On the reliability assessment of trench fieldstop IGBT under ...
Microelectronics Reliability 52 (2012) 124–129
Contents lists available at SciVerse ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/microrel
On the reliability assessment of trench fieldstop IGBT under atmospheric neutron spectrum A.D. Touboul ⇑, L. Foro, F. Wrobel, F. Saigné Université Montpellier 2, IES UMR 5214, Place E. Bataillon, F34095 Montpellier, France
a r t i c l e
i n f o
Article history: Received 11 May 2011 Received in revised form 31 August 2011 Accepted 31 August 2011 Available online 25 September 2011
a b s t r a c t In the early 1990s, active power devices have been shown to be susceptible to radiation-induced failures. Many studies have been focused on cosmic ray-induced power device failures, including even IGBT failures. Till the end of the 1990s, IGBT technologies were susceptible to static or dynamic latch-up, which are respectively occurring in their forward conduction or switching mode. From then on, new technologies, such as trench gate fieldstop IGBTs providing significant improvement in the latching current capability, have been developed. But so far, the impact of atmospheric radiation has not been assessed on actual technologies. With the expected increase of embedded IGBTs in avionics and automotive applications, the impact of neutrons on the functional security of embedded systems has to be quantified. Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction
2. Why atmospheric environment can induce reliability issues?
More than 50 years ago, the first failures induced by radiation were observed on-board satellites in the early 1960s. Even though space is the harsher natural environment regarding radiative constraint and thus the main to be considered, the technological evolution of electronics made elementary devices more and more sensitive to radiation. Nowadays, even particles from atmospheric environment can induce soft or hard failures on embedded electronic systems [1–3]. On the other hand, whatever the environment, from space to ground, power electronics constitutes the core of energy conversion systems, consequently requiring a high reliability level during all their lifetime. This is especially true when considering transportation systems, since failures are unacceptable for those applications. Among components used within power converter systems, Insulated Gate Bipolar Transistors are the most critical ones regarding failure sensitivity due to their high internal electric fields. Facts are IGBTs are even sensitive to atmospheric radiation. Actually, we report in this paper that a single neutron interaction can lead to charged secondary products, themselves able to trigger a failure inside the elementary device. For an elementary up to date IGBT, we have calculated the expected failure in time (FIT) at ground level. This FIT has been obtained at maximum electrical rating and while considering only the high energy part of the neutron energy. This first quantitative result is of prime interest for embedded energy converters-based applications, since a high reliability is needed for their IGBT modules.
2.1. The atmospheric radiative context
⇑ Corresponding author. E-mail address: [email protected] (A.D. Touboul). 0026-2714/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.microrel.2011.08.023
Since the 1980s, it is known that cosmic ray interacting with atmospheric nuclei generate a strong particle shower down to ground as one can see on Fig. 1. The particles which can cause the most significant failures are those experiencing strong interaction, mainly neutrons, protons and pions. Since atmospheric protons and pions are charged particles, they are as well experiencing coulombic interaction with any surrounded electronic cloud. All in all, at sea level, 95% of the particles shower are neutrons [4]. Since neutrons are only experiencing strong interaction, they interact directly with any atom nucleus. Such interaction occurring within the device leads to the creation of secondary charged particles able to initiate very localized currents that can originate in a device failure. In fact, making the simple assumption that the majority of interactions within the device are neutron–silicon reactions, nuclear calculations allow to quantify for each neutron energy the proportion of emitted secondary atoms [5], as reported on Fig. 2. To make it simple, about two silicon recoils will be secondary emitted from the bulk every 10 neutron interactions with the device. Those nuclei are among the heaviest and then the most dangerous secondary emitted particles from the inside of the device, and leading possibly to significant energy deposition. At ground level, the flux of neutron above 1 MeV is about 30 n/cm2/ h. Considering a silicon volume corresponding to the bulk of a typical IGBT, the atmospheric neutron flux will roughly lead to one nuclear interaction every day. Obviously, upper layers will play a role as well in the number of interactions and should be, if possible, taken into account.
A.D. Touboul et al. / Microelectronics Reliability 52 (2012) 124–129
125
Fig. 3. Basic technological structure of a planar gate oxide IGBT. Fig. 1. Schematic view of a cosmic ray shower.
Fig. 2. Proportion of emitted secondary atoms from neutron–silicon interaction.
2.2. Applicative and industrial context As mentioned before, power electronics play a crucial role since switching devices constitute the heart of energy converter systems. Technical progress in power electronics are mainly driven by the development of circuit topologies [6], control techniques [7], fault diagnosis and handling [8–10], and power devices technologies [11,12]. In this section, we will especially focus on this last field. Indeed, when looking at power devices technology progress along the last five decades, one can easily notice that modern power devices have undergone a radical mutation. Silicon Controlled Rectifiers, SCR, and Bipolar Junction Transistors, BJT, left room for Gate Turn-Off thyristors, GTO, power MOSFETs and Insulated Gate Bipolar Transistors, IGBT. Even though this technological mutation of power devices has been favorable to a higher efficiency in handling large currents at high voltage and switching frequency, the technology of IGBTs and of Vertical Diffused Metal Oxide Semiconductor Transistors, VDMOS, make them intrinsically sensitive to some radiationinduced destructive events, namely Single Event Burnout (SEB) and Single Event Latchup (SEL) whose physical mechanisms are described in Section 2.3. 2.3. Atmospheric neutrons-induced destructive single events Atmospheric neutron spectrum-induced system dysfunction has already been observed [1,3,13,14]. The main question to rise is ‘‘what is called a dysfunction’’? We can distinguish soft errors that are non-permanent dysfunctions mainly affecting memories from destructive events which are permanent failures. Since atmospheric spectrum is mainly made of neutrons, that are uncharged particles and therefore very difficult to stop, we will deal through this paper with neutron-induced permanent failures. A lot of studies have been devoted in the 1980s to the understanding of basic mechanisms related to heavy ion-induced catastrophic failures of
Fig. 4. How secondary ions are emitted from the inside of the device after a primary neutron interaction. Effect on the triggering of the IGBT parasitic components.
power MOSFETs and IGBTs. An abundant literature can be found here [13–21]. For all that, in order to improve the clarity of this paper, we will shortly remind within this section the basic mechanisms for neutron-induced SEB and SEL triggering on an elementary IGBT. To this end, the general technology of the IGBT is reported on Fig. 3. The invention of the IGBT structure has been a valuable improvement in the field of power electronic devices. This device, developed in the 1980s [22] combines the advantage of the MOS and bipolar structure. It is basically a PNP bipolar transistor driven by an NMOS transistor as indicated on Fig. 3. Fig. 3 is indicating the expected electrical functioning of the IGBT (i.e. a PNP driven by an NMOS). However, looking at the technological structure of this device allows also revealing the presence of both parasitic NPN bipolar transistor and NPNP Silicon Controlled Rectifier (Thyristor) structure, as reported on Fig. 4. When the former is responsible for SEB triggering, the latter is responsible for latchup. But we have here to distinguish the purely electrical failure modes (static and dynamic latch-up) related to the ability of the device design to withstand out-of-range specifications, from the single particle-induced failure that is affecting a single cell of the device.
