Simulation of Ion Irradiated Power Devices in ATLAS

J. Vobecky and P. Hazdra
Czech Technical University in Prague, Department of Microelectronics, Czech Republic

Introduction

About ten years of evolution was sufficient for ion irradiation technology to become a widely used tool for local carrier lifetime control in power devices. In 1994, the simulation of devices taking into account both the real defect profile resulting from ion irradiation and multi-level Shockley-Read-Hall model was published for the first time [1]. ATLAS has allowed simulation of transient traps since 1995 [2]. The last version of ATLAS (4.0) brought the possibility to account for arbitrary defect spatial distribution. So the development of ion irradiated devices using device simulation is now possible [3]. At present, any application of the simulator requires just a knowledge of the spatial distribution of the defects resulting from irradiation and electrical parameters of the related deep levels [4]. The practical application of this will be presented below for the case of power diode.

Background

Deep levels generated by ion irradiation affect the free-carrier thermal generation-recombination and hence the excess carrier lifetime. The thermal capture and emission of carriers through deep levels located within the bandgap is described in ATLAS by analytical model based on SRH statistics. In case of k independent single-charged acceptor- or donor-like levels the thermal components of the recombination rates Rn and Rp for electrons and holes, are respectively [1-3]:

where Nti is the concentration of the i-th deep level and Gn(p) and Kn(p) are the i-th level emission and capture rates for electrons (holes). The electron occupancy fi of the i-th deep level is calculated from the following balance equation

The charge of traps Dt=(pt-nt) influences the right-hand side of the Poisson equation

Application of this model requires a detail knowledge of deep level parameters. The ATLAS command is

doping trap ascii inf= acceptor/donor \ e.level=... sign= sigp= degen=...

ATLAS also incorporates a model for transient trapping and de-trapping of carriers. For dynamic equilibrium dfi/dt=0 (DC analysis) CPU time is saved because the recombination rate R is unique and may be expressed explicitly by one equation in the following way

Transition between the two states is automatic in ATLAS. There is available a second SRH model (models srh) which uses the static approximation even for transient analysis. In this case only a single G-R centre is considered and the equation above leads to the fairly well known formula

where k is Boltzmann constant, T is the temperature, ni is intrinsic concentration, Ei intrinsic Fermi-level, Et trap level, tn(p)0=1/(sn(p).vn(p).Nt) are electron and hole lifetimes, resp. It is worth reminding that this equation is true only for a single ideal G-R center (Rn=Rp) or dynamic equilibrium, (e.g. ON- or OFF-state).

Defining Trap Parameters

The device under consideration is a p+pnn+ power diode (2.5kV/100A) with length of 370µm between anode and cathode. The detailed device data may be found in [5]. In order to present the simulation capabilities of ATLAS the device under test was virtually irradiated by 10 and 18 MeV 4He2+ ions with the dose of 5x109 cm-2 using the calibrated system for determination of defect distribution [1]. The parameters of deep levels created by penetrating ions were determined by means of the deep level transient spectroscopy (DLTS) and are summarized in Table 1 using the notation of ATLAS. Helium irradiation produces pure damage defects comprising five deep levels within the bandgap which are connected with different charge states of divacancy (E2, E3, H1), acceptor level of vacancy-oxygen VO pair (E1), and donor level of the carbon-vacancy-oxygen CVO complex (H2). The data received from measurements for level positions Et (E.LEVEL) and electron capture cross-sections sign were completed by capture cross-sections for holes sigp presented in reference [4] for the same type of defects.

The influence of individual deep levels on electron lifetime is shown in Figure 1 for both the defect peak (x~180µm) and defect tail (50µm<x<150µm) regions of the 18 MeV irradiation (see Figure 2) using the general lifetime dependence on excess carrier concentration that reads


The thermal component of Rn was calculated from the first equation above for deep level parameters given in Table 1. Figure 1 enables one to compare the lifetime reduction in the defect peak with both the unirradiated region and tail part. Furthermore, t(Dn) as a result of individual and all deep levels implies that only two levels are dominant.

 

Figure 1. Electron lifetime at defect peak (top) and
tail (bottom) vs. injection level (Dn=Dp, static approx.,
T = 300 K, N.I. = no irradiation, ALL=all levels accounted for)

Table 1. Deep levels in FZ n-type silicon irradiated by 4He2+ ions.

The level E1 has the biggest impact on the lifetime decrease with increasing excess carrier concentration above 1014 cm-3 and determines the so-called high-level lifetime. The level E3 is counteracting, so it dominates in decreasing the lifetime below 1014 cm-3 . This is usually referred as a low-level lifetime. For the device under consideration, the E1 level is responsible not only for the magnitude of the DC forward voltage drop (n > 1015 cm-3), but also for the excess carrier recombination within the neutral n-base during the initial part of the turn-off. E3 brings mainly the desirable decrease of charge at the far end of reverse recovery (n > 1015 cm-3). Finally, the figure implies the fact that simulation with the two dominant levels (E1 and E3) gives the same results as with five ones (verified in simulations of reverse recovery). Since the influence of both the double-acceptor (E2) and single donor (H1) levels of divacancy is marginal, the defect can be approximated as a single acceptor E3. Therefore, a problem with inclusion of multiple-charged centers, which are not covered by the current ATLAS SRH model, is avoided.

Device Simulation

Figure 2 shows the excess carrier distribution n + p of both the unirradiated and helium irradiated devices during the ON-state (100A@300K). Figure 3 shows the reverse recovery current and voltage waveforms to be simulated for dc reverse voltage -1000V and dI/dt=-1000A/ms starting from the conditions of Figure 2. The overall behavior of irradiated devices is influenced by position of the defect peak (ion energies) that was intentionally located in two places with different impact on device parameters. The forward voltage drop VF is 0.94, 0.98, and 1.032V @100A for unirradiated devices, 10MeV and 18MeV (dose: 5x109cm-2) irradiations, respectively. The gradual increase of VF with defect peak distance from the anode is in agreement with experiment [5]. The influence of ion irradiation on dynamic behavior is more pronounced. While the unirradiated device shows oscillators, the 18MeV one is even worse. On the other hand, using 10MeV the defect peak placed within the n-base close to the anode softens the diode recovery in agreement with experiment [5]. As a result the removal of the oscillatory behavior takes place.

Figure 2. The profile of VO centers generated by 10 and 18 MeV He2+ 5x109 cm-2 irradiation and
ON-state sums of carrier distributions n+p for unirradiated and irradiated diode (If = 100A).

Figure 3. Current and voltage waveforms of the reverse recovery process (VRM= -1000V, dI/dt= -1000A/ms) for unirradiated and He irradiated diodes (10 and 18MeV@5x109cm-2)

 

Conclusions

It was shown that ATLAS is capable of accurate simulation of ion irradiated power devices. The user should provide the electrical parameters of relevant deep levels and define trap models accordingly in the ATLAS syntax.

References

  1. Hazdra P., Vobecky J.,
    "Accurate Simulation of Fast Ion Irradiated Power Devices",
  2. ATLAS User's Manual, SILVACO International, June 1995
  3. ATLAS User's Manual, SILVACO International, October 1996
  4. Hallén A., Keskitalo N., Masszi F., Nágl V.,
    "Evaluation of Local Lifetime in Proton Irradiated Silicon",
    J. Appl. Phys., Vol.79, pp.3906 - 3914, 1996
  5. J. Vobecky, P. Hazdra, and J. Homola,
    "Optimization of Power Diode by Means of Ion Irradiation",
    IEEE Trans. on Electron Devices, Vol. 43, pp.2283 - 2289, 1996