You are on page 1of 9

A classical experiment revisited: The bounce of balls and superballs

in three dimensions
Antonio Domenecha)
Department of Analytical Chemistry, University of Valencia, Dr. Moliner, 50, 46100 Burjassot (Vale`ncia),
Spain

Received 14 January 2004; accepted 22 July 2004


A description of the inelastic collision of a ball when it bounces on a rigid horizontal surface with
arbitrary initial spin conditions is presented. I consider cases where rebound occurs with and without
sliding, and when the ball grips the surface. Two rebound models are discussed in which friction is
described in terms of the static and dynamic coefficients of friction and the friction is expressed in
terms of horizontal coefficients of restitution. The azimuthal and vertical deviation angles of the ball
after impact are predicted as a function of the incident angle. We present data from an experiment
in which a ball is launched horizontally from the edge of the laboratory bench and then rebounds on
a horizontal surface. Ordinary balls exhibit two rebound regimes, with and without sliding, and can
be satisfactorily described by the first model. Superballs exhibit a grip behavior whose description
requires the use of the second model. 2005 American Association of Physics Teachers.
DOI: 10.1119/1.1794755

I. INTRODUCTION
A common goal in ball sports is to get a ball to bounce at
an oblique angle on a rigid surface. This goal is of interest in
golf and tennis, but also in basketball, soccer, and handball.
The study of collisions also is relevant in the study of nonlinear dynamical systems1 and granular matter.2,3
Approximate solutions of the dynamics of a ball bouncing
on a floor have recently been described. Brancazio4 analyzed
the bounce of a basketball having an initial forward/
backward spin by assuming that the collision is perfectly
elastic and that the ball does not skid just after the impact.
Brody5 studied the bounce of a tennis ball and assumed that
the collision is inelastic in the vertical direction and completely inelastic in the horizontal direction. Garwin6 described the bounce of a superball by assuming that the collision is perfectly elastic in both the vertical and horizontal
directions. More recently, Cross79 described the oblique impact of a ball of mass on a block and considered the impact
to be inelastic. Following Ref. 9, the consideration of friction
forces leads to three possible regimes of motion: pure sliding, slide then roll, and slide then grip. The grip or slip-grip
behavior occurs in some instances, namely, in the rebound of
tennis balls and superballs under certain conditions. The
grip-slip behavior is characterized by the appearance of a
large spin after an impact.6,8,9
Alternatively, Maw et al.10 described the oblique bounce
of a solid elastic sphere in terms of a numerical model. In
this approach, the contact circle is divided into small annuli,
some of which grip the surface and some of which slip leading to results similar to those of Cross.9 The models of Refs.
9 and 10 provide theoretical predictions in close agreement
with experimental data for the oblique bounce of tennis balls,
golf balls, baseballs, and basketballs without initial spin on
smooth and rough surfaces. In this context, Metha and Luck
studied the evolution of a bouncing ball on a vibrating platform and expressed the inelasticity of the impact in terms of
the classical coefficient of restitution.1
In this paper, a ball with an initial arbitrary spin that
bounces obliquely on an infinitely massive surface is studied.
It is well known that a ball launched forward with an initial
28

Am. J. Phys. 73 1, January 2005

http://aapt.org/ajp

forward spin gains speed after the bounce, whereas a ball


initially launched with backspin loses considerable speed after the bounce and may even reverse its direction of motion,
that is, it may bounce backward.4 If an additional spin along
a horizontal axis parallel to the horizontal velocity of the ball
is imparted to the ball, it deviates from the initial direction of
motion. As a result, a variety of rebounds can be obtained.
These effects must be attributed to friction forces acting
during the impact, and were first treated by Whittaker in
190411 and described in detail by Keller.12 Classical approaches to the inelastic impact with friction of rigid bodies
have been developed by Brach13 and Kane and Levinson.14
In these formulations,1214 the inelasticity of the impact is
expressed in terms of the classical coefficient of restitution,
e, which is the negative of the ratio of the relative normal
velocity after impact to that before impact.15
These authors describe two possible regimes of impact
derived from the action of friction forces with and without
sliding between the contacting surfaces. When sliding exists
between the colliding bodies, the friction forces are expressed in Ref. 14 in terms of the classical static and dynamic coefficients of friction and the coefficient of restitution; Ref. 13 uses an equivalent coefficient of friction with
similar meaning and adds a moment coefficient of restitution
to the classical coefficient of restitution. Following Brach,13
the moment coefficient of restitution can be defined as the
ratio of the angular velocities of the colliding bodies after
and before the impact. In both cases it is assumed that for
sufficiently large friction forces, a nonsliding regime is attained; in this regime the relative tangential motion of the
contacting points ceases during the impact.
Following these approaches, the impulsive forces acting
throughout the oblique rebound of a ball can be described as
a normal retarding force, F n , plus a tangential friction force,
F f . For large incident angles, relative tangential motion always must exist during the time of contact between the ball
and the horizontal surface; after the contact the ball slides
throughout the bounce.
The approach presented here compares two alternative formulations of the impact between rigid bodies to describe the
general case in which a ball having an arbitrary spin is
2005 American Association of Physics Teachers

28

launched against an infinitely massive horizontal surface.


