The Twisting Tennis Racket

Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Journal of Dynamics and Differential Equations, Vol. 3, No.

1, 1991

The Twisting Tennis Racket


Mark S. Ashbaugh, 1 Carmen C. Chiconc, 1'3 and Richard H. Cushman 2

Received July 19, 1989

This paper describes, analyzes, and explains a novel twisting phenomenon


which occurs in a triaxial rigid body (such as a tennis racket) when it is rotating
about an axis initially near its unstable intermediate principal axis.

KEY WORDS: Euler equations; rigid body motion; Euler angles; Eulerian
wobble.
1980 AMS(MOS) MATHEMATICS SUBJECT CLASSIFICATIONS: 70E15,
58F05.

1. I N T R O D U C T I O N
The classical treatments of the dynamics of a tennis racket about its inter-
mediate axis fail to describe a remarkable aspect of its motion which is
revealed in the following experiment. Mark the faces of the racket so that
they can be distinguished. Call one rough and the other smooth. Hold the
racket horizontally by its handle with the smooth face up. Toss the racket
into the air attempting to make it rotate about the intermediate axis
(namely, the axis in the plane of the face which is perpendicular to the
handle). After one rotation, catch the racket by the handle. The rough face
will almost always be up! In other words, the racket typically makes a half-
twist about its handle.
The experiment above was shown to one of us (R.C.) by Professor
W. Burke of the University of California at Santa Cruz. The twisting
phenomenon seems to be new. It is not mentioned in a recent article on the
Eulerian wobble (Colley, 1987), in general texts on classical mechanics
(Arnol'd, 1978; Goldstein, 1950; Landau and Lifschitz, 1976), or in

1Department of Mathematics, University of Missouri, Columbia, Missouri 65211.


2Mathematics Institute, Rijksuniversiteit Utrecht, 3508TA Utrecht, The Netherlands.
3To whom correspondence should be addressed.
67
865/3/1-5" 1040-7294/91/0100-0067506.50/0 9 1991 Plenum Publishing Corporation
68 Ashbaugh, Chicone, and Cushman

specialized texts on rigid body motion (Klein and Sommerfeld, 1897-1910;


Webster, 1920).
In this paper we explain the twist by analyzing the equations of
motion of the tennis racket in space. These differential equations, which we
call the full Euler equations (see Section 2), are given in terms of suitably
chosen Euler angles. Our treatment of the twist has two main parts. In the
first part we prove two theorems which show that the handle moves nearly
in a plane and rotates nearly uniformly (see Section 3). The near-planarity
of the motion of the handle allows us to define what we mean when we
say that the racket undergoes a "half-twist" about its handle. Namely, let
(el, ~2, ~3) be a frame which corotates with the racket and lies along its
principal axes (see Fig. 1). We say that the racket performs a half-twist
about its handle (which lies along el), if the vector ~2 (which lies along the
intermediate axis of the racket) crosses from being perpendicular to and
lying on one side of the plane of motion of the handle to being per-
pendicular to and lying on the other side of the plane of motion of the
handle. The fact that the handle rotates nearly uniformly means that when
we stop the experiment, the amount of twist will be reproducible. In the
second part, we discuss how the twist and rotation of the handle are
related. More specifically, we demonstrate that for a high percentage of
initial conditions which lead to near-rotations about the intermediate axis,
the racket will perform a near-half-twist in the time it takes the handle to
make a full rotation. This we do by analyzing a special case where the
motion is along the unstable manifold (see Section 4). In the Appendix we
prove an estimate which gives the size of a region of initial conditions
where a near-half-twist does not take place. These results are then com-
bined with numerical studies which show that, for most suitable initial
conditions, the handle does perform a near-half-twist.

