IEEE ELECTRON DEVICE LETTERS, VOL. 31, NO. 7, JULY 2010
707
Low-Temperature Annealing of Radiation-Induced Degradation in 4H-SiC Bipolar Junction Transistors Anders Hallén, Muhammad Nawaz, Senior Member, IEEE, Carina Zaring, Senior Member, IEEE, Muhammad Usman, Martin Domeij, and Mikael Östling, Fellow, IEEE
Abstract—Radiation hardness is tested for 4H-SiC n-p-n bipolar junction transistors designed for 1200-V breakdown voltage by implanting MeV protons and carbon ions at different doses and energies. The current gain is found to be a very sensitive parameter, and a fluence as low as 1 × 107 cm−2 of 10 MeV 12 C can be clearly detected in the forward-output characteristics, IC (VCE ). At this low dose, no influence of ion radiation is seen in the open-collector characteristics, IB (VEB ), or the reverse bias leakage and breakdown properties. Moreover, by annealing the implanted devices at 420 ◦ C for 30 min, a complete recovery of the electrical characteristics is accomplished. Index Terms—Annealing, bipolar junction transistor (BJT), current gain, point defects, radiation hardness, silicon carbide (SiC). Fig. 1. Layout of the 1200-V 4H-SiC BJT n-p-n structure with mesa-etched epitaxially grown emitters.
I. I NTRODUCTION
S
ILICON carbide (SiC) devices with increasing performance levels have been demonstrated for a number of years, and now, reliability issues, such as long-term stability and radiation hardness, also need to be addressed. Previously, the effects of radiation on SiC diodes [1]–[3] and MESFETs [4] have been reported, but in this letter, we have examined the effect of MeV on the electrical device characteristics of high-voltage 4H-SiC bipolar junction transistors (BJTs). Ions in this energy range have highly nonlinear energy transfer, both for elastic and nonelastic events, but the main effect will be the accumulation of point defects, particularly around the end of the ion path. The objective is to test the device tolerance for mega-electron-volt ions and to understand the mechanisms for degradation. It is shown that, for the doses used here, it is possible to recover the device output characteristics by a thermal anneal at only 420 ◦ C. The implantation-induced point defects that can be responsible for this behavior are also discussed.
Manuscript received March 3, 2010; revised March 23, 2010; accepted March 25, 2010. Date of publication May 6, 2010; date of current version June 25, 2010. The review of this letter was arranged by Editor S.-H. Ryu. A. Hallén, M. Usman, M. Domeij, and M. Östling are with ICT, Department of Microelectronics and Materials Physics, Royal Institute of Technology, 164 40 Kista, Sweden (e-mail:
[email protected];
[email protected]; mdomeij@ kth.se;
[email protected]). M. Nawaz is with the University Graduate Centre (UNIK), 2027 Kjeller, Norway (e-mail:
[email protected]). C. Zaring is with TranSiC AB, 164 40 Kista, Sweden (e-mail: carina.zaring@ transic.com). Color versions of one or more of the figures in this letter are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/LED.2010.2047237
II. E XPERIMENTAL P ROCEDURES The devices, shown in Fig. 1, are epitaxially grown n-p-n structures with mesa-etched emitters designed for 1200 V breakdown voltage and an active area of 3.4 mm2 [5]. The collector layer has a thickness of 15 μm and a nitrogen doping of 4 × 1015 cm−3 , while the p-base region is about 700 nm thick and has a p-doping of 4 × 1017 cm−3 . The mesaetched highly doped emitter structure is about 1 μm thick. The unencapsulated devices are exposed to 2.3 MeV protons, with/without an absorbing Al foil, or 5, 10, or 15 MeV 12 C ions. The ion beam is scanned across the target surface to get a homogeneous coverage, and the beam current is in the range of 10–100 pA. According to SRIM simulations [6], protons without an absorbing foil will reach about 40 μm into the samples and stop somewhere in the substrate. With the foil being placed in front of the samples, the projected proton range is about 10 μm in SiC, and the ions will stop in the low-doped collector region. Since the transistor chips have metallization and passivation layers of various thicknesses covering the surface, and the 1-μm-thick emitter covers about half the chip, there will be a spread of the projected ranges for different areas of the devices of about 2 μm. The 5 MeV 12 C ions will stop in the metallization or passivation layers, while the 10- and 15 MeV 12 C ions will penetrate 2–3 and 6–7 μm, respectively, and end up in the upper part of the collector. The fluences for protons are varied between 1 × 109 and 5 × 1012 cm−2 , while the carbon fluences range from 5 × 106 to 5 × 108 cm−2 . Measurements of the leakage current, breakdown voltage, and transistor characteristics (IC versus VCE ) are performed in pulsed mode (80-μs pulses) at room temperature before and
0741-3106/$26.00 © 2010 IEEE
708
IEEE ELECTRON DEVICE LETTERS, VOL. 31, NO. 7, JULY 2010
Fig. 2. Top six curves show IC versus VCE for three transistors before and after implantation of 10-MeV carbon ions (doses are given in the figure). The five lower IV curves show a device before and after implantation and also for three different 30-min annealing steps at 300 ◦ C, 400 ◦ C, and 420 ◦ C and at a lower base current. The statistical variations on a chip are about 5% and around 10% between different chips.
