Effect of Nitridation on the Regrowth Interface
of AlGaN/GaN Structures Grown by Molecular
Beam Epitaxy on GaN Templates
YUEN-YEE WONG,1,4WEI-CHING HUANG,1HAI-DANG TRINH,1 TSUNG-HSI YANG,2JET-RUNG CHANG,2MICHEAL CHEN,3 and EDWARD YI CHANG1,2
1.—Department of Materials Science and Engineering, National Chiao Tung University, Hsinchu 30010, Taiwan. 2.—Department of Electronics Engineering, National Chiao Tung University, Hsinchu 30010, Taiwan. 3.—ULVAC Taiwan Inc., 8F., No. 5, Keji Rd., Hsinchu Science Park, Hsinchu 30078, Taiwan.4.—e-mail: [email protected]
AlGaN/GaN structures were regrown on GaN templates using plasma-assisted molecular beam epitaxy (PA-MBE). Prior to the regrowth, nitridation was performed using nitrogen plasma in the MBE chamber for different durations (0 min to 30 min). Direct-current measurements on high-electron-mobility transistor devices showed that good pinch-off characteristics and good interdevice isolation were achieved for samples prepared with a 30-min nitridation process. Current–voltage measurements on Schottky barrier diodes also revealed that, for samples prepared without nitridation, the reverse-bias gate leakage current was approximately two orders of magni-tudes larger than that of samples prepared with a 30-min nitridation process. The improvement in the electrical properties is a result of contaminant removal at the regrowth interface which may be induced by the etching effect of nitridation.
Key words: AlGaN/GaN, nitridation, regrowth interfaces, molecular beam epitaxy
INTRODUCTION
AlGaN/GaN high-electron-mobility transistors (HEMTs) are excellent candidates for next-genera-tion high-power and high-frequency applicanext-genera-tions. To be used in these applications, the HEMT devices require high-crystalline-quality and high-resistivity GaN materials. However, due to the lack of a large-size commercial-grade native substrate, GaN materials are usually grown on a foreign substrate such as sapphire, silicon, or silicon carbide. As a result of large lattice and thermal expansion coef-ficient mismatches between the GaN films and these substrates, high threading dislocation (TD) density is generated in the materials and GaN device per-formance is degraded. Dislocations are known to
reduce the carrier mobility1–3and to deteriorate the leakage current4,5 in GaN devices. Recently, advances in hydride vapor-phase epitaxy6 and ammonothermal growth7 methods used to produce high-quality GaN substrates have attracted much attention for direct growth of AlGaN/GaN struc-tures. The homoepitaxial growth has proven useful for achieving AlGaN/GaN structures with better quality2,8 and less inherent stress compared with those grown on foreign substrates. Despite the improvement in crystal quality, HEMT devices regrown on GaN substrates still suffer from inferior electrical properties. This is due to the presence of a parasitic conduction path at the regrowth interface (RI) of the structure as a result of surface contami-nation on the GaN substrate after exposure to air.9,10The contaminants induce charges at the RI, resulting in poor pinch-off characteristics and poor interdevice isolation of HEMT devices.
(Received January 15, 2012; accepted May 19, 2012; published online June 6, 2012)
Ó2012 TMS
enhanced chemical etching of the GaN template. In this work, we reduced the charges at the RI between the AlGaN/GaN structure and the GaN template by using nitrogen plasma treatment, or nitridation. The effects of nitridation time on the electrical and material properties of this structure are also investigated.
