Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
0% found this document useful (0 votes)
30 views20 pages

The Rolling Motion of A Disk On A Horizontal Plane

Download as pdf or txt
Download as pdf or txt
Download as pdf or txt
You are on page 1/ 20

The Rolling Motion of a Disk on a Horizontal Plane

Alexander J. McDonald
Princeton High School, Princeton, New Jersey 08540
Kirk T. McDonald
arXiv:physics/0008227v3 [physics.class-ph] 5 Apr 2002

Joseph Henry Laboratories, Princeton University, Princeton, New Jersey 08544


mcdonald@puphep.princeton.edu

(March 28, 2001)

Abstract
Recent interest in the old problem of the motion of a coin spinning on a tabletop has
focused on mechanisms of dissipation of energy as the angle α of the coin to the table
decreases, while the angular velocity Ω of the point of contact increases. Following a review
of the general equations of motion of a thin disk rolling without slipping on a horizontal
surface, we present results of simple experiment on the time dependence of the motion that
indicate the dominant dissipative power loss to be proportional to the Ω2 up to and including
the last observable cycle.

1 Introduction
This classic problem has been treated by many authors, perhaps in greatest detail but very
succinctly by Routh in article 244 of [1]. About such problems, Lamb [2] has said, “It
is not that the phenomena, though familiar and often interesting, are held to be specially
important, but it was regarded rather as a point of honour to shew how the mathematical
formulation could be effected, even if the solution should prove to be impracticable, or
difficult of interpretation.” Typically, the role of friction was little discussed other than in
relation to “rising” [1, 2, 3, 4, 5, 6]. The present paper is motivated by recent discussions
[7, 8, 9] of friction for small angles of inclination of a spinning disk to the horizontal supporting
surface.
The issues of rolling motion of a disk are introduced in sec. 2 in the larger context of
non-rigid-body motion and rolling motion on curved surfaces, using the science toy “Euler’s
Disk” as an example. In our analysis of the motion of a rigid disk on a horizontal plane we
adopt a vectorial approach as advocated by Milne [10]. The equations of motion assuming
rolling without slipping are deduced in sec. 3, steady motion is discussed in secs. 4 and 5, and
oscillation about steady motion is considered in sec. 7. The case of zero friction is discussed
in secs. 8 and 9, and effects of dynamic friction are discussed in secs. 6, 10 and 11. Section
12 presents a brief summary of the various aspects of the motions discussed in secs. 3-11.

1
2 The Tangent Toy “Euler’s Disk”
An excellent science toy that illustrates the topic of this article is “Euler’s Disk”, distributed
by Tangent Toy Co. [11]. Besides the disk itself, a base is included that appears to be the
key to the superior performance exhibited by this toy. The surface of the base is a thin,
curved layer of glass, glued to a plastic backing. The base rests on three support points to
minimize rocking.
As the disk rolls on the base, the latter is noticeably deformed. If the same disk is rolled
on a smooth, hard surface such as a granite surface plate, the motion dies out more quickly,
and rattling sounds are more prominent. It appears that a small amount of flexibility in the
base is important in damping the perturbations of the rolling motion if long spin times are
to be achieved.
Thus, high-performance rolling motion is not strictly a rigid-body phenomenon. However,
we do not pursue the theme of elasticity further in this paper.
The concave shape of the Tangent Toy base helps center the rolling motion of the disk,
and speeds up the reduction of an initially nonzero radius b to the desirable value of zero.
An analysis of the motion of a disk rolling on a curved surface is more complex than that
of rolling on a horizontal plane [12]. For rolling near the bottom of the sphere, the results
as very similar to those for rolling on a plane. A possibly nonintuitive result is that a disk
can roll stably on the inside of the upper hemisphere of a fixed sphere, as demonstrated in
the motorcycle riding act “The Globe of Death” [13].

3 The Equations of Motion for Rolling Without Slip-


ping
In addition to the ẑ axis which is vertically upwards, we introduce a right-handed coordinate
triad of unit vectors (1̂, 2̂, 3̂) related to the geometry of the disk, as shown in Fig. 1. Axis
1̂ lies along the symmetry axis of the disk. Axis 3̂ is directed from the center of the disk to
the point of contact with the horizontal plane, and makes angle α to that plane. The vector
from the center of the disk to the point of contact is then
a = a3̂. (1)
Axis 2̂ = 3̂ × 1̂ lies in the plane of the disk, and also in the horizontal plane. The sense
of axis 1̂ is chosen so that the component ω1 of the angular velocity vector ~ω of the disk
about this axis is positive. Consequently, axis 2̂ points in the direction of the velocity of the
point of contact. (For the special case where the point of contact does not move, ω1 = 0 and
analysis is unaffected by the choice of direction of axis 1̂.)
Before discussing the dynamics of the problem, a considerable amount can be deduced
from kinematics. The total angular velocity ~ω can be thought of as composed of two parts,
~ω = ~ω123 + ωrel 1̂, (2)
where ~ω123 is the angular velocity of the triad (1̂, 2̂, 3̂), and ωrel 1̂ is the angular velocity
of the disk relative to that triad; the relative angular velocity can only have a component

2
Figure 1: A disk of radius a rolls without slipping on a horizontal plane. The
symmetry axis of the disk is called axis 1̂, and makes angle α to the ẑ axis,
which is vertically upwards, with 0 ≤ α ≤ π. The line from the center of the
disk to the point of contact with the plane is called axis 3̂, which makes angle
α to the horizontal. The horizontal axis 2̂ is defined by 2̂ = 3̂ × 1̂, and the
horizontal axis r̂ is defined by r̂ = 2̂ × ẑ. The angular velocity of the disk
about axis 1̂ is called ω1 , and the angular velocity of the axes (1̂, 2̂, 3̂) about
the vertical is called Ω. The motion of the point of contact is instantaneously
in a circle of radius r. The distance from the axis of this motion to the center
of mass of the disk is labeled b.

