1. INTRODUCTION
Aluminum nitride (AlN) ceramics are widely used in various applications including
microelectronic and optoelectronic devices and heat dissipation systems owing to their
excellent properties, such as their high thermal conductivity and resistivity [1-4]. Moreover, AlN is well known for its extremely high theoretical thermal conductivity
(approximately 320 W/mK). However, due to limitations in their production technology,
it is not possible to entirely prevent impurities and microstructural defects in polycrystalline
AlN. Thus, while the thermal conductivity of AlN is usually lower than its theoretical
value, it still has high values [5-6], making it an excellent heat transfer component for practical use. In addition,
AlN has a coefficient of thermal expansion (CTE) similar to that of Si [7], an important parameter with respect to semiconductor manufacturing. Moreover, because
of its excellent mechanical properties and high corrosion resistance [8-9], AlN is a reliable protective material for use in machine manufacturing and heavy
industry production.
Recently, various surface techniques, such as physical vapor deposition (PVD) and
plasma spraying [10-12], have emerged for the facile fabrication of AlN thin films. Conversely, sintering
is the most common approach for fabricating AlN bulk materials. However, AlN has a
hexagonal crystal structure with extremely strong covalent bonding, and its self-diffusion
coefficient is low, posing challenges to sintering processes. To obtain sintered AlN
ceramics with high densities, higher sintering temperatures and longer sintering times
are required when employing conventional sintering methods, such as pressureless sintering
and hot-pressing sintering. This results in elevated energy consumption and, consequently,
increased manufacturing costs. Spark plasma sintering (SPS), a relatively new sintering
technology, has several advantages including a high heating rate, short sintering
time, and nonpolluting nature [13-14], which can be used to prepare high-density ceramics at relatively low temperatures
within short periods. During the SPS process, the temperature, pressure, and pulsed
direct current are directly varied. For insulating powders such as AlN, the application
of pulsed direct current on the surfaces of the powder particles can induce instantaneous
discharge within the voids between the particles. This leads to Joule heating of the
particles due to the effects of surface activation and purification induced by discharging,
and resulting in an increase in the degree of densification.
By optimizing the sintering parameters, such as the heating rate, sintering temperature,
and holding time, high-density sintered ceramics can be achieved. In this way, AlN
ceramics with densities exceeding 95 % can be prepared by SPS without the need for
sintering aids [15-16]. Research on AlNbased composites fabricated by SPS has attracted widespread attention
in recent years to develop composite materials suitable for a range of applications.
For example, Wang et al. [17] fabricated AlN-W composites for high-average-power electronic tubes. Gao et al.
[18] reported that AlN-based composites containing Mo exhibited better overall performance
than those without the additives. Also, AlN-SiC composite ceramics were fabricated
by SPS for applications involving electromagnetic waves with wavelengths ranging from
1 kHz to 1 MHz [19]. Other composites such as AlNTiN, AlN-BN, and AlN–TiB2 have also been investigated [20-22].
However, studies on AlN-MgO composites have been limited, and there is a shortage
of reference data on these materials. The synthesis of MgAl2O4 and MgAlON ceramics with spinel structures can be achieved through various sintering
methods using AlN, MgO, and Al2O3 as the raw materials. The resulting ceramics exhibit desirable properties, including
high transparency, good slag corrosion resistance, good mechanical properties, and
moderately high CTE values [23-26]. However, the feasibility of obtaining spinel-structured ceramics using AlN and
MgO as the raw materials is still challenging. Moreover, the performance of AlN-MgO
composites needs to be investigated to assess their practical applicability.
Therefore, in this study, we prepared AlN-MgO composites with different compositions
by SPS using AlN and MgO as the raw materials. The phase compositions and microstructures
of the composites were investigated, and their thermal and mechanical properties were
compared with those of pure AlN and MgO samples sintered under the same conditions.
2. EXPERIMENTAL
2.1 Sample preparation
Commercially obtained pure AlN (1μm, ≥97%, Tokuyama Corporation) and MgO (1 μm, 99.5%,
US Research Nanomaterials, Inc.) powders were used as the raw materials in this study.
