Evolutionary 2D organic crystals for optoelectronic transistors and neuromorphic computing

Brain-inspired neuromorphic computing has been extensively researched, taking advantage of increased computer power, the acquisition of massive data, and algorithm optimization. Neuromorphic computing requires mimicking synaptic plasticity and enables near-in-sensor computing. In synaptic transistors, how to elaborate and examine the link between microstructure and characteristics is a major difficulty. Due to the absence of interlayer shielding effects, defect-free interfaces, and wide spectrum responses, reducing the thickness of organic crystals to the 2D limit has a lot of application possibilities in this computing paradigm. This paper presents an update on the progress of 2D organic crystal-based transistors for data storage and neuromorphic computing. The promises and synthesis methodologies of 2D organic crystals (2D OCs) are summarized. Following that, applications of 2D OCs for ferroelectric non-volatile memory, circuit-type optoelectronic synapses, and neuromorphic computing are addressed. Finally, new insights and challenges for the field’s future prospects are presented, pushing the boundaries of neuromorphic computing even farther.


Introduction
Digital technology has advanced in leaps and bounds in recent decades, thanks in large part to increased computer power, the capture of huge data, and algorithm optimization.Despite energy-efficient algorithms for data-centric computational tasks, modern computer architecture is still based on the traditional von Neumann configuration, in which pre-processed data is shuttled between the physically separated central processing unit and memory unit, resulting in unexpected memory wall, time delay, and heat [1,2].Some intriguing technologies dubbed neuromorphic computing and in-memory computing have been proposed and well developed, inspired by computational paradigms with respect to human brains.Neuromorphic computing, as the name implies, is a technique that aims to erase the physical gap between the data-computing unit and the dataprocessing unit, implying that computing takes place where the data is stored.The development of a brain-like model using neurons and synapses is known as neuromorphic computing.Synapses are thought to be the information carriers in biological brain networks [3].Materials scientists and electronic engineers have extensively simulated synaptic plasticity, or the ability to modulate the connection in the synaptic cleft between neurons, using neuromorphic devices.
As an alternative building block, neuromorphic transistors developed from the basic configuration of floating gate flash memory have been explored to linearly control the synaptic weight updating and variation since 1995 [4].Benefiting from the low circuit delay, maturity of quantum tunnelling mechanism and compatibility with current CMOS fabrication process, neuromorphic transistors are close to mass-produced in hardwareimplemented commercialization with low fabrication cost and high integration density, making it possible to perform a data-heavy perception task [5,6].Besides, optical modulation assisted by the gate electrical terminal has been introduced in the conventional transistors as a remote controlling terminal and endowed novel light-programming/electrical-erasing function thanks to its ultrahigh spatial-temporal resolution, enabling and biomimicry of optoelectronic synapse with less energy loss.Finally, despite the tantalizing feasibilities for performance advancements of two-terminal devices, three-terminal have advantages in writing time, ON-OFF ratio without considerations of atomic rearrangement nature and high programming current limitation on the continuous downscaling trend [7,8].
To date, optoelectronic transistors have been reported as promising artificial synapses based on diverse materials systems, including inorganic perovskites, amorphous oxide, biological materials, organic materials, and two-dimensional materials [9][10][11][12][13][14][15].Among the aforementioned materials systems, organic-based transistors have a unique combination of intriguing properties, including wide spectrum response for photoelectrical conversion, tunable molecular structures, low-cost fabrication techniques, flexibility, and abundant deposition conditions [16][17][18][19][20][21][22][23][24][25].There are still issues that must be addressed in order to accelerate the further applications of organic-based optoelectronic synapses for revolutionary neuromorphic computing systems: (i) the internal physical mechanism of charge transport and structure-property relationship of organic molecules with complex microstructures [24,26,27]; (ii) the advancement of reliability and stability under different conditions; (iii) the novel functionalities of organic-based transistors.
2D organic crystals (2D OCs) with adjustable layer numbers have significant advantages over thicker bulk organic materials in research of the aforementioned fundamental challenges.Several layers or perhaps a monolayer of organic molecules are bound by weak van der Waals forces in the microstructures of 2D OCs [28].Because there are no known 'lattice mismatch' difficulties, it is easy to build a vertical heterostructure using multiple 2D layered materials [29,30].The ubiquitous interlayer screening effect that exists in organic bulk materials can be significantly disappeared when the thickness of organic crystals is reduced to less than 10 nm.Furthermore, 2D OCs are attractive candidates for investigating charge transport behaviour in relation to crystal orientations, polaronic relaxation, and layer counts [31,32].At the microscopic scale, the interfacial characteristics and charge carrier behaviour might differ from one layer to the next [33][34][35].These competitive characteristics, when combined with the intrinsic qualities of general organic materials, equip 2D OCs with special functions for enhancing future neuromorphic computing systems.Up to now, there are many works aiming at controlling the thickness of organic crystals to a few layers with good crystallinity and enriched OCs applied in this emerging field (figure 1) [32,[36][37][38][39][40].However, a comprehensive review of highly crystalline 2D organic semiconductors (OSCs) combining the optoelectronic properties and neuromorphic computing applications is still unavailable.Therefore, a recent review consisting 2D OSCs synthesis strategies, promises, characterization, and high-performance neuromorphic devices and their applications in artificial sensory system is not only scientifically important but also urgent to present the development trend in this interdisciplinary field.
In this review, we summarize the recent advancements in data-storage devices and neuromorphic building units based on 2D OCs, which shed light on the superior performance of organic-based transistors at the 2D limit.Then, the state-of-the-art 2D fabrication protocols are explained in detail, with particular emphasis on the synthesis strategies for large scale, high quality and monolayer 2D OCs.Following that, we look at current developments in high-performance ferroelectric non-volatile memory, circuit-type optoelectronic synapse, and neuromorphic computing.Finally, we will discuss the significant difficulties that need to be addressed as well as future research prospects in optoelectronic synapses based on 2D OCs for artificial intelligence systems, such as near-sensor and in-sensor computing.

Promises of 2D organic crystals
Because of their wide spectral range of optoelectronic responsiveness, flexibility, and vast area processability, OSCs are interesting options for manufacturing high-performance optoelectronic devices and circuits [33,[40][41][42][43]. Reducing the thickness of organic crystals to a few layers or even a single monolayer improves performance in a variety of ways: (i) reduce contact resistance and Schottky barrier so that an ohmic contact can be achieved, as depicted in figures 2(c) and (d) [44][45][46][47], (ii) increase carrier mobility, as shown in figures 2(a) and (b) [45,[48][49][50][51][52][53][54][55][56], and (iii) achieve controllable growth [57].Apart from the electrical advancements of transistors, obvious enhancement of internal photo-response in the application of 2D OCs-based optoelectronic transistor compared with that of the same bulk crystals, as shown in figures 2(e) and (f) [48].As presented in figure 2(b), it is notable that thickness modulation becomes negligible beyond some specific conductive layer, leading to a mobility and I on /I off ratio plateau once the percolation is well-formed.It can be concluded that the thickness-dependent performance is related to a bunch of factors, to name a few, including the domain size, crystal orientations, excessive disorders and interfacial trap sites, etc.All these parameters would influence the morphology of dielectric interface.Hence, it is necessary to provide a 2D-based platform for probing carrier transport and structure-property relationship at the low-dimension scale.