The problem can be formulated in terms of classical friction
forces. If we use the classical formulation first developed by
Amontons and further developed by Coulomb,16 friction
forces are treated as proportional to the normal retarding
force. In this case a collision in the sliding regime is obtained
and the ball rebounds with a sliding plus rolling motion.
Alternatively, a generalized non-AmontonsCoulomb regime is described by coefficients of restitution in three dimensions, e x , e y , e z . This description is consistent with the
approaches of Ref. 13 in which a moment coefficient of restitution is defined, Refs. 79 in which horizontal and vertical
coefficients of restitution are introduced,79 and Ref. 3 which
describes the impact of small, nearly identical spherical particles in terms of the Newtonian coefficient of normal restitution and coefficients representative of the frictional properties of contact surfaces with and without negligible sliding.
As in these approaches,714 the coefficients of restitution
and friction will be assumed to be constant. This assumption
is a simplification; the coefficients of restitution and friction
depend on several factors, including the elastic properties of
the materials, surface state, relative velocity, and the ball
diameter.15,17 The validity of this assumption will be discussed in relation to experimental data.
We follow the methodology previously developed for the
impact of disks18 and billiard balls,19 and predict the vertical
and azimuthal angles of rebound from the incident angle for
arbitrary angular rotation rates in each one of the regimes,
sliding, nonsliding, and grip-slip.
The bounce of a uniform sphere launched after rolling
without slipping along a horizontal plane is typically used as
an example of inelastic collisions. In a classic experiment, a
rolling ball is projected horizontally from the edge of the
laboratory bench and then rebounds from a horizontal surface. The horizontal distances x o and x covered prior to the
first and second bounces satisfy the relation x2ex o , where
e is the coefficient of restitution for the impact between the
ball and the floor. A generalized version of this experiment
that can be used in classroom discussions and intermediate
lab experiments is described. Different initial spins are imparted to the incident ball by the collision with an auxiliary
cue ball. The angles of impact and rebound and the distances
x and x o can easily be measured to test the theory.
II. THEORY
A. General equations of motion
Let us consider a homogeneous sphere of mass m and
radius R launched obliquely with initial center-of-mass linear
velocity v o against an infinitely massive horizontal surface.
As can be seen from Fig. 1, if the sphere is launched along
the x axis, the velocity components of the center of mass of
the sphere are v ox v o sin , v oy 0, and v oz v o cos . After the impact, the sphere travels with a velocity v in a direction that may be separated from the xz plane characterized by a vertical angle and an azimuthal angle, . The
rebound velocity components, v x v sin cos , v y
v sin sin , v z v cos , can be derived from the time integrals of the force impulse over the time of impact:

29

F n dtm v cos v o cos ,


Am. J. Phys., Vol. 73, No. 1, January 2005

Fig. 1. Schematic of the bounce of a ball on an infinitely massive horizontal


surface for arbitrary initial spin. V ox , V oy represent the horizontal components of the velocity of the ball just before the rebound event.

F f x dtm v sin cos v o sin ,

F f y dtm v sin sin .

The normal component of the velocity of the point of contact of the sphere after the bounce with the horizontal plane
will be e times its value prior to the impact, that is, v cos
evo cos . Hence, the integrated normal retarding force
becomes:

F n dtm v o 1e cos .

If the incident sphere has initial angular velocities ox , oy ,


oz along the x, y, z axes, the direction of the horizontal
components of the friction force, F f x , F f y , will depend on
the direction of the horizontal components of the velocity at
the point of contact just before the bounce and are given
by v ox v ox R oy , v oy v oy R ox . If we assume, as has
been previously discussed,19 that the direction of the friction
force is the same as that of the horizontal velocity at the
point of contact at the beginning of the bounce see Fig. 1,
we can write F f x F f cos , F f y F f sin , where,
tan

Ffy
R ox

.
F f x v ox R oy

The direction of the friction force depends on the initial spin


as described in Ref. 4. In particular, if the ball is thrown with
pure rolling motion, R ox 0, R oy v o , there is no friction force at the point of contact.
In general, a small deformation occurs at the contacting
region. It is assumed that the normal force acts vertically
through a line passing a distance r behind the center of mass
of the sphere.9 Then, an additional torque appears which will
modify the angular velocity of the ball. For simplicity, it will
be assumed that rR; that is, the deformation effects will be
neglected. Then, conservation of angular momentum
yields,20
I x xo m v y v oy ,

I y oy m v x v ox ,

where I(2/5)mR 2 is the moment of inertia about an axis


through the center of the sphere, and x , y , z , are the
components of the angular velocity of the ball just after the
Antonio Domenech

