Strategies for helium pycnometry density measurements of welded ignimbritic rocks

Stefano Columbu

First

Writing - Original Draft Preparation

;Maurizio Mulas;Filippo Mundula;Raffaello Cioni

2021-01-01

---

Abstract

This study is aimed at recognizing possible sources of dispersion of density measurements of welded ignimbritic rocks by helium pycnometer. The values of bulk density and solid fraction volume, useful for volcanological studies as well as for a complete definition of the mechanical properties of these rocks, commonly used for present or historical buildings, are often tricky to obtain, due to the presence of micropores and the high tortuosity of the porous network. Three samples representative of different lithofacies of Miocene ignimbrites from San Pietro Island (SW Sardinia, Italy), have been used here to tune up a density measurement method. Basing on a statistical study of the density data derived from different analytical strategies, we propose a specific procedure for determining to measure density of welded ignimbritic rocks. Several measurements (runs) are performed on the undisturbed and on powered specimens of the three samples, using two different methods of gas-purging of the pycnometer: flow and pulse purging modes, respectively. The main factors that can produce a source of accidental errors are: uncontrolled fluctuations in temperature of the instrument, pollution of the sample (e.g., moisture), intrinsic petrophysical characteristics (e.g., pore tortuosity) of ignimbritic rocks. Conversely, the main systematic errors introduced in the measurements revealed to be related to the selected time and mode of gas-purging of the instrument during the first steps of the analysis. Using an appropriate sequence of contiguous cycles, accuracy and precision of data can be optimized, thus excluding systematic errors. The results show that in the first cycle, regardless of the purge time (or pulses), the values of density and standard deviations are always higher compared to the second and the third cycles. Basing on the analysis of our data, we propose that the better procedure to obtain an accuracy data is using 10-run cycles and considering the average value of the last seven measures.