- Email: [email protected]
Crystal growth and evaluation of nitrogen and aluminum co-doped N-type 4H-SiC grown by physical vapor transport
Accepted Manuscript Crystal Growth and Evaluation of Nitrogen and Aluminum Co-Doped N-type 4H-SiC Grown by Physical Vapor Transport H. Suo, K. Eto, T. Ise, Y. Tokuda, H. Osawa, H. Tsuchida, T. Kato, H. Okumura PII: DOI: Reference:
S0022-0248(18)30293-8 https://doi.org/10.1016/j.jcrysgro.2018.06.019 CRYS 24641
To appear in:
Journal of Crystal Growth
Received Date: Revised Date: Accepted Date:
28 February 2018 18 June 2018 19 June 2018
Please cite this article as: H. Suo, K. Eto, T. Ise, Y. Tokuda, H. Osawa, H. Tsuchida, T. Kato, H. Okumura, Crystal Growth and Evaluation of Nitrogen and Aluminum Co-Doped N-type 4H-SiC Grown by Physical Vapor Transport, Journal of Crystal Growth (2018), doi: https://doi.org/10.1016/j.jcrysgro.2018.06.019
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Crystal Growth and Evaluation of Nitrogen and Aluminum Co-Doped N-type 4H-SiC Grown by Physical Vapor Transport H. Suo1, 2, K. Eto1, T. Ise1, 3, Y. Tokuda1, 4, H. Osawa2, H. Tsuchida5, T. Kato1, and H. Okumura1 1
National Institute of Advanced Industrial Science and Technology (AIST), Onogawa 16-1, Tsukuba, Ibaraki 305-8569, Japan 2
3
4
5
Abstract
Showa Denko K.K., 1-13-9, Shiba Daimon, Minato-ku, Tokyo 105-8518, Japan
Asahi Diamond Industrial Corporation, 787, Tabi, Ichihara, Chiba, 290-0515, Japan
DENSO CORPORATION, 500-1, Komenokichominamiyama, Nissin, Aichi, 407-0111 Japan
Central Research Institute of Electric Power Industry (CRIEPI), 2-6-1, Nagasaka, Yokosuka, Kanagawa, 240-0196, Japan
N-type 4H-SiC crystals were grown by the physical vapor transport (PVT) method with nitrogen
and aluminum (N–Al) co-doping. By using aluminum carbide powder preannealed in nitrogen gas atmosphere as an aluminum doping source, we obtained highly N–Al co-doped crystals with a nitrogen concentration higher than that in nitrogen-only-doped crystals. The dislocation densities of N-Al co-doped crystals with a high aluminum concentration (> 1 × 1019 cm-3) were found to become higher than those with a low aluminum concentration (< 1 × 1019 cm-3). Moreover, we investigated the expansion velocities of double Shockley-type stacking faults (DSFs) in the N–Al co-doped and the nitrogen-only-doped crystals. We found that the DSF expansion velocities in the N–Al co-doped crystals were lower than those in the nitrogen-only-doped crystals. This difference in the DSF expansion velocity is discussed with respect to the quantum well action model.
Keyword A1. Defect, A1. Doping, A2. Growth from vapor, B2. Semiconducting silicon compounds
1. Introduction Silicon carbide (SiC) has excellent physical and chemical properties as a semiconductor for use in high-power electronics. The strong chemical bonding between silicon and carbon atoms provides this material with a wide bandgap and high critical (breakdown) electric field strength. Recently, many devices for high-power electronics have been demonstrated. In the case of SiC metal-oxide-semiconductor field-effect transistors (MOSFETs) with a breakdown voltage of 600 V class, the specific on-resistances of the devices have reached less than 1 mΩcm2 [1]. However, further on-resistance reduction in such devices is desired. The resistivity
(15–25 mΩcm) of commercially available n-type 4H-SiC substrates cannot be ignored as one of the resistance components in low-on-resistance devices. There is significant demand for SiC substrates with a low-resistivity (< 10 mΩcm), which will reduce the on-resistance of SiC high-power devices. In SiC, nitrogen (N) is used as a donor for n-type doping, whereas aluminium (Al) and boron (B) are employed as acceptors for p-type doping. The concentration of doped nitrogen can be controlled by adjusting the nitrogen partial pressure and/or gas flow rate during the growth of SiC crystals by physical vapor transport (PVT). Several groups have reported on the growth of low-resistivity n-type 4H-SiC bulk crystals [2–8], and the nitrogen concentration and low-resistivity have reached 1.3 × 1020 cm-3 and 1.5 mΩcm, respectively. When the nitrogen concentration exceeds 1–3 × 1019 cm-3, n-type 4H-SiC crystals often degrade owing to the formation of double Shockley-type stacking faults (DSFs) [2–4, 7–11]. Nuclei of DSFs can be introduced by surface damage, for example scratches [11]. The DSFs expand from such nuclei during postgrowth hightemperature treatment in highly nitrogen-doped crystals, which increases the resistivity of the crystals [12]. Thus, DSF formation is a major issue for the practical use of low-resistivity 4H-SiC substrates for high-power devices. The expansion of DSFs is considered to occur when the system free energy of the crystal is reduced by DSF formation in a perfect 4H-SiC crystal. The quantum well action (QWA) model has been proposed to describe the difference in the electronic energy between perfect and DSF-faulted 4H-SiC crystals, especially in crystals with high nitrogen doping concentrations [13–15]. The electronic energy gain ∆ is defined as the difference in the total electron energy between the perfect and DSF-faulted 4H-SiC crystals. The energy state of electrons in a DSF is expressed as a two-dimensional electron gas in a quantum well, and has been calculated to be 0.68 eV below the conduction band minimum [16]. The Fermi energy can become comparable to or exceed the DSF energy state in heavily nitrogen-doped crystals with high donor concentrations. In this case, the electronic energy of the system decreases with the fall of many free electrons in the conduction band to the lower DSF energy states. On the other hand, it has been reported that the DSF formation energy, which is related to the structural strain energy, is greater than zero [17]. When the electronic energy gain ∆ overcomes the increase in the structural strain energy, DSF expansion is induced to reduce the system free energy. It has been reported that the expansion velocity of single Shockley-type stacking faults (SSFs) depends on their electronic energy [18], and it is also considered that the expansion velocity of DSFs is affected by the
electronic energy gain ∆. Although DSF expansion has been observed in lowly nitrogen-doped (< 1 × 1019 cm3
) crystals with a lower electronic energy gain ∆ under external stress [19], such expansion has been observed
in highly nitrogen-doped crystals (> 2 × 1019 cm-3) with a higher electronic energy gain ∆, even without external stress [20, 21]. Recently our group has reported on the growth of high-nitrogen-concentration and low-resistivity (<10 mΩcm) 4H-SiC crystals under nitrogen and aluminum (N–Al) co-doping conditions [22–24]. Despite the high-nitrogen-concentration and low-resistivity of the grown crystals, DSF generation was suppressed by N– Al co-doping in both chemical vapor deposition (CVD) [22] and PVT growth [23, 24]. It has been calculated that N–Al co-doping results in a decrease in the electronic energy gain ∆ [15], and the DSF suppression observed in our aforementioned studies was presumably attributed to this decrease. In addition, low-resistivity nitrogen and boron (N–B) co-doped crystals were grown by the PVT method, and DSF generation in the asgrown state was also suppressed by N–B co-doping [21]. Furthermore, a marked difference in the DSF expansion velocity between N–B co-doped and nitrogen-only-doped crystals during postgrowth hightemperature treatment was confirmed [21]. Not only DSFs, but also other extended defects in bulk crystals can affect device performance. Threading screw dislocations (TSDs), threading edge dislocations (TEDs), and basal plane dislocations (BPDs) are known as major extended defects in bulk crystals [1]. However, little experimental information is available on the relationship between dislocation density and impurity concentration in N–Al co-doped crystals. It was reported that, in CVD growth, an increase in the threading dislocation density is caused by stress relaxation associated with a high aluminum concentration of 3.8 × 1020 cm-3 [25]. In this study, we investigated the relationship between the impurity concentration and heating condition of aluminum carbide (Al4C3) powder that had been preannealed in a nitrogen atmosphere and the relationship between the impurity concentration and dislocation density in grown crystals. We also investigated the velocity of DSF expansion from nuclei introduced by indentation on the surface of N–Al co-doped and nitrogen-only-doped crystals. Finally, on the basis of differences in the DSF expansion velocity of various free electron concentrations, we discussed the effect of N–Al co-doping on DSF expansion with respect to the QWA model.
