US 5,602,049 AGrant
Method of Fabricating a Buried Structure SRAM Cell
Issue Date:1997-02-11
•9 Claims
•4 Drawing Sheets
Abstract
An improved SRAM cell having ultra-high density and methods for fabrication are described. Each SRAM cell, according to the present invention, has its own buried structure, including word lines (i.e., gate regions) and bit lines (i.e., source/drain regions), thus increasing the cell ratio of channel width of cell transistor to that of pass transistor to keep the data stored in the cell transistor more stable without increasing the area per cell. In addition, according to the present invention, the field isolation between active regions is not field oxide but blankly ion-implanted silicon substrate. Therefore, SRAM cells can be densely integrated due to the absence of bird's beak encroachment. Since the present invention has more planar topography, it is easily adapted to the VLSI process, which is always restricted by the limit of resolution of photolithography, thus increasing the degree of integration.
Metadata
Assignee
- United Microelectronics Corporation
Inventors
- Jemmy Wen
- Joe Ko
Application Information
Application Number:US 3229394
Filing Date:1994-10-04
Priority Date:1994-10-04
Art Unit:117
Classifications
IPC:
H01L 2170H01L 2700
Field of Search:
43725747;52;48;60;203;919903-904
Patent Drawings (4 sheets)
Description
Field of the Invention
As indicated in FIG. 1, a 4T SRAM cell comprises two pass transistors 1, two cell transistors 2 and a poly-load 3. In order to keep the data stored in the cell transistors 2 more stable, it is always needed to raise the ratio of a channel width of the cell transistors to that of the pass transistors (it is briefly called "cell ratio", hereafter). For example, a process of fabricating a 64 k CMOS SRAM will be briefly described below. After the formation of a P well 121 and the implantation of a Vt-adjusted ion, the growth of a field oxide 24 proceeds with the formation of a N well 20 simultaneously, as indicated in FIG. 2A. Then, it is sequentially preformed to grow a gate oxide 321 by thermal oxidation, to form a buried contact 51, and to deposit a first polycrystalline silicon layer 21 which fills said buried contact 51. The impurity-doped first polycrystalline silicon layer 21 is, in turn, subject to photoing and etching to form a gate of a MOSFET, as indicated in FIG. 2B. Subsequently, N+/P+ source/drain regions are traditionally formed by ion implantation. Over an entire exposed surface, an inter-poly dielectric layer 61 is deposited and vias 31 are, in turn, formed by photoing and etching. Thereafter, a second polycrystalline silicon layer is deposited and subject to photoing and etching to form a poly-load resistor 4. Finally, a Boro-Phosphosilicate Glass (BPSG) 8 is deposited, as indicated in FIG. 2C. The remaining processes are standard back-end CMOS processes, which include the formation of a contact window, the patterning and etching of a metal layer, and a passivation layer respectively etc.
As MOSFET devices are scaled to submicrometer dimension, their size will become very small and make channel length and width shrink. Thus, it results in a low punch-through voltage, threshold-voltage shift, leakage current, etc. Further, it will be more serious in SRAM cells because it is always necessary to reduce the channel width of the pass transistors for the purpose of raising the cell ratio of SRAM cells. So, the prior art can't raise the cell ratio without deteriorating its electrical characteristics. In addition, the prior art, as shown in FIG. 2C, has a more complex topography; thereby, the devices can't be more densely packed due to the limit of resolution of photolithography. Besides, they have the field oxide between active regions in the prior art, causing extra expenditure of the chip area due to bird's beak encroachment. Therefore, it is needed to provide a method of raising the cell ratio without increasing the area per cell.
Summary of Invention
Therefore, an object of the invention is to provide a buried structure SRAM cell and methods for fabrication which have a higher cell ratio without needing extra area, thereby increasing the integration degree. Another object of the invention is to provide a buried structure SRAM cell and methods for fabrication; therein the channel width of a MOSFET can be controlled without increasing the area per cell.
Another object of the invention is to provide a fieldless buried structure SRAM cell and methods for fabrication which make the SRAM cell additionally packed.
