Potential use of only Yb₂O₃ in producing dense Si₃N₄ ceramics with high thermal conductivity by gas pressure sintering

Abstract

Yb₂O₃ is an efficient sintering additive for enhancing not only thermal conductivity but also the high-temperature mechanical properties of Si₃N₄ ceramics. Here we report the fabrication of dense Si₃N₄ ceramics with high thermal conductivity by the gas pressure sintering of α-Si₃N₄ powder compacts, using only Yb₂O₃ as an additive, at 1900 °C under a nitrogen pressure of 1 MPa. The effects of Yb₂O₃ content, sample packing condition and sintering time on the densification, microstructure and thermal conductivity were investigated. Curves of the density plotted against the Yb₂O₃ content exhibited a characteristic ‘N’ shape with a local minimum at 3 mol% Yb₂O₃ and nearly complete densification below and above this concentration. The effects of the sample packing condition on the densification, microstructure and thermal conductivity strongly depended on the Yb₂O₃ content. The embedded condition led to more complete densification but also to a decrease in thermal conductivity from 119 to 94 W m^-1 K⁻¹ upon 1 mol% Yb₂O₃ addition. The sample packing condition had little effect on the density and thermal conductivity (102–106 W m⁻¹ K⁻¹) at 7 mol% Yb₂O₃. The thermal conductivity value was strongly related to the microstructure.

Introduction

Silicon nitride (Si₃N₄) ceramics have been intensively studied over the past four decades for their high-temperature structural applications owing to the excellent mechanical properties of Si₃N₄ at elevated temperatures. However, little attention was paid to the thermal conductivity of Si₃N₄ ceramics until Haggerty and Lightfoot pointed out in 1995 that Si₃N₄ has an intrinsic thermal conductivity of over 200 W m⁻¹ K⁻¹ at room temperature [1]. A considerable amount of work resulted in significantly increased thermal conductivities of Si₃N₄ ceramics of above 100 W m⁻¹ K⁻¹ [2–7]. Enhanced thermal conductivity opens up new technological applications of Si₃N₄ as substrates for integrated circuits and heat sinks in electronic devices.

Si₃N₄ has two major crystalline forms, α and β. The β form is more suitable for applications because α-Si₃N₄ is chemically unstable and converts to β-Si₃N₄ upon heating and because β-Si₃N₄ grains grow in an elongated manner that reinforces the ceramic matrix [8]. Moreover, β-Si₃N₄ has substantially higher intrinsic thermal conductivity than α-Si₃N₄ [9]. Owing to the strong covalent bonding between silicon and nitrogen atoms, additives are required to densify Si₃N₄ ceramics by the liquid-phase sintering [10]. Self-reinforced β-Si₃N₄ is commonly produced from an α-Si₃N₄ powder that usually contains a small fraction of β-phase (e.g. UBE SN-E10), because the α–β phase transformation promotes the development of large elongated grains. During the sintering, the additives react with the native SiO₂ on the surface of Si₃N₄ particles and with Si₃N₄ to form an eutectic liquid phase through which densification, α–β phase transformation and β-Si₃N₄ grain growth occur. The sintering additives determine the densification, phase transformation, grain growth and characteristics of the grain boundary phase; thus, the use of effective sintering additives is crucial for improving the thermal and mechanical properties of Si₃N₄ ceramics.

The key issues in enhancing the thermal conductivity of β-Si₃N₄ ceramics are the complete densification, purification of β-Si₃N₄ grains and a reduced amount of grain boundary phases; all of which are dependent on the sintering additives and sintering techniques [2, 7]. Dissolved O and Al in the β-Si₃N₄ lattice are the two main impurities that lower the thermal conductivity of β-Si₃N₄ via phonon-defect scattering, and thus the use of SiO₂ and Al₂O₃ additives should be avoided [4, 11]. To enhance the thermal conductivity, sintering additives should play a dual role of promoting densification and removing lattice oxygen. Among the rare-earth oxides, Y₂O₃ and Yb₂O₃ are the most promising additives for enhancing the thermal conductivity of β-Si₃N₄ ceramics [12]. Because of its economic advantage over hot isostatic pressing and the improved thermal and mechanical properties compared with Si₃N₄ ceramics produced by pressureless sintering, gas pressure sintering has been widely used to produce β-Si₃N₄ ceramics with high thermal conductivity. In gas pressure sintering, a higher N₂ pressure allows the sintering of powder compacts at higher temperatures without significant thermal decomposition. Higher temperatures not only favor β-Si₃N₄ grain growth but also make it possible to reduce the amount of sintering additives and to use more stable compounds [3, 6]. However, Yb₂O₃ or Y₂O₃ alone is not efficient in producing dense Si₃N₄ ceramics by gas pressure sintering because of their high melting temperatures and high eutectic temperature with SiO₂. Therefore, Yb₂O₃ or Y₂O₃ is normally used together with another additive, such as MgO [4, 13, 14], MgSiN₂ [16–18] or ZrO₂ [7].

