Transition from noise-induced to self-sustained current spiking generated by a NbOx thin film threshold switch
Transition from noise-induced to self-sustained current spiking generated by a NbOx thin film threshold switch
复制标题
DOI：
10.1063/5.0030918
复制
发表时间：
2021-01-11
影响因子：
4
通讯作者：
Borisov, P.
中科院分区：
文献类型：
--
作者：
Johnson, B. A.;Brahim, K.;Borisov, P.
在具有NbOx薄膜器件的电路中研究了向自持电流振荡的转变，该器件充当人工神经元。在所施加电压的阈值以上，电路开始表现出持续但相当随机的电流尖峰，其随着所施加电压的增加而变得更有规律。实验测量揭示了两个缩放区域的interspike间隔依赖于源极电压：初始对数依赖，过渡到幂律的指数为-1/2。使用人工神经元电路的数值模型，我们观察到与我们的实验数据很好的协议。我们表明，对数标度行为与噪声引起的尖峰，而幂律是由于噪声的自我维持振荡，这将存在于系统中，即使没有噪声。这些发现强调了噪声在NbOx阈值开关的动态中的重要性，并且揭示了它们对输入信号的变化的尖峰响应。所报道的NbOx人工振荡神经元中的尖峰控制方法可以帮助开发用于神经形态神经网络和基于振荡器的计算系统的最优设计。
A transition to self-sustained current oscillations was investigated in a circuit with a NbOx thin film device, acting as an artificial neuron. Above a threshold of the applied voltage, the circuit begins to demonstrate persistent but rather random current spiking, which becomes more regular with increasing applied voltage. Experimental measurements reveal two scaling regions of the interspike interval dependence on source voltage: an initial logarithmic dependence, which transitions to a power law with an exponent of -1/2. Using a numerical model of the artificial neuron circuit, we observe good agreement with our experimental data. We show that the logarithmic scaling behavior is associated with noise-induced spiking, while the power law is attributed to noisy self-sustained oscillations, which would exist in the system even with the absence of noise. These findings stress the importance of noise in the dynamics of NbOx threshold switches and shed light on their spiking response to variation of the input signals. The reported approach to spiking control in NbOx artificial oscillation neurons can help to develop an optimal design for neuromorphic neural networks and oscillator-based computing systems.