Effect of carbon source on the properties of dense α-SiC

Due to its outstanding properties, SiC is a candidate material for use in special applications such as armor. In order to use SiC in these special applications, it is necessary to produce fully dense ceramics. The ability to produce high density materials with superior performance depends on a number of factors. One of these factors is the addition of carbon to aid sintering. In this study, the effect of different carbon sources and ratios on the elastic and mechanical properties of SiC was investigated. Two types of carbon (lamp black and phenolic resin) were added to SiC in different ratios (0%–2% wt.). All samples were sintered via the spark plasma sintering (SPS) method at 1900 °C for 15 min under 50 MPa pressure. Samples made with lamp black were reached full density at 1.0 wt.%C, and the hardness and elastic modulus values were ∼22GPa and 440 GPa, respectively. While samples made with both carbon sources showed similar bulk mechanical properties, the samples made with lamp black showed more consistent microstructures. The carbon from the phonelic resin source did not appear to be as well distributed as that from the lamp black source. The results also confirmed that addition of carbon into SiC was essential to improve the density and other mechanical properties associated with it.

One challenge with silicon carbide is that sintering is difficult due to its strong covalent bonding and low selfdiffusion coeffcient [16,17]. In order to eliminate this problem, either a sintering aid (B 4 C, B, Al, Si, TiC, ZrB 2 , TiB 2 , Al 2 O 3 , Y 2 O 3 , and C) must be used or it must be sintered at high pressure (hot press, spark plasma sintering) and/or high temperature [18][19][20][21][22][23][24][25][26][27][28]. However using oxide additives can cause the formation of a liquid phase at the sintering temperature which can decreases the mechanical properties of SiC due to the residual glassy phase [29,30].
Another important factor affecting the sintering of non-oxide ceramic materials like SiC is the native oxide layer on their powder surface. For silicon carbide, this layer occurs as silicon dioxide (SiO 2 ). The presence of this oxide layer can contribute to coarsening and inhibit densification during sintering [31][32][33]. The addition of boron and carbon as sintering aids seems to be the best candidate to take advantage of the good properties of SiC, as it aids in solid state sintering and does not form a liquid phase. Boron containing additives increase the densification rate of SiC by increasing the self diffusion of the system [34]. On the other hand, carbon eliminates the oxide layer on the SiC surfaces preventing excessive grain growth [20,21,23,25,28]. Stobierski et al showed that both boron and carbon should be added together to SiC in order to produce highly dense parts since glassy phase formation was observed in samples with only boron added [25,34].
The goal of this study was to evaluate the impact of the carbon source and content on the microstructure, hardness, and elastic moduli of dense silicon carbide materials. In order to achieve this objective, silicone carbide powders with additions of 0.5 wt.% boron carbide and 0-2 wt.% carbon added as either lamp black or phenolic resin, were densified by spark plasma sintering. The dense silicone carbide ceramics were then characterized using the Archimedes method, FESEM with EBSD, ultrasound, and Knoop hardness techniques to determine the effects of the different carbon sources on the microstructure and mechanical properties.
For the samples made with lamp black as the carbon source, called the SG-LBC series, SiC, 0.5 wt.%B 4 C and 0-2.0 wt.% lamp black were mixed by ball milling in ethanol with SiC media for 24 h in a HDPE Nalgene bottle. The slurry was then sieved to remove the SiC media and pan dried. For each sample a 50 gram batch of powder was prepared. The SG-LBC series compositions are shown in table 1.
The amount of phenolic resin needed to be able to generate the same amount of carbon as the SG-LBC series samples (0.5%, 1.0%, and 1.5%) was calculated from the char yield. Thermogravimetric analysis (TGA) was performed on the resin, and showed that the char yield was approximately 43% as shown in figure 1. According to the TGA result, the weight of the added phenolic resin should be 2.58 times the amount of carbon desired in the final formulation. The compositions of the samples made with the phenolic resin carbon source, called the SG-PRC series, are shown in table 2.
