Created attachment 683 [details] Exit proton energy distributions after a 60cm C target with a 6.5TeV proton beam I have recently come back to using geant4 after a break of some time and I have found some unexpected behaviour. I have simulated a LHC beam (6.5TeV) impacting a graphite target of depth 60cm, and I am looking at the energy distribution of protons at the exit. In the latest version (10.7.0) there is a quite unusual dip in the energy distribution (between 0 to 1 TeV lost). Tracing back geant4 releases the last good version is 10.4.2. I attach a plot showing this. In this region the distribution should be dominated by single diffraction, and the differential cross section has no such energy dip, so I am not quite sure what is going on. It just looks like single diffraction is missing. From the 10.5 release notes there was an update to both the FTF and QGS models, so I imagine something odd happened there. Does anybody have any ideas? 10.4 matches up with other codes, and what we would expect to see in terms of the energy distribution. The 0-1 TeV energy loss region is the bit that is critical to our needs.
Created attachment 685 [details] Proton target diffraction switched on
Created attachment 686 [details] Proton target and projectile diffraction switched on
Yes, you are right: since Geant4 10.5, target and projectile diffraction has been switched off in the case that either the projectile or the target has A > 10. In your case, with proton projectile and Carbon (A=12) target, diffraction is switched off. Please find in attachment two variants to be applied on top of Geant4 10.7 or 10.7.p01 : G4FTFParameters.cc-g4.10.7-ONE : proton target diffraction switched on (while proton projectile diffraction is still off). G4FTFParameters.cc-g4.10.7-TWO : both proton target and projectile diffraction switched on In order to use one of them, you need to rename it as G4FTFParameters, copy it in: geant4/source/processes/hadronic/models/parton_string/diffraction/src/ and rebuild Geant4. I would suggest to try both variants, and then use the one that describes better you experimental data, or, in absence of the latter, the one which is closer to other simulation codes.
Created attachment 689 [details] 10.7.0-p01 with projectile+target diffraction enabled.
Hi, thanks for the fix. I've built g4 with the projectile + target fix applied, and I have attached the same comparison plot again. Here it now seems like too many protons are produced. I will next test a version with just projectile diffraction on. Is there any reason why diffraction was disabled in 10.5 and later?
Hi, diffraction in the case of hadron-hadron collisions is well established experimentally, and the abundance of data allow to proper check/constrain/tune the phenomenological models that try to model diffraction. For hadron-nucleus (and even more for nucleus-nuclues) interactions the situation is different, with very few experimental measurements available and not yet a generally accepted theoretical prescription or even well established models. To my knowledge, the change in Geant4 10.5 of the diffraction treatment was motivated empirically by a better description of NA61/SHINE proton - Carbon collisions. For these reasons, as said already in my previous post, I would suggest you to use of the three options - 1. Geant4 as it is, i.e. without diffraction; 2. Geant4 with proton target diffraction switched on; 3. Geant4 with proton target and projectile diffraction switched on - the one which is the closest to the "expected/desired" behaviour.
Created attachment 711 [details] Each diffraction option for 11.0 beta vs 10.4
Hi, so I finally ran with each option. None are perfect, but as can be seen in the last attachment, with projectile diffraction off (I'm still not certain what this actually means vs target), the curve is more reasonable, if not still possibly slightly high. If the models were tuned to make a better fit for NA64 (which is somewhat lower energy than the collisions here), then I would say any higher energy pomeron type terms in the models used in g4 need some constraints as well. Is there a specific file that would be a good place to start with when looking at the diffraction differential cross sections? Also, if you (quite correctly) say that current experimental data does not give good constraints on proton-nucleus diffraction, why was diffraction disabled entirely rather than using a best available model approximation? Having one of the largest parts of the hadron-hadron cross sections entirely disabled is going to cause some major problems.
Created attachment 712 [details] Scaled proton target diffraction
Hi, please find in attachment a modified version of G4FTFParameters.cc - to be used on top of 10.7 or 10.7.p01 or 10.7.p02 - which has proton target diffraction switched on (while proton projectile diffraction is left off), and with a scale factor which allows to tune, somehow, the proton target diffraction on your data (or preferred simulation output). Based on your latest comment and plots, it seems that the proton target diffraction as present in 10.7 is too high for you, so the scale factor (that you can find in the attached file by looking at the string "***FOR-JAMES***") that I have introduced should be between 0.0 and 1.0 : currently I put it to 0.5, but you can play with it to find the optimal value for your case. This modified version of G4FTFParameters can be applied - exactly as for the files I gave you before - on top of G4 10.7, or 10.7.p01 or 10.7.p02. Some further information. "Proton target diffraction" means that the projectile proton remains intact, while the target nucleus can fragment. "Proton projectile diffraction" means that the projectile proton can fragment, while the target nucleus remains intact. For your observable (proton energy after the target), what matters most should be indeed "proton target diffraction", as you have found. In Geant4, there is no an explicit "proton target diffraction differential cross section", but with the scale factor I have introduced ad-hoc for you, you should be able to get something similar to what you would like to get, i.e. a way to scale down the probability of proton target diffraction. In principle, you are right by saying that it should be better to leave on the best model available of proton diffraction, rather than not having at all this process. However, in practice, due to the very rough level of the hadronic physics simulations, this is not always true, i.e. there are many observables, such as those of NA61/SHINE, which are better described by Geant4 with proton diffraction switched off. In other words, with hadronic physics simulations, it is like to have a short blanket which inevitably leaves some part uncovered ! I would suggest to adopt a more pragmatic approach, i.e. to strive to get a "reasonable" description of the data, even if not completely satisfactory from the "theoretical" point of view...
