Memory effects using an elemental analyser to combust radiocarbon samples: Failure and recovery

During a measurement run some time ago, we measured, indeed, radiocarbon concentration values somewhat higher than usual in nominally blank samples. These “bad” data could be explained by memory effects. By assuming a constant contribution from the sample of the prior combustion, this effect might be corrected: indeed, by repeating cycles of sequential combustions of standards and blanks, we observed a good reproducibility of the amount of contamination from the previous sample needed to explain the results. However, we were obviously unhappy with the fact itself of such corrections being needed, and several tests were performed to identify the source of contamination and eliminate it. Eventually, we succeeded in finding the cause of this failure and in recovering the full performance of the system.