Technical / research

Researchers from Taiwan's National Cheng Kung University developed RRAM memory based on lead-free Na0.5K0.5NbO₃/HfO perovskite materials. The researchers fabricated a device using low-temperature sputtering, and report high performance and excellent process compatibility. 

The researchers say that the new device exhibits forming-free switching, low operating voltages (−1 V/1.5 V), a high ON/OFF ratio of ∼200, and endurance exceeding 105 cycles.

Researchers from Peking University in Beijing, China, working with the Beijing Advanced Innovation Center for Integrated Circuits, have proposed a protection framework for RRAM-based deep neural network (DNN) accelerators. In this framework, the researchers split accelerator security into two distinct levels. The framework, called OmniGuard, achieves a 1.33x to 4.38x speedup and 1.26x to 2.65x power savings, with a reported 5% energy overhead and 3% area overhead.

RRAM-based computing-in-memory (CIM) accelerators have become one of the mainstream options for running neural network inference on edge devices, as performing the multiply-accumulate operation inside the memory array itself avoids the data movement that dominates the energy budget of conventional von Neumann designs. But edge hardware sits in the field, physically accessible to whoever holds it, and the trained model weights programmed into the RRAM array are usually the most valuable asset on the chip. The researchers note that these security vulnerabilities are a real obstacle to deploying RRAM-based accelerators commercially.

Researchers from the College of Electronic and Optical Engineering at Nanjing University of Posts and Telecommunications (NJUPT) in Nanjing, China, have designed, fabricated and measured a single-pole single-throw (SPST) RF switch built on an RRAM process. The switch operates across the DC-Ku band with a measured insertion loss below 1 dB, and it holds its state without any standing bias.

RF switches sit in the front-end of nearly every wireless system, routing signals between transmit and receive paths. The standard options used today, field-effect transistors, PIN diodes and RF MEMS, all share a drawback: they need a continuous bias voltage or drive current to stay in either the ON or the OFF state. Over long operating periods that translates into meaningful static power draw and thermal management headaches, which is exactly the wrong trade-off for battery-powered IoT nodes and wearables.

Researchers at China's Tsinghua University, together with InnoMem Technologies, have developed a 16Mb embedded RRAM device with a hybrid array structure, built on a 28 nm process, that is highly suited for automotive OTA update applications.

The researchers say that the new 1T1R/2T2R hybrid array structure design enhances the reliability and system compatibility. The researcher used several technologies to increase performance - a load balancing approach to address the IR-drop issue and improve cell uniformity, and a hierarchical column selection strategy to suppress the leakage current in the write path, optimizing write operations at high temperature.

Researchers from the University of Michigan have discovered the role of phase separation in memristor mechanism, thus paving the way to improving RRAM memory. 

Up until now, researchers were not sure why memristors retain their memory for a long time, as current models did not explain this fully. The new researchers explain that in memristors, oxygen ions prefer to be away from the filament and will never diffuse back, even after an indefinite period of time.

Researchers from The University of California, Berkeley, designed a new versatile compact RRAM model that can model different types of RRAM devices such as oxide-RRAM (OxRAM) and conducting-bridge-RRAM (CBRAM). 

The researchers say that their new model unifies the switching mechanisms of these RRAMs into a single framework. The researchers showcase the model’s accuracy in reproducing published experimental device DC and transient characteristics of various RRAM structures. They also demonstrated the model’s efficacy in capturing RRAM variability and conducting 1T1R circuit simulations.

In 2017, Australia-based RRAM developer 4DS Memory signed an agreement with Belgium-based IMEC to develop a transferable manufacturing process for its technology. 4DS now announced that it has renewed its R&D collaboration with the IMEC research institute after receiving positive results from its fourth lot of test chips.

4DS will pay IMEC 1.92 million Euro for the activities in 2024, and the company expects IME to deliver its 5th and 6th batch of 1 Mbit array chips in Q3 2024. The 6th batch will be based on a 20 nm process. 

US-based Tetramem published a paper that shows its form of RRAM-powered analog computing is capable of executing calculations with arbitrary precision. It says the ability to perform high-precision multiplication within single electronic devices that can be readily formed in arrays offers scope to reduce the power consumption of machine learning when based on artificial neural networks.

The Tetramem device is made of a mixture of Al3O2, above a layer of HfO2 sandwiched between a tantalum/titanium top electrode and a platinum bottom electrode. Each of the bilayers is less than 1nm thick so that after being laid down they appear to form a mixed layer rather than two separate continuous layers. The device was fabricated in a 240-nm diameter via above the CMOS peripheral circuitry.

Researchers from Kyushu University and the National Taiwan Normal University developed a new RRAM device, readable through both electrical and optical methods. The device is based on perovskite quantum dots that enable to simultaneously store and visually transmit data.

By integrating a light-emitting electrochemical cell with a resistive random-access memory that are both based on perovskite, the team achieved parallel and synchronous reading of data both electrically and optically in a ‘light-emitting memory.’

The project NEUROTEC (“Neuro-inspired artificial intelligence technologies for the electronics of the future”) was launched in November 2019 to develop innovative "Beyond von Neumann" concepts for highly energy-efficient devices. The two-year project shows the great potential of a future neuromorphic computer.

The project aims to fuse two major technologies - machine learning and artificial neural networks (ANNs) and memristive materials and devices - especially redox-based RRAM and phase change memories (PCM). The project's mandate is to develop a full-range of basic technologies ranging from dedicated material deposition technologies, integration technologies, measurement technologies, the development of simulation and modelling tools, up to the design and realization of novel AI circuits.