Synaptic Iontronic Devices for Brain-Mimicking Functions: Fundamentals and Applications ∥ ∥ Changwei Li, Tianyi Xiong, Ping Yu,* Junjie Fei,* and Lanqun Mao

www.acsabm.org Review

Cite This: ACS Appl. Bio Mater. 2021, 4, 71−84 Read Online

ACCESS Metrics & More Article Recommendations

ABSTRACT: Inspired by the information transmission mechanism in the central nervous systems of life, synapse-mimicking devices have been designed and fabricated for the purpose of breaking the bottleneck of von Neumann architecture and realizing the construction of effective hardware-based artificial intelligence. In this case, synaptic iontronic devices, dealing with current information with ions instead of electrons, have attracted enormous scientific interests owing to their unique characteristics provided by ions, such as the designability of charge carriers and the diversity of chemical regulation. Herein, the basic conception, working mechanism, performance metrics, and advanced applications of synaptic iontronic devices based on three-terminal transistors and two-terminal memristors are systematically reviewed and comprehensively discussed. This Review provides a prospect on how to realize artificial synaptic functions based on the regulation of ions and raises a series of further challenges unsolved in this area.

KEYWORDS: iontronics, memristor, transistor, artiﬁcial synapse, neuromorphic device

1. INTRODUCTION numerous regulation strategies including charge transport,21 22 ff Inspired by the synapse structure in the nervous system noncovalent interactions, and so on. (3) The di erent carrying almost all intelligent characteristics of life in a tiny structure of ions provided them diversity in valencies, sizes, 1 and polarizabilities, and as a result, ion current carries more volume with low energy consumption, research attention has 23 been paid to the fabrication of synapse-mimicking devices for information when compared with electronic current. In this artificial intelligence at the hardware level in order to meet the case, iontronics, a newly emerging interdisciplinary conception requirement of diverse neuromorphic computing tasks such as focusing on the designation of devices based on various 24,25 pattern classification,2 decision making, autonomous driving, behaviors of ions, came into sight for the fabrication of and personal healthcare.3 Hitherto, a variety of synaptic synaptic devices including two-terminal memristors and three- devices based on multifarious materials including carbon terminal transistors in the past decade, and synaptic iontronic − Downloaded via UNIV OF NEW SOUTH WALES on April 19, 2021 at 23:18:54 (UTC). materials,4 6 silicon materials,7,8 perovskite structures,9,10 devices, which simulated the synaptic characteristics with ions, ferromagnetic materials,11,12 and metal oxides13 have been were designed and fabricated based on ion doping/transport/ See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles. − designed and fabricated since the discovery of transistors in redox in gels,26,27 polymers,28 30 ionic liquids,31 confined Bell Laboratories14 and the development of memristors by HP space,32 and so on. 15 Laboratories. The fabrication of these devices provided a To date, development of synaptic devices, especially synaptic new possibility to the designation of integrated circuits and iontronic devices, has been reviewed in previous reports from raised the efficiency of machine learning. different views. For example, Kuzum et al. systematically sorted ff The main di erence between traditional synaptic devices synaptic devices through the differences in materials, and biological synapses is the species of charge carriers: mechanisms, and applications.33 Yang et al. summarized electrons and holes play the vital role in the conduction of memristive devices for computing,16 and He et al. reviewed electronic synaptic devices, while information was mainly ionic-transistor-based synaptic devices.34 Yu et al. concluded transported and processed based on ions in biological synapses.16 Recent attempts have revealed that ions bear omnifarious unique characteristics providing them great Special Issue: Wearable and Biodegradable Sensors potential for application in the designation of synaptic devices: Received: June 30, 2020 (1) The mobility of ions is much lower than that of electrons, Accepted: August 17, 2020 providing them a potential response ability against external Published: August 17, 2020 stimulation such as light,17 temperature,18 and pressure.19,20 (2) Abundant chemical characteristics of ions contributed to

Figure 1. Signal procession in biological synapses and synaptic iontronic devices. (a) Schematic illustration of information transmission in biological (left) and synaptic iontronic devices of different mechanisms (right). Adapted with permission from ref 38. Copyright 2018 WILEY- VCH. (b−d) Synaptic weight changes in synaptic iontronic devices induced by (b) one single spike stimulation and (c) two continuous spike stimulation mimicking paired-pulse facilitation (PPF) in biological synapses (Reprinted with permission from ref 47. Copyright 2015 WILEY- VCH.) and (d) stimulation of continuous spike series mimicking STP−LTP transition. Reproduced from ref 48. Copyright 2012 American Chemical Society. synaptic iontronic devices from structures,35 and Zhu et al. was further demonstrated based on the understanding of reviewed the progress of nanoionic resistive-switching synaptic plasticity in this section. devices.36 However, to the best of our knowledge, little 2.1. Synaptic Plasticity in Biological Synapses and attention has been paid on how ions in synaptic iontronic Synaptic Iontronic Devices. The biological synapse is the devices work and how to regulate ions in these devices, which key junction for signal procession in neural networks. In neural is not only of great importance for the fabrication of iontronic systems, two ways of signaling between synapses formed the devices but also for understanding the characteristics of ions widely connected neural network: ion−ion in electrical fundamentally. synapses and ion−neurotransmitter−ion in chemical synapses. This review systematically summarized the recent progresses Most electrical synapses show simple synaptic behaviors, on of synaptic iontronic devices and was organized into six account of being founded in invertebrates such as fish and sections: The first section briefly introduced the recent crustaceans.37 And the ion−neurotransmitter−ion pathway in progresses in the designation of synaptic iontronic devices. chemical synapses contributed to complex neural functions in The second section focused on the synaptic characteristics higher beings like humans. In this case, understanding and mimicked by the synaptic iontronic devices. The third section mimicking the working mechanism of chemical synapses is of discussed the synaptic iontronic devices based on three great importance for the implantation of synaptic iontronic ff di erent mechanisms, how these devices mimicked the devices. Fundamentally, there are three major components in behaviors of synapses, and how ions contributed to these chemical synapses: presynaptic neuron, synaptic cleft, and mechanisms. The fourth section sorted the evaluation 38 fi postsynaptic neuron (Figure 1a, left panel). In a synaptic standards of the synaptic iontronic devices. The fth section event, ion current signal transmits from the presynaptic neuron introduced the research approach and advanced applications of and drives the release of neurotransmitters to the cleft, further these devices. Finally, a brief summary and the perspective on stimulating the postsynaptic neuron and leading to the this area were presented in the last section. transmission of signal from the presynaptic neuron to the postsynaptic neuron. 2. SYNAPTIC CHARACTERISTICS IN SYNAPTIC Due to the ion−neurotransmitter−ion mechanism in IONTRONIC DEVICES chemical synapses, it was revealed that synaptic characteristics Toward simulating synaptic behaviors in neuronal networks could be regulated by the occurrence of synaptic events: the with synaptic iontronic devices for effective neuromorphic strength of synaptic junctions (i.e., synaptic weight), which is computing, understanding of the synaptic characteristics and related to the efficiency of information processing between how synaptic iontronic devices mimicked these characteristics neurons, changes with the occurrence of synaptic spikes; that is 39−41 should be taken into consideration at first. In this section, one synaptic plasticity. In biological synapses, synaptic of the most fundamental but important synaptic characteristics, plasticity is related to the cellular state of neurons, which is synaptic plasticity, which directly contributed to study and associated with the intracellular and extracellular environment, memory functions of the brain, was comprehensively reviewed such as the Ca2+-dependent nature of the neurotransmitter − to understand the role of ions in biological synapses and release process.42 45 This synaptic plasticity characteristic synaptic iontronic devices. And how to mimic synaptic contributed to the signal procession function of neurons by plasticity of different patterns with synaptic iontronic devices regulating the synapse weight between neurons and outputting