29

bounce. This treatment can easily be extended to balls used


in sports by replacing the moment of inertia of a uniform
solid sphere by that for a thin spherical shell, I
(2/3)mR 2 .
If we combine Eqs. 17, we obtain a description of the
motion of the ball after the bounce provided that the relation
between F f and F n is known. We will consider two descriptions of the rebound, namely, a AmontonsCoulomb-type
friction force and a normal coefficient of restitution the -e
model, and non-AmontonsCoulomb friction force the e-e
model. The latter model introduces coefficients of restitution
in three dimensions, e x , e y , e z , the former being the same
as the usual coefficient of restitution in the normal direction,
e. As a result, the friction force becomes proportional to the
horizontal velocity of the incident ball before the rebound.
The case of impact without sliding can be derived from both
-e and e-e models.
In the following, equations for the linear and angular velocities and rebound and azimuthal angles after the impact
will be given for the case in which ox 0. Note that in all
cases, the time integral of the normal force equals
m v o (1e)cos , where v z e v o cos , and z oz .
B. Rebound with AmontonsCoulomb friction
If we assume that friction can be described by a
AmontonsCoulomb force when sliding exists at the rebound, the friction force can be taken to be times the
normal force, that is, F f F n . We combine this relation
with Eqs. 17 and obtain:
v x v o sin 1e cos cos ,

v y v o 1e cos sin ,

5
R x R ox v o 1e cos sin ,
2
5
R y R oy v o 1e cos cos .
2

10
11

12

and the angle of rebound becomes


tan

tan 1e cos
.
e cos

13

Equations 813 are notably simplified for the most common case in which ox 0, where tan 0, and,
tan 1/e tan 1e cos /e.

14

Note that there are two possible situations. For a large forward spin, v ox R oy 0, cos 1, and the rebound angle
is larger than the impact angle. For backward or small forward spins, v ox R oy 0, cos 1, and the friction force is
opposite to the translational motion. The rebound angle
equals the impact angle when tan (1e)/(1e). As a
special case, when v ox R oy 1, that is, if the incident ball
is launched horizontally with pure rolling motion, no friction
force arises, and then 0 and tan (1/e)tan . This result
comes from Eq. 13 by taking 0.
30

Am. J. Phys., Vol. 73, No. 1, January 2005

tan 5/7e 2/7e R oy / v o sin tan .

15

The transition from the sliding regime of rebound to nonsliding occurs when the friction force F f o F n , where o
is the static coefficient of friction. The transition occurs at a
limiting impact angle, L , given by
tan L 7/2 o 1e cos / 1R oy / v o sin .
16
C. Rebound without AmontonsCoulomb friction
Let us consider friction forces that are not described by the
AmontonsCoulomb law. To obtain a description of such
forces, the horizontal coefficients of restitution, e x , e y , can
be defined in a way similar to the normal coefficient of restitution, ee z . 15 That is, the horizontal restitution coefficients will be defined as the proportionality constants between the relative horizontal components of the velocity of
the point of contact before and after the impact:
e x
e y

v x R y
,
v ox R oy

17

v y R x
.
R ox

18

This formulation replaces the AmontonsCoulomb force by


a friction force defined from these coefficients of restitution.
From Eqs. 17, 17, and 18, we obtain:
v x

5 2
2
e x v o sin 1e x R oy ,
7 7
7

2
v y 1e y R ox ,
7

R x

19
20

2 5
5
R y 1e x v o sin e x R oy ,
7
7 7

The azimuthal angle is given by,

1e sin
,
tan
tan 1e cos

When nonsliding takes place at the rebound, v x R y


0, v y R x 0, and the linear and angular velocities after
the impact can be obtained from Eqs. 17. If R ox 0,
the rebound angle satisfies

2 5
e R ox .
7 7 y

21
22

From Eqs. 1922 we obtain the corresponding general


expressions for the horizontal and vertical angles of rebound:
tan

2/7 1e y R ox
,
2/7
1e

x R oy 5/72/7e x v o sin

23

tan

5/7 2/7 e x tan 2/7 1e x R oy / v o cos


.
e z cos

24
We note that, in general, sliding exists between the points of
contact of the ball and the floor. If nonsliding occurs see
Eqs. 17 and 18, e x 0 and e y 0, and for R xo 0, Eq.
24 reduces to Eq. 15.
Different rebound regimes can be defined depending on
the values of the coefficients of restitution. These regimes
Antonio Domenech

30

els from angle measurements and from the measurement of x


and x o . For rebounds with sliding, the -e model yields

x2e xo 1

1e g
xo ,
u 2o

25

where the sign corresponds to rebounds with v ox R oy


0, and the sign to rebounds with v ox R oy 0. Equation 25 predicts that the ratio x/x o provides a linear function of x o and is equal to 2e at x o 0.
The e-e model leads to the general relation:
x2ex o
Fig. 2. Schematic representation of the experimental arrangement. The ball
travels a distance x o while dropping a vertical distance h, and then rebounds
traveling a distance x along the horizontal surface.

will be discussed for the case in which R ox 0 and R oy


0. Then, 0, that is, there is no horizontal deviation of
the ball after the rebound. We distinguish:
a

The ordinary regime, e x 0, where the absolute value


of v x is larger than R y ; the ball slips along the
bounce and a relatively low spin is acquired during the
impact.
The nonsliding regime, e x 0. In this case the ball
abandons the horizontal surface with pure rolling motion, R y v x . The angle of rebound is tan
(5/7e)tan , in agreement with Eq. 15. The horizontal velocity and the angular velocity become, respectively, v x (5/7) v o sin, R y (5/7) v o sin .
The grip or superball regime, e x 0 in which a
large spin is imparted to the ball throughout the impact:
R yvx .

