Azobenzene-based optoelectronic transistors for neurohybrid building blocks

The study addresses the challenge of creating light-responsive neuromorphic platforms that are biocompatible and mechanically compliant for direct interfacing with biological systems. Traditional inorganic and hybrid optoelectronic devices often require complex processing and may lack properties suitable for biointerfaces (e.g., ionic-electronic coupling, low stiffness, and biocompatibility). Organic photoelectrochemical transistors (OPECTs) offer a promising route to overcome these limitations by leveraging conjugated polymers with mixed ionic-electronic conduction. The authors hypothesize that integrating a photoswitchable azobenzene-functionalized PEDOT:PSS (azo-tz-PEDOT:PSS) as a fully organic, light-responsive gate in an OPECT can enable multimodal operation: (i) emulate retinal visual pathways and (ii) mimic synaptic plasticity (short- and long-term) with light-dependent conditioning and extinction. The purpose is to demonstrate an all-organic, opto- and electro-responsive neurohybrid building block capable of sensing, mimicking, and stimulating biological-like functions with tunable light intensity and electrical bias, thereby advancing bioinspired optoelectronics and neuromorphic interfaces.

The work builds on advances in optoelectronic biointerfaces and neuromorphic devices, including retina-like visual systems and synaptic platforms that use light to modulate device states. Prior OPECT implementations typically required integrating photosensitive hybrid composites at the gate to impart light responsivity, often involving complex material processing and potentially limiting biocompatibility. Photoswitchable molecules such as azobenzenes and spiropyrans have been explored in OFETs/OTFTs for light control of device properties. The authors position their contribution as a fully organic OPECT using a PEDOT:PSS gate covalently functionalized with azobenzene via click chemistry, aiming to simplify processing while enhancing biological compatibility. They note key needs for aqueous operation, electro-mechanical stability, and ionic-electronic transduction, and reference the role of conjugated polymers in enabling these properties. Prior studies on azobenzene isomerization dynamics and photo-modulation inform the expected light-driven behavior, and literature on synaptic plasticity (STP/LTP, PPF) frames the neuromorphic benchmarks used for evaluation.

Materials synthesis and functionalization: - EDOT-N3 synthesis: EDOT-Cl reacted with NaN3 in DMF (reflux 3 h), workup via EtOAc extraction and drying; yield 93%. Verified by 1H NMR and FTIR-ATR. - Azo-alkyne synthesis: 4-(Phenylazo)phenol deprotonated with K2CO3 in degassed acetone, alkylated with propargyl bromide (overnight reflux); yield 95%. Verified by 1H NMR and FTIR-ATR. - Electrodeposition: PEDOT:PSS and N3-PEDOT:PSS films electrochemically polymerized on patterned ITO (WE) using CV. Conditions: PEDOT (0–+1.0 V), N3-EDOT (0–+1.5 V) from PSSNa aqueous suspension containing monomer (0.1 M EDOT or 0.01 M N3-EDOT) at 50 mV s−1 for 10 cycles. Ag/AgCl (3 M NaCl) RE; Pt wire CE. Films rinsed and annealed at 120 °C for 1 h. - Click functionalization (azo-tz-PEDOT:PSS): eN3-PEDOT:PSS films immersed in THF:H2O (1:1) with azo-alkyne (10 mM), CuSO4·5H2O and sodium ascorbate for 24 h; extensive washing with water and THF. Device fabrication: - OPECT: On ITO glass, the gate area was electrodeposited eN3-PEDOT:PSS then click-functionalized to azo-tz-PEDOT:PSS (planar gate). PEDOT:PSS channel prepared from PH1000 with 5 vol% EG, 0.002 vol% DBSA, 1 vol% GOPS; spin-coated at 2000 rpm for 2 min; annealed 140 °C for 1 h. Patterning via O2 plasma with PDMS masks. Channel area 15×7 mm; gate area 13×10 mm; illuminated gate area 20 mm². Devices soaked in DI water overnight prior to use. - Organic resistor for retina circuit: PEDOT:PSS strip on ITO prepared similarly and patterned by O2 plasma (mask 15×7 mm). Characterization techniques: - Spectroscopy and microscopy: FTIR-ATR (4000–600 cm−1), UV-Vis (900–300 nm), XPS (Al Kα), AFM (tapping mode, 5×5 µm²), FESEM with Au sputter coating. - Electrochemistry: CV in PBS (pH 7.4) from −0.4 to +1.0 V at 50 mV s−1; EIS at ±0.01 V in 10 Hz–0.1 Hz. - Electrical testing platform: ARKEO platform (PBS electrolyte throughout). Transfer curves: Vg −0.2 to +0.8 V, Vd −0.8 to +0.1 V, 50 mV s−1. UV illumination at λ=365 nm with intensities 0.56, 1.7, 2.8 W cm−2. Operational protocols: - Light pulsing for charge calculation: Vds = −200 mV, Vgs = 0 mV; alternate 60 s dark/60 s UV at stated intensities; record gate and channel currents; integrate gate current to estimate charge. - Neuromorphic conditioning: Vds = −200 mV, Vg pulse train +300 mV with PW = 500 ms, 1 s, 5 s; light at 365 nm with intensities 0.56, 1.7, 2.8 W cm−2; assess channel conductance variation (ΔG%). PPF: apply two gate pulses (Vg = +300 mV, PW=1 s) with Δt spanning 0.1–11 s; compute PPF index. - LTP: 500 light pulses (365 nm, 2.8 W cm−2, 1 s on/1 s off), then monitor conductance recovery for >30 min in dark. - Extinction via negative bias: After 500 light pulses, apply Vgs = −300 mV for 10 min (or trains of negative pulses; PW=60 s; amplitudes −50 to −350 mV, step −50 mV). Also test consecutive negative pulses −20 to −120 mV under dark vs constant light. Test electrolyte dependence with 100 mM NaCl vs PBS. - Retina OFF-pathway emulation: Series connection of azo-OPECT and PEDOT:PSS resistor; VDD pulsed +400 mV/0 mV; gate held at 0 mV then stepped to +300 mV; test in dark and under UV (365 nm, 2.8 W cm−2). Monitor output voltage (VGC) relative to a firing threshold. - Optoelectronic memory (write/erase): (1) 500 light pulses (365 nm, 2.8 W cm−2, PW=1 s, Δt=1 s) then Vgs=−300 mV for 10 min; (2) 20 light pulses (PW=2 s, Δt=10 s) then negative pulse train (−300 mV, PW=3 s, Δt=10 s). - Atkinson–Shiffrin model: Sensory input with two +200 mV, PW=3 s, Δt=15 s pulses in dark; repeat under low UV (0.56 W cm−2) followed by 10 erasure pulses (−300 mV, PW=5 s, Δt=12 s); repeat under high UV (2.8 W cm−2).

