4.8 Details of ab initio calculations

USPEX employs a powerful two-level parallelization scheme, making its parallel scalability exemplary. The first level of parallelization is performed within structure relaxation codes, the second level of parallelization distributes the calculation over the individuals in the same population (since structures within the same generation are independent of each other).

First, you must specify which code(s) you want to use for structure relaxation and fitness calculation:

$\triangleright$ variable abinitioCode

Meaning: Defines the code used for every optimization step.

Default: 1 for every optimization step (VASP)

Format:

% abinitioCode 3 2 2 1 1 % ENDabinit

Alternative Format:

% abinitioCode 1 1 1 1 1 (14) % ENDabinit

The difference is that here one (or more) stage(s) of calculation for each structure is shown in parentheses. This is very useful when optimizing physical properties: for calculations in parentheses, the structure is symmetrized and represented in the standard crystallographic setting; this allows us to fully use crystal symmetry and make property calculations cheaper and more numerically robust. Before symmetrization and property calculations, the structure must be well relaxed (so the user must make sure that relaxation is well done).

Note 1: the enthalpy is taken from the last stage BEFORE parentheses.

Note 2: Numbers indicate the code used at each step of structure relaxation:

1 — VASP
2 — SIESTA
3 — GULP
4 — LAMMPS
5 — ORCA
6 — DMACRYS
7 — CP2K
8 — Quantum Espresso
9 — FHI-aims
10 — ATK
11 — CASTEP
12 — Tinker
13 — MOPAC
14 — BoltzTraP
15 — DFTB
16 — Gaussian
(only for clusters)
18 — Abinit
19 — CRYSTAL

$\triangleright$ variable KresolStart

Meaning: Specifies the reciprocal-space resolution for $k$ -points generation (units: $2\pi \text {\r{A}}^{-1}$ ).

Default: from 0.2 to 0.08 linearly

Format:

% KresolStart 0.2 0.16 0.12 0.08 % Kresolend

NOTE: You can enter several values (one for each step of structure relaxation), starting with cruder (i.e., larger) values and ending with high resolution. This dramatically speeds up calculations, especially for metals, where very many k-points are needed. This keyblock is important for ab-initio calculations (through some codes, e.g. VASP and SIESTA, now have similar tricks)).

For clusters, 2D-crystals, and surfaces, you have to specify the thickness of the vacuum region around the cluster (or around the surface slab):

$\triangleright$ variable vacuumSize

Meaning: Defines the amount of vacuum added around the structure (closest distance in $\text {\r{A}}$ between neighboring clusters in adjacent unit cells). Used only for surfaces, 2D-crystals, and nanoparticles.

Default: 10 for every step of relaxation

Format:

% vacuumSize 10 10 15 20 20 % EndVacuumSize

$\triangleright$ variable numParallelCalcs

Meaning: Specifies how many structure relaxations you want to run in parallel.

Default: 1

Format:

10 : numParallelCalcs

You need to supply the job submission files or the names of executable files for each code/mode you are using.

$\triangleright$ variable commandExecutable

Meaning: Specifies the name of the job submission files or executables for a given code.

Default: no default, has to be specified by the user.

Format:

% commandExecutable gulp < input > output mpirun -np 8 vasp > out mpirun -np 8 vasp > out mpirun -np 8 vasp > out % EndExecutable

NOTE: Every line corresponds to a stage of relaxation — the first line describes the execution of the first stage of relaxation, etc. For example, abinitioCode equal to “3 1 1 1” means that the first relaxation step will be performed with GULP, while the subsequent steps will be performed using VASP via the command “mpirun -np 8 vasp > out”. If only one line is present in commandExecutable, then the same execution will be performed for all steps of relaxation.

You can actually use USPEX on practically any platform in the remote submission mode. In that case, your workstation will prepare input (including jobs), send them to the remote compute nodes, check when the calculations are complete, get the results back, analyze them, and prepare new input. The amount of data being sent to and fro is not large, so the network does not need to be very fast. Job submission is, of course, machine-dependent.

$\triangleright$ variable whichCluster

Meaning: Specifies the types of job submission.

Possible values (integer):

0 — no-job-script;
1 — local submission;
2 — remote submission.

Default: 0

Format:

1 : whichCluster

$\triangleright$ variable remoteFolder

Meaning: Folder on the remote supercomputer where the calculation will be performed. This keyword is activated only if whichCluster=2.

Default: none

Format:

Blind_test : remoteFolder

NOTE: there is a similar parameter specified in the remote submission file — homeFolder. The actual path to the calculation will be /*remoteFolder*/CalcFolderX where X=1, 2, 3,….

$\triangleright$ variable PhaseDiagram

Meaning: Enables calculation of a phase diagram for calculationType=300 and 301. It gives an idea (crude one — just to get a rough idea!) of which structures may be stable at higher and lower pressures, and a rough idea of transition pressures.

Default: 0

Format:

1 : PhaseDiagram