An electrically and optically controllable memristor with synaptic plasticity based on scalable monolayer ReS2/WS2 heterostructure

doi:10.21203/rs.3.rs-2006369/v1

Download PDF

Article

An electrically and optically controllable memristor with synaptic plasticity based on scalable monolayer ReS₂/WS₂ heterostructure

https://doi.org/10.21203/rs.3.rs-2006369/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Memristors with non-volatile storage performance and simulated synaptic functions are regarded as one of the critical devices to overcome the bottleneck in traditional von Neumann computer architecture. van der Waals heterostructures integrating excellent properties of two-dimensional semiconductor materials, possessing controllable optoelectronic properties and high compatibility with conventional microelectronic technology, have paved a new way for the development of advanced memristors. Herein, we demonstrate a two-dimensional planar memristor with both electrical and optical controllability based on ReS₂/WS₂ van der Waals heterostructure. The device shows a typical unipolar non-volatile behavior with a high R_on/R_off ratio, multiple tunable resistance states, and desirable endurance and retention. It also successfully realizes biological synaptic functions and plasticity, including spike rate-dependent plasticity and paired-pulse facilitation. Furthermore, the developed device shows a significant gate controllability and a remarkable optical tunability. The superior performance is attributed to the unique optoelectronic property and the interlayer interaction in the heterostructure. The research presented here demonstrates the potential of two-dimensional van der Waals heterostructures for high-performance memristor applications and further developments in modelling biological synapses.

Recent years have witnessed a spurt of progress in computing and communications technology. As traditional silicon-based microelectronic devices have struggled to scale up further according to Moore's Law, advanced storage technologies require new materials and device design.^1,2 Memristors can maintain the internal resistance state created by the applied voltage and current,^3,4 making them viable to implement memory computing systems without any program and having attracted tremendous research attention. The preliminary study of memristors mainly involved traditional three-dimensional (3D) materials,^5–7 which are challenging to reduce the size to meet the stringent requirements of future big data and artificial intelligence for high-density integration and low power consumption. Two-dimensional (2D) semiconductors are considered be a focal point for future research due to their ultra-thin thickness, high flexibility, and intriguing optoelectronic properties.^8–10 Rhenium disulfide (ReS₂) is a promising 2D material with unique physical and chemical properties. The soft Re-S covalent bonds are more likely to produce sulfur (S) vacancies, providing an intensive possibility for the development of 2D memristors. The S vacancies have an even lower forming energy and more evident movement under the applied bias,^11–13 which are beneficial to improve the switching performance of memristors.¹² Sifan Li et al. demonstrated a double-ended lateral memristor based on few-layer ReS₂, realizing an amnestic block function with a cycle count of over 100 and a switching ratio of 10².¹⁴ Besides, a simulation of synaptic plasticity was further implemented.

Nevertheless, the retention and durability of memristors based on a single 2D material may be relatively lower than that of traditional 3D devices. Heterostructures can integrate the properties of different materials becoming an alternative to enhance the performance of memristors.^15,16 In order to take advantage of the excellent properties of 2D materials, and also consider the compatibility with conventional microelectronic technology, attention has been targeted on 2D/2D heterostructures with steep potential gradient at the atomic level. The stacked layers in the heterostructure are no longer restricted by lattice matching due to their van der Waals interface. Zhang et al. designed a 2D/2D memristor based on few-layer MoS₂/WS₂ heterostructure, implementing a resistive switching mechanism based on the band structure modulation with a switching ratio of 10⁴.¹⁷ Its higher stability, larger memory window, and superiority in high-density integration further confirm the great potential of 2D van der Waals heterostructures for the memristor applications.

In addition, most of the existing memristors are solely based on the electrical modulation. It does not completely match the practical human stimuli, including vision, hearing, touch, smell and so on. Light as an extra stimulus may also supply a way to fill the gap between the visual system and brain function.¹⁸ Beyond pure electronic devices, further introducing the optical signals as information carriers during the device working has advantages in terms of high bandwidth, low crosstalk, and high connectivity. The optical programmability also facilitates the application in the field of neuromorphic vision.¹⁹ Therefore, developing a memristor with both electrical and optical controllability is significantly crucial. In this regard, van der Waals heterostructures possess another advantage on modulating the optical absorption and exciton recombination through the design of their band alignment.^20–22

In this work, a 2D memristor with planar structure is fabricated based on ReS₂/WS₂ heterostructure through an alignment transfer of chemical vapor deposition (CVD) grown monolayers. The lattice configuration, stacking structure and band alignment are determined by Raman scattering, transmission electron microscopy (TEM), time-resolved photoluminescence (TRPL) and Kelvin atomic force microscopy (KPFM). The memristor properties, such as resistance states, switching ratio, cycle number, and retention time are studied at room temperature, and the synaptic functions of short-term plasticity (STP), spike rate-dependent plasticity (SRDP), and double-pulse (PPF) dissimilation are simulated. Moreover, the special gate and optical control over the resistive switching is achieved, and the related mechanism is further revealed.

