[QE-users] higher total energy in the last stage of vc-relax

[Kevin May]

Hi Mehrdad,

I forgot that you could just set ibrav = 3 instead of using the
conventional BCC unit cell. That way you can just use one atom in the
calculation. Also, don't forget to set a non-zero starting magnetization
and change ecutrho to be 1080. If you make these 3 adjustments to your
previous input file for the bulk BCC Fe it should work.

I did a quick calculation of bulk BCC Fe on my laptop and got 329.27274577
Ry. Using your supercell energies I got a vacancy formation energy of 1.69
eV. Pretty close to the 1.66 eV  reported by Jiang et al. that your
reference uses. They used PW91-GGA, different PAW datasets and different
software so I think that's pretty good. Double-check for yourself though,
of course.

Best,

Kevin May, PhD
Postdoctoral Associate
Department of Materials Science and Engineering
Massachusetts Institute of Technology

On Tue, Jun 18, 2019 at 10:30 PM Kevin May <kmay at mit.edu> wrote:

> Hi Mehrdad,
>
> So you are indeed using an isolated Fe atom as your reference state. That
> is not the conventional chemical potential for elemental iron. To have a
> fair comparison to the calculated value the literature, please try a
> vc-relax calculation of ferromagnetic BCC iron (like this CIF file:
> https://www.crystallography.net/cod/9008536.cif). Divide the energy by 2
> (since there are two atoms per unit cell) to get the chemical potential
> reference for iron. See my comment from the last email:
>
> By* E(single Fe) you mean 1/2 the energy of ferromagnetic BCC Fe (2 Fe
>> per unit cell), rather than an isolated Fe atom, right? That is the
>> chemical potential reference for Fe that was used in calculating the 1.6 eV
>> formation energy value.* And for what it's worth I don't think point
>> defect calculations are that trivial.
>>
>>> 3) thanks very much for the reference Rev. Mod. Phys. 86, 253 (2014).
>>>
>>
>> Regarding chemical potential reference, see *section II.B.2 of this
>> article*.
>>
>
> Best,
> Kevin May, PhD
> Postdoctoral Associate
> Department of Materials Science and Engineering
> Massachusetts Institute of Technology
>
>
>
>
>
> On Tue, Jun 18, 2019 at 11:54 AM mehrdad zamzamian <
> mehrdad.zamzamian at gmail.com> wrote:
>
>> Dear Kevin
>> Hi again
>> Sorry i bothered you many times for asking many questions, but i am
>> really confused about my case
>> After using the paw-pp that you suggested, i obtained E(f-vacancy)~2.0
>> eV!! here in attachment, i gave my input. i used a 128-atom supercell for a
>> perfect crystal (E=32199.76950786 Ry) and by removing a Fe atom, i had
>> E=31870.37251766 Ry and isolated Fe E=329.18792433 Ry. It seems that i
>> cannot obtain a better result than 2 eV. I appreciated if you give me any
>> suggestions. the paper that you mentioned talked about "Ecorr". what
>> exactly is this parameter?
>>
>> On Sat, Jun 1, 2019 at 11:27 PM Kevin May <kmay at mit.edu> wrote:
>>
>>> Hi Mehrdad,
>>>
>>> On Sat, Jun 1, 2019 at 1:58 PM mehrdad zamzamian <
>>> mehrdad.zamzamian at gmail.com> wrote:
>>>
>>>> With regard
>>>> Dear Kevin
>>>> 1) Actually, my reference is computational material science 44(2008)
>>>> 690-694 (although they reported delta(E)= -0.456 eV/atom that i don't know
>>>> what it is. because i said that it must be 1.6 eV according to
>>>> https://doi.org/10.1080/09506608.2018.1560984). i also used
>>>> spin-polarized (0.2 for Fe) but the same result was obtained (not better
>>>> than 2.2 eV). I also used constant volume (by using relax calculation not
>>>> vc_relax), but i had the same results.
>>>>
>>>
>>> The first paper you mention does not calculate vacancy formation energy,
>>> the -0.456 eV/atom is the bulk formation energy of orthorhombic Fe3C from
>>> GGA. The second paper is a review article that cites the paper by Jiang et
>>> al. that I mentioned in my last email for their ~1.6 eV formation energy of
>>> a Fe vacancy.
>>>
>>>> 2) thanks for the proposed links, but in that reference they offered
>>>> two non-consistent pps (C.pbe-n-kjpaw_psl.1.0.0.UPF
>>>> and Fe.pbe-spn-kjpaw_psl.0.2.1.UPF) that i cannot use for Fe3C. i used
>>>> these pp:
>>>> C.pbe-n-rrkjus_psl.1.0.0.UPF
>>>> Fe.pbe-spn-rrkjus_psl.1.0.0.UPF
>>>>
>>>
>>> I'm not sure what you mean by "non-consistent", or why you can't use
>>> those PAW datasets for your calculation. If you need to use ultrasoft for
>>> another reason, I'd still recommend using Fe.pbe-spn-rrkjus_psl.0.2.1.UPF
>>> instead of the psl.1.0.0 version.
>>>
>>> I am really confused about how can i calculate this rather simple
>>>> parameter. i should mention that i calculate the Fe vacancy with:
>>>> E(Fe-vacancy)=E(perfect Fe3C)- E(Fe3C with lack of one Fe atom)-
>>>> E(single Fe)
>>>>
>>>
>>> By E(single Fe) you mean 1/2 the energy of ferromagnetic BCC Fe (2 Fe
>>> per unit cell), rather than an isolated Fe atom, right? That is the
>>> chemical potential reference for Fe that was used in calculating the 1.6 eV
>>> formation energy value. And for what it's worth I don't think point defect
>>> calculations are that trivial.
>>>
>>>> 3) thanks very much for the reference Rev. Mod. Phys. 86, 253 (2014).
>>>>
>>>
>>> Regarding chemical potential reference, see section II.B.2 of this
>>> article.
>>>
>>>
>>>> regards
>>>>
>>>> Seyed Mehrdad Zamzamian
>>>> Sharif University of Technology, Tehran, Iran
>>>> Energy engineering department
>>>> E-mail: mehrdad.zamzamian at gmail.com
>>>>
>>> Good luck,
>>>
>>> Kevin May, PhD
>>> Postdoctoral Associate
>>> Department of Materials Science and Engineering
>>> Massachusetts Institute of Technology
>>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20190619/d3db3976/attachment.html>