LOW VOLTAGE SHORT CHANNEL TRENCH DMOS TRANSISTOR
FIELD OF THE INVENTION This invention pertains to semiconductor devices and more specifically to a trenched DMOS transistor suitable for use as a power transistor and having punch-through elimination, improved safe operating area and threshold control.
DESCRIPTION OF THE PRIOR ART
DMOS transistors are well known and are especially suitable for use as power transistors. Such transistors are typically fabricated by well known semiconductor fabrication techniques. A typical power transistor includes hundreds or thousands of cells formed in a single semiconductor substrate and connected together electrically. Prior art DMOS transistors have several well known deficiencies, including having punch-through i.e. current conduction from the source region to drain region, when such is not desired. Also there is the deficiency of control of the threshold voltage. Additionally, such devices often suffer from excessive resistance of the pinched body region, which tends to cause latchback (i.e. snap back) .
SUMMARY
In accordance with the invention, a vertical DMOS field effect transistor includes (for an N-channel device) a N+ drain region overlain by a P- drift region which in turn is overlain by a P body region which is overlain by a N+ source region. A trench penetrates through the source region, body region and drift region into the drain region and is filled with a conductive polycrystalline silicon gate electrode. The trench penetrating down into the drain region is useful in
accordance with the present invention due to the need to invert the entire trench sidewall surface. A source-body contact overlies the principal surface of the silicon and is in electrical contact with the source region and is also in electrical contact with the body region via a P+ body contact formed in an upper portion of the P body region.
Advantageously this structure provides a power transistor device with a simplified punch-through elimination structure, improved safe operating area, and threshold control. The device has a short channel and a low threshold voltage for e.g. low voltage battery applications.
When the device is reverse-biased, the depletion region spreads from the N+ drain region into the P- drift region. The thickness of this P- drift region in conjunction with the thickness and concentration of the P body region determines the drain-source breakdown voltage. Diffusing the P body region into the P- drift region allows control of both the surface concentration in the channel region and the channel length, resulting in improved threshold control. For safe operating area, the effective body junction depth is the combination of the body and drift regions, resulting in lower Rb' .
When the device is forward biased, the gate voltage easily inverts the surface of the drift region. Since critical electric field has been achieved in the body (channel) region, carriers are injected into the drift region with maximum velocity which results in low resistance for this region.
A complementary P-channel device which is otherwise similar structurally has also been found to be advantageous. In another N-channel embodiment a P+ doped "body plus" region is provided which extends from the principal surface of the semiconductor material
adjacent the source region into the drift region.
BRIEF DESCRIPTION OF THE DRAWINGS
Fig. 1 shows a cross-section of a semiconductor device in accordance with the present invention. Fig. 2 shows a cross-section of a second embodiment of a semiconductor device in accordance with the present invention.
Figs. 3A-3C show process steps for forming a semiconductor device in accordance with the present invention.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT Fig. 1 shows a trenched DMOS transistor in accordance with the present invention. It iε to be understood that this shows one cell of what is typically many cells (as described above) of a transistor. Also conventionally this cross sectional depiction shows various semiconductor regions delineated by lines. It is to be understood that in an actual device there are concentration gradients between these regions. Moreover, Fig. 1, as is true of the other figures in this disclosure, is not to scale and shows only an active portion of the transistor. The surrounding termination region is discussed below. Moreover, the description of various materials, dimensions, doping levels, etc. herein is intended to be illustrative and not limiting; other materials, dimension, and doping levels may also be used in accordance with the present invention as is well known in the field.
The transistor of Fig. 1 in its lower portion includes an N+ doped drain region 20 having typically a dopant concentration of 2 x lθ19/cm3. It is to be understood that Fig. 1 depicts an N-channel device and the complementary P-channel device may also be
fabricated with all the conductivity types being opposite to that shown in Fig. 1. Such a complementary P-channel device would have comparable performance to that of the depicted N-channel device. Overlying the N+ drain region 20 is a P- drift region 22, the thickness of which partially determines the drain/source breakdown voltage of the device. For instance, for a 30 volt device the P- drift region 22 has e.g. a thickness of 2 μm; for a ten volt device a thickness of 1 μm is sufficient. P- drift region 22 is typically an epitaxial layer grown on an N+ doped substrate 20. The drift region 22 is lightly doped, typically having a resistivity of 20 ohm cm and a doping concentration of 7xl0I4/cm3. In the upper portion of the epitaxial layer is a diffused P body (channel) region 26 typically 0.6 to 1.1 μm thick and having a typical compensated surface doping concentration in the range of l to 3 x lθ16/cm3. The thickness and doping level of this body region 26 is important because its properties determine the length of the channel. Also diffused into the upper portion of the epitaxial layer is the N+ source region 30 having a typical thickness of 0.2 to 0.5 micrometer and a surface doping concentration of 5xl019/cm3. Formed also in the upper portion of the epitaxial layer and adjacent the source region 30 is a P+ body contact region 34 allowing ohmic contact to be made to the body region 26. The P+ body contact region 34 has a thickness similar to that of the source region 30 and a typical surface doping concentration of 5xlθI8/cm3.
