BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to semiconductor devices and more particularly to shallow trench isolation in self-aligned FET devices.
2. Description of Related Art
Conventional STI (Shallow Trench Isolation) planarization methods require a planarization mask, extensive measurements, and wafer-to-wafer process customization.
Maskless STI Planarization using Self-Aligned Polysilicon process allows STI to be planarized without a planarization mask, with minimal measurements and no need for wafer to wafer customization and can be designed to be completely ground rule compatible with a gate conductor stack Fill technology.
U.S. Pat. No. 5,173,439 of Dash et al. for "Forming Wide Dielectric-Filled Isolation Trenches In Semi-conductors" shows isolation trenches formed in pad silicon nitride/Si, followed by a silicon oxide layer, followed by a polysilicon layer, followed by Chemical Mechanical Polishing (CMP); followed by silicon oxide RIE (Reactive Ion Etching); followed by CMP down to a pad silicon nitride layer.
U.S. Pat. No. 5,504,033 of Bajor et al. for "Method for Forming Recessed Oxide Isolation Containing Deep And Shallow Trenches" describes in a fifth embodiment, after isolation trenches are dug into silicon nitride over silicon oxide over silicon, the trenches along with horizontal surfaces are first layered with silicon oxide, then sequentially a polysilicon layer; is followed by CMP. This reference appears not to address selective RIE partial silicon oxide etch over pad areas.
U.S. Pat. No. 5,411,913 of Bashir et al. for "Simple Planarized Trench Isolation and Field Oxide Formation using Poly-silicon" where after trenches are dug into pad silicon oxide/silicon nitride and into silicon; a layer of silicon oxide is deposited, followed by a layer of polysilicon, followed by RIE etch back to planarize. This reference appears not to address CMP polishing down to the silicon oxide layer and selective RIE partial silicon oxide etching over pad areas.
U.S. Pat. No. 5,492,858 of Bose et al. for "Shallow Trench Isolation Process Method for High Aspect Ratio Trenches" in which, after isolation trenches are dug through the pad silicon oxide/silicon nitride into silicon, a layer of silicon oxide is deposited and CMP etched back to silicon nitride to provide a planar surface for "active mesa sites." Bose et al. does not address deposition of polysilicon on a silicon oxide layer; CMP polishing down to the silicon oxide layer; and selective RIE partial silicon oxide etching over pad areas.
U.S. Pat. No. 5,494,857 of Cooperman et al. for "Chemical Mechanical Planarization of Shallow Trenches in Semiconductor Substrates" where, after trenches are dug in through pad silicon nitride/silicon oxide, a first silicon oxide layer is deposited, followed by a first silicon etch stop layer, followed by a second silicon oxide layer; followed by a CMP down to pad silicon nitride.
U.S. Pat. No. 5,252,517 of Blalock et al. for "Method Of Conductor Isolation From A Conductive Contact Plug", where, after transistors are completed, a "planarizing insulator layer" is deposited and contact vias are etched down to diffusion areas and filled with polysilicon.
U.S. Pat. No. 5,358,884 of Violette for "Dual Purpose Contact Collector Contact and Isolation Scheme for Advanced BICMOS Processes" shows trenches are dug through silicon nitride into silicon. Silicon oxide is deposited upon the silicon nitride, and CMP is done down to silicon nitride to create a "plurality of mesas."
FIG. 3 shows an isolation region of a prior art MOSFET device 60 with a doped silicon semiconductor substrate 62 on which an STI region 72 has been formed. Above the STI region is formed a gate conductor stack 74 of fill layers comprising a polysilicon layer 64, a silicide layer 68, and a silicon nitride gate insulator layer 70.
See J.-Y. Cheng, T. F. Lei, T. S. Chao, D. L. W. Yen, B. J. Jin, and C. J. Lin "A Novel Planarization of Oxide-Filled Shallow-Trench Isolation" J. Electrochem. Soc., Vol. 144, No.1, (January, 1997) pp. 315-320.
SUMMARY OF THE INVENTION
Maskless STI (MSTI) Planarization for gate conductor Fill Technology is accomplished by designing the AA (active area) mask with dummy active silicon mesas within the holes of gate conductor punch-holes. These dummy active silicon mesas are designed with the same ground rules as the rest of the chip.
