In collaboration with Stanford University, KTH Royal Institute of Technology has developed a material that can be used to make computer components similar to the human brain.
2D titanium carbide was used to make electrochemical random access (ECRAM) memory components. This technology has the potential to complement classical transistor technology and contribute towards the commercialization of powerful computers modeled after the brain’s neural network. These neuromorphic computers are thousands of times more efficient than current computers.
These advancements in computing are possible due to fundamental differences from the current computing architecture. Professor Max Hamedi, the KTH Associate, said that the ECRAM is a component that acts like a synaptic cell in an Artificial Neural Network.
Hamedi states that transistors are not always on and off and that information needs to be transferred back and forth between memory and processor. Instead, these new computers use components that can have multiple states and perform in-memory computing.
KTH and Stanford scientists have been working together to develop better materials for the ECRAM. This component switches by inserting ions in an oxidation channel. It is similar to how our brain works with ions. Materials that can overcome the slow kinetics and poor temperature stability of plastics have been required to make these chips economically viable.
MXene, a two-dimensional (2D) compound made of titanium carbide (Ti3C2Tx), is the key material used in the ECRAM units. Hamedi says that the MXene combines organic chemistry’s high speed with inorganic material integration in one device.
Stanford University’s Alberto Salleo, the co-author, said that MXene ECRRAMs combine the speed and linearity, write noise, switching energy, endurance metrics, and speed necessary for parallel acceleration artificial neural networks.
Salleo states that MXenes, a family of materials with high-temperature stability, can be used to integrate with traditional electronics. They also have a wide composition range to maximize performance. While there are still many hurdles to overcome before people can purchase their neuromorphic computer, Hamedi believes the 2D ECRRAMs represent a significant breakthrough in neuromorphic materials. This could lead to artificial intelligence capable of adapting to confusion and nuance like the brain, with thousands of times less energy consumption. It can make portable devices that can perform heavier computing tasks without relying on the cloud.
