How-to, Tips and Articles: Benefits and Applications of Carbon Matrix Composites

1. Introduction

When writing this chapter one has to ask the question, ‘What are carbon matrix composites?’ There is no simple answer to that question. The most succinct answer I can give is to say that carbon matrix composites (CMCs) are a complex family of advanced materials that consist of carbon or graphite ﬁbers embedded in a carbon or graphite matrix. The carbon ﬁber reinforcement of these materials makes them stronger, tougher, and more resistant to thermal shock than conventional graphite.

The speciﬁc advantages of carbon–carbon are light weight and low density, high strength and stiffness, high thermal con- ductivity, low coefﬁcient of thermal expansion, high fracture toughness, and good fatigue and creep resistance. The combination of low expansion and high thermal conductivity means that carbon–carbon materials also possess excellent thermal shock resistance. Moreover, almost uniquely among materials, the strength of these composites increases with increasing temperature to temperatures well in excess of 2000 °C. Like most systems, a well-engineered carbon–carbon composite can be made to exhibit graceful failure and show pseudo plastic behavior. They also have high thermal and chemical stability in inert environments.

However, in common with all other forms of the element, carbon composites are subject to oxidation at moderately elevated temperatures, T>500 °C in air, and at T>700 °C in steam. Therefore, there is a need to protect them with surface coatings or sealants when used at elevated temperatures in an oxidizing environment and this constitutes one of the major disadvantages of CMCs. The other major drawbacks inhibiting their range of applications are the high manufacturing costs. Carbon–carbon production is expensive not only due to the costs of carbon ﬁbers but also the need to carry out repeat processing cycles irrespective of the manufacturing route followed.

Although both constituents in a carbon–carbon composite are the same element, each constituent can range in structure from carbon to graphite. Graphite contains carbon atoms arranged in a planar-based array in which the atoms are closed packed in hexagonal arrays. They are covalently bonded with sp² hybridization and the bond strength is 524 kJ mol⁻¹. The bonding between basal planes is van der Waal's with strengths of 7 kJ mol⁻¹. Interatomic distances within layers are 0.135 nm but the interlayer spacing is 0.335 nm. Such a structure imparts a high degree of anisotropy and accounts for many of its unusual properties not only as a bulk material but also for the ﬁber and matrix components of carbon–carbon composites.

A range of structures having varying degrees of crystalline perfection are possible in carbon. Solid carbons are generally derived from organic precursors by a pyrolysis process known as carbonization and may exist in graphitic or nongraphic forms. The various deﬁnitions of carbon forms and processes are comprehensively reviewed by Savage (1993). The most important of these are:

Carbonization – the process whereby a material of increasing carbon content is formed from an organic material, usually by pyrolysis, resulting in an almost pure carbon residue at temperatures above 1200 °C.
Graphitization – the solid-state transformation of metastable nongraphic carbon into graphite by the action of heat and/or pressure.

Figure 1 Schematic model of the change in lamellar structure of a graphitizing carbon with increase in heat treatment temperature (after Grifﬁths and Marsh, 1981).

A schematic diagram showing the progression from an irregular structure to a graphitized carbon is shown in Figure 1 (Grifﬁths and Marsh, 1981). A measure of the crystalline perfection is generally obtained from X-ray diffraction and may be characterized in terms of La, the equivalent diameter of crystalline order within the layer planes, and Lc', the mean height of the ordered layer or stack in the c-direction. In the case of carbon ﬁbers additional parameters are needed. Two La dimensions are needed, parallel and perpendicular to the ﬁber axis and a parameter Z which deﬁnes the misorientation of layer planes relative to the ﬁber axis (Johnson, 1980; Guigon et al., 1984; Guigon and Oberlin, 1986). Typical values of La and Lc for well-graphitized carbon can be several hundred nanometers, whereas a poorly graphitized carbon may have values < 10 nm. The net result is that a wide range of crystallographic forms of carbon exist. The anisotropy of graphite, the range of structural forms, the preferred orientation of crystallites, and the effect of other variables can result in a broad range of properties in carbon material. In carbon–carbon composites this range of properties can extend to both constituents. Couple these facts with the variety of processing techniques which can be employed in fabrication and the wide range of ﬁber lay-ups now available in modern processing technology, and it can be seen that enormous ﬂexibility exists in the design of carbon–carbon composites and their resultant properties.

The real birth of carbon–carbon composites came with the development of the ﬁrst carbon ﬁbers with acceptable tensile strength from rayon precursors in 1958–1960; a number of patents were ﬁled in this period (Soltes, 1958; Abbott, 1959; Ford and Mitchell, 1960). The ongoing development of carbon–carbon materials since the 1960s has resulted in the production of ﬁbers from new precursors (PAN and pitch) and the development of methods of depositing matrices for densiﬁcation. However, carbon–carbon materials are expensive and their applications have been limited to specialized applications where their unique combination of properties and their cost can be justiﬁed. The majority of applications are found in the aerospace industry and their application for missile nose-tips, re-entry heat shields, and rocket motor nozzles are well documented and represent the biggest ﬁeld of application in cost terms. Another application relies on their ability to conduct heat away without melting. For example, the Anglo–French Concorde would have been unable to land at commercial airports without the development of high energy consuming carbon–carbon brakes. Apart from aircraft, vehicles such as racing cars and high-speed trains beneﬁt from carbon–carbon brakes. Other areas of application include the nuclear industry (Torus walls for fusion reactors), biomedical applications, and industrial applications such as high-temperature tooling. The speciﬁc applications all necessitate different requirements which may be strength, stiffness, high thermal conductivity, or low thermal expansion (good thermal shock). These in turn inﬂuence the selection of ﬁber, the geometry of the reinforcement, and choice of matrix material, although other con- siderations such as availability and guarantee of supply are also important. Whatever the choices are, the production of composites may be represented by the schematic shown in Figure 2. A number of books and reviews have been written about CMCs (Savage, 1993; Thomas, 1993; Buckley and Edie, 1992).

In order to improve the bonding strength of the interface between carbon/carbon (C/C) composites and Ti–Ni–Si interlayer, SiC nanowires (NWs) were introduced in the interlayer as reinforcement materials (Qian et al., 2015). The SiC NW-reinforced joint was prepared by a two-step technique of electrophoretic deposition (EPD) and hot-pressing. The results show that with the increase of EPD time, the average shear strength of the joint initially increases and then decreases. The shear strength of the joint with EPD for 30 s reaches a maximum value of 30.44±2.28 MPa, which is 29% higher than that of the joint without SiC NWs. The improvement of the shear strength is primarily attributed to the toughening mechanism of SiC NWs by pull-out and bridging, which could play a positive role in extending the crack propagation path by consuming the applied load.

Figure 2 Schematic of the production routes of carbon–carbon ﬁber composites.

Rapid densiﬁcation of thick-walled carbon/carbon (C/C) composite tubes by electrically coupled chemical vapor inﬁltration (EC-CVI) was investigated (Wu et al., 2013). The results show that C/C composites were able to be densiﬁed by EC-CVI rapidly and efﬁciently. A thick-walled C/C composite tube, whose outer diameter, inner diameter, and height was 240, 60, and 200 mm, respectively, was able to be densiﬁed to 1.71 g cm⁻³ after a densiﬁcation time of 400 h. If the outer diameter and inner diameter were increased to 400 and 100 mm, respectively, the ﬁnal density of the component reached 1.50 g cm⁻³ under the same experimental condition. After densiﬁed by EC-CVI, the amount of the pores with a size larger than 100 mm decreased signiﬁcantly, and the porosity was even close to zero for the composites with a density of 1.62 g cm⁻³. It indicates that EC-CVI can effectively ﬁll the large pores in the composites. In the densiﬁed composite by EC-CVI, the pyro carbon on the carbon ﬁbers presented a rough laminar and smooth laminar morphology dominantly.

Multi-wall carbon nanotubes (MWCNTs) were chemically functionalized by 3-aminpropyltriethoxysilane and used to increase the strength and stiffness of an adhesive for joining carbon/carbon (C/C) composites (Zhang et al., 2012). When the content of silanized MWCNTs in the adhesive was 0.2 wt.%, average shear strength of the C/C joint was 10.40 MPa, which was 31% higher than that of neat C/C composites. The adhesive could be cured at room temperature with good heat-resistant property. The MWCNTs reacted with B 4C ﬁller to establish strong B–O–C bond with C/C substrate.

Carbon nanoﬁbers and nanotubes are promising to revolutionize several ﬁelds in material science and are suggested to open the way into nanotechnology (Hammel et al., 2004). People achieved bulk production capacities of high purity carbon nanoﬁbers (CNFs) at low cost by a catalytic chemical vapor deposition (CCVD) process. Reasonably low temperatures and yields of up to several g m⁻² min⁻¹ at more than 70% carbon gas-to-ﬁber conversion rates allow considerable cost reductions. In polypropylene the application of less than 10 wt.%. CNFs reduced the volume resistivity from >1013 𝞨 cm to a value of ~105 𝞨 cm. Different sizing of ﬁbers is normally used for nylon, polycarbonate, and other high temperature applications.

Various applications of carbon nanotubes as components of electrode materials for such electrochemical use as electrochemical capacitors, fuel cells, hydrogen electro sorption, and accumulators have been described (Grzegorz et al., 2011). Generally, carbon nanotubes give exceptional improvement of electrode performance due to their mesoporous and well conducting networks. The cell resistance is drastically reduced and the transport of ions is greatly enhanced. In addition to their good conductivity, carbon nanotubes can be ﬂexible and stretchable which is crucial for cyclability of electrodes, especially if volumetric changes of electrode material occur during operation. Consequently, they serve as excellent support for conducting polymers (e.g., Polyaniline, Polypyrrole) and metal oxides (e.g., MnO2) giving attractive capacitor electrodes. The presence of nanotubes in carbon precursor rich in heteroatoms, for example, nitrogen from melamine or oxygen, also supplies an interesting carbonized composite with good charge propagation for supercapacitor use. However, modiﬁcation of nanotubes can enhance hydrogen storage. On the other hand, carbon nanotubes can serve as an excellent additive to many electrode materials for improvement of conductivity and cell performance. They could be a good support of the catalytic particles for fuel cell application as well.

2. Carbon Fiber Reinforcement

Commercially available carbon ﬁbers can have carbon content in excess of 99.9% and their role is to reinforce the brittle carbon matrix. Since these ﬁbers exert a strong inﬂuence on the important resultant properties of the composite such as its strength and stiffness, it is necessary to understand the properties and microstructure of the ﬁbers. Furthermore, the shape and surface char- acteristics of the ﬁber inﬂuence the ability to manufacture preforms and the deposition behavior of the matrices. Within the scope of this chapter it is inappropriate to attempt a full description of ﬁber manufacture, so the subject will only be covered superﬁcially.

The carbon ﬁber industry uses three different precursor materials; rayon, PAN (poly acrylonitrile), and pitch, either isotropic or mesosphere. Rayon-based ﬁbers were the ﬁrst carbon ﬁbers to be produced commercially by Union Carbide in 1959 and represent the ﬁrst breakthrough by producing a carbon ﬁber using a polymer precursor. The probable mechanism for the conversion of rayon to carbon was worked out by Bacon and Tang (1964). The principle disadvantage of rayon is the low carbon yield of some 20–24%. The morphology of the carbon ﬁbers exhibits a crenulated surface, which is derived from the original precursor. Typical properties of rayon-based carbon ﬁbers are 1.0 GPa tensile strength, 40 GPa modulus, and 2.5% strain to failure. The combination of low carbon yield and poor mechanical properties has meant that rayon-based carbon ﬁbers have not proved competitive in the marketplace. However, they are used in a few applications, mainly in ablative technology, and manufacturing facilities for producing low modulus rayon-based ﬁbers still exist in the USA and France although they are operated at low level.

In the early 1960s, it was discovered that the PAN structure could be stabilized by an oxidation process which assisted controlled thermal decomposition during the carbonization stage to enable the production of carbon ﬁbers with superior mechanical properties to those made from rayon. The oxidative stabilization reaction was ﬁrst described by Watt (1972) and represents the most critical and lengthiest step in the conversion of PAN to carbon ﬁber because of the need to control the considerable heat evolution. After stabilization the oxidized PAN ﬁber is heated in an inert atmosphere, usually high-purity nitrogen. At a temperature between 500 and 600 °C, the organic material becomes carbon char but may still contain a high proportion of heteroatoms. In the case of PAN, a carbon ﬁber still contains as much as 6 wt.% nitrogen after heat treatment to 1000 °C. Heat treatment to 1600 °C is usually required to remove all the residual nitrogen.

Carbonization is usually carried out at temperatures between 1000 and 1500 °C resulting in a change of density from 1.45 to >1.7g cm⁻³. Since the chemical structure of the precursor is only imprecisely known, the exact chain of events during carbonization of PAN is not well deﬁned. A number of reactions have been proposed to account for observations of gas evolution, length changes, strength, and modulus during the conversion to carbon (Hay, 1968; Clarks and Bailey, 1973). At a temperature

between 500 and 600 °C, a change of state from insulator to conductor occurs. This presumably corresponds to the development of local molecular ordering as deﬁned by Oberlin (1984). After this only minor reorganizations of micro texture can occur.

The ﬁbers derived from PAN precursors can generally be divided into three categories:

Low modulus or Type III (E = 120–190 GPa).
Intermediate modulus or Type II (E = 220–250 GPa). These ﬁbers possess the highest tensile strength with good strain to failure (1.2–1.4%).
High modulus or Type I (E = 360–400 GPa). These ﬁbers offer improved stiffness at the expense of strain to failure.

Figure 3 shows how the strength and modulus of PAN-based ﬁbers changes with respect to the heat treatment temperature. The ﬁber types in general offered by most manufacturers are heat treated in the temperature ranges 1000–1200 °C (Type III), 1200–1700 °C (Type II), and >2500 °C (Type I).

Figure 3 Effect of heat treatment temperature on the strength and modulus of PAN-based carbon ﬁbers.

