The indices in the EIG file does _not_ correspond to the indices in the wavefunctions (only for spin-up).
The indices are as follows (lets say 10 eigenvalues per spin, 20 in total): siesta.K1.WF1.UP.* are the first spin-up wavefunction (EIG index 1) siesta.K1.WF1.DOWN.* are the first spin-down wavefunction (EIG index 11) siesta.K1.WF2.UP.* are the second spin-up wavefunction (EIG index 2) siesta.K1.WF2.DOWN.* are the second spin-down wavefunction (EIG index 12) and so on. I hope this clarifies the indexing issue. Currently siesta/denchar enables writing non-existing indices (as you note), they are however meaningless. We have already applied a patch which will be effective in the coming 4.0.2 release. 2018-05-19 0:21 GMT+02:00 Jens Hertkorn <j.hertk...@freenet.de>: > Dear all, > > my question concerns in particular the "numbering" of the wavefunctions > that one gets after using denchar to produce *.WFSX files after a completed > siesta run. I have to get the correct wavefunctions that correspond to the > highest occupied molecular orbital (HOMO) and the lowest occupied molecular > orbital (LUMO), that is- the wavefunctions with the index of the eigenvalue > directly below and directly above the fermi energy, respectively. The > problem I face is the following: Let's say my system would have 20 > eigenvalues, and the index corresponding to HOMO and LUMO are 5,6 (spin up) > and 15,16 (spin down). I would expect denchar to produce spin up > wavefunctions for the lower half (1-10) of the available indices, and spin > down wavefunctions for the upper half (11-20), since apparently that is the > way siesta stores the eigenvalues in the *.EIG file. However, denchar > produces spin up AND spin down wavefunctions for EACH index, i.e. if i were > to write out all wavefunctions with indices 1-20 I would get 40 > wavefunction files, where I should only have 20- that is 10 for spin up and > 10 for spin down. > > So now my question is: what are these other 20 wavefunctions? In > particular, what do the *.DOWN.cube files mean that denchar produces for > the LOWER half of the indices (which should only correspond to spin up > eigenvalues) and what do the *.UP.cube files mean for the UPPER half of the > indices (which should only correspond to spin up eigenvalues) > > I hypothesized that there might be "dummy" indices that denchar writes out > and I made a test with a H2O molecule that I assigned a NetChage -1 (so > that its eigenvalues are different, for different spins) to, to find out > which indices may be "dummy". The H2O gave me 23 eigenvalues per spin / 46 > in total. Then I produced all wavefunctions with indices from 0 to 100, > just to see what happens if I even exceed the number of total eigenvalues. > Denchar still produces wavefunction files that are often only filled with > zeros, but not always... Also, no combination of "dummy" I could think of > made sense with the output denchar produced. I compared all 100 > wavefunctions pairwise, whether the wavefunction data is identical or not, > but I could not identify any of the following "dummy scheme" that I could > understand: > > lower half(spin up) = upper half(spin up) and lower half(spin down) = > upper half(spin down) > > lower half(spin up) = upper half(spin down) and lower half(spin down) = > upper half(spin up). > > In particular I do not understand why denchar still produces wavefunction > files with indices that exceed the number of eigenvalues (and if so, why > the wavefunction data is not only zero!). > > If anyone could tell me what denchar is doing/ how it is numbering > wavefunctions and which ones are dummy/ which ones not, I would be > extremely grateful since I cannot find any documentation about this and > also nothing on the mail archive. > > Sorry if this text was to long, here again in short: > > Denchar produces spin polarized wavefunction output I do not expect / > understand. Denchar can produce more wavefunction files than indices of > wavefunctions exist in a given system and I can't identify a dummy indexing > scheme. Denchar does not order spin up / spin down wavefunction output in > terms of the index in a way I can understand. > > > Regards, > > B. > > > > > > > -- Kind regards Nick
