Fatigue damage propagation and creep behavior on sisal/epoxy composites

The lack of knowledge about the behavior under creep and fatigue limits the use of polymeric composites reinforced with natural fibers. Thus, this work assessed the behavior of epoxy composites reinforced with sisal fibers under tensile, fatigue, and creep tests. Also, thermogravimetry and scanning electron microscopy assessed the sodium hydroxide (NaOH) treatment efficiency in sisal fibers. Further, differential scanning calorimetry determined the degree of cure of the composites, and stereomicroscopy allowed the evaluation of the surface’s fracture. As a result, the tensile strength of the composite was 1.4 times the value of neat epoxy resin after 100,000 cycles in the fatigue test. Moreover, when loaded with 20% of the maximum tensile strength, it is estimated that the composite resists 200,000 h without rupturing by creep. To conclude, the efficient adhesion between sisal fibers and epoxy obtained by NaOH treatment allowed good mechanical behavior to the epoxy composite.


Introduction
Although polymer matrix composites reinforced with carbon, glass, or Kevlar fibers have excellent mechanical properties associated with low weight, the use of these synthetic fibers has adverse effects on the environment since their production requires a large amount of energy [1] . Polymeric composites reinforced with natural fibers from plants are attractive alternatives.
There is a wide variety of plants used as reinforcement in polymeric matrices, for example, sisal, jute, palm, curauá, and others. The fibers used as reinforcement come from different plants parts such as the trunk, leaf, fruit branches, and stem [2] . And even if the adhesion between natural fibers and polymers is low, it can be improved with chemical treatments, removing fiber amorphous and nonpolar components and impurities that repel polymers [3] . The treatment with sodium hydroxide (NaOH) is simple and widely used. Palm fruit branch fibers show about 180% greater tensile strength when treated with a NaOH solution, reaching about 290 MPa [2] . In epoxy composites, natural curauá fibers treated with NaOH improve tensile, flexural, and impact strength by 24%, 44%, and 47%, respectively, comparing untreated samples [4] . This sums up the potential of natural fibers for composite applications.
Recently, several studies have reported the mechanical performance of polymeric composites reinforced with natural fibers, but almost all reports use only tensile, flexural, and impact tests. However, more detailed studies on the behavior of polymeric composites reinforced with natural fibers under cyclic loads or quasi-static loads are still required for structural applications. Since composites reinforced with natural fibers are applied for non-structural purposes, such as panels, engine covers, and internal parts of cars and aircraft [5] .
Fatigue is the leading cause of failure for failures in aircraft components [6] , vehicle storage tanks [7] , machine and rail bearings [8] , among other engineering applications subjected to cyclic loads. Furthermore, the variability in fatigue properties of the same type of composite due to differences in composition and structure properties requires an understanding of the fatigue mechanism in these materials [9] . Experimental studies are still needed to understand damage propagation and failure modes in composites, and in this sense, the analysis of material stiffness during fatigue tests is used [10] . However, the material's stiffness, which should be reduced with increasing damage to the composites, increases in some cases [11] . Therefore, fatigue tests controlled by strain instead of stress have a consistent behavior with the stiffness of composites, which is reduced until the material breaks [12] .
Furthermore, the study of the creep behavior of polymeric composites is also crucial for safe structural design because it presents a time-dependent degradation in stiffness due to strains which can lead to material breakage [13] . The effects of creep are long-term, and research aims to predict the strength of polymer composites through accelerated testing. However, in general, the tests use temperature variation to accelerate the creep effect on the material, which makes it challenging to obtain results [14] . Therefore, the stepped isostress method (SSM) is a form of life prediction in which the temperature is constant, and what varies is the stress, which is systematically applied to the material [15] . SSM effectively tests amorphous or crystalline polymers and reduces test time by 70 times to predict material life [16] .
Although there are reports of epoxy composites reinforced with synthetic fibers under fatigue and creep, few works have studied the performance of composites reinforced with natural fibers. However, natural fibers are promising as reinforcement. For example, hybrid composites of epoxy resin reinforced with fiberglass and kenaf (natural fibers) have fatigue life close to that of material reinforced with only fiberglass [17] . For creep, the stepped isostress method has not yet been reported for composites reinforced with natural fibers. Still, it is promising in predicting the life of polymer composites reinforced with fibers [18] .
Therefore, this innovative work intends to promote natural fibers, especially sisal fibers, and provide valuable and new information about the mechanical performance of sisal fibers/epoxy composites under tension-tension fatigue controlled by strain and creep loading tests using the stepped isostress method.

