Getting NaN in MSDCorrelator plugin

[edwinb-ai]

Hello @toastedcrumpets ,

Once again, thank you for the great software.

I was using the MSDCorrelator in a monodisperse system interacting via the square shoulder potential, and wanted to compute the MSD for short times, however, I am getting NaN constantly, and I am not sure how to debug it. Any help would be appreciated. I can provide configuration files or other output that might be useful in debugging this problem.

Thank you in advance.

[toastedcrumpets]

Hi,
Please give the dynamod command you used to generate the config, and attach the config itself so I can try running it please!
NaNs are very rare, so I'm interested to see what's causing it....

[edwinb-ai]

Hello, thanks for the quick response.

The command that gives NaN's is the following
dynarun config.final.xml.bz2 -c 10000 -o config.end.xml -L MSDCorrelator:Length=10000 -t 0.001
and I have noticed that if I increase the value of -c then it works.

For instance, the following command does not give NaN's
dynarun config.final.xml.bz2 -c 1000000 -o config.end.xml -L MSDCorrelator:Length=1000 -t 0.01
and I thought it might have to do something the the value of Nticks in the expression of the MSD in the documentation.

I am including a configuration that I am using as a txt file since GitHub does not allow me to upload an XML file.
The system is a square shoulder quasicrystal phase.

(https://github.com/user-attachments/files/16069586/config.final.txt)

I am editing this since I forgot to mention something about the initial configuration. I did not use dynamod for the initial configuration, I used a different code to pack my system to the target packing fraction that I needed, and then copied those coordinates to a XML file that I eventually used as a initial configuration. The system itself that I want to simulate and study takes a long time to equilibrate, so I just attached the last equilibrated configuration that I have.

[toastedcrumpets]

Hi, there's a check config mode for dynamod that will check there's no overlaps, or miss configured capture maps, you might want to check your externally generated configs just to be sure there's no accidental diameters where the should be radii etc.

Regarding your problem, is it only NaNs in the correlator inside the output.xml? If so, this is probably just too short a simulation to let the correlator collect it's first results. I can fix this to give a more sensible result, just wanted to make sure it's not giving Nan's elsewhere.

Sorry I've not been able to run this myself yet as I'm travelling/on holiday at the moment.

[edwinb-ai]

Thank you for the tip on the check config mode, will make sure to double check that on my configurations.

Indeed, it is just NaNs inside the output.xml file, directly from the output of the MSDCorrelator.

No worries, thank you for the quick response.