5 2
2
e 1e x R oy .
7 7 x 7

26

Equation 26 predicts that a plot of x/x o as a function of x o


is a straight line with zero slope and a value at the origin that
depends on e x and oy .
For rebounds without sliding, both the -e and e-e models lead to

b c

5
2
x2ex o
7
7

R oy
.
uo

27

Here, the ratio x/x o is independent on x o and depends on the


coefficient of restitution and the initial angular velocity oy .
As expected, both the -e and e-e models predict if a ball
is launched with initial pure rolling motion, that is,
R oy /u o 1, x/x o 2e, in agreement with the well-known
experiment of elementary mechanics. Hence, the postrebound linear and angular velocities of a ball launched with
a particular value of R oy can be calculated from the actual
x and x o values and those measured when that ball is
launched with R oy 1. We denote these values as x * and
x o* and write
v x u o x/x o xo * /x * ,

28

v y g x o /2u o x * /x *
o ,

29

R y R oy 5/2 1 x/x o x *
o /x* ,

30

The ideal elastic frictionless case is given by e x e y


1, e z 1. Accordingly, tan tan , that is, the angle of
rebound equals the angle of impact. This situation is equivalent to that described in the -e model by considering e
1 and 0.
We emphasize that the -e and e-e models differ significantly in the expression for the friction forces. For the case in
which the ball is initially launched without spin, the -e
model predicts an integrated force of m v o (1e z )cos ,
while the e-e model predicts a value of m v o (2/7)(1
e x )sin . Given the implicit assumption that the friction
and restitution coefficients are constant, it appears that there
is no possibility for reconciling the two formulations.

Equations 2830 are independent of the nature of the


friction force and, consequently, can be used for testing the
-e and e-e models from measurements of the linear and
angular velocities. Additionally, such velocity measurements
can be used to determine the dependence of the coefficients
of restitution and friction on different parameters such as the
velocity and the diameter.

D. The horizontal launch experiment

III. EXPERIMENTAL ARRANGEMENT

To test the different models a classic experiment was done.


As depicted in Fig. 2, it consists of the horizontal launch of
a ball from a horizontal track. After the ball leaves the track,
the ball travels through the air and strikes the floor. To impart
different spins to the ball, an auxiliary cue ball was used. As
described in Sec. III, the impact and rebound angles can
easily be measured as well as the horizontal distances covered by the ball before (x o ) and after (x) its rebound. If the
initial velocity of the ball, u o v o sin , remains constant,
experiments with variable height can test the proposed mod-

The experimental arrangement is depicted schematically


in Fig. 2. A manufactured aluminum track width 1 cm was
placed on a lab bench, with one end of the track flush with
the edge of the bench. The balls were first allowed to roll
from rest along the track with the help of an auxiliary slanted
rail. After rolling off the track, they traveled through the air,
finally striking the floor, thus obtaining bounces with initial
forward spin.
To obtain rebounds with initial backspin, a ball, initially at
rest, was struck head-on by an auxiliary cue ball rolling from

31

Am. J. Phys., Vol. 73, No. 1, January 2005

where the integrated friction force can be calculated from:

F f x dtmu o 1 x/x o x *
o /x * .

Antonio Domenech

31

31

the inclined track. The ball was placed at a known distance d


from the end of the track, thereby obtaining different values
of oy with ox 0. To study rebounds without initial spin,
the ball was launched from the edge of the track with the
help of a spring-operated shooter. To obtain the condition
R oy / v ox 1, the track was replaced by a steel tube of diameter 5 cm.
Data were obtained by varying the height of the bench but
ensuring that a constant velocity at the bench edge occurs.
The initial velocity of the ball was adjusted in each series of
experiments to 1.000.05 m/s. In all the experiments, the
incident and rebound angles were measured as well as the
horizontal distances covered by the ball before its rebound.
The trajectories of the balls were determined from photographs recorded by a camera placed either in a vertical and
horizontal position following the method described in Refs.
18 and 19. Steel ball bearings of diameters 1.22, 1.28, 1.51,
1.71, 2.00, 2.50, and 2.85 cm, a golf ball mass 45.6 g, diameter 4.26 cm and a superball mass 46.4 g, diameter 4.60
cm were used. To test the cases in which small and large
surface deformation occurs, experiments were performed on
a ceramic floor, a hard wood panel, and a plastic-type floor.
The effect of air resistance could be neglected because of the
balls small size, high density, and short flight time.21
IV. RESULTS AND DISCUSSION PRIVATE
A. The sliding-nonsliding transition
The existence of a transition from a rebound without sliding to a rebound with sliding was tested. In Fig. 3 the experimental values of the rebound angles are shown for a a
steel bearing, b a golf ball, and c a superball bouncing on
a polished ceramic floor without initial spin (R ox 0,
R oy 0) with different incident angles but the same horizontal velocity at impact. If sliding exists in these circumstances, the -e model predicts from Eq. 14 that plots of
tan versus tan will yield straight lines of slope 1/e and
ordinate at (1e)/e. In contrast, the e-e model predicts
from Eq. 24 that plots of tan versus tan will yield
straight lines passing through the origin with slope 5/7e z
(2/7)(e x /e z ). The nonsliding regime, common to both
models, corresponds to linear tan versus tan plots passing
through the origin with slope 5/7e see Eq. 15.
For the steel bearing Fig. 3a, the data obtained at low
impact angles fit well with Eq. 15 by using e0.78. For
incident angles larger than approximately 24, the data agree
well with the theoretical sliding graph obtained by letting
e0.78 and 0.07 to Eq. 14. The agreement between the
e values calculated for the two regions suggests that the -e
model applies, and the sliding regime occurring at high impact angles.
If we assume that the value of the static coefficient of
friction 0 is close to that of the dynamic coefficient , the
theoretical critical incident angle at which the transition from
one rebound regime to another is close to 24. The data do
not allow for the observation of a discontinuity in the tan
versus tan plot, required by the condition represented by
Eq. 16, thus suggesting that the value of 0 must be close
to that of .
A similar result was observed for the golf ball Fig. 3b.
Here, the data agree with the theoretical prediction with
e0.90, 0.12. With these parameter values, the transi32

