High Temperature Behavior of Si3N4 and Yb2SiO5 Coated Carbon Fibers for Silicon-Nitride CMC


  • The financial support of the German Research Foundation DFG within the graduate School 1229 is gratefully acknowledged.


For oxide-free ceramic matrix composites (CMC), with Si3N4 matrix and carbon fiber reinforcement, for extreme high temperature applications, protective coatings of the C-fibers are investigated. Two different coatings are compared: reactive CVD-derived pure Si3N4 coatings to investigate C-fiber-matrix reactions and powder based Yb-silicate coatings to reveal potential reactions with the Yb-silicate additive serving as sintering aid for Si3N4. The reactivity toward carbon in nitrogen atmosphere is studied in the temperature interval from 20 °C up to 1700 °C. A new ceramic phase – an Yb-carbido-nitiridosilicate, Yb2Si4CN6–is found as product of carbothermal reduction of the Yb-silicate. The carbothermal reduction occurs also with other RE-silicates, RE = Yb, Er, Y, Gd, and Sm while SiC is found as reaction product on carbon fibers coated with pure Si3N4. The oxidation resistance of the coated fibers in air was investigated in the temperature interval up to 1000 °C, and the apparent activation energy of oxidation was analyzed based on DTA-EGA results. The oxidation kinetic reveals a significant increase of onset point of oxidation temperature by up to 150 K for Si3N4 coated short carbon fibers obtained from the reactive CVD coating process. Such fibers have a high application potential for carbon-fiber reinforced Si3N4-CMC. The role of Yb2Si4CN6 as reinforcement for Si3N4-CMC is discussed based on bond strength comparison of carbides (SiC), nitride silicates (SiAlON), and nitrides (Si3N4).

1 Introduction

Carbon fibers, carbon particles as well as carbon nanotubes have found wide application as reinforcing phases like in plastics, CFC, metals, MMC as well as in ceramics, ceramic matrix composites (CMC).[1-4] But the currently available carbon fiber-reinforced non-oxide CMC contain SiC as matrix and carbon fibers as reinforcement. They are commonly obtained by silicon infiltration into porous C-fiber reinforced carbon matrix materials. The major challenge of processing such SiC-CMC is to prevent C-fiber degradation due to reaction with silicon.[4] Attempts to reinforce Si3N4-matrix ceramics with carbon fibers were not successful so far, because of carbo-thermal reactions taking place between the oxide additives used to sinter Si3N4 and carbon.[5] In addition, Si3N4-matrix ceramics are already “self-reinforced” by in situ growth of elongated β-Si3N4 short fibers from solid-state reactions and liquid phase formation between Si3N4 and the oxide additives. However utilization of “in situ” grown sufficiently large β-Si3N4 crystals as reinforcement is limited to certain additive-compositions only and requires very long dwell time at high temperature upon sintering.[6, 7] In addition, the volume fraction of the reinforcement phase cannot be varied but is fixed by the additive, and sintering conditions.

In the present study, the concept of C-fiber reinforced Si3N4-ceramic, Si3N4-C-CMC is revisited based on development of suitable protective coatings for carbon fibers. The coatings have therefore to enable two improvements of carbon-fiber functionality: oxidation resistance along with low reactivity between coating material and carbon fiber. If such coatings would become available, then implementation of a deliberate amount of C-fiber reinforcement into Si3N4-matrix ceramics with any type of additive phase would become feasible. The oxidation resistance of carbon fibers can be increased by coating them with borides (HfB2), carbides (SiC), nitrides (TiN), or aluminates (MgAl2O4).[3, 8-11] In the past, some of these nitrides (TiN), carbides (TiC), or silicides (MoSi2) were also investigated as additives for Si3N4 ceramics.[12] However, the high temperature application of such CMC's is limited to oxygen free atmosphere because of easy oxidation of these additives. In this work, two new model materials were selected for coatings on carbon fibers, namely Si3N4 and Yb2SiO5. The former is intended for the use in almost additive-free Si3N4-ceramics, obtained by new sintering methods like, e.g., field assisted sintering, the latter for Si3N4-ceramics with additives containing rare earth (RE)-silicates.[13] RE-silicates are known for their outstanding oxidation resistance and hydrothermal stability, which is superior over pure SiO2 grown on Si3N4.[14] They are therefore suitable as environmental barrier coatings (EBC) to protect silicon nitride ceramics in high temperature combustion environment.

