Fractographic Investigation of Cryogenic Temperature Mode-II Delamination Behavior of Filament Wound CFRP Laminates with Varied Resin Systems

: This study investigates the mode-II delamination performance of ﬁlament-wound unidirectional composites with different types of epoxies as their matrix phase under room and cryogenic temperatures. A typical vacuum infusion resin, an aerospace-grade cold-curing resin and crack-resistant toughened resin systems were wet-wound with 12K carbon ﬁber tows to manufacture the composite samples. Test samples with a (0) 16 ply sequence were tested according to ASTM D7905-19. The tested samples were investigated via microscopic analysis to assess the failure mechanisms associated with varying the matrix material and temperature. ENF tests at room temperature were found to be susceptible to the inherent variance in the ﬁber architectures along with resin-viscosity-driven ﬁber wetting. Cryogenic conditions induce a shift in the mode-II delamination behavior from a rather complex failure mechanism to a consistent ﬁber/matrix debonding mode with diminishing G IIc values except for the toughened resin system. The provided comprehensive fractographic analysis enables an understanding of the various causes of fracture, which determines the laminate performance. The combined evaluation of the distinctive damage modes reported in this study provides guidance on the conventional wet-winding process, which still remains a volumetrically dominant and viable option for cryogenic applications, particularly for vessels with limited operational durations like sounding rockets


Introduction
The high demands of the aerospace and automotive industries have resulted in the need to develop linerless composite pressure vessels due to their potential weight reduction.The weakest links for any composite structure are matrix-related micro-crack formation and accumulation, which may cause leakage problems and pressure-bearing capacity losses in the vessel even at room temperature [1,2].At cryogenic temperatures, the observation and understanding of the thermomechanical response of matrix-and matrix/fiber-interfacedriven failure modes becomes even more important [3][4][5].
The development of composite pressure vessels requires a combined systematic approach, including composite laminate design, modelling, material testing and largescale manufacturing.
Modelling studies focus on a multi-scaled approach, where constitutive relationships are of concern at different length scales and are frequently studied using finite element analysis [6,7].The aim here is to deduct the vessel performance in order to (i) choose the constituents at the microscale [8], (ii) reinforce the tow architectures at the mesoscale [4,5] and (iii) design related responses of vessels at the macroscale [9].In addition to the mechanical response, works investigating methods for determining the permeation rate and for software development have also been reported [10].Parallel to modelling efforts, experimental studies share the multi-scaled approach and focus on large-scale composite vessel manufacturing/testing, specimen-based laminate testing and constituent testing in a cryogenic environment.
At the macroscale, studies on cryogenic composite pressure vessel have been focused on the permeability characteristics under cryogenic conditions.Some of the most recent works were presented by Flanagan et al. [11], who studied the cryogenic cycling effect on the permeability of PEEK matrix composite pressure vessels considering variations in the pressure, thickness, fiber and matrix type.Grogan [12] worked on the damage and permeability of tape-laid thermoplastic composite tanks experimentally and numerically.Yokozeki [13] evaluated helium leakage on carbon-fiber-reinforced plastic (CFRP) tubes by applying tensile loading to the composite tubes at room temperature (RT) and at cryogenic liquid nitrogen (LN2) temperature.
At the microscale, the constituent thermomechanical properties are also profoundly significant in improving the composite laminate response under cryogenic conditions, especially for matrix materials.For this reason, many research studies have been conducted and continue to be carried out to understand the contribution of matrices to the mechanical response of composite laminates.One of the leading areas of focus on matrix materials is toughened epoxy systems for cryogenic environments.Their molecular movement capability under cryogenic conditions provides improved fracture toughness values, again under cryogenic conditions.This behavior is explained by the "free volume-free space" theory [14].The general trend in the literature is toward toughening epoxy systems with additives/fillers such as MWCNT [15], graphene platelets and rubbery particles [16][17][18].Liu et al. [19] reported that the addition of nanosheet boehmite (AlOOH) provides an increase in the interlaminar fracture toughness under room-temperature conditions; however, it provides no improvement under cryogenic temperatures.The filler type and weight ratio are critical for property maximization, and the optimal values can differ for roomtemperature and cryogenic conditions [20,21].The chemical toughening method represents another method for cryogenic environments, and it reduces transverse cracking under cryogenic conditions [5,22,23].Coronado et al. [22] studied the low-temperature fatigue resistance of untoughened and toughened matrix systems and reported the favorable effect of toughening on mode-I fatigue resistance.Also, damage morphologies such as river markings, broken fibers and hackle patterns have been reported.However, transverse cracking of chemically toughened epoxy systems under cryogenic conditions has not been reported at the meso-mechanical scale with fractographic images.
The effect of manufacturing parameters such as the fiber width and fiber undulation parameters on the mechanical properties and damage modes of composite laminates has also been investigated theoretically [24][25][26].Numerical analyses have also focused on the meso-scale effects of fiber undulations and winding patterns [27][28][29].One recent study conducted with a three-dimensional (3D) repeated unit cell (RUC) at the mesoscale by Pourahmadi [30] indicates that fiber bundle crossovers and undulations can change the mechanical and thermal properties by up to 15.7% when compared to the composite lamination theory.Shen et al. [31] also investigated the effects of fiber crossover and undulation on stiffness failure using numerical and analytical calculations, and they point out that the strain levels are higher by about 1.07-1.13times in the undulation regions compared to the laminate regions.However, all these works are based on mesoscale winding pattern analysis and do not discuss tow undulation effects at the microscale level.
Fracture toughness measurements to assess crack formation and propagation in composite structures have thus become a challenging task due to the above-stated variances in the different length scales.Such effects have been thoroughly investigated at room temperature for several composite structures.Barcikowski and Rybkowska [32] studied the characterization of reactive rubber-added toughened epoxy resins.Rased [33] studied the asymmetrical stacking sequence effect on the fracture toughness values of carbon fiber composites via double-cantilever beam (DCB), end-notched flexure (ENF) and mixedmode bending (MMB) tests.Morais [34] worked on the mode-II behavior of angle-ply glass/polyester filament-wound specimens.Bonhomme et al. [35] investigated the mode-I and mode-II interlaminar fracture surfaces of a unidirectional AS4/8552 carbon/epoxy laminate with fractographic and numerical analyses at room temperature.
Studies on fracture toughness testing under cryogenic temperatures remain rather unexplored when compared with RT testing efforts.Micro-crack progression was investigated by Li et al. [36], who discussed the effect of flexible molecular-structured epoxy on the cryogenic mechanical performance of the matrix and CFRP laminates with tensile and flexural tests under both room-temperature and cryogenic temperatures.Shindo [10] studied the delamination growth of woven glass-reinforced epoxy with mode-II fatigue loadings and four-point ENF tests at cryogenic temperatures.Choi and Sankar [37] performed four-point bending tests on single-notched graphite-epoxy laminates at room and cryogenic temperatures.Although these works have provided exemplary approaches, the current state of the art lacks an experimental approach where resin (ex situ) and laminate (in situ) testing are simultaneously performed.Moreover, the presented fractographic analyses have focused on explaining selected laminate behaviors rather than the effect of the resin choice and how the corresponding fracture mechanisms are affected under cryogenic environments.
Originating from this point, this article aims to highlight the effect of the resin type and manufacturing-related anomalies, such as resin bagging, fiber undulations and fiber wetting, on the mode-II delamination behavior of wet-wound UD composite specimens.Specific attention is given to the fractographic identification of such anomalies, which is of utmost importance for the failure analysis of composite pressure vessels.Three different resin systems were used for neat epoxy tests and for manufacturing CFRP laminate: lowviscosity cold curing (CC), low-viscosity warm-to-hot curing (VI) and toughened epoxy systems (T).Conventional filament winding by a four-axis winding machine was employed.The ENF specimens were prepared from (0) 16 laminates, which were manufactured by filament winding on a flat mandrel.The specimens were tested for their mode-II strain energy release rates (G IIc ) under RT and cryogenic (LN2) conditions.The modes of failure on the filament-wound CFRP laminates manufactured with different resin types and the effect of the temperature on the resin morphologies were investigated by detailed fractography analyses for each scenario.This thorough examination of these fracture surfaces provides a comprehensive understanding of the complex interaction between the behavior of epoxy systems and the damage mechanisms that arise within intricate mesoscale-architecture filament-wound composite laminates.

