You are on page 1of 27

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/278104773

Oblique Bounce of a Rubber Ball

Article in Experimental Mechanics · November 2014


DOI: 10.1007/s11340-014-9938-3

CITATIONS READS

15 744

1 author:

Rod Cross
The University of Sydney
299 PUBLICATIONS 2,878 CITATIONS

SEE PROFILE

All content following this page was uploaded by Rod Cross on 26 June 2017.

The user has requested enhancement of the downloaded file.


Oblique bounce of a rubber ball

Rod Cross
Physics Department, University of Sydney, Sydney, NSW Australia
Email: cross@physics.usyd.edu.au

Keywords: tangential COR, grip-slip, friction, vibration, impact

1
Oblique bounce of a rubber ball

Abstract

Measurements are presented on the oblique impact of a hollow rubber ball incident on a
polished granite surface, and the results are compared with those for a solid rubber superball.
The hollow ball had a much higher coefficient of sliding friction than the superball, resulting
in significant differences in all bounce parameters, at all angles of incidence. The hollow
ball gripped the surface at all observed angles of incidence, resulting in one or two reversals
in the direction of the friction force during the impact. The friction force was measured
directly, as was the rotation speed of the ball during the impact. The results show that
the tangential coefficient of restitution of a ball depends on both the coefficient of sliding
friction and the ratio of the tangential to the normal vibration frequency of the ball.

Introduction

The oblique impact of a ball with another ball or with a flat surface has been studied by
many authors and is relatively well understood [1-24]. Nevertheless, there are some features
of such an impact that are less well understood and that warrant further investigation. Of
particular interest is the behaviour of the ball during the impact itself. Those details are
generally difficult to quantify experimentally, since the impact is usually very short, the
friction force acting on the ball is difficult to measure, and many balls do not remain even
closely spherical during the impact. Consequently, most impact experiments are concerned
primarily with measurements of the outgoing speed, angle and spin of the ball. If the results
are then compared with theoretical predictions, information can be obtained on the elastic
and frictional properties of a ball under conditions where the applied force increases very
rapidly to a large value. Information obtained in this manner for one type of ball does
not necessarily allow for accurate predictions concerning other ball types. For example, the
bounce of a tennis ball may be modelled quite accurately, but the same model may not
provide a good description of the behaviour of a golf or a basketball, even when the changes
in mass, stiffness and ball diameter are taken into account.

In theory, the three outgoing bounce parameters of a particular type of ball (ie speed,
angle and spin) can be predicted analytically in terms of just two impact parameters, these
being the normal coefficient of restitution, ey , and the tangential coefficient of restitution, ex

2
[1-3]. Under conditions where the ball slides throughout the bounce, a more generally useful
second parameter is the coefficient of sliding friction, µ, rather than the tangential coefficient
of restitution. The normal and tangential coefficients of restitution provide measures of the
total energy loss in the ball and the impact surface resulting from deformation of the ball and
the surface in directions respectively normal and tangential to the surface. Both restitution
coefficients are difficult to predict from first principles but are easily measured.

In practice, a third impact parameter may be needed to account for the fact that many
balls are flexible rather than being perfectly rigid. If the ball radius, R, decreases during
the impact then the torque F R arising from the friction force, F , also decreases. If the
normal reaction force, N , is distributed over a finite area of the ball rather than being
concentrated at a single point, and if N acts at a distance D ahead of the centre of mass,
then an opposing torque N D will be generated. Given that R, D and the rotational inertia
of the ball can all vary with time during the bounce, it is difficult to find a simple analytical
model that can usefully be used to accurately describe the bounce of a ball. For that
reason, finite element methods are often used to model the behaviour of a bouncing ball
[4-12]. Alternatively, a simplifying assumption can be made that the combined non-ideal
effects can all be represented empirically by a constant average value of D, combined with
an artificially constant value of R [3]. Any given bounce can then be described in terms
of the three parameters ey , ex and D, none of which necessarily remain constant as the
initial conditions are varied. It is known that ey generally decreases as the incident ball
speed increases, and that it varies with temperature, but it is still widely used to describe
and specify different ball types since the variation with ball speed and temperature usually
remains relatively small.

Measurements of ex are far less common than measurements of ey , for any ball type.
Measurements for small, solid spheres and sports balls can be found in references [1-3, 13-
16]. Similarly, measurements of the friction force acting on a bouncing ball are also rare
but can be found in references [3, 6, 17]. Measurements of D are reported in [3, 18]. In the
present paper, results are extended to consider the bounce of a hollow rubber ball, partly
because the bounce of such a ball has not previously be studied in detail, partly because
hollow balls are relatively common in ball sports, and partly because rubber balls deform
significantly even at low ball speeds. The primary concern was to study the physics of

3
the bounce process under conditions where non-ideal effects might be more readily observed
than with more rigid ball types. All bounce results were obtained when the ball was incident
with relatively small spin (nominally zero) on a horizontal, polished block of granite. For
comparison, results are also presented for a superball.

Oblique bounce model

ω1 ω2
y

x
v1
v2
vy1 vy2
θ1 θ2
vx1 vx2

N
ω
D

FIG. 1: A ball incident obliquely from the left at speed v1 , angle θ1 and with spin ω1 bounces at
speed v2 and angle θ2 with spin ω2 . The positive directions of x and y are as shown, while the spin
is taken to be positive in the clockwise direction. The lower diagram shows the friction force, F ,
the normal reaction force, N and the offset distance D.

The geometry of the oblique impact studied in the present experiment is shown in Fig. 1.
The ball is incident at velocity v1 , at an angle θ1 to the horizontal, and with angular ve-
locity ω1 . It rebounds at velocity v2 , at an angle θ2 and with angular velocity ω2 . The
relevant parameters describing measurements of the bounce of the ball are defined in Fig. 1.
The velocity components vx1 , vy1 , vx2 and vy2 were measured directly from video film with
the help of motion analysis software, as were the angular velocities ω1 and ω2 . The angles
θ1 and θ2 and the speeds v1 and v2 are easily calculated from the measured velocity com-
ponents. From this data, and the known ball radius, R, a number of additional bounce
parameters were calculated, including the normal coefficient of restitution ey = vy2 /vy1 , the
tangential velocity ratio vx2 /vx1 , the ratio of the horizontal impulse to the vertical impulse
COF= (vx1 − vx2 )/(vy2 + vy1 ) and a spin parameter S = Rω2 /vx2 . Even though the ball
was incident in the negative y direction, all velocity components are taken to be positive

4
quantities, including vy1 .

