• Universitat de València

• Carlos David Prado Socorro

The development of technologies such as computer vision, artificial hearing and artificial intelligence has created a growing demand for systems capable of processing and storing large volumes of data quickly and efficiently. However, current devices separate data processing and storage into independent components. This separation not only implies a high energy cost but also makes it difficult to quickly and efficiently handle large amounts of data.

Memristive devices combine information processing and memory storage into a single component, similar to how biological neurons work. However, memristors developed to date offer long-term memory but only allow access to a limited number of electronic states, which restricts their use in neuromorphic computing. Furthermore, they are unstable and poorly reproducible, so there is a great variability of properties between devices.

Memristive devices based on ion migration currently present two major limitations that make their commercial use difficult. First, they need high voltages to operate, which makes the system energy-intensive, unstable, and makes electronic states difficult to control. Second, these systems are memoryless, as information is erased immediately after being written, severely limiting their practical usefulness.

Therefore, it is essential to develop advanced materials that allow data to be pre-processed efficiently directly in the sensor or capture device with a low energy cost, and have an information storage capacity adequate for current technologies.

Researchers at the Universitat de València have developed memristive devices that solve the problem of low electrical conductivity and limited information retention time in memristive systems based on ionic migration. These devices let in-memory computing be implemented with memristors based on ionic migration, eliminating the separation between memory and processing and allowing a system with multiple accessible electronic states that emulate neural processes and manage to perform processing and storage operations simultaneously. This significantly improves system efficiency and reduces latency by eliminating the need to transfer data between separate memory units and processors.

The developed memristive device is capable of processing and retaining information through changes in the electrical conductivity of the system, controlled by the applied voltage history, and can operate at lower voltages than existing memristors.

Furthermore, thanks to its memory capacity, the developed system does not require the constant application of voltage to maintain its state.

Likewise, the biocompatibility of both the organic polymer matrix and the two-dimensional materials of the developed devices permits their implementation in the biodevice market. This allows electronic elements to be included in biological tissues to be used as sensors, markers or prostheses.

The main application of the technology focuses on information preprocessing, emulating the neurosynaptic processes of the brain in microelectronics. In this sense, its main sector of interest is in artificial vision and hearing, where it allows information to be processed and stored in the same device, allowing it to operate continuously and almost instantaneously. Memristive devices can be used in:

• neuromorphic sensors for artificial vision
• neuromorphic processors for artificial hearing
• bionic implants with integrated memory
• edge artificial intelligence devices
• medical diagnostic services

The main advantages of the invention are:

• high energy efficiency and cost reduction
• high operational stability and improvement in ionic mobility
• long-term memory and efficient storage of information

• Patent applied