MOSFET POWER TRANSISTORS AND METHODS
BACKGROUND OF THE INVENTION 5
1. Field of the Invention
The present invention generally relates to MOSFET power transistors and processes for manufacturing MOSFET power transistors.
2. Prior Art
MOSFET transistors have certain characteristics which can be advantageously applied to power applications such as in power supplies, power amplifiers and the like. In particular, a MOS transistor acts as variable resistance 15 device, and when turned on, has a relatively low resistance inversely dependent on the width of the channel of the MOSFET transistor. Consequently for power devices, the on resistance can be reduced as desired by merely increasing the channel width, though of course this also increases the 2Q semiconductor die size required for the transistor. MOSFET devices are also easily controlled, can be fabricated with relatively high breakdown voltages as required, and do not require an on current such as the base current of a junction transistor. 25
Further, MOSFETs are not subject to any current concentrating phenomenon, allowing the fabrication and use of MOSFET devices having a very wide channel region and/or the operation of multiple identical MOSFET transistors connected in parallel without special protection against 30 current hogging. This is in comparison to a pn junction type device wherein the forward conduction voltage drop across the pn junction decreases with increasing temperature. Thus for a large area, high current device, the temperature across the junction may not be perfectly uniform. Accordingly, any 35 higher temperature region will have a lower forward conduction voltage drop, resulting in the hogging of current from other areas of the pn junction in a rapidly escalating manner until there is a local overheating and failure of the junction. Thus unless special protections are provided, large 40 area pn junctions will fail at current densities way below current densities tolerated by small area pn junctions.
Also various semiconductor packaging techniques for increasing the allowable power dissipation of semiconductor devices are known. However, for any particular packaging 45 technique or design, there will be a limit on the power that may be dissipated in the semiconductor device without adverse affects on the semiconductor device. This limit usually is dependent on power or average power dissipated per unit area of the semiconductor device. For a power 50 MOSFET, the current per unit of substrate area can be increased if the on resistance of the MOSFET transistor or transistors occupying that area can be reduced.
In a conventional MOS structure, source and drain regions are formed in the substrate with a channel region 55 having an insulated gate thereover between the source and drain. Such structures are generally used in integrated circuits, though as power devices require a relatively large substrate area. Also known, however, are trench FETs. These devices are formed by etching trenches in a silicon substrate 60 and effectively forming MOSFETs on the sides of the trenches. Such structures have the advantage of providing a greater channel width per unit of substrate area than the more conventional planar structure, thereby having substantial advantages thereover. The present invention provides a 65 new MOSFET structure and method of fabrication, providing even substantially greater channel widths per unit of
substrate area, thereby providing MOSFET structures having a lower on resistance, and thus a higher current carrying capacity per unit area than even prior art trench FETs.
BRIEF SUMMARY OF THE INVENTION
Trench MOSFETs and self aligned processes for fabricating trench MOSFETs are disclosed. These processes produce a higher density of trenches per unit area than can be obtained using prior art masking techniques. The invention self aligns all processing steps (implants, etches, depositions, etc.) to a single mask, thus reducing the pitch of the trenches by the added distances required for multiple masking photolithographic tolerances.
The invention also places the source regions and contacts within the side walls of the trenches, thus eliminating the lateral dimensions required, for masking and source depositions or implants from the top surface, from the pitch of the trenches.
Other aspects and features of the present invention will become apparent to those ordinarily skilled in the art upon review of the following description of specific embodiments of the invention in conjunction with the accompanying claims and figures.
BRIEF DESCRIPTION OF THE DRAWINGS
The features and advantages of the present invention will become apparent from the following detailed description of the present invention in which:
FIGS. 1A through IP are cross-sectional views illustrating steps in an exemplary process for fabricating the diode connected vertical metal oxide semiconductor field effect devices of the present invention.
FIGS. 2A through 2H are cross-sectional views illustrating steps in another exemplary process for fabricating the diode connected vertical metal oxide semiconductor field effect devices of the present invention.
FIG. 3 is a schematic top view of an integrated circuit comprising an array of four interconnected smaller arrays of FETs in accordance with one exemplary embodiment of the present invention.
DETAILED DESCRIPTION OF THE
PREFERRED EMBODIMENTS
In manufacturing the MOSFETs of the present invention, the masking and manufacturing techniques described in U.S. patent application Ser. No. 09/689,074 entitled "METHOD AND APPARATUS FOR PATTERNING FINE DIMENSIONS" filed by Richard A. Metzler on Oct. 12,2000, which is incorporated herein by reference, can also be used to reduce the cost of masking and provide finer lines in the present invention than might otherwise be attainable.