Materials
Three different DGBEA epoxy systems were used for this study.The first one, MARKUT 5025 (CC resin), was an aerospace cold-curing system, and the second one, BASAT 11546 (warm-to-hot-curing vacuum-infusion resin system) (VI resin), was purchased from CET Epoksi Teknolojileri A.Ş. (Istanbul, Turkey).The third one, CTD 7.1 Toughened Epoxy Resin (T resin), was a high-microcrack-fracture-toughness cryogenic resin system purchased from Composite Technology Development Inc. (Lafayette, LA, USA).Specific attention was given to the viscosity values to ensure that the chosen resin types were suitable for the filament-winding process.The characteristics of the purchased resin samples are shown in Table 1.
A-49 12K tow carbon fiber was purchased from Dow-Aksa (Istanbul, Turkey) and used for manufacturing the CFRP laminates (Figure 1).The selected resin systems exhibited significant viscosity differences, which was a tributed to both the resin and hardener viscosities, with the VI resin system having t lowest mixture viscosity and the T system exhibiting the highest mixture viscosity.

Specimen Preparation
First, neat epoxy samples were prepared for both room-temperature and cryogen testing.Before the stoichiometric mixing, all samples were gently mixed for 15 minutes enhance the homogeneity of the resin, hardener and initiator components.These mixtur were then introduced into silicon molds shaped according to ASTM D790-17 [38].T curing of the samples took place in a temperature-and humidity-controlled oven, follow ing the recommended curing specifications provided by the manufacturers (as outlin in Table 1).The Tg values from the in-house manufactured samples fell within an accept ble range when considering the scale of production.
ENF test specimens were manufactured in accordance with ASTM D7905/D7905M 19 [39] (Figure 2).The filament winding of the (0)16 UD laminates was carried out using specially designed mandrel following the ISO 1268-5 [40] guidelines.The winding proce employed a single row of fiber with a 3.1 mm bandwidth (Figure 3a).The winding tensio was set to 20 N, and the average winding speed was 15 rpm.Four hoop layups we wound, and a 12µm Teflon-based insert material was placed at both ends of the mandr (Figure 3b).Subsequently, four additional hoop layups were added to create a symme rical and balanced laminate.The CFRP specimens were then obtained from the manufa tured laminates through water jet cutting, as illustrated in Figure 3c, highlighting the u dulations caused by the filament-winding process.The selected resin systems exhibited significant viscosity differences, which was attributed to both the resin and hardener viscosities, with the VI resin system having the lowest mixture viscosity and the T system exhibiting the highest mixture viscosity.

