Exploration of Effect of Thermal Cycling Treatment on Mechanical and Tribological Performance of Polyimide Composites

Abstract Polyimide (PI) and PI composites reinforced with glass fiber (GF), and aramid fiber (AF) were fabricated with hot-pressing technology. Then polyimide and its composites were treated with a thermal cycling (TC) condition from –50 °C to 150 °C with 300 cycles continuously. Compared with the samples without TC treatment, the fracture morphologies, mechanics, and tribological properties of the samples were studied. The results showed that, based on the fractographic morphologies, the phenomenon of fiber stripping could be obviously observed in the PI composite materials after TC treatment. Compared with before TC treatment, the friction coefficients (FC) and wear rates (WR) of PI and PI composites showed different variations, respectively. In conclusion, TC treatment had a great influence on the morphologies of the worn surface and the composition of transfer film. As the shear force increased, the elemental analysis demonstrated that GF was extruded into the groove.


Introduction
Polymers and polymer composites have attracted huge attention in polymer tribological applications due to their selflubrication property and lightweight characteristic. Owing to the excellent mechanical performance under high temperature and their outstanding chemical and heat resistance properties, polyimide (PI) materials performs well in many high-tech applications such as aerospace vehicles, engines and manufacturing, and others (1).
Due to the synergistic effect of its great modulus, toughness, and high glass-transition temperature (T g ), PI is considered a high-performance polymer. Thus, tribological studies of PI or PI composites have been carried out under dry sliding, lubrication, and finite temperature conditions. In order to improve the load capacity and wear resistance of the composites, different kinds of fiber are used under different sliding conditions (1,2). The excellent interfacial bonding can greatly transfer stress from matrix to fiber, prevent the deformation of matrix, and improve ultimate load performance (3,4). Especially, the interfacial interaction between fiber and matrix becomes a crucial factor to evaluate the performance of the composite in long-term service during the sliding process. Therefore, when moving parts shift in and out of the Earth shadow in space, the influence of temperature becomes a key factor affecting the tribological and mechanical properties. The composites are inevitably subjected to a thermal cycling (TC) condition (from low to high temperature) and a deterioration in mechanical and tribological properties caused by the buildup of stress at the interface (5-7). Thus, the variations of mechanical and tribological properties must be precisely evaluated and investigated to enable the safe and reliable applications of polymer composites.
Previous research studies have found that when composites are treated at high temperature, the difference in expansion coefficient will induce the stress concentration at the interface between the polymer matrix and the inorganic fillers, which finally leads to the deterioration of mechanical (8)(9)(10) and tribological properties of composites. Meanwhile, during sliding process, the extracted fibers are pulled out by repeated shear force, which further reduces the antiwear properties of PI composites. To date, many valuable investigations have been conducted on the tribological properties of PI or its composites, among which the effects at high temperature of fiber (11), lubricants (12)(13)(14)(15), and polymer structures (16,17) have been discussed. The results show that the deterioration of mechanical properties and tribological properties of materials at high temperature can be mitigated by appropriate fillers. By contrast, in a cryogenic environment, the radial and axial contraction of fibers will increase the thermal stresses at the interface. Microcracking then appears due to the accumulation of the stress at the interface between fiber and matrix, when the composite is under extreme cryogenic temperatures such as below -150 C (9-11).
Compared with the softening of the matrix at high temperature and the decrease of adhesion to the fiber interface, shrinkage of the matrix can improve the tensile strength and modulus at low temperature (18)(19)(20)(21). Further, the two aspects of friction during sliding (adhesion aspect and real contact area) will compete with each other, which complicates the tribological performance of different polymer systems (22,23). However, regardless of the effect of environment temperature and contact mode of the friction pairs, the elongation of the fiber and the microcracks at the interface between fiber and matrix will accelerate the deterioration of the tribological properties. The single fiber as the third body leads to severe abrasive wear and microcrack propagation during a sliding process .
In addition, tribological tests of polymer composites have been carried out under certain temperatures and certain loads and rotation speeds (14,28,29). The potential influence of temperature has not yet been fully explored. Meanwhile, as the service conditions become increasingly complex, exploring the effect of the practical application environment on the properties of polymer composite becomes more and more important. However, to the best of our knowledge, the mechanism and performance of friction and wear behavior of PI composites are rarely involved in discussion of the effect of cycling temperature. Therefore, it is worthwhile to explore the tribological properties of PI composites with thermal cycling (TC) treatment.
For this article, a commercial PI named YS-20 is adopted as the polymer matrix. PI composites containing glass fiber (GF) and aramid fiber (AF) were fabricated using hot-pressing technology. GF was treated with the silane coupling agent KH550 before use. The mechanical performance, tribological properties, and microstructure of PI composites before and after TC treatment are analyzed. The flexural mechanisms of TC treatment on tribological properties are discussed.