2. THE F U L L EULER EQUATIONS


The problem, of course, is to explain the twist as a consequence of the
classical mechanics of a rotating tennis racket. Since the only effect of a
uniform gravitational field is to cause a uniform acceleration of the center
of mass of the racket, we can ignore this force in what follows. We
begin by discussing Euler's equations for the components of the angular
momentum M = (M1, M2, M3) of the tennis racket in a noninertial frame
(~,, ~2, ~3) which corotates with the racket and lies along its principal axes
(Fig. 1). We then make a choice of Euler angles and derive the full Euler
equations which describe the evolution of the orientation of the racket in
space.
The components of the angular momentum of a tennis racket about its
The Twisting Tennis Racket 69

~'2
Fig. 1. The principal axes of a tennis racket.

center of mass are governed by Euler's equations (Goldstein, 1950; Landau


and Lifschitz, 1976):

dM~=dt - (~2 - l l
dM2=(%_ 1)
dt \11 ~ MIM3 (1)

dM3=dt - (~ - ~) M1M2

Here I1, 12, 13 are the principal moments of inertia of the racket, which we
assume satisfy
0<11 <12<13 (2)
11 '~I2 (3)
and
Ilq-12~I3 4 (4)

4 For a standard tennis racket such as the Wilson T-2000 the values of I1,/2, and 13
are I1 =0.00121 kg-m 2, I2=0.01638 kg-m 2, and 13=0.01748 kg-m 2 as measured by Brody
(1985). These values are used in all of our numerical examples.
70 Ashbaugh, Chicone, and Cushman

Qualitatively the solutions of (1) lie on the intersection of the energy


ellipsoid
1 / M 21 M s2 + M 3e)

and the angular momentum sphere


2
M = M I +2 M ~ + M ~ (6)

because E and M 2 are conserved quantities (see Fig. 2).


Quantitatively the solutions of Euler's equations are given by Jacobi
elliptic functions (Abramowitz and Stegun, 1964; Gradshteyn and Ryzhik,
1980; Tricomi, 1953; Rauch and Lebowitz, 1973). The nonequilibrium
solutiong of (1) break up into three cases depending on whether
6=M2--212E (7)

is positive, negative, or zero. These solutions are displayed in Table I. In


the table,
/ I1(213E- M2) M/11(13- I2)
A1 = ~/ ~3----~1 ' A'x = ~/12(I3 -- I1)'

/ ( 1 3 -- I2)(M 2 -- 211 E)
Zl=/V ~ +c=Bt+c

/I2(213E-M z) /I2(M2-211 E)
A2= V I3_12 ' A'2= V L ~ '
t M x/(I3 - I2)(I 2 - I1)
722= ~ 2 -- 1113 + c = Bot + c

:/I3(M2--211 E) M/I3(I2-I1)
A3 V ' A;=

z3 = t --./(12 -- I~)(213E-- M 2) + c = B't + c


111213

Table L Solutionsof Euter's Equations

6>0 6=0 6<0

M1 gA 1cn(z~,m) elA~ sech % eA l dn(z3, m')


M2 A2sn(~,, m) ezM tanh z2 A~sn(%, m')
Ma eA 3 dn(zt, m) eleeA'3 sechz2 eA a cn(%, m')
The Twisting Tennis Racket 71

^ A

). ^e 3 C C

^
e2

Fig. 2. Solutions of Euler's equations for a fixed value of energy and angular momentum.

and
m = k 2 = (12 - 11 )(213 E - M 2) 1
(I3-I2)(M2-211E) rn'

The constants c, e, el, e2 are determined by the initial conditions. W h e n


6 r 0 and E, M are fixed, the intersection of the energy ellipsoid (5) with
the angular m o m e n t u m sphere (6) is two closed orbits of (1). The choice
of e is determined as follows. If 6 > 0, then
1, if M 3 > 0
e= - 1, if M 3 < 0

If 6 < O, then
1, if M1 > 0
e= -1, if M1 < 0

W h e n 6 = O, there are four trajectories of (1) which lie on the intersection


of the energy ellipsoid (5) with the planes

II+_ : ~/I3(I2-- I1) M 1 = +x/I1(I3 -12) M3 (8)