Specimen Preparation
First, neat epoxy samples were prepared for both room-temperature and cryogenic testing.Before the stoichiometric mixing, all samples were gently mixed for 15 min to enhance the homogeneity of the resin, hardener and initiator components.These mixtures were then introduced into silicon molds shaped according to ASTM D790-17 [38].The curing of the samples took place in a temperature-and humidity-controlled oven, following the recommended curing specifications provided by the manufacturers (as outlined in Table 1).The T g values from the in-house manufactured samples fell within an acceptable range when considering the scale of production.
ENF test specimens were manufactured in accordance with ASTM D7905/D7905M-19 [39] (Figure 2).The filament winding of the (0) 16 UD laminates was carried out using a specially designed mandrel following the ISO 1268-5 [40] guidelines.The winding process employed a single row of fiber with a 3.1 mm bandwidth (Figure 3a).The winding tension was set to 20 N, and the average winding speed was 15 rpm.Four hoop layups were wound, and a 12 µm Teflon-based insert material was placed at both ends of the mandrel (Figure 3b).Subsequently, four additional hoop layups were added to create a symmetrical and balanced laminate.The CFRP specimens were then obtained from the manufactured laminates through water jet cutting, as illustrated in Figure 3c, highlighting the undulations caused by the filament-winding process.

Mechanical Tests
Initial attention was directed towards conducting cryogenic tests on the neat epoxy samples to investigate the connection between the self-behavior and the in-situ behavior of the selected resins.
Three-point bending tests were conducted on the neat epoxy samples to assess their mechanical responses under both room-temperature and cryogenic conditions.Both the cryogenic and room-temperature three-point bending tests on the cast epoxy samples adhered to the ASTM D790 [38] standards and were performed using a 10 kN Shimadzu AGS-X universal testing machine (Kyoto, Japan) at a crosshead displacement rate of 5mm/min.The force-versus-displacement values were recorded for the subsequent calculations of the parameters including the maximum flexural strain (εmax), maximum flexural stress (σmax) and Eflex flexural modulus.These parameters were determined using the following formulas: where F represents the maximum recorded force in the force-displacement curves, L is the half-span length, B is the specimen width, d is the thickness of the sample and D is the deflection of the beam center [38].A minimum of eight specimens were tested for each case.
For the ENF tests, the same three-point bending fixture was employed, maintaining a constant crosshead displacement rate of 5mm/min.Consequently, the non-precracked

Mechanical Tests
Initial attention was directed towards conducting cryogenic tests on the neat epoxy samples to investigate the connection between the self-behavior and the in-situ behavior of the selected resins.
Three-point bending tests were conducted on the neat epoxy samples to assess their mechanical responses under both room-temperature and cryogenic conditions.Both the cryogenic and room-temperature three-point bending tests on the cast epoxy samples adhered to the ASTM D790 [38] standards and were performed using a 10 kN Shimadzu AGS-X universal testing machine (Kyoto, Japan) at a crosshead displacement rate of 5mm/min.The force-versus-displacement values were recorded for the subsequent calculations of the parameters including the maximum flexural strain (εmax), maximum flexural stress (σmax) and Eflex flexural modulus.These parameters were determined using the following formulas: where F represents the maximum recorded force in the force-displacement curves, L is the half-span length, B is the specimen width, d is the thickness of the sample and D is the deflection of the beam center [38].A minimum of eight specimens were tested for each case.
For the ENF tests, the same three-point bending fixture was employed, maintaining a constant crosshead displacement rate of 5mm/min.Consequently, the non-precracked

Mechanical Tests
Initial attention was directed towards conducting cryogenic tests on the neat epoxy samples to investigate the connection between the self-behavior and the in-situ behavior of the selected resins.
Three-point bending tests were conducted on the neat epoxy samples to assess their mechanical responses under both room-temperature and cryogenic conditions.Both the cryogenic and room-temperature three-point bending tests on the cast epoxy samples adhered to the ASTM D790 [38] standards and were performed using a 10 kN Shimadzu AGS-X universal testing machine (Kyoto, Japan) at a crosshead displacement rate of 5mm/min.The force-versus-displacement values were recorded for the subsequent calculations of the parameters including the maximum flexural strain (ε max ), maximum flexural stress (σ max ) and E flex flexural modulus.These parameters were determined using the following formulas: where F represents the maximum recorded force in the force-displacement curves, L is the half-span length, B is the specimen width, d is the thickness of the sample and D is the deflection of the beam center [38].A minimum of eight specimens were tested for each case.
For the ENF tests, the same three-point bending fixture was employed, maintaining a constant crosshead displacement rate of 5 mm/min.Consequently, the non-precracked G IIc values were calculated based on the direct beam theory, as reported in [41].The formula for calculating G IIc is as follows: where P represents the maximum force, δ is the maximum displacement, a is the delamination crack length, b is the specimen width and L is the span length.Both RT and LN2 tests were carried out (See Figure 4).GIIc values were calculated based on the direct beam theory, as reported in [41].The formula for calculating GIIc is as follows: where P represents the maximum force, δ is the maximum displacement, a is the delamination crack length, b is the specimen width and L is the span length.Both RT and LN2 tests were carried out (See Figure 4).The room-temperature tests were conducted in an environment with a temperature variation ranging from 20-25 °C, while the cryogenic tests involved submerging the specimen in liquid nitrogen (LN2).The LN2 temperature was measured with a thermocouple and maintained at −196 ± 1 °C.Prior to testing in LN2, the specimens were pre-cooled using boiled liquid nitrogen to prevent any thermal shock.

