Neuromorphic Computing via a Bubble

The development of neuromorphic computing systems that can behave exactly like the human brain’s neural networks is one of the hottest areas of research today. To succeed in such development, is developing of the memristor which is a unique electronic component that holds the potential to revolutionize data storage and processing by mimicking the way biological synapses work. These memristors are capable of storing data of the amount of charge that has passed through them are especially useful in artificial intelligence and machine learning applications where rapid learning and adaptive behavior are critical. However, the signalling of current memristors which are mostly made from solid-state materials based on the electrons/holes instead of hydrated ions in electrolyte solution, as in our brain. To this account, recent paper published in Nano Letters and conducted by Yueke Niu, Yu Ma, and led by Professor Yanbo Xie from the Northwestern Polytechnical University, the researchers created a new type of memristor based on a liquid-vapor interface at a microbubble. This innovative approach represents a shift away from the solid-state focus that dominates the memristor field and can operate more flexibly and efficiently. The key in their experiment was the creation of a thin liquid film between a microbubble and a nuclear track membrane (NTM) that could be modulated using an external electrical field. This film’s thickness was the critical factor in generating the memristive behavior which is essentially a change in electrical resistance that depends on the history of the voltage applied to the system. Initially the authors applied varuios voltages to the system and measured the current-voltage (I-V) characteristics over different scanning periods. At very short scanning periods (below 1.6 seconds), the system behaved as a simple resistor with a linear relationship between current and voltage. As the scanning period was increased, however, the behavior changed. The researchers observed a distinct pinched hysteresis loop in the I-V curves which indicated memristive behavior at scanning periods between 1.6 and 51.2 seconds. This phenomenon where the current path depends on the history of the voltage is characteristic of memristors and reflects the device’s ability to “remember” past states. Beyond 51.2 seconds, the system transitioned to a diode-like behavior, with significantly different conductance depending on the direction of the applied voltage. Moreover, the researchers investigated how changes in voltage amplitude affected the device and found that with the increase in the voltage amplitude a more pronounced memory effects achieved with a threshold at around 1 V. At voltages above this threshold, the liquid film’s thickness changed significantly and showed stronger memristive behavior. This is important because this behavior mirrors the action potential in biological synapses where only a sufficiently large signal triggers the synaptic response. Another key finding we believe from the team’s work is evaluation of different salt concentration in the electrolyte solution because salt concentration can directly influence the liquid film’s behavior, with an optimal concentration identified around 0.1 M NaCl. At lower salt concentrations, the film’s thickness increased and reduced the device’s sensitivity to changes in voltage and weakening the memory effect. However, at concentrations above 0.1 M, the higher ionic strength led to excessive screening of charges at the liquid-vapor interface, also dampening the memory effects. The researchers said that this delicate balance in salt concentration was critical to achieve the optimal performance of the soft memristor. They also conducted more experiments to better replicate synapse-like behavior in the memristor which is a critical aspect of neuromorphic computing. They applied periodic voltage pulses and observed how the system responded, particularly its ability to “learn” by adjusting its resistance based on previous electrical stimuli. When negative voltage pulses were applied, the liquid film thickened, and the device’s conductance increased, while positive voltage pulses led to the thinning of the liquid film and caused the conductance to decrease which is similar to synaptic depression. According to the authors, these results demonstrated that the soft memristor could emulate key aspects of neural plasticity, where synapses strengthen or weaken based on activity patterns. Another exciting experiment that was reported by Professor Yanbo Xie and his team was evaluating how the device’s conductance changed over time after being subjected to repeated voltage pulses. They found that the conductance gradually decayed after the stimuli were removed which suggested that the device has a form of short-term memory similar to how biological synapses retain information briefly after stimulation. We believe this data is important for applications in neuromorphic computing where systems need to mimic the brain’s ability to process and store information in a dynamic, adaptive manner. Additionally, the experiments showed that the system’s behavior was repeatable and stable over multiple cycles of operation and of course this repeatability is a critical factor in making the device practical for real-world applications because it ensures that the memristor can reliably perform the same tasks without degradation over time.

In conclusion, Professor Yanbo Xie and his colleagues successfully developed what can be considered first in class memristors that is based on soft fluidic system rather than the conventional solid-state materials which open the door wide to a more flexible and dynamic platform for neuromorphic computing and we believe this innovative design will bring the memristor a step closer to functioning in a manner that truly mimics biological synapses with huge application in artificial intelligence and brain-inspired computing systems. Moreover, the fluidic nature of the new memristor allows for a system that can change and adapt in response to environmental conditions similar to the neural networks in the human brain and it is this adaptability which makes the soft memristor a compelling candidate for use in neuromorphic computing systems, where learning and memory are not only advantageous but essential. Additionally, the new design of fluidic nature of this memristor could make it more suitable for environments where traditional solid-state devices would be impractical such as in flexible electronics or biomedical implants and this responsiveness to voltage and ionic conditions hints at future applications in sensors and bioelectronics, where a more organic and adaptable approach to computing is needed.