Powder mixtures with different weight ratios were ball-milled in ethanol for 18 h
at a rotation speed of 200 rpm using Al2O3 balls of diameter of 6 mm. The ethanol/powders and ethanol/balls ratios were 3:1
and 5:1, respectively. After the ball-milling process, the powders were dried in an
oven at 80 °C for 24 h and then sieved using a stainless steel mesh (200 μm) to prevent
agglomeration. The scanning electron microscopy (SEM) images of the AlN and MgO powders
after ball milling are shown in Fig 1. The AlN powder particles exhibited a spherical shape and had a diameter of approximately
0.5 μm. The MgO powder particles exhibited an irregular block shape of 1-2 μm size.
Before sintering, the powders were placed in a cylindrical graphite die with an inner
diameter of 50 mm. The sintering process was carried out using an SPS apparatus (Dr.
Sinter SPS-825, SPS Syntex Inc.) under a uniaxial pressure of 50 MPa and vacuum pressure
of 20 Pa. The powders were sintered at a temperature of 1600 °C using a heating rate
of 100 °C/min for 20 min. After sintering, the uniaxial pressure was released immediately,
and the sample was allowed to cool to room temperature before being removed from the
vacuum chamber. Based on their compositions, the sintered AlN-MgO composites are denoted
as MgO-20, MgO-40, MgO-60, and MgO-80.
2.2 Sample characterization
The sintered samples were polished to a mirror-like finish and cut before being analyzed.
X-ray diffraction (XRD) analysis (Rigaku Ultima IV, Cu-Kα radiation (λ = 1.54178 Å) operated at 40 kV/30 mA) was performed to analyze the phase compositions
of the samples. The scan range was 2θ = 30° to 80°, and the analysis was performed
using a scan step of 0.02° and scan speed of 2°/min. The microstructures of the sample
were investigated using a transmission electron microscopy (TEM) system (TALOS F200X,
FEI) equipped with energy-dispersive X-ray spectroscopy (EDS). Also, the average grain
sizes of the sintered samples were calculated from TEM images using ImageJ software.
X-ray photoelectron spectroscopy (XPS, AXIS Ultra DLD, Kratos) was used to identify
the chemical states of the samples. The densities (ρ) of the samples were measured using Archimedes’ method. Rod-like sintered samples
with dimensions of 25 × 4 × 3 mm3 were used to determine the CTE values, which were measured with a thermal dilatometer
(DIL-402 C, Netzsch) at 100 °C. The thermal diffusivity (λ) and heat capacity (Cp) values of the sintered samples were measured at the same temperature using a laser
flash measuring system (LFA 467 HyperFlash, Netzsch). The thermal conductivity (κ) values were calculated using the equation κ=ρ>λCp, where ρ is the density, λ is the thermal diffusivity, and Cp is the heat capacity of the materials. The hardness and toughness values of the sintered
samples were measured using a Vickers hardness tester (HM-210, Mitutoyo Corporation)
under a load of 20 N for 15 s. The hardness (Hv) and fracture toughness (KIC) values of the samples were calculated using the following equations:
Hv = 0.1891·F(4a2)-1
KIC = 0.16·(ca-1)-1.5(Hva0.5)
, where F is the load, a is the half-diagonal length of the indentation, and c is
the length from the center of the indentation to the tip of the crack. The final Hv and KIC values listed are the averages of five measurements.
3. RESULTS AND DISCUSSION
The XRD patterns of the sintered samples are shown in Fig 2. Only AlN and MgO phases were detected in the sintered AlN-MgO composites, regardless
of the weight ratio of the raw materials. This indicates that there were no phase
transitions or reactions between MgO and AlN during the sintering process. A diffraction
peak related to Al2O3 was observed in the pattern of the sintered pure AlN ceramic, suggesting the presence
of oxide impurities. Magnified diffraction peaks related to AlN (110) and MgO (200)
are shown in Fig 2 (b) and (c). As can be seen from Fig 2(b), with increasing MgO contents (MgO contents lower than 60 wt%) the AlN (110) peak
of the sintered AlN-MgO composites is shifted to a lower 2θ value compared with the
peak of the pure AlN sample. Similarly, Fig 2(c) shows that the MgO (200) peak of the sintered AlN-MgO composites is shifted to a
higher 2θ value compared with the peak of the pure MgO sample with increasing MgO
contents, (MgO content lower than 80 wt%). These peak shifts may be caused by the
formation of a solid solution phase, indicating that solid solutions were formed in
both the AlN and MgO lattices of the sintered samples with different compositions.
Fig 3 shows the XPS spectra of the sintered AlN-MgO composites with different compositions.
