. During one cycle
the amplitude varies from essentially zero to ~0.05 m. A pattern can also be seen in the traces of
the drive and natural oscillations. They seem to come into correspondence every five cycles.
When the traces coincide, the amplitude of response is growing. However, when the traces are
1
0 0
0
5.48 sec , 0.5, 0.15 , 0 F N b

out of phase, the oscillation amplitude of the mass is decreasing.

This same behavior is more evident in the next plot where the drive frequency is 90% of the
natural frequency. It is important to note that the amplitude of the driving force is the same for
all of these simulations. All that is changing is the frequency of the drive. This time the
maximum amplitude is larger still and the period of modulation has increased to ~10 cycles.

1
0 0
0
5.48 sec , 0.8, 0.15 , 0 F N b

1
0 0
0
5.48 sec , 0.9, 0.15 , 0 F N b

At
0
0.95 this modulating pattern is striking. The oscillations have what is called a beat
pattern. It comes about when two frequencies are present in the problem. The rapid oscillation
occurs at the average of the two frequencies, while the longer oscillation occurs at the difference
of the frequencies. In this case the frequencies differ by 5% or 1/20
th
, therefore the beat period (
1/ f ) is 20 cycles. If you go back and look at the earlier graphs you will see similar behavior.
To generate this beat pattern you cannot simply add the drive oscillation and the natural
oscillation. This would not make any sense as one has the units of force and the other distance.
You cant simple add them. You need to think about what the driving force is doing to the
system.
The system has a natural frequency (in red) that it oscillates at, just like a backyard swing. If the
drive force (cyan) is in perfect sync then it can do positive work all the time, i.e. F dr is always
positive. But if the oscillations frequency is different from the drive frequency, some of the time
they will be out of sync, with the force in one direction and the displacement in the opposite.
During these periods the drive will be doing negative work, reducing the energy of the system
and decreasing the amplitude of oscillation.
Lets look closely at this relationship in the plot with
0
0.98 . The beat period will be fifty
cycles, 1/ 0.02. At time 0 t the drive frequency is very closely matched to the natural
frequency. This is due to the initial conditions used: the mass was at rest, at the equilibrium
position. This being the case 0 F dr and the energy and amplitude of oscillation increase,
essentially linearly at the early times. However as time goes on the two oscillations begin to get
out of phase. The drive force begins to lag the natural oscillation; by 20 oscillations or so, not
1
0 0
0
5.48 sec , 0.95, 0.15 , 0 F N b

nearly as much positive work is being done and the increase in amplitude has slowed markedly.
By the twenty fifth oscillation the two traces are completely out of phase and 0 F dr , and the
drive force is now removing energy from the system. This continues until the amplitude of
oscillation is essentially zero (50 cycles) and the process repeats. This is physically how the beat
pattern is produced for this situation.

So what happens when the drive frequency is equal to the natural frequency? Energy is
continuously being fed into the system and the amplitude of oscillation grows linearly until
something about the system changes, i.e. the system breaks. In the plot below at 23 t sec the
amplitude of oscillation exceeded the length of the spring breaking the simulation.

1
0 0
0
5.48 sec , 0.98, 0.15 , 0 F N b

1
0 0
0
5.48 sec , 1.0, 0.15 , 0 F N b

As the drive frequency is increased above the resonance frequency, the same beat pattern
develops as was present for lower frequencies. This is due to the exact same type of dephasing
conditions leading to times of positive and negative work being done by the external force.

By plotting the maximum amplitude of oscillation against the drive frequency you generate a
resonance curve, shown below. Again, the amplitude of the driving force has remained constant,
what was varied was the frequency of the drive. On the plot are shown the maximum amplitude
1
0 0
0
5.48 sec , 1.05, 0.15 , 0 F N b

1
0 0
0
5.48 sec , 2.0, 0.15 , 0 F N b

data from the previous graphs and the curve given by
0
2 2
0
F
m

as reference. Remember the
solution for this trajectory is
0
0 2 2
0
( ) sin( ) sin( )
F
m
x t A t t

, so the red curve does not

provide the complete solution, however the inhomogeneous amplitude (the part due to the
driving force) does contain the resonance feature. The analytic curve diverges at the resonance,
since there is no damping in the problem. The real data cuts off this singularity when the
amplitude of oscillation exceeds the length of the spring.

