Design and characterisation of high performance pseudo ductile all carbon/epoxy unidirectional hybrid composites.

A variety of thin-ply pseudo-ductile unidirectional interlayer hybrid composite materials comprising high modulus and high strength thin carbon fibre/epoxy prepregs was investigated. The central high modulus carbon plies fragmented and delaminated stably from the outer high strength carbon layers under uniaxial tensile loading in the hybrid materials. These pseudo-ductility mechanisms resulted in favourable, metal-like stress-strain responses featuring pseudo-yielding, a stress plateau and further rise in stress before final failure. The high initial elastic modulus of up to 240 GPa, the early warning and the wide safety margin between damage initiation and final failure make the new hybrids advantageous for safety-critical applications where ultimate performance and low density are key design drivers. A hybrid effect with an increase in the failure strain of the high modulus carbon material was highlighted for very thin plies.


Introduction
Carbon fibre/epoxy composites offer high stiffness and low density resulting in some of the highest achievable specific elastic moduli and tensile strengths among structural materials. These desirable properties make them suitable for applications such as aero-structures, spacecraft, motorsports, high specification sports equipment, where ultimate mechanical performance and lowest possible weight are crucial. Besides their high strength, carbon/epoxy composites generally exhibit unfavourable sudden and brittle failure without sufficient warning and residual integrity, which currently limits their use in many high-volume and safety-critical applications such as mass-produced automotive and construction, therefore improvements are required. The sudden failure characteristic of fibre reinforced composites is usually compensated for by conservative design limits, which hinders component manufacturers from fully exploiting their excellent mechanical properties. The possibility of creating pseudo-ductile carbon fibre reinforced composites assuring safe, progressive failure mechanisms similar to metals' yielding and strain hardening with detectable warning and a wide margin between damage initiation and final failure is therefore of high interest.
The most obvious approach to add ductility to fibre reinforced composites is to replace their intrinsically brittle constituents (i.e. glass, carbon fibres and thermosetting polymer resins) with new, more ductile materials, but the choice of suitable materials is very limited. The focus has been on fibre development as the properties of high performance composites are usually fibre-dominated. Although there are promising new materials such as nanotube fibres [1], the development to make them suitable for structural applications is extremely challenging and their verification and commercialisation is a long process. Low diameter stainless steel fibres with tensile failure strains well beyond 10% were also investigated recently as a new reinforcement for ductile composites with various matrix materials by Allaer et al. [2] and Callens et al. [3]- [5]. Excellent ductility was reported although the density of the obtained composites was at least twice as high as that of carbon/epoxy, which renders them less suitable for lightweight applications.
Architecture modification of laminated composites made of traditional constituents to generate additional strain using different ductility mechanisms is an alternative approach offering improvements in a shorter time scale. The additional strain can be realised e.g. by realignment of off-axis fibres and shearing of the matrix [6], [7] or from excess length due to out of plane waviness [8]. Interface modification on the fibre [9], [10] and on the ply level [11] as well as designed discontinuities [12]- [14] are also suitable for delaying fracture and generating stressstrain non-linearity through controlled damage before final failure.
Hybridisation of commercially available fibres is an established approach to increase the initial modulus of the high strain component and potentially introduce a gradual failure, although it usually results in an unfavourable major load drop when the low strain constituent breaks. Our intention is to use various high stiffness (low failure strain) and high strain (lower stiffness) carbon fibre/epoxy prepregs in thin format to address this issue and fully exploit the benefits of hybridisation for progressive, pseudo-ductile failure. The extensive literature on hybrid composites accumulated since the early 1970s is summarised in a few reviews [15]- [20]. The most practical approach based on the literature is the layer-by-layer or interlayer hybridisation, as intimate mixing of individual continuous fibres (also called intermingling) is not feasible on an industrial scale at the moment.
