Projection on spin states. More...
Functions/Subroutines | |
| subroutine, private | cntrct (nelt, no, ndo, cdo, thres) |
| Throw away determinants with negligible contribution. More... | |
| subroutine, private | dophz (nftw, nocsf, nelt, ndtrf, nconf, indo, ndo, lenndo, icdo, cdo, lencdo, iphz, leniphz, iphz0, leniphz0, nctarg, nctgt, notgt, mrkorb, mdegen, ntgsym, mcont, symtyp, npflg) |
| Phase factor. More... | |
| subroutine, private | dophz0 (nftw, nocsf, nelt, ndtrf, nconf, indo, ndo, lenndo, icdo, cdo, lencdo, iphz, npflg) |
| Phase factor. More... | |
| integer function, private | iphase (nconf, nelt) |
| Sequence phase factor. More... | |
| subroutine, private | mkorbs (nob, nsym, mn, mg, mm, ms, norb, nsrb_in, map, mpos, iposit, nobl, nob0l, symtyp) |
| Compute the orbital table. More... | |
| subroutine, private | pkwf (nod, ieltp, cdo, mdo, idopl, mdop, idcpl, mdcp, nftw, ndo, ndto, len_ndto, ithis_csf) |
| Pack wave function. More... | |
| subroutine, private | pmkorbs (nob, nobe, nsym, mn, mg, mm, ms, nsrb, norb, nsrbd, map, mpos, iposit, symtyp) |
| ? More... | |
| subroutine, private | popnwf (nsrb, nsrbs, nelt, ndtrf, mopmx, mdop, mdcp, mop, mdc, mdo, ndta, nod, nda, idop, idcp, ieltp, flip, nalm) |
| Get open-shell part of determinants. More... | |
| subroutine, private | prjct (nelt, mxss, nodi, ndo, cdi, nodo, cdo, maxcdo, mgvn, iss, isd, thres, r, ndtr, mm, ms, maxndo, symtyp, nsrb) |
| Apply Lowdin projection operator. More... | |
| subroutine, public | projec (sname, megul, symtyp, mgvn, s, sz, r, pin, nocsf, byproj, idiag, npflg, thres, nelt, nsym, nob, ndtrf, nftw, iposit, nob0, nob1, nob01, iscat, ntgsym, notgt, nctgt, mcont, gucont, mrkorb, mdegen, mflag, nobe, nobp, nobv, maxtgsym) |
| Project the wave function. More... | |
| subroutine, private | ptpwf (nftw, nocsf, nelt, ndtrf, nodi, indi, icdi, ndi, cdi) |
| Print the CSFs. More... | |
| integer function, private | qsort (n, a) |
| Sort integer array. More... | |
| subroutine, private | rdwf (nft, k1, nodi, k2, cdi, k3, ndi) |
| Read CSF. More... | |
| subroutine, private | rdwf_getsize (iunit, num_csfs, num_dets, len_dets) |
| Reads the size of the wavefunction. More... | |
| subroutine, private | rfltn (nelt, nodi, ndi, cdi, r, mxnd, ndmxp, thres, nodo, cdo, ndtr, mm, bst) |
| Add mirror-reflected spin-orbitals. More... | |
| real(kind=wp) function, private | snrm2 (n, array, istep) |
| Real strided vector norm. More... | |
| subroutine, private | stmrg (nelt, maxcdo, maxndo, ndo, cdo, nodo, ndi, cdi, bst) |
| Add determinant to a list of determinants. More... | |
| subroutine, private | wfgntr (mgvn, iss, isd, thres, r, symtyp, nelt, nsyml, nob, nobl, nob0l, nobe, norb, nsrb, mn, mg, mm, ms, iposit, map, mpos, nocsf, ndtrf, nodi, ndi, cdi, indil, icdil, maxndi, maxcdi, nodo, ndo, cdo, indo, icdo, maxndo, maxcdo, lenndo, lencdo, npflg, byproj, nftw, nalm) |
| Form spin eigenstates. More... | |
| subroutine, private | wrnfto (sname, mgvn, s, sz, r, pin, norb, nsrb, nocsf, nelt, idiag, nsym, symtyp, nob, ndtrf, nodo, m, icdo, indo, ndo, lndi, cdo, lcdi, nfto, nobl, nx, npflg, thres, iposit, nob0, nob0l, nctarg, ntgsym, notgt, nctgt, mcont, gucont, iphz, nobe, nobp, nobv, maxtgsym) |
| WRNFTO - WRite wavefunction data to unit NFTO. More... | |
| subroutine, private | wrwf (nft, n1, nodo, n2, cdo, n3, ndo) |
| Write wave functions using SPEEDY format. More... | |
Detailed Description