The ﬁrst manufacture of carbon ﬁbers from pitch was begun by Otani and co-workers in the early 1960s (Otani, 1965). Now a number of carbon ﬁbers made from pitch precursors are commercially available. Pitches used as a precursor for carbon ﬁbers are polyatomic molecules obtained as by-products of coal tar and petroleum processing. Since pitch is a thermoplastic it melts on heating so it can easily be melt spun to form pitch ﬁbers. Unfortunately the ﬁbers so formed have a low modulus (~40 GPa m⁻²) and strength (1 GPa m⁻²) as a result of their isotropic structure. In order to improve their properties a very costly and impractical process of hot stretching at 2700–3000 °C is needed (Delmonte, 1981). Hence, their use is limited to nonstructural applications and they are used as cheap ﬁllers in plastics and more recently concrete. Of far more interest and potential for commercial application are ﬁbers developed from mesophase pitch. Mesophase pitch can be produced by thermal or catalytic polymerization of a suitable petroleum or coal tar pitch. The preparation of a mesophase pitch from isotropic pitch typically involves heat treatment at 350–450 °C for periods of about 40 h. Some 45–65% will transform from the isotropic material to an optically anisotropic ﬂuid phase, the mesophase, or liquid crystals. This mesophase pitch may be used to produce high-modulus, high- strength carbon ﬁbers with a highly oriented structure using a variety of conventional spinning methods. Melt spinning, which is most commonly used, involves extruding the molten pitch with a gaseous atmosphere through nozzles directed downward so that the air-cooled ﬁbers are cooled and solidiﬁed (Figure 4). Typical ﬁber diameters would be 8–50 mm. A detailed description of the production of carbon ﬁbers from pitch precursors lies outside the scope of this chapter and the reader is referred to other works (Edie and Diefendorf, 1992).

The ﬁrst successful mesophase pitch-based carbon ﬁber was developed by Otani in the 1970s. In principle mesophase pitch should yield a lower cost, high-performance ﬁber. First, the precursor is relatively cheap at £0.1–0.2 kg⁻¹ compared with £0.5 kg⁻¹ for acrylonitrile. Second, the thermosetting process does not need tension to be applied to the ﬁbers since the spinning process imparts a high degree of molecular orientation as the pitch ﬁlament is being spun. Third, since the pitch-based ﬁber has a structure closer to that of the resultant carbon ﬁber, this implies that less energy is needed to convert to a carbon ﬁber. The carbon yield of pitch is 75–85%, much higher than that of PAN. Despite these obvious advantages high-performance pitch-based carbon ﬁbers currently sell for ⩾d60 kg⁻¹. They are unable to compete with PAN-based ﬁbers except for highly specialized applications requiring extreme stiffness (E4400 GPa), high thermal conductivity, low thermal expansion, and high-temperature oxidation resistance. The problems appear to arise because of processing difﬁculties (Hughes, 1987; Edie and Dunham, 1989).

A comparison of the properties of pitch-, PAN-, and rayon-based ﬁbers is given in Table 1. In general, ﬁber mechanical properties, thermal properties, and preferred basal plane orientation are inter-related. These properties are largely determined by the degree of preferred orientation of graphene layers along the ﬁber axis. Tensile strength is also inﬂuenced by radial structure and the presence of any ﬂaws in the structure. Wetting of the ﬁbers by the composite matrix and the strength of any resultant bond is strongly affected by the orientation of the graphene layers at the surface.

The structure of carbon ﬁbers has been studied by a number of authors using X-ray diffraction and electron microscopy observations (Johnson, 1987; Guigon et al., 1984; Ruland, 1990). The basic structural units are two-dimensional planar arrays associated in layers through p-bonding. Three-dimensional graphite structures result only after the highest temperature heat treatments. In general the graphene layer planes are oriented roughly parallel to the ﬁber axis in folded, convoluted structures. In general the structural aspects that are important are:

Crystallite sizes or coherent lengths parallel to and perpendicular to the carbon layers (La and Lc),
The texture or preferred orientation of carbon layers,
Volume fraction, shape, and orientation of micro voids.

Figure 4 Apparatus for the production of pitch-based ﬁbers. Reproduced with kind permission from Kluwer Academic Publishers from Edie et al., 1990. Carbon Fibers, Filaments and Composites, pp. 43–72.

Table 1 Properties of carbon ﬁbers

Typically the d002 spacing for carbonized ﬁbers is 0.350–0.353 nm, and for graphitized ﬁbers 0.337–0.343 nm. For crystalline graphite d002 is 0.3345 nm.

In general the interlayer spacing decreases with heat treatment temperature while La and Lc both increase and the disorientation of graphene layer planes from the ﬁber axis decreases. Figure 5 shows the effect of heat treatment temperature on the preferred orientation of PAN and mesophase pitch ﬁbers (LeMaistre and Diefendorf, 1973; D’Abate and Diefendorf, 1985). Similar trends are noted except that the curve for PAN is roughly 400 °C higher. It is seen to be difﬁcult to reduce the disorientation of graphene layers in PAN ﬁbers to o101 and so to obtain high moduli. Lc increases with increasing heat treatment temperature as do La║ and LaꞱ (Takaku and Shioya, 1990). The cross-section of a PAN carbon ﬁber can be different from core to skin. This was ﬁrst proposed by Johnson (1987) for Type I PAN ﬁbers. Figure 6 shows the proposed texture model illustrating the duplex structure (skin–core heterogeneity). The formation of the skin is probably a result of layer plane ordering which occurs as the heat treatment temperature is increased.

Unlike PAN-based carbon ﬁbers, pitch based ﬁbers show a variety of microstructures (Edie and Stoner, 1992) which can be produced by varying the spinning conditions of the liquid crystalline precursor.

Figure 5 Preferred orientation of graphene layer planes as a function of heat treatment temperature for PAN- and pitch-derived ﬁbers. Reproduced with kind permission from Kluwer Academic Publishers from ‘Carbon–Carbon Composites,’ 1993, p. 58.

Figure 6 Model for texture of Type 1 PAN-based ﬁber showing skin-core heterogeneity. Reproduced by permission of IOP Publishing Ltd., from J. Phys. D, 1987, 20, 286–291.

Carbon ﬁbers are available from a variety of suppliers in a number of yarns and tows with different speciﬁcations for moduli, strengths, and other properties. Tow sizes vary considerably between manufacturers but there is an increasing move to standardization at 1, 3, 6, 12, and 15 k for the aerospace and sports-goods markets. Larger tows of 24–320 k are produced primarily for automotive applications. They may also be purchased in a variety of forms including continuous tows, woven fabrics, and three-dimensional preforms. The diversity and tailor ability are one of their advantages but also a problem in that a complete evaluation of the various ﬁber types that are available on the market is difﬁcult to obtain.

The simple and continuous deposition of carbon nanotubes onto the surface of carbon ﬁber tows, using ethanol as a dispersive medium, was achieved by the electrophoretic process (Guo and Lu, 2012). The resulting materials showed a uniform distribution of carbon nanotubes on the ﬁber surface. Such a continuous process provides industrial potential for preparing multiscale carbon nanotube–carbon ﬁber reinforcement.

Needle-shaped 3C–SiC nanowires were obtained on the substrate of polyacrylonitrile (PAN) carbon ﬁbers using a simple thermal evaporation of silicon on the carbon ﬁbers (Jianjun et al., 2008). A strong photoluminescence peak located around 468 nm was observed at room temperature. The advantages of this process are that SiC nano needles were grown radially from the substrate of PAN carbon ﬁbers and PAN carbon ﬁber is electrically conductive. This sort of radial SiC nanowire array may ﬁnd practical applications in blue and ultraviolet light emitters, ﬁeld electron emission, reinforced composites, sensors, etc.

Activated carbon ﬁbers (ACF) are obtained mainly by physical activation with steam or carbon dioxide (Maciá-Agulló et al., 2004). ACF with high surface area are obtained by chemical activation with KOH and NaOH. Both chemical agents present different behavior; thus, NaOH developed the highest value of porosity and KOH developed samples with narrower micropore size distribution. The main observation is that by using chemical activation it is possible to obtain similar, or even higher, porosity (~1 ml g⁻¹, ~3000 m² g⁻¹) than by physical activation. However, chemical activation presents two important advantages: (1) a much higher yield (27–47% for chemical activation and 6% physical activation for ~2500 m² g⁻¹ activated carbon ﬁbers) and (2) the surface of the ﬁbers prepared by chemical activation is less damaged than by physical activation.

3. Textile Preforms for Carbon Matrix Composites

The earliest carbon–carbon composites prepared in the 1960s used unidirectional or bidirectional woven fabrics using low- modulus rayon-based carbon ﬁbers. Later, the availability of the ﬁrst PAN-based ﬁbers and ultimately pitch-based ﬁbers elevated these carbon–carbon composites to high-temperature structural materials whose high-temperature strength, toughness, and stiffness far exceeded those of other engineering materials. However, these mechanical properties were found only in the direction of the ﬁbers or, in the case of fabrics, in the plane of the fabric. Properties in a direction away from the ﬁber or fabric are poor, similar to that of monolithic carbon. The need to develop carbon–carbons that were more nearly isotropic in properties led to initiatives to devise methods for multidirectional reinforcement. Although the earliest and still the most popular preform is the orthogonal three-direction, 3D construction in which the ﬁbers are aligned with the X, Y, and Z axes, the technology now exists to construct, using mechanized techniques, more elaborate 4-, 5-, 7-, and 11-directional structures where the ﬁber directions are usually angled with respect to the orthogonal axes. It is worth mentioning at this stage that the two major applications for carbon–carbon are still aircraft brakes and rocket/spacecraft components (Fisher and Thomas, 1993). Whereas the former is a large volume, commercially oriented business where cost is a major factor, the latter is less affected by cost. The industry has therefore developed to the point where a large number of different manufacturing routes have been developed to serve a relatively limited number of applications. Hence, fabrication of a textile preform starts with the selection of a ﬁber, speciﬁcation of textile preform fabrication, and ﬁnally the densiﬁcation route. Although in principle any type of carbon ﬁber may be used, the choice will be governed by the requirements of the ﬁnal product. Low-modulus ﬁbers are easier to handle but if the end product is to be graphitized a high-modulus ﬁber should be used. Hence, at all stages of production of a carbon–carbon composite, to meet design requirements, there is a close cooperation between the textile and material design engineers.

Textile preforming consists of placing the required amounts of carbon ﬁber in the desired arrangement prior to densifying the composite. From the structural geometry point of view, the various levels of ﬁber architecture are classiﬁed into linear, planar, and 3D ﬁbrous assemblies. A detailed description lies outside the scope of this chapter and the reader is referred to excellent articles by Ko (1992) and Cunningham and Thomas (1997). The simplest system is the unidirectional or 1D fabric construction. This arrangement is very suitable for ﬁlament winding and angle-ply tape lay-up structures. The dry process involves winding resin-impregnated carbon ﬁbers on to a rotating mandrel in predetermined patterns (e.g., ±45°). In the wet method the ﬁbers pick up the matrix either by passing through a trough or a metered system. Filament winding permits great control over ﬁber placement and the uniformity of thickness of the structure. So, for example, such arrangements are attractive for the fabrication of rocket motor casings and pressure vessels where interlocking of yarns to make 2-D fabrics permits hundreds of possible combinations (Baillie et al., 1989). The possible combinations may be divided into biaxial and triaxial structures. Biaxial weaves consist of 0° and 90° yarns interlocked in various repeating patterns. The basic geometries are plain weave and satin weave (Figure 7). Plain weave, which is the most highly interlocked, is the tightest basic fabric design and most resistant to in-plane shear movement. Hence, it can be difﬁcult to wet during densiﬁcation. Basket weave is a variation of the plain weave in which warp and ﬁll yarns are paired. Satin weaves are a family of constructions with a minimum of interlacing. In these weaves the ﬁll yarns periodically skip over several yarns giving, for example, four-, ﬁve-, or eight-harness satin weave. Thus, in the four-harness satin weave shown in Figure 7, each ﬁll pick passes over four warp ends and then under one.

In the early 1970s, machine-made triaxially woven fabrics were developed by Dow and Tronﬁeld (1970). The chief advantage of the 0°–60°–120° hexagonal yarn orientation in one plane is a high level of in-plane shear resistance.

Figure 7 Details of plain and eight-harness section weaves.

However, where composites are to be used for structural applications, the low transverse tensile strength and inter laminar shear behavior often mitigates against the use of unidirectional and woven fabric composites. Hence, there was a strong impetus from the 1970s onwards to develop 3-D multidirectional ﬁber reinforcement composites. The major advantage of multidirectional carbon–carbon is the freedom to orient the ﬁbers and their amounts to accommodate the design levels of the component. The disadvantages of such multidirectional fabrication technology are the cost of the preform, size limitation on components that can be produced, and the difﬁculties of impregnating multidimensional arrays. Signiﬁcant progress has been made in recent years in automating 3-D fabric manufacturing by Aerospatiale (Herrick, 1978; Pastenbaugh, 1988), Brochier (Bruno et al., 1986), S.E.P. (Geoghegan, 1988) in France, G.E.C. and Fiber Materials Inc. in the USA (Stover et al., 1971; O’Shea, 1988), and Research Institute for Polymers in Japan (Fukuta et al., 1982). Good descriptions of 3-D fabrication technology are available and the reader is referred to these for more details (Ko et al., 1989; McAllister et al., 1983). The subject will only be covered brieﬂy here. The simplest type of multidirectional preform is based on a 3-D orthogonal construction and normally consists of yarn bundles located on cartesion coordinates X, Y, and Z. In its simplest terms this may be achieved by assembling a set of wires which will conform to the Z- direction. Carbon ﬁber tows housed in multiple carriers will then be passed between the wires alternately in the X- and Y-directions until the desired height is achieved. The vertical wires are then replaced with tows of carbon ﬁber by withdrawing each wire and threading a Z-direction tow into the resultant space. It is not always desirable to have equal numbers of ﬁbers or tows in each direction of the preform. Indeed depending on the application it may be positively advantageous to have more ﬁbers in one direction than in others. A large variety of structures such as those shown in Figure 8 (Lachmann et al., 1978) are possible. Polar weave preforms are used to form cylinders and other shapes of revolution. They are 3-D constructions with yarns oriented on polar coordinates in the radial, axial, and circumferential directions.

