科学家开发出一种人工神经元，其工作电压接近生物神经细胞，使电子电路能够直接与活体组织通信。该装置产生约0.1伏的电脉冲，与自然神经元常见的约70至130毫伏范围相匹配。此前的人工神经元通常需要约0.5伏或更高电压，并消耗约10倍电压和100倍功率，这使得与脆弱的生物信号直接交互变得困难。通过将工作电压降低到接近生物范围，研究团队缩小了电子系统与生物神经系统之间长期存在的能量和信号差距。

该人工神经元的核心是忆阻器，其电阻随电流变化，并通过来自细菌Geobacter sulfurreducens的蛋白纳米线调节，使其在生物电压水平运行。在实验中，该开关约在60毫伏和1.7纳安时启动，然后自动重置，从而产生类似神经脉冲的上升和下降循环。电路随后利用电容器快速充电形成短暂电压峰值，并在充电积累后强制进入短暂的不应期。该机制使输出在每次脉冲后接近零，并能够触发另一个人工神经元，支持构建更复杂的神经网络结构。

实验还表明化学信号可以调节该装置的活动。当钠浓度升高时，电路复位加快并提高发放频率；多巴胺通过石墨烯传感器产生双向响应，在不同剂量下既可增强也可抑制活动。与心肌细胞的测试显示，正常节律不会触发电路，但加速心跳的药物会引发电脉冲。与早期化学人工神经元相比，该设计在化学操作中至少减少100倍能量消耗，并可通过标准硅工艺制造。研究表明低电压神经元能够直接处理来自生物细胞的信号，未来可能减少可穿戴设备中的放大步骤并降低功耗。

Scientists developed an artificial neuron that operates within the same voltage range as biological nerve cells, enabling electronic circuits to communicate directly with living tissue. The device generates electrical spikes around 0.1 volts, closely matching the roughly 70 to 130 millivolt signals typical of natural neurons. Earlier artificial neurons typically required about 0.5 volts or more and used roughly 10 times higher voltage and 100 times more power, preventing direct interaction with delicate biological signals. By reducing the operating voltage to biological levels, the researchers narrowed the long-standing energy and signal gap between electronic systems and neural tissue.

The artificial neuron centers on a memristor whose resistance changes with current and is tuned using protein nanowires from the bacterium Geobacter sulfurreducens to function at biological voltage levels. In tests, the switch activated near 60 millivolts and 1.7 nanoamps, then reset automatically, producing a rise-and-fall cycle similar to a neural spike. A capacitor then rapidly charged to create a brief voltage peak and forced the circuit into a short refractory period as charge accumulated. This mechanism returned the output close to zero after each spike and allowed it to trigger another artificial neuron, enabling the construction of more complex neural network circuits.

Experiments also showed that chemical signals could modulate device activity. Higher sodium levels accelerated the circuit reset and increased firing frequency, while dopamine detected through a graphene sensor produced a bidirectional response that either increased or reduced activity depending on dosage. Tests with cardiomyocytes showed that normal rhythms did not activate the circuit, but a drug that sped up the cells triggered electrical spikes. Compared with earlier chemical artificial neurons, the design used at least 100 times less energy during chemical operation and can be fabricated with standard silicon processes. The results indicate that low-voltage artificial neurons can process signals directly from living cells and may reduce amplification steps and energy use in future wearable or implantable devices.