2. Experimental Crystals were grown using the two-zone heating furnace with induction heating coils to independently heat the SiC source powder and Al4C3 powder [23, 24]. The two-zone heating furnace was described in detail in Refs. _
23 and 24. Three-inch 4° off-axis (0001) C-face 4H-SiC substrates were used as seed crystals. The growth pressure was 5 Torr. Nitrogen and aluminum co-doping was performed by introducing nitrogen gas (N2) and heating the Al4C3 powder. The flow ratio of N2 to argon [N2 / (Ar + N2)] was varied in the range of 5–100% to obtain crystals with different nitrogen concentrations. To investigate the relationship between the aluminum concentration and Al4C3 temperature, the Al4C3 temperature was varied in the range of 1600–1850°C, which was maintained 400–600°C lower than the seed temperature because of the high vapor pressure of Al4C3. Before crystal growth, the Al4C3 powder had been annealed for more than 50 hours in N2 atmosphere, at over 1500°C, for stable sublimation and nitridation of the Al4C3 powder. In general, powder with a large specific surface area (high surface free energy) can easily undergo sublimation. It was reported that the specific surface area of titanium dioxide power is decreased by annealing [26]. Similarly, it is presumed that the specific surface area of Al4C3 powder is decreased by preannealing. Therefore, the stable sublimation of preannealed Al4C3 powder is considered to be caused by decrease in the specific surface area. The composition of the preannealed Al4C3 was analyzed using the X-ray diffraction technique. Four-degree offaxis (0001) substrates were obtained from the grown crystals after wire-saw cutting, grinding, and polishing _
with diamond slurry, and (1100) cross-sectional samples were also obtained from the grown crystals. The nitrogen and aluminum concentrations in the grown crystals were measured by secondary ion mass spectrometry (SIMS). To evaluate their electrical properties, Hall effect measurement with van der Pauw geometry was performed at room temperature, using the 4° off-axis (0001) substrates. As it is difficult to evaluate the dislocation density of highly nitrogen-doped 4° off-axis (0001) substrates (including N–Al codoped crystals) using the conventional molten potassium hydroxide (KOH) etching technique, chemical wet etching in molten KOH + sodium peroxide (Na2O2) [27] at 500°C for 2–3 min was used to evaluate the dislocation densities. The propagation of a dislocation from a seed crystal was investigated by transmission Xray topography (XRT) with 0004 diffractions using a molybdenum Kα1 beam (0.70926 Å). The TSD density was estimated from the number of propagating TSDs in the evaluated area. As it has been reported that carbon
inclusions generate TSDs [6], the carbon inclusion density was also evaluated using a transmission optical microscope. _
To introduce DSF nuclei, indentations were formed on the (1100) surface of cross-sectional samples using a microhardness tester with a diamond indenter. The indentation force was 100 mN. Then, the samples were annealed at 1000°C for 2 hours in an argon atmosphere, to investigate DSF expansion. Molten KOH etching was carried out to observe line-shaped etch pits, which correspond to DSF formation regions. The DSF expansion length was evaluated in terms of the etch-pit length from the indentation point, and the DSF expansion velocity was estimated by dividing the etch-pit length by the annealing time. The DSF expansion velocity was assumed to represent the dislocation velocity of leading partials of the expanding DSFs.
3. Results and discussion 3.1
Crystal growth and dislocation density of N–Al co-doped crystals
We investigated the relationship between the impurity concentration and heating conditions of the preannealed Al4C3 powder. Figure 1 shows the relationship between the nitrogen concentration at nonfaceted regions in the crystal and the N2 to argon flow ratio. In the case of N–Al co-doping, the heating temperature of the preannealed Al4C3 powder was 1750 or 1630°C. The nitrogen-only-doped crystals were grown without Al4C3 powder. The nitrogen concentration increased with increasing the N2 to argon flow ratio, under the three conditions with (1750 or 1630°C) or without using Al4C3 powder. The nitrogen concentrations in the N–Al codoped crystals with the preannealed Al4C3 powder were higher than those in the nitrogen-only-doped crystals at all flow rates. In addition, the nitrogen concentrations increased with increasing Al4C3 temperature. At a growth pressure of 5 Torr, whereas the nitrogen concentrations of nonfaceted regions in the nitrogen-onlydoped crystal did not exceed 3.5 × 1019 cm-3, the nitrogen concentration reached 4.8 × 1019 cm-3 in the N–Al co-doped crystal with the preannealed Al4C3 powder. Nitrogen doping of SiC crystals during PVT growth has been studied by several groups [5–8]. The nitrogen concentration is roughly expressed in terms of a Langmuir isotherm-type equation as
CN = A
Kp , 1 + Kp
(1)
where CN is the nitrogen concentration, A is a proportionality constant, K is a kinetic parameter related to the nitrogen adsorption and desorption processes, and p is the N2 partial pressure in the growth ambient. The dashed line in Fig. 1 is the best-fit theoretical curve obtained on the basis of Eq. (1) for nitrogen-only-doped crystals. The nitrogen concentration is in good agreement with Eq. (1). The nitrogen concentrations in N–Al co-doped crystals were, however, significantly higher than those in the nitrogen-only-doped crystals. Results of X-ray diffraction analysis of the preannealed Al4C3 powder showed that the powder consisted of Al4C3, aluminum nitride (AlN), Al5C3N, and carbon. We consider that N2 was supplied from the preannealed Al4C3 powder by heating, and the amount of supplied N2 increased with increasing Al4C3 powder temperature. When the N2 partial pressure p is very high, the nitrogen concentration approaches 3.7 × 1019 cm-3, using the aforementioned best-fit theoretical curve. However, the nitrogen concentrations in the N–Al co-doped crystals at an Al4C3 temperature of 1750°C exceeded this nitrogen concentration of 3.7 × 1019 cm-3. Therefore, we conclude that the nitrogen incorporation differs between the N–Al co-doped and nitrogen-only-doped crystals grown by the PVT method, beyond the effect of supplying additional N2 from preannealed Al4C3 in the N–Al co-doped growth. There is a possibility that a low formation energy of an acceptor-donor complex [28, 29] is the cause of the difference in nitrogen concentration between N–Al co-doped and nitrogen-only-doped crystals. Figure 2 shows the relationship between the aluminum concentration in the crystal and the Al4C3 temperature. The dashed line is the exponential line fitted to the experimental data. Only the crystal grown at an Al4C3 temperature of 1820°C was p-type, because the high aluminum concentration was higher than the nitrogen concentration. The aluminum concentration increased exponentially with increasing Al4C3 temperature. Aluminum doping during PVT growth has been studied [30, 31], and the aluminum concentration in a grown crystal was shown to have a linear dependence on the aluminum partial pressure during crystal growth. We consider that the observed exponential increase in aluminum concentration with increasing Al4C3 temperature was caused by an exponential increase in the vapor pressure of aluminum from Al4C3 and AlN [32, 33]. This result is consistent with a previous report on p-type growth [29]. Figure 3 shows the relationship between the logarithm of aluminum concentration and the reciprocal of Al4C3 temperature. The logarithm of aluminum concentration shows a linear dependence on the reciprocal of Al4C3 temperature. Therefore, the aluminum
concentration is considered to be mainly affected by aluminum sublimation. Assuming that the aluminum sublimation and aluminum concentration in a grown crystal are proportional, the activation energy of aluminum sublimation from preannealed Al4C3 powder is calculated as 530 kJ/mol from Fig. 3. This activation energy is higher than the reported activation energies of sublimation, which are approximately 180 kJ/mol for Al4C3 [32] and 420 kJ/mol for AlN [33]. In this experiment, it is possible that non-equilibrium and complicated pathways (e.g. sublimation, transport in the furnace, and incorporation) affected the activation energy. Figure 4 shows a photograph of a grown N–Al crystal, with a 3-inch-diameter seed crystal grown at an Al4C3 temperature of 1730°C. The nitrogen concentration was 5.2 × 1019 cm-3 and the aluminum concentration was 2.0 × 1019 cm-3 at a growth thickness of 6 mm. Moreover, the nitrogen concentration was 5.1 × 1019 cm-3 and the aluminum concentration was 1.8 × 1019 cm-3 at a growth thickness of 14 mm. The nonuniformities of nitrogen and aluminum concentrations for each growth thicknesses were less than 10%. The typical growth rate of a N–Al co-doped crystal was 100–150 µmh-1. Crystals with different aluminum concentrations were grown to evaluate the relationship between the dislocation density and aluminum concentration. Figure 5 shows the molten KOH + Na2O2-etched surface of the 4° off-axis (0001) substrate with (a) a higher aluminum (H-Al) concentration and (b) a lower aluminum (L-Al) concentration. The H-Al substrate grown at an Al4C3 temperature of 1750°C, had an aluminum concentration of 2 × 1019 cm-3 and nitrogen concentration of 5 × 1019 cm-3, whereas the L-Al substrate grown at an Al4C3 temperature of 1660°C had an aluminum concentration of 7 × 1018 cm-3 and a nitrogen concentration of 3 × 1019 cm-3. From the shape of the etch pits, threading dislocations and BPDs can be distinguished. It was difficult to clearly distinguish between TSDs and TEDs from the shape and size of the etch pits. The etch pit density (EPD) is the sum of the TSD, TED, and the BPD densities. The EPD and BPD density were determined by averaging their values evaluated at 10 areas in the substrate. The EPD of the H-Al substrate was 5 × 104 cm-2, whereas that of the L-Al substrate was 1 × 104 cm-2. The BPD densities were 3 × 102 cm-2 in the H-Al substrate and 6 × 102 cm-2 in the L-Al substrate. We found that the threading dislocation density clearly increased in the H-Al crystal. No line-shaped etch pits caused by SFs were observed even in the faceted regions with a higher nitrogen concentration in either crystals.