Nishimura et al [19] have shown that Yb₂O₃ results in substantially better sinterability than Y₂O₃. Lee et al [20] have studied the effect of adding Yb₂O₃ on the microstructural uniformity of gas-pressure-sintered Si₃N₄. Although they did not present density data, the microstructure suggested that complete densification was achieved in some of their samples. Yb₂O₃ is also an effective additive for improving the high-temperature mechanical properties of Si₃N₄ ceramics because of the easy formation of the crystalline Yb₄Si₂O₇N₂ phase at the grain boundaries [19, 21]. Thus, we infer that gas pressure sintering can produce dense β-Si₃N₄ ceramics with the addition of only Yb₂O₃, and such ceramics should have high thermal conductivity and excellent high-temperature mechanical properties. These new β-Si₃N₄ ceramics are potentially interesting structural and functional materials.

In this work, we studied the synthesis of dense Si₃N₄ ceramics by gas pressure sintering using only Yb₂O₃ as an additive. The effects of Yb₂O₃ content, sample packing condition and sintering time on the densification, phase transformation, grain boundary phase, microstructure and thermal conductivity of the ceramics were investigated.

Experimental procedure

The starting raw materials were α-Si₃N₄ powder (SN-E10, α> 95 wt%, BET 10.9 m² g⁻¹, 1.29 wt% O, UBE Industries Ltd, Yamaguchi, Japan) and Yb₂O₃ (purity >99.9%, BET 19.3 m² g⁻¹, Nihon Yttrium Co. Ltd, Tokyo, Japan). The amount of added Yb₂O₃was varied in the range of 0.5–7 mol% with respect to Si₃N₄. The calculated compositions and theoretical densities of the starting powder mixtures are listed in table 1; the SiO₂ phase is inherently present on the surface of raw Si₃N₄ particles. The compositions are also indicated in the phase diagram of the Si₃N₄–SiO₂–Yb₂O₃ system [22] (figure 1).

Table 1.

Calculated compositions and theoretical densities of the starting powder mixtures.

Sample No.	Composition (mol%/wt%/vol%)a			ρth (g cm−3)
	Si3N4	SiO2	Yb2O3
A	94.06/96.22/96.63	5.45/2.39/2.89	0.48/1.39/0.48	3.20
B	93.61/94.89/96.17	5.43/2.36/2.87	0.97/2.76/0.96	3.23
C	92.69/92.29/95.23	5.37/2.29/2.85	1.94/5.42/1.92	3.29
D	91.77/89.78/94.30	5.32/2.23/2.82	2.91/7.99/2.89	3.35
E	89.94/85.01/92.42	5.21/2.11/2.76	4.85/12.88/4.81	3.47
F	88.10/80.55/90.55	5.11/2.00/2.71	6.80/17.45/6.74	3.59

^aThe amount of native SiO₂ on the surface of raw Si₃N₄ powders was estimated as 2.42 wt% based on the oxygen content of 1.29 wt% provided by the manufacturer, and the raw Yb₂O₃ powder is assumed to be pure. The component densities used in calculating the theoretical densities of the composites are 3.19 g cm⁻³ for Si₃N₄, 2.65 g cm⁻³ for SiO₂ and 9.28 g cm⁻³ for Yb₂O₃.

Figure 1.

(a) Overall and (b) localized phase diagram of the ternary Si₃N₄–SiO₂–Yb₂O₃ system. (Points A–F represent the compositions with 0.5, 1, 2, 3, 5 and 7 mol% Yb₂O₃, respectively).

The powders were mixed for 1 h using a planetary mill in a Si₃N₄ jar with Si₃N₄ balls and methanol as a milling medium. The slurry was dried in a rotary evaporator at 60 °C and then in a vacuum oven at 110 °C for at least 4 h. A powder mixture of about 3 g was uniaxially pressed in a 20-mm-diameter stainless-steel die and then isostatically cold-pressed at 400 MPa. The green density showed a slight increase from 53.4 to 54.5% of the theoretical value with increasing Yb₂O₃ content from 0.5 to 7 mol%. A boron nitride powder (BN, GP grade, Denki Kagaku Kogyo Co., Tokyo, Japan) was used as the powder bed. We applied two different packing conditions, with samples either placed onto the BN powder bed (SPC-I) or entirely embedded in the BN powder bed (SPC-II). The samples were placed inside a triple-crucible arrangement consisting of a double BN crucible and an outer graphite crucible. The sintering was conducted in a graphite resistance furnace (Multi-500, Fujidempa Kogyo Co. Ltd, Osaka, Japan) as follows: (i) heating to 1000 °C within 1 h in vacuum, (ii) pressurizing the furnace to 1 MPa at 1000 °C with nitrogen gas, (iii) heating at a rate of 10 °C min⁻¹ to the final temperature, (iv) cooling at a rate of 10 °C min⁻¹ to 1200 °C and then natural slow cooling to room temperature.

