Question: Severe blooming event in triplet oxygen

[Zihan Lin]

Hi everyone! I tried to calculate the energy of a triplet oxygen with def2-qzvp basis set using i-FCIQMC method. However, a severe blooming event occured suddenly as the attached figure shows. More than 10^18 psips spawn within 40 iterations. I have changed the timestep to 2*10^-6 but it's useless.

I wonder why it happened and how could I solve it. Thanks a lot for your help!

[william.zeller...@gmail.com]

Hi,

These sort of issues usually originate from bad integrals or a bad compile for the code.

How did you generate the integral dump file and have you run the test suite and ensured everything is passing?

A method to check the integrals would be to inspect the dump file and look for particularly large values of individual integrals.

An unlikely origin could be core electrons, as my anecdotal experience has been that including core electrons in the calculation can result in needing a much lower timestep.

However, I have never seen such an extreme case like this so this seems unlikely.

Some additional information that would be useful to provide would be 1) the lua input script and 2) the "reference" json block from the output (which has lines like "H00" and "F0").