Routines in this module, most prominently the central subroutine projec, read back all CSFs generated in the previous part of CONGEN execution and recombine the determinants to satisfy spin composition rules. While doing that, several new combinations of spin-orbitals (determinants) may be added if there are any open shells where the electrons spins can be freely permuted. Also, the routines apply a threshold for final selection of contributing determinants, so the output can be even smaller than the input (though this mostly signalizes some error in setup).
Function/Subroutine Documentation
◆ cntrct()
|
private |
Scans the store of determinants, discard such whose contribution (multiplication factor) is below given tolerance, and bubbles out the vacated intervals from the storage arrays.
Definition at line 66 of file congen_projection.f90.
Referenced by prjct(), and rfltn().
◆ dophz()
|
private |
Compute phase factor implied by placing continuum spin-orbital after all target spin-orbitals
- Note
- MAL 10/05/2011 Changes made are in order to bring dophz into line with the changes made in 'projec' in order to utilize dynamic memory
Definition at line 107 of file congen_projection.f90.
References iphase().
Referenced by projec().
◆ dophz0()
|
private |
Compute phase factor for target CSFs - given by reordering spin-orbitals in ascending order.
- Todo:
- MAL 10/05/2011: Changes have been made to this subroutine to bring it into line with the changes made to 'projec' and ensure to compliance with F95 standards
Definition at line 232 of file congen_projection.f90.
References iphase().
Referenced by projec().
◆ iphase()
|
private |
Compute phase factor (if any) due to out of sequence ordering of spin-orbitals in CSF stored in nconf.
Definition at line 283 of file congen_projection.f90.
Referenced by dophz(), dophz0(), and projec().
◆ mkorbs()
|
private |
Computes the orbital table which is then used in the projection step. This is called from projec().
Input data:
ISYMTYP Switch for C-inf-v (=0 or 1) / Abelian point group (=2
NOB Number of orbitals per symmetry
NSYM Number of symmetries in the orbital set
NPFLG Flag controlling printing of computed orbital table
Output data:
MN Orbital number associated with each spin-orbital
MG G/U designation for each spin-orbital (C-inf-v only)
Actually this is always zero because C-inf-v does not
distinguish between g/u. It exists because original
version of Alchemy tried to use it for D-inf-h too;
all CI evauation is doen in C-inf-v now because CONGE
converts D-inf-h to C-inf-v data.
MM Symmetry quantum number associated with each spin-orb
MS Spin function ( alpha or beta ) associated with each
spin orbital
Notes:
The orbital table establishes orbital and quantum number data for
each spin orbital in the set.
e.g. C-inf-v symmetry with NSYM=2, NOB=3,1, yields ten spin
orbitals which are designated as follows by this routine:
Spin orb. MN MG MM MS Comments
1 1 0 0 0 1 sigma spin up
2 1 0 0 1 1 sigma spin down
3 2 0 0 0 2 sigma spin up
4 2 0 0 1 2 sigma spin down
5 3 0 0 0 3 sigma spin up
6 3 0 0 1 3 sigma spin down
7 4 0 1 0 1 pi(lambda=+1) spin up
8 4 0 1 1 1 pi(lambda=+1) spin down
9 4 0 -1 0 1 pi(lambda=-1) spin up
10 4 0 -1 1 1 pi(lambda=-1) spin down - Note
- MAL 11/05/2011 : Changes made here are to bring the subroutine into line with the changes that were made in 'projec' in order to utilize dynamic memory and also to comply with the F95 standard
Definition at line 372 of file congen_projection.f90.
