Home > The reactive ion etching characteristics of AlGaN/GaN SLs and etch induced damage study of n-GaN using Cl2/ SiCl4 /Ar plasma

The reactive ion etching characteristics of AlGaN/GaN SLs and etch induced damage study of n-GaN using Cl2/ SiCl4 /Ar plasma

The reactive ion etching characteristics of AlGaN/GaN SLs and etch induced damage study of n-GaN using Cl2/ SiCl4 /Ar plasma

Rui Li, Tao Dai, Ling Zhu, Huapu Pan, Ke Xu, Bei Zhang, ZhiJian Yang, GuoYi Zhang, ZiZhao Gan, Xiaodong Hu*

Research Center for Wide Gap Semiconductors of Peking University,

State Key Laboratory for Artificial Microstructure and Mesoscopic Physics, School of Physics, Peking University, 100871, Beijing, P. R. China

*Email: huxd@pku.edu.cn

Abstract

In this study, the etching characteristics of AlxGa1-xN/GaN superlattices with x=0.11 and 0.21 of Al and n-GaN with Cl2/ SiCl4/Ar plasma using RIE system were investigated. By varying gas ratio and rf power, it was found that SiCl4 is an effective getter to remove residual oxygen in the chamber and has a strong physical sputtering effect to remove the oxide layer during the etching, and a nearly nonselective smooth etching of AlxGa1-xN/GaN SLs with the high etch rate of 220 nm/min could be obtained. X-ray photoelectron spectroscopy (XPS) and Hall measurements were employed together to reveal the correlation between stoichiometry and electrical changes of n-GaN induced by plasma etching. Combining with N2O plasma postetch treatments to restore etched surfaces, those results suggested that oxygen not only influences morphology of the Al-containing samples, but also electrical properties of n-GaN by changing the status of various oxygen related defects which may play crucial roles in determining the nature of the damage.

PACS: 52.77.-j

Keywords: A1. Etching, A1. Surfaces, A1. Defects, A1. doping, B2. Semiconducting III-V materials, B3. Laser diodes

1. Introduction

Due to the inert chemical nature and strong bond energy of group III nitrides, Cl2-based dry etching techniques are indispensable for reliable pattern transfer in the semiconducting III-V materials. However, slow etch rates with rough morphology of AlxGa1−xN layers resultant from aluminum oxidation have often been observed [1-3]. To some extent these problems were resolved by adding BCl3 or CH4 gas to Cl2 plasma [2, 4]. Some articles reported that GaN and AlxGa1−xN etched by ICP in Cl2/Ar plasma on the Si or Ge wafer could also get smooth morphology [3, 5]. So far, an impediment still remains nonselective etching among various binary and ternary nitrides. Extensive studies are required in this area since many optoelectronic devices, especially laser diodes are based on heterostructures of AlxGa1−xN and GaN layers. For a ridge structure laser, etched surfaces need be smooth in order to get better oxidation of the structure and precisely control the residual thickness of upper waveguide layer [6]. An etching process that is highly anisotropic with etch rates independent of layer compositions is most favorable in that circumstance. In this paper, we report on the etching characteristics of high Al-containing AlxGa1−xN/GaN SLs and n-GaN with Cl2/ SiCl4/Ar plasma using reactive-ion-etching (RIE) system. Whereas the conduction properties of n-GaN are governed by either intrinsic or extrinsic defects [7, 8], and plasma exposure typically alters the conduction properties at the surface region by the creation of nonstoichiometric surface and consequent formation of near-surface lattice defects. Etch induced damage study and N2O plasma postetch treatments of n-GaN are also conducted to examine the influence of the plasma exposure on the carrier density at the surface of n-GaN, on which the Ohmic contacts would be formed.

