《复合材料 Composites》课程教学资源（学习资料）第二章 增强体_nicalon-1 A review of the development of three generations of small diameter silicon carbide fibres

40TH ANNIVERSARY J MATER SCI41(006)823-839 A review of the development of three generations of small diameter silicon carbide fibres A.R. BUNSELL A. IANT Ecole des Mines de Paris, Centre des Materiaux, BP87, 91003 Evry Cedex, France Three generations of small diameter ceramic fibres based on polycrystalline silicon carbide have been developed over a period of thirty years. This has been possible due to studies into the relationships between the microstructures and properties of the fibres a variety of techniques have been employed by research teams on three continents the fibres are made by he conversion of polymer precursors to ceramic fibres and all three generations are presently produced commercially. The nature of the precursor and the techniques used for cross-linking have been varied in order to optimise both properties and cost of manufacture. It has been possible to improve the characteristics of the fibres as the processes involved in the cross-linking of the precursor fibres have been better understood and the mechanisms governing both room temperature and high temperature behaviour determined. The result is that, although first generation fibres were limited by a low Young s modulus at room limiting the behaviour of bulk silicon carbide, the third generation fibres shows many ot tho oo temperature and by creep and instability of the structure at temperatures far lower than thos characteristics of stoichiometric silicon carbide. This remarkable improvement in characteristics has been due to a thorough understanding of the materials science governing the behaviour of these fibres which are reinforcements for ceramic matrix composite materials 2006 Springer Science Business Media, Inc. 1 Introduction The PCs obtained in this way could be melt spun to give Silicon carbide fibres, with diameters of around 15 um, weak fibres. Stabilisation was initially by cross-linking were first produced commercially in 1982 by Nippon Car- of the polymer by heating in air, just as in the carbon bon. This industrial production was the direct result of fibre production route using PAN precursors. This was research started in the 1970s and carried out by Professor followed by heating in vacuum at temperatures, generally Yajima and his team at the Tohoku University in Japan. around 1200oC and allowed the first generation of small The approach adopted to produce SiC fibres owed much diameter Sic fibres to be produced. The availability of to the experience gained from the development of carbon these Sic fibres brought rapid interest from the aerospace fibres which involves the spinning of polyacrylonitrile and aero-engine industries as they offered the possibil PAN) precursor fibres, which are stabilised by cro ity of producing ceramic fibre reinforced carbon and ce linking and then pyrolysed under controlled conditions ramic matrix composites materials, capable of being used to give carbon fibres. The starting polymer for produc- as structural materials to higher temperatures than those ing SiC fibres, by necessity, needed to contain silicon and attainable with the best nickel based super-alloys. The at- carbon atoms and the polycarbosilane(PCS), which was traction of silicon carbide is that it is a ceramic, which in chosen as the starting material, comprised these elements bulk form, has a Youngs modulus twice that of steel for arranged in a cyclic form consisting of six atoms, which less than half the density and can be used up to 1600c suggested their arrangement in B-SiC. The synthesis of Although oxidised at high temperature, bulk Sic under the PCS used dimethyldichlorosilane( CH3)2 SiCI which goes surface passive oxidation which protects the bulk of was converted into polydimethylsilane [(CH3) Si]n, by the specimen. However it was found that the characteris dechlorination with metal sodium and which in turn was tics of these first generation fibres were not those of bulk converted into a polycarbosilane polymer by heating in SiC. The fibres possessed a Youngs modulus less than an inert atmosphere at 4000C [1]. The chemical compo- half that expected. The fibres crept at 1000C and above sition of PCS can be simplified as -[SiCH3H-CH2In and degraded above 1250%C. The understanding of the 0022-2461 2006 Springer Science+Business Media, Inc DOI:10.1007/10853-0066566-z 823

40TH ANNIVERSARY J MATER SCI 4 1 (2 0 0 6 ) 8 2 3 –8 3 9 A review of the development of three generations of small diameter silicon carbide fibres A. R. BUNSELL, A. PIANT Ecole des Mines de Paris, Centre des Materiaux, BP87, 91003 Evry Cedex, France ´ Three generations of small diameter ceramic fibres based on polycrystalline silicon carbide have been developed over a period of thirty years. This has been possible due to studies into the relationships between the microstructures and properties of the fibres. A variety of techniques have been employed by research teams on three continents. The fibres are made by the conversion of polymer precursors to ceramic fibres and all three generations are presently produced commercially. The nature of the precursor and the techniques used for cross-linking have been varied in order to optimise both properties and cost of manufacture. It has been possible to improve the characteristics of the fibres as the processes involved in the cross-linking of the precursor fibres have been better understood and the mechanisms governing both room temperature and high temperature behaviour determined. The result is that, although first generation fibres were limited by a low Young’s modulus at room temperature and by creep and instability of the structure at temperatures far lower than those limiting the behaviour of bulk silicon carbide, the third generation fibres shows many of the characteristics of stoichiometric silicon carbide. This remarkable improvement in characteristics has been due to a thorough understanding of the materials science governing the behaviour of these fibres which are reinforcements for ceramic matrix composite materials.
C 2006 Springer Science + Business Media, Inc. 1. Introduction Silicon carbide fibres, with diameters of around 15 µm, were first produced commercially in 1982 by Nippon Car￾bon. This industrial production was the direct result of research started in the 1970s and carried out by Professor Yajima and his team at the Tohoku University in Japan. The approach adopted to produce SiC fibres owed much to the experience gained from the development of carbon fibres which involves the spinning of polyacrylonitrile (PAN) precursor fibres, which are stabilised by cross￾linking and then pyrolysed under controlled conditions to give carbon fibres. The starting polymer for produc￾ing SiC fibres, by necessity, needed to contain silicon and carbon atoms and the polycarbosilane (PCS), which was chosen as the starting material, comprised these elements arranged in a cyclic form consisting of six atoms, which suggested their arrangement in β-SiC. The synthesis of the PCS used dimethyldichlorosilane (CH3)2SiCl2 which was converted into polydimethylsilane [(CH3)2Si]n, by dechlorination with metal sodium and which in turn was converted into a polycarbosilane polymer by heating in an inert atmosphere at 400◦C [1]. The chemical compo￾sition of PCS can be simplified as –[SiCH3H-CH2]n –. The PCS obtained in this way could be melt spun to give weak fibres. Stabilisation was initially by cross-linking of the polymer by heating in air, just as in the carbon fibre production route using PAN precursors. This was followed by heating in vacuum at temperatures, generally around 1200◦C and allowed the first generation of small￾diameter SiC fibres to be produced. The availability of these SiC fibres brought rapid interest from the aerospace and aero-engine industries as they offered the possibil￾ity of producing ceramic fibre reinforced carbon and ce￾ramic matrix composites materials, capable of being used as structural materials to higher temperatures than those attainable with the best nickel based super-alloys. The at￾traction of silicon carbide is that it is a ceramic, which in bulk form, has a Young’s modulus twice that of steel for less than half the density and can be used up to 1600◦C. Although oxidised at high temperature, bulk SiC under￾goes surface passive oxidation which protects the bulk of the specimen. However it was found that the characteris￾tics of these first generation fibres were not those of bulk SiC. The fibres possessed a Young’s modulus less than half that expected. The fibres crept at 1000◦C and above and degraded above 1250◦C. The understanding of the 0022-2461
C 2006 Springer Science + Business Media, Inc. DOI: 10.1007/s10853-006-6566-z 823

40TH ANNIVERSARY material science involved in the processes governing this behaviour and the use of this knowledge to produce fibres CH with greatly enhanced properties have been the preoccu pation of a number of laboratories across the world for the last quarter of a century. The result has been the develop ment of three generations of fibres, the latest of which has produced fibres with properties approaching the limits of what is physically possible with silicon carbide C CH3 Figure I Repeat unit of polycarbosilane(PCS). The numbers allow easy 2. First generation fine Sic fibres The interest in silicon carbide as a reinforcement has prompted the development of several types of fibres in luding fibres with diameters usually greater than 100 um made by CVD onto a core filament and also monocrys- talline short filaments known as whiskers with diameters of the order of l um [2, 3]. The Cvd fibre is finding inter est as a reinforcement for titanium but its large diameter handling processes, which are commonly used with finer exploited in the production of SiC fibres. The numbers refer to the positions fibres used in the majority of composite materials. The Sic whiskers also present serious handling difficulties to- gether with worries about health related problems The announcement of the production of Sic based fi- by the cong diameters in the range of 10 to 20 um,made of polycarbosilane precursors, excited considerable interest amongst those people looking for reinforcements capable of operating in an oxidising atmo- sphere at over 1000oC, above the limits of nickel based alloys[4, 5]. Yajima and his colleagues explored a number of routes to produce polycarbosilane which could be used as a precursor for a ceramic fibre [5, 6]. The difficulty lay in the production of a form of polycarbosilane which Figure 3 The first generation of fine SiC fibres were made by cross-linking could be spun and converted into ceramic filaments. The the precursor PCS with oxygen. decomposition of polydimethylsilane(PDs) which was heated in an autoclave at 470oC for 14 h, was eventu- cross-linked. The PCs precursor was made infusible, in ally chosen as the route for the production of PCS as it the first generation of fibres, by crosslinking in air, in the gave a precursor which, although difficult to spin, could temperature range from 145 to 200oC, which introduced be spun from the melt and converted into a ceramic fibre. oxygen into the polymer, as shown in Fig 3. The con- The repeat element in the chemical structure of polycar- version of the cross-linked precursor fibre into a ceramic bosilane is given in Fig. 1. A steric view of this molecule fibre was explored again by Yajima et al. [7-9] and would show that the cycle of carbon and silicon atoms, others [10]. The cross-linked PCS fibres were insoluble with some bonds removed, is arranged in the form of a in all solvents. Heating the cross-linked precursor up to chair configuration, as illustrated in Fig. 2. This reflects 550C induced the evaporation of low molecular weight the arrangement seen in B-SiC. The numbering of the ar- components in the carbosilane which led to a consider- rangement of the atoms is so that a comparison can be able weight loss but resulted in an increase in molecular made to those in Fig. 1. The groups, shown in Fig. l, weight. Above this temperature and up to around 800oC, numbered 2, 3, 5 and 6 are in the same plane whilst 1 hydrogen and methane were lost from the side groups and 4 are out of this plane "Me"represents the methyl in the PCS leaving behind free carbon and cross-linking group(CH3). Unlike most polymers which are spun into was enhanced. Further heating to 1200oC showed that fibres, it was found that very high molecular weight was gas evolution was almost complete at 1000oC At 1050C not necessarily best and polymers with molecular weights hydrogen was again given off and the XRD patterns be- of around 1500 were eventually used for commercial pro- came sharper indicating greater regularity in the struc- duction. The precursor filaments were then spun from the ture. At and above 1300C the free carbon created by the melt in a nitrogen atmosphere at around 300C and then destruction of the methyl groups reacted with the Si-o 824

