Overcoming conventional trade-offs in strain sensing with physics-guided inverse design of aperiodic architected nanocomposites

Technical Summary: Overcoming Conventional Trade-offs in Strain Sensing with Physics-Guided Inverse Design of Aperiodic Architected Nanocomposites

Problem Statement
Large-strain piezoresistive sensors are essential for soft robotics and wearable electronics but are constrained by an intrinsic trade-off between sensitivity (gauge factor, $GF$), linearity ($R^2$), and usable strain range. Conventional hyperelastic nanocomposites, formed by blending conductive fillers (e.g., carbon nanotubes, CNTs) into elastomeric matrices (e.g., PDMS), exhibit nonlinear electromechanical responses due to geometric deformation and strain-dependent tunneling. This nonlinearity forces a compromise: materials near the percolation threshold offer high sensitivity but poor linearity and limited strain range, while more robust materials offer linearity at the cost of sensitivity. Existing architectural approaches (e.g., auxetic substrates, wrinkled surfaces) often decouple mechanical and sensing layers, failing to create interdependent systems that can fundamentally alter the performance envelope.

Methodology
The authors propose a paradigm shift from material composition optimization to architectural inverse design. The methodology consists of three integrated components:

1. Fabrication of Aperiodic Architected Nanocomposites:

   • Material System: CNT-PDMS composites with three distinct filler concentrations (0.23, 0.55, and 0.82 vol%) were formulated to create constituent materials with varying electromechanical properties (Materials A, B, and C).
   • Printing Technique: Using Direct Ink Writing (DIW) with a dual-syringe setup, the authors fabricated hierarchical structures. Discrete cells with specific concentrations were printed in series to form "branches," which were then connected in parallel to form "networks." This creates aperiodic, spatially graded architectures where cells are mechanically interdependent.
   • Rheology: Inks were engineered to exhibit Herschel–Bulkley fluid behavior (yield stress and shear-thinning) to ensure shape retention and filament stability during printing.

2. Physics-Guided Hierarchical Framework:

   • Modeling: An analytical, physics-based model was developed to predict global sensor responses from local cell behaviors. The model relies on fundamental physical laws:
     • Iso-stress condition: Cells connected in series within a branch experience the same stress, leading to strain redistribution based on local stiffness ($\sigma_{branch} = \sigma_{cell}$).
     • Iso-strain condition: Branches connected in parallel experience the same global strain.
     • Equivalent Resistance: Branch resistance is the sum of series cell resistances; network resistance is the parallel combination of branch resistances.
   • Inverse Design: Instead of trial-and-error, the framework maps how local constitutive relations ($\sigma(\varepsilon)$ and $R(\varepsilon)$) propagate through the hierarchy. This allows for the systematic exploration of a combinatorial design space (thousands of potential architectures) to identify configurations that meet specific target metrics ($GF$, $R^2$, strain range) without costly experimental iteration.

3. Experimental Validation:

   • Synchronized mechanical tensile testing and four-wire electrical resistance measurements were performed on constituent materials, individual branches, and full networks.
   • Real-time on-body testing was conducted using the sensors for respiratory monitoring to validate performance in a practical application.

Key Results

• Emergent Performance: The architected metamaterials demonstrated electromechanical responses that exceeded the "performance envelope" of the homogeneous constituent materials. The system exhibited non-averaging behavior where the global response emerged from strain redistribution and cell interactions.
• Breaking Trade-offs:
  • High Linearity: A metamaterial design achieved a gauge factor of 1.55 with $R^2 > 0.98$ over a 30% strain range. This represents a fourfold increase in sensitivity compared to the most linear homogeneous constituent (Material C, $GF=0.39$) while maintaining extreme linearity.
  • High Sensitivity: Designs optimized for sensitivity (relaxing linearity to $R^2 > 0.85$) achieved a gauge factor of 37.5 over a 38% strain range. This is a tenfold increase compared to the most sensitive homogeneous constituent (Material A, $GF=3.59$). For a broader strain range of 50%, a twofold increase was observed (GF 7.11 vs 3.59).
  • Strain Range: The metamaterials maintained linear responses over significantly larger strain ranges than their homogeneous counterparts.
• Model Accuracy: The physics-guided model successfully predicted the qualitative ranking and semi-quantitative trends of sensor performance. While measured sensitivities were slightly higher than predicted (due to unmodeled geometric overlaps at cell boundaries), the model accurately captured the strain redistribution mechanisms and the expansion of the performance space.
• Application Proof-of-Concept: In respiratory monitoring, the metamaterial sensors demonstrated amplified sensitivity and linearity compared to homogeneous composites, accurately tracking chest expansion with minimal signal distortion.

Significance and Claims
The paper claims a paradigm shift in the design of electromechanical sensors. The significance lies not in maximizing a single metric (such as the gauge factor) but in navigating an expanded performance space made accessible through hierarchical organization.

• From Material to Architecture: The work demonstrates that tailored electromechanical responses can arise from structural organization rather than intrinsic material properties. This moves the field from optimizing chemical composition to engineering network topology.
• Rational Inverse Design: The physics-guided framework transforms sensor development from data-intensive, black-box AI models or costly trial-and-error into an interpretable, systematic design process. It enables the rational selection of architectures for specific application needs, whether that requires high sensitivity, extreme linearity, or strain-insensitive interconnects.
• Universality: The framework is presented as material- and functionality-agnostic. While demonstrated here with CNT/PDMS piezoresistive sensors, the authors state it applies to any hierarchical system where mechanical and functional properties interact, extending to other modes such as transverse piezoresistivity, piezocapacitance, and applications in soft robotics, stretchable transistors, and multi-modal sensing.

The authors conclude that this approach unlocks extraordinary combinations of sensitivity, linearity, and strain range, enabling the rational design of soft electronic systems where performance is dictated by architecture.