mmaps with more than 7 elements

Alloy Theoretic Automated Toolkit (ATAT)
1

Dear all,

Hi. Thank you for providing the amazing tool and sharing the techniques. I am new to ATAT as well as VASP calculation. It would be nice if you could tell me whether mmaps is compatible to calculate equal or more than seven elements. If so, are there any tips to complete the calculations?

Below are the details of my calculations and struggling…

I have been trying to calculate a cation-disordered rocksalt oxide, in which many cations occupy the same cation site. When calculating less than 5 cations, mmaps and VASP nicely calculated the cluster expansion. However, when calculating the structure of more than 7 cations, it looks mmaps suddenly stops creating a new structure to calculate energy after creating tens of them, and eventually I was able to get only ECI of the nearest neighbor dimers whereas I need more ECI values for further distance for the subsequent study, which would require more calculations of the different structures.

I guess mmaps keeps looking for a new (large) structure, but the behavior looks strange to me such that mmaps becomes suddenly completely silent for hours after it stopped while it has been continuously creating structures in a sec or even minute.

As a reference I will share my input files for mmaps as follows:

<lat.in>

2.97784  2.97784  2.97784  60.0000  60.0000  60.0000
1.0 0.0 0.0
0.0 1.0 0.0
0.0 0.0 1.0
     0.00000    0.00000    0.00000   A, B, C, D, E, F, G
     0.50000    0.50000    0.50000   O

, where A, B, C, D, E, F, and G are cation elements, whose valence state can be +1 to +4.

<vasp.wrap>


PREC = Low
LREAL = Auto
ISIF = 6
EDIFF = 1.0e-04
EDIFFG = 1.0e-03
IBRION = 2
NSW = 50
ISMEAR = -5 ; SIGMA = 0.2
NELM = 100 ; NELMIN = 4
ALGO = Normal
ISYM = 0

LWAVE = .FALSE.
LCHARG = .FALSE.
NPAR = 6

LASPH = .TRUE.
KPPRA = 500
USEPOT = PAWPBE
KSCHEME = GAMMA
DOSTATIC

MAGATOM =
  • I know this is not an accurate calculation in terms of DFT, especially in PREC', ISIF’, EDIFF', and EDIFFG’. But, changing these accuracy-related parameters did not solve (apparently) the stopping issue of mmaps.

[typical mmaps command I used]

mmaps -d &

As a reference, when cations A-G were set to Li, Mg, Mn, Fe, Co, Ni, and Cu, I can (barely, namely mmaps got slower and slower to create new structures) get some decent results for the cluster expansion, but when I replaced Co, Ni and Cu to the different elements (say X, Y, and Z), which are my targets and not solely form the rocksalt structure, the mmaps (apparently) stop creating structure at a certain point (after tens of structures) before it `converges’.

Below is what I have tried

  1. I have used crange.in' to limit the composition region to these[i] X, Y,[/i] and [i]Z[/i] in several ways. Because the lowest valence states of X and Y are 4+ and 3+, and therefore, [i]X[/i] and[i] Y [/i]should coexist with Li+ in the rocksalt structure. As for divalent [i]Z[/i], which does not solely form the rocksalt structure, the molar fraction of [i]Z[/i] to the other elements was limited to small. I tried several combinations of the these conditions and boundary values. But, it did not improve the situation of the sudden stop’ of mmaps at a certain point.

  2. I used checkrelax' and exclude some structures whose structures were too relaxed (&gt;0.07) by creating flags [i]n[/i]/error, refresh’, and restart mmaps. When restarting mmaps, I also tried to used -g option.

  3. Changing optional values of mmpas such as -c and -m.

I’d really appreciate it if you could help the calculations.