Figure 6 shows cross-sectional XRT images with 0004 diffractions of a cross-sectional sample of a higher aluminum concentration 2 (H-Al-2) crystal grown at an Al4C3 temperature of 1730°C; the sample had an aluminum concentration of 2 × 1019 cm-3 and a nitrogen concentration of 5 × 1019 cm-3. The thickness of the HAl-2 cross-sectional sample was 300 µm. Dark lines, which correspond to TSDs, can be seen to propagate from the seed crystal to the grown crystal. The TSD density in the H-Al-2 crystal was estimated from the number of TSDs observed in the cross-sectional XRT image. The estimated TSD densities were 8 × 103 cm-2 at a growth thickness of 10 mm and 3 × 103 cm-2 in the vicinity of the grown/seed crystal interface. This significant increase in TSD density was observed only in the H-Al-2 crystal. Figure 7 shows cross-sectional XRT images of a cross-sectional sample of a lower aluminum concentration 2 (L-Al-2) crystal grown at an Al4C3 temperature of 1630°C; the sample had and with an aluminum concentration of 3 × 1018 cm-3 and a nitrogen concentration of 3 × 1019 cm-3. The thickness of the L-Al-2 cross-sectional sample was 215 µm. The estimated TSD densities were 3 × 103 cm-2 at a growth thickness of 10 mm and 3 × 103 cm-2 in the vicinity of the grown/seed crystal interface, with no significant increase in the TSD density observable in this case. The cause of the increase in threading dislocation density in the higher-aluminum-concentration crystals is not yet clearly understood. It has been reported that the dislocation density is increased by stress relaxation at a high aluminum concentration in CVD growth [25]. Similarly, we consider that the increase in threading dislocation density observed here may be related to stress relaxation in higher-aluminum-concentration crystals. Furthermore, it has been reported that TSDs are generated from carbon inclusions [6, 34]. The TSDs generation mechanism has been suggested to involve c-axis lattice misfits that occur when there is overgrowth on carbon inclusions [34]. The carbon inclusion densities of 4–5 × 104 cm-3, observed here, were almost the same in both the H-Al and L-Al-2 cross-sectional samples. The possible mechanism of the TSD density increase is considered that the increasing c-axis strain due to the high aluminum concentration causes lattice misfits during the overgrowth on carbon inclusions.
3.2
Expansion velocity of DSF during postgrowth high-temperature treatment _
Figure 8 shows cross-sectional micrographs of (1 1 00) N–Al co-doped crystal and nitrogen-only-doped crystals after KOH etching. The line-shaped etch pits in Fig. 8 correspond to DSF formation regions.
Expansion of DSFs was confirmed in both types of crystal. The DSF expansion velocities were found to be equivalent on both sides of the indentation points along the basal plane, indicating spontaneous DSF expansion during the postgrowth high-temperature treatment, owing to the lower free energy of the DSFfaulted 4H-SiC. Figure 9 shows the relationship between the DSF expansion velocity and free-electron concentration in the N–Al co-doped at an aluminum concentration of 3 × 1019 cm-3 and nitrogen-only-doped crystals. The DSF expansion velocity increased with increasing free-electron concentration in both types of crystal. Furthermore, the DSF expansion velocities in the N–Al co-doped crystals were lower than those in the nitrogen-only-doped crystals with equivalent free-electron concentrations. A similar decrease in the relative DSF expansion velocity has also been confirmed when comparing N–B co-doped and nitrogen-only-doped crystals [21]. Numerous authors have discussed the formation of DSFs in highly nitrogen-doped 4H-SiC crystals during postgrowth high-temperature treatments [7–15, 19–21]. A large electronic energy gain ∆ indicates instability of the perfect 4H-SiC, and this electronic energy gain ∆ is considered to affect the DSF expansion velocity. It has also been reported that impurities can affect the dislocation velocities and result in the immobilization of dislocations in semiconductor crystals [35–38]. The activation energy of dislocation motion is expected to increase with such immobilization. Therefore, we consider that the low DSF expansion velocity observed here in the N–Al co-doped crystals was due to a reduction in the electronic energy gain ∆ and/or the immobilization of partial dislocations by impurities, caused by N–Al co-doping. The electronic energy gain ∆ has been calculated to increase with increasing nitrogen concentration [13, 15]. Since the free-electron concentration is roughly proportional to the nitrogen concentration, we consider that the DSF expansion velocity in the present study increased with the free-electron concentrations owing to an increase in the electronic energy gain ∆. In a recent study, a reduction in the electronic energy gain ∆ in N–Al co-doped crystals has been calculated [15]. Moreover, it has been suggested that the reduction in electronic energy gain ∆ is owing to electron trapping in the lower energy states of acceptors, which have different energy states in perfect 4H-SiC and DSF-faulted crystals [15, 39]. We thus consider that a reduction in the electronic energy gain ∆, due to N–Al co-doping, helps reduce the DSF expansion velocity in N–Al co-doped crystals. Moreover, the observed suppression of DSF formation during crystal growth may also be attributed to a reduction in the electronic energy gain ∆ caused by N–Al co-doping.
Studies on gallium arsenide have shown the immobilization of dislocations by impurities [35–38], which was explained by the interaction between the dislocations and strain fields created by the impurities. In the present study, however, although the nitrogen concentration was very high (> 3 × 1019 cm-3) in the nitrogen-doped 4H-SiC crystals, the DSF expansion velocity was found to increase with increasing nitrogen concentration, suggesting that the immobilization effect of nitrogen on the partial dislocation of DSFs is relatively weak. On the other hand, the strain field caused by aluminum may be greater than that caused by nitrogen, owing to the large atomic size difference between aluminum and substituted silicon atoms (as discussed in Refs. 21 and 22). In addition, there may be complexes or clusters in the N–Al co-doped crystals, which have been introduced by the nitrogen-aluminum doping (photoluminescence from the N–Al pair in 4H-SiC, as previously reported [40]), and such complexes or clusters may produce larger strain fields. Therefore, some impurity-based immobilization effects, which increase the activation energy of partial dislocations, cannot be ruled out when considering the reduction in DSF expansion velocity in N–Al co-doped crystals. However, similar patterns of the DSF expansion velocity have been confirmed when comparing N–B co-doped and nitrogen-only-doped crystals [21]. Therefore, we presume that a reduction in the electronic energy gain ∆ by donor and acceptor co-doping is responsible for the suppression of DSF formation.
4
Conclusion
We performed N–Al co-doped n-type 4H-SiC crystal growth by the PVT method using a two-zone heating furnace. N–Al co-doping at a high-nitrogen-concentration of 5.2 × 1019 cm-3 and aluminum concentration of 2.0 × 1019 cm-3 was achieved by employing preannealed Al4C3 powder. In the crystal with a higher aluminum concentration, the EPD was high as 5 × 104 cm-2, and an increase in TSD density was only observed in this crystal (and not in the crystal with a lower aluminum concentration). Moreover, we investigated the DSF expansion velocities in the N–Al co-doped and nitrogen-only-doped crystals. Expansion of DSFs was observed in both types of crystal after annealing at 1000°C for 2 hours, and the DSF expansion velocity was found to increase with increasing free-electron concentration. The DSF expansion velocities were lower in the N–Al co-doped crystals than those in the nitrogen-only-doped crystals, with equivalent free-electron concentrations; this difference implies a change in the electronic energy gain ∆ and/or the activation energy for the partial dislocations of DSF, caused by N–Al co-doping.
Acknowledgements This work was supported by Council for Science, Technology and Innovation (CSTI), Cross-ministerial Strategic Innovation Promotion Program (SIP), “Next-generation power electronics/Consistent R&D of nextgeneration SiC power electronics” (funding agency: NEDO). We wish to thank the members of AIST and Showa Denko K.K. for their kind advices.