Am. J. Phys., Vol. 73, No. 1, January 2005

Fig. 3. Experimental data filled circles and theoretical predictions for the
rebound of a a steel bearing, b a golf ball, and c a superball on a
ceramic floor without initial spin. Continuous lines: theoretical plots from
Eq. 15 using a e0.78 and b e0.90. The dotted lines correspond to
Eq. 15 with a e0.78, 0.07 and b e0.90, 0.12. The continuous line in c corresponds to Eq. 24 with e x 0.50.

tion angle is 39. Again, the response at low impact angles


is close to that predicted by the nonsliding condition for both
models, that is, by taking e x 0 with e z 0.90.
However, for the superball, only one well-defined region
was observed in the tan versus tan plots see Fig. 3c,
corresponding to a straight line passing through the origin.
By applying the nonsliding condition, we obtain an unrealistic e z value of 1.17, suggesting that this regime does not
occur under our experimental conditions. If we use e z
0.96, close to that reported for the rebound of superballs
against hard surfaces,6,8,9 our experimental data are consistent with Eq. 24 if we use e x 0.50. These values are reasonable, suggesting that the e-e model holds for the rebound
of superballs.
Antonio Domenech

32

Fig. 4. Comparison of experimental x/x o ratios for steel balls bounced on a


floor with predicted values using e0.78 and 0.07 in Eqs. 25 and 27.
a experimental data S 1 : diameter d1.28 cm, initially in pure rolling
motion along the horizontal track, R oy / v ox 1.60; b experimental data
S 2 : d2.50 cm, initially in pure rolling motion, R oy / v ox 1.09; c experimental data S r : d2.50 cm, R oy / v ox 0; d experimental data S s :
d2.50 cm, R oy / v ox 0.21; e experimental data S o : steel bearing
(d2.50 cm) launched with pure rolling motion with R oy / v ox 1.00. Inset: expected plots for o .

B. Influence of the initial spin


Experimental data concerning the horizontal distances
covered by the ball before and after the rebound provide a
sensitive method for determining not only the transition from
sliding to nonsliding rebounds but also the influence of the
initial spin. This experiment is of pedagogical interest for
illustrating the dependence of the direction of the friction
force on the initial translation/spin conditions.
As shown in Fig. 4, the data for x/x o versus x o for the
rebound of steel bearings on a ceramic floor can be divided
into two linear regions, corresponding to collisions with and
without sliding. Data points S 1 and S 2 in Fig. 4 corresponds
to the collisions of steel bearings of diameter 1.28 and 2.50
cm, respectively, launched on a hard wood surface after rolling along the horizontal track. The balls are launched with
forward spin R oy / v ox equal to the ratio R/R e , 22 where R e
is the effective radius of the sphere, defined as the distance
from its center of mass to the line joining the points of contact with the track. It can easily be demonstrated that R e
(R 2 a 2 /4) 1/2, where a is the width of the track.22 The
data in the sliding region fits Eq. 25 if we let e0.78 and
0.07 in both cases. The data in the nonsliding region also
fit the predicted results if we set e0.78 in Eq. 27. Data S r
corresponds to a steel bearing launched without initial spin
with the help of a spring-operated shooter, whereas data S s in
Fig. 4 corresponds to a steel bearing launched after being
struck by an auxiliary cue ball. Again, the data are consistent
with theory using the above values of e and . For data
33