The liquid phenolic resin was mixed with deionized water and sonicated to help disperse it. After that SiC and B 4 C powders, ammonium hydroxide (0.22 g) and SiC milling media were added to the liquid mixture and ball milled for 24 h. The slurry was then sieved to seperate SiC media from liquid mixture. To remove the excess water, the liquid mixture was filter pressed at 35 psi and dried in an oven at 100°C.  2.2. Sintering of samples 6.5 grams of dry powder mixture were put in a graphite die (20 mm inner diameter) and densified using a Thermal Technology SPS 10-4 spark plasma sintering unit (Thermal Technology, LLC, Santa Rosa, CA, USA). To prevent possible reaction between powder and die, the die and punches were lined with graphite foil.
To densify the SG-LBC series samples, the powder was heated to 1400°C at a 200°C per minute heating rate under vacuum with applied pressure of 50MPa and held under those conditions for 30 min. After the 30 min holding time at 1400°C, the SPS chamber was backfilled with argon and heated to 1900°C at a 200°C per minute heating rate, maintaining 50 MPa applied pressure and held for 15 min at 1900°C. After sintering, the SPS powder was turned off and the sample was allowed to cool.
To densify the SG-PRC series samples, an additional step was added to convert the resin into carbon. The SPS was first heated to 800°C at a 200°C per minute heating rate under vacuum with applied pressure of 20MPa and held for one hour. Afterwards, the same sintering conditions as the SG-LBC samples were applied. The SPS conditions can be seen in figure 2.

Post sintering processing
After densification, the samples were covered with excess graphite foil. This graphite foil was removed by sand blasting using garnet blasting media. Since the samples needed to be smooth for ultrasound analysis, the samples were ground flat using a surface grinder with a 600 grit diamond wheel. The samples were then cut into small pieces using a LECO VC-50 diamond saw (LECO Corporation, St. Joseph, MI, USA), mounted in epoxy resin using a Buehler SimpliMet 1000 automatic mounting press (Buehler, Lake Bluff, IL, USA), and polished to 0.25 μm finish using a Buehler EcoMet 250 automatic grinder/polisher with AutoMet 250 automatic head in preparation for FESM analysis abd microhardness testing. Polished samples were also ion milled using the flatmilling mode with 3 kV acceleration voltage, 80°tilt angle, no offset, and 25 rpm rotation speed for 10 min using a Hitachi IM4000 (Hitachi High-Technologies Corporation, Tokyo, Japan) to prepare for EBSD analysis. Samples were etched using a modified Murakami method (20 g KOH and 20 g K 3 Fe(CN) 6 in 60 ml deionized water) in order to highlight grains boundaries to better view the microstructure with FESEM.

Results and discussion
3.1. Microstructure characterization Figure 3 shows the microstructures of the SG-LBC series of samples. The grain morphology of all samples were similar and showed equiaxed grain shapes. Only their grain size and their porosity differ according to the amount of carbon they contain. The sample made without carbon addition showed significant amounts of porosity. Addition of 0.5 wt.% C helped reduce the porosity, but was not enough to reach full density. It can be seen from the microstructure images that, by adding 1.0 wt.% and 1.5wt.% C, completely dense SiC samples were obtained. Pores were not observed in the structure. However, some porosity was detected again in the sample with 2.0 wt.% C. Due to the native oxide layer of SiC, it is difficult to obtain high density SiC without carbon addition.
To eliminate the oxide layer, it was essential to add sufficent amounts of carbon [8,17]. It can be seen from FESEM images that increasing the amount of carbon from 0 to 1.0 wt.%, decreased the porosity and the SiC ceramic reached the full density at this point (1.0 wt.%C). However, the grain size of the samples increased with the added carbon ratio from 4.08 μm to 5.43 μm, since there were a few pores to prevent the grain growth. With increasing additions of carbon, samples maintained high density but the average grain size of samples was reduced from 5.43 μm to 2.66 μm due to grain boundary pinning by inclusions of the excess carbon. Similar effects could be seen in the literature [25]. The change of average particle size and standard deviation with carbon content is shown in table 3.
