Книги+1 / 2013 [Chandan_Kumar_Sarkar]_Technology_CAD

DEVELOP

etch oxide

$ Screen oxidation... E8020

DIFFUSION TEMPERAT=800 T.FINAL=900 TIME=10 F.O2=0.5 F.N2=9.5 DIFFUSION TEMPERAT=900 TIME=15 F.O2=0.5 F.N2=9.5

DIFFUSION TEMPERAT=900 TIME=5 F.O2=9.0

DIFFUSION TEMPERAT=900 TIME=5 F.O2=9.5

DIFFUSION TEMPERAT=900 TIME=28 F.O2=9.0 F.HCL=0.19 DIFFUSION TEMPERAT=900 TIME=5 F.O2=9.0

DIFFUSION TEMPERAT=900 TIME=30 F.N2=10.0

DIFFUSION TEMPERAT=900 T.FINAL=800 TIME=37.5 F.N2=10 print layers

$ Buried layer implant

implant antimony pearson tilt=7 dose=1.0e15 energy=100 etch oxide all

$ Buried layer drive-in... E0381

DIFFUSION TEMPERAT=800 T.FINAL=1200 TIME=100 F.O2=0.5 F.N2=9.5 DIFFUSION TEMPERAT=1200 TIME=600 F.O2=0.5 F.N2=9.5

DIFFUSION TEMPERAT=1200 T.FINAL=1000 TIME=67 F.O2=9 DIFFUSION TEMPERAT=1000 TIME=20 F.O2=10

DIFFUSION TEMPERAT=1000 TIME=67 F.O2=5.5 F.H2=10.4 DIFFUSION TEMPERAT=1000 TIME=1 F.O2=5.5 F.N2=5.0 DIFFUSION TEMPERAT=1000 T.FINAL=800 TIME=67 F.N2=10 etch oxide all

print layers

$ Epi growth, P-type, 30-60 ohm-cm, 14 um

EPITAXY TIME=14 TEMPERAT=1150 THICKNES=14 dx =.001 ydy=0.0 SPACES=100 +

 RESISTIV BORON=45

$ Pad oxide, tox=500A

DIFFUSION TEMPERAT=800 T.FINAL=900 TIME=10 F.O2=9.0 DIFFUSION TEMPERAT=900 TIME=15 F.O2=9.0

DIFFUSION TEMPERAT=900 TIME=18 F.O2=5.5 F.H2=10.4 DIFFUSION TEMPERAT=900 TIME=1 F.O2=5.5 F.N2=5.0 DIFFUSION TEMPERAT=900 TIME=10 F.N2=10