They are the stable and unstable manifolds of the hyperbolic equilibrium
points B: (0, 2x/~2E, 0) and B': (0, - 2x/~-E2E, 0) (see Fig. 2). The choice of
e2 = _1 determines whether we are on the stable or unstable manifold of
B, while the choice of el determines which of the two branches of these
manifolds we are on.
The Eulerian picture of the m o t i o n of the tennis racket has two
short-comings. The first one is that it is based on the noninertial frame
(el, e2, e3) which is corotating with the body. To remedy this, we recall
that M is a constant of the m o t i o n of the tennis racket when referred to an
inertial frame (X, Y, Z) fixed in space. Thus, a p p a r e n t motions of M along
72 Ashbaugh, Chicone, and Cushman

trajectories of (1) in the noninertial flame translate into rotations of the


triad (~1, ~2, ~3) with respect to M in the fixed spatial frame. In particular,
we can determine the angles between M and each Ei from this picture.
However, we cannot determine the amount of rotation about M. This
second defect can be corrected only by going to the full Euler equations
which describe the motion of the racket in the spatial frame in terms of
Euler angles.
In the remainder of this paper we use the two choices of Euler angles
which are illustrated in Fig. 3. In Euler angles I the vector M, which lies
along the Z-axis, has components

Ml=MsinOcos~/, M2=McosO, M3=MsinOsin~, (9)

in Euler angles H it has components

Ml=McosO, M2=MsinOsintp, M3=MsinOcosO (10)

The full Euler equations for Euler angles I are obtained by first solving the
equations
M1 = ~ sin 0 cos ~, - 0 sin
11

M2 = ~ cos 0 + ~b
12

M3 = ~ sin 0 sin ~ + 0 cos


13

Z /x

e3 ^

\N
II
Fig. 3. Two choicesof Euler angles. N is the line of nodes: the intersection of the ~1-~3plane
in 1 (the E2 k3 in H) with the X-Y plane
The Twisting Tennis Racket 73

for the angular velocities 0, ~, ~ about N, 82, and the Z-axis, respectively,
and then using (9). This gives

--= -M - sin 0 sin ~ cos


dt
&b = M sin2 ~ + M cos2 (11)
dt 13 11
d~ ( M Msin2~b Mcos2~t)
cos 0
d-~= ~2 13 I1
A similar argument shows that the full Euler equations for Euler angles H
are

- - = M - s i n 0 sin ~p cos ~p
dt
m 2
12
&b=MsinZO+)TC~ ~' (12)

dMOd-t( ~ Msin2
OI2 Mc~c ~ ~ )

[see Goldstein (1950) and Landau and Lifschitz (1976) for more details].

3. THE MOTION OF THE HANDLE IN SPACE


In this section we prove two theorems which show that the handle of
the tennis racket rotates nearly uniformly and nearly in a plane.
First we show that the handle moves nearly in a plane. Let

t--*(8,(t), ~2(t), 83(t))


be the time evolution of the frame in Fig. 3 governed by the full Euler
equations (t2). Thinking of ~l(t) as a point moving on the unit sphere, we
show that it stays in a small band about the equator of the sphere, which
lies on the X - Y plane. Let c~(t) be the angle between ~l(t) and the X - Y
plane. We prove the following.

Theorem 1. Given E and M, then for all t ~

0 ~<tan ~(t) ~< /I1(213E- M 2) (13)


X/ I3(M 2 - 211 E)
74 A s h b a u g h , C h i c o n e , and C u s h m a n

Proof. In Euler angles II, it follows that


[Ml(t)l
sin ~(t) = Icos 0(t)l = M

since ~(t)= 17z/2-O(t)l. Therefore we need only look for the maximum
value of IMl(t)l. The argument now breaks up into three cases depending
on whether 6 is positive, negative, or zero.
Suppose that 6 > 0. Then from the second column in Table I and the
fact that Icn(t, m)l ~< 1, we find that the maximum value of sin ~(t) is

