Non-Associated Flow Rule

When materials are pushed beyond their elastic limits, they undergo permanent, irreversible deformation—a phenomenon known as plasticity. While this property allows us to shape metals, it also governs the failure of structures. To predict this behavior, engineers and physicists need rules that define not only when a material will yield but also the direction in which it will deform. The simplest theory, an associated flow rule, elegantly links these two aspects and works perfectly for metals. However, it fails to describe the complex behavior of other critical materials, such as soil, sand, and rock.

This article addresses the knowledge gap by exploring the non-associated flow rule, a more versatile but complex alternative. Across two chapters, you will gain a comprehensive understanding of this crucial concept. The first chapter, "Principles and Mechanisms," delves into the fundamental theory, contrasting the mathematical elegance of associated flow with the practical necessity of non-associated flow and exploring its profound implications for material stability. Following this, the "Applications and Interdisciplinary Connections" chapter grounds the theory in practice, showcasing its vital role in geomechanics and metal plasticity, and examining the significant challenges it poses for computational analysis and structural safety assessment.

Figure 1: An illustration of the yield surface in stress space. For an associated material, the plastic flow direction ($\dot{\boldsymbol{\varepsilon}}^p$) is normal (perpendicular) to the yield surface ($f=0$). For a non-associated material, the flow is normal to a different surface, the plastic potential ($g=\text{const.}$).

Imagine a world without permanent change. You bend a paperclip, and it springs back perfectly. You dent a car fender, and the dent vanishes. This is the world of elasticity, a world that is reversible and tidy. But our world is not like that. Bend a paperclip too far, and it stays bent. This permanent, irreversible deformation is called ​​plasticity​​, and it is what allows us to shape metals, but it is also what causes structures to fail.

To understand plasticity, we need a "rule of the road." We need to know when a material stops being merely elastic and starts to flow like a thick fluid. And once it starts flowing, we need to know which way it will flow. The journey to discover these rules takes us through some of the most elegant and surprisingly subtle ideas in physics.

Let's first build an abstract map to guide us. Instead of a map of cities and roads, imagine a ​​stress space​​. Each point on this map represents a state of stress—a specific combination of pushing, pulling, and twisting that a small piece of material can experience. Somewhere on this map, there is a boundary, a line or a surface. Inside this boundary is the "elastic zone." As long as the stress state stays inside, the material behaves like a perfect spring. But if we push the material hard enough to reach this boundary, it yields. This boundary is aptly named the ​​yield surface​​, and it is described mathematically by a ​​yield function​​, which we can call $f$.

Now for the crucial question: once we're on this boundary and we push just a little bit more, causing the material to flow plastically, in what direction does it deform? The simplest and most beautiful answer nature could have provided is the rule of ​​normality​​. This rule states that the "vector" of plastic deformation is always perpendicular, or ​​normal​​, to the yield surface at the current point of stress.

Having journeyed through the principles and mechanisms of plasticity, we now arrive at a crucial question: where does this theory meet the real world? We have seen that the state of a material—whether it holds its shape or begins to flow—is governed by a yield function, $f$. We have also seen that the character of that flow—its direction and nature—is dictated by a plastic potential, $g$. The simplest and most elegant assumption is that these two functions are one and the same, the case of associated flow. But nature, in its infinite subtlety, is not always so simple.

What happens when a material's flow does not follow the rules prescribed by its own yield surface? This is the realm of the ​​non-associated flow rule​​, where $g \neq f$. It may seem like a mere mathematical complication, but as we shall see, it is a key that unlocks a deeper understanding of real materials and poses profound challenges in engineering and science. The decision to separate the "when" from the "how" of plastic flow is not an arbitrary choice; it is often a direct response to clues from the physical world.

Imagine we are material scientists studying a new metallic alloy. We perform a simple tension test and calibrate a beautiful plasticity model based on an associated flow rule. It works perfectly, predicting the material's response to stretching and compression. We are satisfied. But then, we perform a more complex experiment, subjecting a sheet of the metal to biaxial tension—pulling it in two directions at once. Two strange things happen. First, the material expands slightly in thickness, a phenomenon our model, based on a pressure-insensitive yield surface, forbade. Second, the direction of plastic flow in the plane is skewed; it is not quite perpendicular to the yield surface we so carefully measured. Our elegant model has failed. This discrepancy is not a failure of plasticity theory itself, but a signal from the material that its inner workings are more complex. It is a direct piece of evidence whispering the secret of non-associated flow.

