[Pw_forum] Magnetization in the Adsorption of Molecular O2

[Vic Bermudez]

Hello,

	I'm getting what seem to be very strange results when trying to model the
adsorption of molecular O2 on a semiconductor surface (GaSb). To begin,
everything looks reasonable for free O2, modeled as a cubic lattice of O2
spaced 20 Angstroms apart. The triplet ground state lies 1.08 eV below the
lowest singlet state, in reasonable agreement with the experimental result
of 0.98 eV.
	Now when I place a single O2 molecule near a GaSb surface the preliminary
result is that a "superoxo-like" Ga-O-O species forms during relaxation with
one O bonded to Ga and the other left dangling. With
starting_magnetization(1)=0.17 for each O the adsorption energy is 0.52 eV
with
	total magnetization       =    -1.30 Bohr mag/cell
      absolute magnetization    =     1.42 Bohr mag/cell

If instead I force a triplet by using Tot_Magnetization=2.0 I get a slightly
smaller adsorption energy of 0.49 eV (i.e., a less stable structure). If I
force a singlet (Tot_Magnetization=0.0) I get an even smaller adsorption
energy, 0.43 eV. Adsorption energy is defined here as E(free O2) + E(bare
surface) - E(O2+surface)
	I've tried letting the 2 O atoms be magnetically inequivalent, i.e., using
an O1 and an O2 with the same pseudopotential but with
	starting_magnetization(1)=0.27,
	starting_magnetization(2)=0.07
Atom #1 is the dangling O, which I would expect to have a higher density of
unpaired spin. However, this converges to essentially the same results for
total and absolute magnetization as before.
	Clearly I'm doing something wrong here, since I expect the total spin to be
conserved in adsorption on a diamagnetic surface. The bare-surface
reconstruction I'm studying has all dangling bonds either empty or doubly
occupied and hence has no magnetic properties of its own. This doesn't seem
to be a typical case of spin contamination. From what I've read in the
literature, spin contamination in DFT is usually smaller than in UHF and
also involves the admixing of higher spin states. Hence, a spin-contaminated
triplet would have a total magnetization of >2.0 and not the 1.3 that I'm
getting.
	If it's of use I'm including the input for the run with two different O
spins as a sample. A few of comments may be in order. To save space in this
message I've removed most of the atom positions. The two-dimensional slab
has a dipole moment so a correction dipole is included in the vacuum space.
"nstep" and "electron_maxstep" are set absurdly high to avoid a problem I've
been having with "maximum number of steps exceeded" after only 60 or so
optimizations steps (a different problem, not the subject of this message).
The H5 pseudoatoms are for terminating the unsaturated Ga atoms at the
"bottom" of the slab.

&CONTROL
calculation='relax',
restart_mode='from_scratch',
title='GaSb alpha-4x3 + O2; start with #151846 opt. pos.; make O2
asymmetric',
pseudo_dir="/work1/app/espresso/pseudo",
outdir='/work2/bermudvm/GaSb+O2__outdir',
etot_conv_thr=1.0D-5,
forc_conv_thr=5.0D-4,
nstep=10000,
wf_collect=.TRUE.,
verbosity='default',
tefield=.TRUE.,
dipfield=.TRUE.
/

&SYSTEM
ibrav=8,
a=17.1119841047, b=12.8339880785, c=44.0,
nat=130,
ntyp=5,
ecutwfc=30.0,
ecutrho=300.0,
london=.TRUE.,
occupations='smearing',
smearing='gaussian',
degauss=0.005D0,
nspin=2,
starting_magnetization(3)=0.27,
starting_magnetization(4)=0.07,
edir=3,
eamp=0.0,
emaxpos=0.5,
eopreg=0.1,
input_dft='pw91'
/

&ELECTRONS
electron_maxstep=3000,
conv_thr=1.0D-7,
mixing_mode='local-TF'
/

&IONS
ion_dynamics='bfgs',
upscale=10000.D0
/

ATOMIC_SPECIES
Ga   69.72   Ga.pw91-nsp-van.UPF
Sb  121.75   Sb.pw91-n-van.UPF
O1    15.9994 O.pw91-van_ak.UPF
O2    15.9994 O.pw91-van_ak.UPF
H5    1.25   H.pz-vbc_125.UPF

ATOMIC_POSITIONS angstrom
O1      -1.057077087  -0.134186302   8.773421904
O2      -2.125029769   0.459941835   8.302768681
Sb       2.139040375   4.993070914   6.782713316
Ga       2.132538190  -5.199966759   5.947206011
Ga       6.277129268   1.147167971   5.989478016
.
.
.
.
H5      -7.635543780  -4.277997026  -6.285642281    0   0   0
H5      -3.357548712  -0.000000728  -6.285641379    0   0   0
H5      -3.357548071   4.277997544  -6.285641893    0   0   0
H5      -3.357547485  -4.277997519  -6.285642688    0   0   0

K_POINTS automatic
3 4 1 0 0 0

	I would greatly appreciate any advice regarding this situation.

Thank You,
Vic Bermudez

Victor M. Bermudez
Code 6876
U.S. Naval Research Laboratory
4555 Overlook Ave., S.W.
Washington, DC 20375-5347

Phone: 202-767-6728
FAX: 202-767-1165
E-mail: victor.bermudez at nrl.navy.mil