🔬 condensed matter

A thermomechanical framework for strongly nonlocal continua

This paper establishes a consistent thermomechanical framework for strongly nonlocal continua by deriving integral-type balance equations, conservation laws, and constitutive structures via the principle of virtual power to model size-dependent localization phenomena and prevent mesh dependence.

Original authors: Matthew R. Kuhn

Published 2026-09-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Matthew R. Kuhn

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Materials science often relies on a simplifying assumption: that what happens at a single point inside a solid depends only on the conditions at that exact spot. Imagine a crowd of people; in the traditional view, a person's reaction is determined solely by their own immediate surroundings, ignoring the pressure or movement of those standing a few feet away. This "local" perspective works well for many everyday situations, but it breaks down when materials begin to fail. When a metal bar stretches until it snaps, or when concrete cracks under pressure, the damage does not happen at a single, infinitesimal point. Instead, it concentrates in a specific zone with a measurable width. Traditional models struggle to predict the size of this zone, often suggesting the damage should be infinitely thin, a result that contradicts physical reality and causes computer simulations to fail whenever the grid used to model the material is changed.

To solve this, researchers have long looked toward "nonlocal" theories, which acknowledge that a point in a material is influenced by its neighbors. However, previous attempts to incorporate this influence often required complex, abstract additions to the equations of motion, introducing difficult concepts like "couple stresses" that are hard to visualize or measure. A new study by Matthew R. Kuhn offers a cleaner, more consistent way to describe these interactions. By building a framework from the ground up using the principle of virtual power—a method that considers how a system would respond to tiny, imagined movements—the paper establishes a set of rules where the influence of neighboring points is woven directly into the fundamental laws of motion and energy, rather than tacked on as an afterthought.

The core of this work is a framework that treats the material as a continuous whole where every point feels the average effect of its surroundings. The researchers developed a mathematical "kernel," a function that acts like a weighted average, determining how much influence a neighbor has on a central point. This influence fades with distance, but it ensures that the behavior at any location is a blend of local conditions and the state of the nearby region. Crucially, the paper derives three distinct sources of this nonlocal behavior: the balance of forces (how the material moves), the conservation of energy and entropy (how heat and disorder behave), and the material's own internal structure (how it stretches and deforms). Unlike older approaches that might simply average the strain or temperature after the fact, this framework averages the energy itself. This distinction is vital; it ensures that the laws of thermodynamics hold true without forcing the material to behave in ways that violate the second law of thermodynamics, which dictates that energy dissipation must always be positive.

The study demonstrates that this approach successfully captures the "characteristic size" of deformation patterns. In a classic test case, the researchers modeled a metal bar with a slightly weaker section in the middle. When the bar was stretched, the damage concentrated in the center, but the width of this damaged zone was not arbitrary. Instead, the width was determined by the range of the influence function—the distance over which the material "feels" its neighbors. The simulation showed that whether the bar was short or long, the width of the localized damage zone remained consistent, governed by the material's internal length scale rather than the arbitrary grid used in the computer model. This resolves a long-standing problem where changing the resolution of a simulation would change the predicted outcome, a flaw known as mesh dependence.

Furthermore, the paper clarifies how forces and heat flow across boundaries in these nonlocal materials. In a standard local model, the force pushing on a surface is balanced strictly by the stress right at that surface. In this new framework, the balance is more nuanced: the force at a surface point is balanced by the stress and body forces in the entire neighborhood surrounding that point. This allows the model to handle rough or irregular boundaries where a clear "normal" direction might not exist, a common issue in real-world engineering that often breaks traditional models. The researchers also showed that for slow, steady processes, the complex equations of this nonlocal theory simplify to look very much like the familiar equations used in standard engineering, providing a bridge between the new theory and established practice.

The findings are not merely theoretical; they are demonstrated through a specific example of an elastoplastic bar, a material that stretches and then permanently deforms. The simulation revealed that the peak strength of the bar was slightly higher than the strength of its weakest section alone, because the stronger surrounding material helped support the weak spot through the nonlocal averaging. As the bar continued to stretch, the rate at which it lost strength depended on the length of the bar, a phenomenon that local models cannot capture. The study confirms that by averaging the free energy function over a neighborhood, rather than just averaging the strain, the model produces physically realistic results that are independent of the numerical grid used to solve them.

This work provides a robust, thermodynamically consistent foundation for modeling materials that exhibit size-dependent behavior. It offers a way to predict how cracks form, how damage spreads, and how materials fail without relying on artificial fixes or complex, hard-to-interpret stress terms. By treating the influence of neighboring points as a fundamental part of the material's physics, the framework allows engineers and scientists to simulate failure with a level of reliability that was previously out of reach, ensuring that the predicted size of a fracture zone is a property of the material itself, not an artifact of the computer model.

