Oxygen-Related Border Traps in MOS and GaN Devices D. M. Fleetwood,1,2,* T. Roy,3 X. Shen2, Y. S. Puzyrev2, E. X. Zhang1, R. D. Schrimpf1, and S. T. Pantelides2,1 1
Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville, TN 37235, USA 2 Department of Physics and Astronomy, Vanderbilt University, Nashville, TN 37235, USA 3 Department of Materials Science and Engineering, Georgia Institute of Technology, Atlanta, GA 30332 * Email:
[email protected]
Abstract Oxygen-related border traps cause low-frequency excess (1/f) noise in MOS transistors with SiO2 gate dielectrics and GaN/AlGaN HEMTs. In each case, the noise is associated with a reconfiguration of the microstructure of near-interfacial defects upon charge capture. O vacancies in the near-interfacial SiO2 capture electrons when Si-Si bonds are stretched beyond their equilibrium length, and release electrons when the defect relaxes. These lead to MOS noise at room temperature. Oxygen DX centers in AlGaN exhibit metastable states that differ in energy by ~0.2 eV, which leads to increased noise in GaN/AlGaN HEMTs at cryogenic temperatures. 1. Introduction Defects in microelectronic materials can strongly affect the yield, performance, long-term reliability, and radiation response of microelectronic devices and integrated circuits (ICs). Low-frequency noise measurements can assist the identification of critical defects in silicon-based MOS devices [1]-[3] and in compound-semiconductor [1],[4]-[6] devices. In this paper we show key results from a comprehensive series of experimental and theoretical studies that identify near-interface O-vacancy defects as the border traps causing 1/f noise in MOS transistors, and oxygen DX centers as the defects causing low-temperature noise in GaN/AlGaN high-electron mobility transistors (HEMTs). These results emphasize the critical roles that oxygen-related defects play in a broad variety of microelectronic devices and ICs. 2. Oxygen Vacancies in Si MOS Transistors In a series of studies performed from 1988 through 2002 [2],[3],[7]-[9], a strong correlation was demonstrated between the low-frequency noise of MOS transistors and oxygen vacancies in SiO2. For example, Fig. 1 shows a direct correlation between the pre-irradiation magnitude of the normalized low-frequency noise K and the post-irradiation threshold-voltage shift due to
978-1-4673-2475-5/12/$31.00 ©2012 IEEE
radiation-induced oxide-trap charge (∆VOT). MOS oxide-trap charge correlates strongly with O vacancies in SiO2 [2],[3],[8]-[14]. Density functional theory (DFT) calculations were performed to investigate the properties of common O vacancy defects in the near-interfacial SiO2. The defect microstructures and corresponding electron densities around Si and O atoms calculated by DFT are shown in Fig. 2 for (a) a Si-Si dimer Eδ′ defect, and (b) and (c) two kinds of puckered Eγ′ defects [3],[10]-[17]. The microstructures of these defects differ primarily in the degrees of freedom available to the central Si atoms (between which an O atom would sit, in the absence of the vacancy). These defects can change their local, structural configuration during electron capture or emission events [14]-[17]. DFT calculations for a neutral O vacancy in the dimer configuration demonstrate that the probability of capturing an electron (e.g., from the Si) increases with an
Fig. 1. Normalized noise magnitude K as a function of threshold-voltage shifts due to radiation-induced oxide-trap charge ΔVOT for 3 μm x 16 μm, nMOS transistors with gate oxides of different thickness (A, D: 32 nm; B, E: 48 nm; C: 60 nm) and radiation hardness (A-C hard; D, E soft) processed in the same lot. Noise measurements were performed in the linear region of device operation; values of ΔVOT were obtained from room temperature irradiation to 100 krad(SiO2) in a Co-60 source at a rate of ~ 278 rad(SiO2)/s at an oxide electric field of ~ 3 MV/cm. (After [7], © AIP, 1990.)