Perhaps the most intuitive and widespread application of non-associated flow is in geomechanics—the study of soil, rock, and concrete. Consider a handful of sand or a pile of gravel. When you shear it, the individual grains must roll and slide up and over one another. This forces the assembly to expand in volume, a remarkable phenomenon known as ​​dilatancy​​.

Now, think about this in the context of our plasticity framework. The yielding of soil is governed by internal friction and cohesion, concepts captured in models like the Mohr-Coulomb or Drucker-Prager criteria. The yield function, $f$, is therefore primarily a function of a material's ​​friction angle​​, often denoted $\phi$. This angle determines the stress at which the material begins to slip and fail.

However, the amount of volume expansion—the dilatancy—is a separate physical process related to the geometry of the grains and their packing. It would be a remarkable coincidence if the physics of inter-particle friction were perfectly coupled to the physics of this geometric rearrangement. They are not. A non-associated flow rule provides the perfect tool to decouple them.

We can construct a model where the yield function $f$ uses the friction angle $\phi$, but the plastic potential $g$ uses a separate, independent ​​dilatancy angle​​, $\psi$. The plastic volumetric strain rate, $\dot{\varepsilon}_v^p$, is then directly proportional to a function of this dilatancy angle.

• If $\psi > 0$, the model predicts expansion, correctly capturing the dilatant behavior of dense sands and overconsolidated clays.
• If $\psi = 0$, the plastic flow becomes volume-preserving, or isochoric. This is a good approximation for clays in a critical state.
• If $\psi 0$, the model can even predict plastic compaction, describing the behavior of loose sands.

This freedom to independently tune friction and dilatancy is not just an academic exercise; it is essential for the realistic analysis of foundations, tunnels, dams, and the prediction of landslides. It allows engineers to build models that respect the distinct physical processes at play in the ground beneath our feet.

In stark contrast to soils, the plastic deformation of most metals at moderate strains is almost perfectly volume-preserving. This is because plasticity in crystalline metals is primarily caused by the sliding of atomic planes, a process called dislocation glide, which rearranges atoms without changing the overall volume.

This physical reality is beautifully captured by the associated flow rule when combined with pressure-independent yield criteria like those of von Mises or Tresca. Since these yield functions, $f$, depend only on the deviatoric (shape-changing) part of the stress tensor, their gradients are also purely deviatoric. An associated flow rule ($g=f$) then automatically enforces that the plastic strain rate is purely deviatoric, meaning the plastic volumetric strain rate is zero. Here, the simplicity of $g=f$ is a direct reflection of the underlying physics.

However, the story does not end there. Under extreme loading, metals can begin to exhibit volume change, often due to the nucleation and growth of microscopic voids. This is the precursor to ductile fracture. How can we model this? Once again, the non-associated framework offers an elegant solution. We can retain the pressure-insensitive von Mises yield function, $f$, because the onset of dislocation slip is still largely independent of pressure. But we can introduce a new plastic potential, $g$, that includes a term dependent on the hydrostatic pressure. This allows the model to predict an increase in volume during plastic flow, simulating the opening of voids, without altering the fundamental condition for yielding.

Furthermore, manufacturing processes like rolling can impart a directional texture to metals, making them anisotropic. In such cases, the principal directions of plastic strain may no longer align with the principal directions of the applied stress. While this can be modeled with sophisticated anisotropic yield functions under an associated flow rule, the non-associated framework provides additional latitude to capture these complex directional behaviors, offering a powerful way to separate the description of the yield strength from the direction of flow.

The increased fidelity of non-associated models comes at a price—a computational one. In modern engineering, complex designs are tested using computer simulations, most notably the Finite Element Method (FEM). These methods solve vast systems of nonlinear equations to predict a structure's behavior. The engine of the most powerful solvers is the Newton-Raphson method, which relies on the ​​tangent stiffness matrix​​, $\mathbf{K}$. This matrix describes how the internal forces in the structure respond to infinitesimal changes in deformation.