References
[1] T. Kimoto, Jpn. J. Appl. Phys. 54 (2015) 040103. [2] T. Kato, T. Miura, K. Wada, E. Hozomi, H. Taniguchi, S. Nishizawa, K. Arai, Mater. Sci. Forum 556–557 (2007) 239–242. [3] K. Ohtomo, N. Matsumoto, K. Ashida, T. Kaneko, N. Ohtani, M. Katsuno, S. Sato, H. Tsuge, T. Fujimoto, J. Cryst. Growth 478 (2017) 174–179. [4] K. Kusunoki, K. Kamei, K. Seki, S. Harada, T. Ujihara, J. Cryst. Growth 392 (2014) 60–65. [5] N. Ohtani, M. Katsuno, J. Takahashi, H. Yashiro, M. Kanaya, J. Appl. Phys. 83 (1998) 4487–4490. [6] R. C. Glass, D. Henshall, V. F. Tsvetkov, C. H. Carter, Jr., Phys. Status Solidi B 202 (1997) 149–162. [7] H.-J. Rost, J. Doerschel, K. Irmscher, D. Schulz, D. Siche, J. Cryst. Growth 257 (2003) 75–83. [8] N. Ohtani, M. Katsuno, M. Nakabayashi, T. Fujimoto, H. Tsuge, H. Yashiro, T. Aigo, H. Hirano, T. Hoshino, K. Tatsumi, J. Cryst. Growth 311 (2009) 1475–1481. [9] R. S. Okojie, M. Xhang, P. Pirouz, S. Tumakha, G. Jessen, L. J. Brillson, Appl. Phys. Lett. 79 (2001) 3056–3058. [10] J.Q. Liu, H. J. Chung, T. Kuhr, Q. Li, M. Skowronski, Appl. Phys. Lett. 80 (2002) 2111–2113. [11] K. Irmscher, M. Albrecht, M. Rossberg, H.-J. Rost, D. Siche, G. Wagner, Physica B 376–377 (2006) 338–341. [12] T. Straubinger, R. L. Woodin, T. Witt, J. Shovlin, G. Dolny, P. Sasahara, E. Schmitt, A.-D. Weber, J. B. Casady, J. R. B. Casady, Mater. Sci. Forum 645–648 (2010) 223–226. [13] T. A. Kuhr, J. Q. Liu, H. J. Chung, M. Skowronski, F. Szmulowicz, J. Appl. Phys. 92 (2002) 5863–5871.
[14] M. S. Miao, S. Limpijumnong, W. R. L. Lambrecht, Appl. Phys. Lett. 79 (2001) 4360–4362. [15] C. Taniguchi, A. Ichimura, N. Ohtani, M. Katsuno, T. Fujimoto, S. Sato, H. Tsuge, T. Yano, J. Appl. Phys. 119 (2016) 145704. [16] M. S. Miao, W. R. L. Lambrecht, Phys. Rev. B 85 (2012) 205318. [17] H. Iwata, U, Lindefelt, S. Öberg, P. R. Briddon, J. Appl. Phys. 93 (2003) 1577–1585. [18] K. Maeda, R. Hirano, Y. Sato, M. Tajima, Mater. Sci. Forum 725 (2012) 35–40. [19] H. Idrissi, B. Pichaud, G. Regula, M. Lancin, J. Appl. Phys. 101 (2007) 113533. [20] Y. Tokuda, I. Kamata, N. Hoshino, T. Kato, H. Okumura, T. Kimoto, H. Tsuchida, J. Cryst. Growth 468 (2017) 889–893. [21] H. Suo, K. Eto, T. Ise, Y. Tokuda, H. Osawa, H. Tsuchida, T. Kato, H. Okumura, J. Cryst. Growth 468 (2017) 879–882. [22] K. Kojima, T. Kato, S. Ito, J. Kojima, F. Hirose, Y. Kito, S. Yamauchi, K. Nishikawa, A. Adachi, Mater. Sci. Forum 679–680 (2011) 8–11. [23] T. Kato, K. Eto, S. Takagi, T. Miura, Y. Urakami, H. Kondo, F. Hirose, H. Okumura, Mater. Sci. Forum 778–780 (2014) 47–50. [24] H. Suo, K. Eto, T. Kato, K. Kojima, H. Osawa, H. Okumura, Mater. Sci. Forum 897 (2017) 3–6. [25] S. W. Huh, H. J. Chung, M. Benamara, M. Skowronski, J. J. Sumakeris, M. J. Paisley, J. Appl. Phys. 96 (2004) 4637–4641. [26] Y. Chen, C. Lee, M. Yeng, H. Chiu, J. Cryst. Growth 247 (2003) 363–370. [27] Y. Yao, Y. Ishikawa, Y. Sugawara, K. Sato, K. Danno, H. Suzuki, T. Bessho, S. Yamaguchi, K. Nishikawa, J. Cryst. Growth 364 (2013) 7–10. [28] M. Miyata, Y. Hayafuji, Appl. Phys. Express 1 (2008) 111401. [29] T. Mitani, N. Komatsu, T. Takahashi, T. Kato, T. Ujihara, Y. Matsumoto, K. Kurashige, H. Okumura, Mater. Sci. Forum 821–823 (2015) 9–13. [30] R. Müller, U. Künecke, R. Weingärtner, H. Schmitt, P. Desperrier, P. Wellmann, Mater. Sci. Forum 483– 485 (2005) 31–34. [31] K. Eto, H. Suo, T. Kato, H. Okumura, J. Cryst. Growth 470 (2017) 154–158. [32] C. Qiu, R. Metselaar, J. Alloys Compd. 216 (1994) 55–60.
[33] S. Krukowski, J. Chem. Phys. 117 (2002) 5866–5875. [34] M. Dudley, X. R. Huang, W. Huamg, A. Powell, S. Wang, P. Neudeck, M. Skowronski, Appl. Phys. Lett. 75 (1999) 784–786. [35] I. Yonenaga, K. Sumino, J. Appl. Phys. 62 (1987) 1212–1219. [36] I. Yonenaga, K. Sumino, J. Appl. Phys. 65 (1989) 85–92. [37] I. Yonenaga, K. Sumino, J. Appl. Phys. 71 (1992) 4249–4257. [38] J. R. Patel, P. E. Freeland, Phys. Rev. Lett. 18 (1967) 833–836. [39] M. Ikeda, H. Matsunami, T. Tanaka. Phys Rev. B. 22 (1980) 2842–2854. [40] I. G. Ivanov, A. Ellison, E. Janzén, Mater. Sci. Forum 433–436 (2003) 321–324.
) m c
6
39 1
01 × ( no tia rt ne nco c ne go rit N
5 4 3 2
N-Al co-doped (1750? ) N-Al co-doped (1630? )
1
Nitrogen-only doped Theoretical curve
0 0
20
40
60
80
100
Flow ratio [N2 / (Ar + N2)] (%)
Fig. 1: Relationship between nitrogen concentration and N2 to argon flow ratio [N2 / (Ar + N2)] under the three different conditions; N-Al co-doping (1750 or 1630°C) and nitrogenonly-doping. The dashed line is the best-fit theoretical curve obtained on the basis of Eq. (1), for nitrogen-only-doped crystals. The temperatures in parentheses are the Al4C3 temperatures for each growth condition.
) -3
cm 9 1 01 × ( no tia rt ne cn oc nu ni m ul A
8 7 6
p-type
5 4 3 2 1 0 1550
1600
1650 1700 1750 Al4C3 temperature (ºC)
1800
1850
Fig. 2: Relationship between aluminum concentration and Al4C3 temperature. The dashed line is the best-fit exponential approximation curve. Only the crystal grown at an Al4C3 temperature of 1820°C was p-type.
)] 46 m c( n io ta rt 45 n ec n o c 44 m u n i m u l [A 43 g o L
-3
p-type
42 4.7
4.8 4.9 5 5.1 5.2 Reciprocal of Al4C3 temperature (10000 / K)
Fig. 3: Relationship between logarithm of aluminum concentration and reciprocal of Al4C3 temperature. The dashed line is the best-fit approximation line.
Fig. 4: Photograph of grown N–Al co-doped crystal with 3-inch-diameter seed crystal.
5.3
Fig. 5: Molten KOH + Na2O2-etched (a) H-Al substrate and (b) L-Al substrate surface.
Fig. 6: Cross-sectional transmission XRT images with 0004 diffractions of an H-Al-2 cross-sectional sample: (a) at a growth thickness of 10 mm and (b) in the vicinity of the grown/seed crystal interface.
Fig. 7: Cross-sectional transmission XRT images with 0004 diffractions of an L-Al-2 cross-sectional sample: (a) at a growth thickness of 10 mm and (b) in the vicinity of the grown/seed crystal interface.
Fig. 8: Optical micrographs of line-shaped etch pits _ on the (1100) surface of cross-sectional samples after KOH etching; (a) N–Al co-doped and (b) nitrogenonly-doped.
120
DSF expansion velocity (µm/h)
N-Al co-doped
100
Nitrogen-only doped
80 60 40 20 0 0.0
1.0
2.0
Free-electron concentration
3.0
4.0
( ×1019 cm-3)
Fig. 9: Relationship between DSF expansion velocity and free-electron concentration in N–Al co-doped and nitrogenonly-doped crystals. Error bars indicate the standard error for each free-electron concentration.