At small values of θ1 (quantified below in Eq. (7)) the ball can slide throughout the
impact, in which case COF is equal to the coefficient of sliding friction and S is less than
unity [3]. Analytical solutions of the equations of motion in the x and y directions are then
easily obtained from the relation F = µN where µ is the coefficient of sliding friction. At
large values of θ1 the ball slides for a short distance and then grips the surface when the
tangential speed of the contact region of the ball drops to zero. As shown by Maw et al [13,
19], some parts of the contact region can grip while other parts continue to slide, depending
on the local friction force and the local elastic distortion of the ball. If the ball grips then
the ball bounces off the surface with a value of S approximately equal to or greater than
unity.

If the ball grips the surface during the bounce then the friction force decreases below that
expected from sliding motion. The friction force is then due to static friction and may even
reverse direction during the bounce [3, 13, 17]. As a result, the horizontal impulse is less
than that expected for a sliding ball, and the resulting value of the COF is less than the
coefficient of sliding friction. For that reason, reliable bounce measurements of the coefficient
of sliding friction can be obtained only when θ1 is relatively small [18].

Since there is no simple relation between F and N when a ball is subject to static
friction, there is no analytical solution of the equations of motion. Two simple solutions
have previously been devised to get around this problem. One is to assume that the ball
starts rolling when vx = Rω, in which case F drops to zero and then remains zero until the
ball bounces. This assumption is based on a perfectly rigid ball approximation. The ball
then bounces with vx2 = Rω2 or with S = 1. In those cases where the ball does bounce with
S = 1, the ball is said to bounce in a rolling mode. The other solution is to assume that the
ball bounces with a positive value of ex , where ex is defined by the expression

(vx2 − Rω2 )
ex = − (1)
(vx1 − Rω1 )

The condition ex = 1 gives the result that (vx2 − Rω2 ) = −(vx1 − Rω1 ). The tangential
velocity of the ball at the contact point then reverses direction as a result of the bounce,
without a change in magnitude, and the ball bounces with higher spin than it would if it
bounced in a rolling mode (ie with ex = 0). The ex = 1 approximation can be used to

5
describe the bounce of a superball [20], but most balls have a maximum value of ex less than
0.5.

If a ball slides throughout the bounce then it bounces with Rω2 < vx2 and hence with
S < 1. If also Rω1 < vx1 , then the ball bounces with ex < 0. If the ball grips the surface
during the bounce, then it can bounce with S > 1 and with ex > 0. In the latter case,
the ball is said to overspin. Maximum spin would arise if ex = 1, but such a condition is
not achieved in practice. For an overspinning tennis ball bouncing on a rigid surface, ex
is typically about 0.2 and for an overspinning superball ex is typically about 0.5 [3,16]. A
surprising result with the hollow rubber ball was that it gripped the surface but it bounced
at some angles of incidence with S ≈ 0.8 and ex ≈ −0.1 rather than with S > 1 and ex > 0.

The three equations of motion describing the bounce are F = −mdvx /dt, N = mdvy /dt
and F R − N D = Icm dω/dt where m is the mass of the ball and Icm = αmR2 is the moment
of inertia of the ball about an axis through its centre of mass. For a solid sphere, α = 0.4.
For the thick-walled hollow ball used in the present experiment, α = 0.576. The latter
result was obtained by subtracting the moment of inertia of a small sphere from that of a
larger sphere, the radii of the two spheres differing by the wall thickness of the hollow ball.
Integration of the first two equations over the impact duration yields the results that

∫ F dt = m(vx1 − vx2 ) (2)

∫ N dt = m(vy1 + vy2 ) (3)

If we now assume that R, D and Icm all remain approximately constant during the bounce,
then
R ∫ F dt − D ∫ N dt = Icm (ω2 − ω1 ) (4)

Combining Eqs (2-4) gives

(vx1 − vx2 ) D(1 + ey )vy1


Rω2 = Rω1 + − (5)
α αR

Equation (5) provides a simple method of calculating D from measured bounce parameters,
and shows that the effect of finite, positive D is to reduce the rebound spin below the ideal
value where D = 0 and where R and Icm remain constant. The reduction is typically between
10% and 25% in the experiments described below. In terms of the incident ball parameters,

6
Rω2 is given by

(1 + ex ) D(1 + ey ) (α − ex )
Rω2 = vx1 − vy1 + Rω1 (6)
(1 + α) R(1 + α) (1 + α)

In the pure sliding case, solutions of Eqs. (2)-(4) can be found using the relation F = µN .
It can be shown from these solutions that the ball will slide throughout the impact, when
ω1 = 0 and D = 0, provided that

1
tan θ1 < (7)
(1 + ey )(1 + 1/α)µ

In the pure sliding regime, it is easily shown from Eqs. (2) and (3) that

vx2
= 1 − µ(1 + ey ) tan θ1 (8)
vx1

and hence vx2 /vx1 decreases as θ1 increases until the ball commences to grip the surface.
During the grip phase, Eqs. (1), (2) and (4) indicate that

(1 − αex ) D(1 + ey ) α(1 + ex )


vx2 = vx1 − vy1 + Rω1 (9)
(1 + α) R(1 + α) (1 + α)

If D and ω1 remain small and ex remains approximately constant then vx2 /vx1 will also
remain approximately constant as θ1 increases.

