Acta Materialia 52 (2004) 1423–1438

www.actamat-journals.com

The effect of fibre fractures in the bridging zone of fatigue cracked

Ti–6Al–4V/SiC fibre composites
R. Sinclair a, M. Preuss a, E. Maire b, J.-Y. Buffiere b, P. Bowen c, P.J. Withers a,*

a
Manchester Materials Science Centre, University of Manchester, Grosvenor St, Manchester M1 7HS, UK
b
GEMPPM, Insa de Lyon, 20 Av. A. Einstein, 69621 Villeurbanne, France
c
School of Metallurgy and Materials, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK

Received 13 June 2003; received in revised form 20 November 2003; accepted 24 November 2003

Abstract

The aim of the paper is to study the partitioning of stress between bridging and broken fibres and the nearby matrix in the region
around a fatigue crack in the matrix of a Ti–6Al–4V/SCS–6 SiC fibre composite. This was achieved by using synchrotron X-ray
radiation to perform a combination of high spatial resolution tomographic imaging and strain mapping. The average elastic fibre
strain for each ply was mapped with distance from the crack, ply by ply. Two samples were examined; one in which there were no
broken fibres and one in which some fibres in ply 1 had broken. The contributions of broken and bridging fibres were separated
using a double peak fitting routine. The interfacial stress variation and the extent of interfacial debonding were deduced from the
fibre strain profiles. Contrary to most micromechanical models the interfacial frictional sliding stress was not found to be constant
along the fibre length, but to decrease approximately linearly towards the crack plane. Upon unloading the fibres were found to
undergo reverse sliding at the interface.
Ó 2003 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Synchrotron diffraction; Tomography; Metal matrix composites; Fatigue; Fibre bridging

1. Introduction matrix composites [6] but the divergent beam and long
data acquisition times for small irradiated volumes limit
Reinforcing titanium for application in aeroengines the spatial resolution achievable in practice. Synchro-
has been found to improve high temperature stiffness, tron radiation has proved to be a useful tool for ob-
strength, and fatigue life [1–3]. The factors that control taining this information for Ti/SiC composites [7–12].
fatigue behaviour in fibre composites include crack The intense, high energy, highly parallel X-rays avail-
bridging by fibres, interfacial debonding and sliding and able at the European synchrotron radiation facility
fibre breakage. In order to better understand these (ESRF) in Grenoble, France, are ideal for the materials
processes, it would be helpful to have a method of under study. High X-ray energies are required to give
measuring stress or strain distributions inside the fibre good penetration for relatively large atomic number
and matrix regions as the sample is loaded. Whilst SiC is materials such as titanium (37 % transmission of 40 keV
Raman active, Raman spectroscopy is unsuitable for photons over 1 mm of Ti), and the highly parallel beam
measuring the required distribution of stresses in metal provides excellent spatial resolution. These characteris-
matrix fibre composites because the matrix would need tics mean that it is possible to combine high resolution
to be optically transparent to allow collection of scat- strain measurements with 3D tomographic images of
tering from the embedded fibres [4,5]. Neutron beams unparalleled spatial definition. As a result a much more
provide a means of stress measurement for titanium detailed picture of fibre fracture can be established than
was possible in earlier pioneering tomographic studies
*
Corresponding author. Tel.: +44-161-200-8872; fax: +44-161-200-
[13].
3586. Our previous work [10] was the first to map the fibre
E-mail address: [email protected] (P.J. Withers). strains in the vicinity of bridging SiC fibres embedded in

1359-6454/$30.00 Ó 2003 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2003.11.024
1424 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438

a fatigue cracked Ti–6Al–4V matrix in which none of techniques were used to detect the presence and location
the fibres were broken in the bridging region. The of any fibre failures. A MISTRAS 2001 system with two
present work examines the effect of fibre fractures in the PAC Nano 30 AE sensors glued to the end of each test-
bridging zone. piece was used. The fatigue test was continued for
47,500 cycles for S1 and 20,900 cycles for S2. AE indi-
cated that, as intended, S1 contained no fibre breaks and
2. Experimental methods S2 contained a number of fibre breaks close to the crack
plane. In both cases the matrix fatigue crack was grown
2.1. Materials and fatigue cracking treatment perpendicular to the longitudinal fibre direction.

The composite material comprised 6 plies of SCS–6 2.2. Tomography

SiC fibres. The C coated SiC fibres were 140 lm in di-
ameter with a carbon core, and were coated with Ti– The fatigued test-pieces were cut down to produce
6Al–4V to control interfibre spacing. The composite was dumb-bell shaped specimens using electro discharge
unidirectional containing 35% volume fraction of fibres. machining with the gauge area centred on the crack.
Both sides of the composite were clad with a 0.15 mm Before strain mapping, both were examined by X-ray
layer of monolithic Ti–6Al–4V prior to consolidation by tomography to establish the exact positions of the fibres
high temperature HIPing. The final thickness of the and the extent of the matrix cracks. The tomography
composite was about 1.5 mm. Pieces measuring studies were carried out on the ID19 beam-line at the
50
5
1.5 mm were cut out from the composite plate. ESRF. A monochromatic, highly parallel beam of energy
Spot welds were applied to one of the clad surfaces of 30 keV was used, the CCD camera gave a resolution of 2
the test-pieces to act as crack initiators. lm, the sample-camera distance was 60 mm and the ro-
In general, higher fatigue loads produce shorter tational step size was 0.15° giving 1200 projections over
component life. Jeng et al. [14] summarised several re- 180°. The projections were reconstructed using Fourier
sults to conclude that for a number of types of SiC/Ti transformation algorithms resulting in a resolution or
composites in fatigue there can be three modes of fail- ÔvoxelÕ size of 2.8 lm in the reconstructed data.
ure: the fibres fail first at high loads and low cycles
(region I); the matrix fails first at low loads and high 2.3. Strain mapping
cycles (region III); at intermediate loads and cycles,
there will be failure in the fibres, the interfaces and the The test-pieces were mounted in turn onto an x, y, z
matrix (region II). We used this knowledge to create two stage in experimental hutch 2 of ID11 at the ESRF. The
specimens, with and without broken fibres, by fatiguing X-ray energy chosen for the experiment was 51 keV.
to different stress amplitudes. Using focusing Laue crystal optics, the beam height
Fatigue testing was carried out using an Instron ser- (z-parallel to the fibre axis) was set to 7.5 lm; the width
vo-hydraulic testing machine. The fatigue tests were (x-parallel to the plies) was restricted to 100 lm by using
carried out using three-point bending geometry at room incident slits. Because no slit was used on the diffracted
temperature. Two multi-ply Ti–6Al–4V/SiC fibre speci- beam the gauge volume in the y dimension extended
mens were fatigue cracked. For both test samples the through the thickness of the specimen so as to sample
applied stress ratio, R, was 0.1 (where R is the ratio all of the fibres in a row. A Frelon CCD camera
between the minimum and maximum stress applied (2000
2000 pixels) was used to collect images of the
during a fatigue cycle) and the frequency was 5 Hz. diffraction rings. The camera length was 97 cm and it
In three-point bending the load across the specimen was displaced vertically from the optical axis so as to
cross-section is non-uniform. The front (the spot welded collect the top section of the diffraction rings. Changes
face) is in tension whereas the back is in compression. to this part of the diffraction ring correspond to changes
The stress values are therefore quoted for the front ply in axial strain (ez ).
of fibres, that is, 0.24 mm from the front, ignoring the The diffraction rings were integrated azimuthally over
contribution of residual stress in the fibres (known to be 5° about a vertical line through the centre of each ring.
compressive [10]). For one test-piece (designated S1) the This procedure (known as ÔcakingÕ) ensures better sta-
maximum stress in the front ply of fibres during a stress tistics than a simple radial line scan of intensity. With
cycle was 1000 MPa and for the other test-piece (S2) the the focussed beam, count times of only 5 s per point
maximum stress in the front ply of fibres was 1400 MPa. were required to obtain sufficient diffracted intensity in
The values were chosen to avoid fibre breaks in S1 the experiment. Since the CCD camera recorded a large
(region III) and to introduce some fibre breaks in S2 segment of the diffraction profile, it was possible to
(region II). Fatigue crack growth was monitored by analyse the (0 0 2), (1 0 2), (1 1 0) (1 1 2) and (2 0 1) a-ti-
applying the direct current potential drop technique. tanium (hexagonal phase) reflections, the (2 0 0) b-tita-
During the fatigue loading acoustic emission (AE) nium (cubic phase) reflection, and the (1 0 8) SiC
 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438 1425