It is instructive to take a closer look at the energy of the system and the work performed by the
external driving force. The infinitesimal work performed by the external force is given by
dW F dr . The rate at which work is being done, i.e. the power, is then given by this divided
by dt , i.e.

dW dr
P F F v
dt dt
. It is easy for us to examine the power of the driving force
on the system during a simulation. All we need to do is take the dot product of the external drive
force and the velocity, Power = dot(Force_drive, Momentum/mass) is the VPython syntax, and
graph the result.
This is shown below for
0
0.95 over one beat period. The first plot, as before, shows the
amplitude of oscillation growing and the phase relation between the natural oscillation (red) and
the drive (cyan).

The next plot contains new types of information. The cyan curve shows the instantaneous power
of the external drive force. At the start of the cycle this power is oscillating, yet always positive.
The drive force is feeding energy into the system. The black curve is the integral of this power
(cyan curve) and shows the total work done by the drive and thus the change in energy of the
system. Since this system started with zero kinetic and potential energy, this energy is the total
energy for the system at any point in time. Near 5 t s, a change becomes noticeable. The drive
power is beginning to go negative during parts of a cycle. If you look at the graph above, you
will see that this is the time where the phase relationship between the drive and free oscillation
has slipped by
4
. By 10 seconds the two are completely out of phase. However, instead of

this simply being a time where only negative work is done, the power is actually divided equally
between positive and negative parts during a cycle. This yields almost no net work and the
energy of the system (black curve) does not change substantially during this phase of the cycle.

1
0 0
0
5.48 sec , 0.95, 0.15 , 0 F N b

Continuing on, the drive force starts to have a net negative power and the energy of the system
begins to decrease. This can be seen in both the total energy of the system and the amplitude of
oscillations. Near 23 t s the energy equals zero and the cycle repeats itself.
So we have seen that the external drive provides power to the system, both positive and negative,
and changes the total energy. Can we account for this energy? Of course, it shows up in the
system as kinetic and potential energy. Below is a graph of the three terms in equation
E KE PE : the kinetic energy,
2
2
p
m
, is shown in orange; the potential energy,
2 1
2
k x , in
green; and their sum, the total mechanical energy, in black.

A very important point to note is that the total mechanical energy equals the energy fed into the
system (remember we started this problem with zero energy).
The discussion above about the work done by the driving force on the system is really the key to
understanding resonance and why it occurs. Resonance is an extremely common and important
phenomenon of nature. It is well worth taking the time to understand it well.
Just a couple more examples to further illustrate these general properties of resonance. Lets
look at the case of resonance when the drive frequency equals the natural frequency of the
system. Below are the same three graphs as before.

This shows the amplitude and phase relationship. Only one trace can be seen because the free
oscillation and the drive have identical frequencies. The amplitude increases until it exceeds the
physical limits of the system and it breaks at ~23 s.

The next graph shows the power and the total energy. The power is always positive and the total
energy grows rapidly.
1
0 0
0
5.48 sec , 0.95, 0.15 , 0 F N b

As expected the mechanical energy of the system tracks the total work done by the drive force.

For simplicity we have been examining a system that starts at rest in its equilibrium position, i.e.
the initial total energy is zero. However, all the ideas that we have developed apply just as well
for other initial conditions. Below are the plots from a system that is started away from
equilibrium with an initial velocity. The only important change to note is that the initial energy
of the system is not zero and so the work done by the external force produced changed from this
nonzero value.
Beats are again seen in the amplitude of oscillation.

In this case the external force begins by doing negative work on the system.

Lastly, the total mechanical energy does not start at zero, but is altered from its initial value,
0.0157 J, by the external work.
1
0 0
0
5.48 sec , 0.9, 0.15 , 0 F N b