A1 1 / I I ( 2 1 3 E - M 2)
(14)
M - M ~/ 13-11
which, after a bit of manipulation, gives (13).
Suppose that 6 < 0. Then from the fourth column in Table I and the
fact that 1 - m ~<dn(t, m) ~< 1, we find that the maximum of sin ~(t) is again
given by (14).
Suppose that 6 = 0. Then from the third column in Table I and the fact
that sech(t) ~< 1, we find that the maximum of sin e(t) is given by
A'~ /11(13-12)

which agrees with (14), because M 2-~ 212E in this case. |


If we use the approximations

M2 = 212 E and /3 ~ 11 + / 2

then the right-hand side of (13) becomes

~~--1 ~ I1
9 2 2 ~ -

using (3). Therefore c~(t) is small for a tennis racket whose principal
moments of inertia satisfy (2), (3), and (4). 5
Let fl(t) be the angle that the vector ~2 makes with the Z-axis (in Euler
angles I this is 0). Then the amount of twist that the tennis racket makes
about its handle in time t is

z(t) = fl(t) -- fl(O) ~ fl(t)

5 For the standard tennis racket, substituting M 2 = 212E and the values of Ia,/2, 13 into (13),
we find that c~(t) is at most 4.05 ~
The Twisting Tennis Racket 75

The approximation is valid because the face of the racket is initially nearly
horizontal, fl(0)~ 0. The following theorem shows that the handle rotates
nearly uniformly. Therefore, the amount of twist r(T) after one revolution
is a quantity which can be reproducibly measured. More precisely, we show
that the projection of ~l(t) on the X - Y plane moves almost uniformly.
Using Euler angles II, this means that ~b(t) (which measures the rotation
of the handle) increases nearly linearly with time. We now prove the
following.

Theorem 2. Let O(t), (J(t), t~(t) be solutions of the full Euler equations
(12). Then for all t >>.0,
M M
s-~ t <. ~(t) - ~o <. --I2 t (15)

Moreover, there are constants Jo, J1, 61 >>-O, 62 >/0 such that

-61 <~r (Jot+J1)<~ 62 (16)


and
M M
(17)
-~3 <~J ~ <~-12

Proof. Since 12 < / 3 , (15) follows immediately from the second


equation of (12).
To prove (16), we must use the explicit solutions of Euler's equations
given in Table I. We argue case by case based on the sign of 6. When 6 > 0,
after some straightforward manipulations using (10), the second equation
in (12) reads

dq) M ( M 2 - 2 I I E ) + ( 2 1 3 E - M 2 ) s n Z ( B t + c ' m )
d---/= I3(M 2 - 2IIE) + I~(213E- M z) sn2(Bt + c, m) (18)

Since sn2(t, m) is period 2K, where K = K ( m ) is the complete elliptic


integral
1
o [ (1-- t2)(1-- mt2) ] 1/2 dt

from (18) we see that d4/dt is periodic of period 2K/B. Let Jo be the
average of the right-hand side of (18) over a period, then a straightforward
change of the variable of integration gives

J _Mf~C ( M Z _ Z l a E ) + ( 2 1 3 E _ M Z ) s n 2 ( s , m )
o - K Jo I3(M 2 - 211E) + I1(213E- M 2) sn2(s, m) ds
76 Ashbaugh, Chicone, and Cushman

