Copyright 2011 KIEEME. All rights reserved. http://www.transeem.org32

Regular Paper

TRANSACTIONS ON ELECTRICAL AND ELECTRONIC MATERIALS

Vol. 14, No. 1, pp. 32-35, February 25, 2013

† Author to whom all correspondence should be addressed:E-mail: nykim@kw.ac.kr

Copyright 2013 KIEEME. All rights reserved.This is an open-access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted noncommercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Optimization of Ohmic Contact Metallization Process for AlGaN/GaN High Electron Mobility Transistor

Cong Wang, Sung-Jin Cho, and Nam-Young Kim

Department of Electrical Engineering, Kwangwoon University, Seoul 139-701, Korea

Received May 18 2012, Accepted December 14 2012

In this paper, a manufacturing process was developed for fabricating high-quality AlGaN/GaN high electron mobility transistors (HEMTs) on silicon carbide (SiC) substrates. Various conditions and processing methods regarding the ohmic contact and pre-metal-deposition BCl3 etching processes were evaluated in terms of the device performance. In order to obtain a good ohmic contact performance, we tested a Ti/Al/Ta/Au ohmic contact metallization scheme under different rapid thermal annealing (RTA) temperature and time. A BCl3-based reactive-ion etching (RIE) method was performed before the ohmic metallization, since this approach was shown to produce a better ohmic contact compared to the as-fabricated HEMTs. A HEMT with a 0.5 μm gate length was fabricated using this novel manufacturing process, which exhibits a maximum drain current density of 720 mA/mm and a peak transconductance of 235 mS/mm. The X-band output power density was 6.4 W/mm with a 53% power added efficiency (PAE).

Keywords: AlGaN/GaN, High electron mobility transistor (HEMT), Ohmic contact, SiC substrate, Reactive-ion etching (RIE)

1. INTRODUCTION

Due to their high frequency power handling capability, wide-band gap, high-breakdown voltage, high current density, and high saturation velocity, AlGaN/GaN HEMTs have emerged as a promising candidate for next generation commercial wireless communication systems [1-3]. Most HEMTs are grown on sap-phire [4-6], silicon [7-9], or SiC [10-12] substrates. SiC substrates are presently the best choice for epitaxial growth because of their excellent crystal quality and have been used in devices with the highest output power densities.

Low-resistance ohmic contacts are essential parts of HEMTs and are needed to achieve high current densities and high ex-trinsic gains, which are required to obtain high thermal stability and a smooth surface morphology [13,14]. In this study, we con-sider a SiC-based AlGaN/GaN HEMT fabrication process, which

is highly competitive in terms of device performances. Several optimized processes were presented with the goal of improving the device characteristics. The formation of low resistance ohmic contacts to HEMTs is a key issue in the fabrication process of Al-GaN/GaN HEMTs. These ohmic contacts are required to produce a smooth surface morphology and well-defined edge acuity [15]. Pre-metal-deposition ion etching is also a widely used technique for improving metal to semiconductor contact resistivity in HEMT [16].

In this paper, we present a Ti/Al/Ta/Au-based metallic system that forms ohmic contacts under different annealing conditions, which can provide a reliable solution for AlGaN/GaN HEMT applications with superior electrical, morphological, and micro-structural properties. Furthermore, a BCl3 pre-metal-deposition RIE was used to fabricate an AlGaN/GaN heterostructure for improving metal contact resistance. Standard transmission line model (TLM) and atomic force microscopy (AFM) measurements were used to investigate the Ti/Al/Ta/Au-based ohmic contact structure. When this device was analyzed, a record power density of 6.4 W/mm for a 0.5 μm gate-length AlGaN/GaN HEMT on SiC was achieved which is needed for X-band applications (Fig. 6). This device also exhibited a maximum PAE of 53%, a power gain of 9.0 dB, a current density of 720 mA/mm, and a peak transcon-

pISSN: 1229-7607 eISSN: 2092-7592DOI: http://dx.doi.org/10.4313/TEEM.2013.14.1.32

33Trans. Electr. Electron. Mater. 14(1) 32 (2013): C. Wang et al.

ductance of 235 mS/mm.

