Convergence problem in rare-earth system

FLEUR Forum � User support � Magnetism, Non-collinear calculations, SOC � Convergence problem in rare-earth system

Topic settings

Print view

Jump to forum

Announcements New releases, new features, important bugs User support Installation Inpgen -- the FLEUR input generator Problems while running SCF or other basic tasks DOS, Bandstructures, Relaxations Magnetism, Non-collinear calculations, SOC Wannier functions Other problems FUTURE related problems General discussion Feature requests Physics with FLEUR Open discourse

#1 | Convergence problem in rare-earth system datum

Sat Sep 04, 2021 10:17 am

Dear,
I cannot obtain a convergence result in a magnetic rare-earth system, PrAlGe with SOC and LDAU using FLEUR version of v26. So, What parameters can I modify to get a convergent result ?

Firstly, using the inpgen.x, I can generate the inp and other input documents; Secondly, run fleur.x to generate the charge density; Thirdly, add "&ldaU l=3 u=6.0 j=0.0 l_amf=F /" for Pr atom and modify the itmax=1 in the inp document, and then generate the n_mmp_mat. Fourthly, delete the broyd* and the last line in n_mmp_mat, modify the itmax=51 and run the fleur.x. The distance of charge densities can not convergence.

Thank you very much.

In the following, I post the inp_system document, and attachment is the inp document.

######################################
PrAlGe ! latest template

&input film=f cartesian=f cal_symm=f /

-0.291202873 0.291202873 1.000000000 ! a1
0.291202873 -0.291202873 1.000000000 ! a2
0.291202873 0.291202873 -1.000000000 ! a3 and dvac
13.970477352011365855224 ! lattice constant in bohr. 1 bohr= 0.529177249 ai
1.0 1.0 1.0 ! scale

6
13 0.8326614023758784 0.8326614023758784 0.0000000000000000
13 0.5826614023758784 0.0826614023758784 0.5000000000000000
32 0.9991821765633356 0.9991821765633356 0.0000000000000000
32 0.7491821765633356 0.2491821765633357 0.5000000000000000
59 0.4166564210607859 0.4166564210607859 -0.0000000000000000
59 0.1666564210607859 0.6666564210607858 0.5000000000000000

&soc 0.0 0.0 /

&atom element="Al" id=13 bmu=0.0 /
&atom element="Ge" id=32 bmu=0.0 /
&atom element="Pr" id=59 lo="5s 5p" bmu=4.0 /

&kpt div1=12 div2=12 div3=12 /

&end /

###############################################

xiuxian has attached files to this post

inp.txt

reply

#2 | RE: Convergence problem in rare-earth system datum

Wed Sep 08, 2021 9:22 pm

Dear xiuxian,

Could you give us an impression on how bad the convergence looks like? As mentioned in another discussion I don't use
the old Fleur version. For the new Fleur version I modified the input slightly (and also switched off LDA+U) and see a slowly
converging system, but there seems to be progress. After about 50 iterations I am at a charge density distance of about 0.1.

Does it look much worse for you? What I do and what is probably not done in the old Fleur version is that I delete the mixing
history every 15 iterations (The broyd files in the old Fleur version). That may make a difference.

I also saw that your usage of LDA+U is uncommon in the sense that the J value is 0.0. There are formulations of LDA+U that
abstract from a J and work with an effective U that is slightly different, but typically J is about 0.8 to 1.0. Also please be aware
that the LDA+U approach can stabilize different states, depending on the initial density matrix and how you deal with the
density matrix. It may give hints to inspect which states are occupied. Maybe you get a problematic configuration and have
to guide the LDA+U calculation into another configuration.

reply

#3 | RE: Convergence problem in rare-earth system datum

Thu Sep 16, 2021 9:56 am (Last edited: Thu Sep 16, 2021 9:56 am)

Dear Gregor,

Thank you very much for your valuable comments. In my following calculations, I made two changes. The first change is that I used U=6.8 and J=0.8 in the new calculations; second change is that deleting the mixing history every 15 iterations. After about 150 iterations, I am at a charge densities and spin densities distance of about 0.000001 me/bohr**3. That is a very good results.

Thank you !!!

reply

DMI force theorem � � Missing time-reversal symmetry in soc calculation