This content has been downloaded from IOPscience. Please scroll down to see the full text.
Download details:
IP Address: 140.113.38.11
This content was downloaded on 25/04/2014 at 06:19
Please note that terms and conditions apply.
Electrical Characterization and Transmission Electron Microscopy Assessment of Isolation of
AlGaN/GaN High Electron Mobility Transistors with Oxygen Ion Implantation
View the table of contents for this issue, or go to the journal homepage for more 2010 Jpn. J. Appl. Phys. 49 021001
(http://iopscience.iop.org/1347-4065/49/2R/021001)
annealing, the sheet resistivity was higher than 1012/square, which was attributed to the severe defect interaction eliminating the trapping
centers and reducing the leakage current. A maximum output power density of 5.3 W/mm at Vgs¼ 4 V and Vds¼50 V at 3 GHz was
demonstrated on lag-free HEMTs without field plates on sapphire substrate. #2010 The Japan Society of Applied Physics
DOI: 10.1143/JJAP.49.021001
1. Introduction
GaN based high electron mobility transistors (HEMTs) have attracted considerable attention owing to their excellent performance for microwave, high-power, and high-temper-ature applications. AlGaN/GaN HEMTs have shown much higher output power density at microwave frequencies than GaAs- or Si-based transistors. Device isolation of GaN-based HEMTs is conventionally realized by dry etching to define the device active region.1) However, implantation
isolation has the advantage in the sense that it maintains the planarity of the device, which increases the yield and uniformity of GaN HEMT and monolithic microwave integrated circuit (MMIC) processes.
Implantation isolation has been studied in pure GaN or AlGaN materials using Hþ, Heþ, Nþ, Fþ, Mgþ, Arþ, and
Znþ ions.2–8) The Oþ ion implantation isolation was also
investigated on AlGaAs,9) InAlN,10) and GaN (n-type doping)/GaN materials,11) to study the isolation quality and P/He, Arþ, and Nþion implantations have been carried out in the isolation of AlGaN/GaN HEMTs.12–14)
In this study, multienergy Oþ ion implantation was
applied for isolation in the fabrication of AlGaN/GaN HEMTs. The motivations for this are as follows. Firstly, Oþ
ion implantation isolation has better thermal stability than light atomic mass ion (i.e., Hþor Heþ). Secondly, the lower
implantation incident energy decreases the probability of surface damage of the device and increases the yield. Thirdly, multiple incident energy and higher implantation ion density were used to ensure high quality isolation in both AlGaN Schottky and GaN buffer layers with a good thermal stability. The latter motivations are important in the discussion of isolation processes for full HEMT device processes,15)as compared with isolation tests.
Two major mechanisms concerning oxygen implantation isolation used for GaN-based HEMT isolation need to be confirmed, especially after the thermal annealing treatment. Firstly, is the isolation mechanism a physical damage or chemical compensation? Secondly, the TRIM calculation of atomic displacement only accounts for the ballistic process
and neglects dynamic annealing (i.e., defect interaction process). However, ion-generated point defects that survive after the quenching of collision cascades may migrate through the lattice and experience annihilation and cluster formation.16) For real GaN HEMT application, the post thermal annealing phenomenon should be considered be-cause of subsequent high-temperature processes [300C
plasma-enhanced chemical vapor deposition (PECVD) passivation] and the operational channel temperature of AlGaN/GaN power HEMTs, which may be above 300C.17)
In this paper, we focus on the analysis of an in-house Oþ
implantation process with different postannealing temper-atures (Tas).15) The effects of the ambient temperature and
time on the sheet resistivity were investigated. Tunneling electron microscopy (TEM) analysis was carried out to understand the phenomena in the implanted area after annealing at different temperatures. Finally, the high-performance results of AlGaN/GaN HEMTs using the ion implantation process were demonstrated.
2. Experimental Methods
AlGaN/GaN heterostructures were grown on sapphire by metal–organic chemical vapor deposition (MOCVD) by Hitachi Cable Corporation. They consist of a 2-mm-thick unintentionally doped GaN buffer layer followed by 30-nm-thick undoped Al0:3Ga0:7N Schottky layer. From Hall
measurements, the sheet carrier concentration and electron mobility were determined to be 1 1013cm2and 900 cm2/
(Vs), respectively.
