2 questions on Mpimmaps

Dear Professor Van der Walle,

My name is Eric Seo who use ATAT for multicomponent system. First of all, Thank you for your development of this program. This is interesting program to me.

As I mentioned in subject, I have two questions,

  1. How can I run mpimaps? "mpirun mpimaps" is enough?

  2. Still, for multicomponent system and large system, structure generation takes much time without any output. Can I see some kind of loop output while maps generates best structure?

Thank you.

Best Regards,

You start mpimaps like you would start any mpi code on your machine.
mpirun mpimaps
sounds logical, although sometimes it is
mpirun -np [nb of processors] mpimaps

Another way to speed things up is by removing the -DSLOWENUMALGO option in the makefile and recompiling. Turning on all optimizations (-O3) might help too.

You can also use the -c=… with a number bigger than 3 to force maps to look more at smaller structures.

You could add some trace output in the routine ClusterExpansion::find_best_structure
in mrefine.c++ to see where it gets stuck.

You might want to have a look at sqs2tdb (also in ATAT) as a completely different way to approach the problem.

Thank you for your kind reply.

I’m using mpimmaps by interting it in batch script file. I think it’s same as ‘mpirun -np mpimmaps’.

Also, I’m looking at SQS which is recently updated. It is also interesting to me.

After some tests, I may have more questions to you. :smiley:

And I got another simple question. If ‘error’ is generated in folder with ‘energy’ than that energy value is used for cluster expansion or not?

Thank you

Best Regards,

If there is an "error" file, maps does not read the energy file.
The command clusterexpand does, unless you specify the -e option.

Dear Prof. van de Walle,

Thank you again. :slight_smile:

I got more question after I ran sqs2tdb.

How can I make new lattice with more than ternary system using sqs2tdb?
for example, I want to make FCC_A1 lattice with 4 sub-lattices (1 lattice point and 1 octa, 2 tetrahedral site) but I can’t make it with sqs2tdb because it only generate one lattice with binary system.

Thank you.

Best Regards,

Within the CALPHAD, you don’t necessarily need to make quaternary SQS if you want to get a quaternary phase diagram. The solution model interpolates the ternaries to get quaternary free energies. (You do get better accuracy if you do include quaternary ab initio info.)

Note that you can add your own sqs to the database fairy easily.
It’s true that we haven’t included fcc+interstitial sqs. These can be generated with mcsqs and added to the database.

I should point out that it is probably not a good idea to have a system with both octa and tetra interstitials. These sites are too close to each other, unless the interstitial is a very small atom.

Dear Prof. van de Walle,

Thank you, I will gladly listen your advice about fcc lattice.

I have another question.

Is there any format of sqsgen when we decide the fractions of site for specfic level?

or can I just choose equally divided fractions?

for example,
level=0 a=1 b=1 d=1
level=1 a=0.5,0.5 b=1 d=1
level=1 a=1 b=0.5,0.5 d=1
level=1 a=1 b=1 d=0.5,0.5
level=2 a=0.5,0.5 b=0.5,0.5 d=1
level=2 a=0.5,0.5 b=1 d=0.5,0.5
level=2 a=1 b=0.5,0.5 d=0.5,0.5
level=3 a=0.75,0.25 b=0.5,0.5 d=1

and so on. (part of Perovskite sqsgen.in and additive in bold)

Thank you.

Best Regards,

Equally divided fraction is good. One note: given how the .tdb format works, it is better to have composition 0,0.5,1 on all but one sublattice (the .tdb format does not allow higher order interaction on more than one sublattice).