Oxide Additive-Free α- → β-Si₃N₄ Phase Transformation at 1900 °C

1. Introduction

Because of their high thermal conductivity and excellent mechanical properties, silicon nitride (Si₃N₄) ceramics have been successfully applied as insulated substrates in the semiconductor power modules mounted on inverters assembled in Electric Vehicle (EV) and Hybrid Electric Vehicle (HEV) [1,2]. The material properties of Si₃N₄ceramics are closely related to their microstructures, and a bi-modal structure composed of fine matrix Si₃N₄ grains and elongated large grains is ideal for harmonizing high fracture strength and toughness as well as high thermal conductivity [3–5]. For fabricating such high-performance β-Si₃N₄ ceramics, the rod-like β-Si₃N₄ seeding to starting α-Si₃N₄ powders is a practical way to construct the bi-modal structure composed of fine matrix grains and selectively grown elongated grains originating from the seed crystallites through the liquid phase sintering [6–8]. Regarding the method for preparing rod-like β-Si₃N₄ seed crystallites, Hirao et al. [9] reported the procedure with several steps as follows: (i) heat treatment of α-Si₃N₄, yttria (Y₂O₃) and silica (SiO₂) mixed powders at 1850 °C under N₂ gas pressure of 0.5 MPa to afford rod-like β-Si₃N₄ polycrystallites with Si-Y-O-N glassy phase, (ii) pulverizing the aggregates using a mortar and pestle, (iii) glass phase dissolution by acid treatment followed by neutralization treatment and cleaning, (iv) classifying the resulting isolated seed crystallites and (v) drying. On the other hand, recently, we proposed a facile method to produce rod-like β-Si₃N₄ seed crystallites by the oxide additive-free single-step heat treatment at 1900 ℃ of commercial α-Si₃N₄ powder [10–12]. Subsequently, the bi-modal structure controlling was successfully achieved for the liquid phase-sintered β-Si₃N₄-MgO (1 wt%)-Re₂O₃ (3 wt%) (R = Gd, Yb, Y, La) ceramics by the addition of 10 vol% rod-like β-Si₃N₄ seeds produced at our laboratory, and the resulting ceramics sintered at 1900 ℃ for 40 h exhibited improved thermal conductivity of higher than 110 Wm^-1K^-1 [13].

Generally, it is accepted that the α- ⇀ β-Si₃N₄ phase transformation proceeds through the dissolution-precipitation process in the oxynitride liquid phase formed in-situ at the heat treatment temperature [14,15] and besides the oxide additives [16–18], the initial contents of free silicon [19] and β-Si₃N₄ phase in the starting α-Si₃N₄ powder [20] have been reported as additional factors to accelerate the formation of β-Si₃N₄. On the other hand, the oxide additive-free α- ⇀ β-Si₃N₄ phase transformation has been reported for fully densified Si₃N₄ ceramics fabricated by high-pressure sintering of α-Si₃N₄ powder using techniques such as hot isostatic pressing (HIP) [21–24] and high-pressure and high-temperature (HPHT) sintering [25]. Hou et al. [25] reported that micron-size α-Si₃N₄ was transformed into submicron-sized polycrystallites β-Si₃N₄ under the HPHT condition at 5.5 GPa and temperatures ranging from 1600 to 1900 ℃, and the α- ⇀ β-Si₃N₄ phase transformation was found to complete at 1900 ℃ for 3 min. As a possible mechanism for the observed phase transformation, the nucleation-growth mechanism of liquid-solid phase transformation was suggested since the observed α- ⇀ β-Si₃N₄ phase transformation proceeded at around the melting point of Si₃N₄ under high pressure [25,26]. The high-resolution TEM observation of the sample fabricated by the HPHT sintering revealed full-phase transformation from α- to β-Si₃N_4, while the β-Si₃N₄ crystal grain boundaries were free from secondary crystalline or glassy phase, i.e., there was no evidence for the eutectic liquid-phase formation during the HPHT sintering [25,26].