Material
The polyimide powder used is YS-20 grade commercial polyimide, purchased from Shanghai Research Institute of Synthetic Resins, with YS-20 particle size 75 lm and density 1.4 g/cm 3 . YS-20 is a thermoplastic polymer, and it has high temperature resistance and good thermal stability. Glass fiber (GF) and aramid fiber (AF) were purchased from Jushi Group Co., Ltd. The diameter of GF is 10 lm and the aspect ratio is 10:1. The diameter of AF is 15 lm and the aspect ratio is 3.5:1. The composite material with fiber weight of 15% was prepared by mechanical mixing and ultrasonic dispersing (the power of the ultrasonic machine is 480 W, for the CJ080ST model from Shenzhen Chaojie Technology Industrial Co., Ltd) in industrial ethanol before hot-press molding (since the fibers are short fibers, the cut fibers can be ignored). PI and fiber were pressed hot at 380 C at a pressure of 20 MPa. This process required cyclic ventilation for 60 min. Then the samples were cut into the required size for tribological tests (4 mm Â 4 mm Â 12 mm), mechanical tests (80 mm Â 10 mm Â 4 mm), and dynamic mechanical tests (60 mm Â 10 mm Â 5 mm). Formulations and density of the PI and PI composites are shown in Table 1.

Methods
The TC treatment for PI and its composites was conducted in low/high temperature testing chambers. The program is set up as shown in Figure S1, with a cycle of 9 h and a total of 300 consecutive cycles. First, it was treated at 150 C for 1 h, then cooled to -50 C for 3.5 h, kept at this temperature for 1 h, and then heated to 150 C for 3.5 h. This process was repeated 300 times. During the heating process, the heating rate was approximately 1 C/min. The cooling rate was about 1 C/min at the cooling process. The morphologies of cross-ection and worn surface were characterized using scanning electron microscopy (SEM) and optical microscopy (OM). Thermal gravimetric analysis (TGA) was carried out on a thermal analysis equipment in nitrogen. The dynamic thermomechanical characteristics of PI composites were determined using a dynamic thermomechanical analyzer (DMA8000, PerkinElmer, USA). The relevant tests were completed in the double-cantilever three-point bending mode with a test frequency of 1 Hz, and the composite specimens were heated at 10 C/min and a maximum temperature of 300 C. The specimen size was 10 mm Â 4 mm Â 60 mm. The storage modulus (E') and loss modulus (E") were obtained as a function of scanning temperature. Tribological properties of the PI-based composites were tested at room temperature against a GCrl5 plate (R a ¼ 2.5 lm) with the face-to-face contact mode on a tribometer (Jinan Yihua Tribology Testing Technology Co., Ltd). The friction tests were carried using pin on disc friction; the size of the pin was 4 mm Â 4 mm Â 12 mm, mentioned earlier.
The wear rate (WR) is found by calculating the ratio of wear mass to material density, operating distance, and applied load. Before each test, each polymer sample was polished to ensure face-to-face contact with the steel plate. Before the test, absorbent cottons dipped in acetone and anhydrous ethanol was used to clean all friction pairs. The morphology of the steel plate (R a ¼ 2.5 lm) is shown in Fig. S2. The load selected for the friction test is 1, 2, 3, 4, and 5 MPa, and the sliding speed selected is 1, 2, and 3 m/s. The total length of the friction test was 3600 m, and the time changed with the sliding speed.

TG analysis
To explore the impact of TC on heat resistance properties of PI and PI composites, weight loss curves as a functional of temperature are plotted in Fig. 1. Consistent with our speculation, the PI, PI with aramid fiber (PIA), and PI with glass fiber (PIG) exhibited decomposition temperature at 553 C, 560 C, and 560 C, respectively, before TC treatment, with a tiny variation after treatment. Similarly, the residual weight retention exhibited slight change. The highest residual weight retention rate was observed in PIG, due to the ultrahigh heat resistance of GF compared to AF and pure PI. For PI and PIA, the residual weight retention after TC treatment showed almost no difference from before treatment.

