Processing and characterization of large diameter ceramic SiCN monofilaments from commercial oligosilazanes

This work reports the processing of large diameter ceramic SiCN monofilaments via the precursor route using two chemically different polysilazanes ML33S and HTTS self-synthesized from respective commercially available oligosilazanes. The melt-spinning of continuous polymer fibers with controllable diameters from 35 to 150 mm and their pyrolysis to ceramic SiCN fibers is not influenced by differences in the chemical structure of the polysilazanes. In contrast, the necessary e-beam curing dose is reduced by Si-vinyl groups from 600 kGy for ML33S (vinyl free) to 200 kGy for HTTS derived polymer fibers. The curing step leads to an enhanced handleability important for further pyrolysis at 1100 C in nitrogen and to an increase in ceramic yield. The resulting ceramic SiCN fibers from both systems have similar mechanical and thermal behavior, indicating quite a low influence of the polysilazane type on these properties. For the first time a comprehensive investigation of the effect of fiber diameter on the tensile strength is reported for SiCN fibers. The average strength increases from 800 MPa for 90 mm diameter fibers to 1600 MPa for the 30 mm diameter fibers. Bend Stress Relaxation (BSR) tests demonstrated that no stress relaxation occurs up to 1000 C for SiCN monofilaments and the creep resistance is equal to or better than commercially available SiC monofilaments produced by chemical vapour deposition (CVD). The oxidation resistance is also comparable to commercially available oxygen free CVD SiC fibers (SCS-6). The ceramic fibers in this study were pyrolyzed at low temperature (1100 C) and have high oxygen content (13 to 29 wt%). The high temperature creep resistance and oxidation resistance is expected to improve if the oxygen content is reduced and the pyrolysis temperature increased.


Introduction
Lightweight thermostructural materials are needed in next generation aero turbine engine components to improve the energy efficiency due to weight reduction and increase of the operating temperature.They are also needed for other advanced applications including heat exchangers, rocket propulsion, fusion reactors and UHTC parts for supersonic and hypersonic vehicles.In these applications, ceramic bers play an important role as reinforcements of metallic and ceramic matrices to realize the manufacturing of composite materials with low density and for high temperature applications.
5][6] These high performance bers comprise not only superior mechanical properties, such as tensile strength > 2 GPa and stiffness, but also thermal stability at temperatures higher than 800 C, which are essential for the reinforcement of materials based on metallic and ceramic matrix composites (MMCs and CMCs respectively).
For MMCs applications, such as combustion engines and high efficiency turbines, titanium matrix composites (TMCs) reinforced with CVD SiC monolaments are promising candidates, as they combine the high strength, stiffness and creep resistance of SiC monolaments with the high ductility, toughness and corrosion resistance of titanium alloys. 7,8ecause SiC bers are generally reactive with titanium alloys at high temperatures, fabrication processes that allow a controlled chemical reaction with the metal matrix, mainly dominated by diffusion, enables the formation of a strong ber-matrix interface and good adhesion.This good adhesion is needed for the efficient load transfer from the metal matrix to the SiC bers.The major obstacle for the application of TMCs in the industry is the high material cost of monolament SiC bers.
They are among the most expensive ceramic bers with prices up to $8000 V per kg (11 000 U$ per kg). 2,8,9n contrast, the continuous processing of ne diameter bers from inorganic polymers is still the most attractive method for the manufacturing of cost-effective non-oxide ceramic bers.Ceramic bers from SiCN systems, such as polysilazanes, comprise good oxidation stability up to $1500 C (due to the nitrogen content) and corrosion resistance in acids and bases.Beside this they offer a remarkable cost reduction, due to relatively inexpensive precursors, use of low e-beam doses for their curing and relatively simple processing. 2,10,11n this paper we report on the processing and properties of large diameter SiCN monolaments made from preceramic polymers as an alternative to CVD SiC monolaments.Two different types of meltable polysilazanes with tailored properties were recently developed in our chair. 12,13These polymers open new opportunities for cost effective processing of ceramic SiCN bers with larger diameters as demonstrated in this investigation.
Therefore, the aim of this work was to melt spin tailored polysilazanes bers of large diameter, to investigate the inuence of the electron beam irradiation dose on the curing of melt spun green bers and to study the pyrolysis of these green polymeric bers to ceramic bers.Finally, the mechanical and thermal properties, including strength, strength distribution, creep resistance and oxidation behavior, of these bers have been investigated and compared to commercially available SiC bers.