In our previous study [12], the correlation between the α- ⇀ β-Si₃N₄ phase transformation behavior and the changes in the amount of oxygen and carbon impurities, and the α- ⇀ β-Si₃N₄ phase transformation kinetics at 1900 ℃ were investigated under nitrogen pressure of 980 kPa. Our analytical results revealed that the experimentally observed α- ⇀ β-Si₃N₄ phase transformation was suggested to be promoted by the formation of metastable solid solution between the impurity oxygen and α-Si₃N₄ followed by decomposition and release of oxygen-containing volatile species to afford β-Si₃N₄[12].

In the present study, to investigate the unique phase transformation in more detail, the oxide additive-free α- ⇀ β-Si₃N₄ phase transformation at 1600 to 1900 ℃ is directly compared with the reference powder sample of α-Si₃N₄ doped with 1 mol% yttrium oxide (Y₂O₃) in which the α- ⇀ β-Si₃N₄ phase transformation can be promoted via the liquid phase [27]. Subsequently, the oxide additive-free Si₃N₄ powder samples with various β-phase content are prepared by varying the heat treatment duration time at 1900 ℃. The correlation between the α- ⇀ β-Si₃N₄ phase transformation behavior and changes in the oxygen-impurity content of the selected Si₃N₄ crystal grains of both α- and β-phase and those of their two-grain boundaries is investigated and discussed based on the results obtained by a set of characterization techniques including elemental analysis, X-ray diffraction (XRD), transmission electron microscopy (TEM), high-angle annular dark-field scanning TEM (HAADF-STEM), STEM-energy dispersive X-ray spectroscopy (EDS) and electron backscatter diffraction (EBSD) analyses.

2. Experimental procedures

2.1. Starting powders and heat treatment condition

Commercially available α-Si₃N₄ powder (grade SN-E10, Ube Industries, Ltd., Tokyo, Japan) was used as-received. The specific surface area (SSA), mean particle size (d₅₀) and β-phase content are 11.0 m²g^-1, 0.9 μm and 3.73 %, respectively. The total metal impurity, oxygen and carbon impurity contents are less than 50 ppm, 1.23 and 0.11 wt%, respectively. As a comparison study, to observe the α- ⇀ β-Si₃N₄ transformation through the liquid phase, Y₂O₃ powder (grade UU-type, purity, 99.9 %, SSA: 15.9 m²g^-1, d₅₀: 0.5 μm, Shin-Etsu Chemical, Co. Ltd., Fukui, Japan) [28] as used as an oxide additive: As-received α-Si₃N₄ powder (labeled as E10) was mixed with 1 mol% Y₂O₃ powder and ball milled in ethanol for 2 h and dried at 80 ℃ for 5 h under a flowing nitrogen (N₂). The resulting mixed powder was labeled as E10 +1 mol% Y₂O₃.

A prescribed amount of the E10 and E10+1 mol% Y₂O₃ sample powders were placed in a boron nitride (BN) crucible and heat-treated in a graphite resistance-heated furnace (Model High Multi 5000, Fuji Dempa Kogyo, Co., Ltd., Osaka, Japan) at selected temperatures from 1600 to 1900 °C for 5 h under N₂ gas pressure of 980 kPa. In addition, the effect of the 1900 °C heat-treatment duration time on the morphology and α- ⇀ β-Si₃N₄ phase transformation was studied for up to 20 h.

2.2. Characterizations

The evaluation of oxygen and carbon impurities and the β-phase content of the powder samples was conducted in the same manner as described in our previous report [12].

Morphologies of the powder samples were observed by using a scanning electron microscope (SEM, Model JSM-7900F, JEOL Ltd., Tokyo, Japan.

In this study, TEM/HAADF-STEM, STEM-EDS and EBSD analyses were intensively performed on the powder samples to investigate the α- ⇀ β-Si₃N₄ transformation behavior associated with the changes in the oxygen impurity contents within the Si₃N₄ crystal grains and those at their grain boundaries. TEM specimens were prepared by focused ion beam (FIB) techniques (Model FB-2100, JEOL Ltd., Tokyo, Japan): The powder sample was embedded within epoxy resin and maintained under vacuum for 5 min by using a rotary pomp. Then, the pressure was returned to atmospheric pressure and the mixed compound was cured at room temperature for 12 h. The cured compound composed of the powder sample and resin was mechanically pre-polished. The pre-polished surface of the cured compound was covered with tungsten (W) protective thin film, then the resulting compound was transferred to the FIB equipment for final thinning.