As can be seen from Fig 3(a)–(d), the Mg 2p and O 1s peaks of the sintered AlN-MgO composites exhibit significantly higher bonding energies
(BEs) compared with those of peaks of the sintered pure MgO sample, indicating that
the chemical bonding states between Mg and O were different in the composites. However,
only a minor shift in the BEs was observed for the Al 2p and N 1s peaks. Based on
these results, it can be concluded that a solid solution was primarily formed in the
MgO lattice. Fig 3(e) and (f) show the fitted O 1s peaks of the sintered pure AlN and MgO-60 samples. For the sintered
pure AlN sample, in addition to the peak related to the adsorbed oxygen (BE = 532.1
eV [27]), a peak was also observed at 531.6 eV; this corresponded to the Al-O bond [27]. This confirms the presence of Al2O3 as an impurity in the sample, consistent with the XRD analysis results. With the
sintered MgO-60 sample, the peaks with BE values of 530.8 eV and 531.5 eV could be
ascribed to MgO [28] and magnesium–aluminum composite oxides (MgAlxOy), respectively [27], indicating the formation of a (Mg, Al)O solid solution. The fitted N 1s peaks of
the sintered pure AlN and MgO-40 samples are shown in Fig 3(g) and (h), respectively. In addition to a peak related to the Al-N bond (BE = 396.7 eV [30]), a peak related to the Al-N-O bond was also observed (BE = 396.3 eV) [31] in the spectrum of the sintered MgO-40 samples, indicating the formation of an Al(N,
O) solid solution phase. The oxygen in the AlN solid solution was derived either from
the MgO lattice, the Al2O3 formed on the AlN surface, or the residual oxygen in the sintering chamber. Thus,
the XPS results were consistent with those of the XRD analysis, further confirming
that the different types of solid solutions were formed within the MgO and AlN lattices
with changes in the composition of the sintered AlN-MgO composites.
The calculated relative densities of the sintered samples are summarized in Table 1. Except for pure MgO, all of the sintered samples exhibited relative densities higher
than 95%. As shown in Fig 4(a), SEM surface images showed that the sintered pure MgO sample had a grain size ranging
from 10 μm to more than 100 μm. However, residual pores were present at the grain
junctions and coarse grain boundaries in the sintered pure MgO samples (Fig 4(b) and (c)). It has been reported that MgO ceramics can be sintered readily at temperatures
lower than 1000 °C [32] and MgO undergoes rapid grain growth at sintering temperatures higher than 1200
°C [33]. Hence, it can be concluded that, in this study, the extremely rapid growth of the
MgO grains was induced by the high sintering temperature used during the short sintering
process. This resulted in the formation of pores and the weakened grain boundaries,
leading to a low relative density in the sintered pure MgO sample.
The detailed microstructures of the sintered samples were analyzed using TEM. Fig 5 shows the low-magnification cross-sectional TEM images of the sintered pure MgO,
AlN, and MgO-60 samples. As can be seen from Fig 5(a), no grain boundaries were observed in the sintered TEM sample of pure MgO, indicating
that it had an extremely large grain size (from 10 μm to more than 100 μm), as observed
in Fig 4 (a). Since the average size of the pristine MgO powder particles was approximately 1
μm, the MgO grains grew rapidly during the sintering. Similarly, significant grain
growth was observed with the sintered pure AlN sample, which also showed a significantly
larger grain size (2–3 μm) after sintering (Fig 5(b)).
However, as shown in Fig 5(c), the MgO-60 sample exhibited a finer-grained microstructure. The MgO and AlN grains
of the sample could be distinguished readily based on the results of EDS mapping results
(Fig 5 (d) and (e)): the grain sizes were determined to be 1.0–1.5 μm and 0.5–1.0 μm, respectively.
The average grain sizes of the MgO and AlN in the AlN-MgO composites were determined
from their cross-sectional TEM images, and the results are shown in Fig 6. Neither the MgO and AlN grains showed significant growth during the sintering process,
probably because MgO and AlN can be regarded as secondary phases with respect to each
other. Therefore, it is considered that their grain growth was mutually inhibited
during sintering. In addition, except for the MgO-80 sample, the size of the MgO grains
in the composites decreased with increasing MgO content, while the variations in the
size difference between the MgO and AlN grains was the same as that seen in the MgO
grains. This suggests that the inhibition of grain growth was markedly more pronounced
in the MgO-60 sample, which had a relatively smaller grain size. Relatively larger
differences between the MgO and AlN grains were present in the MgO-80 sample, indicating
that the growth of the MgO grains was impeded with a low AlN content.