126
A.D. Touboul et al. / Microelectronics Reliability 52 (2012) 124–129
Nothing but a technological mitigation can prevent IGBT based energy conversion system from this radiative failure mode. In particular, the sensitive state for those radiative failure modes is the OFF state. Let us say clearly that in the early 1990s, many studies have been focused on cosmic ray-induced power device failures, including even IGBT failures [18,19,23–25]. Till the end of the 1990s, IGBT technologies were susceptible to static or dynamic latch-up, that are respectively occurring in their forward conduction or switching mode. From then on, new technologies, such as trench gate fieldstop IGBTs providing significant improvement in the latching current capability, have been developed. Such a technology is relevant regarding those of interest for present and future applications [26]. But so far, even though some recent studies have reported neutron-induced failures on power devices [27–29], the impact of atmospheric radiation has not been assessed on actual technologies. With the expected increase of embedded IGBTs in avionics and automotive applications, the impact of neutrons on the functional security of embedded systems has to be quantified. To reach this goal, the basics of the radiative failure physical mechanisms have to be reminded. As indicated before, the sensitive state for which destructive events can occur is the OFF-state. In the industrial application, the power device is operating in a dynamic mode. During the switching cycle, the worst case is then when the Anode (Collector) is positively biased and the Gate either reversed biased or grounded. Let us now, for this specific operating state, shortly describe the physical mechanisms related to both SEL and SEB triggering. When single event latchup is a one step process, single event burnout is a two steps phenomenon. First step, which is common to both SEL and SEB, is the ON-state triggering of the parasitic NPN transistor. When a secondary ion is emitted from the inside of the device due to a primary neutron–silicon nuclear reaction, this moving charged atom ionizes the silicon along its path, creating electron–hole pairs. Under the high electric field between Anode and Cathode (Collector and Emitter), electron–hole pairs are separated, as one can see on Fig. 4. Electrons quickly flow to the anode. Holes are flowing to the P diffusion, leading to a drop voltage that forward biases the PN+ junction. The NPN parasitic bipolar transistor is then ON. At this stage, all necessary required conditions for radiative latchup triggering are filled. If the structure is not sensitive to SEL, SEB can still occur. As indicated before, triggering of an SEB is a two steps process. Once the parasitic NPN is ON, if it injects a high enough electron current from NPN emitter to NPN collector, meaning from IGBT cathode to N epitaxial region, a so-called ‘‘Kirk effect’’ is expected [30,31]. A base region spreading occurs into the collector region of the parasitic BJT, leading to a relocation of the electric field from the PN to the NP+ heterojunction. Impact ionization can then occur into the collector P+ region, injecting consequently holes back to the base region, allowing keeping in ON state the parasitic NPN BJT. This way, the single event burnout phenomenon becomes irreversible, leading to a thermal runaway device and thus to a catastrophic failure of the device. Since SEL and SEB are catastrophic failures, their occurrence during application has to be assessed. In particular, contrary to aging failure modes [32], no indicator exist to anticipate the failure. When considering external constraints, either electrical, thermal or radiative, the basics of testing is related to the ability to test devices under the considered constraint with a reasonable accelerated factor. This is already widely used when performing accelerated aging tests, in particular through power cycling tests. Similar accelerated procedures are used for radiation qualification. Let us focus on an example of application, lifetime requirements for power applications in hybrid electric vehicles are listed hereafter: about 15 years of service lifetime for a mileage of 600,000 km, corresponding to 12,000 h of engine on time [33]. Even though atmospheric neutron flux at
ground level is relatively low (about 30 n/cm2/h, which is 300 times lower than at avionic altitudes), the long service lifetime of the hybrid electric vehicle and the number of vehicles in use make it imperative to assess the reliability of embedded power converters submitted to neutron flux. To this end accelerated tests are necessary. 3. Experimental details 3.1. Neutron beam facility The facility used for this study is located at Theodor Svedberg Laboratory (TSL), Uppsala, Sweden. We report here experiment results obtained after irradiation with the 146 MeV quasi-monoenergetic neutron beam facility. The neutron flux delivered by the beam was about 5.3 104 cm2 s1. The relative spectral fluence is reported in Fig. 5. 3.2. Devices under test and beam testing methodology The parts whose sensitivity to neutron has been investigated were trench Insulated Gate Bipolar Transistors in fieldstop technology with a fast recovery anti-parallel diode. Those transistors were assembled in TO-247 packages. Their maximum ratings are VCE = 1200 V, IC = 8 A. During the last three decades, test procedures for the measurement of destructive single event effects on semiconductor devices have been extensively developed for and by the main European and International space actors [34–36]. The main goal was to be able to predict a failure rate during application. Even though those methods are pretty restrictive, all in all they ensure a worst case of failure rate extraction, meaning that during application the reliability of the system can only be better than expected. In fact, being conservative during the test implies rejecting devices that might function during mission, but being sure that devices that passed the test will work all over the mission. Since no dedicated experimental procedure for conducting destructive SEE tests for avionic and ground applications has been developed yet, one has to use first space qualification methods. Indeed, in order to assess the susceptibility of last generation IGBT technologies to destructive single events, there is no choice but to follow the existing test standards. But doing so, one has to be aware that those standards do not take into account the real operating mode of the device during application, nor the mission parameters.
Fig. 5. Relative spectral fluence versus energy for quasi monoenergetic 146 MeV neutrons.
127
A.D. Touboul et al. / Microelectronics Reliability 52 (2012) 124–129
Fig. 8. Integral neutron differential flux at ground level versus neutron energy. Neutron flux above 146 MeV is about 1.3 103 cm2 s1.
Fig. 6. Single event burnout/latchup board test during 146 MeV neutron irradiation.
