ARTICLE IN PRESS Materials Science in Semiconductor Processing 10 (2007) 112–116 Effects of thermal annealing on In-induced metastable defects in InGaN films H Hunga, K.T Lamb, S.J Changa,Ã, H Kuanc, C.H Chend, U.H Liawe a Institute of Microelectronics, Department of Electrical Engineering, National Cheng Kung University, Tainan 701, Taiwan b Department of Information Communication, Leader University, Tainan 701, Taiwan c Department of Electrical Engineering, Far East University, Tainan Country 744, Taiwan d Department of Electronic Engineering, Cheng Shiu University, Kaohsiung 830, Taiwan e Department of Avionics, China Institute of Technology, Hsinchu 312, Taiwan Available online July 2007 Abstract We investigated the effects of thermal annealing on the properties of InGaN layers From secondary ion mass spectroscopy results, it was found that severe In desorption occurred after annealing Photoluminescence and X-ray diffraction results indicate that significant amounts of In vacancy-related defects exist in the annealing samples It was also found that persistent photoconductivity decay time constants were 211, 893 and 1040 s, while the decay exponents were 0.153, 0.120 and 0.213 for the as-grown, 800 1C-annealed and 1000 1C-annealed InGaN epitaxial layers, respectively r 2007 Elsevier Ltd All rights reserved Keywords: InGaN; MOCVD; PPC; XRD; SIMS Introduction Wide band-gap group III-nitride semiconductors have attracted much attention in recent years These materials are potentially useful in various optoelectronic and high-speed device applications, such as blue/green/ultraviolet (UV) light-emitting diodes, laser diodes and high-power electronic devices [1–4] The InGaN/GaN heterostructure has been recognized as the essential structure for most nitridebased devices However, it is well known that the miscibility of InN in GaN is low The lattice mismatch between InN and GaN is also large ÃCorresponding author Tel.: +886 2757575x62391; fax: +886 2761854 E-mail address: changsj@mail.ncku.edu.tw (S.J Chang) 1369-8001/$ - see front matter r 2007 Elsevier Ltd All rights reserved doi:10.1016/j.mssp.2007.05.002 Thus, In-rich InGaN clusters are often formed in InGaN epitaxial layers Due to the compositionally unstable nature of the alloy [5–8], it is still difficult to grow high-quality InGaN epitaxial layers In order to improve the quality of InGaN, we need to reduce dislocation density and understand the nature of defects in these films [9–10] It has been shown that one can use persistent photoconductivity (PPC) and photoluminescence (PL) measurements to evaluate the quality of InGaN films [11–12] PPC measurements could provide us with useful information about the metastable properties of deep-level defects since these defects could be externally excited by shining light onto the epitaxial films Defects that interact with light could thus produce photocurrent that could last for an observably long time [13] It has been reported that PPC ARTICLE IN PRESS H Hung et al / Materials Science in Semiconductor Processing 10 (2007) 112–116 107 Secondary Ion Counts (c/s) Samples used in this study were all grown on (0 0 1) sapphire (Al2O3) substrates by metalorganic chemical vapor deposition system Details of the growth can be found elsewhere [15–18] Briefly, trimethylgallium (TMGa), trimethylindium (TMIn) and ammonia (NH3) were used as indium, gallium and nitrogen sources, respectively Disilane (Si2H6) was used as the n-type doping source We first prepared a low-temperature 25-nm-thick GaN nucleation layer at 500 1C We then raised the reactor temperature to 1050 1C to grow a 4-mmthick Si-doped GaN n-cladding layer Subsequently, the temperature was ramped down to 760 1C to grow an In0.25Ga0.75N layer with an electron concentration of 1018 cmÀ3 The as-grown samples were then furnace annealed in N2 ambient for 10 at 800 or 1000 1C Crystal qualities of these epitaxial layers were then evaluated by room temperature (RT) PL and double-crystal X-ray diffraction (DCXRD) A BioRad rpm 2000 system with a low mW HeCd laser operated at 325 nm was used for PL