The present invention generally concerns fabrication methods and device architectures for use in memory circuits, and more particularly concerns hybrid silicon-on-insulator (SOI) and bulk architectures for use in memory circuits.
There are several semiconductor memory types available for use in constructing memory components for incorporation in computers and other electronic devices. In addition, there are numerous semiconductor fabrication processes available for forming memory components operating in accordance with the several semiconductor memory types. Further, almost continual progress is being made in process and fabrication techniques, resulting in improvements to component speed and stability of operation and reductions in component size and power consumption. In such a situation it is therefore a continuing challenge to adapt emerging process, fabrication and device improvements to semiconductor memories in such a way that the maximum benefit is derived from the improvements.
For example, regarding semiconductor memory types two are most common—dynamic random access memories (hereinafter “DRAMs”) and static random access memories (hereinafter “SRAMs”). A DRAM memory is comprised of DRAM cells which essentially are capacitors for storing charge; the states of the capacitors constitute the memory states of the DRAM cell. DRAMs have relatively high memory densities in comparison to other memory technologies, for example SRAM memories, but this comes at a cost. For various reasons well-known to those skilled in the art, capacitors comprising the memory cells in DRAM devices cannot maintain their charge states in perpetuity and therefore have to be occasionally refreshed in order not to lose their memory state.
In contrast to DRAMs, SRAMs store information in bistable semiconductor circuits. More devices need to be fabricated in order to construct an SRAM memory cell in comparison to a DRAM memory cell, resulting in DRAMs generally achieving better memory density. On the other hand, SRAMs need not be refreshed in the manner of capacitive DRAMs. In addition, SRAMs generally have shorter read/write cycle times. Thus, SRAMs often are used in microprocessors in so-called “cache memory.”
After forming a gate oxide layer on the surface of the silicon regions, gates 116 are patterned and formed at the location of devices 142, 144. Source/drain regions 132 are defined using a standard implant and diffusion step, after forming lightly doped diffusion regions 134 at the gate boundaries, if desired. Device channels 136 are completely isolated from other channels by source/drain diffusions 132 at either end, BOX layer 120 below, gate oxide above and STI (not shown) along the sides of the channel. Further, “halo” regions 133 have been formed between the source/drain regions 132 and channel 136 through a separate diffusion step of the same dopant type used to form the body regions but at a higher concentration.
Ideally, the thin silicon surface layer 130 is no thicker than what is necessary to form a channel 136 between a pair of source/drain diffusions 132. However, in practice, the silicon surface layer can be thicker than the depth of the FET's channel inversion layer. So, when the channel inversion layer forms, i.e., when the FET is turned on, an uninverted layer can remain beneath the channel inversion layer. This uninverted layer remains isolated, resistively, from adjacent regions and any charge that is introduced into the uninverted channel region remains trapped there until it leaks out through junction leakage or is otherwise coupled out. This trapped charge can produce unwanted device channel bias resulting in what is referred to as body effects that are localized to an individual device.
So, these prior art SOI FETs 142, 144 have isolated floating channels (body regions)136 that are not biased by any bias voltage. Thus, the channel bias of any device is dependent upon its current operating state and the device's history, i.e., any remaining charge that has been previously introduced through capacitive coupling or bipolar injection. For typical individual logic circuits such as, decoders, clock buffers, input or output drivers and array output drivers, variations in device characteristics resulting from floating device channels are predicted in device models and are accounted for in chip timing.
Localized body effects present significant problems for CMOS SOI SRAM arrays. This floating body effect allows the body potential and threshold voltage to vary from device to device within a single cell, introducing a use-dependent bias. There are several contributors to this variation or mismatch and body potential is a significant contributor. If the mismatch between devices is sufficiently large the cell will be disturbed during a read or a write operation and even in an idle state. Then data may be lost.
It is known that coupling the bodies of the devices reduces the body potential and threshold mismatch of the devices and so enhances the stability of SRAM cells. Known methods of doing this are to use so-called body-contacted SOI MOSFET transistors. When applied to CMOS SRAM cells, these methods significantly increase cell area and process complexity. The increase in area can be as much as two to three times for each transistor with a small dimension as used in SRAM cells and sums to form at least a two-times-larger SRAM cell. Another drawback is that the parasitic capacitance associated with the polysilicon gate and diffusions of the body-contacted transistor will degrade the SRAM array performance.