Morphological, structural, and electronic properties of the ReS₂/WS₂ heterostructure

For the development of ReS₂/WS₂ memristor, the properties of ReS₂/WS₂ heterostructure are studied at first. The heterostructure is prepared through an alignment transfer of CVD grown ReS₂ and WS₂ monolayers (detailedly in Methods). The optical microscopy in Fig. 1a shows a clear surface all around the sample, and the heterostructure area is vertically stacked with triangular WS₂ on top of hexagonal ReS₂. The crystal and electronic structures of the ReS₂/WS₂ region are characterized through the Raman and photoluminescence (PL) measurements. As shown in Fig. 1b, the Raman spectrum of heterostructure region is coincident with the combination of characteristic peaks from the two monolayers, where the peaks located at 164 cm^− 1 and 216 cm^− 1 respectively correspond to the in-plane (E¹_2g) and out-of-plane (A_1g) vibrational modes from ReS₂,¹¹ and the two peaks at 359 cm^− 1 and 420 cm^− 1 are identified as from monolayer WS₂.^23–25 Different from the superposition of the Raman peaks, the PL spectrum for the heterostructure demonstrates an obvious intensity decay and a 5 nm blue shift with respect to the monolayer WS₂ (625nm), as shown in Fig. 1c, which suggests an interlayer charge transfer and stress interaction.^26,27 The intensity mappings of Raman and PL peaks further reveal the good uniformity of the structures (Figure S1a-c in SI). Accordingly, we infer that a type II heterostructure may be formed between ReS₂ and WS₂.²⁸ The atomic force microscope (AFM) image (Figure S1d in SI) also indicates an existence of interlayer interaction.²⁹

Interlayer interaction related to the band alignment of WS₂ and ReS₂ monolayers is closely associated with the electronic property of the heterostructure. To gain an insight into their band alignment, KPFM is employed to detect the work function (φ_s) of a 60° stacked ReS₂/WS₂ heterostructure on an Au-plated SiO₂/Si substrate. As shown in Fig. 1d and Figure S2a,b in SI, a clear variation of φ_s between ReS₂, WS₂, and the heterostructure areas is observed, in well agreement with the optical contrast. By measuring the potential change in ReS₂/Au and WS₂/Au steps, the φ_s difference is determined to be about − 21 mV between ReS₂ and Au, and about − 66 mV between WS₂ and Au, which indicates a higher Fermi energy level of WS₂ than that of ReS₂ (seeing detail in Figure S2c,d in SI). HRTEM characterization (detailedly in the Methods) is further performed to investigate the lattice and stacking structures of the ReS₂/WS₂ heterostructure. As measured in Fig. 1e, the space between two Re atoms in the Re4-chain (0.27 nm) is smaller than that of adjacent two Re atoms (0.32 nm), consistent with a 1T'-ReS₂ structure, while the distance between neighboring W atoms (0.29 nm) corresponds to 2H-WS₂.³⁰ The FFT image (insert in Fig. 1e) and SAED patterns (Fig. 1f) contain two set of six-fold symmetrical lattices, corresponding to the lattice spacings of 5.5 Å and 2.8 Å, respectively. The lattice structures are identified to the (010) planes of monolayer WS₂ and ReS₂, respectively, confirming a 60° vertical stacking of the heterostructure.

Photoinduced interfacial charge behavior is studied through the TRPL measurements (Fig. 1g). The heterostructure area exhibits a significantly faster intensity decay than that of pristine WS₂. Fitting the TRPL curves by the multi-exponential function:

$$\text{D}\left(\text{t}\right)\text{=}{\text{A}}_{\text{1}}{\text{e}}^{\text{-}\frac{\text{t}}{{\text{τ}}_{\text{1}}}}\text{+}{\text{A}}_{\text{2}}{\text{e}}^{\text{-}\frac{\text{t}}{{\text{τ}}_{\text{2}}}}\text{+}{\text{A}}_{\text{3}}{\text{e}}^{\text{-}\frac{\text{t}}{{\text{τ}}_{\text{3}}}}$$

where D(t) is the exciton concentration, τ_i (i = 1, 2, 3) are the time constants, and A₁, A₂, A₃ symbolize the changes in exciton density due to surface states recombination, radiative recombination, and charge transfer, respectively,³¹ the lifetime of ReS₂/WS₂ heterostructure (about 134 ps) is found significantly shorter than that of monolayer WS₂ (about 700 ps) (seeing detail in Table S1 in SI). Further considering the intensity decay in the PL spectrum (Fig. 1c and Figure S1c), the type II band alignment is confirmed for the ReS₂/WS₂ heterostructure. Accordingly, the band structure is schematically shown in Fig. 1h, where the conduction band minimum and valence band maximum locate in WS₂ and ReS₂, respectively. The facilitated interlayer charge transfer strongly predicts a possibility for electrical modulation for the heterostructure.

