Some Interfacial Problems in Metal Matrix- Carbon Fibre Composites BY S. V. BARNETT, S. J. HARRIS AND J. V. WEAVER Department of Metallurgy and Materials Science, University of Nottingham, Nottingham, NG7 2RD Received 14th July, 1972 Metals reinforced with continuous graphite (Type I) fibres are prone to compatibility and oxida- tion problems when they are subjected to high temperature treatments. To prevent such interactions, the feasibility of using evaporated barrier layers has been examined in a nickel matrix. A number of techniques, e.g., tensile testing, thermal-balance determinations and microscopy have been employed to test the usefulness of this approach. Results have shown that 1 pm thick metal coatings either react vigorously with the fibre or allow diffusion of oxygen or carbon to take place sufficiently quickly so as to delay fibre degradation by only a short time interval.Whilst chemically stable non- metallic coatings tend to prevent diffusion of both carbon and oxygen, they are extremely susceptible to brittle cracking due to differential thermal expansion effects. In this respect, zirconium carbide and boron nitride seem to be affected to a lesser degree than other carbides, nitrides and oxides, and can provide protection over a limited temperature range. The demand for higher specific strength materials which have the capacity to operate in hostile environments, e.g., oxidizing atmospheres above 523 K has en- couraged work on metal matrix composites. Observations 1-4 made on carbon fibre reinforced metals have indicated that a number of fundamental problems exist, each of which requires a solution before such materials can develop their full potential.To a large extent these problems are associated with metal-metal, metal-fibre or TABLE ST STANDARD FREE ENERGIES OF FORMATION OF CARBIDES AND OXIDES 5 9 compound (carbides) ALC3 B4C cr23c6 C0,C Fe3C Ni3C Sic TdC Tic wc ZrC standard free energies of formation at 1300 K/kJ mol-1 - 147 - 43 - 461 + 5 -2 + 25 - 44 - 154 - 167 - 49 - 170 compound (ox ides) A1203 B203 co coo CUO FeO Fe203 Ir02 NiO Si02 Ta205 TiOz wo3 ZrOz Cr203 standard free energies of formation at 1300 K/kJ mol- standard free energies of formation of other compounds at other 1 temperatures/kJ mol-1 - 1686 - 957 BN - 137 at 1300 K - 173 at 900 K - 227 CO -192at 900K - 798 - 143 - 41 - 179 - 487 - 0.04 - 125 Ir07 -70 at 900 K Ni,C +27 at 900 K NiO - 159 at 900 K - 646 - 1480 - 710 -518 - 850 144S .V . BARNETT, S. J . HARRIS A N D f . V . WEAVER 145 composite-environment interfaces, inasmuch as they affect the chemical, physical and mechanical properties of the composite. The chemical stability of carbon fibres in a number of metal matrices may be simply assessed on the basis of the free energy data quoted in table 1. Metals divide themselves into two broad groupings, those that form stable carbides, e.g., Cr, W, V, Zr, etc. and those that either do not form carbides, e.g., Cu, Ir, Ag, etc. or form relatively unstable carbides, e.g., Ni, Co, etc. Principal interest in fabricating carbon fibre composites has centred around aluminium and nickel.Aluminium is an interesting case because on one hand the free energy of formation of its carbide is large and negative but on the other it does not rapidly form this compound at the fibre-metal interface until the temperature is raised close to the melting point of the metal.7 This slow reaction considerably helps the development of a bond between the fibre and matrix, during fabrication, whilst nickel in the absence of any direct chemical reaction with carbon does not establish anything other than a poor bond. TABLE 2.-sOLUBILITY OF CARBON AND OXYGEN IN METALS solubility of solubility of solubility of carbon at 1273 K carbon at 873 K oxygen at 1273 K element atomic % atomic % atomic % Ag A1 Au c o Cr Fe Ir Mn Mo Ni Ti V c u <0.001 (S) < 0.050 (S) <0.30 (S) -0.50 (S) < 0.026 (S) 0.04 (S) 2.00 (S) <0.05 (S) 0.40 (S) - 0.45 (S) <0.001 (L) - < 0.050 (L) - <0.30 (L) - <1.00 (S) - <0.05 (S)* - 10.00 (S) - ~ 0 .0 5 (S) - 1.20 (S) 0.05 1 1.80 (S) 1.00 (S) - 1.60 (S) 0.032 6.70 ( S ) - - - 0.020 * estimated on the basis of evidence given in ref. (16) The direct and excessive formation of carbides at the fibre surface produces a severe reduction in strength. This is not the only compatibility problem associated with metal matrix composites, for metals which do not form stable carbides can degrade carbon fibres under certain conditions, particularly where limited solid solutions exist, see table 2. Contrary evidence exists of the mechanisms by which the solubility aspect manifests itself.Jackson and Majoram have shown, chiefly by X-ray techniques, that both carbonized and graphitized fibres recrystallize to form a three-dimensional graphite when such fibres are encased in a nickel matrix and subjected to heat treatment at temperatures > 1273 K in a vacuum of - 10-2N m-2. Barclay and Bonfield on the other hand have provided evidence of fibres remaining intact and not recrystallized at 1373 K provided the heat treatment was carried out at pressures below 10-4N m-*. Although the explanations of these