The stakes are higher than they might appear. In earthquake engineering, the response of soil layers to seismic waves is governed by two fundamental properties: the dynamic shear modulus, which describes the soil’s stiffness during cyclic loading, and the damping ratio, which quantifies how much vibrational energy the material dissipates as heat and internal friction. As shaking intensifies, soils progressively soften, the shear modulus ratio declines with increasing shear strain, while damping rises. Site response analyses, the computer simulations that predict how ground motion will be amplified or attenuated between bedrock and the surface, depend entirely on accurate curves describing these two relationships. Get them wrong, and every downstream prediction of building response, liquefaction potential, and structural demand inherits the error.

For ordinary sands and clays, engineers have long relied on well-established backbone curves, the skeleton stress-strain relationships from which modulus degradation and damping curves are derived. The Hardin–Drnevich and hyperbolic models dominate practice. But EPS lightweight soil is not an ordinary soil. It is a composite: a cement-stabilized soil matrix embedded with thousands of rigid, closed-cell foam particles that are dramatically softer and lighter than the surrounding medium. Its hysteresis loops, the football-shaped curves traced in stress-strain space during each loading cycle, and its energy dissipation behavior do not necessarily conform to the patterns of granular soils. Previous work by some of the same authors explored Hardin–Drnevich and Davidenkov formulations for this material, but each approach carries limitations when extrapolated across the wide range of strain amplitudes that earthquakes and traffic loads impose.

The Ramberg–Osgood model, originally devised in 1943 to describe the stress-strain curves of aircraft alloys, expresses strain as a function of stress with three parameters that control initial stiffness, yield-like curvature, and the sharpness of the transition to plastic behavior. In its traditional geotechnical form, the model assumes fixed parameters, typically a curve-fitting exponent α and a parameter R that governs the shape of the backbone curve at large strains. The innovation in the new research lies in abandoning that fixed-parameter assumption. Instead, Hou and colleagues let the key parameters vary with the dynamic shear strain γd, extracted directly from the laboratory data at each strain level, allowing the model to track the material’s evolving response across five orders of magnitude of deformation.

The experimental foundation of the work is a systematic program of dynamic triaxial tests. In a dynamic triaxial apparatus, cylindrical specimens of EPS lightweight soil are first consolidated under a confining pressure that simulates in-situ stress conditions, then subjected to cyclic axial loading that induces oscillating shear stresses. Sensors record the stress and strain histories cycle by cycle, from which the researchers computed the dynamic shear modulus and damping ratio as functions of shear strain amplitude. By systematically varying the EPS particle content, the cement mixing ratio, and the confining pressure, the team assembled a comprehensive dataset spanning the material’s compositional and stress-state space. They then verified the model’s applicability by altering both the stress state and the stress path during testing, ensuring that the fitted parameters were not artifacts of a single loading configuration.

The results reveal a rich and quantitatively precise picture of the material’s dynamics. For dynamic shear strains below 1×10⁻³, a threshold region that encompasses the small-strain behavior relevant to many serviceability analyses, the variable parameter R of the Ramberg–Osgood model decreases rapidly with increasing strain before settling into a stable range between 1 and 2 at larger strains. This finding carries practical weight: it defines the domain, (1, +∞), within which R remains physically meaningful for lightweight soil, and it identifies exactly where the parameter stabilizes enough to be treated as a constant in simplified analyses. The companion parameter α1, meanwhile, exhibits a more exotic trajectory: when γd is less than 1×10⁻³, α1 first decreases rapidly and then reverses course, increasing with further strain. This non-monotonic behavior is precisely why fixed-parameter formulations have struggled to represent EPS lightweight soil, and why the variable-parameter approach proves essential.

Perhaps the most striking discovery concerns damping. For conventional soils, the damping ratio curve typically follows a single canonical shape, rising smoothly from a small low-strain value toward a plateau at large strains. The EPS lightweight soil refuses to conform. The team identified two distinct families of damping curves: bell-shaped forms, in which damping rises to a peak and then declines at large strains, and S-shaped forms, in which damping increases, plateaus, and then resumes climbing. The form that a given specimen adopts depends on its composition and stress state, specifically the interplay between EPS particle content, cement dosage, and confining pressure. A bell-shaped damping curve implies that at large strain amplitudes the material actually becomes less efficient at dissipating energy, a behavior with direct consequences for seismic design: assuming monotonic damping could overestimate energy absorption and underestimate the motions transmitted to overlying structures.

The physical origins of these unusual behaviors lie in the material’s microstructure. Cement hydration creates brittle bonds between soil grains and around the embedded EPS beads. Under small cyclic strains, these bonds remain intact and the composite responds nearly elastically, with the soft foam particles contributing to overall compliance. As strain amplitude grows, microcracking progressively severs the cementitious bonds, transferring load to the foam particles and the frictional soil skeleton. The EPS beads themselves deform viscoelastically and can locally debond from the matrix, adding interfacial friction as a dissipation mechanism. The observed evolution of the model parameters, R collapsing in the small-strain range and α1 dipping and recovering, reflects this choreography of bond breakage, particle engagement, and interfacial sliding. Confining pressure suppresses crack opening and interfacial slip, which is why it exerts a systematic influence on both parameters, while higher cement ratios stiffen the matrix and shift the strain thresholds at which these transitions occur.

The broader significance of the study lies in what it enables for earthquake engineering practice. Effective stress and equivalent-linear site response codes, the workhorse tools for predicting ground motion amplification, require input curves for the shear modulus ratio G/Gmax and the damping ratio D as functions of strain. By calibrating the variable-parameter Ramberg–Osgood model specifically for EPS lightweight soil, the researchers have supplied a constitutive backbone that can generate those curves for arbitrary compositions and confining conditions, and, crucially, can reproduce the two damping curve families that fixed-form models miss. The authors demonstrate that because the parameters can be adjusted continuously with γd, the model adequately describes the variation laws of both the dynamic shear modulus ratio and the damping ratio under complex dynamic loads, including the changing stress paths encountered in realistic seismic events.

For engineers weighing EPS lightweight soil as an earthquake mitigation strategy, the message is encouraging. The material’s low unit weight reduces seismic earth pressures on retaining structures and inertial forces on foundations, while its tunable stiffness and dissipative capacity offer a degree of seismic isolation. The new model provides the quantitative grounding needed to exploit those advantages rationally rather than empirically. It also connects to the authors’ broader research program on the material, which includes discrete element simulations of its dynamic failure process under cyclic loading and earlier constitutive work based on alternative models, forming a layered understanding that runs from particle-scale mechanics to field-scale design.

Looking forward, the variable-parameter framework invites several extensions. Calibrating the parameter evolution functions against a wider range of confining pressures, loading frequencies, and long-term cyclic degradation would extend the model’s reach to traffic loading and machine vibration problems, where EPS composite soils are also deployed. Coupling the Ramberg–Osgood backbone with pore pressure generation models could address saturated field conditions. And because the parameters were extracted from dynamic triaxial data, translating them to in-situ conditions, where stress states are anisotropic and heterogeneous, remains a classic challenge in soil dynamics. Still, the study marks a decisive step: EPS particle lightweight soil, a composite born from recycled foam and cement, now possesses a dynamic constitutive model worthy of its engineering promise, one that captures not just how stiff it is, but how it softens, how it dissipates, and how those behaviors intertwine when the ground begins to move.