72 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review different signals when dealing with highly integrated complex 2). In biological synapses, this formation of LTP is related to a neural events.46 long-term change of the synaptic connection strength in a To mimic the behaviors of chemical synapses, especially synaptic plasticity with synaptic iontronic devices, the directional movement of ions in the matrix mimicked the release and transport of neurotransmitter in synaptic clefts, and the conductance of the devices has been taken into consideration, representing synaptic weight in synaptic iontronic devices (Figure 1a, right panel). Pulse signals were conducted mimicking the spike-form synaptic events. In this case, the change of conductivity under spike-form stimulations simulated the change of synaptic weight in biological synapses Figure 2. Schematic illustration of the formation of LTM and the or rather mimicked neural plasticity.37,49 It should be noticed transition of STP toward LTP. that in chemical synapses, signals were transmitted based on ion current and the recognition of the neurotransmitter at the 50 postsynaptic membrane. Therefore, transport and reactions 53−55 of ions are of great potential for tuning the conductivity of neural network. Similarly, for synaptic iontronic devices, synaptic iontronic devices and mimicking the synaptic LTP behavior could be simulated by the change of nonvolatile plasticity. conductivity induced by the accumulation of rehearsal and 2.2. Basic Synaptic Plasticity: Short-Term Plasticity reversible changes (STP mimic) under continuous stimuli. − (STP) and Long-Term Plasticity (LTP). To simplify the Therefore, to mimic this STP LTP transition with synaptic plasticity of the brain, which is a complex integrity of iontronic devices, a hysteretic conductivity and nonvolatile kaleidoscopic synaptic behaviors, the relationship between characteristic are two essential requirements. In synaptic signal potentiation/depression and input signal impulses in iontronic devices, the slow mobility of ions (compared to one single synapse was first taken into consideration. And the switching speed of voltage pulse) contributed to the hysteretic essential synaptic plasticity of one single synapse was divided conductivity, and the diverse chemical characteristics of ions into two patterns according to the time scale difference: STP also provided abundant potential regulation strategies for this − and LTP. As a single short voltage spike was introduced to a reversible irreversible transition. synapse, the synaptic weight changes under the stimulation of 2.4. Advanced Synaptic Plasticity. By using synaptic synaptic spikes and is preserved for a short time period (i.e., iontronic devices mimicking the synapse structures and spike- retention time, few ms∼mins)51 and further returns to the form stimuli simulating synaptic events, synaptic plasticity of original state (Figure 1b).47 And this short-term trend of more complicated patterns had been further emulated, such as retaining at the stimulated state was defined as the STP. spike-time-dependent plasticity (STDP) and spike-rate-de- Further, when this retention trend of synapses rises to a longer pendent plasticity (SRDP). A basic form of STDP focuses on time scale (e.g., hours∼days), LTP, another essential pattern of the temporal order of pre/post-synaptic spikes and was 56 neural plasticity with a longer retention time associated with proposed as a form of Hebbian synaptic learning rules: the formation of long-term memory (LTM), occurs instead.52 Synaptic weight increases when the presynaptic potential When considering STP behaviors, two main forms were arrives before the postsynaptic and drops when the presynaptic 57 observed in biological synapses: paired-pulse facilitation (PPF) potential arrives after the postsynaptic on the contrary. And a and paired-pulse depression (PPD).42 When a second spike is positive correlation between the time difference Δt and fed into the synapse, inducing a stronger postsynaptic current synaptic weight change Δw was revealed in STDP behaviors. (A2) than the previous one (A1), PPF takes place as shown in Meanwhile, SRDP focuses on the relationship between the Figure 1c.47 On the contrary, PPD takes place when the frequency of synaptic spikes and the synaptic weight. second spike is depressed by the previous one. PPF and PPD Stimulation with higher frequency contributed to the increase could be numerically described by a double exponential decay of synaptic weight, while a lower frequency signal contributed − τ 58,59 with the form of PPF (or PPD) = 1 + C1exp( t/ 1)+ to the reduction of synaptic weight. These advanced − τ 42 C2exp( t/ 2), or a simple quantitative evaluation based on patterns of neural plasticity contributed to signal processing the change of the excitatory/inhibitory postsynaptic current mechanism of various patterns to both biological synapses and (EPSC/IPSC) in neurons (i.e., PPF or PPD = A2/A1), where, t synaptic iontronic devices, and ions in these devices provided τ is the time interval between the pair of stimuli, and C1, C2, 1, the hysteretic and nonvolatile characteristics of conductivity. τ 2 are the initial facilitation magnitude and the characteristic Hitherto, the work mechanism of brain is still an unsolved relaxation times of the rapid and slow phases, respectively. puzzle, and a series of other signal procession patterns are still 2.3. STP−LTP Transition in Biological and Artificial not clear and require further study in neuroscience. Synapses. As was mentioned above, LTP contributed to the Furthermore, there have been other rules for neural signal formation of LTM and represents the retentive trend of processing and neural computation, such as associative/ synapses at a longer time scale. It was proven that LTP could nonassociative learning,60,61 pattern learning,62,63 and synaptic be realized based on rehearsal training stimulation and the scaling.64,65 Each pattern of synaptic plasticity has an following accumulation of STP (Figure 1d).48 In a neural associated role in brain, for example, STP for the dynamic network, memory was formed based on this STP−LTP synaptic input in neural network and LTP for the formation of transition. After the input of information stimulation, STP LTM.66,67 By using an individual synaptic device to simulate was formed at first, induced by external stimulation, training, or the synaptic plasticity of various patterns, artificial synaptic learning, leading to the enhancement of neural plasticity and networks could be rationally established to figure out certain finally contributing to the formation of LTP and LTM (Figure tasks and applications.

73 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review

Figure 3. Synaptic ionic transistors with single or multigate/channel structures. (a) Schematic illustration for synaptic ionic transistor based on the doping and electrochemical polymerization/overoxidation of an evolvable channel. (b) With the gate pulse (−0.5 V, 30 times), LTP behaviors of the device described in (a). Reprinted with permission from ref 78. Copyright 2019 The Authors. Published by WILEY-VCH. (c) Schematic ﬂ − illustration of a multiresponsible MoS2 ionic transistor, which could be in uenced by optical, electronic, and ionic signal. (d) STP LTP transition based on electronic (left), iontronic (middle), and photoactive (right) responses of the device in (c). Reprinted with permission from ref 38. Copyright 2018 WILEY-VCH. (e) Schematic illustrations of a biological synaptic integration (upper) and a multi-in-plane-gated transistor structure (lower). Reprinted with permission from ref 80. Copyright 2017 AIP Publishing. (f) Conﬁgurated multichannel PEDOT:PSS device connected by common electrolyte (right) mimicking the global input and local input in biological neural networks. Adapted with permission from ref 81. Copyright 2017 Springer Nature.

3. MECHANISM OF SYNAPTIC IONTRONIC DEVICES conductive polymers, 2D materials, metal oxides, and nano- 69−75 To mimic the characteristics of biological synapses where particles. For example, Malliaras et al. demonstrated information was disseminated in a fJ level low energy cost, neuromorphic characteristics of organic electrochemical miscellaneous materials like conductive polymers, inorganic transistors (OECTs), by using poly(3,4-ethylenedioxy- solid electrolytes, and ionic liquids were taken into thiophene):poly(styrenesulfonate) (PEDOT:PSS) as conducting polymer,76 and synapse-mimicking characteristics including consideration for the fabrication of synaptic iontronic fi devices.26,32,68 In order to develop new materials for synaptic PPD, adaptation, and dynamic ltering were observed and iontronic devices to reduce the energy cost and raise the described. However, the plasticity of this neuromorphic device efficiency of signal procession and dissemination in these was limited. Toward stronger plasticity for LTP simulation, devices, it is of great importance on understanding how ions nonvolatile synaptic transistors were further developed based work as charge carriers and how ions were regulated in these on irreversible changes of devices induced by the ion doping − 76−78 devices. In this section, synaptic iontronic devices based on ion process at the channel electrolyte interface. Burgt et al. doping process at interfaces, ion transport in confined space, reported a neuromorphic OECT with a PEDOT:PSS gate and and redox characteristics of ions were comprehensively PEDOT:PSS/PEI channel, the combination of the OECT discussed in order to show the unique characteristics of ions device and battery-like mechanism contributed to its strong 79 compared with synaptic devices based on electrons. nonvolatile characteristic and low operation voltage. By 3.1. Synaptic Iontronic Devices Based on Ion Doping. applying a well-setting continuous pulse stimulus, a variety of Ionic transistors are a series of three-terminal iontronics where synaptic functions including STP−LTP transition, Pavlovian ion current driven by the gate electrode could be reflected to learning, and image recognition were realized based on this the change of semiconductive channel membrane between the device. Owing to the particularity of the materials, this device source electrode and drain electrode. Inspired by the is flexible to some extent without performance loss, and geometrical similarity between biological synapses and ionic neuromorphic computing could be effectively conducted under transistors, transistor-based synaptic iontronic devices were mechanical deformations. Jennifer et al. demonstrated a designed and fabricated with various materials, such as synaptic OECT device with an evolvable channel. It was

74 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review

Figure 4. Iontronic memristor based on biological nanopore: (a) Schematic illustration of a biological alamethicin-nanopore-based (ion-transport- based) memristor. (b) Hysteretic curves of the memristor described in (a) with different voltage scan rates. (c) SRDP and STDP simulation based on synaptic iontronic device in (a); STDP simulation was conducted by a connection with a nonvolatile memristor. Reproduced from ref 98. Copyright 2018 American Chemical Society. observed that by inducing nonequilibrium doping states at the the channel resistance on the history of the device and semiconductive channel, STP could be obtained by the device, contributed to its application in synaptic iontronics. In this and further LTP could be realized by electrochemical case, how to change the doping process relying on the polymerization/overoxidation of the evolvable channel materi- regulation of the channel membrane/electrolyte interface and al. And a conditioning circuit was performed based on this material selection is the essential problem in the designation synaptic iontronic device (Figure 3a,b).78 and fabrication of synaptic transistors. Moreover, it should be noticed that by stimulating the 3.2. Synaptic Iontronic Devices Based on Ion Trans- channel with other regulators, signal of other sources could port. Inspired by the ion channels in biological synapses as also be received and processed by synaptic transistors. John et well as a series of counterintuitive behaviors of ions in − al. developed a transistor that could be regulated by three confinements,83 88 nanopores/channels of different geometric different modes: electronic, iontronic, and photoactive modes structures and materials have been taken into consideration for (Figure 3c).38 Apart from doping of ions, electron trapping− the fabrication of iontronics devices including ionic detrapping and the photoelectric effect influenced the switches,89,90 ionic amplifiers,91 ionic diodes,92,93 ionic semiconductive channel and also contributed to the change transistors,94 and ionic sensors.95,96 It was revealed that the of output signals. And it was proved that the STP−LTP ion conductivity of nanopores/channels was controlled by a transition could be observed with a specially designated pluses confined area. When a bias potential was applied, ions migrate signal from these three modes (Figure 3d). These multifunc- directionally and show memristive characteristics with the tional synaptic iontronic devices provided potential application regulation of the applied electrical field. In this case, ion- in photosensitive neuron-mimicking and smart response to transport memristors were further designed and fabricated multiple stimulation. based on biological nanopores and solid-state nanopores of Apart from single synaptic transistors, multigate/channel various characteristics.97 structures were also applied to the designation of ionic- To emulate the behavior of biological synapses, protein transistor-based synaptic iontronic devices. The multigate/ nanopores like microtube proteins and ion channels, which are channel structures simulated the neuron network in the brain, almost homologic to ion channels in biological synapses, were and unique characteristics were observed. Qian et al. fabricated first taken into consideration. Sarles et al. fabricated a a multigate transistor device, and the relationship between biological-channel-based memristor with a phospholipid signal input weight and gate-channel distance or angle were membrane modified by alamethicin peptides (Figure 4a).98 systematically discussed. The smaller the interval and distance, With the stimulation of certain voltage bias, alamethicin the greater the change of synaptic weight; the longer the pulse peptides were inserted into the phospholipid membrane duration, the smaller the change in synaptic weight (Figure forming ion-permeable channels and further contributing to 3e).80 Malliaras et al. designed a multichannel OECT device the memristive behavior of devices. This reversible and with one global gate, mimicking the complex connection voltage-driven on−off process contributed to the hysteretic between neurons in brain by a common electrolyte.81 Global current−voltage (I−V) curve and synaptic behavior of the clocks based on the global gate electrode, a soft link based on device (Figure 4b). And synaptic functions including the the interaction among the single channels, and coincidence STP−LTP transition, SRDP, and STDP were realized based detection were demonstrated based on this mechanism (Figure on this nanopore structure in a phospholipid membrane 3f). It should be noticed that integrated signal was given by the (Figure 4c). multitransistor networks, which is of great similarity to However, the fragile intensity of phospholipid membranes biological neural network. limited the further application of protein nanopore structures It was revealed that the switch speed of these ion-conducted for iontronic devices; in this case, synaptic iontronic devices devices is lower than that of electron-conducted metal-oxide- fabricated with solid-state nanopore/channels were taken into semiconductor field-effect transistors (MOSFETs),64 which is consideration. Based on the conductivity difference between induced by the limited velocity of the doping processes of ions electrolyte solution and hydrophobic ionic liquids, Zhang et al. at the interface.82 This characteristic imparts a dependence of designed and fabricated a nanochannel-based ionic memristor

75 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review

Figure 5. Iontronic memristor based on solid-state nanochannels: (a) Schematic illustration of a nanofluidic memristor; aqueous KCl solution and fi fl hydrophobic BmimPF6 were lled in the nanochannel. (b) Hysteretic curves of a nano uidic memristor containing 100 mM KCl solution and fl BmimPF6. (c) LTP and LTD simulation based on device described in (a). (d) Conductivity change of the nano uidic memristor in times of read/ write circulation. Reproduced from ref 32. Copyright 2019 American Chemical Society. (e) Hysteretic curves of a conical nanopore in 10 mM KCl solution under different scan rates. (f) Schematic illustration of the electric double layer polarization in conical nanopores induced by the limitation of ionic mobility under triangle waves, which contributed to the memristive behavior of conical nanopores. Reproduced from ref 99. Copyright 2012 American Chemical Society. with an aqueous solution of KCl and hydrophobic ionic liquid by the polarization of the electrical double layer in nanoscale 32 fi − BmimPF6 at the either side of the channel (Figure 5a). con nement. This hysteresis loops in I V curves indicated that Owing to the mismatch of voltage-changing velocity and the ion transport in nanoconfined solution was of potential − 99 moving velocity of the KCl BmimPF6 interface, which is application in the designation of memristors (Figure 5e,f). related to viscosity of nanofluids and the formation efficiency However, how to convert this transient transport behavior into of the electrical double layer in the KCl filled area, the a nonvolatile memristor was still a great challenge to this hysteretic characteristic was observed in the I−V curve of the mechanism. Raising the influence of the electrical double layer nanochannel (Figure 5b). And simulations of LTP and LTD and reducing the mobility of ions was a potential strategy for were realized based on this nanochannel-based ionic memristor the designation of nonvolatile memristors based on this (Figure 5c). mechanism. Furthermore, it was believed that a series of Although ionic liquid was introduced to the nanochannel for effective regulation methods such as field-effect regulation and the emulation of LTP, ionic liquids, especially hydrophobic noncovalent recognition bear potential application in the ionic liquids, limited their potential application in complex and designation of synaptic iontronics based on ion-transport mild environments (e.g., brain or blood). In this case, the behaviors also.88 fabrication of synaptic iontronic devices in aqueous environ- 3.3. Synaptic Iontronic Devices Based on Ion Redox. ment was still a great challenge.100 Owing to the slow mobility Apart from the transport/doping process of ions, chemical of ions, which could be easily regulated by the external characteristics of ions also contribute to the construction of environment, and the key conductance switching area, which synaptic iontronic devices. Based on the formation/rupture of was controlled by the electrical double layer at the orifice of a conductive pathway originating from the redox processes of the conical nanopore structure, Wang et al. observed the ions, hysteretic and nonvolatile memristors had been designed hysteretic characteristic of conical nanopores, which is induced and fabricated in two different types: valence-changing

76 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review

− Figure 6. Valence-changing memristors with oxygen vacancies. (a) Schematic illustration of the cohesion disruption process of the VO migration pathway. Reprinted with permission from ref 103. Copyright 2012 AIP Publishing. (b) Part of the hysteretic curves and TEM image of a Pt/WOx/ fi W valence-changing memristor. (c) Schematic illustration of the VO dynamics in the device described in (b): the electric- eld-driven VO drift (left) and the spontaneous diffusion (right). (d) STDP simulation based on the device described in (b). Reprinted with permission from ref 68. Copyright 2015 WILEY-VCH. (e) Schematic illustration of VO migration in a GdOx/Co/Pt magnetic valence-changing memristor. (f) Learning and forgetting mimic based on a GdOx/Co/Pt valence-changing memristor. Reprinted with permission from ref 104. Copyright 2019 American Physical Society. memristors and conductive-bridge memristors, or rather anion- Based on the theoretical consideration of valence-changing based and cation-based memristors, respectively. In a valence- ionic memristors, Du et al. fabricated a valence-changing changing memristor, redox processes of ions in oxide-based memristor with a Pd top electrode, WOx switching layer, and insulators, which bear a wide band gap, contributed to the tungsten (W) bottom electrode based on the internal ionic formation/rupture of the conductive path for the oxygen dynamics (Figure 6b). And it was demonstrated that two vacancies (VO), resulting in valence changes in the material transport steps contributed to the modulation of resistance: fi ff and the resistance switch, simulating the change of synaptic (1) electric- eld-driven VO drift and (2) spontaneous di usion 101,102 weight. Analogously, in conductive-bridge memristors, of VO, corresponding to STP and LTP synaptic-mimicking redox of precious metal cations contributed to the formation/ characteristics, respectively (Figure 6c). And further advanced rupture of the metal conductive bridge and changed the STDP and SRDP were further mimicked by rationally tuning resistance of the devices. The chemical characteristics of ions the waveform of the applied stimulation (Figure 6d).68 contributed to the possibility of the redox processes, and the Furthermore, artificial synaptic devices based on oxygen migration of ions in these devices contributed to the hysteretic vacancy migration/redox can also be combined with other characteristic of conductance. resistance modulation methods. Mishra et al. reported a three- For valence-changing memristors where the conductivity terminal magnetic synapse fabricated with magnetic layer Co, was controlled by the redox of anions, density functional Pt, and GdOx. Under the induction of a negative gate bias theory (DFT) results show that the formation and disruption voltage, oxygen vacancies move to the Co layer from the GdOx of oxygen vacancy conducting channels could be triggered by layer, resulting in the valence changing of the Co layer and the injection and removal of carriers (Figure 6a). And based on reducing the overall saturation magnetization of the Co layer the injection location (electrodes or filament at higher bias) of (Figure 6e).104 Simulations of Hebbian STDP and learning/ electrons or holes, resistance switching behaviors of two forgetting were conducted through converting pulse-timing patterns, which are related to the selectivity of dielectric information into pulse-width information (Figure 6f).104 ff materials, could be theoretically elucidated: unipolar switching Di erent from valence-changing memristors where VO plays and bipolar switching. It was also proved that a matched Fermi a key role in the resistance switch mechanism, the resistance of level of the electrode against the VO level is necessary for the conductive-bridge memristors was related to the formation and formation of a valence-changing memristor.103 rupture of the metal filament originating from the reduction/

77 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review

Figure 7. Typical synaptic iontronic devices based on a conductive-bridge mechanism. (a) Schematic illustration of a conductive-bridge-based synaptic iontronic device based on the redox of Ag+/Ag nanoparticles and the formation of a Ag conductive bridge. (b) Schematic illustration of a neural network based on the device described in (a). (c) Measured (blue) and calculated (orange) I−V curves of the device described in (a) with voltage sweeps. (d) Simulated LTP behavior of the device described in (a). Reproduced from ref 55. Copyright 2010 American Chemical Society.