Resistive switching behavior of the ReS₂/WS₂-based memristor

Based on the understanding of the electronic properties, the ReS₂/WS₂-based planar memristor is constructed on a SiO₂/Si substrate using Au as the source and drain electrodes. Each electrode is in contact with both the ReS₂ and WS₂ monolayers simultaneously, as the schematic diagram and the optical image shown in Fig. 2a,b. The I-V measurements of the device exhibit a typical resistive switching behavior, as shown in the blue curve of Fig. 2c. A compliance current is set at 50 µA, followed by a voltage sweeping from 0 V to 4 V. At an applied voltage of about 3 V, the current increases abruptly, completing the "SET" process from high resistance state (HRS) to low resistance state (LRS); after a voltage sweeping from 4 V to 0 V, the device remains LRS. As removing the compliance current setting and performing a voltage sweep from 0 V to 2 V, the device shifts from LRS to HRS at an increasing current, corresponding to the "RESET" process; and the device maintains HRS at the reversed sweep from 2 V to 0 V. Similar performance of the device is found under the negative voltages, as shown in Figure S3 in SI. This electrical property confirms a typical unipolar resistive switching behavior. Different from the previous reported ReS₂ memristor that only exhibited bipolar resistive switching behavior,^14,32,33 the ReS₂/WS₂ unipolar memristor predicts higher switching ratio, higher integration density, and more simplified control circuit. Good reliability is demonstrated during the repeated 100 times switching cycles (the brown cycle curves in Fig. 2c). The set voltage (V_set) is extracted and depicted in the histogram in Fig. 2d. A Gaussian fit suggests that the V_set generally distributes around 2.90 V.

For comparison, the electrical properties of the planar memristors based on pristine monolayer ReS₂ and WS₂ are also studied and shown in Figure S4 and S5 in SI, respectively. The ReS₂-based device exhibits a typical unipolar memristive property that, the resistance jumps from HRS to LRS at a statistical V_set of about 3.19 V, as depicted in the histogram in Fig. 2e. While for the WS₂-based device, the HRS is maintained without significant change even when the voltage exceeds 7.0 V. Obviously, the ReS₂ plays a key role in the resistive switching for the heterostructure, since the ReS₂ and WS₂ layers are in parallel connecting to the Au electrons. By comparison, the ReS₂/WS₂-based planar memristor has a decreased and a more stable V_set value than that of monolayer ReS₂.

Except for the V_set, another crucial factor determining the overall performance of a memristor is the memory window that reflected as R_ON/R_OFF ratio between HRS and LRS. As shown in Fig. 2f, the ReS₂/WS₂-based memristor exhibits a notable R_ON/R_OFF ratio higher than 10⁶, which is superior to most of the existing 2D memristors. ^{14,29,34−36} This value is more than an order of magnitude larger than that of pure ReS₂ (10⁵), indicating a larger memory window when forming the heterostructure. Moreover, the ReS₂/WS₂-based device possesses multiple adjustable resistance states within 5x10² ~ 8x10⁴ Ω by varying the compliance current from 1 µA to 100 µA, as shown in Fig. 2g, demonstrating its great potential for multi-level data storage and multi-state neuromorphic computing.³⁷ Subsequently, a switching cycling test of the heterostructure memristor is conducted over 200 times (Fig. 2h), which shows a reliable resistive switching performance with a clear memory window. A desirable stability of the memristor is also evidenced by the retention time of over 10⁴ s for each HRS and LRS state, as seen in Fig. 2i.

Gate modulation and its mechanism of the ReS₂/WS₂-based memristor

Furthermore, the electrical modulation capability is examined by the three-terminal FET configuration (Fig. 3a). The I-V curves measured under different gate voltages (V_g) are shown in Figures S6 and S7 in SI, and the V_g dependent V_set is summarized in Fig. 3a. As is shown, opposite variation trends for the V_set are found under different V_g directions. For a positive gating (V_g > 0), the V_set increases from 2.9 V to 3.5 V when increasing the V_g from 0 V to 2 V. As the positive V_g increases to 3 V, 4 V, and 5 V, the device is always in a HRS state within the voltage sweep range of 0 ~ 5 V, as shown in Fig. 3b, i.e., the device is blocked from turning on. To be more specific, a negative gating (V_g < 0) significantly reduces the V_set of the device, which is decreased from 2.9 V to 1 V when changing the V_g from − 5 V to − 8 V. While in the V_g range of 0 ~ − 5 V, the V_set is relatively stable. Therefore, the ReS₂/WS₂-based memristor not only possesses a high R_ON/R_OFF ratio, multiple adjustable resistance states, good endurance and desirable retention, but also demonstrates a significant gate controllability.

