CORSIKA8 week

Dec 10, 2018, 11:00 AM → Dec 14, 2018, 5:30 PM Europe/Berlin

vidyo extension: 10738245

vidyo-room url

vidyo room: ngc_meeting

 

Monday, December 10

11:00 AM → 12:00 PM Discussion

We discussed two major points today:

1/ (see Issue 81)

The best logic to select the next process, and to determine the next propagation length. Ideally we have to consider two type of discrete processes: the ones acting on grammage (interaction), and the ones acting on distance/time (decay). The cascade logic should handle both separately, and only in a second step select between those two types. This means, the next process is obtained from two flat random numbers: p_i = Lambda_i / Lambda_tot and p_decay = Lambda_decay / (Lambda_decay+Lambda_tot), or in principle the latter may be included in the former already, saving one random number... Only after process selection the exponential random number for the next step length is determined. 

This is equivalent to the logic we had in mind earlier, but should be more performant. I have to  think a bit more over this, and we can continue to discuss tomorrow.

2/ (see Issue 82)

The geometrical tracking in several volumes must be made robust against numerical effects at the volume boundaries. Right now, with no special care, tracking runs into an infinite loop, when a track is calculated to volume boundary, but for numerical reasons in the end does not mathematically leave the volume. We need to keep track of boundary crossings explicitly. This is important also for boundary processes (mostly optical stuff). There are first ideas how this should work, but it is not 100% clear. We probably need to save a reference to the "active" volume each particle is in at any moment. This even saves some time, since the starting volume of a new trajectory is on the stack and not calculated from the geometry tree. In this way we can explicitly set the starting volume correctly when we did perform a propagation to a volume boundary.

However, the main point is that for milestone2 we don't need a general solution here. We only want to track particles in ONE homogeneous volume. 

 

Tuesday, December 11

2:00 AM → 3:00 AM Discussion Tuesday

2:00 AM Discussion

We should handle "interactions" and "decay" separately. The former works in grammage, the latter in time. The ProcessList interface must be extended to reflect that difference, and in the Cascade::Step function both must be evaluated independently. Only the shorter step-length is selected -> min(interactions, decay, geometry). See Issue 53 and related.

The total inverse interaction length should be evaluated twice. Once at the beginning of Cascade::Step in order to get the step-length, and a second time after all Continuous processes are executed just before selecting the actual discrete process. For decay it is equivalent, but the "branching ratios" are used instead.

Tim mentioned it would be excellent if we can later still move to Trajectory<Helix> at least for cross-checks.

Felix agrees to merge sibyll branch into master. I will look into it today.

Also:

- rename MinStepLength function into MaxStepLength

- rename operator+ for ProcessList into something that makes more clear that ordering matter, i.e. operator<<    . See Issue 91.

Tomorrow we should have a consistent AND improved version with all of that.

Next major topics: environment integration, tracking, and nuclei.

Wednesday, December 12

2:00 PM → 3:00 PM Wednesday

2:00 PM Discussion

Tangy made the point that we should finish soon many of the core concepts. Otherwise they might a) never be changed, b) will it make very hard for more complex applications.

I think this applies mostly to FourVector (issue 62), LorentzBoost (issue 59), Geometry+CoordinateSystems (issue 63).

But also the output module interface is a major and complicated issue (issue 34).

All of this is not planned for the milestone 2, but should be attacked soon.

 

Thursday, December 13

1:15 PM → 3:35 PM Thursday

It was very interesting and important to see the review about electromagnetic interactions by Stephan. This will be an great field for further studies with corsika. Stephan is very interested to contribute to corsika in the future with his specific expertise: electromagnetic interactions, later also 3D cascade equations. 

The situation is more complex than expected. Also for e.m. interactions we have to distinguish multiple energy regimes with different approximations and corrections. There are several available implementations, very likely every single one is incomplete in some sense.

 

1:15 PM Discussion

1:35 PM Electromagnetic interactions

Speaker: Stephan Meighen-Berger

EmCa_Shower.pdf