Effect of Curing Age on Pull-Out Response of Carbon, Steel, and Synthetic Fiber Embedded in Cementitious Mortar Matrix

Natural river sand meeting ASTM C778 (ASTM 2017) and Quickrete portland cement conforming to ASTM C150 Type I (ASTM 2020b) were used. Cement mortar used in this experiment had a total binder content of $811\mathrm{kg}/{\mathrm{m}}^3$. It was produced with a water-binder ratio of 0.4, and the mixing process was operated in accordance with ASTM C305 (ASTM 2020a). The compression strength (on $5\times 5\times 5\mathrm{cm}$ cubes) and tensile strength (on dog-bone samples) of plain cement mortar at 28 days of curing was 55.8 and 4.3 MPa, respectively. To investigate the pull-out response of fiber in cement matrix, hooked-end SF, STRUX BT50 SI and two types of PAN- and pitch-based CF from Mitsubishi chemicals were used (Fig. 1). The properties of each fiber as received from the respective suppliers and the mix design of mortar are presented in Tables 1 and 2, respectively.

Dog-bone shaped specimens were fabricated to measure the pull-out behavior of four types of fiber when embedded within the cementitious matrix. The fiber embedment lengths were 25 mm on both halves of the specimen as suggested by Ratu (2016). For obtaining reliable test data, a minimum of three specimens for every mix was prepared for each test. Carbon fiber bundles were twisted in three different bundle diameters (0.5, 0.8, and 1 mm). The reason to use bundles instead of single fiber was (1) embedding one single carbon fiber (diameter 7–11 μm) in cement mortar was impossible due to a single fiber’s brittleness making it prone to break when hit, and (2) the importance of bundle behavior evaluation. Although different dispersants were introduced to ensure suitable dispersion of carbon fibers in FRCC (Cao and Chung 2001; Chen et al. 1997; Fu et al. 1998; Chuang et al. 2017), the cost of these admixtures is relatively high (Chung 2005), and the homogenous cement paste structure is also reported to be affected by dispersant (Muhua and Roy 1987). In recent years, several alternative approaches were introduced to disperse CF (Gao et al. 2017; Wang et al. 2017), and still a large number of studies report the presence of CF bundles in the matrix (Gao et al. 2017; Bhosale et al. 2019; Deng and Li 2006).

The reason to choose the aforementioned diameters was (1) these values were the smallest diameter with an acceptable accuracy that we could get from the optical microscope secure of fiber breakage, and (2) these dimensions were comparable to SF and SI, which had similar dimensions. As shown in Fig. 2, an optical microscope was used to find the accurate diameter of the twisted fibers. Measurements were done on at least three points on the fiber to validate the correct diameter. Because bundles were twisted, it was assumed that cement paste could not penetrate inside the twisted fibers, and they were considered as a single fiber with an equivalent diameter equal to 0.5, 0.8, and 1 mm. Fibers were also glued together at each end of bundle, so that they did not disperse when put in the mold. During the preparation procedure, a single fiber or fiber bundle (for CF) was positioned inside a small hole in the middle of a thin plastic plate with 1 mm thickness [Figs. 3(a and b)]. Then, half the depth of the molds was filled with mortar and the plastic plate was placed inside the matrix [Fig. 3(c)] such that the length of the fiber was oriented along the loading direction. The last step involved filling the rest of the mold with mortar [Fig. 3(d)]. A summary of sample preparation steps is shown in Fig. 3. Test specimens were cured at room temperature for the initial $24\pm 2\mathrm{h}$. Then, the specimens were demolded and cured in a water bath at $23°\mathrm{C}\pm 3°\mathrm{C}$ until they were tested.

PASCO materials testing machine (MTM) with a load cell capacity of 7,100 N (Fig. 4) and a $1\mathrm{mm}/\min$ crosshead loading rate was used to carry out the pull-out test. The fixture consists of identical upper and lower grips which were fixed to the test machine. During testing, the upward movement of the actuator applies a pull-out force on the upper half of the specimen. The machine is equipped with an optical encoder to measure the displacement of the sample. Force data from the load cell and displacement data from the encoder module can be recorded by a PASCO interface with PASCO data acquisition software (PASCO capstone). PASCO capstone is a data collecting software which can automatically collect and record data with different adjustable frequencies. In this experiment, data values were recorded at a frequency of 25 Hz. The optical encoder in the test machine was used to ensure no preload was applied to the specimen and to minimize the initial seating-based errors while sample gripping.