The mechanism of resistive switching and gate modulation in the ReS₂/WS₂-based planar memristor is described in Fig. 3c. Monolayer ReS₂ and WS₂ grown by the CVD method have been demonstrated to possess n-type conductivity with intrinsic S vacancies and additional electrons. When forming the heterostructure, the electrons will transfer from WS₂ to ReS₂ owing to the type II band alignment (the top panel in Fig. 3c). By applying a forward bias and a compliance current, electrons in the ReS₂ layer transport to the conducting channel to form a current. When the bias voltage reaches a threshold, the conducting channel meets the condition for rapid migration of electrons between electrodes, and the device jumps from HRS to LRS ("SET" process, the middle panel in Fig. 3c). Hence, the formation of conducting filaments by S vacancies is the primary conduction mechanism in the memristor.¹⁴ As grown ReS₂ generally possesses a lower stoichiometric ratio than that of WS₂, and thus has greater possibility to generate S vacancies.^12,38 As such, the resistive switching is dominant by the ReS₂ layer during the device working, while the WS₂ layer maintains HRS. The large accumulation of electrons near the head of the conducting channel causes a noticeable reduction of interface electrons in the ReS₂ layer. Therefore, more electrons will transfer from WS₂ to ReS₂, which explains the faster conductance change of ReS₂/WS₂-based device than that of ReS₂. Simultaneously, owing to the addition of electrons, the resistance at LRS of ReS₂/WS₂-based device is an order of magnitude lower (Fig. 2f), resulting in its higher R_ON/R_OFF ratio. When the voltage with the same polarity is scanned again with no applied compliance current, the effect of Joule heating is greater than that of the voltage between electrodes. As a result, the conducting channel breaks and the S vacancies gradually return to their original state.³⁹ The device resistance jumps back to the HRS ("RESET" process).

When applying a positive V_g during the “SET” process, some electrons in the device are attracted to the interface of ReS₂ and SiO₂ substrate due to the electrostatic equilibrium effect (the bottom panel in Fig. 3c). Fewer electrons could jump around the conducting channel made up of S vacancies, causing an increased V_set threshold required for the resistive switching (Fig. 3a). As the V_g increases further, the consumption of electrons finally blocks out the device from turning on (Fig. 3b). In contrast, by applying a negative V_g, the electrostatic equilibrium effect causes the holes to accumulate at the interface near the SiO₂ substrate. In a small gating range of 0 ~ − 5 V, the consumed holes in ReS₂ may be supplemented by those transferred from WS₂, leading to a relatively stable V_set. Further consumption of holes results in more left electrons for the conduction. The left electrons are active to the conducting channel formed by S vacancies, and the required V_set threshold thus gradually decreases (Fig. 3a).

Simulated synaptic properties of the ReS₂/WS₂-based memristor

The superior electronic performance of the ReS₂/WS₂-based planar memristor sheds light on exploring of its potential applications. We demonstrate that it can simulate partial neuron-based biological synaptic functions, as shown in Fig. 4a. When a presynaptic neuron is stimulated and conducted to a synaptic vesicle, the synaptic vesicle fuses tightly with the presynaptic membrane and causing a rupture. The neurotransmitters within the synaptic vesicles are released into the synaptic space, diffuse to reach the postsynaptic membrane, and thereby induce excitatory or inhibitory modifications in the postsynaptic membrane. For the ReS₂/WS₂-based planar memristor, the S vacancies are comparable to neurotransmitters in biological synapses.

To simulate the biological synaptic plasticity, conductance of ReS₂/WS₂-based memristor is modulated by applying a pulse voltage, with the pulse parameters such as amplitude, width, and interval varied to adjust the synaptic weights. As shown in Figure S8 in SI, at a 1 V reading voltage and a fixed 5 V amplitude, the device current tends to saturate with the continuous application of pulse, when the pulse width and interval increase in equal proportion. This performance exhibits a typical memristor characteric.⁴⁰ Accordingly, the device property is analyzed by acquiring the conductance at the saturated currents for different pulse amplitudes, widths, and intervals. As the results shown in Fig. 4b, the conductance is increased when simultaneously increasing the pulse width and interval from 0.5 s to 5 s, indicating a positive response to a continuously longer stimulation. The conductance can also be modulated by individually adjusting the pulse width or interval. An enlarged pulse width increases the upward trend of the conductance, while an opposite trend is found when increasing the pulse intervals, as shown in Fig. 4c,d. Modulating the memristor characteristics by adjusting pulse width and interval is recognized as the biological synapse SRDP.⁴¹ In addition, the impact of voltage amplitude is shown Fig. 4e, where the device conductance is increased with the enhancing pulse voltage for all the four setting pulse widths and intervals.