Microscopy
The tested ENF specimens were examined from their sides to observe the nature of the cracking events resulting from the temperature change.
Fracture surface analysis of the ENF samples was performed using a Zeiss LEO Supra VP35 (Oberkochen, Germany) field emission scanning electron microscope (SEM) (Sabanci University, Istanbul, Turkey).Prior to the analysis, a thin conductive Au/Pd coating was sputter-deposited onto the samples.The samples collected from both the end of the insert and the fracture surface were investigated in two distinct regions: the pre-crack front (crack initiation zone) and the far regions (crack propagation zone).This allowed for observations and comments on the differences in the crack initiation and propagation during the testing.The examination of the fracture surfaces was conducted in the secondary electron mode with a gun voltage of 5 kV.

Three-Point Bending Tests
Figure 5 depicts the representative flexural stress/strain curves obtained from the neat epoxy sample tests, revealing a transition from elastoplastic behavior at room temperature to fully elastic behavior at cryogenic temperatures.The overall results are summarized in Table 2.When exposed to cryogenic conditions, the increase in Eflex for all three resin systems was approximately 280-400%, which was primarily due to embrittlement.Similarly, the strength values also showed an increase under cryogenic conditions for all three systems when compared to the room temperature values.However, the maximum elongation values exhibited a significant decrease when transitioning to cryogenic levels.This observation suggests that the chain mobility of the resin systems was reduced, The room-temperature tests were conducted in an environment with a temperature variation ranging from 20-25 • C, while the cryogenic tests involved submerging the specimen in liquid nitrogen (LN2).The LN2 temperature was measured with a thermocouple and maintained at −196 ± 1 • C. Prior to testing in LN2, the specimens were pre-cooled using boiled liquid nitrogen to prevent any thermal shock.

Microscopy
The tested ENF specimens were examined from their sides to observe the nature of the cracking events resulting from the temperature change.
Fracture surface analysis of the ENF samples was performed using a Zeiss LEO Supra VP35 (Oberkochen, Germany) field emission scanning electron microscope (SEM) (Sabanci University, Istanbul, Turkey).Prior to the analysis, a thin conductive Au/Pd coating was sputter-deposited onto the samples.The samples collected from both the end of the insert and the fracture surface were investigated in two distinct regions: the pre-crack front (crack initiation zone) and the far regions (crack propagation zone).This allowed for observations and comments on the differences in the crack initiation and propagation during the testing.The examination of the fracture surfaces was conducted in the secondary electron mode with a gun voltage of 5 kV.

Three-Point Bending Tests
Figure 5 depicts the representative flexural stress/strain curves obtained from the neat epoxy sample tests, revealing a transition from elastoplastic behavior at room temperature to fully elastic behavior at cryogenic temperatures.The overall results are summarized in Table 2.When exposed to cryogenic conditions, the increase in E flex for all three resin systems was approximately 280-400%, which was primarily due to embrittlement.Similarly, the strength values also showed an increase under cryogenic conditions for all three systems when compared to the room temperature values.However, the maximum elongation values exhibited a significant decrease when transitioning to cryogenic levels.This observation suggests that the chain mobility of the resin systems was reduced, leading to a corresponding increase in the intermolecular binding forces.These changes directly influenced the measured E flex and elongation values [42][43][44][45].
duced brittleness effect was also observed in the CC resin and VI resin systems.An intriguing point to note is that the Eflex value of the toughened resin system under room-temperature conditions was closely comparable to the cryogenic-temperature Eflex values of the other resin systems.The T system was the system most affected by the cryogenic temperatures, as it exhibited the highest strength value and the lowest strain among all the other resins under cryogenic conditions.

End-Notched Flexure Tests
Figure 6 and Figure 7 display the representative force-elongation and flexural-stressflexural-strain curves of the tested ENF specimens, respectively.The GIIc values for the CC resin system were measured as 1.74 ± 0.17 (kJ/m 2 ) at room temperature and 0.88 ± 0.05 (kJ/m 2 ) at cryogenic temperature.In contrast, such a substantial decrease in the GIIc values was not observed for the VI and T resins (Table 2).The room-temperature GIIc level for the VI resin was 0.97 ± 0.21 (kJ/m 2 ), while the cryogenic-temperature GIIc level was 0.94 ± 0.08 (kJ/m 2 ).For the T resin system, the force-elongation curve exhibited a remarkable similarity under both room and cryogenic conditions, with GIIc values of 1.268 ± 0.091 (kJ/m 2 ) and 1.216 ± 0.121 (kJ/m 2 ) being obtained, respectively.The rationale behind the measured GIIc values is discussed in more detail in the following section on the fractographic analysis.The toughened resin system (T resin) exhibited remarkable toughness during the room-temperature test, enduring the test fixtures when they made contact, which serves as a clear demonstration of the positive toughening effect.However, the toughness effect was lost under cryogenic conditions, and the resin became brittle.This temperature-induced brittleness effect was also observed in the CC resin and VI resin systems.An intriguing point to note is that the E flex value of the toughened resin system under room-temperature conditions was closely comparable to the cryogenic-temperature E flex values of the other resin systems.The T system was the system most affected by the cryogenic temperatures, as it exhibited the highest strength value and the lowest strain among all the other resins under cryogenic conditions.

End-Notched Flexure Tests
Figures 6 and 7 display the representative force-elongation and flexural-stress-flexuralstrain curves of the tested ENF specimens, respectively.The G IIc values for the CC resin system were measured as 1.74 ± 0.17 (kJ/m 2 ) at room temperature and 0.88 ± 0.05 (kJ/m 2 ) at cryogenic temperature.In contrast, such a substantial decrease in the G IIc values was not observed for the VI and T resins (Table 2).The room-temperature G IIc level for the VI resin was 0.97 ± 0.21 (kJ/m 2 ), while the cryogenic-temperature G IIc level was 0.94 ± 0.08 (kJ/m 2 ).For the T resin system, the force-elongation curve exhibited a remarkable similarity under both room and cryogenic conditions, with G IIc values of 1.268 ± 0.091 (kJ/m 2 ) and 1.216 ± 0.121 (kJ/m 2 ) being obtained, respectively.The rationale behind the measured G IIc values is discussed in more detail in the following section on the fractographic analysis.

