Inverse design of chiral functional films by a robotic AI-guided system

Yifan Xie, Shuo Feng, Linxiao Deng, Aoran Cai, Liyu Gan, Zifan Jiang, Peng Yang, Guilin Ye, Zaiqing Liu, Li Wen, Qing Zhu, Wanjun Zhang, Zhanpeng Zhang, Jiahe Li, Zeyu Feng, Chutian Zhang, Wenjie Du, Lixin Xu, Jun Jiang, Xin Chen, Gang ZouNature Communications 2023 · DOI

Abstract

Artificial chiral materials and nanostructures with strong and tuneable chiroptical activities, including sign, magnitude, and wavelength distribution, are useful owing to their potential applications in chiral sensing, enantioselective catalysis, and chiroptical devices. Thus, the inverse design and customized manufacturing of these materials is highly desirable. Here, we use an artificial intelligence (AI) guided robotic chemist to accurately predict chiroptical activities from the experimental absorption spectra and structure/process parameters, and generate chiral films with targeted chiroptical activities across the full visible spectrum. The robotic AI-chemist carries out the entire process, including chiral film construction, characterization, and testing. A machine learned reverse design model using spectrum embedded descriptors is developed to predict optimal structure/process parameters for any targeted chiroptical property. A series of chiral films with a dissymmetry factor as high as 1.9 ($g_{abs} \sim 1.9$) are identified out of more than 100 million possible structures, and their feasible application in circular polarization-selective color filters for multiplex laser display and switchable circularly polarized (CP) luminescence is demonstrated. Our findings not only provide chiral films with the highest reported chiroptical activity, but also have great fundamental value for the inverse design of chiroptical materials.

Introduction

Chirality, or the "handedness" of a structure, is a fundamental property essential for advanced technologies ranging from 3D displays to enantioselective catalysis. While natural systems excel at creating chiral structures, engineering artificial materials with strong, tunable chiroptical responses remains a significant challenge. Organic materials offer flexibility and cost-effectiveness but typically exhibit weak chiroptical activity, with dissymmetry factors ($g_{abs}$) often below 0.1.

To overcome these limitations, researchers utilize "twisted-stacking" methods to amplify optical responses. However, the design space—encompassing material selection, film thickness, stretching strain, and twist angles—is vast, exceeding $10^8$ possible combinations. Navigating this space to find optimal parameters for specific wavelengths is impossible through traditional trial-and-error methods. This study introduces an autonomous "AI-chemist" to bridge this gap.

The Robotic AI-Chemist Platform

The researchers developed an integrated platform that combines a mobile robotic chemist with a computational "brain" to automate the design, manufacture, and testing of chiral functional films.

Overview of the robotic AI-chemist system, including experimentation, inverse design via GAN, and customized manufacture.

The platform operates through a closed-loop workflow:

  1. Robotic Hardware: A mobile robot performs physical tasks such as material selection, dyeing, stretching (to create anisotropy), and stacking layers at precise twist angles. It also autonomously characterizes the films using UV-vis and Circular Dichroism (CD) spectroscopy.
  2. Forward Prediction: A neural network predicts the CD spectrum of a film based on its structural and process parameters, achieving a high coefficient of determination ($R^2 = 0.985$).
  3. Inverse Design (GAN): A Generative Adversarial Network (GAN) allows users to input a desired optical property, which the AI then translates into the exact manufacturing parameters required.
  4. Spectrum Embedded Descriptors: To enhance accuracy, the system uses descriptors that combine physical parameters with experimental absorption spectra, allowing the AI to correlate molecular absorption with macroscopic chirality.

Heterostructure Strategy and Key Insights

A pivotal discovery in this work was the transition from homostructured to heterostructured bilayers. By stacking a transparent anisotropic layer with a dyed anisotropic layer, the system separates the phase-shifting role from the absorption role. This "hetero" approach provides significantly more degrees of freedom to manipulate light-matter interactions compared to traditional identical-layer stacking.

Comparison between homostructure and heterostructure bilayers, highlighting the superior chiroptical activity and tunability of the heterostructure approach.

This strategy enabled the identification of films with a $g_{abs}$ of 1.9, approaching the theoretical maximum of 2.0. The AI-chemist successfully navigated the 100-million-possibility design space using fewer than 1,500 physical experiments, a feat that would take human researchers years to accomplish manually.

Results and Applications

The system's efficiency led to several high-performance applications:

  • Laser Displays: The platform designed circular polarization-selective color filters that achieved an NTSC color gamut of 140%, significantly exceeding the performance of high-end commercial displays.
  • Tunable Luminescence: By pairing chiral films with achiral perovskite quantum dots, the researchers produced switchable circularly polarized luminescence (CPPL) with $g_{lum}$ values up to 1.9, the highest reported for such systems.

Conclusion and Future Outlook

While the system is highly effective within the visible spectrum, its performance at spectral edges (UV < 400 nm and NIR > 700 nm) is currently limited by the absorption properties of the dye molecules used. Future work will focus on expanding the material library to include UV and NIR absorbers. Additionally, while the robotic system is ideal for discovery and customized manufacturing, scaling these heterotwisted structures for mass production remains a key engineering challenge. Nonetheless, this AI-guided approach represents a paradigm shift in the inverse design of functional chiroptical materials.

Summary updated

Abstract

Artificial chiral materials and nanostructures with strong and tuneable chiroptical activities, including sign, magnitude, and wavelength distribution, are useful owing to their potential applications in chiral sensing, enantioselective catalysis, and chiroptical devices. Thus, the inverse design and customized manufacturing of these materials is highly desirable. Here, we use an artificial intelligence (AI) guided robotic chemist to accurately predict chiroptical activities from the experimental absorption spectra and structure/process parameters, and generate chiral films with targeted chiroptical activities across the full visible spectrum. The robotic AI-chemist carries out the entire process, including chiral film construction, characterization, and testing. A machine learned reverse design model using spectrum embedded descriptors is developed to predict optimal structure/process parameters for any targeted chiroptical property. A series of chiral films with a dissymmetry factor as high as 1.9 (g<sub>abs</sub> ~ 1.9) are identified out of more than 100 million possible structures, and their feasible application in circular polarization-selective color filters for multiplex laser display and switchable circularly polarized (CP) luminescence is demonstrated. Our findings not only provide chiral films with the highest reported chiroptical activity, but also have great fundamental value for the inverse design of chiroptical materials.
BibTeX
@inproceedings{xie2023inverse,
  author = {Yifan Xie and Shuo Feng and Linxiao Deng and Aoran Cai and Liyu Gan and Zifan Jiang and Peng Yang and Guilin Ye and Zaiqing Liu and Li Wen and Qing Zhu and Wanjun Zhang and Zhanpeng Zhang and Jiahe Li and Zeyu Feng and Chutian Zhang and Wenjie Du and Lixin Xu and Jun Jiang and Xin Chen and Gang Zou},
  title = {{Inverse design of chiral functional films by a robotic AI-guided system}},
  booktitle = {Nature Communications},
  year = {2023},
  doi = {10.1038/s41467-023-41951-x},
}
Figure