Problem 1702

Problem was identified in G4EmStandardPhysicsWVI constructor and
G4GoudsmithSaundersonMscModel model classes, which were newly introduced
in version 10.1 as "experimental" options. Fixes will be included in the
next patch release. We confirmed all "production" pre-packaged physics 
lists, all physics builders and all other physics constructors in the 
release 10.1 are free from this problem.

Other issues mentioned in the original report regarding other EM
options in 10.1 as well as regarding versions earlier than 10.1 could
not be reproduced. In these cases, we confirmed some misuses of option
parameters in user's custom-made physics list could result in zero 
backscattering in many plots of the original report (slides 22-24) and 
in the summary table (slide 25). This seems already to be partly 
addressed by the original report in their own application (slide 27).

We obtained the following results with the extended example TestEM5
that address to Results on slide 36 (Geant4 10.1) of the original
report. These results of simulated backscattering fraction are for
e- of 10 keV scattered off Silicon, with cut value 1 um.

         Opt0   -   14.2 %
         Opt3   -   14.1 %
         Opt4   -   14.4 %
         WVI    -   18.4 %
         GS     -   14.6 %
         SS     -   17.8 %
         slide36 MC  -    0 (all PhysLists)
         slide36 Data-   ~(19 + 1) %

Here fixes are applied for G4EmStandardPhysicsWVI and
G4GoudsmithSoundersonMscModel classes, while all other physics
lists are out-of-box of release 10.1.

We also obtained the following results this time for gold with the
same 10 keV e- with cut value 1 um.

         Opt0   -   42.8 %
         Opt3   -   42.4 %
         Opt4   -   43.3 %
         WVI    -   52.5 %
         GS     -   42.5 %
         SS     -   52.9 %
    slide36 MC  -    0 (all PhysLists)
    slide36 Data-   ~(48 +- 2) %

Here again, fixes are applied for G4EmStandardPhysicsWVI and
G4GoudsmithSoundersonMscModel classes, while all other physics
lists are out-of-box of release 10.1.

Patch 01 for Geant4 10.1 was recently released (see release
notes http://cern.ch/geant4/support/Patch4.10.1-1.txt). This patch includes fixes in G4EmStandardPhysicsWVI and G4GoudsmithSaundersonMscModel classes addressing backscattering issues.
 
The authors of the publication which originated this bug report (S. H. Kim et al., Validation Test of Geant4 Simulation of Electron Backscattering, IEEE Trans. Nucl. Sci. 62, no. 2, DOI: 10.1109/TNS.2015.2401055, 2015) objected to the use of
the word 'misuse' relating to their code, interpreting this term as indicating a user error.
In our intention, this term indicates merely the fact that, as it has been recognised both in the presentation and in the paper by the authors themselves, some of the problems found came probably from an incorrect use of a class constructor. At the present moment, we cannot determine whether this incorrect usage (that we called 'misuse') derives from user error or poor documentation.
Further investigation aimed at clarifying this point will be possible only when this bug report can be fully verified, by including the complete source code and associated files, to enable us to reproduce the results which have been published.
We very much encourage the authors to attach here a tar-ball of such code at their earliest opportunity.

Message from Maria Grazia Pia

The code suitable to further investigate the behaviour observed in our recent backscattering simulations is available in a public git repository at https://github.com/mariagraziapia/bssim. Documentation is on the way;

We acknowledge the user's code and will investigate what caused the reported problem as soon as we can.

After analysis of the code provided by the authors of the paper, we concluded that this is not the one that has been used to produce the results in the paper. For instance no stepping action nor scoring instructions are present. We reiterate our request to the authors to submit the original code(s) with which they have obtained the results of the paper to enable us to analyse the problems reported.

Message from Maria Grazia Pia

Dear Vladimir Ivanchenko and Federico Carminati,

a draft of the documentation promised in a previous email is publicly available in the code repository https://github.com/mariagraziapia/bssim in the file documentation.docx. Please note that it is still a draft circulating among the co-authors, therefore it is subject to modifications and corrections; nevertheless the substantial information content is not expected to change dramatically. 

As you can see, the documentation illustrates in detail various issues worth to be investigated in depth and how to use the code to reproduce them. All the test cases (approximately 30) discussed in the documentation have been executed with the code supplied in the git repository. The figures illustrating the outcome of the various simulations are also available in electronic format in the git repository to facilitate their appraisal. Do not hesitate to contact me if you need further clarifications to reproduce the test cases illustrated in the documentation and to understand their methodological grounds.

Once you have reproduced all the test cases illustrated in the documentation, have understood the underlying causes of the observed behaviour (which I do not know myself at all) and have devised a way to address them, I would be delighted if you would be so kind to inform us of the outcome of your investigation.  I guess that some issues may not be trivial to dig into, therefore I am ready to exercise all the necessary patience. 

Best regards,
Maria Grazia Pia

We acknowledge receipt of the new extended documentation and macros. We are still missing the scoring code that has produced the plots in the publication. Without this part no investigation is possible. We urge the authors of the paper to provide us with this essential part of their code to allow us starting the analysis of the problem.

On May 6the the Maria Grazia Pia wrote

I placed the code concerning the definition and scoring of hits corresponding to detecting electrons in the "Detector" in the GitHub repository https://github.com/mariagraziapia/bssim.

The bssim application was used to study backscattering. The code and default macro were not changed except one point: few lines were added to save backscattering coefficient and primary energy to a separate file for further analysis.

Geant4 10.1 and 10.1p01 were studied. The bug in the G4Sphere geometry class was identified which happens if G4Sphere is used with a section in theta and with the inner radius zero. The bug is active only if the  initial beam position is in the sphere origin (0,0,0). If beam position is shifted for a tiny value then the bug has no effect.

The bug has been fixed and will be available in the next public release and in patches to previous releases. 

The detailed report is available in the talk: https://indico.cern.ch/event/395144/session/1/contribution/4/material/slides/0.pdf