UniMAP – PSDC INSEP Training Program 2007 School of Microelectronic Engineering UniMAP – PSDC INSEP Training Program 2007 by Syarifah Norfaezah Sabki School

UniMAP – PSDC INSEP Training Program 2007

School of Microelectronic Engineering UniMAP – PSDC INSEP Training Program 2007

by Syarifah Norfaezah Sabki

School of Microelectronic Engineering



• 2D cross-section of wafer– X-coordinate: parallel to the wafer surface– Y-coordinate: depth into the wafer

• Grid structure:– The continous physical process are modeled

numerically by using finite difference (for diffusion) and finite element (for oxide flow) solution techniques.

– Each region is divided into a mesh of non-overlapping triangular elements

– Solution values are calculated at the mesh nodes (at the corners of the triangular elements), value between the nodes are interpolated



• MEDICI solves Poisson’s equation & the current continuity of electrons and holes in two dimensions

• These equations can be extended to include the heat equation and the energy balance equations

• The following modes of analysis can be considered: DC simulation, AC simulation & transient simulation

• A wide range of mobility & recombination/generation models available



• Advanced Application Modules are available– Lattice temperature AAM – solves the heat equation– Optical device AAM – enhanced radiation effects, ray

tracing– Heterojunction device AAM – conduction across a

material boundary with discontinuous energy– Programmable device AAM – allows a charge

boundary condition on a floating electrode– Circuit analysis AAM – allows devices to be treated as

circuit elements in a SPICE type circuit– Anisotropic device AAM – allows anisotropic material

parameters useful in the treatment of SiC type applications



DEVICE STRUCTURE

GENERATING DEVICE STRUCTURE IN MEDICI/DAVINCI



DEVICE STRUCTURE DEFINITION

SEQUENCE OF STATEMENTS: MESH statement X.MESH statements Y.MESH statements Z.MESH statements (Davinci only) ELIMINATE statements (optional) TSUPREM4 statements (optional) REGION statements ELECTRODE statementsPROFILE statements



STRUCTURE INFORMATION

MESH

Initiates a mesh and must appear first when defining a structure. Can be used to import an existing mesh and invoke the Automatic Conforming Boundary (ABC) mesher

X.MESH

Y.MESH

ELIMINATE

Used to specify exact locations of mesh lines. X.MESH & Y.MESH produce a rectangular grid which can be reduced in density by using ELIMINATE to remove excess nodes away from area of interest

TSUPREM4Used to transfer surface features and doping profiles from TSUPREM4 onto existing MEDICI mesh



STRUCTURE INFORMATION

REGIONUsed to define regional properties

where no material data already exists

ELECTRODE Adds location of electrodes to structure

RENAME Rename electrodes or regions

PROFILEAllows addition of doping information either by creating simple profiles or inputting from a process simulator

REGRIDAllows regridding of mesh based on

some internal quantities



DEVICE STRUCTURE: MESH

• The MESH statement initiates the mesh generation or reads a previously generated mesh



[extracted from user guide]

MESH

Initial Mesh Generation{ ( [ { RECTANGULAR | CYLINDRI } ] [DIAG.FLI])

Mesh File Input| (IN.FILE=<c> [QT.FILES=<c>] [PROFILE][ { ASCII.IN | (TSUPREM4 [ ELECT.BOT [Y.TOLER=<n>] [POLY.ELE][X.MIN=<n>] [X.MAX=<n>] [Y.MIN=<n>] [Y.MAX=<n> [FLIP.Y] [SCALE.Y=<n>])| (TIF [ELECT.BOT [Y.TOLER=<n>] [POLY.ELE] ] )}





PARAMETER TYPE DEFINITION DEFAULT

RECTANGU logical

Specifies that the simulation mesh uses rectangular coordinates True

CYLINDRI logical

Specifies that the simulation mesh uses cylindrical coordinates. If this parameter is specified, the horizontal axis represents the radial direction and the vertical axis represents the z-direction

False

DIAG.FLI logical

Specifies that the direction of diagonals is changed about the horizontal center of the grid. If this parameter is false, all diagonals are in the same direction

False



• The X.MESH specifies the placement of nodes in the x direction• Description:

If an initial mesh is being generated, X.MESH and Y.MESH statements should immediately follow the MESH statement

DEVICE STRUCTURE: X.MESH

X.MESH

{LOCATION=<n> | ({ WIDTH=<n> | X.MAX=<n> }

[X.MIN=<n>] )}

[ {NODE=<n> | N.SPACES=<n>} ]

[SPACING=<n> | H2=<n>} ] [H3=<n>] [RATIO=<n>]

