Simple Wet Etching Of GaN
Simple Wet Etching of GaNG. Parish*, P.A. Scali, S.M.R. Spaargaren, B.D. NenerDept. of Electrical and Electronic Engineering, The University of Western AustraliaABSTRACTWe discuss investigations into a contactless UV-enhanced wet etching technique for GaN. The technique utilises theoxidising agent potassium persulfate to consume photogenerated electrons, thus avoiding the need for an electricalcontact to an external cathode. The etch rate is strongly dependent on illumination intensity and uniformity and on thepH of the KOH solution, as is the roughness of the etched surface. The implementation of a dual illumination schemewhereby an additional UVC lamp was used to illuminate only the solution and not the wafer, resulted in an increasedetch rate and smoother etched surface. Finally, the ohmic nature of contacts deposited on n-type GaN that had beenetched in this manner was found to be improved compared to contacts on the unetched surface.Keywords: GaN, gallium nitride, wet etch, UV, PEC, processing, surface1. INTRODUCTIONGaN and its alloys with InGaN and AlGaN form a wide-bandgap semiconductor material system with numerous opticaland electronic device applications. The wavelength range spans from 1.9eV (InN) to 6.2eV (AlN), covering thetechnologically important ultraviolet (UV) and visible spectral ranges. Also, due to the wide bandgap and high bondstrength the material has a high chemical and radiation resistance. With no semiconductor material previouslysatisfying commercial demands for blue, green and UV lasers and light-emitting devices this was the first immediatefocus of GaN research and progress, and these are now commercially available. With a high breakdown field and largepredicted electron saturation velocity, GaN-based materials are also suited for high power, high frequency transistorsfor microwave applications. In this area commercial potential is close to being realized, using modulation doped fieldeffect transistors. The spectral range, which has been utilized with success for light emitting devices, is also beingexploited for detectors in the UV and visible spectrum.Although some device applications have been or are close to being commercially realised, there still remain manyimpediments to the full realisation of the potential of this material system. Material quality has been hindered by lack ofa suitable substrate1 and doping difficulties2. This has meant that other, equally promising, GaN-based devices such asbipolar junction transistors (BJTs), heterojunction bipolar transistors (HBTs) and solar-blind UV photodetectors, so fardemonstrate insufficient performance to challenge more traditional market solutions to their applications.A significant challenge in GaN research is the chemical and physical robustness of (Al,In)GaN, due to the high bondstrengths and wide bandgap. This is advantageous for application of GaN-based devices in harsh environments andunder extreme operating conditions, however it causes great impediments to device processing. Perhaps one of the mostdifficult areas has been etching3. The most common technique used has been chlorine-based reactive ion etching (RIE),however due to the high plasma voltages required this results in damage to the etched surface3,4. Ohmic contacts havebeen another area of processing complexity. Reasons for this include the wide bandgap, surface oxides, and theprevalence of point defects, which compensate doping and affect the surface potential5,6. Ohmic contacts to RIE-etchedsurfaces pose particular problems due to the etch damage sustained.An attractive alternative to ion etching that has been investigated for some time is photoelectrochemical (PEC) etching.In this technique, chemical etching using a base (such as KOH) or acid (such as HCl) is enabled using UV light ofphoton energy greater than the bandgap of the material being etched. Photogenerated holes at the surface excite thesurface Ga to higher oxidation states and this oxide is then dissolved. This technique was first demonstrated for GaN byMinsky et al.7 and has since been investigated by many researchers. Recently, Zhu et al.8 compared GaN n--n rectifiers*giap@ee.uwa.edu.au; phone 61 8 9380 3390; fax 61 8 9380 1095; Dept. of Electrical and Electronic Engineering/TheUniversity of Western Australia, 35 Stirling Hwy Crawley WA 6009 Australia
