Geometric memory in incomplete phase transitions across dimensions

This study extends a geometric model of incomplete solid-state phase transitions to three dimensions, demonstrating that a purely geometric memory effect—where prior transformation history alters subsequent plate-size distributions—is robust across dimensions but significantly stronger in two-dimensional systems than in three-dimensional ones.

The Gist

Imagine a room filled with people trying to grow into perfect squares, cubes, or flat pancakes (lamellae). This paper is about how these shapes grow, bump into each other, and how the room "remembers" a previous attempt to shrink them.

Here is the story of the paper, broken down into simple concepts:

The Setup: A Game of Growing Shapes

The authors created a computer simulation to model how materials change from one solid state to another (like how a metal changes when it gets hot or cold).

• The Players: Instead of atoms, they used simple shapes: Squares (2D), Cubes (3D), and Flat Pancakes (3D Lamellae).
• The Growth: These shapes start tiny and try to grow bigger, like a balloon inflating. They want to reach a specific "maximum size" assigned to them.
• The Problem (Jamming): As they grow, they bump into their neighbors. If a shape tries to grow but hits another shape, it stops. Eventually, the room gets so crowded that no new tiny shapes can even fit in the gaps. This is called the "jamming limit."

The Twist: The "Reverse" Game and the Memory Effect

The real magic happens when the process is reversed.

1. The Reverse Rule: In the real world, small things are often less stable than big things. So, in the simulation, when the process reverses, the smallest shapes disappear first. The big, strong shapes stay behind.
2. The "Arrest" (The Pause Button): Imagine you stop the shrinking process halfway. You say, "Okay, stop! Get rid of everything smaller than size 5, but keep everything size 5 and bigger."
3. The Restart: Now, you start the growing process again from scratch.
   • Because the tiny shapes were removed, the new shapes have to grow into the empty spaces left behind.
   • However, the big shapes that survived are still there. They act like giant boulders in a garden. They block the new shapes from growing into certain spots.
4. The Result (The Memory): When the new shapes finish growing, the final crowd looks different than the first time. There is a specific "hole" in the size distribution where the shapes were removed. The system has "remembered" that you stopped at that specific size.

The Analogy: Think of it like a game of Tetris.

• Round 1: You fill the screen with blocks of all sizes until it's full.
• The Pause: You magically delete all the small blocks, leaving only the big ones floating.
• Round 2: You try to fill the screen again. The new blocks fall in, but they can't fit where the big blocks are. The final pattern of blocks looks different than the first time. The screen "remembers" that you deleted the small blocks.

The Big Discovery: Dimension Matters

The authors tested this in three different "worlds":

1. 2D (Flat Squares): Like a flat sheet of paper.
2. 3D (Cubes): Like a solid block of ice.
3. 3DL (Lamellae): Like thin, flat pancakes stacked in 3D space.

The Finding: The "memory" effect is strongest in the 2D world.

• In the flat world (2D), the shapes block each other very efficiently. When you remove the small ones, the big ones create a very clear, sharp "hole" in the pattern.
• In the 3D world, there is more room to wiggle around. The shapes can squeeze past each other more easily, so the "hole" left by the memory is fuzzier and less distinct.
• The "Pancake" world (3DL) is somewhere in between but behaves a bit differently because the pancakes can block each other from the sides and top/bottom.

How They Measured It

To prove this wasn't just a visual trick, they used two math tools:

1. The "Size-Mass Ratio" (SMR): This is like checking a scale. If you look at the size you stopped at, is there less "stuff" there than the sizes right next to it? If yes, the memory is strong.
2. Shannon Entropy: This is a fancy way of saying "how messy or diverse the crowd is."
   • A perfect mix of all sizes has high entropy (very diverse).
   • When you remove the small ones and restart, the crowd becomes less diverse (lower entropy).
   • They found that the 2D world lost the most diversity, meaning the memory effect was the strongest there.

The "DSC Dip" (The Heat Signal)

In real science, they measure these changes using a machine called a DSC (which measures heat flow).

• The authors simulated this. They found that when the material is heated up again, the heat signal shows a little "dip" or "shoulder" right at the temperature where they stopped the process last time.
• This dip is the physical proof of the memory. It's like the material saying, "I remember stopping here."

The Bottom Line

This paper shows that you don't need complex physics or energy calculations to create a "memory" in a material. You just need geometry.

• If you grow shapes, stop them, remove the small ones, and grow again, the physical blocking of the remaining big shapes creates a permanent record of that stop.
• This geometric memory is strongest in flat, 2D-like situations and gets weaker as you move into 3D space.

The authors suggest this helps explain why some real-world metal ribbons (which are thin and flat, like 2D) show this "thermal memory" effect very clearly, while thick blocks of metal (3D) might show it less clearly. It's all about how the shapes fit together and block each other.

Technical Summary

1. Problem Statement

The paper addresses the Thermal Memory Effect (TME) observed in Shape Memory Alloys (SMAs). TME is the phenomenon where a material "remembers" a specific temperature at which a phase transition (martensite $\leftrightarrow$ austenite) was previously halted. If the transformation is stopped at an arrest temperature ($T_{AT}$), cooled, and then reheated, the calorimetric signal (DSC) shows a distinct dip or shoulder near $T_{AT}$.

While phenomenological and thermodynamic models exist, a direct geometric mechanism explaining this memory has been under-explored. Specifically, the influence of dimensionality (2D vs. 3D vs. quasi-3D lamellar structures) and geometric blocking on the strength of this memory effect has not been systematically analyzed. The authors aim to isolate geometric constraints from complex energetic interactions (like long-range elastic strains) to determine if geometry alone can generate memory.