However excellent intermingling and pseudo-ductility was presented recently with aligned discontinuous fibres [21], [22]. The vast majority of the works presenting mechanical test results of hybrid composites deal with the typical combination of continuous glass and carbon fibres to improve the relatively low failure strain and brittle fracture of carbon fibre composites. A load drop at the low strain component failure and a significant improvement in the carbon component failure strain-usually referred to as a "hybrid effect"-has generally been reported. Different aspects of the hybrid effect are extensively discussed in [20] and [23]. The importance of the correct baseline for carbon layer strain improvement was also emphasised in [23].
The authors recently demonstrated desirable stable pseudo-ductile failure in tension by hybridising unidirectional (UD) standard thickness glass and thin carbon prepreg plies [24], [25] exploiting the fragmentation and stable delamination of the carbon layer. Up to 170% increase of the baseline glass composite modulus and 2.6% pseudo-ductile strain (defined between the final failure strain and the strain on the extrapolated initial slope line at the failure stress, see Table 4.) was achieved. A design framework for the hybrids was also developed in the form of both numerical [26] and analytical tools [27], [28]. These glass/carbon hybrids comprised some of the recently introduced thin-ply carbon prepregs which have generated high interest and have been studied extensively on their own [29]- [36] because of their unique potential to allow for highly dispersed lay-up designs resulting in favourable damage suppression properties.
There are only a few studies in the literature reporting the mechanical behaviour of hybrid composites made of different types of carbon fibres which will be referred to as all-carbon hybrids in this manuscript. Curtis and Browne [37] proposed a hybrid composite architecture where higher performance carbon fibres (T800) are put in the directions in which high stresses are expected (primary direction), and cheaper, lower performance carbon fibres (T300) are utilised in the less important directions (secondary direction). Up to 20% cost saving was anticipated while the mechanical properties of the hybrid composites were maintained at almost the same level as that of the high performance fibre reinforced baseline composites. Naito et al. [38] studied UD interlayer hybrids made of standard thickness (around 0.14 mm) prepreg plies of high strength (HS) and ultra-high modulus (UHM) carbon fibre/epoxy composites. A major load drop at the low strain component fracture was reported for hybrids with approximately 50 vol% of both components, with some load carrying capacity retained. The interlaminar fracture properties of the composites were also examined using specimens with pre-cut UHM carbon plies and release film of various lengths around the cuts. Montagnier and Hochard [39] considered high modulus (HM) and HS carbon reinforced composites for their design study of a drive shaft system. It was concluded that a hybrid composite design may reduce the number of shafts in the complete system and the total weight of the examined helicopter tail rotor driveline. Tsampas et al. [40] studied the compression performance of HS and intermediate modulus (IM) carbon hybrids in a quasi-isotropic lay-up sequence by utilising different types of prepreg plies in specific directions. No clear effect of the hybridisation on the failure loads was identified. Amacher et al, [41] and Czél et al [42] both presented initial results at the same conference showing pseudo-ductility in quasi-isotropic hybrid laminates made of different types of carbon/epoxy. Different fibre types and stacking sequences were investigated by the two groups and both strategies yielded interesting material behaviours.
The aim of this research is to demonstrate that pseudo-ductility can be achieved with all-carbon hybrids to generate high stiffness composites exhibiting gradual failure. This will allow for high specific stiffness by a large increase in the elastic modulus and a significant decrease of the density of the new pseudo-ductile hybrid composites compared to existing glass/carbon ones. The developed materials may be suitable for some of the most demanding safety-critical applications such as aerospace, motorsports and pressure vessels where ultimate stiffness and low weight are key design drivers.

Material and configuration design
This section gives details of the overall concept, the applied materials and the design considerations to assure a stable pseudo-ductile failure of the hybrid laminates.

