supercell and doping

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Tue Sep 11, 2012 9:16 am

PLEASE HELP ME, i want to calculate GS of Ca(0.75)Ba(0.25)B6, i want to have supercell of 28 atoms . What i am trying to do is doping of CaB6, my questions are as follows i dont know how to write my xred because i dont know how to get the atomic positions of my doped system, second question is what will be my mixchal input and i want all to assist in correcting my input file.
This is my input file,
############################################################
# GS of Ca(0.75)Ba(0.25)B6
############################################################

#Datasets: etotal vs acell
ndtset 14 acell: 3*7.50 acell+ 3*0.05

ionmov 3 # Use the modified Broyden algorithm
ntime 10 # Maximum number of Broyden "timesteps"
tolmxf 5.0d-4 # Stopping criterion for the geometry optimization : when
# the residual forces are less than tolmxf, the Broyden
#Basis definitions
ecut 95.
ecutsm 0.5
# pawecutdg 24.
#K-points and sym
nsym 0
occopt 1
ngkpt 4 4 4
nshiftk 1
shiftk 0.5 0.5 0.5

#I/O parameters
optforces 2 optstress 1
prtwf 1 prtden 0 prteig 0
getwfk -1
#Definition of the SCF procedure
nstep 150 # Maximal number of SCF cycles
toldfe 1.0d-6 # Will stop when, twice in a row, the difference
# between two consecutive evaluations of total energy
# differ by less than toldfe (in Hartree)
# This value is way too large for most realistic studies of materials
diemac 12.0 # Although this is not mandatory, it is worth to
# precondition the SCF cycle. The model dielectric
# function used as the standard preconditioner
# is described in the "dielng" input variable section.
# Here, we follow the prescription for bulk silicon.
# tolvrs 1.0d-10
nband 16
# add to conserve old < 6.7.2 behavior for calculating forces at each SCF step

#COMMON INPUT DATA###########################################################
#
# CaB6 lattice structure
npsp 3
ntypat 3
ntypalch 1
mixalch 1 0.25
#ntyppure 1
znucl 5 20 56
natom 8
typat 1 2 2 2 2 2 2 3

#Definition of the unit cell
#acell 3*8.1
angdeg 90 90 90
spgroup 221

# Atomic positions
xred 0.00 0.00 0.00 #Ca1
0.50 0.50 0.20 #B2
0.50 0.50 -0.20 #B2
0.50 0.20 0.50 #B2
0.50 -0.20 0.50 #B2
0.20 0.50 0.50 #B2
0.80 0.50 0.50 #B2
0.00 0.00 0.10 #Ba3

ixc 11
timopt 2
At all, thanks for your usuall assistance.

maxim · Post by **maxim** » Tue Sep 11, 2012 12:06 pm

Hi!

Your ecut is huge (too much, especially for PAW), you should remove "ntime 10" because you already use tolmxf. Your forces probably not converged - to get it investigate dependence of forces on toldfe - it's likely "toldfe 1.0d-6" should be replaced by "toldfe 1.0d-12".
Maybe it is useful to use optcell 0 first, and then (using getxred-1) set optcell 2 (you should set dilatmx also).
Probably you have to decide for yourself whether you want to dope using alchemical psps or not use them. In the latter case you should create a supercell to mimic doping by putting cations in a specific ratio you want. Simultaneous use of supercell and alchemical psps maybe unwise.
I unfortunately did not use alchemical psps, but it seems that: "ntypat 2", "mixalch 0.75 0.25", uncomment acell, "znucl 20 56 5" and change "typat" and maybe "xred" accordingly.

Best wishes,
Maxim

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Wed Sep 12, 2012 9:39 am

Dear Maxim
Thank you, your suggestion works well. One more question, can u pls put me through how to generate supercell.
Thanks.

maxim · Post by **maxim** » Wed Sep 12, 2012 11:25 am

Can you give me a CIF-file (or other file) from which you initially defined the structure?