Dissipation
Quite often the physical system has some dissipation. This is frequently approximated as a force
that is proportional and in opposition to the velocity. It should be emphasized that the reason for
this choice is as much mathematical convenience as physical accuracy. The damping force is
empirically written as
damping
dx
F b
dt
= . Adding this to the oscillatory dynamical equation gives:
2
0 2
sin( )
d x dx
m kx b F t
dt dt
e = + . Rearranging this yields the equation for a damped driven
harmonic oscillator,
2
2 0
0 2
sin( )
F d x dx
x t
dt dt m
e e + + = , where the damping is characterized by
b
m
. This is an equation that can be solved exactly. The solution is:
( ) ( )
( )
0
2
1 1 2
2
2
2 2
0
( ) sin( ) sin( )
t
F
m
x t Ae t t
e e o
e e e
= + + +
+
where as before
2
2
1 0
4
e e = .
The new term
1
2 2
0
( ) tan
e
o e
e e
| |
=
|
\ .
is a phase shift that depends on the frequency of the
driving force. As before, Aand are determined from the initial conditions.
There is one key change in this solution for the trajectory that is not present in the above
undamped problem. Notice that the free oscillation term is now exponentially damped. This is
just as it was for the case of the damped oscillator without a driving force. Again the term
2
t
e

sets the time for the natural oscillations to decay. The time constant for this decay is
2 2m
b
t
= = . With each passing period of time equal to this time constant the amplitude of the
free oscillation decreased by
1
1
0.368
2.718
e
~ = . So after 3t the amplitude of this term is only

5% of the original and after 5t it is down to 0.7%. Just as with the damped free oscillator, there
is one last very important parameter used to characterize the oscillator. It is called the quality
of the system and given the symbol . Q The definition is
0
Q
e
.
Therefore with time the homogeneous solution decays away leaving only the inhomogeneous
driven part. Note that this part of the solution oscillates at exactly the drive frequency.
Therefore after the decay of the homogeneous solution beats are no longer present, the motion is
purely sinusoidal at the drive frequency.
The amplitude of the inhomogeneous term shows the same type of resonance as a function of the
drive frequency as the undamped system did; however this amplitude no longer diverges to
infinity at resonance. This is due to the damping. Energy is always being lost from the system
and the dissipation increases with the velocity of the mass. As the amplitude of oscillation
increases so do the losses and eventually a steady state will be reached.
Lets look at a couple of particular examples. In the next graphs below the drive frequency is
0
0.9 , the damping term 0.013
kg
b
s
this leads to a time constant for the free oscillations of
(2)(0.1)
15
.013
sec. So look that after ~45sec the beats have largely disappeared and what
remains is a pure sinusoidal motion.

The free oscillation term, shown in cyan, is damped in the lower traces.
1
0 0
0
5.48 sec , 0.9, 0.15 , 0.013 F N b

Turning to the energy of the system, the next plot shows the kinetic, potential and total
mechanical energy of the system.

Again you can see that the system settles down to a steady state after the transient free
oscillations are damped away.
It is interesting to break down the energy of the system into two parts: the work done by the
external drive, and now also the work done by the damping term.

The top plots shows the power (cyan) and the total work done on the system by the driving force.
You can see now that this external force continually adds energy to the system even though the
mechanical energy has reached a constant level. Where does this energy go? It counters the
effects of the damping that is continually removing energy from the system by doing negative
work on it. The lower graph shows the power (magenta curve) from the damping force, it is
always negative, and the total work (blue trace) done on the system by this dissipative force.
From examining the slopes of the black and blue energy curves from the external drive and
damping forces, it is clear that the external drive is putting into the system just as much energy as
the damping is removing. This is what is meant by steady-state.
If the frequency of the drive is placed on resonance, the same type of energy balance develops; it
is just that the amplitude of the oscillations is larger.

1
0 0
0
5.48 sec , 1.0, 0.15 , 0.013 F N b

So now for the damped-driven oscillator there is not just one resonance curve but a family of
curves, one for each value of the damping. The amplitude that is recorded for these curves is the
steady state value measured after the transient free oscillations have damped away. Below is a
plot of the resonance curves for a number of damping values.