Let q~l(t)= ~ ( t ) - Jo t. Then ~1 is periodic of period 2K/B, since

=~(t)+jo N(t+s)ds-Jo t+
2K 2K
= ~(t/+ Jo ~ - J o t - Jo ~

=r

Let J1 be the average of ~bl over a period and put ~b2(t)= ~bl(t)- Jl. Then
~b2 is a continuous periodic function with average 0. Hence there are
constants 6~, 62/> 0 such that - 6 1 ~<~b2(t) ~<62 for all t. This establishes
(16). When 6 < 0, the argument can be carried out in essentially the same
manner. When 6 = 0, we have
dqJ=M ( I z - I 1 ) + ( 1 3 - I 2 ) t a n h 2 ( B t + c)
dt I3(12- I~) + Ia(I3- I2) tanhZ(Bt + c)

which can be explicitly integrated to give

~b(t)= C + ~ 2 t - t a n ' L~/I3(I2_I1) tanh(Bt + c)]

where C is a constant of integration. Putting J0 = 34/12 and J1 = C gives


(16) with
,
6, = 62 = t a n - LVI3(I2 - / 1 ) J

To finish the argument we need to prove (17). From the definition

Jo - 2K 3o dt

Substituting t = 2K/B into (15) and then using (19) yields


M 2 K 2K M2K
13 B <~-ff J ~ <<"I 2 B

Multiplying through by B/2K> 0 gives (17). 1

Although the theorems in this section do not prove that the tennis
racket twists, they do show that the handle's motion is close to being
The Twisting Tennis Racket 77

uniform periodic rotation in the X-Y plane. This allows us to make the
definition of half-twist presented in Section 1 meaningful and to formulate
a simple criterion for idealizing the experiment of tossing a tennis racket,
viz., the toss is completed when the handle returns close to its starting
point, as compared to an ideal toss of a racket rotating uniformly in a
plane. Once it is known that the handle returns close to its starting point,
the twist angle is measured by the Euler angle 0 (in Euler angles I).

4. AN EXPLANATION OF THE TWIST


In this section we give an explanation of the twist of the tennis racket,
Informally stated our analysis shows the following.
(a) The twist occurs when M starts near--but not too near (see the
Appendix )--the hyperbolic equilibrium point B on its unstable
manifold.
(b) There is a characteristic time for the half-twist. Moreover, the
time it takes for the handle to complete one revolution is larger
than the characteristic twisting time.
(c) Because B and B' are hyperbolic equilibrium points, trajectories
of Euler's equations traverse neighborhoods of B and B' very
slowly.

These facts constitute our explanation for the twisL Here (a) states
that the twist exist, (b) that the handle will have enough time to twist, and
(c) that the racket is likely to be caught after a half-twist.
We begin by discussing (a). To establish that a half-twist occurs as M
traverses the unstable manifold of B to B', we use Euler angles/. From (8)
and (9) note that the stable and unstable manifolds of B are given by

+/13(12--11)
(20)
tan ~ = --X/I1(13 12)

If our initial 6 = 6o satisfies (20), then (11) reduces to

dt - sin 0 sin 6 cos 6

M
m ~ m
(21)

o
dt
78 Ashbaagh, Chicone, and Cushman

where M = 2~22E. Integrating (21) gives


M
@=~o, r t+{b 0, c o s 0 = +_tanh(Bot+C)

where c is an integration constant and Bo is given in Table I. Note that 0


is the amount of twist around the handle. Even though the time taken to
go from B to B' along the unstable manifold is infinite, most of the twist
takes place in a characteristic time of 2/Bo; see Fig. 4.
In order to verify (b) we note that theorem 2 implies that the time T
needed for the handle to make one revolution is at least 2~I2/M. This is
larger than the characteristic twisting time, because

2 212/" 1113
Bo M 4-(I3-I2)(I2-I,)
212 /12+11 (using the approximation 13 ,~ I1 + I2)
~---M ~/12--11
212
-- (using 11 ~ 12)
M

COSO

B.8
~.?
B.5

0,3-
0.2:
~.~
4.1
-B2
-0.3
-OA,
-0.5~
-0,7
-8.8
-0.9
--B'6
1 . BIj , 1 ' I ' I ' I ' 1 ~ 1 ' 1 ' I ' 1 ' I ~ t

-1.8 -8.8 -ft.6 -I],4 -92 fl,I] fl,2 8.4 0.6 0,8 1.0
Fig. 4. The graph of cos 0 = - t a n h Bot. Here the characteristic time for the twist is
0.35836 sec when E = 0.32333 J. The time for one revolution of the handle is 1 sec.
The Twisting Tennis Racket 79

