Conjugated Backbone-Directed Side Chain-Electrolyte Coupling toward Nonvolatile Artificial Synapse

Conjugated Backbone-Directed Side Chain-Electrolyte Coupling toward Nonvolatile Artificial Synapse | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Conjugated Backbone-Directed Side Chain-Electrolyte Coupling toward Nonvolatile Artificial Synapse Eunho Lee, Junho Sung, Sein Chung, Byeongjun Jeon, Donghwa Lee, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7438138/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Achieving stable and nonvolatile synaptic plasticity remains a central challenge for organic neuromorphic devices. While previous efforts have independently focused on backbone or side chain modifications, the fundamental role of conjugated backbone design in directing side chain–electrolyte coupling has remained elusive. Here, we demonstrate that modulation of thiophene units in the backbone governs the spatial arrangement and ionic accessibility of glycol side chains, thereby enabling strong anion adsorption and long-term retention. Electrolyte-gated organic synaptic transistors (EGOSTs) with extended backbones exhibit pronounced structural reorganization, suppressed ion back-diffusion, and stable nonvolatile characteristics. Backbone-directed side chain–anion coupling was identified as the key mechanism driving enhanced charge transport and persistent doping. As artificial synapses, the device realizes robust neuromorphic functions including paired-pulse facilitation, long-term potentiation/depression, and achieves 94.5% accuracy in artificial neural network simulations. This work establishes conjugated backbone regulation as a facile strategy to control side chain–electrolyte interactions, offering new design principles for nonvolatile synaptic devices and advancing the development of reliable organic neuromorphic computing. Physical sciences/Materials science/Materials for devices/Electronic devices Physical sciences/Materials science/Nanoscale materials/Electronic properties and materials Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction The rapid development of artificial intelligence (AI) systems and the advent of the big data era have faced the fundamental limitations of the conventional von Neumann architecture. The physical separation between the central processing unit (CPU) and memory leads to a severe data bottleneck, which increase latency and energy consumption during large-scale data handling. 1 – 4 This phenomenon poses a critical challenge to the development of next-generation AI models. To overcome these limitations, neuromorphic computing, which mimics the neural networks of the human brain, has emerged as a promising paradigm for information processing. 5 – 7 Neuromorphic systems integrate memory and computation within a single unit, enabling the simultaneous storage and processing of data similar to biological synapses. This architecture facilitates energy-efficient and fast data processing. Accordingly, a major objective of neuromorphic computing is to realize in hardware the parallel processing capabilities of the biological brain that are essential for complex cognitive tasks. To this end, various synaptic devices have been proposed. Although two-terminal memristors are structurally simple, sharing a single path for both reading and writing operations fundamentally causes problems such as crosstalk, unpredictable switching behavior, and the need for complex control circuits. 8 – 12 These limitations highlight the urgent need for alternative synaptic device architectures that can deliver predictable operation, efficient ion modulation, and robust plasticity. Electrolyte-gated organic synaptic transistors (EGOSTs) have emerged as promising three-terminal devices that can overcome the inherent problems of two-terminal structures. 13 – 16 EGOSTs provide predictable synaptic behavior and simplified control because they separate programming operations at the gate from read operations through the source-drain channel. In addition, EGOSTs enable analog control of synaptic weights by utilizing ion doping mechanisms induced by applied voltages, effectively mimicking the plasticity observed in biological synapses. 17 – 21 To achieve high-performance EGOSTs, researchers have mainly pursued two strategies. 22 , 23 Backbone engineering, including donor-acceptor (D-A) design or regiochemistry to regulate the polymer chain and improve its electronic properties, environmental stability, and structural properties, 24 – 35 and side chain engineering, in which alkyl chains are replaced with polar functional groups, such as ethylene glycol, to strengthen and optimize ion transport. 36 – 44 These approaches treat the polymer backbone and side chains as independent design variables, overlooking their interdependence and potential effects. More importantly, the fundamental influence of backbone structure on the side chain spatial arrangement, molecular packing, and ionic accessibility remains unexplored. The backbone and side chain cannot be regarded as completely independent design parameters, since the electronic structure of the backbone inherently dictates the spatial arrangement and accessibility of the side chains to mobile ions. Understanding this interplay is therefore essential for establishing molecular-level design rules for reliable organic synapses. In this study, we systematically investigate an unexplored field to elucidate the fundamental correlation between backbone length and side chain functionality. To exclude interactions caused by complex molecular variables, we designed polymer chains consisting of 2, 3, and 4 thiophenes with identical side chains. This simplified design enables us to decouple confounding factors and isolate how backbone modulation alone reshapes the spatial arrangement of side chains and their accessibility to mobile ions. This strategy induces channel conductance changes by controlling the ion adsorption behavior on the side chains through backbone modulation. In doing so, it is revealed that backbone engineering is not an isolated parameter but a decisive factor that governs side chain–ion coupling, a relationship that has been largely overlooked in prior studies. The increase in the number of thiophene units in the backbone leads to strong interactions between the side chains and ions, which enhances the weight update and non-volatile properties. Our combined theoretical and experimental analyses established a unified picture of how backbone modulation dictates ion behavior. DFT calculations revealed that extending the conjugated backbone enhances the free volume and adsorption energy landscape for anions, while structural and spectroscopic characterizations demonstrated that these strengthened interactions drive crystalline rearrangements and persistent anion retention. Together, these insights show that ion-induced structural reorganization is the fundamental origin of the enhanced electrical stability and nonvolatile synaptic behavior observed in our devices. Further fabricated EGOSTs successfully mimicked biological synaptic properties such as paired-pulse facilitation (PPF) and long-term potentiation/depression (LTP/D). Based on enhanced performance, our device successfully simulated recognition in an artificial neural network (ANN). These findings systematically elucidate a direct molecular-level link between backbone regulation and side chain–ion interactions, providing a rational pathway to realize stable and high-performance synaptic functions. Results In the biological nervous system, signal transmission occurs through neurotransmitter movement between the presynaptic and postsynaptic neurons (Fig. 1 a). When an input spike fires, calcium influx into the presynaptic neuron triggers neurotransmitters release from the synaptic vesicle to the postsynaptic neuron, meditating receptors that in the post-synaptic neuron to generate excitatory post-synaptic potential (EPSP). 45 , 46 To faithfully mimic biological synaptic behavior, we fabricated EGOSTs with a three-terminal architecture (Fig. 1 b). The input pulse applied through the gate electrode generates the movement of mobile ions in the electrolyte, which reacts with the polymer film and causes a change in the current flowing to the source/drain, resulting in the generation of an excitatory post-synaptic current (EPSC). Through the operation of these devices, our artificial synapses can mimic biological synaptic properties such as PPF, and LTP/D. To understand the difference in synaptic characteristics depending on the number of the thiophene units in backbone, we designed artificial synapses using poly[3,3'-bis[2-[2-(2-methoxyethoxy)ethoxy]ethoxy][2,2'-bithiophene]-5,5'-diyl] p(g2T), poly[3,3'-bis[2-[2-(2-methoxyethoxy)ethoxy]ethoxy]-2,2':5',2'‘-terthiophene-5,5’'-diyl] p(g2T -T), and poly[2-(3,3′-bis(2-(2-(2-methoxyethoxy)ethoxy)ethoxy)-[2,2’-bithiophen]-5-yl)thieno[3,2-b]thiophene] p(g2T-TT) (Fig. 1 c). The repeating unit lengths were confirmed via Avogadro to be 7.8, 11.7, and 12.92 Å for p(g2T), p(g2T-T), and p(g2T-TT), respectively, providing a systematic platform for investigating the effect of backbone length on synaptic plasticity. To investigate the electrical characteristics of the fabricated devices, we measured the transfer curve of p(g2T), p(g2T-T), and p(g2T-TT)-based EGOSTs at gate voltages ( V G ) ranging from + 2 V to − 2.5 V under a drain voltage − 1 mV (Fig. 2 a). As shown in Fig. 2 b, several key parameters are extracted including threshold voltage ( V th ), hysteresis window (Δ V hys ), maximum current ( I DS, max ), and transconductance ( g m ) to evaluate the electrical characteristics for each polymer film. The hysteresis window was defined as the full width at half maximum (FWHM) for the I DS, max , and the transconductance was calculated as g m = ∂ I DS /∂ V G . 41,47 The results revealed systematic changes in electrical characteristics with increasing thiophene units in the backbone. p(g2T-TT) exhibited the lowest threshold voltage (− 0.61 V) compared to p(g2T) (− 1.13 V) and p(g2T-T) (− 0.89 V), indicating enhanced charge transport efficiency. As the thiophene in the backbone increased in p(g2T-R) both I DS, max and g m increased progressively. These electrical improvements can be attributed to the increased spacing between side chains as thiophene units are added to the backbone. The expanded intermolecular spacing facilitates more efficient ion doping and reduces the energy barrier for charge transport. 48 In addition, p(g2T-TT) showed the largest hysteresis window, suggesting that the back-diffusion of doped TFSI anions is significantly suppressed, which is crucial for implementing non-volatile synaptic characteristics. 40 , 49 To confirm the statistical validity of these results, we verified the variation across the five devices (Fig. S2). To determine the effect of the backbone design of p(g2T-R)-based polymers on the characteristics of artificial synapses, we observed changes in synaptic plasticity. First, we examined the short-term plasticity (STP) behavior by applying single pulses with varying amplitudes ( V pulse = − 2, − 1.75, − 1.5, and − 1.25 V) and monitoring the EPSC response (Fig. 2 c). p(g2T) and p(g2T-T) exhibited typical STP properties, with EPSC returning to the initial state after single pulse application at all voltages. However, the p(g2T-TT) EGOST maintained EPSC at higher level than the initial state even with single pulse application of − 1.5 V or higher, suggesting the potential of implementing long-term memory characteristics. 50 – 52 Additionally, to further investigate short-term synaptic dynamics, we performed paired-pulse facilitation (PPF) properties by applying a pair of − 2 V pulses with different interval times (Δ t ) (Fig. 2 d). 53 PPF is a phenomenon where the response to the second stimulus (A 2 ) is magnified relative to the response to the first stimulus (A 1 ) during the application of two consecutive stimuli, and the increase can be quantified as PPF index = A 2 /A 1 . All EGOSTs showed the maximum PPF index showed the largest value when Δ t was 20 ms for all EGOSTs, with values decreasing as the interval increased. The PPF decay was fitted using a double exponential function as shown in Fig. 2 e: 54 $$\:PPF=1+{C}_{1}{exp}\left(-\varDelta\:t/{\tau\:}_{1}\right)+{C}_{2}{exp}\left(-\varDelta\:t/{\tau\:}_{2}\right)$$ 1 where, \(\:{C}_{n}\) and \(\:{\tau\:}_{n}\) are parameters for facilitation and relaxation time, respectively. (1 and 2 represent values in the slow phase and rapid phase.) The extracted time constants \(\:{\tau\:}_{1}\) and \(\:{\tau\:}_{2}\) were 70 and 381 ms in p(g2T), 97 and 921 ms in p(g2T-T), and 98 and 2000 ms in p(g2T-TT). These results are similar to previously reported results and biological behavior, suggesting that the STP properties have been successfully implemented. Notably, p(g2T-TT) exhibited the longest decay time, indicating that enhanced synaptic plasticity can be achieved. 40 , 55 To evaluate long-term plasticity (LTP) and non-volatile properties depending on the number of thiophene groups in the backbone, we applied 10 consecutive pulses ( t pulse = t interval = 60 ms) with different voltages ( V pulse = − 1.25, − 1.5, − 1.75, − 2 V). (Fig. 2 f-h). In all EGOSTs, EPSC increased stepwise with pulse application and decreased gradually after the pulse sequence. This phenomenon occurs due to long-term changes in EPSC caused by excessive accumulation of TFSI anions near the polymer/electrolyte interface and partial penetration into the channel as a result of voltage application. Therefore, this trend is further enhanced as the applied voltage increases. p(g2T-TT) showed greater EPSC changes at every amplitude and the most pronounced synaptic plasticity enhancement compared to other EGOSTs. To validate the obtained performance quantitatively and statistically, we determined the memory level defined as EPSC t /EPSC max × 100 (%) and the EPSC response for five devices in each film (see Fig. S3). This change occurs because the back diffusion of ions doped into the film by voltage application is suppressed, thereby maintaining the conductivity of the channel. We also analyzed retention behavior by fitting a tri-exponential function to understand the decrease in EPSC due to ion back-diffusion (Fig. S4): 56 , 57 $$\:{I}_{norm}={I}_{0,\:norm}+{C}_{1}{exp}\left(-\varDelta\:t/{\tau\:}_{1}\right)+{C}_{2}{exp}\left(-\varDelta\:t/{\tau\:}_{2}\right)+{C}_{3}{exp}\left(-\varDelta\:t/{\tau\:}_{3}\right)$$ 2 where, \(\:{C}_{n}\) is a constant; \(\:{I}_{norm}\) is the normalized EPSC; \(\:{I}_{0,\:norm}\) is the initial current; \(\:t\) is time; \(\:{\tau\:}_{1}\) , \(\:{\tau\:}_{2}\) , and \(\:{\tau\:}_{3}\) are the ion relaxation times in the EDL, amorphous region, and crystalline region of the thin film, respectively (Fig. 2 i). The longest EPSC relaxation time, \(\:{\tau\:}_{3}\) , was 7.5 s in p(g2T-TT), which was longer than the relaxation times of 0.84 and 3.88 s in p(g2T) and p(g2T-T), respectively (Fig. 2 j). It means that non-volatile properties can be enhanced by modulating the backbone of the polymer chain. 