The TEM and HAADF-STEM observations were carried out by using an atomic resolution analytical electron microscope with a STEM Cs corrector incorporated as standard (Model JEM-ARM200F, JEOL Ltd., Tokyo, Japan) operating at an acceleration voltage of 200 kV, which allows to perform the HAADF-STEM observation at a high resolution of 78 pm.

EDS analysis was conducted in the STEM mode with a spectrometer (Model JED-2300T, JEOL Ltd., Tokyo, Japan) mounted on the atomic resolution analytical electron microscope. Selected area electron diffraction (SAED) analysis was performed to identify the unit cell structure whether α-Si₃N₄ or β-Si₃N₄ phase and the symmetries (crystal axis and face orientation) of the Si₃N₄ phase.

In this study, HAADF-STEM analysis for determining the atomic configuration in the vicinity of the Si₃N₄ grain boundary was performed under the operation conditions as previously discussed by Clarke and Thomas [29] and conducted by Kleebe et al. [30,31] and Kim et al. [32]: The crystal grain boundary has to be tilted in an edge-on position to the incident beam. Furthermore, the crystal grain boundary should be flat with a low density of interfacial steps, and the crystal grains on either side of the grain boundary should be aligned in an orientation suitable for structure imaging or at least for forming one set of lattice fringes.

In this study, the α-Si₃N₄ crystal grains in the heat-treated powder sample with an especially high β-phase content of 96.5 % were identified by the EBSD analysis (Model Aztec HKL Advanced Symmetry, Oxford Instruments Holdings 2013 Inc., Tokyo, Japan with application equipment: Model Map Sweeper, Oxford Instruments Holdings 2013 Inc., Tokyo, Japan). The EBSD measurements were carried out in low vacuum operating mode which eliminated the coverage of the surface since the gas ionized by the electrons was capable of neutralizing the surface charges. A High-resolution EBSD image on an approximately 6 × 4 μm² area of the sample TEM specimen was obtained under the operation condition with a step size of 20 nm and an acceleration voltage of 20 kV.

3. Results and discussion

3.1.α-⇀β-. Si₃N₄ phase transformation behaviors

3.1.1. Morphological changes and β-phase content

As a reference, the morphologies of the starting α-Si₃N₄ and Y₂O₃ powders used in this study are shown in Figure 1.

The morphology changes and the α- ⇀ β-Si₃N₄ phase transformation behaviors during the heat treatment up to 1900 ℃ of the E10 and E10+1 mol% Y₂O₃ powder samples are summarized and shown in Figure 2 and Figure 3a, respectively. The addition of 1 mol%-Y₂O₃ promoted Si₃N₄ elongated grain growth at 1700 to 1800 ℃ (Figure 2b,c) associated with the completion of α- ⇀ β-Si₃N₄ phase transformation at 1800 ℃ (Figure 3a), which are consistent with those previously observed during the liquid-phase sintering of Si₃N₄-Y₂O_3-SiO₂ system [33]. On the other hand, even after the 1900 ℃ -heat treatment for 5 h, the E10 sample kept the equiaxed morphology (Figure 2f), and the β-Si₃N₄ phase content remained below 25 % (Figure 3a). As we reported previously [12], continuous grain growth proceeded during the extended 1900 °C-heat treatment for up to 20 h to afford hexagonal rod-like grains (Figure 2g–i). Moreover, β-Si₃N₄ phase content increased consistently with the 1900 °C-heat treatment time to reach 100 % at 20 h (Figure 3b).

3.1.2. Relation between chemical composition and β-phase content

The oxygen (O) and carbon (C) impurity contents of as-received and heat-treated α-Si₃N₄ powder samples are listed in Table S1. The contents of O impurity calculated as SiO₂ (mol %) [12] are plotted as a function of heat treatment temperature and 1900 °C-heat treatment time in Figure 3a,b, respectively.

