2.2.4 Double Gate Lateral Inversion Layer Emitter Transistor (DGLILET)

LDMOSFETs have a high switching speed and low switching losses but a relatively low current density in the on-state. On the other hand, IGBTs offer high current densities because of conductivity modulation in the drift region, but have poor switching performance compared to LDMOSFETs. Excess carriers in the drift region generate a tail current in the off-state, which can be removed only by recombination processes. By introducing extra $n^+$-doping in the anode (anode-shorted structure) of the LIGBT the switching performance can be improved drastically, but the anode-shorted LIGBT has an inherent snapback characteristic which is not attractive from a circuit design point of view.

The dual gate lateral inversion layer emitter transistor (DGLILET) is a device that offers controlled carrier injection and fast switching performance. Figure 2.15 shows the cross section of the device. The combination of a shorted-anode structure and an extra gate at the anode gives high switching speed while suppressing the undesirable snapback characteristics. The cathode gate performs the role of the main control gate like in the case of LIGBTs and the anode gate controls carrier injection.

The anode gate which is driven negative with respect to the anode, forms an inversion layer of holes. The flow of electron current underneath the hole inversion layer initiates minority carrier injection driving the device into conductivity modulation. Having the inversion layer in the high resistance $n$-drift region initiates the injection at a very low current density, suppressing the snapback. Moreover, the anode gate acts as a field plate improving the voltage blocking capability of the device.

Full control over hole injection into the device drift region can be achieved by minimizing the length of the $p^+$-anode diffusion. This prevents the potential drop due to lateral electron current flow beneath the $p^+$-diffusion becoming large enough to forward bias the $p^+$-diffusion and $n$-well junction at normal operating current densities. This ensures that injection is taking place only at the hole inversion layer, and thus the anode gate has full control of injection. The device can therefore be switched between unipolar and bipolar modes of conduction. Switching the anode gate off slightly before the actual turn off of the device suppresses injection and improves the turn-off speed of the device. On the other hand, the $p^+$-diffusion can be designed to allow injection of holes into the $n$-well at a higher current density. The injection of holes is initially established by the hole inversion layer and the injection from the $p^+$-diffusion is commenced at a latter stage conserving the snapback suppression. In this design turning off the anode gate before cathode gate might not suppress the injection depending on the operating current density. The on-state is improved due to a higher level of modulation but the turn-off time will be slightly increased.

Figure 2.15: Cross section of the DGLILET.