Chemically Regulated Conical Channel Synapse for Neuromorphic and Sensing Applications

TL;DR

The paper addresses how to realize synaptic-like plasticity in a fluidic iontronic system by coupling ion concentration polarization (ICP) with Langmuir-type surface reactions on a conical channel. The authors solve the coupled Poisson-Nernst-Planck-Stokes equations with Langmuir kinetics, using finite-element simulations and a first-principles analytic approximation to reveal fast voltage-driven charging ($\tau \approx L^2/(12D) \approx 5.6\ \text{ms}$) and slow surface discharging ($kK\rho_B \sim O(0.01)\ \text{s}^{-1}$) that produce STP/LTP. They demonstrate functional analogues of short- and long-term potentiation/depression, frequency-dependent plasticity, and NMDA-like chemical-electrical coincidence detection, including spike-timing-dependent plasticity windows. The work establishes a versatile, tunable platform for chemically regulated neuromorphic iontronics with potential sensing and computing applications.

Abstract

Fluidic iontronics offer a unique capability for emulating the chemical processes found in neurons. We extract multiple distinct chemically regulated synaptic features from an experimentally accessible conical microfluidic channel carrying functionalized surface groups, using finite-element calculations of continuum transport equations. By modeling a Langmuir-type surface reaction on the channel wall we couple fast voltage-induced volumetric salt accumulation with a long-term channel surface charge modulation by means of fast charging and slow discharging. These nonlinear charging dynamics emerge across several orders of magnitude of reaction rates and equilibria, and are understood through an analytic approximation rooted in first-principles. We show how short-and long-term potentiation and depression, frequency-dependent plasticity, and chemical-electrical signal spike-timing dependence and coincidence detection (acting like a chemical-electrical AND logic gate), akin to the NMDA mechanism for Hebbian learning in biological synapses, can all be emulated.

Figures (4)

• Figure 1: Schematic of an azimuthally symmetric conical channel, carrying functionalized surface groups, of length $L$, base and tip radii $R_{\mathrm{b}}$ and $R_{\mathrm{t}}<R_{\mathrm{b}}$, connecting two reservoirs of an aqueous 1:1 background electrolyte with low concentrations of trivalent and divalent reactant ions A$^{3-}$ and B$^{2-}$. Far from the base and tip the concentrations of ion species $i$ read $\rho_i=\rho_{i,\mathrm{b}}+\Delta\rho_i(t)$ and $\rho_i=\rho_{i,\mathrm{b}}$, respectively. At the base, $\Delta\rho_i(t)$ facilitates optional chemical signals and gradients, alongside electric stimulation via an applied voltage $V(t)$.
• Figure 2: Chemically regulated surface charge and consequent conductance changes upon an 1.2 V pulse for $t\in[0.5,1]$ s. (a) The voltage pulse drives fast ICP, altering reactant ion ratios near the wall, charging the surface, and increasing bulk and surface conductance respectively. After the pulse, salt depletes quickly, while slow surface discharge creates the LTP. (b) Average channel surface charge $\overline{\sigma}$ from full finite-element (FE) calculations and our first-principles analytic approximation (AA), showing qualitative agreement. The inset shows $|\partial_t\overline{\sigma}|$ as a function of $\overline{\sigma}/\sigma_{Eq}$ for $V=0$ V (magenta) and $V=1.2$ V (orange). Charging at $V=1.2$ V (red circle) is orders of magnitude faster than discharging at $V=0$ V (green square). (c) Channel conductance $g(t)/g_0$ from full FE calculations. After the pulse the concentration reverts back to equilibrium quickly, forming the volatile STP, while the surface discharges slowly, driving the LTP.
• Figure 3: Channel conductance $g/g_0$ after 40 pulses of 0.8 V with fixed duration 3 ms and varying intervals. Salt accumulation is cumulative when pulses are closely spaced, leading to frequency-dependent charging of the wall and consequent LTP (as shown in Fig. \ref{['fig:Fig2']}(c)), thereby creating FDP.
• Figure 4: (a) Channel conductance $g/g_0$ response to chemical and/or electric signals. A voltage pulse induces LTD, while a chemical signal, i.e temporarily increasing the $\text{A}^{3-}$ base reservoir concentration from $4\cdot10^{-4}$ mM to $0.01$ mM, has a comparatively insignificant effect (magenta). Combined voltage and chemical signals significantly enhance conductance, resulting in LTP (blue). (a, inset) Response in (a) corresponds to a chemical-electrical AND logic gate. (b) Conductance response to chemical and electrical 50 ms signals separated by $\Delta t$ revealing STDP Feldman2000Timing-BasedCortexAbbott2000SynapticBeastBender2006TwoCortex. (c) Schematic of the NMDA coincidence detection mechanism Luscher2012NMDALTP/LTDCotman2003ExcitatoryPlasticityRauschecker1991MechanismsBeyondXia2005ThePlasticity. NMDA receptors (gray) open if glutamate (red) released by the presynaptic neuron binds to them and $\text{Mg}^{2+}$ (green) is removed by postsynaptic depolarization, allowing a $\text{Ca}^{2+}$ (yellow) influx into the postsynaptic neuron and instigate LTP Luscher2012NMDALTP/LTD. Analogously, a simultaneous release of reactant ions at the base and electric signal at the tip increases channel conductance.