In‐Depth Physical Mechanism Analysis of Polymer Artificial Optoelectronic Synapse with High Endurance and Applications of Visual System and Operant Conditioning

Being capable of dealing with both electrical signals and light, artificial optoelectronic synapses are considered to be an important cornerstone of neuromorphic computing. Here, an artificial optoelectronic synapse is reported through a simple solution process using organic poly(3‐hexylthiophene) (P3HT) and a remarkable analog switching characteristic similar to synaptic behavior is observed. The endurance and data retention capability of the P3HT‐based optoelectronic synapse exhibit stable characteristics up to 5000 consecutive cycles and 104 s. Through in‐depth physical mechanism analysis, it is confirmed that the analog switching characteristics of the device are mainly caused by a tunneling mechanism and space charge limited conduction. Furthermore, characterizations such as X‐ray photoelectron spectroscopy and atomic layer deposition prove that the memristive properties of device can be attributed to ion migration. More importantly, the device can co‐modulate the optoelectronic signal and successfully implement a photo‐triggered multi‐signal mode response. Based on this, a 3 × 3 synapse array is developed to demonstrate the potential application of the proposed P3HT‐based optoelectronic synapse in constructing an artificial visual system. Finally, operant conditioning is successfully simulated in the synaptic device. This work provides a reference for the construction of optoelectronic synapses in the neuromorphic visual system.


Introduction
[3][4][5] With the rapid development of artificial intelligence, the visual system of artificial intelligence devices is more and more demanding.Current photodetectors are usually difficult to combine extremely high sensitivity, wide spectra response and high-performance information processing and storage capabilities. [6,7]emristor as the fourth fundamental circuit element offers a great opportunity for realizing high-performance neuromorphic computing.0] As with neuromorphic computing, neuromorphic visual system needs to construct artificial neural network first, and synapses are the key of neural network.Optoelectronic synapses function as both artificial synapses and photosensors, processing detected optical and electrical signals and track the history including signal intensity, the number, duration and frequency of exposure, [11,12] which are key to future advanced visual systems.
Until now, a large variety of materials for optoelectronic synapses have been proposed.[19][20] In addition, most reported optoelectronic synapses often only study the simulation of synaptic function and visual system by optoelectronic signal modulation. [5,21,22]][25] In particular, organic polymer optoelectronic synapses show limited endurance and poor data retention, hindering the practical use of organic optoelectronic synapses.[28] Therefore, it is necessary to design and manufacture organic polymeric artificial optoelectronic synapses with high endurance, and clarify the conduction mechanism of the memristive devices.
Here, we propose and demonstrate a P3HT-based optoelectronic synapse that can exhibit both electrical and light-induced synaptic behaviors, greatly enriching the regulation of synaptic plasticity.Thanks to the good solubility of polymer, the organic layer is prepared by solution spin coating, which is simple and fast.The endurance and data retention capability of P3HT-based optoelectronic synapse exhibit stable characteristics up to 5000 consecutive cycles and 10 4 s.The high endurance expands the practical application of organic optoelectronic synapses.Then, through detailed data fitting and in-depth physical mechanism analysis, it is confirmed that the analog switching characteristics of the device are mainly caused by tunnel mechanism and space charge limited conduction (SCLC).X-ray photoelectron spectroscopy (XPS) and atomic layer deposition (ALD) prove that the memristive properties of device can be attributed to ion migration.Not only that, we successfully tested the synaptic properties of the device, such as paired-pulse facilitation (PPF), post-tetanic potentiation (PTP), short-term potentiation (STP) to long-term transition potentiation (LTP) and spike time-dependent plasticity (STDP), etc.In addition to the basic electrical synaptic behaviors, optoelectronic synaptic devices could also enable light-induced synaptic functions and have a good response to visible light, especially green light, which is very similar to human vision.Further, we constructed a 3×3 array to demonstrate the light response and memory of a simple artificial visual system.Finally, operant conditioning experiments in behaviorist learning theory are simulated due to the multi-signal response of devices.This research provides a reference for the construction of efficient optoelectronic synapses with high endurance, provides a deep theoretical understanding for the physical mechanism of the memristive effect and broadens the way for the development of neuromorphic visual systems.