along 1̂ by definition. The angular velocity of the triad (1̂, 2̂, 3̂) has component α̇ about
the horizontal axis 2̂ (where the dot indicates differentiation with respect to time), and is
defined to have component Ω about the vertical axis ẑ. Since axis 2̂ is always horizontal,
~ω123 has no component along the (horizonal) axis 2̂ × ẑ ≡ r̂. Hence, the angular velocity of
the triad (1̂, 2̂, 3̂) can be written

~ω123 = Ωẑ + α̇2̂ = −Ω cos α1̂ + α̇2̂ − Ω sin α3̂, (3)

noting that
ẑ = − cos α1̂ − sin α3̂, (4)
as can be seen from Fig. 1. The time rates of change of the axes are therefore

d1̂
= ω
~ 123 × 1̂ = −Ω sin α2̂ − α̇3̂, (5)
dt
d2̂
= ω
~ 123 × 2̂ = Ω sin α1̂ − Ω cos α3̂, = −Ωr̂, (6)
dt
d3̂
~ 123 × 3̂ = α̇1̂ + Ω cos α2̂,
= ω (7)
dt

3
where the rotating horizontal axis r̂ is related by

r̂ = 2̂ × ẑ = − sin α1̂ + cos α3̂, (8)

whose time rate of change is


dr̂
= Ω2̂. (9)
dt
Combining eqs. (2) and (3) we write the total angular velocity as

~ω = ω1 1̂ + α̇2̂ − Ω sin α3̂, (10)

where
ω1 = ωrel − Ω cos α. (11)
The (nonholonomic) constraint that the disk rolls without slipping relates the velocity of
the center of mass to the angular velocity vector ~ω of the disk. In particular, the instanta-
neous velocity of the point contact of the disk with the horizontal plane is zero,

vcontact = vcm + ~ω × a = 0. (12)

Hence,
drcm
vcm = = a3̂ × ω
~ = −aα̇1̂ + aω1 2̂, (13)
dt
using eqs. (1) and (10).
Additional kinematic relations can be deduced by noting that the point of contact between
the disk and the horizontal plane can always be considered as moving instantaneously in a
circle whose radius vector we define as r = rr̂ with r ≥ 0, as shown in Fig. 1, and whose
center is defined to have position xA x̂ + yA ŷ where x̂ and ŷ are fixed horizontal unit vectors
in the lab frame. Then, the center of mass of the disk has position

rcm = xA x̂ + yA ŷ + rr̂ − a3̂, (14)

and
drcm
= ẋA x̂ + ẏA ŷ + ṙr̂ − aα̇1̂ + (r − a cos α)Ω2̂. (15)
dt
In the special case of steady motion, ẋA = ẏA = ṙ = α̇ = 0, eqs. (13) and (15) combine to
give
b
ω1 = Ω, (16)
a
where
b = r − a cos α (17)
is the horizontal distance from the axis of the circular motion to the center of mass of the
disk. Thus, for steady motion the “spin” angular velocity ω1 is related to the “precession”
angular velocity Ω according to eq. (16). While ω1 is defined to be nonnegative, length b can
be negative if Ω is negative as well.

4
Except for axis 1̂, the rotating axes (1̂, 2̂, 3̂) are not body axes, but the inertia tensor Iij
is diagonal with respect to them in view of the symmetry of the disk. We write

I11 = 2kma2 , I22 = kma2 = I33 , (18)

which holds for any thin, circularly symmetric disk according to the perpendicular axis
theorem; k = 1/2 for a disk with mass m concentrated at the rim, k = 1/4 for a uniform
disk, etc. The angular momentum Lcm of the disk with respect to its center of mass can
now be written as
~
Lcm = I~ · ω
~ = kma2 (2ω1 1̂ + α̇2̂ − Ω sin α3̂). (19)
Turning at last to the dynamics of the rolling disk, we suppose that the only forces on it
are −mgẑ due to gravity and F at the point of contact with the horizontal plane. For now,
we ignore rolling friction and friction due to the air surrounding the disk.
The equation of motion for the position rcm of the center of mass of the disk is then
d2 rcm
m = F − mgẑ. (20)
dt2
The torque equation of motion for the angular momentum Lcm about the center of mass is
dLcm
= Ncm = a × F. (21)
dt
We eliminate the unknown force F in eq. (21) via eqs. (1) and (20) to find

1 dLcm d2 rcm
+ × 3̂ = g 3̂ × ẑ. (22)
ma dt dt2
This can be expanded using eqs. (4), (5)-(7), (13) and (19) to yield the 1̂, 2̂ and 3̂ components
of the equation of motion,

(2k + 1)ω̇1 + α̇Ω sin α = 0, (23)


g
kΩ2 sin α cos α + (2k + 1)ω1Ω sin α − (k + 1)α̈ = cos α, (24)
a
Ω̇ sin α + 2α̇Ω cos α + 2ω1 α̇ = 0. (25)

These equations agree with those of sec. 244 of Routh [1], noting that his A, C, θ, ψ̇ and ω3
are expressed as ka2 , 2ka2 , α, Ω and −ω1 , respectively, in our notation.
Besides the coordinates (xA , yA ) of the center of motion, we can readily identify only one
other constant of the motion, the total energy
1 2 1 ~
E = T + V = mvcm + ~ω · I~ · ~ω + mgz
2 2
ma2 2g
 
2 2 2 2
= (2k + 1)ω1 + (k + 1)α̇ + kΩ sin α + sin α . (26)
2 a
The time derivative of the energy is consistent with the equations of motion (23)-(25), but
does not provide any independent information.