To verify (c) we linearize Euler's equations at the hyperbolic equi-


librium points B and B' with coordinates (0, _+x/~2E, 0) on the angular
momentum sphere (5). In the tangent plane at B or B' with coordinates
(M1, 0, M3), we obtain the system of linear differential equations

m l = '~ aM3
(22)
M3 = ~ bM1

where

1 1

Choosing new coordinates (4, t/) so that

(22) becomes

(23)

Now choose small segments on t/= d and ~ = - d , called sections,


which meet the coordinate axes and are transverse to the trajectories of
(23) which they intersect (see Fig. 5). The neighborhood transit time T N for

__ AM 3 M3=~ ~,
\ T }_o.,,
# ~ / 1

-axis

! I I\ --~IM3=~-b M 1

at B at B

Fig. 5. A neighborhood of the hyperbolicequilibrium point B.


80 Ashbaugh, Chicone, and Cushman

the trajectory of (23) through P = ( - 6 / x / 2 , 6/x/2) to pass from the section


= - d to the section t/= d is easily computed to be

2 in xf2 d
TN - ~ ~)

Returning to the original coordinates on the energy ellipsoid, the transit


time between the corresponding transverse segments, which to first order
lie at a distance D = d ~/a + b along the stable and unstable manifolds, is
to first approximation TN. Thus the neighborhood transit time for B and B'
is 2TN. The remaining part of the period of the trajectory of (1) through P
is twice the time Ts needed to go from the section on the unstable manifold
at B' to the symmetrically chosen section on the stable manifold of B. By
continuity with respect to parameters, Ts
approaches a finite limit Ts(O)
as
6 goes to 0. In fact Ts(O)represents the transit time between the sections
for the trajectory along the separatrix. Thus, the ratio of the neighborhood
transit time to the period of the trajectory has limiting value 1 as 6 goes
to 0. This makes precise the statement that a point on the trajectory spends
most of its time in a neighborhood of one of the equilibrium points B
o r B'.
This completes our demonstration that the face of the tennis racket
twists about the handle.

5. N U M E R I C A L RESULTS
To show for a large percentage of suitable initial conditions that the
tennis racket does perform a near-half-twist, we have simulated the tennis
racket experiment described in Section 1 on a computer by integrating the
full Euler equations (11). This allows us to simulate performing a large
number of experiments with a variety of initial conditions. We use the
values of 11,/2, and 13 of the standard racket and the value 0.32333 for E.
We choose the initial ~bo from [0, ~) and take small initial values of 00. The
total angular momentum M is then determined by
M2=2E( c~176176 +Tc~ 00 -~ sin2 ~9~- sin2
1 10~3

using (5) and (9). Finally, we choose ~bo so as to make ~1 (that is, the
racket handle) as near to the direction of the positive Y-axis as possible for
the given values of 0o and ~9o. This is accomplished by taking ~bo to be the
angle between (cos ~bo cos 0o, - s i n ~bo) and (1, 0), that is,
( - sin ~bo
~o = arctan \cos ~o cos OoJ
The Twisting Tennis Racket 81

Before we can proceed further we need an estimate of the amount of


scatter about 0 = 0 introduced by a person trying to toss a tennis racket
about its intermediate axis. We make the following experimental observa-
tion: tosses which lead to rotation about the ~3-axis are nearly always
stable. This result is useful, because looking at Fig. 2 we see that significant
deviations from rotation about the ~3-axis lead to trajectories which stray
far away from the equilibrium point. Since such large instabilities are not
observed, we can be sure that, with a little practice, one can toss a racket
so that its axis of rotation is within a few degrees of ~3. We assume that
this tolerance can be carried over to rotations about d2. To be specific, an
error in the toss about 63 of magnitude exceeding

tan - 1 F /I1(I3-I2!]~007073 r a d i a n t 4 . 0 5 ~
LVI3(/2 - L)J "

would lead to wide excursions. Therefore we assume that 0 o e [0, 0.025].