After the 5 h-heat treatment up to 1900 ℃, the SiO₂ and C contents decreased from 5.17 to 2.74 mol% and 1.23 to 0.8 mol%, respectively (Figure 3a and Table S1). Accordingly, the observed increase in the β-Si₃N₄ phase content at 1600 to 1900 ℃ was suggested by the β-Si₃N₄ formation via carbothermal nitridation reaction between the impurity C and SiO₂ generally recognized as a surface oxide layer formed by the impurity oxygen (Eq. (1)) [34]. As another possible route for the β-Si₃N₄ formation, silicon oxynitride (Si₂N₂O) formation and decomposition (Eq. (2) [35]) could be considered since Si₂N₂O was suggested as another possible phase at 1600 to 1800 ℃ according to the phase diagrams of the binary SiO₂-Si₃N₄ system [36,37].

3SiO₂ (s) + 6C (s)+ 2N₂ (g) → β-Si₃N₄ (s) + 6CO (g)

(1)

3Si₂N₂O (s) → β-Si₃N₄ (s) + N₂ (g) + 3SiO (g) (T > 1750 °C)

(2)

During the extended 1900 °C-heat treatment for up to 20 h, both the SiO₂ and C contents decreased consistently with the heating time (Table S1), which revealed that the carbothermal nitridation (Eq. (1)) continuously proceeded, however, this contribution to the increase in the β-Si₃N₄ phase content was essentially small since the amounts of the SiO₂ and C after the 1900 °C-heat treatment for 5 h were as low as 0.64 wt% (2.74 mol% as SiO₂) and 0.07 % (0.8 mol%), respectively (Table S1). It should also be noted that the silicon oxynitride liquid phase formation was excluded at such low SiO₂ contents [36,37].

The experimental results obtained in the present study clearly revealed that the dominant mechanism for the α- ⇀ β-Si₃N₄ phase transformation of the E10 sample at 1900 °C was intrinsically different from that generally discussed for the liquid phase sintering of Si₃N₄ ceramics [38]. In this study, we focused on the role of impurity oxygen in the α- ⇀ β-Si₃N₄ phase transformation, and the relation between the α- ⇀ β-Si₃N₄ transformation behavior and the change in the oxygen impurity content shown in Figure 3b was further studied.

3.2. Nanostructure characterization by HAADF-STEM and STEM-EDS analyses

3.2.1. Phase identification

The changes in the oxygen impurity contents within the Si₃N₄ crystal grains and their two-grain boundaries were intensively studied by the HAADF-STEM observation combined with STEM-EDS analysis. In this study, two powder samples with apparently different β-Si₃N₄ phase content of 53.5 and 96.5 % were selected and labeled as Sample 1 and Sample 2, respectively, as shown in Figure 3b.

First, HAADF-STEM analysis was performed on the selected Si₃N₄ crystal grains for the assignment of α- and β-phase and the identification of the three different kinds of two-grain boundaries, α/α-phase, α/β-phase and β/β-phase.

(a) Sample 1 with β-phase content of 53.5 %

A typical TEM image of the Sample 1 is shown in Figure 4a. Within this image, two grains labeled G1 and G2 were selected.

The HAADF-STEM image of the G1 and the SAED pattern shown in Figure 4b corresponded to the plane orientation adjusted to the [0001] zone axis of α-Si₃N₄. Analogously, that of the G2 shown in Figure 4c corresponded to the plane orientation adjusted to the [01$\overline{1}$1] zone axis of α-Si₃N₄. Accordingly, the two-grain boundary between the G1 and G2 (GB1/2) was assigned as α/α-phase.

