A Study of the Thermomechanical Properties of Carbon FiberPolypropylene Composites

F. W. J. VAN HATTUM and C. A. BERNARDO Department of Polymer Engineering, University of Minho ~ 4800 Guimaraes, Portugal J. C. FINEGAN and G. G. TIBBETTS Department of Physics and Physical Chemistry, General Motors R&D Center Warren, Michigan 48090-9055 R. L. ALIG and M. L. LAKE Applied Sciences Inc. Cedarville, Ohio 45314
Short fiber composites were produced using polypropylene as matrix and polyacrylonitrile (PAN)-based fibers or vapor grown carbon fibers (VGCF) as reinforcement. The strength, stiffness, and coefficient of thermal expansion (CTE) of the composites were measured. The VGCF-composites showed strength and CTE that are competitive with those of conventional PAN-fiber composites, but the stiffness was marginally lower. Micromechanical modeling of the PAN composite properties gives results consistent with the measurements. The models can be used to infer the apparent VGCF-properties from their composites. 1. INTRODUCTION duction of these composites is straightforward using conventional processing technologies. Although the incorporation of the VGCFs in the polymeric matrix led to better mechanical properties, the improvement was marginal. The objectives of the present work are therefore twofold. First, to produce VGCF-thermoplastic composites with thermomechanical properties significantly better than those of the unreinforced polymer and to compare them with conventional composites. Second, to use micromechanical models to directly calculate the properties of short fiber composites. If these models show accurate enough, they can be used to infer the properties of the submicron VGCFs from those of the composite. In the previous investigation (3), it was concluded that the presence of polycyclic aromatic hydrocarbons (PAHs) on the surface of the fibers could cause chemical stress cracking in the polycarbonate matrix. As a consequence, in the work described below, the fibers were heated to drive off the PAHs before incorporation in the matrix. Polypropylene (PP) was chosen, as this polymer has relatively good mechanical properties at a moderate price and is not very sensitive to chemical stress cracking.
683

arbon fibers are widely used in polymer-matrix composites, owing to their good mechanical, thermal and electrical properties. Vapor grown carbon fibers (VGCFs) have recently gained interest, owing to their potential low-cost production and favorable properties. In this way, fibers can be produced at much higher rates and lower costs (1, 2). The VGCFs with greater economic potential are, however, small: the length of the fibers is usually below 100 m and the diameters near 200 nm (2). Because of this, they are sometimes referred to as submicron (diameter) fibers. Although from a cost-property-relation point of view, the submicron VGCFs are commercially attractive, their use poses new problems, as conventional expertise on composite production may in some cases no longer be applied. Moreover, as the minute dimensions of the VGCFs do not allow direct determination of mechanical and thermal properties, the entire modeling process is more difficult. In a previous work (3), the production and mechanical characterization of polycarbonate (PC) thermoplastic composites reinforced with this type of VGCFs was described. It was shown that the pro-

