Bulk Nickel (and possibly other newbie questions)

[Alex]

Hi all,

I am very new to DFT calculations, let alone CP2k, so some level of idiocy on my part should be expected.

As a simple test, I am trying to calculate the binding energy of a Ni atom in a bulk crystal. The relevant portion of the input shown below:

***

&GLOBAL

  PROJECT Ni_inp_test

  RUN_TYPE ENERGY_FORCE

  PRINT_LEVEL LOW

&END GLOBAL

&FORCE_EVAL

  METHOD Quickstep

  &SUBSYS

    &KIND Ni

      ELEMENT Ni

      BASIS_SET DZV-GTH-PADE-q18

      POTENTIAL GTH-PADE-q10

    &END KIND

    &CELL

      A     1.765000    1.765000    0.000000

      B     0.000000    1.765000    1.765000

      C    1.765000    0.000000    1.765000

      PERIODIC XYZ

    &END CELL

    &COORD

      Ni    0.000000000    0.000000000    0.000000000

    &END COORD

  &END SUBSYS

  &DFT

    BASIS_SET_FILE_NAME  BASIS_SET

    POTENTIAL_FILE_NAME  GTH_POTENTIALS

    &QS

      EPS_DEFAULT 1.0E-10

    &END QS

    &MGRID

      NGRIDS 4

      CUTOFF 300

      REL_CUTOFF 60

    &END MGRID

    &XC

      &XC_FUNCTIONAL PADE

      &END XC_FUNCTIONAL

    &END XC

    &SCF

      SCF_GUESS ATOMIC

      EPS_SCF 1.0E-7

      MAX_SCF 300

      ADDED_MOS 10

      &DIAGONALIZATION  ON

        ALGORITHM STANDARD

      &END DIAGONALIZATION

      &MIXING  T

        METHOD BROYDEN_MIXING

        ALPHA 0.4

        NBROYDEN 8

      &END MIXING

      &SMEAR ON

        METHOD FERMI_DIRAC

        ELECTRONIC_TEMPERATURE [K] 300

      &END SMEAR

    &END SCF

  &END DFT

  &PRINT

    &FORCES ON

    &END FORCES

  &END PRINT

&END FORCE_EVAL

***

This yields a total energy of E1=-35.155 a.u. after convergence.

Then I decided to calculate the "vacuum" energy of an isolated atom, input below:

&GLOBAL

  PROJECT Ni_inp_test

  RUN_TYPE ENERGY_FORCE

  PRINT_LEVEL LOW

&END GLOBAL

&FORCE_EVAL

  METHOD Quickstep

  &SUBSYS

    &KIND Ni

      ELEMENT Ni

      BASIS_SET DZV-GTH-PADE-q18

      POTENTIAL GTH-PADE-q10

    &END KIND

    &CELL

      A     30.00000    0.000000    0.000000

      B     0.000000    30.00000    0.000000

      C    0.000000    0.000000    30.00000

    &END CELL

    &COORD

      Ni    0.000000000    0.000000000    0.000000000

    &END COORD

  &END SUBSYS

  &DFT

    BASIS_SET_FILE_NAME  BASIS_SET

    POTENTIAL_FILE_NAME  GTH_POTENTIALS

    &QS

      EPS_DEFAULT 1.0E-10

    &END QS

    &MGRID

      NGRIDS 4

      CUTOFF 300

      REL_CUTOFF 60

    &END MGRID

    &XC

      &XC_FUNCTIONAL PADE

      &END XC_FUNCTIONAL

    &END XC

    &SCF

      SCF_GUESS ATOMIC

      EPS_SCF 1.0E-7

      MAX_SCF 300

      ADDED_MOS 10

      &DIAGONALIZATION  ON

        ALGORITHM STANDARD

      &END DIAGONALIZATION

      &MIXING  T

        METHOD BROYDEN_MIXING

        ALPHA 0.4

        NBROYDEN 8

      &END MIXING

      &SMEAR ON

        METHOD FERMI_DIRAC

        ELECTRONIC_TEMPERATURE [K] 300

      &END SMEAR

    &END SCF

  &END DFT

  &PRINT

    &FORCES ON

    &END FORCES

  &END PRINT

&END FORCE_EVAL

***

This also converges and yields a total energy E2=-34.555 a.u.

Hence, my questions:

1. Is this even the correct way of calculating what I want, including the energy calculations, XC functional, and basis?

2. Should the spin properties be explicitly set in the input? There are none now.

3. Am I setting up the FCC lattice correctly (first input file)? My translation vectors are set by the ABC values, but I have no idea whether this is right.

4. If the first simulation yields the total energy of the system and the FCC lattice implies 12 nearest neighbors, then removing the center would change the total energy by (E1-E2)/6, which isn't the experimental -4.4 eV. Am I completely off track here? :)

Thanks a lot!

[Marcella Iannuzzi]

Hi

Just few remarks.

For the fcc bulk Ni energy, you need to construct a supercell, since there is no k-point sampling,

and check the convergence of the result with system size.