High mobility
He et al used van der Waals epitaxy to fabricate a high-quality few layer dioctylbenzothienobenzothiophene (C 8 -BTBT) growing on graphene or boron nitride without dangling bonds, ensuring quasi-freestanding with low disturbance and a pristine interface for constructing high-performance organic transistors [58].Furthermore, because the crystals' arrangement does not favour carrier transport in the vertical direction, the authors found a very low tunnelling current.Due to the immaculate interface and finite density of charge traps on the h-BN substrate, C 8 -BTBT mobility can still reach up to 10 cm 2 V −1 s −1 despite contact resistance.As shown in figure 3(a), researchers were motivated by the floating-coffee-ring effect that a good solvent on an incompatible antisolvent near the solution edge can have on the growth of extremely uniform C 8 -BTBT at a rapid rate.Wang et al [59] explored this technique to construct C 8 -BTBT-based transistors with increased maximum mobility of 13.0 cm 2 V −1 s −1 (figures 3(b) and (c)).Despite the great mobility given by the aforementioned solution-based processes, producing defect-free ultrathin films without thermal fractures and grain boundaries remains a challenge.These tiny crystal domains and thermally generated defects serve as charge carrier trapping sites, resulting in a device-to-device variance.To eliminate the morphological defects and produce preferential crystal orientation in the monolayer crystal of 2,9-didecyldinaphtho [2,3-b: 2 ,3 -f]thieno [3,2-b]thiophene (C 10 -DNTT), Zhou used a hybrid deposition strategy that combined regular thermal evaporation and ultraslow shearing [60].As depicted in figures 3(d) and (e), the scientists used the ultraslow shearing approach to create a homogeneous millimeter-sized monolayer C 10 -DNTT by specifically fixing the nucleation site and further growing the crystals with a given orientation.The resulting highly ordered monolayer crystal was then used as a template, resulting in the formation of another C 10 -DNTT layer by heat evaporation.Because of the interfacial-interaction, the top C 10 -DNTT layer exhibited the same orientation as the template with the help of the initial monolayer template.Among 45 devices, the resulting device had an averaged mobility of 14.7 cm 2 V −1 s −1 .Compared to that of the referenced thermal evaporation devices, a 100% augment of the average carrier mobility is observed (figure 3(f)).As depicted in figures 3(g) and (h), the output characteristics show that the hybrid thermal evaporation strategy outperforms the other referenced thermal evaporation method.The carrier mobility shows no significant dependence on the gate voltage (figure 3(i)).These inspiring results indicates such a strategy can provide a novel way for well-balanced device performance.Makita et al [61] presented a unique water exfoliation process for producing a virtually flawless single crystal of 3,11-dinonyldinaphtho [2,3-d: 20,30-d0]benzo[1,2-b: 4,5-b0]dithiophene (C 9 -DNBDT-NW).The difference in surface energy can explain such an approach.The hydrophobic alkyl chain in the 2D small molecule causes water infiltration, while the mica substrate has a super-hydrophilic surface with a contact angle of water of 3 • , resulting in water infiltration and an ultrathin 2D C 9 -DNBDT-NW film exfoliation.The destination substrate should be any template substrate with a super-hydrophilic surface in this case.The gadget that was manufactured after that had a high mobility of up to 12 cm 2 V −1 s −1 and an exceptional reliability of higher than 90% with no visible depreciation.

Low contact resistance
Another advantage of reducing the thickness of organic crystals is that it boosts the efficiency of carrier injection and modulation from metal electrodes, which can aid in the discovery of the internal carrier transportation mechanism [64,65].The use of low-resistance ohmic contact shorten device switching times while also lowers operational voltage and power consumption [66,67].The use of a dopant layer at the interface, modifying the work function of metal by introducing a self-assembled monolayer, energy band engineering, and optimizing production procedures have all been examined as ways to reduce contact resistance.He et al [66], for example, used a monolayer C 8 -BTBT channel layer and a prepatterned gold electrode to obtain prominent hole mobility (>30 cm 2 V −1 s −1 ) and an ultralow contact resistance of 0.1 kΩ•cm.The authors also discovered that in-plane carrier transportation has significantly higher mobility than out-of-plane carrier transportation.As a result, the authors indicated that avoiding vertical transit and achieving direct intimate contact with the channel layer are required to reduce contact resistance, which is difficult to do with traditional organic bulk materials.Contact resistance in organic field-effect transistors is divided into two parts: interfacial contact resistance (R c , int ) and bulk injection resistance (R c,bulk ).While R c,int has been studied in recent years, strategies for lowering R c,bulk are rarely reported due to difficulties of finding a balance between thickness and R c,bulk .In concept, inter-layer transport is thought to be substantially weaker than intra-layer transport, contributing to a high R c,bulk , and inefficient charge injection.Jiang et al used a drop-casting synthesis process to construct layered 1,4-bis((50-hexyl-2,20-bithiophen-5-yl)ethynyl)benzene (HTEB) molecular crystals and used them as a semiconducting layer in transistors [52].The contact resistance of different thicknesses of HTEB was examined by the authors, who discovered that contact resistance is proportional to thickness.The authors achieved an ultralow contact resistance of 540 Ω•cm by fabricating HTEB monolayer-based transistors with a 30 nm HfO 2 dielectric layer.Because of the abundance of defects and charge trapping sites, OFETs based on a monolayer channel may have poorer device performance than thicker transistors.Despite the fact that the functional layer is a defect-free monolayer, there is still the possibility of developing defects and trap sites due to the high-energy penetration of metal atoms and the deformation of the molecular packing caused by typical thermal evaporation.As shown in figure 4(a), Peng and his colleagues used a solution-processed monolayer of C 10 -DNTT as the semiconducting layer of the transistors and transferred gold electrodes as the source/drain electrodes, motivated by vanishing defects and trap sites [62].In a tiny V DS biased condition, the combination of nondestructive transferred gold pads and molecular flatness of the semiconductor layer results in a low contact resistance of 40 Ω•cm, as elaborated in figure 4(b).More crucially, the C 10 -DNTT-based transistor had perfect internal gain in the saturation domain, owing to the monolayer C 10 -DNTT's good pinch-off behaviour (figure 4(c)).With the intrinsically higher gm than that of the bulk materials, the organic monolayer-based transistor are competitive candidates in improving inherent gain of devices during the dimensional scaling process.The device performance based on evaporated gold atoms is hampered by the high Fermi level pinning (FLP) effect and the introduction of undesirable gap states, in addition to introducing distortion of molecular packing.To address the issues, numerous investigations have been done.Chemical doping with organic compounds adjusting the thickness of an extra layer to achieve low contact resistance are just a few examples.Yan et al [63] suggested a ground-breaking technology to prominently attenuate the FLP impact and minimize Schottky barrier height (figure 4(e)).As shown in figure 4(d), the Fermi level of 1T-TaSe 2 is close to the LUMO level of 2D F 16 CuPc nanoflake, forming a low Schottky barrier height and ensuring efficient carrier injection.The transfer characteristics of the device with 1T-TaSe 2 and evaporated Au serving as electrodes are presented in figure 4(f).It is noteworthy that the on-state current of device with 1T-TaSe 2 contact is improved by more than three-fold compared to that of the referenced devices with evaporated Au contact and the threshold voltage is smaller than the referenced device based on evaporated Au contact.Then the extracted Schottky barrier height of 1T-TaSe 2 -based device is presented in the figure 4(g), which is the lowest value among the organic-based transistors.According to the relations between calculated trap states energy ET and reduction ratios of ET under different V g (figure 4(h)), the authors proposed that the mid-gap states can be suppressed thanks to the existence of the nonbonding van der Waals interface and the density of trap sites can be reduced since the 1T-TaSe 2 can prevent the semiconducting layer from heat damage and metal percolation.Carriers are easily escaped from the trap sites in the 1T-TaSe 2 -based device with reduced trap state energy, as elaborated in figure 4(i).The authors introduced as the source-drain electrodes in 2D F 16 CuPc-based n-type OFETs to restrain the generation of mid-gap states and lower the Schottky barrier height at the interface, which opens a new path to reduce contact resistance from an energy band perspective.