Energy loss during the impact can be calculated in terms of ey and ex . For a normal
impact with ω1 = 0, the fractional energy loss is 1 − e2y . For an oblique impact with ω1 = 0
and D = 0, it is easy to show from the above relations that the loss in kinetic energy, ∆E,
is given by
α(1 − e2x ) 1 2
 
1 2
∆E = mvx1 + (1 − e2y ) mvy1 , (10)
(1 + α) 2 2
indicating that the energy loss can be separated into vertical and tangential components,
and that ∆E = 0 if ex = 1 and ey = 1. The momentum equations impose no limit on ex ,
but conservation of energy requires that −1 < ex < 1.

Experimental procedure

The results presented in this paper were obtained with two different rubber balls. One
was a hollow Spalding High Bounce rubber ball of mass 44.2 g, diameter 59.4 mm and wall
thickness 4.5 mm. The ball is commonly available in toy shops as a child’s toy, and is similar
to a racquetball but slightly larger in diameter. It is also very similar in mass, diameter

7
and bounce properties to the inner rubber core of a tennis ball. The second ball was a solid
superball of mass 46.2 g and diameter 46.0 mm. Some results for this ball were presented
previously [3], but a more extensive set of results is presented here to compare with the
results for the hollow ball.

Most of the results were obtained by projecting each ball by hand onto a polished granite
slab to impact at a speed between 4 and 7 m/s and over a wide range of angles of incidence
from 12◦ to 90◦ . Some additional results were obtained at speeds up to 15 m/s by projecting
the hollow ball with a Tennis Tutor ball launcher. All results were obtained for balls incident
with nominally zero spin, but in practice a small amount of spin was usually imparted
unintentionally. Since the incident spin affects the outgoing spin, both spins were measured
for every bounce in order to obtain reliable estimates of D.

All impacts were filmed with a Casio EX-F1 video camera. The camera was zoomed in
to enlarge the ball in order to measure its coordinates as accurately as possible. The ball
was illuminated with three 500 W halogen lamps, and was filmed at either 300, 600 or 1200
frames/s depending on the angle of incidence and the desired image size (since fewer pixels
were recorded at the higher frame rates). The shutter speed was set to 1/4000 s to reduce
blurring of the image of the ball. Equator lines drawn on the ball were used to determine
the angular displacement of the ball between frames in order to measure its rotation speed.
Each of the velocity components and the angular velocities ω1 and ω2 were measured with
a maximum estimated error of ±2 %, based on uncertainties in the measured distances and
angles recorded on video film.

Experimental results

Experimental results for the hollow ball are shown in Fig. 2, and corresponding results for
the superball are shown for comparison in Fig. 3. All results in Fig 2 were obtained with
ω1 less than 0.1 ω2 in magnitude, and with v1 between 4.3 m/s and 7.2 m/s. All results in
Fig. 3 were obtained with ω1 less than 0.15 ω2 in magnitude and with v1 between 4.0 m/s
and 6.5 m/s. Despite the variation in ω1 and v1 from one bounce to the next, most of
the results in Figs. 2 and 3 are not sensitive to these variations or are plotted in normalised
form to minimise the effect of the variations. Consequently, most of the scatter in the results
represents normal variations that one finds when measuring bounce parameters even when
ω1 and v1 are held constant.

8
ey
vx ratio

(a) (b)
1.1 1.2

1.0 v y2
ey = 1.0
v y1 θ1
0.9
v x2 / vx1

0.8
0.8 (vx1 - vx2)
0.6 COF =

COF
(vy1 + vy2)
0.7
and

0.4
0.6
ey

v x2 0.215
0.5 0.2
v x1 tan θ1

0.4 0.0
0 20 40 60 80 100 0 20 40 60 80 100
θ1 (degrees) θ1 (degrees)

(c) (d)
1.5 0.8
Rω2
S= 0.7
v x2
v = 4.3 to 7.2 m/s
1.0 0.6 1
- ω 1 ) / v1

0.5
S

0.5 0.4
R ( ω2
and

R ω 2 - v x2
ex = 0.3
v x1 - R ω 1
ex

0.0 0.2

0.1

-0.5 0.0
0 20 40 60 80 100 0 20 40 60 80 100
θ1 (degrees) θ1 (degrees)

(e) (f)
6 1.0

5 N
0.8
Kinetic energy loss fraction

0.6
3
D (mm)

2
0.4

1
0.2
0

-1 0.0
0 20 40 60 80 100 0 20 40 60 80 100
θ1 (degrees) θ1 (degrees)

FIG. 2: Experimental results for the hollow rubber ball incident at nominally zero spin on a
polished granite slab at speeds between 4.3 and 7.2 m/s. Each dot represents a separate bounce,
and solid curves are polynomial fits to the data (apart from the fit in (b)) to guide the eye.

9
(a) (b)
1.1 1.0
v y2
1.0 ey = 0.263
v y1
0.8
tan θ 1
0.9
v x2 / vx1

(vx1 - vx2)
COF =
0.8 0.6 (vy1 + vy2)

θ1

COF
0.7
and

0.4
ey

0.6
v x2 0.2
0.5 v x1

0.4 0.0
0 20 40 60 80 100 0 20 40 60 80 100
θ1 (degrees) θ1 (degrees)
(c) (d)
2.5 1.0

2.0
0.8
- ω 1 ) / v1

1.5 Rω2
S=
v x2 0.6
S

1.0
R ( ω2
and

0.4
ex

0.5

R ω 2 - v x2 0.2 v = 4.0 to 6.5 m/s


0.0 ex = 1
v x1 - R ω 1

-0.5 0.0
0 20 40 60 80 100 0 20 40 60 80 100
θ1 (degrees) θ1 (degrees)

(e) (f)
2.5 0.5

N
2.0 0.4
Kinetic energy loss fraction

1.5 D 0.3
D (mm)

1.0 0.2

0.5 0.1

0.0 0
0 20 40 60 80 100 0 20 40 60 80 100
θ1 (degrees) θ1 (degrees)

FIG. 3: Experimental results for the superball incident at nominally zero spin on a polished granite
slab at speeds between 4.0 and 6.5 m/s. Each dot represents a separate bounce, and solid curves
are polynomial fits to the data (apart from the fit in (b)).