Results and Discussion
Poly(3-hexylthiophene) (P3HT) has good solubility and is suitable for film processing and large area preparation.We prepared a 3 mg mL −1 P3HT solution by dissolving P3HT in chloroform.Then, the P3HT film was spin-coated onto ITO-coated glass (Figure 1a).At last, Al electrodes were deposited on the P3HT film by thermal evaporation to obtain the ITO/P3HT/Al synaptic device.Figure 1b illustrates the schematic diagram of the device and the chemical structure of P3HT.Scanning electron microscopy (SEM) was employed to take a cross-sectional view of the device (Figure 1c).The P3HT layer and the top/bottom electrodes have obvious delamination, and the device structure was clearly visible.In addition, light response is the unique advantage of optoelectronic devices.To investigate the optoelectrical performance of the device, the UV-vis absorption spectra of the P3HT film are given in Figure 1d.It can be seen that the film exhibited significant absorption in the range from 450 to 650 nm, among which the absorption peak of 550 nm green light is the most intense.This will provide information for P3HT-based optoelectronic synapse simulation to select light signals.
Potentiation and depression are the most important properties of artificial synapses, that is, the increase/decrease of conductance in devices corresponding to synaptic currents generated by excitatory/inhibitory signals in biological synapses. [9]The device can simultaneously modulate the conductance state with electrical signals and light.The current-voltage (I-V) characteristics of the P3HT device under 30 cycles of positive (0 → +5→ 0 V) and negative (0 → −5→ 0 V) voltage sweeps are shown in Figure 1e.The scanning current of each path showed clear pinched hysteresis loops, that is, the fingerprint characteristics of a memristive device, [8,[29][30][31] and the conductance of our P3HT device continuously and monotonically increases/decreases under consecutive positive/negative voltage sweeps (Figure 1f).These results indicate that our device has the property of continuous tunable conductance at specific voltages, and this wide range of tunable conductance will be an important basis for achieving multiple synaptic plasticities.The device conforms to the characteristics of an analog memristor and can be used as a potential application of artificial synapses.In an optoelectronic synapse, the change in device conductance is regulated by the optical signal, and the conductance is modulated by the difference in intensity, wavelength, and frequency of the input optical signal. [32,33]In Figure 1g, we applied the device a series of reading voltage pulses (+1 V), which is insufficient to cause a change in the conductance of the device, so the current of the device is maintained at a weak level in the dark environment.After that, the device was illuminated with a light intensity of 5.28 mW cm −2 for a duration of 40 s, and the current (synaptic weight) could be enhanced from 5 to 13 nA.By turning off the light, the current exhibits a gradual decay process instead of recovering rapidly to the initial conductance.This can be explained by the generation and recombination of nonequilibrium carriers under illumination (Figure S1, Supporting Information).Since the work function of ITO and Al electrodes is close to the HOMO level of P3HT, the holes generated by excitons under illumination are easy to migrate under an electric field, resulting in an increase in illumination current.By turning off the light, the photogenerated carriers gradually recombine and eventually return the current to its initial level.Figure 1h shows the current changes of the device under dark condition and 5.28 mW cm −2 light illumination by applying ten consecutive positive voltage pulses (+5 V, 50 ms).It can be clearly seen from the figure that the current level of the device applying both electrical pulse and optical signals is greater than that of the device applying only the electrical pulse.This indicates that illumination can amplify the change in the device conductance under the electrical signal.We have also tested the endurance and retention characteristics of the device under on/off light, as shown in Figure 1i,j.It can be clearly seen that our device has good endurance to withstand at least 5000 cycles and retention characteristics within 10 000 s under light stimulus and dark condition.Our P3HT-based optoelectronic synapse has high endurance and long retention characteristics as compared to previous works reported on both organic and inorganic optoelectronic synaptic devices (Table S1, Supporting Information).The above results indicate that our synaptic devices can achieve synaptic plasticity by changing the conductance level of electrical and optical signals, respectively, and the two signals can be combined to modulate the device.This will diversify the input signal of the device, increase the modulation means of the synaptic device, and greatly enrich the information exchange with the outside world, and the high endurance and long retention characteristics of the device broadens its practical application.
