To planarize semiconductor materials such as gallium nitride (GaN) and silicon carbide with high efficiency, we developed a polarization method that combines ultraviolet irradiation and an abrasive-free polishing method known as catalyst-referred etching (CARE). In this method, the substrate surface is photoelectrochemically oxidized, thus improving the removal rate. Accordingly, an atomically well-ordered surface was obtained at a removal rate 100 times higher than that of the conventional CARE method without ultraviolet irradiation. However, in some cases, for GaN substrates with a high oxygen impurity concentration area, the oxidation rate is nonuniform on the substrate surface, resulting in the formation of a rough surface. In this study, we propose the application of a positive bias to the GaN substrate to suppress the oxidation rate fluctuation. In the positive bias state, the width of the depletion layer generated at the interface of GaN and the etchant becomes uniform on the entire surface regardless of crystallographic fluctuation, thereby achieving a uniform oxidation rate. When only 3.0 V was applied, the oxidation rate was uniform; thus, a flat GaN surface without the footprint originating from crystallographic fluctuations was obtained.
I. INTRODUCTION
Photoelectrochemical (PEC) oxidation proceeds at the step as well as on the terrace.15–17 The oxide materials are expected to be etched immediately because the removal rate is significantly higher than that of GaN and SiC, forming atomic-size pits on the terrace. The peripheral sites of the pit function as additional reaction sites, increasing the removal rate. This method of combining CARE and PEC oxidation is called PEC-CARE. In our previous study, the removal rates of SiC and GaN were found to be 100 times higher than that upon CARE without UV irradiation, and an atomically smooth surface similar to the surface processed by conventional CARE was obtained.15–17 The PEC-CARE method is thus a high-efficiency polishing method, with highly ordered surfaces and high removal rates obtained simultaneously, which is challenging when using other methods.20–25 However, in some cases, for GaN, particularly with extremely low defect densities, surface roughness increased. During the crystal growth of GaN, unintentional impurity is incorporated in the growing bulk. The impurity concentration is nonuniform within the plane owing to the different crystal growth modes within a plane. In particular, the impurity concentration is extremely high in regions grown in a direction different from the c-axis. In this region, the depletion layer is thinner than that in the c-axis-direction-growth region, consequently decreasing the number of holes contributing to surface oxidation. Therefore, the oxidation rate on the surface fluctuates depending on the impurity concentration, resulting in the formation of a rough surface upon PEC-CARE.
To suppress this oxidation rate fluctuation, we propose applying a bias to the GaN substrate to widen the depletion layer. The density of the photogenerated holes is uniform regardless of the impurity concentration, and only the holes in the depletion layer oxidize the surface. Therefore, when the depletion layer width increases beyond the UV-light penetration depth, the number of holes contributing to oxidation is uniform and independent of the impurity concentration, leading to a flat surface. There are no reports about examples of highly efficient smoothing of GaN substrates with impurity concentration distributions using only a pure water-based etchant and catalyst. The advancement of this method and the results of this study could be a breakthrough in the GaN-based device manufacturing process, which is suitable for today’s sustainable society. In this study, we demonstrate the surface polishing of two kinds of GaN substrates, which are prepared by the hydride vapor phase epitaxy (HVPE) method26 and Na-flux method27 using PEC-CARE by applying a positive bias.
II. EXPERIMENTAL SETUP
Figures 1(a) and 1(b) show the overview and cross-sectional image of the PEC-CARE instrument. A fluorine rubber pad with a 50 nm thick Pt layer was fixed in the tank. A neutral phosphate standard solution with a pH of 6.86 (HORIBA powder for neutral phosphate standard solution) was employed as an etchant. This solution has a low UV absorption coefficient and aids in preventing the isotropic dissolution of gallium oxide, which is an amphoteric metal. A UV light source (HAMAMATSU L9588-02A) was positioned below the GaN substrate, which was placed in a holder and pressed against the catalyst immersed in the etchant. The substrate and polishing pad were independently rotated around their respective axes, and the respective processing pressure and relative speed between the substrate and pad were 20 kPa and 10 cm/s, respectively. The UV light reached the GaN surface via the quartz glass plate and through-holes on the polishing pad. The rotation of the polishing pad ensured average, uniform illumination across the entire GaN surface at an intensity of 1.75 mW/cm2 at 365 nm during GaN polishing. The back of the GaN substrate was connected to a Pt wire, which was grounded and served as the cathode, via a DC power source (TAKASAGO LX035-18) and an amplifier (KEITHLEY 2100). As shown in Fig. 1(c), upon applying a positive bias to the GaN substrate, more electrons are attracted to the bulk side, thereby increasing the width of the depletion layer formed at the interface between the GaN surface and etchant. The removed amount was calculated using the photocurrent density measured during PEC-CARE according to Faraday’s law. In each experiment, the amount was set to 200 µg, equal to a 50 nm thick substrate, by adjusting the processing time and monitoring the amount of current applied.