Figure 8 Some common 3-D orthogonal weave constructions. Reproduced with permission of the ASME from Lachman et al., 1978. Proceedings of the International Conference on Composite Materials, pp. 1302–1319.

Although the 3-D structure overcomes the problems of poor mechanical properties perpendicular to 1- and 2-D laminate planes, there remain weaknesses in non-ﬁber directions such as 45° to the X- and Y-axes. In order to overcome this, more complex structures have been devised and built using the same basic machinery and technique. So, for example, a four-directional conﬁguration can be achieved by weaving three sets of tows at 60° between the Z-direction tows. In a ﬁve-directional pentaxial conﬁguration, V, W, X, and Y tows are woven at 45° angles. Further improvements can be made by adding other diagonal tows to the basic 3-D structure; for example, combining either four face diagonals with the three orthogonal leads to a seven-directional conﬁguration (Figure 9; Stover and Latva, 1973). The most complex arrangement yet reported is an 11-directional structure which combines face and body diagonals with the orthogonal. Typically, unidirectional tape can build up a ﬁber volume fraction of ~65%, although theoretically 90% is possible. For a 2-D weave the upper limit for ﬁber volume fraction is 50–60%, whereas for the 3-D structure 40–55% ﬁber volume fractions are typically achievable.

4. Carbon–Carbon Matrix Materials

The matrix is a vitally important element in carbon–carbon composites. It acts as a binder maintaining the alignment of the ﬁbers and ﬁber bundles, also isolating the ﬁbers from one another. It serves as a stress distributor transferring the external load applied to the composite to the reinforcing ﬁlaments. As already indicated, carbon materials can exhibit a wide range of structures and textures ranging from near amorphous to fully graphitic structures. These are controlled by the nature of the precursor material, how it is processed to yield carbon, and the ultimate heat treatment temperature experienced during processing. The choice of carbon ﬁber, its surface morphology, whether it has had any surface treatment, and the weave patterns of the arrays of carbon ﬁbers are also important and lead to a variety of processing options. In general, carbon–carbon composites are processed in one of three ways (or a combination of two of these):

Chemical vapor deposition/inﬁltration,
Impregnation/pyrolysis using resin precursors,
Impregnation/pyrolysis using pitch precursors.

Generally the CVD process occurs at temperatures of typically 800–1500 °C. Pyrolysis of the resin and pitch precursors typically occurs in the same temperature range. Subsequent heat treatments may involve temperatures up to 3000 °C. These three methods lead to very different microstructures in the composite, partly because the method of laying down the carbon is different but mainly because the different precursors yield carbon forms having different structure and properties. A schematic of the pore ﬁlling mechanisms is shown in Figure 10 (Fitzer, 1987). So, for example, CVD lays down the carbon directly onto the ﬁbers, whereas with liquid precursors the carbon is produced in the void between the ﬁbers after a heat treatment. The resultant volumetric shrinkage results in signiﬁcant porosity and pores and shrinkage cracks are common in the matrix. To a certain extent the choice of densiﬁcation method depends on the geometry of the part being processed. Since CVD processing tends to deposit primarily on the near-surface region of the part, it tends to have its greatest uses for thin sections. Thick sections tend to be produced using pitch or resin impregnation. Whatever the production route, multiple impregnations are necessary to achieve high densities and optimum properties (Figure 2).

Figure 9 Seven-direction ﬁber arrangements. Reproduced by permission of the American Carbon Society from ‘Proceedings of the 11th Biennial Conference on Carbon,’ 1973, p. 277.

4.1 Chemical Vapor Deposition

Chemical vapor deposition (CVD) involves the thermal decomposition of a hydrocarbon vapor, usually an inert gas carrier, over the hot substrate on to which the carbon is deposited. Fitzer et al. (1971) have reviewed the thermodynamic stability of a wide range of gaseous precursors. Among the commonly used precursors for carbon deposition one may list CH4, C2HO, C2H4, C2H2, etc., of which methane CH4 is the most widely used. Temperatures in excess of 550 °C are required before carbon deposition is thermodynamically favorable. The kinetics and mechanisms of carbon deposition have been reviewed by Kotlensky et al. (1973) and Bokros et al. (1969). The CVD of carbon from a hydrocarbon gas within a substrate is a complex process. It is controlled by the substrate geometry and deposition conditions, particularly gas ﬂow rate and concentration, pressure, and temperature. Although a full discussion of the CVD of carbon ﬁber preforms lies outside the scope of this chapter, it is pertinent to mention the different types of carbon that have been identiﬁed. These have been reviewed by Bokros et al. (1969) as smooth laminar, isotropic, and rough laminar.

Essentially, smooth laminar is obtained at low temperatures (<1300 °C) and intermediate hydrocarbon concentration and gas pressure, whereas rough laminar occurs at intermediate deposition temperatures and intermediate partial pressure. Isotropic coatings require a high deposition temperature and low partial pressure. Pierson and Liebermann (1975) developed a chemical model for the CVD of carbon in an attempt to rationalize the observed microstructure with experimental conditions. Oh and Lee (1988a,b) summarized the effect of deposition conditions on the microstructure of carbon matrix decomposition from propane and identiﬁes four regions. Essentially, high-temperature low propane conditions favor isotropic deposits and low-temperature high propane concentrations favor a columnar structure (Figure 11). Jacklewski and Diefendorf (1981) essentially conﬁrmed these ﬁndings. Given that the CVD process offers considerable scope for manipulation of the microstructure, what is the optimum microstructure for a carbon–carbon composite? Granoff et al. (1973) concluded that isotropic carbon had too low a density and that smooth laminar carbon was prone to thermal stress microcracking. They concluded that rough laminar produced the optimum microstructure. Oh and Lee (1988a,b) considered that for maximum modulus the smooth laminar deposit was preferred.

Figure 10 Schematic mechanisms of pore ﬁlling and pore blocking by liquid impregnation and CVD (After Fitzer, 1987).

Figure 11 Effect of deposition conditions on microstructure of pyrolytic carbon matrix deposited from propane. (After Oh and Lee, 1988a,b).

Various techniques have been developed to inﬁltrate carbon ﬁber substrates. These include isothermal, thermal gradient, and pressure gradient. The isothermal technique is the most widely used and most brake materials are made by this route (Fisher and Thomas, 1993). As the name implies, the substrate is placed in an even temperature furnace and the reactant gases passed over it. This relies on diffusion in and out of the pores. To avoid sealing the pores off too early, the surface reaction rate needs to be slower than the diffusion rate. The slow rate of deposition is achieved by operating under a reduced pressure (10–100 mbar) and at a low temperature, typically 1100 °C. This means unfortunately that the rates of weight ga+ in are low and hence processing times are very long. Eventually pore blockage occurs so that preforms need to be removed from the furnace, surface machined, and then re-densiﬁed. Hence, to build up the density levels to produce the desired mechanical and thermal properties requires runs of several hundred hours and multiple inﬁltration cycles. In attempts to reduce processing times the thermal gradient and pressure gradient CVI processes were developed (Kotlensky et al., 1973). In the former the part to be inﬁltrated is supported by a mandrel which is itself inductively heated. Thus, the hottest portion of the substrate is the inside surface which is in close proximity to the mandrel. The outer surface is exposed to a cooler environment usually by the proximity of the water-cooled conduction coils. It is important that the ﬁber preform have a low thermal conductivity in order to establish the temperature gradient. The CVI carbon is deposited ﬁrst on the outside surface. As density builds up the densiﬁed substrate couples inductively, begins to heat up, and the densiﬁcation front gradually moves through the component. The big advantage of this process is its relative rapidity and the fact that it can operate at atmospheric pressure. The big disadvantage is that each part needs its own susceptor. Hence, it is mostly applied to rocket components.

The pressure gradient method relies on the forced ﬂow of precursor gas through the pore system of the ﬁber preform, thus removing the surface reaction limitation of the isothermal CVD method. The ﬁber preform is sealed off from the furnace chamber at the base. The hydrocarbon gases are fed into the inner cavity at a positive pressure with respect to the furnace chamber. A pressure difference that forces the gas to ﬂow through the pores is created across the wall of the structure, depositing carbon and exiting as hydrogen. Although high deposition rates may be obtained there are severe drawbacks such as the dependency on robust high temperature pressure seals. The method is not in widespread commercial use.

The chemical vapor deposition (CVD) and the chemical vapor inﬁltration (CVI) processes of carbon materials are reviewed starting from the historical aspects and including the latest developments in the preparation of C/C composites (Delhaes, 2002). A complementary aspect is the structural and physical analysis of the deposited pyro-carbons: they are considered as reproducible metastable phases which are obtained under non-equilibrium thermodynamic conditions. The ﬁnal relevant point concerns the relationship between the process parameters and the type of pyro-carbon.

The synthesis of bulk amounts of high quality single-walled carbon nanotubes (SWNTs) is accomplished by optimizing the chemical compositions and textural properties of the catalyst material used in the chemical vapor deposition (CVD) of methane (Alan et al., 1999). A series of catalysts are derived by systematically varying the catalytic metal compounds and support materials. The optimized catalysts consist of Fe/Mo bimetallic species supported on a novel silica — alumina multicomponent material. The high SWNT yielding catalyst exhibits high surface-area and large meso-pore volume at elevated temperatures. Gram quantities of SWNT materials have been synthesized in B0.5 h using the optimized catalyst material. The nanotube material consists of individual and bundled SWNTs that are free of defects and amorphous carbon coating.

4.2 Liquid Precursors

As an alternative to CVI two types of liquid impregnant are used to densify carbon preforms. The ﬁrst is resins or polymers and the second type is pitch which may be coal tar or petroleum derived. Good reviews of the use of these matrix precursors are given by Rand and Thomas (1997) and Savage (1993). Thermosetting resins are potentially very attractive matrix precursors because they allow conventional polymer matrix composite fabrication techniques to be used prior to carbonization. Although the number of resins that could be used is potentially almost limitless, when processing, variables such as viscosity, cure conditions, shrinkage during carbonization, matrix microstructure but most of all carbon yield are considered, the choice comes down to relatively few. Good precursors are phenolic resins, related resins from furfural alcohol as well as polyimides and polyphenylenes. These tend to have carbon yields of 45–60% (Sandor, 1991) although resins with a char yield of over 80% have been developed and a value as high as 95% has been reported for an aromatic di-acetylene di-polymer (Economy et al., 1992).

The resins normally used are dissolved in an organic solvent or furfural alcohol with an acidic catalyst. Resin matrix composites are usually fabricated from a pre-impregnated woven cloth carbon ﬁber layer, often known as a prepreg. These prepregs are partially cured to a degree of tackiness, cut to size, and then laid up in the desired array. The resin on the impregnated composite is cured or pyrolyzed by heating it to a temperature in the range 350–800 °C. During this curing process a large number of volatiles are expelled such as H2O, H2, CH4, CO, and CO2 (Figure 12; Rand and McEnaney, 1985). Frequently hot pressing at pressures up to 10 MPa and at temperatures in the range 150–350 °C for periods up to 10 h are used to enhance the densiﬁcation process. The pyrolyzed composite is subsequently carbonized at temperatures in excess of 1000 °C.

The carbon materials produced by pyrolysis of cross-linked resins are considered to be non-graphitizable, consisting of small basic structured units on a nanoscale (Jenkins and Kawamura, 1976). Associated with this structure is nanoscale porosity which becomes closed, however, after heat treatment to temperatures over 1000 °C. This texture means that the material is essentially isotropic and no evidence of any graphite structure is observed from X-ray analysis.

However, it has been found that when carbonization or heat treatment is carried out under stress or in the presence of certain catalysts, anisotropic regions begin to develop in the matrix at temperatures as low as 1000 °C (Fishbach and Walkerm, 1971; Noda and Kato, 1965; Oberlin and Rousseau, 1968; Kamiya and Inagaki, 1981). In composites it is believed that there is some interfacial effect occurring between carbon ﬁber and matrix during heat treatment. During carbonization the resin shrinks by as much as 50%, whereas the carbon ﬁbers change very little in dimension. It has been postulated that the driving force for graphitization is the stress accumulation caused by differences in coefﬁcients of thermal expansion between ﬁbers and matrix (Kamiya and Inagaki, 1981). It has been observed by Hishiyama et al. (1974) using a polyfurfuryl alcohol precursor and PAN ﬁbers that the carbon basal planes tend to align along the ﬁber axis and that after heat treatment at temperatures as low as 1800 °C all the matrix carbon between ﬁbers is essentially anisotropic.

Figure 12 Evolution of volatile matter from phenol–formaldehyde resin during carbonization. Reproduced by permission of the Institute of Materials from Br. Ceram. Trans. J., 1985, 84, 157–165.

The above observations relate to untreated carbon ﬁbers. It has been shown that if surface treated ﬁbers are used, weak brittle composites result because the matrices are too strongly bonded to the ﬁbers (Thomas and Walker, 1978a,b).

However, when the composites are weakly bonded, the shrinkage during cooling after carbonization results in the matrix shrinks away from the ﬁbers and matrix cracking. Multiple impregnations are therefore necessary to achieve an acceptable density. Surface machining to open up surface porosity and thus assist further resin impregnation is frequently employed. Sometimes composites are given an intermediate graphitization at up to 2500 °C to aid the reimpregnation process. Because of the relatively low carbon yields multiple impregnations, usually 4–6, are needed to achieve an acceptable density. Unfortunately each successive reimpregnation cycle results in a smaller density increase so a balance needs to be struck between the advantages to be gained from a reimpregnation in terms of improved properties and the cost.

