as an addition to my previous reply..
If you decide to diagonalize the restricted nat_todo matrix.. DO NOT
impose ASR on it.
It does not apply to the truncated matrix..
stefano
On 17/03/20 16:08, Stefano de Gironcoli wrote:
On 17/03/20 15:46, Paolo Giannozzi wrote:
On Mon, Mar 16, 2020 at 1:53 AM Shen, Ziheng <[email protected]
<mailto:[email protected]>> wrote:
1) May I ask is there any literature that can prove it’s feasible
to neglect the small frequencies?
no idea, but you can easily verify whether the neglect of almost-zero
frequencies has any effect: if you impose the ASR on dynamical
matrices, only frequencies of translational modes (and of rotational
ones for a molecule) should change, while all other frequencies
should stay almost the same.
2) I saw you also replied to others that “nat_todo” does not make
any sense, but there were still people claimed that they got some
useful results. I read from some books (like the one written by
D.Sholl) saying that it’s possible to calculate only frequencies
of adsorbates which could save computational resources. Some
other ab initial calculation software (like VASP) also have
similar functions to calculate adsorbates only. Why is “nat_todo”
not working here?
Presumably it doesn't work because it is not sufficient to just
ignore all surface atoms and compute the dynamical matrix restricted
to adsorbate atoms only. I guess there are tricks to approximated
phonons for an adsorbate on a surface by computing a "reduced"
dynamical matrix for the adsorbate only, in which the information on
the adsorbate-surface interactions is "refolded", but I don't know
any. This same question has been asked over and over again in the
list and nobody (that I remember) ever gave an answer.
Paolo
I think that if you have light atoms adsorbed on some heavier material
and are looking for the adsorbate related frequencies then you can
diagonalize the reduced matrix and the frequencies that you obtain
should be variational upper bounds to the real ones....
just pretend you computed the whole matrix but restrict the atomic
motion to a subset of atoms... so to the extent the modes are
localized on the adsorbate with little involvement of the substrate
you should be ok...
but the low lying frequencies (the ones resonant with the substrate
modes) will be completely wrong.
one could start with nat_todo equal to just the adsorbate and then the
first shell, then the second one ... and see how things converge ...
another option (better, but would need some ad hoc coding) would be to
build the dynmat of the nat_todo atoms with interatomic force
constants for the rest of the cell borrowed from some model (the bulk,
or the clean surface) ... and again monitor the convergence as you
include more and more shells of atoms in the nat_todo.
stefano
Best regards
Ziheng Shen
PhD student @ Georgia Institute of Technology
On Mar 15, 2020, at 7:00 AM,
[email protected]
<mailto:[email protected]> wrote:
On Fri, Mar 13, 2020 at 4:22 AM Shen, Ziheng <[email protected]
<mailto:[email protected]>> wrote:
1) When doing frequency analysis for molecules, I expected to get
zero or
extremely small value for the first six frequencies (i.e.
translational &
rotational modes). According to suggestions from those previously
posted
problems, I tried to apply more restrict convergence thresholds and
ASR. It
seems that ASR help a lot to reduce the number. But I still got
frequencies
at ~50 level. Is it possible to completely remove those small
values? Or
are those values small enough to be neglected?
They are small enough to be neglected. They can be removed by
applying the
ASR to the computed dynamical matrix. See the various kinds of ASR
in codes
"dynmat" and "matdyn", in particular the "zero-dim" one. Note that more
sophisticated ASR than "simple" can be surprising slow.
2) My ultimate goal is to perform frequency analysis for adsorbate
so that
I can both determine transition state structures and apply ZPE
corrections.