DIFFUSION TEMPERAT=900 T.FINAL=800 TIME=25 F.N2=10 print layers

$ N-tub mask

DEPOSIT PHOTORESIST NEGATIVE THICKNESS=1

EXPOSE MASK=gateunderdoping

394 Technology Computer Aided Design: Simulation for VLSI MOSFET

DEVELOP

etch nitride

etch oxide thickness=0.02

implant boron pearson tilt=7 dose=2.0e11 energy=100 etch nitride all

$ Gate oxide-2 200A

DEPOSIT PHOTORESIST NEGATIVE THICKNESS=1 EXPOSE MASK=gateoxet

DEVELOP

DIFFUSION TEMPERAT=800 T.FINAL=900 TIME=10 F.O2=0.25 F.N2=10 DIFFUSION TEMPERAT=900 TIME=5 F.O2=0.25 F.N2=10

DIFFUSION TEMPERAT=900 TIME=3 F.O2=9.5

DIFFUSION TEMPERAT=900 TIME=47.5 F.O2=9.5 F.HCL=0.19 DIFFUSION TEMPERAT=900 TIME=3 F.O2=9.5

DIFFUSION TEMPERAT=900 TIME=15 F.N2=10.0

DIFFUSION TEMPERAT=900 T.FINAL=800 TIME=37.5 F.N2=10 print layers

etch Photoresist etch nitride all

$ Gate-poly deposition - 4000A

deposition polysili temperature=625 pressure=1.0 thickness=0.4 concentr

$ Poly doping

DIFFUSION TEMPERAT=950 TIME=20 INERT $ Gate-poly mask

DEPOSITION NITRIDE THICKNES=0.10 CONCENTR DEPOSIT PHOTORESIST POSITIVE THICKNESS=1 EXPOSE MASK=gateoxet

DEVELOP

etch nitride etch polysili etch nitride all

$ N+ mask and implant

DEPOSITION NITRIDE THICKNES=0.15 CONCENTR

DEPOSIT PHOTORESIST NEGATIVE THICKNESS=1 EXPOSE MASK=Nplus

DEVELOP

etch nitride etch oxide

implant arsenic pearson tilt=7 dose=6.0e15 energy=100 implant arsenic pearson tilt=-7 dose=6.0e15 energy=100

etch PHOTORESIST etch nitride all

$ P+ mask and implant

DEPOSITION NITRIDE THICKNES=0.15 CONCENTR DEPOSIT PHOTORESIST NEGATIVE THICKNESS=1 EXPOSE MASK=pplus

DEVELOP

etch nitride etch oxide

IMPLANT BORON PEARSON RP.EFF DOSE=1.0e15 ENERGY=30 etch PHOTORESIST

etch nitride all

$ BPSG deposition

deposition oxide thickness=0.7 concentr

$ BPSG anneal

DIFFUSION TEMPERAT=800 TIME=20 F.N2=10.0 DIFFUSION TEMPERAT=800 TIME=15 F.O2=9.5

DEPOSITION NITRIDE THICKNES=0.15 CONCENTR

extract nitride/oxide y.extract minimum name=CMP1

etch nitride all

ETCH OXIDE START X=0.0 Y=-14.6 ETCH CONTINUE X=5 Y=-14.6

ETCH CONTINUE X=5.0 Y=@CMP1 ETCH DONE X=0.0 Y=@CMP1

$ Contact mask

DEPOSIT PHOTORESIST NEGATIVE THICKNESS=1 EXPOSE MASK=contact

DEVELOP etch oxide

etch PHOTORESIST

$ Metal-1 Deposition

deposition aluminum thickness=0.3 concentr

$ Metal-1 mask

DEPOSIT PHOTORESIST positive THICKNESS=1 EXPOSE MASK=metal1

DEVELOP

etch aluminum etch PHOTORESIST

$ IMD dep

deposition oxide thickness=0.5 concentr

DEPOSITION NITRIDE THICKNES=0.15 CONCENTR

extract nitride/oxide y.extract minimum name=CMP2

etch nitride all

ETCH OXIDE START X=0.0 Y=-15.2

ETCH CONTINUE X=5 Y=-15.2

ETCH CONTINUE X=5.0 Y=@CMP2

ETCH DONE X=0.0 Y=@CMP2

$ metal2 mask

DEPOSIT PHOTORESIST NEGATIVE THICKNESS=1 EXPOSE MASK=metal2

DEVELOP etch oxide

etch PHOTORESIST

$ Metal-2 Deposition

deposition aluminum thickness=0.2 concentr etch PHOTORESIST

$ Metal-2 final mask

DEPOSIT PHOTORESIST positive THICKNESS=1 EXPOSE MASK=metal3

DEVELOP

etch aluminum etch PHOTORESIST

savefile medici out.file=LDNBL.str

8.31  Mask File Named t.tl1

The mask file named t.tl1 is given here because it was used during the fabrication of the device. This mask file should be kept in the same folder where the files are kept, especially the actual device fabrication file MOSFET.inp. The first line identifies the file format that contains the character ‘TL1’ followed by a space and a four-digit binary number. The version of TaurusLayout that created the file is represented by the binary number. The current version of the Taurus-Layout specifies values from 0000 to 0100. Nine masks are being used in this file named gateoxet, nbl, Nplus, pplus, contact, metal1, metal2, metal3, and gateunderdoping, and that is why nine is mentioned in the fourth line of this mask file. This information must be provided before the mask names and their dimensions. Total length of the wafer must also be provided, as it is mentioned here 0 to 5000 or 5 μm by the rule assigned here.

