## Step 2: The NEGF calculation

A. Setting up Lead Device Lead

You can set up the regions , , and in the structural configuration shown in Fig. 31 in the following way:

The geometrical structure of the central region is specified by the following keywords 'Atoms.Number' and 'Atoms.SpeciesAndCoordinates':

     Atoms.Number        18
<Atoms.SpeciesAndCoordinates
1  C  3.000  0.000  0.000  2.0 2.0
.....
18  C 28.500  0.000  0.000  2.0 2.0
Atoms.SpeciesAndCoordinates>


The geometrical structure of the left lead region is specified by the following keywords 'LeftLeadAtoms.Number' and 'LeftLeadAtoms.SpeciesAndCoordinates':

     LeftLeadAtoms.Number         3
<LeftLeadAtoms.SpeciesAndCoordinates
1  C -1.500  0.000  0.000  2.0 2.0
2  C  0.000  0.000  0.000  2.0 2.0
3  C  1.500  0.000  0.000  2.0 2.0
LeftLeadAtoms.SpeciesAndCoordinates>


The geometrical structure of the right lead region is specified by the following keywords 'RightLeadAtoms.Number' and 'RightLeadAtoms.SpeciesAndCoordinates'

     RightLeadAtoms.Number        3
<RightLeadAtoms.SpeciesAndCoordinates
1  C 30.000  0.000  0.000  2.0 2.0
2  C 31.500  0.000  0.000  2.0 2.0
3  C 33.000  0.000  0.000  2.0 2.0
RightLeadAtoms.SpeciesAndCoordinates>

This is the case of carbon chain which is demonstrated in the previous subsection. The central region is formed by 18 carbon atoms, and the left and right regions and contains three carbon atoms, respectively, where every bond length is 1.5 Å. Following the geometrical specification of device and leads, OpenMX will construct an extended central region as shown in Fig. 31. The Green function for the extended central region is self-consistently determined in order to take account of relaxation of electronic structure around the interface between the central region and the region . In addition, we impose two conditions so that the central Green function can be calculated in the NEGF method :
1. The localized basis orbitals in the region overlap with those in the regions and , but do not overlap with those in the regions and .

2. The localized basis orbitals in the ( ) region has no overlap with basis orbitals in the cells beyond the nearest neighboring cells ( ) and ( ).

In our implementation the basis functions are strictly localized in real space because of the generation of basis orbitals by a confinement scheme [30,31]. Therefore, once the localized basis orbitals with specific cutoff radii are chosen for each region, the two conditions can be always satisfied by just adjusting the size of the unit cells for and .

Although the specification of unit cells for the regions , , and is not required, it should be noted that some periodicity is implicitly assumed. The construction of infinite leads is made by employing the unit cells used in the band structure calculations by the step 1, and the informations are stored in a file 'NEGF.filename.hks'. Also, due to the structural configuration shown in Fig. 31, the unit vectors on the bc-plane for the left and right leads should be consistent. Thus, the unit vector on the bc-plane for the extended central region is implicitly assumed to be same as that of the leads. Within the structural limitation, you can set up the structural configuration.

The unit in the specification of the geometrical structure can be given by

    Atoms.SpeciesAndCoordinates.Unit   Ang # Ang|AU


In the NEGF calculation, either 'Ang' or 'AU' for 'Atoms.SpeciesAndCoordinates.Unit' is supported, but 'FRAC' is not.

How OpenMX analyzes the geometrical structure can be confirmed by the standard output as shown below:

   <TRAN_Calc_GridBound>

*******************************************************
The extended cell consists of Left0-Center-Right0.
The cells of left and right reads are connected as.
...|Left2|Left1|Left0-Center-Right0|Right1|Right2...