[MIN.SPAC=<n>] [ SUMMARY ]



DEVICE STRUCTURE: Y.MESH

The following Y.MESH statement specifies the placement of nodes in the y direction

Y.MESH

{LOCATION=<n> | ({DEPTH=<n> | Y.MAX=<n>} [Y.MIN=<n>] ) }

[ {NODE=<n> | N.SPACES=<n>} ]

[ {SPACING=<n> | [MIN.SPAC=<n>]

[SUMMARY]



DEVICE STRUCTURE: REGION

The region statement defines the location of materials in a rectangular mesh

REGION

NAME=<c>

Semiconductor Materials

{ ( { SILICON | GAAS | POLYSILI | GERMANIU | SIC | SEMICOND | SIGE | ALGAAS | A-SILICO | DIAMOND | HGCDTE | INAS | INGAAS | INP | S.OXIDE | ZNSE | ZNTE | ALINAS | GAASP | INGAP | INASP }




Semiconductor material Parameters

[X.MOLE=<n>] [X.END=<n> | X.SLOPE=<n>} {X.LINEAR | Y.LINEAR} ]

)

Insulator Materials

| OXIDE | NITRIDE | SAPPHIRE | OXYNITRI | HFO2 | INSULATO

}




Location{ ( [ {X.MIN=<n> | IX.MIN=<n>} ] [ {X.MAX=<n> | IX.MAX=<n>} ] [ {Y.MIN=<n> | IY.MIN=<n>} [{Y.MAX=<n> | IY.MAX=<n> }] [ { (ROTATE R.INNER=<n> R.OUTER=<n> X.CENTER=<n> Y.CENTER=<n>)

|POLYGON X.POLY=<a> Y.POLY=<a>) } ] ) | (X=<n> Y=<n>)

|CONVERT

}

[VOID]



DEVICE STRUCTURE: ELECTRODE

The ELECTRODE statement specifies the placement of electrodes in a device structure

ELECTRODE

NAME=<c> [VOID]

{ ( [ {TOP | BOTTOM | LEFT | RIGHT | INTERFAC | PERIMETE} ] [ { X.MIN=<n>} ] [X.MAX=<n> | IX.MAX=<n>} ] [ { Y.MIN=<n> | IY.MIN=<n>}] [ {Y.MAX=<n> | IY.MAX=<n>} ] [ { ( ROTATE X.CENTER=<n> Y.CENTER=<n> R.INNER=<n> R.OUTER=<n>) | (POLYGON X.POLY=<a> Y.POLY=<a>) } ] )

| [X=<n> Y=<n>]

| [REGION=<c>] }

[MAJORITY]

Lattice Temperature AAM Parameters

[THERMAL]



DEVICE STRUCTURE: PROFILEThe PROFILE statement defines profiles for impurities and other quantities to be used in the device structure

PROFILE

[REGION=<c>]

[X.MIN=<n>] [ {WIDTH=<n> | X.MAX=<n>} ]

[Y.MIN=<n>] [ {DEPTH=<n> | Y.MAX=<n>} ]

Output Doping File

[OUT.FILE=<c>]

Uniform Profiles

{ (UNIFORM {N-TYPE | P-TYPE | IMPURITY=<c> | OTHER=<c>}

N.PEAK=<n>)

Analytic Profiles

| ( {N-TYPE | P-TYPE IMPURITY=<c> | OTHER=<c>} {N.PEAK=<n> |

DOSE=<n>} { (Y.CHAR=<n> [Y.ERFC] ) | Y.JUNCTI=<n>} {X.CHAR=<n> |

XY.RATIO=<n>} [X.ERFC]



EXAMPLE: CREATING 1D SiGe HBT

$ Structure Generation of 1D SiGe Bipolar

Mesh

X.mesh width=0.5 spaces=1

Y.mesh width=0.1 H2=0.005 Ratio=1.2

Y.mesh width=0.1 H2=0.005

Y.mesh width=0.6 H2=0.005 H2=0.050

Region silicon

Region SiGe Y.min=0.100 y.max=0.125 x.mole=0 x.end=0.2 Y.linear

Region SiGe Y.min=0.125 y.max=0.200 x.mole=0.2

Region SiGe Y.min=0.200 y.max=0.230 x.mole=0.2 x.end=0.0

Electr Name=Emitter Top

Electr Name=base Y.min=0.125 Y.max=0.125 Majority

Electr Name=collector bottom

Profile N-type N.peak=2e16 Uniform

Profile N-type N.peak=5e19 Y.min=0.80 y.char=0.125



EXAMPLE: RESULTS

Basic SiGe Mesh

Corresponding doping profile



DEVICE STRUCTURE

IMPORTING DEVICE STRUCTURE FROM MEDICI/TSUPREM4



MESH STATEMENTS

• IN.FILE – name of input file which contains structure.• Tsuprem4 – logical parameter signaling that IN.FILE was created