FT-IR analysis
In order to analyze the effect of TC treatment on sample, the Fourier-transform infrared (FT-IR) spectra of PI powder and PI sample before and after TC treatment are plotted in Fig. 2. Obviously, the PI without TC treatment is not much different from the PI after TC treatment. The characteristic peaks of PI with treatment revealed a similar value with PI powder. The relaxation of the polymer chain could be demonstrated by the stress variation as a function of temperature for PI sample in Fig. S3. Clearly, compared with before TC treatment, PI after TC treatment showed lower stress. The internal stress caused by the interaction of polymer chains could also be demonstrated by the absorption of the ultraviolet (U)-visible spectra in Fig. S4. The higher the k value of PI before TC treatment, the stronger is the physical interaction, resulting in more physical cross-linking points in the polymer network. Meanwhile, the DSC curves in Fig. S5 also showed the unchanged behavior of the degree of crystallization for PI and PI composites before and after TC treatment.

Bending test
The mechanical properties of PI and PI composites are presented in Fig. 3. Without considering the TC treatment, the fiber-reinforced PI showed higher flexural strength and modulus. All samples presented higher flexural strength compared with that before TC treatment. Except for pure PI, the bending modulus of the composites increased after TC treatment. This is consistent with the variation law of storage modulus obtained from the dynamic mechanical tests. This could be attributed to the cyclization reaction of   imide (33)(34)(35) and the local relaxation of the polymer chain during the TC treatment process (9). The fractographic morphologies of PI and PI composites are presented in Fig. 4. As can be seen from the image, there existed little difference in pure PI before and after TC treatment. The morphology of PI presents the ductile fracture characteristics before and after TC in Fig. 4b. Distinctive microflow and river patterns were clearly observed for pure PI (17), which indicates the higher ductility and lower modulus after thermal treatment. For PIA, the fractographic morphology showed the strong binding between the fiber and the matrix before TC treatment (Fig. 4c), and parental stripping between the fiber and matrix (Fig. 4d) after TC treatment. Compared with the close interaction between fiber and matrix in Fig. 4e, after TC treatment, the visible voids generated by glass fiber pulling out of PIG were obvious (Fig. 4f). Meanwhile, some obvious resin fragments on the fiber surface before TC treatment can be found, compared with the bare surface of the fiber after TC treatment.

