Characterization of reactor fuel burn-up from antineutrino spectral distortions E. Kemp, L.F. G. Gonzalez, T.J.C. Bezerra and B. Miguez for the ANGRA Collaboration State University of Campinas - UNICAMP Physics Institute- Cosmic Rays Department ANGRA Neutrinos Project
So/Si
Neutrino Spectra Parametrization: Precision spectroscopy with reactor anti-neutrinos. Patrick Huber, Thomas Schwetz, hep-ph/ Patrick HuberThomas Schwetz
Simulation Steps Energy draw from selected spectrum Isotope Selection Weighting by cross-section Fitting routine to extract the isotope fraction 1000 events Fuel evolution Static Fuel Energy resolution dE = k.E dE=0
Fitting convergence study: 239Pu fraction Events 239 Pu Fission contribution Perfect energy resolution Static Fuel Assumed rate: 1000/day (Angra expectation)
High Statistics (exposure) Needs Shape comparison: Kolmogorov-Smirnov test –Neutrino spectrum: Composition from normalized Schreckemback’s spectra (235U, 239Pu and 241Pu)
Spectral Distortion: expectations from burn-up Taking the ratio between the spectra measured in the n-th month and the first one, we can observe the distortion induced by the burn-up
Spectral Distortion: expectations from burn-up
Nucifer Simulations: we are in good agreement Thanks to D. Lhuillier !
The Spectral Ratio Fit: an example 6 th Month after reactor starting
Spectral Ratio Fit Red: linear fit Green: 95% C.L. bands
The slope time dependence
Is the slope of R(t) a good indicator for deviations from the expected behavior ? Let’s assume a diversion of 1/3 of the reactor fuel during the 6 th month
Burnup: impact on the spectrum shape with 1/3 of the fuel replaced at half-cycle Day number Fission fraction
Day number Fission fraction Burnup: impact on the spectrum shape with 1/3 of the fuel replaced at half-cycle
The slope time dependence Is it an outlier?
The slope time dependence Yes, with 75% C.L.
Simulation Steps Energy draw from selected spectrum Isotope Selection Weighting by cross-section Χ 2 – KS tests Null hypothesis: No distortion Repeat until Fuel evolution: Poisson-like time interval μ= f.Δt frequency f = f(D,M) D,M: detector distance and mass Static Fuel Energy resolution dE = k.E dE=0 Experimental Data (Shreckemback’s Spectra)
Simulation Steps Energy draw from selected spectrum Isotope Selection Weighting by cross-section Χ 2 – KS tests Null hypothesis: No distortion Repeat until Fuel evolution: Poisson-like time interval μ= f.Δt frequency f = f(D,M) D,M: detector distance and mass Static Fuel Energy resolution dE = k.E dE=0 Experimental Data (Shreckemback’s Spectra)
Simulation Steps Energy draw from selected spectrum Isotope Selection Weighting by cross-section Χ 2 – KS tests Null hypothesis: No distortion Repeat until Fuel evolution: Poisson-like time interval μ= f.Δt frequency f = f(D,M) D,M: detector distance and mass Static Fuel Energy resolution dE = k.E dE=0 Experimental Data (Shreckemback’s Spectra)
Hypothesis Tests Results
Chi^2 vs. KS Chi^2 (is more optimistic…) –More Type II Errors KS test –More Type I Errors See T.J.C. Bezerra, B. Miguez and R.M.Almeida works (poster session) for detailed numbers and generalities on this (including oscillation studies)
Chi^2 vs. KS Is it possible to profit the better from both of the tests? –Fisher’s method: Combination of N different results (p-values) of independent statistical tests resulting in a Chi^2 like quantity with 2K degrees of freedom Next step for this study…
Conclusions Isotopic composition measurements by shape analysis only requires a large number of events –Reduce the time integration: Large time intervals degrades information –High exposure: source luminosity + detector mass+ time Recognition of fuel diversion is possible by observing UNEXPECTED spectral distortions (but, how much?) Required Improvements: –More sophisticated analysis methods to quote the sensitivity in mass of the recognition method –Combining information: Shape + Counting Rates different statistical methods working together – Fisher’s method PCA, LDA: decomposition of a mixed signal (?)
Thank you !