|
|
|
| Figure 1 | Figure 2 | Figure 3 |
Hints & Tips
Q: How can the reverse short channel effect (RSCE) in MOSFETs be simulated using ATHENA and ATLAS? How can the physical effect behind RSCE be tuned ?
A: RSCE in MOSFETs is where the threshold voltage increases with decreasing channel length. At very short channel lengths the normal short channel effect takes over and the threshold voltage decreases.
The cause of the increasing threshold voltage is a non-uniform enhancement of diffusion of the channel implant laterally along the MOS channel. This non-uniformity arises from the extra point defects generated in the source and drain areas of the MOSFET. The source of these point defects is most commonly the damage caused by the heavy n+ and LDD implants. Other possible causes that can be modeled in ATHENA are oxidation or silicidation of the source and drain area.
The amount of implant damage from the source/drain
implants is controlled using the DAM.FACTOR
parameter. The effect of the damage on subsequent
diffusions are modeled in ATHENA using the fully coupled diffusion
model (METHOD FULL.CPL).
A previous hints and tips covered a description of this (Simulation
Standard, Feb 1995).
To model RSCE in ATHENA and ATLAS it is
necessary to construct MOSFETs of different channel lengths. This
can be done either using the MaskViews layout interface, or using
the STRETCH command in
ATHENA or DevEdit. The user should simulate the shortest channel
length up until the polysilicon etch and stretch the device to the
desired length. The FULL.CPL model
is only required for diffusion after the source/drain implants.
Figure 1 shows the result of a threshold voltage simulation
versus gate length for various values of implant damage. VWF was used to automatically
generate and run this experiment. VWF handles the automatic interface to ATLAS
and the extraction of the threshold voltages. Looking horizontally along the
y=0 line it is seen that with zero implant damage the threshold voltage decreases
with decreasing length. No RSCE is seen. However as DAM.FACT
is increased, the threshold voltage starts to rise before
falling at very short lengths. It is clear the size of the RSCE increases with
implant damage factor.
It is also interesting to note that even the threshold voltage for the 20mm long device is affected slightly by the implant damage. This is to be expected from Figure 2 on page 5 which shows point defects diffusing 30mm into the substrate. The lateral diffusion length of point defects should be of a similar order.
Many parameters can be used to tune the fully coupled
diffusion model. The most effective for RSCE is the surface recombination of
the interstitials (KSURF.0). Figure
2 shows threshold voltage versus channel length as a function of KSURF.0
for a fixed DAM.FACT. High
values of KSURF.0 show no RSCE effect
while lower values show strong increases in threshold at lengths around 1.0
mm.
Tuning RSCE using DAM.FACT and
KSURF.0 is possible using ATHENA, ATLAS and
VWF. Users should note that both these parameters will affect process simulation
results such as source/drain junction depth. Figure 3 shows a graph of junction
depth of an arsenic implant after a fixed diffusion as a function of
DAM.FACT and KSURF.0.
For a given measured result for junction depth it is clear
there are a whole set of DAM.FACT and
KSURF.0 combinations that can produce the
correct answer. However the effect of each combination that matches a junction
depth is not the same on RSCE.