Table 1. Performance Metrics of Ionic Synaptic Devices/Memristors

energy mechanism materials device size voltage range consumption retention time function ref biological synapses ions and proteins with tissue fluid 20−40 nm ∼100 mV ∼10 fJ few ms∼years all 33 ∼ × ± valence changing WOx 500 500 nm 1.5 V N/A 5.25 ms EDP, STDP 68 − ∼ ∼ μ valence changing SrTiO3 few tens of nm 4.3 2.2 V 0.026 0.46 fJ 23.5 s SRDP, STDP 113 × − ∼ valence changing TiOx, HfOx 2 2nm 5 2 V 23 nJ N/A LTP, LTD 109 conductive bridge Cu/a-Si/Pt 90 nm −3∼1.6 V N/A ∼30 ms LTP, LTD STDP 105 × ± ∼ conductive bridge SiOxNy/Ag 100 100 nm 3 V N/A 29 ms SRDP, STDP 106 ± ∼ ionic transistor NdNiO3, porous silica 360 nm 4V 1 pJ N/A LTP, LTD 75 ± ∼ ionic transistor SmNiO3, ionic liquid few tens of nm 2.5 V N/A 2.7 h STDP 70 ionic transistor organic nanowire, ionic liquid 300 × 200 nm ±1V ∼12.3 fJ ∼10 s SRDP, STDP 108 nanochannels lipids, alamethicin channel single protein level ±150 mV 15∼1500 pJ 7 ms SRDP, STDP 98 nanochannels KCl solution and ionic liquid 200 × 63 nm ±20 V N/A N/A LTP, LTD 32 oxidation of metal ions/atoms.105 Wang et al. implemented a The simple structure of conductive-bridge-based synaptic conductive-bridge memristor with Pt/Au inert electrodes and memristors contributed to their advantages in fabricating 106 − SiOx/Ag for the dielectric layer. I V curves display a stable integrated arrays toward an artificial neural network. Jo et al. and hysteretic characteristic of this device, and in situ fabricated an artificial neural network based on the integration transmission electron microscopy (TEM) observation showed of Ag conductive-bridge memristors (Figure 7a,b).55 I−V that under a bias potential stimulation, the oxidation of fi + curves with several sweeps got a good t with a simple embedded Ag clusters in the SiOx layer produced Ag to the memristor circuit model (Figure 7c), and the emulation of dielectric layer, and a further reduction reaction around LTP/LTD was well-accomplished based on the crossbar array electrodes induced by the electric field converts Ag+ to (Figure 7d); a good operation life was revealed with optimized nanoscale metal filaments (i.e., conductive bridge) to change conditions. the conductivity of the device. This time-dependent process ff contributed to the hysteresis of the CV curve as well as the Compared to the electronic e ect, which is completely based history-dependent characteristic. It should be noticed that on the physical behavior of electrons, ionic physical/chemical transport dynamic of Ag in this device is of great similarity to modulation is more controllable and designable. The structure the cross-membrane dynamic of Ca2+ in biological synapses. of ionic devices could be easily regulated, and the reactions or Interestingly, this device successfully simulated the appearance transport of ions would cause reversible/irreversible change of of PPD after a period of PPF, and STDP was further realized the materials and conductance in synaptic iontronic devices, based on the connection with a TaOx drift memristor. which is the key of mimicking STP/LTP behaviors.

78 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review

4. PERFORMANCE METRICS IN SYNAPTIC nm sized nanoscale synaptic iontronic devices could be IONTRONIC DEVICES prepared. For example, Pi et al. developed memristor crossbar arrays with a 6 nm half-pitch and 2 nm critical dimension.109 Faced with the requirements of various dimensions toward And image memory was realized based on the array through application in complex integrated circuits, full investigation and programming each device in the array into either on or off quantitative standards of different aspects evaluating the state. performances of synaptic devices were required for the 4.5. Regulating Range of Resistance. For a constructed purpose of mimicking biological synapses or even surpassing fi fl 33 memristor, the regulating range of resistance is xed, re ecting the performances of them. Meanwhile, it should be noticed the resistance change area of the device. To mimic the that there should be different evaluation standards for synaptic ff behaviors of neurons, conductivity changes in various states iontronic devices applied in di erent aspects, for example, a mimicked the change of synaptic weights under bias potential capacity for integration in brain-inspired computing stimulation. It is proved that conductivity changes between and biocompatibility for the formation of sensorimotor − 107 50 and 100% were suitable for the formation of neuro- circuits. In this case, universal evaluation standards for morphic devices.33 synaptic iontronic devices were systematically sorted in order ff To the application in complex integrated circuits, a series of to show what e ects ions have brought to the devices in this other performances such as robustness under different section, and a performance metric evaluating biological ff physical/chemical environments, long-term stability, and synapses and synaptic iontronic devices of di erent mecha- behaviors under noise fluctuation also need to be taken into nisms is given by Table 1. consideration. And the establish of universal quantitative 4.1. Energy Consumption. Energy consumption per standards reflecting these performances still requires research synaptic event is the foremost raised performance metric of a attention in this area. single synaptic device. A synapse-mimicking iontronic device requires comparable energy consumption against a biological 5. APPLICATION IN BRAIN-INSPIRED COMPUTING synapse or even lower. To reduce the energy consumption of AND SENSORIMOTOR CIRCUITS devices, reducing the size and optimizing the designation of Based on synaptic iontronic devices of various mechanisms, a devices were proven to be useful strategies. For example, Xu et series of further applications have been taken into consid- al. designed an organic nanowire (ONW) synaptic transistor eration and contributed to artificial intelligence at the hardware architecture with 1.23 fJ average energy consumption per spike level as well as their further applications in robotic or other in the same condition as the signal input of the biological 110 108 area. The characteristics of iontronic materials contributed synaptic event. For the purpose of setting the desired energy to a series of unique advantages to these applications, which consumption of the devices, rational regulation of the could be classified into memory, digital, and analog neuro- waveform parameters (input amplitude, pulse duration, and 111 ff morphic systems with synaptic devices. In this section, the intervals) was also e ective. typical applications of synaptic iontronic devices in brain- 4.2. Retention Time. The nonvolatile electrical con- inspired computing as well as the formation of artificial neural ductivity of the device is a crucial factor in the construction of fi fl circuits have been comprehensively reviewed, and we focus on arti cial synapses, which is re ected by the retention time. the research approach of the application of ion-based materials. Many studies have reported that a few milliseconds to 10 min 5.1. Brain-Inspired Computing. To emulate the can be defined as STP, and longer to 2 or 3 days can be fi computing mechanism of the central nervous system (CNS), de ned as LTP. Additionally, this ability is related to the simple combination or even deep neural networks (DNNs) properties of the materials used to construct the device, the based on a 2D or 3D integrated array of iontronic synapses and kinetics and thermodynamics of ion-doping, migrating, or other devices have been designed and fabricated with redox processes. nanofabrication techniques. And it was proved that these 4.3. Pulse Voltage Amplitude. Although the biological integrated devices bear extraordinary efficiency when com- ∼ chemical synaptic action potential is only 100 mV, the pared with traditional large-scale integrated circuits, owing to ’ requirements of the device s voltage settings should be taken their functional similarity against synapses as well as their fi into consideration at rst for further application in complex effective computing ability. And a series of brain-inspired integrated circuits. An excessive input voltage may cause computing applications including decision making, pattern breakdown and a irreversibile change of conductivity, which recognition, and arithmetic computing have been effectively could be induced by irreversible chemical/electrochemical realized based on these devices with quite limited device units. processes in the synaptic iontronic devices. Meanwhile, an When it comes to these application of the devices, rational insufficient input voltage might contribute to an unsuitable and effective algorithms are required to be taken into long response time, which is even longer than the retention consideration at first. For example, Prezioso et al. reported time. Therefore, an effective amplitude for the devices decided the training and operation of an integrated neuromorphic the working environment of the devices. network based on metal-oxide memristors.62 And pattern 4.4. Device Dimensions. To match the density of classification experiments were conducted with a fabricated synapses in the brain as well as the highly integrated electronic memristive crossbar: 10 channel signals containing 3*3 binary devices in silicon chips, the size of synaptic devices is a key images and a 10th bias input were introduced to the factor. Moreover, the conductivity range is related to the size of neuromorphic network; synaptic weights of the units were the devices. The size of the devices is decided by the work recorded according to the effective conductance of each mechanism of the synaptic iontronic devices: one single memristor. As the chart of the training algorithm described,62 nanochannel is around 10 nm; most conductive-bridge devices the output of the perceptron f i is calculated, and synaptic are between 10 and 20 nm, followed by ionic transistors with a weights were updated to finish the training through using the size of about 100 nm. Based on nanofabrication technologies, 2 Manhattan update rule and setting the target value of the ith