10
(a) ey and vx2 /vx1

Eight separate graphs are shown for each ball, each as a function of the angle of incidence,
θ1 , to illustrate different aspects of the bounce. As shown in Figs. 2(a) and 3(a), both balls
bounced with a relatively high value of ey , especially at low angles of incidence. The latter
effect can be attributed in part to the fact that ey decreases as vy1 increases, an effect that
is observed with almost all balls. In the present experiment, v1 was held approximately
constant for each ball (within the limits described above), with the result that vy1 increased
as θ1 increased.

Also shown in Figs. 2(a) and 3(a) is the ratio vx2 /vx1 for each ball type. For the hollow
ball, this ratio lies between 0.5 and 0.65 for most angles of incidence, but is more variable at
high angles of incidence. At large θ1 , ω1 was close to zero, and the variabilty was caused by
relatively small changes in D from once bounce to the next. A different result was obtained
for the superball, the vx2 /vx1 ratio decreasing from 0.8 to 0.5 as θ1 increased to 40◦ , and then
remaining approximately constant when θ1 was greater than 40◦ . The different behaviour of
the two balls is due primarily to differences in the coefficient of sliding friction. The hollow
ball had a much higher value of µ, causing the ball to grip the surface at all observed angles
of incidence. The superball gripped the surface when θ1 > 20◦ but it slid throughout the
bounce when θ1 < 20◦ . Additional evidence for this behaviour is shown in Figs. 2(b, c) and
3(b, c).

(b) COF

The variation of COF with θ1 is shown for each ball in Figs. 2(b) and 3(b). COF decreases
to zero at θ1 = 90◦ since then vx1 = vx2 = 0. For the hollow ball, COF exceeds 1.0 at low
θ1 , indicating that µ is greater than 1.0. It was not possible to obtain reliable results at
θ1 < 10◦ , so a reliable estimate of µ was not obtained from the bounce data. A surprisingly
good fit to the COF for the hollow ball can be obtained by assuming that vx2 /vx1 = 0.6 and
vy2 /vy1 = 0.85 at all angles of incidence, in which case COF = 0.215/tan θ1 . The fit is shown
in Fig. 2(b). The results in Fig. 3(b), when combined with those in Fig 3(c), indicate that
µ = 0.45 ± 0.08 for the superball on the smooth granite slab. At high angles of incidence
where vx2 /vx1 ≈ 0.5 and vy2 /vy1 ≈ 0.9, COF ≈ 0.263/ tan θ1 , as shown in Fig. 3(b).

11
(c) S and ex

The variation of S and ex with θ1 is shown in Figs. 2(c) and 3(c). The usual result for most
ball types is that S increases to unity and ex increases to zero as θ1 increases, corresponding
to a transition from sliding at low θ1 to a grip phase at high θ1 . This was the result obtained
for the superball, but a different result was obtained for the hollow ball. In effect, the
transition appeared to occur in reverse order. However, it is clear that the usual transition
would have occurred at lower values of θ1 than those studied in the present experiment,
given the high value of µ. For example, if we assume that µ = 1.5, ey = 0.9 and α = 0.576
for the hollow ball, then the ball would slide throughout the impact only if θ1 < 7.3◦
according to Eq. (7). Evidence presented below shows that the decrease in S below 1.0, and
the corresponding decrease in ex below zero, can be attributed to an early reversal in the
direction of the friction force for this ball.

(d) R(ω2 − ω1 )/v1

Figures 2(d) and 3(d) show the change in the circumferential ball speed, R(ω2 − ω1 ), nor-
malised to the incident speed v1 , vs θ1 . The results show that at any given value of v1 , the
change in ball spin is maximised at θ1 ≈ 35◦ for the superball and at θ1 < 12◦ for the hollow
ball. In general, the superball was found to spin faster than the hollow ball, as expected,
but the hollow ball was found to spin faster than the superball at θ1 < 15◦ .

(e) D offset

The variation of D with θ1 is shown in Figs. 2(e) and 3(e). There was a clear trend for D
to increase at low θ1 or to decrease at high θ1 . Smaller values of D were observed for the
superball. Despite the large scatter in D, the outgoing spin of the ball was not strongly
affected. The reason can be inferred from Eq. (5). The outgoing spin is determined mainly
by the second term on the right side of Eq. (5) and to a smaller extent by the third term.
Given that ey ≈ 0.9 and vx2 /vx1 ≈ 0.55 for both balls, the ratio of the third to the second
term is approximately 4.2D tan θ1 /R. The largest scatter in D occurs when θ1 ≈ 20◦ where
the ratio is then 1.5D/R. Since D varies from about 1 mm to about 5 mm for the hollow ball,
the ratio of the third to the second term varies from about 0.05 to about 0.25. Even when D
is as large as 5 mm, there is only about a 25% reduction in the outgoing spin. Consequently,
the scatter in the results in Figs. 2(d) and 3(d) remains relatively small compared with the

12
corresponding scatter in D.

(f) Kinetic energy loss fraction

The fractional loss of kinetic energy during the impact is shown in Figs 2(f) and 3(f).
The incident kinetic energy is KE1 = 0.5mv12 + 0.5Icm ω12 . The rebound kinetic energy is
KE2 = 0.5mv22 + 0.5Icm ω22 . The fractional energy loss is defined as (KE1 − KE2 )/KE1 .
The fractional energy loss remained relatively constant for the hollow ball, at all angles of
incidence, and increased slightly as θ1 decreased for the superball. The fractional energy
loss was smaller for the superball than for the hollow ball, due to the generally higher values
of ey and ex .

Angular velocity during the impact


In order to measure the rotation speed of the hollow ball during the impact itself, impacts
were recorded on video film at 1200 fps and the incident ball speed was increased so that
a significant rotation of the ball could be observed from one video frame to the next. A
sequence of images at equal time intervals is shown in Fig. 4 for one particular bounce at
v1 = 15.4 m/s. The images highlight several features of the bounce process that have not
previously been observed on video film, as far as the author is aware. They show clearly
that the ball gripped the surface throughout most of the impact period, and they also show
that the whole ball vibrates tangentially when it grips the surface. The vertical dashed lines
in Fig. 4 pass through a fixed point in the granite slab. The bottom of one equator line
remains anchored to this point as the ball rotates, giving the appearance that the ball is
rolling with vx = Rω. However, additional measurements presented below show that the
friction force on the ball reverses direction when the ball grips, and vx is not equal to Rω.