PPF representing the STP is an important physiological phenomenon. It is manifested as the temporal sum of biological synaptic inputs, and can be estimated from the change of synaptic weights as responding to the stimuli of two consecutive pulsed voltages.^42,43 Fig. 4f shows the PPF behavior of the ReS₂/WS₂-based memristor, and the inset illustrates the applied pulse. The PPF can be quantitatively expressed as:⁴⁴

$$\text{PPF=}\frac{\left({\text{B}}_{\text{2}}\text{-}{\text{B}}_{\text{1}}\right)}{{\text{B}}_{\text{1}}}\text{×100%}$$

where B₁ and B₂ are the conductance corresponding to the first and second pulses, respectively. The fitting relationship between PPF data and pulse is as follows:⁴⁵

$$\text{y=}{\text{C}}_{\text{1}}{\text{e}}^{\text{-}\frac{\text{t}}{{\text{α}}_{\text{1}}}}\text{+}{\text{C}}_{\text{2}}{\text{e}}^{\text{-}\frac{\text{t}}{{\text{α}}_{\text{2}}}}$$

which gives α₁ = 0.05 s and α₂ = 2.9 s, corresponding to the fast and slow decay terms, respectively. The enhancement of conductance under successive pulses stimulation tends to decay exponentially with the increasing pulse intervals, consistent with the behavior of biological synapses.

The above studies have demonstrated the effective simulation of biological synaptic function and plasticity in ReS₂/WS₂-based planar memristor, which are all electrical memristor characteristics. While benefiting from the interlayer coupling and charge transfer mechanisms, ReS₂/WS₂ also displays unique photoresponsive behavior as a type II heterostructure,⁴⁶ thus the optoelectronics-inspired memristor characteristics can also be expected. Since the interlayer charge transfer results in a diminution of V_set in the ReS₂/WS₂-based memristor compared with that of the ReS₂-based device (Fig. 2d,e), a light modulation is further performed on the two memristors under different excitation wavelengths (532 nm and 690 nm) and with various powers. The I-V curves for all the cases are acquired, as shown in Figures S9 - S12 in SI, respectively, and the extracted V_set values are illustrated in Fig. 4g. The results for ReS₂/WS₂-based memristor show an interesting wavelength-dependent conductance controllability. Under a 532 nm excitation, its V_set drops from about 3 V to below 1 V, exhibiting a negatively correlated with the luminous power. While a 690 nm excitation basically does not affect the V_set that is almost stable at 3 V even when the luminous power increases to 8 mW. Different from the wavelength-dependent performance for heterostructure memristor, the V_set of ReS₂-based device is basically stable at around 3.1 V under illumination with different powers at both wavelengths. The obtained V_set values are essentially coincident with that measured without the excitation, and larger than that of the ReS₂/WS₂-based memristor with the same conditions. This strongly suggests that, the variation of V_set originates from the property of ReS₂/WS₂ rather than the inherent characteristics of ReS₂. Compared with the single layer, the formation of type II heterostructure enables the optical modulation over its electrical performance. Beyond the most existing memristors which based on only electrical modulation,^47,48 the above results demonstrate an exciting behavior for the optoelectronic memristor based on the ReS₂/WS₂ heterostructure.

The light-tunable synaptic plasticity is then investigated by applying a pulse under 532 nm excitation at different luminous powers. As shown in Fig. 4h, the conductance at all luminous powers maintains an increase trend with the pulse number, and exhibiting a considerable light-sensing behavior that increases steadily with the increasing luminous power from 0 mW to 8 mW. Such power-dependent conductance control indicates the optical tunability on the synaptic weight, predicting a potential for future visual neural applications. Figure 4i exhibits an effective control over the switching time (the time of device current to stabilize when a single pulse is applied) through the modulation of luminous power. For a pulse with a reading voltage of 1 V, an amplitude of 5 V, and a pulse width and interval of both 3 s, the switching time decreases from about 1.8 s to 0.6 s under the 532 nm excitation. This hints at the advanced sensitivity of optically modulated memristor on the neuromorphic applications.

Optical modulation and its mechanism of the ReS₂/WS₂-based memristor

One issue that needs to be clarified is the wavelength-dependent conductance control of the ReS₂/WS₂-based planar memristor. From the perspective of band structure, the mechanism can be described in Fig. 5. Without the optical modulation, the electrons transfer from WS₂ to ReS₂, while the holes transfer in the opposite direction, forming a build-in electric field directed from WS₂ to ReS₂ (Fig. 5a). Under a 690 nm (1.80 eV) laser irradiation, monolayer ReS₂ with a direct bandgap of about 1.65 eV is excited, while WS₂ with a 2.07 eV bandgap is out of the excitation energy (Fig. 5b). Consequently, the produced photogenerated carriers distribute mainly within the ReS₂ layer. Driven by the build-in electric field, the electrons trend to transfer to WS₂, which are blocked by the interfacial potential barrier between the conduction band edges. As a results, the excited electrons and holes mostly will recombine through a radiative or nonradiative process, which essentially does not alter the net carrier distribution in the system. As a result, the V_set of the ReS₂/WS₂-based memristor is essentially unaffected as well as that of the ReS₂-based device, under the modulation of 690 nm illumination and with different luminous powers (Fig. 4g, red and blue lines). For a 532 nm (2.33 eV) laser irradiation, both the ReS₂ and WS₂ layers can be excited, as shown in Fig. 5c. Since the photon energy is closer to the resonance excitation of the optical bandgap of WS₂, a higher excitation efficiency with more photogenerated carriers are produced in WS₂ than in ReS₂. The additional electrons will transfer from WS₂ into ReS₂ through the heterogeneous interface, while the transfer of additional holes can be blocked by the interfacial potential barrier. The increased electrons in ReS₂ make a positive contribution to the composition of the conductive channel by S vacancies. Consequently, the excited electrons increase with increasing luminous power under the 532 nm excitation, and thus reduces the applied voltage required for forming conducting channel in ReS₂/WS₂ heterostructure (Fig. 4g, yellow line).

