Six Facts About ECRAM
- ECRAM is designed to mimic human memory synapses as high-speed RAM with low power consumption for nanotechnology systems.
- Electrochemical Random-Access Memory is designed to be used in Neuromorphic Computing, or computing through technology that mimics human biology, specifically the human brain.
- Currently, the nanomaterial 2D Titanium Carbide MXene that is used to create ECRAM is still in the research and development process.
- Electrochemical RAM is a three-terminal device composed of an insulating electrolyte, a conductive channel, an ionic reservoir, and metal contacts.
- ECRAM cells and nano-sized computer systems are still in the research phase with significant promise and potential.
- IBM, Stanford University, University of Massachusetts Amherst, and Sandia National Laboratories are leading the research and development of ECRAM nanotechnology concerning neuromorphic computing.
Sandia National Laboratories
In 2019, Sandia National Laboratories, in collaboration with Stanford University and the University of Massachusetts Amherst, published a report on discoveries of co-planar organic multilevel cells. Using conductive memory bridge devices, the research team demonstrated parallel programming in up to 3 x 3 arrays.
Individual cells in this demonstration are shown to have 1 MHz read/write cycles.
- IBM, Stanford University
- Original Use
- Synaptic Memory for Artificial Intelligence Deep Neural Networks
- Research and Development, not currently for sale
International Business Machines (IBM)
IBM submitted a report on ECRAM cells in 2019 as well. Using metal-oxide ECRAM cells, the team demonstrated parallel programming and addressing in 2 x 2 arrays.
Individual cells in this demonstration are shown to have a speed of 10 nanoseconds to write pulses.
ECRAM: How It Works
Principle of Operation
Non-volatile memory is leveraged to reduce the frequency of data transfer between data storage and processing units. ECRAM cells are used in the same manner to facilitate faster computing time with a low power cost by eliminating the Von Neumann bottleneck.
Essentially, ECRAM is a nanotechnology that emulates memory synapses in the brain. This allows for the potential to create computing systems that are not only incredibly small but incredibly powerful. Instead of a computer consisting of silicon chips with conductors between components, each component would be made up of cells like in natural biology.
In principle, electrochemical RAM functions in the same way RAM functions. It holds non-volatile memory for processing units to access without having to reach into data storage. In practice, electrochemical RAM cells are digital memory synapses. It may seem like science fiction, but ECRAM cells can be arranged in multi-leveled arrays to allow for similar functionality to the on/off process traditional computer engineers are used to.
As ECRAM is organized into cells rather than larger components, memory is not programmed by the capacity or opacity of the element. Instead, the change of channel conductivity is measured concerning the atomic species used to stress the ECRAM gate.
In simpler terms, stress is caused on the ECRAM cell by a fixed current to drive ions away from or toward the electrolyte channel. In the electrolyte channel, the current charge transfers with free carriers. As the current is inserted into the channel, the ionic charge is neutralized and the atomic species used is intercalated or bonded to the conductive host array.
Due to the multi-level cellular structure, the read function of ECRAM is performed through a separate structure from the write function. This allows for the read operation to occur with limited disturbances.
Due to the size and electrochemical nature of ECRAM cells, the programming speed of ECRAM cells can occur within a nanosecond. Instead of operation speed being limited by physical bandwidth, ECRAM cell operation speed is limited by conductivity, gate capacitance, and current types.
ECRAM cell arrays are placed in a pseudo-crossbar layout. Each cell has a gate access line in common with other devices in a row or column. This way, any change to the electrochemical potential occurs during the ionic exchange between the channel and gate electrode. Due to the relation of electrochemical potential to the array, an open circuit potential exists.
As the components are minuscule, cross-talk between cells in the array that share a gate line is prevented with an access device that isolates each cell in series with the memory element.
ECRAM: Historical Significance
AI computing power, speed, and performance depend on the constraints of software and hardware limitations. With an army of developers working diligently to improve AI function performances in every way imaginable, the largest constraint is due to the limitations of the hardware it runs on. For a time, the problem of scaling AI processes was solved with cloud computing or distributed networks.
To truly achieve the best possible performances from AI, the hardware needs to reach a level similar to human biology. That’s where an investigation into nanomaterials and nanotechnologies plays a massive role. 2D Titanium Carbide MXene is one of the materials being heavily researched for use in high-speed, low-power synapse memory. For AI, synaptic memory can change the landscape of technology forever. Deep Neural Networks (DNN) designed for AI learning tasks and classification may just be the difference between task-oriented pseudo-AI and a fully operational HAL.
While AI is a significant marker for the current era of information technology, there is a much bigger elephant in the room. Nanotechnology is dependent on the same types of system designs used for smartphones, laptops, desktops, tablets, video game consoles, and other computer systems. The upside is that engineers have a great understanding of what is needed to make these systems. The downside is the difficulty of creating nano-sized components that can be molded to any size or shape.
The applications of nanocomputers for wearables are unlimited. From tactile textures to health monitoring technology to structural engineering, nanotechnology can be used in any situation. At least it will be able to after we have the materials science required to create and maintain stable and programmable nanomachines. ECRAM and the 2D Titanium Carbide MXene used to create it are essential components for nanotechnology to advance.