ARTICLE IN PRESS

International Journal of Impact Engineering 30 (2004) 31–53

The effect of temperature on the impact behaviour of

glass/polycarbonate laminates
S.M. Walley*, J.E. Field, P.W. Blair, A.J. Milford
Cavendish Laboratory, Physics and Chemistry of Solids (PCS), Madingley Road, Cambridge CB3 0HE, UK
Received 3 October 2002; received in revised form 11 March 2003

Abstract

A systematic investigation has been performed into the mechanisms that lead to spall of bullet-proof
glass/polycarbonate laminate constructions at low temperatures. A range of techniques were deployed
including liquid jet impact, high speed photography, and a light gas gun. Spall formation was shown to take
ca. ten times longer than a simple shock-wave overlap mechanism would imply. However, a number of
experimental observations including the radii of fracture bands in the glass and fractographic examination
of the edge of the polycarbonate spalls led us to consider a more complex elastic wave analysis: the
reinforcement of Rayleigh surface waves by tensile elastic waves reflected from the various boundaries
within the laminate. This predicted that the radius of the spall should be proportional to the thickness of
the front glass layer. Comparison with the results of full-scale tests confirmed that this was indeed the case.
r 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Bullet-proof; Laminates; Polycarbonate; Glass; Spall

1. Introduction

Glass/polymer laminated constructions are often used as blast-proof safety screens or bullet-
proof windows or windshields [1,2]. A typical bullet-proof construction consists of glass as the
outer layer backed by polycarbonate (PC) with a thin adhesive transparent interlayer (such as a
polyurethane) in between. Ordinary laminated windshields usually have glass instead of PC as the
backing material e.g. Refs. [3–5].
Some constructions have more than one layer of glass [2]. The PC is usually coated with an
abrasion resistant coating. This can be of two main types: a soft coating that heals after being
scratched or a hard coating that resists scratching in the first place. The effects of both types of

*Corresponding author. Fax: +44-1223-350-266.

E-mail address: [email protected] (S.M. Walley).

0734-743X/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0734-743X(03)00046-0
 ARTICLE IN PRESS

32 S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53

coating were examined in this study. The soft coating was an elastometric carbon-based polymer;
the hard coating was a silicon-based polymer (‘siloxane’).
PC has the advantage over other transparent polymers such as polymethylmethacrylate
(PMMA) of being very tough e.g. Refs. [6–25]. If a bullet is fired against such a construction,
although the glass shatters, it is still able to deform and slow the projectile enough so that the
tough PC is able to prevent penetration of the projectile (or other fragments) by bulging
plastically [26–33].
The main problem with PC is that it embrittles with age [34]. This can be due to (i) physical
ageing (the slow approach of an amorphous polymer towards thermodynamic equilibrium) e.g.
Refs. [34–45], (ii) chemical changes due to, for example, ultraviolet (UV) light, ozone, nitrogen
oxides, moisture or solvents e.g. Refs. [46–68], or (iii) physical damage e.g. Refs. [15,69,70]. Such
problems can be alleviated to some extent in a laminate because the glass and interlayer material
reduce the amount of UV reaching the PC, while an abrasion-resistant coating reduces the
diffusion of gases and surface damage [71].
A more serious problem, however, has been identified in service when laminates are impacted at
low temperatures (ca.
7 C or less). In that case, the PC has been observed to spall off a disc of
material several centimetres in diameter. It should be noted that the fracture strain of PC in
compression under 1D stress loading and high rates of strain (>103 s
1) does not change
significantly between room temperature and ca. 100 K [72]. So any change in the behaviour of the
PC must be due to effects of the 3D state of loading or its incorporation into a structure or a
combination of the two.
The main purpose of this study, therefore, was to study the mechanisms that lead to the formation
of spall in PC at low temperatures with the aim of suggesting improvements to the laminate design
to overcome this problem. The approach taken was to study the polymer components separately, the
effect of the heat cycle used in the laminate manufacturing process, the effect of the abrasion-
resistant coatings, and finally complete laminate constructions. Due to size constraints, we studied
small-scale constructions, in the range 50–75 mm diameter/width. The projectiles used were steel
spheres a few millimetres in diameter fired from a laboratory light gas-gun [73].

2. Experimental

2.1. Liquid impact

A convenient and simple technique that can be used to generate an intense localised shock is to
fire a jet or drop of water against a target. The device used in this study is a modified air-rifle in
which the lead slug is used as a high-speed piston to extrude water through a nozzle [74]. This
technique was originally developed to study rain erosion damage [75,76]. Impact velocities up to
1.5 km s
1 can be achieved.
As the front surface of the water drop or jet is in general curved, the contact area initially
expands faster than any wave in either the target material or the water. During this period the
water is effectively inertially confined [75,77,78]. Eventually the shock envelope overtakes the
expanding liquid/solid contact periphery and release waves enter the compressed liquid. At this
stage, the pressure reduces to a much lower value and incompressible flow starts. Damage in the
 ARTICLE IN PRESS

S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53 33

Fig. 1. Schematic diagram of the light gas-gun (not to scale).

target is largely caused by the initial pressure spike, often termed the ‘water hammer’ pressure; see
Ref. [79] for a recent review.
When this compressive pressure pulse reflects off the back surface of the target, it changes
phase. Because of this, the resultant stress in the material eventually becomes tensile. If the tensile
stress exceeds the dynamic failure stress, fracture (or ‘spall’) occurs.

2.2. Solid particle ballistic impact

We have a range of light gas-guns capable of firing solid projectiles at up to 1.5 km s
1. The
firing mechanism used in this study was of the double-diaphragm type [73]. The steel sphere
projectiles were accelerated on the front of a plastic sabot to a pre-selected velocity, dependent on
the firing pressure. The sabot is stopped at the end of the barrel, leaving the sphere to continue in
free flight to strike the target. The impact velocity was measured in free flight using two laser
beams (see Fig. 1).
Low temperature experiments were performed by placing a bath of liquid nitrogen in the
impact chamber above the target. The specimen was therefore cooled to the chosen temperature
by forced convection. A thermocouple was used to monitor the specimen temperature.