Typical destructive SEE test circuit architectures can be found in [35,36]. Dedicated board test used for monitoring collector current during irradiation is presented on Fig. 6. Since single event burnout or latchup result in a short-circuit defect, the failure monitoring can be performed by watching either the power supply current increase or the collector – emitter voltage. Using a high voltage supply, the capacitance basically acts as energy storage for allowing any destructive event to be triggered. As a consequence, RA is the limiting current resistance for protecting the HV supply in case a destructive event occurs. As explained in Section 2, the sensitive state for which destructive events occur is the OFF state. That is why test methods for SEB/SEL measurements recommend biasing collector – emitter voltage at the rated voltage and gate – emitter voltage at 0 V. So did we during irradiation. Since the control room was 100 m far from the irradiation room, only the collector – emitter voltage and the power supply current were monitored. No transient current shape related to the charge deposition was acquired during irradiation. We especially focus here on the Time To Failure (TTF) under neutron beam problematic. 4. Experimental results after exposure, failure analysis IGBT devices coming from the same batch have been irradiated under 146 MeV quasi monoenergetic neutron flux, using the board test described before. The time to failure has then been measured for 13 IGBT devices. Neutron flux under beam was given about 5.3 104 cm2 s1. Such an order of magnitude represents an accelerated flux regarding atmospheric neutron flux at ground level, as one can see on Fig. 7.
However, one has to consider that the failures that have been triggered under 146 MeV QMN beam have a probability of occurrence directly driven by the neutron energy. To make it short, if a 146 MeV neutron can lead to a destructive SEE, all the more so. It means that a best case for failure analysis would be to consider that all neutrons above 146 MeV might lead to (at least) the same failure rate. The assumption that will be used is that the failure rate above 146 MeV is constant. From the atmospheric neutron flux, all the neutrons above 146 MeV will have to be considered. From the integral neutron flux above an energy E, plotted on Fig. 8, we can note that the neutron flux above 146 MeV is about 1.3 103 cm2 s1. We can now define the acceleration factor, AF, as being the ratio between the 146 MeV neutron flux under beam and the integral atmospheric integral neutron flux above 146 MeV.
AF ¼
ubeam ¼ 4 107 uE>146MeV
The time to failure under beam has been reported on Table 1. As a preliminary analysis, we can observe that this set of experimental data exhibits a significant variability, since the range is larger than the sample mean. For this reason, a Weibull distribution is used in order to analyze the impact of neutron flux on the device reliability [37,38]. Fig. 9 reports ln(ln(1F)) versus time to failure experimental data. F is the cumulative failure probability. The best model for characterizing the distribution of time to failure has been investigated based on a Weibull Probability Plot. It looks that a two parameters Weibull model, so-called standard model, allows describing our set of data. The standard Weibull model is given by the distribution function:
F Acc Test ðtÞ ¼ 1 e
g t Acc Test
b
where b is the shape parameter and gAcc Test is the scale parameter. The determination of gAcc Test is pretty straightforward since it is the time to failure value for which 63.2% of the irradiated devices exhibited a failure, whatever the value of b. The graphical extraction gives gAcc Test = 30 s and b = 1. Those parameters allow to describe the cumulative failure rate under beam. The reliability of the device under natural environment can then be easily extracted from the accelerated test. Based on the assumption that the failure mechanisms related to neutron interaction are the same under beam than
Table 1 Failure rate under beam report.
Fig. 7. Atmospheric neutron differential flux at ground level versus neutron energy obtained with QinetiQ Atmospheric Radiation Model (QARM).
IGBT #
1
2
3
4
5
6
7
TTF (s) IGBT # TTF (s)
14 8 14
12 9 12
64 10 64
7 11 7
30 12 30
106 13 106
34 – –
128
A.D. Touboul et al. / Microelectronics Reliability 52 (2012) 124–129
a significant global amount of systems, the reliability margins will have to be tightened. In particular, dedicated test standards for ground level applications will have to be developed. Depending on the application, the derating of the device bias has to be applied during test. By the way, the FIT value might be reduced. Indeed, regarding the expected number of devices embedded in automotive or avionic applications, a specific standard closer to the real operating mode of the device, being less conservative and allowing to reduce reliability margin will have to be brought up. Fig. 9. Weibull distribution of the cumulative failure probability versus time to failure under neutron beam.
5.2. On the assessment of the reliability of the system The extracted shape parameter equal to 1, means that only random failures are considered. This is in agreement with the physical mechanisms of the failure under accelerated constraint, such as the high monoenergetic neutron flux that we used. However, even though there is a good correlation between the failure distribution and the failure mechanisms that have occurred, one can be sure that this distribution does not fit the expected one in ‘‘real life’’. Indeed, the hazard function, often known as the bathtub curve is in our case (b = 1), a constant. Fig. 10. Weibull distribution of the expected cumulative failure probability versus time to failure under atmospheric neutron flux above 146 MeV.
hðtÞ ¼ in the natural atmospheric environment, we can easily extract the cumulative failure distribution of this device at ground level in the natural environment. To this end, let’s now define the expected scale parameter at ground level as follows:
dFðtÞ dt
1 FðtÞ
¼
1
g
b t b1 ¼
1
g
It means that we are only considering the stable period of the device. In particular, IGBTs are supporting during application strong electrical and thermal cycles that can impact the failure modes along life.
g ¼ gAcc Test AF And then,
t b FðtÞ ¼ 1 e ðgÞ ¼ 1 e
t g Acc Test AF
b
The cumulative failure distribution in natural environment is plotted on Fig. 10. 5. Discussion If we consider an exponential distribution with a shape parameter equal to 1, the failure rate is equal to 1/g. Then, the failure in time which is defined as the number of failure over 109 h, is 3000 per device. This result is actually addressing two fields of interest, the accuracy of accelerated standard test of power devices under beam and the assessment of the reliability of the system from the device reliability. 5.1. On the test of active power devices under beam As indicated in Section 3.2, all the existing test methods recommend testing the device to SEB/SEGR in static OFF mode. The main question to rise is ‘‘does the data extracted from this kind of test represent significant results?’’ Indeed, in the application, power devices are mainly used as switch. Very few papers have studied the impact of dynamic operating on the failure rate under radiation on power devices and none on IGBTs [39,40]. The main conclusion of the authors is that the static to dynamic ratio of destructive single events is over a factor 100. Obviously this ratio is supposed to be strongly dependant on the device technology. But it indicates as well that for some critical applications at ground level involving
6. Conclusion In the beginning of this paper we have reminded how the atmospheric environment is naturally radiative. We are indeed permanently submitted to neutrons with a flux about 30 n/cm2/ h. Even though human are immune to this flux, it is not the case for electronics. In particular IGBTs are sensitive to two catastrophic failure mechanisms (SEB and SEL) that can be triggered by a single primary neutron interaction. Looking backward to the long range experience of space actors in the field of device qualification under radiation, one can also look at the past three decades improvement of radiation testing methodology. Following those Single Event Effects testing recommendations, irradiation of IGBTs under 146 MeV neutron beam allowed us to calculate the expected failure in time at ground level. The obtained value, obtained while considering only the high energy part of the neutron spectrum, indicates that even modern IGBTs might be susceptible to destructive single events at ground level. Indeed, the IGBT whose results have been reported here are from a trench gate fieldstop technology that is relevant for modern applications. In particular, considering the large number of energy converters expected to be embedded at ground level in the future makes it important to assess more precisely the impact of the atmospheric environment on the system reliability. This will allow to determine whether the natural radiative constraint get into the reliability margin or not, that means if the IGBT sensitivity to natural atmospheric neutrons will really become a reliability concern. At present, strong efforts have to be done in order to adapt existing test methods for space applications to ground level systems, in order to test devices closer to their real operating mode. As a consequence, such a kind of standard will be less conservative and will allow to reduce reliability margin.