Pitches on the other hand are very attractive precursors for carbon matrices because they have high carbon yields and are graphitizable. Pitches may be derived from petroleum or coal tar residues and are complex mixtures of many organic compounds. The carbon yield depends very much on the composition of the precursor pitch and the pyrolysis conditions. Details of pitch characterization and pyrolysis lie outside the scope of this chapter and have been reviewed by Rand (1987) and Rand et al. (1989). The pyrolysis of pitch occurs in the liquid state via evaporation of species of increasing molecular weights as the temperature is raised. However, at temperatures of 385–400 °C cracking reactions take place, releasing low molecular weight aromatic fragments as the side chains to the poly-aromatic molecules remaining in the ﬂuid are severed. The most signiﬁcant feature in the liquid phase pyrolysis is the development of spheres of liquid crystal or mesophase in the pyrolyzing liquid. These spheres, initially about 100 mm diameter, are precursors of the graphitic structure (Brooks et al., 1968). It has been known for many years that mesophase development can be affected by surfaces. There is a strong tendency for the lamellae to align with respect to the surfaces. This effect is very important in composites where the mesophase aligns preferentially with the ﬁber surface producing a region of preferred orientation which extends out into the matrix surrounding the ﬁbers. However, when the matrix is carbonized under pressure, the orientation effects are modiﬁed even to the extent that alignment of matrix is normal to the ﬁber surface (Evangelides, 1977; Stover et al., 1977). This effect is not fully understood.

Figure 13 HIPIC pressure carbonizing furnace. (After Hosemura and Okamoto, 1991).

The carbonization process can be characterized by the viscosity changes experienced by the pitch. Essentially, heating from 20 °C results in melting of the solid isotropic pitch and a dramatic reduction in viscosity. There is then a negligible change in viscosity until 450 °C when the mesophase begins to develop. Although the rate of mesophase formation is affected by many factors such as composition, temperature, and removal rate of the lower molecular weight fraction, the rate increases as the temperature increases as indeed does the viscosity. Higher temperatures and longer times accelerate the process until the pitch becomes a brittle predominantly crystalline solid.

Mesophase pitches are known to bloat seriously upon carbonization and this has been shown to commence with coalescence of the bulk mesophase (White, 1976; Weinberg et al., 1983). Additionally the yield of carbon from pitches can be greatly increased from ~50 to 90% by charring under high pressure. As a consequence the pitch route for producing a CMC is normally carried out under pressure and heat in an autoclave. This process is usually called HIPIC (hot isostatic pressure impregnation with carbon). Typically (Figure 13) a carbon ﬁber preform is vacuum impregnated with molten pitch, placed inside a metal container, and surrounded by an excess of pitch. This can is then evacuated and sealed. The sealed can is placed inside the HIP unit and then the temperature is raised at a programmed rate to above the melting point of the pitch. The pressure is then increased to around 100 Mbar. As the pitch melts it expands and is forced by isostatic pressure into the pores of the sample. The temperature is then gradually increased to that for pitch carbonization (650–1000 °C). After treatment the preform is cleaned up by machining the surface (Gray and Savage, 1991; McAllister et al., 1983).

In order to improve the anode characteristics of pitch-based carbon ﬁbers for Li secondary batteries, two kinds of heat treatment procedures were adopted, i.e., in vacuum and in a partially oxidizing atmosphere (Kikuchi et al., 1995). Heating to 980 °C in vacuum was found to remove the surface hydroxyls and adsorbed water effectively. This method improved the high current charge-discharge capacities and cyclability of the carbon ﬁbers.

5. Composite Microstructure

The microstructures of the various types of CMCs are quite complex and will differ according to the types of precursor materials and the processing conditions. The microstructure may be inﬂuenced by:

Fiber type,
Preform geometry (1-D, 2-D, ⩾3-D),
Any ﬁber surface treatment,
Number of impregnations/inﬁltrations,
Inclusion of oxidation protection,
Carbonization/graphitization cycles,
Combination of densiﬁcation routes (e.g., CVI and pitch).

The number of possible combinations of parameters is very large. It is not realistic to attempt to produce a catalog of the structures of carbon–carbon so general trends and observations associated with the different variables will be reviewed. The microstructure of CMCs has been reviewed by a number of authors (Murdie et al., 1988; Rand and Thomas, 1997) and the reader is referred to these reviews for a more detailed discussion. Two features are of primary importance in the microstructure of carbon–carbon composites, namely the orientations of the graphite layer planes relative to the ﬁbers and the geometry of the porosity and voids within the composite. The largest scale at which one might view the microstructures of CMCs is the sub- millimeter scale where the composite architecture is deﬁned. The way in which the ﬁber bundles are arranged deﬁnes the overall geometry of the matrix carbon. So, for example, for a 3-D orthogonal arrangement of ﬁbers, roughly one-quarter of the unit cell will be ﬁlled by matrix graphite alone. Increasingly smaller scales of magniﬁcation reveal ﬁber and matrix texture. At the nan- ometer scale the internal structures of the ﬁbers themselves and the structure of the ﬁber–matrix interface are revealed and at very high magniﬁcation features such as the basic structural units and local molecular ordering are seen. In addition to this complex range of features there is a wide range of internal porosity which profoundly affects the properties. These ranges from nearly equiaxed pores such as those formed by bubble percolation during carbonization to high aspect ratio cracks at interfaces. Pores and cracks also range in size from the millimeter scale such as bundle–bundle interface cracks to the nanometer scale such as micro pores between the basic structural units (Oberlin and Thrower, 1989). Jortner (1986) classiﬁes mechanical properties into mini- mechanical properties which are inﬂuenced primarily by composite architecture and micromechanical properties which are inﬂuenced by ﬁber–matrix interfaces.

Recalling Section 4, the two methods of depositing the matrix, either liquid phase using pitch or resin precursors or by inﬁltration with a hydrocarbon gas, generally lead to two different microstructures in the composites. This is in part due to the fact that the method of laying down the carbon is different and partly because the different precursors yield carbon forms having different structure and properties (Fitzer, 1987). Perhaps the most signiﬁcant difference is that the CVD process lays down the carbon directly on to the ﬁbers themselves, whereas with liquid precursors the carbon is produced in the spaces between the ﬁbers after a carbonization heat treatment and this produces a substantial volumetric shrinkage. Thus, the distribution of the carbon matrix in the composite can be very different and pores and shrinkage cracks are common in the matrix. These are ﬁlled by subsequent rein ﬁltrations followed by pyrolysis/heat treatment operations to densify the composite.

The microstructure of matrix carbon produced by CVI depends on the deposition conditions. The kinetics and mechanisms of carbon deposition by chemical vapor inﬁltration have been reviewed by Kotlensky et al. (1973), Bokros et al. (1969), and Fitzer et al. (1971) have reviewed the thermodynamic stability of a wide range of gaseous precursors. The important aspects with respect to the structure of the CVD matrix are the deposition conditions such as concentration, pressure, temperature, as well as substrate geometry. As mentioned earlier the matrix deposited by CVD can be of three different types, characterized by Bokros et al. (1969) as follows:

Smooth laminar. This structure is formed at low temperatures (<1400 °C) and intermediate hydrocarbon concentration and gas pressure. The structure is turbostatic and weakly anisotropic layers are oriented to surround the ﬁber uniformly.
Isotropic. This type normally forms in the temperature range 1400–1900 °C under conditions of higher gas pressure and
hydrocarbon concentration but using a low ﬂow rate. This consists of a somewhat porous carbon deposit comprising an optically isotropic morphology of ﬁne particles of a few micrometers in size.
Rough laminar. At very low temperatures and low hydrocarbon pressures or at very high temperatures only low growth rates are possible and so dense deposits with well-deﬁned growth features are observed. The rough laminar structure comprises a combination of layers with strong optical anisotropy oriented to surround the ﬁber completely.

The major problem in CVI is to achieve a uniform deposition of carbon within the preform. This means that the rate of deposition must be much lower than the rate of inﬁltration of the gas into the porous preform. This constrains the deposition conditions to be both low temperature and low hydrocarbon vapor pressure, thus favoring the rough laminar microstructure. However, in the thermal gradient CVI technique the temperature gradient across the sample will allow the use of higher deposition temperatures so other microstructural types are possible. A particular feature of CVI densiﬁed composites is the presence of elliptical closed pores which are formed when the CVI deposit closes narrow pore necks, thus preventing further inﬁltration.

When carbon matrices are produced from thermosetting resins the resin will usually be used in solution form. Hence, evaporation of the solution will take place and the condensation reactions leading to cross-linking will also release water vapor (Yamashita and Ouchi, 1974). A signiﬁcant volumetric shrinkage occurs. The microstructure of a composite produced from the pyrolysis of a thermoset resin will vary dramatically according to the heat treatment temperature. When carbonized by itself, a thermoset resin forms a glassy, isotropic carbon that is considered as non-graphitizable. This material is made up of basic structural units of small dimensions with no signiﬁcant local molecular ordering. Such a structure can be found in matrix-rich regions of a carbon–carbon composite remote from the ﬁber bundles. However, when the resin is heat treated in composite containing carbon ﬁbers, graphitic material is observed in the region of the interface with the ﬁbers. This is believed to be due to the restraint exerted on the matrix in the vicinity of the ﬁbers so that when shrinkage occurs during carbonization the resultant stresses cause graphitization of the resin carbon. Oberlin and Rousseau (1968) and Kamiya and Inagaki (1981) observed that the graphitic regions tend to form at contact points where stress is concentrated. The graphite microstructure in thermoset-derived matrices is oriented in such a way that the graphite layer planes encircle the ﬁbers (Zimmer and White, 1983; Hishiyama et al., 1974). Similar effects have been noted for phenolic resin and rayon-, pitch-, and PAN derived carbon ﬁbers after graphitization to 3000 °C (Manocha, 1986). The anisotropy was discontinuous in the vicinity of rayon ﬁbers, but in the form of long areas with pitch-based ﬁbers; PAN ﬁbers showed intermediate behavior.

The microstructure of ex-thermoset composites is dominated by large-scale porosity and shrinkage cracks resulting from mass loss and volatilization of the various molecular species during pyrolysis. The interaction of the resin with the ﬁber surface is again important in determining the location of the cracks. The ﬁrst carbonization cycle is critical. If the precursor is well bonded to the ﬁber as, for example, with surface-treated HT ﬁbers, the composite exhibits high shrinkage, extensive cracking in the body of the matrix, and possible ﬁber damage (Fitzer, 1987). Thus, the composite microstructure consists of clumps of ﬁbers which are bonded together but separated by large voids. However, if ﬁber–matrix interaction is poor, for example, with un-surface treated carbon ﬁbers, the matrix shrinks away from the ﬁbers leaving ﬁssures or annular gaps at the interface.

Heat treatment to temperatures above 2000 °C results in further shrinkage of the carbon matrix as it gradually transforms to graphite. The different form and distribution of the porosity developed on carbonization leads to different properties in the composite. The porosity generated forms the locations for further resin penetration on re-inﬁltration.

The second liquid phase inﬁltration route using pitch has many features of similarity to resin inﬁltration and pyrolysis. The major difference is that carbonization is usually carried out under pressure (see Section 4). This reduces volatilization and shrinkage and increases carbon yield from 50 to 85% (McAllister et al., 1983; Huttinger et al., 1989). Matrix carbons resulting from mesophase development within the pitch are inﬂuenced considerably by the composite architecture. It has been known for many years that mesophase development is affected by surfaces. In composites the mesophase tends to align parallel to ﬁber surfaces producing a region of preferred orientation which extends out into the surrounding matrix (Zimmer and White, 1983). Within a ﬁber bundle where the volume fraction of ﬁbers is high, the inter-lamellar matrix is so oriented. This orientation profoundly affects the modulus and thermal conductivity. However, variations in processing conditions have been reported to affect the matrix orientation. Whereas in low-pressure carbonization, alignment of the graphitic layer planes is by viscous ﬂow of the mesophase along the ﬁber surface, when the matrix is carbonized under pressure the orientation effects are modiﬁed. High pressures favor the formation of a transversely aligned matrix where the layer planes are normal to the ﬁber surface (Evangelides, 1977; Stover et al., 1977). This effect is not clearly understood. It appears that the surface character of the ﬁbers is important since this effect was not observed in ﬁbers coated with a thin CVD layer of carbon. Other factors do, however, seem to be important because Murdie et al. (1988) reported that both parallel and transversely oriented graphite were produced in a specimen carbonized under atmospheric pressure.

During the conversion of pitch matrices from a viscous solid to a polycrystalline solid in the temperature range 600–900 °C there is extensive bulk shrinkage. The net result is the formation of cracks which tend to run parallel to the lamellar structure of the solid. Cracks will also form due to the differential contraction between matrix and ﬁber during cool down from the processing temperature. The location of these depends on the strength of the ﬁber–matrix bond. In the case of a strong bond, shrinkage cracks tend to be located within the matrix, whereas with a weaker bond they occur predominantly at the ﬁber–matrix interface.

All CMCs contain pores and cracks which are extremely important features in determining thermal and mechanical properties. A major advantage of the network of pores and cracks are that they provide conduits for further impregnation in the densiﬁcation of carbon–carbon. This is particularly important since the majority of commercially produced composites receive 3–6 reim- pregnations to build up the density. An excellent review by Rand and Thomas (1997) contains a table which compares the characteristic features and properties of carbon matrices from the different precursors. This is reproduced in its entirety in Table 2. The failure behavior and morphology of a carbon–carbon composite (C–C composite) manufactured by isothermal chemical vapor inﬁltration was studied by three-point bending tests, polarized light microscope and scanning electron microscope, respectively (Jing et al., 2014). The C–C composite was reinforced by PAN-based carbon ﬁber aligned in only one direction. Flexural strength and modulus of the composite were 200.9 MPa and 50.5 GPa, respectively. Failure behavior of the unidirectional C–C composite has described in three stages including brittle fracture behavior at beginning, quasi-ductile behavior ﬁnally, and ﬂuctuation behavior between them. Two main kinds of cracks, namely cracks parallel and perpendicular to loading direction alternately resulted in deformation evolution of the composite. The strength of interfacial bonding and cracks orientation played key roles to failure behavior of C–C composite.

