Feature-Guided Inverse Design of Organic A-Site Cations for Perovskites Dimensional Engineering

Organic spacer cations critically influence the dimensionality and stability of low-dimensional perovskites (LDPs), yet current A-site candidate selection remains largely empirical. Herein, we present a machine-learning-driven molecular generation framework based on a Long Short-Term Memory Network model guided by key molecular descriptors, incorporating a Double-Fit strategy to improve dimensional property alignment and structural rationality of LDPs. Our model inversely generates organic cations targeting specific structural dimensions. Subsequent density functional theory calculations identify candidates with favorable thermodynamic stability and configurational features. Experimental synthesis and structural characterization of the resulting perovskites confirm the model's predictive accuracy. This approach provides a rational design paradigm for A-site cations in LDPs and establishes a general platform to accelerate discovery of new organic-inorganic hybrid perovskite materials.