observations are in dispute, it is apparent that the stability of carbon fibres in nickel is dependent on the surface condition of the fibres prior to coating, the gas content of the coating and the vacuum pressure of any subsequent heat-treatment process.Oxidation proteetion at elevated temperatures is a major problem with continuous fibre systems, for even if the metal provides some protection for the fibres, it may be146 METAL MATRIX COMPOSITES difficult to stop local attack where fibres come to the surface, e.g., fibre ends. Stud- ies lo completed on aluminium-matrix composites have indicated that an acceptable oxidation rate was obtained up to 673 K for graphitized fibre reinforcements, provided the fibres were well bonded and fibre ends not exposed. The case with nickel is quite different, and indeed for most metals which have small negative free energies of formation for their oxides (table l), since oxygen effectively ‘‘ diffuses ” to the metal- fibre interface, resulting in oxidation of the carbon.on nickel- graphite fibre composites have shown that within a period of one hour at 873 K the fibres have been attacked severely. Within the present programme of work, attempts have been made to examine methods of controlling some of the reactions that cause degradation of carbon fibres in a nickel matrix. In principle, an assessment of the value of using a barrier layer interleaved between fibre and matrix has been attempted. Such an assessment has involved tensile testing, microscopy and thermobalance determinations on graphitic (Type I) fibres which have been coated with selected metals and compounds and then overcoated with nickel. The selection of barrier layer materials was made on the basis of diffusion data and oxidation resistance.Results obtained TABLE 3.-cARBON DIFFUSION DATA IN METALS AND CARBIDES D at 1273 K/mmz s-1 ref. material DO/IXIIII~ s-1 U/kJ mol-1 HfC Tic0.60 Tic Ta2C NbC M o ~ C W Ta Nb V Zr Cr Fe Ti Ni 5.ox lo6 3.3 x 104 1 .OX 103 1 . 9 ~ 104 1 .OX 105 8 . o ~ 103 1 x 10-3 7 . 0 ~ lo2 1.2x lo2 1.6 1.8 5x 10-1 5x 10-1 3 x 1.7 3 . 2 ~ 10-1 536 477 435 460 347 372 33 1 381 167 163 134 113 109 100 84 79 9x 10-l6 8x 10-l6 4x 10-l6 4x 3 x 10-l2 1 x 10-l1 4 x 10-l2 1 x 4~ 10-15 3 x 10-7 4~ 10-7 2~ 10-5 2~ 10-5 2~ 10-4 3~ 10-5 2x 6x 12 12 12 12 12 12 12 12 13 14 14 14 14 14 14 13 17 Since carbon goes into solid solution in most metals interstitially, its diffusion coefficient tends to be much larger than those found for elements in substitutional solid solutions. An analysis l1 of the diffusion of fibre material into a finite coating has shown that a 1 pm thick barrier layer with a diffusion coefficient of carbon of mm2/s would saturate in approximately one day, whilst a similar thickness of material of diffusion coefficient of mm2/s would saturate in -300 years.Available experimentally determined diffusion coefficients in metals and compounds are given in table 3. It can be seen that only the refractory metal carbides, e.g., HfC, ZrC and TIC have values in the desired range, i.e., to 10-l8 mm2/s. To provide a comparison, iridium was selected as a ductile metallic barrier layer. Al- though the diffusion coefficient for carbon in iridium at 1273 K is not known, Criscione, Volk and Smith l5 obtained a certain amount of data from which it may be obtained if the solid solubility of carbon in iridium is estimated.If the solubility is -0.05 atomic %, then the diffusion coefficient at 1273 K is - mm2/s. ThisS . V . BARNETT, S . J. HARRIS AND J. V . WEAVER 147 value is low when compared with other values of carbon diffusion coefficients in metals. Another criterion for barrier materials was resistance to oxidation. On these grounds, iridium again appears to be an attractive metal, since it has been shown l6 that oxygen does not effectively diffuse through it, and that below 1395 K a protective film of iridium dioxide forms. Alumina was chosen as another material for pre- venting fibre oxidation and carbon diffusion because of its high stability and melting point. Provided alumina is of high purity there is little reaction with nickel at temperatures up to 1373 K.Boron nitride has been used as a means of protecting boron fibres from oxidation in a nickel matrix, although the free energy data (table 1) indicate that the nitride is not highly stable. However, the nitride is more stable than the carbide and therefore should not interact with the fibre. From the oxidation standpoint, the same argument can be applied to the carbides of Ti and Zr, the oxides of these metals being highly stable. EXPERIMENTAL PREPARATION AND HEAT TREATMENT OF COATED FIBRES Small amounts of graphitized fibre (prepared by pyrolysis up to -2973 K) were coated by physical evaporation; the evaporant was heated by a focussed and accelerated beam of electrons. This process was used to apply coats of nickel, aluminium oxide, iridium, zirconium and titanium carbides and boron nitride.Small groups of fibres were spread and fixed to metal frames (150 mmx 60 mm) by using a conductive thermosetting silver preparation. These frames were mounted on a jig which allowed them to rotate within the evacuated bell-jar (vacuum pressure < This rotation was necessary to obtain complete coating on the fibres and thus counteract the line-of-sight nature of the deposition process. Attempts were made to minimize interaction between crucible and evaporant, co-evaporation of crucible material, and contamination of the evaporant beam by the bell-jar atmosphere which could all have affected the properties and composition of the coating. To minimize surface contamination of the carbon fibres prior to coating of high tension voltage was applied across them in an attempt to bake off adsorbed species, etc.Table 4 gives a list of the evaporants together with their chemical composition, physical states and operating conditions for evaporation. N m-2). TABLE 4.