Compression Aging: Dynamics and Mechanisms

• Compression aging is the irreversible evolution of system structure, dynamics, and memory triggered by repeated compressive stresses leading to plastic events and configurational rearrangements.
• Methodologies such as cyclic loading, static stress, and rapid quenches reveal logarithmic or power-law relaxation kinetics in granular, glassy, and AI systems.
• Insights from compression aging guide material design and memory system robustness by linking microscopic rearrangements to macroscopic properties like densification, modulus change, and recall decay.

Compression aging refers to the irreversible evolution of structure, dynamics, or information encoding in a system subjected to repeated or sustained compressive perturbations. Though the term spans physical disordered materials and artificial agent memory systems, a common characteristic is the gradual loss, rearrangement, or transformation of microscopic details as a direct consequence of history-dependent compaction and the mechanisms used to encode or store state under constraint. Compression aging has been extensively characterized in granular matter, structural glasses, metallic glasses, and, more recently, in AI agent memory pipelines, each displaying analogous kinetic, thermodynamic, or informational signatures.

1. Physical Models and Protocols for Compression Aging

Compression aging in physical systems is typically studied using controlled loading protocols—cyclic compressive strains, uniaxial or hydrostatic loading, or mechanical annealing cycles—that probe the system's relaxation and memory. Archetypal protocols include:

• Single- and Multi-Step Compressions: Sudden (step-like) changes in external strain or packing (e.g., gap or volume) with intermediate holds, as used in uniaxially compressed foams, crumpled sheets, and model jammed solids (Lahini et al., 2016, Mandal et al., 2021).
• Cyclic Compression (Mechanical Annealing): Alternating compression–decompression sequences, quantifying evolution in packing fraction, modulus, or microscopic configuration, applied to granular rafts or packings (Shimamoto et al., 2024, Hobson-Rhoades et al., 9 Mar 2026, Yuan et al., 17 Oct 2025).
• Aging under Static Load: Prolonged holding under subyield or moderate stress at fixed temperature, specifically in polymer glasses and metallic glasses, to measure aging or rejuvenation (0704.1840, Priezjev, 2019, Cornet et al., 2023).
• Rapid Quench Protocols: Sudden densification or pressure increases, taking systems below a glass transition or jamming threshold and then monitoring kinetic aging (Gago et al., 2020, Elizondo-Aguilera et al., 2022).
• Memory-Compression Pipelines in Agents: A sequence of session transcripts and memory compactions, with retention/failure statistics monitored over deployment (Zhu et al., 25 May 2026).

All these protocols are designed to generate a slow, often logarithmic or power-law, evolution in relevant observables (e.g., density, modulus, event rate, information recall), enabling systematic measurement of aging effects.

2. Kinetics and Scaling Laws of Compression Aging

Compression aging in both athermal and glassy materials universally manifests as slow, history-dependent relaxation—typically logarithmic or power-law in time or the number of applied loading cycles/taps. Scaling representations include:

• Logarithmic Densification: Compaction proceeds as

$\phi(N) \approx \phi_0 + A \ln(1 + N/N_0)$

over decades in cycle number $N$ (Gago et al., 2020, Hobson-Rhoades et al., 9 Mar 2026, Yuan et al., 17 Oct 2025).

• Decelerating Event Rates: Record-breaking or quake events that drive relaxation occur with rate $r(t) \sim 1/t$, leading to record (log-Poisson) dynamics (Gago et al., 2020).
• Aging Exponents for Relaxation Times: The system's structural relaxation time $\tau$ grows with age (wait time $t_w$) according to:

$\tau(t_w) \sim t_w^{\delta}$

with regimes $\delta<1$ (sub-aging), $\delta=1$ (simple aging), and $\delta'>1$ (hyper-aging) depending on proximity to the glass transition or quench depth (Elizondo-Aguilera et al., 2022).

