Semiconductor device manufacturing: process – Making passive device – Trench capacitor
Reexamination Certificate
1999-02-25
2003-12-23
Utech, Benjamin L. (Department: 1765)
Semiconductor device manufacturing: process
Making passive device
Trench capacitor
C438S398000, C438S745000, C438S753000
Reexamination Certificate
active
06667218
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to semiconductor technology and, more specifically, to the design of capacitors used in dynamic random access memory cells.
BACKGROUND OF THE INVENTION
The invention uses various materials which are electrically either conductive, insulative or semi-conductive, although the completed circuit device itself is usually referred to as a “semiconductor.”
The memory cells of dynamic random access memories (DRAMs) are comprised of two main components: a field-effect transistor and a capacitor. In DRAM cells utilizing a conventional planar capacitor (such as the one depicted in FIG.
1
), far more chip surface area is dedicated to planar capacitor
11
than to field-effect transistor (FET)
12
. The gate
13
of FET
12
and the word line
14
are formed from an etched polycrystalline silicon layer. Bit line
15
connects with access-node junction
16
. Capacitor
11
has a lower plate formed from the n+ silicon substrate extension
17
of storage node junction
18
of FET
12
. Upper capacitor plate (or field plate)
19
is formed from a layer of conductively-doped polycrystalline silicon. Substrate extension
17
is electrically insulated from upper plate
19
by a dielectric layer
20
. Planar capacitors have generally proven adequate for use in DRAM chips up to the one-megabit level. However, planar capacitors constructed with conventional dielectric materials appear to be unusable beyond the one-megabit DRAM level. As component density in memory chips has increased, the shrinkage of cell capacitor size has resulted in a number of problems. Firstly, the alpha-particle component of normal background radiation will generate hole-electron pairs in the n+ silicon substrate plate of a cell capacitor. This phenomena will cause the charge within the affected cell capacitor to rapidly dissipate, resulting in a “soft” error. Secondly, as cell capacitance is reduced, the sense-amp differential signal is reduced. This aggravates noise sensitivity and makes it more difficult to design a sense-amp having appropriate signal selectivity. Thirdly, as cell capacitance is decreased, the cell refresh time must generally be shortened, thus requiring more frequent interruptions for refresh overhead. The difficult goal of a DRAM designer is therefore to increase or, at least, maintain cell capacitance as cell size shrinks, without resorting to processes that reduce product yield or that markedly increase the number of masking and deposition steps in the production process.
Several methods for providing adequate cell capacitance in the face of shrinking cell size are either in use or under investigation. Basically, the efforts fall into two categories. Efforts within the first category are aimed at creating complex three-dimensional capacitors; those within the second are aimed at improving the dielectric of the planar capacitor.
One three-dimensional technique involves the creation of “trench” capacitors in the cell substrate.
FIG. 2
depicts a DRAM cell having a typical trench capacitor
21
. Similar in concept to planar capacitor
11
of
FIG. 1
, the trench is employed to provide greater plate area, and hence, greater capacitance. The lower plate
22
may be formed from the n+ doped silicon substrate or it may be formed from a polysilicon layer which lines a trench cut in the n+ doped silicon substrate. The upper plate
23
is formed from a layer of conductively-doped polycrystalline silicon (poly). Lower plate
22
and upper plate
23
are electrically insulated from each other with a dielectric layer
24
. DRAM chips employing trench capacitors have been built by a number of European, Japanese and U.S. companies, including IBM Corporation, Texas Instruments, Inc., Nippon Electric Company, Toshiba, Matsuchita and Mitsubishi Electric Corporation. There are several problems inherent in the trench design, not the least of which is trench-to-trench capacitive charge leakage which is the result of a parasitic transistor effect between trenches. Another problem is the difficulty of completely cleaning the capacitor trenches during the fabrication process; failure to completely clean a trench will generally result in a defective cell.
Another three-dimensional technique, which is being used by most DRAM manufactures including, Micron Semiconductor, Nippon Electric Company, Samsung, Goldstar, Hyundai, Mitsubishi Electric Corporation, Hitachi, and Fujitsu, Ltd., is the stacking of capacitor plates between dielectric layers on the DRAM cell surface.
FIG. 3
is a graphic representation of a typical DRAM cell having a stacked capacitor
31
. The lower plate
32
is formed from an n-type polycrystalline silicon layer which is in contact with the silicon substrate
33
in the region of the FET storage node junction, while the upper plate
34
is formed from a conductively-doped polycrystalline silicon layer. The two layers are separated by a dielectric layer
35
. Lower or storage node plate
32
and upper plate
34
are both stacked on top of FET
12
and word line
36
, resulting in a high-profile cell which requires more stringent process control for the connection of bit line
37
to access-node junction
38
.
In one variation of the stacked capacitor, which is currently being used by National Electric Company, Micron, Samsung, Matsushita, and other DRAM manufacturers, the storage node plate of the stacked capacitor is a rough polysilicon layer called hemispherical grain (HSG) polysilicon. This layer is formed at a critical temperature and pressure at which an anomalous nucleation occurs, causing the surface to roughen. The HSG polysilicon provides a much larger surface area than planar poly. However the benefits of HSG polysilicon are not fully utilized because the grains, as shown in an enlarged exaggerated cross-sectional view in
FIG. 4
of deposited HSG polysilicon
47
, are so close together that the dielectric layer
48
deposited to overlie the HSG polysilicon
47
bridges between the grains. The dielectric layer
48
in the bridged area is often as thick as 400 angstroms. Therefore a method is needed to reduce the bridging of the dielectric
48
between the grains of the HSG polysilicon
47
while maintaining the increased capacitive area provided by the HSG polysilicon
47
.
SUMMARY OF THE INVENTION
The present invention is applicable to DRAM cell designs, such as the stacked capacitor design heretofore described or a poly-lined trench design, that utilize a conductively-doped polycrystalline layer for the storage node, or lower capacitor plate. The invention is a method for forming HSG polysilicon with reduced dielectric bridging. A doped first polysilicon layer is deposited. A second polysilicon layer is then deposited to overlie the first polysilicon layer. The temperature and pressure of the second polysilicon layer is selected in a range wherein HSG is formed during the deposition. Thus the surface of the second polysilicon layer is roughened as a result of nucleation. Next a wet etch is performed. The aggressiveness of the wet etch is controlled to remove portions of the rough polysilicon and portions of the base polysilicon while retaining a roughened surface and maintaining a large surface area of the storage node capacitor plate. The size of the grains decreases during the wet etch and the distance between the grains increases. A dielectric layer is deposited to overlie the rough polysilicon following the wet etch. Bridging of the dielectric is reduced resulting in the dielectric layer having a uniform thickness over the entire surface of the storage node capacitor plate. Since bridging is reduced the thickness of the dielectric layer in areas formerly bridged is reduced and capacitance is increased.
REFERENCES:
patent: 5130885 (1992-07-01), Fazan et al.
patent: 5134086 (1992-07-01), Ahn
patent: 5162248 (1992-11-01), Dennison et al.
patent: 5170233 (1992-12-01), Liu et al.
patent: 5227322 (1993-07-01), Ko et al.
patent: 5256587 (1993-10-01), Jun et al.
patent: 5278091 (1994-01-01), Fazan et al.
patent: 5304828 (1994-04-
Deo Duy-Vu
Micro)n Technology, Inc.
Schwegman Lundberg Woessner & Kluth P.A.
Utech Benjamin L.
LandOfFree
Enhancing semiconductor structure surface area using HSG and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Enhancing semiconductor structure surface area using HSG and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Enhancing semiconductor structure surface area using HSG and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3139994