2. Experiments

In this study, a 2.5 μm thick undoped n-GaN, two 120 pairs of Mg-doped Al0.11Ga0.89N/GaN and Al0.21Ga0.79N/GaN (25 �/25 �) SLs layers grown on sapphire substrates by metalorganic chemical vapor deposition (MOCVD) were used. The mole fraction of Al in the SLs layers was determined by x-ray diffraction. All the samples were rinsed in a H2SO4:H2O2 (1:1) solution at 125 �C to remove organic contamination and in a HCl: H2O2 (1:1) solution at 75 �C to remove oxide from the surface. Then a 420 nm thick SiO2 layers were deposited on the samples as a mask layer by ANELVA L-450P PECVD system. The mask pattern was defined using conventional photolithography and wet etched by buffered oxide etchant (BOE). The etching processes were conducted using a commercial Load-locked ANEVAL L-451D-L RIE system in which samples were mounted on a 4-inch Si wafer.

Unless otherwise noted, RIE plasma parameters used in this study were: 30 sccm total gas flow with an additional 5 sccm Ar for plasma stabilization, 1 min 20 s or 3 min 30 s of etch time, 2 Pa total process pressure, In order to investigate the influence of plasma composition and rf power on the etch characteristics, two series of experiments were conducted as follows: First, the rf power (13.56 MHz) was held at 70 W, varying the SiCl4 flow from 0 to 30 sccm. Secondly, the rf power was changed from 40 to 110 W,with the gas ratio of Cl2:SiCl4:Ar equal to 26:4:5. After the etching process, Etch rates were calculated from the step heights obtained from Alpha-Step IQ surface profiler before and after the removal of the SiO2 with BOE. The depth was measured at least three points on each sample to avoid incorrect data. A root-mean-square (rms) surface roughness of 3 μm � 3 μm area was examined using an atomic force microscope (AFM) in tapping mode.

To recover the etched surface of n-GaN, N2O plasma postetch treatments were carried out using PECVD under conditions listed in Table I. Carrier density and mobility of the original, etched and postetch treated samples were measured in vander Pauw mode. X-ray photoelectron spectroscopy (XPS) and room temperature photoluminescence (PL) measurements were also performed to study stoichiometry and optical changes.

TABLE I. Postetch treatments of n-GaN etched at Cl2:SiCl4:Ar =26:4:5, power:70W, pressure 2Pa.

Process

No.

Gas flow

(sccm)

Pressure

(Pa)

RF power

(W)

Self bias

(V)

Temperature

(�C)

Time

(min)

A

150

100

35

0

300

10

B

100

70

35

0

300

10

3. Results and discussion

The dependence of the etch rates on the gas flow of SiCl4 at fixing power of 70 W is shown in Fig. 1. As a whole, the etch rates of all samples monotonically decrease as gas flow of SiCl4 increases, while a noticeable turning point at 2 sccm implies that there is a critical SiCl4 flow for the chemical forming and physical removing of byproducts. Compared to the region from 8 sccm to 16 sccm, the etch rates of all samples in the region from 16 sccm to 30 sccm decrease drastically to 20.6-29.5 nm/min. This can be due to the reduction of active Cl radicals and Cl+ ions in the chamber, an evidence of the predominant role of Cl radical species in III-V material etching. It is well accepted that rough morphology of etched AlxGa1−xN is mainly caused by micromask effect of aluminum oxide with huge bond strength of Al–O (21.2 eV/atom) [9, 10]. Previous research has reported that GeClx+, SiClx+ have a strong sputtering effect on GaN etching [5]. Consistently, as shown in Fig. 2, 2 sccm of SiCl4 is enough to scavenge all oxygen and prevent oxidation of aluminum for both AlxGa1-xN/GaN SLs layers, while the presence of SiCl4 in the plasma is insignificant for the smooth etching of GaN.

Fig. 3 presents etch rates of n-GaN, AlxGa1-xN/GaN (x=0.11 and 0.21) SLs layers, and the SiO2 mask as a function of the rf power. With the Cl2:SiCl4:Ar gas ratio held at 26:4:5, the etch rates of all samples monotonically increase as rf power is elevated from 40 W to 110 W. In the low rf power region from 40 W to 60 W, the slight differences in selectivity among n-GaN and AlxGa1-xN/GaN SLs layers are diminished due to the increase of either the ion energy or the density of reactive Cl radicals in the plasma. But continuously increasing rf power up to 90 W, there is a trend to saturate etch rates for all samples, especially for AlxGa1-xN/GaN SLs layers. This can be expected since the decrease of mean free paths and consequent lower ion flux reaching the surface of etched samples by high-frequency collision among ions. The slower etch rate of Al-containing layers is probably related to the higher bond energy of AlN (11.52 eV/atom) compared to that of GaN (8.92 eV/atom) [9, 10]. However, the etch rates of AlxGa1-xN/GaN SLs layers increase sharply by further increasing rf power to 110 W, indicating enhanced sputter desorption by the etchant species such as SiClx, SiClx+.

