High performance quasi-isotropic thin-ply carbon/glass hybrid composites with pseudo-ductile behaviour in all fibre orientations.

This study exploits the potential of thin-ply carbon/glass hybrid laminates to generate high performance Quasi-Isotropic (QI) composite plates that show pseudo-ductility in all fibre orientations under tensile loading, overcoming the inherent brittleness of conventional composites. Two types of QI lay-ups with 45° and 60° intervals, i.e. [45/90/-45/0] and [60/-60/0], were used to fabricate novel architectures of a QI T300-carbon laminate sandwiched between the two halves of a QI S-glass laminate. The fabricated plates were then loaded in all their fibre orientations. The laminates were designed by choosing an appropriate ratio of the carbon thickness to the laminate thickness using a robust analytical damage mode map. The experimental results verified the analytical predictions and showed a desirable pseudo-ductile failure in all the fibre orientations. Microscope images taken through the laminates thickness showed fragmentations (fibre fractures in the carbon layer) appearing only in the 0° carbon plies. A hybrid effect was observed, with an increase in strain and stress to failure of the carbon fibres, which was found to be dependent on the stiffness of the plies separating the 0° carbon plies and the plies adjacent to the 0° carbon plies. Altering the stacking sequence changes the stiffness of the separator and adjacent plies, therefore, leads to changes in the pseudo-ductile characteristics such as the initiation and final failure strains.


Introduction
Polymer-matrix composites use has been continuously increased in engineering applications due to their superior mechanical properties. However, polymer composites show sudden brittle failure, with linear elastic response and little warning before failure. This drawback leads to large values of safety factors in their design.
Hybridisation of different types of fibres is one of the methods that can introduce gradual failure in composite materials [1][2][3][4][5][6][7][8][9][10][11][12][13]. Possible failure modes in a three-layer Uni-Directional (UD) hybrid laminate made from High Strain Material (HSM) and Low Strain Material (LSM) are illustrated schematically in Fig. 1. Two types of catastrophic failure modes occur in conventional hybrid composites; a) a single crack through the whole thickness of the laminate due to a high ratio of LSM to HSM thickness (improperly sized hybrid laminate), which results in a linear stress-strain curve (Fig. 1a), and b) a single fracture in the LSM instantaneously followed by unstable delamination, which appears on the stress-strain graphs as a significant load drop (Fig. 1b). This is a typical failure mode in conventional standard ply thickness hybrid laminates. Fig. 1c shows a desired gradual failure where delamination is suppressed and multiple LSM fractures are obtained followed by stable localised pull-out. This desired pseudoductile behaviour is achievable by selecting appropriate material properties, appropriate values of relative thickness (i.e. proportion of the LSM to HSM) and absolute thickness of the LSM.
Thin carbon plies have superior mechanical properties and lower energy release rates, delaying the propagation of intra-and interlaminar cracks [14][15][16]. Recently, thin-ply UD and QI hybrids with different types of low strain and high strain fibres were introduced that generated the desired nonlinear stress-strain response and pseudo-ductility that avoids the catastrophic failure in laminated composites [17][18][19].  (ɛpd) and (iv) Hybrid effect. ɛpy and σpy are the strain and stress levels at which the tensile response deviates from the initial linear elastic behaviour, respectively. ɛf and σf are the final failure strain and stress values at which the HSM cannot carry any more load, respectively. It usually corresponds with fibre failure in the high strain material. Pseudo-ductile strain is the enhancement in strain achieved as a result of gradual failure and is calculated as the difference between the final failure strain, and the elastic strain based on the initial modulus at the final failure stress. Hybrid effects were reported by different researchers [9][10][11], where they observed higher strain to failure of carbon in their UD glass/carbon hybrids than that of all carbon specimens. Swolfs et al. [20] reviewed the basic mechanisms causing the hybrid effect, with the most significant considered to be thermal residual stresses, altered failure development due to statistical effects on formation of clusters of fibre breaks and dynamic stress concentrations. Wisnom et al. [21] used glass/carbon hybrid composites, rather than all carbon composites, to measure the strain to failure of the baseline carbon plies more accurately, to reduce the variability typically obtained using conventional unidirectional tests and avoid the high values for the hybrid effect values when using these tests as the baseline [9][10][11]. This method addressed the difficulty in measuring the strain to failure due to stress concentrations at the load introduction regions.
They defined hybrid effect as the enhancement in strain and stress to failure of the low strain fibres in the thin-ply hybrid composite, compared to those obtained in hybrid specimens with thick carbon plies where it was found that there is no hybrid effect. The performance of the pseudo-ductile QI layups is influenced not only by the UD laminates' design parameters, but also by the stacking sequence of the plies. Recently, two different stacking approaches were used to fabricate pseudo-ductile hybrids to make a QI hybrid laminate; (i) orientation-blocked method, where the hybrid laminates were fabricated from optimised UD hybrid sub-laminates [22] and (ii) orientation-dispersed method in which the hybrid laminates were made from different non-hybrid multi-directional sub-laminates [23]. A nonlinear stress-strain response was achieved in both cases, however, using the orientationblocked concept, there were significant free-edge delaminations. The orientation-dispersed method is considerably better to decrease interlaminar stresses at the free-edges and to suppress free-edge delamination [23].
Previous studies indicated that the pseudo-ductility is achieved in the QI hybrids in one loading orientation when subjected to tension [22][23]. However, most composite structures are subjected to multiple loading orientations, therefore, the aim of this paper is to study the potential of thin-ply carbon/glass hybrid laminates and the orientation-dispersed method to generate high performance QI composite plates that show pseudo-ductility when loaded in all the fibres orientations under tension. Microscope images taken through the laminates thickness were used to detect fragmentation location and density in the investigated opaque pseudoductile laminates. The results show that the order of the plies and the distance between the 0° carbon and glass plies influenced the damage scenario, the free-edge delamination and the hybrid effect. Altering the stacking sequence influences the stiffness of the inner plies separating the 0° carbon plies and the plies adjacent to the 0° carbon plies, which leads to changes in the pseudo-yield strain and final failure stress/strain. The study introduces for the first time hybrids that fail in a gradual manner when loaded in different directions-avoiding catastrophic failure and free-edge delamination.