The state of grain boundaries has a critical effect on the characteristics of sintered
materials. Two types of grain boundaries were observed in the sintered composites,
and their TEM images are shown in Fig 7. Fig 7 (a) shows a low-density grain boundary in the sintered MgO-20 sample. The corresponding
EDS maps are presented in Fig 7(b) and (c), and show that the low-density grain boundaries were composed of a great part of
oxygen and the low Mg content. Hence, the oxygen-rich interface was more likely attributed
to the residual oxygen remaining between the grains, rather than to the diffusion
of MgO. Fig 7(d) and (e) show TEM images of the grain boundary observed in the sintered MgO-60 sample. A clean
interface was formed between the MgO and AlN grains without the aggregation of oxygen.
The inverse fast Fourier transform (IFFT) result for the marked area, shown in Fig 7(f), reveals an interplanar spacing of 0.473 nm, corresponding to the MgAl2O4 (111) plane. MgAl2O4 is a spinel phase with a cubic structure and is formed by a reaction between MgO
and Al2O3 under certain conditions [34]. Therefore, the formation of the MgAl2O4 spinel phase at the clean interface can probably be attributed to a reaction between
MgO and the Al2O3 layer from the AlN surface.
However, given the presence of AlN grains, it is likely that a MgAlON phase was also
formed at the MgO and AlN grain boundaries. MgAlON has the same structure as the MgAl2O4 spinel phase, and is considered to be a solid solution formed by MgO/MgAl2O4 when they enter the γ-AlON lattice, or because of the dissolution of Al2O3 and AlN in MgAl2O4 [35]. Dai et al. [36] also reported that a MgAlON phase is formed during SPS at 1700 °C when MgO, AlN,
and Al2O3 were used as the raw materials. Therefore, it is likely that, in this study, MgAlON
was formed at the grain boundaries of MgO and AlN. Thus, despite the difficulty in
distinguishing MgAl2O4 from MgAlON, it can be concluded that a spinel phase was formed at the clean interface
between the MgO and AlN grains.
Fig 8 shows the measured thermal conductivities and CTE values of the sintered samples.
The corresponding thermal diffusivity and heat capacity values are listed in Table 2. Sintered pure AlN exhibited the highest thermal conductivity (53 W/mK) of all the
samples, followed by sintered pure MgO (39.7 W/mK). The AlN-MgO composites showed
low thermal conductivities, which varied from 14.9 to 33.3 W/mK. While AlN ceramic
typically shows high thermal conductivity, the AlN ceramic prepared by SPS in this
study exhibited a thermal conductivity of less than 55 W/mK. This could be attributed
to its fine-grained microstructure, meaning it contained more grain boundaries compared
with single-crystal AlN. Grain boundaries are a type of planar defect that scatter
phonons, thereby decreasing thermal conductivity [37].
In addition, oxygen-related defects, including the residual oxygen present at the
interface, the oxide layer on the AlN surface, and the lattice defects induced by
oxygen (substitutions and vacancies), contribute to the reduction of the thermal conductivity
through phonon scattering [38-39]. As a result, the AlN ceramic did not exhibit high thermal conductivity in this
study. The thermal conductivity of the AlN-MgO composites strongly depended on the
MgO/AlN mass ratio and decreased with increasing MgO content. Because MgO has lower
thermal conductivity than AlN, the AlN-MgO composites showed reduced thermal conductivities.
A previous analysis showed that AlN-MgO composites exhibited a fine-grained microstructure,
with solid solutions and an oxygen-rich interface that formed between the grains.
All these factors enhanced phonon scattering and thus contributed to the lowering
of the thermal conductivity of the AlN-MgO composites. Therefore, the low thermal
conductivity of the AlN-MgO composites was a result of both their composition and
microstructure. Considering that the thermal conductivity of the MgO-20 sample was
lower than that of pure MgO, the effect of the microstructure on the thermal conductivity
of the composites was more pronounced in this study.
In the case of the CTE, the pure AlN ceramic showed the lowest CTE value at 4.47 ×
10-6 /K, whereas the pure MgO sample exhibited the highest value, which was 13.06 × 10-6 /K. In general, the microstructure also influences the CTE value, in addition to
the composition. For instance, for the same material system, a sample with finer grains
generally shows a higher CTE value than that of a sample with a coarser microstructure.