Best wishes,
Maxim

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Wed Sep 12, 2012 12:24 pm

Dear Maxim,
this is my initial structure,
0.00 0.00 0.00 #Ca1
0.50 0.50 0.20 #B2
0.50 0.50 -0.20 #B2
0.50 0.20 0.50 #B2
0.50 -0.20 0.50 #B2
0.20 0.50 0.50 #B2
0.80 0.50 0.50 #B2
My system CaB6 is simple cubic.
Thanks.

maxim · Post by **maxim** » Wed Sep 12, 2012 1:31 pm

remove spgroup variable, check your acell parameter - it should be in bohr units and multiolied by 2 because you use 2x2x2 supercell.
here is xred for 2x2x2 supercell:
0.00 0.00 0.00
0.25 0.25 0.10
0.25 0.25 0.40
0.25 0.10 0.25
0.25 0.40 0.25
0.10 0.25 0.25
0.40 0.25 0.25
0.50 0.00 0.00
0.75 0.25 0.10
0.75 0.25 0.40
0.75 0.10 0.25
0.75 0.40 0.25
0.60 0.25 0.25
0.90 0.25 0.25
0.00 0.50 0.00
0.25 0.75 0.10
0.25 0.75 0.40
0.25 0.60 0.25
0.25 0.90 0.25
0.10 0.75 0.25
0.40 0.75 0.25
0.50 0.50 0.00
0.75 0.75 0.10
0.75 0.75 0.40
0.75 0.60 0.25
0.75 0.90 0.25
0.60 0.75 0.25
0.90 0.75 0.25
0.00 0.00 0.50
0.25 0.25 0.60
0.25 0.25 0.90
0.25 0.10 0.75
0.25 0.40 0.75
0.10 0.25 0.75
0.40 0.25 0.75
0.50 0.00 0.50
0.75 0.25 0.60
0.75 0.25 0.90
0.75 0.10 0.75
0.75 0.40 0.75
0.60 0.25 0.75
0.90 0.25 0.75
0.00 0.50 0.50
0.25 0.75 0.60
0.25 0.75 0.90
0.25 0.60 0.75
0.25 0.90 0.75
0.10 0.75 0.75
0.40 0.75 0.75
0.50 0.50 0.50
0.75 0.75 0.60
0.75 0.75 0.90
0.75 0.60 0.75
0.75 0.90 0.75
0.60 0.75 0.75
0.90 0.75 0.75

typat 2 1 1 1 1 1 1 2 1 1 1 1 1 1 2 1 1 1 1 1 1 2 1 1 1 1 1 1 2 1 1 1 1 1 1 2 1 1 1 1 1 1 2 1 1 1 1 1 1 2 1 1 1 1 1 1

this is 2x2x2 supercell for CaB3.
...And for Ca(0.75)Ba(0.25)B6 you may replace now in typat variable above any two "2" by "3" and you should consider different cases of this replacement: structure where Ba atoms are near and structure where Ba atoms are equidistant. You can see this supercell calculation may be quite tedious

. Сalculation of 1x2x2 supercell would be much less tedious, but the supercell will be slightly flat and there will be interaction between the flat parts of the neighboring supercells. Alchemical pseudos perhaps the best way of all.

Best wishes,
Maxim

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Wed Sep 12, 2012 2:40 pm

Dear Maxim,
thanks a lot, i appreciate this reply, but pls tell me the rules to follow whenever i want to generate supercell from unit cell. If there are materials to read please send it to me or to my box bamgbose@physics.unaab.edu.ng . Thanks.

maxim · Post by **maxim** » Wed Sep 12, 2012 8:07 pm

As for me, I did not use special books - first I simply divided xred by 2 (let's denote this set of coordinates by S), then consistently added 0.5 to x-coordinates of S, then to y of S, then to x and y of S, then I added 0.5 to z of all previous coordinates.