40TH ANNIVERSARY material science involved in the processes governing this behaviour and the use of this knowledge to produce fibres with greatly enhanced properties have been the preoccu￾pation of a number of laboratories across the world for the last quarter of a century. The result has been the develop￾ment of three generations of fibres, the latest of which has produced fibres with properties approaching the limits of what is physically possible with silicon carbide. 2. First generation fine SiC fibres The interest in silicon carbide as a reinforcement has prompted the development of several types of fibres in￾cluding fibres with diameters usually greater than 100 µm made by CVD onto a core filament and also monocrys￾talline short filaments, known as whiskers, with diameters of the order of 1 µm [2, 3]. The CVD fibre is finding inter￾est as a reinforcement for titanium but its large diameter makes it unsuitable for weaving and other types of fibre handling processes, which are commonly used with finer fibres used in the majority of composite materials. The SiC whiskers also present serious handling difficulties to￾gether with worries about health related problems. The announcement of the production of SiC based fi- bres having diameters in the range of 10 to 20 µm, made by the conversion of polycarbosilane precursors, excited considerable interest amongst those people looking for reinforcements capable of operating in an oxidising atmo￾sphere at over 1000◦C, above the limits of nickel based alloys [4, 5]. Yajima and his colleagues explored a number of routes to produce polycarbosilane which could be used as a precursor for a ceramic fibre [5, 6]. The difficulty lay in the production of a form of polycarbosilane which could be spun and converted into ceramic filaments. The decomposition of polydimethylsilane (PDS) which was heated in an autoclave at 470◦C for 14 h, was eventu￾ally chosen as the route for the production of PCS as it gave a precursor which, although difficult to spin, could be spun from the melt and converted into a ceramic fibre. The repeat element in the chemical structure of polycar￾bosilane is given in Fig. 1. A steric view of this molecule would show that the cycle of carbon and silicon atoms, with some bonds removed, is arranged in the form of a chair configuration, as illustrated in Fig. 2. This reflects the arrangement seen in β-SiC. The numbering of the ar￾rangement of the atoms is so that a comparison can be made to those in Fig. 1. The groups, shown in Fig. 1, numbered 2, 3, 5 and 6 are in the same plane whilst 1 and 4 are out of this plane. “Me” represents the methyl group (CH3). Unlike most polymers which are spun into fibres, it was found that very high molecular weight was not necessarily best and polymers with molecular weights of around 1500 were eventually used for commercial pro￾duction. The precursor filaments were then spun from the melt in a nitrogen atmosphere at around 300◦C and then Figure 1 Repeat unit of polycarbosilane (PCS). The numbers allow easy reference to Fig. 2. CH Si CH2 SiMe 1 2 3 6 H 4 5 CH2 SiMe 1 Figure 2 The steric conformation of the repeat unit of polycarbosilane exploited in the production of SiC fibres. The numbers refer to the positions shown in Fig. 1. Figure 3 The first generation of fine SiC fibres were made by cross-linking the precursor PCS with oxygen. cross-linked. The PCS precursor was made infusible, in the first generation of fibres, by crosslinking in air, in the temperature range from 145 to 200◦C, which introduced oxygen into the polymer, as shown in Fig. 3. The con￾version of the cross-linked precursor fibre into a ceramic fibre was explored again by Yajima et al. [7–9] and others [10]. The cross-linked PCS fibres were insoluble in all solvents. Heating the cross-linked precursor up to 550◦C induced the evaporation of low molecular weight components in the carbosilanes which led to a consider￾able weight loss but resulted in an increase in molecular weight. Above this temperature and up to around 800◦C, hydrogen and methane were lost from the side groups in the PCS leaving behind free carbon and cross-linking was enhanced. Further heating to 1200◦C showed that gas evolution was almost complete at 1000◦C. At 1050◦C hydrogen was again given off and the XRD patterns be￾came sharper indicating greater regularity in the struc￾ture. At and above 1300◦C the free carbon created by the destruction of the methyl groups reacted with the Si–O 824

40TH ANNIVERSARY TABLE I Compositions of early varieties of first generation SiC fibres TABLE II Compositions, Youngs moduli and densities of first gener- produced by Nippon Carbon ation commercialised SiC fibres hemical composition Nippon Carbon Ube Industries Elemental composition %owt Fibre name Nicalon 200 Tyranno LOX-M C SiO, C Precursor PCS Cured by Oxidation xidation NLP10160 Si(wt%) NLM-1025 C(wt%) 31.6 NLP20254 O(wt%) 117 Ti(wt%) C/Si 1.31 1.36 group with the evolution of Co gas and forming a Si-c Young s modm ys (Pa bond. The nascent B-SiC grains, formed at slightly lower temperatures increased in size and the amorphous struc- ture evolved into a semi-crystalline structure consisting of nano sized B-SiC grains surrounded by a much less ordered phase made up of silicon, carbon and oxygen. [12]. The Tyranno fibres could be made with diameters Heating to 1500 C produced large grain growth, the evo- half that of the Nicalon fibres. The precursors of these first lution of carbon monoxide and the disintegration of the Tyranno fibres were also crosslinked in air. Ube Industries fibr used a code to indicate the oxygen content of the fibres The first fibres of this first generation, which were so that the fibre which was commercialised was known as made available by Nippon Carbon around 1982, were Tyranno LOX-M, with the letter M, which is the thirteenth the Nicalon 100 series but were replaced after about four letter in the alphabet, representing an oxygen content of years by the Nicalon 200 series which became the standard approximately 13%by weight grade for much of the ceramic matrix composite studies By the end of the 1980s the two Japanese companies subsequently undertaken. These fibres had diameters of were producing first generation fine diameter Sic around 15 um but showed variability in diameter along fibres and their compositions and densities are shown their length because of the difficulties of spinning the pre- in Table Il. The details of the composition and the cursor fibres, as can be seen from Fig 4. Table I shows nomenclature used to describe the fibres have changed the approximate chemical composition of these fibre slightly since their initial introduction so that the nicalon Yajima and his colleagues had considered several routes fibres mentioned in Table I are now grouped under the to making SiC fibres and one included the addition of simpler heading of Nicalon 100 or 200 series. Bulk Sic titanium to the PCS so as to give polytitanocarbosilane is however the second hardest material known and is (PTC)[11]. This precursor was obtained by the grafting crystalline, it possesses a Youngs modulus of around of titanium alkoxide, Ti(OR)4, in which=CnH2n+l, onto 400 GPa and a density of 3. 15 g/cm and can be used in the PCS chains. This linked the polymer chains together, air up to 1600oC. At this temperature passive oxidation of increasing molecular weight and its spinability. In 1987, the surface to Sio2 protects it from further degradation another Japanese company, Ube Industries announced the Table II reveals that the properties of the first generation production of Tyranno fibres made from PTC precursors of fine Sic fibres were not those of the bulk material and reported that they had better thermal and chemical This has been shown to be due to the non stoichiometric stability compared with the then-existent Nicalon fibres composition of the fibres, which are rich in carbon and contain oxygen, as shown in Tables I and Il. An alternative approach to producing polymer derived SiC fibres, similar to the first generation fibres which have been developed commercially, has been described by university researchers. In contrast to the manufactur ing technique described above, it has been shown that such fibres could be produced from precursor filaments made 864 from high-molecular-weight PCS [13]. The infusible PCS was prepared by pressure pyrolysis of polydimethylsilane Fine fibres were formed by the dry spinning of concen- trated PCs-based polymer solutions which were then py 78910111213141516118192021 rolised in an inert atmosphere in the temperature range of 1000 to 1200C. The fibres produced in this manner were Figure 4 Range of diameters observed with Nicalon NLM 202 fibre taken reported as possessing very similar properties to those of from the same tow [43 the commercialised first generation fibres [14] 825