Am. J. Phys., Vol. 73, No. 1, January 2005

points S s , with e0.78 in Eq. 27, the data for the nonsliding region lead to R oy / v ox 0.21, which is a reasonable
value.23 Experimental data S o corresponds to the rebound of
a steel bearing launched with pure rolling motion along a
tube, that is, R oy / v ox 1. The data fit the expected curve
for zero friction forces with e0.78. The inset in Fig. 4
shows the theoretical dependence for the corresponding
cases. The jump that separates the linear regions occurs
because the static and kinetic friction coefficients differ from
each other. When the calculations are performed with equal
values of the two friction coefficients, these discontinuities
disappear, but discontinuities in the slope remain at points of
transition from collisions involving sliding at the instant in
which the contact between the sphere and the surface ceases
to those not involving sliding at this instant.
The data for golf balls also agree with the -e model
under our experimental conditions. From our data, no satisfactory estimates of the static coefficient of friction were
obtained because no clear discontinuity was obtained between the lines fitting experimental data to the sliding and
the nonsliding regions. However, the data limited it to more
than 20% of the kinetic coefficient in all cases.
The linearity in the x/x o versus x o plots in Fig. 4 as well
as the tan versus tan plots in Fig. 3 suggests that the
assumption that both friction and restitution coefficients are
constant is reasonably valid under our experimental conditions. Also in agreement with theory, the ordinate of the x/x o
versus x o graphs in the -e region is equal to 2e regardless
of the initial spin conditions and the slope is (1
e)g/u 2o for v ox R oy 0 or (1e)g/u 2o for v ox
R oy 0. The consistency of the values of e and calculated for different steel balls under different initial spin conditions support the idea that these coefficients can be treated
as constants.
C. The superball regime
The post-rebound spin is the quantity most sensitive to
changes in the impact regime. The values of R oy /u o , calculated from x, x o , x * , x *
o see Eq. 30 for a steel bearing
bouncing without initial spin on a wood surface, are shown
in Fig. 5. At low impact angles these values are consistent
with theory for a nonsliding rebound line a, R oy /u o
5/7), whereas for relatively large impact angles, the -e
model Eq. 11 applies with e0.52 and 0.10 line c.
A discontinuity appears in the R oy /u o values near
18. This discontinuity can be associated with the static
coefficient of friction o 0.150.02 see curve d.
However, for the superball, the ratio R oy /u o circles in
Fig. 5 remains consistently equal to the value (R oy /u o
1.07) from the e-e model Eq. 21 with e x 0.50 as can
be seen in Fig. 5 line b. This situation corresponds to grip
behavior. The transition from this grip regime to an ordinary
sliding regime occurs at 76, consistent with the values
e0.96, 0.95, depicted in curve e.
The rebound of superballs is characterized by its large
elasticity restitution coefficient close to one and by the appearance of large friction forces resulting in large postimpact spins. Prior data indicate that the response of superballs cannot be satisfactorily described within the -e
model.9 To test the application of the e-e model, we compare
its predictions with published data on post-rebound linear
Antonio Domenech

33

Fig. 5. Variation of R y /u o for balls bouncing on a wood surface. The balls


were projected horizontally without an initial spin at a constant horizontal
velocity of 1.000.05 m/s. Points: steel ball, diameter 2.50 cm; circles:
superball. a theoretical plots for a nonsliding regime, b e-e grip regime
with e x 0.50; c -e sliding regime with e0.52, 0.10; d e
0.52, 0.15; and e e0.96, 0.95.

and angular velocities. The data of Ref. 9 corresponds to


experiments in which a golf ball and a superball bounce
without initial spin on a hard surface. In contrast with our
experiments, the horizontal velocity of the ball before the
bounce was not constant and, hence, the ratio R y / v o was
used rather than R y /u o .
In Fig. 6 the data taken from Ref. 9 for R y/ v o for a golf

ball filled circles and a superball circles are compared


with the theoretical expectations from the nonsliding regime
a, the e-e model with e x 0.50 line b, and the ideal
response described by Garvin line c.6 The predictions for
the AmontonsCoulomb regime from the -e model Eq.
11 are represented as dotted lines using the and e values
reported in Ref. 9. For the superball e0.96, 1.0 line
f, and for the golf ball e0.90, 0.18 line d. Line
e corresponds to the values e0.90 and 0.22, a value
that is representative of the static coefficient of friction for
the rebounds of the golf ball. The line g marked by arrows
in Fig. 6 corresponds to the variation of R oy /u o with the
incident angle of the golf ball, assuming that a transition
from the -e sliding regime to the nonsliding regime occurs.
As can be seen in Fig. 6, excellent agreement exists between
theory and experimental data using the previously mentioned
values of e, , and o .
The experimental data for the rebound of a superball on a
hard surface reported in Ref. 9 are consistent with predictions of the e-e model Eq. 21 by using e x 0.50, clearly
indicating that this model must be used for rebounds with
large friction. The transition from this regime to a -e sliding regime can occur only at low values of and cannot be
clearly seen in the data in Fig. 6. Similar considerations can
be obtained from the experimental values of (R y
v x )/ v ox e x in Ref. 9.
All these results can be understood by assuming that the
transition from the ordinary sliding regime to a nonsliding
regime occurs when the friction force reaches a value given
by the coefficient of static friction. In ordinary systems, o
and the loss of sliding motion leads to a nonsliding
regime. The behavior of superballs can be interpreted by
assuming that o . In this case, grip-slip behavior is observed. The transition from this grip-slip regime to the ordinary sliding regime occurs when the friction force equals a
critical value that depends on e x . In these circumstances, the
nonsliding regime is absent. These considerations are consistent with the value of the coefficients of static and dynamic
friction estimated here and those reported by Cross8,9 for the
rebound of superballs with hard surfaces. The static coefficient in Ref. 8 was 0.520.04, while the dynamic coefficient
estimated in Ref. 9 was about unity for comparable surfaces.
Under our experimental conditions, the dynamic coefficient
of friction for the superball on different surfaces was estimated to be between 0.850.90, also larger than the reported
static values.8
D. Some implications for sports

