This thread is locked.Only browsing is available.
Top Page > Browsing
Convergence of NEGF in heterostructure
Date: 2017/01/04 20:13
Name: S Dechamps   <samuel.dechamps@student.uclouvain.be>

Dear all,

I'm currently trying to converge a system made of two semi-infinite sheets of graphene, coupled by van der waals interaction (VdW-D3) :

left lead central device right lead
----------|---------------- |
| ----------------|-------------

Nevertheless, the NEGF computations is not converging, even after 3000 iterations; with normRD remaining larger than 0.09, with cyclic variations between 0.09 and 10 (so no monotonically decreasing behavior). Regarding your previous advice for my charge sloshing problem (http://www.openmx-square.org/forum/patio.cgi?mode=view&no=2035), I already put in this case :
scf.Mixing.History 40
scf.Mixing.StartPulay 90
as you can see, with no result for the convergence

How could I attain a converged NEGF computation ?

Best,
S. Dechamps

The input file is the following :

########################################
## SCF parameters ##
########################################
scf.EigenvalueSolver NEGF
scf.Min.Mixing.Weight 0.001
scf.energycutoff 500
scf.Max.Mixing.Weight 0.2
scf.maxIter 36000
scf.criterion 1e-09
scf.XcType GGA-PBE
scf.Mixing.History 40
scf.Init.Mixing.Weight 0.02
scf.Mixing.Type rmm-diisk
scf.Mixing.StartPulay 90
scf.ElectronicTemperature 300.0
scf.Mixing.EveryPulay 1

scf.SpinPolarization on
scf.Kgrid 36 48 1
scf.restart off

########################################
## MD parameters ##
########################################
MD.Type nomd
MD.TimeStep 0.5
MD.maxIter 1
MD.Opt.criterion 0.0001

########################################
## NEGF parameters ##
########################################
NEGF.tran.SCF.skip off

NEGF.Poisson.Solver FFT

NEGF.tran.Analysis on
NEGF.tran.energyrange -1.0 1.0 1.0e-3
NEGF.tran.energydiv 101
NEGF.tran.Kgrid 480 1
NEGF.tran.Channel off # default on
NEGF.Channel.Nkpoint 1 # default=1
<NEGF.Channel.kpoint
0.0 0.0
NEGF.Channel.kpoint>

NEGF.Channel.Nenergy 1 # default=1
<NEGF.Channel.energy
0.0
NEGF.Channel.energy>

NEGF.Channel.Num 3 # defualt=5(for collinear), 10(for Non-collinear)
NEGF.tran.CurrentDensity on

NEGF.filename.hks.l lead-LL_C.hks
NEGF.filename.hks.r lead-RL_C.hks

NEGF.Num.Poles 100
NEGF.scf.Kgrid 48 1
NEGF.scf.Iter.Band 30

NEGF.bias.voltage 0.0
NEGF.bias.neq.energy.step 0.02
NEGF.bias.neq.im.energy 0.01

########################################
## Structure ##
########################################
Atoms.SpeciesAndCoordinates.Unit Ang
LeftLeadAtoms.Number 8
<LeftLeadAtoms.SpeciesAndCoordinates
...
LeftLeadAtoms.SpeciesAndCoordinates>
RightLeadAtoms.Number 8
<RightLeadAtoms.SpeciesAndCoordinates
...
RightLeadAtoms.SpeciesAndCoordinates>
Atoms.Number 40
<Atoms.SpeciesAndCoordinates
...
Atoms.SpeciesAndCoordinates>

Species.Number 1
<Definition.of.Atomic.Species
C C5.0_GGA-s2p2d1 C5.0_GGA
Definition.of.Atomic.Species>

########################################
## VdW ##
########################################
scf.dftD on # on|off, default=off
version.dftD 3 # 2|3, default=2
DFTD3.damp zero # zero|bj, default=bj
DFTD.IntDirection 1 1 0 # default=1 1 1 (1:on 0:off)
メンテ
Page: [1]

Re: Convergence of NEGF in heterostructure ( No.1 )
Date: 2017/01/04 20:20
Name: S Dechamps  <samuel.dechamps@student.uclouvain.be>

For the sake of clarity, I repost the scheme of the L|C|R system :

| left lead | central device | right lead |
|-----------|----------------| |
| |----------------|------------|

with ---- being an atomically thin layer
メンテ
Re: Convergence of NEGF in heterostructure ( No.2 )
Date: 2017/01/05 18:27
Name: Artem Pulkin

Well, as far as I can understand you miss the purpose of the 'central device'. Your system should rather be

[left]|-----central----- |
| ------device-----|[right]

I.e. the edge of the top graphene layer cannot be right above the right lead. Similarly, the edge of the bottom layer cannot be right below the left lead. Both terminations should be away from the leads as much as it is possible. The FFT solver makes no sense in this system either.
メンテ
Re: Convergence of NEGF in heterostructure ( No.3 )
Date: 2017/01/05 18:31
Name: Artem Pulkin

The pseudo-image was trimmed. The correct one is:

[=left=]|=====central=====-----|---------
--------|-----======device=====|[=right=]

'-' corresponds to vacuum
'=' is atomic monolayer
メンテ
Re: Convergence of NEGF in heterostructure ( No.4 )
Date: 2017/01/05 19:10
Name: S Dechamps  <samuel.dechamps@student.uclouvain.be>

Same for my pseudo-image, my device is such as yours :

[=left=]|=====central=====-----|---------
--------|-----======device=====|[=right=]

'-' corresponds to vacuum
'=' is atomic monolayer

with an overlap region of about 2nm
メンテ
Re: Convergence of NEGF in heterostructure ( No.5 )
Date: 2017/01/05 19:39
Name: Artem Pulkin

Well, than the periodic form of the potential screwed up your calculation. What is the total size of the central region?
メンテ
Re: Convergence of NEGF in heterostructure ( No.6 )
Date: 2017/01/05 20:16
Name: S Dechamps  <samuel.dechamps@student.uclouvain.be>

It is the overlap region, plus a conventional cell of graphene on each interface with the leads, so a bit less than 3nm in total.
メンテ

Page: [1]