58 Extended retention measurements over 240 s (Fig. 2 k-l) confirmed that the predicted EPSC levels based on 20 s measurements accurately matched the experimental observations, demonstrating stable long-term memory retention. These results suggest that enhanced plasticity can be sustained stably over a long period of time. These enhanced properties allowed the p(g2T-TT) device to mimic a variety of synaptic behaviors (Fig. S7). Density functional theory (DFT) calculations were performed to compare the adsorption behavior between p(g2T-R)-based polymers and TFSI anions depending on the length of the backbone. First, to focus on the interaction between the backbone and anions, glycol side chains were replaced with methyl groups and the number of rings in the backbone was adjusted accordingly. 59 Anions were then placed near the modified backbone (Fig. S8a). Simplified backbone models showed no clear trend (Fig. S8b). Rather, p(g2T-TT), which exhibited outstanding non-volatile characteristics, showed the lowest adsorption energy. Anions tended to move into areas previously occupied by chains, suggesting dynamic interactions between anions and polymer structures. To analyze the adsorption behavior of TFSI in more detail and investigate the effect of the side chain on p(g2T), p(g2T-T), and p(g2T-TT), we considered a polymer with four chains as a model system (Fig. 3 a). The asymmetric structure of p(g2T-T) results in different interchain distances at the upper ( u ) and lower ( l ) regions. So, we placed two anions in the structure, one positioned between the upper polymer chains and the other between the lower chains. To calculate adsorption energy, the single-point energy of full structure containing both anions and the polymer was compared with the sum of the single-point energies calculated separately for the polymer with one anion and for the isolated second anion. $$\:{E}_{ad}={E}_{polymer+2anions}-({E}_{polymer+1anion\left(u/l\right)}+{E}_{1anion(l/u)})$$ 3 For p(g2T) and p(g2T-TT), the largest adsorption energy among all investigates sites was selected. For p(g2T-T), the lowest adsorption energy identified separately at the upper and lower positions, and the final value was taken as the average of these two values. In the system with side chains and two anions, the highest adsorption energies were observed in the order p(g2T-TT), p(g2T-T), and p(g2T) (Fig. 3 b). This trend suggests that the interaction between anions and polymer chains depends significantly on the polymer structure, particularly the spacing between side chains, consistent with the observed synaptic plasticity. Therefore, it can be inferred that interactions with polymer chains play a crucial role in adsorption behavior, and the length of the backbone is an important factor influencing adsorption energy between the side chain and the anions. These results indicate that p(g2T-TT), which has the highest amount of thiophene, exhibits the strongest interaction with anions, supporting its stronger non-volatile properties. Furthermore, to understand the differences in electrical properties depending on the backbone, we divided the polymer chains into fragments and calculated the Hirshfeld charges for each fragment (Fig. S13). p(g2T) exhibited an extremely localized charge distribution compared to p(g2T-T) and p(g2T-TT). This supports that lower EPSC responses were induced in p(g2T)-based EGOSTs by impeding effective charge transfer. It also suggests that improved electrical properties can be achieved by increasing the length of the thiophene backbone. We performed ultraviolet-visible absorption spectrometric of p(g2T-R)-based films to explore the characteristics changes depending on the doping of anions. A bias of − 2 V was applied for several tens of seconds to induce anion accumulation in the film (Fig. 3 c-e). To carry out quantitative comparison of the doping states, we normalized the spectra and compared the changes in the maximum absorption peaks (Fig. 3 f). p(g2T) and p(g2T-T) showed relatively small decrease in absorption intensity (from 1 to 0.97 and 0.95, respectively) after 30 seconds of bias application. These changes are attributed to the rapid back-diffusion of TFSI anions induced by voltage application. In contrast, the p(g2T-TT) film exhibited a relatively large decrease to 0.88 under bias application. This enhanced doping capability correlates directly with the DFT calculation results and enhanced synaptic plasticity observed in electrical measurements. In addition, chemical analysis was performed using time-of-flight secondary-ion mass spectroscopy (TOF-SIMS) depth profile to verify that the presence of TFSI anions through fluorine atom (F) detection in all polymer films (Fig. S14). p(g2T-TT) showed the strongest and most persistent F intensity during the depth profiling, providing additional evidence for enhanced anion retention consistent with the spectroscopic and electrical results. Grazing-incidence wide-angle X-ray scattering (GIWAXS) analysis were performed to investigate molecular orientation and packing changes upon doping (Fig. 3 g-h). All pristine polymer films exhibited distinct (h00) diffraction peaks in the out-of-plane (OoP) direction, indicating a lamellar layered structure, and (010) peaks in the in-plane (IP) direction, attributed to π–π stacking. These results indicate that all three pristine films have an edge-on orientation. The 1D-GIWAXS line-cut profiles for each film in the IP and OoP directions are shown in Fig. 3 i-j. The lamellar stacking (or π–π stacking) distances for p(g2T), p(g2T-T), and p(g2T-TT) were 16.42 (4.27), 12.46 (4.19), and 11.84 Å (4.10 Å), respectively (Fig. 3 k). As the thiophene units in the backbone increased, causing the distance between side chains to increase, a tendency toward a decrease in lamellar stacking and π–π stacking distances was observed. To observe changes in polymer packing due to doping, GIWAXS analysis was performed after applying a − 2 V bias. p(g2T) and p(g2T-T) showed no significant changes in the lamellar stacking distance upon doping. The π–π stacking distance also decreased with doping. These results are believed to be due to the shortening of the stacking distance as the ion injection amount increases. Conversely, p(g2T-TT) exhibited no distinction between IP and OoP signals upon doping. 60 , 61 This suggests that the increased thiophene units allowed TFSI anions to effectively penetrate the film and disrupt the molecular packing and creating new ion transport pathways. (Fig. 3 l). Based on p(g2T-R), we applied a simple dense packing model to further investigate structural changes resulting from the adjustment of the thiophene unit in the backbone of polymers. 62 Using the packing density model, we calculated the most densely packed areal attachment density for p(g2T), p(g2T-T), and p(g2T-TT) based on the tilt angle of the side chain (Fig. S16). The distance of the repeating units of each polymer was extracted via Avogadro and the distance between chains was determined by the π- π stacking distance, which was measured by GIWAXS. In the density model, p(g2T) was higher than the areal density of vertically arranged polyethylene (PE) (5.4 × 10¹⁴ cm − 2 ) for every angular region. This means that there is not enough free volume between the side chains, which can lead to limited doping by the TFSI anions. In comparison, p(g2T-TT) showed the largest sub-PE region among the three materials. These results are compatible with previous experiments and suggest that ions can approach the polymer backbone, providing conditions for efficient doping and stable doping even under various molecular orientation conditions. The long-term potentiation/depression (LTP/D) characteristics that induce changes in the weight state are important for EGOST to function as an artificial synapse in an artificial neural network (ANN) matrix. The low nonlinearity ( NL ), high dynamic range ( G max / G min ), and number of effective states ( NS eff ) obtained in LTP/D are considered key parameters for high accuracy in computation. To achieve potentiation in our EGOST, we applied 50 consecutive pulses of − 2 V ( t pulse = t interval = 60 ms) (Fig. 4 a). In the LTP region, p(g2T-R)-based devices show a gradual increase in EPSC as the number of pulses increases. p(g2T-TT), which showed the greatest response to TFSI ions, exhibits the largest increase compared to other polymers. To quantitatively compare the characteristics in the potentiation region, | NL |, G max / G min , and NS eff were extracted (Fig. 4 b). | NL | values were extracted by fitting the normalized increase trend of the measured EPSC to a curve defined based on an ideal linear line (Fig. S18 and Note S2). NS eff is calculated as the number of states that are strengthened/weakened beyond the noise level defined as 0.5% of G max − G min . The calculation results showed the lowest | NL | and the largest G max / G min , and NS eff in p(g2T-TT), which is consistent with previous results. To achieve LTD, 50 consecutive pulses of + 2 V ( t pulse = t interval = 60 ms) were applied immediately after potentiation (Fig. 4 c). Upon spike application, all EGOSTs exhibited a marked change in EPSC compared to potentiation. This reduction is attributed to back-diffusion caused by concentration gradients generated by ion doping into the film. As a result, relatively lower G max / G min , and NS eff , and larger | NL | values were observed compared to the LTP region (Fig. 4 d). Nevertheless, p(g2T-TT) showed better characteristics, as it exhibited strong adsorption and doping performance with TFSI. In order to evaluate the characteristics of EGOSTs for actual use in ANNs, it is necessary to consider not only the characteristics of a single LTP/D curve, but also variations such as cycle-to-cycle and device-to-device. We examined changes in the LTP/D curve over 12 cycles to confirm variations in weight updates between cycles in p(g2T-TT)-based artificial synapses (Fig. 4 e). During pulse application, our device exhibited stable LTP/D characteristics, and no significant performance degradation was observed (Fig. 4 f). Additionally, we observed that the device maintained consistent characteristics in metrics such as | NL |, G max / G min , and NS eff (Fig. 4 g). To compare the differences in characteristics between different devices, we performed LTP/D measurements with five p(g2T-TT)-based artificial synapses (Fig. S19). Across the five devices, our device did not show significant performance differences and was found to be statistically stable. Furthermore, to determine the variability and stability of the − 2 V potentiation pulses (P) and + 2 V depression pulses (D) applied, we examined the difference in EPSCs upon application of a randomly arranged pulse sequence of three P and three D (PPPDDD/PDDPPD) (Fig. S20). Both PPPDDD and PDDPPD showed similar levels of EPSCs after the end of the sequence. In the first cycle, the last state of PPPDDD had an error rate of 1.65% for the last state of PDDPPD, and in the 28th cycle, it had an error rate of 1.83%. These results support the reliability of our EGOSTs and show that they can maintain state stability over the long term. Finally, we performed multi-layer perceptron-based ANN simulations to demonstrate the feasibility of neuromorphic computing using our p(g2T-R)-based EGOSTs. To reflect the characteristics of the device, we performed calculations based on various parameters extracted from the LTP/D curve, such as NL , G max / G min , number of states, and variation. To evaluate the accuracy of the ANN simulation, we used the Modified National Institute of Standards and Technology database (MNIST) dataset, which consists of 10 groups of handwritten digits from 0 to 9. The neural network adopted a structure with input, hidden, and output layers (Fig. 4 h). Each layer consisted of 784 input neurons corresponding to the 28 × 28 pixel images of MNIST digits, 100 hidden neurons, and 10 output neurons corresponding to the groups of digits from 0 to 9. To measure the recognition accuracy of handwritten digits, we trained the model using 60,000 randomly selected images and then conducted inference during 10,000 tests. The structure of the parallel crossbar array used for the synaptic weight layer in the simulation is shown in Fig. 4 i. Learning and inference were repeated a total of 125 times, and it was confirmed that recognition accuracy improved in all EGOSTs as the number of learning iterations increased (Fig. 4 j). The simulation results showed that p(g2T) and p(g2T-T) had accuracies below 92%. In contrast, p(g2T-TT) showed a high accuracy of 94.5%. We performed additional simulations to reflect the variability in the statistical and long-term stability of device characteristics, including device-to-device (D2D), cycle-to-cycle (C2C), and conductance variation (Fig. S19c). 