Microstructure and Thermal Conductivity of Sintered Reaction-Bonded Silicon Nitride : The Particle Size Effects of MgO Additive

.e particle size effect of MgO as a sintering additive on the thermal conductivity of sintered reaction-bonded silicon nitride (SRBSN) was investigated. It was revealed that the size of MgO is critical for thermal conductivity with regard to the microstructural evolution process. .at is, the abnormal grain growth promoted by an inhomogeneous liquid-phase distribution led to higher thermal conductivity when coarser MgO was added, whereas a relatively homogeneous liquid-phase distribution induced moderate grain growth and lower thermal conductivity when finer MgO was added.


Introduction
Silicon nitride is considered as a viable component in engines and other parts given its excellent mechanical properties as well as its useful chemical and high-temperature resistance capabilities [1][2][3][4].Meanwhile, since the mid-1990s, Japanese researchers and others have examined its thermal properties.In recent years, sintered reaction-bonded silicon nitride (SRBSN) has been reported to obtain a high thermal conductivity of 177 W•m −1 •K −1 [5].Its excellent mechanical properties and high thermal conductivity make SRBSN an attractive material for high-power electronic devices [6][7][8][9][10].
It is known that the particle size of silicon powder is one of the most critical factors affecting the mechanical properties of the resulting SRBSN [11][12][13].Specifically, the pore size of reaction-bonded silicon nitride (RBSN) is generally dependent on the particle size of the starting silicon powder.In most cases, larger pores in RBSN are difficult to eliminate during the postsintering process, resulting in the degradation of the mechanical properties of SRBSN.erefore, fine silicon powder is suitable as a starting material of SRBSN, and the particle size of silicon can be easily decreased through a milling process [14,15].
Notably, it has also been proved that the grain size is a dominant microstructural factor with regard to the thermal conductivity of SRBSN [16,17].First, the selfpurification of Si 3 N 4 grains with the exclusion of lattice oxygen generally occurs with the grain growth via a solutionreprecipitation process [18].Secondly, because the intrinsic thermal conductivity of the glassy intergranular phase is much lower than that of Si 3 N 4 grains, the elimination of the grain boundaries with grain growth would be beneficial when attempting to realize high thermal conductivity [5,15].
In the present study, we present further improvements of the thermal conductivity of SRBSN based on well-known theories of abnormal grain growth in the research field of silicon nitrides.erefore, a preliminary attempt is carried out to obtain a coarse microstructure in SRBSN by adjusting the particle size of the MgO additive.e coarse MgO additive used here was expected to induce composition fluctuations of the liquid phase and thus abnormal grain growth in the SRBSN.

Experimental Procedure
e silicon (grade 4NE, 99.99%, Vesta Ceramics, Ljungaverk, Sweden), yttria (grade C, H.C. Starck, Goslar, Germany), and magnesia (metal basis 99.99%, Sigma Aldrich) were used as starting powders.In order to control the particle size, each raw powder was crushed in ethanol by planetary milling at 300 RPM for 12 h using ZrO 2 balls and jars.e particle size distributions of each powder were measured using a laser diffraction particle sizing analyzer (Beckman Coulter LS-13320, USA).Also, the specific surface area of specific powder was measured by the BET method (3 Flex, Micromeritics, USA).
e nominal composition of the starting powder mixture was set as Si 3 N 4 : Y 2 O 3 : MgO � 93 : 2 : 5 at a molar ratio assuming that Si is perfectly nitrided.e powder mixtures were ball milled in ethanol in a polyethylene bottle at 200 RPM for 4 h.e dried powders were shaped into V � 15 mm pellets by cold isostatic pressing at 200 MPa.e green compacts were then placed in an alumina tube furnace to fabricate nitrided bodies (RBSN) by flowing a 95%N 2 -5% H 2 mixed gas and heating up to 1450 °C with a heating rate of 2.5 °C/min and without a holding time.e phase identification was conducted by an X-ray diffractometer (D/Max 2200, Rigaku, Japan) using Cu α radiation at 40 kV and 100 mA.en, the nitridation degree was calculated based on the actual and theoretical weight increases of the nitrided compacts [11].After nitridation, the contents of metallic elements (Si, Y, and Mg from the starting powder and Zr from the ZrO 2 milling media) were analyzed by ICP-AES (inductively coupled plasma atomic emission spectrometry, Spectro Flame Modular EOP, SPECTRO) using a destructive approach.Postsintering was carried out at 1900 °C for 6 h under static N 2 pressure of 0.9 MPa.After postsintering, the contents of the metallic elements were analyzed by XRF (Xray fluorescence spectrometry, XRF-1800, SHIMADZU, Japan) nondestructively.In order to observe the microstructure of the sintered body, the surface was initially polished to a 1 µm finish and then plasma etched (SNTEK, Korea) using a gas mixture of CF 4 and O 2 (volume ratio, 46 : 4).e microstructure was observed by a scanning electron microscope (SEM; JSM-5800, Jeol, Tokyo, Japan).e grain sizes were quantitatively analyzed using commercial software (Image-Pro 7.1, Image Cybernetics, USA).e thermal conductivity (K) at room temperature was calculated based on the following equation: where q is the bulk density.e thermal diffusivity (α) was measured using a laser-flash method (LFA437, Netzsch, Germany), and the heat capacity (C p ) was set as a constant value 0.68 J