Experimental procedure
Both polymers, ML33S and HTTS, used for the processing of ceramic SiCN bers are produced by a selective chemical crosslinking of two commercially available liquid oligosilazanes ML33 and HTT1800, respectively, which were purchased from Clariant Advanced Materials GmbH (Sulzbach, Germany) (Fig. 1).
Both tailored polymers are meltable solids, having suitable viscoelasticity and thermal stability up to 170 C for the meltspinning of bers.More details about the synthesis and characterization of ML33S and HTTS, including their rheology and thermal stability, have been reported elsewhere. 12-14

Processing of ceramic SiCN monolaments
Monolaments from ML33S or HTTS polysilazanes were meltspun in a lab-scale gas pressure melt spinning equipment.About 8 g of the polymer were lled in the vessel of the spin equipment, melted at temperatures between 110 and 130 C, evacuated to remove gas bubbles and le at the desired temperature for 15 minutes to ensure a constant and uniform temperature.Aerwards, nitrogen over pressure was applied to allow the melt-spinning of laments through a spinneret followed by winding of the bers.Monolaments with diameter > 70 mm were melt spun through a spinneret with a single capillary having a diameter of 800 mm.To investigate the dependence of the strength of the ceramic bers on the diameter, laments with diameter < 70 mm were also melt spun by using a spinneret with 7 holes, each one with a diameter of 400 mm.
For the investigation of the electron beam irradiation dose needed to render the precursor unmeltable, cylindrical samples having diameter of 20 mm and thickness of 10 mm were prepared from ML33S or HTTS polysilazanes by melting the polymer in nitrogen atmosphere in a silicone mold.The samples were packed under nitrogen atmosphere in a polyethylene (PE) plastic sheet and sent to the company Beta-Gamma-Service GmbH, Saal, Germany.An acceleration voltage of 10 MeV and doses up to 1000 kGy were applied for the electron beam curing experiments of the samples.
The gel fraction of the precursors aer irradiation was determined based on the German standard DIN 16892. 15About 0.25 g of the cured sample powder was solved in toluene and reuxed for 4 hours at 80 C. Aer ltration through a regenerated cellulose lter with a pore size of 0.2 mm, the undissolved fraction remaining on the lter was dried and weighed for the calculation of the gel fraction according to the following equation: where G is the gel fraction in %, m 1 the weight of the cured sample powder before dissolving in toluene and m 2 the weight of the dried undissolved part aer ltration.In order to complement the gel fraction analysis, e-beam cured cylindrical samples were pyrolysed in nitrogen atmosphere up to 1000 C with a heating rate of 10 K min À1 to amorphous SiCN ceramics and their meltability was analysed.
For the investigation of the transition from the polymer to amorphous ceramic SiCN material thermal gravimetric analysis (TGA) was performed, using a Linseis L81 A1550 unit (Linseis, Germany).Approximately 10 mg of the sample was heated from 25 to 1300 C with a heating rate of 5 K min À1 in nitrogen atmosphere.The ceramic yield was calculated from the mass change of the material aer the thermal treatment.
Based on the curing results obtained from the cylindrical samples, the collected monolaments were cured at BGS GmbH, Saal, with an acceleration voltage of 10 MeV and doses up to 600 kGy.Aer curing, the bers were converted into amorphous ceramic SiCN bers by continuous pyrolysis in a tubular furnace (model GERO Gero FA100-500/13) equipped with a bobin at each end of the furnace tube.The pyrolysis was conducted at a maximum temperature of 1100 C in nitrogen atmosphere.Cured green bers were continuously pulled through the furnace using a ber winding unit with a velocity of 0.3 cm min À1 .The elemental composition of the uncured, cured and of the ceramic SiCN bers was analysed at Pascher Microanalytical Laboratory (Remagen, Germany).