The ball compresses vertically during the impact and it also stretches horizontally when
it grips, thereby storing elastic energy as a result of both its horizontal and its vertical
deformation. The result can be seen in Fig. 4, frames 1 and 2, where the front edge of
the ball can be seen to lean forward in a manner similar to that calculated in Ref. 9. The
ball then vibrates in a tangential direction, leading to the result shown in Fig. 4, frame 4,
where the ball leans backwards. The upper part of the ball therefore exerts an oscillating
tangential force on the contact surface, resulting in one or more reversals in the direction of
the static friction force, as described in Sec. 6.

13
0 4

0
+17.2

1 5

+4.3
0 +14.70

2 6

+9.20 +14.00

3 7

+14.20 +13.20

FIG. 4: A sequence of images from video film recorded at 1200 fps showing an impact of the hollow
ball incident at v1 = 15.4 m/s and θ1 = 23◦ on the granite slab. The vertical dashed lines pass
through a fixed point on the granite slab, indicating that the bottom of the ball gripped the surface
during most of the impact period. The indicated angles denote the increase in rotation angle from
one video frame to the next. The bottom of the ball is reflected in the polished granite slab.

The change in rotation angle from one frame to the next is shown in Fig. 4. These
results, plus those recorded before and after the bounce, were used to calculate ω = ∆θ/∆t,
as shown in Fig. 5. The angular velocity increased to a maximum during the bounce and
then decreased before the ball bounced off the surface, with the result that the ball bounced
with S < 1. The latter result implies that there was a significant reversal of the friction
force during the bounce. Direct measurements of F and N are presented in the following
Section.

14
400

350

300

(rad/s)
250

200 Post-bounce

Δθ / Δ t
150
Pre-bounce
100

50

0
-5 0 5 10
t (ms)

FIG. 5: Measured angular velocity of the ball shown in Fig. 4. The ball bounced with vx2 =
8.75 m/s, ω2 = 260 rad/s, and with S = Rω2 /vx2 = 0.89.

Measurements of N and F for the hollow ball

(a) (b)
50 50

40 0
θ1 = 21 v 1 = 2.7 m/s 40 θ1 = 44.50 v 1 = 2.7 m/s

30 30 N
and N (N)

and N (N)

20 F N 20
F
10 10
F (N)

F (N)

0 0

-10 -10

-20 -20
3.3 ms 2.7 ms 4.7ms
-30 -30
-1 0 1 2 3 4 5 6 7 -1 0 1 2 3 4 5 6 7
t (ms) t (ms)

FIG. 6: Measurements of N and F for the hollow rubber ball

Measurements of N and F for the hollow ball are shown in Fig. 6. They were obtained
at a low incident speed, using the same apparatus as described in Ref. [3]. The ball was
projected by hand without spin to impact on a 50 mm x 50 mm x 4 mm piezo plate attached
to the upper surface of a rectangular block of wood. The upper surface of the piezo plate
itself was mechanically protected with a smooth, 0.5 mm thick sheet of G10, an epoxy and
fibreglass insulating material commonly used in printed circuit boards. A smaller piezo plate
was attached to an end face of the wood block, to function as an accelerometer, with an

15
output voltage proportional to F . The wood block was mounted on rollers to minimise the
friction force on the bottom of the block. The piezo outputs were calibrated by filming each
bounce to measure the vertical and horizontal impulse acting on the ball.

As shown in Fig. 6, the normal reaction force on the ball varies with time in the expected
manner, with a peak force approximately proportional to vy1 . The impact duration was
about 6 ms, decreasing slightly as θ1 increased due to the increased vertical speed of the ball
when v1 was held constant. Of greater interest is the time variation in the friction force,
which reversed direction during the bounce at all observed angles of incidence. Initially, F
was greater than N , indicating that the coefficient of sliding friction was approximately 1.8
at low angles of incidence. The ball then gripped the surface, causing F to reverse direction
once at low angles of incidence, and twice at angles of incidence greater than about 21◦ . As
θ1 increased, both the first and second reversals in F occurred earlier in time, as indicated
in Fig. 6.

(a) (b)
3.0 2.5

2.5 Rω Rω
2.0

2.0
ω
and R ω

1.5
and R

1.5
1.0
vx
1.0
vx
vx

vx
0.5
0.5
0
θ1 = 21 v 1 = 2.7 m/s 0.0 θ1 = 44.50 v 1 = 2.7 m/s
0.0

-0.5 -0.5
-1 0 1 2 3 4 5 6 7 -1 0 1 2 3 4 5 6 7
t (ms) t (ms)

FIG. 7: Time variation of vx and Rω deduced from the measurements of F in Fig. 6. Dots and
open circles show experimental results for the two impacts.

If one assumes that D = 0 and that R remains constant during each bounce, then the
time integral of F can be used to estimate the variation of vx and ω during the impact.
Results for the impacts shown in Figs. 6(a) and 6(b) are shown in Figs. 7(a) and 7(b)
respectively. The initial and final values of vx , shown as dots, were used to calibrate the F
signal so the end points of vx in Fig. 7 are defined points. The initial and final values of Rω

16
are shown by open circles. The initial measured value was used to start the integration. The
final values differ slightly from the experimental values, indicating that D is not zero for
these impacts. Nevertheless, there is good qualitative agreement between the experimentally
observed behaviour of ω shown in Fig. 5 and the result shown in Fig. 7(a) at a similar angle
of incidence.

The result shown in Fig. 7(a) indicates that Rω2 < vx2 and hence S < 1 and ex < 0,
despite the fact that the ball gripped the impact surface during the bounce. The same result
is shown in Fig. 2(c) at all angles of incidence from about 15◦ to about 60◦ , in contrast to the
behaviour of the superball. The explanation can be seen by inspection of Fig. 7(a). During
the initial sliding phase, vx decreases and Rω increases, until vx = Rω. The ball then grips
the surface and Rω exceeds vx , as it does for the superball. However, the large reversal in F
then acts to reduce Rω and to increase vx , with the result that the ball bounces with S < 1.

