Electrical Calculations for Suprem4

Scott Irving
Technology Engineering, National Semiconductor Corporation
333 Western Ave., South Portland, Maine 04106-0022

Abstract
2 Macro Use
Suprem4 lacks built in facilities for modeling electrical
parameters. Using Suprem4's shell script facilities this can The section covers the use of the macro, details of it's
be overcome by linking Suprem4 with Suprem3, which operation are in the next section. Macro use simply
does have electrical modeling capabilities. This paper requires the user to add four lines to their input file, wher-
describes a Suprem4 macro file that provides electrical ever an electrical result is required. An example of the
calculations for Suprem4 with minimal complexity for the lines from a file are:
end user.
define ElectTemp 1000
1 Introduction define ElectX 1.0
define ElectDepth 2.0
Feature sizes in advanced semiconductor technologies source S4Electric.s4in
are driving the move from one dimensional simulation to
two dimensions. Within National, the two dimensional The ElectTemp define stores a temperature value to be
simulator of choice is Suprem4 from TMA. used in calculating solid solubilities. The next define,
ElectX, is used to select the x coordinate at which the 1D
In addition to it's two dimensional representation, slice is taken, this may be anywhere within the limits of
Suprem4 incorporates many advanced physical models, the users structure. The final define, ElectDepth, stores the
such as those for the diffusion of interstitials and vacan- maximum depth from which electrical results are
cies, and stress models. However it's output tends to limit extracted. ElectDepth is particularly useful for limiting
it's use to more sophisticated users. The only two current the electrical calculation to the region of actual interest
output methods are via the plot statements, and the print.1d when using very deep simulation structures such as those
statement. If the user wants data these two commands can- required by the TWO.DIM model. All three defines are
not offer, they are forced to link the structure to other tools required, although once the values are stored, multiple
such as Medici, or run separate Suprem3 simulations. calls to the macro may be made with out redoing the
defines. The source statement runs the macro.
Suprem4's predecessor, Suprem3, still has several fea-
tures that are not yet found in Suprem4. Examples of this 3 Principles of Operation
includes the electrical, and extract statements.
The shell command facility of Suprem4 offers the abil-
In this paper we discuss the creation of Suprem4 macro ity to link with virtually any other tools on the UNIX net-
files to greatly enhance the users capabilities, with a mini- work. The shell command is invoked with the ! symbol.
mum of additional complications. For example the following command

Suprem4 and Suprem3 commands are shown in cou- ! ls

rier bold in the following discussion. Unix com-
mands are shown in courier italic. will cause Suprem4 to issue the system ls command to list
the files in the current directory. The UNIX echo
command can be combined with the Suprem4 shell As noted above this is repeated three more times to
command to write information to specific files. For extract all four dopants. The only change required for each
example successive iteration is the dopant on the select statement.
! echo line number 1> a.file At the end of the data extraction we have four files con-
! echo line number 2>> a.file taining the profile data.
creates a two line file called a.file. 3.3 Data Import

The next task is to get this information into Suprem3.

Using this simple feature the s4electric macro file per- Again the ! echo construct is used. This time we write a
forms six steps necessary to link to Suprem3 and generate Suprem3 file. This file is as follows:
results. In order of execution, the steps are initialization,
data extraction, Suprem3 data import, calculation, parsing, initial Si Thick=@{ElectDepth} spac=950
and cleanup. profile arsen active +
file=Electrical.profile.Arsenic
3.1 Initialization profile phosp active +
file=Electrical.profile.Phosphor
The initialization portion of the macro pre-defines sev- profile antim active +
eral variables that make the macro file simple to move file=Electrical.profile.Antimon
between systems. These define calls to awk, Suprem4, and profile boron active +
Suprem3. The Suprem4 define also sets up a temporary file=Electrical.profile.Boron
output file name so the current output file is not overwrit- diffusion temp=1300 time=0.0001
ten. The initialization block next saves the current struc-
ture, and writes a short awk program, the function of The number of spaces is chosen to be large to be as gen-
which will be discussed further below. eral as possible in the structure. The Diffusion statement is
3.2 Data Extraction used to set the temperature high enough to insure that all
dopant passed to Suprem3 is active.
The extraction consists of six lines of code, that are 3.4 Calculation
repeated 4 times, once for arsenic, boron, antimony, and
phosphorus. The first three lines define a short Suprem4 Now we are ready to define the desired calculations, and
input file whose structure is: execute them. The initial version of this program used
init file=Elect.Structure only Suprem3's electrical calculation capabilities. This is
select z=active(arsenic) + the only case I will describe here, although the possibilities
temp=@{ElectTemp} are limited only by Suprem3's capabilities. The zero bias
print x.v=@{ElectX} + electrical calculation is carried out by simply appending a
x.max=@{ElectDepth} PRINT ELECTRICAL statement to the Suprem3 input
file, and executing Suprem3 with the file as input. The
The temperature is included in the select command to result is a Suprem3 output file containing the desired
insure that effects of any included rampdowns are negated. result.