2. Methodology

The authors developed a minimalist, geometry-driven statistical model of solid-state phase transitions. The model operates on a discrete lattice and relies on stochastic nucleation and homothetic (self-similar) growth without explicit energetic terms.

• Geometries Studied:
  • 2D: Square plates.
  • 3D: Cubes.
  • 3DL (Quasi-3D): Square-faced lamellae (plates with fixed thickness) growing within habit planes (XY, XZ, or YZ).
• Forward Transformation (Direct):
  • Nucleation: New "seeds" of minimum size ($L_{min}$) are placed randomly in available empty space.
  • Growth: Each seed grows homothetically toward an intrinsic maximum size ($L_{max}$) chosen randomly from $[L_{min}, L_{max}]$.
  • Constraint: Growth stops if it encounters existing plates (impingement) or system boundaries. The process continues until a "jammed" state is reached where no new $L_{min}$ seeds can fit.
• Reverse Transformation (Incomplete):
  • Modeled as a size-selective disappearance. Smaller plates vanish first (mimicking the instability of high surface-to-volume ratio domains).
  • The process is halted at a specific "arrest size" ($s$), leaving larger plates intact.
• Memory Cycle (Hammering):
  • The system undergoes a cycle: Complete Jamming $\to$ Partial Reverse (Arrest at $s$) $\to$ Regrowth (Direct) $\to$ Partial Reverse (Arrest at $s$ again).
  • Repeating this cycle ("hammering") reinforces the geometric constraint.
• Quantitative Metrics:
  • Size Mass Ratio (SMR): A local metric defined as $SMR(s) = \frac{M_{s-1} + M_{s+1}}{2M_s}$, where $M_k$ is the mass fraction of size $k$. $SMR > 1$ indicates a local depletion (memory) at the arrest size.
  • Shannon Size Entropy ($S$): A global metric measuring the diversity of the plate-size distribution. Lower entropy indicates a narrower, more biased distribution.
  • Simulated DSC Signals: The mass distribution is mapped to a temperature scale to simulate Differential Scanning Calorimetry (DSC) curves, looking for the characteristic "dip" or "shoulder."

3. Key Contributions

1. Extension to Higher Dimensions: The authors successfully generalized a previously established 2D model to 3D cubes and 3DL lamellae, providing a unified framework to compare memory effects across dimensions.
2. Isolation of Geometric Mechanisms: By excluding elastic interactions and explicit thermodynamics, the study proves that geometric blocking alone is sufficient to generate a robust memory effect in first-order solid-solid transformations.
3. Dimensionality Dependence: The study systematically quantifies how dimensionality affects memory strength, revealing that memory is strongest in 2D and weaker in 3D/3DL due to differences in jamming dynamics and available free volume.
4. Quantitative Descriptors: Introduction of the SMR and Shannon entropy as robust tools to quantify the "depth" of the memory imprint and the loss of configurational diversity.

4. Key Results

• Jammed State Distributions:
  • 2D: The initial jammed state has a nearly uniform mass distribution across sizes.
  • 3D: Slightly biased toward smaller sizes.
  • 3DL: Strongly biased toward very small sizes due to lateral blocking in habit planes.
• Memory Effect Visualization:
  • After an incomplete reverse transformation (arrest), the subsequent regrowth shows a distinct depletion of plates at the arrest size ($s$).
  • Simulated DSC: This depletion manifests as a dip or shoulder in the synthetic DSC signal at the temperature corresponding to size $s$.
  • Hammering: Repeating the arrest cycle deepens and narrows this dip, reinforcing the memory.
• Quantitative Findings (SMR & Entropy):
  • SMR: All geometries show $SMR > 1$ after arrest, confirming memory. However, the 2D system exhibits the strongest memory (highest SMR values) compared to 3D and 3DL.
  • Entropy:
    • In 2D, entropy decreases significantly from the initial state to the arrested state, indicating a substantial loss of size diversity.
    • In 3D, entropy changes negligibly after the first arrest (due to the already biased initial distribution).
    • In 3DL, entropy actually increases slightly after the first arrest because the initial distribution was extremely skewed; the arrest "rebalances" the spectrum before hammering reduces it again.
• Dip Depth in DSC: The simulated dip is shallowest in 2D, deeper in 3D, and sharpest in 3DL. However, the relative change and the robustness of the memory imprint (as measured by SMR) are highest in 2D.

5. Significance and Conclusion

• Mechanistic Insight: The paper conclusively demonstrates that the thermal memory effect in SMAs can arise purely from geometric constraints (nucleation, growth, and impingement) without requiring complex energetic coupling.
• Dimensionality Impact: The study suggests that memory effects are inherently stronger in lower-dimensional or ribbon-like geometries (closer to 2D) than in bulk 3D materials. This aligns with experimental observations where TME is often more pronounced in thin ribbons than in bulk samples.
• Practical Application: The model provides a theoretical basis for understanding how incomplete phase transitions alter microstructure. This is relevant for designing SMAs with specific memory capabilities or for using TME as a non-destructive method to record thermal history (e.g., detecting if a component was overheated).
• Future Outlook: The authors suggest that while this model isolates geometry, future work should integrate these geometric findings with elastic and magnetic energy terms to create a more complete physical description of real-world martensitic transformations.

In summary, the paper establishes geometric blocking as a fundamental driver of memory in phase transitions, quantifying that this effect is most potent in two-dimensional systems and diminishes as dimensionality increases and geometric constraints become more complex.