Fig. 6. Plots of R oy / v o versus the incident angle ( 90 ). Comparison of the data for the rebound of a golf ball points and a superball
circles taken from Ref. 6, and theoretical predictions using the nonsliding
condition a, the e-e model Eq. 21 with e x 0.50 b and the Garvin
model c. Dotted lines correspond to the -e sliding regime Eq. 11 with
e0.90, 0.18 d, and e0.90 and 0.22 e and e0.96, 1.0 f.
Line g corresponds to the variation of R oy /u o with the incident angle
of the golf ball, assuming that a transition from the -e sliding regime to
the nonsliding regime occurs.
34

Am. J. Phys., Vol. 73, No. 1, January 2005

From our results, some brief observations concerning the


mechanics of ball sports can be made. Horizontal launch
experiments can be used to determine the mechanical characteristics of a tennis court, thus obtaining the coefficients of
restitution and friction and eventually the conditions for slipping, no slipping, and grip-slip regimes of rebound.
For tennis it is useful to obtain large rebound angles: the
rebound angle depends on the materials for example, clay
and grass and the conditions of the court, given by the values of e and , but also on the angular velocity imparted to
the ball the lift. Figure 7 shows R oy / v ox versus for e
0.80, close to the values reported by Cross9 for the rebound
of tennis balls on hard surfaces. The external sliding regions
Antonio Domenech

34

increases. As a general rule, obtaining a large deviation angle


requires a low angle of impact. Similar considerations apply
for handball, a sport in which lateral deviation effects are
important.
V. DISCUSSION

Fig. 7. The different rebound regimes for R ox 0. Continuous lines correspond to the transition from the -e sliding regime to the nonsliding one for
a e0.80, 0.40, and b e0.80, 0.18. Dotted lines correspond to
the transition from the e-e grip regime with e x 0.50, to the -e sliding
regime with the values of the coefficients of friction and restitution in a.

for which a -e sliding regime must be attained are a e


0.80, 0.40, and b e0.80, 0.10, and are limited
by transitions to the corresponding nonsliding regime continuous lines and to a grip-slip regime with e x 0.50 dotted
lines. Figure 7 exhibits symmetry with respect to
R oy / v ox 1. Obviously, a backspin must be imparted to the
ball to achieve large rebound angles, a condition frequently
desired by tennis players.
Similar considerations can be applied to golf. Initially it is
desirable to hit the ball to a large distance. In this case it is
convenient to impart a forward spin to the ball because the
friction force arising in the rebound increases the linear velocity of the center of mass of the ball after the rebound.
However, for hits in which the ball must be located within
the green, it is convenient to impart a backspin; in this
case the velocity of the center of mass of the ball is decreased; that is, the ball covers a small distance after the
rebound.
In basketball and handball, the players move while bouncing the ball. In these cases it is desirable that the velocity of
the center of mass of the ball remains equal before and after
the bounce, or, equivalently, . This condition can be
obtained by imparting a backspin to the ball with the fingers
as described in Ref. 4. The magnitude of the backspin depends on the restitution and friction coefficients. In these
cases, however, the impact angle is usually low, and we expect that the ordinary nonsliding regime applies, that is, the
rebounds must fall into the nonsliding region of the diagram
in Fig. 7.
In tennis it often is desirable to obtain rebounds in which
the ball is displaced horizontally. In these cases, a spin about
the x axis should be imparted to the ball. If the nonsliding
regime holds, we obtain from Eq. 23 with e x 0 that the
deviation angle is only a function of R ox / v o , and is
independent of the incident angle. The same result is obtained for grip-slip behavior. In contrast, if the -e sliding
regime applies, Eq. 12 predicts that for a given value of
R ox / v o , the deviation angle decreases as the incident angle
35

Am. J. Phys., Vol. 73, No. 1, January 2005

The bounce of a ball on a horizontal surface can be described using two limiting models: the -e model, in which
the friction forces are described in terms of the static and
dynamic coefficients of friction, and the e-e model, in which
the friction forces are expressed in terms of horizontal coefficients of restitution. When the rebound takes place without
sliding at the point of contact, both models lead to identical
predictions.
The classic horizontal launch experiment provides an empirical way of distinguishing between the sliding, nonsliding,
and grip regimes. The data indicate that the rebound of ordinary balls can be described in terms of the transition from a
sliding regime, satisfactorily described by the -e model, to
a nonsliding regime. This situation occurs in the usual case
in which o . The rebound of superballs, however, corresponds to a grip behavior that can be satisfactorily described by the e-e model. This difference appears to be associated with the condition o . The data suggest that
there is a transition from the grip behavior to the ordinary
sliding one while the nonsliding regime is entirely absent.
The horizontal launch experiment can be used to determine the characteristics of courts in several sports. From
pedagogical purposes, the horizontal launch experiment can
be used to illustrate the laws of impact taking into account
inelasticity and friction. Additionally, the experiment discussed in Sec. III is illustrative of the design of suitable
experiments for comparing theoretical models.
Note, that the scope of the models we have discussed is
limited by the assumption of the constancy of the coefficients
of restitution and friction. The fit between theory and experiment suggests that this assumption is reasonable for the conditions of moderate velocity and small surface deformation
imposed here, in agreement with published data.79,13,15,17
The methodology can be used to study the variation of and
e with parameters such as the velocity and the diameter. In
spite of these limitations, the models discussed in this paper
can be considered as plausible and complementary descriptions of rebounds in three dimensions.
a

