Possibilities and limits of axial lifetime control by radiation induced centers in fast recovery diodes

Device simulation, based on an extended recombination model, is used as a design tool for lifetime-controlled power diodes with different lifetime profiles. Homogenous and local recombination center profiles are considered. The sensitivity of important device properties, such as the trade-off between stationary and dynamical characteristics, to the recombination center peak position is investigated. The occurrence of dynamic impatt oscillations is analyzed. INTRODUCTION The use of irradiation techniques for carrier lifetime control is nowadays a commonly accepted strategy for optimizing power device characteristics. In comparison with the conventionally used impurities gold and platinum, the irradiation techniques offer an exact process control and the possibility to realize different lifetime profiles. Today, irradiationbased lifetime adjustment steps are applied to a wide variety of power devices like IGBTs, GTOs and Freewheeling Diodes (FWDs). The improvement and optimization of radiated devices was usually done through timeand cost-consuming experiments. Using a device simulation tool with an appropriate extended recombination model allows a significant reduction of the necessary experimental efforts. Based on center parameters determined in earlier work [13,14], the effects of different lifetime profiles on the stationary and dynamical characteristics of fast recovery diodes are studied in this work using 2D device simulation. The validity of the results is demonstrated by comparison with measurements taken on manufactured samples. LIFETIME CONTROL In this work, device simulation is used for the investigation of the influence of various lifetime profiles on the properties of fast silicon power diodes with a blocking voltage of 1.2kV. Table 1 gives an overview of the studied basic profile types. Additionally, the irradiation dose was varied in a dose range as given in table 1. All manufactured samples were annealed using identical conditions at a temperature of more than 300°C for one hour. The nominal current density of the devices is app. 170A/cm. SIMULATION MODEL Irradiation generates centers with different energy levels in the band gap of Silicon. Each level may act as an effective recombination center where the total recombination rate results from the emission and capture processes of each single level as illustrated in Figure 1. The implementation of this extended model, which includes the complete trap dynamics, is fundamental for an appropriate simulation of such devices [4,21]. For all simulations, the 2D device simulator TeSCA has been used [3]. This simulation system solves the three fundamental semiconductor equations (the Poisson equation as well as the electron and hole current continuity equation). For the consideration of deep traps, additional terms are necessary. In the Poisson equation (1), the charged recombination centers are considered. The thermal capture and emission processes of carriers via the deep levels within the band gap lead to additional recombination terms in the continuity equations (2) and (3). Here, the terms R and G refer to further recombination and generation mechanisms, such as auger recombination or avalanche. The occupancies of the acceptor and donor traps are evaluated from the balance equations (4) and (5) according to the relations of equations (6) and (7). The emission rates are calculated from the position of the recombination center within the band gap, the capture rates, the entropy factors and the temperature, as given in equations (8) to (11). Based on this extended recombination model, the behavior of radiated devices is predicted with qualitatively and quantitatively good results [12,13,14]. RECOMBINATION CENTER DATA Recombination Center Properties For simulation purposes it is necessary not only to implement an appropriate recombination model but also to know the parameters of the radiation-induced centers as well as their temperature dependencies. Even though a lot of publications deal with recombination center parameter determination [1,5-7,18,22], reliable data were not available until recently due to the sophisticated and therefore fault sensitive character of the necessary measurements. Table 2 shows the properties of the recombination-relevant centers, as used in the simulations, according to previous work [13,14]. There, the fundamental properties of the radiationinduced centers are determined by DLTS measurements [8]. Figure 2 shows, as an example, the majority and minority carrier DLTS spectrum measured at type E100. Due to the applied annealing step, the center E(90K) controls the high-level lifetime. Figure 3 gives the change of the calculated carrier lifetime under high-injection condition with the annealing of the acceptor-like radiation-induced centers. Power devices are usually operated at high injection levels in on-state and turn-on/turn-off. Therefore, it is necessary to know the parameters of E(90K) exactly to allow correct simulations. Since the electron capture rate is small compared to the hole capture rate of E(90K), it is possible to use measurements of the high-level lifetime for an estimation of the temperature-dependent electron capture rate [14]. These measurements are based on the well-known OCVD (Open Circuit Voltage Decay) technique [9]. Due to the comparatively shallow energetic position of E(90K) within the band gap of silicon, optical excitation of carriers