Two-inch high-quality (001) diamond heteroepitaxial growth on sapphire (112̄0) misoriented substrate by step-flow mode
To cite this article: Seong-Woo Kim et al 2021 Appl. Phys. Express 14 115501
Seong-Woo Kim1* , Ryota Takaya2 , Shintaro Hirano1 , and Makoto Kasu2
1 Adamant Namiki Precision Jewel Co., Ltd. Adachi, Tokyo, 123-8511 Japan
2 Department of Electrical and Electronic Engineering, Saga University, 1 Honjomachi, Saga, 840-8502 Japan
* E-mail: s-kim@ad-na.com
Received September 3, 2021; revised September 16, 2021; accepted September 21, 2021; published online October 5, 2021
Two-inch-diameter high-quality free-standing (001) diamond layers were grown on misoriented (1120) sapphire. The substrate misorientation allows step-flow growth, and tensile stress is released in the diamond layer. Consequently, the diamond layer delaminates naturally from the substrate without cracking. For the diamond grown on the sapphire misoriented by 7° toward the [1100] direction, the widths of the (004) and (311) X-ray rocking curves were 98.35 and 175.3 arcsec, respectively, the lowest ever reported. The curvature radius of the diamond was 99.64 cm in the [1100] direction and 260.21 cm in the [0001] direction of the substrate, the highest ever reported. © 2021 The Japan Society of Applied Physics
Diamond power semiconductor devices exhibit characteristics beyond those of SiC and GaN.1,2) This is because the diamond possesses exceptional physical properties, such as high breakdown electric field (>10 MV cm−1 ), extremely high thermal conductivity (22 W cm−1 K−1 ),3) and high carrier mobility (4500 and 3800 cm2 V−1 s −1 for electrons and holes, respectively).4) Ueda and Kasu et al. demonstrated high RF power performance in diamond, such as a maximum frequency of oscillation of 120 GHz5) and RF output power density of 2.1 W mm−1 at 1 GHz.6) Recently, we demonstrated an FET fabricated on our grown heteroepitaxial diamonds, Kenzan Diamond® , with a low specific on-resistance of 19.74 mΩ·cm2 , a high breakdown voltage of 2608 V, and Baliga’s figure of merit output power of 344.7 MW cm−2 , which are the highest ever reported in diamond7) and modulation-doped diamond FETs on Kenzan Diamond® . 8)
For semiconductor device research and development in corporate laboratories, a high-quality 2inch diameter wafer is essential. However, the size of commercially available single crystal diamond grown by high-pressure high-temperature method9,10) is 4 × 4 mm2 . Therefore, diamond heteroepitaxial growth has been extensively studied for several decades.11–29)
Compared to other materials, Ir is considered the best substrate for diamond heteroepitaxial growth because it offers the highest diamond nucleation density (>1 × 108 cm−2 ).17) However, Ir single crystals are expensive and not commercially available. Therefore, Ir/YSZ (yttria-stabilized zirconia)/Si20–22) and Ir/MgO23–25) have been extensively studied. Schreck et al.22) grew a 92 mm sized heteroepitaxial diamond on Ir/YSZ/Si substrate, and reported a FWHM of the X-ray rocking curve (XRC) (004) of 230 arcsec, and a (311) XRC of 432 arcsec.21,22) However, there have been no reports on semiconductor devices and characterization of their heteroepitaxial diamond. Aida and Kim et al. reported a FWHM of the XRC (004) of 252 arcsec on Ir/MgO substrate.25) They utilized the microneedle (Kenzan) process to delaminate the grown diamond layer from the substrate without cracking.
