Series of publications are describing annually varying periodic fluctuations in nuclear decay data. Air and space pilots of the US air and space forces have analysed nuclear decay rates displaying annual fluctuations, measured in metrology institutes (http://arxiv.org/abs/1007.0924v1).

We were interested to see at what level of accuracy we can measure nuclear half-life with germanium semiconductor detectors (HPGe) in the Experimental Physics Department at Bem tér. It is customary and overwhelmingly abundant to assume Poisson distribution for the measured x-ray and gamma ray spectra, and deriving the true input rate (N) and its statistical uncertainty SQRT(N), from the Poisson distribution. However, the preamplifier-signal processing chain has dead times and applies discriminators, as a consequence disturbs the randomness. In addition, other distributions are contributing, like noise, electronic disturbances, and improperly developed signals, therefore the assumption of Poisson distribution is invalid for the measured spectra. We have used two new approaches, where the inversion problem is not posed, and there is no need to assume the Poisson distribution. The half-life of 68Ga (T1/2≈68 min) was measured using three HPGe detectors simultaneously, in two runs. One method was a dead time and discriminator free time interval histogram analysis, with almost no pile up. This method found half of the measurements useless, although the energy distribution spectra have not revealed systematic effects, or un-deterministic behaviours of the detectors. We could have determined (erroneous) half-life with a very small uncertainty (10^-4). For those measurements, where neither the collection of accepted and the rejected energy distribution spectra approach (second method), or the time interval histogram analysis, accompanied with rudimentary energy distribution spectra have not indicated other contribution, or un-deterministic behavior of the detector, the time interval histogram analysis yielded an order of magnitude smaller values than SQRT(N) for the statistical uncertainty. Although the possibility to measure with an order of magnitude smaller uncertainty is important, in our opinion the additional method to observe systematic effects is also significant.

The exponential analysis is an ill-posed problem. We have learned at well-posed problems that the more data points are measured on the curve, the more accurately it can be fitted. To the contrary, the excess to the necessary data points makes the inversion problem less stable at the ill-posed problem.

We take the liberty to conclude from these exercise, that the strong contradictions in the PIXE data base could be resolved using the proper statistical uncertainty (notSQRT(N)).

Lecture in Hungarian.
Before the event (from 10:30) we offer a cup of tea and some cookies.