79 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review

Figure 8. Image recognition, memory applications for brain-inspired computing. (a) Schematic illustration of the computing strategy toward a handwritten digit-recognition task based on a neural network fabricated with synaptic nanochannels. Reproduced from ref 32. Copyright 2019 American Chemical Society. (b) Memory of a 5*7 binary image based on a nonvolatile synaptic transistor array. Adapted with permission from ref 112. Copyright 2018 WILEY-VCH. (c) Input of binary images based on spike-form inputs of different amplitude and frequency and 5*7 image memory based on an oxygen vacancy memristor. Reproduced from ref 113. Copyright 2019 American Chemical Society. output for the nth input pattern. Zhang et al. realized actuators, the formation of artificial reflex arcs could also be handwritten digit recognition by building a convolutional constructed in this way. By outputting a biomimic response neural network (CNN) with a nanochannel-based ionic based on the calculated stimulation from a simple signal memristor (Figure 8a).32 The Manhattan update rule was provided by the artificial receptor, enormous potential taken into application while an online supervised learning applications were proposed and demonstrated with these algorithm was employed, and a benchmark pattern recognition smart devices. Furthermore, toward the designation of with an accuracy of 94% was realized. wearable or implantable artificial reflex arcs, biocompatibility With the assistance of an algorithm, the realization of and soft materials were required for the fabrication of artificial effective neuromorphic computing also requires synaptic reflex arcs. And soft materials like hydrogels and biocompatible iontronic devices with suitable performances. For example, polymers have been widely applied in synaptic iontronic the effective accomplishment of image memory tasks requires a devices,115 contributing to their great potential in the nonvolatile characteristic of synaptic iontronic devices. Jin et al. formation of artificial reflex arcs. fabricated a 5*7 array based on tungsten oxide synaptic The artificial reflex arc was demonstrated by Lee et al. by transistors, and the memory of 5*7 pixel images was realized rational assembly of a photodetector, artificial synapse, owing to the nonvolatile behavior of the devices (Figure transimpedance circuit, and artificial muscle actuator.114 8b).112 Li et al. mimicked the selective activation in the parietal When optical stimulation was inputted to the photodetector, cortex through a neuromorphic network composed of oxide- a carbon-nanotube-based stretchable organic nanowire syn- tunnel-junction-based individual synapses: three letters were aptic transistor transmitted the corresponding current signal input into the selected individual pixels with different memory from the photodetector to the actuator and further led to the modes in time order, and the memory mode difference bending of polymer actuator. The synaptic transistor responds regulated by the pulse amplitude as well as frequency to electrical signals from the photodetector, and the resistance contributed to the image memory (Figure 8c).113 And the was regulated by the ion effect. Apart from the STP oxide tunnel junction provided a low energy consumption to characteristic demonstrated by the observation of PPF, LTP the device. characteristics like spike-number-dependent plasticity were 5.2. Sensorimotor System Mimicking. Toward the revealed, contributing to an asymptotic tension of the artificial needs of the disabled and the designation of smart robots as actuator with the increase of stimulation number, which is well as further mimicking the complex behaviors of life, analogous to biological muscle tension responses. establishment of artificial reflex arcs with synaptic devices came Except for sensorimotor mimics totally based on artificial into the sight of researchers. Like the biological reflex arcs devices, biological−artificial hybrid synaptic iontronic devices containing three major elements: receptors, neurons, and came into sight recently; Scott et al. designed a biohybrid

80 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review synapse based on PC-12 cells and organic transistors, and the chemical designation, biological understanding, and theoretical behaviors of dopamine neurons were mimicked by this assistance from computer science. These biomimicking devices biohybrid synapse.107 An organic-transistor-based postsynaptic bear the unlimited potential of an ideal platform for membrane received a dopamine signal released by PC-12 cells information storage and procession, implantable/wearable and showed typical synaptic behaviors. This biohybrid strategy devices, and the further realization of hardware artificial provided a new possibility for the designation of synaptic intelligence. iontronic devices. The well-designed ionic transistor provided a long-term neuroplasticity potential of the device with an ion ■ AUTHOR INFORMATION doping mechanism, and the electrolyte-gating nature of the Corresponding Authors device provided a possibility of introducing live cells to the Ping Yu − Beijing National Laboratory for Molecular Sciences, devices, which requires a precise and complex electrolyte-based Key Laboratory of Analytical Chemistry for Living Biosystems, environment. Institute of Chemistry, the Chinese Academy of Sciences (CAS), 6. CONCLUSION Beijing 100190, China; University of Chinese Academy of Sciences, Beijing 100049, China; orcid.org/0000-0002- Inspired by the signal transition in biological synapses, synaptic 6096-1933; Email: [email protected] iontronic devices have been designed and fabricated by Junjie Fei − Key Laboratory of Environmentally Friendly manipulating ions through doping, transporting them in Chemistry and Applications of Ministry of Education, College of confined space, and redox reaction. Various characteristics of Chemistry, Xiangtan University, Xiangtan 411105, China; biological synapses, including retention states, synaptic Email: [email protected] plasticity, and more complicated behaviors, were successfully emulated based on these iontronic devices. By further Authors integration with other devices, brain-inspired computation Changwei Li − Key Laboratory of Environmentally Friendly and artificial reflex arcs were constructed based on these Chemistry and Applications of Ministry of Education, College of synaptic iontronic devices. Although great progress had been Chemistry, Xiangtan University, Xiangtan 411105, China; made in this area over the past decade, a series of challenges Beijing National Laboratory for Molecular Sciences, Key still exists in this research area: (1) Although the synaptic Laboratory of Analytical Chemistry for Living Biosystems, iontronic devices could be integrated at the nm level by Institute of Chemistry, the Chinese Academy of Sciences (CAS), nanofabrication methods like chemical vapor deposition, the Beijing 100190, China problems of integration of synaptic iontronic devices, such as Tianyi Xiong − Beijing National Laboratory for Molecular cross-talk between artificial synapses and large scale integra- Sciences, Key Laboratory of Analytical Chemistry for Living tion, still existed. For a single iontronic device, the application Biosystems, Institute of Chemistry, the Chinese Academy of of new materials like nanowires or nanopores has the potential Sciences (CAS), Beijing 100190, China; University of Chinese of integration at a large scale, but the problems for the Academy of Sciences, Beijing 100049, China application of these devices also existed, such as long-term Lanqun Mao − Beijing National Laboratory for Molecular stability, EPSC fluctuation, environmental specification, and Sciences, Key Laboratory of Analytical Chemistry for Living response velocity. (2) There are still a series of unique Biosystems, Institute of Chemistry, the Chinese Academy of characteristics of ions that require further consideration for the Sciences (CAS), Beijing 100190, China; University of Chinese design of synaptic iontronic devices: First, cations and anions Academy of Sciences, Beijing 100049, China; orcid.org/ work separately as charge carriers in electrolyte solutions and 0000-0001-8286-9321 molten electrolytes, while electrons and holes recombined ff Complete contact information is available at: instead. In this case, precise and e ective regulation strategies https://pubs.acs.org/10.1021/acsabm.0c00806 should be developed based on this unique characteristic of free “ ” ions. Furthermore, ions were colorful and bear a series of Author Contributions ff ∥ di erent characteristics. These characteristics have the C.L. and T.X. contributed to this Review equally. potential of carrying more information, which has been observed in biological nervous systems. Therefore, how to Notes fi “read” and process the information with a variety of ions The authors declare no competing nancial interest. remains a great challenge. The combination of synaptic iontronic devices with other stimulation, such as optical ■ ACKNOWLEDGMENTS stimulation or chemical stimulation, has the potential of This work was supported by the Beijing National Natural dealing complex information carried by ions at the same time. Science Foundation (JQ19009), the National Natural Science (3) Although a series of applications based on these synaptic Foundation of China (21775151 for P.Y., 21790390, 21790391 iontronic devices have been realized, including brain- for L.M.), the National Basic Research Program of China mimicking computation and the formation of artificial neural (2018YFE0200800), and the Chinese Academy of Sciences circuits, most applications of these devices were limited in (QYZDJSSW-SLH030). basic model study, such as Pavlov experiments or pattern recognition. The realization of more complex neuronal ■ REFERENCES functions requires the rational integration of devices of large (1) Harris, J. J.; Jolivet, R.; Attwell, D. Synaptic Energy Use and scale as well as the assistance of computer science and Supply. Neuron 2012, 75 (5), 762−777. electronic engineering. (2) Wang, Z.; Joshi, S.; Savel’ev, S.; Song, W.; Midya, R.; Li, Y.; Rao, In summary, the fabrication and regulation of synaptic M.; Yan, P.; Asapu, S.; Zhuo, Y.; Jiang, H.; Lin, P.; Li, C.; Yoon, J.; iontronic devices is an interdisciplinary and fast-growing area. Upadhyay, N. K.; Zhang, J.; Hu, M.; Strachan, J. P.; Barnell, M.; Wu, Development of this area requires a deepening dialogue among Q.; Wu, H.; Williams, R. S.; Xia, Q.; Yang, J. Fully Memristive Neural