-EVAPORATION CONDITIONS FOR COATINGS physical coating electron electron state of thickness gun beam evaporant pm potential /kV current /mA comments material nickel iridium alumina boron nitride zirconium carbide titanium carbide sheet 0.5-1.0 3.0 60 good smooth coatings wire 0.5-1.0 3.2 80 good even coatings granules 0.5-1.0 2.5 50 easy operation to powder - 1.0 3 .O 80 charge did not melt but produce even coatings turned from white to grey carbides was difficult, coating rates were required powder - 1.0 3.5 110 evaporation of both powder - 1.0 3.4 100 particularly where fast Boron nitride coatings were also applied by the passage of the fibres through an aqueous solution,* containing boron salts.Coating was achieved by immersing bundles of fifteen fibres into the solution. After immersion, the fibres were transferred to an alumina holder148 METAL MATRIX COMPOSITES ready for treatment in the furnace. The specimens were introduced slowly into the hot zone (1273 K) of the nitriding furnace to allow slow drying of the fibre and coating. Electrodeposited overcoats of nickel were obtained from a sulphamate solution. To facilitate plating on to non-conducting fibre coatings a thin layer of physically evaporated nickel was put down prior to immersion in the electrolyte.Specimens were subjected to heat treatment in bundles, each of which contained indi- vidually coated fibres. The bundles were inserted into alumina tubes which were in turn supported in a graphitic carbon block; the whole assembly was placed in a tube furnace, which was controlled to k 2 K . The presence of the carbon block acted as a getter for residual oxygen in the furnace tube. Sufficient time was allowed for outgassing at temper- atures below 623 K, whilst the vacuum was maintained at a pressure below 5 x 10-3N m-2. Heat treatment in oxidizing conditions was carried out in a similar furnace arrangement, with the exception that the furnace tube remained open to the atmosphere to enable a free supply of air to flow over the specimen.COATED FIBRE STRENGTH TESTS Tensile tests were carried out on micro-composites; a technique used by Jackson and Majomn2 Each composite was prepared from approximately fifteen fibres after coating and heat treatment by applying a few drops of Araldite MY753/HY951 resin and curing. The failure load of each composite was measured on a straining frame with a load cell of 9 N capacity. Throughout this work, strengths are quoted as bundle breaking loads rather than stress values. This is done for three reasons. First, the strength changes are large and are immediately apparent from the bundle breaking load. Second, there is the problem of counting and measuring the diameter of the fibres. Although attempts were made to keep the bundles at fifteen fibres, the handling problems resulted in an inaccuracy of & 1 fibre, which must in part account for the spread in results obtained.The mean fibre diameter was determined as 8.8 pm for the batch of fibres, with a standard deviation of rf: 0.40 pm. Finally, it must be remembered that a composite was being tested and some allowance would have to be made for the strength of the matrix. This would be of particular importance in testing of coated fibres. However, the neglect of this factor is reasonable since the coatings were thin, i.e., 0.1-0.5 pm, and no evidence was found of abnormally high strengths with as-coated fibres. OXIDATION KINETICS Experiments were performed using an automatic recording thermobalance which allowed continuous monitoring of the weight of the sample.Initial tests using as-received graphite fibres established a rate dependence on the initial sample weight, for this reason a minimum sample weight of 50mg was imposed. The thermobalance measured to an accuracy of k0.l mg. MICROSCOPY Coated fibres, before and after heat treatment, were examined in the scanning and trans- mission electron microscopes (S.E.M. and T.E.M.). Small pieces of fibre were trapped between two grids before insertion in the T.E.M., fibre ends, edges and coatings could then be examined in shadow. To examine the fibre surfaces of nickel overcoated specimens after treatment, cross and taper sections were prepared for optical microscopy. It became necessary to remove the thick nickel electrodeposit for S.E.M. observation and this was achieved by etching the bundles in a dilute