Controllable growth
Organic crystals are known for their self-assembling properties from monomers, which have been extensively studied in recent years for the following reasons: (i) a bottom-up synthesis strategy for precisely controlled growth, (ii) high quality with a low ratio of trap sites to grain boundaries [68][69][70].The regulated expansion of size and thickness dimension are the remaining barriers to advancing promising applications.For example, using the self-limited organic molecular beam epitaxy method, the strong and tunable van der Waals interaction between graphene and C 8 -BTBT organic molecules easily drives the ordered stacking of monomers into a well-defined layer with molecular flatness and precise thickness [71].Yamamura et al [72] presented that an improved meniscus-driven solution-based synthesis technique can boost scaling-up of homogeneous ultrathin films named as 3,11-dioctyldinaphtho [2,3- The edge of the meniscus enhances the evaporation of solvent, which propels the nucleation of crystals from a supersaturated zone in the proposed approaches.The authors also confirmed that the centimeter-sized C 8 -DNBDT-NW single crystal is large enough to cover hundreds of transistors.Besides, the authors exploited the C 8 -DNBDT-NW-based device to realize the rectification characteristics via a simple circuit diagram.Technically, the rate of solute deposition is directly proportional to the amount of solvent evaporation.As a result, process temperature, shearing rate, and solution concentration were tuned to precisely regulate the quantity of monolayer to multilayer single-crystalline organic layers with wafer-scale coverage.The size of crystals is mostly impeded during the growth process based on solution-epitaxial self-assembly methods by the poor dispersion of organic molecules on the water surface and the weak connections between the organic molecules.Wang et al [73] revealed another self-assembly approach generated by graphene quantum dots (GQDs) for a centimeter-sized 2,7-didecyl benzothienobenzothiopene (C 10 -BTBT) monolayer, as presented in figure 5(a).Figure 5(b) depicts that the molecular stacking characteristics can be tuned by different PH values of GQD solutions since the π-π packing interactions between the molecules and the GQDs is much stronger than that of the molecular interlayers.Based on numerous characteristics, the approach takes advantage of both the benefits of self-assembly on water surfaces and the epitaxial growth induced by graphene: (i) the introduction of GQDs facilitates the spreading of C 10 -BTBT molecules on the water surface by modulating the PH value of the solution; (ii) the nucleation energy can be significantly reduced, resulting in a strong cohesive force to aggregate the molecules together thanks to stacking interactions (figure 5(c)), which opens the door to large-area growth of monolayer or even multilayer organic crystals.

Fabrication of 2D organic crystals
In the realm of organic electronics, the concept of 2D molecular crystals, which is essentially a single crystal molecule with molecular-level thickness that is connected through intermolecular van der Waals contacts, has evolved in recent years.2D OCs are ideal semiconductor materials in the manufacture of electronic devices because they lack grain boundaries, have fewer flaws, and have a long-range ordered crystal structure and intrinsic properties [74].As a result, we need to create simple and effective organic crystal development methods as well as suitable crystal growth methods to prepare high-quality single crystals.Many preparation strategies have been proposed since the definition of 2D OCs (figure 6) [19,59,69,[75][76][77][78][79].The procedures can be classified into two groups based on the condition of matter: gas-phase preparation and liquid-phase preparation.Each method has its own set of benefits and disadvantages.When it comes to developing highquality organic single crystals, it is vital to consider the physical and chemical qualities of different materials, as well as crystallization properties and application needs.The fabrication of 2D OCs in the gas phase is a relatively established procedure.The vapor-phase method of growing single crystals produces superior quality and can be utilized for materials with limited solubility, but it has higher material thermal stability and growth equipment requirements, as well as lower yields.Physical vapour transport (PVT) has been used to create organic crystals since the 1990s, and it is now one of the most frequent ways for making micro-nano crystals from insoluble OSC materials [80].The material is sublimated in the high-temperature zone, then transported to the condensation zone to become saturated vapour, which will develop into crystals after condensation and nucleation.The ratio of the growth vapour pressure in the low temperature area to the equilibrium vapour pressure is the driving factor for crystal growth.Many other factors influence the crystal growth process, including the raw material's sublimation temperature (T s ), the crystal growth temperature (T c ), the substrate's surface properties, the placement position, the temperature gradient in the tube, the system pressure, the carrier gas purity and speed, and the raw material's dosage and purity, among other things.As a result, by altering these parameters, we can generate a variety of crystal shapes and morphologies.Temperature is usually the most effective and controlled adjusting element.Park et al [75] used the PVD process to modify the sublimation temperature of the pentacene raw material, but all other parameters stayed the same.1D wire and 2D disk morphologies may be achieved at 350 • C and 325 • C, respectively, as shown in figures 7(b) and (c).The temperature is near to the sublimation point, according to the experiment.It is conducive to the production of a thermodynamically balanced crystal morphology (2D disk) at that time; it is conducive to the formation of a 1D wire morphology at a higher sublimation temperature.Although the PVT method produces high-quality crystals with controllable crystal shape and morphology by altering process parameters, the raw material utilization rate is often low, vacuum and carrier gas are required, and the cost is expensive.
The liquid-phase preparation method excels in the development of high-quality organic crystals due to the structural features of organic molecules [81,82].Not only does the liquid-phase preparation process require little equipment and mild temperatures, but it may also be done in a vast area at once.Dip coating [79,[83][84][85][86], spin coating [77,[87][88][89][90][91], Langmuir-Blodgett (LB) self-assembly [92][93][94], drop casting [69,95,96], pen writing [25,97], solution epitaxial growth [78,98], floating-coffee-ring-driven assembly [59], solution shearing [19,99,100], inkjet printing [101], and other liquid phase preparation methods are among them.Non-directional liquid phase preparation methods include spin coating and drop casting [82], which are both widely used liquid phase preparation methods.Dropping the solution onto the substrate's surface is the instillation method.The substance naturally assembles into a crystal due to the solvent's inherent volatilization.The gradual self-assembly process allows for the ordered arrangement of molecules, the formation of a thermodynamically stable stacked structure, and the fabrication of the finest quality single crystal or crystal film.For the first time, Jiang et al [69] used a solution drip assembly method to create a 2D organic single crystal with a configurable number of layers.The thickness of HTEB can be decreased to a single layer while preserving an excellent long-range ordered arrangement.This 2D single crystal can also be applied to any process substrate.The spin-coating approach, as opposed to the drip method, can acquire the molecular arrangement faster, the solvent volatilizes faster, and it saves time.The thickness of the film is commonly controlled in the traditional spin coating technique by altering the rotation speed or solution concentration during spin coating.Zhang et al [77] optimized the spin-coating process by distributing the solution while the spincoating machine motor was running at high speed, allowing for perfect control of the semiconductor layer thickness.The high-speed camera clearly shows that the solution generates an ultra-thin layer in 0.03 s during spin-coating, as shown in figure 7(d).The film thickness of the NDI(2OD)(4tBuPh)-DTYM2 film thins as the solution concentration falls at a fixed speed of 7000 rpm, and may be regulated to a single layer with a thickness of roughly 2 nm.Experiments (see figure 7(e)) have shown that this tweaked spin coating process is capable of ultra-thin films of various conjugated polymers and small molecule semiconductors on both hydrophobic and hydrophilic substrates.
Uncontrolled migration of organic molecules in the solvent and random crystallization behaviour will result in inescapable holes, flaws, and grain boundaries in the crystalline film, and these crystalline features would substantially inhibit crystal liquid phase growth [114].Exploration of device performance and research into the intrinsic features of OSCs are highly needed.Wang et al [59] suggested a new liquid phase preparation method called 'floating coffee ring effect' that integrated the coffee ring effect with anti-solvent-assisted crystallization technology design.The OSC C 8 -BTBT was dissolved in a mixed solvent containing anisole as the main solvent and a minor amount of p-anisaldehyde as the anti-solvent in the experiment.A little amount of solution is dripped onto the SiO 2 /Si substrate, and the liquid is dragged at a high speed in one direction on the substrate by the air created by the mechanical pump.The air next to the droplet surface and relative to the solvent atmosphere is unsaturated as the droplet dragged by the airflow moves swiftly on the substrate, resulting in a quicker evaporation rate and thinner liquid at the droplet's edge.The morphology may be obtained simultaneously, and the C 8 -BTBT molecules can fully travel to the droplet's edge to form a 2D film with a diameter of around 220 μm.The production of a floating coffee ring effect on the anti-solvent at the droplet's edge is caused by the flow of the main solvent driven by volatilization, and such a liquid-liquid interface is an appropriate platform for C 8 -BTBT molecules to unite to create a film (see figures 8(a)-(c)).The solution shear method, as a representative of the meniscus deposition method, can efficiently regulate the arrangement of molecules in the crystalline film, and the resulting films have outstanding electrical properties [82].The solution shearing process involves placing the top shearing blade and the bottom fixed substrate at an inclination angle with the growth solution in between, and moving the top shearing blade horizontally in one direction using surface tension to stretch and traction the growth solution.A 2D OSC crystal is created in this manner.The growth of the crystal is influenced by the shear rate, the concentration of the material solution, and the growth temperature.Giri et al [99] employed the solution shear approach to deposit a TIPS-pentacene crystal film that was highly crystalline and organized.Stress stretching, which can produce a stronger intermolecular-conjugation and enhance charge transfer, is employed for the first time in a non-synthetic technique to alter the crystal structure.TIPS-pentacene transistor mobility is increased from 0.8 cm 2 V −1 s −1 (unstressed regulation) to 4.6 cm 2 V −1 s −1 (stress regulation).However, because the generated film is a strip-shaped crystal, this method cannot be used in a large-area integration process.The dual solution-shearing approach, presented by Peng et al [19], is an improvement on the solution-shearing method.C 10 -DNTT is dissolved in tetralin using this approach, and the concentration of the solution is near to the solvent's solubility limit.The substrate temperature is 85 • C when the solution is sheared for the first time (lower than the second time), and single crystals with various orientations and layers can be formed at this time, as shown in figure 8(d).To further improve the film, a second solution shear was performed with the substrate temperature elevated to 90 • C. The solvent can dissolve more C 10 -DNTT molecules as the substrate temperature rises, as depicted in figure 8(e).The adhesion interaction between the C 10 -DNTT layers is weak, while the adhesion between the bottom layer and the substrate is high, allowing the organic molecules in the upper layer to dissolve and recrystallize (see figure 8(f)) as a result, the growth process of the thin film may be successfully controlled by altering the substrate deposition temperature and solution concentration to generate a single-layer organic crystal with a vast area of centimeter scale.2D OCs are ideal semiconductor materials for transistors, sensors, and low-cost flexible devices due to their unique chemical, mechanical, and optoelectronic capabilities [23,30,[115][116][117][118][119][120].Table 1 shows the comparison of devices prepared by different methods of 2D OCs.The preparation procedure gives a firm platform for scholars to analyse materials more thoroughly.The ideal preparation method is to select one that is appropriate for the material.Whether a liquid phase preparation approach or a gas phase preparation method is used, both have proven to be effective in producing high-quality organic crystals.