The TRIM software was used to simulate the implantation process.2)Figure 1 shows the simulated depth profile of the
distribution of the implanted Oþions, and damage vacancies
created by the implantation. The lower ion concentration near the wafer surface, below 50 nm [Fig. 1(a)], is of concern in the AlGaN/GaN HEMT process. Nevertheless, Fig. 1(b) shows a different shape vacancy depth profile without the surface low vacancy concentration problem.
The HEMT devices were 2 50 0:6 mm2 gates defined
in the middle of the 4 mm source-drain spacing. The epiwafers were first cleaned using a standard degreasing procedure and a standard RCA clean. Ohmic contacts were formed by e-beam evaporation of a Ti/Al/Ni/Au multilayer E-mail address: edc@mail.nctu.edu.tw
followed by rapid thermal annealing (RTA) in a nitrogen environment at 850C for 30 s. The ohmic contact process
step was done prior to the implantation isolation process. A typical contact resistance of 0.2 mm was measured on chip using TLM patterns. For the transistor process, the photoresist S1818 was used as an implantation mask to define the active region of the isolation testing sample and devices. All the samples were subjected to Oþ ion implantation with implantation energies of 25, 50, and 75 keV, and the dose is 5 1014cm2 for each energy. The
0.6-mm-long gates were defined by electron beam lithog-raphy, and the Ni/Au gate metallization was deposited by e-beam evaporation. The transistors were passivated using SiNx by PECVD at 300C, followed by the etching of
probing windows using a fluorine-based RIE process.18)
The isolation test structure used in this work consisted of two 100-mm-wide ohmic contacts with a separation of 5 mm.
On these samples, the implantation process was carried out after the ohmic contact process without any active region definition.
Atomic force microscopy (AFM) analysis was utilized after implantation and post high temperature process to investigate the surface morphology. The TEM specimens were prepared by manual lapping before fine polishing by Ar-ion milling (Gatan precision ion polishing system). The TEM investigation was performed using the JEOL JEM-2010F FEG.
3. Results and Discussion 3.1 Resistivity testing
The sheet resistivity (sh) vs ambient temperature (Ta) was
measured on the isolation test structures after different annealing times [Fig. 2(a)]. sh is higher than 1012/square
up to at least 450C annealing. After 100 h of annealing at
300C, the implanted material
shwas 4:3 1012/square,
which is higher than that of the sample with 1 h of annealing. (a)
(b)
Fig. 1. Depth profile of the distribution of (a) implanted Oþions and
(b) damaged vacancy created from the implantation. Estimated with the TRIM software.
(a)
(b)
Fig. 2. (a) Sheet resistivity vs annealing temperature with different annealing times. (b) Arrhenius plot showing activation energy (Ea)
calculated from linear fitting curve of sheet resistivity with different annealing temperatures.
function of the density of the trapping centers.20) Pearton
et al. claimed that the stable high shafter high-temperature
annealing can be explained on the basis of the chemical compensation of the electrons by deep O-related acceptors like in the AlGaAs material.11,19) By relying on those
investigations, the variation of sh with different Tas in our
case can be completely explained on the basis of the damage-induced compensation. We use TEM analysis to investigate the physical damage-induced phenomenon in the specimens after annealing with our optimized HEMT implantation isolation process.15)
The sheet resistances of these samples were also measured as a function of temperature, and a temperature-activated behavior was observed. From this, we obtained activation energies of 0.43, 0.52, 0.12, and 0.03 eV at 300, 450, 600, and 850C, respectively [Fig. 2(b)].
3.2 Material analysis
The rms of surface morphology on our wafers before implantation is 0.44 nm. The AFM analysis (Fig. 3) shows very good surface morphology after implantation and even after 1 h of high temperature (950C) postannealing. This is
important to achieve good yield and uniformity in GaN HEMT and MMIC processes.
