J O U R N A L O F M A T E R I A L S S C I E N C E : M A T E R I A L S I N E L E C T R O N I C S 1 6 (2 0 0 5 ) 463 – 467
Effects of high temperature electron irradiation on trench-IGBT M. NAKABAYASHI 1 , H. OHYAMA 2 , N. HANANO 3 , E. SIMOEN 4 , C. CLAEYS 4, 5 , K. TAKAKURA 2 , T. IWATA 2 , T. KUDOU 2 , M. YONEOKA 2 1 Renesas Technology Corporation, 4-1 Mizuhara, Itami, Hyogo, 664-0005, Japan 2 Kumamoto National College of Technology, 2659-2 Suya Nishigoshi Kumamoto, 861-1102, Japan E-mail:
[email protected] 3 Renesas Eastern Japan Semiconductor, Inc., 1-1, Nishiyokote, Takasaki, Gunma, 370-0021, Japan 4 IMEC, Kapeldreef 75, B-3001 Leuven, Belgium 5 E.E. Dept., KU Leuven, Kasteelpark Arenberg 10, B-3001 Leuven, Belgium The degradation of the electrical properties of IGBTs (Insulated-Gate Bipolar Transistors) by 2-MeV electron irradiation at high-temperatures was studied. The irradiation temperatures were 25, 100, 200 and 300 ◦ C, and the fluence was fixed at 1015 e/cm2 . For most experimental conditions, the threshold voltage (VTH ) is observed to recover partially during 100 ◦ C and the saturation voltage (VCEsat ) increase is most pronounced at 100 ◦ C. It is shown that in this irradiation temperature range, the temperature dependency of the voltage shift due to the radiation-induced interface traps (Vit ) and the voltage shift due to the radiation-induced oxide traps (Vot ) are different, which suggests a possible degradation mechanism for these phenomena. It is tentatively suggested that the diffusion of hydrogen released from the gate by the 2-MeV electrons passivates the interface traps during the C 2005 Springer Science + Business Media, Inc. high temperature irradiation.
1. Introduction The insulated-gate bipolar transistor (IGBT) is a unique device that combines the low forward voltage drop of a power bipolar transistor and the high input impedance of a power metal-oxide-semiconductor field effect transistor (MOSFET). Turn-off losses in IGBTs are largely dependent on the lifetime of excess charge carriers. In indirect band-gap semiconductors such as silicon, the recombination of excess charge carriers occurs predominantly through deep energy levels within the band-gap. Several techniques have been used for introducing deep energy levels in order to control the carrier lifetime in silicon. An early method was based on the diffusion of impurities, such as Au or Pt, which acted as recombination sites, [1]. The minority-carrier lifetime, τ , decreases linearly with the impurity concentration. The doping technique requires the diffusion of these impurities at high temperature, usually in excess of 800 ◦ C. Since the contact metallization for devices cannot endure these temperatures, the lifetime control must be performed before complete device fabrication. In recent technologies, high-energy electron irradiation is used, [2]. For the manufacture of IGBTs with various turn-off times, usually 2–3 MeV irradiation is applied. The collisions of electrons displace silicon atoms, and vacancyrelated point defects are formed. The electron irradiation introduces a uniform distribution of defects, some of which are efficient recombination centers, with both
the A-center [3] (oxygen-vacancy complex) and the divacancy [4] being cited. In more recent years the development of local lifetime control techniques using proton and helium irradiation has allowed the introduction of a very thin and well-localized reduced lifetime region in silicon devices, [5–7]. But these methods are more expensive than electron irradiation because it does not allow to irradiate several wafers at the same time, [8]. The electron irradiation technique and post anneal are used to control the IGBT switching performances, enabling to manufacture production series of IGBT’s with various turnoff times for several applications. However, a problem, which occurs when performing the electron irradiation and anneal is that an excessive amount of energy and time, and thus money, are consumed in the device manufacturing. High temperature irradiation might have the possibility of process reduction. This paper shows the response of the threshold voltage (VTH ), lifetime (τ ), saturation voltage (VCesat ), breakdown voltage (BVCES ) and leakage current (ICES ) for IGBT’s irradiated at room and high temperature by 2-MeV electrons.
C 2005 Springer Science + Business Media, Inc. 0957–4522
463
2. Experimental The device structure used in this study is a 400 V/150 A punch-through Trench IGBT with the stripe trench gate structure encapsulated in a 8 pin flat plastic package.
T A B L E I Measurement condition of steady-state characteristics
Figure 1 Cross-sectional view of the trench IGBT.