A further objective of the invention is to provide a buried structure SRAM cell and methods for fabrication having more planar topography, which also adds to the overall cell density.
Each SRAM cell according to the present invention has its own buried structure including word lines (i.e. gate regions) and bit lines (i.e. source/drain regions), thus increasing the ratio of the channel width of the cell transistors to that of the pass transistors to keep the data stored in the cell transistors more stable without increasing the area per cell. In addition, according to the present invention, the field isolation between active regions is not a field oxide but blankly an ion-implanted silicon substrate, so the SRAM cell can be integrated with high density due to the absence of bird's beak encroachment. Since the present invention has more planar topography than the prior art, it is easily adapted to the VLSI process which is always restricted by the limit of resolution of photolithography, thus increasing the degree of integration.
Brief Description of the Drawings
The invention itself, however, as well as other features and advantages thereof, will be best understood by reference to the detailed description which follows, read in conjunction with the accompanying drawings, wherein:
FIG. 1 is a 4T (Transistor) SRAM cell;
FIG. 2A through 2C are schematic cross-sectional views of several crucial steps in a conventional 64 K SRAM CMOS process;
FIG. 3A through 3G, 3G', 3G" and 3H are schematic cross-sectional views of several crucial steps according to the present invention;
FIG. 4 is a lay out of a MOSFET manufactured according to the present invention;
FIG. 5 is a cross-sectional view from the A-A' line of FIG. 4;
FIG. 6A is a cross-sectional view from the B-B' line of FIG. 4, wherein the MOSFET is used as a pass transistor; and
FIG. 6B is a cross-sectional view from the B-B' line of FIG. 4, wherein the MOSFET is used as a cell transistor.
Description of the Preferred Embodiment
Although the operating principle and the fabricating process of DRAM cells are quite different to that of SRAM cells, there are some similarities between these two devices. For example, in order to reduce the area per call and yet maintain the capacitance of a capacitor, most commercial DRAM cells have three dimensional trenched capacitors. A wide range of capacitor-trenched DRAM cells have been proposed; such as, see Baglee, et al., U.S. Pat. No. 4,271,987, "Trench Capacitor Process for High Density Dynamic RAM", issued Jan. 26, 1988. Similarly, the principle goal of the invention is to raise the cell ratio (it is defined in the beginning of this article) without needing extra area occupied by the cells, thus, a buried-gate MOSFET structure is now proposed to substitute the conventional diffusion structure.
A process for fabricating buried structure SRAM cells is disclosed. First of all, referring to FIG. 3A, a well (indicated by dashed line) of a predetermined conductivity type (i.e. N Type or P Type) is first formed on a silicon substrate 10. Then, the entire substrate 10 is subject to a thermal oxidation cycle to grow a pad oxide 11 to a thickness of approximately 800 Angstroms. A layer of photoresist is then patterned and the pad oxide 11 is etched to provide source/drain trenches 12 by anisotropic etching to a depth of 1-5 .mu.m. This anisotropic etching may be accomplished by a suitable plasma source of etchants and ions. After the removal of photoresist, walls of the source/drain trenches 12 are then doped with arsenic for the N well or Boron for the P well to raise the punch-through voltage between the bit lines using one of many processes (not shown in FIG. 3A). For example, an angled ion implantation with impurity ions at an angle of 8.degree. while rotating the substrate 10 may be performed. Other suitable doping techniques such as deposition of doped materials and diffusion of dopants into the substrate may suitably be used. The walls of the source/drain trenches 12 plus the silicon substrate between the source/drain trenches 12 doped simultaneously are used to replace the field oxide for the field isolation of active regions. So, there is no bird's beak encroachment problem in the invention, thereby increasing the cell density. Subsequently, the entire substrate 10 is subject to a thermal cycle to grow a silicon dioxide layer at the walls of the source/drain trenches 12 to a thickness of approximately 100 to 600 Angstroms, which should be thinner than the pad oxide 11.
