以下文章来源于Artificial Synapse ，作者Synapse
研究背景
人工神经系统，能够实现神经和认知功能，在广泛的应用中是非常可取的，如神经康复，人机互动，人工反射弧和智能机器人。柔软和可拉伸的神经形态电子设备将使它们能够抵抗钝性损伤(例如，冲击、压缩和压力)，适应不可预测的变形情况。此外这种设备将提高信号采集的保真度和佩戴的舒适度，因为它们可以无缝地附着在人体皮肤上。因此人们为实现可拉伸的人工神经做了很多努力，它们可以表现出可塑性行为，感知环境信号(如触觉、温度和光线)，其至控制生物运动反应。在生物神经系统中，同一神经元的突触在同一刺激下可以表达相反形式的可塑性。具有这种双语功能的单一神经元可以模仿生物神经元的多维调整性，并通过促进系统整合、实现并行信息处理和降低功耗来提高突触计算性能。只有少数研究报道了具有双门或异质结结构的双极性突触晶体管，可以实现双语突触行为。然而，这些研究需要复杂的器件结构，而且大多数器件是刚性的。此外，突触电位的增强和抑制是通过在严格控制下切换工作电压来实现的，并且通常依赖于一个额外的活性终端来调节载流子的相反极性。为了模仿高级生物体的行为并彻底改变大规模并行信息处理，人们非常需要能够表现出双语突触行为而不需要调制终端的可拉伸神经形态器件。
此外，尽管可拉伸电子器件是柔软的，但仍然容易受到严重的机械损伤(如穿刺、撕裂和切割)，这严重限制了它们在实际应用中的可靠性和寿命。因此，理想的可拉伸人工神经元还应该拥有自愈特性，在机械损伤后恢复其结构和特性。然而，由于在保留突触功能的同时整合自愈材料的挑战，还没有工作实际证明自愈和内在可拉伸的神经形态设备。
研究成果
兴奋性和抑制性神经递质在生物突触上的共存和相互作用使双语交流成为可能，成为哺乳动物的生物体适应、内部稳定以及行为和情感调节的生理基础。神经形态电子学有望模仿生物神经系统的双语功能，用于人工神经机器人和神经康复。在此，北京大学深圳研究张敏副教授团队提出了一个双语双向的人工神经管阵列，它利用离子迁移和静电耦合特性，在本质上可拉伸和自愈的聚脉聚氨醋弹性体和碳纳米管电极之间，通过范德瓦尔斯集成实现。该神经元在不同的操作阶段对同一刺激表现出抑制或增效行为，并实现了四象限的信息处理能力。这些特性使得模拟复杂的神经形态过程成为可能，这些过程涉及双语双向反应，如戒断或成瘾反应，以及基于阵列的自动刷新。此外，该神经元阵列是一种自我修复的神经形态电子设备，即使在 50%的机械压力下也能有效运作，并能在经历机械损伤后2小时内主动恢复运作。此外，双语双向可拉伸自愈式神经管可以模拟从运动皮层到肌肉的协调神经信号传输，并通过应变调制整合本体感知，类似于生物肌肉纺锤体。所提出的神经管的特性、结构、运行机制和神经综合功能标志着下一代神经康复和神经机器人的神经形态电子学的进步。相关研究以“Bilingual Bidirectional Stretchable Self-Healing Neuristors with Proprioception”为题发表在期刊ACS Nano期刊上。
图文导读
Figure 1. (a) Schematic illustration depicting the bilingual neurologic synaptic behaviors involved in complex processes in the mammalian nervous system. (b) Structure diagram of the BBSSN array: (i) SEM image of M-CNTs (scale bar, 300 nm), (ii) schematic of ion transport in PUU, and (iii) a 7 × 7 array of BBSSNs on the skin. (c) Schematic diagram illustrating the utilization of disposable IGZO as a sacrificial layer to protect the PDMS/PUU from damage during photolithography. (d) Schematic illustration showing the vdW integration between M[1]CNT electrodes and PDMS/PUU. Photographs of the BBSSN array under (e) stretching and (f) poking.
Figure 2. (a) Schematic illustration of the dynamic balance and migration of sodium ions and potassium ions across the cell membrane for signal transmission in biological neurons. (b) I−V hysteresis curves measured in sweep cycles of 0 to 1 V and 0 to −1 V for BBSSNs. Bilingual bidirectional PSC consisting of EPSC/IPSC and DPSC/HPSC triggered by (c) positive and (d) negative presynaptic pulse, respectively. Inset: schematic illustrations of ion dynamic diffusion and the electric field formed by external pulse voltage and internal interface ions, respectively. (e) SADP induced by presynaptic negative/positive pulses with different amplitudes but the same pulse duration. (f) The peak values of DPSC and HPSC with different pulse amplitudes simulate both the excitatory and inhibitory synapse modes.
