Non-Associative Flow

When materials are pushed beyond their elastic limits, they deform permanently—a behavior known as plasticity. A central question in mechanics is predicting the "direction" of this plastic flow. For decades, the elegant theory of associative flow, where the deformation rule is directly derived from the yield criterion, has successfully described ductile metals. This model is underpinned by principles of stability and maximum energy dissipation, predicting that metals deform at a constant volume. However, this tidy picture breaks down when applied to a vast class of "frictional" materials, such as soil, rock, and concrete. For these materials, the associative rule incorrectly predicts their volume change during deformation, creating a significant gap between theory and experimental reality. This discrepancy necessitates a more nuanced approach to accurately model the world around us. This article delves into the solution: the theory of non-associative flow. In the following chapters, we will explore its principles, mechanisms, applications, and interdisciplinary connections, revealing the trade-offs involved in sacrificing simplicity for accuracy and demonstrating its critical importance in fields from geomechanics to computational engineering.

Imagine the state of stress within a material as a point on a map. On this map, there is a boundary, a kind of "sound barrier" for solids. This is the ​​yield surface​​. As long as the stress point stays within this boundary, the material behaves elastically, like a perfect spring; if you unload it, it snaps right back to its original shape. But if you push the stress state to the edge of this boundary, something profound happens. The material yields. It begins to flow, to deform irreversibly. Bend a paperclip past its limit, and it stays bent. This permanent deformation is what we call ​​plasticity​​.

The most fundamental question we can ask is this: when the material starts to flow, in which "direction" does it deform? The answer to this question is a journey that reveals a beautiful interplay between mathematical elegance, physical principles, and the messy reality of the materials that build our world.

A remarkably elegant and powerful idea, known as the ​​principle of maximum plastic dissipation​​, provides a natural first answer. It suggests that of all the possible ways a material could deform, it chooses the path that dissipates energy at the highest possible rate. This principle, closely related to ​​Drucker's stability postulate​​, isn't just an arbitrary guess; it's a statement about the inherent stability of a material. A material that follows this rule won't spontaneously collapse or release energy when prodded.

The geometric consequence of this principle is wonderfully simple. It dictates that the "direction" of plastic flow—represented by a vector of plastic strain rates—must be perpendicular, or ​​normal​​, to the yield surface at the current stress point. This is the celebrated ​​normality rule​​.

Since the rule for how the material flows is derived directly from the function defining the yield surface, $f$, we call this an ​​associative flow rule​​. In this elegant picture, the yield function $f$ wears two hats: it defines the boundary of elastic behavior ($f=0$), and it also serves as a "potential," let's call it $g$, that dictates the direction of flow. For associative flow, we simply have $g=f$.

For a vast and crucial class of materials—the ductile metals that form our bridges, cars, and airplanes—this associative picture is a monumental success. A key characteristic of most metals is that their tendency to yield is largely unaffected by hydrostatic pressure. You can squeeze a block of steel from all sides, and it won't yield any more easily. Its plasticity is a response to shear.

The ​​von Mises​​, or ​​$J_2$ plasticity theory​​, captures this behavior with a beautifully simple yield surface: an infinitely long cylinder in the abstract space of stresses. What happens when we apply our associative normality rule to this cylinder? The normal vector at any point on the cylinder's surface always points purely radially outward, with no component along the cylinder's axis (the hydrostatic axis).

The stunning consequence is that the plastic strain rate is predicted to have no volumetric component. This means that when a metal deforms plastically, its volume does not change. Plastic flow is ​​isochoric​​, or volume-preserving. Squeeze a metal bar in one direction, and it will expand in the other two to keep its total volume constant. This prediction aligns almost perfectly with countless experiments. It is a triumph of continuum mechanics, where a simple rule born from a deep principle predicts a key material behavior with stunning accuracy.

But nature's repertoire is far richer. Let's turn our gaze from a steel beam to a pile of sand, a block of concrete, or the rock deep within the earth. These materials are different. Their strength depends critically on how much pressure they are under. The more you squeeze a handful of sand, the stronger it becomes in shear. These are known as ​​frictional materials​​.

To model them, the cylindrical yield surface of von Mises is no longer adequate. We need a yield surface shaped more like a cone, as described by the ​​Drucker-Prager​​ or ​​Mohr-Coulomb​​ criteria. The slope of this cone is directly related to the material's internal ​​friction angle​​, denoted by $\phi$.