6. Thermal Properties of Carbon Matrix Composites

There is considerable interest in developing high thermal conductivity carbon–carbon composites. These would have many important applications such as ﬁrst wall tiles for nuclear fusion reactors, hypersonic aircraft, thermal radiator panels, and electronic heat sinks. To this end, much attention has been devoted to the development of high thermal conductivity carbon ﬁbers. However, with the wide range of possible ﬁber–matrix combinations, a very large number of processing permutations are possible. This wide range of options means not only that a wide range of thermal conductivity is possible, but there is the opportunity to 'tailor' the thermophysical properties of carbon materials. It is generally accepted that carbon and graphite are phonon conductors and >99% of heat is transported by phonons or quantized lattice vibrations (Kelly et al., 1969). These may be scattered by other lattice vibrations (Umklapp scattering) or by any defects present in the crystal lattice. These may be added in reciprocal manner

where 𝞴tot is the measured thermal conductivity, 𝞴u is the contribution due to 𝞢 𝞴p.d. scattering, and lp:d: represents the additive contributions due to phonon scattering by point defects.

Table 2 Comparison of characteristic features and properties of carbon matrices from vapor phase, pitch, and resin precursors

Graphite is also unique among materials in having an extremely high anistropy of thermal conductivity. Within the layer planes the thermal conductivity is more than 200 times the out-of-plane thermal conductivity (Taylor, 1966) making graphite a very good thermal conductor in two directions and a virtual insulator in the third. Hence, the thermal conductivity of any graphite assembly is critically dependent on the orientation of the layer planes.

Kelly and co-workers in an excellent series of papers have calculated the thermal conductivity contribution for scattering due to the various defect types present in graphite and these have been summarized in a review article (Kelly et al., 1969). The most signiﬁcant contribution is grain boundary scattering (lg:b:). Of the remainder only isotope scattering makes a contribution amounting to some 1%. However, as postulated by Taylor et al. (1968), the conduction in polycrystalline graphite will be affected by voids, porosity, and the tortuous heat conduction path which depends on the orientation of layer planes. These they group together as a porosity/tortuosity factor, a. Hence, the thermal conductivity may be written

The mean free path for Umklapp scattering has an exponential temperature dependence (Taylor, 1966) which gives way to a linear dependence of thermal resistivity with temperature only at very high temperatures. The relative contributions to thermal conductivity depend on the crystallite size and this has been presented by Kelly and Gilchrist (1969) and Taylor et al. (1993). The net effect is that, as the crystallite size decreases, the thermal conductivity decreases and the thermal conductivity peak shifts to a higher temperature (Figure 14).

Bokros et al. (1969) shows the temperature dependence of the thermal conductivity for crystalline and noncrystalline carbon material (Figure 15). Thus to a ﬁrst approximation, knowledge of the crystallite size would enable a prediction of the thermal conductivity and the porosity/tortuosity factor gives a good indication of the alignment of layer planes in any speciﬁc orientation. These arguments would apply to both ﬁbers and matrix in carbon–carbon composites.

If we ﬁrst consider the ﬁbers then there is a strong thrust to developing high thermal conductivity carbon ﬁbers and a number of pitch based ﬁbers are now offered with claimed thermal conductivity as high as 1200 W m⁻¹ K⁻¹, although no information is available showing how these were measured. There is in fact surprisingly little published information on determinations of the thermal conductivity of carbon ﬁbers, which is critically dependent on the orientation of layer planes parallel to the ﬁber axis.

Figure 14 Predicted thermal conductivity of perfectly oriented fully dense polycrystalline graphite as a function of grain size La. Reproduced by permission of Pion from High Temperatures–High Pressures, 1993, 25, 443–450.

Figure 15 Temperature dependence of thermal conductivity for crystalline and noncrystalline carbon (after Bokros et al., 1969).

This in turn is dependent on ﬁber type and the temperature to which the ﬁber has been heat-treated. Volga et al. (1973) have measured an increase in room temperature thermal conductivity from 30 W m⁻¹ K⁻¹ for ﬁbers heat treated at 1400 °C to 320 W m⁻¹ K⁻¹ for ﬁbers heat treated to 2600 °C. In general, low modulus, high-strength ﬁbers have thermal conductivities in the range 5–50 W m⁻¹ K⁻¹, whereas the thermal conductivity of high-modulus ﬁbers is signiﬁcantly higher, up to 200 W m⁻¹ K⁻¹. Mesophase pitch ﬁbers generally exhibit higher thermal conductivities than PAN-based ﬁbers for similar processing conditions. While a number of room temperature measurements have been reported, the only systematic investigation of the variation in thermal conductivity with temperature of a range of carbon ﬁbers was carried out by Taylor et al. (1993) who measured the thermal diffusivity of four PAN-based ﬁbers and three pitch-derived ﬁbers. The results for seven of the ﬁbers are shown in Figure 16 and the thermal conductivities at 400 K for the derived crystallite sizes parallel to the ﬁber axis are shown in Figure 17.

In contrast, very little is known about the transverse thermal conductivity of carbon ﬁbers but it has been variously estimated to be 2–10 W m⁻¹ K⁻¹ (Pilling et al., 1979; Whittaker et al., 1990a,b).

When ﬁbers are incorporated into a carbon matrix to make a composite, the resultant thermal conductivity is not only inﬂuenced by the ﬁber architecture but also the geometry of any porosity. A composite is therefore a ternary system composed of ﬁbers, matrix, and porosity. The latter acts as a barrier to heat ﬂow so the geometry and orientation dependence of any porosity is crucial and orientation dependence is clearly inﬂuenced by ﬁber architecture (Mottram and Taylor, 1991). However, there is a common misconception that the thermal conductivity of a carbon composite is dominated by the ﬁber conductivity and that to produce a high conductivity composite it is necessary to produce an ultrahigh conductivity ﬁber. This is only partly true. Whittaker and Taylor (1990) have measured the thermal conductivity of a series of 1- and 2-D composites produced by Societe Europeane de Propulsion (SEP) made from PAN ﬁbers and densiﬁed by CVD. After each of three CVD inﬁltrations, the materials were graphitized at 2800 °C. The nominal ﬁber volume fraction was 28% for 2-D samples and 22% for 1-D samples. The measurements of thermal diffusivity were carried out over the temperature range 300–3000 K (Whittaker et al., 1990a) and careful microstructural characterization was carried out (Whittaker et al., 1990b) to determine crystallite sizes of ﬁbers and matrix and porosity geometry.

Parallel to the ﬁbers in the 1-D composite the conductivity components are given by

Figure 16 Thermal conductivity of carbon ﬁbers derived from thermal diffusivity measurements. Reproduced by permission of Pion from High Temperatures–High Pressures, 1993, 25, 443–450.

Figure 17 Thermal conductivity at 400 K (l 400) as a function of calculated grain size for seven carbon ﬁbers. Reproduced by permission of Pion from High Temperatures–High Pressures, 1993, 25, 443–450.

The data normal to the ﬁbers were interpreted in a novel way by Whittaker and Taylor (1990) using the equation derived by Bruggemann (1935) for heat ﬂow in a two-phase composite where VD is the volume fraction of discontinuous phase, 𝞴D is thꞱe thermal conductivity of the discontinuous phase, 𝞴C is the thermal conductivity of the continuous phase, 𝞴 is the composite thermal conductivity, and X is a shape factor which describes the geometry of the discontinuous phase. This equation was used twice, ﬁrst to eliminate porosity and calculate the solid component conductivity, and second to separate the conductivity components 𝞴FꞱ and 𝞴MꞱ. The results are shown in Figure 18 that exhibit number of surprising features. First the thermal conductivity of the matrix parallel to the ﬁbers is signiﬁcantly higher than that of the ﬁbers themselves. Second and perhaps even more surprisingly, the matrix thermal conductivity perpendicular to the ﬁber at room temperature is at 250 W m⁻¹ K⁻¹ more than twice that of the PAN ﬁber. Clearly, orientation effects are important and it is possible to manipulate the transverse thermal conductivity, certainly of 1- and 2-D composites, by control of matrix precursor and processing conditions, particularly the ﬁnal heat treatment temperature, since more graphitic carbon is associated with higher thermal conductivity. The effect of different matrix precursors has received relatively little attention but is worth considering since matrices may be made by one of the three production routes (CVD, pitch, or resin) or any combination of these.

Lieberman and Peierson (1973) made several investigations into CVD process–property relationships for different carbon composites. Using a 100 kg m⁻³ felt preform, they prepared a range of composites from CH4/H2 using a thermal gradient inﬁltration technique. Each composite was given a ﬁnal heat treatment at 3000 °C. The correlation of matrix microstructure with thermal conductivity is shown in Table 3. The results clearly show that the rough laminar microstructure exhibits the highest

conductivity and is more graphitizable than the smooth laminar and isotropic microstructure.

Curlee and Lieberman (1973) reported thermal conductivity values for a ﬁlament-wound CVD densiﬁed tube where either CH4 or C6H6 was used as the hydrocarbon feedstock. Although both matrices possessed the same smooth laminar microstructure the composites prepared from benzene were found to have a thermal conductivity 80–120% higher than those derived from methane. This was attributed to the relative ability to graphitize of the matrices.

Figure 18 Thermal conductivity components for ﬁber and matrix deposited by CVD. Reproduced by permission of the Royal Society from Proc. R. Soc. London, 1990, 430, 167–181.

Table 3 Relation between CVC matrix microstructure and thermal conductivity

For thermoset materials it is generally considered that ﬁbers will dominate the thermal conductivity due to the low thermal conductivity of the non-graphitizable carbon. Pitch-derived CMCs are generally processed by the HIPIC technique using repeated cycles to build up an acceptably high density. Pitch matrices are considered to be highly graphitizable but although data has been presented for a number of 2 and 3-D composites, the author has been unable to ﬁnd any data for 1-D composites.

For 2- and 3-D composites the complex geometry of the porosity and the orthogonal ﬁber orientations means that the thermal conductivity in the principal directions is lower than that of 1-D composites parallel to the ﬁbers.

Whittaker et al. (1990a,b) noted values at 500 K of 190 W m⁻¹ K⁻¹ parallel to the cloth plane and 60 W m⁻¹ K⁻¹ perpendicular to the cloth for composites densiﬁed by CVD. These are to be compared with their values of 300 W m⁻¹ K⁻¹ at 500 K parallel to the ﬁber axis for their 1-D composites. A number of measurements have been made of the thermal conductivities of 3-D constructions. Values cited by McAllister et al. (1976) for the X–Y directions of two 3-D pierced fabric constructions are compared with the X–Y direction of a 3-D ﬁne weave orthogonal composite in Figure 19. A lower thermal conductivity is noted for the pierced fabric composite made using low-modulus ﬁbers but a higher thermal conductivity is noted for the pierced fabric composite made using high-modulus ﬁbers. Lee and Taylor (1975) quoted rather higher values at room temperature of 115–123 W m⁻¹ K⁻¹ for the X–Y and 66 W m⁻¹ K⁻¹ for the Z-directions of a similarly pierced fabric 3-D block.

Recognizing the lack of detailed information on the inﬂuence of different matrix types on the thermal conductivity of CMCs, Rand and coworkers have undertaken an investigation whereby 1-D composites are densiﬁed using pitch or resin with or without coke additions and subsequently heat-treated to a series of temperatures in the range 1000–2400 °C. 1-D composites containing ~42–48% ﬁbers and a range of different ﬁbers are being studied. Although this investigation is still not complete a number of interesting features are beginning to emerge as exempliﬁed by Figure 20, which shows the thermal conductivity of a series of 1-D composites made from pitch-derived ﬁbers which have been heat-treated to 2400 °C. The addition of coke to either pitch or resin precursor matrices slightly decreases the longitudinal thermal conductivity but signiﬁcantly increases the transverse thermal conductivity.

The thermal expansion of a graphite single crystal is unique in that, parallel to the layer planes; the expansion coefﬁcient is very low. It is negative at 20 °C, rising to ~1 × 10⁻⁶ K⁻¹ at T = °C. However, normal to the layer planes a very high thermal expansion coefﬁcient of ~26 × 10⁻⁶ K⁺¹ is recorded at high temperatures. Hence, carbon ﬁbers for which the layer planes are arranged parallel to the ﬁber axis have a nearly zero coefﬁcient of expansion parallel to the ﬁber axis. However, normal to the ﬁber axis a much higher expansion coefﬁcient of ~10 × 10—6 is noted (Fitzer, 1987). The expansion coefﬁcient of a 1-D composite will be restrained by the ﬁber expansion but also inﬂuenced by the layer plane orientations within the matrix. The differential thermal contraction between ﬁber and matrix on cooling from a high temperature is responsible for the devel- opment of a network of cracks parallel to the ﬁber and is generally held to be beneﬁcial in providing access for further impregnation to improve densiﬁcation. Therefore, a considerable amount of the thermal expansion on heating from ambient will be used initially in ﬁlling voids. However, in composites made up of ﬁber bundles the expansion will be strongly affected by the geometry of the ﬁber bundles. In Figure 21 the thermal expansion of a series of four reinforced carbon composites is shown which illustrates the nonlinearity of the thermal expansion (Fitzer, 1987). In general, values quoted over the interval 20–250 °C range from 0 to 1.0 × 10⁻⁶ K⁻¹ and from 2 to 4 × 10⁻⁶ K⁻¹ over the temperature range 20–2500 °C in 3-D composites (McAllister et al., 1983).

Figure 19 Thermal conductivity of some 3-D carbon–carbons in the X–Y direction (after McAllister et al., 1976).

Figure 20 Thermal conductivity of 1-D composites parallel and perpendicular to ﬁber axis for composite made up of 43–47% ﬁbers and matrices deposited by pitch, resin, and resin and coke (Rand, Taylor, Appleyard, Ly, and Zhang, Unpublished work).

Figure 21 Thermal expansion behavior of different rod-like prepregs (after Fitzer, 1987).

7. Mechanical Properties

The stiffness and strength of carbon–carbons are dominated by the ﬁbers and consequently the speciﬁc arrangement of ﬁbers in the composite architecture has a strong inﬂuence on these properties.

Carbon–carbon composites are very complex as a result of the physical and chemical changes and interactions that can occur during processing. Hence, the mechanical properties are dependent on a whole host of factors among which the most important are (Cunningham and Thomas, 1997):

Fiber type: rayon, PAN, pitch, high strength, high modulus,
Type of construction: weave, pierced fabric, rigidized tow,
Number of ﬁber directions: 1-D, 2-D, 3-D, etc.,
Fiber volume fractions in each orientation,
Matrix precursor: resin, pitch, CVD,
(Final density of ﬁnished composite.