N-Al co-doped n-type 4H-SiC crystals were grown using the PVT method. The relation between impurity concentration and Al4C3 temperature was investigated. The threading dislocation density of highly N-Al co-doped crystal was high. The expansion velocities of the DSFs in the N-Al co-doped crystals were low.
S0022-0248(18)30293-8 https://doi.org/10.1016/j.jcrysgro.2018.06.019 CRYS 24641
To appear in:
Journal of Crystal Growth
Received Date: Revised Date: Accepted Date:
28 February 2018 18 June 2018 19 June 2018
Please cite this article as: H. Suo, K. Eto, T. Ise, Y. Tokuda, H. Osawa, H. Tsuchida, T. Kato, H. Okumura, Crystal Growth and Evaluation of Nitrogen and Aluminum Co-Doped N-type 4H-SiC Grown by Physical Vapor Transport, Journal of Crystal Growth (2018), doi: https://doi.org/10.1016/j.jcrysgro.2018.06.019
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Crystal Growth and Evaluation of Nitrogen and Aluminum Co-Doped N-type 4H-SiC Grown by Physical Vapor Transport H. Suo1, 2, K. Eto1, T. Ise1, 3, Y. Tokuda1, 4, H. Osawa2, H. Tsuchida5, T. Kato1, and H. Okumura1 1
National Institute of Advanced Industrial Science and Technology (AIST), Onogawa 16-1, Tsukuba, Ibaraki 305-8569, Japan 2
3
4
5
Abstract
Showa Denko K.K., 1-13-9, Shiba Daimon, Minato-ku, Tokyo 105-8518, Japan
Asahi Diamond Industrial Corporation, 787, Tabi, Ichihara, Chiba, 290-0515, Japan
DENSO CORPORATION, 500-1, Komenokichominamiyama, Nissin, Aichi, 407-0111 Japan
Central Research Institute of Electric Power Industry (CRIEPI), 2-6-1, Nagasaka, Yokosuka, Kanagawa, 240-0196, Japan
N-type 4H-SiC crystals were grown by the physical vapor transport (PVT) method with nitrogen
and aluminum (N–Al) co-doping. By using aluminum carbide powder preannealed in nitrogen gas atmosphere as an aluminum doping source, we obtained highly N–Al co-doped crystals with a nitrogen concentration higher than that in nitrogen-only-doped crystals. The dislocation densities of N-Al co-doped crystals with a high aluminum concentration (> 1 × 1019 cm-3) were found to become higher than those with a low aluminum concentration (< 1 × 1019 cm-3). Moreover, we investigated the expansion velocities of double Shockley-type stacking faults (DSFs) in the N–Al co-doped and the nitrogen-only-doped crystals. We found that the DSF expansion velocities in the N–Al co-doped crystals were lower than those in the nitrogen-only-doped crystals. This difference in the DSF expansion velocity is discussed with respect to the quantum well action model.
Keyword A1. Defect, A1. Doping, A2. Growth from vapor, B2. Semiconducting silicon compounds
1. Introduction Silicon carbide (SiC) has excellent physical and chemical properties as a semiconductor for use in high-power electronics. The strong chemical bonding between silicon and carbon atoms provides this material with a wide bandgap and high critical (breakdown) electric field strength. Recently, many devices for high-power electronics have been demonstrated. In the case of SiC metal-oxide-semiconductor field-effect transistors (MOSFETs) with a breakdown voltage of 600 V class, the specific on-resistances of the devices have reached less than 1 mΩcm2 [1]. However, further on-resistance reduction in such devices is desired. The resistivity
(15–25 mΩcm) of commercially available n-type 4H-SiC substrates cannot be ignored as one of the resistance components in low-on-resistance devices. There is significant demand for SiC substrates with a low-resistivity (< 10 mΩcm), which will reduce the on-resistance of SiC high-power devices. In SiC, nitrogen (N) is used as a donor for n-type doping, whereas aluminium (Al) and boron (B) are employed as acceptors for p-type doping. The concentration of doped nitrogen can be controlled by adjusting the nitrogen partial pressure and/or gas flow rate during the growth of SiC crystals by physical vapor transport (PVT). Several groups have reported on the growth of low-resistivity n-type 4H-SiC bulk crystals [2–8], and the nitrogen concentration and low-resistivity have reached 1.3 × 1020 cm-3 and 1.5 mΩcm, respectively. When the nitrogen concentration exceeds 1–3 × 1019 cm-3, n-type 4H-SiC crystals often degrade owing to the formation of double Shockley-type stacking faults (DSFs) [2–4, 7–11]. Nuclei of DSFs can be introduced by surface damage, for example scratches [11]. The DSFs expand from such nuclei during postgrowth hightemperature treatment in highly nitrogen-doped crystals, which increases the resistivity of the crystals [12]. Thus, DSF formation is a major issue for the practical use of low-resistivity 4H-SiC substrates for high-power devices. The expansion of DSFs is considered to occur when the system free energy of the crystal is reduced by DSF formation in a perfect 4H-SiC crystal. The quantum well action (QWA) model has been proposed to describe the difference in the electronic energy between perfect and DSF-faulted 4H-SiC crystals, especially in crystals with high nitrogen doping concentrations [13–15]. The electronic energy gain ∆ is defined as the difference in the total electron energy between the perfect and DSF-faulted 4H-SiC crystals. The energy state of electrons in a DSF is expressed as a two-dimensional electron gas in a quantum well, and has been calculated to be 0.68 eV below the conduction band minimum [16]. The Fermi energy can become comparable to or exceed the DSF energy state in heavily nitrogen-doped crystals with high donor concentrations. In this case, the electronic energy of the system decreases with the fall of many free electrons in the conduction band to the lower DSF energy states. On the other hand, it has been reported that the DSF formation energy, which is related to the structural strain energy, is greater than zero [17]. When the electronic energy gain ∆ overcomes the increase in the structural strain energy, DSF expansion is induced to reduce the system free energy. It has been reported that the expansion velocity of single Shockley-type stacking faults (SSFs) depends on their electronic energy [18], and it is also considered that the expansion velocity of DSFs is affected by the
electronic energy gain ∆. Although DSF expansion has been observed in lowly nitrogen-doped (< 1 × 1019 cm3
) crystals with a lower electronic energy gain ∆ under external stress [19], such expansion has been observed
in highly nitrogen-doped crystals (> 2 × 1019 cm-3) with a higher electronic energy gain ∆, even without external stress [20, 21]. Recently our group has reported on the growth of high-nitrogen-concentration and low-resistivity (<10 mΩcm) 4H-SiC crystals under nitrogen and aluminum (N–Al) co-doping conditions [22–24]. Despite the high-nitrogen-concentration and low-resistivity of the grown crystals, DSF generation was suppressed by N– Al co-doping in both chemical vapor deposition (CVD) [22] and PVT growth [23, 24]. It has been calculated that N–Al co-doping results in a decrease in the electronic energy gain ∆ [15], and the DSF suppression observed in our aforementioned studies was presumably attributed to this decrease. In addition, low-resistivity nitrogen and boron (N–B) co-doped crystals were grown by the PVT method, and DSF generation in the asgrown state was also suppressed by N–B co-doping [21]. Furthermore, a marked difference in the DSF expansion velocity between N–B co-doped and nitrogen-only-doped crystals during postgrowth hightemperature treatment was confirmed [21]. Not only DSFs, but also other extended defects in bulk crystals can affect device performance. Threading screw dislocations (TSDs), threading edge dislocations (TEDs), and basal plane dislocations (BPDs) are known as major extended defects in bulk crystals [1]. However, little experimental information is available on the relationship between dislocation density and impurity concentration in N–Al co-doped crystals. It was reported that, in CVD growth, an increase in the threading dislocation density is caused by stress relaxation associated with a high aluminum concentration of 3.8 × 1020 cm-3 [25]. In this study, we investigated the relationship between the impurity concentration and heating condition of aluminum carbide (Al4C3) powder that had been preannealed in a nitrogen atmosphere and the relationship between the impurity concentration and dislocation density in grown crystals. We also investigated the velocity of DSF expansion from nuclei introduced by indentation on the surface of N–Al co-doped and nitrogen-only-doped crystals. Finally, on the basis of differences in the DSF expansion velocity of various free electron concentrations, we discussed the effect of N–Al co-doping on DSF expansion with respect to the QWA model.
