Dear FLUKA expert,
I want to know how to apply the variance reduction technique for isotope decay?
I am currently using the BIASING card, and when I use it for the dose measurement of prompt and delayed radiation of 230MeV protons, it works very well and can estimate the dose for every place in the proton treatment center. However, when I use it to calculate isotope dose, it seems to be inadequate. I have tried using the new AUTOIMBS card, and the calculation results have improved to some extent, but they are still far from being reliable.
I am currently using the source_newgen.f composed of many radioactive isotopes, and I need to measure the dose outside the concrete shielding, but it only reached the outside of the lead shielding, and even did not penetrate the lead shielding.
I reviewed the documents and forums of advanced courses, but I don’t know how to apply LAM-BIAS cards to the dose measurement of isotope decay. It is not as easy to understand as the BIASING card. Can you give me some suggestive help?
Looking at your geometry, I suspect that your issue is not about statistics, but about physics. Your source is simply not energetic enough to penetrate the shieldings that you have.
Could you please provide the results you already have for the unbiased vs region biased vs automatically biased?
If you have just one point in my mind, you can bias/adjust the decay direction towards that specific point.
If you do Interaction length biasing on the concrete, then your particles could generate more secondaries (with less individual weights), some possibly exiting the shielding.
Subsequently, I adjusted the distribution of the source term to cover a larger area within the lead shield (although this did not seem to affect the results). These are the results from the run using manual regional biasing with 3e8 primary particles.
The points I am interested in include the maximum dose points outside the lead shield and outside the concrete room, which means that the entire modeling area must be considered.
I also have a few questions regarding automatic biasing:
If my region of interest is the entire modeled area, do I only need to input the largest RPP in the WHAT(3-6) of the AUTOIMBS card? But I have tried this method, and it did not work. I still need to select an area within the RPP to make it work.
From my understanding, the more primary particles used in automatic biasing, the better the results. Therefore, I set 5E7 primary particles in the calculation. Here are some of its result files. 3-1_01005_AUTOIMBS_statistical_checks_1.dat (16.3 KB)
Additionally, I am experimenting with the LAM-BIAS card. As you can see, does this setup provide beneficial results? (If possible, I will set it to air and concrete)
In principle, any RPP body should work, however it should not be too large (more exactly, the number of voxels for the automatic calculation should not be too large). See Note 2.
However, if your RPP is the entire geometry, then this defies the concept of biasing, which is supposed to focus in one particular region/direction. I would suggest to identify one location of interest and bias in that region/direction.
In general, the more primaries, the more precise the results thanks to statistical convergence. For automatic biasing, the importance of each voxel is computed during the run; that is to say, if you don’t have enough primaries, then the voxels’ importance might be undercomputed. In your statistical checks, 3 tests are not passing. I suggest to run more primaries, as per Note 8. Moreover, your automatic biasing seems to work correctly.
In your scenario, the LAM-BIAS cards will have minimal/negligible impact. If you are interested in the residual dose outside the shielding, then this will come mostly from the EMF component (as you suspected), but the LAM-BIAS cards will not help: they impact the nuclear interactions, which are not activated for photons, electrons and positrons in the absence of the PHOTONUC card, and which anyway have a very small cross section.
My recommendation is to continue with automatic biasing and run more primaries.
