Yielding behavior of glasses under asymmetric cyclic deformation

Monoj Adhikari, Muhittin Mungan, Srikanth SastryarXiv 2022 · DOI · arXiv

TL;DR

Researchers have discovered that glasses behave fundamentally differently when "pushed" in only one direction (asymmetric cyclic shear) compared to standard back-and-forth (symmetric) loading. In well-annealed glasses, they found a unique "pre-yielding" state where the material deforms permanently but doesn't yet flow, ultimately leading to a surprising reversal: under these conditions, high-quality, well-rested glasses actually break sooner than low-quality ones as the system size increases.

What Problem Does This Paper Address?

Predicting how amorphous solids—like metallic glasses, polymers, or even foams—fail under stress is a major challenge in materials science. Most research focuses on "symmetric" cyclic loading (straining the material to $+\gamma$ and then $-\gamma$). However, real-world scenarios, such as the mechanical fatigue of structural components, often involve "asymmetric" cycles (straining from $0$ to $+\gamma$).

The authors investigate how the "history" of a glass (how well it was annealed or "rested") and the size of the sample influence its yielding behavior under these lopsided cycles. They specifically look for the point where a solid stops behaving like a solid and starts flowing like a liquid (yielding).

Methods

The team used Athermal Quasistatic Shear (AQS) simulations. In this approach, the material is deformed in tiny increments, and after each step, the system's energy is minimized to find the new stable configuration. This effectively removes the "noise" of temperature, allowing researchers to see the pure mechanical response of the atomic structure.

They modeled a standard "benchmark" glass: the 80:20 Kob-Andersen binary Lennard-Jones mixture. To understand the impact of preparation, they created samples with different degrees of annealing, ranging from "poorly annealed" (high-energy, unstable) to "well-annealed" (low-energy, stable). They also scaled the system size from 200 atoms up to 64,000 atoms to see how these effects hold up in larger materials.

Datasets

While this is a physics simulation paper rather than a machine learning one, the "data" consists of simulated atomic trajectories and energy states:

  • System Sizes ($N$): 200, 400, 800, 4000, 8000, 32000, and 64000 particles.
  • Annealing Levels ($E_{init}$): Five distinct energy levels representing different cooling rates.
  • Protocol: Asymmetric cycles ($0 \to \gamma_{max} \to 0$).

Key Insights and Ideas

  • The "Symmetric Shift" in Poor Glasses: Poorly annealed glasses are "messy" enough that they can rearrange themselves during the first few cycles. They effectively shift their internal reference point so that an asymmetric $0$ to $0.15$ cycle eventually feels like a symmetric $-0.075$ to $+0.075$ cycle.
  • The Intermediate Regime: Well-annealed glasses are too stable to shift easily. Instead, they enter a strange intermediate regime. As the strain increases, the internal stress reaches a peak and then drops—usually a sign of breaking—but the atoms don't actually start diffusing (flowing). They are stuck in a state of "plasticity without yielding."
  • The Yielding Reversal: In symmetric loading, a better-annealed glass is always stronger (yields at a higher strain). The most shocking finding here is that for large systems under asymmetric loading, this trend flips: well-annealed glasses yield at lower strain amplitudes than poorly annealed ones.
  • Mesostate Validation: The authors showed that a simplified "mesostate model"—which treats the glass as a series of energy wells—can qualitatively replicate these complex results, suggesting the behavior is a fundamental property of the energy landscape of glasses.

Key Results

  • Quantitative Yielding: For poorly annealed glasses, yielding occurs at $\gamma_{max} \approx 0.15$. For well-annealed glasses in small systems, a stress peak appears at $\gamma \approx 0.1$, but true diffusive yielding doesn't happen until $\gamma \approx 0.155$.
  • System Size Scaling: As the number of particles ($N$) increases, the window of the "intermediate regime" narrows. The onset of flow ($\gamma_{diff}$) moves closer to the stress peak ($\gamma_{peak}$).
  • The "Aha" Moment: At the largest system sizes ($N=64,000$), the yield strain for the most stable glass drops significantly, falling below the yield strain of the unstable, poorly annealed samples.

Limitations and Future Work

The authors acknowledge that while the narrowing of the intermediate regime is clear, they cannot "conclusively demonstrate" that it vanishes entirely at infinite system sizes (the thermodynamic limit), though the trend strongly suggests it.

Additionally, the study is performed at absolute zero temperature (AQS). Future work would need to explore how thermal fluctuations (room temperature) might "smear" these sharp transitions or affect the observed yield strain reversal. This discovery has significant implications for the study of fatigue failure, suggesting that "better" materials might actually be more prone to certain types of cyclic failure than previously thought.

Summary updated

Abstract

We consider the yielding behaviour of a model glass subjected to asymmetric cyclic shear deformation, wherein the applied strain varies between 0 and a maximum value $γ_{\rm max}$, and study its dependence on the degree of annealing of the glass and system size. The yielding behaviour of well annealed glasses (unlike poorly annealed glasses) display striking differences from the symmetric case, with the emergence of an intermediate strain regime with substantial plasticity but no yielding. The observed behaviour is satisfactorily captured by a recently proposed model. For larger system sizes, the intermediate strain regime narrows, leading to a remarkable reversal of yield strain with annealing.
BibTeX
@misc{adhikari2022yielding,
  author = {Monoj Adhikari and Muhittin Mungan and Srikanth Sastry},
  title = {{Yielding behavior of glasses under asymmetric cyclic deformation}},
  year = {2022},
  eprint = {2201.06535},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2201.06535},
}
Figure