40TH ANNIVERSARY T A B L E I Compositions of early varieties of first generation SiC fibres produced by Nippon Carbon Elemental composition %wt Chemical composition %wt Fibre type Si C O SiC SiO2 C NLP-101 60 27 13 69 24 7 NLM-102 54 34 12 63 21.5 15.5 NLP-202 54 37 9 66 17 17 group with the evolution of CO gas and forming a Si–C bond. The nascent β-SiC grains, formed at slightly lower temperatures increased in size and the amorphous struc￾ture evolved into a semi-crystalline structure consisting of nano sized β-SiC grains surrounded by a much less ordered phase made up of silicon, carbon and oxygen. Heating to 1500◦C produced large grain growth, the evo￾lution of carbon monoxide and the disintegration of the fibre. The first fibres of this first generation, which were made available by Nippon Carbon around 1982, were the Nicalon 100 series but were replaced after about four years by the Nicalon 200 series which became the standard grade for much of the ceramic matrix composite studies subsequently undertaken. These fibres had diameters of around 15 µm but showed variability in diameter along their length because of the difficulties of spinning the pre￾cursor fibres, as can be seen from Fig. 4. Table I shows the approximate chemical composition of these fibres. Yajima and his colleagues had considered several routes to making SiC fibres and one included the addition of titanium to the PCS so as to give polytitanocarbosilane (PTC) [11]. This precursor was obtained by the grafting of titanium alkoxide, Ti(OR)4, in which R=CnH2n+1, onto the PCS chains. This linked the polymer chains together, increasing molecular weight and its spinability. In 1987, another Japanese company, Ube Industries announced the production of Tyranno fibres made from PTC precursors and reported that they had better thermal and chemical stability compared with the then-existent Nicalon fibres 0 2 4 6 8 10 12 14 16 18 20 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 24 25 Diameter (µm) % Figure 4 Range of diameters observed with Nicalon NLM 202 fibre taken from the same tow [43]. T AB L E I I Compositions, Young’s moduli and densities of first gener￾ation commercialised SiC fibres Producer Nippon Carbon Ube Industries Fibre name Nicalon 200 Tyranno LOX-M Precursor PCS PTC Cured by Oxidation Oxidation Si (wt%) 56.6 54 C (wt%) 31.7 31.6 O (wt%) 11.7 12.4 Ti (wt%) 0 2.0 C/Si 1.31 1.36 Young’s modulus (GPa) 200 185 Density (g/cm3) 2.55 2.37 [12]. The Tyranno fibres could be made with diameters half that of the Nicalon fibres. The precursors of these first Tyranno fibres were also crosslinked in air. Ube Industries used a code to indicate the oxygen content of the fibres so that the fibre which was commercialised was known as Tyranno LOX-M, with the letter M, which is the thirteenth letter in the alphabet, representing an oxygen content of approximately 13% by weight. By the end of the 1980s the two Japanese companies were producing first generation fine diameter SiC fibres and their compositions and densities are shown in Table II. The details of the composition and the nomenclature used to describe the fibres have changed slightly since their initial introduction so that the Nicalon fibres mentioned in Table I are now grouped under the simpler heading of Nicalon 100 or 200 series. Bulk SiC is however the second hardest material known and is crystalline, it possesses a Young’s modulus of around 400 GPa and a density of 3.15 g/cm3 and can be used in air up to 1600◦C. At this temperature passive oxidation of the surface to SiO2 protects it from further degradation. Table II reveals that the properties of the first generation of fine SiC fibres were not those of the bulk material. This has been shown to be due to the non stoichiometric composition of the fibres, which are rich in carbon and contain oxygen, as shown in Tables I and II. An alternative approach to producing polymer derived SiC fibres, similar to the first generation fibres which have been developed commercially, has been described by university researchers. In contrast to the manufactur￾ing technique described above, it has been shown that such fibres could be produced from precursor filaments made from high-molecular-weight PCS [13]. The infusible PCS was prepared by pressure pyrolysis of polydimethylsilane. Fine fibres were formed by the dry spinning of concen￾trated PCS-based polymer solutions which were then py￾rolised in an inert atmosphere in the temperature range of 1000 to 1200◦C. The fibres produced in this manner were reported as possessing very similar properties to those of the commercialised first generation fibres [14]. 825

40TH ANNIVERSARY TABLE III Details of manufacture, elemental composition and approximate cost of all three generations of SiC based fibres maximum Cross linki Elemental composition Trade mark Manufacture temperature (wt%) (g/cm)(um)(US S/g First Gen. Nicalon 200 Nippon Carbon Oxygen 1200°C 56si+32C+120 2.5514 2000 Tyranno LOX-M Ube Ind. 1200°C 54Si+32C+120+2T24811 Second Hi-Nicalon Electron °C 1300° 625Si+37C+0502.7412 8000 Gen irradiation Tyranno LOX-E Ube Ind Electron 1300°C 55Si+37.5+5.50+2T2.3911 radiation Tyranno ZM Ube Ind Oxygen 1300°C 57Si+345C+7.50+248 11 1500 Tyranno ZE Ube Ind Electron 1300°C 585Si+38.5C+20+2.5511 Third Gen. Tyranno SA 68Si+32C+0.6Al Tyranno SA 3 Ind 68Si+32C+06Al Sylramic COI ceramics 67Si+29c+0.80+3.0 23B+04N+2.1T Sylramic iB COI Ceramics Oxygen >1700°CN/A 3.05 002 Hi-Nicalon Type-S Nippon Carbon Electron 1500°C 69Si+31C+0.20 3.05 3000 irradiation TABLE IV Details of mechanical and thermal properties of all three generations of SiC based fibres Thermal expansion Room om temperature coefficient, ppm/C(to emal conductivity W/m tempera gs modulus Trade mark Manufacturer 1000°C)[1 K[15 rength(GPa)(GPa) First gen Nicalon 200 Nippon Carbon 3.2 Second Gen. Hi-Nicalon Nippon Carbon 3.5 Tyranno LOX-E Ube Ind Tyranno ZM Third gen Tyranno SA3 Sylramic iBN Hi-Nicalon Type-s Nippon Carbon NA 3. Mechanical behaviour of first generation Sic The behaviour of the first generation fibres remains lin- fibres early elastic up to 1250C but short term strength begins to At room temperature the fibres show linearly elastic be- fall around 1000oC. There is a difference, particularly for haviour. The variation in fibre diameter along individual the Tyranno LOX-M fibre, when tested in air or in argon, fibres makes the measurement of stress and modulus in- with an earlier onset of strength reduction being observed herently difficult which explains some discrepancies in the when the fibres are tested in air indicating a higher sen- published data for these fibres which in any case have been sitivity to carbon oxidation of the surfaces. However the improved since their introduction. Some typical property oxidation of the Nicalon 100 series fibres could be bene- data for first generation SiC fibres can be found in Tables ficial particularly under long term loading conditions as it II and IV. It should be noted that the Tyranno LOX-M, al- slowed internal decomposition of the fibres [17]. Growth nerally available. Fig. 5 shows of silica is observed on the surfaces of both fibre the fracture morphology of a first generation Nicalon fi- they are heated in air at 1200C and above in air. This bre broken in tension [15]. It can be seen that the fracture layer can have an irregular thickness along the fibre and suggests a glassy structure of the fibre. A critical stress pores are formed at the silica/SiC fibre interface and pores intensity factor K lc of 2 MPa.m was determined for the can be formed at 1450C which induce local decohesion Nicalon NLM-202 fibre which is more characteristic of of the silica layer from the fibre. These pores are pro glass than bulk Sic [16] duced by the outgassing of carbon monoxide from the