Properties of ceramic SiCN monolaments
Tensile tests were carried out for both the cured green and ceramic SiCN monolaments at ambient conditions using a universal mechanical testing machine (Inspekt Mini, Hegewald & Peschke GmbH, Germany).From each batch 40 samples were selected and their tensile strength was determined.According to DIN-ENV 1007-4 individual laments were glued onto a split cardboard frame with a gauge length of 25 mm. 16A 20 N load cell was used with a cross-head speed of 4 mm min À1 .Fiber diameters were determined prior to the mechanical test with an optical microscope (Zeiss, Axiotech 100) equipped with a camera for digital analysis of the picture.Aer the tensile test, laments that broke in the extremities of the split cardboard frame were excluded from the analysis.
In order to measure the creep resistance of the amorphous ceramic SiCN bers, the Bend Stress Relaxation (BSR) method developed by Morscher et al. 17 was used for single mono-laments with diameter > 70 mm.Fibers were tied into a loop with a radius R 0 of 10 mm at room temperature and heated up in air or nitrogen atmosphere for 1 hour to temperatures ranging from 900 to 1200 C. Aer the thermal treatment of the sample, the ber loop was cut at room temperature and the resulting radius of curvature R a measured.To have a quantitative measurement of the creep resistance of the bers, the BSR ratio (m BSR (T, t) ¼ 1 À R 0 /R a ) was calculated.For the evaluation of m BSR , 4 samples of each ceramic SiCN ber charge were used for each test condition.The mean value of the m BSR of these samples was used for the discussion.
The oxidation resistance of the ceramic SiCN bers was investigated using the TGA equipment (Linseis, L81/1550), and compared with commercial ceramic SiC bers, type SCS 6. Monolaments were cut in samples of 2 cm in length and placed in an alumina crucible.The samples were heated up to a maximum temperature ranging from 900 to 1300 C, heating rate of 10 K min À1 and a dwell time up to 12 hours under owing oxygen (5 L min À1 ).Aer oxidation, the morphology and thickness of the oxide layer on the ber was analysed by scanning electron microscopy (SEM, Zeiss, 1540EsB) coupled with an Energy Dispersive X-ray Spectrometer (EDS, Thermo Scien-tic, NORAN System 6).