For low damping, 55 Q , the red curve, the peak is very similar to the undamped case. The
maximum in system response occurs at
0
1
. However, as the damping is increased, not only

is the peak amplitude decreased, but the peak position shifts to lower frequency. This is shown
by the black dots and curve. This curve is found by determining the maximum of
1
2 2 2 2
2
0
1
(( ) ( ) )
. By 2 Q , the resonance is quite flat and the frequency with maximum
Damping coefficient
1
(s )

0.1 red
0.2 orange
0.4 yellow
0.8 green
1.6 cyan
3.2 blue
6.4 magenta
12.8 black

Damping coefficient
1
(s )

0.1 red
0.2 orange
0.4 yellow
0.8 green
1.6 cyan
3.2 blue
6.4 magenta
12.8 black

Quality Q
55 red
27 orange
14 yellow
6.9 green
3.4 cyan
1.7 blue
0.9 magenta
0.4 black

Quality Q
55 red
27 orange
14 yellow
6.9 green
3.4 cyan
1.7 blue
0.9 magenta
0.4 black

response is much reduced. There is a critical damping
0
2
2 or 0.7
2
crit crit
Q

where the
maximum response occurs at 0 . In other words, at this damping and higher, the response of
the system simply decreases with increasing frequency of external drive. The damping is so
large that there is no longer a resonance. Sometimes this is a desirable feature of the system.
High Qsystems have quality factors of
3 6
10 10 , examples include clocks, tuning forks and
electronic tuners. Extremely accurate systems can go as high as
10
10 Q . On the other hand,
damping systems like the shock absorbers on your car have
1
2
Q .
There is a very important and deep connection between how an oscillatory system responds
freely to an impulse and how it responds to a continuous sinusoidal drive. These two properties
are connected through the quality Q of the system. We have previously examine a damped
non-driven oscillator and learned that theQcharacterizes the decay of the oscillations in terms of
the number of cycles.
We are going to compare two systems that are identical, with 0.1 m kg, 3 k N/m,
0
1 F N,
except for the damping. The natural frequency is
0
5.48 rad/sec. The heavily damped system
will have a damping constant 0.06 kg/s b , giving
1
0.6 s ,

and a
5.48
9.1
0.6
Q . The other
system has less damping with a 0.01, 0.1 b and a higher 55 Q .

Now take the more heavily damped system and apply an impulse, i.e. perturb the system for a
short time. The system will oscillate and the amplitude will decay away with time as shown
below.

The red trace shows the envelope function
2
t
e

. The time constant for the decay of the
oscillation is
2
which in this case is

2
3.3
0.6
s. This is the time it takes for the
oscillations to decay to
1
e
of its original value. Since

0 0
Q
Q

therefore the time

constant
0
2Q
. So the quality factor determines how long the system with ring after being
disturbed. High quality systems will oscillate a very long time, a good example is a fine wine
glass, tap it with your fingernail and it will ting for a long time, many hundreds of oscillations.
Now consider the energy in the system. Due to the dissipation in the system, the energy
provided by the impulse decays away:
1
0
5.48 sec , 0.6, m 0.1

Again the red curve is an exponential decay with the same time constant, 3.3 t = s, as above.
What is wrong? The energy decays much faster than this. The difference is that the energy
depends on the square of the amplitude, or velocity. As the amplitude decreases by a factor of 2,
the energy has decreased by a factor of 4. So the proper envelope function is
2
2
t
t
e e
| |
=
|
\ .
; this
is shown in magenta and is an excellent fit. So again, the time constant for decay can be written
as a function of Q,
0
1
2
energy
Q t
t
e
= = = .
Just to demonstrate where the energy goes, the next graph plots the power and energy lost to the
damping force. After the decay, the blue curve shows that -0.044 J has been lost. Now look at
the previous graph to see how much total energy the system started with. It is exactly the same,
so after a few time constants have passed essentially all the energy has been dissipated by the
damping.

The exact same analysis can be done with the low damping case. The plots are below,
remember
1
0.01 kg/s, 0.1 s b

and a higher 55 Q .

The higher Qis, the longer the system will ring. Here the time constant is 20 s.
1
0
5.48 sec , 0.1, m 0.1

Again, we see that the time constant for the energy is one half (10 sec) that for the amplitude
decay. Similarly, the energy lost shows up in the work done by the drag force.

What you should have gained from the discussion above is the understanding of how the quality
factor Q plays a role in the dynamics of a system perturbed and allowed to oscillate freely. High
Q systems ring for a long time, and the time constant is determined by the natural frequency and
Q,
0
2Q
. The energy of system decays away with a characteristic time that is one half that of
the amplitude. Lastly, energy is lost due to the negative work of the dissipative drag force.

