Thermal-assisted contactless photoelectrochemical etching for GaN

Advanced contactless photoelectrochemical etching for GaN was conducted under the condition that the sulfate radicals (SO4·−) as the oxidizing agent were mainly produced from the S2O82− ions by heat. The generation rate of SO4·− was determined from the titration curve of the pH in the mixed solutions between KOH (aq.) and K2S2O8 (aq.); it clearly increased with an increase in the S2O82− ion concentration. The highest etching rate of >25 nm min−1 was obtained in the “alkali-free” electrolyte of 0.25 mol dm−3 (NH4)2S2O8 (aq.) at 80 °C, which was approximately 10 times higher than that reported by previous studies.

G allium nitride (GaN) is widely used for electronic devices such as mobile base stations, to meet the 5G application demands. 1) In addition, GaN power devices have recently attracted considerable research attention as energy-saving solutions, because of their low specific on-resistance (R on ) coupled with a high breakdown voltage (V B ). 2) These advantages arise from GaN's high-electron-drift velocity and high breakdown field, compared with those of Si or GaAs. 3) The etching process is essential for fabricating GaN power device structures, e.g. isolations, mesas, trenches, and gate recesses. In general, GaN is etched by inductively coupled plasma reactive ion etching (ICP-RIE). 4,5) ICP-RIE is known as a high-throughput process, but it sometimes damages the GaN surface. Recently, it was reported that the plasma damage layer of the GaN surface can be removed using a photo-assisted electrochemical (PEC) etching process. 6) In addition, PEC etching can be applied as a damage-less process to the mesa fabrication of GaN pn junction diodes, 7,8) vertical trenches, 9) and gate recesses for highelectron-mobility transistors (HEMTs). 10,11) Moreover, it produces a high device yield and good performance, compared with devices fabricated using conventional dry-etching processes. 8) Thus, recently, this PEC-etching feature has received considerable attention.
Many reports about GaN PEC etching have been published since 1996. [12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31] Photo-assisted anodic oxidation is the basis of the PEC etching of GaN. 13,14,16,18,20,[22][23][24]29) In addition, recently, simple contactless PEC (CL-PEC) etching methods have been reported such as simply dipping the sample into the solution under UVC illumination. [25][26][27][28][29][30][31][32][33] In the usual CL-PEC etching process, an electrolyte containing an oxidizing agent, K 2 S 2 O 8 (aq.), is used, which absorbs UVC of less than 310 nm. 26,[32][33][34] In a previous study, we chose a deep-UV flexible surface light source with luminous array film (LAFi) technology, which has a low cost and low energy consumption at the wavelength of 260 nm. Furthermore, we succeeded in fabricating a precisely controlled recess-gate HEMT structure by CL-PEC etching. The details of the recent CL-PEC etching technologies are described in the related papers. 35,36) We believe that this CL-PEC technology is one of the best solutions for fabricating recess-gate HEMTs on (GaN/SiC/Si) substrates from the viewpoint of the reproducibility of the thickness of the residual barrier layer, the smooth roughness of the etched surface, and the low process cost. 35,36) The CL-PEC etching rate is 0.5-1.0 nm min −1 , which is a sufficiently good rate for the recess process compared with the advanced low-damage dry-etching process, such as neutral beam etching 37,38) and atomic layer etching. 39,40) However, this etching rate is not sufficient to achieve the through-via fabrication process, which requires an etching rate of more than 100 nm min −1 . In this paper, we describe the approach for a considerably enhanced CL-PEC etching rate of GaN.
The overall CL-PEC reaction of GaN in the present redox system could be represented as follows: where the description of intermediate products (such as Ga 2 O 3 ) is omitted for simplification. As described in reaction (1), photo-generated holes were used for the oxidation of the GaN surface, whereas photo-generated electrons were consumed by the sulfate radicals (SO 4 ·− ) making it contactless. 26 4 ·− in terms of redox potential with respect to the conduction band and the valence band edge of GaN. 29) In addition, the SO 4 ·− radicals behave as hole-injector into valence band of GaN and/or the electron-acceptor from conduction band. 29 The photo-generated SO 4 ·− was also used for the cathode reaction of GaN CL-PEC etching. From the viewpoint of the radical supply, it was simply better to use dense K 2 S 2 O 8 (aq.).