The synapses are the connection points of two adjacent neurons. [34]The presynaptic neuron causes membrane potential changes in the postsynaptic membrane by releasing neurotransmitters, generating a postsynaptic current (PSC) that excites or inhibits the postsynaptic neuron, thus allowing information to be transmitted across a large-scale neural network (Figure 2a). [35]Notably, synaptic connection strength increases or decreases depending on its activity and maintains this change over time, which is called synaptic plasticity. [36,37]Short-term plasticity (STP) and long-term plasticity (LTP) can be classified according to the length of retention capability. [16,38]Paired-pulse facilitation (PPF) is an important form in STP can be demonstrated by applying two successive presynaptic spikes. [39,40]When two successive stimuli reach the presynaptic neuron, the second stimulus will cause a larger postsynaptic response.Obviously, the PPF effect depends directly on the time interval (Δt) between two stimuli, and a smaller Δt leads to a larger PPF.In our artificial synapse, we successfully simulated PPF behavior in our artificial synapse (Figure 2b).The PPF can be depicted via the evolution of a PPF index (defined as (A 2 − A 1 ) /A 2 × 100%), obviously increases the time interval, the increment decreases and finally approaches zero.Subsequently, we input electrical pulses (+5 V, 50 ms) into the device at different frequencies (0.1-10 Hz), and record the corresponding current (Figure 2c).The higher the input frequency, the denser the signal, and the faster the postsynaptic current increases and ultimately causes the synaptic to change for a long time, known as long-term plasticity (LTP).Furthermore, extending the pulse duration and increasing the number of pulse inputs can also lead to the STP-to-LTP transition.We applied a series of fixed simulation pulses (+5 V) to the device with different pulse widths (20, 50, 100, and 200 ms), recorded the current levels, as shown in Figure S2, Supporting Information.It can be seen that the device reaches higher current levels and decreases more slowly as the pulse width increases.This means that synapses produce stronger membrane potential changes that last longer, enabling the transition from short-term memory (STM) to long-term memory (LTM).In our artificial synaptic device, we applied different numbers of identical voltage pulses (5, 10, 15, and 20) with a fixed amplitude of +5 V, a width of 50 ms, and then recorded the EPSC level of the device at a read voltage of +1 V. Figure 2d shows the synaptic weight plotted with respect to retention time after different numbers of presynaptic spikes.The synaptic weight exhibits a rapid decay at the beginning, but there is a pronounced slowing down of the final decay as the number of presynaptic spikes increases.The STP-to-LTP transition has also been demonstrated.In addition, spike time-dependent plasticity (STDP) is an important neuromorphic activity for learning and memorizing which is determined by the relative timing of presynaptic and postsynaptic spikes.By applied two correlated spikes with different pulse intervals ranging from 10 to 100 ms, respectively, to emulate the STDP function in the P3HT-based artificial synapse.The variation of synaptic weight change [ΔW = (W 1 − W 0 ) / W 0 ] as a function of Δt is shown in Figure 2e,f, where W 1 and W 0 represent the post-and pre-spike conductance, respectively.Two STDP rules of asymmetric Hebbian and asymmetric anti-Hebbian were implemented, and ΔW was increased when |Δt| was small.