which had been etched by RIE to those etched by PEC technique. The latter had superior properties, confirming theadvantages of the PEC process.A somewhat modified PEC process was introduced by Bardwell et al. in 19999. They employed the oxidising agentpotassium persulfate to consume photogenerated electrons. This circumvented the need for the electrical contact used inconventional photoelectrochemical techniques. Maher et al10 later reported success in optimising the technique,achieving minimal increase in roughness (1nm increase in RMS roughness compared to pre-etch surface) with amoderate etch rate of 2nm/min.This paper will detail the investigations at UWA of this contactless UV-enhanced wet etching technique for GaN. Wewill discuss the investigation of the effect of etch parameters such as illumination intensity and wavelength, pH of thesolution, and method of illumination. Atomic force microscopy (AFM) of the surface before and after etching was usedto assess both the etch rate and the quality of the etched surface. Finally, the quality of ohmic contacts deposited on ntype GaN that has been etched in this manner will be addressed.2. EXPERIMENTALSeveral different experimental schemes were used during the experiments. Wafers with a Pt mask pattern (Pt thickness15, 50 or 150nm) were illuminated in a stirred KOH/K2S2O8 solution. Adjustment of molar concentration of the KOHwas used to vary solution pH. A variety of light sources were used, either separately or in combination. A HeCd laserwas used, with either an intensity of 1W/cm2 or 10mW/cm2 at 325nm. The latter was achieved by using a beamexpander consisting of a concave-convex lens combination (Galilean principle). Also used was a 253.7nm low pressureHg vapour discharge lamp (with a parabolic reflector to produce parallel output) that had an intensity of 1.1mW/cm2 atthe distance used in the experiments. Wafers were either held in place using a plastic clip (for illumination from theside) or placed face-up on a perforated plastic basket above the stirrer (for illumination from above).All etches were performed on pieces of GaN taken from the same 2” template. The template (on sapphire substrate) was2µm thick and Si-doped with a carrier concentration of 1-2x1018cm-3. High-resolution x-ray diffraction of templatesgrown under similar conditions showed an on-axis (0002) full-width-half-max (FWHM) of approximately 250arcsecand an off-axis (10 1 2) FWHM of approximately 550arcsec. The RMS roughness as measured by AFM was less than1nm. AFM images were taken on DI Nanoscope (III) operated in contact mode. Optical microscope images wereobtained using an Olympus IP11 Digital Microscope and field-effect secondary electron microscope (FE-SEM) imageswere obtained using a JEOL JSM-6300F.3. RESULTS AND DISCUSSION2.1 IlluminationThe significance of the choice of light source is evident when reviewing the results of Bardwell et al. In their recentpaper11, they discuss the particular importance of the wavelength range for this contactless technique. They foundgeneration of OH and/or SO4- radicals from the K2S2O8 was required to facilitate the cathodic consumption ofphotogenerated electrons. These radicals are by-products of the photolysis of K2S2O8 by illumination at or below310nm. They also believe these radicals are produced via a catalytic reaction at the mask surface if Pt is used as themask. Use of a Pt mask was found to not only enhance the etch rate when illumination below 310nm was used, but toenable etching at higher wavelengths (such as 365nm).The first etch achieved in this research was with the HeCd laser beam at the full intensity of 1W/cm2. The solution pHwas 12.2, with 0.02M K2S2O8. The sample was etched for 10 minutes and was found to be etched to the substrate,indicating a etch rate of at least 200nm/min. This rapid etch resulted in an extremely rough etched surface, with an RMSroughness of over 300nm when measured over a 100µm2 area. Figure 1 shows a 30µm by 30µm image of the etchborder. The large etch height and roughness meant that a clear image could not be obtained. The 15nm thick Pt maskhas not been removed and it can be seen that near the edges the mask appears to be delaminating. This occurred formany of the wafers and the cause is still under investigation.
