ABINIT Discussion Forums

Hi ,

May I ask if there is any way to convert a dx or cube file with charge density to _DEN file.
I need to do this for preparing input files for aim to do bader.

Isn't it the other way around? You have a _DEN and want a .cube for bader? If so you can use cut3d - option 10 is OpenDx and option 14 gives you a cube, at least on the most recent public Abinit (6.12.3).

There is a fortran utility code named parchg2den.f90(http://lmtoelf.wikidot.com/hirschfeld-i-analysis) used to convert the vasp partial charge density file to abinit format.
Is there is any kind of such file to convert cube and dx file to abinit file format (DEN)

Ah sorry I thought you wanted to use *bader* (http://theory.cm.utexas.edu/bader/) to do bader analysis, not aim. As far as I know, there are no utilities to convert from a cube or dx to _DEN. It wouldn't be generally of much use to have such a utility apart from when you're trying to graft the output of one code to another without a formal interface (as here).

Since you have the cube files, my advice would be to use the bader code and not aim.

Thank you so much for your advice.

May I clear my idea in more detail, I have valence charge density in a cube format file I want to
augment it with the core one (atom.fc files generated from fhi98 code ) provided by the abinit
site and do bader analysis using aim.
Using Bader code will not help me because I only have valence density not all- electron density, But
this can be done using aim utility from abinit if I could convert the cube file to _DEN file then using aim
to do bader analysis.

I have no idea about the exact of data contained in the _DEN file but the cube file it is easy to note that
it contain the structure and the density in 3D grid.

Some of the _DEN file data are so confusing such as the rprim ,no .bands, kpoints, ecut ...etc which
can not be found in dx or cube format.

So is in theory is such conversion is applicable ?
and did aim utility use such data (i.e. rprim ,no .bands, kpoints, ecut ) ?

I will really appreciate any help or idea in this topic.

Many thanks in advance

Hi,

Ok, so you want all-electron bader. You have a couple of options. The first (probably not very helpful) is that if you generate the density via an actual abinit calculation rather than just taking the .cube file, you can get an all-electron density. To do this you need to run a PAW calculation and set pawprtden to 2 (I think). This requires some extra data files which is very annoying. Or you can use pawprtden 1 which at least gives you the on-site PAW contributions in the density so you don't just have the pseudovalence charge.

If you really insist on taking the charge from this .cube file you have and augmenting it (I'd like to hear why you need to do this) with the core density files from abinit.org, you need to write a conversion utility yourself. The _DEN file format is actually fairly simple but it is an unformatted fortran binary file. The rprim, numbands, kpoints, ecut etc parameters are all part of the standard ABINIT unformatted output header as described here. I don't actually know if aim uses anything from the header - look in the aim source code - but if so, you are in a spot of bother unless you know more about your system.

The point is that a .cube file by default contains structural information and field information but no information about a *calculation*, which is what the Abinit _DEN file does contain precisely because _DEN files are not only used as outputs but also as chainable inputs in the Abinit code.

How about we back up a minute. Why don't you tell us why you want to use the .cube, what molecule/crystal it is, how many atoms, etc - ie, what physical problem are you actually working on. Maybe there's a better way entirely.

Dear Dr Kane O'Donnell,

Thank for the valuable notes.

If aim (provided by abinit) depend on the charge density
{ all-electron = valence (from _DEN) + core (from *.fc) }
and structure (from a _DEN file) then this utility will
the only one software that can help me doing bader analysis.

I am working on a linear scalling DFT code that deals with large numbers
atoms it is very fast but only outputs the charge density in .cube file format
(no ELF or WFN data ), such charge density (which is a valence density) should
be augmented with core density to do bader analysis.

I wrote my own code (in fortran )to add core to valence density and use them as
all-electron density for bader code (from Henkelman group) it just provide
bader volumes and charges such code can not help me do energy decompsition analysis.

I used some of codes that depends on experimental charge density but they give
wrong topology that can not be dependent in my work.

My next step is to move toward abinit aim code and use it to do bader analysis
I start with molecule of 10 atoms (may be increased up to 500 atoms or more),
I am working on energy decomposition analysis using AIM theory.

OK. Well, the best route might be to make a 'shadow' _DEN file - generate a simple Abinit input file for your model system (use the latest Avogadro GUI - it has a quick Abinit input file generator) and run a single SCF step. This will generate a _DEN file. Then write a fortran code to use the header from the shadow _DEN and graft it onto a new _DEN except using the augmented .cube density re-formatted into unformatted fortran binary as specified in the Abinit help file link I posted yesterday. You will need to be a bit careful about the grid densities - the safest way is probably to interpolate the .cube density onto a grid of the same density as contained in the _DEN header (the grid dimensions are given in the Abinit output file). Alternatively you could modify the header show the same grid density as the .cube grid but this might lead to computational problems if AIM uses the grid density to generate a plane wave basis or something like that (I don't know).

I have a lot of code for dealing with Abinit outputs (and other codes) at my github (github.com/kaneod). In particular, libabitools has a routine for reading a header from an Abinit unformatted binary file, and for reading the density. Also, esc_lib.py has a lot of functions for reading/writing things like Abinit input files and .cubes and so on (for python). Maybe you can use some of these to get you started. All the code is GPL'd so watch the licensing if you plan to redistribute.

Of course, by this stage it's almost at the point where you should fork AIM and re-write it's input routines to read a .cube!

Out of interest, what is the name of the O(N) code?

Sorry for late reply,

and thanks for the notes about doing such kinds of test calculations, your code looks very good start point for writing my own one I hope you don't mind to use them.
below is the link for more details about the code I use it in my calculations
http://www.sciencedirect.com/science/article/pii/S0010465508004414

Ah, yeah, it's ONETEP. Great code! Why don't you just ask Nick to write an all-electron .cube output? ONETEP does PAW these days doesn't it?

Not yet (as far as I know).