It is here that the beautiful associative story starts to unravel. If we faithfully apply the normality rule, the plastic flow must be normal to this conical surface. Because the cone's walls are sloped, the normal vector points not only outwards in shear but also along the pressure axis. The associative rule predicts that as a frictional material is sheared, it must expand in volume. This phenomenon is called ​​dilatancy​​, and it is entirely real! Watch a densely packed bag of sand being squashed, and you will see it bulge and expand.

The problem, however, is a quantitative one. The associative rule inexorably links the amount of dilation to the internal friction; a higher friction angle must lead to a higher rate of volume expansion. But when we go to the laboratory and test these materials, we find that while they do dilate, they do so far less than the associative theory predicts. The theory is beautiful, but for these materials, it is wrong.

How do we resolve this crisis? We must perform what you might call a "great decoupling." We break the assumption that a single function governs both the onset of yielding and the direction of flow.

We introduce a second, distinct function: the ​​plastic potential​​, $g$. The yield function $f$ still defines the boundary for when plasticity starts ($f=0$), but the direction of flow is now governed by the normal to the surface of the plastic potential, $g$. The flow rule becomes $\dot{\boldsymbol{\varepsilon}}^p \propto \partial g / \partial \boldsymbol{\sigma}$. When $g$ and $f$ are not the same function, the flow is called ​​non-associative​​.

This is an act of inspired pragmatism. It allows us to tailor the plastic potential $g$ to match the observed flow behavior, while the yield function $f$ remains dedicated to matching the material's strength. For our frictional materials, we can now choose a plastic potential $g$ that is also a cone, but one with a gentler slope governed by a separate ​​dilation angle​​, $\psi$, which is smaller than the friction angle, $\phi$.

By making $\psi \phi$, the normal to the plastic potential surface is "tilted" differently from the normal to the yield surface. This directly reduces the predicted volumetric expansion, bringing the model's predictions for materials from dense sand to cohesive clays back into alignment with experimental reality. We have fixed our theory by sacrificing the elegant but restrictive assumption of associativity.

This clever solution, however, is not a free lunch. By decoupling yield and flow, we step outside the tidy framework of associativity and must face the consequences.

First, the ​​loss of a beautiful principle​​. A non-associative model, by definition, violates the principle of maximum plastic dissipation. This also means that the material's stability, as defined by Drucker, is no longer automatically guaranteed by the constitutive law. The model could, under certain conditions, become unstable in ways that were impossible in an associative world.

Second, and more alarmingly, is the ​​danger of thermodynamic heresy​​. The second law of thermodynamics, codified in the ​​Clausius-Duhem inequality​​, is an inviolable law of physics. It insists that a material cannot create energy out of thin air; the internal dissipation of energy during a process like plastic flow must be non-negative. With associative flow, this was guaranteed. With non-associative flow, it is not. A carelessly chosen plastic potential $g$ can lead to a model that predicts negative dissipation—a physical impossibility. This imposes strict limits on how much the plastic potential $g$ can deviate from the yield function $f$. To build robust models, scientists have developed more advanced frameworks, such as that of ​​Generalized Standard Materials​​, which ensure thermodynamic consistency by considering dissipation in a higher-dimensional space of forces and fluxes.

Finally, there are ​​computational headaches​​. In the world of computer simulations that power modern engineering, such as the Finite Element Method, associative models generate a symmetric system of equations (specifically, a symmetric ​​consistent tangent operator​​). This mathematical symmetry is a godsend for numerical algorithms, allowing for incredibly efficient and robust solutions. Non-associative flow breaks this symmetry. The resulting numerical problem becomes significantly harder and more expensive to solve.

The story of non-associative flow is a perfect miniature of the scientific process itself. We begin with an elegant, unifying theory, test it against nature, find its limits, and then adapt it to be more truthful. But this adaptation is not without cost, forcing us to look deeper at the foundational principles of stability, thermodynamics, and computation. It is a beautiful lesson that in the quest to describe our world, there is a constant, fascinating trade-off between simplicity, accuracy, and complexity.

Now that we have grappled with the principles of non-associative flow, this notion that a material can yield according to one rule but deform plastically according to another, you might be asking a very sensible question: "So what?" Is this just a mathematical subtlety, a peculiar curio for theoreticians to ponder? The answer is a resounding no. This departure from the simple, elegant picture of associated plasticity is not a mere abstraction. It is a fundamental truth about the world around us, a truth written in the behavior of soil, the failure of rocks, the properties of modern polymers, and even in the very code that engineers write to design our world's infrastructure. To ignore non-associativity is to misunderstand how many materials actually work. Let’s embark on a journey to see where this "unruly" behavior shows up and why it matters so profoundly.