The microstructures of the SG-PRC series of samples can be seen in figure 4. The sample made without C, with 0.5 and 1.0 wt.% C also showed significant amounts of porosity. The sample made with addition of 1.5wt.% C is almost fully dense (>99%), and had very few small pores. The grain morphology and the average grain size of these samples were quite similar. The density values and average grain sizes with standard deviation can be seen in table 4. The pirimary difference between these four samples was the amount of porosity. Adding 1.0 wt.% C enabled to achieve high density for the LBC series. The same result was expected in the PRC series, but this did not happen. When carbon is added to the SiC, the C reacts with the oxide on the surface of the SiC at high temperature and forms volatile oxygen containing species, which are removed via gas transport. In cases where a small amount of added C is not sufficient to reduce the oxygen content, it is prevented densification. In such a  case, it is expected that no carbon will remain in the microstructure as all of the carbon will react with oxygen. However, this was not the case for the SG-PRC series, as there were visible carbon residues in the microstructures of the samples. It can be thought that the carbon from this source accumulates in some regions due to inhomogeneous mixing, and reduces the oxygen content locally, but after sintering, leaves residual carbon in those regions. In areas where there was less carbon, the oxide is not completly removed and results in residual porosity. When comparing both series, the SG-PRC series showed higher amounts of porosity than SG-LBC series when the same amounts of carbon was added. This suggested that the phonelic resin may not mixed as well as the lamp black or that more of the phelonic resin might be lost during sintering than predicted by the TGA, so the remaining carbon was not enough to fully remove the surface oxide layer. EBSD maps of the SG-LBC series of samples can be seen in figure 5. The red color in the maps indicates the 6H-SiC polytype and the green color the 4H-SiC polytype. Black color indicates areas where no SiC phase could be detected because of the presence of secondary phases, pores, roughness, grain boundaries or other factors. It can be seen from the EBSD maps that the samples had a predominantly 6H polytype, and a small proportion of the 4H polytype. The phase fraction of 6H and 4H of SG-LBC series can be seen in table 3. As the phase fractions were determined by the areas of each color in the image, the sum of the 6H and 4H polytypes were not 100% due to the presence of the black areas. The raw SiC powder's 4H/6H ratio was 0.11, and the ratio of 4H/6H for the SG-LBC series were between 0.14-0.24 which were slightly higher than the starting powder ratio. It can be concluded that addition of lamp black carbon had effect on the conversion of 6H polytype to 4H polytype. This inference is also supported by the literature [5].  A similar situation to the SG-LBC series is observed in these samples. Samples mostly contain the 6H-SiC polytypes with smaller amounts of 4H-SiC polytype with the ratio of 4H/6H between 0.15-0.23. As seen in the EBSD maps and FESEM images, all samples showed primarily equiaxed grains of mostly 6H-SiC polytype. This result is supported by previous research the literature shows that 6H-SiC polytype smaller and equiaxed grains and 4H-SiC polytype shows large and elongated grains [8].

Elastic and mechanical properties characterization
The densities and elastic modulus values of the SG-LBC series samples are shown in table 3. When looking at sample densities, it can be seen that 0.5 wt.% C was insufficient to remove the oxide layer and achieve densification, it only reached theoretical density of 96%. The highest density (>99%) was obtained at 1.0 wt.% C addition. This was sufficient to remove the oxide layer from the SiC surface. When 1.5 wt.% C was added, a slightly decrease in density was seen. However, since the high density was already reached at 1.0 wt.% C, the 1.5 wt.% C was more than the carbon required to remove oxygen from the SiC surface, and this excess carbon caused a decrease in the average grain size. At the highest C levels, there was excess carbon residues in the microstructure. Because the theoretical density of carbon is lower than that of SiC, residual carbon in the structure caused the density of the sample to decrease. Lomello et al sintered SiC without additives by SPS at 1900°C for 5 min and under 70 MPa pressure, but due to the high oxygen content of the powder and not adding carbon to remove it, only 96% density was achieved [19].
Similar effects can be seen in the elastic modulus of the samples. The highest elastic modulus was achieved at 1.0 wt.% C since it had the highest density. The elastic modulus of materials has close relationship with materials density and secondary phase inclusions. It can be seen from the table that lower elastic modulus were obtained at low carbon additions because of porosity, while the modulus values of samples with larger carbon additions were reduced by the residual carbon. The 0 wt.% C and 0.5 wt.% C are very similar in density and microstructure and that the difference in elastic modulus is probably due to variability in the ultrasound measurement. The standard deviation value of this elastic modulus is ±5. Asimilar effect can be seen with conventional sintering method, since SiC sintered without applied pressure could not reach high density and had pores in the microstructure and could only reach the elastic modulus of 409 GPa [11].
