Fig 1 - available via license: Creative Commons Attribution-NonCommercial 4.0 International
Content may be subject to copyright.
![Organic ECRAMs using ion gels enable submicrosecond switching in vacuum. (A) ECRAM device schematic. (B) Chemical structures of the channel/gate (left) and electrolyte (right) materials. The blue circle on 1-ethylimidazolium bis(trifluoromethylsulfonyl)imide (EIM:TFSI) highlights the hydrogen that renders EIM:TFSI protic. (C) Resistive switching characteristics of ECRAM with PEDOT:PSS as the channel/gate material and Aquivion as the electrolyte rapidly deteriorate when going from 20% relative humidity (RH) in N 2 atmosphere (black) to 2 × 10 −4 mbar vacuum (gray). (D) Cycling of ECRAM with PEDOT:PSS as the channel/gate material and EIM:TFSI poly(vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP) as the electrolyte operating in vacuum. (E) Cycling of ECRAM with poly(2-(3,3-bis(2-(2-(2-methoxyethoxy)ethoxy)ethoxy)-[2,2-bithiophen]-5-yl)thieno[3,2-b]thiophene) [p(g2T-TT)] as the channel/gate material and 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide (EMIM:TFSI) PVDF-HFP as the electrolyte operating in vacuum. Inset shows normalized channel conductance G SD /G min for PEDOT:PSS-based (blue) and p(g2T-TT)-based (red) ECRAMs.](publication/342688278/figure/fig1/AS:909571946184704@1593870259268/Organic-ECRAMs-using-ion-gels-enable-submicrosecond-switching-in-vacuum-A-ECRAM-device.png)
Organic ECRAMs using ion gels enable submicrosecond switching in vacuum. (A) ECRAM device schematic. (B) Chemical structures of the channel/gate (left) and electrolyte (right) materials. The blue circle on 1-ethylimidazolium bis(trifluoromethylsulfonyl)imide (EIM:TFSI) highlights the hydrogen that renders EIM:TFSI protic. (C) Resistive switching characteristics of ECRAM with PEDOT:PSS as the channel/gate material and Aquivion as the electrolyte rapidly deteriorate when going from 20% relative humidity (RH) in N 2 atmosphere (black) to 2 × 10 −4 mbar vacuum (gray). (D) Cycling of ECRAM with PEDOT:PSS as the channel/gate material and EIM:TFSI poly(vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP) as the electrolyte operating in vacuum. (E) Cycling of ECRAM with poly(2-(3,3-bis(2-(2-(2-methoxyethoxy)ethoxy)ethoxy)-[2,2-bithiophen]-5-yl)thieno[3,2-b]thiophene) [p(g2T-TT)] as the channel/gate material and 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide (EMIM:TFSI) PVDF-HFP as the electrolyte operating in vacuum. Inset shows normalized channel conductance G SD /G min for PEDOT:PSS-based (blue) and p(g2T-TT)-based (red) ECRAMs.
Source publication
Devices with tunable resistance are highly sought after for neuromorphic computing. Conventional resistive memories, however, suffer from nonlinear and asymmetric resistance tuning and excessive write noise, degrading artificial neural network (ANN) accelerator performance. Emerging electrochemical random-access memories (ECRAMs) display write line...
Contexts in source publication
Context 1
... exchange membranes require extensive hydration to conduct protons (13,14). Organic ECRAMs (Fig. 1A) made using poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) (Fig. 1B) as the channel/gate material and a commercially available perfluorosulfonic acid ionomer membrane (Aquivion; see Materials and Methods) as electrolyte do not exhibit any conductance modulation when operated in moderate (2 × 10 −4 mbar) vacuum (Fig. ...
Context 2
... exchange membranes require extensive hydration to conduct protons (13,14). Organic ECRAMs (Fig. 1A) made using poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) (Fig. 1B) as the channel/gate material and a commercially available perfluorosulfonic acid ionomer membrane (Aquivion; see Materials and Methods) as electrolyte do not exhibit any conductance modulation when operated in moderate (2 × 10 −4 mbar) vacuum (Fig. 1C). This limitation renders the use of conventional proton exchange membranes ...