Figure 3. (a) Schematic illustration of synaptic transmission in a biological neuron system. Postsynaptic current triggered by a pair of (b)positive presynaptic pulses (V = 0.5 V, W = 200 ms) and (c) negative presynaptic pulses (V = −0.5 V, W = 200 ms) with a time interval of 0.5s. A1 and A2 represent the peak values of the initial and second DPSC, respectively, while B1 and B2 represent the peak values of the initial and second HPSC, respectively. The PPD index [(A2 − A1)/A1] and PPF index [(B2 − B1)/B1] are functions of the pulse interval time (Δt)under (d) positive paired pulses and (e) negative paired pulses, and both of them are fitted with double exponential functions. (f)Representative values of the PSC when BBSSNs are stimulated by a −1.2 V pulse with a pulse duration of 200 ms, which simulates the withdrawal response to addictive substances. (g) The peak values of excitatory and inhibitory PSCs, represent the adaptation and enhancement of tolerance behaviors. The initial three stimuli are −1 V pulses with a duration of 200 ms, followed by a fourth stimulus of −2V pulse with the same duration, and the time interval between each stimulus is 2 s.
Figure 4. (a) Chemical structure of PUU. (b) FTIR spectrum and (c) Raman spectrum of the PUU. (d) Reversible dynamic disulfide exchange and multiple hydrogen bond interactions involved in the self-healing process. (e) Optical microscope images of the self-healing process of PUU at different time points after removal of the damage (scale bar, 100 μm). (f) The stress−strain behavior of PUU specimens before cutting (blue line) and after a 2 h healing period (green line). (Inset: (i) the original dumbbell-shaped specimens of PUU, (ii) the above sample being split in half, and (iii) the sample self-healing through simple contact at room temperature.) (g) In-situ conductance characterization for the self-healable M-CNT network as a function of time while undergoing multiple cuts (a constant and continuous voltage of 500 mV is applied during the measurement). (Inset: schematic representation of the self-healing mechanism of electrical conductivity in M-CNT network films spread on a self-healing substrate through dynamic reconstruction.) (h) Real-time and automatic refresh of the letters N, E, and L at a 7 × 7 post-healing BBSSN array without requiring an additional pulse to erase. (i) The stable repeated potentiation and depression processes of the post-healing BBSSNs in response to successive presynaptic pulses.
Figure 5. (a) Schematic of the neurological pathways of the reflex arc, including proprioception, which involves (i) coordination of the arm’s synergist and antagonist muscles driven by the excitatory synapse and inhibitory synapse, (ii) the muscle spindle, which protects muscles from injury caused by overstretching, and (iii) the integration of a strain perception-based artificial muscle spindle into BBSSNs. (b) The HPSC/EPSC and (c) DPSC/IPSC in response to 10 successive presynaptic pulses (V = −0.5 V, W = 200 ms) with different frequencies in BBSSNs. The SRDP index (A10/A1 × 100% and B10/B1 × 100%) is depicted as functions of presynaptic pulse rate, demonstrating the (d)high-pass filters and (e) low-pass filters, respectively. (f) HPSC/EPSC and DPSC/IPSC induced by multiple negative presynaptic pulses (V =−0.5 V, W = 200 ms) under different levels of applied mechanical strain ranging from 0% to 50%. (g) The HPSC/EPSC and DPSC/IPSC under different numbers of negative presynaptic pulses. (h) The HPSC/EPSC and DPSC/IPSC induced by multiple negative presynaptic pulses under different stretching cycles with the same tensile deformation. The dashed circles and arrows point to the relevant y-axis.
总结与展望
作者设计并制作了可实现双语双向响应的内在可拉伸和自修复的人工神经元，在模仿一系列复杂的生物行为方面显示出非凡的潜力，如抑郁和促进共存、戒断和成瘾反应以及自动刷新功能。BBSSNs 不仅可以将相反的神经信号协同传递给肌肉，而且还可以通过发射率调节肌肉运动和力量，类似于生物的肌肉纤维招募反应。此外，它们可以通过整合突触功能和应变感知来实现本体感知，以感知身体姿势并防止肌肉过度拉伸。BBSSNs 有很好的可伸展性即使在50%的应变后仍能保持各种形式的突触可塑性。此外，BBSSNs 拥有自我修复能力允许人工神经元在 2小时内从机械损伤中恢复。它们还显示了加强对生物系统理解的机会和潜力，改善了神经机器人与物理环境的互动，并恢复了神经系统疾病患者的情绪调节和运动功能。
文献链接
Bilingual Bidirectional Stretchable Self-Healing Neuristors with Proprioception
https://doi.org/10.1021/acsnano.3c03212
转自：“i学术i科研”微信公众号
如有侵权，请联系本站删除！