Other contributory factors which are more difﬁcult to quantify are: changes in ﬁber properties during heat treatment, ﬁber damage during construction and densiﬁcation, ﬁber matrix bonding, and thermal expansion differences between ﬁber and matrix. Because so many different factors need to be taken into consideration, comparisons between different sets of published data must be treated very circumspectly. Nevertheless, some valid comparisons can be made.

Parallel to the ﬁber axis of a unidirectional composite the tensile strength may be expressed by the rule of mixtures

Similarly, the elastic modulus may be expressed by

The failure strain (ɛm) of the matrix in carbon–carbon composites is smaller than the failure strain (ɛf) of the ﬁbers, so that in a well-bonded composite, failure occurs when the matrix fails. If a strong bond exists between ﬁber and matrix, transverse cracks will penetrate the ﬁbers causing failure at a strain characteristic of the matrix and well below that of the ﬁber. Catastrophic brittle failure will occur and the full effect of the reinforcing potential of the ﬁbers will not be realized. The principal factor in avoiding brittle failure is poor adhesion between ﬁber and matrix. There is a considerable body of evidence which shows that in weakly bonded composites the longitudinal strength of unidirectional composites is higher than in well-bonded composites (Thomas and Walker, 1978a,b; Manocha et al., 1988; Fitzer and Huttner, 1981). This type of behavior is associated with surface treatment of the ﬁbers to promote bonding. Signiﬁcant strength reductions have been noted by Manocha et al. (1988) for surface treated and untreated Toray M40 ﬁbers, by Thomas and Walker (1978a,b) for phenolic resin precursors reinforced with three types of commercially available carbon ﬁber, and by Fitzer et al. (1980b) for composites reinforced with Sigriﬁl HF and Sigriﬁl HM ﬁbers. The work of Fitzer and co-workers is interesting in that, relative to the material reinforced with no oxidized ﬁbers, after initial carbonization increasing oxidation improved the strength of material reinforced with HM ﬁbers but decreased the strength of the material containing the high-strength ﬁber. The reasonable conclusion is that the HF ﬁbers reacted to produce a strong bond with the resin matrix, whereas the HM ﬁber did not. Another curious result discussed by Fitzer and Huttner (1981) was that, for composites which fail at the matrix failure strain, the strength can be greater when stiffer, usually weaker ﬁbers are used as the reinforcement. It is argued that when failure is matrix dominated differences in thermal expansion between ﬁber and matrix can cause the matrix to be pre-stressed in tension on cooling from carbonization temperatures. Since matrix and ﬁber must be well bonded to pre-stress the matrix, failure occurs at lower strains and reduced strengths are expected. However, as noted by Thomas and Walker (1978a,b), matrix-dominated properties of unidirectional composites such as modulus, transverse ﬂexural strength, and inter-laminar shear strength of well-bonded phenolic char matrix composites are superior to those of less well-bonded composites.

When the modulus of the matrix is much less than that of the ﬁbers, eqn [6] reduces to Ec = EfVf and this is a reasonable approximation for polymer matrix composites. However, Perry and Adams (1974), using a variety of ﬁber types in composites made using a variety of resin impregnant, measured modulus values that were much larger. Further evidence for the contribution of the matrix to the stiffness of CMCs has been provided by a number of authors (Evangelides, 1977; Fitzer and Huttner, 1981). Fitzer and Huttner, for example, reported modulus values for pitch/char matrix composites that were twice that computed by considering ﬁbers alone, implying an equal contribution from the matrix. A substantial body of evidence is beginning to emerge to show that an appreciable degree of preferred orientation develops on heat-treating pitch, resin, or CVD matrix precursors. This directionality signiﬁcantly inﬂuences the modulus values of the resultant composites.

In a similar fashion the out-of-plane and transverse properties of 1-D reinforced composites depend mostly on the properties of the matrix or matrix–ﬁber bond. As previously noted by Thomas and Walker (1978a,b) in their studies of CVD densiﬁed carbonized phenolic resin chars reinforced with commercially available ﬁbers; surface treatment of the ﬁbers improved both the transverse ﬂexural strength and inter-laminar shear. Unfortunately this improvement which always remained even after heat treatment was at the expense of a reduced axial strength. The results of Perry and Adams (1974), who studied three types of resin matrices and different ﬁber types, support the conclusion that larger values of transverse strength, modulus, and inter-laminar shear strength are obtained in composites reinforced with well-bonded ﬁbers. Although the foregoing discussion has highlighted the fact that a wide variety of properties are possible, it is perhaps instructive to summarize typical properties. Typical properties of 1-D composites are summarized in Table 4.

The results reported in the literature are difﬁcult to interpret because investigating one parameter while maintaining all other factors constant is difﬁcult if not impossible. While it is possible to optimize the matrix microstructure to improve matrix- dominated properties, the greatest improvement in properties can be obtained by placing ﬁbers in the appropriate orientations.

The simplest arrangement is the simple 2-D reinforcement which may be achieved by layering ﬁbers in 0°, 90°, ±45° orientations, or by the use of woven carbon fabrics. However, the maximum Vf in any given direction is 0.39 (c.f. 0.78 Vf for 1-D).

While this results in a composite with virtually isotropic properties in two orthogonal directions, the strength and modulus are lower than for the 1D composite. Most bidirectional carbon–carbons are in fact made from woven cloth, which, since fabrics contain undulating ﬁbers, means some loss of strength compared to simple cross-ply lay-up unidirectional types. Thus, a signiﬁcant factor in the mechanical properties of woven laminates is crimping, the displacement of ﬁll yarns from the laminate direction as they cross over the warp yarn. Manocha and Bahl (1988) studied the effect of weave pattern on the mechanical properties of 2-D carbon–carbon and demonstrated convincingly that the strength of a woven carbon–carbon with an eight- harness section weave was much greater than that of a composite made using the same ﬁbers but with a plain weave. Jortner (1989) showed that the tensile strength of woven carbon–carbon was inversely proportional to si𝞱 where 𝞱 is the mean crimp angle. A similar relationship was derived by Pollock (1990). However, although the properties in the X- and Y-directions may be adequate the properties through the thickness are still inadequate (Crocker and McEnaney, 1991).

Table 4 Typical properties of unidirectional carbon–carbon composites

Figure 22 Pseudoplastic behavior of carbon ﬁber reinforced carbon composites (after Fitzer, 1987).

Table 5 Comparison of X–Y properties of 3-D orthogonal and pierced fabric

Three-dimensional composites have been designed to improve the poor through-thickness properties of 1- and 2-D composites. Since the maximum possible Vf in any given direction is only 0.20, this means a reduced effectiveness in strength reinforcement. Figure 22 (Fitzer, 1987) demonstrates the relative achievable strengths for 1-, 2-, and 3-D reinforcement but also shows how pseudo plastic behavior may be realized especially for 3-D composites. Two types of 3-D composites are available, pierced weave (or pseudo 3-D) and orthogonal 3-D reinforced. McAllister et al. (1976) compared the mechanical properties in the X–Y direction of 3-D composites produced by pierced fabric and orthogonal weaving using the same high-modulus ﬁbers. The results shown in Table 5 show that tensile strengths and Young's moduli are similar although the ﬁber content in the pierced fabric composite is higher (see also Table 6). In the same study the authors reported an increase in strength from 80 to 98 MPa when the Z component was changed from dry tow to preformed rods containing the same number of ﬁbers. This effect is attributable to ﬁber damage in piercing with dry tow compared to the protection afforded by the resin matrix using pultruded rods.

Table 6 Effect of ﬁber bundle orientation on tensile strength of carbon–carbons

Figure 23 Stress–strain behavior of CVD carbon–carbon composites prepared from 10% propane at various temperatures. Reproduced by permission of Elsevier Science Ltd., from Carbon, 1988, 26, 763–768.

The mechanical properties of CMCs are very much dominated by the properties of the ﬁbers which depend on ﬁber type and processing. Apart from the obvious deduction that a key parameter is optimum ﬁber utilization with respect to the axis of the load, the ﬁber surface affects the ﬁber–matrix interaction which plays a key role in the densiﬁcation process. Manocha and Bahl (1988) found that high-modulus ﬁbers lead to better densiﬁed composites than high-strength or IM ﬁbers.

Perry and Adams (1976) reported tensile properties of a 1:1:2 ﬁne-weaves, 3-D carbon–carbon woven with high-modulus rayon ﬁbers and densiﬁed by multiple resin impregnation and carbonization cycles. The tensile strength and modulus in the Z-direction after 13 impregnation cycles (192 and 88 GPa) were only slightly higher than after seven cycles (177 and 84 GPa). However, after graphitization these values fell to 105 and 57 GPa, whereas the same properties in the X- and Y- direction increased slightly. This was attributed to shrinkage-induced gaps around the Z ﬁber tows. Shrinkage of the carbon matrix after graphitization is well known and the use of intermediate graphitization cycles improves the efﬁciency of subsequent impregnation by improving access to porosity.

As shown in Section 4, the structure of the carbon matrix is very much inﬂuenced by processing variables. This holds true for all three matrix precursors and while much of the information available remains proprietary some information detailing the effect on mechanical properties is available in the open literature. For CVD graphite, Oh and Lee (1988a,b) carried out a series of densiﬁcation experiments of a carbon fabric using a reactant gas concentration of 10% propane at temperatures between 1100 and 1400 °C. In material inﬁltrated at 1100 °C, the matrix was well inﬁltrated and well bonded to the ﬁbers, whereas at 1400 °C the matrix was loosely bonded to the ﬁbers and high porosity was observed. The bulk density decreased from 1790 to 1370 kg m⁻³ with increasing deposition temperature. The stress–strain behavior of the composites is shown in Figure 23. Increasing the concentration of reactant gas at constant temperature can result in increased composite bulk density and optical activity of the microstructures (Oh and Lee, 1989). Stress–strain curves change from catastrophic failure to a stepwise pseudo plastic failure as the hydrocarbon content is increased. A higher degree of preferred orientation of the matrix relative to the ﬁber surface was observed as the propane concentration was increased, resulting in a weaker bond between ﬁber and matrix.

Figure 24 Effect of heat treatment temperature on the bend strength of a composite made from a furan precursor (after Kimura et al., 1982).

The mechanical properties of thermoset-derived CMCs depend very much on the selection of resin and whether or not the ﬁber has been surface treated. The highest mechanical properties are achieved by using non-surface treated HM PAN or mesophase pitch-based ﬁbers. The requirements for a suitable thermoset resin matrix precursor are:

High carbon yield using simple pyrolysis conditions,
(Shrinkage on carbonization should not damage the carbon ﬁbers,
The carbon matrix formed should contain open rather than closed porosity.

The thermosetting resins usually employed have a carbon yield of 50–60% by weight. A single impregnation would result in a density of 1300–1400 kg m⁻³. Typically, 4–6 repeat impregnations would be required to raise the densities to 1700–1800 kg m⁻³. A major inﬂuence on the mechanical properties is the heat treatment temperature (Kimura et al., 1982) which can be summarized in Figure 24. At a heat treatment temperature below 2000 °C the strength is low and brittle fracture the predominant mode. At heat treatment temperatures above 2400 °C the presence of a graphitic component is observed at the ﬁber–matrix interface resulting in pseudo plastic fracture behavior. Processing to 2800 °C and above tends to reduce the strength due to thermal damage to the ﬁbers. This effect was also noted by Perry and Adams (1974) who, as noted earlier, reported a decrease in the tensile properties in the Z-direction of 1:1:2 ﬁber weave carbon–carbon composites after graphitization.

The effect of processing on the mechanical properties of pitch-precursor CMCs is difﬁcult to assess since no systematic results are available in the open literature. The mechanical and other properties are controlled not only by the carbon yield, 35% at ambient pressure to >80% at 100 bar (Fitzer and Terweisch, 1973), but also by the microstructure. Pitch carbonized under low pressures results in a well-graphitized carbon (Rand and Thomas, 1997) with a sheath structure parallel to the surface, whereas a transverse oriented matrix structure possessing more isotropic properties has been observed when high pressures have been applied during carbonization. As a general rule multiple impregnation cycles increase the ﬂexural strength of a pitch-based carbon composite (Fitzer et al., 1980c). A ﬂexural strength of 1000 MPa can be achieved after four impregnation/carbonization cycles although after four graphitization cycles this drops to 700 MPa. This is due to thermal contraction between isotropic matrix and ﬁber generating shrinkage stresses resulting in crack formation between ﬁber and matrix and weak bonding. Graphitization, by generating such cracks, aids densiﬁcation but results in a lower composite strength.

One of the main factors in determining the strength of carbon–carbon is the porosity. The strength of polycrystalline graphite and glassy carbon materials may be expressed in terms of porosity by a simple empirical Knudsen equation

where 𝞼o is the strength with zero porosity, 𝞫 is a constant, and 𝞺 is the porosity. A similar relationship has been shown in carbon–carbon materials (Kimura et al., 1978). The work of Kimura et al. (Figure 25) clearly demonstrates the inﬂuence of matrix type which shows quite dramatically that the geometry of the porosity is very important.

One feature that is particularly affected by heat treatment is of course the modulus of the composite. An increase in heat treatment temperature causes a rapid increase in modulus in the ﬁber direction of a unidirectional material (Yasuda et al., 1980a, b). In contrast the modulus perpendicular to the ﬁber axis falls slightly (Figure 26).

One of the greatest advantages of carbon-matrix composites is their ability, like that of carbon itself, to retain high speciﬁc strength and stiffness to high temperatures far in excess of the maximum useful temperatures for other ceramic materials.

Figure 25 The effect of porosity on the realizable percentage of ﬁber strength in a CMC (after Kimura et al., 1978).

Figure 26 Relationship between Young's modulus and heat treatment temperature for a 1-D carbon ﬁber composite (after Yasuda et al., 1980a).

