Journal of the Korean Physical Society, Vol. 42, No. 3, March 2003, pp. 358362

Dry Etching of InGaN/GaN Multiple Quantum-Well LED Structures inInductively Coupled Cl2/Ar Plasmas

H. J. Park, R. J. Choi and Y. B. Hahn

School of Chemical Engineering and Technology, SemiconductorPhysics Research Center, Chonbuk National University, Chonju 561-756

Y. H. ImFocus Center-New York,

Rensselaer; Interconnections for Gigascale Integration,Rensselaer Polytechnic Institute, U.S.A.

A. YoshikawaCenter for Frontier Electronics and Photonics, Chiba University, Chiba 263-8522, Japan

(Received 29 August 2002, in final form 23 December 2002)

A parametric study of the etch characteristics of InGaN/GaN multiple quantum-well (MQW)light-emitting diode (LED) structures has been carried out with Cl2/Ar in an inductively coupledplasma (ICP) discharge. The etch rate increased with Cl2 concentration up to 60 % and remainedrelatively constant at higher Cl2 percentages. It also increased monotonically with the ICP sourceand the rf chuck powers. The attainable etch rate for the InGaN/GaN MQW structures was about4,500 A/min under moderate ICP conditions: 700 W ICP source power, 100 W chuck power, 10mTorr, and 50 % Cl2. The experimental results overall showed that the dominant etch mechanism ofthe ICP etching of the InGaN/GaN MQW LED structures was an energetic ion-enhanced chemicaletching.

PACS numbers: 52.75.RKeywords: InGaN/GaN, Multiple quantum wells, Light-emitting diodes, Plasma etching

I. INTRODUCTION

Wide band gap III-nitride semiconductors have at-tracted great attention because of their successful appli-cations to photonic devices such as light-emitting diodes(LEDs) and laser diodes (LDs) [13]. In recent years,there has been remarkable progress for blue LEDs andLDs by utilizing InGaN/GaN multiple quantum wells asactive materials [4,5]. All of the LEDs and a majorityof the LDs have a ridge waveguide structure in whichthe mesas are formed by using a dry etching technique[6]. Hence, the fabrication of these GaN-based optoelec-tonic devices depends on dry etching either totally orpartially. Plasma etching techniques have been predom-inantly used in the patterning of the III-nitrides. Sincethe III-nitrides have relatively strong bond energies (InN:7.72 eV, GaN: 8.92 eV, AlN: 11.52 eV), a conventionalreactive ion etching (RIE) system utilizing a capacitivelycoupled plasma source cannot reach fast etch rates [7].

Recently, the most significant advancement in dry

E-mail: ybhahn@moak.chonbuk.ac.kr

etching has been the utilization of high-density plasmas,such as inductively coupled plasmas (ICPs) and electroncyclotron resonance (ECR) systems, in which the plasmadensity (or the ion flux) and the ion energy are controlledindependently [712]. Several research groups reportedthe etch characteristics of the III-nitrides by using high-density plasmas for epitaxially grown films. It is nowwell known that such a high-density plasma etcher canreadily produce etch rates greater than 6000 A/min forsingle layer films of the III-nitrides under conditions ofhigh ion flux and ion energy [1321]. However, littlework has been reported for the dry etching of multiplylayered heterostructures, such as the InGaN/GaN mul-tiple quantum-well (MQW) LED structures.

In this work, to elucidate the etch characteristics ofthe InGaN/GaN MQW LED structures grown by usingmetal-organic chemical vapor deposition, we carried outa parametric study of mesa etching with Cl2/Ar ICPs.The effects of the etch gas concentration, the pressure,the rf chuck power, and the ICP source on the etch rates,the dc bias, the plasma species, the surface chemistry,and the surface morphology were investigated. The In-

-358-

Dry Etching of InGaN/GaN Multiple Quantum-Well LED Structures H. J. Park et al. -359-

Fig. 1. Schematic illustration of the InGaN/GaN multiplequantum well LED structure grown by metal-organic chemi-cal vapor deposition.