Materials
Sisal fibers were supplied by Sisalsul Industry and Commerce LTD (Bahia, Brazil) and treated using sodium hydroxide (NaOH) 97% pure (Greentec) aqueous solution. The polymer matrix is constituted by epoxy resin AR 260, Di-Glycidyl Ether of Bisphenol A (DGEBA), and hardener agent AH 260, Triethylenetetramine (TETA), both supplied by E-composites Commerce of Composite Materials LTD (Rio de Janeiro, Brazil).

Prepregs and laminated composites obtaining
A superficial treatment was carried out for the sisal fibers soaking them in a 5% (w/v) sodium hydroxide (NaOH) solution with a ratio of 10:1 (solution: fiber). Then, the mix (fibers and sodium hydroxide solution) was kept at 80°C for 2 h under constant stirring. This condition followed the best result from a recent study developed by Teixeira et al. [19] . After, the fibers were washed with distilled water until pH 7. Finally, the fibers were dried at room temperature (~ 25°C) for 96 h and subsequently in an oven (32 °C) for 24 h [20] .
The epoxy prepregs were obtained with 15 wt.% of fibers by the hand lay-up method [21] . This fiber content was chosen based on a previous work, where the authors studied the influence of cure agent, treatment, and fiber content by using statistical analysis [22] . Sisal fibers were placed unidirectionally in a steel mold of 200 x 220 mm. The epoxy system was prepared by mixing the resin and hardener in a ratio of 100:21 (epoxy resin: hardener agent), as recommended by the supplier. Then, the resin was spread on the fibers with a roller. The prepreg was kept at room temperature until it reached stage B, also known as the gel stage, and then kept under refrigeration at -18ºC.
Laminated composites were prepared with three unidirectional prepreg layers. The layers were placed in a hydraulic press under a load of 0.5 tons for 1 h. Then, the material was cured at room temperature for 24 h. The postcure was carried out in the oven at 60ºC for 12 h. A neat epoxy resin plate was prepared for a comparison with the composite materials.

Thermal characterization
Untreated and NaOH-treated sisal fibers, neat uncured epoxy, untreated, and NaOH-treated sisal 15 wt.%/epoxy prepregs were analyzed in a simultaneous TGA-DSC thermal analyzer (SDT Q600, TA Instruments, USA). The samples weighing 10 ± 1 mg were deposited on an alumina pan. The analysis was from 20 to 600°C with a heating rate of 5°C/min under a nitrogen atmosphere at a 50 mL/min flow rate.
TGA was used to assess the surface treatment effect on sisal fibers, and DSC to calculate the degree of cure in the composites, α, determined by Equation 1: where ∆Hp is the partial enthalpy (integral area of the prepregs cure peaks), and ∆Ht is the total enthalpy of the cure reaction (integral area of the uncured epoxy resin cure peak).

Microscopic analysis
The surface morphology of the treated and untreated sisal fibers was analyzed by scanning electron microscopy (SEM) (TM-4000Plus, Hitachi, Japan), using 15 kV voltage and 400 × magnification. After the rupture in tensile tests, the fracture region of neat epoxy and NaOH-sisal 15 wt.%/epoxy composite specimens were analyzed under a stereomicroscope (DFC-700T, Leica, Germany). The investigated samples were tested before fatigue tests, and 31.5 x magnification and incident light were used.