Material synthesis and photophysics: - Successful covalent functionalization of electrodeposited N3-PEDOT:PSS with azobenzene via Cu(I)-catalyzed azide–alkyne cycloaddition confirmed by FTIR-ATR and XPS; AFM showed retained globular morphology with slight surface smoothing after click reaction. - UV-Vis: trans-azobenzene absorption at 347 nm; cis-related band at 434 nm. UV irradiation (365 nm, 5 min) decreased the 347 nm band and slightly increased 434 nm, indicating trans→cis isomerization. Reverse isomerization achieved with heat (80 °C) or blue light (445 nm, 440 mW) with full recovery but different kinetics. - Electrochemistry: CV exhibited a reduction peak at −0.12 V (attributed to N=N reduction influenced by H2PO4−/protons) and an oxidation peak at +0.57 V; UV exposure reduced hysteresis consistent with photoisomerization/functionalization effects. EIS showed reduced capacitance for azo-tz-PEDOT:PSS vs pristine PEDOT:PSS, consistent with an additional molecular spacer layer at the interface. Device photoresponse and mechanism: - In OPECTs using azo-tz-PEDOT:PSS as a planar gate, transfer/output characteristics showed that UV illumination (365 nm, 5 min) shifted Ids to lower values, unlike pristine PEDOT:PSS, indicating light-induced charge transfer/trapping between azobenzene and PEDOT:PSS. - DFT/TD-DFT on azo-tz-PEDOT model (cis conformer) showed HOMO localized on PEDOT backbone and LUMO on azobenzene moiety; UV excitation enables electron transfer from azo LUMO to PEDOT HOMO, leaving holes on azo units, polarizing the gate and decreasing channel current. Calculations confirmed the nature of UV-vis transitions for cis/trans. - Ids inversely correlated with light intensity (20% = 0.56 W cm−2; 60% = 1.7 W cm−2; 100% = 2.8 W cm−2). Under alternating 60 s dark/60 s UV, gate current displayed sustained faradaic behavior during illumination (no rapid decay), and channel current decreased rapidly with light and recovered in dark. Integrated gate charge was stable across initial pulses (e.g., 0.054 ± 0.023, 0.044 ± 0.036, 0.042 ± 0.042 µC/s; N=5). First-pulse gate charge scaled with light intensity: 0.020 ± 0.012 (20%), 0.058 ± 0.040 (60%), 0.051 ± 0.022 µC/s (100%; N=3). Retina pathway emulation: - A biohybrid retinal circuit (azo-OPECT + PEDOT:PSS resistor in series) mapped photoreceptors (gate), BCs (electrolyte–channel interface), and GCs (resistor). In dark, the GC-equivalent output (VGC) fired continuously, mimicking OFF pathway depolarization due to glutamate release. - Lateral pathway emulation via gate bias (V1 up to +300 mV) decreased VGC to ~15% (firing threshold set by a fixed level). Under light, the GC firing slowed/ceased: output dropped by ~18% (vertical OFF pathway), and with lateral input (+300 mV) an additional ~20% reduction occurred, totaling ~−43% relative decrease. Synaptic plasticity and memory: - Light-enabled conditioning: During pre-synaptic-like gate pulses (Vg=+300 mV, PW=5 s, Vds=−200 mV), UV exposure (365 nm, 6 min) decreased post-synaptic channel conductance (ΔG% = 12.5 ± 3.1; N=3). After light removal, conductance partially recovered (ΔG% = −2.7 ± 3.1; N=3), indicating reversible conditioning due to photoinduced gate polarization and cation diffusion back to electrolyte. - Conditioning increased from 20% to 60% light intensity across PWs (500 ms, 1 s, 5 s), saturating at 100% intensity; little dependence on electrical PW at fixed light intensity. - STP via PPF: Under UV, PPF index reached ≤99% when the second pulse followed within ~5 s; in dark, potentiation observed mainly for Δt < 2 s. Overall, PPF indexes under light/dark were comparable, indicating STP primarily driven by electrical stimulation. - LTP: 500 UV pulses yielded stepwise conductance decreases with slow, partial recovery persisting beyond 30 min in dark, demonstrating long-term memory retention. - Extinction/erasure via negative gate bias: After light conditioning, applying Vgs = −300 mV increased channel conductance (erasure). Under dark, significant Ids increase occurred at −120 mV; under constant light, Ids initially decreased at −20 and −40 mV (due to polarization) then increased for amplitudes ≥ −80 mV. Mechanism: electrochemical reduction of azobenzene to hydrazobenzene at modest negative bias consumes protons, attracting them from the channel through the electrolyte; PSS counterions rebind PEDOT, increasing channel conductance. Using 100 mM NaCl (lower proton availability) reduced modulation vs PBS, corroborating proton involvement. - Optoelectronic memory: Demonstrated write (light pulse trains) and erase (negative voltage pulse trains) operations in the same device. Atkinson–Shiffrin model emulated: sensory memory (sub-threshold); working memory (low-intensity light caused threshold crossing reversible by erasure pulses); long-term memory (high-intensity light produced persistent state surpassing threshold despite erasure pulses).