DMA analysis
Dynamic mechanical properties of PI and composites are plotted in Fig. 5. It is interesting that the species presented two distinct transitions near 232 C and 260 C, except for the relaxation at 105-110 C in the tand curves, which were not observed in our previous research (32). The transition at low temperature diminished or shifted to high temperature for PI and PIG, and disappeared for PIA after TC treatment, respectively. Meanwhile, the beginning storage modulus increased slightly for PIA and PIG, which was consistent with the bending test results. According to the results of polymer mechanical properties, thermoplastic polymers with rigid structures have a higher storage modulus and glass transition temperature (33)(34)(35). Considering the relaxation of the polymer, it can be concluded that the cyclic reaction of imide occurs with the molecular chain relaxation during TC treatment process. Finally, the equilibrium of the polymer chain relaxation during the cyclization of imide leads to the variation of mechanical properties of PI and PI composites. Moreover, the different transition temperatures of the loss modulus before and after TC treatment demonstrate the physical state of the polymer chain before and after TC treatment. At the same time, the loss modulus curves also presented a convex transition peak before TC treatment at 260 C, and after TC treatment the transition peak of the loss modulus became sharp. The main transition peak changed from 232 C to 260 C by TC treatment. Thus, the polymer chain relaxation and cyclization imide can be identified based on the comprehensive analysis just described. Figure 6 exhibits the tribological results of PI composites at 2 m/s with the load increasing. Overall, tribological properties of the samples presented different sceneries before and after TC treatment. For PI without fillers, the tribological test can only be performed at 1 MPa, 2 m/s, before TC treatment due to friction vibration. Similarly, the test for PIA without TC treatment failed when pressure exceeded 2 MPa. In this case, failure means friction failure, which is judged by the friction coefficient exceeding 1 or the friction material breaking during friction. By contrast, PIG presented superior tribological performance. All tests could be successfully conducted under the given conditions of PIG without TC treatment. Friction coefficients (FC) and wear rates (WR) of PIG decreased as the load increasing. Compared with the result before treatment, the tribological tests of PI    mm 3 Á(NÁm) À1 to 0.966 Â 10 À6 mm 3 Á(NÁm) À1 , from 0.639 Â 10 À6 mm 3 Á(NÁm) À1 to 0.842 Â 10 À6 mm 3 Á(NÁm) À1 , and from 0.598 Â 10 À6 mm 3 Á(NÁm) À1 to 0.657 Â 10 À6 mm 3 Á(NÁm) À1 , respectively. Unlike for PIG, the WR of PIA increased with the increase of load, but FC decreased with the increase of load. Similar scenarios can also be found for  PI after treatment. The distinctive variation can be attributed to the difference in mechanical properties as shown in Fig. 6. PI reinforced with glass fiber presented slightly higher bending strength and better ductile property, which implied the superior load capacity during the sliding test. The tribological properties of samples are plotted in Fig. 7 before and after TC treatment under 2 MPa with velocity rising from 1 m/s to 3 m/s. From the figures, it is obvious that FC and WR decreased with increasing velocity for PI, PIA, and PIG. The tribological test for PI failed when velocity exceeded 1 m/s before TC treatment and 2 m/s after TC treatment. This is similar to the variation in Fig. 6, where WR of PI and PIA increased with increasing velocity, regardless of the effect of TC treatment. Interestingly, PI and PIA can tolerate increasingly harsh conditions after TC treatment, compared with their performance before TC treatment. Similar to the variation in Fig. 6, PIG proceeded well under the given conditions. FC and WR of PIG decreased with increasing velocity before and after TC treatment. With the velocity increasing from 1 m/s to 3 m/s, the FC increased slightly from 0.372 to 0.374, from 0.264 to 0.327, and from 0.236 to 0.296, respectively. WR of PIG increased from 1.123 Â 10 À6 mm 3 Á(NÁm) À1 to 1.345 Â 10 À6 mm 3 Á(NÁm) À1 , from 0.92Â 10 À6 mm 3 Á(NÁm) À1 to 1.258 Â 10 À6 mm 3 Á(NÁm) À1 , and from 0.834 Â 10 À6 mm 3 Á(NÁm) À1 to 1.142 Â 10 À6 mm 3 Á(NÁm) À1 , respectively. This may be due to the accumulation of friction heat during sliding. Due to the softening of polymer at the sliding interface, the fibers were pulled out easily, which resulted in abrasive friction during sliding process at the interface (29)(30)(31). This can be demonstrated in the worn surface analysis section. Similar to the FC, WR also increased due to the scratching action of inorganic filler at the sliding interface. This can be confirmed by the worn surface morphology of samples in Fig. 8 and Fig. 9.

Tribological properties
The worn surfaces of PI and PIA before and after TC treatment are presented in Fig. 8. From the SEM images, it is apparent that there is little difference in the image before and after TC treatment. This means the effect of TC treatment can be ignored under relatively lower load and velocity conditions. However, from the images in Fig. 8c, d, e, and f, a slight scratch can be clearly found on the worn surface of samples with the increasing PV value. Under a higher PV condition, the deformation and accumulation of friction heat provided an easy shear property to the polymer at the sliding surface, which lead to peeling off the large piece and a scrape trace on the worn surface. Under the action of shear force, the fragments stripped from the surface are repeatedly extruded to form a transfer film on the corresponding surface.
Compared with PI and PIA, the PIG worn surface was significantly different before and after TC treatment at 2 MPa, 2 m/s, and 5 MPa, 2 m/s. At 2 Mpa and 2 m/s, the TC-treated PIG worn surface (see Fig. 9b, e) showed obvious grooves and pits compared with the TC-untreated PIG worn surface (see Fig. 9a). The suggestion that the fibers were pulled out can be verified by the deep groove in Fig.  9e. However, there is no obvious debonding between fiber and matrix in Fig. 9a. With the load increasing, debonding appeared due to the severe shear force before TC treatment (see Fig. 9c), under 5MPa, 2 m/s. Meanwhile, the accumulation of debris around the fiber caused by the severe shear can be easily distinguished on the worn surface under 5 MPa, 2 m/s (see Fig. 9d, g, h). In addition, obvious fracture of the fiber was observed on the worn surface under the 5 MPa, 2 m/s condition after TC treatment (see Fig. 9d, g), which may act as a third body that causes severe wear during sliding. Similarly, with the velocity increasing, the friction heat also can promote the formation of the friction platform on the worn surface in Fig. 9g, h (36). The severe friction caused by fiber fracture can also be shown from the transfer film on the counterpart (see Fig. 10a). Point element analysis was used to detect the strip of the worn surface, as shown in Fig. 10. Compared with the surface of the unworn steel plate (see Fig. S2), the appearance of C, N, and O can be ascribed to the transfer of polymer. The break and transfer of glass fiber in the matrix was found based on the Si element, which corresponded to the fiber pull-out and breakage during the sliding process, under 5 MPa, 2 m/s, after TC treatment.
With the increase of load, the severe shear also squeezed the fractured fiber into fine particles, which were transferred to the steel plate with the transfer of polymer, forming the lumpy transfer film on the steel plate surface. Meanwhile, the transferred debris acted as lubricant additives according to the analysis of the formation mechanism of transfer film by Zhang (1,24,37). From the mapping of element distribution in Fig. 11, it is obvious that the brightness of Si and O and the darkness of Fe in the grooves in the figure were very consistent, which presented cogent evidence for the presence of fiber debris in the grooves. In addition, the