after beam exposure, as well as after thermal annealing of the samples for 30 min at 300 ◦ C, 350 ◦ C, 400 ◦ C, and 420 ◦ C. III. R ESULTS AND D ISCUSSION The top six curves in Fig. 2 show IC versus VCE for three transistors before and after implantation of 10 MeV carbon ions using the doses of 1 × 107 , 5 × 107 , and 1 × 108 cm−2 and the base currents of 140, 120, and 120 mA, respectively. The output characteristics (IC versus VCE ) are very sensitive to implantations, and even at these extremely low doses, there is a clear reduction in gain after the implantation compared to the nonimplanted devices. By comparing the curves, it is evident that the reduction is directly linked to the dose. For the highest dose shown (1 × 108 cm−2 ), the reduction in saturated collector current is about 25%. The leakage and breakdown characteristics of these transistors, as well as the open-collector characteristics [IB (VEB )], are not affected by the implantations at these doses. Similar results are seen in transistors implanted by 15 MeV 12 C and proton-implanted samples, although higher doses are required for these beams to reach a similar reduction of the output characteristics. However, for the lowest carbon energy (5 MeV), no significant effects on the electrical behavior of the transistors can be seen. This is probably due to the fact that the ion energy was not sufficient to penetrate the metallization or oxide passivation layers and that no damage is produced in the SiC layer. The proton implantations are shown in Fig. 3, which displays the relative reduction of current gain [β = IC /IB ] as a function of fluence for the two proton implantations, with/without an absorbing Al foil. In this figure, it is clear that a drastic change of amplification occurs only for doses that are well above 1010 cm−2 and that the devices implanted without a foil are less affected by protons, e.g., about
Fig. 3. Reduction of the current amplification factor (β) as a function of 2.3-MeV proton fluence. The full line shows results from devices implanted with an absorbing Al foil, and the dashed line shows devices implanted without a foil.
ten times higher dose is needed to reduce the amplification to below 80% of the original value for these devices. The results shown in Figs. 2 and 3 can be interpreted in the following two-step model. At low fluences, the degradation of the gain is mainly caused by an increased number of defects in the base region and also possibly by an increased density of surface states at the interface between the passivation layer and the base. Other device properties, such as reverse bias leakage or breakdown, are not yet affected. The larger reduction in β occurs, however, at higher doses, which, we suggest, results in a high compensating defect concentration in the collector, where the doping is 100 times lower than that in the base. The defects are predominantly created around the end of range of the ions, and measurements by deep-level transient spectroscopy (DLTS) [7] and capacitance–voltage (CV )[8], as well as simulations with SRIM, show that the peak defect concentrations in the collector for the doses used here are, in fact, exceeding the collector doping concentration of 4 × 1015 cm−3 . This strong degradation effect has also been seen in device characteristics for He-implantated BJT structures, using higher doses [9]. A typical annealing behavior of the collector current is shown in Fig. 2 by the five lower IV curves for a transistor implanted at 10 MeV using a dose of 108 cm−2 , i.e., the low-dose regime. The measurements are shown for a base current of 60 mA. Surprisingly, a nearly complete recovery of the characteristics is found for an annealing temperature as low as 420 ◦ C. This is a highly remarkable result considering that SiC is known to be a thermally very stable material and, for instance, diffusion of major dopants requires temperatures over 1500 ◦ C. Furthermore, many of the defects created in SiC by energetic radiation are known to be stable at temperatures even beyond 1500 ◦ C, for instance, the notorious Z1 /Z2 [10] defect positioned
HALLÉN et al.: LOW-TEMPERATURE ANNEALING OF RADIATION-INDUCED DEGRADATION
0.68 eV below the conduction band and the EH6 /EH7 [11] defect complex at EC -1.5 eV in 4H-SiC. These defects have relatively large charge carrier capture cross sections and have been pointed out as two major recombination and generation centers in 4H-SiC [12]. Since these defects are known to be stable at such high temperatures, it seems unlikely that they are involved in the reduction and recovery of the transistor characteristics seen in Fig. 2. It has been shown by DLTS that a handful of electrically active defects exist in the upper part of the bandgap after lowdose irradiation of 4H-SiC [13]. A common feature appears to be that they anneal at relatively low temperatures, starting already from room temperatures up to about 500 ◦ C. If these defects are the ones mostly responsible for the reduced β at low doses, annealing up to 420 ◦ C would certainly lead to the recovery seen in Fig. 2. IV. C ONCLUSION It has been shown that the current amplification β is a very sensitive parameter for monitoring the effect of low doses of MeV ions impinging on 4H-SiC BJT’s. The formation of defects in the base and SiC/dielectric interface are likely to be responsible for the degradation of transistor output characteristics in this low-fluence regime. It is furthermore shown that, for doses leading to < 25% reduction of the collector current, it is possible to fully recover the devices by thermal annealing at only 420 ◦ C. This low temperature excludes the well-known Z1 /Z2 and EH6 /EH7 defects as a cause of the degradation. Instead, we attribute the gain reduction to a family of deep levels in the upper part of the bandgap, which anneal at much lower temperatures. The major reduction of the gain occurring at higher fluence is, however, due to the formation of a highly compensated region in the lowly doped collector, which will reduce the collector current.