Processing of polymer derived ceramic SiCN bers
Based on the molecular, thermal and rheological analysis of the polysilazanes manufactured by controlled cross-linking, 12,13 continuous green bers were processed in air, using the labscale melt spinning equipment.Aer the green bers were spun in air and le for 3 hours on the spool in contact with air, the oxygen content of the bers remains <0.5 wt%.The oxygen incorporation is due to the reaction between the surface of the ber with the moisture in the environment 18 as described by Hacker, 11 who used the self-synthesized ABSE polysilazane for processing of ceramic SiCN bers.Fig. 2 shows a large diameter melt spun monolament wound on a spool.
For the processing of green bers, based on prior knowledge in our laboratory, a viscosity of 5 Â 10 3 Pa s was chosen to ensure a continuous viscoelastic owability of the polymer.Based on rheological measurements, this required a processing temperature of about 50 K higher than the glass temperatures of the polymers. 12,13In addition to the viscosity and the spinneret hole diameter, the nal diameter of the green bers is controlled by the pull-off speed.Polymer monolaments having diameters between $35 and 150 mm were successfully processed by adjusting the temperature (and hence the viscosity of the melt) between 110 and 120 C and by varying the pull-off speed between 10 and 30 m min À1 .
The e-beam irradiation is not only crucial to turn the green bers into infusible solids, but also improves their exibility and handleability for further pyrolysis into ceramic bers.
Commercially available SiC bers require high electron beam doses, between 7500 and 20 000 kGy. 19 The irradiation dose to cross-link Hi-Nicalon bers from polycarbosilane (PCS) can be reduced by the addition of vinyl sources, such as polyvinylsilane. 20,21But the work about the inuence of the irradiation dose on the cross-linking behavior of polyorganosilazanes is rather limited. 22,23herefore the present investigations to determine the minimum e-beam dose are based on scientic works regarding the development of ceramic SiCN bers from ABSE polysilazane. 22In order to quantify the curing results, the gel content of the polymer ML33S and HTTS bers aer irradiation was measured (using eqn (01)).
Fig. 3 demonstrates that the curing behavior of the polysilazanes strongly depends on their functional groups.Due to the absence of Si-vinyl groups in ML33S, the required dose for its curing is higher than that for HTTS.Hence, an e-beam dose $600 kGy is required to reach a gel content of about 60% and to render the ML33S green bers infusible.Melting tests performed with the polymers aer curing conrm that ML33S samples treated with e-beam doses greater than 600 kGy maintain their shape (remain infusible).
HTTS already reaches a gel content of nearly 100% aer treatment with a low dose of only 200 kGy due to the highly reactive Si-vinyl groups.In contrast to the curing of ML33S, all HTTS samples showed a yellowish appearance, which is probably a sign of the formation of reactive free radicals or conjugated bondings.The melting test conrmed that all cured HTTS samples were already infusible even aer treating with a very low dose of 200 kGy, which is enough to obtain infusible HTTS green bers.
With the additional cross-linking of the precursors by using e-beam irradiation, their ceramic yield slightly increases due to the further reduction of volatile oligomers.In Fig. 4a the TGA curve indicates that a dose of only 200 kGy to cure ML33S has small inuence on the reduction of released volatile oligomers, occurring between 200 and 400 C. However with doses $400 kGy this reduction is more pronounced and the ceramic yield for cured ML33S materials increases to up to $78 wt% in comparison to a ceramic yield of 70 wt% for the uncured ML33S.However the increase in the ceramic yield for the uncured (78 wt%) to the e-beam cured HTTS samples ($82 wt%) is not signicant (Fig. 4b).
These results clearly illustrate that the Si-vinyl groups reduce the necessary electron beam dose for the curing of HTTS derived green bers signicantly in comparison to the less reactive ML33S derived green bers.Furthermore, while both types of melt spun uncured green bers are brittle, especially the cured green bers from HTTS appear more exible and stronger, which is important for the subsequent pyrolysis process.Aer exposure to an e-beam dose of 600 kGy, cured green bers from HTTS have a mean tensile strength of 65 MPa, while cured ML33S derived green bers reached a mean strength of only 43 MPa.
Concerning the continuous pyrolysis of the cured bers into ceramic bers, a minimum diameter for the take-up spool used to pull the ceramic bers leaving the furnace was chosen to avoid that the bers with d > 70 mm break on the spool.For this purpose, prior to the continuous pyrolysis, about 1 meter of the cured monolaments with diameter $ 100 mm were cut in samples with 20 cm length, deposited on a graphite boat and pyrolysed in a furnace at 1000 C in nitrogen atmosphere for 1 hour.The pyrolysed monolaments were bent until they broke and the mean value of the bending diameter prior to the breaking was taken as the minimum spool diameter used to pull the SiCN bers.
Aer the continuous pyrolysis in a horizontal tubular furnace in nitrogen atmosphere the resulting ceramic SiCN bers leave the furnace passing through the guide pulley before they are nally pulled on a spool with a diameter of 160 mm by a motor operated coiling unit (Fig. 5a and b).Fig. 5b shows a bundle of 3 pyrolysed ceramic monolaments with a length of more than 10 meters on the spool.
For all investigated ber batches continuous pyrolysis lead to amorphous ceramic SiCN bers with regular and smooth surface.SEM micrograph of HTTS derived ceramic bers conrms diameters up to 100 mm, with a dense core and without noticeable defects or pores (Fig. 6).Due to the similarity, micrographs of ML33S derived ceramic bers are not shown.
The oxygen content of the processed amorphous ceramic SiCN bers is similar or higher than for Nicalon bers (Table 1).Although the oxygen content of the uncured polymers is only <0.5 wt%, aer their melt-spinning in air and curing with electron beam irradiation in inert atmosphere it increases to 3.5 and 4.7 wt% for ML33S and HTTS derived green bers, respectively.Aer pyrolysis in nitrogen atmosphere the oxygen content of ML33S derived ceramic bers reaches $13 wt% while HTTS derived ceramic bers contain $29 wt% oxygen.This high increase in oxygen content is due to the handling and long exposure of the cured bers in air before they are pulled through the furnace at a very low velocity of 0.3 cm min À1 , as described in Chapter 2.1.
During exposure in air, the nitrogen of the silazanes is replaced by oxygen. 18This fact becomes evident especially for HTTS derived bers as a consequence of the higher content of moisture sensitive Si-H groups.Beside this, vinyl groups are present in this precursor, which are activated during e-beam curing, distinguishable on the change in color from colorless to yellowish due to the formation of reactive radicals.Clearly, this enhanced reactivity also leads to an additional formation of Si-O bondings. 21,22More details about the inuence of oxygen content in the creep and oxidation stability of the processed SiCN bers will be later on discussed in Chapters 3.2.2 and 3.2.3.

