研究背景
神经形态计算从生物原理中获得灵感，创造出高能效的人工神经元和突触，实现类似大脑的学习和记忆，并可用于解决高度复杂的任务。忆阻器通常被视为神经形态系统中人工突触最普遍的设备级元件之一，因为它们能够以生物方式实现突触的可塑性，并在建立节能电路方面具有巨大的潜力，提供高性能的硬件平合来进行高级信息处理。人们提出了大量的模拟记忆体材料，如金属氧化物、非晶硅、过氧化物、有机材料等。特别是有机聚合物忆阻器已经被广泛研究，原因包括低成本和易加工、可调控的性能和灵活性等。
有机聚合物忆阻器是一种金属-绝缘体-金属(MIM)夹层结构，利用金属-有机混合聚合物、聚合物纳米复合材料和聚合物电解质等材料，已被探索用于实现模拟忆阻器的连续变化的器件电导。然而，大多数报道的有机聚合物忆阻器往往表现出低热耐受性，导致器件性能下降。特别是，热稳定性是有机聚合物忆阻器作为人工突触器件应用于极端温度条件下的神经形态系统的关键因素，如行星探测机器人和深空探测器。关于有机聚合物记忆体的热稳定性的报道屈指可数，而且所展示的工作温度范围仍未达到先进应用的要求包括在军事和航空航天领域的应用。
研究成果
神经形态电子学的灵感来自于大脑的工作方式，为成功实现智能人工系统带来了巨大的希望。在几个神经形态的硬件问题中，在极端温度下的强大设备功能对实际应用特别重要。鉴于用于人工突触的有机记忆体已经在室温下被证明，在极低或极高温度下实现强大的器件性能仍然是一个巨大的挑战。在这项工作中，作者通过调整基于溶液的有机聚合物记忆器的功能来解决温度问题。优化后的记忆体在低温和高温环境下都表现出了可靠的性能。利用X射线光电子能谱(XPS)飞行时间二次离子质谱 (TOF-SIMS)深度剖析，通过比较新鲜和写入的有机聚合物记忆体的组成概况，揭示了器件的工作机制。由外加电压诱导的可逆离子迁移有助于记忆体的特征开关行为。在此，在极端温度下实现的稳健的记忆体反应和经过验证的器件工作机制都将明显加快记忆体在神经形态系统中的发展。相关研究以“Polymeric memristor based artificial synapses with ultra-wide operating temperature”为题发表在Advanced Materials期刊上。
图文导读
Figure 1. Schematics of (a) two-terminal memristive device in ITO/MDMO-PPV/Al configuration and (b) molecular structure of MDMO-PPV. (c) Cross-sectional SEM image of MDMO-PPV memristor. Scale bar: 100 nm. (d) I–V characteristics of the MDMO-PPV memristor under positive (left) and negative (right) voltage sweeps. (e) Currents extracted from the I-V curves at ± 8 V plotted as a function of number of cycles.
Figure 2. (a) Schematic illustration of a biological synapse. (b) Current dependence recorded during the application of ten stimulation pulses with different pulse amplitudes (7, 8, and 9 V). (c) PPF index (defined as (A2-A1)/A1*100 %) versus pulse interval by applying a paired pulse. (d) Current responses to ten identical stimulation pulses (8 V, 100 ms) at different frequencies (0.1 Hz, 0.5 Hz, 1.0 Hz, 2.0 Hz, 3.3 Hz, and 5.0 Hz). (e)Current responses to presynaptic pulses with different pulse widths (0.1 s, 0.2 s, 0.5 s, and 1.0 s). The EPSC response was continuously monitored using a reading voltage of 0.2 V for 1 s. STDP implementation in the MDMO-PPV memristor-based artificial synapse. Asymmetric STDP of (f) Hebbian learning rule and (g) anti-Hebbian learning rule. The shape of presynaptic and postsynaptic spikes used in the STDP measurements is shown as insets. (h) STP–LTP transition. LTP were triggered by applying consecutive spikes (N = 5 – 70, V = 8 V, ΔT = 100 ms), where ΔT denotes spike duration. The retention data was recorded at a reading voltage of 0.2 V. (i) Relaxation time τ as a function of number of presynaptic spikes.
Figure 3. XPS depth profiles of the MDMO-PPV memristor in (a) initial state, (b) LRS, and (c) HRS. Schematic diagrams illustrating the proposed memristive switching mechanisms of the MDMO-PPV memristor in (d) initial state, (e) LRS, and (f) HRS. ToF-SIMS 3D images of the (g) initial state, (h) LRS, and (i) HRS MDMO-PPV memristors.
Figure 4. (a) Current response to successive spikes (Vpos=8 V, Vread=1 V, ΔT =100 ms) measured in real time at different temperatures (173 K- 473 K). (b) The conductance change of the device measured after thermal annealing at a high temperature of 573 K. (c) The changes of I1 and I10/I1 for MDMO-PPV memristor after storing at 77 K for different periods. I10/I1 is used to determine the ability of an artificial synapse to change in strength in response to external voltage. (d) Device stability characterized at different temperatures ranging from 77 K to 573 K, with three sets of data recorded at each temperature. (e) Summary of limiting temperature for memristors reported in the literature.
总结与展望
利用有机聚合物MDMO-PPV，作者成功地展示了一个人工突触装置，它类似于生物突触的特性，具有良好的突触模拟能力。XPS 和TOF-SIMS 深度剖析验证了可逆的离子迁移，从而产生了特征性的记忆器反应。此外，该无封装突触装置具有超宽的温度适用性(77 K至573 K)。在此，该研究填补了人工突触器件在极端温度条件下应用的空白，并为未来神经形态电子学在各个领域的应用铺平了道路，如自动机器人和航空航天探索。
文献链接
Polymeric memristor based artificial synapses with ultra-wide operating temperature
https://doi.org/10.1002/adma.202209728
转自：“i学术i科研”微信公众号
如有侵权，请联系本站删除！