The pull-out behavior of fiber is based on different factors including the physio-chemical adhesion, friction, and mechanical resistance (Naaman and Najm 1991). In this test, pull-out behavior of fiber is evaluated by analyzing force-displacement curves. Similar to previous studies (Wu et al. 2018; Bhutta et al. 2017), pull-out force versus displacement curves are used to reflect the amount of load and total displacement of fibers being pulled out; the effect of fiber type and matrix age on the pull-out mechanism is evaluated by comparing pull-out results. All types of fibers are pulled out at 7, 14, 28, and 56 days.

According to Le et al. (2018), a typical pull-out load versus slip curve for smooth steel fiber consists of the elastic stage, the partial debonding stage and the frictional stage. There is a linear elastic fiber-matrix bond in the elastic stage that is followed by a gradual separation in the adhesion between fiber and matrix, which is called partial debonding phase and finally results in complete debonding. In the final stage and after debonding of the fiber, the pull-out resistance is only governed by friction (Le et al. 2018). Analyzing the three mentioned stages will help to get a broad insight into pull-out behavior of different types of fibers.

Different parameters can evaluate the pull-out behavior and resistance of fibers. Maximum fiber stress (${\sigma }_{f,\max }$) is a vital parameter to assess the performance of fibers in concrete. If the maximum fiber stress exceeds its tensile strength, the fiber will be fractured before complete pull-out. The maximum fiber stress can be calculated using Eq. (1)where ${\sigma }_{f,\max }$ = maximum fiber stress; $P_{\max }$ = maximum pull-out force; and $A_f$ = cross-sectional area of fibers (in samples with CF, diameter of CF bundle is used to calculate the cross-sectional area).

Average bond strength (${\tau }_{av}$) between the fiber and the matrix is calculated using Eq. (2) (Kanda and Li 1998)where ${\tau }_{av}$ = average bond strength; $d_f$ = fiber diameter; and $L_e$ = initial embedded length of the fiber.

Fiber pull-out energy ($W_p$) is the area under the pull-out force versus displacement curve up to the point where the pull-out load becomes zero; it can be calculated using Eq. (3). Interface toughness of fibers is one of the most critical factors that evaluates performance of fiber and enhances the ductility of FRC. Interface toughness is defined as the mechanical energy consumed during fiber pull-out and is determined by integrating the area under the pull-out force-displacement curvewhere $W_p$ = fiber pull-out work; and $P(s)$ = pull-out force at a certain slip.

Interfacial bond strength at the interface between the fiber and matrix can be evaluated by calculating equivalent bond strength (${\tau }_{eq}$) using Eq. (4) (assuming shear strength is evenly distributed over the length of fiber) (Kim et al. 2007)

The results of maximum pull-out load ($P_{\max }$), standard deviation of pull-out loads (SD), displacement at maximum pull-out load (${\mathrm{\Delta }}_{p\max }$), fiber pull-out energy ($W_p$), maximum fiber stress (${\sigma }_{f,\max }$), average bond strength (${\tau }_{av}$), and equivalent bond strength (${\tau }_{eq}$) are given in Table 3 in which numbers after PA and PI represent the diameter of the twisted bundle of fiber in millimeters. For example, PA0.5 represents mortar mix with PAN-based CF and 0.5 mm diameter.

Pull-out behavior of four types of fibers was investigated. Fig. 5 shows optical microscopy images of fibers after being pulled out. According to the image, after the pull-out test, cement paste was still attached to the surface of steel fiber in some areas of the surface, which is proposed to be an indication of strong fiber-matrix bonding. In hooked end SF, the failure was a combination of matrix spalling and fiber-matrix debonding. However, in CF and SI, fiber-matrix debonding was the only and predominant failure in CF and SI.