3. Results

3.1. Liquid impact on coated PC

PC discs of nominal thickness 3 mm and coated with a carbon-based polymeric abrasion-

resistant coating were impacted with a water jet travelling at a velocity of 700 m s
1 at
 ARTICLE IN PRESS

34 S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53

temperatures of
50 C,
30 C,
10 C and +17 C [80]. The nozzle used and stand-off distance
were such that the contact diameter was ca. 3 mm, contact pressure was ca. 1 GPa, and the shock
pressure duration was ca. 1 ms. Care was taken to remove any films of ice or frost from the
specimens before impact. Liquid jet impacts were also performed on siloxane-coated PC discs at
two temperatures:
50 C and +17 C. A minimum of six independent impacts were made at each
temperature.
The impact sites were recorded using optical photography and profilometry. Selected sites were
sectioned, polished and photographed in section.
Examples of the damage produced by the liquid jets on the polymer-coated PC specimens at the
four different temperatures are shown in Fig. 2. The damage sites consist of three main regions:
(a) a central depression within the contact area of the liquid jet; (b) a ring where the coating has
been eroded away by radial jetting of the liquid after confinement of the water has ended; (c) an
annular region of plastic deformation beneath the eroded coating.
A typical profilometer trace of a crater formed at
10 C is shown in Fig. 3a. It is clear that the
volume of material displaced above the original surface level is less than that displaced below.
There are two possible explanations: (a) the rear surface of the disc could have bulged out; (b) the
material beneath the crater could have densified. Experiments on ball impact on polyethylene
showed that densification of up to 10% is possible in that material [81]. A graph of the crater
depth (measured by profilometry) versus temperature is presented in Fig. 3b. Clearly the lower the
temperature, the shallower the crater. This is simply due to the flow stress of PC at high strain
rates being higher at lower temperatures [82].

Fig. 2. Optical photographs of selected liquid impact damage sites on carbon-based polymer-coated PC specimens at:
(a) 17 C, (b)
10 C, (c)
30 C, (d)
50 C. ‘a’ is central depression, ‘b’ points to area of coating removal, ‘c’ points to
a ring of plastic deformation. From Ref. [80].
 ARTICLE IN PRESS

S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53 35

Fig. 3. (a). Typical profilometer trace obtained from a PC specimen subjected to liquid impact at 700 m s
1 and
10 C.
(b). Plot of depth of crater versus temperature. ‘a’ is central depression, ‘b’ points to region of coating removal, ‘c’
points to zone of plastic deformation. From Ref. [80].

Cross-sections of two specimens impacted at +17 C and
30 C are shown in Fig. 4. Spall
damage can be clearly seen. It is also clear that the extent of spall damage is greater at the lower
temperature.
The damage was found to increase if multiple impacts were performed on the same site. At
room temperature, all the 3 mm thick PC specimens, whether coated or uncoated, were perforated
by the time five impacts had been carried out. But at
50 C, none of the specimens were
perforated after five impacts. However, a radial crack formed at the rear surface at this
temperature, this being ca. 2 mm long in a specimen with a hard coating and ca. 1 mm long in a
specimen with a soft coating.
The damage to siloxane-coated specimens was essentially the same. However, the coating
exhibited an extra mode of failure: a series of fine cracks. These were also seen in specimens
impacted by solid particles. A photograph of these fine cracks is presented later in the paper.

3.2. Solid particle ballistic impact on uncoated PC

Fifty millimetre diameter uncoated PC discs of two different thicknesses (2 and 3 mm)
were supported by an annular steel holder of inner diameter 37 mm. Four mm diameter steel
 ARTICLE IN PRESS

36 S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53

Fig. 4. Cross-section of PC specimens that were subjected to liquid impact at 700 m s
1 and temperatures of: (a) +17 C
and (b)
30 C. ‘a’ points to central depression, ‘b’ points to coating removal, ‘s’ points to spall. From Ref. [80].

spheres were fired at these targets using the light gas-gun described above. At room tem-
perature (+20 C), the penetration velocity of the 2 mm thick targets was 310 m s
1 whereas
for the 3 mm thick specimens it was 340 m s
1. The resulting holes were always smaller
than the spheres, indicating viscoelastic recovery had taken place, a well-known phenomena in
polymers e.g. Ref. [83]. As the temperature was reduced, two changes in behaviour were
noted. First, at
25 C, the hole formed in a 3 mm thick specimen was much ‘cleaner’, as if a
sharp punch had been used. At temperatures between
80 C and
40 C penetration did
not occur even at velocities of 360 m s
1 (the limit at the time for the gun used) for 3 mm
thick discs. The penetration velocity for 2 mm thick discs at
55 C was measured as 325 m s
1.
The conclusion therefore is that the ballistic impact resistance of PC is slightly higher at
low temperatures. This is consistent with the observations of Walley et al. [72,82] that the
fracture strain of PC in compression at high strain rates is the same at ca. 100 K as at room
temperature although the flow stress is higher. Photographs of selected impact sites are presented
in Fig. 5.
 ARTICLE IN PRESS

S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53 37

Fig. 5. Optical photographs of solid particle impact sites on uncoated PC discs at: (a) 20 C, 350 m s
1; and (b)
80 C,
360 m s
1 (birefringence fringes are shown around the impact site using polarisers).