Concept
The basic concept behind pseudo-ductility in thin-ply UD interlayer hybrid composites is the exploitation of two damage mechanisms i.e. fragmentation of the low strain material (LSM) and dispersed, stable delamination or stable pull-out of the LSM from the high strain material (HSM) layers local to the LSM cracks as presented earlier in [24], [28]. Fragmentation in this paper refers to the damage process where multiple fractures take place stably in a UD composite ply or layer loaded in tension parallel to the fibres. The fractures in the carbon plies were observed earlier [25] in UD glass/carbon hybrids visually through the translucent glass plies, which is not possible in the allcarbon/epoxy composites. It can be detected here as an almost horizontal plateau in the stress/strain response, when more and more fractures take place in the LSM upon further increase of displacement accompanied by only a small increase in stress. The fragmentation of the LSM is only achievable if the energy release rate at first LSM fracture is lower than the mode II fracture toughness of the interface between the LSM and HSM layers so that global unstable delamination of the layers is prevented. Instead, delamination may either be completely absent, or present in a dispersed, stable, localised form due to the low energy available to drive it. Therefore thin constituent prepregs are utilised to limit the energy released at the first LSM fracture. The resulting expected stress-strain responses of the studied UD hybrids are given in Fig. 1. The key features of the expected response are a quasilinear initial part, a flat, wide stress plateau and a second rising part before final failure. This metal-like failure character assures a warning sign which can be easily detected e.g. by acoustic emission [43] and provides a significant stress and strain margin after the first fracture of the LSM layer. These are large improvements to the brittle failure of conventional UD composites. In this study the previously demonstrated thin-ply hybrid concept is applied to all-carbon hybrids.

Materials
The materials considered for design, and used for the experiments were thin carbon/epoxy prepregs from North Thin Ply Technology made of different type intermediate modulus (IM), high modulus (HM) and ultrahigh modulus (UHM) carbon fibres. The choice of fibres in the utilised prepregs was driven by a previous parametric study [28] revealing that the higher the difference between the moduli of the hybridised materials, the larger the pseudo-ductile strain that may be achieved. This is the main reason for hybridising IM carbon fibres with HM and UHM carbon fibres. The matrix in all utilised prepregs was North TPT's ThinPreg 120 EPHTg-402 a 120°C cure toughened epoxy. Basic properties of the applied fibres and prepreg systems can be found in Tables 1 and 2. Please note, that the volume fraction of the thin prepregs were about 50%, 10 absolute% lower than that of standard thickness high performance composites, which limits the maximum achievable stiffness of the developed pseudo-ductile configurations.

