Homework 1: # 1.21, 2.7, 2.

12
Michael Good
Sept 3, 2004

1.21. Two mass points of mass m1 and m2 are connected by a string passing
through a hole in a smooth table so that m1 rests on the table surface and
m2 hangs suspended. Assuming m2 moves only in a vertical line, what are the
generalized coordinates for the system? Write the Lagrange equations for the
system and, if possible, discuss the physical significance any of them might have.
Reduce the problem to a single second-order differential equation and obtain a
first integral of the equation. What is its physical significance? (Consider the
motion only until m1 reaches the hole.)
Answer:

The generalized coordinates for the system are , the angle m1 moves round
on the table, and r the length of the string from the hole to m1 . The whole
motion of the system can be described by just these coordinates. To write the
Lagrangian, we will want the kinetic and potential energies.
1 1
T = m2 r 2 + m1 (r 2 + r2 2 )
2 2

V = m2 g(R r)
The kinetic energy is just the addition of both masses, while V is obtained
so that V = mgR when r = 0 and so that V = 0 when r = R.
1 1
L=T V = (m2 + m1 )r 2 + m1 r2 2 + m2 g(R r)
2 2
To find the Lagrangian equations or equations of motion, solve for each
component:
L
=0

L
= m1 r2

d L d = m1 r2 + 2m1 rr
= (m1 r2 )

dt dt

1
 Thus
m1 r(r + 2r)
=0
and
L
= m2 g + m1 r2
r
L
= (m2 + m1 )r
r
d L
= (m2 + m1 ) r
dt r
Thus
m2 g m1 r2 + (m2 + m1 )
r=0
Therefore our equations of motion are:
d = m1 r(r + 2r)
(m1 r2 ) =0
dt
m2 g m1 r2 + (m2 + m1 ) r=0
See that m1 r2 is constant, because dt
d = 0. It is angular momentum.
(m1 r2 )
Now the Lagrangian can be put in terms of angular momentum and will lend the
problem to interpretation. We have = l/m1 r2 , where l is angular momentum.
The equation of motion

m2 g m1 r2 + (m2 + m1 )
r=0
becomes

l2
r
(m1 + m2 ) + m2 g = 0
m1 r3
The problem has been reduced to a single non-linear second-order differential
equation. The next step is a nice one to notice. If you take the first integral
you get

1 l2
(m1 + m2 )r 2 + + m2 gr + C = 0
2 2m1 r2
To see this, check by assuming that C = m2 gR:

d 1 l2 l2
( (m1 + m2 )r 2 + m 2 g(R r)) = (m1 + m2 )r
r r + m2 g r = 0
dt 2 2m1 r2 m1 r3

l2
r
(m1 + m2 ) + m2 g = 0
m1 r3
Because this term is T plus V , this is the total energy, and because its time
derivative is constant, energy is conserved.

2
2.7 In Example 2 of Section 2.1 we considered the problem of the minimum
surface of revolution. Examine the symmetric case x1 = x2 , y2 = y1 > 0,
and express the condition for the parameter a as a transcendental equation in
terms of the dimensionless quantities k = x2 /a, and = y2 /x2 . Show that
for greater than a certain value 0 two values of k are possible, for = 0
only one value of k is possible, while if < 0 no real value of k (or a) can be
found, so that no catenary solution exists in this region. Find the value of 0 ,
numerically if necessary.
Answer:

Starting with Goldsteins form for a catenary, in section 2.2, not 2.1,
yb
x = a cosh
a
and recognizing by symmetry that the soap film problem and the catenary
problem are the same. In Marion and Thorton this is made clear (pg 222).
Also, in a similar way to MathWorlds analysis of a surface of revolution, it is
clear that y and x, when interchanged, change the shape of the catenary to be
about the x-axis.
xb
y = a cosh
a
To preserve symmetry, x1 = x2 and y2 = y1 . This switch makes sense
because if you hang a rope from two points, its going to hang between the
points with a droopy curve, and fall straight down after the points. This shaped
revolved around the x-axis looks like a horizontal worm hole. This is the classic
catenary curve, or catenoid shape. The two shapes are physically equivalent,
and take on different mathematical forms. With this, we see that
x1 b x2 b
y1 = a cosh y2 = a cosh
a a
holds. The two endpoints are (x0 , y0 ) and (x0 , y0 ).
Thus
x0 b x0 b
y0 = a cosh = a cosh
a a
and because

cosh(x) = cosh(x)
we have
x0 + b x0 b
cosh = cosh
a a

x0 + b = x0 b

3
 b = b b=0
By symmetry, with the center of the shape or rings at the origin, b = 0,
simplifies the problem to a much nicer form:
x
y = a cosh
a
Including our end points:
x0
y0 = a cosh
a
In terms of the dimensionless quantities,
1
= x2 /a = = y2 /x2

,
the equation is
x0
y0 = a cosh
a
y0 a cosh
=
x0 x0
y0 cosh()
=
x0
cosh()
=

The minimum value of in terms of can be found by taking the derivative,
and setting to zero:
d 1 1 cosh
0= ( cosh ) = sinh
d 2
cosh
sinh =

= coth
Thus, solved numerically, 1.2. Plugging this in to find 0 , the value is:

0 1.51
Since
1
0 =
0
,
0 .66

4
 This symmetric but physically equivalent example is not what the problem
asked for, but I think its interesting. If we start at Goldsteins equation, again,
only this time recognize b = 0 due to symmetry from the start, the solution
actually follows more quickly.
y
x = a cosh
a
x xy
= cosh
a ax
Using, the dimensional quantities defined in the problem,
x2 y2
k= =
a x2
we have

k = cosh k
Taking the derivative with respect to k,

1 = 0 sinh k0
Using the hyperbolic identity,

cosh2 A sinh2 A = 1
a more manageable expression in terms of k and becomes apparent,
1
k2 =1
02
1
0 =
k2 1
Plug this into k = cosh k
k
k = cosh
k2 1
Solving this numerically for k yields,

k 1.81
Since
1
0 = 0 .66
k2 1
If < 0 , two values of k are possible. If > 0 , no real values of k exist,
but if = 0 then only k 1.81 will work. This graph is arccosh(k)/k =
and looks like a little hill. It can be graphed by typing acosh(x)/x on a free
applet at http://www.tacoma.ctc.edu/home/jkim/gcalc.html.

5
 2.12 The term generalized mechanics has come to designate a variety of classical
mechanics in which the Lagrangian contains time derivatives of qi higher than
the first. Problems for which triple dot x = f (x, x,
x, t) have been referred to
as jerky mechanics. Such equations of motion have interesting applications
in chaos theory (cf. Chapter 11). By applying the mehtods of the calculus
of variations, show that if there is a Lagrangian of the form L(qi , qi qi , t), and
Hamiltons principle holds with the zero variation of both qi and qi at the end
points, then the corresponding Euler-Lagrange equations are

d2 L d L L
2
( ) ( )+ = 0, i = 1, 2, ..., n.
dt qi dt qi qi
Apply this result to the Lagrangian
m k
q q q 2 L=
2 2
Do you recognize the equations of motion?
Answer:

If there is a Lagrangian of the form

L = L(qi , qi , qi , t)
and Hamiltons principle holds with the zero variation of both qi and qi at
the end points, then we have:
Z 2
I= L(qi , qi , qi , t)dt
1
and
Z 2
I X L qi L qi L qi
d = ( d + d + d)dt
1 i
q i i q
i i qi i
To make life easier, were going to assume the Einstein summation con-
vention, as well as drop the indexes entirely. In analogy with the differential
quantity, Goldstein Equation (2.12), we have
q
q = d

Applying this we have
Z 2
L L q L q
I = ( q + d + d)dt
1 q q q

6
 The indexes are invisible and the two far terms are begging for some mathe-
matical manipulation. Integration by parts on the middle term yields, in analogy
to Goldstein page 44,
2 2 Z 2
L 2 q
Z
L q  q d L
dt = ( )dt
1 q t q 1 1 dt q
This first term on the right is zero because the condition exists that all the
varied curves pass through the fixed end points and thus the partial derivative
of q wrt to at x1 and x2 vanish. Substituting back in, we have:
Z 2
L d L L q
I = ( q q + d)dt
1 q dt q q
q
Where we used the definition q = d again. Now the last term needs
attention. This requires integration by parts twice. Here goes:
2 2 Z 2 2
L 2 q 
Z
L q q d L
q
dt =
q
t  t
(
dt q
)dt
1 1 1
R R
Where we used vdu = uv vdu as before. The first term vanishes once
again, and we are still left with another integration by parts problem. Turn the
crank again.
2 2 Z 2
2 q d L q d2 L
Z
d L q 
( )dt = dt
1 t dt q dt q 1 1 dt2 q
First term vanishes for the third time, and we have
2 2
q d2 L
Z Z
L q
dt = dt
1 q 1 dt2 q
Plugging back in finally, and using the definition of our q, we get closer
2
d2 L
Z
L d L
I = ( q q + 2 q)dt
1 q dt q dt q
Gathering qs, throwing our summation sign and indexs back in, and ap-
plying Hamilitons principle:
2
d2 L
Z X L d L
I = ( + 2 )qi dt = 0
1 i
qi dt qi dt qi
We know that since q variables are independent, the variations qi are in-
dependent and we can apply the calculus of variations lemma, (Goldstein, Eq.
2.10) and see that I = 0 requires that the coefficients of qi separately vanish,
one by one:

L d L d2 L
+ 2 =0 i = 1, 2, ...n.
qi dt qi dt qi

7
 Applying this result to the Lagrangian,
1 k
L = mq q q 2
2 2
yields
L 1
= m
q kq
q 2
d L
=0
dt q
d2 L d d 1 d 1 1
= m
= ( ( mq)) = ( mq) q
dt2 q dt dt 2 dt 2 2
Adding them up:

m
q kq = 0
This is interesting because this equation of motion is just Hookes Law. This
crazy looking Lagrangian yields the same equation for simple harmonic motion
using the jerky form of Lagrangians equations. Its interesting to notice that
if the familiar Lagrangian for a simple harmonic oscillator (SHO) plus an extra
term is used, the original Lagrangian can be obtained.
d mq q
L = LSHO + ( )
dt 2

mq2 kq 2 d mq q
L= + ( )
2 2 dt 2

mq2 kq 2 mq q mq2
L=
2 2 2 2

mq q kq 2
L=
2 2
This extra term, dt d
( mq q
2 ) probably represents constraint. The generalized
force of constraint is the Lagrange multipliers term that is added to the original
form of Lagranges equations.