Further nanostructure characterization was performed on Sample 1 and the results are summarized and shown in Figure 5. Three grains labeled G3, G4 and G5 in the BF-STEM image shown in Figure 5a were selected. The HAADF-STEM image and the SAED pattern analyses resulted in assigning the G3 as α-Si₃N₄ (Figure 5b) and the G4 as β-Si₃N₄ (Figure 5c), while the G5 (Figure 5d) exhibited a typical HAADF-STEM image of the basal plane observed from the [0001] direction of β-Si₃N₄ [39], indicating the GB3/4 and GB3/5 as α/β-phase.

On the other hand, the selected grain labeled G6 in the BF-STEM image in Figure 5e exhibited a typical HAADF-STEM image of the prismatic plane observed from the [10$\overline{1}$0] direction of β-Si₃N₄ [40] (Figure 5f), and those labeled G7 and G8 in Figure 5e were also assigned as β-Si₃N₄ (Figure 5g,h), thus it was identified that both the GB6/7 and GB6/8 as β/β-phase.

(b) Sample 2 with β-phase content of 96.5 %

To find out α-Si₃N₄ crystal grains and the two-grain boundary of α/β-phase in Sample 2 with a high β-phase content of 96.5 %, the TEM-EBSD analysis was performed as shown in Figure 6a in which the yellow colored and blue colored grains were β-Si₃N₄ and α-Si₃N₄, respectively. Within the matrix composed of β-Si₃N₄ grains, the α-Si₃N₄ grains were located randomly, and their grain sizes were in the range of about 0.2 to 0.8 μm, while those of β-Si₃N₄ grains were approximately 0.15 to 1.00 μm.

Based on the results obtained by the EBSD combined with BF-STEM analyses (Figure 6a,b), three grains labeled G9, G10 and G11 were selected and confirmed as α-Si₃N₄, β-Si₃N₄ and β-Si₃N₄, respectively　(Figure 6c–e). As a result, both the GB9/10 and GB9/11 were assigned as α/β-phase. Further analysis was performed for several two-grain boundaries in Sample 2, and as a typical result, the BF-STEM image, HAADF-STEM image, and SAED pattern of the selected grains labeled G12 and G13 were shown in Figure 7.

3.2.2. Oxygen impurity contents within Si₃N₄ crystal grains and at two-grain boundaries

Figure 4d,e show a HAADF-STEM image near the α/α-phase grain boundary between the G1 and G2 (GB1/2), and the result of STEM-EDS line scanning analysis to detect the changes in the O element near the grain boundary of the two grains, respectively. The HAADF-STEM image showed no obvious existence of glassy phases at the two-grain boundary, however, the detective count for O element was highest at the X-ray scan distance of 9.5 nm. Then, this position was fixed as the center of the two-grain boundary, and for measuring the oxygen impurity content, the STEM-EDS analysis was performed for the square area with 4.0 nm width and 8.0 nm length in each Si₃N₄ crystal grain, G1 and G2 as well as for the square area with 0.3 nm width and 2.0 nm length in the two-grain boundary, GB1/2 (Figure S1a). The center position of the former square sites was set at 7 nm from the center of the two-grain boundary as indicated by the red arrows in Figure 4d,e.

The spectra obtained by the STEM-EDS area analysis for the G1, G2 and GB1/2 are shown in Figure S2. The O element was detected as a minor peak in each spectrum, and the impurity O contents of the G1, G2 and GB1/2 were evaluated as 1.2, 1.0 and 1.0 wt%, respectively (Table S2).

A set of the HAADF-STEM observation near the two-grain boundary accompanied by the STEM-EDS line scanning and area analysis for O element was performed on all the labeled grains G3 to G13 and their two-grain boundaries (Figure 8 and Figure 9 and Figure S1).

The HAADF-STEM image near the two-grain boundary GB3/4 shown in Figure 8a revealed the G3 and G4 grains in the edge-on condition. Here, when the grain boundary was under edge-on conditions, the incident direction of the electron beam was under off-Bragg conditions for each grain. On the other hand, it was difficult to obtain an edge-on condition for the grain boundary GB3/5, therefore, the observation was conducted by adjusting to the orientation of the G5 grain from the [0001] direction of β-Si₃N₄ crystal to obtain the clear image as shown in Figure 5d.