5
4 Steady Motion
For steady motion, α̇ = α̈ = Ω̇ = ω̇1 = 0, and we define αsteady = α0 , Ωsteady = Ω0
and ω1,steady = ω10 . The equations of motion (23) and (25) are now trivially satisfied, and
eq. (24) becomes
g
kΩ20 sin α0 cos α0 + (2k + 1)ω10 Ω0 sin α0 = cos α0 , (27)
a
A special case of steady motion is α0 = π/2, corresponding to the plane of the disk being
vertical. In this case, eq. (27) requires that ω10 Ω0 = 0. If Ω0 = 0, the disk rolls along a
straight line and ω10 is the rolling angular velocity. If ω10 = 0, the disk spins in place about
the vertical axis with angular velocity Ω0 .
For α0 6= π/2, the angular velocity Ω0 ẑ of the axes about the vertical must be nonzero.
We can then replace ω10 by the radius b of the horizontal circular motion of the center of
mass using eqs. (16)-(17):
b r
 
ω10 = Ω0 = Ω0 − cos α0 . (28)
a a
Inserting this in (27), we find
g cot α0 g cot α0
Ω20 = = . (29)
ka cos α0 + (2k + 1)b (2k + 1)r − (k + 1)a cos α0

For π/2 < α0 < π the denominator of eq. (29) is positive, since r is positive by definition,
but the numerator is negative. Hence, Ω0 is imaginary, and steady motion is not possible in
this quadrant of angle α0 .
For 0 < α0 < π/2, Ω0 is real and steady motion is possible so long as

ak cos α0
b>− . (30)
2k + 1
In addition to the commonly observed case of b > 0, steady motion is possible with small
negative values of b
A famous special case is when b = 0, and the center of mass of the disk is at rest. Here,
eq. (29) becomes
g
Ω20 = , (31)
ak sin α0
and ω10 = 0 according to eq. (28), so that

ωrel = Ω0 cos α0 , (32)

recalling eq. (11). Also, the total angular velocity becomes simply ω
~ = −Ω0 sin α0 3̂ according
to eq. (10), so the instantaneous axis of rotation is axis 3 which contains the center of mass
and the point of contact, both of which are instantaneously at rest.

6
5 Shorter Analysis of Steady Motion with b = 0
The analysis of a spinning disk (for example, a coin) whose center is at rest can be shortened
considerably by noting at the outset that in this case axis 3̂ is the instantaneous axis of
rotation. Then, the angular velocity is ~ω = ω 3̂, and the angular momentum is simply

L = I33 ω 3̂ = kma2 ω 3̂. (33)

Since the center of mass is at rest, the contact force F is just mgẑ, so the torque about the
center of mass is
dL
N = a3̂ × mgẑ = . (34)
dt
We see that the equation of motion for L has the form
dL ~
= Ω0 × L, (35)
dt
where
~ 0 = − g ẑ.
Ω (36)
akω
Thus, the angular momentum, and the coin precesses about the vertical at rate Ω0 .
A second relation between ~ω and Ω~ 0 is obtained from eqs. (2) and (3) by noting that
~ 0 , so that
~ω123 = Ω
~ω = (−Ω0 cos α0 + ωrel )1̂ − Ω0 sin α0 3̂ = ω 3̂, (37)
using eq. (4). Hence,
ω = −Ω0 sin α0 , (38)
and the angular velocity ω1 about the symmetry axis vanishes, so that

ωrel = Ω0 cos α0 . (39)

Combining eqs. (36) and (38), we again find that


g
Ω20 = , (40)
ak sin α0
As α0 approaches zero, the angular velocity of the point of contact becomes very large, and
one hears a high-frequency sound associated with the spinning coin. However, a prominent
aspect of what one sees is the rotation of the figure on the face of the coin, whose angular
velocity Ω0 −ωrel = Ω0 (1−cos α0 ) approaches zero. The total angular velocity ω also vanishes
as α0 → 0.

6 Radial Slippage During “Steady” Motion


The contact force F during steady motion at a small angle α0 is obtained from eqs. (6), (13),
(20), (28) and (31) as
b
F = mgẑ − mgr̂. (41)
ak sin α0

7
The horizontal component of force F is due to static friction at the point of contact. The
coefficient µ of friction must therefore satisfy

|b|
µ≥ , (42)
ak sin α0
otherwise the disk will slip in the direction opposite to the radius vector b. Since coefficient
µ is typically one or less, slippage will occur whenever ak sin α0 < ∼ |b|. As the disk loses
energy and angle α decreases, the slippage will reduce |b| as well. The trajectory of the
center of the disk will be a kind of inward spiral leading toward b = 0 for small α.
If distance b is negative, it must obey |b| < ak cos α0 /(2k + 1) according to eq. (30). In
this case, eq. (42) becomes
cot α0
µ≥ , (43)
2k + 1
which could be satisfied for a uniform disk only for α0 > ∼ π/3. Motion with negative b is
likely to be observed only briefly before large radial slippage when α0 is large reduces b to
zero.
Once b is zero, the contact force is purely vertical, according to eq. (41). Surprisingly,
the condition of rolling without slipping can be maintained in this special case without any
friction at the point of contact. Hence, an analysis of the motion with b = 0 could be made
with the assumption of zero friction, as discussed in secs. 8-9. In practice, there will always
be some friction, aspects of which are further discussed in secs. 10-11. From the argument
here, we see that if b = 0, there is no frictional force to oppose the radial slippage that
accompanies a change in angle α.