Processes for fabricating exemplary embodiments of the present invention are presented herein in relation to the cross-sectional views of FIGS. 1A through IP and FIGS. 2A-2N. A MOSFET in accordance with the present invention may typically be thought of as a plurality of very wide channel MOSFET devices (such as may extend over a significant part of, or even most of, a chip dimension or even a die dimension) located in very close proximity to each other and coupled in parallel as a result of the fabrication process itself.
Referring now to FIGS. 1A through IP, cross-sectional views illustrating steps of an exemplary process for fabricating the metal oxide semiconductor field effect devices (MOSFETs) of the present invention are illustrated. The
3
MOSFETs are preferably (but not necessarily) formed using linear pedestals, such as may extend over a significant part of a chip or die dimension, or even over most or essentially all of the associated die dimension. In that regard, FIGS. 1A through IN show cross sections of only a portion of a wafer 5 illustrating the formation of a limited number of commonly connected MOSFET devices. Typically, additional commonly connected devices are simultaneously formed by simultaneous replication of the commonly connected devices across larger portions or each entire die on the 10 semiconductor wafer. Also typically each commonly connected group of multiple devices will incorporate some form of edge termination to provide the breakdown voltage as needed, as is well known in the art.
In the Figures to follow, for purposes of clarity, relative :5 dimensions are not necessarily accurately depicted. Further, while the various layers are generally identified by their respective identification numerals, again for purposes of clarity, each and every Figure is not necessarily to the same scale. 20
For the starting point in the preferred process, a commercially available substrate 100 (n-type preferably) having an epitaxial layer 101 of the same conductivity type thereon. This layer is used to form an increased reverse bias breakdown voltage for the MOSFET devices. Then a sheet 25 implant is used to provide a further n-type layer 114 that ultimately will become the drain regions for the MOSFETs. This layer may be formed using an Arsenic implant with a dose of about 3el4 at about 15 KeV. This is followed by the
30
growth of layer 102 of the opposite conductivity type directly on top of the layer 114, forming a pn junction therebetween. In one embodiment with an N-type substrate, the N-type silicon epitaxial layer 101 has a resistivity of approximately 1.1 ohm-cm and a thickness of approximately 3 fan in order to achieve a reverse bias breakdown voltage 35 of about forty-five volts. The second P-type layer 102 has a resistivity of 0.25 ohm-cm, 0.6 fim thick that is selected to determine the MOSFET threshold.
In the case of an N-type silicon substrate, the lower or 4Q backside surface of the substrate 100 may form the drain while a portion of the top surface (layer 114) of the substrate 100 may form the source. In the case of a P-type silicon substrate, preferably the source and drain terminals are not
reversed, and the lower or backside surface of the substrate „.
45
100 may again form the drain while a portion of the top surface (layer 114 again) of the substrate 100 may form the source, though of course the doping and implants that one might use change in the later case.
A thin oxide layer 103 is grown on the surface of the layer 50 102 as shown in FIG. 1A to randomize the sheet implant which is to follow. The thin oxide 103 is typically three hundred Angstroms (300 A) in thickness. The sheet implant that follows requires no masking by a mask, but rather is composed of ions that are implanted over the entire wafer. 55 The sheet implant is to provide a good ohmic contact for the P-type back gate region of the MOSFETs. The implant is a Boron implant at about 4xl015 atoms per cm2 with an energy of 15 KeV. Then a layer of oxide 104 is applied on top of the thin oxide 103 across the wafer. 60
Referring to FIG. IB, completion of a first masking and etching step is illustrated. The oxide layer 104 is patterned using a mask and areas are etched away, including parts of the thin oxide 103, to leave linear construction pedestals 104 on top of the thin oxide 103. In one embodiment, the 65 construction pedestals 104 are approximately 0.5 micron (/an) high.
4
FIG. IB illustrates a cross-section of four of a plurality of linear pedestals 104 that are formed across the silicon wafer. The dimensions of the rectangular cylindrical construction pedestals 104 in one embodiment are approximately 0.4 microns in width, approximately 1.0 micron in height, with a pitch of approximately 2.0 microns. It is understood that these dimensions can be altered in coordination with any adjustment in the implantation levels in order to provide similar device physics for the MOSFETs. Region 110 of the wafer is shown on an expanded scale in FIG. 1C, with the subsequent Figures showing details of the processing for and around adjacent pedestals 104.
