General

Alloy Theoretic Automated Toolkit (ATAT)
1

Hi Axel,
Where do I specify alternative pseudopotential to be used in CE? Is it in lat.in (e.g., Ti vs Ti_pv)?

Thanks.
Gautam

2

That would work but the cleanest way is to put

SUBATOM = s/Ti/Ti_pv/g

in your vasp.wrap file.
That way you can change it easily.

3

Hi Axel,
I tried in lat.in by adding Ti_pv after each atom position, and it seems to work.

I have two questions, this time lat.in itself.
(i) I have a case of Orthorhombic structure for which I would like to do CE. It has the Pearson symbol oC4, sp. gr. #64, and has only one Wyckoff position. If I run cellcvrt -s option, will it generate smaller cell?

(ii) In this case, should I run maps or mmaps? I used conventional unit cell as lat.in and ran maps, then killed the job. In the process, it generated a message "For coupled cluster expansions use mmaps instead". What does it mean? When should I use maps and mmaps. A related issue: when should I use conventional cell and when a smaller cell?

Thanks,
Gautam

4

If you give the conventional orthorhombic cell to cellcvrt -s when the spacegroup has is oC4, the code will output a smaller primitive unit cell (by a factor 2).

If all your sites with multiple occupation have the same two species, you can use maps. Otherwise mmaps is needed. (If you think maps should work, but it tells you to use mmaps, check for spelling errors in the species.)

5

Hi Axel,
Thanks. Indeed, after running cellcvrt -s option it generated a smaller cell with two atoms, then I had to run cellcvrt with -abc option to get the output in lat.in format.

Yesterday, I forgot to ask another important question regarding the output of mc code. Currently, do you output the configurational entropy? If so, which column and what is the unit?

A couple years ago I asked the same question to Marcel Sluiter about his imrcvm code, but his answer was imrcvm is not programmed to handle arbitrary pairs, triplets and quads (as in ATAT). Re-programming his code to handle arbitrary pairs, triplets and quads would involve significant re-writing and time. So, I was wondering if ATAT currently does the same. If not, would it be possible to output configurational entropy in future release of ATAT?

Thanks,
Gautam

6

In general Monte Carlo does not directly give a configurational entropy. A way to get it is to obtain the free energy via thermodynamic integration (see https://arxiv.org/abs/cond-mat/0201473 or https://dx.doi.org/10.1088/0965-0393/10/5/304 ). And the you can get S=[-dF/dT] at constant composition.

7

Hi Axel,
Thus far, I have using ATAT3.0. But, now I understand there is a newer version! Where do I get it from?

Regarding ATAT:
(i) If I do CE of any property, say energy, volume etc, is there a command which gives me the property at a composition I want? Say energy(of mixing) at A_0.38B_0.62 or A_0.182B_0.232_C_0.586? Similarly, for volume etc. If so, please let me know the on-line command.

In addtion, I have a couple of questions regarding phonon DOS:
(i) What are the general features one should look for in a phonon DOS, especially regarding phase stability, phase transformation etc matter the most;
(ii) Related to (i), when partial phonon DOS is(are) needed to understand the same.

Thanks,
Gautam

8

  1. New versions are continuously posted at https://alum.mit.edu/www/avdw/atat/
    You have the choice between an older "Stable version" or a "Beta version of the latest update"

  2. create a structure file str.out at the composition of interest (in could be ordered - only its composition matters). I am assuming your also have lat.in clusters.out and eci.out files.
    Then use

corrdump -rnd -eci=eci.out -c
  1. (i) no idea, sorry, other than just all frequencies generally moving up means lower entropy and lower stability (and vice-versa)
    (ii) I find projected phonon DOS not so useful, but I’ve added the feature upon request by users.
9

Hi Axel,
Thanks. I will download the latest atat package from the site you mentioned.

Sorry, I have problem understanding your reply. Specifically, what is meant by Code: Select All?
CODE: SELECT ALL
corrdump -rnd -eci=eci.out -c
If I generate a structure at composition of my interest, then the above command will give me "energy of mixing". Is it what you meant? Similarly, if I want volume of that structure I use -eci=eci_volume.out, and so on. Is it correct?