Ferroelectric memory
The ferroelectric field effect is explained by manipulating the surface potential while taking use of the spontaneous electric polarization of a ferroelectric contact with a semiconducting layer [121].A typical ferroelectric field effect transistor (FEFET) has a configuration similar to that of a normal metal-oxide-semiconductor field-effect transistor, but with an extra ferroelectric layer in close proximity to the dielectric layer [122].Technically, such a transistor can be used as a non-volatile memory device, storing binary data in the form of polarization directions (up and down), which sparked extensive research in the latter two decades of the twentieth century.However, due to semiconductor fabrication methods, particularly those related to etching and dimension scaling beyond the 130 nm technological node, the successful commercialization of FEFET was never realized.Since the discovery of ferroelectricity in crystalline hafnium silicon oxide in 2011, which is compatible with cutting-edge fabrication technology, a second wave of research interest in FEFETs has sparked [123,124].Many works on molecular crystals-based FEFETs have resulted in astonishing advances and fundamental device physics throughout the last few decades.Using a multilayer ferroelectric dielectric layer, Xu et al first constructed a multi-bit FEFET memory on a paper substrate.There are four distinct bit states and a high mobility of 0.92 cm −2 V −1 s −1 .The physical realization of high-density data-storage devices, regardless of substrate choice, is one of the most important applications [125].
Li et al from Nanjing University conducted a series of experiments concentrating on FEFET memory based on two-dimensional molecular crystals (2DMCs).In 2016, Li's group used the floating-coffee-ring effect to achieve single crystalline C 8 -BTBT on a large scale [59].On the other hand, C 8 -BTBT deposition on the ferroelectric polymer poly(vinylidene fluoride-trifluoroethylene) P(VDF-TrFE) surface differs from earlier research.In 2017, Li's team developed 2DMCs using a solution-based method at room temperature (C 8 -BTBT served as the p-type semiconducting channel layer, while P(VDF-TrFE) served as the dielectric material) [127].Because this technique can generate a double layer of 10 nm on the P(VDF-TrFE) surface induced by a vertical phase separation, the scientists added a little amount of poly (methyl methacrylate) (PMMA) to the solution to optimize the shape of the ferroelectric material.The charge injection from metal electrodes becomes effective owing to the ultrathin nature of 2DMCs, resulting in a fast switching-on time of 2.9 ms.More crucially, because P(VDF-TrFE) is the most common functional material for ferroelectric organic field effect transistors, the authors' technique opens a new way to optimize the performance of solution-based FEFET memory.The impacts of thermal annealing on the structural reorganization and ferroelectric polarization essence of ultrathin P(VDF-TrFE) were then investigated in a system [128].The film has a highly-ordered antiparallel dipole packing and a faster decay period after a 40 • C annealing treatment.The authors proposed that the electrical polarization of P(VDF-TrFE) can be attributed to such a highly ordered orientation of the chains and molecular dipoles because the ferroelectric feature of P(VDF-TrFE) is highly related to the orientation degree of the electric dipoles and the backbone chains.The ultrathin P(VDF-TrFE) film is now more suitable for use in FEFET data storage devices as a result of in-depth investigations.Semiconducting materials in the 2D dimensions can significantly enhance charge injection and switching speed when compared to bulk functional materials.Using an antisolvent-assisted solution-based method, Li's team successfully deposited a highly uniform 2D P(VDF-TrFE) layer on a high-κ Al 2 O 3 substrate with a thickness of 5 nm to suppress gate leakage current in 2019 (figure 9(a)) [44].The semiconductor layer possessed a high area and atomical flatness as a result of this approach, which was confirmed by AFM characterizations.It is worth noting that instead of using evaporation techniques, the source and drain electrodes were changed to a bottom-gate topcontact layout.Even at a low scanning voltage of and superior subthreshold swing of 41 and 97 mv/dec, the device showed an apparent hysteresis loop, as illustrated in figure 9(b).Furthermore, without gate bias, a high on/off ratio of 10 5 was observed, which is desirable in FEFET non-volatile memory.In comparison to the bulk functional layer, the contact resistance can be reduced to 303 Ω cm, signifying good thermal management, power consumption and device stability (figure 9(c)).As depicted in figure 9(d), Pei et al introduced a floating platinum metal layer for separation of electron injection into the semiconducting channel layer and the floating gate layer, intercoupling the lateral current at the semiconducting/dielectric interface by modulating the dipole polarization of the ferroelectric layer, motivated by prompting the on/off ratio [126].Platinum was used as a floating gate to store charges and modify the polarization of the ferroelectric layer, as illustrated in figure 9(e).Figure 9(f) presents that an increased ON/OFF state distinction when the C 8 -BTBT molecular layers shrink to bilayer.Electrons injected from the top electrodes spontaneously split into the floating layer and collect at the semiconductor/dielectric interface when a negative bias is applied to the gate electrode.As shown in figures 9(g) and (h), the electrons dividing into the floating gate cause polarization, and as a result, lateral charge transportation, which corresponds to the memory's ON state.When a positive bias is supplied, electrons escape from the floating gate and change polarized directions, causing charge depletion and completing the information erasure process.The retention test shows no obvious performance degradation even after 1200 s, demonstrating a good device stability, as depicted in figure 9(i).In addition, when the thickness of the ferroelectric layer reduced, an optimal buffering layer of Al 2 O 3 was formed to trap the holes and provided an atomically smooth morphology, reducing the leak current in the vertical direction.As a result, by combining a high-quality organic channel layer with a ferroelectric with exceptional polarizations, the lateral current was boosted and an unbelievable on/off ratio was attained.Further research is required to better understand the device's mechanics and improve the device's performance.

