ABINIT Discussion Forums

I have problem in calculating the elastic constants of Ni3Al alloys.
I have attached the input file with this mail and after running the programme we get output in which elastic constants are (here the Rigid-atom elastic tensor is for first set of kpoints ngkpt1 2 2 2)
Rigid-atom elastic tensor , in cartesian coordinates,
j1 j2 matrix element
dir pert dir pert real part imaginary part

1 7 1 7 ******************** 0.0000000000
1 7 1 8 -23774766.9581691027 0.0000000000

2 7 1 7 ******************** 0.0000000000
2 7 1 8 ******************** 0.0000000000

3 7 1 7 ******************** 0.0000000000
3 7 1 8 ******************** 0.0000000000

1 8 1 7 6905595.2596016508 0.0000000000
1 8 1 8 ******************** 0.0000000000

2 8 1 7 -3774893.7480101618 0.0000000000
2 8 1 8 6905580.1166111240 0.0000000000

3 8 1 7 -3774905.1635063933 0.0000000000
3 8 1 8 6905581.4872329356 0.0000000000

Where I have mistake, kindly advice

with regards,
Ashish

ngkpt 2 2 2 is completely unconverged. It's normal that the elastic constants are wrong and even exploding. Get a decent grid for the ground state, then worry about converging it more for the elastic constants. For a metal you basically never go below 8x8x8, unless you have massive supercells.

matthieu

Again, I have tried with ecut 20 Ha, ngkpt 8x8x8, nband 20, tsmear .02 Ha ( ni3al.in file enclosed) the elastic constants in Ha/(Bohr)3 is

Rigid-atom elastic tensor , in cartesian coordinates,
j1 j2 matrix element
dir pert dir pert real part imaginary part

1 7 1 7 0.1384157620 0.0000000000
1 7 1 8 0.0000000607 0.0000000000

2 7 1 7 0.0475563520 0.0000000000
2 7 1 8 0.0000000461 0.0000000000

3 7 1 7 0.0475563517 0.0000000000
3 7 1 8 0.0000000461 0.0000000000

1 8 1 7 0.0000000000 0.0000000000
1 8 1 8 0.0480231507 0.0000000000

2 8 1 7 -0.0000000000 0.0000000000
2 8 1 8 -0.0000000000 0.0000000000

3 8 1 7 -0.0000000000 0.0000000000
3 8 1 8 -0.0000000000 0.0000000000

Which is again not close to the expt. ones. again I have change tsmear from .02 Ha to .01 Ha but the results are not much different.
For Higher values of ngkpt, more than 8x8x8 there is a memory problem in the pc.
kindly, advice how can I overcome with my problem.


Ashish

Did you check convergence of the total energy with respect to the cutoff? in my experience you will get good elastic constants but cutoff energy has to be quite good as well as the K-mesh. After defining the energy cutoff, it will be nice if you try to make a convergence of the elastic constants as function of the K mesh. By the way, how good is the cell parameter with respect to experiment?

regards

-aldo.

What are the experimental values you are comparing to? As Aldo says your k-points may still be very under-converged. For some metals you need up to 16x16x16 or 20x20x20 k-points.

matthieu

Sir,
the experimental values for Ni3Al are
c11=230GPa
c12=150GPa
c44=131GPa

I will try as per your suggestion.

regards,
Ashish