by TSUPREM4• TIF – logical parameter signaling that IN.FILE is in universal (TIF)

format • ELECT.BOT – logical flag signaling that the structure bottom

(substrate) electrode is supposed to be appended to the structure• POLY.ELEC – logical parameter signaling that all polysilicon

regions in the imported structure are to be converted to electrode

NOTE: Once Poly Region is converted to Electrode, its doping information is lost and intrinsic work function of 4.6eV is assign to it



• From TSUPREM4 MESH in.file=s4filename tsuprem4 elec.bot poly.elec

y.max=3RENAME electr oldname=1 newname=sourceRENAME electr oldname=2 newname=drainSAVE mesh out.file=mdfile

• From previous MEDICI execution

MESH in.file=mdfile

EXAMPLE: IMPORTING STRUCTURE FILE



• Default structure depth in TSUPREM4 is 200m. Use Y.MAX or alternatively TRUNCATE the device within TSUPREM4 first

• X.SPLIT, WIDTH and N.SPACES allow the structure to be expanded at point x.split by an amount width and subdivided into n.spaces. A typical use of this would be to model various channel lengths without repeating the process simulation

MESH ADJUSTMENT



• REGRID statement• Regrid doping log ratio=2 in.file=test.dop

smooth=1

Which test for the log of the doping being greater than 2 between mesh points. It uses a doping file stored from the original PROFILE statement so that information on doping is not lost through successive refinements. A number of different techniques from smooth=-1 to 2 can be selected (-1 is usually the best)

• Regrid potential ratio=1.1• Regrid min.carr ratio=2 log smooth=-1

MESH ADJUSTMENT



REGRID



• Increasing mesh density results in increasing accuracy of potential and carrier concentrations

• Care must be taken in aligning the mesh to the current flow

• High density mesh needs computing space and time

MESH ISSUES



CHOICE OF MODELS : RECOMBINATION & GENERATION

MODEL DESCRIPTION

SRH Shockley – Read – Hall recombination

CONSRH SRH + concentration dependent lifetime

Note: lattice temp dependence can also be modeled by specifying non-zero values of EXN.TAU and EXP.TAU on the MATERIAL statement (Lattice temp AAM only)

AUGER Auger recombination

R.TUNNEL SRH including tunneling in presence of strong electric fields

IMPACT.I Classic Chynoweth expression

II.TEMP Invokes a temperature based version of the impact ionization model for use with the energy balance model



CHOICE OF MODELS : MOBILITY

MODEL LOW FIELD

TRANSVERSE FIELD

PARALLEL FIELD

COMMENTS

CCSMOB Carrier-carrier scattering

CONMOB Concentration dependence from tables 300K

ANALYTIC Analytic alternative to CONMOB with temp. dependence

PHUMOB Carrier-carrier scattering, different donor and acceptor scattering, screening, useful for bipolars

LSMMOB Treats surface scattering and bulk effects

GMCMOB Modified LSMMOB to include impurity scattering



CHOICE OF MODELS : MOBILITY

MODEL LOW FIELD

TRANSVERSE FIELD

PARALLEL FIELD

COMMENTS

SRFMOB Basic and enhanced model for surface scattering. Requires vertical grid spacing > inversion layer

SRFMOB2

UNIMOB Needs rectangular grid in inversion layer – models surface scattering

PRPMOB General model for degradation of mobility with transverse electric field – applies all over –not just at surface

TFLDMOB Univ. Texas mobility model

FLDMOB Carrier heating and velocity saturation effects

HPMOB Accounts for both parallel and perpendicular field dependence



CHOICE OF MODELS : ENERGY GAP & CARRIER DENSITY

MODEL0 DESCRIPTION

FERMIDIRFermi Dirac statistics instead of Boltzman. Recommended to be used in conjunction with:

INCOMPLE Incomplete ionization of impurities

BGNBandgap narrowing modelling – especially important for bipolars

QM.PHILI

Accounts for quantum mechanical effects in MOSFET inversion layers using Van Dort’s bandgap widening model. Implemented as a shift in the energy gap just as in BGN modeling



CHOICE OF MODELS : ENERGY BALANCE

MODEL DESCRIPTION

ET.MODELUses the energy transport model where the spatial derivative of the mobility is included in the diffusion term of the current equation

COMP.ETInvokes an energy balance eq. suitable for compound material such as GaAs

TMPMOBA carrier temperature based mobility – alternative to FLDMOB

EF.TMPSolves effective electric fields exactly in Si instead of approx for use in TMPMOB

TMPTAUWInvokes an electron temperature model for the electron energy relaxation



CHOICE OF MODELS : ENERGY BALANCE

MODEL DESCRIPTION

II.TEMPUses the energy transport model where the spatial derivative of the mobility is included in the diffusion term of the current equation

EFI.TMPInvokes an energy balance eq. suitable for compound material such as GaAs



MODEL DECISION: MOS

• Use mobility model specifically calibrated on MOSFETS as surface scattering effects are a dominant feature such as CONMOB LSMMOB FLDMOB

• For <0.2m technologies, one of the newer models i.e UNIMOB, GMCMOM or TFLDMOB should be considered i.e TFLDMOB (for NMOS)

• When modeling breakdown CONSRH, IMPACT.I are important

• AUGER and BGN which has a small effect on the source/drain resistance can be included but both of these will not significantly impact the results



• Carrier-carrier scattering is a more important mechanism for bipolars and PHUMOB would be a good choice. Bandgap narrowing and the recombination mechanisms are also important so a full set would be:

CONMOB PHUMOB AUGER CONSRH BGN IMPACT.I

• Change the lifetimes and bandgap coefficients on the material statement:

material silicon

v0bgn=n0.bgn=con.bgn=taun=taup=

• For a general device, then an all purpose choice would be:

CONMOB FLDMOB PRPMOB CONSRH AUGER BGN IMPACT

MODEL DECISION: BIPOLAR



SOLUTION TECHNIQUESTATEMENTS

STATEMENTS DESCRIPTION

SYMBOLICSelects with equations to solve as well as the method of the solution either coupled (Newton) or de-coupled (Gummel)

METHODControl the iteration process – number of iterations use of numerical damping, selection of linear solver

LOGTo open the file which will contain terminal values calculated during the solution process

SOLVE Starts the solution process either DC, AC or transient



SOLUTION TECHNIQUE: SYMBOLIC

• Solve only Poisson’s equationsymbolic carr=0

• Solve Poisson’s equation and electron-current continuity equation using Gummel’s method

symbolic carr=1 electron gummel• Solve Poisson’s equation and electron-current

continuity equation using coupled methodsymbolic carr=1 electron newton

• Solve Poisson’s equation and both hole and electron Drift-Diffusion (DD) equations

symbolic carr=2 newton



SOLUTION TECHNIQUE: METHOD & LOG

• Method – contains more than fifty parameters, only a few are normally used

• Itlimit, which controls the number of iterations which are tried before the bias is cut back by the program

method itlimit=100• Log

log outfile=drain.ivl (filename)



SOLUTION TECHNIQUE: SOLVE

• There are two fundamental rules when using the solve statement:– At the beginning of the simulation, all electrode potentials are set to

0V– Terminal values stay unchanged until they are addressed by the

next solve statement. In other words, terminal values are not implicity reset to their initial values in subsequent solve statements

• When the program solves for a new bias condition, it must rely on an initial guess. There are three types (initial, previous, project) which are automatically selected by the program

• Rules for succesful solution strategy:– Specify all models (with the possible exception of impact.i before

the first solve statement– Build-up solution gradually




• DC ANALYSIS• Apply 1V gate electrode

solve v(gate)=1• Ramp voltage of gate electrode at 1V interval for 5 times

solve elec=gate vstep=1 nstep=5• Ramp current of base while applying 5V at collector

solve elec=base istep=1e-6 nstep=10

v(collector)=5




• TRANSIENT ANALYSIS

solve v(base)=1 tstep=1e-13 tstop=1e9

• To define a pulse we need two solve statements:Solve v(base)=1 tstep=1e-13 tstop=1e-9

Solve v(base)=0 tstep=1e-13 tstop=5e-9

V

tstop tstop t



SOLUTION TECHNIQUE:CONVERGENCE ISSUE

• The primary causes of non-convergence are:– Poor initial guess – bias step too large (for some structures even

0.1V can be too large)– Lack of necessary physical models– Poor simulation grid– Depletion layer touching the electrode

V-error

px.tol

itlimit #of iterations

Iter V-error

1 3.4567e+4

2 2.7543e+02

3 1.6734e+00

4 1.0000e+00

5 1.0000e+00

… 1.0000e+00

20 1.0000e+00

Documents

UniMAP – PSDC INSEP Training Program 2007 School of Microelectronic Engineering UniMAP – PSDC INSEP Training Program 2007 by Syarifah Norfaezah Sabki School