63 , 64 To consider the variability caused by repetitive pulse application, we calculated the relative standard deviation of weight update variation, defined as the percentage of the conductance range ( G max − G min ), using the LTP/D curves measured over 12 cycles. To take care of differences between devices, we calculated the relative standard deviation of the NL values from the five measured devices. We adopted the NL values in the LTD region, which showed larger values, to be conservative in variability. For conductance variation, we used both the relative standard deviations of G max and G min from the five different devices. The corresponding values for D2D, C2C, and conductance variation were 1.288%, 0.974, 1.17 × 10 − 7 ( G max ), and 2.79 × 10 − 8 ( G max ), respectively, and were applied for simulation. Despite taking all of these variations into account, p(g2T-TT) showed a high accuracy of 94.2% (Fig. 4 k). This is close to the accuracy achieved by devices with ideal conditions. To confirm the digit classification in the artificial neural network, we extracted the confusion matrix from the p(g2T-TT) EGOST-based simulation (Fig. 4 l-n). It shows what number the target digit was recognized as in the inference. The more inferences it made to that number, the darker blue color it showed. The first epochs showed random results, but by 115 epochs it successfully made an inference corresponding to the target, showing a diagonal pattern. In the initial epochs, the outcomes were random; however, by 115 epochs the network correctly inferred the target, yielding a distinct diagonal pattern. This was also seen in simulations that considered variation. These results suggest that the modulation of thiophene units in the backbone can be applied to neuromorphic computing by achieving enhanced synaptic plasticity and weight updates. In other words, p(g2T-TT) EGOSTs present a novel strategy to enhance neuromorphic computing performance and can be utilized for pattern identification, such as number recognition based on artificial neural networks. Discussion In conclusion, we first demonstrate that regulating the conjugated backbone architecture directly governs side chain–electrolyte coupling, providing the molecular basis for enhanced ion retention and long-term nonvolatile synaptic behavior. The p(g2T-TT)-based EGOSTs with the longest backbone exhibit the strongest interaction with ions compared to other devices, achieving distinct structural and electrical property changes. DFT and electrical analysis results confirmed that as the backbone lengthens, it exhibits strong adsorption energy between the anion and side chain, leading to long-term non-volatile characteristics over 240 seconds. As an artificial synapse, it achieves enhanced neuromorphic performance such as LTP/D and achieves 94.5% accuracy in ANN-based recognition simulation. These findings suggest that by controlling the adsorption behavior between the side chain and anion through backbone regulation, enhanced synaptic plasticity reinforcement and weight update can be achieved. It also presents new possibilities for design strategies that are limited to independent modulation of side chains and backbones. Methods Materials All materials directly use after purchase including: poly[3,3'-bis[2-[2-(2-methoxyethoxy)ethoxy]ethoxy][2,2'-bithiophene]-5,5'-diyl] p(g2T), poly[3,3'-bis[2-[2-(2-methoxyethoxy)ethoxy]ethoxy]-2,2':5',2'‘-terthiophene-5,5’'-diyl] p(g2T -T), and poly[2-(3,3′-bis(2-(2-(2-methoxyethoxy)ethoxy)ethoxy)-[2,2’-bithiophen]-5-yl)thieno[3,2-b]thiophene] p(g2T-TT) were purchased from Derthon Co. LTD (China), diethylmethyl (2-methoxyethyl) ammonium bis(trifluoromethylsulfonyl)imide (DEME-TFSI) was purchased from Sigma-Aldrich (USA). Fabrication of EGOST Our artificial synapses were fabricated on boron-doped p-type Si wafers covered with 300 nm of SiO₂. To ensure uniform film formation, wafers cut into 1.5 × 1.5 cm 2 pieces were sonicated in acetone and isopropyl alcohol (IPA) and then washed with DI water and N 2 gas blow. The polymer thin film was spin-coated at 1000 rpm for 60 seconds using a solution prepared at 5 mg mL − 1 in chloroform solvent. The electrodes were then patterned using a shadow mask via a thermal evaporator and deposited with 40 nm of gold in a high vacuum environment (≈ 10 − 6 Torr). Finally, as shown in Fig. S1 , the ionic liquid was drop-cast between the source/drain electrodes. Measurement and Analysis The electrical characteristics of the fabricated devices were measured using a Keithley 4200A-SCS (Tektronix, USA) semiconductor parameter analyzer at 10 − 3 A compliance, ambient and darkroom conditions. The UV absorption spectra of each thin film were obtained using a Lambda 465 (PerkinElmer, USA) UV/vis spectrophotometer. GIWAXS was conducted at the Pohang Accelerator Laboratory (PAL), South Korea, utilizing synchrotron radiation from the 9A beamline. Time-of-flight secondary-ion mass spectrometry (TOF-SIMS) analysis was performed using M6 (IONTOF GmbH, Germany). Devices with a bias of − 2 V applied for 20 seconds were used for GIWAXS and TOF-SIMS analysis dependent on doping. Artificial Neural Network The recognition simulation was performed in a Linux environment based on the open-source NeuroSim_V3.0. The artificial neural network had a multi-layer perceptron structure and learned and inferred MNIST digits. Parameters such as NL , G max / G min , and variation were considered to reflect the characteristics of EGOSTs. Declarations Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request. Acknowledgements This work was supported by a National Research Foundation of Korea (NRF) grant funded by the Ministry of Science and ICT (MSIT) (No. RS-2022-NR072040) and Korea Planning & Evolution Institute of Industrial Technology (RS-2024-00420537) grant funded by the Ministry of Trade, Industry & Energy (MOTIE). Author Information Author and Affiliations Department of Chemical and Biomolecular Engineering, Seoul National University of Science and Technology, Seoul 01811, Republic of Korea Junho Sung, Byeongjun Jeon, Donghwa Lee, Yoohyeon Jang, Bumjoon Seo & Eunho Lee Department of Chemical Engineering, Pohang University of Science and Technology, Pohang 37673, Republic of Korea Sein Chung & Jiyeong Shin Department of Chemical Engineering, Kumoh National Institute of Technology, Gumi 39177, Republic of Korea Myungjin An Contributions J.S., S.C., and E.L. conceived and designed the project. J.S., D.L., and Y.J. designed the experiments, conducted the experiments. B.J. and B.S. performed the computational analysis of the adsorption behavior and charge distribution. S.C. and J.S.(Shin) investigated the crystallinity of organic thin films. J.S. and B.J. wrote the manuscript. S.C., B.S. and E.L. revised the manuscript. All authors participated in data analysis and discussions. Contributions Corresponding authors Correspondence to Bumjoon Seo or Eunho Lee Ethics declarations Competing interests The authors declare no competing interests. References Beyond von Neumann (2020) Nat Nanotechnol 15:507–507 Liu C et al (2020) Two-dimensional materials for next-generation computing technologies. Nat Nanotechnol 15:545–557 Sangwan VK, Hersam MC (2020) Neuromorphic nanoelectronic materials. Nat Nanotechnol 15:517–528 Bae J, Lee SJ (2024) Current Progress in the Application of Artificial Intelligence for Nuclear Power Plant Operation. Korean J Chem Eng 41:2851–2870 Sebastian A, Le Gallo M, Khaddam-Aljameh R, Eleftheriou E (2020) Memory devices and applications for in-memory computing. Nat Nanotechnol 15:529–544 Schuman CD et al (2022) Opportunities for neuromorphic computing algorithms and applications. Nat Comput Sci 2:10–19 Boosting AI (2025) with neuromorphic computing. Nat. Comput. Sci. 5, 1–2 Pi S et al (2019) Memristor crossbar arrays with 6-nm half-pitch and 2-nm critical dimension. Nat Nanotechnol 14:35–39 Lee H-S et al (2020) Dual-Gated MoS2 Memtransistor Crossbar Array. Adv Funct Mater 30:2003683 Feng X et al (2021) Self-Selective Multi-Terminal Memtransistor Crossbar Array for In-Memory Computing. ACS Nano 15:1764–1774 Chen Z et al (2024) Cross-layer transmission realized by light-emitting memristor for constructing ultra-deep neural network with transfer learning ability. Nat Commun 15:1930 Lee S, Jung HS (2024) Cutting-Edge Advances in Perovskite Photovoltaic Devices and Applications. Korean J Chem Eng 41:3703–3715 Rivnay J et al (2018) Organic electrochemical transistors. Nat Rev Mater 3:17086 Yao B-W et al (2021) Non-Volatile Electrolyte-Gated Transistors Based on Graphdiyne/MoS2 with Robust Stability for Low-Power Neuromorphic Computing and Logic-In-Memory. Adv Funct Mater 31:2100069 Torricelli F et al (2021) Electrolyte-gated transistors for enhanced performance bioelectronics. Nat Rev Methods Primer 1:66 Zhang W et al (2024) Electrolyte-Gated Transistor Array (20 × 20) with Low-Programming Interference Based on Coplanar Gate Structure for Unsupervised Learning. Small Sci 4:2300306 Jang H, Bae GY, Kim SH, Sung J, Lee E (2024) Crosslinking-induced anion transport control for enhancing linearity in organic synaptic devices. Mater Horiz 11:4638–4650 Li J et al (2024) High-Density Artificial Synapse Array Consisting of Homogeneous Electrolyte-Gated Transistors. Adv Sci 11:2305430 Sung J et al (2025) Modulating Alkyl Groups in Copolymer to Control Ion Transport in Electrolyte-Gated Organic Transistors for Neuromorphic Computing. Small Struct 6:2400319 Liang Y et al (2025) Bioinspired Electrolyte-Gated Organic Synaptic Transistors: From Fundamental Requirements to Applications. Nano-Micro Lett 17:198 Ji J et al (2025) Single-transistor organic electrochemical neurons. Nat Commun 16:4334 Wang Y et al (2024) Designing organic mixed conductors for electrochemical transistor applications. Nat Rev Mater 9:249–265 Bardagot O et al (2025) Over Tenfold Increase in Current Amplification Due to Anisotropic Polymer Chain Alignment in Organic Electrochemical Transistors. Adv Mater 37:2420323 Ding B et al (2021) Influence of Backbone Curvature on the Organic Electrochemical Transistor Performance of Glycolated Donor–Acceptor Conjugated Polymers. Angew Chem Int Ed 60:19679–19684 Hallani RK et al (2021) Regiochemistry-Driven Organic Electrochemical Transistor Performance Enhancement in Ethylene Glycol-Functionalized Polythiophenes. J Am Chem Soc 143:11007–11018 Moser M et al (2021) Polaron Delocalization in Donor–Acceptor Polymers and its Impact on Organic Electrochemical Transistor Performance. Angew Chem Int Ed 60:7777–7785 Bardagot O et al (2025) Balancing Electroactive Backbone and Oligo(Ethylene Oxy) Side-Chain Content Improves Stability and Performance of Soluble PEDOT Copolymers in Organic Electrochemical Transistors. Adv Funct Mater 35:2412554 Chen Z et al (2025) π-Conjugated Polymers for High-Performance Organic Electrochemical Transistors: Molecular Design Strategies, Applications and Perspectives. Angew Chem Int Ed 64:e202423013 Wu H-Y et al (2023) Stable organic electrochemical neurons based on p-type and n-type ladder polymers. Mater Horiz 10:4213–4223 Giovannitti A et al (2020) Energetic Control of Redox-Active Polymers toward Safe Organic Bioelectronic Materials. Adv Mater 32:1908047 Sung M-J et al (2025) Organic Artificial Nerves: Neuromorphic Robotics and Bioelectronics. Chem Rev 125:2625–2664 Zhang X et al (2025) Backbone Engineering of Bithiophene Imide Dimer-Based Polymeric Mixed Ionic-Electronic Conductors for High-Performance n-Type Organic Electrochemical Transistors. Small 21:2408716 Zhang C et al (2025) Polythiophenes for High-Performance N-type Organic Electrochemical Transistors. Adv Funct Mater 35:2419706 Zhou H, Ye W, Jiang J, Wang Z (2024) Recent advances on surface modification of non-oxide photocatalysts towards efficient CO2 conversion. Carbon Lett 34:1569–1591 Penyan WL, Jasim LS (2024) Preparation and characterization of a novel pH-sensitive pectin-grafted-poly (acrylic acid-co-itaconic acid)/MWCNTs-COOH hydrogel composite for adsorption and controlled release of diclofenac sodium. Carbon Lett 34:1413–1429 Moser M et al (2020) Side Chain Redistribution as a Strategy to Boost Organic Electrochemical Transistor Performance and Stability. Adv Mater 32:2002748 Moser M et al (2020) Ethylene Glycol-Based Side Chain Length Engineering in Polythiophenes and its Impact on Organic Electrochemical Transistor Performance. Chem Mater 32:6618–6628 Jo I-Y et al (2024) High-Performance Organic Electrochemical Transistors Achieved by Optimizing Structural and Energetic Ordering of Diketopyrrolopyrrole-Based Polymers. Adv Mater 36:2307402 Tseng H-S et al (2024) Strategically tailoring ethylene glycol side chains with bridged-carbonyl ester in polythiophene-based organic electrochemical transistors for bioelectronics. Chem Eng J 486:150371 Sung M-J et al (2024) Overcoming the Trade-off Between Efficient Electrochemical Doping and High State Retention in Electrolyte-Gated Organic Synaptic Transistors. Adv Funct Mater 34:2312546 Sung J et al (2025) Improving ion uptake in artificial synapses through facilitated diffusion mechanisms. Mater Horiz 12:5225–5235 Kwon O et al (2024) Optimizing Molecular Weight of Polyethylene Glycol as an Additive for Stabilizing Zn Metal Anode in Aqueous Electrolyte. Korean J Chem Eng 41:539–544 Wang P et al (2024) Theoretically and Experimentally Exploring the Inhibition Effect of Imidazole Sulfonic Acid Derivatives with Multiple Adsorption Sites in 1.0 mol/L HCl. Korean J Chem Eng 41:2117–2132 An SH, Kim KY, Chung CW, Lee JU (2024) Development of cement nanocomposites reinforced by carbon nanotube dispersion using superplasticizers. Carbon Lett 34:1481–1494 London M, Schreibman A, Häusser M, Larkum ME, Segev I (2002) The information efficacy of a synapse. Nat Neurosci 5:332–340 Tang A-H et al (2016) A trans-synaptic nanocolumn aligns neurotransmitter release to receptors. Nature 536:210–214 Sung J et al (2025) Co-Stimuli-Driven 2D WSe2 Optoelectronic Synapses for Neuromorphic Computing. Small 21:2504024 Tan E et al (2022) The Role of Long-Alkyl-Group Spacers in Glycolated Copolymers for High-Performance Organic Electrochemical Transistors. Adv Mater 34:2202574 Oh S-J et al (2024) Non-volatile and Stretchable Polyvinyl Chloride-Based Solid-State Electrolyte for Capacitive Energy Storage. Korean J Chem Eng 41:1861–1869 Zeng X et al An Interlayer Strategy for Low-Voltage Thin-Film Organic Electrochemical Transistors. Small Methods n/a, 2500322 Lee CM, Kim Y, Kim W, Lee E, Lee EK (2025) High-Performance Synaptic Devices Based on Cross-linked Organic Electrochemical Transistors with Dual Ion Gel. Adv Funct Mater 35:2417539 Choi Y et al (2023) Physically defined long-term and short-term synapses for the development of reconfigurable analog-type operators capable of performing health care tasks. Sci Adv 9:eadg5946 Wang H et al (2018) A Ferroelectric/Electrochemical Modulated Organic Synapse for Ultraflexible, Artificial Visual-Perception System. Adv Mater 30:1803961 Kwon SM et al (2019) Environment-Adaptable Artificial Visual Perception Behaviors Using a Light-Adjustable Optoelectronic Neuromorphic Device Array. Adv Mater 31:1906433 Park SM, Won Y, Oh JH, Lee EK (2025) Enhancement of Synaptic Behavior in Organic Electrochemical Transistors via the Introduction of Layer-by-Layer Grown Metal-Organic Framework. Adv Mater Technol 10:2401316 Seo D-G et al (2019) Versatile neuromorphic electronics by modulating synaptic decay of single organic synaptic transistor: From artificial neural networks to neuro-prosthetics. Nano Energy 65:104035 Lee D et al (2025) Inter-Ion Mutual Repulsion Control for Nonvolatile Artificial Synapse. Adv Funct Mater 35:2412012 Roe DG et al (2024) Reconfigurable Logic Gates Capable of Device-Level Parallel Processing Through Multi-Input Synaptic Device. Adv Funct Mater 34:2312988 Luo X et al (2022) Circularly Recyclable Polymers Featuring Topochemically Weakened Carbon–Carbon Bonds. J Am Chem Soc 144:16588–16597 Choi W et al (2023) Unravelling Disorder Effects on Thermoelectric Properties of Semicrystalline Polymers in a Wide Range of Doping Levels. Small Methods 7:2201145 Kim J et al (2023) Remarkable