Problems have been encountered in other areas as well. Advances have been made in fundamental substrate fabrication techniques which now permit portions of a substrate to be fabricated in silicon with different crystal orientations. It has long been known that PFETs experience improved performance when fabricated in (110) crystal orientation silicon due to the increased mobility of the majority carrier (holes) in (110) crystal orientation silicon. It has only become possible recently to form (110) crystal orientation silicon regions no larger than the PFET devices themselves so that such regions can be incorporated in an otherwise (100) crystal orientation substrate without sacrificing device density. Although advantageous, the hybrid substrate technology must be used judiciously so that the potential improvements in device performance achievable with such technology are not blunted through sub-optimal decisions concerning other design issues, for example, overcoming the floating body effect.
Those skilled in the art thus desire CMOS SRAM cell architectures that overcome the problems associated with the floating body effect without sacrificing the gains made by fabricating the device in SOI, for example, improved read/write speeds and lower power consumption. In particular, such an improved CMOS SRAM cell architecture would have improved stability, and experience far fewer anomalies during read/write operations.
In addition those skilled in the art desire improved SRAM cell layouts that derive increased benefit from linked body technology; in particular, those skilled in the art desire SRAM cell layouts that reduce the resistance encountered in devices having linked bodies.
Further, those skilled in the art also desire the judicious use of state of the art device structures to improve the performance of the logic and memory portions of SRAM memories or microprocessors. In particular, those skilled in the art desire the application of state of the art device structures to improve the speed of logic operations in the logic portion of an SRAM memory or microprocessor and the stability of the memory portion of the SRAM memory or microprocessor.
A first embodiment of the present invention comprises an SRAM array comprising a plurality of SRAM cells, each of said SRAM cells comprising: a pair of cross-coupled CMOS inverters in a surface silicon layer disposed on an SOI buried-oxide layer, where each cross-coupled inverter comprises an NFET and a PFET; a pair of NFET pass gates selectively coupling a pair of bit lines to said cross-coupled CMOS inverters; and where at least two adjacent NFETs of the SRAM cell share a leakage path between body regions, and where the at least two adjacent NFETs have a source/drain diffusion region and a leakage path diffusion region under the source/drain diffusion region positioned between their respective body regions, wherein the source/drain diffusion region extends fractionally into the surface silicon layer and the leakage path diffusion region extends from a bottom of the source/drain diffusion down to the SOI buried-oxide layer, and where the leakage path diffusion region is counter-doped with the same dopant type as the source/drain diffusion but at relatively lower concentrations than the source/drain diffusion, thereby presenting a lower barrier to junction leakage than the source/drain regions.
A second embodiment of the present invention comprises a pair of adjacent SRAM cells in an SRAM array, the pair comprising a first SRAM cell and a second SRAM cell, each of the adjacent SRAM cells comprising: a pair of cross-coupled CMOS inverters in a surface silicon layer disposed on an SOI buried oxide layer, where each cross-coupled inverter comprises an NFET and a PFET; a pair of NFET pass gates selectively coupling a pair of bit lines to said cross-coupled CMOS inverters; and where at least one of the NFETs from the first SRAM cell and at least one of the NFETs from the second SRAM cell share a leakage path between body regions, the respective NFETs sharing a leakage path being adjacent to one another and where the at least two adjacent NFETs have a source/drain diffusion region and a leakage path diffusion region under the source/drain diffusion region positioned between their respective body regions, wherein the source/drain diffusion region extends fractionally into the surface silicon layer and the leakage path diffusion region extends from a bottom of the source/drain diffusion down to the SOI buried-oxide layer, and where the leakage path diffusion region is counter-doped with the same dopant type as the source/drain diffusion but at relatively lower concentrations than the source/drain diffusion, thereby presenting a lower barrier to junction leakage than the source/drain regions.