Dash et al. discusses STI planarization using polysilicon but differs from the present invention in several ways with respect to the level of the silicon oxide fill and the level of the polysilicon. The silicon oxide RIE used after polysilicon CMP does not include the required break-thru step and stops on the silicon nitride instead of in the silicon oxide. Both the break-thru step and having the RIE stop in the silicon oxide were found to be essential in creating a manufacturable process.
None of the patents discussed above describes the active area (AA) fill concept.
In accordance with this invention, a method is provided for manufacture of a semiconductor FET device employing a Shallow Trench Isolation (STI) comprising the following steps. Provide a doped silicon semiconductor substrate coated with a pad structure on the surface thereof. Form raised active silicon device areas and dummy active silicon mesas capped with pad structures from the doped silicon semiconductor substrate and the pad structure. Then deposit a conformal blanket silicon oxide layer on the device with conformal projections above the mesas. Deposit a conformal blanket polysilicon layer on the blanket silicon oxide layer with additional conformal projections above the mesas. Perform chemical mechanical polishing of the blanket polysilicon layer to remove the additional conformal projections until the silicon oxide layer is exposed over the pad structures. Then follows selective RIE partial etching of the conformal silicon oxide layer over the mesas, followed in turn by chemical mechanical polishing of the conformal blanket silicon oxide layer using pad silicon nitride as an etch stop and converting the silicon oxide layer into a planar silicon oxide layer.
Preferably, the pad structures are composed of silicon nitride; or the pad structures are composed of a lower layer of silicon oxide capped with an upper layer of silicon nitride.
It is also preferred that after the second chemical mechanical polishing in step one strips away the pad structures from the device and then forms gate oxide layers above the surfaces of the substrate exposed by stripping away the pad structures.
In addition, after formation of the gate oxide layer P-wells and N-wells are formed in the substrate beneath the gate oxide layer and the silicon oxide layer.
Following formation of the wells there is a step of blanket deposition of a gate conductor layer composed of a polysilicon sublayer and a silicide sublayer upon the device followed by blanket deposition of a dielectric layer, followed by patterning and etching of windows down to active device areas and dummy areas in the substrate followed by the step of formation of FET devices and dummy devices by ion implantation of a dose of source/drain dopant ions into the active device areas and the dummy areas in the P-wells and the N-wells.
In accordance with another aspect of this invention, a Shallow Trench Isolation (STI) semiconductor FET device comprises a doped silicon semiconductor substrate with raised active silicon device areas and dummy active silicon mesas capped with a gate oxide layer and the substrate being coated with a planarized silicon oxide layer elsewhere. There are P-wells and N-wells formed in the substrate beneath the gate oxide layers and the silicon oxide layer, and a gate conductor layer and a dielectric layer formed over the gate oxide layers and the silicon oxide layer patterned into active devices, and dummy devices.
In accordance with another aspect of this invention, a Shallow Trench Isolation (STI) semiconductor FET device comprises a doped silicon semiconductor substrate with raised active silicon device areas and dummy active silicon mesas capped with a gate oxide layer and the substrate being coated with a planarized silicon oxide layer elsewhere, a gate conductor layer and a dielectric layer formed over the gate oxide layers, and polysilicon and dielectric layers being formed above the silicon oxide layer which are then patterned into dummy devices surrounding the mesas providing a pattern of punch hole vias.
Preferably, the pad structure is stripped from the device and a gate oxide layer is formed above the surface of the substrate exposed as the pad structure is stripped away.
Preferably, FET devices with gate structure are formed on the surface of the device with gate conductor structures and dummy structures formed on the surface of the planar silicon oxide layer. "Gate Conductor (GC) stack Filli" over trenches, and etch of the fill to produce vias for vertical contacts to diffusion areas on active sites.
BRIEF DESCRIPTION OF THE DRAWINGS
The foregoing and other aspects and advantages of this invention are explained and described below with reference to the accompanying drawings, in which:
FIGS. 1A-1O illustrate a process of manufacturing a Shallow Trench Isolation (STI) device in accordance with this invention.
FIGS. 2A-2G illustrate a process of manufacturing a Shallow Trench Isolation (STI) device in accordance with this invention with deep trench capacitor structures.
FIG. 3 shows an isolation region of a prior art MOSFET device.
FIG. 4 shows a perspective view of the device in accordance with this invention with a dummy area in which the vias reach down into dummy regions where no active devices have been formed. The structure is otherwise the same as the device described in FIGS. 1A-10.
