We have to fulfill two contradictory requirements. It is the responsibility of each individual process to produce secondary particles according to its own capabilities. On the other hand, it is only the Geant4 kernel (i.e., tracking) which can ensure an overall coherence of the simulation. The general principles in Geant4 are the following: Each process has its intrinsic limit(s) to produce secondary particles. All particles produced (and accepted) will be tracked up to zero range. Each particle has a suggested cut in range (which is converted to energy for all materials), and defined via a SetCut() method (see ). Points 1 and 2 imply that the cut associated with the particle is a (recommended) production threshold of secondary particles. As already mentioned, each kind of particle has a suggested production threshold. Some of the processes will not use this threshold (e.g., decay), while other processes will use it as a default value for their intrinsic limits (e.g., ionisation and bremsstrahlung). See to see how to set the production threshold. The DoIt methods of each process can produce secondary particles. Two cases can happen: a process sets its intrinsic limit greater than or equal to the recommended production threshold. OK. Nothing has to be done (nothing can be done !). a process sets its intrinsic limit smaller than the production threshold (for instance 0). The list of secondaries is sent to the SteppingManager via a ParticleChange object. Before being recopied to the temporary stack for later tracking, the particles below the production threshold will be kept or deleted according to the safe mechanism explained hereafter. The ParticleDefinition (or ParticleWithCuts) has a boolean data member: ApplyCut. ApplyCut is OFF: do nothing. All the secondaries are stacked (and then tracked later on), regardless of their initial energy. The Geant4 kernel respects the best that the physics can do, but neglects the overall coherence and the efficiency. Energy conservation is respected as far as the processes know how to handle correctly the particles they produced! ApplyCut in ON: the TrackingManager checks the range of each secondary against the production threshold and against the safety. The particle is stacked if range > min(cut,safety). If not, check if the process has nevertheless set the flag ``good for tracking'' and then stack it (see below for the explanation of the GoodForTracking flag). If not, recuperate its kinetic energy in the localEnergyDeposit, and set tkin=0. Then check in the ProcessManager if the vector of ProcessAtRest is not empty. If yes, stack the particle for performing the ``Action At Rest'' later. If not, and only in this case, abandon this secondary. With this sophisticated mechanism we have the global cut that we wanted, but with energy conservation, and we respect boundary constraint (safety) and the wishes of the processes (via ``good for tracking''). A process may have good reasons to produce particles below the recommended threshold: checking the range of the secondary versus geometrical quantities like safety may allow one to realize the possibility that the produced particle, even below threshold, will reach a sensitive part of the detector; another example is the gamma conversion: the positron is always produced, even at zero energy, for further annihilation. These secondary particles are sent to the ``Stepping Manager'' with a flag GoodForTracking to pass the filter explained in the previous section (even when ApplyCut is ON). The cuts in stopping range allow one to say that the energy has been released at the correct space position, limiting the approximation within a given distance. On the contrary, cuts in energy imply accuracies of the energy depositions which depend on the material. In summary, we do not have tracking cuts; we only have production thresholds in range. All particles produced and accepted are tracked up to zero range. It must be clear that the overall coherency that we provide cannot go beyond the capability of processes to produce particles down to the recommended threshold. In other words a process can produce the secondaries down to the recommended threshold, and by interrogating the geometry, or by realizing when mass-to-energy conversion can occur, recognize when particles below the threshold have to be produced. One may need to cut given particle types in given volumes for optimisation reasons. This decision is under user control, and can happen for particles during tracking as well. The user must be able to apply these special cuts only for the desired particles and in the desired volumes, without introducing an overhead for all the rest. The approach is as follows: special user cuts are registered in the UserLimits class (or its descendant), which is associated with the logical volume class. The current default list is: max allowed step size max total track length max total time of flight min kinetic energy min remaining range The user can instantiate a UserLimits object only for the desired logical volumes and do the association. The first item (max step size) is automatically taken into account by the G4 kernel while the others items must be managed by the user, as explained below. Example(see novice/N02): in the Tracker region, in order to force the step size not to exceed 1/10 of the Tracker thickness, it is enough to put the following code in DetectorConstruction::Construct(): G4double maxStep = 0.1*TrackerLength; logicTracker->SetUserLimits(new G4UserLimits(maxStep)); and in PhysicsList, the process G4StepLimiter needs to be attached to each particle's process manager where step limitation in the Tracker region is required: // Step limitation seen as a process G4StepLimiter* stepLimiter = new G4StepLimiter(); pmanager->AddDiscreteProcess(StepLimiter); The G4UserLimits class is in source/global/management. Concerning the others cuts, the user must define dedicaced process(es). He registers this process (or its descendant) only for the desired particles in their process manager. He can apply his cuts in the DoIt of this process, since, via G4Track, he can access the logical volume and UserLimits. An example of such process (called UserSpecialCuts) is provided in the repository, but not inserted in any process manager of any particle. Example: neutrons. One may need to abandon the tracking of neutrons after a given time of flight (or a charged particle in a magnetic field after a given total track length ... etc ...). Example(see novice/N02): in the Tracker region, in order to force the total time of flight of the neutrons not to exceed 10 milliseconds, put the following code in DetectorConstruction::Construct(): G4double maxTime = 10*ms; logicTracker->SetUserLimits(new G4UserLimits(DBL_MAX,DBL_MAX,maxTime)); and put the following code in N02PhysicsList: G4ProcessManager* pmanager = G4Neutron::Neutron->GetProcessManager(); pmanager->AddProcess(new G4UserSpecialCuts(),-1,-1,1); (The default G4UserSpecialCuts class is in source/processes/transportation.)