POLYMER COMPOSITES, OCTOBER 1999, Vol. 20, No. 5

F. W. J. Van Hattum, C. A. Bernardo, J. C. Finegan, G. G. Tibbetts, R. L. Alig, and M. L. Lake For comparison, polypropylene composites reinforced with PAN-based carbon fibers were also produced and investigated. The micromechanical models found in the literature (46, 13) were validated on these PP-PAN composites to calculate the elastic modulus, strength, and CTE. 2. EXPERIMENTAL 2.1 Materials and Sample Preparation The VGCFs, Applied Sciences Pyrograf III, were produced using a floating catalyst method developed by Tibbetts, Gorkiewicz, and Alig (2). A catalyst such as iron pentacarbonyl was injected into the flowing hydrocarbon gas, where it nucleates and grows the fiber. The fibers moved through the reactor with the gas stream and were collected at the exit. Montells Moplen F30G polypropylene, TENAXs PAN-based HTA 5131 carbon fibers, and the VGCFs were processed in a Leistritz LSM 30.34 twin-screw extruder to obtain composites with a fiber volume fraction of 15%. The VGCFs were kept in an oven at 200C for 2 hours prior to processing, to ensure that any PAHs remaining on the fibers surface had vanished. The extrudate was granulated to obtain composite granules with a length of some millimeters. Prior to further processing, batches with fiber volume fractions of 10% and 5% were obtained by mixing the granulate with unreinforced polypropylene. Tensile bars, adapted from ASTM D638M, were subsequently injection molded using a Klockner Ferromatic FM20 injection molding machine. The processing conditions were kept constant for the different materials, however higher injection pressure and melt temperature had to be used for the PP-VGCF composites. In this way, PP-PANfiber and PP-VGCF composite tensile bars were obtained with fiber volume fractions of 5%, 10%, and 15% respectively. Tensile bars of unreinforced polypropylene were similarly produced. The processing conditions are given in Table 1. 2.2 Material Characterization PP-PAN Composites In the modeling work referred to above (6), the present PP-PAN composites and PAN fibers were fully characterized. In order to obtain the mechanical properties of the composites and the virgin material, tensile tests were performed according to the ASTM D638M standard, using an Instron 4505 universal testing machine. Tensile bars were tested at a crosshead speed of 5 mm/min. The PAN-fiber properties were determined by testing the fibers according to ASTM D 3379-89 in an Instron 1122 universal testing machine with a 5 N load cell. Using a crosshead speed of 0.5 mm/min and 5 different gauge lengths (5, 10, 20, 40, and 80 mm), 20 samples at each gauge length were tested. The diameters of the PAN-fibers were measured prior to testing using a laser diffraction technique. In the present work, the longitudinal CTE of the composites was also determined, since we were particularly interested in the lowest coefficient of thermal expansion and the longitudinal CTE is usually lower than the transverse CTE. The coefficient of thermal expansion in the longitudinal direction was measured with a Thermal Mechanical Analyzer 2940, fabricated by TA Instruments, which is capable of heating or cooling samples while applying a predetermined force. A quartz probe, placed in contact with the sample, is used to determine linear or volumetric changes, at any selected temperature. Samples of material having a maximum height of 25 mm and maximum diameter of 10 mm were cut with parallel faces, and the thermal expansion coefficients were measured from 30C to 120C. Fiber orientation measurements were made on polished cross sections of the PAN-composites cut parallel to the flow at the center of the composite tensile bars, using the method described by Bay and Tucker (7). To determine the fiber length, the matrix of some of the PP-PAN tensile bars was burned off in an oven, the fibers spread on a glass slide and fiber lengths measured under an optical microscope using an Image Analysis system. The fiber volume fractions for each sample were determined by density measurements. PP-VGCF Composites Because of the small fiber size in the PP-VGCF composites, traditional methods to determine fiber properties cannot be applied. Earlier work on a different type of longer VGCFs (diameters and lengths in the micrometer and centimeter range, respectively) has revealed the large dependence of the properties on fiber shape, and the occurrence of a large number of fiber shapes (8, 9). Overall, the properties showed a qualitative dependence similar to that observed with conventional carbon fibers. The determination of the composites strength, stiffness, CTE, and fiber volume fraction was done using the same methods as with the PP-PAN composites. However, because of their fiber size, conventional measurement of the fiber orientation state in the PP-VGCF composites was not possible. The fiber lengths of the VGCFs were determined by oxidizing the composite in air at 400C overnight. The SEM samples were prepared by first squeezing a glue dot onto an aluminum sample stub

Table 1. Operational Conditions Used in the Injection Molding. PPPAN Injection pressure (bar) Hold pressure (bar) Hold time (s) Cool time (s) Mold temperature (C) Melt temperature (C) 25 20 6 20 60 230 PPVGCF 40 30 4 20 60 250

Note: the pressures are those of the hydraulic system.