References congen_data::nftw.
Referenced by projec().
◆ pkwf()
|
private |
Reformats (packs) the CSF expression into the style used throughout the rest of Alchemy, that is as a set of replacements from the reference determinant. Adds this to the end of the array ntdo() from location "ndo".
On entry to this routine we have the CSF defined for us as follows (from the projection step):
- there are "nod" determinants
- each determinant is of length "ieltp".
- "cdo" contains the coefficient which multiplies each determinant. This is derived from the coupling process.
- the determinants are stored in "mdo", as a list of spin orbitals - so it is of lenth nod*ieltp
The above information is complemented by the analysis in the calling routine which classifies spin orbitals in this CSF wrt the reference determinant:
- "idopl" is the number of spin orbs in the reference det but not present in this CSF; "mdop()" is the list of those spin orbitals
- "idcpl" is the number of spin orbs in this CSF but not present in the reference determinant; "mdcp()" is the list of those spin orbitals.
So, given all of the above information, the determinants in "mdo" are processed and each expressed in the format
- number of replacements from ref determinant
- list of replaced spin orbitals
- list of replacing spin orbitals
The output is placed into array "ndto". The length available in "ndto" is passed into the routing in "len_ndto" and this is monitored to be sure we do not overflow it.
During the process, it may be necessary to multiply "cdo" by -1 as we order spin orbitals. cdo() holds the coefficients associated with each determinant. These were constructed earlier in the spin projection - note we only receive the relevant "piece" of the cdo(0 array in the arglist, the bit for this CSF, not all of it for all CSFs, as is the case with "ndto()". "nftw" is the logical unit for the printer
- Note
- MAL 11/05/2011: Changes made here are to bring the subroutine into line with the changes that were made in 'projec' in order to utilize dynamic memory and to comply with the F95 standard.
Definition at line 655 of file congen_projection.f90.
Referenced by wfgntr().
◆ pmkorbs()
|
private |
Definition at line 809 of file congen_projection.f90.
References congen_data::nftw.
Referenced by projec().
◆ popnwf()
|
private |
Fills the array mop with subset of specification of every determinant for the current CSF. Only spin-orbitals that are in open shells are used, as the rest does not participate in spin composition done later in prjct. The written spin-orbitals are also immediately sorted in non-descending order. The even or odd number of swaps needed for sorting is returned (as .false. and .true., respectively) via the logical array flip, so that the sign of determinant coefficients within this CSF can be adjusted to the used order.
OUTPUT IDOP NO OF SO IN DR BUT NOT IN DC
MDOP(NELT) SO IN DR BUT NOT IN DC
IDCP NO OF SO IN DC BUT NOT IN DR
MDCP(NELT) SO IN DC BUT NOT IN DR
IELTP NO OF SO IN OPEN SHELLS FOR A DTR
MOP(MOPMX) SO- Note
- In the original code there was a common block
/OWF/ IDOP,IDCP,IELTP
which was used to pass three integer values back to the caling routine. these have now been placed into the argument list; correspondingly the routine WFGNTR has been modified. It is the only routine whihc calls this one. the purpose of the variables is as follows:IDOP holds the number of entries in MDOP IDCP holds the number of entries in MDCP IELTP is the number of electrons in open shell
Definition at line 949 of file congen_projection.f90.
References qsort().
Referenced by wfgntr().
