ATKError: Exceeded maximum number of self-consistent iterations.

[njuxyh]

I want to calculate the transport spetrum of graphene nanoribbon--c-atom--graphene nanoribbon, but the scf does not convergence.
i have searched the forum, and found some hints, but i am not sure where is my problem. so i wish someone could help me. 3x!

 # TwoProbe Calculation
# ----------------------------------------------------------------
........

# sc 481 : q =  252.11354 e  dRho =  1.0295E+02
# sc 482 : q =  254.01931 e  dRho =  9.4704E+02
# sc 483 : q =  250.41971 e  dRho =  1.0058E+03
# sc 484 : q =  252.76199 e  dRho =  6.1260E+02
# sc 485 : q =  255.75515 e  dRho =  5.3518E+02
# sc 486 : q =  251.57264 e  dRho =  2.4662E+02
# sc 487 : q =  251.12636 e  dRho =  4.2365E+02
# sc 488 : q =  252.31944 e  dRho =  4.2701E+02
# sc 489 : q =  251.69348 e  dRho =  9.0235E+01
# sc 490 : q =  252.34374 e  dRho =  6.3624E+01
# sc 491 : q =  253.15964 e  dRho =  2.9300E+02
# sc 492 : q =  250.92071 e  dRho =  2.0976E+02
# sc 493 : q =  255.20466 e  dRho =  3.3770E+02
# sc 494 : q =  252.40583 e  dRho =  3.7540E+02
# sc 495 : q =  251.49326 e  dRho =  6.8605E+02
# sc 496 : q =  249.82777 e  dRho =  6.8713E+02
# sc 497 : q =  252.42359 e  dRho =  2.2527E+02
# sc 498 : q =  249.17996 e  dRho =  3.0114E+02
# sc 499 : q =  250.88531 e  dRho =  2.9827E+02
# sc 500 : q =  251.38814 e  dRho =  1.8827E+02
Traceback (most recent call last):
  File "<string>", line 27, in ?
ATKError: Exceeded maximum number of self-consistent iterations.
Traceback (most recent call last):
  File "<string>", line 27, in ?
ATKError: Exceeded maximum number of self-consistent iterations.
Traceback (most recent call last):
  File "<string>", line 27, in ?Traceback (most recent call last):
  File "<string>", line 27, in ?
ATKError: Exceeded maximum number of self-consistent iterations.
Traceback (most recent call last):
  File "<string>", line 27, in ?
ATKError: Exceeded maximum number of self-consistent iterations.
Traceback (most recent call last):
  File "<string>", line 27, in ?
ATKError: Exceeded maximum number of self-consistent iterations.
Traceback (most recent call last):
  File "<string>", line 27, in ?
ATKError: Exceeded maximum number of self-consistent iterations.
Traceback (most recent call last):
  File "<string>", line 27, in ?
ATKError: Exceeded maximum number of self-consistent iterations.
Traceback (most recent call last):
  File "<string>", line 27, in ?
ATKError: Exceeded maximum number of self-consistent iterations.
                                            

ATKError: Exceeded maximum number of self-consistent iterations.

my input file :
from ATK.TwoProbe import *
# Read the atomic configuration from a VNL file
vnl_file = VNLFile("GNR-three-C-GNR-opt.vnl")
configurations = vnl_file.readAtomicConfigurations()
two_probe_conf = configurations["GNR-three-C-GNR-opt"]
# Setting the electrode parameters
bz_int_param = brillouinZoneIntegrationParameters( (1,1,100) )
electrode_parameters = ElectrodeParameters(
brillouin_zone_integration_parameters= bz_int_param
)
# Creating the TwoProbeMethod
ele_den_para = electronDensityParameters(mesh_cutoff=150*Units.Ry)
twoprobe_method = TwoProbeMethod(
(electrode_parameters,electrode_parameters),
electron_density_parameters=ele_den_para,
basis_set_parameters=basisSetParameters(
type=DoubleZetaDoublePolarized),
iteration_control_parameters = iterationControlParameters(
tolerance = 1e-4,
max_steps=500)
)
# Execute the Self Consistent Field
scf = executeSelfConsistentCalculation(
two_probe_conf,
twoprobe_method,
runtime_parameters=runtimeParameters(verbosity_level=1,
checkpoint_filename='GNR-three-C-GNR-opt.nc'))
 and GNR-three-C-GNR-opt.vnl is in attatachment

[zh]

The atomic configuration of your system seems  very simple, however the bad convergence happened. It may be some of unreasonable values specified for the parameters your input file. For example, zigzag graphene nanoribbon with mono-hydrogen-terminated edge C is favorable to the antiferromagnetic state. You need to perform the spin-polarized calculation for your configuration. Other thing needed to be taken care is the width of graphene nanoribbons.

[Anders Blom]

Your atoms are outside the unit cell, that's never a good idea.

[njuxyh]

Your atoms are outside the unit cell, that's never a good idea.

i have move the atoms in the unit cell. but still it can not convengence.

it seems a serious probblem from the DRHO, it has no trend to reduce at all.

where is?......., i just want to reproduce others work, they mentiond they took a non-spin-polarized calculation. so, i think it is not the spin issues.

[Anders Blom]

Hard to say anything without access to your system. In general, ATK 10.8 offers much better convergence, in particular for spin-polarized systems.

[zh]

where is?......., i just want to reproduce others work, they mentiond they took a non-spin-polarized calculation. so, i think it is not the spin issues.

Please show the reference for the other work mentioned by you. If possible, post here your the script file in order to let me check it. It's well known that in nonmagnetic state of zigzag graphene ribbons there are a sharp peak at the Fermi level in the density of states (DOS).

[njuxyh]

my scprit is *.vnl in attachment, you can access my structure. and
scf script as well as the reference is J. AM. CHEM. SOC. 2010, 132, 11481–11486. in their paper, they said:
We focus only on the charge transport property in this work
since the spin-polarized calculation produces similar results for
the oscillatory conductance.

[zh]

Through reading the reference mentioned by you, an issue in the atomic configuration of your two-probe system is that the surface layers are too small. As clearly claimed by that reference (see below quote), the authors used a wider surface layers.  In your case, only one and two surface layers are used for the left and right side, respectively.

"Enough surface layers are essential in order to properly screen the induced electric field between two electrodes separated by a short distance. Our calculations show that the transport property of the carbon chains does not change appreciably when there are four or more surface layers"

[zh]

I have updated your two-probe system by increasing the surface layers and found that it can be converged well. Please see the attached file.

[njuxyh]

thank you very much .
i will read it