Best wishes,
Maxim

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Thu Sep 13, 2012 11:14 am

Dear Maxim,
Thanks a lot, but your explanation is not clear to me. I want you to please shed more light on this, if u divide xred by 2 and add 0.5 your third line of supercell suppose to be 0.75 for x-axis and soon. please and please shed more light.
Thanks.

maxim · Post by **maxim** » Thu Sep 13, 2012 3:33 pm

Let us denote the xred of original cell by "x y z". Then the supercell will be:
x/2 y/2 z/2
x/2+0.5 y/2 z/2
x/2 y/2+0.5 z/2
x/2+0.5 y/2+0.5 z/2
x/2 y/2 z/2+0.5
x/2+0.5 y/2 z/2+0.5
x/2 y/2+0.5 z/2+0.5
x/2+0.5 y/2+0.5 z/2+0.5

Best wishes,
Maxim

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Fri Sep 14, 2012 2:54 am

Dear Maxim,
Thanks a lot, i can write out xred of supercell. Now look at my input file, i want to do optimization of lattice parameter, and plot etot vs acell in order to obtain a parabola curve, but my total energy continues to decrease without stopping at a particular point and then increases. What could be the problem please help, this is my input file,
############################################################
# GS of Ca(0.75)La(0.25)B6
############################################################

#Datasets: etotal vs acell
ndtset 18 acell: 3*6.50 acell+ 3*0.05

ionmov 3 # Use the modified Broyden algorithm
ntime 10 # Maximum number of Broyden "timesteps"
tolmxf 5.0d-4 # Stopping criterion for the geometry optimization : when
# the residual forces are less than tolmxf, the Broyden
#Basis definitions
ecut 95.
ecutsm 0.5
# pawecutdg 24.
#K-points and sym
nsym 0
occopt 1
ngkpt 4 4 4
nshiftk 1
shiftk 0.5 0.5 0.5

#I/O parameters
optforces 2 optstress 1
prtwf 1 prtden 0 prteig 0
getwfk -1
#Definition of the SCF procedure
nstep 150 # Maximal number of SCF cycles
toldfe 1.0d-12 # Will stop when, twice in a row, the difference
# between two consecutive evaluations of total energy
# differ by less than toldfe (in Hartree)
# This value is way too large for most realistic studies of materials
diemac 12.0 # Although this is not mandatory, it is worth to
# precondition the SCF cycle. The model dielectric
# function used as the standard preconditioner
# is described in the "dielng" input variable section.
# Here, we follow the prescription for bulk silicon.
# tolvrs 1.0d-10
nband 10
# add to conserve old < 6.7.2 behavior for calculating forces at each SCF step

#COMMON INPUT DATA###########################################################
#alchemical input
npsp 3
ntypalch 1
mixalch 0.75 0.25
#
# CaB6 lattice structure
ntypat 2
znucl 5 20 56
natom 7
typat 1 2 2 2 2 2 2

#Definition of the unit cell
#acell 3*8.1
angdeg 90 90 90
spgroup 221

# Atomic positions
xred 0.00 0.00 0.00 #Ca1
0.50 0.50 0.20 #B2
0.50 0.50 -0.20 #B2
0.50 0.20 0.50 #B2
0.50 -0.20 0.50 #B2
0.20 0.50 0.50 #B2
0.80 0.50 0.50 #B2

ixc 11
timopt 2

maxim · Post by **maxim** » Fri Sep 14, 2012 10:50 pm

Hi!

but my total energy continues to decrease without stopping at a particular point and then increases

- give, please, more explanation.
Remove "ntime 10" - If you already use tolmxf as stopping criterion, there is no need to use ntime. Set ionmov to 2. How many electrons does La pseudopotential have (about 15 and more)? Increase nband to accommodate all electrons of all psp you have in the cell (2 electrons per 1 band - 10 bands maybe not enough). Your ecut is too high - this unnecessarily loads your PC - perform convergence in ecut.

Best wishes,
Maxim

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Fri Sep 14, 2012 11:17 pm

Dear Maxim,
I say big thank you, i love talking to you on this forum. I will effect the correction and get back to you.
Thanks.

maxim · Post by **maxim** » Sun Sep 16, 2012 11:28 am

I forgot to tell you - you have to check the convergence in ngkpt - it may (or may not) have a bit higher value for the cubic crystals (perform convergence in forces test - run job with different values of ngkpt and plot dependences of forces on ngkpt). Also, is your system FCC, BCC or smth else? - depending on this your shiftk will have different values - see description of shiftk variable.