Ball vibration

The results in Fig. 7 indicate that vx = Rω several times during each impact. The first
occurrence coincides with the time at which F is a maximum in Figs. 6(a) and 6(b), and
can be taken as the end of the initial sliding phase of the impact, given that F/N remains
approximately constant up to that point. The subsequent behaviour of the ball, during
the grip phase, depends on the normal and tangential elastic response of the ball while the
contact region on the surface remains at rest. That response was investigated experimentally
by measuring the vibration frequencies of the ball when the bottom of the hollow ball was
attached to a piezo plate with double-sided adhesive tape. The ball was impacted with a
small steel ball incident either vertically or horizontally on the rubber ball, and the results
are shown in Figs. 8(a) and (b).

The vibration period in the vertical direction was approximately 11 ms, consistent with
the fact that the impact duration of the ball was 6.0 ms at low impact speeds. The relatively
light damping of the oscillations in the vertical direction is consistent with the relatively high
normal coefficient of restitution for this ball. By contrast, a hollow rubber squash ball has
a low COR and exhibits heavy damping when excited in this manner [21]. The vibration
period in the horizontal direction was approximately 5 ms, consistent with the observed
oscillations in the friction force shown in Fig. 6. The ratio of the tangential to normal

17
vibration frequencies is therefore about 2.2 for this ball, higher than the theoretical estimate
quoted by Maw et al [13]. The latter authors considered the impact of an elastic sphere
with a half space of the same material, in which case tangential vibrations occur at the
contact surface. In the present case, the contact region remains at rest and the centre of
mass vibrates tangentially with respect to the contact surface. The results in Figs. 6 and 8
indicate an increase in tangential stiffness when a sphere impacts on a rigid surface.

(a) Vertical impact (b) Horizontal impact


60 30
Rubber ball

40 Rubber ball 20 Tape


Tape
Piezo
Piezo
Piezo signal (mV)

Piezo signal (mV)


20 10

0 0

-20 -10

11 ms 5 ms
-40 -20
-5 0 5 10 15 20 25 30 -5 0 5 10 15 20
t (ms) t (ms)

(c) Vertical impact (b) Horizontal impact


100 10
Superball Superball
3.5 ms

50 4.5 ms 5
Piezo signal (mV)

Piezo signal (mV)

0 0

-50 -5
25 ms

-100 -10
-5 0 5 10 15 20 25 30 -10 0 10 20 30 40 50 60 70
t (ms) t (ms)

FIG. 8: Response of the hollow rubber ball and the superball attached to a piezo plate, when
impacted by a small steel ball incident vertically or horizontally.

For comparison, the vibration response of the superball is shown in Figs. 8(c) and (d).
Since the superball was relatively stiff, the bottom of the ball was glued with epoxy to
a rigid plate and the plate was attached to the piezo plate with double-sided tape. The

18
vertical response is approximately as expected, apart from the fact that the impact duration
of the ball was about 3.4 ms at incident speeds less than 2 m/s, whereas the half period
of oscillation shown in Fig. 8(c) is 2.25 ms. During a vertical bounce, compression and
expansion of the bottom of the ball occurs on a slower time scale than the propagation
time of a compression wave across the whole ball. This result can be explained by the fact
that the contact region of a solid sphere is relatively soft compared with the stiffness of the
central part of the ball. A mass-spring model of the effect is described in Ref. [22]. The
horizontal impact shown in Fig. 8(d) excited a tangential vibration with period 3.5 ms, as
well as a much lower frequency tangential vibration with period 25 ms. The latter vibration
is analogous to the cantilever or “diving board” mode of a beam rigidly supported at one end
and free at the other end. The ratio of the higher frequency tangential to normal vibration
frequencies was 1.28 for the superball, consistent with the fact that the friction force was
observed to reverse relatively late in time for the superball [3].

Discussion

The experimental results presented in this paper represent a more detailed examination
of the impact of a ball than is usually found in studies of this type. They show that
the main features can be explained in terms of a simple analytical model, together with
additional experimental observations related to the bounce process. Nevertheless, there are
some features that remain unexplained in this study.

Normal coefficient of restitution, ey

An intriguing result is that the normal coefficient of restitution was found to approach unity
at low angles of incidence, for both ball types. A similar effect was observed previously
with a tennis ball incident at low angles on a tennis court, particularly on surfaces with
a high coefficient of sliding friction [18]. The explanation offered in that case may also
be relevant in the present case. When a ball impacts obliquely on a horizontal surface, it
stretches horizontally when it grips the surface, particularly at low angles of incidence where
the tangential velocity is relatively high. Elastic energy stored as a result of tangential
deformation may then be partly converted to vertical rebound energy as the ball rapidly
recovers its spherical shape, resulting in an additional vertical force on the ball. This process
differs from that described in Refs. 23 and 24, where it was found that ey can be greater

19
than 1.0 if an incident ball deforms the impact surface.

vx2 /vx1 ratio

Another intriguing result is that the vx2 /vx1 ratio remains relatively constant at high angles
of incidence, apart from some results in Fig. 2(a) when θ1 > 70◦ . The same result was
found in the previous tennis ball experiment [18]. Such a result is expected when a rigid
ball bounces in a rolling mode, since there is no change in vx2 if the friction force drops to
zero. However, it is not obvious why the vx2 /vx1 ratio should also remain constant for a
flexible ball, particularly when there is a large reversal in the friction force during the bounce.
Nevertheless, Eq. (9) indicates that vx2 /vx1 will remain relatively constant, provided that D
is small and that ex remains approximately constant. The latter conditions were generally
satisfied by both balls at high angles of incidence. For a rolling ball, ex = 0. For the hollow
ball, ex = 0 ± 0.2, while ex = 0.55 ± 0.05 for the superball at high angles of incidence. Since
ex was higher for the superball, its vx2 /vx1 ratio was lower than that of the hollow ball.

