Neural morphological devices have vast application prospects. Besides accomplishing computational tasks such as perception and recognition through efficient parallel computing modes, these devices are essential biomimetic components for next-generation intelligent sensing systems (such as electronic skins) in robotics. However, despite substantial progress, the current biomimetic nociceptors based on three-terminal vertical transistors and two-terminal memristors are disadvantaged by unclear model transitions, unstable switching, and excessive write noise. Here we develop an easily manufactured, energy-economical, and reconfigurable optical nociceptor with a double-layer Si₃N₄/SiO₂ tunneling junction. The resulting device operates all common synaptic functions under non-injured conditions, such as excitatory postsynaptic current (EPSC), facilitation, dynamic paired-pulse facilitation (PPF), and consolidation, with a low energy consumption density (33.5 fJ/μm²). The device additionally realizes an intensity- and rehearsal-triggered adaptive mode jump resembling a biological alarm system.

Our nociceptor featuring bi-mode synaptic operations is based on lateral back-to-back Schottky junctions (B-B SJ), in which spontaneous trappings at the SiO₂ interface can be modulated to host completely programmable neuromorphic functions. B-B SJs are easily fabricated and include a Si₃N₄ tunneling layer inserted at the semiconductor/SiO₂ interface (Fig. 1). The Si₃N₄ tunneling junction modulates its own tunneling modes under different stimulation patterns, behaving as an adaptive switch that controls the electron flux reaching the SiO₂ interface (Fig. 2). When the light intensity or pulse number is sufficient, the photogating field located at both sides of Si₃N₄ can reach a threshold by successively accumulating electrons at the SiO₂ interface, enabling the bi-mode feature of the tunneling mode transition.

To demonstrate adaptive jumping between the bi-mode operations, a set of single pulses with differing light power was applied to the device as shown in Fig. 4(a). The EPSC exhibited a threshold-jumping characteristic. Subsequently, we tested a series of synaptic functions at the low-threshold stage under a low-intensity stimulation of 5 μW/cm2, which can be viewed as non-injured stimulation. A single-pulse EPSC output corresponding to the transition from a high-resistance state (HRS) to a low-resistance state (LRS) was followed by a decay from the LRS to HRS [Fig. 4(b)]. PPF was also demonstrated in the device. After fitting the interval-dependent PPF behavior as shown in Fig. 4(c), two characteristic decay times (16.7 ms and 840.0 ms) similar to those in biological systems were obtained. The stable state of the device was enhanced after 12 continuous pulses and the duration of this state was prolonged under repetitive stimulus [Fig. 5(a)]. Similarly, continuous light pulses at higher intensity (10.4 μW/cm2) triggered clear mode jumping in the device with a threshold of approximately 20.5 pA [Fig. 5(b)]. Moreover, this mode transition was highly reversible, indicating no structural damage to the device.

We present a novel optical nociceptor device that performs efficient resistance switching through optical modulation of the Schottky barrier, along with bi-mode synaptic operation using a bilayer Si₃N₄/SiO₂ tunneling junction. In the low-threshold stage, the device behaves like an optical synapse, replicating some common synaptic functions. Interestingly, the device can adaptively jump to the high-threshold stage under high-intensity repetitive stimuli, resembling the threshold alarm characteristics of pain perception in biological systems. Moreover, the high reversibility of this mode transition confirms the physical feasibility of reconfigurable nociceptors.