According to the etch results mentioned above, we chose the optimized condition of Cl2:SiCl4:Ar 26:4:5, power 70 W and pressure 2 Pa for the damage study of n-GaN. The surface stoichiometries of n-GaN samples were studied by comparing the Ga (3d), N (1s), C (1s), O (1s), and Cl (2p) peaks in the XPS spectra. As Ga 3d:N 1s ratio shown in Table II, It is evident that N atoms are preferentially removed by the plasma etchant resulting in a gallium rich surface, which would give rise to the creation of VN at the surface region [11]. However, carrier density decreases to 83.4% of the original sample suggests that the considerable compensation with increase of O concentration has occurred, which can mainly be attributed to newly formed VGa-ON complexes, a deep level acceptor in n-GaN [12]. Due to the larger bond strength of Ga-O (14.7 eV/atom) compared to that of Ga–N (8.92 eV/atom), O atom is more likely to occupy dangling-bond states of Ga-rich etched surface during surface reconstruction, besides the plasma can also promote VGa-ON complexes formation by exposing the inner parts of epitaxial layer with high density of intrinsic VGa during the etching [13]. The evidence of the formation of VGa-ON complexes (not shown) is provided by the integrated PL intensity ratio of the band-edge and yellow luminescence transitions (IBE /IYE) [14] decreases from 0.277 to 0.157. It can be further proved by the FWHM of the Ga 3d core peaks before and after plasma exposure, as shown by the XPS spectra in Fig. 4 are 1.258 and 1.422 eV respectively. The broader peak of the etched one is attributed to the multiple splitting of the core level, which occurs as the system has unpaired electrons in the valence band [15]. This certainly is the case because the VGa-ON complex is doubly charged while the VGa-(ON)2 complex is singly charged.

TABLET II. XPS and Hall data of the original, etched and N2O plasma treated samples.

Sample

Ga(%)

N(%)

C(%)

O(%)

Cl(%)

Ga/N ratio

Ga/O ratio

μ (cm2/ν�s)

n (�1018cm-3)

original

46.01

32.31

16.13

5.55

-

1.424

8.306

211.52

4.549

etched

47.65

25.40

13.53

10.11

3.30

1.876

4.713

241.28

3.795

(A)

30.23

23.43

19.98

26.36

-

1.290

1.147

221.84

4.272

(B)

33.15

11.76

25.41

29.69

-

2.819

1.117

225.78

3.880

To recover the damaged surface of n-GaN, N2O plasma postetch treatments were performed to improve electrical properties. Sample (A) and sample (B) split from the same etched sample respectively underwent two different treating processes of A and B listed in table I. By comparing XPS data of the two samples after treatments, there are opposite changes in the % composition of N (1s) between the two samples, indicating the enhanced sputtering of GaN due to larger mean free paths of ions in process B. The preferential loss of nitrogen in the plasma exposure can be depicted by two processes as below: [16]

N2O+e-→NO+N+ e-,�� N+NO→N2+O.

The carrier density of (A) increased up to 94% higher than that of (B) up to 85% of original sample, while the Ga 3d:O 1s ratios are almost the same. The status of oxygen related defects differs from each other, (B) must have more compensative level compared to (A). The peak positions of Ga 3d shift from 20.45 eV of the etched to 20.20 eV of (B) and 20.15 eV of (A) respectively, and the FWHMs of two samples are nearly equal (1.396 eV vs. 1.402 eV), suggesting the concentrations of VGa-ON complexes may be the same. Because of plasma sputtering, the equilibrium concentrations of various defects in etched samples are modified after treatments. For (A), the newly incorporated O may prevailingly exist in ON which is believed to be the candidate donor for the auto doping within n-GaN [6]. But for (B), the concentration of VGa-ON complexes with possibly substantial amounts of OI which is predicted to be a double-charge acceptor [11] having no impact on broadening of the Ga 3d peak, is comparable to that of ON. The discrepancies in the results are attributed to the competition among acceptors of VGa-ON complexes and OI and donors of ON except commonly believed VN.

