DSperse: Multidisciplinary Dispersion Frameworks

• DSperse is a multifaceted concept encompassing specialized dispersion phenomena in porous media, polymer blends, zero-knowledge ML, and CFD particle simulations.
• In porous media, DSperse leverages diffusiophoresis to modulate colloidal transport via salt gradients, yielding regimes of enhanced or suppressed dispersion.
• DSperse frameworks offer practical benefits such as efficient ML verification with slice-based ZK proofs, on-demand polymer dispersion, and accurate simulation of particle-fluid interactions.

DSperse denotes multiple specialized concepts and frameworks across diverse scientific and technical domains, most notably: (1) diffusiophoretically altered colloid dispersion in porous media (“DSperse mechanism”) (Pujari et al., 6 May 2026); (2) a slice-based framework for selective zero-knowledge proof (ZKP) verification in distributed machine learning inference (“DSperse framework”) (Ivanov et al., 9 Aug 2025); (3) the DSperse phenomenon in polymer blends, referring to rapid aqueous dispersion after phase separation (Zhu et al., 2015); and (4) a direct numerical simulation method for dispersed particles in compressible fluid dynamics (historically spelled “DSperse” as an acronym) (Tatsumi et al., 2012). Below, each is detailed in its primary technical context.

1. Diffusiophoretic Dispersion in Porous Media

DSperse, in the context of colloidal transport, describes the phenomenon whereby solute gradients modulate macroscopic colloid dispersion within porous media via diffusiophoresis. In such systems, each particle acquires a drift velocity $v_{dp} = D_{dp} \nabla \!\ln c$ (with $D_{dp}$ the phoretic mobility and $c$ the salt concentration field), leading to sign-dependent effects on dispersion (Pujari et al., 6 May 2026).

Key features:

• Attractive phoresis ($D_{dp}>0$): Particles migrate up salt gradients, which paradoxically enhances longitudinal dispersion, even producing bimodal (split peak) distributions by amplifying the residence-time differences between fast and slow streamlines.
• Repulsive phoresis ($D_{dp}<0$): Particles move down salt gradients, suppressing dispersion and resulting in more compact blobs.
• Minimal two-layer model: Captures the physics using coupled plug flows (velocities $u_1 < u_2$), where the effective macroscopic dispersion is

$D_\mathrm{eff} = D_c + \frac{\gamma^2}{2k}$

with $D_c$ the colloidal diffusivity, $\gamma = (u_2 - u_1)/2$, and $k$ the combined transverse exchange rate (diffusion + phoresis).

• Disorder: Introducing randomness in the porous geometry (disorder parameter $D_{dp}$0) tunes the regime from shear-dominated (dispersion enhancement/splitting) to diffusion-dominated (dispersion suppression/homogenization).

The DSperse effect is thus a direct consequence of diffusiophoretically modulated cross-channel exchange, fundamentally altering classical Taylor–Aris dispersion predictions.

2. DSperse in Zero-Knowledge Machine Learning Verification

DSperse is a modular framework targeting trust-minimized, scalable verification of distributed ML inference via selective, slice-based ZK proof generation (Ivanov et al., 9 Aug 2025). Instead of monolithic model circuitization—which is often intractable—DSperse enables partitioning the inference pipeline into verifiable “slices,” each implemented as a distinct circuit and ZK proof.

Technical structure:

• Core entities: Orchestrator (schedules/model partitioning), Provers (compute slice, generate per-slice proof), Verifiers (slice-level proof validation), and optionally, audit/replication layers.
• Slice formalism: A slice $D_{dp}$1, where $D_{dp}$2 is an arithmetic circuit for a sub-model $D_{dp}$3, $D_{dp}$4 are public/activation inputs, $D_{dp}$5 are outputs.
• Proof composition: DSperse is agnostic to SNARK/STARK backend; proof/verification time scales with slice circuit size. Audit logs, redundant execution, or staking enforce global (cross-slice) consistency.
• Empirical performance: Experiments using LeNet-5 on CIFAR inputs show that per-slice proofing reduces memory footprint by 40–60%, and total proof time by up to 70%, with negligible accuracy penalty (logit discrepancy $D_{dp}$6 for EZKL).

DSperse is especially suited where high-leverage sub-computations must be proved verifiably, but full-model ZKP remains prohibitive.

3. DSperse Phenomenon in Recrystallized Polymer Blends

DSperse, as observed in recrystallized poly(ethylene glycol)–poly(lactic acid) (PEG–PLA) blends, refers to the rapid and complete dispersion of films in aqueous environments, initiated by thermally induced phase separation (Zhu et al., 2015).

