When structure does not imply symmetry

1. Problem Statement

Continuum mechanics and constitutive modeling traditionally operate under the paradigm that microstructural alignment dictates macroscopic mechanical symmetry. The prevailing assumption is that if a material possesses a visible, anisotropic microstructure (such as aligned fibers or filaments), its mechanical response must also be anisotropic (direction-dependent). This assumption is rooted in invariant-based constitutive laws where fiber-dependent invariants (e.g., $I_4, I_5$) are introduced to account for preferred directions.

However, this assumption is challenged by observations in various soft materials (e.g., polymers, cartilage, brain tissue) where microstructural organization does not always translate into direction-dependent macroscopic behavior. The paper addresses a central question: Does anisotropic microstructure necessarily produce anisotropic mechanical response and sensory perception? Specifically, it investigates whether visual structural alignment is a sufficient predictor of material symmetry classes in complex soft materials.

2. Methodology

The authors employed a combined approach of controlled mechanical experimentation and data-driven automated model discovery using three distinct fungi-based protein systems.

A. Material Systems

Three materials were selected to represent a spectrum of microstructural organization:

1. Mushroom Root Mycelium: Large sheet-like structures with limited alignment.
2. Fruiting Body: Densely connected fibers with a pronounced preferred orientation.
3. Protein Mycelium Blend: A heterogeneous mix of locally aligned regions and large voids.

B. Experimental Strategy

• Sample Preparation: Samples were cut in two orthogonal orientations relative to the dominant structural direction: in-plane (parallel to fiber alignment) and cross-plane (perpendicular).
• Mechanical Testing: For each material and orientation, $n=10$ samples underwent quasi-static testing up to 10% strain in three modes:
  • Uniaxial Tension
  • Uniaxial Compression
  • Simple Shear
  • Total: 180 tests ($3 \text{ materials} \times 2 \text{ orientations} \times 3 \text{ modes} \times 10 \text{ samples}$).
• Sensory Evaluation: A panel of human participants ($n=16-21$) evaluated the texture (softness, chewiness, fibrousness, etc.) of the materials in both orientations using a 5-point Likert scale.
• Microscopy: Light microscopy was used to characterize the hierarchical microstructure (hyphae networks, cell walls) to confirm structural anisotropy.

C. Data-Driven Model Discovery

Instead of prescribing a symmetry class a priori, the authors used a Constitutive Neural Network to discover the governing symmetry directly from the data.

• Framework: Finite deformation continuum mechanics using strain energy density functions ($\psi$) expressed as a sparse combination of scalar invariants.
• Invariants:
  • Isotropic: $I_1$ (total stretch), $I_2$ (shear-like distortion), $I_3$ (volume change).
  • Anisotropic: $I_4$ (fiber stretch squared), $I_5$ (nonlinear fiber stretch).
• Algorithm: A neural network with 12 hidden terms (linear, quadratic, exponential) was trained on the stress-strain data. L1 regularization (Lasso) was applied to enforce sparsity, effectively "pruning" irrelevant terms. If fiber-dependent terms ($I_4, I_5$) were pruned, the material was identified as isotropic; if retained, it was anisotropic.

3. Key Results

A. Mechanical Response & Symmetry Classes

The experimental data revealed that structurally similar fibrous materials exhibited distinct symmetry classes:

1. Mycelium (Strongly Anisotropic):
   • Showed significant directional differences in stiffness (e.g., Tensile stiffness: 106 kPa in-plane vs. 29 kPa cross-plane; $p < 0.01$).
   • Discovered Model: $\psi = 5.07[I_1 - 3] + 1.64\langle I_5 - 1 \rangle^2$.
   • Interpretation: The model retained the anisotropic $I_5$ term, confirming strong mechanical anisotropy.
2. Fruiting Body (Transitional/Weakly Anisotropic):
   • Showed moderate directional differences (Tensile stress: 4.9 kPa in-plane vs. 3.5 kPa cross-plane), but many differences were not statistically significant ($p > 0.05$).
   • Discovered Model: $\psi = 4.42[I_1 - 3] + 0.69[\exp(0.70\langle I_5 - 1 \rangle^2) - 1]$.
   • Interpretation: The anisotropic term was retained but weighted significantly lower than in the mycelium, placing it in a transitional regime.
3. Protein Mycelium Blend (Effectively Isotropic):
   • Showed nearly identical stress-strain curves for in-plane and cross-plane orientations across all loading modes ($p > 0.05$).
   • Discovered Model: $\psi = 14.30[I_1 - 3]$.
   • Interpretation: The model reduced to a pure Neo-Hookean isotropic form. Despite visible fibrousness, the fiber-dependent invariants were pruned, indicating mechanical isotropy.

B. Sensory Perception

• Anisotropy in Perception: Mycelium and Fruiting Body showed direction-dependent differences in specific attributes (e.g., softness, chewiness).
• Isotropy in Perception: The Protein Mycelium Blend was perceived as isotropic.
• Decoupling of Fibrousness: Crucially, the attribute of "fibrousness" showed no statistically significant directional dependence for any of the three products ($p > 0.05$). This suggests that perceived fibrousness is not a direct function of microstructural alignment or stiffness anisotropy but emerges from a complex combination of cues.

C. Model Discovery Validation

The automated discovery successfully identified the correct symmetry class without human bias:

• It correctly identified the need for $I_5$ terms for the anisotropic materials.
• It correctly discarded $I_4$ and $I_5$ terms for the isotropic blend, proving that structure alone does not dictate the necessary invariants for the constitutive law.

4. Key Contributions

1. Refutation of Structure-Symmetry Equivalence: The study provides empirical evidence that visible microstructural anisotropy (fibrous networks) does not universally imply macroscopic mechanical anisotropy.
2. Data-Driven Symmetry Identification: It establishes a framework using constitutive neural networks and L1 regularization to infer material symmetry directly from experimental data, removing the need for a priori assumptions about symmetry groups.
3. Mechanical vs. Sensory Decoupling: It demonstrates that mechanical anisotropy and sensory perception of "fibrousness" are not strictly coupled; a material can be mechanically isotropic yet structurally fibrous, and perceived fibrousness may not vary with loading direction.
4. Application to Food Science: The findings offer a new paradigm for designing plant-based meat alternatives, suggesting that achieving a specific texture (isotropic vs. anisotropic) requires tuning the mechanical response, not just the visual microstructure.

5. Significance

This work challenges a fundamental tenet of continuum mechanics and soft matter physics: that geometry dictates symmetry. By showing that symmetry emerges from the mechanical response rather than the geometry itself, the paper:

• Prevents the misclassification of materials in engineering and biology (e.g., assuming brain tissue or cartilage is anisotropic based solely on fiber imaging).
• Provides a robust, generalizable methodology for characterizing complex soft materials where traditional invariant-based modeling might fail due to incorrect symmetry assumptions.
• Offers critical insights for the food industry, indicating that "meaty" textures can be engineered through mechanical isotropy even in fibrous systems, broadening the design space for sustainable protein alternatives.