Context 3
... ECRAMs (Fig. 1A) made using poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) (Fig. 1B) as the channel/gate material and a commercially available perfluorosulfonic acid ionomer membrane (Aquivion; see Materials and Methods) as electrolyte do not exhibit any conductance modulation when operated in moderate (2 × 10 −4 mbar) vacuum (Fig. 1C). This limitation renders the use of conventional proton exchange membranes incompatible with dry environments and thus integration into ANN accelerators, which is the first technological challenge. The second technological challenge stems from water evaporation from the proton exchange membrane at the temperatures generated in ...
Context 4
... we demonstrate that solid electrolytes made by infiltrating an electrically insulating polymer with common ionic liquids (Fig. 1B) (15) enable organic ECRAMs that are programmable in vacuum at a low voltage of ±1 V using submicrosecond pulses, while still displaying linear resistive switching spanning 100× distinct states (Fig. 1, D and E). Size scaling enables faster switching down to 20 ns (our instrument limit). In this work, we obtain accurate readout of the ...
Context 5
... we demonstrate that solid electrolytes made by infiltrating an electrically insulating polymer with common ionic liquids (Fig. 1B) (15) enable organic ECRAMs that are programmable in vacuum at a low voltage of ±1 V using submicrosecond pulses, while still displaying linear resistive switching spanning 100× distinct states (Fig. 1, D and E). Size scaling enables faster switching down to 20 ns (our instrument limit). In this work, we obtain accurate readout of the ECRAM channel conductance G SD as well as the amount of injected charge Q per write pulse by operating the devices using two access transistors (write and read select) (see fig. S1 for the measurement ...
Context 6
... spanning 100× distinct states (Fig. 1, D and E). Size scaling enables faster switching down to 20 ns (our instrument limit). In this work, we obtain accurate readout of the ECRAM channel conductance G SD as well as the amount of injected charge Q per write pulse by operating the devices using two access transistors (write and read select) (see fig. S1 for the measurement schematic and a more detailed explanation). We use ion gel electrolytes similar to those reported previously (15) by mixing the polymeric insulator poly(vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP) with common ionic liquids, such as 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide (EMIM: TFSI) ...
Context 7
... We use ion gel electrolytes similar to those reported previously (15) by mixing the polymeric insulator poly(vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP) with common ionic liquids, such as 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide (EMIM: TFSI) or 1-ethylimidazolium bis(trifluoromethylsulfonyl)imide (EIM:TFSI) (Fig. ...
Context 8
... ion gel-based ECRAMs using PEDOT:PSS as the channel/gate material perform as well as previously reported Nafion-based PEDOT:PSS devices had (6, 9) in all respects, they still suffer from a limited dynamic range (<2×) between the highest/lowest conductance states (Fig. 1D). We improve the device dynamic range by replacing PEDOT:PSS with the intrinsic semiconducting polymer poly(2-(3,3-bis(2-(2-(2- (Fig. 1B), which was recently developed to operate in enhancement-mode organic electrochemical transistors (OECTs) (16). In contrast to PEDOT:PSS, p(g2T-TT) is highly resistive before electrochemical gating and ...
Context 9
... material perform as well as previously reported Nafion-based PEDOT:PSS devices had (6, 9) in all respects, they still suffer from a limited dynamic range (<2×) between the highest/lowest conductance states (Fig. 1D). We improve the device dynamic range by replacing PEDOT:PSS with the intrinsic semiconducting polymer poly(2-(3,3-bis(2-(2-(2- (Fig. 1B), which was recently developed to operate in enhancement-mode organic electrochemical transistors (OECTs) (16). In contrast to PEDOT:PSS, p(g2T-TT) is highly resistive before electrochemical gating and enables a higher dynamic range (~4×) even under considerably shorter ±1-V 300-ns write pulses than needed for PEDOT:PSS (Fig. 1, D and ...
Context 10
... (Fig. 1B), which was recently developed to operate in enhancement-mode organic electrochemical transistors (OECTs) (16). In contrast to PEDOT:PSS, p(g2T-TT) is highly resistive before electrochemical gating and enables a higher dynamic range (~4×) even under considerably shorter ±1-V 300-ns write pulses than needed for PEDOT:PSS (Fig. 1, D and E). The large improvement in ECRAM dynamic range is evident when comparing the ECRAM channel conductance G SD normalized by its minimum value G SD /G min (Fig. 1E, inset). The p(g2T-TT) device linearity and signal-to-noise ratio (G 2 / 2 > 100, where G is the conductance update and is its standard deviation) fulfill the ...
