Optimization error!!!

[hieido]

Dear All,
Have a nice day.

I have a question about calculation in Parallel system.

I'm optimizing the structure of Silicon electrode with very simple molecule-dithiolbenzen.
with condition is:
device_configuration = OptimizeGeometry(
        device_configuration,
        max_forces=0.05*eV/Ang,
        max_stress=0.05*eV/Ang**3,
        max_steps=200,
        max_step_length=0.5*Ang,
        trajectory_filename=None,
        optimizer_method=QuasiNewton(),
And all the output file have a same source output.
After optimization was done, the output file includes: log file and NetCDF file. I checked the log file, it showed the
Timing:                          Total     Per Step        %

--------------------------------------------------------------------------------

Density Matrix          :   10587.96 s      38.64 s      19.02% |=============|
Constant Terms          :    1805.71 s      11.73 s       3.24% |=|
Real Space Integral     :    1285.27 s       2.79 s       2.31% |=|
Real Space Basis        :     466.25 s       2.99 s       0.84% ||
Basis Set Generation    :     443.67 s       2.84 s       0.80% ||
Valence Density         :     417.10 s       0.90 s       0.75% ||
Setting Density Matrix  :     165.79 s      41.45 s       0.30% |
Diagonalization         :     162.84 s       5.25 s       0.29% |
Exchange-Correlation    :     138.13 s       0.30 s       0.25% |
Hartree Potential       :      80.69 s       0.18 s       0.14% |
Mixing                  :      36.58 s       0.12 s       0.07% |
Difference Density      :      35.23 s       1.04 s       0.06% |
Core Density            :      12.83 s       0.38 s       0.02% |
Neutral Atom Potential  :       2.63 s       0.88 s       0.00% |
File IO, nlsave         :       0.06 s       0.06 s       0.00% |
Hubbard Term            :       0.00 s       0.00 s       0.00% |
--------------------------------------------------------------------------------
Total                   :   55680.73 s (15h28m00.73s)
I think the calculation was done!
but when I check the netCDF file just have only one the old configuration, the new configuration isn't including.

Could somebody help me! plzz!
 

[Anders Blom]

For some reason you have not saved the optimized geometry to the NetCDF file. The script should contain an "nlsave" statement at the end, to save the optimized version. You have by mistake removed it (or never inserted it, in case you wrote this part by hand).

[hieido]

I will send to you the input Py file can you check for me?

[Anders Blom]

If this script run to the end, it should give 2 device configurations in the file. Check the log file, did it actually converge properly, or are there any error messages?

[hieido]

I attached the log file below!
I checked log file there no error messenger, but i saw that in the log file they mention about:
Calculation Converged in 0 steps...                         
.... So how to fix this error? could you help me!

                 

[zh]

Your log file shows the following information:
1). The self-consistent calculations for the initial configuration (i.e. "Forces step =   0")  are converged within 22 steps;
2). The self-consistent calculations in the 2nd geometry optimization step (i.e., Forces step = 1") are not converged within 99 steps. There is a warning information shown in line  16098, i.e., "Warning: The calculation did not converge to the requested tolerance!" Note that the total energy is given in a positive value.  To solve the self-consistent calculations of electrons during the geometry optimization, you may tune the parameters in "IterationControlParameters()".
3). The geometry of your device configuration has been updated by 153 steps (see, the final "Force step = 152").
4). The problem is that the self-consistent calculations for the sequential configurations (i.e., those in the forces steps from 2 to 152) proceed abnormally, because they are converged within 0 step (i.e., " Calculation Converged in 0 steps" as shown in your log file).

[hieido]

Thanks you so much! ...
but the error still exits...
How can fix this problem?
should I change something on the set up process ...? what is the best of condition to optimized the two probe device?
plzzz! help me!

[Anders Blom]

The configuration is too minimalistic. You need at least 2-3 layers of Si more in the central region. Right now all the Si is constrained in the the optimization because it is part of the electrode extension. Go back to the Builder and use Device Tools>Central Region Size.

[hieido]

Dear Anders Blom.

I tried to increase the layer in the central region like your idea: back to Builder and use device tools. However, my calculation still not work! it's still Calculation Converged in 0 steps
Do you have another idea to solve this error?
Or all I have to do is keep increasing the central region's size until successfully optimated?
and another thing is the log file was very big. it's almost 104Mb compare with the NETCDF file just only 37Mb... is this normally?
And How to reduce the time calcultion? because it's take almost 3 days!

I'm waiting for your reply ...
Plzz!
Thanks in advance.
 

[hieido]

Dear all,
I got the error messenger from the re-calculation.
I checked all the structure but cannot find out where is problem ...
1- in log file: it keeps mentioning about the "  Calculation Converged in 0 steps "
2- this time I checked the error messenger:
" An electrode-like region in the central-region in a two-probe system may not contain
atoms not present in the corresponding electrode. The right electrode has 45 atoms defined,
while the right electrode-like region in the central-region contains 46 atoms. "

Could you please explain about the information in the error find for me?
I attached the py file and the error find.
plzz help me! thanks

[zh]

You should fix the atoms in the electrode region and their corresponding ones in the center  region.

[Anders Blom]

In the latest file you attached, the central region is still too small (too few layers of Si). What you have now is that the "electrode copies", i.e. the parts of the central region which are identical to the electrodes, is all the Si in the central region. So, the only atoms which can move in the optimization is the molecule. But during the optimization, it may happen that the S atoms come too close to the Si, and then this error message you see will be thrown. To fix this, add 2-3 layers of Si as suggested above, and things should work better.