Results and Discussion
e particle sizes and compositions of the powder batches are summarized in Table 1.After high-energy planetary milling for 12 h, the particle sizes of Si and MgO were clearly decreased, while that of Y 2 O 3 had not obviously changed.In addition, the specific surface area of Y 2 O 3 was measured by BET adsorption isotherm model and found to be almost same (13.8m 2 •g −1 and 13.1 m 2 •g −1 ) regardless of milling time.Specifically, the particles of the pulverized MgO (1.7 µm) were approximately three times smaller than those of the as-received MgO powder (5.9 µm).
e nitridation degrees of the Si compacts and the relative densities of the Si compacts, RBSN, and SRBSN are plotted in Figure 1.Both specimens exhibited nitration degrees higher than 90% and were free from residual Si based on XRD analyses (Figure 2).e relative densities of the RBSN samples increased compared to those of the corresponding Si compacts.is is attributed to the filling of internal pores with the transformation from Si to Si 3 N 4 (weight gain ∼66.5%) [20].
e relative density of the SRBSN specimens exceeded 98% regardless of the particle size of MgO.
e microstructures of the SRBSN specimens are presented in Figures 3(a) and 3(b).It was noted that the grain size of Si 3 N 4 was affected by the characteristics of the MgO additive.Specifically, coarser MgO additive was advantageous for promoting the grain growth.Generally, the homogeneity of powder mixtures deteriorates with an increase in the starting particle size.Based on the above findings of the sintering of Si 3 N 4 , the SRBSN specimens in the current research using coarse MgO additives resulted in an inhomogeneous distribution of the liquid phase and enhanced abnormal grain growth.e different grain sizes of the Si 3 N 4 in the SRBSN samples were clearly identified, as shown in Figures 3(c) and 3(d), using Image-Pro 7.1 software.
It is highly consistent and was very interesting to observe as well that the inhomogeneous microstructure featuring abnormally grown large elongated grains could be successfully tailored using coarse sintering additives.e critical factor proposed in this study is the particle size of the MgO additive.In previous studies, the number of irregularly shaped boundaries between β-Si 3 N 4 grains was shown to increase with a decrease in the MgO content [21], and Mg species became volatile at a high postsintering temperature of 1900 °C.Other work reported that the vaporization loss of Mg was especially severe when the total weight loss of the samples exceeded 10 wt.% [22].In other words, the absolute amount of residual Mg species and consequent composition of the liquid phase at the sintering temperature are affected by the volatilization of the MgO additive.In order to determine the dependency of the evaporation loss of MgO on its particle size, the residual contents of Mg species in both the RBSN and SRBSN samples were measured by ICP and XRF, respectively.ese results are shown in Table 2.

2
Advances in Materials Science and Engineering Except for the differences in numbers caused by the apparatus employed, it was revealed that the compositions of the two specimens were nearly identical regardless of the size difference in the MgO additive.Generally, it is postulated that coarser ceramic powders in a powder mixture are dispersed unevenly.erefore, one of the possible causes of the abnormal grain growth is that the eutectic melting point and the viscosity of the liquid phase in the MgO-rich regions were most likely much lower than in the MgO-free regions [23].erefore, the Si 3 N 4 grains in the MgOrich regions would preferentially grow and then act as seeds during the postsintering process, resulting in a coarser microstructure with abnormal grain growth.e area fractions of Si 3 N 4 grains in different sizes in the SRBSN samples are plotted in Figure 4.It was noted that the area fraction of small Si 3 N 4 grains (0∼2 µm 2 ) was dominated by the particle size of silicon, while that of large grains (10∼50 µm 2 and 50∼100 µm 2 ) was strongly affected by the particle size of the MgO additive.e SRBSN sample with a high area fraction of large Si 3 N 4 grains was expected to have high thermal conductivity, as both the lattice oxygen content and the number of grain boundaries decrease with the grain growth of Si 3 N 4 .
e thermal diffusivity and thermal conductivity are shown in Table 3. High thermal conductivity of 87.8 W•m −1 •K −1 was achieved in the sample doped with the  Advances in Materials Science and Engineering coarse MgO additive, while a low value of 77.3 W•m −1 •K −1 was obtained in the sample doped with ne MgO.erefore, it is reasonable to conclude that the coarser MgO additive contributed to the inhomogeneous distribution of the liquid phase in the SRBSN sample during the postsintering process, which has a signi cant in uence on the high thermal conductivity by promoting the abnormal grain growth.

Conclusion
In this study, the e ects of the particle size of MgO when used as an additive on the microstructure and thermal conductivity of SRBSN were systematically investigated.
e coarser MgO became unevenly distributed in the powder mixture, leading to composition uctuations of the liquid phase during the postsintering process.Consequently, the Si 3 N 4 grains around the MgO-rich regions preferentially grow to a large size, promoted by the abnormal grain growth and thus resulting in a higher fraction of larger Si 3 N 4 grains in the nal SRBSN product.As a result, high thermal conductivity of 87.8 W•m −1 •K −1 was successfully obtained in the SRBSN sample with coarse MgO additive.

Figure 2 :Figure 1 :
Figure 2: XRD patterns and phase analysis for the RBSN (a) and SRBSN (b) specimens.e values (R N ) in the figure indicate the nitridation degree for RBSN specimens after the nitriding process.

Figure 4 :
Figure 4: Comparison of area fraction of the SRBSN fabricated by the postsintering of RBSN.

Table 1 :
e particle size and composition of the specimens.

TABLE 2 :
Analysis of elemental contents for both RBSN and SRBSN.