Error about runtime_parameters

[Anders Blom]

Please post the error message

[Tom]

Traceback (most recent call last):
  File "<string>", line 624, in ?
ATKError: inverse(DZMatrix const &) : Could not LU factorize!

[Anders Blom]

It's hard to say what exactly caused this error message, but the system doesn't look proper for a couple of reasons.

• There is too little vacuum in the X direction
• There is too much vacuum in the Y direction
• There are two atoms towards the end of the right electrode coordinate list with deviating coordinates (suddenly x=5, y=5)
• Both electrodes are too short, at least two repetitions in the Z direction will be necessary
• The right "surface region" is also too short, and will not provide enough screening. It should be about as long as the left one, in Z, I think

[Tom]

 Thanks very much for your help .I will have a try.

[wilson]

Hello everyone:
     I am a freshman of atk, I came across the same error when I do some calculation on a metal-molecule-metal system using Nickel as the electrodes. What is strange is, the error only came out when I doing spin-polarized calculation on this system, when I doing spin-unpolarized calculation on the same system, everything is OK. No matter how I change the initial spin of the elements, the same error still came out just at the start of twoprobe calculation. I'm not sure about the reason of the error but it do bothers me a lot, can anyone help me?  My input files is enclosed. Thanks a lot!

The error message:
# ----------------------------------------------------------------
# TwoProbe Calculation
# ----------------------------------------------------------------
# sc  0 : q =  706.01149 e
Traceback (most recent call last):
  File "<string>", line 323, in ?
ATKError: inverse(DZMatrix const &) : Could not LU factorize!
Traceback (most recent call last):
  File "<string>", line 323, in ?
ATKError: inverse(DZMatrix const &) : Could not LU factorize!
Traceback (most recent call last):
  File "<string>", line 323, in ?
ATKError: inverse(DZMatrix const &) : Could not LU factorize!
Traceback (most recent call last):
  File "ni11144s1.py", line 323, in ?
    runtime_parameters = runtime_parameters
ATKError: inverse(DZMatrix const &) : Could not LU factorize!

[Anders Blom]

I'd suggest running a bulk calculation for spin-polarized Ni to find out a suitable value for the initial spin.

[wilson]

Thanks for you reply, Anders Bolm. Actually, I had already do a bulk calculation for the spin-polarized Ni (follow the instruction of the atk manual on spin polarized iron), the obtained spin polarization is about 0.13. I have also tried many different initial spin for Ni, but the error still always unavoidable
.

[Anders Blom]

Tricky problem, there is no obvious mistake in the geometry setup or parameters...

Try running it with NeutralAtoms instead of EquivalentBulk as initial density.

[wilson]

Thank you, Anders Blom! Really very tricky problem, I have also trid to running it with NeutralAtoms as initial density or change the type of XC function, but the error will continue to appear all the same…… I guess is it possible that there are some problem with the basis of Nickel atom?

[Nordland]

I have only seen this problem coming from two cases:
1) The alignment of the atoms causes them to be on top of one another.
2) The constraint "DensityMatrix" is used, and the device regions converges to zero, and hence this errors follows.

[Anders Blom]

Wilson, I think you mentioned you can run the calculation without spin (on the exact same system???). If so, we can rule out 1)

For 2), the script actually uses the "Off" constraint.

But I agree, those are the usual suspects...

I think there might be a 3) - if the atoms in the central region are not ordered in Z. But usually only when it applies to the surfaces, not the molecule. But it's worth a try. That is, sort the atoms in the central to be in increasing order in Z. Note that you will have to do this very carefully (it's not easy!) to move each element with the coordinates. Best to extract both to some Excel sheet or so and sort it there.

Also, in general, I fear there are too few screening layers in this system, at least to use DensityMatrix, might be ok for Off.

[wilson]

Hi,Anders, the calculation can indeed complete correctly for exactly the same system without spin. The error will come out when I give an initial spin to Nickel.
I'm sorry but I'm not quite understand "the atoms in the central region are not ordered in Z.", do you mean that I should try to enlarge the distance between atomic layer of the surface region in Z direction? or I can increase the number of screening layers?

[Anders Blom]

No, I just mean sort them in order of increasing Z coordinate. It's a long shot, but since the calculation fails quite quickly, it's worth trying.

[wilson]

Ok, Anders, I will try, Thank you!