Unfortunately, very little has been published on the high-temperature mechanical properties possibly for reasons of military sensitivity. The Young's modulus of a carbon–carbon previously heat treated to 1800 °C showed a progressive decrease with increasing test temperature to 1400 °C with a rapid decrease above this temperature. However, for a more graphitic carbon–carbon heat treated to 2600 °C, the modulus increased with test temperature to 1000 °C before decreasing (Hill et al., 1974). A similar observation of increasing modulus, this time to 1500 °C, has been noted by Fitzer and Heym (1977) and Kimura et al. (1982).

There is no clear trend of strength with test temperature. Fitzer and Terweisch (1972) found ﬂexural strength to be a weak function of temperature with a shallow minimum at 1000 °C. Kimura et al. (1982) noted a 10% increase in ﬂexural strength on heating from room temperature to 1500 °C. Thomas and Walker (1978a,b) found a similar weak temperature dependence but with a shallow maximum at 1000 °C. The decrease above this temperature was accompanied by a marked increase in failure strain particularly above 1400 °C. Sato et al. (1989) reported that the strengths of a woven bidirectional carbon–carbon and a nonwoven felt-based composite increased progressively with test temperature to 2400 °C matching a similar trend for polycrystalline graphite.

Figure 27 Fatigue curve for a CMC. Reproduced by permission of VDI Verlag from KunstoffTechnick, 1981, 87–107.

For one of the felt composites there was a dramatic jump in tensile strength of some 60% accompanied by a large increase in strain to failure from testing at 1600 to 2400 °C.

Relatively little data has been published in the open literature on the fatigue behavior of carbon–carbon. Fitzer and Heym (1981) showed a lifetime of 107 cycles at a stress of 40% of the static bending strength (Figure 27). Savage (1993) discusses whether the pore opening and closing mechanism which occurs during cyclic loading will continue indeﬁnitely and observes that the fracture strength of the matrix will be exceeded in local regions, thus causing fragmentation. He comments that this will cause material loss from the composite as dust and notes that ‘dusting out’ has been observed at high temperatures (B1400 °C) in centrifugally loaded fan blades.

Since carbon matrices have a large number of internal cracks or voids as a consequence of the fabrication process, for any speciﬁc application the fracture toughness values are of considerable interest. Unfortunately, very few data have been published. Fracture toughness is effectively the resistance to propagation of a crack through a body. Consequently the various fracture toughness parameters depend strongly on the type of carbon ﬁber used and the orientation of the initial crack with respect to the ﬁber architecture. In 2-D fabric reinforced composites, severe crack blunting and delamination are observed when crack propa- gation is perpendicular to the ﬁbers (Kim et al., 1985). Unfortunately, parallel to the plane of the cloth very low fracture toughness values are noted. Typical values of R, the resistance to crack growth, are ~50 000 J m⁻² normal to the laminae but only 60–95 for inter-laminar crack growth (Rmili et al., 1990; Sakai et al., 1991). Figure 28 (Sakai et al., 1991) shows the crack growth resistance curves for polycrystalline graphite and felt and ﬁber reinforced composites. The crack growth resistance for the ﬁber reinforced composite shows KR an order of magnitude higher increasing to 25 MPa m^0,5 from its initial Klc value of 7 MPa m^1/2. The occurrence of progressive cracking indicated by stepwise unloading of the stress–strain curve (Figure 22) results from a high degree of ﬁber pull-out and makes an extremely tough material (Fitzer, 1987).

It is generally accepted that the maximum use temperature of carbon materials is limited to ~2000 °C by the onset of creep (Green et al., 1970). CMCs were developed for defense/aerospace application to push this limit higher. Unfortunately, very little data has been published possibly because of the military sensitivity of the work. However, high-temperature creep of CMCs is characterized by an initial transient followed by a steady-state creep rate which increases progressively with increasing test temperature in the range 2060–2600 °C (Sines et al., 1989; Feldman, 1983).

The mechanical property and failure mechanism of carbon–carbon braided composites (C–Cs) bolted joints structure sub- jected to unidirectional tensile load were studied by the experimental method and numerical analysis (Yuling et al., 2015). The braided C–Cs bolted joints with the single-bolt single-lap (SBS) and double-bolt single-lap (DBS) were tested. The dominant failure modes for both C–Cs SBS and DBS joint conﬁgurations were bearing failure and net-tension. Parametric studies were implemented by ﬁnite element (FE) analysis to classify the effects of geometric parameters including the joint width (W), edge distance (e), and the bolt pitch (p) on the SBS and DBS joint conﬁgurations.

Various strengths of carbon–carbon composites (C/Cs) are comprehensively reviewed (Hiroshi et al., 2005). The topics reviewed include tensile, shear, compressive, and fatigue strength as well as ﬁber/matrix interfacial strength of C/Cs. When data are available, high temperature properties, including creep behavior, are presented. Since C/Cs have extremely low ﬁber/matrix interfacial strength td, the interfacial fracture plays important roles in all of the fracture processes dealt in this review. The low td was found to divide tensile fracture units into small bundles, to seriously degrade both shear and compressive strength, and to improve fatigue performance.

Figure 28 Crack growth resistance curves for carbon and carbon–carbon composites. Reproduced by permission of Elsevier Science Ltd., from Comp. Sci. Technol., 1991, 40, 231–250.

A series of carbon ﬁber laminate bearing versus bypass load tests were carried out investigating the effect of clearance, laminate lay-up, washer contact size and clamping force value (Rosales-Iriarte et al., 2011). Explanations of the underlying effects that inﬂuence the results have been described. The results compare well with literature but hole clearance was shown to increase joint strength when bypass loads are dominant, which seems counter intuitive, and has not been reported in literature.

8. Oxidation Protection

The most severe drawback inhibiting the use of CMCs is their susceptibility to oxidation above 500 °C. This becomes progressively more severe until at about 800 °C the rate of oxidation is limited only by diffusion of oxygen through the surrounding gas to the specimen surface. A detailed discussion of the kinetics of oxidation of carbon lies outside the scope of this chapter and the reader is referred to more specialist reviews (McKee et al., 1981). However, the predominant reaction is

Yasuda et al. (1980a,b) studied the oxidation behavior of a number of carbon matrix composites. The relationship between weight loss and time at a number of different temperatures is shown in Figure 29 for a composite containing 67 wt.% ﬁbers heat- treated to 2800 °C. They found the matrix to be more reactive with the ﬁbers oxidizing at a slower rate. These reaction rates are faster than for pyrolytic graphite and isotropic carbon. The oxidation of composites preferentially attacks sites of high energy such as ﬁber–matrix interfaces. The rate of oxidation is increased by an increase in operating temperature but reduced by an increase in heat treatment temperature of the composite. The latter observation is interpreted as being due to a reduction in the degree of retained impurities, relaxation of the carbonization stress, and reduction of reactive edge sites, despite an increase in the fraction of open pores (Chang and Rhee, 1978).

When evaluating prospective thermal protection systems, a number of important factors and associated system requirements need to be considered. These are summarized in Figure 30 (Strife and Sheehan, 1988) and have been reviewed on a number of occasions (Bines and Thomas, 1992; Westwood et al., 1996). The primary aim is to apply a coating which isolates the composite from the environment. In order to achieve this, the coating system must have at least one major component that acts as an efﬁcient barrier to oxygen. The primary oxygen barrier should have low oxygen permeability and its aim is to totally encapsulate the carbon, ideally with no defects through which the oxidizing species can ingress. Optimally a material can be used which forms an in situ oxide. It is, however, equally important to minimize the diffusion of carbon outwards from the substrate to avoid carbo thermic reduction of any oxides that may be present. It is also important to consider mechanical compatibility and the avoidance of coating spallation is a key issue.

Figure 29 Oxidation behavior of a CMC in air. Reproduced by permission of the Japan Society for Composite Materials from Trans. J. S. C. M., 1980, 6(1), 14–23.

Figure 30 Design considerations for an integrated oxidation protection system. Reprinted with permission of the American Ceramic Society, PO Box 6136, Westerville, OH 43086–6136. Copyright: 1988 by the American Ceramic Society. All rights reserved.

The coefﬁcient of thermal expansion of a carbon matrix substrate is very low compared to that of the bulk of refractory ceramics that may form part of coating systems. Any applied coatings are likely to contain micro cracks since the coating process(es) are carried out at elevated temperature. Moreover, the composite will:

Cycle from ambient to the working temperature, thereby generating more thermomechanical stresses,
Spend minutes or hours at that temperature.

Therefore, it is imperative that the coating possess a self-healing capability. The most successful solutions to date have been to incorporate a glass or a glass-forming compound which can ﬂow into and seal any cracks in the primary coating. It is necessary to seal the cracks which develop in the temperature range from 400 °C (the oxidation threshold of the coating) to the micro cracking temperature of the primary oxygen barrier. It is also necessary to establish a good adherence between the substrate and the coating and between the different layers of the system. The mechanical properties of the coating also need to be considered since the coating must be able to withstand any stresses generated at the surface of the component. Ideally, a low modulus is desirable to accommodate expansion mismatch strain during thermal cycling.

It is generally accepted that this combination of properties cannot be met by any single material and that to protect a carbon matrix composite over a wide range of temperatures a multilayer system is required. The main advantage of this methodology is to associate the speciﬁc advantages of each layer while limiting their drawbacks. Individual layers are strategically stacked with respect to the substrate to provide protection over the whole temperature range of interest. While a wide range of different systems have been developed and patented, the most successful systems are basically composed of three ingredients: bond layer, functionally active layer(s), and primary, erosion-resistant over layers or oxygen barriers (Figure 31). The most common primary oxygen barriers are SiC and Si3N4. They are both refractory and oxidation resistant due to the formation of a skin of SiO2 on oxidation. Silica exhibits a relatively low vapor pressure to temperatures as high as 1650 °C as well as low oxygen diffusivity (Schick, 1960;

Figure 31 Schematic of a multilayer thermal protection system.

Sucow, 1963). However, although at high temperatures the viscosity is low enough to allow SiO2 to ﬂow into and seal any cracks, below 1150 °C the viscosity of the glass is too high to afford any self-sealing properties. Many other hard oxide ceramics have been considered as potential outer layers. Sheehan (1989) has calculated that a value of 10—3 mm is an appropriate maximum vapor pressure for a material to be used as an erosion layer. Westwood et al. (1996) discuss the suitability of these materials and review the attempts to utilize some of them. They consider that none is ideal. One nonceramic that has been considered is iridium (Sheehan, 1989). Although excellent protection was achieved in the temperature range 2000–2100 °C for short periods, the major difﬁculty is the cost of iridium and the difﬁculty of fabricating good quality coatings.

To prevent oxygen reaching the substrate and to be capable of sealing cracks which inevitably form during service, the optimum solutions appear to be functionally active layers. Glassy systems have been considered for such use for many years and the ﬁrst patent was issued by the National Carbon Company as early as 1934 (Johnson, 1934). The use of borate glasses has been extensively studied and the viscosity of B2O3 in the temperature range 600–1100 °C coupled with its tendency to wet SiC and Si3N4 makes it an excellent sealant in this temperature range. However, the usefulness of borate glasses is limited by vaporization above 1000 °C (McKee, 1986) and moisture sensitivity (Chang and Wilcox, 1971). Volatilization can be reduced by increasing the viscosity of borate glasses. This may be achieved by adding up to 25% of a refractory oxide. For example, ZrO2 or HfO2 may be added for the temperature range 1200–1600 °C (Gray, 1986). To provide protection at temperatures of >1100 °C silica may be employed. However, its viscosity is too high at temperatures below 1100 °C to close cracks effectively. To overcome the deﬁciencies of simple systems, more complex systems have been developed such as TiO2–SiO2–B2O3 (Gray, 1988) and P2O5–SiO2–Al2O3 (Tawil et al., 1993).

A more recent approach to the use of glass sealants is the use of functionally active layers which include glass-forming compounds which form glasses when oxidized. The advantage of such layers is to actively absorb the oxygen (Figure 31). Several boron and silicon containing compounds in various combinations have been suggested for use in functionally active layers. These include B4C (Sheehan, 1989), TiB2 (Courtois et al., 1991), and MoSi2, although the latter is subject to a spallation known as MoSi2 pest (Lin et al., 1994), and a ductile to brittle transformation at ~1000 °C (Jeng and Lavernia, 1994). Bond layers are designed primarily to reduce the thermal expansion mismatch between the substrate and functionally active layers but also to prevent the outward diffusion of carbon from the substrate. The most commonly used bond coat materials are SiC and Si3N4 which may be applied either by CVD or using a slurry technique. Although in principle it would appear that the lower coefﬁcient of thermal expansion of Si3N4 would be an advantage, it has been found that no signiﬁcant improvement over SiC was found (Barrett et al., 1989).

While many simple systems have been described in the literature they are limited in their ability to provide protection and will not be considered here. The emphasis in the last decade has shifted to the development of multilayer systems, understanding the principles of thermal protection, and predicting behavior. An early notable success was the development of thermal protection for the nose cone and leading edges of the space shuttle (Rogers et al., 1976; Shuford, 1984a,b). Here the carbon–carbon component is encapsulated in a powder pack comprising SiC, silicon, and Al2O3 and then heated to 1750–1850 °C for a period of 4–7 h. This produces a coating thickness of 125–170 𝞵m. A combination of reactions generates high-temperature species (eqn [8]) which then interact at the carbon surface to form silicon carbide (eqn [9]).

To seal the cracks and pores they are ﬁlled with a mixture of glazes derived from tetraethyl Orto-silicate (TEOS), silicon carbide powder, alumina powder, and aluminum phosphate (Shuford, 1984a,b). This approach is typical of those where the protective properties of refractory coatings may be improved dramatically by covering with a glassy layer of a B2O3 based glass (McKee, 1987; Huttinger and Greil, 1992), SiO2, or more complex systems such as a borate glass containing ZrSiO4 particles (Cranmer, 1989).

Figure 32 Multilayer oxidation protection system developed by Bentson et al. (1989).