4. Conclusion

From our current studies, it was found that SiCl4 as an oxygen getter within plasma has a strong physical sputtering effect, which has great impact on the surface morphology of Al-containing layers. By optimizing radio-frequency power and plasma composition to reach a compromise between chemical and physical mechanism, a nearly nonselective etch with smooth morphology can be obtained. Under such condition the etch rate is approximately 220 nm/min for the Al0.21Ga0.79N/GaN SLs layers, the highest etch rate ever known. Hall measurements and XPS analysis of etched and postetch N2O plasma treated n-GaN samples, strongly suggested that oxygen not only influences morphology of the Al-containing layers, but also the electrical properties of n-GaN at the surface by altering the status of its related defects during plasma exposure.

Acknowledgements

This work was supported by the National High Technology Program of China under Grant Nos. 2005AA31G020, the National Natural Science Foundation of China under Grant Nos. 60477011, 60476028, 60406007 and 60276010.

References

[1] D. Buttari et al., Appl. Phys. Lett. 83, 4779 (2003).

[2] T. Wu, Z. Hao, G. Tang, and Y. Luo, Jpn. J. Appl. Phys., Part 2 42, L257 (2003).

[3] Evgeny Zhirnov, Sergei Stepanov, Wang Nang Wang, Y. G. Shreter, D. V. Takhin, and N. I. Bochkareva, J.

Vac. Sci. Technol. A 22, 2336 (2004).

[4] D. Basak, K. Yamashita, T. Sugahara, D. Nakagawa, Q. Fareed, K. Nishino, and S. Sakai, Jpn. J. Appl. Phys.,

Part 1 38, 42 (1999).

[5] Tatsuhiro Urushido, Harumasa Yoshida, Hideto Miyake, and Kazumasa Hiramatsu, Jpn. J. Appl. Phys., Part 2 41, L31 (2002).

[6] Joon Seop Kwak, T. Jang, K. K. Choi, Y. J. Sung, Y. H. Kim, S. Chae, S. N. Lee, K. H. Ha, O. H. Nam, and Y. Park, phys. stat. sol. (a) 201, No. 12, 2649–2652 (2004)

[7] Chris G. Van de Walle and J�rg Neugebauer, J. Appl. Phys. 95, 3851 (2004).

[8] D. C. Look, D. C. Reynolds, J. W. Hemsky, J. R. Sizelove, R. L. Jones, and R. J. Molnar, Phys. Rev. Lett. 79, 2273 (1997).

[9] S. J. Pearton, R. J. Shul, and F. Ren, MRS Internet J. Nitride Semicond. Res. 5, 11 (2000).

[10] J. Lee, K. Chang, I. Lee, and S. Park, J. Vac. Sci. Technol. B 18, 1409 (2000).

[11] S. J. Chua, H. W. Choi, J. Zhang, and P. Li, Phys. Rev. B 64, 205302 (2001).

[12] A. F. Wright, J. Appl. Phys. 98, 103531 (2005).

[13] D. K. Johnstone, S. Akarca-Biyikli, J. Xie, Y. Fu, C. W. Litton, and H. Morkoc, Proc. SPIE 6121, 61210N (2006)

[14] J�rg Neugebauer and Chris G. Van de Walle, Appl. Phys. Lett. 69, 503 (1996).

[15] D. Briggs and M.P. Seah, Practical Surface Analysis, 2nd edition John Wiley & Sons, Chichester, (1993), Vol. 1.

[16] Hiroaki Matsui, Hiromasa Saeki, Tomoji Kawai, Hitoshi Tabata, and Bunsho Mizobuchi, J. Appl. Phys. 95, 5882 (2004).

Recent Documents:

Set Home | Add to Favorites

All Rights Reserved Powered by Free Document Search and Download

Copyright © 2011
This site does not host pdf,doc,ppt,xls,rtf,txt files all document are the property of their respective owners. complaint#downhi.com
TOP