Materials and specimen design
As listed in Table 1, the HSM is standard thickness UD S-glass/913 epoxy prepreg supplied by Hexcel and the LSM is thin carbon prepreg (SkyFlex USN020A) from SK Chemicals (South Korea) that has T300 carbon fibres made by Toray. Please note that these are different from the TR30 fibres in prepreg with the same manufacturer's designation tested previously [19].
The corresponding matrix was SK Chemical's type K 50 epoxy resin. The resin systems in the hybrid laminates were 120 o C curing epoxies, which were found to be compatible. a) Assumed to be equal to E-glass properties used in [24]. b) This value is based on T300 Data Sheet [25]. c) The transverse properties were assumed to be equal to the similar USN020A prepreg utilised in [26], as the data was not available. Table 2 gives information about the hybrid specimen types and the sequences that they were laid up in two different configurations of QI lay-ups, i.e. ±45QI (45/90/-45/0) and ±60QI (60/-60/0). From Table 2 and as shown in Fig. 3, adding the specified angles to the orientation of each sub-laminate, in specimens of type 1 and type 5, produces the other ±60QI and ±45QI layups, when the 0 direction is defined as the loading direction. By changing the angles at which each of the two quasi-isotropic configurations was tested, the behaviour in all the fibre directions of both types of laminates can be studied.  The layups were chosen using appropriate values of relative thickness (i.e. proportion of the low strain fibre plies) and absolute thickness of the carbon fibre plies. Pseudo-ductile response is achieved by suppression of catastrophic delamination and appearance of damage modes of (i) fragmentation in the carbon plies and (ii) local delamination, before the final failure of the high strain glass plies. This has been done through using damage mode maps, similar to those for UD laminates [18]. These show how the damage depends on the relative and absolute thickness of the different materials. To do so, the multi-directional glass and carbon sublaminates were assumed to be homogenised before crack development. Cracks are assumed to propagate through the thickness of the whole carbon epoxy laminate. The equivalent in-plane moduli of both glass and carbon QI laminates, calculated using Classical Laminate Theory, are given in Table 1. The failure strain of the homogenised materials is assumed to be equal to the fibre failure strain of the 0° layers. This is an acceptable assumption as the stiffness reduction in the deterioration process of the homogenised QI laminate is mainly due to 0° carbon layer fragmentation and the final failure of the QI glass is because of 0° layer glass failure. Although this assumption may not be completely accurate one due to hybrid effects, it is a reasonable basis to design the configurations. Most of the strength and stiffness of each layer is coming from the 0° layers and fragmentation/failure of this layer means a big drop in the stiffness of the remaining parts or failure of the whole laminate. Interfacial fracture toughness between the glass and carbon layers, GIIc, was assumed to be 1 N/mm [18]. The damage mode map with calculated boundaries between the different regions is illustrated in Fig. 4

Test procedure
Tensile testing of the hybrid laminates was performed under uniaxial tensile loading and displacement control using a crosshead speed of 2 mm/min on a computer controlled Instron 8801 type 100 kN rated universal hydraulic test machine with wedge-type hydraulic grips. A 25 kN load cell was used for better resolution in the expected load range. At least 6 specimens of each type were tested to check the repeatability of the results. To measure the strains over a nominal gauge length of 130 mm, an Imetrum video gauge system was used by tracking a dotted pattern applied on the specimen face.

