1. PSEUDOPOTENTIALS

Reference (a bit old but explains how they are generated: 
Bachelet, Hamann, Schluter, Phys Rev B 26, 4199 (1982)

What is the difference between the "old" nonlocal pseudopotentials and 
the "new" ultrasoft pseudopotentials?  What are the energy cutoffs 
associated with each?

			COARSE	MEDIUM	FINE
	H_00.recpot	400	600	800
	H_00.usp	200	270	340
	Si_00.recpot   	125	200	300
	Si_00.usp	 80	120	160	
	O_00.recpot	330	450	500
	O_01.recpot	600	750	900
	O_02.recpot	620	780	900
	O_03.recpot	620	780	900 
	O_00.usp	260	300	340


COMMENTS IN THE H_00.recpot FILE:
h2001 - very old..
This is a pure (1/r) potential of Hydrogen. It comes from many people's 
experience that using an pseudopotential for H do not necessarily
have real advantage.
Convergence testing (5x5x5 A box, LDA).
Eunrel refers to 0.765 bond length (ADF result).
===================================
Ecut    E unrel     Erel       Bond
200    -29.67023  -29.70076   0.813
400    -30.45969  -30.48472   0.804
600    -30.71480  -30.72192   0.787
800    -30.82470  -30.82791   0.780
1000   -30.88436  -30.88717   0.779
1200   -30.92240  -30.92518   0.779

COMMENTS IN THE H_00.usp FILE:
Ultrasoft potential generated using the setting
suggested by Prof. Lee group (H_mhl_01).
H2 dimer, orthorombic cell, a=6.05, b=5.95, c=6.00 Angstrom
Fractional coordinates: 
(0.612294   0.622585   0.617397) and (0.544730   0.553885   0.549270)
=============================================================
Ecut           Etot       dE            Force on atom 1
(eV)           (eV)    (eV/atom)            (eV/A)
-------------------------------------------------------------
200 (COARSE)  -30.540    0.324    1.18260   1.12637   1.08127
220           -30.652    0.268    1.19001   1.15062   1.19479
270 (MEDIUM)  -30.805    0.191    1.09293   1.08480   1.09594
280           -30.846    0.171    1.09470   1.09690   1.09281
320           -30.932    0.128    1.11031   1.11356   1.11282
340 (FINE)    -30.965    0.111    1.11338   1.11467   1.10942
380 (PRECISE) -31.032    0.078    1.13022   1.13198   1.12783
400           -31.048    0.070    1.13129   1.13361   1.13356
450           -31.097    0.045    1.13282   1.13608   1.13237
800           -31.187             1.06065   1.06116   1.06096
=============================================================
                         Validation test
                         ---------------
#1 H2 dimer, exp. bond length 0.7414, CASTEP (GGA, PRECISE)
gives 0.7422 (+0.1%)






 
2.  Choice of Supercell

Build a model for the Si(100) surface.
First, load the bulk Si model:
	load model
	Cerius2-Models
	semiconductors
	Si.msi

For an accurate structure, you would then do a geometry
optimization on the bulk, allowing the cell parameters to 
vary so as to optimize the Si-Si bond length.   
The cell parameter that results from a geometry optimization
of bulk Si using GGA  and a 500 eV cutoff is 5.395693 Angstroms.

Next, build a slab from the bulk:
	click on surface builder
	specify 100 
	click on the plus sign to get a new model window
	now when you click "cleave" you will see the result
	you can click "cleave" over and over until you like what you get
	
How many layers to you want to have in the slab?
Are you going to have to use hydrogens to terminate dangling bonds?
(How many layers are you going to keep "frozen" in the bulk positions?
How many Si atoms do you want in each layer?
How much vaccuum will you need between slabs?

By default, the supercell here has only 1 Si atom per layer
If you are going to want to build the reconstructed Si(100)-(2x1)
surface (or other periodicities) you will probably want more Si
atoms per layer.  To build a supercell that will contain one Si
dimer, try the following:

	specify 6 A depth (gives 5 Si layers)
	specify (1,2) for the surface cell display range
	click non-periodic superstructure
	switch to crystal builder
	click build crystal
	alter the cell parameters so that bond
	lengths between Si atoms in neighboring
	cells are identical to bond lengths between
	Si atoms in the same supercell (use visualization)
	Use the measuring tool to figure out what cell parameter
	to specify for "a" in order to get the amount of vaccuum
	you want - generally 6-9 Angstroms is enough. 




Measuring a bond length across supercells



When you have appropriate cell parameters (for bond lengths 
and also for the right amount of vaccuum) you can put on the
hydrogens to terminate the dangling bonds on the bottom and 
do the reconstruction to make a surface Si-Si dimer. 

Adding the hydrogens to terminate the bottom of the slab:
	Click "unbuild crystal".
	Open the sketcher (from the build pull down menu).
	Click on the button labeled "Hydrogen".
	now click on the Si atoms that you wish to terminate with H
	It will add only two to some of them but three to others -
	you will need to delete one from those Si that got three and
	then adjust the angles.
	Two the angles, open the bond geometry box from the Move 
	pull down menu.
	Measure the bond torsion for a sequence of four atoms, 
	finishing with one of the hydrogens that is in the correct
	place. 
	Now do the same thing for the H atoms that you want to move
	change the torsional angle to the desired amount - it should
	move the atoms that you clicked last. 
	(To select more than one atom, hold the shift button while i
	you click). 


Making the Si-Si dimer for the reconstructed surface:
        Unbuild the crystal (if it is not already done).
	Pressing "ctrl" and the middle mouse button lets
	you move any selected atom or atoms in the plane of 
	the screen. 
	To control what plane you are moving the atoms in, it
	is helpful to orient the model with respect to the screen. 
	Simulataneously press "home", "alt" and "r" on the keyboard.
	(That will bring the model into a position where the supercell
	boundary planes are orthogonal to the viewing plane).
	To add the dimer bond, use the 3-D sketcher. 
	Rebuild the crystal and you're ready for a calculation using castep. 

Note:

	If you want to really use this model to do a geometry optimization, 
	here are some values that are pretty close to what the minimum
	energy structure might be:
	Cell Parameters:
	a=7.630600
	b=3.815300
	c=16.000000
	Si-Si dimer length 2.3 Angstroms
	Si-Si "back bond" length 2.4 Angstroms
	(the "back bond" is between the Si dimer atoms and the next lower layer)
	Si-H bond length for terminating hydrogens: 1.35 Angstroms should be ok. 


Non-periodic model
Periodic Supercell