Mechanical and thermal properties of ceramic SiCN bers
3.2.1.Room temperature tensile strength and strength distribution.We measured the tensile strength of the pyrolyzed bers which fail in a brittle manner.The strength of brittle materials is controlled by the size, shape and orientation (relative to stress) of the largest aw.For brittle bers, ber diameter, ber surface and the microstructure of the material have been shown to be the main factors that control the tensile strength.Therefore, the general observation that the measured strength of brittle bers increases as the diameter decreases, according to earlier works from Griffith 24 with glass bers, was tested with the processed ML33S and HTTS derived ceramic SiCN bers with a broad range of diameters ($30 to 100 mm).Due to the fact that the ceramic SiCN bers are amorphous ceramic materials, the effect of ber diameter is expected to be similar to that for the behavior of glass bers. 24ig. 7 shows the tensile strength of ML33S and HTTS derived ceramic SiCN bers as a function of the ber diameter.In the case of ML33S derived ceramic bers, there is signicant scatter in strength for bers with a diameter between 40 and 50 mm, but the tendency to higher strength values with the reduction of the diameter is clearly illustrated.Such scatter results from the not optimized melt-spinning process.With this simple lab-scale gas pressure melt spinning equipment it is impossible to remove all gas bubbles from the polymer melt, which lead to pores within the resulting bers.However, especially for HTTS derived ceramic bers the effect of ber diameter on tensile strength is much clear due to lower scatter.For both ceramic ber types the tensile strength values are similar for the same diameter.While bers with about 100 mm have a tensile strength of $0.8 GPa, thinner bers with a diameter of 30 mm have a mean tensile strength as high as $1.5 GPa.A further reduction of the ber diameter should lead to tensile strength values higher than 2 GPa.For comparison, the tensile strength of commercially available SCS-6 bers is about 3.9 GPa. 2 This very high value is a result of the CVD process, which leads to nearly defect free SiC   bers.Although the tensile strength value of the prepared ceramic SiCN bers is still lower, the continuous processing of thick SiCN bers with diameter > 70 mm and tensile strength of $0.8 GPa is a signicant result and a crucial precondition for their use as reinforcement in metal matrix composites.
In order to better understand the mechanical behavior of these ceramic SiCN bers, Weibull distribution analysis was used to analyze the fracture strength.Considering that identical specimens of brittle materials, such as ceramic bers, show a large variation of tensile fracture stresses, the cumulative distribution function proposed by Weibull 25 was used for a statistical characterization of the tensile strength for ceramic SiCN bers made from ML33S and HTTS.The single parameter Weibull distribution leads to a failure probability given by: where m is the Weibull modulus, V the tested volume, s the failure strength and V 0 and s 0 are scale parameters.If the tested volume is constant, i.e. when the gauge length and the diameter of the bers are assumed constant, eqn (02) can be rearranged and reduced to: The Weibull analysis of brittle fracture is based on the weakest link theory, which assumes that all materials contain inhomogeneities, such as pores or aws, distributed randomly with a certain density per unit volume.If these defects are the origin of material's fracture then tensile failure is determined by the largest critical defect. 25,26n Fig. 8, the results of the Weibull analysis on the strength of the ceramic SiCN bers are presented.Although it is known that the variation in diameter inuences the volume variation when the tensile strength is measured using the same gauge length (in this work being 25 mm) the single mode Weibull analysis was used and it was assumed that there is no volume variation as the ber diameter changes.
The Weibull parameters, modulus m and scale parameter s 0 are presented in Table 2.The Weibull parameters are similar for the two types of bers.In spite of the diameter variation from 30 to 100 mm, the single mode Weibull distribution ts the strength data reasonably well.The mean strength of ML33S and HTTS derived ceramic bers is about 1 GPa.
For further analysis of the strength distribution, the strength results were grouped in two ranges -ber diameter less than 70 mm and greater than 70 mm.Weibull analysis was conducted separately for these two groups and the results are shown in Fig. 9.As can be seen, the Weibull modulus of higher diameter  bers from ML33S and HTTS achieve values of m > 8, which implies a high reproducibility of the tensile strength of thick monolaments.For thinner bers, the Weibull modulus specially in the case of ML33S derived bers is lower, being m ¼ $6, reecting the high scatter in the strength values for bers with diameters of about 50 mm.In accordance with Griffith's observations and analysis, the average strength of the thin bers was higher than that of thick bers.The Weibull parameters for the thick and the thin ceramic bers made from the two precursors are summarized in Table 3.
As the strength of the SiCN ceramic bers derived from ML33S or HTTS show similar values, it appears that the strength of the ceramic SiCN bers is independent on the tensile strength of the cured green bers.The results obtained in this work indicate that the strength of the uncured or cured green bers is closely related to the chemical properties of the polymer, particularly the molecular weight, the rheology and the functional reactive groups.But in relation to ceramic SiCN bers, the type of precursor, ML33S or HTTS, does not have signicant inuence on the strength.
A discussion focused on the dependency of the tensile strength of a polymer derived amorphous ceramic ber on the ber diameter, ber surface and microstructure is rather complex.The strength distribution will depend on the polymer system, the processing history of the green ber, their curing and pyrolysis conditions.Further investigations are needed including careful analysis of fracture origin in order to improve the strength and strength distribution of bers made from precursors.
3.2.2.Creep response.The use of ceramic SiCN bers as reinforcements for MMCs and CMCs for thermomechanical applications also requires the investigation of their creep behavior, which was characterized by measuring their Bend Stress Relaxation (BSR) in air and inert atmosphere.
The creep behavior of polymer derived ceramic SiCN bers, as well as SiC bers, is correlated to their microstructure and mechanical stability.][29][30][31][32][33] Fig. 10 demonstrates that both ber types derived from ML33S and HTTS have similar creep resistance either in air as well as in nitrogen atmosphere.No stress relaxation occurs up to 1000 C. Comparing the creep response of these bers to other SiCN bers, similar behavior is observed for ABSE polycarbosilazane derived SiCN bers reported by Hacker,11 which were pyrolysed at 1300 C containing 16 wt% of oxygen.When the test temperature for the ML33S and HTTS derived ceramic bers is increased to 1100 C the creep relaxation starts leading to decreased m BSR values.For creep test temperatures higher than 1100 C, which exceed the processing temperature of the bers, the m BSR values drops dramatically due to rearrangements within the amorphous SiCN phases.