3.3. Solid particle ballistic impact on coated PC discs

Four millimetre diameter steel spheres were fired at PC discs faced with two types of abrasion
resistant coatings at impact speeds X340 m s
1 and at temperatures down to
80 C. The most
striking difference between the coated and uncoated specimens was that below some critical
temperature, the discs shattered. The critical temperature T for the specimens coated with the
carbon-based material was Tp6075 C, whereas that for the specimens coated with the silicon-
based material was
30 C to
45 C (specimens impacted below
15 C cracked but did not
shatter). Penetration was necessary for brittle fracture in both cases i.e. the impact velocity had to
be higher than ca. 350 m s
1.
For the specimens coated with the carbon-based coating, there was a transition range of around
10 C i.e. above
55 C, none of the specimens shattered, below
65 C all the specimens
shattered, but between
65 C and
55 C, some fractured and some did not. The heat treatment
necessary to bond the glass to the interlayer material was also found to be necessary to embrittle
 ARTICLE IN PRESS

38 S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53

the coated discs. Specimens that had the carbon-based abrasion resistant coating but which had
not been subjected to a heat cycle did not shatter when impacted at 360 m s
1 in the temperature
range
50 C to
70 C. It should be noted that uncoated PC discs that had been subjected to the
heat cycle were not brittle: such discs proved capable of stopping 4 mm diameter steel spheres
travelling at 360 m s
1 at a temperature of
67 C. We conclude that there is some interaction
between the soft carbon-based abrasion resistant coating and the PC during the heat cycle that
lowers the ballistic impact toughness. We emphasize that neither the soft coating on its own nor
the heat treatment on its own embrittle the PC within the temperature range the laminate
construction is likely to see in service. Photographs of selected impact sites on coated discs are
presented in Fig. 6.
One extra feature seen only on the siloxane-coated PC discs were hairline cracks in the coating
(see Fig. 7). These were circumferential on the impact side (not illustrated) and radial on the
reverse side (see Fig. 7a). This is consistent with the Rayleigh surface wave causing circumferential
cracks on the front surface and bending stresses the radial cracks on the rear. The radial cracks
propagated out to within about 2 mm of the specimen holder ring and then turned through 90 .
Two ‘singular’ points were observed at about 120 to each other where the direction of crack
turning changed by 180 . As the crack turning was close to the supporting ring, it was probably
caused by their interaction with bending modes of the disc rather than with reflected elastic waves.
Such patterns of cracks in brittle surface layers have been reported by Durelli et al. [84] as giving
useful information about the state of stress, both static and dynamic, in various engineering
structures.

3.4. Solid particle ballistic impact studies on PC discs with polyurethane adhesive layer

A vital part of the bullet-proof glass/PC laminate structure is the adhesive interlayer that bonds
the two materials together. A number of different polymeric adhesives can be used [2], but in this
study we examined only one, namely a polyurethane.
Three millimetre thick discs of PC with a 1 mm thick polyurethane layer bonded to them were
impacted with 4 mm diameter steel spheres. The impacts took place on the side with the
polyurethane layer. Above around
30 C, the two-layer construction was capable of stopping the
spheres at velocities up to 355 m s
1 with very little damage. Below
30 C, the PC spalled with a
piece of the polymer about 20 mm in diameter being knocked out. The polyurethane suffered very
little damage, having only a pinhole left in it where the sphere had passed through (see Fig. 8).

3.5. Double impacts

Glass/PC laminate constructions are required in service to be able to stop two bullets impacting
four or more inches (100 mm) apart. For this reason, we performed a number of double-impact
experiments. As we were using scaled-down specimens, the average distance between impacts was
about 25 mm.
We found there were distinct temperature and geometry effects to do with whether a crack
initiated from the first impact site during a second impact on the same specimen. Square PC
specimens (75 mm2, 3 mm thick) only cracked if the temperature were less than
20 C (impact
speed 350 m s
1; see Fig. 9). By contrast, 50 mm circular PC discs always cracked even at room
 ARTICLE IN PRESS

S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53 39

Fig. 6. Optical photographs of solid particle impact sites on carbon-based abrasion-resistant coated PC discs at various
temperatures. 4 mm diameter steel spheres were used. (a) No heat treatment, 360 m s
1,
53 C (birefringence fringes are
shown around the impact site using polarisers); (b) with heat treatment,
67 C. Diameter of specimen: 50 mm.

temperature (20 C) for an impact speed of 350 m s
1. It was also found that the second impact did
not need to penetrate the disc to initiate a crack at the first impact site: a crack could form from
the first impact site for an impact speed as low as 220 m s
1 for the second impactor. Conversely, if
the first impact did not penetrate, cracks did not form even if the second impact did. We thus
conclude that penetration of the PC plate produces damage sites from which cracks can grow if
stress waves or plate bending modes are excited by a subsequent impact. The fact that the shape of
the plate matters in this phenomenon implies that plate bending modes are the main factor in
causing cracks to grow from the damage sites as stress waves propagating out from an impact site
 ARTICLE IN PRESS

40 S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53

Fig. 7. Optical photographs of solid particle impact site (260 m s
1) on siloxane-coated PC discs showing radial hairline
cracks on rear surface: (a) close to impact site; and (b) close to specimen support ring.

would not be affected by the shape of the plate, and any stress-wave focussing effects due to
reflections would depend on the relative positions of the impact sites.

3.6. Solid particle ballistic impact studies on glass/polyurethane/PC laminates

As the geometry of specimens had previously been found to have an effect on the damage
processes, circular discs of various diameters were tested, namely 50, 100, and 152 mm. These were
compared with 100 mm2 plates. In all cases, the thickness (in mm) of the glass/polyurethane/PC
layers was 3/1/3. Impacts were performed at 350 m s
1 with 6 mm diameter steel spheres over a
range of temperatures in order to find the critical temperature below which spall took place.
Table 1 summarises the results. It can be clearly seen that increasing the diameter of the circular
laminate discs lowered the spall temperature. Also the square specimen of the same linear
dimension had a lower spall temperature than a circular specimen.
In the contact area, the glass was intensely fragmented down to powder size. A small hole was
punched through the interlayer. Further out, there were both radial and circumferential crack
 ARTICLE IN PRESS

S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53 41

Fig. 8. Optical photographs of spalling of a 50 mm diameter polyurethane-coated PC specimen impacted at 350 m s
1
and
30 C. (a) shows the disc and (b) shows the spall.