EXPERIMENTAL PROCEDURES Unintentionally doped GaN templates approxi-mately 1.5 lm thick were first prepared on a sap-phire substrate using a metalorganic chemical
vapor deposition system (MOCVD; EMCORE
D-180). In this study, the GaN templates were prepared from the same MOCVD growth run to ensure that they had identical properties. After the growth, the GaN templates were removed from the chamber and left exposed to air for approximately 1 week before regrowth of AlGaN/GaN structures using plasma-assisted molecular beam epitaxy (PA-MBE; ULVAC). In the MBE chamber, the templates were annealed briefly at 800°C (for a few minutes) until the streaky reflection high-energy electron deflection patterns became sharp and bright. Then, the substrate temperature was low-ered to 740°C for the regrowth of the AlGaN/GaN structures. Prior to the regrowth, a nitridation process was performed using nitrogen plasma in the MBE chamber. Three samples with 0 min, 15 min, and 30 min nitridation treatment, designated as samples A, B, and C, respectively, were prepared for investigation of the nitridation effect on the RI. During the nitridation process, two plasma gener-ators with power set at 350 W were used to enhance the nitrogen radical supply. After that, an identical AlGaN/GaN structure (25 nm/400 nm) was grown on these templates. The Al composition in the
spacing with 2-lm gate length. Mesa isolation was achieved with Cl2/Ar inductively coupled plasma etching to depth of 150 nm, which is well below the two-dimensional electron gas (2DEG) channel. The circular SBDs consisted of a Schottky contact with diameter of 50 lm at the center of diode, and an ohmic contact formed as the outer ring. The spacing between the Schottky and ohmic contacts was also 50 lm. For material characterization, high-resolution x-ray diffraction (HRXRD; Bede D1 system) was used to characterize the GaN crystal quality. Meanwhile, time-of-flight secondary-ion mass spectroscopy (SIMS; ION-TOF, TOF.SIMS IV) was used to determine the impurities present in the structure. In addition, atomic force microscopy (AFM; Digital Instrument, D3100) and Auger elec-tron spectroscopy (AES; Microlab 350) were also employed to check the effect of nitridation on the bare GaN template surfaces.
RESULTS AND DISCUSSION
Hall measurement results for AlGaN/GaN struc-tures are presented in TableI. The sheet carrier concentrations (Ns) of samples A and B are much higher than that of sample C. However, from both simulation14 and experimental15results, typical Ns values for similar material structures are always less than 1.5 9 1013cm2. This suggests that the excess carriers in samples A and B should be con-tributed from outside the 2DEG channels. As a result, the electron mobility is also significantly lower for these samples. On the other hand, the Hall result of the sample prepared with a 30-min nitridation process easily corresponds to that of a good-quality AlGaN/GaN structure.
Figure1shows the direct-current (DC) measure-ment results of the HEMT devices. It can be seen
Table I. Hall measurement results of AlGaN/GaN structures prepared with different nitridation times
Sample Nitridation Time (min) Sheet Carrier Concentration (cm22) Mobility (cm2/V s)
A 0 2.86 9 1013 422
B 15 2.27 9 1013 603
that the electrical properties of the regrown HEMT structures are sensitively affected by the nitridation time. Figure1a, b suggests that devices fabricated on both samples A and B had poor electrical prop-erties. Only the sample prepared with a 30-min nitridation process was pinched-off well and had good gate modulation characteristics (Fig.1c). Fur-ther examinations were also carried out to deter-mine the leakage currents at the sample buffers. The buffer leakage currents were measured from the drain pads of two neighboring devices in these samples, and the results are shown in Fig.1d.
Compared with sample C, samples A and B suffered from leaky buffers, which explains the poor perfor-mance of the HEMT devices. Both the Hall and DC measurements suggest that parasitic conduction paths are present at the RI of samples A and B.
Large leakage currents were not only observed in the HEMT devices but were also observed in the SBDs fabricated on samples A and B. Figure 2
shows the SBD characteristics of all three samples in this study. The leakage current, measured at 10 V reverse bias, was decreased by approximately two orders of magnitude when the nitridation time
Fig. 1. DC characteristics of HEMT devices prepared with (a) 0 min, (b) 15 min, and (c) 30 min nitridation. (d) The interdevice leakage currents of these samples (color online).