Circuit-type synapse
Circuit-type optoelectronic synapses are predicted to transform optical input signals to electrical output and replicate the synaptic plasticity of human brains.Circuit-type optoelectronic synapses are made by merging various basic functional blocks, such as a phototransistor, a load transistor, and an electrical synapse with no photoactive function [129].Due to the apparent lack of photosensitive elements, the load transistor and simply electrical synapse seldom display direct photoelectricity connection when the input light stimulus is applied to the sensor element.The electrical bias intensity is highly proportional to the phototransistor's conductance state, which is modified by light stimuli.The electrical bias signal is then sent through the interconnecting wire to the controlling terminal of the three-terminal transistor-based or two-terminal memristor-based synapse.Finally, different synaptic device principles can be used to switch the electrical synapse's synaptic weight.Even if the optical stimulus is turned off, the electrical synapse's updated synaptic weight remains intact for a long time.As a result of the photo-to-electrical conversion and synaptic weight shift of a pure electrical synapse, a unique optoelectronic synapse may be built using such a circuit-based architecture to carry on device conductance switching.
2D OSs with well-defined interfaces and precisely adjusted thicknesses have various of potential for physical implementation of phototransistors because their tunable energy gap can cover a large spectrum range from near-infrared to ultraviolet [53,54,[130][131][132][133].Low photogating efficiency, low carrier mobility, and short photo-induced exciton diffusion distance in bulk organic materials, on the other hand, impede the development of highly efficient and fast response phototransistors [134][135][136][137].In order to ensure photo-generated exciton separation, the thickness of the photoactive layer should be lowered below 10 nm [138].Furthermore, the interlayer screening effect found in thick photoactive materials impedes photocurrent decay characteristics.Guo et al built a few-lay N,N -(1-hexyl)-1,4,5,8-naphthalenetetracarboxydiamide (NDI-C 6 )/C 8 -BTBT van der Waals heterojunction to overcome the exciton-diffusion bottleneck and restrict exciton separation/recombination in the year 2020 [134] (Inset of figure 10(a)).As presented in figure 10(a), the authors extracted the photoresponsivity and sensitivity to evaluate the device's performance, which are 1.78 A/W and 50, respectively.The device also had a quick rise time of 4 ms and a fast decay time of 6 ms (figure 10(b)).When the heterostructure's thickness is reduced to less than ten nanometers, the exciton diffusion process is significantly sped up, and exciton separation/recombination can be restricted to the atomically interfacial layers, taking use of the strong built-in field at the interface (figure 10(c)).Another solution is to design a defect-free interface between OSCs and high-mobility materials.Liu et al [139] proposed a van der Waals epitaxy of 2D OSCs (C 8 -BTBT) on graphene for phototransistors.The wide spectrum sensitivity of OSCs and the high mobility of graphene can be combined in this way.The authors discovered a 25 ms photoresponse time and a gain of 10 8 .The scientists discovered that monolayer-based phototransistors may respond 30 times faster than multilayer-based phototransistors when compared to multilayers C 8 -BTBT with precisely controlled thickness.The approach for precisely regulated growth of large area 2D OCs with high crystallinity, on the other hand, is an impediment to phototransistor application.Because the contact between the molecule and the substrate becomes non-negligible at the 2D limit, the crystallization process is highly sensitive to external factors.Cao et al recently discovered that the Gibbs-Curie-Wulff equation governs the formation of 2D OCs via the solution approach, necessitating thermodynamically controlled circumstances such as high temperature, seed layers, and oversaturated solution [48].The authors then presented a seed crystal drop-casting approach to assist the lateral growth of 1,4-bis (4-methylstyryl) benzene (p-MSB) on a thermally oxidized Si substrate with a controllable layer thickness and reach near to a surface energy minimum state (figure 10(d)).Generally, there are three interactions including inter-layer interaction (I inter-layer ), molecule-substrate interaction (I substrate ) and intra-lay interaction (I intra-layer ) in an organic molecular stacking process.Differ from the bulk counterparts, the predominant factors in the 2D molecular packing process are I substrate and I intra-layer , respectively, as elaborated in figure 10(e).It is worth mentioning that a distinct 2D limit effect was discovered, which significantly increased the photoelectrical response of the as-fabricated device.Experiments and theoretical calculations revealed that a compacted molecule stack of p-MSB is produced by a weak interaction with the SiO 2 substrate and a strong stacking between layers.As a result, the overlapping molecular orbitals form a continuous network, delocalizing charge carriers and changing the band structure (figures 10(f) and (g)).This internal essence boosted the internal photoresponsivity up to 2-3 orders of magnitude greater than that of 3D bulk crystals (figures 10(h) and (i)), paving the way for high-performance phototransistors based on 2D OCs.Despite the fascinating properties of phototransistors, novel technologies for compact manufacture and quasi-linear synaptic weight switching performance have yet to be established, and additional research is needed before high-density optoelectronic neuromorphic computing systems can be implemented.

