# Lattice Thermal conductivity calculation error

I have been trying to calculate Lattice Thermal Conductivity on my system using QuantumEspresso. I used phonopy and phono3py.

I was successful to run these codes on Si system for practice. Now when I am trying to solve this on my material it is showing error.

Since I am doing it on my own and new to it, I think I don't know how to change parameters for calculations.

Please let me know which parameters I can change here (eg: conv_thr, mesh size, cutoff energy) and if I can take these parameters in non converged values since we are taking supercell in initial displacement calculation.

If there is any paper or written material on this, I would like to read it.

These are the errors I am getting in the step where I have to get output of the input of displacement positions. I have taken 2 2 2 supercell which is giving me 255 values and 4 4 4 supercell which is giving values as high as 1875. please let me know upto which supercell I can get my further correct calculations.

Following are the errors I am getting in the output file

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Error in routine cdiaghg (211):
S matrix not positive definite
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

stopping ...

job aborted:
[ranks] message

[0] application aborted
aborting MPI_COMM_WORLD (comm=0x44000000), error 1, comm rank 0

[1-3] terminated

---- error analysis -----

[0] on DESKTOP-683NPIM
pw aborted the job. abort code 1

---- error analysis -----

error 2

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Error in routine c_bands (1):
too many bands are not converged
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

stopping ...

job aborted:
[ranks] message

[0] terminated

[1] application aborted
aborting MPI_COMM_WORLD (comm=0x44000000), error 1, comm rank 1

[2] application aborted
aborting MPI_COMM_WORLD (comm=0x44000000), error 1, comm rank 2

[3] application aborted
aborting MPI_COMM_WORLD (comm=0x44000000), error 1, comm rank 3

---- error analysis -----

[1-3] on DESKTOP-683NPIM
pw aborted the job. abort code 1

---- error analysis -----

The command prompt errors-

MPI_ABORT was invoked on rank 3 in communicator MPI_COMM_WORLD
with errorcode 1.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.


Note: I also think this is also because of low RAM. What system requirements should be kept in mind?

For the geometry in header file

&control
calculation='scf'
tprnfor = .true.
pseudo_dir = '.'
outdir='outt'
/
&system
ibrav = 0,
nat = 24,
ntyp = 3,
occupations = 'smearing'
smearing = 'gaussian'
degauss = 0.01
ecutwfc = 100
/
&electrons
diagonalization = 'cg'
conv_thr = 1.0d-10,
mixing_beta = 0.7d0,
/
K_POINTS {automatic}
2 2 2 1 1 1


My geometry optimization converged values for this material was at ecutoff=40, K_points=16*16*16. I also need to confirm if I can take here 2 2 2 1 1 1, at 161616 is the converged value and the computation will increase.

• It may help to see what you initial inputs are, along with the geometry of the system that is failing.
– Tyberius
Nov 23, 2021 at 14:45
• This thread on RG might help: researchgate.net/post/S-matrix-not-positive-definite. The input file you've posted here lacks the atomic types/positions and cell parameters section. If that's not the case then maybe increasing the energy cutoff might help (don't know why that is, saw it in the RG link). Nov 28, 2021 at 10:12
• @HitanshuSachania The atomic types/positions & Cell parameters are in the displacement inputs which are 510 which I got when I did the original scf phonopy calculation again. Now i did the calculations on a different system - Windows with 32, processors. It did not give errors but it is taking a long time to complete just one calculation. It is thus not looking feasible to do all 510 calculations, considering I am not sure about 2 2 2 1 1 1 mesh size. Nov 29, 2021 at 10:16

1. To remove this error Error in routine cdiaghg (211): S matrix not positive definite Use PAW pseudopotential files instead of USPP. Make sure the exchange correlational functional are same for all elements

2. Increase Energy cut off value. (I am still facing problems in completing the calculations, it's taking too long to complete one calculation. Need a solution for this.)

3. The system requirement for the processors I used on a different system which did not give errors was - mpiexec -n -32. If there are some other changes which can optimize this, I want to know about them.

Lastly, i am still not sure if i can take the mesh size 2 2 2 1 1 1 here, but above are the things which might help you solve some errors. I still need to complete these calculations to get Lattice thermal conductivity.

Edit 1 Adding some useful papers for reference (I found)

1. Esfarjani K, Stokes HT. Method to extract anharmonic force constants from first principles calculations. Phys Rev B 2008;77:144112.
2. Parlinski K, Li ZQ, Kawazoe Y. First-principles determination of the soft mode in cubic ZrO2. Phys Rev Lett 1997;78:4063.
3. Togo A, Chaput L, Tanaka I. Distributions of phonon lifetimes in Brillouin zones. Phys Rev B 2015;91:94306. http://dx.doi.org/10.1103/PhysRevB.91.094306.