"Impossible to build the kernel"

[Daniele Varsano]

Dear Subhayan,
I've found the time to look at your files and contrary to what I told you before, the code is reading the QP correction correctly.
Looking at IP spectrum (4th column) you can see that you have low energy transition coming from the fact that you have a very different correction to the first unoccupied states (16 and 17) with respect to the subsequent states.
It seems that these states have very different character (e.g. bound and unbound states). This gives you many "mainly dark" transitions at low energy (around 3.5-4 eV) that can be the origin of the negative excitation energy once the e-h attraction is turned on.

Code: Select all

     1.00000     15.00000     -4.38046     -1.02524      1.80882      1.00000
     1.00000     15.00000     -3.99366     -1.12946      0.78070     -1.00000
     1.00000     16.00000     -0.43496     -0.83409      1.38500      1.00000
     1.00000     16.00000      0.53357      3.84547     -0.58916     -1.00000
     1.00000     17.00000      0.77064      4.11526     -0.18895      1.00000
     1.00000     17.00000      1.01088      4.00725     -0.51342     -1.00000
     1.00000     18.00000      1.15196      1.58374     -0.39048      1.00000
     1.00000     18.00000      1.21029      1.55171     -0.39459     -1.00000
     1.00000     19.00000      1.82056      1.23486     -0.31337      1.00000
     1.00000     19.00000      1.87109      1.16537     -0.30937     -1.00000
     1.00000     20.00000      1.99069      1.16519     -0.30045      1.00000

Now, looking at your scf input, I can see you are modelling a triclinic crystal of dilute "benzene" molecules, but considering only 1 k point.
There is a particular reason for this choice?
In case you need to model an isolated molecule, I strongly suggest you to consider to build a simple supercell with enough vacuum adding the assume_isolated="mt" keyword, in this way you will have the correct electrostatic and vacuum level. Next, you may want also to consider to use a cutoff coulomb interaction in yambo to avoid replica interactions and accelerate convergence wrt the volume supercell.
If, otherwise you need to model a molecular crystal, are you sure that the gamma approximation is correct?
Moreover, the size of the supercell is set by some experimental evidence, or it is just a "large enough" cell? I suspect that if you have unbound states their energies will strongly depend on the volume of your supercell.

Best,
Daniele