In summary, the microstructural investigation of CVI matrix composites indicates

1. Pyrolytic carbon tends to deposit preferentially on the surface of the composite being fabricated, thereby blocking surface porosity.

2. The structure of carbon produced from CVI deposits can vary from isotropic to highly oriented anisotropic, depending on processing conditions.

3. Bonding between isotropic CVI matrices and fibers is stronger than the bonding between lamellar CVI matrices and fibers.

4. The lamellar CVI deposits contain many narrow, slit-shaped microfissures that are generally <1 μm in length and <0.1 um in width. The isotropic CVI deposits contained no cracks.

Influence of Matrix on Composite Properties

General Background

When selecting a resin, metal, ceramic, or carbon-matrix composite for a given application, three criteria are generally used: the composite must have the desired physical and mechanical properties; it must be capable of being processed or manufactured into the desired shape; and it must be economical to produce.

For most composites, the primary consideration for a given application concerns the properties of the reinforcing fibers. The mechanical properties (strength and modulus) of carbon fibers are also related to the physical properties (thermal and electrical conductivity and coefficient of thermal expansion).

Because of the interrelationships that exist between the microstructure and the elastic and physical properties (thermal and electrical conductivity and coefficient of thermal expansion), the choice of a fiber based on one property usually determines the value of the other properties. For instance, the microstructure of very highmodulus fibers usually consists of long fibrils that are almost perfectly aligned parallel to the fiber axis. As a result, the transverse modulus will be relatively low, the thermal and electrical conductivity will be high in the longitudinal direction, and the thermal expansion coefficient will be small or negative.

A composite matrix usually serves to protect the reinforcement fibers from damage or reaction with the environment, to provide some measure of support in compression, to provide adequate matrix-dominated properties, and to provide a continuity of material. This last property is important in electrical and thermal applications and is particularly important in mechanical applications since load must be transferred to the fibers through the matrix. In this respect, a load can be transferred to the fibers across a chemically or physically bonded interface or across a mechanically interlocked one formed by the matrix shrinking onto and thereby gripping the fiber surface.

The properties of the matrix dominate the properties of composites in any direction in which fibers are not aligned. Such properties include the transverse tensile strength and modulus, interlaminar shear strength, thermal expansion coefficient, and electrical conductivity.

Composite materials reinforced with continuous fibers are complex materials that, for a given weight, exhibit specific mechanical properties almost always superior to those exhibited by conventional metals and alloys. Three basic composite types can be conveniently discussed in terms of maximum temperature of use: reinforced polymers, which are restricted to use at relatively low temperatures; reinforced metals, which can be used at intermediate temperatures; and reinforced carbons and ceramics, which can be used up to very high temperatures.

In order for the mechanical properties of a continuous fiber-reinforced composite to be superior to an unreinforced material, the modulus and strength of the fiber reinforcement must be greater than the matrix. In addition, there must be chemical, physical, or mechanical bonds formed between the fibers and matrix that are strong enough to transfer load between individual fibers and between fiber layers.

In estimating the mechanical properties of specific composites, assumptions are made with respect to the contribution of the matrix. For instance, the elastic modulus of typical polymers is usually so small that the contribution to the mechanical properties of the composite is ignored. Conversely, the modulus of metals is much larger; thus, the stiffness and strength of the composite will reflect a significant contribution to the tensile modulus. The modulus of polycrystalline carbon is small; however, the basal planes of carbon within the matrix of a composite can become highly oriented with respect to the fiber axis, showing a large increase in modulus. The modulus of carbon is very large in the direction of its basal plane, but very small in the out-of-plane direction; thus, any contribution to the modulus of the composite will depend on the alignment of the basal planes. In the previous section of this chapter, we have shown that a carbonaceous matrix highly oriented with respect to the axis of the fibers is most often produced from a pitch precursor. A range of structures (isotropic and anisotropic) can be produced from CVI and resin. However, basal plane alignment can be more easily achieved in CVI.

Elastic Modulus

When measured parallel to the fiber direction, the elastic modulus of a unidirectionally reinforced composite can be calculated to a first approximation from the rule of mixtures

where Em and Ef are the moduli of the matrix and fiber measured parallel to the fiber axis and Vƒ is the fiber volume fraction. When the modulus of the matrix is much smaller than the fibers, equation (1) reduces to

The above expressions give reasonable values for resin and metal matrix composites. Experiments carried out by Perry and Adams (ref. 97), using a variety of fiber types with SC-1008, FF-26 phenolic resins, or polyphenylquinoxaline and various furfural or furfural blends as impregnants, measured modulus values for CC which were much larger.