684

POLYMER COMPOSITES, OCTOBER 1999, Vol. 20, No. 5

Thermomechanical Properties of Carbon Fiber-PP Composites

Fig. 1. Normalized experimental modulus at 1% strain.

and then scraping it off with forceps to leave an area of adhesive so thin that fibers could not submerge in it. Next, a small wad of fibers held in a pair of forceps was scraped over the adhesive area to disperse and spread the individual fibers. Gold was then sputterdeposited on the stub to make it conductive. From micrographs of regions where the fibers were well dispersed, the lengths of at least 50 fibers were measured and averaged. 2.3 Micromechanical Modeling Micromechanical models to predict stiffness and CTE of short fiber composites have already been successfully applied (4, 5). In addition, the authors recently derived and successfully applied a model to predict strength in

short fiber composites (6), following the same general method used for stiffness- and CTE-prediction. As a first step, all these models derive expressions for the properties of a unidirectionally aligned short fiber composite as a function of fiber length. The expressions for the unidirectional stiffness are directly derived from the matrix and fiber stiffness by the wellknown Halpin-Tsai equations (10). The expressions for the unidirectional CTEs are given by Schaperys equations as adapted by Halpin (11). These expressions contain not only the underlying CTEs of matrix and fiber, but also the above unidirectional composite stiffness, thus slightly complicating the overall expressions. The unidirectional strengths are given by relations based on the Kelly-Tyson theory (12). This theory

Fig. 2. Normalized experimental strength.

POLYMER COMPOSITES, OCTOBER 1999, Vol. 20, No. 5

685

F. W. J. Van Hattum, C. A. Bernardo, J. C. Finegan, G. G. Tibbetts, R. L. Alig, and M. L. Lake

Table 2. Composites Test Results. Fiber Fraction Fiber volume fraction (%) Fiber length ( m) Tensile modulus at 1% strain (MPa) Tensile Strength (MPa) CTE (106/C) PPPAN PPVGCF PPPAN PPVGCF PPPAN PPVGCF PPPAN PPVGCF PPPAN PPVGCF 0% 1343 1646 30.4 32.9 114.0 111.3 5% 4.5 4.7 161.7 3.8 2821 2584 36.6 40.3 31.5 57.4 10% 9.6 9.7 148.3 3.8 4294 3466 42.6 46.5 20.8 39.8 15% 14.7 15.6 161.3 3.8 4591 3685 49.7 48.5 16.9 26.2

assumes that an applied load is transferred to the fibers by means of a shear force at the fiber-matrix interface. A critical minimum fiber length is thus needed to build up sufficient strength to fracture the fiber. If the fibers in a composite are shorter than this critical length, fiber fracture will not occur. Hence, the strength of such a unidirectional composite is not determined by the fiber strength, as one would expect, but by the strength of the interface between fiber and matrix, the so-called interfacial shear strength. By substituting the average fiber length found in the composite in these expressions or, alternatively, integrating the expressions over the whole range of fiber lengths, the unidirectional composite properties can be obtained. The properties of the composite are then taken as an average of these unidirectional properties over all directions, weighted by the orientation distribution function, which describes the fiber orientation. This is often referred to as orientation averaging. The unidirectional properties are used to construct the stiffness, thermal expansion and strength tensors. These tensors can be orientation averaged following the

method described by Advani and Tucker (13), to derive the final composite properties (5, 6). In the present work, micromechanical models are used to predict the properties of the PP-PAN composites. The applicability of the models is verified by comparison with experimental data. Because of the difficulty of determining the orientation of the VGCF in composites, the above models were not applied to VGCFs at this stage. However, when shown to be valid, the micromechanical models can be used to determine the VGCF properties, if their orientation in the composites is known. 3. RESULTS AND DISCUSSION 3.1 Material Characterization The properties of the composites are given in Table 2. It should be noted that the properties for the unreinforced polymer (fiber fraction 0%) differ for the PPPAN and PP-VGCF composites. This is due to the differences in processing conditions used, as explained in Section 2.1 (see also Table 1). For ease of comparison, Figs. 1 through 3 depict, therefore, the normal-

Fig. 3. Normalized experimental CTE.