81 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review

Networks for Pattern Classification with Unsupervised Learning. Nat. (24) Han, J. H.; Kim, K. B.; Kim, H. C.; Chung, T. D. Ionic Circuits Electron. 2018, 1 (2), 137−145. Based on Polyelectrolyte Diodes on a Microchip. Angew. Chem., Int. (3) Xia, Q.; Yang, J. J. Memristive Crossbar Arrays for Brain-Inspired Ed. 2009, 48 (21), 3830−3833. Computing. Nat. Mater. 2019, 18 (4), 309−323. (25) Yang, C.; Suo, Z. Hydrogel Ionotronics. Nat. Rev. Mater. 2018, (4) Chai, Y.; Wu, Y.; Takei, K.; Chen, H.; Yu, S.; Chan, P. H.; Javey, 3 (6), 125−142. A.; Wong, H. S. P. Nanoscale Bipolar and Complementary Resistive (26) Kong, L.; Sun, J.; Qian, C.; Wang, C.; Yang, J.; Gao, Y. Switching Memory Based on Amorphous Carbon. IEEE Trans. Spatially-Correlated Neuron Transistors with Ion-Gel Gating for − Electron Devices 2011, 58 (11), 3933−3939. Brain-Inspired Applications. Org. Electron. 2017, 44,25 31. (5) Porro, S.; Accornero, E.; Pirri, C. F.; Ricciardi, C. Memristive (27) Kong, L.; Sun, J.; Qian, C.; Gou, G.; He, Y.; Yang, J.; Gao, Y. devices based on graphene oxide. Carbon 2015, 85, 383−396. Ion-gel Gated Field-Effect Transistors with Solution-Processed Oxide (6) Feng, P.; Xu, W.; Yang, Y.; Wan, X.; Shi, Y.; Wan, Q.; Zhao, J.; Semiconductors for Bioinspired Artificial Synapses. Org. Electron. − Cui, Z. Printed Neuromorphic Devices Based on Printed Carbon 2016, 39,64 70. Nanotube Thin-Film Transistors. Adv. Funct. Mater. 2017, 27 (5), (28) Park, Y.; Lee, J. S. Artificial Synapses with Short- and Long- Term Memory for Spiking Neural Networks Based on Renewable 1604447. − (7) Yin, L.; Han, C.; Zhang, Q.; Ni, Z.; Zhao, S.; Wang, K.; Li, D.; Materials. ACS Nano 2017, 11 (9), 8962 8969. Xu, M.; Wu, H.; Pi, X.; Yang, D. Synaptic Silicon-Nanocrystal (29) Chen, Y.; Liu, G.; Wang, C.; Zhang, W.; Li, R.; Wang, L. Polymer Memristor for Information Storage and Neuromorphic Phototransistors for Neuromorphic Computing. Nano Energy 2019, Applications. Mater. Horiz. 2014, 1 (5), 489−506. 63, 103859. (30) Zhang, C.; Tai, Y. T.; Shang, J.; Liu, G.; Wang, K.; Hsu, C.; Yi, (8) Zhang, K.; He, X.; Liu, Y.; Yu, P.; Fei, J.; Mao, L. Highly X.; Yang, X.; Xue, W.; Tan, H.; Guo, S.; Pan, L.; Li, R. W. Synaptic Selective Cerebral ATP Assay Based on Micrometer Scale Ion Plasticity and Learning Behaviours in Flexible Artificial Synapse Based Current Rectification at Polyimidazolium-Modified Micropipettes. − − on Polymer/Viologen System. J. Mater. Chem. C 2016, 4 (15), 3217 Anal. Chem. 2017, 89 (12), 6794 6799. 3223. (9) De Souza, R. A. Oxygen Diffusion in SrTiO3 and Related (31) Kong, L.; Sun, J.; Qian, C.; Fu, Y.; Wang, J.; Yang, J.; Gao, Y. − Perovskite Oxides. Adv. Funct. Mater. 2015, 25 (40), 6326 6342. Long-Term Synaptic Plasticity Simulated in Ionic Liquid/Polymer (10) Yang, J. M.; Choi, E. S.; Kim, S. Y.; Kim, J. H.; Park, J. H.; Park, Hybrid Electrolyte Gated Organic Transistors. Org. Electron. 2017, 47, N. G. Perovskite-Related (CH3NH3)3Sb2Br9 for Forming-Free 126−132. Memristor and Low-Energy-Consuming Neuromorphic Computing. (32) Zhang, P.; Xia, M.; Zhuge, F.; Zhou, Y.; Wang, Z.; Dong, B.; Nanoscale 2019, 11 (13), 6453−6461. Fu, Y.; Yang, K.; Li, Y.; He, Y.; Scheicher, R. H.; Miao, X. S. (11) Luo, Z.; Apachitei, G.; Yang, M.; Peters, J. J. P.; Sanchez, A. M.; Nanochannel-Based Transport in an Interfacial Memristor Can Alexe, M. Bi-Ferroic Memristive Properties of Multiferroic Tunnel Emulate the Analog Weight Modulation of Synapses. Nano Lett. Junctions. Appl. Phys. Lett. 2018, 112 (10), 102905. 2019, 19 (7), 4279−4286. (12) Hai, P. N.; Tanaka, M. Memristive Magnetic Tunnel Junctions (33) Kuzum, D.; Yu, S.; Wong, H. S. P. Synaptic Electronics: with MnAs Nanoparticles. Appl. Phys. Lett. 2015, 107 (12), 122404. Materials, Devices and Applications. Nanotechnology 2013, 24 (38), (13) Serb, A.; Bill, J.; Khiat, A.; Berdan, R.; Legenstein, R.; 382001. Prodromakis, T. Unsupervised Learning in Probabilistic Neural (34) He, Y.; Yang, Y.; Nie, S.; Liu, R.; Wan, Q. Electric-Double- Networks with Multi-State Metal-Oxide Memristive Synapses. Nat. Layer Transistors for Synaptic Devices and Neuromorphic Systems. J. Commun. 2016, 7, 12611. Mater. Chem. C 2018, 6 (20), 5336−5352. (14) Shockley, W. The Theory of P-N Junctions in Semiconductors (35) Yu, F.; Zhu, L. Ionotronic Neuromorphic Devices for Bionic and P-N Junction Transistors. Bell Syst. Tech. J. 1949, 28 (3), 435− Neural Network Applications. Phys. Status Solidi RRL 2019, 13 (6), 489. 1800674. (15) Strukov, D. B.; Snider, G. S.; Stewart, D. R.; Williams, R. S. The (36) Zhu, X.; Lee, S.; Lu, W. Nanoionic Resistive-Switching Devices. Missing Memristor Found. Nature 2008, 453 (7191), 80−83. Adv. Electron. Mater. 2019, 5 (9), 1900184. (16) Yang, J. J.; Strukov, D. B.; Stewart, D. R. Memristive Devices (37) Pereda, A. E. Electrical Synapses and Their Functional for Computing. Nat. Nanotechnol. 2013, 8 (1), 13−24. Interactions with Chemical Synapses. Nat. Rev. Neurosci. 2014, 15 − (17) Yang, J.; Hu, X.; Kong, X.; Jia, P.; Ji, D.; Quan, D.; Wang, L.; (4), 250 263. Wen, Q.; Lu, D.; Wu, J.; Jiang, L.; Guo, W. Photo-Induced Ultrafast (38) John, R. A.; Liu, F.; Chien, N. A.; Kulkarni, M. R.; Zhu, C.; Fu, Active Ion Transport through Graphene Oxide Membranes. Nat. Q.; Basu, A.; Liu, Z.; Mathews, N. Synergistic Gating of Electro-Iono- Commun. 2019, 10 (1), 1171. Photoactive 2d Chalcogenide Neuristors: Coexistence of Hebbian and (18) Perera, R. T.; Johnson, R. P.; Edwards, M. A.; White, H. S. Homeostatic Synaptic Metaplasticity. Adv. Mater. 2018, 30 (25), Effect of the Electric Double Layer on the Activation Energy of Ion 1800220. (39) Choquet, D.; Triller, A. The Dynamic Synapse. Neuron 2013, Transport in Conical Nanopores. J. Phys. Chem. C 2015, 119 (43), − − 80 (3), 691 703. 24299 24306. (40) McAllister, A. K.; Katz, L. C.; Lo, D. Neurotrophins and (19) Jubin, L.; Poggioli, A.; Siria, A.; Bocquet, L. Dramatic Pressure- Synaptic Plasticity. Annu. Rev. Neurosci. 1999, 22, 295−318. Sensitive Ion Conduction in Conical Nanopores. Proc. Natl. Acad. Sci. − (41) Feldman, D. E. Synaptic Mechanisms for Plasticity in U. S. A. 2018, 115 (16), 4063 4068. Neocortex. Annu. Rev. Neurosci. 2009, 32,33−55. (20) Lan, W.; Holden, D. A.; White, H. S. Pressure-Dependent Ion (42) Zucker, R. S.; Regehr, W. G. Short-term Synaptic Plasticity. Current Rectification in Conical-Shaped Glass Nanopores. J. Am. Annu. Rev. Physiol. 2002, 64, 355−405. − Chem. Soc. 2011, 133 (34), 13300 13303. (43) Varela, J. A.; Sen, K.; Gibson, J.; Fost, J.; Abbott, L. F.; Nelson, (21) Kumar, M.; Ban, D. K.; Kim, S. M.; Kim, J.; Wong, C. P. S. B. A Quantitative Description of Short-term Plasticity at Excitatory Vertically Aligned WS2 Layers for High-Performing Memristors and Synapses in Layer 2/3 of Rat Primary Visual Cortex. J. Neurosci. 1997, Artificial Synapses. Adv. Electron. Mater. 2019, 5 (10), 1900467. 17 (20), 7926−7940. (22) He, X.; Zhang, K.; Liu, Y.; Wu, F.; Yu, P.; Mao, L. Chaotropic (44) Tsodyks, M. V.; Markram, H. The Neural Code between Monovalent Anion-Induced Rectification Inversion at Nanopipettes Neocortical Pyramidal Neurons depends on Neurotransmitter Release Modified by Polyimidazolium Brushes. Angew. Chem., Int. Ed. 2018, Probability. Proc. Natl. Acad. Sci. U. S. A. 1997, 94 (2), 719−723. 57 (17), 4590−4593. (45) Dittman, J. S.; Kreitzer, A. C.; Regehr, W. G. Interplay between (23) Bocquet, L. Nanofluidics Coming of Age. Nat. Mater. 2020, 19 Facilitation, Depression, and Residual Calcium at Three Presynaptic (3), 254−256. Terminals. J. Neurosci. 2000, 20 (4), 1374−1385.