Technical Summary: A Thermomechanical Framework for Strongly Nonlocal Continua

Problem Statement
The paper addresses the limitations of classical local continuum mechanics, where stress at a point depends solely on the deformation gradient at that same point. While weakly nonlocal theories (e.g., higher-gradient models) relax this by including higher-order gradients, they often require the introduction of abstruse higher-order stresses and generalized tractions, complicating boundary conditions and numerical implementation. Strongly nonlocal theories, where a point's response depends on the state of its entire neighborhood, offer a more nuanced description of material behavior, particularly for phenomena with intrinsic length scales (e.g., strain localization, shear bands) and for regularizing numerical solutions to prevent mesh dependence. However, a consistent thermomechanical framework for such media—integrating balance laws, thermodynamic principles, and constitutive relations without relying on local approximations—has been lacking. The paper seeks to develop a coherent, integral-type nonlocal framework that avoids the complexities of higher-order stress tensors while rigorously adhering to the laws of thermodynamics.

Methodology
The framework is constructed using the principle of virtual power, applied to a broad class of virtual velocities (both mechanical and thermal). This approach, rooted in the works of Piola, Hellinger, and Germain, treats force as an emergent dual of motion rather than a primitive object.

Virtual Power Formalism: The author defines a Hilbert space of virtual mechanical velocities ( $\hat{v}$ ) and virtual thermal velocities ( $\hat{\theta}$ ). A linear integral operator $L$ , utilizing a kernel function (mollifier) $\phi(x, x')$ , maps these velocities to a space of gradients. This operator replaces the Dirac delta function used in local theories, effectively averaging quantities over a neighborhood.
Derivation of Balance Laws: By applying the postulates of virtual power (equating external and internal virtual power to inertial power, and requiring internal power to vanish for rigid-body motions), the author derives:
- Momentum Balance: Unlike local theories where stress and traction are linked by separate equations ( $\nabla \cdot \mathbf{T} + \mathbf{b} = \rho \dot{\mathbf{v}}$ and $\mathbf{t} = \mathbf{T}\mathbf{n}$ ), the nonlocal balance collapses into a single integral condition. This condition governs both the interior and the boundary, implying that traction at a point is balanced by the averaged stress and body forces in the surrounding neighborhood.
- Thermodynamic Laws: The first and second laws are derived by transitioning from virtual to actual velocities. The author introduces a "hidden" rate of kinetic energy ( $\dot{R}_K$ ) representing meso-scale fluctuations, which is assumed negligible for quasi-static processes.
Constitutive Framework: Following the Coleman-Noll and Green-Naghdi approaches, the author defines the Helmholtz free energy $\Psi$ as a spatial average of a local energy function over the kernel's support. Crucially, they derive stress, entropy, and dissipative forces as averaged derivatives of this free energy with respect to strain, temperature, and internal variables. The paper explicitly cautions against constitutive models that simply average the state variables (strain, temperature) before applying the energy function, demonstrating that such an approach leads to thermodynamic inconsistencies.

Key Contributions

Unified Nonlocal Balance: The derivation of a single, integral-based balance equation that simultaneously satisfies equilibrium in the interior and on the boundary, even for non-smooth surfaces where normal vectors are ill-defined. This eliminates the need for higher-order stress tensors.
Thermodynamic Consistency: A rigorous derivation of the second law (Clausius-Duhem inequality) for strongly nonlocal media, establishing that the dissipation rate is governed by the nonlocal average of internal entropy production.
Constitutive Distinction: A clear theoretical distinction between averaging the energy function (thermodynamically consistent) versus averaging the state variables (thermodynamically inconsistent). The paper demonstrates that the former is required to correctly define stress and entropy in nonlocal media.
Kernel Constraints: Identification of necessary properties for the kernel function $\phi$ (e.g., symmetry, normalization, compact support) to ensure objectivity, symmetry of the stress tensor, and the recovery of classical local balance equations under specific smoothness conditions.

Results
The framework is applied to a one-dimensional example: a prismatic bar with a central weakened section undergoing elasto-plastic deformation with softening.

Regularization: The nonlocal model successfully prevents the mesh-dependence and ill-posedness observed in local models when softening occurs. The localization zone acquires a characteristic width determined by the kernel's support radius, rather than collapsing to a single element.
Size Independence: Simulations of bars with different total lengths (0.5L, L, 2L) but identical weakened sections and kernel widths show that the width of the strain localization zone remains constant (approximately four times the kernel width). This confirms that the nonlocal framework imparts an intrinsic length scale to the deformation pattern.
Numerical Robustness: The results are shown to be insensitive to grid fineness (tested with N=40, 100, 200), validating the regularization capability of the integral formulation.

Significance and Claims
The paper claims to provide a consistent thermomechanical framework where nonlocality is intrinsic to the balance laws, conservation principles, and constitutive structure, rather than being an ad-hoc addition to local equations.

It asserts that by viewing force as a dual of motion via virtual power, one can derive nonlocal balance equations that naturally handle non-smooth boundaries and indistinct normal directions.
It emphasizes that the "strongly nonlocal" hypothesis, where behavior depends on the neighborhood, offers a practical alternative to higher-gradient theories by avoiding the need for generalized stresses and complex boundary conditions.
The author modestly claims that while the framework assumes quasi-static processes (neglecting residual meso-scale kinetic energy), it successfully captures the essential physics of strain localization and provides a robust numerical tool for materials exhibiting softening and damage. The work serves as a foundational step for modeling generalized continua where long-range internal interactions are significant.

More like this