A layer of polycrystalline silicon 21 is then deposited by low pressure chemical vapor deposition of silane to a thickness of approximately 2000 to 4000 Angstroms. By the use of etching back technique, the polycrystalline silicone layer 21 filled in the source/drain trenches 12 is left to a depth of approximately 0.5 to 2 micrometer, and then a step of wet etching follows for a period of a short time to remove the silicon dioxide layer of the trench walls to form a residual silicon dioxide layer 131, as shown in FIG. 3B.
There is some thickness of the pad oxide 11 left during said wet etching to be used as the end point detection; and this is why the silicon dioxide layer of the trench walls should be thinner than the pad oxide. Simultaneously, a residual silicon dioxide 131 is formed at the bottom of the source/drain trenches 12 to be used as a blocking layer to prevent leakage of the current from the bit lines, also shown in FIG. 3B. Thereafter, a layer of polycrystalline silicon 21 is again deposited to a thickness of approximately 2000 to 4000 Angstroms and is doped by in situ doping to provide reduction of resistivity and a diffusion source of impurity. In order to form a shallow P-N junction, the polycrystalline silicone layer 21 is doped with BF2 in the case of the N well, or As75 in the case of the P well because of their low diffusivity in silicon. And, this shallow P-N junction provides the prevention of punch-through between the bit lines, thereby providing a greater integration degree. Then, a drive-in step proceeds to form said shallow P-N junction. Subsequently, the polycrystalline silicon layer 21 is again etched back by reactive ion etching (RIE) while the pad oxide 11 is used as the end point detecting during etching. The resultant structure is shown in FIG. 3C.
Then, the pad oxide 11 is removed by wet etching. A layer of photoresist is, in turn, patterned and the polycrystalline silicon layer 21, and the silicon substrate 10, are etched to provide gate trenches 311, as shown in FIG. 3D. The depth of the gate trenches 311 is approximately 0.4 to 1.8 micrometer. It should be mentioned that there must have some distance between the bottom of the gate trenches 311 and the top of the residual silicon dioxide 131, to form a channel region.
Thereafter, a series of processes controlling the channel width of the pass transistors follows. A SOG (Spin On Glass) insulating layer 312 is deposited to a thickness of approximately 500 to 4000 Angstroms to fill the gate trenches 311, as shown in FIG. 3E. The thickness of the SOG insulating layer 312 is used to control the channel width of the pass transistors. A photoresist layer is then patterned to leave only the pass transistors region exposed; and the resultant entire substrate is, in turn, subject to an ion implantation with the impurity having the same conductivity type as the well region to form a channel stop region 3111. Simultaneously, the SOG insulating layer 312 and its thickness provide an implantation mask for formation of the channel region of the pass transistors at the bottom and corners of the gate trenches 311. The resultant structure is also shown in FIG. 3E. The photoresist layer is then removed using standard photoresist removal processes such as wet removal or ashing. Subsequently, the SOG insulating layer 312 and the pad oxide 11 are simultaneously removed by wet etching. Then, a process of two-steps ion implantation is used to adjust the threshold voltage of the pass transistors and the cell transistors; wherein, the two-steps ion implantion means that one angle is 0.degree., that is, normal to the substrate and the other is approximate 15.degree. to 60.degree., depending on the width and depth of the gate trenches 311. A silicon dioxide layer is then formed by thermal oxidation to a thickness of approximately 80 to 200 Angstroms and used as a gate insulating layer 321. Also, there grows a thicker silicon dioxide layer at a surface of the bit-line polycrystalline silicon, simultaneously, because of the higher rate of oxidation of the polycrystalline silicon. The resultant structure is shown in FIG. 3F.
A layer of photoresist is then patterned and the gate insulating layer 321 is etched to provide a buried contact 51, designated G point shown is FIG. 1. FIG. 3G' is a cross-sectional view along E-E' line of FIG. 3G, wherein the buried contact 51 is the region where the bit lines (i.e. the first polycrystalline silicon) and a polyresistor can be connected. Besides, FIG. 3G" is a cross-sectional view along D-D' line of FIG. 3G.
