The fundamental issue of reconstructing a porous medium is examined anew in this paper, thanks to a sample of low-porosity Fontainebleau sandstone that has been analyzed by computed microtomography. Various geometric properties are determined on the experimental sample. A statistical property, namely, the probability density of the covering radius, is determined. This is used in order to reconstruct a porous medium by means of a Poissonian generation of polydisperse spheres. In a second part, the properties of the real experimental sample and of the reconstructed one are compared. The most important success of the present reconstruction technique is the fact that the numerical sample percolates despite its low porosity. Moreover, other geometrical features and conductivity are found to be in good agreement.

• Received 17 January 2001

DOI:https://doi.org/10.1103/PhysRevE.63.061307