2. EXPERIMENTS

The AlGaN/GaN epi-layers were grown using metal organic chemical vapour deposition (MOCVD) on a 4-in SiC substrate. The device fabrication started with an initial passivation layer consisting of SiO2 deposited by plasma-enhanced chemical va-por deposition (PECVD). The mesa isolation process was then performed using the RIE. Before the ohmic contact was estab-lished, BCl3 RIE treatment was carried out for 10, 20, and 30 s etches. The source and drain ohmic contacts, with a source-drain spacing Lsd of 6 μm, were achieved by the evaporation of Ti/Al/Ta/Au (20/80/40/100 nm) followed by an RTA at 850, 900, and 950℃ for 30 and 60 s, respectively. The specific contact resistanc-es (ρc) were recorded based on TLM measurements; the contact metals were separated by 2, 4, 8, 16, and 32 μm. The circular TLM data indicated a ρc of below 10-5 ohm·cm2 with a sheet resistance of around 350 ohm·cm2. After that, a Ni/Au (40/400 nm) gamma gate with a length of 0.5 μm, was formed between the source and drain ohmic contacts. The passivation process was performed after all of the aforementioned steps, followed by the forming of vias to open the metal contacts. DC current-voltage (I-V) and RF performance was measured to characterize the electrical properties of the device. The DC I-V measurements employed a VDS = 0 to 15 V and a VGS = 2 to -5 V. The load-pull measurements were performed at 10 GHz using an active load-pull setup under probes, permitting one to get the optimum load impedance, particularly for the smaller transistors. The cross-sectional sche-matic and the layout diagram of an AlGaN/GaN HEMT on a SiC substrate is shown in Fig. 1.

2.1 Ohmic contact process

The metallization scheme uses Ti/Al/Ta/Au with a thickness of 20/80/40/100 nm. A deoxydation process using HCl/H2O (1:10) for 20 s and distilled water cleaning for 480 s, and smooth O2/H2 plasma-etching was performed using a ULVAC microwave asher with a gas mixing rate of 9,000/450 sccm, a working pres-sure of 2 Torr, a chuck temperature of 80℃, and an RF power of 550 W for 20 s in order to remove the native oxides in the ohmic contact area. The metals were deposited by e-beam evaporation, followed by a lift-off process. After lift-off, the samples were se-quentially annealed under an N2 flow at annealing temperatures ranging from 850℃ to 950℃ over 30 or 60 s. Figure 2 shows the behavior of the ρc as a function of the annealing temperature and the annealing time for all of the samples. A better ohmic contact

was observed when the annealing time increased from 30 s to 60 s. In addition, the contact resistance decreased as the annealing temperature increased from 850℃ to 900℃; the lowest value of 3.25 × 10-5 ohm·cm2 was obtained at an annealing temperature of 900℃ and an annealing time of 60 s. In the fabricated ohmic contact metallization structure, Ti is essential because it partici-pates in the reaction at the interface with nitrides to form TiN or AlTi2N layers by high-temperature RTA, and nitride vacancies are

Fig. 1. The schematic diagram of the AlGaN/GaN HEMT on the SiC substrate: (a) cross-sectional view, and (b) layout.

1. Undoped AlN, ~0.1 um

3. Undoped GaN, ~50 nm

2. Undoped GaN buffer, ~1.2 um

SiC substrate

DrainSource Source

2-DEG

G G

2 um 3.5 um

200 um

75 um

4. Undoped AlGaN spacer, ~3 nm5. 4.5E18 Si-doped AlGaN, ~15 nm6. Undoped AlGaN cap layer, ~5 nm

(a) (b)

(e)

(d)(c)

(f)

Fig. 2. ρc of the Ti/Al/Ta/Au ohmic contact metallization scheme as a function of the RTA temperature and annealing time.

850 900 9501E-6

1E-5

1E-4

1E-3

Annealing Time : 30 sec Annealing Time : 60 sec

Annealing Temperature [oC]

Spec

ific

Cont

act R

esis

tanc

e [o

hm·c

m2 ]

Fig. 3. 3D AFM images of the ohmic contact surface morphology under various RTA conditions: (a) 850℃ for 30 s with a RMS value of 51.5 nm, (b) 850℃ for 60 s with a RMS value of 101.2 nm, (c) 900℃ for 30 s with a RMS value of 105.7 nm, (d) 900℃ for 60 s with an RMS value of 83.1 nm, (e) 950℃ for 30 s with an RMS value of 133.6 nm, and (f) 950℃ for 60 s with an RMS value of 146.1 nm. The images are 3D AFM with a scan area of 10 μm × 10 μm.