reflection. Other reflections of a-titanium and SiC a computer controlled microtensile rig equipped with a
overlapped with each other making the analysis of their 2 kN load cell. Note that the loading was carried out in a
peak positions difficult. simple tensile manner rather than under three-point
The 2h diffraction peak positions were obtained from bending. However since the samples were cracked on
the caked diffraction data by Gaussian peak fitting. one side, some bending moment was unavoidable.
Elastic strain was derived using the Bragg equation to Measurements on S1 were taken at 1000 and 0 MPa,
convert 2h data into d-spacing. An additional sample and on S2 at 1400, 1000, and 0 MPa. The extra loading
was prepared to serve as a stress-free lattice spacing (d0 ) step at 1000 MPa for S2 provided a comparison between
specimen by electro-polishing one end of an off-cut of the two samples using the same load. These loads were
the composite material in order to remove the matrix calculated for the composite, ignoring residual stress,
material and leave the exposed fibres stress-free. but taking into account the area of the crack measured
The elastic lattice strain e was then calculated using using tomography. At each load level diffraction data
the following equation: were collected as the sample stage was translated in 10
lm steps along the fibre direction ðzÞ, so as to scan a
d  d0
e¼ : ð1Þ region 2 mm above and below the crack plane. The
d0 (1 0 8) SiC peak was used to calculate the fibre strain
Given that it is very difficult to produce a represen- variation averaged over the fibres within each ply for S1.
tative stress-free unreinforced matrix sample, the matrix The strain in the matrix was also derived using the (1 0 2)
strain was assumed to be zero at the free surfaces of the Ti peak from the same diffraction patterns.
crack in the unloaded test-pieces. Because the specimens
contained 6 plies with 5–6 fibres in each ply, the fact that
the sampling gauge spanned the whole thickness means 3. Results and discussion
that each measured point represents the strain averaged
over the regions within all the fibres through the ply 3.1. Tomography
(thickness) within the 7.5
100 lm window centred on
the co-ordinates x, z. Each test-piece was scanned across The reconstructed tomographs shown in Figs. 1 and 2
the beam during loading which was applied in situ using were recorded at zero applied load. In Fig. 1(a) the

Fig. 1. Reconstructed tomographic images of S1. (a) 3D tomograph in which the location of ply 1 marked; the fibres are 0.14 mm in diameter. (b) The
same volume rotated so as to look along the fibre direction, displayed in transparent mode. The dashed line represents the size of the illuminating
beam. (c) A y–z slice across ply 1 as marked on (a). (d) An x–z slice corresponding to the position marked on (a).
1426 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438

Fig. 2. Reconstructed tomographic images of S2. (a) Three-dimensional tomograph in which the location of ply 1 marked; the fibres are 0.14 mm in
diameter. (b) The same volume rotated so as to look along the fibre direction, displayed in transparent mode. The dashed line represents the size of
the illuminating beam. (c) A z–y slice through ply 1 as marked on (a), with an enlarged section showing the first two fibres on the left in the main
picture. (d) An x–z slice at the position marked on (a).

location of the spot weld is visible near the centre of the notch in [10] which gave a crack of equal extent across
cracked (front) side. It is clear that two cracks have the specimen width. The matrix–fibre interfaces in
initiated from points near the top and bottom of the Fig. 1(b) are visible as rings. It can be seen that each ring
spot weld. It can also be seen that several fibres were is broken at the sides and this is due to a centring error
bisected in the process of electro discharge machining of artefact during reconstruction. It arises when the infor-
the sides. If the remaining parts of the fibre were smaller mation that the reconstruction uses about the centre of
than half the original diameter they tended to fall out rotation of the specimen is incorrect by a small fraction
during cutting. of the pixel size of the detector. The phenomenon is
When viewing tomographs it can be helpful to vary described fully in [15].
the ÔopacityÕ of regions of differing attenuation using the In the 2D slices shown in Fig. 1(c) and (d), dark
volume viewing software. In Fig. 1(b) areas of low areas correspond to low attenuation and light areas to
density such as the crack were set to opaque (white) and high attenuation, so that the matrix is shown as light
higher density areas such as fibres and matrix were set to regions, the fibres are intermediate in colour and the
transparent. Interfaces are also in high contrast due to carbon cores of the fibres are darker. In Fig. 1(c) the
Fresnel fringing in the original radiographs. In Fig. 1(b) two cracks differ in z-position by approximately 0.14
the extent of and overlap between the two cracks that mm. In Fig. 1(d) overlap of the cracks can be seen
were visible on the specimen surface in Fig. 1(a) can be between the specimen surface and ply 1. Beyond ply 1
seen. The right-hand crack extends to x ¼ 0:51 mm in the two cracks have amalgamated to form a single
the centre of the specimen and 0.45 mm at the side of the crack. The tomographic resolution was not fine enough
specimen. The left-hand crack extends to 0.45 mm near to resolve interfacial debonding. In Fig. 1(d) the dis-
the centre of the specimen and 0.30 mm at the side. The tance from the surface of the specimen to the centre of
differences in crack extent near the centre and sides of the first row of fibres is approximately 0.24 mm. The
the specimen are due to the initiation at a spot weld, positions of each fibre relative to one corner of the
creating a curved crack front, as opposed to a linear specimen areas were carefully recorded, in order to
 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438 1427

facilitate finding the plies during the strain scanning and unloading propping the crack open. A shadow can
experiments. be seen in the section of fibre in the second ply (second
That S2 contained a single undulating crack is clear from the left in Fig. 2(d)) but it was known from evi-
from Fig. 2. On the upper face of the reconstructed dence discussed above that this fibre was not broken.
volume several rings can be seen faintly due to a re- The third fibre from the left has a mottled appearance
construction artefact, described in [15]. For CCD to- due to the ring artefact.
mography it can arise because not all the pixels on the In summary, care must be taken when interpreting
CCD are equally sensitive [16]. Fig. 2(b) shows the ex- sections from tomographic reconstructions, but by
tent of cracking. Between some of the fibres the crack comparing Figs. 1 and 2 it can be seen that S1 contains a
front appears to be curved suggesting that the crack is pair of parallel overlapping matrix cracks that have
longest between fibres. The greatest extent of the crack is coalesced upon passing ply 1 and that S2 contains a
0.73 mm, located near to the y-position of the spot weld. single undulating crack. Three fibres in ply 1 of S2 are
The extent of the crack was least at the right side of the fractured in the vicinity of the matrix fatigue crack,
specimen, at 0.67 mm. while no fibre breaks were found in S1. It can be seen
In the transparent rendering in Fig. 2(b) the three that the matrix cracks have reached the second ply of
fibres on the left of ply 1 (bottom) have a lighter ap- fibres in S1 and the third ply of fibres in S2. The crack
pearance than the other fibres, similar in appearance to lengths range from 0.30 to 0.47 mm in S1 and from 0.67
the matrix crack. The sixth (partial) fibre in ply 1 and to 0.73 mm in S2.
the first (also partial) fibre in ply 2 are similarly affected.
It was concluded from this and other views that these 3.2. Measurement of unstrained lattice parameters
fibres contained breaks near the matrix crack plane, and
that the others were unbroken. The morphology of the When deriving strain from lattice parameter mea-
fibre fractures can be seen for the three fibres on the left surements, it is important to know the strain free lattice
of ply 1 in Fig. 2(c) lying about 1 fibre diameter above spacing d0 . Representative examples of diffraction peaks
the plane of the matrix crack. A second break is also collected from the exposed fibre end and the embedded
visible in the first fibre from the left at the matrix crack fibre end of the d0 specimen are shown in Fig. 3. It can
plane. The (partial) fibre on the right of the figure also be seen that the only SiC reflection that is not over-
has a crack. At first glance it appears that all the fibres lapped by titanium reflections is the (1 0 8) SiC peak,
are also broken in the crack plane due to ÔshadowsÕ which was subsequently used for all strain measure-
across each fibre. However, these shadows are in fact ments of the SiC fibres. It was found that the fibre peaks
artefacts arising from the large matrix crack surround- corresponding to planes normal to the fibre direction
ing the fibres. The shadows are of much lower contrast were shifted slightly to higher 2h values when embedded
than the fibre breaks. In addition, the absence of fibre in the matrix compared to the positions when exposed.
fractures in the fourth and fifth fibres from the left This is due to thermal residual strain in the composite,
concurs with the evidence from Fig. 2(b). It should also
be noted that shadows in the fibres were visible in
Fig. 1(c) at the matrix crack planes, even though it was 12
(102)
known from the acoustic emission data that no fibres composite
(110)
were broken in S1. The three left most fibres in ply 1 are 10 exposed fibres
Intensity (arbitrary units)

