EGS parameters.

EGS is dependant on a number of non-physical parameters, and it is important to either prove that these parameters do not have any effect on the results that are predicted by the code, or to determine the correct setting for them. There are five such parameters, namely:

1. The step length - the maximum energy loss in any one ``step'' of the shower. The default step length is determined from several variables (including the geometric constraints, the probability of a discrete interaction, AE, AP among others), and was originally designed to obtain good agreement with data from high energy physics experiments. At these energies, the default gives rise to very small energy losses per step. However, this is not the case at low energies (about 20 MeV or less) where the default gives rise to permissable energy losses of approaching 25%. Correctly setting the step length in EGS is important because too long a step length can introduce inaccuracies as the program is then sampling quantities, such as the stopping power and the multiple scattering angle (both of which are energy dependant), at large energy differences. Conversely, if the maximum step length is too short then multiple scattering starts to become ``switched off''. This lower limit is dependant on the material, and drops with increasing Z. This is more fully discussed by the authors of the EGS4 manual.

2. AE, AP - the parameters that describe the boundary between discrete and continuous energy loss for electrons and gamma rays respectively. The program will discard particles with energies below these cutoffs, and secondary particles (such as knock on electrons, or bremsstrahlung gamma rays) will not be produced if their energy would be below this energy - instead this energy loss is factored into the continuous energy loss.

3. ECUT, PCUT - the energies below which electrons and gamma rays are automatically discarded. These are not quite the same as AE and AP, as these parameters have no bearing on whether the program creates a particle, only on whether it is discarded.

The predicted number of Cerenkov photons produced by an electron can be shown to have only a weak dependance on the value of the step length (changing the maximum energy loss per step from 30% to 3% reduces the predicted Cerenkov yield from a 5 MeV electron by $\sim 2\%$), and can thus be set to conform to other demands. The value of the step length is set in SNOMAN by another EGS routine called FIXTMX - this is not shown in figure 13.1, as its use is optional. If a non zero value of the parameter ESTEPE is set in the EGS4 bank, then this routine will be called, and the maximum energy loss per step will be set equal to ESTEPE. Previous work [6] has shown that in order to simulate the multiple scattering angles properly, a value of approximately 3% should be set. The disadvantage of this procedure is that the code will take fractionally longer to execute as the EGS code is performing of the order of 5 times the number of steps it would do if the default step length were used.

Deciding how to set the other four parameters is more difficult. It is relatively easy to show that the yield of Cerenkov photons is not dependant on these energy cutoffs, provided that they are set below the Cerenkov threshold. Thus they may be set according to other considerations, the first of which is that there can be little point in tracking electrons once they have dropped below the Cerenkov threshold - the electrons will produce no further Cerenkov photons, so to continue following them is a waste of valuable CPU time. However, there is a second, more subtle, effect in play, arising from the simulation (or lack thereof) of straggling by the EGS code.

As an example, assume that the value of AE is set to 100keV. In this case, the creation of knock on electrons with more than this energy will be treated as discrete interactions, and the subsequent $\delta$-rays followed by the code, whilst all knock on electrons that might have been created with energies below this threshold are instead accounted for in the continuous energy loss. This means that when tracking an electron, we find it will always fall into one of two classes - either it will suffer no discrete interactions, and lose all its energy through continuous energy loss (and thus travel the same distance as all other electrons in this class), or it will suffer one or more discrete interactions, in which case it will travel a shorter distance, and it will lose at least 100 keV in the process. This means that there will be an non physical gap in (for example) the track length distribution between the electrons that suffer no discrete interactions, and those that lose at least 100keV through discrete interactions.

This argument implies that it is better to set the value of AE as low as possible in order to simulate straggling more accurately. The price payed for this is in the speed of the simulation. The solution to these problems is to set AE to a low value (10 keV has been selected), and ECUT to a value just less than the Cerenkov threshold. All electrons created between these limits are placed on the stack, but are then immediately discarded when the code checks their energy against the value of ECUT. This is slightly less efficient than setting a high value of AE, but removes a number of possible artifacts from the simulation.

Other considerations come into play if one is interested in other than Cerenkov production. Two obvious examples are neutron transport and energy deposit. If one is interested in photodisintegration, then EGS4 may as well discard all particles when they reach the appropriate (ECUT) threshold of roughly 2.2 MeV. However, all the remaining energy is dumped in the current region during a discard operation, and so calculating the energy deposition in a low density medium requires a low value of ECUT - otherwise, particles that might penetrate through the medium and discard part of their energy elsewhere will be treated as if they had dumped all their energy in the low density region, leading to an anomalously high energy deposit.

These changes can be achieved via the EGS4 titles file, which sets defaults of ECUT and allows over-rides. One may either change the default, or just tinker with the value for one medium.