Frequency-Dependent Synapse Weight Tuning in 1S1R with a Short-Term Plasticity TiO_x-Based Exponential Selector

Abstract

Short-term plasticity (STP) is an important synaptic characteristic in the hardware implementation of artificial neural networks (ANN), as it enables the temporal information processing (TIP) capability. However, the STP feature is rather challenging to reproduce from a single nonvolatile resistive random-access memory (RRAM) element, as it requires a certain degree of volatility. In this work, a Pt/TiO x /Pt exponential selector is introduced not only to suppress the sneak current but also to enable the TIP feature in a one selector-one RRAM (1S1R) synaptic device. Our measurements reveal that the exponential selector exhibits the STP characteristic, while a Pt/HfO x /Ti RRAM enables the long-term memory capability of the synapse. Thereafter, we experimentally demonstrated pulse frequency-dependent multilevel switching in the 1S1R device, exhibiting the TIP capability of the developed 1S1R synapse. The observed STP of the selector is strongly influenced by the bottom metal–oxide interface, in which Ar plasma treatment on the bottom Pt electrode resulted in the annihilation of the STP feature in the selector. A mechanism is thus proposed to explain the observed STP, using the local electric field enhancement induced at the metal–oxide interface coupled with the drift-diffusion model of mobile O2– and Ti3+ ions. This work therefore provides a reliable means of producing the STP feature in a 1S1R device, which demonstrates the TIP capability sought after in hardware-based ANN.

The Effect of Carbon Doping on the Crystal Structure and Electrical Properties of Sb2Te3

“…In this case, we used a voltage pulse with a pulse width of 500 ns to RESET the PCRAM in a low resistance state, and the continuously adjustable resistance value can be obtained as shown in Figure 10 a. With this change, maybe we can realize several basic synaptic functions at the cell level, including long-term plasticity (LTP) [ 41 , 42 ], short-term plasticity (STP) [ 41 , 42 ], spike timing-dependent plasticity (STDP) [ 43 , 44 ], and spike rate-dependent plasticity (SRDP) [ 44 , 45 ], and maybe can also realize more complex or higher-order learning behaviors at the network level, such as supervised learning [ 46 ] and associative learning [ 47 ], as well as non-von Neumann architecture of in-memory computing [ 48 , 49 ]. In general, for this phenomenon of continuous resistance change, the resistance drift caused by the widening of the band gap due to the structural relaxation (SR) of amorphous Sb 2 Te 3 is a great obstacle to multilevel storage, neuromorphic learning and in-memory computing.…”

The impact of oxygen vacancy defect density on the nonlinearity and short-term plasticity of TiO2-based exponential selector

“…In our previous work, the short-term plasticity of the Pt/ TiO 2 /Pt exponential selector can be explained by the drift and diffusion of mobile Ti 3+ and O 2− ions at positive and negative voltages, respectively [18]. As the sputtering pressure increases, the composition of non-lattice oxygen increases.…”

“…Chao Du et al utilised the short-term plasticity of a WO x -based dynamic memristor in a reservoir computing system to perform digit recognition [17]. In addition, in our previous work, the short-term plasticity of the Pt/TiO x /Pt exponential selector was demonstrated to enable the stimulation ratedependent weight tuning in the one selector-one RRAM (1S1R) synapse [18]. Samuel Shin et al tuned the short-term plasticity of an organic mixed ionic-electronic conductor memristor by adding LiClO 4 salts [19].…”

Artificial Neuron Devices

“…The RRAM conductance, which can be modulated by electrical pulse of different polarities, is just analogous to the synaptic weight updating. In addition, the multibit conductions of RRAM can mimic the multilevel synaptic weight. , Nonvolatile RRAM devices are well suited for emulating long-term synaptic plasticity, , whereas volatile devices are ideal for simulating short-term plasticity. , When using memristor to emulate neuronal functions, compared to nonvolatile device-based neurons, artificial neurons based on volatile devices require a simpler auxiliary circuit because the nonvolatile one typically needs a comparator to establish a threshold value and requires an extra circuit to reset the device to its original state, which incurs high area and energy costs. , …”