Context 11
... is highly resistive before electrochemical gating and enables a higher dynamic range (~4×) even under considerably shorter ±1-V 300-ns write pulses than needed for PEDOT:PSS (Fig. 1, D and E). The large improvement in ECRAM dynamic range is evident when comparing the ECRAM channel conductance G SD normalized by its minimum value G SD /G min (Fig. 1E, inset). The p(g2T-TT) device linearity and signal-to-noise ratio (G 2 / 2 > 100, where G is the conductance update and is its standard deviation) fulfill the requirements for high ANN accuracy (4). Hence, p(g2T-TT) outperforms PEDOT:PSS due to its larger dynamic range, lower write energy, and faster ...
Context 12
... the cycling characteristics of all devices in the array shift with temperature by the same amount, ANN computation will remain unaffected. Figure 2C shows that p(g2T-TT) requires considerably less charge than PEDOT:PSS to attain a similar sized update G SD /G o , enabling the use of considerably shorter write pulses, as corroborated earlier in Fig. 1 (D and E). p(g2T-TT) therefore outperforms PEDOT:PSS not only in terms of larger dynamic range (Fig. 1, D and E) and temperature resilience (Fig. 2, A and B) but also speed (Fig. ...
Context 13
... computation will remain unaffected. Figure 2C shows that p(g2T-TT) requires considerably less charge than PEDOT:PSS to attain a similar sized update G SD /G o , enabling the use of considerably shorter write pulses, as corroborated earlier in Fig. 1 (D and E). p(g2T-TT) therefore outperforms PEDOT:PSS not only in terms of larger dynamic range (Fig. 1, D and E) and temperature resilience (Fig. 2, A and B) but also speed (Fig. ...
Context 14
... find that PEDOT:PSS devices using the aprotic ionic liquid EMIM: TFSI are barely operational ( fig. S6), while those made with the protic (26,27) EIM:TFSI exhibit excellent performance (Figs. 1 to 3). Similar to aqueous electrolyte-gated organic neuromorphic devices (6), we thus hypothesize that protons diffusing in the confined environment of the ion gel and the conjugated polymer microstructure play an important role in the switching of our ion gel-based ECRAMs. ...
Context 15
... are known to diffuse fast and with low activation barriers, especially in hydrogen-bonded networks, where the Grotthuss mechanism is active (28). p(g2T-TT) devices, on the other hand, show excellent performance with both EMIM:TFSI-based (Figs. 1 to 3) and EIM:TFSI-based ion gels ( fig. S6), suggesting the need for a more detailed understanding of the switching mechanisms and their relationship to proton transport that are beyond the scope of this work. ...
Context 16
... C layer was then peeled off in ambient to confine the organic semiconductor only in the photolithographically defined ECRAM channel and gate regions. The wafer dies were gently rinsed in deionized water to eliminate residual soap and were subsequently dried using a N 2 gun. The ion gel solution was then drop-cast on top, as schematically shown in Fig. ...
Citations
... holes and ions. This polymer serves as a versatile platform for various functionalities and is suitable for both short-and long-term devices depending on the probing conditions 34,42 . Hence, the organic neuromorphic circuit allows for monolithic integration of both volatile and non-volatile functionalities with the same polymer as the channel material of the transistors. ...
... Hence, the organic neuromorphic circuit allows for monolithic integration of both volatile and non-volatile functionalities with the same polymer as the channel material of the transistors. It exhibits a wide range of well-defined conductance states (with a > 100 on/off ratio), high linearity, sensitivity to gate pulses (ranging from μS to mS), and stability (>10^9 write-read operations) 42,43 . The low-voltage operation (≤ ± 1 V) and compatibility with solution-based processing methods contribute to high energy efficiency and cost-effectiveness. ...
... Reactive ion etching with O2 plasma is used to carve out the channel and corresponding gates. The semiconducting polymer p(g2T-TT) is synthesized according to (41) and prepared and applied following the procedure in 42,43 . p(g2T-TT) is solved in chloroform (3 mg/ml) inside an N2-filled glove box and spin-cast inside the N2filled glove box at 1000 rpm for 1 min. ...
Biological systems interact directly with the environment and learn by receiving multimodal feedback via sensory stimuli that shape the formation of internal neuronal representations. Drawing inspiration from biological concepts such as exploration and sensory processing that eventually lead to behavioral conditioning, we present a robotic system handling objects through multimodal learning. A small-scale organic neuromorphic circuit locally integrates and adaptively processes multimodal sensory stimuli, enabling the robot to interact intelligently with its surroundings. The real-time handling of sensory stimuli via low-voltage organic neuromorphic devices with synaptic functionality forms multimodal associative connections that lead to behavioral conditioning, and thus the robot learns to avoid potentially dangerous objects. This work demonstrates that adaptive neuro-inspired circuitry with multifunctional organic materials, can accommodate locally efficient bio-inspired learning for advancing intelligent robotics.
