Not enough known energies to fit CE

Alloy Theoretic Automated Toolkit (ATAT)
1

Dear Prof. is it possible to "Not enough known energies to fit CE" on maps.log even after 60 ab-initio calculations?
I am using your example in Al,Ti fcc structure and vasp as ab calculator.
below are what I did:
makelat Al,Ti fcc
cd AlTi_fcc
maps -d &
touch ready
…

touch ready

cd 0
runstruct_vasp
…
…
and finally
pollmach runstruct_vasp

I will highly appreciate any clue for this.

2

I am not able to get anything even after 100 structures! I am working on the same AlTi example, please help. Below are my input files:

vasp.wrap file
[INCAR]
PREC = high
ISMEAR = -1
SIGMA = 0.1
NSW = 41
IBRION = 2
ISIF = 3
KPPRA = 1000
DOSTATIC

lat.in file
4.157788 0.000000 0.000000
0.000000 4.157788 0.000000
0.000000 0.000000 4.157788
0.000000 0.500000 0.500000
0.500000 0.000000 0.500000
0.500000 0.500000 0.000000
0.000000 0.000000 0.000000 Al,Ti

3

Sorry for the delay!

What do see if you type

ls */energy */error
cat fit.out

(I need this to see what the problem could be.)

4

Thanks Prof. for the reply, later I realized the same.

Now I am on the way to calculate the Phonon spectrums (trying both, spring model and bond stiffness vs bond length) . I am following Rev. Mod. Phys 74, 11 (2002). Will you recommend me any other papers directly related to the code?

Sincerely,
Rajendra

5

Perhaps this:
https://arxiv.org/abs/0906.1608
Enjoy!

6

I’m not sure how you finally addressed the problem, but I’m having the same thing here. I went to 240 structures using the same lines given above, and it still shows "Not enough known energies to fit CE".

> cat */energy */error

-.37306703E+01
-.35999114E+02
-.35826698E+02
-.39523148E+02
...
-.40030153E+02
-.43304649E+02
-.26934668E+02

Each folder has an empty error file as well. There is no fit.out in the root directory.

The way I carried out these calculations was: feeding the code (after interactive testing) below to a job submission script:

maps -d &
pollmach runstruct_vasp

From the output log of the stdout when it runs out of walltime, quite frequently, such error messages show up:

Running in /nfs/02/utl0268/terencelz/local/ATAT/AlTi_fcc/137
Finding best cluster expansion...
done!
sort: write failed: standard output: Broken pipe
sort: write error
Done running in /nfs/02/utl0268/terencelz/local/ATAT/AlTi_fcc/137
Finding best cluster expansion...
done!
Finding best structure...
done!

If I let it run in the interactive "debugging" mode, these errors don’t show up. I assume it’s just an IO error and doesn’t matter? Or it could be the culprit.

7

I tried it on another supercomputer (recompiling and installing atat) and it still went all the way to more than one hundred structures without being enough.

Scratching my head…

8

If each numbered sub directory has file called error in it, that means the code saw an error in the vasp output files and the energy was ignored by maps. If there are no errors in your vasp runs then delete the "error" files and maps should work normally. Btw, iwill soon post an update for the run struct_vasp command that prevents this problem in the first place.

9

Thank you very much for the help. maps is finally moving. I saw the empty error file was mentioned in the manual, but I was not able to make the connection back then.

Maybe instead of touching an empty error file, write some words in it… So that one will not interpret it as no error when seeing one.

I’m not sure if this is the reason that triggered it, but I enabled KPAR > 1 parallelism and VASP always gives me warnings about this.

checkerr_vasp has this cryptic "errmsg" function I have no idea what is…

10

I’ve fix the problem in the " Beta version of the latest update " on ATAT web page. Version 3.14 and after should not have this problem (which was due, as you noticed, to VASP now displaying a warning whenever the NPAR parameter may not be optimal…)

11

As a follow up question. With more than 200 structures and their energy files, maps is working. It keeps spitting out cluster expansions:

1 1 0 0.0356947
1 1 1 0 0.0353892
1 1 1 1 0 0.0350984
1 2 0 0 0 0.0339581
1 2 1 0 0 0.0335953
1 2 1 1 0 0.033729
...
1 88 23 35 0 0.0312753
1 88 23 85 0 3.40282e+38
1 88 23 106 0 3.40282e+38
1 88 43 0 0 0.0294409
1 88 43 1 0 0.027938
1 88 43 3 0 0.028847

However, it never ends and maps.log still shows "Not enough known energies to fit CE."

12

It will update the log once the search loop is complete. For 200 structure it may take some time… Sorry. If search seems to restart multiple times, perhaps try to limit the composition range where ground states have to be exact with -c0 and -c1 option.