A.D. Touboul et al. / Microelectronics Reliability 52 (2012) 124–129
References [1] Leray J. Effects of atmospheric neutrons on devices, at sea level and in avionics embedded systems. Microelectron Reliab 2007;47:1827–35. [2] Normand E et al. Neutron-induced single event burnout in high voltage electronics. IEEE Trans Nucl Sci 2002;44:2358–66. [3] Ziegler J, Lanford W. The effect of sea level cosmic rays on electronic devices. J Appl Phys 2009;52:4305–12. [4] Ziegler J. Terrestrial cosmic ray intensities. IBM J Res Dev 2010;42:117–40. [5] Wrobel F, Saigné F. MC-ORACLE: A tool for predicting soft error rate. Comput Phys Commun 2011;182:317–21. [6] Azzopardi S et al. Assessment of the trench IGBT reliability: low temperature experimental characterization. Microelectron Reliab 2005;45:1700–5. [7] Cortes P et al. Predictive control in power electronics and drives. IEEE Trans Ind Electron 2008;55:4312–24. [8] Turpin C et al. Fault management of multicell converters. IEEE Trans Ind Electron 2002;49:988–97. [9] Akin B et al. A simple real-time fault signature monitoring tool for motor drive embedded fault diagnosis systems. IEEE Trans Ind Electron 2010;PP(99):1. [10] Karimi S et al. FPGA-based real-time power converter failure diagnosis for wind energy conversion systems. IEEE Trans Ind Electron 2008;55: 4299–308. [11] Bose B. Technology trends in microcomputer control of electrical machines. IEEE Trans Ind Electron 2002;35:160–77. [12] Bose B. Power electronics – an emerging technology. IEEE Trans Ind Electron 2002;36:403–12. [13] Wrobel T, Azarewicz J. High dose rate burnout in silicon epitaxial transistors. IEEE Trans Nucl Sci 2007;27:1411–5. [14] Wrobel T et al. Current induced avalanche in epitaxial structures. IEEE Trans Nucl Sci 2007;32:3991–5. [15] Allenspach M et al. SEGR and SEB in n-channel power MOSFETs. IEEE Trans Nucl Sci 2002;43:2927–31. [16] Dachs C et al. Simulation aided hardening of N-channel power MOSFETs to prevent single event burnout. IEEE Trans Nucl Sci 2002;42:1935–9. [17] Johnson G et al. Simulating single-event burnout of n-channel power MOSFET’s. IEEE Trans Electron Devices 2002;40:1001–8. [18] Lorfevre E et al. Heavy ion induced failures in a power IGBT. IEEE Trans Nucl Sci 2002;44:2353–7. [19] Lorfevre E et al. Failure mode of different irradiated power IGBT structures; 2002. p. 516–9. [20] Lorfevre E et al. Cell design modifications to harden an N-channel power IGBT against single event latchup. IEEE Trans Nucl Sci 2002;46:1410–4.
129
[21] Lorfevre E et al. SEB occurrence in a VIP: influence of the epi-substrate junction; 2002. p. 557–60. [22] Baliga B et al. The insulated gate transistor: a new three-terminal MOScontrolled bipolar power device. IEEE Trans Electron Devices 2005;31: 821–8. [23] Findeisen C et al. Extrapolation of cosmic ray induced failures from test to field conditions for IGBT modules. Microelectron Reliab 1998;38:1335–9. [24] Kabza H et al. Cosmic radiation as a cause for power device failure and possible countermeasures; 2002. p. 9–12. [25] Zeller H. Cosmic ray induced failures in high power semiconductor devices. Solid State Electron 1995;38:2041–6. [26] Bauer J et al. New IGBT development for traction drive and wind power; 2010. p. 768–72. [27] Nishida S et al. Cosmic ray ruggedness of IGBTs for hybrid vehicles; 2010. p. 129–32. [28] Nishida S et al. Cosmic ray ruggedness of power semiconductor devices for hybrid vehicles. IEEJ Trans Electron Inf Syst 2010;130:934–8. [29] Shoji T et al. Neutron induced single-event burnout of IGBT; 2010. p. 142–8. [30] Kirk Jr C. A theory of transistor cutoff frequency falloff at high current densities. IRE Trans Electron Dev 2005;9:164–74. [31] Sze S, Ng K. Physics of semiconductor devices. Wiley-Blackwell; 2007. [32] Xiong Y et al. Prognostic and warning system for power-electronic modules in electric, hybrid electric, and fuel-cell vehicles. IEEE Trans Ind Electron 2008;55:2268–76. [33] Scheuermann U. Reliability challenges of automotive power electronics. Microelectron Reliab 2009;49:1319–25. [34] ASTM F1192-00. In: Standard guide for the measurement of Single Event Phenomena (SEP) induced by heavy ion irradiation of semiconductor devices, editor. vol. F1192. 2006. [35] Standard EJ. JESD57. In: Test procedures for the measurement of single-event effects in semiconductor devices from heavy ion irradiation, editor. 1996. [36] Defense DO. MIL-STD-750E test methods for semiconductor devices. In: Method 1080 single-event burnout and single-event gate rupture ed; 2006. [37] Abernethy R. The new Weibull handbook. Gulf Publication; 1996. [38] Murthy D et al. Weibull models. Wiley-Interscience; 2004. [39] Calvel P et al. Comparison of experimental measurements of power MOSFET SEBs in dynamic and static modes. IEEE Trans Nucl Sci 2002;38: 1310–4. [40] Stassinopoulos E et al. Charge generation by heavy ions in power MOSFETs, burnout space predictions and dynamic SEB sensitivity. IEEE Trans Nucl Sci 2002;39:1704–11.