increasing separation of the Si1-Si0 bond at the center of this complex, as shown schematically in Fig. 3. At an equilibrium spacing of ~0.25-0.30 nm in bulk SiO2, the electron trapping level for the neutral dimer in Fig. 2(a) is near the SiO2 conduction band. However, if the Si-Si bond is stretched to ~0.35-0.4 nm, near-midgap states can open up that can metastably capture an electron. The energetics of electron capture become more favorable with increasing Si-Si spacing [3],[16],[17]. When an electron is captured by a dimer defect in the stretched configuration, the defect naturally relaxes to reduce the bond length, favoring electron emission, leading to the observed noise [3].
3. Oxygen Centers in Compound Semiconductors Fig. 4 shows 1/f noise magnitude vs. temperature for GaN/AlGaN HEMTs that were fabricated on SiC substrates using plasma-assisted molecular beam epitaxy (PAMBE) and metal-organic chemical vapor deposition (MOCVD) [18]. The energy scale on the upper x-axis is estimated using the method of Dutta and Horn [19]. The PAMBE growth was performed under Ga-rich and N-rich conditions. The MOCVD requires the presence of ammonia gas to reduce tri-methyl gallium to form GaN, so this fabrication method leads to characteristic defects that are similar to NH3-rich PAMBE growth conditions [18],[20],[21]. A clear peak in the noise is observed at low temperatures, especially for the N-rich MBE and MOCVD devices. The noise magnitude is significantly lower at higher temperatures. 0.00
-14
2
Svd*f/T (V /K)
10
Fig. 2. Schematic illustrations of unpaired electron densities (gray regions) and atomic configurations of (a) a dimer O vacancy center associated with the Eδ′ defect, and (b) and (c) two relaxed O vacancy centers associated with the Eγ′ defect. (After [3], © IEEE, 2002.)
0.23
E0 (eV)
0.47
0.70
0.94
1.17
N-rich MBE Ga-rich MBE MOCVD
-15
10
-16
10
0
200
400
T (K)
Fig. 4. Normalized low-frequency noise magnitude at f = 10 Hz as a function of temperature and processing environment for GaN/AlGaN HEMTs. The gate length of the devices is 0.7 µm; LGD=1.2 µm and LGS=0.7 µm. The devices are 150 µm wide. (After [18], © AIP, 2011.)
Equilibrium Dimer (SiO2 bulk)
EF Stretched Dimers (E decreases with increasing Si-Si separation)
Si
SiO2 Fig. 3. Schematic illustration of the energy levels of dimer O vacancies in bulk SiO2 with equilibrium Si-Si atomic spacing of ~0.25-0.30 nm and near-interface dimers with stretched Si-Si spacing of ~0.35-0.40 nm and a distribution of energy levels near midgap. (After [3], © IEEE, 2002.)
To understand the likely origin of the low-temperature peak in the noise, Fig. 5 shows (a) the results of DFT calculations of defects in the ternary alloy Al0.30Ga0.70N which includes (b) an oxygen DX center. The thermal excitation of an ON-1 DX center can lead to a physical reconfiguration of the defect and emission of an electron to the AlGaN conduction band, which converts the negatively charged DX center into a neutral ON. Further release of an electron results in a positively charged ON+1. These results are consistent with calculations of the configuration energy of an oxygen DX center in AlN and GaN by McCluskey, et al. [22]. The crossover point of the energy curve of O-1 with that of O0 in the inset of Fig. 5(a) shows that the energy barrier for this thermal excitation is ~0.25 eV. The emission of an electron from O0, resulting in ON+1, has essentially no energy barrier.