◆ prjct()
|
private |
This routine applies the Lowdin projection operator. More details can be found in the literature at for example:
Nelson F Beebe and Sten Lucil, J Phys B: At Mol Phys, Vol. 8, Issue 14, 1975, p2320
This routine is called when a CSF is found to have two, or more electrons in open shells. Each pair of spin-orbitals in each determinant is examined and potentially used to create a new determinant. Thus the output expression for the CSF
nodo, cdo(), ndo()
may be much larger than the input
nodi, cdi(), ndo()
Note reuse of ndo() here. cdi() is used an an extendable buffer too and must have more than "nodi" elements.
The generated list of determinants is examined for any with very small coefficients (thres) and these are removed. Thus the the list may grow and shrink in this routine.
Of course the nature of the projection process is controlled by the quantum numbers input.
The routine terminates with an error message if any error conditions are found.
Definition at line 1310 of file congen_projection.f90.
References cntrct(), qsort(), rfltn(), snrm2(), and stmrg().
Referenced by wfgntr().
◆ projec()
| subroutine, public congen_projection::projec | ( | character(len=80) | sname, |
| integer | megul, | ||
| integer | symtyp, | ||
| integer | mgvn, | ||
| real(kind=wp) | s, | ||
| real(kind=wp) | sz, | ||
| real(kind=wp) | r, | ||
| real(kind=wp) | pin, | ||
| integer | nocsf, | ||
| integer | byproj, | ||
| integer | idiag, | ||
| integer, dimension(6) | npflg, | ||
| real(kind=wp) | thres, | ||
| integer | nelt, | ||
| integer | nsym, | ||
| integer, dimension(nsym) | nob, | ||
| integer, dimension(nelt) | ndtrf, | ||
| integer | nftw, | ||
| integer | iposit, | ||
| integer, dimension(nsym) | nob0, | ||
| integer, dimension(*) | nob1, | ||
| integer, dimension(nsym) | nob01, | ||
| integer, intent(in) | iscat, | ||
| integer, intent(inout) | ntgsym, | ||
| integer, dimension(ntgsym) | notgt, | ||
| integer, dimension(ntgsym) | nctgt, | ||
| integer, dimension(ntgsym) | mcont, | ||
| integer, dimension(ntgsym) | gucont, | ||
| integer, dimension(ntgsym) | mrkorb, | ||
| integer, dimension(ntgsym) | mdegen, | ||
| integer, intent(in) | mflag, | ||
| integer, dimension(nsym) | nobe, | ||
| integer, dimension(nsym) | nobp, | ||
| integer, dimension(nsym) | nobv, | ||
| integer | maxtgsym | ||
| ) |
The subroutine projec controls the projection of the wavefunctions and writes out the final wavefunctions plus header information for future use.
- Note
- MAL 06/05/11 PROJEC has been considerably modified to take advantage of dynamic memory allocation.
Definition at line 1646 of file congen_projection.f90.
References dophz(), dophz0(), iphase(), mkorbs(), pmkorbs(), ptpwf(), rdwf(), rdwf_getsize(), wfgntr(), wrnfto(), and wrwf().
Referenced by congen_driver::csfgen().
◆ ptpwf()
|
private |
This provides a complete text dump of all CSFs in terms of their packed determinants.
- Parameters
-
nftw File unit for text output of the program. nocsf Number of wave functions (CSFs). nelt Number of electrons (and size of ndtrf).ndtrf Reference determinant (spinorbitals per electron). nodi Array with number of determinants per CSF. indi Array with offsets in ndiper CSF.icdi Array with offsets in cdiper CSF.ndi Packed determinants. cdi Deteminant coefficients.
Definition at line 2614 of file congen_projection.f90.
Referenced by projec().
◆ qsort()
|
private |
- Date
- 2018
Sort array in-place in non-descending order. Currently implemented as a stable insertion sort. This algorithm is advantageous for short and almost sorted arrays, which is the case for augmented open-shell-only subsets of determinants in CONGEN, typically resulting in asymptotic complexity of O(n).
- Parameters
-
n Length of the array to sort. a Integer array to sort.
- Returns
- Number of swaps done. This is needed elsewhere to update determinant signs.
