Study outlines a readout framework for photodetectors made from 2D materials

A team of researchers, most of them at Sichuan University in Chengdu, China, has published a framework meant to help photodetectors built from two-dimensional (2D) materials move from single-device lab studies towards use in larger systems. The open-access paper appeared in Nature Communications. The journal marks it as an early release of peer-reviewed, accepted research that may still be edited before the final Version of Record replaces it.

Photodetectors are light-sensing devices that produce an electrical current, called a photocurrent, when light reaches them. According to the authors, research on 2D-material detectors has mostly aimed at making them more responsive, while another problem holds them back: trap-governed persistent photoconductivity (PPC). Put simply, defects in the material shape how the current behaves, so the output depends on the detector's recent history. The authors describe drift, hysteresis and fatigue that builds up over time. In their view this leaves a mismatch between what the device delivers and what the connected system expects, and slows the step from device to working system.

Instead of handling these effects only as imperfections, the team recasts them as information about the device's state, which can be defined, corrected and passed on to electronics. The framework is called cross-layer because it covers several levels together: analysing the defect mechanisms in the material, defining signals that change over time (non-stationary signals), compensating for dynamic distortion, and engineering the readout interface that links the detector to the rest of the system. The researchers applied it to detectors made from three 2D materials: MoS2, WSe2 and PtSe2.

According to the abstract, restructuring the interface and stabilising the readout allowed the team to map nW-level photocurrents, which vary with the device's state, onto a mixed-signal readout interface, the circuitry that turns the detector's small currents into system output. The group then extended the method to arrays of detectors that share one unified output and to fusing signals from several sensors. A frequency-guided way of representing the data was evaluated in a machine-vision task; the abstract does not give performance figures for that test.

The authors conclude that the response that PPC drives, which depends on what the device experienced before, should be read as “a state-encoded signal” that links defect behaviour, readout and system tasks, not merely as a flaw of the individual device. They describe the work as an engineering route from device-level studies to system-level applications; the abstract does not discuss products or timelines. Funding came from the National Natural Science Foundation of China, the National Key Research and Development Program of China and the Sichuan Science and Technology Foundation, and the authors declare no competing interests.