Scientists led by Professor Nazek El-Atab of the Smart, Advanced Memory devices and Applications Lab (SAMA), Computer Electrical Mathematical Science and Engineering Division, King Abdullah University of Science and Technology (KAUST), Thuwal, Saudi Arabia, and colleagues have created a MOSCap device using Hafnium diselenide (HfSe2) that mimics neuron-like adaptive behavior and memory retention. Their findings were published in a recent study in Light Science & Applications.
Neuromorphic systems based on the structure of biological neurons. Image Credit: Light: Science & Applications
The separation of sensing, processing, and memory operations causes traditional computing systems to struggle with dynamic adaptation, which results in excessive latency and energy consumption. By simulating biological brain networks, neuromorphic computing presents a viable approach that makes data processing quicker, more energy-efficient, and more adaptive.
Neuromorphic devices can transcend traditional architecture limits by combining sensing, processing, and memory capabilities into a single device. Artificial neurons and synapses are frequently implemented using materials with programmable electrical properties or optoelectronic devices, providing a versatile foundation for developing innovative computing solutions.
Their findings enhance the field of neuromorphic technology, which aims to emulate the brain's highly efficient data processing and adaptive abilities.
The researchers accomplished this by incorporating two-dimensional HfSe2 nanosheets into the MOSCap structure. This allows the device to sense and retain light information in both “charge trapping and memcapacitive behavior within the same MOSCap device, whose threshold voltage and capacitance vary based on the light intensity,” as the researchers stated.
Electrical characterization experiments revealed a large memory window and strong memory retention, with the device preserving data stability under stressful conditions like high temperatures.
“The memory window of the device remained above the failure threshold for 106 seconds at 60–80 °C,” researchers noted, demonstrating its dependability in real-world scenarios.
Due to an efficient charge-trapping mechanism, the MOSCap also demonstrated the ability to retain data after the absence of light stimuli, highlighting its potential for energy-efficient, optoelectronic non-volatile memory.
The MOSCap architecture enables device reconfigurability by “tuning memcapacitor volatility based on biasing conditions, enabling the transition from volatile light sensing to non-volatile optical data retention,” according to the scientists.
This is a big step forward in the evolution of neuromorphic devices, displaying optoelectronic synapse capabilities and enabling “stimulus-associated learning” in which “the responsiveness of the device to light across the entire visible spectrum is notable,” according to the KAUST researchers.
One significant advantage of this innovation is the utilization of capacitive synapses, which work in the charge domain. This results in lower power consumption and leakage currents than memristive synapses. According to the KAUST researchers, capacitive synapses are excellent for small, high-density memory applications because they require little static power, enable 3D stacking, and have little sneak-path current leakage.
The researchers propose a particularly interesting application for this adaptive MOSCap in astronomy, specifically the detection of exoplanets using changes in light intensity. By incorporating the device into a leaky integrate-and-fire (LIF) neuron model, the researchers proved that the MOSCap may change firing patterns in response to light fluctuations, potentially simplifying discovering exoplanets transiting faraway stars.
“These dynamic optoelectronic neurons showed exceptional capabilities for detecting exoplanets based on their light intensity,” the researchers pointed out, emphasizing the neurons’ integration into a spiking neural network (SNN).
The MOSCap device has varied capabilities, making it a significant achievement in the field of neuromorphic technology. This achievement has the potential to spur future innovation in the development of artificial systems capable of responding to and learning from environmental cues in the same dynamic manner as biological neurons.
This study was funded by the King Abdullah University of Science and Technology (KAUST) Baseline Fund and the KAUST Transition Award in Semiconductors (Award No. FCC/1/5939).
Journal Reference:
Koolen, C. D., et. al. (2024) Scalable synthesis of Cu-cluster catalysts via spark ablation for the electrochemical conversion of CO2 to acetaldehyde. Light Science & Applications. doi.org/10.1038/s44160-024-00705-3