[Wien] Where am I making mistake in LDA+U MAE calculation
Ahmad Asali
hajiahmadagha at gmail.com
Fri Jun 14 00:20:29 CEST 2019
Hello,
1. First of all, as Prof. Blaha says always use the lowest symmetry. This
means always run initso_lapw in ALL magnetic directions before any SCF.
2. obviously, any follow-up step should use the converged charges and
densities. so yes, use the files created before, if you are continuing the
calculation.
3. the values of (U, J) are given in RYDBERG in WIEN2k. most papers write
the values in eV. make sure you have the right values. for f-electrons of
Sm in Sm2Co17, the values for (U, J) should be roughly about (5.7, 0.7) eV.
turn this into RYDBERG and test them. change the value in 0.2 eV steps to
find best agreement. the +U calculation is not ab-initio (unless you
calculate (U,J) from Raccah parameters).
4. WIEN2k user guide suggests using U_eff = (U-J) and then J=0. my
experience suggests you use both U and J, which works better for
f-electrons.
5. it is my personal experience, that starting with non-spin polarized
calculation is better than starting with spin-polarized. formally, DFT only
guarantees an energy minimum if there is no spin-polarization. this means
that starting with runsp_lapw could find a local minimum but not the
absolute minimum. then the quantum numbers of angular momentum would be
wrong and the magnetic moments would be false. so I suggest you start with
the non-spin polarized calculation, then go to spin-polarized (SRC) and
then SOC.
a) init_lapw + initso_lapw + x lapwdm b) run_lapw -orb, c) runsp_lapw -orb,
d) runsp_lapw -so -orb.
6. you can also test the following: include -orb only after the convergence
of SOC. that means
a) run_lapw, b) runsp_lapw, c) runsp_lapw -so, d) x lapwdm, e) runsp_lapw
-so -orb
On Tue, Jun 11, 2019 at 5:23 PM Tuvshin D <tuvshin1230 at gmail.com> wrote:
> Truly appreciated
>
>
>
> [image: Mailtrack]
> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality5&> Sender
> notified by
> Mailtrack
> <https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality5&> 06/12/19,
> 12:21:56 AM
>
> On Wed, Jun 12, 2019 at 12:13 AM <tran at theochem.tuwien.ac.at> wrote:
>
>> Yes this is that. P1 means only one symmetry operation (identity) in
>> struct file. The goal is to make all calculations
>> (for the different directions) with the same struct file.
>>
>>
>> On Tuesday 2019-06-11 17:07, Tuvshin D wrote:
>>
>> >Date: Tue, 11 Jun 2019 17:07:38
>> >From: Tuvshin D <tuvshin1230 at gmail.com>
>> >Reply-To: A Mailing list for WIEN2k users <
>> wien at zeus.theochem.tuwien.ac.at>
>> >To: A Mailing list for WIEN2k users <wien at zeus.theochem.tuwien.ac.at>
>> >Subject: Re: [Wien] Where am I making mistake in LDA+U MAE calculation
>> >
>> >Thanks you sir, lowest symmetry in P1 means it's better to not group
>> >individual atoms right? For my example of Sm2Fe17, I should make struct
>> file
>> >with 19 individual atoms instead of 5 that x symm of init generates to
>> me
>> >(or 8 in certain direction)
>> >
>> >My inaccuracy comes from that my struct file changed in -so calculation.
>> >Best way to prevent is use P1 with ungrouped all atoms, is it correct?
>> >
>> >
>> >
>> >
>> >Mailtrack
>> >Sender notified by
>> >Mailtrack 06/11/19, 11:51:42 PM
>> >[5ecdf98e77e1337fcb068fb5d8e03e6b0c061794.png?u=3834765]
>> >
>> >On Tue, Jun 11, 2019 at 11:29 PM Peter Blaha <
>> pblaha at theochem.tuwien.ac.at>
>> >wrote:
>> > Definitely, MAE calculations should ALWAYS be done with ONE
>> > struct file
>> > of lowest symmetry (eventually in P1) to avoid any possible
>> > biases.