• Observable-Specific Superposition: For mechanical observables, compliance or mean-square displacement curves can be collapsed by time–age superposition, evidencing a rigid aging of relaxation spectra (0704.1840).
• Informational Decay in Compression Memory Pipelines: In agents, metrics such as keyword survival $m(t)$ or chain recall decay smoothly in session index $N$0, with a definable half-life $N$1 and decay slope that depend on the compression ratio $N$2 and memory policy (Zhu et al., 25 May 2026).

These dynamical laws are often accompanied by emergent universality in two-time scaling variables ($N$3) and are insensitive to microscopic details of dissipation or constraint enforcement.

3. Microscopic Mechanisms and Structural Evolution

Compression aging leads to qualitatively new structures and energy landscapes. Mechanisms include:

• Irreversible Plastic Events: Local rearrangements (plastic events or stress-drops) that lower internal energy, leading to densification and increased coordination in the aging regime (Gago et al., 2020, Hobson-Rhoades et al., 9 Mar 2026).
• Size Segregation and Energy Landscape Transitions: In polydisperse spheres under extensive annealing, compression induces size-dependent effective attractions, promoting segregation of like-sized particles through energy-driven migration (Shimamoto et al., 2024).
• Competition between Aging and Rejuvenation: In glasses or alloys, sustained load or pressure can produce either aging (relaxation to deeper minima) or rejuvenation (activation into higher-energy basins), depending on stress and temperature (Cornet et al., 2023, Priezjev, 2019).
• Void Dynamics in Cohesive Granular Rafts: Slow collapse of large voids and accompanying non-affine particle motion decrease heterogeneity and increase global density (Hobson-Rhoades et al., 9 Mar 2026).
• Frictional Mediation in Granular Materials: Friction modulates the balance between intermittent avalanche-like rearrangements and continuous creep-driven compaction, yielding reentrant transitions between aging and fluidized regimes (Yuan et al., 17 Oct 2025).
• Information Filtering and Loss in Agent Memory: Systematic loss of low-frequency or rare details occurs at each memory write as full transcripts are recursively compacted, modeled as decay in retention probability with increasing age and compression (Zhu et al., 25 May 2026).

The interplay between microscopic event statistics, collective rearrangements, and evolving mechanical or informational constraints determines the macroscopic aging pathway in each context.

4. Memory, Hysteresis, and History Dependence

Compression aging is strongly path-dependent, exhibiting pronounced hysteresis and memory effects:

• Memory Encoding in Stress Response: Under two-step (Kovacs-like) protocols, systems show non-monotonic (overshoot) relaxation, with the time to peak response scaling linearly with the waiting time at prior compression. This allows explicit readout of "age" (Lahini et al., 2016, Mandal et al., 2021).
• Hysteretic State Evolution: Pressure–packing curves and structural observables differ on compression and decompression, marking irreversible history encoded in configuration and contacts (Shimamoto et al., 2024, Cornet et al., 2023).
• Protocol-Dependent Final States: The details of compaction, annealing ramp speeds, or compression amplitudes select among nonergodic glassy states, leading to protocol-dependent endstate densities and mechanical properties (Gago et al., 2020, Elizondo-Aguilera et al., 2022).
• Information Memory in Agents: Dependency-directed graphs and counterfactual probes separate error sources by stage (write, read, utilization), allowing precise attribution of loss to compression, interference, or downstream utilization aging (Zhu et al., 25 May 2026).

These effects demonstrate that compression aging is fundamentally non-Markovian, with the current state determined not just by instantaneous control parameters but by the full loading or processing history.