Weight loss was determined from the weight change of the sample during sintering. Bulk density was measured by the Archimedes method in distilled water. Phase identification was conducted on the cross sections of sintered samples by powder x-ray diffractometry (XRD, Model RINT 2500, Rigaku Co., Tokyo, Japan) with CuKα radiation. The microstructure was examined by scanning electron microscopy (SEM, Model JSM-5600, JEOL, Tokyo, Japan) on the fracture surfaces coated with a 30-nm-thick gold layer.

To measure thermal conductivity, disk specimens with diameter of 10 mm and thickness of 3 mm were prepared by grinding the sintered samples. Thermal diffusivity was measured by a laser-flash method (Model TC-7000, ULVAC, Yokohama, Japan). Prior to measurements, the surfaces of the specimens were sputter-coated with a 60-nm-thick gold layer to avoid direct transmission of the laser pulse, followed by a subsequent coating of colloidal graphite to enhance the absorption of the flash energy. Thermal conductivity (k) was calculated as

where ρ, C_p and α are the bulk density, specific heat and thermal diffusivity of the disk specimen, respectively. A specific heat of 0.68 J (g K)⁻¹ [6] was adopted in this work.

Results and discussion

Sintering at 1800 °C

To investigate the densification, α–β phase transformation and grain growth at the initial stage of sintering, we interrupted the sintering (SPC-I) when the temperature reached 1800 °C without holding the sample at this temperature. This process is hereafter referred to as ‘initial sintering’. Figure 2 reveals that both densification and a phase transformation occurred, consistent with the lowest eutectic liquid temperature of ∼1650 °C in the Yb₂O₃–SiO₂ system [23] and the findings of Yang et al [24]. Relative densities of 69–74% were achieved in all samples. The density had a maximum at 1 mol% Yb₂O₃, decreased with increasing Yb₂O₃ content, and increased again at 7 mol% Yb₂O₃. The β-phase fraction was ∼70 wt% for ≤3 mol% of Yb₂O₃ and reached 100% at a higher Yb₂O₃ content, i.e, the α–β phase transformation was completed earlier than the densification. The highest weight loss of 1.6 wt% upon sintering was observed in the case of 0.5 mol% Yb₂O₃. The weight loss decreased with increasing Yb₂O₃ content, and a weight gain occurred at 5 mol% Yb₂O₃ or higher. The reason for this is discussed below.

Figure 2.

Relative density and β/(α+β) ratio as a function of Yb₂O₃ content for the samples rapidly sintered at 1800 °C under sample packing condition I.

When a liquid phase is formed, densification proceeds by particle rearrangement followed by dissolution–reprecipitation [10]. Because the phase transformation occurred, the observed densification should be due to the combination of particle rearrangement and dissolution–reprecipitation. Owing to the limited amount of the liquid phase in this study, the dissolution–reprecipitation process should play a dominant role in the densification. The phase transformation promoted by the addition of Yb₂O₃ suggests that the dissolution–reprecipitation process is diffusion controlled. The Yb³⁺ cation is primarily a network modifier of Yb–Si–O–N glass allowing the rapid diffusion of Si and N ions through the liquid phase [25]. A higher Yb₂O₃ content produces a higher volume of the liquid phase; thus, so the solution–reprecipitation process is accelerated resulting in a faster phase transformation.

No crystalline secondary phases could be detected by XRD in the samples with the addition of <7 mol% Yb₂O₃. However, at 7 mol% Yb₂O₃, both the Yb₂SiO₅ (Yb₂O₃:SiO₂ molar ratio of 1) and Yb₄Si₂O₇N₂ (Yb₂O₃:SiO₂=4:1) phases were observed. This is consistent with the fact that composition F is located in the Si₃N₄–Yb₂SiO₅–Yb₄Si₂O₇N₂ compatibility triangle (figure 1(b)).

Samples with 0.5 or 1 mol% Yb₂O₃ had ∼65 wt% β-phase and mostly consisted of fine equiaxial grains with only a few elongated β-Si₃N₄ grains (figures 3(a) and (b)). This illustrates that no significant β-Si₃N₄ grain growth occurred. The β-phase content was similar upon the addition of 3 mol% Yb₂O₃; however, most grains became elongated, and β-Si₃N₄ crystallites larger than several micrometers can be seen in figure 3(c). At 7 mol% of Yb₂O₃, the phase transformation was completed, almost all β-Si₃N₄ grains became elongated (figure 3(d)), and the grain size was larger than that in 3 mol% samples. The promotion of β-Si₃N₄ grain growth upon increasing Yb₂O₃ content is similar to the findings of Park et al [21]. These results suggest that both the phase transformation and grain growth are dominated by the diffusion-controlled solution–reprecipitation mechanism.