by means of laser light pulses was used to generate a large density of excess carriers to fulfill the high-injection condition [14]. The composition of the recombination centers detected in electron-radiated silicon depends on the irradiation parameters. Figure 4 shows the dependence of the trap density on electron energy and dose. In case of the lowest irradiation energy, E(230K) is not detectable. This recombination center is already vanished due to the annealing process. Under the applied conditions, the dependence of the generated centers is approximately linear to the irradiation dose. Recombination Center Profile Estimation DLTS measurements were also used for the determination of the concentration profiles in the helium-radiated samples. Due to the high concentrations, an additional second annealing step was necessary to allow the profile measurement of E(230K) to be performed. Figure 5 shows the determined center distribution and the approximation based on a simple gaussian profile as used in the device simulations. For the traps H(195K) and E(90K), the same profile is assumed. The peak concentrations of these traps are estimated from the comparison of the results of lifetime measurements and measurements of the DLTSand junction capacitance at the helium-radiated samples with and without an additional annealing step. Further work must improve the possibilities and the accuracy of recombination center profile measurement. THE INFLUENCE OF LIFETIME CONTROL ON DEVICE PROPERTIES Forward Characteristics Figure 6 shows the forward characteristics of the different types E (electron irradiation), H (helium irradiation) and EH (combination of the electron irradiation of type E and the helium irradiation of type H) as a result of simulation and measurement. Beside a satisfying agreement between measurement and simulation results, figure 6 clearly indicates the influence of the different lifetime profiles on the forward voltage characteristics. Figure 7 shows the impact of the temperature on the forward voltage at nominal current. This temperature dependency is strongly influenced by the properties of the recombination centers due to the temperature dependent capture rates of the dominant recombination center E(90K). The comparison of the measured and simulated characteristics of type E shows a good accordance. The deviations in case of the types H and EH are most probably caused by the uncertainties of the recombination center profile measurements. Due to the higher forward voltage drop in type EH, other effects, like self-heating caused by recombination heat, may cause an additional error. Nevertheless, figure 7 shows the change in the temperature coefficient due to the different irradiation types. Therefore, lifetime control offers a chance to slightly tune the temperature coefficient. This is important since a positive temperature coefficient simplifies the paralleling of power devices. In difference to results recently published by other authors, as in [20], no further adjustment of recombination center parameters, especially the capture rates of the dominant center E(90K), was needed due to the sophisticated measurement techniques used [14]. Turn-Off Characteristics Figure 8 gives a comparison of the reverse recovery current peak IRRM for the different types E, H and EH. Figure 9 shows the comparison of the stored charge QRR. As in case of the forward voltage dependencies, these figures give clear evidence to the effects of the different irradiation processes. Obviously, type EH offers the best properties with respect to a low reverse recovery current peak IRRM and the lowest stored charge QRR. In case of type H deviations between measurement and simulation are caused by a larger tail current in the measurement compared to the simulation, and by non-calibrating the simulation parameters of the avalanche generation model. Under common operating conditions, a sinusoidal current lower than the nominal current is often switched in usual topologies. Due to the reduced number of stored carriers in the lowdoped region of the freewheeling diode, this is a critical condition for the device. As on example of type E, shown in figure 10, low current may cause a snap-off in the reverse current. This leads to overvoltages and/or oscillations due to parasitic inductances. The use of device simulation offers an opportunity for an optimization of the device design to avoid such a behavior. The Influence of Recombination Center Peak Position As commonly known, the combination of local and homogenous lifetime control is one possibility to realize fast FWDs with soft recovery behavior and a high dynamical ruggedness [10]. If local lifetime adjustment is applied, the peak position xp-xj of the recombination center profile, as illustrated in figure 11, controls the trade-off between forward losses and the stored charge as well as the trade-off between forward losses and reverse recovery current maximum. In this investigation, the peak position of a constant recombination center profile was moved along the vertical axis, while homogenous base lifetime has been reduced as in case of an applied electron irradiation. Figure 12 shows the trade-off between forward voltage drop and the reverse recovery current maximum while figure 13 shows the trade-off between forward losses and the stored charge, both in dependence of the recombination center peak position, as a result of device simulation. A minimum is found at a recombination center peak position close to the pn-junction