YMnO3 local magnetic moment with no spin-splitting in the local dos ?

[mfechner]

Dear elk community,

I recently performed calculations for the hexganoal manganite YMnO3 using elk within its non-colinear magnetic ground state (see elk.in below). My calculation converged nicely and I obtained a spin magnetic moment of 3.2 mub on Mn which corresponds to the expected oxidation state of Mn3+. I then analyzed the density of states and local dos and found the following picture see fig1 with no spin splitting of the Mn-d states. Hence from this data I would expect a magnetic moment of 0 ? So does somebody has a suggestion for this case ? Every comment is welcome. So far I could not ruled out that there is a problem in the code as well

best regards

Michael

PS there is no problem in summing the dos hence I attach the files as well.

--> elk.in

tasks
0

swidth
1E-4

dft+u
1 1
2 2 0.03674932 0.00000000 : is, l, U, J

spinpol
.true.

reducebf
0.7

ngridk
4 4 2

xctype
3

rgkmax
8.000000000

gmaxvr
20.00000000

trimvg
.true.

lmaxapw
8

lmaxvr
8

nempty
6

epspot
0.1000000000E-06

epsengy
0.1000000000E-05

!This is the elkstructure
scale
1.8897261329

avec
6.1553001404 0.0000000000 0.0000000000
-3.0776500702 5.3306462895 0.0000000000
0.0000000000 0.0000000000 11.4026002884

atoms
3 !NUMBER OF SPECIES
'Y.in' !KIND
6 !NUMBER OF ATOMS
0.000000000 0.000000000 0.268900007 0.000000000 0.000000000 0.000000000
0.000000000 0.000000000 0.768900037 0.000000000 0.000000000 0.000000000
0.333333343 0.666666687 0.229000002 0.000000000 0.000000000 0.000000000
0.666666627 0.333333313 0.728999972 0.000000000 0.000000000 0.000000000
0.333333313 0.666666627 0.728999972 0.000000000 0.000000000 0.000000000
0.666666687 0.333333343 0.229000002 0.000000000 0.000000000 0.000000000

'Mn.in' !KIND
6 !NUMBER OF ATOMS
0.320800006 0.000000000 0.000000000 -3.000000000 0.000000000 0.000000000
0.000000000 0.320800006 0.000000000 1.500000000 -2.598076211 0.000000000
0.679199994 0.679199994 0.000000000 1.500000000 2.598076211 0.000000000
0.679199994 0.000000000 0.500000000 -3.000000000 0.000000000 0.000000000
0.000000000 0.679199994 0.500000000 1.500000000 -2.598076211 0.000000000
0.320800006 0.320800006 0.500000000 1.500000000 2.598076211 0.000000000

'O.in' !KIND
18 !NUMBER OF ATOMS
0.310000002 0.000000000 0.162100002 0.000000000 0.000000000 0.000000000
0.000000000 0.310000002 0.162100002 0.000000000 0.000000000 0.000000000
0.689999998 0.689999998 0.162100002 0.000000000 0.000000000 0.000000000
0.689999998 0.000000000 0.662100017 0.000000000 0.000000000 0.000000000
0.000000000 0.689999998 0.662100017 0.000000000 0.000000000 0.000000000
0.310000002 0.310000002 0.662100017 0.000000000 0.000000000 0.000000000
0.638899982 0.000000000 0.336699992 0.000000000 0.000000000 0.000000000
0.000000000 0.638899982 0.336699992 0.000000000 0.000000000 0.000000000
0.361100018 0.361100018 0.336699992 0.000000000 0.000000000 0.000000000
0.361100018 0.000000000 0.836699963 0.000000000 0.000000000 0.000000000
0.000000000 0.361100018 0.836699963 0.000000000 0.000000000 0.000000000
0.638899982 0.638899982 0.836699963 0.000000000 0.000000000 0.000000000
0.000000000 0.000000000 0.000000000 0.000000000 0.000000000 0.000000000
0.000000000 0.000000000 0.500000000 0.000000000 0.000000000 0.000000000
0.333333343 0.666666687 0.013300000 0.000000000 0.000000000 0.000000000
0.666666627 0.333333313 0.513300002 0.000000000 0.000000000 0.000000000
0.333333313 0.666666627 0.513300002 0.000000000 0.000000000 0.000000000
0.666666687 0.333333343 0.013300000 0.000000000 0.000000000 0.000000000

Link to all other dos files raw data

https://www.dropbox.com/sh/mr0ugn8cvd9gkit/AAB6rfX6GQjk6LjBq98YpRLEa?dl=0

[Lars Nordström]

Hi Michael,

is it not that the code does not know what quantization axis it ought to split the states, so it uses z,
and you probably only have moments along x and y, right?

There are remedies for this, but I am uncertain if they are in the official versions.

In short, instead of calculating the spin-split DOS for a fixed axis the full 2x2 matrix g(e) should be printed out and you can afterwards split it along any choosen axis with help of appropriate combination of Pauli principle:
D_up/down(e)=Tr {(1 +/- n.sigma)/2 g(e)}
where n is the local quantization axis, usually chosen as the local moment direction.

You can have a look if such a procedure is somehow included in your version of ELK. I have a slightly more general approach included in at least one developer-version, or you do the programming yourself – it is fairly trivial, and as I think useful.

Take care,
Lars