7 Small Oscillations about Steady Motion


We now consider oscillations at angular frequency ̟ about steady motion, assuming that
the disk rolls without slipping. We suppose that α, Ω and ω1 have the form

α = α0 + ǫ cos ̟t, (44)


Ω = Ω0 + δ cos ̟t, (45)
ω1 = ω10 + γ cos ̟t, (46)

where ǫ, δ and γ are small constants. Inserting these in the equation of motion (25) and
equating terms of first order of smallness, we find that

δ=− (Ω0 cos α0 + ω10 ). (47)
sin α0
From this as well as from eq. (44), we see that ǫ/ sin α0 ≪ 1 for small oscillations. Similarly,
eq. (23) leads to
Ω0 sin α0
γ = −ǫ , (48)
2k + 1

8
and eq. (24) leads to
ǫ̟ 2 (k + 1) = −(2k + 1)(ǫω10 Ω0 cos α0 + γΩ0 sin α0 + δω10 sin α0 ) + ǫkΩ20 (1 − 2 cos2 α0 )
g
−2δkΩ0 sin α0 cos α0 − ǫ sin α0 . (49)
a
Combining eqs. (47)-(49), we obtain
̟ 2 (k + 1) = Ω20 (k(1 + 2 cos2 α0 ) + sin2 α0 ) + (6k + 1)ω10 Ω0 cos α0
2 g
+2(2k + 1)ω10 − sin α0 , (50)
a
which agrees with Routh [1], noting that our k, Ω0 , and ω10 are his k 2 , µ, and −n.
For the special case of a wheel rolling in a straight line, α0 = π/2, Ω0 = 0, and
g
̟ 2 (k + 1) = 2(2k + 1)ω102
− . (51)
a
The rolling is stable only if
2 g
ω10 > . (52)
2(2k + 1)a
Another special case is that of a disk spinning about a vertical diameter, for which
α0 = π/2 and ω10 and b are zero. Then, eq. (50) indicates that the spinning is stable only
for
g
s
|Ω0 | > , (53)
a(k + 1)
which has been called the condition for “sleeping”. Otherwise, angle α decreases when
perturbed, and the motion of the disc becomes that of the more general case. Further
discussion of this special case is given in the following section.
Returning to the general analysis of eq. (50), we eliminate ω10 using eq. (28) and replace
the term (g/a) sin α0 via eq. (29) to find
̟2 sin2 α0
!
2 2 b
(k + 1) = 3k cos α0 + sin α0 + (6k + 1) cos α0 − (2k + 1)
Ω20 a cos α0
2
b
+ 2 2 (2k + 1). (54)
a
q
The term in eq. (54) in large parentheses is negative for α0 > tan−1 (6k + 1)/(2k + 1),
which is about 60◦ for a uniform disk. Hence, for positive b the motion is unstable for large
α0 , and the disk will appear fall over quickly into a rolling motion with α0 <
∼ 60 , after which

α0 will decrease more slowly due to the radial slippage discussed in sec. 5, until b becomes
very small. The subsequent motion at small α0 is considered further in sec. 11.
The motion with negative b is always stable against small oscillations, but the radial
slippage is large as noted in sec. 5.
For motion with b ≪ a, such as for a spinning coin whose center is nearly fixed, the
frequency of small oscillation is given by
s
̟ 3k cos2 α0 + sin2 α0
= . (55)
Ω0 k+1

9
For small angles this becomes s
̟ 3k
≈ . (56)
Ω0 k+1
q
For a uniform disk with k = 1/4, the frequency ̟ of small oscillation approaches 3/5Ω0 =
0.77Ω0 , while for a hoop with k = 1/2, ̟ → Ω0 as α0 → 0.
The effect of this small oscillation of a spinning coin is to produce a kind of rattling sound
during which the frequency sounds a bit “wrong”. This may be particularly noticeable if a
surface imperfection suddenly excites the oscillation to a somewhat larger amplitude.
The radial slippage of the point of contact discussed in sec. 5 will be enhanced by the
rattling, which requires a larger peak frictional force to maintain slip-free motion.
As angle α0 approaches zero, the slippage keeps the radius b of order a sin α0 . For small
α0 , b ≈ α0 a and eq. (54) gives the frequency of small oscillation as
s
3k + (6k + 1)α0
̟ ≈ Ω0 . (57)
k+1
For a uniform disk, k = 1/4, and eq. (57) gives
s
3 + 10α0
̟ ≈ Ω0 . (58)
5
When α0 ≈ 0.2 rad, the oscillation and rotation frequencies are nearly identical, at which
time a very low frequency beat can be discerned in the nutations of the disk. Once α0
drops below about 0.1 rad, the low-frequency nutation disappears and the disk settles into
a motion
q in which the center of mass hardly appears to move, and the rotation frequency
Ω0 ≈ g/akα0 grows very large.
For a hoop (k = 1/2), the low-frequency beat will be prominent for angles α near zero.