Each atom in the extended cell is assigned as follows:
where '12' and '2' mean that they are in 'Left0', and
'12' has overlap with atoms in the Left1,
and '13' and '3' mean that they are in 'Right0', and
'13' has overlap with atoms in the 'Right1', and also
'1' means atom in the 'Center'.
********************************************************

Atom1  = 12 Atom2  =  2 Atom3  =  1 Atom4  =  1 Atom5  =  1 Atom6  =  1 Atom7  =  1
Atom8  =  1 Atom9  =  1 Atom10 =  1 Atom11 =  1 Atom12 =  1 Atom13 =  1 Atom14 =  1
Atom15 =  1 Atom16 =  1 Atom17 =  1 Atom18 =  1 Atom19 =  1 Atom20 =  1 Atom21 =  3
Atom22 = 13


The atoms in the extended cell consisting of are assigned by the numbers, where '12' and '2' mean that they are in , and '12' has overlap with atoms in , and '13' and '3' mean that they are in , and '13' has overlap with atoms in , and also '1' means atom in . By checking the analysis you may confirm whether the structure is properly constructed or not.

B. Keywords

The NEGF calculation of the step 2 is performed by the keyword 'scf.EigenvalueSolver'

    scf.EigenvalueSolver       NEGF


For the NEGF calculation the following keywords are newly added.

    NEGF.filename.hks.l     lead-chain.hks
NEGF.filename.hks.r     lead-chain.hks

NEGF.Num.Poles             100       # defalut=150
NEGF.scf.Kgrid             1 1       # defalut=1 1

NEGF.bias.voltage          0.0       # default=0.0 (eV)
NEGF.bias.neq.im.energy    0.01      # default=0.01 (eV)
NEGF.bias.neq.energy.step  0.02      # default=0.02 (eV)


An explanation for each keyword is given below.

    NEGF.filename.hks.l     lead-chain.hks
NEGF.filename.hks.r     lead-chain.hks


The files containing information of leads are specified by the above two keywords, where 'NEGF.filename.hks.l' and 'NEGF.filename.hks.r' are for the left and right leads, respectively.

    NEGF.Num.Poles             100       # defalut=150


The equilibrium density matrix is evaluated by a contour integration method [58,59]. The number of poles used in the method is specified by the keyword 'NEGF.Num.Poles'.

    NEGF.scf.Kgrid             1 1       # defalut=1 1


The numbers of k-points to discretize the reciprocal vectors and are specified by the keyword 'NEGF.scf.Kgrid'. Since no periodicity is assumed along the a-axis, you do not need to specify that for the a-axis.

    NEGF.scf.Iter.Band          6        # defalut=6


It would be better to use the conventional diagonalization method for a few SCF steps in the initial SCF iterations by assuming a periodicity along the a-axis as well as b- and c-axes. The procedure is effective to avoid an erratic charge distribution which is a serious problem in the self-consistent NEGF method. The number of first SCF steps for which the conventional diagonalization method is applied is controlled by the keyword 'NEGF.scf.Iter.Band'. Up to and including the SCF steps specified by 'NEGF.scf.Iter.Band', the conventional diagonalization method is used and then onward, the solver is switched from the conventional method to the NEGF method. The default is 6.

    NEGF.bias.voltage          0.0       # default=0.0 (eV)


The source-drain bias voltage applied to the left and right leads is specified by the keyword 'NEGF.bias.voltage' in units of eV, corresponding to Volt. Noting that only the difference between applied bias voltages has physical meaning, you only have to give a single value as the source-drain bias voltage.

    NEGF.bias.neq.im.energy    0.01      # default=0.01 (eV)
NEGF.bias.neq.energy.step  0.02      # default=0.02 (eV)