Figure 3.

SEM micrographs of fracture surfaces in the samples with (a) 0.5, (b) 1, (c) 3 and (d) 7 mol% Yb₂O₃ sintered at 1800 °C.

Sintering at 1900 °C

Weight loss, bulk density and bulk color

As shown in figure 4(a), the weight loss increased with increasing holding time regardless of the Yb₂O₃ content. However, the effect of Yb₂O₃ content on the weight loss appeared to depend on the holding time. For 2 h holding, the weight loss decreased with increasing Yb₂O₃ content, and a similar result was obtained for rapid sintering at 1800 °C under the same sample packing condition (SPC-I). In the case of holding for 12 h, the weight loss increased with increasing Yb₂O₃ content from 0.5 to 5 mol%. A weight gain occurred upon further increasing in the Yb₂O₃ content from 5 to 7 mol%. The embedded condition (SPC-II) had no effect on the weight loss at <5 mol% Yb₂O₃ but tended to increase the weight loss at >5 mol% Yb₂O₃. This is because some of the Yb-containing oxynitride liquid was transported into the surrounding BN powder bed by the capillary force [26].

Figure 4.

(a) Weight loss and (b) relative density as a function of Yb₂O₃ content for the samples sintered at 1900 °C (holding times 2 and 12 h) under different sample packing conditions (SPCs). All density measurements were conducted immediately after taking samples out of the furnace.

During the gas pressure sintering of Si₃N₄, the weight loss is dominated by the reaction [27]

According to reaction (2), if SiO₂ is depleted then the estimated weight loss decreases from 4.25 to 3.56 wt% with the increasing Yb₂O₃ content from 0.5 to 7 mol%, that is, the highest weight loss is about 4.25 wt% for 0.5 mol% Yb₂O₃ addition. This weight loss is close to the observed values of 4.0–4.2 wt% for 12 h holding. With increasing Yb₂O₃ content, the lower SiO₂ fraction is responsible for the slight decrease in weight loss during rapid sintering at 1800 °C or 2 h sintering at 1900 °C. Our previous work revealed that the Yb content decreased from 4.93 to 2.56 wt% during sintering at 1900 °C for 12 h, corresponding to a loss of 48 wt%; this loss was slightly affected by the type of magnesium compound which was used as the second additive [13]. Thus, the increase in weight loss with increasing Yb₂O₃ content should be associated with the enhanced evaporation of the Yb-containing grain-boundary phases during prolonged sintering. Similar findings were reported by Lee et al [20].

The weight gain at a Yb₂O₃ content of 5 mol% or higher was attributed to the chemical dissolution of nitrogen gas in the oxynitride liquid through substitution for bonded oxygen via the reaction [28, 29]

The addition of Yb₂O₃ increases the contents of oxynitride liquid and dissolved nitrogen. Meanwhile, the SiO vapor generated by reaction (3) can inhibit the progress of reaction (2) and thus reduce the weight loss.

As shown in figure 4(b), the relative density increased with increasing Yb₂O₃ content from 0.5 to 1 mol%, remained constant between 1 and 2 mol%, sharply decreased from 2 to 3 mol% and increased again at higher Yb₂O₃ concentrations. Density depended on the sintering time, particularly for Yb₂O₃ content of 5 and 7 mol%. For 7 mol% Yb₂O₃, it increased with sintering time from 86% (2 h) to 97% (12 h), reaching nearly complete densification. Once the phase transformation is completed, the further densification is due to the combination of Ostwald ripening and coalescence via the solution–precipitation process. Moreover, the embedded condition did not affect the density when Yb₂O₃ addition was ≽5 mol% but significantly increased it at <5 mol%. The embedded condition allowed the samples with 1–2 mol% Yb₂O₃ to achieve relative densities of >95% after holding for 12 h.

During the sintering, the sample interacts with the reduced atmosphere and thus usually exhibits either a porous or dense outer surface layer, depending on the processing parameters. A direct and simple approach to understanding this feature is to measure the density of samples before and after surface grinding.

Before being ground into disk specimens with 10 mm diameter and 3 mm thickness for thermal conductivity measurements, the sintered samples had diameters of 14.4–15.5 mm and thicknesses of 5.0–6.2 mm. The grinding allowed the samples to achieve relative densities of >97% and even 100% with the addition of <3 mol% Yb₂O₃ (figure 5(a)); for example, the sample with 2 mol% Yb₂O₃ sintered under SPC-I showed a large increase in relative density from ∼88 to ∼100%. This result is attributed to the removal of open pores by grinding (figure 5(b)); this indicates that these samples had at least two layers—a dense inner layer and a porous surface layer. In the case of 3 mol% Yb₂O₃ addition, the grinding led to an increase in relative density from 73 to 77% under SPC-I but a decrease from 87 to 80% under SPC-II. However, at 7 mol% Yb₂O₃, grinding led to a very slight decrease in relative density (∼97%, figure 5(a)) regardless of the packing condition, suggesting the formation of a dense outer surface layer.

