Initial Guess: Difference between revisions

From Flooxs
Jump to navigation Jump to search
No edit summary
 
Line 5: Line 5:
  proc InitialGuess {Doping} {  
  proc InitialGuess {Doping} {  
  sel z= {(Doping>0.0)  
  sel z= {(Doping>0.0)  
  ?  ( 0.025*log( (Doping+1.0e10) / 1.0e10))
  ?  ( 0.025*log( (Doping+1.0e-10) / 1.0e10))
  :  (-0.025*log(-(Doping+1.0e10) / 1.0e10))} name = DevPsi
  :  (-0.025*log(-(Doping+1.0e-10) / 1.0e10))} name = DevPsi
  sel z=1.0e10*exp(DevPsi/0.025) name=Elec
  sel z=1.0e10*exp(DevPsi/0.025) name=Elec
  sel z=1.0e10*exp(-DevPsi/0.025) name=Hole
  sel z=1.0e10*exp(-DevPsi/0.025) name=Hole
Line 16: Line 16:
  proc InitialGuessQF {Doping} {
  proc InitialGuessQF {Doping} {
  sel z= {(Doping>0.0)  
  sel z= {(Doping>0.0)  
  ?  ( 0.025*log( (Doping+1.0e10) / 1.0e10))
  ?  ( 0.025*log( (Doping+1.0e-10) / 1.0e10))
  :  (-0.025*log(-(Doping+1.0e10) / 1.0e10))} name = DevPsi store
  :  (-0.025*log(-(Doping+1.0e-10) / 1.0e10))} name = DevPsi store
  sel z=1.0e10*exp((DevPsi)/0.025) name=Elec store
  sel z=1.0e10*exp((DevPsi)/0.025) name=Elec store
  sel z=1.0e10*exp((-DevPsi)/0.025) name=Hole store
  sel z=1.0e10*exp((-DevPsi)/0.025) name=Hole store
Line 27: Line 27:
  proc InitialGuessSB {Doping} {
  proc InitialGuessSB {Doping} {
  sel z= {(Doping>0.0)  
  sel z= {(Doping>0.0)  
  ?  ( 0.025*log( (Doping+1.0e10) / 1.0e10))
  ?  ( 0.025*log( (Doping+1.0e-10) / 1.0e10))
  :  (-0.025*log(-(Doping+1.0e10) / 1.0e10))} name = DevPsi store
  :  (-0.025*log(-(Doping+1.0e-10) / 1.0e10))} name = DevPsi store
  sel z=1.0e10*exp((DevPsi)/0.025) name=Elec  
  sel z=1.0e10*exp((DevPsi)/0.025) name=Elec  
  sel z=1.0e10*exp((-DevPsi)/0.025) name=Hole  
  sel z=1.0e10*exp((-DevPsi)/0.025) name=Hole  
Line 34: Line 34:
  sel z=1.0 name=SBp store
  sel z=1.0 name=SBp store
  }
  }
= Poisson-Only Solve as Initial Guess =
= Poisson-Only Solve as Initial Guess =
Sometimes it is very hard to get convergence on your first try. For example, when Elec and Hole concentrations are very low. In this case, you can improve your initial guess by doing a Poisson-Only solve first (a DevPsi only solve). Then you can use your improved DevPsi solution as an initial guess to your full solve:
Sometimes it is very hard to get convergence on your first try. For example, when Elec and Hole concentrations are very low. In this case, you can improve your initial guess by doing a Poisson-Only solve first (a DevPsi only solve). Then you can use your improved DevPsi solution as an initial guess to your full solve:

Latest revision as of 00:08, 11 January 2011

Common Procedures

Initial guess assumes charge neutrality

For Scharfetter-Gummel (Elec Hole DevPsi) solution

proc InitialGuess {Doping} { 
	sel z= {(Doping>0.0) 
		?  ( 0.025*log( (Doping+1.0e-10) / 1.0e10))
		:  (-0.025*log(-(Doping+1.0e-10) / 1.0e10))} name = DevPsi
	sel z=1.0e10*exp(DevPsi/0.025) name=Elec
	sel z=1.0e10*exp(-DevPsi/0.025) name=Hole
	sel z=0.0 name=Qfn
	sel z=0.0 name=Qfp
} 

For Quasi-Fermi Solution (Qfn Qfp DevPsi) solution

proc InitialGuessQF {Doping} {
	sel z= {(Doping>0.0) 
		?  ( 0.025*log( (Doping+1.0e-10) / 1.0e10))
		:  (-0.025*log(-(Doping+1.0e-10) / 1.0e10))} name = DevPsi store
	sel z=1.0e10*exp((DevPsi)/0.025) name=Elec store
	sel z=1.0e10*exp((-DevPsi)/0.025) name=Hole store
	sel z=0.0 name=Qfn store
	sel z=0.0 name=Qfp store
}

For Slotboom (SBn SBp DevPsi) solution

proc InitialGuessSB {Doping} {
	sel z= {(Doping>0.0) 
		?  ( 0.025*log( (Doping+1.0e-10) / 1.0e10))
		:  (-0.025*log(-(Doping+1.0e-10) / 1.0e10))} name = DevPsi store
	sel z=1.0e10*exp((DevPsi)/0.025) name=Elec 
	sel z=1.0e10*exp((-DevPsi)/0.025) name=Hole 
	sel z=1.0 name=SBn store
	sel z=1.0 name=SBp store
}

Poisson-Only Solve as Initial Guess

Sometimes it is very hard to get convergence on your first try. For example, when Elec and Hole concentrations are very low. In this case, you can improve your initial guess by doing a Poisson-Only solve first (a DevPsi only solve). Then you can use your improved DevPsi solution as an initial guess to your full solve:

#===================================ponly===========================================#
# Solution Variables
solution add name=DevPsi pde   solve negative damp;#electrostatic potential (Device Psi)
solution add name=Elec   const solve val = "(Elec0)";# electron concentration /cm^3
solution add name=Hole   const solve val = "(Hole0)";# hole concentration /cm^3
pdbSetBoolean PoissonOnly 1;#this flag controls if statements in supporting procedures

# Bulk Equations
AddPoisson 	Oxide
AddPoisson	Silicon
BandTerms Oxide     Semiconductor;# sets Ec, Ev, nQFL, pQFL, Elec0, Hole0
BandTerms Silicon   Semiconductor
SetDoping 

#Poisson only - no cont. eqns - just add charge to Poisson in Mats that have charge
AddCharge Oxide   "-Elec+Hole"
AddCharge Silicon "-Elec+Hole" 

# Boundary Conditions
solution name=DevPsi continuous 

# Contact Equations
MetalContact top Aluminum
OhmicContact bot Silicon Ntype 

# Initial Conditions - bias the Contacts to 0 
contact name=top voltage supply=0.0 
contact name=bot voltage supply=0.0

#Initial guess
puts "You are Guessing"
sel z=-(4.8) name=DevPsi;#this value may be different depending on where your "0" is

# Poisson-Only Solution (Finally we SS Solve)
device init 

# Plots
PlotBands x.v=0 y.v=0
PlotConc ponly 
 
#end of Poisson only
pdbSetBoolean PoissonOnly 0;#turn off ponly flag
 	
#========================zerobias================================#
# Solution Variables - change all to pde
solution name=DevPsi    pde   solve negative damp;#electrostatic potential (Device Psi)
solution name=Elec	 pde   solve !negative;# electron concentration /cm^3
solution name=Hole	 pde   solve !negative;# hole concentration /cm^3 

# Bulk Equations
AddPoisson 	Oxide
AddPoisson	Silicon
BandTerms Oxide     Semiconductor;# sets Ec, Ev, nQFL, pQFL
BandTerms Silicon   Semiconductor
SetDoping 

AddSpecies	Hole		Oxide		MOB	1.0e-5	+
AddSpecies	Elec		Oxide		MOB	20.0	-
AddSpecies	Hole		Silicon		MOB	150.0	+
AddSpecies	Elec		Silicon		MOB	300.0	- 

# Boundary Conditions
solution name=DevPsi continuous 

# Contact Equations
MetalContact top Aluminum 
OhmicContact bot Silicon Ntype

#Initial Guess 
    #use DevPsi from ponly solve above
    sel z=Elec0 name=Elec;#is a fcn of DevPsi
    sel z=Hole0 name=Hole;#is a fcn of DevPsi
pdbSetDouble Math iterLimit 50;#give SS a chance
device

#PlotBands x.v=0 y.v=0
#PlotConc zerobias

struct outf=init.str

supporting procedures