When a finite source-drain bias voltage is applied, a part of the density matrix is contributed by the non-equilibrium Green function. Since the non-equilibrium Green function is not analytic in general in the complex plane, the contour integration method used for the equilibrium Green function cannot be applied. Thus, in the current implementation the non-equilibrium Green function is evaluated on the real axis with a small imaginary part using a simple rectangular quadrature scheme. Then, the imaginary part is given by the keyword 'NEGF.bias.neq.im.energy' and the step width is given by the keyword 'NEGF.bias.neq.energy.step' in units of eV. In most cases, the default values are sufficient, while the detailed analysis of the convergence property can be found in Ref. . How many energy points on the real axis are used for the evaluation of the non-equilibrium Green function can be confirmed in the standard output and the file 'System.Name.out'. In case of 'NEGF-Chain.dat', if the bias voltage of 0.5 V is applied, you will see in the standard output that the energy points of 120 are allocated for the calculation as follows:
   Intrinsic chemical potential (eV) of the leads
Left lead:  -7.752843837400
Right lead: -7.752843837400
add voltage =  0.0000 (eV) to the  left lead: new ChemP (eV):  -7.7528
add voltage =  0.5000 (eV) to the right lead: new ChemP (eV):  -7.2528

Parameters for the integration of the non-equilibrium part
lower bound:           -8.706843837400 (eV)
upper bound:           -6.298843837400 (eV)
energy step:            0.020000000000 (eV)
number of steps:           120

The total number of energy points where the Green function is evaluated is given by the sum of the number of poles and the number of energy points on the real axis determined by the two keywords 'NEGF.bias.neq.im.energy' and 'NEGF.bias.neq.energy.step', and you should notice that the computational time is proportional to the total number of energy points.

    NEGF.Poisson.Solver       FD     # FD|FFT, default=FD


In the NEGF method, the electrostatic potential is calculated by either a finite difference plus two dimensional FFT (FD)  or three dimensional FFT (FFT) . The choice of the Poisson solver is specified by the keyword 'NEGF.Poisson.Solver'. Both the methods provide similar electrostatic potentials for non-polar systems, while the difference can be large for polar systems. The former is a proper choice in a sense that the eletrostatic potential at the boundaries between the leads and the central region should be the same as that in the calculations of the step 1 for the leads, while the SCF convergence seems to be rather easily obtained by the latter. The default is FD.

C. SCF criterion

In the NEGF method, the SCF criterion given by the keyword 'scf.criterion' is applied to the residual norm between the input and output charge densities 'NormRD', while in the other cases 'dUele' is monitored. See also the keyword 'NEGF.scf.Iter.Band'.

D. Gate bias voltage

In our implementation, the gate voltage is treated by adding an electric potential defined by where is a constant value corresponding to the gate voltage, and is specified by the keyword 'NEGF.gate.voltage' as follows:
      NEGF.gate.voltage   1.0    # default=0.0 (in eV) the center of the region , and the length of the unit vector along a-axis for the region . Due to the form of the equation, the applied gate voltage affects mainly the region in the central region . The electric potential may resemble the potential produced by the image charges .

E. Density of States (DOS)

In the NEGF calculation, the density of states can be calculated by setting the following keywords:

    Dos.fileout                 on              # on|off, default=off
NEGF.Dos.energyrange     -15.0 25.0 5.0e-3  #default=-10.0 10.0 5.0e-3 (eV)
NEGF.Dos.energy.div        200              # default=200
NEGF.Dos.Kgrid             1 1              # default=1 1


When you want to calculate DOS, the keyword 'Dos.fileout' should be set to 'on' as usual. Also, the energy range where DOS is calculated is given by the keyword 'NEGF.Dos.energyrange', where the first and second numbers correspond to the lower and upper bounds, and the third number is an imaginary number used for smearing out DOS. The energy range specified by 'NEGF.Dos.energyrange' is divided by the number specified by the keyword 'NEGF.Dos.energy.div'. The numbers of k-points to discretize the reciprocal vectors and are specified by the keyword 'NEGF.Dos.Kgrid'. The set of numbers given by 'NEGF.Dos.Kgrid' tends to be larger than that by 'NEGF.scf.Kgrid' because of computational efficiency. After the NEGF calculation with these parameters, you will find two files 'System.Name.Dos.val' and 'System.Name.Dos.vec', and can analyze those by the same procedure as usual. Also, it should be noted that the origin of energy is set to the chemical potential of the left lead.

2016-04-03