As most suitable method to prepare Si3N4 coatings, chemical vapor deposition of silicon followed by a subsequent nitridation was selected and solid state reaction starting from nano-particulate powders for the Yb2SiO5 coating, as described in detail in the experimental part of the paper. The CVD-process is performed at ambient pressure with the help of a microwave plasma, as recently developed in our chair for coating short carbon fibers with AlN in a fluidized bed reactor, as described elsewhere.[15] In this work, the coating process established for AlN is modified to produce Si3N4 coatings.

The reactions of the fibers and coatings are investigated by means of dynamic thermo-gravimetric and evolved gas analysis (TGA-EGA), SEM-EDX, and XRD, up to 1500 °C in synthetic air as well as in N2. The Si3N4 and Yb2SiO5 coatings shift the onset point of oxidation T00 to higher temperatures in comparison to the uncoated substrates and reduce the oxidation rates. With a non-isothermal oxidation model with constant heating rate described in the experimental part, also the apparent activation energy Qapp and exponential prefactor Aapp could be determined from the single experiments. The investigations of the reaction products after the heat treatment in N2 reveal that Yb2Si4CN6 has formed. The expected mechanical properties of this new class of material called “nitridocarbidosilicates” is discussed by means of bond strength comparison with SiC, SiALONs, and Si3N4 and is now under investigation by the authors.[16, 17]

2 Experimental Section

2.1 Materials and Chemicals

Two different qualities of carbon fibers are coated with either Si3N4 or Yb2SiO5, with the aim, to enhance their oxidation resistance. Loose carbon short fibers (C25 S003 EPY, SGL Carbon, Germany, l = 3 mm, diameter = 7 μm, the ≈3% epoxy sizing was burned of in air before usage) further denoted as F1 and “endless” carbon fibers (denoted as F2) that are fabricated into felts (Sigratherm GFA10, SGL Carbon) are used. The first coating is a microwave plasma CVD process for the deposition of Si on the fibers with subsequent nitridation step at 1300 °C in N2 atmosphere in a high temperature furnace (Thermal Technology, Germany), to form Si3N4. Trichlorosilane (SiHCl3, 99%, Sigma–Aldrich GmbH, Germany) is used as Si-precursor that decomposes in the microwave plasma at temperature between 600 and 1000 °C. The setup of the microwave fluidized bed reactor is published elsewhere.[15] The highly oxidation resistant Yb2SiO5 is deposited by a dip coating process with aqueous solutions containing Yb-nitrate (Yb(NO3)3 · xH2O, 99.99%, Chempur GmbH, Germany) and colloidal SiO2 (Ludox AS-30 colloidal silica, Sigma–Aldrich GmbH). For the dip coating process, the fibers were pretreated in H2O2–HNO3 solutions, to introduce hydrophilic surface groups. To fix the Yb2SiO5 coating, the coated fibers are dried in N2 atmosphere at 1000 °C.

2.2 Characterization

The morphology and the elemental composition of the coating and the fibers is investigated by SEM-EDX (JSM-840A, JEOL, Japan; INCA 4.05 EDS, Oxford Instruments Microanalysis Limited, UK) before and after coating as well as after the heat treatment in air and N2. The crystalline phases are determined by powder XRD measurements (X'Pert-MPD System, Phillips Analytical X-Ray B.V., The Netherlands). The oxidation behavior is measured in synthetic air (20 vol% O2) with a STA device (STA-449C Jupiter, Netsch, Germany). Alumina crucibles without lids were used, and the flow rate is set to 3 Nl h−1. The heating rate h = dT/dt is set to 5 K min−1. The weight fraction of the coatings is calculated by the ratio of burn of mass to initial mass. The onset point of oxidation T00 is randomly chosen to be the temperature where the relative mass drops below 90% of the initial mass m0 although the oxidation starts at lower temperatures. This value seems to be more reliable, because it can be easier distinguished than 99% weight loss for instance, which can also result from a poor calibration or vibrational or thermal fluctuations during the measurement.

