Mimicking efferent nerves using a graphdiyne-based artificial synapse with multiple ion diffusion dynamics

Complex human nervous systems are compact, parallel, and reliable, inspiring artificial systems for neuromorphic computing, bioinspired sensorimotor systems, brain–machine interfaces, and prosthetics. Somatosensory nerves transfer signals through synapses to achieve perception, memory, and motion, motivating synapse imitation as building blocks for neural processing. Various device structures (MIM switches, electrolyte/semiconductor heterojunctions, multiterminal transistors) have been used to emulate synaptic transmission, with ion-migration configurations particularly promising for bioinspired ionotronic and biohybrid systems. However, new materials and device responsivity, including thermal and environmental stability, require further exploration to achieve biomimetic functionality. Graphdiyne (GDY), a π-conjugated carbon allotrope with triangular pores and sp-hybridized carbon atoms, offers storage sites and rapid diffusion channels for alkali ions with low diffusion barriers, suggesting potential for GDY-based artificial synapses (GASs) that mimic synaptic cleft transmission. Inspired by biological motor neurons, the authors propose a junction-type GAS by coupling a GDY film with solid-state electrolytes to emulate short-term plasticity (postsynaptic current, PPF, dynamic filtering) with ultralow pulse responsiveness and femtowatt energy consumption, and to demonstrate real-time information integration, parallel processing, and signal transduction to actuators.

The study situates itself within neuromorphic and biohybrid research where artificial synapses emulate biological plasticity. Prior approaches include metal/insulator/metal resistive switches, electrolyte/semiconductor heterojunctions, and multiterminal transistors. Ion-migration-based devices are highlighted for their suitability in ionotronic sensorimotor and biohybrid systems. Graphdiyne has been advanced for batteries, catalysis, solar cells, nonlinear optics, electronics, and biomedicine due to its optoelectronic properties and biocompatibility. Its triangular pore network and sp-carbon afford adsorption sites and low-barrier diffusion paths for Li+ and Na+ (and perchlorate) enabling surface adsorption and interlayer insertion, indicating advantages for synaptic emulation. Gaps remain in understanding new materials’ working principles and in ensuring device responsivity and stability under heat and humidity.

GDY synthesis and characterization: Two-dimensional graphdiyne (GDY) was synthesized from hexakis[(trimethylsilyl)ethynyl]benzene (HEB-TMS) via liquid/liquid interfacial coupling. HEB-TMS (8 mg) was dissolved in degassed dichloromethane (120 mL); tetrabutylammonium fluoride (100 µL, 1 M in THF) was added under Ar and stirred (15 min). For interfacial growth, HEB in dichloromethane (10 mL, 0.1 mM) was added to a glass cylinder; deionized water (12 mL) was overlaid; a mixture (8 mL) of Cu(OAc)2 (0.01 M) and pyridine (0.25 M) was dropped into the aqueous phase. After >24 h, a brown film formed at the interface, collected, filtered (nylon membrane, 100 nm), washed with HCl (1 M, 10 mL) and water (10 mL). GDY was dispersed in DMF (5 mg/10 mL), ultrasonicated (10 min), spin-coated (80 µL, 800 rpm) onto cleaned Si substrates, and annealed (80 °C, 20 min) to form ~400 nm films. Characterization included AFM, SEM, TEM, HRTEM, SAED, Raman (532 nm), and XPS to confirm morphology, crystallinity, and sp/sp2 carbon composition.

Device fabrication (GAS): Solid polymer electrolytes (Li-SPE and Na-SPE) were prepared by dissolving PEO (0.8 g) with LiClO4 or NaClO4 (0.1 g) in acetonitrile (10 mL). Electrolytes were spin-coated on GDY films and annealed at 90 °C (20 min) in a N2 glove box. Au-dot top electrodes were deposited to complete Li-GAS and Na-GAS devices. Electrical measurements were performed with a Keithley 4200A in N2 (H2O and O2 <0.1 ppm). Thermal stability was tested by heating from 30 to 80 °C (10 °C steps, 10 min each), and humidity stability by conditioning at 15–65% RH (1 h each), followed by glove box testing. I–V sweeps and pulse protocols (positive/negative pulses with varied amplitudes, durations, frequencies) were applied to probe short-term plasticity.