82 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review

(46) Bliss, T. V. P.; Lomo, T. Long-Lasting Potentiation of Synaptic (71) Alibart, F.; Pleutin, S.; Guerin,́ D.; Novembre, C.; Lenfant, S.; Transmission in Dentate Area of Anesthetized Rabbit Following Lmimouni, K.; Gamrat, C.; Vuillaume, D. An Organic Nanoparticle Stimulation of Perforant Path. J. Physiol. 1973, 232 (2), 331−356. Transistor Behaving as a Biological Spiking Synapse. Adv. Funct. (47) Liu, Y.; Zhu, L.; Feng, P.; Shi, Y.; Wan, Q. Freestanding Mater. 2010, 20 (2), 330−337. Artificial Synapses Based on Laterally Proton-Coupled Transistors on (72) Rivnay, J.; Inal, S.; Salleo, A.; Owens, R. M.; Berggren, M.; Chitosan Membranes. Adv. Mater. 2015, 27 (37), 5599−5604. Malliaras, G. G. Organic Electrochemical Transistors. Nat. Rev. Mater. (48) Kuzum, D.; Jeyasingh, R. G. D.; Lee, B.; Wong, H. S. P. 2018, 3 (2), 17086. Nanoelectronic Programmable Synapses Based on Phase Change (73) Bernards, D. A.; Malliaras, G. G. Steady-State and Transient Materials for Brain-Inspired Computing. Nano Lett. 2012, 12 (5), Behavior of Organic Electrochemical Transistors. Adv. Funct. Mater. 2179−2186. 2007, 17 (17), 3538−3544. (49) Abbott, L. F.; Regehr, W. G. Synaptic Computation. Nature (74) Yang, Y.; Wen, J.; Guo, L.; Wan, X.; Du, P.; Feng, P.; Shi, Y.; 2004, 431 (7010), 796−803. Wan, Q. Long-Term Synaptic Plasticity Emulated in Modified (50) Malinow, R.; Malenka, R. C. Ampa Receptor Trafficking and Graphene Oxide Electrolyte Gated Izo-Based Thin-Film Transistors. − Synaptic Plasticity. Annu. Rev. Neurosci. 2002, 25, 103 126. ACS Appl. Mater. Interfaces 2016, 8 (44), 30281−30286. (51)Fioravante,D.;Regehr,W.G.Short-TermFormsof (75) Oh, C.; Jo, M.; Son, J. All-Solid-State Synaptic Transistors with − Presynaptic Plasticity. Curr. Opin. Neurobiol. 2011, 21 (2), 269 274. High-Temperature Stability Using Proton Pump Gating of Strongly (52) Bear, M. F.; Malenka, R. C. Synaptic Plasticity: LTP and LTD. Correlated Materials. ACS Appl. Mater. Interfaces 2019, 11 (17), − Curr. Opin. Neurobiol. 1994, 4 (3), 389 399. 15733−15740. (53) Martin, S. J.; Grimwood, P. D.; Morris, R. G. M. Synaptic (76) Gkoupidenis, P.; Schaefer, N.; Garlan, B.; Malliaras, G. G. Plasticity and Memory: An Evaluation of the Hypothesis. Annu. Rev. Neuromorphic Functions in PEDOT:PSS Organic Electrochemical − Neurosci. 2000, 23, 649 711. Transistors. Adv. Mater. 2015, 27 (44), 7176. (54) Bliss, T. V. P.; Collingridge, G. L. A Synaptic Model of (77) Gkoupidenis, P.; Schaefer, N.; Strakosas, X.; Fairfield, J. A.; Memory-Long-Term Potentiation in the Hippocampus. Nature 1993, − Malliaras, G. G. Synaptic Plasticity Functions in an Organic 361 (6407), 31 39. Electrochemical Transistor. Appl. Phys. Lett. 2015, 107 (26), 263302. (55) Jo, S.; Chang, T.; Ebong, I.; Bhadviya, B. B.; Mazumder, P.; Lu, (78) Gerasimov, J. Y.; Gabrielsson, R.; Forchheimer, R.; Stavrinidou, W. Nanoscale Memristor Device as Synapse in Neuromorphic − E.; Simon, D. T.; Berggren, M.; Fabiano, S. An Evolvable Organic Systems. Nano Lett. 2010, 10 (4), 1297 1301. Electrochemical Transistor for Neuromorphic Applications. Adv. Sci. (56) Caporale, N.; Dan, Y. Spike Timing-Dependent Plasticity: A − 2019, 6 (7), 1801339. Hebbian Learning Rule. Annu. Rev. Neurosci. 2008, 31,25 46. (79) van de Burgt, Y.; Lubberman, E.; Fuller, E. J.; Keene, S. T.; (57) Bi, G.; Poo, M. Synaptic Modifications in Cultured Hippo- Faria, G. C.; Agarwal, S.; Marinella, M. J.; Alec Talin, A.; Salleo, A. A campal Neurons: Dependence on Spike Timing, Synaptic Strength, Non-Volatile Organic Electrochemical Device as a Low-Voltage and Postsynaptic Cell Type. J. Neurosci. 1998, 18 (24), 10464−10472. Artificial Synapse for Neuromorphic Computing. Nat. Mater. 2017, 16 (58) Bear, M. F.; Cooper, L. N.; Ebner, F. F. A Physiological-Basis (4), 414−418. for a Theory of Synapse Modification. Science 1987, 237 (4810), 42− (80) Qian, C.; Kong, L.; Yang, J.; Gao, Y.; Sun, J. Multi-Gate 48. Organic Neuron Transistors for Spatiotemporal Information Process- (59) Cooper, L. N.; Bear, M. F. Opinion the Bcm Theory of Synapse 2017 Modification at 30: Interaction of Theory with Experiment. Nat. Rev. ing. Appl. Phys. Lett. , 110 (8), 083302. (81) Gkoupidenis, P.; Koutsouras, D. A.; Malliaras, G. G. Neurosci. 2012, 13 (11), 798−810. (60) Wexler, K. A review of John R. Anderson's language, memory, Neuromorphic Device Architectures with Global Connectivity − through Electrolyte Gating. Nat. Commun. 2017, 8, 15448. and thought. Cognition 1978, 6, 327 351. ́ (61) Pavlov, P. I. Conditioned Reflexes: An Investigation of the (82) Bai, L.; Elosegui, C. G.; Li, W.; Yu, P.; Fei, J.; Mao, L. Biological Physiological Activity of the Cerebral Cortex. Ann. Neurosci. 2010, 17 Applications of Organic Electrochemical Transistors: Electrochemical (3), 136−141. Biosensors and Electro-physiology Recording. Front. Chem. 2019, 7, (62) Prezioso, M.; Merrikh-Bayat, F.; Hoskins, B. D.; Adam, G. C.; 313. Likharev, K. K.; Strukov, D. B. Training Andoperation of an (83) He, X.; Zhang, K.; Li, T.; Jiang, Y.; Yu, P.; Mao, L. Micrometer- Integrated Neuromorphic Network Based on Metal-Oxide Mem- Scale Ion Current Rectification at Polyelectrolyte Brush-Modified − ristors. Nature 2015, 521 (7550), 61−64. Micropipets. J. Am. Chem. Soc. 2017, 139 (4), 1396 1399. (63) Hu, L.; Fu, S.; Chen, Y.; Cao, H.; Liang, L.; Zhang, H.; Gao, J.; (84) Powell, M. R.; Sullivan, M.; Vlassiouk, I.; Constantin, D.; Sudre, Wang, J.; Zhuge, F. Ultrasensitive Memristive Synapses Based on O.; Martens, C. C.; Eisenberg, R. S.; Siwy, Z. S. Nanoprecipitation- − Lightly Oxidized Sulfide Films. Adv. Mater. 2017, 29 (24), 1606927. Assisted Ion Current Oscillations. Nat. Nanotechnol. 2008, 3 (1), 51 (64) Turrigiano, G. G.; Nelson, S. B. Homeostatic Plasticity in the 57. Developing Nervous System. Nat. Rev. Neurosci. 2004, 5 (2), 97−107. (85) Acar, E. T.; Buchsbaum, S. F.; Combs, C.; Fornasiero, F.; Siwy, (65) Turrigiano, G. G.; Leslie, K. R.; Desai, N. S.; Rutherford, L. C.; Z. S. Biomimetic Potassium-Selective Nanopores. Sci. Adv. 2019, 5 Nelson, S. B. Activity-Dependent Scaling of Quantal Amplitude in (2), No. eaav2568. Neocortical Neurons. Nature 1998, 391 (6670), 892−896. (86) Luo, L.; Holden, D. A.; White, H. S. Negative Differential (66) Abbott, L. F.; Nelson, S. B. Synaptic Plasticity: Taming the Electrolyte Resistance in a Solid-State Nanopore Resulting from Beast. Nat. Neurosci. 2000, 3 (11), 1178−1183. Electroosmotic Flow Bistability. ACS Nano 2014, 8 (3), 3023−3030. (67) Ho, V. M.; Lee, J.; Martin, K. The Cell Biology of Synaptic (87) Xiong, T.; Zhang, K.; Jiang, Y.; Yu, P.; Mao, L. Ion Current Plasticity. Science 2011, 334 (6056), 623−628. Rectification: From Nanoscale to Microscale. Sci. China: Chem. 2019, (68) Du, C.; Ma, W.; Chang, T.; Sheridan, P.; Lu, W. Biorealistic 62 (10), 1346−1359. Implementation of Synaptic Functions with Oxide Memristors (88) Prakash, S.; Conlisk, A. T. Field Effect Nanofluidics. Lab Chip through Internal Ionic Dynamics. Adv. Funct. Mater. 2015, 25 (27), 2016, 16 (20), 3855−3865. 4290−4299. (89) Li, T.; He, X.; Yu, P.; Mao, L. A Bioinspired Light-Controlled (69) Zhu, L.; Wan, C.; Guo, L.; Shi, Y.; Wan, Q. Artificial Synapse Ionic Switch Based on Nanopipettes. Electroanalysis 2015, 27 (4), Network on Inorganic Proton Conductor for Neuromorphic Systems. 879−883. Nat. Commun. 2014, 5 (1), 3158. (90) Xie, G.; Li, P.; Zhao, Z.; Zhu, Z.; Kong, X.; Zhang, Z.; Xiao, K.; (70) Shi, J.; Ha, S.; Zhou, Y.; Schoofs, F.; Ramanathan, S. A Wen, L.; Jiang, L. Light- and Electric-Field-Controlled Wetting Correlated Nickelate Synaptic Transistor. Nat. Commun. 2013, 4, Behavior in Nanochannels for Regulating Nanoconfined Mass 2676. Transport. J. Am. Chem. Soc. 2018, 140 (13), 4552−4559.