Trans. Electr. Electron. Mater. 14(1) 32 (2013): C. Wang et al.34

simultaneously formed at the AlGaN/GaN surface. At the same time, the diffused Ti and Al reduce the native gallium oxide on the AlGaN/GaN surface. Au is applied as an outer layer to pre-vent the oxidation of the Ti/Al metals during the RTA process. In addition, a diffusion barrier layer Ta is applied to prevent or min-imize the Au upper layer from diffusing downward. The ohmic contact is formed by the reaction between the Ti and the AlGaN, which produces TiN. In addition, there is a high potential barrier between the Ti and the AlGaN. The Al atoms diffuse through the Ti layer to enable a low work function for the Al-Ti with an inter-metallic phase to relieve the barrier. When the annealing tem-perature is low and the annealing time is insufficient, the ohmic contact property is poor due to the limited amount of Al and Ti. With an increase in the annealing temperature and annealing time, a great deal of Al diffuses into the AlGaN surface and reacts with the Ti, resulting in a reduction in the resistance. Significant intermetallic diffusions - an in-diffusion of Au towards the AlGaN interface and an out-diffusion of Al towards the contact surface occur when the temperature increases beyond 900℃, which causes a degradation in the contact characteristics. The surface morphologies of the ohmic contact metallization at different annealing temperatures are shown in Fig. 3. In addition, consid-erable differences were observed in their surface morphology. The surface morphology was smooth after annealing at 850℃ or 900℃. But after annealing at 950℃, the ohmic contact surface morphology was degraded. Various sized semispherical bulges were found to be randomly distributed on the surface, including some that cracked and broke apart. The smoother surface was probably due to the reaction between the different metals and no excessive Al remained. Therefore, a temperature as high as 900℃ was deemed to be the best, since the lowest contact resistance and a relatively smooth surface morphology was obtained under these conditions.

2.2 Pre-metal-deposition BCl3 etching process

The pre-metal-deposition BCl3 process was implemented be-fore the ohmic contact formation using an inductively coupled plasma (ICP) etching system. The ICP etching conditions were as follows: 100 W input power, 5 sccm BCl3 flow, and 10 mTorr pressure. Three different etching times (10, 20, and 30 s) were used and the results were compared to the un-etched samples. An initial decrease in ρc was observed with an increase in the etching time. As shown in Fig. 4, the ρc decreased from 3.25 × 10-5 ohm·cm2 for un-etched ohmics to about 7.74 × 10-6 ohm·cm2 for 20 s etch. A slight increase in ρc was observed when the etch time

was 30 s, which was most likely due to the non-uniformities in AlGaN thickness and composition. Thus, a RIE time of around 20 s was determined to be optimal.

3. RESULTS AND DISCUSSION

In this study, optimal RTA conditions and a pre-metal-depo-sition etching method was identified to minimize the contact resistance. To achieve this, the properties of Au/Ta/Al/Ti/AlGaN/GaN/AlN-SiC under different annealing conditions were investi-gated. In these experiments, the annealing process was shown to affect Al diffusion to the AlGaN surface to create N vacancies and were found to simultaneously react with the Ti and Au resulting in the formation of TiAl and AuAl phases. The concentration of N vacancies led to heavy doping and, hence, a reduction in the contact resistance. An initial decrease in the contact resistance with etching time was observed due to the removal of an oxide surface layer and/or to an increase in tunneling transport with a decrease in the AlGaN thickness. The presence of a dissimilar surface layer was confirmed by an initial non-uniform behavior in etch depth as a function of etch time. In order to obtain an efficient large signal operation, it is critical to fabricate HEMTs with a low parasitic gate and drain currents as well as minimized carrier trapping. The high voltage stability of the HEMTs was investigated in three-terminal configuration with the breakdown voltage defined at a current density of 1 mA/mm. Figure 5 shows the gate-drain breakdown voltage of the AlGaN/GaN HEMTs. The gate-drain breakdown voltage was better than 140 V and even at 100 V drain bias the parasitic drain currents under the pinch-off conditions were well below 1 mA/mm. This result demonstrates the excellent high-voltage stability of the HEMT.

Fig. 4. Measured ρc as a function of the etching conditions. The aver-aged contact resistances over 10 TLMs were 3.25 × 10-5 ohm·cm2, 9.27 × 10-6 ohm·cm2, 7.74 × 10-6 ohm·cm2, and 5.15 × 10-5 ohm·cm2 for 0, 10, 20, and 30 s etches, respectively.