A third alternate embodiment of the present invention comprises a pair of adjacent SRAM cells in an SRAM array, the pair comprising a first SRAM cell and a second SRAM cell, each of the adjacent SRAM cells comprising: a pair of cross-coupled CMOS inverters in a surface silicon layer disposed on an SOI buried oxide layer, where each cross-coupled inverter comprises an NFET and a PFET; a pair of NFET pass gates selectively coupling a pair of bit lines to said cross-coupled inverters; and where at least one of the PFETs from the first SRAM cell and at least one of the PFETs from the second SRAM cell share a leakage path between body regions, the respective PFETs sharing a leakage path being adjacent to one another and where the at least two adjacent PFETs have a source/drain diffusion region and a leakage path diffusion region under the source/drain diffusion region positioned between their respective body regions, wherein the source/drain diffusion region extends fractionally into the surface silicon layer and the leakage path diffusion region extends from a bottom of the source/drain diffusion down to the SOI buried-oxide layer, and where the leakage path diffusion region is counter-doped with the same dopant type as the source/drain diffusion but at relatively lower concentrations than the source/drain diffusion, thereby presenting a lower barrier to junction leakage than the source/drain regions.
A fourth alternate embodiment of the present invention comprises a pair of adjacent SRAM cells in an SRAM array, the pair comprising a first SRAM cell and a second SRAM cell, where each of the SRAM cells have two longitudinal and two lateral sides, the adjacent SRAM cells sharing a longitudinal side, each of the adjacent SRAM cells comprising: a pair of cross-coupled CMOS inverters in a surface silicon layer disposed on an SOI buried-oxide layer, wherein the cross-coupled CMOS inverters each comprise an NFET and a PFET; a pair of NFET pass gates selectively coupling a pair of bit lines to said cross-coupled CMOS inverters, and wherein one each of the pass gate NFETs and inverter NFETs are positioned along each of the lateral sides of the SRAM cell, whereby the pass gate NFET and inverter NFET positioned on the same lateral side of the SRAM cell comprise a pair and have body regions linked with leakage path diffusion regions formed beneath adjacent shallow source/drain diffusions wherein the shallow source/drain diffusion region extends fractionally into the surface silicon layer and the leakage path diffusion region extends from a bottom of the source/drain diffusion down to the SOI buried-oxide layer, and where the leakage path diffusion region is counter-doped with the same dopant type as the source/drain diffusion but at relatively lower concentrations than the source/drain diffusion, thereby presenting a lower barrier to junction leakage than the source/drain regions.
A fifth alternate embodiment of the present invention comprises an SRAM array comprising a plurality of SRAM cells organized in rows and columns, wherein each of the SRAM cells have two longitudinal sides and two lateral sides, the SRAM cells further comprising: a pair of cross-coupled CMOS inverters in a surface silicon layer disposed on an SOI buried-oxide layer, where each cross-coupled inverter comprises an NFET and a PFET; a pair of NFET pass gates selectively coupling a pair of bit lines to said cross-coupled CMOS inverters; and where at least two adjacent NFETs of the SRAM cell share a leakage path between body regions, and where the at least two adjacent NFETs have a source/drain diffusion region and a leakage path diffusion region under the source/drain diffusion region positioned between their respective body regions, wherein the source/drain diffusion region extends fractionally into the surface silicon layer and the leakage path diffusion region extends from a bottom of the source/drain diffusion down to the SOI buried-oxide layer, wherein the leakage path diffusion region is counter-doped with the same dopant type as the source/drain diffusion but at relatively lower concentrations than the source/drain diffusion, thereby presenting a lower barrier to junction leakage than the source/drain regions, and where each of the SRAM cells arrayed in a particular row of the SRAM array share longitudinal sides with two other SRAM cells positioned in the same row, except for at least two of the SRAM cells one longitudinal side of each coincides with a termination point of the row, and where the at least two adjacent NFETs of each SRAM cell arrayed in the particular row of the SRAM array having body regions linked by the leakage path diffusion region have their body regions further linked to the body regions of NFETs contained in adjacent SRAM cells sharing longitudinal sides with the SRAM cell with leakage path diffusion regions positioned beneath adjacent shallow source/drain diffusion regions, except for the at least two of the SRAM cells having one longitudinal side coinciding with the termination point of the row which have at least one pair of NFETs having a body region linked to the body regions of NFETs positioned in one SRAM cell on a longitudinal side opposite from the termination point of the row with leakage path diffusion regions positioned beneath adjacent shallow source/drain diffusion regions; and whereby a continuous chain of NFETS having body regions linked with leakage path diffusion regions positioned beneath adjacent shallow source/drain diffusion regions exists across the particular row of the SRAM array.