Actuator fabrication: An electrolyte film was formed by dissolving PVDF-HFP and EMIBF4 in DMF (2 mL, 60 °C, 1 day), solution-cast in N2 at room temperature (1 day), peeled, and sandwiched between CNT electrodes pressed at 70 °C (2 min). Actuators were aged under reduced pressure (1 day) and cut to 20 × 2 mm2 strips.

Circuit integration (artificial efferent nerve): The GAS bottom electrode was connected to an operational amplifier circuit to convert device current to actuator-driving voltage. The op-amp output was used to operate the polymer/CNT actuator. Single- and dual-synapse inputs were realized by applying independent presynaptic pulse trains to two GAS devices whose outputs were summed via the circuit to drive a single actuator.

Computational methods (DFT): First-principles calculations were performed with VASP at the PBE-GGA level. Optimized GDY lattice parameters: a = b = 9.46 Å. Band structures and PDOS were computed for pristine GDY and Li/Na-adsorbed GDY. Adsorption sites evaluated included benzene center (A) and triangular pore center (B); total energies favored B (Li: −155.22 eV vs −154.69 eV; Na: −155.13 eV vs −154.11 eV). Charge density differences indicated electron transfer from Li/Na to GDY. Diffusion barriers for a single Li or Na ion moving between adjacent stable adsorption positions on GDY were calculated (Li: 0.54 eV; Na: 0.72 eV).

• GAS architecture: Junction-type electrolyte/GDY synapse (Li-GAS and Na-GAS) emulates synaptic cleft signaling via Li+/Na+ migration, accumulation, and intercalation.
• Electrical characteristics: Repeated I–V sweeps showed negative differential resistance (NDR): pronounced in Li-GAS, and two NDR regions in Na-GAS (low potential <2 V more pronounced), attributed to interfacial ion fields and electrochemical doping/dedoping; narrowing the scan window suppressed NDR (pseudocapacitance-dominated response). Li-GAS exhibited larger current response, consistent with higher Li storage.
• Short-term plasticity: Spike-voltage-dependent plasticity from 0.5–4 V pulses produced peak postsynaptic currents from sub-nA to tens of nA. Single positive/negative spikes yielded transient currents decaying to ±0.2 nA in <6 s (bidirectional volatility). Paired-pulse facilitation (PPF) decreased with inter-spike interval; Na-GAS showed higher facilitation indices than Li-GAS. Spike-rate-dependent plasticity (SRDP): higher-rate negative pulse trains increased gain (A10/A1×100%). Li-GAS showed stronger dynamic filtering. Rapid discharge attributed to ion migration/interface depolarization enabled real-time imaging in a 9×9 array (letters G, D, Y). Nonidentical pulse sequences produced repeatable, amplitude-correlated responses; 0.22 s sufficed to encode amplitude-dependent patterns.
• Ultralow-voltage operation and energy: Plasticity persisted down to 80–20 mV in Li-GAS; Na-GAS operated at ≤20 mV with clear responses. Average energy per synaptic event corresponded to 16.7 fW power, orders of magnitude below biological synapses and competitive among two-terminal devices.
• Computational insights: Pristine GDY bandgap ~0.50 eV; Li/Na adsorption shifted the Fermi level upward, enhancing conductivity; charge transfer from Li/Na to GDY confirmed. Calculated diffusion barriers: Li 0.54 eV, Na 0.72 eV on GDY surface, explaining easier Li+ migration and larger synaptic weights experimentally.
• Thermal and environmental stability: After heating up to 353 K (80 °C), I–V characteristics remained stable with a more pronounced bulge near ~2.05 V; current windows were largely unchanged. With increasing RH, current windows changed gradually; curves remained stable up to ~35% RH. Li-GAS was more sensitive to temperature/humidity. Electrolyte films showed no severe damage after exposure to ≤353 K and ≤65% RH.
• Dendritic integration and parallel processing: Two-synapse spatiotemporal pairing produced nonlinear output enhancement; when one synapse preceded the other by <4 s (ΔT<0), outputs increased to ~340% (Li-GAS) and ~250% (Na-GAS) of single-synapse outputs. Gains from dual-synapse integration exceeded those from repeated pulses or increased duration on a single synapse, enabling logic operations based on duration-dependent thresholds. Shunting inhibition was demonstrated by concurrent excitatory/inhibitory inputs, progressively suppressing output with increasing inhibitory amplitude.
• Frequency identification and actuation: GAS identified presynaptic frequency content up to ~5.6 Hz via peak-to-valley spacings and increased postsynaptic current with frequency. Artificial efferent nerves integrating one vs two synaptic inputs drove polymer/CNT actuators to bend with different amplitudes (e.g., ~10° single input vs ~20° dual input at 0.8 Hz), demonstrating real-time parallel integration and motor actuation.