Our first stop is right under our feet. Consider a pile of sand, a bag of gravel, or the soil supporting a skyscraper. These are granular materials. Imagine shearing such a material—that is, sliding one layer over another. The individual grains are not smooth spheres that can just glide past each other. They are jagged and interlocked. To shear the material, the grains in one layer must ride up and over the grains in the layer below. This forced "riding up" causes the entire material to expand in volume. This phenomenon is called ​​dilatancy​​, and it is a hallmark of dense granular materials.

Herein lies the essential seed of non-associativity. What determines when the soil begins to yield? It is primarily the friction between the grains. When the shear stress becomes large enough to overcome this friction, the grains start to slip. This is governed by the material's ​​friction angle​​, which you might call $\phi$. But what determines how much the material expands as it shears? This is a question of kinematics, of the geometry of the grains and how they are packed. It has a different physical origin from friction. We can capture this volumetric change with a ​​dilatancy angle​​, let's call it $\psi$.

In a real soil, there is no a priori reason why the condition for overcoming friction ($\phi$) should be identical to the condition governing the geometry of expansion ($\psi$). In fact, experiments overwhelmingly show they are different. Therefore, the yield function, which depends on $\phi$, cannot be the same as the plastic potential, which dictates the flow and depends on $\psi$. This is the quintessential example of non-associative flow. The rule for "go" (yield) is different from the rule for "how to go" (flow). Setting $\psi=0$ describes a material that shears at a constant volume (isochoric flow), while $\psi > 0$ describes dilation. By separating these two angles, engineers can accurately model the complex pressure-sensitive strength and volume changes that are critical for predicting landslides, designing stable foundations, and constructing tunnels. The earth, in a sense, "breathes" as it deforms, and non-associativity gives us the language to describe it.

We don't need to dig in the dirt to find this behavior. Modern materials like glassy polymers also play by these rules. Imagine taking a piece of hard plastic and subjecting it to immense hydrostatic pressure, then testing its yield strength. By carefully measuring the yield stress at different pressures and independently measuring how the material's volume changes as it starts to yield plastically, we can test our theory. Hypothetical experiments of this kind, based on real material behavior, consistently show that the parameter controlling the pressure-sensitivity of the yield strength is different from the parameter controlling the plastic volume change. The data forces our hand: to be faithful to reality, we must adopt a non-associative framework.

When a block of soil or rock is compressed to its limit, it doesn't just uniformly bulge. Instead, it often fails along a remarkably narrow band, a distinct plane of intense shear. This process is called ​​strain localization​​, and these failure zones are known as shear bands. You can see them in the fault lines after an earthquake or in the failure patterns of concrete under compression. The orientation of these bands is not random; it is a deep clue about the physics of the material.

And here, non-associativity plays a starring role. Imagine again our material at the point of failure. There is a conflict within. On one hand, the state of stress "wants" to cause slip on the plane where the shear-to-normal stress ratio is at its critical value—the plane of maximum stress obliquity. The orientation of this plane is dictated purely by the yield criterion, governed by our old friend, the friction angle $\phi$. Let's call this the "static" preference.

On the other hand, for a shear band to form, there must be a kinematically compatible way for deformation to occur. The material can't just deform in any way it pleases. The direction of plastic flow, governed by the plastic potential and its dilatancy angle $\psi$, dictates the planes along which deformation can occur without requiring any extension. Let's call this the "kinematic" preference.

When the flow is associative ($\phi=\psi$), a wonderful thing happens: the static and kinematic preferences align perfectly. The plane of maximum stress ratio is also a plane of zero extension. But in a non-associative material ($\phi \neq \psi$), these two planes are different. The material is torn between what is statically favorable and what is kinematically possible. What does it do? It compromises. The rigorous mathematical theory of material instability (the acoustic tensor criterion) shows that the shear band forms at an orientation that is, beautifully, the arithmetic mean of the angles predicted by the static and kinematic criteria alone. For a plane strain condition, the angle of the shear band, $\beta$, with respect to the direction of major principal stress, is given by:

This is a remarkable result. It tells us that the very angle of a geological fault or a crack in a dam is a physical manifestation of the material's non-associative nature. That "scar" is the material's solution to an internal conflict.

So, we have a theory that describes real materials and their failure. The next logical step is to put this theory into a computer to simulate and design things—a tunnel, a car part, an airplane wing. We do this using powerful tools like the Finite Element Method (FEM). This method breaks a complex structure down into small pieces (elements) and solves the equations of force balance iteratively. At the heart of the solution process—for a nonlinear problem like plasticity—is a giant matrix called the ​​tangent stiffness matrix​​. You can think of it as the Jacobian of the system; it tells the computer how the internal forces in the structure will change in response to a small change in displacements. It provides the "map" for the solver to navigate toward the correct solution.