nitric-sulphuric acid mixture.The fibres were recovered by filtration. RESULTS TENSILE TESTS AFTER HEAT TREATMENT IN VACUO The bundle breaking loads obtained for plain carbon fibres after heat treatment for one day at temperatures up to 1423 K in uacuo fall in the range 100 gf to 190 gf * Boric acid and urea (19 g of each) were melted at 415 K in air ; the resultant liquor was made up to 250 ml with distilled water.S . V . BARNETT, S . J . HARRIS A N D J . V . WEAVER 149 (equivalent to an individual fibre breaking stress range of 1070-2040 MN/m2). With iridium coatings there was no obvious reduction in breaking load even when the fibres had been held for 36 h at 1433 K, see fig. 1. However, the imposition of an overcoat of electrodeposited nickel over the iridium significantly changed the strength of the mini-composite at temperatures in excess of 1273 K, see fig.2. After a one day treatment at 1373 K a complete loss in strength had occurred. For comparison purposes, a series of specimens in which the fibres had been directly coated with nickel (by physical vapour deposition) were tested and are plotted on fig. 2. 1401 8 100 z 8. 8. 8. 691 m 3 6 hours El e temperature/"C FIG. 1 .-Breaking load for mini-composites containing iridium coated fibres after vacuum heat treatment. The horizontal lines refer to the loads obtained on uncoated Type I graphite fibres tested under similar conditions. The tensile behaviour of alumina coated samples are shown in fig. 3, together with samples which had been coated with nickel.The single alumina coating does not appear to reduce the bundle breaking load when such values are compared with those associated with uncoated fibres. However, the nickel electrodeposited samples became more and more difficult to handle as the heat treatment temperature increased. A definite strength reduction had occurred in samples treated at 1323 K and a number of " no-load " failures occurred with fibres heated to 1373 K.150 METAL MATRIX COMPOSITES 200 L, 8 180 - 160 - 140 - 8 s g 8 0 - a 60- 40 - 20 - I I *. 800 900 I000 1 roo temperature/"C FIG. 2.-Breaking load for mini-composites containing fibres which have been (a) nickel coated and (b) iridium coated, followed by an overcoat of nickel, and vacuum heat treated for 24 h.TENSILE TESTS AFTER OXIDATION TREATMENT Tests carried out on plain carbon fibres after four-hour periods of treatment showed that there was a marked degradation at 773 K and a complete loss of strength at 873 K. Iridium coated samples gave almost exactly similar results, except that some fibres were at least capable of being tested after the 873 K treatment, see fig. 4. Nickel coated fibres appeared to be degraded at lower temperatures than either the uncoated or iridium coated samples, see fig. 5. However, statistically there was no significant difference between results, i.e., a " t " significance test showed that there was a greater than 70 % probability that the differences in the 773 K test results occurred by chance. The plating of an alumina coating on the fibres did not prevent the breaking load of the fifteen fibre bundles from falling after they had been treated at temperatures in the range 673 to 873 K, see fig.5. Again there was evidence that some strength could be retained after 4 h at 873 K ; the range of the breaking loads were somewhat higher than those obtained on the iridium coated samples.S . V . BARNETT, S . J . HARRIS AND J . V . WEAVER 151 The tensile test results obtained on boron nitride coated fibres are shown in fig. 6. The fibres in this series of experiments had been coated by immersion in the boron salts solution (held at 323 K) and nitrided at 1273 K. To remove any contribution ------- .--- 4 : I, .-\- - - -7- - - -- - - - ------ - - '. .. Uncoated. D-43 Alumina-coated - Nickel - c o o f e d 1 i.1 . . \ ! 24-h vacuum heat-treatment temperature/"C FIG. 3 .-Comparison of breaking loads obtained on mini-composites containing (a) uncoated, (b) nickel coated, and (c) alumina coated fibres after vacuum heat treatment for 24 h. from the -0.1 pm thick boron nitride, the heat treated fibres were ultrasonically cleaned. The appearance of a limited number of testable specimens after treatment for 4 h at 873 K was encouraging; it was noted that three control samples of un- coated fibres did have sufficient strength to be tested in this particular set of experi- ments. OXIDATION BEHAVIOUR USING THE THERMOBALANCE The results of the thermobalance oxidation (at 868 K) of fibres coated with boron nitride from a continuously stirred and heated solution are summarized in fig.7. It would appear from these results that there is no advantage in increasing the thickness of the boron nitride coating, i.e., from 1 vol % (0.02 pm thick) to 19 vol % (0.5 pm thick). The points on the graph represent the calculated carbon loss assuming that152 METAL MATRIX COMPOSITES I 40t -I I I 1 I m 200 300 400 500 600 temperature/"(= FIG. 4.