Using these initial conditions (0o, 0 o ) e [0, 0.025] x [0, n), we
integrated (11) until the angle of the projection of Or(t) on the X - Y plane
increased by 2n. In other words, we kept track of the polar angle of the
vector

( - sin 0 cos ~b- cos 0 cos 0 sin ~, - sin 0 sin ~b+ cos ~ cos 0 cos ~b)

Finally, we checked the value of O(t) at the stopping time. If ~2(t) was
within 27 ~ of the negative Z-axis, i.e., within 15% of completing a half-
twist, then we considered that a near-half-twist had occurred; otherwise it
had not. The results of the numerical experiments are presented in Fig. 6.
Figure 6 suggests that, to a good approximation, the region of no
near-half-twist is a strip of constant width at an angle 0 = 1.5001 radians
and centered on the stable manifold through B. Assuming this to be correct
we can derive the half-width 0.004426 for the strip from our knowledge
of how the twist occurs along the unstable manifold. This compares
reasonably well with the rigorous bound 0~< 0.00019 for the no near-half-
twist region found in the Appendix. In addition, one can use the initial con-
ditions (00, 0o) = (0.0044, 0) to show by numerical integration that the
near-half-twist occurs. This confirms Fig. 6.
Moreover, we have computed the expected percentage of the time that
the twist occurs, using the initial conditions in Fig. 6, our twist criterion,
and the formula

success ratio = 52 ......... sin 0i(60i)(60i)


Za,l poi,ts sin Oi(~Oi)(~Oi )
82 Ashbaugh, Chicone, and Cushman

Iql3

M1

Fig. 6. The black dots are those points (0 cos ~,, 0 sin 0) corresponding to initial conditions
(0, 0) in [-0,0.025] x [-0, 27z) where the racket makes a near-half-twist according to our
criterion.

We obtained the value 0.804. F u r t h e r m o r e , assuming a uniform distribu-


tion of initial angular m o m e n t u m vectors within the region displayed in
Fig. 6, the expected a m o u n t of twist is given by the fomula

expected a m o u n t of twist = ~au points (angle of twist) sin 0i(60,-)(6~i)


~ , n poi.ts sin Ot(fOi)(f~b~)

We obtained the value 2.769 radians ( ~ 1 5 9 ~ This demonstrates that a


twist is quite likely in our experiment.
The TwistingTennisRacket 83

APPENDIX

In this Appendix we give an estimate of the radius of a small ball of


initial conditions about unstable equilibrium point B such that each initial
condition in the ball results in the tennis racket not performing a near-half-
twist. The existence of such a ball about B follows from
(a) the continuous dependence of solutions of Euler's equations on
initial conditions and
(b) the fact that, if the initial condition is the unstable equilibrium,
then the racket rotates uniformly about its intermediate axis and
does not twist.
To estimate the region of no near-half-twist, we introduce the function

f(M1, M2, M3) = (~--~- / ~ ) M~ + ( ~ - ~3) M3S (24)

on the energy ellipsoid


E --/--I[M~ M s M~
~"2+
= 2 \ I 1 + I2 13 //
Computing the Lie derivative o f f with respect to the vector field defined
by Euler's equations gives

f= -4\i1 I2]\I2 M1M2M3 (25)

Using IM21 ~ M and the inequality

[M1M31 <~'~ M 1+

where

/I1 (13 -- I s)
(r = X//--~2
" 11)

we obtain the estimate

1 1
(26)

Here B o is given in Table I. Integrating (26) gives

f(t) ~<f(0) e 2"~ (27)


84 Ashbaugh, Chicone, and Cushman

By (15) the largest time required for the projection of the racket handle to
have completed one revolution is