0 10 20 30

1E-5

1E-4

Etching Time [s]

Annealing Temperature : 900 oC & Time : 60 s

Spec

ific

Cont

act R

esis

tanc

e [o

hm·c

m2 ]

Fig. 6. Electrical measurements: (a) DC transfer characteristics of a 200 × 0.5 μm2 dual gate HEMT at a drain source voltage of 10 V, and (b) power performances at 10 GHz of the same device when biased at 25 V.

Fig. 5. The gate-drain breakdown voltage of the HEMT.

0 20 40 60 80 100 120 1400.0

0.2

0.4

0.6

0.8

1.0

VGD (V)

Gate-Drain Breakdown Voltage

I GD (m

A/m

m)

VGD(V)

I GD(m

A/m

m)

35Trans. Electr. Electron. Mater. 14(1) 32 (2013): C. Wang et al.

The SiC substrate with the high-quality MOCVD-deposited epi-layers and the optimized ohmic contact process enabled us to develop high-performance AlGaN/GaN HEMTs. The DC and RF properties were determined to evaluate the device perfor-mance. The DC transfer characteristics of the 2 × (200 × 0.5) μm2 HEMT recorded at a drain source voltage of 10 V are shown in Fig. 6 (a). A maximum drain-source current of 720 mA, a peak transconductance of 235 mS/mm, and a threshold voltage of -3.6 V were achieved. Large signal measurements were performed using a load-pull system at 10 GHz. The device was biased at a drain bias of 25 V. The measured input power (Pin) versus the out-put power (Pout) responses of the device are illustrated in Fig. 6 (b), which indicate an output power density of 6.4 W/mm, PAE of 53.2% at 20.2 dBm Pin with an associated gain (Gp) of 9.0 dB. Ta-ble 1 shows the X-band AlGaN/GaN HEMTs fabricated using the proposed manufacturing process compared to previous works. To the best of our knowledge, these devices show the best output power density and PAE for AlGaN/GaN HEMTs on SiC substrates with a gate length of 0.5 μm. This superior performance was at-tributed to the optimized ohmic contact process and good mate-rial quality of the SiC substrate under high frequency operations.

4. CONCLUSIONS

In this study, a 0.5 μm gamma-gate AlGaN/GaN HEMTs fab-ricated on SiC substrates was evaluated. In order to meet the performance improvement requirements needed for high power applications in microwave systems, we proposed using an opti-mized ohmic contact process that minimizes the specific contact resistance. The DC and RF performances of this device were shown to be excellent. Based on these results, this ohmic contact manufacturing process holds great promise for use in fabricating HEMTs for X-band high-power applications.

ACKNOWLEDGMENTS

This research was supported by the National Research Foun-dation of Korea (NRF) and a Grant supported from the Korean government (MEST) No. 2012R1A1A2004366 and No. 2012-0009224. This work was also supported by a Research Grant of Kwangwoon University in 2012.

REFERENCES

[1] M. Faqir, G. Verzellesi, G. Meneghesso, E. Zanoni, and F. Fan-

tini, IEEE Trans. Electron Devices 55, 1592 (2008) [DOI: http://dx.doi.org/10.1109/TED.2008.924437].

[2] S. Y. Park, T. H. Lee, and M. J. Kim, Trans. Electr. Electron. Mater. 11, 49 (2010) [DOI: http://dx.doi.org/10.4313/TEEM.2010.11.2.049].

[3] Y. S. Kim, O. G. Seok, M. K. Han, and M. W. Ha, Trans. Electr. Electron. Mater. 12, 148 (2011) [DOI: http://dx.doi.org/10.4313/TEEM.2011.12.4.148].

[4] M. Werquin, C. Gaquiere, Y. Guhel, N. Vellas, D. Theron et al, Electron. Lett. 41, 46 (2005) [DOI: http://dx.doi.org/10.1049/el:20056735].

[5] S. Arulkumaran, T. Egawa, and H. Ishikawa, Solid State Electron 49, 1632 (2005) [DOI: http://dx.doi.org/10.1016/j.sse.2005.08.014].

[6] Y. Liu, T. Egawa, and H. Jiang, Electron. Lett. 42, 884 (2006) [DOI: http://dx.doi.org/10.1049/el:20061150].

