Praseodymium Perovskite Substitution & Energy Fluctuations

[luke_pretzie]

I am trying to determine the consequences of replacing the Lanthanum in a perovskite - K2La2Ti3O10 - with Praseodymium - K2Pr2Ti3O10.

The Lanthanum perovskite was simple enough to work with. I found its POSCAR file at https://next-gen.materialsproject.org/m ... s/mp-6548/ and performed convergence tests on the bulk structure using ISIF=3. The following INCAR file was used:
---
ENCUT = 400

ALGO = Normal

EDIFF = 1e-6

EDIFFG = -0.05

IBRION = 2

GGA = PE

ISPIN = 1

ISMEAR = 1

SIGMA = 0.1

LCHARG = .FALSE

LWAVE = .FALSE

LREAL = Auto #(.FALSE, .TRUE or Auto)

NSW = 100

NELM = 100

ISIF = 3

NPAR = 8
---
---
Now the KPOINT grid density:
---
---
Gamma-point only
0
Monkhorst Pack
6 6 3
0 0 0
---
---
Lastly, the POSCAR:
---
---
K La Ti O
1.00000000000000
3.8876621302235486 -0.0000000000000000 0.0000000000000002
0.0000000000000006 3.8876621302235486 0.0000000000000002
-0.0000000000000000 -0.0000000000000000 29.6325787428460679
K La Ti O
4 4 6 20
Direct
-0.0000000000000000 -0.0000000000000000 0.2918359623112963
0.5000000000000000 0.5000000000000000 0.2081640376887037
0.5000000000000000 0.5000000000000000 0.7918359623112964
0.0000000000000000 0.0000000000000000 0.7081640376887036
0.0000000000000000 0.0000000000000000 0.4283179818116329
0.5000000000000000 0.5000000000000000 0.0716820181883671
0.5000000000000000 0.5000000000000000 0.9283179818116328
-0.0000000000000000 -0.0000000000000000 0.5716820181883672
0.0000000000000000 0.0000000000000000 0.1403709899647160
0.5000000000000000 0.5000000000000000 0.3596290100352841
-0.0000000000000000 0.0000000000000000 -0.0000000000000000
0.5000000000000000 0.5000000000000000 0.6403709899647160
-0.0000000000000000 -0.0000000000000000 0.8596290100352840
0.5000000000000000 0.5000000000000000 0.5000000000000000
0.5000000000000000 0.5000000000000000 0.2995903599533972
-0.0000000000000000 -0.0000000000000000 0.2004096400466028
0.5000000000000000 -0.0000000000000000 0.3731016614096467
-0.0000000000000000 0.5000000000000000 0.3731016614096467
0.0000000000000000 0.5000000000000000 0.1268983385903533
0.5000000000000000 0.0000000000000000 0.1268983385903533
0.5000000000000000 0.5000000000000000 0.4350746280993091
0.5000000000000000 0.0000000000000000 -0.0000000000000000
-0.0000000000000000 0.5000000000000000 -0.0000000000000000
-0.0000000000000000 -0.0000000000000000 0.0649253719006909
0.0000000000000000 0.0000000000000000 0.7995903599533972
0.5000000000000000 0.5000000000000000 0.7004096400466028
-0.0000000000000000 0.5000000000000000 0.8731016614096467
0.5000000000000000 -0.0000000000000000 0.8731016614096467
0.5000000000000000 0.0000000000000000 0.6268983385903533
0.0000000000000000 0.5000000000000000 0.6268983385903533
0.0000000000000000 0.0000000000000000 0.9350746280993092
-0.0000000000000000 0.5000000000000000 0.5000000000000000
0.5000000000000000 0.0000000000000000 0.5000000000000000
0.5000000000000000 0.5000000000000000 0.5649253719006908
---
Running this job results in a converged, well-behaved system within a reasonable time frame. If, however, I instead use Pr in place of La, there appears to be no electronic convergence, but rather a "fluctuation" in the "dE' value. Here's a snippet of the OSZICAR output:

N E dE d eps ncg rms rms(c)
DAV: 1 0.145975798243E+04 0.14598E+04 -0.97119E+03 53392 0.104E+02
DAV: 2 -0.227526068078E+03 -0.16873E+04 -0.12296E+04 53160 0.398E+01
DAV: 3 -0.338308220787E+03 -0.11078E+03 -0.10814E+03 58560 0.120E+01
DAV: 4 -0.338736500520E+03 -0.42828E+00 -0.42450E+00 60856 0.664E-01
DAV: 5 -0.338737683750E+03 -0.11832E-02 -0.11811E-02 56144 0.293E-02 0.455E+01
DAV: 6 -0.283146680534E+03 0.55591E+02 -0.26598E+00 54232 0.357E+00 0.296E+01
DAV: 7 -0.277952611639E+03 0.51941E+01 -0.42640E+00 53808 0.265E+00 0.187E+01
DAV: 8 -0.275346150199E+03 0.26065E+01 -0.13713E+01 52128 0.979E-01 0.136E+01
DAV: 9 -0.274696063193E+03 0.65009E+00 -0.79191E+00 53464 0.327E-01 0.108E+01
DAV: 10 -0.276766466914E+03 -0.20704E+01 -0.42097E-01 52328 0.300E-01 0.829E+00
DAV: 11 -0.274591445134E+03 0.21750E+01 -0.10070E+00 52992 0.538E-01 0.958E+00
DAV: 12 -0.276409398089E+03 -0.18180E+01 -0.13526E-01 52296 0.354E-01 0.847E+00
DAV: 13 -0.273936839996E+03 0.24726E+01 -0.90424E-01 51408 0.295E-01 0.689E+00
DAV: 14 -0.276017007945E+03 -0.20802E+01 -0.31021E-01 51032 0.258E-01 0.570E+00
DAV: 15 -0.273919387049E+03 0.20976E+01 -0.38756E-01 51024 0.203E-01 0.626E+00
DAV: 16 -0.275887861136E+03 -0.19685E+01 -0.17828E-01 50664 0.225E-01 0.479E+00
DAV: 17 -0.273907121630E+03 0.19807E+01 -0.81824E-02 50624 0.190E-01 0.596E+00
DAV: 18 -0.275867266720E+03 -0.19601E+01 -0.84742E-02 50544 0.213E-01 0.496E+00
DAV: 19 -0.273899500314E+03 0.19678E+01 -0.12054E-01 50488 0.189E-01 0.575E+00
DAV: 20 -0.275885777553E+03 -0.19863E+01 -0.13249E-01 50456 0.214E-01 0.489E+00
DAV: 21 -0.273883030188E+03 0.20027E+01 -0.71822E-02 50464 0.187E-01 0.563E+00
DAV: 22 -0.275914909469E+03 -0.20319E+01 -0.16231E-01 50384 0.214E-01 0.495E+00
---
So far, following the advice given at https://www.researchgate.net/post/Energ ... went-wrong , as well as increasing the NBANDS value, has yet to resolve the issue. Any further advice would be appreciated.

[svijay]

Dear luke_pretzie,

Yes indeed it does look like the energy is jumping between two values and hence not converging. Have you tried a different ALGO (maybe ALGO=All?)

Sudarshan

[luke_pretzie]

Hello Sudarshan,

I had tried ALGO=VeryFast, but that did not work. ALGO=All does seem to work! Thank you