PADE is probably not an optimal choice for the XC functional, anyway, you should use consistent potential and basis set,

i.e. for the same number of valence electrons (in your input, the PP is for 10 v.e. and the BS for 18)

If you don't specify in in the input, no spin polarisation is considered.

Regards

Marcella

[Alex]

HI Marcella,

Thank you. Potential and basis set aside (I can fix that easily), I was under the impression that the ABC settings were indeed defining the supercell, which then, combined with PERIODIC XYZ would give me the desired setup overall. What are you suggesting instead?

Thanks,

Alex

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/llUH48_Y3vY/unsubscribe.
To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at http://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

[Alex]

Also, I just redid the calculation with 

    BASIS_SET DZV-GTH-PADE-q18

   POTENTIAL GTH-PADE-q18

Same energy within I believe five decimal places. As far as the correct choice of the XC-potential-basis combination, what would be better in my case? As I've said before, I am really new to this... Any reference to that effect would be great.

Thank you.

[Ralph Koitz]

Hi,

To use spin polarization, you should specify the keyword UKS in the DFT section and define a MULTIPLICITY.

You may also want to turn on RELAX_MULTIPLICITY in that section.

To use a supercell, you can either manually repeat your coordinates and load the entire supercell as input coordinates, or specify MULTIPLE_UNIT_CELL in both the CELL and TOPOLOGY sections of the SUBSYS.

A 6x6x6 supercell should be okay to start out with, but you may want to check convergence.

We have had decent experience with revPBE as an XC Functional for Ni [1]. BEEFvdW, a more recent addition to cp2k also gives very good bulk nickel, but it's currently somewhat of a beta-version and you need to compile with libxc to get it to work.

Best, 

Ralph

[1] Gomez-Diaz et al, Theor Chim Acta, 2013

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.

[Alex]

Hi Ralph,

This is all very useful, thanks a bunch. I added cell multiplicity explicitly, as you suggested. 

Since I am not the owner of the installation (I'm running it on a remote cluster), I won't be able to compile anything myself, so I'll stick with what's there now (I think we have Debian binaries). I am looking at the reference you provided... For PBE/revPBE XC, what potential and basis did you use, exactly (assuming they are part of the package)? Also, did you use s=1/2 for the multiplicity in this case, i.e. the MULTIPLICITY param would be 2?..

Thank you,

Alex

[Ralph Koitz]

I used the following 

      BASIS_SET  DZVP-MOLOPT-SR-GTH

      POTENTIAL  GTH-PBE-q18

and set the multiplicity to n(atom)+1, basically s=1/2 for each atom in the supercell.

[Alex]

Okay, this is a bit weird. Just for testing, I decided to get the total self-energy for Ni, just one atom in vacuum following Ralph's suggestions.

***

&KIND Ni

      ELEMENT   Ni

      BASIS_SET DZVP-MOLOPT-SR-GTH

      POTENTIAL GTH-PBE-q18

    &END KIND

    &CELL

      A     10.00000    0.000000    0.000000

      B     0.000000    10.00000    0.000000

      C     0.000000    0.000000    10.00000

    &END CELL

    &COORD

      Ni    0.000000000    0.000000000    0.000000000

    &END COORD

  &END SUBSYS

  &DFT

    BASIS_SET_FILE_NAME  BASIS_SET2

    POTENTIAL_FILE_NAME  GTH_POTENTIALS2

    UKS                  TRUE

    MULTIPLICITY         2

***

The message I get with above is "try a different multiplicity." 

The manual states:

"MULTIPLICITY {Integer}

Two times the total spin plus one. Specify 3 for a triplet, 4 for a quartet, and so on. Default is 1 (singlet) for an even number and 2 (doublet) for an odd number of electrons.  [Edit]

This optional keyword cannot be repeated and it expects precisely one integer."

Here, we have 28 electrons, but setting multiplicity to 1 is not even an option, the simulation crashes. Any suggestions?

[SRKC Sharma Yamijala]

Hi,

It seems you have given MULTIPLICITY = 2, in your input. Try with '1'.

********************************************************
Chaitanya Sharma,
Prof. Pati's group,
Chemistry and Physics Materials unit,
JNCASR, BANGLORE,
Lab:: (080-2208) 2581, 2809
https://sites.google.com/site/sharmasrkcyamijala/
*********************************************************

[Alex]

Thanks for the reply, but please read my entire post above. 

[Ralph Koitz]

Hi,

I tried to calculate a single Ni atom (PBE, dzvp-molopt, 500 Ry cutoff). It works for both MULTIPLICITY 1 and MULTIPLICITY 3, although the latter doesn't seem to converge with the parameters I'm using.

You mentioned that your calculation crashes... Do you get any kind of error message? How far does it go?

Best,

rk

[Alex]

Today, I tried with '1' and it converged, but the total force is nonzero at the end. My cutoff was at 300 Ry, so I am trying with 500 now. So far, it's been about an hour on two xeon cores and not a single energy update. The tail -f .out file just reads

...

 SCF WAVEFUNCTION OPTIMIZATION

  Step     Update method      Time    Convergence         Total energy    Change

  ------------------------------------------------------------------------------