• Process: Films are isothermally annealed above PEG and PLA melting points, then slowly cooled, permitting PLA spherulite crystallization with PEG segregating into percolating, amorphous channels.
• Mechanism: Upon immersion in water, PEG dissolves rapidly (Fickian kinetics with $D_{dp}$7 m$D_{dp}$8s$D_{dp}$9), severing the PLA network and dispersing discrete PLA spherulites as colloidal particulates in 6–8 h for 30 wt% PEG.
• Critical threshold: DSperse requires PEG content above the percolation threshold (typically $c$0 25–30 wt %); below this, the film remains intact.
• Control variables: Cooling rates, PEG/PLA ratio, and crystallization kinetics can modulate dispersion time and completeness.
• Applications: Enables “on-demand” dissolvable matrices for tissue scaffolds, controlled-release, and transient electronics.

The DSperse event unambiguously correlates with microstructural channels imaged by SEM, and is absent in amorphous/quench-cast blends or at sub-percolation PEG fractions.

4. DSperse in Direct Numerical Simulation of Dispersed Particles

In the context of computational fluid dynamics, DSperse (Direct-numerical-Simulation of Dispersed particles in a comPReSSible fluiD) designates a diffuse-interface formulation for coupling rigid particles and compressible fluids (Tatsumi et al., 2012).

• Core equations: Compressible Navier–Stokes for barotropic Newtonian fluids with momentum/density fields $c$1, and smoothed particle profiles $c$2 to mediate particle–fluid coupling.
• Numerics: A split semi-implicit time-stepping advances sound, viscosity, and rigid-body motion independently. Hydrodynamic forces/torques derive from the momentum exchanged inside diffuse interfaces.
• Validation: The scheme faithfully reproduces analytic velocity relaxation curves for impulsively forced spheres, captures hydrodynamic memory kernels, and recovers correct thermal velocity autocorrelation (VACF) when stochastic stress is included.
• Physical insights: Demonstrates phenomena such as added-mass jumps, compressible viscous backtracking, and the interplay of compressibility parameter $c$3.

This approach enables mesoscale resolution of dense, thermally fluctuating, many-particle suspensions in compressible solvents—a regime otherwise inaccessible to classical lattice-Boltzmann or Lagrangian methods.

5. DSperse in Polymer Melt Dynamics: Dispersity Effects

While not always spelled “DSperse,” dynamic effects linked to dispersity (denoted $c$4 or $c$5) in polymer systems overlap terminologically due to their focus on dispersed states or events (Tejuosho et al., 2024). Dispersity here quantifies the molecular weight distribution breadth, with

$c$6

where $c$7 and $c$8 are weight- and number-average molecular weights, respectively.

• Static effects: For fixed $c$9, end-to-end length and gyration radius remain “ideal” and independent of $D_{dp}>0$0.
• Dynamic effects: As $D_{dp}>0$1 rises, long-chain diffusion constants increase (up to $D_{dp}>0$2 for $D_{dp}>0$3), consistent with constraint-release—the rapid relaxation of short chains transiently collapses tube confinement for long chains, accelerating reptation.
• Kinetics: Entanglement times decrease at high $D_{dp}>0$4 and $D_{dp}>0$5, yet remain unchanged at low $D_{dp}>0$6 or $D_{dp}>0$7.
• Structure factor: K-space analysis reveals a diminishing Kratky minimum (apparent local stiffening) as $D_{dp}>0$8 increases.

A plausible implication is that controlling $D_{dp}>0$9 in melt processing enables precise tuning of viscoelastic and relaxation properties without altering chain chemistry or blending ratios.

6. Synthesis, Distinctions, and Nomenclature

“DSperse” as a term is domain-specific and context-dependent:

• In soft matter and porous media transport, it denotes a phoresis-modified dispersion process (Pujari et al., 6 May 2026).
• In cryptography/ML, DSperse formalizes strategic, slice-based ZKP partitioning (Ivanov et al., 9 Aug 2025).
• In polymer science, DSperse references both a rapid phase-induced dispersion event in blends (Zhu et al., 2015) and the influence of dispersity-distribution on melt dynamics (Tejuosho et al., 2024).
• In CFD/numerics, it labels a simulation scheme for coupled particle–fluid compressible flows (Tatsumi et al., 2012).

Each definition underpins conceptually distinct mathematical, algorithmic, and physical mechanisms, but all relate to the emergent properties of complex, distributed, or multiphase systems under the influence of disorder, selective interactions, or modularization. No unifying nomenclature across these domains is consistently established beyond the coincidence of “DSperse” as a label.