Analogously, the HAADF-STEM image analyses for the area nearby the GB6/7 (Figure 8c) and GB6/8 (Figure 8d) were conducted by adjusting to the orientation of the G6 grain [10$\overline{1}$0] direction of β-Si₃N₄ crystal to obtain clear images as shown in Figure 5f, and those for the area nearby the GB9/10 (Figure 9a) and GB9/11 (Figure 9b) were conducted by adjusting to the orientation of the grain G9 [12$\overline{3}$1] direction of β-Si₃N₄ crystal to obtain clear image as shown in Figure 6c. On the other hand, the HAADF-STEM image analysis for the area near the GB12/13 were conducted by adjusting to the orientation of the grain G13 [10$\overline{1}$0] direction of β-Si₃N₄ crystal to obtain a clear image as shown in Figure 9c.

All the HAADF-STEM images showed no obvious existence of glassy phases at the two-grain boundary. Then, the impurity O contents were evaluated in the same manner as shown in Figure S1. The impurity O contents are listed in Table S2 and compared with those for the G1, G2, and GB1/2 in Figure 10.

Due to the low impurity O contents (Table S1), It was indeed difficult to precisely evaluate the amount of impurity O by the present STEM-EDS area analysis. In this study, the relative difference in the amount of the impurity O was discussed for the samples. The results obtained for Sample 1 suggest that, regardless of the phase combination, impurity O content at the two-grain boundary is higher than those within the two Si₃N₄ crystal grains (Figure 10a). Sample 2 in which the β-phase content reaches 96.5% (Figure 10b) also shows a similar tendency, however, the impurity O contents measured for the Si₃N₄ crystal grains are varied: In Sample 1, the impurity O contents in the β-Si₃N₄ crystal grains are 0.4 to 0.9 wt% and found to be lower than those measured for the α-Si₃N₄ crystal grains (approximately 1.0 wt%). On the other hand, in Sample 2, in addition to the α-Si₃N₄ crystal grain G9 (1.5 wt%), particular β-Si₃N₄ crystal grains such as the G10 and G11 which form two-grain boundary with the α-Si₃N₄ crystal grain G9 exhibit relatively high impurity O content above 1.0 wt%, while other β-Si₃N₄ crystal grains such as G12 and G13 show much lower impurity O contents of 0.5 and 0.6 wt %, respectively.

The experimental results obtained in the present study suggest that the impurity oxygen was initially distributed uniformly both in the α-Si₃N₄ crystal grains and at the grain boundaries. Then, the 1900 °C -heat treatment promoted oxygen diffusion from Si₃N₄ crystal grains to their grain boundaries associated with the α- ⇀ β-phase transformation of the α-Si₃N₄ crystal grains.

Around the final stage of the α- ⇀ β-phase transformation reaching 96.5 % β-phase (Figure 10b), the impurity O seemed to be concentrated within the local area composed of α-Si₃N₄ crystal grain remained, the β-Si₃N₄ crystal grains surrounding the α-Si₃N₄ crystal grain and their two-grain boundaries. The α-Si₃N₄ crystal grain might act as an oxygen scavenger to increase the impurity O content within the local area since α-Si₃N₄ has been reported as an oxygen-stabilized phase in an approximate formula of Si_11.5N₁₅O_0.5 [41,42]. However, the subsequent oxygen diffusion from the β-Si₃N₄ crystal grains to their grain boundaries proceeded, which reduced the resulting impurity O content in the β-Si₃N₄ crystal grains as low as 0.5 to 0.6 wt%. The impurity O concentrated at the grain boundaries was thought to be released from the sample via the grain boundary diffusion, and the resulting total impurity O content of the powder sample measured by the inert gas fusion method [12] was reduced to 0.12 wt% (Table S1).