Best wishes, Maxim

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Sat Sep 22, 2012 11:31 am

Dear Maxim,
Thank you very much. I can infer from your replies that i have to do ecut convergence, follow by ngkpt convergence and finally check my shiftk, but i want you to please look at my input file again and see if there are other things to put or adjust.

#Datasets: etotal vs acell
ndtset 14 acell: 3*6.5 acell+ 3*0.05

ionmov 2 # Use the modified Broyden algorithm
tolmxf 5.0d-4 # Stopping criterion for the geometry optimization : when
# the residual forces are less than tolmxf, the Broyden
#Basis definitions
ecut 65.
ecutsm 0.5
# pawecutdg 24.
#K-points and sym
nsym 0
occopt 1
ngkpt 4 4 4
nshiftk 1
shiftk 0.5 0.5 0.5

#I/O parameters
optforces 2 optstress 1
prtwf 1 prtden 0 prteig 0
getwfk -1
#Definition of the SCF procedure
nstep 150 # Maximal number of SCF cycles
toldfe 1.0d-12 # Will stop when, twice in a row, the difference
# between two consecutive evaluations of total energy
# differ by less than toldfe (in Hartree)
# This value is way too large for most realistic studies of materials
diemac 12.0 # Although this is not mandatory, it is worth to
# precondition the SCF cycle. The model dielectric
# function used as the standard preconditioner
# is described in the "dielng" input variable section.
# Here, we follow the prescription for bulk silicon.
# tolvrs 1.0d-10
nband 25
# add to conserve old < 6.7.2 behavior for calculating forces at each SCF step

#COMMON INPUT DATA###########################################################
#alchemical input
npsp 3
ntypalch 1
mixalch 0.75 0.25
#
# CaB6 lattice structure
ntypat 2
znucl 5 20 56
natom 7
typat 1 2 2 2 2 2 2

#Definition of the unit cell
#acell 3*8.1
angdeg 90 90 90
spgroup 221

# Atomic positions
xred 0.00 0.00 0.00 #Ca1
0.50 0.50 0.20 #B2
0.50 0.50 -0.20 #B2
0.50 0.20 0.50 #B2
0.50 -0.20 0.50 #B2
0.20 0.50 0.50 #B2
0.80 0.50 0.50 #B2

ixc 11
timopt 2

Maxim once again thank you very much.

maxim · Post by **maxim** » Sat Sep 22, 2012 12:46 pm

Dear BAMGBOSE,

It seems that everything is correct.
Increase nstep to 1000. Is not critical, but if your object is an insulator, then set diemac to 4. Remove spgroup variable - abinit will recognize space group automatically. You can make sure that you select the correct value of nband in the following way: run input with removed nband variable (abinit will allocate enought bands automatically) and look at "occ" in section "outvars: echo values of variables after computation" or "occupation numbers for kpt#" in *.out file and then set nband to the number of occupied bands. Maybe you need higher value of ngkpt (6x6x6) (have you performed convergence in forces test? - run job with different values of ngkpt and plot dependences of forces on ngkpt).
Maybe later you should try spin-poolarized calculation if this is your case (see the corresponding tutorial in abinit).