I tried to use ?nat_todo? to fix the surface atoms and only did
calculation
for adsorbate (CH in my case). I got crazy result (~10000 cm-1)
when using
large tr2_ph, and got improved results when I decrease the
threshold. But I
still got fairly large translational & rotational frequencies like
below
I don't think you will obtain anything sensible by fixing the
surface atoms
and making the calculation for the adsorbate atoms only
Paolo
freq ( 1) = -25.618746 [THz] = -854.549399 [cm-1]
freq ( 2) = -7.333895 [THz] = -244.632409 [cm-1]
freq ( 3) = -6.696884 [THz] = -223.383991 [cm-1]
freq ( 4) = -6.248674 [THz] = -208.433322 [cm-1]
freq ( 5) = -4.947831 [THz] = -165.041892 [cm-1]
freq ( 6) = -2.014699 [THz] = -67.203109 [cm-1]
freq ( 37) = 0.571458 [THz] = 19.061786 [cm-1]
freq ( 38) = 5.754719 [THz] = 191.956759 [cm-1]
freq ( 39) = 16.488930 [THz] = 550.011494 [cm-1]
freq ( 40) = 16.563150 [THz] = 552.487199 [cm-1]
freq ( 41) = 18.255969 [THz] = 608.953585 [cm-1]
freq ( 42) = 56.121326 [THz] = 1872.005923 [cm-1]
What does negative translational frequencies indicate, is it
possible to
eliminate these imaginary numbers (like using more restrict threshold)?
And does my result indicate that my structure is most probably not a
transition state since all the other frequencies are positive?
I?m attaching the input file of pw.x &ph.x below:
=========================scf input, structure obtained from
neb.x========================
&CONTROL
Calculation='scf',
restart_mode='from_scratch',
prefix = "Ni_ch_ts"
outdir = "./ts/tmp",
pseudo_dir = "./pseudo",
tstress = .true.
verbosity = 'high'
tefield = .true.
dipfield = .true.
/
&SYSTEM
ibrav = 0,
nat = 14,
ntyp = 3,
ecutwfc = 65,
ecutrho = 650,
Occupations='smearing',
smearing='mp',
degauss=0.01,
nspin=2,
starting_magnetization(1)=0.2,
eamp = 0.0
edir = 3
emaxpos = 0.95
eopreg = 0.05
/
&ELECTRONS
electron_maxstep=250,
conv_thr = 1.D-10,
mixing_beta = 0.1,
/
ATOMIC_SPECIES
Ni 58.69 ni_pbe_v1.4.uspp.F.UPF
C 12 C.pbe-n-kjpaw_psl.1.0.0.UPF
H 1 H.pbe-kjpaw_psl.1.0.0.UPF
CELL_PARAMETERS { angstrom }
4.9667177200 0.0000000000 0.0000000000
2.4833588600 4.3013037190 0.0000000000
0.0000000000 0.0000000000 20.000000000
ATOMIC_POSITIONS { angstrom }
Ni 0.0000000000 0.0000000000 7.9723500000
Ni 1.2416800000 2.1506500000 7.9723500000
Ni 2.4833600000 0.0000000000 7.9723500000
Ni 3.7250400000 2.1506500000 7.9723500000
Ni 2.4833600000 1.4337700000 10.0000000000
Ni 3.7250400000 3.5844200000 10.0000000000
Ni 4.9667200000 1.4337700000 10.0000000000
Ni 6.2084000000 3.5844200000 10.0000000000
Ni 1.2281125212 0.7413038538 12.1124602290
Ni 2.4833613443 2.9495126082 12.0722664849
Ni 3.7386101535 0.7413040204 12.1124604793
Ni 4.9667205066 2.8946406480 11.9560276983
C 2.4833610952 1.4972020489 13.1166864267
H 2.4833559794 3.1940064219 13.5465576823
K_POINTS { automatic }
6 6 1 0 0 0
=======================================ph.x
input=======================================
phonons of CH on metal Ni at Gamma
&inputph
tr2_ph=1.0d-16,
prefix='Ni_ch_ts',
epsil=.false.,
amass(1)=58.69,
amass(2)=12.011,
amass(3)=1.0,
alpha_mix(1)=0.1,
outdir='./tmp/',
fildyn='CH.dynG',
nat_todo= 2,
/
0.0 0.0 0.0
13 14
Thanks in advance for anyone that could give suggestions to me!
Best regards
Ziheng Shen
PhD student @ Georgia Institute of Technology
_______________________________________________
Quantum ESPRESSO is supported by MaX
(www.max-centre.eu/quantum-espresso
<http://www.max-centre.eu/quantum-espresso>)
users mailing list [email protected]
<mailto:[email protected]>
https://lists.quantum-espresso.org/mailman/listinfo/users
--
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso)
users mailing [email protected]
https://lists.quantum-espresso.org/mailman/listinfo/users
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso)
users mailing list [email protected]
https://lists.quantum-espresso.org/mailman/listinfo/users
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso)
users mailing list [email protected]
https://lists.quantum-espresso.org/mailman/listinfo/users