measurement, and a Bede QC2A system was used for DCXRD measurement Secondary ion mass spectroscopy (SIMS) was also used to evaluate distribution profiles of In, Ga and N atoms before and after thermal annealing After these measurements, Ti/Al contacts were deposited onto the sample surfaces to serve as contact electrodes Photocurrent measurements were subsequently performed using HeCd laser as the excitation source at various temperatures In order to measure transient responses, we switched the HeCd laser ON and OFF during PPC measurements We also applied a constant DC voltage onto the samples and measured current transients by an HP 4156 semiconductor parameter analyzer Figs 1(a)–(c) show SIMS profiles of as-grown, 800 1C-annealed and 1000 1C-annealed InGaN 106 N Ga In As-grown 105 104 103 102 101 100 200 400 600 Depth (nm) 800 1000 b 107 106 Secondary Ion Counts (c/s) Experiments Results and discussion N Ga In Annealed at 800 °C 105 104 103 102 101 100 200 400 600 Depth (nm) 800 1000 c 107 Secondary Ion Counts (c/s) effects in GaN-related alloys originate from random local fluctuations of alloy composition [14,15] However, no report on the annealing effect of PPC can be found in the literature to our knowledge For nitridebased devices, thermal annealing is often performed to activate Mg in p-GaN and/or for ohmic contact alloying In this study, we annealed InGaN epitaxial layers at various temperatures The effects of thermal annealing on the InGaN films will be discussed PCC measurements were then performed on these samples Measured electron-capture energies in these annealed InGaN films were also reported 113 N Ga In 106 Annealed at 1000 °C 105 104 103 102 101 100 200 400 600 Depth (nm) 800 1000 Fig SIMS profiles of (a) as-grown, (b) 800 1C-annealed and (c) 1000 1C-annealed InGaN epitaxial layers prepared in this study ARTICLE IN PRESS H Hung et al / Materials Science in Semiconductor Processing 10 (2007) 112–116 This again can be attributed to In desorption during annealing The inset of Fig shows a typical transient response of our samples It can be seen that fall-time is much longer than rise-time, which indicates that the PPC effect indeed exists in our samples Fig shows current transients measured at 80 K for the as-grown and annealed samples as we turned OFF the HeCd laser It was found that we could clearly observe PPC effect from all these three samples As we turned off the excitation, it was found that we could fit the current transients by the following stretched-exponential function [17]: I PPC ðtÞ ẳ expẵt=tịb 0obo1ị, as-grown 800 1000 1.0 0.8 (1) light off Photocurrent (a.u.) epitaxial layers prepared in this study It can be seen clearly that In atoms could only be observed near the sample surfaces, which agrees well with our initial design However, it was found that relative intensity of In decreased after annealing, particularly for the 1000 1C-annealed samples Such a result suggests that In atoms were desorbed from the top InGaN layers after annealing As a result, average In composition in the InGaN epitaxial layers should decrease Furthermore, defects related to In vacancies should also be generated after annealing Fig 2(a) shows measured PL profiles of the three samples It was found that PL intensity decreased while PL full-width at half-maximum increased after annealing These results also suggest that the quality of the samples was degraded after annealing It was also found that PL peak position shifted to the long-wavelength side This is probably due to the fact that deep-level-related luminescence became dominant for the annealed samples due to the increased defect density It is well known that defect-related yellow luminescence (YL) is often observed from GaN epitaxial layers [11] Instead of In clusters [16], we believe that the deep-level-related broad luminescence peak at 585 nm observed from the annealed samples is similar to the previously reported YL Fig 2(b) shows DCXRD spectra of as-grown, 800 1C-annealed and 1000 1C-annealed InGaN epitaxial layers prepared in this study It was found that we could clearly observe