2. Experimental Crystals were grown using the two-zone heating furnace with induction heating coils to independently heat the SiC source powder and Al4C3 powder [23, 24]. The two-zone heating furnace was described in detail in Refs. _
23 and 24. Three-inch 4° off-axis (0001) C-face 4H-SiC substrates were used as seed crystals. The growth pressure was 5 Torr. Nitrogen and aluminum co-doping was performed by introducing nitrogen gas (N2) and heating the Al4C3 powder. The flow ratio of N2 to argon [N2 / (Ar + N2)] was varied in the range of 5–100% to obtain crystals with different nitrogen concentrations. To investigate the relationship between the aluminum concentration and Al4C3 temperature, the Al4C3 temperature was varied in the range of 1600–1850°C, which was maintained 400–600°C lower than the seed temperature because of the high vapor pressure of Al4C3. Before crystal growth, the Al4C3 powder had been annealed for more than 50 hours in N2 atmosphere, at over 1500°C, for stable sublimation and nitridation of the Al4C3 powder. In general, powder with a large specific surface area (high surface free energy) can easily undergo sublimation. It was reported that the specific surface area of titanium dioxide power is decreased by annealing [26]. Similarly, it is presumed that the specific surface area of Al4C3 powder is decreased by preannealing. Therefore, the stable sublimation of preannealed Al4C3 powder is considered to be caused by decrease in the specific surface area. The composition of the preannealed Al4C3 was analyzed using the X-ray diffraction technique. Four-degree offaxis (0001) substrates were obtained from the grown crystals after wire-saw cutting, grinding, and polishing _
with diamond slurry, and (1100) cross-sectional samples were also obtained from the grown crystals. The nitrogen and aluminum concentrations in the grown crystals were measured by secondary ion mass spectrometry (SIMS). To evaluate their electrical properties, Hall effect measurement with van der Pauw geometry was performed at room temperature, using the 4° off-axis (0001) substrates. As it is difficult to evaluate the dislocation density of highly nitrogen-doped 4° off-axis (0001) substrates (including N–Al codoped crystals) using the conventional molten potassium hydroxide (KOH) etching technique, chemical wet etching in molten KOH + sodium peroxide (Na2O2) [27] at 500°C for 2–3 min was used to evaluate the dislocation densities. The propagation of a dislocation from a seed crystal was investigated by transmission Xray topography (XRT) with 0004 diffractions using a molybdenum Kα1 beam (0.70926 Å). The TSD density was estimated from the number of propagating TSDs in the evaluated area. As it has been reported that carbon
inclusions generate TSDs [6], the carbon inclusion density was also evaluated using a transmission optical microscope. _
To introduce DSF nuclei, indentations were formed on the (1100) surface of cross-sectional samples using a microhardness tester with a diamond indenter. The indentation force was 100 mN. Then, the samples were annealed at 1000°C for 2 hours in an argon atmosphere, to investigate DSF expansion. Molten KOH etching was carried out to observe line-shaped etch pits, which correspond to DSF formation regions. The DSF expansion length was evaluated in terms of the etch-pit length from the indentation point, and the DSF expansion velocity was estimated by dividing the etch-pit length by the annealing time. The DSF expansion velocity was assumed to represent the dislocation velocity of leading partials of the expanding DSFs.
3. Results and discussion 3.1
Crystal growth and dislocation density of N–Al co-doped crystals
We investigated the relationship between the impurity concentration and heating conditions of the preannealed Al4C3 powder. Figure 1 shows the relationship between the nitrogen concentration at nonfaceted regions in the crystal and the N2 to argon flow ratio. In the case of N–Al co-doping, the heating temperature of the preannealed Al4C3 powder was 1750 or 1630°C. The nitrogen-only-doped crystals were grown without Al4C3 powder. The nitrogen concentration increased with increasing the N2 to argon flow ratio, under the three conditions with (1750 or 1630°C) or without using Al4C3 powder. The nitrogen concentrations in the N–Al codoped crystals with the preannealed Al4C3 powder were higher than those in the nitrogen-only-doped crystals at all flow rates. In addition, the nitrogen concentrations increased with increasing Al4C3 temperature. At a growth pressure of 5 Torr, whereas the nitrogen concentrations of nonfaceted regions in the nitrogen-onlydoped crystal did not exceed 3.5 × 1019 cm-3, the nitrogen concentration reached 4.8 × 1019 cm-3 in the N–Al co-doped crystal with the preannealed Al4C3 powder. Nitrogen doping of SiC crystals during PVT growth has been studied by several groups [5–8]. The nitrogen concentration is roughly expressed in terms of a Langmuir isotherm-type equation as
CN = A
Kp , 1 + Kp
(1)
where CN is the nitrogen concentration, A is a proportionality constant, K is a kinetic parameter related to the nitrogen adsorption and desorption processes, and p is the N2 partial pressure in the growth ambient. The dashed line in Fig. 1 is the best-fit theoretical curve obtained on the basis of Eq. (1) for nitrogen-only-doped crystals. The nitrogen concentration is in good agreement with Eq. (1). The nitrogen concentrations in N–Al co-doped crystals were, however, significantly higher than those in the nitrogen-only-doped crystals. Results of X-ray diffraction analysis of the preannealed Al4C3 powder showed that the powder consisted of Al4C3, aluminum nitride (AlN), Al5C3N, and carbon. We consider that N2 was supplied from the preannealed Al4C3 powder by heating, and the amount of supplied N2 increased with increasing Al4C3 powder temperature. When the N2 partial pressure p is very high, the nitrogen concentration approaches 3.7 × 1019 cm-3, using the aforementioned best-fit theoretical curve. However, the nitrogen concentrations in the N–Al co-doped crystals at an Al4C3 temperature of 1750°C exceeded this nitrogen concentration of 3.7 × 1019 cm-3. Therefore, we conclude that the nitrogen incorporation differs between the N–Al co-doped and nitrogen-only-doped crystals grown by the PVT method, beyond the effect of supplying additional N2 from preannealed Al4C3 in the N–Al co-doped growth. There is a possibility that a low formation energy of an acceptor-donor complex [28, 29] is the cause of the difference in nitrogen concentration between N–Al co-doped and nitrogen-only-doped crystals. Figure 2 shows the relationship between the aluminum concentration in the crystal and the Al4C3 temperature. The dashed line is the exponential line fitted to the experimental data. Only the crystal grown at an Al4C3 temperature of 1820°C was p-type, because the high aluminum concentration was higher than the nitrogen concentration. The aluminum concentration increased exponentially with increasing Al4C3 temperature. Aluminum doping during PVT growth has been studied [30, 31], and the aluminum concentration in a grown crystal was shown to have a linear dependence on the aluminum partial pressure during crystal growth. We consider that the observed exponential increase in aluminum concentration with increasing Al4C3 temperature was caused by an exponential increase in the vapor pressure of aluminum from Al4C3 and AlN [32, 33]. This result is consistent with a previous report on p-type growth [29]. Figure 3 shows the relationship between the logarithm of aluminum concentration and the reciprocal of Al4C3 temperature. The logarithm of aluminum concentration shows a linear dependence on the reciprocal of Al4C3 temperature. Therefore, the aluminum
concentration is considered to be mainly affected by aluminum sublimation. Assuming that the aluminum sublimation and aluminum concentration in a grown crystal are proportional, the activation energy of aluminum sublimation from preannealed Al4C3 powder is calculated as 530 kJ/mol from Fig. 3. This activation energy is higher than the reported activation energies of sublimation, which are approximately 180 kJ/mol for Al4C3 [32] and 420 kJ/mol for AlN [33]. In this experiment, it is possible that non-equilibrium and complicated pathways (e.g. sublimation, transport in the furnace, and incorporation) affected the activation energy. Figure 4 shows a photograph of a grown N–Al crystal, with a 3-inch-diameter seed crystal grown at an Al4C3 temperature of 1730°C. The nitrogen concentration was 5.2 × 1019 cm-3 and the aluminum concentration was 2.0 × 1019 cm-3 at a growth thickness of 6 mm. Moreover, the nitrogen concentration was 5.1 × 1019 cm-3 and the aluminum concentration was 1.8 × 1019 cm-3 at a growth thickness of 14 mm. The nonuniformities of nitrogen and aluminum concentrations for each growth thicknesses were less than 10%. The typical growth rate of a N–Al co-doped crystal was 100–150 µmh-1. Crystals with different aluminum concentrations were grown to evaluate the relationship between the dislocation density and aluminum concentration. Figure 5 shows the molten KOH + Na2O2-etched surface of the 4° off-axis (0001) substrate with (a) a higher aluminum (H-Al) concentration and (b) a lower aluminum (L-Al) concentration. The H-Al substrate grown at an Al4C3 temperature of 1750°C, had an aluminum concentration of 2 × 1019 cm-3 and nitrogen concentration of 5 × 1019 cm-3, whereas the L-Al substrate grown at an Al4C3 temperature of 1660°C had an aluminum concentration of 7 × 1018 cm-3 and a nitrogen concentration of 3 × 1019 cm-3. From the shape of the etch pits, threading dislocations and BPDs can be distinguished. It was difficult to clearly distinguish between TSDs and TEDs from the shape and size of the etch pits. The etch pit density (EPD) is the sum of the TSD, TED, and the BPD densities. The EPD and BPD density were determined by averaging their values evaluated at 10 areas in the substrate. The EPD of the H-Al substrate was 5 × 104 cm-2, whereas that of the L-Al substrate was 1 × 104 cm-2. The BPD densities were 3 × 102 cm-2 in the H-Al substrate and 6 × 102 cm-2 in the L-Al substrate. We found that the threading dislocation density clearly increased in the H-Al crystal. No line-shaped etch pits caused by SFs were observed even in the faceted regions with a higher nitrogen concentration in either crystals.