Figure 1: AFM image of GaN:Si etched using 325nm laser illumination of 1W/cm2. The left side of the image is the unetched region,including the 15nm Pt mask. The greyscale is 2.5µm.Another GaN piece was etched using the expanded beam with an intensity of 10mW/cm2, and a similar solution pH(though a slightly stronger K2S2O8 concentration of 0.05M). In comparison to the first etch, an etch rate of around5nm/min was observed. (Note that this is only an approximation. Exact determination of etch heights by AFM was verydifficult, due to the roughness of the etched surface, and because the etch steps were not abrupt.) Thus a two orders ofmagnitude reduction in intensity resulted in a two orders of magnitude reduction in etch rate. This extends the results ofBardwell et al.9, who reported a linear dependence of etch rate on intensity using an unfiltered Hg lamp with intensityvaried between 1 and 25mW/cm2 at 365nm. This dependence has also been seen under certain situations by researchersinvestigating standard PEC etching of GaN7,12. Youtsey et al.12 noted that it indicates the reaction rate is proportional tothe electron-hole pair generation (carrier-limited regime). At higher light intensities they observed a saturation of thisrelationship. For example, for a solution of 0.01M KOH, saturation was observed at an intensity of 15mW/cm2 at365nm. They suggested that this saturation was indicative of a diffusion-limited regime, in which the etch rate waslimited by the rate of arrival from the solution of reactants needed to dissolve the surface oxide.The fact that a saturation in the intensity-etch rate relationship was not observed over two orders of magnitude variationin light intensity (despite a similar KOH concentration to Youtsey et al.’s experiments) requires careful consideration ofthe different mechanisms involved in the contactless etching technique. Perhaps it is more instructive to compare thesurface oxide formation- and dissolution-limited regimes (which correspond to the carrier- and diffusion-limitedregimes in the above discussion). In the case of the contactless technique, the oxide formation mechanism depends notonly on the generation of electrons and holes by above-bandgap photon energy, but also on the availability of reactantsin the solution to consume the electrons as the holes oxidise the surface. When using the 325nm illumination, theproduction of the necessary radicals occurs only because the Pt acts as a catalyst. However they would only beproduced locally on the illuminated mask, and not in the rest of the solution. Perhaps therefore, the dependence on theintensity is due to a limitation in the production of these radicals which continues to limit the progression of the etchrate even at much higher intensities (at the higher intensity the beam diameter was on the order of 1mm or less,providing an even smaller area of radical production). The etch rates at similar pH and intensity are certainly muchslower than those obtained from PEC (at least an order of magnitude lower), which would appear to support this theory.This strong dependence of etch rate on light intensity has tremendous implications for the experimental setup andchoice of illumination source. The etch uniformity will obviously be dependent on the uniformity of the light intensity.Thus, the use of the laser beam, even with the beam expander in place, resulted in non-uniform etching across the waferarea (pieces were sized approximately 8mm by 8mm or larger). This was due to the Gaussian distribution of the beam.Furthermore, imperfections in the lenses used for the beam expander resulted in localised areas where there was noetching. Another aspect which affected the etch uniformity was the method of light projection onto the wafer. When thelight was projected through the side of the beaker with the wafer held vertically, imperfections in the glass also led to apatterning effect of the etch. This is evident in the image given in Figure 2.