After that, we applied ten consecutive positive voltage pulses (+5 V and +6 V) into the device, respectively, and added optical signals to the electrical pulses as mixed signals to record the response current of the device, respectively (Figure 2g).It is not difficult to see that the current generated by adding the optical signal is greater than the current generated by the individual electrical signal, and the increment is greater than +5 V at +6 V.This means that the addition of optical signals can assist electrical impulses to achieve greater changes in synaptic plasticity.Then we investigated the unique role of optical signals in synaptic plasticity.Two consecutive light pulses with 1 s duration and 5.28 mW cm −2 intensity are applied to the device, as depicted in Figure 2h.The light-induced PPF exhibits similar characteristics with the electrical-induced PPF, in the increase of pulse interval time, the PPF index exponentially decays, which is consistent with the learning rules of biological synapse. [41]imilar to PPF, post-tetanic potentiation (PTP) refers to the phenomenon that the synaptic weight increases in a certain period of time under repeated high-frequency stimulation.The simulation of PTP with/without light is shown in Figure S3, Supporting Information.The current response of the device under various light intensity levels (1.98, 3.56, and 5.28 mW cm −2 ) as depicted in Figure 2i demonstrates the intensity-dependent current response of the device.Ten consecutive light pulses with a pulse duration of 1 s and interval of 1 s at the reading voltage of +1 V were applied.Noting that a higher light intensity leads to a larger current response and a slower decay trend.It can be seen that the current decay trend is also dependent on the light intensity, similar to the postsynaptic current caused by different stimulus intensity.Furthermore, benefiting from the broad absorption of P3HT in visible light region, the current response of the device under different wavelengths of light as illustrated in Figure 2j.Red (650 nm), green (550 nm), and blue (450 nm) light were input into the device with +5 V electrical pulses, respectively.The results showed that EPSC of the device under green light is the maximum, which is consistent with the absorption spectrum of P3HT.The above results indicate that our artificial synapse can realize the co-modulation of op-tical and electrical signals, enrich the modulation means of the synapse.
We carried out a series of experiments to elucidate the switching mechanism.First, we measured the area dependence I-V characteristics for P3HT device.Figure S4, Supporting Information shows the cell area dependence of currents in the devices with different areas ranging from 100 × 100 to 100 × 1000 μm 2 .We note that the current increases with increasing device area, indicating a homogeneous change in device conductance and hence ruling out a filamentary switching mechanism. [42,43]Next, we explore the cause of the change of conductance of the device under electric signal.Since we replaced the top electrode with Ag (device structure ITO/P3HT/Ag), the device showed a single resistance and no change in resistance value (Figure S5, Supporting Information), we speculated that the Al electrode might be the key.In the air, Al can spontaneously form thin natural oxides of 2 to 4 nm, and when we deposited Al, residual oxygen in the chamber could also react with it, so the Al electrode inevitably contained AlO x .It has been reported that the migration of oxygen ions in AlO x under electric field will lead to the change of device resistance value, thus realizing artificial synapse. [24,44]To test this hypothesis, we used atomic deposition equipment (ALD) to deposit a 3.6 nm AlO x layer on P3HT film, followed by deposition of Ag electrode (Figure 3a).Sure enough, the I-V curve of the device with this structure is enhanced in the positive direction and decreased in the negative direction, reproducing the change of conductance under the electric field (Figure S6, Supporting Information).Figure 3b shows the current variation when the device was programmed by positive pulses (+5 V) followed by negative pulses (−5 V).We find that the current is gradually strengthened/weakened with increasing number of positive/negative pulses.Not only that, to reveal the origin of resistive switching behavior, X-ray photoelectron spectroscopy (XPS) analysis of bottom ITO electrode was further performed.For metal oxides, certain conditions (such as high temperatures) cause oxygen to detach from the lattice and form oxygen vacancies.Oxygen vacancy is a common defect in metal oxides, and ITO is no exception.The O 1s peak of ITO electrode in initial state is fitted and divided into two peaks (Figure 3c).The peak at 530.13 eV was lattice oxygen (35.50%), and the peak at 531.21 eV was chemically adsorption of oxygen (64.50%) at oxygen vacancy position.This method of peak separation has been reported. [45,46]The XPS spectra of O 1s peak of ITO electrode after voltage stimulation (+5 V) were obtained by the same method (Figure 3d).It can be found that the peak content of oxygen vacancy increased to 66.05%.After voltage stimulation (−5 V), the proportion of oxygen vacancy peak content decreased to 65.03% (Figure 3e).As a comparison, XPS spectra of In 3d and Sn 3d were also analyzed, and the results showed that the position and content of the peaks were almost unchanged (Figure S7, Supporting Information).Thus, we obtain the mechanism model of device conductance change (Figure 3f).Under the forward bias, oxygen ions in Al electrode migrate to ITO electrode and are captured by oxygen vacancy.This process reduces the AlO x in the Al electrode and thus increases the conductance.The reduction of conductance under negative bias can be interpreted as the reverse process of this mechanism.We also studied the role of compliance currents (I cc ) during I-V sweeps.Figure S8, Supporting Information, shows the I-V curves of the device under different I cc .It can be seen that P3HT device requires I cc > 0.04 mA to generate an electric field that is high enough to drive the ion migration in the device, and therefore increasing the device conductance.