the closest to the position of the spot weld. The fact that (101)
breaks in these fibres occurred at the same z distance 8 (103)
from the crack plane suggests that the breaks may be co-
6
ordinated. During the fatigue process, a fibre break
(200)
occurring in one of these fibres may have caused a local
4
stress concentration thus overloading the neighbouring (100) (108)
fibres at the same z position. The other fibre fractures (112)
2 (201)
occurred in fibres that were damaged by dissection
during cutting.
0
Fractured fibre 3 (in Fig. 2(c)) can also be seen to the 4 6 8 10 12
left in Fig. 2(d) from which a difference in crack opening 2θ (º)
between the crack-tip and near the specimen surface is
also evident. Since the tomograph was recorded at zero Fig. 3. ÔCakedÕ synchrotron X-ray diffraction spectra for SiC fibres
exposed by electropolishing away the matrix, and the Ti/SiC com-
load this must arise either from, a permanent set caused
posite, with the beam centred on a ply of fibres. In both cases the
by matrix plasticity, local relief of tensile matrix thermal samples were oriented such that the scattering planes were normal to
stresses, or from the difference between forward and the fibre direction (the composite profile has been displaced upwards
reverse fibre–matrix interface sliding between loading by 0.5 a.u. so it can be distinguished).
1428 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438

which is relieved when the matrix has been etched away. sition may depress the stress at the peak. Third, Poisson
The axial residual fibre strain was measured to be effects mean that the radial clamping stress may be lower
)0.15% and compares well with the )0.15% strain re- near the crack where the fibre stresses are highest, re-
ported in [6]. It is equivalent to a fibre stress of around ducing the sliding friction stress. Finally, it is possible
650 MPa (calculated in [6] using a fibre stiffness of 415 that the interface may be worn by repeated sliding
GPa). To determine a strain free reference for the ma- during the fatigue loading process lowering the interfa-
trix, far field measurements of the matrix and fibre peak cial shear stress. The first possible explanation could be
positions in the fatigued specimens in the unloaded examined by measuring the strain in the individual
condition were made. The fibre residual strain was fibres. The second reason is a likely contributor since it
found to be )0.15%, as found in the d0 specimen. The was observed from the load controller that the load re-
matrix residual strain was then calculated on the basis of laxed by about 5% between the initial application of the
stress balance between the phases. The matrix residual load and the measurement of the plies.
strain was deduced to be +0.16% (using a volume frac- In Fig. 4(a) the fibre strain profile is somewhat
tion of fibres of 0.25, measured from the tomograph asymmetric about z ¼ 0 mm. For sample S1 the z ¼ 0
including the cladding layer) and this was used to pro- mm position has been defined to correspond to the
vide a strain free diffraction peak position for the ma- lower of the two matrix cracks depicted in Fig. 1(d) and
trix. Using values for the coefficients of thermal the upper crack is therefore located at z ¼ þ0:14 mm.
expansion given in [6] and the method outlined in [17], This may be the cause some of the rounding of the peak
the equivalent stress free temperature to which this in the strain profile referred to above, although it is
corresponds is calculated as 650 °C which is 50 °C lower unlikely to be the sole reason for the rounded strain
than that recorded in [6]. profile. There is also asymmetry in the nature of the
slope changes further from the peak strain. For exam-
3.3. Fibre strain measurements at Kmax ple, there is a discernible change in slope at z ¼ 0:7 mm
whereas no such slope change is visible on the positive
Fig. 4(a) shows the lattice strain variations in the side. This may be because interfacial debonding occurs
region of the first ply for S1. Note that the fibre and to different extents in the different fibres in the ply,
matrix strain profiles are complementary in that where leading to a smearing out of the strain profiles.
the matrix strain is small in the vicinity of the crack, the By comparison with S1 the profiles in Fig. 4(b) show
bridging fibre strain is correspondingly large. a much smaller peak fibre strain in the vicinity of the
While in both cases the peak in fibre strain is found at crack even though the applied stress is higher. The to-
the crack plane, the peak is rather diffuse in contrast to mography results indicate that ply 1 of S2 contains three
those predicted by finite element modelling of bonded broken and two intact fibres. Therefore the diffraction
and debonded interfaces as outlined in [10], or by shear- data collected from this ply are an average of the con-
lag theory for bridging fibres [18–21]. There are at least tributions of these fibres. In the region of maximum
four reasons why this might be. First, the crack and/or fibre strain for S2 two diffraction peaks were discernable
debonding position along the length of the embedded (Fig. 5). This suggests that the contributions from the
fibres may vary in each of the six fibres in the ply. Sec- highly stressed unbroken fibres and the lowly stressed
ond, stress relaxation during the period of data acqui- broken ones are separable. To separate the contribu-

0. 7 0. 7

0. 6 0. 6
matrix
0. 5 0. 5
matrix
strain (%)

0. 4 0. 4
strain (%)

0. 3 0. 3

0. 2 fibres 0. 2

0. 1 0. 1 fibres

0. 0 0. 0
-0.1 -0.1
-2 -1 0 1 2 -2 -1 0 1 2
(a) z (mm) (b) z (mm)

Fig. 4. Ply 1 axial lattice strain variations along z for (1 0 8) fibre and (1 0 2) matrix (a) for S1 stressed to 1000 MPa and (b) S2 stressed to 1400 MPa.
 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438 1429

1. 0 1. 0

0. 8 0. 8
intact
intensity (arbitrary units)

strain (%)
0. 6
0. 6

0. 4
0. 4 average

0. 2
0. 2
broken
0. 0
0. 0
8. 8 8. 9 9. 0 9. 1 -2 -1 0 1 2
(a) 2θ (º) (b) axial distance from crack plane (mm)

Fig. 5. (a) A typical (1 0 8) SiC diffraction peak profile acquired close to the matrix crack plane (z ¼ 0) for ply 1 of S2 at 1400 MPa; the error for each
point is smaller than the size of the markers and (b) the resulting strains deduced from the broad (intact) and narrow (broken fibres) component
Gaussian peaks.

tions from the broken and intact fibres, the double peaks contours in Fig. 6(a) delineate rounded triangular re-
were fitted to two Gaussian peaks and the results are gions, enclosing the crack tip whereas the 0.2% contour
shown in Fig. 5(b). has two lobes. These arise because the constituent ply
The difference between broken and intact fibres is
clear; the intact fibres have a peak strain of 1% strain.
This corresponds well with elastic strains up to 1.4%
ply ply
recorded during the progressive loading of a single fibre
composite sample prior to fracture [8,22]. This is con- 1 2 3 4 5 6 1 2 3 4 5
2. 0
sistent with the mean fracture strength of 5100  665
MPa recorded for single SCS-6 fibres with 1 mm gauge
lengths [23] in contrast to lower mean strengths recorded

0 .2
1. 5
0 .1

for longer gauge lengths, e.g. 4521  497 for 20 mm

gauge length. It is clear from Fig. 5(b) that, as one would 0. 9
expect, in the loaded condition, the maximum strain in 1. 0
3
0.
0. 8
the intact fibres in ply 1 is much higher than for the first 0 .2
2

ply in S1, both because there are fewer intact fibres in 0. 7

0 .1
0.

