Transient Memory via Local Pressure Heterogeneity in Non-Equilibrium Systems

Dimensional Memory Encoding (DME) is presented as a precise operational framework for identifying transient physical memory in driven non-equilibrium systems. In this work, “dimensional” refers to information distributed across spatial degrees of freedom. It does not refer to hidden dimensions, quantum storage, biological memory or consciousness. The framework is tested in a two-dimensional overdamped Brownian system of 72 repulsive soft disks at area fraction φ = 0.68. Two identical localized compression pulses are applied at fixed left and right sites. The two histories contain the same sites, amplitudes, widths, durations, ramp functions and total schedule, differing only in temporal order: left-then-right or right-then-left. The study uses 120 paired initial configurations, 240 trained trajectories and 1,200 post-drive states recorded at five relaxation delays. After all external forcing is removed, bulk potential energy, global pressure and protocol work do not reliably identify the pulse order. Local virial-pressure geometry, however, distinguishes the histories with balanced accuracy 0.908, with a 95% interval of 0.873–0.942, compared with 0.863 for local density. Spatially shuffling the pressure values while preserving their complete distribution reduces decoding to chance, demonstrating that the historical information depends on spatial arrangement. A later weak mechanical probe also produces an order-dependent response. The signal progressively decays during undriven relaxation, identifying a finite memory window. The central conclusion is that local interaction-pressure geometry can act as a transient, distributed and mechanically readable carrier of temporal-order information within the declared Brownian soft-particle model. The work does not claim consciousness, hidden dimensions, quantum memory, a new physical force or universal memory in all matter.