Fig. 2. (a) I–V characteristics of AlGaN/GaN SBDs. The inset illustrates the structure of the SBDs. (b) Reverse-bias leakage currents measured at10 V. Inset shows the current leaking path in the SBDs (color online).
was increased from 0 min to 30 min (Fig.2b). As illustrated in the inset of Fig.2b, the leakage cur-rent in the SBDs may flow vertically across the AlGaN/GaN structure and laterally across the RI. For GaN materials, TDs, especially screw TDs, are generally known to form the conduction paths for leakage current in the vertical direction.5,13 To compare the crystal quality of the samples used in this study, HRXRD analyses were performed. The rocking curves (x scans) for both the GaN sym-metric plane (002) and asymsym-metric plane (102) were scanned, and their full-widths at half-maximum (FWHM) are summarized in Table II. The HRXRD results suggested that all three samples had the same crystal quality, as the FWHM values were similar. This means that the nitridation time did not affect the crystal quality of the regrown GaN materials. Therefore, the improvement of the leak-age current by nitridation may be attributed to the increase of the resistivity in both the RI and the TDs. The reason for the increase of resistivity in the TDs will be discussed later.
Meanwhile, from the measurement of the SBD forward currents, the characteristics of the fabri-cated SBD devices can also be determined. The current density in the forward direction can be expressed using the following equation:16
J¼ AT2expðq/b=kTÞ exp qV=nkTð Þ;
for V kT=q; ð1Þ
where J is the current density, A** is the effective Richardson constant, T is absolute temperature, q is the electron charge, /b is the barrier height, k is Boltzmann’s constant, V is the applied voltage, and n is the ideality factor. In this study, a Richardson constant of 24 was used for the AlGaN layer.14The /b and n values of the samples can therefore be calculated and are listed in TableII. It is shown that the SBD fabricated on sample C had the highest barrier height of 1.05 eV. As the nitridation time decreased, the barrier height decreased, and this explains the large leakage currents for SBDs fabricated on samples A and B. Besides, the SBDs on sample C also had ideality factor of 1.36, which is quite closed to the value of 1 for an ideal SBD. The large leakage current at the Schottky contact can also explain the poor gate control ability observed in the HEMT devices for samples A and B.
To explain the origin of the leakage currents in the HEMT and SBD devices, SIMS analysis was performed to check the unintentional impurities in the samples. Figure3a shows that the peak con-centrations of silicon, oxygen, and carbon were observed at the RI of sample A. These impurities, especially Si and O, are believed to induce parasitic charges at the RI and thus degrade the material electrical properties.10,11 In contrast, these impuri-ties were effectively reduced by the 30-min nitrida-tion process, as seen in sample C (Fig.3b).
To gain an insight into the effect of nitridation on the GaN template surface, AFM and AES analyses
were performed. AFM images reveal that the tem-plate surface without nitridation (Fig.4a) is rela-tively smooth compared with the template surface after a 30-min nitridation process (Fig.4b). The root-mean-square (rms) roughness of these sam-ples was 0.39 nm and 0.86 nm, respectively. The roughening of the nitridated sample could be induced by the etching effect in the nitrogen plas-ma. Frayssinet et al.17 have shown that the GaN film can be etched away slightly in the nitrogen plasma. The etching effect may cause the removal of Si, O, and C impurities at the RI, as determined from the SIMS results. As a result, the parasitic charges at the RI were also reduced. On the other hand, many tiny particles were observed on the GaN template surface without nitridation. To determine the chemical compositions of these par-ticles, the template surfaces were characterized using AES. The surface elements of Ga and N for samples prepared with 0 min and 30 min nitrida-tion were compared, and their ratios are also shown in Fig.4. AES performed over an area of 10 lm 9 10 lm showed higher Ga/N ratio (7%) on the sample surface prepared without nitridation. In a normal growth condition, the difference in Ga/N ratio of these GaN templates (maybe in the form of N-vacancy), which were prepared in the same MOCVD growth run, should be much smaller than 1019 cm3(0.1%), well below the detection limit of Auger analysis. So, it is believed that the large dif-ference in Ga/N for different samples was a result of the N2 plasma treatment. The larger Ga/N ratio found on sample without nitridation implies that the tiny particles observed by AFM are Ga droplets. A small amount of Ga might have resulted from
GaN decomposition upon high-temperature
annealing in the MBE chamber.18 The Ga atoms formed into droplets after the sample was cooled. During the regrowth, however, these excess Ga atoms can be incorporated into the screw TDs in the regrown structure and increase the conductivity of
the TDs.19 Therefore, large leakage currents are observed in the fabricated devices, as shown above, even at low bias voltages, due to the for-mation of low-resistance paths in the vertical direction. This result may explain why in situ thermal treatment alone was not sufficient to remove the RI charge.9 It suffices to say that both the impurities and Ga atoms were removed at the RI after the nitridation and therefore the leakage currents in the devices were reduced. In the presence study, however, it is hard to tell whether the impurities or the Ga atoms contribute more to the leakage current. Additional and careful examinations should be conducted to distinguish these two effects. Besides, more work should also be performed to find an optimum nitridation con-dition based on the interplay between plasma power, substrate temperature, and nitridation duration.