The oxidation of carbon is a single step reaction and CO2 is assumed to be the only gaseous species evolving, so the reaction velocity can be described by Equation (1) whereas the reaction rate constant k is defined as an Arrhenius function given in (2) with the kinetic parameters A, the exponential prefactor and Q the activation energy

display math(1)
display math(2)

The activation energy Q is the measure of the energy necessary for the reaction to occur and the pre-exponential factor A is the measure of how many particles posses this energy, or the possibility that the reaction can occur. The fractional mass conversion is defined as (3) (m, l, r = mass, length, radius of the fibers, index 0 means initial condition):

display math(3)

The oxidation process takes place only at the accessible fiber surface. In case of uncoated fibers, it is equal to the skin surface (2πrl). The two basal areas (r2π) can be neglected without making any pronounced mistake. The radius reduces according to (3) assuming l = l0 because of the high aspect ratio of the fibers. The coating can be assumed to cover the whole surface, but as soon as cracks or pores open at around T00 the oxidation starts. In this case, the accessible fiber surface becomes equal to two times the basal area (r2π). The length of the coated fibers reduces during the oxidation according to (3) but r = r0. These simple geometric assumptions, together with the heating rate h = dT/dt and Equation (1) and (2) give two differential equations for the oxidation velocities (4) and (5) for uncoated and coated fibers, respectively that differ by a factor of four only. According to Sima-Ella et al. these equations can be integrated assuming that Q and A are constant together with an approximation given by Coats and Redfern that leads to expressions for 1 − u(T), see Equation (6) and (7).[18] These equations can be used to fit the measurement data so the apparent kinetic constants Qapp and Aapp are calculated from the best fitting values. Equation (6) is used for uncoated fibers but (7) for all coated fibers that show a pronounced shift in T00

display math(4)
display math(5)
display math(6)
display math(7)

3 Results

3.1 Thermal Analysis

The thermal analyses reveal a different oxidation behavior for the two different fiber substrates F1 and F2. The onset point of oxidation of the felt F2 is roughly 100 °C higher at 899 K than 814 K for F1. The shift of T00 for the two different fiber substrates coated with Si3N4 or Yb2SiO5 are collected in Figure 1. All coatings shift T00 to higher temperatures. The smallest shift is observed for the Yb2SiO5 coated F2 fibers that reach a maximum of only ΔT00 = 54 K for 60 wt% coating. The coating of the F2 fibers with the same amount of 60 wt% Si3N4 shift T00 to a value of ΔT00 = 127 K. One has to take into account that the Si3N4 coating is approximately 1 µm thick compared to only 0.4 μm for the Yb2SiO5 coating due to their different densities (math formula, math formula). The largest shift of ΔT00 = 151 K is reached for F1 fibers coated with 25 wt% Si3N4.

Figure 1.

Onset temperature of oxidation T00 of F1 and F2 fibers, coated with various amounts of Yb2SiO5 or Si3N4.

In Figure 2, the reaction velocities du/dT are plotted versus u(T) for some representative oxidation measurements. After reaching T00, the oxidation starts and the velocities show linear increasing behavior at the early stage. The reaction constants kl for the linear regime can be obtained by the slope and are listed in Table 1 for all measurements. After reaching a certain weight loss of approximately 50 or 75% for the uncoated or coated fibers, respectively, the reaction velocities reach their maximum and then decrease rapidly to zero, that is finally reached at 100% weight loss of course. The linear reaction velocity for F1 = 0.036 K−1 is lower than for F2 = 0.057 K−1 and all coated samples show a lower reaction velocity than the uncoated fibers.

Figure 2.

Oxidation velocities of F1 and F2 fibers and Si3N4 as well as Yb2SiO5 coated fiber substrates versus fractional mass conversion u(T).

Table 1. Table of the kinetic parameters Qapp and ln Aapp for all investigated fiber samples as well as the onset point of oxidation T00 and the linear reaction constant kl
SubstrateCoatingQapp [kJ mol−1]ln Aapp [s−1]kl [K−1]T00 [K]
F1 85 ± 5130.036 ± 4.67 × 10−4814 ± 5
F121 wt% Si3N4190 ± 7170.011 ± 2.34 × 10−4955 ± 5
F125 wt% Si3N4117 ± 9150.020 ± 5.06 × 10−4965 ± 5
F2 108 ± 12170.057 ± 6.56 × 10−4899 ± 5
F243 wt% Si3N4206 ± 13180.017 ± 3.10 × 10−4984 ± 5
F246 wt% Si3N4183 ± 5170.013 ± 1.34 × 10−4999 ± 5
F262 wt% Si3N4187 ± 11180.014 ± 3.42 × 10−41026 ± 5
F228 wt% Yb2SiO5115 ± 18100.017 ± 1.52 × 10−4928 ± 5
F259 wt% Yb2SiO5162 ± 5150.017 ± 1.52 × 10−4953 ± 5

