SPIN polarized calculation

Message

baki · #1 Post by **baki** » Wed Jul 05, 2006 12:00 pm

Dear SIR
Recently I performed some calculation on methanol dimer and dissociated dimer on Rh(111) surface. If I did the calculation on dissociated dimer with ISPIN=1, than I got the following results
Free energy of the ion-electron system (eV)
---------------------------------------------------
alpha Z PSCENC = 1684.28533988
Ewald energy TEWEN = 82004.20797398
-1/2 Hartree DENC = -114958.56869277
-V(xc)+E(xc) XCENC = -730.72877171
PAW double counting = 26261.76492910 -24341.88700501
entropy T*S EENTRO = -0.02470724
eigenvalues EBANDS = -1161.67607508
atomic energy EATOM = 30852.76794302
---------------------------------------------------
free energy TOTEN = -389.85906582 eV

energy without entropy = -389.83435858 energy(sigma->0) = -389.85083008

If I did the calculation with ISPIN=2, I got totally different result
alpha Z PSCENC = 1684.28533988
Ewald energy TEWEN = 82004.20797398
-1/2 Hartree DENC = -114897.54109769
-V(xc)+E(xc) XCENC = 66.85307337
PAW double counting = 26266.17142720 -24346.74991612
entropy T*S EENTRO = -0.02289073
eigenvalues EBANDS = -1222.00939195
atomic energy EATOM = 30852.76794302
---------------------------------------------------
free energy TOTEN = 407.96246097 eV

energy without entropy = 407.98535169 energy(sigma->0) = 407.97009121
The main differnce between the two calculation is in the
"-V(xc)+E(xc) XCENC " data.
I think here is something wrong. Can you give me some advice to improve this calculation
best
imre

Veronika · #2 Post by **Veronika** » Thu Jul 06, 2006 5:44 am

Your second calculation is wrong anyway, you have a positive total energy!

#3 Post by **admin** » Tue Jul 11, 2006 8:38 am

please reduce the compilation optimization of xcspin.F to the lowest possible level, (maybe your compiler has over-optimized this part of code) and see if the result of the spin-polarized run improves.
As a test, I strongly recommend to do a run for a small system which is known to be magnetic.

baki · #4 Post by **baki** » Wed Jul 12, 2006 10:52 am

Dear Admin
At teh end we solved the problem. On Itanium cluster with O2 optimalisation some calculation give wrong results. After I changed to -O0 , it works fine
best
baki