Electronic mail: antonio.domenech@uv.es


A. Metha and J. M. Luck, Novel temporal behavior of a nonlinear dynamic system: The completely inelastic bouncing ball, Phys. Rev. Lett.
65, 393396 1990.
2
A. Metha and G. C. Barker, The dynamics of sand, Rep. Prog. Phys. 57,
383 416 1994.
3
A. Lorenz, C. Tuozzolo, and M. Y. Louge, Measurements of impact
properties of small, nearly spherical particles, Exp. Mech. 37, 292298
1997.
4
P. J. Brancazio, Physics of basketball, Am. J. Phys. 49, 356 365 1981.
5
H. Brody, Thats how the ball bounces, Phys. Teach. 22, 494 497
1984.
6
R. Garwin, Kinematics of ultraelastic rough ball, Am. J. Phys. 37,
88 92 1969.
7
R. Cross, The bounce of a ball, Am. J. Phys. 67, 222227 1999.
8
R. Cross, Measurements of the horizontal coefficient of restitution for a
superball and a tennis ball, Am. J. Phys. 70, 482 489 2002.
9
R. Cross, Grip-slip behavior of a bouncing ball, Am. J. Phys. 70, 1093
1102 2002.
10
N. Maw, J. R. Barber, and J. N. Fawcett, The role of elastic tangential
compliance in oblique impact, J. Lubr. Technol. 103, 74 80 1981.
1

Antonio Domenech

35

11

E. T. Whittaker, A Treatise on the Analytical Dynamics of Particles and


Rigid Bodies Cambridge University Press, Cambridge, 1904, p. 232.
12
J. B. Keller, Impact with friction, J. Appl. Mech. 53, 13 1986.
13
R. M. Brach, Friction, restitution, and energy loss in planar collisions, J.
Appl. Mech. 51, 164 170 1984.
14
T. R. Kane and D. A. Levinson, An explicit solution of the general
two-body collision problem, Comput. Mech. 2, 75 87 1987.
15
G. Barnes, Study of collisions. Part I: A survey of the periodical literature, Am. J. Phys. 26, 5 8 1958; ibid. Study of collisions. Part II:
Survey of the textbooks, Am. J. Phys. 26, 912 1958.
16
G. Amontons, De la resistance causee dans les machines, Memoires de
lAcademie Royale A, 275282 1699; briefly described in J. Krim, Resource Letter: FMMLS-1: Friction at macroscopic and microscopic length
scales, Am. J. Phys. 70, 890 896 2002.
17
H. L. Armstrong, How dry friction really behaves, Am. J. Phys. 53,
910911 1985.
18
A. Domenech and M. T. Domenech, Analysis of two-disc collisions,
Eur. J. Phys. 14, 177183 1993.
19
A. Domenech and M. T. Domenech, Oblique impact of rolling spheres: A

generalization of billiard-ball collisions, Rev. Mex. Fis. 44, 611 618


1998.
20
Strictly speaking, friction associated with the rotation of the sphere around
a vertical axis, resulting in a decrease of z , appears. This effect has been
neglected here Ref. 10.
21
R. A. Bachman, Sphere rolling down a grooved track, Am. J. Phys. 53,
765767 1985.
22
For the case of pure rolling motion along a horizontal track, the term
R oy /u o equals the ratio of the effective radius of gyration and the radius
of the ball. See, for instance, D. E. Shaw and F. Wunderlich, Study of the
slipping of a rolling sphere, Am. J. Phys. 52, 9971000 1984; R. L.
Chaplin and M. G. Miller, Coefficient of friction for a sphere, Am. J.
Phys. 52, 1108 1111 1984.
23
For the case of the head-on impact of spheres along a horizontal track,
R oy /u o must vary from (5/2) S , where S is the coefficient of friction between the spheres, to R/Re, when the ball reaches the pure rolling
motion along the track. See A. Domenech and E. Casasus, Frontal impact
of rolling spheres, Phys. Educ. 26, 186 189 1991.

Steam Turbine. Models of reciprocating steam engines from the first half of the 20th century are fairly common, but this turbine-type engine in the
collection at Hobart and William Smith Colleges is unique. A 400 W electric heater acting on the water in the boiler produces the jet of steam from the nozzle
at the right. The turbine and its blades are covered with wire mesh to protect inquiring fingers. A pulley, hidden behind the turbine disk, is used to take off
the rotary motion. The apparatus is listed at $8.50 in the 1940 Central Scientific Company catalogue, and uses the same boiler and safety-valve mechanism
as the accompanying reciprocating steam engine model. Photograph and notes by Thomas B. Greenslade, Jr., Kenyon College

36

Am. J. Phys., Vol. 73, No. 1, January 2005

Antonio Domenech

36

You might also like