After the removal of the photoresist layer, a second polycrystalline silicon layer is deposited to a thickness of over-flowing the gate trenches 311 and is doped by in situ doping with phosphorus impurity for a N MOSFET, or with Boron impurity for a P MOSFET. Using the silicon dioxide located at a top surface of the substrate 10, as the end point detection, the second poly crystalline silicon layer is etched back to form a gate electrode 4 of the MOSFET. An inter-poly insulating film 61 is then deposited by chemical vapor deposition to provide electrical insulating between the two polycrystalline silicon layer; a layer of photoresist is, in turn, patterned to provide the vias 62 connecting the gate electrode 4 and the poly-load. Over the entire exposed substrate a third polycrystalline silicon layer 71 is deposited and, in turn, subject to photoing and etching to form the poly-load of the SRAM cells, designated 3, as shown in FIG. 1. The remaining processes are conventional back-end CMOS processes; that is, a sequential process of depositing, photoing, and etching of an insulating layer 8, a metal layer 81, and a passivation layer respectively. The resultant structure is shown in FIG. 3H.
With reference now to FIG. 4, there is shown a lay-out of a MOSFET and it may be a pass transistor or a cell transistor in the present invention. Wherein the source/drain regions (bit lines) and gate regions (word lines) are shown by single-declined lines and double declined lines respectively. FIG. 5 is a cross-sectional view from the A-A' line of FIG. 4; wherein, there is some suitable distance between the gate insulating layer 321 and the residual silicon dioxide 131 to provide the formation of the channel region: otherwise when a inversion layer occurs beneath the gate insulating layer 321, a carrier will be blocked by the residual silicon dioxide 131 and can't reach the source/drain regions, therefore, the MOSFET would lose its function. Besides, the source and drain regions (the bit lines) are formed by the first polycrystalline silicon 21, not the diffusion region in the prior art, thereby forming the shallow junction due to the low diffusivity of impurity doped in the first polycrystalline silicon 21, which allows the invention having a greater packing density. Also, there is the residual silicon dioxide 131 surrounding most the source/drain regions to prevent the leakage of current from it, which results in the raising of the punch-through voltage.
FIG. 6A is a cross-sectional view from the B-B' line of FIG. 4; wherein the MOSFET is used as a pass transistor of the SRAM cells. It should be mentioned that the carrier will only pass the bottom of the gate trenches 311. Now, suppose that the depth and width of the gate trenches 311 have L and X value respectively, as indicated in FIG. 6A. The effective channel width of the pass transistor would be restricted to X value by the channel stop region 3111. In the other cases, shown in FIG. 6B in which the MOSFET is used as a cell transistor, the effective channel width of the cell transistor would be 2L+X value, due to not having the channel stop region 3111 at a side-wall of the gate trenches 311. It is evident that the channel width of the cell transistor or the pass transistor and their ratio (i.e. cell ratio) can be easily controlled according to the invention without needing extra area, thereby increasing the stability of the data stores in the cell transistor; in other words, making the data read from the SRAM cells more stable without distortion.
Beside the several advantages disclosed above, compared with the prior art, there are other advantageous features of the invention, as listed below. (1) There remains a self-aligned buried gate structure, because some of the first polycrystalline silicon 21 (i.e. the source/drain regions) is etched to provide a part of the gate trenches 311 and the depth of the gate trenches is much more shallow than that of the source/drain regions, as evident from FIG. 5. Thus, the invention also reserves a high cell density of the self-aligned process due to no aliment tolerances between the gate and source/drain regions. (2) The SRAM cells according to the invention have more planar topography due to the buried source/drain regions and the gate regions. Since the prior art, as shown in FIG. 2C, has a more complex topography the accurance of exposure and developing of photoresist, especially at a recessed region, is not easily controlled due to the limit of resolution of photolithography. In the invention, the above-mentioned problem will not be encountered, therefore, increases the integration degree. (3) There is no field oxide between active regions. In the invention, a blank ion implantation for rasing the punch through voltage is used to substitute the field oxide as field isolation in the prior art. So, the expenditure of the chip area, due to bird's beak encroachment encountered in the prior art, will not happen in the invention, thereby providing a greater packing density.
