Modulating Thermally Activated Multiexcitonic De‐Trapping in Zero‐Dimensional Tin‐Halide Perovskites for Deep‐Learning‐Assisted Temperature Mapping

ABSTRACT Low‐dimensional metal halides are promising for remote optical thermometry due to their strong thermally‐quenched luminescence from self‐trapped excitons (STEs), yet achieving rational control of STE‐specific thermal responses and integrating them with deep learning for robust temperature mapping remains challenging. Here, it is unveiled thermally activated multiexcitonic de‐trapping dynamics in Bi 3+ /Te 4+ co‐doped 0D Cs 2 SnCl 6 . By tuning the spatial proximity between BiCl 6 and TeCl 6 octahedra, a thermally induced backward energy transfer (BET) from Te 4+ ‐induced STEs to Bi 3+ ‐induced STEs emerges at elevated temperatures. Temperature‐dependent Raman spectroscopy reveals that at high Te 4+ doping levels, TeCl 6 ‐related vibrational modes undergo more rapid thermal attenuation than SnCl 6 ‐related modes, indicating enhanced local lattice dynamics and increased dynamic disorder upon heating. This thermally softened coordination environment strengthens electron‐phonon coupling and facilitates BET processes, leading to rapid thermal quenching of Te 4+ ‐related STE emission. These phenomena enable distinct thermal‐response tuning of dual STE emissions, providing a versatile approach to modulate sensitivity in ratiometric, fluorescence‐lifetime‐based, and colorimetric optical thermometry. By combining pronounced thermochromic behavior with deep‐learning‐based image analysis, we establish a robust colorimetric thermometry platform for an encapsulated light‐emitting diode with 1.12°C resolution and 15 ms response. This work showcases a practical pathway toward intelligent thermal sensing in device applications.