All of these concepts are related to the freely oscillating system. There is nothing here about a
driven system. So what role does the quality play in a driven system? The first thing to
examine is the resonance curves for the two systems. The high Qsystem has a large well
defined peak with an amplitude at resonance that is greater than 50 times that at low frequencies.
By comparison, the more highly damped system has a much broader and less distinct resonance
peak. These are qualitative features; can the quality Q be quantitatively connected to the
behavior of the driven system?

The answer to this question is yes, and there are two common ways of doing this. The first comes
from looking at the resonance curves. An obvious dimensionless quantity that differentiates
these curves is the resonant frequency divided by the width of the peak,
0

. This quantity is
clearly larger in the case of the higher Q, but what is the exact relationship?
You start by picking a somewhat arbitrary point to measure the width, typically this taken to be a
50% of the maximum peak of the resonance curve. This is shown on the resonance curve in
black. The ratio of the length of these lines looks about like 6, the ratio of the Qfactors, but can
we be more precise? Of course, but it does take some algebra. I will just quote these results,
they are straightforward to derive.
Damping
= 0.1 55 Q red
= 0.6 9.1 Q blue

50%

The frequency of the maximum amplitude is:
1
2
max 0 2
1
1
2Q
e e
| |
=
|
\ .

The maximum amplitude is:
max 0 1
2
2
1
1
4
Q
A A
Q
=
| |
|
\ .
where
0 0
0
0
F F
A
k me
= =
The amplitude is given by
0
0 1
2
0
2
0
2
0
1
A A
Q
e
e
e
e
e e
e e
| |
=
|
\ .
(
| |
( +
|
( \ .

Now you need to find the width at 50% of
max
A . This means solving for
0
e
e
in the following
equation:
0
1 1
2
2 2
0
2
2
0
1
2
1
1
1
4
Q
Q
Q
e
e
e e
e e
=
| | (
| |
| ( +
|
\ .
( \ .

. Solving for this general case looks tedious.
Lets look at the limit where Q >> 1, this not need to be too high, greater than 5 works well.

After several pages of algebra, you find that in this highQlimit
0
2
Q e
e
=
A
. Measuring the full-
width at half amplitude for the two resonance curves gives
, 9.1 0
0.206
blue Q
e e
=
A = and
, 55 0
0.032
red Q
e e
=
A = . For the high damping case
0
1
4.85
0.206
e
e
= =
A
. The Q for this case was
9.1 so the agreement
0
2
Q e
e
=
A
is quite reasonable
9.1
4.85 4.55
2
~ = . Turning to the high Qcase,
here the measured
0
1
30.3
0.033
e
e
= =
A
this is to be compared with
55
27.5
2 2
Q
= = , also
agreement within 10%.

So this connects the decay of the free oscillation amplitude to the width of the dynamic
resonance curve. By measuring the decay properties of the perturbed oscillator you can
learn about the width of the resonance for the driven system. That is important.
There is one last very physical way of thinking about the quality of the system. When the free
system is perturbed and starts to oscillate, energy is removed during every cycle. The fraction of
the energy that is removed each cycle,
E
E
, is constant. (This is what it means to have an

exponential decay.) We can measure this fractional energy lost in a cycle.

The plots below shows one cycle from the 9.1 Q heavily damped system. The arrow indicates
the measured period.

You can see, below, that the total energy decreased markedly during this period. To determine
E
E
the data from the start and the end of a period is listed in the table below. Any period
would work for the analysis.
T

Here is the data from the initial and final points in the cycle:
Time (s) x position (m) Energy = KE + PE (J) Energy dissipated (J)
0.911 0.1603 0.029259 -0.018561
2.051 0.2011 0.014684 -0.033136
delta 0.014575 0.014575

The change in energy is 0.014575 J and since there is such a large fractional change I will use the
average energy over the period. (For higher Qsystems using this average is not important since
there is a smaller fractional change.)
0.02926 0.01468
1.51
2*0.01456
ave
E
E

. Now we know that (1 )

T
cycle
E E e

where T is a period.
For small losses (1 (1 )) E E T E T so
0
1 1
2 2
E Q
E T

9.1
1.45 1.51
2 2
ave
E Q
E

. This is quite good agreement considering that this is a rather low

Q system.