Figure 4 shows cross-sectional bright-field TEM (XTEM) images after 1 h of annealing at different Tas. Some defect
clusters are observed in the implanted region of the as-implanted sample, especially close to the AlGaN-GaN interface [Fig. 4(a)]. However, much clearer defect clusters can be noted on the sample postannealed at 300C
[Fig. 4(b)]. With higher Ta, a gradual annihilation of the
defect clusters is observed [Figs. 4(c) and 4(d)]. These
for AlGaN/GaN HEMTs that have operation channel temperatures of at least up to 300C.17)
3.3 Device performance
HEMTs with 2 50 0:6 mm2 gates defined in the middle
of the 4 mm source-drain spacing were processed on the same AlGaN/GaN structure grown on a sapphire substrate. Comparing the DC and pulsed current–voltage (I–V) characteristics of the HEMTs, with short 100 ns pulse time and 0.1% duty cycle, is an effective way to clarify the gate-lag and drain-gate-lag effects at the same time.21,22) Figure 5
shows three different measurements: DC, pulsed from Vgs¼
0 V, Vds¼0 V (pulsed #1), and pulsed from Vgs¼Vpinch,
Vds ¼20 V (mimicking a class B operation, pulsed #2).22)
The pulsed #1 measurement shows a higher current than DC owing to the absence of a self-heating effect. The pulsed #2 measurement shows almost no current collapse. On the basis of DC, and pulsed I–V measurements, we conclude that the Oþ ion implantation isolation does not
introduce trapping problems. The extrinsic cut-off frequency ( fT) and maximum oscillation frequency ( fmax) calculated
from s-parameters measured up to 50 GHz VNA were 33 and 57 GHz, respectively. The large signal performance of the devices was determined by CW load–pull measurements at 3 GHz without active cooling. The saturated output power density is 4.5 W/mm with 51.5% power added efficiency (PAE) at Vds¼30 V, Vgs¼ 4 V. The highest output
power density was 5.3 W/mm at Vds¼50 V, Vgs ¼ 4 V
(Fig. 6).15)
4. Conclusions
The multienergy oxygen ion implantation process was investigated by Monte Carlo computer simulation (TRIM), sheet resistivity and thermal stability isolation test, material analysis (AFM, TEM), and AlGaN/GaN HEMT demon-stration. There is a clear correlation between the sheet resistivity (after different Tas) and TEM analysis results.
From XTEM investigation results, the defect cluster for-mation changed with different Tas affecting the ion
gen-erated point defects interaction. Fewer implant-induced point defects reduce the probability of trapped electrons to hop from one site to another and to be excited to the conduction band. Finally, the devices were processed and characterized to demonstrate the compability of the multi-energy oxygen ion implantation process in dispersion-free high-frequency and high-power AlGaN/GaN HEMT proc-essing and operation.
(a) (b)
Fig. 3. AFM surface morphology analysis after implantation, the scan area is 4mm2. (a) Without thermal annealing, the wafer showed R
rmsof
0.34 nm. (b) After 950C and 1 h of postannealing, the wafer showed R rms
Acknowledgements
The authors at NCTU would like to acknowledge the Min-istry of Education, the MinMin-istry of Economic Affairs, and the National Science Council of the Republic of China for
sup-porting this research under the contracts NSC 94-2752-E-009-001-PAE and 94-EC-17-A-05-S1-020. We also thank the Hitachi Cable Corp. for providing the high-quality wafer and NTHU Nuclear Science and Technology Development Cen-ter/Instrument Division for carrying out stable implantation.
Fig. 5. Pulsed I–V characteristics of Oþ-implanted HEMT under
differ-ent quiescdiffer-ent states (Vgs, Vds) (0 V, 0 V) (dashed line), (Vpinch, 20 V)
(dotted line), and DC (solid line).
Fig. 6. Power sweep at Vds¼50 V, Vgs¼ 4 V, showing 5.3 W/mm
saturated output power density.
(a)
(c)
(b)
(d)
Fig. 4. Cross-sectional bright-field TEM images of (a) as-implanted without annealing, (b) after 300C, (c) after 600C, and (d) after 850C
postannealing for 1 h.
10) S. J. Pearton, J. C. Zolper, R. J. Shul, and F. Ren:J. Appl. Phys. 94 (2003) 1662.
11) G. Dang, X. A. Cao, F. Ren, S. J. Pearton, J. Han, A. G. Baca, and R. J. Shul:J. Vac. Sci. Technol. B 17 (1999) 2015.
12) G. Hanington, Y. M. Hsin, Q. Z. Liu, P. M. Asbeck, S. S. Lau, M. Asif Khan, J. W. Yang, and Q. Chen:Electron. Lett. 34 (1998) 193.
Phys. Lett. 52 (1988) 395.
20) S. M. Sze:J. Appl. Phys. 38 (1967) 2951.
21) V. Desmaris, M. Rudzinski, N. Rorsman, P. R. Larssen, H. Zirath, T. C. Rodle, and H. F. F. Jos:IEEE Trans. Electron Devices 53 (2006) 2413. 22) R. Chu, Y. Zhou, J. Liu, D. Wang, K. J. Chen, and K. M. Lau:IEEE
