BACKGROUND OF THE INVENTION
This application claims the benefit of U.S. Provisional Application No. 60/235,533, filed Sep. 26, 2000.
This invention relates to MOSgated devices and more specifically relates to a novel structure for a MOSgated device with a reduced gate-to-drain capacitance.
- BRIEF SUMMARY OF THE INVENTION
The periodic charging/discharging of the parasitic gate-drain (Miller) capacitance of a MOSFET during each switching cycle is known to increase the power dissipated within the conventional trench-gate MOSFET structures. This decreases the efficiency of the power system and limits the operating frequency of the MOSFETs. Furthermore, the gate-drain capacitance makes the MOSFET susceptible to spurious turn-on, induced by a rapidly changing drain voltage (dv/dt induced turn-on). Sometimes, this leads to the destructive failure of the MOSFET. The novel reverse source-drain FET structures of this invention significantly reduces the gate-drain capacitance and susceptibility to spurious turn-on.
BRIEF DESCRIPTION OF THE DRAWINGS
A significant portion of the gate-drain capacitance (Cgd) of a conventional MOSFET is the oxide capacitance (Cox) in the gate-drain overlap area. The oxide capacitance (Cox) is inversely proportional to the oxide thickness (tox) in the gate-drain overlap region and directly proportional to the gate-drain overlap area. In accordance with the invention, the arrangement of the source and drain electrodes (with respect to the gate electrode) is reversed. The reversed source-drain FET structure significantly lowers the gate-drain capacitance by reducing the overlap area of gate and drain and by increasing the oxide thickness between the gate and the drain electrodes.
FIG. 1 is a cross-sectional diagram of one-half of a cell of a conventional trench gate MOSFET.
FIG. 2 is a cross-section like that of FIG. 1, but for a prior art depletion mode MOSFET with an accumulation channel (hereinafter an ACCUFET).
FIG. 3 is a cross-section like that of FIGS. 1 and 2, for a conventional MOSFET adapted for a-c operation, hereinafter an INVFET.
FIG. 4 is a cross-section like that of FIGS. 1, 2 and 3, for a reversed source-drain ACCUFET of the invention.
DETAILED DESCRIPTION OF THE DRAWINGS
FIG. 5 is a cross-section like that of FIGS. 1 to 4 for a reversed source-drain INVFET of the invention.
The unit half-cell cross-sections of the conventional MOSFET, ACCUFET and the INVFET structures are shown in FIGS. 1, 2 and 3 respectively. In all of the conventional FET structures of FIGS. 1, 2 and 3, a non epi monocrystelline substrate 5 of any desired thickness has an epitaxially grown layer 12 grown thereon which consists of an N− drift region 6 (FIGS. 1 and 2) a P− base region 7 (FIGS. 1 and 3) and N+ source region 8. A drain electrode 10 is formed at the bottom surface of region 5 and is thus connected to the silicon epitaxially grown body 12 while the source electrode 11 is formed at the top surface of body 12. A trench 20 is formed in the silicon 12 and is lined with gate oxide 13. A conductive polyslicon gate 21 fills the trench 20 and defines an invertible channel region along its vertical height and between source region 8 and along the P− base 7 or N− drift region 6. The thickness of the gate oxide 13 in the gate-drain overlap area (at the bottom of trench 20) is tox,g which is generally less than about 500 Å. Since, tox,g is an order of magnitude smaller than the top isolation oxide layer 23(conventionally about 4000 Å), the capacitance portion of Cgd of the conventional MOSFET structures of FIGS. 1, 2 and 3 is an order of magnitude larger than that of the reverse source-drain MOSFETs of FIGS. 4 and 5 of the invention as will be seen. Consequently, the switching losses of the conventional FETs of FIGS. 1, 2 and 3 are higher than the MOSFET structures of FIGS. 4 and 5. Furthermore, due to a higher gate-drain capacitance, the convention FETs of FIGS. 1, 2 and 3 are more susceptible to the dv/dt induced turn-on.
Unit half-cell cross-sections of the proposed reverse source-drain ACCUFET and INVFET of the invention are shown in FIGS. 4 and 5, respectively. In these trench-gate MOSFET structures, similar numerals designate similar components. The drain electrode 10, however, is formed at the top surface of body 12 while the source electrode 11 is formed on its bottom surface. The thick (>4000 Å) oxide 23 over the gate polysilicon 21 within the trenches 20 now isolates the gate electrode 21 from the planar drain electrode 10. Thus, as seen in FIGS. 4 and 5, the drain electrode 10 overlaps the gate electrode 21 in areas with varying oxide thickness. The very short vertical overlap between the N+ drain diffusion 30 and the gate electrode 21 determines the overlap area with an underlying oxide thickness of tox,g while the horizontal overlap of the drain and gate electrodes 10 and 21 respectively in the trench region determines the overlap area with an underlying oxide thickness of tox,iso. In the rest of the overlap area, the underlying oxide thickness is in between tox,g and tox,iso.
The design considerations, the operational physics and the static performance of the proposed reversed source-drain MOSFET structures are identical to those of the conventional ACCUFET and INVFET structures. However, the switching performance (turn-on and turn-off times) of the proposed FET structures of FIGS. 4 and 5 are significantly superior to the conventional ACCUFET and INVFET structures. This is because the drain voltage fall time (during turn-on) and voltage rise time (during turn-off) are directly proportional to the gate-drain capacitance (Cgd). Consequently, the voltage rise time and fall times of the reverse source-drain FET structures are smaller than those of the conventional ACCUFET and INVFET structure.
The device of FIG. 4 is a depletion mode device and, in normal operation, the source 11 is at ground potential and the drain 10 is at a positive potential. The device turns off when gate 21 is shorted to the source 11. Thus, the work function difference between the P− type gate 21 and N− drift region 6, gate 21 then cause the N− drift region 6 to deplete and provide blocking voltage.
The device of FIG. 5 is an N channel device but will act as an a-c device in view of its NPN junction pattern.
It should be noted that both of the structures of FIGS. 4 and 5 will act as an a-c device in view of the symmetric structure N+/N−/N+ in FIG. 4 and N+/P/N+in FIG. 5.
Further, both structures of FIGS. 4 and 5 could have a trench bottom oxide thicker than the gate oxide for further reduction of the Qg (gate charge) thru a reduction in G-S capacitance. This is shown by the dotted line boundary 40 in FIGS. 4 and 5 for a thicker tox,g.
Although the present invention has been described in relation to particular embodiments thereof, many other variations and modifications and other uses will become apparent to those skilled in the art. It is preferred, therefore, that the present invention be limited not by the specific disclosure herein, but only by the appended claims.