Contents lists available at SciVerse ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/microrel
On the reliability assessment of trench fieldstop IGBT under atmospheric neutron spectrum A.D. Touboul ⇑, L. Foro, F. Wrobel, F. Saigné Université Montpellier 2, IES UMR 5214, Place E. Bataillon, F34095 Montpellier, France
a r t i c l e
i n f o
Article history: Received 11 May 2011 Received in revised form 31 August 2011 Accepted 31 August 2011 Available online 25 September 2011
a b s t r a c t In the early 1990s, active power devices have been shown to be susceptible to radiation-induced failures. Many studies have been focused on cosmic ray-induced power device failures, including even IGBT failures. Till the end of the 1990s, IGBT technologies were susceptible to static or dynamic latch-up, which are respectively occurring in their forward conduction or switching mode. From then on, new technologies, such as trench gate fieldstop IGBTs providing significant improvement in the latching current capability, have been developed. But so far, the impact of atmospheric radiation has not been assessed on actual technologies. With the expected increase of embedded IGBTs in avionics and automotive applications, the impact of neutrons on the functional security of embedded systems has to be quantified. Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction
2. Why atmospheric environment can induce reliability issues?
More than 50 years ago, the first failures induced by radiation were observed on-board satellites in the early 1960s. Even though space is the harsher natural environment regarding radiative constraint and thus the main to be considered, the technological evolution of electronics made elementary devices more and more sensitive to radiation. Nowadays, even particles from atmospheric environment can induce soft or hard failures on embedded electronic systems [1–3]. On the other hand, whatever the environment, from space to ground, power electronics constitutes the core of energy conversion systems, consequently requiring a high reliability level during all their lifetime. This is especially true when considering transportation systems, since failures are unacceptable for those applications. Among components used within power converter systems, Insulated Gate Bipolar Transistors are the most critical ones regarding failure sensitivity due to their high internal electric fields. Facts are IGBTs are even sensitive to atmospheric radiation. Actually, we report in this paper that a single neutron interaction can lead to charged secondary products, themselves able to trigger a failure inside the elementary device. For an elementary up to date IGBT, we have calculated the expected failure in time (FIT) at ground level. This FIT has been obtained at maximum electrical rating and while considering only the high energy part of the neutron energy. This first quantitative result is of prime interest for embedded energy converters-based applications, since a high reliability is needed for their IGBT modules.
2.1. The atmospheric radiative context
⇑ Corresponding author. E-mail address: [email protected] (A.D. Touboul). 0026-2714/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.microrel.2011.08.023
Since the 1980s, it is known that cosmic ray interacting with atmospheric nuclei generate a strong particle shower down to ground as one can see on Fig. 1. The particles which can cause the most significant failures are those experiencing strong interaction, mainly neutrons, protons and pions. Since atmospheric protons and pions are charged particles, they are as well experiencing coulombic interaction with any surrounded electronic cloud. All in all, at sea level, 95% of the particles shower are neutrons [4]. Since neutrons are only experiencing strong interaction, they interact directly with any atom nucleus. Such interaction occurring within the device leads to the creation of secondary charged particles able to initiate very localized currents that can originate in a device failure. In fact, making the simple assumption that the majority of interactions within the device are neutron–silicon reactions, nuclear calculations allow to quantify for each neutron energy the proportion of emitted secondary atoms [5], as reported on Fig. 2. To make it simple, about two silicon recoils will be secondary emitted from the bulk every 10 neutron interactions with the device. Those nuclei are among the heaviest and then the most dangerous secondary emitted particles from the inside of the device, and leading possibly to significant energy deposition. At ground level, the flux of neutron above 1 MeV is about 30 n/cm2/ h. Considering a silicon volume corresponding to the bulk of a typical IGBT, the atmospheric neutron flux will roughly lead to one nuclear interaction every day. Obviously, upper layers will play a role as well in the number of interactions and should be, if possible, taken into account.
A.D. Touboul et al. / Microelectronics Reliability 52 (2012) 124–129
125
Fig. 3. Basic technological structure of a planar gate oxide IGBT. Fig. 1. Schematic view of a cosmic ray shower.
Fig. 2. Proportion of emitted secondary atoms from neutron–silicon interaction.
2.2. Applicative and industrial context As mentioned before, power electronics play a crucial role since switching devices constitute the heart of energy converter systems. Technical progress in power electronics are mainly driven by the development of circuit topologies [6], control techniques [7], fault diagnosis and handling [8–10], and power devices technologies [11,12]. In this section, we will especially focus on this last field. Indeed, when looking at power devices technology progress along the last five decades, one can easily notice that modern power devices have undergone a radical mutation. Silicon Controlled Rectifiers, SCR, and Bipolar Junction Transistors, BJT, left room for Gate Turn-Off thyristors, GTO, power MOSFETs and Insulated Gate Bipolar Transistors, IGBT. Even though this technological mutation of power devices has been favorable to a higher efficiency in handling large currents at high voltage and switching frequency, the technology of IGBTs and of Vertical Diffused Metal Oxide Semiconductor Transistors, VDMOS, make them intrinsically sensitive to some radiationinduced destructive events, namely Single Event Burnout (SEB) and Single Event Latchup (SEL) whose physical mechanisms are described in Section 2.3. 2.3. Atmospheric neutrons-induced destructive single events Atmospheric neutron spectrum-induced system dysfunction has already been observed [1,3,13,14]. The main question to rise is ‘‘what is called a dysfunction’’? We can distinguish soft errors that are non-permanent dysfunctions mainly affecting memories from destructive events which are permanent failures. Since atmospheric spectrum is mainly made of neutrons, that are uncharged particles and therefore very difficult to stop, we will deal through this paper with neutron-induced permanent failures. A lot of studies have been devoted in the 1980s to the understanding of basic mechanisms related to heavy ion-induced catastrophic failures of
Fig. 4. How secondary ions are emitted from the inside of the device after a primary neutron interaction. Effect on the triggering of the IGBT parasitic components.
power MOSFETs and IGBTs. An abundant literature can be found here [13–21]. For all that, in order to improve the clarity of this paper, we will shortly remind within this section the basic mechanisms for neutron-induced SEB and SEL triggering on an elementary IGBT. To this end, the general technology of the IGBT is reported on Fig. 3. The invention of the IGBT structure has been a valuable improvement in the field of power electronic devices. This device, developed in the 1980s [22] combines the advantage of the MOS and bipolar structure. It is basically a PNP bipolar transistor driven by an NMOS transistor as indicated on Fig. 3. Fig. 3 is indicating the expected electrical functioning of the IGBT (i.e. a PNP driven by an NMOS). However, looking at the technological structure of this device allows also revealing the presence of both parasitic NPN bipolar transistor and NPNP Silicon Controlled Rectifier (Thyristor) structure, as reported on Fig. 4. When the former is responsible for SEB triggering, the latter is responsible for latchup. But we have here to distinguish the purely electrical failure modes (static and dynamic latch-up) related to the ability of the device design to withstand out-of-range specifications, from the single particle-induced failure that is affecting a single cell of the device.
