Hi forum,
I've been using abinit 6.0.4 and calculating orthorombic perovskite structure (Space group Pbam , #55) relaxzation of PbZrO3. The calculation employed Broyden structural optimization within ntime parameter = 30 . The result this calculation as :
fconv : WARNING -
ntime= 30 was not enough Broyd/MD steps to converge gradients:
max grad (force/stress) = 3.5432E-04 > tolmxf= 5.0000E-05 ha/bohr (free atoms)
I don' known about suitable value of ntime parameter setting calculation.
Please,help me.
Part of input files :
#Structure
acell 11.1191 22.2742 15.5543
optcell 2
ionmov 2
ntime 30
dilatmx 1.05
ecutsm 0.5
xred 0.2960 0.0970 0.0000
0.7040 0.9030 0.0000
0.7960 0.4030 0.0000
0.2040 0.5970 0.0000
0.2780 0.1560 0.5000
0.7220 0.8440 0.5000
0.7780 0.3440 0.5000
0.2220 0.6560 0.5000
0.0360 0.2621 0.2200
0.9640 0.7379 0.2200
0.5360 0.2379 0.7800
0.4640 0.7621 0.7800
0.9640 0.7379 0.7800
0.0360 0.2621 0.7800
0.4640 0.7621 0.2200
0.5360 0.2379 0.2200
0.0000 0.5000 0.2970
0.5000 0.0000 0.7030
0.0000 0.5000 0.7030
0.5000 0.0000 0.2970
0.0000 0.0000 0.2700
0.5000 0.5000 0.7300
0.0000 0.0000 0.7300
0.5000 0.5000 0.2700
0.6991 0.1299 0.0000
0.3009 0.8701 0.0000
0.1991 0.3701 0.0000
0.8009 0.6299 0.0000
0.7066 0.1235 0.5000
0.2934 0.8765 0.5000
0.2066 0.3765 0.5000
0.7934 0.6235 0.5000
0.2417 0.1236 0.2490
0.7583 0.8764 0.2490
0.7417 0.3764 0.7510
0.2583 0.6236 0.7510
0.7583 0.8764 0.7510
0.2417 0.1236 0.7510
0.2583 0.6236 0.2490
0.7417 0.3764 0.2490
rprim 1 0 0
0 1 0
0 0 1
natom 40
ntypat 3
znucl 82 40 8
typat 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2
#K-point
kptopt 1
ngkpt 2 2 2
#SCF
ecut 50
nstep 50
toldfe 1.0d-6
Suitable value of ntime parameter setting calculation
Moderator: bguster
Re: Suitable value of ntime parameter setting calculation
The input variable ntime is just a safe guard to avoid doing useless calculation in case there is no convergency. It just stop the geometry loop after a given number of iterations.
You should give a look at your stopping criterion for geometry optimisation (i.e. max of forces) along geometry iterations to see if it's nicely decreasing or not.
If it's decreasing indeed, just increase ntime to be able to reach geometry convergency. On the contrary, if the max of forces is not decreasing or occilating too much, you should tune your calculations. Issues in geometry calculations usually comes from badly converged electronic states. Indeed forces being the derivatives of energy, they are very sensitive to energy convergency. If you are in that case, change toldfe to make it more strict, like 1e-9, or better use toldff (see the doc).
You should give a look at your stopping criterion for geometry optimisation (i.e. max of forces) along geometry iterations to see if it's nicely decreasing or not.
If it's decreasing indeed, just increase ntime to be able to reach geometry convergency. On the contrary, if the max of forces is not decreasing or occilating too much, you should tune your calculations. Issues in geometry calculations usually comes from badly converged electronic states. Indeed forces being the derivatives of energy, they are very sensitive to energy convergency. If you are in that case, change toldfe to make it more strict, like 1e-9, or better use toldff (see the doc).