The Yb2SiO5 coatings reduce the oxidation velocities down to values of kl= 0.017 K−1. An increasing amount of Yb2SiO5 coating from 28 to 59 wt% does not further decrease the velocity. The Si3N4 coatings on the F2 fibers decrease the linear reaction velocities to kl= 0.019 K−1 for 11 wt% further to the minimum value of kl= 0.013 K−1 for 46 wt% Si3N4, which is also the lowest value measured for all coated fiber substrates. A similar behavior already seen for the F2 fibers is obtained from the coated F1 fiber substrates. The oxidation velocities decrease from the maximum value kl= 0.036 K−1 to kl= 0.020 K−1 for 25 wt% further to the minimum value of kl= 0.011 K−1 for 21 wt% Si3N4 coatings.

All oxidation measurements are fitted with Equation (6), respectively Equation (7) if they show a pronounced shift in T00. Some of these measurements and the fitted curves are shown in Figure 3. One can see that the calculated data fit well with the measurements. The kinetic values Qapp and Aapp obtained from the best fitting values are shown in Table 1.

Figure 3.

Fractional mass conversions 1 − u(T) versus absolute temperature T of uncoated F1 and F2 fibers, as well as Yb2SiO5 and Si3N4 coated F1 and F2 fiber substrates. The best fits for the measurements are also shown. The values for Qapp and ln Aapp obtained by the best fitting values are collected in Table 1.

The activation energy for F1 is determined to be 85 kJ mol−1 and is lower than 108 kJ mol−1 for the F2 fibers. The apparent activation energies for F1 fibers coated with Si3N4 scatter and are higher, between 190 and 117 kJ mol−1. Qapp also increases for the Yb2SiO5 coated F2 fibers with increasing coating to a maximum value of 162 kJ mol−1 for 60 wt% Yb2SiO5. The values for Qapp of the F2 fibers coated with Si3N4 also increase with increasing coating thickness up to 206 kJ mol−1. The apparent exponential prefactors ln Aapp are also determined and vary in a wide range between 10 and 18 s−1. The difficulty to determine ln Aapp is already known from the Arrhenius plotting of oxidation measurements. With this fitting method, the errors for Aapp are also very high and are therefore not shown in Table 1.

3.2 Microstructure Analysis

The two used carbon substrates show different surface structures (see Figure 4). The surface of F2 appears smoother than of F1. Some particles stick on both surfaces but consisting of carbon only according to EDX analysis.

Figure 4.

SEM image of F1 (left) and the F2 fiber substrates (right).

The microwave plasma treatment, prior to the Si deposition removes these particles from the surface and increase the roughness of the fibers significantly (Figure 5, left). During the deposition, silicon grows island like on the roughened fiber surface and increasing deposition times finally give a total coverage of the fiber surface when the Si islands start to overgrow each other (Figure 5, middle). In the EDX, small amounts of chlorine and oxygen impurities are detected, coming from the TCS-precursor. The nitridation of the silicon coating does not further change the roughness of the coating (Figure 5, right), but the nitrogen content approaches the composition of Si3N4 and only small amounts of oxygen impurities can still be detected but no chlorine.

Figure 5.

F2 fiber surface after plasma treatment (left) and Si-coated F2 surface (middle). The Si3N4-coated F2 fiber surface after the nitridation of the Si coating is shown in the right image. The elemental compositions according to the EDX analyses are also shown.

After the oxidation experiments, see Figure 6 (left), most parts of the Si3N4 coating stay intact leafing elongated hollow tubes behind, without endcups but no pronounced pores or cracks are visible in the coating. The coating thickness after 10 min deposition time reaches around 1 μm for approximately 50 wt% Si3N4 coating, as expected, but is still thin compared to the fiber diameter of 8 μm. According to the EDX analysis, silicon, carbon, and nitrogen are found in the former coating. The oxygen content slightly increased from 1.5 to 2.5 at% and originates from SiO2 that has formed during the oxidation process. Some carbon can still be detected, suggesting the formation of small amounts of SiC during the heat treatment. The heat treatment in nitrogen (see Figure 6, right) reveals no significant change in the surface structure and according to the EDX carbon, nitrogen, and silicon are detectable only.

Figure 6.

Hollow Si3N4 tubes, after oxidation experiments (left) and after heat treatment in N2 (right). The elemental compositions according to the EDX analyses are also shown.