... To address the challenge of the absence of a CMOS-compatible allsolid-state electrolyte, a nanoporous phosphosilicate glass (PSG) solid electrolyte was proposed by Alamo's group [70]. Solid electrolytes based on organic materials are susceptible to chemical and thermal instability during CMOS fabrication procedures [71]. In Fig. 3e, a three-terminal EIS with a WO 3 channel, a PSG electrolyte layer, and a Pd gate reservoir demonstrates a fully CMOS-compatible process flow. ...
... Polymer-based intercalation devices have demonstrated submicrosecond switching speeds but were limited by their incompatibility with the semiconductor industry. For suitable on-chip fabrication in Si electronics, similar to 2D materials, polymer-based intercalation materials need to provide integration stability [40,71]. Therefore, 2D titanium carbide (Ti 3 C 2 T x ) MXene was employed by Melians et al. for high-performance intercalation devices compatible with Si fabrication processes [75]. ...
... When most OMIECs come into contact with solvent or liquid electrolytes, they swell, and the solvated ion transport occurs at an accelerated rate 2 . In some OMIEC systems, proton conduction can occur even more rapidly through the Grotthuss mechanism 104 . This mechanism involves the hopping of protons between hydronium and water molecules within a hydrogen-bonded network. ...
... To achieve rapid switching, a successful strategy involves utilizing internally ion-gated OECTs, embedded with small mobile ions within the film, which forgoes the need for an external electrolyte and reduces the distance and time required for ion transit 8 . Ionic liquids that permeate the semiconductor before biasing were shown to enhance the device bandwidth 104 . ...
... The ability to finely tune the OMIEC conductance with external bias and the rapid transition of the film between distinct doping states in response to sub-microsecond voltage pulses have enabled their use in electrochemical random-access memory (ECRAM) devices. However, for integrating these devices into hardware-based artificial neural networks, it is essential for the film to remain in one doping state without the bias for an extended period to enhance long-term memory 56,104 . This can be achieved through synthetic design 70,115 and microstructure optimization 30,116 . ...
The organic electrochemical transistor (OECT) has emerged as the core component of specialized bioelectronic technologies, such as neural interfaces and sensors of disease biomarkers. At the heart of the OECT is its channel, made of an organic mixed ionic–electronic conductor (OMIEC). The chemical structure of the OMIEC governs the electronic, optical and mechanical traits of OECT, with even subtle structural tweaks leading to sizeable functional disparities. In this Review, we summarize the recent progress in OECT device development while underscoring the critical role of OMIEC selection in steering diverse applications. Our narrative charts the milestones in materials exploration, tracing the evolution of the field in parallel with polymer chemistry breakthroughs. We emphasize how materials design has enabled new device operation mechanisms by adding features such as biocompatibility, stretchability, stimuli response and memory retention to OMIECs. We also highlight the obstacles that must be surmounted to translate OECT-based devices from laboratory instruments into tangible real-world technologies.
... Reactive ion etching with O2 plasma is used to carve out the channel and corresponding gates. The semiconducting polymer p(g2T-TT) is synthesized according to (41) and prepared and applied following the procedure in 41 42 . p(g2T-TT) is solved in chloroform (3 mg/ml) inside an N2-filled glove box and spin-cast inside the N2-filled glove box at 1000 rpm for 1 min. ...
Biological systems interact directly with the environment and learn by receiving multimodal feedback via sensory stimuli that shape the formation of internal neuronal representations. Drawing inspiration from biological concepts such as exploration and sensory processing that eventually lead to behavioral conditioning, we present a robotic system handling objects through multimodal learning. A small-scale organic neuromorphic circuit locally integrates and adaptively processes multimodal sensory stimuli, enabling the robot to interact intelligently with its surroundings. The real-time handling of sensory stimuli via low-voltage organic neuromorphic devices with synaptic functionality forms multimodal associative connections that lead to behavioral conditioning, and thus the robot learns to avoid potentially dangerous objects.
This work demonstrates that adaptive neuro-inspired circuitry with multifunctional organic materials, can accommodate locally efficient bio-inspired learning for advancing intelligent robotics.