The cross-sectional view, not to scale, of one from the thousands of cells comprised in the IGBT is shown in Fig. 1. The thickness and the resistivity of the nbase are 55 µm and 20 cm, respectively. The N+ epi buffer layer is 10 µm and has an average resistivity of 0.15 cm. The trench depth and width are typically 4.2 and 1.4 µm, respectively. The trench was formed using a special fluorine based dry etch process. The thickness of the gate oxide was 25 nm. 2-MeV electron irradiation was performed at the Dinamitron linear electron accelerator at Takasaki Japan Atomic Energy. Perpendicular irradiation to the sample was performed without applied bias to the IGBT. The sample temperature was fixed at 25, 100, 200 and 300 ◦ C and was controlled by a panel heater, mounted in a chamber. The number of samples is five for each irradiation temperature. Electron irradiation was performed through a 50 µm thick Ti window. The electron fluence is 1015 e/cm2 . Fig. 2 shows a schematic view of the high-temperature electron irradiation set-up. After irradiation, the samples were cooled immediately to avoid high temperature annealing of the irradiation damage. Steady-state and switching experimental characteristics are obtained from electrical measure-
Figure 2 Experimental set-up for the high-temperature electron irradiation.
464
Measurement items
Measurement condition
IGES+ IGES− BVCES ICES VCEast VTH
VGE = 10 V VGE = −10 V IC = 1 mA VCE = 10 V Ic = 130 A, VGE = 4 V I D = 1 mA, VCE = 10 V
ments at room temperature. The measurement conditions of the steady-state characteristics are shown in Table I and the switching characteristics were measured using a 5 V gate pulse at one shot with an anode current of 130A through the device before and after irradiation, using a TS3710 measurement system (Tektronix). The lifetime values were determined from the exponentially decaying portion of the turn-off current waveform due to carrier recombination in the epitaxial layer. And the average of measurement data was analyzed. 3. Results and discussion Some IGBT parameters are shown in Figs. 3 and 4 as a function of the irradiation temperature. In Fig. 3 the
Figure 3 Variation of Lifetime, BVCES and ICES after a 1 × 1015 e/cm2 irradiation at different temperatures. The pre-rad value of lifetime, BVCES and ICES are 719 ns, 550 V and 0.006 µA, respectively.
Figure 4 Variation of VTH and VCEsat after a 1015 e/cm2 irradiation at different temperatures. The pre-rad value of VTH and VCEsat are 1.0 V and 3.31 V, respectively.
lifetime and BVCES increase with increasing the irradiation •• temperature, while ICES decreases. For a 300 ◦ C irradiation, the reduction of the lifetime is only 28% of the starting value. This result suggests that the creation and recovery of the radiation damage proceeds simultaneously at high temperatures. On the other hand, Fig. 4 shows VTH and VCEsat as a function of the irradiation temperature. It is found that a partial VTH recovery and a pronounced change of VCEsat are observed in the case of the 100 ◦ C irradiation. For a further increase of the irradiation temperature, the VTH shifts in more negative direction and VCEsat is recovered. In our previous study concerning radiation damage in Si bipolar devices after high-temperature irradiation, the degradation of the device performance for high-temperature irradiation is smaller than that for room temperature irradiation [9]. This phenomenon is the same as the results of Fig. 3 and also leads to the conclusion that the radiation damage in the substrate is annealed out or that less electrically active defects are formed during high-temperature irradiation. Therefore, it is necessary to consider other contributions than the silicon substrate for explaining the results of Fig. 4. It is well known that the irradiation induced positive charge trapped in the gate dielectric and negative charge trapped at the Si-SiO2 interface, cause a corresponding negative and positive VTH shift for n-MOSFETs. Therefore, in order to identify the possible degradation mechanism, the charge separation technique [10] was applied to the subthreshold current-voltage characteristics. The subthreshold transfer characteristics of an IGBT were measured in the saturation region (VCE = 20 V) only, because the IGBT had a forward biased pn junction. The equipment used for electrical characterization is a semiconductor parameter analyzer (HP4156A). The measured transfer characteristics of IGBTs in function of the irradiation temperature are shown in Fig. 5. The contribution of trapped-oxide charge to the threshold shift simply causes a parallel shift of the entire subthreshold current curve to the left. When the bands are
bent by an amount b (midgap condition), the donor traps fall below the Fermi level, and the acceptor traps fall above the Fermi level, [10]. Therefore, the shift between the subthreshold curves at the midgap voltage (VMG ) represents the shift due to trapped charge in the oxide. The value of VMG is determined from the measured subthreshold characteristics at a drain current level IMG . The drain current (IMG ) in the subthreshold region at midgap condition is given by [11]:
IMG
β εs k 2 T 2 = √ Cox 2 q 2 L D
b =
Na kT ln q ni
ni Na
2 1−e
q VD kT
e
qb kT
kT qb (1) (2)
where β is the gain factor which derived from voltagecurrent characteristic, Cox is the gate oxide capacitance per unit area, L D is the Debye length, Na is the concentration of acceptors in the channel, n i is the intrinsic carrier density, εs , is the silicon permittivity, b is Fermi bulk potential, k is Boltzmann’s constant, T is absolute temperature, q is the electronic charge and VD is the drain voltage. To apply this equation to an IGBT, VD is replaced by VCE -0.7. Because the IGBT doping concentration along the channel is non-uniform, an average value of Na (4 × 1016 cm−3 ) is used. Consequently, the contribution to the gate voltage shift due to trapped charge in the oxide is given by: Vot = VMG − VMGbefore
(3)
where values of VMG are determined from measured subthreshold characteristics (Fig. 5) at drain current level IMG , determined by using the Equation 1. On the other hand, a number of acceptor interface traps falls below the Fermi level increase due to band bending from midgap to threshold. Therefore, the subthreshold current curve between midgap and threshold is stretched out, and the stretched out voltage Vso is defined as [10]: Vso = VTH − VMG
(4)
Consequently, the shift in the threshold voltage due to interface traps at each irradiation temperature is the difference in Vso : Vit = Vso − Vsobefore
(5)
The change of the threshold voltage VTH follows from combining the Equations 3 and 5. Hence, VTH is given by: VTH = Vot + Vit
Figure 5 High-temperature irradiation induced shift of subthreshold characteristics in an IGBT in the saturation regime.