686

POLYMER COMPOSITES, OCTOBER 1999, Vol. 20, No. 5

Thermomechanical Properties of Carbon Fiber-PP Composites

Table 3. Material Parameters Used for the Evaluation of the Model. Material Parameters Matrix modulus (GPa) Matrix Poisson ratio Matrix yield stress (MPa) Matrix stress at fiber failure strain1 (MPa) Matrix CTE (106/C) Fiber modulus (GPa) Fiber diameter ( m) Fiber Poisson ratio Fiber tensile strength2 (MPa) Fiber CTE (106/C) Critical fiber length ( m) Fiber volume fraction (%) Average fiber length ( m)
1Calculated 2Fiber

PPPAN 0% 1.343 0.40 30.4 20.8 114.0

PPPAN 5%

PPPAN 10%

PPPAN 15%

PAN

218 7.2 0.26 4940.1 l 0.1554 0.1 1012 4.5 161.7 9.6 148.3 14.7 161.3

from fiber failure strain and matrix modulus. length l in mm.

ized (that is, divided by the matrix value) stiffness, strength and CTE, respectively. In these Figures, the normalized properties of the PC-VGCF composites (3) are also shown. From the Table and the Figures it is easily observed that the VGCF-composites show strength properties comparable to those of the PP-PAN composites. Stiffness values are lower, whereas the CTEs are slightly higher. Other work (14), however, has shown that 16% (v/v) PP-VGCF composites can be 300% stiffer than the matrix. This increase compared with the present work can possibly be attributed to better fiber alignment and less fiber length degradation. The dependence of the properties on fiber volume fraction is similar for both PP-VGCF and PP-PAN composites. From the Figures it is clear that the strength and CTE (and possibly stiffness) of the VGCF-composites are competitive with properties of composites based on PAN-fibers. 3.2 Modeling The experimental data were used to model the PPPAN composite properties. The data used for modeling the properties are summarized in Table 3. A typical fiber orientation distribution through the short dimension transverse to the injection direction is shown in Fig. 4, for the PP-PAN composite with 15% fiber volume fraction. In this Figure, a11 and a22 represent the well-known 2nd order orientation tensor components (13), giving the relative orientation of the fibers around the 1-axis (flow-direction) and 2-axis (cross-flow direction), respectively. The skin-core structure, reported for many injection molded short fiber composites, can

be readily observed in this Figure. From the properties of fiber and matrix shown in Table 2, stiffness, strength, and CTE for the PP-PAN specimens in the flow direction were calculated using the methods described in Refs 13, 6, and 5, respectively. The predictions, together with the experimental values, are given in Table 4. As already shown in Ref. 6, the predicted values for strength and stiffness for the PP-PAN composites are within 10% of the experimental values. The CTE-predictions lead to greater discrepancies. These errors, however, are consistent with results found in work of other authors (5). The dependence of all composite properties on fiber volume fraction, as observed experimentally, is well described by the models. Because of the problems related to conventional fiber orientation measurements, the current modeling cannot be applied to the submicron VGCF-composites. Alternative methods of measuring fiber orientation, like X-ray scattering, could be considered. In that case, the micromechanical models can be inversely applied to the PP-VGCF composites to derive the apparent VGCF properties. However, the models are based on various assumptions, the most critical one being that the fibers are cylindrical in shape. This is a very restrictive hypothesis, as it is well known that because of the production method, submicron VGCFs are not cylindrical (15). Hence, strict conclusions could not be drawn with this method. The properties derived would only correspond to those of a model fiber. If used as reinforcement in a polypropylene matrix, this ideal fiber, as-

Table 4. Experimental and Predicted Values of the Thermomechanical Properties. PPPAN 5% Strength (MPa) Modulus (MPa) CTE (106/C) Experimental Predicted Experimental Predicted Experimental Predicted 36.6 35.0 2821 2787 31.5 51.3 PPPAN 10% 42.6 45.2 4294 4140 20.8 32.3 PPPAN 15% 49.7 51.4 4591 5043 16.9 24.1

POLYMER COMPOSITES, OCTOBER 1999, Vol. 20, No. 5

687

F. W. J. Van Hattum, C. A. Bernardo, J. C. Finegan, G. G. Tibbetts, R. L. Alig, and M. L. Lake

Fig. 4. Typical fiber orientation distribution through-thickness, for PP-PAN 15%.