conductivity enhancement in P-doped polythiophenes via rational engineering of polymer-dopant interactions. Mater Today Adv 18:100360 Kline RJ et al (2007) Critical Role of Side-Chain Attachment Density on the Order and Device Performance of Polythiophenes. Macromolecules 40:7960–7965 Chen P-Y, Peng X, Yu S, NeuroSim+ (2017) An integrated device-to-algorithm framework for benchmarking synaptic devices and array architectures. in 2017 IEEE International Electron Devices Meeting (IEDM) 6.1.1–6.1.4 10.1109/IEDM.2017.8268337 Chen P-Y, Peng X, Yu S, NeuroSim: (2018) A Circuit-Level Macro Model for Benchmarking Neuro-Inspired Architectures in Online Learning. IEEE Trans Comput -Aided Des Integr Circuits Syst 37:3067–3080 Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryInformationEunhoLee.docx Conjugated Backbone-Directed Side Chain-Electrolyte Coupling toward Nonvolatile Artificial Synapse Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7438138","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":504435737,"identity":"d3ca3682-ac74-4de0-85d5-58ce1614a218","order_by":0,"name":"Eunho Lee","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1klEQVRIie3RsQrCMBCA4SuBTFezVor0FSKCk+irNBQ661ZRsFI4R1cHX6YU6qK74FIfQQTRRUwdRSLdHPJDOAh85CAANts/5rFlVc+WPux9gz+Jk8l6ct6AkNeMCH9J03syUBRk+XUMwwDwUBlJe5vTCfexIs4jfwNRN3VX0kjkUdHJoUIT7DMEFoLg5sVGmkwez5qImyaL30R6isBN369wTYoQXDIT76gyH8u4Rzzu+Sh3XcLSTMQmyi/3+aCzZsX5isksEBibyeee+oMaAZvNZrN97QWbyDg6PMNO7QAAAABJRU5ErkJggg==","orcid":"","institution":"Seoul National University of Science and Technology","correspondingAuthor":true,"prefix":"","firstName":"Eunho","middleName":"","lastName":"Lee","suffix":""},{"id":504435738,"identity":"25b4b2ac-34a1-43ee-9f3b-ae3b5f4ef901","order_by":1,"name":"Junho Sung","email":"","orcid":"https://orcid.org/0009-0002-0513-2072","institution":"Seoul National University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Junho","middleName":"","lastName":"Sung","suffix":""},{"id":504435739,"identity":"bbaaed56-a9ff-4237-b38c-6716ee3195f2","order_by":2,"name":"Sein Chung","email":"","orcid":"https://orcid.org/0000-0003-3953-5208","institution":"POSTECH","correspondingAuthor":false,"prefix":"","firstName":"Sein","middleName":"","lastName":"Chung","suffix":""},{"id":504435740,"identity":"1649b58d-ef98-4691-9279-a2735b59f03d","order_by":3,"name":"Byeongjun Jeon","email":"","orcid":"","institution":"Seoul National University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Byeongjun","middleName":"","lastName":"Jeon","suffix":""},{"id":504435741,"identity":"0c2f46d4-a5d6-4050-a6d6-39dbad3733f6","order_by":4,"name":"Donghwa Lee","email":"","orcid":"","institution":"Seoul National University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Donghwa","middleName":"","lastName":"Lee","suffix":""},{"id":504435742,"identity":"b6c74204-c2f8-4c83-bd23-713d95d3e3ea","order_by":5,"name":"Myungjin An","email":"","orcid":"","institution":"Kumoh National Institute of Technology","correspondingAuthor":false,"prefix":"","firstName":"Myungjin","middleName":"","lastName":"An","suffix":""},{"id":504435743,"identity":"ac115f89-0a39-43df-815b-3c2f5568afed","order_by":6,"name":"Jihyeong Shin","email":"","orcid":"","institution":"Pohang University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Jihyeong","middleName":"","lastName":"Shin","suffix":""},{"id":504435744,"identity":"c8682923-0a71-47c1-9215-2d724957d0a3","order_by":7,"name":"Yoohyeon Jang","email":"","orcid":"","institution":"Seoul National University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Yoohyeon","middleName":"","lastName":"Jang","suffix":""},{"id":504435745,"identity":"6886207b-b96a-41b7-9c8c-f22d05d368c2","order_by":8,"name":"Bumjoon Seo","email":"","orcid":"https://orcid.org/0000-0002-5029-1593","institution":"Purdue University West Lafayette","correspondingAuthor":false,"prefix":"","firstName":"Bumjoon","middleName":"","lastName":"Seo","suffix":""}],"badges":[],"createdAt":"2025-08-23 03:00:42","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7438138/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7438138/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":89893224,"identity":"062f8045-6cba-480d-a17c-81c92c11c1ce","added_by":"auto","created_at":"2025-08-26 07:58:17","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":182509,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eBio-inspired artificial synapse design and polymeric structures for neuromorphic devices.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea\u003c/strong\u003e) A Schematic illustration of signal transduction process in biological nervous system consisting of pre-/postsynaptic neurons and synapses; EPSPs are triggered by the transfer of neurotransmitters upon the firing of the input spikes. \u003cstrong\u003eb\u003c/strong\u003e) The artificial synapse-based p(g2T-R) having the architecture of a 3-terminal transistor; EPSCs are evoked by the movements of ions in response to the applied pulses. \u003cstrong\u003ec\u003c/strong\u003e) Chemical structures of p(g2T), p(g2T-T), and p(g2T-TT) polymers varying the number of thiophene groups in the backbone.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7438138/v1/f2cc1740ed83ff3857600cd9.png"},{"id":89893604,"identity":"8a6f019b-8089-4e93-bd62-4b2affade0eb","added_by":"auto","created_at":"2025-08-26 08:06:18","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":231544,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSynaptic plasticity characteristics of polymer-based transistor devices.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea\u003c/strong\u003e) Transfer characteristics of p(g2T), p(g2T-T), and p(g2T-TT) devices under \u003cem\u003eV\u003c/em\u003e\u003csub\u003eDS\u003c/sub\u003e = −1 mV. Lines and dashed lines represent drain current and gate current, respectively. \u003cstrong\u003eb\u003c/strong\u003e) The Obtained hysteresis window (Δ\u003cem\u003eV\u003c/em\u003e\u003csub\u003ehys\u003c/sub\u003e), threshold voltage (\u003cem\u003eV\u003c/em\u003e\u003csub\u003eth\u003c/sub\u003e), maximum current (\u003cem\u003eI\u003c/em\u003e\u003csub\u003eD, max\u003c/sub\u003e), and transconductance (\u003cem\u003eg\u003c/em\u003e\u003csub\u003em\u003c/sub\u003e) from the transfer curves of each device. \u003cstrong\u003ec\u003c/strong\u003e) Short-term plasticity (STP) properties obtained from synaptic devices with a single pulse (\u003cem\u003eV\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e = −2, −1.75, −1.5, and −1.25 V). \u003cstrong\u003ed\u003c/strong\u003e) ESPC response to applying two continuous pulses (\u003cem\u003et\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e = 60 ms, \u003cem\u003et\u003c/em\u003e\u003csub\u003einterval\u003c/sub\u003e = 60 ms , \u003cem\u003eV\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e = −2 V). \u003cstrong\u003ee\u003c/strong\u003e) Changes in PPF index (A\u003csub\u003e2\u003c/sub\u003e/A\u003csub\u003e1\u003c/sub\u003e) at each device as a function of pulse application time intervals from 20 to 3000 ms. \u003cstrong\u003ef-h\u003c/strong\u003e) Long-term plasticity (LTP) curves measured with 10 consecutive pulses (\u003cem\u003et\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e = 60 ms, \u003cem\u003et\u003c/em\u003e\u003csub\u003einterval\u003c/sub\u003e = 60 ms) of p(g2T), p(g2T-T), and p(g2T-TT) devices. \u003cstrong\u003ei\u003c/strong\u003e) Schematic illustration of the de-doping time constant τ\u003csub\u003e1\u003c/sub\u003e (electrical double layer), τ\u003csub\u003e2\u003c/sub\u003e (amorphous region), and τ\u003csub\u003e3\u003c/sub\u003e (crystalline region) extracted from the tri-exponential function. \u003cstrong\u003ej\u003c/strong\u003e) Calculated values of the time constant (τ\u003csub\u003e1\u003c/sub\u003e, τ\u003csub\u003e2\u003c/sub\u003e, and τ\u003csub\u003e3\u003c/sub\u003e) for each material. \u003cstrong\u003ek\u003c/strong\u003e) Comparison of predicted and experimental memory levels, derived from the EPSC decay for p(g2T), p(g2T-T), and p(g2T-TT) devices. \u003cstrong\u003el\u003c/strong\u003e) Comparison between predicted and measured memory levels at 240 seconds.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7438138/v1/bc4f8fe32e492a18a70fa275.png"},{"id":89893608,"identity":"a9ab1650-2651-4352-8421-0c9e4e2d7543","added_by":"auto","created_at":"2025-08-26 08:06:18","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":346352,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSpectroscopic and crystallographic analysis of organic films upon doping of ions.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea\u003c/strong\u003e) Schematic illustration of the optimized conformation with two adsorbed anions. \u003cstrong\u003eb\u003c/strong\u003e) adsorption energy between anions and each polymer conformation. Changes in the UV-vis spectra of \u003cstrong\u003ec\u003c/strong\u003e) p(g2T), \u003cstrong\u003ed\u003c/strong\u003e) p(g2T-T), and \u003cstrong\u003ee\u003c/strong\u003e) p(g2T-TT) upon bias application of −2 V. \u003cstrong\u003ef\u003c/strong\u003e) Normalized peak absorbance over time as a function of bias application. 2D grazing incidence wide-angle X-ray scattering (GIWAXS) images of \u003cstrong\u003eg\u003c/strong\u003e) pristine and h) doped films via pulse application. 1D GIWAXS patterns of i) pristine and \u003cstrong\u003ej\u003c/strong\u003e) doped films of each polymer, extracted in the in-plane and out-of-plane directions. \u003cstrong\u003ek\u003c/strong\u003e) π-π stacking distance from the IP (010) peak and lamellar stacking distance from the OoP (100) peak of pristine polymer films. \u003cstrong\u003el\u003c/strong\u003e) Schematic of ion penetration into the p(g2T-R)-based polymer structure with doping.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7438138/v1/49e41c8ada5e4187ab4dba36.png"},{"id":89893605,"identity":"2b08e6c0-c0d6-4cc2-b00a-f70b0e766dd7","added_by":"auto","created_at":"2025-08-26 08:06:18","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":373287,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePulse-based weight modulation and pattern recognition performance of polymer synaptic networks.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea\u003c/strong\u003e) Long-term potentiation (LTP) curves with application 50 consecutive pulses (\u003cem\u003eV\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e = −2 V, \u003cem\u003et\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e = 60 ms, \u003cem\u003et\u003c/em\u003e\u003csub\u003einterval\u003c/sub\u003e = 60 ms.) of p(g2T-R) based devices. \u003cstrong\u003eb\u003c/strong\u003e) Number of effective state (\u003cem\u003eNS\u003c/em\u003e\u003csub\u003eeff\u003c/sub\u003e), \u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e/\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e, and nonlinearity (\u003cem\u003eNL\u003c/em\u003e) values in the LTP region. \u003cstrong\u003ec\u003c/strong\u003e) The properties of long-term depression (LTD) when applied to 50 continuous pulses (\u003cem\u003eV\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e = +2 V, \u003cem\u003et\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e = 60 ms, \u003cem\u003et\u003c/em\u003e\u003csub\u003einterval\u003c/sub\u003e = 60 ms.) of p(g2T-R) based devices. \u003cstrong\u003ed\u003c/strong\u003e) \u003cem\u003eNS\u003c/em\u003e\u003csub\u003eeff\u003c/sub\u003e, \u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e/\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e, and \u003cem\u003eNL\u003c/em\u003e values extracted from LTD curves. \u003cstrong\u003ee\u003c/strong\u003e) LTP/D characteristics obtained from 1200 pulses applied to p(g2T-TT) artificial synapse. \u003cstrong\u003ef\u003c/strong\u003e) Cycle-to-cycle variation of LTP/D curves over 12 cycles. \u003cstrong\u003eg\u003c/strong\u003e) Trends in \u003cem\u003eNS\u003c/em\u003e\u003csub\u003eeff\u003c/sub\u003e, \u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e/\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e, and \u003cem\u003eNL\u003c/em\u003e\u003csub\u003eLTP/D\u003c/sub\u003e for repeated curves. \u003cstrong\u003eh\u003c/strong\u003e) Multilayer perceptron-based ANN structure with 784 neurons for input layer, 100 neurons for hidden layer, and 10 neurons for output layer. \u003cstrong\u003ei\u003c/strong\u003e) The circuit diagram of artificial synapse-based crossbar array. \u003cstrong\u003ej\u003c/strong\u003e) Recognition accuracy based on p(g2T-R) artificial synapses. \u003cstrong\u003ek\u003c/strong\u003e) Comparison of the highest recognition accuracy between the ideal device, the regular p(g2T-TT) device, and the p(g2T-TT) device with variation considered. Confusion matrix for recognition accuracy of MNIST digit on p(g2T-TT) device at \u003cstrong\u003el\u003c/strong\u003e) 1 epoch, \u003cstrong\u003em\u003c/strong\u003e) 110 epochs, and \u003cstrong\u003en\u003c/strong\u003e) considering variation.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7438138/v1/d03d59e050bf4c9189dda73a.png"},{"id":90656808,"identity":"bb53f3f6-87a9-401b-80ee-16ed01a65aab","added_by":"auto","created_at":"2025-09-05 10:19:20","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1850375,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7438138/v1/b7658cc2-002d-4ca4-afaa-e928205e0764.pdf"},{"id":89893229,"identity":"12df1963-4f9d-401b-b084-61c52224140f","added_by":"auto","created_at":"2025-08-26 07:58:18","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":31541856,"visible":true,"origin":"","legend":"Conjugated Backbone-Directed Side Chain-Electrolyte Coupling toward Nonvolatile Artificial Synapse","description":"","filename":"SupplementaryInformationEunhoLee.docx","url":"https://assets-eu.researchsquare.com/files/rs-7438138/v1/861868cf51004f1b09b35b52.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Conjugated Backbone-Directed Side Chain-Electrolyte Coupling toward Nonvolatile Artificial Synapse","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe rapid development of artificial intelligence (AI) systems and the advent of the big data era have faced the fundamental limitations of the conventional von Neumann architecture. The physical separation between the central processing unit (CPU) and memory leads to a severe data bottleneck, which increase latency and energy consumption during large-scale data handling.\u003csup\u003e\u003cspan additionalcitationids=\"CR2 CR3\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e This phenomenon poses a critical challenge to the development of next-generation AI models. To overcome these limitations, neuromorphic computing, which mimics the neural networks of the human brain, has emerged as a promising paradigm for information processing.