Scientists from ITMO University and Harbin Engineering University have developed a perovskite nanomeemristor capable of withstanding more than 1,500 rewrite cycles and maintaining stability for months. This was reported in the press service of ITMO. The new component promises to accelerate the creation of ultra-compact and energy-efficient processors for artificial intelligence and machine learning.
A memristor is an electronic element that changes resistance depending on the current passing through it. It can be used both for storing information and for processing it with minimal energy consumption, inferior in this parameter to conventional silicon transistors. Perovskites, promising materials for microelectronics, could not previously be used in real devices due to instability. The new development of scientists from ITMO and the Ioffe Institute solves this problem.
The nanomeemristor is made in the form of single-crystal nanocubes of cesium lead bromide — one of the most chemically resistant perovskites based on lead halide. The monocube is located between indium tin oxide and boron-doped diamond — chemically inert electrodes that ensure switching stability. The dimensions of the element are only 130–160 nanometers, and the power consumption is 70–80 nanowatts, which is several times less than that of analogues.
According to researchers, the technology allows creating not only single memristors, but also scalable circuits (crossbars) for implementing logic and switching semiconductors. This opens the way to the production of compact neuromorphic processors capable of processing data with high speed and low power consumption — a key requirement for artificial intelligence applications.