Gautam

10

Correct (sorry for overlooking this post for so long!).

The "CODE: Select All" comes from the forum software to allow you to easily copy it - ignore.

11

Hi Axel,
I bought a new laptop. In had to do it because I do not have access to PSC clusters where I have been running ATAT. XSEDE has a new rule that if I do not have NSF funding, which I don not, I can not get XSEDE allocation. As a result, they rejected my proposal three times!

I am trying to install ATAT in my new linux laptop. Would you please tell me how to bind VASP and MAPS for running CE. Of course, I do remember I have create something to define the path of POTCAR file(s). Other than that, what else do I need to do?

Thanks,
Gautam

12

As explained at https://www.brown.edu/Departments/Engineering/Labs/avdw/atat/manual/node11.html
you can just run ezvasp
and it will guide you through the steps of configuring.

13

Hi Axel,
I am planning to write a paper on what I did for the Ni-Re and Ni-Ru systems (long ago and my first attempt to use ATAT). At this time, the plan is to prepare for an Acta Mater paper. But, I have a short question:
I have calculated heat of mixing of fcc, bcc, and hcp solid solutions (i.e., random solid solutions). W/o vibrational contributions, we all know that the temperature scale is not correct. At the same time, if we do Calphad modeling we have to assign a temperature. So, the question is what temperature do we assign?

I am thinking to take the highest solidus temperature. Does it make sense to you?

Please, let me know your opinion.

Thanks,
Gautam

14

I would assign it a very low T (room T perhaps) because it that limit vibrational effects are truly small so the ab initial data is correct.
Note that if your free energy model is linear in T, then it doesn’t matter what T is (but you probably know that!)

15

Hi Axel,
Thanks for the hints!
Actually, it makes life easier! Because, we have bcc (virtual phase) which does not show up in equilibrium phase diagram, hence, there are no phase boundaries involving the bcc phase! In that sense, assigning room temp to calculated heat of mixing (random solid solution) makes life easier!

But, did you get confused "heat of mixing" with "free energy of mixing"? Note, I am talking about "heat of mixing" of Ni-Re and Ni-Ru alloys! My experience with Al-TM alloys (Acta Mater Paper in 2008) is that MC simulation of all phases exhibit SRO/clustering effect, and only at very high temperature they exhibit random solid solution behavior! But, now you are suggesting me to take T=300 to assign. Hence, I am confused. Even for a random solid solution if we consider "heat of mixing" in the POP file in Thermo-Calc, we have to define the temperature (to make degree of freedom zero along with comp, pressure, and n=1).

Please take a look in our Acta Mater 2008 paper. We discussed in the issue of temp and calculated heat of mixing via MC simulation (in several binary solid solutions). Would you please clarify the issue?

Thanks,
Gautam

16

Hi Axel,
I installed atat_3_04 in a NERSC computer, and followed all instructions described in ATAT usre guide. But, when I try to submit the job for CE, it does not accept the script file! Of course, I will consult NERSC people tomorrow about the problem. However, I do not understand where do I mention the path for VASP executable in the script file!

A script file that works (or, accepted by the same computer) looks like

#!/bin/csh
#PBS -q regular
#PBS -l mppwidth=96
#PBS -l mppnppn=24
#PBS -l walltime=24:00:00
#PBS -N sigmaNi
#PBS -e sigma-N.err
#PBS -o sigma-N.out
#PBS -V

change to the appropriate directory

cd /global/u1/g/gghosh/TCP/sigma
#cp 5.5264.poscar POSCAR
aprun -n 96 /global/u1/g/gghosh/hopper2/vasp.4.6.28/vasp.4.6/vasp-Hopper2
mv CONTCAR 5.5264.contcar
mv OSZICAR 5.5264.oszicar
mv OUTCAR 5.5264.outcar

Here, the path of VASP executable is clearly define. My question is, where do I define a similar path in the script file for CE?

