Lagrangian Modeling of History Dependent Fluid Memory

This direction is split into other directions: Lagrangian history tracking. Its last version stays readable and citable here.

Version 1 of 119 papers28 conceptsfrozen 2026-10-04

In classical fluid dynamics, we often assume that a fluid's behavior depends only on its current state—its instantaneous velocity and pressure. This "Markovian" view works well for water or air, but it fails for the complex materials that dominate our world: polymer melts, foams, blood, and even turbulent air. These materials possess memory. Their current stress is not just a function of how they are being deformed now, but of how they were stretched, stirred, or squeezed in the past.

Recent research has converged on a powerful strategy for capturing this memory: moving away from fixed grids (Eulerian) toward particle-based (Lagrangian) frameworks that naturally "carry" their history with them.

The Bookkeeping Problem of Fluid Memory

To model a material with memory, such as a viscoelastic polymer, you must solve an integral constitutive model. This requires knowing the entire deformation history of a specific "chunk" of fluid.

In a traditional grid-based simulation, this is a nightmare. As fluid flows through a fixed mesh, the "chunk" moves from cell to cell. To calculate its current stress, you would have to trace its path backward in time through the grid to find its previous states. As noted by Luca Santelli, Adolfo Vázquez-Quesada, and Marco Ellero in Smoothed Particle Hydrodynamics simulations of integral multi-mode and fractional viscoelastic models, this back-tracking is computationally expensive and prone to errors.

The Lagrangian solution is to treat the fluid as a collection of discrete particles. In methods like Smoothed Particle Hydrodynamics (SPH), each particle represents a physical element of the fluid. Because the particle is the fluid element, it can simply store its own deformation history as an internal variable. This inherent bookkeeping eliminates the need for complex mapping, making it much easier to implement sophisticated models like the fractional Maxwell or K-BKZ models used in industrial polymer processing.

Multiscale Memory without Equations

Sometimes, we don't even have a mathematical formula (a constitutive equation) for a material's memory. For example, in "Oobleck" or discontinuous shear-thickening fluids, the memory emerges from the microscopic interactions of suspended particles.

Nicolás Moreno and Marco Ellero address this in Generalized Lagrangian Heterogenous Multiscale Modeling of Complex Fluids. They propose a Lagrangian Heterogeneous Multiscale Method (LHMM). Instead of using a pre-defined equation, each macroscopic fluid particle carries its own "micro-simulator." As the macro-particle moves, it runs a tiny molecular-level simulation to determine its current stress. This "constitutive-free" approach ensures that the memory of past deformations is preserved perfectly because the microscopic state (the "memory") travels with the macroscopic particle.

Learning the Hidden Correlations of Turbulence

Memory isn't just for "squishy" materials; it also appears in the chaotic motion of particles in turbulent flows. A bubble in a storm doesn't just react to the local wind; its trajectory is influenced by the "history" of the eddies it has passed through.

Petr Karnakov, Themistoklis P. Sapsis, and colleagues demonstrate in Bubbles in Turbulent Flows: Data-driven, kinematic models with memory terms that neural networks can be trained to "remember" these flow conditions. Similarly, Xander de Wit and Federico Toschi use the Mori-Zwanzig formalism—a concept from statistical mechanics—to show that these complex, hidden interactions manifest as memory effects. By using data-driven models to capture these history-dependent correlations, researchers can predict long-term turbulent statistics far more accurately than models that only look at the "now."

Why Memory Matters for Failure

Understanding memory is critical for predicting when a material will fail. Soft glassy materials, like foams or gels, store a "directional signature" of their history. Elisabeth Agoritsas and Roberto Benzi point out in Ductile-to-brittle transition and yielding in soft amorphous materials that this rheological memory is so persistent that standard methods for "resetting" a material (like high-speed stirring) often fail to erase it.

This persistent memory can lead to counter-intuitive results. For instance, in Anomalous Shear Stress Growth During Relaxation of a Soft Glass, researchers found that internal stress can actually increase while a material is supposedly relaxing, simply because different internal regions "remember" and relax at different rates.

The Future: Generative Rheology

The field is now moving toward "neural surrogates" that can learn these complex histories from data. Maedeh Saberi and Safa Jamali introduced RheOFormer, a generative transformer model. Much like how a language model predicts the next word based on the history of a sentence, RheOFormer uses attention mechanisms to predict the next state of a fluid based on its history of deformation.

By treating fluid dynamics as a sequence-modeling problem, these researchers are bridging the gap between the rigorous physics of Wenjie Deng and the practical needs of industrial manufacturing and soft robotics.

Papers in this version

History & citationVersion 1 · 2026-10-04

Split into other directions · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_lagrangian_modeling_of_history_dependent_497047c0,
  title = {Lagrangian Modeling of History Dependent Fluid Memory},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/lagrangian-modeling-of-history-dependent-fluid-memory/rdr_497047c0752d4840a861023fc0d08713},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_lagrangian_modeling_of_history_dependent_497047c0_v1,
  title = {Lagrangian Modeling of History Dependent Fluid Memory},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/lagrangian-modeling-of-history-dependent-fluid-memory/rdr_497047c0752d4840a861023fc0d08713/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}