40TH ANNIVERSARY T A B L E I I I Details of manufacture, elemental composition and approximate cost of all three generations of SiC based fibres Trade mark Manufacturer Cross linking method Approximate maximum production temperature Elemental composition (wt%) Density (g/cm3) Average diameter (µm) Cost (US $/kg) First Gen. Nicalon 200 Nippon Carbon Oxygen 1200◦C 56Si + 32C + 12O 2.55 14 2000 Tyranno LOX-M Ube Ind. Oxygen 1200◦C 54Si + 32C +12O + 2Ti 2.48 11 1250 Second Gen. Hi-Nicalon Nippon Carbon Electron irradiation 1300◦C 62.5Si + 37C +O.5O 2.74 12 8000 Tyranno LOX-E Ube Ind. Electron irradiation 1300◦C 55Si + 37.5 +5.5O + 2Ti 2.39 11 N/A Tyranno ZM Ube Ind. Oxygen 1300◦C 57Si + 34.5C + 7.5O + 1Zr 2.48 11 1500 Tyranno ZE Ube Ind. Electron irradiation 1300◦C 58.5Si + 38.5C + 2O + 1Zr 2.55 11 N/A Third Gen. Tyranno SA 1 Ube Ind. Oxygen >1700◦C 68Si + 32C +0.6Al 3.02 11 N/A Tyranno SA 3 Ube Ind. Oxygen >1700◦C 68Si + 32C +0.6Al 3.1 7.5 5000 Sylramic COI ceramics Oxygen >1700◦C 67Si + 29C +0.8O + 2.3B +0.4N +2.1Ti 3.05 10 10000 Sylramic iBN COI Ceramics Oxygen >1700◦C N/A 3.05 10 >10000 Hi-Nicalon Type-S Nippon Carbon Electron irradiation >1500◦C 69Si + 31C + 0.2O 3.05 12 13000 T A B L E I V Details of mechanical and thermal properties of all three generations of SiC based fibres Trade mark Manufacturer Thermal expansion coefficient, ppm/◦C (to 1000◦C) [15] Room temperature axial thermal conductivity W/m K [15] Room temperature strength (GPa) Room temperature Young’s modulus (GPa) First Gen. Nicalon 200 Nippon Carbon 3.2 3 3 200 Tyranno LOX-M Ube Ind. 3.1 1.5 3.3 185 Second Gen. Hi-Nicalon Nippon Carbon 3.5 8 2.8 270 Tyranno LOX-E Ube Ind. NA NA 2.9 200 Tyranno ZM Ube Ind. NA 2.5 3.4 200 Tyranno ZE Ube Ind. NA NA 3.5 233 Third Gen. Tyranno SA1 Ube Ind. NA 65 2.8 375 Tyranno SA3 Ube Ind. NA 65 2.9 375 Sylramic COI Ceramics 5.4 46 3.2 400 Sylramic iBN COI Ceramics 5.4 >46 3.5 400 Hi-Nicalon Type-S Nippon Carbon NA 18 2.5 400 3. Mechanical behaviour of first generation SiC fibres At room temperature the fibres show linearly elastic be￾haviour. The variation in fibre diameter along individual fibres makes the measurement of stress and modulus in￾herently difficult which explains some discrepancies in the published data for these fibres which in any case have been improved since their introduction. Some typical property data for first generation SiC fibres can be found in Tables III and IV. It should be noted that the Tyranno LOX-M, al￾though still made, is not generally available. Fig. 5 shows the fracture morphology of a first generation Nicalon fi- bre broken in tension [15]. It can be seen that the fracture suggests a glassy structure of the fibre. A critical stress intensity factor K1c of 2 MPa.m1/2 was determined for the Nicalon NLM-202 fibre which is more characteristic of glass than bulk SiC [16]. The behaviour of the first generation fibres remains lin￾early elastic up to 1250◦C but short term strength begins to fall around 1000◦C. There is a difference, particularly for the Tyranno LOX-M fibre, when tested in air or in argon, with an earlier onset of strength reduction being observed when the fibres are tested in air indicating a higher sen￾sitivity to carbon oxidation of the surfaces. However the oxidation of the Nicalon 100 series fibres could be bene- ficial particularly under long term loading conditions as it slowed internal decomposition of the fibres [17]. Growth of silica is observed on the surfaces of both fibres when they are heated in air at 1200◦C and above in air. This layer can have an irregular thickness along the fibre and pores are formed at the silica/SiC fibre interface and pores can be formed at 1450◦C which induce local decohesion of the silica layer from the fibre. These pores are pro￾duced by the outgassing of carbon monoxide from the 826

40TH ANNIVERSARY grain growth depended on the temperature but the time to stabilisation corresponded to the period of primary creep observed at higher loads. At temperatures above 1150oC, for the Nicalon 200 fibres and above 1050. for the served that it was possible to measure an initial shrinkag followed by positive creep showing that two mechanisms were in competition. The variation of steady state creep rate with temperature, T, and applied stress, o, can often be modelled by: 5um where A is a constant depending on the material. The stress exponent, n, and the activation energy for creep, 2, can be deduced from creep experiments and their val- Figure 5 Fracture morphology of a first generation Nicalon fibre [15]- ues can suggest the processes involved. Below 1200 C, the stress component is near to unity for Nicalon 200 fi interior of the fibre. Failure surfaces remain brittle at high bres revealing Newtonian creep, most probably caused by temperature but new types of defects are seen compared grain boundary sliding and controlled by the oxygen rich to those found at room temperature. Local chemical in- homogeneities at the fibres surfaces such as carbon rich increases to 2 as the intergranular phase decomposes and grain sliding becomes more difficult. zones are preferentially decomposed or oxidized giving The two mechanisms which were in competition were rise to porous weak regions The fibres were seen to creep from around 1000C[17] therefore grain growth and perfection and grain bound although, when first observed, this was considered a con- ary sliding. The results are that the creep curves of first troversial observation as most studies indicated that the generation fibres show primary creep which lasts several fibres shrank on heating above this temperature. These hours followed by secondary creep with no tertiary creep stage up to 1250C Shrinkage could be reduced in these latter conclusions were based on fibres heat treated un- fibres by heat treatment at 1200 C in Argon for 5 h which der no imposed load however the creep observations were induces densification and an increase in Youngs modulus made on fibres subjected to loads at temperature. This of around 15 GPa. The creep activation energy was found still leads to some confusion in the literature with some researchers referring to fibre stability as reflecting the to be around 250kJ/mol characteristics of the fibres, tested at room temperature As the applied stress was increased, at a given tem- after heat treatment. The present authors prefer to consider perature, the period during which shrinkage was ob- the characteristics under load at high temperature, which served decreased until a stress was reached at which only positive creep was measured although, as inti better reflects possible end use conditions. The strength mated above, the processes inducing shrinkage still and elastic moduli characteristics of a material are or- occurred and controlled the period of primary creep ten altered at high temperature with respect to those at Table v shows the stresses as a function of temperature room temperature, because of reversible mechanisms at below which shrinkage has been observed for the two the level of the finest structure of the material and so do not become apparent when the temperature is lowered. fibres Exposure to high temperatures can also, of course, induce Creep curves of the first generation fibres revealed the irreversible changes in the fibre structure which also need existence of stress level thresholds, defined as a creep rate to be understood. The testing conditions for defining sta bility at temperature will be explained as necessary in this TABLE V Maximum stress at which shrinkage has been detected in first ge eneration SiC based fibres Further studies on the first generation fibres revealed that. under low loads. the fibres did shrink but, under Temperature(C) Nicalon 200 Tyranno LOX-M higher loads they crept. This behaviour was observed from around 1000oc. studies on the nicalon 100 series fibres which first revealed this behaviour showed that shrinkage 150 0.34Gl could be attributed to B-Sic grain growth in the fibres 1350 0.18 0. 08 GPa hich stabilised with grain sizes around 3 nm. The rate of 1450 006Gl 0.19 827

40TH ANNIVERSARY Figure 5 Fracture morphology of a first generation Nicalon fibre [15]. interior of the fibre. Failure surfaces remain brittle at high temperature but new types of defects are seen compared to those found at room temperature. Local chemical in￾homogeneities at the fibres’ surfaces such as carbon rich zones are preferentially decomposed or oxidized giving rise to porous weak regions. The fibres were seen to creep from around 1000◦C [17], although, when first observed, this was considered a con￾troversial observation as most studies indicated that the fibres shrank on heating above this temperature. These latter conclusions were based on fibres heat treated un￾der no imposed load however the creep observations were made on fibres subjected to loads at temperature. This still leads to some confusion in the literature with some researchers referring to fibre stability as reflecting the characteristics of the fibres, tested at room temperature, after heat treatment. The present authors prefer to consider the characteristics under load at high temperature, which better reflects possible end use conditions. The strength and elastic moduli characteristics of a material are of￾ten altered at high temperature with respect to those at room temperature, because of reversible mechanisms at the level of the finest structure of the material and so do not become apparent when the temperature is lowered. Exposure to high temperatures can also, of course, induce irreversible changes in the fibre structure which also need to be understood. The testing conditions for defining sta￾bility at temperature will be explained as necessary in this paper. Further studies on the first generation fibres revealed that, under low loads, the fibres did shrink but, under higher loads they crept. This behaviour was observed from around 1000◦C. Studies on the Nicalon 100 series fibres which first revealed this behaviour showed that shrinkage could be attributed to β-SiC grain growth in the fibres which stabilised with grain sizes around 3 nm. The rate of grain growth depended on the temperature but the time to stabilisation corresponded to the period of primary creep observed at higher loads. At temperatures above 1150◦C, for the Nicalon 200 fibres and above 1050◦C, for the Tyranno LOX-M fibres, and under low loads, it was ob￾served that it was possible to measure an initial shrinkage followed by positive creep showing that two mechanisms were in competition. The variation of steady state creep rate with temperature, T, and applied stress, σ, can often be modelled by: ε˙ = Aσn exp
−Q RT where A is a constant depending on the material. The stress exponent, n, and the activation energy for creep, Q, can be deduced from creep experiments and their val￾ues can suggest the processes involved. Below 1200◦C, the stress component is near to unity for Nicalon 200 fi- bres revealing Newtonian creep, most probably caused by grain boundary sliding and controlled by the oxygen rich intergranular phase. Above 1200◦C, the stress component increases to 2 as the intergranular phase decomposes and grain sliding becomes more difficult. The two mechanisms which were in competition were therefore grain growth and perfection and grain bound￾ary sliding. The results are that the creep curves of first generation fibres show primary creep which lasts several hours followed by secondary creep with no tertiary creep stage up to 1250◦C. Shrinkage could be reduced in these fibres by heat treatment at 1200◦C in Argon for 5 h which induces densification and an increase in Young’s modulus of around 15 GPa. The creep activation energy was found to be around 250 kJ/mol. As the applied stress was increased, at a given tem￾perature, the period during which shrinkage was ob￾served decreased until a stress was reached at which only positive creep was measured although, as inti￾mated above, the processes inducing shrinkage still occurred and controlled the period of primary creep. Table V shows the stresses as a function of temperature below which shrinkage has been observed for the two fibres. Creep curves of the first generation fibres revealed the existence of stress level thresholds, defined as a creep rate T A B L E V Maximum stress at which shrinkage has been detected in first generation SiC based fibres Temperature (◦C) Nicalon 200 Tyranno LOX-M 1050 - 1.00 GPa 1150 - 0.65 GPa 1250 0.34 GPa 0.40 Gpa 1350 0.18 GPa 0.08 GPa 1450 0.06 GPa 0.19 827