Neuromorphic computing
To address the ever-growing issues caused by the traditional von-Neumann architecture, neuromorphic devices are essential for developing matrix computing paradigms that benefit from their excellent parallel computing capabilities, self-adaptability, and ultra-low power consumption [140][141][142].The emerging 2DMCs, which have an atomically layered structure, are expected to effectively eliminate the ubiquitous interlayer shielding effect and short-channel effects found in bulk materials, allowing for efficient electrical gating, photogating, and photo-response of artificial synaptic devices [71,127,143].Organic crystals with a 2D thickness are potential candidates for use in integrated neural networks in this regard.
There have been a number of reports on bio-inspired synaptic devices, including three-terminal/multiterminal transistors and two-terminal memristors.Photo-programming, as an unconventional stimulus, enables for precise control of synaptic devices on a spatial and temporal scale.Taking advantage of the elaborate energy band engineering between 2D inorganic and OSCs, Wang et al first demonstrated a multi-functional synaptic transistor with both electrical and optical tunability and gate modulation based on a molybdenum disulfide (MoS 2 )/perylene-3,4,9,10-tetracarboxylic dianhydride (PTCDA) heterostructure, which successfully simulates abundant synaptic behaviours in a single device, including dynamic filtering, short-term plasticity and long-term plasticity (LTP) [144].Because of the superior light absorption characteristics of 2D MoS 2 and PTCDA, the fully 2D system achieved the lowest inhibition and maximum facilitation ratios of 3 percent and 500 percent under successive electrical pulse trains, and the long-term synaptic weight change was significantly improved up to 60, far superior to that of the reported works.
Furthermore, numerous investigations on 2D metal-organic framework (MOF) materials have been carried out for prospective application sectors like as drug delivery, catalysis, and electronic devices, among others.Because of the easily accessible active regions in the molecular structure, 2D MOFs are expected to have significant charge trapping capabilities.However, few researches have been done to see if the 2D MOF might be used as the active layer in simulating the human synapse.As a result, Liu and his colleagues presented a light-stimulated MOF-based (Cu-THPP, THPP = 5,10,15,20-tetrakis(4-hydroxyphenyl)-21H, 23H-porphine) synaptic device with remarkable synaptic plasticity characteristics and good stability under air exposure, accelerating the advancement of neuromorphic devices in practical applications [145].Ding et al [146] used a 2D MOF, Zn-TCPP (TCPP: tetrakis(4-carboxyphenyl)porphyrin), as the charge trapping layer to replicate neural plasticity and dynamic filtering (figures 11(a) and (b)).With pyrrolidone, the 2D MOF nanosheets serve as floating gates (PVPy, which was utilized to act as the dielectric materials).As shown in figures 11(c) and (d), the scientists successfully showed a number of essential synaptic functions as well as diverse postsynaptic current responses to temperature changes, paving the way for the development of a neuromorphic computer system in the future.COF, a sort of 2D layered porous polymer covalent organic framework, has shown significant promise as function layers for neuromorphic device construction.COF's highly ordered structure allows for long-range interaction along the stacking direction, which helps with charge transport [148].Since active sites are included into the framework and nanopores, COF exhibits improved charge separation and ambipolar carrier transport behaviour [149].Unfortunately, due to the issues including processability, thickness, stability, and crystallinity of thin films, integrating 2D boronate ester covalent organic frameworks (2D BECOF) into logic and data-storage devices remains a significant issue.Park et al initially produced a 2D BECOF single-crystalline film with a domain size of 60 μm 2 using an interfacial synthesis approach supported by an anionic monolayer surfactant [147] (figure 11(e)).As shown in figure 11(e), the charges accumulate at the 2D BECOF/SiO 2 interface when the positive gate voltage is applied.And the charge can be maintained by the residual polarization even if the gate voltage sweeps back to 0 V, contributing to a significant memory window.Figure 11(f) depicts that the threshold voltage of device based on Si NW does not shift during the scanning process, while the threshold voltage shifts to the left when the device sweeps back.Based on the memory effect, the 2D BECOF/SiNW hybrid transistor was then successfully used to create an artificial synapse with learning-forgetting-rehearsal properties (figures 11(g)-(i)).2D OSC-based synaptic transistors with an ultrathin heterojunction interface can produce excessive bandwidth and ultra-fast switching-on time, resulting in a significant increase in the performance of the light-dark current ratio, light response, and detection rate, which is essential for the practical application of optical synaptic devices.To ensure high-mobility and high switching-on stability of carriers, the channel layer of conventional transistors is continuously tuned without subgap defect states.Under external stimuli, these high-ordered crystalline-based transistors displayed transient switching of device functioning.Once the electrical/optical signals are interrupted, the conductance quickly recovers to its initial state with fast recovery behaviours.In ordinary transistors, there are no additional components to sustain external signal responses and update analogue synaptic weight.As a result, additional components are required to offer the future transistor-based synapses with persistent conductance switching dynamics and analogue conductance updating capabilities.Fang et al recently published a light-stimulated artificial synaptic device based on 2D OSC deposited at room temperature using the solution epitaxy approach [151].Because of its outstanding solubility and crystallinity, the p-type organic small molecule 2,7-dioctyl benzothieno [3,2-b] benzothiophene (C 8 -BTBT) was used as the synapse's channel layer.The solution-based epitaxy technology allows 2D organic thin films to develop directly on the water surface, eliminating the dielectric effect and simplifying the lightstimulated operating mechanism.The device had remarkable device performance with a high I light /I dark ratio of up to 1.7 × 10 5 , a superior photoresponsivity of 3 × 10 3 A W −1 , and a competitive detectivity of over 10 14 Jones.More importantly, the authors successfully realized synaptic functions based on the 2D OSC by introducing the interfacial charge trapping effect at the dielectric/channel interface, which may shed new light on introducing intriguing 2D OSCs due to their environmental stability and simple preparation process for constructing neuromorphic computing systems.Despite its numerous advantages, the development of high-crystallinity and large-scale 2D conjugated polymers (2D CPs) is hampered by their very complicated microstructures and molecular interactions, which necessitate the development of regulated growth methods.Zhang et al recently presented a controlled method for obtaining centimeter-scale 2D CPs using appropriate solvents and surfactants [152].Furthermore, the typical solution-based epitaxial approach takes a long time to produce crystals, ranging from one day to several days.Because of the spatial restriction effect, Zhang et al used a drop-casting approach to create 2D CPs-based photonic synaptic transistors that successfully executed normal synaptic functions.The scientists claimed that the growth of centimeter-scale 2D CPs can be attributed to a synergistic effect caused by using the right solvent and surfactant, which can be applied to growth of other polymers as well.Photogenerated carriers are induced and released at the metal/semiconductor and semiconductor/dielectric interfaces for optically regulated synaptic plasticity in two-terminal vertical memristors and three-terminal transistors.However, carriers are transported at the interface as a result of the electrical field, which might cause an unwanted charge combination process and decrease synaptic efficacy.This conundrum necessitates a paradigm shift in principle design and innovative physics.Yang et al has developed an optically controlled artificial synapse in which the Schottky barrier may be sensitively modified when exposed to UV light.The intrinsic ultrathin nature of 2D OSCs, enables for increased trapping manipulation of the Schottky barrier in planar diodes [153].The quick recovery of conductance state is attributed to 'shallow trapping' at the semiconductor/dielectric contact, according to the authors.Furthermore, by increasing the irradiation intensity and frequency, deep trapping can be achieved to preserve LTP, prompting fresh ways for emulating light signals in cognition and processing.Zhang et al created a spectrum-dependent optically modulated flash memory to replicate visual learning and memory behaviours that take advantage of 2D polymer's wide spectrum responses [150] (figures 12(a) and(b)).The photosensitive charge trapping layer was created using a 2D polymer that was produced at the air/water interface using the Schiff base condensation technique.Because of the extraordinarily sensitive photo-response of the 2D polymer, which approaches that of a true synaptic event, the device had a low working voltage of −0.1 V and a competitively low energy consumption of 0.29 pJ per spike.The 2D polymer-based bionic synapse demonstrated robust synaptic characteristics with no performance degradation across a large temperature range, as expected.The authors demonstrated a transition from short-term memory to long-term memory by altering the frequency, pulse number, and breadth of laser pulses.Photogenerated carriers can be introduced into the 2D polymer layer and then stored to modify synaptic weight under varied light irradiations.The authors assumed that 400 and 430 nm light could be used to represent a happy mood and a neutral mood, respectively, based on the features of learning, memory, and forgetting under various emotional states.They also assumed that 430 nm light could be used to represent a bad mood, based on the features of learning, memory, and forgetting under various emotional states.Then, using a letter recognition simulation, they effectively replicated the impact of emotions on learning, memory, and forgetting behaviours.In addition, they also demonstrated emulation of Pavlov's dog behaviour based on the neuronal plasticity and photosensitive properties.As depicted in 8d, the authors introduced 400 nm light as the signal of 'food' (Pre 1 in figure 12(c) and 430 nm light (Pre 2 in figure 12(c)) as the signal of 'bell ring', respectively.Each of the optical signals can produce various PSC variation owing to the incident energy difference.After a simultaneous training with food and ring stimuli, the bell ring signals with lower incident energy produce a significant PSC variation beyond the preset value, as depicted in figure 12(e).After conditioning, the carrier escaped from the trap sites and the PSC change could not reach beyond the threshold value under the bell ring stimuli (figure 12(f)).Furthermore, KPFM characterization demonstrated that the transporting of photogenerated holes to the pentacene layer enhance the surface potential of the semiconducting layer under optical stimulation, resulting in synaptic weight modulation and reducing energy consumption.In contrast to traditional digital imaging and recognition systems, which rely heavily on various digital components in conjunction with complex optical modules due to a large amount of data sampling, storage, pre-processing, and transmission, the human vision recognition system, despite its simplicity, has unique advantages of high efficiency and accuracy thanks to the distinct characteristics of the matching between the curved retina and the focal plane.Choi and his colleagues used a 2D heterostructure of MoS 2 and poly (1,3,5-trimethyl-1,3,5trivinyl cyclotrisiloxane) (pV3D3-PTr) to create an integrated module that included a curved artificial image sensor array and a plano-convex lens [154].(Figures 13(a) and (b)) Long-term potentiation is realized under continuous optical stimuli thanks to the accumulated photocurrent (figure 13(c)).As shown in figures 13(d) and (e), the integrated module received a pre-processed picture from a set of noisy light signals that were free of other unwanted components, significantly improving the efficiency of the aberration-free data acquisition and identification operation by taking advantage of the nearly linear time-dependent photoinduced current and prolonged self-recovery relaxation.The device also showed a significant difference in relaxation and retention times, which is connected to carriers trapped at the interface, and successfully raised the contrast of the pre-conditioned image, as presented in figures 13(f)-(h).The remaining contrast can be instantly eliminated under a positive gate voltage, if there is necessary.Therefore, the succedent image acquisition process and pre-processing operation can be realized without any interference after the aforementioned step.Theoretical analysis of the heterostructure features revealed that the substantial exciton binding energy and type-II band alignment cause the separation of electron-hole pairs, resulting in photocurrent due to photogating effects.Such an integrated module might boost the efficiency for the data acquisition and pre-processing process thanks to a reasonable device design and an intelligent choice of functional materials, which enlighten efficient computer vision techniques.
The majority of previously published works focuses on demonstrating basic synaptic plasticity, although tactics for increasing accuracy are rarely offered.Yang et al discovered that modulating the channel layer thickness increased the pattern recognition rate significantly.The authors claim that flaws in thinner films have a major role in preventing charge carrier transfer at the interface, causing synaptic weights to rise linearly [155].These preparatory studies will inspire new pathways to comprehending the basic properties of 2D molecular crystals and extending the bounds of novel devices, thanks to the quick advancement and extensive research of 2D molecular crystals.

