ABINIT Discussion Forums

Hi,

Working with nanotubes I've come across strange problems when trying optimize the interwall distance between the tubes. In my case they are (3,3)'s in a hexagonal arrangement (cross section of the unit cell is a parallelogram, and each corner of the parallelogram contains about a quarter of a nanotube).

1. For some interwall distances, even in 500 datasets the calculation fails to converge while for others it converges nicely within 50.
2. As I decrease the interwall distances over a wide range (2.4 to 3.4 A), the energy just decreases, from -71.12 to below -71.13 while according to the literature it should have a minimum between 3 and 3.2 A.

Can someone offer a suggestion? I'll post my input file below.

Thanks in Advance,

MG

# Number of Data Sets

ndtset 5

# Definition of the unit cell

acell: 6.4 6.4 2.463 angstrom
acell+ 0.1 0.1 0.0 angstrom

rprim 1 0 0

0.5 sqrt(0.75) 0.0

0 0 1

# Definition of the atom types

ntypat 1

znucl 6

ixc 1

# Definition of the atoms

natom 12

typat *1

xangst

1.3082219e+000 -1.5590782e+000 0.0000000e+000
2.0043123e+000 -3.5341431e-001 0.0000000e+000
2.0043123e+000 3.5341435e-001 1.2305009e+000
1.3082219e+000 1.5590782e+000 1.2305009e+000
6.9609033e-001 1.9124925e+000 0.0000000e+000
-6.9609038e-001 1.9124925e+000 0.0000000e+000
-1.3082220e+000 1.5590782e+000 1.2305009e+000
-2.0043123e+000 3.5341431e-001 1.2305009e+000
-2.0043123e+000 -3.5341436e-001 0.0000000e+000
-1.3082219e+000 -1.5590782e+000 0.0000000e+000
-6.9609033e-001 -1.9124925e+000 1.2305009e+000
6.9609037e-001 -1.9124925e+000 1.2305009e+000



# Definition of the planewave basis set

ecut 22

# Definition of the SCF procedure

nstep 500

toldfe 1.0e-12

diemac 12.0

# Usual self-consistent calculation

kptopt 1
shiftk 0 0 0
ngkpt 4 4 4

prtden 1

MGulian wrote:Hi,

Working with nanotubes I've come across strange problems when trying optimize the interwall distance between the tubes. In my case they are (3,3)'s in a hexagonal arrangement (cross section of the unit cell is a parallelogram, and each corner of the parallelogram contains about a quarter of a nanotube).

1. For some interwall distances, even in 500 datasets the calculation fails to converge while for others it converges nicely within 50.
2. As I decrease the interwall distances over a wide range (2.4 to 3.4 A), the energy just decreases, from -71.12 to below -71.13 while according to the literature it should have a minimum between 3 and 3.2 A.

Hello Gulian,

The most important point you did not include in your post: are you using LDA or GGA pseudopotentials (the ixc 1 in the input is correct)? If you check the literature on graphite/CNT you'll see that GGA does not bind graphene planes or CNT. LDA does, at roughly the correct 3.5A, although for reasons that are not completely correct (compensation of "covalent-like" overbinding for the VdW binding which is completely absent).

a few points:
1) 22Ha is a bit low for carbon - depends on your pseudopotential

2) (3,3) tubes are very small and very curved - they will be quite reactive a priori, and these things are only synthesized in zeolites, they are never freestanding or in bundles. This may explain why they bind instead of sitting at a nice 3 Angstr

3) I don't know what you mean by "in 500 datasets". You mean SCF steps, or geometry relaxation steps? If it's scf, you might use iprcel 45

4) the kpoints along the tube axis are probably insufficient

5) check that "angstrom" is correctly interpreted by abinit (and that acell in bohr are correct in your input). I always use "Angstr", but I know some combinations are not read correctly.

Matthieu

Can someone offer a suggestion? I'll post my input file below.

Thanks in Advance,

MG

# Number of Data Sets

ndtset 5

# Definition of the unit cell

acell: 6.4 6.4 6.4 angstrom
acell+ 0.1 0.1 0.1 angstrom

rprim 1 0 0

0.5 sqrt(0.75) 0.0

0 0 1

# Definition of the atom types

ntypat 1

znucl 6

ixc 1

# Definition of the atoms

natom 12

typat *1

xangst

1.3082219e+000 -1.5590782e+000 0.0000000e+000
2.0043123e+000 -3.5341431e-001 0.0000000e+000
2.0043123e+000 3.5341435e-001 1.2305009e+000
1.3082219e+000 1.5590782e+000 1.2305009e+000
6.9609033e-001 1.9124925e+000 0.0000000e+000
-6.9609038e-001 1.9124925e+000 0.0000000e+000
-1.3082220e+000 1.5590782e+000 1.2305009e+000
-2.0043123e+000 3.5341431e-001 1.2305009e+000
-2.0043123e+000 -3.5341436e-001 0.0000000e+000
-1.3082219e+000 -1.5590782e+000 0.0000000e+000
-6.9609033e-001 -1.9124925e+000 1.2305009e+000
6.9609037e-001 -1.9124925e+000 1.2305009e+000



# Definition of the planewave basis set

ecut 22

# Definition of the SCF procedure

nstep 500

toldfe 1.0e-12

diemac 12.0

# Usual self-consistent calculation

kptopt 1
shiftk 0 0 0
ngkpt 4 4 4

prtden 1

Dear Matthieu,

Thanks for your reply. I apologize it took me so long to acknowledge it, but we've had 1 meter of snow here in D.C (power issues...).

I did mean SFC cycles, not datasets, and I am using a norm-conserving PSP of the Troullier-Martins type (.pspnc) I appreciate for your notifications about the cutoff energy and the iprcel parameter, and I hope that they will improve my modeling.

Regarding the K-Point sampling, if I understood correctly, you meant that the k-points are sufficient in the b-x and b-y directions, but not in the b-z direction (so that I should have 4x4x16, say, instead of 4x4x4). Actually, I was wondering whether the k-point grid should be of the form 1x1xN for a nanotube, as other scientists have told me. Is there an essential difference between these two approaches, at least for relaxation?

Thanks for sharing with me your expertise,

Mamikon

If you are simulating a bundle, there will be some band dispersion in the perpendicular direction, so you do need to converge ngkpt in all directions. What is certain is that the tube direction will need a denser sampling.

For the rest (ecut...) you need to do your own convergence, and don't trust published results, even if they claim to use the same pseudopotential (that's personal experience there).

Matthieu

Using "iprcel 45" fixed the SCF convergence issues, and it turns out that the major problem was just a mix-up on my part of PSP file names... With better k-point sampling (and the correct PSP file), as you suggested, it relaxes to the value in the literature.

Thanks for the help and your recent advice,
Mamikon