The capture of one electron by ON+1 has an energy barrier of ~0.3 eV at the AlGaN/GaN interface, as also shown in the inset of Fig. 5(a). This process involves an electron that is thermally excited from the Fermi level in the GaN, which tunnels to the empty level of the O+1 defect, ~0.1 eV below the AlGaN conduction band. The second electron capture has an energy barrier of ~0.35 eV, when a thermally excited electron tunnels to the unfilled level of O0, ~0.05 eV below the AlGaN conduction band. The subsequent structural relaxation results again in an ON-1 DX center. So this reversible cycle leads to enhanced noise at cryogenic temperatures in GaN/AlGaN HEMTs. (a)
Oxygen-related defect reconfiguration does not always lead primarily to noise in Si MOS or compound semiconductor devices. For example, metastable configurations of substitutional and interstitial oxygen complexes can lead to recoverable (with annealing) charge trapping in InAs/AlSb HEMTs [23]-[25], as illustrated in Fig. 7. This charge trapping and annealing is quite similar to the noise processes we outlined above. However, the energy barriers that must be overcome for defect reconfiguration to occur in the InAs/AlSb HEMTs are higher than for the Si MOS devices and GaN/AlGaN HEMTs, so the time constants for charge exchange are on the order of hours or days (interpreted as charge trapping and de-trapping) instead of seconds (interpreted as noise).
(b)
Fig. 5. (a) Defect energy of substitutional O in AlGaN as a function of the distance from the ideal lattice site, for (b) an O DX configuration; O is red, Al is blue, N is gray, and Ga is green in (b). Black squares represent transition points between the charge states of ON during the electron emission. The dashed line shows that the energy barrier to emit one electron from a negatively charged O DX center is ~0.25 eV. (After [18], © AIP, 2011.)
Fig. 7. Threshold-voltage and peak transconductance gm shift of a 2 20 m InAs/AlSb HEMT with 100 nm gate length and 2 m source-drain spacing. Devices were stressed at Vg = –0.5 V and Vds = 0.4 V for 5 h. Annealing results at room temperature are shown. The device recovers nearly completely in 2 days. (After [25], © IEEE, 2011.)
4. Summary O-vacancy centers in Si MOS devices and oxygen DX centers in GaN/AlGaN HEMTs lead to enhanced low-frequency noise. Substitutional and interstitial oxygen complexes lead to recoverable charge trapping in InAs/AlSb HEMTs. These results emphasize the wide range of O-related defects that contribute to the degradation of the performance, reliability, and radiation response of microelectronic devices and materials. Understanding and reducing the densities of these defects continues to be an important challenge. Acknowledgments This work was supported in part by the Air Force Office of Scientific Research, the Office of Naval Research and the US Navy, and the Defense Threat Reduction Agency. We thank G. Koblmueller, R. Chu, C. Poblenz, N. Fichtenbaum, C. S. Suh, U. K. Mishra, and J. S. Speck for providing the GaN/AlGaN HEMTs, and S. Dasgupta for stimulating discussions. References [1] L. K. J. Vandamme, IEEE Trans. Electron Dev., 41, p. 2176 (1994). [2] D. M. Fleetwood, T. L. Meisenheimer, and J. H. Scofield, IEEE Trans. Electron Dev., 41, p. 1953 (1994). [3] D. M. Fleetwood, H. D. Xiong, Z. Y. Lu, C. J. Nicklaw, J. A. Felix, R. D. Schrimpf, and S. T. Pantelides, IEEE Trans. Nucl. Sci., 49, p. 2674 (2002). [4] M. E. Levinshtein, S. L. Rumyantsev, R. Gaska, J. W. Wang, and M. S. Shur, Appl. Phys. Lett., 73, p. 1089 (1998). [5] D. Kuksenkov, H. Temkin, R. Gaska, and J. W. Yang, IEEE Electron Dev. Lett., 7, p. 222 (1998). [6] S. L. Rumyantsev, Y. Deng, E. Borovitskaya, A. Dmitriev, W. Knap, N. Pala, M. S. Shur, M. E. Levinshtein, M. Asif Khan, G. Simin, J. Yang, and X. Hu, J. Appl. Phys., 92, p. 4726 (2002). [7] J. H. Scofield, T. P. Doerr, and D. M. Fleetwood, IEEE Trans. Nucl. Sci., 36, p. 1946 (1989). [8] D. M. Fleetwood and J. H. Scofield, Phys. Rev. Lett., 64, p. 579 (1990). [9] D. M. Fleetwood, W. L. Warren, M. R. Shaneyfelt, R. A. B. Devine, and J. H. Scofield, J. Non-Crystalline Solids, 187, p. 199 (1995). [10] P. M. Lenahan and P. V. Dressendorfer, J. Appl. Phys., 55, p. 3495 (1984).
