QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: yangzw1985 on March 7, 2009, 13:37

Title: transmission spectrum
Post by: yangzw1985 on March 7, 2009, 13:37
Hi, Everyone! please have a look at the attached calculated results, In my opinion, it is not right. can anyone help check out why all of the results are almost zero? what's the matter?
Thank you!!
Title: Re: transmission spectrum
Post by: Nordland on March 7, 2009, 13:52
The problem is that you have lost your charge, as it is seen from the final step.
Quote
#-------------------------------------------------------------------------------
# Total Charge =   -0.01290
#-------------------------------------------------------------------------------
# sc 14 : q =   -0.01290 e  Etot = 216651.44959 Ry  dRho = 6.3174E-008  dEtot = 2.0729E-005 Ry

The are a couple of ways to resolve this issue, desribed for example in these posts:
--> Please give me advice--Charge missing problem. (http://quantumwise.com/forum/index.php?topic=73.0)
--> Is the current value resonable? (http://quantumwise.com/forum/index.php?topic=112.0)

Edit:
Correct the link for first post.
Title: Re: transmission spectrum
Post by: yangzw1985 on March 7, 2009, 14:04
I have tried with the second a few days ago, but it does't works. the results are the same. I really can not find out the problem. can you check it for me as quickly as possible?
And I also need a confirm, My calculated results is right or not? thanks!
Title: Re: transmission spectrum
Post by: Nordland on March 7, 2009, 14:15
Your results are not correct as the system has converged to an unphysical state with no electrons - an artificial saddle point (http://en.wikipedia.org/wiki/Saddle_point) of Non-Equilibrium Green's Function theory (http://cobweb.ecn.purdue.edu/~datta/negf1.htm).

You will to find out why your system is converging to this saddle point.
Title: Re: transmission spectrum
Post by: yangzw1985 on March 7, 2009, 14:35
what should we do if we want to solve the problem, reset the paremeters,or ?
but we also can not know our recalculated results is right or not?
So, can you check out the problem for me ? Thank you !
Title: Re: transmission spectrum
Post by: Nordland on March 7, 2009, 14:46
Yes I would try to use different parameters:

Your parameters for the twocenter integral parameters are very poor:
Quote
# -----------------------------------------------------------------------------
# Two Center Integral Parameters
# -----------------------------------------------------------------------------
Cutoff           = 100.0 Rydberg
Number Of Points = 100

Use the default instead, as they are very good and don't take alot extra time in the calculation:
Quote
# -----------------------------------------------------------------------------
# Two Center Integral Parameters
# -----------------------------------------------------------------------------
Cutoff           = 2500.0 Rydberg
Number Of Points = 1024

Also you are using a bias of 2 Volt:
Quote
# -----------------------------------------------------------------------------
# Electrode Voltages
# -----------------------------------------------------------------------------
Voltage at Electrode 0 = 1.00 V
Voltage at Electrode 1 = -1.00 V
If you are having this problem, start out at zero bias and calculate the transmission spectrum there.
Then in small steps you can increase the bias.
Quote
# -----------------------------------------------------------------------------
# Electrode Voltages
# -----------------------------------------------------------------------------
Voltage at Electrode 0 = 0.00 V
Voltage at Electrode 1 = 0.00 V

Your electrode contraint is the best in ATK 2008.02:
Quote
# -----------------------------------------------------------------------------
# TwoProbe Algorithm Parameters
# -----------------------------------------------------------------------------
Electrode Constraint = ElectrodeConstraints.Off
Initial Density Type = InitialDensityType.EquivalentBulk

However if you have ATK 2008.10 the new constraint DensityMatrix is superior to handle these things:
Quote
# -----------------------------------------------------------------------------
# TwoProbe Algorithm Parameters
# -----------------------------------------------------------------------------
Electrode Constraint = ElectrodeConstraints.DensityMatrix
Initial Density Type = InitialDensityType.EquivalentBulk

This should improve the calculation alot, however if the problem persit try the following:
--> The mesh cutoff is 100*Ry, try using 200*Ry instead
--> Use more k-points in the kx,ky plane as it tends to help a lot on the convergence of these system.
--> Carefully inspect your system to see if the geometry is correct, as it might also be a result of atoms being placed ontop of one another.
Title: Re: transmission spectrum
Post by: yangzw1985 on March 7, 2009, 14:54

Thank you very much. but if we use more accuracy settings, there is always something happens incidently, and the calculation failed again and again.
we wii try you suggestions.
and if you solved the problem, please tell me as soon as possible.
Thanks in advance!
Title: Re: transmission spectrum
Post by: Nordland on March 7, 2009, 14:58
If you are interessed, and you are not interested in sharing the device publicly, you can send it to me personally. Then I can try to see
what I can do to make this baby converge :)
Title: Re: transmission spectrum
Post by: yangzw1985 on March 7, 2009, 15:13
you can down load it in this topic "the transimission spectrum in a two probe system ",
the model is the same in the topic.
perhaps, you have already download it before.
 liznoliclusterwur48(12.76).py (17.93 KB - downloaded 8 times.)
thanks.

I have to go to my dorminary. see you!
Title: Re: transmission spectrum
Post by: yangzw1985 on April 8, 2009, 10:06
Hi,Nordland! Have you checked out the problem for me? why the value of results is so small?