-Breaking load for mini-composites containing iridium coated fibres after heat-treatment in air for 4 h. the boron nitride weight remains constant. Standard uncoated fibres suffered complete loss in weight in -20 h whilst fibres with 1 vol % boron nitride coatings have lost only some 10 % of their weight at this stage. OPTICAL AND ELECTRON MICROSCOPY PLAIN CARBON FIBRES Scanning electron microscope examination showed the fibres to be essentially circular in cross-section with regular ridges on the surface.Shadow micrographs of fibre edges and fracture ends in the transmission electron microscope (T.E.M.) showed the edges to be essentially smooth and the fibrilar substructure to be in evi- dence at thin projections on the fracture surface. Electron diffraction patterns obtained from fibre edges and ends showed a limited angle of arcing on the 002 and 004 rings, and streaking in the 10 and 11 bands in the (001) direction, see fig. 8. After treatment in air for 4 h at 673 K, slight pitting of the fibre surface was observed in the S.E.M. As the temperature increased up to 873 K the incidence and depth of pitting increased markedly. COATED FIBRES SUBJECTED TO VACUUM HEAT TREATMENT Prior to heat treatment, all types of coating were examined in the S.E.M.to When fibres were first coated with ensure that they were of a continuous nature.S . V . BARNETT, S . J . HARRIS AND J . V . WEAVER 0 153 f I , I I I i a 0 ? I I I I I I I 1 I , I I I I I I i 7 I 1 d l ? t j I I I I I I I I I I I I I I I 1 I I I I I t I I I I I I I A )--. Uncooted. c)----o Alumina-coated &---A Nickel-coated I I 1 1 300 400 500 600 4-h air heat-treatment temperature/"(= FIG. 5.-Comparison of breaking loads obtained on mini-composites containing (a) uncoated, (b) nickel coated, and (c) alumina coated fibres after heat treatment in air for 4 h. alumina and subsequently with nickel, the two layers were discernable in the S.E.M., see fig.9a. The nickel and alumina thickness in this particular case were 0.2 and 0.5 pm respectively. On occasions when nickel was electrodeposited around an entire fifteen fibre bundle, cross sections of this type of micro-composite were examined by optical microscopy, see fig. 96. During treatment in vacuum for one day from 1123 K and upwards, agglomeration of thin vapour-deposited nickel coatings occurred, and at 1273 K there was complete spheroidization, see fig. 10a. In these specimens surface pits were observed on carbon fibre surfaces. Transmission electron micrographs of the surface regions between the nickel spheroids showed that a number of growths existed, see fig. lob. At higher magnification, fig. lOc, a fine grained layer was observed to be attached in places to the fibre surface together with a number of separate particles or flakes which were not in general associated with the fibre surface.Diffraction patterns were obtained partic- ularly from the fine grained material, see fig. lOd, and calculated " d " spacing indi- cated that it was graphite. The diffraction rings obtained were unlike those associated154 200 180 160 ru M 3 140- 9 8 \ 4 2 120- z w ru 0 100- 4 - .g 80- 24 9 2 60 40 20 METAL MATRIX COMPOSITES m rn As-coated . o BN-coated (0.lymthick ) ultrasonically cleaned of coating after oxidation 0 Uncoated standards - - - 00 0 0 0 - 0 - 0 0 - 0 I I I I NONE 400 500 600 0 B. BB 00 0 m 0 0 0 0 0 0 0 0 O n 00 0 0 0 0 0 0 0 heat treatment temperature/"C FIG. 6.-Breaking loads obtained on mini-composites containing boron nitride coated fibres after heat treatment in air for 4 h.test time in hours 10 x ) 30 40 50 60 80 100 120 I40 0 Uncoated fibres standard. I. vol.percent cmting . 0 4 . - - - - 19. " " . CO.lp4 FIG. 7.-Oxidation tests on boron nitride coated fibres in the thermo-balance at 868 K represented as percentage carbon loss from the fibre.FIG. 8.-Electron diffraction pattern from very thin area of fibre surface showing 002 and 004 arcs. The fibre axis parallel to the 002 arcs. [To face page 154FKG. 9.--(a) As deposited duplex nickel- FIG. 9.-(h) Specimens with electrodeposited aIuminia coating on a gr-aphite fibre, 0.2 p.m nickel overcoatings, sho\iing the distribution and-0.5 pin thick respectively. Magnification of a bundle of fifteen fibres.Magnification ;< 6000. x 600. FIG. 10.-(a) Nickel coated fibre heat treated in vacuum for one day at 1273 K. Magni- FIG. lO.-(b) Graphite film associated with nickel globules (as in fig. I O U ) as observed in fication x 12 500. transmission electron niicruscope. Yagni- fication x 1500. FIG. 10.--(c) Higher magnification picture of a FIG. lO.-(d) F,lectron diffraction pattern regicn shown in fig. JOh showing particle or flak: graphite within the film. Magnification x 20 000. obtained from a region shown in fig. 1Oc.FIG. 11 .-Iridium coated fibre after heat FIG. 12.-Scanning electron micrograph of treatment in vacuum for one day at 1373 K. zirconium carbide coatings after heat treat- Magnification x 8000. ment in vacuum at 1273 K. The nickel overcoat has been etched away.(a) Showing longitudinal cracks. Magnification x 3000. FIG. 12.-(b) Showing circumferential cracks. Magnification x 6100. FIG. 12.-(c) Showing absence of attack on the fibre surface where the coating has been re- moved. Magnification x 7200. FIG. 13.-Scanning electron micrograph of pits on a carbon fibre surface after heat treatment in air under a poor carbide coating and nickel FIG. 14.