Simulation of Irradiation Effects in IGBTs with ATLAS

Omar Elmazria, Jean-Pierre Charles and Bernard Lepley
Laboratoire Interfaces Composants et Microlectronique Centre Lorrain d'Optique et d'Electronique des Solides Supelec and Universite de Metz


Designs for Insulated Gate Bipolar Transistors (IGBTs) are typically optimized for a particular range of operating frequencies. The process generally used to improve the switching speed is the exposition of IGBTs to radiation doses. In this article we describe an approach to simulate the effects of exposure to 3 MeV electrons [1] in IGBTs with the ATLAS device simulator [2]. ATLAS includes complete physics required to perform detailed analysis of the behavior of various IGBT designs. This simulation explains variations observed in the IGBT characteristics after radiation. The simulated results are in excellent agreement with measurement for these devices.


Device Architecture

The impurity profile of the simulated IGBT [3] is given in Figure 1. The initial structure doping (in cm-3) is: n++=9.31019, P=2.7.1017 N-=1.5.1014, N+=1017, P+=1019. The unirradiated IGBT is taken as a reference: IGBT (reference). The radiation exposure is simulated here as a uniform introduction of negative carriers throughout all device layers. Two radiation doses are simulated:

Radiation 1: 1.5.1015e-/cm-3 injected
Radiation 2: 2:1.2.1017e-/cm-1injected

These densities correspond to the three IGBTs tested.


Figure 1. Cross section showing donor and
acceptor concentration. The shallow n++ region
is present at the surface.



DC Characteristics

The Iak=f(Vgk) characteristics simulated with ATLAS are shown in Figure 2 for the three IGBTs. The first dose shifts the curve towards more positive Vg (the threshold voltage value Vth is increased) while the second dose shifts it towards negative Vg values. It can also be observed in the figure that the slope decreases when the radiation dose increases [1]. The slope variations express channel mobility variations.


Figure 2. The threshold voltage and slope of the
lak vs. Vg are shifted by the irradation process.


The epitaxial n-layer is the most seriously damaged layer for the first dose. This n-doping increment has the same effect that a positive substrate bias would have in a MOSFET. As a result, the curve is shifted towards higher Vg values and the mobility decreases. For the second dose all the layers are affected by the irradiation. As the n-layer concentration increment shifts the curve towards positive Vg values, the P layer concentration decrease shifts it the other way. In this case the P layer influence is more important, so the curve is shifted towards lower Vg values. The slope, related to the mobility, varies in the same way as for the first dose but in larger proportion. The electron concentrations in the n-and n++wells increase and produce an increment in the electron density in the channel, hence the mobility decrease.


Switching Characteristics and Latching

The biggest limitation to the turn-off speed of an IGBT is the lifetime of minority carriers in the N epitaxial layer base of the PNP transistor. Since this base is not accessible, external drive circuitry cannot be used to improve the switching time. Hence the need for another means of lifetime reduction: the exposure of IGBTs to radiation doses to produce a reduction of the minority carrier lifetime.

The current densities are different for these IGBTs, and the comparison of the switching times is not easy. Therefore we have simulated the switching waveforms for a value of Vg where the current densities are approximately equal (Figure 3). This shows clearly that increasing the dose improves transistor speed.


Figure 3. IGBT switching behavior with
simliar current shows the improved
speed of the irradated devices.


As shown in the cross-section of Figure 1, the IGBT is made of four alternate P-N-P-N layers. Its equivalent circuit [4] is given in Figure 4, where Rp is the P-layer resistance and the NPN is a parasitic bipolar transistor. The dynamic latching could occur when a high density of hole current (collector current) flows in Rp, making the bias of the parasitic NPN higher than the threshold value (~ 0.6 V).



Figure 4. Electrical equivalent circuit of IGBT s
howing the parasitic thyristor structure.




The start of the latch-up can be observed in Figure 5 for different IGBTs with the same Vg (Vg=20V). The anode current value for which the latch-up begins is higher for IGBT (reference) than the IGBT (radiation2) because the P layer doping is higher for IGBT (reference) and its Rp resistance is lower. The latch-up threshold current remains the same when the value of Vg changes (IGBT (radiation 2) with Vg=16V and 20V).


Figure 5. Simulated start up of latch-up
for the three thyristors.



The simulated complete characteristics after latch-up is shown in Figure 6. A classical thyristor characteristic is seen. The simulation provides the characteristics beyond the safe operating area of the IGBT and shows the full breakdown and snapback behavior.


Figure 6. Complete simulated
latch-up characteristics




The effects of radiation in IGBTs were simulated with ATLAS. The changes in current-voltage characteristics and switching time values are in good agreement with measured data. ATLAS provides a complete simulation of the associated improvement in switching time characteristics. This change is related to a degradation due to a decrease of the latch-up threshold current.




[1] B. J. BALIGA
"Switching Speed Enhancement in Insulated Gate Transistors by Electron Irradiation",
IEEE Trans ED, Vol ED-31, No 12, Dec 84.

[2] SILVACO International,
ATLAS User's Manual.

[3] N. Iwamuro and al
"Numerical Analysis of Short-Circuit Safe Operating Area for p-Channel and n-Channel IGBT's",
IEEE Trans ED, Vol 38, No 2, February 1991.

[4] P. ALOISI & Motorola Semiconducteurs
"Un Composant de Puissance Dcouvrir (L'IGBT et ses ALTERNATIVES)",
Seminaire Technique du 22 Septembre, 1989, SEE-Club 13.