8 Disk Spinning About a Vertical Diameter


When a disc is spinning about a vertical diameter the condition of contact with the horizontal
surface is not obviously that of rolling without slipping, which requires nonzero static friction.
Olsson has suggested that there is zero friction between the disk and the surface in this case
[14].
If there is no friction, all forces on the disc are vertical. Then, the center of mass moves
only vertically, and there is no vertical torque component about the center of mass, so the
vertical component Lz of angular momentum is constant.
The equations of motion in the absence of friction can be found by the method of sec. 3,
writing the position of the center of mass as

rcm = a sin αẑ. (59)

Using this in eq. (22), the 1̂, 2̂ and 3̂ components of the equation of motion are

ω̇1 = 0, (60)

10
g
(kΩ2 + α̇2 ) sin α cos α + 2kω1 Ω sin α − (k + cos2 α)α̈ = cos α, (61)
a
Ω̇ sin α + 2α̇Ω cos α + 2ω1 α̇ = 0. (62)

According to eq. (60), the angular velocity ω1 about the symmetry axis of the disk is constant.
Then, eq. (62) can be multiplied by kma2 sin α and integrated to give

kma2 (Ω sin2 α − 2ω1 cos α) = Lz = constant, (63)

recalling eq. (19).


In the case of motion of a disk with no friction we find five constants of the motion,
xcm , ycm , ω1 , Lz and the total energy E, in contrast to the case of rolling without slipping
in which the only (known) constants of the motion are the energy E and the coordinates
(xA , yA ) of the center of motion.
For spinning about a vertical diameter, α = π/2 and ω1 = 0. For small perturbations
about this motion we write α = π/2 − ǫ, and for small ǫ eq. (61) becomes
g
 
2
ǫ̈ + Ω − ǫ = 0. (64)
ak
Hence, in the case of no friction, spinning about a vertical diameter is stable for
g
r
|Ω| > . (65)
ak
q
For a uniform disk with k = 1/4, this stability condition is that |Ω| > 2 g/a.
In contrast,
q the stability
q condition (53) for a uniform disk that rolls without slipping is
that |Ω| > 2 g/5a ≈ 0.9 g/a.
As the stability conditions (53) and (65) differ by more than a factor of two for a uniform
disk, there is hope of distinguishing between them experimentally.
We conducted several tests in which a U.S. quarter dollar was spun initially about a
vertical diameter on a vinyl floor, on a sheet of glossy paper, and on the glass surface of the
base of the Tangent Toy Euler’s Disk. (The Euler’s Disk is so thick that when spun about
a vertical diameter it comes to rest without falling over.)
We found it essentially impossible to spin a coin such that there is no motion of its center
of mass. Rather, the center of mass moves slowly in a spiral before the coin falls over into
the “steady” motion with small b described in sec. 5. A centripetal force is required for
such spiral motion, and so friction cannot be entirely neglected. The occasional observation
of “rising”, as discussed further in sec. 10, is additional evidence for the role of friction in
nearly vertical spinning.
Analysis of frames taken with a digital video camera [15] at 30 frames per second with
exposure time 1/8000 s did not reveal a sharp transition from spinning of a coin nearly
vertically about a diameter to the settling motion of sec.q5, but in our judgment the transition
point for Ω in several data sets was in the range 1.5-3 g/a. This suggests that during the
spinning about a nearly vertical diameter friction plays only a small role, as advocated by
Olsson [14].

11
9 Small Oscillations About Steady Motion with No
Friction
It is interesting to pursue the consequences of the equations of motion (60)-(62), deduced
assuming no friction, when angle α is different from π/2. For motion that has evolved from
α = π/2 initially, we expect the constant ω1 to be zero still. Then, eq. (61) indicates that
the value of Ω0 for steady motion at angle α0 is the same as that of eq. (31). This was
anticipated in sec. 5, where it was noted that for b = 0 no friction is required to enforce the
condition of rolling without slipping.
We consider small oscillations about steady motion at angle α0 of the form
α = α0 + ǫ cos ̟t, (66)
Ω = Ω0 + δ cos ̟t, (67)
where ǫ and δ are small constants. Inserting these in the equation of motion (62) and
equating terms of first order of smallness, we find that
δ = −2ǫΩ0 cot α0 , (68)
which is the same as eq. (47) with ω1 = 0, since eqs. (25) and (62) are the same. Similarly,
eq. (61) leads to
g
ǫ̟ 2 (k + cos2 α0 ) = ǫkΩ20 (1 − 2 cos2 α0 ) − 2δkΩ0 sin α0 cos α0 − ǫ sin α0 . (69)
a
Combining eqs. (68)-(69), we obtain
g
̟ 2(k + cos2 α0 ) = kΩ20 (1 + 2 cos2 α0 ) − sin α0 = 3kΩ20 cos2 α0 , (70)
a
using eq. (31). The ratio of the frequency ̟ of small oscillations to the frequency Ω0 of
rotation about the vertical axis for α0 < π/2 is
s
̟ 3k
= cos α0 , (71)
Ω0 k + cos2 α0
which differs somewhat from the result (55) obtained assuming rolling without slipping. For
very small α0 , both eq. (55) and (71) take on the same limiting value (56).
Because of the similarity of the results for small oscillations about steady motion with
b = 0 for either assumption of no friction or rolling without slipping, it will be hard to
distinguish experimentally which condition is the more realistic, but the distinction is of
little consequence.

