Dear all:
Could someone give some explaination about the diferences between nbandkss and nband, also between ecuteps and ecutwfn?
From papers,it is not so obvious to find some information about the nbandkss except nband which control the empty band,also
it is no so easy to find some information about ecutwfn except ecuteps which controls the dimension of chi^0_KS.
Any idea would be great!
Yuanxun
difference:(nbandkss) and (ecuteps Vs. ecutwfn)
Moderators: gmatteo, bruneval, rigna
Re: difference:(nbandkss) and (ecuteps Vs. ecutwfn)
Dear Yuanxun,
nband is the number of bands to be used to calculate chi_0 if optdriver=3 or to calculate \Sigma_c if optdriver=4.
nbandkss is the number of bands to be stored in the wavefunction file called KSS file.
So that nband is always lower than nbandkss.
ecutwfn is the cutoff energy used to truncate the wavefunction expansion in planewaves. A good policy if your computer is fast enough is to constrain ecutwfn=ecut.
regards,
Fabien
nband is the number of bands to be used to calculate chi_0 if optdriver=3 or to calculate \Sigma_c if optdriver=4.
nbandkss is the number of bands to be stored in the wavefunction file called KSS file.
So that nband is always lower than nbandkss.
ecutwfn is the cutoff energy used to truncate the wavefunction expansion in planewaves. A good policy if your computer is fast enough is to constrain ecutwfn=ecut.
regards,
Fabien

 Posts: 138
 Joined: Sat Aug 15, 2009 12:45 am
Re: difference:(nbandkss) and (ecuteps Vs. ecutwfn)
Hello,
One comment to Fabien's answer is that when you construct the KSS file in a ground state calculation (so with optdriver 0 instead of optdriver 3 or 4), the nband parameter is the number of bands you use in your ground state calculations (the states you want to define) and nbandkss is the number of bands you will store in the KSS file. Usually the last unoccupied bands are difficult to converge so you use nbandkss less than nband in this case and you set nbdbuf to the difference of the two.
For example, you will use nband 30 with nbdbuf 4 and nbandkss 26 with some tolerance parameter that will be checked only on the first (nband  nbdbuf) bands to generate your KSS file.
For the GW calculations, as Fabien said, nband is always less than nbandkss.
David
One comment to Fabien's answer is that when you construct the KSS file in a ground state calculation (so with optdriver 0 instead of optdriver 3 or 4), the nband parameter is the number of bands you use in your ground state calculations (the states you want to define) and nbandkss is the number of bands you will store in the KSS file. Usually the last unoccupied bands are difficult to converge so you use nbandkss less than nband in this case and you set nbdbuf to the difference of the two.
For example, you will use nband 30 with nbdbuf 4 and nbandkss 26 with some tolerance parameter that will be checked only on the first (nband  nbdbuf) bands to generate your KSS file.
For the GW calculations, as Fabien said, nband is always less than nbandkss.
David