In summary, a scalable 2D planar memristor based on ReS₂/WS₂ heterostructure is constructed through an alignment transfer of CVD grown ReS₂ and WS₂ monolayers. The type II band structure with strong interlayer interaction is confirmed by Raman, TRPL, HRTEM and KPFM characterizations for the heterostructure. The fabricated memristor exhibits a typical unipolar resistive switching property, with the high R_ON/R_OFF ratio up to 10⁶, tunable resistance states (10²–10⁴ Ω) for multi-level data storage, high gate controllability, clearly extended endurance, and good retention. It is also successfully in simulation of partially biological synaptic functions, including STP, SRDP and PPF, demonstrating the potential applications in neural networks and complex system. Besides the electrical modulation, optical control over the resistive switching is further achieved by tuning the wavelength and power of the exciting light. The optoelectronic property and interlayer interaction are revealed to be the mechanism for the superior performance of the heterostructure device, as well as its electrical and optical controllability. All these findings demonstrate a potential for the development of energy-efficient multi-terminal devices based on 2D van der Waals heterostructures and enable the further exploration of neurofunctional devices with optical programmability.

Preparation of ReS₂ /WS₂ Heterostructure.

2D monolayer ReS₂ crystals are grown at atmospheric pressure, using the CVD growth methods as our previous reports.^49,50 The setup of growth systems and the growth processes are provided detailedly in Figure S13 in SI. The ReS₂/WS₂ heterostructure with special stacking configuration (60° vertical stacking) is prepared through the typical wet transfer method. The poly (methyl methacrylate) (PMMA) is used as the supporting film assisting to peel off the monolayer ReS₂ and WS₂ from mica and sapphire substrates, respectively. The transfer process is described detailedly in Figure S14 in SI.

Characterizations.

Raman and PL spectra are recorded using a WITec alpha 300RA confocal spectrometer system with a laser wavelength of 488 nm and a beam size of about 1 µm. The step size of the Raman and PL mapping is 800 nm. A SPA400-Nanonavi AFM is used to measure the thickness of ReS₂/WS₂ heterostructure. TEM measurements, including HRTEM and SAED, are performed on a field emission TEM (JEM-2100) at an accelerating voltage of 200 kV. Surface potential of ReS₂/WS₂ is measured by the Kelvin Probe Module of the SPA400-Nanonavi AFM after is transferred to an Au-coated SiO₂/Si substrate. The photoinduced interfacial charge behavior of the samples are recorded by the TRPL spectra (HORIBA, MicOS) at a PL peak of 620 nm.

Device Fabrication and Measurements.

ReS₂/WS₂, ReS₂ and WS₂ layers are transferred to SiO₂/Si substrates, respectively, through the above transfer method. Designed ReS₂/WS₂-based, ReS₂-based and WS₂-based devices are patterned by the direct-write laser photolithography technique using an ML microwriter. A transmission channel with a source-drain spacing of 7 µm is defined by depositing 50 nm Au films as the source and drain in the patterned region using magnetron sputtering. The fabricated devices are annealed at 200°C in mixed Ar (a flow rate of 100 sccm) and hydrogen (H₂) (a flow rate of 3 sccm) gas environment for 30 min to improve the quality of the heterogeneous interface as well as the metal-semi-contact. The electrical properties of the devices are measured in ambient atmosphere condition at room temperature, and under dark, 532 nm and 690 nm laser irradiation conditions, respectively, using a semiconductor parameter analyzer (Agilent B2912A source-meter unit system).

Acknowledgements

The work is funded by the National Science Fund for Excellent Young Scholars (grant no. 62022068), the National Natural Science Foundation of China (grant nos. 61874092, 61974123, and 61804129), Science and Technology Project of Fujian Province of China (grant no. 2019H0002).