sumed cylindrical and characterized by its length and diameter, would lead to composite properties matching those of the PP-VGCF properties. Nevertheless, this method would still allow direct comparison of the VGCF with conventional fibers. In fact, in this way one can readily determine what properties a conventional fiber should have to obtain composite properties equivalent to those of the PP-VGCF composites. Furthermore, in previous work (6), the authors have shown that in the PP-PAN composites, the interfacial shear strength rather than fiber strength dominates composite strength, owing to the short fiber lengths (see Section 2.3). In the PP-VGCF composites this should also be the case, considering the average fiber length found experimentally (see Table 2). Therefore, using inverse modeling, an actual fiber strength value for the VGCFs cannot be derived. Instead, the interfacial shear strength between fiber and matrix can be obtained. In this way, the apparent properties of the VGCFs can be determined, allowing comparison with conventional fibers. CONCLUSION In the present work it has been shown that VGCFthermoplastic composites can be produced, with strength and CTE that are competitive with those of conventional PAN-fiber composites. Micromechanical models have been successfully applied to predict the strength and stiffness of PP-PAN composites. Modeling the CTE in the same way yielded slightly worse results. The models can be used to infer the apparent VGCF-properties from their thermoplastic composites, if the exact orientation of the fibers in the composites is known. ACKNOWLEDGMENTS This work was supported by the European Economic Community, through the Human Capital and

Mobility Programme, under Grant Number CHCRXCT940457. F. W. J. van Hattum acknowledges the personal grant received under the same contract. J. C. Finegan would like to thank the NIST ATP for support under co-operative agreement number 70NANB5H1173, Vapor-Grown Carbon Fiber Composites for Automotive Applications. The support of Tenax Fibers, Germany, which kindly supplied the PAN-fibers, is also acknowledged. REFERENCES
1. G. G. Tibbetts, and M. G. Devour, U. S. Patent No. 4,565,684 (1986). 2. G. G. Tibbetts, D. W. Gorkiewicz, and R. L. Alig, Carbon, 31, 809 (1993). 3. O. S. Carneiro, J. A. Covas, C. A. Bernardo, G. Caldeira, F. W. J. van Hattum, J.-M. Ting, R. L. Alig, and M. L. Lake, Compos. Sci. Tech., 58, 401 (1998). 4. P. H. Foss, J. P. Harris, J. F. OGara, L. P. Inzinna, E. W. Liang, C. M. Dunbar, C. L. Tucker III, and K. F. Heitzman, SPE ANTEC Tech. Papers, 42, 501 (1996). 5. C. W. Camacho, C. L. Tucker III, S. Yalva, and R. L. McGee, Polym. Compos., 11, 229 (1990). 6. F. W. J. van Hattum and C. A. Bernardo, Polym. Compos., 20, 524 (1999). 7. R. S. Bay and C. L. Tucker III, Polym. Eng. Sci., 32, 240 (1992). 8. F. W. J. van Hattum, J. M. Benito-Romero, A. Madroero, and C. A. Bernardo, Carbon, 35, 1175 (1997). 9. F. W. J. van Hattum, C. A. Bernardo, Ph. Serp, and J. L. Figueiredo, Carbon, 35, 860 (1997). 10. J. C. Halpin and J. L. Kardos, Polym. Eng. Sci., 16, 344 (1976). 11. J. C. Halpin, J. Compos. Mater., 3, 732 (1969). 12. A. Kelly and W. R. Tyson, J. Mech. Phys. Solids, 13, 329 (1965). 13. S. G. Advani and C. L. Tucker III, J. Rheol., 31, 751 (1987). 14. J.-M. Ting, D. J. Burton, R. L. Alig, and P. Flem, Ext. Abs. 23 rd Biennial Conference on Carbon, p. 294, American Carbon Society, State College, Pa. (1997). 15. G. G. Tibbetts, C. A. Bernardo, D. W. Gorkiewicz, and R. L. Alig, Carbon, 32, 569 (1994).

688

POLYMER COMPOSITES, OCTOBER 1999, Vol. 20, No. 5