The fourth line of the block runs the Suprem4 program, 3.5 Parsing
generating an output file containing 1D profile data for
arsenic. The profile generated can not be read by Suprem3 The electrical data has been successfully generated,
yet. The fifth line of the block uses the UNIX grep com- however the current state is undesirable. The electrical
mand to extract the silicon doping information to a tempo- output is in file separate from our Suprem4 output, with no
rary file. apparent causal connection to the Suprem4 output. If we
execute the electrical macro again from this Suprem4 sim-
The silicon interface in Suprem4 need not have a depth ulation we will also lose the first run data. The parsing
value of zero, later use of the Suprem3 PROFILE state- block solves this by extracting the results of the Suprem3
ment however will require it. The final line of the block electrical statement, and putting it into the Suprem4 output
uses the previously defined awk program to zero the inter- file.
face point.
 The Suprem3 electrical output begins at line 50 of the however for buried layers with a highly resistive epi
s3out file. The lines are extracted using the UNIX sed layer, it may lead to a prediction that is lower than the
editor, and piped to awk where each line is pre-pended measured sheet resistance. Simulations with Medici can
with a $S4Elec:, and saved to a temporary file. Next be used to show that this is indeed the case.
the Suprem4 macro issues a Suprem4 source com-
mand with the temporary file as input, which causes the Due to the nature of the method used to extract the
data to be listed in the output file. data from Suprem4, it is imperative that the be only a
single layer of silicon present. If more than one layer is
It is in the parsing stage that creativity is required as present, the extraction will remove any non-silicon lay-
more Suprem3 calculations such as VT's, capacitance, ers in the 1D slice, and pass the remains on to Suprem3
SRP's or junction depths are added to the a macro. In as a single piece of silicon. The following figure shows
the general case it is difficult to determine the relevent the effect graphically. This is only a minor limitation
lines, so that they can be extracted and included in the
output file. If the user has a specialized task, and struc-
Si
ture to simulate, then it simply becomes a matter of Si
counting lines in the Suprem3 output file. It is the gen- non-Si
eralized case that is difficult. Si
Si
3.6 Cleanup Suprem3
Suprem4
As a final step the macro uses a Suprem4 shell com-
since most simulations contain only a single layer of sil-
mand to remove all temporary files created.
icon. It is possible to overcome this by adding a mini-
mum depth parameter for the profile extraction. This
4 Results
will not be discussed here.
Using simple simulations that insure that the physical
Although Suprem3 now contains dynamic clustering
models used by both Suprem3 and Suprem4 are similar,
models, Suprem4 does not. The use of a fixed tempera-
I have compared results and found them to be within 1-
ture to set solid solubilities is a problem if more than
2%. It must be emphasized that comparisons are only
one diffusion region exceeds solid solubility, and those
valid if the same models are used in both Suprem3 and
regions have seen different thermal histories. This may
Suprem4, and the grid structure is sufficient in both.
force the user to set solid solubility for one or more
The technique has also been extended to the generation
regions either too high or too low, since only a single
of SRP plots for correlation, however this module is not
temperature setting is available.
yet ready for general release.

5 Limitations 6 Future Expansion

There are several limitations to this method, many of The currently released module contains only electrical
which can be overcome by the inclusion of a 1D electri- calculations. A module to generate SRP data has been
cal calculation directly into Suprem4. created, and tested, but the output needs to be cleaned
up prior to release. Future modules are envisioned
Due to Suprem3 the electrical calculation is inher- which perform 1D capacitance calculations, 1D VT cal-
ently 1D. It is crucial that the user selects a X position culations, and extract depletion region edges.
to such that contours of doping would be constant for a
fixed Y depth. Data generated through curved regions Beyond these it is possible to use other tools such as
can be interpreted, but it's reliability is unknown. Medici, for calculations. For example of Medici could
be used for an accurate calculation of sheet resistance
Suprem3 has a built in limitation to the electrical cal- (four point probe results) for deep buried layers, by
culation. Suprem3 does the electrical calculation by cal- including vertical resistance elements. This would be
culating the conductance at each node, and adding the useful from Suprem3 also, but unfortunately Suprem3 is
conductances within a diffusion region. Suprem3 prints unable to execute shell commands.
the inverse as the sheet resistance. This ignores the ver-
tical components normally present in a sheet resistance
measurement. For many structures this is unimportant,
7 Challange to Reader The use of Suprem4's shell command capabilities pro-
vides an extremely powerful capability. The above
To conclude I offer a challenge to readers to create example concerns itself with the linking of Suprem3.
more Suprem4 macros that are general in nature, and to However, using the same feature it is possible to add in
share them via the TCAD bulletin board. Specialized links to Depict, Medici, Minimos and others. The possi-
modules would also be appreciated since other users bilities are very promising.
can adapt them to their needs.