Figure 2: Microscope image of GaN:Si etched using 325nm laser illumination of 10mW/cm2, with light incident through the side ofthe beaker holding the solution and the wafer. The black areas correspond to the etched regions.The black areas correspond to the (rough) etched surface. It can be seen that the light was patterned in stripes, withsome circular diffraction-like patterns also present. It should be noted that the non-uniform light distribution incident onthe wafer was not visible to the eye. However, the resulting etch pattern was clearly visible. When a mirror was used toreflect the light directly onto the wafer (held horizontally) from above, this particular patterning did not occur.As already mentioned, several different illumination schemes were investigated in this work. An experiment wasdevised in which two GaN pieces were etched using the laser light at 325nm incident directly on the (horizontal) wafer,under identical conditions except that for the second piece the solution was illuminated from the side using the 253.7nmlamp. The pH was 12.5 with 0.05M K2S2O8, and a 50nm Pt mask was used. The placement of the lamp on the sidemeant that it would enable photolysis of the K2S2O8 without contributing to photogeneration of carriers in the GaN.This resulted in two effects. The etch rate was increased nearly threefold from approximately 5nm/min to approximately13nm/min. Conversely, the RMS roughness was reduced from 130nm to 70nm in a 4µm2 area. These resultscomplement both Bardwell et al.’s results and the above results regarding etch rate dependence on light intensity. Theincreased production of the necessary radicals via illumination of the entire solution with the 253.7nm light would haveincreased the etch rate and moved the etching away from the oxide-formation-limited regime to the oxide-dissolutionlimited regime. Youtsey et al. had shown that etching in the latter regime (the diffusion-limited regime) resulted in asmoother etch, which is consistent with these results.2.2 Solution pHBardwell et al.9 had found that the etch rate was dependent on solution pH, with a peak etch rate occurring at 12.7.However, no information was given as to the variation in etch quality. Our investigations therefore included a series ofetches in solutions of different pH, for which both etch rate and etched surface roughness were measured. Figure 3 plotsthe results of these measurements. For this series the expanded 325nm laser beam and a K2S2O8 molarity of 0.05M wereused, with a Pt mask thickness of 150nm. The etch time was 20 minutes. Unfortunately the illumination was through theside of the beaker, resulting in the etch patterns seen in Figure 2. Etch rate therefore had to be estimated by measuringthe step heights between the dark (etched) and light (non-etched) regions. There was significant variation in themeasured heights on a given wafer, so the etch rates given are representative of the differences in etch rate as pH wasvaried, rather than being measurements of the absolute etch rate. It can be seen from the figure that the variation in etchrate with pH, which matches that seen by Bardwell et al., is closely matched by the variation in etch roughness. That is,as the pH was decreased, the etch rate decreased, but so did the etch roughness. This again agrees with the PEC etchfindings of Youtsey et al., that by moving to the oxide-dissolution- (diffusion-) limited regime the smoothness wasimproved. The reason for the increase in etch rate at lower solution pH values (less than 12) is yet to be determined.
160201401812014100128010860640RMS Roughness (nm)Etch Rate (nm/min)16420211.812.012.212.412.612.813.0Etching solution pHFigure 3: Etch rate and roughness of GaN:Si etched using 325nm laser illumination of 10mW/cm2, as a function of solution pH(KOH concentration). Light was incident through the side of the beaker holding the solution.2.3 Pt maskThere are several as-yet unresolved issues regarding the Pt mask. As was evident in Figure 1, the mask frequentlyappeared to be delaminating at the edges. The FE-SEM image in Figure 4a is an extreme example of cracking andpeeling of a mask that was 15nm thick. In another experiment, two pieces with different Pt mask thicknesses wereetched under identical conditions using the expanded 325nm laser beam. In that case, the 15nm mask appearedunaffected however peeling occurred on the second piece, which had a mask thickness of 150nm. A microscope imageof this is given in Figure 4b. The affected area was limited to where the mask had been illuminated during etching. It isworthwhile to note that there was no difference in etch rate or etched surface roughness for the two mask thicknesses.Figure 4: a) FESEM image of 15nm Pt mask after 30 minutes etching. b) Microscope image of 150nm Pt mask after 20 minutesetching. In both cases 325nm laser illumination of 10mW/cm2 was used.Another perplexing problem is that in most cases no etching occurred immediately in the vicinity of the mask edge.This can be seen in Figures 5a and 5b. Figure 5a is an optical microscope image of a section of the mask for a waferetched using the expanded laser beam incident from above. Figure 5b is an AFM image in the vicinity of the mask edgeof the wafer etched using both the laser beam and the 253.7nm lamp. (These two wafers were those discussed insubsection 2.1)
Figure 5: a) Microscope image of GaN:Si etched using 325nm laser illumination of 10mW/cm2. The black area corresponds to theetched region b). AFM image of GaN:Si etched using 325nm laser illumination of 10mW/cm2 and additional illumination of thesolution by a 253.7nm lamp at 1.1mW/cm2. The greyscale is 400nm.In both images, it can be seen that there is a distinct gap between the etched region and the mask edge. This gap was notuniform, and not always present. Often there was a gradual increase in etch rate away from the gap, which meant thatthe steps weren’t abrupt. This further compounded the difficulty in determining etch rates.2.4 Ohmic contactsAlthough the etch technique has not yet been perfected, preliminary investigations were made as to the nature of ohmiccontacts to the etched surface. The pH 12.13 wafer from the pH series described in subsection 2.2 was used. Alcontacts were deposited by thermal evaporation on both the etched and unetched regions. The contacts were notannealed. I-V measurements were made between adjacent contacts in each of the regions. Typical curves for eachregion are shown in Figure 6. It can be seen from the linearity that the contacts to the etch region are more ohmic.Obviously, the roughness of the etched surface is beneficial rather than detrimental to the nature of the contact.Furthermore there are no compensating defects introduced by the etching process.0.100.08etched0.06unetchedCurrent 012345Bias Voltage (V)Figure 6: I-V characteristics between adjacent Al contacts deposited on etched and unetched regions of GaN:Si.
