Matveyev Yury, Kirtaev Roman, Fetisova Alena, Zakharchenko Sergey, Negrov Dmitry, Zenkevich Andrey
Moscow Institute of Physics and Technology, Dolgoprudny, 141700, Russia.
Nanoscale Res Lett. 2016 Dec;11(1):147. doi: 10.1186/s11671-016-1360-6. Epub 2016 Mar 15.
Crossbar resistive switching devices down to 40 × 40 nm(2) in size comprising 3-nm-thick HfO2 layers are forming-free and exhibit up to 10(5) switching cycles. Four-nanometer-thick devices display the ability of gradual switching in both directions, thus emulating long-term potentiation/depression properties akin to biological synapses. Both forming-free and gradual switching properties are modeled in terms of oxygen vacancy generation in an ultrathin HfO2 layer. By applying the voltage pulses to the opposite electrodes of nanodevices with the shape emulating spikes in biological neurons, spike-timing-dependent plasticity functionality is demonstrated. Thus, the fabricated memristors in crossbar geometry are promising candidates for hardware implementation of hybrid CMOS-neuron/memristor-synapse neural networks.
尺寸小至40×40纳米（2）、包含3纳米厚氧化铪层的交叉电阻式开关器件无需形成过程，并展现出高达10（5）次的开关循环。四纳米厚的器件显示出双向渐变开关的能力，从而模拟类似于生物突触的长期增强/抑制特性。无需形成过程和渐变开关特性均根据超薄氧化铪层中氧空位的产生进行建模。通过向具有模拟生物神经元尖峰形状的纳米器件的相对电极施加电压脉冲，证明了尖峰时间依赖可塑性功能。因此，交叉结构中制造的忆阻器是混合CMOS-神经元/忆阻器-突触神经网络硬件实现的有前途的候选者。