The dip coated Yb2SiO5 films are uniform and do not change the morphology of the fibers (see Figure 7), but they are much thinner compared to the Si3N4 coatings. No cracks are visible directly after dipping, but after the drying step at 1000 °C in N2 (see Figure 7, right). The EDX analyses show the presence of ytterbium, silicon, oxygen, carbon, and nitrogen. The nitrogen content probably results from YbN formation during the drying step in N2.

Figure 7.

Yb2SiO5 coating on F2 fiber substrate, directly after the dip coating (left) as well as after the drying step (right). The elemental composition according to the EDX analysis is also shown.

After the oxidation experiments, no hollow tubes can be found but porous Yb2SiO5 skeletons (see Figure 8, left). The carbon content found in the EDX results from the carbon sample holder shining through the skeleton and does not result from the Yb2SiO5 skeleton itself. After the heat treatment in N2, the Yb2SiO5 coating totally changes its morphology (see Figure 8, right). Needle like Yb2Si4CN6 crystals start to grow on the fiber surface, containing carbon, silicon, ytterbium, and nitrogen only.

Figure 8.

Oxidized Yb2SiO5 coated F2 fibres (left) as well as after heat treatment in N2 (right). The elemental compositions according to the EDX analyses are also shown.

3.3 Phase Analysis

The powder diffraction measurements (see Figure 9 and 10) show that both fiber substrates consist of amorphous carbon resulting in two broad, amorphous peaks around 25° and 43° 2θ although the F2 fibers are said to be, at least partially, graphitized. After the silicon deposition by the microwave CVD process, Si is the only crystalline phase (see Figure 9A). After nitridation, the Si transforms into Si3N4, which appears in the α- and β-modification, and SiC (see Figure 9B). After the oxidation tests (Figure 9C), no qualitative changes in the phase composition are observed although the relative amount of SiC has reduced and amorphous SiO2 has formed, according to the EDX, but which is difficult to be detected by XRD. After heat treatment in N2 (Figure 9D), the qualitative phase composition does not change much, too, but during the oxidation test as well as the heat treatment in N2 some α-Si3N4 transforms into the high temperature β-modification.

Figure 9.

XRD measurements of F2 fiber substrates coated with Si (A), and after the nitridation of the Si into Si3N4 (B). The XRD pattern (C) shows the oxidized Si3N4 coated F2 fibers and pattern (D) shows the heat-treated Si3N4 coated F2 fibers in N2 atmosphere.

Figure 10.

XRD pattern of Yb2SiO5 coated F2 fibers directly after the dip coating (A). The XRD pattern of the coated fibers after the oxidation tests is shown in (B) and after the heat treatment in N2 is shown in (C).

It was not possible to exactly determine the phases of the Yb2SiO5 coated fibers directly after coating (see Figure 10A) because the coating thickness is very small and the fixing step at 1000 °C in N2 only removes the volatile species but is not high enough to crystallize the high melting Yb2SiO5. The oxidation in air (Figure 10B) removes the carbon and lead to Yb2SiO5 formation as expected but small amounts of Yb2Si2O7 and unreacted Yb2O3 and SiC are visible, too. The heat treatment in N2 (Figure 10C), leads to the formation of the so-called nitridocarbidosilicate crystals with the composition Yb2Si4CN6.

4 Discussion

4.1 Coating

The dip coating process of Yb2SiO5 is an easy process that became possible when the fibers where pretreated in H2O2–HNO3 solution, to make them hydrophilic. However, the coating thickness is very low and contains cracks (see Figure 7). These cracks appear after the heat treatment step, where the deposited material was fixed at the fibers. The reason is the evaporation of volatile species like H2O or NOx as well as the shrinkage of the coating and different coefficients of thermal expansion between coating and fiber.