Definition at line 2657 of file congen_projection.f90.
Referenced by popnwf(), prjct(), and rfltn().
◆ rdwf()
|
private |
Routine congen_distribution::wfn stores the CSF data in buffers of fixed size.
When the buffers are full, they are emptied to disk and reused. This is why we can have several sets to read here.
Definition at line 2699 of file congen_projection.f90.
Referenced by projec().
◆ rdwf_getsize()
|
private |
Reads the information giving the size of the data used for the wavefunction, for example the number of determinants, but does not read the actual data, such as the determinants.
This is really just a stripped down version of the routine rdwf which reads the full data.
- Parameters
-
iunit The logical unit on which the wavefucntion data is located. num_csfs On return, number of CSFs stored in the file. num_dets On return, number of determinants summed over all CSFs in the file. len_dets On return, total summed length of all arrays of packed determinants in the file.
- Note
- MAL 10/05/2011: This subroutine is new to congen and is included to bring congen into line with the changes that were made in projec in order to utilize dynamic memory
Definition at line 2748 of file congen_projection.f90.
Referenced by projec().
◆ rfltn()
|
private |
◆ snrm2()
|
private |
This is actually an in-house version of the BLAS level 1 routine of the same name, included here to avoid dependency on BLAS library.
- Parameters
-
n Length of the vector array Vector of real numbers istep Stride
- Returns
- Square of L2 norm of the vector.
Definition at line 2844 of file congen_projection.f90.
Referenced by prjct(), and wfgntr().
◆ stmrg()
|
private |
Given an existing list of determinants in cdo()/ndo() and a new single determinant cdi/ndi(), the new determinant is merged into the list and the list extended if necessary.
This operation is used during spin-projection of a CSF.
- Note
- MAL 16/05/11 : Modified to bring the subroutine into line with the changes made to projec
- Parameters
-
nelt Number of electrons in each det. maxcdo Dimension of cdo. maxndo Dimension of ndo. ndo List of determinants each with "nelt" spin-orbitals. cdo Coefficient for each det in ndo. nodo On input is the number of determinants in cdo/ndo. Will be updated for output if the data in cdi/ndi new entry. ndi A single det of "nelt" spin orbs which has to be merged into ndo. cdi Single coefficient going with the single determinant defined in ndi. bst Binary search tree used for fast localization of determinants.
Definition at line 2888 of file congen_projection.f90.
Referenced by prjct(), and rfltn().
◆ wfgntr()
|
private |
Takes the wavefunction generated by CSFGEN and transforms it to be fully in accord with the spin quantum numbers of the system.
- Note
- MAL 10/05/2011: The changes made to wfgntr have made so as to make the subroutine compatible with the changes made to its calling subroutine, projec.
Definition at line 3003 of file congen_projection.f90.
References pkwf(), popnwf(), prjct(), and snrm2().
Referenced by projec().
◆ wrnfto()
|
private |
- Note
- MAL 10/05/2011 : This subroutine has been changed to bring it into line with the changes that have been made to 'projec' in order to utilize dynamic memory. 'wrnfto' has also been modified in order to comply to the F95 standard
Definition at line 3615 of file congen_projection.f90.
Referenced by projec().
◆ wrwf()
|
private |
This subroutine writes all determinants in the current CSF to a binary file that has the following format:
[WP] n1, [INTEGER] nodo(1:n1)
[WP] n2, [WP] cdo(1:n2)
[WP] n3, [INTEGER] ndo(1:n3)On entry, the file will be rewindws to its beginning. On return, the output file is rewinded to its beginning, again.
- Parameters
-
nft An open binary file for output. n1 Length of nodo(number of determinants).nodo Determinant sizes. n2 Length of cdo(number of determinants).cdo Determinant factors within the CSF. n3 Length of ndo(number of integers defining the determinants).ndo Packed determinants.
Definition at line 3739 of file congen_projection.f90.
Referenced by projec().