systems (see Fig. 10a). At certain radii, the circumferential cracks concentrated into intense bands
of fracture, and it was this observation that first gave us a clue as to the mechanism of spall
formation (see below). As the temperature of the test was lowered, the glass fragmentation
became less (Fig. 10b–d), consistent with the observation that PC and polyurethane become more
rigid due to the b mechanical loss peak becoming activated (Fig. 11; see also Refs. [39,85–89]) so
that the laminate bends less.
Examination of the recovered spall fragments showed that the diameter of the attached
interlayer was always less than or equal to the diameter of the PC fragment (see Fig. 12). The
 ARTICLE IN PRESS

42 S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53

Fig. 9. Double impacts on 75 mm2 specimens at 350 m s
1: (a)
15 C, (b)
20 C, (c)
25 C, and (d)
60 C;
(birefringence fringes are shown around the impact sites using polarisers).

Table 1
Effect of specimen size and geometry on spall temperature
Specimen type Spall temperature/ C
50 mm diameter disc
2571
100 mm diameter disc
2871
152 mm diameter disc
3371
100 mm2
3671

spalls were roughly circular in shape. Fractographic examination of the edge of the spalls showed
that cracks initiated at between 3 and 6 independent sites (see Fig. 13). Irregularities in the shape
of the spall fragment are due to these sites not being exactly at the same radius (see Fig. 16a).
 ARTICLE IN PRESS

S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53 43

Fig. 10. (a) Result of the impact of a 6 mm diameter steel ball on a 50 mm diameter laminate disc at 350 m s
1 at a
temperature of
25 C. Photograph shows system of radial and circumferential cracks along with intense fracture bands
at a radius of 30 mm. (b–d) Photographs of 152 mm diameter laminate discs after impact with a 6 mm diameter steel ball
at 350 m s
1 at the following temperatures: (b) +20 C; (c)
28.5 C; and (d)
34 C. Note that as the temperature is
reduced, the amount of cracking in the glass also reduces due to the increased rigidity of the polymer backing. Spall of
the PC/polyurethane took place in the coldest specimen, and the hole is labelled S.
 ARTICLE IN PRESS

44 S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53

100

80 -20 °C, 9000±500 s-1

Average stress/MPa
60
RT, 7900±250 s-1

40

20
+70 °C, 8200±200 s-1

0
0 0.2 0.4 0.6 0.8 1
True strain

Fig. 11. Average high strain-rate stress-strain curves for a polyurethane used in a laminate construction as a function of
temperature. Each curve is an average of four data sets. The data was obtained using a split Hopkinson pressure bar.
The dotted lines indicate the standard deviation of the data.

An important observation was that the radius of the spall is not related to the diameter of the
specimen. Rather it is related to the thickness of the glass front-plate.
High-speed photography (Hadland Imacon) was used to determine the time for spall formation.
If this were due to the conventional tensile elastic wave overlap mechanism identified by
Hopkinson [90,91] and Rinehart [92–94] for metals and Kolsky [95,96] for polymers, the
formation time would be close to the double wave transit time of shock waves through the
laminate. As the wave velocities in glass, polyurethane, and PC are approximately 5.6, 1.5, and
2.5 mm ms
1, respectively, and the thicknesses of component layers are 3/1/3 mm, the spall
formation time would be ca. 4.8 ms via this mechanism.
Fig. 14 gives an example of a high-speed photographic sequence of spall formation. Note
that the 7 mm thick laminate is mounted in a plate holder 20.5 mm thick. The distance from the
LH edge of the holder to the front surface of the laminate was 10.0 mm, and the distance from
the rear of the laminate to the RH edge of the plate holder was 3.5 mm. By extrapolating
the motion of the ball and spall, the time for spall formation can be calculated. In three
experiments performed, this time was 3873 ms, larger by nearly an order of magnitude than the
time calculated using a simple 1D wave analysis. The exit velocities of the spalls were typically ca.
200 m s
1. Optical techniques can be used to examine the local wave propagation and global
bending modes of laminates [97,98], but these would be difficult to apply at cryogenic
temperatures due to frost and vapour problems. A full analysis of the interaction of the projectile
and subsequent plate bending and oscillations requires frictional and 3D effects to be taken into
consideration [99].
The photographic sequences give clear evidence that the spall formation time is longer by an
order of magnitude than what would be expected on the basis of the classic elastic wave overlap
mechanism. This combined with fractographic evidence of crack initiation sites in the PC and
fracture bands in the glass led us to postulate the following mechanism for spall formation in these
laminated constructions.
 ARTICLE IN PRESS

S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53 45

Fig. 12. (a) Photograph of a typical spall fragment. Note that the radius of the interlayer is smaller than that of the PC.
The fractured glass has imprinted and cut the interlayer material. There are several fractures under the interlayer
material at, for example, F and F0 . These fractures are associated with fracture bands in the glass. The outer boundary is
made up of several segments due to cracks initiating at more than one site. ‘A’ marks the boundary between two cracks.
(b) Photograph of a spall fragment produced by the impact of a 6 mm diameter steel sphere at 350 m s
1 on a 100 mm
diameter disc at
29.5 C. The lines AB in (a) and (b) was where a plastic hinge existed before separation.

4. Discussion

The radial cracks in the glass nucleate at the rear surface of the glass plate as it bends under the
impact of the ball. These cracks travel at about one-third of the longitudinal elastic wave velocity.
 ARTICLE IN PRESS

46 S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53

Fig. 13. Photographs of the periphery of a 3 mm thick PC spall. (a) ‘O’ marks a fracture initiation site on the top
surface of the PC. XY marks a line where the fracture paused as it encountered a region in compression. (b) and (c) are
further views around the periphery of the spall between two fracture initiation sites O and O0 (note the sequence does
not go all round the fragment). The line AB in (b) is the boundary between the two cracks initiated at O and O0 .

