Вы находитесь на странице: 1из 9

The Falling Chain Problem

Benjamin Clouser and Eric Oberla


December 4, 2008

Abstract
We present an experimental confirmation that the falling chain problem is best characterized by
inelatic collisions between succesive links in the chain. Theoretical treatments which assume energy
conservation as the chain falls exhibit limiting behavior that differs from those that do not. Our
experiment exploits these differences to decisively show that energy is not conserved and inelastic
collisions dominate.

1 Theoretical Motivations
A length of chain hangs at rest over the edge of a table, and the rest of the chain is bunched up as close to
the edge of the table as possible. The chain has length l, linear mass density , and the amount hanging
over the edge of the table is x0 . We wish to determine the behavior of the chain once it is released from
rest and begins to fall. We will now present an overview of two methods for deriving the equations of
motion.

1.1 Assuming Energy Conservation


The first derivation assumes that energy is conserved as the chain falls from the table. In effect, this
means that one assumes that the collisions between the chain links are elastic. We first write the poten-
tial and kinetic energies:
Z Z Z
1
U = dU = mgdx = g xdx = gx2 (1)
2
and
1 1
T = mx2 = xx2 (2)
2 2
Now we write the Lagrangian:
1 1
L=T U = xx2 + gx2 (3)
2 2
We define the Hamiltonian:
1 1
H = xpx L = xx gx2 (4)
2 2
The Hamiltonian is constant in time, so:
1 1
E= xx gx2 (5)
2 2
which in turn implies that
C
x2 gx = (6)
x
where C is a constant. Applying the boundary condition x(x0 ) = 0 leads to

x20
x2 = v 2 = g(x ) (7)
x

1
1.2 Assuming Inelastic Collisions
The above derivation relies on the unlikely assumption that energy is conserved in the system. Each
collision between successive links in the chain is inelastic, since initially one link is moving and the other
is stationary, and after the collision both links are moving at the same speed. We know from the previous
derivation that the momentum of the moving part of the chain is p = xx and the potential of the chain
dp
is U = 21 gx2 . With the knowledge that F = ( U x ) and F = dt , we can write:

d(xx)
= x2 + xx = gx (8)
dt
Using the relation that x = dv/dt = (dv/dx)(dx/dt) = v(dv/dx) we see that

1 d 2 2
(v x ) = gx2 (9)
2 dx
Integrating and applying the boundary conditions we arrive at

2g x3
x2 = v 2 = (x 20 ) (10)
3 x

1.3 Observable Behavior


We wish to experimentally show that the falling chain is a non-conservative system and obeys the
equations of motion derived assuming inelastic collisions. Thusly, we must determine which physical
consequences of the equations are in starkest contrast in order to best determine which regime dominates.
Although it is tempting to study the motion in the limit where x0 is comparable to x, this turns out to
unfeasible due to the limitations of our experimental apparatus. We are then led to study the motion
in the limit where x  x0 . Taking the derivative with respect to time for both equations of motion we
have:

N on Conservative Conservative
d 2 2 x3 d 2 x2
[x = g(x 02 )] [x = g(x 0 )]
dt 3 x dt x
2 x30 x20
2xx = g(x + 2 2 x) 2xx = g(x + 2 x)
3 x x
1 x3 1 x20
x = g(1 + 03 ) x = g(1 + 2 )
3 x 2 2x
In the limit x  x0 we see that these accelerations become constant, with:
1
x = g (11)
3
and
1
x = g (12)
2
for the non-conservative and conservative derivations, respectively.
Figure 1 graphs x vs. xx0 for both derivations and clearly shows the limiting behavior. Since the
accelerations become constant in this limit, we see in Figure 2 that v 2 becomes linear with the appropriate
slope in this limit as well. These are the behaviors we will seek in our experiment. Whether or not the
acceleration in our data goes as g3 or g2 when x  x0 will tell us whether or not energy is conserved.

2
8
E not conserved
E conserved
a = g/3
7 a = g/2

Acceleration [m/s2]
4

0
1 2 3 4 5 6
x/x0

Figure 1: Acceleration as a function of x/x0 for both derivations.

5
E not conserved, x0 = 0.1
E conserved, x0 = 0.1

3
v2 [m2/s2]

0
0.1 0.2 0.3 0.4 0.5 0.6
Length of Chain off Table, x [m]

Figure 2: Velocity squared as a function of x for both derivations. x is the length of chain over the edge
of the table.

2 Experimental Setup
We performed two experiments to test energy conservation of the falling chain. For the first experiment,
we used the set-up presented in Figure 3. A metal linked chain (m = 37.2 g, l = 64.5 cm) was coiled on
the edge of a table and a length x0 was allowed to drop over the edge. A string was tied to the link that
was immediately over the table edge and strung over a pulley with a light counterweight (m = 2.8 g)
on the other side. This counterweight keeps the string taut as it passes over the pulley, which is also a
rotary encoder and allows us to take position, velocity and acceleration data with a computer. We used
the Logger Pro software to take the data. The rotary encoder consists of 512 spokes, which block and

3
unblock 2 LEDs located inside the encoder. Having 2 LEDs instead of just one allows the software to
assign a direction to the displacement it measures.

