The careful hydrostatic weighing measurements with alternative preparations performed could not verify the working hypothesis of an existing very thin air layer on objects immersed into water, a problem that occurs whenever primary density determinations are carried out. Most probably, the disturbing buoyancy effects we thought were caused by not de-aerating are instead most probably due to small convection currents in the water bath. These can be induced in two ways. One reason is a slight temperature difference between the artefact and the water. The other is due to the fact of not having a perfectly stable and isotropic temperature distribution. Our results indicate that the experiments suffered of both especially an ongoing convection despite a very stable vertical temperature gradient in the water (1mK/cm). The driving force is believed coming from small temperature variations in the surrounding air producing both additive and subtractive overlaying friction forces as a function of time. This effect has so far not been reported as an important error contribution. The idea of stirring the bath in order to achieve temperature homogeneity is not an alternative, as this would not allow
performing weighing at all.
Due to these findings, the planned experiments studying the air layer effect in liquids differing in surface tension and on various materials and surface roughness are not meaningful to carry out. It seems that the mentioned effect is enlarged in hydrometer calibration. We suggest instead a different goal for the project, namely to carefully study the temperature field and its behaviour in time both in the active and buffering water bath as well as in the surrounding air to indirectly determine the dynamic forces on the artefact under weighing. It is further important to repeat the measurements with some of the earlier artefacts at some controlled higher and lower temperatures than the test water. This would produce experimental data for modelling the convection effect on the hydrostatic weighing result. The model then could be used both for correction and uncertainty estimation purposes, an important aspect in hydrostatic weighing as this effect is of the same order than the stated measurement uncertainty so far.