1. Introduction

The predominant plasma chemistries for dry etching of III-nitrides are based on Cl₂ [a] Reference Pearton, Shul, Pankove and Moustakas[2] Reference Shul and Pearton[3] Reference Adesida, Mahajan, Andideh, Asif Khan, Olsen and Kuznia[4] Reference Zhang, Ramer, Brown, Zheng, Lester and Hersee[5] Reference Vartuli, MacKenzie, Lee, Abernathy, Pearton and Shul[6] Reference Shul, McClellan, Casalnuovo, Rieger, Pearton, Constantine, Barratt, Karlicek, Tran and Schurman[7] Reference Pearton, Abernathy and Ren[8]. These have worked well for fabrication of laser and light-emitting diodes where the etching is non-selective Reference Nakamura, Mukai and Senoh[9] Reference Nakamura, Senoh, Nagahama, Iwasa, Yamada, Matsushita, Kiyoku and Sugimoto[10]. As attention turns to application of nitride materials in high power, high temperature electronics, there is increasing need for plasma chemistries that will remove In-based nitrides from underlying AlGaN alloy layers with high selectivity. It is expected that InN or InGaN contact layers will be necessary to produce acceptable contact resistance in transistor or thyristor devices. Besides Cl₂-based mixtures, there have been some reports of Br₂(in the form of HBr) Reference Ping, Adesida and Khan[11] Reference Pearton, Abernathy and Vartuli[12] and I₂(in the form of HI) Reference Pearton, Abernathy and Vartuli[12] plasma etching of nitrides. In particular it has long been recognized that InI_x etch products have higher volatility than the corresponding InCl_x species, making iodine an attractive etchant for InGaN alloys.

In this paper we describe use of two new dry etch chemistries for nitrides, namely BI₃ and BBr₃. Both are found to provide high etch selectivity for InN over both GaN and AlN. Vartuli et al. Reference Vartuli, Pearton, MacKenzie, Abernathy and Shul[13] previously reported selectivities of ~6 for InN over GaN in CH₄/H₂ Electron Cyclotron Resonance(ECR) plasmas, whereas Cl₂/Ar, ICl/Ar and IBr/Ar chemistries all showed values less than unity. Subsequently, Shul et al. Reference Shul, Willison, Bridges, Han, Lee, Pearton, Abernathy, MacKenzie, Donovan, Zhang and Lester[14] Reference Shul, Willison, Bridges, Han, Lee, Pearton, Abernathy, MacKenzie and Donovan[15] investigated selective Inductively Coupled Plasma(ICP) etching of group III-nitrides in BCl₃ and Cl₂ mixtures with addition of Ar, N₂, SF₆ and H₂, and found selectivities as high as 8 for GaN/AlN and 6.5 for GaN/InN under optimum conditions. Our new results for BI₃ and BBr₃ mean that there are now available chemistries that will allow the full range of desired etching properties, i.e. non-selective, selective for In-based nitrides over GaN and AlN, and selective for GaN over AlN and In-based materials. We also find that BI₃ and BBr₃ produce smooth surface morphologies and anisotropic etched sidewalls.

2. Experimental

The epitaxial films were grown on c-plane α-Al₂O₃ by either Metal Organic Chemical Vapor Deposition(GaN) at 1040°C, or by Metal Organic Molecular Beam Epitaxy Reference Abernathy[16] (InN and AlN) at 600°C and 800°C, respectively. The layers were 1.2−3.0μm thick and were nominally undoped(n ~ 6 × 10¹⁶cm⁻³ for GaN, n ~ 10²⁰cm⁻³ for InN and resistive, > 10⁸Ω˙cm, for AlN).

BI₃ is a white crystalline solid with a melting point of ~40°C, while BBr₃ is a red liquid with a boiling point of 91.2°C. Approximately 50g of each was placed in a quartz container within a stainless steel vacuum vessel heated to ~45°C to increase the vapor pressure of the reactants. The resultant flow rates were in the range of 5−10 standard cubic centimeters per minute(sccm). Etching was performed in a Plasma Therm 790 system in which the samples are thermally bonded to a Si carrier wafer mechanically clamped to an rf-biased(13.56MHz, 450W), He backside cooled chuck. The 3 turn coil ICP source operates at 2MHz and powers up to 1500W. The process pressure was held constant at 5mTorr. Typically Ar was added to the gas flow to facilitate plasma ignition and enhance the physical component of the etching.

Etch rates were obtained from stylus profilometry measurements of the features, while scanning electron microscopy(SEM) and atomic force microscopy(AFM) were used to examine surface morphology.

3. Results and Discussion

Discharges were readily sustained with both BI₃ and BBr₃. Pure BI₃ plasmas were blue-white, and optical emission spectra revealed many strong transitions in the range 380−440nm(Figure 1, top), which are related to atomic iodine. Similarly, pure BBr₃ discharges were slightly darker in appearance, but also showed many atomic transitions dominated by lines at 484nm, 658nm and 826nm(Figure 1, bottom).