\u003csup\u003e\u003cspan additionalcitationids=\"CR6\" citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e Neuromorphic systems integrate memory and computation within a single unit, enabling the simultaneous storage and processing of data similar to biological synapses. This architecture facilitates energy-efficient and fast data processing. Accordingly, a major objective of neuromorphic computing is to realize in hardware the parallel processing capabilities of the biological brain that are essential for complex cognitive tasks. To this end, various synaptic devices have been proposed. Although two-terminal memristors are structurally simple, sharing a single path for both reading and writing operations fundamentally causes problems such as crosstalk, unpredictable switching behavior, and the need for complex control circuits.\u003csup\u003e\u003cspan additionalcitationids=\"CR9 CR10 CR11\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e These limitations highlight the urgent need for alternative synaptic device architectures that can deliver predictable operation, efficient ion modulation, and robust plasticity.\u003c/p\u003e\u003cp\u003eElectrolyte-gated organic synaptic transistors (EGOSTs) have emerged as promising three-terminal devices that can overcome the inherent problems of two-terminal structures.\u003csup\u003e\u003cspan additionalcitationids=\"CR14 CR15\" citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e EGOSTs provide predictable synaptic behavior and simplified control because they separate programming operations at the gate from read operations through the source-drain channel. In addition, EGOSTs enable analog control of synaptic weights by utilizing ion doping mechanisms induced by applied voltages, effectively mimicking the plasticity observed in biological synapses.\u003csup\u003e\u003cspan additionalcitationids=\"CR18 CR19 CR20\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e To achieve high-performance EGOSTs, researchers have mainly pursued two strategies.\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e Backbone engineering, including donor-acceptor (D-A) design or regiochemistry to regulate the polymer chain and improve its electronic properties, environmental stability, and structural properties,\u003csup\u003e\u003cspan additionalcitationids=\"CR25 CR26 CR27 CR28 CR29 CR30 CR31 CR32 CR33 CR34\" citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e and side chain engineering, in which alkyl chains are replaced with polar functional groups, such as ethylene glycol, to strengthen and optimize ion transport.\u003csup\u003e\u003cspan additionalcitationids=\"CR37 CR38 CR39 CR40 CR41 CR42 CR43\" citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e These approaches treat the polymer backbone and side chains as independent design variables, overlooking their interdependence and potential effects. More importantly, the fundamental influence of backbone structure on the side chain spatial arrangement, molecular packing, and ionic accessibility remains unexplored. The backbone and side chain cannot be regarded as completely independent design parameters, since the electronic structure of the backbone inherently dictates the spatial arrangement and accessibility of the side chains to mobile ions. Understanding this interplay is therefore essential for establishing molecular-level design rules for reliable organic synapses.\u003c/p\u003e\u003cp\u003eIn this study, we systematically investigate an unexplored field to elucidate the fundamental correlation between backbone length and side chain functionality. To exclude interactions caused by complex molecular variables, we designed polymer chains consisting of 2, 3, and 4 thiophenes with identical side chains. This simplified design enables us to decouple confounding factors and isolate how backbone modulation alone reshapes the spatial arrangement of side chains and their accessibility to mobile ions. This strategy induces channel conductance changes by controlling the ion adsorption behavior on the side chains through backbone modulation. In doing so, it is revealed that backbone engineering is not an isolated parameter but a decisive factor that governs side chain\u0026ndash;ion coupling, a relationship that has been largely overlooked in prior studies. The increase in the number of thiophene units in the backbone leads to strong interactions between the side chains and ions, which enhances the weight update and non-volatile properties. Our combined theoretical and experimental analyses established a unified picture of how backbone modulation dictates ion behavior. DFT calculations revealed that extending the conjugated backbone enhances the free volume and adsorption energy landscape for anions, while structural and spectroscopic characterizations demonstrated that these strengthened interactions drive crystalline rearrangements and persistent anion retention. Together, these insights show that ion-induced structural reorganization is the fundamental origin of the enhanced electrical stability and nonvolatile synaptic behavior observed in our devices. Further fabricated EGOSTs successfully mimicked biological synaptic properties such as paired-pulse facilitation (PPF) and long-term potentiation/depression (LTP/D). Based on enhanced performance, our device successfully simulated recognition in an artificial neural network (ANN). These findings systematically elucidate a direct molecular-level link between backbone regulation and side chain\u0026ndash;ion interactions, providing a rational pathway to realize stable and high-performance synaptic functions.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eIn the biological nervous system, signal transmission occurs through neurotransmitter movement between the presynaptic and postsynaptic neurons (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ea). When an input spike fires, calcium influx into the presynaptic neuron triggers neurotransmitters release from the synaptic vesicle to the postsynaptic neuron, meditating receptors that in the post-synaptic neuron to generate excitatory post-synaptic potential (EPSP).\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e45\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e To faithfully mimic biological synaptic behavior, we fabricated EGOSTs with a three-terminal architecture (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003eb). The input pulse applied through the gate electrode generates the movement of mobile ions in the electrolyte, which reacts with the polymer film and causes a change in the current flowing to the source/drain, resulting in the generation of an excitatory post-synaptic current (EPSC). Through the operation of these devices, our artificial synapses can mimic biological synaptic properties such as PPF, and LTP/D. To understand the difference in synaptic characteristics depending on the number of the thiophene units in backbone, we designed artificial synapses using poly[3,3\u0026apos;-bis[2-[2-(2-methoxyethoxy)ethoxy]ethoxy][2,2\u0026apos;-bithiophene]-5,5\u0026apos;-diyl] p(g2T), poly[3,3\u0026apos;-bis[2-[2-(2-methoxyethoxy)ethoxy]ethoxy]-2,2\u0026apos;:5\u0026apos;,2\u0026apos;\u0026lsquo;-terthiophene-5,5\u0026rsquo;\u0026apos;-diyl] p(g2T -T), and poly[2-(3,3\u0026prime;-bis(2-(2-(2-methoxyethoxy)ethoxy)ethoxy)-[2,2\u0026rsquo;-bithiophen]-5-yl)thieno[3,2-b]thiophene] p(g2T-TT) (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ec). The repeating unit lengths were confirmed via Avogadro to be 7.8, 11.7, and 12.92 \u0026Aring; for p(g2T), p(g2T-T), and p(g2T-TT), respectively, providing a systematic platform for investigating the effect of backbone length on synaptic plasticity.\u003c/p\u003e\n\u003cp\u003eTo investigate the electrical characteristics of the fabricated devices, we measured the transfer curve of p(g2T), p(g2T-T), and p(g2T-TT)-based EGOSTs at gate voltages (\u003cem\u003eV\u003c/em\u003e\u003csub\u003eG\u003c/sub\u003e) ranging from +\u0026thinsp;2 V to \u0026minus;\u0026thinsp;2.5 V under a drain voltage \u0026minus;\u0026thinsp;1 mV (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ea). As shown in Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eb, several key parameters are extracted including threshold voltage (\u003cem\u003eV\u003c/em\u003e\u003csub\u003eth\u003c/sub\u003e), hysteresis window (\u0026Delta;\u003cem\u003eV\u003c/em\u003e\u003csub\u003ehys\u003c/sub\u003e), maximum current (\u003cem\u003eI\u003c/em\u003e\u003csub\u003eDS, max\u003c/sub\u003e), and transconductance (\u003cem\u003eg\u003c/em\u003e\u003csub\u003em\u003c/sub\u003e) to evaluate the electrical characteristics for each polymer film. The hysteresis window was defined as the full width at half maximum (FWHM) for the \u003cem\u003eI\u003c/em\u003e\u003csub\u003eDS, max\u003c/sub\u003e, and the transconductance was calculated as \u003cem\u003eg\u003c/em\u003e\u003csub\u003em\u003c/sub\u003e = \u0026part;\u003cem\u003eI\u003c/em\u003e\u003csub\u003eDS\u003c/sub\u003e/\u0026part;\u003cem\u003eV\u003c/em\u003e\u003csub\u003eG\u003c/sub\u003e.\u003csup\u003e41,47\u003c/sup\u003e The results revealed systematic changes in electrical characteristics with increasing thiophene units in the backbone. p(g2T-TT) exhibited the lowest threshold voltage (\u0026minus;\u0026thinsp;0.61 V) compared to p(g2T) (\u0026minus;\u0026thinsp;1.13 V) and p(g2T-T) (\u0026minus;\u0026thinsp;0.89 V), indicating enhanced charge transport efficiency. As the thiophene in the backbone increased in p(g2T-R) both \u003cem\u003eI\u003c/em\u003e\u003csub\u003eDS, max\u003c/sub\u003e and \u003cem\u003eg\u003c/em\u003e\u003csub\u003em\u003c/sub\u003e increased progressively. These electrical improvements can be attributed to the increased spacing between side chains as thiophene units are added to the backbone. The expanded intermolecular spacing facilitates more efficient ion doping and reduces the energy barrier for charge transport.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e In addition, p(g2T-TT) showed the largest hysteresis window, suggesting that the back-diffusion of doped TFSI anions is significantly suppressed, which is crucial for implementing non-volatile synaptic characteristics.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e40\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e To confirm the statistical validity of these results, we verified the variation across the five devices (Fig. S2).\u003c/p\u003e\n\u003cp\u003eTo determine the effect of the backbone design of p(g2T-R)-based polymers on the characteristics of artificial synapses, we observed changes in synaptic plasticity. First, we examined the short-term plasticity (STP) behavior by applying single pulses with varying amplitudes (\u003cem\u003eV\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;2, \u0026minus;\u0026thinsp;1.75, \u0026minus;\u0026thinsp;1.5, and \u0026minus;\u0026thinsp;1.25 V) and monitoring the EPSC response (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ec). p(g2T) and p(g2T-T) exhibited typical STP properties, with EPSC returning to the initial state after single pulse application at all voltages. However, the p(g2T-TT) EGOST maintained EPSC at higher level than the initial state even with single pulse application of \u0026minus;\u0026thinsp;1.5 V or higher, suggesting the potential of implementing long-term memory characteristics.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e50\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e Additionally, to further investigate short-term synaptic dynamics, we performed paired-pulse facilitation (PPF) properties by applying a pair of \u0026minus;\u0026thinsp;2 V pulses with different interval times (\u0026Delta;\u003cem\u003et\u003c/em\u003e) (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ed).\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e PPF is a phenomenon where the response to the second stimulus (A\u003csub\u003e2\u003c/sub\u003e) is magnified relative to the response to the first stimulus (A\u003csub\u003e1\u003c/sub\u003e) during the application of two consecutive stimuli, and the increase can be quantified as PPF index\u0026thinsp;=\u0026thinsp;A\u003csub\u003e2\u003c/sub\u003e/A\u003csub\u003e1\u003c/sub\u003e. All EGOSTs showed the maximum PPF index showed the largest value when \u0026Delta;\u003cem\u003et\u003c/em\u003e was 20 ms for all EGOSTs, with values decreasing as the interval increased. The PPF decay was fitted using a double exponential function as shown in Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ee:\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e$$\\:PPF=1+{C}_{1}{exp}\\left(-\\varDelta\\:t/{\\tau\\:}_{1}\\right)+{C}_{2}{exp}\\left(-\\varDelta\\:t/{\\tau\\:}_{2}\\right)$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003ewhere, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{C}_{n}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\tau\\:}_{n}\\)\u003c/span\u003e\u003c/span\u003e are parameters for facilitation and relaxation time, respectively. (1 and 2 represent values in the slow phase and rapid phase.) The extracted time constants \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\tau\\:}_{1}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\tau\\:}_{2}\\)\u003c/span\u003e\u003c/span\u003e were 70 and 381 ms in p(g2T), 97 and 921 ms in p(g2T-T), and 98 and 2000 ms in p(g2T-TT). These results are similar to previously reported results and biological behavior, suggesting that the STP properties have been successfully implemented. Notably, p(g2T-TT) exhibited the longest decay time, indicating that enhanced synaptic plasticity can be achieved.