Validation of TALYS

TALYS has been tested both formally, to check the computational robustness, and by comparison of calculated results with experimental data.

1. Validation with sample cases

The proof of the pudding is in the eating. The proof that TALYS is able to produce physical meaningful results lies in a comparison with experimental data.  Therefore,  we give many different examples of the use of TALYS in which comparisons with experimental data are shown; see sample cases.

2. Validation of level density models:

Generally, there are two different ways to use TALYS:

  • A very detailed calculation with various adjusted parameters and choices for nuclear models, so that specific experimental data are reproduced.
  • Large scale, default calculations for many nuclides, in which case  adjustment to experimental data is impossible or, for the moment, inpracticable.

These two approaches are strongly linked. Since nuclear model calculations and fits to experiments generally require many adjustable parameters, it is important that these parameters all remain within physically acceptable boundaries.  Although this can never be 100% guaranteed, reliability of the code can be ensured to a large extent if we can demonstrate that so-called default,  or ``blind'' calculations, produce results that are not too far from the truth  (where ``truth'' is supposed to be represented by high-quality experiments).  Starting from these default parameters, usually only one or a few parameters need to be adjusted to obtain the best possible result, after which we can be pretty confident that we did not end up in some strange corner of parameter space. To test the stability of TALYS, we have performed neutron-induced reaction calculations up to 40 MeV for 8 different level density models, and each level density model with a set of local parameters (i.e. level density parameters adjusted to discrete levels and mean resonance spacings per nucleus) and global parameters. All other parameters (optical model, gamma strength functions, pre-equilibrium etc.) were kept at their global values. For all stable nuclides, we have collected the (n,gamma), (n,2n), (n,p), (n,d), and (n,alpha) cross sections from the TALYS output and compared them automatically with whatever the EXFOR database was prepared to give us. It should be understood that for very light and fissile targets, the results can be either unphysical or/and wildly fluctuating.

The results consist of 5 sets of figures with local level density parameters and 5 sets for global level density parameters.

Download level density plots here

3. Statistical analysis of cross sections (SACS, JUKO Research)

Another "horizontal" validation of the TALYS results is provided by the so called Statistical Analysis of Cross Sections, performed by Jura Kopecky of JUKO Research. TALYS has been used to generate more than 60,000 cross section excitation functions for the EAF-2005/2007 activation data libraries. An automatic comparison of the calculated results and the EXFOR database has been made. Below is a collection of figures showing the deviation of blind TALYS calculations from the experimental data.
Note: the EAF team uses a different numbering for TALYS versions. Their "TALYS-6" is
equal to the previous beta release TALYS-0.72.

Validation_TALYS6.doc

Next, for all targets with A > 20, and excluding the actinides, some characteristics of the excitation functions have been plotted. These are the maximum cross section, half-width at maximum, and the energy of the maximum

Validation_SACS_TALYS6.doc

4. Robustness test with DRIP:

One way to test a nuclear model code is to let it calculate results for a huge number of nuclides, and the whole range of projectiles and energies. We have written a little code DRIP, which is not released, that launches TALYS for complete calculations for all nuclides, from dripline to dripline. Besides checking whether the code crashes, visual inspection of many curves simultaneously (e.g., 50 (n,2n) excitation curves), may reveal non-smooth behaviour. Various problems in the code and the nuclear structure database have been discovered with this approach.

5. Robustness test with MONKEY:

A TALYS calculationcan be steered with more than 200 keywords in the input. Some of them are flags, others have values which cover a quasi-continuous range between minimally and maximally allowed values. The total number of possible input files for TALYS is huge, but only a small subset corresponds to physical meaningful cases. Nevertheless, to test the computational robustness of the code, we wrote a program MONKEY which remotely simulates TALYS in the hands of an evil user: it creates an input file using all keywords with random values which remain within the corresponding allowed (broad) range. We have tested that TALYS completes 1000 cases without crashing. This may not be a full-proof test on the Fortran level, but MONKEY has helped us to discover several bugs.