Conclusion and outlook
From the internal structure-property relationship to implementations of in-memory computer technologies, tremendous progress has been made in the field of phototransistors based on low-dimensional organic crystals over the last decade.We summarized and analyzed previous work in this emerging field, focussing on the 2D OC promises, efficient synthesis strategies for large-scale 2D OCs with high uniformity and crystallinity, photoelectrical responses in phototransistors, their hybrid configuration with other electrical components, and groundbreaking optoelectrical synapses.Despite significant progress, 2D OCs-based optoelectronic synapses are still in the early stages of development, and current research on large-scale 2D OCs is far from mature.Generally, there is a convention performance discrepancy between inorganic transistors and 2D OCs counterparts.Therefore, to boost the advancement of 2D OCs transistors from concept genesis to commercial maturity, 2D OCs transistors should be explored as the extended functional devices in the perspective of flexibility, biocompatibility and photosensitivity.Thus, the ability to manipulate the fast turn-on time, photoresponsivity and large conductance range is highly intriguing for realizing 2D OCs-based hardware implementation of artificial sensory system.Besides, access to the electrical, mechanical, chemical and optical modulation enables 2D OCs transistor integration in various scenario.Cycling endurance and device stability should also be sufficiently increased, especially for long-term and high-frequency operation.Finally, device reproducibility of 2D OCs transistors is highly needed, resulting to an array integration success.All these features call for further in-depth investigations and developments from the materials' point of view.The intending research directions of 2D OCs-based memory devices should include the following aspects.First, due to structural asymmetry and highly-ordered molecular configuration, it is crucial to elaborate the crystal engineering for the growth of 2D OCs in the preferable crystalline orientation since anisotropic carrier transport would affect the ultimate electrical performance.Second, organic crystals at the 2D limit are often dominated by trap sites and structural disorders and far away from the intrinsic material property, which leads to the atomically thin films operate completely different from bulk ones.Therefore, precisely layered growth of organic crystals is imperative to endow novel functions to the artificial synapses despite the advancement of controllable growth along the lateral direction.Third, accidental doping originated from chemical adsorption of oxygen and water, and thermal fluctuations would degrade the device's uniformity and stability.Specific encapsulation technologies and fabrication strategies with stable organic monomers could help improve device reproducibility and endurance to desired levels.Fourth, new strategies for developing large-scale n-type 2D semiconductors with good stability is urgently required for constructing high-performance logic units.Finally, more mechanism investigations should be done to gain a better understanding of materials design and circuit improvements.2D OCs with distinct optoelectronic responses can assist progress neuromorphic computing and even expand the border to the realm of near-in-sensor computing by enriching 2D organic materials systems and providing complete insights into relevant scientific difficulties.

Figure 1 .
Figure 1.Representative structures of organic molecules involved for optoelectronic transistor.

Figure 2 .
Figure 2. (a) Summary of mobilities of 2D OCs OFET involved in this review, comparisons of 2D limit and bulk organic crystals in the same work are depicted.(b) Plots of the mobility, threshold voltage and I on /I off ratio of P(NDI2OD-T2)-based transistor versus different number of layers.(b) Reprinted from [50], Copyright (2012), with permission from Wiley-VCH.(c) Comparison of contact resistance and Schottky barrier height versus different gate bias of monolayer and multilayer F 4 BDOPV-2T-based transistor.(c) and (d) Reprinted from [45], Copyright (2019), with permission from Wiley-VCH.External photoresponsivity under different number of molecular layers in both (e) off and (f) on state.(e) and (f) Reprinted from [48], Copyright (2019), with permission from Springer Nature.

