My Community

I am relaxing nitrogen on graphene. I set the accuracy EDIFFG=-.3 first. the structure gets relaxed but as I increase the accuracy to EDIFFG=-.2 the structure gets distorted in the sense that some middle portion of graphene goes in z direction and nitrogen gets sandwiched. I think there is problem in POSCAR file. I am sending POSCAR file here. Please suggest me what to do?
Ngraphene
1.00000000000000
12.7955000000000005 0.0000000000000000 0.0000000000000000
0.0000000000000000 9.8377599999999994 0.0000000000000000
0.0000000000000000 0.0000000000000000 10.0000000000000000
48 1
cart
0.789370 4.951600 0.000000
3.630600 4.951640 0.000000
5.054580 4.951650 0.000000
7.895470 4.951650 0.000000
9.319370 4.951630 0.000000
12.160690 4.951610 0.000000
1.498050 6.181320 0.000000
2.922080 6.181330 0.000000
5.763060 6.181370 0.000000
7.186980 6.181370 0.000000
10.027900 6.181350 0.000000
11.452020 6.181320 0.000000
0.789360 7.411040 0.000000
3.630590 7.411070 0.000000
5.054580 7.411090 0.000000
7.895470 7.411090 0.000000
9.319360 7.411070 0.000000
12.160680 7.411040 0.000000
1.498060 8.640770 0.000000
2.922080 8.640790 0.000000
5.763070 8.640810 0.000000
7.186980 8.640820 0.000000
10.027910 8.640790 0.000000
11.452020 8.640770 0.000000
0.789370 0.032730 0.000000
3.630590 0.032760 0.000000
5.054580 0.032780 0.000000
7.895470 0.032780 0.000000
9.319370 0.032770 0.000000
12.160690 0.032740 0.000000
1.498060 1.262460 0.000000
2.922090 1.262480 0.000000
5.763070 1.262510 0.000000
7.186990 1.262510 0.000000
10.027910 1.262490 0.000000
11.452020 1.262470 0.000000
0.789370 2.492180 0.000000
3.630600 2.492210 0.000000
5.054590 2.492230 0.000000
7.895470 2.492230 0.000000
9.319370 2.492210 0.000000
12.160690 2.492180 0.000000
1.498050 3.721900 0.000000
2.922080 3.721920 0.000000
5.763070 3.721950 0.000000
7.186980 3.721940 0.000000
10.027900 3.721930 0.000000
11.452020 3.721900 0.000000
6.477300 4.900000 2.000000

EDIFFG has no effect on the behavior of the relaxation, if just gives the convergence breaking condition. Your POSCAR looks perfectly OK, but please have a look at the forces on the atoms evolving during the relaxation run. Maybe POTIM is set too large, such that one of the geometry extrapolations 'overshoots'. If you want the graphene sheet to remain perfectly flat, please consider to use selective dynamics and hold the z - component of the C atoms fixed.