Figure 5.

Effect of surface grinding on (a) relative density and (b) porosity of the samples sintered at 1900 °C (holding time 12 h) under different SPCs. Before being ground into disks with diameter of 10 mm and thickness of 3 mm, the sintered samples had diameters of 14.4–15.5 mm and thicknesses of 5.0–6.2 mm.

After cutting, the cross sections of all samples exhibited an inhomogeneous color. This inhomogeneity depended on the Yb₂O₃ content, sintering time and sample packing condition, as shown in figure 6. With the addition of 1 mol% Yb₂O₃, the bulk color changed from dark gray to light gray and then to white from inside to outside. The embedded condition led to an increase in the thickness of the dark gray inner layer from ∼58 to ∼78% of the sample thickness. With the addition of 7 mol% Yb₂O₃, the bulk color changed from inside to outside in the sequence very dark gray, gray, white, gray. The embedded condition led to an increase in the thickness of the very dark gray inner layer from ∼26 to ∼48% of the sample thickness. If we assume that the very dark gray layer and gray layer constitute the whole inner region, then the thickness of the inner layer increased from ∼68 to ∼75% of the sample thickness in the SPC-II sample. The largest thickness was about 500 μm for the white surface layer and ∼200 μm for the outermost gray surface layer, which corresponded to ∼10 and ∼4% of the sample thickness, respectively.

Figure 6.

Optical photographs of polished cross sections of the samples with (a, c) 1 mol% and (b, d) 7 mol% Yb₂O₃, sintered at 1900 °C (holding time 12 h) under SPC-I (a, c) and SPC-II (b, d).

The gray color in Si₃N₄ ceramics is primarily associated with the formation of free Si at the grain boundaries during sintering [30, 31]. The following three reactions in the oxynitride liquid have been proposed to account for the formation of free Si in gas-pressure-sintered Si₃N₄ [31]:

The formed SiO species will not only exist in the gaseous phase but are also dissolved in the oxynitride liquid. The formation of free Si in gas-pressure-sintered Si₃N₄ was demonstrated by the weight increase during the second stage of sintering under a nitrogen pressure of 5 MPa when a BN crucible was used [29]. The gray color becomes darker with increasing amount of free Si. The stability of free Si formed in the oxynitride liquid upon cooling depends on the diffusion rate of nitrogen through the bulk Si₃N₄ ceramic. In the open-pore region, the diffusion path is sufficiently short for nitrogen to bond with Si or directly with SiO to form Si₃N₄, thus resulting in a white color. Density data and SEM observation confirmed that the dark gray layer was the dense region, whereas the white surface layer corresponded to the porous region.

Crystalline secondary phases

No α-Si₃N₄ phase was detected by XRD in all samples sintered at 1900 °C, indicating a complete phase transformation. In addition to the major β-Si₃N₄ phase, crystalline secondary phases were observed in all samples. The identification results are summarized in table 2. In the case of 2 h holding under SPC-I, only Yb₄Si₂O₇N₂ (among the possible Y-Si phases) was observed for 0.5 and 1 mol% Yb₂O₃, and both Yb₄Si₂O₇N₂ and Yb₂Si₂O₇ (Yb₂O₃:SiO₂=1:2) were detected at >2 mol% Yb₂O₃, with Yb₄Si₂O₇N₂ being predominant. More Yb₄Si₂O₇N₂ was formed at a higher Yb₂O₃ content. In the case of 12 h holding under SPC-I, both the Yb₄Si₂O₇N₂ and Yb₂Si₂O₇ phases were observed for 2–5 mol% Yb₂O₃, and the Yb₄Si₂O₇N₂ phase was again dominant. At 3 and 7 mol% Yb₂O₃, in addition to the Yb₄Si₂O₇N₂ and Yb₂Si₂O₇ phases, the Yb₂Si₃O₅N₂ phase appeared, and it dominated the Yb₄Si₂O₇N₂ phase at 7 mol% of Yb₂O₃. In the case of 12 h holding under SPC-II, only the Yb₄Si₂O₇N₂ phase was observed for 0.5 and 1 mol% Yb₂O₃, and both the Yb₄Si₂O₇N₂ and Yb₂Si₂O₇ phases were present at ≽2 mol% Yb₂O₃, similar to the case of 2 h holding under SPC-I.

Table 2.

XRD identification results of crystalline secondary phases in the samples sintered at 1900 °C (holding times 2 and 12 h) under different sample packing conditions (SPCs).