Although the preferred embodiment of the invention has been disclosed for illustrative purposes, those skilled in the art will appreciate that various modifications, additions and substitutions are possible, without departing from the scope and spirit of the invention, as disclosed in the accompanying claims.
Claims
We claim:
1. A method of fabricating a buried structure SRAM cell comprising the steps of: forming a well of a first conductivity type in a silicon substrate; forming a pad oxide over said substrate; masking and then etching said pad oxide to form source/drain trenches in said substrate; introducing dopants with said first conductivity type into said substrate to raise punch-through voltage between source and drain trenches; forming a silicon dioxide layer at a surface of said source/drain trenches; filling said source/drain trenches and covering an entire substrate with a first polycrystalline silicon layer formed by deposition method; etching back said first polycrystalline silicon layer to form recesses at said source/drain trenches; partially removing said silicon dioxide layer to form a residual silicon dioxide layer by a wet etching and then filling said recesses and covering an entire substrate with a doped first polycrystalline silicon layer, wherein dopants being of a second conductivity type to form a P-N junction by diffusion; etching back said doped first polycrystalline layer to form source/drain regions of a MOSFET in said recesses using said pad oxide as end point detection; removing said pad oxide, masking said substrate and then etching said doped first polycrystalline silicon layer and substrate to form gate trenches; filling said gate trenches with a thickness of a dielectric layer to control a channel width of a pass transistor of SRAM cells; masking said gate trenches so as to leave only the gate trenches of said pass transistor exposed and then implanting dopants of said first conductivity type to form a channel stop region at a wall of said gate trenches of said pass transistor; removing said making and dielectric layer filled on gate trenches, then ion-implanting the gate trenches to adjust a threshold voltage of said MOSFET; then thermally growing a gate oxide over an entire surface of the substrate followed by masking and etching the gate oxide to form a buried contact region; depositing a doped second polycrystalline silicon to fill said gate trenches and cover the entire substrate; etching back said doped second polycrystalline silicon to form a gate electrode of said MOSFET, while the gate oxide on the surface of the substrate being used as end point detection; depositing an inter-poly dielectric layer; masking and etching said inter-poly dielectric layer to form vias in which poly-loads can be connected to the gate electrode and source/drain regions; and depositing a third polycrystalline silicon over the substrate and then patterned to form poly-loads of SRAM cells.
2. The method according to claim 1 wherein the silicon dioxide layer is thermally grown silicon dioxide.
3. The method according to claim 1 wherein the silicon dioxide layer at the surface of the source/drain trenches is thermally grown silicon dioxide and its thickness is thinner than that of the pad oxide.
4. The method according to claim 1 wherein the source/drain trenches and gate trenches are formed by reactive ion etching.
5. The method according to claim 1 wherein the doped first polycrystalline silicon is doped with BF.sub.2 in case of N well, or As.sub.75 in case of p well to form a shallow P-N junction due to the low diffusivity of BF.sub.2 and As.sub.75 in silicon.
6. The method according to claim 1 wherein a doping process of the first and second doped polycrystalline silicon is performed by one of in situ doping and ion-implantation methods.
7. The method according to claim 1 further comprising a step of forming a channel region having a distance between the gate oxide and residual silicon dioxide layer to prevent the blockage of carriers by the residual silicon dioxide layer from reaching the source/drain region when the inversion layer occurs beneath gate oxide.
8. The method according to claim 1 wherein the step for ion-implanting the gate trenches to adjust the threshold voltage of said MOSFET includes two angled steps of implantation; that is, one is 0.degree., and the other is about 15.degree. to 60.degree. from normal to the substrate depending on the width and depth of the gate trenches.
9. The method according to claim 1 further comprising of the deposition of a Boro-Phosphosilicate Glass (BPSG) as an insulating layer, the formation of a contact window, the patterning and etching of a metal layer and a passivation layer respectively.