Design of hybrid laminates
The following design criteria were identified earlier [24] to assure stable pseudo-ductile failure for UD glass/carbon interlayer hybrids. These criteria were adopted here for the preliminary design of similar interlayer hybrid configurations with new sets of constituent prepregs: (i) The outer, HSM layers (see fig. 3) need to be thick and strong enough to take the full load after low strain material fracture and pull-out. A sufficient margin should also be allowed to account for stress concentration in the HSM layer due to the LSM layer fracture which was not initially considered in this approach, but was shown not to be a major effect for similar interlayer hybrid configurations [27]. Inequality (1) is useful for choosing the high strain constituent material and gives the minimum strength of the HSM layer for given layer thicknesses and initial moduli.
Where E1 is the initial modulus of the HSM layers, E2 is the initial modulus of the LSM layer, t1 is the thickness of one HSM layer, t2 is the thickness of the LSM layer as shown on Fig. 3, σ1b is the breaking stress of the HSM layers, σ2b is the breaking stress of the LSM layer.
Inequality (1) can be rearranged in another form which is more practical for stacking sequence design. Inequality (2) gives the minimum thickness of one HSM layer for a given material pair and LSM thickness required to avoid premature failure at LSM layer fracture.
Where ε1b and ε2b are the breaking strain of the HSM and LSM layers respectively. If, as a first approximation, the breaking strains are assumed to be equal to the failure strains of the fibres, a lower bound for the HSM layer thickness can be obtained.
(ii) The energy release rate (GII) at the expected failure strain of the LSM layer must be lower than the mode II fracture toughness (GIIC) of the interface to avoid catastrophic delamination of the central LSM layer after its first fracture. This criterion assures the condition for fragmentation and stable pull-out of the LSM layer.
The preliminary design based on the properties of the available prepreg materials and the above criterion resulted in the configurations summarised in Table 3. The minimum required thickness of the HSM layer is exceeded in each configuration and the predicted mode II energy release rate GII at LSM failure strain in all cases is significantly lower than 0.5 N/mm which is the estimated fracture toughness GIIC of the tested materials. The critical value is based on observations during previous tests on similar hybrid configurations made of prepregs comprising the same epoxy matrix system [44]. A change from stable to unstable failure was observed earlier in configurations having GII>0.5 N/mm therefore this value was adopted as an approximation for GIIC. Based on the above discussion, all configurations are expected to fail in a progressive, pseudo-ductile way showing LSM fragmentation and stable pull-out. Two scaled configurations of T1000 and XN-80 fibres (one with blocked and one with dispersed lay-up sequence) are tested to analyse the effect of layer thicknesses and the corresponding energy release rate on the failure process. The third configuration of the same material pair was included to show the effect of the relative LSM thickness (to full thickness). The reason for testing a plate made of two identical sub-laminates in the 50 GSM XN-80 dispersed configuration, is that this material combination was tested early in the experimental study and our design was initially cautious as it is hard to machine and handle very thin laminates. It was confirmed later when other types were tested that it is possible to handle thin laminates of less than 0.2 mm thickness.
The thermal residual strain in the LSM was calculated using the moduli of the prepregs, the 100°C difference between cure and room temperatures, the coefficients of thermal expansion (CTE) of the fibres given in Table 1 and a typical value for the epoxy from the literature and included in Table 3 as well. A more advanced design tool based on the analytical criteria was also applied to the available material combinations to predict the expected failure mechanisms. The so-called damage-mode maps were developed by the authors recently and full details on how to generate the maps can be found in [27], [28].
The horizontal axis of each map in Fig. 2 shows the relative LSM layer thickness with regards to the full thickness of the hybrid laminate and the vertical one shows the absolute thickness of the LSM layer so every point on the map represents a specific hybrid configuration of a given material pair. The maps show the expected damage modes and the achievable pseudo-ductile strains for the selected configurations marked with circles. The pseudo-ductile strain (see Table 4. for graphical definition) is a good measure of improvement in failure type against sudden catastrophic fracture of the whole hybrid laminate therefore it was selected to indicate the ductility performance of the designed material configurations. The coloured regions of the maps indicate favourable pseudo-ductile damage process and the white regions show either premature HSM layer failure or catastrophic delamination. Each damage mode map in Fig. 2 was generated with a GIIC=0.5 N/mm estimated fracture toughness and the first LSM layer fracture was assumed to take place at the strains used for energy release rate calculation and given in Table 3. The HSM layer failure which always leads to final failure of the hybrid laminate depends on the variation of strain along the fragment lengths and the volume of material. This was estimated using a statistical strength distribution based on a typical Weibull modulus of 30 and reference failure strain of 2.2% for a reference volume of 1 mm 3 for the T1000 plies. This approach follows the one used successfully in [27] which should give conservative design configurations. A stress concentration factor of Kt=1.08 was assumed around the fractured low strain layer as explained in [27] for more accurate failure type prediction, but those at the end-tab region were not considered, therefore the predicted pseudo-ductile strains represent upper bounds. can take place in the LSM layer until they saturate, (4) LSM fragmentation and stable delamination: the energy release rate is lower than the fracture toughness at LSM failure but exceeds that after saturation of the LSM fragmentation before final failure. Therefore the fragmented LSM segments are pulled-out stably from the HSM layers (see Fig. 1).
Only one map was generated for T1000/XN-80 hybrids, where vf=49.4% was used for both thickness XN-80 prepregs. A reduced equivalent thickness was calculated for the lower volume fraction 63 g/m 2 XN-80 prepreg type as shown in Table 2 keeping the ply stiffness constant. The damage mode maps predict pseudoductile failure for all five designed configurations. Three of them fall into the LSM fragmentation only region, but they are very close to the boundary of LSM fragmentation and stable delamination region where the remaining two configurations sit and where the best pseudo-ductile properties are achievable. Please note that more optimal design was hindered by the limited choice of available thin carbon/epoxy prepregs (e.g. layer thicknesses in the hybrids were governed by available ply thicknesses). The designed configurations were deemed safe enough against premature or unstable failure and they were promising desirable properties despite the limitations and therefore they were all manufactured and tested. The damage mode maps proved to be powerful tools to design optimal hybrids laminates with desired behaviour and to predict the failure mode of a given configuration.