-Scanning electron micrograph of car- bon fibre surface after heat treatment at 873 Kin air under a boron nitride coating, indicating ab- overcoat. Magnification x 6100. sence of surfacebutting. Magnification x 6000.S . V . BARNETT, S . J . HARRIS AND J . V. WEAVER 155 with the original fibre, i.e., there was no preferred orientation, suggesting that the graphite was of three-dimensional form.Iridium also agglomerated in vacuum at temperatures in excess of 1273 K, see fig. 11. The size of the agglomerates was much smaller than with nickel and there was no apparent disruption or pitting of the fibre surface, i.e., the ridges originally present could easily be seen. Electron diffraction evidence of three-dimensional graphite on the surface could not be found. However, fibres coated with iridium and nickel showed a similar tendency towards coating agglomeration, after treatment at 1323 K for one day. Growths were observed between the agglomerates and electron diffraction from these regions gave three-dimensional graphite patterns. Examination of vacuum heat-treated alumina coated fibres in the S.E.M. showed that there was a separation of the coating from the fibre.Various degrees of spalling resulting from circumferential and longitudinal cracking of the alumina coating were noted. Electron diffraction showed that the preferred orientation of the fibre was retained even with specimens heated for 24 h at 1373 K ; the remains of the coating gave rise to diffraction patterns which were indexed as y-alumina. A series of specimens was prepared with zirconium carbide coatings which were in turn overcoated with nickel in the form of a mini-composite ; these were subjected to heat treatment for 96 h in the temperature range 1173 K to 1323 K. After this treat- ment, taper samples were cut from the mini-composite and examined in the optical microscope. It was noted that some of the fibres had seLerely degraded and others had not been affected at all. These observations may be explained in terms of the variable stoichiometry of the original zirconium carbide coatings.In a number of cases, where the nickel was etched away leaving the carbide coating on the fibre, the S.E.M. showed that both longitudinal and circumferential cracks, fig. 12 a-c, have occurred in the coating, but in the main the coating is still adherent to the fibre surface. Where sufficient coating has been removed, the fibre surface was not observed to be damaged. Titanium carbide coatings did not prevent degradation of the fibres after treatment in excess of 1273 K. OXIDATION HEAT TREATMENT After four hours at 673 K, the nickel coat had cracked; longitudinal and radial cracks were observed.At 773 K only radial cracks were observed, and X-ray diffraction analysis on the coating showed that it consisted almost entirely of nickel oxide. Iridium coated fibres suffered no visible damage after treatment at 673 K, at 773 K slight pitting has occurred, but at 873 K the coating had cracked radially and longitudinally. Beneath the regions where the coating had exfoliated, the fibre surface was badly pitted. Electron diffraction patterns obtained on the specimens were indexed, the " d " spacings closely corresponded to those quoted for iridium dioxide. Fibres which had been coated with alumina showed evidence of large scale cracking and spalling after treatment at 873 K, where such disruption had occurred pitting of the fibre surface had taken place, see fig.13. Similar behaviour was noted where a nickel overcoat had been used on top of the alumina ; nickel oxide had formed which had behaved in a brittle manner in conjunction with the alumina. Boron nitride coatings were shown to be rather uneven in thickness ; undoubtedly this was due to the chemical means by which the material had been applied. After treatment for four hours at 773 K and 873 K, there was no evidence of fibre attack on a large scale. When evaporated boron nitride coatings existed under a nickel overcoat for four hours at 773 K, no evidence could be found of attack on the fibre surface, i.e., after the coatings had been removed by ultrasonic agitation, see fig. 14.156 METAL MATRIX COMPOSITES DISCUSSION Barrier layers may be broadly characterized into those that agglomerate, i.e., ductile metals, and those that tend to crack and spall, i.e., brittle non-metals.METALLIC BARRIER LAYERS The results obtained on the nickel coated graphite fibres support the findings of Jackson and Marjoram,' indicating that graphite fibres recrystallize into a three dimensional form after heat treatment above 1273 K and under a pressure of 5 x N m-2. Both electrodeposited and evaporated nickel coats appear to bring about this fibre degradation process. The electron microscope technique shows two types of graphite growing between the nickel spheroids, a fine grained film and flakes. The fine grained film appears to have been left behind as the nickel lowered its surface energy and agglomerated; it is possible that a surface diffusion process has taken place on the nickel as it begins to form the spheroids. The