They therefore meet the reflected tensile elastic wave about halfway to the boundary changing the
nature of the fracture pattern (see Fig. 10 and the discussion in Ref. [100]). The glass within this
region is crossed by radially directed cracks and hence has little cohesion. The recovered
specimens therefore have often lost most of the glass within a region whose radius is half that of
the total. Outside this region, the glass is fragmented into smaller pieces which can lock together
like a jigsaw. Note that the glass fracture pattern is strongly imprinted onto the soft polyurethane
interlayer (Fig. 15). Where the fracture is particularly severe, the interlayer can be cut right
through introducing defects into the surface of the PC.
However, the most important system of cracks within the glass from the point of view of
spall formation in the PC plus interlayer are the fracture ‘bands’ (see Fig. 10a in particular). This
 ARTICLE IN PRESS

S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53 47

Fig. 14. Selected frames from a high-speed framing sequence taken with a Hadland Imacon camera. The framing
interval was set to 17 ms. The numbers in each frame give the order the camera took them in. Note that frames 3–6 have
been removed as nothing of significance was happening in them. The laminate is held in a plate holder (labelled ‘p’)
which was 20.5 mm thick. The ball (labelled ‘b’) in frame 1 approaches from the left. The spall (labelled ‘s’) is ejected
from the right. Note that the spall rotates around one edge due to the formation of a plastic hinge.

Fig. 15. Imprint of glass fracture pattern on polyurethane interlayer material after the impact of a 6 mm steel sphere at
350 m s
1 on a 50 mm diameter laminate disc at
24 C.
 ARTICLE IN PRESS

48 S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53

Fig. 16. Reflexion and reinforcement of impact-generated stress waves in a thin plate (from Ref. [75]).

form of fracture was first discussed by Bowden and Field [75], and their analysis is summarised
here.
When an impact or other dynamic event takes place at (or close to) the surface of a
solid, several types of bulk and surface elastic waves are excited [101–104]; see also Fig. 16.
The most important waves from the point of view of the formation of fracture bands are the
bulk elastic longitudinal cl and shear cs waves and the surface Rayleigh cR waves. These
travel at markedly different speeds: cl ¼ 5750; cs ¼ 3370; and cR ¼ 3100 m s
1 for soda-lime
glass. What this difference in speeds means is that the longitudinal wave reflected off the back
surface as a tensile wave can reinforce the Rayleigh surface wave at certain radii, opening up
surface flaws [105]. Several possibilities exist, and the full analysis published by Bowden and
Field [75] allowed them, for instance, to take account of the effect of the finite area of
the projectile contact zone and the influence of Poisson’s ratio of the materials. The most
important prediction of their analysis for the work presented in this paper is that the diameter
of the spall should be proportional to the thickness of the glass. Evidence supporting this
conclusion was the fact that the spall diameter of a full-scale construction (total thickness 33 mm)
was 45 mm.
 ARTICLE IN PRESS

S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53 49

5. Conclusions

(i) We have shown that small-scale experiments can reproduce the main features of the
mechanisms that lead to spall formation in laminated bullet-proof window constructions at
low temperatures.
(ii) Geometric effects were identified which affect the critical temperature for spall formation.
These effects also explain why small-scale specimens have a lower spall temperature than full-
scale constructions.
(iii) An elastic wave analysis showed that reinforcement of the Rayleigh surface wave in the glass
should take place at distinct radii from the point of impact. This lead to bands of intense
fracture which are able to penetrate the interlayer and produce surface flaws on the PC.
(iv) Bending of the laminate causes cracks to initiate from these surface flaws if the temperature is
low enough. These cracks then propagate in a ring around the impact point leading to
detachment of a spall fragment, whose typical formation time is 38 ms for the laminate
constructions investigated.
(v) One way of preventing spall formation would be to have a double layer of PC at the rear of
the construction. The first PC layer would spall, but cracks would not be able to propagate
through to the rearmost layer from the front glass layer.

Acknowledgements

The research was supported by Pilkingtons plc who also provided specimens and EPSRC (then
SERC) who gave a grant for the Imacon camera. D. Hand and C.W.G. Hall of Pilkingtons are
thanked for their interest.

References

[1] Toqueboeuf W, Mortaigne B, Cottenot C. Dynamic behaviour of polycarbonate/polyurethane multi-layer for

transparent armor. J Phys IV France 7 Colloq. C3 (EURODYMAT 97) 1997;499–504.
[2] Hall CWG, Harper RA, Snelling JP. Bullet resistant glazings for automotive vehicles, In: Proceedings of the
International Body Engineering Conference and Exposition. Vol. 3, Warrendale, PA: Publ. SAE International,
1998. p. 332–7.
[3] Grant PV, Cantwell WJ, McKenzie H, Corkhill P. The damage threshold of laminated glass structures. Int
J Impact Eng 1998;21:737–46.
[4] Oda J, Zang MY. Analysis of impact fracture behavior of laminated glass of bilayer type using discrete element
method. Key Eng Mater 1998;145–149:349–54.
[5] Rahul-Kumar P, Jagota A, Bennison SJ, Saigal S. Interfacial failures in a compressive shear strength test of glass/
polymer laminates. Int J Solids Struct 2000;37:7281–305.
[6] Golden JH, Hammant BL, Hazell EA. Effects of molecular weight and strain rate on the flexural properties of
polycarbonate. J Appl Polym Sci 1968;12:557–69.
[7] Bauwens JC. Relation between the compression yield stress and the mechanical loss peak of bisphenol-
A-polycarbonate in the b transition range. J Mater Sci 1972;7:577–84.
[8] Allen G, Morley DCW, Williams T. The impact strength of polycarbonate. J Mater Sci 1973;8:1449–52.
 ARTICLE IN PRESS