Author Contributions

Yaping Wu and Feihong Huang conceived the project and designed experiments. Feihong Huang prepared samples, fabricated devices, performed Raman and PL measurements, carried AFM and KPFM measurements and measured the performance of devices. Jun Yin performed TRPL measurements. Li Chen performed TEM and SAED measurements. Yaping Wu, Zhiming Wu, Xu Li and Junyong Kang provided the Fund support. Chunmiao Zhang and Feiya Xu assisted laboratory management. Yaping Wu, Feihong Huang and Congming Ke analyzed the data. Feihong Huang and Yaping Wu wrote the manuscript. All authors discussed the results and contributed to the final version of the manuscript.

Conflict of Interest

The authors declare no conflict of interest.

Supporting Information

Supporting Information is available from Light: Science & Applications or from the author.

Zhu, X. & Lee, J. & Lu, W. D. Iodine Vacancy Redistribution in Organic-Inorganic Halide Perovskite Films and Resistive Switching Effects. Adv. Mater. 29, 1700527 (2017).
Chen, S. & Lou, Z. & Chen, D. & Shen, G. An Artificial Flexible Visual Memory System Based on an UV-Motivated Memristor. Adv. Mater. 30, 1705400 (2018).
Yang, J. J. & Strukov, D. B. & Stewart, D. R., Memristive devices for computing. Nat Nanotechnol. 2013, 8, 1324.
Indiveri, G. & Linares-Barranco, B. & Legenstein, R. & Deligeorgis, G. & Prodromakis, T. Integration of nanoscale memristor synapses in neuromorphic computing architectures. Nanotechnol. 24, 384010 (2013).
Yan, X. et al. Memristor with Ag-Cluster-Doped TiO₂ Films as Artificial Synapse for Neuroinspired Computing. Adv. Funct. Mater. 28, 1705320 (2018).
Wedig, A. et al. Nanoscale cation motion in TaO_x, HfO_x and TiO_x memristive systems. Nat. Nanotechnol. 11, 67 (2015).
Wang, Z. et al. Engineering incremental resistive switching in TaO_x based memristors for brain-inspired computing. Nanoscale 8, 14015-22 (2016).
Tian, H. et al. Graphene Dynamic Synapse with Modulatable Plasticity. Nano Lett. 15, 8013-9 (2015).
Tian, H. et al. Emulating Bilingual Synaptic Response Using a Junction-Based Artificial Synaptic Device. ACS Nano 11, 7156-7163 (2017).
Yang, C. M. et al. Bidirectional All‐Optical Synapses Based on a 2D Bi₂O₂Se/Graphene Hybrid Structure for Multifunctional Optoelectronics. Adv. Funct. Mater. 30, 2001598 (2020).
Tongay, S. et al. Monolayer behaviour in bulk ReS₂ due to electronic and vibrational decoupling. Nat. Commun. 5, 3252 (2014).
Horzum, S. et al. Formation and stability of point defects in monolayer rhenium disulfide. Phys. Rev. B 89, 155433 (2014).
Shim, J. et al. High-Performance 2D Rhenium Disulfide (ReS₂) Transistors and Photodetectors by Oxygen Plasma Treatment. Adv. Mater. 28, 6985-92 (2016).
Li, S. et al. Electron-beam-irradiated rhenium disulfide memristors with low variability for neuromorphic computing. npj 2D Mater. and Applications 5, 1 (2021).
Özçelik, V. O. et al. Band alignment of two-dimensional semiconductors for designing heterostructures with momentum space matching. Phys. Rev. B 94, 035125 (2016).
Wang, Q. & Dou, K. & Shi, X. Band alignment in multilayered semiconductor homojunctions supported on metals. J. Mater. Chem. C 8, 959-967 (2020).
Zhang, W. et al. An ultrathin memristor based on a two-dimensional WS₂/MoS₂ heterojunction. Nanoscale 13, 11497-11504 (2021).
Li, H. et al. Fully photon modulated heterostructure for neuromorphic computing. Nano Energy 65, 104000 (2019).
Situ, G. Deep holography. Light: Advanced Manufacturing 3, 13 (2022).
Kim, C. O. et al. High photoresponsivity in an all-graphene p-n vertical junction photodetector. Nat. Commun. 5, 3249 (2014).
Yuan, H. et al. Polarization-sensitive broadband photodetector using a black phosphorus vertical p-n junction. Nat. Nanotechnol. 10, 707-13 (2015).
Nugera, F. A. et al. Bandgap Engineering in 2D Lateral Heterostructures of Transition Metal Dichalcogenides via Controlled Alloying. Small 18, 2106600 (2022).
Cai, Z. & Liu, B. & Zou, X. & Cheng, H. M. Chemical Vapor Deposition Growth and Applications of Two-Dimensional Materials and Their Heterostructures. Chem. Rev. 118, 6091-6133 (2018).
Xu, Z. et al. CVD controlled growth of large-scale WS₂ monolayers. RSC Advances 9, 29628-29635 (2019).