Figure 1. Optical emission spectra from pure BI₃(top) or BBr₃(bottom) discharges under ICP conditions.

Figure 2 shows etch rates(top), etch yields(center) and selectivities for InN over both GaN and AlN(bottom) as a function of BI₃ percentage in BI₃/Ar discharges with fixed source power(750W) and rf chuck power(150W). Note that dc chuck self-bias decreases as BI₃ content increases, indicating that the ion density in the plasma is increasing under these conditions, and that BI₃ is therefore more easily ionized than Ar. The InN etch rate is monotonically dependent on BI₃ content, indicating the presence of a strong chemical component to its etching. By contrast AlN and GaN show insignificant rates until ~50% BI₃ where values of ~500Å·min⁻¹ for AlN and ~1700Å·min⁻¹ for GaN are obtained. Further increasing the BI₃ content in discharges actually leads to a reduction in etch rate for those two materials. There are at least two possible explanations for this result. First, the corresponding fall-off in chuck self-bias and hence ion energy under these conditions may more than compensate for the increased active iodine available. Second, since GaI_x and AlI_x etch products are not that volatile, a selvedge or reaction layer may form involving these species that quenches further etching. A precedent for the latter mechanism is reactive ion etching of InP in Cl₂ plasmas, where etching does not proceed unless elevated sample temperatures or high dc biases are used to facilitate removal of the InCl₃ etch product Reference Lee, Hong and Pearton[17]. Due to the behavior of GaN and AlN etch rates with BI₃ percentage, the InN selectivity to both materials initially increases but also goes through a minimum. Note however that selectivities of > 100 can be achieved for both InN/AlN and InN/GaN.

Figure 2. Nitride etch rates(top) and yields(center) and etch selectivities for InN/AlN and InN/GaN(bottom) in BI₃/Ar discharges(750W source power, 150W rf chuck power, 5mTorr) as a function of BI₃ content.

Similar data is shown in Figure 3 for BBr₃/Ar discharges with fixed source power(750W) and rf chuck power(350W). We needed higher rf powers to initiate etching with BBr₃ than with BI₃. Another difference is that now dc self-bias increases with BBr₃ content, indicating that it is less readily ionized than Ar. The etch rate of InN again increases with boron halide content, while GaN shows significant rates(~1800Å·min⁻¹) only for pure BBr₃ discharges. By contrast, AlN shows very low etch rates over the whole range of conditions investigated. Maximum selectivities of ~100:1 for InN/AlN and ~7.5:1 for InN/GaN are obtained.

Figure 3. Nitride etch rates(top) and yields(center) and etch selectivities for InN/AlN and InN/GaN(bottom) in BBr₃/Ar discharges(750W source power, 350W rf chuck power, 5mTorr) as a function of BBr₃ content.

One feature of the etching with these two new chemicals was the good surface morphologies obtained. Figure 4 shows examples of AFM scans(10 × 10μm²) of GaN before and after BI₃/Ar etching with different BI₃ percentages, while Figure 5 shows similar data for BBr₃/Ar etched samples. While there are clearly differences in the resulting surfaces, with pits or hillocks evident in some cases, the most important result is that all of the etched surfaces have lower root-mean-square(RMS) roughness than the control value. Figure 6 shows the dependence of RMS values for both chemistries as a function of discharge composition. This type of surface smoothing has been reported previously for GaN Reference Vartuli, Pearton, Lee, Hong, MacKenzie, Abernathy and Shul[18], and ascribed to the angular dependence of ion milling rates producing faster removal of sharp features. We were able to obtain AFM data over a much narrower range of conditions for InN because of the much higher etch rates and consequent difficulty in etching to a pre-determinant depth for AFM measurements, but the surfaces were also quite good for this material.

Figure 4. AFM scans of GaN surfaces before and after etching in 750W source power, 150W rf chuck power BI₃/Ar discharges as a function of BI₃ content.

Figure 5. AFM scans of GaN surfaces before and after etching in 750W source power, 350W rf chuck power BBr₃/Ar discharges as a function of BBr₃ content.

Figure 6. Dependence of GaN normalized etched surface roughness on boron halide percentage in BI₃/Ar or BBr₃/Ar discharges.

In high density plasma sources, ion density is basically controlled by the power applied to the source, while ion energy is mostly dependent on applied rf chuck power. If the latter is fixed, then increasing the source power will reduce the chuck self-bias. Figure 7 shows that source power had a significant effect only on InN etch rate for 4BI₃/6Ar discharges at fixed rf power(150W). Etch yields are quite low, even for InN, and under the best conditions about 9 incident ions on average are required to remove one In and one N atom. The etch selectivity for InN/AlN and InN/GaN increases with source power and reaches ~100 at 750W.

Figure 7. Nitride etch rates(top) and yields(center) and etch selectivities for InN/AlN and InN/GaN(bottom) in 4BI₃/6Ar discharges(150W rf chuck power, 5mTorr) as a function of ICP source power.