In CL-PEC etching, the SO 4 ·− radical supply probably limited the etching rate, although the efficiency of the cathode reaction seemed to be very low around 1%-2%. 41) The low efficiency seems to be originated the short lifetime of SO 4 ·− radical, and the spatially far SO 4 ·− radicals from GaN surface does not contribute to the cathode reaction. Alternatively, the following reaction would occur by mainly consuming the sulfate radicals (SO 4 ·− ) in the solutions: Thus, the SO 4 ·− generation rate was estimated for the pH change in the solution by using Eqs. (3) and (4). The titration curve described as a function of the consumption (production) rate of the OH − (H + ) ion x = d[H + ]/dt and time t at the temperature T by 33) = -- where t n is the time it takes for the reaction to reach the neutralization point with pH = −0.5 × log 10 K w (T), which corresponds to [OH − ] = [H + ]. Based on these fundamental, we measured the transmittance of the dense K 2 S 2 O 8 (aq.) for choosing the light source and the titration curve for confirming the SO 4 ·− generation by heat. Figure 1 shows the transmittance of K 2 S 2 O 8 (aq.) in the range of 0-0.175 M (=mol dm −3 ), which was measured using a UV-visible spectrometer (Shimadzu, UV-1700) with a 10 mm path cell. Each K 2 S 2 O 8 (aq.) sample was prepared by dissolving K 2 S 2 O 8 powder [molecular weight (MW) = 270.33 g/mol] into water at room temperature, to prevent SO 4 ·− from forming because of the heat. [42][43][44] It shows the UVC absorption and the absorption edge shifting to a longer wavelength with the increasing K 2 S 2 O 8 (aq.) concentration. In a previous study, we chose the 0.025 M K 2 S 2 O 8 (aq.) concentration, which corresponded to half of the UVC with the wavelength of 260 nm consumed for the SO 4 ·− formation. 32,33,35,36) Then, the remaining half intensity was absorbed into the GaN surface for the photo-generated hole-electron-pair formation. However, Fig. 1 indicates that the UVC was not transparent when the K 2 S 2 O 8 (aq.) concentration was more than 0.125 M. In contrast, UVA with the wavelength of 365 nm, which corresponded to the bandgap of GaN, was transparent to dense K 2 S 2 O 8 (aq.). That is, if the SO 4 ·− generation rate by heat was comparable to that by UVC, the CL-PEC etching rate could be considerably enhanced because the hole generation into GaN would be available by UVA even in the electrolyte including dense K 2 S 2 O 8 (aq.). Thus, the approach for achieving a higher etching rate was simply based on a higher SO 4 ·− supply. As the first titration experiment, a mixed solution of 100:100 ml was prepared to measure the pH and the temperature while elevating the temperature, using a multi pH-meter (Eutech Instruments, PC700). The initial pH values of the 1:1 mixture of (0.001, 0.01, 0.1, 1.0) M KOH (aq.) and 0.05 M K 2 S 2 O 8 (aq.) were 4.4, 11.8, 13.0, and 13.9, respectively, at room temperature. Thus, we chose the 1:1 mixture of 0.01 M KOH (aq.) and 0.05 M K 2 S 2 O 8 (aq.) for this measurement to obtain the appropriate titration curve. Figure 2 shows the relationship between the pH and the temperature in the 1:1 mixed solutions of 0.01 M KOH (aq.) and 0.05 M K 2 S 2 O 8 (aq.). The inset represents the elevating temperature curve and the pH against time. It clearly indicates that a temperature elevation of more than 70°C was required for SO 4 ·− generation from S 2 O 8 2− ions by heat. The gradually decrease in pH in the range of 25°C-65°C was explained by the change in the ionic product for water K w (T). The calculated pH values as the neutralization point were 7.48, 7.0, 6.63, and 6.35 at each temperature of 0°C, 25°C, 50°C, and 75°C, respectively. 45) The blue broken line represents the calculation according to pH(T) = −log 10 (K w (T)/[OH − ] 25°C ) in Fig. 2.
As the second experiment, the mixed solution was prepared by directly dissolving K 2 S 2 O 8 powder into KOH (aq.) at 70°C. The pH values were measured within 10 s. Note that the temperature deviation after dissolving K 2 S 2 O 8 powder was only 1°C-2°C. Figure 3 shows the titration curves in the 1:1 mixed solutions of 0.01 M KOH (aq.) and 0.05-0.15 M K 2 S 2 O 8 (aq.) at 70°C. The solid and dashed lines corresponded to the fitting results, which were based on Eqs. (5) and (6) with the parameters from Table I. The fitting parameters were the consumption (production) rate of the  OH − (H + ) ions, x, in the basic and the acidic regions, respectively. This indicated that the SO 4 ·− generation rate probably increased with an increase in the S 2 O 8 2− ions in the solution, although the generation rate differed between the basic and the acidic region. 42) The difference in the basic and the acidic solution was larger than that of UVC, and the reason for this difference is still not clear. These results suggested that a higher etching rate would be obtained in a dense solution involving S 2 8 2− ion source because of the following benefits. Firstly, its high water solubility enabled us to prepare a dense solution at room temperature. Secondly, it is an "alkali-free" electrolyte, which minimized the contamination to other processes. In addition, the initial pH values of the electrolytes prepared by using only the persulfate salts were acidic in the range of 4.5-3.8, and it was appropriate for preventing the dramatic change in the SO 4 ·− generation rate between the basic and the acidic solutions.
