Difference between revisions of "CPMD"
Jump to navigation
Jump to search
import>Jss43 |
import>Jss43 |
||
(2 intermediate revisions by the same user not shown) | |||
Line 5: | Line 5: | ||
* &QMC and &SYM sections to the input file. |
* &QMC and &SYM sections to the input file. |
||
* Calculate symmetry of the states. |
* Calculate symmetry of the states. |
||
− | * Compute integrals of the type <math> |
+ | * Compute integrals of the type <math>\langle i|v_{xc}|j \rangle</math> and <math>\langle ij|kl \rangle</math>. |
* Integration with NECI code (mainly qmc* subroutines, plus antisymint.F). NECI code resides in the NECI subdirectory. |
* Integration with NECI code (mainly qmc* subroutines, plus antisymint.F). NECI code resides in the NECI subdirectory. |
||
+ | Current work focuses on the implementation of the above with k-point sampling. |
||
− | = Input file options = |
||
+ | --[[User:jss43|james]] 14:22, 14 December 2006 (GMT) |
||
⚫ | |||
− | The following are in addition to the options given in the manual for CPMD v3.11.1. |
||
− | + | = Input file options = |
|
⚫ | |||
− | * QMC |
||
− | Run NECI code using Kohn-Sham orbitals generated by CPMD. Need to perform a wavefunction optimization calclulation first---requires the RESTART file and the QMC input file (default filename: inp.) |
||
+ | --[[User:jss43|james]] 14:22, 14 December 2006 (GMT) |
||
− | * 4-INDEX INTEGRALS |
||
− | Compute the integrals <math><i|v_{xc}|j></math> and <math><ij|kl></math> if not doing a QMC run. (Calculated on the fly or as an initialisation for QMC jobs.) |
||
⚫ | |||
− | * RESTART SYMMETRY 2-INDEX 4-INDEX |
||
− | Read in the symmetry, <math>v_{xc}</math> and 4-index integrals respectively, from the files SYMMETRY, 2INDEXINT and 4INDEXINT. |
||
+ | Please add! |
||
− | == &QMC section == |
||
− | |||
− | Specify options relevant to a QMC job. |
||
− | |||
− | * INPUT FILE <filename> |
||
− | QMC input filename. |
||
− | |||
− | * MEMORY ALLOCATE <Memory in MB> |
||
− | Specify the size of memory block to use to store wavefunctions in real space in MB. Default is 500MB. |
||
− | |||
− | * INTEGRALS FAST |
||
− | Compute the four-index integrals quicker using a coarser mesh (currently not working---don't use). |
||
− | |||
− | * INTEGRALS PRINT |
||
− | Print out the 'non-special' <math><ab|cd></math> integrals to the file abcdint. |
||
− | |||
− | == &SYM section == |
||
− | |||
− | Options for symmetry code. |
||
− | |||
− | * CALC SYM [ALL,OFF] |
||
− | ALL: Calculate characters due to rotational symmetry for all operations rather than by classes (default). If the system is non-symmorphic, then characters are found for each operation automatically. |
||
− | |||
− | OFF: Turns symmetry off (by setting each state to be totally symmetric). |
||
− | |||
− | * INTERPOLATE <NSHELLS> |
||
− | A symmetry operation may not necessarily move us from one mesh point to another. A Lagrange interpolation process is used to find the wavefunction by using NSHELLS (default 1) of boxes around the point (i.e. an interpolation of order 2*NSHELLS-1). The scaling is not so nice, but you shouldn't really need to ever go above 2 (and 1 is almost always good enough). |
||
− | |||
− | * INT STEP <SYMINTSTEP> |
||
− | The calculation of characters by <psi|R|psi> over the real space mesh is fairly costly. Integrating using a step size of SYMINTSTEP (default 1) can speed things up significantly for large high-symmetry systems. Integrating using every other point (INT STEP 2) is a good value---there appears to be little loss in accuracy. SYMINTSTEP=4 is less good (i.e. really shouldn't be used). A check is made the the specified value is consumerate with the real space mesh (if it isn't, then SYMINTSTEP=1 is set). |
||
− | |||
− | * TRANS SYM OFF |
||
− | Turn off translational symmetry for supercells. (Not applicable if space group symmetry is used.) |
||
− | |||
− | * SPACE GROUP [OFF] |
||
− | Use rotational and translational symmetry (if applicable) separately, rather than all the symmetry operations of the space group. |
||
− | |||
− | * DEGEN TOL <DEGEN TOL> |
||
− | Specify the tolerance for two states to be considered degenerate. Default is 1.d-6. |
||
− | |||
− | == &QMC section == |
||
− | |||
− | Specify options relevant to a QMC job. |
||
− | |||
− | * INPUT FILE <filename> |
||
− | QMC input filename. |
||
− | |||
− | * MEMORY ALLOCATE <Memory in MB> |
||
− | Specify the size of memory block to use to store wavefunctions in real space in MB. Default is 500MB. |
||
− | |||
− | * INTEGRALS FAST |
||
− | Compute the four-index integrals quicker using a coarser mesh (currently not working---don't use). |
||
− | |||
− | * INTEGRALS PRINT |
||
− | Print out the ''non-special'' <math><ab|cd></math> integrals to the file abcdint. |
||
− | |||
− | &SYM section: |
||
− | Options for symmetry code. |
||
− | |||
− | * CALC SYM [ALL,OFF] |
||
− | ALL: Calculate characters due to rotational symmetry for all operations rather than by classes (default). If the system is non-symmorphic, then characters are found for each operation automatically. |
||
− | |||
− | OFF: Turns symmetry off (by setting each state to be totally symmetric). |
||
− | |||
⚫ |
Latest revision as of 12:03, 31 January 2007
Local Modifications
Local modifications to CPMD:
- &QMC and &SYM sections to the input file.
- Calculate symmetry of the states.
- Compute integrals of the type <math>\langle i|v_{xc}|j \rangle</math> and <math>\langle ij|kl \rangle</math>.
- Integration with NECI code (mainly qmc* subroutines, plus antisymint.F). NECI code resides in the NECI subdirectory.
Current work focuses on the implementation of the above with k-point sampling.
--james 14:22, 14 December 2006 (GMT)
Input file options
Many additional options are now available in the local code. These are covered in the LOCAL_MODS file in the repository and in CPMDInput.
--james 14:22, 14 December 2006 (GMT)
Technical tricks
Please add!