Basically similar trends were observed for 4BBr₃/6Ar discharges as a function of source power, as shown in Figure 8. InN etch rates and etch yields are lower than with BI₃/Ar, and therefore selectivities of ~30 for InN/AlN and ~40 for InN/GaN were obtained. There is a minimum in the InN/GaN data around 750W source power, which may result from a competition between increased etch rate of GaN due to higher flux, and desorption of the active bromine by ion-assistance at still higher fluxes.

Figure 8. Nitride etch rates(top) and yields(center) and etch selectivities for InN/AlN and InN/GaN(bottom) in 4BBr₃/6Ar discharges(350W rf chuck power, 5mTorr) as a function of ICP source power.

As mentioned above, incident ion energy in high density plasmas can be controlled by the rf chuck power. The resultant dc self-bias is the potential through which ions are accelerated as they cross the plasma sheath. Ion energy is then the sum of plasma potential (typically 20−30eV), plus the dc chuck bias. Figure 9 shows the dependence of etch rate, etch yield and InN/AlN and InN/GaN selectivity on rf chuck power for 4BI₃/6Ar discharges at fixed source power(750W). While GaN and AlN etch rates increase only at the highest chuck powers investigated, the InN etch rate increases rapidly to 250W, indicating a strong ion-assisted component to the etching, and then decreases at higher powers. This behavior produces corresponding maxima in both etch yield and selectivity. This type of behavior is quite common to high density plasma etching of III-V materials, where the etching is predominantly ion-assisted desorption of somewhat volatile products, with insignificant rates under ion-free conditions Reference Pearton[19]. In this scenario, at very high ion energies, the active etching species(iodine neutrals in this case) can be removed by sputtering before they have a chance to complete the reaction with substrate atoms.

Figure 9. Nitride etch rates(top) and yields(center) and etch selectivities for InN/AlN and InN/GaN(bottom) in 4BI₃/6Ar discharges(750W source power, 5mTorr) as a function of rf chuck power.

Similar data is shown in Figure 10 for BBr₃/Ar mixtures, as a function of rf chuck power. For this chemistry the InN etch rate saturates and we did not observe a reduction, although this might be expected to occur if higher powers could be applied(our power supply is limited to 450W). GaN does show an etch rate maximum at ~350W, producing a minimum in the resultant InN/GaN selectivity. Note again that the InN etch yields for BI₃/Ar are much higher than for BBr₃/Ar.

Figure 10. Nitride etch rates(top) and yields(center) and etch selectivities for InN/AlN and InN/GaN(bottom) in 4BBr₃/6Ar discharges(750W source, 5mTorr) as a function of rf chuck power.

AFM scans for InN before and after etching in BI₃/Ar discharges as a function of applied rf chuck power at fixed plasma composition and source power are shown in Figure 11. Note that the RMS surface roughness remain roughly constant and similar to that of the control value. Similar data is shown in Figure 12 for samples etched in BBr₃/Ar at different rf chuck powers. In this case, there is consistent smoothing of the surface occurring, as shown in the data in Figure 13. The fact that the smoothing effect decreases at higher powers is probably not unexpected, given that very high ion energies should eventually produce preferential sputtering of the higher N atoms.

Figure 11. AFM scans of InN surfaces before and after etching in 750W source power, 4BI₃/6Ar discharges as a function of rf chuck power.

Figure 12. AFM scans of InN surfaces before and after etching in 750W source power, 4BBr₃/6Ar discharges as a function of rf chuck power.

Figure 13. Dependence of InN normalized etched surface roughness on rf chuck power in 4BI₃/6Ar discharges(750W source power, 5mTorr).

Finally, we found the etched features to have vertical sidewalls which is expected given the ion-assisted nature of the etch mechanism. Figure 14 shows a few examples of SEM micrographs taken from GaN samples etched in either BI₃/Ar or BBr₃/Ar, using a SiN_x mask. The striations on the feature sidewalls originate from roughness on the initial photoresist mask used to pattern the SiN_x, and then this is replicated into the GaN.

Figure 14. SEM micrographs of features etched into GaN using either BI₃/Ar or BBr₃/Ar discharges. The SiN_x masks are still in place.

4. Summary and Conclusions

Two new plasma chemistries have been examined for etching III-nitrides. BI₃ produces etch rates for InN as high as 7,500 Å·min⁻¹ under ICP conditions, whereas the maximum rate with BBr₃ is ~5,500 Å·min⁻¹. The rate for AlN are low under all conditions, while GaN rates up to 1700−1800 Å·min⁻¹ can be obtained in both mixtures. Under optimum conditions etch selectivities of ~100 for InN over AlN and GaN were achieved in BI₃ chemistries, while in BBr₃ maximum values of ~100 for InN/AlN and ~25 for InN/GaN were obtained. These are the highest values reported for high density conditions, and result from the good volatility of InI_x etch products. The etched surface morphologies of GaN and InN were also very good, having similar or even lower RMS roughness than control samples. Both of these plasma chemistries appear useful for selective etch processes in nitride electronic device fabrication.