The work addresses how to emulate biological synaptic transmission and integration with a material system that supports fast ion migration, low-voltage operation, and environmental robustness. By leveraging GDY’s porous, sp-carbon network for Li+/Na+ adsorption and diffusion, the GAS achieves essential short-term plasticity (PPF, SRDP, dynamic filtering) with ultralow millivolt operation and femtowatt-level energy, surpassing biological synapses in energy efficiency. The devices maintain functionality under elevated temperature and moderate humidity, supporting practical deployment. Parallel processing across multiple synapses yields nonlinear integration gains and shunting inhibition, enabling dynamic logic and spatiotemporal learning akin to dendritic processing. The GAS can also infer input frequency content up to ~5.6 Hz, and, when interfaced with an amplifier and ionic polymer actuator, performs efferent motor functions with graded bending proportional to integrated synaptic output. These findings demonstrate a viable path toward bioinspired peripheral nervous systems and neuromorphic sensorimotor platforms, with GDY’s biocompatibility suggesting promise for biohybrid interfacing.

This study demonstrates the first graphdiyne-based artificial synapse (GAS) that emulates short-term synaptic plasticity with ultralow-voltage, femtowatt-level operation, robust thermal/environmental stability, and real-time parallel processing and integration. DFT analyses corroborate facile Li+/Na+ ion diffusion and charge transfer on GDY, explaining experimental performance. The GAS integrates multiple presynaptic inputs to realize dynamic logic, shunting inhibition, frequency identification (≤5.6 Hz), and drives artificial muscles as an artificial efferent nerve with graded actuation. The approach offers a compact, low-power processing element for soft electronics, neurorobotics, smart prostheses, and brain–computer interface biohybrids. Future work should address scalable synthesis of single-/multilayer GDY, broaden environmental operating windows, extend frequency bandwidth, and explore direct coupling with biological neurons and all-GDY efferent nerve systems combining GDY actuators.

• Materials scalability: Large-scale preparation of single- and multilayer GDY remains challenging, potentially limiting device uniformity and integration density.
• Environmental tolerance: While devices remained functional after exposure to ≤353 K and ≤65% RH, stable I–V behavior was emphasized up to ~35% RH, indicating performance sensitivity at higher humidity; Li-GAS was more sensitive to temperature/humidity changes than Na-GAS.
• Bandwidth: Frequency identification and reliable SRDP demonstrations were shown up to ~5.6 Hz, suggesting limited operational bandwidth for higher-frequency neuromorphic signaling without further optimization.
• Variability and conditioning: Initial I–V sweeps showed evolving behavior before stabilization, and Na-GAS required a few cycles for repeatability at low voltages, indicating potential device-to-device or cycle-to-cycle variability.