Moving to the 55 Q system

T
E
ave
E

Time (s) x position (m) Energy = KE + PE (J) Energy dissipated (J)
0.911 0.1273 0.044755 -0.003852
2.051 0.1369 0.039896 -0.008702
delta 0.004859 0.00485

T
E
ave
E
T
0.04476 0.03990 55
8.71 8.76
2*0.004859 2 2
E Q
E

Again this is excellent agreement. So we have been able to relate the loss in energy per cycle to
the Qof the system.
Here is a summary of the relationships for the quality factor Q:
Definition Decay of
free
oscillation
amplitude
Decay of free
oscillation
energy
Fractional
width of
resonance
peak
Fractional
loss of energy
per cycle
Maximum
Amplitude at
resonance
0 0
m
Q
b

0
2Q

0
Q

0
2
Q

2
E Q
E

max 0
A A Q

All of these properties of the oscillator are proportional to Q. They relate the decay of the free
oscillation and the resonance characteristics of the driven system. Fundamentally, Q

tells us
about the ability of the system to store energy.
Energy Dissipation in an Oscillating System
We have learned that oscillations are a natural response of a system to being disturbed from a
stable equilibrium. If the amplitude of the oscillation is small then harmonic motion results, for
larger amplitude anharminicities enter. If a system is disturbed it will continue to oscillate unless
there is some other force to do negative work on the system and remove mechanical energy.
Quite often the physical system has appreciable dissipation. This is frequently approximated as a
force that is proportional and in opposition to the velocity. It should be emphasized that the
reason for this choice is as much mathematical convenience as physical accuracy. Remember
the differential equation for a harmonic oscillator:
2
2
d x
m kx
dt
. The left-hand side is the rate of
change of momentum (with a constant mass) and the right-hand side is the restoring force
proportional to the displacement. To include damping an empirical force acting on the particle is
added. The damping force is frequently written as
damping
dx
F b
dt
. Adding this to the
oscillatory dynamical equation gives:
2
2
d x dx
m kx b
dt dt
. Rearranging this yields the equation
for a damped harmonic oscillator,
2
2
0 2
0
d x dx
x
dt dt
, where the system is characterized by
0
and
b k
m m
. This is an equation that can be solved exactly. The solution is:
2
( ) sin( )
t
x t Ae t
where
2
2
0
4
. The two constants Aand are, as before with the

undamped oscillator, determined from the initial conditions.

Below is a numerical simulation for a damped harmonic oscillator that is started from an initial
position L L with 0.2 L L . The system parameters are
0.1 kg, 3 N/m, 0.01 kg/s. m k b This means that the relevant parameters are damping
constant
1
0.1 s
b
m

and free angular frequency
1
0
5.48 s
k
m

. Note that both of
these parameter have the same units,
1
s
. You already know what

0
controls, the frequency of
oscillation, but what about ?

The new feature in the solution
2
( ) sin( )
t
x t Ae t
= +

is the exponential decay. The term
2
t
e

sets the time for the natural oscillations to die away. The time constant for this decay is
2 2m
b
t
= = . The red envelope functions shown in the graph are,

2
0
( )
t
f t L Le

= A , and show
the exponential decrease in oscillation amplitude around the equilibrium position. For the
simulation parameters above, the decay time constant is 20 s. With each passing period of time
equal to this time constant the amplitude of the free oscillation decreased by
1
1
0.368
2.718
e
~ = of the initial amplitude. So after 3t the amplitude of this term is only 5%

of the original and after 5t it is down to 0.7%.

It is also instructive to look at the energy of the system. This is shown in the plot below. The
black line shows the total mechanical energy, while kinetic energy is shown in orange and
potential energy is shown in green. Obviously energy is being lost from the system. Where does
it go?

This question can be answered by looking at the work done by the damping force, bv . But
before looking at the work, lets examine the power due to this force. The infinitesimal work
performed by the damping force is given by dW F dr . The rate at which work is being done,
i.e. the power, is then given by this divided by dt , i.e.

dW dr
P F F v
dt dt
. To get the
total work you integrate the power,
0
( )
t
W t Pdt
. This is presented in the graph below.

The power is shown in magenta. It oscillates as the velocity of the particle changes. The units
for power are Joules/s, or Watts (W), and in these plots the power is not shown to scale. What is
shown properly is the integral of the power, in blue, giving the work done by the damping force.
The important thing to recognize is that the loss of mechanical energy is equal to the work done
by the dissipation. This is seen by comparing the black total energy line in the previous graph to
the work done by the damping force shown in blue. They are simply a shifted version of one
another.

