Ground-State Descriptor Enables Machine Learning-Assisted Virtual Screening of AIE-Active Mechanofluorochromic...

Aggregation-induced emission mechanofluorochromic (AIE-MFC) molecules with high-contrast are in high demand for pressure-sensing devices and optoelectronic devices. However, developing AIE-MFC molecules with high-contrast beyond 100 nm still highly relies on scientific intuition through experimental and traditional trial-and-error methods. Herein, we establish a new electronic descriptor E_DAHL (the orbital energy of the donor and acceptor under ground-state) to characterize mechanofluorochromic (MFC) properties (preground and postground fluorescence emission wavelengths Oλ_em,max and Gλ_em,max). And an MFC property prediction model with high prediction accuracy and less computational complexity is established by combining E_DAHL with molecular descriptors as inputs. Subsequently, we constructed a closed-loop approach that integrates machine learning (ML), density functional theory (DFT), high-throughput virtual screening (HTVS), and experiment to accelerate the discovery of AIE-MFC molecules with high contrast in a large space. After processing, 37 new high-contrast AIE-MFC molecules beyond 100 nm are rapidly screened out of more than 10000 candidates. Notably, experiments were used to further confirm the reliability of the workflow, opening an avenue for screening of high-contrast AIE-MFC molecule. We believe the new electronic descriptor may be extended to discover other molecules constructed from donors and acceptors with different properties.