5. Regimes, Control Parameters, and Crossover Phenomena

Compression aging exhibits distinct regimes depending on system parameters:

• Weak vs. Strong Cycling/Compression: For granular packings, weak cyclic loading causes monotonic compaction; strong loading induces segregation, modulus reduction, and net expansion in polydisperse assemblies, governed by a threshold in maximum packing fraction or pressure amplitude (Shimamoto et al., 2024).
• Aging, Fluidized, and Elastic Regimes in Frictional Granular Media: The (strain amplitude, friction) state diagram delineates domains with logarithmic aging, rapid fluidization, or immediate elastic locking. Friction controls stabilization at small μ and fluidization at large μ, resulting in non-monotonic (reentrant) dependence (Yuan et al., 17 Oct 2025).
• Transition Between Aging and Rejuvenation: In metallic glasses and amorphous alloys, the applied pressure or static load can push the system from an aging-dominated to a rejuvenation-dominated regime, which can be mapped in the (stress, temperature) or (pressure, temperature) parameter space (Priezjev, 2019, Cornet et al., 2023).
• Compression Policy Control in Agent Memory: The structure and aggressiveness of the memory compaction policy (e.g., lossy vs. careful prompt) directly determine the half-life of fact recall and the onset of catastrophic aging in information retention (Zhu et al., 25 May 2026).

These crossovers are often sharp and can be predicted or fit quantitatively, e.g., via scaling exponents or threshold criteria.

6. Practical Implications and Applications

Compression aging has substantial implications for materials design, reliability engineering, and agent deployment:

• Material Property Tuning: Controlled loading can program the mechanical response of amorphous solids or granular composites (e.g., stiffness, yield strength) by selecting the appropriate position in the aging–rejuvenation space (Priezjev, 2019, Cornet et al., 2023).
• Memory System Robustness: In deployed agents, fact retention can be prolonged by augmenting memory policies (careful prompts, typed-state overlays), and diagnostic probes can localize compression-induced failures, informing mitigation strategies (Zhu et al., 25 May 2026).
• Structural Evolution under Repeated Use: Rafts, soils, and powders subjected to repeated loading undergo slow evolution that alters mechanical performance (e.g., increased strength, embrittlement, or loss of ductility), requiring adaptation in engineering cycles (Hobson-Rhoades et al., 9 Mar 2026).
• Non-Monotonic Aging Phenomena: Expansion following compaction (as opposed to unidirectional densification) in polydisperse granular systems points to the need for careful control of compression amplitude in applications such as powder compaction, soil stability, or mechanical metamaterials (Shimamoto et al., 2024).

A plausible implication is that any system reliant on compressive compaction for functionality or reliability—physical or informational—must account for long-term, history-dependent, and regime-specific aging kinetics.

7. Universality, Theoretical Perspectives, and Broader Context

Compression aging bridges concepts across glassy dynamics, athermal relaxation, and information theory:

• Universality in Relaxation Laws: Logarithmic compaction, 1/t event rates, and power-law aging exponents appear across granular, polymeric, colloidal, metallic, and agent-based systems (Gago et al., 2020, Elizondo-Aguilera et al., 2022, Zhu et al., 25 May 2026).
• Record Dynamics and Log-Poisson Statistics: Rare, irreversible events (“record quakes”) drive aging universally in nonergodic systems, and two-time scaling collapses reflect a deep connection between noise-driven and deterministic relaxation (Gago et al., 2020).
• Linear Response and Memory Kernels: Non-monotonic (overshoot) effects in Kovacs-like protocols are generic consequences of slow, broad-spectrum relaxation and do not require true aging (breaking of time-translation invariance) (Mandal et al., 2021).
• Scaling Theories: Mode-coupling theory (MCT) and Onsager’s irreversible process framework provide explicit aging exponent predictions directly linked to underlying static structure (Elizondo-Aguilera et al., 2022).
• Agent Compression Aging as Computational Analogue: The recursive compaction and memory decay in persistent agents mirrors physical aging, with compression ratio and information filtering acting analogously to energy dissipation or mechanical rearrangement (Zhu et al., 25 May 2026).

These theoretical and phenomenological frameworks unify disparate manifestations of compression aging, emphasizing its importance in both natural and engineered systems.