[7] S. Jia, Y. Dikme, D. Wang, K. J. Chen, and K. M. Lau, IEEE Elec-tron Device Lett. 26, 130 (2005) [DOI: http://dx.doi.org/10.1109/LED.2004.842647].

[8] S. Joblot, Y. Cordier, F. Semond, P. Lorenzini, S. Chenot et al, Electron. Lett. 42, 117 (2006) [DOI: http://dx.doi.org/10.1049/el:20063688].

[9] C. Wang and N. Y. Kim, Nanoscale Res. Lett. 7, 107 (2012) [DOI: http://dx.doi.org/10.1186/1556-276X-7-107].

[10] J. W. Lee, V. Kumar, and I. Adesida, Jpn. J. Appl. Phys. 45, 13 (2005) [DOI: http://dx.doi.org/10.1143/JJAP.45.13].

[11] N. Sarazin, E. Morvan, M. A. di Forte Poisson, M. Oualli, C. Gaquiere et al, IEEE Electron Device Lett. 31, 11 (2010) [DOI: http://dx.doi.org/10.1109/LED.2009.2035145].

[12] C. Wang, W. S. Lee, S. J. Cho, and N. Y. Kim, Electron. Lett. 48, 405 (2012) [DOI: http://dx.doi.org/10.1049/el.2012.0130].

[13] M. Piazza, C. Dua, M. Oualli, E. Morvan, and D. Carisetti, Microelectron. Reliab. 49, 1222 (2009) [DOI: http://dx.doi.org/10.1016/j.microrel.2009.06.043].

[14] R. M. Gong, J. Y. Wang, S. H. Liu, Z. H. Dong, M. Yu et al, Appl. Phys. Lett. 97, 062115 (2010) [DOI: http://dx.doi.org/10.1063/1.3479928].

[15] W. S. Lau, J. B. H. Tan, and B. P. Singh, Microelectron. Reliab. 49, 558 (2009). DOI: 10.1016/j.microrel.2009.02.010

[16] D. Buttari, A. Chini, G. Meneghesso, E. Zanoni, B. Moran et al, IEEE Electron Device Lett. 23, 76 (2002) [DOI: http://dx.doi.org/10.1109/55.981311].

[17] J. C. Gerbedoen, A. Soltani, S. Joblot, J. C. D. Jaeger, C. Gaquiere et al, IEEE Trans. Electron Devices 57, 1497 (2010) [DOI: http://dx.doi.org/10.1109/TED.2010.2048792].

[18] A. Minko, V. Hoel, E. Morvan, B. Grimbert, A. Soltani, E. Delos et al, IEEE Electron Device Lett. 25, 453 (2004). DOI: 10.1109/LED.2004.830272

[19] K. D. Chabak, J. K. Gillespie, V. Miller, A. Crespo, J. Roussos et al, IEEE Electron Device Lett. 31, 99 (2010) [DOI: http://dx.doi.org/10.1109/LED.2009.2036574].

[20] C. Wang, Y. L. He, X. F. Zheng, Y. Hao, X. H. Ma, and J. C. Zhang, J. Semiconductors 33, 034003 (2012) [DOI: http://dx.doi.org/10.1088/1674-4926/33/3/034003].

[21] X. L. Wang, G. X. Hu, Z. Y. Ma, J. X. Ran, C. M. Wang et al J. Cryst. Growth 298, 835 (2007) [DOI: http://dx.doi.org/10.1016/j.jcrysgro.2006.10.219].

[22] N. Defrance, V. Hoel, Y. Douvry, J. C. D. Jaeger, C. Gaquiere et al, IEEE Electron Device Lett. 30, 596 (2009) [DOI: http://dx.doi.org/10.1109/LED.2009.2019972].

Table 1. The summary of the published X-band AlGaN/GaN HEMTs on different substrates.

Works Substrate Power density PAE Gain[17] Si (001) 2.9 W/mm 20% 7.0 dB[18] Si (111) 1.9 W/mm 18% 10.0 dB[19] Diamond 2.08 W/mm 44.1% 14.6 dB[20] Sapphire 4.44 W/mm 49% 8.5 dB[21] SiC 4.1 W/mm 33.6% 6.13 dB[22] SiC 3.58 W/mm 25% 15 dB[22] SiC 5.06 W/mm 34.7% 11.8 dB

This work SiC 6.4 W/mm 53.2% 9.0 dB