Real-Time Time Series Processing Achieves Ultra-Low Power with IonTronic Memristor Circuits

Researchers are tackling the significant challenge of energy consumption in real-time time series processing with a novel approach utilising iontronic memristor circuits. T. M. Kamsma, Y. Gu, and C. Spitoni, alongside M. Dijkstra, Y. Xie, R. van Roij and colleagues from the Institute for Theoretical Physics, Utrecht University, present a pathway towards circuits that promise exceptionally low power usage , exceeding five orders of magnitude less than conventional solid-state systems. This work directly addresses a key hurdle in the field of iontronics, providing concrete performance comparisons on established benchmark tasks and demonstrating the potential for ultra-low-power applications in processing signals of natural origin. Their open-source pyontronics package integrates modelling of Kirchhoff-governed circuits with dynamic internal voltages, offering a powerful tool for future exploration and development in this rapidly expanding paradigm.

Iontronic Reservoir Computing via Kirchhoff Circuit Simulation

Scientists have identified iontronic neuromorphic computing as a rapidly expanding paradigm. Researchers are tackling the significant challenge of energy consumption in real-time time series processing with a novel approach utilising iontronic memristor circuits. This work directly addresses a key hurdle in the field of iontronics, providing concrete performance comparisons on established benchmark tasks and demonstrating the potential for ultra-low-power applications in processing signals of natural origin. The emergence of angstrom-confined iontronic devices enables ultra-low power consumption with dynamics and memory timescales intrinsically aligned with signals of natural origin, presenting a challenge for conventional (solid-state) neuromorphic materials.
However, comparisons to earlier conventional substrates and evaluations of concrete application domains remain a challenge for iontronics. Researchers propose a pathway toward iontronic circuits. Scientists that can address established time series benchmark tasks, enabling performance comparisons and highlighting possible application domains for efficient real-time time series processing. We model a Kirchhoff-governed circuit with iontronic memristors as edges, while the dynamic internal voltages serve as output vector for a linear readout function, during which energy consumption is also logged. All these aspects are integrated into the open-source pyontronics package.

Without requiring input encoding or virtual timing mechanisms, our simulations demonstrate prediction performance comparable to various earlier solid-state reservoirs, notably with an exceptionally low energy consumption of over 5 orders of magnitude lower. These Results suggest a pathway of iontronic technologies for ultra-low-power real-time neuromorphic computation. The emerging field of iontronic neuromorphic computing has grown at an extraordinary rate in recent years. Consequently, exciting demonstrations of iontronic computing have been presented recently, including fluidic crossbar arrays, (integrated) logic circuits, and reservoir computing.

These initial advancements increase the achieved functionality of iontronic circuitry for computing purposes, although they still remain at relatively simple levels for now. Consequently, power consumption estimates of fluidic circuits remain sparse as the focus is often on the power of individual device components. Moreover, power comparisons to other substrates on established tasks are challenging as many time-series benchmarks for comparisons remain out of reach. Thus, highlighting potential avenues for applications of iontronic neuromorphic devices remains difficult, especially when this ultimately has to be compared to the ultra-high performances of well-established solid-state technologies.

Here we aim to both advance the theoretical guidance towards interacting ion- tronic circuits, including circuit-wide power estimates for per- forming established time-series benchmarks, and to advance the mapping out of potential domains of relevance for ion- tronic technologies. We highlight possible applications in efficient real-time processing of time series by exploiting the combined ingre- dients of (i) intrinsic alignment of the ms to s time scales of iontronic devices with those of signals of natural origin, (ii) iontronic dynamics that align well with established reservoir computing protocols, and (iii) the possibility that these features can scale down to angstrom-confined iontronic de- vices of ultra-low power consumption. Especially the first feature of natural timescales is notable, because achiev- ing this within conventional materials proves challenging and many conventional neuromorphic circuits require additional hardware to artificially slow down its components or have to rely on a virtual time that is detached from real-time. This requires additional overhead and energy consumption.

Iontronic memristors on the other hand can scale down to angstrom confined devices, enabling low power consumption per channel, while retaining relevant millisecond to second timescales. There- fore, while the Results in this work will be in the ∼1−10ms regime, all Results should equivalently hold for slower inputs of e.g0.1−100s time scales (or anything in between), opening the window for an excitingly wide range of time series. Although our Results are in principle aimed to be representative for a wide class of iontronic memristors, we base the conductance properties here on experimentally realised ion- tronic memristors, consisting of conical ion channels embed- ded in a membrane.

Iontronic circuits deliver ultra-low power time series prediction

Scientists achieved a groundbreaking reduction in energy consumption using iontronic circuits for real-time time series processing.The research team modelled a Kirchhoff-governed circuit with iontronic memristors, demonstrating prediction performance comparable to existing solid-state reservoirs, but with an energy consumption over five orders of magnitude lower. Experiments revealed that this exceptional efficiency stems from the intrinsic alignment of iontronic device timescales, ranging from milliseconds to seconds, with the natural timescales of many signals. Data shows the team successfully integrated all aspects of their work into the open-source pyontronics package, facilitating further research and development in this emerging field.

The team measured the steady-state conductance of 12μm thick membranes, observing voltage-dependent behaviour crucial for memristive function. Detailed dynamic current-voltage measurements, conducted on membranes of both 2.5μm and 12μm thicknesses, revealed clear memristive hysteresis loops across varying frequencies. These loops, confirmed by the team’s state-variable model (SVM) theory with timescales of 1ms and 10ms for the short and long channels respectively, demonstrate the dynamic conductance changes essential for information processing. Results demonstrate the ability to directly apply time series inputs as imposed voltages, with the evolving internal voltages serving as the output vector for a linear readout function.

Measurements confirm that the modelled fluidic circuit, containing 15 iontronic memristors, accurately reflects the behaviour of a wide class of iontronic memristors. The researchers calculated total internal Ohmic power dissipation based on the conductances and voltages within the iontronic circuit, providing a comprehensive energy consumption profile. Specifically, the team’s simulations show that the system requires no input encoding or virtual timing mechanisms, further contributing to its ultra-low power operation. The breakthrough delivers a pathway towards ultra-low-power neuromorphic computation, potentially enabling efficient real-time processing of time series data in diverse applications.

The study highlights the potential of angstrom-confined iontronic devices, which retain relevant millisecond to second timescales while enabling exceptionally low power consumption per channel. Tests prove the feasibility of this approach, building upon previously demonstrated experimental results and realistically achievable technologies. The team’s work provides a concrete foundation for exploring applications beyond traditional neuromorphic computing, including areas where biological compatibility, ion selectivity, and chemical regulation are paramount. The pyontronics package facilitates the exploration of these possibilities, offering a powerful tool for simulating and analysing iontronic circuits.

Iontronic Memristors Enable Picowatt Time Series Prediction

Scientists have demonstrated a simulated fluidic circuit utilising iontronic memristors capable of performing time series prediction with remarkably low energy consumption. This research establishes a pathway towards ultra-low-power real-time time series processing by modelling a Kirchhoff-governed circuit where iontronic memristors function as edges and dynamic internal voltages serve as output vectors for a linear readout function. Simulations reveal performance comparable to existing solid-state reservoirs, achieving an energy consumption over five orders of magnitude lower, estimated at approximately 16 picowatts for the full circuit. The significance of these findings lies in the potential for iontronic technologies to address the growing demand for energy-efficient computing.