For materials governed by an associated flow rule, the underlying mathematical structure is deeply elegant. The constitutive equations can be derived from a single potential, which ensures that the material's tangent stiffness tensor, $\mathbb{C}^{ep}$, is ​​symmetric​​. When assembled, this results in a global tangent stiffness matrix $\mathbf{K}$ that is also symmetric. This symmetry is a tremendous gift. Symmetric linear systems can be solved with extraordinarily fast and memory-efficient algorithms.

However, when we introduce a non-associated flow rule, this beautiful symmetry is broken. The very fact that the direction of flow ($\partial g / \partial \boldsymbol{\sigma}$) is not aligned with the normal to the yield surface ($\partial f / \partial \boldsymbol{\sigma}$) leads inexorably to a ​​non-symmetric​​ tangent stiffness tensor $\mathbb{C}^{ep}$ and, consequently, a non-symmetric global matrix $\mathbf{K}$,.

This forces a difficult choice upon the computational engineer:

1. Use the exact, non-symmetric tangent matrix. This preserves the celebrated quadratic convergence of the Newton-Raphson method, meaning the solution is found in very few iterations. However, it requires the use of more complex, slower, and more memory-intensive non-symmetric linear solvers (like GMRES or a direct LU factorization).
2. Use an approximation, for instance, by simply taking the symmetric part of the true tangent matrix. This allows the use of fast symmetric solvers, but because the matrix is no longer the exact Jacobian, the convergence of the algorithm degrades from quadratic to, at best, linear. The simulation will require many more iterations to reach a solution, if it converges at all.

This trade-off between accuracy, speed, and robustness is a central theme in computational mechanics. The choice to use a non-associated model is a choice to embrace a more complex computational reality in exchange for a more faithful description of the physical world.

Perhaps the most profound and startling consequence of non-associativity lies in the classical theory of structural stability, known as ​​limit analysis​​. For decades, engineers have used this elegant theory to calculate the ultimate collapse load of structures. One of its cornerstones is the ​​kinematic (or upper-bound) theorem​​. This theorem allows one to postulate a failure mechanism and calculate a corresponding load. It guarantees that this calculated load is an upper bound to the true collapse load—the structure will fail at or below this load.

The mathematical proof of this powerful theorem rests on a crucial assumption: the principle of maximum plastic dissipation. This principle states that for a given rate of plastic deformation, the actual stress state in the material is one that maximizes the rate of energy dissipation. This principle is mathematically equivalent to the associated flow rule.

What happens if the material follows a non-associated flow rule? The link is broken. The actual stress state no longer maximizes the dissipation for a given plastic strain rate. In fact, the actual dissipation is strictly less than the maximum possible value assumed by the classical theorem,.

The consequence is alarming: the classical upper-bound theorem becomes ​​unsafe​​. It calculates a collapse load based on a fictitious material that is "stronger" (dissipates more energy) than the real non-associated material. The theorem might yield a load prediction that is higher than the true collapse load. Applying this result uncritically could lead an engineer to believe a structure is safe when, in reality, it is poised for catastrophic failure. The non-associated nature of the material has undermined one of the fundamental safety nets of classical structural analysis. This sobering realization demonstrates that non-associativity is not just a detail; it is a feature that can fundamentally alter our predictions about safety and collapse.

Our exploration has revealed the non-associated flow rule to be far more than a mathematical footnote. It is a powerful and necessary concept that appears whenever the "when" of plastic flow is physically distinct from the "how". It allows us to model the dilatancy of soil, the subtle volume changes in damaged metals, and the complex directional response of anisotropic materials.

This added realism, however, demands a price. It complicates our numerical algorithms by destroying a fundamental symmetry, and it challenges the foundations of classical stability theorems. Yet, in this challenge lies its true value. The non-associated flow rule forces us to confront the beautiful complexity of the material world and to refine our tools to meet it. It represents an elegant separation of concepts that gives our models the flexibility to be not just mathematically consistent, but physically true.