The density values of the SG-PRC series samples and elastic modulus values are shown below in table 4. It can be seen from the results that increasing amounts of added carbon results in an increased in density. While only 95% of the theoretical density was achieved when 0.5 wt.% C was added, a theoretical density of>99% could be reached when 1.5 wt.% C was added.
Looking at the work of Guillard et al, it is clearly seen how important the additives are in the sintering of SiC. SiC samples sintered using SPS at 75 MPa pressure, 5 min at a temperature that could be considered as high as 1850°C , only 92% density was achieved [30].
Load-knoop hardness curves for SG-LBC series samples are shown below in figure 7. It can be clearly seen in the graph that the hardness of the samples decreased with the increasing loads. This behavior is seen in ceramic materials in general [35]. The effect of carbon addition on the hardness of SiC also follows the similar trend to the  effect on the elastic modulus. Hardness values increased with the density of the SiC, and obtained the highest value at 1.0 wt.% C. As increasing the carbon addition from 0.5 to 1.0 wt.%, first hardness value increased from 19.89 to 21.98 GPa at 1000 g load. Then increased the carbon content from 1.0 to 1.5 wt.%, the hardness values slightly decreased from 21.98 to 21.34 GPa at 1000 g load. Increasing the carbon addition further, at 2.0 wt.% C, the hardnes value was 20.83 GPa at 1000 g load. At 0.5 and 2.0 wt.% C the low hardness values were due to the effects of porosity and residual carbon. This effect of porosity has also been seen in other studies [19]. It can be concluded that just enough carbon should be added to SiC to produce dense samples with better elastic and mechanical properties, as residual carbon also has a negative effect on SiC mechanical properties.
Load-knoop hardness curves for SG-PRC series samples are shown below in figure 8. By increasing the amount of added of carbon from 0.5 to 1.5 wt.%, the hardness values increased from 19.35 to 21.43 GPa at 1000 g load. This is again because of the increasing densities of samples with addition of carbon. The highest value was obtained at 1.5 wt.%C. It has been seen in previous studies that with the addition of more than 1.5-2 wt.% carbon to SiC, residual carbon is easily seen in the microstructure and reduced the hardness of SiC [17,36]. Therefore, in this study, phenolic resin was added to SiC at a maximum of 1.5wt.%. This was slightly different from the SG-LBC series where the highest hardness was obtained at 1% C before the hardness decreased as more carbon was added. This may be because more of the carbon from the phenolic resin was lost during processing than anticipated or because the phenolic resin was not homogeneously distributed so that it could effectively remove all of the surface oxide.

Conclusions
To investigate the effect of different carbon additives on the properties of SiC, two different carbon sources were added to SiC with 0.50 wt.% B 4 C. The carbon sources were particuate lamp black (LBC), and liquid phenolic resin (PRC). All samples were densified by SPS at 1900°C for 15 min under 50 MPa applied pressure in argon atmosphere. This sintering temperature and time were sufficient for high density SiC production.
The general results can be listed as follows; • Samples made with low carbon content or without carbon showed significant amounts of porosity.
• Low elastic, and mechcanical properties were obtained at low carbon addition because of porosity, or with high amount of carbon addition due to residual carbon.
• Addition of either type of carbon had affect the conversion of 6H to 4H SiC polytype.
• Addition of carbon of both types affected the density, average grain size, hardness, and elastic properties.
For the SG-LBC series, the highest hardness, elastic modulus, and density were obtained at 1.0 wt.% C additon, and the values were ∼22 GPa, 440 GPa, and >99%, respectively. For the SG-PRC series the highest hardness, elastic modulus, and density were obtained at 1.5 wt.% C additon, and the values were 21.43 GPa, 441 GPa, and >99%, respectively. Although the results were similar for both series, differences in microstructural properties were observed. The results suggest that the phonelic resin was not able to be mixed as well as the lamp black carbon. These results showed that uniform addition of carbon to SiC is essential for obtaining high density and other properties associated with it. However, only enough carbon must be added to the SiC to remove the surface oxide layer as the presence of residual carbon can adversely affects the mechanical properties of SiC.