Berkdemir, A. et al. Identification of individual and few layers of WS₂ using Raman Spectroscopy. Sci. Rep. 3, 1755 (2013).
Wang, F. et al. Strain-induced phonon shifts in tungsten disulfide nanoplatelets and nanotubes. 2D Materials 4, 015007 (2016).
Zhang, G. & Zhang, Y. W. Strain effects on thermoelectric properties of two-dimensional materials. Mech. Mater. 91, 382-398 (2015).
Zhang, T. et al. Twinned growth behaviour of two-dimensional materials. Nat. Commun. 7, 13911 (2016).
Jia, Z. et al. Grain wall boundaries in centimeter-scale continuous monolayer WS₂ film grown by chemical vapor deposition. Nanotechnol. 29, 255705 (2018).
Liu, D. et al. Synthesis of 2H‐1T′ WS₂‐ReS₂ Heterophase Structures with Atomically Sharp Interface via Hydrogen‐Triggered One‐Pot Growth. Adv. Funct. Mater. 30, 1910169 (2020).
Yuan, Y. et al. Growth of CdSe/MoS₂ vertical heterostructures for fast visible-wavelength photodetectors. J. Alloys Compd. 815, 152309 (2020).
Nguyen, D. A. et al. Electrically and Optically Controllable p-n Junction Memtransistor Based on an Al₂O₃ Encapsulated 2D Te/ReS₂ van der Waals Heterostructure. Small Methods 5, e2101303 (2021).
Pam, M. E. et al. Interface-Modulated Resistive Switching in Mo-Irradiated ReS₂ for Neuromorphic Computing. Adv. Mater. 34, e2202722 (2022).
Yan, X. et al. Vacancy-Induced Synaptic Behavior in 2D WS₂ Nanosheet-Based Memristor for Low-Power Neuromorphic Computing. Small 15, 1901423 (2019).
Jadwiszczak, J. et al. MoS₂ Memtransistors Fabricated by Localized Helium Ion Beam Irradiation. ACS Nano 13, 14262-14273 (2019).
Xu, R. et al. Vertical MoS₂ Double-Layer Memristor with Electrochemical Metallization as an Atomic-Scale Synapse with Switching Thresholds Approaching 100 mV. Nano Lett. 19, 2411-2417 (2019).
Feng, X. et al. A Fully Printed Flexible MoS₂ Memristive Artificial Synapse with Femtojoule Switching Energy. Advanced Electronic Materials 5, 1900740 (2019).
Vogel, E. M. & Robinson, J. A. Two-dimensional layered transition-metal dichalcogenides for versatile properties and applications. MRS Bull. 40, 558-563 (2015).
Ahn & D. C. K. S. S. S. E. Electrical observations of filamentary conductions for the resistive memory switching in NiO films. Appl. Phys. Lett. 88, 202102 (2006).
Ren, Y. et al. Phosphorene nano-heterostructure based memristors with broadband response synaptic plasticity. J. Mater. Chem. C 6, 9383-9393 (2018).
Chen, Y. et al. Polymer memristor for information storage and neuromorphic applications. Materials Horizons 1, 489 (2014).
Wang, Z. et al. Memristors with diffusive dynamics as synaptic emulators for neuromorphic computing. Nat. Mater. 16, 101-108 (2017).
Kumar, M. et al. Vertically Aligned WS₂ Layers for High‐Performing Memristors and Artificial Synapses. Adv. Electron. Mater. 5, 1900467 (2019).
Zhong, Y.-N. & Wang, T. & Gao, X. & Xu, J.-L. & Wang, S.-D. Synapse-Like Organic Thin Film Memristors. Adv. Funct. Mater. 28, 1800854 (2018).
Zucker, R. S. & Regehr, W. G. Short-term synaptic plasticity. Annu. Rev. Physiol. 64, 355-405 (2002).
Tang, Y. et al. Distinctive Interfacial Charge Behavior and Versatile Photoresponse Performance in Isotropic/Anisotropic WS₂/ReS₂ Heterojunctions. ACS Appl. Mater. Interfaces 12, 53475-53483 (2020).
Yang, Y. et al. Three-terminal memtransistors based on two-dimensional layered gallium selenide nanosheets for potential low-power electronics applications. Nano Energy 57, 566-573 (2019).
Sangwan, V. K. et al. Gate-tunable memristive phenomena mediated by grain boundaries in single-layer MoS₂. Nat. Nanotechnol. 10, 403-6 (2015).
He, Q. et al. Deeply Exploring Anisotropic Evolution toward Large-Scale Growth of Monolayer ReS₂. ACS Appl. Mater. Interfaces 12, 2862-2870 (2020).
Chen, J. et al. Synthesis of Wafer-Scale Monolayer WS₂ Crystals toward the Application in Integrated Electronic Devices. ACS Appl. Mater. Interfaces 11, 19381-19387 (2019).

(Not answered)

SupplementaryInformation.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

An electrically and optically controllable memristor with synaptic plasticity based on scalable monolayer ReS₂/WS₂ heterostructure

Status:

Version 1

Abstract

Figures

Introduction

Results

Discussion

Materials And Methods

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1