A sixth alternate embodiment of the present invention comprises a microprocessor fabricated on a CMOS hybrid orientation substrate, wherein the microprocessor comprises a logic portion and a cache memory portion, wherein the cache memory portion further comprises at least one CMOS SRAM array and where: the logic portion comprises, in part, PFETs fabricated in (110) crystal orientation bulk silicon regions and NFETs fabricated in (100) crystal orientation SOI silicon regions, wherein the NFETs in the logic portion have floating body regions; and the CMOS SRAM array comprises a plurality of CMOS SRAM cells comprising, in part, PFETs fabricated in (110) crystal orientation silicon regions and NFETs fabricated in (100) crystal orientation SOI silicon regions, wherein at least a portion of the NFETs in the CMOS SRAM cells have body regions linked to body regions of adjacent NFETs with leakage path diffusion regions formed beneath adjacent shallow source/drain diffusions wherein the source/drain diffusion regions extend fractionally into a surface silicon layer and the leakage path diffusion regions extends from bottoms of the source/drain diffusions down to an SOI buried-oxide layer, and where the leakage path diffusion regions are counter-doped with the same dopant type as the source/drain diffusions but at relatively lower concentrations than the source/drain diffusions, thereby presenting a lower barrier to junction leakage than the source/drain regions.
A seventh alternate embodiment comprises a method of forming an SRAM array comprising a plurality of SRAM cells, each of the SRAM cells comprising a pair of cross-coupled CMOS inverters in a surface silicon layer disposed on an SOI buried-oxide layer, where each cross-coupled inverter comprises an NFET and a PFET and the SRAM cell further comprises a pair of NFET pass gates selectively coupling a pair of bit lines to said cross-coupled CMOS inverters, the method comprising: forming a buried oxide layer in a silicon wafer, the buried oxide layer positioned between a surface silicon layer and a silicon substrate; forming a plurality of PFET and NFET gates above body regions in the surface silicon layer; forming a leakage path diffusion region between at least a pair of adjacent NFET body regions wherein the leakage path diffusion region is counter-doped with a same dopant type as a shallow source/drain diffusion to be formed in another step but at relatively lower concentration than the shallow source/drain diffusions, thereby presenting a lower barrier to junction leakage than the source/drain regions, the leakage path diffusion region extending to the buried oxide layer; and forming the shallow source/drain diffusions above the leakage path diffusion regions, the shallow source/drain diffusions extending fractionally into the surface silicon layer.
An eighth alternate embodiment of the present invention comprises an SRAM memory comprising: peripheral logic fabricated in a high-performance silicon substrate; an SRAM array comprised of a plurality of SRAM cells, wherein the SRAM cells are arrayed in rows and columns and further comprise: a pair of cross-coupled CMOS inverters in a surface silicon layer disposed on an SOI buried-oxide layer, wherein the cross-coupled CMOS inverters each comprise an NFET and a PFET; a pair of NFET pass gates selectively coupling a pair of bit lines to said cross-coupled CMOS inverters, where body regions of NFETs arrayed along a column of SRAM cells coinciding with a bit line are linked by leakage path diffusion regions beneath adjacent shallow source drain diffusion regions, thereby forming a chain of linked body regions.
In one variant of the eighth alternate embodiment, the high-performance silicon substrate of the peripheral logic comprises a strained silicon region.
In another variant of the eighth alternate embodiment, the high-performance silicon substrate of the peripheral logic comprises a hybrid orientation substrate, where the NFETs are fabricated in (100) crystal orientation silicon regions and PFETs are fabricated in (110) crystal orientation silicon regions.