A trench 38, typically 2 to 3 μm deep and 1 μm wide, penetrates into the drain region 20. Trench 38 is conventionally lined with a gate oxide layer 42 and filled with a doped polysilicon gate electrode 46. A layer 48 of boro-phosphosilicate glass (BPSG) overlies and insulates the upper portion of the conductive gate
electrode 46. A conventional aluminum silicon metallization layer 52 overlies the BPSG layer and electrically contacts both the source region 30 and the body contact region 34. Also conventionally present is a passivation layer (not shown) overlying the metallization layer and a drain metallization layer 56 formed on the lower portion of the substrate to electrically contact the drain region 20.
The structure shown in Fig. 1 therefore includes, arranged vertically, semiconductor regions including a source region 30, a body region 26, a drift region 22 and a drain region 20. There is known in the art a lateral DMOS with similar semiconductor regions but of course arranged laterally rather than vertically. Advantageously, the vertical device of Figure 1 saves considerable chip "real estate" (surface area) over a comparable lateral device.
Advantageously the present device allows spreading of the breakdown between the drain and the drift region and not across the channel. Thus, a shorter channel is formed than in other types of semiconductor devices. In the case of the device of Fig. 1, the channel is in the body region 26 between the source region 30 and the drift region 22. This device therefore allows use of a short channel without the problem of punch-through. As is known, in prior art trenched DMOS devices when one grows gate oxide over an N- doped drift region, surface accumulation occurs, i.e. the N- region becomes more N-type and this undesirably compensates the P-body region and shortens the channel even more, leading to punch through. The present device avoids this by providing an enhancement in the opposite direction because the P- drift region 22 depletes at the conduction surface next to the gate oxide, so that the channel experiences less effect from the redistribution of charge at the conduction surface.
The present device also avoids the problem of latchback (snap back) which is typically caused by the resistance Rb' of the body region 26. When the device is reverse biased, the leakage current through Rb' causes a voltage gradient along the source/P body junction. When this junction becomes forward biased, the NPN parasitic transistor latches. The parasitic transistor is the NPN (bipolar) transistor formed by source 30, body and drift regions 26 and 22, and drain region 20. Since the present device has a vertically wider effective body (regions 22, 26) , Rb' advantageously is decreased.
When the device of Fig. 1 is reverse biased, the depletion region spreads from the drain region 20 into the drift region 22. The thickness of the drift region 22 determines the drain to source breakdown voltage. Formation by diffusion of the body region 26 into the drift region 22 improves control of both the surface concentration and the body region concentration and the resulting channel length, and improves the threshold control. For safe operating area, the effective body junction depth is that of the combined drift 22 and body regions 26, resulting in lower Rb' .
When the device of Fig. 1 is forward biased, the gate voltage easily inverts the conduction surface next to the gate oxide of the drift region 22. Since critical electric field has been achieved in the channel (body) region 26, carriers are injected into the drift region 22 with maximum velocity, resulting in lower resistance for the drift region.
A top side geometry (not shown) suitable for the device of Fig. 1 is any of the well known types of cells, i.e. circular, rectangular, hexagonal, linear, etc. A second embodiment in accordance with the present invention is shown in Fig. 2, the structure of which is
generally identical to those of Fig. 1 (although two trenches are shown rather than one for greater understanding) . Thus trench 38B, gate electrode 46, and BPSG layer 48A in Fig. 2 correspond to structures 38, 46, 48 in Fig. 1 and trench 38B, gate electrode 46B, and BPSG layer 48B are the second trench and associated structures. Fig. 2 shows the additional P+ "body plus" region 62A, 62B formed between two portions of the source region 30 and extending at portion 62A not only into the upper portion of the body region 26 to serve as a body contact, but also extending down at portion 62B into the drift region 22. The doping level of body plus region 62A, 62B is the same or even heavier than that of the body contact region 34 in Fig. 1.
Fabrication of the embodiment of Fig. 2 is compatible with processes already used in the semiconductor industry and provides better control (prevents latchback) of the NPN parasitic transistor present. However, the embodiment of Fig. 2 has a potential detriment in that the breakdown voltage may be compromised by the distance from the body plus region 62B to the drain region 20 i.e., these two regions approach relatively close together, hence providing a potential breakdown path.
Also shown in Fig. 2 is an example of a termination structure which in this case is a polysilicon field plate 66 in the righthand portion of the drawing. The field plate 66 is located on the principal surface of the semiconductor substrate and is in contact with the P+ region 34 to the right of trench 38A. P+ region 34 is a channel stop to prevent conduction. A metallization layer 52B overlies field plate 60, but is isolated from metallization layer 52. This termination structure is also suitable for use with the transistor of Fig. 1.
Another termination structure (not shown) suitable for use with the transistors of Fig. 1 and Fig. 2 is a trench which penetrates (like the trenches in the active region) through the epitaxial layer down into the drain region but having a dummy cell (one not having any N+ source region) in the (exterior) termination portion of the transistor. This second termination structure would typically include a conductive polysilicon gate runner (not shown) which connects to the polysilicon in the termination trench and to the active gate electrodes. One advantage therefore in accordance with the present invention is the ability to use such relatively simple termination structures. A process flow to fabricate a transistor as in Fig. 1 (or Fig. 2) is described hereinafter with reference to Fig. 3A and following. It is to be understood that this process flow is illustrative and not limiting and other process flows may also be used to fabricate a transistor in accordance with the present invention.
