Вы находитесь на странице: 1из 4

HW 3 EE669

1. For oxidation in 1D, show that at small oxide thickness t ox, the oxidation is
oxide thickness independent. What is the critical thickness for transition from
reaction to diffusion limited oxidation? Speculate how can the critical
thickness be modified to ensure reaction limited oxidation occurs at higher
thickness? Conversely how can it be reduced.
2. Argue that in reaction limited oxidation, the oxidation rate is geometry
independent i.e. it is same for planar Si, cylindrical nanowire or spherical
nanoparticle. Can the above be explained easily by resistor model of
oxidation? What is the O-concentration drop across the oxide?
3. Argue that the reaction rate is larger for the same oxide thickness t ox, if the
silicon is a nanowire of radius r compared to planar Si. What would be the
reaction rate for a Si nanoparticle of radius r? Can the above be explained
easily by resistor model of oxidation? What is the O-concentration drop across
the oxide?
4. Show that error function based diffusion is a convolution of a step function
with a Gaussian impulse response.
5. Using the MATLAB code below plot the following;
Comment that the Fourier transform of the pulse input concentration profile is
reasonable.
Modify the standard deviation s by 2X and 0.5x from given value and plot the
c vs. x and c vs. w plots. Relate the s values to a given diffusion length. At
constant temperature, how much would we need to vary time relatively to
get the different diffusion lengths. At constant time, how much would we
need to vary temperate relatively to get the different diffusion lengths. At
constant time and temperature, how much would we need to vary Ea
(assuming same Do) relatively to get the different diffusion lengths assuming
we can chose different species to diffuse and hence select different Ea. Do
engineers have the flexibility to choose oxidizing species during oxidation?
The answer is yes. There are many species that we can use other than O 2.
Examples are O-radical, NO, N2O, NO2 , UV excited O-atom or O2 molecule etc.
These will all diffuse. O-ions (or other oxidizing ions) may also be used. We
will study their dynamics later.
clear all; %close all;
% This code is written by Udayan Ganguly to show that the diffusion profile
% of a step function (or any other function) in 1D is given by the
% convolution of that function with the gaussian impulse response

xstep=0.01; % setting discretization


s=0.05/xstep; %<<< Modify the standard deviation

x=[-1:xstep:1]; %defining the x-vector

y=abs(x)<0.5; % defining the square pulse initial profile i.e. y vector


w=[-1:xstep:1]./xstep; % define spatial frequency
filter=1*exp(-(w./s).^2); % define gaussian filter in frequency space
ffty=fft(y); % fourier transform on y
newF=ffty.*fftshift(filter); % convolution is equivalent to multiplication in
frequency space;
newf=(ifft(newF)); % inverse fourier transform to real space
% plotting section
figure (1); plot(x,y, 'r',x, newf, '-'); xlabel ('distance'); ylabel ('conc');
axis([-1 1 -0.2 1.2]); hold on;
figure (2); plot(w, fftshift(real(ffty)),'b.-', w, 100*(filter), 'r',w,
fftshift(newF),'k'); xlabel ('spatial frequency'); ylabel ('conc');

6. In the 2D version of the above code,


a) Run the following code with elliptical implanted profile enabled (as given
by default). Modify the diffusion length. From the graphs, read off the
decay length of doping in nm / decade (Which plot is better for this Figure 31 or Figure 41?)
b) Use the MODIFY THIS SECTION space to enable the box shaped doping
profiles. Argue that we get the LDD based source/drain structure based on
the input box profile in a symmetric way. Argue that this simulation is
reasonable if we assume that the Si surface lies at y=0 and the Si extends
for y<0 while it can be air at y>0, then the profile looks like the LDD
based source drain. Argue that at y=0 no concentration is exchanged from
y>0 to y<0 and vice versa. Are there other such lines where no
concentration is exchanged across the line? Identify if yes.
% This is a code written by Udayan Ganguly to calcualte 2D diffusion
% profiles from any input (as implanted) profile
clear all; close all;
sigma=0.01; % Diffusion length
% box shaped doping regions
xpos=1; xlim=0.2;
xpos2=1; xlim2=0.5
ypos=0; ylim=0.4;
ypos2=0; ylim2=0.04
p=2; % ellipsity
%---------% Discretize and create vectors in XY and generate grid
%---------x=[-2:0.01:2];
y=[-1:0.01:1];
[X,Y]=meshgrid(x,y);

%---------% Generate filer


%---------Zfilter=exp(-(X.^2+Y.^2)./sigma);
%---------% Choose as implanted geometry <<< SECTION TO MODIFY
%---------% for ellipse ; enable below else comment out
Z=1*(sqrt(X.^2+p*Y.^2)<0.5);
%% Enable this for box profile
%Z=1e-3+1e2*((abs(X+xpos)<xlim)&(abs(Y-ypos)<ylim))+1e2*((abs(Xxpos)<xlim)&(abs(Y-ypos)<ylim))+
0.1e2*((abs(X+xpos2)<xlim2)&(abs(Y+ypos2)<ylim2))+0.1e2*((abs(Xxpos2)<xlim2)&(abs(Y+ypos2)<ylim2));
%---------% Use convolution in spatial frequency domain in 2D
%---------smallz=fft2(Z); % Fourier transform on initial as implanted profile
smallzfilter=fft2(Zfilter); % Fourier transform of filter
small=smallz.*smallzfilter; % F*Z in spatial frequency domain
SMALL=ifft2(small); % inverse fourier transform
SMALLreal=real(fftshift(SMALL)); % taking only real parts (due to FFT command)
%---------% Plotting the results
%---------figure(1);meshc(X,Y,Z); xlabel ('x'); ylabel ('y'); zlabel ('doping
(linear)'); title ('as implanted');
figure(11);contour(X,Y,Z); colorbar; xlabel ('x'); ylabel ('y'); zlabel
('doping (linear)'); title ('as implanted');
figure(2);meshc(X,Y,real(smallz))
figure(21);meshc(X,Y,Zfilter); xlabel ('x'); ylabel ('y'); zlabel ('doping
(linear)'); title ('as implanted');
figure(22);contour(X,Y,Zfilter); colorbar; xlabel ('x'); ylabel ('y'); zlabel
('doping (linear)'); title ('as implanted');
figure(31); contour(X,Y,abs(SMALLreal)); xlabel ('x'); ylabel ('y'); zlabel
('doping (linear)'); title ('after diffusion (lin)');
figure(3);meshc(X,Y,SMALLreal); colorbar; xlabel ('x'); ylabel ('y'); zlabel
('doping (linear)'); title ('after diffusion (lin)');
figure(4);meshc(X,Y,log(abs((SMALLreal)))); xlabel ('x'); ylabel ('y');
zlabel ('doping (log)'); title ('after diffusion (log)');
figure(41);contour(X,Y,log10(abs((SMALLreal)))); colorbar; xlabel ('x');
ylabel ('y'); zlabel ('doping (log)'); title ('after diffusion (log)');

Вам также может понравиться