Figure 6 shows cross-sectional XRT images with 0004 diffractions of a cross-sectional sample of a higher aluminum concentration 2 (H-Al-2) crystal grown at an Al4C3 temperature of 1730°C; the sample had an aluminum concentration of 2 × 1019 cm-3 and a nitrogen concentration of 5 × 1019 cm-3. The thickness of the HAl-2 cross-sectional sample was 300 µm. Dark lines, which correspond to TSDs, can be seen to propagate from the seed crystal to the grown crystal. The TSD density in the H-Al-2 crystal was estimated from the number of TSDs observed in the cross-sectional XRT image. The estimated TSD densities were 8 × 103 cm-2 at a growth thickness of 10 mm and 3 × 103 cm-2 in the vicinity of the grown/seed crystal interface. This significant increase in TSD density was observed only in the H-Al-2 crystal. Figure 7 shows cross-sectional XRT images of a cross-sectional sample of a lower aluminum concentration 2 (L-Al-2) crystal grown at an Al4C3 temperature of 1630°C; the sample had and with an aluminum concentration of 3 × 1018 cm-3 and a nitrogen concentration of 3 × 1019 cm-3. The thickness of the L-Al-2 cross-sectional sample was 215 µm. The estimated TSD densities were 3 × 103 cm-2 at a growth thickness of 10 mm and 3 × 103 cm-2 in the vicinity of the grown/seed crystal interface, with no significant increase in the TSD density observable in this case. The cause of the increase in threading dislocation density in the higher-aluminum-concentration crystals is not yet clearly understood. It has been reported that the dislocation density is increased by stress relaxation at a high aluminum concentration in CVD growth [25]. Similarly, we consider that the increase in threading dislocation density observed here may be related to stress relaxation in higher-aluminum-concentration crystals. Furthermore, it has been reported that TSDs are generated from carbon inclusions [6, 34]. The TSDs generation mechanism has been suggested to involve c-axis lattice misfits that occur when there is overgrowth on carbon inclusions [34]. The carbon inclusion densities of 4–5 × 104 cm-3, observed here, were almost the same in both the H-Al and L-Al-2 cross-sectional samples. The possible mechanism of the TSD density increase is considered that the increasing c-axis strain due to the high aluminum concentration causes lattice misfits during the overgrowth on carbon inclusions.
3.2
Expansion velocity of DSF during postgrowth high-temperature treatment _
Figure 8 shows cross-sectional micrographs of (1 1 00) N–Al co-doped crystal and nitrogen-only-doped crystals after KOH etching. The line-shaped etch pits in Fig. 8 correspond to DSF formation regions.
Expansion of DSFs was confirmed in both types of crystal. The DSF expansion velocities were found to be equivalent on both sides of the indentation points along the basal plane, indicating spontaneous DSF expansion during the postgrowth high-temperature treatment, owing to the lower free energy of the DSFfaulted 4H-SiC. Figure 9 shows the relationship between the DSF expansion velocity and free-electron concentration in the N–Al co-doped at an aluminum concentration of 3 × 1019 cm-3 and nitrogen-only-doped crystals. The DSF expansion velocity increased with increasing free-electron concentration in both types of crystal. Furthermore, the DSF expansion velocities in the N–Al co-doped crystals were lower than those in the nitrogen-only-doped crystals with equivalent free-electron concentrations. A similar decrease in the relative DSF expansion velocity has also been confirmed when comparing N–B co-doped and nitrogen-only-doped crystals [21]. Numerous authors have discussed the formation of DSFs in highly nitrogen-doped 4H-SiC crystals during postgrowth high-temperature treatments [7–15, 19–21]. A large electronic energy gain ∆ indicates instability of the perfect 4H-SiC, and this electronic energy gain ∆ is considered to affect the DSF expansion velocity. It has also been reported that impurities can affect the dislocation velocities and result in the immobilization of dislocations in semiconductor crystals [35–38]. The activation energy of dislocation motion is expected to increase with such immobilization. Therefore, we consider that the low DSF expansion velocity observed here in the N–Al co-doped crystals was due to a reduction in the electronic energy gain ∆ and/or the immobilization of partial dislocations by impurities, caused by N–Al co-doping. The electronic energy gain ∆ has been calculated to increase with increasing nitrogen concentration [13, 15]. Since the free-electron concentration is roughly proportional to the nitrogen concentration, we consider that the DSF expansion velocity in the present study increased with the free-electron concentrations owing to an increase in the electronic energy gain ∆. In a recent study, a reduction in the electronic energy gain ∆ in N–Al co-doped crystals has been calculated [15]. Moreover, it has been suggested that the reduction in electronic energy gain ∆ is owing to electron trapping in the lower energy states of acceptors, which have different energy states in perfect 4H-SiC and DSF-faulted crystals [15, 39]. We thus consider that a reduction in the electronic energy gain ∆, due to N–Al co-doping, helps reduce the DSF expansion velocity in N–Al co-doped crystals. Moreover, the observed suppression of DSF formation during crystal growth may also be attributed to a reduction in the electronic energy gain ∆ caused by N–Al co-doping.
Studies on gallium arsenide have shown the immobilization of dislocations by impurities [35–38], which was explained by the interaction between the dislocations and strain fields created by the impurities. In the present study, however, although the nitrogen concentration was very high (> 3 × 1019 cm-3) in the nitrogen-doped 4H-SiC crystals, the DSF expansion velocity was found to increase with increasing nitrogen concentration, suggesting that the immobilization effect of nitrogen on the partial dislocation of DSFs is relatively weak. On the other hand, the strain field caused by aluminum may be greater than that caused by nitrogen, owing to the large atomic size difference between aluminum and substituted silicon atoms (as discussed in Refs. 21 and 22). In addition, there may be complexes or clusters in the N–Al co-doped crystals, which have been introduced by the nitrogen-aluminum doping (photoluminescence from the N–Al pair in 4H-SiC, as previously reported [40]), and such complexes or clusters may produce larger strain fields. Therefore, some impurity-based immobilization effects, which increase the activation energy of partial dislocations, cannot be ruled out when considering the reduction in DSF expansion velocity in N–Al co-doped crystals. However, similar patterns of the DSF expansion velocity have been confirmed when comparing N–B co-doped and nitrogen-only-doped crystals [21]. Therefore, we presume that a reduction in the electronic energy gain ∆ by donor and acceptor co-doping is responsible for the suppression of DSF formation.
4
Conclusion
We performed N–Al co-doped n-type 4H-SiC crystal growth by the PVT method using a two-zone heating furnace. N–Al co-doping at a high-nitrogen-concentration of 5.2 × 1019 cm-3 and aluminum concentration of 2.0 × 1019 cm-3 was achieved by employing preannealed Al4C3 powder. In the crystal with a higher aluminum concentration, the EPD was high as 5 × 104 cm-2, and an increase in TSD density was only observed in this crystal (and not in the crystal with a lower aluminum concentration). Moreover, we investigated the DSF expansion velocities in the N–Al co-doped and nitrogen-only-doped crystals. Expansion of DSFs was observed in both types of crystal after annealing at 1000°C for 2 hours, and the DSF expansion velocity was found to increase with increasing free-electron concentration. The DSF expansion velocities were lower in the N–Al co-doped crystals than those in the nitrogen-only-doped crystals, with equivalent free-electron concentrations; this difference implies a change in the electronic energy gain ∆ and/or the activation energy for the partial dislocations of DSF, caused by N–Al co-doping.
Acknowledgements This work was supported by Council for Science, Technology and Innovation (CSTI), Cross-ministerial Strategic Innovation Promotion Program (SIP), “Next-generation power electronics/Consistent R&D of nextgeneration SiC power electronics” (funding agency: NEDO). We wish to thank the members of AIST and Showa Denko K.K. for their kind advices.
