Automatic centrality class generation

[NGoetz]

Future versions should have the ability to run the initial condition many times, in order to select a range from x% to y% of the most energy rich initial conditions as an approximation of centrality classes. The rest of the workflow would only be initialized with the chosen initial conditions.

[AxelKrypton]

Just to clarify a bit, is this a way to specify a parameter scan? If I understand it correctly, at the moment, you need to specify the list of energies by hand in the IC section, giving a list of values {Scan: {Values: [e_1, e_2, ...]}}, where e_1, e_2, ... are the desired energies. Or do you mean something else?

[NGoetz]

No this is orthogonal. In experiment, they don't know the distance the nuclei had. This is estimated by multiplicity. We can use a shortcut. Imagine we want 5% of thr events wirh most multiplicity. We can run IC minimum bias and selecr the 5% with the most energy.

[AxelKrypton]

I am not sure to follow you here. 😅 What would you need to achieve the wished IC initialisation in terms of parameters? Or is this more involved than merely setting correctly values in the input file? 🤔

[NGoetz]

This is considerably more involved.

1. We need a way to specify in the configs that I want to draw x initial conditions from the centrality class a%-b%, which we generate using y raw ICs (for example, if we want 5 events from 0-5%, x is 5 and y has to be at least 100)
2. We have to enforce that the IC config is minimum bias, that means, does not specify the impact parameter
3. Then we have to run the IC with the same config y times
4. We have to use a python script to calculate the energy of each of the IC final states, and remove the ones which have too much or too less energy for the centrality class
5. Then, we execute the further stages on the remaining ICs

[AxelKrypton]

Now I follow you, thanks. Your list of steps is basically a perfect reference to then work on this in the future. A couple of remarks:

• I'd not fix y as number in advance, but I'd try to implement the logic as: "After every IC run I check if the energy is a good one and increase x by 1. Repeat till desired value of x is reached (possibly with a cutoff on y)."

• One should think how to make the handler handle this. For the scan we decided to come up with a two-step process: I first prepare the scan and then run everything. Here it might make sense to also split the task as it follows:

  1. Run IC till x good IC are created and then
  2. prepare all hybrid handler input files for the following phases for each IC.

  This might also be done in a way such that the step ii. is taken out immediately once the good IC is produced (better in case of later IC failures).

  We then need to decide what to do with the bad IC. Would it be sensible to simple throw them away or should these be stored somewhere? Maybe leave the choice to the user?