A beaker with a diameter of 47 mm was set on a hot plate under the Hg lamp. The "alkali-free" electrolyte of (NH 4 ) 2 S 2 O 8 (aq.) was set at a depth of 10 mm, which corresponded to 17.3 ml. Compared with the titration measurement, we used a small amount of the electrolyte for preventing the NH 3 vapor from damaging the optics of the mask aligner system. The pH meter (Horiba, LAQUAtwin) and a thermocouple were used for this small setup. The electrolyte was also stirred at 200 rpm for minimizing the temperature un-uniformity. The sample chip was lifted up by using small 0.4-mm-thick sapphire chips, because the rear side of the sample worked as the cathode. 35) The two-inch free-standing GaN substrates used in this study were produced by our void-assisted separation method. [47][48][49] The epitaxial layers with Schottky barrier diode (SBD) structures were grown by metal-organic vapor-phase epitaxy on n-GaN substrates. The SBDs consisted of an n-GaN layer with a nominal Si concentration of 0.9 × 10 16 cm −3 and a 13 μm thickness.
SiO 2 was used for the etching masks. A 330-nm-thick SiO 2 mask was prepared by spin-on-glass and patterned by buffered hydrofluoric acid (BHF) with a photoresist mask. 32,35,36) The epitaxial wafer was cut into small pieces of approximately 6 mm × 6 mm. The etching depth was measured using a surface profiler (Sloan Dektak3 ST). Figure 4 shows the elevating temperature curve and the pH against time in this thermal-assisted CL-PEC etching experiment. The concentration of the electrolyte was 0.025 M and 0.25 M (NH 4 ) 2 S 2 O 8 (aq.), respectively. The pH smoothly decreased with time at the elevated temperature even when there was some deviation from the target temperature. Then, all of the samples were put in the electrolyte 10 min after switching the heater on, which corresponded to approximately 5 min after reaching the etching temperature. The etching time was set to 5 min for the elevated temperature and 30/60 min for the room temperature experiments, which corresponded to an etching depth in the range of 8-154 nm. Figure 5 shows the relationship between the etching rate and the electrolyte temperature. CL-PEC etching was observed even at room temperature, because the Hg lamp included a UVC component of approximately 2 mW cm −2 at the wavelength of 254 nm. In contrast, the etching rate clearly increased at an elevated temperature of more than 50°C, which was consistent with the titration measurement, as shown in Fig. 2. A higher etching rate was also obtained in the case of the dense 0.25 M (NH 4 ) 2 S 2 O 8 (aq.) electrolyte. These results clearly indicated that a higher SO 4 ·− generation rate led to a higher etching rate. Note that the reason for the low etching rate in the case of 0.025 M (NH 4 ) 2 S 2 O 8 (aq.) at elevated temperature was not clear, but it seemed to be caused by the exhaustion of S 2 O 8 2− ions in the solutions. The etched surface roughness gradually increased with increasing the etching depth even in this thermal-assisted CL-PEC etching. It indicates that the removing (etching) rate of  Table I. surface oxidized GaN is probably faster than that of the anodic oxidation.
In conclusion, we succeeded in considerably increasing the etching rate of GaN by using thermal-assisted contactless photoelectrochemical (CL-PEC) etching. This advanced CL-PEC etching of GaN was conducted under the condition that the sulfate radicals (SO 4 ·− ) as the oxidizing agent were mainly produced from S 2 O 8 2− ions by heat. The generation rate of SO 4 ·− was determined from the titration curve of the pH in the mixed solutions between 0.01 M KOH (aq.) and 0.05-0.15 M K 2 S 2 O 8 (aq.), and it clearly increased with an increase in the concentration of S 2 O 8 2− ions in the solutions. The highest etching rate of more than 25 nm min −1 was obtained in the "alkali-free" electrolyte of 0.25 M (NH 4 ) 2 S 2 O 8 (aq.) at 80°C, which was approximately 10 times higher than that reported by previous studies because of the high generation rate of SO 4 ·− . The higher etching rate could simply achieve the throughvia fabrication process. Moreover, these results indicated that the generation of the hole in GaN and that of SO 4 ·− in the electrolyte could be controlled independently in thermalassisted CL-PEC etching. Thus, it could also be applied to a lift-off process using a narrow band-gap InGaN layer, 50) which is fundamentally impossible for conventional CL-PEC etching because of the etching of the GaN device layer. In other words, thermal-assisted CL-PEC etching is a cuttingedge GaN wet process.