126
A.D. Touboul et al. / Microelectronics Reliability 52 (2012) 124–129
Nothing but a technological mitigation can prevent IGBT based energy conversion system from this radiative failure mode. In particular, the sensitive state for those radiative failure modes is the OFF state. Let us say clearly that in the early 1990s, many studies have been focused on cosmic ray-induced power device failures, including even IGBT failures [18,19,23–25]. Till the end of the 1990s, IGBT technologies were susceptible to static or dynamic latch-up, that are respectively occurring in their forward conduction or switching mode. From then on, new technologies, such as trench gate fieldstop IGBTs providing significant improvement in the latching current capability, have been developed. Such a technology is relevant regarding those of interest for present and future applications [26]. But so far, even though some recent studies have reported neutron-induced failures on power devices [27–29], the impact of atmospheric radiation has not been assessed on actual technologies. With the expected increase of embedded IGBTs in avionics and automotive applications, the impact of neutrons on the functional security of embedded systems has to be quantified. To reach this goal, the basics of the radiative failure physical mechanisms have to be reminded. As indicated before, the sensitive state for which destructive events can occur is the OFF-state. In the industrial application, the power device is operating in a dynamic mode. During the switching cycle, the worst case is then when the Anode (Collector) is positively biased and the Gate either reversed biased or grounded. Let us now, for this specific operating state, shortly describe the physical mechanisms related to both SEL and SEB triggering. When single event latchup is a one step process, single event burnout is a two steps phenomenon. First step, which is common to both SEL and SEB, is the ON-state triggering of the parasitic NPN transistor. When a secondary ion is emitted from the inside of the device due to a primary neutron–silicon nuclear reaction, this moving charged atom ionizes the silicon along its path, creating electron–hole pairs. Under the high electric field between Anode and Cathode (Collector and Emitter), electron–hole pairs are separated, as one can see on Fig. 4. Electrons quickly flow to the anode. Holes are flowing to the P diffusion, leading to a drop voltage that forward biases the PN+ junction. The NPN parasitic bipolar transistor is then ON. At this stage, all necessary required conditions for radiative latchup triggering are filled. If the structure is not sensitive to SEL, SEB can still occur. As indicated before, triggering of an SEB is a two steps process. Once the parasitic NPN is ON, if it injects a high enough electron current from NPN emitter to NPN collector, meaning from IGBT cathode to N epitaxial region, a so-called ‘‘Kirk effect’’ is expected [30,31]. A base region spreading occurs into the collector region of the parasitic BJT, leading to a relocation of the electric field from the PN to the NP+ heterojunction. Impact ionization can then occur into the collector P+ region, injecting consequently holes back to the base region, allowing keeping in ON state the parasitic NPN BJT. This way, the single event burnout phenomenon becomes irreversible, leading to a thermal runaway device and thus to a catastrophic failure of the device. Since SEL and SEB are catastrophic failures, their occurrence during application has to be assessed. In particular, contrary to aging failure modes [32], no indicator exist to anticipate the failure. When considering external constraints, either electrical, thermal or radiative, the basics of testing is related to the ability to test devices under the considered constraint with a reasonable accelerated factor. This is already widely used when performing accelerated aging tests, in particular through power cycling tests. Similar accelerated procedures are used for radiation qualification. Let us focus on an example of application, lifetime requirements for power applications in hybrid electric vehicles are listed hereafter: about 15 years of service lifetime for a mileage of 600,000 km, corresponding to 12,000 h of engine on time [33]. Even though atmospheric neutron flux at
ground level is relatively low (about 30 n/cm2/h, which is 300 times lower than at avionic altitudes), the long service lifetime of the hybrid electric vehicle and the number of vehicles in use make it imperative to assess the reliability of embedded power converters submitted to neutron flux. To this end accelerated tests are necessary. 3. Experimental details 3.1. Neutron beam facility The facility used for this study is located at Theodor Svedberg Laboratory (TSL), Uppsala, Sweden. We report here experiment results obtained after irradiation with the 146 MeV quasi-monoenergetic neutron beam facility. The neutron flux delivered by the beam was about 5.3 104 cm2 s1. The relative spectral fluence is reported in Fig. 5. 3.2. Devices under test and beam testing methodology The parts whose sensitivity to neutron has been investigated were trench Insulated Gate Bipolar Transistors in fieldstop technology with a fast recovery anti-parallel diode. Those transistors were assembled in TO-247 packages. Their maximum ratings are VCE = 1200 V, IC = 8 A. During the last three decades, test procedures for the measurement of destructive single event effects on semiconductor devices have been extensively developed for and by the main European and International space actors [34–36]. The main goal was to be able to predict a failure rate during application. Even though those methods are pretty restrictive, all in all they ensure a worst case of failure rate extraction, meaning that during application the reliability of the system can only be better than expected. In fact, being conservative during the test implies rejecting devices that might function during mission, but being sure that devices that passed the test will work all over the mission. Since no dedicated experimental procedure for conducting destructive SEE tests for avionic and ground applications has been developed yet, one has to use first space qualification methods. Indeed, in order to assess the susceptibility of last generation IGBT technologies to destructive single events, there is no choice but to follow the existing test standards. But doing so, one has to be aware that those standards do not take into account the real operating mode of the device during application, nor the mission parameters.
Fig. 5. Relative spectral fluence versus energy for quasi monoenergetic 146 MeV neutrons.
127
A.D. Touboul et al. / Microelectronics Reliability 52 (2012) 124–129
Fig. 8. Integral neutron differential flux at ground level versus neutron energy. Neutron flux above 146 MeV is about 1.3 103 cm2 s1.
Fig. 6. Single event burnout/latchup board test during 146 MeV neutron irradiation.