(a) Schematic and (b) cross-sectional profile of the PEC-CARE instrument. The GaN substrate is placed in the holder and pushed toward the polishing pad. The back of the GaN substrate is connected to the Pt wire immersed in the etchant via an amperemeter and a DC power source. The Pt wire is grounded, and it functions as the cathode. (c) Detailed view of the interface between the GaN surface and etchant within the red square shown in (b). At the interface between the GaN surface and etchant, a depletion layer is formed, and the electrons and holes generated within the depletion layer are transferred to the bulk and surface layers, respectively.
(a) Schematic and (b) cross-sectional profile of the PEC-CARE instrument. The GaN substrate is placed in the holder and pushed toward the polishing pad. The back of the GaN substrate is connected to the Pt wire immersed in the etchant via an amperemeter and a DC power source. The Pt wire is grounded, and it functions as the cathode. (c) Detailed view of the interface between the GaN surface and etchant within the red square shown in (b). At the interface between the GaN surface and etchant, a depletion layer is formed, and the electrons and holes generated within the depletion layer are transferred to the bulk and surface layers, respectively.
Two commercial freestanding non-doped GaN (0001) substrates prepared by the HVPE and Na-flux methods were used as a sample. These are subsequently referred to as HVPE-GaN and Na-flux GaN, respectively. Secondary ion mass spectrometry (SIMS) measurements showed that the main unintentional impurity was oxygen and that the conductivity type of each substrate was n-type. The surface smoothness/roughness, with root-mean-square (RMS) and peak-to-valley (PV) values, was obtained by phase-shift interference microscopy (ZYGO NewView 7100). The optical interferometry images of these substrates are shown in Figs. 2(a) and 2(d). The distribution of the impurity concentration was evaluated using cathodoluminescence (CL) with a scanning electron microscope (HORIBA DF-100-OU), and the obtained CL images are shown in Figs. 2(b) and 2(e). The emission intensity increases with the oxygen impurity concentration because oxygen acts as a donor and promotes the recombination of the electron–hole pairs.28 The region not grown along the c-axis (denoted the {}-growth region in HVPE-GaN and {101}-growth region in Na-flux-GaN) is observed as a bright area in the CL image owing to its higher oxygen impurity concentration than the region grown along the c-axis [denoted the c-growth region, Figs. 2(c) and 2(f)].
(a) Optical interferometric profile image (PV 26.21 nm, RMS 0.49 nm) and (b) CL image and (c) CL image with labeled areas with different etching properties for reference of the pre-processed surface of the HVPE-GaN substrate. (d) Optical interferometric profile image (PV 37.17 nm, RMS 0.98 nm) and (e) CL image and (f) CL image with labeled areas with different etching properties for reference of the pre-processed surface of the Na-flux-GaN substrate.
(a) Optical interferometric profile image (PV 26.21 nm, RMS 0.49 nm) and (b) CL image and (c) CL image with labeled areas with different etching properties for reference of the pre-processed surface of the HVPE-GaN substrate. (d) Optical interferometric profile image (PV 37.17 nm, RMS 0.98 nm) and (e) CL image and (f) CL image with labeled areas with different etching properties for reference of the pre-processed surface of the Na-flux-GaN substrate.
III. RESULTS AND DISCUSSION
Figures 3(a)–3(e) show the distribution of the removed amount on the HVPE-GaN substrate polished by PEC-CARE at several bias voltages. These images were obtained by subtracting the height distribution of the etched surface from that the pre-processed surface. A color scale indicates the relative removed amount; the higher the value, the larger the relative etched amount. As the applied bias increased from 0.0 to 10.0 V, the difference in the removed amounts in c-growth and {102}-growth regions decreased, particularly at 10.0 V where the difference was less than 1 nm. Upon continuing the polishing at an applied voltage of 10.0 V, as shown in Fig. 3(f), surface roughness RMS was significantly reduced from 0.49 to 0.14 nm, leading to the formation of a flat surface without crystallographic damage and a footprint caused by the nonuniform crystallographic properties. Surface smoothing is observed in five randomly selected locations on the substrate, and the average RMS decreased from 0.48 to 0.17 nm. The AFM image in Fig. 3(g) shows that the etched HVPE-GaN substrate surface had atomic smoothness, including a step-and-terrace structure. At 10.0 V, the removal rate increased to 1000 nm/h, which was ten times higher than that obtained through PEC-CARE without an applied voltage. It is considered that the expansion of the depletion layer leads to an increasing number of holes assisting surface oxidation. In contrast, in the case of the Na-flux-GaN substrate, the difference in the removed amounts is >10 nm even if the bias was set to 10.0 V, as shown in Figs. 4(a) and 4(b).