40TH ANNIVERSARY lower than 10-10 s-, which reduced as the temperature the methyl groups present in the PCS and PTC poly increased mers. The presence of oxygen in the precursor fibres Similar mechanical data have been reported for the lab- induced the out-gassing of carbon oxides between 400 oratory produced fibres produced from high-molecular- and 600 C at the beginning of the pyrolysis so reduc- weight PCS, both in tension and in creep[18] ing the final carbon content of the ceramic fibre most The creep data given in this paper, which come from important was the understanding that the presence of the authors'laboratory, have been obtained using ma- the oxygen produced an amorphous phase in which the hines dedicated to the high temperature testing of small small Sic grains and the free carbon aggregates were diameter filaments These machines have been described embedded elsewhere [19] and are capable of conducting tensile, re- The first fibre of this generation to be studied by XRD laxation, creep and, if necessary, fatigue tests on very and TEM was the Nicalon NLM-102 fibre with a 16 wt% small diameter filaments over a wide range of temper- content [17]. This study was followed by others on the atures. Other research teams have conducted creep tests same type of fibre and gave greater information on the using dead weights supported by the fibres. An alternative bonding of the oxygen atoms with neighbouring silicon and rapid way of ranking the creep or relaxation behaviour and carbon atoms [21, 22]. These studies revealed that of ceramic fibres is to put them into the form of a loop the fibre consisted of very small B-SiC grains with an around a former and to heat treat them. If the fibre re- average size, as calculated from the XRD patterns using mains perfectly elastic during the test its form returns to the Sherrer method, of 1.7 nm in size and these were its original, straight form on cooling and removal from embedded in an oxygen rich amorphous silicate phase the former. However if any anelastic behaviour occurs the The B-Sic grains were the only crystalline phase de- fibre retains a curvature. The degree of the induced cur- tected by XRD and electron diffraction. The early grade vature is taken as a mean of ranking the probable creep was later replaced by what was called the ceramic grade behaviour of the fibre compared to that of other fibres Nicalon NLM-202 fibre, the composition of which dif- [20] fered slightly from the earlier grade as it had a lower oxygen content of 12%. Studies on this fibre, which be- came the standard and most widely studied first gener- tion fibre, by TEM, allowed the structure of the free 3.1. Compositions and microstructures carbon to be determined and the mean composition of of first generation fibres the amorphous phase to be proposed [23]. Originally the Fig. 5 shows a typical failure surface of a first genera- oxygen was presumed to be included in an amorphous tion Sic fibre. There is no sign of any granular texture silica phase. Further studies showed that only a fraction so that the fibre appears to be glassy and possibly amor- of the oxygen was in the form of SiO2 and the rest was phous, as was originally concluded. However, wide angle present in the fibre in the form of a ternary phase SiO, Cy X-ray diffraction(XRD) studies and later examination by [21]. The fibre therefore seen to be composed of a transmission electron microscopy (TEM) showed that the continuum of SiO, C, tetrahedra with x +y=4.The first generation fibres were nano-crystalline with sizes of Nicalon 200 grade fibre with a 1l wt. oxygen content B-Sic grains generally being in the range 1.7 to 2 nm was determined as being composed of SiC4 tetrahedra although some variation occurred in early fibres due, pre- formed into small crystallites of 1.4 nm with a diamond sumably, to optimisation by the manufacturers in the du- like structure separated by two Sio, Cy tetrahedra (x+ ration of pyrolysis [17]. O), whereas the 100 grade fibres with 16 wt. of oxygen The compositions by weight percentages and densi- had a lower crystallinity. The Nicalon 200 grade fibre was s of the fibres are given in Table Il, together with the composed of 55 wt% B-SiC grains, 40% of an intergran carbon-to-silicon atomic weight ratios. It is clear that the ular phase with a mean composition of SiO,, 15 Co85 and first generation fibres were far from being stoichiomet- 5% of randomly oriented free carbon aggregates, I nm ric silicon carbide and that they contained an excess of in size. Coo2 lattice fringe images showed small stacks oxygen and carbon. This observation was the beginning of two fringes around 0.7 nm in size suggesting that the of an understanding of the easons why the fibres had basic structural unit(BSU) was a face-to-face associa- such different characteristics from those of bulk Sic. Al- tion of aromatic rings, called dicoronenes, in which the though the manufacture of these fibres from PCS and hydrogen-to-carbon atomic ratio is 0.5. With such a model PTC was analogous to that of carbon fibres made from for the microstructure of the fibre, a porosity level of PAN, an important difference quickly became evident, more than 2% was calculated [23]. Other authors pro which was that the oxygen, used to crosslink and re posed that the intergranular phase should be written as der infusible the precursors before pyrolysis, introduced Sio, Cl-xp which suggests that the composition could into the PAN fibres, is entirely removed at high temper- vary continuously from SiC to SiOz as the oxygen concen- atures but that used to stabilise the pcs and ptc fibres ation varied [24]. This gave a composition by weight of remains. The excess carbon in the SiC fibres came from 56% SiC, 10%C and 34% SiO.1 Co44

40TH ANNIVERSARY lower than 10−10 s−1, which reduced as the temperature increased. Similar mechanical data have been reported for the lab￾oratory produced fibres produced from high-molecular￾weight PCS, both in tension and in creep [18]. The creep data given in this paper, which come from the authors’ laboratory, have been obtained using ma￾chines dedicated to the high temperature testing of small diameter filaments. These machines have been described elsewhere [19] and are capable of conducting tensile, re￾laxation, creep and, if necessary, fatigue tests on very small diameter filaments over a wide range of temper￾atures. Other research teams have conducted creep tests using dead weights supported by the fibres. An alternative and rapid way of ranking the creep or relaxation behaviour of ceramic fibres is to put them into the form of a loop around a former and to heat treat them. If the fibre re￾mains perfectly elastic during the test its form returns to its original, straight form on cooling and removal from the former. However if any anelastic behaviour occurs the fibre retains a curvature. The degree of the induced cur￾vature is taken as a mean of ranking the probable creep behaviour of the fibre compared to that of other fibres [20]. 3.1. Compositions and microstructures of first generation fibres Fig. 5 shows a typical failure surface of a first genera￾tion SiC fibre. There is no sign of any granular texture so that the fibre appears to be glassy and possibly amor￾phous, as was originally concluded. However, wide angle X-ray diffraction (XRD) studies and later examination by transmission electron microscopy (TEM) showed that the first generation fibres were nano-crystalline with sizes of β-SiC grains generally being in the range 1.7 to 2 nm although some variation occurred in early fibres due, pre￾sumably, to optimisation by the manufacturers in the du￾ration of pyrolysis [17]. The compositions by weight percentages and densi￾ties of the fibres are given in Table II, together with the carbon-to-silicon atomic weight ratios. It is clear that the first generation fibres were far from being stoichiomet￾ric silicon carbide and that they contained an excess of oxygen and carbon. This observation was the beginning of an understanding of the reasons why the fibres had such different characteristics from those of bulk SiC. Al￾though the manufacture of these fibres from PCS and PTC was analogous to that of carbon fibres made from PAN, an important difference quickly became evident, which was that the oxygen, used to crosslink and ren￾der infusible the precursors before pyrolysis, introduced into the PAN fibres, is entirely removed at high temper￾atures but that used to stabilise the PCS and PTC fibres remains. The excess carbon in the SiC fibres came from the methyl groups present in the PCS and PTC poly￾mers. The presence of oxygen in the precursor fibres induced the out-gassing of carbon oxides between 400 and 600◦C at the beginning of the pyrolysis so reduc￾ing the final carbon content of the ceramic fibre. Most important was the understanding that the presence of the oxygen produced an amorphous phase in which the small SiC grains and the free carbon aggregates were embedded. The first fibre of this generation to be studied by XRD and TEM was the Nicalon NLM-102 fibre with a 16 wt.% content [17]. This study was followed by others on the same type of fibre and gave greater information on the bonding of the oxygen atoms with neighbouring silicon and carbon atoms [21, 22]. These studies revealed that the fibre consisted of very small β-SiC grains with an average size, as calculated from the XRD patterns using the Sherrer method, of 1.7 nm in size and these were embedded in an oxygen rich amorphous silicate phase. The β-SiC grains were the only crystalline phase de￾tected by XRD and electron diffraction. The early grade was later replaced by what was called the ceramic grade Nicalon NLM-202 fibre, the composition of which dif￾fered slightly from the earlier grade as it had a lower oxygen content of 12%. Studies on this fibre, which be￾came the standard and most widely studied first gener￾ation fibre, by TEM, allowed the structure of the free carbon to be determined and the mean composition of the amorphous phase to be proposed [23]. Originally the oxygen was presumed to be included in an amorphous silica phase. Further studies showed that only a fraction of the oxygen was in the form of SiO2 and the rest was present in the fibre in the form of a ternary phase SiOxCy [21]. The fibre was therefore seen to be composed of a continuum of SiOxCy tetrahedra with x + y = 4. The Nicalon 200 grade fibre with a 11 wt.% oxygen content was determined as being composed of SiC4 tetrahedra formed into small crystallites of 1.4 nm with a diamond￾like structure separated by two SiOxCy tetrahedra (x = 0), whereas the 100 grade fibres with 16 wt.% of oxygen had a lower crystallinity. The Nicalon 200 grade fibre was composed of 55 wt.% β-SiC grains, 40% of an intergran￾ular phase with a mean composition of SiO1.15C0.85 and 5% of randomly oriented free carbon aggregates, 1 nm in size. C002 lattice fringe images showed small stacks of two fringes around 0.7 nm in size suggesting that the basic structural unit (BSU) was a face-to-face associa￾tion of aromatic rings, called dicoronenes, in which the hydrogen-to-carbon atomic ratio is 0.5. With such a model for the microstructure of the fibre, a porosity level of more than 2% was calculated [23]. Other authors pro￾posed that the intergranular phase should be written as SiOxC1−x/2 which suggests that the composition could vary continuously from SiC to SiO2 as the oxygen concen￾tration varied [24]. This gave a composition by weight of 56% SiC, 10% C and 34% SiO1.12C0.44. 828

