V. QUALITATIVE EXPERIMENTS
Several experiments were conducted to assist in the interpretation of the above measurements. The first was to squash four large superballs in a shallow wood box in such a way that the balls could be dragged across a surface without rolling, and lead bricks could be placed on the upturned box to increase the normal reaction force. When a small horizontal force is applied to the box, the bottom of the balls remain stuck to the surface by static friction. However, the box itself was observed to move horizontally through a small distance due to the fact that the balls stretch slightly in the horizontal direction. At a sufficiently large value of the horizontal force, the balls release their grip and start to slide in the direction of the force. A ball can therefore store elastic energy due to the fact that it stretches horizontally when subject to a horizontal force. A central feature of the bounce model of Ref. 5 is that when a ball releases its grip and starts to slide, it does so progressively. The edge of the contact area slips first, while the central part of the contact area remains stuck because the normal reaction force acting in the central region is larger than that at the edge of the contact area. As the horizontal force increases, the area that is stuck shrinks until the whole contact area slides. A dynamic version of the above experiment was performed by gluing a tennis ball to a 260-g wood block and attaching a small piezo disk to the side of the ball. A 12-kg load was placed on top of the ball while the glue dried so that a large circular area of the ball adhered to the block. After drying, the block was rotated by 90° and dropped on a horizontal surface to excite tangential oscillations in the ball. In this orientation the piezo disk was at the bottom of the ball and the ball vibrated in a vertical direction. The excitation of high frequency transverse modes was minimized because the ball did not compress in a direction perpendicular to the vertical surface of the wood block. The ball was observed to undergo five damped oscillations with a period of 5.2 ms. Because the impact duration of a low speed tennis ball is about 6 ms, the ball can undergo a fraction more than one complete cycle of tangential oscillation during the bounce. The data shown in Fig. 3(b) are consistent with this result. In Fig. 3(a), the friction force reverses direction only once, indicating that the bottom of the ball gripped the surface at a later stage of the impact than in Fig. 3(b). It is well known that each contact point on the circumference of a rigid ball comes to rest momentarily on a surface when the ball rolls. When the ball is flexible and subject to a vertical force, the ball squashes and the contact point enlarges to a flat, circular area. To investigate whether some or all contact points remain at rest when the ball rolls, a large and relatively soft rubber ball was rolled on a table under a thick plate of glass to observe the effects by eye. The ball was marked by a series of dots around a circumference and a straight line was drawn with a felt pen across the width of the glass plate. The dots and the line on the plate were made to coincide by pushing down on the glass plate, and then the plate was pushed sideways by hand to allow the ball to roll. Each dot remained attached to the line without slipping until it reached the edge of the contact area and rotated away from the glass. Consequently, a squashed ball can roll on a surface in such a way that all points in contact with the surface remain at rest on the surface. When the same experiment was repeated with a tennis ball, the dots on the ball gradually slipped behind the line as the ball rolled forwards, indicating that the low coefficient of friction between a tennis ball and the glass plate allowed both grip and slip to occur. In the absence of friction the plate would simply slide across the top of the ball and the ball would not roll forwards.