(a) Distribution of the removed amount (PV 26.64 nm, RMS 1.24 nm) from the HVPE-GaN substrate etched by PEC-CARE and (b) distribution of the removed amount with a label indicating the region with different etching properties. Distribution of the removed amount from the HVPE-GaN substrate etched by PEC-CARE under the condition of applying bias of (c) 1.5 V (PV 15.83 nm, RMS 1.07 nm), (d) 3.0 V (9.36 nm, RMS 0.63 nm), and (e) 10.0 V (PV 6.89 nm, RMS 0.42 nm). (f) Optical interferometric profile image (PV 1.07 nm, RMS 0.14 nm) and (g) AFM image (PV 1.14 nm, RMS 0.12 nm) of the HVPE-GaN substrate surface after PEC-CARE processing at 10.0 V.
(a) Distribution of the removed amount (PV 26.64 nm, RMS 1.24 nm) from the HVPE-GaN substrate etched by PEC-CARE and (b) distribution of the removed amount with a label indicating the region with different etching properties. Distribution of the removed amount from the HVPE-GaN substrate etched by PEC-CARE under the condition of applying bias of (c) 1.5 V (PV 15.83 nm, RMS 1.07 nm), (d) 3.0 V (9.36 nm, RMS 0.63 nm), and (e) 10.0 V (PV 6.89 nm, RMS 0.42 nm). (f) Optical interferometric profile image (PV 1.07 nm, RMS 0.14 nm) and (g) AFM image (PV 1.14 nm, RMS 0.12 nm) of the HVPE-GaN substrate surface after PEC-CARE processing at 10.0 V.
(a) Distribution of the removed amount on the Na-flux-GaN substrate etched by PEC-CARE under the condition of applying 10.0 V (PV 18.10 nm, RMS 5.54 nm) and (b) removed amount distribution with a label indicating the region of different etching properties.
(a) Distribution of the removed amount on the Na-flux-GaN substrate etched by PEC-CARE under the condition of applying 10.0 V (PV 18.10 nm, RMS 5.54 nm) and (b) removed amount distribution with a label indicating the region of different etching properties.
Estimated widths of the depletion layers of the (a) HVPE-GaN and (b) Na-flux-GaN substrates. The vertical and horizontal axes, respectively, represent the depletion layer width and applied bias. The circles and squares, respectively, indicate the depletion layer widths of the c-growth and {102} or {101}-growth regions. (c) Spectra of the irradiated UV light before (gray line) and after (black line) application of the filter. The distributions of the amounts removed from the Na-flux-GaN substrate, which was simultaneously etched by applying 3.0 V and UV light irradiation (d) with the filter (PV 62.34 nm, RMS 12.71 nm) and (e) without the filter (PV 1.40 nm, RMS 0.12 nm). (f) Optical interferometric profile (PV 8.58 nm, RMS 0.41 nm) and (g) AFM images (PV 0.77 nm, RMS 0.08 nm) of the surface of the Na-flux-GaN substrate after PEC-CARE processing at 3.0 V with the filter.
Estimated widths of the depletion layers of the (a) HVPE-GaN and (b) Na-flux-GaN substrates. The vertical and horizontal axes, respectively, represent the depletion layer width and applied bias. The circles and squares, respectively, indicate the depletion layer widths of the c-growth and {102} or {101}-growth regions. (c) Spectra of the irradiated UV light before (gray line) and after (black line) application of the filter. The distributions of the amounts removed from the Na-flux-GaN substrate, which was simultaneously etched by applying 3.0 V and UV light irradiation (d) with the filter (PV 62.34 nm, RMS 12.71 nm) and (e) without the filter (PV 1.40 nm, RMS 0.12 nm). (f) Optical interferometric profile (PV 8.58 nm, RMS 0.41 nm) and (g) AFM images (PV 0.77 nm, RMS 0.08 nm) of the surface of the Na-flux-GaN substrate after PEC-CARE processing at 3.0 V with the filter.