The atmospheric pressure microwave plasma assisted CVD process to deposit AlN onto carbonaceous particles could be successfully applied and modified, to deposit also Si on the C-fiber surface. The microwave power is absorbed by the carbon, heating the fibers directly and above a certain power level, the thermal energy is even high enough to emit free electrons that can be accelerated in the electromagnetic field and ignite a plasma. So the TCS precursor molecules decompose not only at the hot fiber surface, like in the conventional CVD processes, but mainly in the plasma. The deposition of Si on the fiber surface seems to be rather an “infiltration” of the fiber bunches with Si particles that have already formed in the plasma. This observation is consistent with the SEM images where Si islands can be seen growing together with increasing deposition times. Although high growth rates of around 0.1 μm min−1 can be achieved in this CVD process, a draw back is, that the Si layer thickness reduces when going from the substrate surface to the inside of the fiber bundles. It is also difficult to get a real dense Si coating, because the Si islands touch each other only punctually that favors a porous coating. The heat treatment increases the density of the Si layer due to sintering of the Si islands and the volume increase of 21% during nitridation of Si into Si3N4. No Si is found in the XRD after nitridation but only Si3N4 and small amounts of SiC that is assumed to have formed in the contact region between C and Si. The carbon–Si3N4 bond strength seems to be high, because of the rough surface of the plasma treated fibers, so the thermal tension that sets up due to the different coefficients of thermal expansion can easily be transferred from the fiber to the coating or vice versa. Because of this, the highest tension values or the highest probability that cracks occur is expected at the fiber ends.

4.2 Oxidation Behavior

To our knowledge, both fiber substrates are so called PAN-derived carbon fibers but the F2 fibers are additionally graphitized, that means they were heat treated in inert atmosphere at temperatures between 1500 and 3000 °C, making them more “graphitic.” According to our XRD measurements, only amorphous carbon is visible but the more graphitic properties of the F2 substrates results in a higher onset point of oxidation T00. The oxidation rates behave in the opposite way, F2 oxidizes faster (kl= 0.057 K−1) than F1 (kl= 0.036 K−1). The reason might be that the F1 fibers are made for reinforcing purposes (good mechanical properties) but F2 for thermal insulating applications (good thermal properties). The shift of T00 is an important value because it reveals how good the coating encapsulates the fibers, but as soon as cracks or pores open, oxygen can react with the fibers. Therefore, we decided to use non-isothermal oxidation tests with constant, moderate heating rates, which are closer to reality than classical isothermal oxidation tests. The maximum T00 is dependent on the strength of the coating material and can be estimated according to Equation (8), which is valid for thin coatings. For typical values of reaction bonded Si3N4 (RBSN) with E ≈ 100 GPa, σ ≈ 300 MPa, ν ≈ 0.2 and Δα ≈ 2.3 × 10−6 K−1 ΔT00 can be estimated to be ≈1000 K, which agrees well with our results (see Table 1).

display math(8)

The oxidation cannot be prevented above T00 but the reaction velocities are reduced. We could show with a simple geometrical model that a maximum reduction by a factor of four can be achieved. The oxidation velocities of fibers coated with Si3N4 are reduced by this factor but not the Yb2SiO5 coated fibers. That is because the Yb2SiO5 coatings always contain cracks unlike Si3N4, so Si3N4 coatings remain as hollow tubes whereas the Yb2SiO5 remain as a porous skeleton. Free convection also supports the transport of the oxidation products in case of non-coated, or to some extent Yb2SiO5 coated fibers, which is not possible when the fibers oxidize inside a tube made of Si3N4. One has also take into account, that the oxidation mechanism can change from CO2 to CO formation when the oxidation takes places above ≈1000 K according to the Boudouard equilibrium. To get Qapp and Aapp from the single oxidation tests, the developed non-isothermal kinetic Equation (6) and (7) were fitted to the measurement data. The obtained values for uncoated fibers: F1 = 85 kJ mol−1; F2 = 108 kJ mol−1 are somewhat small but in the range of values published in literature.[18-20] According to the EDX and XRD measurements, the coatings themselves do not oxidize significantly, so in general Qapp and Aapp should not change, because carbon is the only oxidizing species. Gas diffusion processes through or inside the Si3N4 tubes or Yb2SiO5 skeletons increase Qapp in the range of several 10 kJ mol−1, which can explain some but not all our results where we have an increase of more than 60–100 kJ mol−1, see Table 1.[21] In some samples, the coating thicknesses were inhomogeneous, and we could see also larger particles of Yb2SiO5 or Si3N4 placed in the contact points of different fibers. That means the calculation of the wt% coating content from the burn off mass contains an error and should be improved. We could show that our non-isothermal model validates the measurements in a satisfying way, although it does not take into account possible changes of the oxidation mechanism according to the Boudouard equilibrium or diffusion processes, so there is still plenty of room for improvements of the model, so all values determined in this work should rather be taken as good approximations.