Macro-Scale Interpretation of the Main Mode of Failure
The effect of tow undulation was evident from the cross-sectional views of the tested specimens, as shown in Figure 2c.The side view of the tested specimens (Figure 8) revealed a linear crack propagation pattern at room temperature, which was replaced with a zig-zag crack propagation under LN2 conditions.The linear crack propagation was somewhat expected from the ENF tests due to the propagation of a mode-II crack along the interlaminar plane [33].The periodic zig-zag crack propagation in the LN2 environment, on the other hand, suggests the coexistence of intra-and interlaminar damage propagation.The reduction in the GIIc values, along with the presence of intra-laminar cracking, clarified the dominant mode of failure as fiber/matrix debonding under LN2 conditions due to the extreme cold affecting the regular resin types.These observations are linked to the significant difference in the thermal expansion coefficients between the tested resins and carbon fibers [10,36,37].

Macro-Scale Interpretation of the Main Mode of Failure
The effect of tow undulation was evident from the cross-sectional views of the tested specimens, as shown in Figure 2c.The side view of the tested specimens (Figure 8) revealed a linear crack propagation pattern at room temperature, which was replaced with a zig-zag crack propagation under LN2 conditions.The linear crack propagation was somewhat expected from the ENF tests due to the propagation of a mode-II crack along the interlaminar plane [33].The periodic zig-zag crack propagation in the LN2 environment, on the other hand, suggests the coexistence of intra-and interlaminar damage propagation.The reduction in the GIIc values, along with the presence of intra-laminar cracking, clarified the dominant mode of failure as fiber/matrix debonding under LN2 conditions due to the extreme cold affecting the regular resin types.These observations are linked to the significant difference in the thermal expansion coefficients between the tested resins and carbon fibers [10,36,37].

Macro-Scale Interpretation of the Main Mode of Failure
The effect of tow undulation was evident from the cross-sectional views of the tested specimens, as shown in Figure 2c.The side view of the tested specimens (Figure 8) revealed a linear crack propagation pattern at room temperature, which was replaced with a zig-zag crack propagation under LN2 conditions.The linear crack propagation was somewhat expected from the ENF tests due to the propagation of a mode-II crack along the interlaminar plane [33].The periodic zig-zag crack propagation in the LN2 environment, on the other hand, suggests the coexistence of intra-and interlaminar damage propagation.The reduction in the G IIc values, along with the presence of intra-laminar cracking, clarified the dominant mode of failure as fiber/matrix debonding under LN2 conditions due to the extreme cold affecting the regular resin types.These observations are linked to the significant difference in the thermal expansion coefficients between the tested resins and carbon fibers [10,36,37].

Macro-Scale Interpretation of the Main Mode of Failure
The effect of tow undulation was evident from the cross-sectional views of the te specimens, as shown in Figure 2c.The side view of the tested specimens (Figure 8 vealed a linear crack propagation pattern at room temperature, which was replaced a zig-zag crack propagation under LN2 conditions.The linear crack propagation somewhat expected from the ENF tests due to the propagation of a mode-II crack a the interlaminar plane [33].The periodic zig-zag crack propagation in the LN2 env ment, on the other hand, suggests the coexistence of intra-and interlaminar damage p agation.The reduction in the GIIc values, along with the presence of intra-laminar crac clarified the dominant mode of failure as fiber/matrix debonding under LN2 condi due to the extreme cold affecting the regular resin types.These observations are link the significant difference in the thermal expansion coefficients between the tested r and carbon fibers [10,36,37].Pre-crack insertion during conventional filament winding can be rather challenging when compared to other mold-based manufacturing techniques (e.g., pre-preg laying, RTM).In this process, the entire manufacturing operation must be temporarily halted, and pre-cut non-sticking films need to be introduced at the crack-initiation regions on one part of the flat mold.During this pause, the amount of liquid resin on the pre-tensioned carbon fibers can change due to variations in the resin viscosities.As the process is continuous, the tows subjected to this interruption form the upper part of the interlaminar region.Consequently, the amount of resin on the interlaminar plane may vary, leading to the occurrence of resin bagging [22].To illustrate this effect, the pre-crack fronts of the selected samples were analyzed using SEM.
The primary characteristic observed on the pre-crack fronts of all three types of samples that fractured at room temperature was the presence of resin bagging (Figure 9A-C).This issue led to the formation of periodic tow undulation.The undulated tow structure created hills on the fracture surface, resulting in non-planar crack propagation.This, in turn, left nearly perpendicular shear cusps at the side of the hills and caused simultaneous fiber/matrix debonding to occur on the plateau.The damage shape of the matrix was determined by the local stress state and the orientation of the hackles, indicating the direction of crack propagation [46].For the toughened T resin system, local out-of-plane crack deflections revealed instances of crack jumping to other planes, and non-uniform crack initiation occurred at the crack front.These characteristics are unique to the conventional filament-winding process [30,31,33].
On the far region of the same undulated tow (Figure 9A-C), larger hackle formations characteristic of mode-II crack propagation and fiber/matrix debonding regions were demonstrated for the variance in the resin volume fraction along the interlaminar plane.The fundamental reason for this was determined to be the limited control of the perpendicular resin flow and tow overlapping [47,48] during the wet-winding process.In the investigated cases, the resin flow problem was even more problematic due to the pre-crack sticking-filminsertion process, where a total process stop was required.
For cryogenic conditions, the surface view was mostly flat and smooth, which suggests fiber/matrix debonding without any plastic deformation (Figure 9D,E).The effect of tow undulations was minor.However, again, for the toughened case, matrix hackle patterns and fiber/matrix debonding damage modes can be seen together.Also, the tow misalignment suggests that the tow movement was still effective for this toughened resin system due to the sample's plastic deformation capability under cryogenic conditions (Figure 9F).

