Charge transfer in Al1000-C7-Al1000 wire

[jjhskang]

Hi!

I am maiking test runs for the systems alread investigated by others. As one of them, I tried to reproduce the result for Al100 wire-C7 -Al1000 wire in PRB, 65, 165401  (2002). After the run is finished, I find that the total charge of the scattering region is 118.71e, while its neutral correspondent would give 121e. If I interprete it correctly, there is a charge transfer of 2.29e from the scattering region to the eelctrode, which is certainly unphysical and different from the result reported in the paper!!

# ----------------------------------------------------------------
# TwoProbe Calculation
# ----------------------------------------------------------------
# sc  0 : q =  121.06767 e
# sc  1 : q =  178.81648 e  dRho =  1.7683E+00
# sc  2 : q =   15.39491 e  dRho =  3.5080E+00
# sc  3 : q =  101.96733 e  dRho =  3.3972E+00
# sc  4 : q =  110.44791 e  dRho =  4.2365E-01
# sc  5 : q =  121.46240 e  dRho =  3.3486E-01
# sc  6 : q =  122.31578 e  dRho =  1.2091E-01
# sc  7 : q =  123.56391 e  dRho =  9.9452E-01
# sc  8 : q =  121.74473 e  dRho =  1.6857E-01
# sc  9 : q =  120.09753 e  dRho =  2.2583E-01
# sc 10 : q =  117.54981 e  dRho =  3.2823E-01
# sc 11 : q =  118.42281 e  dRho =  1.1353E-01

--..........................................................

# sc 35 : q =  118.70779 e  Etot = -960.12974 Ry  dRho =  3.0487E-05  dEtot = -2.6901E-05 Ry
# sc 36 : q =  118.70790 e  Etot = -960.12976 Ry  dRho =  3.3733E-05  dEtot = -1.5025E-05 Ry
# sc 37 : q =  118.70789 e  Etot = -960.12977 Ry  dRho =  2.6802E-05  dEtot = -1.4921E-05 Ry





Do you have any suggestion for this problem??

For youe convenience, I attach two python scripts. On is for setting up the configuration, wjile the other is for the scf calculation.

[Anders Blom]

An explanation for the unphysical result might be the use of SingleZeta basis set, which normally cannot be expected to give really good results. But then again, the paper uses single zetas too...

Up front, it should be noticed carefully that the results in that paper were not calculated using ATK. So, it can be hard to expect an exact match of the results.

Also, the conclusion in the paper regarding the charge transfer is slightly different. It states there is an excess charge on the carbon atoms, while the q shown in the SCF loop is for the whole scattering region (including the Al layers). So, it is necessary to investigate the individual Mulliken populations, esp. for the carbon atoms, to see where the charges go.

[Anders Blom]

It would also be interesting to compare the transmission spectra.

[jjhskang]

Yes, they also used SinglZeta. I used the exactly the same parameters.

As you pointed out, change transfer is defined for the carbon chain. Howerver. I cannot imagine the situation in which the scattering region is so much devoid of electrons!!

Is'nt there any chance that I've missed some states by not inlcuding them in the real axis integration of the contour integral? Or some other reason?

[Anders Blom]

I doubt the contour is the issue, provided you used default parameters (I saw that you tried to tune them a bit, this might be dangerous).

One problem, which is far from obvious here, is that the "q" reported in the SCF loop is actually not the sum of the Mulliken charges. So, I would not conclude immediately that there is a large electron deficit without first calculating the Mulliken populations.