A ninth alternate embodiment of the present invention comprises an SRAM memory comprising: peripheral logic comprised of CMOS NFETs and PFETs, where the NFETs are fabricated in bulk silicon regions and the PFETs are fabricated in SOI silicon regions, where body regions of the PFETs are floating; and an SRAM array comprised of a plurality of SRAM cells, wherein the SRAM cells are arrayed in rows and columns and further comprise: a pair of cross-coupled CMOS inverters in a surface silicon layer disposed on an SOI buried-oxide layer, wherein the cross-coupled CMOS inverters each comprise an NFET and a PFET; a pair of NFET pass gates selectively coupling a pair of bit lines to said cross-coupled CMOS inverters; where body regions of NFETs along a column of SRAM cells coinciding with a bit line are linked by leakage path diffusion regions beneath adjacent shallow source drain diffusion regions, thereby forming a chain of linked body regions.
A tenth alternate embodiment of the present invention comprises a method of forming an SRAM memory comprised of an SRAM array portion and a peripheral logic portion, where the SRAM array portion is comprised of a plurality of SRAM cells, and where each of the SRAM cells further comprises a pair of cross-coupled CMOS inverters in a surface silicon layer disposed on an SOI buried-oxide layer, where each cross-coupled inverter comprises an NFET and a PFET and the SRAM cell further comprises a pair of NFET pass gates coupling a pair of bit lines to the cross-coupled CMOS inverters, the method comprising: forming a high-performance silicon substrate portion in a silicon wafer; forming circuits comprising the peripheral logic portion of the SRAM memory in the high-performance silicon substrate portion of the silicon wafer; forming the SRAM array portion of the SRAM memory by: forming a buried oxide layer in the silicon wafer, the buried oxide layer positioned between a surface silicon layer and a silicon substrate; forming a plurality of PFET and NFET gates above body regions in the surface silicon layer; forming a leakage path diffusion region between at least a pair of adjacent NFET body regions wherein the leakage path diffusion regions are counter-doped with a same dopant type as a shallow as a shallow source/drain diffusion to be formed in another step but at a relatively lower concentration than the shallow source/drain diffusions, thereby presenting a lower barrier to junction leakage than the source/drain regions, the leakage path diffusion region extending to the buried oxide layer; and forming the shallow source/drain diffusions above the leakage path diffusion regions, the shallow source/drain diffusions extending fractionally into the surface silicon layer.
In one variant of the tenth alternate embodiment the high-performance silicon substrate portion comprises a hybrid orientation substrate having (100) crystal orientation silicon regions and (110) crystal orientation silicon regions.
In another variant of the tenth alternate embodiment, the high-performance silicon substrate portion comprises a strained silicon region.
Thus it is seen that embodiments of the present invention overcome limitations of the prior art. Known device structures suitable for use in overcoming the floating body effect in SOI SRAM arrays suffer from a number of drawbacks. In particular, one known method comprising the use of body-contacted SOI MOSFET transistors significantly increases SRAM cell area and processing complexity. The present invention significantly reduces the need for increased SRAM cell area and further reduces the resistance of the leakage path used to join the body regions through an improved SRAM cell layout.
In addition, embodiments of the present invention improve the overall performance of SRAM memories by applying high-performance substrate technologies to peripheral logic portions of SRAM memories to improve the speed of operation of such portions, while applying linked body technology to the memory array portion of the SRAM memory to improve the stability of the array. This can be accomplished with differing high-performance substrate technologies; for example, strained silicon substrates or hybrid orientation substrates. Strained silicon substrates, which are relatively less expensive to fabricate when compared to hybrid orientation substrates, can be used in applications where improved performance is sought. In other applications where state-of-the-art performance is sought, hybrid orientation substrates can be used to achieve the fastest possible operation for the peripheral logic portions of SRAM memories.