4.3 Formation of Carbidonitridosilicates

The reactions between the Si3N4 coating and the C-fibers are weak as expected. Besides small changes in the SiC and SiO2 content, no significant changes could be observed even at temperatures up to 1500 °C neither in N2 nor in synthetic air. The Yb2SiO5 coated fibers however show a totally different behavior. In synthetic air, Yb2SiO5 reveals its good oxidation stability that is well known and already used for EBC applications. But to our big surprise, a totally new class of material has formed when the Yb2SiO5 coated C-fibers were heat-treated in N2 atmosphere. In a systematic investigation, we produced reaction couples containing different RE oxides RE2O3 (RE = Yb, Er, Y, Gd, and Sm) and SiO2 in different molar ratios mixed with carbon. The XRD results for all different RE reveal a homolog group of patterns see Figure 11, so finally we could determine the peaks coming from so called carbidonitridosilicates RE2Si4CN6.

Figure 11.

XRD pattern for the RE2O3–SiO2–carbon reaction couples for RE = Yb, Er, Y, Gd, and Sm.

We were able to solve the crystal structures for the RE2Si4CN6 phases starting from Ho2Si4CN6 (PDF#: 04-011-935) by simple exchanging the Ho atoms by the RE atoms. In Figure 12, the refined crystal structure for Yb2Si4CN6 is shown, for example, which fits perfectly and finally proves the existence of the nitridocarbidosilicate crystals growing on the C-fiber surface.

Figure 12.

XRD pattern of single-phase Yb2Si4CN6 and Rietveld refinement. The R values and goodness of fit (GOF) are also shown.

To our knowledge RE2Si4CN6 have never been considered as a possible binder phase for Si3N4 ceramics. But according to the crystal structure, and bond strength considerations the nitridocarbidosilicates should have very promising properties. In Si3N4, for example, nitrogen atoms connect three [SiN4]-tetrahedron centers but in SiC the carbon atoms connect already four [SiC4]-tetrahedrons. A higher degree of condensation is known to improve the mechanical properties. Additionally, the bond strength of Si[BOND]C is stronger than Si[BOND]N or Si[BOND]O bonds. Therefore, materials like nitridocarbidosilicates who contain both, Si[BOND]C and Si[BOND]N bonds, can be expected to give stronger ceramics than Si3N4 or SiAlONs for instance because Si[BOND]N or Si[BOND]O bonds are replaced by Si[BOND]C bonds. Although their promising crystal structure, to our knowledge no mechanical or thermal properties of RE2Si4CN6 are published so far, and are therefore under intensive investigation by the authors, to be used as binder phase in Si3N4 ceramics for example.

5 Conclusion

We could successful adopt and modify the atmospheric pressure microwave CVD coating procedure developed for AlN coating to produce Si3N4 coatings on different fiber substrates. The dip coating with Yb2SiO5 was also possible, but the deposited layers are thin compared to the Si3N4 layers, and crack formation could not be prevented so far. By simple geometrical assumptions, we successfully expanded the kinetic model for non-isothermal oxidation tests suggested by Sima-Ella et al. to be used for coated fibers, too. The reduction of the oxidation velocities of the coated fibers could be explained by the model leading to a maximum decrease by a factor of four. By fitting our oxidation measurements with the model equations, we were also able to determine Qapp and Aapp of the uncoated as well as the coated fibers. The obtained values for the uncoated fibers are in the range expected for carbonaceous materials. The increase in the apparent kinetic parameters for the coated fibers is due to gas diffusion processes through or inside the Si3N4 tubes or the porous Yb2SiO5 skeletons as well as a change in the reaction mechanism (Boudouard equilibrium). The model does not take into account these diffusion processes or the Boudouard equilibrium so far and should be further improved. We observed shifts of T00 whose maximum is related to the strength and the difference in coefficients of thermal expansion of the coating and the fiber. The Si3N4 coating seems to be very resistant against reactions with the carbon fibers as well as against oxidation. Interestingly, the Yb2SiO5 strongly reacts with the carbon fibers, forming the new class of material the so-called nitridocarbidosilicate Yb2Si4CN6. We were also able to show that the formation of nitridocarbidosilicates is not restricted to Yb, but also works for Y, Sm, Gd, and Er. According to bond strength considerations in Yb2Si4CN6, Si–O and Si–N bonds are replaced by Si–C bonds and therefore a higher degree of condensation is achieved, so we predict that the mechanical and thermal properties of these nitridosilicates are superior compared to Si3N4 or SiAlONs. That is why the nitridocarbidosilicates are now under intensive investigation by the authors.