YPP avehole produces no volumetric XSF/Cube output for a large 2D conjugated system

[qipan]

I am observing a reproducible issue with the YPP avehole + wavefunction real-space output for a relatively large periodic 2D conjugated organic framework (72 atoms/unit cell, referred to as TpPa-1 in my calculations).

I first observed the problem with Yambo/YPP 5.2.4 on the DTU cluster. I then repeated the same calculation using my own Yambo/YPP 5.3.0 installation, and obtained the same behaviour.

For this system, YPP successfully reads the BSE eigenstates, constructs a 75 × 75 × 12 real-space mesh, processes the requested excitonic state, reaches 3D Plot [100%], and terminates with Game Over.

However:

avehole + Format="c" produces no volumetric Gaussian-Cube grid;
avehole + Format="x" produces no DATAGRID_3D.

As an A/B control, using the same BSE calculation, same excitonic state, same FFTGvecs=30 Ry, and the same real-space mesh, but replacing avehole with a fixed hole position:

fixed-hole + Cube produces a complete standard Gaussian Cube;
fixed-hole + XSF produces a complete standard ASCII XSF with DATAGRID_3D.

Therefore, the BSE database, wavefunctions, FFT grid and the general XSF/Cube output routines appear to be functional. The failure seems specifically associated with the avehole real-space-output path.

The YPP report for the larger system shows approximately:

154923 wavefunction components
1536 BSS eigenstates

I also tested the small official hBN-2D example. In that case, avehole does generate non-empty real-space data. However, with the DTU YPP 5.2.4 build (MPI + OpenMP + SLK + HDF5_MPI_IO), files requested with Format="x" or Format="c" are written as HDF5 containers with .xsf / .cube extensions, rather than standard ASCII XSF / Gaussian-Cube files. The corresponding real-space datasets (xsf_values / cube_values) are present and non-empty.

I would therefore like to ask:

Is there any known issue or size-related limitation in the avehole real-space-density path?
Could the behaviour depend on the number of BSE transitions/eigenstates?
Is there a recommended alternative way to obtain averaged electron/hole densities from an existing BSE calculation?
For the hBN case, is the HDF5-backed .xsf/.cube output expected with an HDF5 MPI I/O build?

[qipan]

In 5.3.0 the avehole run terminates normally and creates the output containers, but the volumetric datasets are not written. In the XSF case only header-related HDF5 datasets are present; in the Cube case the HDF5 file contains only the root group.

[Daniele Varsano]

Dear Qipan,

let me ask to sign your post with your name and affiliation, you can do once for all by filling your signature profile. This is a rule of the forum.

Next, please post your question in the correct sub-forum, in this case post processing ypp and not compilation.

We will have a look at the issue, anyway can you post your input and report file? You can attach in the post by rename the files as .txt

Finally, the hole averaged wavefunction (as averaged in the whole space) is a periodic function and does not provide you information on the exciton localisation, I would say that it is useful for molecules but less informative that the fixed hole for periodic systems.

Best,

Daniele