Overall mechanical results
The desired pseudo-ductile response was achieved, with consistent stress-strain curves, without any catastrophic delamination and with obvious plateau as illustrated in Figs. 5-7. The carbon fibres started to fragment near the knee-point which was then followed by a plateau due to stable progressive carbon plies fragmentation and local delamination of the carbon fragments.
After the plateau, the stress started to increase as fragmentation was fully saturated, the contribution of the carbon plies to the specimen stiffness was diminished and the glass plies carried most of the increase in load. The failure behaviour was in good agreement with that expected from the design considerations. It should be noted that the hybrid effect is not considered in the damage mode map results shown in Fig. 4, and the manufacturer's quoted failure strain for T300 carbon (1.5 %) was used for the calculations. Table 3 gives some of the important features of the curve such as the initial elastic modulus, pseudo-yield strain, and pseudo-ductile strain of these hybrid configurations. Pseudo-yield strain values were defined as the intersection of two lines fitted to the stress-strain graph before and after the fragmentation initiation point. The elastic response of the layups is identical, as expected; however, due to different hybrid effects, the average pseudo-yield strains and the average pseudo-yield stresses are slightly different.

Hybrid effect
It has been shown that the hybrid effect is strongly dependent on the ply thickness in UD hybrids [16]. As will be discussed in Section 3.3., the fragmentations appeared only in the 0° carbon plies. The thickness of single 0° carbon plies is the same for the ±45QI and ±60QI laminates, however they have different hybrid effects due to the different parameters such as numbers and stacking sequences of plies in the investigated configurations. The stiffness of the plies separating the two 0° carbon plies on either side of the symmetry plane and the stiffness of the outer plies adjacent to the 0° carbon plies are different for the various tested configurations, as illustrated schematically in Fig. 8. The stiffnesses of the separator plies and the adjacent plies are worked out using the mechanical properties listed in Table 1 and reported in Table 3. These stiffness values are evaluated by multiplying the modulus of the separator plies along the loading direction by the thickness of the plies. The shear deformation is constrained by the rest of the laminate, therefore, the longitudinal stiffness of the unbalanced 45s and 60s plies are considered to be the same as the balanced ±45s and ±60s plies, respectively.
Since both the adjacent and separator plies are thought to influence the behaviour, a further column is included in Table 3 where these are added together. Although this is somewhat arbitrary, it does facilitate a comparison of their potential combined effect. For the ±60QI/0° and ±45QI/0° laminates, the nearby 0° glass ply has a significantly higher stiffness compared to the off-axis glass layers. So, when it is next to the fragmenting 0° carbon layer, the stress concentration at a single carbon fibre break would be lower as the glass ply can significantly contribute to carrying load around a subcritical cluster of broken fibres [21]. This explains the highest average pseudo-yield strain in the ±60QI/0° (1.80 %) and the ±45QI/0° (1.90 %) specimens. These strain values are significantly higher than the manufacturer's quoted 1.5 % failure strain [25] and the pseudo-yield strain values for the other configurations listed in Table   3. The higher pseudo-yield strain in the ±45QI/0° laminate, compared to the ±60QI/0° laminate, is due to the thicker, stiffer and stronger separator laminate between the two 0° plies in these specimens. The higher ratio of the stiff 0° glass layer to the overall thickness in the ±60QI/0° results in a higher final failure strain, compared to the ±45QI/0° laminate. More details can be found in [23].
For the other ±60QI and ±45QI laminates, the 0° carbon sub-laminate is adjacent to the ±60° and ±45° carbon layers with lower stiffness in the loading direction, respectively. Therefore, as listed in Table 3, the pseudo-yield strain values for these laminates is lower than the ±60QI/0° and ±45QI/0° laminates. Table 3 shows the different values for the hybrid effect of the investigated configurations. These values are measured relative to the ±60QI/+60° laminate that has the lowest hybrid effect. The pseudo-yield strain of ±60QI/+60° (1.6%) is higher than the manufacturer's quoted 1.5 % failure strain [26]. The reason for this higher value is partially as a result of eliminating stress concentrations at the grips using the hybrid configuration that results in consistent gauge length failures, but may include a small hybrid effect as the 0˚ plies are still quite thin. Using the failure strain of ±60QI/+60° as the reference allows a comparison of the relative hybrid effects between the different layups. There is a need for further experimental, numerical and analytical investigations to study these hybrid effects more comprehensively.   6. Results of the tensile tests for the ±45QI specimens. Fig. 7. Comparison of the tensile results for the ±45QI/0°and the ±60QI/0° specimens.  Fig. 8. Schematic of the investigated layups, the red lines on the 0° carbon plies are representative of fragmentations.

