Total Monte Carlo and related applications of the

Total Monte Carlo and related applications of the TALYS code system Arjan Koning NRG Petten, the Netherlands Technical Meeting on Neutron Cross. Section Covariances September 27 -30 2010, IAEA, Vienna

Contents • Introduction: TALYS code system • Implications and possibilities: - Large scale nuclear data library production (TENDL) - “Total” Monte Carlo uncertainty propagation - Random search for the best data library • Conclusions 2

TALYS code system A loop over nuclear physics, data libraries, processing and applications: • • • Resonance parameters + uncertainties An EXFOR database with more uncertainties than errors The TALYS code The Reference Input Parameter Library (RIPL) Software for remaining reaction types (nubar, fns + unc. ) For many nuclides: A set of adjusted model parameters + uncertainties + “non-physical evaluation actions” All major world libraries The ENDF-6 formatting code TEFAL NJOY, MCNP(X) + other codes A script that drives everything The secret: Insist on absolute reproducibility 3

4

5

Nuclear data scheme + covariances +Uncertainties Resonance Parameters. TARES Determ. code +Covariances Output +Covariances ENDF Gen. purpose file NJOY -K-eff MCNP -Neutron flux -Etc. Experimental data (EXFOR) Output Nucl. model parameters +Covariances TEFAL TALYS +Covariances ENDF/EAF Activ. file PROC. CODE +(Co)variances FISPACT -activation - transmutation +Covariances Other (ORIGEN) +Uncertainties Monte Carlo: 1000 TALYS runs TASMAN 6

Uncertainties for Cu isotopes 7

Application 1: TENDL TALYS Evaluated Nuclear Data Library, www. talys. eu/tendl 2009 • n, p, d, t , h, a and g libraries in ENDF-6 format • 2400 nuclides (all with lifetime > 1 sec. ) up to 200 Me. V • Neutrons: complete covariance data (MF 31 -MF 35) • MCNP-libraries (n, p and d) and multi-group covariances (n only) • Production time: 2 months (40 processors) Strategy: • Always ensure completeness, global improvement in 2010, 2011. . • Extra effort for important nuclides, especially when high precision is required (e. g. actinides): adjusted parameters (data fitting). These input files per nuclide are stored for future use. • All libraries are always reproducible from scratch • The ENDF-6 libraries are created, not manually touched • Zeroing in on the truth for the whole nuclide chart at once 8

TENDL: Complete ENDF-6 data libraries MF 1: description and average fission quantities MF 2: resonance data MF 3: cross sections MF 4: angular distributions MF 5: energy spectra MF 6: double-differential spectra, particle yields and residual products MF 8 -10: isomeric cross sections and ratios MF 12 -15: gamma yields, spectra and angular distributions MF 31: covariances of average fission quantities (TENDL-2010) MF 32: covariances of resonance parameters MF 33: covariances of cross sections MF 34: covariances of angular distributions MF 35: covariances of fission neutron spectra (TENDL-2010) and particle spectra (TENDL-2011) MF 40: covariances of isomeric data (TENDL-2011) 9

IAEA covariance visualisation system (V. Zerkin) 10

11

Application 2: “Total” Monte Carlo • • • Propagating covariance data is an approximation of true uncertainty propagation (especially regarding ENDF-6 format limitations) Covariance data requires extra processing and “satellite software” for application codes Alternative: Create an ENDF-6 file for each random sample and finish the entire physics-to-application loop. (Koning and Rochman, Ann Nuc En 35, 2024 (2008) 12

Nuclear data scheme + covariances +Uncertainties Resonance Parameters. TARES Determ. code +Covariances Output +Covariances ENDF Gen. purpose file NJOY -K-eff MCNP -Neutron flux -Etc. Experimental data (EXFOR) Output Nucl. model parameters +Covariances TEFAL TALYS +Covariances ENDF/EAF Activ. file PROC. CODE +(Co)variances FISPACT -activation - transmutation +Covariances Other (ORIGEN) +Uncertainties Monte Carlo: 1000 TALYS runs TASMAN 13

Nuclear data scheme: Total Monte Carlo +Uncertainties Resonance Parameters. Determ. code TARES ENDF Gen. purpose file Output NJOY -K-eff MCNP -Neutron flux -Etc. Experimental data (EXFOR) Output Nucl. model parameters +Covariances TEFAL ENDF/EAF Activ. file PROC. CODE FISPACT - activation - transmutation +Covariances Other codes TALYS +Uncertainties TASMAN Monte Carlo: 1000 runs of all codes 14

15

16

17

18

19

20

21

Application: criticality benchmarks Total of 60000 random ENDF-6 files Sometimes deviation from Gaussian shape Rochman, Koning, van der Marck Ann Nuc En 36, 810 (2009) Yields uncertainties on benchmarks 22

23

Covariance versus Total Monte Carlo Advantages: - Relatively quick - Use in sensitivity study - Easier release (TENDL) Disadvantages: - Approximative (cross-correlations) - No covariance for gamma production, DDX (MF 36), etc. - Requires special processing - Requires covariance software for application codes Advantages: - Exact - Requires only “main” software Disadvantages: - (Computer) time consuming - Backward (sensitivity) route not obvious 24

25

Application: SFR void coefficient • • • KALIMER-600 Sodium Fast Reactor (Korea) Total Monte Carlo with MCNP and FISPACT Uncertainties due to transport libraries only, but for all materials Sensitivity profiles with MCNP K-eff, void coefficient, burn-up and radiotoxicity using TMC 26

27

The total uncertainty is underestimated. Uncertainties for: • Activation cross sections • Fission yield data • Decay data Are not (yet) taken into account. 28

29

TMC: Other possibilities • • • Random thermal scattering data libraries (? ) Random decay data libraries Random fission yield libraries Normalization to experimental data or other nuclear data libraries at the basic input level (TENDL-2010) Optimization to integral benchmarks using e. g. simulated annealing (“search for the best random file”) 30

31

Optimization of Pu-239 • • Select 120 ICSBEP benchmarks Create 630 random Pu-239 libraries, all within, or closely around, the uncertainty bands Do a total of 120 x 630 =75600 MCNP criticality calculations Do another 120 x 4 calculations: 32

Optimization of Pu-239 33

Optimization of Pu-239 • • 6% of libraries have lower chi-2 than JEFF-3. 1 Library #307 has the lowest 34

35

Conclusions • To improve evaluated libraries, TMC is an easier tool than covariances + perturbation + sensitivity • However, the world wants covariances, and they get covariances (TENDL) • With a reproducible automated system, almost anything is possible. After some years of serious software development we can now fork into various branches: - TALYS Evaluated Nuclear Data Library (TENDL) including complete covariance data (MF 31 -35) - Total Monte Carlo uncertainty propagation - Nuclear data library optimization - Other applications (not discussed here) The results of all improvements in uncertainly handling (UMC, model uncertainties, etc. ) will be directly visible 36