Neutron imaging unveils heterogeneous flow patterns in homogeneous porous media and limitations of Darcy-Richards models

We applied neutron imaging techniques to unveil pore scale flow processes occurring during desaturation of a homogeneous, saturated sand pack. For this purpose, we used a small glass flow cell (2 cm x 2 cm x 0.1 cm) filled with pure, artificial S250 quartz sand. The pore space of the sand was initially fully saturated with double distilled water (DDW). The saturated flow cell was subjected to a series of suction phases with increasing suction tensions to extract water via a bottom outlet, controlled by a vacuum pump. In the first phase, a tension of 0.016 MPa (low suction) was applied for 248 min, followed by 1.14 MPa (mid suction) for 227 min, and finally, 10 MPa (high suction) for 397 min. Throughout the entire duration of the experiment, the flow cell was continuously exposed to neutrons. A back-end detector collected the neutron beams passing through the different matters (sand, water, air) contained in the flow cell and generated snapshot images of the internal pore structure and the water distribution with a pixel resolution of 5 µm at one-minute intervals.

The resulting images revealed that water did not redistribute homogeneously during the desaturation of the flow cell, over dimensions of a few millimeters. Despite using “perfectly homogeneous” sand under initially fully water-saturated, controlled conditions, heterogeneous patterns of stable water pockets were observed inside the pore space of the sand, where water became immobilized.

These experiments demonstrate that truly homogeneous flow does not occur even under controlled laboratory conditions in a “perfectly homogeneous” porous medium.

Subsequent simulations of the experiments with common Darcy-Richards models showed that the macroscopic 1D desaturation time series of the flow cell could be realistically depicted. However, even after parameter calibration and the manual addition of heterogeneity, the microscopic, heterogeneous 2D distribution of water observed inside the flow cell could not be reproduced.

This highlights limitations on the applicability of Darcy-Richards models, which may be effective at a macroscopic level but simultaneously fail to represent accurately the internal dynamics of the system. This insight is crucial for the application of Darcy-Richards models and the interpretation of their results.