Microscopic observations
Microscope top view images from specimens interrupted before the final failure, taken through the laminate thickness by grinding and polishing layers one-by-one from off the top, showed that the fragmentations were only in the 0° carbon plies. As an instance, Fig. 9 shows microscopy pictures from the polished surface of the ±60QI/-60° laminate at different depths through the thickness. Similar results were obtained for all the investigated specimens. Fig. 10 shows the fragmentation pattern in the 0° carbon layer over the width of the investigated layups after removing the top layers. The highest fragmentation crack density and pseudo-yield strain value (1.9 %) with the smallest average fragmentation spacing is observed in the ±45QI/0° laminate which is believed to be due to the high combined stiffness of both the separator and adjacent plies (9958 N/mm). The ±60QI/0° laminate has the second-highest fragmentation cracks density and pseudo-yield strain value (1.8 %) with the second-highest combined stiffness value (7765 N/mm). The ±45QI/-45° laminate has the third-highest combined stiffness value (2042 N/mm) and the third-highest pseudo-yield strain value (1.7 %).
The other ±60QI and ±45QI laminates, have significantly lower combined stiffness values compared with the ±60QI/0° and ±45QI/0° laminates. As a result, the fragmentation cracks densities in these laminates are lower than the ±60QI/0° and ±45QI/0° laminates. Another interesting observation is the angle of the oblique cracks in the 0° carbon layer. As shown in Fig. 10, the overall fragmentation path angle is oblique, and it is affected by the other carbon layers adjacent to the 0° carbon layer. The oblique angle is between the transverse direction and the orientation of the inner adjacent carbon layer. When the 0° carbon plies are in the midplane, the oblique angle is between the transverse direction and the orientation of the layers adjacent to the 0° carbon plies. the 0° carbon ply with fragmentation and d) the 0° carbon ply, microscopy picture with higher zoom.  was distinguished. The finer damage pattern in the ±45 QI/0°, illustrated in Fig. 10, is hardly visible through the thick glass layers (Fig. 11 (5)). For the ±45 QI/-45° and ±45 QI/+45°, the carbon plies covering the fragmentations hindered the visibility of the damage (Fig. 11 (6), (8)). The 90° glass layer is on the top of the ±45 QI/+45° laminate and the matrix cracking is clearly observable parallel to the fibres in this layer. It will be shown that the energy release rates are significantly higher in the ±45QI/+90° laminate, compared with the other laminates, which caused free-edge delaminations, just after a few carbon ply fragmentations, and resulted in the different appearance ( Fig. 11 (7)) and the lowest final failure stress value (Fig. 6.).

Free-edge delamination
There was not any significant free-edge delamination for the investigated specimens except for the ±45 QI/90° specimen, hence the observable cracks on the ±45 QI/90° are due to the freeedge delaminations and fragmentations (See Fig. 12  delamination. This is consistent with the experimental observations and the side view picture taken from the ±45QI/+90° layup illustrated in Fig. 12c, where free-edge delamination initiated at interface 5 after the first fragmentations within the carbon laminate.

Conclusions
The aim of this work was to generate high performance QI composite laminates that show pseudo-ductility in all the fibres orientations under tensile loading. This aim was successfully achieved by designing, manufacturing and testing novel QI thin-ply carbon/glass hybrid laminates. The laminates were made with orientation-dispersed layups in which non-hybrid multi-directional S-glass/epoxy and T300-carbon/epoxy sub-laminates were put together to build the hybrid laminate. The following conclusions are drawn: 1. The investigated QI laminates showed the desired pseudo-ductility in all the fibre directions, with an obvious plateau and without any catastrophic delamination before the final failure. The orientation-dispersed concept successfully suppressed free-edge delamination, which was an unwanted damage mode in the previously investigated QI laminates fabricated using the orientation-blocked concept. Pseudo-ductile strains of up to 1.3 % were achieved, with strengths up to 504 MPa.
2. Microscopic observation showed carbon ply fragmentations just in the 0° carbon plies with overall fragmentation path angles that are oblique. The angle is affected by the other carbon layers adjacent to the 0° carbon layer, and is between the transverse direction and the orientation of the adjacent carbon layers.
3. A significant hybrid effect was observed, with a pseudo-yield strain corresponding to carbon ply failure of 1.90 % for the ±45QI laminates with the 0 carbon plies adjacent to the 0 glass plies, compared with only 1.6 % in some of the other ±60QI configurations. The effect was found to be dependent on the stiffness of the inner plies separating the 0° carbon plies and the plies adjacent to the 0° carbon plies. Altering the stacking sequence influences the stiffness of the separator and adjacent plies, therefore, leads to changes in the pseudo-yield strain and final failure stress/strain.