Analysis of effect of TC treatment
According to the changes in dynamic mechanical properties and cross-section microstructure of the samples before and after TC treatment, the composites show different performances after TC treatment in spite of the failure test. After TC treatment, PI and PIA can be used well in friction testing under a high PV value. However, the comparison analysis of tribological properties of PI and PIA before and after TC treatment was not feasible under 2 m/s with different loads, due to the failure of friction tests.  In order to comprehensively analyze the effects of load and velocity on tribological performance of PIA and PIG before and after TC treatment, the variations of FC and WR of PIG and PIA after TC treatment were examined under the same PV conditions (2 MPa, 1 m/s and 1 MPa, 2 m/s), as shown in Fig. 12. As shown in Fig. 12(a), FC variations of PIA showed an increasing trend with increasing PV value. That means a slight effect of velocity or load on the FC performance of PIA, as the PV value increases. In contrast to FC variation, the WR of PIA presented an increasing trend with increasing PV value (Fig. 12b). Especially, WR increased significantly as the load increased to 3 MPa, when the PV value was 6 MPaÁm/s. This led us to speculate that WR was more closely related to load than to velocity as PV values increase. However, in contrast to PIA, the FC and WR of PIG presents different variation trends. When PV ¼ 6 MPaÁm/s, FC decreased sharply with the increase of load, and the relationship between FC and speed was basically linear and inversely proportional, which indicated that PIG was more dependent on load than velocity (37). Meanwhile, with the increase of PV, WR curves at different slopes showed different dependence on load or speed. It is clear that, similar to the effect on FC, the effect of load on PIG WR was huge compared to the effect on speed. This can also be demonstrated by the severely worn surface in Fig.  9g, h.

Conclusion
In this study, PI composites with aramid fiber and glass fiber were fabricated. The mechanical and tribological properties of these composites before and after TC treatment are investigated to explore the effect of TC treatment on PI and PI composites. Some conclusions are drawn as follows: 1. The results demonstrate that both bending strength and bending modulus are increased; however, the bending modulus does not change much. Results found the explicit divestiture of fiber in PI resin appears after TC treatment for PI composites based on the fractographic morphologies due to different coefficients of thermal expansion. The bending modulus increased slightly, while the bending strength increased too much after TC treatment. This is consistent with the chain growth observed during TC treatment in dynamic thermal mechanical experiments. 2. Friction coefficients (FC) and wear rates (WR) displayed different variations for PI and PI composites, respectively, compared to that before TC treatment. For PI and PIA, most tribological tests failed due to the fluctuation during the sliding process, especially tests before TC treatment. However, FC and WR increased after TC treatment for PIG. Load presented a great impact on tribological performance based on the variation under different PV values. 3. TC treatment has a great influence on the morphology of worn surface and composition of transfer film. As the shear force increased, the GF squeezed into the grooves, as demonstrated by the elemental analysis. Especially for PIG, the worn surface shows more craters caused by fiber pulling out from the matrix.

Electronic supplementary material
The online version of this article contains supplementary material, which is available to authorized users.