Last, due to the spatial presence of variable distributions of ionic species, the I−V characteristic curve of the device was further fitted.The experimental I−V data were characterized using the Simmons equation, which is widely used to calculate the barrier parameters in the tunneling layer of a device, and the I−V relationship can be expressed through the tunneling barrier.At a very low bias, the barrier is rectangular, and can be simplified to a direct tunneling, as the following equation: [47] I ∝ V exp where d is the barrier width, m* is the electron effective mass, ϕ B is the barrier height, and ℏ is the reduced Planck's constant.When the applied bias voltage is greater than the barrier height ϕ B , the barrier transitions from trapezoidal to triangular, and the I−V relationship can be described as the Fowler−Nordheim tunneling (Equation ( 2)) [48] According to Equations ( 1) and ( 2), the conduction mechanism of the device was investigated by fitting the I-V curves in the HRS (blue scatter plot) and LRS states (red scatter plot), and the I−V characteristic curve can be divided into two parts according to the fitting results, in which the transition voltage (V trans ) was used as the inflection point, as shown in Figure 3g.The relationship with a negative slope in Figure 3g confirms that the conduction mechanism is dominated by FNT at the relatively large-voltage region.The plot with positive and negative slope are clearly distinguished, with the increase of the bias voltage, the shape of the tunneling barrier is gradually changed from rectangle to trapezoid.When the bias voltage exceeds the V trans , the barrier becomes triangular, and an obvious transition from direct tunneling (DT) to Fowler−Nordheim tunneling (FNT) occurs.Furthermore, we plotted the I-V curves in the HRS (blue scatter plot) and LRS states (red scatter plot) in dual-logarithmic scales by fitting the hysteresis loops using the space charge limited current (SCLC) model described by the equation: I = V n , where  is a constant, V is the applied voltage, n is an exponential parameter associated with the trapped charge state.When the operating voltage gradually increases to +5 V, the current increases exponentially as I ∝ V n , indicating the conduction mechanism in the device follows the SCLC theory.As seen in Figure 3h, in the HRS state, the slope increases from 1.60 to 5.34; in the LRS state, the slope increases from 1.73 to 4.25; this is attributed to the increased free charge carrier concentration and charges are gradually filled from shallow to deep traps with the gradually increasing of the corresponding n.The device conductivity gradually increases as the concentration of trapped carriers increases after their release from applying voltage, resulting in a gradually enhanced response current that depends on the electrical operation history.Such a behavior reflects the typical response characteristics of our memristor.In addition to the above, we further plotted the I-V curves in the HRS (blue scatter plot) and LRS states (red scatter plot) with the proposed conduction mechanism models, including thermionic emission (ln(I) ∝ V 1/2 ) and Poole-Frenkel emission (ln (I/V) ∝ V 1/2 ). [47]Figure S9, Supporting Information, is the fitting result with thermionic emission.For a standard thermionic emission, the plot of ln(I) versus V 1/2 should be linear.And Figure S10, Supporting Information, is the fitting result with Poole-Frenkel emission.For the Poole-Frenkel emission, a plot of ln(I/V) versus V 1/2 should be linear.But in Figures S9 and S10, Supporting Information, nonlinear plot can be clearly observed as shown in the circle region, the fitted curve results with low goodness-of-fit, especially in the LRS state, which does not conform to the thermionic emission and Poole-Frenkel emission conduction mechanisms.In summary, the HRS and LRS transport property in P3HT device should be a joint effect of several conduction mechanisms, such as tunneling and SCLC mechanism.