Typical destructive SEE test circuit architectures can be found in [35,36]. Dedicated board test used for monitoring collector current during irradiation is presented on Fig. 6. Since single event burnout or latchup result in a short-circuit defect, the failure monitoring can be performed by watching either the power supply current increase or the collector – emitter voltage. Using a high voltage supply, the capacitance basically acts as energy storage for allowing any destructive event to be triggered. As a consequence, RA is the limiting current resistance for protecting the HV supply in case a destructive event occurs. As explained in Section 2, the sensitive state for which destructive events occur is the OFF state. That is why test methods for SEB/SEL measurements recommend biasing collector – emitter voltage at the rated voltage and gate – emitter voltage at 0 V. So did we during irradiation. Since the control room was 100 m far from the irradiation room, only the collector – emitter voltage and the power supply current were monitored. No transient current shape related to the charge deposition was acquired during irradiation. We especially focus here on the Time To Failure (TTF) under neutron beam problematic. 4. Experimental results after exposure, failure analysis IGBT devices coming from the same batch have been irradiated under 146 MeV quasi monoenergetic neutron flux, using the board test described before. The time to failure has then been measured for 13 IGBT devices. Neutron flux under beam was given about 5.3 104 cm2 s1. Such an order of magnitude represents an accelerated flux regarding atmospheric neutron flux at ground level, as one can see on Fig. 7.
However, one has to consider that the failures that have been triggered under 146 MeV QMN beam have a probability of occurrence directly driven by the neutron energy. To make it short, if a 146 MeV neutron can lead to a destructive SEE, all the more so. It means that a best case for failure analysis would be to consider that all neutrons above 146 MeV might lead to (at least) the same failure rate. The assumption that will be used is that the failure rate above 146 MeV is constant. From the atmospheric neutron flux, all the neutrons above 146 MeV will have to be considered. From the integral neutron flux above an energy E, plotted on Fig. 8, we can note that the neutron flux above 146 MeV is about 1.3 103 cm2 s1. We can now define the acceleration factor, AF, as being the ratio between the 146 MeV neutron flux under beam and the integral atmospheric integral neutron flux above 146 MeV.
AF ¼
ubeam ¼ 4 107 uE>146MeV
The time to failure under beam has been reported on Table 1. As a preliminary analysis, we can observe that this set of experimental data exhibits a significant variability, since the range is larger than the sample mean. For this reason, a Weibull distribution is used in order to analyze the impact of neutron flux on the device reliability [37,38]. Fig. 9 reports ln(ln(1F)) versus time to failure experimental data. F is the cumulative failure probability. The best model for characterizing the distribution of time to failure has been investigated based on a Weibull Probability Plot. It looks that a two parameters Weibull model, so-called standard model, allows describing our set of data. The standard Weibull model is given by the distribution function:
F Acc Test ðtÞ ¼ 1 e
g t Acc Test
b
where b is the shape parameter and gAcc Test is the scale parameter. The determination of gAcc Test is pretty straightforward since it is the time to failure value for which 63.2% of the irradiated devices exhibited a failure, whatever the value of b. The graphical extraction gives gAcc Test = 30 s and b = 1. Those parameters allow to describe the cumulative failure rate under beam. The reliability of the device under natural environment can then be easily extracted from the accelerated test. Based on the assumption that the failure mechanisms related to neutron interaction are the same under beam than
Table 1 Failure rate under beam report.
Fig. 7. Atmospheric neutron differential flux at ground level versus neutron energy obtained with QinetiQ Atmospheric Radiation Model (QARM).
IGBT #
1
2
3
4
5
6
7
TTF (s) IGBT # TTF (s)
14 8 14
12 9 12
64 10 64
7 11 7
30 12 30
106 13 106
34 – –
128
A.D. Touboul et al. / Microelectronics Reliability 52 (2012) 124–129
a significant global amount of systems, the reliability margins will have to be tightened. In particular, dedicated test standards for ground level applications will have to be developed. Depending on the application, the derating of the device bias has to be applied during test. By the way, the FIT value might be reduced. Indeed, regarding the expected number of devices embedded in automotive or avionic applications, a specific standard closer to the real operating mode of the device, being less conservative and allowing to reduce reliability margin will have to be brought up. Fig. 9. Weibull distribution of the cumulative failure probability versus time to failure under neutron beam.
5.2. On the assessment of the reliability of the system The extracted shape parameter equal to 1, means that only random failures are considered. This is in agreement with the physical mechanisms of the failure under accelerated constraint, such as the high monoenergetic neutron flux that we used. However, even though there is a good correlation between the failure distribution and the failure mechanisms that have occurred, one can be sure that this distribution does not fit the expected one in ‘‘real life’’. Indeed, the hazard function, often known as the bathtub curve is in our case (b = 1), a constant. Fig. 10. Weibull distribution of the expected cumulative failure probability versus time to failure under atmospheric neutron flux above 146 MeV.
hðtÞ ¼ in the natural atmospheric environment, we can easily extract the cumulative failure distribution of this device at ground level in the natural environment. To this end, let’s now define the expected scale parameter at ground level as follows:
dFðtÞ dt
1 FðtÞ
¼
1
g
b t b1 ¼
1
g
It means that we are only considering the stable period of the device. In particular, IGBTs are supporting during application strong electrical and thermal cycles that can impact the failure modes along life.
g ¼ gAcc Test AF And then,
t b FðtÞ ¼ 1 e ðgÞ ¼ 1 e
t g Acc Test AF
b
The cumulative failure distribution in natural environment is plotted on Fig. 10. 5. Discussion If we consider an exponential distribution with a shape parameter equal to 1, the failure rate is equal to 1/g. Then, the failure in time which is defined as the number of failure over 109 h, is 3000 per device. This result is actually addressing two fields of interest, the accuracy of accelerated standard test of power devices under beam and the assessment of the reliability of the system from the device reliability. 5.1. On the test of active power devices under beam As indicated in Section 3.2, all the existing test methods recommend testing the device to SEB/SEGR in static OFF mode. The main question to rise is ‘‘does the data extracted from this kind of test represent significant results?’’ Indeed, in the application, power devices are mainly used as switch. Very few papers have studied the impact of dynamic operating on the failure rate under radiation on power devices and none on IGBTs [39,40]. The main conclusion of the authors is that the static to dynamic ratio of destructive single events is over a factor 100. Obviously this ratio is supposed to be strongly dependant on the device technology. But it indicates as well that for some critical applications at ground level involving
6. Conclusion In the beginning of this paper we have reminded how the atmospheric environment is naturally radiative. We are indeed permanently submitted to neutrons with a flux about 30 n/cm2/ h. Even though human are immune to this flux, it is not the case for electronics. In particular IGBTs are sensitive to two catastrophic failure mechanisms (SEB and SEL) that can be triggered by a single primary neutron interaction. Looking backward to the long range experience of space actors in the field of device qualification under radiation, one can also look at the past three decades improvement of radiation testing methodology. Following those Single Event Effects testing recommendations, irradiation of IGBTs under 146 MeV neutron beam allowed us to calculate the expected failure in time at ground level. The obtained value, obtained while considering only the high energy part of the neutron spectrum, indicates that even modern IGBTs might be susceptible to destructive single events at ground level. Indeed, the IGBT whose results have been reported here are from a trench gate fieldstop technology that is relevant for modern applications. In particular, considering the large number of energy converters expected to be embedded at ground level in the future makes it important to assess more precisely the impact of the atmospheric environment on the system reliability. This will allow to determine whether the natural radiative constraint get into the reliability margin or not, that means if the IGBT sensitivity to natural atmospheric neutrons will really become a reliability concern. At present, strong efforts have to be done in order to adapt existing test methods for space applications to ground level systems, in order to test devices closer to their real operating mode. As a consequence, such a kind of standard will be less conservative and will allow to reduce reliability margin.
