A Gate-Tunable Thermal Persistent Photocurrent Device for In-Sensor Spiking Neural Networks

Area of Science:

• Materials Science
• Condensed Matter Physics
• Device Physics

Background:

• Persistent photocurrent is common in van der Waals (vdW) heterostructures.
• Its microscopic origin is often attributed to trap-assisted photogating but remains unclear.

Purpose of the Study:

• To clarify the mechanism of persistent photocurrent in a gate-tunable MoS2/black phosphorus (BP) p-n heterojunction.
• To investigate the role of self-heating versus trap-assisted photogating.

Main Methods:

• Combined DC and lock-in measurements with time-resolved decay.
• Simultaneous measurement of general (DC) and net photocurrent.
• Tuning the heterostructure between p-n and n-n configurations using a back gate.
• Application of a leaky integrate-and-fire model for neuromorphic network demonstration.

Main Results:

• DC photocurrent exhibits long decay lifetimes (τ > 100s) dependent on gate voltage (Vg) and bias, inconsistent with trap-only models.
• Long-lived response is dominated by self-heating from majority-carrier recombination in forward-biased p-n junctions, generating thermoelectric and bolometric currents.
• Back gate effectively switches the thermal channel on/off by tuning the heterostructure configuration.

Conclusions:

• Persistent photocurrent in MoS2/BP heterostructures is primarily caused by recombination-induced self-heating, not trap-assisted photogating.
• This thermal mechanism can be controlled by gate voltage, enabling applications in neuromorphic computing.
• Demonstrated an in-sensor spiking neural network with high accuracy (91.95%) on MNIST using thermally engineered devices.