Further, the present invention judiciously applies state of the art device structures to accomplish improved overall performance for microprocessors. In particular, the present invention applies hybrid orientation technology in combination with advances in overcoming the floating body effect to achieve an overall improvement in microprocessor performance. In particular, fabrication of the logic portion of a microprocessor in CMOS with NFETs in (100) crystal orientation SOI silicon where the NFETs have floating bodies and PFETs in (110) crystal orientation bulk silicon improves the operating speed of the logic portion of the microprocessor.
In conclusion, the foregoing summary of the alternate embodiments of the present invention is exemplary and non-limiting. For example, one of ordinary skill in the art will understand that one or more aspects or steps from one alternate embodiment can be combined with one or more aspects or steps from another alternate embodiment to create a new embodiment within the scope of the present invention.
The foregoing and other aspects of these teachings are made more evident in the following Detailed Description of the Preferred Embodiments, when read in conjunction with the attached Drawing Figures, wherein:
Supply asymmetry switches 154 selectively provide a higher, offset voltage, mutually exclusively, to one or the other of the cell supply line pairs. A bit decode circuit 156 decodes a bit address to select one of N columns 158 of cells in the array 152. Each of the N columns 158 of cells in the array 152 is connected to one of N pairs of column supply lines. A word decoder 160 selects a row of cells by driving one of M local word lines 162. So, in this example, the M by N array is addressed by coincidence of a selected row 162 with a selected column 158. During a read, bit select 164, which may include a sensing capability, selects one column 158 and buffers and re-drives data that is stored in the selected cells in that column 158. An active/passive supply couple, e.g., located with supply asymmetry switches 154 or with the bit select 164, passes a nominal supply voltage to the array 152; and when appropriate, allows the supply asymmetry switches 154 to mutually exclusively pass an offset voltage to one or the other of a pair of column supply lines. Examples of suitable supply couples include a resistor, diode or FET connected between the array supply and each of the column supply lines. Data input/output (I/O) drivers 166 receive input data and drive selected data from the bit select 164, for example, off chip. Clock logic 168 provides local timing and glue logic 169 provides local control, for example, read/write select, address gating and buffering, etc. In the present application, “SRAM array” generally refers to array portion 152 of SRAM memory 150 and “peripheral logic” refers to the remaining portions of the SRAM memory. Further, the SRAM memory 150 depicted in
Normally, the supply asymmetry switches 154 are open. Matched supply voltages (nominal) are provided to each pair of column supply lines to maintain cell symmetry. During accesses, supply asymmetric switches 154 switch the higher offset voltage onto one side of the cell in each column being accessed. Thus, the higher offset voltage unbalances the voltage on each pair of column supply lines, making the cells in each unbalanced column 158 asymmetric during the access. In particular, the offset voltage is switched such that the imbalance or asymmetry favors any data state being stored/read. So, the imbalance facilitates writing and reading data to and from preferred embodiment storage cells.
Each SRAM cell 200 is written by pulling one of the bit line pair 250, 260 high and the other low while holding word line 270 high so that both access transistors 230, 240 are on; and, then pulling the word line 270 low to turn off the access transistors 230, 240, trapping the state of the bit lines in the cross coupled inverters 210, 220. The SRAM cell 200 is read by pre-charging the bit lines 250,260 to a known state; driving the word line 270 high which couples the cross coupled inverters 210, 220 through the access transistors 230, 240 to the bit line pair 250, 260; and, then, measuring the resulting voltage difference on the bit line pair 250, 260. The signal on the bit line pair 250, 260 increases with time toward a final state wherein each one of the pair 250, 260 may be, ultimately, a full up level and a full down level. However, to improve performance, the voltage difference is sensed well before the difference reaches its ultimate value.
As noted hereinabove, in a prior art bulk CMOS technology 214C, 224C, 230C and 340C were tied to GND and, 212C and 222C were tied to Vhi biasing the respective devices. However, in the prior art SOI process of
The body potential mismatch problem is solved in this invention by linking the bodies of adjacent devices together with a leakage path diffusion positioned beneath a shallow source/drain diffusion. This avoids butting of the deep source and drain implants against the backside of the silicon film in an SOI device. It creates shallow source and drain implants with a leakage path diffusion region near the backside of the SOI silicon film to allow the two bodies of neighboring devices to electrically connect to each other via the leakage path. In some embodiments made in accordance with this invention, the body of a PFET is linked to an adjacent PFET in a neighbor cell using shallow source and drain implants and a leakage path diffusion region. The bodies of NFETs are linked along the bit line and further connected to ground at two sides of the bit line row. By doing this, the threshold voltage mismatches between pull down and pass gate NFETs and between PFETs are reduced.