A.D. Touboul et al. / Microelectronics Reliability 52 (2012) 124–129
References [1] Leray J. Effects of atmospheric neutrons on devices, at sea level and in avionics embedded systems. Microelectron Reliab 2007;47:1827–35. [2] Normand E et al. Neutron-induced single event burnout in high voltage electronics. IEEE Trans Nucl Sci 2002;44:2358–66. [3] Ziegler J, Lanford W. The effect of sea level cosmic rays on electronic devices. J Appl Phys 2009;52:4305–12. [4] Ziegler J. Terrestrial cosmic ray intensities. IBM J Res Dev 2010;42:117–40. [5] Wrobel F, Saigné F. MC-ORACLE: A tool for predicting soft error rate. Comput Phys Commun 2011;182:317–21. [6] Azzopardi S et al. Assessment of the trench IGBT reliability: low temperature experimental characterization. Microelectron Reliab 2005;45:1700–5. [7] Cortes P et al. Predictive control in power electronics and drives. IEEE Trans Ind Electron 2008;55:4312–24. [8] Turpin C et al. Fault management of multicell converters. IEEE Trans Ind Electron 2002;49:988–97. [9] Akin B et al. A simple real-time fault signature monitoring tool for motor drive embedded fault diagnosis systems. IEEE Trans Ind Electron 2010;PP(99):1. [10] Karimi S et al. FPGA-based real-time power converter failure diagnosis for wind energy conversion systems. IEEE Trans Ind Electron 2008;55: 4299–308. [11] Bose B. Technology trends in microcomputer control of electrical machines. IEEE Trans Ind Electron 2002;35:160–77. [12] Bose B. Power electronics – an emerging technology. IEEE Trans Ind Electron 2002;36:403–12. [13] Wrobel T, Azarewicz J. High dose rate burnout in silicon epitaxial transistors. IEEE Trans Nucl Sci 2007;27:1411–5. [14] Wrobel T et al. Current induced avalanche in epitaxial structures. IEEE Trans Nucl Sci 2007;32:3991–5. [15] Allenspach M et al. SEGR and SEB in n-channel power MOSFETs. IEEE Trans Nucl Sci 2002;43:2927–31. [16] Dachs C et al. Simulation aided hardening of N-channel power MOSFETs to prevent single event burnout. IEEE Trans Nucl Sci 2002;42:1935–9. [17] Johnson G et al. Simulating single-event burnout of n-channel power MOSFET’s. IEEE Trans Electron Devices 2002;40:1001–8. [18] Lorfevre E et al. Heavy ion induced failures in a power IGBT. IEEE Trans Nucl Sci 2002;44:2353–7. [19] Lorfevre E et al. Failure mode of different irradiated power IGBT structures; 2002. p. 516–9. [20] Lorfevre E et al. Cell design modifications to harden an N-channel power IGBT against single event latchup. IEEE Trans Nucl Sci 2002;46:1410–4.
129
[21] Lorfevre E et al. SEB occurrence in a VIP: influence of the epi-substrate junction; 2002. p. 557–60. [22] Baliga B et al. The insulated gate transistor: a new three-terminal MOScontrolled bipolar power device. IEEE Trans Electron Devices 2005;31: 821–8. [23] Findeisen C et al. Extrapolation of cosmic ray induced failures from test to field conditions for IGBT modules. Microelectron Reliab 1998;38:1335–9. [24] Kabza H et al. Cosmic radiation as a cause for power device failure and possible countermeasures; 2002. p. 9–12. [25] Zeller H. Cosmic ray induced failures in high power semiconductor devices. Solid State Electron 1995;38:2041–6. [26] Bauer J et al. New IGBT development for traction drive and wind power; 2010. p. 768–72. [27] Nishida S et al. Cosmic ray ruggedness of IGBTs for hybrid vehicles; 2010. p. 129–32. [28] Nishida S et al. Cosmic ray ruggedness of power semiconductor devices for hybrid vehicles. IEEJ Trans Electron Inf Syst 2010;130:934–8. [29] Shoji T et al. Neutron induced single-event burnout of IGBT; 2010. p. 142–8. [30] Kirk Jr C. A theory of transistor cutoff frequency falloff at high current densities. IRE Trans Electron Dev 2005;9:164–74. [31] Sze S, Ng K. Physics of semiconductor devices. Wiley-Blackwell; 2007. [32] Xiong Y et al. Prognostic and warning system for power-electronic modules in electric, hybrid electric, and fuel-cell vehicles. IEEE Trans Ind Electron 2008;55:2268–76. [33] Scheuermann U. Reliability challenges of automotive power electronics. Microelectron Reliab 2009;49:1319–25. [34] ASTM F1192-00. In: Standard guide for the measurement of Single Event Phenomena (SEP) induced by heavy ion irradiation of semiconductor devices, editor. vol. F1192. 2006. [35] Standard EJ. JESD57. In: Test procedures for the measurement of single-event effects in semiconductor devices from heavy ion irradiation, editor. 1996. [36] Defense DO. MIL-STD-750E test methods for semiconductor devices. In: Method 1080 single-event burnout and single-event gate rupture ed; 2006. [37] Abernethy R. The new Weibull handbook. Gulf Publication; 1996. [38] Murthy D et al. Weibull models. Wiley-Interscience; 2004. [39] Calvel P et al. Comparison of experimental measurements of power MOSFET SEBs in dynamic and static modes. IEEE Trans Nucl Sci 2002;38: 1310–4. [40] Stassinopoulos E et al. Charge generation by heavy ions in power MOSFETs, burnout space predictions and dynamic SEB sensitivity. IEEE Trans Nucl Sci 2002;39:1704–11.