Referring now to FIG. ID, a portion of the substrate 100 between pedestals 104 has been etched away to a depth of about 1500 A, forming shallow silicon trenches 108 and substrate pedestals 109 in layer 102. This etch step may be a Reactive Ion Etch (RIE, a directional etch) commonly used in silicon processing to form trench metal-oxidesemiconductor (MOS) transistors and capacitors. An N-type anode contact implant is now performed, typically using Arsenic, with a dose of about 3E15 and an energy of about 15 KeV. This provides regions 112 around the periphery of each pedestal that will become the source of the device.
Referring now to FIG. IE, additional silicon directional etching has been performed to a depth of 0.6 fan. Preferably this etch extends entirely through layer 114 between pedestals for improved breakdown, but alternatively may stop within layer 114.
This leaves a source region 112 at the top of the silicon pedestal structure, and drain region 114 near the bottom of the pedestal in contact with the N-type epitaxial layer 101 if used (as shown), or to the N-type substrate if an N-type epitaxial layer is not used. (Use of region 114 is not mandatory, but may be useful however to lower the on resistance of the devices.) If desired, a threshold adjustment implant may now be performed, if desired.
As presented in FIG. IF, the remaining part of the oxide pedestals 104 is stripped away, and a layer of gate oxide 116, 100 A thick, and a layer of heavily doped poly silicon 118 are conformally deposited on the device in the first stage of the gate formation. (Alternatively, a metal layer may be deposited and processed much like the poly silicon layer to ultimately form metal gates for the FETs.)
The next process is the RIE etching (a directional etch) of the poly silicon layer, stopping on the oxide layer, to leave sidewall gate regions 118 shown in FIG. 1G. This is followed by a Boron implant that forms a P-type region 120 to block the potential from the bottom of the gate oxide when the MOSFET is off.
A further poly silicon RIE etch is performed as shown in FIG. 1H. This moves the top of the poly gates 118 down to allow access to the source regions 112. This is followed by the deposition of a gate interconnect material 126 as shown in FIG. II. This layer can be poly silicon, a silicide (such as tungsten silicide), a barrier metallization (such as Titanium Nitride), or a metal (such as copper). This layer is to provide a low resistance path to all of the gate regions, which will be accessed at an external bond pad region. The pedestal spacing (less than twice the deposited thickness) is such that the region between the pedestals is completely filled. A CMP process may now be performed to planarize the interconnect material surface, if desired.
FIG. 1J shows the results of a subsequent reactive ion etch to remove the excess interconnect material 126, and to countersink the interconnect material surface into the region between pedestals. The interconnect material completely
5
covers the gates 118, but is isolated from the source region by the gate oxide layer 116. Then a gate isolation layer 128 is deposited in a fashion to again fill the trench spacing, as shown in FIG. IK. This layer is an insulating layer such as Si02. A reactive ion etch is again applied to remove the 5 excess deposited oxide as shown in FIG. 1L. This etch also removes the residual oxide on top and on the upper side of the exposed pedestal. FIG. IN shows the finished device with a barrier metal layer 122 (FIG. 1M), and a top metal layer 124 applied over the structure. 10
Not shown in the foregoing Figures is the drain contact for the MOSFET device, and the manner of bringing out of the source and gate contacts. In some applications, the drain contact is made by depositing a metallization layer 130 on the back of the substrate 100 opposite the vertical MOSFET :5 devices 132, as shown in FIG. 10. In other applications, the vertical MOSFET devices 132 are formed in a well 134, and the drain contact is made by depositing a metallization layer 130 on the well surface adjacent or surrounding the vertical MOSFET devices, as shown very schematically in FIG. IP. 20 The gate contact may be made by bring out (exposing) the gate metallization layer in other locations on the chip for connection to a gate lead for the device.
One of the advantages of the foregoing process is that the process is effectively self aligning, in that masking steps are 25 held to a bare minimum, and a very high density MOSFET structure results without any corresponding masking accuracy required. In particular, the initial masking step to define the pedestals 104 of FIG. IB, being the first masking step, need not be accurately aligned. Thereafter, there may be no 30 masking steps until the FETs are fully fabricated, as shown in FIG. IN, after which masking may be used to expose the source and gate connections for connection to the outside world. Note however that the mask alignment for any such purposes may be approximate only, and need not be of the 35 scale and accuracy with which the FETs are defined and fabricated by the self aligned process. Alternatively, masking may be used during the FET fabrication process itself to provide for what will ultimately form the gate and source contacts for connection to the outside world. Here again, any 40 such masking need not be of the scale and accuracy with which the FETs are defined and fabricated by the self aligned process.