Best wishes, Maxim

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Mon Sep 24, 2012 3:41 pm

Dear Maxim,
Thanks for your assistance and prompt response all along. I have done all the correction you suggested, but the total energy versus lattice parameter does not give me a parabola curve.
If you are to do this kind of calculation, how will your input file look like or please prepare input file for this problem or something similar.
Thanks

maxim · Post by **maxim** » Tue Sep 25, 2012 9:07 am

Dear BAMGBOSE,

I do not know what it was, I did not use alchemical pseudos. Try to construct a parabola for occupations with Ca or with Ba only. Parabola can not be obtained for simultaneous occupation only because the doping creates a new phase, different from you set. If so, then for this simultaneous occupation use internal optimization (optcell=0), then start the cell shape and size optimization from the cell with relaxed ionic positions (optcell=2). In order to obtain parabola in this case you should (at each point of parabola) use not only optcell=0, but constant-volume optimization optcell=3. Always set nsym=1 to break the symmetries, because otherwise symmetry finder of abinit will recognize a cubic structure, and will strive to maintain it as a cubic. First, to get successful geometry optimization you should obtain correct forces - have you performed convergence in forces test? - run job (with cell parameter close to experimental) with different values of ngkpt and plot dependences of forces on ngkpt. Maybe you need higher value of ngkpt (6x6x6).
Another reason may be pseudos - try another psps (not GGA fhi, but LDA fhi with ixc=1, etc.).

Best wishes, Maxim

maxim · Post by **maxim** » Sat Sep 29, 2012 10:27 am

Hello BAMGBOSE!

I've edited your file. First, are your pseudos GGA? If so, ixc 11 is correct. If your pseudos are LDA, then remove ixc.
Try to check whether this gives a parabola first (check whether I have correctly entered correct parameters acell for CaB6 from here http://www.crystallography.net/informat ... if=1010302):

Code: Select all

#Datasets: etotal vs acell
ndtset 14 acell: 3*7.592 acell+ 3*0.05 

ionmov 2 # Use the modified Broyden algorithm 
tolmxf 5.0d-5 # Stopping criterion for the geometry optimization : when 
# the residual forces are less than tolmxf, the Broyden 
#Basis definitions
ecut 65.
ecutsm 0.5
# pawecutdg 24.
#K-points and sym
nsym 1 
occopt 1
ngkpt 6 6 6 
nshiftk 1
shiftk 0.5 0.5 0.5

#I/O parameters
optforces 2 optstress 1
prtwf 1 prtden 0 prteig 0
getwfk -1
#Definition of the SCF procedure
nstep 1000 # Maximal number of SCF cycles
toldfe 1.0d-12 # Will stop when, twice in a row, the difference 
# between two consecutive evaluations of total energy 
# differ by less than toldfe (in Hartree) 
# This value is way too large for most realistic studies of materials
#diemac 12.0 # Although this is not mandatory, it is worth to
# precondition the SCF cycle. The model dielectric
# function used as the standard preconditioner
# is described in the "dielng" input variable section.
# Here, we follow the prescription for bulk silicon.
# tolvrs 1.0d-10
#nband 25
# add to conserve old < 6.7.2 behavior for calculating forces at each SCF step



#COMMON INPUT DATA###########################################################
#alchemical input
#npsp 3
#ntypalch 1
#mixalch 0.75 0.25 
#
# CaB6 lattice structure
ntypat 2
znucl 5 20
natom 7
typat 1 2 2 2 2 2 2 

#Definition of the unit cell
#acell 3*8.1
angdeg 90 90 90
#spgroup 221

# Atomic positions 
xred 0.00 0.00 0.00 #Ca1
0.50 0.50 0.20 #B2
0.50 0.50 -0.20 #B2
0.50 0.20 0.50 #B2
0.50 -0.20 0.50 #B2
0.20 0.50 0.50 #B2
0.80 0.50 0.50 #B2


ixc 11
timopt 2

Tell me about your results. If no parabola, then maybe ramove ixc and try LDA pseudos.

Best wishes, Maxim

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Sun Sep 30, 2012 6:12 pm

Dear Maxim,
I appreciate your assistance and i will get back to as soon as i am through with the work u gave me.
Thanks.

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Mon Oct 01, 2012 7:57 am

Dear Maxim,
I did the calculation you gave and i got parabola with minimum energy of -21.9348Ha and the corresponding lattice parameter(acell) is 7.792Bohr. Now maxim remember i am working on Barium doped CaB6 and not pure CaB6, how do i go about this and i don't want alchemical method.
Thanks a lot.

maxim · Post by **maxim** » Mon Oct 01, 2012 4:41 pm

Hello!

