how to run with the same pseudos in siesta and abinit

How to run with the same pseudos in Siesta and Abinit

Objectives

Run examples with the same pseudos (same decomposition in local part and Kleinman-Bylander projectors) in Siesta and Abinit.

Compare total energies

Download the last versions of both codes, Siesta and Abinit

Regarding the Abinit code, you can download the required version from:

http://personales.unican.es/junqueraj/Abinit.tar.gz

But the merge of the relevant subroutines into the main trunk will be done soon

Regarding Siesta, the code is available at the usual web site:

http://www.icmab.es/siesta

A few modifications to be done before running:Siesta

Edit the file atom.F in the Src directory

1. Replace “nrval” by “nrwf” in the call to schro_eq inside the subroutine KBgen (file atom.F)

2. Replace “nrval” by “nrwf” in the call to ghost inside the subroutine KBgen (file atom.F)

3. Increase the default of Rmax_kb_default to 60.0 bohrs (file atom.F)

Definition of the Kleinman-Bylander projectors

The normalized Kleinman-Bylander projectors are given by

where

For the sake of simplicity, we assume here only one projector per angular momentum shell. If more than one is used, they must be orthogonalized

and are the eigenstates of the semilocal pseudopotential (screened by the pseudovalence charge density).

X. Gonze et al., Phys. Rev. B 44, 8503 (1991)

Note that these are the radial part of the wave function multiplied by ,

Definition of the Kleinman-Bylander projectors (old choice in Siesta for the atomic eigenstates)

In the standard version of Siesta, the Schrödinger equation for the isolated

atom while generating the KB projectors is solved inside a box

whose size is determined by nrval.This is usually a very large radius (of

the order of 120 bohrs)

Then, this wave functions is normalized inside a sphere of a much

smaller radius, determined by Rmax_KB (default value = 6.0 bohr)

Definition of the Kleinman-Bylander projectors (new choice in Siesta for the atomic eigenstates)

In the new version of Siesta, everything is consistent, and the Schrödinger equation and the normalization are

solved with respect the same boundary conditions

Almost no change in total energies observed, but the Kleinman-Bylander energies might be very different, specially for unbounded orbitals

A few modifications to be done before running:Abinit

Edit the file src/65_psp/psp5nl.F90

1. Uncomment the last two lines for the sake of comparing Kleinman-Bylander energies and cosines with the ones obtained with Siesta

./configure --with-trio-flavor=netcdf+etsf_io+fox

2. Remember to compile the code enabling the FOX library

Examples to run Siesta and Abinit with the same pseudos

1. Visit the web page:http://personales.unican.es/junqueraj

and follow the links:TeachingMétodos Computacionales en Estructura de la MateriaHand-on sessionsPseudos

2. Download the Pseudos and input files for both codes

3. Untar the ball file$ tar –xvf Siesta-Abinit.tar

This will generate a directory called Comparison-Siesta-Abinit with 4 directories:$ cd Comparison-Siesta-Abinit$ ls -ltr$ Si (example for a covalent semiconductor, LDA)$ Al (example for a sp-metal, LDA)$ Au (example for a noble metal, includes d-orbital, LDA)$ Fe (example for a transition metal, includes NLCC, GGA)

Examples to run Siesta and Abinit with the same pseudos

In every subdirectory it can be found:$ cd Si$ ls –ltr$ Pseudo (files to generate and test the pseudopotential)$ Optimized-Basis (files to optimize the basis set)$ Runsiesta (files to run Siesta)$ Runabinit (files to run Abinit)

How to generate and test a norm-conserving pseudopotential

Generate the pseudopotential using the ATM code as usual, following the notes in the Tutorial

“How to generate a norm conserving pseudopotential”

Copy the input file in the corresponding atom/Tutorial/PS_Generation directory and run

The pseudopotentials will be on the same parent directory: .vps (unformatted) (required to test the pseudopotential) .psf (formatted) .xml (in XML format) (required to run Abinit)

Remember to test the pseudopotential using the ATM code as usual, following the notes in the Tutorial

“How to test the transferability of a norm conserving pseudopotential”

Running the energy versus lattice constant curve in Siesta

Run the energy versus lattice constant curve in Siesta as usual. You can use both the .psf or the .xml pseudopotential.

Follow the rules given in the tutorial“Lattice constant, bulk modulus, and equilibrium energy of solids”

The input file has been prepared for you (file Si.fdf).Since we are interested in compare the performance of the basis set, it is

important to converge all the rest of approximations (Mesh Cutoff, k-point grid, etc.) as much as possible

At the end, we would be able to write a file (here called Si.siesta.latcon.dat) that looks like this:

These data have been obtained with a double-zeta plus polarization basis

set, optimized at the theoretical lattice constant with a pressure of 0.05 GPa

Running the energy versus cutoff energy in Abinit

To check the equivalent cutoff energy in Abinit:

1. We run the same system (same lattice vectors and internal coordinates) at the same level of approximations (same exchange and correlation functional, Monkhorst-Pack mesh etc.) at a given lattice constant.