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e40\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n\u003cp\u003eTo evaluate long-term plasticity (LTP) and non-volatile properties depending on the number of thiophene groups in the backbone, we applied 10 consecutive pulses (\u003cem\u003et\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e = \u003cem\u003et\u003c/em\u003e\u003csub\u003einterval\u003c/sub\u003e = 60 ms) with different voltages (\u003cem\u003eV\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;1.25, \u0026minus;\u0026thinsp;1.5, \u0026minus;\u0026thinsp;1.75, \u0026minus;\u0026thinsp;2 V). (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ef-h). In all EGOSTs, EPSC increased stepwise with pulse application and decreased gradually after the pulse sequence. This phenomenon occurs due to long-term changes in EPSC caused by excessive accumulation of TFSI anions near the polymer/electrolyte interface and partial penetration into the channel as a result of voltage application. Therefore, this trend is further enhanced as the applied voltage increases. p(g2T-TT) showed greater EPSC changes at every amplitude and the most pronounced synaptic plasticity enhancement compared to other EGOSTs. To validate the obtained performance quantitatively and statistically, we determined the memory level defined as EPSC\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e/EPSC\u003csub\u003e\u003cem\u003emax\u003c/em\u003e\u003c/sub\u003e \u0026times; 100 (%) and the EPSC response for five devices in each film (see Fig. S3). This change occurs because the back diffusion of ions doped into the film by voltage application is suppressed, thereby maintaining the conductivity of the channel. We also analyzed retention behavior by fitting a tri-exponential function to understand the decrease in EPSC due to ion back-diffusion (Fig. S4):\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e56\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e$$\\:{I}_{norm}={I}_{0,\\:norm}+{C}_{1}{exp}\\left(-\\varDelta\\:t/{\\tau\\:}_{1}\\right)+{C}_{2}{exp}\\left(-\\varDelta\\:t/{\\tau\\:}_{2}\\right)+{C}_{3}{exp}\\left(-\\varDelta\\:t/{\\tau\\:}_{3}\\right)$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003ewhere, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{C}_{n}\\)\u003c/span\u003e\u003c/span\u003e is a constant; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{I}_{norm}\\)\u003c/span\u003e\u003c/span\u003e is the normalized EPSC; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{I}_{0,\\:norm}\\)\u003c/span\u003e\u003c/span\u003e is the initial current; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:t\\)\u003c/span\u003e\u003c/span\u003e is time; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\tau\\:}_{1}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\tau\\:}_{2}\\)\u003c/span\u003e\u003c/span\u003e, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\tau\\:}_{3}\\)\u003c/span\u003e\u003c/span\u003e are the ion relaxation times in the EDL, amorphous region, and crystalline region of the thin film, respectively (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ei). The longest EPSC relaxation time, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\tau\\:}_{3}\\)\u003c/span\u003e\u003c/span\u003e, was 7.5 s in p(g2T-TT), which was longer than the relaxation times of 0.84 and 3.88 s in p(g2T) and p(g2T-T), respectively (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ej). It means that non-volatile properties can be enhanced by modulating the backbone of the polymer chain.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e Extended retention measurements over 240 s (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ek-l) confirmed that the predicted EPSC levels based on 20 s measurements accurately matched the experimental observations, demonstrating stable long-term memory retention. These results suggest that enhanced plasticity can be sustained stably over a long period of time. These enhanced properties allowed the p(g2T-TT) device to mimic a variety of synaptic behaviors (Fig. S7).\u003c/p\u003e\n\u003cp\u003eDensity functional theory (DFT) calculations were performed to compare the adsorption behavior between p(g2T-R)-based polymers and TFSI anions depending on the length of the backbone. First, to focus on the interaction between the backbone and anions, glycol side chains were replaced with methyl groups and the number of rings in the backbone was adjusted accordingly.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e Anions were then placed near the modified backbone (Fig. S8a). Simplified backbone models showed no clear trend (Fig. S8b). Rather, p(g2T-TT), which exhibited outstanding non-volatile characteristics, showed the lowest adsorption energy. Anions tended to move into areas previously occupied by chains, suggesting dynamic interactions between anions and polymer structures.\u003c/p\u003e\n\u003cp\u003eTo analyze the adsorption behavior of TFSI in more detail and investigate the effect of the side chain on p(g2T), p(g2T-T), and p(g2T-TT), we considered a polymer with four chains as a model system (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ea). The asymmetric structure of p(g2T-T) results in different interchain distances at the upper (\u003cem\u003eu\u003c/em\u003e) and lower (\u003cem\u003el\u003c/em\u003e) regions. So, we placed two anions in the structure, one positioned between the upper polymer chains and the other between the lower chains. To calculate adsorption energy, the single-point energy of full structure containing both anions and the polymer was compared with the sum of the single-point energies calculated separately for the polymer with one anion and for the isolated second anion.\u003c/p\u003e\n\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e$$\\:{E}_{ad}={E}_{polymer+2anions}-({E}_{polymer+1anion\\left(u/l\\right)}+{E}_{1anion(l/u)})$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003eFor p(g2T) and p(g2T-TT), the largest adsorption energy among all investigates sites was selected. For p(g2T-T), the lowest adsorption energy identified separately at the upper and lower positions, and the final value was taken as the average of these two values. In the system with side chains and two anions, the highest adsorption energies were observed in the order p(g2T-TT), p(g2T-T), and p(g2T) (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eb). This trend suggests that the interaction between anions and polymer chains depends significantly on the polymer structure, particularly the spacing between side chains, consistent with the observed synaptic plasticity. Therefore, it can be inferred that interactions with polymer chains play a crucial role in adsorption behavior, and the length of the backbone is an important factor influencing adsorption energy between the side chain and the anions. These results indicate that p(g2T-TT), which has the highest amount of thiophene, exhibits the strongest interaction with anions, supporting its stronger non-volatile properties. Furthermore, to understand the differences in electrical properties depending on the backbone, we divided the polymer chains into fragments and calculated the Hirshfeld charges for each fragment (Fig. S13). p(g2T) exhibited an extremely localized charge distribution compared to p(g2T-T) and p(g2T-TT). This supports that lower EPSC responses were induced in p(g2T)-based EGOSTs by impeding effective charge transfer. It also suggests that improved electrical properties can be achieved by increasing the length of the thiophene backbone.\u003c/p\u003e\n\u003cp\u003eWe performed ultraviolet-visible absorption spectrometric of p(g2T-R)-based films to explore the characteristics changes depending on the doping of anions. A bias of \u0026minus;\u0026thinsp;2 V was applied for several tens of seconds to induce anion accumulation in the film (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ec-e). To carry out quantitative comparison of the doping states, we normalized the spectra and compared the changes in the maximum absorption peaks (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ef). p(g2T) and p(g2T-T) showed relatively small decrease in absorption intensity (from 1 to 0.97 and 0.95, respectively) after 30 seconds of bias application. These changes are attributed to the rapid back-diffusion of TFSI anions induced by voltage application. In contrast, the p(g2T-TT) film exhibited a relatively large decrease to 0.88 under bias application. This enhanced doping capability correlates directly with the DFT calculation results and enhanced synaptic plasticity observed in electrical measurements. In addition, chemical analysis was performed using time-of-flight secondary-ion mass spectroscopy (TOF-SIMS) depth profile to verify that the presence of TFSI anions through fluorine atom (F) detection in all polymer films (Fig. S14). p(g2T-TT) showed the strongest and most persistent F intensity during the depth profiling, providing additional evidence for enhanced anion retention consistent with the spectroscopic and electrical results.\u003c/p\u003e\n\u003cp\u003eGrazing-incidence wide-angle X-ray scattering (GIWAXS) analysis were performed to investigate molecular orientation and packing changes upon doping (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eg-h). All pristine polymer films exhibited distinct (h00) diffraction peaks in the out-of-plane (OoP) direction, indicating a lamellar layered structure, and (010) peaks in the in-plane (IP) direction, attributed to \u0026pi;\u0026ndash;\u0026pi; stacking. These results indicate that all three pristine films have an edge-on orientation. The 1D-GIWAXS line-cut profiles for each film in the IP and OoP directions are shown in Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ei-j. The lamellar stacking (or \u0026pi;\u0026ndash;\u0026pi; stacking) distances for p(g2T), p(g2T-T), and p(g2T-TT) were 16.42 (4.27), 12.46 (4.19), and 11.84 \u0026Aring; (4.10 \u0026Aring;), respectively (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ek). As the thiophene units in the backbone increased, causing the distance between side chains to increase, a tendency toward a decrease in lamellar stacking and \u0026pi;\u0026ndash;\u0026pi; stacking distances was observed. To observe changes in polymer packing due to doping, GIWAXS analysis was performed after applying a \u0026minus;\u0026thinsp;2 V bias. p(g2T) and p(g2T-T) showed no significant changes in the lamellar stacking distance upon doping. The \u0026pi;\u0026ndash;\u0026pi; stacking distance also decreased with doping. These results are believed to be due to the shortening of the stacking distance as the ion injection amount increases. Conversely, p(g2T-TT) exhibited no distinction between IP and OoP signals upon doping.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e60\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e This suggests that the increased thiophene units allowed TFSI anions to effectively penetrate the film and disrupt the molecular packing and creating new ion transport pathways. (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003el).\u003c/p\u003e\n\u003cp\u003eBased on p(g2T-R), we applied a simple dense packing model to further investigate structural changes resulting from the adjustment of the thiophene unit in the backbone of polymers.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e Using the packing density model, we calculated the most densely packed areal attachment density for p(g2T), p(g2T-T), and p(g2T-TT) based on the tilt angle of the side chain (Fig. S16). The distance of the repeating units of each polymer was extracted via Avogadro and the distance between chains was determined by the \u0026pi;- \u0026pi; stacking distance, which was measured by GIWAXS. In the density model, p(g2T) was higher than the areal density of vertically arranged polyethylene (PE) (5.4 \u0026times; 10\u0026sup1;⁴ cm\u003csup\u003e\u0026minus;\u0026thinsp;2\u003c/sup\u003e) for every angular region. This means that there is not enough free volume between the side chains, which can lead to limited doping by the TFSI anions. In comparison, p(g2T-TT) showed the largest sub-PE region among the three materials. These results are compatible with previous experiments and suggest that ions can approach the polymer backbone, providing conditions for efficient doping and stable doping even under various molecular orientation conditions.\u003c/p\u003e\n\u003cp\u003eThe long-term potentiation/depression (LTP/D) characteristics that induce changes in the weight state are important for EGOST to function as an artificial synapse in an artificial neural network (ANN) matrix. The low nonlinearity (\u003cem\u003eNL\u003c/em\u003e), high dynamic range (\u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e/\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e), and number of effective states (\u003cem\u003eNS\u003c/em\u003e\u003csub\u003eeff\u003c/sub\u003e) obtained in LTP/D are considered key parameters for high accuracy in computation. To achieve potentiation in our EGOST, we applied 50 consecutive pulses of \u0026minus;\u0026thinsp;2 V (\u003cem\u003et\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e = \u003cem\u003et\u003c/em\u003e\u003csub\u003einterval\u003c/sub\u003e = 60 ms) (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ea). In the LTP region, p(g2T-R)-based devices show a gradual increase in EPSC as the number of pulses increases. p(g2T-TT), which showed the greatest response to TFSI ions, exhibits the largest increase compared to other polymers. To quantitatively compare the characteristics in the potentiation region, |\u003cem\u003eNL\u003c/em\u003e|, \u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e/\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e, and \u003cem\u003eNS\u003c/em\u003e\u003csub\u003eeff\u003c/sub\u003e were extracted (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eb). |\u003cem\u003eNL\u003c/em\u003e| values were extracted by fitting the normalized increase trend of the measured EPSC to a curve defined based on an ideal linear line (Fig. S18 and Note S2). \u003cem\u003eNS\u003c/em\u003e\u003csub\u003eeff\u003c/sub\u003e is calculated as the number of states that are strengthened/weakened beyond the noise level defined as 0.5% of \u003cem\u003eG\u003c/em\u003e\u003csub\u003e\u003cem\u003emax\u003c/em\u003e\u003c/sub\u003e\u0026minus;\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e. The calculation results showed the lowest |\u003cem\u003eNL\u003c/em\u003e| and the largest \u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e/\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e, and \u003cem\u003eNS\u003c/em\u003e\u003csub\u003eeff\u003c/sub\u003e in p(g2T-TT), which is consistent with previous results. To achieve LTD, 50 consecutive pulses of +\u0026thinsp;2 V (\u003cem\u003et\u003c/em\u003e\u003csub\u003epulse\u003c/sub\u003e = \u003cem\u003et\u003c/em\u003e\u003csub\u003einterval\u003c/sub\u003e = 60 ms) were applied immediately after potentiation (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ec). Upon spike application, all EGOSTs exhibited a marked change in EPSC compared to potentiation. This reduction is attributed to back-diffusion caused by concentration gradients generated by ion doping into the film. As a result, relatively lower \u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e/\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e, and \u003cem\u003eNS\u003c/em\u003e\u003csub\u003eeff\u003c/sub\u003e, and larger |\u003cem\u003eNL\u003c/em\u003e| values were observed compared to the LTP region (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ed). Nevertheless, p(g2T-TT) showed better characteristics, as it exhibited strong adsorption and doping performance with TFSI.