octave-secs2d 0.0.8-5 / usr / share / octave / packages / secs2d-0.0.8 / DDGOX / DDGOXnlpoisson.m

function [V,n,p,res,niter] = DDGOXnlpoisson (mesh,Dsides,Sinodes,SiDnodes,...
                                             Sielements,Vin,nin,pin,...
                                             Fnin,Fpin,D,l2,l2ox,...
                                             toll,maxit,verbose)

%  
%   [V,n,p,res,niter] = DDGOXnlpoisson (mesh,Dsides,Sinodes,Vin,nin,pin,...
%                                             Fnin,Fpin,D,l2,l2ox,toll,maxit,verbose)
%
%  solves $$ -\lambda^2 V'' + (n(V,Fn) - p(V,Fp) -D)$$
%


% This file is part of 
%
%            SECS2D - A 2-D Drift--Diffusion Semiconductor Device Simulator
%         -------------------------------------------------------------------
%            Copyright (C) 2004-2006  Carlo de Falco
%
%
%
%  SECS2D is free software; you can redistribute it and/or modify
%  it under the terms of the GNU General Public License as published by
%  the Free Software Foundation; either version 2 of the License, or
%  (at your option) any later version.
%
%  SECS2D is distributed in the hope that it will be useful,
%  but WITHOUT ANY WARRANTY; without even the implied warranty of
%  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
%  GNU General Public License for more details.
%
%  You should have received a copy of the GNU General Public License
%  along with SECS2D; If not, see <http://www.gnu.org/licenses/>.

global DDGOXNLPOISSON_LAP DDGOXNLPOISSON_MASS DDGOXNLPOISSON_RHS 

%% Set some useful constants
dampit 		= 10;
dampcoeff	= 2;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% convert input vectors to columns
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

if Ucolumns(D)>Urows(D)
	D=D';
end


if Ucolumns(nin)>Urows(nin)
	nin=nin';
end

if Ucolumns(pin)>Urows(pin)
	pin=pin';
end

if Ucolumns(Vin)>Urows(Vin)
	Vin=Vin';
end 

if Ucolumns(Fnin)>Urows(Fnin)
	Fnin=Fnin';
end 

if Ucolumns(Fpin)>Urows(Fpin)
	Fpin=Fpin';
end 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% setup FEM data structures
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

nodes=mesh.p;
elements=mesh.t;
Nnodes = length(nodes);
Nelements = length(elements);

% Set list of nodes with Dirichelet BCs
Dnodes = Unodesonside(mesh,Dsides);

% Set values of Dirichelet BCs
Bc     = zeros(length(Dnodes),1);
% Set list of nodes without Dirichelet BCs
Varnodes = setdiff([1:Nnodes],Dnodes);


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
%% 		initialization:
%% 		we're going to solve
%% 		$$ - \lambda^2 (\delta V)'' +  (\frac{\partial n}{\partial V} - \frac{\partial p}{\partial V})= -R $$
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%% set $$ n_1 = nin $$ and $$ V = Vin $$
V = Vin;
Fn = Fnin;
Fp = Fpin;
n = exp(V(Sinodes)-Fn);
p = exp(-V(Sinodes)+Fp);
n(SiDnodes) = nin(SiDnodes);
p(SiDnodes) = pin(SiDnodes);


%%%
%%% Compute LHS matrices
%%%

%% let's compute  FEM approximation of $$ L = -  \frac{d^2}{x^2} $$
if (isempty(DDGOXNLPOISSON_LAP))
	coeff = l2ox * ones(Nelements,1);
	coeff(Sielements)=l2;
	DDGOXNLPOISSON_LAP = Ucomplap (mesh,coeff);
end

%% compute $$ Mv = ( n + p)  $$
%% and the (lumped) mass matrix M
if (isempty(DDGOXNLPOISSON_MASS))
    coeffe = zeros(Nelements,1);
    coeffe(Sielements)=1;
	DDGOXNLPOISSON_MASS = Ucompmass2(mesh,ones(Nnodes,1),coeffe);
end
freecarr=zeros(Nnodes,1);
freecarr(Sinodes)=(n + p);
Mv      =  freecarr;
M       =  DDGOXNLPOISSON_MASS*spdiag(Mv);

%%%
%%% Compute RHS vector (-residual)
%%%

%% now compute $$ T0 = \frac{q}{\epsilon} (n - p - D) $$
if (isempty(DDGOXNLPOISSON_RHS))
    coeffe = zeros(Nelements,1);
    coeffe(Sielements)=1;
	DDGOXNLPOISSON_RHS = Ucompconst (mesh,ones(Nnodes,1),coeffe);
end
totcharge = zeros(Nnodes,1);
totcharge(Sinodes)=(n - p - D);
Tv0   = totcharge;
T0    = Tv0 .* DDGOXNLPOISSON_RHS;

%% now we're ready to build LHS matrix and RHS of the linear system for 1st Newton step
A 		= DDGOXNLPOISSON_LAP + M;
R 		= DDGOXNLPOISSON_LAP * V  + T0; 

%% Apply boundary conditions
A (Dnodes,:) = [];
A (:,Dnodes) = [];
R(Dnodes)  = [];

%% we need $$ \norm{R_1} $$ and $$ \norm{R_k} $$ for the convergence test   
normr(1)	=  norm(R,inf);
relresnorm 	= 1;
reldVnorm   = 1;
normrnew	= normr(1);
dV          = V*0;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
%% START OF THE NEWTON CYCLE
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
for newtit=1:maxit
  if (verbose>0)
    fprintf(1,'\n***\nNewton iteration: %d, reldVnorm = %e\n***\n',newtit,reldVnorm);
  end

%  A(1,end)=realmin;
  dV(Varnodes) =(A)\(-R);
  dV(Dnodes)=0;
  
  
  %%%%%%%%%%%%%%%%%%
	%% Start of th damping procedure
	%%%%%%%%%%%%%%%%%%
	tk = 1;
	for dit = 1:dampit
		if (verbose>0)
			fprintf(1,'\ndamping iteration: %d, residual norm = %e\n',dit,normrnew);
		end
		Vnew   = V + tk * dV;
		
		n = exp(Vnew(Sinodes)-Fn);
		p = exp(-Vnew(Sinodes)+Fp);
		n(SiDnodes) = nin(SiDnodes);
		p(SiDnodes) = pin(SiDnodes);
		
		
		%%%
		%%% Compute LHS matrices
		%%%
		
		%% let's compute  FEM approximation of $$ L = -  \frac{d^2}{x^2} $$
		%L      = Ucomplap (mesh,ones(Nelements,1));
		
		%% compute $$ Mv =  ( n + p)  $$
		%% and the (lumped) mass matrix M
		freecarr=zeros(Nnodes,1);
		freecarr(Sinodes)=(n + p);  
		Mv   =  freecarr;
		M       =  DDGOXNLPOISSON_MASS*spdiag(Mv);
		
		%%%
		%%% Compute RHS vector (-residual)
		%%%
		
		%% now compute $$ T0 = \frac{q}{\epsilon} (n - p - D) $$
		totcharge( Sinodes)=(n - p - D);
		Tv0   = totcharge;
		T0    = Tv0 .* DDGOXNLPOISSON_RHS;%T0    = Ucompconst (mesh,Tv0,ones(Nelements,1));
		
		%% now we're ready to build LHS matrix and RHS of the linear system for 1st Newton step
		A 		= DDGOXNLPOISSON_LAP + M;
		R 		= DDGOXNLPOISSON_LAP * Vnew  + T0; 
		
		%% Apply boundary conditions
		A (Dnodes,:) = [];
		A (:,Dnodes) = [];
		R(Dnodes)  = [];
		
		%% compute $$ | R_{k+1} | $$ for the convergence test
		normrnew= norm(R,inf);
		
		% check if more damping is needed
		if (normrnew > normr(newtit))
			tk = tk/dampcoeff;
		else
			if (verbose>0)
				fprintf(1,'\nexiting damping cycle because residual norm = %e \n-----------\n',normrnew);
			end		
			break
		end	
	end
    
	V		            = Vnew;	
	normr(newtit+1) 	= normrnew;
	dVnorm              = norm(tk*dV,inf);
	pause(.1);
    % check if convergence has been reached
	reldVnorm           = dVnorm / norm(V,inf);
	if (reldVnorm <= toll)
		if(verbose>0)
			fprintf(1,'\nexiting newton cycle because reldVnorm= %e \n',reldVnorm);
		end
		break
	end

end

res = normr;
niter = newtit;