The apparently anomalous behaviour of vx2 /vx1 shown in Fig. 2(a) when θ1 > 70◦ can
be attributed to small variations in D. Small variations in ω1 had a negligible effect, as
estimated from Eq. (9). However, small variations in D have a relatively large effect on vx2
when vx1 is small and vy1 is large, as indicated by Eq. (9). In fact, the largest values of
vx2 /vx1 in Fig. 2(a) were all associated with small negative values of D around −0.5 mm,
and the smallest values of vx2 /vx1 were all associated with D values around +1.0 mm. It
is possible that small asymmetries in the mass distribution of the ball contributed to these
small variations in D.

Offset distance, D

The source of the large scatter in D at low angles of incidence was not established, nor was
it possible to distinguish the separate contributions to D arising from either a change in R
or an offset in N . At low angles of incidence, D was not well correlated with any observed
bounce parameter, including the COF. When considering all observed angles of incidence,
the strongest correlations were those with θ1 and with ω2 or vx1 , as shown in Fig. 9 which
includes some low angle of incidence data obtained at higher incident ball speeds than shown
in Fig. 2. The two plots in Fig. 9 are very similar since ω2 and vx1 were closely correlated
for the hollow ball. The leading edge of the ball would rotate faster into the impact surface

20
at high ω2 , giving a larger offset in the normal reaction force, but there is still considerable
scatter in the D values even when ω2 remains approximately constant. Similarly, D could
be enhanced if the ball leans further forward, as expected at high vx1 , due to an increase in
the offset in N . The change in R during an impact is largest at large θ1 , since vy1 and the
resulting ball compression increases with θ1 . Given that D decreased as θ1 increased, for
both balls, it is unlikely that the observed D values can be explained entirely by a change
in R alone.
(a) (b)
7 7

6 6

5 5

4 4
D (mm)
D (mm)

3 3

2 2

1 1

0 0

-1 -1
0 50 100 150 200 250 300 0 5 10 15
ω2 (rad/s) v x1 (m/s)

FIG. 9: Offset distance D vs (a) ω2 and (b) vx1 for the hollow rubber ball. The solid lines are
quadratic fits to the experimental data.

Tangential coefficient of restitution, ex

It is clear from the experimental results in Figs. 2 and 3 that the high value of ex for the
superball acts to enhance its rebound spin. The physical mechanism behind the enhanced
spin, as well as the enhanced value of ex , is apparent from the results in Figs. 4-7. As shown
in Figs. 6 and 7, the friction force does not drop instantly to zero when the ball grips, but it
decreases slowly to zero as a result of the forward stretch of the ball. The angular velocity
of the ball continues to increase while F and the resulting torque remains positive, and it
can increase to a value where Rω is significantly larger than vx . When the friction force
reverses, the angular velocity begins to decrease and vx begins to increase. In the examples
shown in Figs. 5 and 7(a), the angular velocity decreases to such an extent that the ball
rebounds with Rω2 < vx2 . However, a superball bounces with Rω2 about twice as large as
vx2 since the reversal in the friction force occurs near the end of the impact rather than

21
about half way through the impact [3]. The difference in the two balls can be explained
partly by the fact that the ratio of tangential to normal vibration frequency was higher for
the hollow ball, and partly by the fact that the hollow ball had a much higher coefficient of
sliding friction and therefore had a shorter sliding phase.

Given that vx decreases and ω increases by amounts proportional to the horizontal im-
pulse, it is clear that S and ex will be maximised if the reversal in the direction of F is
delayed as long as possible after the ball grips. If the ball grips early and if F reverses at or
near the end of the impact, then the horizontal impulse will be relatively large. However, if
the ball grips early, then F will tend to reverse early if the tangential vibration frequency
remains constant. If F reverses well before the end of the impact then the horizontal impulse
is reduced. That was the case with the hollow ball, with the result that ex was actually
negative at angles of incidence between 15◦ and 60◦ .

Grip phase

It is conventionally assumed that a ball incident on a surface will initially slide with vx > Rω
and will then roll or grip when vx = Rω. If that were the case then the contact region of
the ball should slide backwards during the grip phase while vx < Rω, as indicated in Fig. 7.
However, there is no evidence of any backward sliding in the video images presented in
Fig. 4, despite the fact that the angular velocity shown in Fig. 5 is qualitatively consistent
with the calculations shown in Fig. 7. That is, the evidence shows shows that vx was less
than Rω during the grip phase but there was no sliding, except perhaps near the end of
the impact period when the ball bounced with S > 1 or with S < 1. A relatively simple
explanation can be given in terms of the change in R during the impact, as indicated in
Fig. 10.

ω
vx
G
r

FIG. 10: During the grip phase, vx = rω

While the ball is compressed, the centre of mass, G, will be located at a distance r < R

22
above the horizontal impact surface, and translates horizontally at velocity vx . A point P on
the ball’s surface, directly below G, will be instantaneously at rest if vx = rω. Point P will
remain at rest on the surface as the ball translates horizontally, provided that all points in
the ball at the same horizontal level as G are also translating with horizontal speed vx = rω.
Consequently, vx < Rω while the ball grips the surface. Translation of the ball during the
grip phase can then be likened to the motion of a military tank on caterpillar track. That
is, the track remains at rest on the ground as the tank moves forward. A bouncing ball
is different since the centre of mass of the ball moves vertically as well as horizontally. As
the centre of mass rises, r increases and rω can then exceed vx , causing the ball to slide
backwards on the surface. During the grip phase, F < µN . In order to slide backwards, N
must also decrease so that the ball can release its grip when F = µN . During the bounce,
points to the right of G also rotate down towards the surface, while points to the left of G
also rotate away from the surface. Consequently, the normal reaction force on the right side
of the ball can be expected to be larger than the reaction force on the left side.

Conclusions

The results presented in this paper support and extend previous observations of bouncing
balls. It was shown that the outgoing ball speed, spin and angle (or the equivalent ratio
vy2 /vx2 ) can be described analytically in terms of the incident conditions and three bounce
parameters ey , ex (or µ) and D, although all three parameters vary with the angle of inci-
dence. Nevertheless, the variation with angle is well-behaved and can be explained in terms
of a transition from pure sliding at low angles of incidence to grip at high angles of incidence.