5
0.
0. 5
the ply and the higher load to support. The broken fibres 0. 6
0 .6

collectively exhibit a plateau region at zero strain

0 .3

0. 5
z (mm)

around 0.2 mm with compressive strain either side. 0. 0

0 .2
0 .4

0 .1

The reason for this plateau may be the differences in the 0. 4

z positions of the breaks along the individual fibres of 0. 3
0. 2
6
0.

the ply as discussed above (see Fig. 2(c)). -0.5

3
0.

0 .3

0. 2
The individual ply profiles have been assembled to
0 .2
4
0.

0. 1
0 .1

produce area maps to facilitate the comparison between

-1.0
S1 and S2. The bending component caused by the 0
0 .2

opening of the crack is clear from the variation in fibre

strains across the crack plane. The fibre stress in the strain (%)
-1.5
0 .2
0 .1

cracked region is further concentrated by a shedding of

load from matrix to fibre occurring by shear lag near the
matrix crack. In fact the fibres in plies 1 and 2 of S1 have -2.0
a strain greater than 0.4% near the crack plane (z ¼ 0) 0. 5 1. 0 0. 5 1. 0
compared to less than 0.15% at large z. For S1 the ply 1 (a) x (mm) (b) x (mm)
fibres exhibit a strain concentration of 3.7 (peak strain:
Fig. 6. Axial (1 0 8) elastic fibre strain contour plots for intact fibres for
far field strain) and ply 2 fibres 3.5 the values for the (a) S1 stressed at 1000 MPa (b) S2 stressed at 1400 MPa. The data are
fibres ahead of the crack being 2.8 (ply 3), 1.7 (ply 4), rotated through 90° compared to the line profiles shown in Fig. 4 with
1.03 (ply 5) and 0.48 (ply 6). The 0.4% and 0.3% strain the crack positions marked by arrows.
1430 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438

profiles for the plies ahead of the crack tip show two
maxima, one either side of the crack plane, with the
spacing between the maxima widening with distance
ahead of the crack. If the 0.2% contour is taken to be the
extent of the region in which the fibres is loaded sig-
nificantly above the far field case, then the maximum
dimensions of the region are 2 mm in the z direction and
0.5 mm in the x direction.
Fig. 6(b) shows that the extent and intensity of the
overloaded region is significantly greater in S2 than in
S1. The strain in the intact fibres in ply 1 reaches almost
1.0% at z ¼ 0 mm compared to about 0.15% far from
the crack. This maximum value is approaching the 1.4%
fibre fracture strain measured in fracture [8,22]. The
elastic strain concentration in this ply is 6.0 in the crack
plane. The strain concentration in ply 2 is 4.3 and in ply
3 it is 3.7 (the matrix crack extended to ply 3). Ahead of
the crack tip the stress concentrations are 3.1 (ply 4) and
1.3 (ply 5). Except for the very high stresses (strains) in
the few intact fibres in ply 1 of S2 these values are
comparable with those deduced from near surface fibres
by Davidson [24]. He noted that because of the larger
slip lengths associated with those nearest the notch the
bridging stresses fall-off less sharply than would other-
wise be the case. The notched SiC fibre/Ti–6Al–4V
specimens in [10] also showed a triangular region of fibre
overload in the locality of a fatigue crack. This was
shown to coincide with a region of fibre–matrix sliding Fig. 7. Axial Ti (1 0 2) elastic matrix strain contour plots for (a) S1 at
evidenced by changes in the interfacial shear stress data. 1000 MPa and (b) S2 at 1400 MPa. The cracks are shown as black
arrows.

3.4. Matrix strain

1. 0
The average axial matrix strain in the vicinity of the
fibres was measured using the (1 0 2) reflection because
0. 8 3b
of its high multiplicity. The resulting data are shown in
Fig. 7 as maps, which facilitate a comparison with the
SiC data, and as line scans for ply 3 in S2 in Fig. 8,
strain (%)

0. 6
showing the effect of the variation in crack extent across
the specimen width on the matrix strain. The Ti data are
0. 4
significantly noisier than the SiC data. This is due to the
3a
larger grain size, which results in the X-ray beam sam-
pling a smaller number of grains within the gauge 0. 2
volume.
The axial matrix strain field for S1 exhibits a trian-
0. 0
gular shaped region in which load is shed as the crack is
-2 -1 0 1 2
approached (Fig. 7(a)). This region of load shedding is
z (mm)
complementary to the region of load concentration in
Fig. 6(a). The extent of this region is partly determined Fig. 8. Ti (1 0 2) elastic axial strain as a function of distance from the
by the geometry of the crack, but also by the extent to crack plane at S2 ply 3 (x ¼ 0:64 mm) at 1400 MPa load, separated
which the fibre matrix interfacial strength determines the into intact (3a) and cracked (3b) data sets.
rate at which the matrix stress builds up to the far-field
value with distance from the crack plane. The matrix and these were separated out to give two datasets.
reloading zone for S2 (Fig. 7(b)) is larger than for S1. These probably result from the fact that the crack front
The data summarised in Fig. 8 provide a more de- position varies from 0.67 to 0.73 mm across the width
tailed picture of the matrix in the vicinity of ply 3 of S2. of ply 3 and thus varies across the 100 lm-wide gauge
In this case two matrix diffraction peaks were observed volume sampled for ply 3. A large stress concentration
 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438 1431

is seen in 3b consistent with stress concentration at the


sinh½nðLe  zÞ=r
crack tip, while the minimum in 3a falls to approxi- rf ðzÞ ¼ rf ðð1  mÞLe Þ ; ð3Þ
sinh½nmLe =r
mately zero strain consistent with an open crack; as a
result these have been labelled 3a (crack wake) and 3b 2sf ð1  mÞLe
with rf ðð1  mÞLe Þ ¼ rf ð0Þ 
(crack tip). r
Em
and n2 ¼ ;
3.5. Interfacial shear stress and debonding Ef ð1 þ mm Þ lnðR=rÞ
where mm is the matrix PoissonÕs ratio. The fibre is as-
Interfacial shear strength is a controlling factor in sumed to have a circular cross section with radius r, and
determining fatigue performance. In some systems the the mean fibre spacing is R.
shear stress is solely frictional and the maximum shear The stress in the sliding region is given by
stress is simply the value at which sliding begins. In
others an interfacial ÔbondÕ must first be broken before 2sf z
rf ðzÞ ¼ rf ð0Þ  : ð4Þ
frictional sliding can occur. If the interfacial shear r
strength is too high, the fibres in a composite tend to
break co-operatively under the steep stress concentration Under this model, the fibre stress falls because the
near the crack to form a flat fatigue crack, too low and shear stress is constant until the point of transition be-
the stress transfer to the fibre is too inefficient and the tween the bonded and sliding segments. At this point the
interface cracks run very large distances into the com- interfacial shear stress may rise abruptly before the fibre
posite. It is therefore important to measure the interfacial and interface stresses decay with distance into the bon-
shear response for the composite materials under study. ded region of the composite. Upon increasing the load
It can be shown by balancing stress across a section at the free end (z ¼ 0), because the interfacial stress in
of the fibre dz that the variation in interfacial shear stress the bonded zone is limited to smax , further increases in
(sÞ is directly proportional to the gradient of fibre strain interfacial shear stress would cause the transition point
with distance as shown in Eq. (2) [21]: between frictional and perfect bonding to move along
the interface, increasing the debonded length. The fibre
Ef  r de stresses for the ply 1 fibres in samples S1 and S2 are
s¼  ; ð2Þ
2 dz shown in Fig. 9(a) and (b) from which the transitions
where Ef is fibre YoungÕs modulus, r is fibre radius and between bonded and debonded regions have been de-
de=dz is the rate of change of fibre strain along the fibre. duced and best fit solutions using the above equations
This relation means that we can derive the interfacial derived. It is clear that the profiles are broadly in ac-
shear stress variation directly from the fibre strain cordance with such a model with Em , Ef , mm , mm given by
profiles. 115, 345 GPa, 0.34 and 0.2, respectively. The fibre
In our case it is appropriate to consider two interfa- spacing parameter R was measured from Fig. 1 to be
cial conditions; bonded regions for which no sliding has 0.12 mm for a fibre in ply 2.
occurred and sliding regions which have experienced The corresponding interfacial shear stress profiles for
sliding at some time. In order for a ÔbondedÕ region to the ply 1 fibres were calculated from the corresponding
begin sliding a shear strength smax must be exceeded, fibre strain data in Fig. 9(a) and (b) using Eq. (2). Since
while regions which have already slipped may resume the strain data contained some scatter, a degree of
sliding when a frictional sliding stress sf is exceeded. It is smoothing was necessary. This was performed using a
a matter of debate whether smax is greater than or simply cubic spline method. It should be noted that the abso-
equal to sf , while almost all the models in the literature lute values of s are plotted, that is, the direction of
assume that sf is constant. In this case, the rate of sliding is ignored. Considering sample S1, distant from
change of stress in the fibre is linear and expressed by the crack plane (z < 0:7 mm and >1.1 mm) the re-
Eq. (2). Piggott [21] proposed a model to describe the sponse is consistent with a ÔbondedÕ interface (Eq. (3)).
distribution of fibre strain and interfacial shear stress in Nearer to the crack plane, the data are consistent with
bonded fibres. Despite its simplicity it has been shown to regions of approximately constant interfacial stress
fit data well when compared with point-to-point strain characteristic of a transition from bonding to debonding
distributions measured using various techniques such as at )0.7 mm [8]. In this case the data (circles) suggests
Raman and fluorescence spectroscopy [18–20] and syn- that smax is in the order of 80 MPa for )ve z, but a
chrotron strain mapping [8,22]. It has been modified to distinct value is not so clear for +ve z. The debonded
apply to several different geometries of embedded and plateau region is consistent with an approximately
partially embedded fibres. Consider an embedded length constant frictional sliding stress (sf ) of the order of 50
Le of which a fraction m remains bonded then the fibre MPa. The best fit modelled interfacial shear stress curve
stress rf in the bonded section (z > mLe ) varies accord- is broadly similar to the data obtained from the cubic
ing to [17] spline method. However the smax value obtained is
1432 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438

