QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: songll0808 on September 16, 2010, 10:45

Title: An old question-Pulay mixing inversion failed. Using only last step
Post by: songll0808 on September 16, 2010, 10:45

Hi everyone, I have set up a two probe system. When I make scf calculation, the calculation would finish and give the message:Pulay mixing inversion failed. Using only last step.

I have check my system, and I can make sure that the structure is all right. I also improve the accuracy of the energy contour integral parameters, but the results is the same.

Thank for any replies.

Title: Re: An old question-Pulay mixing inversion failed. Using only last step
Post by: Nordland on September 16, 2010, 11:13

Have you removed anything in the output?

What are you mixing factor?

Something is strange. It appears to converged in the first step for the two-probe geometry - so perfectly that the pulay mixer warns. I have never seen this before,
 and therefore I suspect that something is wrong.

Title: Re: An old question-Pulay mixing inversion failed. Using only last step
Post by: fanjiaping on September 20, 2010, 15:42

 hi nordland:
       how can you know the system converged from the result?

Title: Re: An old question-Pulay mixing inversion failed. Using only last step
Post by: Anders Blom on September 20, 2010, 16:37

It converged numerically in 4 steps, that's what the log file shows. In step 4 you see

Code

# sc  4 : q =  412.90491 e  Etot = -4308.82930 Ry  dRho =  0.0000E+00  dEtot =  0.0000E+00 Ry

which is below the tolerance for dEtot.

However, having a convergence to exactly dE=0 and dRho=0 is very suspicious and shows something is wrong. Indeed, as has been discussed separately offline, the geometry was not correctly set up. With the correct geometry, all works fine.