Figure 3: Experimental setup 1. The figure shows the string looped over the rotary encoder and attached
to the counterweight on one end and the chain at the other.

We tested this set-up by varying x0 and the configuration of the chain on the table. Two configurations
were tried: 1) s-coil, where the chain was coiled in a repeated s shape on the table edge and 2) piled,
where the chain was piled on top of itself on the table edge. Piling the chain seems to result in very
uneven uncoiling, which is noticeable in the data. Putting the chain in an s-coil is more repeatable and
results in more regular data.
One problem with this method is that it is not valid when x x0 . When this is the case, the length
of chain x0 hanging over the table has a mass which is fairly close to the mass of the counterweight.
When the chain starts falling, the counterweight holds up the x0 portion of the chain, while the rest of
the chain continues to fall quickly to the ground. In effect, we end up measuring only the fact that the
x0 portion of the chain weighs about as much as the counterweight.
Another problem with the first experimental set-up seemed to be the large interaction between the
chain and the table edge. This was not taken into account by our theory, so we wanted to try an
experiment that eliminated this issue. Our second method is shown in Figure 4. In this set-up, we used
a beaded chain (see previous figure) that was draped over the pulley located at a height, h = 18.8 cm,
above the table. The beaded chain had mass, m = 37.4 g, streched length, ls = 94.5 cm and compressed
length lc = 71.8 cm.
Although this setup seemed to avoid the problems of the chain interacting with the tables edge, it
still does not correctly model the problem in the x x0 limit. We had great difficulty keeping the beaded
chain running over the pulley during the entirety of its fall. Any irregularites in the motion as the chain
unpiled itself were enough for the chain to jump the tracks and foul up the data. In order to combat this
we constructed a channel of sorts out of two index cards. These cards were attached to either side of the
rotary encoders pulley and did help to keep the chain in place. In reality, though, the channel did not
address the root issue of the chain bouncing off the pulley, it only ensured that the chain landed again
on the pulley after it bounced off.

4
Figure 4: Experimental setup 2. The beaded chain is looped over the rotary encoder, which spins and
directly records its movement. Call h the distance from the table up to the pulley, and x0 the amount
of chain hanging from the top of the pulley. A mathematical treatment of this problem shows that it
results in the same limiting behavior as the original setup. The cut up index cards on the rotary encoder
served as channels to keep the chain on the pulley.

The theory behind this experiment is slightly different than the previous set-up, as Newtons Second
Law becomes:
d
[(x + h)x] = g(x h)
(13)
dt
After some manipulation, integration, and application of initial conditions, equation (13) becomes:
1 1
(x + h)2 x2 = g(x x0 )(x2 + x0 x + x20 3h2 ) (14)
2 3
Examining this equation leads us to conclude that if x0 = 2h, then the acceleration is always constant
in time with a magnitude of g3 . Although this behavior is interesting, and our data did reflect it to some
extent, it is clearly not contained in either of the equations derived to describe the standard falling chain
problem. That said, this configuration still tends towards x = 31 g in the x  x0 limit.1

3 Data
Figure 5 below shows velocity data taken by our first experimental method with four different initial
conditions. We took data with x0 = 10.5 cm and 30 cm for both s-coil and pile configurations of the
chain. We concluded both from observing the chain fall as well as the presented data that the s-coil
method limited unwanted friction of the uncoiling chain. Careful inspection of the velocity data for the
chain initially in a pile shows oscillatory behavior. We surmise that this behavior is due to the chain
unpiling in an uneven manner. Accordingly, data from the s-coil method will be used to compare with
theoretical predictions.
1 Wong, C. H., S. H. Youn, K. Yasui, The falling chain of Hopkins, Tait, Steele, and Cayley.

5
160
x0 = 10.5 cm, pile
x0 = 30 cm, pile
x0 = 10.5 cm, s-coil
140 x0 = 30 cm, s-coil

120

100

Velocity [cm/s]
80

60

40

20

0
0 5 10 15 20 25 30 35 40
Length of Chain off Table [cm]

Figure 5: 4 data sets from our first experimental configuration. These data compare piled and s-coiled
chain configurations for x0 = 10.5 cm and 30 cm.

x0 = 3.2 cm
x0 = 5.2 cm
100 x0 = 11.2 cm
x0 = 21.2 cm

80
Velocity [cm/s]

60

40

20

0
0 5 10 15 20 25 30 35 40
Length of Chain off Table [cm]

Figure 6: Data sets for our second experimental setup. Note that the x0 values quoted on the graph
are actually x0 h. Data were taken with x0 h = 3.2 cm, 5.2 cm, 11.2 cm, and 21.2 cm. This setup
exhibits the oscillatory beat behavoir much more strongly than the first setup. Once again, we believe
these beats to be due to the chain unpiling unevenly. This claim is supported by the fact that the beats
seem to occur over a characteristic distance of 7 8 cm, which is roughly twice the diameter of the chain
pile.

