[Pw_forum] Hydrogen dissociation test

[Axel Kohlmeyer]

On Mon, 16 Feb 2009, Green Power wrote:

GP> Hi, All,
GP> I wonder how to do broken symmetry calculations for a hydrogen molecule with
GP> a separation larger than 3.0 a.u. using pw.x code. As I understand,  the i
GP> the region larger then 3.0 a.u. separation, the restricted wavefountion
GP> method, for instance the RHF, would fail, and UHF method or correlation

please do not confuse pw.x with a "normal" quantum chemistry program
(like gamess, gaussian etc.). if you want to scan the breaking of an
H-H bond, you have to use a multi-determinant method. pw.x (and 
cp.x) use(s) DFT exclusively.

GP> method is needed. I don't know how to break the symmetry of alpha spin and
GP> beta spin using pw.x code. I tried to do that by spin-polarized calculations

you have to assign two hydrogen atom types. H1, H2 and then initialize
their magnetization in opposing directions. 

GP> using the following input. The energy is 0.05 a.u.lower than the
GP> sipn-unpolarized method at he hydrogen separation of 2.4333672 angstom. But
GP> at the sepration of 2.0100254, the spin-polarized method gave much energy
GP> then sipn-unpolarized. Could you please check my input file? Mabye I have
GP> smething wrong.
GP> 
GP> Another question is about the binding energy of hydrgen molecule. I
GP> calculated the total engergy of H2 at 0.74 angstrom bond distanced and the
GP> energy H atom, and comapred the energy difference, I got a binding energy
GP> much higher than the experimental value of 436 kj/mol. I guess I did not

what experiment? have you computed the minimum H-H distance first?

GP> calculate the energy of Hydrogen atom properly. Could you give me any
GP> suggestion how to do that?

doing this kind of calculation is not as simple as with a code using
an atom centered basis set. you are always computing a supercell (and
its vacuum) and you have to make sure that this cell is large enough, 
so that periodic interactions don't matter, as well as you have to make 
sure that atoms stay at their relative positions, since with plane waves 
and a small(ish) cutoff, the total energy depends to some degree on the 
position of an atom in space (so-called ripples). finally, you should 
factor in that you are using an LDA pseudopotential, and LDA has a
tendency to overbind. not sure whether that explains your observations,
since you don't say how much your result deviates.

in general, i suggest you have an extensive look at the mailing list
archives which contain many answers and explanations to your kind of
beginner questions.

cheers,
   axel.

GP> Thank you.
GP> 
GP> Qian
GP> 
GP> 
GP> *INPUT: Spin-polarized calculation*
GP> &control
GP> /
GP> &system
GP>     ibrav              = 1
GP>     celldm(1)       = 30.0
GP>     nat                = 2
GP>     ntyp              = 1
GP>     nosym=.true.,
GP>     starting_magnetization(1)=1.0
GP>     nspin =2
GP>     ecutwfc         = 60.0
GP>     occupations = 'smearing'
GP>     degauss = 0.01
GP>     smearing = 'gaussian'
GP> /
GP> &electrons
GP>     conv_thr        = 1.0d-8
GP> /
GP> ATOMIC_SPECIES
GP>     H   1.00    H.vbc.UPF
GP> ATOMIC_POSITIONS {angstrom}
GP>    H   0.0             0.0    0.0
GP>    H   2.4333672   0.0   0.0
GP> K_POINTS  AUTOMATIC
GP> 1   1   1   0   0   0
GP> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.