(6)
Fig. 6 shows the results of the charge separation. From Vot and Vit , the increase in the number of trapped charges in the oxide Not and the change in interface 465
Figure 8 Correlation between the change in lifetime and in leakage current with Nit for high temperature irradiations. Figure 6 High-temperature irradiation induced threshold voltage shift (VTH ), and contributions of gate oxide charge (Vot ) and interface traps (Vit ) in an IGBT.
trap density Nit can be calculated from: Not = Cox |Vot |q
(7)
Nit = Cox Vit /q
(8)
High-temperature irradiation induced changes in gate oxide charge and interface trap density in an IGBT are shown in Fig. 7. From Figs. 6 and 7, it is found that the temperature dependence of the gate oxide charges and interface trap density are different. It is found that the gate oxide charges change logarithmically and the interface trap density changes linearly with the irradiation temperature, as empirical formulae denoted in Fig. 7. For this reason, the threshold voltage (VTH ) is observed to recover at 100 ◦ C. The following explanation for the Nit decrease with increasing irradiation temperature is tentatively suggested: the diffusion of hydrogen atom released from the gate or gate-oxide interface by the 2-MeV electrons passivates the interface traps during high temperature irradiation. The activa-
tion energy of this process is found to be 0.094 eV. Luo and Sah reported that interface states were annealed out during chip bonding between 85 and 380 ◦ C with the thermal activation energy of 0.15 ± 0.02 eV, [12]. The likely origin of the pronounced change of VCEsat at the 100 ◦ C irradiation is the positive VTH shift and the more pronounced decrease of β, shown in Fig. 5. At the same time, it is found that change in leakage current and in lifetime ((1/τ ) = 1/τpost − 1/τpre ) both correlate with Nit , as shown in Fig. 8. 4. Conclusions The following conclusions are obtained by this study, (1) It is found that the temperature dependence of gate oxide charges and the interface trap density are different. The gate oxide charges change logarithmically with the irradiation temperature and the interface trap density changes linearly. Therefore, VTH is observed to partially recover during 100 ◦ C irradiation. (2) It is tentatively suggested that the diffusion of hydrogen released from the gate by the 2-MeV electrons passivates the interface traps during the high temperature irradiation, which could explain the Nit decrease with irradiation temperature increase. The corresponding activation energy is found to be 0.094 eV. (3) The origin of the pronounced change of VCEsat during the 100 ◦ C irradiation is the smaller VTH and the more pronounced decrease of β. (4) It is found that the leakage current and lifetime correlate with Nit . References 1. B . J . B A L I G A , in “Modern Power Devices” (Wiley, New York,
Figure 7 High-temperature irradiation induced changes in gate oxide charge and interface trap density.
466
1987) p. 52. 2. S . B R O T H E R T O N and P . B R A D L E Y , J. Appl. Phys. 53 (1992) 5720. 3. D . B I E L L E - D A S P E T , Solid-State Electron. 16 (1973) 1103. 4. A . O . E V W A R A Y E and B . J . B A L I G A , J. Electrochem. Soc. 124 (1977) 913. 5. Y . K O N I S H I , Y . O N I S H I , S . M O M O T A and K . S A K U R A I , Proc. of ISPSD’96 23–26 (1996) 335. 6. K . M O C H I Z U K I , K . I S H I I , M . T A K E D A , H . H A G I N O and T . Y A M A D A , Proc. of ISPSD’97 26–29 (1997) 237.
7. J . V O B E C K Y and P . H A Z D R A , Proc. of ISPSD’96 23–26 (1996) 161. ¨ GL, M. 8. M . S C H M I T T , H . - J . S C H U L Z E , A . S C H L O ¨ R G E R , A . W I L L M E R O T H , G . D E B O Y and G . VOSSEB U W A C H U T K A , Proc. of ISPSD’02 4–7 (2002) 229. 9. H . O H Y A M A , T . H I R A O , E . S I M O E N , C . C L A E Y S , M . N A K A B A Y A S H I and S . O N O D A , Solid State Phenomena 82–84 (2002) 465.
10. P . J . M C W H O R T E R and P . S . W I N O K U R , Appl. Phys. Lett. 48 (1986) 133. 11. S . M . S Z E , in “Physics of Semiconductor Devices” (New York, Wiley, 1981) p. 446. 12. M . S . C . L U O and C . T . S A , J. Appl. Phys. 62 (1987) 4940.
Received 15 September 2004 and accepted 27 January 2005
467