Tension-tension fatigue
Tension-tension fatigue tests were carried out according to the American Society for Testing and Materials Standard (ASTM), ASTM D3479 [23] . The test was performed on a universal testing machine (8801, Instron, USA) with 100 kN load cell. The test specimen dimensions were 140 × 25 × 3 mm. Tabs of epoxy reinforced with glass fibers of 40 x 25 mm were attached during the composites manufacturing. The fatigue tests were carried out with a frequency of 5 Hz, strain ratio R = 0.1, and controlled by strain. The cyclic load was applied in a sinusoidal waveform with three maximum strain levels (ε max ), 0.14, 0.10, and 0.02%. For each strain level, one NaOH-sisal 15 wt.%/epoxy composite specimen and two neat epoxy specimens were analyzed. The pressure on the grips was 15 bar.
The evolution of the dynamic stiffness of the materials during the fatigue tests, E f , is calculated by the slope of the secant that links the lowest and the highest peaks in the stress-strain hysteresis loop of each loading cycle. Then it is normalized by the dynamic stiffness of the first cycle, E. The damage propagation during fatigue tests, D, was calculated by D = 1 -(E f /E) [24] , where E f and E are the same mentioned in dynamic stiffness. Dynamic stiffness values and accumulated damage were plotted as a function of loading cycles. The x-axis, N/Nf, presented the number of cycles, N, divided by the maximum number of loading cycles applied to the materials, Nf.

Tensile tests
Tensile tests were performed according to ASTM D3039 [25] , in two different steps: 1) Tensile before fatigue tests: To measure the maximum tensile strength and determine the maximum stress levels used in creep tests. Five specimens of NaOH-sisal 15 wt.%/epoxy composites specimens (200 × 20 × 3 mm) were tested before fatigue tests using a test machine (Electropuls E10000, Instron, USA) equipped with 10 kN load cell at a speed 2 mm/min. 2) Tensile after fatigue tests: To determine the remaining tensile strength and properties on epoxy composites reinforced with sisal fibers and neat epoxy samples that reached the run-out on fatigue tests, defined as 100,000 cycles. For each strain level in fatigue tests, one NaOH-sisal 15 wt.%/epoxy composite specimen and two neat epoxy specimens were tested. The specimen size was 140 × 25 × 3 mm, manufactured with tabs, and the tests were performed using a test machine (8801, Instron, USA) equipped with a 100 kN load cell at a speed of 0.3 mm/min. In these tests, the presence of tabs influenced the premature rupture of the material, so the execution speed was lower than that of the previous tensile test, used to determine the stress levels of the creep test.

Creep
Creep tests were performed using a test machine (ElectroPuls E10000, Instron, USA) equipped with a 10 kN load cell and mechanical grips. The NaOH-sisal 15 wt.%/ epoxy composite rectangular specimens (200 × 20 × 3 mm) were manufactured without tabs and tested before fatigue tests. The creep test was performed according to the Stepped Isostress Method (SSM) [15] . Different loads are applied to a specimen during the same test at 25oC. The loads were changed at a staggering rate: 20% (12.
where is the strain, is the time, and are materials constants, 0 is the point where the i-th fitting curve crosses the abscissa. Second, the rescaling shift, each step of the curve is discounted by 0 . This shifts each step to the left. Finally, in horizontal shift, the curve points are divided by a displacement factor, as obtained by Equation 3: where C 0 is the coefficient C I of the first level of strain. The creep curves are plotted with these strains vs the log time, estimating the strain after long periods.

Effect of chemical treatment on sisal fibers' morphology
The structural morphology of sisal fibers for untreated and NaOH-treated sisal fibers can be seen in SEM images in Figure 1. The chemical treatment removed part of the hemicellulose layers on the surface of sisal fibers, presented in Figure 1a, facilitating the adhesion between the fibers and the polymeric matrix. Thus, bundles of cellulose fibrils, a component with a high percentage of crystallinity and mechanical resistance, are exposed, as shown in Figure 1b. The adhesion between the reinforcement and the polymeric matrix could improve due to the significant surface roughness of fibers and the effective contact area [26,27] . In previous work, the treatment conditions applied to natural fibers (5% solution, for 2 h at 80°C) improved its crystallinity. Also, the hemicellulose was removed without losses in cellulose content [19] . Figures 2a and b show the TG and DTG curves for samples of untreated and NaOH-treated sisal fibers, respectively. The degradation steps are indicated in the TG curves. The first step, from 30-105 °C for untreated and 30-150 °C for treated fibers, is associated with loss of moisture absorbed by the fibers (about 5% weight loss). In treated fibers, the range of temperature is higher, as the treatment increases its moisture absorption [28] . In the second step, the degradation of fiber components begins at 160°C for the untreated fiber and refers to the degradation of hemicellulose (about 20% weight loss) followed by cellulose degradation (about 40% loss weight) in the third step.

