Author Topic: hard to convergence for the SPIN-POLARIZED transport of GNR-wires-GNR two probe  (Read 4109 times)

0 Members and 1 Guest are viewing this topic.

Offline njuxyh

  • QuantumATK Guru
  • ****
  • Posts: 163
  • Reputation: 0
    • View Profile
hi , i have read a paper appl.phys.lett.96.042104 (2010),
before, i have run non-spin transport calculation, it goes well, but when try to study the spin-polarized transport of GNR-nanowires-GNR two probe system, it has errors:
it is a problem that is disscussed in this forum before. 1)structure geometry problem, but it seems that my system has not,for non-spin transport calculation, it goes well 2)
the initial_spin  problem, i am  not sure my system has or not, but i change the initical spin from 1~0.5, the error still came out as below.so i don not where is my problem.
my input script is in attatchment. in fact my non-spin script is revised by zh last time, i only add the spin related part in the sript.
any suggestion is apppreciated. thanks a lot !!!!!!

# sc 26 : Fermi Energy =   -0.29992 Ry  Etot = -1084.00920 Ry  dRho =  4.6250E-03  dEtot = -3.6937E-04 Ry
# sc 27 : Fermi Energy =   -0.29992 Ry  Etot = -1084.00932 Ry  dRho =  1.2625E-03  dEtot = -1.2192E-04 Ry
# sc 28 : Fermi Energy =   -0.29994 Ry  Etot = -1084.00941 Ry  dRho =  8.9892E-04  dEtot = -9.1762E-05 Ry
# sc 29 : Fermi Energy =   -0.29996 Ry  Etot = -1084.00950 Ry  dRho =  9.6766E-04  dEtot = -8.7967E-05 Ry
# sc 30 : Fermi Energy =   -0.29999 Ry  Etot = -1084.00956 Ry  dRho =  6.1229E-04  dEtot = -6.0620E-05 Ry
# sc 31 : Fermi Energy =   -0.29998 Ry  Etot = -1084.00962 Ry  dRho =  7.9344E-04  dEtot = -6.1285E-05 Ry
# sc 32 : Fermi Energy =   -0.29998 Ry  Etot = -1084.00962 Ry  dRho =  6.7926E-05  dEtot =  1.0508E-06 Ry
# ----------------------------------------------------------------
# TwoProbe Calculation
# ----------------------------------------------------------------
# sc  0 : q =  322.00362 e
# sc  1 : q =  341.96025 e  dRho =  3.5328E+01
Traceback (most recent call last):
  File "<string>", line 365, in ?
ATKError: inverse(DZMatrix const &) : Could not LU factorize!

Offline zh

  • Supreme QuantumATK Wizard
  • *****
  • Posts: 1141
  • Reputation: 24
    • View Profile
You had better check the setup of initial spins for the electrodes and the scattering region:
the spin configuration for carbon atoms in the left electrode should be exactly same with the one for carbon atoms in the left screening layers in scattering region. Also for the right side.

Offline njuxyh

  • QuantumATK Guru
  • ****
  • Posts: 163
  • Reputation: 0
    • View Profile
hi,
thank you zh,
i have check the spin configuration for carbon atoms in the left electrode should be exactly same with the one for carbon atoms in the left screening layers in scattering region. Also for the right side.
yes they are, because i use a small code below to make sure above the situation , but the error still exsits. i do not know why. would you  give me further hints, thanks  a lot !

scattering_region_initial_scaled_spin = [0,]*len(scattering_elements)
 
for index in range(len(scattering_elements)):
        if (scattering_elements[index]==Carbon) :
                if (scattering_coordinates[index][1]> y1):
                      scattering_region_initial_scaled_spin[index]=0.5
                elif  (scattering_coordinates[index][1]< y2):
                       scattering_region_initial_scaled_spin[index]=-0.5
......
electrode_region_initial_scaled_spin = [0,]*len(electrode_elements)
for index in range(len(electrode_elements)):
       if (electrode_elements[index]==Carbon) :
                if (electrode_coordinates[index][1]> y1):
                       electrode_region_initial_scaled_spin[index]=0.5
                elif  (electrode_coordinates[index][1]< y2):
                       electrode_region_initial_scaled_spin[index]=-0.5

Offline zh

  • Supreme QuantumATK Wizard
  • *****
  • Posts: 1141
  • Reputation: 24
    • View Profile
Please try the attached file, which was modified from your file.
The GNR is shifted to the middle of supercell (especially in the y direction). The initial spins for the edge C atoms in the top side are set as 0.5, the edge C atoms in the down side are set as -0.5. The SZP basis set is used.

It seems that only the use of SZP could get a converged calculation.