Different failure modes were observed during pull-out test. In SF and SI samples, the failure mode was complete pull-out, while in CF, as it is shown in Fig. 6, different failure patterns were observed including fiber fracture and fiber pull-out. The complete pull-out failure mode in SF and SI indicates that that bonding between fiber and matrix is weaker compared to the strength of the matrix or fiber (Singh et al. 2004; Richardson 2005). Complete pull-out failure mode is a desirable failure mode to prevent sudden brittle failure in structures. However, the fracture failure mode reveals that the interfacial bonding is stronger than fiber tensile strength.

Fig. 7 is presented to show the failure pattern of different fibers. Note that each curve in the figure is a representative curve indicating a typical failure behavior of that specific fiber while being pulled out from the matrix. According to the results, the failure pattern in different fibers can be summarized as follows. In CF, a sudden drop in the pull-out load occurred after reaching the peak load. Considering the hydrophobic nature of CF, as well as the straight geometry of the fiber, the sudden drop at around 1 mm of CF bundle slip is attributed to the poor interfacial bonding between the bundle and the matrix. After deboning, friction is the predominant mechanism in pull-out behavior (Le et al. 2018) and as the slip increases, the pull-out load decreases due to the reduced friction area. However, because of the geometry of CF, the friction is much lower in CF bundles compared to SF and SI, which results in lower postcrack energy absorption. In SF, the decrease in the post peak load was not as sudden as for CF similar to what was found by Nieuwoudt and Boshoff (2017). After reaching the peak load, the increase in slip was followed by a strain softening behavior with a smooth decrease in pull-out load until anchorage of the fiber was straightened, which occurred at slip of approximately 5 mm. In SI, the slope of the curve before peak load was even smoother than SF, and after reaching the peak load, a slip-hardening behavior was observed leading to zero load at around 4.5 mm of fiber slip. The reason for slip-hardening behavior is the friction bonding, which is mainly dependent on the properties of fiber rather than concrete properties (Nanni et al. 1995). Hence, fiber geometry is one of the important factors in fiber bonding behavior, and by improving the geometrical shape of fiber, the energy absorption of fibers can be enhanced significantly. For example, crimping fibers is an effective way of improving the postcrack behavior of SI resulting in slip-hardening behavior after the maximum pull-out force is reached.

The detailed results of the pull-out test are presented in Table 3. During the preparation and testing of CF, some samples were damaged or broken due to the high vulnerability of fibers to any kind of impacts (missing values are presented in Table 3). These impacts were produced with sample handling, pretest preparation, testing, and also due to the matrix-to-matrix adherence of the two parts of the dog-bone sample. Although attempts were made to install the plastic plates in such a way that fibers were the only holders of the two parts of the briquette, still some small portion of the matrix attached to each other and needed to be broken. The load needed to break the cement matrix by the MTM was very low, but it could damage or break the bundle. Another reason for brittle behavior of CF is the high loading rate of MTM ($1\mathrm{mm}/\min$). The future scope of the carbon fiber pull-out test will be (1) increase in number of samples to make sure enough results will be provided in case of fiber damage; (2) decrease in loading rate to avoid brittle fiber breakage; and (3) fix the divider plate inside the mold in order to make sure there is no matrix-to-matrix adherence.

According to the bond strength results, SF had the highest bond strength and pull-out energy, followed by SI and then CF. According to the literature, the superior bond behavior will result in better fracture behavior (Bhosale et al. 2019; Deng and Li 2006). At 28 days of curing, steel fibers were able to carry an average of 358 N force, while this amount was 122, 155, and 156 N for SI, PA1, and PI1 at 28 days, respectively. This means steel fiber could carry more than twice the amount of force compared to other fibers, and the other three fiber types showed comparable results in terms of maximum pull-out force. However, for both S and SI samples, the maximum load was at 3.5 mm of slip, while for PA1 and PI1, the displacement at maximum load was 0.8 and 0.5, respectively.

