| Summary: | Particle disappearance on the boundary | ||
|---|---|---|---|
| Product: | Geant4 | Reporter: | novikova |
| Component: | geometry | Assignee: | Gabriele Cosmo <Gabriele.Cosmo> |
| Status: | CLOSED FIXED | ||
| Severity: | minor | ||
| Priority: | P2 | ||
| Version: | 4.1 | ||
| Hardware: | PC | ||
| OS: | Windows | ||
The problem is currently under investigation. The problem is now fixed in internal tag "procman-V05-00-01" and the fix is included in public release 5.1. Thanks again for the detailed report. |
I have a case of disappearing of an energetic muon at the boundary of a volume: Step# X(mm) Y(mm) Z(mm) KinE(MeV) dE(MeV) StepLeng TrackLeng NextVolume ProcName 0 -151 -32.4 -1e+003 500 0 0 0 worldVolumePhysical initStep 1 -151 -32.4 -85 500 0 915 915 CsIPhysical Transportation 2 -151 -32.6 -55.3 484 15.4 29.7 945 CsIPhysical MuIoni :----- List of 2ndaries - #SpawnInStep= 1(Rest= 0,Along= 0,Post= 1), #SpawnTotal= 1 --------------- : -151 -32.6 -55.3 0.9 e- :----------------------------------------------------------------- EndOf2ndaries Info --------------- 3 -152 -30.3 85 412 71.3 140 1.09e+003 worldVolumePhysical Transportation 4 -152 -30.3 85 412 0 0 1.09e+003 OutOfWorld MuIoni My word volume is 5m x 5m x 5m and is filled with vacuum (H with the lowest possible density). In the middle of it, a rectangular box of CsI (35cm x 35cm x 17cm) is located. The z-face (z=-85mm) of this chunk is illuminated with a beam of muons with uniform density. Muons are shot in the direction parallel to z- axis, towards increasing z. The propagation of one of this muons is illustrated in the printout shown above. At z = -85 mm the muon passes into CsI chunk. At z = 85 mm (the boundary between CsI and the vacuum) the muon is expected to come out of the CsI into the world. But it doesn't. It experiences an ionization interaction, in which is doesn't loose any energy (see the printout) and doesn't produce any secondaries (see the printout). Instead, it disappears right then and there, together with 412MeV of kinetic energy it possessed. Step #4 was the last step for this particle. The conditions in the "if" statement identify the case when the particle disappears on the boundary. If you have such UserSteppingAction() function, just shoot energetic muons into a chunk of something heavy. I think that sooner or later you will find a disappearing particle. I want to add more to the discussion of this bug. First, I did see this disappearance not only for muons, but for electrons only. I think the bug doesn't depend on the particle type. I am pretty sure that the reason for the bug is somehow related to the fact that when a particle is trying to cross a boundary, for example, exit from a heavy chunk, it sometimes ends up having coordinates EXACTLY on the edge of a heavy chunk, but the pointer pStepData- >GetPostStepPoint()->GetPhysicalVolume() points to the world volume. Thus, the particle is still considered to be INSIDE the chunk by some parts of the code, and outside the chunk by other parts of the code. I observed another phenomenon related (I think) to the same bug. We here (at NRL) are calling it "virtual reflection" of a particle (as opposed to "particle disappearance" bug). The printout (verbose level 2) is shown below, the file with verbose level 5 is attached (output-tracking5-virtualReflection.txt). Step# X(mm) Y(mm) Z(mm) KinE(MeV) dE(MeV) StepLeng TrackLeng NextVolume ProcName 0 -46.3 -174 -1e+003 100 0 0 0 worldVolumePhysical initStep 1 -46.3 -174 -85 100 0 915 915 CsIPhysical Transportation 2 -46.4 -174 -66.6 87.3 12 18.4 933 CsIPhysical MuIoni :----- List of 2ndaries - #SpawnInStep= 1(Rest= 0,Along= 0,Post= 1), #SpawnTotal= 1 --------------- : -46.4 -174 -66.6 0.773 e- :----------------------------------------------------------------- EndOf2ndaries Info --------------- 3 -46.2 -175 -45.7 74.4 12.9 20.9 954 CsIPhysical MuIoni 4 -46 -175 -28.9 62.2 12.2 16.8 971 CsIPhysical MuIoni 5 -46 -175 -25 58.9 2.43 3.9 975 CsIPhysical MuIoni :----- List of 2ndaries - #SpawnInStep= 1(Rest= 0,Along= 0,Post= 1), #SpawnTotal= 2 --------------- : -46 -175 -25 0.824 e- :----------------------------------------------------------------- EndOf2ndaries Info --------------- 6 -45.7 -175 -12.8 49.8 9.1 12.3 987 CsIPhysical MuIoni 7 -44.1 -175 -3.12 41.7 8.15 9.79 997 CsIPhysical MuIoni 8 -43.2 -175 3.68 35.2 6.49 6.88 1e+003 worldVolumePhysical Transportation 9 -43.2 -175 3.68 35.2 0 0 1e+003 CsIPhysical Transportation 10 -42.7 -175 8.34 31 4.19 4.7 1.01e+003 CsIPhysical msc 11 -41.8 -175 12.9 26.2 4.75 4.7 1.01e+003 CsIPhysical msc 12 -41.3 -175 15.2 23.5 2.7 2.35 1.02e+003 worldVolumePhysical Transportation 13 546 -751 2.5e+003 23.5 0 2.62e+003 3.63e+003 OutOfWorld Transportation Here is how it happens. The muon enters the CsI block close to one of the edges and is doing ionization and multiple scattering while propagating very close to one of the volume boundaries (generally going in the increasing Z direction). Then in one (or more) of the interactions, it ends up exactly on the edge of the block, but this time it approaches the boundary practically sliding along it (as oppose to the perpendicular approach in the case of the first bug ("disappearing particle")). The printout above shows what was happening to the particle that entered the CsI block going along z axis. The entrance point happened to be close to the xz plane of the CsI chunk, so that the particle propagates close edge of the chunk which is located at y=-175.00. Now pay attention to the steps 8 and 9 in the printout above. After ionization in step 8, the particle appears to have the coordinates EXACTLY on the edge of the chunk (y=-175). (Btw, I checked in the debugger, the y-coordinate are exactly -175., i.e., (the_current_y - (-175)) = 0 (it is exactly zero, and not 1e-10 or some such number.)) In a way, it still belongs to the chunk, but the pointer to the current volume at step 8 shows "worldVolumePhysical." During the next step, the particle "reenters" the chunk (although it never quite left it), and the pointer of the current volume is back to "CsIPhysical." What PHYSICAL process turned the particle around? If it DID exit and was in the "worldVolumePhysical," how could it go back to "CsIPhysical"? If we consider it as staying inside the chunk and never making it to the outside, then how come the current volume was "worldVolumePhysical" in one of the steps? This phenomenon of "virtual reflection" off the boundary is probably related to the "disappearance" bug. Taken by itself, the "virtual reflection" phenomenon makes following the propagation of single particles more complex that it probably should be. My conclusion, at this time, is that the interpretation of the pointer to the current volume and the coordinates of the particle at any given time should be consistent and treated consistently by different parts of Geant code. Suppose there is a particle that is leaving a volume and ** at this moment ** having coordinates exactly at the boundary of this volume. If the current volume for such a particle is the one it is going to enter (not the one it just left), then all the code has obey by this fact. What happens now is that, although the current volume of a particle "on the boundary" is "outside" volume, the physical processes are still processing this particle as if it is "inside." As a result, we have such things as virtual reflections and disappearance. Sincerely, -- Elena.