Specimen geometry
The specimens tested within the study were UD, parallel edge end-tabbed tensile specimens. Nominal specimen dimensions were 240/160/20/h mm overall length/Lf -free length/W-width/h-variable thickness respectively. Fig. 3 shows the geometric parameters on the side and top view schematics of a tensile specimen. Fig. 3. Schematic of the specimen geometry

Specimen manufacturing
The interlayer hybrid specimens were made by stacking the specified prepreg layers on top of each other, vacuum bagging the composite plate and curing it in an autoclave according to the recommended cure cycle of the constituent prepregs (2 hours at 120°C and 0.7 MPa pressure). The individual specimens were cut from 300x300 mm plates with a diamond cutting wheel in the case of specimens thicker than 0.2 mm and thinner laminates were cut with a CNC laser cutter. The edges of specimens cut with the laser were finished with P240 grit size sandpaper on a flat board. Finally 40 mm long, 2 mm thick cross-ply glass/epoxy tabs were bonded to the ends of the specimens with a two component epoxy adhesive.

Test method
Testing of the parallel edge specimens was executed under uniaxial tensile loading and displacement control at a crosshead speed of 2 mm/min on a computer controlled Instron 8801 type 100 kN rated universal hydraulic test machine with a regularly calibrated 25 kN load cell and wedge type hydraulic grips. The grip pressure was kept at the minimum required to avoid slippage of the specimens in the grips during loading to minimise the through thickness compressive stress at the end of the specimens. Strains were measured using an Imetrum video extensometer, with a nominal gauge length of 130 mm. A minimum of five specimens were tested from each configuration. predicted by the corresponding damage mode map (see Fig. 2a). This can result in a dispersed spatial distribution of initiating damage with several small cracks growing slowly at the same time while the stiffness of the hybrid is gradually reduced. The shape of the curves confirms the fragmentation of the LSM layer (plateau region) and