50 S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53

[9] Bauwens-Crowet C. The compression yield behaviour of PMMA over a wide range of temperature and strain-
rates. J Mater Sci 1973;8:968–79.
[10] Bauwens-Crowet C, Ots J-M, Bauwens JC. The strain-rate and temperature dependence of polycarbonate in
tension, tensile creep and impact tests. J Mater Sci 1974;9:1197–201.
[11] Sacher E. The impact strength of polycarbonate. J Macromol Sci B: Phys 1974;9:163–7.
[12] Vincent PI. Impact strength and mechanical loss in thermoplastics. Polymer 1974;15:111–6.
[13] Illinger JL, Lewis RW, Barr DB. Effects of adhesive structure on impact resistance and optical properties of
acrylic/polycarbonate laminates. In: Lee L-H, editor. Adhesion Science and Technology 1. New York: Plenum
Press, 1975. p. 217–32.
[14] Parvin M, Williams JG. Ductile-brittle fracture transitions in polycarbonate. Int J Fract 1975;11:963–72.
[15] Ravetti R, Gerberich WW, Hutchinson TE. Toughness, fracture markings, and losses in bisphenol-A
polycarbonate and high strain rate. J Mater Sci 1975;10:1441–8.
[16] Sacher E. Secondary structural motions in polycarbonate 2: Identification of the motions and their relation to
impact strength. J Macromol Sci B: Phys 1975;11:403–10.
[17] Goldsmith W, Katsamanis F. Crack propagation in plastics due to impact. In: Kawata K, Shioiri J, editors. High
velocity deformation of solids. Berlin: Springer, 1978. p. 195–207.
[18] Hillig WB. Response of polycarbonate, poly(methylmethacrylate), and epoxy resins to dynamic contact 1
Loading experimental results. Polym Eng Sci 1985;25:339–47.
[19] Fleck NA, Stronge WJ, Liu JH. High strain-rate shear response of PC and PMMA. Proc R Soc Lond A
1990;429:459–79.
[20] Rietsch F, Bouette B. The compression yield behaviour of PC over a wide range of strain rates and temperatures.
Eur Polym J 1990;26:1071–5.
[21] Sahraoui S, Lataillade JL. Deformation and fracture of PMMA at high rates of loading. J Appl Polym Sci
1994;51:1527–32.
[22] Walley SM, Xing D, Field JE. Mechanical properties of three transparent polymers in compression at a very high
rate of strain. In: Williams JG, Pavan A, editors. Impact and dynamic fracture of polymers and composites.
London: Mechanical Engineering Publications Ltd, 1995. p. 289–303.
[23] Chiou K-C, Hsu H-C, Chang F-C. Precrack hysteresis energy in determining polycarbonate ductile-brittle
transition 4: effect of strain rate. J Appl Polym Sci 1997;65:655–66.
[24] Grellmann W, Lach R. Toughness and relaxation behaviour of polymethylmethacrylate, polystyrene, and
polycarbonate (in German). Angew Makromol Chem 1997;253:27–50.
[25] Wu J, Xiao C, Yee AF, Klug CA, Schaefer J. Controlling molecular mobility and ductile-brittle transitions of
polycarbonate copolymers. J Polym Sci B: Polym Phys 2001;39:1730–40.
[26] Fleck NA, Wright SC, Liu JH, Stronge WJ. Ballistic perforation of polycarbonate sheet and its high strain rate
response. J Phys France 49 Colloq. C3 (DYMAT 88) 1988;145–150.
[27] Wright SC, Fleck NA, Stronge WJ. Ballistic impact of polycarbonate: an experimental investigation. Int J Impact
Eng 1993;13:1–20.
[28] Flocker FW, Dharani LR. Stresses in laminated glass subject to low velocity impact. Eng Struct 1997;19:
851–6.
[29] Flocker FW, Dharani LR. Modeling interply debonding in laminated architectural glass subject to low velocity
impact. Struct Eng Mech 1998;6:485–96.
[30] Bourne NK, Forde LC, Millett JCF, Field JE. Impact and penetration of a borosilicate glass. J Phys IV France 7
Colloq. C3 (EURODYMAT 97) 1997;157–62.
[31] Forde LC. Ballistic impact of rods, Ph.D. thesis, University of Cambridge, 2000.
[32] Kajon G, Monteleone L, Steindler R. Monitored ballistic tests on shockproof sandwich glasses 2. In: Proceedings
of the International Congress on Experimental Mechanics. Society for Experimental Mechanics, Bethel, CT,
2000. p. 822–5.
[33] Kajon G, Monteleone L, Steindler R. Monitored ballistic tests on shockproof sandwich glasses. Exper Tech
2001;25(5):27–31.
[34] Zurimendi JA, Biddlestone F, Hay JN, Haward RN. Physical factors affecting the impact strength of
polycarbonate. J Mater Sci 1982;17:199–203.
 ARTICLE IN PRESS

