Why Monte Carlo? Re: Calculation of uncertainty in EU determination

[Kerembaev@cs.com]

Eric, 

I had experience with Pu-239 and I haven't done particulate calculations for 
U-234, U-235 and U-238 samples and but I do not see any significant 
difference from the alpha spectroscopy point of view. I had analyzed some 
Chernobyl samples a few years ago.

It is my understanding, that there is a problem in the sample preparation 
error/uncertanty.
Most of the times there are the biggest uncertainty.

Your approach would be correct only if you have 100% reproducibility in the 
sample preparation stage.
However, in the real life there is always uncertainty related to the 
preparation stage.

<<< 
 I do not understand why the determination of uncertainty in the %U-235
 enrichment calculation requires Monte Carlo or other special techniques.
 Why does the fact that U-234, U-235 and U-238 add up to 100% pose a problem?
 It seems that standard propagation of error techniques can be used.
 The following approach is used to calculate enriched uranium (%U-235):
 1.  By alpha spec, the activity concentration (usually in pCi/g) is measured
 for each of the 3 isotopes.  The total propagated uncertainty (TPU) is
 provided with each analytical result.
>>>>>>>
uncertainty in the %U-235
enrichment calculation requires Monte Carlo or other special techniques
Why does the fact that U-234, U-235 and U-238 add up to 100% pose a problem?
>>>

.......By using of all those methods we generally are trying to simulated 
distribution of uncertainties (using simulated model of distribution) 
associated with preparation stage, sampling technique, human error, etc.; 
includes bad hair days too. :-)

In the other words, there is a systematic error/uncertanity associated with 
preparation and measurements and non a systematic associated with changing 
conditions in WHOLE process, which we can control to some degree. It is in 
some way similar case with ALARA.
In this particulate case non systematic error/uncertanity plays significant 
role (deposits significant part in uncertainty as a whole). 
So we need to use something more than "traditional" calculations to determine 
the uncertainty. Something like Monte Carlo simulated numerical method or 
other more simple methods for each particulate measurement method (sampling + 
preparation + measurements + data processing).

If you move from counting plated sources to soil/dust samples reproducibility 
worsen significantly.

Radiochemistry is a little different ball game but the general idea will be 
the same.

Emil
kerembaev@cs.com
************************************************************************
The RADSAFE Frequently Asked Questions list, archives and subscription
information can be accessed at http://www.ehs.uiuc.edu/~rad/radsafe.html