Application of Sodium Polyacrylate Superabsorbent Polymer on Moisture Stability of Clay and Prediction of the Suction Potential of the Mixture.

Abstract

Superabsorbent polymers (SAPs), such as sodium polyacrylate (PAAS), are the neutralized form of poly (acrylic acid) and belong to a class of materials characterized by three-dimensional networks of flexible polymer chains with exceptional water absorption and retention capacities. Due to these properties, PAAS has been widely used in agriculture as a soil water conditioner. This study investigates the potential of PAAS as a soil stabilizer for infrastructure applications. Three suction measurement techniques, namely the axis translation method, osmotic technique, and vapor equilibrium method, were employed to determine the suction behavior of PAAS and soil-PAAS mixtures over a wide range of water contents and to develop their soil-water characteristic curves (SWCCs). The investigation covers the full suction spectrum, from near-complete dryness to full saturation. Experimental results show that the SWCC of PAAS exhibits the three characteristic zones commonly observed in soils: boundary, transition, and residual zones. However, when PAAS is mixed with soil, the boundary and transition zones disappear from the SWCCs of the soil-PAAS mixtures. This behavior is attributed to the suppression of PAAS suction capacity caused by soil confinement. Unlike agricultural applications, where SAP particles are relatively unconstrained, engineering applications typically involve highly compacted soils that restrict polymer expansion and water absorption. The study also evaluates the feasibility of predicting the suction behavior of soil-PAAS mixtures using numerical modeling techniques based on limited experimental datasets. Among the methods considered, Lagrange interpolation and K-nearest neighbors (KNN) produced prediction models with errors below 10%. In contrast, deep neural network models demonstrated lower predictive accuracy, primarily due to the limited size of the available dataset.