CONCLUSIONS
The effects of nitridation on GaN templates prior to regrowth of AlGaN/GaN structures by MBE were investigated. AlGaN/GaN HEMTs with good pinch-off characteristics and good interdevice isolation, and SBDs with reduced leakage currents at reverse bias, were fabricated on samples prepared with 30 min of nitridation. From the material analyses, the nitridation process did not affect the crystal quality of the regrown materials but was found to be useful for removing the charges at the RI by removing the impurities, such as Si, O, C, and Ga, on the GaN template surface. The removal of Si, O, and C by nitridation could be attributed to the nitrogen plasma etching process. The nitridation is also believed to have eliminated the excess Ga on the template surface, probably by the formation of new GaN material at the RI. Apparently, the nitridation step used in this study is in contrast to other plasma treatments, such as Ar and He
Fig. 4. AFM images of GaN templates after (a) 0 min and (b) 30 min of nitridation. The Ga/N ratios shown in the AFM images were obtained from AES analysis.
2. C. Skierbiszewski, K. Dybko, W. Knap, M. Siekacz, W. Krupczynski, G. Nowak, M. Bockowski, J. Lusakowski, Z.R. Wasilewski, D. Maude, T. Suski, and S. Porowski, Appl. Phys. Lett. 86, 102106 (2005).
3. N.G. Weimann, L.F. Eastman, D. Doppalapudi, H.M. Ng, and T.D. Moustakas, J. Appl. Phys. 83, 3656 (1998). 4. K.H. Lee, S.J. Chang, P.C. Chang, Y.C. Wang, and C.H.
Kuo, J. Electrochem. Soc. 155, H716 (2008).
5. J.W.P. Hsu, M.J. Manfra, D.V. Lang, S. Richter, S.N.G. Chu, A.M. Sergent, R.N. Kleiman, L.N. Pfeiffer, and R.J. Molnar, Appl. Phys. Lett. 78, 1685 (2001).
6. C.-C. Tsai, C.-S. Chang, and T.-Y. Chen, Appl. Phys. Lett. 80, 3718 (2002).
Heidelberg: Springer, 2008), pp. 91–138.
16. R. Williams, Modern GaAs Processing Methods (Norweed, MA: Artech House, 1990), pp. 241–270.
17. E. Frayssinet, P. Prystawko, M. Leszczynski, J. Domagala, W. Knap, and J.L. Robert, Phys. Status Solidi (a) 181, 151 (2000).
18. S. Fernandez-Garrido, G. Koblmuller, E. Calleja, and J.S. Speck, J. Appl. Phys. 104, 033541 (2008).
19. J.W.P. Hsu, M.J. Manfra, S.N.G. Chu, C.H. Chen, L.N. Pfeiffer, and R.J. Molnar, Appl. Phys. Lett. 78, 3980 (2001). 20. C. Eddy and B. Molnar, J. Electron. Mater. 28, 314 (1999). 21. Z.-Q. Fang, D.C. Look, X.-L. Wang, J. Han, F.A. Khan, and I.
