I am new to ABINIT and I am trying to calculate the nonlinear optical coefficients of a crystal. By following the examples in Static non-linear properties, I was able to successfully run the tutorial series such as tnlo_1.abi.
However, when I modified the input to use my own crystal structure (taken from Materials Project, mp-28264), the output file tnlo_4.abo reports the following error:
— !ERROR
src_file: m_ddb_diel.F90
src_line: 173
mpi_rank: 0
message: |
The lowest mode appears to be a “true” negative mode,
and not an acoustic mode. This precludes the computation
of the frequency-dependent dielectric tensor.
Action : likely there is no action to be taken, although you,
could try to raise your convergence parameters (ecut and k-points).
For your information, here are the four lowest frequencies :
-3.82650133E-02 -3.82398469E-02 -3.81771228E-02 -3.81586129E-02
you are calculating the NL optical properties based on phonon and dielectric responses. If the phonons are not stable, the system is not in its thermodynamical ground state and it does not make sense to proceed with the NLO calculation.
You should first relax the structure properly (check forces are well below 1.e-4 Ha/bohr, indicatively- yours are now massive > 2 Ha/bohr : see fcart). Your atomic positions are given with few digits of accuracy, so I suspect they come from MP or something. Needs to be refined, and you are also changing the xc functional, which means it may no longer be relaxed if the structure is PBE. If you have a delicate structure, you may want to run everything with PBE or PBEsol, the structure and phonons may be better, then just do the last dtset (NLO) with LDA, if the response you want is not coded with GGA (to be checked).
If the structure is relaxed, and the phonons are still imaginary, then you might have some symmetry breaking modes to look at: you are at a local maximum instead of minimum of ETOT. This may be for physical reasons (e.g. ferroelectricity) or numerical reasons if there is a breaking of the acoustic sum rule by your pseudopotentials or something else (cf tutorials).
In passing I would set shift to 0 0 0. The additional 0.5 is not a big improvement, and means Gamma is not included explicitly.