#!/bin/csh
#PBS -q regular
#PBS -l mppwidth=96
#PBS -l mppnppn=24
#PBS -l walltime=48:00:00
#PBS -j eo
#PBS -q batch
#PBS -N Ni-Re-BCC-CE
#PBS -o nire-bcc-CE.out
#PBS -e nire-bcc-CE.err

set echo

change to appropriate directory

cd /global/u1/g/gghosh/nire-bcc
maps -d &
pollmach -f runstruct_vasp "mpirun -np 96" >& pollmach.log

Please, let me know.
Thanks,
Gautam

17

Hi Axel,
I just got a new linux laptop! Also, the cluster at PSC has changed (Blacklight has been replaced by Greenfield). So, I unpacked ATAT-3 package. Now, what else do I need to do run cluster expansion? Is it just ezvasp in atat dir? Or, something else? Can you gave me detailed instructions? Im sure, I need to create .ezvasp.rc to define the path of POTCAR file! I do not recall anything else. Nor do I have a written record.

Thanks very much.
Gautam

18

Hi Axel,
I just submitted a job in the new cluster at PSC.
Could I get help in preparing the script file to run MAPS. The script for a VASP job looks like

#!/bin/csh
#PBS -l nodes=1:ppn=32
#PBS -l walltime=4:00:00
#PBS -j oe
#PBS -q batch

module load ifort

setenv TMPDIR $SCRATCH_RAMDISK
cd /crucible/pscstaff/gghosh/tests

cp $HOME/tests/input/* .
mpirun -np $PBS_NP $HOME/VASP/vasp.4.6/vasp
cp * $HOME/tests/output

So, what would the script file to run CE?
Thanks for the help.
Gautam

19

Try:
#!/bin/csh
#PBS -l nodes=1:ppn=32
#PBS -l walltime=4:00:00
#PBS -j oe
#PBS -q batch

module load ifort

setenv TMPDIR $SCRATCH_RAMDISK
cd /crucible/pscstaff/gghosh/tests

cp $HOME/tests/input/* .
maps -d &
pollmach runstruct_vasp mpirun -np $PBS_NP

you need to do this manually: cp * $HOME/tests/output

You need to configure your .ezvasp.rc, e.g. something like:
#!/bin/csh
#enter name of vasp executable here
set VASPCMD="$HOME/VASP/vasp.4.6/vasp"
#enter the directories containing the pseudopotentials here
set POTLDA="/usr/local/lib/pot_lda"
set POTGGA="/usr/local/lib/pot_gga"
set POTPAWLDA="/usr/local/lib/paw_lda"
set POTPAWGGA="/usr/local/lib/paw_gga"
set POTPAWPBE="/usr/local/lib/paw_pbe"

20

Hi Axel,
I should have posted following questions here, at ATAT forum instead of regular e-mail. Its too late, and I am sorry. So, I am posting those questions again. Accordingly, you may reply here so that others have access too.

My computational resources at NU is none/very litlle, so I have no other option but to use external resources to run CE. Specifically, I plan to run CE using one of the NERSC facility. For a single VASP run, the job script reads

gghosh@cori01:~/calphad-2016/hP2-Ru> cat script
#!/bin/bash -l

#SBATCH -p regular
#SBATCH -N 1
#SBATCH -t 4:00:00
#SBATCH -J hP2-Ru

cd $SLURM_SUBMIT_Dir /global/homes/g/gghosh/calphad-2016/hP2-Ru
srun -n 32 /global/homes/g/gghosh/vasp/cori/vasp.5.3/vasp

Now, how do I run CE in this resource? Could you modify the above script? Which part of the modified job script is machine/OS dependent?
The turnaround for jobs with 1 or 2 nodes (32 or 64 cores) is rather short. So, I should be able to run CE jobs rather quickly. That is better than nothing.

I have unpacked your ATAT package atat3_04.tar. Do you have a newer varsion of atat? I have licensed VASP-5.3, and as you know the POTCAR files come in a format different from earlier format. In VASp-5.3 POTCARs come as xx.Z! Will atat extract POTCAR of the element xx by itself? if I specify the element xx in lat.in?

For the CE, I include lat.in, vasp.wrap, error, what else?

Thanks,
Gautam