\u003c/p\u003e\n\u003cp\u003eIn order to evaluate the characteristics of EGOSTs for actual use in ANNs, it is necessary to consider not only the characteristics of a single LTP/D curve, but also variations such as cycle-to-cycle and device-to-device. We examined changes in the LTP/D curve over 12 cycles to confirm variations in weight updates between cycles in p(g2T-TT)-based artificial synapses (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ee). During pulse application, our device exhibited stable LTP/D characteristics, and no significant performance degradation was observed (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ef). Additionally, we observed that the device maintained consistent characteristics in metrics such as |\u003cem\u003eNL\u003c/em\u003e|, \u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e/\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e, and \u003cem\u003eNS\u003c/em\u003e\u003csub\u003eeff\u003c/sub\u003e (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eg). To compare the differences in characteristics between different devices, we performed LTP/D measurements with five p(g2T-TT)-based artificial synapses (Fig. S19). Across the five devices, our device did not show significant performance differences and was found to be statistically stable. Furthermore, to determine the variability and stability of the \u0026minus;\u0026thinsp;2 V potentiation pulses (P) and +\u0026thinsp;2 V depression pulses (D) applied, we examined the difference in EPSCs upon application of a randomly arranged pulse sequence of three P and three D (PPPDDD/PDDPPD) (Fig. S20). Both PPPDDD and PDDPPD showed similar levels of EPSCs after the end of the sequence. In the first cycle, the last state of PPPDDD had an error rate of 1.65% for the last state of PDDPPD, and in the 28th cycle, it had an error rate of 1.83%. These results support the reliability of our EGOSTs and show that they can maintain state stability over the long term.\u003c/p\u003e\n\u003cp\u003eFinally, we performed multi-layer perceptron-based ANN simulations to demonstrate the feasibility of neuromorphic computing using our p(g2T-R)-based EGOSTs. To reflect the characteristics of the device, we performed calculations based on various parameters extracted from the LTP/D curve, such as \u003cem\u003eNL\u003c/em\u003e, \u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e/\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e, number of states, and variation. To evaluate the accuracy of the ANN simulation, we used the Modified National Institute of Standards and Technology database (MNIST) dataset, which consists of 10 groups of handwritten digits from 0 to 9. The neural network adopted a structure with input, hidden, and output layers (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eh). Each layer consisted of 784 input neurons corresponding to the 28 \u0026times; 28 pixel images of MNIST digits, 100 hidden neurons, and 10 output neurons corresponding to the groups of digits from 0 to 9. To measure the recognition accuracy of handwritten digits, we trained the model using 60,000 randomly selected images and then conducted inference during 10,000 tests. The structure of the parallel crossbar array used for the synaptic weight layer in the simulation is shown in Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ei. Learning and inference were repeated a total of 125 times, and it was confirmed that recognition accuracy improved in all EGOSTs as the number of learning iterations increased (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ej). The simulation results showed that p(g2T) and p(g2T-T) had accuracies below 92%. In contrast, p(g2T-TT) showed a high accuracy of 94.5%.\u003c/p\u003e\n\u003cp\u003eWe performed additional simulations to reflect the variability in the statistical and long-term stability of device characteristics, including device-to-device (D2D), cycle-to-cycle (C2C), and conductance variation (Fig. S19c).\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e63\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e To consider the variability caused by repetitive pulse application, we calculated the relative standard deviation of weight update variation, defined as the percentage of the conductance range (\u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e\u0026minus;\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e), using the LTP/D curves measured over 12 cycles. To take care of differences between devices, we calculated the relative standard deviation of the \u003cem\u003eNL\u003c/em\u003e values from the five measured devices. We adopted the \u003cem\u003eNL\u003c/em\u003e values in the LTD region, which showed larger values, to be conservative in variability. For conductance variation, we used both the relative standard deviations of \u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e and \u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e from the five different devices. The corresponding values for D2D, C2C, and conductance variation were 1.288%, 0.974, 1.17 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;7\u003c/sup\u003e (\u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e), and 2.79 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;8\u003c/sup\u003e (\u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e), respectively, and were applied for simulation. Despite taking all of these variations into account, p(g2T-TT) showed a high accuracy of 94.2% (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ek). This is close to the accuracy achieved by devices with ideal conditions. To confirm the digit classification in the artificial neural network, we extracted the confusion matrix from the p(g2T-TT) EGOST-based simulation (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003el-n). It shows what number the target digit was recognized as in the inference. The more inferences it made to that number, the darker blue color it showed. The first epochs showed random results, but by 115 epochs it successfully made an inference corresponding to the target, showing a diagonal pattern. In the initial epochs, the outcomes were random; however, by 115 epochs the network correctly inferred the target, yielding a distinct diagonal pattern. This was also seen in simulations that considered variation. These results suggest that the modulation of thiophene units in the backbone can be applied to neuromorphic computing by achieving enhanced synaptic plasticity and weight updates. In other words, p(g2T-TT) EGOSTs present a novel strategy to enhance neuromorphic computing performance and can be utilized for pattern identification, such as number recognition based on artificial neural networks.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn conclusion, we first demonstrate that regulating the conjugated backbone architecture directly governs side chain\u0026ndash;electrolyte coupling, providing the molecular basis for enhanced ion retention and long-term nonvolatile synaptic behavior. The p(g2T-TT)-based EGOSTs with the longest backbone exhibit the strongest interaction with ions compared to other devices, achieving distinct structural and electrical property changes. DFT and electrical analysis results confirmed that as the backbone lengthens, it exhibits strong adsorption energy between the anion and side chain, leading to long-term non-volatile characteristics over 240 seconds. As an artificial synapse, it achieves enhanced neuromorphic performance such as LTP/D and achieves 94.5% accuracy in ANN-based recognition simulation. These findings suggest that by controlling the adsorption behavior between the side chain and anion through backbone regulation, enhanced synaptic plasticity reinforcement and weight update can be achieved. It also presents new possibilities for design strategies that are limited to independent modulation of side chains and backbones.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eMaterials\u003c/p\u003e\u003cp\u003eAll materials directly use after purchase including: poly[3,3'-bis[2-[2-(2-methoxyethoxy)ethoxy]ethoxy][2,2'-bithiophene]-5,5'-diyl] p(g2T), poly[3,3'-bis[2-[2-(2-methoxyethoxy)ethoxy]ethoxy]-2,2':5',2'\u0026lsquo;-terthiophene-5,5\u0026rsquo;'-diyl] p(g2T -T), and poly[2-(3,3\u0026prime;-bis(2-(2-(2-methoxyethoxy)ethoxy)ethoxy)-[2,2\u0026rsquo;-bithiophen]-5-yl)thieno[3,2-b]thiophene] p(g2T-TT) were purchased from Derthon Co. LTD (China), diethylmethyl (2-methoxyethyl) ammonium bis(trifluoromethylsulfonyl)imide (DEME-TFSI) was purchased from Sigma-Aldrich (USA).\u003c/p\u003e\u003cp\u003eFabrication of EGOST\u003c/p\u003e\u003cp\u003eOur artificial synapses were fabricated on boron-doped p-type Si wafers covered with 300 nm of SiO₂. To ensure uniform film formation, wafers cut into 1.5 \u0026times; 1.5 cm\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e pieces were sonicated in acetone and isopropyl alcohol (IPA) and then washed with DI water and N\u003csub\u003e2\u003c/sub\u003e gas blow. The polymer thin film was spin-coated at 1000 rpm for 60 seconds using a solution prepared at 5 mg mL\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e in chloroform solvent. The electrodes were then patterned using a shadow mask via a thermal evaporator and deposited with 40 nm of gold in a high vacuum environment (\u0026asymp;\u0026thinsp;10\u003csup\u003e\u0026minus;\u0026thinsp;6\u003c/sup\u003e Torr). Finally, as shown in Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e, the ionic liquid was drop-cast between the source/drain electrodes.\u003c/p\u003e\u003cp\u003eMeasurement and Analysis\u003c/p\u003e\u003cp\u003eThe electrical characteristics of the fabricated devices were measured using a Keithley 4200A-SCS (Tektronix, USA) semiconductor parameter analyzer at 10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e A compliance, ambient and darkroom conditions. The UV absorption spectra of each thin film were obtained using a Lambda 465 (PerkinElmer, USA) UV/vis spectrophotometer. GIWAXS was conducted at the Pohang Accelerator Laboratory (PAL), South Korea, utilizing synchrotron radiation from the 9A beamline. Time-of-flight secondary-ion mass spectrometry (TOF-SIMS) analysis was performed using M6 (IONTOF GmbH, Germany). Devices with a bias of \u0026minus;\u0026thinsp;2 V applied for 20 seconds were used for GIWAXS and TOF-SIMS analysis dependent on doping.\u003c/p\u003e\u003cp\u003eArtificial Neural Network\u003c/p\u003e\u003cp\u003eThe recognition simulation was performed in a Linux environment based on the open-source NeuroSim_V3.0. The artificial neural network had a multi-layer perceptron structure and learned and inferred MNIST digits. Parameters such as \u003cem\u003eNL\u003c/em\u003e, \u003cem\u003eG\u003c/em\u003e\u003csub\u003emax\u003c/sub\u003e/\u003cem\u003eG\u003c/em\u003e\u003csub\u003emin\u003c/sub\u003e, and variation were considered to reflect the characteristics of EGOSTs.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of this study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by a National Research Foundation of Korea (NRF) grant funded by the Ministry of Science and ICT (MSIT) (No. RS-2022-NR072040) and Korea Planning \u0026amp; Evolution Institute of Industrial Technology (RS-2024-00420537) grant funded by the Ministry of Trade, Industry \u0026amp; Energy (MOTIE).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAuthor and Affiliations\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Chemical and Biomolecular Engineering, Seoul National University of Science and Technology, Seoul 01811, Republic of Korea\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJunho Sung, Byeongjun Jeon, Donghwa Lee, Yoohyeon Jang, Bumjoon Seo \u0026amp; Eunho Lee\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Chemical Engineering, Pohang University of Science and Technology, Pohang 37673, Republic of Korea\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSein Chung \u0026amp; Jiyeong Shin\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDepartment of Chemical Engineering, Kumoh National Institute of Technology, Gumi 39177, Republic of Korea\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMyungjin An\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eContributions\u003c/p\u003e\n\u003cp\u003eJ.S., S.C., and E.L. conceived and designed the project. J.S., D.L., and Y.J. designed the experiments, conducted the experiments. B.J. and B.S. performed the computational analysis of the adsorption behavior and charge distribution. S.C. and J.S.(Shin) investigated the crystallinity of organic thin films. J.S. and B.J. wrote the manuscript. S.C., B.S. and E.L. revised the manuscript. All authors participated in data analysis and discussions.