DESCRIPTION OF THE PREFERRED EMBODIMENT
FIGS. 1A-10 illustrate a process of manufacturing a Shallow Trench Isolation (STI) device 10 in accordance with this invention.
FIG. 1A shows the device 10 in an early stage of manufacture. The device 10 is formed on a P-doped silicon substrate 11 upon which pad silicon dioxide/silicon nitride pad layer segments 14/14' have been formed. The silicon nitride in pad layer segments 14/14' is to serve as an etch stop. In the pad layers segments 14/14' the silicon dioxide layer has a thickness from about 50 .ANG. to about 150 .ANG. and the silicon nitride layer has a thickness from about 1,000 .ANG. to about 1,500 .ANG.. The active area mask 15/15' has been formed on the surface of silicon nitride layer segments 14/14' to protect the pad silicon dioxide/silicon nitride pad layer segments 14/14' and the silicon mesas 12/12' formed from the substrate 11 beneath the mask 15/15' during etching of the pad layers 14/14' into the pattern of mask 15/15', and during subsequent etching through mask 15/15' to form a set of shallow trenches 9/9'/9" in substrate 11 to a depth H below the surface of the pad layer segments 14/14'. The depth H is from about 0.35 .mu.m to about 0.48 .mu.m below the upper surface of the pad layer segments 14/14'.
FIG. 1B shows the device 10 of FIG. 1A after the mask 15/15' has been stripped away from the device 10, leaving the substrate 11 with the raised (mesa) active areas 12, 12' covered with structures comprising pad layer segments 14, 14'. The space between the structures comprising pad layer segments 14, 14' is a width W from about 0.25 .mu.m to about 250 .mu.m.
FIG. 1C shows the device 10 of FIG. 1B after deposition of a silicon dioxide layer 22 (having a thickness from about 4,800 .ANG. to about 5,600 .ANG.) on the device 10 covering the shallow trenches 9/9'/9" and the structures comprising pad layer segments 14/14' and the mesas 12/12'. Next the silicon dioxide layer 22 on device 10 is covered by deposition of a blanket polysilicon layer 24 on silicon oxide layer 22. Layer 24 has a thickness from about 4,000 .ANG. to about 4,800 .ANG..
FIG. 1D shows the device 10 of FIG. 1C after CMP (Chemical Mechanical) Polishing of blanket polysilicon layer 24 down to those portions of silicon dioxide layer 22 which are exposed because they overlie the remaining portions of the pad layer segments 14/14' above the mesas 12/12'. Thus the only portions of silicon dioxide layer 22 which are exposed by the CMP step are those directly above the mesas 12/12' as the layer 24 remains in place aside from the mesas 12/12'.
FIG. 1E shows the device 10 of FIG. 1D after selective RIE partial etching of the exposed surface of the silicon dioxide layer 22 forming hollows 22' and 22" above the pad layer segments 14/14' above the mesas 12/12'. The etching removes a thickness from about 3,200 .ANG. to about 4,700 .ANG. of the silicon dioxide layer 22 over the remaining areas of pad layer 14/14'.
FIG. 1F shows the device 10 of FIG. 1E after removal of the remainder of the polysilicon layer 24 with a selective etchant which removes the polysilicon layer 24 while leaving the silicon dioxide structure 22 with hollows 22'/22" intact. In this case a thickness from about 200 .ANG. to about 4,300 .ANG. of polysilicon layer 24 is removed.
FIG. 1G shows the device 10 of FIG. 1F after a CMP process was used for about 50 seconds to about 70 seconds planarizing silicon dioxide layer 22 and clearing away the silicon nitride portions of pad layer segments 14/14'. Thus the CMP process leaves the surface of device 10 as a planarized surface of silicon dioxide layer 22.
FIG. 1H shows the device 10 of FIG. 1G after the silicon nitride etch stop and the silicon dioxide layers of the pad layer 14/14' have been stripped from device 10 leaving openings 24/24' (where layer 14/14' had been) in planar silicon dioxide layer 22 down to the surfaces of the mesas 12/12' exposed between the remaining portions of the silicon dioxide layer 22.
FIG. 1I shows the device 10 of FIG. 1H after "gate" sacrificial silicon dioxide gate segments 30/30' about 125 .ANG. thick have been formed above the mesas 12/12' by the conventional process of oxidation of the exposed surface of the substrate 11. Then V.sub.T implants are made through the sacrificial silicon dioxide gate segments 30/30' into the substrate 11.