Next we examine to the same situation, only the damping coefficient, b , is increase by a factor
of 5. The simulations results are shown below. Clearly, the increased damping causes the
oscillations to decay more rapidly than before.

The decay time constant
1
b
is inversely proportional to b . Since b has increased by 5 the
decay time has been reduced a factor of 5.

There is one last very important parameter used to characterize an oscillator. It is called the
quality of the system and given the symbol . Q The definition is
0
Q

. This dimensionless
parameter characterizes how rapidly, in terms of number of oscillations, energy is lost in the
system. After Q oscillations the amplitude has been reduced to e

, or ~4%, of its original
value. This come from the period
0
2
T

, so the time for Q cycles is

0
0 0
2 2 2
QT Q

, and thus the exponential decay factor for t QT is
2
2
e e

. So
high quality systems ring for a large number of oscillations, while system with a low Q damp
out after a small number of oscillations. In naturally occurring systems the Q can range from
less than 1 to over
9
10 . The dimensionless parameter Qand the resonant frequency
0
are the
two most frequently cited parameters for oscillatory systems.

To complete the discussion, the plot below show a system with strong damping and a Qof 2.2.

It is certainly possible to damp a system so heavily that there are no oscillations at all.
Sometimes this is exactly what is needed. This can be seen from the oscillation frequency
2
2
0
4
. When
0
2
, or alternatively 2 b km , becomes imaginary and there are

no oscillations. An alternate way to write the oscillation angular frequency is
0 2
1
1
4Q
.
There are three possibilities: 1/ 2 Q and the system oscillates, 1/ 2 Q and the system has no
oscillations and 1/ 2 Q . These three behaviors are respectively called underdamped,
overdamped and critical damping. In the critically damped case the amplitude asymptotes to
zero the most rapidly.

The physics description low quality system is not a value judgment about its usefulness.
Rather it a characterization about how well the system stores energy.

Kinematics and Dynamics, An Analytic Approach

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Kinematics and Dynamics, An Analytic Approach

Uploaded by

Copyright:

Available Formats

Kinematics and dynamics, an analytic approach

Given ( ) a t find ( ) v t and ( ) r t

. From our model, the drift velocity is constant in

. This means that the momentum

, which is a kinematical definition and is certainly correct. What is no

.Instead, the correct relativistic expression is

, the expression reduces to v cv ~

. This is the case for an ultrarelativistic

, V is the voltage between the

, however the definition of the momentum for a particle is changed to

= . This just returns us to the Newtonian definition for momentum, P mv = . At the

and the acceleration (red) is out of phase by .

= . The general solution is the sum of these, plus

= . Having now done all of this we have

= + + into a new form

, but we cant write g = ]Jg.

|cos0. Putting this all together

linear motion) and (curvature)

. So that describes the change in

, we can now use the nonrelativistic momentum to find,

linear motion) and = m = m

. Clearly these are opposite

and noting that r r yeilds,

and the velocty of the CM,

. Since the two blocks must move together,

. Solving this for

The spring constant for a parallel combination of

. Not all forces can have their work done

. In other words the work done by the force

. If this path independence holds, then the force

= . It may concern you that this function is always negative.

where h is the height above the surface

where 0 r and , 0 . Dimensionally

and setting equal to zero and solving yields

. Note that dimensionally

. We can now label these on the potential energy

= + . This is really completely an approximate form that only takes into

. This solution has two parts. The first we have seen

, so the red curve does not

. By 10 seconds the two are completely out of phase. However, instead of

~ = . So after 3t the amplitude of this term is only

. However, as the damping is increased, not only

which in this case is

of its original value. Since

therefore the time

, is constant. (This is what it means to have an

. Now we know that (1 )

. This is quite good agreement considering that this is a rather low

. The two constants Aand are, as before with the

. You already know what

= = . The red envelope functions shown in the graph are,

~ = of the initial amplitude. So after 3t the amplitude of this term is only 5%

. This is presented in the graph below.

, so the time for Q cycles is

, or alternatively 2 b km , becomes imaginary and there are

You might also like