3.3. Role of impurity oxygen in α- ⇀ β-Si₃N₄ phase transformation

It has been suggested that the formation of α-Si₃N₄ is attributed to kinetic reasons and metastable compared to β-Si₃N₄ [43,44]. As one possible pathway for the α- ⇀ β-Si₃N₄ phase transformation, Zakorzhevskii suggested that, at low impurity O content (0.5 to 0.7 wt%), the impurity O dissolves into the crystal lattice of α-Si₃N₄ to afford a metastable solid solution [45] which is the driving force of α- ⇀ β-Si₃N₄ phase transformation below the theoretical α-Si₃N₄ dissociation temperature [36]:

α-Si₃N₄ crystal → 3Si + 2N₂ (dissociation) → β-Si₃N₄ crystallization

(3)

The impurity O content of the E10 sample powder after the 1900 °C -heat treatment for 5 h was as low as 0.64 wt% (2.74 mol% as SiO₂), and thus silicon oxynitride liquid phase formation was excluded at such low SiO₂ content [37,38], indeed the 1900 °C -heat treated powder samples exhibited clear Si₃N₄ crystal grain boundaries without secondary glassy phases (Figure 4, Figure 8 and Figure 9), and the final impurity O content after the 1900 °C-heat treatment for 20 h reduced to be 0.12 % (Table S1). Moreover, the α- ⇀ β-Si₃N₄ phase transformation experimentally observed in the present study proceeded mainly at 1900 °C under N₂ pressure of 980 kPa, which was 33 °C lower than the theoretical α-Si₃N₄ dissociation temperature (1933 °C [46]). These results strongly suggest that the oxide additive-free α- ⇀ β-Si₃N₄ phase transformation observed in our present study is dominantly governed by the reaction Eq. (4) promoted by the metastable solid solution formation between the oxygen impurity (SiO₂) and α-Si₃N₄ followed by dissociation to afford thermodynamically favorable β-Si₃N₄ accompanied by the release of oxygen.

α-Si₃N₄ crystal + xSiO₂ → metastable solid solution →

(4)

[3Si + 2N₂ + 1/2xO₂] → β-Si₃N₄ crystal + 1/2xO₂ ↑

(4)

4. Conclusions

In this study, high-purity commercial α-Si₃N₄ powder with an initial β-phase content of 3.7 %, total metal impurities below 50 ppm, and oxygen and carbon impurity contents of 1.23 and 0.11 wt%, respectively was used to investigate the oxide additive-free α- ⇀ β-Si₃N₄ phase transformation at 1600 to 1900 °C under nitrogen pressure of 980 kPa. The results can be summarized as follows:

(1)

The α- ⇀ β-Si₃N₄ phase transformation was found to proceed mainly at 1900 °C, and the extensive heat treatment for up to 20 h achieved full transformation to afford rod-like β-Si₃N₄ polycrystallites. This transformation temperature was found to be 33 °C lower than the theoretical α-Si₃N₄ dissociation temperature.

(2)

The impurity O contents of the powder samples after the 1900 °C -heat treatment for 5 h and 20 h were measured to be 0.64 and 0.12 wt %, respectively. At such lower oxygen contents, the silicon oxynitride liquid phase formation at 1900 °C was excluded according to the phase diagram reported for the binary Si₃N₄-SiO₂ system.

(3)

The HAADF-STEM, as well as the STEM-EDS analyses performed on the 1900 °C heat-treated powder samples, also revealed no evidence for the formation of the secondary crystalline or glassy phases at the Si₃N₄ two-grain boundaries.

(4)

The HAADF-STEM/STEM-EDS analyses performed on the area near the two-grain boundary clarified that, at the β-phase content of 53. 5 %, regardless of the phase combination, the impurity oxygen content at the two-grain boundary was higher than those within the two Si₃N₄ crystal grains.

(5)

As a possible dominant mechanism, the oxide additive-free α- ⇀ β-Si₃N₄ phase transformation at 1900 °C was suggested to be governed by the formation of metastable solid solution between the α-Si₃N₄ and the impurity oxygen remained at approximately 0.6 wt%, which promoted the dissociation below the theoretical α-Si₃N₄ dissociation temperature to afford thermodynamically favorable β-Si₃N₄. Along with the β-Si₃N₄ formation, the impurity oxygen was concentrated at the Si₃N₄ crystal grain boundaries and subsequently released from the sample via the grain boundary diffusion.