Yb2O3 content	Crystalline secondary phases (intensity∗)
(mol%)	SPC-I, 1900 °C–2 h	SPC-I, 1900 °C–12 h	SPC-II, 1900 °C–12 h
0.5	Yb4Si2O7N2 (×)	Yb4Si2O7N2 (×),Yb2Si2O7 ((×))	Yb4Si2O7N2 (×)
1	Yb4Si2O7N2 (×)	Yb4Si2O7N2 (×), Yb2Si2O7 ((×))	Yb4Si2O7N2 (×)
2	Yb4Si2O7N2 (×), Yb2Si2O7 ((×))	Yb4Si2O7N2 (××), Yb2Si2O7 (×)	Yb4Si2O7N2 (××), Yb2Si2O7 ((×))
3	Yb4Si2O7N2(×××), Yb2Si2O7 (×)	Yb4Si2O7N2(×××), Yb2Si3O5N2 (××)	Yb4Si2O7N2(×), Yb2Si2O7 (×)
		Yb2Si2O7(×)
5	Yb4Si2O7N2(×××), Yb2Si2O7 (×)	Yb4Si2O7N2(×××), Yb2Si2O7 (××)	Yb4Si2O7N2 (××), Yb2Si2O7 (××)
7	Yb4Si2O7N2 (×××), Yb2Si2O7 (×)	Yb4Si2O7N2 (×××), Yb2Si3O5N2 (×××)	Yb4Si2O7N2 (×××), Yb2Si2O7 (××)
		Yb2Si2O7 (×)

^∗××× = major content, ×× = small content, × smaller content, (×) = traces.

It is worth mentioning that the valence of Yb is +3 in Yb₄Si₂O₇N₂ and +2 in Yb₂Si₃O₅N₂. The formation of Yb₄Si₂O₇N₂ is attributed to the following reaction:

Wills et al [32] reported the production of Yb₂Si₃O₅N₂ in reactions between Si₃N₄ and lanthanide oxides. Tanaka et al [33] also reported Yb₂Si₃O₅N₂ formation in MgSiN₂ hot-pressed with Yb₂O₃ and Si₃N₄ and explained it by the following reaction between Si₃N₄ and Yb₂O₃:

Although no YbN phase was detected by XRD, the formation of Yb₂Si₃O₅N₂ in this study probably results from reaction (8). However, it is unclear why Yb₂Si₃O₅N₂ was formed only under SPC-I.

Yang et al [24] also observed only Yb₄Si₂O₇N₂ among the possible yttrium-silicon phases in porous Si₃N₄ sintered with 1.0–7.5 wt% Yb₂O₃. Park et al [21] reported a transformation from Yb₂Si₂O₇ to Yb₄Si₂O₇N₂ in hot-pressed Si₃N₄ with increasing Yb₂O₃ content from 4 to 8 wt%; however, only Yb₄Si₂O₇N₂ was detected at a higher Yb₂O₃ content. Our results indicate that the formation of the Yb₄Si₂O₇N₂ phase is independent of the Yb₂O₃ content upon sintering at 1900 °C. We interpret this observation in terms of the evaporation of native SiO₂ from the surface of Si₃N₄ particles, which is most likely to shift the starting composition to within the Si₃N₄–Yb₂SiO₅–Yb₄Si₂O₇N₂ compatibility triangle. During the rapid sintering at 1800 °C, the minimum evaporation of SiO₂ allows the formation of crystalline secondary phases in accordance with the compatibility triangle.

Microstructure

Figure 7 reveals the typical three-layer structure formed in the sintered samples with 7 mol% Yb₂O₃. The structure is characterized by a dense inner region, followed by a porous surface layer of ∼500 μmthickness and an outermost dense surface layer of ∼200 μm thickness. This layering correlates with the inhomogeneous coloration presented in figure 6.

Figure 7.

SEM micrographs showing the typical three-layer structure in the sintered sample with 7 mol% Yb₂O₃ regardless of the SPC.

As shown in figure 8, in the case of 2 h holding under SPC-I, the samples with 1 and 3 mol% Yb₂O₃ exhibited the finest and coarsest microstructures, respectively. Almost all large grains were elongated in samples with 1 and 7 mol% Yb₂O₃; the elongated grains have a similar width of ∼4 μm and length of ∼40 μm in these two samples. The structure of the 1 mol% Yb₂O₃ sample was most bimodal, that is, it had the largest size difference between the fine matrix and the large elongated grains.

Figure 8.

SEM micrographs of fracture surfaces of the samples with (a) 1, (b) 3 and (d) 7 mol% Yb₂O₃ sintered at 1900 ^oC (holding time 2 h) under SPC- I.