Crack Propagation Surfaces and Resin Morphologies
4.2.1.Pre-Crack Region (Crack Initiation Zone) Pre-crack insertion during conventional filament winding can be rather challenging when compared to other mold-based manufacturing techniques (e.g., pre-preg laying RTM).In this process, the entire manufacturing operation must be temporarily halted and pre-cut non-sticking films need to be introduced at the crack-initiation regions on on part of the flat mold.During this pause, the amount of liquid resin on the pre-tensioned carbon fibers can change due to variations in the resin viscosities.As the process is con tinuous, the tows subjected to this interruption form the upper part of the interlamina region.Consequently, the amount of resin on the interlaminar plane may vary, leading to the occurrence of resin bagging [22].To illustrate this effect, the pre-crack fronts of th selected samples were analyzed using SEM.
The primary characteristic observed on the pre-crack fronts of all three types of sam ples that fractured at room temperature was the presence of resin bagging (Figure 9A-C) This issue led to the formation of periodic tow undulation.The undulated tow structur created hills on the fracture surface, resulting in non-planar mode-II crack propagation This, in turn, left nearly perpendicular shear cusps at the side of the hills and caused sim ultaneous fiber/matrix debonding to occur on the plateau.The damage shape of the matrix was determined by the local stress state and the orientation of the hackles, indicating th direction of crack propagation [46].For the toughened T resin system, local out-of-plan crack deflections revealed instances of crack jumping to other planes, and non-uniform crack initiation occurred at the crack front.These characteristics are unique to the conven tional filament-winding process [30,31,33].
On the far region of the same undulated tow (Figure 9A-C), larger hackle formation characteristic of mode-II crack propagation and fiber/matrix debonding regions wer demonstrated for the variance in the resin volume fraction along the interlaminar plane The fundamental reason for this was determined to be the limited control of the perpen dicular resin flow and tow overlapping [47,48] during the wet-winding process.In th investigated cases, the resin flow problem was even more problematic due to the pre crack sticking-film-insertion process, where a total process stop was required.
For cryogenic conditions, the surface view was mostly flat and smooth, which sug gests fiber/matrix debonding without any plastic deformation (Figure 9D,E).The effect o tow undulations was minor.However, again, for the toughened case, matrix hackle pat terns and fiber/matrix debonding damage modes can be seen together.Also, the tow mis alignment suggests that the tow movement was still effective for this toughened resin sys tem due to the sample's plastic deformation capability under cryogenic conditions (Figur 9F).

Far-Crack Region (Crack Propagation Zone)
Low-magnification images of the fracture surfaces of the investigated samples are presented for all three specimen types in Figure 10.A fundamental difference in the CFRP VI resin laminate was the lower fiber wettability observed on the delamination surface.In contrast, the size and quantity of the hackle patterns (serrations) [46] on the CFRP-CC resin and CFRP-T resin surfaces were notably greater than those observed in the CFRP-V resin laminate.This discrepancy accounts for the variations in measured GIIc values a room temperature.Another valuable observation was the coexistence of interlaminar de lamination (leaving hackle marks in between the continuous fibers) and fiber/matrix debonding failure events for all three types of specimens (as highlighted in Figure 10A, C E).The continuous carbon fiber surfaces of the CFRP-CC resin laminates appeared rela tively smoother when compared to CFRP-VI resin, indicating poorer fiber/matrix adhe sion in the former.In contrast, the T resin system exhibited regions of the fiber/matrix interface that were well preserved, indicating strong fiber/matrix adhesion.Even in cases of fiber breakage, the presence of strong fiber/matrix adhesion was evident, leading to fiber rupturing (Figure 10E).
Furthermore, the presence of perpendicular hackle patterns (Figure 10A) suggests that the meso-structure resulting from the winding process induced non-planar crack propagation.This, in turn, led to an increase in the measured GIIc values due to the tow interlocking effect.The frequency of such patterns was relatively lower in the case of the CFRP-VI resin (Figure 10C).When zoomed in (Figure 10B), the morphology of hackle pat terns revealed that the in-situ behavior of the CC resin was much more ductile when com pared to the VI resin.This is in contrast with the three-point bending results (Table 2).This