Similar to Nicalon SiC bers, 35 ML33S and HTTS derived ceramic bers contain free carbon content up to $20 mol% (Table 4), high oxygen content (see Table 1) but they were pyrolysed at lower temperatures in comparison to Hi-Nicalon or Hi-Nicalon S bers. 37,38However, Nicalon bers already suffer stress relaxation at 1000 C with the evolution of SiO and CO gases due to its microstructure, composed of an amorphous   SiCO phase, a free carbon phase and small 2-5 nm b-SiC crystallites, compromising its creep resistance. 35,36n fact the creep resistance of SiCN bers derived from ML33S and HTTS is similar to that of Hi-Nicalon SiC bers, although they contain lower levels of oxygen and are processed at higher temperatures of $1300 C. The better creep resistance of SiCN bers in comparison to the oxygen rich Nicalon bers is due to the stability of the amorphous SiCN phase against crystallization up to 1400 C.An increase of the pyrolysis temperature for the processing of ML33S and HTTS derived bers up to $1300 C should also lead to a further increase of the creep resistance, as already demonstrated for ABSE derived ceramic SiCN bers with about 16 wt% oxygen. 11s the focus of the present work was to develop large diameter SiCN monolaments, it is instructive to compare the creep resistance of these bers with other large diameter SiC bers.Fig. 11 shows the creep resistance of large diameter (>70 mm) SiCN bers from ML33S and HTTS in comparison to commercially available CVD SiC bers.It can be noted that the m BSR modulus of the ceramic SiCN bers prepared in this work is similar to the commercially available SCS-6 bers and signicantly better than that of Sigma bers.The stress relaxation of Sigma bers starts at 900 C and the m BSR modulus is 0.8.At 1100 C the creep resistance is very low and the BSR modulus is only $0.1.
Sigma 1156 bers are composed of stoichiometric SiC grain near the tungsten core and excess silicon near the surface, which is responsible for the drop of the creep resistance at T > 900 C. 40 In contrast, the SiC in the SCS-6 bers is nearly stoichiometric and chemically stable up to high temperatures.The creep behavior of these bers in the temperature region 1100 < T < 1300 C is dominated by microstructural changes in the complex composite structure of these bers.SCS-6 bers contain concentric regions with different SiC compositions, free carbon at the grain boundaries and a carbon-rich surface coating.This composite structure induces microstructural changes at high temperatures due to diffusion of free silicon from the near-stoichiometric region with b-SiC grains into the carbon-rich surface coating, reducing its thermal stability.][41] The results in this chapter indicate that the more costeffective precursor route for the processing of large diameter ML33S and HTTS derived ceramic SiCN bers make their use as reinforcement of MMCs very attractive, despite their still limited mechanical properties.Although the maximum pyrolysis temperature of the polymer derived ceramic bers prepared in this work is only 1100 C, due to the limited maximum pyrolysis temperature of the furnace, their creep resistance is already comparable to SCS-6 bers, which are currently the most preferred ceramic monolaments to reinforce MMCs in spite of their high cost.
3.2.3.Oxidation resistance.The well known high oxidation resistance of Si, SiC and Si 3 N 4 ceramics is due to the growth of a silica lm in the passive regime, protecting the materials against continued oxidation.For all these Si-materials the permeation of O 2 through SiO 2 governs the oxidation rate.However, in Si 3 N 4 ceramics the thickness of the silica layer in the passive regime is lower than that in Si and SiC under similar thermal and atmospheric conditions.During the oxidation of Si 3 N 4 a duplex Si 2 N 2 O/SiO 2 layer is formed.The silicon oxynitride layer acts as a diffusion barrier during oxidation, leading to better oxidation resistance of Si 3 N 4 compared to SiC.For the oxidation tests performed in this work, at temperatures between 900 and 1100 C in air a thin and smooth passivating SiO 2 coating forms on the surface of the bers (CVD SCS-6, ML33S and HTTS derived ceramic bers), protecting or reducing the oxidation of the material at these temperatures.
Aer annealing of the SCS-6 bers in air at 1200 C, an outer SiO 2 layer with a thickness of 2.5 mm is formed (Fig. 12a).Cracks and pores within the oxide layer already appear aer this oxidation test.Additionally, the carbon core is completely oxidized.
In contrast, the oxide layer of the ML33S and HTTS derived bers formed at the same conditions is smooth and defect free with a thickness between 1.5 and 1.8 mm (Fig. 12b and c and 13).
If ML33S or HTTS derived ceramic bers are exposed to air at 1300 C, the outer surface of the SiO 2 layer with a thickness of $3.5 mm shows rst signs of cracks (Fig. 14b and c).For comparison, the micrographs of the SCS-6 ber also oxidized at 1300 C are shown in Fig. 14a.The thickness of the resulting oxide layer is with about 3.5 mm comparable to our bers.
The oxidation behavior of the SiCN bers most likely follows the same mechanisms as reported by Chollon and Mocaer 10,43 and later on discussed by Hacker 11 and Kokott. 34The ML33S and HTTS derived ceramic bers can be considered as SiCN(O) systems, due to their oxygen contents >10 wt%.Therefore, the oxidation process in summary follows eqn (04): (with 0 # x # x + y # 1) for the amorphous SiC x N 4y/3 O 2(1ÀxÀy) phase and, if present, for the free carbon: During the oxidation of SiCN(O) systems a complex process occurs, involving the oxidation mechanism of both SiC and Si 3 N 4 , such as simultaneous evaporation of CO and N 2 by diffusion  through the oxide layer.According to Chollon, 10 the high oxidation stability of SiCN(O) bers is due to the formation of a SiCNO interface (composition SiC (xÀdx) N 4(yÀdy)/3 O 2(1ÀxÀy+dx+dy) ) between the ber core and the outer SiO 2 layer.This interface enables a continuous variation of the concentration from initial SiCN(O) composition of the ceramic ber until SiO 2 .Decisive for the oxidation stability is the necessary activation energy to further oxidize the SiCNO interface, which in this case is considered to be a combination of Si 3 N 4 , SiC and SiO 2 .The activation energy of the reaction of oxygen with Si 3 N 4 is 330-490 kJ mol À1 , which is much higher than with SiC, being 90-140 kJ mol À1 .Therefore, the higher the content of Si-N bondings in the SiCN(O) system, the higher is the necessary activation energy for the oxidation reaction of the ber and nally their oxidation stability.
Thus the oxidation tests indicate that the oxidation resistance of ML33S and HTTS derived ceramic SiCN bers is higher or comparable to the commercially available SCS-6 bers despite their low processing temperature of only 1100 C. MMCs and CMCs are generally used to combine thermomechanical stability with oxidation resistance at high temperatures.If voids or cracks are present (or generated) in the matrix, the contact with oxygen degrades the reinforcement, depending on its oxidation resistance.Therefore, carbon bers, thus also the carbon core of SCS-6 bers (see Fig. 12a and 14a), are not suitable for high temperature applications in an oxidative atmosphere.In contrast ceramic SiCN bers are completely protected with passivating oxide layers, which allows their use also in air at higher temperatures.