4. SUMMARYWe have presented investigations into a contactless UV-enhanced wet etching technique for GaN using a K2S2O8/KOHetch solution. This etch is a two-step process. Above-bandgap UV light creates electron-hole pairs in the GaN, whichresults in oxidation of the surface atoms by the photogenerated holes. The oxide is then dissolved in the KOH sol
enable etching at higher wavelengths (such as 365nm). The first etch achieved in this research was with the HeCd laser beam at the full intensity of 1W/cm2. The solution pH was 12.2, with 0.02M K2S2O8. The sample was etched for 10 minutes and was found to be etched to the substrate, indicating a etch rate of at least 200nm/min.
Etching is a process of removing material from the substrate’s surface. In general, there are two categories that etching can be divided into dry etching, and wet etching. The focus of this section will be solely on dry etching. Wet etching, a process where the substrate is submerged
etching is usually faster than the rates for many dry etching processes and can easily be changed by varying temperature or the concentration of active species. Wet Etch Synonyms: chemical etching, liquid etching Definition: Wet etching is a material removal process that uses liquid chemicals or etchants to remove materials from a wafer.
Review: GaAs etching overview; wet and dry etching; Ref. (Ashby, C.I.H., 1990a) Review: InP wet chemical etching; with (1) defect or damage revealing etchant table, (2) polishing etchant table, and (3) pattern etchant table; Ref. (Adachi, S., 1990b) Review: wet and dry chemical etching of GaAs; classifies wet etchants as non-electrolyte (those with
Plasma Etching Page 2 OUTLINE Introduction Plasma Etching Metrics – Isotropic, Anisotropic, Selectivity, Aspect Ratio, Etch Bias Plasma and Wet Etch Summary The Plasma State - Plasma composition, DC & RF Plasma Plasma Etching Processes - The principle of plasma etching, Etching Si and SiO2 with CF4
Al etching start and thus to diff erent etching depths or times (Fig. 118). The formation of hydrogen in the etching reaction is also problematic for a homogeneous etching result. The constantly produced H 2 bubbles stick to the surface and block the etching process through a sup
Dry plasma etching has become the dominant patterning technique for the group-III nitrides, due to the shortcom-ings in wet chemical etching. Plasma etching proceeds by either physical sputtering, chemical reaction, or a combination of the two often referred to as ion-assisted plasma etching, Physical sputtering is dominated by the
After etching, line width and the length only for the tail along the circuit was measured to obtain etching factor. Etching factor was defined as shown in Figure 3. From the etching test results, etching factor improved when matte side surface roughness decreased. Flatter foil seemed to be better to create narrower traces.
Photoelectrochemical (PEC) etching is a rapid and inexpensive means of etching GaN, InGaN, and related materials for micro-electro-mechanical systems (MEMS) applications. In this work, we show that bandgap engineering of GaN/InGaN heterostructures can be used to exert substantial control over PEC etching and achieve strain-free cantilevers.