>> >
>> > Usually initso will change your struct file and reduce symmetry.
>> > Take
>> > the reduced symmetry file and put another magnetization
>> > direction.
>> > Repeat such that at the end no further symmetry change appears
>> > in any of
>> > your desired directions.
>> >
>> > With this struct file do a non-SO calculation with -orb (PS: You
>> > should
>> > NEVER use -orb right after init_lapw, but always converge
>> > first with
>> > GGA, then create new dmatup/dn files (x lapwdm -up/dn); save and
>> > then
>> > continue with -orb.
>> >
>> > Obviously, LDA+U can lead to different (meta-stable) solutions
>> > and then
>> > a comparison of total energies is not possible.
>> >
>> > On 6/9/19 6:13 AM, Tuvshin D wrote:
>> > > Dear WIEN2k users, while my normal MAE calculations are being
>> > well
>> > > achieved, LDA+U or inclusion of Orbital Potential methods
>> > giving not so
>> > > reliable results, makes me wonder if I'm doing correct or not.
>> > I'd
>> > > really appreciate if anyone with an experience on MAE
>> > calculations make
>> > > quick skim through my steps and point out where did I
>> > went wrong. System
>> > > is SmFe and calculating DM and U only on Sm atom. Full list of
>> > my given
>> > > commands are included.
>> > >
>> > > 1. I make directory, bring struct file and run (init_lapw)
>> > > 2. Set proper case.indm case.indmc and case.inorb files and
>> > run
>> > > (runsp_lapw -p -orb -ec 0.000001 -cc 0.000001 -fc 0.001 -i 500
>> > -NI)
>> > > 3. After reached convergence, (save_lapw -d name) to save
>> > results.
>> > > 4. Make 2 new directory for each magnetization directions and
>> > copy above
>> > > result to them.
>> > > 5. Run initso_lapw for setup on each of directions.
>> > > 6. Now run (runsp_lapw -p -so -orb -ec 0.000001 -cc 0.000001
>> > -fc 0.001
>> > > -i 500 -NI)
>> > > 7. Get MAE from difference between energy of 2 directions
>> > (from bottom
>> > > of case.scf) as -orb already calculated DM.
>> > >
>> > > Is there any wrong steps, if I were to run SO first then scf,
>> > what would
>> > > be its step, or should I include -orb after normal scf.
>> > >
>> > > Thanks for your kind attention, best of all.
>> > >
>> > >
>> > >
>> > >
>> > > Mailtrack
>> > ><
>> https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=si
>> > gnaturevirality5&>
>> > > Sender notified by
>> > > Mailtrack
>> > ><
>> https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=si
>> > gnaturevirality5&>
>> > > 06/09/19, 1:10:23 PM
>> > >
>> > >
>> > > _______________________________________________
>> > > Wien mailing list
>> > > Wien at zeus.theochem.tuwien.ac.at
>> > > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>> > > SEARCH the MAILING-LIST at:
>> >
>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>> > >
>> >
>> > --
>> >
>> > P.Blaha
>> >
>> --------------------------------------------------------------------------
>> > Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060
>> > Vienna
>> > Phone: +43-1-58801-165300 FAX: +43-1-58801-165982
>> > Email: blaha at theochem.tuwien.ac.at WIEN2k:
>> > http://www.wien2k.at
>> > WWW: http://www.imc.tuwien.ac.at/TC_Blaha
>> >
>> --------------------------------------------------------------------------
>> > _______________________________________________
>> > Wien mailing list
>> > Wien at zeus.theochem.tuwien.ac.at
>> > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>> > SEARCH the MAILING-LIST at:
>> >
>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>> >
>> >
>> >_______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>> SEARCH the MAILING-LIST at:
>> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:
> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20190614/2b1072d0/attachment.html>
More information about the Wien
mailing list