BS Seg. Fault

[mdvorak]

Dear Claudio,
Thank you for your help. When I issued "yambo -b -o b -y d -V resp," no variable denoting the electron-hole energy range appeared with respect to the BS kernel. There were ranges related to Xp and Xs, and changing these did not change the matrix size. I think my Abinit input is very standard and is posted below.

# Crystalline silicon
#
chksymbreak 0
ndtset 2
istwfk2 110*1
symmorphi 0


kptopt 1
ngkpt 4 4 4
nshiftk 4
shiftk 0.5 0.5 0.5
0.5 0.0 0.0
0.0 0.5 0.0
0.0 0.0 0.5

toldfe 1.0d-12

# Print the KSS file
iscf2 -2
getden2 -1
tolwfr2 1.0d-10
nband2 150
nbandkss2 150

#Definition of the unit cell
acell 3*7.2125683
rprim 0.000000 0.707107 0.707107
0.707107 0.000000 0.707107
0.707107 0.707107 0.000000

#Definition of the atom types
ntypat 1
znucl 14

#Definition of the atoms
natom 2
typat 1 1
xred
0.0 0.0 0.0
1/4 1/4 1/4

#Definition of the planewave basis set
ecut 20.0
ecutsm 1.0d-7
nband1 150

#Definition of the SCF procedure
nstep 50
diemac 12.0

I think the solution to my problem may be QP interpolation. I have experimented with this, and increasing the qpoints decreases the binding energy. Is the following the proper way to implement the interpolation?

KfnQPdb= "E < ./SAVE/ndb.QP <iu"

Then I simply change the number of Qpnts from the default to a larger value? This seems to be working. Can you elaborate a bit on how the interpolated values are implemented in the BS calculation? Is there a way to interpolate only near the band edges or must it be done uniformly across the BZ?

EDIT: I am not using sizes for NGsBlkXp and BSENGBlk of 1 either, though this is shown in my old input files. I have not rigorously tested their convergence, but am using values of near 100 for each of these.

Thank you,
Marc

[claudio]

Dear Mark

1) try the Yambo version that is now on subversion (revision 18) where there is the variable BSEEhEny

2) do not confuse QP = quasi-particle with the q-points, the interpolation is for quasi-particles not q-points

3) try to do the calculation with only one grid

nshiftk 1
shiftk 0.0 0.0 0.0

and increase the number of k-points to see if you converge.
You can use the multi-shift grid for the density, this is ok

Claudio