The shallow source drain implant is performed by blocking the source/drain area of SRAM FETs from the normal deep source drain implant and defining separate lithographic regions in the SRAM arrays coinciding with the leakage path diffusion regions to receive less energy or smaller doses of implanted species. By opening contacts on source drain diffusions of NFETs at both sides of one bit line row, the bodies of N-type devices can be tied to a fixed potential, such as ground, for enhanced performance. Performance depends on the effectiveness of body linking.
To demonstrate the invention at work, a cross-section of the body linked and grounded pull down and pass gate devices is shown in
After forming a gate oxide layer on the surface of the silicon regions, gates 316 are patterned and formed at the location of devices 342, 344. Source/drain regions 332 are defined using a standard implant and diffusion step, after forming lightly doped diffusion regions 334 at the gate boundaries, if desired. In various embodiments halo regions 333 can be formed by a separate diffusion step of the same species as the body region. In other embodiments halo regions can be deleted. In contrast to the device architecture depicted in
Layouts of the body-linked SOI SRAM array/cells are illustrated in
In
A particular advantage of the embodiment depicted in
Bodies of P-type devices are linked in a similar way by using a leakage path diffusion region P+ implant mask “WP” in the regions 616 depicted in
Tying the connected body chain of N-type devices to ground is implemented by performing a normal P+ source drain implant on the outside diffusions of edge cells 339 (
SRAM cell stability is studied using Access Disturb Margin approach. The stability dependence on threshold voltage mismatch is plotted in
A method for fabricating a CMOS SRAM array in accordance with the present invention is depicted in
In one variant of the method depicted in
In another variant of the method depicted in
In other embodiments of the invention improved overall performance for SRAM memories is achieved by applying high-performance substrate technologies to peripheral logic portions of an SRAM memory in combination with the linked body technology applied to the memory array portion of the SRAM memory. For example, with reference to
An exemplary method for fabricating an SRAM memory having a peripheral logic portion implemented in a high-performance silicon substrate is depicted in
One of ordinary skill in the art will understand that the ordering of the steps in
In one variant of the method of
In another variant of the present invention, the high-performance silicon substrate portion comprises a strained silicon region.
Thus it is seen that the foregoing description has provided by way of exemplary and non-limiting examples a full and informative description of the best method and apparatus presently contemplated by the inventors for creating CMOS SOI SRAM arrays having body regions of FETs linked with leakage path diffusion regions. One skilled in the art will appreciate that the various embodiments described herein can be practiced individually; in combination with one or more other embodiments described herein; or in combination with SOI CMOS SRAM architectures differing from those described herein. Further, one skilled in the art will appreciate that the present invention can be practiced by other than the described embodiments; that these described embodiments are presented for the purposes of illustration and not of limitation; and that the present invention is therefore limited only by the claims which follow.
Number | Name | Date | Kind |
---|---|---|---|
4965213 | Blake | Oct 1990 | A |
5906951 | Chu et al. | May 1999 | A |
6059895 | Chu et al. | May 2000 | A |
6316808 | Smith, III | Nov 2001 | B1 |
6344671 | Mandelman et al. | Feb 2002 | B1 |
6387739 | Smith, III | May 2002 | B1 |
6603156 | Rim | Aug 2003 | B2 |
6621727 | Chen | Sep 2003 | B2 |
6624459 | Dachtera et al. | Sep 2003 | B1 |
6627952 | Wollesen | Sep 2003 | B1 |
6642536 | Xiang et al. | Nov 2003 | B1 |
6661253 | Lee et al. | Dec 2003 | B1 |
Number | Date | Country | |
---|---|---|---|
20060157788 A1 | Jul 2006 | US |