709
R EFERENCES [1] Z. Luo, T. Chen, A. C. Ahyi, A. K. Sutton, B. M. Haugerud, J. D. Cressler, D. C. Sheridan, J. R. Williams, P. W. Marshall, and R. A. Reed, “Proton radiation effects in 4H-SiC diodes and MOS capacitors,” IEEE Trans. Nucl. Sci., vol. 51, no. 6, pp. 3748–3752, Dec. 2004. [2] F. Moscatelli, A. Scorzoni, A. Poggi, M. Bruzzi, S. Sciortino, S. Lagomarsino, G. Wagner, I. Mandic, and R. Nipoti, “Radiation hardness after very high neutron irradiation of minimum ionizing particle detectors based on 4H-SiC junctions,” IEEE Trans. Nucl. Sci., vol. 53, no. 3, pp. 1557–1563, Jun. 2006. [3] Z. Luo, T. Chen, J. D. Cressler, D. C. Sheridan, J. R. Williams, R. A. Reed, and P. W. Marshall, “Impact of proton irradiation on the static and dynamic characteristics of high-voltage 4H-SiC JBS switching diodes,” IEEE Trans. Nucl. Sci., vol. 50, no. 6, pp. 1821–1826, Dec. 2003. [4] C. Brisset, O. Noblanc, C. Picard, F. Joffre, and C. Brylinski, “4H-SiC MESFETs behavior after high dose irradiation,” IEEE Trans. Nucl. Sci., vol. 47, no. 3, pp. 598–603, Jun. 2000. [5] H.-S. Lee, M. Domeij, C. M. Zetterling, M. Östling, F. Allerstam, and E. Ö. Sveinbjörnsson, “1200-V 5.2- mΩcm2 4H-SiC BJTs with a high common-emitter current gain,” IEEE Electron Device Lett., vol. 28, no. 11, pp. 1007–1009, Nov. 2007. [6] [Online]. Available: http://www.srim.org/ [7] D. Åberg, A. Hallén, and B. G. Svensson, “Low-dose ion implanted epitaxial 4H–SiC investigated by deep level transient spectroscopy,” Phys. B, vol. 273/274, pp. 672–676, Dec. 1999. [8] D. Åberg, A. Hallén, P. Pellegrino, and B. G. Svensson, “Nitrogen deactivation by implantation-induced defects in 4H–SiC n-type epitaxial layers,” Appl. Phys. Lett., vol. 78, no. 19, pp. 2908–2910, May 2001. [9] M. Usman, A. Hallén, R. Ghandi, and M. Domeij, “Effect of 3.0 MeV helium implantation on electrical characteristics of 4H-SiC BJTs,” Phys. Scr., 2009, submitted for publication. [10] T. Dalibor, C. Peppermüller, G. Pensl, S. Sridhara, R. P. Devaty, W. J. Choyke, A. Itoh, T. Kimoto, and H. Matsunami, “Defect centers in ion-implanted 4H silicon carbide,” Inst. Phys. Conf. Ser., vol. 142, p. 517, 1996. [11] C. Hemmingsson, N. T. Son, O. Kordina, J. P. Bergman, E. Janzén, J. L. Lindström, S. Savage, and N. Nordell, “Deep level defects in electron-irradiated 4H SiC epitaxial layers,” J. Appl. Phys., vol. 81, no. 9, pp. 6155–6159, May 1997. [12] L. Storasta, H. Tsuchida, and T. Miyazawa, “Enhanced annealing of the Z1/2 defect in 4H-SiC epilayers,” J. Appl. Phys., vol. 103, no. 1, p. 013 705, Jan. 2008. [13] P. Lévêque, D. Martin, B. G. Svensson, and A. Hallén, “Defect evolution in proton-irradiated 4H SiC investigated by deep level transient spectroscopy,” Mater. Sci. Forum, vol. 433–436, pp. 419–423, 2003.