0. 5 120

100

Interfacial shear stress (MPa)

0. 4

80
0. 3
strain (%)

60
0. 2
40
model
0. 1 strain data 20

0. 0 0
-2 -1 0 1 2 -2 -1 0 1 2
(a) z (mm) (c) z (mm)

1. 0 450
400

Interfacial shear stress (MPa)

0. 8 350
300
0. 6
strain (%)

250
200
0. 4 strain data
model 1 150
model 2 100
0. 2
average
50
0. 0 0
-2 -1 0 1 2 -2 -1 0 1 2
(b) z (mm) (d) z (mm)

Fig. 9. Axial (a) fibre elastic strain and (c) interfacial shear stress profiles for ply 1 of S1 at 1000 MPa and (b) and (d) the corresponding variations for
the intact fibres in ply 1 of S2 at 1400 MPa. For the axial strains the circles represent the strain data, for the shear stresses the circles represent the
stress calculated from the smoothed strain data, in all cases the continuous lines represent the best-fit shear-lag model.

higher (110 MPa cf. 80 MPa from the spline method plitude consistent with earlier work (at constant R ratio)
(circles)). This arises because while the spline method [25,26].
smoothes out noise in the data it also smoothes out From the fits to the data, it would seem that there is
changes in gradient that arise from real physical limited evidence for a debonding threshold smax which is
phenomena. greater than the sliding frictional stress. If a constant
The shear stress data for ply 1 of S2 are summarised friction stress is appropriate then a value of around 50–
in Fig. 9(d). The need to separate the contributions of 70 MPa would be most appropriate. When interpreting
unbroken and broken fibres has led to more scatter than the preceding shear stress results it must be borne in
for S1. In this case the bonded regions extend to within mind that the signal is an average over all the fibres in a
z ¼ 1:25 and 1.1/1.3 mm of the matrix crack. The ply which may act to wash out the threshold peak if the
double maxima may arise from the two intact fibres in debonding lengths are different for each fibre in the ply.
the ply having different extents of debonding on the +ve This is especially likely for the 6 fibres in ply 1 of S1. Of
side of the crack. The effect of superimposing two deb- course it may be simply that there is no debonding
onding events at different locations would be to reduce threshold stress smax . In static tests [8,22] a much higher
the magnitude of both, as is observed. However, caution friction stress was observed (200 MPa) and although a
must be exercised in interpreting these results due to the threshold stress in s ðsmax ) was introduced into the
high degree of scatter. The levels of sf in the frictional modelling at the transition between debonding and
regions vary between 50 and 70 MPa. This is similar to bonded regions, the gauge sampling length along the
what was observed in the S1 suggesting that frictional fibre (50 lm) was too large to verify whether one ex-
sliding stress is not strongly dependent on fatigue am- isted in practice. Piggott [21] on the other hand has
 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438 1433

predicted the absence of a smax above sf for metal matrix

composites. In our case it might arise from the threshold
stress required to unseat the rough interface from its
original location [27], or the drop between smax and sf
might simply represent the difference between the stres-
ses required to slide pristine (200 MPa) and worn (50
MPa) interfaces.
Although a shear lag model coupled with a constant
friction stress in the debonded zone captures the essence
of the fibre strain variations for the first plies, it is clear
from Fig. 9(c) and (d) that the agreement in the sliding
regime is poor, with significantly lower frictional stresses
nearer the crack.
In Fig. 10, the variation in interfacial shear stress in
the loaded condition is shown for all the scans (approx. Fig. 11. The interfacial shear stress profile in 3D relief and as stress
one scan per ply) for the two samples. From this it is contours for a 4 mm deep by 3.5 mm wide Ti–6Al–4V/SCS–6 SiC
clear that the sliding lengths are clearly delineated by composite containing fibres bridging a crack of length 1.64 mm grown
pffiffiffiffi
wedge-shaped region, but that contrary to the simple at Kmax (32 MPa m) (1.25 kN) for 6,020,000 cycles [10]. The scan was
model, the shear stress within these regions is not measured using a 70 lm (axial) by 200 lm (transverse) gauge window.
The crack had progressed to the fifth ribbon in the upper figure.
constant. Instead the shear stress falls almost linearly
from the boundary with the bonded region to almost is possible for the interfacial shear stress to decrease
zero near the crack plane (note that the crack opening towards the crack even if there is no wear of the in-
is very small on this plot). Furthermore, the gradient terfaces due to the reduction in the radial clamping
appears to be steeper for fibres nearer to the crack-tip forces caused by the Poisson contraction under the
(at least for Fig. 10(b)). This effect had been noted axial tension. This effect can be examined analytically.
before in a similar experiment involving a longer crack The radial clamping stress (rDT f;radial ) is normally around
and ascribed to interface wear [10]. As a consequence 300 MPa due to thermal residual stresses [6,28], but
a poor fit was achieved with an FE model using a this is reduced by Em mf rf =Ef [17]. Assuming a tradi-
Coulomb friction law. That data is replotted here as a tional Colulomb friction (constant l) the fibre stress
3D relief plot in Fig. 11. The linear gradation in shear would be expected to vary as
stress towards the crack is very clear with the rise in


Ef DT 2lEm mf
shear stress lowest nearest the notch. In this case the rf ðzÞ ¼ r 1  exp z
maximum shear stress values at the boundary with the Em mf f;radial rEf


bonded region are approximately 70–80 MPa com- 2lEm mf
þ rf ðz ¼ 0Þ exp z ð5Þ
pared with about 100 MPa for S2 (neglecting ply 1). It rEf

260
240
220
200

150 150 160

180
)

125
x ( mm

125 140
120
100 100
75 75 100
50 50
25 25 80
0 0
60
5 6

5 6 7

40
1.0

1.0
x (mm )

20
3 4
ply

ply

20
2 3 4

20 20 20
0
0.5

0.5

20 40
40
1 2

ISS (MPa)