In addition, an N-well mask 31" has been formed over the device 11 with a N-well window 31'" over the sacrificial silicon dioxide gate segments 30/30' through which N type dopant ions 31' are ion implanted into the surface of substrate 11 below the gate segment 30 to form an N-well 34.
FIG. 1J shows the device 10 of FIG. 1I after N-well mask 31" has been stripped away and a P-well mask 32" has been formed over the device 11 with a P-well window 32'" over the sacrificial silicon oxide gate segment 30' through which P type dopant ions 32' are ion implanted into the surface of substrate 11 below the gate segment 30 forming a P-well region 36.
FIG. 1K shows the device 10 of FIG. 1J after the sacrificial silicon oxide gate segments 30/30' have been stripped away by etching which also thins the silicon dioxide layer 22 into thinned planar oxide larger 22' to be coplanar with the surface of device 11. Then conventional gate silicon oxide (gate oxide) layer segments 38/38' (about 100 .ANG. thick) are formed on the surface of the mesas 12/12' within the recently enlarged openings 24/24'.
FIG. 1L shows the device 10 of FIG. 1K after the thinned planar oxide layer 22' and the gate oxide layer segments 38/38' have been coated with a doped polysilicon layer 40 preferably about 1000 .ANG. thick, with a thickness range from about 500 .ANG. to about 2,000 .ANG.. Polysilicon layer 40 is coated with a silicide layer 42 preferably a tungsten silicide layer about 800 .ANG. thick, with a thickness range from about 500 .ANG. to about 2,000 .ANG.. Tungsten silicide layer 42 is coated with a silicon dioxide or silicon nitride gate insulator layer 44 preferably about 2800 .ANG. thick, with a thickness range from about 2,000 .ANG. to about 4,000 .ANG..
Next, the device is coated with a photoresist gate stack mask 46 with openings 48A, 48B therethrough over the ends of gate oxide layer segment 38 over N-well 34 and opening 48C therethrough over gate oxide layer segment 38' P-well 36.
FIG. 1M shows the device 10 of FIG. 1L after the introducing the RIE etchant through openings 48A, 48B, and 48C down through gate insulator layer 44 etching openings 50A, 50B, and 50C therein extending down through tungsten silicide layer 42 and doped polysilicon layer 40 to expose the surface of the gate oxide layer segments 38/38' leaving a gate conductor stack 51 over N-well 34 with source/drain windows on either side and a dummy window 50C exposing the P-well for ion implanting subsequently, as indicated by arrows 58 in FIG. 10 which is described below.
FIG. 1N shows the device 10 FIG. 1M after ion implanting the P+ dopant source/drain regions 56S/56D below the silicon oxide segment 38 self-aligned with the gate conductor stack 51.
FIG. 10 shows the device 10 of FIG. 1N after ion implanting N+ dopant into N+ region 58 through the dummy opening 50C, as indicated by arrow 58, to form N+ region 58(N+) in P-well 36, below silicon oxide segment 38'.
By designing dummy active silicon mesas in STI regions within the gate conductor stack by filling punch-hole areas, a maskless STI (MSTI) planarization process previously can be realized on products incorporating STI and gate conductor stack Fill technologies.
FIGS. 2A-2G illustrate a process of manufacturing a Shallow Trench Isolation (STI) device 10 in accordance with this invention with deep trench capacitor structures 16. Corresponding structures in FIGS. 2A-2G are the same as those in FIGS. 1A-1G and the descriptions thereof apply to a device 10 which includes the deep trench capacitor structures 16 initially at the beginning of the process of manufacture.
FIG. 4 shows a perspective, sectional view of a portion of the device 10 in accordance with this invention with a dummy area in which the vias 54 reach down into dummy regions where no active devices have been formed. The structure is otherwise the same as the device described in FIGS. 1A-10. The device 10 is formed on the P-doped silicon substrate 11 upon which silicon mesas 12 have been formed from the substrate 11 between the recesses therein containing the shallow trenches filled with the silicon oxide regions 22 upon which dummy gate conductor stacks of polysilicon layer 40, silicide layer 42 and the silicon dioxide or silicon nitride dielectric layer 44 have been formed. Between the dummy conductor stacks are the vias 54 (punch holes) which extend down to the top of the mesas 12.