Vision is the most important human sense (Figure 4a).Nine artificial synapses devices are created and arranged to form a 3 × 3 array to demonstrate the potential for application in artificial visual systems.First, we input red (650 nm), green (550 nm), and blue (450 nm) primary colors illumination into the corresponding devices of 3 × 3 array (light intensity of 5.28 mW cm −2 , duration time 5 s), and monitor the PSC level by a series of reading voltage pulses (+1 V).The color in the array represents the varieties of PSC.The PSC induced by green illumination was significantly larger than that of red and blue illumination, and the PSC retention of green illumination was also longer in the subsequent attenuation process (Figure 4b).Our array has good absorption and recognition of visible light, almost no obvious response to ultraviolet (365 nm) illumination and attenuation is faster (Figure S11, Supporting Information).This suggests that our artificial synapses have a certain filtering effect, automatically ignoring harmful ultraviolet light in the environment and being sensitive only to visible light, which is particularly important for the visual system.Then we further simulate the effect of light input frequency on the visual system.Take green (550 nm) illumination as input, we used the identical photonic pulses with various frequencies at 1, 0.5, and 0.2 Hz to investigate the device, respectively, and also monitor the varies of PSC (Figure 4c).It was found that higher input frequency resulted in higher levels and longer retention time of PSC, which is consistent with high frequency stimulation causing synaptic long-term memory.We also simulated the learning-forgetting-relearning processes of visual memory (Figure 4d).When the device is illuminated for 2 s as a learning signal, the visual memory is almost forgotten and not retained at all after 10 s.When the same signal is applied again, the device will achieve a stronger PSC than the first time, forming a stronger visual memory.The effect is also evident over longer time spans (Figure S12, Supporting Information).These results indicate that our artificial synapse has a good ability to recognize and filter different optical signals, and can basically realize the function of artificial visual system, which has great application potential.
In addition to simulating synaptic behavior, our artificial synapse also simulates operant conditioning.Operant conditioning, also known as instrumental conditioning, is used to distinguish "elicited response" and "spontaneous response." [49]Simply put, in operant conditioning behaviors are modified by the effect they produce (i.e., reward or punishment). [50]We have recreated this process using artificial synapses.In positive reinforcement, a response or behavior is strengthened by rewards, leading to the repetition of desired behavior. [51]The reward is a reinforcing stimulus.Specifically, a hungry rat was placed in a box.The box contained a lever on the side, and as the rat moved about the box, it would accidentally knock the lever.Immediately it did so a food pellet would drop into a container next to the lever.The rats quickly learned to go straight to the lever after a few times of being put in the box.The consequence of receiving food if they pressed the lever ensured that they would repeat the action again and again.After a lot of experimentation, the rat spontaneously learned to press lever.This behavior is spontaneous learning in rats.Operant behavior is said to be "voluntary."In our device, we apply +1 V electrical pulse to monitor the current level, which is small and represents the rat's "exploration" process.The +5 V electrical pulse represented that the rat accidentally pressed the lever to obtain food.At this point, the current level increased under the reading voltage.From the results, we set the 20 nA current level as a threshold above which the rat would press the lever to establish an association with the food.After repeated "learning," the rat learned to press the lever spontaneously, and the electric current greater than 20 nA, above the threshold.The subsequent current decay represents the spontaneous elimination of this connection (Figure S13a, Supporting Information).Thanks to the optical response characteristics of our device, the reinforcement learning process caused by the reinforcement stimulus is realized.When the +5 V electrical pulse and optical signal were fed into the device, the "learning" process was significantly accelerated, increasing the frequency of that behavior and the en-hanced memory was formed, and the memory retention time was longer (Figure S13b, Supporting Information).Apparently, the food was a reinforcer in our experiment, increasing the probability that the rat would press the lever.Therefore, in our experiment, the addition of optical signal can increase the frequency of the rat pressing the lever to obtain food, resulting in positive reinforcement.In another experiment, Skinner put a rat in his Skinner box and then subject it to an unpleasant electric current which caused it some discomfort.As the rat moved about the box it would accidentally knock the lever.Immediately it did so the electric current would be switched off.The rats quickly learned to go straight to the lever after a few times of being put in the box.The consequence of escaping the electric current ensured that they would repeat the action again and again.After a lot of experimentation, the rat was able to associate pressing the lever with stopping the shock.In our device, continuous input of +5 and −5 V electrical pulse simulates electric shock.The −5 V electrical pulse was then withdrawn and only the +5 V electrical pulse was retained, at which point the current increased and then exceeded the threshold, indicating that the rat had established a connection (Figure S13c, Supporting Information).Negative reinforcement (also known as escape) occurs when a behavior (response) is followed by the removal of an aversive stimulus, thereby increasing the original behavior's frequency.In this case, the shock is a negative reinforcement, and the point is that canceling it increases the probability of a reaction happening.We successfully simulated Skinner's operant conditioning experiment and expanded the application of synaptic equipment in learning theory.

