Hi Marcos,
thank you for the suggestion! In fact I had (naively) compiled 'atm' with
gfort and just be recompiling it with ifort 9.0, Siesta stopped complaining
about ghost states in N.
I would have never ever thougt of a compiler issue. My siesta is compiled with
ifort already.
Now I just have to test that the BN lattice constant makes sense :)
cheers, Serena
On Jun 7 2006, Marcos Verissimo Alves wrote:
Hi Serena,
This is quite strange, because I have used the same pseudo for some test
PtN calculations and SIESTA didn't complain about ghost states. Actually,
the results were very good for the bulk modulus and bond length,
comparable to plane wave calculations. Are the warnings on the N pseudo?
One suggestion I give you (maybe it's nonsense, but it might be worth
giving it a shot) is to check the optimization level of your siesta
compilation. If you are using ifort, which is what many people do, upgrade
to the latest version and use at most O2 optimization. I have seen once a
post in the list by a person who had a serious mistake in his calculations
and after recompiling with O2 optimization, the error disappeared and his
results made sense.
Hope this helps.
Cheers,
Marcos
---------------------------- Original Message ----------------------------
Subject: [SIESTA-L] N pseudopotential
From: "Serena Povia" <[EMAIL PROTECTED]>
Date: Tue, June 6, 2006 11:28 pm
To: SIESTA-L@LISTSERV.UAM.ES
--------------------------------------------------------------------------
Dear Siesters,
I'm working on the pseudopotential of N. I'm interested in LDA, Troullier
Martin flavour (tm2). I already tryed to generate a pseudo using the input
file present in the database and contributed by Maria Vicotria
Fernandez-Serra and changing it to tm2. The potential looks nice but
siesta complains about a ghost state in the L=0 orbital, where I can see
no kink. I played around a bit with the r_c and I came to the conclusion
that for radiuses r_c(L=0)<1.50 siesta complains about ghost states in L=0
irrespectively to the R_core/R_core_flag parameters. so one of the many
versions of the potential that I generated looks like -----------------
pe Nitrogen
tm2 2.0 # PS flavor, logder R
n=N c=car
0.0 0.0 0.0 0.0 0.0 0.0
1 4
2 0 2.00 0.00 # 2s2
2 1 3.00 0.00 # 2p1
3 2 0.00 0.00 # 3d0
4 3 0.00 0.00 # 4f0
1.50 1.10 1.90 1.90 0.00 0.50
----------------
I tryed to use the above potential in the claculation of the hexagonal BN
sheet and I got a lattice constant which is 2% larger than the acceped one
(~2.50 Ang).
does anyone have experience with N and with BN hexagonal structures (flat
sheet and/or nanotubes)? (I vaguely trust my generated B pseudo)
does anyone have any suggestion about the partial core correction? or has
anyone noticed ghost states in the L=0 component of nitrogen?
thank you for your help, Serena
--
_
_ _ (o_ Serena Povia
(o- (o- //\ (001) 617-253-2063, [EMAIL PROTECTED], [EMAIL PROTECTED]
(/)_ (\)_ V_/_ room 13-5122, 77 Massachusetts Ave, Cambridge 02139, MA,
USA
Never, Never, Never Quit.
- Winston Churchill