Thermal characterization
The untreated fibers showed two degradation peaks at 280 °C and 340 °C (DTG curves), corresponding to hemicellulose and cellulose degradation, respectively [29] . The treated fibers, however, have a single cellulose degradation stage (about 60% weight loss), which starts at 230 °C, indicating an increase in the thermal stability of the fibers by 50 °C. The increase in the thermal stability of the fibers happens because the alkali treatment with NaOH removes hemicellulose, which has low thermal stability [30] .
In previous work, the components losses are followed by chemical characterization and Fourier transform infrared spectroscopy, confirming the effects of the treatments on hemicellulose and cellulose thermal degradation. The degradation steps (TG curves), for untreated (fourth step) and treated fibers (third step) are linked to lignin degradation (about 12% weight loss) [29] . A residue of 23% in weight of the carbonized fibers was observed. In our previous work with the same epoxy and fiber content, 15 wt.%, also presented high thermal stability [31] .
Through the DSC analyzes, it is possible to observe the exothermic peaks of the prepregs. Figure 2c shows the DSC curves for sisal (NaOH treated and untreated) 15 wt.%/epoxy prepregs compared to uncured neat epoxy resin. The first exothermic peak, around 40 °C, is related to the glass transition temperature of epoxy and agrees with what is reported in the literature [32] . The exothermic peaks around 100°C are related to the curing reactions of the crosslinks in the epoxy resin [21] . Enthalpy is the energy released by the system in the formation of crosslinks, calculated by the area under the peak curing temperature in DSC curves delimited by the baselines (dashed lines). The degree of cure of prepregs is inverse to the enthalpy value. The higher degree of cure is related to the more significant crosslinking number in the epoxy resin matrix and the elevated mechanical resistance of the composite [33] . The high crosslink content in epoxy resins reduces the strain rate of the material, increases its recovery capacity, and affects the fatigue strength of epoxy resin-based materials [34] .
The degree of cure of NaOH-sisal/epoxy prepreg was 51.55%, 1.7 times higher than that of untreated sisal/epoxy prepregs (30.36%). The presence of hydroxyl groups, OH, increases the number of cross-links established between the resin and the curing agent [35] . The chemical treatment exposes the cellulose chains of the natural fibers and their OH groups, hence the higher concentration of cross-links in the treated sisal-reinforced prepreg. In addition, the treatment on sisal fibers allows a higher crosslinked interphase due to the intermediate module between reinforcement and polymer [36] . Undoubtedly, it influenced the better fatigue and tensile strength for NaOH-sisal 15 wt.%/epoxy composites compared to neat epoxy, as discussed in the following section.