Conclusion
In conclusion, we report an optoelectronic synapse based on P3HT, which is prepared by solution spin-coating.P3HT-based optoelectronic synapse has excellent stability, with endurance and data retention capability characteristics up to 5000 consecutive cycles and 10 4 s.The high endurance expands the practical application of organic optoelectronic synapses.Then, the switching mechanism of P3HT was systematically studied.Through detailed physical mechanism analysis and characterization methods such as XPS, it was confirmed that the analog switching characteristics of the device were mainly caused by tunnel mechanism and SCLC and induced by oxygen ions migration.This provides an innovative approach for the study of the conduction mechanism.Moreover, we have successfully tested the synaptic properties of the devices, such as PPF, PTP, STP-LTP, and STDP, using electrical pulses and optical signals.Because the absorption spectrum of P3HT is in the visible light range, synaptic devices have a good response to visible light, especially green light.In addition, a 3 × 3 array is constructed to demonstrate the optical response and memory of a simple artificial visual system.Finally, operant conditioning experiments in behaviorist learning theory are simulated due to the multisignal response of devices.This study provides a reference for the construction of efficient optoelectronic synapses with high endurance, provides a deep understanding for the physical mechanism of the memristive effect, and broadens the way for the construction of optoelectronic synapses in the neuromorphic visual system.

Experimental Section
Device Fabrications: The P3HT memristor was fabricated in a configuration of ITO/P3HT/Al, in which a 42-nm-thick P3HT film by solution spin coating.An 8 × 8 crossbar array with a total of 64 P3HT memristors was fabricated.Glass substrates with ITO strip electrodes were cleaned by acetone, ethanol, and deionized water for 10 min each, and then baked in an oven at 120 °C for 30 min, followed by a UV-ozone treatment for 10 min.Poly(3-hexylthiophene) (P3HT) was purchased from Sigma-Aldirch Co., Ltd., and chloroform was used as the solvent to prepare a solution of 3 mg mL −1 .The prepared solution was deposited at 500 rpm for 9 s + 3000 rpm for 30 s onto an ITO-coated glass substrate using a spin coating method.The substrate was then annealed in an oven at 60 °C for 30 min.Al top electrodes were finally thermally evaporated onto the P3HT layer through patterned shadow masks (strip type: 100, 250, 500, and 1000 μm, in vacuum (≈10 −4 Pa).The thickness of each evaporated layer was monitored in real time in a vacuum chamber by a quartz crystal vibrometer.The ITO/P3HT/Ag, devices were fabricated via the same procedure.3.6 nm ALD-Al 2 O 3 layer was deposited at 150 °C using Al(CH 3 ) 3 and H 2 O as precursors to fabricated ITO/P3HT/AlO x (3.6 nm) / Ag device.