GaN/GaN MQW LED structures showed an overall etchbehavior different from that for single-layered GaN films.

II. EXPERIMENT

InxGa1xN/GaN MQW LED structures were grownon c-plane sapphire substrates by using a metal-organicchemical vapor deposition system. Trimethylgallium(TMGa), trimethylindium (TMIn), ammonia (NH3),and silane (SIH4) were used as the precursors of Ga,In, N, and Si, respectively. Before the nitride films weregrown, the substrates loaded into the reactor were ther-mally cleaned in a hydrogen atmosphere at 1100 C for10 min. A GaN nucleation layer with a thickness of 25nm was grown on the cleaned substrate at 560 C, and4-m-thick GaN:Si was then grown at 1100 C. Figure1 shows the InGaN/GaN MQW LED structure grownat 750 C, having a 6-period InGaN(2 nm)/GaN(6 nm)quantum well structure. Finally a 0.25-m-thick Mg-doped p-GaN layer was grown at 1100 C on the top ofthe MQWs.

Dry etching was performed on samples (1010 mm2)of the InGaN/GaN MQW LED wafer in a planar-type ICP system (Vacuum Science ICP etcher, VSICP-1250A), in which the ICP source operated at 13.56 MHz.The temperature of the back-side-cooled chuck was heldat 25 C. The ion energy was controlled by using anapplied radiofrequency (rf) chuck power at 13.56 MHz.The Cl2/Ar mixture with a total gas flow rate of 40standard cubic centimeters per minute (sccm) was in-jected into the etcher through electronic mass flow con-trollers. For the etch rate experiments, the sampleswere patterned with AZ 4330 photoresists, and etchdepths were obtained from stylus profilometry measure-ments of the etched samples after removal of the photore-

Fig. 2. Effect of reactor pressure on (top) the etch rate ofthe InGaN/GaN MQW LED structure and the dc bias and(bottom) the OES peak intensities of the plasma species.

sists. Plasma species were analyzed using optical emis-sion spectroscopy (OES). The surface morphology andthe near-surface chemistry were examined using atomicforce microscopy (AFM) operating in the tapping modewith a Si tip and Auger electron microscopy (AES), re-spectively.

III. RESULTS AND DISCUSSION

Figure 2 shows the effects of the reactor pressure onthe (top) etch rate and dc bias and on the (bottom)OES peak intensities of of Cl, Cl+, Cl+2 and Ar

+ plasmaspecies. The pressure was varied from 5 to 30 mTorr at700 W ICP source power, 100 W rf chuck power, and50 % Cl2 (20 sccm Ar/20 sccm Cl2). The etch rate ofthe InGaN/GaN MQW LED structures increased up to10 mTorr, and decreased monotonically at higher pres-sures (>10 mTorr). By contrast, the dc bias increasedsubstantially with pressure. The maximum etch rate at10 mTorr was 4,500 A/min, which is much lower thanthe etch rates for single-layered GaN films reported else-where [1721]. The OES analysis revealed that the in-tensities of chlorine ions decreased, but those of Cl andAr somewhat increased. Combining the OES intensitydata for Cl+ and Cl+2 with the etch rates, we may draw

-360- Journal of the Korean Physical Society, Vol. 42, No. 3, March 2003

Fig. 3. Effect of Cl2 concentration on (top) the etch rateof the InGaN/GaN MQW LED structure and the dc bias and(bottom) the OES peak intensities of the plasma species.

the conclusion that at higher pressures, reactive chlorineions may play more important roles than neutral chlo-rine and inert argon, leading to a chlorine -ion-controlledetching mechanism. By contrast, at lower pressures, theetch rate is controlled by the neutral flux (i.e., by thepressure). Hence, the decrease in the etch rate at higherpressures can be attributed either to an insufficient sup-ply of chlorine ions or the shorter mean free paths ofplasma species, or a combination of the two.