10 “Rising” of a Rotating Disk When Nearly Vertical


(α ≈ π/2)
A rotating disk can exhibit “rising” when launched with spin about a nearly vertical diameter,
provided there is slippage at the point of contact with the horizontal plane. That is, the

12
plane of the disc may rise first towards the vertical, before eventually falling towards the
horizontal.
The rising of tops appears to have been considered by Euler, but rather inconclusively.
The present explanation based on sliding friction can be traced to a note by “H.T.” in 1839
[3].
Briefly, we consider motion that is primarilyq rotation about a nearly vertical diameter.
The angular velocity about the vertical is Ω > g/ak, large enough so that “sleeping” at the
vertical is possible in the absence of friction. The needed sliding friction depends on angular
velocity component ω1 = bΩ/a being nonzero, which implies that the center of mass moves
in a circle of radius b ≪ a in the present case. Then, ω1 ≪ Ω, and the angular momentum
(19) is L ≈ −Ω3̂, which is almost vertically upwards (see Fig. 1). Rising depends on slippage
of the disk at the point of contact such that the lowermost point on the disk is not at rest
but moves with velocity −ǫaω1 2̂, which is opposite to the direction of motion of the center of
mass. Corresponding to this slippage, the horizontal surface exerts friction Fs 2̂ on the disk,
with Fs > 0. The related torque, Ns = a3̂ × Fs 2̂ = −aFs 1̂, pushes the angular momentum
towards the vertical, and the center of mass of the disk rises.
The torque needed for rising exists in principle even for a disk of zero thickness, provided
there is sliding friction at the point of contact.
The most dramatic form of rising motion is that of a “tippe” top, which has recently
been reviewed by Gray and Nickel [6].

11 Friction at Very Small α


In practice, the motion of a spinning disk appears to cease rather abruptly for a small value
of the angle α, corresponding to large precession angular velocity Ω. If the motion continued,
the velocity Ωa of the point of contact would eventually exceed the speed of sound.
This suggests that air friction may play a role in the motion at very small α, as has been
discussed recently by Moffatt [7, 8, 9].
When the rolling motion ceases, the disk seems to float for a moment, and then settle
onto the horizontal surface. It appears that the upward contact force Fz vanished, and
the disk lost contact with the surface. From eqs. (13) and (20), we see that for small α,
Fz ≈ mg + maα̈. Since the height of the center of mass above the surface is h ≈ aα for
small α, we recognize that the disk loses contact with the surface when the center of mass
is falling with acceleration g.
Moffatt invites us to relate the power P dissipated by friction to the rate of change dU/dt
of total energy of the disk. For a disk moving with b = 0 at a small angle α(t),
1 1 1 3
U = mḣ2 + I33 ω 2 + mgh ≈ ma2 α̇2 + magα, (72)
2 2 2 2
using eq. (38) and assuming that eq. (40) holds adiabatically. Then,

dU 3 5
P = ≈ ma2 α̇α̈ + mag α̇ ≈ mag α̇, (73)
dt 2 2

13
where the second approximation holds when Fz ≈ 0 and maα̈ ≈ mg.
For the dissipation of energy we need a model. First, we consider rolling friction, taken to
be the effect of inelastic collisions between the disk and the horizontal surface. For example,
suppose the surface has small bumps with average spacing δ and average height h = ǫδ. We
also suppose that the disk dissipates energy mgh = mgǫδ when passing over a bump. The
time taken for the rotating disk to pass over a bump is δ/aΩ (at small α), so the rate of
dissipation of energy to rolling friction is
mgǫδ
P =− = −ǫmagΩ. (74)
δ/aΩ
A generalized form of velocity-dependent friction could be written as

P = −ǫmagΩβ , (75)

where the drag force varies with (angular) velocity as Ωβ−1 . A rolling frictional force propor-
tional to the velocity of the contact point corresponds to β = 2; an air drag force proportional
to the square of the velocity corresponds to β = 3. The model of Moffatt [7] emphasizes
the viscous shear of the air between the disk and the supporting horizontal surface, and
corresponds to β = 4. A revised version of Moffatt’s model reportedly [9] corresponds to
β = 2.5.
Equating the frictional power loss to the rate of change (73) of the energy of the disk, we
find
2ǫ 2ǫ g β/2 1
 
α̇ = − Ωβ ≈ − , (76)
5 5 ak αβ/2
using Ω2 ≈ g/akα from eq. (31), which integrates to give
β/2
ǫ(β + 2) g

(β+2)/2
α = (t0 − t), (77)
5 ak
and !2/(β+2)  β/(β+2)
ǫ(β + 2) g
α= (t0 − t)2/(β+2) . (78)
5 ak
In this model, the angular velocity Ω obeys
!1/(β+2) 1/(β+2)
5g/ǫ(β + 2)ak C

Ω= ≡ , (79)
t0 − t t0 − t

which exhibits what is called by Moffatt a “finite-time singularity” at time t0 [7] for any
value of β greater than −2.
A premise of this analysis is that it will cease to hold when the disk loses contact with
the surface, i.e., when Fz = 0, at which time α̈ = −g/a, or equivalently d2 (1/Ω2 )/dt2 = −k.
Taking the derivative of eq. (79), the maximum angular velocity is
!1/2(β+1)
k(β + 2)2
Ωmax = C 1/(β+1) , (80)

14
which occurs at time tmax given by
!(β+2)/2(β+1)

t0 − tmax = C −1/(β+1) . (81)
k(β + 2)2
In Moffatt’s model based on viscous drag of the air between the disc and the surface [7],
β = 4,
1/3
2πηa