flake graphite was not in the main associated with the fibre surface and must have gone through a process involving dissolution, diffusion and precipitation in the nickel phase, i.e., rather than an in-situ recrystallization process on the fibre.Iridium has been shown to be compatible with carbon fibres for up to 36 h at 1373 K. However, agglomeration is observed to take place, which would not commend it as a barrier layer. No evidence could be found of three-dimensional graphite growths or of fibre weakening. When iridium was interleaved between the nickel and the fibre, the fibres were appreciably weakened and three-dimensional diffraction patterns for graphite were once again observed.Such a process could have taken place either by diffusion of nickel through iridium, or diffusion of carbon through the nickel after iridium agglomeration had taken place. Data are not available on nickel solubility and diffusion in iridium. However, using the data given in tables 2 and 3 together with an assumed diffusion coefficient of - 10-lo mm2/s for Ni in Ir at 1373 K, it may be shown that 99.9 % carbon saturation of the 1 pm iridium layer would have taken place in -20 s, whilst a 50/50 atomic % Ni/Ir alloy would have resulted in -200 s. The carbon solubility of the Ni/Ir alloy may increase markedly with rising nickel content, i.e., if note is taken of the solubility levels of this element in the two metals concerned, see table 2.Since the carbon solubility in iridium is at least an order of magnitude smaller than that of carbon in nickel, it could be simply assumed that simple carbon dissolution was the rate controlling degradation process in nickel-carbon fibre composites. However, Barclay and Bonfield have shown that this process does not degrade fibres at temperatures < 1353 K, and that the impurity content of the fibre-nickel coating and the furnace atmosphere are all important. It is known that oxygen effectively diffuses through nickel and its oxide to attack carbon fibres in the oxidation studies. Therefore, small concentrations of oxygen at the nickel-fibre interface could effectively " catalyze " the dissolution process. Iridium forms a protective oxide and could provide a barrier to oxygen diffusion, hence preventing oxygen from reaching the fibre surface ; thus the damaging aspects of the dissolution process never effectively begin.When nickel is placed on top of the iridium and interdiffusion takes place, the possibility of oxygen permeating the coatings may once again increase, thus leading to fibre degradation. If oxygen does play some role in the dissolution process it is not clearly defined at present, since on thermodynamic and experimental evi- dence l 8 oxygen dissolved in nickel and/or nickel oxide in the presence of carbon in solid solution reacts to form carbon monoxide above 723 K. In the solid state theS. V . BARNETT, S. J . HARRIS AND J . V . WEAVER 157 difficult step in this process is the formation of gas bubbles; however, no evidence has yet been found of such bubbles in coated fibres.Assuming dissolution of carbon takes place, such a process would cease after the limit of solid solubility has been reached, and precipitation would not normally take place, unless the unstable carbide of nickel forms. However, a driving force for precipitation may result from the following : (a) Differences in solubilities between carbonized fibre, graphite fibre and three dimensional graphite, in general a more stable phase has a lower solubility than a less stable phase, i.e., stability of three dimensional graphite > graphitized fibres > carbonized fibre. (b) High local concentrations of impurities, e.g., oxygen, sulphur, etc., in the coating changing the solubilities of carbon in nickel.Finally impurities, in particular compounds occluded as boundary films, etc., may act as heterogeneous nucleation sites for three-dimensional graphite, otherwise nucleation away from the fibre surface would be difficult without a large precipitation driving force. Treatment of nickel coated fibres in an oxidizing atmosphere results in a significant drop in strength at 773 K. The weakening of the fibres is associated with local pitting of the fibre surface, suggesting that some surface sites are more vulnerable to attack. Above 673 K clear evidence exists of nickel oxide formation. The oxide cracks circumferentially and allows attack to take place directly on the fibre surface. Spalling of the iridium coating was observed after treatment at 873 K and evidence was found of iridium dioxide.Some degree of protection was afforded by iridium up to 873 K and, where failures had occurred, the thickness of the iridium may have been insufficient to prevent oxide formation through the section. NON-METALLIC COATINGS The observations made on alumina and the carbides together with nickel and iridium oxides, indicate that crack formation, followed by spalling, are the initial steps in the failure of these brittle barrier layers. Some coatings appear to be more susceptible to this type of failure, e.g., alumina > titanium carbide> zirconium carbide> boron nitride. Three possible reasons for this type of failure are : (i) Oxygen ingress through pores in the coating resulting in gaseous products that expand and lift the coating away.