for both dependencies. Dynamic Effects Furthermore, device simulation holds potential to avoid disturbing dynamic effects. As an example, figure 14 shows the reverse recovery measurement of type E45, where impatt (impact ionization transit-time) oscillations appear. The measurement was done using a conventional double-pulse method. The oscillations are caused by the temporarily positively-charged donors H(195K) which enhance the effective doping and therefore reduce the reverse blocking capability. Consequently, avalanche breakdown occurs at the pn-junction region and generates electrons. These electrons counterbalance the positive donors and hence stop the avalanche generation of carriers. Due to the electric field, the electrons are transported to the nn-junction and again, avalanche generation starts at the pn-junction. The impatt oscillations stop as soon as the positive donors are discharged and the device is again able to withstand the reverse voltage [11]. The frequency of the oscillations depends on the transit-time of the electron carrier flow through the low-doped region of the device. Thus, the oscillation frequency is defined by the carrier saturation velocity vd, depending on the strength of the electric field, and the width of the low-doped region wB (equation 12). For simulation purposes, we use an emulation of the measurement setup. The simulation circuit consists of the discrete freewheeling diode, a time-variable serial resistance instead of the IGBT and a small inductance. This emulation was used to decrease the necessary computing time since only the diode has to be considered. Figure 15 shows a simulation of type E45 (electron dose d= 2 15 cm 10 1 − ⋅ ) where impatt oscillations are observed. Figure 16 shows the electron carrier distribution at different points in time as a result of device simulation, while figure 17 shows the change of the electrical field with time to exemplify the physical processes in the device. The avoidance of these high-frequency oscillations is necessary because of their adverse influence on drive control units and because of EMC (electromagnetic compatibility) issues. According to previous work [11], the threshold voltage of the impatt oscillation mainly depends on the concentration of the donor-state H(195K), the reverse voltage and the temperature. Figure 18 shows the threshold voltage VDI of type E45 in dependence of temperature for different irradiation doses. The measured values from [11] are compared with the simulation results. The agreement is sufficient. Additionally, an analytical estimation arrives from the discharging of the centers H(195K) due to the thermal emission of previously captured holes as described in detail in [11]. Figure 18 also includes the results of this estimation, based on an abrupt pn-junction, the ionization coefficients published in [2,16] with the temperature dependency given in [17] and a triangular-shaped electrical field (as shown in figure 17). Obviously, the temperature dependency of the threshold voltage of impatt oscillation can be predicted using simulation. Thus, Figure 18 indicates the potential of device simulation if deep centers are considered. CONCLUSION To consider lifetime killing effects in device simulation, the use of an extended recombination model including full trap dynamics is necessary. The introduction of several recombination centers with different properties into simulation allows the correct description of recombination processes under different conditions. The previously determined center parameters, which were used for the simulations in this work, explain the temperature dependencies of stationary and dynamic characteristics. Based on these results, device simulation is used as a tool for device design. The influence of different recombination center profiles on the stationary and dynamic properties of freewheeling diodes is studied. It is shown that for the realization of a fast and soft freewheeling diode, the optimal position of the recombination center peak is located close to the pn-junction which is in agreement with previously published results [10,15,19]. Furthermore, the physically correct description of the trap dynamics allows an appropriate simulation of dynamic impatt oscillations. This effect is caused by temporarily positivelycharged donor-states which reduce the blocking capability of the device. The consideration of the donor-states is necessary in case of high-dose electron, proton or helium ion irradiation to prevent high-frequency oscillations. Consequently, the use of high-energy particles for carrier lifetime control is limited due to the formation of these unavoidable defects. Therefore, device simulation may be used as a powerful tool in the development and optimization of power devices as well as in the explanation of their behavior. ACKNOWLEDGEMENTS The authors wish to thank the scientists, especially Dr. Nürnberg and Prof. Gajewski, from the Weierstrass Institute for Applied Analysis and Stochastics in Berlin, who developed the device simulator TeSCA. Moreover, the authors are grateful for their support and the addition of new features and algorithms into the simulation system. The Authors would like to thank also Ms. Pellkofer and Mr. Umland for their support in sample preparation and measurement assistance. We also wish to thank Prof. Wagemann from the Technical University of Berlin for the opportunity to perform DLTS measurements. This work has been supported by grants of the Deutsche Forschungsgemeinschaft.