The findings validate that covalently integrating azobenzene into PEDOT:PSS yields a fully organic, light-responsive gate material that enables OPECTs to operate as multimodal neurohybrid devices. Spectroscopy and electrochemical data, supported by DFT/TD-DFT, reveal a mechanistic basis in which UV light drives charge transfer from azobenzene to the PEDOT backbone, polarizing the gate and modulating the channel via ionic interactions at the electrolyte interface. This light-driven modulation is intensity-dependent and produces sustained faradaic gate currents, which directly translate into tunable channel conductance in aqueous media. By embedding the azo-OPECT in a simple series circuit, the device mimics retinal OFF pathways with both vertical (light-driven) and lateral (gate-bias-controlled) features, reproducing the decrease in ganglion-cell-like output under illumination and the adjustable gain via lateral inputs. Neuromorphically, the device expresses core synaptic functions: short-term facilitation (PPF) largely governed by electrical pulse timing, and long-term plasticity induced by repeated light exposure, with selective erasure achievable via negative gate bias that modulates proton migration. The dependence on light intensity and electrolyte proton concentration provides additional control levers. Together, these results demonstrate that all-organic OPECTs with photoswitchable gates can bridge sensing, processing, and memory operations, addressing the initial goal of creating bio-compliant, multimodal building blocks for interfacing and emulating neural functions.

This work introduces a fully organic azobenzene-functionalized PEDOT:PSS integrated as a planar gate in an OPECT, delivering robust light-responsiveness in aqueous environments. The device exhibits: (i) intensity-dependent optical gating via photoinduced charge transfer and gate polarization; (ii) emulation of retinal OFF pathways, including lateral gain control; (iii) neuromorphic behavior with both STP (PPF) and LTP; and (iv) optoelectronic memory with light-enabled conditioning (write) and negative-bias-driven extinction (erase). The approach simplifies material integration compared to hybrid photoactive gates and leverages the biophysical advantages of conjugated polymers. Potential future directions include: optimizing device geometry to tune light-mediated gating efficiency and operation speed; extending spectral responsivity to visible/NIR wavelengths for enhanced biocompatibility; integrating with living neuronal tissues for closed-loop interfacing; engineering electrolyte composition for selective ionic control; and scaling arrays for neuromorphic vision and computing applications.

- Illumination used UV (365 nm), which may limit direct in vivo or cell-interfacing applications due to potential phototoxicity; spectral shifting to visible/NIR would improve bio-compatibility. - The device response entails partial recovery after extensive light stimulation (LTP), which, while desirable for memory, may limit rapid full reversibility in certain sensing modes. - Performance depends on electrolyte proton availability (PBS vs 100 mM NaCl), indicating environmental sensitivity that could affect reproducibility across biological media. - Operation speed and gating efficiency are geometry-dependent and may require optimization for faster or lower-energy operation. - Quantitative mapping of stability and cycling endurance over very long time scales was not detailed and remains to be assessed.