Results and discussion
gives the desired warning through sub-critical damage accumulation that is detectable e.g. by acoustic emission [43] or stiffness monitoring before final failure. The stress at damage initiation (first knee of the curves) was 1400 MPa on average which is high compared to other pseudo-ductile composites (around 1000 MPa for UD glass/carbon [25]) or ductile metals.  Table 4.)-was 0.58%, significantly higher than 0.5% which is given in the manufacturer's data sheet for the YSH-70A fibres. This indicates that there may be a hybrid effect increasing the LSM layer failure strain in this configuration because the development of broken fibre clusters could have been delayed by the low thickness of the LSM ply in the centre of the hybrid laminate. The LSM fibre failure strain given by the manufacturer is however not a satisfactory baseline to claim a hybrid effect. The importance of the LSM baseline strain and the types of possible hybrid effects are discussed extensively in [23].  Table 4) with the prediction of the corresponding damage mode map (Fig. 2c) indicating LSM fragmentation and stable delamination. The deviation in the predicted and measured pseudo-ductile strains is mostly attributed to the stress concentrations around the end-tabs affecting the final failure strain of the hybrids by HSM fracture. This effect was not implemented in the damage mode maps but could cause premature failure of the hybrids.
The second 50 GSM XN-80 dispersed configuration has two sub-laminates of hybrid blocks with half the thickness of the blocked type. This means that the relative LSM thickness of the dispersed and the blocked configurations is the same but the absolute thickness of the hybrid sub-laminates as well as the LSM layer is halved (see Fig. 2c). The thinner hybrid blocks of the dispersed type resulted in only half the amount of energy released at LSM fracture than in case of the blocked configuration (see Table 3). The pseudo-ductile strain of this configuration was limited by the early final failure which was most probably due to high stress concentrations around LSM layer fractures caused by the predicted low probability for stable pull-out capable of releasing the stress-concentrations (see Fig. 2c). This explanation is based on earlier numerical analysis of the effect of LSM absolute thickness on the stress concentration around LSM cracks in scaled UD glass/carbon hybrid laminates discussed in [26]. It was also reported elsewhere by the authors [23], that the thickness of the LSM layer has a significant effect on the damage initiation strain of the interlayer hybrids, because the development of broken The third 63 GSM XN-80 configuration had only a slightly lower energy release rate than the dispersed one and therefore is suitable to highlight the effect of the different relative LSM thickness (see Fig. 2c), which is most obvious from the lower initial modulus. Since the LSM ratio was less in this configuration than in the others, the length of the plateau and the change between initial and final slope of the stress-strain curves -corresponding to the contribution of the LSM layer to the hybrid stiffness-was lower. Although the energy release rate of this configuration at 0.4% strain was lower than that of the dispersed type, this hybrid was more likely to delaminate stably due to its higher HSM thickness, which could withstand stress-concentrations until higher strains at which stable pull-out took place. The obtained pseudo-ductile strain value was well predicted by the damage mode maps and it was in between that of the other two configurations of the same materials. It was noted that this type had the highest pseudo-yield strain of all XN80 configurations. This may have been caused by less damage introduced to the 63 g/m 2 XN80 prepreg when it was spread compared with the thinner 50 g/m 2 type. More stable stressstrain traces of the 63 GSM XN-80 specimens confirm that there may have been a wider margin between the actual and the critical energy release rates in this configuration than in others. Table 4 summarises the results of the tested configurations and gives the definitions of the pseudoductility parameters. The initial modulus values were evaluated with the nominal thicknesses based on assuming constant fibre areal densities and fibre volume fractions to exclude the effect of thickness variation due to small variations in resin distribution (see Table 3). The pseudo-yield points were evaluated at each individual specimen's stress-strain curve at the intersection of lines fitted to the initial linear and the plateau parts of the curves and the average pseudo-ductile strains were determined graphically by fitting lines to the aggregated test graphs of each series. It is worth noting that the pseudo-yield strain of the 50 GSM XN-80 blocked configuration was only 0.4%, similar to the value obtained earlier in [25] for the same thickness XN-80/epoxy block in S-glass/epoxy. The higher pseudo-yield strain of the 50 GSM XN-80 dispersed configuration is attributed to a hybrid effect due to the lower LSM thickness. In the case of the 63 GSM XN-80 type, the pseudo-yield strain could have been higher due to less damage introduced to the thicker XN80 prepreg during the tow spreading process. However, the fibre damage in different thickness thin plies and its effect on ply failure was not investigated therefore this hypothesis could not be confirmed. The pseudo-yield strains of the M55 and the YSH-70A types both exceed the strain of the LSM fibres given by the manufacturer, which could be attributed to a hybrid effect due to low thickness as well, but we do not have sufficient baseline LSM strains to claim this with confidence. The

Conclusions
The following conclusions were drawn from the study of thin-ply unidirectional all-carbon/epoxy interlayer hybrid composites:  A series of new high performance, unidirectional all-carbon/epoxy interlayer hybrid configurations were presented. Pseudo-ductility was demonstrated with all five tested configurations with high initial moduli providing up to 70% increase to the high strain material baseline. Well defined, wide plateaus were observed due to low strain material fragmentation which could be exploited as a warning sign before final failure if monitored e.g. by acoustic emission.
 The M55 configuration provided high initial modulus of 160 GPa together with 0.83% pseudo-yield (i.e. damage initiation) strain and 1400 MPa pseudo-yield stress, which is the highest pseudo-yield stress we have achieved with pseudo-ductile UD hybrids so far. The 50 GSM XN-80 blocked type material exhibited outstanding 245 GPa initial modulus, 990 MPa pseudo-yield stress and high, 0.94% pseudo-ductile strain.
These figures demonstrate the wide range of properties achievable and highlight the potential to tailor the mechanical response of the proposed hybrid materials.
 The scaled configurations of 50 GSM XN-80 blocked and dispersed types confirmed the key role of the low strain material layer thickness in controlling the hybrid effect in the unidirectional all-carbon/epoxy hybrid material.
 Good agreement was achieved between predicted and experimentally observed failure modes of the unidirectional carbon/epoxy interlayer hybrid composites. The predictions for the pseudo-ductile strains were reasonable. The presented study validated the damage mode map approach with a new set of materials and highlighted its merit for design and prediction.

Acknowledgement
This TPT for supplying materials for this research. All data required for reproducibility are provided within the paper.