I am running a simulation on checking the efficiency of a neutron counter designed for GCR induced neutron monitoring, which operates by utilizing the neutron-Boron-10 interaction. I am using DETECT to acquire the number of “counts” recorded by the detector. By comparing the counts for different setup of the detector, I hope to investigate the effect of various parameters on the efficiency of the detector.
This process was going fine until I decided to look at how the back scattering caused by the ground(dirt) affects the counts. However, it takes a long time to get meaningful results because of the scale of my simulation. To my understanding, no biasing is allowed since the DETECT card has to be ran under full analogue simulation in order to get meaningful results. I would like to know if there’s any other scoring method I should be looking at so I could speed things up a bit?
Hello Martin,
another way to assess the efficiency of the Rem-counter is to score the neutron fluence spectrum inside of the detector, and to fold (integral multiplication) it with the cross section for the neutron capture reaction on Boron 10. In this calculation, you may use biassing.
The neutron spectrum is scored with USRTRK, it is normalised to a single neutron. You can find the neutron capture cross section of Boron in one of the libraries ENDF, JEFF or JENDL (available at Nuclear Energy Agency (NEA) - Data Bank ). These data files are not obvious to read and you will need some help or study the manuals carefully. And you have to normalise the result correctly, with the number of Boron-10 nuclei in the gas.
Good luck,
Thomas
Thank you for your help. I’ll play around with it to figure out how to implement it.
Nevertheless, just to be sure, do I need to go through this process tracking to all major contributors (electrons, protons and gamma) to the energy deposition inside the detector? Since all charged particles that enters the detector will deposit energy into detector through ionization and thus contribute to the signal. GCR induced neutrons have span a large range of energy (from thermal up to 400 GeV), so I fear just looking at the neutron-Boron-10 interaction would not be enough.
no worries, scoring the thermal neutron capture reaction on Boron, emitting an alpha, is enough. It’s Q-Value (released energy) is much higher than that from all other contribution, fast neutrons and photons/electrons included. In the electronics of Rem-counter, one sets a lower threshold of deposited energy before registering a count, cutting off all the photon/electron interections.
Thank you for your explanation, I’ll go through the fluence spectrum method and compare the result to the one using DETECT card to see if everything’s ok.
I finished the simulation by employing your suggestion. The result is very promising, I got way better statistics with less time.
Nevertheless, I have a question regarding the USRTRK card. I failed to find information to help understand what is the difference between USRTRK vs. USRBDX. The fluence spectrum I acquired from USRTRK is awfully similar to the one I got from USRBDX a few weeks ago. What is difference between track-length fluence and boundary crossing fluence?
USRTRACK gives a differential fluence averaged over a region volume, such as your detector, and so looks like better fitting your purposes, according to Thomas @totto 's suggestion. USRBDX gives a differential fluence averaged over a region boundary, such as the detector surface. The fact that the respective neutron spectra are quite similar is not surprising, though.
Thanks for your detailed explanation. I understand it now. I feared it might be the average of the fluence caused I was worrying about that it might be wrong to use this to represent what is actually happening in the detector.
But considering how neutrons behave in this specific detector, this is a good approximation I guess.
I finished the simulations with the same geometry and everything except one is using DETECT under fully analogue mode and the other is using USRTRK with importance biasing option. I was hoping I can do a side by side comparison to make sure I set up the simulation correctly. However, the results are off by a lot.
What I did is simply sum up the y-values from the DETECT card for the analogue one, I capped it at 60 MeV since ENDF does not have n(B-10,a)* cross section data above that value. For the biasing one, I first multiply the y-values in each bin by their bin widths, then I multiply them by their corresponding n(B-10,a)* cross sections I interpolated from ENDF’s library and then I multiply it by the total number of B-10 atoms present in the entire region. Is this the right way of doing it?
Also, since I am doing this conversion myself, does it mean that I can simply put VACUUM in the gas region to save some CPU time? Will this affect the neutron fluence scored by USRTRK?
Sorry to bother you with so many questions, thanks again.
Best regards,
Martin
P.S. The attached are my input files for the simulation. NM64.inp (5.7 KB) NM64_Bias.inp (6.5 KB)
Without looking in detail, did you take into account that the neutron spectrum that FLUKA records natively is a differential spectrum, with a bin-content per GeV ? You have to multiply the bin-value by the bin-width to get the “counts per bin”, which you can then fold with the cross section.
Filling the detctor with gas or vacuum should not make a big difference for scoring the neutron-spectrum
I shall recheck my calculations to make sure I di not mess up on Monday. Nevertheless, I did multiply the bin-width then multiply the cross section piece by piece energy-wise. The normalized counts I’m getting is 10 times higher than the one I got from DETECT so it seems the results are not comparable. I wonder if it is because of the biasing I used.
Hello Martin,
searching in my “archives” (> 15 years old), I have found that the response of the n(10B,alpha), n(6Li,alpha) or n, (3He,alpha) reactions was always calculated by folding the neutron fluence in the detector volume with the nuclear reaction cross section. This can be done after the FLUKA run with the scored neutron spectrum, or at run-time with a suitable COMSCW subroutine.
I now wonder if DETECT even captures the nuclear binding energy released in the n(10B, alpha) reaction (the Q-value), or only the energy loss by the neutron.
Can a FLUKA-code expert make a statement on this ?
the use of biasing is perfectly legitimate when calculating the neutron spectrum and has no vocation to produce an unusable spectrum (otherwise it would not be really useful);
the DETECT scoring, itself requiring a fully analogue run with no biasing at all, is indeed meant to capture the energy actually deposited in the detector region, in particular the kinetic energy released by the alpha and 7Li products (deriving from the neutron reaction Q-value), which in principle may also leave the gas and deposit a fraction of it elsewhere;
the procedure described above by @KillMartin for folding the resulting differential neutron fluence (GeV-1 cm-2) with the relevant cross section (cm2) looks correct in order to get the number of counts per primary neutron. In this regard, I assume that you input in the USRTRACK card the detector volume in cm3, and you took the actual bin-width (that is predefined for low energy neutrons) in GeV from the resulting spectrum file.
One may wonder about the statistical quality of the DETECT results, but the factor 10 underestimation is intriguing and shall be clarified (I did not find the time to run your input files).
Thank you for your detailed explanation. Yes, I did put in the volume (which is approximated by flair) of the gas region and did the bin-width multiplication by myself. I think the only thing could contribute to such disagreement is probably I did something wrong in my input file when setting up the simulation. I did use a lot of DISCARD, PART-THR and BIASING to save CPU time.