the InGaNrelated XRD peak from the as-grown sample However, intensities of InGaN-related XRD peaks seemed to decrease and eventually merged into the GaN main peak for the annealed samples Normalized Photocurrent (a.u.) 114 0.6 light on 10 15 Time (sec) 20 0.4 0.2 20 40 60 Time (sec) 80 100 Fig Current transients measured at 80 K for the as-grown and annealed samples as we turned OFF the HeCd laser The inset shows typical transient response of our samples 104 as-grown 800 1000 As-grown 103 1.2 × 107 Intensity (a.u) Intensity (a.u) 1.4 × 107 1.0 × 107 102 Annealed at 1000 101 8.0 × 106 6.0 × 106 500 Annealed at 800 100 525 550 575 600 Wavelength (nm) 625 650 -3000 -1500 1500 Omega-2Theada (arcsec) Fig Measured (a) PL profiles and (b) DCXRD spectra of the three samples 3000 ARTICLE IN PRESS H Hung et al / Materials Science in Semiconductor Processing 10 (2007) 112–116 Table PPC decay curve parameters for as-grown and annealed InGaN epitaxial layers Sample As-grown Annealed at 800 1C Annealed at 1000 1C Decay time constant t (s) Decay exponent b Electroncapture barrier DE (MeV) 211 893 1040 0.153 0.120 0.213 98 743 247 non-radiative recombination centers became larger due to severe In desorption Thus, we observed a much larger decay exponent b and a faster decay for the 1000 1C-annealed sample To clarify these points, we need to perform experiments such as deep-level transient spectroscopy Such experiments are under way and the results will be reported separately 6.0 As-grown In (τ) 5.5 ∆E = 98meV 5.0 4.5 1000/T (K-1) b Annealed at 800°C 16 In (τ) 14 12 ∆E = 743meV 10 6 10 1000/T (K-1) 12 14 c Annealed at 1000°C In (τ) where IPPC(0) is the initial photocurrent, t is the PPC decay time constant and b is the decay exponent For comparison, we normalized the current by IPPC(t) ¼ [I(t)ÀId]/[I(0)ÀId] [18] In other words, the current was normalized to unity at the time when illumination was switched OFF Here, I(t) is the current measured at time t, I(0) is the current immediately taken after the termination of the excitation source and Id is the initial dark current It has been shown previously that PPC effects of p-type GaN and InGaN/GaN quantum wells were related to DX-like deep-level defects and In-induced potential fluctuation [7,8] Since In atoms were desorbed after annealing, we observed significant PPC effect from the annealed samples due to severe In fluctuation and the large number of defect-induced deep-level centers Using the normalized current transients shown in Fig 3, we can calculate the PPC decay time constant t and decay exponent b As listed in Table 1, it was found that the PPC decay time constants were 211, 893 and 1040 s, while the decay exponents were 0.153, 0.120 and 0.213 for the as-grown, 800 1C-annealed and 1000 1C-annealed InGaN epitaxial layers, respectively Compared with the as-grown sample, it was found that PPC decay time constant t was larger while decay exponent b was smaller for the 800 1Cannealed sample These values result in slower decay for the 800 1C-annealed sample In contrast, it was found that PPC decay time constant t and decay exponent b of the 1000 1C-annealed sample were both larger than those observed from the as-grown sample These values result in faster decay for the 1000 1Cannealed sample The exact reasons for these observations were not clear yet It is possible that In vacancy-related deep-level centers could result in slower decay for the 800 1C-annealed sample On the other hand, the average In composition in the InGaN layer was much smaller for the 1000 1Cannealed sample It is possible that the number of 115 ∆E = 247meV 10 12 14 1000/T (K-1) Fig Arrhenius plots of t for the (a) as-grown, (b) 800 1Cannealed and (c) 1000 1C-annealed InGaN epitaxial layers ARTICLE IN PRESS H Hung et al / Materials Science in Semiconductor Processing 10 (2007) 112–116 116 Conclusions Ec Total Energy D-As grown D-1000 °C Ev ∆E D-800 °C In summary, the effects of thermal annealing on the properties of InGaN layers were investigated From SIMS results, it was found that severe In desorption occurred after annealing PL and