TL1 0100

1e3

  0 5000 9

gateoxet 1

purpose it is necessary to create an interface between TSUPREM-4 and Medici. The output structure (having extension .str) file is saved during TSUPREM-4 simulation in such a way that it can be used in the Medici script file. It is done by the expression ‘savefile medici out.file = LDNBL.str’ of the last line of the TSUPREM-4 script. This file ‘LDNBL.str’ is basically created to incorporate the device structure in Medici for its output characteristics analysis. The same file is called to read the simulation meshing information by the following statement.

MESH IN.FILE=LDNBL.str TSUPREM4 ELEC.BOT POLY.ELEC Y.MAX=10

8.35  Rename Electrodes from TSUPREM-4 to Standard Names

Three metal contacts of the simulated device will be named as 1, 2, and 3 by default at the end of the TSUPREM-4 simulation. After the simulation the program numbered the left-most metal contact (Source and Bulk in this case) as 1. In accordance with this numbering scheme, Table 8.1 represents the electrode names and their coordinate positions.

The minimum and maximum x and y positions of different electrodes will be automatically created at the end of the TSUPREM-4 simulation.

The electrode name 1 will be renamed Source by the expression ‘RENAME ELECTR OLDNAME = 1 NEWNAME = Source’ in Medici. Similarly, electrodes 2 and 3 will be renamed as Gate and Drain. After renaming the electrodes to the actual standard device name, the mesh file has been saved with the new electrode name by the expression ‘SAVE MESH OUT.FILE = BVNBL’ in the ext line. The following commands will plot the device structure in the screen with proper labeling as mentioned in the expressions.

RENAME ELECTR OLDNAME=1 NEWNAME=Source

RENAME ELECTR OLDNAME=2 NEWNAME=Gate

RENAME ELECTR OLDNAME=3 NEWNAME=Drain

SAVE MESH OUT.FILE=BVNBL

PLOT.2D GRID FILL TITLE=“Structure from TSUPREM-4” PLOT.1D DOPING LOG X.START=0 X.END=0 Y.START=0 Y.END=2 +    POINTS BOT=1E14 TOP=1E21 TITLE=“S/D Profile” PLOT.1D DOPING LOG X.START=1.8 X.END=1.8 Y.START=0 Y.END=2

TABLE 8.1

Electrode Names and Their Coordinate Positions

+    POINTS BOT=1E14 TOP=1E19 TITLE=“Channel Profile” PLOT.2D BOUND FILL L.ELEC=-1 TITLE=“Impurity Contours” CONTOUR DOPING LOG MIN=14 MAX=20 DEL=1 COLOR=2

CONTOUR DOPING LOG MIN=-20 MAX=-14 DEL=1 COLOR=1 LINE=2

8.36  Major Physical Models

For accurate simulations, a number of physical models are incorporated into the program. Depending on the type of device structure and the device type models like recombination, mobility, band-gap narrowing, band-to-band tunneling, photogeneration, impact ionization, and lifetime can be incorporated.

• Medici also includes semiconductor statistics like Boltzmann and Fermi-Dirac statistics including the incomplete ionization of impurities.

•Different recombination models like SRH, Auger, direct, surface recombinations, and concentration-dependent lifetimes can be incorporated.

•Carrier mobility and scattering are very important phenomena in the mechanism of electrical transport of the device. Mobility models can be divided into two major categories:

•Low and high field mobility

•Surface scattering and electron hole scattering

Low field mobility models are

1.Constant mobility that can be specified with the MUN0 and MUP0 parameters for electron and hole.

2.Concentration dependent mobility can be incorporated with the CONMOB parameter.

3.Either of the two analytic mobility models can be incorporated with the ANALYTIC or ARORA parameters.

4.Carrier-carrier scattering mobility can be incorporated with the CCSMOB parameter.

5.Philips unified mobility can be incorporated with the PHUMOB parameter.