(ii) Volume changes during oxidation of the coating (not alumina) producing sufficient internal stressing to cause coating failure. (iii) Excessive stressing due to thermal mismatch between coating and fibre during cooling. Mechanism (iii) appears to be the most important because of the cracks which have been observed in vacuum heat treated samples, although mechanisms (i) and (ii) may still be relevant in oxidizing conditions. Thermal mismatch results from differences between the moduli of elasticity, Poisson’s ratios, and thermal expansion coefficients of the fibre and coating. Coating thickness also may play a significant part in the process. Longitudinal and circumferential cracks have been observed in both alumina and carbide coatings.An examination of the values of the linear coefficients of expansion of alumina, titanium carbide, zirconium carbide and carbon fibres along the fibre axis, indicates that compressive longitudinal stresses will be set up in virtually all coating materials on heating, resulting in a brittle failure around the circumference of the coated fibre. Alumina has the greatest mismatch with respect to the fibre, followed by titanium carbide and then zirconium carbide. Of the other physical properties to be considered, Poisson’s ratio and elastic modulus do not differ greatly from one coating material to another.158 METAL MATRIX COMPOSITES Boron nitride may be unique in that having a graphite-like structure, it also possesses similar anisotropic properties to the fibres.Accepting this, the stress condition in a boron nitride coating will be dependent upon there being an epitaxial relationship between coating and fibre. APPLICATION OF BARRIER LAYERS The desirable properties of an effective barrier layer for a carbon fibre reinforced nickel may be summarized as a material that possesses a low carbon diffusion rate, a low diffusion rate in the nickel matrix, a low solubility in the matrix, resistance to oxidation, and low oxygen permeability, together with thermal expansion, Poisson ratio and elastic modulus values that are compatible with the fibre. On the basis of diffusion and solubility data, the majority, if not all, metals and alloys may be ex- cluded. The non-metals give some encouragement, particularly in respect of the degree of protection afforded by zirconium carbide under vacuum conditions at 1373 K and boron nitride at > 873 K in air.However, there are appreciable diffi- culties in applying these coating materials to - 8 pm carbon fibres ; obviously evaporation techniques are not ideally suited to coating multi-fibre tows. Consider- ation may be given to the use of such coatings on a duplex system of fibres proposed by M0r1ey.l~ Here, the barrier layer would act as an outer tube on a core prepared from a bundle of carbon fibres, thus reducing the total surface area for coating in any given composite. By controlling the properties of the fibre core/barrier layer and the barrier/metal matrix interfaces it may be possible to introduce a system of weak and strong bonded interfaces which could lead to higher works of fracture in these materials without impairing transverse strength. We are grateful to the late Dr. E. Holmes and Dr. A. A. Baker for many helpful discussions and to the Ministry of Defence for providing a research contract under which part of this work was financed. One of the authors (SVB) gratefully acknow- ledges the award of a studentship by the Science Research Council. The authors are also indebted to Prof. J. S. L1. Leach for the provision of laboratory facilities. P. W. Jackson, Met. Eng. Quart., 1969, 9, 22. P. W. Jackson and J. R. Marjoram, J. Mat. Sci., 1970, 5,9. P. W. Jackson, D. M. Braddick and P. J. Walker, J. Mat. Sci., 1971, 6,419. J. M. Evans and D. M. Braddick, Corrosion Sci., 1971, 11, 611. 0. Kubaschewski, E. L1. Evans and C. B. Alcock, Metallurgical Thermochemistry (Pergamon, London, 4th Edn, 1967), p. 421. J. F. Elliot and M. Gleiser, Thermochemistry for Steelmaking (Addison Wesley, Massachusetts, 1960), p. 131. G. Blankenburgs, Int. Conf. on Interfaces, Melbourne, 1969. * M. Hansen, Constitution of Binary Alloys (McGraw-Hill, New York, 1958). R. B. Barclay and W. Bonfield, J. Mat. Sci., 1971, 6, 1076. lo A. A. Baker, P. W. Jackson and C. Shipman, Fibre Sci. Techn., 1972, 5, 285. l1 S. V. Barnett, PhB. Thesis (University of Nottingham, 1970). l 2 R. A. Andrievsky, V. S. Eremeev, V. N. Zagryazkin and A. Panov, Zzvest. Akad. Nauk. SSSR, l 3 I. I. Kovenskii in Diflusion in bcc Materials (A.S.M., Ohio, 1965). l4 L. V. Pavlinov and V. N. Bykov, Phys. Metal Metallogr., 1965, 19, 73. lS J. M. Criscione, H. F. Volk and A. F. Smith, A.Z.A.A. J., 1966, 4, 1791. J. C. Chaston, Platinum Met. Rev., 1965, 9, 51. l7 J. J. Lander, H. E. Kern and A. L. Beach, J. Appl. Phys., 1952, 23, 1305. D. M. Braddick and S . J. Harris, Trans. Znst. Met. Finishing, 1972, 50, 46. l9 J. G. Morley, Proc. Roy. Soc. A , 1970, 319, 117. Neorg. Materialy, 1967, 3, 2158.