Artículos
Motores de simulación numérica y arquitecturas de aprendizaje profundo para el diseño inverso de metamateriales multicapa
Resumen
El diseño inverso de metamateriales multicapa —estructuras artificiales con propiedades electromagnéticas, acústicas, térmicas o mecánicas diseñadas a medida— constituye un problema de optimización de alta dimensionalidad que ha impulsado la integración del aprendizaje profundo en la fotónica computacional y la ciencia de materiales. Este artículo presenta una revisión sistemática de la literatura realizada siguiendo el protocolo PRISMA 2020, orientada a mapear los métodos de simulación numérica empleados como motores directos o generadores de datos de entrenamiento dentro de flujos de trabajo de diseño inverso asistido por aprendizaje profundo. A partir de 392 registros recuperados de Scopus y Web of Science en el periodo 2010–2025, un proceso de cribado multietapa basado en un índice heurístico de priorización denominado CARS (Citation-Adjusted Relevance Score) permitió identificar 44 artículos centrales. Se identificaron once métodos numéricos distintos, organizados en dos capas conceptuales: motores clásicos de simulación (método de elementos finitos, optimización topológica, expansión en modos guiados, método adjunto y algoritmos genéticos) y arquitecturas de aprendizaje profundo (CNN en tándem, transformer, GAN, PINN y redes neuronales difractivas). Los resultados evidencian un predominio del dominio electromagnético-fotónico (72 %), un crecimiento acelerado de las publicaciones basadas en aprendizaje profundo a partir de 2022, y una tendencia clara hacia flujos híbridos que combinan física, datos y aprendizaje. La revisión concluye proponiendo una hoja de ruta experimental priorizada para futuras investigaciones.
Palabras clave
Cómo citar
Estadísticas
Referencias
Cai, Y., Huang, Y., Feng, N., & Wang, H. (2023). Improved transformer-based target matching of terahertz broadband reflective metasurfaces. Optics Express, 31(26), 44084–44095. https://doi.org/10.1364/OE.508956
Cao, L., Wu, J., Zhang, Z., & Zhang, C. (2024). Design optimization of elastic metamaterials with multilayered honeycomb structure. Applied Thermal Engineering, 236, 121516. https://doi.org/10.1016/j.applthermaleng.2023.121516
Chen, Y., Zhu, J., Xie, Y., Feng, N., & Liu, Q. H. (2019). Smart inverse design of graphene-based photonic metamaterials by an adaptive artificial neural network. Nanoscale, 11(19), 9749–9755. https://doi.org/10.1039/C9NR01315F
Fu, S. M., Zhong, Y. K., Ju, N. P., & Tu, M. H. (2016). Broadband polarization-insensitive metamaterial perfect absorbers using topology optimization. IEEE Photonics Journal, 8(5), 1–11. https://doi.org/10.1109/JPHOT.2016.2602092
García-Esteban, J. J., Bravo-Abad, J., & Cuevas, J. C. (2021). Deep learning for the modeling and inverse design of radiative heat transfer. Physical Review Applied, 16(6), 064006. https://doi.org/10.1103/PhysRevApplied.16.064006
Hodge, J. A., Mishra, K. V., & Zaghloul, A. I. (2020). Joint multi-layer GAN-based design of tensorial RF metasurfaces. IEEE Transactions on Antennas and Propagation, 68(3), 1646–1658. https://doi.org/10.1109/TAP.2019.2948818
Hu, Y., Zhang, W., Chen, Y., & Zhao, Z. (2024). Deep-learning-assisted inverse design of polarization-multiplexed structural metaoptics. Light: Science & Applications, 13, 88. https://doi.org/10.1038/s41377-024-01409-1
Jia, Y., Qian, C., Fan, Z., & Ding, Y. (2022). In situ customized illusion enabled by global metasurface reconfiguration. ACS Nano, 16(3), 3752–3758. https://doi.org/10.1021/acsnano.1c10592
Jiang, X., Yuan, H., He, X., & Du, M. (2023). Implementing of infrared camouflage with thermal management based on inverse design and hierarchical metamaterial. Nanophotonics, 12(14), 2891–2902. https://doi.org/10.1515/nanoph-2023-0137
