Semiconductor device manufacturing: process – Making field effect device having pair of active regions... – Having insulated gate
Reexamination Certificate
1999-03-11
2001-10-02
Whitehead, Jr., Carl (Department: 2822)
Semiconductor device manufacturing: process
Making field effect device having pair of active regions...
Having insulated gate
C438S248000, C438S249000, C438S386000, C438S391000, C438S392000
Reexamination Certificate
active
06297086
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to an improved method of fabricating a semiconductor memory device, and in particular to an improved method of fabricating a deep trench dynamic random access memory (DRAM) capacitor structure wherein an excimer laser anneal step is employed in activating the gate doping impurities and the source/drain diffusion regions of the structure. Under appropriate conditions, the excimer laser anneal step may substantially reduce the extent of decomposition of any temperature sensitive high dielectric constant material that may be used in fabricating the DRAM trench capacitor.
BACKGROUND OF THE INVENTION
In dynamic random access memory (DRAM) cell manufacturing, a key issue is the size of the overall cell. As the integration densities increase, it is desirable in the semiconductor industry to decrease the storage capacitor size while maintaining the charge storage capacity. One approach to this problem in the prior art is to utilize a deep trench capacitor. Such capacitor structures have reduced surface space while maintaining the charge storage capacity of the capacitor.
One problem associated with the formation of deep trench capacitors for semiconductor memory cells such as DRAMs is how to incorporate temperature sensitive high dielectric constant materials such as barium strontium titanium oxide (BSTO) into the storage node of the DRAM capacitor device.
Another problem associated with the formation of deep trench storage capacitors is the effect of the buried-strap outdiffusion on the array MOSFET electrical characteristics. Presently, DRAM technology falls into two main categories, use of stacked capacitor storage elements and use of deep trench storage capacitors. Each approach has certain advantages and disadvantages relative to each other. For example, deep trench technology results in improved planarization of the various layers of the structure which facilitates the ultra-fine lithographic and etching processes required for today's DRAMs.
FIG. 1
is a cross-sectional diagram showing the structure of a prior art DRAM cell
61
having a trench storage capacitor
60
and an array MOSFET
62
. The DRAM cell
61
occupies an area of about 7F
2
of the surface area of the DRAM IC, F being the minimum feature size which is photolithographically defined for features on the IC. The DRAM cell
61
is characterized by a design dimension
63
which is defined as the lateral distance between the near edge
64
of the trench storage capacitor
60
and the far edge
66
of the gate conductor
68
. In an existing 7F
2
DRAM cell design, this dimension
63
is designed to have a value of nominally 1.5 F, with the width of the gate conductor being 1.0 F and the nominal (designed) distance between the near edge of the gate conductor and the trench storage capacitor
60
being 0.5 F.
The most widely used trench storage DRAM technology utilizes a buried-strap (see
FIG. 1
) to form the connection between the array MOSFET and the storage capacitor
60
. The buried-strap has a diffusion associated with it which extends vertically and laterally away from the interfacial opening between the trench storage capacitor
60
and the silicon substrate (see FIG.
1
). The diffusion is formed by the outdiffusion of dopants (i.e. arsenic) from the N+ polysilicon in the storage trench into the adjacent single crystal silicon substrate. The depth and lateral extent of this buried-strap outdiffusion is highly detrimental to the scalability of the array MOSFET.
In past DRAM generations, when the state of the art minimum feature size, F, was larger than approximately 0.5 &mgr;m, the presence of the buried-strap outdiffusion did not pose much of an electrical problem for the array MOSFET. However, with present day DRAM designs approaching minimum feature size equal to 0.15 &mgr;m, and the typical buried-strap outdiffusion distance being greater than 50 nm from the interface between the N+ polysilicon in the deep trench, it is likely that the buried-strap outdiffused junction may extend under the gate conductor (wordline). This is likely to occur since, in addition to the large outdiffusion relative to the dimension
63
, there is significant misalignment tolerance between the gate conductor and the deep storage trench. As shown in
FIG. 1
, the encroachment of the buried-strap outdiffusion upon the array MOSFET is characterized by the parameter &dgr;, which is the distance between the bitline (BL) diffusion and the buried-strap (BS) outdiffusion, expressed as a percentage of the trench edge to gate conductor edge dimension
63
.
FIG. 2
illustrates how the device off-current increases with decreasing &dgr;. To assure that the off-current objective is met under all circumstances, the channel doping of the MOSFET must be raised. However, increased channel doping results in increased junction leakage and degraded device performance, thus degrading data retention time. Therefore, it is highly desirable to make the value of &dgr; as large as possible to minimize these deleterious effects.
The distance between the bitline diffusion and the buried-strap diffusion is determined by the layout groundrules, process tolerances (overlay and feature size) and the amount of buried-strap outdiffusion from the trench storage capacitor. This critical distance is illustrated by the parameter &dgr; in the prior art deep trench DRAM cell shown in FIG.
1
. The amount of buried-strap outdiffusion is principally determined by the thermal budget that the buried-strap encounters in the course of the chip fabrication process. The thermal budget is a function of the square root of the sum of the products of the diffusivity, D, of the strap dopant impurity and the effective amount of time spent at each high temperature step which contributes significantly to the diffusion.
For prior art deep trench storage capacitor DRAM cells, the high temperature processing steps which contribute to the strap outdiffusion from the storage capacitor polysilicon typically consist of STI oxidation, STI densification, gate sacrificial oxidation, transfer gate oxidation, gate conductor sidewall oxidation, and junction anneals. The combined thermal budget of these high temperature processes results in an arsenic outdiffusion from the storage trench polysilicon into the silicon substrate which typically ranges from 50 to 100 nm. This means that for high-density trench DRAM cell designs having a minimum feature size of 1.5 &mgr;m, the distance &dgr; typically ranges from 0.08 to 0.13 &mgr;m; &dgr; typically ranges from 35% to 60% of the design distance between the storage trench edge and the far edge of the gate conductor, for prior art DRAM cells having a layout area of 7 minimum features squared (7F
2
) per bit and a minimum feature size, F, equal to 0.15 &mgr;m. As seen from
FIG. 2
, the off-current can vary by more than 100×over this range of variation in &dgr;.
In a possible fabrication approach, a deep trench is first formed in a semiconductor substrate or wafer and then the deep trench is filled with a temperature sensitive high dielectric constant material. After filling the deep trench with the temperature sensitive high dielectric constant material, shallow trench isolation (STI) regions and gate conductor (GC) stacks are typically formed. A problem with such an approach is that the high temperatures used in fabricating the STI regions and the GC stacks adversely affect the temperature sensitive high dielectric constant material used in filling the deep trench and contribute to the buried-strap outdiffusion. Specifically, the length of time at the high processing temperatures employed in fabricating the STI regions and the GC stacks cause decomposition of the temperature sensitive high dielectric constant material and add to the thermal budget of the strap outdiffusion. The high dielectric material and its by-products thus formed may diffuse and interact with the underlying semiconductor material.
A partial solution to the above problem has been provided in related application U.S
Hegde Suryanarayan G.
Lee Kam Leung
Mandelman Jack A.
Radens Carl J.
Capella, Esq. Steven
International Business Machines - Corporation
Jr. Carl Whitehead
Scully Scott Murphy & Presser
Thomas Toniae M.
LandOfFree
Application of excimer laser anneal to DRAM processing does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Application of excimer laser anneal to DRAM processing, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Application of excimer laser anneal to DRAM processing will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2562698