2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5 -2.0 2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5 -2.0
(a) Distance from crack (mm) (b) Distance from crack (mm)

Fig. 10. Interfacial shear stress contour plots for (a) S1 at 1000 MPa and (b) the intact fibres of S2 at 1400 MPa in 3D relief (above) to emphasise the
shear stress variation within the debonded region, and as contours (below) for quantitative interpretation. The cracks are shown as white arrows. The
graphs were made using the spline-smoothed data. A uniform friction stress would be represented by single triangle shaped plateau at a fixed height.
The cracks had progressed to the second/third and the third/fourth ribbons for samples S1 and S2, respectively.
1434 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438

while the interfacial shear stress varies as measured a decrease in fibre surface roughness within


Em m f
 600 lm of a fibre fracture for a fibre extracted from a
DT
sf ðzÞ ¼ l rf;radial  rf ðzÞ : ð6Þ fatigue sample. Further they have deduced that a
Ef
significant decrease in sliding stress occurs when the
Assuming that the constant sliding strength measured cyclic sliding distance is greater than 1lm; comparison
in fibre fragmentation (for which Poisson effects would be with [22] shows this to be in agreement with our
expected to be less marked and would even increase the sliding lengths. There is evidence in Fig. 10 that in-
radial clamping force) is also representative of fatigue terface sliding occurs up to 400 lm ahead of the
then the sliding coefficient l would be 2/3. The variations current matrix crack position.
in shear stress predicted on this basis are shown in Fig. 12. The sliding (debonding) lengths for all the plies are
It is clear from this that Poisson effects are not negligible summarised in Fig. 13. It should be remembered that
halving the shear stress at the crack for the S1 ply 1 fibres the debonding lengths are not dependent on the model
and having an even more marked effect on the few intact used to interpret the curves, but were deduced simply
fibres in ply 1 of S2. Nevertheless, the decrease in sliding from the point of transition from a constant or rising
stress observed in Figs. 10 and 11 are larger than can be interfacial strength to a falling curve (shear lag style)
explained solely in terms of the Poisson effects. This typical of a well bonded interface with distance from
conclusion is corroborated by the fact that the FE model the crack plane. The sliding lengths measured here are
[10] which was based upon a Coulomb law underesti- comparable to those measured in previous measure-
mated the breadth of the sliding length. ments made using synchrotron diffraction [10]. Previ-
As a result the interfacial shear stresses are con- ously, few direct measurements of the sliding length
siderably below those expected on the basis of the have been possible although Davidson did etch viewing
frictional sliding stress recorded for a pristine fibre. windows in the matrix to deduce whether the fibres
This effect has been well documented previously and is were sliding or not [24]. These direct measurements as
almost certainly because fatigue causes interface deg- well as coating damage revealed by etching [24,34],
radation due to the forward and reverse sliding of the have also indicated a V shaped sliding region in the
interface [8,29,30]. Those regions nearest the crack will wake of the crack with a maximum sliding length of
have been exposed to the greatest number and am- about 1 mm on either side for the fibre nearest the
plitude of forward and reverse sliding cycles. Marshall notch (although Davidson had 6–10 fibres bridging the
et al. [31] discussing the effect of repeated load cycles crack). He also found a threshold for sliding ahead of
on the interface in some SiC/Ti3 Al composites, found the crack tip. By comparison, critical fibre lengths of
that the load required to cause sliding of a fully 1–1.5mm have been measured for SCS-6 fibres in
debonded fibre decreased with increased number of Ti–6Al–4V in 1991 [35]. Gradual improvements in the
sliding cycles in common with others [32]. This effect fibre strength since 1991 would mean the critical fibre
has been included in models previously and a linear length today would be greater and should be of the
degradation does seem the most reasonable in the order of twice the maximum debonding lengths re-
light of our work [11,30]. Tanaka et al. [33] have corded here.

Fig. 12. The interfacial shear stress on the basis of l ¼ 200 MPa for each ply, based on the fibre bridging stresses for each ply at z ¼ 0. The results for
S1 are shown with continuous lines; those for S2 with dashed lines. The ply 1 S1 line shows a particularly large Poisson effect due to the very high
stresses acting on the two intact fibres.
 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438 1435

ply ply Fig. 14(b), the M-shape first appears when the load is
1 2 3 4 5 6 1 2 3 4 5 decreased from 1400 to 1000 MPa, and then widens
further as the load is completely removed. Reverse
1. 0 sliding was considered by Bennett and Young [20] in
the case of crack bridging, modifying an analysis by
Marshall [31] for fibre pull-out. It occurs because during
0. 5 loading fibres in the debonded regions slide out of the
matrix to accommodate the crack opening. This gener-
z (mm)

0. 0
ates a misfit near the crack plane during unloading.
Refs. [11,20,31] consider the case of loading, unloading
and reloading of the composites. During unloading re-
-0.5 verse interfacial slip occurs in the frictional region ini-
tiating at the crack plane causing sliding opposite to that
which occurred during loading. Since the gradient in
-1.0
Fig. 14 appears to be similar to that for forward sliding
it seems that the reverse sliding stress is approximately
0. 5 1. 0 0. 5 1. 0 the same as the forward one. As unloading progresses
(a) x (mm) (b) x (mm)
the extent of reverse sliding increases, but conventional
debonded frictional hysteresis means that it does not completely
reverse sliding (1000MPa) return to the original fibre matrix position, i.e. the for-
reverse sliding (0MPa) ward sliding debonded length is greater than the reverse
one [22]. Fig. 13 summarises the reverse sliding lengths
Fig. 13. Sliding lengths for each ply, deduced from fibre strain profiles. measured for each ply for the 2 samples. Note also that
S2 ply 1 was subject to difficulty in separating the contributions of the reverse sliding is such that the stress in fibres in ply 1
broken and intact fibres at low loads and the reverse sliding lengths does not reverse into compression at z ¼ 0 after un-
upon unloading are not shown.
loading, but instead the fibres are approximately un-
strained near the crack plane. For fibres subject to
3.6. Partial unloading compressive residual strain and which tend to hold the
crack open, one might expect to find some compression
For single fibre fragmentation tests of Ti/SiC fibre near the crack plane. Perhaps a reason for the lack of
systems [8,22] it was found that reverse sliding occurs on compressive strain here is due to the permanent bend in
unloading giving rise to characteristic ÔW Õ elastic strain the specimen. Further from the crack plane the original
profiles for fibre fragments. In the present case of matrix compressive thermal fibre residual strains ()0.15%) are
cracking a characteristic ÔMÕ axial fibre strain profile retained after unloading indicating that there has been
develops upon unloading (Fig. 14). This is a direct result negligible generalised yielding of the matrix during the
of progressive reverse sliding. It initiates from near the cycle characteristic of high cycle fatigue conditions.
matrix crack plane and works progressively backwards During the next fatigue cycle interfacial slip occurs
into the interior. For sample S2, as is clear from again in the forward direction during reloading, starting

0. 6 0. 8

0. 5 0. 7
0. 6 1400 MPa
0. 4
0. 5
strain (%)

0. 3
1000 MPa 0. 4
strain (%)

0. 2 0. 3 1000 MPa

0. 1 0. 2
0. 1
0. 0
0 MPa
0. 0
-0.1 0 MPa
-0.1
-0.2 -0.2
-2 -1 0 1 2 -2 -1 0 1 2
(a) z (mm) (b) z (mm)