This is because the CTE of the grain boundaries is significantly greater than that
of the grains themselves [40]. In the present study, the CTE of the AlN-MgO composites increased from approximately
7.20 ×10-6 /K to 10.73 ×10-6 /K with increasing MgO content. However, the opposite was observed in the case of
the MgO-40 sample, which showed a slightly lower CTE (approximately 6.49 ×10-6 /K) than that of the MgO-20 sample (approximately 7.20 × 10-6 /K). Since the relative densities and microstructures of the sintered composites
were similar, the exact reason for the abnormal results observed in the case of MgO-20
and MgO-40 remains unclear and requires further study. Overall, however, the effect
of the composition on the CTE was more significant than that of microstructure.
The mechanical properties of the sintered ceramics were evaluated using the Vickers
hardness test. Fig 9 shows the calculated hardness and fracture toughness values of the sintered pure
AlN, MgO, and AlN-MgO composites. The pure AlN sample exhibited the highest hardness
value at approximately 14 GPa. In contrast, the MgO sample exhibited an extremely
low hardness at approximately 4 GPa. It was previously reported that MgO ceramics
prepared by SPS exhibit a wide range of hardness values, from 8–13 GPa, based on the
microstructure of the material [41]. Therefore, the low hardness of the MgO ceramic fabricated in this study may have
been caused by its coarse grains. The hardness values of the AlN-MgO composites decreased
with increasing MgO content; however, the opposite was observed in the case of the
MgO-60 sample, which exhibited the highest hardness (12.86 GPa) of all the composites,
close to that of the pure AlN sample. Thus, the hardness of the AlN-MgO composites
depended on factors other than their composition.
The pure AlN sample also exhibited a low fracture toughness of approximately 1.9 MPa·m0.5. With increasing MgO content, the toughness of the AlN-MgO composites first increased,
reaching the maximum at a MgO content of 60 wt. %, and then decreased. The indentation
morphology of the sintered pure MgO sample was irregular (see Fig 10), which made determining its fracture toughness difficult. The AlN and MgO-60 samples
exhibited clear Vickers indentations, with cracks initiating from the diagonal ends
of the indentations. In the case of the sintered pure MgO sample, however, extensive
cracks were observed originating from the edges of the indentations, indicating poor
crack-extension resistance. In Fig 10 (d) it can be seen that the cracks have mainly propagated along the grain boundaries
of the sintered pure MgO sample, suggesting the grain boundary cohesion of MgO was
weak and resulted in the intergranular fracture. These results indicate the large
grain size and coarse grain boundaries of the sintered pure MgO sample were responsible
for its poor mechanical properties. Therefore, the primary hardening and toughening
mechanisms for the MgO-60 sample are grain refinement and strengthening. In addition,
the formation of a solid solution also helps improve mechanical properties. In summary,
the MgO-60 sample exhibited the best overall mechanical properties in this study.
5. CONCLUSIONS
In this study, AlN-MgO composites with different compositions were fabricated using
SPS. The effects of the composition on their structure, thermal properties, and mechanical
properties were investigated. Pure AlN and MgO were also sintered under the same conditions
for reference. The results indicated that the AlN-MgO composites were composed of
AlN and MgO phases, and different solid solutions were formed within the MgO and AlN
lattices. Compared with the sintered pure samples, the AlN-MgO composites showed finer
grain microstructures owing to the inhibition of grain growth between the two phases.
TEM analysis indicated that two types of grain boundaries were formed in the composites:
oxygen-rich low-density grain boundaries and clean boundaries with a spinel phase.
The sintered pure AlN sample showed the highest thermal conductivity of 53 W/mK and
lowest CTE of 4.47 × 10-6 /K at 100 °C. Conversely, the thermal conductivity and CTE of the AlN-MgO composites
decreased from 33.3 to 14.9 W/mK and generally increased from 6.49 ×10-6 to 10.73 ×10-6 /K with increasing MgO content. The composite with an MgO content of 60 wt. % exhibited
the best overall mechanical properties, including a hardness of 12.86 GPa and toughness
of 3.15 MPa·m0.5, attributed to its fine microstructure. It also demonstrated a thermal conductivity
of 16.8 W/mK and a CTE of 8.14 ×10-6 /K. Thus, the variations in the thermal and mechanical properties of the AlN-MgO
composites can be attributed to the combined effect of their composition and microstructure.