As shown in figure 9, in the case of 12 h holding, the microstructure was unaffected by the sample packing condition but was affected by the Yb₂O₃ content. The samples with 1 and 7 mol% Yb₂O₃ were dense, whereas the 3 mol% Yb₂O₃ material was porous; the 1 mol% Yb₂O₃ sample had the coarsest and the most bimodal microstructure, whereas the 7 mol% Yb₂O₃ sample exhibited the finest microstructure. In the sample with 1 mol% Yb₂O₃, the largest elongated grains had a width of about 17 μm and length ∼100 μm. These results illustrate that the abnormal grain growth of β-Si₃N₄ was significantly promoted by the addition of 1 mol% Yb₂O₃ but hindered by the addition of 7 mol% Yb₂O₃. With the addition of 3 mol% Yb₂O₃, the β-Si₃N₄ grains grew in a normal manner. This is similar to the effect of MgSiN₂ addition on the microstructure development of Si₃N₄ when Y₂O₃ is used as an additive [34].

Figure 9.

SEM photographs of fracture surfaces of the samples with (a, d) 1, (b, e) 3 and (c, f) 7 mol% Yb₂O₃ sintered at 1900 ^oC (holding time 12 h) under SPC-I (a–c) and SPC-II (d–f).

The above results indicate that after the α–β phase transformation is complete, the grain growth of β-Si₃N₄ is determined by the amount and size of the large elongated β-Si₃N₄ grains, specifically, by the difference in size between the large β-Si₃N₄ grains and the matrix grains [35, 36]. If the difference exceeds a critical value, then abnormal grain growth occurs; normal grain growth is observed otherwise. The exaggerated grain growth in the samples with <3 mol% Yb₂O₃ compared with samples having >3 mol% Yb₂O₃ is due to the enhanced bimodal microstructure that developed in the early stage of sintering. With the addition of 3 mol% Yb₂O₃, normal grain growth dominates, resulting in a strong porous skeleton composed of elongated grains of similar size. The steric impingement effect hinders and even stops the densification when the amount of liquid phase is insufficient [34, 37].

Thermal conductivity

As shown in figure 10(a), the samples sintered under SPC-I had the highest thermal diffusivity of 0.59 cm² s⁻¹ at 0.5 mol% Yb₂O₃. The thermal diffusivity generally decreased with the Yb₂O₃ content, except for the 3 mol% Yb₂O₃ sample which also exhibited the lowest density of <80% (figure 5(a)). In the samples sintered under SPC-II, the thermal diffusivity was almost independent of the Yb₂O₃ content and had a value of ∼0.45 cm² s⁻¹, except for the 3 mol% Yb₂O₃ sample. The effect of the sample packing condition on the thermal diffusivity was dependent on the Yb₂O₃ content. With the addition of <3 mol% Yb₂O₃, the embedded condition led to a large reduction in thermal diffusivity, and this reduction became more significant with decreasing Yb₂O₃ content. However, with the addition of ≽3 mol% Yb₂O₃, the sample packing condition had little effect on the thermal diffusivity.

Figure 10.

(a) Thermal diffusivity, α, and (b) thermal conductivity, κ, as functions of Yb₂O₃ content for the samples sintered at 1900 °C (holding time 12 h) under different SPCs.

As shown in figure 10(b), in the samples sintered under SPC-I the thermal conductivity had the highest values of 115–119 W m⁻¹ K⁻¹ at Yb₂O₃ contents of 0.5–2 mol% and exhibited the lowest value of 67 W m⁻¹ K⁻¹ at 3 mol% Yb₂O₃. When the Yb₂O₃ content was further increased to 7 mol%, the thermal conductivity reached 102 W m⁻¹ K⁻¹. Under SPC-II, at low Yb₂O₃ contents of 0.5–2 mol% the thermal conductivity was significantly reduced to near 100 W m⁻¹ K⁻¹. However, at higher Yb₂O₃ contents of ≽ 3 mol%, the thermal conductivities were close to the values obtained under SPC-I and reached 106 W m⁻¹ K⁻¹ at 7 mol% Yb₂O₃.

Porosity is known to lower the thermal conductivity of ceramics. Only the samples with 1, 2 and 7 mol% Yb₂O₃ were almost fully densified, whereas those with other amounts of Yb₂O₃ were more porous. Considering that most pores are closed and isolated and assuming that the thermal conductivity of pores is zero, the thermal conductivity without pores, κ_c0, can be calculated using the following modified Maxwell equation [38]:

where κ_c is the thermal conductivity of the ceramic sample and V_p is the porosity, given by V_p=1−φ (relative density). The thermal conductivities corrected for porosity are added in figure 10(b). The correction compensated for the decrease at 3 mol% Yb₂O₃, resulting in a more regular dependence.