It is well known that the sliding phase of a ball can be described theoretically in terms of
the coefficient of sliding friction and the normal coefficient of restitution. Less well known
is the fact that non-ideal effects can also influence the outcome, particularly if the ball is
flexible. The latter effects were quantified in the present paper in terms of the parameter D,
which was found to be relatively large and strongly variable at low angles of incidence, for
both of the balls studied. The origin of that variability was not established in the present
paper.

Measurements of the spin and the friction force during the bounce of the hollow rubber
ball shed new light on the grip phase of the bounce. Video film of the bounce was especially

23
revealing since it showed that there was negligible slip during the grip phase and that the
whole ball vibrates tangentially when it grips. The behaviour of the friction force is directly
related to tangential vibration of the ball. The hollow rubber ball had a high coefficient
of sliding friction and therefore gripped the surface early on during the bounce. It also
had a relatively high tangential vibration frequency, resulting in an early reversal of the
friction force at all angles of incidence and a second reversal at high angles of incidence.
The consequence for this ball was that ex was relatively low at all angles of incidence.

By contrast, ex was relatively high for the superball, a result that can be explained in
terms of its low tangential vibration frequency. In addition, the coefficient of friction was
relatively low, meaning that the ball gripped the surface at a later stage of the bounce than
the hollow ball, thereby delaying the reversal of the friction force until a relatively late stage
of the bounce. As a result, the horizontal impulse on the ball was not strongly affected by
the reversal in the friction force, and the tangential speed at the surface of the ball increased
to more than twice the horizontal speed of the center of mass at angles of incidence greater
than 40◦ . The latter result, when compared with the result for the hollow ball, suggests
that the ex value of a ball could potentially be tailored to suit any given application by an
appropriate choice of the coefficient of sliding friction and an an appropriate choice of the
mass and stiffness distribution of the ball.

References

1. Foerster S, Louge M, Chang H, Allia K (1994) Phys. Fluids 6:1108-1115


2. Kharaz A, Gorham D, Salman A (2001) An experimental study of the elastic rebound of
spheres, Powder Technology 120:281-291
3. Cross R (2002) Grip-slip behavior of a bouncing ball, Am. J. Phys. 70(11):1093-1102
4. Smith L, Duris J (2009) Progress and challenges in numerically modeling solid sports
balls with applications to softballs. J Sports Sci 27(4):353360
5. Ranga D, Strangwood M (2010) Finite element modelling of the quasi-static and dynamic
behaviour of a solid sports ball based on component material properties. Procedia Eng
2:32873292
6. Rezaei A, Verhelst R, Van Paepegem W, Degriecka J (2011) Finite element modelling
and experimental study of oblique soccer ball bounce, J. Sports Sc. 29(11):1201-1213

24
7. Carre M J, James D M, Haake S J (2004) Impact of a non-homogeneous sphere on a rigid
surface, Proc. Instn. Mech. Eng. Part C: J. Mech. Eng Sci. 218(3):273-281
8. Goodwill S R, Haake S J (2004) Modelling of tennis ball impacts on a rigid surface, Proc.
Instn. Mech. Eng. Part C: J. Mech. Eng Sci. 218:1139-1153
9. Goodwill S R, Kirk R, Haake S J (2005) Experimental and finite element analysis of a
tennis ball impact on a rigid surface, Sports Eng 8:145-158
10. Haake S J, Carre M J, Kirk R, Goodwill S R (2005) Oblique impact of thick walled
pressurized spheres as used in tennis, Proc. Instn. Mech. Eng. Part C: J. Mech. Eng Sci.
219(11):1179-1189
11. Tanaka K, Sato F, Oodaira H, Teranishi Y, Sato F, Ujihashi S (2006) Construction of
the Finite-Element Models of Golf Balls and Simulations of Their Collisions, Proc. IMechE
Vol. 220 Part L: J. Materials: Design and Applications, 220:13-22
12. Tanaka K, Matsuoka K, Fujita S, Teranishi Y, Ujihashi S (2012) Construction of a finite
element model for collisions of a golf ball with a club during swing, Proc. Instit. Mech.
Eng., Part P: Jnl of Sports Eng and Tech, 226(2):96-106
13. Maw N, Barber J R, Fawcett J N (1981) The role of elastic tangential compliance in
oblique impact, Jnl Lubrication Technology 103:74-80
14. Labous L, Rosato A, Dave R (1997) Measurements of collisional properties of spheres
using high-speed video analysis, Phys Rev E 56(5):5717-5725
15. Lorenz A, Tuozzolo C, Louge M (1997) Measurement of impact properties of small,
nearly spherical particles, Experimental Mechanics 37(3):292-298
16. Cross R (2002) Measurements of the horizontal coefficient of restitution for a superball
and a tennis ball, Am. J. Phys. 70(5):482-489
17. Garland P, Rogers R (2009) An Experimental Study of Contact Forces During Oblique
Elastic Impact, J App Mech 76:031015
18. Cross R (2003) Measurements of the horizontal and vertical speeds of tennis courts,
Sports Eng. 6:93-109
19. Maw N, Barber J R, Fawcett J N (1976) The oblique impact of elastic spheres, Wear
38(1):101-114
20. Garwin R (1969) Kinematics of an ultraelastic rough ball, Am. J. Phys. 37(1):88-92
21. Cross R (1999) Impact behavior of hollow balls, Am. J. Phys. 82(3):189-195
22. Cross R (2008) Differences between bouncing balls, springs, and rods, Am. J. Phys.

25
76(10):908-915
23. Kuninaka H, Hayakawa H (2004) Anomalous Behavior of the Coefficient of Normal
Restitution in Oblique Impact, Phys. Rev Letts. 93(15):154301-1-4
24. Louge M, Adams M (2002) Anomalous behavior of normal kinematic restitution in the
oblique impacts of a hard sphere on an elastoplastic plate, Phys Rev E, 65:021303-1-6

26

View publication stats

You might also like