Are your pseudos GGA? If so, ixc 11 is correct. If your pseudos are LDA, then remove ixc.
Try this:

Code: Select all

#Datasets: etotal vs acell
ndtset 28  udtset 2 14
acell:? 3*7.592 acell+? 3*0.05 

#Internal optim. first, then constant-volume optim.
getxcart?2 1
optcell?1 0
ionmov?1 2
optcell?2 3
ionmov?2 2
ntime 40 # Maximum number of Broyden "timesteps"
tolmxf 5.0d-5 # Stopping criterion for the geometry optimization : when
dilatmx 1.05

ionmov 2 # Use the modified Broyden algorithm 
tolmxf 5.0d-5 # Stopping criterion for the geometry optimization : when 
# the residual forces are less than tolmxf, the Broyden 
#Basis definitions
ecut 65.
ecutsm 0.5
# pawecutdg 24.
#K-points and sym
nsym 1 
occopt 1
ngkpt 6 6 6 
nshiftk 1
shiftk 0.5 0.5 0.5

#I/O parameters
optforces 2 optstress 1
prtwf 1 prtden 0 prteig 0
getwfk -1
#Definition of the SCF procedure
nstep 1000 # Maximal number of SCF cycles
toldfe 1.0d-12 # Will stop when, twice in a row, the difference 
# between two consecutive evaluations of total energy 
# differ by less than toldfe (in Hartree) 
# This value is way too large for most realistic studies of materials
#diemac 12.0 # Although this is not mandatory, it is worth to
# precondition the SCF cycle. The model dielectric
# function used as the standard preconditioner
# is described in the "dielng" input variable section.
# Here, we follow the prescription for bulk silicon.
# tolvrs 1.0d-10
#nband 25
# add to conserve old < 6.7.2 behavior for calculating forces at each SCF step



#COMMON INPUT DATA###########################################################
#alchemical input
npsp 3
ntypalch 1
mixalch 0.95 0.05 
#
# CaB6 lattice structure
ntypat 2
znucl 5 20 56
natom 7
typat 1 2 2 2 2 2 2 

#Definition of the unit cell
#acell 3*8.1
angdeg 90 90 90
#spgroup 221

# Atomic positions 
xred 0.00 0.00 0.00 #Ca1
0.50 0.50 0.20 #B2
0.50 0.50 -0.20 #B2
0.50 0.20 0.50 #B2
0.50 -0.20 0.50 #B2
0.20 0.50 0.50 #B2
0.80 0.50 0.50 #B2


ixc 11
timopt 2

When you build parabola here look at *2 datasets.
If ok, then try mixalch 0.85 0.15 and finally try mixalch 0.75 0.25

Best wishes, Maxim

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Mon Oct 01, 2012 5:25 pm

Dear Maxim,
Thank you very much, i appreciate your kind gesture. Once again thank you, i will do the calculation as instructed and get back to you soonest.
Thanks.

BAMGBOSE, M. K. · Post by **BAMGBOSE, M. K.** » Sat Oct 06, 2012 12:51 pm

Dear Maxim,
The file you sent to me will take weeks to run on my computer. I want you to link me up with high performing computer that has abinit code, in my country electricity supply is not stable, so please help me out. In the file you send to me udtset value is 14 but according to abinit manual udtset value should not be more than 9, may be i should adjust this value and that of ndtset.
Thanks alot.

maxim · Post by **maxim** » Sat Oct 06, 2012 1:59 pm

BAMGBOSE, M. K. wrote:Dear Maxim,
I want you to link me up with high performing computer that has abinit code, in my country electricity supply is not stable.
Thanks alot.

I do not have acess to HPCs, but you can divide this file into small datasets and run some of them on nanohub and your PC simultaneously (note that FHI and maybe LDA pseudos are only available there); to save your PC from damage use UPS.

Best wishes, Maxim.

ABINIT Discussion Forums

supercell and doping

supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping

Re: supercell and doping