XRD results indicate that significant amounts of In vacancy-related defects exist in the annealing samples It was also found that the PPC decay time constants were 211, 893 and 1040 s, while the decay exponents were 0.153, 0.120 and 0.213 for the as-grown, 800 1C-annealed and 1000 1C-annealed InGaN epitaxial layers, respectively Configuration Coordinate Q Fig Configuration-coordinate diagram for defect centers in InGaN after annealing at different temperature The observation of PPC effect implies that there is an insufficient amount of energy for carriers to overcome a capture barrier DE created by localized defects, preventing recapture of electrons by deep-levelrelated non-radiative recombination centers We thus performed current transient measurements and calculated PPC decay time constant t and decay exponent b at various temperatures For most III–V and II–IV semiconductor materials, it has been shown that the temperature-dependent PPC decay time constant t could be fitted well by t ¼ t0 exp[DE/kT], where t0 is the high-temperature limit of the time constant while DE is the capture barrier [19–22] Figs 4(a)–(c) show the Arrhenius plots of t for the as-grown, 800 C-annealed and 1000 1C-annealed InGaN epitaxial layers, respectively From these figures, we can determine the electron-capture energy, DE (i.e., the energy barrier for electrons to relax to ground state), of the samples As listed in Table 1, it was found that the electron-capture energies DE were 98, 743 and 247 MeV for the as-grown, 800 1C-annealed and 1000 1C-annealed InGaN epitaxial layers, respectively Fig shows the configuration-coordinate diagram of the fabricated samples [23] The vertical axis of Fig symbolizes the electronics and the strain energy of the defect center while the horizontal axis symbolizes the position of the defect and its neighboring atoms After annealing, we believe that the vacancies left by the vanished In atoms will change the configuration of the system Deep-level defect centers also shift toward the associated shallow states after annealing Thus, we also observed a change in capture energy, DE, after annealing due to the change of system configuration and the change of electron transition between its ground state and metastable state Reference [1] Nakamura S, Senoh M, Iwasa N, Nagahama S Appl Phys Lett 1995;67:1868 [2] Nakamura S, Fasol G The blue laser diode Berlin, Germany: Springer, 1997 [3] Wu LW, Chang SJ, Wen TC, Su YK, Chen JF, Lai WC, et al IEEE J Quantum Electron 2002;38:446 [4] Nakamura S, Senoh M, Nagahama SI, Iwasa N, Yamada T, Matsushita T, et al Jpn J Appl Phys Lett 1998;37:L309 [5] Dang XZ, Wang CD, Yu ET, Boutros KS, Redwing JM Appl Phys Lett 2001;72:2745 [6] Chen CH, Hang DR, Chen WH, Chen YF Appl Phys Lett 2003;82:1884 [7] Johnson C, Lin JY, Jiang HX, Khan MA, Sun CJ Appl Phys Lett 1996;68:1808 [8] Yang HC, Lin TY, Chen YF Appl Phys Lett 2001;78:338 [9] Zeng KC, Lin JY, Jiang HX, Salavador A, Popovici G, Tang H, et al Appl Phys Lett 1997;71:1368 [10] Deguchi T, Shikanai A, Torii K, Sota T, Chichibu S, Nakamura S Appl Phys Lett 1998;72:3329 [11] Reddy CV, Balakrishnan K, Okumura H, Yoshida S Appl Phys Lett 1998;73:244 [12] Grandjean N, Massies J, Leroux M Appl Phys Lett 1991;68:43 [13] Li JZ, Lin JY, Jiang HX, Geisz JF, Kurtz SR Appl Phys Lett 1999;75:1899 [14] Chang MCY, Tsang KO, Li EH, DenBaars SP MRS Internet J Nitride Semicond Res 1999;4:6d [15] Romano LT, McCluskey MD, Krusor BS, Bour DP, Chua C, Brennan S, et al J Cryst Growth 1998;189:38 [16] Feng SW, Lin EC, Tang TY, Cheng YC, Wang HC, Yang CC Appl Phys Lett 2003;83:3906 [17] Li JZ, Lin JY, Jiang HX J Appl Phys 1997;82:1227 [18] Lee YC, Shu GW, Chou WC, Shen JL, Uen WY Solid State Commun 2002;123:421 [19] Chen YF, Huang SF, Chen WS Phys Rev B 1991;44:12478 [20] Lin JY, Dissanayake A, Brown G, Jiang HX Phys Rev B 1990;42:5855 [21] Lin JY, Dissanayake A, Jiang HX Solid State Commun 1993;87:787 [22] Dissanayake A, Huang SX, Jiang HX, Lin JY Phys Rev B 1991;44:13343 [23] Redfield D, Bube RH Photoinduced defects in semiconductors Cambridge: Cambridge University Press; 2006