Laminated composites characterization
Figures 3a and 3b shows the stiffness loss of neat epoxy and NaOH-sisal 15 wt.%/epoxy composite during the fatigue tests, respectively. For all materials, a dynamic stiffness loss during the fatigue tests is observed. This behavior is as expected, presenting the stiffness reduction [37] . When comparing the materials, at 0.14% of maximum strain, the stiffness loss was lower, about 3% for the composite. However, the composite was more resistant to fatigue than neat epoxy because the fibers transfer the applied load and provide a barrier preventing crack propagation in composites [38] . The stiffness is reduced because only part of the energy released in the loading is recovered [39] .  Figures 3c and 3d show the accumulated damage during fatigue tests of neat epoxy and NaOH-sisal 15 wt.%/epoxy composite, respectively. The presented damage curve is expected because damage to epoxy composites is inversely proportional to degradation in stiffness [40] . This is also reflected in a consistent damage curve representation. Strain control prevents the false increase in dynamic stiffness by residual strains observed for some epoxy composites reinforced with natural fibers in fatigue tests [12] .
After the run-out in fatigue tests, the specimens were tested under tensile. The properties of neat epoxy and NaOH-sisal 15 wt.%/epoxy composite are shown in Table 1, where ε max is the maximum strain applied during fatigue tests cycles. The addition of treated sisal fibers to the epoxy resin increased the tensile strength by about 40%, achieving 47 MPa against 34 MPa of neat epoxy. The found values are as expected for tensile strength of epoxy composites reinforced with aligned sisal fibers, 51.5 ± 5.6 MPa [41] , and superior to epoxy composites reinforced with jute fabrics, 40 MPa of tensile strength [42] . In general, composites reinforced with unidirectional fibers have better tensile properties than those reinforced with woven fabrics due to the undulations in the yarn which form the fabric structure [43] . In addition, the NaOH-sisal 15 wt.%/epoxy composite presented superior tensile strength comparing epoxy composites reinforced with 10 wt.% of untreated sisal fibers, by 50% [44] .
In addition, the tested specimens broke at the base of tabs. Silva et al. [21] , reported a 40% higher tensile strength for NaOH-treated sisal fibers (15 wt.%) epoxy composites in samples manufactured without tabs and not subjected to cycles, showing the negative influence of these conditions on the tensile properties of the material. The presence of tabs can cause the premature rupture of the specimens due to the increase in the concentration of stresses caused by geometric discontinuities among the parts [45] .
The stress level used in creep tests was also determined by testing NaOH-sisal 15 wt.%/epoxy under tensile. The samples presented 64.86 ± 12.20 MPa on tensile strength and 5.04 ± 0.49 GPa on Young's Modulus, as shown in Table 1. This tensile strength is similar to the value given for epoxy composites reinforced with glass fibers [46] .
Stereomicroscopy was used to analyze the morphology of surface fracture of the neat epoxy and NaOH-sisal 15 wt.%/ epoxy composites materials after the tensile tests, as shown in Figure 4. The specimens were manufactured without tabs and tested before fatigue tests. The micrograph of neat epoxy resin, Figure 4a, shows the cracks propagation without material flow and the absence of voids and defects in the epoxy resin. This region is flat, not reflective, has riverbed patterns, and demonstrates its brittle characteristics [47] . The river lines with a smooth surface are due to the excellent bonding of the epoxy resin [48] . Figure 4b shows the NaOH-sisal 15 wt.%/epoxy composite, and it is possible to observe the fiber rupture. Compared with neat epoxy resin, the best mechanical performance is linked to the efficient adhesion between sisal fiber and epoxy  resin. In addition, voids are observed, defects caused by the release of volatiles and trapping of the gases generated in the resin curing reactions during the manufacturing process.
The presence of voids in the epoxy resin reduces its tensile properties, but resin degassing with an ultrasonic bath, as was done in this work, improves its tensile strength [49] . Also, in recent literature, alkali-pretreatment jute fibers (natural fibers) reduce the number of epoxy composites voids and increase their tensile strength [50] . Figure 5a shows the strain vs time curve obtained for NaOH-sisal/epoxy composites after the creep test. It is possible to observe that strain increases with each step's time (20,25,30,35, and 40% of UTS). Figures 5b and 5c show the vertical and rescaling shifts, respectively. Finally, Figure 5d shows the Stepped Isostress Method (SSM) curve and, according to this, it is valid to state that with 20% UTS the material resists for 200,000 hours without undergoing a creep break. This is due to the insertion of natural fibers in the composite, as the incorporation of fibers can successfully reduce creep since the tension transmitted to the matrix can be transferred to the fibers before the material breaks [51] .

Conclusions
In conclusion, the NaOH treatment removed hemicellulose and increased its thermal stability. Also, the NaOH-sisal 15wt.%/epoxy composites present the best strength under fatigue and tensile compared to neat epoxy. This behavior can be explained by adding unidirectional sisal fibers with high tensile strength when the load is applied on its longitudinal axis. Due to the NaOH treatment of the sisal fibers, the adhesion between the fiber and the epoxy matrix was also efficient in distributing loads applied to the composite. The more significant crosslinks number was observed for the NaOH-sisal 15 wt.%/epoxy compared to the neat epoxy. The higher crosslink content also contributed to the higher fatigue and tensile strength of the material. These characteristics were also significant in the resistance of the composite against creep.