The thermal conductivity of dense β-Si₃N₄ ceramics is determined by the β-Si₃N₄ grains and grain boundaries [11]. The secondary phases consist of isolated glassy pockets located at the triple grain boundary junctions and the continuous grain boundary film located between the grains. The grain boundary phases have thermal conductivity of as low as ∼1 W m⁻¹ K⁻¹, and thus should lower the overall value. The thermal conductivity is enhanced by the purification of β-Si₃N₄ grains and by lowering the amount of secondary phases. The grain growth allows the dissolution of the less pure small β-Si₃N₄ grains and the precipitation of the purer large β-Si₃N₄ grains through the liquid phase, thus enhancing the thermal conductivity of β-Si₃N₄ grains. Not only the amount, but also the distribution of the secondary phases affects the thermal conductivity—a continuous distribution has a more adverse effect than localization in the form of isolated glass pockets [39, 40]. The thermal conductivity is reduced with the increasing thickness and amount of the grain boundary film [18, 41]. According to the observed microstructure, the decrease in thermal conductivity with increasing Yb₂O₃ content is attributed to the increasing amounts of secondary phases and the grain boundary film, as well as the higher concentration of lattice defects in the β-Si₃N₄ grains, similar to our previous results for MgSiN₂ addition [34]. Because the average grain size is larger than 1 μm in all cases, the effect of grain size on thermal conductivity should be less important in the present work.

As the sample packing condition had no effect on the grain size, the purification of β-Si₃N₄ grains is responsible for the change in thermal conductivity. The results of our previous studies suggest that the embedding condition had a strong effect on the removal of lattice oxygen by reaction (2) and thus on the thermal conductivity of β-Si₃N₄ ceramics [42]. The embedded condition suppresses the evaporation of SiO, which is indicative of decreased weight loss, thereby lowering the thermal conductivity. The sample packing condition did not change the weight loss in the case of ⩽3 mol% Yb₂O addition, but slightly increased the weight loss for samples with >3 mol% Yb₂O₃ (figure 4(a)). This result can be understood from the microstructure: the samples with <3 mol% Yb₂O₃ had a porous outermost surface, making the removal of lattice oxygen by reaction (2) very sensitive to the sample packing condition. SPC-I promoted the removal of lattice oxygen and thus enhanced the thermal conductivity in comparison with SPC-II. However, samples with >3 mol% Yb₂O₃ exhibited a dense outermost surface, making the removal of lattice oxygen less sensitive to the sample packing condition; thus, the thermal conductivity remained unchanged. SPC-II suppressed the evaporation of SiO generated by reaction (2) and thus enhanced the densification, as evidenced by the large increase in the thickness of the dark-gray inner layer (figures 6(a) and (c)). Under SPC-I, the unchanged weight loss is attributed to the nitridation of SiO in the outer porous surface region. The embedded condition hindered the removal of lattice oxygen and thus lowered the thermal conductivity. Nevertheless, the embedded condition was efficient in producing high densities of >95% of the theoretical value and thermal conductivities of ∼100 W m⁻¹ K⁻¹ upon the addition of 1–2 mol% Yb₂O₃. From a commercial viewpoint, the adoption of embedded conditions is recommended for producing dense Si₃N₄ ceramics with high thermal conductivity (>100 W m⁻¹ K⁻¹) via gas pressure sintering, and the recommended Yb₂O₃ content is either 1–2 mol% or ≽7 mol%.

Conclusions

1. Curves of the density of Si₃N₄ ceramics plotted against Yb₂O₃ content exhibited a characteristic ‘N’ shape with a local minimum at 3 mol% and complete densification at lower and higher Yb₂O₃ concentrations.

2. Both the α–β phase transformation and β-Si₃N₄ grain growth were dominated by the diffusion-controlled dissolution–reprecipitation mechanism and were promoted at the early stage of sintering by increasing the Yb₂O₃ content. At the final stage of sintering, the amount and size of large elongated β-Si₃N₄ grains dominated the grain growth, thereby resulting in enhanced abnormal grain growth with decreasing Yb₂O₃ content.

3. The thermal conductivity of pore-free Si₃N₄ decreased with increasing Yb₂O₃ content for non-embedded samples but slightly increased when the samples were entirely embedded in BN powder during the sintering.

4. The sample packing condition had no significant effect on the grain size but a strong effect on the thermal conductivity. Embedding decreased the thermal conductivity from 115–119 W m⁻¹ K⁻¹ to about 100 W m⁻¹ K⁻¹ when 0.5–2 mol% Yb₂O₃ was added, whereas the thermal conductivity was constant at 106 W m⁻¹ K⁻¹ when 7 mol% Yb₂O₃ was added. This is due to the formation of an outer porous surface and an outer dense surface in the former and latter cases, respectively.

5. Our results suggest that dense Si₃N₄ ceramics with high thermal conductivity and good high-temperature mechanical properties can be produced by gas pressure sintering with the addition of only Yb₂O₃.