1. A conventional silicon substrate 20 is provided in Fig. 3A which is N+ doped to a concentration of 2xl019/cm3. 2. An epitaxial layer 22 approximately 4 μm thick is grown on the upper portion of the substrate. This epitaxial layer is lightly P- doped and has a resistivity of 20 ohm.cm.
3. An oxide layer (not shown) is grown 6,000 A thick over the principal surface of the epitaxial layer. The first mask layer (not shown) is formed over the surface of this oxide layer. The mask is patterned and the oxide layer then etched. This mask thereby defines the active portions of the transistor. 4. In the second mask step, another mask layer 70 is formed on the principal surface and patterned to
define the locations of the gate trenches.
5. The gate trench is 38 anisotropically etched to a depth of 2 to 3 μm and a width of 1 μm in Fig. 3B and mask layer 70 is stripped. 6. Gate oxide layer 42 is grown to a thickness of 100 to 70θA over the sidewalls and floors of the trench and also over the principal surface of the substrate.
7. Polycrystalline silicon 72 (polysilicon) is deposited in the trench 38 and over the silicon principal surface to a thickness of 15,OOθA and on the drain 20 surface of the silicon.
8. The polysilicon is then removed from the backside (drain) surface of the substrate, and any oxide on the backside surface is also removed.
9. In Fig. 3C, the polysilicon is then etched back (planarized) to a final thickness of 5,50θA.
10. The polysilicon is N-doped using for instance phosphorous. 11. Next is the polysilicon masking step in which a photoresist mask layer (not shown) is formed thereover and patterned to define the gate electrodes and gate runners. In this step the polysilicon, after the mask is patterned, is etched down so that the polysilicon gate electrode 46 does do not protrude above the level of the substrate at the trench 38 (the polysilicon 46 is planarized with the silicon principal surface) . The photoresist is then stripped.
12. Next is a blanket implant of the P body region 26 using a dose of 5 x 1013 to 1014/cm2 of boron at 50 KeV energy.
13. The P body region 26 is diffused (driven in) so as to form a 50θA thick oxide layer (not shown) during the drive-in procesε. The doping concentration of the P body region 26 is intended to be in one embodiment 1 to 3 x 1016/cm3 doping level at its surface
next to the gate oxide.
An unmasked (blanket) etch is performed to etch back this 500 A oxide layer to a thickness of 250 to 300 A. 14. Next is the formation and patterning of the N+ source region mask (not shown) .
15. The source region 30 is implanted using this source region mask using a dose of 8 x 1015/cm2 at 80 KeV of arsenic. 16. The source region 30 is diffused (driven in) so as to grow a l,60θA thick oxide layer.
17. The BPSG layer is deposited; the BPSG is doped lightly with phosphorous and boron. (This step and the remaining steps are not depicted, being conventional) .
18. The BPSG layer is flowed.
19. Next is the formation and patterning of the contact mask. This determines the locationε of the P+ body contact regionε. 20. Uεing the contact openingε, the P+ body contact regionε are implanted. For the embodiment of Fig. 2 where it is desired to have the body plus region, this implantation is at a dose of 3xl0I5/cm2 at 50 KeV of boron so as to achieve a final surface concentration of 1019/cm3. This forms the ohmic body contact. (For the Fig. 1 embodiment this implantation is at a dose of 1015/cm2 at 50 KeV.)
21. The BPSG is reflowed (smoothed out) . This reflow step also activates the P+ body contact region implant.
22. An unmasked oxide etch is performed to clear out the contact holes in the BPSG layer, to remove the oxide and any material present due to the reflow process. 23. An aluminum silicon metal layer is deposited to a thickness of e.g. 2.8 μm over the entire
structure .
24. A metal mask layer is formed and patterned and the metal layer is etched accordingly to define the metallization. 25. A passivation layer of PSG is formed over the entire structure.
26. A pad mask layer is formed and patterned and the PSG passivation layer patterned thereby to expose the contact pads. 27. The aluminum/silicon is alloyed.
28. Next there is a back lap of the backside of the substrate.
29. Last is the backside metallization deposit to form the drain electrode. While particular structures and processes are discloεed herein, these are not intended to be limiting. Furthermore, a transistor in accordance with the present invention may be used for applications other than the above-described low voltage application; as is well known, the voltage which the device will withstand is typically limited by the trench and semiconductor region configurations. Particular advantages in accordance with the present invention are that one can achieve a lower threshold voltage and short channel without punch through; improved threshold control due to the P drift region; and the ability to use a relatively simple termination εtructure, rather than the more complex termination εtructureε often used with field effect tranεistors for power applications. This disclosure is illustrative and not limiting; further modifications will be apparent to one skilled in the art in light of this disclosure and are intended to fall within the scope of the appended claims.