40TH ANNIVERSARY The existence of porosity in the first generation ceramic If improved fibres were to be made, which could go fibres was proposed so as to account for the outgassing some way to meeting the promise of a pure SiC, it was of the hundreds of volumes of gas per volume of fibre, clear that a more stoichiometric composition had to be which occurs during the transformation from an organic achieved. As mentioned above. it had been seen that the to a mineral structure, which occurs during the manu- presence of oxygen in the fibre resulted in a poorly or- facture of the fibre [25]. It was reasoned that nanometric ganised phase in which the Sic grains were embedded channels must exist in the fibres during the transformation It was clear that the oxygen content had to be consider stage as diffusion through a solid phase would be too slow. ably reduced if improvements were to be made. As the These nanochannels collat further heating above the primary reason that the oxygen was in the fibres was that temperature at which the gas is evolved. A porosity of it was introduced to render the precursor fibres infusible 6.2% was shown to exist experimentally by X-ray scatter- it was clear that other means of cross-linking the polymer ing measurements on Nicalon NLM-202 fibres pyrolized precursors had to be investigated at1400°C[26,27 The result of the studies on the first generation fibres as the conclusion that it was the non-stoichiometric com- 4. Second generation small diameter SiC fibres position of the fibres which were limiting their physical The route adopted to reduce the oxygen content in the characteristics. The presence of the amorphous intergran- second generation of small diameter SiC fibres was to in the fibres after pyrolysis making a Si-o-C phase. The render the PCS precursor fibre infusible. This was possible low fraction of a granular Sic phase accounted for the by using different types of irradiation which could interact Youngs modulus of the fibres being only half that of with the precursor polymers to produce free radicals and bulk SiC. The amorphous phase also explained why the gaseous products by the scission of the chemical bonds fibres began to lose strength and creep at temperatures of Si-CH3, Si-H and C-H. This allowed Si-Si and Si-C around 1000 or 1100oc whereas bulk sic would be bonds to be formed. a number of different types of radi pected to resist to higher temperatures [28]. Heating to ation were investigated by both fibre producers working above 1500.C induced rapid grain growth and outgassing in collaboration with the Japanese Atomic Energy Re- of Sio and Co which came from the oxygen in the amor- search Institute. Gamma irradiation was investigated but phous Si-o-C phase and also the free carbon in the fibre, finally electron radiation in a helium atmosphere was used as described by the following relationship to make the second generation fibres. The cross-linking step was followed by heat treatment at 327 C for a short time to eliminate the remaining free radicals which SiC,O,- SiC()+ Sio(g)+Co(g) apped in the irradiated precursor fibre [35]. An example of how the hydrogen atom, bonded to the silicon atom, as It had been noticed in early studies that the rate of shown in Fig. 2, is removed by electron bombardment, so strength loss of first generation fibres was lower in as to allow direct bonding between the two silicon atoms idizing atmospheres than in an inert argon atmosphere in neighbouring molecules, is shown in Fig. 6[36] [17, 29, 30]. It was concluded that heating the fibres in air Below 550 C cross-linking between main chains dom- produced a silica coating which hindered outgassing of inates and is induced by the dehydrogenation condensa the products of the decomposition of the Si-o-C phase; tion of the Si-H groups. From 550oC to 800C, the side this was further studied by heating fibres made from both chains on the cross-linked polymer begin to decompose the pCS and PtC precursors over a wide range of oxygen and CH4 and H2 are given off producing an inorganic fi partial pressures and in carbon dioxide [31, 32]. These bre. Above 800C and up to 1000C, hydrogen is given authors also showed that the formation of a barrier coat- off, most probably associated with the decomposition ing of Sic laid down by chemical vapour deposition on the fibre surface suppressed gas evolution and slowed the degradation process. They also showed that, what had been presumed in earlier studies was actually the case and that a solid layer of Sio2 formed on the surface of the Nicalon fibres and a Sioz layer with a small amount of TiO2 was formed on the surfaces of the Tyranno fibres It should however be noted that increasing the pressure an argon environment also hindered the onset of decomposition. When the first generation fibres wer heated in Argon at a pressure of 138 MPa the onset of Figure 6 Direct cross-linking of the PCS precursor polymer by irradiation fibre weight loss increased from 1200 to 1500C [34]

40TH ANNIVERSARY The existence of porosity in the first generation ceramic fibres was proposed so as to account for the outgassing of the hundreds of volumes of gas per volume of fibre, which occurs during the transformation from an organic to a mineral structure, which occurs during the manu￾facture of the fibre [25]. It was reasoned that nanometric channels must exist in the fibres during the transformation stage as diffusion through a solid phase would be too slow. These nanochannels collapse on further heating above the temperature at which the gas is evolved. A porosity of 6.2% was shown to exist experimentally by X-ray scatter￾ing measurements on Nicalon NLM-202 fibres pyrolized at 1400◦C [26, 27]. The result of the studies on the first generation fibres was the conclusion that it was the non-stoichiometric com￾position of the fibres which were limiting their physical characteristics. The presence of the amorphous intergran￾ular phase was clearly due to the oxygen which remained in the fibres after pyrolysis making a Si–O–C phase. The low fraction of a granular SiC phase accounted for the Young’s modulus of the fibres being only half that of bulk SiC. The amorphous phase also explained why the fibres began to lose strength and creep at temperatures around 1000 or 1100◦C whereas bulk SiC would be ex￾pected to resist to higher temperatures [28]. Heating to above 1500◦C induced rapid grain growth and outgassing of SiO and CO which came from the oxygen in the amor￾phous Si–O–C phase and also the free carbon in the fibre, as described by the following relationship. SiCxOy → SiC (s) + SiO(g) + CO(g) It had been noticed in early studies that the rate of strength loss of first generation fibres was lower in ox￾idizing atmospheres than in an inert argon atmosphere [17, 29, 30]. It was concluded that heating the fibres in air produced a silica coating which hindered outgassing of the products of the decomposition of the Si–O–C phase; this was further studied by heating fibres made from both the PCS and PTC precursors over a wide range of oxygen partial pressures and in carbon dioxide [31, 32]. These authors also showed that the formation of a barrier coat￾ing of SiC laid down by chemical vapour deposition on the fibre surface suppressed gas evolution and slowed the degradation process. They also showed that, what had been presumed in earlier studies was actually the case and that a solid layer of SiO2 formed on the surface of the Nicalon fibres and a SiO2 layer with a small amount of TiO2 was formed on the surfaces of the Tyranno fibres [33]. It should however be noted that increasing the pressure of an argon environment also hindered the onset of decomposition. When the first generation fibres were heated in Argon at a pressure of 138 MPa the onset of fibre weight loss increased from 1200 to 1500◦C [34]. If improved fibres were to be made, which could go some way to meeting the promise of a pure SiC, it was clear that a more stoichiometric composition had to be achieved. As mentioned above, it had been seen that the presence of oxygen in the fibre resulted in a poorly or￾ganised phase in which the SiC grains were embedded. It was clear that the oxygen content had to be consider￾ably reduced if improvements were to be made. As the primary reason that the oxygen was in the fibres was that it was introduced to render the precursor fibres infusible, it was clear that other means of cross-linking the polymer precursors had to be investigated. 4. Second generation small diameter SiC fibres The route adopted to reduce the oxygen content in the second generation of small diameter SiC fibres was to eliminate the oxygen induced cross-linking step used to render the PCS precursor fibre infusible. This was possible by using different types of irradiation which could interact with the precursor polymers to produce free radicals and gaseous products by the scission of the chemical bonds of Si–CH3, Si–H and C–H. This allowed Si–Si and Si–C bonds to be formed. A number of different types of radi￾ation were investigated by both fibre producers working in collaboration with the Japanese Atomic Energy Re￾search Institute. Gamma irradiation was investigated but finally electron radiation in a helium atmosphere was used to make the second generation fibres. The cross-linking step was followed by heat treatment at 327◦C for a short time to eliminate the remaining free radicals which were trapped in the irradiated precursor fibre [35]. An example of how the hydrogen atom, bonded to the silicon atom, as shown in Fig. 2, is removed by electron bombardment, so as to allow direct bonding between the two silicon atoms in neighbouring molecules, is shown in Fig. 6 [36]. Below 550◦C cross-linking between main chains dom￾inates and is induced by the dehydrogenation condensa￾tion of the Si–H groups. From 550◦C to 800◦C, the side chains on the cross-linked polymer begin to decompose and CH4 and H2 are given off producing an inorganic fi- bre. Above 800◦C and up to 1000◦C, hydrogen is given off, most probably associated with the decomposition of Figure 6 Direct cross-linking of the PCS precursor polymer by irradiation curing. 829