Conclusions
Ceramic SiCN bers with diameters > 70 mm were successfully produced via melt-spinning of two recently developed solid polysilazanes from our chair, ML33S and HTTS, curing in inert atmosphere with electron beam irradiation doses up to 1000 kGy and continuous pyrolysis in inert atmosphere.Aer curing with an electron beam dose of 600 KGy, green bers from ML33S achieve a tensile strength of 43 MPa and a gel content of 60%.Green bers from HTTS cured with the same electron beam dose show higher strength values of 65 MPa and a gel content of 100%.This is a consequence of the more reactive Sivinyl groups in HTTS, which assures a high level of cross-linking aer electron beam treatment already with a low dose of only 200 kGy.The resulting cured green bers are stronger and more exible, which is very important for subsequent handling and pyrolysis to ceramic SiCN bers.
The tensile strength distribution of ceramic SiCN bers is related to their ber diameter.Ceramic SiCN bers with diameters > 70 mm achieve a tensile strength of 800 MPa, while ceramic SiCN bers with diameter < 70 mm yield strength values of $1.6 GPa.The Weibull modulus for the large diameter bers is rather high ($8) indicating that the strength variability is rather low.
The creep behavior of the ceramic SiCN bers is inuenced by their composition, mainly oxygen and free carbon contents, as well as by the pyrolysis temperature.No stress relaxation is observed for ML33S and HTTS derived ceramic SiCN bers up to 1000 C. Up to a temperature of 1100 C, their creep behavior is similar to that of Nicalon and SCS-6 bers.At higher temperatures the creep resistance decreases remarkably due to the low pyrolysis temperature of only 1100 C and the high oxygen contents (12.9 and 28.6 wt%).
The oxidation resistance of ceramic SiCN bers from ML33S and HTTS is similar or superior than SCS-6 bers.Aer a heat treatment in air at 1200 C for 12 hours, the surface of the ML33S and HTTS derived ceramic bers is protected by a close defect free protective silica layer, while the oxidation layer of SCS-6 bers already show cracks and pores.
For the rst time we demonstrated the continuous processing of thick ceramic SiCN bers by using the PDC technology.There is a high potential to improve the ceramic ber properties by optimizing the processing parameters, reducing the oxygen content and increasing the pyrolysis temperature.Therefore the developed technology offers a promising route to produce thick ceramic bers for e.g. the reinforcement of MMCs, especially titanium matrix composites.