Far-Crack Region (Crack Propagation Zone)
Low-magnification images of the fracture surfaces of the investigated samples are presented for all three specimen types in Figure 10.A fundamental difference in the CFRP-VI resin laminate was the lower fiber wettability observed on the delamination surface.In contrast, the size and quantity of the hackle patterns (serrations) [46] on the CFRP-CC resin and CFRP-T resin surfaces were notably greater than those observed in the CFRP-VI resin laminate.This discrepancy accounts for the variations in measured G IIc values at room temperature.Another valuable observation was the coexistence of interlaminar delamination (leaving hackle marks in between the continuous fibers) and fiber/matrix debonding failure events for all three types of specimens (as highlighted in Figure 10A,C,E).The continuous carbon fiber surfaces of the CFRP-CC resin laminates appeared relatively smoother when compared to CFRP-VI resin, indicating poorer fiber/matrix adhesion in the former.In contrast, the T resin system exhibited regions of the fiber/matrix interface that were well preserved, indicating strong fiber/matrix adhesion.Even in cases of fiber breakage, the presence of strong fiber/matrix adhesion was evident, leading to fiber rupturing (Figure 10E).
Furthermore, the presence of perpendicular hackle patterns (Figure 10A) suggests that the meso-structure resulting from the winding process induced non-planar crack propagation.This, in turn, led to an increase in the measured G IIc values due to the tow interlocking effect.The frequency of such patterns was relatively lower in the case of the CFRP-VI resin (Figure 10C).When zoomed in (Figure 10B), the morphology of hackle patterns revealed that the in-situ behavior of the CC resin was much more ductile when compared to the VI resin.This is in contrast with the three-point bending results (Table 2).This contradiction suggests that the higher strain values measured for the CC resin system during the ENF testing (Figure 7) were mainly due to the low fiber volume fraction ratio and also the undulations in the tow structures, whose movement was enabled at room temperature.On the other hand, the fiber/resin interaction was much more favorable in the case of the VI resin and the CC resin, where the surfaces of the carbon fibers were not as smooth, which caused a lower effect on the mode-II crack propagation (Figure 10D,E).The toughened system's resin damage morphology showed crack deflection in the investigated region (Figure 10F).The toughness effect revealed its contribution to the damage mode of the resin system due to the strong fiber/matrix adhesion and high shear deformation capability (Figure 10F).temperature.On the other hand, the fiber/resin interaction was much more favorable in the case of the VI resin and the CC resin, where the surfaces of the carbon fibers were not as smooth, which caused a lower effect on the mode-II crack propagation (Figure 10D,E).
The toughened system's resin damage morphology showed crack deflection in the investigated region (Figure 10F).The toughness effect revealed its contribution to the damage mode of the resin system due to the strong fiber/matrix adhesion and high shear deformation capability (Figure 10F).In contrast to the tests conducted at room temperature, the quantity of resin failure marks on the samples at cryogenic temperatures was notably smaller and less densely distributed, indicating that resin failure and tow undulations had a reduced impact on the fracture of the CC resin system (Figure 11A).Consequently, the zig-zag crack propagation (Figure 7B) observed on the macroscale was primarily attributed to fiber/matrix debonding.However, in the case of the toughened resin system, the rate of resin failure was significantly higher on the damaged surface, where the resin remained effective under

Effect of LN2 on Multi-Scaled Interactions
In contrast to the tests conducted at room temperature, the quantity of resin failure marks on the samples at cryogenic temperatures was notably smaller and less densely distributed, indicating that resin failure and tow undulations had a reduced impact on the fracture of the CC resin system (Figure 11A).Consequently, the zig-zag crack propa-gation (Figure 7B) observed on the macroscale was primarily attributed to fiber/matrix debonding.However, in the case of the toughened resin system, the rate of resin failure was significantly higher on the damaged surface, where the resin remained effective under cryogenic conditions (Figure 11C).The overall appearance of the fracture surface of the CFRP-VI resin laminates (Figure 11B) closely resembled that of the CFRP-CC laminates, which further supports the similarity in the measured G IIc values.
ductile hackles under cryogenic conditions, alongside strong fiber/matrix adhesion.This explains the equivalent GIIc values obtained between the RT and LN2 conditions for this resin system.
When further zoomed in, the shape of the hackles on the CFRP-CC resin laminate unveiled an example failure mode unique to the cryogenic testing.This mode involved in fiber/matrix interface cracking and fiber shear rupture (Figure 11G) due to the extreme cold and large difference in the thermal expansion coefficients between the constituents [10].This situation may have caused a self-cracking event prior to the loading, potentially leading to a reduction in the GIIc values.This failure mode was not observed in the CFRP-VI resin and CFRP-T resin laminates, where better fiber/matrix adhesion was achieved.In the case of the CFRP-T resin system, the strong fiber/matrix adhesion resulted in tow cracking and splitting on the fiber (Figure 11I).Furthermore, the ductile resin failure morphology revealed the cryogenic performance capability of this resin system (Figure 11I).When zoomed in (Figure 11D,E), a notable shift in the resin morphologies for the CC resin and VI resin was evident under cryogenic conditions.In this context, the fiber/matrix interfaces displayed smaller and thinner hackles, a consequence of the brittleness effect [46].Conversely, the dominance of resin damage was apparent for the T resin system, indicating the contribution of the resin to the G IIc values under cryogenic conditions (Figure 11F).An enhanced quality of fiber/matrix bonding was once again observed in the CFRP-VI resin and CFRP-T resin laminates.This was particularly relevant for the cryogenic fracture of the laminates, as the primary mode of failure was associated with fiber/matrix debonding.Notably, the toughened CFRP-T resin still showed large fractured ductile hackles under cryogenic conditions, alongside strong fiber/matrix adhesion.This explains the equivalent G IIc values obtained between the RT and LN2 conditions for this resin system.
When further zoomed in, the shape of the hackles on the CFRP-CC resin laminate unveiled an example failure mode unique to the cryogenic testing.This mode involved in fiber/matrix interface cracking and fiber shear rupture (Figure 11G) due to the extreme cold and large difference in the thermal expansion coefficients between the constituents [10].This situation may have caused a self-cracking event prior to the loading, potentially leading to a reduction in the G IIc values.This failure mode was not observed in the CFRP-VI resin and CFRP-T resin laminates, where better fiber/matrix adhesion was achieved.In the case of the CFRP-T resin system, the strong fiber/matrix adhesion resulted in tow cracking and splitting on the fiber (Figure 11I).Furthermore, the ductile resin failure morphology revealed the cryogenic performance capability of this resin system (Figure 11I).