Compared to MgO and YSZ, sapphire has the advantages of high crystal quality, low cost, and large wafer availability (up to 8 inches). For the edge-defined film-fed growth method, by using a seed crystal with (1120) A-planes,8 inch diameter (1120) A-planes sapphire substrates can be grown. Furthermore, sapphire has a small coefficient of thermal expansion (CTE) (4.2 and 5.3 × 10−6 K−1 along the a axis and the c axis, respectively) compared to MgO (12.8 × 10−6 K−1 ); therefore, sapphire has a smaller mismatch of CTE with diamond (1.5 × 10−6 K−1 ). Dai et al. reported epitaxial (001) diamond growth on (001) Ir/(1120) sapphire substrates.26,27) Samoto et al. reported diamond growth on Ir/sapphire substrates for various surface orientations and misorientations; however, their epitaxial relation was not shown.28) Tang et al. investigated residual stress in diamonds as a result of heteroepitaxial overgrowth on a sapphire substrate.29) Recently, Kim et al. reported a 1 inch high-quality free-standing heteroepitaxial (001) diamond grown on a (001) Ir buffer layer/(1120) A-plane, just-oriented sapphire substrate.30)
In this letter, we demonstrate a 2 inch diameter high quality free-standing (001) diamond layer grown on a sapphire misoriented (1120) A-plane substrate in step-flow mode. Using the misoriented sapphire substrate, step-flow growth in diamond is realized, the tensile stress in the diamond layer is released, the curvature radius increases (bending reduces), and the diamond layer delaminates naturally from the substrate without the microneedle technique, and without cracking.
Figure 1 shows the fabrication procedure for free-standing heteroepitaxial diamond grown on Ir/misoriented sapphire. The growth parameters were described in Ref. 30. First, a sapphire (α-Al2O3) (1120) misoriented substrate was prepared [Fig. 1(a)]. The misorientation angle was changed from 0 to 7°, and the misorientation direction was the [1100] m-direction or [0001] c-direction. Then, an approximately 1 μm thick Ir buffer layer was deposited on the sapphire substrate using a sputtering method. As shown in Fig. 1(b), a bias-enhanced nucleation (BEN)31) for diamond nucleation was performed on the Ir buffer layer using DC plasma CVD. H2-diluted CH4 was used as the gas source. Here, the substrate was negatively biased so that positively charged CH3 radicals with kinetic energy were bombarded onto the Ir surface promoting diamond nucleation. As shown in Fig. 1(c), the diamond layer was grown on the BEN-treated Ir buffer. The CH4/(CH4 + H2) gas ratio was 5.5%. The substrate temperature was set to 1000 °C. N2 gas was slightlyadded. The diamond growth rate was 16.5 μm h−1 . The grown diamond thickness was in the range of 800– 1000 μm. Owing to the misoriented substrate, the diamond growth proceeds in the step-flow mode. This leads to a decrease in residual stress in the diamond layer. Unlike diamond growth on a just-oriented sapphire substrate reported previously,30) in this process, we did not use vertically aligned pillars, so-called microneedles (Kenzan), for delamination of the diamond layer from the substrate. Thus, as shown in Fig. 1(d), because the CTE of sapphire is much higher than that of diamond, during cooling the sapphire substrate contracts more than the grown diamond layer; consequently, the diamond layer with the Ir buffer layer is naturally delaminated from the sapphire substrate without any cracking.
Figure 2 shows the X-ray diffraction (XRD) results of the diamond/Ir/misoriented sapphire (1120) structure. A diamond layer was grown for 60.5 h, and thus, its estimated diamond thickness was approximately 1070 μm. As shown in Fig. 2(a), the XRD 2θ/ω-scan results display the Ir (002) (2θ = 47.334°), Ir (004) (2θ = 106.687°), and diamond (004) (2θ = 119.28°) diffraction peaks. The sapphire substrate was misoriented by 3° toward the [1100] m-direction. The sapphire substrate was fragmented before this measurement.
The results indicate that even if the misoriented sapphire substrate is used, the diamond (001) layer and Ir (001) buffer layer are epitaxially grown in the sapphire [1120] direction.
