Nuclear mass forecasting: can observed pattern determine mass values?

International audienceTheoretical prediction of nuclear masses is analyzed as a pattern recognition problem on the N-Z plane. A global pattern is observed by plotting the differences between measured masses and Liquid Drop Model (LDM) predictions. After unfolding the data by removing the smooth LDM mass contributions, the remaining microscopic effects have proved difficult to model, although they display a striking pattern. These deviations carry information related to shell closures, nuclear deformation and the residual nuclear interactions. In the present work the more than 2000 known nuclear masses are studied as an array in the N-Z plane viewed through a mask, behind which the approximately 7000 unknown unstable nuclei that can exist between the proton and neutron drip lines are hidden. Employing a Fourier transform deconvolution method these masses can be predicted. Measured masses are reconstructed with and r.m.s. error of less than 200 keV. The existence of an island of stability around (Z\approx$ 116, N\approx 194) is strongly suggested. Other potential applications of the present approach are outlined