Neural-field design of broadband Rayleigh-wave carpet cloaks under microstructure realisability constraints
Abstract
Transformation elasticity provides appropriate material distributions for elastodynamic cloaks but the required stiffness tensors generally violate the minor symmetries of Cauchy elasticity and are difficult to realise using conventional materials. Existing approaches restore these symmetries by modifying the transformed tensor, producing only an approximate cloak. In this work rather than modifying the transformed tensor we seek the best performing cloak within the class of Cauchy materials. We formulate 2D Rayleigh wave carpet cloak design as an optimisation problem governed by partial differential equations. Using a coordinate based neural-field and a differentiable finite element model solver we optimise symmetric stiffness and density fields by minimising wave field distortion. Both single frequency and broadband optimisation are considered, with the broadband model trained over multiple frequencies. Physical realisability is addressed using a database of homogenised microstructures through conditional diffusion, neural-field inverse design, and nearest-neighbour selection. FEM simulations show that the optimised Cauchy design approaches the ideal transformation-based cloak. After projection onto explicit microstructures, the homogenised representation recovers approximately 97% of the defect free reference surface-displacement magnitude, while direct FEM simulation of the fully resolved microstructured geometry recovers approximately 76%
Related
- Breaking Rayleigh’s curse: broadband frequency super-resolution via a quantum harmonic oscillator
- Quantization-Aware Nonlinear Diffractive Neural Network for Robust Optical Image Reconstruction under Fabrication Constraints
- The S-function at arbitrary order: hyperplanes of symmetries and the realisability condition
- Neural Network Guided Parameter Space Constraints for Gravitational Wave Searches from Binary Black Holes
- Tighter Dark Matter Constraints from the Projected Mass Method: A Neural Network Enhanced Method for Galaxy Groups and Clusters
BibTeX
@misc{aznaurov2026neuralfield,
title = {{Neural-field design of broadband Rayleigh-wave carpet cloaks under microstructure realisability constraints}},
author = {David Aznaurov and Davit Piliposyan and Danila Rukhovich and Sebastien Guenneau},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.05163},
eprint = {2609.05163},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.05163},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - Neural-field design of broadband Rayleigh-wave carpet cloaks under microstructure realisability constraints
AU - David Aznaurov
AU - Davit Piliposyan
AU - Danila Rukhovich
AU - Sebastien Guenneau
PY - 2026
DA - 2026/09/04
PB - arXiv
DO - 10.48550/arxiv.2609.05163
UR - https://arxiv.org/abs/2609.05163
AB - Transformation elasticity provides appropriate material distributions for elastodynamic cloaks but the required stiffness tensors generally violate the minor symmetries of Cauchy elasticity and are difficult to realise using conventional materials. Existing approaches restore these symmetries by modifying the transformed tensor, producing only an approximate cloak. In this work rather than modifying the transformed tensor we seek the best performing cloak within the class of Cauchy materials. We formulate 2D Rayleigh wave carpet cloak design as an optimisation problem governed by partial differential equations. Using a coordinate based neural-field and a differentiable finite element model solver we optimise symmetric stiffness and density fields by minimising wave field distortion. Both single frequency and broadband optimisation are considered, with the broadband model trained over multiple frequencies. Physical realisability is addressed using a database of homogenised microstructures through conditional diffusion, neural-field inverse design, and nearest-neighbour selection. FEM simulations show that the optimised Cauchy design approaches the ideal transformation-based cloak. After projection onto explicit microstructures, the homogenised representation recovers approximately 97% of the defect free reference surface-displacement magnitude, while direct FEM simulation of the fully resolved microstructured geometry recovers approximately 76%
ER -