... 22,23 Furthermore, 2D van der Waals heterostructures have complex fabrication requirements, 24,25 and organic materials suffer from low speed and are incompatible with standard CMOS technology. 26,27 Among the numerous materials available for incorporation in neuromorphic networks, transparent conductive oxides (TCOs) have recently emerged as the most suitable for certain tasks due to their ability to deliver nonlinearity and bistability under both electrical signaling and optical power coupled to the waveguide. 28,29 Consequently, TCOs offer a dual-mode operation, making operation conditions flexible. ...
Fully CMOS-compatible photonic memory holding devices hold a potential in the development of ultrafast artificial neural networks.
Leveraging the benefits of photonics such as high-bandwidth, low latencies, low-energy interconnect, and high speed, they can overcome the
existing limits of electronic processing. To satisfy all these requirements, a photonic platform is proposed that combines low-loss nitride-rich
silicon as a guide and low-loss transparent conductive oxides as an active material that can provide high nonlinearity and bistability under
both electrical and optical signals.
... 24 Moreover, it was shown that when IL ions prediffused inside a glycolated p-type film, they improved the speed, independent of their proton-transporting nature. 19,25 This strategy was first shown by Spyropoulos et al., who encapsulated mobile ions inside PEDOT:PSS to improve the device speed. 26 While these findings indicate that IL gating can enhance the speed of OECTs for certain electrolyte-channel combinations, the underlying physical phenomena behind these changes are not fully understood. ...
The organic electrochemical transistor (OECT) is a biosignal transducer known for its high amplification but relatively slow operation. Here, we demonstrate that the use of an ionic liquid as the dielectric medium significantly improves the switching speed of a p-type enhancement-mode OECT, regardless of the gate electrode used. The OECT response time with the ionic liquid improves up to ca. 41-fold and 46-fold for the silver/silver chloride (Ag/AgCl) and gold (Au) gates, respectively, compared with devices gated with the phosphate buffered saline (PBS) solution. Notably, the transistor gain remains uncompromised, and its maximum is reached at lower voltages compared to those of PBS-gated devices with Ag/AgCl as the gate electrode. Through ultraviolet–visible spectroscopy and etching X-ray photoelectron spectroscopy characterizations, we reveal that the enhanced bandwidth is associated with the prediffused ionic liquid inside the polymer, leading to a higher doping level compared to PBS. Using the ionic liquid-gated OECTs, we successfully detect electrocardiography (ECG) signals, which exhibit a complete waveform with well-distinguished features and a stable signal baseline. By integrating nonaqueous electrolytes that enhance the device bandwidth, we unlock the potential of enhancement-mode OECTs for physiological signal acquisition and other real-time biosignal monitoring applications.
... 29,30 Depending on the specific application, synaptic functions with different decay constants are required for synaptic devices because synaptic functions with different decay constants can affect the temporal dynamics of the network and its ability to perform certain computations, such as short-term memory or pattern recognition. 31 Synaptic decay refers to the gradual decrease in the strength of a synaptic connection over time. For example, different decay constants can help the network distinguish between phonemes that have similar temporal patterns but different durations. ...
A liquid Ga-based synaptic device with two-terminal electrodes is demonstrated in NaOH solutions at 50 °C. The proposed electrochemical redox device using the liquid Ga electrode in the NaOH solution can emulate various biological synapses that require different decay constants. The device exhibits a wide range of current decay times from 60 to 320 ms at different NaOH mole concentrations from 0.2 to 1.6 M. This research marks a step forward in the development of flexible and biocompatible neuromorphic devices that can be utilized for a range of applications where different synaptic strengths are required lasting from a few milliseconds to seconds.
... Up to now, most of the efforts on material platform for a neuromorphic computing focused on magnetic alloys, metal oxides, phase change materials (PCM), ferroelectrics, 2D van der Waals materials or organic materials [4]. However, they suffer either from a slow phase transition and thus slow switching time (PCM) [15][16][17], very high operation voltage related to the high coercive field (ferroelectrics) [18,19], high power operation (metal oxides for valence change memory) [20,21], very poor device-to-device stability resulting in deviation of the switching voltage and currents that finally provides to failure (electrochemical metallization cells) [22,23], complex fabrication (2D van der Waals heterostructures) [24,25] or low speed and incompatibility with standard CMOS technology (organic materials) [26,27]. Transparent conductive oxides that belong to the epsilon-near-zero (ENZ) materials proved to be an excellent material for electro-optic modulation and other optoelectronic applications due to the large permittivity tunability under an applied voltage or a light illumination [28][29][30][31][32][33][34][35]. ...