Within the remit of multilayer protection systems, a number of combinations have been developed. Bentson et al. (1989) have patented a complex four-layer system comprising (Figure 32):

An inner sealant comprising a boron-rich layer and a zirconia source,
An outer sealant layer comprising precursors to a complex borate glass and a granular refractory material. A preferred composition is 30% B₄C, 5 wt.% SiO2, 15 wt.% Li2ZrO3, 30 wt.% SiC, and 20 wt.% pitch,
An inner coating, 5–25 𝞵m thick, of B₄C,
An outer coating comprising 100–300 𝞵m of SiC applied by CVD or CVI.

This was successfully used on a carbon–carbon composite thermally cycled up to 1460 °C. Another protective system developed by Dietrich (1991) shown in Figure 33 comprises an initial boride sealant applied by slurry painting or CVD. Next is the primary oxygen barrier applied by CVD of stoichiometric or siliconized, SiC. The ﬁnal step is a borosilicate over glaze.

Barrett et al. (1989) evaluated the oxidation performance of coated carbon–carbon composites with or without a glassy overcoat. The optimum protection system comprised a nonstoichiometric B4C/SiC layer covered with a siliconized Sic layer and ﬁnally a borosilicate over glaze.

A multilayer system developed for C/SiC matrix composites but applicable to carbon–carbon has been developed by Goujard et al. (1994). In the former all three layers were deposited by CVD and comprised a thick (120–140 mm) inner SiC layer a thin (10–15 𝞵m) B4C layer and a 40–60 𝞵m outer layer of SiC. A system developed by Franc and Macret (1990) comprises an inner layer of SiC, an intermediate layer of AlN, and an outer layer of Al2O3. It is claimed that alternative materials may be used such as HfO2 or ZrO2 for the outer layer and TiB2, HfN, ZrC, Pt, or Ir for the intermediate layer.

An idealized four-layer coating has been proposed by Strife and Sheehan (1988) for protecting carbon–carbon at temperatures in excess of 1800 °C. The inner layer is refractory carbide to act as a diffusion barrier between the carbon–carbon and refractory oxides. Candidate materials may include TaC, TiCHfC, or ZrC because they all have low carbon diffusivities. Above this is an inner refractory oxide, a modiﬁed silica glass, and an outer layer of refractory oxide.

Candidate materials for the refractory oxide include ZrO2, HfO2, Y2O3, and ThO2. The trend toward designing multilayer surface coatings from ﬁrst principles by a judicious evaluation of candidate materials and an assessment of their likely performance for a speciﬁc application is likely to continue. An interesting approach has recently been pioneered by combining thermodynamic calculations and ﬁnite element modeling of candidate coating systems (Webster et al., 1997a,b; Westwood et al., 1996, 1997) for integrated oxidation protection. Potentially desirable combinations of system components are identiﬁed and the ﬁrst stage is to carry out thermodynamic calculations to predict (Webster et al., 1997a,b):

The chemical compatibility between any combination of functional layer components,
The chemical compatibility of interfaces between layers,
Simulation of the exposure of each layer to oxygen.

Figure 33 Multilayer oxidation protection system developed by Dietrich (1991).

Figure 34 (a) Calculated Si–B–C phase diagram isothermal section at 1600 °C. (b) Calculated B–C phase diagram (after Westwood et al., 1997).

Once a suitable combination of materials has been identiﬁed a ﬁnite element model is applied to calculate the mechanical stresses in the substrate and the protective coating system (Westwood et al., 1996). This will enable one to determine whether the stresses are sufﬁciently high to cause cracking of either the whole coating or particular layers and allow one to tailor the thicknesses of different layers to minimize the stresses. An example of the value of such an approach was illustrated by considering the system: an SiC bond layer, a 50% SiB4 + 50% B4C functional layer, and an SiC erosion protection layer. The compatibility between the two functional layer components was predicted by considering isothermal sections of the Si–B–C ternary phase diagram at 1200 and 1600 °C (Figure 34(a)). However, consideration of the B–C binary phase diagram shows that the B4C phase ﬁeld spans the range 10–20 mol., %C (Figure 34(b)). Where stoichiometric or carbon-rich B4C is present, no tie line exists between SiB6 and B4C in the ternary system. Hence, it is thermodynamically feasible at both 1200 and 1600 °C for excess carbon to react with SiB6 to form a nonstoichiometric boride SiBn þ B4C. The thermodynamically stable phases are SiB6 + SiBn + B4C and there is a liquid silicon boride phase at 1600 °C. Hence, to avoid the occurrence of such reactions, it is necessary to use boron-rich B4C. The sequence of reactions on exposure to oxygen is predicted to be

unoxidized boron in SiB6 is predicted to form a nonstoichiometric boride or be taken up by B4C

As the oxygen concentration increases further, the oxidation of B4C and SiB6 to form B2O3, SiO2, and CO2 gas is predicted

It is important to optimize the composition of the functional layer since the relative amounts of SiB6 þ B4C will determine the borosilicate glass composition and hence its viscosity and ability to ﬂow to seal cracks. An example of the beneﬁts of ﬁnite element modeling are also shown by Westwood et al. (1996) who considered three different pairs of phases containing B and SiC in the functional layer, namely, SiB6 + TiB2, B4C + MoSi2, and ZrB2 + TiSi2. In all combinations of these systems the stresses are such that all layers will crack in a system comprising a 40 𝞵m outer SiC layer, a 60 𝞵m thick functional layer, and a 20 𝞵m inner bond layer of SiC. Halving the coating thicknesses increases the stresses, whereas doubling the thickness reduces the stresses. A recent notable success of this approach was the development of a coating system based on SiC þ a mixed layer of Y2SiO5 and Y2SiO7 which protected a carbon ﬁber composite for 53 h at 1600 °C (Webster et al., 1997a,b).

An alternative method for applying additional protection to CMCs is to introduce oxygen inhibitors or getters into the carbon matrix during lay-up and densiﬁcation cycles. The most successful of these is boron, although silicon and titanium compounds such as SiC, Ti5Si3, and TiB2 may be used (Woodley, 1968; McKee et al., 1984; Gray, 1990). A number of so-called ‘inhibited prepregs’ are commercially available. However, while these reduce the reactivity of the composite with air, they only become effective after a signiﬁcant fraction of the composite has been gasiﬁed (McKee, 1988).

In order to protect carbon/carbon (C/C) composites from oxidation, a silicon carbide coating has been produced by a two-step pack cementation technique (Qian et al., 2005). XRD analysis showed that the inner coating obtained from the ﬁrst step was a 𝞫-SiC layer and the outer coating formed by the second step pack cementation was composed of 𝞪-SiC, Si, and 𝞫-SiC. The as-received coating provided excellent oxidation resistance, proving able to protect C/C composites from oxidation for more than 310 h at 1773 K in air. The weight loss of the coated C/C composites was considered to arise from the debonding of some glass and the formation of bubble holes on the coating surface.

To reduce the residual thermal stress between the carbon ﬁber-reinforced carbon (C/C) composites and the SiC coating layer, functionally graded materials (FGM) consisting of a C/SiC compositionally graded layer (C/SiC interlayer) were adopted (Kim et al., 2005). After designing the compositional distribution of the C/SiC interlayer which can relieve the thermal stress effectively, the deposition conditions of the entire compositional range of the C/SiC composites were determined using a thermodynamic calculation. According to the design and calculation the C/SiC interlayer and the SiC outer layer were deposited on the C/C composites by a low pressure chemical vapor deposition (LPCVD) method at deposition temperatures of 1100 and 1300 °C.

The stress calculation and the experimental results suggested that the SiC-rich compositional proﬁle in the FGM layer is the most effective for relieving the thermal stress and increasing the oxidation resistance.

An oxidation protective double layered coating was deposited on a carbon–carbon composite (C/C) using a simple and low cost method (Smeacetto et al., 2002). A surface modiﬁcation of the C/C was obtained by direct reaction of liquid silicon with the C/C, promoting the formation of a 5–10 mm 𝞫-SiC layer on the composite surface. The inner layer, in contact with the C/C, is a composite made with a barium borosilicate glass matrix (SABB) and boron carbide particles; the outer layer is another composite layer formed by a SABB glass matrix and yttrium oxide particles. The layers are deposited by a slurry technique. Oxidation tests were carried out in a furnace in air, in order to verify the coating stability and its effectiveness. No mass loss of the C/C composite was observed after 100 h at 1200 °C, while after 150 h at 1300 °C the C/C mass loss did not exceed 1%.

9. Applications of Carbon Matrix Composites

The most high proﬁle use of CMCs is as the nose cone and leading edges of the US space shuttle. Carbon–carbon is also being considered for other hypersonic plane applications such as the leading edges of the US National Aerospace plane (McConnell, 1990) and for gas turbine components. Some 63% by volume of the carbon–carbon produced in the world is used in aircraft braking systems. These were ﬁrst ﬁtted to Concorde in 1973. Advances in technology since the 1970s have now reduced the cost of carbon–carbon from ~£550 to ~£100 kg⁻¹. It is now therefore commercially advantageous to employ carbon–carbon brakes in civil subsonic aircraft and, for example, they are now speciﬁed for the Boeing 747–400, 757, 767, and 777 airliners, and all the Airbus family. On the Boeing 767 airliner, using carbon–carbon produces a weight saving of 400 kg over conventional steel brake systems and the increased durability permits 3000 jet aircraft landings compared to 1500 for metal rotors. Brake disks are required to provide the frictional torque to stop the aircraft and to absorb the several hundred mega joules of heat generated during braking.

Friction between disks causes them to heat up to ~500 °C with surface temperatures reaching 2000 °C. Hence, the materials used must exhibit good thermal shock resistance. The high thermal conductivity and low coefﬁcient of thermal expansion make carbon–carbon a very appropriate choice. Carbon–carbon composites were ﬁrst introduced into formula 1 racing in the early 1980s by the Braham team. Since that time they have become universally used in the sport for brakes and clutches. Two manufacturers provide brakes to the formula 1 circuit, S.G.L. Ltd., and S.E.P. The necessity to dissipate large quantities of energy in braking trains has also led to their use in high-speed passenger trains such as TGV. However, it is unlikely that carbon–carbon would be used in passenger cars in the foreseeable future because of costs.

The aerospace ﬁeld continues to be one of the primary areas for the use of carbon–carbon. In addition to the space shuttle highlighted earlier, carbon–carbon has been used in solid propellant rocket nozzles, in exit cones (Grenie et al., 1987), and as ablative nose tips and heat shields for reentry vehicles. Since on average a rocket motor burns for about 30 s the demands on structural materials for nozzles and exit cones are short-lived but intense. Oxidation is generally not a problem and a controlled ablation is often built into the structure. Dense carbon–carbon is preferred because of its superior ablation resistance. Initially, 2-D weave exit cones were used but 3-D weaving technology developed in France by Brochier and Aerospatiale has gradually sup- plemented these. Again although the space shuttle has perhaps the best known carbon–carbon re-entry heat shield, the greatest number of parts which account for 11% by volume, 37% by value of all carbon–carbon produced are used in the nose cones of ballistic missiles. The thermal and mechanical loading during reentry is such that 3-D carbon–carbon is the most convenient material. The major advantages of carbon–carbon are the high thermal conductivity which eliminates thermomechanical overload thereby avoiding surface cracking and the high heat capacity which means that the component effectively operates as a heat sink. Carbon–carbon components are being used for the ﬁrst wall tiles of fusion reactors such as Tokomak Test reactor (TFTR), the Japan Atomic Energy Research Institute's JT60, and the joint European Torus JET. These use multidirectional ﬁrst wall bumper limiter tiles, RF limiter tiles, and ﬁrst wall diverters. Some of these are required to function at continuous temperatures of 2200 °C with occasional spikes to 3300 °C. The next generation of plasma fusion reactors such as The Burning Plasma Experiment (BPX) and the International Thermonuclear Experimental reactor (ITER) will require advanced carbon–carbon composites possessing very high thermal conductivities to cope with the anticipated severe heat loads.

Elemental carbon is known to have the best biocompatibility of all known materials (Bokros, 1977) and is compatible with bones, blood, and soft tissue. This excellent biocompatibility plus the ability to tailor a modulus to be similar to that of bone make carbon–carbon an attractive material for areas of implant surgery. However, due to the long elapsed time to obtain licenses to use such materials the medical applications are limited and face stiff competition from thermoplastic resin matrix composites which are cheaper and easier to work with. There are, however, reports in the literature of its use as a ﬁxation for carbon ﬁber artiﬁcial ligaments and bone plates in osteon synthesis and endo-prosthesis (Claes et al., 1980). Other applications include furnace heating elements and changing stages, hot press dies, and ‘gob’ interceptors used in glassmaking. Possible future applications that are being actively researched are hypersonic vehicle airframe structures, space structures, and engine components for gas turbines.

Textile-reinforced thermoplastic composites offer huge application potentials for a rapid manufacturing of components with versatile possibilities of integrating functions (Hufenbach et al., 2011). However, an application of these new materials requires the knowledge of the directional dependent material properties. For the new group of multi-layered ﬂatbed weft-knitted glass ﬁber/ polypropylene composites (MKF-GF/PP), tensile tests under different temperatures and test velocities have been carried out as well as Charpy impact tests, open hole tension tests and dynamic-mechanical analysis. The mechanical properties of MKF-GF/PP and unidirectional GF/PP composites with tailored ﬁber surface and interphase, respectively, have been compared to those of woven GF/PP composites and GF/PP composites made of non-crimp fabrics (NCF) as a benchmark.

In development of an air-turbo-ramjet engine with an expander cycle (ATREX) for a space plane, application of carbon–carbon (C/C) composites plays an important role to achieve high performance (Yasuo et al., 2002). Dovetail joints are one of the key structures to realize the turbine system made of C/C composites. Tensile tests were carried out on simpliﬁed dovetail joint models. Comparison between the experimental and the calculated results suggests that fracturing of the dovetail joint was controlled by the average shear stress, which implies that the shear stress concentration on the shoulder was relaxed during the fracture process. It was also shown that the dovetail joint made of C/C composites is feasible for use in the ATREX engine.

Source: Rama Sreekanth - National Institute of Science and Technology, Berhampur

Advertisement

The 10 largest coal producers and exporters in Indonesia:

Click Here for Top World Clean Coal Technologies Contractors

Sunday, March 1, 2020

Benefits and Applications of Carbon Matrix Composites