4 Results
Results from our experiment can be seen in the next figures. We first consider the findings from experi-
mental setup 1.

6
25000

20000

15000

v2 [cm2/s2]
10000

x0 = 5 cm
5000 x0 = 30 cm
fit to x0=5 cm, v(x) = 711.1x - 4012
fit to x0=30 cm, v(x) = 760.7x + 2172

0
0 5 10 15 20 25 30 35 40
Length of Chain off Table, x [cm]

Figure 7: Plot of v 2 vs. x, the length of chain that has fallen off the table. When energy is not conserved,
equation (10) shows that v 2 (x) approaches a linear function with slope 23 g as x  x0 . A linear fit was
performed in this region. Our results showed a slope of 7.11 sm2 for x0 = 5 cm and 7.61 sm2 for x0 = 30 cm.
The expected result, 6.53 sm2 , is somewhat consistent with our data. It makes sense that value obtained
for x0 = 5 cm is closer to the expected value than that for x0 = 30 cm since the limit xx0 0 is more
easily attained.

160

140

120 x0 = 5 cm
x0 = 30 cm
fit (x0= 5 cm) v(t) = 332.9t - 102.4
fit (x0= 30 cm) v(t) = 372.1t + 27.2
100
Velocity [cm/s]

80

60

40

20

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time [s]

Figure 8: Plot of v vs. t. In our analysis, we apply our previous discussion of the limiting behavior of the
acceleration. Our data is fit with a line in the asymptotic region and the results are shown on the plot.
The results of the fit gave a slope of 3.32 sm2 for x0 = 5 cm and 3.72 sm2 for x0 = 30 cm. The expected
value, as shown in equation (11), is g3 or 3.27 sm2 for inelastic collisions. Our results confirm this.

7
2

1.5

Velocity [m/s]
1 x0 = 10.5 cm, s-coil
Theory, x0 = 10.5 cm
x0 = 30 cm, s-coil
Theory, x0 = 30 cm

0.5

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Length of Chain off Table [m]

Figure 9: Comparison of theoretical data with experimental data in a plot of v(x). Our data shows
the same shape and asymptotic behavior as the theory, but is consistently less by what appears to be
a constant offset. This is most likely the result of some systematic friction and is the reason why we
wanted to focus on the limiting region in the first place.

100

80
Velocity [cm/s]

60

40 x0 = 3.2 cm
x0 = 11.2 cm
x0 = 21.2 cm
fit (x0=3.2 cm) v(t)=186.0t-67.4
fit (x0=11.2 cm) v(t)=208.0t-29.7
fit (x0=21.2 cm) v(t)=238.1t-15.4
20

0
0 0.2 0.4 0.6 0.8 1
Time [s]

Figure 10: Results from experimental setup 2. We plotted v vs. t and fitted the data with a line in
the asymptotic region. As with the first set-up, we expect a value of a = g3 in this region, but our
data suggests something 2 times less. We conclude that our first set up was more consistent. By
attempting to eliminate interactions between the table edge and the chain, we appear to have introduced
more friction by dragging the beaded chain over the pulley.

5 Conclusion
We have presented our solution to the falling chain problem. The theory behind the falling chain suggests
that it will reach a constant acceleration once a sufficient length has fallen off the table. This acceleration

8
depends on whether energy is taken to be conserved (a = g2 ) or not (a = g3 ) in the equations of motion.
After trying several experimental setups, we settled for the two presented here. The data from our
first experimental setup gave us more consistent results. Upon fitting a v vs. t plot of our data in the
linear region, we found a constant acceleration that was close to g3 . A comprehensive error analysis was
not performed in this experiment so it is difficult to determine the specific range of our uncertainty.
However, our data was repeatable and it is realistic to conclude our experiment showed energy is not
conserved.
There was a clear discrepancy between the data from the two setups and several factors are likely to
blame. In setup 2, we used a beaded chain that was allowed to fall over a pulley, where the rotation of
the pulley gave us our data. Since our data gave results less than expected, it is possible that the chain
slid on the pulley and some motion data was lost (ie v > r). By placing a counterweighted string over
the pulley as in setup 1, it was more likely that v = r and the rotation of the pulley gave an accurate
description of the motion of the chain.
Our results, however surprising, seem to confirm the treatment of the falling chain as a series of in-
elastic collisions. The experimental setups were somewhat crude, but still gave reasonable and repeatable
data. Our first experimental setup was subjected to different initial conditions and considerable distur-
bances yet still managed to yield data that was consistent in the asymptotic, suggesting that behavior
in the limit x  x0 is fairly robust.

6 Acknowledgements
We would like to express our gratitude to Van Bistrow for graciously allowing us to use University
equipment and laboratory space.

Вам также может понравиться