Investigation of the WIMP annual modulation signature 289in our experiment is given by PMT fast single photoelectrons with decay
times of the order of tens of nanoseconds, while the scintillation pulses have
decay times of the order of hundreds of nanoseconds. The large difference in
decay times and the relatively large number of available photoelectrons response
(5.5–7.5 photoelectron/keV depending on the detector) ensures effective noise
rejection; see, e.g., [12] for details. To investigate quantitatively the possible role
of a noise tail in the data after noise rejection on the annual modulation result,
the hardware rate,RHj, of each detector above a single photoelectron, can be
considered. The distribution of&j(RHj−〈RHj〉)shows a Gaussian behaviour
withσ = 0 .6% and 0.4% for DAMA/NaI-3 and DAMA/NaI-4, respectively,
values well in agreement with those expected on the basis of simple statistical
arguments. Moreover, by fitting its time behaviour in both data periods including
a WIMP-like modulated term a modulation amplitude compatible with zero
( 0. 04 ± 0. 12 )× 10 −^2 Hz, is obtained. From this value, considering also the
typical noise contribution to the hardware rate of the nine detectors, the upper
limit on the noise relative modulation amplitude has been derived to be [15] less
than
- 6 × 10 −^3 Hz
9 × 0 .10 Hz
1. 8 × 10 −^3 (90% C.L.).
This shows that even in the worst hypothetical case of a 10% contamination of the
residual noise—after rejection—in the counting rate, the noise contribution to the
modulation amplitude in the lowest energy bins would be less than 1. 8 × 10 −^4 of
the total counting rate, that is a possible noise modulation could account only for
less than 1% of the annual modulation amplitude observed in [9]. In conclusion,
there is no evidence that a hypothetical tail of residual noise after rejection plays
any role in the results.
The behaviour of the efficiencies during the whole data-taking periods
has also been investigated; their possible time variation depends essentially on
the stability of the cut efficiencies, which are regularly measured by dedicated
calibrations [9, 15]. In this way, the unlikely idea of a possible role played by the
efficiency values in the observed effect in [7–9] has also been ruled out [9, 15].
In order to verify the absence of any significant background modulation, the
measured energy distribution in energy regions not of interest for the WIMP–
nucleus elastic scattering has been investigated [7–9, 13]. For this purpose, we
have considered the rate integrated above 90 keV,R 90 , as a function of time. The
distributions of the percentage variations ofR 90 with respect to their mean values
for all the crystals during the whole DAMA/NaI-3 and DAMA/NaI-4 running
periods show cumulative Gaussian behaviour withσ 1%, well accounted for
by the statistical spread expected from the used sampling time [9, 15]. This result
excludes any significant background variation. Moreover, including a WIMP-like
modulation in the analysis of the time behaviour ofR 90 , an amplitude compatible
with zero is found in both the running periods:−( 0. 11 ± 0. 33 )cpd kg−^1 and
−( 0. 35 ± 0. 32 )cpd kg−^1. This excludes the presence of a background modulation
in the whole energy spectrum at a level much lower than the effect found in the