Fig. 2
Fig.2(a) Melt spinning of polymer green monofilament and a spool with wound monofilaments.SEM micrographs of (b) the cross-section and (c) the surface of ML33S derived green monofilaments.

Fig. 4
Fig.4The dependence of mass change during pyrolysis on the e-beam dose of (a) ML33S and (b) HTTS polysilazanes.

Fig. 3
Fig. 3 Gel content of ML33S and HTTS polysilazanes after curing with electron beam doses up to 1000 kGy and comparison with the melting test of the respective cured samples.

Fig. 6
Fig. 6 SEM micrographs of (a) the cross section and (b) surface of amorphous ceramic SiCN fibers from HTTS.

Fig. 5
Fig. 5 (a) Ceramic SiCN fibers from HTTS with diameter of $100 mm leaving the tubular furnace and (b) continuously pulled on the spool.

Fig. 7
Fig.7Effect of fiber diameter on the tensile strength for (a) ML33S or (b) HTTS derived ceramic SiCN fibers compared to glass fibers.24

Fig. 10
Fig. 10 Bend stress relaxation modulus (m BSR ) versus temperature for polymer derived ceramic fibers (dwell time at temperature: 1 hour).

Fig. 13
Fig. 13 Energy dispersive spectroscopy (EDS) (a) oxygen map and (b) silicon map of the cross section (SEM micrograph) of a (c) ML33S derived ceramic fiber after 12 hours at 1200 C in air with (d) Si K and O K linescan.

Fig. 12
Fig. 12 SEM micrographs of the surface and cross section of ceramic fibers after 12 hours at 1200 C in air: (a) CVD SCS-6 fiber, (b) ML33S and (c) HTTS derived ceramic SiCN fibers.

Fig. 14
Fig. 14 SEM micrographs of the surface and cross section of ceramic fibers after 12 hours at 1300 C in air: (a) CVD SCS-6 fiber, (b) ML33S and (c) HTTS derived ceramic SiCN fibers.

Table 1
Elemental composition and empirical formula of synthesized polysilazanes, their respective cured green fibers and ceramic fibers

Table 3
Mean strength and Weibull parameters of ceramic SiCN fibers

Table 2
Mean strength and Weibull parameters of ceramic SiCN fibers

Table 4
Microstructural features of the ceramic SiCN monofilaments