Figure 3 .
Figure 3. (a) Schematic diagram of the growth of C 8 -BTBT via the floating-coffee-ring strategy.(b) 3D schematic of the bilayer C 8 -BTBT-based transistor with a bottom-gate top-contact configuration.(c) The relations between the average/maximum carrier mobilities with different layer numbers.(a)-(c) Reprinted from [59], Copyright (2016), with permission from Wiley-VCH.(d) Schematic representation of the ultraslow shearing strategy.(e) Schematic of the growth of the highly ordered C 10 -DNTT crystals by the ultraslow shearing method.(f) Histogram of the carrier mobility.(g) and (h) Output curves of the devices with/without thermal evaporation templates.(i) The extracted relationships between the mobility and the gate voltage.(d)-(i) Reprinted from [60], Copyright (2016), with permission from Wiley-VCH.

Figure 4 .
Figure 4. (a) Schematic illustration of the cross-sections monolayer C 10 -DNTT-based transistor with a bottom-gate top contact configuration.(b) Extracted contact resistance and carrier mobility of the monolayer-based device under various gate overdrive voltages.(c) Output characteristics of transistors based on monolayer C 10 -DNTT crystals with Vov from −0.1 V to −12.1 V. (a)-(c) Reprinted from [62], Copyright (2020), with permission from Wiley-VCH.(d) Energy band diagrams of Au, 1T-TaSe 2 , and F 16 CuPc.(e) Schematic structure of bottom-gate top gate transistors based on F 16 CuPc nanoflake.(f) Typical transfer curves of the device with 1T-TaSe 2 (green line) and evaporated Au pad (purple line) in linear (solid line) and log (dash line) scales.(g) Extracted Schottky barrier height as a function of V g of the device based on 1T-TaSe 2 contact.(h) Calculated trap states energy ET and reduction ratios of ET as a function of V g for the two kinds of contacts.(i) Scheme of operating mechanism of transistors with 1T-TaSe 2 and Au contacts.(d)-(i) Reprinted from [63], Copyright (2021), with permission from Wiley-VCH.

Figure 5 .
Figure 5. (a) Schematic diagrams of the self-assembly growth of monolayer C 10 -BTBT induced by GQDs.(b) Homologous layer stacking characteristics of C 10 -BTBT with different PH values of GQD solutions from 3 to 12. (c) Schematic illustrations of decisive qualities of GQDs during the large-area growth of C 10 -BTBT.(a)-(c) Reprinted from [73], Copyright (2020), with permission from Wiley-VCH.

Figure 7 .
Figure 7. (a) A schematic depiction of the equipment used to synthesize 1D wires and 2D disks via physical vapour deposition methods.(b) and (c) SEM images of 1D wires and 2D disks of pentacene generated at 350 • C and 325 • C (scale bar: 5 μm).The inset is the AFM image of the respective height profile.(a)-(c) Reprinted from [75], Copyright (2012), with permission from Wiley-VCH.(d) The spin-coating process captured by a high-speed camera at different times.(e) AFM image of NDI(2OD) (4tBuPh)-DTYM2 film with different thickness of 4-50 nm.(d) and (e) Reprinted from [77], Copyright (2013), with permission from Wiley-VCH.

Figure 8 .
Figure 8.(a) Optical images and schematic diagrams of the coating process of single crystal C 8 -BTBT prepared by floating-coffee-ring-driven assembly technology at different time periods.(b) AFM image of monolayer C 8 -BTBT.The height contour line in the figure corresponds to the white dashed line in the AFM image.(c) Optical microscope image of the monolayer C 8 -BTBT.(a)-(c) Reprinted from [19], Copyright (2016), with permission from Wiley-VCH.(d) and (e) POM images of C 10 -DNTT crystals after the first and second solution shearing (scale bar: 200 μm).(f) AFM images of highly crystallized C 10 -DNTT monolayers.The white arrow indicates the direction of solution shearing (scale bar: 2 μm).(d)-(f) Reprinted from [59], Copyright (2017), with permission from Wiley-VCH.

Figure 9 .
Figure 9. (a) 3D illustration of 2D C 8 -BTBT-based FEFET with a bottom-gate top-contact configuration.(b) Typical output characteristics of the FEFET memory.(c) Power consumptions of different operations for the 2D C 8 -BTBT-based FEFET memory device.(a)-(c) Reprinted from [44], Copyright (2019), with permission from American Chemical Society.(d) and (e) Schematic and operation mechanism of the FEFET memory with bilayer C 8 -BTBT as the semiconducting layer and P(VDF-TrFE) as the ferroelectric layer.(f) I -V characteristics in FEFET memory with different C 8 -BTBT layer numbers.(g) and (h) Schematic demonstration of the internal mechanism of the C 8 -BTBT-based FEFET memory device in both (g) on and (h) off states.(i) Retention test of the FEFET memory based on bilayer C 8 -BTBT.(d)-(i) Reprinted from [126], Copyright (2021), with permission from Elsevier.

Figure 10 .
Figure 10.(a) Transfer curves of the phototransistors with an NDI-C 6 /C 8 -BTBT van der Waals heterojunction with/without UV light illumination.Inset shows the schematic of corresponding device structure.(b) The photo-switching process involving the set and reset operation.(c) Energy band diagram and charge transfer process of the NDI-C 6 /C 8 -BTBT van der Waals heterojunction.(a)-(c) Reprinted from [134], Copyright (2020), with permission from Wiley-VCH.(d) The 3D schematic configuration of the phototransistor based on 2D p-MSB crystals, and the molecular structure of perylene, α-6T and p-MSB.(e) Molecular stacking patterns of 2D p-MSB on h-BN/SiO 2 substrate.(f) Calculated molecular orbitals on the intermolecular bonding states for the upper and bottom molecular layers.(g) Band dispersions for the upper and bottom molecular layers.(h) and (i) External photoresponsivity under different number of molecular layers in both (h) off and (i) on state.(d)-(i) Reprinted from [48], Copyright (2019), with permission from Springer Nature.

Figure 12 .
Figure 12.(a) The schematic of a biological synapse.(b) 3D illustration of the synaptic transistor based on 2D polymer.(c) Schematic demonstration of the classical Pavlov's dog experiment.(d)-(f) Simulation of classical Pavlov's associative learning process driven by optical signals of 400 nm as unconditioned stimuli and 450 nm as conditioned signals; (d) before conditioning step, (e) during conditioning process, (f) after conditioning step.(a)-(f) Reprinted from [150], Copyright (2021), with permission from Wiley-VCH.

Figure 13 .
Figure 13.(a) 3D illustration of the synaptic device with a hybrid heterostructure of MoS 2 /pV3D3-PTr.(b) Optical microscope image of the device.(c) Long-term potentiation triggered by optical pulses.(d) Schematic demonstration of the image inputs and data-processing process based on a 3 × 3 pV3D3-PTr device array.(e) Normalized photo-induced current obtained at each individual pixel in the array.(f) The image acquisition process.(g) Pre-processed image measured by the integrated module.(h) The decay process of photocurrents.(i) The erasure operation by applying an electrical signal.(a)-(i) Reprinted from [154], Copyright (2020), with permission from Springer Nature.

Table 1 .
Performance comparison of 2D OC devices prepared by different methods.