While neuromorphic hardware allows semiconductors to process temporal data directly, traditional hardware devices have fixed, single response speeds once fabricated. Electron Trapping Layer: Controls the device’s recovery speed across multiple non-volatile levels without requiring a continuous external power supply. Drastic Error Reduction: When predicting complex data containing a mix of fast and slow changes, the PDM reduced prediction errors by up to 40 times compared to traditional fixed-response hardware. Input-Scale Adaptation: The device dynamically configures its response to match varying input speeds, such as fluctuating handwriting pace or object movement, eliminating the need for complex software preprocessing. Array integration and commercial viabilityPublished in Nature Communications, the study verified the hardware’s practical viability by constructing a functional PDM array in collaboration with researchers from Samsung Electronics’ Semiconductor R&D Centre: