Heterosynaptic Plasticity and Neuromorphic Boolean Logic Enabled by Ferroelectric Polarization Modulated Schottky Diodes

Abstract

Neuromorphic computing employs a great number of artificial synapses which transfer information between neurons. Conventional two‐ or three‐terminal artificial synapses with homosynaptic plasticity suffer from a positive feedback loop problem. Synapses with heterosynaptic plasticity are thus required to perform learning, processing and modulating simultaneously. Here, complementary metal‐oxide‐semiconductor compatible artificial synapses based on ferroelectric polarization modulated Schottky diodes (FEMOD) on silicon, which enables heterosynaptic plasticity with multi‐functionalities, high endurance, low power consumption, and high speed, are presented. High accuracy is obtained in the supervised learning simulation of artificial neural networks due to the large number of conductance states, good linearity, and small variations of FEMOD synapses. Boolean functions are demonstrated with only one or two FEMOD devices operating at low voltage and low power consumption. The proposed device structure performs multi‐functions of biological synapse and Boolean logic, thus provides high potential for the future large scale and low power neuromorphic computing applications.

Engineered Vertically Stacked NSFET Charge-Trapping Synapse for Neuromorphic Applications

“…The table shows that the proposed device achieves comparable image classification accuracy, nonlinearity LTP and LTD values, and a higher dynamic range than those in Ref . Compared with Ref − the engineered NSFET device has less linearity in conductance values, but it can still achieve comparable image classification accuracy and a higher dynamic range and PPF index. The estimated results convey that the e-NSFET can be a promising candidate for future neuromorphic applications.…”

“…Table benchmarks the results with the published work ,− in terms of dynamic range, nonlinearity values of LTP and LTD, accuracy, and PPF index %. The table shows that the proposed device achieves comparable image classification accuracy, nonlinearity LTP and LTD values, and a higher dynamic range than those in Ref .…”

“…Creating a more effective computing paradigm could potentially address the scaling limitations of complementary metal oxide semiconductor (CMOS) devices. In the pursuit of overcoming the von Neumann data transfer bottleneck, there is considerable interest in adopting an in-memory computing (IMC) configuration as the next generation of computation. − In IMC, neuromorphic devices replicate human behavior (short-term potentiation, long-term potentiation (LTP)/depression (LTD), spike time-dependent plasticity (STDP), paired-pulse facilitation (PPF)/depression (PPD)), which is used for parallel computation and adaptive learning. − In order to realize this behavior, artificial neuromorphic devices such as synapses and neurons are essential building blocks in artificial neural networks (ANNs), conventional neural networks (CNNs), and artificial intelligence (AI). − The requirement to get high accuracy in neuromorphic systems requires developing an ideal device with ideal synaptic properties …”

Heterosynaptic Plasticity in a Vertical Two-Terminal Synaptic Device

“…The von Neumann bottleneck caused by enormous data processing can be solved by neuromorphic computing systems which mimic the working system of the brain. , The neuromorphic system has typically focused on homosynaptic plasticity, which occurs only at the synaptic device in response to spiking signals from pre- and postsynaptic neurons. However, homosynaptic plasticity has a positive feedback problem that increases the probability of synapses becoming further potentiated and depressed. − As a result, synaptic weights tend to either increase to their maximum value or decrease to zero, leading to system instability. In contrast, heterosynaptic plasticity, which plays a crucial role in biological brain systems, can be simultaneously induced at each synapse following intense postsynaptic activity, avoiding the positive feedback problem and stabilizing the postsynaptic neuron’s activity. − To realize the heterosynaptic plasticity in neuromorphic systems, it is necessary to adjust the range of synaptic weight updates between the pre- and postsynaptic neurons.…”

“…However, homosynaptic plasticity has a positive feedback problem that increases the probability of synapses becoming further potentiated and depressed. − As a result, synaptic weights tend to either increase to their maximum value or decrease to zero, leading to system instability. In contrast, heterosynaptic plasticity, which plays a crucial role in biological brain systems, can be simultaneously induced at each synapse following intense postsynaptic activity, avoiding the positive feedback problem and stabilizing the postsynaptic neuron’s activity. − To realize the heterosynaptic plasticity in neuromorphic systems, it is necessary to adjust the range of synaptic weight updates between the pre- and postsynaptic neurons. Incorporating heterosynaptic plasticity into neuromorphic systems may result in more stable and robust neural activity, potentially enhancing the development of more sophisticated and intelligent neuromorphic computing systems.…”