Conclusions
The mode-II delamination behavior of filament-wound CFRP laminates with different resin types was extensively investigated under both room-temperature and cryogenic conditions by employing a detailed fractographic analysis.Notably, the G IIc values of the CC resin system exhibited a significant reduction, decreasing from 1.74 ± 0.17 (kJ/m 2 ) at room temperature to 0.88 ± 0.05 (kJ/m 2 ) at cryogenic temperature.In contrast, both the VI and T resin systems displayed relatively stable G IIc values between room and cryogenic temperature.The measured G IIc values of the VI resin were 0.97 ± 0.21 (kJ/m 2 ) at room temperature and 0.94 ± 0.08 (kJ/m 2 ) at cryogenic temperature.The G IIc values for the T resin system were 1.268 ± 0.091 (kJ/m 2 ) under room-temperature conditions and 1.216 ± 0.121 (kJ/m 2 ) under cryogenic conditions.The following conclusions were drawn based on the experimental findings of this study:

•
The three-point bending tests on the neat epoxy specimens demonstrated that cryogenic temperatures led to a decrease in the elongation and an increase in the strength and stiffness due to embrittlement.It became apparent that the neat epoxy's performance may not accurately reflect the cryogenic performance of the composites due to the dominant influence of fiber/matrix interactions on the damage behavior.Consequently, experimental efforts focusing on single-fiber tow/resin interactions may be favored.

•
The mode-II delamination behavior was notably affected by inherent resin bagging and tow undulation problems associated with conventional wet winding.Their presence caused complex 3D crack propagation.Tow movement at room temperature provided resin ductility, contributing to an increase in the G IIc levels through a tow interlocking mechanism.

•
Matrix shear failure, fiber/matrix debonding and even fiber fracture events were observed during crack propagation at room temperature.

•
Under cryogenic temperatures, the dominant failure mode shifted towards fiber/matrix debonding, characterized by visible zig-zag crack propagation, as seen from side views of the specimens.

•
The VI resin, with the lowest mixture viscosity, suffered from a wetting problem that increased the contribution of the undulated fiber architecture (tow interlocking).This mechanism explained the similar G IIc values measured at both room and cryogenic temperatures, with the embrittlement events observed for the neat resin system having a limited impact on the laminate behavior.

•
The CC resin system, with a moderate mixture viscosity, enabled better wetting of the interlaminar plane.This was associated with the pre-crack insertion process, where the overall winding process was temporarily halted.The combination of the undulated fiber architecture and matrix phase contributed to the highest G IIc levels measured at room temperature.However, this case also experienced the most significant G IIc reduction under cryogenic temperatures, which may be attributed to resin embrittlement.The fractographic analyses suggested that nanoscale matrix cracking, fiber shear fracturing and fiber/matrix debonding events collectively contributed the ultimate fracture.

•
The T resin system, with the highest mixture viscosity, likewise offered good wetting.Parallel to the VI resin system, the G IIc values of the T resin system remained stable under cryogenic conditions.The fractographic analyses indicated better fiber/matrix adhesion under cryogenic temperatures.This fact simultaneously allowed for an equal contribution of the matrix phase during mode-II crack propagation.Notably, typical shear hackles inherent to mode-II crack propagation were observed on the fracture surfaces of the samples tested under cryogenic conditions.This situation was not present in any of the other resins due to the fact that the damage progression under cryogenic temperatures was always initiated by fiber/matrix debonding followed by other failure events.
In summary, the reported results and conclusions underline the key considerations related to material selection and manufacturing-related issues that influence the performance of filament-wound laminates under cryogenic conditions.Although such problems may be overcome by the use of more advanced AFP techniques, the presented interpretation of the delamination behavior of wet-wound laminates may provide guidance for further material development.With such efforts, pressure vessels with limited operational lifetimes (e.g., sounding rockets) could still be efficiently manufactured through conventional filament winding, a method that remains a dominant and cost-effective manufacturing approach in terms of volume.

Figure 5 .
Figure 5. Representative flexural stress-strain curves of three-point bending tests of neat epoxy.

Figure 5 .
Figure 5. Representative flexural stress-strain curves of three-point bending tests of neat epoxy.

Figure 8 .
Figure 8. Side views of tested specimens (orange arrows indicate crack propagation direction): (A) CFRP-CC resin at RT with linear crack propagation; (B) CFRP-CC resin at LN2 with zig-zag crack propagation.

Figure 8 .
Figure 8. Side views of tested specimens (orange arrows indicate crack propagation direction): (A) CFRP-CC resin at RT with linear crack propagation; (B) CFRP-CC resin at LN2 with zig-zag crack propagation.

Figure 8 .
Figure 8. Side views of tested specimens (orange arrows indicate crack propagation direction CFRP-CC resin at RT with linear crack propagation; (B) CFRP-CC resin at LN2 with zig-zag propagation.

Figure 8 .
Figure 8. Side views of tested specimens (orange arrows indicate crack propagation direction): (A) CFRP-CC resin at RT with linear crack propagation; (B) CFRP-CC resin at LN2 with zig-zag crack propagation.

Table 1 .
Epoxy systems, pot life, curing cycle and viscosity properties.

Table 1 .
Epoxy systems, pot life, curing cycle and viscosity properties.

Epoxy System Pot Life at 23 • C (hours) Nominal Curing Cycle Resin Viscosity (cPs) Hardener Viscosity (cPs) Measured T g * ( • C) Data Sheet T g ( • C)
* Measurements were obtained from manufactured ENF samples.

Table 2 .
Three-point bending and ENF results of epoxy specimens at room and LN2 temperatures.