Advances in photodetector-based miniaturized computational spectrometers

Spectral analysis underpins materials science, biomedicine and many other fields, yet conventional spectrometers rely on bulky dispersive elements such as prisms and gratings. This dependence increases system size, complicates optical layouts and constrains portable and on-chip implementations. Computational spectroscopy addresses this trade-off by combining detector-response encoding with algorithmic reconstruction. Photodetector-based architectures generate wavelength-dependent responses through material properties, external modulation, carrier dynamics, or micro- and nanostructures tightly integrated with the detector. They can therefore substantially reduce reliance on external dispersive elements and, in some systems, eliminate them. Here we review recent advances in photodetector-based miniaturized computational spectrometers and establish a two-level framework linking encoding architectures to material platforms. We classify these systems into four categories: passive spatial, active temporal, transient-response and hybrid encoding. By examining relationships among response matrices, data acquisition and reconstruction algorithms, we systematically compare major reconstruction strategies. We further analyze encoding mechanisms and performance differences across silicon, two-dimensional materials, organic semiconductors and perovskites. Overall, these systems are shifting from the optimization of individual device metrics toward application-oriented system-level co-design, with the aim of combining reliable spectral acquisition with miniaturization, stability, energy efficiency and manufacturability.