Fig. 14. Fibre strain profiles during unloading of (a) S1 ply 1 and (b) S2 ply 2.
1436 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438

once more from the crack plane. Our data are reminis- strength for systems such as ours where several inter-
cent of those shown in [20]. The more jagged appearance faces fail simultaneously (between inner and outer C
of the strain profiles there may have been due to sticking layers, outer C layer and reaction zone and reaction
points during loading; these would have been smeared zone and matrix) [35]. It is interesting to subdivide the
out by averaging over all the fibres in a ply. In Fig. 14 interface measurements on Ti–6Al–4V in the literature and
the responses have rounded transitions between the to compare these to the values measured here (Table 1). It
forward and reverse sliding regions. This may be due to would appear that methods such as pull-out and fatigue
a transition region between forward and reverse sliding cracking which hold the fibre in greater tension than the
regions in which the interfacial shear stress is too low to matrix record lower interfacial stresses than those which
cause sliding in either direction (below sf ). Other factors tend to increase the radial clamping stress, such as push-
may also contribute to, and widen, the rounded transi- out and fibre fragmentation. It is clear that the interfacial
tion including differences in sliding lengths for different shear stress drops by at least a half and often very much
fibres in the ply (averaging), and interface wear due to more when it is fatigued. Clearly the extent to which this
repeated forward and reverse sliding. occurs is a function of the severity of the fretting fatigue.
On the whole, the maximum interface sliding stresses
reported for this system measured directly from syn-
4. Discussion chrotron X-ray measurements for both the pristine and
worn fibre–matrix interfaces (200 and 50–20 MPa,
Previously researchers have deduced the interfacial respectively) are about two times higher than those
shear stress on the basis of indirect measurements with measured elsewhere by mechanical evaluation, e.g. from
the aid of micromechanical models. Nearly all the 60–80 MPa (pristine) to 10 MPa (worn) [25] measured by
models have assumed a constant interfacial shear stress fatiguing bridging fibre crack test-pieces cut out from
all the way along the fibre. On this basis very large centre notch samples. This may be a real effect due to the
variations in interfacial shear stress have been found for examination of different systems and manufacturing
SCS-6 fibres in Ti matrices. Some of this is almost cer- routes, or it may be due to the fact that the latter methods
tainly due to differences in processing conditions and smooth out the variations in interfacial shear stress along
different matrices, whilst different testing geometries also the fibre to give an overall average value. One advantage
contribute to the scatter. In addition the condition of the of the current technique is that it can measure the inter-
interface is crucial since pristine and worn interfaces face strength as a function of position from the matrix
appear to have very different strengths. Finally, it had crack in situ under tension without the need for a mi-
been postulated that the shear stress should vary along cromechanical model. It should also be remembered that
the sliding fibre length as a consequence of Poisson ef- our synchrotron measurements are averaged over all the
fects [17], interface roughness, fibre–fibre interactions fibres in a ply. However, in recent synchrotron X-ray
[27]. Wear had been expected to affect the interfacial measurements we have been able to map the strain in

Table 1
Interfacial shear strengths and sliding stresses
Method smax (MPa) Sliding stress (MPa) Notes Reference
Strain mapping (fatigue) 200 0–70 2–3 bridged fibres This work
Strain mapping (fatigue) 0–90 3 or 4 bridged fibres [10]
C crack growth rates 9 4 mm cracks (200–444 MPa) Ti–6Al–4V [26]
COD measurements 7–3 Ti–6Al–4V [37]
COD measurements 245 Ti–6Al–4V [24]
Crack growth rates 1.7 TIMETAL 21S [24]
Fragmentation 180 Ti–6Al–4V [38]
Push-out 160 88 Ti–6Al–4V [35]
Push-out 160 Ti–6Al–4V [39]
Strain mapping (fragmentation) 200 [8]
Push-out 180 TIMETAL 21S [40]
Pull-out multifibre compliance 60–80 Ti–6Al–4V/Sigma fibre pristine fibres [25]
Pull-out multifibre compliance 10 Ti–6Al–4V/Sigma fibre worn fibres [25]
Pull-out multifibre 55–60 Ti–15V–3Cr–3Al–3Sn, pristine [30]
Pull-out 5 1 Ti, spray deposit [41]
Centre notch fatigue 15–35 Ti–15V–3Cr–3Al–3Sn, fatigued [29]
Push-out 90 Ti–6Al–4V, pristine [42]
Push-out 20 Ti–6Al–4V, fatigued [42]
Push-out 160 Ti–6Al–4V [35]
Push-out 160 Ti–6Al–4V [35]
 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438 1437

individual fibres within the composite [36]. Those mea- of the crack. No single representative interfacial shear
surements confirm the shear stress values measured here stress has been observed. Instead the frictional sliding
as well as identifying additional features such as stick-slip shear stress appears to decrease linearly towards the
regions. These effects are smeared out in the current crack plane the gradient of which appears to reduce with
measurements and may lead to a slight decrease in the increasing sliding length (i.e. nearer to the notch). The
shear stress averaged over a ply. effect of the reduction in the thermal clamping force
It is not clear the extent to which the fact that the caused by the Poisson contraction of the fibre is only
assumed constant interfacial shear stress is not observed partly responsible for this decrease in frictional sliding
in practice affects the interfacial shear strengths mea- stress.
sured by other techniques. Given the approximately Reverse sliding has been observed over the debonded
linear variation an average value of the shear stress may regions of the fibre matrix interface during unloading.
indeed capture the important micromechanics of the At the present time, the results are averages over all the
situation with respect to fatigue crack growth rate ex- fibres in the line of sight of the X-ray beam. In the near
periments. Our work would seem to indicate that the future recently acquired results will be published in
average interfacial shear stress is higher for fibres nearer which individual fibre strains in the multiple ply com-
the crack-tip. posite specimen have been measured to show effects such
Finally, our current results provide only limited evi- as slip stick not smeared out by the averaging process
dence for a threshold interfacial stress smax which must [36]. Finite element models as a means of interpret-
be exceeded to initiate debonding. However, given that ing the strain variations in fatigue loading are being
the sliding stress for a pristine fibre interface is at least developed.
five times higher than on a worn one, even if there is no
chemical bond which must be broken to initiate sliding
one would expect the continually forward and reverse
sliding regime to be significantly below the highest stress Acknowledgements
attained in the perfectly bonded region.
The authors would like to thank Dr. Ann Terry
(ESRF) and Dr. Larry Margulies (Risoe) for experi-
5. Conclusions mental assistance, Rolls Royce for donating the mate-
rials used and Dr. Tim Doel and Julien Laffargue
Load partitioning has been observed directly and in (Birmingham University) for pre-fatiguing the speci-
detail in Ti–6Al–4V/SiC fibre composite test-pieces mens. P.J.W. would also like to thank the referee whose
containing fatigue cracks with and without broken fi- helpful comments prompted a re-examination of the
bres. The stress distributions in bridging fibres and in interfacial shear stress variations. The work is part of a
the matrix ahead of the crack have been accurately project funded by EPSRC. P.J.W. is grateful for a Royal
measured. Diffraction profiles for the fibres in ply 1 of Society-Wolfson Award.
S2 comprised a double peak near the crack plane (see
Fig. 4), indicative of two states of fibre strain. When
deconvolved the data can be interpreted on the basis of
two fibre populations; one lowly stressed (broken) and References
one highly stressed (intact). This was not the case for
other plies containing only intact fibres. Tomography [1] Rao KTV, Siu SC, Ritchie RO. Failure mechanisms in SiC-fiber
has shown that ply 1 of S2 contained two unbroken fi- reinforced 6061 aluminium alloy composites under monotonic and
bres and three broken ones. The fibre bridging stress is cyclic loading. Met Trans A 1993;24:721–33.
[2] Cotterill PJ, Bowen P. Fatigue crack growth in a Ti-15-3 matrix/
largest for the fibres nearest the notch, with the stress in SCS-6 fibre material matrix composite under tension–tension
the two intact fibres in ply 1 of S2 very near to the mean loading. Mater Sci Technol 1996;12:523–9.
fibre breaking strength, as one might expect given the [3] Barney C, Ibbotson A, Bowen P. Experimental characterization of
failure of the other fibres in the ply. The fibre bridging fibre failure and its influence on crack growth resistance in fibre
stress falls fairly gradually from the ply near the notch to reinforced titanium metal matrix composites. Mater Sci Technol
1998;14:658–68.
that nearest the crack-tip because of the counteracting [4] Ward Y, Young RJ, Shatwell RA. A microstructural study of
effect of shorter debonding lengths. Some debonding silicon carbide fibres through the use of Raman microscopy. J
occurs ahead of the crack tip. Mater Sci 2001;36:55–66.
A triangular sliding zone above and below the crack [5] Yang X, Young RJ. Model ceramic fiber-reinforced glass
plane can be seen in the fibres and the matrix strains in composites – residual thermal-stresses. Composites 1994;25:488–
93.
which the interface shear stress is substantially below [6] Withers PJ, Clarke AP. A neutron diffraction study of load
that of a pristine sliding interface (200 MPa). The sliding partitioning in continuous Ti/SiC composites. Acta Mater
length near the notch is approximately 1 mm either side 1998;46:6585–98.
1438 R. Sinclair et al. / Acta Materialia 52 (2004) 1423–1438