Further evidence for the large contribution of matrix to the stiffness of CC composites has been reported by other workers (refs. 98 through 101). For instance,

Fitzer and Hüttner (ref. 98) reported modulus values for pitch-char matrix materials that were twice that computed using equation (2), implying that the stiffnesses of the matrix and fiber were equal. The modulus values for resin char matrices were 40 percent higher. Further confirmation of this large stiffness contribution of the matrix was provided by the present authors in testing a unidirectionally reinforced sample of PAN fiber T-300 reinforced pitch material cut from a graphitized multidirectionally reinforced composite. The composite sample containing about 50 volume percent fiber reinforcement exhibited a modulus value of 45 Msi (310 GPa). Since the work reported by Becker (ref. 102) indicates that the modulus of graphitized T-300 fibers is about 58 Msi (400 GPa), it can be inferred from equation (1) that the modulus of the carbon matrix in the direction of the fiber axis was 32 Msi (220 GPa).

In the previous section dealing with microstructures, we established that an appreciable degree of preferred orientation develops on heat treating pitch, resin (or CVI) matrix precursors. Therefore, we concluded that this directionality significantly influences the modulus values of the resulting composite, and for a given volume fraction of fibers with similar bonding, the strong orientation tendency of pitch and highly oriented CVI-based matrix materials will produce the stiffest composite when measured parallel to the fibers.

Tensile Strength

In a mechanically loaded composite, it is usually assumed that the fibers and the matrix are bonded together and that no differential movement exists between them. The strain in each part of the composite can then be considered identical and equal to the total strain on the composite. For an elastic response, the stress on a unidirectionally reinforced composite loaded parallel to the fiber axis σe can be computed by multiplying the elastic modulus in that direction E, by the measured strain c, i.e.,

The failure strain of a polymer or a metal em is substantially larger than that of the reinforcing fibers ef, and failure of this type composite occurs when the fibers break. Equation (4) is therefore used to predict failure of the composite by assuming at fracture m = f. In contrast, the failure strain of the matrix em in most CC's is smaller than the failure strain of the fibers eƒ and, in a well-bonded composite, failure occurs when the the matrix fails. Failure occurs rapidly and results from the catastrophic propagation of a crack initiated at some weak point in the composite.

The failure of CC composites can again be discussed with reference to equation (4). If it is assumed that the maximum failure strain of carbon matrices is approximately 0.3 percent, then the composite will fail at that strain. Also, if it is assumed, for simplicity, that the matrix carries no load, the strength of a carbon matrix reinforced with 50 volume percent of 75 Msi (517 GPa) high modulus fibers would be about 112 ksi (772 MPa). Conversely, the strength of the same matrix reinforced with 30 Msi (206 GPa) low modulus fibers would be about 45 ksi (310 MPa). Since higher modulus fibers are usually weaker, we note the curious result discussed by Fitzer and Hüttner (ref. 98) that for composites that fail at the matrix failure strain, the strengths can be larger when stiffer; usually weaker fibers are used as the reinforcement. The failure of a composite whose failure is matrix-dominated can occur at lower strains when differences in thermal expansion between the fiber and matrix cause the matrix to be prestressed in tension on cooling from carbonization temperatures. Since the matrix and fiber must be bonded together in order to prestress the matrix, very low strengths might be expected from well-bonded materials. Thomas and Walker (ref. 103) studied phenolic resin precursors reinforced with three types of commercially available carbon fibers. Some of the fibers were surface treated to promote bonding; some were not. In every case, the matrix-dominated properties (transverse flexural strength, modulus, and interlaminar shear) of well-bonded phenolic char-matrix composites were superior to those exhibited by less well-bonded composites. Unfortunately, the longitudinal strengths of the well-bonded material were extremely poor and the work of fracture was low. In contrast, the matrix-dominated properties were poor, and the longitudinal strength of fully processed CC composites reinforced with nonsurface treated fibers was superior to those containing surface-treated fibers.

An additional series of experiments were performed by Fitzer et al. (ref. 104) who studied the effect of the degree of fiber oxidation on the mechanical properties of carbonized resin materials reinforced with SIGRAFIL HF and SIGRAFIL HM carbon fibers. After initial carbonization, but before subsequent densification, they found, relative to material reinforced with nonoxidized fibers, that increasing oxidation improved the strength of the material reinforced with the HM fibers, but decreased the strength of the material containing the high-strength fiber.

Manocha et al. (ref. 105) performed a similar experiment using surface-treated Toray M40 fibers and a matrix of furfural alcohol condensate. It was found that after the initial carbonization, the strength of the composite containing the surfacetreated fibers (oxidized) was poorer than that containing untreated fibers. The effect was similar to that observed by Fitzer et al. on SIGRAFIL HF materials. Manocha et al. graphitized their first carbonized material without densifying it. The strength of the surface-treated fiber composites increased by a factor of 3 or more while the strength of the nonsurface-treated materials decreased by nearly the same amount. In effect, Manocha showed that the composite reinforced with the surface-treated fibers exhibited the highest strength when graphitized.