Maskless STI Planarization Using Self-Aligned Polysilicon.
For the ideally scalable ULSI CMOS device, process control of the isolation technology is crucial. Device isolation needs to provide an abrupt active-to-isolation transition with sufficient isolation depth and to provide a planar wafer surface. This must be achieved with a wide process window at a low cost.
Several alternative planarization techniques have been proposed. For example, LOCOS is inexpensive but suffers from insulator thinning at narrow dimensions, bird's beak formation, field-implant encroachment, and creates significant wafer topography. Poly-Buffered LOCOS and Poly-encapsulated LOCOS improved the bird's beak formation but still result in a narrow channel effect that increases device Vts.
Shallow Trench Isolation (STI) provides an abrupt active-to-isolation transition without bird's beak formation with a minimum impact on device characteristics or topography. However, the process often requires extensive measurements and wafer to wafer process customization to control, i.e., a resist planarization mask used for fabrication of a 16 Mb DRAM, and has a higher cost than LOCOS based methods.
A manufacturable STI planarization process using Self-Aligned polysilicon and a planarization mask provides a stable and reliable process with a robust process window. It does not require extensive inline measurements or wafer to wafer process customization to control.
The Self-Aligned Poly-silicon planarization process can be greatly simplified by use of mesas of active silicon within the STI regions in accordance with this invention. This allows the planarization mask, CMP stop silicon oxide layer deposition, and the CMP stop silicon oxide layer etch to be completely eliminated. This is compatible with a gate conductor stack with fill technology by using a `punch-hole` gate conductor stack mask. This allows large area STI regions to have active silicon mesas placed within the gate conductor stack-fill `punch-hole`areas, thus creating additional polish stop areas to prevent CMP dishing in the STI regions during planarization.
Maskless STI Planarization Using Self-Aligned Polysilicon.
Process
1. Substrates start with an STI of depth "H" with maximum width W. "H" depends on device design requirements. "W" depends on planarization distance of CMP pad used for polishing sacrificial polysilicon (.about.30-50 .apprxeq..mu.m for an IC1000 CMP Pad);
2. STI shapes are designed to have active silicon mesas placed within the gate conductor stack-Fill "punch-hole" areas;
3. Deposition of silicon oxide layer to be planarized of thickness "H"+.apprxeq.25% Ox; over initial pad structure (e.g. silicon nitride);
4. Deposition of polysilicon sacrificial layer of thickness .apprxeq."H" over silicon oxide layer;
5. CMP polysilicon layer stopping on silicon oxide layer;
6. Using a combination of non-selective & selective silicon oxide/polysilicon RIE, etch the silicon oxide over the initial pad structure to be coplanar with the top of the silicon oxide in the large STI areas. The RIE will leave a thin layer of polysilicon over the STI regions;
7. Selectively strip remaining polysilicon.
8. CMP remaining silicon oxide down to initial pad structure.
Supporting Data
Cost of Ownership of STI Planarization Using Self-Aligned Poly-silicon
The cost has been calculated to decrease from about $145 to about $117 for the STI planarization module using MSTI.
Operations Summary
Much development effort was invested in creating a manufacturable STI planarization process using self-aligned polysilicon for this process. Data was collected and analyzed from hundreds of integrated product lots showing that the method of this invention for STI planarization creates a stable process with a large process window.
Simplifying the self-aligned polysilicon STI planarization process by eliminating a planarization mask allows the polysilicon CMP stop silicon oxide layer and polysilicon CMP stop silicon oxide etch processes to be eliminated also.
By designing active silicon mesas within large STI regions (within the gate conductor stack-Fill punch-hole areas in a gate conductor stack-Fill technology) and limiting the largest STI width to the planarization distance of CMP pad used for polishing the sacrificial polysilicon (-30-50 .apprxeq..mu.m for an IC1000 CMP pad), MSTI can be easily implemented.
There is a gate conductor (GC) stack with fill over trenches, and the fill is etched to produce vias for vertical contacts to diffusion areas on active sites.
While this invention has been described in terms of the above specific embodiment(s), those skilled in the art will recognize that the invention can be practiced with modifications within the spirit and scope of the appended claims, i.e. that changes can be made in form and detail, without departing from the spirit and scope of the invention. Accordingly all such changes come within the purview of the present invention and the invention encompasses the subject matter of the claims which follow.