40TH ANNIVERSARY C-H bonds remaining in the PCs Above 1000oC, B-Sic 280 Hi-Nicalon grains, free carbon aggregates and a poorly organised in tergranular phase comprising of silicon and carbon atoms are formed [37, 38]. The presence of the free carbon ag- 220 gregates is considered to hinder grain growth so allowing 200 the Sic grains to remain small during fibre production 180 which takes place at around 1300C as indicated in Table 160 NLM202 The radiation cross-linking process, although costly, was successful in reducing the oxygen content of fibres 100 produced from PCS precursors to 0.5 wt. and gave rise 15 to the Hi-Nicalon fibre produced by Nippon Carbon [39, Failure Stress(GPa)at Lo= 25mm 40] Figure& The second generation fibres showed increased Youngs moduli When the radiation process was applied to the Ptc compared to the first generation fibres[42] precursor fibres, it was found that, although the oxygen content fell, it was only reduced to around 5 wt % The explanation was that the Ti addition to the PCs, so as to to graft the zirconium onto the PCS polymer contained make the PTC polymer, was achieved by adding a titanium less oxygen than was the case with titanium alkoxide alkoxide, Ti(OR)4, which added oxygen by another route [41]. Two types of fibre were produced. The Tyranno ZM than that of oxygen cross-linking. This alkoxide grouping fibre was cured by oxidation and was commercially pro formed links between the PCS molecules through Ti-o duced and an experimental Tyranno ZE fibres cured by bonds which were stable up to 700C. Above this temper- electron irradiation, which was not commercialised. The ature Ti-C began to be formed. Intermediate species such former fibre was produced with an oxygen concentration as TiC,O, were produced with the formation of SiC Oy. of 10 wt. and also 7.8 wt.%. The stability of the inter The result was that, although cross-linking occurred by granular phase which contained zirconium was said to be the process described above for the PCS fibres, there re- improved, compared to the titanium containing intergran- mained a significant presence of oxygen in the Tyranno ular phase as weight loss began at 1600 C for the Tyranno LOX-E fibre which Ube Industries produced by this ZM fibre which was 200oC higher than for the Tyrannd method. Fig. 7 shows the atomic carbon to silicon ratios LOX-M. The Tyranno ZM fibre found a market, for some and oxygen contents for the first and second generation time, in filters for engines running on sulphur rich diesel fibres Ube Industries took the decision not to commercially produce the Tyranno LOX-E fibre as the irradiation pro- 5. Mechanical behaviour of second generation cess was expensive and the improvement in properties Sic fibres not sufficiently adequate. The company produced other The reduction of the oxygen content in the fibres, made fibres with the intention of reducing the oxygen content by irradiating PCS, produced Hi-Nicalon fibres which by changing the metal added to the PCS polymer and the were approximately 35% stiffer than the first generation titanium was replaced by zirconium. This substitution re- fibres, as can be seen from Fig 8. However the increase duced the oxygen concentration as the compound used was not nearly as significant for the Tyranno LOX-E fibre which, unlike the Hi-Nicalon, was not developed as commercial second generation fibre. There was no change in appearance of the fracture surfaces between the first and econd generation fibres which remained devoid of any obvious signs of an ordered microstructure, as seen in the scanning electron microscope Strength retention at high temperatures of the Hi- Nicalon fibre was improved, when compared to the first generation fibres [42]. These second generation fibres re LOx-M Oxidation curing mained linearly elastic up to 1350C which was 100C higher than the first generation fibres and room strength NLM was retained up to around 1200oC in both argon and air, although the Tyranno LOX-E fibres, heated in air, showed 0 2 4 6 8 10 12 14 a fall in strength at temperatures lower than 1000oC. This o % wt was not observed in argon, suggesting a sensitivity to car- Figure7 The carbon/silicon atomic ratios and oxygen contents in the first bon oxidation at the surface of the fibres. Such sensitivity and second generation SiC fibres [42]. could be seen in the Hi-Nicalon fibres but at higher tem- 830

40TH ANNIVERSARY C–H bonds remaining in the PCS. Above 1000◦C, β-SiC grains, free carbon aggregates and a poorly organised in￾tergranular phase comprising of silicon and carbon atoms are formed [37, 38]. The presence of the free carbon ag￾gregates is considered to hinder grain growth so allowing the SiC grains to remain small during fibre production which takes place at around 1300◦C as indicated in Table III. The radiation cross-linking process, although costly, was successful in reducing the oxygen content of fibres produced from PCS precursors to 0.5 wt.% and gave rise to the Hi-Nicalon fibre produced by Nippon Carbon [39, 40]. When the radiation process was applied to the PTC precursor fibres, it was found that, although the oxygen content fell, it was only reduced to around 5 wt.%. The explanation was that the Ti addition to the PCS, so as to make the PTC polymer, was achieved by adding a titanium alkoxide, Ti(OR)4, which added oxygen by another route than that of oxygen cross-linking. This alkoxide grouping formed links between the PCS molecules through Ti–O bonds which were stable up to 700◦C. Above this temper￾ature Ti–C began to be formed. Intermediate species such as TiCxOy were produced with the formation of SiCxOy. The result was that, although cross-linking occurred by the process described above for the PCS fibres, there re￾mained a significant presence of oxygen in the Tyranno LOX-E fibre which Ube Industries produced by this method. Fig. 7 shows the atomic carbon to silicon ratios and oxygen contents for the first and second generation fibres. Ube Industries took the decision not to commercially produce the Tyranno LOX-E fibre as the irradiation pro￾cess was expensive and the improvement in properties not sufficiently adequate. The company produced other fibres with the intention of reducing the oxygen content by changing the metal added to the PCS polymer and the titanium was replaced by zirconium. This substitution re￾duced the oxygen concentration as the compound used Figure 7 The carbon/silicon atomic ratios and oxygen contents in the first and second generation SiC fibres [42]. Figure 8 The second generation fibres showed increased Young’s moduli compared to the first generation fibres [42]. to graft the zirconium onto the PCS polymer contained less oxygen than was the case with titanium alkoxide [41]. Two types of fibre were produced. The Tyranno ZM fibre was cured by oxidation and was commercially pro￾duced and an experimental Tyranno ZE fibres cured by electron irradiation, which was not commercialised. The former fibre was produced with an oxygen concentration of 10 wt.% and also 7.8 wt.%. The stability of the inter￾granular phase which contained zirconium was said to be improved, compared to the titanium containing intergran￾ular phase, as weight loss began at 1600◦C for the Tyranno ZM fibre which was 200◦C higher than for the Tyranno LOX-M. The Tyranno ZM fibre found a market, for some time, in filters for engines running on sulphur rich diesel. 5. Mechanical behaviour of second generation SiC fibres The reduction of the oxygen content in the fibres, made by irradiating PCS, produced Hi-Nicalon fibres which were approximately 35% stiffer than the first generation fibres, as can be seen from Fig. 8. However the increase was not nearly as significant for the Tyranno LOX-E fibre which, unlike the Hi-Nicalon, was not developed as a commercial second generation fibre. There was no change in appearance of the fracture surfaces between the first and second generation fibres which remained devoid of any obvious signs of an ordered microstructure, as seen in the scanning electron microscope. Strength retention at high temperatures of the Hi￾Nicalon fibre was improved, when compared to the first generation fibres [42]. These second generation fibres re￾mained linearly elastic up to 1350◦C which was 100◦C higher than the first generation fibres and room strength was retained up to around 1200◦C in both argon and air, although the Tyranno LOX-E fibres, heated in air, showed a fall in strength at temperatures lower than 1000◦C. This was not observed in argon, suggesting a sensitivity to car￾bon oxidation at the surface of the fibres. Such sensitivity could be seen in the Hi-Nicalon fibres but at higher tem- 830

40TH ANNIVERSARY peratures, as is revealed by Fig. 9 which shows such a 1.00 E-06 fibre broken in tension at 1400, which is 100C higher than the temperature used for fibre manufacture. Growth of a silica layer was observed on the surfaces of the fibres when heated in air at 1200C and pores at the silica and SiC fibre interface were observed at 1450C induced by the decomposition products of the fibre 100E-07 The Hi-Nicalon fibre was found not to creep below 1000C and to possess the highest creep threshold and ◆ ryran the lowest strain rate of the first and second generation ◆ Tyranno LOX-E口 fibres. The Tyranno LOX-E fibre was found to creep from o Nicalon 200 950C. The strain rates of the fibres, under an applied ●Hi- Nicalon stress of 1 GPa at 1050%C. were 10-10 s-l for the hi Nicalon fibre, 6x 10-s- for the Tyranno LOX-E fibre, 1.00E 2 s-l for the Nicalon 200 fibre and 2 x10-10s for the Tyranno LOX-M fibre, although it should be noted Figure 10 Creep rates at 1400oC for the first and second generation fibres that the latter fibre showed much higher creep rates at [42 slightly higher stresses. The creep rates of all the first and second generation fibres were very similar at 1400 C, Tyranno ZE fibre when compared to the Hi-Nicalon fibre at around 5 x 10s under a stress of 0.3 GPa, as [43] can be seen from Fig. 10. The activation energy for the The mechanical properties of the second generation SiC could be explained by grain boundary sliding accommo- dated by interface-controlled diffusion mechanisms. The carbon layers, between the grains, could facilitate thi 5.1. Compositions and microstructure sliding. Ultimate failure in creep of the fibres was found of second generation fibres not to be due to a lack of accommodation, as is the case The removal of oxygen from the cross-linking process for bulk ceramics, but related to surface defects, such used to make the Hi-Nicalon fibres, made from electron as cavities or porous zones growing from local chemical irradiated PCS precursors, resulted in an increase in size heterogeneities, as shown in Fig. 9 of the B-Sic grains which were seen, in the as received The Tyranno ZM and Tyranno ZE fibres showed im- fibre, to be in the range of 5 to 10 nm, as shown in the proved strength retention at high temperatures, com- TEM micrograph shown in Fig. 11[44, 45].This accounts pared to their titanium containing homologues. Sim- for the slightly higher thermal conductivity of these fibres lar strength retention at 1400C was seen by the compared to the first generation fibres, as can be seen I um gure 9 Hi-Nicalon fibre broken at 1400C showing a large pore perhaps produced by oxidation of carbon