S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53 51

[35] Golden JH, Hammant BL, Hazell EA. The effect of thermal pretreatment on the strength of polycarbonate.
J Appl Polym Sci 1967;11:1571–9.
[36] Struik LCE. Physical ageing in amorphous polymers and other materials. Amsterdam: Elsevier, 1978.
[37] Bauwens-Crowet C, Bauwens J-C. Annealing of polycarbonate below the glass transition: quantitative
interpretation of the effect on yield stress and differential scanning calorimetry measurements. Polymer
1982;23:1599–604.
[38] Sharko KK, Aniskevich AN, Yanson YO. Effect of physical aging on deformation characteristics of
polycarbonate. Mech Compos Mater 1982;18:628–30.
[39] Bauwens-Crowet C, Bauwens J-C. Effect of thermal history on the tensile yield stress of polycarbonate in the beta
transition range. Polymer 1983;24:921–4.
[40] Hill AJ, Heater KJ, Keller A. The effects of physical aging in polycarbonate. J Polym Sci B: Polym Phys
1990;28:387–405.
[41] Wimberger-Friedl R, de Bruin JG. The very long-term volume recovery of polycarbonate: Is self-retardation
finite? Macromolecules 1996;29:4992–7.
[42] Bauwens-Crowet C. Long term physical ageing of polycarbonate at room temperature: dynamic mechanical
measurements. J Mater Sci 1999;34:1701–9.
[43] Hutchinson JM, Smith S, Horne B, Gourlay GM. Physical aging of polycarbonate: enthalpy relaxation, creep
response, and yielding behavior. Macromolecules 1999;32:5046–61.
[44] Orreindy S, Rincon! GA. Physical aging of bisphenol-A polycarbonate. J Appl Polym Sci 1999;74:1646–8.
[45] Araki O, Horie M, Masuda T. Physical aging of polycarbonate investigated by dynamic viscoelasticity. J Polym
Sci B: Polym Phys 2001;39:337–41.
[46] Wyzgoski MG, Yeh GSY. Origin of impact strength in polycarbonate 1: effect of crystallization and residual
solvent. Int J Polym Mater 1974;3:133–48.
[47] Wyzgoski MG, Yeh GSY. Origin of impact strength in polycarbonate 2: effect of thermal treatments. Int J Polym
Mater 1974;3:149–63.
[48] Wyzgoski MG, Yeh GSY. Origin of impact strength in polycarbonate 3: effect of liquids and orientation.
J Macromol Sci B: Phys 1974;10:441–76.
[49] Ohishi F, Nakamura S, Koyama D, Minabe K, Fujisawa Y, Tsuruga Y. Factors affecting dynamic durability of
polycarbonate under bending, torsional, and impacting fatigue and in cavitation erosion and dynamic solvent
cracking conditions. J Appl Polym Sci 1976;20:79–92.
[50] Gardner RJ, Martin JR. Humid aging of plastics: effect of molecular weight on mechanical properties and
fracture morphology of polycarbonate. J Appl Polym Sci 1979;24:1269–80.
[51] Factor A, Chu ML. The role of oxygen in the photo-ageing of bisphenol-A polycarbonate. Polym Degrad Stab
1980;2:203–23.
[52] Clark DT, Munro HS. Surface aspects of the photodegradation of bisphenol-A polycarbonate in oxygen and
nitrogen atmospheres as revealed by ESCA. Polym Degrad Stab 1982;4:441–58.
[53] Heath JBR, Gould RW. Degradation of the bird impact resistance of polycarbonate. SAMPE Quart 1982;
14(1):20–9.
[54] Sherman ES, Ram A, Kenig S. Tensile failure of weathered polycarbonate. Polym Eng Sci 1982;22:457–65.
[55] Clark DT, Munro HS. Surface aspects of the photodegradation of bisphenol-A polycarbonate in oxygen and air
as a function of relative humidity as revealed by ESCA. Polym Degrad Stab 1983;5:227–36.
[56] Hurley CJ, Zahavi J, Schmitt Jr GF. UV radiation exposure effects on the rain erosion resistance of coated
monolithic polycarbonate transparencies. Wear 1983;87:9–28.
[57] Clark DT, Munro HS. Surface and bulk aspects of the natural and artificial photoageing of bisphenol-A
polycarbonate as revealed by ESCA and difference UV spectroscopy. Polym Degrad Stab 1984;8:195–211.
[58] Munro HS, Allaker RS. Wavelength dependence of the surface photo-oxidation of bisphenol-A polycarbonate.
Polym Degrad Stab 1985;11:349–58.
[59] Andrady AL, Searle ND, Crewdson LFE. Wavelength sensitivity of unstabilized and UV stabilized
polycarbonate to solar simulated radiation. Polym Degrad Stab 1992;35:235–47.
[60] Adams MR, Garton A. Surface modification of bisphenol-A polycarbonate by far-UV radiation 2: in air. Polym
Degrad Stab 1993;42:145–51.
 ARTICLE IN PRESS