83 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84 ACS Applied Bio Materials www.acsabm.org Review

(91) Lucas, R. A.; Lin, C.; Baker, L. A.; Siwy, Z. S. Ionic Amplifying (111) Mazumder, P.; Kang, S.; Waser, R. Memristors: Devices, Circuits Inspired by Electronics and Biology. Nat. Commun. 2020, 11 Models, and Applications. Proc. IEEE 2012, 100 (6), 1911−1919. (1), 1568. (112) Yang, J.; Ge, C.; Du, J.; Huang, H.; He, M.; Wang, C.; Lu, H.; (92) Vlassiouk, I.; Siwy, Z. S. Nanofluidic Diode. Nano Lett. 2007, 7 Yang, G.; Jin, K. Artificial Synapses Emulated by an Electrolyte-Gated (3), 552−556. Tungsten-Oxide Transistor. Adv. Mater. 2018, 30 (34), 1801548. (93) Plett, T. S.; Cai, W.; Le Thai, M.; Vlassiouk, I. V.; Penner, R. (113) Li, J.; Ge, C.; Lu, H.; Guo, H.; Guo, E.; He, M.; Wang, C.; M.; Siwy, Z. S. Solid-State Ionic Diodes Demonstrated in Conical Yang, G.; Jin, K. Energy-Efficient Artificial Synapses Based on Oxide Nanopores. J. Phys. Chem. C 2017, 121 (11), 6170−6176. Tunnel Junctions. ACS Appl. Mater. Interfaces 2019, 11 (46), 43473− (94) Nam, S. W.; Rooks, M. J.; Kim, K. B.; Rossnagel, S. M. Ionic 43479. Field Effect Transistors with Sub-10 nm Multiple Nanopores. Nano (114) Lee, Y.; Oh, J. Y.; Xu, W.; Kim, O.; Kim, T. R.; Kang, J.; Kim, Lett. 2009, 9 (5), 2044−2048. Y.; Son, D.; Tok, J. B. H.; Park, M. J.; Bao, Z.; Lee, T. W. Stretchable (95) Lu, S.; Peng, Y.; Ying, Y.; Long, Y. Electrochemical Sensing at a Organic Optoelectronic Sensorimotor Synapse. Sci. Adv. 2018, 4 (11), Confined Space. Anal. Chem. 2020, 92 (8), 5621−5644. No. eaat7387. (96) Ying, Y.; Long, Y. Nanopore-Based Single-Biomolecule (115) Yu, C.; Guo, H.; Cui, K.; Li, X.; Ye, Y.; Kurokawa, T.; Gong, J. Interfaces: From Information to Knowledge. J. Am. Chem. Soc. Hydrogels as Dynamic Memory with Forgetting Ability. Proc. Natl. − 2019, 141 (40), 15720−15729. Acad. Sci. U.S.A. 2020, 117, 18962 18968. (97) Lee, J.; Du, C.; Sun, K.; Kioupakis, E.; Lu, W. Tuning Ionic Transport in Memristive Devices by Graphene with Engineered Nanopores. ACS Nano 2016, 10 (3), 3571−3579. (98) Najem, J. S.; Taylor, G. J.; Weiss, R. J.; Hasan, M. S.; Rose, G.; Schuman, C. D.; Belianinov, A.; Collier, C. P.; Sarles, S. A. Memristive Ion Channel-Doped Biomembranes as Synaptic Mimics. ACS Nano 2018, 12 (5), 4702−4711. (99) Wang, D.; Kvetny, M.; Liu, J.; Brown, W.; Li, Y.; Wang, G. Transmembrane Potential across Single Conical Nanopores and Resulting Memristive and Memcapacitive Ion Transport. J. Am. Chem. Soc. 2012, 134 (8), 3651−3654. (100) Wang, Z.; Wang, L.; Nagai, M.; Xie, L.; Yi, M.; Huang, W. Nanoionics-Enabled Memristive Devices: Strategies and Materials for Neuromorphic Applications. Adv. Electron. Mater. 2017, 3 (7), 1600510. (101) Yang, J.; Pickett, M. D.; Li, X. M.; Ohlberg, D. A. A.; Stewart, D. R.; Williams, R. S. Memristive Switching Mechanism for Metal/ Oxide/Metal Nanodevices. Nat. Nanotechnol. 2008, 3 (7), 429−433. (102) Yang, J. J.; Miao, F.; Pickett, M. D.; Ohlberg, D. A. A.; Stewart, D. R.; Lau, C. N.; Williams, R. S. The Mechanism of Electroforming of Metal Oxide Memristive Switches. Nanotechnology 2009, 20 (21), 215201. (103) Kamiya, K.; Young Yang, M. Y.; Park, S. G.; Magyari-Kope, B.; Nishi, Y.; Niwa, M.; Shiraishi, K. On-Off Switching Mechanism of Resistive-Random-Access-Memories Based on the Formation and Disruption of Oxygen Vacancy Conducting Channels. Appl. Phys. Lett. 2012, 100 (7), 073502. (104) Mishra, R.; Kumar, D.; Yang, H. Oxygen-Migration-Based Spintronic Device Emulating a Biological Synapse. Phys. Rev. Appl. 2019, 11 (5), 054065. (105) Zhang, X.; Liu, S.; Zhao, X.; Wu, F.; Wu, Q.; Wang, W.; Cao, R.; Fang, Y.; Lv, H.; Long, S.; Liu, Q.; Liu, M. Emulating Short-Term and Long-Term Plasticity of Bio-Synapse Based on Cu/a-Si/Pt Memristor. IEEE Electron Device Lett. 2017, 38 (9), 1208−1211. (106) Wang, Z.; Joshi, S.; Savel’ev, S. E.; Jiang, H.; Midya, R.; Lin, P.; Hu, M.; Ge, N.; Strachan, J. P.; Li, Z.; Wu, Q.; Barnell, M.; Li, G.; Xin, H.; Williams, R. S.; Xia, Q.; Yang, J. Memristors with Diffusive Dynamics as Synaptic Emulators for Neuromorphic Computing. Nat. Mater. 2017, 16 (1), 101−108. (107) Keene, S. T.; Lubrano, C.; Kazemzadeh, S.; Melianas, A.; Tuchman, Y.; Polino, G.; Scognamiglio, P.; Cina, L.; Salleo, A.; van de Burgt, Y.; Santoro, F. A Biohybrid Synapse with Neurotransmitter- mediated Plasticity. Nat. Mater. 2020, 19, 969−973. (108) Xu, W.; Min, S.; Hwang, H.; Lee, T. W. Organic Core-Sheath Nanowire Artificial Synapses with Femtojoule Energy Consumption. Sci. Adv. 2016, 2 (6), No. e1501326. (109) Pi, S.; Li, C.; Jiang, H.; Xia, W.; Xin, H.; Yang, J.; Xia, Q. Memristor Crossbar Arrays with 6-nm Half-Pitch and 2-nm Critical Dimension. Nat. Nanotechnol. 2019, 14 (1), 35−39. (110) Lee, Y.; Lee, T. W. Organic Synapses for Neuromorphic Electronics: From Brain-Inspired Computing to Sensorimotor Nervetronics. Acc. Chem. Res. 2019, 52 (4), 964−974.

84 https://dx.doi.org/10.1021/acsabm.0c00806 ACS Appl. Bio Mater. 2021, 4, 71−84