Now, for materials with associative flow, a profound and beautiful property exists. The entire set of rules for the material's response—elasticity, yielding, and plastic flow—can be derived from a single, scalar-valued potential function, much like how forces in a gravitational field can be derived from a gravitational potential energy function. This "variational" structure is not just mathematically elegant; it guarantees that the resulting tangent stiffness matrix is ​​symmetric​​.

A symmetric matrix is a wonderful thing to have. It's computationally cheaper to store and factorize, and it allows us to use incredibly efficient and robust iterative solvers, with the Conjugate Gradient (CG) method being the celebrated workhorse.

Non-associativity destroys this beautiful picture. By decoupling the flow rule ($g$) from the yield function ($f$), we lose the existence of a single overarching potential. The system of equations can no longer be seen as the gradient of a single scalar function. And the consequence of this is immediate and severe: the consistent tangent stiffness-like operator, $\mathbb{C}^{ep}$, becomes ​​non-symmetric​​.

When we assemble the global stiffness matrix for our entire structure, this non-symmetry at the material level infects the global matrix, making it non-symmetric as well. What does this mean in practice? It means we must throw our favorite tool, the CG method, out the window; it simply doesn't work for non-symmetric systems. We are forced to use more general, more complex, and often significantly more computationally expensive solvers like the Generalized Minimal Residual (GMRES) or Bi-Conjugate Gradient Stabilized (BiCGStab) methods. If we use a direct solver, we can't use a fast Cholesky factorization; we must resort to a more general LU factorization.

Engineers might be tempted to cheat. They could say, "Let's use the real non-associative rule to update the stresses, but for the tangent matrix, let's just pretend the flow is associative to get a symmetric matrix and use our fast solver." This is a common strategy, but it comes at a cost. The method is no longer a true Newton-Raphson iteration; it becomes a "quasi-Newton" method. And by using an incorrect Jacobian, one generally sacrifices the prized quadratic convergence of Newton's method, leading to many more iterations, and often requires extra machinery to prevent the solver from diverging entirely. The ghost of non-associativity haunts our computations, forcing a choice between accuracy, speed, and robustness.

Finally, let's ascend to one of the most elegant concepts in structural mechanics: the ​​shakedown theorems​​. Imagine a structure, like a bridge or an offshore platform, subjected to repeated, cyclical loading (traffic, waves, wind). Will it eventually "settle down" and respond elastically to these loads, or will plastic deformation accumulate with each cycle, leading to failure by "ratcheting" or low-cycle fatigue?

For materials with associative plasticity, classical theory provides two powerful theorems that give a definitive answer. Melan's static theorem gives an upper bound on a safe load, based on finding a residual stress field that keeps the total stress always within the yield limit. Koiter's kinematic theorem gives a lower bound, based on analyzing possible failure mechanisms. For "well-behaved" associative materials, these two bounds coincide. There is a single, crisp shakedown limit. This beautiful equality is a statement of strong duality.

This elegant duality relies on a cornerstone of associated plasticity: the principle of maximum plastic dissipation. This principle states that, for a given yield surface, the actual plastic dissipation is maximized. Non-associative flow violates this very principle. The actual plastic work being done can be less (or in some cases, more) than what would be predicted by an associative model with the same yield surface.

What is the consequence? Melan's theorem, which cares only about equilibrium and the yield surface, remains perfectly valid. It continues to provide a safe, conservative upper bound on the loads a structure can withstand. It doesn't care about the flow rule. But Koiter's theorem, whose proof is intimately tied to the flow rule and the maximum dissipation principle, is invalidated. It no longer guarantees a lower bound on the shakedown limit. A "duality gap" can open up between the static and kinematic predictions.

The practical implication is a loss of certainty. We lose the comfort of a single, definitive shakedown limit. Our analysis of structures under cyclic loading, critical for long-term safety, becomes more complex and often more conservative, because the beautiful symmetry of the theory has been broken by the material's refusal to flow in the "expected" direction.

From the mundane expansion of shearing soil to the highly abstract foundations of structural stability, the decision of a material to follow one rule for yielding and another for flowing sends ripples through physics and engineering. It is a compelling reminder that the rich complexity of the real world often defies our initial desire for the simplest possible model. The true beauty of physics is not just in identifying elegant symmetries, but also in understanding the profound and intricate consequences that unfold when those symmetries are—for very good physical reasons—broken.