Figure 3 presents the effect of Cl2 concentration onthe (top) etch rate and dc bias and on the (bottom)OES peak intensities of the plasma species of Cl, Cl+,Cl+2 and Ar

+. The plasma conditions were 700 W ICPsource power, 150 W rf chuck power, and 10 mTorr. Theetch rate increased with the Cl2 concentration and re-mained relatively constant at higher concentrations (>60%). Although not illustrated, a pure argon plasma re-sulted in a very slow etch rate (

Dry Etching of InGaN/GaN Multiple Quantum-Well LED Structures H. J. Park et al. -361-

Fig. 5. Effect of rf chuck power on the etch rate of theInGaN/GaN MQW LED structure and the dc bias.

Fig. 6. AES surface scans from the as-grown and theetched GaN films at 700 W ICP, 100 W rf, 10 mTorr, and 50% Cl2.

in the etch rate. Although not illustrated, the etch ratewithout applying the chuck power applied was less than200 A/min.

Figure 6 shows the AES surface scans from the as-grown and the etched p- and n-type GaN films, on whichelectrode metals were deposited for p-type and n-typeOhmic contacts, respectively. The etched p-GaN showedalmost the same surface stoichiometry as the unetchedcontrol sample, indicating equal rates of removal of groupIII and group V elements. By contrast, n-GaN showedsome depletion of nitrogen from the etched surface, butthis actually improved the contact resistance for n-typeOhmic contact. The oxygen and the carbon result fromexposure to ambient during the sample transfer from theICP etcher to the AES room.

The surface roughness of the etched InGaN/GaNquantum well structures was examined using AFM, andthe results for the etched surface of n-GaN in the mesastructure are shown in Fig. 7. The root mean square(rms) roughness increased with the chuck power because

Fig. 7. AFM images of the etched surfaces of n-GaN inthe LED mesa structures etched at 700 W ICP, 10 mTorr,and 50 % Cl2 for various rf chcuk powers.

of an increase in the ion-bombarding energy while theetched samples showed overall quite smooth surfaces(rms = 0.621.26 nm) up to 150 W. Although not il-lustrated, the surface morphology of the etched p-GaNwas similar to that of n-GaN.

Based on the results obtained in this work, we mayconclude that the dominant etch mechanism of the ICPetching of the InGaN/GaN MQW LED structures is anenergetic ion-enhanced chemical etching. The attainableetch rate of the InGaN/GaN MQW structure was about4,500 A/min under moderate ICP conditions: 700 WICP source power, 100 W chuck power, 10 mTorr, and 50% Cl2. However, this rate is much lower than etch ratesreported for single-layer GaN films. This indicates thatfor a desired pattern transfer of the InGaN/GaN MQWLED structure to be realized, the etch process has to beoptimized based on a parametric study for MQW LEDstructures rather than on one for single-layer GaN films.

IV. SUMMARY AND CONCLUSIONS

Dry etching of InGaN/GaN MQW LED structures hasbeen carried out with Cl2/Ar in an inductively coupled

-362- Journal of the Korean Physical Society, Vol. 42, No. 3, March 2003

plasma discharge. The effects of the etch gas concentra-tion, the ICP source power, the rf chuck power, and thepressure on the etch rate, the dc bias, the plasma species,the near-surface stoichiometry, and the surface roughnesswere examined. At higher pressures (>10 mTorr), reac-tive chlorine ions played more important roles than neu-tral chlorine and inert argon, leading to a reactive ionetching mechanism. The etch rate of the InGaN/GaNMQW LEDs increased up to a Cl2 concentration of 60% and remained relatively constant at higher Cl2 per-centages. It also increased monotonically with the ICPsource and rf chuck powers. The attainable etch ratefor the InGaN/GaN MQW structure was about 4,500A/min under moderate ICP conditions: 700 W ICPsource power, 100 W chuck power, 10 mTorr, and 50% Cl2. Equal removal rates were obtained for group-IIIand group-V elements for p-GaN, but some depletion ofnitrogen from the etched surface was seen for n-GaN.The surface of the etched InGaN/GaN MQW structureshowed a quite smooth surface (rms = 0.62 1.26 nm).The experimental results overall showed that the dom-inant etch mechanism for ICP etching of InGaN/GaNMQW LED structures was an energetic ion-enhancedchemical etching.