[11] W. L. Warren, M. R. Shaneyfelt, D. M. Fleetwood, J. R. Schwank, P. S. Winokur, and R. A. B. Devine, IEEE Trans. Nucl. Sci., 41, p. 1817 (1994). [12] A. J. Lelis, T. R. Oldham, H. E. Boesch, Jr., and F. B. McLean, IEEE Trans. Nucl. Sci., 36, p. 1808 (1989). [13] J. F. Conley, Jr., P. M. Lenahan, H. L. Evans, R. K. Lowry, and T. J. Morthorst, J. Appl. Phys., 76, p. 2872 (1994). [14] J. K. Rudra and W. B. Fowler, Phys. Rev. B, 35, p. 8223 (1987). [15] S. P. Karna, A. C. Pineda, R. D. Pugh, W. M. Shedd, and T. R. Oldham, IEEE Trans. Nucl. Sci., 47, p. 2316 (2000). [16] Z. Y. Lu, C. J. Nicklaw, D. M. Fleetwood, R. D. Schrimpf, and S. T. Pantelides, Phys. Rev. Lett., 89, p. 285505 (2002). [17] C. J. Nicklaw, Z. Y. Lu, D. M. Fleetwood, R. D. Schrimpf, and S. T. Pantelides, IEEE Trans. Nucl. Sci., 49, p. 2667 (2002). [18] T. Roy, E. X. Zhang, Y. S. Puzyrev, X. Shen, D. M. Fleetwood, R. D. Schrimpf, G. Koblmueller, R. Chu, C. Poblenz, N. Fichtenbaum, C. S. Suh, U. K. Mishra, J. S. Speck, and S. T. Pantelides, Appl. Phys. Lett., 99, p. 203501 (2011). [19] P. Dutta and P. M. Horn, Rev. Mod. Phys., 53, p. 497 (1981). [20] T. Roy, Y. S. Puzyrev, B. R. Tuttle, D. M. Fleetwood, R. D. Schrimpf, D. F. Brown, U. K. Mishra, and S. T. Pantelides, Appl. Phys. Lett., 96, p. 133503 (2010). [21] Y. S. Puzyrev, B. R. Tuttle, R. D. Schrimpf, D. M. Fleetwood, and S. T. Pantelides, Appl. Phys. Lett., 96, p. 053505 (2010). [22] M. D. McCluskey, N. M. Johnson, C. G. Van de Walle, D. P. Bour, and M. Kneissl, Phys. Rev. Lett., 80, p. 4008 (1998). [23] X. Shen, S. DasGupta, R. A. Reed, R. D. Schrimpf, D. M. Fleetwood, and S. T. Pantelides, J. Appl. Phys., 108, p. 114505 (2010). [24] S. Dasgupta, X. Shen, R. D. Schrimpf, R. A. Reed, S. T. Pantelides, D. M. Fleetwood, J. I. Bergman, and B. Brar, IEEE Trans. Electron Dev., 58, p. 1499 (2011). [25] S. T. Pantelides, Y. Puzyrev, X. Shen, T. Roy, S. Dasgupta, B. R. Tuttle, D. M. Fleetwood, and R. D. Schrimpf, Microelectron. Engineering, 90, p. 3 (2012).
