“…When σ 0 is such that the simulated signal S is below the measured signal B 0 , B ′ adds to the signal S, completing it to B 0 and the likelihood is unchanged. With increasing σ 0 , when S starts to exceed the function B 0 , B ′ becomes zero, and we calculate the likelihood using S alone in such bins, leading to a new likelihood L. Following the CRESST collaboration prescription, σ 0 excluded at 90% CL is taken to be the value of σ 0 giving ln L−ln L 0 = −1.28 2 /2 [2], since 10% of the area of a normalized Gaussian distribution of width σ is 1.28σ above than the peak. We show the window obtained this way in Figure 2 and for the low mass range, in Figure 1.…”