Here it has been written for you (file Si.input.convergence)

Diamond structure at the lattice constant that minimizes the energy in Siesta

6 × 6 × 6 Monkhorst-Pack mesh

Ceperley-Alder (LDA) functional

Running the energy versus cutoff energy in Abinit

To check the equivalent cutoff energy in Abinit:

1. We run the same system (same lattice vectors and internal coordinates) at the same level of approximations (same exchange and correlation functional, Monkhorst-Pack mesh etc.) at a given lattice constant.

Here it has been written for you (file Si.input.convergence)

2. Change the cutoff energy for the plane waves

4. Run the code

3. Edit the .files file and select the input file and the pseudo file (in XML format)

Running the energy versus cutoff energy in Abinit: bulk Si (covalent semiconductor)

Write the total energy as a function of the cutoff energy and edit the corresponding file that should look like this

Equivalent PW cutoff a DZP basis set at 5.38 Å

Running the energy versus lattice constant curve in Abinit

1. Same input as before but…

… setting the plane wave cutoff to the equivalent one to a DZP basis set

… and changing the lattice constant embracing the minimum

2. Change in the .files the name of the input file

3. Run the code

Running the energy versus lattice constant curve in Abinit

Write the total energy as a function of the lattice constant and edit the corresponding file that should look like this

Comparing the pseudopotential in Siesta and Abinit: bulk Si (covalent semiconductor)

To be totally sure that we have run Siesta and Abinit with the same peudopotential operator, i.e. with the same decomposition in local part and

Kleinman-Bylander projectors:

1. Edit one of the output files in Siesta and search for the following lines:

2. Edit the log file in Abinit and search for the following lines:

The Kleinman-Bylander energies and cosines should be the same upto numerical

roundoff errors

Note: In Siesta they are written in Ry and in

Abinit they are in Ha.

Comparing the energy versus lattice constant in Siesta and Abinit: bulk Si (covalent semiconductor)

Comparing the pseudopotential in Siesta and Abinit: bulk Al (sp metal)

To be totally sure that we have run Siesta and Abinit with the same peudopotential operator, i.e. with the same decomposition in local part and

Kleinman-Bylander projectors:

Comparing the pseudopotential in Siesta and Abinit: bulk Al (sp metal)

For the case of metallic system, besides the k-point sampling we have to pay particular attention to the occupation option

Siesta Abinit

Default: Fermi-Dirac

Also, as explained in the Tutorial“Convergence of electronic and structural properties of a metal with respect to

the k-point sampling: bulk Al”we should look at the Free Energy and not to the Kohn-Sham energy

Lattice constant 3.97 Å

Running the energy versus cutoff energy in Abinit: bulk Al (a sp metal)

Comparing the energy versus lattice constant in Siesta and Abinit: bulk Al (sp metal)Basis set of Siesta: DZP optimized with a pressure of 0.001 GPa at the theoretical

lattice constant of 3.97 Å)Plane wave cutoff in Abinit: 8.97 Ha

Comparing the pseudopotential in Siesta and Abinit: bulk Au (a noble metal)

To be totally sure that we have run Siesta and Abinit with the same peudopotential operator, i.e. with the same decomposition in local part and

Kleinman-Bylander projectors:

Lattice constant 4.08 Å

Running the energy versus cutoff energy in Abinit: bulk Au (a noble metal)

Comparing the energy versus lattice constant in Siesta and Abinit: bulk Au (a noble metal)

Basis set of Siesta: DZP optimized with a pressure of 0.02 GPa at the theoretical lattice constant of 4.08 Å

Plane wave cutoff in Abinit: 17.432 Ha

Comparing the pseudopotential in Siesta and Abinit: bulk Fe (a magnetic transition metal)

To be totally sure that we have run Siesta and Abinit with the same peudopotential operator, i.e. with the same decomposition in local part and

Kleinman-Bylander projectors:

Comparing the pseudopotential in Siesta and Abinit: bulk Fe (a magnetic transition metal)

For the case of metallic system, besides the k-point sampling we have to pay particular attention to the occupation option.

Now, besides:-The system is spin polarized-We use a GGA functional-We include non-linear partial core corrections in the pseudo

Siesta Abinit

Lattice constant 2.87 Å

Running the energy versus cutoff energy in Abinit: bulk Fe (a magnetic transition metal)

Comparing the energy versus lattice constant in Siesta and Abinit: bulk Fe(magnetic transition metal)

Basis set of Siesta: DZP optimized without pressure at the experimental lattice constant of 2.87 Å

Plane wave cutoff in Abinit: 34.82 Ha