[7] Maire E, Owen RA, Buffiere J-Y, Withers PJ. A synchrotron reinforced with unidirectional SiC fibers. In: Bucinell RB, editor.
X-ray study of a Ti/SiCf composite during in situ straining. Acta Composite materials: fatigue & fracture, 1330. ASTM STP; 1998.
Mater 2001;49:153–63. p. 122–44.
[8] Preuss M, Withers PJ, Maire E, Buffiere J-Y. SiC single fibre full- [27] Cox BN. Life prediction for bridged fiber cracks. In: Johnson WS,
fragmentation during straining in a Ti–6Al–4V matrix studied by Larsen JM, Cox BN, editors. Life prediction methodology for Ti
synchrotron X-rays. Acta Mater 2002;50:3175–90. matrix composites, 1253. ASTM STP; 1996. p. 552–72.
[9] Preuss M, Rauchs G, Withers PJ, Maire E, Buffiere J-Y. [28] Wright PK, Nimmer R, Smith G, Sensmeier M, Brun M. The
Interfacial shear strength of Ti/SiC fibre composites measured influence of the interface on mechanical behaviour of Ti–6Al–
by synchrotron strain measurement. Composites A 4V/SCS-6 composites. In: Lin RY et al., editors. Interfaces
2002;S33:1381–5. in metal–ceramic composites. Anaheim, CA: TMS; 1990.
[10] Preuss M, Rauchs G, Doel T, Steuwer A, Bowen P, Withers PJ. p. 559–81.
Measurements of fibre bridging during fatigue crack growth in Ti/ [29] Walls DP, Bao G, Zok FW. Mode I fatigue cracking in a fiber
SiC fibre metal matrix composites. Acta Mater 2003;51:1045–57. reinforced metal matrix composite. Acta Metall Mater
[11] Rauchs G, Thomason PF, Withers PJ. Finite element modelling of 1993;41:2061–71.
frictional bridging during fatigue crack growth in fibre-reinforced [30] Walls DP, Zok FW. Interfacial fatigue in a fiber reinforced metal
metal matrix composites. Comp Mater Sci 2002;25:166–73. matrix composite. Acta Metall Mater 1994;42:2675–81.
[12] MacDonald S, Preuss M, Maire E, Buffiere J-Y, Mummery PM, [31] Marshall DB, Shaw MC, Morris WL. Measurement of interfacial
Withers PJ. X-ray tomographic imaging of Ti/SiC composites. J debonding and sliding resistance in fibre reinforced intermetallics.
Microsc 2003;209:102–12. Acta Metall Mater 1992;40(3):443–54.
[13] Stock SR. X-ray microtomography of materials. Int Mater Rev [32] Zheng D, Ghonem H. High temperature/high frequency fatigue
1999;44:141–64. crack growth in Ti metal matrix composites. In: Johnson WS,
[14] Jeng SM, Alassoeur P, Yang J-M. Fracture mechanisms of fiber- Larsen JM, Cox BN, editors. Life prediction methodology for Ti
reinforced titanium alloy matrix composites, Part IV: low cycle matrix composites, 1253. ASTM STP; 1996. p. 137–63.
fatigue. Mater Sci Eng 1991;A148:67–77. [33] Tanaka Y, Kagawa Y, Liu Y-F, Masuda C. Interface damage
[15] Schneberk DJ, Azevedo SG, Martz HE, Skeate MF. Sources mechanism during high temperature fatigue test in SiC fibre-
of error in industrial tomographic reconstructions. Mater Eval reinforced alloy matrix composite. Mater Sci Eng A 2001;314:
1990;48:609–17. 110–7.
[16] Davis GR, Elliott JC. X-ray microtomography scanner using [34] Larsen JM, Jira JR, John R, Ashbaugh NE. Crack bridging effects
time-delay integration for elimination of ring artefacts in the in notch fatigue of SCS-6/Ti Metal 21S composite laminates. In:
reconstructed image. Nucl Instr Meth A 1997;394:157–62. Johnson WS, Larsen JM, Cox BN, editors. Life prediction
[17] Clyne TW, Withers PJ. An introduction to metal matrix methodology for Ti matrix composites, 1253. ASTM STP; 1996.
composites. Cambridge Solid State Series. Cambridge: Cambridge p. 114–36.
University Press; 1993. p. 509. [Also in Chinese]. [35] Yang JM, Jeng SM, Yang CJ. Fracture mechanisms of fibre-
[18] Heppenstall-Butler M, Bannister DJ, Young RJ. A study of reinforced titanium alloy matrix composites, Part I: interfacial
transcrystalline polypropylene/single-aramid-fibre pull-out behav- behaviour. Mater Sci Eng 1991;138:155–67.
iour using Raman spectroscopy. Composites A: Appl [36] Sinclair R, Tan KTW, Preuss M, Withers PJ. Imaging of elastic
1996;S27:833–8. strain and damage fibre by fibre in a Ti/SiC fibre composite, 2003
[19] Bannister DJ, Andrews MC, Cervenka AJ, Young RJ. Analysis of (submitted).
the single-fiber pull-out test by means of Raman-spectroscopy. 2. [37] Hutson A, John R, Jira JR. Effect of temperature on fiber/matrix
Micromechanics of deformation for an aramid/epoxy system. interface sliding stress in SCS-6/TI METAL 21S. Scripta Mater
Compos Sci Technol 1995;53:411–21. 1999;40:529–35.
[20] Bennett JA, Young RJ. The effect of fibre–matrix adhesion upon [38] Le Petitcorps Y, Pailler R, Naslain R. The fibre/matrix interfacial
crack bridging in fibre reinforced composites. Composites A-Appl shear strength in titanium alloy matrix composites reinforced by
1998;S29:1071–81. SiC or B CVD filaments. Comp Sci Tech 1989;35:207–14.
[21] Piggott MR. Load bearing fibre composites. Oxford: Pergamon; [39] Majumdar BS,Warrier SG,Miracle DB, In: Briant CL,Carter CB,
1980. Hall EL, editors. Interface Engineering for Optim, Properties, vol.
[22] Rauchs G, Preuss M, Withers PJ. Micromechanical analysis of 485. 1977, p. 185.
internal stress development during single-fibre fragmentation [40] Blatt D, Jira JR, Larsen JM. Time dependent degradation of fiber/
testing of Ti/SiCf . Acta Mater 2002;50:2477–90. matrix interfaces under fatigue of SCS-6/Ti metal 21S. Scripta
[23] Bowen P, Lafargue J. SCS-6 fibre strengths. University of Metall Mater 1995;33(6):939–44.
Birmingham; 2003. [41] Kieschke RR, Clyne TW. Control over interfacial bond strength
[24] Davidson DL. The micromechanics of fatigue crack growth at in Ti/SiC fibrous composites. In: Liaw PK, Gungor MN, editors.
25C in Ti–6Al–4V reinforced with SCS-6 fibers. Met Trans A Fundamental relationships between microstructures and mechan-
1992;23:865–79. ical properties of metal–matrix composites. Indianapolis: TMS;
[25] Connell SJ, Zok FW. Measurement of cyclic bridging law in a Ti 1990. p. 325–40.
matrix composite and its application to simulating crack growth. [42] Warren PD, Mackin TJ, Evans AG. Design, analysis and
Acta Metall Mater 1997;45(12):5203–11. application of an improved push-through test for the measure-
[26] John R, Jira JR, Larsen JM. Effect of stress and geometry on ment of interface properties in composites. Acta Metall Mater
fatigue crack growth perpendicular to fibers in Ti–6Al–4V 1992;40(6):1243–9.