ACKNOWLEDGMENTS

This work was supported by the Basic Research Pro-gram of the Korea Science and Engineering Foundation(R01-2000-000-00330-0) through Chonbuk National Uni-versity.

REFERENCES

[1] S. Nakamura, T. Mokai and M. Senoh, Jpn. Appl. Phys.30, 1998 (1991).

[2] S. Nakamura, T. Mokai and M. Senoh, Appl. Phys. Lett.64, 1687 (1994).

[3] S. Nakamura, M. Senoh, S. Magahama, N. Iwasa, T.Yamada, T. Matsushita, H. Kiooku and Y. Sugimoto,Jpn. J. Appl. Phys. 35, L217 (1996).

[4] S. Nakamura, M. Senoh, N. Iwasa and S. Nagaham, Jpn.J. Appl. Phys. Part 1, 34, 797 (1995).

[5] F. A. Ponce and D. P. Bour, Nature 386, 351 (1997).[6] S. Nakamura, in GaN and Related Materials, edited by

S. J. Pearton (Gordon and Breach, New York, 1997).[7] Y. B. Hahn, D. C. Hays, S. M. Donovan, C. R. Aber-

nathy, J. Han, R. J. Shul, H. Cho, K. B. Jung and S. J.Pearton, J. Vac. Sci. Technol. A 17, 763 (1999).

[8] Y. B. Hahn and S. J. Pearton, Korean J. Chem. Eng.17, 304 (2000).

[9] R. J. Shul, in GaN and Related Materials, edited by S.J. Pearton (Gordon and Breach, New York, 1997).

[10] Y.-H. Im, Y. B. Hahn and S. J. Pearton, J. Vac. Sci.Technol. B 19, 701 (2001).

[11] Yeon-Ho Im and Yoon-Bong Hahn, Korean J. Chem.Eng. 19, 347 (2002).

[12] Jin Su Park, Tae Hee Kim, Chang Sun Choi and Yoon-Bong Hahn, Korean J. Chem. Eng. 19, 486 (2002).

[13] S. A. Smith, C. A. Woldren, M. D. Bremser, A. D.Hansen and R. F. Davis, Appl. Phys. Lett. 71, 3631(1997).

[14] R. J. Shul, G. B. McClellan, S. J. Pearton, C. R. Aber-nathy, C. Constantine and C. Barratt, Electron. Lett.32, 1408 (1996).

[15] D. C. Hays, C. R. Abernathy, W. S. Hobson, S. J.Pearton, J. Han, R. J. Shul, H. Cho, K. B. Jung, F.Ren and Y. B. Hahn, Mater. Res. Soc. Symp. Proc. 573,281 (1999).

[16] Y. B. Hahn, D. C. Hays, S. M. Donovan, C. R. Aber-nathy, J. Han, R. J. Shul, H. Cho, K. B. Jung and S. J.Pearton, J. Vac. Sci. Technol. A 17, 763 (1999).

[17] Y. H. Im, J. S. Park, Y. B. Hahn, K. S. Nahm and Y. S.Lee, J. Vac. Sci. Technol. A 18, 2169 (2000).

[18] B. C. Cho, Y. H. Im, J. S. Park, T. Jeong and Y. B.Hahn, J. Korean Phys. Soc. 37, 23 (2000).

[19] B. C. Cho, Y. H. Im, Y. B. Hahn, K. S. Nahm, Y.-S.Lee and S. J. Pearton, J. Electrochem. Soc. 147, 3914(2000).

[20] Y. H. Im, C. S. Choi and Y. B. Hahn, J. Korean Phys.Soc. 39, 617 (2001).

[21] Y. B. Hahn, B.-C. Cho, Y,-H. Im, K. S. Nahm and Y. S.Lee, J. Vac. Sci. Technol. A 19, 1277 (2001).