40TH ANNIVERSARY peratures, as is revealed by Fig. 9 which shows such a fibre broken in tension at 1400◦C, which is 100◦C higher than the temperature used for fibre manufacture. Growth of a silica layer was observed on the surfaces of the fibres when heated in air at 1200◦C and pores at the silica and SiC fibre interface were observed at 1450◦C induced by the decomposition products of the fibre. The Hi-Nicalon fibre was found not to creep below 1000◦C and to possess the highest creep threshold and the lowest strain rate of the first and second generation fibres. The Tyranno LOX-E fibre was found to creep from 950◦C. The strain rates of the fibres, under an applied stress of 1 GPa at 1050◦C, were 10−10 s−1 for the Hi￾Nicalon fibre, 6 × 10−8 s−1 for the Tyranno LOX-E fibre, 2 × 10−8 s−1 for the Nicalon 200 fibre and 2 × 10−10 s−1 for the Tyranno LOX-M fibre, although it should be noted that the latter fibre showed much higher creep rates at slightly higher stresses. The creep rates of all the first and second generation fibres were very similar at 1400◦C, at around 5 × 10−7 s−1 under a stress of 0.3 GPa, as can be seen from Fig. 10. The activation energy for the creep of Hi-Nicalon fibres was around 360 kJ/mol and could be explained by grain boundary sliding accommo￾dated by interface-controlled diffusion mechanisms. The carbon layers, between the grains, could facilitate this sliding. Ultimate failure in creep of the fibres was found not to be due to a lack of accommodation, as is the case for bulk ceramics, but related to surface defects, such as cavities or porous zones growing from local chemical heterogeneities, as shown in Fig. 9. The Tyranno ZM and Tyranno ZE fibres showed im￾proved strength retention at high temperatures, com￾pared to their titanium containing homologues. Sim￾ilar strength retention at 1400◦C was seen by the Figure 10 Creep rates at 1400◦C for the first and second generation fibres [42]. Tyranno ZE fibre when compared to the Hi-Nicalon fibre [43]. The mechanical properties of the second generation SiC based fibres are given in Table IV. 5.1. Compositions and microstructures of second generation fibres The removal of oxygen from the cross-linking process used to make the Hi-Nicalon fibres, made from electron irradiated PCS precursors, resulted in an increase in size of the β-SiC grains which were seen, in the as received fibre, to be in the range of 5 to 10 nm, as shown in the TEM micrograph shown in Fig. 11 [44, 45]. This accounts for the slightly higher thermal conductivity of these fibres compared to the first generation fibres, as can be seen in Figure 9 Hi-Nicalon fibre broken at 1400◦C showing a large pore perhaps produced by oxidation of carbon. 831

40TH ANNIVERSARY Figure 11 High Resolution Transmission Electron Microscopy image of a Hi-Nicalon fibre, showing B-SiC grains of about 10 nm, surrounded by poorly a organised Si-C phase, and turbostratic carbon[42] Table IV. The microstructure can be seen to have consisted Ti compounds were found indicating that the grains must of areas of well ordered SiC surrounded by Si and C atoms be separated by a Si-C-Ti-O phase which were not completely crystallized into B-SiC grains. It became clear, therefore, that the oxygen content The Hi-Nicalon fibre contained a higher ratio of carbon was the key factor in controlling the microstructures atoms to silicon atoms than in the first generation fibres and ultimately the mechanical properties of the polymer as the absence of oxygen did not allow the excess carbon derived SiC fibres. The oxygen cured, first generation to be oxidized. The free carbon can be seen in Fig. Il and fibres, the Nicalon 200 and Tyranno LOX-M fibres, was composed of 4 to 10 distorted layers with some aggre- possessed similar grain sizes, of about 2 nm, and gates measuring up to 5 nm. The small amount of oxygen properties. However the Hi-Nicalon and Tyranno LOX-E boundaries. The Hi-Nicalon fibre would be composed by different behaviours. The reduction in oxygen content weight of 85% SiC, 11%C and 4% of SiCo86O.29 if the to 5 wt. in the Tyranno LOX-E fibre did not produc oxicarbide phase is taken to be composed of SiO Cl significant change when compared to the first generation with three carbon atoms to every one oxygen atom. It fibres however the Hi-Nicalon with only 0.5 wt %o can be was calculated that the non-crystallized Sic in the Hi- seen to be a very different fibre Nicalon fibre represented around 26% by weight of the Although bulk Sic can be used to 1600oC, the oxygen fibre [46]. The absence of a significant amount of oxygen rich SiC fibres became too brittle to be handled when removed the amorphous phase present in the first gener- heated above 1500C and were reduced to a powder ation fibres and allowed larger Sic grains to form at the at 1800oC by the degradation of the silicon oxicarbide higher manufacturing temperature used. Their size how- intergranular phase [47]. Thermodynamic analyses have ever was limited by the excess free carbon in the fibre. This shown that the solid solution SiCa OB, coexisting with microstructure explains the observed increase in Youngs SiOz and C, were unstable above 1150oC in an inert modulus of these fibres when compared to first generation atmosphere [48]. Its decomposition produced CO and fibres be seen from Table iv Sio and free carbon. The Sio gas reacted with the free The microstructure of the Tyranno LOX-E fibres was carbon to give SiC and CO [23]. The absence of an very similar to that of the first generation Tyranno LOX- oxicarbide intergranular phase in the Hi-Nicalon fibre M fibres with SiC grains having a mean size of approx clearly helped its high temperature stability and the larger imately 2 nm, although some larger grains up to 5 nm percentage of crystalline Sic accounted for its higher were seen. The free carbon aggregates were around 1- Youngs modulus when compared to the first generation 3 nm and, as with the first generation Tyranno fibres, no fibres and the still oxygen rich Tyranno LOX-E fibre

40TH ANNIVERSARY Figure 11 High Resolution Transmission Electron Microscopy image of a Hi-Nicalon fibre, showing β-SiC grains of about 10 nm, surrounded by poorly a organised Si–C phase, and turbostratic carbon [42]. Table IV. The microstructure can be seen to have consisted of areas of well ordered SiC surrounded by Si and C atoms which were not completely crystallized into β-SiC grains. The Hi-Nicalon fibre contained a higher ratio of carbon atoms to silicon atoms than in the first generation fibres as the absence of oxygen did not allow the excess carbon to be oxidized. The free carbon can be seen in Fig. 11 and was composed of 4 to 10 distorted layers with some aggre￾gates measuring up to 5 nm. The small amount of oxygen in these fibres was presumed to be present at the SiC/SiC boundaries. The Hi-Nicalon fibre would be composed by weight of 85% SiC, 11% C and 4% of SiC0.86O0.29 if the oxicarbide phase is taken to be composed of SiOxC1−x/2 with three carbon atoms to every one oxygen atom. It was calculated that the non-crystallized SiC in the Hi￾Nicalon fibre represented around 26% by weight of the fibre [46]. The absence of a significant amount of oxygen removed the amorphous phase present in the first gener￾ation fibres and allowed larger SiC grains to form at the higher manufacturing temperature used. Their size how￾ever was limited by the excess free carbon in the fibre. This microstructure explains the observed increase in Young’s modulus of these fibres when compared to first generation fibres, as can be seen from Table IV. The microstructure of the Tyranno LOX-E fibres was very similar to that of the first generation Tyranno LOX￾M fibres with SiC grains having a mean size of approx￾imately 2 nm, although some larger grains up to 5 nm were seen. The free carbon aggregates were around 1– 3 nm and, as with the first generation Tyranno fibres, no Ti compounds were found indicating that the grains must be separated by a Si–C–Ti–O phase. It became clear, therefore, that the oxygen content was the key factor in controlling the microstructures and ultimately the mechanical properties of the polymer derived SiC fibres. The oxygen cured, first generation fibres, the Nicalon 200 and Tyranno LOX-M fibres, possessed similar grain sizes, of about 2 nm, and properties. However the Hi-Nicalon and Tyranno LOX-E fibres had distinctly different microstructures and showed different behaviours. The reduction in oxygen content to 5 wt.% in the Tyranno LOX-E fibre did not produce a significant change when compared to the first generation fibres however the Hi-Nicalon with only 0.5 wt.% can be seen to be a very different fibre. Although bulk SiC can be used to 1600◦C, the oxygen rich SiC fibres became too brittle to be handled when heated above 1500◦C and were reduced to a powder at 1800◦C by the degradation of the silicon oxicarbide intergranular phase [47]. Thermodynamic analyses have shown that the solid solution SiCαOβ, coexisting with SiO2 and C, were unstable above 1150◦C in an inert atmosphere [48]. Its decomposition produced CO and SiO and free carbon. The SiO gas reacted with the free carbon to give SiC and CO [23]. The absence of an oxicarbide intergranular phase in the Hi-Nicalon fibre clearly helped its high temperature stability and the larger percentage of crystalline SiC accounted for its higher Young’s modulus when compared to the first generation fibres and the still oxygen rich Tyranno LOX-E fibre. 832