Further, the foregoing process has the advantage of pro- 4J viding the source contact by way of a metal layer over the FETs. This allows the use of a thick conductive layer for the source contact, minimizing the resistance of this conductor.
Now referring to FIGS. 2A through 2H, an alternate process for forming the MOSFETs of the present invention 50 may be seen. In this process, the initial steps of FIGS. 2A through 2C are substantially the same as for that described with respect to FIG. 1, specifically FIGS. 1A through 1C. Accordingly, FIGS. 2A through 2C use the same numerical identifications as FIGS. 1A through 1C. However, a nitride 55 layer is provided between the thin oxide layer (103 in the earlier Figures) and the thick oxide layer (104 in the earlier Figures). For purposes of the present Figures, the thin oxide layer 103 of the earlier Figures may be considered to comprise a thin oxide layer with a nitride layer thereon. 60
Next, a layer of poly silicon 201 is deposited (FIG. 2D), followed by a directional etch to remove the poly silicon from the horizontal surfaces, leaving the poly silicon sidewalls 201 of FIG. 2E on the oxide pedestals 104. Removal of the oxide pedestals 104 (FIG. 2F) leaves the sidewall 65 regions 201 for use as construction pedestals in place of the oxide construction pedestals 104 of the previous fabrication
6
method. An N-type anode contact implant is now performed, typically using Arsenic, with a dose of about 3E15 and an energy of about 15 KeV. This provides regions 112 around the periphery of each pedestal that will become the source of the device, as in FIG. ID. It also forms similar regions at the top of the poly silicon construction pedestals 201. Thereafter, a further directional etch is performed, as illustrated in FIG. 2H (like that illustrated in FIG. IE). This leaves a source region 112 at the top of the silicon pedestal structure, and drain region 114 near the bottom of the pedestal in contact with the N-type epitaxial layer 101 if used (as shown), or to the N-type substrate if an N-type epitaxial layer is not used. (Use of region 114 is not mandatory, but may be useful however to lower the on resistance of the devices.) The etch also removes regions 112 at the top of the construction pedestals, and may remove most if not all of the remaining material of the original construction pedestals, though some vertical thickness of the construction pedestals may remain. Otherwise, processing to completion of the FETs may proceed as before described, resulting in FETs of substantially the same structure as in the previously described embodiment.
This embodiment of the process preserves the advantages of the process being effectively self aligning and of providing the source contact by way of a continuous thick conductive layer over the FETs. It also has the advantage that the sidewalk used for the construction pedestals may be of a smaller scale and more accurately controlled in width than is possible by definition by masking processes. Consequently, this together with the fact that a single oxide construction pedestal of the prior process yields two construction pedestals in this process, yields an even greater density MOSFET by this method, providing an even greater "on" current density per unit area of semiconductor substrate.
In a completed part in accordance with the exemplary fabrication methods such as illustrated in FIG. IN, region 120, being of the opposite conductivity type from the epitaxial layer 101, will provide electrical isolation of the gate interconnect material 126 from the substrate 100. Consequently, in embodiments having the region 120, the thin oxide 116 need not extend over the substrate between gate regions on adjacent pedestals, or even under the gate, and therefore may be removed from those areas at appropriate stages of the processing if desired. Similarly, in embodiments having the thin oxide 116 extending under the gate regions and over the substrate between gate regions on adjacent pedestals, formation of the regions 120 is optional. The presence of both regions 120 and the thin oxide between gate regions on adjacent pedestals is preferred however, as both help lower the gate capacitance.
There has been disclosed herein certain exemplary methods of fabricating diode configured vertical MOSFETs, as well as the MOSFETs so made. It will be noted that in the exemplary processes there is a single masking step having only a trivial alignment requirement with respect to the semiconductor substrate on which the MOSFETs are to be formed. Thereafter, each additional step is self-aligning with respect to the prior step, thereby eliminating multiple masks, and particularly the critical alignment requirement between individual masks of mask sets typically used for the fabrication of semiconductor devices. This simplifies processing, increases yields and reduces wafer to wafer device variations due to differences in mask alignment. Also while not a limitation of the present invention, as stated before, preferably the construction pedestals and thus the resulting NMOSFETs are straight and parallel, as opposed to hollow
|