References
[1] T. Kimoto, Jpn. J. Appl. Phys. 54 (2015) 040103. [2] T. Kato, T. Miura, K. Wada, E. Hozomi, H. Taniguchi, S. Nishizawa, K. Arai, Mater. Sci. Forum 556–557 (2007) 239–242. [3] K. Ohtomo, N. Matsumoto, K. Ashida, T. Kaneko, N. Ohtani, M. Katsuno, S. Sato, H. Tsuge, T. Fujimoto, J. Cryst. Growth 478 (2017) 174–179. [4] K. Kusunoki, K. Kamei, K. Seki, S. Harada, T. Ujihara, J. Cryst. Growth 392 (2014) 60–65. [5] N. Ohtani, M. Katsuno, J. Takahashi, H. Yashiro, M. Kanaya, J. Appl. Phys. 83 (1998) 4487–4490. [6] R. C. Glass, D. Henshall, V. F. Tsvetkov, C. H. Carter, Jr., Phys. Status Solidi B 202 (1997) 149–162. [7] H.-J. Rost, J. Doerschel, K. Irmscher, D. Schulz, D. Siche, J. Cryst. Growth 257 (2003) 75–83. [8] N. Ohtani, M. Katsuno, M. Nakabayashi, T. Fujimoto, H. Tsuge, H. Yashiro, T. Aigo, H. Hirano, T. Hoshino, K. Tatsumi, J. Cryst. Growth 311 (2009) 1475–1481. [9] R. S. Okojie, M. Xhang, P. Pirouz, S. Tumakha, G. Jessen, L. J. Brillson, Appl. Phys. Lett. 79 (2001) 3056–3058. [10] J.Q. Liu, H. J. Chung, T. Kuhr, Q. Li, M. Skowronski, Appl. Phys. Lett. 80 (2002) 2111–2113. [11] K. Irmscher, M. Albrecht, M. Rossberg, H.-J. Rost, D. Siche, G. Wagner, Physica B 376–377 (2006) 338–341. [12] T. Straubinger, R. L. Woodin, T. Witt, J. Shovlin, G. Dolny, P. Sasahara, E. Schmitt, A.-D. Weber, J. B. Casady, J. R. B. Casady, Mater. Sci. Forum 645–648 (2010) 223–226. [13] T. A. Kuhr, J. Q. Liu, H. J. Chung, M. Skowronski, F. Szmulowicz, J. Appl. Phys. 92 (2002) 5863–5871.
[14] M. S. Miao, S. Limpijumnong, W. R. L. Lambrecht, Appl. Phys. Lett. 79 (2001) 4360–4362. [15] C. Taniguchi, A. Ichimura, N. Ohtani, M. Katsuno, T. Fujimoto, S. Sato, H. Tsuge, T. Yano, J. Appl. Phys. 119 (2016) 145704. [16] M. S. Miao, W. R. L. Lambrecht, Phys. Rev. B 85 (2012) 205318. [17] H. Iwata, U, Lindefelt, S. Öberg, P. R. Briddon, J. Appl. Phys. 93 (2003) 1577–1585. [18] K. Maeda, R. Hirano, Y. Sato, M. Tajima, Mater. Sci. Forum 725 (2012) 35–40. [19] H. Idrissi, B. Pichaud, G. Regula, M. Lancin, J. Appl. Phys. 101 (2007) 113533. [20] Y. Tokuda, I. Kamata, N. Hoshino, T. Kato, H. Okumura, T. Kimoto, H. Tsuchida, J. Cryst. Growth 468 (2017) 889–893. [21] H. Suo, K. Eto, T. Ise, Y. Tokuda, H. Osawa, H. Tsuchida, T. Kato, H. Okumura, J. Cryst. Growth 468 (2017) 879–882. [22] K. Kojima, T. Kato, S. Ito, J. Kojima, F. Hirose, Y. Kito, S. Yamauchi, K. Nishikawa, A. Adachi, Mater. Sci. Forum 679–680 (2011) 8–11. [23] T. Kato, K. Eto, S. Takagi, T. Miura, Y. Urakami, H. Kondo, F. Hirose, H. Okumura, Mater. Sci. Forum 778–780 (2014) 47–50. [24] H. Suo, K. Eto, T. Kato, K. Kojima, H. Osawa, H. Okumura, Mater. Sci. Forum 897 (2017) 3–6. [25] S. W. Huh, H. J. Chung, M. Benamara, M. Skowronski, J. J. Sumakeris, M. J. Paisley, J. Appl. Phys. 96 (2004) 4637–4641. [26] Y. Chen, C. Lee, M. Yeng, H. Chiu, J. Cryst. Growth 247 (2003) 363–370. [27] Y. Yao, Y. Ishikawa, Y. Sugawara, K. Sato, K. Danno, H. Suzuki, T. Bessho, S. Yamaguchi, K. Nishikawa, J. Cryst. Growth 364 (2013) 7–10. [28] M. Miyata, Y. Hayafuji, Appl. Phys. Express 1 (2008) 111401. [29] T. Mitani, N. Komatsu, T. Takahashi, T. Kato, T. Ujihara, Y. Matsumoto, K. Kurashige, H. Okumura, Mater. Sci. Forum 821–823 (2015) 9–13. [30] R. Müller, U. Künecke, R. Weingärtner, H. Schmitt, P. Desperrier, P. Wellmann, Mater. Sci. Forum 483– 485 (2005) 31–34. [31] K. Eto, H. Suo, T. Kato, H. Okumura, J. Cryst. Growth 470 (2017) 154–158. [32] C. Qiu, R. Metselaar, J. Alloys Compd. 216 (1994) 55–60.
[33] S. Krukowski, J. Chem. Phys. 117 (2002) 5866–5875. [34] M. Dudley, X. R. Huang, W. Huamg, A. Powell, S. Wang, P. Neudeck, M. Skowronski, Appl. Phys. Lett. 75 (1999) 784–786. [35] I. Yonenaga, K. Sumino, J. Appl. Phys. 62 (1987) 1212–1219. [36] I. Yonenaga, K. Sumino, J. Appl. Phys. 65 (1989) 85–92. [37] I. Yonenaga, K. Sumino, J. Appl. Phys. 71 (1992) 4249–4257. [38] J. R. Patel, P. E. Freeland, Phys. Rev. Lett. 18 (1967) 833–836. [39] M. Ikeda, H. Matsunami, T. Tanaka. Phys Rev. B. 22 (1980) 2842–2854. [40] I. G. Ivanov, A. Ellison, E. Janzén, Mater. Sci. Forum 433–436 (2003) 321–324.
) m c
6
39 1
01 × ( no tia rt ne nco c ne go rit N
5 4 3 2
N-Al co-doped (1750? ) N-Al co-doped (1630? )
1
Nitrogen-only doped Theoretical curve
0 0
20
40
60
80
100
Flow ratio [N2 / (Ar + N2)] (%)
Fig. 1: Relationship between nitrogen concentration and N2 to argon flow ratio [N2 / (Ar + N2)] under the three different conditions; N-Al co-doping (1750 or 1630°C) and nitrogenonly-doping. The dashed line is the best-fit theoretical curve obtained on the basis of Eq. (1), for nitrogen-only-doped crystals. The temperatures in parentheses are the Al4C3 temperatures for each growth condition.
) -3
cm 9 1 01 × ( no tia rt ne cn oc nu ni m ul A
8 7 6
p-type
5 4 3 2 1 0 1550
1600
1650 1700 1750 Al4C3 temperature (ºC)
1800
1850
Fig. 2: Relationship between aluminum concentration and Al4C3 temperature. The dashed line is the best-fit exponential approximation curve. Only the crystal grown at an Al4C3 temperature of 1820°C was p-type.
)] 46 m c( n io ta rt 45 n ec n o c 44 m u n i m u l [A 43 g o L
-3
p-type
42 4.7
4.8 4.9 5 5.1 5.2 Reciprocal of Al4C3 temperature (10000 / K)
Fig. 3: Relationship between logarithm of aluminum concentration and reciprocal of Al4C3 temperature. The dashed line is the best-fit approximation line.
Fig. 4: Photograph of grown N–Al co-doped crystal with 3-inch-diameter seed crystal.
5.3
Fig. 5: Molten KOH + Na2O2-etched (a) H-Al substrate and (b) L-Al substrate surface.
Fig. 6: Cross-sectional transmission XRT images with 0004 diffractions of an H-Al-2 cross-sectional sample: (a) at a growth thickness of 10 mm and (b) in the vicinity of the grown/seed crystal interface.
Fig. 7: Cross-sectional transmission XRT images with 0004 diffractions of an L-Al-2 cross-sectional sample: (a) at a growth thickness of 10 mm and (b) in the vicinity of the grown/seed crystal interface.
Fig. 8: Optical micrographs of line-shaped etch pits _ on the (1100) surface of cross-sectional samples after KOH etching; (a) N–Al co-doped and (b) nitrogenonly-doped.
120
DSF expansion velocity (µm/h)
N-Al co-doped
100
Nitrogen-only doped
80 60 40 20 0 0.0
1.0
2.0
Free-electron concentration
3.0
4.0
( ×1019 cm-3)
Fig. 9: Relationship between DSF expansion velocity and free-electron concentration in N–Al co-doped and nitrogenonly-doped crystals. Error bars indicate the standard error for each free-electron concentration.
N-Al co-doped n-type 4H-SiC crystals were grown using the PVT method. The relation between impurity concentration and Al4C3 temperature was investigated. The threading dislocation density of highly N-Al co-doped crystal was high. The expansion velocities of the DSFs in the N-Al co-doped crystals were low.