Parameters used to calculate the depletion layer width.
| ɛs | 9.5 | ||
| ɛ0 (Fm−1) | 8.8 × 10−12 | ||
| VD (V) | 1.7 | ||
| e (C) | 1.6 × 10−19 | ||
| HVPE-GaN | c-Growth region | 7.0 × 1015–1.0 × 1017 | |
| N (cm−3) | {}-Growth region | 1.0 × 1018–3.0 × 1019 | |
| Na-flux-GaN | c-Growth region | 5.0 × 1016–5.0 × 1017 | |
| {}-Growth region | 1.0 × 1018–1.0 × 1020 | ||
| ɛs | 9.5 | ||
| ɛ0 (Fm−1) | 8.8 × 10−12 | ||
| VD (V) | 1.7 | ||
| e (C) | 1.6 × 10−19 | ||
| HVPE-GaN | c-Growth region | 7.0 × 1015–1.0 × 1017 | |
| N (cm−3) | {}-Growth region | 1.0 × 1018–3.0 × 1019 | |
| Na-flux-GaN | c-Growth region | 5.0 × 1016–5.0 × 1017 | |
| {}-Growth region | 1.0 × 1018–1.0 × 1020 | ||
Figure 5(a) shows that the depletion layer can be expanded more than the UV-light penetration depth in the HVPE-GaN substrate at 10.0 V bias, resulting in a uniform oxidation rate, which is in agreement with the processing result. Conversely, even if 10.0 V is applied to the Na-flux-GaN substrate, Fig. 5(b) indicates that the width of the depletion layer remains lower than the UV-light penetration depth because the oxygen impurity concentration of the {101}-growth region is considerably high. Although a significantly higher bias can expand the depletion layer, the application of such a high bias voltage is undesirable in order to avoid the formation of gaseous products, which negatively affects the process condition. To overcome this problem, in addition to bias application, we attempted to employ UV light with a shorter wavelength, causing the UV-light penetration depth to be lower than the depletion layer width. The high-pass filter (HAMAMATSU A9616-11), which was installed over the light source, limits the wavelength of the irradiated light to <250 nm, as shown in Fig. 5(c), shortening the penetration depth of the UV light from 75 to 40 nm. When the penetration depth is 40 nm, as indicated by the red line shown in Fig. 5(b), only the application of 3.0 V suppresses the fluctuation in the oxidation rate. Figures 5(d) and 5(e) show the distributions of the amounts removed after PEC-CARE processing using UV irradiation with different penetration depths at an applied potential of 3.0 V. Upon shortening the penetration depth of the UV light, even if the applied bias potential is 3.0 V, the fluctuation in the oxidation rate is suppressed, as shown in Fig. 5(e). After PEC-CARE processing, RMS decreases from 0.98 to 0.41 nm, as indicated by the optical interferometry image shown in Fig. 5(f), and the average RMS, as calculated based on five randomly selected locations, is improved from 0.98 to 0.45 nm. Figure 5(g) shows the AFM image of the atomic smoothness, which reveals a stepped terrace, similar to that observed following CARE processing without UV irradiation and bias application.
IV. CONCLUSION
We have discovered that by adjusting the GaN potential based on the depletion layer’s width and the UV light’s penetration depth, the GaN surface can be efficiently planarized. In recent years, the search for conditions that can reduce defects and oxygen impurity concentrations has been ongoing. By merely adjusting the penetration depth of the irradiated light on a GaN substrate subjected to PEC-CARE, we can obtain an atomically flat surface with a high removal rate. This advancement can boost the productivity and enhance the quality of GaN-based devices.
ACKNOWLEDGMENTS
This research was partially supported by the Japan Society for the Promotion of Science (Grant No. 21H05004).
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
D. Toh: Conceptualization (lead); Data curation (lead); Formal analysis (equal); Methodology (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). K. Kayao: Data curation (equal); Formal analysis (equal); Writing – original draft (equal); Writing – review & editing (equal). R. Ohnishi: Conceptualization (equal); Data curation (equal); Formal analysis (lead); Methodology (lead); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). A. I. Osaka: Data curation (equal); Formal analysis (equal). K. Yamauchi: Methodology (lead); Writing – original draft (equal); Writing – review & editing (equal). Y. Sano: Conceptualization (equal); Methodology (equal); Writing – original draft (supporting); Writing – review & editing (equal).
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request.