TransmissionSpectrum suddenly using too much load

[scicalc]

I have a problem with TransmissionSpectrum of a large bulk configuration. The system is 960 carbon and hydrogen atoms total. I am calculating using the semi-empirical Huckel method. I did not suffer any problems setting up the bulk configuration. However, when running the TransmissionSpectrum calculation, the cpu usage suddenly spikes several minutes after starting. If I watch `top` carefully, I notice that  about 2~5 minutes after the calculation begins, a spike in cpu usage suddenly exceeds the available cores on my server. This eventually crashes the server and it becomes unable to stop the rogue atkpython processes.

Is there any way to get more verbosity in order to debug what exactly is causing this sudden spike? The only information I have is that it is after the TransmissionSpectrum calculation begins. I have reduced the history number of the SCF calculations, but this did not change it. What steps could I take to better understand what is causing this and then correct this problem? Thank you for the help.

Also, the script used to simulate worked fine with smaller systems ~300 atoms with no reduction in SCF history number.

The OS is CentOS and ATKPython is 12.8.2

Also, I had forgotten about the forums, so I have accidentally submitted this problem through the contact page. Sorry about that, but I suppose it belongs in the forums more.

[Nordland]

If it is a bulk configuration you are doing it for, I understand it right away. The algorithm for doing it from a bulk configuration is calculation-wise very though unless the system is small. It performs an order N^3 operation to get the transmissoin, while the device algorithm uses a order N algorithm. So therefore I would convert your bulk system into a device system, and then it should run much faster.

[scicalc]

If it is a bulk configuration you are doing it for, I understand it right away. The algorithm for doing it from a bulk configuration is calculation-wise very though unless the system is small. It performs an order N^3 operation to get the transmissoin, while the device algorithm uses a order N algorithm. So therefore I would convert your bulk system into a device system, and then it should run much faster.

Thank you for the help! I will try to move my systems over to DeviceConfiguration. However, the main motivation for using Bulk was that I could reduce the total number of unitcells in the simulation, which, right now, is only 3. I hope that doesn't slow it down further.

[Nordland]

If your unitcell is long enough, you might be able to get along with an electrode length of 1 and a central region of length 3, and then it will be much faster to do the device simulation than the bulk calculation.