High field mobility models are

1.Field-dependent mobility that can be incorporated with the FLDMOB parameter.

2.Caughey-Thomas mobility can be incorporated with the FLDMOB = 1 parameter.

3.Gallium arsenide–like mobility can be incorporated with the FLDMOB = 2 parameter.

4.Hewlett-Packard mobility can be incorporated with the HPMOB parameter.

Surface scattering mobility models are

1.Surface mobility model can be incorporated with the SRFMOB parameter.

2.Enhanced surface mobility model can be incorporated with the SRFMOB2 parameter.

3.Perpendicular field dependent mobility model can be incorporated with the PRPMOB parameter.

4.Lombardi surface mobility model can be incorporated with the LSMMOB parameter.

5.A number of MOS inversion layer models are available through the parameters UNIMOB, LSMMOB, GMCMOB, SHIRAMOB, and TFLDMOB.

One or more models to be included in a Medici simulation can be specified in models statement. For this device simulation models such as lsmmob, fldmob, auger, bgn, btbt, fermi, incomplete, energy.l and high.dop have been incorporated by the ‘models lsmmob fldmob auger bgn btbt fermi incomplete energy.l high. dop’ statement below.

models lsmmob fldmob auger bgn btbt fermi incomplete energy.l high.dop

8.37  Initial Guess/Convergence and Solution Methods

Depending on the device structure and the range of its operation, one solution method may not be optimal in all cases. Several possibilities can arise for different cases, like at zero bias a Poisson alone is sufficient. For MOSFET, as the device is a unipolar type, only one carrier needs to be solved for its I-V characteristics, though in bipolar and MOSFET breakdown simulations, both carriers are needed. For small geometry devices where the electric field changes rapidly, carrier energy balance may be added to see hot-carrier effect. Solving the lattice heat equation is essential when the device heating effect is important.

The equation that needs to be solved is specified on the SYMBOLIC or SYMB statement.

8.38  Nonlinear System Solutions and Current-Voltage Analysis

For nonlinear system, the solutions are performed by two widely used iteration methods. The methods are de-coupled solutions (Gummel’s method) and coupled solutions (Newton’s method). Newton’s method with Gaussian elimination of the Jacobian is by far the most stable method of solution. For low current solutions, Gummel’s method offers an attractive alternative to inverting the full Jacobian.

Either approach involves solving several large linear systems of equations. The total number of equations in each system is on the order of one to four times the number of grid points, depending on the number of device equations being solved for.

Several ideas are common to all methods of solving the equations. They are convergence rate, error norms, convergence criteria, error norms selection, linear solution options, and initial guess.

The nonlinear iteration converges usually at a linear rate or at a quadratic rate. At a linear rate, the error decreases by about the same factor at each iteration.

The convergence is rapid in the quadratic method, as the error is approximately squared at each iteration. Hence Gummel’s method, which is linear in most cases, is less accurate than Newton’s method, which is quadratic.

Medici uses six types of initial guesses named initial, previous, local, project, p.local, and post-regrid initial.

For tracing the I-V curves, bias step and number of steps are specified on the SOLVE statement by VSTEP and NSTEP for a corresponding electrode.

SYMB GUMMEL CARR=1 ELECTRON

METHOD ICCG DAMPED itlimit=40 stack=10 cont.stk

SOLVE initial V(Source)=0.0 V(Gate)=0.0 V(Drain)=0.005 SYMB NEWTON CARR=1 ELECTRON

METHOD AUTONR N.DAMP N.DVLIM=0.5

SOLVE PREVIOUS V(Source)=0.0 V(Gate)=2 V(Drain)=0.0

LOG OUT.FILE=BVNBLlog

SOLVE V(Drain)=0.0 ELEC=Drain VSTEP=0.2 NSTEP=50

8.39  Post-Processing and Parameter Extraction

After the fabrication of the device, fabricated and simulated device results are necessary, which is possible by the post-processing of the data by the following commands:

Print: The command print will print specific quantities at points within a defined area.