2rcI3
t 0 --
M

Therefore if f(0) is sufficiently small, by (27)f(t) cannot be too large for


all t e [0, to].
To complete the argument we must find a bound on f(t) which
precludes the occurrence of a near-half-twist. Toward this goal, observe
that along a solution of Euler's equations the values o f f ( t ) = 2 determine
a family of ellipses o~;.in the M1 - M3 plane. If for every t e [0, to], gx lies
in the interior of the ellipse
m 2 2
~: 2E = 1 + M3 (28)
11 13

which is the intersection of the energy ellipsoid (5) and the plane { m 2 = 0 } ,
then the solution M(t) of Euler's equations on the energy ellipsoid does not
cross the {M2 = 0} plane. Thus no near-half-twist can occur. Since

13 - / 2 ~<I2 -- 11

which follows from the approximations (3) and (4), we have

I 3 J \ 11 + 13 /t

For d~ to lie in the interior of g, we need

Therefore if

f(O)<213E(1-1~ e - 2~~176 (29)


\12 I3J
no near-half-twist will occur. Using Euler angles I (9) one finds that

sn00)sin0o
The Twisting Tennis Racket 85

where (0o, ~bo) are the initial values of (0, ~b). Combining (29) and (30), we
see that no near-half-twist occurs if

Msinuo< /
/211E(I3-12)
~ exp[-2~/
13/(I3-I2)(I2-I,!]
/~ _~ (3i)

For the standard tennis racket with E = 0.32333 J and M 2 = 212E,(25)


gives
0o ~<0.000190 radian

This bound is rigorous when M2<212E. This covers 95% of all initial
conditions due to the small size of the acute angle sectors at B in Fig. 2.
Even when M 2 > 212Ethe corrections to this estimate will be very small.

ACKNOWLEDGMENTS
We are grateful to Howard Brody of the University of Pennsylvania,
Enoch Durban of Princeton University, and Doug Wisbey of Wilson
Sporting Goods for sharing their knowledge of the moments of inertia of
various tennis rackets with us. We would also like to thank William Burke
of the University of California at Santa Cruz and Tom Kane of Stanford
University for their helpful criticism of the first draft of this paper.
Our computer experiments were carried out on an IBM 4381-R14
using the IMSL routine DGEAR.
This research was supported by The Air Force Office of Scientific
Research under Grant AF-AFOSR-89-0078 (C.C.C.).

REFERENCES
Abramowitz, M., and Stegun, I. A. (eds.) (1964). Handbook of Mathematical Functions,
Applied Mathematics Series, Vol. 55, National Bureau of Standards, Washington, D.C.
Arnol'd V. I. (1978). Mathematical Methods of Classical Mechanics, Springer-Verlag,
New York.
Brody, H. (1985). The moment of inertia of a tennis racket. Phys. Teach. April, 213-216.
Colley, S. J. (1987). The tumbling box. Am. Math. Month. 94, 62-68. [See also (1987). Letter
to the editor. Am. Math. Month. 94, 646.]
Goldstein, H. (1950). Classical Mechanics, Addison-Wesley, Reading, Mass.
Gradshteyn, I. S., and Ryzhik, I. M. (1980). Table of Integrals, Series and Products, corrected
and enlarged ed., Academic Press, New York.
Klein, F., and Sommerfeld, A. (.1897-1910). (Yber die Theorie des Kreisels (4 vols,), B.G.
Teubner, Leipzig.
Landau, L. D., and Lifshitz, E. M. (1976). Mechanics, 3rd ed., Pergamon Press, Oxford.
Rauch, H., and Lebowitz, A. (1973). Elliptic Functions, Theta Functions, and Riemann
Surfaces, Williams and Wilkins, Baltimore.
Tricomi, F. G. (1953). Differential Equations, Blackie and Son, London.
Webster, A. G. (1920). The Dynamics of Particles and of Rigid, Elastic, and Fluid Bodies,
Stechert-Hafner, New York.

865/3/1-6

You might also like