Figure 2(b) shows the XRC j-scan results for the diamond grown on the Ir/sapphire substrate. The j angle is the same for the diamond, Ir, and sapphire measurements. Sapphire {3300} diffraction peaks (ω = 4.079°, χ = 30.000°) from the sapphire (1120) substrate were observed at j = 88.48 and 268.00°. This indicates that 88.48 and 268.00° correspond to
the [1100] and [1100] directions of the sapphire (1120) substrate, respectively. Subsequently, the Ir {111} diffraction peaks from the Ir buffer layer (ω = 20.336°) were measured and Ir {111} diffraction peaks appeared at j = 0.28, 90.06, 180.29, and 269.90°, which correspond to the [110], [110], [110], and [110] directions of the Ir (001) buffer layer, respectively. Finally, the diamond {111} diffraction peaks from the diamond layer (ω = 21.965°) were measured and {111} diffraction peaks appeared at j = –0.573, 89.46, 179.43, and 269.42°, indicating that the j angles correspond to the [110], [110], [110], [110] directions of the diamond (001) layer, respectively. As shown in Fig. 2(c), the epitaxial relation between diamond, Ir buffer, and sapphire substrate isdetermined to be diamond (001) [110]//Ir (001) [110]// sapphire (1120) [0001]. This epitaxial relation is the same as that on a just-oriented sapphire substrate.30)
The misorientation-angle dependence of the FWHM of the XRC diamond (004) and (311) diffractions for the [0001] c- and [1100] m-misorientation directions is shown in Fig. 3(a). As reported previously, the FWHMs for zero misorientation (just-oriented) substrate of diamond (004) and (311) were 113.4, 234.0 arcsec, respectively,30) where the microneedle technique was used. Conversely, for the diamond without microneedle technique case, the FWHMs for zero misorientation of diamond (004) and (311) were 325– 363 and 655 arcsec, respectively. As the misorientation angle, α, increased toward either the [0001] or [1100] direction, the FWHM decreased drastically, indicating an improvement in the crystal quality. As a result, the FWHMs for α = 7° in the m-misorientation direction were the lowest. For the case of the just-oriented substrate, after diamond growth, during cooling [Fig. 1(d)], the sapphire substrate tended to break catastrophically because of its residual stress. However, for the case of the misoriented substrate, as the misorientation angle increased, the sapphire substrate exhibited small fractures.
Figure 3(b) shows that for 7°-misorientation toward the [1100] m-direction, the XRCs of the (004) and (311) diffractions show the FWHM for the (004) and (311) diffractionsas low as 98.35 and 175.3 arcsec, respectively. These values are lower than the FWHMs for a zero misorientation angle with the microneedle technique and the lowest among heteroepitaxial diamonds.
Figure 3(c) shows the FWHM values of the XRC of the diamond (004) diffraction peak on the 2 inch diameter diamond wafer; the values ranged from 99 to 152 arcsec.
The crystal quality was relatively uniform on the 2 inch diameter diamond wafer.
Figure 3(d) shows the XRCs of the Ir (004) and (311) diffraction peaks at 3° in the [1100] m-misorientation direction. The FWHMs of the respective peaks were 397.8 and 414.0 arcsec, respectively. The FWHMs of the Ir (004) peak were approximately half of the previous value of 550.8 arcsec on our just-oriented substrate.30) This interesting result means that the misorientation effect appears even in the Ir buffer layer prior to diamond growth.
Next, we investigated the crystal quality along the surface. Figure 3(e) shows the XRD in-plane 2θχ–f scan and XRC in plane f scan of the diamond (220) diffraction peaks of thediamond (001) layer misoriented by 7° toward the [1100] direction. The incident X-ray direction is set toward the [0001] direction. The FWHM value of the peak angle of 2θχ–f scan is 1641.6 arcsec. This indicates the distribution of the lattice constant, a, along the surface. The lattice constant, a, was determined from the 2θχ peak angle of 75.262° to be 3.56824 Å. which is a slightly higher than the reported value (3.567 Å). Here, the Cu Kα1 X-ray wavelength of 1.540929 Å was used. This indicates that the diamond layer undergoes residual tensile strain. In contrast, the FWHM value of the XRC in-plane f scan of the diamond (220) diffraction peak (the width of crystal rotation along the growth direction) was 662.4 arcsec.
Figure 3(f) shows the X-ray pole figure of diamond {111} diffractions from the diamond (001) layer grown on the Ir (001)/sapphire substrate misoriented by 5° toward the [1100] direction. Four {111} peaks appeared in the [±1 ± 10] direction, and no additional peaks due to twinning were observed. This four-fold symmetry proves the single crystallinity of the diamond layer without twinning.