Knightley, T., Yakovlev, A. B., & Pacheco-Peña, V. (2023). Neural network design of multilayer metamaterial for temporal differentiation. Advanced Optical Materials, 11(12), 2202655. https://doi.org/10.1002/adom.202202655
Lin, Z., Groever, B., Capasso, F., Rodriguez, A. W., & Lončar, M. (2018). Topology-optimized multilayered metaoptics. Physical Review Applied, 9(4), 044030. https://doi.org/10.1103/PhysRevApplied.9.044030
Lininger, A., Hinczewski, M., & Strangi, G. (2021). General inverse design of layered thin-film materials with convolutional neural networks. ACS Photonics, 8(12), 3641–3650. https://doi.org/10.1021/acsphotonics.1c01498
Liu, W., Li, Z., Cheng, H., & Chen, S. (2022). Integrated photonic metasystem for image classifications at telecommunication wavelength. Nature Communications, 13(1), 123. https://doi.org/10.1038/s41467-021-27784-0
Liu, Y., Wang, Y., Ren, H., & Meng, Z. (2024). Ultrastiff metamaterials generated through a multilayer strategy and topology optimization. Nature Communications, 15, 4518. https://doi.org/10.1038/s41467-024-48930-6
Luo, Y., Mengu, D., & Yardimci, N. T. (2019). Design of task-specific optical systems using broadband diffractive neural networks. Light: Science & Applications, 8(1), 112. https://doi.org/10.1038/s41377-019-0223-1
Ma, J., Huang, Y., Pu, M., Xu, M., & Xu, X. (2020). Inverse design of broadband metasurface absorber based on convolutional neural network. Optics Express, 28(20), 29021–29031. https://doi.org/10.1364/OE.401360
Mansouree, M., Kwon, H., & Arbabi, E. (2020). Multifunctional 2.5D metastructures enabled by adjoint optimization. Optica, 7(1), 77–84. https://doi.org/10.1364/OPTICA.373835
Naseri, P., & Hum, S. V. (2021). A generative machine learning-based approach for inverse design of multilayer metasurfaces. IEEE Transactions on Antennas and Propagation, 69(9), 5725–5739. https://doi.org/10.1109/TAP.2021.3060142
Noureen, S., Syed, I. H., & Ijaz, S. (2023). Physics-driven tandem inverse design neural network for efficient optical metasurfaces. Nanomaterials, 13(3), 1030. https://doi.org/10.3390/nano13031030
Page, M. J., Moher, D., & Bossuyt, P. M. (2021). PRISMA 2020 explanation and elaboration: Updated guidance and exemplars for reporting systematic reviews. BMJ, 372, n160. https://doi.org/10.1136/bmj.n160
Peralta, I., & Fachinotti, V. D. (2017). Optimization-based design of heat flux manipulation devices with emphasis on fabricability. Scientific Reports, 7(1), 6086. https://doi.org/10.1038/s41598-017-06565-6
Teng, Y., Li, C., Li, S., & Xiao, L. (2023). Efficient design method for terahertz broadband metasurface patterns via deep learning. Optics Express, 31(18), 29638–29651. https://doi.org/10.1364/OE.498079
Wang, L., Dong, J., Zhang, W., & Zhao, Z. (2023). Deep learning assisted optimization of metasurface for multi-band compatible camouflage. Optics Express, 31(4), 6320–6333. https://doi.org/10.1364/OE.481765
Weeger, O. (2022). Isogeometric sizing and shape optimization of 3D beams and lattice structures at large deformations. Structural and Multidisciplinary Optimization, 65(2), 43. https://doi.org/10.1007/s00158-021-03131-7
Xiong, J., Shen, J., & Gao, Y. (2023). Real-time on-demand design of circuit-analog plasmonic stack metamaterials. Optics Express, 31(8), 13114–13127. https://doi.org/10.1364/OE.484834
Yago, D., Sal-Anglada, G., & Roca, D. (2024). Machine learning in solid mechanics: Application to acoustic metamaterials. Computer Methods in Applied Mechanics and Engineering, 421, 116739. https://doi.org/10.1016/j.cma.2024.116739
Zanotti, S., Minkov, M., Nigro, D., Fan, S., & Andreani, L. C. (2024). Legume: A free implementation of the guided-mode expansion method for photonic structures. SoftwareX, 26, 101665. https://doi.org/10.1016/j.softx.2024.101665
Zhu, L., Zhang, C., Guo, J., & Ding, Y. (2022). Deep learning for electromagnetically induced transparency (EIT) metasurface design. Optics Express, 30(18), 32319–32330. https://doi.org/10.1364/OE.465503