Photonics integrated circuits have a huge potential to serve as a framework for a new class of information processing machines and can enable ultrafast artificial neural networks. They can overcome the existing speed and power limits of the electronic processing elements and provide additional benefits of photonics such as high-bandwidth, sub-nanosecond latencies and low-energy interconnect credentials leading to a new paradigm called neuromorphic photonics. The main obstacle to realize such a task is a lack of proper material platform that imposes serious requirements on the architecture of the network. Here we suggest and justify that transparent conductive oxides can be an excellent candidate for such a task as they provide a nonlinearity and bistability under both optical and electrical inputs.
... Conventional computing systems based on von Neuman architecture are impractical for processing large amounts of unstructured data as the processor needs to communicate with the memory to perform each operation, leading to wasted energy and time resources (21)(22)(23). On the other hand, emerging neuromorphic computing systems that can perform parallel operations with merged memory and processing units, such as the brain, serve as an attractive solution (24)(25)(26)(27)(28). ...
Extracting valuable information from the overflowing data is a critical yet challenging task. Dealing with high volumes of biometric data, which are often unstructured, nonstatic, and ambiguous, requires extensive computer resources and data specialists. Emerging neuromorphic computing technologies that mimic the data processing properties of biological neural networks offer a promising solution for handling overflowing data. Here, the development of an electrolyte-gated organic transistor featuring a selective transition from short-term to long-term plasticity of the biological synapse is presented. The memory behaviors of the synaptic device were precisely modulated by restricting ion penetration through an organic channel via photochemical reactions of the cross-linking molecules. Furthermore, the applicability of the memory-controlled synaptic device was verified by constructing a reconfigurable synaptic logic gate for implementing a medical algorithm without further weight-update process. Last, the presented neuromorphic device demonstrated feasibility to handle biometric information with various update periods and perform health care tasks.
... Among organic mixed ionic electronic conductors (OMIECs), poly(3,4ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) is the most widely used conductive polymer [4] that has been used to fabricate organic neuromorphic devices [5]. PEDOT:PSS-based artificial synapses were adapted to be screen printable [6] and temperature resilient [7]. Devices with nanofiber channels show ultra-fast response times [8]. ...
... However, self-discharge in organic artificial synapses prevents them from maintaining stable conductance states over long periods. This effect can be partially mitigated by careful device design [16] but gets worse for small, scaleddown devices [7] that are required for many applications. Therefore, the weights of such an organic neural network degrade and information and trained behavior are lost over times as short as a few minutes [17]. ...
Organic neuromorphic devices can accelerate neural networks and integrate with biological systems. Devices based on the biocompatible and conductive polymer PEDOT:PSS are fast, require low amounts of energy and perform well in crossbar simulations. However, parasitic electrochemical reactions lead to self-discharge and the fading of learned conductance states over time. This limits a neural network's operating time and requires complex compensation mechanisms. Spiking neural networks take inspiration from biology to implement local and always-on learning. We show that these spiking neural networks can function on organic neuromorphic hardware and compensate for self-discharge by continuously relearning and reinforcing forgotten states. In this work, we use a high-resolution charge transport model to describe the behavior of organic neuromorphic devices and create a computationally-efficient surrogate model. By integrating the surrogate model into a Brian 2 simulation, we can describe the behavior of spiking neural networks on organic neuromorphic hardware. A biologically-plausible two-layer network for recognizing 28x28 pixel MNIST images is trained and observed during self-discharge. The network achieves, for its size, competitive recognition results up to 82.5%. Building the network with forgetful devices yields superior accuracy during training with 84.5% compared to ideal devices. However, the trained networks without active spike-timing-dependent plasticity quickly lose their predictive performance. We show that online learning can keep the performance at a steady level close to the initial accuracy, even for idle rates of up to 90%. This performance is maintained when the output neuron's labels are not revalidated for up to 24 hours. These findings reconfirm the potential of organic neuromorphic devices for brain-inspired computing. Their biocompatibility and the demonstrated adaptability to spiking neural networks open the path toward close integration with multi-electrode arrays, drug-delivery devices, and other bio-interfacing systems as either full organic or hybrid organic-inorganic systems.