Optoelectronic Measurement: All optoelectronic properties and synaptic functions of the device were performed in an ambient air using a probe station and a Keithley 4200 s semiconductor parameter analyzer system equipped with programming test software by the authors.The probe station was equipped with an LED light source with a constant wavelength range (410-800 nm) and an OPS light source with adjustable wavelength and brightness.The voltage signals designed for specific learning rules were applied to the ITO electrode, and the Al electrode were grounded.The test process was carried out in air (test temperature 25 °C, humidity 40%).
Material Characterizations: The cross-sectional view of the memristor was characterized by field-emission scanning electron microscopy (Hitachi S4800-SEM).Optical absorption spectra of the active layer were collected by UV−vis spectrophotometer (PerkinElmer Lambda35).XPS measurements were performed using X-ray photoelectron spectroscopy (Kratos AXIS Supra) with monochromated Al K- source.The XPS peaks were fitted using a convolution of Gaussian Lorentzian peaks for each component.The background signal of the peaks was removed using a Shirley background model.All the binding energies were calibrated with reference to the C 1s peak at 284.8 eV.

Figure 1 .
Figure 1.Schematics of a) P3HT film preparation process and b) two-terminal memristive device in ITO/P3HT/Al and the configuration molecular structure of P3HT.c) Cross-sectional SEM image of P3HT memristive device.Scale bar: 100 nm.d) The UV-vis absorption spectrum of the P3HT film.e) I-V characteristics of the P3HT memristive device respond to positive (+5 V, left) and negative (−5 V, right) voltage sweeps.f) Gradual current change of the P3HT memristive device with a series voltage pulses (+5 V for potentiation, −5 V for depression).g) The light-potentiation process with a single light pulse (intensity: 5.28 mW cm −2 , duration: 40 s), and electrical erase process with a voltage pulse (amplitude: +1 V) of the P3HT memristive device.h) Gradual current change of the P3HT memristive device under dark condition and 5.28 mW cm −2 light illumination by applying ten consecutive positive voltage pulses (+5 V, 50 ms).i) Endurance of the P3HT memristive device under 10 000 continuous stimuli pulses (+5 V/−5 V, 100 ms) withstand 5000 cycles under light stimulus and dark condition.j) Current variation of P3HT memristive device under continuous stimuli of +5 V/−5 V pulses within in 10 000 s under light stimulus and dark condition.

Figure 2 .
Figure 2. a) Schematic illustration of neuron and synapse.b) PPF index (defined as (A 2 − A 1 ) / A 1 × 100%) versus pulse interval by applying a paired pulse.c) Current responses to ten identical stimulation pulses (+5 V, 50 ms) at different frequencies (0.1-10 Hz).d) STP-LTP transition.LTP were triggered by applying consecutive spikes (N = 5 -20, V = +5 V, ΔT = 50 ms), where ΔT denotes spike duration.The retention data recorded at a read voltage of +1 V. STDP implementation in the P3HT memristor-based artificial synapse.Asymmetric STDP of e) Hebbian learning rule and f) anti-Hebbian learning rule.The shape of presynaptic and postsynaptic spikes used in the STDP measurements is shown as insets.g) Current dependence recorded during the application of ten stimulation pulses under on/off light with different pulse amplitudes (+5 and +6 V). h) The light-induced PPF function of the device.Inset: Schematic diagram of the measurement of light-induced PPF.i) Current responses to ten consecutive light pulses under the various light intensity (1.98, 3.56, and 5.28 mW cm −2 , duration: 1 s, interval: 1 s), followed by conductance decay after light is off.j) Current responses to different wavelengths of light (red: 650 nm, green: 550 nm, and bule: 450 nm).

Figure 3 .
Figure 3. a) Schematics of two-terminal memristive device in ITO / P3HT / AlO x (3.6 nm) / Ag. b) Currents extracted from the I-V curves at ±5 V plotted as a function of number of cycles.O 1s XPS spectra collected from the ITO electrode at c) initial, d) LRS, and e) HRS.f) Schematic diagrams illustrating the proposed memristive switching mechanisms in P3HT memristors.Fitted I-V curves of the g) first (HRS, blue scatter plot) and 30th (LRS, red scatter plot) cycle with tunneling and h) SCLC conduction mechanisms.