α= (t0 − t) , (82)
m
where η = 1.8 × 10−4 g-cm−1 -s is the viscosity of air, and
!1/6
g m/2πηa
r
Ω= . (83)
ak t0 − t
This model is notable for having no free parameters.
The main distinguishing feature between the various models for friction is the different
time dependences (79) for the angular velocity Ω as angle α decreases.
A recent report [8] indicates that the total times of spin of coins in vacuum and in air are
similar, which suggests that air drag is not the dominant mechanism of energy dissipation.
Such results do not preclude that air drag could be important for disks of better surface
quality, and hence lower rolling friction, or that air drag becomes important only during the
high-frequency motion as time t approaches t0 .
To help determine whether any of the above models corresponds to the practical physics,
we have performed an experiment using a Tangent Toy Euler’s Disk [11]. The spinning disk
was illuminated by a flashlight whose beam was reflected off the surface of the disk onto a
phototransistor [16] whose output was recorded by a digital oscilloscope [17] at 5 kS/s. The
record length of 50,000 samples permitted the last 10 seconds of the spin history of the disk
to be recorded, as shown in Figs. 2-4.
The analysis of the data shown in Figs. 2-4 consisted of identifying the time ti of the
peak of cycle i of oscillation as the mean of the times of the rising and falling edges of the
waveform at roughly one half the peak height. The average angular frequency for each cycle
was calculated as Ωi = 2π/(ti+1 − ti ), as shown in Fig. 5, and the rate of change of angular
frequency was calculated as dΩi /dt = 2(Ωi+1 − Ωi )/(ti+2 − ti ). The angular frequency of the
last analyzable cycle was Ωmax = 680 Hz.
The data can be conveniently compared to the result (79) in the form
1/(β+2)
1 t0 − t

= (84)
Ω C
via a log-log plot of 1/Ω vs. t0 − t, given an hypothesis as to t0 . Inspection of Fig. 4 suggests
that t0 is in the range 7.26-7.28 s for our data sample. Figures 6 and 7 show plots of 1/Ω vs.
t0 − t for t0 = 7.26 and 7.28 s, respectively. The straight lines are not fits to the data, but
illustrate the behavior expected according to eq. (79) for various values of parameter β.
A larger value of t0 has the effect of lowering the apparent value of parameter β for the
last few cycles of the motion. Figure 4 suggests that t0 could hardly be less than 7.26 s, for
which case a value of β = 2 would fit the entire data sample rather well.

15
Figure 2: A 10-s record at 5 kS/s of the spinning of a Tangent Toy Euler’s
Disk [11] as observed by a phototransistor that detected light reflected off the
disk.

1.5

1.0
V

0.5

0.0
7.0 7.1 7.2 7.3 7.4
t (sec)

Figure 3: The last 0.25 s of the history of the spinning disk shown in Fig. 2.

In view of the uncertainty in assigning a value to the time t0 , it is interesting to ask at


what time t does the time remaining equal exactly one period of the motion, i.e., when does
t0 − t = 2π/Ω(t)? For β = 2, the answer from eq. (79) is when Ω = (C/2π)1/3 . From Fig. 6
we estimate that C −1/4 = 0.0055, and hence Ω ≈ 560 Hz when the remaining time of the
motion is 2π/Ω = 0.011 s. Recall that the last cycle analyzable in our data sample yielded
a value of 680 Hz for Ω. The preceding analysis tells us that time t0 cannot be more than
about 0.01 s after the last observable peak in the data, which suggests that t0 is closer to
7.26 than to 7.28 s, and that Fig. 6 is the relevant representation of the experiment.
For β = 2, the spinning disk is predicted by eq. (81) to lose contact with the horizontal
surface when t0 − t = C −1/3 = 0.001 s for C −1/4 = 0.0055. The instantaneous angular
frequency at that time is predicted by eq. (80) to be Ω = C 1/3 = 1030 Hz. These values are,
of course, beyond those for the last analyzable cycle in the data.
The question as to the value of t0 can be avoided by noting [18] that the time derivative

16
1.0

0.9

0.8

V
0.7

0.6

0.5
7.20 7.22 7.24 7.26 7.28 7.30
t (sec)

Figure 4: The last 0.06 s of the history of the spinning disk shown in Figs. 2
and 3.

800

600
Ω (Hz)

400

200

0
0 2 4 6 8
t (sec)

Figure 5: Ω vs. t deduced from the data shown in Figs. 2-4.

of eq. (79) yields the relation


dΩ
∝ Ωβ+3 . (85)
dt
However, dΩ/dt must be calculated from differences of differences of the times of the peaks
in the data, so is subject to greater uncertainty than is Ω. Figure 8 shows a log-log plot of
dΩ/dt vs. Ω together with straight lines illustrating the expected behavior for various values
of β. Again, β = 2 is a consistent description of the entire data sample. The value of β = 0
suggested by Fig. 7 based on t0 = 7.28 s is quite inconsistent with Fig. 8.
The results of our experiment on the time history of the motion of a spinning disk are
not definitive, but are rather consistent with the dissipated power being proportional to the
square of the velocity of the point of contact, and hence the dissipative force varying linearly
with velocity. Our experiment cannot determine whether during the 0.01 s beyond the last
full cycle of the motion an additional dissipative mechanism such as air friction with power
loss proportional to the fourth power of the velocity [7] became important.

17
–2
10

β =1

1/Ω
β =2

β =4

–3
Straight lines: Ω = [ C /( 7.26 – t ) ] 1/( β + 2)
10
–3 –2 –1
10 10 10 10 0 10 1
7.26 – t (sec)

Figure 6: 1/Ω vs. t0 − t for t0 = 7.26 s, using the data shown in Figs. 2-4. The
straight lines illustrate the behavior expected according to eq. (79) for various
values of parameter β.

–2
10

β =0
1/Ω

β =2

Straight lines: Ω = [ C /( 7.28 – t ) ] 1/( β + 2)


–3
10
–3 –2 –1
10 10 10 10 0 10 1
7.28 – t (sec)

Figure 7: 1/Ω vs. t0 − t for t0 = 7.28 s, using the data shown in Figs. 2-4.