Figure 4 shows the curvature of the diamond layer; thus, the diamond (004) XRC peak angle, ω, at different positions, X, from the diamond (001) layer grown on the Ir (001) buffer/ sapphire (1120) substrate misoriented by 7° toward the [1100] m-direction. The XRC of the (004) peak angle shifted to a higher angle. This implies that the diamond layer is convex upward, because the CTE of sapphire substrate is higher than that of the diamond layer; therefore, during cooling after diamond growth, sapphire substrate contracts more than the diamond layer. Thus, the residual stress remained after delamination. In the [1100] m-direction of the substrate, ω was 52.72126° at Y = −8.0 mm and 53.64174° at Y = 8.0 mm; by the least root-square method, the best-fitted curvature radius was determined to be 99.64 cm. Interestingly, in the [0001] direction of the substrate (perpendicular to the misorientation direction), ω was 57.97505° at X = −3.0 mm and 58.10732° at X = 3.0 mm; surprisingly, the curvature radius was determined to be 260.21 cm. This value is much longer than our previous value of approximately 90.6 cm for diamond on just-oriented substrate.30) These results additionally support that the misorientation effect releases the tensile stress in the diamond layer.
Figure 5 shows the scanning transmission electron microscopy (STEM) plan-view bright-field image of the free-standing (001) heteroepitaxial diamond. The sapphire substrate was misoriented by 7° in the [1100] m-direction. Under the observation conditions, all types of dislocations should be visible. A total of 13 dislocations were observed in the range of 7.8 μm × 6.5 μm. The threading dislocation density was determined to be 2.6 × 107cm−2 , which is twice as high as our previous value for a just-oriented substrate (1.4 × 107 cm−2 ).30) The crystal quality is at the highest level in the heteroepitaxial diamond. Interestingly, the dislocations appear to run in an oblique direction, and thus, they are not perpendicular to the diamond surface. Considering a sample thickness of 1 μm and the length of 0.3 μm of each dislocation observed from the surface, the dislocation angle was 17° to the growth direction(perpendicular to the surface), which was much higher than the substrate misorientation angle of 7°. This indicates that the propagation of threading dislocations is also influenced by the misorientation effect, that is, the step-flow growth mode.
Figure 6 shows the atomic force microscopy (AFM) image of diamond grown on Ir/sapphire (1120) misoriented by 5° toward the [1100] m-direction. Five bunched steps are approximately 80–240 nm height (130 nm average) and the terraces are 8 μm wide. The bunched steps were aligned in the [0001] direction of the substrate, thus the surface was misoriented in the [1100] m-direction. This bunched step structure proves that step-flow growth occurs on the diamond surface.
Figure 7 shows a photograph of a 2 inch diamond freestanding heteroepitaxial layer grown on the sapphiresubstrate misoriented by 5° toward the [0001] c-direction. No cracks were observed. The brown color originates from infrared absorption due to nitrogen impurities. We want to emphasize that this wafer was obtained without the microneedle technique. Heteroepitaxial diamond wafer fabrication without the microneedle technique is much simpler and less costly.
In conclusion, 2 inch diameter free-standing high-quality (001) diamond layers on misoriented (1120) A-plane sapphire substrates were grown in the step-flow mode. Using a misoriented substrate, the step-flow growth occurred and effectively decreased the tensile stress in the diamond. Consequently, without the microneedle technique, the natural delamination of the grown diamond layer from the substrate is possible. As the substrate misorientation angle increased, the XRC FWHM of diamond (004) and (311) diffractionsdecreased, thus the crystal quality improved. For diamond grown on a sapphire substrate with 7° misorientation toward the [1100] direction, the widths of the (004) and (311) XRC were 98.35 and 175.3 arcsec, respectively, which are the lowest ever reported. The curvature radius of diamond was 99.64 and 260.21 cm in the [1100] and [0001] direction of the substrate, respectively. The dislocation density was 2.6 × 107 cm−2 , and inclined dislocations were observed. This bunched-step structure proves that step-flow growth occurred on the diamond surface.
Acknowledgments The authors appreciate Dr. Saha Niloy Chandra (Saga University) for his fruitful discussions. This work was partially supported by the Japan Society for the Promotion of Science (JSPS) Grants-in-Aid for Scientific Research No. 19H02616; the Collaborative Research Program of the Research Institute for Applied Mechanics, Kyushu University, and the Institute of Ocean Energy, Saga University.