Another factor to assess the performance of fiber is the amount of energy that can be absorbed by fibers. As shown in Table 3, the average fiber pull-out energy obtained was 1,050, 277, 15, 36, 47, 50, 50, and 55 N·mm for S, SI, PA0.5, PA0.8, PA1, PI0.5, PI0.8, and PI1 at 28 days, respectively. The results showed the superior behavior of SF to absorb energy as well as a suitable behavior of SI and thus, in order to provide an equal reinforcement effect to a cement-based composite, a much greater amount of carbon fiber will be needed. The reason is mainly due to the fact that hooked-end steel fibers anchor at the ends and crimped fibers anchor along the length while in straight filament fibers the frictional shear stress at the interface determines the energy absorption capacity of fibers.

Fig. 8 shows the tensile strength of fibers generated by force. According to the results, PI0.5 had the highest tensile strength; notably it was higher that SF and SI and thicker bundles. The increase in tensile strength with the decrease in diameter of the bundle approves the importance of fiber dispersion in order to maximize the capacity of CF. However, by comparing fiber tensile strength and energy absorption values, fibers with better tensile strength did not result in better energy absorption while fibers with better shear or bonding strength had improved energy absorption and after crack behavior. Although the tensile strength generated by force in all carbon fiber samples is lower than the fiber tensile strength, a few carbon fibers were fractured before complete pull-out. This is mainly due to the initiation of fracture of single fibers around the bundle, which are adhered to cement matrix and then the gradual breakage of inner fibers until all of fibers are fractured.

Fig. 9 shows maximum pull-out force versus matrix age of fibers. In case of the effect of matrix age, an increase in specimens curing age led to an improvement in fiber tensile strength in all fiber types except SF. The reason why SF didn’t show the same pattern as in CF and SI is because of an important additional major contributor to pull-out resistance, that is, mechanical anchorage (Tai and El-Tawil 2017; Le et al. 2018). This means when the hooked end fiber is pulled out, a large amount of energy is needed for the fiber to become straightened, and hence the interlock between the anchorage and the matrix is one of the determining factors in pull-out behavior of fiber.

The improvement in interfacial bond properties with an increase in curing age is mainly due to the production of hydration products especially calcium silicate hydrate and development of microstructure of interfacial transition zone (ITZ) as the degree of hydration increases over time. For carbon fibers, after curing for 7 days, all tests ended with the fibers being pulled out from the mortar, which indicated the weak bonding between fiber and matrix at early ages. As the curing age increased to 28 and 56 days, a few fiber samples were fractured while being pulled out, which indicates improvement in fiber bonding at later ages.

The fiber pull-out test is important since it provides information about interfacial properties of carbon fiber in the cement matrix. The pull-out force versus slip curves of various specimens is shown in Fig. 10. Although the test machine was equipped with an optical encoder to minimize the gripping error, a few samples still show some preloading values in the pull-out force versus slip curve. These errors, which can be seen in Fig. 10(a), were removed during the analysis.

The pull-out curves revealed the characteristics of the bonding properties of each type of fiber. In CF samples, after failure of fiber (after $P_{\max }$), the pull-out force decreased with increase in slip and no slip-hardening or slip-softening was observed, which is in conformity with previous studies (Lu et al. 2018). In synthetic macrofibers, the initial incline of the force-slip curve is linear to the point of failure and then follows a smooth decrease in the amount of force until fiber failure. In steel fibers due to the anchorage at the end of the fiber, the cement matrix will be damaged while fibers are being pulled out from the matrix and as a result several drops in load prior to final fiber failure are seen in the force-slip curve (Fig. 10). Having a smooth surface is the main reason for a sudden drop in the pull-out force of carbon fibers.

Although steel and synthetic fibers showed a much better energy absorption capacity, the bonding behavior of carbon fibers is comparable with other fibers. Average bond strength (${\tau }_{av}$) is considered as an important factor to evaluate the ITZ quality between fiber and matrix. According to Table 3, after steel fibers, most carbon fibers showed a comparable bond strength compared with synthetic fibers. In a few samples, such as PI0.5, bond strength was reported to be higher compared to bond strength of SI. Although the effect of fiber bundle diameter on bond strength didn’t show a particular and clear trend, it can be interpreted that as the age of samples increases, a better bonding is formed between fibers and matrix in all four types of fibers.