12 Summary of the Motion of a Disk Spun Initially


About a Vertical Diameter
If a uniform disk is given a large initial angular velocity about a vertical diameter, and
the initial horizontal velocity of the center of mass is very small, the disk will “sleep” until
friction at the point of contact reduces the angular velocity below that of condition (65)
(secs. 8). The disk will then appear to fall over rather quickly into a rocking motion with
angle α < 90◦ (sec. 9). After this, the vertical angular velocity Ω will increase ever more
rapidly, while angle α decreases (sec. 5), until the disk loses contact with the table at a value
of α of a few degrees (sec. 11). The disk then quickly settles on to the horizontal surface.
One hears sound at frequency Ω/2π, which becomes dramatically higher until the sound

18
10 5
Straight lines: dΩ / dt ~ Ω β + 3
10 4

β =2
10 3

dΩ/ dt
β =1

10 2

10 1 β =4

10 0
10 2 10 3
Ω (Hz)

Figure 8: dΩ/dt vs. Ω for the data shown in Figs. 2-4. The straight lines
illustrate the behavior expected according to eq. (85) for various values of
parameter β.

abruptly ceases. But if one observes a figure on the face of the disk, this rotates every more
slowly and seems almost to have stopped moving before the sounds ceases (sec. 4).
If the initial motion of the disk included a nonzero initial velocity in addition to the spin
about a vertical diameter, the center of mass will initially move in a circle (sec. 4). If the
initial vertical angular velocity is small, the disc will roll in a large circle, tilting slightly
inwards until the rolling angular velocity ω1 drops below that of condition (52). While in
most cases the angle α of the disk will then quickly drop to 60◦ or so (sec. 6), occasionally
α will rise back towards 90◦ before falling (sec. 9). As the disk rolls and spins, the center
of mass traces an inward spiral on average (sec. 5), but nutations about this spiral can be
seen, often accompanied by a rattling sound (sec. 6). The nutation is especially prominent
for α ≈ 10 − 15◦ at which time a very low beat frequency between that of primary spin and
that of the small oscillation can be observed (sec. 6). As α decreases below this, the radius
of the circle traced by the center of mass becomes very small, and the subsequent motion is
that of a disk without horizontal center of mass motion (secs. 4 and 8).
The authors thank A. Chatterjee, C.G. Gray, H.K. Moffatt, M.G. Olsson and A. Ruina
for insightful correspondence on this topic.

13 References
[1] E.J. Routh, The Advanced Part of a Treatise on the Dynamics of a System of Rigid
Bodies, 6th ed. (Macmillan, London, 1905; reprinted by Dover Publications, New York,
1955).

[2] H. Lamb, Higher Mechanics (Cambridge U. Press, Cambridge, 1920), p. 156.

[3] H.T., Note on the Theory of the Spinning Top, Camb. Math. J. 1, 42-44 (1839). A
reprint edition of this journal, published in 1845 under the supervision of W. Thomson,

19
includes an index that identifies the author of this paper as Archibald Smith, who was
the first editor of the Camb. Math. J. (T. Crilly, private communication). This paper is
alluded to in ref. 1 of [6].

[4] J.H. Jellett, A Treatise on the Theory of Friction (Hodges, Foster and Co., Dublin;
Macmillan, London 1872), Chap. VIII, Sec. I.

[5] E.G. Gallop, On the Rise of a Spinning Top, Trans. Camb. Phil. Soc. 19, 356-373 (1904).

[6] C.G. Gray and B.G. Nickel, Constants of the motion for nonslipping tippe tops and
other tops with round pegs, Am. J. Phys. 68, 821-828 (2000).

[7] H.K. Moffatt, Euler’s disk and its finite-time singularity, Nature 404, 833-834 (2000).

[8] G. van den Engh, P. Nelson, J. Roach, Numismatic gyrations, Nature 408, 540 (2000).

[9] H.K. Moffatt, Moffatt replies, Nature 408, 540 (2000).

[10] E.A. Milne, Vectorial Mechanics (Interscience Publishers, New York, 1948).

[11] J. Bendik, The Official Euler’s Disk Website, http://www.eulersdisk.com/


Tangent Toy Co., P.O. Box 436, Sausalito, CA 94966, http://www.tangenttoy.com/

[12] K.T. McDonald, Circular Orbits Inside the Globe of Death,


http://xxx.lanl.gov/abs/physics/0008226

[13] Some websites featuring the Globe of Death are


http://www.globeofdeath.com/home.html
“The Urias” at http://www.ringling.com/bios/
http://www.sciencejoywagon.com/physicszone/lesson/03circ/sphear/sphear.htm

[14] M.G. Olsson, Coin Spinning On a Table, Am. J. Phys. 40, 1543-1545 (1972).

[15] Panasonic Model PVDV910D, Matsushita Electric Corporation of America, One Pana-
sonic Way, Secaucus, NJ 07094, http://www.panasonic.com/

[16] Model 253, Taos Inc., Plano, TX 75074,


http://www.taosinc.com/pdf/TSL253.pdf

[17] Model TDS744A, Tektronix Inc., Beaverton, OR 97077,


http://www.tektronix.com/

[18] A. Chatterjee, private communication. He performed (unpublished) experiments in


which a 20-cm-diameter disk was photographed with a high-speed camera, yielding
measurements of angle α and time t for each cycle. A plot of dα/dt vs. Ω suggested that
β was considerably less than 4.

20

You might also like