52 S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53

[61] Torikai A, Mitsuoka T, Fueki K. Wavelength sensitivity of the photoinduced reaction in polycarbonate. J Polym
Sci A: Polym Chem 1993;31:2785–8.
[62] Kim A, Bosynak CP, Chudnovsky A. Effects of weathering, scale, and rate of loading on polycarbonate fracture
toughness. J Appl Polym Sci 1994;51:1841–8.
[63] Ghorbel I, Thominette F, Spiteri P, Verdu J. Hydrolytic aging of polycarbonate 1: physical aspects. J Appl Polym
Sci 1995;55:163–72.
[64] Ghorbel I, Akele N, Thominette F, Spiteri P, Verdu J. Hydrolytic aging of polycarbonate 2: hydrolysis kinetics,
effect of static stresses. J Appl Polym Sci 1995;55:173–80.
[65] Risch BG, Wilkes GL. Effects of physical aging and carbon dioxide absorption in bisphenol-A polycarbonate.
J Appl Polym Sci 1995;56:1511–8.
[66] Rivaton A. Recent advances in bisphenol-A polycarbonate photodegradation. Polym Degrad Stab 1995;49:
163–79.
[67] Turton TJ, White JR. Degradation depth profiles and fracture of UV exposed polycarbonate. Plast Rubb
Compos 2001;30:175–82.
[68] Tjandraatmadja GF, Burn LS, Jollands MC. Evaluation of commercial polycarbonate optical properties after
QUV-A radiation: the role of humidity in photodegradation. Polym Degrad Stab 2002;78:435–48.
[69] Blaga A, Yasmasaki RS. Surface microcracking induced by weathering of polycarbonate sheet. J Mater Sci
1976;11:1513–20.
[70] Petrie SP, Dibenedetto AT, Miltz J. The effects of stress cracking on the fracture toughness of polycarbonate.
Polym Eng Sci 1980;20:385–92.
[71] Park KJ, Chin EY. Effect of diamond-like carbon thin film deposition on the resistance of polycarbonate to
radiation-induced degradation. Polym Degrad Stab 2000;68:93–6.
[72] Walley SM, Field JE, Pope PH, Safford NA. The rapid deformation behaviour of various polymers. J Phys III
France 1991;1:1889–925.
[73] Hutchings IM, Winter RE. A simple small-bore laboratory gas-gun. J Phys E: Sci Instrum 1974;8:84–6.
[74] Bowden FP, Brunton JH. The deformation of solids by liquid impact at supersonic speeds. Proc R Soc Lond A
1961;263:433–50.
[75] Bowden FP, Field JE. The brittle fracture of solids by liquid impact, by solid impact and by shock. Proc R Soc
Lond A 1964;282:331–52.
[76] Field JE. Stress waves, deformation and fracture caused by liquid impact. Phil Trans R Soc Lond A 1966;260:
86–93.
[77] Lesser MB, Field JE. Liquid impact. Ann Rev Fluid Mech 1983;15:97–122.
[78] Lesser M. Thirty years of liquid impact research: A tutorial review. Wear 1995;186–187:28–34.
[79] Field JE. Liquid impact: Theory experiment applications. Wear 1999;233–235:1–12.
[80] Blair PW. The liquid impact behaviour of composites and some infra-red transparent materials. Ph.D. thesis,
University of Cambridge, 1981.
[81] Walley SM, Field JE. The erosion and deformation of polyethylene by solid-particle impact. Phil Trans R Soc
Lond A 1987;321:277–303.
[82] Walley SM, Field JE, Safford NA. A comparison of the high strain rate behaviour in compression of polymers at
300 K and 100 K. J Phys IV France 1 Colloq. C3 (DYMAT 91) 1991;185–90.
[83] Bernatskii AD, Rabinovich AL. Strain in certain crosslinked polymers. Polym Sci USSR 1964;6:1166–75.
[84] Durelli AJ, Hall J, Stern F. Brittle coating 1: Stresses and strains in brittle coatings, In: Kobayashi AS, editor.
Handbook on experimental mechanics (Second Revised Edition), New York: VCH Publishers Inc., 1993. p. 491–515.
[85] Steer P, Rietsch F, Lataillade J-L, Marchand A, El Bounia N-E. Viscoplasticite! dynamique de polycarbonate aux
grandes vitesses de d!eformation. Relations structure-propri!et!es. J Phys France 46 Colloq. C5 (DYMAT 85)
1985;415–23.
[86] Cayssials F, Lataillade JL. Effect of the secondary transition on the behavior of epoxy adhesive joints at high
rates of loading. J Adhesion 1996;58:281–98.
[87] Vidal B, Lataillade J-L, Collombet F, Bacon C. M!ethodologie pour le dimensionnement a" l’impact d’un
empilement lin!eaire e! lastique et visco!elastique. J Phys IV France 7 Colloq. C3 (EURODYMAT 97) (1997)
779–84.
 ARTICLE IN PRESS

S.M. Walley et al. / International Journal of Impact Engineering 30 (2004) 31–53 53

[88] Mart!ınez MA, Chocron IS, Rodr!ıguez J, S!anchez G!alvez V, Sastre LA. Confined compression of elastic adhesives
at high rates of strain. Int J Adhesion Adhes 1998;18:375–83.
[89] Yu XX, Crocombe AD, Richardson G. Material modelling for rate-dependent adhesives. Int J Adhesion Adhes
2001;21:197–210.
[90] Hopkinson B. The effects of momentary stresses in metals. Proc R Soc Lond 1905;74:498–506.
[91] Hopkinson B. The pressure of a blow. In: Ewing JA, Larmor J, editors. The Scientific Papers of Bertram
Hopkinson. Cambridge: Cambridge University Press, 1921. p. 423–37.
[92] Rinehart JS. Scabbing of metals under explosive attack: multiple scabbing. J Appl Phys 1952;23:1229–33.
[93] Rinehart JS, Pearson J. Behavior of metals under impulsive loads. Cleveland, OH: American Society for Metals,
1954.
[94] Rinehart JS. Fracturing by spalling. Wear 1964;7:315–29.
[95] Kolsky H, Shi YY. Fractures produced by stress pulses in glass-like solids. Proc Phys Soc Lond 1958;72:447–53.
[96] Kolsky H. Fractures produced by stress waves. In: Averbach BL, Felbeck DK, Hahn GT, Thomas DA, editors.
Fracture. New York: Wiley, 1959. p. 281–96.
[97] Aprahamian R, Evenson DA, Mixson JS, Jacoby JL. Holographic study of propagating transverse waves in
plates. Exper Mech 1971;11:357–62.
[98] F.allstron K-E, Gustavsson H, Molin N-E, W(ahlin A. Transient bending waves in plates studied by hologram
interferometry. Exper Mech 1989;29:378–87.
[99] Billon N, Maurin I, Germain Y. Characterization of ductile polymers at high strain rates: Experimental and
theoretical considerations. In: Emri I, Knauss WG, editors. Proceedings of the First International Conference on
Mechanics of Time-Dependent Materials. Bethel, CT: Society of Experimental Mechanics, 1995. pp. 293–8.
[100] Field JE. Dynamic fracture: its study and application. Contemp Phys 1971;12:1–31.
[101] Miller GF, Pursey H. On the partition of energy between elastic waves in a semi-infinite solid. Proc R Soc Lond A
1955;233:55–69.
[102] Woods RD. Screening of surface waves in soils. ASCE: J Soil Mech Foundations Div 1968;94:951–79.
[103] Graff KF. Wave Motion in Elastic Solids. Oxford: Clarendon Press, 1975. p. 356.
[104] Drucker DC. Approximate calculations of spall and cratering in high speed impact. In: Kawata K, Shioiri J,
editors. High velocity deformation of solids. Berlin: Springer, 1978. p. 214–27.
[105] Seward CR, Field JE, Coad EJ. Liquid impact erosion of bulk diamond, diamond composites and diamond
coatings. J Hard Mater 1994;5:49–62.