\u003c/p\u003e\n\u003cp\u003eContributions Corresponding authors\u003c/p\u003e\n\u003cp\u003eCorrespondence to Bumjoon Seo or Eunho Lee\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics declarations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCompeting interests\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBeyond von Neumann (2020) Nat Nanotechnol 15:507\u0026ndash;507\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu C et al (2020) Two-dimensional materials for next-generation computing technologies. Nat Nanotechnol 15:545\u0026ndash;557\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSangwan VK, Hersam MC (2020) Neuromorphic nanoelectronic materials. Nat Nanotechnol 15:517\u0026ndash;528\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBae J, Lee SJ (2024) Current Progress in the Application of Artificial Intelligence for Nuclear Power Plant Operation. Korean J Chem Eng 41:2851\u0026ndash;2870\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSebastian A, Le Gallo M, Khaddam-Aljameh R, Eleftheriou E (2020) Memory devices and applications for in-memory computing. Nat Nanotechnol 15:529\u0026ndash;544\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchuman CD et al (2022) Opportunities for neuromorphic computing algorithms and applications. Nat Comput Sci 2:10\u0026ndash;19\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBoosting AI (2025) with neuromorphic computing. Nat. Comput. Sci. 5, 1\u0026ndash;2\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePi S et al (2019) Memristor crossbar arrays with 6-nm half-pitch and 2-nm critical dimension. Nat Nanotechnol 14:35\u0026ndash;39\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLee H-S et al (2020) Dual-Gated MoS2 Memtransistor Crossbar Array. Adv Funct Mater 30:2003683\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFeng X et al (2021) Self-Selective Multi-Terminal Memtransistor Crossbar Array for In-Memory Computing. ACS Nano 15:1764\u0026ndash;1774\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen Z et al (2024) Cross-layer transmission realized by light-emitting memristor for constructing ultra-deep neural network with transfer learning ability. Nat Commun 15:1930\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLee S, Jung HS (2024) Cutting-Edge Advances in Perovskite Photovoltaic Devices and Applications. Korean J Chem Eng 41:3703\u0026ndash;3715\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRivnay J et al (2018) Organic electrochemical transistors. Nat Rev Mater 3:17086\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYao B-W et al (2021) Non-Volatile Electrolyte-Gated Transistors Based on Graphdiyne/MoS2 with Robust Stability for Low-Power Neuromorphic Computing and Logic-In-Memory. Adv Funct Mater 31:2100069\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTorricelli F et al (2021) Electrolyte-gated transistors for enhanced performance bioelectronics. Nat Rev Methods Primer 1:66\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang W et al (2024) Electrolyte-Gated Transistor Array (20 \u0026times; 20) with Low-Programming Interference Based on Coplanar Gate Structure for Unsupervised Learning. Small Sci 4:2300306\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJang H, Bae GY, Kim SH, Sung J, Lee E (2024) Crosslinking-induced anion transport control for enhancing linearity in organic synaptic devices. Mater Horiz 11:4638\u0026ndash;4650\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi J et al (2024) High-Density Artificial Synapse Array Consisting of Homogeneous Electrolyte-Gated Transistors. Adv Sci 11:2305430\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSung J et al (2025) Modulating Alkyl Groups in Copolymer to Control Ion Transport in Electrolyte-Gated Organic Transistors for Neuromorphic Computing. Small Struct 6:2400319\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiang Y et al (2025) Bioinspired Electrolyte-Gated Organic Synaptic Transistors: From Fundamental Requirements to Applications. Nano-Micro Lett 17:198\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJi J et al (2025) Single-transistor organic electrochemical neurons. Nat Commun 16:4334\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang Y et al (2024) Designing organic mixed conductors for electrochemical transistor applications. Nat Rev Mater 9:249\u0026ndash;265\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBardagot O et al (2025) Over Tenfold Increase in Current Amplification Due to Anisotropic Polymer Chain Alignment in Organic Electrochemical Transistors. Adv Mater 37:2420323\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDing B et al (2021) Influence of Backbone Curvature on the Organic Electrochemical Transistor Performance of Glycolated Donor\u0026ndash;Acceptor Conjugated Polymers. Angew Chem Int Ed 60:19679\u0026ndash;19684\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHallani RK et al (2021) Regiochemistry-Driven Organic Electrochemical Transistor Performance Enhancement in Ethylene Glycol-Functionalized Polythiophenes. J Am Chem Soc 143:11007\u0026ndash;11018\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMoser M et al (2021) Polaron Delocalization in Donor\u0026ndash;Acceptor Polymers and its Impact on Organic Electrochemical Transistor Performance. Angew Chem Int Ed 60:7777\u0026ndash;7785\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBardagot O et al (2025) Balancing Electroactive Backbone and Oligo(Ethylene Oxy) Side-Chain Content Improves Stability and Performance of Soluble PEDOT Copolymers in Organic Electrochemical Transistors. Adv Funct Mater 35:2412554\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen Z et al (2025) π-Conjugated Polymers for High-Performance Organic Electrochemical Transistors: Molecular Design Strategies, Applications and Perspectives. Angew Chem Int Ed 64:e202423013\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWu H-Y et al (2023) Stable organic electrochemical neurons based on p-type and n-type ladder polymers. Mater Horiz 10:4213\u0026ndash;4223\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGiovannitti A et al (2020) Energetic Control of Redox-Active Polymers toward Safe Organic Bioelectronic Materials. Adv Mater 32:1908047\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSung M-J et al (2025) Organic Artificial Nerves: Neuromorphic Robotics and Bioelectronics. Chem Rev 125:2625\u0026ndash;2664\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang X et al (2025) Backbone Engineering of Bithiophene Imide Dimer-Based Polymeric Mixed Ionic-Electronic Conductors for High-Performance n-Type Organic Electrochemical Transistors. Small 21:2408716\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang C et al (2025) Polythiophenes for High-Performance N-type Organic Electrochemical Transistors. Adv Funct Mater 35:2419706\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhou H, Ye W, Jiang J, Wang Z (2024) Recent advances on surface modification of non-oxide photocatalysts towards efficient CO2 conversion. Carbon Lett 34:1569\u0026ndash;1591\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePenyan WL, Jasim LS (2024) Preparation and characterization of a novel pH-sensitive pectin-grafted-poly (acrylic acid-co-itaconic acid)/MWCNTs-COOH hydrogel composite for adsorption and controlled release of diclofenac sodium. Carbon Lett 34:1413\u0026ndash;1429\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMoser M et al (2020) Side Chain Redistribution as a Strategy to Boost Organic Electrochemical Transistor Performance and Stability. Adv Mater 32:2002748\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMoser M et al (2020) Ethylene Glycol-Based Side Chain Length Engineering in Polythiophenes and its Impact on Organic Electrochemical Transistor Performance. Chem Mater 32:6618\u0026ndash;6628\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJo I-Y et al (2024) High-Performance Organic Electrochemical Transistors Achieved by Optimizing Structural and Energetic Ordering of Diketopyrrolopyrrole-Based Polymers. Adv Mater 36:2307402\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTseng H-S et al (2024) Strategically tailoring ethylene glycol side chains with bridged-carbonyl ester in polythiophene-based organic electrochemical transistors for bioelectronics. Chem Eng J 486:150371\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSung M-J et al (2024) Overcoming the Trade-off Between Efficient Electrochemical Doping and High State Retention in Electrolyte-Gated Organic Synaptic Transistors. Adv Funct Mater 34:2312546\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSung J et al (2025) Improving ion uptake in artificial synapses through facilitated diffusion mechanisms. Mater Horiz 12:5225\u0026ndash;5235\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKwon O et al (2024) Optimizing Molecular Weight of Polyethylene Glycol as an Additive for Stabilizing Zn Metal Anode in Aqueous Electrolyte. Korean J Chem Eng 41:539\u0026ndash;544\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang P et al (2024) Theoretically and Experimentally Exploring the Inhibition Effect of Imidazole Sulfonic Acid Derivatives with Multiple Adsorption Sites in 1.0 mol/L HCl. Korean J Chem Eng 41:2117\u0026ndash;2132\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAn SH, Kim KY, Chung CW, Lee JU (2024) Development of cement nanocomposites reinforced by carbon nanotube dispersion using superplasticizers. Carbon Lett 34:1481\u0026ndash;1494\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLondon M, Schreibman A, H\u0026auml;usser M, Larkum ME, Segev I (2002) The information efficacy of a synapse. Nat Neurosci 5:332\u0026ndash;340\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTang A-H et al (2016) A trans-synaptic nanocolumn aligns neurotransmitter release to receptors. Nature 536:210\u0026ndash;214\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSung J et al (2025) Co-Stimuli-Driven 2D WSe2 Optoelectronic Synapses for Neuromorphic Computing. Small 21:2504024\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTan E et al (2022) The Role of Long-Alkyl-Group Spacers in Glycolated Copolymers for High-Performance Organic Electrochemical Transistors. Adv Mater 34:2202574\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOh S-J et al (2024) Non-volatile and Stretchable Polyvinyl Chloride-Based Solid-State Electrolyte for Capacitive Energy Storage. Korean J Chem Eng 41:1861\u0026ndash;1869\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZeng X et al An Interlayer Strategy for Low-Voltage Thin-Film Organic Electrochemical Transistors. Small Methods n/a, 2500322\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLee CM, Kim Y, Kim W, Lee E, Lee EK (2025) High-Performance Synaptic Devices Based on Cross-linked Organic Electrochemical Transistors with Dual Ion Gel. Adv Funct Mater 35:2417539\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChoi Y et al (2023) Physically defined long-term and short-term synapses for the development of reconfigurable analog-type operators capable of performing health care tasks. Sci Adv 9:eadg5946\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang H et al (2018) A Ferroelectric/Electrochemical Modulated Organic Synapse for Ultraflexible, Artificial Visual-Perception System. Adv Mater 30:1803961\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKwon SM et al (2019) Environment-Adaptable Artificial Visual Perception Behaviors Using a Light-Adjustable Optoelectronic Neuromorphic Device Array. Adv Mater 31:1906433\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePark SM, Won Y, Oh JH, Lee EK (2025) Enhancement of Synaptic Behavior in Organic Electrochemical Transistors via the Introduction of Layer-by-Layer Grown Metal-Organic Framework. Adv Mater Technol 10:2401316\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSeo D-G et al (2019) Versatile neuromorphic electronics by modulating synaptic decay of single organic synaptic transistor: From artificial neural networks to neuro-prosthetics. Nano Energy 65:104035\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLee D et al (2025) Inter-Ion Mutual Repulsion Control for Nonvolatile Artificial Synapse. Adv Funct Mater 35:2412012\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRoe DG et al (2024) Reconfigurable Logic Gates Capable of Device-Level Parallel Processing Through Multi-Input Synaptic Device. Adv Funct Mater 34:2312988\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLuo X et al (2022) Circularly Recyclable Polymers Featuring Topochemically Weakened Carbon\u0026ndash;Carbon Bonds. J Am Chem Soc 144:16588\u0026ndash;16597\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChoi W et al (2023) Unravelling Disorder Effects on Thermoelectric Properties of Semicrystalline Polymers in a Wide Range of Doping Levels. Small Methods 7:2201145\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKim J et al (2023) Remarkable conductivity enhancement in P-doped polythiophenes via rational engineering of polymer-dopant interactions. Mater Today Adv 18:100360\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKline RJ et al (2007) Critical Role of Side-Chain Attachment Density on the Order and Device Performance of Polythiophenes. Macromolecules 40:7960\u0026ndash;7965\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen P-Y, Peng X, Yu S, NeuroSim+ (2017) An integrated device-to-algorithm framework for benchmarking synaptic devices and array architectures. in 2017 IEEE International Electron Devices Meeting (IEDM) 6.1.1\u0026ndash;6.1.4 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/IEDM.2017.8268337\u003c/span\u003e\u003cspan address=\"10.1109/IEDM.2017.8268337\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen P-Y, Peng X, Yu S, NeuroSim: (2018) A Circuit-Level Macro Model for Benchmarking Neuro-Inspired Architectures in Online Learning. IEEE Trans Comput -Aided Des Integr Circuits Syst 37:3067\u0026ndash;3080\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7438138/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7438138/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAchieving stable and nonvolatile synaptic plasticity remains a central challenge for organic neuromorphic devices. While previous efforts have independently focused on backbone or side chain modifications, the fundamental role of conjugated backbone design in directing side chain\u0026ndash;electrolyte coupling has remained elusive. Here, we demonstrate that modulation of thiophene units in the backbone governs the spatial arrangement and ionic accessibility of glycol side chains, thereby enabling strong anion adsorption and long-term retention. Electrolyte-gated organic synaptic transistors (EGOSTs) with extended backbones exhibit pronounced structural reorganization, suppressed ion back-diffusion, and stable nonvolatile characteristics. Backbone-directed side chain\u0026ndash;anion coupling was identified as the key mechanism driving enhanced charge transport and persistent doping. As artificial synapses, the device realizes robust neuromorphic functions including paired-pulse facilitation, long-term potentiation/depression, and achieves 94.5% accuracy in artificial neural network simulations. This work establishes conjugated backbone regulation as a facile strategy to control side chain\u0026ndash;electrolyte interactions, offering new design principles for nonvolatile synaptic devices and advancing the development of reliable organic neuromorphic computing.\u003c/p\u003e","manuscriptTitle":"Conjugated Backbone-Directed Side Chain-Electrolyte Coupling toward Nonvolatile Artificial Synapse","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-26 07:58:12","doi":"10.21203/rs.3.rs-7438138/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"dc588606-01f9-4edc-a3e7-2eb314d04eb0","owner":[],"postedDate":"August 26th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":53595066,"name":"Physical sciences/Materials science/Materials for devices/Electronic devices"},{"id":53595067,"name":"Physical sciences/Materials science/Nanoscale materials/Electronic properties and materials"}],"tags":[],"updatedAt":"2025-09-05T10:11:11+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-26 07:58:12","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7438138","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7438138","identity":"rs-7438138","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}